{ "best_metric": null, "best_model_checkpoint": null, "epoch": 0.2395954829596032, "global_step": 6000, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 1.0000000000000001e-07, "loss": 1.0488, "step": 1 }, { "epoch": 0.0, "learning_rate": 2.0000000000000002e-07, "loss": 1.1027, "step": 2 }, { "epoch": 0.0, "learning_rate": 3.0000000000000004e-07, "loss": 1.2865, "step": 3 }, { "epoch": 0.0, "learning_rate": 4.0000000000000003e-07, "loss": 1.0681, "step": 4 }, { "epoch": 0.0, "learning_rate": 5.000000000000001e-07, "loss": 1.1824, "step": 5 }, { "epoch": 0.0, "learning_rate": 6.000000000000001e-07, "loss": 1.1611, "step": 6 }, { "epoch": 0.0, "learning_rate": 7.000000000000001e-07, "loss": 1.0395, "step": 7 }, { "epoch": 0.0, "learning_rate": 8.000000000000001e-07, "loss": 1.1625, "step": 8 }, { "epoch": 0.0, "learning_rate": 9.000000000000001e-07, "loss": 1.1697, "step": 9 }, { "epoch": 0.0, "learning_rate": 1.0000000000000002e-06, "loss": 1.233, "step": 10 }, { "epoch": 0.0, "learning_rate": 1.1e-06, "loss": 1.1196, "step": 11 }, { "epoch": 0.0, "learning_rate": 1.2000000000000002e-06, "loss": 1.028, "step": 12 }, { "epoch": 0.0, "learning_rate": 1.3e-06, "loss": 1.079, "step": 13 }, { "epoch": 0.0, "learning_rate": 1.4000000000000001e-06, "loss": 1.161, "step": 14 }, { "epoch": 0.0, "learning_rate": 1.5e-06, "loss": 1.1258, "step": 15 }, { "epoch": 0.0, "learning_rate": 1.6000000000000001e-06, "loss": 1.1793, "step": 16 }, { "epoch": 0.0, "learning_rate": 1.7000000000000002e-06, "loss": 1.2195, "step": 17 }, { "epoch": 0.0, "learning_rate": 1.8000000000000001e-06, "loss": 1.0552, "step": 18 }, { "epoch": 0.0, "learning_rate": 1.9000000000000002e-06, "loss": 0.9736, "step": 19 }, { "epoch": 0.0, "learning_rate": 2.0000000000000003e-06, "loss": 1.2038, "step": 20 }, { "epoch": 0.0, "learning_rate": 2.1000000000000002e-06, "loss": 1.1958, "step": 21 }, { "epoch": 0.0, "learning_rate": 2.2e-06, "loss": 0.9723, "step": 22 }, { "epoch": 0.0, "learning_rate": 2.3000000000000004e-06, "loss": 1.0598, "step": 23 }, { "epoch": 0.0, "learning_rate": 2.4000000000000003e-06, "loss": 0.8501, "step": 24 }, { "epoch": 0.0, "learning_rate": 2.5e-06, "loss": 0.8728, "step": 25 }, { "epoch": 0.0, "learning_rate": 2.6e-06, "loss": 1.1256, "step": 26 }, { "epoch": 0.0, "learning_rate": 2.7000000000000004e-06, "loss": 0.9702, "step": 27 }, { "epoch": 0.0, "learning_rate": 2.8000000000000003e-06, "loss": 1.1258, "step": 28 }, { "epoch": 0.0, "learning_rate": 2.9e-06, "loss": 1.0507, "step": 29 }, { "epoch": 0.0, "learning_rate": 3e-06, "loss": 0.9789, "step": 30 }, { "epoch": 0.0, "learning_rate": 3.1000000000000004e-06, "loss": 0.8359, "step": 31 }, { "epoch": 0.0, "learning_rate": 3.2000000000000003e-06, "loss": 0.9653, "step": 32 }, { "epoch": 0.0, "learning_rate": 3.3000000000000006e-06, "loss": 0.8996, "step": 33 }, { "epoch": 0.0, "learning_rate": 3.4000000000000005e-06, "loss": 1.0018, "step": 34 }, { "epoch": 0.0, "learning_rate": 3.5e-06, "loss": 0.8889, "step": 35 }, { "epoch": 0.0, "learning_rate": 3.6000000000000003e-06, "loss": 0.9844, "step": 36 }, { "epoch": 0.0, "learning_rate": 3.7e-06, "loss": 0.9551, "step": 37 }, { "epoch": 0.0, "learning_rate": 3.8000000000000005e-06, "loss": 0.8327, "step": 38 }, { "epoch": 0.0, "learning_rate": 3.900000000000001e-06, "loss": 0.8914, "step": 39 }, { "epoch": 0.0, "learning_rate": 4.000000000000001e-06, "loss": 0.8989, "step": 40 }, { "epoch": 0.0, "learning_rate": 4.1e-06, "loss": 0.703, "step": 41 }, { "epoch": 0.0, "learning_rate": 4.2000000000000004e-06, "loss": 0.884, "step": 42 }, { "epoch": 0.0, "learning_rate": 4.3e-06, "loss": 0.7519, "step": 43 }, { "epoch": 0.0, "learning_rate": 4.4e-06, "loss": 0.8547, "step": 44 }, { "epoch": 0.0, "learning_rate": 4.5e-06, "loss": 0.8231, "step": 45 }, { "epoch": 0.0, "learning_rate": 4.600000000000001e-06, "loss": 0.7351, "step": 46 }, { "epoch": 0.0, "learning_rate": 4.7e-06, "loss": 0.7906, "step": 47 }, { "epoch": 0.0, "learning_rate": 4.800000000000001e-06, "loss": 0.7365, "step": 48 }, { "epoch": 0.0, "learning_rate": 4.9000000000000005e-06, "loss": 0.7611, "step": 49 }, { "epoch": 0.0, "learning_rate": 5e-06, "loss": 0.6853, "step": 50 }, { "epoch": 0.0, "learning_rate": 5.1e-06, "loss": 0.6458, "step": 51 }, { "epoch": 0.0, "learning_rate": 5.2e-06, "loss": 0.712, "step": 52 }, { "epoch": 0.0, "learning_rate": 5.300000000000001e-06, "loss": 0.6755, "step": 53 }, { "epoch": 0.0, "learning_rate": 5.400000000000001e-06, "loss": 0.6172, "step": 54 }, { "epoch": 0.0, "learning_rate": 5.500000000000001e-06, "loss": 0.6864, "step": 55 }, { "epoch": 0.0, "learning_rate": 5.600000000000001e-06, "loss": 0.6047, "step": 56 }, { "epoch": 0.0, "learning_rate": 5.7e-06, "loss": 0.5952, "step": 57 }, { "epoch": 0.0, "learning_rate": 5.8e-06, "loss": 0.6368, "step": 58 }, { "epoch": 0.0, "learning_rate": 5.9e-06, "loss": 0.5363, "step": 59 }, { "epoch": 0.0, "learning_rate": 6e-06, "loss": 0.5781, "step": 60 }, { "epoch": 0.0, "learning_rate": 6.1e-06, "loss": 0.6068, "step": 61 }, { "epoch": 0.0, "learning_rate": 6.200000000000001e-06, "loss": 0.5856, "step": 62 }, { "epoch": 0.0, "learning_rate": 6.300000000000001e-06, "loss": 0.5719, "step": 63 }, { "epoch": 0.0, "learning_rate": 6.4000000000000006e-06, "loss": 0.5862, "step": 64 }, { "epoch": 0.0, "learning_rate": 6.5000000000000004e-06, "loss": 0.5424, "step": 65 }, { "epoch": 0.0, "learning_rate": 6.600000000000001e-06, "loss": 0.4901, "step": 66 }, { "epoch": 0.0, "learning_rate": 6.700000000000001e-06, "loss": 0.4572, "step": 67 }, { "epoch": 0.0, "learning_rate": 6.800000000000001e-06, "loss": 0.4476, "step": 68 }, { "epoch": 0.0, "learning_rate": 6.9e-06, "loss": 0.5266, "step": 69 }, { "epoch": 0.0, "learning_rate": 7e-06, "loss": 0.5051, "step": 70 }, { "epoch": 0.0, "learning_rate": 7.100000000000001e-06, "loss": 0.4577, "step": 71 }, { "epoch": 0.0, "learning_rate": 7.2000000000000005e-06, "loss": 0.4483, "step": 72 }, { "epoch": 0.0, "learning_rate": 7.3e-06, "loss": 0.5268, "step": 73 }, { "epoch": 0.0, "learning_rate": 7.4e-06, "loss": 0.4774, "step": 74 }, { "epoch": 0.0, "learning_rate": 7.500000000000001e-06, "loss": 0.3889, "step": 75 }, { "epoch": 0.0, "learning_rate": 7.600000000000001e-06, "loss": 0.5319, "step": 76 }, { "epoch": 0.0, "learning_rate": 7.7e-06, "loss": 0.4035, "step": 77 }, { "epoch": 0.0, "learning_rate": 7.800000000000002e-06, "loss": 0.5946, "step": 78 }, { "epoch": 0.0, "learning_rate": 7.9e-06, "loss": 0.4474, "step": 79 }, { "epoch": 0.0, "learning_rate": 8.000000000000001e-06, "loss": 0.4996, "step": 80 }, { "epoch": 0.0, "learning_rate": 8.1e-06, "loss": 0.6067, "step": 81 }, { "epoch": 0.0, "learning_rate": 8.2e-06, "loss": 0.4301, "step": 82 }, { "epoch": 0.0, "learning_rate": 8.3e-06, "loss": 0.4119, "step": 83 }, { "epoch": 0.0, "learning_rate": 8.400000000000001e-06, "loss": 0.4311, "step": 84 }, { "epoch": 0.0, "learning_rate": 8.5e-06, "loss": 0.4479, "step": 85 }, { "epoch": 0.0, "learning_rate": 8.6e-06, "loss": 0.3555, "step": 86 }, { "epoch": 0.0, "learning_rate": 8.700000000000001e-06, "loss": 0.4877, "step": 87 }, { "epoch": 0.0, "learning_rate": 8.8e-06, "loss": 0.3912, "step": 88 }, { "epoch": 0.0, "learning_rate": 8.900000000000001e-06, "loss": 0.3578, "step": 89 }, { "epoch": 0.0, "learning_rate": 9e-06, "loss": 0.3706, "step": 90 }, { "epoch": 0.0, "learning_rate": 9.100000000000001e-06, "loss": 0.3889, "step": 91 }, { "epoch": 0.0, "learning_rate": 9.200000000000002e-06, "loss": 0.4256, "step": 92 }, { "epoch": 0.0, "learning_rate": 9.3e-06, "loss": 0.3372, "step": 93 }, { "epoch": 0.0, "learning_rate": 9.4e-06, "loss": 0.4404, "step": 94 }, { "epoch": 0.0, "learning_rate": 9.5e-06, "loss": 0.3325, "step": 95 }, { "epoch": 0.0, "learning_rate": 9.600000000000001e-06, "loss": 0.4104, "step": 96 }, { "epoch": 0.0, "learning_rate": 9.7e-06, "loss": 0.4069, "step": 97 }, { "epoch": 0.0, "learning_rate": 9.800000000000001e-06, "loss": 0.3977, "step": 98 }, { "epoch": 0.0, "learning_rate": 9.9e-06, "loss": 0.47, "step": 99 }, { "epoch": 0.0, "learning_rate": 1e-05, "loss": 0.3292, "step": 100 }, { "epoch": 0.0, "learning_rate": 9.999866702212744e-06, "loss": 0.4137, "step": 101 }, { "epoch": 0.0, "learning_rate": 9.999733404425488e-06, "loss": 0.3597, "step": 102 }, { "epoch": 0.0, "learning_rate": 9.999600106638231e-06, "loss": 0.3905, "step": 103 }, { "epoch": 0.0, "learning_rate": 9.999466808850974e-06, "loss": 0.3795, "step": 104 }, { "epoch": 0.0, "learning_rate": 9.999333511063717e-06, "loss": 0.3564, "step": 105 }, { "epoch": 0.0, "learning_rate": 9.999200213276461e-06, "loss": 0.4117, "step": 106 }, { "epoch": 0.0, "learning_rate": 9.999066915489204e-06, "loss": 0.3322, "step": 107 }, { "epoch": 0.0, "learning_rate": 9.998933617701947e-06, "loss": 0.3209, "step": 108 }, { "epoch": 0.0, "learning_rate": 9.99880031991469e-06, "loss": 0.2899, "step": 109 }, { "epoch": 0.0, "learning_rate": 9.998667022127434e-06, "loss": 0.3324, "step": 110 }, { "epoch": 0.0, "learning_rate": 9.998533724340177e-06, "loss": 0.4101, "step": 111 }, { "epoch": 0.0, "learning_rate": 9.99840042655292e-06, "loss": 0.4075, "step": 112 }, { "epoch": 0.0, "learning_rate": 9.998267128765663e-06, "loss": 0.2901, "step": 113 }, { "epoch": 0.0, "learning_rate": 9.998133830978407e-06, "loss": 0.3618, "step": 114 }, { "epoch": 0.0, "learning_rate": 9.99800053319115e-06, "loss": 0.3754, "step": 115 }, { "epoch": 0.0, "learning_rate": 9.997867235403893e-06, "loss": 0.2869, "step": 116 }, { "epoch": 0.0, "learning_rate": 9.997733937616636e-06, "loss": 0.4456, "step": 117 }, { "epoch": 0.0, "learning_rate": 9.99760063982938e-06, "loss": 0.3589, "step": 118 }, { "epoch": 0.0, "learning_rate": 9.997467342042123e-06, "loss": 0.3046, "step": 119 }, { "epoch": 0.0, "learning_rate": 9.997334044254866e-06, "loss": 0.3358, "step": 120 }, { "epoch": 0.0, "learning_rate": 9.997200746467608e-06, "loss": 0.3456, "step": 121 }, { "epoch": 0.0, "learning_rate": 9.997067448680353e-06, "loss": 0.3451, "step": 122 }, { "epoch": 0.0, "learning_rate": 9.996934150893096e-06, "loss": 0.3562, "step": 123 }, { "epoch": 0.0, "learning_rate": 9.996800853105839e-06, "loss": 0.301, "step": 124 }, { "epoch": 0.0, "learning_rate": 9.996667555318581e-06, "loss": 0.379, "step": 125 }, { "epoch": 0.01, "learning_rate": 9.996534257531326e-06, "loss": 0.3841, "step": 126 }, { "epoch": 0.01, "learning_rate": 9.996400959744069e-06, "loss": 0.3027, "step": 127 }, { "epoch": 0.01, "learning_rate": 9.996267661956812e-06, "loss": 0.3442, "step": 128 }, { "epoch": 0.01, "learning_rate": 9.996134364169554e-06, "loss": 0.3396, "step": 129 }, { "epoch": 0.01, "learning_rate": 9.996001066382299e-06, "loss": 0.3589, "step": 130 }, { "epoch": 0.01, "learning_rate": 9.995867768595042e-06, "loss": 0.3797, "step": 131 }, { "epoch": 0.01, "learning_rate": 9.995734470807785e-06, "loss": 0.3317, "step": 132 }, { "epoch": 0.01, "learning_rate": 9.995601173020527e-06, "loss": 0.3804, "step": 133 }, { "epoch": 0.01, "learning_rate": 9.995467875233272e-06, "loss": 0.3244, "step": 134 }, { "epoch": 0.01, "learning_rate": 9.995334577446015e-06, "loss": 0.3035, "step": 135 }, { "epoch": 0.01, "learning_rate": 9.995201279658758e-06, "loss": 0.3823, "step": 136 }, { "epoch": 0.01, "learning_rate": 9.995067981871502e-06, "loss": 0.3087, "step": 137 }, { "epoch": 0.01, "learning_rate": 9.994934684084245e-06, "loss": 0.3275, "step": 138 }, { "epoch": 0.01, "learning_rate": 9.994801386296988e-06, "loss": 0.318, "step": 139 }, { "epoch": 0.01, "learning_rate": 9.99466808850973e-06, "loss": 0.3507, "step": 140 }, { "epoch": 0.01, "learning_rate": 9.994534790722475e-06, "loss": 0.3125, "step": 141 }, { "epoch": 0.01, "learning_rate": 9.994401492935218e-06, "loss": 0.28, "step": 142 }, { "epoch": 0.01, "learning_rate": 9.99426819514796e-06, "loss": 0.2708, "step": 143 }, { "epoch": 0.01, "learning_rate": 9.994134897360705e-06, "loss": 0.2941, "step": 144 }, { "epoch": 0.01, "learning_rate": 9.994001599573448e-06, "loss": 0.339, "step": 145 }, { "epoch": 0.01, "learning_rate": 9.99386830178619e-06, "loss": 0.2508, "step": 146 }, { "epoch": 0.01, "learning_rate": 9.993735003998934e-06, "loss": 0.2983, "step": 147 }, { "epoch": 0.01, "learning_rate": 9.993601706211678e-06, "loss": 0.3023, "step": 148 }, { "epoch": 0.01, "learning_rate": 9.993468408424421e-06, "loss": 0.3689, "step": 149 }, { "epoch": 0.01, "learning_rate": 9.993335110637164e-06, "loss": 0.2918, "step": 150 }, { "epoch": 0.01, "learning_rate": 9.993201812849908e-06, "loss": 0.3157, "step": 151 }, { "epoch": 0.01, "learning_rate": 9.993068515062651e-06, "loss": 0.3471, "step": 152 }, { "epoch": 0.01, "learning_rate": 9.992935217275394e-06, "loss": 0.2485, "step": 153 }, { "epoch": 0.01, "learning_rate": 9.992801919488137e-06, "loss": 0.2765, "step": 154 }, { "epoch": 0.01, "learning_rate": 9.992668621700881e-06, "loss": 0.2795, "step": 155 }, { "epoch": 0.01, "learning_rate": 9.992535323913624e-06, "loss": 0.3035, "step": 156 }, { "epoch": 0.01, "learning_rate": 9.992402026126367e-06, "loss": 0.3065, "step": 157 }, { "epoch": 0.01, "learning_rate": 9.992268728339111e-06, "loss": 0.2798, "step": 158 }, { "epoch": 0.01, "learning_rate": 9.992135430551854e-06, "loss": 0.2647, "step": 159 }, { "epoch": 0.01, "learning_rate": 9.992002132764597e-06, "loss": 0.3191, "step": 160 }, { "epoch": 0.01, "learning_rate": 9.99186883497734e-06, "loss": 0.2611, "step": 161 }, { "epoch": 0.01, "learning_rate": 9.991735537190084e-06, "loss": 0.2834, "step": 162 }, { "epoch": 0.01, "learning_rate": 9.991602239402827e-06, "loss": 0.3579, "step": 163 }, { "epoch": 0.01, "learning_rate": 9.99146894161557e-06, "loss": 0.2927, "step": 164 }, { "epoch": 0.01, "learning_rate": 9.991335643828315e-06, "loss": 0.3214, "step": 165 }, { "epoch": 0.01, "learning_rate": 9.991202346041057e-06, "loss": 0.2955, "step": 166 }, { "epoch": 0.01, "learning_rate": 9.9910690482538e-06, "loss": 0.304, "step": 167 }, { "epoch": 0.01, "learning_rate": 9.990935750466543e-06, "loss": 0.3134, "step": 168 }, { "epoch": 0.01, "learning_rate": 9.990802452679288e-06, "loss": 0.2993, "step": 169 }, { "epoch": 0.01, "learning_rate": 9.99066915489203e-06, "loss": 0.2541, "step": 170 }, { "epoch": 0.01, "learning_rate": 9.990535857104773e-06, "loss": 0.2284, "step": 171 }, { "epoch": 0.01, "learning_rate": 9.990402559317516e-06, "loss": 0.2952, "step": 172 }, { "epoch": 0.01, "learning_rate": 9.99026926153026e-06, "loss": 0.258, "step": 173 }, { "epoch": 0.01, "learning_rate": 9.990135963743003e-06, "loss": 0.2711, "step": 174 }, { "epoch": 0.01, "learning_rate": 9.990002665955746e-06, "loss": 0.2751, "step": 175 }, { "epoch": 0.01, "learning_rate": 9.989869368168489e-06, "loss": 0.2949, "step": 176 }, { "epoch": 0.01, "learning_rate": 9.989736070381233e-06, "loss": 0.2647, "step": 177 }, { "epoch": 0.01, "learning_rate": 9.989602772593976e-06, "loss": 0.2864, "step": 178 }, { "epoch": 0.01, "learning_rate": 9.989469474806719e-06, "loss": 0.2175, "step": 179 }, { "epoch": 0.01, "learning_rate": 9.989336177019462e-06, "loss": 0.2628, "step": 180 }, { "epoch": 0.01, "learning_rate": 9.989202879232206e-06, "loss": 0.3199, "step": 181 }, { "epoch": 0.01, "learning_rate": 9.98906958144495e-06, "loss": 0.2391, "step": 182 }, { "epoch": 0.01, "learning_rate": 9.988936283657692e-06, "loss": 0.2495, "step": 183 }, { "epoch": 0.01, "learning_rate": 9.988802985870435e-06, "loss": 0.2704, "step": 184 }, { "epoch": 0.01, "learning_rate": 9.98866968808318e-06, "loss": 0.2108, "step": 185 }, { "epoch": 0.01, "learning_rate": 9.988536390295922e-06, "loss": 0.3196, "step": 186 }, { "epoch": 0.01, "learning_rate": 9.988403092508665e-06, "loss": 0.276, "step": 187 }, { "epoch": 0.01, "learning_rate": 9.988269794721408e-06, "loss": 0.3185, "step": 188 }, { "epoch": 0.01, "learning_rate": 9.988136496934152e-06, "loss": 0.234, "step": 189 }, { "epoch": 0.01, "learning_rate": 9.988003199146895e-06, "loss": 0.2377, "step": 190 }, { "epoch": 0.01, "learning_rate": 9.987869901359638e-06, "loss": 0.2593, "step": 191 }, { "epoch": 0.01, "learning_rate": 9.98773660357238e-06, "loss": 0.3055, "step": 192 }, { "epoch": 0.01, "learning_rate": 9.987603305785125e-06, "loss": 0.2497, "step": 193 }, { "epoch": 0.01, "learning_rate": 9.987470007997868e-06, "loss": 0.2585, "step": 194 }, { "epoch": 0.01, "learning_rate": 9.987336710210611e-06, "loss": 0.2356, "step": 195 }, { "epoch": 0.01, "learning_rate": 9.987203412423354e-06, "loss": 0.2526, "step": 196 }, { "epoch": 0.01, "learning_rate": 9.987070114636098e-06, "loss": 0.306, "step": 197 }, { "epoch": 0.01, "learning_rate": 9.986936816848841e-06, "loss": 0.2766, "step": 198 }, { "epoch": 0.01, "learning_rate": 9.986803519061584e-06, "loss": 0.2694, "step": 199 }, { "epoch": 0.01, "learning_rate": 9.986670221274327e-06, "loss": 0.2712, "step": 200 }, { "epoch": 0.01, "learning_rate": 9.986536923487071e-06, "loss": 0.2296, "step": 201 }, { "epoch": 0.01, "learning_rate": 9.986403625699814e-06, "loss": 0.2322, "step": 202 }, { "epoch": 0.01, "learning_rate": 9.986270327912557e-06, "loss": 0.308, "step": 203 }, { "epoch": 0.01, "learning_rate": 9.9861370301253e-06, "loss": 0.2734, "step": 204 }, { "epoch": 0.01, "learning_rate": 9.986003732338044e-06, "loss": 0.373, "step": 205 }, { "epoch": 0.01, "learning_rate": 9.985870434550787e-06, "loss": 0.2614, "step": 206 }, { "epoch": 0.01, "learning_rate": 9.98573713676353e-06, "loss": 0.2577, "step": 207 }, { "epoch": 0.01, "learning_rate": 9.985603838976273e-06, "loss": 0.2611, "step": 208 }, { "epoch": 0.01, "learning_rate": 9.985470541189017e-06, "loss": 0.2841, "step": 209 }, { "epoch": 0.01, "learning_rate": 9.98533724340176e-06, "loss": 0.2529, "step": 210 }, { "epoch": 0.01, "learning_rate": 9.985203945614503e-06, "loss": 0.3231, "step": 211 }, { "epoch": 0.01, "learning_rate": 9.985070647827246e-06, "loss": 0.2378, "step": 212 }, { "epoch": 0.01, "learning_rate": 9.98493735003999e-06, "loss": 0.2943, "step": 213 }, { "epoch": 0.01, "learning_rate": 9.984804052252733e-06, "loss": 0.2543, "step": 214 }, { "epoch": 0.01, "learning_rate": 9.984670754465476e-06, "loss": 0.2852, "step": 215 }, { "epoch": 0.01, "learning_rate": 9.98453745667822e-06, "loss": 0.2654, "step": 216 }, { "epoch": 0.01, "learning_rate": 9.984404158890963e-06, "loss": 0.2501, "step": 217 }, { "epoch": 0.01, "learning_rate": 9.984270861103706e-06, "loss": 0.2566, "step": 218 }, { "epoch": 0.01, "learning_rate": 9.984137563316449e-06, "loss": 0.2683, "step": 219 }, { "epoch": 0.01, "learning_rate": 9.984004265529193e-06, "loss": 0.2632, "step": 220 }, { "epoch": 0.01, "learning_rate": 9.983870967741936e-06, "loss": 0.2654, "step": 221 }, { "epoch": 0.01, "learning_rate": 9.983737669954679e-06, "loss": 0.2296, "step": 222 }, { "epoch": 0.01, "learning_rate": 9.983604372167422e-06, "loss": 0.2424, "step": 223 }, { "epoch": 0.01, "learning_rate": 9.983471074380166e-06, "loss": 0.2829, "step": 224 }, { "epoch": 0.01, "learning_rate": 9.983337776592909e-06, "loss": 0.2626, "step": 225 }, { "epoch": 0.01, "learning_rate": 9.983204478805652e-06, "loss": 0.2232, "step": 226 }, { "epoch": 0.01, "learning_rate": 9.983071181018395e-06, "loss": 0.2274, "step": 227 }, { "epoch": 0.01, "learning_rate": 9.98293788323114e-06, "loss": 0.314, "step": 228 }, { "epoch": 0.01, "learning_rate": 9.982804585443882e-06, "loss": 0.2423, "step": 229 }, { "epoch": 0.01, "learning_rate": 9.982671287656625e-06, "loss": 0.2245, "step": 230 }, { "epoch": 0.01, "learning_rate": 9.98253798986937e-06, "loss": 0.2669, "step": 231 }, { "epoch": 0.01, "learning_rate": 9.982404692082112e-06, "loss": 0.2575, "step": 232 }, { "epoch": 0.01, "learning_rate": 9.982271394294855e-06, "loss": 0.2434, "step": 233 }, { "epoch": 0.01, "learning_rate": 9.982138096507598e-06, "loss": 0.2384, "step": 234 }, { "epoch": 0.01, "learning_rate": 9.982004798720342e-06, "loss": 0.2143, "step": 235 }, { "epoch": 0.01, "learning_rate": 9.981871500933085e-06, "loss": 0.2146, "step": 236 }, { "epoch": 0.01, "learning_rate": 9.981738203145828e-06, "loss": 0.2148, "step": 237 }, { "epoch": 0.01, "learning_rate": 9.981604905358573e-06, "loss": 0.2109, "step": 238 }, { "epoch": 0.01, "learning_rate": 9.981471607571315e-06, "loss": 0.207, "step": 239 }, { "epoch": 0.01, "learning_rate": 9.981338309784058e-06, "loss": 0.2262, "step": 240 }, { "epoch": 0.01, "learning_rate": 9.981205011996801e-06, "loss": 0.2663, "step": 241 }, { "epoch": 0.01, "learning_rate": 9.981071714209545e-06, "loss": 0.2131, "step": 242 }, { "epoch": 0.01, "learning_rate": 9.980938416422288e-06, "loss": 0.2106, "step": 243 }, { "epoch": 0.01, "learning_rate": 9.980805118635031e-06, "loss": 0.2212, "step": 244 }, { "epoch": 0.01, "learning_rate": 9.980671820847776e-06, "loss": 0.2108, "step": 245 }, { "epoch": 0.01, "learning_rate": 9.980538523060518e-06, "loss": 0.3081, "step": 246 }, { "epoch": 0.01, "learning_rate": 9.980405225273261e-06, "loss": 0.2827, "step": 247 }, { "epoch": 0.01, "learning_rate": 9.980271927486004e-06, "loss": 0.2791, "step": 248 }, { "epoch": 0.01, "learning_rate": 9.980138629698749e-06, "loss": 0.2734, "step": 249 }, { "epoch": 0.01, "learning_rate": 9.980005331911491e-06, "loss": 0.244, "step": 250 }, { "epoch": 0.01, "learning_rate": 9.979872034124234e-06, "loss": 0.2329, "step": 251 }, { "epoch": 0.01, "learning_rate": 9.979738736336979e-06, "loss": 0.2721, "step": 252 }, { "epoch": 0.01, "learning_rate": 9.979605438549722e-06, "loss": 0.1978, "step": 253 }, { "epoch": 0.01, "learning_rate": 9.979472140762464e-06, "loss": 0.2301, "step": 254 }, { "epoch": 0.01, "learning_rate": 9.979338842975207e-06, "loss": 0.171, "step": 255 }, { "epoch": 0.01, "learning_rate": 9.979205545187952e-06, "loss": 0.2484, "step": 256 }, { "epoch": 0.01, "learning_rate": 9.979072247400695e-06, "loss": 0.1941, "step": 257 }, { "epoch": 0.01, "learning_rate": 9.978938949613437e-06, "loss": 0.298, "step": 258 }, { "epoch": 0.01, "learning_rate": 9.97880565182618e-06, "loss": 0.2461, "step": 259 }, { "epoch": 0.01, "learning_rate": 9.978672354038925e-06, "loss": 0.2526, "step": 260 }, { "epoch": 0.01, "learning_rate": 9.978539056251668e-06, "loss": 0.1747, "step": 261 }, { "epoch": 0.01, "learning_rate": 9.97840575846441e-06, "loss": 0.2062, "step": 262 }, { "epoch": 0.01, "learning_rate": 9.978272460677153e-06, "loss": 0.2439, "step": 263 }, { "epoch": 0.01, "learning_rate": 9.978139162889898e-06, "loss": 0.273, "step": 264 }, { "epoch": 0.01, "learning_rate": 9.97800586510264e-06, "loss": 0.2838, "step": 265 }, { "epoch": 0.01, "learning_rate": 9.977872567315383e-06, "loss": 0.217, "step": 266 }, { "epoch": 0.01, "learning_rate": 9.977739269528126e-06, "loss": 0.2484, "step": 267 }, { "epoch": 0.01, "learning_rate": 9.97760597174087e-06, "loss": 0.2254, "step": 268 }, { "epoch": 0.01, "learning_rate": 9.977472673953613e-06, "loss": 0.2451, "step": 269 }, { "epoch": 0.01, "learning_rate": 9.977339376166356e-06, "loss": 0.2048, "step": 270 }, { "epoch": 0.01, "learning_rate": 9.9772060783791e-06, "loss": 0.3033, "step": 271 }, { "epoch": 0.01, "learning_rate": 9.977072780591844e-06, "loss": 0.2393, "step": 272 }, { "epoch": 0.01, "learning_rate": 9.976939482804586e-06, "loss": 0.2174, "step": 273 }, { "epoch": 0.01, "learning_rate": 9.97680618501733e-06, "loss": 0.2839, "step": 274 }, { "epoch": 0.01, "learning_rate": 9.976672887230074e-06, "loss": 0.2567, "step": 275 }, { "epoch": 0.01, "learning_rate": 9.976539589442817e-06, "loss": 0.2696, "step": 276 }, { "epoch": 0.01, "learning_rate": 9.97640629165556e-06, "loss": 0.265, "step": 277 }, { "epoch": 0.01, "learning_rate": 9.976272993868302e-06, "loss": 0.2521, "step": 278 }, { "epoch": 0.01, "learning_rate": 9.976139696081047e-06, "loss": 0.2795, "step": 279 }, { "epoch": 0.01, "learning_rate": 9.97600639829379e-06, "loss": 0.2528, "step": 280 }, { "epoch": 0.01, "learning_rate": 9.975873100506532e-06, "loss": 0.2622, "step": 281 }, { "epoch": 0.01, "learning_rate": 9.975739802719275e-06, "loss": 0.2223, "step": 282 }, { "epoch": 0.01, "learning_rate": 9.97560650493202e-06, "loss": 0.1791, "step": 283 }, { "epoch": 0.01, "learning_rate": 9.975473207144763e-06, "loss": 0.1852, "step": 284 }, { "epoch": 0.01, "learning_rate": 9.975339909357505e-06, "loss": 0.208, "step": 285 }, { "epoch": 0.01, "learning_rate": 9.975206611570248e-06, "loss": 0.2116, "step": 286 }, { "epoch": 0.01, "learning_rate": 9.975073313782993e-06, "loss": 0.2073, "step": 287 }, { "epoch": 0.01, "learning_rate": 9.974940015995735e-06, "loss": 0.2499, "step": 288 }, { "epoch": 0.01, "learning_rate": 9.974806718208478e-06, "loss": 0.2036, "step": 289 }, { "epoch": 0.01, "learning_rate": 9.974673420421221e-06, "loss": 0.2002, "step": 290 }, { "epoch": 0.01, "learning_rate": 9.974540122633966e-06, "loss": 0.2174, "step": 291 }, { "epoch": 0.01, "learning_rate": 9.974406824846708e-06, "loss": 0.2696, "step": 292 }, { "epoch": 0.01, "learning_rate": 9.974273527059451e-06, "loss": 0.2361, "step": 293 }, { "epoch": 0.01, "learning_rate": 9.974140229272194e-06, "loss": 0.2007, "step": 294 }, { "epoch": 0.01, "learning_rate": 9.974006931484939e-06, "loss": 0.2188, "step": 295 }, { "epoch": 0.01, "learning_rate": 9.973873633697681e-06, "loss": 0.2078, "step": 296 }, { "epoch": 0.01, "learning_rate": 9.973740335910424e-06, "loss": 0.2195, "step": 297 }, { "epoch": 0.01, "learning_rate": 9.973607038123167e-06, "loss": 0.1892, "step": 298 }, { "epoch": 0.01, "learning_rate": 9.973473740335912e-06, "loss": 0.2222, "step": 299 }, { "epoch": 0.01, "learning_rate": 9.973340442548654e-06, "loss": 0.2315, "step": 300 }, { "epoch": 0.01, "learning_rate": 9.973207144761397e-06, "loss": 0.2185, "step": 301 }, { "epoch": 0.01, "learning_rate": 9.97307384697414e-06, "loss": 0.2148, "step": 302 }, { "epoch": 0.01, "learning_rate": 9.972940549186885e-06, "loss": 0.2228, "step": 303 }, { "epoch": 0.01, "learning_rate": 9.972807251399627e-06, "loss": 0.1759, "step": 304 }, { "epoch": 0.01, "learning_rate": 9.97267395361237e-06, "loss": 0.1981, "step": 305 }, { "epoch": 0.01, "learning_rate": 9.972540655825113e-06, "loss": 0.2366, "step": 306 }, { "epoch": 0.01, "learning_rate": 9.972407358037858e-06, "loss": 0.2247, "step": 307 }, { "epoch": 0.01, "learning_rate": 9.9722740602506e-06, "loss": 0.2328, "step": 308 }, { "epoch": 0.01, "learning_rate": 9.972140762463343e-06, "loss": 0.2712, "step": 309 }, { "epoch": 0.01, "learning_rate": 9.972007464676086e-06, "loss": 0.2111, "step": 310 }, { "epoch": 0.01, "learning_rate": 9.97187416688883e-06, "loss": 0.2068, "step": 311 }, { "epoch": 0.01, "learning_rate": 9.971740869101573e-06, "loss": 0.1966, "step": 312 }, { "epoch": 0.01, "learning_rate": 9.971607571314316e-06, "loss": 0.2331, "step": 313 }, { "epoch": 0.01, "learning_rate": 9.971474273527059e-06, "loss": 0.1854, "step": 314 }, { "epoch": 0.01, "learning_rate": 9.971340975739803e-06, "loss": 0.2479, "step": 315 }, { "epoch": 0.01, "learning_rate": 9.971207677952546e-06, "loss": 0.2219, "step": 316 }, { "epoch": 0.01, "learning_rate": 9.971074380165289e-06, "loss": 0.1735, "step": 317 }, { "epoch": 0.01, "learning_rate": 9.970941082378034e-06, "loss": 0.2171, "step": 318 }, { "epoch": 0.01, "learning_rate": 9.970807784590776e-06, "loss": 0.1955, "step": 319 }, { "epoch": 0.01, "learning_rate": 9.97067448680352e-06, "loss": 0.1766, "step": 320 }, { "epoch": 0.01, "learning_rate": 9.970541189016262e-06, "loss": 0.2598, "step": 321 }, { "epoch": 0.01, "learning_rate": 9.970407891229007e-06, "loss": 0.1983, "step": 322 }, { "epoch": 0.01, "learning_rate": 9.97027459344175e-06, "loss": 0.2088, "step": 323 }, { "epoch": 0.01, "learning_rate": 9.970141295654492e-06, "loss": 0.2033, "step": 324 }, { "epoch": 0.01, "learning_rate": 9.970007997867237e-06, "loss": 0.2359, "step": 325 }, { "epoch": 0.01, "learning_rate": 9.96987470007998e-06, "loss": 0.2355, "step": 326 }, { "epoch": 0.01, "learning_rate": 9.969741402292722e-06, "loss": 0.2323, "step": 327 }, { "epoch": 0.01, "learning_rate": 9.969608104505465e-06, "loss": 0.2265, "step": 328 }, { "epoch": 0.01, "learning_rate": 9.96947480671821e-06, "loss": 0.2111, "step": 329 }, { "epoch": 0.01, "learning_rate": 9.969341508930953e-06, "loss": 0.2175, "step": 330 }, { "epoch": 0.01, "learning_rate": 9.969208211143695e-06, "loss": 0.2193, "step": 331 }, { "epoch": 0.01, "learning_rate": 9.96907491335644e-06, "loss": 0.2515, "step": 332 }, { "epoch": 0.01, "learning_rate": 9.968941615569183e-06, "loss": 0.1859, "step": 333 }, { "epoch": 0.01, "learning_rate": 9.968808317781925e-06, "loss": 0.236, "step": 334 }, { "epoch": 0.01, "learning_rate": 9.968675019994668e-06, "loss": 0.2299, "step": 335 }, { "epoch": 0.01, "learning_rate": 9.968541722207413e-06, "loss": 0.2232, "step": 336 }, { "epoch": 0.01, "learning_rate": 9.968408424420156e-06, "loss": 0.1819, "step": 337 }, { "epoch": 0.01, "learning_rate": 9.968275126632898e-06, "loss": 0.1866, "step": 338 }, { "epoch": 0.01, "learning_rate": 9.968141828845643e-06, "loss": 0.2574, "step": 339 }, { "epoch": 0.01, "learning_rate": 9.968008531058386e-06, "loss": 0.2325, "step": 340 }, { "epoch": 0.01, "learning_rate": 9.967875233271129e-06, "loss": 0.2212, "step": 341 }, { "epoch": 0.01, "learning_rate": 9.967741935483871e-06, "loss": 0.2192, "step": 342 }, { "epoch": 0.01, "learning_rate": 9.967608637696616e-06, "loss": 0.1918, "step": 343 }, { "epoch": 0.01, "learning_rate": 9.967475339909359e-06, "loss": 0.2243, "step": 344 }, { "epoch": 0.01, "learning_rate": 9.967342042122102e-06, "loss": 0.2391, "step": 345 }, { "epoch": 0.01, "learning_rate": 9.967208744334846e-06, "loss": 0.2287, "step": 346 }, { "epoch": 0.01, "learning_rate": 9.967075446547589e-06, "loss": 0.192, "step": 347 }, { "epoch": 0.01, "learning_rate": 9.966942148760332e-06, "loss": 0.2215, "step": 348 }, { "epoch": 0.01, "learning_rate": 9.966808850973075e-06, "loss": 0.1705, "step": 349 }, { "epoch": 0.01, "learning_rate": 9.966675553185819e-06, "loss": 0.1775, "step": 350 }, { "epoch": 0.01, "learning_rate": 9.966542255398562e-06, "loss": 0.2048, "step": 351 }, { "epoch": 0.01, "learning_rate": 9.966408957611305e-06, "loss": 0.2141, "step": 352 }, { "epoch": 0.01, "learning_rate": 9.966275659824048e-06, "loss": 0.2265, "step": 353 }, { "epoch": 0.01, "learning_rate": 9.966142362036792e-06, "loss": 0.2002, "step": 354 }, { "epoch": 0.01, "learning_rate": 9.966009064249535e-06, "loss": 0.2152, "step": 355 }, { "epoch": 0.01, "learning_rate": 9.965875766462278e-06, "loss": 0.1991, "step": 356 }, { "epoch": 0.01, "learning_rate": 9.96574246867502e-06, "loss": 0.1985, "step": 357 }, { "epoch": 0.01, "learning_rate": 9.965609170887765e-06, "loss": 0.2076, "step": 358 }, { "epoch": 0.01, "learning_rate": 9.965475873100508e-06, "loss": 0.2167, "step": 359 }, { "epoch": 0.01, "learning_rate": 9.96534257531325e-06, "loss": 0.239, "step": 360 }, { "epoch": 0.01, "learning_rate": 9.965209277525993e-06, "loss": 0.1731, "step": 361 }, { "epoch": 0.01, "learning_rate": 9.965075979738738e-06, "loss": 0.2006, "step": 362 }, { "epoch": 0.01, "learning_rate": 9.96494268195148e-06, "loss": 0.1882, "step": 363 }, { "epoch": 0.01, "learning_rate": 9.964809384164224e-06, "loss": 0.2155, "step": 364 }, { "epoch": 0.01, "learning_rate": 9.964676086376966e-06, "loss": 0.1793, "step": 365 }, { "epoch": 0.01, "learning_rate": 9.964542788589711e-06, "loss": 0.2205, "step": 366 }, { "epoch": 0.01, "learning_rate": 9.964409490802454e-06, "loss": 0.1717, "step": 367 }, { "epoch": 0.01, "learning_rate": 9.964276193015197e-06, "loss": 0.1876, "step": 368 }, { "epoch": 0.01, "learning_rate": 9.96414289522794e-06, "loss": 0.1951, "step": 369 }, { "epoch": 0.01, "learning_rate": 9.964009597440684e-06, "loss": 0.1272, "step": 370 }, { "epoch": 0.01, "learning_rate": 9.963876299653427e-06, "loss": 0.1715, "step": 371 }, { "epoch": 0.01, "learning_rate": 9.96374300186617e-06, "loss": 0.1494, "step": 372 }, { "epoch": 0.01, "learning_rate": 9.963609704078912e-06, "loss": 0.1957, "step": 373 }, { "epoch": 0.01, "learning_rate": 9.963476406291657e-06, "loss": 0.2448, "step": 374 }, { "epoch": 0.01, "learning_rate": 9.9633431085044e-06, "loss": 0.1905, "step": 375 }, { "epoch": 0.02, "learning_rate": 9.963209810717143e-06, "loss": 0.2125, "step": 376 }, { "epoch": 0.02, "learning_rate": 9.963076512929885e-06, "loss": 0.2128, "step": 377 }, { "epoch": 0.02, "learning_rate": 9.96294321514263e-06, "loss": 0.2136, "step": 378 }, { "epoch": 0.02, "learning_rate": 9.962809917355373e-06, "loss": 0.216, "step": 379 }, { "epoch": 0.02, "learning_rate": 9.962676619568116e-06, "loss": 0.1928, "step": 380 }, { "epoch": 0.02, "learning_rate": 9.962543321780858e-06, "loss": 0.2432, "step": 381 }, { "epoch": 0.02, "learning_rate": 9.962410023993603e-06, "loss": 0.1724, "step": 382 }, { "epoch": 0.02, "learning_rate": 9.962276726206346e-06, "loss": 0.2409, "step": 383 }, { "epoch": 0.02, "learning_rate": 9.962143428419088e-06, "loss": 0.1632, "step": 384 }, { "epoch": 0.02, "learning_rate": 9.962010130631833e-06, "loss": 0.1845, "step": 385 }, { "epoch": 0.02, "learning_rate": 9.961876832844576e-06, "loss": 0.2127, "step": 386 }, { "epoch": 0.02, "learning_rate": 9.961743535057319e-06, "loss": 0.2163, "step": 387 }, { "epoch": 0.02, "learning_rate": 9.961610237270061e-06, "loss": 0.2409, "step": 388 }, { "epoch": 0.02, "learning_rate": 9.961476939482806e-06, "loss": 0.2279, "step": 389 }, { "epoch": 0.02, "learning_rate": 9.961343641695549e-06, "loss": 0.2303, "step": 390 }, { "epoch": 0.02, "learning_rate": 9.961210343908292e-06, "loss": 0.2285, "step": 391 }, { "epoch": 0.02, "learning_rate": 9.961077046121034e-06, "loss": 0.1884, "step": 392 }, { "epoch": 0.02, "learning_rate": 9.960943748333779e-06, "loss": 0.1826, "step": 393 }, { "epoch": 0.02, "learning_rate": 9.960810450546522e-06, "loss": 0.192, "step": 394 }, { "epoch": 0.02, "learning_rate": 9.960677152759265e-06, "loss": 0.2114, "step": 395 }, { "epoch": 0.02, "learning_rate": 9.960543854972007e-06, "loss": 0.1949, "step": 396 }, { "epoch": 0.02, "learning_rate": 9.960410557184752e-06, "loss": 0.1871, "step": 397 }, { "epoch": 0.02, "learning_rate": 9.960277259397495e-06, "loss": 0.1848, "step": 398 }, { "epoch": 0.02, "learning_rate": 9.960143961610238e-06, "loss": 0.1901, "step": 399 }, { "epoch": 0.02, "learning_rate": 9.96001066382298e-06, "loss": 0.2432, "step": 400 }, { "epoch": 0.02, "learning_rate": 9.959877366035725e-06, "loss": 0.2295, "step": 401 }, { "epoch": 0.02, "learning_rate": 9.959744068248468e-06, "loss": 0.2045, "step": 402 }, { "epoch": 0.02, "learning_rate": 9.95961077046121e-06, "loss": 0.2009, "step": 403 }, { "epoch": 0.02, "learning_rate": 9.959477472673953e-06, "loss": 0.2044, "step": 404 }, { "epoch": 0.02, "learning_rate": 9.959344174886698e-06, "loss": 0.2251, "step": 405 }, { "epoch": 0.02, "learning_rate": 9.95921087709944e-06, "loss": 0.2486, "step": 406 }, { "epoch": 0.02, "learning_rate": 9.959077579312183e-06, "loss": 0.1897, "step": 407 }, { "epoch": 0.02, "learning_rate": 9.958944281524926e-06, "loss": 0.2208, "step": 408 }, { "epoch": 0.02, "learning_rate": 9.95881098373767e-06, "loss": 0.2308, "step": 409 }, { "epoch": 0.02, "learning_rate": 9.958677685950414e-06, "loss": 0.2256, "step": 410 }, { "epoch": 0.02, "learning_rate": 9.958544388163156e-06, "loss": 0.1759, "step": 411 }, { "epoch": 0.02, "learning_rate": 9.958411090375901e-06, "loss": 0.2234, "step": 412 }, { "epoch": 0.02, "learning_rate": 9.958277792588644e-06, "loss": 0.2309, "step": 413 }, { "epoch": 0.02, "learning_rate": 9.958144494801387e-06, "loss": 0.1627, "step": 414 }, { "epoch": 0.02, "learning_rate": 9.95801119701413e-06, "loss": 0.1762, "step": 415 }, { "epoch": 0.02, "learning_rate": 9.957877899226874e-06, "loss": 0.2459, "step": 416 }, { "epoch": 0.02, "learning_rate": 9.957744601439617e-06, "loss": 0.2292, "step": 417 }, { "epoch": 0.02, "learning_rate": 9.95761130365236e-06, "loss": 0.1528, "step": 418 }, { "epoch": 0.02, "learning_rate": 9.957478005865104e-06, "loss": 0.204, "step": 419 }, { "epoch": 0.02, "learning_rate": 9.957344708077847e-06, "loss": 0.2003, "step": 420 }, { "epoch": 0.02, "learning_rate": 9.95721141029059e-06, "loss": 0.166, "step": 421 }, { "epoch": 0.02, "learning_rate": 9.957078112503333e-06, "loss": 0.1674, "step": 422 }, { "epoch": 0.02, "learning_rate": 9.956944814716077e-06, "loss": 0.1642, "step": 423 }, { "epoch": 0.02, "learning_rate": 9.95681151692882e-06, "loss": 0.1706, "step": 424 }, { "epoch": 0.02, "learning_rate": 9.956678219141563e-06, "loss": 0.1557, "step": 425 }, { "epoch": 0.02, "learning_rate": 9.956544921354307e-06, "loss": 0.1716, "step": 426 }, { "epoch": 0.02, "learning_rate": 9.95641162356705e-06, "loss": 0.1799, "step": 427 }, { "epoch": 0.02, "learning_rate": 9.956278325779793e-06, "loss": 0.1879, "step": 428 }, { "epoch": 0.02, "learning_rate": 9.956145027992536e-06, "loss": 0.2326, "step": 429 }, { "epoch": 0.02, "learning_rate": 9.95601173020528e-06, "loss": 0.2191, "step": 430 }, { "epoch": 0.02, "learning_rate": 9.955878432418023e-06, "loss": 0.2121, "step": 431 }, { "epoch": 0.02, "learning_rate": 9.955745134630766e-06, "loss": 0.2485, "step": 432 }, { "epoch": 0.02, "learning_rate": 9.95561183684351e-06, "loss": 0.2327, "step": 433 }, { "epoch": 0.02, "learning_rate": 9.955478539056253e-06, "loss": 0.1954, "step": 434 }, { "epoch": 0.02, "learning_rate": 9.955345241268996e-06, "loss": 0.1432, "step": 435 }, { "epoch": 0.02, "learning_rate": 9.955211943481739e-06, "loss": 0.2018, "step": 436 }, { "epoch": 0.02, "learning_rate": 9.955078645694483e-06, "loss": 0.1769, "step": 437 }, { "epoch": 0.02, "learning_rate": 9.954945347907226e-06, "loss": 0.205, "step": 438 }, { "epoch": 0.02, "learning_rate": 9.954812050119969e-06, "loss": 0.1834, "step": 439 }, { "epoch": 0.02, "learning_rate": 9.954678752332713e-06, "loss": 0.1641, "step": 440 }, { "epoch": 0.02, "learning_rate": 9.954545454545456e-06, "loss": 0.2168, "step": 441 }, { "epoch": 0.02, "learning_rate": 9.954412156758199e-06, "loss": 0.2191, "step": 442 }, { "epoch": 0.02, "learning_rate": 9.954278858970942e-06, "loss": 0.236, "step": 443 }, { "epoch": 0.02, "learning_rate": 9.954145561183686e-06, "loss": 0.2504, "step": 444 }, { "epoch": 0.02, "learning_rate": 9.95401226339643e-06, "loss": 0.1767, "step": 445 }, { "epoch": 0.02, "learning_rate": 9.953878965609172e-06, "loss": 0.2016, "step": 446 }, { "epoch": 0.02, "learning_rate": 9.953745667821915e-06, "loss": 0.139, "step": 447 }, { "epoch": 0.02, "learning_rate": 9.95361237003466e-06, "loss": 0.1341, "step": 448 }, { "epoch": 0.02, "learning_rate": 9.953479072247402e-06, "loss": 0.2284, "step": 449 }, { "epoch": 0.02, "learning_rate": 9.953345774460145e-06, "loss": 0.168, "step": 450 }, { "epoch": 0.02, "learning_rate": 9.953212476672888e-06, "loss": 0.1672, "step": 451 }, { "epoch": 0.02, "learning_rate": 9.953079178885632e-06, "loss": 0.238, "step": 452 }, { "epoch": 0.02, "learning_rate": 9.952945881098375e-06, "loss": 0.1706, "step": 453 }, { "epoch": 0.02, "learning_rate": 9.952812583311118e-06, "loss": 0.1689, "step": 454 }, { "epoch": 0.02, "learning_rate": 9.95267928552386e-06, "loss": 0.2128, "step": 455 }, { "epoch": 0.02, "learning_rate": 9.952545987736605e-06, "loss": 0.2185, "step": 456 }, { "epoch": 0.02, "learning_rate": 9.952412689949348e-06, "loss": 0.1851, "step": 457 }, { "epoch": 0.02, "learning_rate": 9.952279392162091e-06, "loss": 0.1793, "step": 458 }, { "epoch": 0.02, "learning_rate": 9.952146094374834e-06, "loss": 0.1969, "step": 459 }, { "epoch": 0.02, "learning_rate": 9.952012796587578e-06, "loss": 0.17, "step": 460 }, { "epoch": 0.02, "learning_rate": 9.951879498800321e-06, "loss": 0.1812, "step": 461 }, { "epoch": 0.02, "learning_rate": 9.951746201013064e-06, "loss": 0.1882, "step": 462 }, { "epoch": 0.02, "learning_rate": 9.951612903225807e-06, "loss": 0.1678, "step": 463 }, { "epoch": 0.02, "learning_rate": 9.951479605438551e-06, "loss": 0.2108, "step": 464 }, { "epoch": 0.02, "learning_rate": 9.951346307651294e-06, "loss": 0.1847, "step": 465 }, { "epoch": 0.02, "learning_rate": 9.951213009864037e-06, "loss": 0.1964, "step": 466 }, { "epoch": 0.02, "learning_rate": 9.95107971207678e-06, "loss": 0.2409, "step": 467 }, { "epoch": 0.02, "learning_rate": 9.950946414289524e-06, "loss": 0.1728, "step": 468 }, { "epoch": 0.02, "learning_rate": 9.950813116502267e-06, "loss": 0.1888, "step": 469 }, { "epoch": 0.02, "learning_rate": 9.95067981871501e-06, "loss": 0.201, "step": 470 }, { "epoch": 0.02, "learning_rate": 9.950546520927753e-06, "loss": 0.1965, "step": 471 }, { "epoch": 0.02, "learning_rate": 9.950413223140497e-06, "loss": 0.1726, "step": 472 }, { "epoch": 0.02, "learning_rate": 9.95027992535324e-06, "loss": 0.1448, "step": 473 }, { "epoch": 0.02, "learning_rate": 9.950146627565983e-06, "loss": 0.1819, "step": 474 }, { "epoch": 0.02, "learning_rate": 9.950013329778726e-06, "loss": 0.1928, "step": 475 }, { "epoch": 0.02, "learning_rate": 9.94988003199147e-06, "loss": 0.1982, "step": 476 }, { "epoch": 0.02, "learning_rate": 9.949746734204213e-06, "loss": 0.2022, "step": 477 }, { "epoch": 0.02, "learning_rate": 9.949613436416956e-06, "loss": 0.1753, "step": 478 }, { "epoch": 0.02, "learning_rate": 9.949480138629699e-06, "loss": 0.2047, "step": 479 }, { "epoch": 0.02, "learning_rate": 9.949346840842443e-06, "loss": 0.1592, "step": 480 }, { "epoch": 0.02, "learning_rate": 9.949213543055186e-06, "loss": 0.2255, "step": 481 }, { "epoch": 0.02, "learning_rate": 9.949080245267929e-06, "loss": 0.1797, "step": 482 }, { "epoch": 0.02, "learning_rate": 9.948946947480672e-06, "loss": 0.2411, "step": 483 }, { "epoch": 0.02, "learning_rate": 9.948813649693416e-06, "loss": 0.2061, "step": 484 }, { "epoch": 0.02, "learning_rate": 9.948680351906159e-06, "loss": 0.2177, "step": 485 }, { "epoch": 0.02, "learning_rate": 9.948547054118902e-06, "loss": 0.2384, "step": 486 }, { "epoch": 0.02, "learning_rate": 9.948413756331645e-06, "loss": 0.1846, "step": 487 }, { "epoch": 0.02, "learning_rate": 9.948280458544389e-06, "loss": 0.241, "step": 488 }, { "epoch": 0.02, "learning_rate": 9.948147160757132e-06, "loss": 0.2319, "step": 489 }, { "epoch": 0.02, "learning_rate": 9.948013862969875e-06, "loss": 0.1587, "step": 490 }, { "epoch": 0.02, "learning_rate": 9.947880565182618e-06, "loss": 0.1702, "step": 491 }, { "epoch": 0.02, "learning_rate": 9.947747267395362e-06, "loss": 0.183, "step": 492 }, { "epoch": 0.02, "learning_rate": 9.947613969608105e-06, "loss": 0.1297, "step": 493 }, { "epoch": 0.02, "learning_rate": 9.947480671820848e-06, "loss": 0.2208, "step": 494 }, { "epoch": 0.02, "learning_rate": 9.947347374033592e-06, "loss": 0.2033, "step": 495 }, { "epoch": 0.02, "learning_rate": 9.947214076246335e-06, "loss": 0.179, "step": 496 }, { "epoch": 0.02, "learning_rate": 9.947080778459078e-06, "loss": 0.204, "step": 497 }, { "epoch": 0.02, "learning_rate": 9.94694748067182e-06, "loss": 0.1981, "step": 498 }, { "epoch": 0.02, "learning_rate": 9.946814182884565e-06, "loss": 0.175, "step": 499 }, { "epoch": 0.02, "learning_rate": 9.946680885097308e-06, "loss": 0.239, "step": 500 }, { "epoch": 0.02, "learning_rate": 9.94654758731005e-06, "loss": 0.1766, "step": 501 }, { "epoch": 0.02, "learning_rate": 9.946414289522794e-06, "loss": 0.2208, "step": 502 }, { "epoch": 0.02, "learning_rate": 9.946280991735538e-06, "loss": 0.1757, "step": 503 }, { "epoch": 0.02, "learning_rate": 9.946147693948281e-06, "loss": 0.1888, "step": 504 }, { "epoch": 0.02, "learning_rate": 9.946014396161024e-06, "loss": 0.1662, "step": 505 }, { "epoch": 0.02, "learning_rate": 9.945881098373768e-06, "loss": 0.167, "step": 506 }, { "epoch": 0.02, "learning_rate": 9.945747800586511e-06, "loss": 0.179, "step": 507 }, { "epoch": 0.02, "learning_rate": 9.945614502799254e-06, "loss": 0.2478, "step": 508 }, { "epoch": 0.02, "learning_rate": 9.945481205011997e-06, "loss": 0.1778, "step": 509 }, { "epoch": 0.02, "learning_rate": 9.945347907224741e-06, "loss": 0.1596, "step": 510 }, { "epoch": 0.02, "learning_rate": 9.945214609437484e-06, "loss": 0.1743, "step": 511 }, { "epoch": 0.02, "learning_rate": 9.945081311650227e-06, "loss": 0.1909, "step": 512 }, { "epoch": 0.02, "learning_rate": 9.944948013862971e-06, "loss": 0.2093, "step": 513 }, { "epoch": 0.02, "learning_rate": 9.944814716075714e-06, "loss": 0.2196, "step": 514 }, { "epoch": 0.02, "learning_rate": 9.944681418288457e-06, "loss": 0.1832, "step": 515 }, { "epoch": 0.02, "learning_rate": 9.9445481205012e-06, "loss": 0.1884, "step": 516 }, { "epoch": 0.02, "learning_rate": 9.944414822713944e-06, "loss": 0.1987, "step": 517 }, { "epoch": 0.02, "learning_rate": 9.944281524926687e-06, "loss": 0.1826, "step": 518 }, { "epoch": 0.02, "learning_rate": 9.94414822713943e-06, "loss": 0.2217, "step": 519 }, { "epoch": 0.02, "learning_rate": 9.944014929352175e-06, "loss": 0.1992, "step": 520 }, { "epoch": 0.02, "learning_rate": 9.943881631564917e-06, "loss": 0.1945, "step": 521 }, { "epoch": 0.02, "learning_rate": 9.94374833377766e-06, "loss": 0.1686, "step": 522 }, { "epoch": 0.02, "learning_rate": 9.943615035990403e-06, "loss": 0.1931, "step": 523 }, { "epoch": 0.02, "learning_rate": 9.943481738203148e-06, "loss": 0.1533, "step": 524 }, { "epoch": 0.02, "learning_rate": 9.94334844041589e-06, "loss": 0.211, "step": 525 }, { "epoch": 0.02, "learning_rate": 9.943215142628633e-06, "loss": 0.2226, "step": 526 }, { "epoch": 0.02, "learning_rate": 9.943081844841378e-06, "loss": 0.1583, "step": 527 }, { "epoch": 0.02, "learning_rate": 9.94294854705412e-06, "loss": 0.2328, "step": 528 }, { "epoch": 0.02, "learning_rate": 9.942815249266863e-06, "loss": 0.2064, "step": 529 }, { "epoch": 0.02, "learning_rate": 9.942681951479606e-06, "loss": 0.2281, "step": 530 }, { "epoch": 0.02, "learning_rate": 9.94254865369235e-06, "loss": 0.1782, "step": 531 }, { "epoch": 0.02, "learning_rate": 9.942415355905093e-06, "loss": 0.1613, "step": 532 }, { "epoch": 0.02, "learning_rate": 9.942282058117836e-06, "loss": 0.1661, "step": 533 }, { "epoch": 0.02, "learning_rate": 9.942148760330579e-06, "loss": 0.1514, "step": 534 }, { "epoch": 0.02, "learning_rate": 9.942015462543324e-06, "loss": 0.2058, "step": 535 }, { "epoch": 0.02, "learning_rate": 9.941882164756066e-06, "loss": 0.2076, "step": 536 }, { "epoch": 0.02, "learning_rate": 9.94174886696881e-06, "loss": 0.1974, "step": 537 }, { "epoch": 0.02, "learning_rate": 9.941615569181552e-06, "loss": 0.2201, "step": 538 }, { "epoch": 0.02, "learning_rate": 9.941482271394297e-06, "loss": 0.1856, "step": 539 }, { "epoch": 0.02, "learning_rate": 9.94134897360704e-06, "loss": 0.2188, "step": 540 }, { "epoch": 0.02, "learning_rate": 9.941215675819782e-06, "loss": 0.1779, "step": 541 }, { "epoch": 0.02, "learning_rate": 9.941082378032525e-06, "loss": 0.1967, "step": 542 }, { "epoch": 0.02, "learning_rate": 9.94094908024527e-06, "loss": 0.1733, "step": 543 }, { "epoch": 0.02, "learning_rate": 9.940815782458012e-06, "loss": 0.1181, "step": 544 }, { "epoch": 0.02, "learning_rate": 9.940682484670755e-06, "loss": 0.2046, "step": 545 }, { "epoch": 0.02, "learning_rate": 9.940549186883498e-06, "loss": 0.1776, "step": 546 }, { "epoch": 0.02, "learning_rate": 9.940415889096243e-06, "loss": 0.1466, "step": 547 }, { "epoch": 0.02, "learning_rate": 9.940282591308985e-06, "loss": 0.2233, "step": 548 }, { "epoch": 0.02, "learning_rate": 9.940149293521728e-06, "loss": 0.1821, "step": 549 }, { "epoch": 0.02, "learning_rate": 9.940015995734471e-06, "loss": 0.1532, "step": 550 }, { "epoch": 0.02, "learning_rate": 9.939882697947215e-06, "loss": 0.1437, "step": 551 }, { "epoch": 0.02, "learning_rate": 9.939749400159958e-06, "loss": 0.1371, "step": 552 }, { "epoch": 0.02, "learning_rate": 9.939616102372701e-06, "loss": 0.1858, "step": 553 }, { "epoch": 0.02, "learning_rate": 9.939482804585446e-06, "loss": 0.1922, "step": 554 }, { "epoch": 0.02, "learning_rate": 9.939349506798188e-06, "loss": 0.1766, "step": 555 }, { "epoch": 0.02, "learning_rate": 9.939216209010931e-06, "loss": 0.1887, "step": 556 }, { "epoch": 0.02, "learning_rate": 9.939082911223674e-06, "loss": 0.1168, "step": 557 }, { "epoch": 0.02, "learning_rate": 9.938949613436419e-06, "loss": 0.2044, "step": 558 }, { "epoch": 0.02, "learning_rate": 9.938816315649161e-06, "loss": 0.1898, "step": 559 }, { "epoch": 0.02, "learning_rate": 9.938683017861904e-06, "loss": 0.1608, "step": 560 }, { "epoch": 0.02, "learning_rate": 9.938549720074647e-06, "loss": 0.1969, "step": 561 }, { "epoch": 0.02, "learning_rate": 9.938416422287392e-06, "loss": 0.2167, "step": 562 }, { "epoch": 0.02, "learning_rate": 9.938283124500134e-06, "loss": 0.1404, "step": 563 }, { "epoch": 0.02, "learning_rate": 9.938149826712877e-06, "loss": 0.1834, "step": 564 }, { "epoch": 0.02, "learning_rate": 9.93801652892562e-06, "loss": 0.183, "step": 565 }, { "epoch": 0.02, "learning_rate": 9.937883231138365e-06, "loss": 0.1832, "step": 566 }, { "epoch": 0.02, "learning_rate": 9.937749933351107e-06, "loss": 0.1732, "step": 567 }, { "epoch": 0.02, "learning_rate": 9.93761663556385e-06, "loss": 0.1467, "step": 568 }, { "epoch": 0.02, "learning_rate": 9.937483337776593e-06, "loss": 0.1815, "step": 569 }, { "epoch": 0.02, "learning_rate": 9.937350039989338e-06, "loss": 0.2087, "step": 570 }, { "epoch": 0.02, "learning_rate": 9.93721674220208e-06, "loss": 0.1751, "step": 571 }, { "epoch": 0.02, "learning_rate": 9.937083444414823e-06, "loss": 0.1982, "step": 572 }, { "epoch": 0.02, "learning_rate": 9.936950146627566e-06, "loss": 0.1867, "step": 573 }, { "epoch": 0.02, "learning_rate": 9.93681684884031e-06, "loss": 0.2222, "step": 574 }, { "epoch": 0.02, "learning_rate": 9.936683551053053e-06, "loss": 0.1978, "step": 575 }, { "epoch": 0.02, "learning_rate": 9.936550253265796e-06, "loss": 0.1563, "step": 576 }, { "epoch": 0.02, "learning_rate": 9.936416955478539e-06, "loss": 0.1833, "step": 577 }, { "epoch": 0.02, "learning_rate": 9.936283657691283e-06, "loss": 0.1777, "step": 578 }, { "epoch": 0.02, "learning_rate": 9.936150359904026e-06, "loss": 0.1659, "step": 579 }, { "epoch": 0.02, "learning_rate": 9.936017062116769e-06, "loss": 0.1671, "step": 580 }, { "epoch": 0.02, "learning_rate": 9.935883764329512e-06, "loss": 0.1773, "step": 581 }, { "epoch": 0.02, "learning_rate": 9.935750466542256e-06, "loss": 0.1646, "step": 582 }, { "epoch": 0.02, "learning_rate": 9.935617168755e-06, "loss": 0.1771, "step": 583 }, { "epoch": 0.02, "learning_rate": 9.935483870967742e-06, "loss": 0.2103, "step": 584 }, { "epoch": 0.02, "learning_rate": 9.935350573180485e-06, "loss": 0.1426, "step": 585 }, { "epoch": 0.02, "learning_rate": 9.93521727539323e-06, "loss": 0.2256, "step": 586 }, { "epoch": 0.02, "learning_rate": 9.935083977605972e-06, "loss": 0.1647, "step": 587 }, { "epoch": 0.02, "learning_rate": 9.934950679818715e-06, "loss": 0.1421, "step": 588 }, { "epoch": 0.02, "learning_rate": 9.934817382031458e-06, "loss": 0.1872, "step": 589 }, { "epoch": 0.02, "learning_rate": 9.934684084244202e-06, "loss": 0.2479, "step": 590 }, { "epoch": 0.02, "learning_rate": 9.934550786456945e-06, "loss": 0.1702, "step": 591 }, { "epoch": 0.02, "learning_rate": 9.934417488669688e-06, "loss": 0.1582, "step": 592 }, { "epoch": 0.02, "learning_rate": 9.934284190882433e-06, "loss": 0.1654, "step": 593 }, { "epoch": 0.02, "learning_rate": 9.934150893095175e-06, "loss": 0.1743, "step": 594 }, { "epoch": 0.02, "learning_rate": 9.934017595307918e-06, "loss": 0.1397, "step": 595 }, { "epoch": 0.02, "learning_rate": 9.933884297520661e-06, "loss": 0.1599, "step": 596 }, { "epoch": 0.02, "learning_rate": 9.933750999733405e-06, "loss": 0.2328, "step": 597 }, { "epoch": 0.02, "learning_rate": 9.933617701946148e-06, "loss": 0.2086, "step": 598 }, { "epoch": 0.02, "learning_rate": 9.933484404158891e-06, "loss": 0.154, "step": 599 }, { "epoch": 0.02, "learning_rate": 9.933351106371636e-06, "loss": 0.155, "step": 600 }, { "epoch": 0.02, "learning_rate": 9.933217808584378e-06, "loss": 0.1941, "step": 601 }, { "epoch": 0.02, "learning_rate": 9.933084510797121e-06, "loss": 0.1988, "step": 602 }, { "epoch": 0.02, "learning_rate": 9.932951213009864e-06, "loss": 0.1939, "step": 603 }, { "epoch": 0.02, "learning_rate": 9.932817915222609e-06, "loss": 0.2051, "step": 604 }, { "epoch": 0.02, "learning_rate": 9.932684617435351e-06, "loss": 0.1781, "step": 605 }, { "epoch": 0.02, "learning_rate": 9.932551319648094e-06, "loss": 0.1732, "step": 606 }, { "epoch": 0.02, "learning_rate": 9.932418021860839e-06, "loss": 0.1651, "step": 607 }, { "epoch": 0.02, "learning_rate": 9.932284724073582e-06, "loss": 0.1776, "step": 608 }, { "epoch": 0.02, "learning_rate": 9.932151426286324e-06, "loss": 0.1871, "step": 609 }, { "epoch": 0.02, "learning_rate": 9.932018128499067e-06, "loss": 0.1868, "step": 610 }, { "epoch": 0.02, "learning_rate": 9.931884830711812e-06, "loss": 0.2196, "step": 611 }, { "epoch": 0.02, "learning_rate": 9.931751532924555e-06, "loss": 0.198, "step": 612 }, { "epoch": 0.02, "learning_rate": 9.931618235137297e-06, "loss": 0.1652, "step": 613 }, { "epoch": 0.02, "learning_rate": 9.931484937350042e-06, "loss": 0.1611, "step": 614 }, { "epoch": 0.02, "learning_rate": 9.931351639562785e-06, "loss": 0.1824, "step": 615 }, { "epoch": 0.02, "learning_rate": 9.931218341775528e-06, "loss": 0.2027, "step": 616 }, { "epoch": 0.02, "learning_rate": 9.93108504398827e-06, "loss": 0.1958, "step": 617 }, { "epoch": 0.02, "learning_rate": 9.930951746201015e-06, "loss": 0.237, "step": 618 }, { "epoch": 0.02, "learning_rate": 9.930818448413758e-06, "loss": 0.1859, "step": 619 }, { "epoch": 0.02, "learning_rate": 9.9306851506265e-06, "loss": 0.1719, "step": 620 }, { "epoch": 0.02, "learning_rate": 9.930551852839245e-06, "loss": 0.2071, "step": 621 }, { "epoch": 0.02, "learning_rate": 9.930418555051988e-06, "loss": 0.1644, "step": 622 }, { "epoch": 0.02, "learning_rate": 9.93028525726473e-06, "loss": 0.1656, "step": 623 }, { "epoch": 0.02, "learning_rate": 9.930151959477473e-06, "loss": 0.1571, "step": 624 }, { "epoch": 0.02, "learning_rate": 9.930018661690218e-06, "loss": 0.1605, "step": 625 }, { "epoch": 0.02, "learning_rate": 9.92988536390296e-06, "loss": 0.1395, "step": 626 }, { "epoch": 0.03, "learning_rate": 9.929752066115704e-06, "loss": 0.1533, "step": 627 }, { "epoch": 0.03, "learning_rate": 9.929618768328446e-06, "loss": 0.1287, "step": 628 }, { "epoch": 0.03, "learning_rate": 9.929485470541191e-06, "loss": 0.2039, "step": 629 }, { "epoch": 0.03, "learning_rate": 9.929352172753934e-06, "loss": 0.22, "step": 630 }, { "epoch": 0.03, "learning_rate": 9.929218874966677e-06, "loss": 0.1955, "step": 631 }, { "epoch": 0.03, "learning_rate": 9.92908557717942e-06, "loss": 0.1811, "step": 632 }, { "epoch": 0.03, "learning_rate": 9.928952279392164e-06, "loss": 0.2012, "step": 633 }, { "epoch": 0.03, "learning_rate": 9.928818981604907e-06, "loss": 0.1704, "step": 634 }, { "epoch": 0.03, "learning_rate": 9.92868568381765e-06, "loss": 0.1996, "step": 635 }, { "epoch": 0.03, "learning_rate": 9.928552386030392e-06, "loss": 0.2665, "step": 636 }, { "epoch": 0.03, "learning_rate": 9.928419088243137e-06, "loss": 0.2115, "step": 637 }, { "epoch": 0.03, "learning_rate": 9.92828579045588e-06, "loss": 0.1573, "step": 638 }, { "epoch": 0.03, "learning_rate": 9.928152492668623e-06, "loss": 0.1969, "step": 639 }, { "epoch": 0.03, "learning_rate": 9.928019194881365e-06, "loss": 0.2111, "step": 640 }, { "epoch": 0.03, "learning_rate": 9.92788589709411e-06, "loss": 0.1911, "step": 641 }, { "epoch": 0.03, "learning_rate": 9.927752599306853e-06, "loss": 0.1601, "step": 642 }, { "epoch": 0.03, "learning_rate": 9.927619301519595e-06, "loss": 0.163, "step": 643 }, { "epoch": 0.03, "learning_rate": 9.927486003732338e-06, "loss": 0.1691, "step": 644 }, { "epoch": 0.03, "learning_rate": 9.927352705945083e-06, "loss": 0.1781, "step": 645 }, { "epoch": 0.03, "learning_rate": 9.927219408157826e-06, "loss": 0.166, "step": 646 }, { "epoch": 0.03, "learning_rate": 9.927086110370568e-06, "loss": 0.1948, "step": 647 }, { "epoch": 0.03, "learning_rate": 9.926952812583311e-06, "loss": 0.1901, "step": 648 }, { "epoch": 0.03, "learning_rate": 9.926819514796056e-06, "loss": 0.2003, "step": 649 }, { "epoch": 0.03, "learning_rate": 9.926686217008799e-06, "loss": 0.1498, "step": 650 }, { "epoch": 0.03, "learning_rate": 9.926552919221541e-06, "loss": 0.2387, "step": 651 }, { "epoch": 0.03, "learning_rate": 9.926419621434284e-06, "loss": 0.1618, "step": 652 }, { "epoch": 0.03, "learning_rate": 9.926286323647029e-06, "loss": 0.2223, "step": 653 }, { "epoch": 0.03, "learning_rate": 9.926153025859772e-06, "loss": 0.1542, "step": 654 }, { "epoch": 0.03, "learning_rate": 9.926019728072514e-06, "loss": 0.1616, "step": 655 }, { "epoch": 0.03, "learning_rate": 9.925886430285257e-06, "loss": 0.1934, "step": 656 }, { "epoch": 0.03, "learning_rate": 9.925753132498002e-06, "loss": 0.1437, "step": 657 }, { "epoch": 0.03, "learning_rate": 9.925619834710745e-06, "loss": 0.172, "step": 658 }, { "epoch": 0.03, "learning_rate": 9.925486536923487e-06, "loss": 0.2211, "step": 659 }, { "epoch": 0.03, "learning_rate": 9.92535323913623e-06, "loss": 0.1705, "step": 660 }, { "epoch": 0.03, "learning_rate": 9.925219941348975e-06, "loss": 0.2009, "step": 661 }, { "epoch": 0.03, "learning_rate": 9.925086643561718e-06, "loss": 0.2043, "step": 662 }, { "epoch": 0.03, "learning_rate": 9.92495334577446e-06, "loss": 0.1469, "step": 663 }, { "epoch": 0.03, "learning_rate": 9.924820047987205e-06, "loss": 0.1422, "step": 664 }, { "epoch": 0.03, "learning_rate": 9.924686750199948e-06, "loss": 0.1584, "step": 665 }, { "epoch": 0.03, "learning_rate": 9.92455345241269e-06, "loss": 0.1479, "step": 666 }, { "epoch": 0.03, "learning_rate": 9.924420154625433e-06, "loss": 0.1627, "step": 667 }, { "epoch": 0.03, "learning_rate": 9.924286856838178e-06, "loss": 0.165, "step": 668 }, { "epoch": 0.03, "learning_rate": 9.92415355905092e-06, "loss": 0.1532, "step": 669 }, { "epoch": 0.03, "learning_rate": 9.924020261263663e-06, "loss": 0.1558, "step": 670 }, { "epoch": 0.03, "learning_rate": 9.923886963476406e-06, "loss": 0.1519, "step": 671 }, { "epoch": 0.03, "learning_rate": 9.92375366568915e-06, "loss": 0.1666, "step": 672 }, { "epoch": 0.03, "learning_rate": 9.923620367901894e-06, "loss": 0.1732, "step": 673 }, { "epoch": 0.03, "learning_rate": 9.923487070114636e-06, "loss": 0.1691, "step": 674 }, { "epoch": 0.03, "learning_rate": 9.92335377232738e-06, "loss": 0.1647, "step": 675 }, { "epoch": 0.03, "learning_rate": 9.923220474540124e-06, "loss": 0.1806, "step": 676 }, { "epoch": 0.03, "learning_rate": 9.923087176752867e-06, "loss": 0.1578, "step": 677 }, { "epoch": 0.03, "learning_rate": 9.92295387896561e-06, "loss": 0.1852, "step": 678 }, { "epoch": 0.03, "learning_rate": 9.922820581178352e-06, "loss": 0.1855, "step": 679 }, { "epoch": 0.03, "learning_rate": 9.922687283391097e-06, "loss": 0.1756, "step": 680 }, { "epoch": 0.03, "learning_rate": 9.92255398560384e-06, "loss": 0.1588, "step": 681 }, { "epoch": 0.03, "learning_rate": 9.922420687816582e-06, "loss": 0.1507, "step": 682 }, { "epoch": 0.03, "learning_rate": 9.922287390029325e-06, "loss": 0.2001, "step": 683 }, { "epoch": 0.03, "learning_rate": 9.92215409224207e-06, "loss": 0.1605, "step": 684 }, { "epoch": 0.03, "learning_rate": 9.922020794454813e-06, "loss": 0.1402, "step": 685 }, { "epoch": 0.03, "learning_rate": 9.921887496667555e-06, "loss": 0.1552, "step": 686 }, { "epoch": 0.03, "learning_rate": 9.921754198880298e-06, "loss": 0.2262, "step": 687 }, { "epoch": 0.03, "learning_rate": 9.921620901093043e-06, "loss": 0.169, "step": 688 }, { "epoch": 0.03, "learning_rate": 9.921487603305785e-06, "loss": 0.2193, "step": 689 }, { "epoch": 0.03, "learning_rate": 9.921354305518528e-06, "loss": 0.2318, "step": 690 }, { "epoch": 0.03, "learning_rate": 9.921221007731273e-06, "loss": 0.2113, "step": 691 }, { "epoch": 0.03, "learning_rate": 9.921087709944016e-06, "loss": 0.1774, "step": 692 }, { "epoch": 0.03, "learning_rate": 9.920954412156758e-06, "loss": 0.1587, "step": 693 }, { "epoch": 0.03, "learning_rate": 9.920821114369503e-06, "loss": 0.1783, "step": 694 }, { "epoch": 0.03, "learning_rate": 9.920687816582246e-06, "loss": 0.1698, "step": 695 }, { "epoch": 0.03, "learning_rate": 9.920554518794989e-06, "loss": 0.158, "step": 696 }, { "epoch": 0.03, "learning_rate": 9.920421221007731e-06, "loss": 0.1701, "step": 697 }, { "epoch": 0.03, "learning_rate": 9.920287923220476e-06, "loss": 0.1478, "step": 698 }, { "epoch": 0.03, "learning_rate": 9.920154625433219e-06, "loss": 0.1531, "step": 699 }, { "epoch": 0.03, "learning_rate": 9.920021327645962e-06, "loss": 0.1941, "step": 700 }, { "epoch": 0.03, "learning_rate": 9.919888029858706e-06, "loss": 0.194, "step": 701 }, { "epoch": 0.03, "learning_rate": 9.919754732071449e-06, "loss": 0.1539, "step": 702 }, { "epoch": 0.03, "learning_rate": 9.919621434284192e-06, "loss": 0.175, "step": 703 }, { "epoch": 0.03, "learning_rate": 9.919488136496935e-06, "loss": 0.2185, "step": 704 }, { "epoch": 0.03, "learning_rate": 9.919354838709679e-06, "loss": 0.1409, "step": 705 }, { "epoch": 0.03, "learning_rate": 9.919221540922422e-06, "loss": 0.1573, "step": 706 }, { "epoch": 0.03, "learning_rate": 9.919088243135165e-06, "loss": 0.1512, "step": 707 }, { "epoch": 0.03, "learning_rate": 9.91895494534791e-06, "loss": 0.1903, "step": 708 }, { "epoch": 0.03, "learning_rate": 9.918821647560652e-06, "loss": 0.1657, "step": 709 }, { "epoch": 0.03, "learning_rate": 9.918688349773395e-06, "loss": 0.173, "step": 710 }, { "epoch": 0.03, "learning_rate": 9.918555051986138e-06, "loss": 0.1924, "step": 711 }, { "epoch": 0.03, "learning_rate": 9.918421754198882e-06, "loss": 0.198, "step": 712 }, { "epoch": 0.03, "learning_rate": 9.918288456411625e-06, "loss": 0.1581, "step": 713 }, { "epoch": 0.03, "learning_rate": 9.918155158624368e-06, "loss": 0.1938, "step": 714 }, { "epoch": 0.03, "learning_rate": 9.91802186083711e-06, "loss": 0.1719, "step": 715 }, { "epoch": 0.03, "learning_rate": 9.917888563049855e-06, "loss": 0.1976, "step": 716 }, { "epoch": 0.03, "learning_rate": 9.917755265262598e-06, "loss": 0.1244, "step": 717 }, { "epoch": 0.03, "learning_rate": 9.91762196747534e-06, "loss": 0.1582, "step": 718 }, { "epoch": 0.03, "learning_rate": 9.917488669688084e-06, "loss": 0.1243, "step": 719 }, { "epoch": 0.03, "learning_rate": 9.917355371900828e-06, "loss": 0.1607, "step": 720 }, { "epoch": 0.03, "learning_rate": 9.917222074113571e-06, "loss": 0.1691, "step": 721 }, { "epoch": 0.03, "learning_rate": 9.917088776326314e-06, "loss": 0.1509, "step": 722 }, { "epoch": 0.03, "learning_rate": 9.916955478539058e-06, "loss": 0.1372, "step": 723 }, { "epoch": 0.03, "learning_rate": 9.916822180751801e-06, "loss": 0.1876, "step": 724 }, { "epoch": 0.03, "learning_rate": 9.916688882964544e-06, "loss": 0.192, "step": 725 }, { "epoch": 0.03, "learning_rate": 9.916555585177287e-06, "loss": 0.1273, "step": 726 }, { "epoch": 0.03, "learning_rate": 9.916422287390031e-06, "loss": 0.1551, "step": 727 }, { "epoch": 0.03, "learning_rate": 9.916288989602774e-06, "loss": 0.195, "step": 728 }, { "epoch": 0.03, "learning_rate": 9.916155691815517e-06, "loss": 0.1152, "step": 729 }, { "epoch": 0.03, "learning_rate": 9.91602239402826e-06, "loss": 0.1634, "step": 730 }, { "epoch": 0.03, "learning_rate": 9.915889096241004e-06, "loss": 0.1675, "step": 731 }, { "epoch": 0.03, "learning_rate": 9.915755798453747e-06, "loss": 0.1427, "step": 732 }, { "epoch": 0.03, "learning_rate": 9.91562250066649e-06, "loss": 0.1708, "step": 733 }, { "epoch": 0.03, "learning_rate": 9.915489202879233e-06, "loss": 0.1793, "step": 734 }, { "epoch": 0.03, "learning_rate": 9.915355905091977e-06, "loss": 0.1736, "step": 735 }, { "epoch": 0.03, "learning_rate": 9.91522260730472e-06, "loss": 0.168, "step": 736 }, { "epoch": 0.03, "learning_rate": 9.915089309517463e-06, "loss": 0.1774, "step": 737 }, { "epoch": 0.03, "learning_rate": 9.914956011730206e-06, "loss": 0.1505, "step": 738 }, { "epoch": 0.03, "learning_rate": 9.91482271394295e-06, "loss": 0.1394, "step": 739 }, { "epoch": 0.03, "learning_rate": 9.914689416155693e-06, "loss": 0.147, "step": 740 }, { "epoch": 0.03, "learning_rate": 9.914556118368436e-06, "loss": 0.1287, "step": 741 }, { "epoch": 0.03, "learning_rate": 9.914422820581179e-06, "loss": 0.1769, "step": 742 }, { "epoch": 0.03, "learning_rate": 9.914289522793923e-06, "loss": 0.2391, "step": 743 }, { "epoch": 0.03, "learning_rate": 9.914156225006666e-06, "loss": 0.2356, "step": 744 }, { "epoch": 0.03, "learning_rate": 9.914022927219409e-06, "loss": 0.1548, "step": 745 }, { "epoch": 0.03, "learning_rate": 9.913889629432152e-06, "loss": 0.199, "step": 746 }, { "epoch": 0.03, "learning_rate": 9.913756331644896e-06, "loss": 0.155, "step": 747 }, { "epoch": 0.03, "learning_rate": 9.913623033857639e-06, "loss": 0.1913, "step": 748 }, { "epoch": 0.03, "learning_rate": 9.913489736070382e-06, "loss": 0.1378, "step": 749 }, { "epoch": 0.03, "learning_rate": 9.913356438283125e-06, "loss": 0.1195, "step": 750 }, { "epoch": 0.03, "learning_rate": 9.913223140495869e-06, "loss": 0.1736, "step": 751 }, { "epoch": 0.03, "learning_rate": 9.913089842708612e-06, "loss": 0.1635, "step": 752 }, { "epoch": 0.03, "learning_rate": 9.912956544921355e-06, "loss": 0.1774, "step": 753 }, { "epoch": 0.03, "learning_rate": 9.912823247134098e-06, "loss": 0.1656, "step": 754 }, { "epoch": 0.03, "learning_rate": 9.912689949346842e-06, "loss": 0.1455, "step": 755 }, { "epoch": 0.03, "learning_rate": 9.912556651559585e-06, "loss": 0.1396, "step": 756 }, { "epoch": 0.03, "learning_rate": 9.912423353772328e-06, "loss": 0.2479, "step": 757 }, { "epoch": 0.03, "learning_rate": 9.91229005598507e-06, "loss": 0.1565, "step": 758 }, { "epoch": 0.03, "learning_rate": 9.912156758197815e-06, "loss": 0.1911, "step": 759 }, { "epoch": 0.03, "learning_rate": 9.912023460410558e-06, "loss": 0.2161, "step": 760 }, { "epoch": 0.03, "learning_rate": 9.9118901626233e-06, "loss": 0.1451, "step": 761 }, { "epoch": 0.03, "learning_rate": 9.911756864836043e-06, "loss": 0.2208, "step": 762 }, { "epoch": 0.03, "learning_rate": 9.911623567048788e-06, "loss": 0.1458, "step": 763 }, { "epoch": 0.03, "learning_rate": 9.91149026926153e-06, "loss": 0.202, "step": 764 }, { "epoch": 0.03, "learning_rate": 9.911356971474274e-06, "loss": 0.1551, "step": 765 }, { "epoch": 0.03, "learning_rate": 9.911223673687016e-06, "loss": 0.176, "step": 766 }, { "epoch": 0.03, "learning_rate": 9.911090375899761e-06, "loss": 0.1734, "step": 767 }, { "epoch": 0.03, "learning_rate": 9.910957078112504e-06, "loss": 0.1802, "step": 768 }, { "epoch": 0.03, "learning_rate": 9.910823780325247e-06, "loss": 0.1511, "step": 769 }, { "epoch": 0.03, "learning_rate": 9.91069048253799e-06, "loss": 0.1885, "step": 770 }, { "epoch": 0.03, "learning_rate": 9.910557184750734e-06, "loss": 0.1912, "step": 771 }, { "epoch": 0.03, "learning_rate": 9.910423886963477e-06, "loss": 0.1953, "step": 772 }, { "epoch": 0.03, "learning_rate": 9.91029058917622e-06, "loss": 0.1476, "step": 773 }, { "epoch": 0.03, "learning_rate": 9.910157291388962e-06, "loss": 0.1713, "step": 774 }, { "epoch": 0.03, "learning_rate": 9.910023993601707e-06, "loss": 0.1687, "step": 775 }, { "epoch": 0.03, "learning_rate": 9.90989069581445e-06, "loss": 0.1525, "step": 776 }, { "epoch": 0.03, "learning_rate": 9.909757398027193e-06, "loss": 0.1545, "step": 777 }, { "epoch": 0.03, "learning_rate": 9.909624100239937e-06, "loss": 0.1789, "step": 778 }, { "epoch": 0.03, "learning_rate": 9.90949080245268e-06, "loss": 0.1804, "step": 779 }, { "epoch": 0.03, "learning_rate": 9.909357504665423e-06, "loss": 0.1265, "step": 780 }, { "epoch": 0.03, "learning_rate": 9.909224206878165e-06, "loss": 0.1676, "step": 781 }, { "epoch": 0.03, "learning_rate": 9.90909090909091e-06, "loss": 0.2089, "step": 782 }, { "epoch": 0.03, "learning_rate": 9.908957611303653e-06, "loss": 0.196, "step": 783 }, { "epoch": 0.03, "learning_rate": 9.908824313516396e-06, "loss": 0.1953, "step": 784 }, { "epoch": 0.03, "learning_rate": 9.90869101572914e-06, "loss": 0.2139, "step": 785 }, { "epoch": 0.03, "learning_rate": 9.908557717941883e-06, "loss": 0.1778, "step": 786 }, { "epoch": 0.03, "learning_rate": 9.908424420154626e-06, "loss": 0.1574, "step": 787 }, { "epoch": 0.03, "learning_rate": 9.908291122367369e-06, "loss": 0.1877, "step": 788 }, { "epoch": 0.03, "learning_rate": 9.908157824580113e-06, "loss": 0.1747, "step": 789 }, { "epoch": 0.03, "learning_rate": 9.908024526792856e-06, "loss": 0.1891, "step": 790 }, { "epoch": 0.03, "learning_rate": 9.907891229005599e-06, "loss": 0.1897, "step": 791 }, { "epoch": 0.03, "learning_rate": 9.907757931218343e-06, "loss": 0.1134, "step": 792 }, { "epoch": 0.03, "learning_rate": 9.907624633431086e-06, "loss": 0.15, "step": 793 }, { "epoch": 0.03, "learning_rate": 9.907491335643829e-06, "loss": 0.1782, "step": 794 }, { "epoch": 0.03, "learning_rate": 9.907358037856572e-06, "loss": 0.1685, "step": 795 }, { "epoch": 0.03, "learning_rate": 9.907224740069316e-06, "loss": 0.1636, "step": 796 }, { "epoch": 0.03, "learning_rate": 9.907091442282059e-06, "loss": 0.1961, "step": 797 }, { "epoch": 0.03, "learning_rate": 9.906958144494802e-06, "loss": 0.1867, "step": 798 }, { "epoch": 0.03, "learning_rate": 9.906824846707546e-06, "loss": 0.1404, "step": 799 }, { "epoch": 0.03, "learning_rate": 9.90669154892029e-06, "loss": 0.1646, "step": 800 }, { "epoch": 0.03, "learning_rate": 9.906558251133032e-06, "loss": 0.1574, "step": 801 }, { "epoch": 0.03, "learning_rate": 9.906424953345775e-06, "loss": 0.1505, "step": 802 }, { "epoch": 0.03, "learning_rate": 9.90629165555852e-06, "loss": 0.1298, "step": 803 }, { "epoch": 0.03, "learning_rate": 9.906158357771262e-06, "loss": 0.1465, "step": 804 }, { "epoch": 0.03, "learning_rate": 9.906025059984005e-06, "loss": 0.1591, "step": 805 }, { "epoch": 0.03, "learning_rate": 9.90589176219675e-06, "loss": 0.1406, "step": 806 }, { "epoch": 0.03, "learning_rate": 9.905758464409492e-06, "loss": 0.2217, "step": 807 }, { "epoch": 0.03, "learning_rate": 9.905625166622235e-06, "loss": 0.1988, "step": 808 }, { "epoch": 0.03, "learning_rate": 9.905491868834978e-06, "loss": 0.152, "step": 809 }, { "epoch": 0.03, "learning_rate": 9.905358571047722e-06, "loss": 0.1957, "step": 810 }, { "epoch": 0.03, "learning_rate": 9.905225273260465e-06, "loss": 0.1404, "step": 811 }, { "epoch": 0.03, "learning_rate": 9.905091975473208e-06, "loss": 0.137, "step": 812 }, { "epoch": 0.03, "learning_rate": 9.904958677685951e-06, "loss": 0.1345, "step": 813 }, { "epoch": 0.03, "learning_rate": 9.904825379898695e-06, "loss": 0.1551, "step": 814 }, { "epoch": 0.03, "learning_rate": 9.904692082111438e-06, "loss": 0.2091, "step": 815 }, { "epoch": 0.03, "learning_rate": 9.904558784324181e-06, "loss": 0.1576, "step": 816 }, { "epoch": 0.03, "learning_rate": 9.904425486536924e-06, "loss": 0.137, "step": 817 }, { "epoch": 0.03, "learning_rate": 9.904292188749668e-06, "loss": 0.1705, "step": 818 }, { "epoch": 0.03, "learning_rate": 9.904158890962411e-06, "loss": 0.1848, "step": 819 }, { "epoch": 0.03, "learning_rate": 9.904025593175154e-06, "loss": 0.1729, "step": 820 }, { "epoch": 0.03, "learning_rate": 9.903892295387897e-06, "loss": 0.1907, "step": 821 }, { "epoch": 0.03, "learning_rate": 9.903758997600641e-06, "loss": 0.1956, "step": 822 }, { "epoch": 0.03, "learning_rate": 9.903625699813384e-06, "loss": 0.1829, "step": 823 }, { "epoch": 0.03, "learning_rate": 9.903492402026127e-06, "loss": 0.1181, "step": 824 }, { "epoch": 0.03, "learning_rate": 9.90335910423887e-06, "loss": 0.1403, "step": 825 }, { "epoch": 0.03, "learning_rate": 9.903225806451614e-06, "loss": 0.165, "step": 826 }, { "epoch": 0.03, "learning_rate": 9.903092508664357e-06, "loss": 0.1747, "step": 827 }, { "epoch": 0.03, "learning_rate": 9.9029592108771e-06, "loss": 0.1495, "step": 828 }, { "epoch": 0.03, "learning_rate": 9.902825913089843e-06, "loss": 0.1512, "step": 829 }, { "epoch": 0.03, "learning_rate": 9.902692615302587e-06, "loss": 0.1567, "step": 830 }, { "epoch": 0.03, "learning_rate": 9.90255931751533e-06, "loss": 0.1674, "step": 831 }, { "epoch": 0.03, "learning_rate": 9.902426019728073e-06, "loss": 0.1558, "step": 832 }, { "epoch": 0.03, "learning_rate": 9.902292721940817e-06, "loss": 0.143, "step": 833 }, { "epoch": 0.03, "learning_rate": 9.90215942415356e-06, "loss": 0.1314, "step": 834 }, { "epoch": 0.03, "learning_rate": 9.902026126366303e-06, "loss": 0.1732, "step": 835 }, { "epoch": 0.03, "learning_rate": 9.901892828579046e-06, "loss": 0.1208, "step": 836 }, { "epoch": 0.03, "learning_rate": 9.90175953079179e-06, "loss": 0.1235, "step": 837 }, { "epoch": 0.03, "learning_rate": 9.901626233004533e-06, "loss": 0.1178, "step": 838 }, { "epoch": 0.03, "learning_rate": 9.901492935217276e-06, "loss": 0.1831, "step": 839 }, { "epoch": 0.03, "learning_rate": 9.901359637430019e-06, "loss": 0.1846, "step": 840 }, { "epoch": 0.03, "learning_rate": 9.901226339642763e-06, "loss": 0.1738, "step": 841 }, { "epoch": 0.03, "learning_rate": 9.901093041855506e-06, "loss": 0.1703, "step": 842 }, { "epoch": 0.03, "learning_rate": 9.900959744068249e-06, "loss": 0.1429, "step": 843 }, { "epoch": 0.03, "learning_rate": 9.900826446280992e-06, "loss": 0.1382, "step": 844 }, { "epoch": 0.03, "learning_rate": 9.900693148493736e-06, "loss": 0.154, "step": 845 }, { "epoch": 0.03, "learning_rate": 9.90055985070648e-06, "loss": 0.1789, "step": 846 }, { "epoch": 0.03, "learning_rate": 9.900426552919222e-06, "loss": 0.1653, "step": 847 }, { "epoch": 0.03, "learning_rate": 9.900293255131965e-06, "loss": 0.1556, "step": 848 }, { "epoch": 0.03, "learning_rate": 9.90015995734471e-06, "loss": 0.1255, "step": 849 }, { "epoch": 0.03, "learning_rate": 9.900026659557452e-06, "loss": 0.1751, "step": 850 }, { "epoch": 0.03, "learning_rate": 9.899893361770195e-06, "loss": 0.1888, "step": 851 }, { "epoch": 0.03, "learning_rate": 9.899760063982938e-06, "loss": 0.1519, "step": 852 }, { "epoch": 0.03, "learning_rate": 9.899626766195682e-06, "loss": 0.1854, "step": 853 }, { "epoch": 0.03, "learning_rate": 9.899493468408425e-06, "loss": 0.1454, "step": 854 }, { "epoch": 0.03, "learning_rate": 9.899360170621168e-06, "loss": 0.1543, "step": 855 }, { "epoch": 0.03, "learning_rate": 9.89922687283391e-06, "loss": 0.1302, "step": 856 }, { "epoch": 0.03, "learning_rate": 9.899093575046655e-06, "loss": 0.1612, "step": 857 }, { "epoch": 0.03, "learning_rate": 9.898960277259398e-06, "loss": 0.1413, "step": 858 }, { "epoch": 0.03, "learning_rate": 9.898826979472141e-06, "loss": 0.1887, "step": 859 }, { "epoch": 0.03, "learning_rate": 9.898693681684884e-06, "loss": 0.1204, "step": 860 }, { "epoch": 0.03, "learning_rate": 9.898560383897628e-06, "loss": 0.2046, "step": 861 }, { "epoch": 0.03, "learning_rate": 9.898427086110371e-06, "loss": 0.1633, "step": 862 }, { "epoch": 0.03, "learning_rate": 9.898293788323114e-06, "loss": 0.1758, "step": 863 }, { "epoch": 0.03, "learning_rate": 9.898160490535857e-06, "loss": 0.1626, "step": 864 }, { "epoch": 0.03, "learning_rate": 9.898027192748601e-06, "loss": 0.1279, "step": 865 }, { "epoch": 0.03, "learning_rate": 9.897893894961344e-06, "loss": 0.1581, "step": 866 }, { "epoch": 0.03, "learning_rate": 9.897760597174087e-06, "loss": 0.177, "step": 867 }, { "epoch": 0.03, "learning_rate": 9.89762729938683e-06, "loss": 0.157, "step": 868 }, { "epoch": 0.03, "learning_rate": 9.897494001599574e-06, "loss": 0.1084, "step": 869 }, { "epoch": 0.03, "learning_rate": 9.897360703812317e-06, "loss": 0.1831, "step": 870 }, { "epoch": 0.03, "learning_rate": 9.89722740602506e-06, "loss": 0.1663, "step": 871 }, { "epoch": 0.03, "learning_rate": 9.897094108237804e-06, "loss": 0.176, "step": 872 }, { "epoch": 0.03, "learning_rate": 9.896960810450547e-06, "loss": 0.1734, "step": 873 }, { "epoch": 0.03, "learning_rate": 9.89682751266329e-06, "loss": 0.1528, "step": 874 }, { "epoch": 0.03, "learning_rate": 9.896694214876033e-06, "loss": 0.1932, "step": 875 }, { "epoch": 0.03, "learning_rate": 9.896560917088777e-06, "loss": 0.1835, "step": 876 }, { "epoch": 0.04, "learning_rate": 9.89642761930152e-06, "loss": 0.1523, "step": 877 }, { "epoch": 0.04, "learning_rate": 9.896294321514263e-06, "loss": 0.1522, "step": 878 }, { "epoch": 0.04, "learning_rate": 9.896161023727007e-06, "loss": 0.1885, "step": 879 }, { "epoch": 0.04, "learning_rate": 9.89602772593975e-06, "loss": 0.1921, "step": 880 }, { "epoch": 0.04, "learning_rate": 9.895894428152493e-06, "loss": 0.2162, "step": 881 }, { "epoch": 0.04, "learning_rate": 9.895761130365236e-06, "loss": 0.1617, "step": 882 }, { "epoch": 0.04, "learning_rate": 9.89562783257798e-06, "loss": 0.1266, "step": 883 }, { "epoch": 0.04, "learning_rate": 9.895494534790723e-06, "loss": 0.1402, "step": 884 }, { "epoch": 0.04, "learning_rate": 9.895361237003466e-06, "loss": 0.1738, "step": 885 }, { "epoch": 0.04, "learning_rate": 9.89522793921621e-06, "loss": 0.1417, "step": 886 }, { "epoch": 0.04, "learning_rate": 9.895094641428953e-06, "loss": 0.2088, "step": 887 }, { "epoch": 0.04, "learning_rate": 9.894961343641696e-06, "loss": 0.166, "step": 888 }, { "epoch": 0.04, "learning_rate": 9.894828045854439e-06, "loss": 0.165, "step": 889 }, { "epoch": 0.04, "learning_rate": 9.894694748067184e-06, "loss": 0.1643, "step": 890 }, { "epoch": 0.04, "learning_rate": 9.894561450279926e-06, "loss": 0.1648, "step": 891 }, { "epoch": 0.04, "learning_rate": 9.89442815249267e-06, "loss": 0.154, "step": 892 }, { "epoch": 0.04, "learning_rate": 9.894294854705414e-06, "loss": 0.1594, "step": 893 }, { "epoch": 0.04, "learning_rate": 9.894161556918157e-06, "loss": 0.1798, "step": 894 }, { "epoch": 0.04, "learning_rate": 9.8940282591309e-06, "loss": 0.1141, "step": 895 }, { "epoch": 0.04, "learning_rate": 9.893894961343642e-06, "loss": 0.1377, "step": 896 }, { "epoch": 0.04, "learning_rate": 9.893761663556387e-06, "loss": 0.1692, "step": 897 }, { "epoch": 0.04, "learning_rate": 9.89362836576913e-06, "loss": 0.1552, "step": 898 }, { "epoch": 0.04, "learning_rate": 9.893495067981872e-06, "loss": 0.167, "step": 899 }, { "epoch": 0.04, "learning_rate": 9.893361770194617e-06, "loss": 0.146, "step": 900 }, { "epoch": 0.04, "learning_rate": 9.89322847240736e-06, "loss": 0.1355, "step": 901 }, { "epoch": 0.04, "learning_rate": 9.893095174620103e-06, "loss": 0.1307, "step": 902 }, { "epoch": 0.04, "learning_rate": 9.892961876832845e-06, "loss": 0.1999, "step": 903 }, { "epoch": 0.04, "learning_rate": 9.89282857904559e-06, "loss": 0.1594, "step": 904 }, { "epoch": 0.04, "learning_rate": 9.892695281258333e-06, "loss": 0.1921, "step": 905 }, { "epoch": 0.04, "learning_rate": 9.892561983471075e-06, "loss": 0.1769, "step": 906 }, { "epoch": 0.04, "learning_rate": 9.892428685683818e-06, "loss": 0.1411, "step": 907 }, { "epoch": 0.04, "learning_rate": 9.892295387896563e-06, "loss": 0.1889, "step": 908 }, { "epoch": 0.04, "learning_rate": 9.892162090109306e-06, "loss": 0.1375, "step": 909 }, { "epoch": 0.04, "learning_rate": 9.892028792322048e-06, "loss": 0.1348, "step": 910 }, { "epoch": 0.04, "learning_rate": 9.891895494534791e-06, "loss": 0.1877, "step": 911 }, { "epoch": 0.04, "learning_rate": 9.891762196747536e-06, "loss": 0.1247, "step": 912 }, { "epoch": 0.04, "learning_rate": 9.891628898960279e-06, "loss": 0.126, "step": 913 }, { "epoch": 0.04, "learning_rate": 9.891495601173021e-06, "loss": 0.1373, "step": 914 }, { "epoch": 0.04, "learning_rate": 9.891362303385764e-06, "loss": 0.1543, "step": 915 }, { "epoch": 0.04, "learning_rate": 9.891229005598509e-06, "loss": 0.1688, "step": 916 }, { "epoch": 0.04, "learning_rate": 9.891095707811252e-06, "loss": 0.1385, "step": 917 }, { "epoch": 0.04, "learning_rate": 9.890962410023994e-06, "loss": 0.1391, "step": 918 }, { "epoch": 0.04, "learning_rate": 9.890829112236737e-06, "loss": 0.1521, "step": 919 }, { "epoch": 0.04, "learning_rate": 9.890695814449482e-06, "loss": 0.1645, "step": 920 }, { "epoch": 0.04, "learning_rate": 9.890562516662225e-06, "loss": 0.1725, "step": 921 }, { "epoch": 0.04, "learning_rate": 9.890429218874967e-06, "loss": 0.1651, "step": 922 }, { "epoch": 0.04, "learning_rate": 9.89029592108771e-06, "loss": 0.1621, "step": 923 }, { "epoch": 0.04, "learning_rate": 9.890162623300455e-06, "loss": 0.1441, "step": 924 }, { "epoch": 0.04, "learning_rate": 9.890029325513198e-06, "loss": 0.1976, "step": 925 }, { "epoch": 0.04, "learning_rate": 9.88989602772594e-06, "loss": 0.1737, "step": 926 }, { "epoch": 0.04, "learning_rate": 9.889762729938683e-06, "loss": 0.139, "step": 927 }, { "epoch": 0.04, "learning_rate": 9.889629432151428e-06, "loss": 0.1533, "step": 928 }, { "epoch": 0.04, "learning_rate": 9.88949613436417e-06, "loss": 0.1576, "step": 929 }, { "epoch": 0.04, "learning_rate": 9.889362836576913e-06, "loss": 0.1375, "step": 930 }, { "epoch": 0.04, "learning_rate": 9.889229538789656e-06, "loss": 0.2009, "step": 931 }, { "epoch": 0.04, "learning_rate": 9.8890962410024e-06, "loss": 0.1409, "step": 932 }, { "epoch": 0.04, "learning_rate": 9.888962943215143e-06, "loss": 0.1687, "step": 933 }, { "epoch": 0.04, "learning_rate": 9.888829645427886e-06, "loss": 0.2027, "step": 934 }, { "epoch": 0.04, "learning_rate": 9.888696347640629e-06, "loss": 0.1772, "step": 935 }, { "epoch": 0.04, "learning_rate": 9.888563049853374e-06, "loss": 0.1393, "step": 936 }, { "epoch": 0.04, "learning_rate": 9.888429752066116e-06, "loss": 0.1818, "step": 937 }, { "epoch": 0.04, "learning_rate": 9.88829645427886e-06, "loss": 0.1454, "step": 938 }, { "epoch": 0.04, "learning_rate": 9.888163156491602e-06, "loss": 0.1536, "step": 939 }, { "epoch": 0.04, "learning_rate": 9.888029858704347e-06, "loss": 0.139, "step": 940 }, { "epoch": 0.04, "learning_rate": 9.88789656091709e-06, "loss": 0.1459, "step": 941 }, { "epoch": 0.04, "learning_rate": 9.887763263129832e-06, "loss": 0.1655, "step": 942 }, { "epoch": 0.04, "learning_rate": 9.887629965342575e-06, "loss": 0.1524, "step": 943 }, { "epoch": 0.04, "learning_rate": 9.88749666755532e-06, "loss": 0.1415, "step": 944 }, { "epoch": 0.04, "learning_rate": 9.887363369768062e-06, "loss": 0.1269, "step": 945 }, { "epoch": 0.04, "learning_rate": 9.887230071980805e-06, "loss": 0.1898, "step": 946 }, { "epoch": 0.04, "learning_rate": 9.88709677419355e-06, "loss": 0.1836, "step": 947 }, { "epoch": 0.04, "learning_rate": 9.886963476406293e-06, "loss": 0.137, "step": 948 }, { "epoch": 0.04, "learning_rate": 9.886830178619035e-06, "loss": 0.1741, "step": 949 }, { "epoch": 0.04, "learning_rate": 9.886696880831778e-06, "loss": 0.1577, "step": 950 }, { "epoch": 0.04, "learning_rate": 9.886563583044523e-06, "loss": 0.1221, "step": 951 }, { "epoch": 0.04, "learning_rate": 9.886430285257265e-06, "loss": 0.1395, "step": 952 }, { "epoch": 0.04, "learning_rate": 9.886296987470008e-06, "loss": 0.1467, "step": 953 }, { "epoch": 0.04, "learning_rate": 9.886163689682751e-06, "loss": 0.106, "step": 954 }, { "epoch": 0.04, "learning_rate": 9.886030391895496e-06, "loss": 0.1216, "step": 955 }, { "epoch": 0.04, "learning_rate": 9.885897094108238e-06, "loss": 0.1313, "step": 956 }, { "epoch": 0.04, "learning_rate": 9.885763796320981e-06, "loss": 0.1573, "step": 957 }, { "epoch": 0.04, "learning_rate": 9.885630498533724e-06, "loss": 0.1614, "step": 958 }, { "epoch": 0.04, "learning_rate": 9.885497200746469e-06, "loss": 0.1203, "step": 959 }, { "epoch": 0.04, "learning_rate": 9.885363902959211e-06, "loss": 0.1798, "step": 960 }, { "epoch": 0.04, "learning_rate": 9.885230605171954e-06, "loss": 0.1304, "step": 961 }, { "epoch": 0.04, "learning_rate": 9.885097307384697e-06, "loss": 0.1326, "step": 962 }, { "epoch": 0.04, "learning_rate": 9.884964009597442e-06, "loss": 0.1463, "step": 963 }, { "epoch": 0.04, "learning_rate": 9.884830711810184e-06, "loss": 0.2028, "step": 964 }, { "epoch": 0.04, "learning_rate": 9.884697414022927e-06, "loss": 0.1749, "step": 965 }, { "epoch": 0.04, "learning_rate": 9.884564116235672e-06, "loss": 0.1686, "step": 966 }, { "epoch": 0.04, "learning_rate": 9.884430818448415e-06, "loss": 0.1643, "step": 967 }, { "epoch": 0.04, "learning_rate": 9.884297520661157e-06, "loss": 0.1534, "step": 968 }, { "epoch": 0.04, "learning_rate": 9.8841642228739e-06, "loss": 0.165, "step": 969 }, { "epoch": 0.04, "learning_rate": 9.884030925086645e-06, "loss": 0.1495, "step": 970 }, { "epoch": 0.04, "learning_rate": 9.883897627299388e-06, "loss": 0.1731, "step": 971 }, { "epoch": 0.04, "learning_rate": 9.88376432951213e-06, "loss": 0.146, "step": 972 }, { "epoch": 0.04, "learning_rate": 9.883631031724875e-06, "loss": 0.2028, "step": 973 }, { "epoch": 0.04, "learning_rate": 9.883497733937618e-06, "loss": 0.1138, "step": 974 }, { "epoch": 0.04, "learning_rate": 9.88336443615036e-06, "loss": 0.1541, "step": 975 }, { "epoch": 0.04, "learning_rate": 9.883231138363103e-06, "loss": 0.14, "step": 976 }, { "epoch": 0.04, "learning_rate": 9.883097840575848e-06, "loss": 0.149, "step": 977 }, { "epoch": 0.04, "learning_rate": 9.88296454278859e-06, "loss": 0.1504, "step": 978 }, { "epoch": 0.04, "learning_rate": 9.882831245001333e-06, "loss": 0.1369, "step": 979 }, { "epoch": 0.04, "learning_rate": 9.882697947214078e-06, "loss": 0.1189, "step": 980 }, { "epoch": 0.04, "learning_rate": 9.88256464942682e-06, "loss": 0.156, "step": 981 }, { "epoch": 0.04, "learning_rate": 9.882431351639564e-06, "loss": 0.1359, "step": 982 }, { "epoch": 0.04, "learning_rate": 9.882298053852306e-06, "loss": 0.1437, "step": 983 }, { "epoch": 0.04, "learning_rate": 9.882164756065051e-06, "loss": 0.1457, "step": 984 }, { "epoch": 0.04, "learning_rate": 9.882031458277794e-06, "loss": 0.1843, "step": 985 }, { "epoch": 0.04, "learning_rate": 9.881898160490537e-06, "loss": 0.1888, "step": 986 }, { "epoch": 0.04, "learning_rate": 9.881764862703281e-06, "loss": 0.1687, "step": 987 }, { "epoch": 0.04, "learning_rate": 9.881631564916024e-06, "loss": 0.1469, "step": 988 }, { "epoch": 0.04, "learning_rate": 9.881498267128767e-06, "loss": 0.1636, "step": 989 }, { "epoch": 0.04, "learning_rate": 9.88136496934151e-06, "loss": 0.1467, "step": 990 }, { "epoch": 0.04, "learning_rate": 9.881231671554254e-06, "loss": 0.1439, "step": 991 }, { "epoch": 0.04, "learning_rate": 9.881098373766997e-06, "loss": 0.1905, "step": 992 }, { "epoch": 0.04, "learning_rate": 9.88096507597974e-06, "loss": 0.1454, "step": 993 }, { "epoch": 0.04, "learning_rate": 9.880831778192483e-06, "loss": 0.1379, "step": 994 }, { "epoch": 0.04, "learning_rate": 9.880698480405227e-06, "loss": 0.1412, "step": 995 }, { "epoch": 0.04, "learning_rate": 9.88056518261797e-06, "loss": 0.121, "step": 996 }, { "epoch": 0.04, "learning_rate": 9.880431884830713e-06, "loss": 0.1289, "step": 997 }, { "epoch": 0.04, "learning_rate": 9.880298587043455e-06, "loss": 0.1393, "step": 998 }, { "epoch": 0.04, "learning_rate": 9.8801652892562e-06, "loss": 0.2252, "step": 999 }, { "epoch": 0.04, "learning_rate": 9.880031991468943e-06, "loss": 0.1656, "step": 1000 }, { "epoch": 0.04, "learning_rate": 9.879898693681686e-06, "loss": 0.1667, "step": 1001 }, { "epoch": 0.04, "learning_rate": 9.87976539589443e-06, "loss": 0.1027, "step": 1002 }, { "epoch": 0.04, "learning_rate": 9.879632098107173e-06, "loss": 0.1558, "step": 1003 }, { "epoch": 0.04, "learning_rate": 9.879498800319916e-06, "loss": 0.1078, "step": 1004 }, { "epoch": 0.04, "learning_rate": 9.879365502532659e-06, "loss": 0.1656, "step": 1005 }, { "epoch": 0.04, "learning_rate": 9.879232204745403e-06, "loss": 0.1897, "step": 1006 }, { "epoch": 0.04, "learning_rate": 9.879098906958146e-06, "loss": 0.1746, "step": 1007 }, { "epoch": 0.04, "learning_rate": 9.878965609170889e-06, "loss": 0.122, "step": 1008 }, { "epoch": 0.04, "learning_rate": 9.878832311383632e-06, "loss": 0.1619, "step": 1009 }, { "epoch": 0.04, "learning_rate": 9.878699013596376e-06, "loss": 0.1567, "step": 1010 }, { "epoch": 0.04, "learning_rate": 9.878565715809119e-06, "loss": 0.17, "step": 1011 }, { "epoch": 0.04, "learning_rate": 9.878432418021862e-06, "loss": 0.1647, "step": 1012 }, { "epoch": 0.04, "learning_rate": 9.878299120234605e-06, "loss": 0.1527, "step": 1013 }, { "epoch": 0.04, "learning_rate": 9.878165822447349e-06, "loss": 0.1189, "step": 1014 }, { "epoch": 0.04, "learning_rate": 9.878032524660092e-06, "loss": 0.1575, "step": 1015 }, { "epoch": 0.04, "learning_rate": 9.877899226872835e-06, "loss": 0.1438, "step": 1016 }, { "epoch": 0.04, "learning_rate": 9.877765929085578e-06, "loss": 0.1432, "step": 1017 }, { "epoch": 0.04, "learning_rate": 9.877632631298322e-06, "loss": 0.1305, "step": 1018 }, { "epoch": 0.04, "learning_rate": 9.877499333511065e-06, "loss": 0.146, "step": 1019 }, { "epoch": 0.04, "learning_rate": 9.877366035723808e-06, "loss": 0.1498, "step": 1020 }, { "epoch": 0.04, "learning_rate": 9.87723273793655e-06, "loss": 0.2041, "step": 1021 }, { "epoch": 0.04, "learning_rate": 9.877099440149295e-06, "loss": 0.1359, "step": 1022 }, { "epoch": 0.04, "learning_rate": 9.876966142362038e-06, "loss": 0.1605, "step": 1023 }, { "epoch": 0.04, "learning_rate": 9.87683284457478e-06, "loss": 0.1448, "step": 1024 }, { "epoch": 0.04, "learning_rate": 9.876699546787523e-06, "loss": 0.1058, "step": 1025 }, { "epoch": 0.04, "learning_rate": 9.876566249000268e-06, "loss": 0.1357, "step": 1026 }, { "epoch": 0.04, "learning_rate": 9.87643295121301e-06, "loss": 0.122, "step": 1027 }, { "epoch": 0.04, "learning_rate": 9.876299653425754e-06, "loss": 0.1558, "step": 1028 }, { "epoch": 0.04, "learning_rate": 9.876166355638496e-06, "loss": 0.1536, "step": 1029 }, { "epoch": 0.04, "learning_rate": 9.876033057851241e-06, "loss": 0.1356, "step": 1030 }, { "epoch": 0.04, "learning_rate": 9.875899760063984e-06, "loss": 0.1609, "step": 1031 }, { "epoch": 0.04, "learning_rate": 9.875766462276727e-06, "loss": 0.1555, "step": 1032 }, { "epoch": 0.04, "learning_rate": 9.87563316448947e-06, "loss": 0.1918, "step": 1033 }, { "epoch": 0.04, "learning_rate": 9.875499866702214e-06, "loss": 0.1461, "step": 1034 }, { "epoch": 0.04, "learning_rate": 9.875366568914957e-06, "loss": 0.1577, "step": 1035 }, { "epoch": 0.04, "learning_rate": 9.8752332711277e-06, "loss": 0.1437, "step": 1036 }, { "epoch": 0.04, "learning_rate": 9.875099973340442e-06, "loss": 0.1584, "step": 1037 }, { "epoch": 0.04, "learning_rate": 9.874966675553187e-06, "loss": 0.1548, "step": 1038 }, { "epoch": 0.04, "learning_rate": 9.87483337776593e-06, "loss": 0.1649, "step": 1039 }, { "epoch": 0.04, "learning_rate": 9.874700079978673e-06, "loss": 0.1379, "step": 1040 }, { "epoch": 0.04, "learning_rate": 9.874566782191415e-06, "loss": 0.1658, "step": 1041 }, { "epoch": 0.04, "learning_rate": 9.87443348440416e-06, "loss": 0.1699, "step": 1042 }, { "epoch": 0.04, "learning_rate": 9.874300186616903e-06, "loss": 0.1507, "step": 1043 }, { "epoch": 0.04, "learning_rate": 9.874166888829645e-06, "loss": 0.1701, "step": 1044 }, { "epoch": 0.04, "learning_rate": 9.874033591042388e-06, "loss": 0.1712, "step": 1045 }, { "epoch": 0.04, "learning_rate": 9.873900293255133e-06, "loss": 0.1389, "step": 1046 }, { "epoch": 0.04, "learning_rate": 9.873766995467876e-06, "loss": 0.1768, "step": 1047 }, { "epoch": 0.04, "learning_rate": 9.873633697680618e-06, "loss": 0.1199, "step": 1048 }, { "epoch": 0.04, "learning_rate": 9.873500399893361e-06, "loss": 0.207, "step": 1049 }, { "epoch": 0.04, "learning_rate": 9.873367102106106e-06, "loss": 0.1657, "step": 1050 }, { "epoch": 0.04, "learning_rate": 9.873233804318849e-06, "loss": 0.1794, "step": 1051 }, { "epoch": 0.04, "learning_rate": 9.873100506531591e-06, "loss": 0.2125, "step": 1052 }, { "epoch": 0.04, "learning_rate": 9.872967208744336e-06, "loss": 0.1473, "step": 1053 }, { "epoch": 0.04, "learning_rate": 9.872833910957079e-06, "loss": 0.1697, "step": 1054 }, { "epoch": 0.04, "learning_rate": 9.872700613169822e-06, "loss": 0.1718, "step": 1055 }, { "epoch": 0.04, "learning_rate": 9.872567315382564e-06, "loss": 0.1392, "step": 1056 }, { "epoch": 0.04, "learning_rate": 9.872434017595309e-06, "loss": 0.1626, "step": 1057 }, { "epoch": 0.04, "learning_rate": 9.872300719808052e-06, "loss": 0.1343, "step": 1058 }, { "epoch": 0.04, "learning_rate": 9.872167422020795e-06, "loss": 0.1896, "step": 1059 }, { "epoch": 0.04, "learning_rate": 9.872034124233539e-06, "loss": 0.1608, "step": 1060 }, { "epoch": 0.04, "learning_rate": 9.871900826446282e-06, "loss": 0.1361, "step": 1061 }, { "epoch": 0.04, "learning_rate": 9.871767528659025e-06, "loss": 0.1741, "step": 1062 }, { "epoch": 0.04, "learning_rate": 9.871634230871768e-06, "loss": 0.157, "step": 1063 }, { "epoch": 0.04, "learning_rate": 9.871500933084512e-06, "loss": 0.1663, "step": 1064 }, { "epoch": 0.04, "learning_rate": 9.871367635297255e-06, "loss": 0.1395, "step": 1065 }, { "epoch": 0.04, "learning_rate": 9.871234337509998e-06, "loss": 0.1365, "step": 1066 }, { "epoch": 0.04, "learning_rate": 9.871101039722742e-06, "loss": 0.1467, "step": 1067 }, { "epoch": 0.04, "learning_rate": 9.870967741935485e-06, "loss": 0.1983, "step": 1068 }, { "epoch": 0.04, "learning_rate": 9.870834444148228e-06, "loss": 0.1699, "step": 1069 }, { "epoch": 0.04, "learning_rate": 9.87070114636097e-06, "loss": 0.1379, "step": 1070 }, { "epoch": 0.04, "learning_rate": 9.870567848573715e-06, "loss": 0.1887, "step": 1071 }, { "epoch": 0.04, "learning_rate": 9.870434550786458e-06, "loss": 0.1458, "step": 1072 }, { "epoch": 0.04, "learning_rate": 9.8703012529992e-06, "loss": 0.1449, "step": 1073 }, { "epoch": 0.04, "learning_rate": 9.870167955211945e-06, "loss": 0.1263, "step": 1074 }, { "epoch": 0.04, "learning_rate": 9.870034657424688e-06, "loss": 0.1219, "step": 1075 }, { "epoch": 0.04, "learning_rate": 9.869901359637431e-06, "loss": 0.1675, "step": 1076 }, { "epoch": 0.04, "learning_rate": 9.869768061850174e-06, "loss": 0.1802, "step": 1077 }, { "epoch": 0.04, "learning_rate": 9.869634764062918e-06, "loss": 0.1632, "step": 1078 }, { "epoch": 0.04, "learning_rate": 9.869501466275661e-06, "loss": 0.1478, "step": 1079 }, { "epoch": 0.04, "learning_rate": 9.869368168488404e-06, "loss": 0.2206, "step": 1080 }, { "epoch": 0.04, "learning_rate": 9.869234870701148e-06, "loss": 0.1582, "step": 1081 }, { "epoch": 0.04, "learning_rate": 9.869101572913891e-06, "loss": 0.1417, "step": 1082 }, { "epoch": 0.04, "learning_rate": 9.868968275126634e-06, "loss": 0.1686, "step": 1083 }, { "epoch": 0.04, "learning_rate": 9.868834977339377e-06, "loss": 0.1417, "step": 1084 }, { "epoch": 0.04, "learning_rate": 9.868701679552121e-06, "loss": 0.1921, "step": 1085 }, { "epoch": 0.04, "learning_rate": 9.868568381764864e-06, "loss": 0.1885, "step": 1086 }, { "epoch": 0.04, "learning_rate": 9.868435083977607e-06, "loss": 0.1454, "step": 1087 }, { "epoch": 0.04, "learning_rate": 9.86830178619035e-06, "loss": 0.1564, "step": 1088 }, { "epoch": 0.04, "learning_rate": 9.868168488403094e-06, "loss": 0.1156, "step": 1089 }, { "epoch": 0.04, "learning_rate": 9.868035190615837e-06, "loss": 0.1387, "step": 1090 }, { "epoch": 0.04, "learning_rate": 9.86790189282858e-06, "loss": 0.1473, "step": 1091 }, { "epoch": 0.04, "learning_rate": 9.867768595041323e-06, "loss": 0.1313, "step": 1092 }, { "epoch": 0.04, "learning_rate": 9.867635297254067e-06, "loss": 0.1569, "step": 1093 }, { "epoch": 0.04, "learning_rate": 9.86750199946681e-06, "loss": 0.1378, "step": 1094 }, { "epoch": 0.04, "learning_rate": 9.867368701679553e-06, "loss": 0.1263, "step": 1095 }, { "epoch": 0.04, "learning_rate": 9.867235403892296e-06, "loss": 0.1853, "step": 1096 }, { "epoch": 0.04, "learning_rate": 9.86710210610504e-06, "loss": 0.1157, "step": 1097 }, { "epoch": 0.04, "learning_rate": 9.866968808317783e-06, "loss": 0.1615, "step": 1098 }, { "epoch": 0.04, "learning_rate": 9.866835510530526e-06, "loss": 0.1504, "step": 1099 }, { "epoch": 0.04, "learning_rate": 9.866702212743269e-06, "loss": 0.2074, "step": 1100 }, { "epoch": 0.04, "learning_rate": 9.866568914956013e-06, "loss": 0.1364, "step": 1101 }, { "epoch": 0.04, "learning_rate": 9.866435617168756e-06, "loss": 0.1417, "step": 1102 }, { "epoch": 0.04, "learning_rate": 9.866302319381499e-06, "loss": 0.1986, "step": 1103 }, { "epoch": 0.04, "learning_rate": 9.866169021594242e-06, "loss": 0.1466, "step": 1104 }, { "epoch": 0.04, "learning_rate": 9.866035723806986e-06, "loss": 0.0855, "step": 1105 }, { "epoch": 0.04, "learning_rate": 9.865902426019729e-06, "loss": 0.1666, "step": 1106 }, { "epoch": 0.04, "learning_rate": 9.865769128232472e-06, "loss": 0.1177, "step": 1107 }, { "epoch": 0.04, "learning_rate": 9.865635830445215e-06, "loss": 0.1829, "step": 1108 }, { "epoch": 0.04, "learning_rate": 9.86550253265796e-06, "loss": 0.1063, "step": 1109 }, { "epoch": 0.04, "learning_rate": 9.865369234870702e-06, "loss": 0.1191, "step": 1110 }, { "epoch": 0.04, "learning_rate": 9.865235937083445e-06, "loss": 0.1326, "step": 1111 }, { "epoch": 0.04, "learning_rate": 9.865102639296188e-06, "loss": 0.1228, "step": 1112 }, { "epoch": 0.04, "learning_rate": 9.864969341508932e-06, "loss": 0.1285, "step": 1113 }, { "epoch": 0.04, "learning_rate": 9.864836043721675e-06, "loss": 0.1195, "step": 1114 }, { "epoch": 0.04, "learning_rate": 9.864702745934418e-06, "loss": 0.222, "step": 1115 }, { "epoch": 0.04, "learning_rate": 9.864569448147162e-06, "loss": 0.1306, "step": 1116 }, { "epoch": 0.04, "learning_rate": 9.864436150359905e-06, "loss": 0.1237, "step": 1117 }, { "epoch": 0.04, "learning_rate": 9.864302852572648e-06, "loss": 0.1729, "step": 1118 }, { "epoch": 0.04, "learning_rate": 9.86416955478539e-06, "loss": 0.1586, "step": 1119 }, { "epoch": 0.04, "learning_rate": 9.864036256998135e-06, "loss": 0.1675, "step": 1120 }, { "epoch": 0.04, "learning_rate": 9.863902959210878e-06, "loss": 0.1437, "step": 1121 }, { "epoch": 0.04, "learning_rate": 9.863769661423621e-06, "loss": 0.143, "step": 1122 }, { "epoch": 0.04, "learning_rate": 9.863636363636364e-06, "loss": 0.1511, "step": 1123 }, { "epoch": 0.04, "learning_rate": 9.863503065849108e-06, "loss": 0.1275, "step": 1124 }, { "epoch": 0.04, "learning_rate": 9.863369768061851e-06, "loss": 0.1726, "step": 1125 }, { "epoch": 0.04, "learning_rate": 9.863236470274594e-06, "loss": 0.1357, "step": 1126 }, { "epoch": 0.05, "learning_rate": 9.863103172487337e-06, "loss": 0.1299, "step": 1127 }, { "epoch": 0.05, "learning_rate": 9.862969874700081e-06, "loss": 0.1288, "step": 1128 }, { "epoch": 0.05, "learning_rate": 9.862836576912824e-06, "loss": 0.1784, "step": 1129 }, { "epoch": 0.05, "learning_rate": 9.862703279125567e-06, "loss": 0.1633, "step": 1130 }, { "epoch": 0.05, "learning_rate": 9.86256998133831e-06, "loss": 0.1857, "step": 1131 }, { "epoch": 0.05, "learning_rate": 9.862436683551054e-06, "loss": 0.1715, "step": 1132 }, { "epoch": 0.05, "learning_rate": 9.862303385763797e-06, "loss": 0.1449, "step": 1133 }, { "epoch": 0.05, "learning_rate": 9.86217008797654e-06, "loss": 0.1308, "step": 1134 }, { "epoch": 0.05, "learning_rate": 9.862036790189283e-06, "loss": 0.1578, "step": 1135 }, { "epoch": 0.05, "learning_rate": 9.861903492402027e-06, "loss": 0.1262, "step": 1136 }, { "epoch": 0.05, "learning_rate": 9.86177019461477e-06, "loss": 0.1631, "step": 1137 }, { "epoch": 0.05, "learning_rate": 9.861636896827513e-06, "loss": 0.1701, "step": 1138 }, { "epoch": 0.05, "learning_rate": 9.861503599040256e-06, "loss": 0.1764, "step": 1139 }, { "epoch": 0.05, "learning_rate": 9.861370301253e-06, "loss": 0.1564, "step": 1140 }, { "epoch": 0.05, "learning_rate": 9.861237003465743e-06, "loss": 0.1483, "step": 1141 }, { "epoch": 0.05, "learning_rate": 9.861103705678486e-06, "loss": 0.1488, "step": 1142 }, { "epoch": 0.05, "learning_rate": 9.860970407891229e-06, "loss": 0.1641, "step": 1143 }, { "epoch": 0.05, "learning_rate": 9.860837110103973e-06, "loss": 0.2019, "step": 1144 }, { "epoch": 0.05, "learning_rate": 9.860703812316716e-06, "loss": 0.2056, "step": 1145 }, { "epoch": 0.05, "learning_rate": 9.860570514529459e-06, "loss": 0.1638, "step": 1146 }, { "epoch": 0.05, "learning_rate": 9.860437216742203e-06, "loss": 0.1602, "step": 1147 }, { "epoch": 0.05, "learning_rate": 9.860303918954946e-06, "loss": 0.1615, "step": 1148 }, { "epoch": 0.05, "learning_rate": 9.860170621167689e-06, "loss": 0.1504, "step": 1149 }, { "epoch": 0.05, "learning_rate": 9.860037323380432e-06, "loss": 0.1617, "step": 1150 }, { "epoch": 0.05, "learning_rate": 9.859904025593176e-06, "loss": 0.205, "step": 1151 }, { "epoch": 0.05, "learning_rate": 9.859770727805919e-06, "loss": 0.1519, "step": 1152 }, { "epoch": 0.05, "learning_rate": 9.859637430018662e-06, "loss": 0.154, "step": 1153 }, { "epoch": 0.05, "learning_rate": 9.859504132231406e-06, "loss": 0.1129, "step": 1154 }, { "epoch": 0.05, "learning_rate": 9.85937083444415e-06, "loss": 0.1418, "step": 1155 }, { "epoch": 0.05, "learning_rate": 9.859237536656892e-06, "loss": 0.1684, "step": 1156 }, { "epoch": 0.05, "learning_rate": 9.859104238869635e-06, "loss": 0.1165, "step": 1157 }, { "epoch": 0.05, "learning_rate": 9.85897094108238e-06, "loss": 0.1074, "step": 1158 }, { "epoch": 0.05, "learning_rate": 9.858837643295122e-06, "loss": 0.1393, "step": 1159 }, { "epoch": 0.05, "learning_rate": 9.858704345507865e-06, "loss": 0.1383, "step": 1160 }, { "epoch": 0.05, "learning_rate": 9.85857104772061e-06, "loss": 0.1408, "step": 1161 }, { "epoch": 0.05, "learning_rate": 9.858437749933352e-06, "loss": 0.1456, "step": 1162 }, { "epoch": 0.05, "learning_rate": 9.858304452146095e-06, "loss": 0.1634, "step": 1163 }, { "epoch": 0.05, "learning_rate": 9.858171154358838e-06, "loss": 0.1793, "step": 1164 }, { "epoch": 0.05, "learning_rate": 9.858037856571582e-06, "loss": 0.1347, "step": 1165 }, { "epoch": 0.05, "learning_rate": 9.857904558784325e-06, "loss": 0.1362, "step": 1166 }, { "epoch": 0.05, "learning_rate": 9.857771260997068e-06, "loss": 0.1466, "step": 1167 }, { "epoch": 0.05, "learning_rate": 9.857637963209813e-06, "loss": 0.1216, "step": 1168 }, { "epoch": 0.05, "learning_rate": 9.857504665422555e-06, "loss": 0.112, "step": 1169 }, { "epoch": 0.05, "learning_rate": 9.857371367635298e-06, "loss": 0.2115, "step": 1170 }, { "epoch": 0.05, "learning_rate": 9.857238069848041e-06, "loss": 0.1558, "step": 1171 }, { "epoch": 0.05, "learning_rate": 9.857104772060786e-06, "loss": 0.144, "step": 1172 }, { "epoch": 0.05, "learning_rate": 9.856971474273528e-06, "loss": 0.1741, "step": 1173 }, { "epoch": 0.05, "learning_rate": 9.856838176486271e-06, "loss": 0.146, "step": 1174 }, { "epoch": 0.05, "learning_rate": 9.856704878699016e-06, "loss": 0.1828, "step": 1175 }, { "epoch": 0.05, "learning_rate": 9.856571580911759e-06, "loss": 0.1692, "step": 1176 }, { "epoch": 0.05, "learning_rate": 9.856438283124501e-06, "loss": 0.1688, "step": 1177 }, { "epoch": 0.05, "learning_rate": 9.856304985337244e-06, "loss": 0.1398, "step": 1178 }, { "epoch": 0.05, "learning_rate": 9.856171687549989e-06, "loss": 0.1627, "step": 1179 }, { "epoch": 0.05, "learning_rate": 9.856038389762732e-06, "loss": 0.1726, "step": 1180 }, { "epoch": 0.05, "learning_rate": 9.855905091975474e-06, "loss": 0.1704, "step": 1181 }, { "epoch": 0.05, "learning_rate": 9.855771794188217e-06, "loss": 0.1605, "step": 1182 }, { "epoch": 0.05, "learning_rate": 9.855638496400962e-06, "loss": 0.1453, "step": 1183 }, { "epoch": 0.05, "learning_rate": 9.855505198613705e-06, "loss": 0.2029, "step": 1184 }, { "epoch": 0.05, "learning_rate": 9.855371900826447e-06, "loss": 0.1377, "step": 1185 }, { "epoch": 0.05, "learning_rate": 9.85523860303919e-06, "loss": 0.1554, "step": 1186 }, { "epoch": 0.05, "learning_rate": 9.855105305251935e-06, "loss": 0.1576, "step": 1187 }, { "epoch": 0.05, "learning_rate": 9.854972007464677e-06, "loss": 0.1807, "step": 1188 }, { "epoch": 0.05, "learning_rate": 9.85483870967742e-06, "loss": 0.1733, "step": 1189 }, { "epoch": 0.05, "learning_rate": 9.854705411890163e-06, "loss": 0.2, "step": 1190 }, { "epoch": 0.05, "learning_rate": 9.854572114102908e-06, "loss": 0.1706, "step": 1191 }, { "epoch": 0.05, "learning_rate": 9.85443881631565e-06, "loss": 0.1726, "step": 1192 }, { "epoch": 0.05, "learning_rate": 9.854305518528393e-06, "loss": 0.1466, "step": 1193 }, { "epoch": 0.05, "learning_rate": 9.854172220741136e-06, "loss": 0.1799, "step": 1194 }, { "epoch": 0.05, "learning_rate": 9.85403892295388e-06, "loss": 0.1938, "step": 1195 }, { "epoch": 0.05, "learning_rate": 9.853905625166623e-06, "loss": 0.1515, "step": 1196 }, { "epoch": 0.05, "learning_rate": 9.853772327379366e-06, "loss": 0.1618, "step": 1197 }, { "epoch": 0.05, "learning_rate": 9.853639029592109e-06, "loss": 0.1426, "step": 1198 }, { "epoch": 0.05, "learning_rate": 9.853505731804854e-06, "loss": 0.1661, "step": 1199 }, { "epoch": 0.05, "learning_rate": 9.853372434017596e-06, "loss": 0.1523, "step": 1200 }, { "epoch": 0.05, "learning_rate": 9.85323913623034e-06, "loss": 0.1323, "step": 1201 }, { "epoch": 0.05, "learning_rate": 9.853105838443082e-06, "loss": 0.1203, "step": 1202 }, { "epoch": 0.05, "learning_rate": 9.852972540655827e-06, "loss": 0.1625, "step": 1203 }, { "epoch": 0.05, "learning_rate": 9.85283924286857e-06, "loss": 0.123, "step": 1204 }, { "epoch": 0.05, "learning_rate": 9.852705945081312e-06, "loss": 0.1426, "step": 1205 }, { "epoch": 0.05, "learning_rate": 9.852572647294055e-06, "loss": 0.1143, "step": 1206 }, { "epoch": 0.05, "learning_rate": 9.8524393495068e-06, "loss": 0.13, "step": 1207 }, { "epoch": 0.05, "learning_rate": 9.852306051719542e-06, "loss": 0.1552, "step": 1208 }, { "epoch": 0.05, "learning_rate": 9.852172753932285e-06, "loss": 0.1566, "step": 1209 }, { "epoch": 0.05, "learning_rate": 9.852039456145028e-06, "loss": 0.0947, "step": 1210 }, { "epoch": 0.05, "learning_rate": 9.851906158357772e-06, "loss": 0.1307, "step": 1211 }, { "epoch": 0.05, "learning_rate": 9.851772860570515e-06, "loss": 0.1406, "step": 1212 }, { "epoch": 0.05, "learning_rate": 9.851639562783258e-06, "loss": 0.1437, "step": 1213 }, { "epoch": 0.05, "learning_rate": 9.851506264996001e-06, "loss": 0.1866, "step": 1214 }, { "epoch": 0.05, "learning_rate": 9.851372967208745e-06, "loss": 0.1453, "step": 1215 }, { "epoch": 0.05, "learning_rate": 9.851239669421488e-06, "loss": 0.1545, "step": 1216 }, { "epoch": 0.05, "learning_rate": 9.851106371634231e-06, "loss": 0.1434, "step": 1217 }, { "epoch": 0.05, "learning_rate": 9.850973073846974e-06, "loss": 0.176, "step": 1218 }, { "epoch": 0.05, "learning_rate": 9.850839776059718e-06, "loss": 0.1788, "step": 1219 }, { "epoch": 0.05, "learning_rate": 9.850706478272461e-06, "loss": 0.1622, "step": 1220 }, { "epoch": 0.05, "learning_rate": 9.850573180485204e-06, "loss": 0.1235, "step": 1221 }, { "epoch": 0.05, "learning_rate": 9.850439882697947e-06, "loss": 0.1851, "step": 1222 }, { "epoch": 0.05, "learning_rate": 9.850306584910691e-06, "loss": 0.1311, "step": 1223 }, { "epoch": 0.05, "learning_rate": 9.850173287123434e-06, "loss": 0.1262, "step": 1224 }, { "epoch": 0.05, "learning_rate": 9.850039989336177e-06, "loss": 0.1642, "step": 1225 }, { "epoch": 0.05, "learning_rate": 9.84990669154892e-06, "loss": 0.148, "step": 1226 }, { "epoch": 0.05, "learning_rate": 9.849773393761664e-06, "loss": 0.124, "step": 1227 }, { "epoch": 0.05, "learning_rate": 9.849640095974407e-06, "loss": 0.1405, "step": 1228 }, { "epoch": 0.05, "learning_rate": 9.84950679818715e-06, "loss": 0.1409, "step": 1229 }, { "epoch": 0.05, "learning_rate": 9.849373500399895e-06, "loss": 0.1779, "step": 1230 }, { "epoch": 0.05, "learning_rate": 9.849240202612637e-06, "loss": 0.1166, "step": 1231 }, { "epoch": 0.05, "learning_rate": 9.84910690482538e-06, "loss": 0.1556, "step": 1232 }, { "epoch": 0.05, "learning_rate": 9.848973607038123e-06, "loss": 0.1369, "step": 1233 }, { "epoch": 0.05, "learning_rate": 9.848840309250867e-06, "loss": 0.1366, "step": 1234 }, { "epoch": 0.05, "learning_rate": 9.84870701146361e-06, "loss": 0.1624, "step": 1235 }, { "epoch": 0.05, "learning_rate": 9.848573713676353e-06, "loss": 0.1705, "step": 1236 }, { "epoch": 0.05, "learning_rate": 9.848440415889096e-06, "loss": 0.1357, "step": 1237 }, { "epoch": 0.05, "learning_rate": 9.84830711810184e-06, "loss": 0.1097, "step": 1238 }, { "epoch": 0.05, "learning_rate": 9.848173820314583e-06, "loss": 0.149, "step": 1239 }, { "epoch": 0.05, "learning_rate": 9.848040522527326e-06, "loss": 0.1186, "step": 1240 }, { "epoch": 0.05, "learning_rate": 9.84790722474007e-06, "loss": 0.1092, "step": 1241 }, { "epoch": 0.05, "learning_rate": 9.847773926952813e-06, "loss": 0.1508, "step": 1242 }, { "epoch": 0.05, "learning_rate": 9.847640629165556e-06, "loss": 0.168, "step": 1243 }, { "epoch": 0.05, "learning_rate": 9.847507331378299e-06, "loss": 0.2179, "step": 1244 }, { "epoch": 0.05, "learning_rate": 9.847374033591044e-06, "loss": 0.1226, "step": 1245 }, { "epoch": 0.05, "learning_rate": 9.847240735803786e-06, "loss": 0.1152, "step": 1246 }, { "epoch": 0.05, "learning_rate": 9.84710743801653e-06, "loss": 0.1134, "step": 1247 }, { "epoch": 0.05, "learning_rate": 9.846974140229274e-06, "loss": 0.1271, "step": 1248 }, { "epoch": 0.05, "learning_rate": 9.846840842442017e-06, "loss": 0.1748, "step": 1249 }, { "epoch": 0.05, "learning_rate": 9.84670754465476e-06, "loss": 0.171, "step": 1250 }, { "epoch": 0.05, "learning_rate": 9.846574246867502e-06, "loss": 0.1708, "step": 1251 }, { "epoch": 0.05, "learning_rate": 9.846440949080247e-06, "loss": 0.1494, "step": 1252 }, { "epoch": 0.05, "learning_rate": 9.84630765129299e-06, "loss": 0.1298, "step": 1253 }, { "epoch": 0.05, "learning_rate": 9.846174353505732e-06, "loss": 0.1634, "step": 1254 }, { "epoch": 0.05, "learning_rate": 9.846041055718477e-06, "loss": 0.1409, "step": 1255 }, { "epoch": 0.05, "learning_rate": 9.84590775793122e-06, "loss": 0.1239, "step": 1256 }, { "epoch": 0.05, "learning_rate": 9.845774460143962e-06, "loss": 0.1435, "step": 1257 }, { "epoch": 0.05, "learning_rate": 9.845641162356705e-06, "loss": 0.1456, "step": 1258 }, { "epoch": 0.05, "learning_rate": 9.84550786456945e-06, "loss": 0.1657, "step": 1259 }, { "epoch": 0.05, "learning_rate": 9.845374566782193e-06, "loss": 0.149, "step": 1260 }, { "epoch": 0.05, "learning_rate": 9.845241268994935e-06, "loss": 0.1228, "step": 1261 }, { "epoch": 0.05, "learning_rate": 9.84510797120768e-06, "loss": 0.1432, "step": 1262 }, { "epoch": 0.05, "learning_rate": 9.844974673420423e-06, "loss": 0.1437, "step": 1263 }, { "epoch": 0.05, "learning_rate": 9.844841375633166e-06, "loss": 0.1698, "step": 1264 }, { "epoch": 0.05, "learning_rate": 9.844708077845908e-06, "loss": 0.1397, "step": 1265 }, { "epoch": 0.05, "learning_rate": 9.844574780058653e-06, "loss": 0.1638, "step": 1266 }, { "epoch": 0.05, "learning_rate": 9.844441482271396e-06, "loss": 0.1684, "step": 1267 }, { "epoch": 0.05, "learning_rate": 9.844308184484139e-06, "loss": 0.1382, "step": 1268 }, { "epoch": 0.05, "learning_rate": 9.844174886696881e-06, "loss": 0.1315, "step": 1269 }, { "epoch": 0.05, "learning_rate": 9.844041588909626e-06, "loss": 0.1407, "step": 1270 }, { "epoch": 0.05, "learning_rate": 9.843908291122369e-06, "loss": 0.2015, "step": 1271 }, { "epoch": 0.05, "learning_rate": 9.843774993335112e-06, "loss": 0.2166, "step": 1272 }, { "epoch": 0.05, "learning_rate": 9.843641695547854e-06, "loss": 0.1576, "step": 1273 }, { "epoch": 0.05, "learning_rate": 9.843508397760599e-06, "loss": 0.1552, "step": 1274 }, { "epoch": 0.05, "learning_rate": 9.843375099973342e-06, "loss": 0.0959, "step": 1275 }, { "epoch": 0.05, "learning_rate": 9.843241802186085e-06, "loss": 0.1151, "step": 1276 }, { "epoch": 0.05, "learning_rate": 9.843108504398827e-06, "loss": 0.155, "step": 1277 }, { "epoch": 0.05, "learning_rate": 9.842975206611572e-06, "loss": 0.1543, "step": 1278 }, { "epoch": 0.05, "learning_rate": 9.842841908824315e-06, "loss": 0.1537, "step": 1279 }, { "epoch": 0.05, "learning_rate": 9.842708611037057e-06, "loss": 0.1683, "step": 1280 }, { "epoch": 0.05, "learning_rate": 9.8425753132498e-06, "loss": 0.1513, "step": 1281 }, { "epoch": 0.05, "learning_rate": 9.842442015462545e-06, "loss": 0.1458, "step": 1282 }, { "epoch": 0.05, "learning_rate": 9.842308717675288e-06, "loss": 0.1487, "step": 1283 }, { "epoch": 0.05, "learning_rate": 9.84217541988803e-06, "loss": 0.1414, "step": 1284 }, { "epoch": 0.05, "learning_rate": 9.842042122100775e-06, "loss": 0.1622, "step": 1285 }, { "epoch": 0.05, "learning_rate": 9.841908824313518e-06, "loss": 0.1318, "step": 1286 }, { "epoch": 0.05, "learning_rate": 9.84177552652626e-06, "loss": 0.0979, "step": 1287 }, { "epoch": 0.05, "learning_rate": 9.841642228739003e-06, "loss": 0.1429, "step": 1288 }, { "epoch": 0.05, "learning_rate": 9.841508930951748e-06, "loss": 0.1356, "step": 1289 }, { "epoch": 0.05, "learning_rate": 9.84137563316449e-06, "loss": 0.1025, "step": 1290 }, { "epoch": 0.05, "learning_rate": 9.841242335377234e-06, "loss": 0.1603, "step": 1291 }, { "epoch": 0.05, "learning_rate": 9.841109037589976e-06, "loss": 0.1569, "step": 1292 }, { "epoch": 0.05, "learning_rate": 9.840975739802721e-06, "loss": 0.1443, "step": 1293 }, { "epoch": 0.05, "learning_rate": 9.840842442015464e-06, "loss": 0.1304, "step": 1294 }, { "epoch": 0.05, "learning_rate": 9.840709144228207e-06, "loss": 0.1415, "step": 1295 }, { "epoch": 0.05, "learning_rate": 9.84057584644095e-06, "loss": 0.1522, "step": 1296 }, { "epoch": 0.05, "learning_rate": 9.840442548653694e-06, "loss": 0.1168, "step": 1297 }, { "epoch": 0.05, "learning_rate": 9.840309250866437e-06, "loss": 0.1443, "step": 1298 }, { "epoch": 0.05, "learning_rate": 9.84017595307918e-06, "loss": 0.178, "step": 1299 }, { "epoch": 0.05, "learning_rate": 9.840042655291922e-06, "loss": 0.2247, "step": 1300 }, { "epoch": 0.05, "learning_rate": 9.839909357504667e-06, "loss": 0.1626, "step": 1301 }, { "epoch": 0.05, "learning_rate": 9.83977605971741e-06, "loss": 0.1468, "step": 1302 }, { "epoch": 0.05, "learning_rate": 9.839642761930152e-06, "loss": 0.0967, "step": 1303 }, { "epoch": 0.05, "learning_rate": 9.839509464142895e-06, "loss": 0.1152, "step": 1304 }, { "epoch": 0.05, "learning_rate": 9.83937616635564e-06, "loss": 0.1769, "step": 1305 }, { "epoch": 0.05, "learning_rate": 9.839242868568383e-06, "loss": 0.1338, "step": 1306 }, { "epoch": 0.05, "learning_rate": 9.839109570781125e-06, "loss": 0.1554, "step": 1307 }, { "epoch": 0.05, "learning_rate": 9.838976272993868e-06, "loss": 0.1667, "step": 1308 }, { "epoch": 0.05, "learning_rate": 9.838842975206613e-06, "loss": 0.122, "step": 1309 }, { "epoch": 0.05, "learning_rate": 9.838709677419356e-06, "loss": 0.16, "step": 1310 }, { "epoch": 0.05, "learning_rate": 9.838576379632098e-06, "loss": 0.1302, "step": 1311 }, { "epoch": 0.05, "learning_rate": 9.838443081844841e-06, "loss": 0.1379, "step": 1312 }, { "epoch": 0.05, "learning_rate": 9.838309784057586e-06, "loss": 0.1398, "step": 1313 }, { "epoch": 0.05, "learning_rate": 9.838176486270329e-06, "loss": 0.1267, "step": 1314 }, { "epoch": 0.05, "learning_rate": 9.838043188483071e-06, "loss": 0.1651, "step": 1315 }, { "epoch": 0.05, "learning_rate": 9.837909890695814e-06, "loss": 0.1082, "step": 1316 }, { "epoch": 0.05, "learning_rate": 9.837776592908559e-06, "loss": 0.1274, "step": 1317 }, { "epoch": 0.05, "learning_rate": 9.837643295121302e-06, "loss": 0.1604, "step": 1318 }, { "epoch": 0.05, "learning_rate": 9.837509997334044e-06, "loss": 0.1613, "step": 1319 }, { "epoch": 0.05, "learning_rate": 9.837376699546787e-06, "loss": 0.098, "step": 1320 }, { "epoch": 0.05, "learning_rate": 9.837243401759532e-06, "loss": 0.1148, "step": 1321 }, { "epoch": 0.05, "learning_rate": 9.837110103972275e-06, "loss": 0.1866, "step": 1322 }, { "epoch": 0.05, "learning_rate": 9.836976806185017e-06, "loss": 0.191, "step": 1323 }, { "epoch": 0.05, "learning_rate": 9.83684350839776e-06, "loss": 0.1493, "step": 1324 }, { "epoch": 0.05, "learning_rate": 9.836710210610505e-06, "loss": 0.1271, "step": 1325 }, { "epoch": 0.05, "learning_rate": 9.836576912823247e-06, "loss": 0.135, "step": 1326 }, { "epoch": 0.05, "learning_rate": 9.83644361503599e-06, "loss": 0.1167, "step": 1327 }, { "epoch": 0.05, "learning_rate": 9.836310317248733e-06, "loss": 0.1408, "step": 1328 }, { "epoch": 0.05, "learning_rate": 9.836177019461478e-06, "loss": 0.1655, "step": 1329 }, { "epoch": 0.05, "learning_rate": 9.83604372167422e-06, "loss": 0.1438, "step": 1330 }, { "epoch": 0.05, "learning_rate": 9.835910423886963e-06, "loss": 0.218, "step": 1331 }, { "epoch": 0.05, "learning_rate": 9.835777126099708e-06, "loss": 0.1425, "step": 1332 }, { "epoch": 0.05, "learning_rate": 9.83564382831245e-06, "loss": 0.1564, "step": 1333 }, { "epoch": 0.05, "learning_rate": 9.835510530525193e-06, "loss": 0.1394, "step": 1334 }, { "epoch": 0.05, "learning_rate": 9.835377232737936e-06, "loss": 0.1369, "step": 1335 }, { "epoch": 0.05, "learning_rate": 9.83524393495068e-06, "loss": 0.1384, "step": 1336 }, { "epoch": 0.05, "learning_rate": 9.835110637163424e-06, "loss": 0.1463, "step": 1337 }, { "epoch": 0.05, "learning_rate": 9.834977339376166e-06, "loss": 0.1379, "step": 1338 }, { "epoch": 0.05, "learning_rate": 9.834844041588911e-06, "loss": 0.1415, "step": 1339 }, { "epoch": 0.05, "learning_rate": 9.834710743801654e-06, "loss": 0.1824, "step": 1340 }, { "epoch": 0.05, "learning_rate": 9.834577446014397e-06, "loss": 0.1994, "step": 1341 }, { "epoch": 0.05, "learning_rate": 9.83444414822714e-06, "loss": 0.149, "step": 1342 }, { "epoch": 0.05, "learning_rate": 9.834310850439884e-06, "loss": 0.1501, "step": 1343 }, { "epoch": 0.05, "learning_rate": 9.834177552652627e-06, "loss": 0.1468, "step": 1344 }, { "epoch": 0.05, "learning_rate": 9.83404425486537e-06, "loss": 0.1843, "step": 1345 }, { "epoch": 0.05, "learning_rate": 9.833910957078114e-06, "loss": 0.1689, "step": 1346 }, { "epoch": 0.05, "learning_rate": 9.833777659290857e-06, "loss": 0.097, "step": 1347 }, { "epoch": 0.05, "learning_rate": 9.8336443615036e-06, "loss": 0.1179, "step": 1348 }, { "epoch": 0.05, "learning_rate": 9.833511063716342e-06, "loss": 0.1773, "step": 1349 }, { "epoch": 0.05, "learning_rate": 9.833377765929087e-06, "loss": 0.111, "step": 1350 }, { "epoch": 0.05, "learning_rate": 9.83324446814183e-06, "loss": 0.1147, "step": 1351 }, { "epoch": 0.05, "learning_rate": 9.833111170354573e-06, "loss": 0.1492, "step": 1352 }, { "epoch": 0.05, "learning_rate": 9.832977872567317e-06, "loss": 0.1335, "step": 1353 }, { "epoch": 0.05, "learning_rate": 9.83284457478006e-06, "loss": 0.1356, "step": 1354 }, { "epoch": 0.05, "learning_rate": 9.832711276992803e-06, "loss": 0.1394, "step": 1355 }, { "epoch": 0.05, "learning_rate": 9.832577979205546e-06, "loss": 0.1976, "step": 1356 }, { "epoch": 0.05, "learning_rate": 9.83244468141829e-06, "loss": 0.1188, "step": 1357 }, { "epoch": 0.05, "learning_rate": 9.832311383631033e-06, "loss": 0.1266, "step": 1358 }, { "epoch": 0.05, "learning_rate": 9.832178085843776e-06, "loss": 0.1649, "step": 1359 }, { "epoch": 0.05, "learning_rate": 9.83204478805652e-06, "loss": 0.1351, "step": 1360 }, { "epoch": 0.05, "learning_rate": 9.831911490269263e-06, "loss": 0.1883, "step": 1361 }, { "epoch": 0.05, "learning_rate": 9.831778192482006e-06, "loss": 0.1735, "step": 1362 }, { "epoch": 0.05, "learning_rate": 9.831644894694749e-06, "loss": 0.1577, "step": 1363 }, { "epoch": 0.05, "learning_rate": 9.831511596907493e-06, "loss": 0.1288, "step": 1364 }, { "epoch": 0.05, "learning_rate": 9.831378299120236e-06, "loss": 0.1658, "step": 1365 }, { "epoch": 0.05, "learning_rate": 9.831245001332979e-06, "loss": 0.1476, "step": 1366 }, { "epoch": 0.05, "learning_rate": 9.831111703545722e-06, "loss": 0.1014, "step": 1367 }, { "epoch": 0.05, "learning_rate": 9.830978405758466e-06, "loss": 0.1267, "step": 1368 }, { "epoch": 0.05, "learning_rate": 9.830845107971209e-06, "loss": 0.1506, "step": 1369 }, { "epoch": 0.05, "learning_rate": 9.830711810183952e-06, "loss": 0.1443, "step": 1370 }, { "epoch": 0.05, "learning_rate": 9.830578512396695e-06, "loss": 0.1313, "step": 1371 }, { "epoch": 0.05, "learning_rate": 9.83044521460944e-06, "loss": 0.1227, "step": 1372 }, { "epoch": 0.05, "learning_rate": 9.830311916822182e-06, "loss": 0.1404, "step": 1373 }, { "epoch": 0.05, "learning_rate": 9.830178619034925e-06, "loss": 0.1297, "step": 1374 }, { "epoch": 0.05, "learning_rate": 9.830045321247668e-06, "loss": 0.1788, "step": 1375 }, { "epoch": 0.05, "learning_rate": 9.829912023460412e-06, "loss": 0.164, "step": 1376 }, { "epoch": 0.05, "learning_rate": 9.829778725673155e-06, "loss": 0.117, "step": 1377 }, { "epoch": 0.06, "learning_rate": 9.829645427885898e-06, "loss": 0.1179, "step": 1378 }, { "epoch": 0.06, "learning_rate": 9.82951213009864e-06, "loss": 0.139, "step": 1379 }, { "epoch": 0.06, "learning_rate": 9.829378832311385e-06, "loss": 0.2051, "step": 1380 }, { "epoch": 0.06, "learning_rate": 9.829245534524128e-06, "loss": 0.1343, "step": 1381 }, { "epoch": 0.06, "learning_rate": 9.82911223673687e-06, "loss": 0.1447, "step": 1382 }, { "epoch": 0.06, "learning_rate": 9.828978938949614e-06, "loss": 0.1441, "step": 1383 }, { "epoch": 0.06, "learning_rate": 9.828845641162358e-06, "loss": 0.1306, "step": 1384 }, { "epoch": 0.06, "learning_rate": 9.828712343375101e-06, "loss": 0.1501, "step": 1385 }, { "epoch": 0.06, "learning_rate": 9.828579045587844e-06, "loss": 0.1437, "step": 1386 }, { "epoch": 0.06, "learning_rate": 9.828445747800587e-06, "loss": 0.1449, "step": 1387 }, { "epoch": 0.06, "learning_rate": 9.828312450013331e-06, "loss": 0.1147, "step": 1388 }, { "epoch": 0.06, "learning_rate": 9.828179152226074e-06, "loss": 0.1477, "step": 1389 }, { "epoch": 0.06, "learning_rate": 9.828045854438817e-06, "loss": 0.1484, "step": 1390 }, { "epoch": 0.06, "learning_rate": 9.82791255665156e-06, "loss": 0.1157, "step": 1391 }, { "epoch": 0.06, "learning_rate": 9.827779258864304e-06, "loss": 0.1364, "step": 1392 }, { "epoch": 0.06, "learning_rate": 9.827645961077047e-06, "loss": 0.1453, "step": 1393 }, { "epoch": 0.06, "learning_rate": 9.82751266328979e-06, "loss": 0.1527, "step": 1394 }, { "epoch": 0.06, "learning_rate": 9.827379365502533e-06, "loss": 0.1373, "step": 1395 }, { "epoch": 0.06, "learning_rate": 9.827246067715277e-06, "loss": 0.131, "step": 1396 }, { "epoch": 0.06, "learning_rate": 9.82711276992802e-06, "loss": 0.1874, "step": 1397 }, { "epoch": 0.06, "learning_rate": 9.826979472140763e-06, "loss": 0.1285, "step": 1398 }, { "epoch": 0.06, "learning_rate": 9.826846174353507e-06, "loss": 0.1975, "step": 1399 }, { "epoch": 0.06, "learning_rate": 9.82671287656625e-06, "loss": 0.1625, "step": 1400 }, { "epoch": 0.06, "learning_rate": 9.826579578778993e-06, "loss": 0.2216, "step": 1401 }, { "epoch": 0.06, "learning_rate": 9.826446280991736e-06, "loss": 0.1442, "step": 1402 }, { "epoch": 0.06, "learning_rate": 9.82631298320448e-06, "loss": 0.121, "step": 1403 }, { "epoch": 0.06, "learning_rate": 9.826179685417223e-06, "loss": 0.1845, "step": 1404 }, { "epoch": 0.06, "learning_rate": 9.826046387629966e-06, "loss": 0.1393, "step": 1405 }, { "epoch": 0.06, "learning_rate": 9.825913089842709e-06, "loss": 0.1302, "step": 1406 }, { "epoch": 0.06, "learning_rate": 9.825779792055453e-06, "loss": 0.1867, "step": 1407 }, { "epoch": 0.06, "learning_rate": 9.825646494268196e-06, "loss": 0.1665, "step": 1408 }, { "epoch": 0.06, "learning_rate": 9.825513196480939e-06, "loss": 0.1191, "step": 1409 }, { "epoch": 0.06, "learning_rate": 9.825379898693682e-06, "loss": 0.1181, "step": 1410 }, { "epoch": 0.06, "learning_rate": 9.825246600906426e-06, "loss": 0.1961, "step": 1411 }, { "epoch": 0.06, "learning_rate": 9.825113303119169e-06, "loss": 0.1187, "step": 1412 }, { "epoch": 0.06, "learning_rate": 9.824980005331912e-06, "loss": 0.1631, "step": 1413 }, { "epoch": 0.06, "learning_rate": 9.824846707544655e-06, "loss": 0.1239, "step": 1414 }, { "epoch": 0.06, "learning_rate": 9.824713409757399e-06, "loss": 0.1156, "step": 1415 }, { "epoch": 0.06, "learning_rate": 9.824580111970142e-06, "loss": 0.1391, "step": 1416 }, { "epoch": 0.06, "learning_rate": 9.824446814182885e-06, "loss": 0.1536, "step": 1417 }, { "epoch": 0.06, "learning_rate": 9.824313516395628e-06, "loss": 0.1198, "step": 1418 }, { "epoch": 0.06, "learning_rate": 9.824180218608372e-06, "loss": 0.1751, "step": 1419 }, { "epoch": 0.06, "learning_rate": 9.824046920821115e-06, "loss": 0.1563, "step": 1420 }, { "epoch": 0.06, "learning_rate": 9.823913623033858e-06, "loss": 0.1035, "step": 1421 }, { "epoch": 0.06, "learning_rate": 9.8237803252466e-06, "loss": 0.1358, "step": 1422 }, { "epoch": 0.06, "learning_rate": 9.823647027459345e-06, "loss": 0.1284, "step": 1423 }, { "epoch": 0.06, "learning_rate": 9.823513729672088e-06, "loss": 0.1242, "step": 1424 }, { "epoch": 0.06, "learning_rate": 9.82338043188483e-06, "loss": 0.165, "step": 1425 }, { "epoch": 0.06, "learning_rate": 9.823247134097575e-06, "loss": 0.107, "step": 1426 }, { "epoch": 0.06, "learning_rate": 9.823113836310318e-06, "loss": 0.163, "step": 1427 }, { "epoch": 0.06, "learning_rate": 9.82298053852306e-06, "loss": 0.1406, "step": 1428 }, { "epoch": 0.06, "learning_rate": 9.822847240735804e-06, "loss": 0.1245, "step": 1429 }, { "epoch": 0.06, "learning_rate": 9.822713942948548e-06, "loss": 0.126, "step": 1430 }, { "epoch": 0.06, "learning_rate": 9.822580645161291e-06, "loss": 0.1406, "step": 1431 }, { "epoch": 0.06, "learning_rate": 9.822447347374034e-06, "loss": 0.1518, "step": 1432 }, { "epoch": 0.06, "learning_rate": 9.822314049586778e-06, "loss": 0.1498, "step": 1433 }, { "epoch": 0.06, "learning_rate": 9.822180751799521e-06, "loss": 0.1804, "step": 1434 }, { "epoch": 0.06, "learning_rate": 9.822047454012264e-06, "loss": 0.1049, "step": 1435 }, { "epoch": 0.06, "learning_rate": 9.821914156225007e-06, "loss": 0.1829, "step": 1436 }, { "epoch": 0.06, "learning_rate": 9.821780858437751e-06, "loss": 0.1576, "step": 1437 }, { "epoch": 0.06, "learning_rate": 9.821647560650494e-06, "loss": 0.1616, "step": 1438 }, { "epoch": 0.06, "learning_rate": 9.821514262863237e-06, "loss": 0.1432, "step": 1439 }, { "epoch": 0.06, "learning_rate": 9.821380965075981e-06, "loss": 0.1437, "step": 1440 }, { "epoch": 0.06, "learning_rate": 9.821247667288724e-06, "loss": 0.1554, "step": 1441 }, { "epoch": 0.06, "learning_rate": 9.821114369501467e-06, "loss": 0.1769, "step": 1442 }, { "epoch": 0.06, "learning_rate": 9.82098107171421e-06, "loss": 0.1379, "step": 1443 }, { "epoch": 0.06, "learning_rate": 9.820847773926954e-06, "loss": 0.1534, "step": 1444 }, { "epoch": 0.06, "learning_rate": 9.820714476139697e-06, "loss": 0.1527, "step": 1445 }, { "epoch": 0.06, "learning_rate": 9.82058117835244e-06, "loss": 0.1452, "step": 1446 }, { "epoch": 0.06, "learning_rate": 9.820447880565185e-06, "loss": 0.1046, "step": 1447 }, { "epoch": 0.06, "learning_rate": 9.820314582777927e-06, "loss": 0.1395, "step": 1448 }, { "epoch": 0.06, "learning_rate": 9.82018128499067e-06, "loss": 0.1899, "step": 1449 }, { "epoch": 0.06, "learning_rate": 9.820047987203413e-06, "loss": 0.1275, "step": 1450 }, { "epoch": 0.06, "learning_rate": 9.819914689416157e-06, "loss": 0.1513, "step": 1451 }, { "epoch": 0.06, "learning_rate": 9.8197813916289e-06, "loss": 0.1013, "step": 1452 }, { "epoch": 0.06, "learning_rate": 9.819648093841643e-06, "loss": 0.1444, "step": 1453 }, { "epoch": 0.06, "learning_rate": 9.819514796054388e-06, "loss": 0.1307, "step": 1454 }, { "epoch": 0.06, "learning_rate": 9.81938149826713e-06, "loss": 0.0958, "step": 1455 }, { "epoch": 0.06, "learning_rate": 9.819248200479873e-06, "loss": 0.128, "step": 1456 }, { "epoch": 0.06, "learning_rate": 9.819114902692616e-06, "loss": 0.191, "step": 1457 }, { "epoch": 0.06, "learning_rate": 9.81898160490536e-06, "loss": 0.1289, "step": 1458 }, { "epoch": 0.06, "learning_rate": 9.818848307118103e-06, "loss": 0.1145, "step": 1459 }, { "epoch": 0.06, "learning_rate": 9.818715009330846e-06, "loss": 0.1153, "step": 1460 }, { "epoch": 0.06, "learning_rate": 9.818581711543589e-06, "loss": 0.1446, "step": 1461 }, { "epoch": 0.06, "learning_rate": 9.818448413756334e-06, "loss": 0.1664, "step": 1462 }, { "epoch": 0.06, "learning_rate": 9.818315115969076e-06, "loss": 0.115, "step": 1463 }, { "epoch": 0.06, "learning_rate": 9.81818181818182e-06, "loss": 0.1544, "step": 1464 }, { "epoch": 0.06, "learning_rate": 9.818048520394562e-06, "loss": 0.1226, "step": 1465 }, { "epoch": 0.06, "learning_rate": 9.817915222607307e-06, "loss": 0.1252, "step": 1466 }, { "epoch": 0.06, "learning_rate": 9.81778192482005e-06, "loss": 0.1397, "step": 1467 }, { "epoch": 0.06, "learning_rate": 9.817648627032792e-06, "loss": 0.1209, "step": 1468 }, { "epoch": 0.06, "learning_rate": 9.817515329245535e-06, "loss": 0.1615, "step": 1469 }, { "epoch": 0.06, "learning_rate": 9.81738203145828e-06, "loss": 0.1518, "step": 1470 }, { "epoch": 0.06, "learning_rate": 9.817248733671022e-06, "loss": 0.1117, "step": 1471 }, { "epoch": 0.06, "learning_rate": 9.817115435883765e-06, "loss": 0.1666, "step": 1472 }, { "epoch": 0.06, "learning_rate": 9.816982138096508e-06, "loss": 0.146, "step": 1473 }, { "epoch": 0.06, "learning_rate": 9.816848840309252e-06, "loss": 0.1318, "step": 1474 }, { "epoch": 0.06, "learning_rate": 9.816715542521995e-06, "loss": 0.1566, "step": 1475 }, { "epoch": 0.06, "learning_rate": 9.816582244734738e-06, "loss": 0.1113, "step": 1476 }, { "epoch": 0.06, "learning_rate": 9.816448946947481e-06, "loss": 0.1291, "step": 1477 }, { "epoch": 0.06, "learning_rate": 9.816315649160225e-06, "loss": 0.1657, "step": 1478 }, { "epoch": 0.06, "learning_rate": 9.816182351372968e-06, "loss": 0.2098, "step": 1479 }, { "epoch": 0.06, "learning_rate": 9.816049053585711e-06, "loss": 0.1511, "step": 1480 }, { "epoch": 0.06, "learning_rate": 9.815915755798454e-06, "loss": 0.1592, "step": 1481 }, { "epoch": 0.06, "learning_rate": 9.815782458011198e-06, "loss": 0.1497, "step": 1482 }, { "epoch": 0.06, "learning_rate": 9.815649160223941e-06, "loss": 0.1518, "step": 1483 }, { "epoch": 0.06, "learning_rate": 9.815515862436684e-06, "loss": 0.1462, "step": 1484 }, { "epoch": 0.06, "learning_rate": 9.815382564649427e-06, "loss": 0.1473, "step": 1485 }, { "epoch": 0.06, "learning_rate": 9.815249266862171e-06, "loss": 0.1468, "step": 1486 }, { "epoch": 0.06, "learning_rate": 9.815115969074914e-06, "loss": 0.118, "step": 1487 }, { "epoch": 0.06, "learning_rate": 9.814982671287657e-06, "loss": 0.111, "step": 1488 }, { "epoch": 0.06, "learning_rate": 9.8148493735004e-06, "loss": 0.1452, "step": 1489 }, { "epoch": 0.06, "learning_rate": 9.814716075713144e-06, "loss": 0.1109, "step": 1490 }, { "epoch": 0.06, "learning_rate": 9.814582777925887e-06, "loss": 0.1761, "step": 1491 }, { "epoch": 0.06, "learning_rate": 9.81444948013863e-06, "loss": 0.1219, "step": 1492 }, { "epoch": 0.06, "learning_rate": 9.814316182351373e-06, "loss": 0.1749, "step": 1493 }, { "epoch": 0.06, "learning_rate": 9.814182884564117e-06, "loss": 0.1278, "step": 1494 }, { "epoch": 0.06, "learning_rate": 9.81404958677686e-06, "loss": 0.1392, "step": 1495 }, { "epoch": 0.06, "learning_rate": 9.813916288989603e-06, "loss": 0.216, "step": 1496 }, { "epoch": 0.06, "learning_rate": 9.813782991202346e-06, "loss": 0.1389, "step": 1497 }, { "epoch": 0.06, "learning_rate": 9.81364969341509e-06, "loss": 0.1542, "step": 1498 }, { "epoch": 0.06, "learning_rate": 9.813516395627833e-06, "loss": 0.1655, "step": 1499 }, { "epoch": 0.06, "learning_rate": 9.813383097840576e-06, "loss": 0.158, "step": 1500 }, { "epoch": 0.06, "learning_rate": 9.813249800053319e-06, "loss": 0.1367, "step": 1501 }, { "epoch": 0.06, "learning_rate": 9.813116502266063e-06, "loss": 0.1287, "step": 1502 }, { "epoch": 0.06, "learning_rate": 9.812983204478806e-06, "loss": 0.1525, "step": 1503 }, { "epoch": 0.06, "learning_rate": 9.812849906691549e-06, "loss": 0.1405, "step": 1504 }, { "epoch": 0.06, "learning_rate": 9.812716608904292e-06, "loss": 0.1071, "step": 1505 }, { "epoch": 0.06, "learning_rate": 9.812583311117036e-06, "loss": 0.1519, "step": 1506 }, { "epoch": 0.06, "learning_rate": 9.812450013329779e-06, "loss": 0.1303, "step": 1507 }, { "epoch": 0.06, "learning_rate": 9.812316715542522e-06, "loss": 0.1194, "step": 1508 }, { "epoch": 0.06, "learning_rate": 9.812183417755266e-06, "loss": 0.1356, "step": 1509 }, { "epoch": 0.06, "learning_rate": 9.81205011996801e-06, "loss": 0.1597, "step": 1510 }, { "epoch": 0.06, "learning_rate": 9.811916822180752e-06, "loss": 0.129, "step": 1511 }, { "epoch": 0.06, "learning_rate": 9.811783524393495e-06, "loss": 0.1333, "step": 1512 }, { "epoch": 0.06, "learning_rate": 9.81165022660624e-06, "loss": 0.1241, "step": 1513 }, { "epoch": 0.06, "learning_rate": 9.811516928818982e-06, "loss": 0.1581, "step": 1514 }, { "epoch": 0.06, "learning_rate": 9.811383631031725e-06, "loss": 0.1279, "step": 1515 }, { "epoch": 0.06, "learning_rate": 9.811250333244468e-06, "loss": 0.1723, "step": 1516 }, { "epoch": 0.06, "learning_rate": 9.811117035457212e-06, "loss": 0.2155, "step": 1517 }, { "epoch": 0.06, "learning_rate": 9.810983737669955e-06, "loss": 0.1333, "step": 1518 }, { "epoch": 0.06, "learning_rate": 9.810850439882698e-06, "loss": 0.1721, "step": 1519 }, { "epoch": 0.06, "learning_rate": 9.810717142095442e-06, "loss": 0.1272, "step": 1520 }, { "epoch": 0.06, "learning_rate": 9.810583844308185e-06, "loss": 0.1474, "step": 1521 }, { "epoch": 0.06, "learning_rate": 9.810450546520928e-06, "loss": 0.1538, "step": 1522 }, { "epoch": 0.06, "learning_rate": 9.810317248733671e-06, "loss": 0.1513, "step": 1523 }, { "epoch": 0.06, "learning_rate": 9.810183950946415e-06, "loss": 0.1312, "step": 1524 }, { "epoch": 0.06, "learning_rate": 9.810050653159158e-06, "loss": 0.1216, "step": 1525 }, { "epoch": 0.06, "learning_rate": 9.809917355371901e-06, "loss": 0.143, "step": 1526 }, { "epoch": 0.06, "learning_rate": 9.809784057584646e-06, "loss": 0.1626, "step": 1527 }, { "epoch": 0.06, "learning_rate": 9.809650759797388e-06, "loss": 0.1183, "step": 1528 }, { "epoch": 0.06, "learning_rate": 9.809517462010131e-06, "loss": 0.1417, "step": 1529 }, { "epoch": 0.06, "learning_rate": 9.809384164222874e-06, "loss": 0.1144, "step": 1530 }, { "epoch": 0.06, "learning_rate": 9.809250866435619e-06, "loss": 0.1325, "step": 1531 }, { "epoch": 0.06, "learning_rate": 9.809117568648361e-06, "loss": 0.1181, "step": 1532 }, { "epoch": 0.06, "learning_rate": 9.808984270861104e-06, "loss": 0.0977, "step": 1533 }, { "epoch": 0.06, "learning_rate": 9.808850973073849e-06, "loss": 0.1595, "step": 1534 }, { "epoch": 0.06, "learning_rate": 9.808717675286592e-06, "loss": 0.1598, "step": 1535 }, { "epoch": 0.06, "learning_rate": 9.808584377499334e-06, "loss": 0.1638, "step": 1536 }, { "epoch": 0.06, "learning_rate": 9.808451079712077e-06, "loss": 0.0842, "step": 1537 }, { "epoch": 0.06, "learning_rate": 9.808317781924822e-06, "loss": 0.1765, "step": 1538 }, { "epoch": 0.06, "learning_rate": 9.808184484137565e-06, "loss": 0.1466, "step": 1539 }, { "epoch": 0.06, "learning_rate": 9.808051186350307e-06, "loss": 0.1186, "step": 1540 }, { "epoch": 0.06, "learning_rate": 9.807917888563052e-06, "loss": 0.1364, "step": 1541 }, { "epoch": 0.06, "learning_rate": 9.807784590775795e-06, "loss": 0.1395, "step": 1542 }, { "epoch": 0.06, "learning_rate": 9.807651292988537e-06, "loss": 0.172, "step": 1543 }, { "epoch": 0.06, "learning_rate": 9.80751799520128e-06, "loss": 0.1899, "step": 1544 }, { "epoch": 0.06, "learning_rate": 9.807384697414025e-06, "loss": 0.1408, "step": 1545 }, { "epoch": 0.06, "learning_rate": 9.807251399626768e-06, "loss": 0.148, "step": 1546 }, { "epoch": 0.06, "learning_rate": 9.80711810183951e-06, "loss": 0.2109, "step": 1547 }, { "epoch": 0.06, "learning_rate": 9.806984804052253e-06, "loss": 0.1319, "step": 1548 }, { "epoch": 0.06, "learning_rate": 9.806851506264998e-06, "loss": 0.1039, "step": 1549 }, { "epoch": 0.06, "learning_rate": 9.80671820847774e-06, "loss": 0.1406, "step": 1550 }, { "epoch": 0.06, "learning_rate": 9.806584910690483e-06, "loss": 0.1135, "step": 1551 }, { "epoch": 0.06, "learning_rate": 9.806451612903226e-06, "loss": 0.1363, "step": 1552 }, { "epoch": 0.06, "learning_rate": 9.80631831511597e-06, "loss": 0.096, "step": 1553 }, { "epoch": 0.06, "learning_rate": 9.806185017328714e-06, "loss": 0.1893, "step": 1554 }, { "epoch": 0.06, "learning_rate": 9.806051719541456e-06, "loss": 0.1662, "step": 1555 }, { "epoch": 0.06, "learning_rate": 9.8059184217542e-06, "loss": 0.1331, "step": 1556 }, { "epoch": 0.06, "learning_rate": 9.805785123966944e-06, "loss": 0.1287, "step": 1557 }, { "epoch": 0.06, "learning_rate": 9.805651826179687e-06, "loss": 0.1048, "step": 1558 }, { "epoch": 0.06, "learning_rate": 9.80551852839243e-06, "loss": 0.1321, "step": 1559 }, { "epoch": 0.06, "learning_rate": 9.805385230605172e-06, "loss": 0.1536, "step": 1560 }, { "epoch": 0.06, "learning_rate": 9.805251932817917e-06, "loss": 0.1155, "step": 1561 }, { "epoch": 0.06, "learning_rate": 9.80511863503066e-06, "loss": 0.1313, "step": 1562 }, { "epoch": 0.06, "learning_rate": 9.804985337243402e-06, "loss": 0.12, "step": 1563 }, { "epoch": 0.06, "learning_rate": 9.804852039456145e-06, "loss": 0.1154, "step": 1564 }, { "epoch": 0.06, "learning_rate": 9.80471874166889e-06, "loss": 0.1153, "step": 1565 }, { "epoch": 0.06, "learning_rate": 9.804585443881632e-06, "loss": 0.1384, "step": 1566 }, { "epoch": 0.06, "learning_rate": 9.804452146094375e-06, "loss": 0.1785, "step": 1567 }, { "epoch": 0.06, "learning_rate": 9.80431884830712e-06, "loss": 0.1313, "step": 1568 }, { "epoch": 0.06, "learning_rate": 9.804185550519863e-06, "loss": 0.1414, "step": 1569 }, { "epoch": 0.06, "learning_rate": 9.804052252732605e-06, "loss": 0.2052, "step": 1570 }, { "epoch": 0.06, "learning_rate": 9.803918954945348e-06, "loss": 0.1409, "step": 1571 }, { "epoch": 0.06, "learning_rate": 9.803785657158093e-06, "loss": 0.1208, "step": 1572 }, { "epoch": 0.06, "learning_rate": 9.803652359370836e-06, "loss": 0.1406, "step": 1573 }, { "epoch": 0.06, "learning_rate": 9.803519061583578e-06, "loss": 0.1629, "step": 1574 }, { "epoch": 0.06, "learning_rate": 9.803385763796321e-06, "loss": 0.1563, "step": 1575 }, { "epoch": 0.06, "learning_rate": 9.803252466009066e-06, "loss": 0.1592, "step": 1576 }, { "epoch": 0.06, "learning_rate": 9.803119168221809e-06, "loss": 0.0897, "step": 1577 }, { "epoch": 0.06, "learning_rate": 9.802985870434551e-06, "loss": 0.1139, "step": 1578 }, { "epoch": 0.06, "learning_rate": 9.802852572647294e-06, "loss": 0.1311, "step": 1579 }, { "epoch": 0.06, "learning_rate": 9.802719274860039e-06, "loss": 0.1239, "step": 1580 }, { "epoch": 0.06, "learning_rate": 9.802585977072782e-06, "loss": 0.1483, "step": 1581 }, { "epoch": 0.06, "learning_rate": 9.802452679285524e-06, "loss": 0.1473, "step": 1582 }, { "epoch": 0.06, "learning_rate": 9.802319381498267e-06, "loss": 0.1188, "step": 1583 }, { "epoch": 0.06, "learning_rate": 9.802186083711012e-06, "loss": 0.1188, "step": 1584 }, { "epoch": 0.06, "learning_rate": 9.802052785923755e-06, "loss": 0.152, "step": 1585 }, { "epoch": 0.06, "learning_rate": 9.801919488136497e-06, "loss": 0.1287, "step": 1586 }, { "epoch": 0.06, "learning_rate": 9.80178619034924e-06, "loss": 0.1736, "step": 1587 }, { "epoch": 0.06, "learning_rate": 9.801652892561985e-06, "loss": 0.1341, "step": 1588 }, { "epoch": 0.06, "learning_rate": 9.801519594774727e-06, "loss": 0.1133, "step": 1589 }, { "epoch": 0.06, "learning_rate": 9.80138629698747e-06, "loss": 0.1328, "step": 1590 }, { "epoch": 0.06, "learning_rate": 9.801252999200213e-06, "loss": 0.1319, "step": 1591 }, { "epoch": 0.06, "learning_rate": 9.801119701412958e-06, "loss": 0.1637, "step": 1592 }, { "epoch": 0.06, "learning_rate": 9.8009864036257e-06, "loss": 0.1449, "step": 1593 }, { "epoch": 0.06, "learning_rate": 9.800853105838443e-06, "loss": 0.1446, "step": 1594 }, { "epoch": 0.06, "learning_rate": 9.800719808051186e-06, "loss": 0.0939, "step": 1595 }, { "epoch": 0.06, "learning_rate": 9.80058651026393e-06, "loss": 0.1674, "step": 1596 }, { "epoch": 0.06, "learning_rate": 9.800453212476673e-06, "loss": 0.1422, "step": 1597 }, { "epoch": 0.06, "learning_rate": 9.800319914689416e-06, "loss": 0.1544, "step": 1598 }, { "epoch": 0.06, "learning_rate": 9.800186616902159e-06, "loss": 0.1548, "step": 1599 }, { "epoch": 0.06, "learning_rate": 9.800053319114904e-06, "loss": 0.1698, "step": 1600 }, { "epoch": 0.06, "learning_rate": 9.799920021327646e-06, "loss": 0.1414, "step": 1601 }, { "epoch": 0.06, "learning_rate": 9.79978672354039e-06, "loss": 0.1149, "step": 1602 }, { "epoch": 0.06, "learning_rate": 9.799653425753132e-06, "loss": 0.1485, "step": 1603 }, { "epoch": 0.06, "learning_rate": 9.799520127965877e-06, "loss": 0.1259, "step": 1604 }, { "epoch": 0.06, "learning_rate": 9.79938683017862e-06, "loss": 0.1121, "step": 1605 }, { "epoch": 0.06, "learning_rate": 9.799253532391362e-06, "loss": 0.1419, "step": 1606 }, { "epoch": 0.06, "learning_rate": 9.799120234604107e-06, "loss": 0.147, "step": 1607 }, { "epoch": 0.06, "learning_rate": 9.79898693681685e-06, "loss": 0.0957, "step": 1608 }, { "epoch": 0.06, "learning_rate": 9.798853639029592e-06, "loss": 0.1392, "step": 1609 }, { "epoch": 0.06, "learning_rate": 9.798720341242335e-06, "loss": 0.1481, "step": 1610 }, { "epoch": 0.06, "learning_rate": 9.79858704345508e-06, "loss": 0.1757, "step": 1611 }, { "epoch": 0.06, "learning_rate": 9.798453745667822e-06, "loss": 0.1667, "step": 1612 }, { "epoch": 0.06, "learning_rate": 9.798320447880565e-06, "loss": 0.143, "step": 1613 }, { "epoch": 0.06, "learning_rate": 9.79818715009331e-06, "loss": 0.1292, "step": 1614 }, { "epoch": 0.06, "learning_rate": 9.798053852306053e-06, "loss": 0.1445, "step": 1615 }, { "epoch": 0.06, "learning_rate": 9.797920554518795e-06, "loss": 0.1322, "step": 1616 }, { "epoch": 0.06, "learning_rate": 9.797787256731538e-06, "loss": 0.0963, "step": 1617 }, { "epoch": 0.06, "learning_rate": 9.797653958944283e-06, "loss": 0.1444, "step": 1618 }, { "epoch": 0.06, "learning_rate": 9.797520661157026e-06, "loss": 0.1227, "step": 1619 }, { "epoch": 0.06, "learning_rate": 9.797387363369768e-06, "loss": 0.12, "step": 1620 }, { "epoch": 0.06, "learning_rate": 9.797254065582513e-06, "loss": 0.1356, "step": 1621 }, { "epoch": 0.06, "learning_rate": 9.797120767795256e-06, "loss": 0.1424, "step": 1622 }, { "epoch": 0.06, "learning_rate": 9.796987470007999e-06, "loss": 0.159, "step": 1623 }, { "epoch": 0.06, "learning_rate": 9.796854172220741e-06, "loss": 0.1148, "step": 1624 }, { "epoch": 0.06, "learning_rate": 9.796720874433486e-06, "loss": 0.1069, "step": 1625 }, { "epoch": 0.06, "learning_rate": 9.796587576646229e-06, "loss": 0.1265, "step": 1626 }, { "epoch": 0.06, "learning_rate": 9.796454278858972e-06, "loss": 0.1181, "step": 1627 }, { "epoch": 0.07, "learning_rate": 9.796320981071716e-06, "loss": 0.1216, "step": 1628 }, { "epoch": 0.07, "learning_rate": 9.796187683284459e-06, "loss": 0.1486, "step": 1629 }, { "epoch": 0.07, "learning_rate": 9.796054385497202e-06, "loss": 0.1136, "step": 1630 }, { "epoch": 0.07, "learning_rate": 9.795921087709945e-06, "loss": 0.1384, "step": 1631 }, { "epoch": 0.07, "learning_rate": 9.795787789922689e-06, "loss": 0.1474, "step": 1632 }, { "epoch": 0.07, "learning_rate": 9.795654492135432e-06, "loss": 0.1579, "step": 1633 }, { "epoch": 0.07, "learning_rate": 9.795521194348175e-06, "loss": 0.1364, "step": 1634 }, { "epoch": 0.07, "learning_rate": 9.79538789656092e-06, "loss": 0.1174, "step": 1635 }, { "epoch": 0.07, "learning_rate": 9.795254598773662e-06, "loss": 0.1488, "step": 1636 }, { "epoch": 0.07, "learning_rate": 9.795121300986405e-06, "loss": 0.1453, "step": 1637 }, { "epoch": 0.07, "learning_rate": 9.794988003199148e-06, "loss": 0.159, "step": 1638 }, { "epoch": 0.07, "learning_rate": 9.794854705411892e-06, "loss": 0.1027, "step": 1639 }, { "epoch": 0.07, "learning_rate": 9.794721407624635e-06, "loss": 0.1115, "step": 1640 }, { "epoch": 0.07, "learning_rate": 9.794588109837378e-06, "loss": 0.1476, "step": 1641 }, { "epoch": 0.07, "learning_rate": 9.79445481205012e-06, "loss": 0.1531, "step": 1642 }, { "epoch": 0.07, "learning_rate": 9.794321514262865e-06, "loss": 0.1294, "step": 1643 }, { "epoch": 0.07, "learning_rate": 9.794188216475608e-06, "loss": 0.1569, "step": 1644 }, { "epoch": 0.07, "learning_rate": 9.79405491868835e-06, "loss": 0.1246, "step": 1645 }, { "epoch": 0.07, "learning_rate": 9.793921620901094e-06, "loss": 0.1588, "step": 1646 }, { "epoch": 0.07, "learning_rate": 9.793788323113838e-06, "loss": 0.146, "step": 1647 }, { "epoch": 0.07, "learning_rate": 9.793655025326581e-06, "loss": 0.1474, "step": 1648 }, { "epoch": 0.07, "learning_rate": 9.793521727539324e-06, "loss": 0.1509, "step": 1649 }, { "epoch": 0.07, "learning_rate": 9.793388429752067e-06, "loss": 0.0878, "step": 1650 }, { "epoch": 0.07, "learning_rate": 9.793255131964811e-06, "loss": 0.1768, "step": 1651 }, { "epoch": 0.07, "learning_rate": 9.793121834177554e-06, "loss": 0.0965, "step": 1652 }, { "epoch": 0.07, "learning_rate": 9.792988536390297e-06, "loss": 0.1104, "step": 1653 }, { "epoch": 0.07, "learning_rate": 9.79285523860304e-06, "loss": 0.1612, "step": 1654 }, { "epoch": 0.07, "learning_rate": 9.792721940815784e-06, "loss": 0.1361, "step": 1655 }, { "epoch": 0.07, "learning_rate": 9.792588643028527e-06, "loss": 0.1401, "step": 1656 }, { "epoch": 0.07, "learning_rate": 9.79245534524127e-06, "loss": 0.1355, "step": 1657 }, { "epoch": 0.07, "learning_rate": 9.792322047454012e-06, "loss": 0.1558, "step": 1658 }, { "epoch": 0.07, "learning_rate": 9.792188749666757e-06, "loss": 0.1973, "step": 1659 }, { "epoch": 0.07, "learning_rate": 9.7920554518795e-06, "loss": 0.1393, "step": 1660 }, { "epoch": 0.07, "learning_rate": 9.791922154092243e-06, "loss": 0.1168, "step": 1661 }, { "epoch": 0.07, "learning_rate": 9.791788856304985e-06, "loss": 0.1036, "step": 1662 }, { "epoch": 0.07, "learning_rate": 9.79165555851773e-06, "loss": 0.164, "step": 1663 }, { "epoch": 0.07, "learning_rate": 9.791522260730473e-06, "loss": 0.1338, "step": 1664 }, { "epoch": 0.07, "learning_rate": 9.791388962943216e-06, "loss": 0.1511, "step": 1665 }, { "epoch": 0.07, "learning_rate": 9.791255665155958e-06, "loss": 0.1175, "step": 1666 }, { "epoch": 0.07, "learning_rate": 9.791122367368703e-06, "loss": 0.1763, "step": 1667 }, { "epoch": 0.07, "learning_rate": 9.790989069581446e-06, "loss": 0.1217, "step": 1668 }, { "epoch": 0.07, "learning_rate": 9.790855771794189e-06, "loss": 0.1055, "step": 1669 }, { "epoch": 0.07, "learning_rate": 9.790722474006931e-06, "loss": 0.1193, "step": 1670 }, { "epoch": 0.07, "learning_rate": 9.790589176219676e-06, "loss": 0.1046, "step": 1671 }, { "epoch": 0.07, "learning_rate": 9.790455878432419e-06, "loss": 0.1131, "step": 1672 }, { "epoch": 0.07, "learning_rate": 9.790322580645162e-06, "loss": 0.1414, "step": 1673 }, { "epoch": 0.07, "learning_rate": 9.790189282857904e-06, "loss": 0.1326, "step": 1674 }, { "epoch": 0.07, "learning_rate": 9.790055985070649e-06, "loss": 0.1659, "step": 1675 }, { "epoch": 0.07, "learning_rate": 9.789922687283392e-06, "loss": 0.0931, "step": 1676 }, { "epoch": 0.07, "learning_rate": 9.789789389496135e-06, "loss": 0.1283, "step": 1677 }, { "epoch": 0.07, "learning_rate": 9.789656091708879e-06, "loss": 0.1311, "step": 1678 }, { "epoch": 0.07, "learning_rate": 9.789522793921622e-06, "loss": 0.1428, "step": 1679 }, { "epoch": 0.07, "learning_rate": 9.789389496134365e-06, "loss": 0.1506, "step": 1680 }, { "epoch": 0.07, "learning_rate": 9.789256198347107e-06, "loss": 0.16, "step": 1681 }, { "epoch": 0.07, "learning_rate": 9.789122900559852e-06, "loss": 0.1282, "step": 1682 }, { "epoch": 0.07, "learning_rate": 9.788989602772595e-06, "loss": 0.0957, "step": 1683 }, { "epoch": 0.07, "learning_rate": 9.788856304985338e-06, "loss": 0.1457, "step": 1684 }, { "epoch": 0.07, "learning_rate": 9.78872300719808e-06, "loss": 0.1581, "step": 1685 }, { "epoch": 0.07, "learning_rate": 9.788589709410825e-06, "loss": 0.1442, "step": 1686 }, { "epoch": 0.07, "learning_rate": 9.788456411623568e-06, "loss": 0.1172, "step": 1687 }, { "epoch": 0.07, "learning_rate": 9.78832311383631e-06, "loss": 0.1446, "step": 1688 }, { "epoch": 0.07, "learning_rate": 9.788189816049053e-06, "loss": 0.0777, "step": 1689 }, { "epoch": 0.07, "learning_rate": 9.788056518261798e-06, "loss": 0.17, "step": 1690 }, { "epoch": 0.07, "learning_rate": 9.78792322047454e-06, "loss": 0.1509, "step": 1691 }, { "epoch": 0.07, "learning_rate": 9.787789922687284e-06, "loss": 0.108, "step": 1692 }, { "epoch": 0.07, "learning_rate": 9.787656624900026e-06, "loss": 0.1631, "step": 1693 }, { "epoch": 0.07, "learning_rate": 9.787523327112771e-06, "loss": 0.1112, "step": 1694 }, { "epoch": 0.07, "learning_rate": 9.787390029325514e-06, "loss": 0.136, "step": 1695 }, { "epoch": 0.07, "learning_rate": 9.787256731538257e-06, "loss": 0.1447, "step": 1696 }, { "epoch": 0.07, "learning_rate": 9.787123433751e-06, "loss": 0.1584, "step": 1697 }, { "epoch": 0.07, "learning_rate": 9.786990135963744e-06, "loss": 0.1452, "step": 1698 }, { "epoch": 0.07, "learning_rate": 9.786856838176487e-06, "loss": 0.1256, "step": 1699 }, { "epoch": 0.07, "learning_rate": 9.78672354038923e-06, "loss": 0.1198, "step": 1700 }, { "epoch": 0.07, "learning_rate": 9.786590242601974e-06, "loss": 0.1098, "step": 1701 }, { "epoch": 0.07, "learning_rate": 9.786456944814717e-06, "loss": 0.1539, "step": 1702 }, { "epoch": 0.07, "learning_rate": 9.78632364702746e-06, "loss": 0.138, "step": 1703 }, { "epoch": 0.07, "learning_rate": 9.786190349240202e-06, "loss": 0.138, "step": 1704 }, { "epoch": 0.07, "learning_rate": 9.786057051452947e-06, "loss": 0.1094, "step": 1705 }, { "epoch": 0.07, "learning_rate": 9.78592375366569e-06, "loss": 0.143, "step": 1706 }, { "epoch": 0.07, "learning_rate": 9.785790455878433e-06, "loss": 0.1189, "step": 1707 }, { "epoch": 0.07, "learning_rate": 9.785657158091177e-06, "loss": 0.1411, "step": 1708 }, { "epoch": 0.07, "learning_rate": 9.78552386030392e-06, "loss": 0.129, "step": 1709 }, { "epoch": 0.07, "learning_rate": 9.785390562516663e-06, "loss": 0.1183, "step": 1710 }, { "epoch": 0.07, "learning_rate": 9.785257264729406e-06, "loss": 0.1298, "step": 1711 }, { "epoch": 0.07, "learning_rate": 9.78512396694215e-06, "loss": 0.1565, "step": 1712 }, { "epoch": 0.07, "learning_rate": 9.784990669154893e-06, "loss": 0.1159, "step": 1713 }, { "epoch": 0.07, "learning_rate": 9.784857371367636e-06, "loss": 0.1205, "step": 1714 }, { "epoch": 0.07, "learning_rate": 9.78472407358038e-06, "loss": 0.1759, "step": 1715 }, { "epoch": 0.07, "learning_rate": 9.784590775793123e-06, "loss": 0.158, "step": 1716 }, { "epoch": 0.07, "learning_rate": 9.784457478005866e-06, "loss": 0.1755, "step": 1717 }, { "epoch": 0.07, "learning_rate": 9.784324180218609e-06, "loss": 0.1219, "step": 1718 }, { "epoch": 0.07, "learning_rate": 9.784190882431353e-06, "loss": 0.1663, "step": 1719 }, { "epoch": 0.07, "learning_rate": 9.784057584644096e-06, "loss": 0.1365, "step": 1720 }, { "epoch": 0.07, "learning_rate": 9.783924286856839e-06, "loss": 0.1323, "step": 1721 }, { "epoch": 0.07, "learning_rate": 9.783790989069583e-06, "loss": 0.1096, "step": 1722 }, { "epoch": 0.07, "learning_rate": 9.783657691282326e-06, "loss": 0.1769, "step": 1723 }, { "epoch": 0.07, "learning_rate": 9.783524393495069e-06, "loss": 0.1117, "step": 1724 }, { "epoch": 0.07, "learning_rate": 9.783391095707812e-06, "loss": 0.1006, "step": 1725 }, { "epoch": 0.07, "learning_rate": 9.783257797920556e-06, "loss": 0.1517, "step": 1726 }, { "epoch": 0.07, "learning_rate": 9.7831245001333e-06, "loss": 0.1064, "step": 1727 }, { "epoch": 0.07, "learning_rate": 9.782991202346042e-06, "loss": 0.1322, "step": 1728 }, { "epoch": 0.07, "learning_rate": 9.782857904558785e-06, "loss": 0.0953, "step": 1729 }, { "epoch": 0.07, "learning_rate": 9.78272460677153e-06, "loss": 0.1349, "step": 1730 }, { "epoch": 0.07, "learning_rate": 9.782591308984272e-06, "loss": 0.1127, "step": 1731 }, { "epoch": 0.07, "learning_rate": 9.782458011197015e-06, "loss": 0.1106, "step": 1732 }, { "epoch": 0.07, "learning_rate": 9.782324713409758e-06, "loss": 0.1498, "step": 1733 }, { "epoch": 0.07, "learning_rate": 9.782191415622502e-06, "loss": 0.1063, "step": 1734 }, { "epoch": 0.07, "learning_rate": 9.782058117835245e-06, "loss": 0.1406, "step": 1735 }, { "epoch": 0.07, "learning_rate": 9.781924820047988e-06, "loss": 0.1521, "step": 1736 }, { "epoch": 0.07, "learning_rate": 9.781791522260732e-06, "loss": 0.1195, "step": 1737 }, { "epoch": 0.07, "learning_rate": 9.781658224473475e-06, "loss": 0.1207, "step": 1738 }, { "epoch": 0.07, "learning_rate": 9.781524926686218e-06, "loss": 0.1391, "step": 1739 }, { "epoch": 0.07, "learning_rate": 9.781391628898961e-06, "loss": 0.1398, "step": 1740 }, { "epoch": 0.07, "learning_rate": 9.781258331111705e-06, "loss": 0.1228, "step": 1741 }, { "epoch": 0.07, "learning_rate": 9.781125033324448e-06, "loss": 0.1363, "step": 1742 }, { "epoch": 0.07, "learning_rate": 9.780991735537191e-06, "loss": 0.1378, "step": 1743 }, { "epoch": 0.07, "learning_rate": 9.780858437749934e-06, "loss": 0.1119, "step": 1744 }, { "epoch": 0.07, "learning_rate": 9.780725139962678e-06, "loss": 0.155, "step": 1745 }, { "epoch": 0.07, "learning_rate": 9.780591842175421e-06, "loss": 0.1424, "step": 1746 }, { "epoch": 0.07, "learning_rate": 9.780458544388164e-06, "loss": 0.1536, "step": 1747 }, { "epoch": 0.07, "learning_rate": 9.780325246600907e-06, "loss": 0.0943, "step": 1748 }, { "epoch": 0.07, "learning_rate": 9.780191948813651e-06, "loss": 0.1506, "step": 1749 }, { "epoch": 0.07, "learning_rate": 9.780058651026394e-06, "loss": 0.1132, "step": 1750 }, { "epoch": 0.07, "learning_rate": 9.779925353239137e-06, "loss": 0.1422, "step": 1751 }, { "epoch": 0.07, "learning_rate": 9.77979205545188e-06, "loss": 0.1338, "step": 1752 }, { "epoch": 0.07, "learning_rate": 9.779658757664624e-06, "loss": 0.1814, "step": 1753 }, { "epoch": 0.07, "learning_rate": 9.779525459877367e-06, "loss": 0.12, "step": 1754 }, { "epoch": 0.07, "learning_rate": 9.77939216209011e-06, "loss": 0.1689, "step": 1755 }, { "epoch": 0.07, "learning_rate": 9.779258864302853e-06, "loss": 0.1059, "step": 1756 }, { "epoch": 0.07, "learning_rate": 9.779125566515597e-06, "loss": 0.1362, "step": 1757 }, { "epoch": 0.07, "learning_rate": 9.77899226872834e-06, "loss": 0.1212, "step": 1758 }, { "epoch": 0.07, "learning_rate": 9.778858970941083e-06, "loss": 0.1392, "step": 1759 }, { "epoch": 0.07, "learning_rate": 9.778725673153826e-06, "loss": 0.1904, "step": 1760 }, { "epoch": 0.07, "learning_rate": 9.77859237536657e-06, "loss": 0.1618, "step": 1761 }, { "epoch": 0.07, "learning_rate": 9.778459077579313e-06, "loss": 0.121, "step": 1762 }, { "epoch": 0.07, "learning_rate": 9.778325779792056e-06, "loss": 0.1183, "step": 1763 }, { "epoch": 0.07, "learning_rate": 9.778192482004799e-06, "loss": 0.138, "step": 1764 }, { "epoch": 0.07, "learning_rate": 9.778059184217543e-06, "loss": 0.1561, "step": 1765 }, { "epoch": 0.07, "learning_rate": 9.777925886430286e-06, "loss": 0.1445, "step": 1766 }, { "epoch": 0.07, "learning_rate": 9.777792588643029e-06, "loss": 0.1366, "step": 1767 }, { "epoch": 0.07, "learning_rate": 9.777659290855772e-06, "loss": 0.1508, "step": 1768 }, { "epoch": 0.07, "learning_rate": 9.777525993068516e-06, "loss": 0.1167, "step": 1769 }, { "epoch": 0.07, "learning_rate": 9.777392695281259e-06, "loss": 0.1271, "step": 1770 }, { "epoch": 0.07, "learning_rate": 9.777259397494002e-06, "loss": 0.139, "step": 1771 }, { "epoch": 0.07, "learning_rate": 9.777126099706745e-06, "loss": 0.1475, "step": 1772 }, { "epoch": 0.07, "learning_rate": 9.77699280191949e-06, "loss": 0.1098, "step": 1773 }, { "epoch": 0.07, "learning_rate": 9.776859504132232e-06, "loss": 0.1701, "step": 1774 }, { "epoch": 0.07, "learning_rate": 9.776726206344975e-06, "loss": 0.1366, "step": 1775 }, { "epoch": 0.07, "learning_rate": 9.776592908557718e-06, "loss": 0.1266, "step": 1776 }, { "epoch": 0.07, "learning_rate": 9.776459610770462e-06, "loss": 0.1086, "step": 1777 }, { "epoch": 0.07, "learning_rate": 9.776326312983205e-06, "loss": 0.1463, "step": 1778 }, { "epoch": 0.07, "learning_rate": 9.776193015195948e-06, "loss": 0.1563, "step": 1779 }, { "epoch": 0.07, "learning_rate": 9.77605971740869e-06, "loss": 0.1593, "step": 1780 }, { "epoch": 0.07, "learning_rate": 9.775926419621435e-06, "loss": 0.1155, "step": 1781 }, { "epoch": 0.07, "learning_rate": 9.775793121834178e-06, "loss": 0.1376, "step": 1782 }, { "epoch": 0.07, "learning_rate": 9.77565982404692e-06, "loss": 0.135, "step": 1783 }, { "epoch": 0.07, "learning_rate": 9.775526526259664e-06, "loss": 0.1665, "step": 1784 }, { "epoch": 0.07, "learning_rate": 9.775393228472408e-06, "loss": 0.111, "step": 1785 }, { "epoch": 0.07, "learning_rate": 9.775259930685151e-06, "loss": 0.1414, "step": 1786 }, { "epoch": 0.07, "learning_rate": 9.775126632897894e-06, "loss": 0.1146, "step": 1787 }, { "epoch": 0.07, "learning_rate": 9.774993335110638e-06, "loss": 0.1146, "step": 1788 }, { "epoch": 0.07, "learning_rate": 9.774860037323381e-06, "loss": 0.1142, "step": 1789 }, { "epoch": 0.07, "learning_rate": 9.774726739536124e-06, "loss": 0.1382, "step": 1790 }, { "epoch": 0.07, "learning_rate": 9.774593441748867e-06, "loss": 0.1291, "step": 1791 }, { "epoch": 0.07, "learning_rate": 9.774460143961611e-06, "loss": 0.1205, "step": 1792 }, { "epoch": 0.07, "learning_rate": 9.774326846174354e-06, "loss": 0.1495, "step": 1793 }, { "epoch": 0.07, "learning_rate": 9.774193548387097e-06, "loss": 0.1579, "step": 1794 }, { "epoch": 0.07, "learning_rate": 9.774060250599841e-06, "loss": 0.1323, "step": 1795 }, { "epoch": 0.07, "learning_rate": 9.773926952812584e-06, "loss": 0.1395, "step": 1796 }, { "epoch": 0.07, "learning_rate": 9.773793655025327e-06, "loss": 0.1179, "step": 1797 }, { "epoch": 0.07, "learning_rate": 9.77366035723807e-06, "loss": 0.1135, "step": 1798 }, { "epoch": 0.07, "learning_rate": 9.773527059450814e-06, "loss": 0.2038, "step": 1799 }, { "epoch": 0.07, "learning_rate": 9.773393761663557e-06, "loss": 0.1902, "step": 1800 }, { "epoch": 0.07, "learning_rate": 9.7732604638763e-06, "loss": 0.119, "step": 1801 }, { "epoch": 0.07, "learning_rate": 9.773127166089044e-06, "loss": 0.1534, "step": 1802 }, { "epoch": 0.07, "learning_rate": 9.772993868301787e-06, "loss": 0.1078, "step": 1803 }, { "epoch": 0.07, "learning_rate": 9.77286057051453e-06, "loss": 0.0912, "step": 1804 }, { "epoch": 0.07, "learning_rate": 9.772727272727273e-06, "loss": 0.1102, "step": 1805 }, { "epoch": 0.07, "learning_rate": 9.772593974940017e-06, "loss": 0.1384, "step": 1806 }, { "epoch": 0.07, "learning_rate": 9.77246067715276e-06, "loss": 0.1403, "step": 1807 }, { "epoch": 0.07, "learning_rate": 9.772327379365503e-06, "loss": 0.1266, "step": 1808 }, { "epoch": 0.07, "learning_rate": 9.772194081578248e-06, "loss": 0.1436, "step": 1809 }, { "epoch": 0.07, "learning_rate": 9.77206078379099e-06, "loss": 0.167, "step": 1810 }, { "epoch": 0.07, "learning_rate": 9.771927486003733e-06, "loss": 0.1505, "step": 1811 }, { "epoch": 0.07, "learning_rate": 9.771794188216476e-06, "loss": 0.1627, "step": 1812 }, { "epoch": 0.07, "learning_rate": 9.77166089042922e-06, "loss": 0.1118, "step": 1813 }, { "epoch": 0.07, "learning_rate": 9.771527592641963e-06, "loss": 0.1544, "step": 1814 }, { "epoch": 0.07, "learning_rate": 9.771394294854706e-06, "loss": 0.1682, "step": 1815 }, { "epoch": 0.07, "learning_rate": 9.77126099706745e-06, "loss": 0.1495, "step": 1816 }, { "epoch": 0.07, "learning_rate": 9.771127699280194e-06, "loss": 0.1757, "step": 1817 }, { "epoch": 0.07, "learning_rate": 9.770994401492936e-06, "loss": 0.1483, "step": 1818 }, { "epoch": 0.07, "learning_rate": 9.77086110370568e-06, "loss": 0.107, "step": 1819 }, { "epoch": 0.07, "learning_rate": 9.770727805918424e-06, "loss": 0.1152, "step": 1820 }, { "epoch": 0.07, "learning_rate": 9.770594508131167e-06, "loss": 0.1475, "step": 1821 }, { "epoch": 0.07, "learning_rate": 9.77046121034391e-06, "loss": 0.1418, "step": 1822 }, { "epoch": 0.07, "learning_rate": 9.770327912556652e-06, "loss": 0.146, "step": 1823 }, { "epoch": 0.07, "learning_rate": 9.770194614769397e-06, "loss": 0.1206, "step": 1824 }, { "epoch": 0.07, "learning_rate": 9.77006131698214e-06, "loss": 0.1792, "step": 1825 }, { "epoch": 0.07, "learning_rate": 9.769928019194882e-06, "loss": 0.1259, "step": 1826 }, { "epoch": 0.07, "learning_rate": 9.769794721407625e-06, "loss": 0.1304, "step": 1827 }, { "epoch": 0.07, "learning_rate": 9.76966142362037e-06, "loss": 0.1385, "step": 1828 }, { "epoch": 0.07, "learning_rate": 9.769528125833112e-06, "loss": 0.1205, "step": 1829 }, { "epoch": 0.07, "learning_rate": 9.769394828045855e-06, "loss": 0.0905, "step": 1830 }, { "epoch": 0.07, "learning_rate": 9.769261530258598e-06, "loss": 0.137, "step": 1831 }, { "epoch": 0.07, "learning_rate": 9.769128232471343e-06, "loss": 0.1113, "step": 1832 }, { "epoch": 0.07, "learning_rate": 9.768994934684085e-06, "loss": 0.0988, "step": 1833 }, { "epoch": 0.07, "learning_rate": 9.768861636896828e-06, "loss": 0.1314, "step": 1834 }, { "epoch": 0.07, "learning_rate": 9.768728339109571e-06, "loss": 0.1535, "step": 1835 }, { "epoch": 0.07, "learning_rate": 9.768595041322316e-06, "loss": 0.1246, "step": 1836 }, { "epoch": 0.07, "learning_rate": 9.768461743535058e-06, "loss": 0.1274, "step": 1837 }, { "epoch": 0.07, "learning_rate": 9.768328445747801e-06, "loss": 0.1472, "step": 1838 }, { "epoch": 0.07, "learning_rate": 9.768195147960544e-06, "loss": 0.1248, "step": 1839 }, { "epoch": 0.07, "learning_rate": 9.768061850173289e-06, "loss": 0.1242, "step": 1840 }, { "epoch": 0.07, "learning_rate": 9.767928552386031e-06, "loss": 0.1134, "step": 1841 }, { "epoch": 0.07, "learning_rate": 9.767795254598774e-06, "loss": 0.1543, "step": 1842 }, { "epoch": 0.07, "learning_rate": 9.767661956811517e-06, "loss": 0.174, "step": 1843 }, { "epoch": 0.07, "learning_rate": 9.767528659024262e-06, "loss": 0.1264, "step": 1844 }, { "epoch": 0.07, "learning_rate": 9.767395361237004e-06, "loss": 0.1095, "step": 1845 }, { "epoch": 0.07, "learning_rate": 9.767262063449747e-06, "loss": 0.1232, "step": 1846 }, { "epoch": 0.07, "learning_rate": 9.767128765662492e-06, "loss": 0.0957, "step": 1847 }, { "epoch": 0.07, "learning_rate": 9.766995467875234e-06, "loss": 0.1329, "step": 1848 }, { "epoch": 0.07, "learning_rate": 9.766862170087977e-06, "loss": 0.0959, "step": 1849 }, { "epoch": 0.07, "learning_rate": 9.76672887230072e-06, "loss": 0.1291, "step": 1850 }, { "epoch": 0.07, "learning_rate": 9.766595574513465e-06, "loss": 0.12, "step": 1851 }, { "epoch": 0.07, "learning_rate": 9.766462276726207e-06, "loss": 0.1454, "step": 1852 }, { "epoch": 0.07, "learning_rate": 9.76632897893895e-06, "loss": 0.1606, "step": 1853 }, { "epoch": 0.07, "learning_rate": 9.766195681151693e-06, "loss": 0.1409, "step": 1854 }, { "epoch": 0.07, "learning_rate": 9.766062383364438e-06, "loss": 0.1063, "step": 1855 }, { "epoch": 0.07, "learning_rate": 9.76592908557718e-06, "loss": 0.1121, "step": 1856 }, { "epoch": 0.07, "learning_rate": 9.765795787789923e-06, "loss": 0.1548, "step": 1857 }, { "epoch": 0.07, "learning_rate": 9.765662490002666e-06, "loss": 0.1909, "step": 1858 }, { "epoch": 0.07, "learning_rate": 9.76552919221541e-06, "loss": 0.115, "step": 1859 }, { "epoch": 0.07, "learning_rate": 9.765395894428153e-06, "loss": 0.1573, "step": 1860 }, { "epoch": 0.07, "learning_rate": 9.765262596640896e-06, "loss": 0.1227, "step": 1861 }, { "epoch": 0.07, "learning_rate": 9.765129298853639e-06, "loss": 0.1872, "step": 1862 }, { "epoch": 0.07, "learning_rate": 9.764996001066384e-06, "loss": 0.1069, "step": 1863 }, { "epoch": 0.07, "learning_rate": 9.764862703279126e-06, "loss": 0.1331, "step": 1864 }, { "epoch": 0.07, "learning_rate": 9.76472940549187e-06, "loss": 0.1046, "step": 1865 }, { "epoch": 0.07, "learning_rate": 9.764596107704612e-06, "loss": 0.1954, "step": 1866 }, { "epoch": 0.07, "learning_rate": 9.764462809917357e-06, "loss": 0.103, "step": 1867 }, { "epoch": 0.07, "learning_rate": 9.7643295121301e-06, "loss": 0.1296, "step": 1868 }, { "epoch": 0.07, "learning_rate": 9.764196214342842e-06, "loss": 0.1528, "step": 1869 }, { "epoch": 0.07, "learning_rate": 9.764062916555585e-06, "loss": 0.1346, "step": 1870 }, { "epoch": 0.07, "learning_rate": 9.76392961876833e-06, "loss": 0.0964, "step": 1871 }, { "epoch": 0.07, "learning_rate": 9.763796320981072e-06, "loss": 0.1312, "step": 1872 }, { "epoch": 0.07, "learning_rate": 9.763663023193815e-06, "loss": 0.1098, "step": 1873 }, { "epoch": 0.07, "learning_rate": 9.763529725406558e-06, "loss": 0.1417, "step": 1874 }, { "epoch": 0.07, "learning_rate": 9.763396427619302e-06, "loss": 0.1186, "step": 1875 }, { "epoch": 0.07, "learning_rate": 9.763263129832045e-06, "loss": 0.1102, "step": 1876 }, { "epoch": 0.07, "learning_rate": 9.763129832044788e-06, "loss": 0.1238, "step": 1877 }, { "epoch": 0.07, "learning_rate": 9.762996534257531e-06, "loss": 0.1405, "step": 1878 }, { "epoch": 0.08, "learning_rate": 9.762863236470275e-06, "loss": 0.1778, "step": 1879 }, { "epoch": 0.08, "learning_rate": 9.762729938683018e-06, "loss": 0.1354, "step": 1880 }, { "epoch": 0.08, "learning_rate": 9.762596640895761e-06, "loss": 0.1317, "step": 1881 }, { "epoch": 0.08, "learning_rate": 9.762463343108506e-06, "loss": 0.1562, "step": 1882 }, { "epoch": 0.08, "learning_rate": 9.762330045321248e-06, "loss": 0.1528, "step": 1883 }, { "epoch": 0.08, "learning_rate": 9.762196747533991e-06, "loss": 0.1447, "step": 1884 }, { "epoch": 0.08, "learning_rate": 9.762063449746734e-06, "loss": 0.1243, "step": 1885 }, { "epoch": 0.08, "learning_rate": 9.761930151959479e-06, "loss": 0.1213, "step": 1886 }, { "epoch": 0.08, "learning_rate": 9.761796854172221e-06, "loss": 0.1456, "step": 1887 }, { "epoch": 0.08, "learning_rate": 9.761663556384964e-06, "loss": 0.1662, "step": 1888 }, { "epoch": 0.08, "learning_rate": 9.761530258597709e-06, "loss": 0.1223, "step": 1889 }, { "epoch": 0.08, "learning_rate": 9.761396960810452e-06, "loss": 0.1864, "step": 1890 }, { "epoch": 0.08, "learning_rate": 9.761263663023194e-06, "loss": 0.1207, "step": 1891 }, { "epoch": 0.08, "learning_rate": 9.761130365235937e-06, "loss": 0.1477, "step": 1892 }, { "epoch": 0.08, "learning_rate": 9.760997067448682e-06, "loss": 0.0921, "step": 1893 }, { "epoch": 0.08, "learning_rate": 9.760863769661424e-06, "loss": 0.1602, "step": 1894 }, { "epoch": 0.08, "learning_rate": 9.760730471874167e-06, "loss": 0.1454, "step": 1895 }, { "epoch": 0.08, "learning_rate": 9.760597174086912e-06, "loss": 0.0986, "step": 1896 }, { "epoch": 0.08, "learning_rate": 9.760463876299655e-06, "loss": 0.1073, "step": 1897 }, { "epoch": 0.08, "learning_rate": 9.760330578512397e-06, "loss": 0.1261, "step": 1898 }, { "epoch": 0.08, "learning_rate": 9.76019728072514e-06, "loss": 0.1233, "step": 1899 }, { "epoch": 0.08, "learning_rate": 9.760063982937885e-06, "loss": 0.123, "step": 1900 }, { "epoch": 0.08, "learning_rate": 9.759930685150628e-06, "loss": 0.1512, "step": 1901 }, { "epoch": 0.08, "learning_rate": 9.75979738736337e-06, "loss": 0.0907, "step": 1902 }, { "epoch": 0.08, "learning_rate": 9.759664089576115e-06, "loss": 0.1465, "step": 1903 }, { "epoch": 0.08, "learning_rate": 9.759530791788858e-06, "loss": 0.0993, "step": 1904 }, { "epoch": 0.08, "learning_rate": 9.7593974940016e-06, "loss": 0.1419, "step": 1905 }, { "epoch": 0.08, "learning_rate": 9.759264196214343e-06, "loss": 0.0921, "step": 1906 }, { "epoch": 0.08, "learning_rate": 9.759130898427088e-06, "loss": 0.1457, "step": 1907 }, { "epoch": 0.08, "learning_rate": 9.75899760063983e-06, "loss": 0.1238, "step": 1908 }, { "epoch": 0.08, "learning_rate": 9.758864302852574e-06, "loss": 0.1355, "step": 1909 }, { "epoch": 0.08, "learning_rate": 9.758731005065318e-06, "loss": 0.1516, "step": 1910 }, { "epoch": 0.08, "learning_rate": 9.758597707278061e-06, "loss": 0.1598, "step": 1911 }, { "epoch": 0.08, "learning_rate": 9.758464409490804e-06, "loss": 0.1406, "step": 1912 }, { "epoch": 0.08, "learning_rate": 9.758331111703547e-06, "loss": 0.1293, "step": 1913 }, { "epoch": 0.08, "learning_rate": 9.758197813916291e-06, "loss": 0.096, "step": 1914 }, { "epoch": 0.08, "learning_rate": 9.758064516129034e-06, "loss": 0.1782, "step": 1915 }, { "epoch": 0.08, "learning_rate": 9.757931218341777e-06, "loss": 0.1232, "step": 1916 }, { "epoch": 0.08, "learning_rate": 9.75779792055452e-06, "loss": 0.1581, "step": 1917 }, { "epoch": 0.08, "learning_rate": 9.757664622767264e-06, "loss": 0.1308, "step": 1918 }, { "epoch": 0.08, "learning_rate": 9.757531324980007e-06, "loss": 0.1211, "step": 1919 }, { "epoch": 0.08, "learning_rate": 9.75739802719275e-06, "loss": 0.1317, "step": 1920 }, { "epoch": 0.08, "learning_rate": 9.757264729405492e-06, "loss": 0.1216, "step": 1921 }, { "epoch": 0.08, "learning_rate": 9.757131431618237e-06, "loss": 0.1807, "step": 1922 }, { "epoch": 0.08, "learning_rate": 9.75699813383098e-06, "loss": 0.1097, "step": 1923 }, { "epoch": 0.08, "learning_rate": 9.756864836043723e-06, "loss": 0.147, "step": 1924 }, { "epoch": 0.08, "learning_rate": 9.756731538256465e-06, "loss": 0.1341, "step": 1925 }, { "epoch": 0.08, "learning_rate": 9.75659824046921e-06, "loss": 0.1203, "step": 1926 }, { "epoch": 0.08, "learning_rate": 9.756464942681953e-06, "loss": 0.1331, "step": 1927 }, { "epoch": 0.08, "learning_rate": 9.756331644894696e-06, "loss": 0.1524, "step": 1928 }, { "epoch": 0.08, "learning_rate": 9.756198347107438e-06, "loss": 0.1224, "step": 1929 }, { "epoch": 0.08, "learning_rate": 9.756065049320183e-06, "loss": 0.1262, "step": 1930 }, { "epoch": 0.08, "learning_rate": 9.755931751532926e-06, "loss": 0.1611, "step": 1931 }, { "epoch": 0.08, "learning_rate": 9.755798453745669e-06, "loss": 0.098, "step": 1932 }, { "epoch": 0.08, "learning_rate": 9.755665155958411e-06, "loss": 0.1169, "step": 1933 }, { "epoch": 0.08, "learning_rate": 9.755531858171156e-06, "loss": 0.1093, "step": 1934 }, { "epoch": 0.08, "learning_rate": 9.755398560383899e-06, "loss": 0.2282, "step": 1935 }, { "epoch": 0.08, "learning_rate": 9.755265262596642e-06, "loss": 0.1635, "step": 1936 }, { "epoch": 0.08, "learning_rate": 9.755131964809384e-06, "loss": 0.1107, "step": 1937 }, { "epoch": 0.08, "learning_rate": 9.754998667022129e-06, "loss": 0.1308, "step": 1938 }, { "epoch": 0.08, "learning_rate": 9.754865369234872e-06, "loss": 0.1069, "step": 1939 }, { "epoch": 0.08, "learning_rate": 9.754732071447615e-06, "loss": 0.1555, "step": 1940 }, { "epoch": 0.08, "learning_rate": 9.754598773660357e-06, "loss": 0.1802, "step": 1941 }, { "epoch": 0.08, "learning_rate": 9.754465475873102e-06, "loss": 0.1327, "step": 1942 }, { "epoch": 0.08, "learning_rate": 9.754332178085845e-06, "loss": 0.1564, "step": 1943 }, { "epoch": 0.08, "learning_rate": 9.754198880298587e-06, "loss": 0.1397, "step": 1944 }, { "epoch": 0.08, "learning_rate": 9.75406558251133e-06, "loss": 0.1214, "step": 1945 }, { "epoch": 0.08, "learning_rate": 9.753932284724075e-06, "loss": 0.1168, "step": 1946 }, { "epoch": 0.08, "learning_rate": 9.753798986936818e-06, "loss": 0.1174, "step": 1947 }, { "epoch": 0.08, "learning_rate": 9.75366568914956e-06, "loss": 0.102, "step": 1948 }, { "epoch": 0.08, "learning_rate": 9.753532391362303e-06, "loss": 0.1305, "step": 1949 }, { "epoch": 0.08, "learning_rate": 9.753399093575048e-06, "loss": 0.1607, "step": 1950 }, { "epoch": 0.08, "learning_rate": 9.75326579578779e-06, "loss": 0.1181, "step": 1951 }, { "epoch": 0.08, "learning_rate": 9.753132498000533e-06, "loss": 0.1243, "step": 1952 }, { "epoch": 0.08, "learning_rate": 9.752999200213276e-06, "loss": 0.1055, "step": 1953 }, { "epoch": 0.08, "learning_rate": 9.75286590242602e-06, "loss": 0.0914, "step": 1954 }, { "epoch": 0.08, "learning_rate": 9.752732604638764e-06, "loss": 0.1298, "step": 1955 }, { "epoch": 0.08, "learning_rate": 9.752599306851506e-06, "loss": 0.1078, "step": 1956 }, { "epoch": 0.08, "learning_rate": 9.75246600906425e-06, "loss": 0.1526, "step": 1957 }, { "epoch": 0.08, "learning_rate": 9.752332711276994e-06, "loss": 0.126, "step": 1958 }, { "epoch": 0.08, "learning_rate": 9.752199413489737e-06, "loss": 0.1312, "step": 1959 }, { "epoch": 0.08, "learning_rate": 9.75206611570248e-06, "loss": 0.1199, "step": 1960 }, { "epoch": 0.08, "learning_rate": 9.751932817915224e-06, "loss": 0.1334, "step": 1961 }, { "epoch": 0.08, "learning_rate": 9.751799520127967e-06, "loss": 0.1161, "step": 1962 }, { "epoch": 0.08, "learning_rate": 9.75166622234071e-06, "loss": 0.1195, "step": 1963 }, { "epoch": 0.08, "learning_rate": 9.751532924553452e-06, "loss": 0.1148, "step": 1964 }, { "epoch": 0.08, "learning_rate": 9.751399626766197e-06, "loss": 0.1487, "step": 1965 }, { "epoch": 0.08, "learning_rate": 9.75126632897894e-06, "loss": 0.0934, "step": 1966 }, { "epoch": 0.08, "learning_rate": 9.751133031191682e-06, "loss": 0.1215, "step": 1967 }, { "epoch": 0.08, "learning_rate": 9.750999733404425e-06, "loss": 0.1432, "step": 1968 }, { "epoch": 0.08, "learning_rate": 9.75086643561717e-06, "loss": 0.1575, "step": 1969 }, { "epoch": 0.08, "learning_rate": 9.750733137829913e-06, "loss": 0.1273, "step": 1970 }, { "epoch": 0.08, "learning_rate": 9.750599840042655e-06, "loss": 0.1115, "step": 1971 }, { "epoch": 0.08, "learning_rate": 9.750466542255398e-06, "loss": 0.1316, "step": 1972 }, { "epoch": 0.08, "learning_rate": 9.750333244468143e-06, "loss": 0.1476, "step": 1973 }, { "epoch": 0.08, "learning_rate": 9.750199946680886e-06, "loss": 0.1389, "step": 1974 }, { "epoch": 0.08, "learning_rate": 9.750066648893628e-06, "loss": 0.1126, "step": 1975 }, { "epoch": 0.08, "learning_rate": 9.749933351106371e-06, "loss": 0.1639, "step": 1976 }, { "epoch": 0.08, "learning_rate": 9.749800053319116e-06, "loss": 0.1397, "step": 1977 }, { "epoch": 0.08, "learning_rate": 9.749666755531859e-06, "loss": 0.1299, "step": 1978 }, { "epoch": 0.08, "learning_rate": 9.749533457744601e-06, "loss": 0.1223, "step": 1979 }, { "epoch": 0.08, "learning_rate": 9.749400159957346e-06, "loss": 0.1389, "step": 1980 }, { "epoch": 0.08, "learning_rate": 9.749266862170089e-06, "loss": 0.1702, "step": 1981 }, { "epoch": 0.08, "learning_rate": 9.749133564382832e-06, "loss": 0.118, "step": 1982 }, { "epoch": 0.08, "learning_rate": 9.749000266595574e-06, "loss": 0.152, "step": 1983 }, { "epoch": 0.08, "learning_rate": 9.748866968808319e-06, "loss": 0.145, "step": 1984 }, { "epoch": 0.08, "learning_rate": 9.748733671021062e-06, "loss": 0.1117, "step": 1985 }, { "epoch": 0.08, "learning_rate": 9.748600373233805e-06, "loss": 0.1094, "step": 1986 }, { "epoch": 0.08, "learning_rate": 9.748467075446549e-06, "loss": 0.1378, "step": 1987 }, { "epoch": 0.08, "learning_rate": 9.748333777659292e-06, "loss": 0.1125, "step": 1988 }, { "epoch": 0.08, "learning_rate": 9.748200479872035e-06, "loss": 0.1638, "step": 1989 }, { "epoch": 0.08, "learning_rate": 9.748067182084777e-06, "loss": 0.182, "step": 1990 }, { "epoch": 0.08, "learning_rate": 9.747933884297522e-06, "loss": 0.1632, "step": 1991 }, { "epoch": 0.08, "learning_rate": 9.747800586510265e-06, "loss": 0.1147, "step": 1992 }, { "epoch": 0.08, "learning_rate": 9.747667288723008e-06, "loss": 0.1459, "step": 1993 }, { "epoch": 0.08, "learning_rate": 9.747533990935752e-06, "loss": 0.1286, "step": 1994 }, { "epoch": 0.08, "learning_rate": 9.747400693148495e-06, "loss": 0.1447, "step": 1995 }, { "epoch": 0.08, "learning_rate": 9.747267395361238e-06, "loss": 0.1019, "step": 1996 }, { "epoch": 0.08, "learning_rate": 9.74713409757398e-06, "loss": 0.1079, "step": 1997 }, { "epoch": 0.08, "learning_rate": 9.747000799786725e-06, "loss": 0.1023, "step": 1998 }, { "epoch": 0.08, "learning_rate": 9.746867501999468e-06, "loss": 0.1252, "step": 1999 }, { "epoch": 0.08, "learning_rate": 9.74673420421221e-06, "loss": 0.1265, "step": 2000 }, { "epoch": 0.08, "learning_rate": 9.746600906424955e-06, "loss": 0.152, "step": 2001 }, { "epoch": 0.08, "learning_rate": 9.746467608637698e-06, "loss": 0.1737, "step": 2002 }, { "epoch": 0.08, "learning_rate": 9.746334310850441e-06, "loss": 0.1302, "step": 2003 }, { "epoch": 0.08, "learning_rate": 9.746201013063184e-06, "loss": 0.1398, "step": 2004 }, { "epoch": 0.08, "learning_rate": 9.746067715275928e-06, "loss": 0.142, "step": 2005 }, { "epoch": 0.08, "learning_rate": 9.745934417488671e-06, "loss": 0.1437, "step": 2006 }, { "epoch": 0.08, "learning_rate": 9.745801119701414e-06, "loss": 0.1283, "step": 2007 }, { "epoch": 0.08, "learning_rate": 9.745667821914157e-06, "loss": 0.1607, "step": 2008 }, { "epoch": 0.08, "learning_rate": 9.745534524126901e-06, "loss": 0.1137, "step": 2009 }, { "epoch": 0.08, "learning_rate": 9.745401226339644e-06, "loss": 0.1553, "step": 2010 }, { "epoch": 0.08, "learning_rate": 9.745267928552387e-06, "loss": 0.1159, "step": 2011 }, { "epoch": 0.08, "learning_rate": 9.74513463076513e-06, "loss": 0.092, "step": 2012 }, { "epoch": 0.08, "learning_rate": 9.745001332977874e-06, "loss": 0.1683, "step": 2013 }, { "epoch": 0.08, "learning_rate": 9.744868035190617e-06, "loss": 0.1111, "step": 2014 }, { "epoch": 0.08, "learning_rate": 9.74473473740336e-06, "loss": 0.1324, "step": 2015 }, { "epoch": 0.08, "learning_rate": 9.744601439616104e-06, "loss": 0.109, "step": 2016 }, { "epoch": 0.08, "learning_rate": 9.744468141828847e-06, "loss": 0.1253, "step": 2017 }, { "epoch": 0.08, "learning_rate": 9.74433484404159e-06, "loss": 0.111, "step": 2018 }, { "epoch": 0.08, "learning_rate": 9.744201546254333e-06, "loss": 0.1493, "step": 2019 }, { "epoch": 0.08, "learning_rate": 9.744068248467077e-06, "loss": 0.1208, "step": 2020 }, { "epoch": 0.08, "learning_rate": 9.74393495067982e-06, "loss": 0.1318, "step": 2021 }, { "epoch": 0.08, "learning_rate": 9.743801652892563e-06, "loss": 0.1168, "step": 2022 }, { "epoch": 0.08, "learning_rate": 9.743668355105306e-06, "loss": 0.0886, "step": 2023 }, { "epoch": 0.08, "learning_rate": 9.74353505731805e-06, "loss": 0.1234, "step": 2024 }, { "epoch": 0.08, "learning_rate": 9.743401759530793e-06, "loss": 0.1899, "step": 2025 }, { "epoch": 0.08, "learning_rate": 9.743268461743536e-06, "loss": 0.1265, "step": 2026 }, { "epoch": 0.08, "learning_rate": 9.743135163956279e-06, "loss": 0.1014, "step": 2027 }, { "epoch": 0.08, "learning_rate": 9.743001866169023e-06, "loss": 0.1042, "step": 2028 }, { "epoch": 0.08, "learning_rate": 9.742868568381766e-06, "loss": 0.1498, "step": 2029 }, { "epoch": 0.08, "learning_rate": 9.742735270594509e-06, "loss": 0.1261, "step": 2030 }, { "epoch": 0.08, "learning_rate": 9.742601972807252e-06, "loss": 0.1411, "step": 2031 }, { "epoch": 0.08, "learning_rate": 9.742468675019996e-06, "loss": 0.151, "step": 2032 }, { "epoch": 0.08, "learning_rate": 9.742335377232739e-06, "loss": 0.1392, "step": 2033 }, { "epoch": 0.08, "learning_rate": 9.742202079445482e-06, "loss": 0.1163, "step": 2034 }, { "epoch": 0.08, "learning_rate": 9.742068781658225e-06, "loss": 0.1659, "step": 2035 }, { "epoch": 0.08, "learning_rate": 9.74193548387097e-06, "loss": 0.2141, "step": 2036 }, { "epoch": 0.08, "learning_rate": 9.741802186083712e-06, "loss": 0.1776, "step": 2037 }, { "epoch": 0.08, "learning_rate": 9.741668888296455e-06, "loss": 0.1016, "step": 2038 }, { "epoch": 0.08, "learning_rate": 9.741535590509198e-06, "loss": 0.0897, "step": 2039 }, { "epoch": 0.08, "learning_rate": 9.741402292721942e-06, "loss": 0.1122, "step": 2040 }, { "epoch": 0.08, "learning_rate": 9.741268994934685e-06, "loss": 0.1201, "step": 2041 }, { "epoch": 0.08, "learning_rate": 9.741135697147428e-06, "loss": 0.1289, "step": 2042 }, { "epoch": 0.08, "learning_rate": 9.74100239936017e-06, "loss": 0.1059, "step": 2043 }, { "epoch": 0.08, "learning_rate": 9.740869101572915e-06, "loss": 0.0845, "step": 2044 }, { "epoch": 0.08, "learning_rate": 9.740735803785658e-06, "loss": 0.1439, "step": 2045 }, { "epoch": 0.08, "learning_rate": 9.7406025059984e-06, "loss": 0.136, "step": 2046 }, { "epoch": 0.08, "learning_rate": 9.740469208211144e-06, "loss": 0.1126, "step": 2047 }, { "epoch": 0.08, "learning_rate": 9.740335910423888e-06, "loss": 0.1516, "step": 2048 }, { "epoch": 0.08, "learning_rate": 9.740202612636631e-06, "loss": 0.1301, "step": 2049 }, { "epoch": 0.08, "learning_rate": 9.740069314849374e-06, "loss": 0.1605, "step": 2050 }, { "epoch": 0.08, "learning_rate": 9.739936017062117e-06, "loss": 0.1062, "step": 2051 }, { "epoch": 0.08, "learning_rate": 9.739802719274861e-06, "loss": 0.1162, "step": 2052 }, { "epoch": 0.08, "learning_rate": 9.739669421487604e-06, "loss": 0.1009, "step": 2053 }, { "epoch": 0.08, "learning_rate": 9.739536123700347e-06, "loss": 0.1483, "step": 2054 }, { "epoch": 0.08, "learning_rate": 9.73940282591309e-06, "loss": 0.0846, "step": 2055 }, { "epoch": 0.08, "learning_rate": 9.739269528125834e-06, "loss": 0.1309, "step": 2056 }, { "epoch": 0.08, "learning_rate": 9.739136230338577e-06, "loss": 0.0959, "step": 2057 }, { "epoch": 0.08, "learning_rate": 9.73900293255132e-06, "loss": 0.1103, "step": 2058 }, { "epoch": 0.08, "learning_rate": 9.738869634764062e-06, "loss": 0.1448, "step": 2059 }, { "epoch": 0.08, "learning_rate": 9.738736336976807e-06, "loss": 0.0846, "step": 2060 }, { "epoch": 0.08, "learning_rate": 9.73860303918955e-06, "loss": 0.1239, "step": 2061 }, { "epoch": 0.08, "learning_rate": 9.738469741402293e-06, "loss": 0.1405, "step": 2062 }, { "epoch": 0.08, "learning_rate": 9.738336443615035e-06, "loss": 0.1084, "step": 2063 }, { "epoch": 0.08, "learning_rate": 9.73820314582778e-06, "loss": 0.1286, "step": 2064 }, { "epoch": 0.08, "learning_rate": 9.738069848040523e-06, "loss": 0.1372, "step": 2065 }, { "epoch": 0.08, "learning_rate": 9.737936550253266e-06, "loss": 0.0977, "step": 2066 }, { "epoch": 0.08, "learning_rate": 9.73780325246601e-06, "loss": 0.1504, "step": 2067 }, { "epoch": 0.08, "learning_rate": 9.737669954678753e-06, "loss": 0.0934, "step": 2068 }, { "epoch": 0.08, "learning_rate": 9.737536656891496e-06, "loss": 0.1323, "step": 2069 }, { "epoch": 0.08, "learning_rate": 9.737403359104239e-06, "loss": 0.1554, "step": 2070 }, { "epoch": 0.08, "learning_rate": 9.737270061316983e-06, "loss": 0.1157, "step": 2071 }, { "epoch": 0.08, "learning_rate": 9.737136763529726e-06, "loss": 0.1569, "step": 2072 }, { "epoch": 0.08, "learning_rate": 9.737003465742469e-06, "loss": 0.1217, "step": 2073 }, { "epoch": 0.08, "learning_rate": 9.736870167955213e-06, "loss": 0.1317, "step": 2074 }, { "epoch": 0.08, "learning_rate": 9.736736870167956e-06, "loss": 0.126, "step": 2075 }, { "epoch": 0.08, "learning_rate": 9.736603572380699e-06, "loss": 0.1192, "step": 2076 }, { "epoch": 0.08, "learning_rate": 9.736470274593442e-06, "loss": 0.1483, "step": 2077 }, { "epoch": 0.08, "learning_rate": 9.736336976806186e-06, "loss": 0.1427, "step": 2078 }, { "epoch": 0.08, "learning_rate": 9.736203679018929e-06, "loss": 0.1232, "step": 2079 }, { "epoch": 0.08, "learning_rate": 9.736070381231672e-06, "loss": 0.1311, "step": 2080 }, { "epoch": 0.08, "learning_rate": 9.735937083444416e-06, "loss": 0.1222, "step": 2081 }, { "epoch": 0.08, "learning_rate": 9.73580378565716e-06, "loss": 0.1212, "step": 2082 }, { "epoch": 0.08, "learning_rate": 9.735670487869902e-06, "loss": 0.1038, "step": 2083 }, { "epoch": 0.08, "learning_rate": 9.735537190082645e-06, "loss": 0.1458, "step": 2084 }, { "epoch": 0.08, "learning_rate": 9.73540389229539e-06, "loss": 0.1092, "step": 2085 }, { "epoch": 0.08, "learning_rate": 9.735270594508132e-06, "loss": 0.1262, "step": 2086 }, { "epoch": 0.08, "learning_rate": 9.735137296720875e-06, "loss": 0.1153, "step": 2087 }, { "epoch": 0.08, "learning_rate": 9.73500399893362e-06, "loss": 0.1174, "step": 2088 }, { "epoch": 0.08, "learning_rate": 9.734870701146362e-06, "loss": 0.1031, "step": 2089 }, { "epoch": 0.08, "learning_rate": 9.734737403359105e-06, "loss": 0.0967, "step": 2090 }, { "epoch": 0.08, "learning_rate": 9.734604105571848e-06, "loss": 0.1313, "step": 2091 }, { "epoch": 0.08, "learning_rate": 9.734470807784592e-06, "loss": 0.1373, "step": 2092 }, { "epoch": 0.08, "learning_rate": 9.734337509997335e-06, "loss": 0.1115, "step": 2093 }, { "epoch": 0.08, "learning_rate": 9.734204212210078e-06, "loss": 0.1507, "step": 2094 }, { "epoch": 0.08, "learning_rate": 9.734070914422823e-06, "loss": 0.1505, "step": 2095 }, { "epoch": 0.08, "learning_rate": 9.733937616635565e-06, "loss": 0.1243, "step": 2096 }, { "epoch": 0.08, "learning_rate": 9.733804318848308e-06, "loss": 0.1649, "step": 2097 }, { "epoch": 0.08, "learning_rate": 9.733671021061051e-06, "loss": 0.124, "step": 2098 }, { "epoch": 0.08, "learning_rate": 9.733537723273796e-06, "loss": 0.125, "step": 2099 }, { "epoch": 0.08, "learning_rate": 9.733404425486538e-06, "loss": 0.138, "step": 2100 }, { "epoch": 0.08, "learning_rate": 9.733271127699281e-06, "loss": 0.1014, "step": 2101 }, { "epoch": 0.08, "learning_rate": 9.733137829912024e-06, "loss": 0.1536, "step": 2102 }, { "epoch": 0.08, "learning_rate": 9.733004532124769e-06, "loss": 0.1865, "step": 2103 }, { "epoch": 0.08, "learning_rate": 9.732871234337511e-06, "loss": 0.1425, "step": 2104 }, { "epoch": 0.08, "learning_rate": 9.732737936550254e-06, "loss": 0.1651, "step": 2105 }, { "epoch": 0.08, "learning_rate": 9.732604638762997e-06, "loss": 0.1046, "step": 2106 }, { "epoch": 0.08, "learning_rate": 9.732471340975742e-06, "loss": 0.1701, "step": 2107 }, { "epoch": 0.08, "learning_rate": 9.732338043188484e-06, "loss": 0.1356, "step": 2108 }, { "epoch": 0.08, "learning_rate": 9.732204745401227e-06, "loss": 0.1451, "step": 2109 }, { "epoch": 0.08, "learning_rate": 9.73207144761397e-06, "loss": 0.1216, "step": 2110 }, { "epoch": 0.08, "learning_rate": 9.731938149826714e-06, "loss": 0.1562, "step": 2111 }, { "epoch": 0.08, "learning_rate": 9.731804852039457e-06, "loss": 0.1224, "step": 2112 }, { "epoch": 0.08, "learning_rate": 9.7316715542522e-06, "loss": 0.0921, "step": 2113 }, { "epoch": 0.08, "learning_rate": 9.731538256464943e-06, "loss": 0.1142, "step": 2114 }, { "epoch": 0.08, "learning_rate": 9.731404958677687e-06, "loss": 0.1319, "step": 2115 }, { "epoch": 0.08, "learning_rate": 9.73127166089043e-06, "loss": 0.1533, "step": 2116 }, { "epoch": 0.08, "learning_rate": 9.731138363103173e-06, "loss": 0.1773, "step": 2117 }, { "epoch": 0.08, "learning_rate": 9.731005065315916e-06, "loss": 0.1268, "step": 2118 }, { "epoch": 0.08, "learning_rate": 9.73087176752866e-06, "loss": 0.1233, "step": 2119 }, { "epoch": 0.08, "learning_rate": 9.730738469741403e-06, "loss": 0.1082, "step": 2120 }, { "epoch": 0.08, "learning_rate": 9.730605171954146e-06, "loss": 0.1402, "step": 2121 }, { "epoch": 0.08, "learning_rate": 9.730471874166889e-06, "loss": 0.1084, "step": 2122 }, { "epoch": 0.08, "learning_rate": 9.730338576379633e-06, "loss": 0.1352, "step": 2123 }, { "epoch": 0.08, "learning_rate": 9.730205278592376e-06, "loss": 0.1464, "step": 2124 }, { "epoch": 0.08, "learning_rate": 9.730071980805119e-06, "loss": 0.1202, "step": 2125 }, { "epoch": 0.08, "learning_rate": 9.729938683017862e-06, "loss": 0.1292, "step": 2126 }, { "epoch": 0.08, "learning_rate": 9.729805385230606e-06, "loss": 0.1113, "step": 2127 }, { "epoch": 0.08, "learning_rate": 9.72967208744335e-06, "loss": 0.1326, "step": 2128 }, { "epoch": 0.09, "learning_rate": 9.729538789656092e-06, "loss": 0.0887, "step": 2129 }, { "epoch": 0.09, "learning_rate": 9.729405491868837e-06, "loss": 0.145, "step": 2130 }, { "epoch": 0.09, "learning_rate": 9.72927219408158e-06, "loss": 0.1211, "step": 2131 }, { "epoch": 0.09, "learning_rate": 9.729138896294322e-06, "loss": 0.1509, "step": 2132 }, { "epoch": 0.09, "learning_rate": 9.729005598507065e-06, "loss": 0.133, "step": 2133 }, { "epoch": 0.09, "learning_rate": 9.72887230071981e-06, "loss": 0.0972, "step": 2134 }, { "epoch": 0.09, "learning_rate": 9.728739002932552e-06, "loss": 0.1394, "step": 2135 }, { "epoch": 0.09, "learning_rate": 9.728605705145295e-06, "loss": 0.1622, "step": 2136 }, { "epoch": 0.09, "learning_rate": 9.728472407358038e-06, "loss": 0.1159, "step": 2137 }, { "epoch": 0.09, "learning_rate": 9.728339109570782e-06, "loss": 0.1404, "step": 2138 }, { "epoch": 0.09, "learning_rate": 9.728205811783525e-06, "loss": 0.0883, "step": 2139 }, { "epoch": 0.09, "learning_rate": 9.728072513996268e-06, "loss": 0.1164, "step": 2140 }, { "epoch": 0.09, "learning_rate": 9.727939216209011e-06, "loss": 0.12, "step": 2141 }, { "epoch": 0.09, "learning_rate": 9.727805918421755e-06, "loss": 0.1356, "step": 2142 }, { "epoch": 0.09, "learning_rate": 9.727672620634498e-06, "loss": 0.1226, "step": 2143 }, { "epoch": 0.09, "learning_rate": 9.727539322847241e-06, "loss": 0.2081, "step": 2144 }, { "epoch": 0.09, "learning_rate": 9.727406025059984e-06, "loss": 0.1468, "step": 2145 }, { "epoch": 0.09, "learning_rate": 9.727272727272728e-06, "loss": 0.1634, "step": 2146 }, { "epoch": 0.09, "learning_rate": 9.727139429485471e-06, "loss": 0.135, "step": 2147 }, { "epoch": 0.09, "learning_rate": 9.727006131698214e-06, "loss": 0.1445, "step": 2148 }, { "epoch": 0.09, "learning_rate": 9.726872833910957e-06, "loss": 0.1411, "step": 2149 }, { "epoch": 0.09, "learning_rate": 9.726739536123701e-06, "loss": 0.1244, "step": 2150 }, { "epoch": 0.09, "learning_rate": 9.726606238336444e-06, "loss": 0.1516, "step": 2151 }, { "epoch": 0.09, "learning_rate": 9.726472940549187e-06, "loss": 0.137, "step": 2152 }, { "epoch": 0.09, "learning_rate": 9.72633964276193e-06, "loss": 0.0863, "step": 2153 }, { "epoch": 0.09, "learning_rate": 9.726206344974674e-06, "loss": 0.0895, "step": 2154 }, { "epoch": 0.09, "learning_rate": 9.726073047187417e-06, "loss": 0.135, "step": 2155 }, { "epoch": 0.09, "learning_rate": 9.72593974940016e-06, "loss": 0.1499, "step": 2156 }, { "epoch": 0.09, "learning_rate": 9.725806451612903e-06, "loss": 0.125, "step": 2157 }, { "epoch": 0.09, "learning_rate": 9.725673153825647e-06, "loss": 0.1397, "step": 2158 }, { "epoch": 0.09, "learning_rate": 9.72553985603839e-06, "loss": 0.1215, "step": 2159 }, { "epoch": 0.09, "learning_rate": 9.725406558251133e-06, "loss": 0.117, "step": 2160 }, { "epoch": 0.09, "learning_rate": 9.725273260463877e-06, "loss": 0.1159, "step": 2161 }, { "epoch": 0.09, "learning_rate": 9.72513996267662e-06, "loss": 0.1108, "step": 2162 }, { "epoch": 0.09, "learning_rate": 9.725006664889363e-06, "loss": 0.1352, "step": 2163 }, { "epoch": 0.09, "learning_rate": 9.724873367102106e-06, "loss": 0.1298, "step": 2164 }, { "epoch": 0.09, "learning_rate": 9.72474006931485e-06, "loss": 0.0934, "step": 2165 }, { "epoch": 0.09, "learning_rate": 9.724606771527593e-06, "loss": 0.1384, "step": 2166 }, { "epoch": 0.09, "learning_rate": 9.724473473740336e-06, "loss": 0.1015, "step": 2167 }, { "epoch": 0.09, "learning_rate": 9.72434017595308e-06, "loss": 0.122, "step": 2168 }, { "epoch": 0.09, "learning_rate": 9.724206878165823e-06, "loss": 0.1132, "step": 2169 }, { "epoch": 0.09, "learning_rate": 9.724073580378566e-06, "loss": 0.1103, "step": 2170 }, { "epoch": 0.09, "learning_rate": 9.723940282591309e-06, "loss": 0.1221, "step": 2171 }, { "epoch": 0.09, "learning_rate": 9.723806984804054e-06, "loss": 0.129, "step": 2172 }, { "epoch": 0.09, "learning_rate": 9.723673687016796e-06, "loss": 0.095, "step": 2173 }, { "epoch": 0.09, "learning_rate": 9.72354038922954e-06, "loss": 0.1196, "step": 2174 }, { "epoch": 0.09, "learning_rate": 9.723407091442284e-06, "loss": 0.1518, "step": 2175 }, { "epoch": 0.09, "learning_rate": 9.723273793655027e-06, "loss": 0.1067, "step": 2176 }, { "epoch": 0.09, "learning_rate": 9.72314049586777e-06, "loss": 0.1528, "step": 2177 }, { "epoch": 0.09, "learning_rate": 9.723007198080512e-06, "loss": 0.119, "step": 2178 }, { "epoch": 0.09, "learning_rate": 9.722873900293257e-06, "loss": 0.1026, "step": 2179 }, { "epoch": 0.09, "learning_rate": 9.722740602506e-06, "loss": 0.1365, "step": 2180 }, { "epoch": 0.09, "learning_rate": 9.722607304718742e-06, "loss": 0.146, "step": 2181 }, { "epoch": 0.09, "learning_rate": 9.722474006931487e-06, "loss": 0.1439, "step": 2182 }, { "epoch": 0.09, "learning_rate": 9.72234070914423e-06, "loss": 0.0935, "step": 2183 }, { "epoch": 0.09, "learning_rate": 9.722207411356972e-06, "loss": 0.113, "step": 2184 }, { "epoch": 0.09, "learning_rate": 9.722074113569715e-06, "loss": 0.1226, "step": 2185 }, { "epoch": 0.09, "learning_rate": 9.72194081578246e-06, "loss": 0.1286, "step": 2186 }, { "epoch": 0.09, "learning_rate": 9.721807517995203e-06, "loss": 0.1043, "step": 2187 }, { "epoch": 0.09, "learning_rate": 9.721674220207945e-06, "loss": 0.1557, "step": 2188 }, { "epoch": 0.09, "learning_rate": 9.72154092242069e-06, "loss": 0.1515, "step": 2189 }, { "epoch": 0.09, "learning_rate": 9.721407624633433e-06, "loss": 0.1204, "step": 2190 }, { "epoch": 0.09, "learning_rate": 9.721274326846176e-06, "loss": 0.0865, "step": 2191 }, { "epoch": 0.09, "learning_rate": 9.721141029058918e-06, "loss": 0.1234, "step": 2192 }, { "epoch": 0.09, "learning_rate": 9.721007731271663e-06, "loss": 0.0997, "step": 2193 }, { "epoch": 0.09, "learning_rate": 9.720874433484406e-06, "loss": 0.1075, "step": 2194 }, { "epoch": 0.09, "learning_rate": 9.720741135697149e-06, "loss": 0.1163, "step": 2195 }, { "epoch": 0.09, "learning_rate": 9.720607837909891e-06, "loss": 0.1645, "step": 2196 }, { "epoch": 0.09, "learning_rate": 9.720474540122636e-06, "loss": 0.1265, "step": 2197 }, { "epoch": 0.09, "learning_rate": 9.720341242335379e-06, "loss": 0.13, "step": 2198 }, { "epoch": 0.09, "learning_rate": 9.720207944548122e-06, "loss": 0.1119, "step": 2199 }, { "epoch": 0.09, "learning_rate": 9.720074646760864e-06, "loss": 0.1338, "step": 2200 }, { "epoch": 0.09, "learning_rate": 9.719941348973609e-06, "loss": 0.1236, "step": 2201 }, { "epoch": 0.09, "learning_rate": 9.719808051186352e-06, "loss": 0.1225, "step": 2202 }, { "epoch": 0.09, "learning_rate": 9.719674753399094e-06, "loss": 0.128, "step": 2203 }, { "epoch": 0.09, "learning_rate": 9.719541455611837e-06, "loss": 0.0845, "step": 2204 }, { "epoch": 0.09, "learning_rate": 9.719408157824582e-06, "loss": 0.1166, "step": 2205 }, { "epoch": 0.09, "learning_rate": 9.719274860037325e-06, "loss": 0.1787, "step": 2206 }, { "epoch": 0.09, "learning_rate": 9.719141562250067e-06, "loss": 0.1141, "step": 2207 }, { "epoch": 0.09, "learning_rate": 9.71900826446281e-06, "loss": 0.1653, "step": 2208 }, { "epoch": 0.09, "learning_rate": 9.718874966675555e-06, "loss": 0.1087, "step": 2209 }, { "epoch": 0.09, "learning_rate": 9.718741668888298e-06, "loss": 0.1391, "step": 2210 }, { "epoch": 0.09, "learning_rate": 9.71860837110104e-06, "loss": 0.161, "step": 2211 }, { "epoch": 0.09, "learning_rate": 9.718475073313783e-06, "loss": 0.1224, "step": 2212 }, { "epoch": 0.09, "learning_rate": 9.718341775526528e-06, "loss": 0.1015, "step": 2213 }, { "epoch": 0.09, "learning_rate": 9.71820847773927e-06, "loss": 0.1009, "step": 2214 }, { "epoch": 0.09, "learning_rate": 9.718075179952013e-06, "loss": 0.0886, "step": 2215 }, { "epoch": 0.09, "learning_rate": 9.717941882164756e-06, "loss": 0.1379, "step": 2216 }, { "epoch": 0.09, "learning_rate": 9.7178085843775e-06, "loss": 0.1317, "step": 2217 }, { "epoch": 0.09, "learning_rate": 9.717675286590244e-06, "loss": 0.1152, "step": 2218 }, { "epoch": 0.09, "learning_rate": 9.717541988802986e-06, "loss": 0.1021, "step": 2219 }, { "epoch": 0.09, "learning_rate": 9.71740869101573e-06, "loss": 0.1932, "step": 2220 }, { "epoch": 0.09, "learning_rate": 9.717275393228474e-06, "loss": 0.1122, "step": 2221 }, { "epoch": 0.09, "learning_rate": 9.717142095441217e-06, "loss": 0.1156, "step": 2222 }, { "epoch": 0.09, "learning_rate": 9.71700879765396e-06, "loss": 0.1641, "step": 2223 }, { "epoch": 0.09, "learning_rate": 9.716875499866702e-06, "loss": 0.0979, "step": 2224 }, { "epoch": 0.09, "learning_rate": 9.716742202079447e-06, "loss": 0.1247, "step": 2225 }, { "epoch": 0.09, "learning_rate": 9.71660890429219e-06, "loss": 0.1125, "step": 2226 }, { "epoch": 0.09, "learning_rate": 9.716475606504932e-06, "loss": 0.1129, "step": 2227 }, { "epoch": 0.09, "learning_rate": 9.716342308717675e-06, "loss": 0.1222, "step": 2228 }, { "epoch": 0.09, "learning_rate": 9.71620901093042e-06, "loss": 0.1329, "step": 2229 }, { "epoch": 0.09, "learning_rate": 9.716075713143162e-06, "loss": 0.1195, "step": 2230 }, { "epoch": 0.09, "learning_rate": 9.715942415355905e-06, "loss": 0.102, "step": 2231 }, { "epoch": 0.09, "learning_rate": 9.715809117568648e-06, "loss": 0.1405, "step": 2232 }, { "epoch": 0.09, "learning_rate": 9.715675819781393e-06, "loss": 0.178, "step": 2233 }, { "epoch": 0.09, "learning_rate": 9.715542521994135e-06, "loss": 0.1061, "step": 2234 }, { "epoch": 0.09, "learning_rate": 9.715409224206878e-06, "loss": 0.1514, "step": 2235 }, { "epoch": 0.09, "learning_rate": 9.715275926419621e-06, "loss": 0.11, "step": 2236 }, { "epoch": 0.09, "learning_rate": 9.715142628632366e-06, "loss": 0.1555, "step": 2237 }, { "epoch": 0.09, "learning_rate": 9.715009330845108e-06, "loss": 0.1583, "step": 2238 }, { "epoch": 0.09, "learning_rate": 9.714876033057851e-06, "loss": 0.1616, "step": 2239 }, { "epoch": 0.09, "learning_rate": 9.714742735270594e-06, "loss": 0.1238, "step": 2240 }, { "epoch": 0.09, "learning_rate": 9.714609437483339e-06, "loss": 0.1404, "step": 2241 }, { "epoch": 0.09, "learning_rate": 9.714476139696081e-06, "loss": 0.1305, "step": 2242 }, { "epoch": 0.09, "learning_rate": 9.714342841908824e-06, "loss": 0.1051, "step": 2243 }, { "epoch": 0.09, "learning_rate": 9.714209544121569e-06, "loss": 0.0934, "step": 2244 }, { "epoch": 0.09, "learning_rate": 9.714076246334312e-06, "loss": 0.1325, "step": 2245 }, { "epoch": 0.09, "learning_rate": 9.713942948547054e-06, "loss": 0.1249, "step": 2246 }, { "epoch": 0.09, "learning_rate": 9.713809650759797e-06, "loss": 0.1253, "step": 2247 }, { "epoch": 0.09, "learning_rate": 9.713676352972542e-06, "loss": 0.1091, "step": 2248 }, { "epoch": 0.09, "learning_rate": 9.713543055185284e-06, "loss": 0.1365, "step": 2249 }, { "epoch": 0.09, "learning_rate": 9.713409757398027e-06, "loss": 0.1682, "step": 2250 }, { "epoch": 0.09, "learning_rate": 9.71327645961077e-06, "loss": 0.1163, "step": 2251 }, { "epoch": 0.09, "learning_rate": 9.713143161823515e-06, "loss": 0.1492, "step": 2252 }, { "epoch": 0.09, "learning_rate": 9.713009864036257e-06, "loss": 0.1192, "step": 2253 }, { "epoch": 0.09, "learning_rate": 9.712876566249e-06, "loss": 0.127, "step": 2254 }, { "epoch": 0.09, "learning_rate": 9.712743268461745e-06, "loss": 0.1173, "step": 2255 }, { "epoch": 0.09, "learning_rate": 9.712609970674488e-06, "loss": 0.1577, "step": 2256 }, { "epoch": 0.09, "learning_rate": 9.71247667288723e-06, "loss": 0.1141, "step": 2257 }, { "epoch": 0.09, "learning_rate": 9.712343375099973e-06, "loss": 0.122, "step": 2258 }, { "epoch": 0.09, "learning_rate": 9.712210077312718e-06, "loss": 0.1325, "step": 2259 }, { "epoch": 0.09, "learning_rate": 9.71207677952546e-06, "loss": 0.265, "step": 2260 }, { "epoch": 