{ "best_metric": null, "best_model_checkpoint": null, "epoch": 2.0, "global_step": 11066, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 2.9997288993312853e-05, "loss": 5.8131, "step": 1 }, { "epoch": 0.0, "learning_rate": 2.99945779866257e-05, "loss": 5.867, "step": 2 }, { "epoch": 0.0, "learning_rate": 2.9991866979938553e-05, "loss": 5.7093, "step": 3 }, { "epoch": 0.0, "learning_rate": 2.99891559732514e-05, "loss": 5.6477, "step": 4 }, { "epoch": 0.0, "learning_rate": 2.998644496656425e-05, "loss": 5.5229, "step": 5 }, { "epoch": 0.0, "learning_rate": 2.99837339598771e-05, "loss": 5.4867, "step": 6 }, { "epoch": 0.0, "learning_rate": 2.998102295318995e-05, "loss": 5.4317, "step": 7 }, { "epoch": 0.0, "learning_rate": 2.9978311946502802e-05, "loss": 5.3849, "step": 8 }, { "epoch": 0.0, "learning_rate": 2.9975600939815654e-05, "loss": 5.2677, "step": 9 }, { "epoch": 0.0, "learning_rate": 2.9972889933128502e-05, "loss": 5.1013, "step": 10 }, { "epoch": 0.0, "learning_rate": 2.9970178926441354e-05, "loss": 4.9517, "step": 11 }, { "epoch": 0.0, "learning_rate": 2.9967467919754202e-05, "loss": 4.8092, "step": 12 }, { "epoch": 0.0, "learning_rate": 2.996475691306705e-05, "loss": 4.9024, "step": 13 }, { "epoch": 0.0, "learning_rate": 2.9962045906379903e-05, "loss": 4.7373, "step": 14 }, { "epoch": 0.0, "learning_rate": 2.9959334899692754e-05, "loss": 4.8521, "step": 15 }, { "epoch": 0.0, "learning_rate": 2.9956623893005603e-05, "loss": 4.6115, "step": 16 }, { "epoch": 0.0, "learning_rate": 2.9953912886318455e-05, "loss": 4.8838, "step": 17 }, { "epoch": 0.0, "learning_rate": 2.9951201879631303e-05, "loss": 4.4024, "step": 18 }, { "epoch": 0.0, "learning_rate": 2.9948490872944155e-05, "loss": 4.4759, "step": 19 }, { "epoch": 0.0, "learning_rate": 2.9945779866257003e-05, "loss": 4.518, "step": 20 }, { "epoch": 0.0, "learning_rate": 2.9943068859569852e-05, "loss": 4.563, "step": 21 }, { "epoch": 0.0, "learning_rate": 2.9940357852882704e-05, "loss": 4.5919, "step": 22 }, { "epoch": 0.0, "learning_rate": 2.9937646846195556e-05, "loss": 4.7354, "step": 23 }, { "epoch": 0.0, "learning_rate": 2.9934935839508404e-05, "loss": 4.5677, "step": 24 }, { "epoch": 0.0, "learning_rate": 2.9932224832821256e-05, "loss": 4.2246, "step": 25 }, { "epoch": 0.0, "learning_rate": 2.9929513826134108e-05, "loss": 4.3559, "step": 26 }, { "epoch": 0.0, "learning_rate": 2.9926802819446956e-05, "loss": 3.9653, "step": 27 }, { "epoch": 0.01, "learning_rate": 2.9924091812759805e-05, "loss": 4.5756, "step": 28 }, { "epoch": 0.01, "learning_rate": 2.9921380806072653e-05, "loss": 4.1574, "step": 29 }, { "epoch": 0.01, "learning_rate": 2.9918669799385505e-05, "loss": 4.3055, "step": 30 }, { "epoch": 0.01, "learning_rate": 2.9915958792698357e-05, "loss": 4.4877, "step": 31 }, { "epoch": 