0.09, "learning_rate": 9.711943481738203e-06, "loss": 0.1778, "step": 2261 }, { "epoch": 0.09, "learning_rate": 9.711810183950948e-06, "loss": 0.1511, "step": 2262 }, { "epoch": 0.09, "learning_rate": 9.71167688616369e-06, "loss": 0.1194, "step": 2263 }, { "epoch": 0.09, "learning_rate": 9.711543588376434e-06, "loss": 0.1033, "step": 2264 }, { "epoch": 0.09, "learning_rate": 9.711410290589176e-06, "loss": 0.1107, "step": 2265 }, { "epoch": 0.09, "learning_rate": 9.711276992801921e-06, "loss": 0.1545, "step": 2266 }, { "epoch": 0.09, "learning_rate": 9.711143695014664e-06, "loss": 0.1084, "step": 2267 }, { "epoch": 0.09, "learning_rate": 9.711010397227407e-06, "loss": 0.1276, "step": 2268 }, { "epoch": 0.09, "learning_rate": 9.710877099440151e-06, "loss": 0.1136, "step": 2269 }, { "epoch": 0.09, "learning_rate": 9.710743801652894e-06, "loss": 0.1053, "step": 2270 }, { "epoch": 0.09, "learning_rate": 9.710610503865637e-06, "loss": 0.119, "step": 2271 }, { "epoch": 0.09, "learning_rate": 9.71047720607838e-06, "loss": 0.1589, "step": 2272 }, { "epoch": 0.09, "learning_rate": 9.710343908291124e-06, "loss": 0.1136, "step": 2273 }, { "epoch": 0.09, "learning_rate": 9.710210610503867e-06, "loss": 0.1477, "step": 2274 }, { "epoch": 0.09, "learning_rate": 9.71007731271661e-06, "loss": 0.0959, "step": 2275 }, { "epoch": 0.09, "learning_rate": 9.709944014929354e-06, "loss": 0.1073, "step": 2276 }, { "epoch": 0.09, "learning_rate": 9.709810717142097e-06, "loss": 0.1137, "step": 2277 }, { "epoch": 0.09, "learning_rate": 9.70967741935484e-06, "loss": 0.117, "step": 2278 }, { "epoch": 0.09, "learning_rate": 9.709544121567583e-06, "loss": 0.1249, "step": 2279 }, { "epoch": 0.09, "learning_rate": 9.709410823780327e-06, "loss": 0.1814, "step": 2280 }, { "epoch": 0.09, "learning_rate": 9.70927752599307e-06, "loss": 0.1244, "step": 2281 }, { "epoch": 0.09, "learning_rate": 9.709144228205813e-06, "loss": 0.1344, "step": 2282 }, { "epoch": 0.09, "learning_rate": 9.709010930418556e-06, "loss": 0.1335, "step": 2283 }, { "epoch": 0.09, "learning_rate": 9.7088776326313e-06, "loss": 0.1338, "step": 2284 }, { "epoch": 0.09, "learning_rate": 9.708744334844043e-06, "loss": 0.0972, "step": 2285 }, { "epoch": 0.09, "learning_rate": 9.708611037056786e-06, "loss": 0.1444, "step": 2286 }, { "epoch": 0.09, "learning_rate": 9.708477739269529e-06, "loss": 0.1246, "step": 2287 }, { "epoch": 0.09, "learning_rate": 9.708344441482273e-06, "loss": 0.1062, "step": 2288 }, { "epoch": 0.09, "learning_rate": 9.708211143695016e-06, "loss": 0.1209, "step": 2289 }, { "epoch": 0.09, "learning_rate": 9.708077845907759e-06, "loss": 0.1677, "step": 2290 }, { "epoch": 0.09, "learning_rate": 9.707944548120502e-06, "loss": 0.127, "step": 2291 }, { "epoch": 0.09, "learning_rate": 9.707811250333246e-06, "loss": 0.1708, "step": 2292 }, { "epoch": 0.09, "learning_rate": 9.707677952545989e-06, "loss": 0.1468, "step": 2293 }, { "epoch": 0.09, "learning_rate": 9.707544654758732e-06, "loss": 0.1246, "step": 2294 }, { "epoch": 0.09, "learning_rate": 9.707411356971474e-06, "loss": 0.0963, "step": 2295 }, { "epoch": 0.09, "learning_rate": 9.707278059184219e-06, "loss": 0.1028, "step": 2296 }, { "epoch": 0.09, "learning_rate": 9.707144761396962e-06, "loss": 0.1691, "step": 2297 }, { "epoch": 0.09, "learning_rate": 9.707011463609705e-06, "loss": 0.1136, "step": 2298 }, { "epoch": 0.09, "learning_rate": 9.70687816582245e-06, "loss": 0.1161, "step": 2299 }, { "epoch": 0.09, "learning_rate": 9.706744868035192e-06, "loss": 0.1248, "step": 2300 }, { "epoch": 0.09, "learning_rate": 9.706611570247935e-06, "loss": 0.149, "step": 2301 }, { "epoch": 0.09, "learning_rate": 9.706478272460678e-06, "loss": 0.1277, "step": 2302 }, { "epoch": 0.09, "learning_rate": 9.706344974673422e-06, "loss": 0.1427, "step": 2303 }, { "epoch": 0.09, "learning_rate": 9.706211676886165e-06, "loss": 0.1808, "step": 2304 }, { "epoch": 0.09, "learning_rate": 9.706078379098908e-06, "loss": 0.1468, "step": 2305 }, { "epoch": 0.09, "learning_rate": 9.70594508131165e-06, "loss": 0.1415, "step": 2306 }, { "epoch": 0.09, "learning_rate": 9.705811783524395e-06, "loss": 0.1225, "step": 2307 }, { "epoch": 0.09, "learning_rate": 9.705678485737138e-06, "loss": 0.1066, "step": 2308 }, { "epoch": 0.09, "learning_rate": 9.70554518794988e-06, "loss": 0.0824, "step": 2309 }, { "epoch": 0.09, "learning_rate": 9.705411890162624e-06, "loss": 0.1344, "step": 2310 }, { "epoch": 0.09, "learning_rate": 9.705278592375368e-06, "loss": 0.1665, "step": 2311 }, { "epoch": 0.09, "learning_rate": 9.705145294588111e-06, "loss": 0.1205, "step": 2312 }, { "epoch": 0.09, "learning_rate": 9.705011996800854e-06, "loss": 0.1141, "step": 2313 }, { "epoch": 0.09, "learning_rate": 9.704878699013597e-06, "loss": 0.1145, "step": 2314 }, { "epoch": 0.09, "learning_rate": 9.704745401226341e-06, "loss": 0.1317, "step": 2315 }, { "epoch": 0.09, "learning_rate": 9.704612103439084e-06, "loss": 0.1173, "step": 2316 }, { "epoch": 0.09, "learning_rate": 9.704478805651827e-06, "loss": 0.1274, "step": 2317 }, { "epoch": 0.09, "learning_rate": 9.70434550786457e-06, "loss": 0.091, "step": 2318 }, { "epoch": 0.09, "learning_rate": 9.704212210077314e-06, "loss": 0.1195, "step": 2319 }, { "epoch": 0.09, "learning_rate": 9.704078912290057e-06, "loss": 0.1015, "step": 2320 }, { "epoch": 0.09, "learning_rate": 9.7039456145028e-06, "loss": 0.1615, "step": 2321 }, { "epoch": 0.09, "learning_rate": 9.703812316715542e-06, "loss": 0.1357, "step": 2322 }, { "epoch": 0.09, "learning_rate": 9.703679018928287e-06, "loss": 0.121, "step": 2323 }, { "epoch": 0.09, "learning_rate": 9.70354572114103e-06, "loss": 0.1574, "step": 2324 }, { "epoch": 0.09, "learning_rate": 9.703412423353773e-06, "loss": 0.1341, "step": 2325 }, { "epoch": 0.09, "learning_rate": 9.703279125566515e-06, "loss": 0.1087, "step": 2326 }, { "epoch": 0.09, "learning_rate": 9.70314582777926e-06, "loss": 0.1087, "step": 2327 }, { "epoch": 0.09, "learning_rate": 9.703012529992003e-06, "loss": 0.1257, "step": 2328 }, { "epoch": 0.09, "learning_rate": 9.702879232204746e-06, "loss": 0.1246, "step": 2329 }, { "epoch": 0.09, "learning_rate": 9.702745934417488e-06, "loss": 0.124, "step": 2330 }, { "epoch": 0.09, "learning_rate": 9.702612636630233e-06, "loss": 0.1435, "step": 2331 }, { "epoch": 0.09, "learning_rate": 9.702479338842976e-06, "loss": 0.1345, "step": 2332 }, { "epoch": 0.09, "learning_rate": 9.702346041055719e-06, "loss": 0.1726, "step": 2333 }, { "epoch": 0.09, "learning_rate": 9.702212743268461e-06, "loss": 0.0985, "step": 2334 }, { "epoch": 0.09, "learning_rate": 9.702079445481206e-06, "loss": 0.1115, "step": 2335 }, { "epoch": 0.09, "learning_rate": 9.701946147693949e-06, "loss": 0.1226, "step": 2336 }, { "epoch": 0.09, "learning_rate": 9.701812849906692e-06, "loss": 0.1175, "step": 2337 }, { "epoch": 0.09, "learning_rate": 9.701679552119434e-06, "loss": 0.1271, "step": 2338 }, { "epoch": 0.09, "learning_rate": 9.701546254332179e-06, "loss": 0.1261, "step": 2339 }, { "epoch": 0.09, "learning_rate": 9.701412956544922e-06, "loss": 0.1134, "step": 2340 }, { "epoch": 0.09, "learning_rate": 9.701279658757664e-06, "loss": 0.1401, "step": 2341 }, { "epoch": 0.09, "learning_rate": 9.701146360970409e-06, "loss": 0.1013, "step": 2342 }, { "epoch": 0.09, "learning_rate": 9.701013063183152e-06, "loss": 0.1345, "step": 2343 }, { "epoch": 0.09, "learning_rate": 9.700879765395895e-06, "loss": 0.1216, "step": 2344 }, { "epoch": 0.09, "learning_rate": 9.700746467608637e-06, "loss": 0.1348, "step": 2345 }, { "epoch": 0.09, "learning_rate": 9.700613169821382e-06, "loss": 0.1103, "step": 2346 }, { "epoch": 0.09, "learning_rate": 9.700479872034125e-06, "loss": 0.1227, "step": 2347 }, { "epoch": 0.09, "learning_rate": 9.700346574246868e-06, "loss": 0.132, "step": 2348 }, { "epoch": 0.09, "learning_rate": 9.700213276459612e-06, "loss": 0.1242, "step": 2349 }, { "epoch": 0.09, "learning_rate": 9.700079978672355e-06, "loss": 0.0911, "step": 2350 }, { "epoch": 0.09, "learning_rate": 9.699946680885098e-06, "loss": 0.1149, "step": 2351 }, { "epoch": 0.09, "learning_rate": 9.69981338309784e-06, "loss": 0.0975, "step": 2352 }, { "epoch": 0.09, "learning_rate": 9.699680085310585e-06, "loss": 0.1285, "step": 2353 }, { "epoch": 0.09, "learning_rate": 9.699546787523328e-06, "loss": 0.1599, "step": 2354 }, { "epoch": 0.09, "learning_rate": 9.69941348973607e-06, "loss": 0.1588, "step": 2355 }, { "epoch": 0.09, "learning_rate": 9.699280191948815e-06, "loss": 0.1119, "step": 2356 }, { "epoch": 0.09, "learning_rate": 9.699146894161558e-06, "loss": 0.1386, "step": 2357 }, { "epoch": 0.09, "learning_rate": 9.699013596374301e-06, "loss": 0.1824, "step": 2358 }, { "epoch": 0.09, "learning_rate": 9.698880298587044e-06, "loss": 0.1029, "step": 2359 }, { "epoch": 0.09, "learning_rate": 9.698747000799788e-06, "loss": 0.1512, "step": 2360 }, { "epoch": 0.09, "learning_rate": 9.698613703012531e-06, "loss": 0.1156, "step": 2361 }, { "epoch": 0.09, "learning_rate": 9.698480405225274e-06, "loss": 0.1169, "step": 2362 }, { "epoch": 0.09, "learning_rate": 9.698347107438018e-06, "loss": 0.1595, "step": 2363 }, { "epoch": 0.09, "learning_rate": 9.698213809650761e-06, "loss": 0.0976, "step": 2364 }, { "epoch": 0.09, "learning_rate": 9.698080511863504e-06, "loss": 0.1426, "step": 2365 }, { "epoch": 0.09, "learning_rate": 9.697947214076247e-06, "loss": 0.0931, "step": 2366 }, { "epoch": 0.09, "learning_rate": 9.697813916288991e-06, "loss": 0.1368, "step": 2367 }, { "epoch": 0.09, "learning_rate": 9.697680618501734e-06, "loss": 0.1148, "step": 2368 }, { "epoch": 0.09, "learning_rate": 9.697547320714477e-06, "loss": 0.1021, "step": 2369 }, { "epoch": 0.09, "learning_rate": 9.697414022927221e-06, "loss": 0.1649, "step": 2370 }, { "epoch": 0.09, "learning_rate": 9.697280725139964e-06, "loss": 0.1333, "step": 2371 }, { "epoch": 0.09, "learning_rate": 9.697147427352707e-06, "loss": 0.1124, "step": 2372 }, { "epoch": 0.09, "learning_rate": 9.69701412956545e-06, "loss": 0.1162, "step": 2373 }, { "epoch": 0.09, "learning_rate": 9.696880831778194e-06, "loss": 0.1002, "step": 2374 }, { "epoch": 0.09, "learning_rate": 9.696747533990937e-06, "loss": 0.1342, "step": 2375 }, { "epoch": 0.09, "learning_rate": 9.69661423620368e-06, "loss": 0.1111, "step": 2376 }, { "epoch": 0.09, "learning_rate": 9.696480938416423e-06, "loss": 0.1215, "step": 2377 }, { "epoch": 0.09, "learning_rate": 9.696347640629167e-06, "loss": 0.1777, "step": 2378 }, { "epoch": 0.1, "learning_rate": 9.69621434284191e-06, "loss": 0.1532, "step": 2379 }, { "epoch": 0.1, "learning_rate": 9.696081045054653e-06, "loss": 0.1615, "step": 2380 }, { "epoch": 0.1, "learning_rate": 9.695947747267396e-06, "loss": 0.0897, "step": 2381 }, { "epoch": 0.1, "learning_rate": 9.69581444948014e-06, "loss": 0.1297, "step": 2382 }, { "epoch": 0.1, "learning_rate": 9.695681151692883e-06, "loss": 0.1441, "step": 2383 }, { "epoch": 0.1, "learning_rate": 9.695547853905626e-06, "loss": 0.1142, "step": 2384 }, { "epoch": 0.1, "learning_rate": 9.695414556118369e-06, "loss": 0.0901, "step": 2385 }, { "epoch": 0.1, "learning_rate": 9.695281258331113e-06, "loss": 0.1396, "step": 2386 }, { "epoch": 0.1, "learning_rate": 9.695147960543856e-06, "loss": 0.1316, "step": 2387 }, { "epoch": 0.1, "learning_rate": 9.695014662756599e-06, "loss": 0.1302, "step": 2388 }, { "epoch": 0.1, "learning_rate": 9.694881364969342e-06, "loss": 0.1772, "step": 2389 }, { "epoch": 0.1, "learning_rate": 9.694748067182086e-06, "loss": 0.1004, "step": 2390 }, { "epoch": 0.1, "learning_rate": 9.69461476939483e-06, "loss": 0.1022, "step": 2391 }, { "epoch": 0.1, "learning_rate": 9.694481471607572e-06, "loss": 0.12, "step": 2392 }, { "epoch": 0.1, "learning_rate": 9.694348173820315e-06, "loss": 0.1222, "step": 2393 }, { "epoch": 0.1, "learning_rate": 9.69421487603306e-06, "loss": 0.1092, "step": 2394 }, { "epoch": 0.1, "learning_rate": 9.694081578245802e-06, "loss": 0.117, "step": 2395 }, { "epoch": 0.1, "learning_rate": 9.693948280458545e-06, "loss": 0.0963, "step": 2396 }, { "epoch": 0.1, "learning_rate": 9.693814982671288e-06, "loss": 0.1107, "step": 2397 }, { "epoch": 0.1, "learning_rate": 9.693681684884032e-06, "loss": 0.1583, "step": 2398 }, { "epoch": 0.1, "learning_rate": 9.693548387096775e-06, "loss": 0.1098, "step": 2399 }, { "epoch": 0.1, "learning_rate": 9.693415089309518e-06, "loss": 0.1299, "step": 2400 }, { "epoch": 0.1, "learning_rate": 9.69328179152226e-06, "loss": 0.1174, "step": 2401 }, { "epoch": 0.1, "learning_rate": 9.693148493735005e-06, "loss": 0.1274, "step": 2402 }, { "epoch": 0.1, "learning_rate": 9.693015195947748e-06, "loss": 0.1326, "step": 2403 }, { "epoch": 0.1, "learning_rate": 9.692881898160491e-06, "loss": 0.1111, "step": 2404 }, { "epoch": 0.1, "learning_rate": 9.692748600373234e-06, "loss": 0.1211, "step": 2405 }, { "epoch": 0.1, "learning_rate": 9.692615302585978e-06, "loss": 0.1154, "step": 2406 }, { "epoch": 0.1, "learning_rate": 9.692482004798721e-06, "loss": 0.0936, "step": 2407 }, { "epoch": 0.1, "learning_rate": 9.692348707011464e-06, "loss": 0.0862, "step": 2408 }, { "epoch": 0.1, "learning_rate": 9.692215409224207e-06, "loss": 0.1432, "step": 2409 }, { "epoch": 0.1, "learning_rate": 9.692082111436951e-06, "loss": 0.0839, "step": 2410 }, { "epoch": 0.1, "learning_rate": 9.691948813649694e-06, "loss": 0.1292, "step": 2411 }, { "epoch": 0.1, "learning_rate": 9.691815515862437e-06, "loss": 0.1309, "step": 2412 }, { "epoch": 0.1, "learning_rate": 9.691682218075181e-06, "loss": 0.1396, "step": 2413 }, { "epoch": 0.1, "learning_rate": 9.691548920287924e-06, "loss": 0.122, "step": 2414 }, { "epoch": 0.1, "learning_rate": 9.691415622500667e-06, "loss": 0.1455, "step": 2415 }, { "epoch": 0.1, "learning_rate": 9.69128232471341e-06, "loss": 0.1105, "step": 2416 }, { "epoch": 0.1, "learning_rate": 9.691149026926154e-06, "loss": 0.1099, "step": 2417 }, { "epoch": 0.1, "learning_rate": 9.691015729138897e-06, "loss": 0.1278, "step": 2418 }, { "epoch": 0.1, "learning_rate": 9.69088243135164e-06, "loss": 0.1065, "step": 2419 }, { "epoch": 0.1, "learning_rate": 9.690749133564383e-06, "loss": 0.1188, "step": 2420 }, { "epoch": 0.1, "learning_rate": 9.690615835777127e-06, "loss": 0.1328, "step": 2421 }, { "epoch": 0.1, "learning_rate": 9.69048253798987e-06, "loss": 0.1242, "step": 2422 }, { "epoch": 0.1, "learning_rate": 9.690349240202613e-06, "loss": 0.1563, "step": 2423 }, { "epoch": 0.1, "learning_rate": 9.690215942415356e-06, "loss": 0.1031, "step": 2424 }, { "epoch": 0.1, "learning_rate": 9.6900826446281e-06, "loss": 0.1438, "step": 2425 }, { "epoch": 0.1, "learning_rate": 9.689949346840843e-06, "loss": 0.1011, "step": 2426 }, { "epoch": 0.1, "learning_rate": 9.689816049053586e-06, "loss": 0.1375, "step": 2427 }, { "epoch": 0.1, "learning_rate": 9.689682751266329e-06, "loss": 0.1088, "step": 2428 }, { "epoch": 0.1, "learning_rate": 9.689549453479073e-06, "loss": 0.1324, "step": 2429 }, { "epoch": 0.1, "learning_rate": 9.689416155691816e-06, "loss": 0.1605, "step": 2430 }, { "epoch": 0.1, "learning_rate": 9.689282857904559e-06, "loss": 0.1576, "step": 2431 }, { "epoch": 0.1, "learning_rate": 9.689149560117302e-06, "loss": 0.146, "step": 2432 }, { "epoch": 0.1, "learning_rate": 9.689016262330046e-06, "loss": 0.1257, "step": 2433 }, { "epoch": 0.1, "learning_rate": 9.688882964542789e-06, "loss": 0.1344, "step": 2434 }, { "epoch": 0.1, "learning_rate": 9.688749666755532e-06, "loss": 0.1017, "step": 2435 }, { "epoch": 0.1, "learning_rate": 9.688616368968276e-06, "loss": 0.1412, "step": 2436 }, { "epoch": 0.1, "learning_rate": 9.68848307118102e-06, "loss": 0.1103, "step": 2437 }, { "epoch": 0.1, "learning_rate": 9.688349773393762e-06, "loss": 0.147, "step": 2438 }, { "epoch": 0.1, "learning_rate": 9.688216475606505e-06, "loss": 0.128, "step": 2439 }, { "epoch": 0.1, "learning_rate": 9.68808317781925e-06, "loss": 0.1409, "step": 2440 }, { "epoch": 0.1, "learning_rate": 9.687949880031992e-06, "loss": 0.1265, "step": 2441 }, { "epoch": 0.1, "learning_rate": 9.687816582244735e-06, "loss": 0.1199, "step": 2442 }, { "epoch": 0.1, "learning_rate": 9.68768328445748e-06, "loss": 0.141, "step": 2443 }, { "epoch": 0.1, "learning_rate": 9.687549986670222e-06, "loss": 0.1505, "step": 2444 }, { "epoch": 0.1, "learning_rate": 9.687416688882965e-06, "loss": 0.1231, "step": 2445 }, { "epoch": 0.1, "learning_rate": 9.687283391095708e-06, "loss": 0.1366, "step": 2446 }, { "epoch": 0.1, "learning_rate": 9.687150093308452e-06, "loss": 0.1227, "step": 2447 }, { "epoch": 0.1, "learning_rate": 9.687016795521195e-06, "loss": 0.1264, "step": 2448 }, { "epoch": 0.1, "learning_rate": 9.686883497733938e-06, "loss": 0.0953, "step": 2449 }, { "epoch": 0.1, "learning_rate": 9.686750199946683e-06, "loss": 0.1123, "step": 2450 }, { "epoch": 0.1, "learning_rate": 9.686616902159425e-06, "loss": 0.124, "step": 2451 }, { "epoch": 0.1, "learning_rate": 9.686483604372168e-06, "loss": 0.1039, "step": 2452 }, { "epoch": 0.1, "learning_rate": 9.686350306584911e-06, "loss": 0.0966, "step": 2453 }, { "epoch": 0.1, "learning_rate": 9.686217008797656e-06, "loss": 0.0945, "step": 2454 }, { "epoch": 0.1, "learning_rate": 9.686083711010398e-06, "loss": 0.1286, "step": 2455 }, { "epoch": 0.1, "learning_rate": 9.685950413223141e-06, "loss": 0.1031, "step": 2456 }, { "epoch": 0.1, "learning_rate": 9.685817115435886e-06, "loss": 0.1481, "step": 2457 }, { "epoch": 0.1, "learning_rate": 9.685683817648629e-06, "loss": 0.131, "step": 2458 }, { "epoch": 0.1, "learning_rate": 9.685550519861371e-06, "loss": 0.0919, "step": 2459 }, { "epoch": 0.1, "learning_rate": 9.685417222074114e-06, "loss": 0.1016, "step": 2460 }, { "epoch": 0.1, "learning_rate": 9.685283924286859e-06, "loss": 0.1264, "step": 2461 }, { "epoch": 0.1, "learning_rate": 9.685150626499602e-06, "loss": 0.1731, "step": 2462 }, { "epoch": 0.1, "learning_rate": 9.685017328712344e-06, "loss": 0.149, "step": 2463 }, { "epoch": 0.1, "learning_rate": 9.684884030925087e-06, "loss": 0.1187, "step": 2464 }, { "epoch": 0.1, "learning_rate": 9.684750733137832e-06, "loss": 0.0924, "step": 2465 }, { "epoch": 0.1, "learning_rate": 9.684617435350574e-06, "loss": 0.1264, "step": 2466 }, { "epoch": 0.1, "learning_rate": 9.684484137563317e-06, "loss": 0.1148, "step": 2467 }, { "epoch": 0.1, "learning_rate": 9.684350839776062e-06, "loss": 0.1698, "step": 2468 }, { "epoch": 0.1, "learning_rate": 9.684217541988805e-06, "loss": 0.1387, "step": 2469 }, { "epoch": 0.1, "learning_rate": 9.684084244201547e-06, "loss": 0.1262, "step": 2470 }, { "epoch": 0.1, "learning_rate": 9.68395094641429e-06, "loss": 0.0939, "step": 2471 }, { "epoch": 0.1, "learning_rate": 9.683817648627035e-06, "loss": 0.089, "step": 2472 }, { "epoch": 0.1, "learning_rate": 9.683684350839778e-06, "loss": 0.1508, "step": 2473 }, { "epoch": 0.1, "learning_rate": 9.68355105305252e-06, "loss": 0.1116, "step": 2474 }, { "epoch": 0.1, "learning_rate": 9.683417755265263e-06, "loss": 0.1167, "step": 2475 }, { "epoch": 0.1, "learning_rate": 9.683284457478008e-06, "loss": 0.1376, "step": 2476 }, { "epoch": 0.1, "learning_rate": 9.68315115969075e-06, "loss": 0.1304, "step": 2477 }, { "epoch": 0.1, "learning_rate": 9.683017861903493e-06, "loss": 0.1285, "step": 2478 }, { "epoch": 0.1, "learning_rate": 9.682884564116236e-06, "loss": 0.1072, "step": 2479 }, { "epoch": 0.1, "learning_rate": 9.68275126632898e-06, "loss": 0.1628, "step": 2480 }, { "epoch": 0.1, "learning_rate": 9.682617968541724e-06, "loss": 0.1125, "step": 2481 }, { "epoch": 0.1, "learning_rate": 9.682484670754466e-06, "loss": 0.1536, "step": 2482 }, { "epoch": 0.1, "learning_rate": 9.68235137296721e-06, "loss": 0.1491, "step": 2483 }, { "epoch": 0.1, "learning_rate": 9.682218075179954e-06, "loss": 0.1227, "step": 2484 }, { "epoch": 0.1, "learning_rate": 9.682084777392697e-06, "loss": 0.145, "step": 2485 }, { "epoch": 0.1, "learning_rate": 9.68195147960544e-06, "loss": 0.182, "step": 2486 }, { "epoch": 0.1, "learning_rate": 9.681818181818182e-06, "loss": 0.1251, "step": 2487 }, { "epoch": 0.1, "learning_rate": 9.681684884030927e-06, "loss": 0.1469, "step": 2488 }, { "epoch": 0.1, "learning_rate": 9.68155158624367e-06, "loss": 0.1424, "step": 2489 }, { "epoch": 0.1, "learning_rate": 9.681418288456412e-06, "loss": 0.1137, "step": 2490 }, { "epoch": 0.1, "learning_rate": 9.681284990669155e-06, "loss": 0.1239, "step": 2491 }, { "epoch": 0.1, "learning_rate": 9.6811516928819e-06, "loss": 0.1496, "step": 2492 }, { "epoch": 0.1, "learning_rate": 9.681018395094642e-06, "loss": 0.1523, "step": 2493 }, { "epoch": 0.1, "learning_rate": 9.680885097307385e-06, "loss": 0.1268, "step": 2494 }, { "epoch": 0.1, "learning_rate": 9.680751799520128e-06, "loss": 0.1243, "step": 2495 }, { "epoch": 0.1, "learning_rate": 9.680618501732873e-06, "loss": 0.0969, "step": 2496 }, { "epoch": 0.1, "learning_rate": 9.680485203945615e-06, "loss": 0.1159, "step": 2497 }, { "epoch": 0.1, "learning_rate": 9.680351906158358e-06, "loss": 0.1494, "step": 2498 }, { "epoch": 0.1, "learning_rate": 9.680218608371101e-06, "loss": 0.1, "step": 2499 }, { "epoch": 0.1, "learning_rate": 9.680085310583846e-06, "loss": 0.118, "step": 2500 }, { "epoch": 0.1, "learning_rate": 9.679977607762642e-06, "loss": 0.1028, "step": 2501 }, { "epoch": 0.1, "learning_rate": 9.679844320635513e-06, "loss": 0.1338, "step": 2502 }, { "epoch": 0.1, "learning_rate": 9.679711033508384e-06, "loss": 0.1349, "step": 2503 }, { "epoch": 0.1, "learning_rate": 9.679577746381255e-06, "loss": 0.0982, "step": 2504 }, { "epoch": 0.1, "learning_rate": 9.679444459254126e-06, "loss": 0.1297, "step": 2505 }, { "epoch": 0.1, "learning_rate": 9.679311172126997e-06, "loss": 0.1502, "step": 2506 }, { "epoch": 0.1, "learning_rate": 9.679177884999867e-06, "loss": 0.1259, "step": 2507 }, { "epoch": 0.1, "learning_rate": 9.67904459787274e-06, "loss": 0.1286, "step": 2508 }, { "epoch": 0.1, "learning_rate": 9.678911310745609e-06, "loss": 0.1267, "step": 2509 }, { "epoch": 0.1, "learning_rate": 9.67877802361848e-06, "loss": 0.1256, "step": 2510 }, { "epoch": 0.1, "learning_rate": 9.67864473649135e-06, "loss": 0.0817, "step": 2511 }, { "epoch": 0.1, "learning_rate": 9.67851144936422e-06, "loss": 0.1014, "step": 2512 }, { "epoch": 0.1, "learning_rate": 9.678378162237093e-06, "loss": 0.1521, "step": 2513 }, { "epoch": 0.1, "learning_rate": 9.678244875109962e-06, "loss": 0.1212, "step": 2514 }, { "epoch": 0.1, "learning_rate": 9.678111587982833e-06, "loss": 0.1235, "step": 2515 }, { "epoch": 0.1, "learning_rate": 9.677978300855704e-06, "loss": 0.1469, "step": 2516 }, { "epoch": 0.1, "learning_rate": 9.677845013728575e-06, "loss": 0.1264, "step": 2517 }, { "epoch": 0.1, "learning_rate": 9.677711726601446e-06, "loss": 0.1223, "step": 2518 }, { "epoch": 0.1, "learning_rate": 9.677578439474317e-06, "loss": 0.1248, "step": 2519 }, { "epoch": 0.1, "learning_rate": 9.677445152347186e-06, "loss": 0.0948, "step": 2520 }, { "epoch": 0.1, "learning_rate": 9.677311865220059e-06, "loss": 0.1079, "step": 2521 }, { "epoch": 0.1, "learning_rate": 9.677178578092928e-06, "loss": 0.1051, "step": 2522 }, { "epoch": 0.1, "learning_rate": 9.6770452909658e-06, "loss": 0.1234, "step": 2523 }, { "epoch": 0.1, "learning_rate": 9.67691200383867e-06, "loss": 0.1031, "step": 2524 }, { "epoch": 0.1, "learning_rate": 9.676778716711541e-06, "loss": 0.1776, "step": 2525 }, { "epoch": 0.1, "learning_rate": 9.676645429584412e-06, "loss": 0.135, "step": 2526 }, { "epoch": 0.1, "learning_rate": 9.676512142457283e-06, "loss": 0.1418, "step": 2527 }, { "epoch": 0.1, "learning_rate": 9.676378855330153e-06, "loss": 0.1415, "step": 2528 }, { "epoch": 0.1, "learning_rate": 9.676245568203024e-06, "loss": 0.0914, "step": 2529 }, { "epoch": 0.1, "learning_rate": 9.676112281075895e-06, "loss": 0.1372, "step": 2530 }, { "epoch": 0.1, "learning_rate": 9.675978993948766e-06, "loss": 0.09, "step": 2531 }, { "epoch": 0.1, "learning_rate": 9.675845706821637e-06, "loss": 0.1206, "step": 2532 }, { "epoch": 0.1, "learning_rate": 9.675712419694506e-06, "loss": 0.1214, "step": 2533 }, { "epoch": 0.1, "learning_rate": 9.675579132567377e-06, "loss": 0.1523, "step": 2534 }, { "epoch": 0.1, "learning_rate": 9.675445845440248e-06, "loss": 0.0985, "step": 2535 }, { "epoch": 0.1, "learning_rate": 9.675312558313119e-06, "loss": 0.1104, "step": 2536 }, { "epoch": 0.1, "learning_rate": 9.67517927118599e-06, "loss": 0.1133, "step": 2537 }, { "epoch": 0.1, "learning_rate": 9.67504598405886e-06, "loss": 0.1225, "step": 2538 }, { "epoch": 0.1, "learning_rate": 9.67491269693173e-06, "loss": 0.0875, "step": 2539 }, { "epoch": 0.1, "learning_rate": 9.674779409804603e-06, "loss": 0.1395, "step": 2540 }, { "epoch": 0.1, "learning_rate": 9.674646122677472e-06, "loss": 0.1228, "step": 2541 }, { "epoch": 0.1, "learning_rate": 9.674512835550343e-06, "loss": 0.1279, "step": 2542 }, { "epoch": 0.1, "learning_rate": 9.674379548423214e-06, "loss": 0.1314, "step": 2543 }, { "epoch": 0.1, "learning_rate": 9.674246261296085e-06, "loss": 0.1411, "step": 2544 }, { "epoch": 0.1, "learning_rate": 9.674112974168956e-06, "loss": 0.1043, "step": 2545 }, { "epoch": 0.1, "learning_rate": 9.673979687041827e-06, "loss": 0.1121, "step": 2546 }, { "epoch": 0.1, "learning_rate": 9.673846399914696e-06, "loss": 0.1204, "step": 2547 }, { "epoch": 0.1, "learning_rate": 9.673713112787569e-06, "loss": 0.1217, "step": 2548 }, { "epoch": 0.1, "learning_rate": 9.673579825660438e-06, "loss": 0.1181, "step": 2549 }, { "epoch": 0.1, "learning_rate": 9.67344653853331e-06, "loss": 0.1112, "step": 2550 }, { "epoch": 0.1, "learning_rate": 9.67331325140618e-06, "loss": 0.0999, "step": 2551 }, { "epoch": 0.1, "learning_rate": 9.67317996427905e-06, "loss": 0.1303, "step": 2552 }, { "epoch": 0.1, "learning_rate": 9.673046677151922e-06, "loss": 0.1425, "step": 2553 }, { "epoch": 0.1, "learning_rate": 9.672913390024792e-06, "loss": 0.1425, "step": 2554 }, { "epoch": 0.1, "learning_rate": 9.672780102897663e-06, "loss": 0.0989, "step": 2555 }, { "epoch": 0.1, "learning_rate": 9.672646815770534e-06, "loss": 0.1312, "step": 2556 }, { "epoch": 0.1, "learning_rate": 9.672513528643405e-06, "loss": 0.1256, "step": 2557 }, { "epoch": 0.1, "learning_rate": 9.672380241516276e-06, "loss": 0.0756, "step": 2558 }, { "epoch": 0.1, "learning_rate": 9.672246954389147e-06, "loss": 0.0883, "step": 2559 }, { "epoch": 0.1, "learning_rate": 9.672113667262016e-06, "loss": 0.1152, "step": 2560 }, { "epoch": 0.1, "learning_rate": 9.671980380134887e-06, "loss": 0.1745, "step": 2561 }, { "epoch": 0.1, "learning_rate": 9.671847093007758e-06, "loss": 0.069, "step": 2562 }, { "epoch": 0.1, "learning_rate": 9.671713805880629e-06, "loss": 0.1237, "step": 2563 }, { "epoch": 0.1, "learning_rate": 9.6715805187535e-06, "loss": 0.1041, "step": 2564 }, { "epoch": 0.1, "learning_rate": 9.671447231626371e-06, "loss": 0.1237, "step": 2565 }, { "epoch": 0.1, "learning_rate": 9.67131394449924e-06, "loss": 0.121, "step": 2566 }, { "epoch": 0.1, "learning_rate": 9.671180657372113e-06, "loss": 0.107, "step": 2567 }, { "epoch": 0.1, "learning_rate": 9.671047370244982e-06, "loss": 0.1311, "step": 2568 }, { "epoch": 0.1, "learning_rate": 9.670914083117853e-06, "loss": 0.1127, "step": 2569 }, { "epoch": 0.1, "learning_rate": 9.670780795990724e-06, "loss": 0.1052, "step": 2570 }, { "epoch": 0.1, "learning_rate": 9.670647508863593e-06, "loss": 0.1236, "step": 2571 }, { "epoch": 0.1, "learning_rate": 9.670514221736466e-06, "loss": 0.1122, "step": 2572 }, { "epoch": 0.1, "learning_rate": 9.670380934609335e-06, "loss": 0.1811, "step": 2573 }, { "epoch": 0.1, "learning_rate": 9.670247647482206e-06, "loss": 0.1365, "step": 2574 }, { "epoch": 0.1, "learning_rate": 9.670114360355077e-06, "loss": 0.1058, "step": 2575 }, { "epoch": 0.1, "learning_rate": 9.669981073227948e-06, "loss": 0.1149, "step": 2576 }, { "epoch": 0.1, "learning_rate": 9.66984778610082e-06, "loss": 0.1152, "step": 2577 }, { "epoch": 0.1, "learning_rate": 9.66971449897369e-06, "loss": 0.1317, "step": 2578 }, { "epoch": 0.1, "learning_rate": 9.66958121184656e-06, "loss": 0.1804, "step": 2579 }, { "epoch": 0.1, "learning_rate": 9.669447924719432e-06, "loss": 0.1458, "step": 2580 }, { "epoch": 0.1, "learning_rate": 9.669314637592302e-06, "loss": 0.0738, "step": 2581 }, { "epoch": 0.1, "learning_rate": 9.669181350465173e-06, "loss": 0.1077, "step": 2582 }, { "epoch": 0.1, "learning_rate": 9.669048063338044e-06, "loss": 0.1308, "step": 2583 }, { "epoch": 0.1, "learning_rate": 9.668914776210915e-06, "loss": 0.1011, "step": 2584 }, { "epoch": 0.1, "learning_rate": 9.668781489083786e-06, "loss": 0.0956, "step": 2585 }, { "epoch": 0.1, "learning_rate": 9.668648201956657e-06, "loss": 0.1632, "step": 2586 }, { "epoch": 0.1, "learning_rate": 9.668514914829526e-06, "loss": 0.1233, "step": 2587 }, { "epoch": 0.1, "learning_rate": 9.668381627702397e-06, "loss": 0.1322, "step": 2588 }, { "epoch": 0.1, "learning_rate": 9.668248340575268e-06, "loss": 0.1068, "step": 2589 }, { "epoch": 0.1, "learning_rate": 9.668115053448139e-06, "loss": 0.108, "step": 2590 }, { "epoch": 0.1, "learning_rate": 9.66798176632101e-06, "loss": 0.1404, "step": 2591 }, { "epoch": 0.1, "learning_rate": 9.66784847919388e-06, "loss": 0.1212, "step": 2592 }, { "epoch": 0.1, "learning_rate": 9.667715192066752e-06, "loss": 0.1317, "step": 2593 }, { "epoch": 0.1, "learning_rate": 9.667581904939621e-06, "loss": 0.1455, "step": 2594 }, { "epoch": 0.1, "learning_rate": 9.667448617812492e-06, "loss": 0.1543, "step": 2595 }, { "epoch": 0.1, "learning_rate": 9.667315330685363e-06, "loss": 0.0841, "step": 2596 }, { "epoch": 0.1, "learning_rate": 9.667182043558234e-06, "loss": 0.148, "step": 2597 }, { "epoch": 0.1, "learning_rate": 9.667048756431105e-06, "loss": 0.1148, "step": 2598 }, { "epoch": 0.1, "learning_rate": 9.666915469303976e-06, "loss": 0.0847, "step": 2599 }, { "epoch": 0.1, "learning_rate": 9.666782182176846e-06, "loss": 0.1309, "step": 2600 }, { "epoch": 0.1, "learning_rate": 9.666648895049717e-06, "loss": 0.125, "step": 2601 }, { "epoch": 0.1, "learning_rate": 9.666515607922588e-06, "loss": 0.1285, "step": 2602 }, { "epoch": 0.1, "learning_rate": 9.666382320795459e-06, "loss": 0.116, "step": 2603 }, { "epoch": 0.1, "learning_rate": 9.66624903366833e-06, "loss": 0.1123, "step": 2604 }, { "epoch": 0.1, "learning_rate": 9.6661157465412e-06, "loss": 0.1311, "step": 2605 }, { "epoch": 0.1, "learning_rate": 9.66598245941407e-06, "loss": 0.1348, "step": 2606 }, { "epoch": 0.1, "learning_rate": 9.665849172286942e-06, "loss": 0.1168, "step": 2607 }, { "epoch": 0.1, "learning_rate": 9.665715885159812e-06, "loss": 0.1247, "step": 2608 }, { "epoch": 0.1, "learning_rate": 9.665582598032683e-06, "loss": 0.0976, "step": 2609 }, { "epoch": 0.1, "learning_rate": 9.665449310905554e-06, "loss": 0.1209, "step": 2610 }, { "epoch": 0.1, "learning_rate": 9.665316023778423e-06, "loss": 0.1655, "step": 2611 }, { "epoch": 0.1, "learning_rate": 9.665182736651296e-06, "loss": 0.1112, "step": 2612 }, { "epoch": 0.1, "learning_rate": 9.665049449524165e-06, "loss": 0.1237, "step": 2613 }, { "epoch": 0.1, "learning_rate": 9.664916162397036e-06, "loss": 0.1378, "step": 2614 }, { "epoch": 0.1, "learning_rate": 9.664782875269907e-06, "loss": 0.1597, "step": 2615 }, { "epoch": 0.1, "learning_rate": 9.664649588142778e-06, "loss": 0.1195, "step": 2616 }, { "epoch": 0.1, "learning_rate": 9.664516301015649e-06, "loss": 0.0984, "step": 2617 }, { "epoch": 0.1, "learning_rate": 9.66438301388852e-06, "loss": 0.1194, "step": 2618 }, { "epoch": 0.1, "learning_rate": 9.66424972676139e-06, "loss": 0.1, "step": 2619 }, { "epoch": 0.1, "learning_rate": 9.664116439634262e-06, "loss": 0.1721, "step": 2620 }, { "epoch": 0.1, "learning_rate": 9.663983152507131e-06, "loss": 0.1138, "step": 2621 }, { "epoch": 0.1, "learning_rate": 9.663849865380002e-06, "loss": 0.1061, "step": 2622 }, { "epoch": 0.1, "learning_rate": 9.663716578252873e-06, "loss": 0.1218, "step": 2623 }, { "epoch": 0.1, "learning_rate": 9.663583291125744e-06, "loss": 0.107, "step": 2624 }, { "epoch": 0.1, "learning_rate": 9.663450003998615e-06, "loss": 0.0953, "step": 2625 }, { "epoch": 0.1, "learning_rate": 9.663316716871486e-06, "loss": 0.1272, "step": 2626 }, { "epoch": 0.1, "learning_rate": 9.663183429744356e-06, "loss": 0.14, "step": 2627 }, { "epoch": 0.1, "learning_rate": 9.663050142617227e-06, "loss": 0.1603, "step": 2628 }, { "epoch": 0.1, "learning_rate": 9.662916855490098e-06, "loss": 0.1583, "step": 2629 }, { "epoch": 0.11, "learning_rate": 9.662783568362969e-06, "loss": 0.1194, "step": 2630 }, { "epoch": 0.11, "learning_rate": 9.66265028123584e-06, "loss": 0.1352, "step": 2631 }, { "epoch": 0.11, "learning_rate": 9.662516994108709e-06, "loss": 0.1048, "step": 2632 }, { "epoch": 0.11, "learning_rate": 9.66238370698158e-06, "loss": 0.1027, "step": 2633 }, { "epoch": 0.11, "learning_rate": 9.662250419854451e-06, "loss": 0.0896, "step": 2634 }, { "epoch": 0.11, "learning_rate": 9.662117132727322e-06, "loss": 0.1172, "step": 2635 }, { "epoch": 0.11, "learning_rate": 9.661983845600193e-06, "loss": 0.0974, "step": 2636 }, { "epoch": 0.11, "learning_rate": 9.661850558473064e-06, "loss": 0.1064, "step": 2637 }, { "epoch": 0.11, "learning_rate": 9.661717271345933e-06, "loss": 0.1019, "step": 2638 }, { "epoch": 0.11, "learning_rate": 9.661583984218806e-06, "loss": 0.1385, "step": 2639 }, { "epoch": 0.11, "learning_rate": 9.661450697091675e-06, "loss": 0.1075, "step": 2640 }, { "epoch": 0.11, "learning_rate": 9.661317409964546e-06, "loss": 0.1226, "step": 2641 }, { "epoch": 0.11, "learning_rate": 9.661184122837417e-06, "loss": 0.1084, "step": 2642 }, { "epoch": 0.11, "learning_rate": 9.661050835710288e-06, "loss": 0.1254, "step": 2643 }, { "epoch": 0.11, "learning_rate": 9.660917548583159e-06, "loss": 0.1305, "step": 2644 }, { "epoch": 0.11, "learning_rate": 9.66078426145603e-06, "loss": 0.1247, "step": 2645 }, { "epoch": 0.11, "learning_rate": 9.6606509743289e-06, "loss": 0.1685, "step": 2646 }, { "epoch": 0.11, "learning_rate": 9.66051768720177e-06, "loss": 0.136, "step": 2647 }, { "epoch": 0.11, "learning_rate": 9.660384400074641e-06, "loss": 0.0965, "step": 2648 }, { "epoch": 0.11, "learning_rate": 9.660251112947512e-06, "loss": 0.1596, "step": 2649 }, { "epoch": 0.11, "learning_rate": 9.660117825820383e-06, "loss": 0.1403, "step": 2650 }, { "epoch": 0.11, "learning_rate": 9.659984538693253e-06, "loss": 0.1205, "step": 2651 }, { "epoch": 0.11, "learning_rate": 9.659851251566125e-06, "loss": 0.1384, "step": 2652 }, { "epoch": 0.11, "learning_rate": 9.659717964438995e-06, "loss": 0.1107, "step": 2653 }, { "epoch": 0.11, "learning_rate": 9.659584677311866e-06, "loss": 0.1366, "step": 2654 }, { "epoch": 0.11, "learning_rate": 9.659451390184737e-06, "loss": 0.1027, "step": 2655 }, { "epoch": 0.11, "learning_rate": 9.659318103057608e-06, "loss": 0.1318, "step": 2656 }, { "epoch": 0.11, "learning_rate": 9.659184815930479e-06, "loss": 0.143, "step": 2657 }, { "epoch": 0.11, "learning_rate": 9.65905152880335e-06, "loss": 0.1873, "step": 2658 }, { "epoch": 0.11, "learning_rate": 9.658918241676219e-06, "loss": 0.0923, "step": 2659 }, { "epoch": 0.11, "learning_rate": 9.658784954549092e-06, "loss": 0.1334, "step": 2660 }, { "epoch": 0.11, "learning_rate": 9.658651667421961e-06, "loss": 0.1038, "step": 2661 }, { "epoch": 0.11, "learning_rate": 9.658518380294832e-06, "loss": 0.1282, "step": 2662 }, { "epoch": 0.11, "learning_rate": 9.658385093167703e-06, "loss": 0.119, "step": 2663 }, { "epoch": 0.11, "learning_rate": 9.658251806040574e-06, "loss": 0.1237, "step": 2664 }, { "epoch": 0.11, "learning_rate": 9.658118518913445e-06, "loss": 0.0961, "step": 2665 }, { "epoch": 0.11, "learning_rate": 9.657985231786316e-06, "loss": 0.1044, "step": 2666 }, { "epoch": 0.11, "learning_rate": 9.657851944659185e-06, "loss": 0.107, "step": 2667 }, { "epoch": 0.11, "learning_rate": 9.657718657532056e-06, "loss": 0.1449, "step": 2668 }, { "epoch": 0.11, "learning_rate": 9.657585370404927e-06, "loss": 0.1017, "step": 2669 }, { "epoch": 0.11, "learning_rate": 9.657452083277797e-06, "loss": 0.1077, "step": 2670 }, { "epoch": 0.11, "learning_rate": 9.65731879615067e-06, "loss": 0.1139, "step": 2671 }, { "epoch": 0.11, "learning_rate": 9.657185509023539e-06, "loss": 0.1489, "step": 2672 }, { "epoch": 0.11, "learning_rate": 9.65705222189641e-06, "loss": 0.1618, "step": 2673 }, { "epoch": 0.11, "learning_rate": 9.65691893476928e-06, "loss": 0.1702, "step": 2674 }, { "epoch": 0.11, "learning_rate": 9.656785647642152e-06, "loss": 0.1606, "step": 2675 }, { "epoch": 0.11, "learning_rate": 9.656652360515022e-06, "loss": 0.1439, "step": 2676 }, { "epoch": 0.11, "learning_rate": 9.656519073387893e-06, "loss": 0.093, "step": 2677 }, { "epoch": 0.11, "learning_rate": 9.656385786260763e-06, "loss": 0.1269, "step": 2678 }, { "epoch": 0.11, "learning_rate": 9.656252499133635e-06, "loss": 0.1302, "step": 2679 }, { "epoch": 0.11, "learning_rate": 9.656119212006505e-06, "loss": 0.1153, "step": 2680 }, { "epoch": 0.11, "learning_rate": 9.655985924879376e-06, "loss": 0.0822, "step": 2681 }, { "epoch": 0.11, "learning_rate": 9.655852637752247e-06, "loss": 0.1256, "step": 2682 }, { "epoch": 0.11, "learning_rate": 9.655719350625118e-06, "loss": 0.1444, "step": 2683 }, { "epoch": 0.11, "learning_rate": 9.655586063497989e-06, "loss": 0.1342, "step": 2684 }, { "epoch": 0.11, "learning_rate": 9.65545277637086e-06, "loss": 0.1698, "step": 2685 }, { "epoch": 0.11, "learning_rate": 9.655319489243729e-06, "loss": 0.1206, "step": 2686 }, { "epoch": 0.11, "learning_rate": 9.6551862021166e-06, "loss": 0.0842, "step": 2687 }, { "epoch": 0.11, "learning_rate": 9.655052914989471e-06, "loss": 0.1669, "step": 2688 }, { "epoch": 0.11, "learning_rate": 9.654919627862342e-06, "loss": 0.1067, "step": 2689 }, { "epoch": 0.11, "learning_rate": 9.654786340735213e-06, "loss": 0.1066, "step": 2690 }, { "epoch": 0.11, "learning_rate": 9.654653053608082e-06, "loss": 0.1394, "step": 2691 }, { "epoch": 0.11, "learning_rate": 9.654519766480955e-06, "loss": 0.1152, "step": 2692 }, { "epoch": 0.11, "learning_rate": 9.654386479353824e-06, "loss": 0.1274, "step": 2693 }, { "epoch": 0.11, "learning_rate": 9.654253192226695e-06, "loss": 0.1385, "step": 2694 }, { "epoch": 0.11, "learning_rate": 9.654119905099566e-06, "loss": 0.1088, "step": 2695 }, { "epoch": 0.11, "learning_rate": 9.653986617972437e-06, "loss": 0.1418, "step": 2696 }, { "epoch": 0.11, "learning_rate": 9.653853330845308e-06, "loss": 0.1588, "step": 2697 }, { "epoch": 0.11, "learning_rate": 9.65372004371818e-06, "loss": 0.1274, "step": 2698 }, { "epoch": 0.11, "learning_rate": 9.653586756591049e-06, "loss": 0.1632, "step": 2699 }, { "epoch": 0.11, "learning_rate": 9.65345346946392e-06, "loss": 0.0996, "step": 2700 }, { "epoch": 0.11, "learning_rate": 9.65332018233679e-06, "loss": 0.0899, "step": 2701 }, { "epoch": 0.11, "learning_rate": 9.653186895209662e-06, "loss": 0.1086, "step": 2702 }, { "epoch": 0.11, "learning_rate": 9.653053608082533e-06, "loss": 0.1366, "step": 2703 }, { "epoch": 0.11, "learning_rate": 9.652920320955404e-06, "loss": 0.1056, "step": 2704 }, { "epoch": 0.11, "learning_rate": 9.652787033828273e-06, "loss": 0.1278, "step": 2705 }, { "epoch": 0.11, "learning_rate": 9.652653746701144e-06, "loss": 0.2091, "step": 2706 }, { "epoch": 0.11, "learning_rate": 9.652520459574015e-06, "loss": 0.0965, "step": 2707 }, { "epoch": 0.11, "learning_rate": 9.652387172446886e-06, "loss": 0.1623, "step": 2708 }, { "epoch": 0.11, "learning_rate": 9.652253885319757e-06, "loss": 0.133, "step": 2709 }, { "epoch": 0.11, "learning_rate": 9.652120598192626e-06, "loss": 0.1384, "step": 2710 }, { "epoch": 0.11, "learning_rate": 9.651987311065499e-06, "loss": 0.1277, "step": 2711 }, { "epoch": 0.11, "learning_rate": 9.651854023938368e-06, "loss": 0.1699, "step": 2712 }, { "epoch": 0.11, "learning_rate": 9.651720736811239e-06, "loss": 0.1501, "step": 2713 }, { "epoch": 0.11, "learning_rate": 9.65158744968411e-06, "loss": 0.174, "step": 2714 }, { "epoch": 0.11, "learning_rate": 9.651454162556981e-06, "loss": 0.1115, "step": 2715 }, { "epoch": 0.11, "learning_rate": 9.651320875429852e-06, "loss": 0.1249, "step": 2716 }, { "epoch": 0.11, "learning_rate": 9.651187588302723e-06, "loss": 0.1202, "step": 2717 }, { "epoch": 0.11, "learning_rate": 9.651054301175592e-06, "loss": 0.113, "step": 2718 }, { "epoch": 0.11, "learning_rate": 9.650921014048465e-06, "loss": 0.076, "step": 2719 }, { "epoch": 0.11, "learning_rate": 9.650787726921334e-06, "loss": 0.099, "step": 2720 }, { "epoch": 0.11, "learning_rate": 9.650654439794205e-06, "loss": 0.1554, "step": 2721 }, { "epoch": 0.11, "learning_rate": 9.650521152667076e-06, "loss": 0.1487, "step": 2722 }, { "epoch": 0.11, "learning_rate": 9.650387865539947e-06, "loss": 0.1224, "step": 2723 }, { "epoch": 0.11, "learning_rate": 9.650254578412818e-06, "loss": 0.0899, "step": 2724 }, { "epoch": 0.11, "learning_rate": 9.650121291285688e-06, "loss": 0.1292, "step": 2725 }, { "epoch": 0.11, "learning_rate": 9.649988004158559e-06, "loss": 0.1045, "step": 2726 }, { "epoch": 0.11, "learning_rate": 9.64985471703143e-06, "loss": 0.1002, "step": 2727 }, { "epoch": 0.11, "learning_rate": 9.6497214299043e-06, "loss": 0.1473, "step": 2728 }, { "epoch": 0.11, "learning_rate": 9.649588142777172e-06, "loss": 0.1435, "step": 2729 }, { "epoch": 0.11, "learning_rate": 9.649454855650043e-06, "loss": 0.1099, "step": 2730 }, { "epoch": 0.11, "learning_rate": 9.649321568522912e-06, "loss": 0.0937, "step": 2731 }, { "epoch": 0.11, "learning_rate": 9.649188281395783e-06, "loss": 0.1254, "step": 2732 }, { "epoch": 0.11, "learning_rate": 9.649054994268654e-06, "loss": 0.1777, "step": 2733 }, { "epoch": 0.11, "learning_rate": 9.648921707141525e-06, "loss": 0.1296, "step": 2734 }, { "epoch": 0.11, "learning_rate": 9.648788420014396e-06, "loss": 0.1016, "step": 2735 }, { "epoch": 0.11, "learning_rate": 9.648655132887267e-06, "loss": 0.1285, "step": 2736 }, { "epoch": 0.11, "learning_rate": 9.648521845760136e-06, "loss": 0.1334, "step": 2737 }, { "epoch": 0.11, "learning_rate": 9.648388558633009e-06, "loss": 0.1241, "step": 2738 }, { "epoch": 0.11, "learning_rate": 9.648255271505878e-06, "loss": 0.1187, "step": 2739 }, { "epoch": 0.11, "learning_rate": 9.64812198437875e-06, "loss": 0.1305, "step": 2740 }, { "epoch": 0.11, "learning_rate": 9.64798869725162e-06, "loss": 0.1022, "step": 2741 }, { "epoch": 0.11, "learning_rate": 9.647855410124491e-06, "loss": 0.1209, "step": 2742 }, { "epoch": 0.11, "learning_rate": 9.647722122997362e-06, "loss": 0.1688, "step": 2743 }, { "epoch": 0.11, "learning_rate": 9.647588835870233e-06, "loss": 0.132, "step": 2744 }, { "epoch": 0.11, "learning_rate": 9.647455548743102e-06, "loss": 0.1223, "step": 2745 }, { "epoch": 0.11, "learning_rate": 9.647322261615973e-06, "loss": 0.1168, "step": 2746 }, { "epoch": 0.11, "learning_rate": 9.647188974488844e-06, "loss": 0.1173, "step": 2747 }, { "epoch": 0.11, "learning_rate": 9.647055687361715e-06, "loss": 0.1235, "step": 2748 }, { "epoch": 0.11, "learning_rate": 9.646922400234586e-06, "loss": 0.1084, "step": 2749 }, { "epoch": 0.11, "learning_rate": 9.646789113107456e-06, "loss": 0.0767, "step": 2750 }, { "epoch": 0.11, "learning_rate": 9.646655825980328e-06, "loss": 0.1402, "step": 2751 }, { "epoch": 0.11, "learning_rate": 9.646522538853198e-06, "loss": 0.1238, "step": 2752 }, { "epoch": 0.11, "learning_rate": 9.646389251726069e-06, "loss": 0.1496, "step": 2753 }, { "epoch": 0.11, "learning_rate": 9.64625596459894e-06, "loss": 0.1232, "step": 2754 }, { "epoch": 0.11, "learning_rate": 9.64612267747181e-06, "loss": 0.1352, "step": 2755 }, { "epoch": 0.11, "learning_rate": 9.645989390344682e-06, "loss": 0.0772, "step": 2756 }, { "epoch": 0.11, "learning_rate": 9.645856103217553e-06, "loss": 0.1321, "step": 2757 }, { "epoch": 0.11, "learning_rate": 9.645722816090422e-06, "loss": 0.1264, "step": 2758 }, { "epoch": 0.11, "learning_rate": 9.645589528963295e-06, "loss": 0.0939, "step": 2759 }, { "epoch": 0.11, "learning_rate": 9.645456241836164e-06, "loss": 0.1489, "step": 2760 }, { "epoch": 0.11, "learning_rate": 9.645322954709035e-06, "loss": 0.0825, "step": 2761 }, { "epoch": 0.11, "learning_rate": 9.645189667581906e-06, "loss": 0.0886, "step": 2762 }, { "epoch": 0.11, "learning_rate": 9.645056380454777e-06, "loss": 0.1005, "step": 2763 }, { "epoch": 0.11, "learning_rate": 9.644923093327648e-06, "loss": 0.1449, "step": 2764 }, { "epoch": 0.11, "learning_rate": 9.644789806200517e-06, "loss": 0.173, "step": 2765 }, { "epoch": 0.11, "learning_rate": 9.644656519073388e-06, "loss": 0.1454, "step": 2766 }, { "epoch": 0.11, "learning_rate": 9.64452323194626e-06, "loss": 0.1016, "step": 2767 }, { "epoch": 0.11, "learning_rate": 9.64438994481913e-06, "loss": 0.1079, "step": 2768 }, { "epoch": 0.11, "learning_rate": 9.644256657692001e-06, "loss": 0.1042, "step": 2769 }, { "epoch": 0.11, "learning_rate": 9.644123370564872e-06, "loss": 0.1171, "step": 2770 }, { "epoch": 0.11, "learning_rate": 9.643990083437742e-06, "loss": 0.1254, "step": 2771 }, { "epoch": 0.11, "learning_rate": 9.643856796310613e-06, "loss": 0.091, "step": 2772 }, { "epoch": 0.11, "learning_rate": 9.643723509183484e-06, "loss": 0.129, "step": 2773 }, { "epoch": 0.11, "learning_rate": 9.643590222056355e-06, "loss": 0.1071, "step": 2774 }, { "epoch": 0.11, "learning_rate": 9.643456934929226e-06, "loss": 0.1131, "step": 2775 }, { "epoch": 0.11, "learning_rate": 9.643323647802097e-06, "loss": 0.1186, "step": 2776 }, { "epoch": 0.11, "learning_rate": 9.643190360674966e-06, "loss": 0.1202, "step": 2777 }, { "epoch": 0.11, "learning_rate": 9.643057073547839e-06, "loss": 0.1051, "step": 2778 }, { "epoch": 0.11, "learning_rate": 9.642923786420708e-06, "loss": 0.1467, "step": 2779 }, { "epoch": 0.11, "learning_rate": 9.642790499293579e-06, "loss": 0.1158, "step": 2780 }, { "epoch": 0.11, "learning_rate": 9.64265721216645e-06, "loss": 0.1027, "step": 2781 }, { "epoch": 0.11, "learning_rate": 9.64252392503932e-06, "loss": 0.1438, "step": 2782 }, { "epoch": 0.11, "learning_rate": 9.642390637912192e-06, "loss": 0.1173, "step": 2783 }, { "epoch": 0.11, "learning_rate": 9.642257350785061e-06, "loss": 0.1324, "step": 2784 }, { "epoch": 0.11, "learning_rate": 9.642124063657932e-06, "loss": 0.0922, "step": 2785 }, { "epoch": 0.11, "learning_rate": 9.641990776530803e-06, "loss": 0.1327, "step": 2786 }, { "epoch": 0.11, "learning_rate": 9.641857489403674e-06, "loss": 0.1178, "step": 2787 }, { "epoch": 0.11, "learning_rate": 9.641724202276545e-06, "loss": 0.0882, "step": 2788 }, { "epoch": 0.11, "learning_rate": 9.641590915149416e-06, "loss": 0.1305, "step": 2789 }, { "epoch": 0.11, "learning_rate": 9.641457628022285e-06, "loss": 0.1492, "step": 2790 }, { "epoch": 0.11, "learning_rate": 9.641324340895158e-06, "loss": 0.1172, "step": 2791 }, { "epoch": 0.11, "learning_rate": 9.641191053768027e-06, "loss": 0.0947, "step": 2792 }, { "epoch": 0.11, "learning_rate": 9.641057766640898e-06, "loss": 0.1324, "step": 2793 }, { "epoch": 0.11, "learning_rate": 9.64092447951377e-06, "loss": 0.123, "step": 2794 }, { "epoch": 0.11, "learning_rate": 9.64079119238664e-06, "loss": 0.1284, "step": 2795 }, { "epoch": 0.11, "learning_rate": 9.640657905259511e-06, "loss": 0.1038, "step": 2796 }, { "epoch": 0.11, "learning_rate": 9.640524618132382e-06, "loss": 0.1363, "step": 2797 }, { "epoch": 0.11, "learning_rate": 9.640391331005252e-06, "loss": 0.1149, "step": 2798 }, { "epoch": 0.11, "learning_rate": 9.640258043878123e-06, "loss": 0.1627, "step": 2799 }, { "epoch": 0.11, "learning_rate": 9.640124756750994e-06, "loss": 0.088, "step": 2800 }, { "epoch": 0.11, "learning_rate": 9.639991469623865e-06, "loss": 0.158, "step": 2801 }, { "epoch": 0.11, "learning_rate": 9.639858182496736e-06, "loss": 0.1672, "step": 2802 }, { "epoch": 0.11, "learning_rate": 9.639724895369607e-06, "loss": 0.1288, "step": 2803 }, { "epoch": 0.11, "learning_rate": 9.639591608242476e-06, "loss": 0.1495, "step": 2804 }, { "epoch": 0.11, "learning_rate": 9.639458321115347e-06, "loss": 0.175, "step": 2805 }, { "epoch": 0.11, "learning_rate": 9.639325033988218e-06, "loss": 0.1299, "step": 2806 }, { "epoch": 0.11, "learning_rate": 9.639191746861089e-06, "loss": 0.0827, "step": 2807 }, { "epoch": 0.11, "learning_rate": 9.63905845973396e-06, "loss": 0.1118, "step": 2808 }, { "epoch": 0.11, "learning_rate": 9.63892517260683e-06, "loss": 0.1092, "step": 2809 }, { "epoch": 0.11, "learning_rate": 9.638791885479702e-06, "loss": 0.1561, "step": 2810 }, { "epoch": 0.11, "learning_rate": 9.638658598352571e-06, "loss": 0.1331, "step": 2811 }, { "epoch": 0.11, "learning_rate": 9.638525311225442e-06, "loss": 0.1267, "step": 2812 }, { "epoch": 0.11, "learning_rate": 9.638392024098313e-06, "loss": 0.0908, "step": 2813 }, { "epoch": 0.11, "learning_rate": 9.638258736971184e-06, "loss": 0.1394, "step": 2814 }, { "epoch": 0.11, "learning_rate": 9.638125449844055e-06, "loss": 0.1391, "step": 2815 }, { "epoch": 0.11, "learning_rate": 9.637992162716926e-06, "loss": 0.1696, "step": 2816 }, { "epoch": 0.11, "learning_rate": 9.637858875589795e-06, "loss": 0.1505, "step": 2817 }, { "epoch": 0.11, "learning_rate": 9.637725588462668e-06, "loss": 0.1061, "step": 2818 }, { "epoch": 0.11, "learning_rate": 9.637592301335537e-06, "loss": 0.138, "step": 2819 }, { "epoch": 0.11, "learning_rate": 9.637459014208408e-06, "loss": 0.1189, "step": 2820 }, { "epoch": 0.11, "learning_rate": 9.63732572708128e-06, "loss": 0.0893, "step": 2821 }, { "epoch": 0.11, "learning_rate": 9.63719243995415e-06, "loss": 0.1202, "step": 2822 }, { "epoch": 0.11, "learning_rate": 9.637059152827021e-06, "loss": 0.1077, "step": 2823 }, { "epoch": 0.11, "learning_rate": 9.63692586569989e-06, "loss": 0.1494, "step": 2824 }, { "epoch": 0.11, "learning_rate": 9.636792578572762e-06, "loss": 0.1354, "step": 2825 }, { "epoch": 0.11, "learning_rate": 9.636659291445633e-06, "loss": 0.1206, "step": 2826 }, { "epoch": 0.11, "learning_rate": 9.636526004318504e-06, "loss": 0.093, "step": 2827 }, { "epoch": 0.11, "learning_rate": 9.636392717191375e-06, "loss": 0.1022, "step": 2828 }, { "epoch": 0.11, "learning_rate": 9.636259430064246e-06, "loss": 0.1132, "step": 2829 }, { "epoch": 0.11, "learning_rate": 9.636126142937115e-06, "loss": 0.1367, "step": 2830 }, { "epoch": 0.11, "learning_rate": 9.635992855809988e-06, "loss": 0.1197, "step": 2831 }, { "epoch": 0.11, "learning_rate": 9.635859568682857e-06, "loss": 0.1003, "step": 2832 }, { "epoch": 0.11, "learning_rate": 9.635726281555728e-06, "loss": 0.1183, "step": 2833 }, { "epoch": 0.11, "learning_rate": 9.635592994428599e-06, "loss": 0.1273, "step": 2834 }, { "epoch": 0.11, "learning_rate": 9.63545970730147e-06, "loss": 0.1134, "step": 2835 }, { "epoch": 0.11, "learning_rate": 9.635326420174341e-06, "loss": 0.1863, "step": 2836 }, { "epoch": 0.11, "learning_rate": 9.635193133047212e-06, "loss": 0.128, "step": 2837 }, { "epoch": 0.11, "learning_rate": 9.635059845920081e-06, "loss": 0.136, "step": 2838 }, { "epoch": 0.11, "learning_rate": 9.634926558792952e-06, "loss": 0.1043, "step": 2839 }, { "epoch": 0.11, "learning_rate": 9.634793271665823e-06, "loss": 0.1425, "step": 2840 }, { "epoch": 0.11, "learning_rate": 9.634659984538694e-06, "loss": 0.1012, "step": 2841 }, { "epoch": 0.11, "learning_rate": 9.634526697411565e-06, "loss": 0.1192, "step": 2842 }, { "epoch": 0.11, "learning_rate": 9.634393410284435e-06, "loss": 0.1174, "step": 2843 }, { "epoch": 0.11, "learning_rate": 9.634260123157306e-06, "loss": 0.1288, "step": 2844 }, { "epoch": 0.11, "learning_rate": 9.634126836030177e-06, "loss": 0.1116, "step": 2845 }, { "epoch": 0.11, "learning_rate": 9.633993548903048e-06, "loss": 0.1165, "step": 2846 }, { "epoch": 0.11, "learning_rate": 9.633860261775919e-06, "loss": 0.1049, "step": 2847 }, { "epoch": 0.11, "learning_rate": 9.63372697464879e-06, "loss": 0.1573, "step": 2848 }, { "epoch": 0.11, "learning_rate": 9.633593687521659e-06, "loss": 0.1338, "step": 2849 }, { "epoch": 0.11, "learning_rate": 9.633460400394532e-06, "loss": 0.1057, "step": 2850 }, { "epoch": 0.11, "learning_rate": 9.6333271132674e-06, "loss": 0.0895, "step": 2851 }, { "epoch": 0.11, "learning_rate": 9.633193826140272e-06, "loss": 0.1103, "step": 2852 }, { "epoch": 0.11, "learning_rate": 9.633060539013143e-06, "loss": 0.097, "step": 2853 }, { "epoch": 0.11, "learning_rate": 9.632927251886014e-06, "loss": 0.1374, "step": 2854 }, { "epoch": 0.11, "learning_rate": 9.632793964758885e-06, "loss": 0.1261, "step": 2855 }, { "epoch": 0.11, "learning_rate": 9.632660677631756e-06, "loss": 0.11, "step": 2856 }, { "epoch": 0.11, "learning_rate": 9.632527390504625e-06, "loss": 0.1253, "step": 2857 }, { "epoch": 0.11, "learning_rate": 9.632394103377498e-06, "loss": 0.1421, "step": 2858 }, { "epoch": 0.11, "learning_rate": 9.632260816250367e-06, "loss": 0.0851, "step": 2859 }, { "epoch": 0.11, "learning_rate": 9.632127529123238e-06, "loss": 0.1087, "step": 2860 }, { "epoch": 0.11, "learning_rate": 9.631994241996109e-06, "loss": 0.1146, "step": 2861 }, { "epoch": 0.11, "learning_rate": 9.63186095486898e-06, "loss": 0.1025, "step": 2862 }, { "epoch": 0.11, "learning_rate": 9.631727667741851e-06, "loss": 0.0913, "step": 2863 }, { "epoch": 0.11, "learning_rate": 9.63159438061472e-06, "loss": 0.1189, "step": 2864 }, { "epoch": 0.11, "learning_rate": 9.631461093487591e-06, "loss": 0.1468, "step": 2865 }, { "epoch": 0.11, "learning_rate": 9.631327806360462e-06, "loss": 0.108, "step": 2866 }, { "epoch": 0.11, "learning_rate": 9.631194519233333e-06, "loss": 0.0933, "step": 2867 }, { "epoch": 0.11, "learning_rate": 9.631061232106204e-06, "loss": 0.1506, "step": 2868 }, { "epoch": 0.11, "learning_rate": 9.630927944979075e-06, "loss": 0.1116, "step": 2869 }, { "epoch": 0.11, "learning_rate": 9.630794657851945e-06, "loss": 0.119, "step": 2870 }, { "epoch": 0.11, "learning_rate": 9.630661370724816e-06, "loss": 0.1353, "step": 2871 }, { "epoch": 0.11, "learning_rate": 9.630528083597687e-06, "loss": 0.1064, "step": 2872 }, { "epoch": 0.11, "learning_rate": 9.630394796470558e-06, "loss": 0.1204, "step": 2873 }, { "epoch": 0.11, "learning_rate": 9.630261509343429e-06, "loss": 0.1704, "step": 2874 }, { "epoch": 0.11, "learning_rate": 9.6301282222163e-06, "loss": 0.1378, "step": 2875 }, { "epoch": 0.11, "learning_rate": 9.629994935089169e-06, "loss": 0.124, "step": 2876 }, { "epoch": 0.11, "learning_rate": 9.629861647962042e-06, "loss": 0.1071, "step": 2877 }, { "epoch": 0.11, "learning_rate": 9.629728360834911e-06, "loss": 0.1125, "step": 2878 }, { "epoch": 0.11, "learning_rate": 9.629595073707782e-06, "loss": 0.1344, "step": 2879 }, { "epoch": 0.12, "learning_rate": 9.629461786580653e-06, "loss": 0.1089, "step": 2880 }, { "epoch": 0.12, "learning_rate": 9.629328499453524e-06, "loss": 0.128, "step": 2881 }, { "epoch": 0.12, "learning_rate": 9.629195212326395e-06, "loss": 0.1271, "step": 2882 }, { "epoch": 0.12, "learning_rate": 9.629061925199264e-06, "loss": 0.1379, "step": 2883 }, { "epoch": 0.12, "learning_rate": 9.628928638072135e-06, "loss": 0.1077, "step": 2884 }, { "epoch": 0.12, "learning_rate": 9.628795350945006e-06, "loss": 0.1254, "step": 2885 }, { "epoch": 0.12, "learning_rate": 9.628662063817877e-06, "loss": 0.141, "step": 2886 }, { "epoch": 0.12, "learning_rate": 9.628528776690748e-06, "loss": 0.1176, "step": 2887 }, { "epoch": 0.12, "learning_rate": 9.62839548956362e-06, "loss": 0.1297, "step": 2888 }, { "epoch": 0.12, "learning_rate": 9.628262202436488e-06, "loss": 0.1213, "step": 2889 }, { "epoch": 0.12, "learning_rate": 9.628128915309361e-06, "loss": 0.1248, "step": 2890 }, { "epoch": 0.12, "learning_rate": 9.62799562818223e-06, "loss": 0.1116, "step": 2891 }, { "epoch": 0.12, "learning_rate": 9.627862341055101e-06, "loss": 0.1541, "step": 2892 }, { "epoch": 0.12, "learning_rate": 9.627729053927972e-06, "loss": 0.1277, "step": 2893 }, { "epoch": 0.12, "learning_rate": 9.627595766800843e-06, "loss": 0.1285, "step": 2894 }, { "epoch": 0.12, "learning_rate": 9.627462479673714e-06, "loss": 0.1374, "step": 2895 }, { "epoch": 0.12, "learning_rate": 9.627329192546585e-06, "loss": 0.1329, "step": 2896 }, { "epoch": 0.12, "learning_rate": 9.627195905419455e-06, "loss": 0.1248, "step": 2897 }, { "epoch": 0.12, "learning_rate": 9.627062618292327e-06, "loss": 0.0983, "step": 2898 }, { "epoch": 0.12, "learning_rate": 9.626929331165197e-06, "loss": 0.14, "step": 2899 }, { "epoch": 0.12, "learning_rate": 9.626796044038068e-06, "loss": 0.0901, "step": 2900 }, { "epoch": 0.12, "learning_rate": 9.626662756910939e-06, "loss": 0.1171, "step": 2901 }, { "epoch": 0.12, "learning_rate": 9.626529469783808e-06, "loss": 0.224, "step": 2902 }, { "epoch": 0.12, "learning_rate": 9.62639618265668e-06, "loss": 0.1487, "step": 2903 }, { "epoch": 0.12, "learning_rate": 9.62626289552955e-06, "loss": 0.1531, "step": 2904 }, { "epoch": 0.12, "learning_rate": 9.626129608402421e-06, "loss": 0.1089, "step": 2905 }, { "epoch": 0.12, "learning_rate": 9.625996321275292e-06, "loss": 0.1711, "step": 2906 }, { "epoch": 0.12, "learning_rate": 9.625863034148163e-06, "loss": 0.1259, "step": 2907 }, { "epoch": 0.12, "learning_rate": 9.625729747021034e-06, "loss": 0.0944, "step": 2908 }, { "epoch": 0.12, "learning_rate": 9.625596459893905e-06, "loss": 0.087, "step": 2909 }, { "epoch": 0.12, "learning_rate": 9.625463172766774e-06, "loss": 0.1201, "step": 2910 }, { "epoch": 0.12, "learning_rate": 9.625329885639645e-06, "loss": 0.1111, "step": 2911 }, { "epoch": 0.12, "learning_rate": 9.625196598512516e-06, "loss": 0.1118, "step": 2912 }, { "epoch": 0.12, "learning_rate": 9.625063311385387e-06, "loss": 0.1692, "step": 2913 }, { "epoch": 0.12, "learning_rate": 9.624930024258258e-06, "loss": 0.1174, "step": 2914 }, { "epoch": 0.12, "learning_rate": 9.62479673713113e-06, "loss": 0.1211, "step": 2915 }, { "epoch": 0.12, "learning_rate": 9.624663450003999e-06, "loss": 0.1655, "step": 2916 }, { "epoch": 0.12, "learning_rate": 9.624530162876871e-06, "loss": 0.1288, "step": 2917 }, { "epoch": 0.12, "learning_rate": 9.62439687574974e-06, "loss": 0.0905, "step": 2918 }, { "epoch": 0.12, "learning_rate": 9.624263588622612e-06, "loss": 0.1103, "step": 2919 }, { "epoch": 0.12, "learning_rate": 9.624130301495483e-06, "loss": 0.1142, "step": 2920 }, { "epoch": 0.12, "learning_rate": 9.623997014368352e-06, "loss": 0.0854, "step": 2921 }, { "epoch": 0.12, "learning_rate": 9.623863727241225e-06, "loss": 0.1133, "step": 2922 }, { "epoch": 0.12, "learning_rate": 9.623730440114094e-06, "loss": 0.0765, "step": 2923 }, { "epoch": 0.12, "learning_rate": 9.623597152986965e-06, "loss": 0.1108, "step": 2924 }, { "epoch": 0.12, "learning_rate": 9.623463865859836e-06, "loss": 0.1331, "step": 2925 }, { "epoch": 0.12, "learning_rate": 9.623330578732707e-06, "loss": 0.1208, "step": 2926 }, { "epoch": 0.12, "learning_rate": 9.623197291605578e-06, "loss": 0.1389, "step": 2927 }, { "epoch": 0.12, "learning_rate": 9.623064004478449e-06, "loss": 0.0961, "step": 2928 }, { "epoch": 0.12, "learning_rate": 9.622930717351318e-06, "loss": 0.1349, "step": 2929 }, { "epoch": 0.12, "learning_rate": 9.62279743022419e-06, "loss": 0.1153, "step": 2930 }, { "epoch": 0.12, "learning_rate": 9.62266414309706e-06, "loss": 0.1155, "step": 2931 }, { "epoch": 0.12, "learning_rate": 9.622530855969931e-06, "loss": 0.1419, "step": 2932 }, { "epoch": 0.12, "learning_rate": 9.622397568842802e-06, "loss": 0.124, "step": 2933 }, { "epoch": 0.12, "learning_rate": 9.622264281715673e-06, "loss": 0.1126, "step": 2934 }, { "epoch": 0.12, "learning_rate": 9.622130994588544e-06, "loss": 0.1371, "step": 2935 }, { "epoch": 0.12, "learning_rate": 9.621997707461415e-06, "loss": 0.1041, "step": 2936 }, { "epoch": 0.12, "learning_rate": 9.621864420334284e-06, "loss": 0.1171, "step": 2937 }, { "epoch": 0.12, "learning_rate": 9.621731133207155e-06, "loss": 0.1332, "step": 2938 }, { "epoch": 0.12, "learning_rate": 9.621597846080026e-06, "loss": 0.154, "step": 2939 }, { "epoch": 0.12, "learning_rate": 9.621464558952897e-06, "loss": 0.1084, "step": 2940 }, { "epoch": 0.12, "learning_rate": 9.621331271825768e-06, "loss": 0.1305, "step": 2941 }, { "epoch": 0.12, "learning_rate": 9.621197984698638e-06, "loss": 0.1557, "step": 2942 }, { "epoch": 0.12, "learning_rate": 9.621064697571509e-06, "loss": 0.1039, "step": 2943 }, { "epoch": 0.12, "learning_rate": 9.62093141044438e-06, "loss": 0.1648, "step": 2944 }, { "epoch": 0.12, "learning_rate": 9.62079812331725e-06, "loss": 0.1242, "step": 2945 }, { "epoch": 0.12, "learning_rate": 9.620664836190122e-06, "loss": 0.0929, "step": 2946 }, { "epoch": 0.12, "learning_rate": 9.620531549062993e-06, "loss": 0.1499, "step": 2947 }, { "epoch": 0.12, "learning_rate": 9.620398261935862e-06, "loss": 0.1124, "step": 2948 }, { "epoch": 0.12, "learning_rate": 9.620264974808735e-06, "loss": 0.1062, "step": 2949 }, { "epoch": 0.12, "learning_rate": 9.620131687681604e-06, "loss": 0.1652, "step": 2950 }, { "epoch": 0.12, "learning_rate": 9.619998400554475e-06, "loss": 0.0926, "step": 2951 }, { "epoch": 0.12, "learning_rate": 9.619865113427346e-06, "loss": 0.1388, "step": 2952 }, { "epoch": 0.12, "learning_rate": 9.619731826300217e-06, "loss": 0.1356, "step": 2953 }, { "epoch": 0.12, "learning_rate": 9.619598539173088e-06, "loss": 0.1352, "step": 2954 }, { "epoch": 0.12, "learning_rate": 9.619465252045959e-06, "loss": 0.1019, "step": 2955 }, { "epoch": 0.12, "learning_rate": 9.619331964918828e-06, "loss": 0.1471, "step": 2956 }, { "epoch": 0.12, "learning_rate": 9.619198677791701e-06, "loss": 0.111, "step": 2957 }, { "epoch": 0.12, "learning_rate": 9.61906539066457e-06, "loss": 0.092, "step": 2958 }, { "epoch": 0.12, "learning_rate": 9.618932103537441e-06, "loss": 0.132, "step": 2959 }, { "epoch": 0.12, "learning_rate": 9.618798816410312e-06, "loss": 0.0965, "step": 2960 }, { "epoch": 0.12, "learning_rate": 9.618665529283181e-06, "loss": 0.1689, "step": 2961 }, { "epoch": 0.12, "learning_rate": 9.618532242156054e-06, "loss": 0.1282, "step": 2962 }, { "epoch": 0.12, "learning_rate": 9.618398955028923e-06, "loss": 0.1575, "step": 2963 }, { "epoch": 0.12, "learning_rate": 9.618265667901794e-06, "loss": 0.1075, "step": 2964 }, { "epoch": 0.12, "learning_rate": 9.618132380774665e-06, "loss": 0.0909, "step": 2965 }, { "epoch": 0.12, "learning_rate": 9.617999093647536e-06, "loss": 0.1409, "step": 2966 }, { "epoch": 0.12, "learning_rate": 9.617865806520407e-06, "loss": 0.1043, "step": 2967 }, { "epoch": 0.12, "learning_rate": 9.617732519393278e-06, "loss": 0.14, "step": 2968 }, { "epoch": 0.12, "learning_rate": 9.617599232266148e-06, "loss": 0.0988, "step": 2969 }, { "epoch": 0.12, "learning_rate": 9.617465945139019e-06, "loss": 0.1071, "step": 2970 }, { "epoch": 0.12, "learning_rate": 9.61733265801189e-06, "loss": 0.0905, "step": 2971 }, { "epoch": 0.12, "learning_rate": 9.61719937088476e-06, "loss": 0.1128, "step": 2972 }, { "epoch": 0.12, "learning_rate": 9.617066083757632e-06, "loss": 0.1511, "step": 2973 }, { "epoch": 0.12, "learning_rate": 9.616932796630503e-06, "loss": 0.1291, "step": 2974 }, { "epoch": 0.12, "learning_rate": 9.616799509503372e-06, "loss": 0.1051, "step": 2975 }, { "epoch": 0.12, "learning_rate": 9.616666222376245e-06, "loss": 0.1081, "step": 2976 }, { "epoch": 0.12, "learning_rate": 9.616532935249114e-06, "loss": 0.1059, "step": 2977 }, { "epoch": 0.12, "learning_rate": 9.616399648121985e-06, "loss": 0.1569, "step": 2978 }, { "epoch": 0.12, "learning_rate": 9.616266360994856e-06, "loss": 0.0989, "step": 2979 }, { "epoch": 0.12, "learning_rate": 9.616133073867725e-06, "loss": 0.113, "step": 2980 }, { "epoch": 0.12, "learning_rate": 9.615999786740598e-06, "loss": 0.0901, "step": 2981 }, { "epoch": 0.12, "learning_rate": 9.615866499613467e-06, "loss": 0.1136, "step": 2982 }, { "epoch": 0.12, "learning_rate": 9.615733212486338e-06, "loss": 0.1386, "step": 2983 }, { "epoch": 0.12, "learning_rate": 9.61559992535921e-06, "loss": 0.1315, "step": 2984 }, { "epoch": 0.12, "learning_rate": 9.61546663823208e-06, "loss": 0.1106, "step": 2985 }, { "epoch": 0.12, "learning_rate": 9.615333351104951e-06, "loss": 0.1306, "step": 2986 }, { "epoch": 0.12, "learning_rate": 9.615200063977822e-06, "loss": 0.1254, "step": 2987 }, { "epoch": 0.12, "learning_rate": 9.615066776850692e-06, "loss": 0.1567, "step": 2988 }, { "epoch": 0.12, "learning_rate": 9.614933489723564e-06, "loss": 0.1245, "step": 2989 }, { "epoch": 0.12, "learning_rate": 9.614800202596434e-06, "loss": 0.1418, "step": 2990 }, { "epoch": 0.12, "learning_rate": 9.614666915469305e-06, "loss": 0.1258, "step": 2991 }, { "epoch": 0.12, "learning_rate": 9.614533628342176e-06, "loss": 0.1067, "step": 2992 }, { "epoch": 0.12, "learning_rate": 9.614400341215047e-06, "loss": 0.106, "step": 2993 }, { "epoch": 0.12, "learning_rate": 9.614267054087917e-06, "loss": 0.0893, "step": 2994 }, { "epoch": 0.12, "learning_rate": 9.614133766960788e-06, "loss": 0.106, "step": 2995 }, { "epoch": 0.12, "learning_rate": 9.614000479833658e-06, "loss": 0.1508, "step": 2996 }, { "epoch": 0.12, "learning_rate": 9.61386719270653e-06, "loss": 0.1488, "step": 2997 }, { "epoch": 0.12, "learning_rate": 9.6137339055794e-06, "loss": 0.1498, "step": 2998 }, { "epoch": 0.12, "learning_rate": 9.61360061845227e-06, "loss": 0.1165, "step": 2999 }, { "epoch": 0.12, "learning_rate": 9.613467331325142e-06, "loss": 0.1083, "step": 3000 }, { "epoch": 0.12, "learning_rate": 9.613334044198011e-06, "loss": 0.1203, "step": 3001 }, { "epoch": 0.12, "learning_rate": 9.613200757070884e-06, "loss": 0.1534, "step": 3002 }, { "epoch": 0.12, "learning_rate": 9.613067469943753e-06, "loss": 0.0943, "step": 3003 }, { "epoch": 0.12, "learning_rate": 9.612934182816624e-06, "loss": 0.1143, "step": 3004 }, { "epoch": 0.12, "learning_rate": 9.612800895689495e-06, "loss": 0.1242, "step": 3005 }, { "epoch": 0.12, "learning_rate": 9.612667608562366e-06, "loss": 0.0931, "step": 3006 }, { "epoch": 0.12, "learning_rate": 9.612534321435237e-06, "loss": 0.1048, "step": 3007 }, { "epoch": 0.12, "learning_rate": 9.612401034308108e-06, "loss": 0.1184, "step": 3008 }, { "epoch": 0.12, "learning_rate": 9.612267747180977e-06, "loss": 0.1318, "step": 3009 }, { "epoch": 0.12, "learning_rate": 9.612134460053848e-06, "loss": 0.1338, "step": 3010 }, { "epoch": 0.12, "learning_rate": 9.61200117292672e-06, "loss": 0.1125, "step": 3011 }, { "epoch": 0.12, "learning_rate": 9.61186788579959e-06, "loss": 0.1178, "step": 3012 }, { "epoch": 0.12, "learning_rate": 9.611734598672461e-06, "loss": 0.0959, "step": 3013 }, { "epoch": 0.12, "learning_rate": 9.611601311545332e-06, "loss": 0.1114, "step": 3014 }, { "epoch": 0.12, "learning_rate": 9.611468024418202e-06, "loss": 0.1396, "step": 3015 }, { "epoch": 0.12, "learning_rate": 9.611334737291074e-06, "loss": 0.1316, "step": 3016 }, { "epoch": 0.12, "learning_rate": 9.611201450163944e-06, "loss": 0.1005, "step": 3017 }, { "epoch": 0.12, "learning_rate": 9.611068163036815e-06, "loss": 0.1285, "step": 3018 }, { "epoch": 0.12, "learning_rate": 9.610934875909686e-06, "loss": 0.1133, "step": 3019 }, { "epoch": 0.12, "learning_rate": 9.610801588782555e-06, "loss": 0.1083, "step": 3020 }, { "epoch": 0.12, "learning_rate": 9.610668301655428e-06, "loss": 0.1527, "step": 3021 }, { "epoch": 0.12, "learning_rate": 9.610535014528297e-06, "loss": 0.0856, "step": 3022 }, { "epoch": 0.12, "learning_rate": 9.610401727401168e-06, "loss": 0.1393, "step": 3023 }, { "epoch": 0.12, "learning_rate": 9.610268440274039e-06, "loss": 0.1366, "step": 3024 }, { "epoch": 0.12, "learning_rate": 9.61013515314691e-06, "loss": 0.0786, "step": 3025 }, { "epoch": 0.12, "learning_rate": 9.610001866019781e-06, "loss": 0.119, "step": 3026 }, { "epoch": 0.12, "learning_rate": 9.609868578892652e-06, "loss": 0.1305, "step": 3027 }, { "epoch": 0.12, "learning_rate": 9.609735291765521e-06, "loss": 0.1516, "step": 3028 }, { "epoch": 0.12, "learning_rate": 9.609602004638394e-06, "loss": 0.1466, "step": 3029 }, { "epoch": 0.12, "learning_rate": 9.609468717511263e-06, "loss": 0.1222, "step": 3030 }, { "epoch": 0.12, "learning_rate": 9.609335430384134e-06, "loss": 0.1237, "step": 3031 }, { "epoch": 0.12, "learning_rate": 9.609202143257005e-06, "loss": 0.1063, "step": 3032 }, { "epoch": 0.12, "learning_rate": 9.609068856129876e-06, "loss": 0.1089, "step": 3033 }, { "epoch": 0.12, "learning_rate": 9.608935569002747e-06, "loss": 0.1411, "step": 3034 }, { "epoch": 0.12, "learning_rate": 9.608802281875618e-06, "loss": 0.1027, "step": 3035 }, { "epoch": 0.12, "learning_rate": 9.608668994748487e-06, "loss": 0.0898, "step": 3036 }, { "epoch": 0.12, "learning_rate": 9.608535707621358e-06, "loss": 0.115, "step": 3037 }, { "epoch": 0.12, "learning_rate": 9.60840242049423e-06, "loss": 0.1595, "step": 3038 }, { "epoch": 0.12, "learning_rate": 9.6082691333671e-06, "loss": 0.1269, "step": 3039 }, { "epoch": 0.12, "learning_rate": 9.608135846239971e-06, "loss": 0.1364, "step": 3040 }, { "epoch": 0.12, "learning_rate": 9.60800255911284e-06, "loss": 0.1365, "step": 3041 }, { "epoch": 0.12, "learning_rate": 9.607869271985712e-06, "loss": 0.1125, "step": 3042 }, { "epoch": 0.12, "learning_rate": 9.607735984858583e-06, "loss": 0.1864, "step": 3043 }, { "epoch": 0.12, "learning_rate": 9.607602697731454e-06, "loss": 0.1531, "step": 3044 }, { "epoch": 0.12, "learning_rate": 9.607469410604325e-06, "loss": 0.1296, "step": 3045 }, { "epoch": 0.12, "learning_rate": 9.607336123477196e-06, "loss": 0.1246, "step": 3046 }, { "epoch": 0.12, "learning_rate": 9.607202836350065e-06, "loss": 0.1237, "step": 3047 }, { "epoch": 0.12, "learning_rate": 9.607069549222938e-06, "loss": 0.1325, "step": 3048 }, { "epoch": 0.12, "learning_rate": 9.606936262095807e-06, "loss": 0.1513, "step": 3049 }, { "epoch": 0.12, "learning_rate": 9.606802974968678e-06, "loss": 0.1269, "step": 3050 }, { "epoch": 0.12, "learning_rate": 9.606669687841549e-06, "loss": 0.1262, "step": 3051 }, { "epoch": 0.12, "learning_rate": 9.60653640071442e-06, "loss": 0.1192, "step": 3052 }, { "epoch": 0.12, "learning_rate": 9.606403113587291e-06, "loss": 0.1227, "step": 3053 }, { "epoch": 0.12, "learning_rate": 9.606269826460162e-06, "loss": 0.1167, "step": 3054 }, { "epoch": 0.12, "learning_rate": 9.606136539333031e-06, "loss": 0.1288, "step": 3055 }, { "epoch": 0.12, "learning_rate": 9.606003252205904e-06, "loss": 0.1412, "step": 3056 }, { "epoch": 0.12, "learning_rate": 9.605869965078773e-06, "loss": 0.1048, "step": 3057 }, { "epoch": 0.12, "learning_rate": 9.605736677951644e-06, "loss": 0.2186, "step": 3058 }, { "epoch": 0.12, "learning_rate": 9.605603390824515e-06, "loss": 0.1774, "step": 3059 }, { "epoch": 0.12, "learning_rate": 9.605470103697385e-06, "loss": 0.1419, "step": 3060 }, { "epoch": 0.12, "learning_rate": 9.605336816570257e-06, "loss": 0.0833, "step": 3061 }, { "epoch": 0.12, "learning_rate": 9.605203529443127e-06, "loss": 0.1468, "step": 3062 }, { "epoch": 0.12, "learning_rate": 9.605070242315997e-06, "loss": 0.1259, "step": 3063 }, { "epoch": 0.12, "learning_rate": 9.604936955188868e-06, "loss": 0.1157, "step": 3064 }, { "epoch": 0.12, "learning_rate": 9.60480366806174e-06, "loss": 0.116, "step": 3065 }, { "epoch": 0.12, "learning_rate": 9.60467038093461e-06, "loss": 0.1213, "step": 3066 }, { "epoch": 0.12, "learning_rate": 9.604537093807481e-06, "loss": 0.132, "step": 3067 }, { "epoch": 0.12, "learning_rate": 9.60440380668035e-06, "loss": 0.1215, "step": 3068 }, { "epoch": 0.12, "learning_rate": 9.604270519553223e-06, "loss": 0.141, "step": 3069 }, { "epoch": 0.12, "learning_rate": 9.604137232426093e-06, "loss": 0.1288, "step": 3070 }, { "epoch": 0.12, "learning_rate": 9.604003945298964e-06, "loss": 0.1604, "step": 3071 }, { "epoch": 0.12, "learning_rate": 9.603870658171835e-06, "loss": 0.1711, "step": 3072 }, { "epoch": 0.12, "learning_rate": 9.603737371044706e-06, "loss": 0.1262, "step": 3073 }, { "epoch": 0.12, "learning_rate": 9.603604083917577e-06, "loss": 0.0948, "step": 3074 }, { "epoch": 0.12, "learning_rate": 9.603470796790448e-06, "loss": 0.1259, "step": 3075 }, { "epoch": 0.12, "learning_rate": 9.603337509663317e-06, "loss": 0.2001, "step": 3076 }, { "epoch": 0.12, "learning_rate": 9.603204222536188e-06, "loss": 0.137, "step": 3077 }, { "epoch": 0.12, "learning_rate": 9.603070935409059e-06, "loss": 0.1197, "step": 3078 }, { "epoch": 0.12, "learning_rate": 9.60293764828193e-06, "loss": 0.0853, "step": 3079 }, { "epoch": 0.12, "learning_rate": 9.602804361154801e-06, "loss": 0.1417, "step": 3080 }, { "epoch": 0.12, "learning_rate": 9.60267107402767e-06, "loss": 0.0949, "step": 3081 }, { "epoch": 0.12, "learning_rate": 9.602537786900541e-06, "loss": 0.1293, "step": 3082 }, { "epoch": 0.12, "learning_rate": 9.602404499773412e-06, "loss": 0.1375, "step": 3083 }, { "epoch": 0.12, "learning_rate": 9.602271212646283e-06, "loss": 0.1027, "step": 3084 }, { "epoch": 0.12, "learning_rate": 9.602137925519154e-06, "loss": 0.1092, "step": 3085 }, { "epoch": 0.12, "learning_rate": 9.602004638392025e-06, "loss": 0.1382, "step": 3086 }, { "epoch": 0.12, "learning_rate": 9.601871351264895e-06, "loss": 0.1086, "step": 3087 }, { "epoch": 0.12, "learning_rate": 9.601738064137767e-06, "loss": 0.1307, "step": 3088 }, { "epoch": 0.12, "learning_rate": 9.601604777010637e-06, "loss": 0.125, "step": 3089 }, { "epoch": 0.12, "learning_rate": 9.601471489883508e-06, "loss": 0.1377, "step": 3090 }, { "epoch": 0.12, "learning_rate": 9.601338202756379e-06, "loss": 0.1252, "step": 3091 }, { "epoch": 0.12, "learning_rate": 9.60120491562925e-06, "loss": 0.1165, "step": 3092 }, { "epoch": 0.12, "learning_rate": 9.60107162850212e-06, "loss": 0.0851, "step": 3093 }, { "epoch": 0.12, "learning_rate": 9.600938341374992e-06, "loss": 0.1233, "step": 3094 }, { "epoch": 0.12, "learning_rate": 9.600805054247861e-06, "loss": 0.1212, "step": 3095 }, { "epoch": 0.12, "learning_rate": 9.600671767120734e-06, "loss": 0.0971, "step": 3096 }, { "epoch": 0.12, "learning_rate": 9.600538479993603e-06, "loss": 0.1253, "step": 3097 }, { "epoch": 0.12, "learning_rate": 9.600405192866474e-06, "loss": 0.1358, "step": 3098 }, { "epoch": 0.12, "learning_rate": 9.600271905739345e-06, "loss": 0.1462, "step": 3099 }, { "epoch": 0.12, "learning_rate": 9.600138618612214e-06, "loss": 0.1179, "step": 3100 }, { "epoch": 0.12, "learning_rate": 9.600005331485087e-06, "loss": 0.1453, "step": 3101 }, { "epoch": 0.12, "learning_rate": 9.599872044357956e-06, "loss": 0.071, "step": 3102 }, { "epoch": 0.12, "learning_rate": 9.599738757230827e-06, "loss": 0.1156, "step": 3103 }, { "epoch": 0.12, "learning_rate": 9.599605470103698e-06, "loss": 0.144, "step": 3104 }, { "epoch": 0.12, "learning_rate": 9.599472182976569e-06, "loss": 0.0823, "step": 3105 }, { "epoch": 0.12, "learning_rate": 9.59933889584944e-06, "loss": 0.1396, "step": 3106 }, { "epoch": 0.12, "learning_rate": 9.599205608722311e-06, "loss": 0.1883, "step": 3107 }, { "epoch": 0.12, "learning_rate": 9.59907232159518e-06, "loss": 0.0852, "step": 3108 }, { "epoch": 0.12, "learning_rate": 9.598939034468051e-06, "loss": 0.1282, "step": 3109 }, { "epoch": 0.12, "learning_rate": 9.598805747340922e-06, "loss": 0.1114, "step": 3110 }, { "epoch": 0.12, "learning_rate": 9.598672460213793e-06, "loss": 0.1009, "step": 3111 }, { "epoch": 0.12, "learning_rate": 9.598539173086664e-06, "loss": 0.1226, "step": 3112 }, { "epoch": 0.12, "learning_rate": 9.598405885959535e-06, "loss": 0.112, "step": 3113 }, { "epoch": 0.12, "learning_rate": 9.598272598832405e-06, "loss": 0.1071, "step": 3114 }, { "epoch": 0.12, "learning_rate": 9.598139311705277e-06, "loss": 0.1568, "step": 3115 }, { "epoch": 0.12, "learning_rate": 9.598006024578147e-06, "loss": 0.1239, "step": 3116 }, { "epoch": 0.12, "learning_rate": 9.597872737451018e-06, "loss": 0.1292, "step": 3117 }, { "epoch": 0.12, "learning_rate": 9.597739450323889e-06, "loss": 0.1417, "step": 3118 }, { "epoch": 0.12, "learning_rate": 9.597606163196758e-06, "loss": 0.0819, "step": 3119 }, { "epoch": 0.12, "learning_rate": 9.59747287606963e-06, "loss": 0.1207, "step": 3120 }, { "epoch": 0.12, "learning_rate": 9.5973395889425e-06, "loss": 0.0945, "step": 3121 }, { "epoch": 0.12, "learning_rate": 9.597206301815371e-06, "loss": 0.0917, "step": 3122 }, { "epoch": 0.12, "learning_rate": 9.597073014688242e-06, "loss": 0.1218, "step": 3123 }, { "epoch": 0.12, "learning_rate": 9.596939727561113e-06, "loss": 0.1144, "step": 3124 }, { "epoch": 0.12, "learning_rate": 9.596806440433984e-06, "loss": 0.11, "step": 3125 }, { "epoch": 0.12, "learning_rate": 9.596673153306855e-06, "loss": 0.1114, "step": 3126 }, { "epoch": 0.12, "learning_rate": 9.596539866179724e-06, "loss": 0.0925, "step": 3127 }, { "epoch": 0.12, "learning_rate": 9.596406579052597e-06, "loss": 0.1331, "step": 3128 }, { "epoch": 0.12, "learning_rate": 9.596273291925466e-06, "loss": 0.111, "step": 3129 }, { "epoch": 0.12, "learning_rate": 9.596140004798337e-06, "loss": 0.1327, "step": 3130 }, { "epoch": 0.13, "learning_rate": 9.596006717671208e-06, "loss": 0.1006, "step": 3131 }, { "epoch": 0.13, "learning_rate": 9.59587343054408e-06, "loss": 0.1264, "step": 3132 }, { "epoch": 0.13, "learning_rate": 9.59574014341695e-06, "loss": 0.1344, "step": 3133 }, { "epoch": 0.13, "learning_rate": 9.595606856289821e-06, "loss": 0.1162, "step": 3134 }, { "epoch": 0.13, "learning_rate": 9.59547356916269e-06, "loss": 0.115, "step": 3135 }, { "epoch": 0.13, "learning_rate": 9.595340282035563e-06, "loss": 0.1342, "step": 3136 }, { "epoch": 0.13, "learning_rate": 9.595206994908432e-06, "loss": 0.1047, "step": 3137 }, { "epoch": 0.13, "learning_rate": 9.595073707781303e-06, "loss": 0.0804, "step": 3138 }, { "epoch": 0.13, "learning_rate": 9.594940420654174e-06, "loss": 0.158, "step": 3139 }, { "epoch": 0.13, "learning_rate": 9.594807133527044e-06, "loss": 0.1056, "step": 3140 }, { "epoch": 0.13, "learning_rate": 9.594673846399916e-06, "loss": 0.1261, "step": 3141 }, { "epoch": 0.13, "learning_rate": 9.594540559272786e-06, "loss": 0.1305, "step": 3142 }, { "epoch": 0.13, "learning_rate": 9.594407272145657e-06, "loss": 0.1316, "step": 3143 }, { "epoch": 0.13, "learning_rate": 9.594273985018528e-06, "loss": 0.1095, "step": 3144 }, { "epoch": 0.13, "learning_rate": 9.594140697891399e-06, "loss": 0.1728, "step": 3145 }, { "epoch": 0.13, "learning_rate": 9.59400741076427e-06, "loss": 0.1505, "step": 3146 }, { "epoch": 0.13, "learning_rate": 9.59387412363714e-06, "loss": 0.0781, "step": 3147 }, { "epoch": 0.13, "learning_rate": 9.59374083651001e-06, "loss": 0.0959, "step": 3148 }, { "epoch": 0.13, "learning_rate": 9.593607549382881e-06, "loss": 0.083, "step": 3149 }, { "epoch": 0.13, "learning_rate": 9.593474262255752e-06, "loss": 0.1099, "step": 3150 }, { "epoch": 0.13, "learning_rate": 9.593340975128623e-06, "loss": 0.1223, "step": 3151 }, { "epoch": 0.13, "learning_rate": 9.593207688001494e-06, "loss": 0.1118, "step": 3152 }, { "epoch": 0.13, "learning_rate": 9.593074400874365e-06, "loss": 0.1136, "step": 3153 }, { "epoch": 0.13, "learning_rate": 9.592941113747234e-06, "loss": 0.1015, "step": 3154 }, { "epoch": 0.13, "learning_rate": 9.592807826620107e-06, "loss": 0.1588, "step": 3155 }, { "epoch": 0.13, "learning_rate": 9.592674539492976e-06, "loss": 0.092, "step": 3156 }, { "epoch": 0.13, "learning_rate": 9.592541252365847e-06, "loss": 0.0999, "step": 3157 }, { "epoch": 0.13, "learning_rate": 9.592407965238718e-06, "loss": 0.0949, "step": 3158 }, { "epoch": 0.13, "learning_rate": 9.592274678111588e-06, "loss": 0.162, "step": 3159 }, { "epoch": 0.13, "learning_rate": 9.59214139098446e-06, "loss": 0.1565, "step": 3160 }, { "epoch": 0.13, "learning_rate": 9.59200810385733e-06, "loss": 0.1097, "step": 3161 }, { "epoch": 0.13, "learning_rate": 9.5918748167302e-06, "loss": 0.1576, "step": 3162 }, { "epoch": 0.13, "learning_rate": 9.591741529603072e-06, "loss": 0.0847, "step": 3163 }, { "epoch": 0.13, "learning_rate": 9.591608242475943e-06, "loss": 0.1309, "step": 3164 }, { "epoch": 0.13, "learning_rate": 9.591474955348814e-06, "loss": 0.1232, "step": 3165 }, { "epoch": 0.13, "learning_rate": 9.591341668221685e-06, "loss": 0.1069, "step": 3166 }, { "epoch": 0.13, "learning_rate": 9.591208381094554e-06, "loss": 0.1074, "step": 3167 }, { "epoch": 0.13, "learning_rate": 9.591075093967427e-06, "loss": 0.161, "step": 3168 }, { "epoch": 0.13, "learning_rate": 9.590941806840296e-06, "loss": 0.1209, "step": 3169 }, { "epoch": 0.13, "learning_rate": 9.590808519713167e-06, "loss": 0.1014, "step": 3170 }, { "epoch": 0.13, "learning_rate": 9.590675232586038e-06, "loss": 0.1141, "step": 3171 }, { "epoch": 0.13, "learning_rate": 9.590541945458909e-06, "loss": 0.113, "step": 3172 }, { "epoch": 0.13, "learning_rate": 9.59040865833178e-06, "loss": 0.1162, "step": 3173 }, { "epoch": 0.13, "learning_rate": 9.59027537120465e-06, "loss": 0.126, "step": 3174 }, { "epoch": 0.13, "learning_rate": 9.59014208407752e-06, "loss": 0.1189, "step": 3175 }, { "epoch": 0.13, "learning_rate": 9.590008796950391e-06, "loss": 0.0945, "step": 3176 }, { "epoch": 0.13, "learning_rate": 9.589875509823262e-06, "loss": 0.1351, "step": 3177 }, { "epoch": 0.13, "learning_rate": 9.589742222696133e-06, "loss": 0.1595, "step": 3178 }, { "epoch": 0.13, "learning_rate": 9.589608935569004e-06, "loss": 0.1201, "step": 3179 }, { "epoch": 0.13, "learning_rate": 9.589475648441873e-06, "loss": 0.1165, "step": 3180 }, { "epoch": 0.13, "learning_rate": 9.589342361314744e-06, "loss": 0.1419, "step": 3181 }, { "epoch": 0.13, "learning_rate": 9.589209074187615e-06, "loss": 0.1097, "step": 3182 }, { "epoch": 0.13, "learning_rate": 9.589075787060486e-06, "loss": 0.138, "step": 3183 }, { "epoch": 0.13, "learning_rate": 9.588942499933357e-06, "loss": 0.1359, "step": 3184 }, { "epoch": 0.13, "learning_rate": 9.588809212806228e-06, "loss": 0.0932, "step": 3185 }, { "epoch": 0.13, "learning_rate": 9.588675925679098e-06, "loss": 0.1296, "step": 3186 }, { "epoch": 0.13, "learning_rate": 9.58854263855197e-06, "loss": 0.1183, "step": 3187 }, { "epoch": 0.13, "learning_rate": 9.58840935142484e-06, "loss": 0.1137, "step": 3188 }, { "epoch": 0.13, "learning_rate": 9.58827606429771e-06, "loss": 0.159, "step": 3189 }, { "epoch": 0.13, "learning_rate": 9.588142777170582e-06, "loss": 0.119, "step": 3190 }, { "epoch": 0.13, "learning_rate": 9.588009490043453e-06, "loss": 0.1423, "step": 3191 }, { "epoch": 0.13, "learning_rate": 9.587876202916324e-06, "loss": 0.1492, "step": 3192 }, { "epoch": 0.13, "learning_rate": 9.587742915789195e-06, "loss": 0.1336, "step": 3193 }, { "epoch": 0.13, "learning_rate": 9.587609628662064e-06, "loss": 0.1292, "step": 3194 }, { "epoch": 0.13, "learning_rate": 9.587476341534937e-06, "loss": 0.0944, "step": 3195 }, { "epoch": 0.13, "learning_rate": 9.587343054407806e-06, "loss": 0.1143, "step": 3196 }, { "epoch": 0.13, "learning_rate": 9.587209767280677e-06, "loss": 0.1048, "step": 3197 }, { "epoch": 0.13, "learning_rate": 9.587076480153548e-06, "loss": 0.1272, "step": 3198 }, { "epoch": 0.13, "learning_rate": 9.586943193026417e-06, "loss": 0.1092, "step": 3199 }, { "epoch": 0.13, "learning_rate": 9.58680990589929e-06, "loss": 0.1547, "step": 3200 }, { "epoch": 0.13, "learning_rate": 9.58667661877216e-06, "loss": 0.14, "step": 3201 }, { "epoch": 0.13, "learning_rate": 9.58654333164503e-06, "loss": 0.136, "step": 3202 }, { "epoch": 0.13, "learning_rate": 9.586410044517901e-06, "loss": 0.1474, "step": 3203 }, { "epoch": 0.13, "learning_rate": 9.586276757390772e-06, "loss": 0.1278, "step": 3204 }, { "epoch": 0.13, "learning_rate": 9.586143470263643e-06, "loss": 0.1481, "step": 3205 }, { "epoch": 0.13, "learning_rate": 9.586010183136514e-06, "loss": 0.0944, "step": 3206 }, { "epoch": 0.13, "learning_rate": 9.585876896009383e-06, "loss": 0.1084, "step": 3207 }, { "epoch": 0.13, "learning_rate": 9.585743608882254e-06, "loss": 0.0949, "step": 3208 }, { "epoch": 0.13, "learning_rate": 9.585610321755125e-06, "loss": 0.1441, "step": 3209 }, { "epoch": 0.13, "learning_rate": 9.585477034627996e-06, "loss": 0.1097, "step": 3210 }, { "epoch": 0.13, "learning_rate": 9.585343747500867e-06, "loss": 0.1193, "step": 3211 }, { "epoch": 0.13, "learning_rate": 9.585210460373738e-06, "loss": 0.0789, "step": 3212 }, { "epoch": 0.13, "learning_rate": 9.585077173246608e-06, "loss": 0.0884, "step": 3213 }, { "epoch": 0.13, "learning_rate": 9.58494388611948e-06, "loss": 0.1598, "step": 3214 }, { "epoch": 0.13, "learning_rate": 9.58481059899235e-06, "loss": 0.1213, "step": 3215 }, { "epoch": 0.13, "learning_rate": 9.58467731186522e-06, "loss": 0.1396, "step": 3216 }, { "epoch": 0.13, "learning_rate": 9.584544024738092e-06, "loss": 0.0995, "step": 3217 }, { "epoch": 0.13, "learning_rate": 9.584410737610961e-06, "loss": 0.1083, "step": 3218 }, { "epoch": 0.13, "learning_rate": 9.584277450483834e-06, "loss": 0.1546, "step": 3219 }, { "epoch": 0.13, "learning_rate": 9.584144163356703e-06, "loss": 0.1061, "step": 3220 }, { "epoch": 0.13, "learning_rate": 9.584010876229574e-06, "loss": 0.1364, "step": 3221 }, { "epoch": 0.13, "learning_rate": 9.583877589102445e-06, "loss": 0.1339, "step": 3222 }, { "epoch": 0.13, "learning_rate": 9.583744301975316e-06, "loss": 0.0761, "step": 3223 }, { "epoch": 0.13, "learning_rate": 9.583611014848187e-06, "loss": 0.1044, "step": 3224 }, { "epoch": 0.13, "learning_rate": 9.583477727721058e-06, "loss": 0.1406, "step": 3225 }, { "epoch": 0.13, "learning_rate": 9.583344440593927e-06, "loss": 0.1244, "step": 3226 }, { "epoch": 0.13, "learning_rate": 9.5832111534668e-06, "loss": 0.132, "step": 3227 }, { "epoch": 0.13, "learning_rate": 9.58307786633967e-06, "loss": 0.1375, "step": 3228 }, { "epoch": 0.13, "learning_rate": 9.58294457921254e-06, "loss": 0.1404, "step": 3229 }, { "epoch": 0.13, "learning_rate": 9.582811292085411e-06, "loss": 0.0899, "step": 3230 }, { "epoch": 0.13, "learning_rate": 9.582678004958282e-06, "loss": 0.0911, "step": 3231 }, { "epoch": 0.13, "learning_rate": 9.582544717831153e-06, "loss": 0.1315, "step": 3232 }, { "epoch": 0.13, "learning_rate": 9.582411430704024e-06, "loss": 0.1287, "step": 3233 }, { "epoch": 0.13, "learning_rate": 9.582278143576894e-06, "loss": 0.1231, "step": 3234 }, { "epoch": 0.13, "learning_rate": 9.582144856449765e-06, "loss": 0.1144, "step": 3235 }, { "epoch": 0.13, "learning_rate": 9.582011569322636e-06, "loss": 0.1521, "step": 3236 }, { "epoch": 0.13, "learning_rate": 9.581878282195507e-06, "loss": 0.1222, "step": 3237 }, { "epoch": 0.13, "learning_rate": 9.581744995068378e-06, "loss": 0.121, "step": 3238 }, { "epoch": 0.13, "learning_rate": 9.581611707941247e-06, "loss": 0.1494, "step": 3239 }, { "epoch": 0.13, "learning_rate": 9.58147842081412e-06, "loss": 0.114, "step": 3240 }, { "epoch": 0.13, "learning_rate": 9.581345133686989e-06, "loss": 0.1639, "step": 3241 }, { "epoch": 0.13, "learning_rate": 9.58121184655986e-06, "loss": 0.1773, "step": 3242 }, { "epoch": 0.13, "learning_rate": 9.58107855943273e-06, "loss": 0.1299, "step": 3243 }, { "epoch": 0.13, "learning_rate": 9.580945272305602e-06, "loss": 0.112, "step": 3244 }, { "epoch": 0.13, "learning_rate": 9.580811985178473e-06, "loss": 0.1155, "step": 3245 }, { "epoch": 0.13, "learning_rate": 9.580678698051344e-06, "loss": 0.1478, "step": 3246 }, { "epoch": 0.13, "learning_rate": 9.580545410924213e-06, "loss": 0.133, "step": 3247 }, { "epoch": 0.13, "learning_rate": 9.580412123797084e-06, "loss": 0.1227, "step": 3248 }, { "epoch": 0.13, "learning_rate": 9.580278836669955e-06, "loss": 0.0883, "step": 3249 }, { "epoch": 0.13, "learning_rate": 9.580145549542826e-06, "loss": 0.0899, "step": 3250 }, { "epoch": 0.13, "learning_rate": 9.580012262415697e-06, "loss": 0.1378, "step": 3251 }, { "epoch": 0.13, "learning_rate": 9.579878975288568e-06, "loss": 0.1023, "step": 3252 }, { "epoch": 0.13, "learning_rate": 9.579745688161437e-06, "loss": 0.1561, "step": 3253 }, { "epoch": 0.13, "learning_rate": 9.579612401034308e-06, "loss": 0.1317, "step": 3254 }, { "epoch": 0.13, "learning_rate": 9.57947911390718e-06, "loss": 0.0876, "step": 3255 }, { "epoch": 0.13, "learning_rate": 9.57934582678005e-06, "loss": 0.0901, "step": 3256 }, { "epoch": 0.13, "learning_rate": 9.579212539652921e-06, "loss": 0.1477, "step": 3257 }, { "epoch": 0.13, "learning_rate": 9.57907925252579e-06, "loss": 0.0944, "step": 3258 }, { "epoch": 0.13, "learning_rate": 9.578945965398663e-06, "loss": 0.1576, "step": 3259 }, { "epoch": 0.13, "learning_rate": 9.578812678271533e-06, "loss": 0.1283, "step": 3260 }, { "epoch": 0.13, "learning_rate": 9.578679391144404e-06, "loss": 0.0841, "step": 3261 }, { "epoch": 0.13, "learning_rate": 9.578546104017275e-06, "loss": 0.0932, "step": 3262 }, { "epoch": 0.13, "learning_rate": 9.578412816890146e-06, "loss": 0.1077, "step": 3263 }, { "epoch": 0.13, "learning_rate": 9.578279529763017e-06, "loss": 0.0972, "step": 3264 }, { "epoch": 0.13, "learning_rate": 9.578146242635888e-06, "loss": 0.1578, "step": 3265 }, { "epoch": 0.13, "learning_rate": 9.578012955508757e-06, "loss": 0.1276, "step": 3266 }, { "epoch": 0.13, "learning_rate": 9.57787966838163e-06, "loss": 0.1411, "step": 3267 }, { "epoch": 0.13, "learning_rate": 9.577746381254499e-06, "loss": 0.1246, "step": 3268 }, { "epoch": 0.13, "learning_rate": 9.57761309412737e-06, "loss": 0.1105, "step": 3269 }, { "epoch": 0.13, "learning_rate": 9.577479807000241e-06, "loss": 0.1529, "step": 3270 }, { "epoch": 0.13, "learning_rate": 9.577346519873112e-06, "loss": 0.1387, "step": 3271 }, { "epoch": 0.13, "learning_rate": 9.577213232745983e-06, "loss": 0.14, "step": 3272 }, { "epoch": 0.13, "learning_rate": 9.577079945618854e-06, "loss": 0.0896, "step": 3273 }, { "epoch": 0.13, "learning_rate": 9.576946658491723e-06, "loss": 0.0796, "step": 3274 }, { "epoch": 0.13, "learning_rate": 9.576813371364594e-06, "loss": 0.128, "step": 3275 }, { "epoch": 0.13, "learning_rate": 9.576680084237465e-06, "loss": 0.1151, "step": 3276 }, { "epoch": 0.13, "learning_rate": 9.576546797110336e-06, "loss": 0.1204, "step": 3277 }, { "epoch": 0.13, "learning_rate": 9.576413509983207e-06, "loss": 0.1243, "step": 3278 }, { "epoch": 0.13, "learning_rate": 9.576280222856076e-06, "loss": 0.1025, "step": 3279 }, { "epoch": 0.13, "learning_rate": 9.576146935728947e-06, "loss": 0.1278, "step": 3280 }, { "epoch": 0.13, "learning_rate": 9.576013648601818e-06, "loss": 0.1044, "step": 3281 }, { "epoch": 0.13, "learning_rate": 9.57588036147469e-06, "loss": 0.098, "step": 3282 }, { "epoch": 0.13, "learning_rate": 9.57574707434756e-06, "loss": 0.1508, "step": 3283 }, { "epoch": 0.13, "learning_rate": 9.575613787220431e-06, "loss": 0.0936, "step": 3284 }, { "epoch": 0.13, "learning_rate": 9.5754805000933e-06, "loss": 0.0914, "step": 3285 }, { "epoch": 0.13, "learning_rate": 9.575347212966173e-06, "loss": 0.1391, "step": 3286 }, { "epoch": 0.13, "learning_rate": 9.575213925839043e-06, "loss": 0.1249, "step": 3287 }, { "epoch": 0.13, "learning_rate": 9.575080638711914e-06, "loss": 0.1059, "step": 3288 }, { "epoch": 0.13, "learning_rate": 9.574947351584785e-06, "loss": 0.1421, "step": 3289 }, { "epoch": 0.13, "learning_rate": 9.574814064457656e-06, "loss": 0.1091, "step": 3290 }, { "epoch": 0.13, "learning_rate": 9.574680777330527e-06, "loss": 0.0978, "step": 3291 }, { "epoch": 0.13, "learning_rate": 9.574547490203398e-06, "loss": 0.1403, "step": 3292 }, { "epoch": 0.13, "learning_rate": 9.574414203076267e-06, "loss": 0.1477, "step": 3293 }, { "epoch": 0.13, "learning_rate": 9.574280915949138e-06, "loss": 0.1105, "step": 3294 }, { "epoch": 0.13, "learning_rate": 9.574147628822009e-06, "loss": 0.0974, "step": 3295 }, { "epoch": 0.13, "learning_rate": 9.57401434169488e-06, "loss": 0.1285, "step": 3296 }, { "epoch": 0.13, "learning_rate": 9.573881054567751e-06, "loss": 0.1845, "step": 3297 }, { "epoch": 0.13, "learning_rate": 9.57374776744062e-06, "loss": 0.0995, "step": 3298 }, { "epoch": 0.13, "learning_rate": 9.573614480313493e-06, "loss": 0.0964, "step": 3299 }, { "epoch": 0.13, "learning_rate": 9.573481193186362e-06, "loss": 0.1108, "step": 3300 }, { "epoch": 0.13, "learning_rate": 9.573347906059233e-06, "loss": 0.1316, "step": 3301 }, { "epoch": 0.13, "learning_rate": 9.573214618932104e-06, "loss": 0.1581, "step": 3302 }, { "epoch": 0.13, "learning_rate": 9.573081331804975e-06, "loss": 0.106, "step": 3303 }, { "epoch": 0.13, "learning_rate": 9.572948044677846e-06, "loss": 0.1335, "step": 3304 }, { "epoch": 0.13, "learning_rate": 9.572814757550717e-06, "loss": 0.1715, "step": 3305 }, { "epoch": 0.13, "learning_rate": 9.572681470423587e-06, "loss": 0.0995, "step": 3306 }, { "epoch": 0.13, "learning_rate": 9.57254818329646e-06, "loss": 0.1052, "step": 3307 }, { "epoch": 0.13, "learning_rate": 9.572414896169329e-06, "loss": 0.116, "step": 3308 }, { "epoch": 0.13, "learning_rate": 9.5722816090422e-06, "loss": 0.1636, "step": 3309 }, { "epoch": 0.13, "learning_rate": 9.57214832191507e-06, "loss": 0.1387, "step": 3310 }, { "epoch": 0.13, "learning_rate": 9.572015034787942e-06, "loss": 0.1098, "step": 3311 }, { "epoch": 0.13, "learning_rate": 9.571881747660812e-06, "loss": 0.1218, "step": 3312 }, { "epoch": 0.13, "learning_rate": 9.571748460533682e-06, "loss": 0.1551, "step": 3313 }, { "epoch": 0.13, "learning_rate": 9.571615173406553e-06, "loss": 0.1087, "step": 3314 }, { "epoch": 0.13, "learning_rate": 9.571481886279424e-06, "loss": 0.1239, "step": 3315 }, { "epoch": 0.13, "learning_rate": 9.571348599152295e-06, "loss": 0.1096, "step": 3316 }, { "epoch": 0.13, "learning_rate": 9.571215312025166e-06, "loss": 0.1029, "step": 3317 }, { "epoch": 0.13, "learning_rate": 9.571082024898037e-06, "loss": 0.1928, "step": 3318 }, { "epoch": 0.13, "learning_rate": 9.570948737770906e-06, "loss": 0.0942, "step": 3319 }, { "epoch": 0.13, "learning_rate": 9.570815450643777e-06, "loss": 0.122, "step": 3320 }, { "epoch": 0.13, "learning_rate": 9.570682163516648e-06, "loss": 0.1329, "step": 3321 }, { "epoch": 0.13, "learning_rate": 9.570548876389519e-06, "loss": 0.0859, "step": 3322 }, { "epoch": 0.13, "learning_rate": 9.57041558926239e-06, "loss": 0.1124, "step": 3323 }, { "epoch": 0.13, "learning_rate": 9.570282302135261e-06, "loss": 0.1249, "step": 3324 }, { "epoch": 0.13, "learning_rate": 9.57014901500813e-06, "loss": 0.1109, "step": 3325 }, { "epoch": 0.13, "learning_rate": 9.570015727881003e-06, "loss": 0.1562, "step": 3326 }, { "epoch": 0.13, "learning_rate": 9.569882440753872e-06, "loss": 0.1543, "step": 3327 }, { "epoch": 0.13, "learning_rate": 9.569749153626743e-06, "loss": 0.1373, "step": 3328 }, { "epoch": 0.13, "learning_rate": 9.569615866499614e-06, "loss": 0.114, "step": 3329 }, { "epoch": 0.13, "learning_rate": 9.569482579372485e-06, "loss": 0.0981, "step": 3330 }, { "epoch": 0.13, "learning_rate": 9.569349292245356e-06, "loss": 0.1815, "step": 3331 }, { "epoch": 0.13, "learning_rate": 9.569216005118227e-06, "loss": 0.1298, "step": 3332 }, { "epoch": 0.13, "learning_rate": 9.569082717991097e-06, "loss": 0.1031, "step": 3333 }, { "epoch": 0.13, "learning_rate": 9.568949430863968e-06, "loss": 0.1129, "step": 3334 }, { "epoch": 0.13, "learning_rate": 9.568816143736839e-06, "loss": 0.1374, "step": 3335 }, { "epoch": 0.13, "learning_rate": 9.56868285660971e-06, "loss": 0.1359, "step": 3336 }, { "epoch": 0.13, "learning_rate": 9.56854956948258e-06, "loss": 0.0809, "step": 3337 }, { "epoch": 0.13, "learning_rate": 9.56841628235545e-06, "loss": 0.1457, "step": 3338 }, { "epoch": 0.13, "learning_rate": 9.568282995228323e-06, "loss": 0.205, "step": 3339 }, { "epoch": 0.13, "learning_rate": 9.568149708101192e-06, "loss": 0.1195, "step": 3340 }, { "epoch": 0.13, "learning_rate": 9.568016420974063e-06, "loss": 0.1155, "step": 3341 }, { "epoch": 0.13, "learning_rate": 9.567883133846934e-06, "loss": 0.0997, "step": 3342 }, { "epoch": 0.13, "learning_rate": 9.567749846719805e-06, "loss": 0.1215, "step": 3343 }, { "epoch": 0.13, "learning_rate": 9.567616559592676e-06, "loss": 0.1302, "step": 3344 }, { "epoch": 0.13, "learning_rate": 9.567483272465547e-06, "loss": 0.119, "step": 3345 }, { "epoch": 0.13, "learning_rate": 9.567349985338416e-06, "loss": 0.1228, "step": 3346 }, { "epoch": 0.13, "learning_rate": 9.567216698211287e-06, "loss": 0.1076, "step": 3347 }, { "epoch": 0.13, "learning_rate": 9.567083411084158e-06, "loss": 0.1109, "step": 3348 }, { "epoch": 0.13, "learning_rate": 9.566950123957029e-06, "loss": 0.1361, "step": 3349 }, { "epoch": 0.13, "learning_rate": 9.5668168368299e-06, "loss": 0.1096, "step": 3350 }, { "epoch": 0.13, "learning_rate": 9.566683549702771e-06, "loss": 0.1321, "step": 3351 }, { "epoch": 0.13, "learning_rate": 9.56655026257564e-06, "loss": 0.1325, "step": 3352 }, { "epoch": 0.13, "learning_rate": 9.566416975448511e-06, "loss": 0.1519, "step": 3353 }, { "epoch": 0.13, "learning_rate": 9.566283688321382e-06, "loss": 0.1128, "step": 3354 }, { "epoch": 0.13, "learning_rate": 9.566150401194253e-06, "loss": 0.1028, "step": 3355 }, { "epoch": 0.13, "learning_rate": 9.566017114067124e-06, "loss": 0.1129, "step": 3356 }, { "epoch": 0.13, "learning_rate": 9.565883826939994e-06, "loss": 0.1196, "step": 3357 }, { "epoch": 0.13, "learning_rate": 9.565750539812866e-06, "loss": 0.0843, "step": 3358 }, { "epoch": 0.13, "learning_rate": 9.565617252685736e-06, "loss": 0.1031, "step": 3359 }, { "epoch": 0.13, "learning_rate": 9.565483965558607e-06, "loss": 0.1272, "step": 3360 }, { "epoch": 0.13, "learning_rate": 9.565350678431478e-06, "loss": 0.0957, "step": 3361 }, { "epoch": 0.13, "learning_rate": 9.565217391304349e-06, "loss": 0.126, "step": 3362 }, { "epoch": 0.13, "learning_rate": 9.56508410417722e-06, "loss": 0.1198, "step": 3363 }, { "epoch": 0.13, "learning_rate": 9.56495081705009e-06, "loss": 0.107, "step": 3364 }, { "epoch": 0.13, "learning_rate": 9.56481752992296e-06, "loss": 0.1076, "step": 3365 }, { "epoch": 0.13, "learning_rate": 9.564684242795833e-06, "loss": 0.1115, "step": 3366 }, { "epoch": 0.13, "learning_rate": 9.564550955668702e-06, "loss": 0.0693, "step": 3367 }, { "epoch": 0.13, "learning_rate": 9.564417668541573e-06, "loss": 0.0987, "step": 3368 }, { "epoch": 0.13, "learning_rate": 9.564284381414444e-06, "loss": 0.103, "step": 3369 }, { "epoch": 0.13, "learning_rate": 9.564151094287315e-06, "loss": 0.0829, "step": 3370 }, { "epoch": 0.13, "learning_rate": 9.564017807160186e-06, "loss": 0.1154, "step": 3371 }, { "epoch": 0.13, "learning_rate": 9.563884520033055e-06, "loss": 0.0881, "step": 3372 }, { "epoch": 0.13, "learning_rate": 9.563751232905926e-06, "loss": 0.0818, "step": 3373 }, { "epoch": 0.13, "learning_rate": 9.563617945778797e-06, "loss": 0.1105, "step": 3374 }, { "epoch": 0.13, "learning_rate": 9.563484658651668e-06, "loss": 0.1506, "step": 3375 }, { "epoch": 0.13, "learning_rate": 9.56335137152454e-06, "loss": 0.0936, "step": 3376 }, { "epoch": 0.13, "learning_rate": 9.56321808439741e-06, "loss": 0.1242, "step": 3377 }, { "epoch": 0.13, "learning_rate": 9.56308479727028e-06, "loss": 0.0746, "step": 3378 }, { "epoch": 0.13, "learning_rate": 9.562951510143152e-06, "loss": 0.1118, "step": 3379 }, { "epoch": 0.13, "learning_rate": 9.562818223016022e-06, "loss": 0.1135, "step": 3380 }, { "epoch": 0.14, "learning_rate": 9.562684935888892e-06, "loss": 0.1369, "step": 3381 }, { "epoch": 0.14, "learning_rate": 9.562551648761763e-06, "loss": 0.1274, "step": 3382 }, { "epoch": 0.14, "learning_rate": 9.562418361634634e-06, "loss": 0.0969, "step": 3383 }, { "epoch": 0.14, "learning_rate": 9.562285074507505e-06, "loss": 0.1319, "step": 3384 }, { "epoch": 0.14, "learning_rate": 9.562151787380376e-06, "loss": 0.1483, "step": 3385 }, { "epoch": 0.14, "learning_rate": 9.562018500253246e-06, "loss": 0.1308, "step": 3386 }, { "epoch": 0.14, "learning_rate": 9.561885213126117e-06, "loss": 0.1108, "step": 3387 }, { "epoch": 0.14, "learning_rate": 9.561751925998988e-06, "loss": 0.159, "step": 3388 }, { "epoch": 0.14, "learning_rate": 9.561618638871859e-06, "loss": 0.1395, "step": 3389 }, { "epoch": 0.14, "learning_rate": 9.56148535174473e-06, "loss": 0.1236, "step": 3390 }, { "epoch": 0.14, "learning_rate": 9.561352064617599e-06, "loss": 0.1052, "step": 3391 }, { "epoch": 0.14, "learning_rate": 9.56121877749047e-06, "loss": 0.1264, "step": 3392 }, { "epoch": 0.14, "learning_rate": 9.561085490363341e-06, "loss": 0.1364, "step": 3393 }, { "epoch": 0.14, "learning_rate": 9.560952203236212e-06, "loss": 0.1027, "step": 3394 }, { "epoch": 0.14, "learning_rate": 9.560818916109083e-06, "loss": 0.119, "step": 3395 }, { "epoch": 0.14, "learning_rate": 9.560685628981954e-06, "loss": 0.1289, "step": 3396 }, { "epoch": 0.14, "learning_rate": 9.560552341854823e-06, "loss": 0.1719, "step": 3397 }, { "epoch": 0.14, "learning_rate": 9.560419054727696e-06, "loss": 0.1393, "step": 3398 }, { "epoch": 0.14, "learning_rate": 9.560285767600565e-06, "loss": 0.1077, "step": 3399 }, { "epoch": 0.14, "learning_rate": 9.560152480473436e-06, "loss": 0.123, "step": 3400 }, { "epoch": 0.14, "learning_rate": 9.560019193346307e-06, "loss": 0.1228, "step": 3401 }, { "epoch": 0.14, "learning_rate": 9.559885906219178e-06, "loss": 0.1678, "step": 3402 }, { "epoch": 0.14, "learning_rate": 9.55975261909205e-06, "loss": 0.0833, "step": 3403 }, { "epoch": 0.14, "learning_rate": 9.55961933196492e-06, "loss": 0.1324, "step": 3404 }, { "epoch": 0.14, "learning_rate": 9.55948604483779e-06, "loss": 0.1406, "step": 3405 }, { "epoch": 0.14, "learning_rate": 9.559352757710662e-06, "loss": 0.1374, "step": 3406 }, { "epoch": 0.14, "learning_rate": 9.559219470583532e-06, "loss": 0.1077, "step": 3407 }, { "epoch": 0.14, "learning_rate": 9.559086183456403e-06, "loss": 0.1049, "step": 3408 }, { "epoch": 0.14, "learning_rate": 9.558952896329274e-06, "loss": 0.0846, "step": 3409 }, { "epoch": 0.14, "learning_rate": 9.558819609202145e-06, "loss": 0.1242, "step": 3410 }, { "epoch": 0.14, "learning_rate": 9.558686322075016e-06, "loss": 0.1466, "step": 3411 }, { "epoch": 0.14, "learning_rate": 9.558553034947885e-06, "loss": 0.1413, "step": 3412 }, { "epoch": 0.14, "learning_rate": 9.558419747820756e-06, "loss": 0.1301, "step": 3413 }, { "epoch": 0.14, "learning_rate": 9.558286460693627e-06, "loss": 0.149, "step": 3414 }, { "epoch": 0.14, "learning_rate": 9.558153173566498e-06, "loss": 0.0957, "step": 3415 }, { "epoch": 0.14, "learning_rate": 9.558019886439369e-06, "loss": 0.1076, "step": 3416 }, { "epoch": 0.14, "learning_rate": 9.55788659931224e-06, "loss": 0.1114, "step": 3417 }, { "epoch": 0.14, "learning_rate": 9.557753312185109e-06, "loss": 0.1089, "step": 3418 }, { "epoch": 0.14, "learning_rate": 9.55762002505798e-06, "loss": 0.1195, "step": 3419 }, { "epoch": 0.14, "learning_rate": 9.557486737930851e-06, "loss": 0.1225, "step": 3420 }, { "epoch": 0.14, "learning_rate": 9.557353450803722e-06, "loss": 0.1582, "step": 3421 }, { "epoch": 0.14, "learning_rate": 9.557220163676593e-06, "loss": 0.1161, "step": 3422 }, { "epoch": 0.14, "learning_rate": 9.557086876549464e-06, "loss": 0.1049, "step": 3423 }, { "epoch": 0.14, "learning_rate": 9.556953589422333e-06, "loss": 0.0862, "step": 3424 }, { "epoch": 0.14, "learning_rate": 9.556820302295206e-06, "loss": 0.1207, "step": 3425 }, { "epoch": 0.14, "learning_rate": 9.556687015168075e-06, "loss": 0.1314, "step": 3426 }, { "epoch": 0.14, "learning_rate": 9.556553728040946e-06, "loss": 0.1019, "step": 3427 }, { "epoch": 0.14, "learning_rate": 9.556420440913817e-06, "loss": 0.1045, "step": 3428 }, { "epoch": 0.14, "learning_rate": 9.556287153786688e-06, "loss": 0.0889, "step": 3429 }, { "epoch": 0.14, "learning_rate": 9.55615386665956e-06, "loss": 0.1302, "step": 3430 }, { "epoch": 0.14, "learning_rate": 9.556020579532429e-06, "loss": 0.0885, "step": 3431 }, { "epoch": 0.14, "learning_rate": 9.5558872924053e-06, "loss": 0.148, "step": 3432 }, { "epoch": 0.14, "learning_rate": 9.55575400527817e-06, "loss": 0.0905, "step": 3433 }, { "epoch": 0.14, "learning_rate": 9.555620718151042e-06, "loss": 0.1386, "step": 3434 }, { "epoch": 0.14, "learning_rate": 9.555487431023913e-06, "loss": 0.119, "step": 3435 }, { "epoch": 0.14, "learning_rate": 9.555354143896784e-06, "loss": 0.1317, "step": 3436 }, { "epoch": 0.14, "learning_rate": 9.555220856769653e-06, "loss": 0.1021, "step": 3437 }, { "epoch": 0.14, "learning_rate": 9.555087569642526e-06, "loss": 0.1395, "step": 3438 }, { "epoch": 0.14, "learning_rate": 9.554954282515395e-06, "loss": 0.136, "step": 3439 }, { "epoch": 0.14, "learning_rate": 9.554820995388266e-06, "loss": 0.1418, "step": 3440 }, { "epoch": 0.14, "learning_rate": 9.554687708261137e-06, "loss": 0.1299, "step": 3441 }, { "epoch": 0.14, "learning_rate": 9.554554421134008e-06, "loss": 0.099, "step": 3442 }, { "epoch": 0.14, "learning_rate": 9.554421134006879e-06, "loss": 0.1499, "step": 3443 }, { "epoch": 0.14, "learning_rate": 9.55428784687975e-06, "loss": 0.1902, "step": 3444 }, { "epoch": 0.14, "learning_rate": 9.55415455975262e-06, "loss": 0.1596, "step": 3445 }, { "epoch": 0.14, "learning_rate": 9.55402127262549e-06, "loss": 0.1703, "step": 3446 }, { "epoch": 0.14, "learning_rate": 9.553887985498361e-06, "loss": 0.1062, "step": 3447 }, { "epoch": 0.14, "learning_rate": 9.553754698371232e-06, "loss": 0.0932, "step": 3448 }, { "epoch": 0.14, "learning_rate": 9.553621411244103e-06, "loss": 0.1343, "step": 3449 }, { "epoch": 0.14, "learning_rate": 9.553488124116972e-06, "loss": 0.102, "step": 3450 }, { "epoch": 0.14, "learning_rate": 9.553354836989843e-06, "loss": 0.1569, "step": 3451 }, { "epoch": 0.14, "learning_rate": 9.553221549862714e-06, "loss": 0.125, "step": 3452 }, { "epoch": 0.14, "learning_rate": 9.553088262735585e-06, "loss": 0.1111, "step": 3453 }, { "epoch": 0.14, "learning_rate": 9.552954975608456e-06, "loss": 0.1274, "step": 3454 }, { "epoch": 0.14, "learning_rate": 9.552821688481327e-06, "loss": 0.1713, "step": 3455 }, { "epoch": 0.14, "learning_rate": 9.552688401354197e-06, "loss": 0.1163, "step": 3456 }, { "epoch": 0.14, "learning_rate": 9.55255511422707e-06, "loss": 0.0991, "step": 3457 }, { "epoch": 0.14, "learning_rate": 9.552421827099939e-06, "loss": 0.1229, "step": 3458 }, { "epoch": 0.14, "learning_rate": 9.55228853997281e-06, "loss": 0.1104, "step": 3459 }, { "epoch": 0.14, "learning_rate": 9.55215525284568e-06, "loss": 0.1247, "step": 3460 }, { "epoch": 0.14, "learning_rate": 9.552021965718552e-06, "loss": 0.1656, "step": 3461 }, { "epoch": 0.14, "learning_rate": 9.551888678591423e-06, "loss": 0.145, "step": 3462 }, { "epoch": 0.14, "learning_rate": 9.551755391464294e-06, "loss": 0.1054, "step": 3463 }, { "epoch": 0.14, "learning_rate": 9.551622104337163e-06, "loss": 0.1337, "step": 3464 }, { "epoch": 0.14, "learning_rate": 9.551488817210036e-06, "loss": 0.1196, "step": 3465 }, { "epoch": 0.14, "learning_rate": 9.551355530082905e-06, "loss": 0.111, "step": 3466 }, { "epoch": 0.14, "learning_rate": 9.551222242955776e-06, "loss": 0.0944, "step": 3467 }, { "epoch": 0.14, "learning_rate": 9.551088955828647e-06, "loss": 0.1171, "step": 3468 }, { "epoch": 0.14, "learning_rate": 9.550955668701518e-06, "loss": 0.139, "step": 3469 }, { "epoch": 0.14, "learning_rate": 9.550822381574389e-06, "loss": 0.1224, "step": 3470 }, { "epoch": 0.14, "learning_rate": 9.550689094447258e-06, "loss": 0.1073, "step": 3471 }, { "epoch": 0.14, "learning_rate": 9.55055580732013e-06, "loss": 0.1339, "step": 3472 }, { "epoch": 0.14, "learning_rate": 9.550422520193e-06, "loss": 0.1727, "step": 3473 }, { "epoch": 0.14, "learning_rate": 9.550289233065871e-06, "loss": 0.1168, "step": 3474 }, { "epoch": 0.14, "learning_rate": 9.550155945938742e-06, "loss": 0.0937, "step": 3475 }, { "epoch": 0.14, "learning_rate": 9.550022658811613e-06, "loss": 0.143, "step": 3476 }, { "epoch": 0.14, "learning_rate": 9.549889371684483e-06, "loss": 0.1161, "step": 3477 }, { "epoch": 0.14, "learning_rate": 9.549756084557355e-06, "loss": 0.0919, "step": 3478 }, { "epoch": 0.14, "learning_rate": 9.549622797430225e-06, "loss": 0.1326, "step": 3479 }, { "epoch": 0.14, "learning_rate": 9.549489510303096e-06, "loss": 0.1221, "step": 3480 }, { "epoch": 0.14, "learning_rate": 9.549356223175967e-06, "loss": 0.0896, "step": 3481 }, { "epoch": 0.14, "learning_rate": 9.549222936048838e-06, "loss": 0.1164, "step": 3482 }, { "epoch": 0.14, "learning_rate": 9.549089648921709e-06, "loss": 0.1141, "step": 3483 }, { "epoch": 0.14, "learning_rate": 9.54895636179458e-06, "loss": 0.1639, "step": 3484 }, { "epoch": 0.14, "learning_rate": 9.548823074667449e-06, "loss": 0.1067, "step": 3485 }, { "epoch": 0.14, "learning_rate": 9.54868978754032e-06, "loss": 0.1136, "step": 3486 }, { "epoch": 0.14, "learning_rate": 9.54855650041319e-06, "loss": 0.0973, "step": 3487 }, { "epoch": 0.14, "learning_rate": 9.548423213286062e-06, "loss": 0.1385, "step": 3488 }, { "epoch": 0.14, "learning_rate": 9.548289926158933e-06, "loss": 0.1368, "step": 3489 }, { "epoch": 0.14, "learning_rate": 9.548156639031802e-06, "loss": 0.1297, "step": 3490 }, { "epoch": 0.14, "learning_rate": 9.548023351904673e-06, "loss": 0.138, "step": 3491 }, { "epoch": 0.14, "learning_rate": 9.547890064777544e-06, "loss": 0.141, "step": 3492 }, { "epoch": 0.14, "learning_rate": 9.547756777650415e-06, "loss": 0.1393, "step": 3493 }, { "epoch": 0.14, "learning_rate": 9.547623490523286e-06, "loss": 0.1035, "step": 3494 }, { "epoch": 0.14, "learning_rate": 9.547490203396157e-06, "loss": 0.0807, "step": 3495 }, { "epoch": 0.14, "learning_rate": 9.547356916269026e-06, "loss": 0.1191, "step": 3496 }, { "epoch": 0.14, "learning_rate": 9.547223629141899e-06, "loss": 0.1483, "step": 3497 }, { "epoch": 0.14, "learning_rate": 9.547090342014768e-06, "loss": 0.1417, "step": 3498 }, { "epoch": 0.14, "learning_rate": 9.54695705488764e-06, "loss": 0.1014, "step": 3499 }, { "epoch": 0.14, "learning_rate": 9.54682376776051e-06, "loss": 0.1249, "step": 3500 }, { "epoch": 0.14, "learning_rate": 9.546690480633381e-06, "loss": 0.1079, "step": 3501 }, { "epoch": 0.14, "learning_rate": 9.546557193506252e-06, "loss": 0.1256, "step": 3502 }, { "epoch": 0.14, "learning_rate": 9.546423906379123e-06, "loss": 0.0883, "step": 3503 }, { "epoch": 0.14, "learning_rate": 9.546290619251993e-06, "loss": 0.1061, "step": 3504 }, { "epoch": 0.14, "learning_rate": 9.546157332124865e-06, "loss": 0.1117, "step": 3505 }, { "epoch": 0.14, "learning_rate": 9.546024044997735e-06, "loss": 0.1534, "step": 3506 }, { "epoch": 0.14, "learning_rate": 9.545890757870606e-06, "loss": 0.1148, "step": 3507 }, { "epoch": 0.14, "learning_rate": 9.545757470743477e-06, "loss": 0.0987, "step": 3508 }, { "epoch": 0.14, "learning_rate": 9.545624183616346e-06, "loss": 0.1302, "step": 3509 }, { "epoch": 0.14, "learning_rate": 9.545490896489219e-06, "loss": 0.1038, "step": 3510 }, { "epoch": 0.14, "learning_rate": 9.545357609362088e-06, "loss": 0.1427, "step": 3511 }, { "epoch": 0.14, "learning_rate": 9.545224322234959e-06, "loss": 0.0984, "step": 3512 }, { "epoch": 0.14, "learning_rate": 9.54509103510783e-06, "loss": 0.1041, "step": 3513 }, { "epoch": 0.14, "learning_rate": 9.544957747980701e-06, "loss": 0.1279, "step": 3514 }, { "epoch": 0.14, "learning_rate": 9.544824460853572e-06, "loss": 0.078, "step": 3515 }, { "epoch": 0.14, "learning_rate": 9.544691173726443e-06, "loss": 0.1033, "step": 3516 }, { "epoch": 0.14, "learning_rate": 9.544557886599312e-06, "loss": 0.172, "step": 3517 }, { "epoch": 0.14, "learning_rate": 9.544424599472183e-06, "loss": 0.1039, "step": 3518 }, { "epoch": 0.14, "learning_rate": 9.544291312345054e-06, "loss": 0.1082, "step": 3519 }, { "epoch": 0.14, "learning_rate": 9.544158025217925e-06, "loss": 0.1304, "step": 3520 }, { "epoch": 0.14, "learning_rate": 9.544024738090796e-06, "loss": 0.1229, "step": 3521 }, { "epoch": 0.14, "learning_rate": 9.543891450963667e-06, "loss": 0.1037, "step": 3522 }, { "epoch": 0.14, "learning_rate": 9.543758163836536e-06, "loss": 0.128, "step": 3523 }, { "epoch": 0.14, "learning_rate": 9.54362487670941e-06, "loss": 0.1512, "step": 3524 }, { "epoch": 0.14, "learning_rate": 9.543491589582278e-06, "loss": 0.111, "step": 3525 }, { "epoch": 0.14, "learning_rate": 9.54335830245515e-06, "loss": 0.1196, "step": 3526 }, { "epoch": 0.14, "learning_rate": 9.54322501532802e-06, "loss": 0.1199, "step": 3527 }, { "epoch": 0.14, "learning_rate": 9.543091728200891e-06, "loss": 0.0867, "step": 3528 }, { "epoch": 0.14, "learning_rate": 9.542958441073762e-06, "loss": 0.1213, "step": 3529 }, { "epoch": 0.14, "learning_rate": 9.542825153946632e-06, "loss": 0.1137, "step": 3530 }, { "epoch": 0.14, "learning_rate": 9.542691866819503e-06, "loss": 0.1406, "step": 3531 }, { "epoch": 0.14, "learning_rate": 9.542558579692374e-06, "loss": 0.0914, "step": 3532 }, { "epoch": 0.14, "learning_rate": 9.542425292565245e-06, "loss": 0.1408, "step": 3533 }, { "epoch": 0.14, "learning_rate": 9.542292005438116e-06, "loss": 0.1417, "step": 3534 }, { "epoch": 0.14, "learning_rate": 9.542158718310987e-06, "loss": 0.0962, "step": 3535 }, { "epoch": 0.14, "learning_rate": 9.542025431183856e-06, "loss": 0.154, "step": 3536 }, { "epoch": 0.14, "learning_rate": 9.541892144056729e-06, "loss": 0.1463, "step": 3537 }, { "epoch": 0.14, "learning_rate": 9.541758856929598e-06, "loss": 0.089, "step": 3538 }, { "epoch": 0.14, "learning_rate": 9.541625569802469e-06, "loss": 0.0691, "step": 3539 }, { "epoch": 0.14, "learning_rate": 9.54149228267534e-06, "loss": 0.0567, "step": 3540 }, { "epoch": 0.14, "learning_rate": 9.541358995548211e-06, "loss": 0.1235, "step": 3541 }, { "epoch": 0.14, "learning_rate": 9.541225708421082e-06, "loss": 0.1661, "step": 3542 }, { "epoch": 0.14, "learning_rate": 9.541092421293953e-06, "loss": 0.0914, "step": 3543 }, { "epoch": 0.14, "learning_rate": 9.540959134166822e-06, "loss": 0.1225, "step": 3544 }, { "epoch": 0.14, "learning_rate": 9.540825847039695e-06, "loss": 0.1607, "step": 3545 }, { "epoch": 0.14, "learning_rate": 9.540692559912564e-06, "loss": 0.1341, "step": 3546 }, { "epoch": 0.14, "learning_rate": 9.540559272785435e-06, "loss": 0.1475, "step": 3547 }, { "epoch": 0.14, "learning_rate": 9.540425985658306e-06, "loss": 0.1369, "step": 3548 }, { "epoch": 0.14, "learning_rate": 9.540292698531176e-06, "loss": 0.1402, "step": 3549 }, { "epoch": 0.14, "learning_rate": 9.540159411404048e-06, "loss": 0.1222, "step": 3550 }, { "epoch": 0.14, "learning_rate": 9.540026124276918e-06, "loss": 0.2039, "step": 3551 }, { "epoch": 0.14, "learning_rate": 9.539892837149789e-06, "loss": 0.0672, "step": 3552 }, { "epoch": 0.14, "learning_rate": 9.53975955002266e-06, "loss": 0.1225, "step": 3553 }, { "epoch": 0.14, "learning_rate": 9.53962626289553e-06, "loss": 0.1477, "step": 3554 }, { "epoch": 0.14, "learning_rate": 9.539492975768402e-06, "loss": 0.092, "step": 3555 }, { "epoch": 0.14, "learning_rate": 9.539359688641273e-06, "loss": 0.1416, "step": 3556 }, { "epoch": 0.14, "learning_rate": 9.539226401514142e-06, "loss": 0.1612, "step": 3557 }, { "epoch": 0.14, "learning_rate": 9.539093114387013e-06, "loss": 0.1252, "step": 3558 }, { "epoch": 0.14, "learning_rate": 9.538959827259884e-06, "loss": 0.1134, "step": 3559 }, { "epoch": 0.14, "learning_rate": 9.538826540132755e-06, "loss": 0.1571, "step": 3560 }, { "epoch": 0.14, "learning_rate": 9.538693253005626e-06, "loss": 0.1746, "step": 3561 }, { "epoch": 0.14, "learning_rate": 9.538559965878497e-06, "loss": 0.1206, "step": 3562 }, { "epoch": 0.14, "learning_rate": 9.538426678751366e-06, "loss": 0.1103, "step": 3563 }, { "epoch": 0.14, "learning_rate": 9.538293391624239e-06, "loss": 0.1274, "step": 3564 }, { "epoch": 0.14, "learning_rate": 9.538160104497108e-06, "loss": 0.1239, "step": 3565 }, { "epoch": 0.14, "learning_rate": 9.538026817369979e-06, "loss": 0.1392, "step": 3566 }, { "epoch": 0.14, "learning_rate": 9.53789353024285e-06, "loss": 0.1105, "step": 3567 }, { "epoch": 0.14, "learning_rate": 9.53776024311572e-06, "loss": 0.106, "step": 3568 }, { "epoch": 0.14, "learning_rate": 9.537626955988592e-06, "loss": 0.1476, "step": 3569 }, { "epoch": 0.14, "learning_rate": 9.537493668861461e-06, "loss": 0.0957, "step": 3570 }, { "epoch": 0.14, "learning_rate": 9.537360381734332e-06, "loss": 0.1111, "step": 3571 }, { "epoch": 0.14, "learning_rate": 9.537227094607203e-06, "loss": 0.1153, "step": 3572 }, { "epoch": 0.14, "learning_rate": 9.537093807480074e-06, "loss": 0.1095, "step": 3573 }, { "epoch": 0.14, "learning_rate": 9.536960520352945e-06, "loss": 0.1091, "step": 3574 }, { "epoch": 0.14, "learning_rate": 9.536827233225816e-06, "loss": 0.1485, "step": 3575 }, { "epoch": 0.14, "learning_rate": 9.536693946098686e-06, "loss": 0.1298, "step": 3576 }, { "epoch": 0.14, "learning_rate": 9.536560658971558e-06, "loss": 0.1125, "step": 3577 }, { "epoch": 0.14, "learning_rate": 9.536427371844428e-06, "loss": 0.0743, "step": 3578 }, { "epoch": 0.14, "learning_rate": 9.536294084717299e-06, "loss": 0.1306, "step": 3579 }, { "epoch": 0.14, "learning_rate": 9.53616079759017e-06, "loss": 0.0939, "step": 3580 }, { "epoch": 0.14, "learning_rate": 9.53602751046304e-06, "loss": 0.1505, "step": 3581 }, { "epoch": 0.14, "learning_rate": 9.535894223335912e-06, "loss": 0.1361, "step": 3582 }, { "epoch": 0.14, "learning_rate": 9.535760936208783e-06, "loss": 0.0859, "step": 3583 }, { "epoch": 0.14, "learning_rate": 9.535627649081652e-06, "loss": 0.1323, "step": 3584 }, { "epoch": 0.14, "learning_rate": 9.535494361954523e-06, "loss": 0.1112, "step": 3585 }, { "epoch": 0.14, "learning_rate": 9.535361074827394e-06, "loss": 0.1105, "step": 3586 }, { "epoch": 0.14, "learning_rate": 9.535227787700265e-06, "loss": 0.1031, "step": 3587 }, { "epoch": 0.14, "learning_rate": 9.535094500573136e-06, "loss": 0.1231, "step": 3588 }, { "epoch": 0.14, "learning_rate": 9.534961213446005e-06, "loss": 0.0917, "step": 3589 }, { "epoch": 0.14, "learning_rate": 9.534827926318876e-06, "loss": 0.1189, "step": 3590 }, { "epoch": 0.14, "learning_rate": 9.534694639191747e-06, "loss": 0.1284, "step": 3591 }, { "epoch": 0.14, "learning_rate": 9.534561352064618e-06, "loss": 0.1239, "step": 3592 }, { "epoch": 0.14, "learning_rate": 9.53442806493749e-06, "loss": 0.2135, "step": 3593 }, { "epoch": 0.14, "learning_rate": 9.53429477781036e-06, "loss": 0.0916, "step": 3594 }, { "epoch": 0.14, "learning_rate": 9.53416149068323e-06, "loss": 0.1015, "step": 3595 }, { "epoch": 0.14, "learning_rate": 9.534028203556102e-06, "loss": 0.1774, "step": 3596 }, { "epoch": 0.14, "learning_rate": 9.533894916428971e-06, "loss": 0.104, "step": 3597 }, { "epoch": 0.14, "learning_rate": 9.533761629301842e-06, "loss": 0.1051, "step": 3598 }, { "epoch": 0.14, "learning_rate": 9.533628342174713e-06, "loss": 0.1217, "step": 3599 }, { "epoch": 0.14, "learning_rate": 9.533495055047584e-06, "loss": 0.1386, "step": 3600 }, { "epoch": 0.14, "learning_rate": 9.533361767920455e-06, "loss": 0.0939, "step": 3601 }, { "epoch": 0.14, "learning_rate": 9.533228480793326e-06, "loss": 0.0847, "step": 3602 }, { "epoch": 0.14, "learning_rate": 9.533095193666196e-06, "loss": 0.1035, "step": 3603 }, { "epoch": 0.14, "learning_rate": 9.532961906539068e-06, "loss": 0.1247, "step": 3604 }, { "epoch": 0.14, "learning_rate": 9.532828619411938e-06, "loss": 0.0824, "step": 3605 }, { "epoch": 0.14, "learning_rate": 9.532695332284809e-06, "loss": 0.1209, "step": 3606 }, { "epoch": 0.14, "learning_rate": 9.53256204515768e-06, "loss": 0.1523, "step": 3607 }, { "epoch": 0.14, "learning_rate": 9.532428758030549e-06, "loss": 0.1493, "step": 3608 }, { "epoch": 0.14, "learning_rate": 9.532295470903422e-06, "loss": 0.0864, "step": 3609 }, { "epoch": 0.14, "learning_rate": 9.532162183776291e-06, "loss": 0.1256, "step": 3610 }, { "epoch": 0.14, "learning_rate": 9.532028896649162e-06, "loss": 0.1208, "step": 3611 }, { "epoch": 0.14, "learning_rate": 9.531895609522033e-06, "loss": 0.1056, "step": 3612 }, { "epoch": 0.14, "learning_rate": 9.531762322394904e-06, "loss": 0.0933, "step": 3613 }, { "epoch": 0.14, "learning_rate": 9.531629035267775e-06, "loss": 0.1012, "step": 3614 }, { "epoch": 0.14, "learning_rate": 9.531495748140646e-06, "loss": 0.1096, "step": 3615 }, { "epoch": 0.14, "learning_rate": 9.531362461013515e-06, "loss": 0.0875, "step": 3616 }, { "epoch": 0.14, "learning_rate": 9.531229173886388e-06, "loss": 0.1069, "step": 3617 }, { "epoch": 0.14, "learning_rate": 9.531095886759257e-06, "loss": 0.1023, "step": 3618 }, { "epoch": 0.14, "learning_rate": 9.530962599632128e-06, "loss": 0.132, "step": 3619 }, { "epoch": 0.14, "learning_rate": 9.530829312505e-06, "loss": 0.0902, "step": 3620 }, { "epoch": 0.14, "learning_rate": 9.53069602537787e-06, "loss": 0.0885, "step": 3621 }, { "epoch": 0.14, "learning_rate": 9.530562738250741e-06, "loss": 0.1298, "step": 3622 }, { "epoch": 0.14, "learning_rate": 9.530429451123612e-06, "loss": 0.091, "step": 3623 }, { "epoch": 0.14, "learning_rate": 9.530296163996482e-06, "loss": 0.1368, "step": 3624 }, { "epoch": 0.14, "learning_rate": 9.530162876869353e-06, "loss": 0.1479, "step": 3625 }, { "epoch": 0.14, "learning_rate": 9.530029589742224e-06, "loss": 0.1096, "step": 3626 }, { "epoch": 0.14, "learning_rate": 9.529896302615095e-06, "loss": 0.1031, "step": 3627 }, { "epoch": 0.14, "learning_rate": 9.529763015487966e-06, "loss": 0.082, "step": 3628 }, { "epoch": 0.14, "learning_rate": 9.529629728360835e-06, "loss": 0.1076, "step": 3629 }, { "epoch": 0.14, "learning_rate": 9.529496441233706e-06, "loss": 0.1481, "step": 3630 }, { "epoch": 0.14, "learning_rate": 9.529363154106577e-06, "loss": 0.1302, "step": 3631 }, { "epoch": 0.15, "learning_rate": 9.529229866979448e-06, "loss": 0.1023, "step": 3632 }, { "epoch": 0.15, "learning_rate": 9.529096579852319e-06, "loss": 0.1328, "step": 3633 }, { "epoch": 0.15, "learning_rate": 9.52896329272519e-06, "loss": 0.125, "step": 3634 }, { "epoch": 0.15, "learning_rate": 9.528830005598059e-06, "loss": 0.1498, "step": 3635 }, { "epoch": 0.15, "learning_rate": 9.528696718470932e-06, "loss": 0.1439, "step": 3636 }, { "epoch": 0.15, "learning_rate": 9.528563431343801e-06, "loss": 0.0935, "step": 3637 }, { "epoch": 0.15, "learning_rate": 9.528430144216672e-06, "loss": 0.1265, "step": 3638 }, { "epoch": 0.15, "learning_rate": 9.528296857089543e-06, "loss": 0.0791, "step": 3639 }, { "epoch": 0.15, "learning_rate": 9.528163569962414e-06, "loss": 0.1002, "step": 3640 }, { "epoch": 0.15, "learning_rate": 9.528030282835285e-06, "loss": 0.1422, "step": 3641 }, { "epoch": 0.15, "learning_rate": 9.527896995708156e-06, "loss": 0.1323, "step": 3642 }, { "epoch": 0.15, "learning_rate": 9.527763708581025e-06, "loss": 0.1027, "step": 3643 }, { "epoch": 0.15, "learning_rate": 9.527630421453898e-06, "loss": 0.1082, "step": 3644 }, { "epoch": 0.15, "learning_rate": 9.527497134326767e-06, "loss": 0.118, "step": 3645 }, { "epoch": 0.15, "learning_rate": 9.527363847199638e-06, "loss": 0.132, "step": 3646 }, { "epoch": 0.15, "learning_rate": 9.52723056007251e-06, "loss": 0.0966, "step": 3647 }, { "epoch": 0.15, "learning_rate": 9.527097272945379e-06, "loss": 0.1234, "step": 3648 }, { "epoch": 0.15, "learning_rate": 9.526963985818251e-06, "loss": 0.131, "step": 3649 }, { "epoch": 0.15, "learning_rate": 9.52683069869112e-06, "loss": 0.1144, "step": 3650 }, { "epoch": 0.15, "learning_rate": 9.526697411563992e-06, "loss": 0.1546, "step": 3651 }, { "epoch": 0.15, "learning_rate": 9.526564124436863e-06, "loss": 0.1495, "step": 3652 }, { "epoch": 0.15, "learning_rate": 9.526430837309734e-06, "loss": 0.0921, "step": 3653 }, { "epoch": 0.15, "learning_rate": 9.526297550182605e-06, "loss": 0.1341, "step": 3654 }, { "epoch": 0.15, "learning_rate": 9.526164263055476e-06, "loss": 0.1039, "step": 3655 }, { "epoch": 0.15, "learning_rate": 9.526030975928345e-06, "loss": 0.1147, "step": 3656 }, { "epoch": 0.15, "learning_rate": 9.525897688801216e-06, "loss": 0.0899, "step": 3657 }, { "epoch": 0.15, "learning_rate": 9.525764401674087e-06, "loss": 0.1039, "step": 3658 }, { "epoch": 0.15, "learning_rate": 9.525631114546958e-06, "loss": 0.1228, "step": 3659 }, { "epoch": 0.15, "learning_rate": 9.525497827419829e-06, "loss": 0.1062, "step": 3660 }, { "epoch": 0.15, "learning_rate": 9.5253645402927e-06, "loss": 0.0837, "step": 3661 }, { "epoch": 0.15, "learning_rate": 9.52523125316557e-06, "loss": 0.1274, "step": 3662 }, { "epoch": 0.15, "learning_rate": 9.525097966038442e-06, "loss": 0.1274, "step": 3663 }, { "epoch": 0.15, "learning_rate": 9.524964678911311e-06, "loss": 0.1261, "step": 3664 }, { "epoch": 0.15, "learning_rate": 9.524831391784182e-06, "loss": 0.0804, "step": 3665 }, { "epoch": 0.15, "learning_rate": 9.524698104657053e-06, "loss": 0.0968, "step": 3666 }, { "epoch": 0.15, "learning_rate": 9.524564817529922e-06, "loss": 0.1584, "step": 3667 }, { "epoch": 0.15, "learning_rate": 9.524431530402795e-06, "loss": 0.1142, "step": 3668 }, { "epoch": 0.15, "learning_rate": 9.524298243275664e-06, "loss": 0.1305, "step": 3669 }, { "epoch": 0.15, "learning_rate": 9.524164956148535e-06, "loss": 0.1562, "step": 3670 }, { "epoch": 0.15, "learning_rate": 9.524031669021406e-06, "loss": 0.1163, "step": 3671 }, { "epoch": 0.15, "learning_rate": 9.523898381894277e-06, "loss": 0.1114, "step": 3672 }, { "epoch": 0.15, "learning_rate": 9.523765094767148e-06, "loss": 0.1143, "step": 3673 }, { "epoch": 0.15, "learning_rate": 9.52363180764002e-06, "loss": 0.1158, "step": 3674 }, { "epoch": 0.15, "learning_rate": 9.523498520512889e-06, "loss": 0.1172, "step": 3675 }, { "epoch": 0.15, "learning_rate": 9.523365233385761e-06, "loss": 0.1285, "step": 3676 }, { "epoch": 0.15, "learning_rate": 9.52323194625863e-06, "loss": 0.1377, "step": 3677 }, { "epoch": 0.15, "learning_rate": 9.523098659131502e-06, "loss": 0.1004, "step": 3678 }, { "epoch": 0.15, "learning_rate": 9.522965372004373e-06, "loss": 0.0921, "step": 3679 }, { "epoch": 0.15, "learning_rate": 9.522832084877244e-06, "loss": 0.1054, "step": 3680 }, { "epoch": 0.15, "learning_rate": 9.522698797750115e-06, "loss": 0.1743, "step": 3681 }, { "epoch": 0.15, "learning_rate": 9.522565510622986e-06, "loss": 0.1025, "step": 3682 }, { "epoch": 0.15, "learning_rate": 9.522432223495855e-06, "loss": 0.1234, "step": 3683 }, { "epoch": 0.15, "learning_rate": 9.522298936368728e-06, "loss": 0.1014, "step": 3684 }, { "epoch": 0.15, "learning_rate": 9.522165649241597e-06, "loss": 0.1362, "step": 3685 }, { "epoch": 0.15, "learning_rate": 9.522032362114468e-06, "loss": 0.1021, "step": 3686 }, { "epoch": 0.15, "learning_rate": 9.521899074987339e-06, "loss": 0.1668, "step": 3687 }, { "epoch": 0.15, "learning_rate": 9.521765787860208e-06, "loss": 0.1417, "step": 3688 }, { "epoch": 0.15, "learning_rate": 9.521632500733081e-06, "loss": 0.1624, "step": 3689 }, { "epoch": 0.15, "learning_rate": 9.52149921360595e-06, "loss": 0.138, "step": 3690 }, { "epoch": 0.15, "learning_rate": 9.521365926478821e-06, "loss": 0.1158, "step": 3691 }, { "epoch": 0.15, "learning_rate": 9.521232639351692e-06, "loss": 0.1025, "step": 3692 }, { "epoch": 0.15, "learning_rate": 9.521099352224563e-06, "loss": 0.1138, "step": 3693 }, { "epoch": 0.15, "learning_rate": 9.520966065097434e-06, "loss": 0.1464, "step": 3694 }, { "epoch": 0.15, "learning_rate": 9.520832777970305e-06, "loss": 0.1399, "step": 3695 }, { "epoch": 0.15, "learning_rate": 9.520699490843175e-06, "loss": 0.1165, "step": 3696 }, { "epoch": 0.15, "learning_rate": 9.520566203716046e-06, "loss": 0.0987, "step": 3697 }, { "epoch": 0.15, "learning_rate": 9.520432916588917e-06, "loss": 0.1155, "step": 3698 }, { "epoch": 0.15, "learning_rate": 9.520299629461788e-06, "loss": 0.1019, "step": 3699 }, { "epoch": 0.15, "learning_rate": 9.520166342334658e-06, "loss": 0.0793, "step": 3700 }, { "epoch": 0.15, "learning_rate": 9.52003305520753e-06, "loss": 0.1471, "step": 3701 }, { "epoch": 0.15, "learning_rate": 9.519899768080399e-06, "loss": 0.1224, "step": 3702 }, { "epoch": 0.15, "learning_rate": 9.519766480953271e-06, "loss": 0.0657, "step": 3703 }, { "epoch": 0.15, "learning_rate": 9.51963319382614e-06, "loss": 0.143, "step": 3704 }, { "epoch": 0.15, "learning_rate": 9.519499906699012e-06, "loss": 0.1104, "step": 3705 }, { "epoch": 0.15, "learning_rate": 9.519366619571883e-06, "loss": 0.1437, "step": 3706 }, { "epoch": 0.15, "learning_rate": 9.519233332444752e-06, "loss": 0.1747, "step": 3707 }, { "epoch": 0.15, "learning_rate": 9.519100045317625e-06, "loss": 0.1437, "step": 3708 }, { "epoch": 0.15, "learning_rate": 9.518966758190494e-06, "loss": 0.1514, "step": 3709 }, { "epoch": 0.15, "learning_rate": 9.518833471063365e-06, "loss": 0.1194, "step": 3710 }, { "epoch": 0.15, "learning_rate": 9.518700183936236e-06, "loss": 0.121, "step": 3711 }, { "epoch": 0.15, "learning_rate": 9.518566896809107e-06, "loss": 0.0959, "step": 3712 }, { "epoch": 0.15, "learning_rate": 9.518433609681978e-06, "loss": 0.1371, "step": 3713 }, { "epoch": 0.15, "learning_rate": 9.518300322554849e-06, "loss": 0.111, "step": 3714 }, { "epoch": 0.15, "learning_rate": 9.518167035427718e-06, "loss": 0.0909, "step": 3715 }, { "epoch": 0.15, "learning_rate": 9.518033748300591e-06, "loss": 0.1287, "step": 3716 }, { "epoch": 0.15, "learning_rate": 9.51790046117346e-06, "loss": 0.0917, "step": 3717 }, { "epoch": 0.15, "learning_rate": 9.517767174046331e-06, "loss": 0.1308, "step": 3718 }, { "epoch": 0.15, "learning_rate": 9.517633886919202e-06, "loss": 0.1311, "step": 3719 }, { "epoch": 0.15, "learning_rate": 9.517500599792073e-06, "loss": 0.1327, "step": 3720 }, { "epoch": 0.15, "learning_rate": 9.517367312664944e-06, "loss": 0.1371, "step": 3721 }, { "epoch": 0.15, "learning_rate": 9.517234025537815e-06, "loss": 0.11, "step": 3722 }, { "epoch": 0.15, "learning_rate": 9.517100738410685e-06, "loss": 0.1421, "step": 3723 }, { "epoch": 0.15, "learning_rate": 9.516967451283556e-06, "loss": 0.1925, "step": 3724 }, { "epoch": 0.15, "learning_rate": 9.516834164156427e-06, "loss": 0.1617, "step": 3725 }, { "epoch": 0.15, "learning_rate": 9.516700877029298e-06, "loss": 0.1261, "step": 3726 }, { "epoch": 0.15, "learning_rate": 9.516567589902169e-06, "loss": 0.1043, "step": 3727 }, { "epoch": 0.15, "learning_rate": 9.516434302775038e-06, "loss": 0.1466, "step": 3728 }, { "epoch": 0.15, "learning_rate": 9.516301015647909e-06, "loss": 0.1294, "step": 3729 }, { "epoch": 0.15, "learning_rate": 9.51616772852078e-06, "loss": 0.1059, "step": 3730 }, { "epoch": 0.15, "learning_rate": 9.516034441393651e-06, "loss": 0.0884, "step": 3731 }, { "epoch": 0.15, "learning_rate": 9.515901154266522e-06, "loss": 0.0979, "step": 3732 }, { "epoch": 0.15, "learning_rate": 9.515767867139393e-06, "loss": 0.1256, "step": 3733 }, { "epoch": 0.15, "learning_rate": 9.515634580012262e-06, "loss": 0.1128, "step": 3734 }, { "epoch": 0.15, "learning_rate": 9.515501292885135e-06, "loss": 0.1331, "step": 3735 }, { "epoch": 0.15, "learning_rate": 9.515368005758004e-06, "loss": 0.1233, "step": 3736 }, { "epoch": 0.15, "learning_rate": 9.515234718630875e-06, "loss": 0.1511, "step": 3737 }, { "epoch": 0.15, "learning_rate": 9.515101431503746e-06, "loss": 0.1001, "step": 3738 }, { "epoch": 0.15, "learning_rate": 9.514968144376617e-06, "loss": 0.1286, "step": 3739 }, { "epoch": 0.15, "learning_rate": 9.514834857249488e-06, "loss": 0.1372, "step": 3740 }, { "epoch": 0.15, "learning_rate": 9.514701570122359e-06, "loss": 0.1122, "step": 3741 }, { "epoch": 0.15, "learning_rate": 9.514568282995228e-06, "loss": 0.0951, "step": 3742 }, { "epoch": 0.15, "learning_rate": 9.514434995868101e-06, "loss": 0.1289, "step": 3743 }, { "epoch": 0.15, "learning_rate": 9.51430170874097e-06, "loss": 0.1141, "step": 3744 }, { "epoch": 0.15, "learning_rate": 9.514168421613841e-06, "loss": 0.1185, "step": 3745 }, { "epoch": 0.15, "learning_rate": 9.514035134486712e-06, "loss": 0.1073, "step": 3746 }, { "epoch": 0.15, "learning_rate": 9.513901847359582e-06, "loss": 0.0911, "step": 3747 }, { "epoch": 0.15, "learning_rate": 9.513768560232454e-06, "loss": 0.0983, "step": 3748 }, { "epoch": 0.15, "learning_rate": 9.513635273105324e-06, "loss": 0.0923, "step": 3749 }, { "epoch": 0.15, "learning_rate": 9.513501985978195e-06, "loss": 0.1269, "step": 3750 }, { "epoch": 0.15, "learning_rate": 9.513368698851066e-06, "loss": 0.095, "step": 3751 }, { "epoch": 0.15, "learning_rate": 9.513235411723937e-06, "loss": 0.0954, "step": 3752 }, { "epoch": 0.15, "learning_rate": 9.513102124596808e-06, "loss": 0.1787, "step": 3753 }, { "epoch": 0.15, "learning_rate": 9.512968837469679e-06, "loss": 0.1324, "step": 3754 }, { "epoch": 0.15, "learning_rate": 9.512835550342548e-06, "loss": 0.1024, "step": 3755 }, { "epoch": 0.15, "learning_rate": 9.512702263215419e-06, "loss": 0.1415, "step": 3756 }, { "epoch": 0.15, "learning_rate": 9.51256897608829e-06, "loss": 0.1117, "step": 3757 }, { "epoch": 0.15, "learning_rate": 9.512435688961161e-06, "loss": 0.1486, "step": 3758 }, { "epoch": 0.15, "learning_rate": 9.512302401834032e-06, "loss": 0.1115, "step": 3759 }, { "epoch": 0.15, "learning_rate": 9.512169114706903e-06, "loss": 0.1232, "step": 3760 }, { "epoch": 0.15, "learning_rate": 9.512035827579772e-06, "loss": 0.1327, "step": 3761 }, { "epoch": 0.15, "learning_rate": 9.511902540452645e-06, "loss": 0.1153, "step": 3762 }, { "epoch": 0.15, "learning_rate": 9.511769253325514e-06, "loss": 0.1457, "step": 3763 }, { "epoch": 0.15, "learning_rate": 9.511635966198385e-06, "loss": 0.1461, "step": 3764 }, { "epoch": 0.15, "learning_rate": 9.511502679071256e-06, "loss": 0.0777, "step": 3765 }, { "epoch": 0.15, "learning_rate": 9.511369391944126e-06, "loss": 0.0824, "step": 3766 }, { "epoch": 0.15, "learning_rate": 9.511236104816998e-06, "loss": 0.118, "step": 3767 }, { "epoch": 0.15, "learning_rate": 9.511102817689868e-06, "loss": 0.1425, "step": 3768 }, { "epoch": 0.15, "learning_rate": 9.510969530562738e-06, "loss": 0.1219, "step": 3769 }, { "epoch": 0.15, "learning_rate": 9.51083624343561e-06, "loss": 0.099, "step": 3770 }, { "epoch": 0.15, "learning_rate": 9.51070295630848e-06, "loss": 0.1291, "step": 3771 }, { "epoch": 0.15, "learning_rate": 9.510569669181351e-06, "loss": 0.1318, "step": 3772 }, { "epoch": 0.15, "learning_rate": 9.510436382054222e-06, "loss": 0.0826, "step": 3773 }, { "epoch": 0.15, "learning_rate": 9.510303094927092e-06, "loss": 0.1299, "step": 3774 }, { "epoch": 0.15, "learning_rate": 9.510169807799964e-06, "loss": 0.1008, "step": 3775 }, { "epoch": 0.15, "learning_rate": 9.510036520672834e-06, "loss": 0.1241, "step": 3776 }, { "epoch": 0.15, "learning_rate": 9.509903233545705e-06, "loss": 0.079, "step": 3777 }, { "epoch": 0.15, "learning_rate": 9.509769946418576e-06, "loss": 0.1035, "step": 3778 }, { "epoch": 0.15, "learning_rate": 9.509636659291447e-06, "loss": 0.1319, "step": 3779 }, { "epoch": 0.15, "learning_rate": 9.509503372164318e-06, "loss": 0.1143, "step": 3780 }, { "epoch": 0.15, "learning_rate": 9.509370085037189e-06, "loss": 0.167, "step": 3781 }, { "epoch": 0.15, "learning_rate": 9.509236797910058e-06, "loss": 0.0881, "step": 3782 }, { "epoch": 0.15, "learning_rate": 9.509103510782929e-06, "loss": 0.1372, "step": 3783 }, { "epoch": 0.15, "learning_rate": 9.5089702236558e-06, "loss": 0.1121, "step": 3784 }, { "epoch": 0.15, "learning_rate": 9.508836936528671e-06, "loss": 0.1344, "step": 3785 }, { "epoch": 0.15, "learning_rate": 9.508703649401542e-06, "loss": 0.1581, "step": 3786 }, { "epoch": 0.15, "learning_rate": 9.508570362274411e-06, "loss": 0.1131, "step": 3787 }, { "epoch": 0.15, "learning_rate": 9.508437075147284e-06, "loss": 0.1165, "step": 3788 }, { "epoch": 0.15, "learning_rate": 9.508303788020153e-06, "loss": 0.1312, "step": 3789 }, { "epoch": 0.15, "learning_rate": 9.508170500893024e-06, "loss": 0.1204, "step": 3790 }, { "epoch": 0.15, "learning_rate": 9.508037213765895e-06, "loss": 0.1173, "step": 3791 }, { "epoch": 0.15, "learning_rate": 9.507903926638766e-06, "loss": 0.1023, "step": 3792 }, { "epoch": 0.15, "learning_rate": 9.507770639511637e-06, "loss": 0.1007, "step": 3793 }, { "epoch": 0.15, "learning_rate": 9.507637352384508e-06, "loss": 0.0975, "step": 3794 }, { "epoch": 0.15, "learning_rate": 9.507504065257378e-06, "loss": 0.1068, "step": 3795 }, { "epoch": 0.15, "learning_rate": 9.507370778130249e-06, "loss": 0.1376, "step": 3796 }, { "epoch": 0.15, "learning_rate": 9.50723749100312e-06, "loss": 0.1017, "step": 3797 }, { "epoch": 0.15, "learning_rate": 9.50710420387599e-06, "loss": 0.1787, "step": 3798 }, { "epoch": 0.15, "learning_rate": 9.506970916748862e-06, "loss": 0.1194, "step": 3799 }, { "epoch": 0.15, "learning_rate": 9.506837629621733e-06, "loss": 0.0881, "step": 3800 }, { "epoch": 0.15, "learning_rate": 9.506704342494602e-06, "loss": 0.1108, "step": 3801 }, { "epoch": 0.15, "learning_rate": 9.506571055367475e-06, "loss": 0.1029, "step": 3802 }, { "epoch": 0.15, "learning_rate": 9.506437768240344e-06, "loss": 0.1538, "step": 3803 }, { "epoch": 0.15, "learning_rate": 9.506304481113215e-06, "loss": 0.1236, "step": 3804 }, { "epoch": 0.15, "learning_rate": 9.506171193986086e-06, "loss": 0.1386, "step": 3805 }, { "epoch": 0.15, "learning_rate": 9.506037906858955e-06, "loss": 0.1229, "step": 3806 }, { "epoch": 0.15, "learning_rate": 9.505904619731828e-06, "loss": 0.122, "step": 3807 }, { "epoch": 0.15, "learning_rate": 9.505771332604697e-06, "loss": 0.108, "step": 3808 }, { "epoch": 0.15, "learning_rate": 9.505638045477568e-06, "loss": 0.1086, "step": 3809 }, { "epoch": 0.15, "learning_rate": 9.505504758350439e-06, "loss": 0.1052, "step": 3810 }, { "epoch": 0.15, "learning_rate": 9.50537147122331e-06, "loss": 0.1117, "step": 3811 }, { "epoch": 0.15, "learning_rate": 9.505238184096181e-06, "loss": 0.0724, "step": 3812 }, { "epoch": 0.15, "learning_rate": 9.505104896969052e-06, "loss": 0.1293, "step": 3813 }, { "epoch": 0.15, "learning_rate": 9.504971609841921e-06, "loss": 0.0936, "step": 3814 }, { "epoch": 0.15, "learning_rate": 9.504838322714794e-06, "loss": 0.1421, "step": 3815 }, { "epoch": 0.15, "learning_rate": 9.504705035587663e-06, "loss": 0.1509, "step": 3816 }, { "epoch": 0.15, "learning_rate": 9.504571748460534e-06, "loss": 0.1472, "step": 3817 }, { "epoch": 0.15, "learning_rate": 9.504438461333405e-06, "loss": 0.1722, "step": 3818 }, { "epoch": 0.15, "learning_rate": 9.504305174206276e-06, "loss": 0.1599, "step": 3819 }, { "epoch": 0.15, "learning_rate": 9.504171887079147e-06, "loss": 0.1273, "step": 3820 }, { "epoch": 0.15, "learning_rate": 9.504038599952018e-06, "loss": 0.1369, "step": 3821 }, { "epoch": 0.15, "learning_rate": 9.503905312824888e-06, "loss": 0.0965, "step": 3822 }, { "epoch": 0.15, "learning_rate": 9.503772025697759e-06, "loss": 0.1176, "step": 3823 }, { "epoch": 0.15, "learning_rate": 9.50363873857063e-06, "loss": 0.1351, "step": 3824 }, { "epoch": 0.15, "learning_rate": 9.5035054514435e-06, "loss": 0.0837, "step": 3825 }, { "epoch": 0.15, "learning_rate": 9.503372164316372e-06, "loss": 0.1186, "step": 3826 }, { "epoch": 0.15, "learning_rate": 9.503238877189241e-06, "loss": 0.1187, "step": 3827 }, { "epoch": 0.15, "learning_rate": 9.503105590062112e-06, "loss": 0.1206, "step": 3828 }, { "epoch": 0.15, "learning_rate": 9.502972302934983e-06, "loss": 0.0987, "step": 3829 }, { "epoch": 0.15, "learning_rate": 9.502839015807854e-06, "loss": 0.1133, "step": 3830 }, { "epoch": 0.15, "learning_rate": 9.502705728680725e-06, "loss": 0.1059, "step": 3831 }, { "epoch": 0.15, "learning_rate": 9.502572441553596e-06, "loss": 0.1, "step": 3832 }, { "epoch": 0.15, "learning_rate": 9.502439154426465e-06, "loss": 0.1011, "step": 3833 }, { "epoch": 0.15, "learning_rate": 9.502305867299338e-06, "loss": 0.1001, "step": 3834 }, { "epoch": 0.15, "learning_rate": 9.502172580172207e-06, "loss": 0.075, "step": 3835 }, { "epoch": 0.15, "learning_rate": 9.502039293045078e-06, "loss": 0.1303, "step": 3836 }, { "epoch": 0.15, "learning_rate": 9.50190600591795e-06, "loss": 0.1384, "step": 3837 }, { "epoch": 0.15, "learning_rate": 9.50177271879082e-06, "loss": 0.0961, "step": 3838 }, { "epoch": 0.15, "learning_rate": 9.501639431663691e-06, "loss": 0.1278, "step": 3839 }, { "epoch": 0.15, "learning_rate": 9.501506144536562e-06, "loss": 0.1463, "step": 3840 }, { "epoch": 0.15, "learning_rate": 9.501372857409431e-06, "loss": 0.1162, "step": 3841 }, { "epoch": 0.15, "learning_rate": 9.501239570282302e-06, "loss": 0.0809, "step": 3842 }, { "epoch": 0.15, "learning_rate": 9.501106283155173e-06, "loss": 0.1305, "step": 3843 }, { "epoch": 0.15, "learning_rate": 9.500972996028044e-06, "loss": 0.0905, "step": 3844 }, { "epoch": 0.15, "learning_rate": 9.500839708900915e-06, "loss": 0.1223, "step": 3845 }, { "epoch": 0.15, "learning_rate": 9.500706421773785e-06, "loss": 0.1245, "step": 3846 }, { "epoch": 0.15, "learning_rate": 9.500573134646657e-06, "loss": 0.1205, "step": 3847 }, { "epoch": 0.15, "learning_rate": 9.500439847519527e-06, "loss": 0.119, "step": 3848 }, { "epoch": 0.15, "learning_rate": 9.500306560392398e-06, "loss": 0.0886, "step": 3849 }, { "epoch": 0.15, "learning_rate": 9.500173273265269e-06, "loss": 0.1018, "step": 3850 }, { "epoch": 0.15, "learning_rate": 9.50003998613814e-06, "loss": 0.1404, "step": 3851 }, { "epoch": 0.15, "learning_rate": 9.49990669901101e-06, "loss": 0.1009, "step": 3852 }, { "epoch": 0.15, "learning_rate": 9.499773411883882e-06, "loss": 0.0963, "step": 3853 }, { "epoch": 0.15, "learning_rate": 9.499640124756751e-06, "loss": 0.1296, "step": 3854 }, { "epoch": 0.15, "learning_rate": 9.499506837629624e-06, "loss": 0.1098, "step": 3855 }, { "epoch": 0.15, "learning_rate": 9.499373550502493e-06, "loss": 0.1271, "step": 3856 }, { "epoch": 0.15, "learning_rate": 9.499240263375364e-06, "loss": 0.1117, "step": 3857 }, { "epoch": 0.15, "learning_rate": 9.499106976248235e-06, "loss": 0.113, "step": 3858 }, { "epoch": 0.15, "learning_rate": 9.498973689121106e-06, "loss": 0.0944, "step": 3859 }, { "epoch": 0.15, "learning_rate": 9.498840401993977e-06, "loss": 0.1082, "step": 3860 }, { "epoch": 0.15, "learning_rate": 9.498707114866848e-06, "loss": 0.1781, "step": 3861 }, { "epoch": 0.15, "learning_rate": 9.498573827739717e-06, "loss": 0.1151, "step": 3862 }, { "epoch": 0.15, "learning_rate": 9.498440540612588e-06, "loss": 0.1177, "step": 3863 }, { "epoch": 0.15, "learning_rate": 9.49830725348546e-06, "loss": 0.0949, "step": 3864 }, { "epoch": 0.15, "learning_rate": 9.49817396635833e-06, "loss": 0.1001, "step": 3865 }, { "epoch": 0.15, "learning_rate": 9.498040679231201e-06, "loss": 0.16, "step": 3866 }, { "epoch": 0.15, "learning_rate": 9.49790739210407e-06, "loss": 0.1245, "step": 3867 }, { "epoch": 0.15, "learning_rate": 9.497774104976942e-06, "loss": 0.1166, "step": 3868 }, { "epoch": 0.15, "learning_rate": 9.497640817849813e-06, "loss": 0.1649, "step": 3869 }, { "epoch": 0.15, "learning_rate": 9.497507530722684e-06, "loss": 0.1263, "step": 3870 }, { "epoch": 0.15, "learning_rate": 9.497374243595555e-06, "loss": 0.1202, "step": 3871 }, { "epoch": 0.15, "learning_rate": 9.497240956468426e-06, "loss": 0.1197, "step": 3872 }, { "epoch": 0.15, "learning_rate": 9.497107669341295e-06, "loss": 0.1657, "step": 3873 }, { "epoch": 0.15, "learning_rate": 9.496974382214168e-06, "loss": 0.1371, "step": 3874 }, { "epoch": 0.15, "learning_rate": 9.496841095087037e-06, "loss": 0.1337, "step": 3875 }, { "epoch": 0.15, "learning_rate": 9.496707807959908e-06, "loss": 0.0966, "step": 3876 }, { "epoch": 0.15, "learning_rate": 9.496574520832779e-06, "loss": 0.1043, "step": 3877 }, { "epoch": 0.15, "learning_rate": 9.49644123370565e-06, "loss": 0.1091, "step": 3878 }, { "epoch": 0.15, "learning_rate": 9.49630794657852e-06, "loss": 0.0881, "step": 3879 }, { "epoch": 0.15, "learning_rate": 9.496174659451392e-06, "loss": 0.0897, "step": 3880 }, { "epoch": 0.15, "learning_rate": 9.496041372324261e-06, "loss": 0.1471, "step": 3881 }, { "epoch": 0.16, "learning_rate": 9.495908085197132e-06, "loss": 0.1086, "step": 3882 }, { "epoch": 0.16, "learning_rate": 9.495774798070003e-06, "loss": 0.1181, "step": 3883 }, { "epoch": 0.16, "learning_rate": 9.495641510942874e-06, "loss": 0.1344, "step": 3884 }, { "epoch": 0.16, "learning_rate": 9.495508223815745e-06, "loss": 0.1229, "step": 3885 }, { "epoch": 0.16, "learning_rate": 9.495374936688614e-06, "loss": 0.1028, "step": 3886 }, { "epoch": 0.16, "learning_rate": 9.495241649561487e-06, "loss": 0.0991, "step": 3887 }, { "epoch": 0.16, "learning_rate": 9.495108362434356e-06, "loss": 0.11, "step": 3888 }, { "epoch": 0.16, "learning_rate": 9.494975075307227e-06, "loss": 0.1502, "step": 3889 }, { "epoch": 0.16, "learning_rate": 9.494841788180098e-06, "loss": 0.114, "step": 3890 }, { "epoch": 0.16, "learning_rate": 9.49470850105297e-06, "loss": 0.1143, "step": 3891 }, { "epoch": 0.16, "learning_rate": 9.49457521392584e-06, "loss": 0.1556, "step": 3892 }, { "epoch": 0.16, "learning_rate": 9.494441926798711e-06, "loss": 0.1236, "step": 3893 }, { "epoch": 0.16, "learning_rate": 9.49430863967158e-06, "loss": 0.0686, "step": 3894 }, { "epoch": 0.16, "learning_rate": 9.494175352544452e-06, "loss": 0.1307, "step": 3895 }, { "epoch": 0.16, "learning_rate": 9.494042065417323e-06, "loss": 0.1188, "step": 3896 }, { "epoch": 0.16, "learning_rate": 9.493908778290194e-06, "loss": 0.0958, "step": 3897 }, { "epoch": 0.16, "learning_rate": 9.493775491163065e-06, "loss": 0.1182, "step": 3898 }, { "epoch": 0.16, "learning_rate": 9.493642204035936e-06, "loss": 0.1383, "step": 3899 }, { "epoch": 0.16, "learning_rate": 9.493508916908805e-06, "loss": 0.1084, "step": 3900 }, { "epoch": 0.16, "learning_rate": 9.493375629781676e-06, "loss": 0.1001, "step": 3901 }, { "epoch": 0.16, "learning_rate": 9.493242342654547e-06, "loss": 0.1586, "step": 3902 }, { "epoch": 0.16, "learning_rate": 9.493109055527418e-06, "loss": 0.1488, "step": 3903 }, { "epoch": 0.16, "learning_rate": 9.492975768400289e-06, "loss": 0.1235, "step": 3904 }, { "epoch": 0.16, "learning_rate": 9.492842481273158e-06, "loss": 0.1381, "step": 3905 }, { "epoch": 0.16, "learning_rate": 9.492709194146031e-06, "loss": 0.073, "step": 3906 }, { "epoch": 0.16, "learning_rate": 9.4925759070189e-06, "loss": 0.1255, "step": 3907 }, { "epoch": 0.16, "learning_rate": 9.492442619891771e-06, "loss": 0.1199, "step": 3908 }, { "epoch": 0.16, "learning_rate": 9.492309332764642e-06, "loss": 0.1454, "step": 3909 }, { "epoch": 0.16, "learning_rate": 9.492176045637513e-06, "loss": 0.1158, "step": 3910 }, { "epoch": 0.16, "learning_rate": 9.492042758510384e-06, "loss": 0.0841, "step": 3911 }, { "epoch": 0.16, "learning_rate": 9.491909471383255e-06, "loss": 0.1202, "step": 3912 }, { "epoch": 0.16, "learning_rate": 9.491776184256124e-06, "loss": 0.1284, "step": 3913 }, { "epoch": 0.16, "learning_rate": 9.491642897128997e-06, "loss": 0.0858, "step": 3914 }, { "epoch": 0.16, "learning_rate": 9.491509610001866e-06, "loss": 0.0972, "step": 3915 }, { "epoch": 0.16, "learning_rate": 9.491376322874737e-06, "loss": 0.0977, "step": 3916 }, { "epoch": 0.16, "learning_rate": 9.491243035747608e-06, "loss": 0.1348, "step": 3917 }, { "epoch": 0.16, "learning_rate": 9.49110974862048e-06, "loss": 0.1377, "step": 3918 }, { "epoch": 0.16, "learning_rate": 9.49097646149335e-06, "loss": 0.1228, "step": 3919 }, { "epoch": 0.16, "learning_rate": 9.49084317436622e-06, "loss": 0.1302, "step": 3920 }, { "epoch": 0.16, "learning_rate": 9.49070988723909e-06, "loss": 0.1415, "step": 3921 }, { "epoch": 0.16, "learning_rate": 9.490576600111962e-06, "loss": 0.0971, "step": 3922 }, { "epoch": 0.16, "learning_rate": 9.490443312984833e-06, "loss": 0.0967, "step": 3923 }, { "epoch": 0.16, "learning_rate": 9.490310025857704e-06, "loss": 0.1502, "step": 3924 }, { "epoch": 0.16, "learning_rate": 9.490176738730575e-06, "loss": 0.1361, "step": 3925 }, { "epoch": 0.16, "learning_rate": 9.490043451603444e-06, "loss": 0.076, "step": 3926 }, { "epoch": 0.16, "learning_rate": 9.489910164476317e-06, "loss": 0.0815, "step": 3927 }, { "epoch": 0.16, "learning_rate": 9.489776877349186e-06, "loss": 0.1283, "step": 3928 }, { "epoch": 0.16, "learning_rate": 9.489643590222057e-06, "loss": 0.0954, "step": 3929 }, { "epoch": 0.16, "learning_rate": 9.489510303094928e-06, "loss": 0.1196, "step": 3930 }, { "epoch": 0.16, "learning_rate": 9.489377015967799e-06, "loss": 0.0953, "step": 3931 }, { "epoch": 0.16, "learning_rate": 9.48924372884067e-06, "loss": 0.1242, "step": 3932 }, { "epoch": 0.16, "learning_rate": 9.489110441713541e-06, "loss": 0.1232, "step": 3933 }, { "epoch": 0.16, "learning_rate": 9.48897715458641e-06, "loss": 0.1006, "step": 3934 }, { "epoch": 0.16, "learning_rate": 9.488843867459281e-06, "loss": 0.1136, "step": 3935 }, { "epoch": 0.16, "learning_rate": 9.488710580332152e-06, "loss": 0.132, "step": 3936 }, { "epoch": 0.16, "learning_rate": 9.488577293205023e-06, "loss": 0.1288, "step": 3937 }, { "epoch": 0.16, "learning_rate": 9.488444006077894e-06, "loss": 0.126, "step": 3938 }, { "epoch": 0.16, "learning_rate": 9.488310718950765e-06, "loss": 0.1448, "step": 3939 }, { "epoch": 0.16, "learning_rate": 9.488177431823635e-06, "loss": 0.1044, "step": 3940 }, { "epoch": 0.16, "learning_rate": 9.488044144696506e-06, "loss": 0.1256, "step": 3941 }, { "epoch": 0.16, "learning_rate": 9.487910857569377e-06, "loss": 0.0859, "step": 3942 }, { "epoch": 0.16, "learning_rate": 9.487777570442248e-06, "loss": 0.1068, "step": 3943 }, { "epoch": 0.16, "learning_rate": 9.487644283315119e-06, "loss": 0.1484, "step": 3944 }, { "epoch": 0.16, "learning_rate": 9.487510996187988e-06, "loss": 0.0754, "step": 3945 }, { "epoch": 0.16, "learning_rate": 9.48737770906086e-06, "loss": 0.1384, "step": 3946 }, { "epoch": 0.16, "learning_rate": 9.48724442193373e-06, "loss": 0.139, "step": 3947 }, { "epoch": 0.16, "learning_rate": 9.4871111348066e-06, "loss": 0.0932, "step": 3948 }, { "epoch": 0.16, "learning_rate": 9.486977847679472e-06, "loss": 0.126, "step": 3949 }, { "epoch": 0.16, "learning_rate": 9.486844560552343e-06, "loss": 0.1111, "step": 3950 }, { "epoch": 0.16, "learning_rate": 9.486711273425214e-06, "loss": 0.1286, "step": 3951 }, { "epoch": 0.16, "learning_rate": 9.486577986298085e-06, "loss": 0.1115, "step": 3952 }, { "epoch": 0.16, "learning_rate": 9.486444699170954e-06, "loss": 0.1102, "step": 3953 }, { "epoch": 0.16, "learning_rate": 9.486311412043827e-06, "loss": 0.1184, "step": 3954 }, { "epoch": 0.16, "learning_rate": 9.486178124916696e-06, "loss": 0.1265, "step": 3955 }, { "epoch": 0.16, "learning_rate": 9.486044837789567e-06, "loss": 0.1201, "step": 3956 }, { "epoch": 0.16, "learning_rate": 9.485911550662438e-06, "loss": 0.0937, "step": 3957 }, { "epoch": 0.16, "learning_rate": 9.485778263535309e-06, "loss": 0.1279, "step": 3958 }, { "epoch": 0.16, "learning_rate": 9.48564497640818e-06, "loss": 0.0993, "step": 3959 }, { "epoch": 0.16, "learning_rate": 9.48551168928105e-06, "loss": 0.0907, "step": 3960 }, { "epoch": 0.16, "learning_rate": 9.48537840215392e-06, "loss": 0.0948, "step": 3961 }, { "epoch": 0.16, "learning_rate": 9.485245115026791e-06, "loss": 0.0808, "step": 3962 }, { "epoch": 0.16, "learning_rate": 9.485111827899662e-06, "loss": 0.1256, "step": 3963 }, { "epoch": 0.16, "learning_rate": 9.484978540772533e-06, "loss": 0.0897, "step": 3964 }, { "epoch": 0.16, "learning_rate": 9.484845253645404e-06, "loss": 0.1309, "step": 3965 }, { "epoch": 0.16, "learning_rate": 9.484711966518274e-06, "loss": 0.1649, "step": 3966 }, { "epoch": 0.16, "learning_rate": 9.484578679391145e-06, "loss": 0.0932, "step": 3967 }, { "epoch": 0.16, "learning_rate": 9.484445392264016e-06, "loss": 0.1455, "step": 3968 }, { "epoch": 0.16, "learning_rate": 9.484312105136887e-06, "loss": 0.0738, "step": 3969 }, { "epoch": 0.16, "learning_rate": 9.484178818009758e-06, "loss": 0.1055, "step": 3970 }, { "epoch": 0.16, "learning_rate": 9.484045530882629e-06, "loss": 0.152, "step": 3971 }, { "epoch": 0.16, "learning_rate": 9.483912243755498e-06, "loss": 0.0963, "step": 3972 }, { "epoch": 0.16, "learning_rate": 9.48377895662837e-06, "loss": 0.1034, "step": 3973 }, { "epoch": 0.16, "learning_rate": 9.48364566950124e-06, "loss": 0.123, "step": 3974 }, { "epoch": 0.16, "learning_rate": 9.483512382374111e-06, "loss": 0.0947, "step": 3975 }, { "epoch": 0.16, "learning_rate": 9.483379095246982e-06, "loss": 0.0938, "step": 3976 }, { "epoch": 0.16, "learning_rate": 9.483245808119853e-06, "loss": 0.0975, "step": 3977 }, { "epoch": 0.16, "learning_rate": 9.483112520992724e-06, "loss": 0.1592, "step": 3978 }, { "epoch": 0.16, "learning_rate": 9.482979233865593e-06, "loss": 0.1298, "step": 3979 }, { "epoch": 0.16, "learning_rate": 9.482845946738464e-06, "loss": 0.1592, "step": 3980 }, { "epoch": 0.16, "learning_rate": 9.482712659611335e-06, "loss": 0.1235, "step": 3981 }, { "epoch": 0.16, "learning_rate": 9.482579372484206e-06, "loss": 0.1288, "step": 3982 }, { "epoch": 0.16, "learning_rate": 9.482446085357077e-06, "loss": 0.115, "step": 3983 }, { "epoch": 0.16, "learning_rate": 9.482312798229948e-06, "loss": 0.1258, "step": 3984 }, { "epoch": 0.16, "learning_rate": 9.482179511102817e-06, "loss": 0.1234, "step": 3985 }, { "epoch": 0.16, "learning_rate": 9.48204622397569e-06, "loss": 0.1239, "step": 3986 }, { "epoch": 0.16, "learning_rate": 9.48191293684856e-06, "loss": 0.0988, "step": 3987 }, { "epoch": 0.16, "learning_rate": 9.48177964972143e-06, "loss": 0.1216, "step": 3988 }, { "epoch": 0.16, "learning_rate": 9.481646362594301e-06, "loss": 0.1237, "step": 3989 }, { "epoch": 0.16, "learning_rate": 9.481513075467172e-06, "loss": 0.1019, "step": 3990 }, { "epoch": 0.16, "learning_rate": 9.481379788340043e-06, "loss": 0.1068, "step": 3991 }, { "epoch": 0.16, "learning_rate": 9.481246501212914e-06, "loss": 0.083, "step": 3992 }, { "epoch": 0.16, "learning_rate": 9.481113214085784e-06, "loss": 0.0922, "step": 3993 }, { "epoch": 0.16, "learning_rate": 9.480979926958655e-06, "loss": 0.1127, "step": 3994 }, { "epoch": 0.16, "learning_rate": 9.480846639831526e-06, "loss": 0.1234, "step": 3995 }, { "epoch": 0.16, "learning_rate": 9.480713352704397e-06, "loss": 0.1189, "step": 3996 }, { "epoch": 0.16, "learning_rate": 9.480580065577268e-06, "loss": 0.1275, "step": 3997 }, { "epoch": 0.16, "learning_rate": 9.480446778450139e-06, "loss": 0.0902, "step": 3998 }, { "epoch": 0.16, "learning_rate": 9.480313491323008e-06, "loss": 0.1044, "step": 3999 }, { "epoch": 0.16, "learning_rate": 9.480180204195879e-06, "loss": 0.1142, "step": 4000 }, { "epoch": 0.16, "learning_rate": 9.48004691706875e-06, "loss": 0.1245, "step": 4001 }, { "epoch": 0.16, "learning_rate": 9.479913629941621e-06, "loss": 0.1327, "step": 4002 }, { "epoch": 0.16, "learning_rate": 9.479780342814492e-06, "loss": 0.1327, "step": 4003 }, { "epoch": 0.16, "learning_rate": 9.479647055687361e-06, "loss": 0.0986, "step": 4004 }, { "epoch": 0.16, "learning_rate": 9.479513768560234e-06, "loss": 0.1076, "step": 4005 }, { "epoch": 0.16, "learning_rate": 9.479380481433103e-06, "loss": 0.0715, "step": 4006 }, { "epoch": 0.16, "learning_rate": 9.479247194305974e-06, "loss": 0.1333, "step": 4007 }, { "epoch": 0.16, "learning_rate": 9.479113907178845e-06, "loss": 0.1127, "step": 4008 }, { "epoch": 0.16, "learning_rate": 9.478980620051716e-06, "loss": 0.0956, "step": 4009 }, { "epoch": 0.16, "learning_rate": 9.478847332924587e-06, "loss": 0.1055, "step": 4010 }, { "epoch": 0.16, "learning_rate": 9.478714045797458e-06, "loss": 0.0943, "step": 4011 }, { "epoch": 0.16, "learning_rate": 9.478580758670328e-06, "loss": 0.1042, "step": 4012 }, { "epoch": 0.16, "learning_rate": 9.4784474715432e-06, "loss": 0.1295, "step": 4013 }, { "epoch": 0.16, "learning_rate": 9.47831418441607e-06, "loss": 0.1358, "step": 4014 }, { "epoch": 0.16, "learning_rate": 9.47818089728894e-06, "loss": 0.0958, "step": 4015 }, { "epoch": 0.16, "learning_rate": 9.478047610161812e-06, "loss": 0.1225, "step": 4016 }, { "epoch": 0.16, "learning_rate": 9.477914323034683e-06, "loss": 0.1111, "step": 4017 }, { "epoch": 0.16, "learning_rate": 9.477781035907553e-06, "loss": 0.0982, "step": 4018 }, { "epoch": 0.16, "learning_rate": 9.477647748780423e-06, "loss": 0.1692, "step": 4019 }, { "epoch": 0.16, "learning_rate": 9.477514461653294e-06, "loss": 0.0923, "step": 4020 }, { "epoch": 0.16, "learning_rate": 9.477381174526165e-06, "loss": 0.1057, "step": 4021 }, { "epoch": 0.16, "learning_rate": 9.477247887399036e-06, "loss": 0.0944, "step": 4022 }, { "epoch": 0.16, "learning_rate": 9.477114600271907e-06, "loss": 0.1262, "step": 4023 }, { "epoch": 0.16, "learning_rate": 9.476981313144778e-06, "loss": 0.1126, "step": 4024 }, { "epoch": 0.16, "learning_rate": 9.476848026017647e-06, "loss": 0.1467, "step": 4025 }, { "epoch": 0.16, "learning_rate": 9.47671473889052e-06, "loss": 0.0956, "step": 4026 }, { "epoch": 0.16, "learning_rate": 9.476581451763389e-06, "loss": 0.1078, "step": 4027 }, { "epoch": 0.16, "learning_rate": 9.47644816463626e-06, "loss": 0.1045, "step": 4028 }, { "epoch": 0.16, "learning_rate": 9.476314877509131e-06, "loss": 0.1055, "step": 4029 }, { "epoch": 0.16, "learning_rate": 9.476181590382002e-06, "loss": 0.1274, "step": 4030 }, { "epoch": 0.16, "learning_rate": 9.476048303254873e-06, "loss": 0.1113, "step": 4031 }, { "epoch": 0.16, "learning_rate": 9.475915016127744e-06, "loss": 0.1554, "step": 4032 }, { "epoch": 0.16, "learning_rate": 9.475781729000613e-06, "loss": 0.1212, "step": 4033 }, { "epoch": 0.16, "learning_rate": 9.475648441873484e-06, "loss": 0.136, "step": 4034 }, { "epoch": 0.16, "learning_rate": 9.475515154746355e-06, "loss": 0.1006, "step": 4035 }, { "epoch": 0.16, "learning_rate": 9.475381867619226e-06, "loss": 0.1074, "step": 4036 }, { "epoch": 0.16, "learning_rate": 9.475248580492097e-06, "loss": 0.1236, "step": 4037 }, { "epoch": 0.16, "learning_rate": 9.475115293364967e-06, "loss": 0.1274, "step": 4038 }, { "epoch": 0.16, "learning_rate": 9.474982006237838e-06, "loss": 0.1199, "step": 4039 }, { "epoch": 0.16, "learning_rate": 9.474848719110709e-06, "loss": 0.0884, "step": 4040 }, { "epoch": 0.16, "learning_rate": 9.47471543198358e-06, "loss": 0.1144, "step": 4041 }, { "epoch": 0.16, "learning_rate": 9.47458214485645e-06, "loss": 0.1489, "step": 4042 }, { "epoch": 0.16, "learning_rate": 9.474448857729322e-06, "loss": 0.1257, "step": 4043 }, { "epoch": 0.16, "learning_rate": 9.474315570602191e-06, "loss": 0.0759, "step": 4044 }, { "epoch": 0.16, "learning_rate": 9.474182283475064e-06, "loss": 0.1337, "step": 4045 }, { "epoch": 0.16, "learning_rate": 9.474048996347933e-06, "loss": 0.1055, "step": 4046 }, { "epoch": 0.16, "learning_rate": 9.473915709220804e-06, "loss": 0.1183, "step": 4047 }, { "epoch": 0.16, "learning_rate": 9.473782422093675e-06, "loss": 0.1399, "step": 4048 }, { "epoch": 0.16, "learning_rate": 9.473649134966546e-06, "loss": 0.1115, "step": 4049 }, { "epoch": 0.16, "learning_rate": 9.473515847839417e-06, "loss": 0.118, "step": 4050 }, { "epoch": 0.16, "learning_rate": 9.473382560712288e-06, "loss": 0.0985, "step": 4051 }, { "epoch": 0.16, "learning_rate": 9.473249273585157e-06, "loss": 0.0992, "step": 4052 }, { "epoch": 0.16, "learning_rate": 9.47311598645803e-06, "loss": 0.1543, "step": 4053 }, { "epoch": 0.16, "learning_rate": 9.472982699330899e-06, "loss": 0.1055, "step": 4054 }, { "epoch": 0.16, "learning_rate": 9.47284941220377e-06, "loss": 0.189, "step": 4055 }, { "epoch": 0.16, "learning_rate": 9.472716125076641e-06, "loss": 0.1125, "step": 4056 }, { "epoch": 0.16, "learning_rate": 9.47258283794951e-06, "loss": 0.1192, "step": 4057 }, { "epoch": 0.16, "learning_rate": 9.472449550822383e-06, "loss": 0.1321, "step": 4058 }, { "epoch": 0.16, "learning_rate": 9.472316263695252e-06, "loss": 0.1164, "step": 4059 }, { "epoch": 0.16, "learning_rate": 9.472182976568123e-06, "loss": 0.1164, "step": 4060 }, { "epoch": 0.16, "learning_rate": 9.472049689440994e-06, "loss": 0.1186, "step": 4061 }, { "epoch": 0.16, "learning_rate": 9.471916402313865e-06, "loss": 0.1018, "step": 4062 }, { "epoch": 0.16, "learning_rate": 9.471783115186736e-06, "loss": 0.0989, "step": 4063 }, { "epoch": 0.16, "learning_rate": 9.471649828059607e-06, "loss": 0.1116, "step": 4064 }, { "epoch": 0.16, "learning_rate": 9.471516540932477e-06, "loss": 0.0886, "step": 4065 }, { "epoch": 0.16, "learning_rate": 9.471383253805348e-06, "loss": 0.1137, "step": 4066 }, { "epoch": 0.16, "learning_rate": 9.471249966678219e-06, "loss": 0.0922, "step": 4067 }, { "epoch": 0.16, "learning_rate": 9.47111667955109e-06, "loss": 0.1129, "step": 4068 }, { "epoch": 0.16, "learning_rate": 9.47098339242396e-06, "loss": 0.0979, "step": 4069 }, { "epoch": 0.16, "learning_rate": 9.470850105296832e-06, "loss": 0.1157, "step": 4070 }, { "epoch": 0.16, "learning_rate": 9.470716818169701e-06, "loss": 0.1398, "step": 4071 }, { "epoch": 0.16, "learning_rate": 9.470583531042574e-06, "loss": 0.1074, "step": 4072 }, { "epoch": 0.16, "learning_rate": 9.470450243915443e-06, "loss": 0.1416, "step": 4073 }, { "epoch": 0.16, "learning_rate": 9.470316956788314e-06, "loss": 0.1363, "step": 4074 }, { "epoch": 0.16, "learning_rate": 9.470183669661185e-06, "loss": 0.0978, "step": 4075 }, { "epoch": 0.16, "learning_rate": 9.470050382534056e-06, "loss": 0.1417, "step": 4076 }, { "epoch": 0.16, "learning_rate": 9.469917095406927e-06, "loss": 0.1465, "step": 4077 }, { "epoch": 0.16, "learning_rate": 9.469783808279796e-06, "loss": 0.1156, "step": 4078 }, { "epoch": 0.16, "learning_rate": 9.469650521152667e-06, "loss": 0.1069, "step": 4079 }, { "epoch": 0.16, "learning_rate": 9.469517234025538e-06, "loss": 0.1533, "step": 4080 }, { "epoch": 0.16, "learning_rate": 9.46938394689841e-06, "loss": 0.1097, "step": 4081 }, { "epoch": 0.16, "learning_rate": 9.46925065977128e-06, "loss": 0.1207, "step": 4082 }, { "epoch": 0.16, "learning_rate": 9.469117372644151e-06, "loss": 0.0978, "step": 4083 }, { "epoch": 0.16, "learning_rate": 9.46898408551702e-06, "loss": 0.0977, "step": 4084 }, { "epoch": 0.16, "learning_rate": 9.468850798389893e-06, "loss": 0.101, "step": 4085 }, { "epoch": 0.16, "learning_rate": 9.468717511262763e-06, "loss": 0.0865, "step": 4086 }, { "epoch": 0.16, "learning_rate": 9.468584224135633e-06, "loss": 0.1076, "step": 4087 }, { "epoch": 0.16, "learning_rate": 9.468450937008504e-06, "loss": 0.0889, "step": 4088 }, { "epoch": 0.16, "learning_rate": 9.468317649881375e-06, "loss": 0.0926, "step": 4089 }, { "epoch": 0.16, "learning_rate": 9.468184362754246e-06, "loss": 0.1196, "step": 4090 }, { "epoch": 0.16, "learning_rate": 9.468051075627117e-06, "loss": 0.1109, "step": 4091 }, { "epoch": 0.16, "learning_rate": 9.467917788499987e-06, "loss": 0.1044, "step": 4092 }, { "epoch": 0.16, "learning_rate": 9.46778450137286e-06, "loss": 0.1459, "step": 4093 }, { "epoch": 0.16, "learning_rate": 9.467651214245729e-06, "loss": 0.1211, "step": 4094 }, { "epoch": 0.16, "learning_rate": 9.4675179271186e-06, "loss": 0.1391, "step": 4095 }, { "epoch": 0.16, "learning_rate": 9.46738463999147e-06, "loss": 0.1149, "step": 4096 }, { "epoch": 0.16, "learning_rate": 9.46725135286434e-06, "loss": 0.1477, "step": 4097 }, { "epoch": 0.16, "learning_rate": 9.467118065737213e-06, "loss": 0.1187, "step": 4098 }, { "epoch": 0.16, "learning_rate": 9.466984778610082e-06, "loss": 0.1103, "step": 4099 }, { "epoch": 0.16, "learning_rate": 9.466851491482953e-06, "loss": 0.0893, "step": 4100 }, { "epoch": 0.16, "learning_rate": 9.466718204355824e-06, "loss": 0.0978, "step": 4101 }, { "epoch": 0.16, "learning_rate": 9.466584917228695e-06, "loss": 0.0934, "step": 4102 }, { "epoch": 0.16, "learning_rate": 9.466451630101566e-06, "loss": 0.1154, "step": 4103 }, { "epoch": 0.16, "learning_rate": 9.466318342974437e-06, "loss": 0.0836, "step": 4104 }, { "epoch": 0.16, "learning_rate": 9.466185055847306e-06, "loss": 0.1781, "step": 4105 }, { "epoch": 0.16, "learning_rate": 9.466051768720177e-06, "loss": 0.1261, "step": 4106 }, { "epoch": 0.16, "learning_rate": 9.465918481593048e-06, "loss": 0.1187, "step": 4107 }, { "epoch": 0.16, "learning_rate": 9.46578519446592e-06, "loss": 0.1338, "step": 4108 }, { "epoch": 0.16, "learning_rate": 9.46565190733879e-06, "loss": 0.1082, "step": 4109 }, { "epoch": 0.16, "learning_rate": 9.465518620211661e-06, "loss": 0.1104, "step": 4110 }, { "epoch": 0.16, "learning_rate": 9.46538533308453e-06, "loss": 0.0856, "step": 4111 }, { "epoch": 0.16, "learning_rate": 9.465252045957403e-06, "loss": 0.1384, "step": 4112 }, { "epoch": 0.16, "learning_rate": 9.465118758830273e-06, "loss": 0.1213, "step": 4113 }, { "epoch": 0.16, "learning_rate": 9.464985471703144e-06, "loss": 0.0767, "step": 4114 }, { "epoch": 0.16, "learning_rate": 9.464852184576015e-06, "loss": 0.1305, "step": 4115 }, { "epoch": 0.16, "learning_rate": 9.464718897448884e-06, "loss": 0.0808, "step": 4116 }, { "epoch": 0.16, "learning_rate": 9.464585610321757e-06, "loss": 0.1143, "step": 4117 }, { "epoch": 0.16, "learning_rate": 9.464452323194626e-06, "loss": 0.1018, "step": 4118 }, { "epoch": 0.16, "learning_rate": 9.464319036067497e-06, "loss": 0.1301, "step": 4119 }, { "epoch": 0.16, "learning_rate": 9.464185748940368e-06, "loss": 0.1223, "step": 4120 }, { "epoch": 0.16, "learning_rate": 9.464052461813239e-06, "loss": 0.1341, "step": 4121 }, { "epoch": 0.16, "learning_rate": 9.46391917468611e-06, "loss": 0.0893, "step": 4122 }, { "epoch": 0.16, "learning_rate": 9.46378588755898e-06, "loss": 0.1241, "step": 4123 }, { "epoch": 0.16, "learning_rate": 9.46365260043185e-06, "loss": 0.091, "step": 4124 }, { "epoch": 0.16, "learning_rate": 9.463519313304723e-06, "loss": 0.1201, "step": 4125 }, { "epoch": 0.16, "learning_rate": 9.463386026177592e-06, "loss": 0.1326, "step": 4126 }, { "epoch": 0.16, "learning_rate": 9.463252739050463e-06, "loss": 0.1541, "step": 4127 }, { "epoch": 0.16, "learning_rate": 9.463119451923334e-06, "loss": 0.0904, "step": 4128 }, { "epoch": 0.16, "learning_rate": 9.462986164796205e-06, "loss": 0.1223, "step": 4129 }, { "epoch": 0.16, "learning_rate": 9.462852877669076e-06, "loss": 0.1129, "step": 4130 }, { "epoch": 0.16, "learning_rate": 9.462719590541947e-06, "loss": 0.1444, "step": 4131 }, { "epoch": 0.17, "learning_rate": 9.462586303414816e-06, "loss": 0.1655, "step": 4132 }, { "epoch": 0.17, "learning_rate": 9.462453016287687e-06, "loss": 0.1469, "step": 4133 }, { "epoch": 0.17, "learning_rate": 9.462319729160558e-06, "loss": 0.1101, "step": 4134 }, { "epoch": 0.17, "learning_rate": 9.46218644203343e-06, "loss": 0.1195, "step": 4135 }, { "epoch": 0.17, "learning_rate": 9.4620531549063e-06, "loss": 0.1195, "step": 4136 }, { "epoch": 0.17, "learning_rate": 9.46191986777917e-06, "loss": 0.1255, "step": 4137 }, { "epoch": 0.17, "learning_rate": 9.46178658065204e-06, "loss": 0.1122, "step": 4138 }, { "epoch": 0.17, "learning_rate": 9.461653293524912e-06, "loss": 0.166, "step": 4139 }, { "epoch": 0.17, "learning_rate": 9.461520006397783e-06, "loss": 0.1066, "step": 4140 }, { "epoch": 0.17, "learning_rate": 9.461386719270654e-06, "loss": 0.1379, "step": 4141 }, { "epoch": 0.17, "learning_rate": 9.461253432143525e-06, "loss": 0.0993, "step": 4142 }, { "epoch": 0.17, "learning_rate": 9.461120145016394e-06, "loss": 0.0955, "step": 4143 }, { "epoch": 0.17, "learning_rate": 9.460986857889267e-06, "loss": 0.1377, "step": 4144 }, { "epoch": 0.17, "learning_rate": 9.460853570762136e-06, "loss": 0.1834, "step": 4145 }, { "epoch": 0.17, "learning_rate": 9.460720283635007e-06, "loss": 0.0767, "step": 4146 }, { "epoch": 0.17, "learning_rate": 9.460586996507878e-06, "loss": 0.1103, "step": 4147 }, { "epoch": 0.17, "learning_rate": 9.460453709380749e-06, "loss": 0.0999, "step": 4148 }, { "epoch": 0.17, "learning_rate": 9.46032042225362e-06, "loss": 0.1098, "step": 4149 }, { "epoch": 0.17, "learning_rate": 9.460187135126491e-06, "loss": 0.0991, "step": 4150 }, { "epoch": 0.17, "learning_rate": 9.46005384799936e-06, "loss": 0.109, "step": 4151 }, { "epoch": 0.17, "learning_rate": 9.459920560872233e-06, "loss": 0.1153, "step": 4152 }, { "epoch": 0.17, "learning_rate": 9.459787273745102e-06, "loss": 0.0885, "step": 4153 }, { "epoch": 0.17, "learning_rate": 9.459653986617973e-06, "loss": 0.1117, "step": 4154 }, { "epoch": 0.17, "learning_rate": 9.459520699490844e-06, "loss": 0.1406, "step": 4155 }, { "epoch": 0.17, "learning_rate": 9.459387412363713e-06, "loss": 0.1195, "step": 4156 }, { "epoch": 0.17, "learning_rate": 9.459254125236586e-06, "loss": 0.0948, "step": 4157 }, { "epoch": 0.17, "learning_rate": 9.459120838109455e-06, "loss": 0.1328, "step": 4158 }, { "epoch": 0.17, "learning_rate": 9.458987550982326e-06, "loss": 0.1077, "step": 4159 }, { "epoch": 0.17, "learning_rate": 9.458854263855197e-06, "loss": 0.1226, "step": 4160 }, { "epoch": 0.17, "learning_rate": 9.458720976728068e-06, "loss": 0.1135, "step": 4161 }, { "epoch": 0.17, "learning_rate": 9.45858768960094e-06, "loss": 0.0735, "step": 4162 }, { "epoch": 0.17, "learning_rate": 9.45845440247381e-06, "loss": 0.1361, "step": 4163 }, { "epoch": 0.17, "learning_rate": 9.45832111534668e-06, "loss": 0.1133, "step": 4164 }, { "epoch": 0.17, "learning_rate": 9.458187828219552e-06, "loss": 0.1324, "step": 4165 }, { "epoch": 0.17, "learning_rate": 9.458054541092422e-06, "loss": 0.1041, "step": 4166 }, { "epoch": 0.17, "learning_rate": 9.457921253965293e-06, "loss": 0.1292, "step": 4167 }, { "epoch": 0.17, "learning_rate": 9.457787966838164e-06, "loss": 0.1253, "step": 4168 }, { "epoch": 0.17, "learning_rate": 9.457654679711035e-06, "loss": 0.1268, "step": 4169 }, { "epoch": 0.17, "learning_rate": 9.457521392583906e-06, "loss": 0.0935, "step": 4170 }, { "epoch": 0.17, "learning_rate": 9.457388105456777e-06, "loss": 0.1571, "step": 4171 }, { "epoch": 0.17, "learning_rate": 9.457254818329646e-06, "loss": 0.1512, "step": 4172 }, { "epoch": 0.17, "learning_rate": 9.457121531202517e-06, "loss": 0.1244, "step": 4173 }, { "epoch": 0.17, "learning_rate": 9.456988244075388e-06, "loss": 0.1435, "step": 4174 }, { "epoch": 0.17, "learning_rate": 9.456854956948259e-06, "loss": 0.1674, "step": 4175 }, { "epoch": 0.17, "learning_rate": 9.45672166982113e-06, "loss": 0.1232, "step": 4176 }, { "epoch": 0.17, "learning_rate": 9.456588382694e-06, "loss": 0.0968, "step": 4177 }, { "epoch": 0.17, "learning_rate": 9.45645509556687e-06, "loss": 0.1291, "step": 4178 }, { "epoch": 0.17, "learning_rate": 9.456321808439741e-06, "loss": 0.1151, "step": 4179 }, { "epoch": 0.17, "learning_rate": 9.456188521312612e-06, "loss": 0.0859, "step": 4180 }, { "epoch": 0.17, "learning_rate": 9.456055234185483e-06, "loss": 0.1277, "step": 4181 }, { "epoch": 0.17, "learning_rate": 9.455921947058354e-06, "loss": 0.1215, "step": 4182 }, { "epoch": 0.17, "learning_rate": 9.455788659931224e-06, "loss": 0.1547, "step": 4183 }, { "epoch": 0.17, "learning_rate": 9.455655372804096e-06, "loss": 0.0955, "step": 4184 }, { "epoch": 0.17, "learning_rate": 9.455522085676966e-06, "loss": 0.1086, "step": 4185 }, { "epoch": 0.17, "learning_rate": 9.455388798549837e-06, "loss": 0.1701, "step": 4186 }, { "epoch": 0.17, "learning_rate": 9.455255511422708e-06, "loss": 0.1039, "step": 4187 }, { "epoch": 0.17, "learning_rate": 9.455122224295579e-06, "loss": 0.072, "step": 4188 }, { "epoch": 0.17, "learning_rate": 9.45498893716845e-06, "loss": 0.1265, "step": 4189 }, { "epoch": 0.17, "learning_rate": 9.45485565004132e-06, "loss": 0.1107, "step": 4190 }, { "epoch": 0.17, "learning_rate": 9.45472236291419e-06, "loss": 0.1276, "step": 4191 }, { "epoch": 0.17, "learning_rate": 9.454589075787063e-06, "loss": 0.1378, "step": 4192 }, { "epoch": 0.17, "learning_rate": 9.454455788659932e-06, "loss": 0.0978, "step": 4193 }, { "epoch": 0.17, "learning_rate": 9.454322501532803e-06, "loss": 0.1097, "step": 4194 }, { "epoch": 0.17, "learning_rate": 9.454189214405674e-06, "loss": 0.1082, "step": 4195 }, { "epoch": 0.17, "learning_rate": 9.454055927278543e-06, "loss": 0.1525, "step": 4196 }, { "epoch": 0.17, "learning_rate": 9.453922640151416e-06, "loss": 0.1185, "step": 4197 }, { "epoch": 0.17, "learning_rate": 9.453789353024285e-06, "loss": 0.1032, "step": 4198 }, { "epoch": 0.17, "learning_rate": 9.453656065897156e-06, "loss": 0.0902, "step": 4199 }, { "epoch": 0.17, "learning_rate": 9.453522778770027e-06, "loss": 0.1146, "step": 4200 }, { "epoch": 0.17, "learning_rate": 9.453389491642898e-06, "loss": 0.1581, "step": 4201 }, { "epoch": 0.17, "learning_rate": 9.453256204515769e-06, "loss": 0.1641, "step": 4202 }, { "epoch": 0.17, "learning_rate": 9.45312291738864e-06, "loss": 0.1324, "step": 4203 }, { "epoch": 0.17, "learning_rate": 9.45298963026151e-06, "loss": 0.1616, "step": 4204 }, { "epoch": 0.17, "learning_rate": 9.45285634313438e-06, "loss": 0.1098, "step": 4205 }, { "epoch": 0.17, "learning_rate": 9.452723056007251e-06, "loss": 0.1397, "step": 4206 }, { "epoch": 0.17, "learning_rate": 9.452589768880122e-06, "loss": 0.1253, "step": 4207 }, { "epoch": 0.17, "learning_rate": 9.452456481752993e-06, "loss": 0.1096, "step": 4208 }, { "epoch": 0.17, "learning_rate": 9.452323194625864e-06, "loss": 0.1092, "step": 4209 }, { "epoch": 0.17, "learning_rate": 9.452189907498734e-06, "loss": 0.121, "step": 4210 }, { "epoch": 0.17, "learning_rate": 9.452056620371606e-06, "loss": 0.1187, "step": 4211 }, { "epoch": 0.17, "learning_rate": 9.451923333244476e-06, "loss": 0.1351, "step": 4212 }, { "epoch": 0.17, "learning_rate": 9.451790046117347e-06, "loss": 0.1287, "step": 4213 }, { "epoch": 0.17, "learning_rate": 9.451656758990218e-06, "loss": 0.1163, "step": 4214 }, { "epoch": 0.17, "learning_rate": 9.451523471863087e-06, "loss": 0.118, "step": 4215 }, { "epoch": 0.17, "learning_rate": 9.45139018473596e-06, "loss": 0.1196, "step": 4216 }, { "epoch": 0.17, "learning_rate": 9.451256897608829e-06, "loss": 0.1353, "step": 4217 }, { "epoch": 0.17, "learning_rate": 9.4511236104817e-06, "loss": 0.1337, "step": 4218 }, { "epoch": 0.17, "learning_rate": 9.450990323354571e-06, "loss": 0.0921, "step": 4219 }, { "epoch": 0.17, "learning_rate": 9.450857036227442e-06, "loss": 0.1488, "step": 4220 }, { "epoch": 0.17, "learning_rate": 9.450723749100313e-06, "loss": 0.0803, "step": 4221 }, { "epoch": 0.17, "learning_rate": 9.450590461973184e-06, "loss": 0.108, "step": 4222 }, { "epoch": 0.17, "learning_rate": 9.450457174846053e-06, "loss": 0.1361, "step": 4223 }, { "epoch": 0.17, "learning_rate": 9.450323887718926e-06, "loss": 0.1074, "step": 4224 }, { "epoch": 0.17, "learning_rate": 9.450190600591795e-06, "loss": 0.0801, "step": 4225 }, { "epoch": 0.17, "learning_rate": 9.450057313464666e-06, "loss": 0.1216, "step": 4226 }, { "epoch": 0.17, "learning_rate": 9.449924026337537e-06, "loss": 0.1509, "step": 4227 }, { "epoch": 0.17, "learning_rate": 9.449790739210408e-06, "loss": 0.1046, "step": 4228 }, { "epoch": 0.17, "learning_rate": 9.44965745208328e-06, "loss": 0.0977, "step": 4229 }, { "epoch": 0.17, "learning_rate": 9.44952416495615e-06, "loss": 0.0834, "step": 4230 }, { "epoch": 0.17, "learning_rate": 9.44939087782902e-06, "loss": 0.1232, "step": 4231 }, { "epoch": 0.17, "learning_rate": 9.44925759070189e-06, "loss": 0.1062, "step": 4232 }, { "epoch": 0.17, "learning_rate": 9.449124303574761e-06, "loss": 0.0821, "step": 4233 }, { "epoch": 0.17, "learning_rate": 9.448991016447632e-06, "loss": 0.0978, "step": 4234 }, { "epoch": 0.17, "learning_rate": 9.448857729320503e-06, "loss": 0.1243, "step": 4235 }, { "epoch": 0.17, "learning_rate": 9.448724442193373e-06, "loss": 0.178, "step": 4236 }, { "epoch": 0.17, "learning_rate": 9.448591155066244e-06, "loss": 0.127, "step": 4237 }, { "epoch": 0.17, "learning_rate": 9.448457867939115e-06, "loss": 0.1588, "step": 4238 }, { "epoch": 0.17, "learning_rate": 9.448324580811986e-06, "loss": 0.0953, "step": 4239 }, { "epoch": 0.17, "learning_rate": 9.448191293684857e-06, "loss": 0.1072, "step": 4240 }, { "epoch": 0.17, "learning_rate": 9.448058006557728e-06, "loss": 0.0931, "step": 4241 }, { "epoch": 0.17, "learning_rate": 9.447924719430597e-06, "loss": 0.1183, "step": 4242 }, { "epoch": 0.17, "learning_rate": 9.44779143230347e-06, "loss": 0.106, "step": 4243 }, { "epoch": 0.17, "learning_rate": 9.447658145176339e-06, "loss": 0.1472, "step": 4244 }, { "epoch": 0.17, "learning_rate": 9.44752485804921e-06, "loss": 0.1154, "step": 4245 }, { "epoch": 0.17, "learning_rate": 9.447391570922081e-06, "loss": 0.1137, "step": 4246 }, { "epoch": 0.17, "learning_rate": 9.447258283794952e-06, "loss": 0.1079, "step": 4247 }, { "epoch": 0.17, "learning_rate": 9.447124996667823e-06, "loss": 0.0986, "step": 4248 }, { "epoch": 0.17, "learning_rate": 9.446991709540694e-06, "loss": 0.1, "step": 4249 }, { "epoch": 0.17, "learning_rate": 9.446858422413563e-06, "loss": 0.1219, "step": 4250 }, { "epoch": 0.17, "learning_rate": 9.446725135286436e-06, "loss": 0.1167, "step": 4251 }, { "epoch": 0.17, "learning_rate": 9.446591848159305e-06, "loss": 0.118, "step": 4252 }, { "epoch": 0.17, "learning_rate": 9.446458561032176e-06, "loss": 0.0836, "step": 4253 }, { "epoch": 0.17, "learning_rate": 9.446325273905047e-06, "loss": 0.117, "step": 4254 }, { "epoch": 0.17, "learning_rate": 9.446191986777917e-06, "loss": 0.1029, "step": 4255 }, { "epoch": 0.17, "learning_rate": 9.44605869965079e-06, "loss": 0.1129, "step": 4256 }, { "epoch": 0.17, "learning_rate": 9.445925412523659e-06, "loss": 0.1329, "step": 4257 }, { "epoch": 0.17, "learning_rate": 9.44579212539653e-06, "loss": 0.1348, "step": 4258 }, { "epoch": 0.17, "learning_rate": 9.4456588382694e-06, "loss": 0.1273, "step": 4259 }, { "epoch": 0.17, "learning_rate": 9.445525551142272e-06, "loss": 0.0751, "step": 4260 }, { "epoch": 0.17, "learning_rate": 9.445392264015143e-06, "loss": 0.0969, "step": 4261 }, { "epoch": 0.17, "learning_rate": 9.445258976888014e-06, "loss": 0.1072, "step": 4262 }, { "epoch": 0.17, "learning_rate": 9.445125689760883e-06, "loss": 0.0864, "step": 4263 }, { "epoch": 0.17, "learning_rate": 9.444992402633756e-06, "loss": 0.1075, "step": 4264 }, { "epoch": 0.17, "learning_rate": 9.444859115506625e-06, "loss": 0.0921, "step": 4265 }, { "epoch": 0.17, "learning_rate": 9.444725828379496e-06, "loss": 0.1551, "step": 4266 }, { "epoch": 0.17, "learning_rate": 9.444592541252367e-06, "loss": 0.1366, "step": 4267 }, { "epoch": 0.17, "learning_rate": 9.444459254125238e-06, "loss": 0.1575, "step": 4268 }, { "epoch": 0.17, "learning_rate": 9.444325966998109e-06, "loss": 0.0841, "step": 4269 }, { "epoch": 0.17, "learning_rate": 9.44419267987098e-06, "loss": 0.1502, "step": 4270 }, { "epoch": 0.17, "learning_rate": 9.444059392743849e-06, "loss": 0.1057, "step": 4271 }, { "epoch": 0.17, "learning_rate": 9.44392610561672e-06, "loss": 0.1591, "step": 4272 }, { "epoch": 0.17, "learning_rate": 9.443792818489591e-06, "loss": 0.1015, "step": 4273 }, { "epoch": 0.17, "learning_rate": 9.443659531362462e-06, "loss": 0.1385, "step": 4274 }, { "epoch": 0.17, "learning_rate": 9.443526244235333e-06, "loss": 0.0894, "step": 4275 }, { "epoch": 0.17, "learning_rate": 9.443392957108202e-06, "loss": 0.0739, "step": 4276 }, { "epoch": 0.17, "learning_rate": 9.443259669981073e-06, "loss": 0.1337, "step": 4277 }, { "epoch": 0.17, "learning_rate": 9.443126382853944e-06, "loss": 0.113, "step": 4278 }, { "epoch": 0.17, "learning_rate": 9.442993095726815e-06, "loss": 0.113, "step": 4279 }, { "epoch": 0.17, "learning_rate": 9.442859808599686e-06, "loss": 0.1316, "step": 4280 }, { "epoch": 0.17, "learning_rate": 9.442726521472557e-06, "loss": 0.1108, "step": 4281 }, { "epoch": 0.17, "learning_rate": 9.442593234345427e-06, "loss": 0.1263, "step": 4282 }, { "epoch": 0.17, "learning_rate": 9.4424599472183e-06, "loss": 0.0876, "step": 4283 }, { "epoch": 0.17, "learning_rate": 9.442326660091169e-06, "loss": 0.1523, "step": 4284 }, { "epoch": 0.17, "learning_rate": 9.44219337296404e-06, "loss": 0.1486, "step": 4285 }, { "epoch": 0.17, "learning_rate": 9.44206008583691e-06, "loss": 0.1575, "step": 4286 }, { "epoch": 0.17, "learning_rate": 9.441926798709782e-06, "loss": 0.1328, "step": 4287 }, { "epoch": 0.17, "learning_rate": 9.441793511582653e-06, "loss": 0.1623, "step": 4288 }, { "epoch": 0.17, "learning_rate": 9.441660224455524e-06, "loss": 0.0796, "step": 4289 }, { "epoch": 0.17, "learning_rate": 9.441526937328393e-06, "loss": 0.1175, "step": 4290 }, { "epoch": 0.17, "learning_rate": 9.441393650201266e-06, "loss": 0.1222, "step": 4291 }, { "epoch": 0.17, "learning_rate": 9.441260363074135e-06, "loss": 0.1032, "step": 4292 }, { "epoch": 0.17, "learning_rate": 9.441127075947006e-06, "loss": 0.1418, "step": 4293 }, { "epoch": 0.17, "learning_rate": 9.440993788819877e-06, "loss": 0.0863, "step": 4294 }, { "epoch": 0.17, "learning_rate": 9.440860501692746e-06, "loss": 0.113, "step": 4295 }, { "epoch": 0.17, "learning_rate": 9.440727214565619e-06, "loss": 0.0878, "step": 4296 }, { "epoch": 0.17, "learning_rate": 9.440593927438488e-06, "loss": 0.1394, "step": 4297 }, { "epoch": 0.17, "learning_rate": 9.44046064031136e-06, "loss": 0.0942, "step": 4298 }, { "epoch": 0.17, "learning_rate": 9.44032735318423e-06, "loss": 0.1098, "step": 4299 }, { "epoch": 0.17, "learning_rate": 9.440194066057101e-06, "loss": 0.0977, "step": 4300 }, { "epoch": 0.17, "learning_rate": 9.440060778929972e-06, "loss": 0.1054, "step": 4301 }, { "epoch": 0.17, "learning_rate": 9.439927491802843e-06, "loss": 0.1179, "step": 4302 }, { "epoch": 0.17, "learning_rate": 9.439794204675712e-06, "loss": 0.1027, "step": 4303 }, { "epoch": 0.17, "learning_rate": 9.439660917548583e-06, "loss": 0.1146, "step": 4304 }, { "epoch": 0.17, "learning_rate": 9.439527630421454e-06, "loss": 0.1403, "step": 4305 }, { "epoch": 0.17, "learning_rate": 9.439394343294325e-06, "loss": 0.1538, "step": 4306 }, { "epoch": 0.17, "learning_rate": 9.439261056167196e-06, "loss": 0.1972, "step": 4307 }, { "epoch": 0.17, "learning_rate": 9.439127769040067e-06, "loss": 0.0647, "step": 4308 }, { "epoch": 0.17, "learning_rate": 9.438994481912937e-06, "loss": 0.1266, "step": 4309 }, { "epoch": 0.17, "learning_rate": 9.43886119478581e-06, "loss": 0.1224, "step": 4310 }, { "epoch": 0.17, "learning_rate": 9.438727907658679e-06, "loss": 0.0982, "step": 4311 }, { "epoch": 0.17, "learning_rate": 9.43859462053155e-06, "loss": 0.106, "step": 4312 }, { "epoch": 0.17, "learning_rate": 9.43846133340442e-06, "loss": 0.1531, "step": 4313 }, { "epoch": 0.17, "learning_rate": 9.43832804627729e-06, "loss": 0.0908, "step": 4314 }, { "epoch": 0.17, "learning_rate": 9.438194759150163e-06, "loss": 0.1143, "step": 4315 }, { "epoch": 0.17, "learning_rate": 9.438061472023032e-06, "loss": 0.0952, "step": 4316 }, { "epoch": 0.17, "learning_rate": 9.437928184895903e-06, "loss": 0.1094, "step": 4317 }, { "epoch": 0.17, "learning_rate": 9.437794897768774e-06, "loss": 0.0992, "step": 4318 }, { "epoch": 0.17, "learning_rate": 9.437661610641645e-06, "loss": 0.0938, "step": 4319 }, { "epoch": 0.17, "learning_rate": 9.437528323514516e-06, "loss": 0.1349, "step": 4320 }, { "epoch": 0.17, "learning_rate": 9.437395036387387e-06, "loss": 0.1119, "step": 4321 }, { "epoch": 0.17, "learning_rate": 9.437261749260256e-06, "loss": 0.1015, "step": 4322 }, { "epoch": 0.17, "learning_rate": 9.437128462133129e-06, "loss": 0.1471, "step": 4323 }, { "epoch": 0.17, "learning_rate": 9.436995175005998e-06, "loss": 0.1361, "step": 4324 }, { "epoch": 0.17, "learning_rate": 9.43686188787887e-06, "loss": 0.1159, "step": 4325 }, { "epoch": 0.17, "learning_rate": 9.43672860075174e-06, "loss": 0.1151, "step": 4326 }, { "epoch": 0.17, "learning_rate": 9.436595313624611e-06, "loss": 0.1141, "step": 4327 }, { "epoch": 0.17, "learning_rate": 9.436462026497482e-06, "loss": 0.1277, "step": 4328 }, { "epoch": 0.17, "learning_rate": 9.436328739370353e-06, "loss": 0.1175, "step": 4329 }, { "epoch": 0.17, "learning_rate": 9.436195452243223e-06, "loss": 0.1298, "step": 4330 }, { "epoch": 0.17, "learning_rate": 9.436062165116095e-06, "loss": 0.0632, "step": 4331 }, { "epoch": 0.17, "learning_rate": 9.435928877988965e-06, "loss": 0.0846, "step": 4332 }, { "epoch": 0.17, "learning_rate": 9.435795590861836e-06, "loss": 0.0867, "step": 4333 }, { "epoch": 0.17, "learning_rate": 9.435662303734707e-06, "loss": 0.0732, "step": 4334 }, { "epoch": 0.17, "learning_rate": 9.435529016607576e-06, "loss": 0.1349, "step": 4335 }, { "epoch": 0.17, "learning_rate": 9.435395729480448e-06, "loss": 0.1215, "step": 4336 }, { "epoch": 0.17, "learning_rate": 9.435262442353318e-06, "loss": 0.1189, "step": 4337 }, { "epoch": 0.17, "learning_rate": 9.435129155226189e-06, "loss": 0.119, "step": 4338 }, { "epoch": 0.17, "learning_rate": 9.43499586809906e-06, "loss": 0.143, "step": 4339 }, { "epoch": 0.17, "learning_rate": 9.43486258097193e-06, "loss": 0.1077, "step": 4340 }, { "epoch": 0.17, "learning_rate": 9.434729293844802e-06, "loss": 0.134, "step": 4341 }, { "epoch": 0.17, "learning_rate": 9.434596006717673e-06, "loss": 0.1158, "step": 4342 }, { "epoch": 0.17, "learning_rate": 9.434462719590542e-06, "loss": 0.1206, "step": 4343 }, { "epoch": 0.17, "learning_rate": 9.434329432463413e-06, "loss": 0.128, "step": 4344 }, { "epoch": 0.17, "learning_rate": 9.434196145336284e-06, "loss": 0.1341, "step": 4345 }, { "epoch": 0.17, "learning_rate": 9.434062858209155e-06, "loss": 0.1754, "step": 4346 }, { "epoch": 0.17, "learning_rate": 9.433929571082026e-06, "loss": 0.1366, "step": 4347 }, { "epoch": 0.17, "learning_rate": 9.433796283954897e-06, "loss": 0.0941, "step": 4348 }, { "epoch": 0.17, "learning_rate": 9.433662996827766e-06, "loss": 0.1472, "step": 4349 }, { "epoch": 0.17, "learning_rate": 9.433529709700639e-06, "loss": 0.1373, "step": 4350 }, { "epoch": 0.17, "learning_rate": 9.433396422573508e-06, "loss": 0.1071, "step": 4351 }, { "epoch": 0.17, "learning_rate": 9.43326313544638e-06, "loss": 0.1504, "step": 4352 }, { "epoch": 0.17, "learning_rate": 9.43312984831925e-06, "loss": 0.0992, "step": 4353 }, { "epoch": 0.17, "learning_rate": 9.43299656119212e-06, "loss": 0.1399, "step": 4354 }, { "epoch": 0.17, "learning_rate": 9.432863274064992e-06, "loss": 0.1207, "step": 4355 }, { "epoch": 0.17, "learning_rate": 9.432729986937862e-06, "loss": 0.1423, "step": 4356 }, { "epoch": 0.17, "learning_rate": 9.432596699810733e-06, "loss": 0.1441, "step": 4357 }, { "epoch": 0.17, "learning_rate": 9.432463412683604e-06, "loss": 0.098, "step": 4358 }, { "epoch": 0.17, "learning_rate": 9.432330125556475e-06, "loss": 0.1005, "step": 4359 }, { "epoch": 0.17, "learning_rate": 9.432196838429346e-06, "loss": 0.0937, "step": 4360 }, { "epoch": 0.17, "learning_rate": 9.432063551302217e-06, "loss": 0.0986, "step": 4361 }, { "epoch": 0.17, "learning_rate": 9.431930264175086e-06, "loss": 0.1419, "step": 4362 }, { "epoch": 0.17, "learning_rate": 9.431796977047959e-06, "loss": 0.0959, "step": 4363 }, { "epoch": 0.17, "learning_rate": 9.431663689920828e-06, "loss": 0.1147, "step": 4364 }, { "epoch": 0.17, "learning_rate": 9.431530402793699e-06, "loss": 0.0953, "step": 4365 }, { "epoch": 0.17, "learning_rate": 9.43139711566657e-06, "loss": 0.1144, "step": 4366 }, { "epoch": 0.17, "learning_rate": 9.431263828539441e-06, "loss": 0.0813, "step": 4367 }, { "epoch": 0.17, "learning_rate": 9.431130541412312e-06, "loss": 0.1358, "step": 4368 }, { "epoch": 0.17, "learning_rate": 9.430997254285183e-06, "loss": 0.1442, "step": 4369 }, { "epoch": 0.17, "learning_rate": 9.430863967158052e-06, "loss": 0.0974, "step": 4370 }, { "epoch": 0.17, "learning_rate": 9.430730680030923e-06, "loss": 0.1337, "step": 4371 }, { "epoch": 0.17, "learning_rate": 9.430597392903794e-06, "loss": 0.1027, "step": 4372 }, { "epoch": 0.17, "learning_rate": 9.430464105776665e-06, "loss": 0.0903, "step": 4373 }, { "epoch": 0.17, "learning_rate": 9.430330818649536e-06, "loss": 0.1128, "step": 4374 }, { "epoch": 0.17, "learning_rate": 9.430197531522405e-06, "loss": 0.1056, "step": 4375 }, { "epoch": 0.17, "learning_rate": 9.430064244395276e-06, "loss": 0.0838, "step": 4376 }, { "epoch": 0.17, "learning_rate": 9.429930957268147e-06, "loss": 0.1147, "step": 4377 }, { "epoch": 0.17, "learning_rate": 9.429797670141018e-06, "loss": 0.101, "step": 4378 }, { "epoch": 0.17, "learning_rate": 9.42966438301389e-06, "loss": 0.0979, "step": 4379 }, { "epoch": 0.17, "learning_rate": 9.42953109588676e-06, "loss": 0.1144, "step": 4380 }, { "epoch": 0.17, "learning_rate": 9.42939780875963e-06, "loss": 0.0977, "step": 4381 }, { "epoch": 0.17, "learning_rate": 9.429264521632502e-06, "loss": 0.1192, "step": 4382 }, { "epoch": 0.18, "learning_rate": 9.429131234505372e-06, "loss": 0.1589, "step": 4383 }, { "epoch": 0.18, "learning_rate": 9.428997947378243e-06, "loss": 0.1145, "step": 4384 }, { "epoch": 0.18, "learning_rate": 9.428864660251114e-06, "loss": 0.0983, "step": 4385 }, { "epoch": 0.18, "learning_rate": 9.428731373123985e-06, "loss": 0.136, "step": 4386 }, { "epoch": 0.18, "learning_rate": 9.428598085996856e-06, "loss": 0.1347, "step": 4387 }, { "epoch": 0.18, "learning_rate": 9.428464798869727e-06, "loss": 0.1031, "step": 4388 }, { "epoch": 0.18, "learning_rate": 9.428331511742596e-06, "loss": 0.1252, "step": 4389 }, { "epoch": 0.18, "learning_rate": 9.428198224615467e-06, "loss": 0.1653, "step": 4390 }, { "epoch": 0.18, "learning_rate": 9.428064937488338e-06, "loss": 0.0891, "step": 4391 }, { "epoch": 0.18, "learning_rate": 9.427931650361209e-06, "loss": 0.108, "step": 4392 }, { "epoch": 0.18, "learning_rate": 9.42779836323408e-06, "loss": 0.1543, "step": 4393 }, { "epoch": 0.18, "learning_rate": 9.42766507610695e-06, "loss": 0.1608, "step": 4394 }, { "epoch": 0.18, "learning_rate": 9.427531788979822e-06, "loss": 0.1021, "step": 4395 }, { "epoch": 0.18, "learning_rate": 9.427398501852691e-06, "loss": 0.1389, "step": 4396 }, { "epoch": 0.18, "learning_rate": 9.427265214725562e-06, "loss": 0.1132, "step": 4397 }, { "epoch": 0.18, "learning_rate": 9.427131927598433e-06, "loss": 0.1036, "step": 4398 }, { "epoch": 0.18, "learning_rate": 9.426998640471304e-06, "loss": 0.1409, "step": 4399 }, { "epoch": 0.18, "learning_rate": 9.426865353344175e-06, "loss": 0.1242, "step": 4400 }, { "epoch": 0.18, "learning_rate": 9.426732066217046e-06, "loss": 0.1142, "step": 4401 }, { "epoch": 0.18, "learning_rate": 9.426598779089916e-06, "loss": 0.1032, "step": 4402 }, { "epoch": 0.18, "learning_rate": 9.426465491962788e-06, "loss": 0.1104, "step": 4403 }, { "epoch": 0.18, "learning_rate": 9.426332204835658e-06, "loss": 0.0933, "step": 4404 }, { "epoch": 0.18, "learning_rate": 9.426198917708528e-06, "loss": 0.1148, "step": 4405 }, { "epoch": 0.18, "learning_rate": 9.4260656305814e-06, "loss": 0.1198, "step": 4406 }, { "epoch": 0.18, "learning_rate": 9.42593234345427e-06, "loss": 0.0892, "step": 4407 }, { "epoch": 0.18, "learning_rate": 9.425799056327141e-06, "loss": 0.1091, "step": 4408 }, { "epoch": 0.18, "learning_rate": 9.425665769200012e-06, "loss": 0.1274, "step": 4409 }, { "epoch": 0.18, "learning_rate": 9.425532482072882e-06, "loss": 0.0851, "step": 4410 }, { "epoch": 0.18, "learning_rate": 9.425399194945753e-06, "loss": 0.115, "step": 4411 }, { "epoch": 0.18, "learning_rate": 9.425265907818624e-06, "loss": 0.1407, "step": 4412 }, { "epoch": 0.18, "learning_rate": 9.425132620691495e-06, "loss": 0.1073, "step": 4413 }, { "epoch": 0.18, "learning_rate": 9.424999333564366e-06, "loss": 0.1161, "step": 4414 }, { "epoch": 0.18, "learning_rate": 9.424866046437235e-06, "loss": 0.1078, "step": 4415 }, { "epoch": 0.18, "learning_rate": 9.424732759310106e-06, "loss": 0.0933, "step": 4416 }, { "epoch": 0.18, "learning_rate": 9.424599472182977e-06, "loss": 0.0875, "step": 4417 }, { "epoch": 0.18, "learning_rate": 9.424466185055848e-06, "loss": 0.122, "step": 4418 }, { "epoch": 0.18, "learning_rate": 9.424332897928719e-06, "loss": 0.1602, "step": 4419 }, { "epoch": 0.18, "learning_rate": 9.42419961080159e-06, "loss": 0.1218, "step": 4420 }, { "epoch": 0.18, "learning_rate": 9.42406632367446e-06, "loss": 0.0731, "step": 4421 }, { "epoch": 0.18, "learning_rate": 9.423933036547332e-06, "loss": 0.0957, "step": 4422 }, { "epoch": 0.18, "learning_rate": 9.423799749420201e-06, "loss": 0.0999, "step": 4423 }, { "epoch": 0.18, "learning_rate": 9.423666462293072e-06, "loss": 0.1456, "step": 4424 }, { "epoch": 0.18, "learning_rate": 9.423533175165943e-06, "loss": 0.1463, "step": 4425 }, { "epoch": 0.18, "learning_rate": 9.423399888038814e-06, "loss": 0.2008, "step": 4426 }, { "epoch": 0.18, "learning_rate": 9.423266600911685e-06, "loss": 0.1396, "step": 4427 }, { "epoch": 0.18, "learning_rate": 9.423133313784556e-06, "loss": 0.1099, "step": 4428 }, { "epoch": 0.18, "learning_rate": 9.423000026657426e-06, "loss": 0.1371, "step": 4429 }, { "epoch": 0.18, "learning_rate": 9.422866739530297e-06, "loss": 0.1081, "step": 4430 }, { "epoch": 0.18, "learning_rate": 9.422733452403168e-06, "loss": 0.1216, "step": 4431 }, { "epoch": 0.18, "learning_rate": 9.422600165276039e-06, "loss": 0.1383, "step": 4432 }, { "epoch": 0.18, "learning_rate": 9.42246687814891e-06, "loss": 0.1221, "step": 4433 }, { "epoch": 0.18, "learning_rate": 9.422333591021779e-06, "loss": 0.0956, "step": 4434 }, { "epoch": 0.18, "learning_rate": 9.422200303894652e-06, "loss": 0.1003, "step": 4435 }, { "epoch": 0.18, "learning_rate": 9.422067016767521e-06, "loss": 0.1451, "step": 4436 }, { "epoch": 0.18, "learning_rate": 9.421933729640392e-06, "loss": 0.1778, "step": 4437 }, { "epoch": 0.18, "learning_rate": 9.421800442513263e-06, "loss": 0.1145, "step": 4438 }, { "epoch": 0.18, "learning_rate": 9.421667155386134e-06, "loss": 0.1056, "step": 4439 }, { "epoch": 0.18, "learning_rate": 9.421533868259005e-06, "loss": 0.1238, "step": 4440 }, { "epoch": 0.18, "learning_rate": 9.421400581131876e-06, "loss": 0.0981, "step": 4441 }, { "epoch": 0.18, "learning_rate": 9.421267294004745e-06, "loss": 0.0971, "step": 4442 }, { "epoch": 0.18, "learning_rate": 9.421134006877616e-06, "loss": 0.1038, "step": 4443 }, { "epoch": 0.18, "learning_rate": 9.421000719750487e-06, "loss": 0.1408, "step": 4444 }, { "epoch": 0.18, "learning_rate": 9.420867432623358e-06, "loss": 0.1327, "step": 4445 }, { "epoch": 0.18, "learning_rate": 9.420734145496229e-06, "loss": 0.1482, "step": 4446 }, { "epoch": 0.18, "learning_rate": 9.4206008583691e-06, "loss": 0.1171, "step": 4447 }, { "epoch": 0.18, "learning_rate": 9.42046757124197e-06, "loss": 0.0753, "step": 4448 }, { "epoch": 0.18, "learning_rate": 9.42033428411484e-06, "loss": 0.0879, "step": 4449 }, { "epoch": 0.18, "learning_rate": 9.420200996987711e-06, "loss": 0.1145, "step": 4450 }, { "epoch": 0.18, "learning_rate": 9.420067709860582e-06, "loss": 0.1698, "step": 4451 }, { "epoch": 0.18, "learning_rate": 9.419934422733453e-06, "loss": 0.1001, "step": 4452 }, { "epoch": 0.18, "learning_rate": 9.419801135606323e-06, "loss": 0.1106, "step": 4453 }, { "epoch": 0.18, "learning_rate": 9.419667848479195e-06, "loss": 0.072, "step": 4454 }, { "epoch": 0.18, "learning_rate": 9.419534561352065e-06, "loss": 0.1529, "step": 4455 }, { "epoch": 0.18, "learning_rate": 9.419401274224936e-06, "loss": 0.0975, "step": 4456 }, { "epoch": 0.18, "learning_rate": 9.419267987097807e-06, "loss": 0.0978, "step": 4457 }, { "epoch": 0.18, "learning_rate": 9.419134699970678e-06, "loss": 0.1262, "step": 4458 }, { "epoch": 0.18, "learning_rate": 9.419001412843549e-06, "loss": 0.087, "step": 4459 }, { "epoch": 0.18, "learning_rate": 9.41886812571642e-06, "loss": 0.1225, "step": 4460 }, { "epoch": 0.18, "learning_rate": 9.418734838589289e-06, "loss": 0.1306, "step": 4461 }, { "epoch": 0.18, "learning_rate": 9.418601551462162e-06, "loss": 0.1134, "step": 4462 }, { "epoch": 0.18, "learning_rate": 9.418468264335031e-06, "loss": 0.113, "step": 4463 }, { "epoch": 0.18, "learning_rate": 9.418334977207902e-06, "loss": 0.1054, "step": 4464 }, { "epoch": 0.18, "learning_rate": 9.418201690080773e-06, "loss": 0.1492, "step": 4465 }, { "epoch": 0.18, "learning_rate": 9.418068402953644e-06, "loss": 0.0866, "step": 4466 }, { "epoch": 0.18, "learning_rate": 9.417935115826515e-06, "loss": 0.1427, "step": 4467 }, { "epoch": 0.18, "learning_rate": 9.417801828699386e-06, "loss": 0.2002, "step": 4468 }, { "epoch": 0.18, "learning_rate": 9.417668541572255e-06, "loss": 0.1099, "step": 4469 }, { "epoch": 0.18, "learning_rate": 9.417535254445126e-06, "loss": 0.0831, "step": 4470 }, { "epoch": 0.18, "learning_rate": 9.417401967317997e-06, "loss": 0.0931, "step": 4471 }, { "epoch": 0.18, "learning_rate": 9.417268680190868e-06, "loss": 0.1112, "step": 4472 }, { "epoch": 0.18, "learning_rate": 9.41713539306374e-06, "loss": 0.0957, "step": 4473 }, { "epoch": 0.18, "learning_rate": 9.417002105936608e-06, "loss": 0.1541, "step": 4474 }, { "epoch": 0.18, "learning_rate": 9.41686881880948e-06, "loss": 0.1044, "step": 4475 }, { "epoch": 0.18, "learning_rate": 9.41673553168235e-06, "loss": 0.0993, "step": 4476 }, { "epoch": 0.18, "learning_rate": 9.416602244555221e-06, "loss": 0.0842, "step": 4477 }, { "epoch": 0.18, "learning_rate": 9.416468957428092e-06, "loss": 0.1129, "step": 4478 }, { "epoch": 0.18, "learning_rate": 9.416335670300963e-06, "loss": 0.1164, "step": 4479 }, { "epoch": 0.18, "learning_rate": 9.416202383173833e-06, "loss": 0.1517, "step": 4480 }, { "epoch": 0.18, "learning_rate": 9.416069096046705e-06, "loss": 0.0929, "step": 4481 }, { "epoch": 0.18, "learning_rate": 9.415935808919575e-06, "loss": 0.0692, "step": 4482 }, { "epoch": 0.18, "learning_rate": 9.415802521792446e-06, "loss": 0.0861, "step": 4483 }, { "epoch": 0.18, "learning_rate": 9.415669234665317e-06, "loss": 0.0906, "step": 4484 }, { "epoch": 0.18, "learning_rate": 9.415535947538188e-06, "loss": 0.0764, "step": 4485 }, { "epoch": 0.18, "learning_rate": 9.415402660411059e-06, "loss": 0.1149, "step": 4486 }, { "epoch": 0.18, "learning_rate": 9.41526937328393e-06, "loss": 0.1647, "step": 4487 }, { "epoch": 0.18, "learning_rate": 9.415136086156799e-06, "loss": 0.0837, "step": 4488 }, { "epoch": 0.18, "learning_rate": 9.41500279902967e-06, "loss": 0.1102, "step": 4489 }, { "epoch": 0.18, "learning_rate": 9.414869511902541e-06, "loss": 0.098, "step": 4490 }, { "epoch": 0.18, "learning_rate": 9.414736224775412e-06, "loss": 0.1719, "step": 4491 }, { "epoch": 0.18, "learning_rate": 9.414602937648283e-06, "loss": 0.0999, "step": 4492 }, { "epoch": 0.18, "learning_rate": 9.414469650521152e-06, "loss": 0.123, "step": 4493 }, { "epoch": 0.18, "learning_rate": 9.414336363394025e-06, "loss": 0.0923, "step": 4494 }, { "epoch": 0.18, "learning_rate": 9.414203076266894e-06, "loss": 0.1432, "step": 4495 }, { "epoch": 0.18, "learning_rate": 9.414069789139765e-06, "loss": 0.1045, "step": 4496 }, { "epoch": 0.18, "learning_rate": 9.413936502012636e-06, "loss": 0.074, "step": 4497 }, { "epoch": 0.18, "learning_rate": 9.413803214885507e-06, "loss": 0.0982, "step": 4498 }, { "epoch": 0.18, "learning_rate": 9.413669927758378e-06, "loss": 0.105, "step": 4499 }, { "epoch": 0.18, "learning_rate": 9.41353664063125e-06, "loss": 0.1075, "step": 4500 }, { "epoch": 0.18, "learning_rate": 9.413403353504119e-06, "loss": 0.1337, "step": 4501 }, { "epoch": 0.18, "learning_rate": 9.413270066376991e-06, "loss": 0.137, "step": 4502 }, { "epoch": 0.18, "learning_rate": 9.41313677924986e-06, "loss": 0.1239, "step": 4503 }, { "epoch": 0.18, "learning_rate": 9.413003492122732e-06, "loss": 0.1619, "step": 4504 }, { "epoch": 0.18, "learning_rate": 9.412870204995603e-06, "loss": 0.1076, "step": 4505 }, { "epoch": 0.18, "learning_rate": 9.412736917868474e-06, "loss": 0.0933, "step": 4506 }, { "epoch": 0.18, "learning_rate": 9.412603630741345e-06, "loss": 0.1102, "step": 4507 }, { "epoch": 0.18, "learning_rate": 9.412470343614214e-06, "loss": 0.1067, "step": 4508 }, { "epoch": 0.18, "learning_rate": 9.412337056487085e-06, "loss": 0.1169, "step": 4509 }, { "epoch": 0.18, "learning_rate": 9.412203769359956e-06, "loss": 0.158, "step": 4510 }, { "epoch": 0.18, "learning_rate": 9.412070482232827e-06, "loss": 0.1012, "step": 4511 }, { "epoch": 0.18, "learning_rate": 9.411937195105698e-06, "loss": 0.0811, "step": 4512 }, { "epoch": 0.18, "learning_rate": 9.411803907978569e-06, "loss": 0.1241, "step": 4513 }, { "epoch": 0.18, "learning_rate": 9.411670620851438e-06, "loss": 0.1184, "step": 4514 }, { "epoch": 0.18, "learning_rate": 9.411537333724309e-06, "loss": 0.1343, "step": 4515 }, { "epoch": 0.18, "learning_rate": 9.41140404659718e-06, "loss": 0.1024, "step": 4516 }, { "epoch": 0.18, "learning_rate": 9.411270759470051e-06, "loss": 0.0849, "step": 4517 }, { "epoch": 0.18, "learning_rate": 9.411137472342922e-06, "loss": 0.0701, "step": 4518 }, { "epoch": 0.18, "learning_rate": 9.411004185215793e-06, "loss": 0.1187, "step": 4519 }, { "epoch": 0.18, "learning_rate": 9.410870898088662e-06, "loss": 0.1444, "step": 4520 }, { "epoch": 0.18, "learning_rate": 9.410737610961535e-06, "loss": 0.1239, "step": 4521 }, { "epoch": 0.18, "learning_rate": 9.410604323834404e-06, "loss": 0.1497, "step": 4522 }, { "epoch": 0.18, "learning_rate": 9.410471036707275e-06, "loss": 0.1279, "step": 4523 }, { "epoch": 0.18, "learning_rate": 9.410337749580146e-06, "loss": 0.0719, "step": 4524 }, { "epoch": 0.18, "learning_rate": 9.410204462453017e-06, "loss": 0.1458, "step": 4525 }, { "epoch": 0.18, "learning_rate": 9.410071175325888e-06, "loss": 0.1025, "step": 4526 }, { "epoch": 0.18, "learning_rate": 9.40993788819876e-06, "loss": 0.1532, "step": 4527 }, { "epoch": 0.18, "learning_rate": 9.409804601071629e-06, "loss": 0.0844, "step": 4528 }, { "epoch": 0.18, "learning_rate": 9.4096713139445e-06, "loss": 0.1208, "step": 4529 }, { "epoch": 0.18, "learning_rate": 9.40953802681737e-06, "loss": 0.1466, "step": 4530 }, { "epoch": 0.18, "learning_rate": 9.409404739690242e-06, "loss": 0.086, "step": 4531 }, { "epoch": 0.18, "learning_rate": 9.409271452563113e-06, "loss": 0.1276, "step": 4532 }, { "epoch": 0.18, "learning_rate": 9.409138165435982e-06, "loss": 0.1354, "step": 4533 }, { "epoch": 0.18, "learning_rate": 9.409004878308855e-06, "loss": 0.1044, "step": 4534 }, { "epoch": 0.18, "learning_rate": 9.408871591181724e-06, "loss": 0.1052, "step": 4535 }, { "epoch": 0.18, "learning_rate": 9.408738304054595e-06, "loss": 0.1046, "step": 4536 }, { "epoch": 0.18, "learning_rate": 9.408605016927466e-06, "loss": 0.1366, "step": 4537 }, { "epoch": 0.18, "learning_rate": 9.408471729800337e-06, "loss": 0.1069, "step": 4538 }, { "epoch": 0.18, "learning_rate": 9.408338442673208e-06, "loss": 0.1458, "step": 4539 }, { "epoch": 0.18, "learning_rate": 9.408205155546079e-06, "loss": 0.0935, "step": 4540 }, { "epoch": 0.18, "learning_rate": 9.408071868418948e-06, "loss": 0.1234, "step": 4541 }, { "epoch": 0.18, "learning_rate": 9.40793858129182e-06, "loss": 0.0929, "step": 4542 }, { "epoch": 0.18, "learning_rate": 9.40780529416469e-06, "loss": 0.1545, "step": 4543 }, { "epoch": 0.18, "learning_rate": 9.407672007037561e-06, "loss": 0.0768, "step": 4544 }, { "epoch": 0.18, "learning_rate": 9.407538719910432e-06, "loss": 0.1431, "step": 4545 }, { "epoch": 0.18, "learning_rate": 9.407405432783303e-06, "loss": 0.1356, "step": 4546 }, { "epoch": 0.18, "learning_rate": 9.407272145656172e-06, "loss": 0.0916, "step": 4547 }, { "epoch": 0.18, "learning_rate": 9.407138858529043e-06, "loss": 0.0882, "step": 4548 }, { "epoch": 0.18, "learning_rate": 9.407005571401914e-06, "loss": 0.0962, "step": 4549 }, { "epoch": 0.18, "learning_rate": 9.406872284274785e-06, "loss": 0.1004, "step": 4550 }, { "epoch": 0.18, "learning_rate": 9.406738997147656e-06, "loss": 0.0941, "step": 4551 }, { "epoch": 0.18, "learning_rate": 9.406605710020526e-06, "loss": 0.1208, "step": 4552 }, { "epoch": 0.18, "learning_rate": 9.406472422893398e-06, "loss": 0.1233, "step": 4553 }, { "epoch": 0.18, "learning_rate": 9.406339135766268e-06, "loss": 0.1023, "step": 4554 }, { "epoch": 0.18, "learning_rate": 9.406205848639139e-06, "loss": 0.0928, "step": 4555 }, { "epoch": 0.18, "learning_rate": 9.40607256151201e-06, "loss": 0.115, "step": 4556 }, { "epoch": 0.18, "learning_rate": 9.40593927438488e-06, "loss": 0.1146, "step": 4557 }, { "epoch": 0.18, "learning_rate": 9.405805987257752e-06, "loss": 0.0792, "step": 4558 }, { "epoch": 0.18, "learning_rate": 9.405672700130623e-06, "loss": 0.085, "step": 4559 }, { "epoch": 0.18, "learning_rate": 9.405539413003492e-06, "loss": 0.1245, "step": 4560 }, { "epoch": 0.18, "learning_rate": 9.405406125876365e-06, "loss": 0.1283, "step": 4561 }, { "epoch": 0.18, "learning_rate": 9.405272838749234e-06, "loss": 0.0811, "step": 4562 }, { "epoch": 0.18, "learning_rate": 9.405139551622105e-06, "loss": 0.1473, "step": 4563 }, { "epoch": 0.18, "learning_rate": 9.405006264494976e-06, "loss": 0.1145, "step": 4564 }, { "epoch": 0.18, "learning_rate": 9.404872977367847e-06, "loss": 0.1017, "step": 4565 }, { "epoch": 0.18, "learning_rate": 9.404739690240718e-06, "loss": 0.0977, "step": 4566 }, { "epoch": 0.18, "learning_rate": 9.404606403113587e-06, "loss": 0.1116, "step": 4567 }, { "epoch": 0.18, "learning_rate": 9.404473115986458e-06, "loss": 0.0936, "step": 4568 }, { "epoch": 0.18, "learning_rate": 9.40433982885933e-06, "loss": 0.0833, "step": 4569 }, { "epoch": 0.18, "learning_rate": 9.4042065417322e-06, "loss": 0.1365, "step": 4570 }, { "epoch": 0.18, "learning_rate": 9.404073254605071e-06, "loss": 0.0951, "step": 4571 }, { "epoch": 0.18, "learning_rate": 9.403939967477942e-06, "loss": 0.1021, "step": 4572 }, { "epoch": 0.18, "learning_rate": 9.403806680350812e-06, "loss": 0.118, "step": 4573 }, { "epoch": 0.18, "learning_rate": 9.403673393223684e-06, "loss": 0.0925, "step": 4574 }, { "epoch": 0.18, "learning_rate": 9.403540106096554e-06, "loss": 0.0833, "step": 4575 }, { "epoch": 0.18, "learning_rate": 9.403406818969425e-06, "loss": 0.1452, "step": 4576 }, { "epoch": 0.18, "learning_rate": 9.403273531842296e-06, "loss": 0.1173, "step": 4577 }, { "epoch": 0.18, "learning_rate": 9.403140244715167e-06, "loss": 0.1162, "step": 4578 }, { "epoch": 0.18, "learning_rate": 9.403006957588038e-06, "loss": 0.1185, "step": 4579 }, { "epoch": 0.18, "learning_rate": 9.402873670460909e-06, "loss": 0.0844, "step": 4580 }, { "epoch": 0.18, "learning_rate": 9.402740383333778e-06, "loss": 0.1364, "step": 4581 }, { "epoch": 0.18, "learning_rate": 9.402607096206649e-06, "loss": 0.0871, "step": 4582 }, { "epoch": 0.18, "learning_rate": 9.40247380907952e-06, "loss": 0.108, "step": 4583 }, { "epoch": 0.18, "learning_rate": 9.40234052195239e-06, "loss": 0.1164, "step": 4584 }, { "epoch": 0.18, "learning_rate": 9.402207234825262e-06, "loss": 0.1433, "step": 4585 }, { "epoch": 0.18, "learning_rate": 9.402073947698131e-06, "loss": 0.1296, "step": 4586 }, { "epoch": 0.18, "learning_rate": 9.401940660571002e-06, "loss": 0.1182, "step": 4587 }, { "epoch": 0.18, "learning_rate": 9.401807373443873e-06, "loss": 0.0812, "step": 4588 }, { "epoch": 0.18, "learning_rate": 9.401674086316744e-06, "loss": 0.0941, "step": 4589 }, { "epoch": 0.18, "learning_rate": 9.401540799189615e-06, "loss": 0.1108, "step": 4590 }, { "epoch": 0.18, "learning_rate": 9.401407512062486e-06, "loss": 0.1721, "step": 4591 }, { "epoch": 0.18, "learning_rate": 9.401274224935355e-06, "loss": 0.099, "step": 4592 }, { "epoch": 0.18, "learning_rate": 9.401140937808228e-06, "loss": 0.1152, "step": 4593 }, { "epoch": 0.18, "learning_rate": 9.401007650681097e-06, "loss": 0.1169, "step": 4594 }, { "epoch": 0.18, "learning_rate": 9.400874363553968e-06, "loss": 0.1007, "step": 4595 }, { "epoch": 0.18, "learning_rate": 9.40074107642684e-06, "loss": 0.1319, "step": 4596 }, { "epoch": 0.18, "learning_rate": 9.40060778929971e-06, "loss": 0.0942, "step": 4597 }, { "epoch": 0.18, "learning_rate": 9.400474502172581e-06, "loss": 0.1067, "step": 4598 }, { "epoch": 0.18, "learning_rate": 9.400341215045452e-06, "loss": 0.0945, "step": 4599 }, { "epoch": 0.18, "learning_rate": 9.400207927918322e-06, "loss": 0.0925, "step": 4600 }, { "epoch": 0.18, "learning_rate": 9.400074640791194e-06, "loss": 0.129, "step": 4601 }, { "epoch": 0.18, "learning_rate": 9.399941353664064e-06, "loss": 0.1699, "step": 4602 }, { "epoch": 0.18, "learning_rate": 9.399808066536935e-06, "loss": 0.1295, "step": 4603 }, { "epoch": 0.18, "learning_rate": 9.399674779409806e-06, "loss": 0.1411, "step": 4604 }, { "epoch": 0.18, "learning_rate": 9.399541492282677e-06, "loss": 0.1025, "step": 4605 }, { "epoch": 0.18, "learning_rate": 9.399408205155548e-06, "loss": 0.0908, "step": 4606 }, { "epoch": 0.18, "learning_rate": 9.399274918028417e-06, "loss": 0.1117, "step": 4607 }, { "epoch": 0.18, "learning_rate": 9.399141630901288e-06, "loss": 0.1541, "step": 4608 }, { "epoch": 0.18, "learning_rate": 9.399008343774159e-06, "loss": 0.0832, "step": 4609 }, { "epoch": 0.18, "learning_rate": 9.39887505664703e-06, "loss": 0.1024, "step": 4610 }, { "epoch": 0.18, "learning_rate": 9.398741769519901e-06, "loss": 0.122, "step": 4611 }, { "epoch": 0.18, "learning_rate": 9.398608482392772e-06, "loss": 0.1198, "step": 4612 }, { "epoch": 0.18, "learning_rate": 9.398475195265641e-06, "loss": 0.1073, "step": 4613 }, { "epoch": 0.18, "learning_rate": 9.398341908138512e-06, "loss": 0.0731, "step": 4614 }, { "epoch": 0.18, "learning_rate": 9.398208621011383e-06, "loss": 0.115, "step": 4615 }, { "epoch": 0.18, "learning_rate": 9.398075333884254e-06, "loss": 0.1041, "step": 4616 }, { "epoch": 0.18, "learning_rate": 9.397942046757125e-06, "loss": 0.09, "step": 4617 }, { "epoch": 0.18, "learning_rate": 9.397808759629996e-06, "loss": 0.1395, "step": 4618 }, { "epoch": 0.18, "learning_rate": 9.397675472502865e-06, "loss": 0.117, "step": 4619 }, { "epoch": 0.18, "learning_rate": 9.397542185375738e-06, "loss": 0.0953, "step": 4620 }, { "epoch": 0.18, "learning_rate": 9.397408898248607e-06, "loss": 0.1129, "step": 4621 }, { "epoch": 0.18, "learning_rate": 9.397275611121478e-06, "loss": 0.0982, "step": 4622 }, { "epoch": 0.18, "learning_rate": 9.39714232399435e-06, "loss": 0.1257, "step": 4623 }, { "epoch": 0.18, "learning_rate": 9.39700903686722e-06, "loss": 0.1224, "step": 4624 }, { "epoch": 0.18, "learning_rate": 9.396875749740091e-06, "loss": 0.117, "step": 4625 }, { "epoch": 0.18, "learning_rate": 9.39674246261296e-06, "loss": 0.1647, "step": 4626 }, { "epoch": 0.18, "learning_rate": 9.396609175485832e-06, "loss": 0.0971, "step": 4627 }, { "epoch": 0.18, "learning_rate": 9.396475888358703e-06, "loss": 0.1703, "step": 4628 }, { "epoch": 0.18, "learning_rate": 9.396342601231574e-06, "loss": 0.1123, "step": 4629 }, { "epoch": 0.18, "learning_rate": 9.396209314104445e-06, "loss": 0.0726, "step": 4630 }, { "epoch": 0.18, "learning_rate": 9.396076026977316e-06, "loss": 0.1078, "step": 4631 }, { "epoch": 0.18, "learning_rate": 9.395942739850185e-06, "loss": 0.0957, "step": 4632 }, { "epoch": 0.19, "learning_rate": 9.395809452723058e-06, "loss": 0.0979, "step": 4633 }, { "epoch": 0.19, "learning_rate": 9.395676165595927e-06, "loss": 0.0793, "step": 4634 }, { "epoch": 0.19, "learning_rate": 9.395542878468798e-06, "loss": 0.1275, "step": 4635 }, { "epoch": 0.19, "learning_rate": 9.395409591341669e-06, "loss": 0.1174, "step": 4636 }, { "epoch": 0.19, "learning_rate": 9.39527630421454e-06, "loss": 0.114, "step": 4637 }, { "epoch": 0.19, "learning_rate": 9.395143017087411e-06, "loss": 0.1569, "step": 4638 }, { "epoch": 0.19, "learning_rate": 9.395009729960282e-06, "loss": 0.1034, "step": 4639 }, { "epoch": 0.19, "learning_rate": 9.394876442833151e-06, "loss": 0.108, "step": 4640 }, { "epoch": 0.19, "learning_rate": 9.394743155706024e-06, "loss": 0.1044, "step": 4641 }, { "epoch": 0.19, "learning_rate": 9.394609868578893e-06, "loss": 0.1265, "step": 4642 }, { "epoch": 0.19, "learning_rate": 9.394476581451764e-06, "loss": 0.1223, "step": 4643 }, { "epoch": 0.19, "learning_rate": 9.394343294324635e-06, "loss": 0.1362, "step": 4644 }, { "epoch": 0.19, "learning_rate": 9.394210007197505e-06, "loss": 0.0952, "step": 4645 }, { "epoch": 0.19, "learning_rate": 9.394076720070377e-06, "loss": 0.1094, "step": 4646 }, { "epoch": 0.19, "learning_rate": 9.393943432943247e-06, "loss": 0.1292, "step": 4647 }, { "epoch": 0.19, "learning_rate": 9.393810145816118e-06, "loss": 0.111, "step": 4648 }, { "epoch": 0.19, "learning_rate": 9.393676858688989e-06, "loss": 0.1326, "step": 4649 }, { "epoch": 0.19, "learning_rate": 9.39354357156186e-06, "loss": 0.1097, "step": 4650 }, { "epoch": 0.19, "learning_rate": 9.39341028443473e-06, "loss": 0.1172, "step": 4651 }, { "epoch": 0.19, "learning_rate": 9.393276997307602e-06, "loss": 0.1086, "step": 4652 }, { "epoch": 0.19, "learning_rate": 9.39314371018047e-06, "loss": 0.1017, "step": 4653 }, { "epoch": 0.19, "learning_rate": 9.393010423053342e-06, "loss": 0.1169, "step": 4654 }, { "epoch": 0.19, "learning_rate": 9.392877135926213e-06, "loss": 0.0838, "step": 4655 }, { "epoch": 0.19, "learning_rate": 9.392743848799084e-06, "loss": 0.1351, "step": 4656 }, { "epoch": 0.19, "learning_rate": 9.392610561671955e-06, "loss": 0.1376, "step": 4657 }, { "epoch": 0.19, "learning_rate": 9.392477274544826e-06, "loss": 0.1026, "step": 4658 }, { "epoch": 0.19, "learning_rate": 9.392343987417695e-06, "loss": 0.0756, "step": 4659 }, { "epoch": 0.19, "learning_rate": 9.392210700290568e-06, "loss": 0.113, "step": 4660 }, { "epoch": 0.19, "learning_rate": 9.392077413163437e-06, "loss": 0.0855, "step": 4661 }, { "epoch": 0.19, "learning_rate": 9.391944126036308e-06, "loss": 0.1107, "step": 4662 }, { "epoch": 0.19, "learning_rate": 9.391810838909179e-06, "loss": 0.1038, "step": 4663 }, { "epoch": 0.19, "learning_rate": 9.39167755178205e-06, "loss": 0.1203, "step": 4664 }, { "epoch": 0.19, "learning_rate": 9.391544264654921e-06, "loss": 0.0826, "step": 4665 }, { "epoch": 0.19, "learning_rate": 9.39141097752779e-06, "loss": 0.0881, "step": 4666 }, { "epoch": 0.19, "learning_rate": 9.391277690400661e-06, "loss": 0.1175, "step": 4667 }, { "epoch": 0.19, "learning_rate": 9.391144403273532e-06, "loss": 0.1065, "step": 4668 }, { "epoch": 0.19, "learning_rate": 9.391011116146403e-06, "loss": 0.1224, "step": 4669 }, { "epoch": 0.19, "learning_rate": 9.390877829019274e-06, "loss": 0.0956, "step": 4670 }, { "epoch": 0.19, "learning_rate": 9.390744541892145e-06, "loss": 0.0851, "step": 4671 }, { "epoch": 0.19, "learning_rate": 9.390611254765015e-06, "loss": 0.0862, "step": 4672 }, { "epoch": 0.19, "learning_rate": 9.390477967637887e-06, "loss": 0.1195, "step": 4673 }, { "epoch": 0.19, "learning_rate": 9.390344680510757e-06, "loss": 0.1465, "step": 4674 }, { "epoch": 0.19, "learning_rate": 9.390211393383628e-06, "loss": 0.1132, "step": 4675 }, { "epoch": 0.19, "learning_rate": 9.390078106256499e-06, "loss": 0.1353, "step": 4676 }, { "epoch": 0.19, "learning_rate": 9.38994481912937e-06, "loss": 0.1391, "step": 4677 }, { "epoch": 0.19, "learning_rate": 9.38981153200224e-06, "loss": 0.0825, "step": 4678 }, { "epoch": 0.19, "learning_rate": 9.389678244875112e-06, "loss": 0.1246, "step": 4679 }, { "epoch": 0.19, "learning_rate": 9.389544957747981e-06, "loss": 0.1294, "step": 4680 }, { "epoch": 0.19, "learning_rate": 9.389411670620852e-06, "loss": 0.1422, "step": 4681 }, { "epoch": 0.19, "learning_rate": 9.389278383493723e-06, "loss": 0.0885, "step": 4682 }, { "epoch": 0.19, "learning_rate": 9.389145096366594e-06, "loss": 0.1342, "step": 4683 }, { "epoch": 0.19, "learning_rate": 9.389011809239465e-06, "loss": 0.0886, "step": 4684 }, { "epoch": 0.19, "learning_rate": 9.388878522112334e-06, "loss": 0.105, "step": 4685 }, { "epoch": 0.19, "learning_rate": 9.388745234985205e-06, "loss": 0.1025, "step": 4686 }, { "epoch": 0.19, "learning_rate": 9.388611947858076e-06, "loss": 0.1152, "step": 4687 }, { "epoch": 0.19, "learning_rate": 9.388478660730947e-06, "loss": 0.1354, "step": 4688 }, { "epoch": 0.19, "learning_rate": 9.388345373603818e-06, "loss": 0.095, "step": 4689 }, { "epoch": 0.19, "learning_rate": 9.388212086476689e-06, "loss": 0.1386, "step": 4690 }, { "epoch": 0.19, "learning_rate": 9.388078799349558e-06, "loss": 0.0822, "step": 4691 }, { "epoch": 0.19, "learning_rate": 9.387945512222431e-06, "loss": 0.0925, "step": 4692 }, { "epoch": 0.19, "learning_rate": 9.3878122250953e-06, "loss": 0.1315, "step": 4693 }, { "epoch": 0.19, "learning_rate": 9.387678937968171e-06, "loss": 0.1336, "step": 4694 }, { "epoch": 0.19, "learning_rate": 9.387545650841042e-06, "loss": 0.074, "step": 4695 }, { "epoch": 0.19, "learning_rate": 9.387412363713913e-06, "loss": 0.136, "step": 4696 }, { "epoch": 0.19, "learning_rate": 9.387279076586784e-06, "loss": 0.0927, "step": 4697 }, { "epoch": 0.19, "learning_rate": 9.387145789459655e-06, "loss": 0.102, "step": 4698 }, { "epoch": 0.19, "learning_rate": 9.387012502332525e-06, "loss": 0.1138, "step": 4699 }, { "epoch": 0.19, "learning_rate": 9.386879215205397e-06, "loss": 0.1125, "step": 4700 }, { "epoch": 0.19, "learning_rate": 9.386745928078267e-06, "loss": 0.1016, "step": 4701 }, { "epoch": 0.19, "learning_rate": 9.386612640951138e-06, "loss": 0.1669, "step": 4702 }, { "epoch": 0.19, "learning_rate": 9.386479353824009e-06, "loss": 0.1319, "step": 4703 }, { "epoch": 0.19, "learning_rate": 9.386346066696878e-06, "loss": 0.1468, "step": 4704 }, { "epoch": 0.19, "learning_rate": 9.38621277956975e-06, "loss": 0.0777, "step": 4705 }, { "epoch": 0.19, "learning_rate": 9.38607949244262e-06, "loss": 0.1643, "step": 4706 }, { "epoch": 0.19, "learning_rate": 9.385946205315491e-06, "loss": 0.1245, "step": 4707 }, { "epoch": 0.19, "learning_rate": 9.385812918188362e-06, "loss": 0.106, "step": 4708 }, { "epoch": 0.19, "learning_rate": 9.385679631061233e-06, "loss": 0.1074, "step": 4709 }, { "epoch": 0.19, "learning_rate": 9.385546343934104e-06, "loss": 0.0859, "step": 4710 }, { "epoch": 0.19, "learning_rate": 9.385413056806975e-06, "loss": 0.1338, "step": 4711 }, { "epoch": 0.19, "learning_rate": 9.385279769679844e-06, "loss": 0.1148, "step": 4712 }, { "epoch": 0.19, "learning_rate": 9.385146482552717e-06, "loss": 0.0783, "step": 4713 }, { "epoch": 0.19, "learning_rate": 9.385013195425586e-06, "loss": 0.0867, "step": 4714 }, { "epoch": 0.19, "learning_rate": 9.384879908298457e-06, "loss": 0.1324, "step": 4715 }, { "epoch": 0.19, "learning_rate": 9.384746621171328e-06, "loss": 0.1161, "step": 4716 }, { "epoch": 0.19, "learning_rate": 9.3846133340442e-06, "loss": 0.1113, "step": 4717 }, { "epoch": 0.19, "learning_rate": 9.38448004691707e-06, "loss": 0.1129, "step": 4718 }, { "epoch": 0.19, "learning_rate": 9.384346759789941e-06, "loss": 0.0717, "step": 4719 }, { "epoch": 0.19, "learning_rate": 9.38421347266281e-06, "loss": 0.0627, "step": 4720 }, { "epoch": 0.19, "learning_rate": 9.384080185535682e-06, "loss": 0.076, "step": 4721 }, { "epoch": 0.19, "learning_rate": 9.383946898408553e-06, "loss": 0.1332, "step": 4722 }, { "epoch": 0.19, "learning_rate": 9.383813611281424e-06, "loss": 0.1145, "step": 4723 }, { "epoch": 0.19, "learning_rate": 9.383680324154294e-06, "loss": 0.1583, "step": 4724 }, { "epoch": 0.19, "learning_rate": 9.383547037027164e-06, "loss": 0.1034, "step": 4725 }, { "epoch": 0.19, "learning_rate": 9.383413749900035e-06, "loss": 0.0888, "step": 4726 }, { "epoch": 0.19, "learning_rate": 9.383280462772906e-06, "loss": 0.0981, "step": 4727 }, { "epoch": 0.19, "learning_rate": 9.383147175645777e-06, "loss": 0.1238, "step": 4728 }, { "epoch": 0.19, "learning_rate": 9.383013888518648e-06, "loss": 0.1148, "step": 4729 }, { "epoch": 0.19, "learning_rate": 9.382880601391519e-06, "loss": 0.137, "step": 4730 }, { "epoch": 0.19, "learning_rate": 9.382747314264388e-06, "loss": 0.1386, "step": 4731 }, { "epoch": 0.19, "learning_rate": 9.38261402713726e-06, "loss": 0.132, "step": 4732 }, { "epoch": 0.19, "learning_rate": 9.38248074001013e-06, "loss": 0.0945, "step": 4733 }, { "epoch": 0.19, "learning_rate": 9.382347452883001e-06, "loss": 0.098, "step": 4734 }, { "epoch": 0.19, "learning_rate": 9.382214165755872e-06, "loss": 0.1628, "step": 4735 }, { "epoch": 0.19, "learning_rate": 9.382080878628743e-06, "loss": 0.1409, "step": 4736 }, { "epoch": 0.19, "learning_rate": 9.381947591501614e-06, "loss": 0.0997, "step": 4737 }, { "epoch": 0.19, "learning_rate": 9.381814304374485e-06, "loss": 0.1371, "step": 4738 }, { "epoch": 0.19, "learning_rate": 9.381681017247354e-06, "loss": 0.0808, "step": 4739 }, { "epoch": 0.19, "learning_rate": 9.381547730120227e-06, "loss": 0.1213, "step": 4740 }, { "epoch": 0.19, "learning_rate": 9.381414442993096e-06, "loss": 0.0935, "step": 4741 }, { "epoch": 0.19, "learning_rate": 9.381281155865967e-06, "loss": 0.1144, "step": 4742 }, { "epoch": 0.19, "learning_rate": 9.381147868738838e-06, "loss": 0.1679, "step": 4743 }, { "epoch": 0.19, "learning_rate": 9.381014581611708e-06, "loss": 0.1444, "step": 4744 }, { "epoch": 0.19, "learning_rate": 9.38088129448458e-06, "loss": 0.1222, "step": 4745 }, { "epoch": 0.19, "learning_rate": 9.38074800735745e-06, "loss": 0.1141, "step": 4746 }, { "epoch": 0.19, "learning_rate": 9.38061472023032e-06, "loss": 0.0971, "step": 4747 }, { "epoch": 0.19, "learning_rate": 9.380481433103192e-06, "loss": 0.0756, "step": 4748 }, { "epoch": 0.19, "learning_rate": 9.380348145976063e-06, "loss": 0.1216, "step": 4749 }, { "epoch": 0.19, "learning_rate": 9.380214858848934e-06, "loss": 0.1174, "step": 4750 }, { "epoch": 0.19, "learning_rate": 9.380081571721805e-06, "loss": 0.0829, "step": 4751 }, { "epoch": 0.19, "learning_rate": 9.379948284594674e-06, "loss": 0.0829, "step": 4752 }, { "epoch": 0.19, "learning_rate": 9.379814997467545e-06, "loss": 0.1015, "step": 4753 }, { "epoch": 0.19, "learning_rate": 9.379681710340416e-06, "loss": 0.1255, "step": 4754 }, { "epoch": 0.19, "learning_rate": 9.379548423213287e-06, "loss": 0.1301, "step": 4755 }, { "epoch": 0.19, "learning_rate": 9.379415136086158e-06, "loss": 0.1246, "step": 4756 }, { "epoch": 0.19, "learning_rate": 9.379281848959029e-06, "loss": 0.1084, "step": 4757 }, { "epoch": 0.19, "learning_rate": 9.379148561831898e-06, "loss": 0.1336, "step": 4758 }, { "epoch": 0.19, "learning_rate": 9.37901527470477e-06, "loss": 0.126, "step": 4759 }, { "epoch": 0.19, "learning_rate": 9.37888198757764e-06, "loss": 0.1093, "step": 4760 }, { "epoch": 0.19, "learning_rate": 9.378748700450511e-06, "loss": 0.1273, "step": 4761 }, { "epoch": 0.19, "learning_rate": 9.378615413323382e-06, "loss": 0.1199, "step": 4762 }, { "epoch": 0.19, "learning_rate": 9.378482126196251e-06, "loss": 0.1076, "step": 4763 }, { "epoch": 0.19, "learning_rate": 9.378348839069124e-06, "loss": 0.0865, "step": 4764 }, { "epoch": 0.19, "learning_rate": 9.378215551941993e-06, "loss": 0.1059, "step": 4765 }, { "epoch": 0.19, "learning_rate": 9.378082264814864e-06, "loss": 0.1151, "step": 4766 }, { "epoch": 0.19, "learning_rate": 9.377948977687735e-06, "loss": 0.1228, "step": 4767 }, { "epoch": 0.19, "learning_rate": 9.377815690560606e-06, "loss": 0.1904, "step": 4768 }, { "epoch": 0.19, "learning_rate": 9.377682403433477e-06, "loss": 0.109, "step": 4769 }, { "epoch": 0.19, "learning_rate": 9.377549116306348e-06, "loss": 0.1219, "step": 4770 }, { "epoch": 0.19, "learning_rate": 9.377415829179218e-06, "loss": 0.1178, "step": 4771 }, { "epoch": 0.19, "learning_rate": 9.37728254205209e-06, "loss": 0.1259, "step": 4772 }, { "epoch": 0.19, "learning_rate": 9.37714925492496e-06, "loss": 0.1035, "step": 4773 }, { "epoch": 0.19, "learning_rate": 9.37701596779783e-06, "loss": 0.0903, "step": 4774 }, { "epoch": 0.19, "learning_rate": 9.376882680670702e-06, "loss": 0.091, "step": 4775 }, { "epoch": 0.19, "learning_rate": 9.376749393543573e-06, "loss": 0.0873, "step": 4776 }, { "epoch": 0.19, "learning_rate": 9.376616106416444e-06, "loss": 0.1145, "step": 4777 }, { "epoch": 0.19, "learning_rate": 9.376482819289315e-06, "loss": 0.1196, "step": 4778 }, { "epoch": 0.19, "learning_rate": 9.376349532162184e-06, "loss": 0.1138, "step": 4779 }, { "epoch": 0.19, "learning_rate": 9.376216245035055e-06, "loss": 0.095, "step": 4780 }, { "epoch": 0.19, "learning_rate": 9.376082957907926e-06, "loss": 0.1059, "step": 4781 }, { "epoch": 0.19, "learning_rate": 9.375949670780797e-06, "loss": 0.1074, "step": 4782 }, { "epoch": 0.19, "learning_rate": 9.375816383653668e-06, "loss": 0.1122, "step": 4783 }, { "epoch": 0.19, "learning_rate": 9.375683096526537e-06, "loss": 0.104, "step": 4784 }, { "epoch": 0.19, "learning_rate": 9.375549809399408e-06, "loss": 0.1091, "step": 4785 }, { "epoch": 0.19, "learning_rate": 9.37541652227228e-06, "loss": 0.1149, "step": 4786 }, { "epoch": 0.19, "learning_rate": 9.37528323514515e-06, "loss": 0.1403, "step": 4787 }, { "epoch": 0.19, "learning_rate": 9.375149948018021e-06, "loss": 0.1297, "step": 4788 }, { "epoch": 0.19, "learning_rate": 9.375016660890892e-06, "loss": 0.092, "step": 4789 }, { "epoch": 0.19, "learning_rate": 9.374883373763762e-06, "loss": 0.1189, "step": 4790 }, { "epoch": 0.19, "learning_rate": 9.374750086636634e-06, "loss": 0.1721, "step": 4791 }, { "epoch": 0.19, "learning_rate": 9.374616799509504e-06, "loss": 0.0842, "step": 4792 }, { "epoch": 0.19, "learning_rate": 9.374483512382374e-06, "loss": 0.1219, "step": 4793 }, { "epoch": 0.19, "learning_rate": 9.374350225255245e-06, "loss": 0.1413, "step": 4794 }, { "epoch": 0.19, "learning_rate": 9.374216938128116e-06, "loss": 0.0655, "step": 4795 }, { "epoch": 0.19, "learning_rate": 9.374083651000987e-06, "loss": 0.1524, "step": 4796 }, { "epoch": 0.19, "learning_rate": 9.373950363873858e-06, "loss": 0.1184, "step": 4797 }, { "epoch": 0.19, "learning_rate": 9.373817076746728e-06, "loss": 0.1225, "step": 4798 }, { "epoch": 0.19, "learning_rate": 9.3736837896196e-06, "loss": 0.1291, "step": 4799 }, { "epoch": 0.19, "learning_rate": 9.37355050249247e-06, "loss": 0.1062, "step": 4800 }, { "epoch": 0.19, "learning_rate": 9.37341721536534e-06, "loss": 0.1186, "step": 4801 }, { "epoch": 0.19, "learning_rate": 9.373283928238212e-06, "loss": 0.1362, "step": 4802 }, { "epoch": 0.19, "learning_rate": 9.373150641111081e-06, "loss": 0.1153, "step": 4803 }, { "epoch": 0.19, "learning_rate": 9.373017353983954e-06, "loss": 0.108, "step": 4804 }, { "epoch": 0.19, "learning_rate": 9.372884066856823e-06, "loss": 0.1126, "step": 4805 }, { "epoch": 0.19, "learning_rate": 9.372750779729694e-06, "loss": 0.116, "step": 4806 }, { "epoch": 0.19, "learning_rate": 9.372617492602565e-06, "loss": 0.0812, "step": 4807 }, { "epoch": 0.19, "learning_rate": 9.372484205475436e-06, "loss": 0.1047, "step": 4808 }, { "epoch": 0.19, "learning_rate": 9.372350918348307e-06, "loss": 0.1292, "step": 4809 }, { "epoch": 0.19, "learning_rate": 9.372217631221178e-06, "loss": 0.1091, "step": 4810 }, { "epoch": 0.19, "learning_rate": 9.372084344094047e-06, "loss": 0.1692, "step": 4811 }, { "epoch": 0.19, "learning_rate": 9.37195105696692e-06, "loss": 0.1521, "step": 4812 }, { "epoch": 0.19, "learning_rate": 9.37181776983979e-06, "loss": 0.1104, "step": 4813 }, { "epoch": 0.19, "learning_rate": 9.37168448271266e-06, "loss": 0.072, "step": 4814 }, { "epoch": 0.19, "learning_rate": 9.371551195585531e-06, "loss": 0.1198, "step": 4815 }, { "epoch": 0.19, "learning_rate": 9.371417908458402e-06, "loss": 0.089, "step": 4816 }, { "epoch": 0.19, "learning_rate": 9.371284621331273e-06, "loss": 0.1568, "step": 4817 }, { "epoch": 0.19, "learning_rate": 9.371151334204144e-06, "loss": 0.1097, "step": 4818 }, { "epoch": 0.19, "learning_rate": 9.371018047077014e-06, "loss": 0.1125, "step": 4819 }, { "epoch": 0.19, "learning_rate": 9.370884759949885e-06, "loss": 0.0836, "step": 4820 }, { "epoch": 0.19, "learning_rate": 9.370751472822756e-06, "loss": 0.0874, "step": 4821 }, { "epoch": 0.19, "learning_rate": 9.370618185695627e-06, "loss": 0.0917, "step": 4822 }, { "epoch": 0.19, "learning_rate": 9.370484898568498e-06, "loss": 0.1138, "step": 4823 }, { "epoch": 0.19, "learning_rate": 9.370351611441367e-06, "loss": 0.1344, "step": 4824 }, { "epoch": 0.19, "learning_rate": 9.370218324314238e-06, "loss": 0.1335, "step": 4825 }, { "epoch": 0.19, "learning_rate": 9.370085037187109e-06, "loss": 0.1311, "step": 4826 }, { "epoch": 0.19, "learning_rate": 9.36995175005998e-06, "loss": 0.1285, "step": 4827 }, { "epoch": 0.19, "learning_rate": 9.36981846293285e-06, "loss": 0.105, "step": 4828 }, { "epoch": 0.19, "learning_rate": 9.369685175805722e-06, "loss": 0.0854, "step": 4829 }, { "epoch": 0.19, "learning_rate": 9.369551888678591e-06, "loss": 0.0942, "step": 4830 }, { "epoch": 0.19, "learning_rate": 9.369418601551464e-06, "loss": 0.0971, "step": 4831 }, { "epoch": 0.19, "learning_rate": 9.369285314424333e-06, "loss": 0.0829, "step": 4832 }, { "epoch": 0.19, "learning_rate": 9.369152027297204e-06, "loss": 0.0878, "step": 4833 }, { "epoch": 0.19, "learning_rate": 9.369018740170075e-06, "loss": 0.1252, "step": 4834 }, { "epoch": 0.19, "learning_rate": 9.368885453042946e-06, "loss": 0.089, "step": 4835 }, { "epoch": 0.19, "learning_rate": 9.368752165915817e-06, "loss": 0.0864, "step": 4836 }, { "epoch": 0.19, "learning_rate": 9.368618878788688e-06, "loss": 0.114, "step": 4837 }, { "epoch": 0.19, "learning_rate": 9.368485591661557e-06, "loss": 0.0918, "step": 4838 }, { "epoch": 0.19, "learning_rate": 9.36835230453443e-06, "loss": 0.1184, "step": 4839 }, { "epoch": 0.19, "learning_rate": 9.3682190174073e-06, "loss": 0.0823, "step": 4840 }, { "epoch": 0.19, "learning_rate": 9.36808573028017e-06, "loss": 0.1368, "step": 4841 }, { "epoch": 0.19, "learning_rate": 9.367952443153041e-06, "loss": 0.0964, "step": 4842 }, { "epoch": 0.19, "learning_rate": 9.36781915602591e-06, "loss": 0.1285, "step": 4843 }, { "epoch": 0.19, "learning_rate": 9.367685868898783e-06, "loss": 0.0942, "step": 4844 }, { "epoch": 0.19, "learning_rate": 9.367552581771653e-06, "loss": 0.1096, "step": 4845 }, { "epoch": 0.19, "learning_rate": 9.367419294644524e-06, "loss": 0.0749, "step": 4846 }, { "epoch": 0.19, "learning_rate": 9.367286007517395e-06, "loss": 0.1329, "step": 4847 }, { "epoch": 0.19, "learning_rate": 9.367152720390266e-06, "loss": 0.1223, "step": 4848 }, { "epoch": 0.19, "learning_rate": 9.367019433263137e-06, "loss": 0.0889, "step": 4849 }, { "epoch": 0.19, "learning_rate": 9.366886146136008e-06, "loss": 0.1195, "step": 4850 }, { "epoch": 0.19, "learning_rate": 9.366752859008877e-06, "loss": 0.1341, "step": 4851 }, { "epoch": 0.19, "learning_rate": 9.366619571881748e-06, "loss": 0.0862, "step": 4852 }, { "epoch": 0.19, "learning_rate": 9.366486284754619e-06, "loss": 0.1007, "step": 4853 }, { "epoch": 0.19, "learning_rate": 9.36635299762749e-06, "loss": 0.115, "step": 4854 }, { "epoch": 0.19, "learning_rate": 9.366219710500361e-06, "loss": 0.0823, "step": 4855 }, { "epoch": 0.19, "learning_rate": 9.366086423373232e-06, "loss": 0.0966, "step": 4856 }, { "epoch": 0.19, "learning_rate": 9.365953136246101e-06, "loss": 0.1079, "step": 4857 }, { "epoch": 0.19, "learning_rate": 9.365819849118974e-06, "loss": 0.1063, "step": 4858 }, { "epoch": 0.19, "learning_rate": 9.365686561991843e-06, "loss": 0.1483, "step": 4859 }, { "epoch": 0.19, "learning_rate": 9.365553274864714e-06, "loss": 0.1056, "step": 4860 }, { "epoch": 0.19, "learning_rate": 9.365419987737585e-06, "loss": 0.1048, "step": 4861 }, { "epoch": 0.19, "learning_rate": 9.365286700610454e-06, "loss": 0.1414, "step": 4862 }, { "epoch": 0.19, "learning_rate": 9.365153413483327e-06, "loss": 0.098, "step": 4863 }, { "epoch": 0.19, "learning_rate": 9.365020126356196e-06, "loss": 0.1269, "step": 4864 }, { "epoch": 0.19, "learning_rate": 9.364886839229067e-06, "loss": 0.0964, "step": 4865 }, { "epoch": 0.19, "learning_rate": 9.364753552101938e-06, "loss": 0.1164, "step": 4866 }, { "epoch": 0.19, "learning_rate": 9.36462026497481e-06, "loss": 0.121, "step": 4867 }, { "epoch": 0.19, "learning_rate": 9.36448697784768e-06, "loss": 0.1122, "step": 4868 }, { "epoch": 0.19, "learning_rate": 9.364353690720551e-06, "loss": 0.1455, "step": 4869 }, { "epoch": 0.19, "learning_rate": 9.36422040359342e-06, "loss": 0.0931, "step": 4870 }, { "epoch": 0.19, "learning_rate": 9.364087116466293e-06, "loss": 0.1065, "step": 4871 }, { "epoch": 0.19, "learning_rate": 9.363953829339163e-06, "loss": 0.0692, "step": 4872 }, { "epoch": 0.19, "learning_rate": 9.363820542212034e-06, "loss": 0.1132, "step": 4873 }, { "epoch": 0.19, "learning_rate": 9.363687255084905e-06, "loss": 0.1077, "step": 4874 }, { "epoch": 0.19, "learning_rate": 9.363553967957776e-06, "loss": 0.1146, "step": 4875 }, { "epoch": 0.19, "learning_rate": 9.363420680830647e-06, "loss": 0.0889, "step": 4876 }, { "epoch": 0.19, "learning_rate": 9.363287393703518e-06, "loss": 0.0793, "step": 4877 }, { "epoch": 0.19, "learning_rate": 9.363154106576387e-06, "loss": 0.1191, "step": 4878 }, { "epoch": 0.19, "learning_rate": 9.36302081944926e-06, "loss": 0.0982, "step": 4879 }, { "epoch": 0.19, "learning_rate": 9.362887532322129e-06, "loss": 0.1248, "step": 4880 }, { "epoch": 0.19, "learning_rate": 9.362754245195e-06, "loss": 0.1207, "step": 4881 }, { "epoch": 0.19, "learning_rate": 9.362620958067871e-06, "loss": 0.1184, "step": 4882 }, { "epoch": 0.19, "learning_rate": 9.36248767094074e-06, "loss": 0.1359, "step": 4883 }, { "epoch": 0.2, "learning_rate": 9.362354383813613e-06, "loss": 0.0998, "step": 4884 }, { "epoch": 0.2, "learning_rate": 9.362221096686482e-06, "loss": 0.1094, "step": 4885 }, { "epoch": 0.2, "learning_rate": 9.362087809559353e-06, "loss": 0.1091, "step": 4886 }, { "epoch": 0.2, "learning_rate": 9.361954522432224e-06, "loss": 0.1061, "step": 4887 }, { "epoch": 0.2, "learning_rate": 9.361821235305095e-06, "loss": 0.1329, "step": 4888 }, { "epoch": 0.2, "learning_rate": 9.361687948177966e-06, "loss": 0.0866, "step": 4889 }, { "epoch": 0.2, "learning_rate": 9.361554661050837e-06, "loss": 0.1311, "step": 4890 }, { "epoch": 0.2, "learning_rate": 9.361421373923707e-06, "loss": 0.1397, "step": 4891 }, { "epoch": 0.2, "learning_rate": 9.361288086796578e-06, "loss": 0.0827, "step": 4892 }, { "epoch": 0.2, "learning_rate": 9.361154799669449e-06, "loss": 0.1442, "step": 4893 }, { "epoch": 0.2, "learning_rate": 9.36102151254232e-06, "loss": 0.1096, "step": 4894 }, { "epoch": 0.2, "learning_rate": 9.36088822541519e-06, "loss": 0.0784, "step": 4895 }, { "epoch": 0.2, "learning_rate": 9.360754938288062e-06, "loss": 0.0994, "step": 4896 }, { "epoch": 0.2, "learning_rate": 9.36062165116093e-06, "loss": 0.1171, "step": 4897 }, { "epoch": 0.2, "learning_rate": 9.360488364033804e-06, "loss": 0.14, "step": 4898 }, { "epoch": 0.2, "learning_rate": 9.360355076906673e-06, "loss": 0.1525, "step": 4899 }, { "epoch": 0.2, "learning_rate": 9.360221789779544e-06, "loss": 0.1159, "step": 4900 }, { "epoch": 0.2, "learning_rate": 9.360088502652415e-06, "loss": 0.0958, "step": 4901 }, { "epoch": 0.2, "learning_rate": 9.359955215525284e-06, "loss": 0.0904, "step": 4902 }, { "epoch": 0.2, "learning_rate": 9.359821928398157e-06, "loss": 0.1269, "step": 4903 }, { "epoch": 0.2, "learning_rate": 9.359688641271026e-06, "loss": 0.0922, "step": 4904 }, { "epoch": 0.2, "learning_rate": 9.359555354143897e-06, "loss": 0.0891, "step": 4905 }, { "epoch": 0.2, "learning_rate": 9.359422067016768e-06, "loss": 0.1056, "step": 4906 }, { "epoch": 0.2, "learning_rate": 9.359288779889639e-06, "loss": 0.1018, "step": 4907 }, { "epoch": 0.2, "learning_rate": 9.35915549276251e-06, "loss": 0.1068, "step": 4908 }, { "epoch": 0.2, "learning_rate": 9.359022205635381e-06, "loss": 0.104, "step": 4909 }, { "epoch": 0.2, "learning_rate": 9.35888891850825e-06, "loss": 0.1482, "step": 4910 }, { "epoch": 0.2, "learning_rate": 9.358755631381123e-06, "loss": 0.1028, "step": 4911 }, { "epoch": 0.2, "learning_rate": 9.358622344253992e-06, "loss": 0.1345, "step": 4912 }, { "epoch": 0.2, "learning_rate": 9.358489057126863e-06, "loss": 0.0827, "step": 4913 }, { "epoch": 0.2, "learning_rate": 9.358355769999734e-06, "loss": 0.1238, "step": 4914 }, { "epoch": 0.2, "learning_rate": 9.358222482872605e-06, "loss": 0.1439, "step": 4915 }, { "epoch": 0.2, "learning_rate": 9.358089195745476e-06, "loss": 0.1081, "step": 4916 }, { "epoch": 0.2, "learning_rate": 9.357955908618347e-06, "loss": 0.1277, "step": 4917 }, { "epoch": 0.2, "learning_rate": 9.357822621491217e-06, "loss": 0.1467, "step": 4918 }, { "epoch": 0.2, "learning_rate": 9.357689334364088e-06, "loss": 0.1241, "step": 4919 }, { "epoch": 0.2, "learning_rate": 9.357556047236959e-06, "loss": 0.1083, "step": 4920 }, { "epoch": 0.2, "learning_rate": 9.35742276010983e-06, "loss": 0.1717, "step": 4921 }, { "epoch": 0.2, "learning_rate": 9.3572894729827e-06, "loss": 0.1092, "step": 4922 }, { "epoch": 0.2, "learning_rate": 9.35715618585557e-06, "loss": 0.0819, "step": 4923 }, { "epoch": 0.2, "learning_rate": 9.357022898728441e-06, "loss": 0.1281, "step": 4924 }, { "epoch": 0.2, "learning_rate": 9.356889611601312e-06, "loss": 0.1148, "step": 4925 }, { "epoch": 0.2, "learning_rate": 9.356756324474183e-06, "loss": 0.1329, "step": 4926 }, { "epoch": 0.2, "learning_rate": 9.356623037347054e-06, "loss": 0.133, "step": 4927 }, { "epoch": 0.2, "learning_rate": 9.356489750219925e-06, "loss": 0.1589, "step": 4928 }, { "epoch": 0.2, "learning_rate": 9.356356463092794e-06, "loss": 0.1036, "step": 4929 }, { "epoch": 0.2, "learning_rate": 9.356223175965667e-06, "loss": 0.0942, "step": 4930 }, { "epoch": 0.2, "learning_rate": 9.356089888838536e-06, "loss": 0.0809, "step": 4931 }, { "epoch": 0.2, "learning_rate": 9.355956601711407e-06, "loss": 0.0996, "step": 4932 }, { "epoch": 0.2, "learning_rate": 9.355823314584278e-06, "loss": 0.1014, "step": 4933 }, { "epoch": 0.2, "learning_rate": 9.35569002745715e-06, "loss": 0.1248, "step": 4934 }, { "epoch": 0.2, "learning_rate": 9.35555674033002e-06, "loss": 0.1112, "step": 4935 }, { "epoch": 0.2, "learning_rate": 9.355423453202891e-06, "loss": 0.1393, "step": 4936 }, { "epoch": 0.2, "learning_rate": 9.35529016607576e-06, "loss": 0.0892, "step": 4937 }, { "epoch": 0.2, "learning_rate": 9.355156878948633e-06, "loss": 0.1073, "step": 4938 }, { "epoch": 0.2, "learning_rate": 9.355023591821502e-06, "loss": 0.1123, "step": 4939 }, { "epoch": 0.2, "learning_rate": 9.354890304694373e-06, "loss": 0.1461, "step": 4940 }, { "epoch": 0.2, "learning_rate": 9.354757017567244e-06, "loss": 0.1035, "step": 4941 }, { "epoch": 0.2, "learning_rate": 9.354623730440114e-06, "loss": 0.1475, "step": 4942 }, { "epoch": 0.2, "learning_rate": 9.354490443312986e-06, "loss": 0.1641, "step": 4943 }, { "epoch": 0.2, "learning_rate": 9.354357156185856e-06, "loss": 0.1175, "step": 4944 }, { "epoch": 0.2, "learning_rate": 9.354223869058727e-06, "loss": 0.1108, "step": 4945 }, { "epoch": 0.2, "learning_rate": 9.354090581931598e-06, "loss": 0.125, "step": 4946 }, { "epoch": 0.2, "learning_rate": 9.353957294804469e-06, "loss": 0.1254, "step": 4947 }, { "epoch": 0.2, "learning_rate": 9.35382400767734e-06, "loss": 0.146, "step": 4948 }, { "epoch": 0.2, "learning_rate": 9.35369072055021e-06, "loss": 0.128, "step": 4949 }, { "epoch": 0.2, "learning_rate": 9.35355743342308e-06, "loss": 0.0764, "step": 4950 }, { "epoch": 0.2, "learning_rate": 9.353424146295953e-06, "loss": 0.1078, "step": 4951 }, { "epoch": 0.2, "learning_rate": 9.353290859168822e-06, "loss": 0.1198, "step": 4952 }, { "epoch": 0.2, "learning_rate": 9.353157572041693e-06, "loss": 0.0837, "step": 4953 }, { "epoch": 0.2, "learning_rate": 9.353024284914564e-06, "loss": 0.113, "step": 4954 }, { "epoch": 0.2, "learning_rate": 9.352890997787435e-06, "loss": 0.122, "step": 4955 }, { "epoch": 0.2, "learning_rate": 9.352757710660306e-06, "loss": 0.1014, "step": 4956 }, { "epoch": 0.2, "learning_rate": 9.352624423533177e-06, "loss": 0.091, "step": 4957 }, { "epoch": 0.2, "learning_rate": 9.352491136406046e-06, "loss": 0.1292, "step": 4958 }, { "epoch": 0.2, "learning_rate": 9.352357849278917e-06, "loss": 0.1276, "step": 4959 }, { "epoch": 0.2, "learning_rate": 9.352224562151788e-06, "loss": 0.1273, "step": 4960 }, { "epoch": 0.2, "learning_rate": 9.35209127502466e-06, "loss": 0.1216, "step": 4961 }, { "epoch": 0.2, "learning_rate": 9.35195798789753e-06, "loss": 0.0942, "step": 4962 }, { "epoch": 0.2, "learning_rate": 9.3518247007704e-06, "loss": 0.1034, "step": 4963 }, { "epoch": 0.2, "learning_rate": 9.35169141364327e-06, "loss": 0.0712, "step": 4964 }, { "epoch": 0.2, "learning_rate": 9.351558126516142e-06, "loss": 0.1607, "step": 4965 }, { "epoch": 0.2, "learning_rate": 9.351424839389013e-06, "loss": 0.1217, "step": 4966 }, { "epoch": 0.2, "learning_rate": 9.351291552261884e-06, "loss": 0.1174, "step": 4967 }, { "epoch": 0.2, "learning_rate": 9.351158265134755e-06, "loss": 0.1061, "step": 4968 }, { "epoch": 0.2, "learning_rate": 9.351024978007624e-06, "loss": 0.1189, "step": 4969 }, { "epoch": 0.2, "learning_rate": 9.350891690880497e-06, "loss": 0.1091, "step": 4970 }, { "epoch": 0.2, "learning_rate": 9.350758403753366e-06, "loss": 0.067, "step": 4971 }, { "epoch": 0.2, "learning_rate": 9.350625116626237e-06, "loss": 0.1014, "step": 4972 }, { "epoch": 0.2, "learning_rate": 9.350491829499108e-06, "loss": 0.1013, "step": 4973 }, { "epoch": 0.2, "learning_rate": 9.350358542371979e-06, "loss": 0.1367, "step": 4974 }, { "epoch": 0.2, "learning_rate": 9.35022525524485e-06, "loss": 0.1649, "step": 4975 }, { "epoch": 0.2, "learning_rate": 9.35009196811772e-06, "loss": 0.099, "step": 4976 }, { "epoch": 0.2, "learning_rate": 9.34995868099059e-06, "loss": 0.0863, "step": 4977 }, { "epoch": 0.2, "learning_rate": 9.349825393863461e-06, "loss": 0.0991, "step": 4978 }, { "epoch": 0.2, "learning_rate": 9.349692106736332e-06, "loss": 0.1042, "step": 4979 }, { "epoch": 0.2, "learning_rate": 9.349558819609203e-06, "loss": 0.113, "step": 4980 }, { "epoch": 0.2, "learning_rate": 9.349425532482074e-06, "loss": 0.1287, "step": 4981 }, { "epoch": 0.2, "learning_rate": 9.349292245354943e-06, "loss": 0.1366, "step": 4982 }, { "epoch": 0.2, "learning_rate": 9.349158958227816e-06, "loss": 0.1065, "step": 4983 }, { "epoch": 0.2, "learning_rate": 9.349025671100685e-06, "loss": 0.1439, "step": 4984 }, { "epoch": 0.2, "learning_rate": 9.348892383973556e-06, "loss": 0.1249, "step": 4985 }, { "epoch": 0.2, "learning_rate": 9.348759096846427e-06, "loss": 0.1127, "step": 4986 }, { "epoch": 0.2, "learning_rate": 9.348625809719298e-06, "loss": 0.1062, "step": 4987 }, { "epoch": 0.2, "learning_rate": 9.34849252259217e-06, "loss": 0.1394, "step": 4988 }, { "epoch": 0.2, "learning_rate": 9.34835923546504e-06, "loss": 0.1157, "step": 4989 }, { "epoch": 0.2, "learning_rate": 9.34822594833791e-06, "loss": 0.0848, "step": 4990 }, { "epoch": 0.2, "learning_rate": 9.34809266121078e-06, "loss": 0.0827, "step": 4991 }, { "epoch": 0.2, "learning_rate": 9.347959374083652e-06, "loss": 0.136, "step": 4992 }, { "epoch": 0.2, "learning_rate": 9.347826086956523e-06, "loss": 0.0915, "step": 4993 }, { "epoch": 0.2, "learning_rate": 9.347692799829394e-06, "loss": 0.1347, "step": 4994 }, { "epoch": 0.2, "learning_rate": 9.347559512702265e-06, "loss": 0.0981, "step": 4995 }, { "epoch": 0.2, "learning_rate": 9.347426225575134e-06, "loss": 0.1071, "step": 4996 }, { "epoch": 0.2, "learning_rate": 9.347292938448007e-06, "loss": 0.0968, "step": 4997 }, { "epoch": 0.2, "learning_rate": 9.347159651320876e-06, "loss": 0.115, "step": 4998 }, { "epoch": 0.2, "learning_rate": 9.347026364193747e-06, "loss": 0.0835, "step": 4999 }, { "epoch": 0.2, "learning_rate": 9.346893077066618e-06, "loss": 0.1573, "step": 5000 }, { "epoch": 0.2, "learning_rate": 9.346759789939487e-06, "loss": 0.1129, "step": 5001 }, { "epoch": 0.2, "learning_rate": 9.34662650281236e-06, "loss": 0.1026, "step": 5002 }, { "epoch": 0.2, "learning_rate": 9.34649321568523e-06, "loss": 0.137, "step": 5003 }, { "epoch": 0.2, "learning_rate": 9.3463599285581e-06, "loss": 0.0749, "step": 5004 }, { "epoch": 0.2, "learning_rate": 9.346226641430971e-06, "loss": 0.0794, "step": 5005 }, { "epoch": 0.2, "learning_rate": 9.346093354303842e-06, "loss": 0.1302, "step": 5006 }, { "epoch": 0.2, "learning_rate": 9.345960067176713e-06, "loss": 0.1107, "step": 5007 }, { "epoch": 0.2, "learning_rate": 9.345826780049584e-06, "loss": 0.1324, "step": 5008 }, { "epoch": 0.2, "learning_rate": 9.345693492922453e-06, "loss": 0.1287, "step": 5009 }, { "epoch": 0.2, "learning_rate": 9.345560205795326e-06, "loss": 0.1852, "step": 5010 }, { "epoch": 0.2, "learning_rate": 9.345426918668195e-06, "loss": 0.0966, "step": 5011 }, { "epoch": 0.2, "learning_rate": 9.345293631541066e-06, "loss": 0.1636, "step": 5012 }, { "epoch": 0.2, "learning_rate": 9.345160344413937e-06, "loss": 0.0892, "step": 5013 }, { "epoch": 0.2, "learning_rate": 9.345027057286808e-06, "loss": 0.1456, "step": 5014 }, { "epoch": 0.2, "learning_rate": 9.34489377015968e-06, "loss": 0.0756, "step": 5015 }, { "epoch": 0.2, "learning_rate": 9.34476048303255e-06, "loss": 0.1161, "step": 5016 }, { "epoch": 0.2, "learning_rate": 9.34462719590542e-06, "loss": 0.1165, "step": 5017 }, { "epoch": 0.2, "learning_rate": 9.34449390877829e-06, "loss": 0.1325, "step": 5018 }, { "epoch": 0.2, "learning_rate": 9.344360621651162e-06, "loss": 0.1287, "step": 5019 }, { "epoch": 0.2, "learning_rate": 9.344227334524033e-06, "loss": 0.1284, "step": 5020 }, { "epoch": 0.2, "learning_rate": 9.344094047396904e-06, "loss": 0.0893, "step": 5021 }, { "epoch": 0.2, "learning_rate": 9.343960760269773e-06, "loss": 0.1359, "step": 5022 }, { "epoch": 0.2, "learning_rate": 9.343827473142644e-06, "loss": 0.0707, "step": 5023 }, { "epoch": 0.2, "learning_rate": 9.343694186015515e-06, "loss": 0.0762, "step": 5024 }, { "epoch": 0.2, "learning_rate": 9.343560898888386e-06, "loss": 0.1738, "step": 5025 }, { "epoch": 0.2, "learning_rate": 9.343427611761257e-06, "loss": 0.1011, "step": 5026 }, { "epoch": 0.2, "learning_rate": 9.343294324634128e-06, "loss": 0.122, "step": 5027 }, { "epoch": 0.2, "learning_rate": 9.343161037506997e-06, "loss": 0.1097, "step": 5028 }, { "epoch": 0.2, "learning_rate": 9.34302775037987e-06, "loss": 0.1004, "step": 5029 }, { "epoch": 0.2, "learning_rate": 9.34289446325274e-06, "loss": 0.143, "step": 5030 }, { "epoch": 0.2, "learning_rate": 9.34276117612561e-06, "loss": 0.1663, "step": 5031 }, { "epoch": 0.2, "learning_rate": 9.342627888998481e-06, "loss": 0.1166, "step": 5032 }, { "epoch": 0.2, "learning_rate": 9.342494601871352e-06, "loss": 0.152, "step": 5033 }, { "epoch": 0.2, "learning_rate": 9.342361314744223e-06, "loss": 0.0766, "step": 5034 }, { "epoch": 0.2, "learning_rate": 9.342228027617094e-06, "loss": 0.1065, "step": 5035 }, { "epoch": 0.2, "learning_rate": 9.342094740489964e-06, "loss": 0.0916, "step": 5036 }, { "epoch": 0.2, "learning_rate": 9.341961453362835e-06, "loss": 0.0961, "step": 5037 }, { "epoch": 0.2, "learning_rate": 9.341828166235706e-06, "loss": 0.0795, "step": 5038 }, { "epoch": 0.2, "learning_rate": 9.341694879108577e-06, "loss": 0.0962, "step": 5039 }, { "epoch": 0.2, "learning_rate": 9.341561591981448e-06, "loss": 0.084, "step": 5040 }, { "epoch": 0.2, "learning_rate": 9.341428304854317e-06, "loss": 0.0817, "step": 5041 }, { "epoch": 0.2, "learning_rate": 9.34129501772719e-06, "loss": 0.1109, "step": 5042 }, { "epoch": 0.2, "learning_rate": 9.341161730600059e-06, "loss": 0.1045, "step": 5043 }, { "epoch": 0.2, "learning_rate": 9.34102844347293e-06, "loss": 0.0795, "step": 5044 }, { "epoch": 0.2, "learning_rate": 9.3408951563458e-06, "loss": 0.1107, "step": 5045 }, { "epoch": 0.2, "learning_rate": 9.340761869218672e-06, "loss": 0.0899, "step": 5046 }, { "epoch": 0.2, "learning_rate": 9.340628582091543e-06, "loss": 0.1104, "step": 5047 }, { "epoch": 0.2, "learning_rate": 9.340495294964414e-06, "loss": 0.1016, "step": 5048 }, { "epoch": 0.2, "learning_rate": 9.340362007837283e-06, "loss": 0.0987, "step": 5049 }, { "epoch": 0.2, "learning_rate": 9.340228720710156e-06, "loss": 0.1366, "step": 5050 }, { "epoch": 0.2, "learning_rate": 9.340095433583025e-06, "loss": 0.1069, "step": 5051 }, { "epoch": 0.2, "learning_rate": 9.339962146455896e-06, "loss": 0.1257, "step": 5052 }, { "epoch": 0.2, "learning_rate": 9.339828859328767e-06, "loss": 0.1393, "step": 5053 }, { "epoch": 0.2, "learning_rate": 9.339695572201638e-06, "loss": 0.1147, "step": 5054 }, { "epoch": 0.2, "learning_rate": 9.339562285074509e-06, "loss": 0.0934, "step": 5055 }, { "epoch": 0.2, "learning_rate": 9.33942899794738e-06, "loss": 0.0748, "step": 5056 }, { "epoch": 0.2, "learning_rate": 9.33929571082025e-06, "loss": 0.1435, "step": 5057 }, { "epoch": 0.2, "learning_rate": 9.33916242369312e-06, "loss": 0.1001, "step": 5058 }, { "epoch": 0.2, "learning_rate": 9.339029136565991e-06, "loss": 0.0989, "step": 5059 }, { "epoch": 0.2, "learning_rate": 9.338895849438862e-06, "loss": 0.0975, "step": 5060 }, { "epoch": 0.2, "learning_rate": 9.338762562311733e-06, "loss": 0.0919, "step": 5061 }, { "epoch": 0.2, "learning_rate": 9.338629275184603e-06, "loss": 0.1622, "step": 5062 }, { "epoch": 0.2, "learning_rate": 9.338495988057474e-06, "loss": 0.102, "step": 5063 }, { "epoch": 0.2, "learning_rate": 9.338362700930345e-06, "loss": 0.0975, "step": 5064 }, { "epoch": 0.2, "learning_rate": 9.338229413803216e-06, "loss": 0.1013, "step": 5065 }, { "epoch": 0.2, "learning_rate": 9.338096126676087e-06, "loss": 0.132, "step": 5066 }, { "epoch": 0.2, "learning_rate": 9.337962839548958e-06, "loss": 0.123, "step": 5067 }, { "epoch": 0.2, "learning_rate": 9.337829552421827e-06, "loss": 0.0734, "step": 5068 }, { "epoch": 0.2, "learning_rate": 9.3376962652947e-06, "loss": 0.0911, "step": 5069 }, { "epoch": 0.2, "learning_rate": 9.337562978167569e-06, "loss": 0.1052, "step": 5070 }, { "epoch": 0.2, "learning_rate": 9.33742969104044e-06, "loss": 0.1263, "step": 5071 }, { "epoch": 0.2, "learning_rate": 9.337296403913311e-06, "loss": 0.1262, "step": 5072 }, { "epoch": 0.2, "learning_rate": 9.337163116786182e-06, "loss": 0.1319, "step": 5073 }, { "epoch": 0.2, "learning_rate": 9.337029829659053e-06, "loss": 0.0702, "step": 5074 }, { "epoch": 0.2, "learning_rate": 9.336896542531924e-06, "loss": 0.1196, "step": 5075 }, { "epoch": 0.2, "learning_rate": 9.336763255404793e-06, "loss": 0.1269, "step": 5076 }, { "epoch": 0.2, "learning_rate": 9.336629968277664e-06, "loss": 0.0898, "step": 5077 }, { "epoch": 0.2, "learning_rate": 9.336496681150535e-06, "loss": 0.1072, "step": 5078 }, { "epoch": 0.2, "learning_rate": 9.336363394023406e-06, "loss": 0.0991, "step": 5079 }, { "epoch": 0.2, "learning_rate": 9.336230106896277e-06, "loss": 0.1251, "step": 5080 }, { "epoch": 0.2, "learning_rate": 9.336096819769146e-06, "loss": 0.1166, "step": 5081 }, { "epoch": 0.2, "learning_rate": 9.335963532642019e-06, "loss": 0.0876, "step": 5082 }, { "epoch": 0.2, "learning_rate": 9.335830245514888e-06, "loss": 0.1023, "step": 5083 }, { "epoch": 0.2, "learning_rate": 9.33569695838776e-06, "loss": 0.0978, "step": 5084 }, { "epoch": 0.2, "learning_rate": 9.33556367126063e-06, "loss": 0.1287, "step": 5085 }, { "epoch": 0.2, "learning_rate": 9.335430384133501e-06, "loss": 0.1054, "step": 5086 }, { "epoch": 0.2, "learning_rate": 9.335297097006372e-06, "loss": 0.1168, "step": 5087 }, { "epoch": 0.2, "learning_rate": 9.335163809879243e-06, "loss": 0.109, "step": 5088 }, { "epoch": 0.2, "learning_rate": 9.335030522752113e-06, "loss": 0.1137, "step": 5089 }, { "epoch": 0.2, "learning_rate": 9.334897235624984e-06, "loss": 0.0884, "step": 5090 }, { "epoch": 0.2, "learning_rate": 9.334763948497855e-06, "loss": 0.1049, "step": 5091 }, { "epoch": 0.2, "learning_rate": 9.334630661370726e-06, "loss": 0.1442, "step": 5092 }, { "epoch": 0.2, "learning_rate": 9.334497374243597e-06, "loss": 0.1328, "step": 5093 }, { "epoch": 0.2, "learning_rate": 9.334364087116468e-06, "loss": 0.1223, "step": 5094 }, { "epoch": 0.2, "learning_rate": 9.334230799989337e-06, "loss": 0.1144, "step": 5095 }, { "epoch": 0.2, "learning_rate": 9.334097512862208e-06, "loss": 0.0989, "step": 5096 }, { "epoch": 0.2, "learning_rate": 9.333964225735079e-06, "loss": 0.1487, "step": 5097 }, { "epoch": 0.2, "learning_rate": 9.33383093860795e-06, "loss": 0.0961, "step": 5098 }, { "epoch": 0.2, "learning_rate": 9.333697651480821e-06, "loss": 0.0868, "step": 5099 }, { "epoch": 0.2, "learning_rate": 9.33356436435369e-06, "loss": 0.1094, "step": 5100 }, { "epoch": 0.2, "learning_rate": 9.333431077226563e-06, "loss": 0.1209, "step": 5101 }, { "epoch": 0.2, "learning_rate": 9.333297790099432e-06, "loss": 0.0771, "step": 5102 }, { "epoch": 0.2, "learning_rate": 9.333164502972303e-06, "loss": 0.1225, "step": 5103 }, { "epoch": 0.2, "learning_rate": 9.333031215845174e-06, "loss": 0.0997, "step": 5104 }, { "epoch": 0.2, "learning_rate": 9.332897928718045e-06, "loss": 0.1648, "step": 5105 }, { "epoch": 0.2, "learning_rate": 9.332764641590916e-06, "loss": 0.1077, "step": 5106 }, { "epoch": 0.2, "learning_rate": 9.332631354463787e-06, "loss": 0.1056, "step": 5107 }, { "epoch": 0.2, "learning_rate": 9.332498067336657e-06, "loss": 0.1385, "step": 5108 }, { "epoch": 0.2, "learning_rate": 9.33236478020953e-06, "loss": 0.0981, "step": 5109 }, { "epoch": 0.2, "learning_rate": 9.332231493082399e-06, "loss": 0.1297, "step": 5110 }, { "epoch": 0.2, "learning_rate": 9.33209820595527e-06, "loss": 0.1184, "step": 5111 }, { "epoch": 0.2, "learning_rate": 9.33196491882814e-06, "loss": 0.1046, "step": 5112 }, { "epoch": 0.2, "learning_rate": 9.331831631701011e-06, "loss": 0.1108, "step": 5113 }, { "epoch": 0.2, "learning_rate": 9.331698344573882e-06, "loss": 0.0764, "step": 5114 }, { "epoch": 0.2, "learning_rate": 9.331565057446752e-06, "loss": 0.0992, "step": 5115 }, { "epoch": 0.2, "learning_rate": 9.331431770319623e-06, "loss": 0.0941, "step": 5116 }, { "epoch": 0.2, "learning_rate": 9.331298483192494e-06, "loss": 0.1329, "step": 5117 }, { "epoch": 0.2, "learning_rate": 9.331165196065365e-06, "loss": 0.0899, "step": 5118 }, { "epoch": 0.2, "learning_rate": 9.331031908938236e-06, "loss": 0.1458, "step": 5119 }, { "epoch": 0.2, "learning_rate": 9.330898621811107e-06, "loss": 0.0868, "step": 5120 }, { "epoch": 0.2, "learning_rate": 9.330765334683976e-06, "loss": 0.1203, "step": 5121 }, { "epoch": 0.2, "learning_rate": 9.330632047556849e-06, "loss": 0.1172, "step": 5122 }, { "epoch": 0.2, "learning_rate": 9.330498760429718e-06, "loss": 0.1461, "step": 5123 }, { "epoch": 0.2, "learning_rate": 9.330365473302589e-06, "loss": 0.1092, "step": 5124 }, { "epoch": 0.2, "learning_rate": 9.33023218617546e-06, "loss": 0.1079, "step": 5125 }, { "epoch": 0.2, "learning_rate": 9.330098899048331e-06, "loss": 0.1022, "step": 5126 }, { "epoch": 0.2, "learning_rate": 9.329965611921202e-06, "loss": 0.112, "step": 5127 }, { "epoch": 0.2, "learning_rate": 9.329832324794073e-06, "loss": 0.0939, "step": 5128 }, { "epoch": 0.2, "learning_rate": 9.329699037666942e-06, "loss": 0.1132, "step": 5129 }, { "epoch": 0.2, "learning_rate": 9.329565750539813e-06, "loss": 0.0853, "step": 5130 }, { "epoch": 0.2, "learning_rate": 9.329432463412684e-06, "loss": 0.0884, "step": 5131 }, { "epoch": 0.2, "learning_rate": 9.329299176285555e-06, "loss": 0.0686, "step": 5132 }, { "epoch": 0.2, "learning_rate": 9.329165889158426e-06, "loss": 0.1175, "step": 5133 }, { "epoch": 0.21, "learning_rate": 9.329032602031297e-06, "loss": 0.1644, "step": 5134 }, { "epoch": 0.21, "learning_rate": 9.328899314904167e-06, "loss": 0.074, "step": 5135 }, { "epoch": 0.21, "learning_rate": 9.328766027777038e-06, "loss": 0.1481, "step": 5136 }, { "epoch": 0.21, "learning_rate": 9.328632740649909e-06, "loss": 0.1316, "step": 5137 }, { "epoch": 0.21, "learning_rate": 9.32849945352278e-06, "loss": 0.1279, "step": 5138 }, { "epoch": 0.21, "learning_rate": 9.32836616639565e-06, "loss": 0.1382, "step": 5139 }, { "epoch": 0.21, "learning_rate": 9.32823287926852e-06, "loss": 0.1091, "step": 5140 }, { "epoch": 0.21, "learning_rate": 9.328099592141393e-06, "loss": 0.1231, "step": 5141 }, { "epoch": 0.21, "learning_rate": 9.327966305014262e-06, "loss": 0.1023, "step": 5142 }, { "epoch": 0.21, "learning_rate": 9.327833017887133e-06, "loss": 0.1137, "step": 5143 }, { "epoch": 0.21, "learning_rate": 9.327699730760004e-06, "loss": 0.1356, "step": 5144 }, { "epoch": 0.21, "learning_rate": 9.327566443632875e-06, "loss": 0.0758, "step": 5145 }, { "epoch": 0.21, "learning_rate": 9.327433156505746e-06, "loss": 0.1122, "step": 5146 }, { "epoch": 0.21, "learning_rate": 9.327299869378617e-06, "loss": 0.1391, "step": 5147 }, { "epoch": 0.21, "learning_rate": 9.327166582251486e-06, "loss": 0.0722, "step": 5148 }, { "epoch": 0.21, "learning_rate": 9.327033295124359e-06, "loss": 0.1198, "step": 5149 }, { "epoch": 0.21, "learning_rate": 9.326900007997228e-06, "loss": 0.1054, "step": 5150 }, { "epoch": 0.21, "learning_rate": 9.326766720870099e-06, "loss": 0.1375, "step": 5151 }, { "epoch": 0.21, "learning_rate": 9.32663343374297e-06, "loss": 0.104, "step": 5152 }, { "epoch": 0.21, "learning_rate": 9.326500146615841e-06, "loss": 0.1427, "step": 5153 }, { "epoch": 0.21, "learning_rate": 9.326366859488712e-06, "loss": 0.1223, "step": 5154 }, { "epoch": 0.21, "learning_rate": 9.326233572361581e-06, "loss": 0.1188, "step": 5155 }, { "epoch": 0.21, "learning_rate": 9.326100285234452e-06, "loss": 0.1282, "step": 5156 }, { "epoch": 0.21, "learning_rate": 9.325966998107323e-06, "loss": 0.0833, "step": 5157 }, { "epoch": 0.21, "learning_rate": 9.325833710980194e-06, "loss": 0.1439, "step": 5158 }, { "epoch": 0.21, "learning_rate": 9.325700423853065e-06, "loss": 0.0972, "step": 5159 }, { "epoch": 0.21, "learning_rate": 9.325567136725936e-06, "loss": 0.1484, "step": 5160 }, { "epoch": 0.21, "learning_rate": 9.325433849598806e-06, "loss": 0.117, "step": 5161 }, { "epoch": 0.21, "learning_rate": 9.325300562471677e-06, "loss": 0.149, "step": 5162 }, { "epoch": 0.21, "learning_rate": 9.325167275344548e-06, "loss": 0.1205, "step": 5163 }, { "epoch": 0.21, "learning_rate": 9.325033988217419e-06, "loss": 0.1487, "step": 5164 }, { "epoch": 0.21, "learning_rate": 9.32490070109029e-06, "loss": 0.1012, "step": 5165 }, { "epoch": 0.21, "learning_rate": 9.32476741396316e-06, "loss": 0.1257, "step": 5166 }, { "epoch": 0.21, "learning_rate": 9.32463412683603e-06, "loss": 0.1391, "step": 5167 }, { "epoch": 0.21, "learning_rate": 9.324500839708903e-06, "loss": 0.1042, "step": 5168 }, { "epoch": 0.21, "learning_rate": 9.324367552581772e-06, "loss": 0.1076, "step": 5169 }, { "epoch": 0.21, "learning_rate": 9.324234265454643e-06, "loss": 0.1237, "step": 5170 }, { "epoch": 0.21, "learning_rate": 9.324100978327514e-06, "loss": 0.1333, "step": 5171 }, { "epoch": 0.21, "learning_rate": 9.323967691200385e-06, "loss": 0.105, "step": 5172 }, { "epoch": 0.21, "learning_rate": 9.323834404073256e-06, "loss": 0.0824, "step": 5173 }, { "epoch": 0.21, "learning_rate": 9.323701116946125e-06, "loss": 0.1414, "step": 5174 }, { "epoch": 0.21, "learning_rate": 9.323567829818996e-06, "loss": 0.1068, "step": 5175 }, { "epoch": 0.21, "learning_rate": 9.323434542691867e-06, "loss": 0.1203, "step": 5176 }, { "epoch": 0.21, "learning_rate": 9.323301255564738e-06, "loss": 0.1228, "step": 5177 }, { "epoch": 0.21, "learning_rate": 9.32316796843761e-06, "loss": 0.1249, "step": 5178 }, { "epoch": 0.21, "learning_rate": 9.32303468131048e-06, "loss": 0.1225, "step": 5179 }, { "epoch": 0.21, "learning_rate": 9.32290139418335e-06, "loss": 0.1546, "step": 5180 }, { "epoch": 0.21, "learning_rate": 9.322768107056222e-06, "loss": 0.1058, "step": 5181 }, { "epoch": 0.21, "learning_rate": 9.322634819929091e-06, "loss": 0.0918, "step": 5182 }, { "epoch": 0.21, "learning_rate": 9.322501532801962e-06, "loss": 0.0913, "step": 5183 }, { "epoch": 0.21, "learning_rate": 9.322368245674833e-06, "loss": 0.1363, "step": 5184 }, { "epoch": 0.21, "learning_rate": 9.322234958547704e-06, "loss": 0.1475, "step": 5185 }, { "epoch": 0.21, "learning_rate": 9.322101671420575e-06, "loss": 0.1081, "step": 5186 }, { "epoch": 0.21, "learning_rate": 9.321968384293446e-06, "loss": 0.09, "step": 5187 }, { "epoch": 0.21, "learning_rate": 9.321835097166316e-06, "loss": 0.1036, "step": 5188 }, { "epoch": 0.21, "learning_rate": 9.321701810039188e-06, "loss": 0.1036, "step": 5189 }, { "epoch": 0.21, "learning_rate": 9.321568522912058e-06, "loss": 0.097, "step": 5190 }, { "epoch": 0.21, "learning_rate": 9.321435235784929e-06, "loss": 0.0971, "step": 5191 }, { "epoch": 0.21, "learning_rate": 9.3213019486578e-06, "loss": 0.106, "step": 5192 }, { "epoch": 0.21, "learning_rate": 9.32116866153067e-06, "loss": 0.1103, "step": 5193 }, { "epoch": 0.21, "learning_rate": 9.321035374403542e-06, "loss": 0.0874, "step": 5194 }, { "epoch": 0.21, "learning_rate": 9.320902087276411e-06, "loss": 0.1009, "step": 5195 }, { "epoch": 0.21, "learning_rate": 9.320768800149282e-06, "loss": 0.0727, "step": 5196 }, { "epoch": 0.21, "learning_rate": 9.320635513022153e-06, "loss": 0.1041, "step": 5197 }, { "epoch": 0.21, "learning_rate": 9.320502225895024e-06, "loss": 0.1325, "step": 5198 }, { "epoch": 0.21, "learning_rate": 9.320368938767895e-06, "loss": 0.091, "step": 5199 }, { "epoch": 0.21, "learning_rate": 9.320235651640766e-06, "loss": 0.1282, "step": 5200 }, { "epoch": 0.21, "learning_rate": 9.320102364513635e-06, "loss": 0.0773, "step": 5201 }, { "epoch": 0.21, "learning_rate": 9.319969077386506e-06, "loss": 0.133, "step": 5202 }, { "epoch": 0.21, "learning_rate": 9.319835790259377e-06, "loss": 0.1183, "step": 5203 }, { "epoch": 0.21, "learning_rate": 9.319702503132248e-06, "loss": 0.1234, "step": 5204 }, { "epoch": 0.21, "learning_rate": 9.31956921600512e-06, "loss": 0.1165, "step": 5205 }, { "epoch": 0.21, "learning_rate": 9.31943592887799e-06, "loss": 0.0758, "step": 5206 }, { "epoch": 0.21, "learning_rate": 9.31930264175086e-06, "loss": 0.1093, "step": 5207 }, { "epoch": 0.21, "learning_rate": 9.319169354623732e-06, "loss": 0.1356, "step": 5208 }, { "epoch": 0.21, "learning_rate": 9.319036067496602e-06, "loss": 0.1191, "step": 5209 }, { "epoch": 0.21, "learning_rate": 9.318902780369473e-06, "loss": 0.1106, "step": 5210 }, { "epoch": 0.21, "learning_rate": 9.318769493242344e-06, "loss": 0.1074, "step": 5211 }, { "epoch": 0.21, "learning_rate": 9.318636206115215e-06, "loss": 0.1274, "step": 5212 }, { "epoch": 0.21, "learning_rate": 9.318502918988086e-06, "loss": 0.1074, "step": 5213 }, { "epoch": 0.21, "learning_rate": 9.318369631860955e-06, "loss": 0.1965, "step": 5214 }, { "epoch": 0.21, "learning_rate": 9.318236344733826e-06, "loss": 0.1157, "step": 5215 }, { "epoch": 0.21, "learning_rate": 9.318103057606697e-06, "loss": 0.0985, "step": 5216 }, { "epoch": 0.21, "learning_rate": 9.317969770479568e-06, "loss": 0.1127, "step": 5217 }, { "epoch": 0.21, "learning_rate": 9.317836483352439e-06, "loss": 0.1035, "step": 5218 }, { "epoch": 0.21, "learning_rate": 9.31770319622531e-06, "loss": 0.1377, "step": 5219 }, { "epoch": 0.21, "learning_rate": 9.317569909098179e-06, "loss": 0.1024, "step": 5220 }, { "epoch": 0.21, "learning_rate": 9.317436621971052e-06, "loss": 0.0902, "step": 5221 }, { "epoch": 0.21, "learning_rate": 9.317303334843921e-06, "loss": 0.1433, "step": 5222 }, { "epoch": 0.21, "learning_rate": 9.317170047716792e-06, "loss": 0.1295, "step": 5223 }, { "epoch": 0.21, "learning_rate": 9.317036760589663e-06, "loss": 0.1277, "step": 5224 }, { "epoch": 0.21, "learning_rate": 9.316903473462534e-06, "loss": 0.1175, "step": 5225 }, { "epoch": 0.21, "learning_rate": 9.316770186335405e-06, "loss": 0.1273, "step": 5226 }, { "epoch": 0.21, "learning_rate": 9.316636899208276e-06, "loss": 0.1694, "step": 5227 }, { "epoch": 0.21, "learning_rate": 9.316503612081145e-06, "loss": 0.1634, "step": 5228 }, { "epoch": 0.21, "learning_rate": 9.316370324954016e-06, "loss": 0.1273, "step": 5229 }, { "epoch": 0.21, "learning_rate": 9.316237037826887e-06, "loss": 0.1257, "step": 5230 }, { "epoch": 0.21, "learning_rate": 9.316103750699758e-06, "loss": 0.0989, "step": 5231 }, { "epoch": 0.21, "learning_rate": 9.31597046357263e-06, "loss": 0.1154, "step": 5232 }, { "epoch": 0.21, "learning_rate": 9.315837176445499e-06, "loss": 0.1203, "step": 5233 }, { "epoch": 0.21, "learning_rate": 9.31570388931837e-06, "loss": 0.1097, "step": 5234 }, { "epoch": 0.21, "learning_rate": 9.31557060219124e-06, "loss": 0.1072, "step": 5235 }, { "epoch": 0.21, "learning_rate": 9.315437315064112e-06, "loss": 0.1349, "step": 5236 }, { "epoch": 0.21, "learning_rate": 9.315304027936983e-06, "loss": 0.0858, "step": 5237 }, { "epoch": 0.21, "learning_rate": 9.315170740809854e-06, "loss": 0.1212, "step": 5238 }, { "epoch": 0.21, "learning_rate": 9.315037453682723e-06, "loss": 0.1462, "step": 5239 }, { "epoch": 0.21, "learning_rate": 9.314904166555596e-06, "loss": 0.1252, "step": 5240 }, { "epoch": 0.21, "learning_rate": 9.314770879428465e-06, "loss": 0.1032, "step": 5241 }, { "epoch": 0.21, "learning_rate": 9.314637592301336e-06, "loss": 0.1289, "step": 5242 }, { "epoch": 0.21, "learning_rate": 9.314504305174207e-06, "loss": 0.1084, "step": 5243 }, { "epoch": 0.21, "learning_rate": 9.314371018047078e-06, "loss": 0.0998, "step": 5244 }, { "epoch": 0.21, "learning_rate": 9.314237730919949e-06, "loss": 0.1087, "step": 5245 }, { "epoch": 0.21, "learning_rate": 9.31410444379282e-06, "loss": 0.1248, "step": 5246 }, { "epoch": 0.21, "learning_rate": 9.31397115666569e-06, "loss": 0.0901, "step": 5247 }, { "epoch": 0.21, "learning_rate": 9.313837869538562e-06, "loss": 0.1125, "step": 5248 }, { "epoch": 0.21, "learning_rate": 9.313704582411431e-06, "loss": 0.074, "step": 5249 }, { "epoch": 0.21, "learning_rate": 9.313571295284302e-06, "loss": 0.1473, "step": 5250 }, { "epoch": 0.21, "learning_rate": 9.313438008157173e-06, "loss": 0.0848, "step": 5251 }, { "epoch": 0.21, "learning_rate": 9.313304721030042e-06, "loss": 0.1641, "step": 5252 }, { "epoch": 0.21, "learning_rate": 9.313171433902915e-06, "loss": 0.0733, "step": 5253 }, { "epoch": 0.21, "learning_rate": 9.313038146775784e-06, "loss": 0.1829, "step": 5254 }, { "epoch": 0.21, "learning_rate": 9.312904859648655e-06, "loss": 0.1184, "step": 5255 }, { "epoch": 0.21, "learning_rate": 9.312771572521526e-06, "loss": 0.2204, "step": 5256 }, { "epoch": 0.21, "learning_rate": 9.312638285394397e-06, "loss": 0.1542, "step": 5257 }, { "epoch": 0.21, "learning_rate": 9.312504998267268e-06, "loss": 0.1417, "step": 5258 }, { "epoch": 0.21, "learning_rate": 9.31237171114014e-06, "loss": 0.1349, "step": 5259 }, { "epoch": 0.21, "learning_rate": 9.312238424013009e-06, "loss": 0.1003, "step": 5260 }, { "epoch": 0.21, "learning_rate": 9.31210513688588e-06, "loss": 0.1267, "step": 5261 }, { "epoch": 0.21, "learning_rate": 9.31197184975875e-06, "loss": 0.1123, "step": 5262 }, { "epoch": 0.21, "learning_rate": 9.311838562631622e-06, "loss": 0.1171, "step": 5263 }, { "epoch": 0.21, "learning_rate": 9.311705275504493e-06, "loss": 0.107, "step": 5264 }, { "epoch": 0.21, "learning_rate": 9.311571988377364e-06, "loss": 0.1388, "step": 5265 }, { "epoch": 0.21, "learning_rate": 9.311438701250233e-06, "loss": 0.1024, "step": 5266 }, { "epoch": 0.21, "learning_rate": 9.311305414123106e-06, "loss": 0.12, "step": 5267 }, { "epoch": 0.21, "learning_rate": 9.311172126995975e-06, "loss": 0.1124, "step": 5268 }, { "epoch": 0.21, "learning_rate": 9.311038839868846e-06, "loss": 0.1098, "step": 5269 }, { "epoch": 0.21, "learning_rate": 9.310905552741717e-06, "loss": 0.1, "step": 5270 }, { "epoch": 0.21, "learning_rate": 9.310772265614588e-06, "loss": 0.0893, "step": 5271 }, { "epoch": 0.21, "learning_rate": 9.310638978487459e-06, "loss": 0.1283, "step": 5272 }, { "epoch": 0.21, "learning_rate": 9.310505691360328e-06, "loss": 0.1265, "step": 5273 }, { "epoch": 0.21, "learning_rate": 9.3103724042332e-06, "loss": 0.1182, "step": 5274 }, { "epoch": 0.21, "learning_rate": 9.31023911710607e-06, "loss": 0.1086, "step": 5275 }, { "epoch": 0.21, "learning_rate": 9.310105829978941e-06, "loss": 0.1015, "step": 5276 }, { "epoch": 0.21, "learning_rate": 9.309972542851812e-06, "loss": 0.0857, "step": 5277 }, { "epoch": 0.21, "learning_rate": 9.309839255724683e-06, "loss": 0.1046, "step": 5278 }, { "epoch": 0.21, "learning_rate": 9.309705968597553e-06, "loss": 0.1188, "step": 5279 }, { "epoch": 0.21, "learning_rate": 9.309572681470425e-06, "loss": 0.0995, "step": 5280 }, { "epoch": 0.21, "learning_rate": 9.309439394343295e-06, "loss": 0.0979, "step": 5281 }, { "epoch": 0.21, "learning_rate": 9.309306107216166e-06, "loss": 0.084, "step": 5282 }, { "epoch": 0.21, "learning_rate": 9.309172820089037e-06, "loss": 0.0922, "step": 5283 }, { "epoch": 0.21, "learning_rate": 9.309039532961908e-06, "loss": 0.081, "step": 5284 }, { "epoch": 0.21, "learning_rate": 9.308906245834779e-06, "loss": 0.1049, "step": 5285 }, { "epoch": 0.21, "learning_rate": 9.30877295870765e-06, "loss": 0.1426, "step": 5286 }, { "epoch": 0.21, "learning_rate": 9.308639671580519e-06, "loss": 0.1027, "step": 5287 }, { "epoch": 0.21, "learning_rate": 9.308506384453392e-06, "loss": 0.1449, "step": 5288 }, { "epoch": 0.21, "learning_rate": 9.30837309732626e-06, "loss": 0.1017, "step": 5289 }, { "epoch": 0.21, "learning_rate": 9.308239810199132e-06, "loss": 0.1205, "step": 5290 }, { "epoch": 0.21, "learning_rate": 9.308106523072003e-06, "loss": 0.0758, "step": 5291 }, { "epoch": 0.21, "learning_rate": 9.307973235944872e-06, "loss": 0.079, "step": 5292 }, { "epoch": 0.21, "learning_rate": 9.307839948817745e-06, "loss": 0.0966, "step": 5293 }, { "epoch": 0.21, "learning_rate": 9.307706661690614e-06, "loss": 0.1039, "step": 5294 }, { "epoch": 0.21, "learning_rate": 9.307573374563485e-06, "loss": 0.1221, "step": 5295 }, { "epoch": 0.21, "learning_rate": 9.307440087436356e-06, "loss": 0.0984, "step": 5296 }, { "epoch": 0.21, "learning_rate": 9.307306800309227e-06, "loss": 0.1015, "step": 5297 }, { "epoch": 0.21, "learning_rate": 9.307173513182098e-06, "loss": 0.0788, "step": 5298 }, { "epoch": 0.21, "learning_rate": 9.307040226054969e-06, "loss": 0.0962, "step": 5299 }, { "epoch": 0.21, "learning_rate": 9.306906938927838e-06, "loss": 0.098, "step": 5300 }, { "epoch": 0.21, "learning_rate": 9.30677365180071e-06, "loss": 0.1352, "step": 5301 }, { "epoch": 0.21, "learning_rate": 9.30664036467358e-06, "loss": 0.1421, "step": 5302 }, { "epoch": 0.21, "learning_rate": 9.306507077546451e-06, "loss": 0.1107, "step": 5303 }, { "epoch": 0.21, "learning_rate": 9.306373790419322e-06, "loss": 0.1362, "step": 5304 }, { "epoch": 0.21, "learning_rate": 9.306240503292193e-06, "loss": 0.113, "step": 5305 }, { "epoch": 0.21, "learning_rate": 9.306107216165063e-06, "loss": 0.1105, "step": 5306 }, { "epoch": 0.21, "learning_rate": 9.305973929037935e-06, "loss": 0.1231, "step": 5307 }, { "epoch": 0.21, "learning_rate": 9.305840641910805e-06, "loss": 0.1451, "step": 5308 }, { "epoch": 0.21, "learning_rate": 9.305707354783676e-06, "loss": 0.1073, "step": 5309 }, { "epoch": 0.21, "learning_rate": 9.305574067656547e-06, "loss": 0.1162, "step": 5310 }, { "epoch": 0.21, "learning_rate": 9.305440780529416e-06, "loss": 0.1117, "step": 5311 }, { "epoch": 0.21, "learning_rate": 9.305307493402289e-06, "loss": 0.1378, "step": 5312 }, { "epoch": 0.21, "learning_rate": 9.305174206275158e-06, "loss": 0.1369, "step": 5313 }, { "epoch": 0.21, "learning_rate": 9.305040919148029e-06, "loss": 0.1259, "step": 5314 }, { "epoch": 0.21, "learning_rate": 9.3049076320209e-06, "loss": 0.0766, "step": 5315 }, { "epoch": 0.21, "learning_rate": 9.304774344893771e-06, "loss": 0.1422, "step": 5316 }, { "epoch": 0.21, "learning_rate": 9.304641057766642e-06, "loss": 0.1555, "step": 5317 }, { "epoch": 0.21, "learning_rate": 9.304507770639513e-06, "loss": 0.0998, "step": 5318 }, { "epoch": 0.21, "learning_rate": 9.304374483512382e-06, "loss": 0.0762, "step": 5319 }, { "epoch": 0.21, "learning_rate": 9.304241196385255e-06, "loss": 0.0861, "step": 5320 }, { "epoch": 0.21, "learning_rate": 9.304107909258124e-06, "loss": 0.1524, "step": 5321 }, { "epoch": 0.21, "learning_rate": 9.303974622130995e-06, "loss": 0.1141, "step": 5322 }, { "epoch": 0.21, "learning_rate": 9.303841335003866e-06, "loss": 0.1009, "step": 5323 }, { "epoch": 0.21, "learning_rate": 9.303708047876737e-06, "loss": 0.101, "step": 5324 }, { "epoch": 0.21, "learning_rate": 9.303574760749608e-06, "loss": 0.0878, "step": 5325 }, { "epoch": 0.21, "learning_rate": 9.303441473622479e-06, "loss": 0.128, "step": 5326 }, { "epoch": 0.21, "learning_rate": 9.303308186495348e-06, "loss": 0.107, "step": 5327 }, { "epoch": 0.21, "learning_rate": 9.30317489936822e-06, "loss": 0.0967, "step": 5328 }, { "epoch": 0.21, "learning_rate": 9.30304161224109e-06, "loss": 0.0953, "step": 5329 }, { "epoch": 0.21, "learning_rate": 9.302908325113961e-06, "loss": 0.0862, "step": 5330 }, { "epoch": 0.21, "learning_rate": 9.302775037986832e-06, "loss": 0.1502, "step": 5331 }, { "epoch": 0.21, "learning_rate": 9.302641750859702e-06, "loss": 0.1387, "step": 5332 }, { "epoch": 0.21, "learning_rate": 9.302508463732573e-06, "loss": 0.1028, "step": 5333 }, { "epoch": 0.21, "learning_rate": 9.302375176605444e-06, "loss": 0.1328, "step": 5334 }, { "epoch": 0.21, "learning_rate": 9.302241889478315e-06, "loss": 0.1246, "step": 5335 }, { "epoch": 0.21, "learning_rate": 9.302108602351186e-06, "loss": 0.1144, "step": 5336 }, { "epoch": 0.21, "learning_rate": 9.301975315224057e-06, "loss": 0.0929, "step": 5337 }, { "epoch": 0.21, "learning_rate": 9.301842028096926e-06, "loss": 0.1675, "step": 5338 }, { "epoch": 0.21, "learning_rate": 9.301708740969799e-06, "loss": 0.131, "step": 5339 }, { "epoch": 0.21, "learning_rate": 9.301575453842668e-06, "loss": 0.1102, "step": 5340 }, { "epoch": 0.21, "learning_rate": 9.301442166715539e-06, "loss": 0.1249, "step": 5341 }, { "epoch": 0.21, "learning_rate": 9.30130887958841e-06, "loss": 0.1227, "step": 5342 }, { "epoch": 0.21, "learning_rate": 9.301175592461281e-06, "loss": 0.0875, "step": 5343 }, { "epoch": 0.21, "learning_rate": 9.301042305334152e-06, "loss": 0.0938, "step": 5344 }, { "epoch": 0.21, "learning_rate": 9.300909018207023e-06, "loss": 0.1217, "step": 5345 }, { "epoch": 0.21, "learning_rate": 9.300775731079892e-06, "loss": 0.1516, "step": 5346 }, { "epoch": 0.21, "learning_rate": 9.300642443952765e-06, "loss": 0.1831, "step": 5347 }, { "epoch": 0.21, "learning_rate": 9.300509156825634e-06, "loss": 0.15, "step": 5348 }, { "epoch": 0.21, "learning_rate": 9.300375869698505e-06, "loss": 0.113, "step": 5349 }, { "epoch": 0.21, "learning_rate": 9.300242582571376e-06, "loss": 0.0778, "step": 5350 }, { "epoch": 0.21, "learning_rate": 9.300109295444246e-06, "loss": 0.1005, "step": 5351 }, { "epoch": 0.21, "learning_rate": 9.299976008317118e-06, "loss": 0.0897, "step": 5352 }, { "epoch": 0.21, "learning_rate": 9.299842721189988e-06, "loss": 0.1161, "step": 5353 }, { "epoch": 0.21, "learning_rate": 9.299709434062859e-06, "loss": 0.1298, "step": 5354 }, { "epoch": 0.21, "learning_rate": 9.29957614693573e-06, "loss": 0.1196, "step": 5355 }, { "epoch": 0.21, "learning_rate": 9.2994428598086e-06, "loss": 0.1003, "step": 5356 }, { "epoch": 0.21, "learning_rate": 9.299309572681472e-06, "loss": 0.1146, "step": 5357 }, { "epoch": 0.21, "learning_rate": 9.299176285554343e-06, "loss": 0.0878, "step": 5358 }, { "epoch": 0.21, "learning_rate": 9.299042998427212e-06, "loss": 0.1075, "step": 5359 }, { "epoch": 0.21, "learning_rate": 9.298909711300084e-06, "loss": 0.1379, "step": 5360 }, { "epoch": 0.21, "learning_rate": 9.298776424172954e-06, "loss": 0.1657, "step": 5361 }, { "epoch": 0.21, "learning_rate": 9.298643137045825e-06, "loss": 0.1049, "step": 5362 }, { "epoch": 0.21, "learning_rate": 9.298509849918696e-06, "loss": 0.1086, "step": 5363 }, { "epoch": 0.21, "learning_rate": 9.298376562791567e-06, "loss": 0.1199, "step": 5364 }, { "epoch": 0.21, "learning_rate": 9.298243275664438e-06, "loss": 0.1043, "step": 5365 }, { "epoch": 0.21, "learning_rate": 9.298109988537309e-06, "loss": 0.1019, "step": 5366 }, { "epoch": 0.21, "learning_rate": 9.297976701410178e-06, "loss": 0.1115, "step": 5367 }, { "epoch": 0.21, "learning_rate": 9.297843414283049e-06, "loss": 0.1156, "step": 5368 }, { "epoch": 0.21, "learning_rate": 9.29771012715592e-06, "loss": 0.0871, "step": 5369 }, { "epoch": 0.21, "learning_rate": 9.297576840028791e-06, "loss": 0.0916, "step": 5370 }, { "epoch": 0.21, "learning_rate": 9.297443552901662e-06, "loss": 0.1079, "step": 5371 }, { "epoch": 0.21, "learning_rate": 9.297310265774531e-06, "loss": 0.126, "step": 5372 }, { "epoch": 0.21, "learning_rate": 9.297176978647402e-06, "loss": 0.1554, "step": 5373 }, { "epoch": 0.21, "learning_rate": 9.297043691520273e-06, "loss": 0.1216, "step": 5374 }, { "epoch": 0.21, "learning_rate": 9.296910404393144e-06, "loss": 0.1044, "step": 5375 }, { "epoch": 0.21, "learning_rate": 9.296777117266015e-06, "loss": 0.1521, "step": 5376 }, { "epoch": 0.21, "learning_rate": 9.296643830138886e-06, "loss": 0.1059, "step": 5377 }, { "epoch": 0.21, "learning_rate": 9.296510543011756e-06, "loss": 0.0924, "step": 5378 }, { "epoch": 0.21, "learning_rate": 9.296377255884628e-06, "loss": 0.0881, "step": 5379 }, { "epoch": 0.21, "learning_rate": 9.296243968757498e-06, "loss": 0.0989, "step": 5380 }, { "epoch": 0.21, "learning_rate": 9.296110681630369e-06, "loss": 0.1382, "step": 5381 }, { "epoch": 0.21, "learning_rate": 9.29597739450324e-06, "loss": 0.0998, "step": 5382 }, { "epoch": 0.21, "learning_rate": 9.29584410737611e-06, "loss": 0.1231, "step": 5383 }, { "epoch": 0.21, "learning_rate": 9.295710820248982e-06, "loss": 0.0802, "step": 5384 }, { "epoch": 0.22, "learning_rate": 9.295577533121853e-06, "loss": 0.1045, "step": 5385 }, { "epoch": 0.22, "learning_rate": 9.295444245994722e-06, "loss": 0.1859, "step": 5386 }, { "epoch": 0.22, "learning_rate": 9.295310958867595e-06, "loss": 0.0958, "step": 5387 }, { "epoch": 0.22, "learning_rate": 9.295177671740464e-06, "loss": 0.1063, "step": 5388 }, { "epoch": 0.22, "learning_rate": 9.295044384613335e-06, "loss": 0.1203, "step": 5389 }, { "epoch": 0.22, "learning_rate": 9.294911097486206e-06, "loss": 0.112, "step": 5390 }, { "epoch": 0.22, "learning_rate": 9.294777810359075e-06, "loss": 0.1031, "step": 5391 }, { "epoch": 0.22, "learning_rate": 9.294644523231948e-06, "loss": 0.1411, "step": 5392 }, { "epoch": 0.22, "learning_rate": 9.294511236104817e-06, "loss": 0.0849, "step": 5393 }, { "epoch": 0.22, "learning_rate": 9.294377948977688e-06, "loss": 0.1411, "step": 5394 }, { "epoch": 0.22, "learning_rate": 9.294244661850559e-06, "loss": 0.0945, "step": 5395 }, { "epoch": 0.22, "learning_rate": 9.29411137472343e-06, "loss": 0.0817, "step": 5396 }, { "epoch": 0.22, "learning_rate": 9.293978087596301e-06, "loss": 0.086, "step": 5397 }, { "epoch": 0.22, "learning_rate": 9.293844800469172e-06, "loss": 0.1543, "step": 5398 }, { "epoch": 0.22, "learning_rate": 9.293711513342041e-06, "loss": 0.1196, "step": 5399 }, { "epoch": 0.22, "learning_rate": 9.293578226214912e-06, "loss": 0.1264, "step": 5400 }, { "epoch": 0.22, "learning_rate": 9.293444939087783e-06, "loss": 0.1254, "step": 5401 }, { "epoch": 0.22, "learning_rate": 9.293311651960654e-06, "loss": 0.0923, "step": 5402 }, { "epoch": 0.22, "learning_rate": 9.293178364833525e-06, "loss": 0.0998, "step": 5403 }, { "epoch": 0.22, "learning_rate": 9.293045077706396e-06, "loss": 0.1503, "step": 5404 }, { "epoch": 0.22, "learning_rate": 9.292911790579266e-06, "loss": 0.0875, "step": 5405 }, { "epoch": 0.22, "learning_rate": 9.292778503452138e-06, "loss": 0.0875, "step": 5406 }, { "epoch": 0.22, "learning_rate": 9.292645216325008e-06, "loss": 0.1281, "step": 5407 }, { "epoch": 0.22, "learning_rate": 9.292511929197879e-06, "loss": 0.1239, "step": 5408 }, { "epoch": 0.22, "learning_rate": 9.29237864207075e-06, "loss": 0.1236, "step": 5409 }, { "epoch": 0.22, "learning_rate": 9.292245354943619e-06, "loss": 0.0979, "step": 5410 }, { "epoch": 0.22, "learning_rate": 9.292112067816492e-06, "loss": 0.1426, "step": 5411 }, { "epoch": 0.22, "learning_rate": 9.291978780689361e-06, "loss": 0.111, "step": 5412 }, { "epoch": 0.22, "learning_rate": 9.291845493562232e-06, "loss": 0.1487, "step": 5413 }, { "epoch": 0.22, "learning_rate": 9.291712206435103e-06, "loss": 0.0848, "step": 5414 }, { "epoch": 0.22, "learning_rate": 9.291578919307974e-06, "loss": 0.1228, "step": 5415 }, { "epoch": 0.22, "learning_rate": 9.291445632180845e-06, "loss": 0.0923, "step": 5416 }, { "epoch": 0.22, "learning_rate": 9.291312345053716e-06, "loss": 0.163, "step": 5417 }, { "epoch": 0.22, "learning_rate": 9.291179057926585e-06, "loss": 0.1037, "step": 5418 }, { "epoch": 0.22, "learning_rate": 9.291045770799458e-06, "loss": 0.091, "step": 5419 }, { "epoch": 0.22, "learning_rate": 9.290912483672327e-06, "loss": 0.0997, "step": 5420 }, { "epoch": 0.22, "learning_rate": 9.290779196545198e-06, "loss": 0.1173, "step": 5421 }, { "epoch": 0.22, "learning_rate": 9.29064590941807e-06, "loss": 0.1291, "step": 5422 }, { "epoch": 0.22, "learning_rate": 9.29051262229094e-06, "loss": 0.084, "step": 5423 }, { "epoch": 0.22, "learning_rate": 9.290379335163811e-06, "loss": 0.0905, "step": 5424 }, { "epoch": 0.22, "learning_rate": 9.290246048036682e-06, "loss": 0.1156, "step": 5425 }, { "epoch": 0.22, "learning_rate": 9.290112760909552e-06, "loss": 0.1479, "step": 5426 }, { "epoch": 0.22, "learning_rate": 9.289979473782424e-06, "loss": 0.1047, "step": 5427 }, { "epoch": 0.22, "learning_rate": 9.289846186655294e-06, "loss": 0.0822, "step": 5428 }, { "epoch": 0.22, "learning_rate": 9.289712899528164e-06, "loss": 0.0804, "step": 5429 }, { "epoch": 0.22, "learning_rate": 9.289579612401035e-06, "loss": 0.1167, "step": 5430 }, { "epoch": 0.22, "learning_rate": 9.289446325273905e-06, "loss": 0.124, "step": 5431 }, { "epoch": 0.22, "learning_rate": 9.289313038146777e-06, "loss": 0.1133, "step": 5432 }, { "epoch": 0.22, "learning_rate": 9.289179751019647e-06, "loss": 0.1087, "step": 5433 }, { "epoch": 0.22, "learning_rate": 9.289046463892518e-06, "loss": 0.0906, "step": 5434 }, { "epoch": 0.22, "learning_rate": 9.288913176765389e-06, "loss": 0.1266, "step": 5435 }, { "epoch": 0.22, "learning_rate": 9.28877988963826e-06, "loss": 0.106, "step": 5436 }, { "epoch": 0.22, "learning_rate": 9.28864660251113e-06, "loss": 0.0825, "step": 5437 }, { "epoch": 0.22, "learning_rate": 9.288513315384002e-06, "loss": 0.1275, "step": 5438 }, { "epoch": 0.22, "learning_rate": 9.288380028256871e-06, "loss": 0.1666, "step": 5439 }, { "epoch": 0.22, "learning_rate": 9.288246741129742e-06, "loss": 0.095, "step": 5440 }, { "epoch": 0.22, "learning_rate": 9.288113454002613e-06, "loss": 0.0894, "step": 5441 }, { "epoch": 0.22, "learning_rate": 9.287980166875484e-06, "loss": 0.1197, "step": 5442 }, { "epoch": 0.22, "learning_rate": 9.287846879748355e-06, "loss": 0.1115, "step": 5443 }, { "epoch": 0.22, "learning_rate": 9.287713592621226e-06, "loss": 0.0983, "step": 5444 }, { "epoch": 0.22, "learning_rate": 9.287580305494095e-06, "loss": 0.1283, "step": 5445 }, { "epoch": 0.22, "learning_rate": 9.287447018366968e-06, "loss": 0.0999, "step": 5446 }, { "epoch": 0.22, "learning_rate": 9.287313731239837e-06, "loss": 0.1098, "step": 5447 }, { "epoch": 0.22, "learning_rate": 9.287180444112708e-06, "loss": 0.1133, "step": 5448 }, { "epoch": 0.22, "learning_rate": 9.28704715698558e-06, "loss": 0.1093, "step": 5449 }, { "epoch": 0.22, "learning_rate": 9.286913869858449e-06, "loss": 0.1453, "step": 5450 }, { "epoch": 0.22, "learning_rate": 9.286780582731321e-06, "loss": 0.0864, "step": 5451 }, { "epoch": 0.22, "learning_rate": 9.28664729560419e-06, "loss": 0.0909, "step": 5452 }, { "epoch": 0.22, "learning_rate": 9.286514008477062e-06, "loss": 0.0826, "step": 5453 }, { "epoch": 0.22, "learning_rate": 9.286380721349933e-06, "loss": 0.158, "step": 5454 }, { "epoch": 0.22, "learning_rate": 9.286247434222804e-06, "loss": 0.0991, "step": 5455 }, { "epoch": 0.22, "learning_rate": 9.286114147095675e-06, "loss": 0.1262, "step": 5456 }, { "epoch": 0.22, "learning_rate": 9.285980859968546e-06, "loss": 0.1036, "step": 5457 }, { "epoch": 0.22, "learning_rate": 9.285847572841415e-06, "loss": 0.0858, "step": 5458 }, { "epoch": 0.22, "learning_rate": 9.285714285714288e-06, "loss": 0.1193, "step": 5459 }, { "epoch": 0.22, "learning_rate": 9.285580998587157e-06, "loss": 0.0958, "step": 5460 }, { "epoch": 0.22, "learning_rate": 9.285447711460028e-06, "loss": 0.1403, "step": 5461 }, { "epoch": 0.22, "learning_rate": 9.285314424332899e-06, "loss": 0.0819, "step": 5462 }, { "epoch": 0.22, "learning_rate": 9.28518113720577e-06, "loss": 0.1274, "step": 5463 }, { "epoch": 0.22, "learning_rate": 9.28504785007864e-06, "loss": 0.1174, "step": 5464 }, { "epoch": 0.22, "learning_rate": 9.284914562951512e-06, "loss": 0.1233, "step": 5465 }, { "epoch": 0.22, "learning_rate": 9.284781275824381e-06, "loss": 0.0783, "step": 5466 }, { "epoch": 0.22, "learning_rate": 9.284647988697252e-06, "loss": 0.0633, "step": 5467 }, { "epoch": 0.22, "learning_rate": 9.284514701570123e-06, "loss": 0.1143, "step": 5468 }, { "epoch": 0.22, "learning_rate": 9.284381414442994e-06, "loss": 0.1033, "step": 5469 }, { "epoch": 0.22, "learning_rate": 9.284248127315865e-06, "loss": 0.1184, "step": 5470 }, { "epoch": 0.22, "learning_rate": 9.284114840188734e-06, "loss": 0.0981, "step": 5471 }, { "epoch": 0.22, "learning_rate": 9.283981553061605e-06, "loss": 0.1058, "step": 5472 }, { "epoch": 0.22, "learning_rate": 9.283848265934476e-06, "loss": 0.0868, "step": 5473 }, { "epoch": 0.22, "learning_rate": 9.283714978807347e-06, "loss": 0.1099, "step": 5474 }, { "epoch": 0.22, "learning_rate": 9.283581691680218e-06, "loss": 0.1092, "step": 5475 }, { "epoch": 0.22, "learning_rate": 9.28344840455309e-06, "loss": 0.1356, "step": 5476 }, { "epoch": 0.22, "learning_rate": 9.283315117425959e-06, "loss": 0.1262, "step": 5477 }, { "epoch": 0.22, "learning_rate": 9.283181830298831e-06, "loss": 0.1197, "step": 5478 }, { "epoch": 0.22, "learning_rate": 9.2830485431717e-06, "loss": 0.1016, "step": 5479 }, { "epoch": 0.22, "learning_rate": 9.282915256044572e-06, "loss": 0.1168, "step": 5480 }, { "epoch": 0.22, "learning_rate": 9.282781968917443e-06, "loss": 0.102, "step": 5481 }, { "epoch": 0.22, "learning_rate": 9.282648681790314e-06, "loss": 0.1014, "step": 5482 }, { "epoch": 0.22, "learning_rate": 9.282515394663185e-06, "loss": 0.1123, "step": 5483 }, { "epoch": 0.22, "learning_rate": 9.282382107536056e-06, "loss": 0.108, "step": 5484 }, { "epoch": 0.22, "learning_rate": 9.282248820408925e-06, "loss": 0.1239, "step": 5485 }, { "epoch": 0.22, "learning_rate": 9.282115533281798e-06, "loss": 0.0947, "step": 5486 }, { "epoch": 0.22, "learning_rate": 9.281982246154667e-06, "loss": 0.1189, "step": 5487 }, { "epoch": 0.22, "learning_rate": 9.281848959027538e-06, "loss": 0.1495, "step": 5488 }, { "epoch": 0.22, "learning_rate": 9.281715671900409e-06, "loss": 0.1192, "step": 5489 }, { "epoch": 0.22, "learning_rate": 9.281582384773278e-06, "loss": 0.1136, "step": 5490 }, { "epoch": 0.22, "learning_rate": 9.281449097646151e-06, "loss": 0.0815, "step": 5491 }, { "epoch": 0.22, "learning_rate": 9.28131581051902e-06, "loss": 0.0939, "step": 5492 }, { "epoch": 0.22, "learning_rate": 9.281182523391891e-06, "loss": 0.0981, "step": 5493 }, { "epoch": 0.22, "learning_rate": 9.281049236264762e-06, "loss": 0.1385, "step": 5494 }, { "epoch": 0.22, "learning_rate": 9.280915949137633e-06, "loss": 0.0746, "step": 5495 }, { "epoch": 0.22, "learning_rate": 9.280782662010504e-06, "loss": 0.0984, "step": 5496 }, { "epoch": 0.22, "learning_rate": 9.280649374883375e-06, "loss": 0.1157, "step": 5497 }, { "epoch": 0.22, "learning_rate": 9.280516087756244e-06, "loss": 0.0979, "step": 5498 }, { "epoch": 0.22, "learning_rate": 9.280382800629115e-06, "loss": 0.097, "step": 5499 }, { "epoch": 0.22, "learning_rate": 9.280249513501986e-06, "loss": 0.096, "step": 5500 }, { "epoch": 0.22, "learning_rate": 9.280116226374857e-06, "loss": 0.0754, "step": 5501 }, { "epoch": 0.22, "learning_rate": 9.279982939247728e-06, "loss": 0.103, "step": 5502 }, { "epoch": 0.22, "learning_rate": 9.2798496521206e-06, "loss": 0.1133, "step": 5503 }, { "epoch": 0.22, "learning_rate": 9.279716364993469e-06, "loss": 0.1351, "step": 5504 }, { "epoch": 0.22, "learning_rate": 9.279583077866341e-06, "loss": 0.1298, "step": 5505 }, { "epoch": 0.22, "learning_rate": 9.27944979073921e-06, "loss": 0.0887, "step": 5506 }, { "epoch": 0.22, "learning_rate": 9.279316503612082e-06, "loss": 0.1069, "step": 5507 }, { "epoch": 0.22, "learning_rate": 9.279183216484953e-06, "loss": 0.1081, "step": 5508 }, { "epoch": 0.22, "learning_rate": 9.279049929357822e-06, "loss": 0.1241, "step": 5509 }, { "epoch": 0.22, "learning_rate": 9.278916642230695e-06, "loss": 0.0963, "step": 5510 }, { "epoch": 0.22, "learning_rate": 9.278783355103564e-06, "loss": 0.1403, "step": 5511 }, { "epoch": 0.22, "learning_rate": 9.278650067976435e-06, "loss": 0.1072, "step": 5512 }, { "epoch": 0.22, "learning_rate": 9.278516780849306e-06, "loss": 0.117, "step": 5513 }, { "epoch": 0.22, "learning_rate": 9.278383493722177e-06, "loss": 0.1021, "step": 5514 }, { "epoch": 0.22, "learning_rate": 9.278250206595048e-06, "loss": 0.1041, "step": 5515 }, { "epoch": 0.22, "learning_rate": 9.278116919467919e-06, "loss": 0.1134, "step": 5516 }, { "epoch": 0.22, "learning_rate": 9.277983632340788e-06, "loss": 0.1029, "step": 5517 }, { "epoch": 0.22, "learning_rate": 9.277850345213661e-06, "loss": 0.1526, "step": 5518 }, { "epoch": 0.22, "learning_rate": 9.27771705808653e-06, "loss": 0.0935, "step": 5519 }, { "epoch": 0.22, "learning_rate": 9.277583770959401e-06, "loss": 0.1166, "step": 5520 }, { "epoch": 0.22, "learning_rate": 9.277450483832272e-06, "loss": 0.0974, "step": 5521 }, { "epoch": 0.22, "learning_rate": 9.277317196705143e-06, "loss": 0.1032, "step": 5522 }, { "epoch": 0.22, "learning_rate": 9.277183909578014e-06, "loss": 0.1077, "step": 5523 }, { "epoch": 0.22, "learning_rate": 9.277050622450885e-06, "loss": 0.0762, "step": 5524 }, { "epoch": 0.22, "learning_rate": 9.276917335323755e-06, "loss": 0.1317, "step": 5525 }, { "epoch": 0.22, "learning_rate": 9.276784048196627e-06, "loss": 0.0869, "step": 5526 }, { "epoch": 0.22, "learning_rate": 9.276650761069497e-06, "loss": 0.0873, "step": 5527 }, { "epoch": 0.22, "learning_rate": 9.276517473942368e-06, "loss": 0.1137, "step": 5528 }, { "epoch": 0.22, "learning_rate": 9.276384186815239e-06, "loss": 0.0961, "step": 5529 }, { "epoch": 0.22, "learning_rate": 9.276250899688108e-06, "loss": 0.1319, "step": 5530 }, { "epoch": 0.22, "learning_rate": 9.27611761256098e-06, "loss": 0.1045, "step": 5531 }, { "epoch": 0.22, "learning_rate": 9.27598432543385e-06, "loss": 0.0813, "step": 5532 }, { "epoch": 0.22, "learning_rate": 9.27585103830672e-06, "loss": 0.1193, "step": 5533 }, { "epoch": 0.22, "learning_rate": 9.275717751179592e-06, "loss": 0.1139, "step": 5534 }, { "epoch": 0.22, "learning_rate": 9.275584464052463e-06, "loss": 0.1238, "step": 5535 }, { "epoch": 0.22, "learning_rate": 9.275451176925334e-06, "loss": 0.1175, "step": 5536 }, { "epoch": 0.22, "learning_rate": 9.275317889798205e-06, "loss": 0.1038, "step": 5537 }, { "epoch": 0.22, "learning_rate": 9.275184602671074e-06, "loss": 0.0991, "step": 5538 }, { "epoch": 0.22, "learning_rate": 9.275051315543945e-06, "loss": 0.1172, "step": 5539 }, { "epoch": 0.22, "learning_rate": 9.274918028416816e-06, "loss": 0.1367, "step": 5540 }, { "epoch": 0.22, "learning_rate": 9.274784741289687e-06, "loss": 0.1647, "step": 5541 }, { "epoch": 0.22, "learning_rate": 9.274651454162558e-06, "loss": 0.1176, "step": 5542 }, { "epoch": 0.22, "learning_rate": 9.274518167035429e-06, "loss": 0.0978, "step": 5543 }, { "epoch": 0.22, "learning_rate": 9.274384879908298e-06, "loss": 0.124, "step": 5544 }, { "epoch": 0.22, "learning_rate": 9.274251592781171e-06, "loss": 0.1001, "step": 5545 }, { "epoch": 0.22, "learning_rate": 9.27411830565404e-06, "loss": 0.1039, "step": 5546 }, { "epoch": 0.22, "learning_rate": 9.273985018526911e-06, "loss": 0.1051, "step": 5547 }, { "epoch": 0.22, "learning_rate": 9.273851731399782e-06, "loss": 0.073, "step": 5548 }, { "epoch": 0.22, "learning_rate": 9.273718444272652e-06, "loss": 0.0687, "step": 5549 }, { "epoch": 0.22, "learning_rate": 9.273585157145524e-06, "loss": 0.113, "step": 5550 }, { "epoch": 0.22, "learning_rate": 9.273451870018394e-06, "loss": 0.0649, "step": 5551 }, { "epoch": 0.22, "learning_rate": 9.273318582891265e-06, "loss": 0.0866, "step": 5552 }, { "epoch": 0.22, "learning_rate": 9.273185295764136e-06, "loss": 0.1137, "step": 5553 }, { "epoch": 0.22, "learning_rate": 9.273052008637007e-06, "loss": 0.1302, "step": 5554 }, { "epoch": 0.22, "learning_rate": 9.272918721509878e-06, "loss": 0.1533, "step": 5555 }, { "epoch": 0.22, "learning_rate": 9.272785434382749e-06, "loss": 0.117, "step": 5556 }, { "epoch": 0.22, "learning_rate": 9.272652147255618e-06, "loss": 0.0947, "step": 5557 }, { "epoch": 0.22, "learning_rate": 9.27251886012849e-06, "loss": 0.1051, "step": 5558 }, { "epoch": 0.22, "learning_rate": 9.27238557300136e-06, "loss": 0.1247, "step": 5559 }, { "epoch": 0.22, "learning_rate": 9.272252285874231e-06, "loss": 0.099, "step": 5560 }, { "epoch": 0.22, "learning_rate": 9.272118998747102e-06, "loss": 0.1332, "step": 5561 }, { "epoch": 0.22, "learning_rate": 9.271985711619973e-06, "loss": 0.1007, "step": 5562 }, { "epoch": 0.22, "learning_rate": 9.271852424492844e-06, "loss": 0.0819, "step": 5563 }, { "epoch": 0.22, "learning_rate": 9.271719137365715e-06, "loss": 0.0907, "step": 5564 }, { "epoch": 0.22, "learning_rate": 9.271585850238584e-06, "loss": 0.0598, "step": 5565 }, { "epoch": 0.22, "learning_rate": 9.271452563111455e-06, "loss": 0.1303, "step": 5566 }, { "epoch": 0.22, "learning_rate": 9.271319275984326e-06, "loss": 0.1484, "step": 5567 }, { "epoch": 0.22, "learning_rate": 9.271185988857197e-06, "loss": 0.1006, "step": 5568 }, { "epoch": 0.22, "learning_rate": 9.271052701730068e-06, "loss": 0.1154, "step": 5569 }, { "epoch": 0.22, "learning_rate": 9.270919414602937e-06, "loss": 0.0935, "step": 5570 }, { "epoch": 0.22, "learning_rate": 9.270786127475808e-06, "loss": 0.1282, "step": 5571 }, { "epoch": 0.22, "learning_rate": 9.27065284034868e-06, "loss": 0.0781, "step": 5572 }, { "epoch": 0.22, "learning_rate": 9.27051955322155e-06, "loss": 0.0953, "step": 5573 }, { "epoch": 0.22, "learning_rate": 9.270386266094421e-06, "loss": 0.1062, "step": 5574 }, { "epoch": 0.22, "learning_rate": 9.270252978967292e-06, "loss": 0.0788, "step": 5575 }, { "epoch": 0.22, "learning_rate": 9.270119691840162e-06, "loss": 0.1012, "step": 5576 }, { "epoch": 0.22, "learning_rate": 9.269986404713034e-06, "loss": 0.1299, "step": 5577 }, { "epoch": 0.22, "learning_rate": 9.269853117585904e-06, "loss": 0.1373, "step": 5578 }, { "epoch": 0.22, "learning_rate": 9.269719830458775e-06, "loss": 0.0964, "step": 5579 }, { "epoch": 0.22, "learning_rate": 9.269586543331646e-06, "loss": 0.0946, "step": 5580 }, { "epoch": 0.22, "learning_rate": 9.269453256204517e-06, "loss": 0.1293, "step": 5581 }, { "epoch": 0.22, "learning_rate": 9.269319969077388e-06, "loss": 0.1184, "step": 5582 }, { "epoch": 0.22, "learning_rate": 9.269186681950259e-06, "loss": 0.1161, "step": 5583 }, { "epoch": 0.22, "learning_rate": 9.269053394823128e-06, "loss": 0.1479, "step": 5584 }, { "epoch": 0.22, "learning_rate": 9.268920107695999e-06, "loss": 0.0937, "step": 5585 }, { "epoch": 0.22, "learning_rate": 9.26878682056887e-06, "loss": 0.1177, "step": 5586 }, { "epoch": 0.22, "learning_rate": 9.268653533441741e-06, "loss": 0.1166, "step": 5587 }, { "epoch": 0.22, "learning_rate": 9.268520246314612e-06, "loss": 0.0903, "step": 5588 }, { "epoch": 0.22, "learning_rate": 9.268386959187481e-06, "loss": 0.0955, "step": 5589 }, { "epoch": 0.22, "learning_rate": 9.268253672060354e-06, "loss": 0.1016, "step": 5590 }, { "epoch": 0.22, "learning_rate": 9.268120384933223e-06, "loss": 0.0934, "step": 5591 }, { "epoch": 0.22, "learning_rate": 9.267987097806094e-06, "loss": 0.0893, "step": 5592 }, { "epoch": 0.22, "learning_rate": 9.267853810678965e-06, "loss": 0.1122, "step": 5593 }, { "epoch": 0.22, "learning_rate": 9.267720523551836e-06, "loss": 0.1117, "step": 5594 }, { "epoch": 0.22, "learning_rate": 9.267587236424707e-06, "loss": 0.0965, "step": 5595 }, { "epoch": 0.22, "learning_rate": 9.267453949297578e-06, "loss": 0.1288, "step": 5596 }, { "epoch": 0.22, "learning_rate": 9.267320662170448e-06, "loss": 0.1026, "step": 5597 }, { "epoch": 0.22, "learning_rate": 9.26718737504332e-06, "loss": 0.0852, "step": 5598 }, { "epoch": 0.22, "learning_rate": 9.26705408791619e-06, "loss": 0.0993, "step": 5599 }, { "epoch": 0.22, "learning_rate": 9.26692080078906e-06, "loss": 0.1094, "step": 5600 }, { "epoch": 0.22, "learning_rate": 9.266787513661932e-06, "loss": 0.1063, "step": 5601 }, { "epoch": 0.22, "learning_rate": 9.266654226534803e-06, "loss": 0.121, "step": 5602 }, { "epoch": 0.22, "learning_rate": 9.266520939407674e-06, "loss": 0.0856, "step": 5603 }, { "epoch": 0.22, "learning_rate": 9.266387652280545e-06, "loss": 0.0979, "step": 5604 }, { "epoch": 0.22, "learning_rate": 9.266254365153414e-06, "loss": 0.1356, "step": 5605 }, { "epoch": 0.22, "learning_rate": 9.266121078026285e-06, "loss": 0.1102, "step": 5606 }, { "epoch": 0.22, "learning_rate": 9.265987790899156e-06, "loss": 0.0729, "step": 5607 }, { "epoch": 0.22, "learning_rate": 9.265854503772027e-06, "loss": 0.1111, "step": 5608 }, { "epoch": 0.22, "learning_rate": 9.265721216644898e-06, "loss": 0.1296, "step": 5609 }, { "epoch": 0.22, "learning_rate": 9.265587929517767e-06, "loss": 0.1279, "step": 5610 }, { "epoch": 0.22, "learning_rate": 9.265454642390638e-06, "loss": 0.1052, "step": 5611 }, { "epoch": 0.22, "learning_rate": 9.265321355263509e-06, "loss": 0.1348, "step": 5612 }, { "epoch": 0.22, "learning_rate": 9.26518806813638e-06, "loss": 0.1241, "step": 5613 }, { "epoch": 0.22, "learning_rate": 9.265054781009251e-06, "loss": 0.1203, "step": 5614 }, { "epoch": 0.22, "learning_rate": 9.264921493882122e-06, "loss": 0.1029, "step": 5615 }, { "epoch": 0.22, "learning_rate": 9.264788206754991e-06, "loss": 0.0818, "step": 5616 }, { "epoch": 0.22, "learning_rate": 9.264654919627864e-06, "loss": 0.1043, "step": 5617 }, { "epoch": 0.22, "learning_rate": 9.264521632500733e-06, "loss": 0.0894, "step": 5618 }, { "epoch": 0.22, "learning_rate": 9.264388345373604e-06, "loss": 0.1609, "step": 5619 }, { "epoch": 0.22, "learning_rate": 9.264255058246475e-06, "loss": 0.0869, "step": 5620 }, { "epoch": 0.22, "learning_rate": 9.264121771119346e-06, "loss": 0.1236, "step": 5621 }, { "epoch": 0.22, "learning_rate": 9.263988483992217e-06, "loss": 0.1573, "step": 5622 }, { "epoch": 0.22, "learning_rate": 9.263855196865088e-06, "loss": 0.097, "step": 5623 }, { "epoch": 0.22, "learning_rate": 9.263721909737958e-06, "loss": 0.1626, "step": 5624 }, { "epoch": 0.22, "learning_rate": 9.263588622610829e-06, "loss": 0.104, "step": 5625 }, { "epoch": 0.22, "learning_rate": 9.2634553354837e-06, "loss": 0.094, "step": 5626 }, { "epoch": 0.22, "learning_rate": 9.26332204835657e-06, "loss": 0.0926, "step": 5627 }, { "epoch": 0.22, "learning_rate": 9.263188761229442e-06, "loss": 0.1187, "step": 5628 }, { "epoch": 0.22, "learning_rate": 9.263055474102311e-06, "loss": 0.117, "step": 5629 }, { "epoch": 0.22, "learning_rate": 9.262922186975184e-06, "loss": 0.141, "step": 5630 }, { "epoch": 0.22, "learning_rate": 9.262788899848053e-06, "loss": 0.0906, "step": 5631 }, { "epoch": 0.22, "learning_rate": 9.262655612720924e-06, "loss": 0.1083, "step": 5632 }, { "epoch": 0.22, "learning_rate": 9.262522325593795e-06, "loss": 0.1154, "step": 5633 }, { "epoch": 0.22, "learning_rate": 9.262389038466666e-06, "loss": 0.122, "step": 5634 }, { "epoch": 0.23, "learning_rate": 9.262255751339537e-06, "loss": 0.1477, "step": 5635 }, { "epoch": 0.23, "learning_rate": 9.262122464212408e-06, "loss": 0.1159, "step": 5636 }, { "epoch": 0.23, "learning_rate": 9.261989177085277e-06, "loss": 0.1052, "step": 5637 }, { "epoch": 0.23, "learning_rate": 9.261855889958148e-06, "loss": 0.1252, "step": 5638 }, { "epoch": 0.23, "learning_rate": 9.26172260283102e-06, "loss": 0.1058, "step": 5639 }, { "epoch": 0.23, "learning_rate": 9.26158931570389e-06, "loss": 0.1206, "step": 5640 }, { "epoch": 0.23, "learning_rate": 9.261456028576761e-06, "loss": 0.0673, "step": 5641 }, { "epoch": 0.23, "learning_rate": 9.261322741449632e-06, "loss": 0.0812, "step": 5642 }, { "epoch": 0.23, "learning_rate": 9.261189454322501e-06, "loss": 0.0837, "step": 5643 }, { "epoch": 0.23, "learning_rate": 9.261056167195372e-06, "loss": 0.0999, "step": 5644 }, { "epoch": 0.23, "learning_rate": 9.260922880068243e-06, "loss": 0.1107, "step": 5645 }, { "epoch": 0.23, "learning_rate": 9.260789592941114e-06, "loss": 0.1187, "step": 5646 }, { "epoch": 0.23, "learning_rate": 9.260656305813985e-06, "loss": 0.1181, "step": 5647 }, { "epoch": 0.23, "learning_rate": 9.260523018686855e-06, "loss": 0.091, "step": 5648 }, { "epoch": 0.23, "learning_rate": 9.260389731559727e-06, "loss": 0.0981, "step": 5649 }, { "epoch": 0.23, "learning_rate": 9.260256444432597e-06, "loss": 0.1175, "step": 5650 }, { "epoch": 0.23, "learning_rate": 9.260123157305468e-06, "loss": 0.0886, "step": 5651 }, { "epoch": 0.23, "learning_rate": 9.259989870178339e-06, "loss": 0.1357, "step": 5652 }, { "epoch": 0.23, "learning_rate": 9.25985658305121e-06, "loss": 0.0869, "step": 5653 }, { "epoch": 0.23, "learning_rate": 9.25972329592408e-06, "loss": 0.0925, "step": 5654 }, { "epoch": 0.23, "learning_rate": 9.259590008796952e-06, "loss": 0.1154, "step": 5655 }, { "epoch": 0.23, "learning_rate": 9.259456721669821e-06, "loss": 0.1141, "step": 5656 }, { "epoch": 0.23, "learning_rate": 9.259323434542694e-06, "loss": 0.0806, "step": 5657 }, { "epoch": 0.23, "learning_rate": 9.259190147415563e-06, "loss": 0.1053, "step": 5658 }, { "epoch": 0.23, "learning_rate": 9.259056860288434e-06, "loss": 0.1226, "step": 5659 }, { "epoch": 0.23, "learning_rate": 9.258923573161305e-06, "loss": 0.1118, "step": 5660 }, { "epoch": 0.23, "learning_rate": 9.258790286034176e-06, "loss": 0.1301, "step": 5661 }, { "epoch": 0.23, "learning_rate": 9.258656998907047e-06, "loss": 0.0879, "step": 5662 }, { "epoch": 0.23, "learning_rate": 9.258523711779918e-06, "loss": 0.1123, "step": 5663 }, { "epoch": 0.23, "learning_rate": 9.258390424652787e-06, "loss": 0.0994, "step": 5664 }, { "epoch": 0.23, "learning_rate": 9.258257137525658e-06, "loss": 0.1228, "step": 5665 }, { "epoch": 0.23, "learning_rate": 9.25812385039853e-06, "loss": 0.1348, "step": 5666 }, { "epoch": 0.23, "learning_rate": 9.2579905632714e-06, "loss": 0.0993, "step": 5667 }, { "epoch": 0.23, "learning_rate": 9.257857276144271e-06, "loss": 0.1229, "step": 5668 }, { "epoch": 0.23, "learning_rate": 9.25772398901714e-06, "loss": 0.1088, "step": 5669 }, { "epoch": 0.23, "learning_rate": 9.257590701890013e-06, "loss": 0.1331, "step": 5670 }, { "epoch": 0.23, "learning_rate": 9.257457414762883e-06, "loss": 0.1074, "step": 5671 }, { "epoch": 0.23, "learning_rate": 9.257324127635754e-06, "loss": 0.1155, "step": 5672 }, { "epoch": 0.23, "learning_rate": 9.257190840508625e-06, "loss": 0.1084, "step": 5673 }, { "epoch": 0.23, "learning_rate": 9.257057553381496e-06, "loss": 0.1114, "step": 5674 }, { "epoch": 0.23, "learning_rate": 9.256924266254367e-06, "loss": 0.095, "step": 5675 }, { "epoch": 0.23, "learning_rate": 9.256790979127238e-06, "loss": 0.1588, "step": 5676 }, { "epoch": 0.23, "learning_rate": 9.256657692000107e-06, "loss": 0.159, "step": 5677 }, { "epoch": 0.23, "learning_rate": 9.256524404872978e-06, "loss": 0.1149, "step": 5678 }, { "epoch": 0.23, "learning_rate": 9.256391117745849e-06, "loss": 0.1048, "step": 5679 }, { "epoch": 0.23, "learning_rate": 9.25625783061872e-06, "loss": 0.0884, "step": 5680 }, { "epoch": 0.23, "learning_rate": 9.25612454349159e-06, "loss": 0.103, "step": 5681 }, { "epoch": 0.23, "learning_rate": 9.255991256364462e-06, "loss": 0.0963, "step": 5682 }, { "epoch": 0.23, "learning_rate": 9.255857969237331e-06, "loss": 0.1499, "step": 5683 }, { "epoch": 0.23, "learning_rate": 9.255724682110202e-06, "loss": 0.1192, "step": 5684 }, { "epoch": 0.23, "learning_rate": 9.255591394983073e-06, "loss": 0.1033, "step": 5685 }, { "epoch": 0.23, "learning_rate": 9.255458107855944e-06, "loss": 0.1057, "step": 5686 }, { "epoch": 0.23, "learning_rate": 9.255324820728815e-06, "loss": 0.0909, "step": 5687 }, { "epoch": 0.23, "learning_rate": 9.255191533601684e-06, "loss": 0.1088, "step": 5688 }, { "epoch": 0.23, "learning_rate": 9.255058246474557e-06, "loss": 0.1183, "step": 5689 }, { "epoch": 0.23, "learning_rate": 9.254924959347426e-06, "loss": 0.0714, "step": 5690 }, { "epoch": 0.23, "learning_rate": 9.254791672220297e-06, "loss": 0.085, "step": 5691 }, { "epoch": 0.23, "learning_rate": 9.254658385093168e-06, "loss": 0.1993, "step": 5692 }, { "epoch": 0.23, "learning_rate": 9.25452509796604e-06, "loss": 0.1047, "step": 5693 }, { "epoch": 0.23, "learning_rate": 9.25439181083891e-06, "loss": 0.1168, "step": 5694 }, { "epoch": 0.23, "learning_rate": 9.254258523711781e-06, "loss": 0.1083, "step": 5695 }, { "epoch": 0.23, "learning_rate": 9.25412523658465e-06, "loss": 0.1393, "step": 5696 }, { "epoch": 0.23, "learning_rate": 9.253991949457523e-06, "loss": 0.088, "step": 5697 }, { "epoch": 0.23, "learning_rate": 9.253858662330393e-06, "loss": 0.0812, "step": 5698 }, { "epoch": 0.23, "learning_rate": 9.253725375203264e-06, "loss": 0.1367, "step": 5699 }, { "epoch": 0.23, "learning_rate": 9.253592088076135e-06, "loss": 0.1472, "step": 5700 }, { "epoch": 0.23, "learning_rate": 9.253458800949006e-06, "loss": 0.0793, "step": 5701 }, { "epoch": 0.23, "learning_rate": 9.253325513821877e-06, "loss": 0.1355, "step": 5702 }, { "epoch": 0.23, "learning_rate": 9.253192226694746e-06, "loss": 0.1168, "step": 5703 }, { "epoch": 0.23, "learning_rate": 9.253058939567617e-06, "loss": 0.1276, "step": 5704 }, { "epoch": 0.23, "learning_rate": 9.252925652440488e-06, "loss": 0.1291, "step": 5705 }, { "epoch": 0.23, "learning_rate": 9.252792365313359e-06, "loss": 0.1134, "step": 5706 }, { "epoch": 0.23, "learning_rate": 9.25265907818623e-06, "loss": 0.0916, "step": 5707 }, { "epoch": 0.23, "learning_rate": 9.252525791059101e-06, "loss": 0.1141, "step": 5708 }, { "epoch": 0.23, "learning_rate": 9.25239250393197e-06, "loss": 0.1023, "step": 5709 }, { "epoch": 0.23, "learning_rate": 9.252259216804841e-06, "loss": 0.1018, "step": 5710 }, { "epoch": 0.23, "learning_rate": 9.252125929677712e-06, "loss": 0.0665, "step": 5711 }, { "epoch": 0.23, "learning_rate": 9.251992642550583e-06, "loss": 0.1097, "step": 5712 }, { "epoch": 0.23, "learning_rate": 9.251859355423454e-06, "loss": 0.0977, "step": 5713 }, { "epoch": 0.23, "learning_rate": 9.251726068296325e-06, "loss": 0.112, "step": 5714 }, { "epoch": 0.23, "learning_rate": 9.251592781169194e-06, "loss": 0.1164, "step": 5715 }, { "epoch": 0.23, "learning_rate": 9.251459494042067e-06, "loss": 0.0914, "step": 5716 }, { "epoch": 0.23, "learning_rate": 9.251326206914936e-06, "loss": 0.1408, "step": 5717 }, { "epoch": 0.23, "learning_rate": 9.251192919787807e-06, "loss": 0.1447, "step": 5718 }, { "epoch": 0.23, "learning_rate": 9.251059632660678e-06, "loss": 0.1121, "step": 5719 }, { "epoch": 0.23, "learning_rate": 9.25092634553355e-06, "loss": 0.1073, "step": 5720 }, { "epoch": 0.23, "learning_rate": 9.25079305840642e-06, "loss": 0.0817, "step": 5721 }, { "epoch": 0.23, "learning_rate": 9.250659771279291e-06, "loss": 0.1348, "step": 5722 }, { "epoch": 0.23, "learning_rate": 9.25052648415216e-06, "loss": 0.0826, "step": 5723 }, { "epoch": 0.23, "learning_rate": 9.250393197025032e-06, "loss": 0.102, "step": 5724 }, { "epoch": 0.23, "learning_rate": 9.250259909897903e-06, "loss": 0.0833, "step": 5725 }, { "epoch": 0.23, "learning_rate": 9.250126622770774e-06, "loss": 0.101, "step": 5726 }, { "epoch": 0.23, "learning_rate": 9.249993335643645e-06, "loss": 0.114, "step": 5727 }, { "epoch": 0.23, "learning_rate": 9.249860048516514e-06, "loss": 0.1033, "step": 5728 }, { "epoch": 0.23, "learning_rate": 9.249726761389387e-06, "loss": 0.1706, "step": 5729 }, { "epoch": 0.23, "learning_rate": 9.249593474262256e-06, "loss": 0.0911, "step": 5730 }, { "epoch": 0.23, "learning_rate": 9.249460187135127e-06, "loss": 0.1416, "step": 5731 }, { "epoch": 0.23, "learning_rate": 9.249326900007998e-06, "loss": 0.1064, "step": 5732 }, { "epoch": 0.23, "learning_rate": 9.249193612880869e-06, "loss": 0.0853, "step": 5733 }, { "epoch": 0.23, "learning_rate": 9.24906032575374e-06, "loss": 0.128, "step": 5734 }, { "epoch": 0.23, "learning_rate": 9.248927038626611e-06, "loss": 0.1026, "step": 5735 }, { "epoch": 0.23, "learning_rate": 9.24879375149948e-06, "loss": 0.1187, "step": 5736 }, { "epoch": 0.23, "learning_rate": 9.248660464372353e-06, "loss": 0.0844, "step": 5737 }, { "epoch": 0.23, "learning_rate": 9.248527177245222e-06, "loss": 0.1327, "step": 5738 }, { "epoch": 0.23, "learning_rate": 9.248393890118093e-06, "loss": 0.135, "step": 5739 }, { "epoch": 0.23, "learning_rate": 9.248260602990964e-06, "loss": 0.1279, "step": 5740 }, { "epoch": 0.23, "learning_rate": 9.248127315863835e-06, "loss": 0.0814, "step": 5741 }, { "epoch": 0.23, "learning_rate": 9.247994028736706e-06, "loss": 0.1477, "step": 5742 }, { "epoch": 0.23, "learning_rate": 9.247860741609576e-06, "loss": 0.1051, "step": 5743 }, { "epoch": 0.23, "learning_rate": 9.247727454482447e-06, "loss": 0.1128, "step": 5744 }, { "epoch": 0.23, "learning_rate": 9.247594167355318e-06, "loss": 0.1086, "step": 5745 }, { "epoch": 0.23, "learning_rate": 9.247460880228189e-06, "loss": 0.1221, "step": 5746 }, { "epoch": 0.23, "learning_rate": 9.24732759310106e-06, "loss": 0.1013, "step": 5747 }, { "epoch": 0.23, "learning_rate": 9.24719430597393e-06, "loss": 0.0906, "step": 5748 }, { "epoch": 0.23, "learning_rate": 9.2470610188468e-06, "loss": 0.1179, "step": 5749 }, { "epoch": 0.23, "learning_rate": 9.24692773171967e-06, "loss": 0.1195, "step": 5750 }, { "epoch": 0.23, "learning_rate": 9.246794444592542e-06, "loss": 0.1085, "step": 5751 }, { "epoch": 0.23, "learning_rate": 9.246661157465413e-06, "loss": 0.1452, "step": 5752 }, { "epoch": 0.23, "learning_rate": 9.246527870338284e-06, "loss": 0.1091, "step": 5753 }, { "epoch": 0.23, "learning_rate": 9.246394583211155e-06, "loss": 0.1059, "step": 5754 }, { "epoch": 0.23, "learning_rate": 9.246261296084024e-06, "loss": 0.0782, "step": 5755 }, { "epoch": 0.23, "learning_rate": 9.246128008956897e-06, "loss": 0.1201, "step": 5756 }, { "epoch": 0.23, "learning_rate": 9.245994721829766e-06, "loss": 0.1093, "step": 5757 }, { "epoch": 0.23, "learning_rate": 9.245861434702637e-06, "loss": 0.163, "step": 5758 }, { "epoch": 0.23, "learning_rate": 9.245728147575508e-06, "loss": 0.1509, "step": 5759 }, { "epoch": 0.23, "learning_rate": 9.245594860448379e-06, "loss": 0.1225, "step": 5760 }, { "epoch": 0.23, "learning_rate": 9.24546157332125e-06, "loss": 0.1102, "step": 5761 }, { "epoch": 0.23, "learning_rate": 9.24532828619412e-06, "loss": 0.1387, "step": 5762 }, { "epoch": 0.23, "learning_rate": 9.24519499906699e-06, "loss": 0.0825, "step": 5763 }, { "epoch": 0.23, "learning_rate": 9.245061711939861e-06, "loss": 0.1174, "step": 5764 }, { "epoch": 0.23, "learning_rate": 9.244928424812732e-06, "loss": 0.1026, "step": 5765 }, { "epoch": 0.23, "learning_rate": 9.244795137685603e-06, "loss": 0.0916, "step": 5766 }, { "epoch": 0.23, "learning_rate": 9.244661850558474e-06, "loss": 0.096, "step": 5767 }, { "epoch": 0.23, "learning_rate": 9.244528563431344e-06, "loss": 0.1044, "step": 5768 }, { "epoch": 0.23, "learning_rate": 9.244395276304216e-06, "loss": 0.0865, "step": 5769 }, { "epoch": 0.23, "learning_rate": 9.244261989177086e-06, "loss": 0.1177, "step": 5770 }, { "epoch": 0.23, "learning_rate": 9.244128702049957e-06, "loss": 0.1571, "step": 5771 }, { "epoch": 0.23, "learning_rate": 9.243995414922828e-06, "loss": 0.1286, "step": 5772 }, { "epoch": 0.23, "learning_rate": 9.243862127795699e-06, "loss": 0.156, "step": 5773 }, { "epoch": 0.23, "learning_rate": 9.24372884066857e-06, "loss": 0.0782, "step": 5774 }, { "epoch": 0.23, "learning_rate": 9.24359555354144e-06, "loss": 0.0998, "step": 5775 }, { "epoch": 0.23, "learning_rate": 9.24346226641431e-06, "loss": 0.1039, "step": 5776 }, { "epoch": 0.23, "learning_rate": 9.243328979287181e-06, "loss": 0.1258, "step": 5777 }, { "epoch": 0.23, "learning_rate": 9.243195692160052e-06, "loss": 0.0957, "step": 5778 }, { "epoch": 0.23, "learning_rate": 9.243062405032923e-06, "loss": 0.1071, "step": 5779 }, { "epoch": 0.23, "learning_rate": 9.242929117905794e-06, "loss": 0.1115, "step": 5780 }, { "epoch": 0.23, "learning_rate": 9.242795830778663e-06, "loss": 0.1306, "step": 5781 }, { "epoch": 0.23, "learning_rate": 9.242662543651534e-06, "loss": 0.1181, "step": 5782 }, { "epoch": 0.23, "learning_rate": 9.242529256524405e-06, "loss": 0.1178, "step": 5783 }, { "epoch": 0.23, "learning_rate": 9.242395969397276e-06, "loss": 0.1028, "step": 5784 }, { "epoch": 0.23, "learning_rate": 9.242262682270147e-06, "loss": 0.1266, "step": 5785 }, { "epoch": 0.23, "learning_rate": 9.242129395143018e-06, "loss": 0.1333, "step": 5786 }, { "epoch": 0.23, "learning_rate": 9.241996108015887e-06, "loss": 0.1121, "step": 5787 }, { "epoch": 0.23, "learning_rate": 9.24186282088876e-06, "loss": 0.1221, "step": 5788 }, { "epoch": 0.23, "learning_rate": 9.24172953376163e-06, "loss": 0.1386, "step": 5789 }, { "epoch": 0.23, "learning_rate": 9.2415962466345e-06, "loss": 0.1074, "step": 5790 }, { "epoch": 0.23, "learning_rate": 9.241462959507371e-06, "loss": 0.149, "step": 5791 }, { "epoch": 0.23, "learning_rate": 9.241329672380242e-06, "loss": 0.106, "step": 5792 }, { "epoch": 0.23, "learning_rate": 9.241196385253113e-06, "loss": 0.1061, "step": 5793 }, { "epoch": 0.23, "learning_rate": 9.241063098125984e-06, "loss": 0.0739, "step": 5794 }, { "epoch": 0.23, "learning_rate": 9.240929810998854e-06, "loss": 0.1004, "step": 5795 }, { "epoch": 0.23, "learning_rate": 9.240796523871726e-06, "loss": 0.1341, "step": 5796 }, { "epoch": 0.23, "learning_rate": 9.240663236744596e-06, "loss": 0.0829, "step": 5797 }, { "epoch": 0.23, "learning_rate": 9.240529949617467e-06, "loss": 0.121, "step": 5798 }, { "epoch": 0.23, "learning_rate": 9.240396662490338e-06, "loss": 0.1081, "step": 5799 }, { "epoch": 0.23, "learning_rate": 9.240263375363209e-06, "loss": 0.098, "step": 5800 }, { "epoch": 0.23, "learning_rate": 9.24013008823608e-06, "loss": 0.0979, "step": 5801 }, { "epoch": 0.23, "learning_rate": 9.239996801108949e-06, "loss": 0.1208, "step": 5802 }, { "epoch": 0.23, "learning_rate": 9.23986351398182e-06, "loss": 0.0958, "step": 5803 }, { "epoch": 0.23, "learning_rate": 9.239730226854691e-06, "loss": 0.0567, "step": 5804 }, { "epoch": 0.23, "learning_rate": 9.239596939727562e-06, "loss": 0.1218, "step": 5805 }, { "epoch": 0.23, "learning_rate": 9.239463652600433e-06, "loss": 0.0794, "step": 5806 }, { "epoch": 0.23, "learning_rate": 9.239330365473304e-06, "loss": 0.1223, "step": 5807 }, { "epoch": 0.23, "learning_rate": 9.239197078346173e-06, "loss": 0.1013, "step": 5808 }, { "epoch": 0.23, "learning_rate": 9.239063791219044e-06, "loss": 0.1139, "step": 5809 }, { "epoch": 0.23, "learning_rate": 9.238930504091915e-06, "loss": 0.1239, "step": 5810 }, { "epoch": 0.23, "learning_rate": 9.238797216964786e-06, "loss": 0.0975, "step": 5811 }, { "epoch": 0.23, "learning_rate": 9.238663929837657e-06, "loss": 0.1326, "step": 5812 }, { "epoch": 0.23, "learning_rate": 9.238530642710528e-06, "loss": 0.1328, "step": 5813 }, { "epoch": 0.23, "learning_rate": 9.238397355583398e-06, "loss": 0.129, "step": 5814 }, { "epoch": 0.23, "learning_rate": 9.23826406845627e-06, "loss": 0.1882, "step": 5815 }, { "epoch": 0.23, "learning_rate": 9.23813078132914e-06, "loss": 0.0948, "step": 5816 }, { "epoch": 0.23, "learning_rate": 9.23799749420201e-06, "loss": 0.1056, "step": 5817 }, { "epoch": 0.23, "learning_rate": 9.237864207074881e-06, "loss": 0.1181, "step": 5818 }, { "epoch": 0.23, "learning_rate": 9.237730919947752e-06, "loss": 0.107, "step": 5819 }, { "epoch": 0.23, "learning_rate": 9.237597632820623e-06, "loss": 0.0953, "step": 5820 }, { "epoch": 0.23, "learning_rate": 9.237464345693493e-06, "loss": 0.0996, "step": 5821 }, { "epoch": 0.23, "learning_rate": 9.237331058566364e-06, "loss": 0.0844, "step": 5822 }, { "epoch": 0.23, "learning_rate": 9.237197771439235e-06, "loss": 0.0903, "step": 5823 }, { "epoch": 0.23, "learning_rate": 9.237064484312106e-06, "loss": 0.0845, "step": 5824 }, { "epoch": 0.23, "learning_rate": 9.236931197184977e-06, "loss": 0.1373, "step": 5825 }, { "epoch": 0.23, "learning_rate": 9.236797910057848e-06, "loss": 0.1202, "step": 5826 }, { "epoch": 0.23, "learning_rate": 9.236664622930717e-06, "loss": 0.1105, "step": 5827 }, { "epoch": 0.23, "learning_rate": 9.23653133580359e-06, "loss": 0.137, "step": 5828 }, { "epoch": 0.23, "learning_rate": 9.236398048676459e-06, "loss": 0.1133, "step": 5829 }, { "epoch": 0.23, "learning_rate": 9.23626476154933e-06, "loss": 0.1054, "step": 5830 }, { "epoch": 0.23, "learning_rate": 9.236131474422201e-06, "loss": 0.0894, "step": 5831 }, { "epoch": 0.23, "learning_rate": 9.235998187295072e-06, "loss": 0.1108, "step": 5832 }, { "epoch": 0.23, "learning_rate": 9.235864900167943e-06, "loss": 0.0982, "step": 5833 }, { "epoch": 0.23, "learning_rate": 9.235731613040814e-06, "loss": 0.1046, "step": 5834 }, { "epoch": 0.23, "learning_rate": 9.235598325913683e-06, "loss": 0.1178, "step": 5835 }, { "epoch": 0.23, "learning_rate": 9.235465038786556e-06, "loss": 0.1485, "step": 5836 }, { "epoch": 0.23, "learning_rate": 9.235331751659425e-06, "loss": 0.112, "step": 5837 }, { "epoch": 0.23, "learning_rate": 9.235198464532296e-06, "loss": 0.175, "step": 5838 }, { "epoch": 0.23, "learning_rate": 9.235065177405167e-06, "loss": 0.1131, "step": 5839 }, { "epoch": 0.23, "learning_rate": 9.234931890278037e-06, "loss": 0.1055, "step": 5840 }, { "epoch": 0.23, "learning_rate": 9.23479860315091e-06, "loss": 0.1345, "step": 5841 }, { "epoch": 0.23, "learning_rate": 9.234665316023779e-06, "loss": 0.1044, "step": 5842 }, { "epoch": 0.23, "learning_rate": 9.23453202889665e-06, "loss": 0.1344, "step": 5843 }, { "epoch": 0.23, "learning_rate": 9.23439874176952e-06, "loss": 0.1449, "step": 5844 }, { "epoch": 0.23, "learning_rate": 9.234265454642392e-06, "loss": 0.089, "step": 5845 }, { "epoch": 0.23, "learning_rate": 9.234132167515263e-06, "loss": 0.131, "step": 5846 }, { "epoch": 0.23, "learning_rate": 9.233998880388134e-06, "loss": 0.102, "step": 5847 }, { "epoch": 0.23, "learning_rate": 9.233865593261003e-06, "loss": 0.1284, "step": 5848 }, { "epoch": 0.23, "learning_rate": 9.233732306133874e-06, "loss": 0.1078, "step": 5849 }, { "epoch": 0.23, "learning_rate": 9.233599019006745e-06, "loss": 0.1216, "step": 5850 }, { "epoch": 0.23, "learning_rate": 9.233465731879616e-06, "loss": 0.1359, "step": 5851 }, { "epoch": 0.23, "learning_rate": 9.233332444752487e-06, "loss": 0.152, "step": 5852 }, { "epoch": 0.23, "learning_rate": 9.233199157625358e-06, "loss": 0.1295, "step": 5853 }, { "epoch": 0.23, "learning_rate": 9.233065870498227e-06, "loss": 0.0873, "step": 5854 }, { "epoch": 0.23, "learning_rate": 9.2329325833711e-06, "loss": 0.1088, "step": 5855 }, { "epoch": 0.23, "learning_rate": 9.232799296243969e-06, "loss": 0.0954, "step": 5856 }, { "epoch": 0.23, "learning_rate": 9.23266600911684e-06, "loss": 0.0877, "step": 5857 }, { "epoch": 0.23, "learning_rate": 9.232532721989711e-06, "loss": 0.1331, "step": 5858 }, { "epoch": 0.23, "learning_rate": 9.232399434862582e-06, "loss": 0.1209, "step": 5859 }, { "epoch": 0.23, "learning_rate": 9.232266147735453e-06, "loss": 0.1145, "step": 5860 }, { "epoch": 0.23, "learning_rate": 9.232132860608322e-06, "loss": 0.074, "step": 5861 }, { "epoch": 0.23, "learning_rate": 9.231999573481193e-06, "loss": 0.0996, "step": 5862 }, { "epoch": 0.23, "learning_rate": 9.231866286354064e-06, "loss": 0.1011, "step": 5863 }, { "epoch": 0.23, "learning_rate": 9.231732999226935e-06, "loss": 0.0849, "step": 5864 }, { "epoch": 0.23, "learning_rate": 9.231599712099806e-06, "loss": 0.1772, "step": 5865 }, { "epoch": 0.23, "learning_rate": 9.231466424972677e-06, "loss": 0.1334, "step": 5866 }, { "epoch": 0.23, "learning_rate": 9.231333137845547e-06, "loss": 0.0925, "step": 5867 }, { "epoch": 0.23, "learning_rate": 9.23119985071842e-06, "loss": 0.1622, "step": 5868 }, { "epoch": 0.23, "learning_rate": 9.231066563591289e-06, "loss": 0.1166, "step": 5869 }, { "epoch": 0.23, "learning_rate": 9.23093327646416e-06, "loss": 0.164, "step": 5870 }, { "epoch": 0.23, "learning_rate": 9.23079998933703e-06, "loss": 0.1198, "step": 5871 }, { "epoch": 0.23, "learning_rate": 9.230666702209902e-06, "loss": 0.128, "step": 5872 }, { "epoch": 0.23, "learning_rate": 9.230533415082773e-06, "loss": 0.1035, "step": 5873 }, { "epoch": 0.23, "learning_rate": 9.230400127955644e-06, "loss": 0.1177, "step": 5874 }, { "epoch": 0.23, "learning_rate": 9.230266840828513e-06, "loss": 0.1299, "step": 5875 }, { "epoch": 0.23, "learning_rate": 9.230133553701384e-06, "loss": 0.1269, "step": 5876 }, { "epoch": 0.23, "learning_rate": 9.230000266574255e-06, "loss": 0.1297, "step": 5877 }, { "epoch": 0.23, "learning_rate": 9.229866979447126e-06, "loss": 0.1341, "step": 5878 }, { "epoch": 0.23, "learning_rate": 9.229733692319997e-06, "loss": 0.1644, "step": 5879 }, { "epoch": 0.23, "learning_rate": 9.229600405192866e-06, "loss": 0.0999, "step": 5880 }, { "epoch": 0.23, "learning_rate": 9.229467118065737e-06, "loss": 0.0735, "step": 5881 }, { "epoch": 0.23, "learning_rate": 9.229333830938608e-06, "loss": 0.0837, "step": 5882 }, { "epoch": 0.23, "learning_rate": 9.22920054381148e-06, "loss": 0.1279, "step": 5883 }, { "epoch": 0.23, "learning_rate": 9.22906725668435e-06, "loss": 0.1182, "step": 5884 }, { "epoch": 0.24, "learning_rate": 9.228933969557221e-06, "loss": 0.0967, "step": 5885 }, { "epoch": 0.24, "learning_rate": 9.22880068243009e-06, "loss": 0.1075, "step": 5886 }, { "epoch": 0.24, "learning_rate": 9.228667395302963e-06, "loss": 0.1334, "step": 5887 }, { "epoch": 0.24, "learning_rate": 9.228534108175832e-06, "loss": 0.1123, "step": 5888 }, { "epoch": 0.24, "learning_rate": 9.228400821048703e-06, "loss": 0.0722, "step": 5889 }, { "epoch": 0.24, "learning_rate": 9.228267533921574e-06, "loss": 0.1126, "step": 5890 }, { "epoch": 0.24, "learning_rate": 9.228134246794445e-06, "loss": 0.1298, "step": 5891 }, { "epoch": 0.24, "learning_rate": 9.228000959667316e-06, "loss": 0.1285, "step": 5892 }, { "epoch": 0.24, "learning_rate": 9.227867672540187e-06, "loss": 0.1268, "step": 5893 }, { "epoch": 0.24, "learning_rate": 9.227734385413057e-06, "loss": 0.1167, "step": 5894 }, { "epoch": 0.24, "learning_rate": 9.22760109828593e-06, "loss": 0.0772, "step": 5895 }, { "epoch": 0.24, "learning_rate": 9.227467811158799e-06, "loss": 0.107, "step": 5896 }, { "epoch": 0.24, "learning_rate": 9.22733452403167e-06, "loss": 0.0872, "step": 5897 }, { "epoch": 0.24, "learning_rate": 9.22720123690454e-06, "loss": 0.1646, "step": 5898 }, { "epoch": 0.24, "learning_rate": 9.22706794977741e-06, "loss": 0.1247, "step": 5899 }, { "epoch": 0.24, "learning_rate": 9.226934662650283e-06, "loss": 0.1085, "step": 5900 }, { "epoch": 0.24, "learning_rate": 9.226801375523152e-06, "loss": 0.1259, "step": 5901 }, { "epoch": 0.24, "learning_rate": 9.226668088396023e-06, "loss": 0.1118, "step": 5902 }, { "epoch": 0.24, "learning_rate": 9.226534801268894e-06, "loss": 0.1163, "step": 5903 }, { "epoch": 0.24, "learning_rate": 9.226401514141765e-06, "loss": 0.0968, "step": 5904 }, { "epoch": 0.24, "learning_rate": 9.226268227014636e-06, "loss": 0.0751, "step": 5905 }, { "epoch": 0.24, "learning_rate": 9.226134939887507e-06, "loss": 0.1232, "step": 5906 }, { "epoch": 0.24, "learning_rate": 9.226001652760376e-06, "loss": 0.0955, "step": 5907 }, { "epoch": 0.24, "learning_rate": 9.225868365633249e-06, "loss": 0.1597, "step": 5908 }, { "epoch": 0.24, "learning_rate": 9.225735078506118e-06, "loss": 0.1154, "step": 5909 }, { "epoch": 0.24, "learning_rate": 9.22560179137899e-06, "loss": 0.135, "step": 5910 }, { "epoch": 0.24, "learning_rate": 9.22546850425186e-06, "loss": 0.1026, "step": 5911 }, { "epoch": 0.24, "learning_rate": 9.225335217124731e-06, "loss": 0.1281, "step": 5912 }, { "epoch": 0.24, "learning_rate": 9.225201929997602e-06, "loss": 0.1961, "step": 5913 }, { "epoch": 0.24, "learning_rate": 9.225068642870473e-06, "loss": 0.1361, "step": 5914 }, { "epoch": 0.24, "learning_rate": 9.224935355743343e-06, "loss": 0.0807, "step": 5915 }, { "epoch": 0.24, "learning_rate": 9.224802068616214e-06, "loss": 0.1486, "step": 5916 }, { "epoch": 0.24, "learning_rate": 9.224668781489085e-06, "loss": 0.1898, "step": 5917 }, { "epoch": 0.24, "learning_rate": 9.224535494361956e-06, "loss": 0.1178, "step": 5918 }, { "epoch": 0.24, "learning_rate": 9.224402207234827e-06, "loss": 0.1359, "step": 5919 }, { "epoch": 0.24, "learning_rate": 9.224268920107696e-06, "loss": 0.1055, "step": 5920 }, { "epoch": 0.24, "learning_rate": 9.224135632980567e-06, "loss": 0.0882, "step": 5921 }, { "epoch": 0.24, "learning_rate": 9.224002345853438e-06, "loss": 0.1195, "step": 5922 }, { "epoch": 0.24, "learning_rate": 9.223869058726309e-06, "loss": 0.0848, "step": 5923 }, { "epoch": 0.24, "learning_rate": 9.22373577159918e-06, "loss": 0.1326, "step": 5924 }, { "epoch": 0.24, "learning_rate": 9.22360248447205e-06, "loss": 0.1306, "step": 5925 }, { "epoch": 0.24, "learning_rate": 9.22346919734492e-06, "loss": 0.1427, "step": 5926 }, { "epoch": 0.24, "learning_rate": 9.223335910217793e-06, "loss": 0.0888, "step": 5927 }, { "epoch": 0.24, "learning_rate": 9.223202623090662e-06, "loss": 0.1055, "step": 5928 }, { "epoch": 0.24, "learning_rate": 9.223069335963533e-06, "loss": 0.091, "step": 5929 }, { "epoch": 0.24, "learning_rate": 9.222936048836404e-06, "loss": 0.0867, "step": 5930 }, { "epoch": 0.24, "learning_rate": 9.222802761709275e-06, "loss": 0.0833, "step": 5931 }, { "epoch": 0.24, "learning_rate": 9.222669474582146e-06, "loss": 0.0864, "step": 5932 }, { "epoch": 0.24, "learning_rate": 9.222536187455017e-06, "loss": 0.1077, "step": 5933 }, { "epoch": 0.24, "learning_rate": 9.222402900327886e-06, "loss": 0.098, "step": 5934 }, { "epoch": 0.24, "learning_rate": 9.222269613200759e-06, "loss": 0.0853, "step": 5935 }, { "epoch": 0.24, "learning_rate": 9.222136326073628e-06, "loss": 0.1151, "step": 5936 }, { "epoch": 0.24, "learning_rate": 9.2220030389465e-06, "loss": 0.0961, "step": 5937 }, { "epoch": 0.24, "learning_rate": 9.22186975181937e-06, "loss": 0.1215, "step": 5938 }, { "epoch": 0.24, "learning_rate": 9.22173646469224e-06, "loss": 0.1235, "step": 5939 }, { "epoch": 0.24, "learning_rate": 9.221603177565112e-06, "loss": 0.0859, "step": 5940 }, { "epoch": 0.24, "learning_rate": 9.221469890437982e-06, "loss": 0.1308, "step": 5941 }, { "epoch": 0.24, "learning_rate": 9.221336603310853e-06, "loss": 0.1233, "step": 5942 }, { "epoch": 0.24, "learning_rate": 9.221203316183724e-06, "loss": 0.122, "step": 5943 }, { "epoch": 0.24, "learning_rate": 9.221070029056595e-06, "loss": 0.0945, "step": 5944 }, { "epoch": 0.24, "learning_rate": 9.220936741929466e-06, "loss": 0.1204, "step": 5945 }, { "epoch": 0.24, "learning_rate": 9.220803454802337e-06, "loss": 0.1177, "step": 5946 }, { "epoch": 0.24, "learning_rate": 9.220670167675206e-06, "loss": 0.1381, "step": 5947 }, { "epoch": 0.24, "learning_rate": 9.220536880548077e-06, "loss": 0.0811, "step": 5948 }, { "epoch": 0.24, "learning_rate": 9.220403593420948e-06, "loss": 0.0873, "step": 5949 }, { "epoch": 0.24, "learning_rate": 9.220270306293819e-06, "loss": 0.1142, "step": 5950 }, { "epoch": 0.24, "learning_rate": 9.22013701916669e-06, "loss": 0.1366, "step": 5951 }, { "epoch": 0.24, "learning_rate": 9.220003732039561e-06, "loss": 0.1171, "step": 5952 }, { "epoch": 0.24, "learning_rate": 9.21987044491243e-06, "loss": 0.1142, "step": 5953 }, { "epoch": 0.24, "learning_rate": 9.219737157785303e-06, "loss": 0.1521, "step": 5954 }, { "epoch": 0.24, "learning_rate": 9.219603870658172e-06, "loss": 0.119, "step": 5955 }, { "epoch": 0.24, "learning_rate": 9.219470583531043e-06, "loss": 0.132, "step": 5956 }, { "epoch": 0.24, "learning_rate": 9.219337296403914e-06, "loss": 0.0982, "step": 5957 }, { "epoch": 0.24, "learning_rate": 9.219204009276783e-06, "loss": 0.1425, "step": 5958 }, { "epoch": 0.24, "learning_rate": 9.219070722149656e-06, "loss": 0.111, "step": 5959 }, { "epoch": 0.24, "learning_rate": 9.218937435022525e-06, "loss": 0.1103, "step": 5960 }, { "epoch": 0.24, "learning_rate": 9.218804147895396e-06, "loss": 0.0783, "step": 5961 }, { "epoch": 0.24, "learning_rate": 9.218670860768267e-06, "loss": 0.1076, "step": 5962 }, { "epoch": 0.24, "learning_rate": 9.218537573641138e-06, "loss": 0.131, "step": 5963 }, { "epoch": 0.24, "learning_rate": 9.21840428651401e-06, "loss": 0.1177, "step": 5964 }, { "epoch": 0.24, "learning_rate": 9.21827099938688e-06, "loss": 0.0917, "step": 5965 }, { "epoch": 0.24, "learning_rate": 9.21813771225975e-06, "loss": 0.0983, "step": 5966 }, { "epoch": 0.24, "learning_rate": 9.218004425132622e-06, "loss": 0.1191, "step": 5967 }, { "epoch": 0.24, "learning_rate": 9.217871138005492e-06, "loss": 0.105, "step": 5968 }, { "epoch": 0.24, "learning_rate": 9.217737850878363e-06, "loss": 0.1467, "step": 5969 }, { "epoch": 0.24, "learning_rate": 9.217604563751234e-06, "loss": 0.095, "step": 5970 }, { "epoch": 0.24, "learning_rate": 9.217471276624105e-06, "loss": 0.0904, "step": 5971 }, { "epoch": 0.24, "learning_rate": 9.217337989496976e-06, "loss": 0.0978, "step": 5972 }, { "epoch": 0.24, "learning_rate": 9.217204702369847e-06, "loss": 0.151, "step": 5973 }, { "epoch": 0.24, "learning_rate": 9.217071415242716e-06, "loss": 0.1406, "step": 5974 }, { "epoch": 0.24, "learning_rate": 9.216938128115589e-06, "loss": 0.0842, "step": 5975 }, { "epoch": 0.24, "learning_rate": 9.216804840988458e-06, "loss": 0.0954, "step": 5976 }, { "epoch": 0.24, "learning_rate": 9.216671553861329e-06, "loss": 0.1453, "step": 5977 }, { "epoch": 0.24, "learning_rate": 9.2165382667342e-06, "loss": 0.1462, "step": 5978 }, { "epoch": 0.24, "learning_rate": 9.21640497960707e-06, "loss": 0.1297, "step": 5979 }, { "epoch": 0.24, "learning_rate": 9.216271692479942e-06, "loss": 0.1359, "step": 5980 }, { "epoch": 0.24, "learning_rate": 9.216138405352811e-06, "loss": 0.1407, "step": 5981 }, { "epoch": 0.24, "learning_rate": 9.216005118225682e-06, "loss": 0.1006, "step": 5982 }, { "epoch": 0.24, "learning_rate": 9.215871831098553e-06, "loss": 0.1347, "step": 5983 }, { "epoch": 0.24, "learning_rate": 9.215738543971424e-06, "loss": 0.1264, "step": 5984 }, { "epoch": 0.24, "learning_rate": 9.215605256844295e-06, "loss": 0.1096, "step": 5985 }, { "epoch": 0.24, "learning_rate": 9.215471969717166e-06, "loss": 0.1352, "step": 5986 }, { "epoch": 0.24, "learning_rate": 9.215338682590036e-06, "loss": 0.0858, "step": 5987 }, { "epoch": 0.24, "learning_rate": 9.215205395462907e-06, "loss": 0.1241, "step": 5988 }, { "epoch": 0.24, "learning_rate": 9.215072108335778e-06, "loss": 0.1072, "step": 5989 }, { "epoch": 0.24, "learning_rate": 9.214938821208649e-06, "loss": 0.1077, "step": 5990 }, { "epoch": 0.24, "learning_rate": 9.21480553408152e-06, "loss": 0.0853, "step": 5991 }, { "epoch": 0.24, "learning_rate": 9.21467224695439e-06, "loss": 0.1042, "step": 5992 }, { "epoch": 0.24, "learning_rate": 9.21453895982726e-06, "loss": 0.1278, "step": 5993 }, { "epoch": 0.24, "learning_rate": 9.214405672700133e-06, "loss": 0.0986, "step": 5994 }, { "epoch": 0.24, "learning_rate": 9.214272385573002e-06, "loss": 0.1163, "step": 5995 }, { "epoch": 0.24, "learning_rate": 9.214139098445873e-06, "loss": 0.1462, "step": 5996 }, { "epoch": 0.24, "learning_rate": 9.214005811318744e-06, "loss": 0.1131, "step": 5997 }, { "epoch": 0.24, "learning_rate": 9.213872524191613e-06, "loss": 0.0925, "step": 5998 }, { "epoch": 0.24, "learning_rate": 9.213739237064486e-06, "loss": 0.1135, "step": 5999 }, { "epoch": 0.24, "learning_rate": 9.213605949937355e-06, "loss": 0.1286, "step": 6000 } ], "max_steps": 75126, "num_train_epochs": 3, "total_flos": 7.671059653304844e+18, "trial_name": null, "trial_params": null }