0.01, "learning_rate": 2.9913247786011205e-05, "loss": 4.3467, "step": 32 }, { "epoch": 0.01, "learning_rate": 2.9910536779324057e-05, "loss": 4.5641, "step": 33 }, { "epoch": 0.01, "learning_rate": 2.990782577263691e-05, "loss": 4.4136, "step": 34 }, { "epoch": 0.01, "learning_rate": 2.9905114765949757e-05, "loss": 4.1937, "step": 35 }, { "epoch": 0.01, "learning_rate": 2.9902403759262606e-05, "loss": 3.8617, "step": 36 }, { "epoch": 0.01, "learning_rate": 2.9899692752575457e-05, "loss": 4.1304, "step": 37 }, { "epoch": 0.01, "learning_rate": 2.9896981745888306e-05, "loss": 3.829, "step": 38 }, { "epoch": 0.01, "learning_rate": 2.9894270739201158e-05, "loss": 4.1963, "step": 39 }, { "epoch": 0.01, "learning_rate": 2.9891559732514006e-05, "loss": 4.2229, "step": 40 }, { "epoch": 0.01, "learning_rate": 2.9888848725826858e-05, "loss": 3.7992, "step": 41 }, { "epoch": 0.01, "learning_rate": 2.988613771913971e-05, "loss": 4.1811, "step": 42 }, { "epoch": 0.01, "learning_rate": 2.9883426712452558e-05, "loss": 3.9828, "step": 43 }, { "epoch": 0.01, "learning_rate": 2.9880715705765407e-05, "loss": 4.2949, "step": 44 }, { "epoch": 0.01, "learning_rate": 2.987800469907826e-05, "loss": 3.5624, "step": 45 }, { "epoch": 0.01, "learning_rate": 2.9875293692391107e-05, "loss": 3.7444, "step": 46 }, { "epoch": 0.01, "learning_rate": 2.987258268570396e-05, "loss": 3.5932, "step": 47 }, { "epoch": 0.01, "learning_rate": 2.986987167901681e-05, "loss": 3.7105, "step": 48 }, { "epoch": 0.01, "learning_rate": 2.986716067232966e-05, "loss": 4.0077, "step": 49 }, { "epoch": 0.01, "learning_rate": 2.986444966564251e-05, "loss": 3.5106, "step": 50 }, { "epoch": 0.01, "learning_rate": 2.986173865895536e-05, "loss": 4.0515, "step": 51 }, { "epoch": 0.01, "learning_rate": 2.9859027652268208e-05, "loss": 3.5935, "step": 52 }, { "epoch": 0.01, "learning_rate": 2.985631664558106e-05, "loss": 3.942, "step": 53 }, { "epoch": 0.01, "learning_rate": 2.9853605638893908e-05, "loss": 3.8536, "step": 54 }, { "epoch": 0.01, "learning_rate": 2.985089463220676e-05, "loss": 3.3275, "step": 55 }, { "epoch": 0.01, "learning_rate": 2.984818362551961e-05, "loss": 4.062, "step": 56 }, { "epoch": 0.01, "learning_rate": 2.984547261883246e-05, "loss": 3.7084, "step": 57 }, { "epoch": 0.01, "learning_rate": 2.9842761612145312e-05, "loss": 3.5701, "step": 58 }, { "epoch": 0.01, "learning_rate": 2.9840050605458164e-05, "loss": 3.3679, "step": 59 }, { "epoch": 0.01, "learning_rate": 2.983733959877101e-05, "loss": 4.0951, "step": 60 }, { "epoch": 0.01, "learning_rate": 2.983462859208386e-05, "loss": 3.4606, "step": 61 }, { "epoch": 0.01, "learning_rate": 2.983191758539671e-05, "loss": 2.9702, "step": 62 }, { "epoch": 0.01, "learning_rate": 2.982920657870956e-05, "loss": 3.2061, "step": 63 }, { "epoch": 0.01, "learning_rate": 2.9826495572022413e-05, "loss": 3.5631, "step": 64 }, { "epoch": 0.01, "learning_rate": 2.982378456533526e-05, "loss": 3.2902, "step": 65 }, { "epoch": 0.01, "learning_rate": 2.9821073558648113e-05, "loss": 3.5982, "step": 66 }, { "epoch": 0.01, "learning_rate": 2.9818362551960965e-05, "loss": 3.7596, "step": 67 }, { "epoch": 0.01, "learning_rate": 2.981565154527381e-05, "loss": 3.1705, "step": 68 }, { "epoch": 0.01, "learning_rate": 2.9812940538586662e-05, "loss": 3.4415, "step": 69 }, { "epoch": 0.01, "learning_rate": 2.9810229531899514e-05, "loss": 2.5929, "step": 70 }, { "epoch": 0.01, "learning_rate": 2.9807518525212362e-05, "loss": 3.0434, "step": 71 }, { "epoch": 0.01, "learning_rate": 2.9804807518525214e-05, "loss": 3.2459, "step": 72 }, { "epoch": 0.01, "learning_rate": 2.9802096511838062e-05, "loss": 3.2382, "step": 73 }, { "epoch": 0.01, "learning_rate": 2.9799385505150914e-05, "loss": 3.4536, "step": 74 }, { "epoch": 0.01, "learning_rate": 2.9796674498463766e-05, "loss": 2.386, "step": 75 }, { "epoch": 0.01, "learning_rate": 2.979396349177661e-05, "loss": 3.9266, "step": 76 }, { "epoch": 0.01, "learning_rate": 2.9791252485089463e-05, "loss": 2.8747, "step": 77 }, { "epoch": 0.01, "learning_rate": 2.9788541478402315e-05, "loss": 3.4947, "step": 78 }, { "epoch": 0.01, "learning_rate": 2.9785830471715163e-05, "loss": 2.6419, "step": 79 }, { "epoch": 0.01, "learning_rate": 2.9783119465028015e-05, "loss": 3.3084, "step": 80 }, { "epoch": 0.01, "learning_rate": 2.9780408458340867e-05, "loss": 3.5509, "step": 81 }, { "epoch": 0.01, "learning_rate": 2.9777697451653715e-05, "loss": 3.3718, "step": 82 }, { "epoch": 0.02, "learning_rate": 2.9774986444966567e-05, "loss": 2.629, "step": 83 }, { "epoch": 0.02, "learning_rate": 2.9772275438279415e-05, "loss": 2.9046, "step": 84 }, { "epoch": 0.02, "learning_rate": 2.9769564431592264e-05, "loss": 3.0044, "step": 85 }, { "epoch": 0.02, "learning_rate": 2.9766853424905116e-05, "loss": 2.694, "step": 86 }, { "epoch": 0.02, "learning_rate": 2.9764142418217964e-05, "loss": 2.9554, "step": 87 }, { "epoch": 0.02, "learning_rate": 2.9761431411530816e-05, "loss": 3.4671, "step": 88 }, { "epoch": 0.02, "learning_rate": 2.9758720404843668e-05, "loss": 2.7424, "step": 89 }, { "epoch": 0.02, "learning_rate": 2.9756009398156516e-05, "loss": 2.9786, "step": 90 }, { "epoch": 0.02, "learning_rate": 2.9753298391469368e-05, "loss": 3.1256, "step": 91 }, { "epoch": 0.02, "learning_rate": 2.9750587384782217e-05, "loss": 2.6842, "step": 92 }, { "epoch": 0.02, "learning_rate": 2.9747876378095065e-05, "loss": 2.5909, "step": 93 }, { "epoch": 0.02, "learning_rate": 2.9745165371407917e-05, "loss": 2.9799, "step": 94 }, { "epoch": 0.02, "learning_rate": 2.974245436472077e-05, "loss": 3.1519, "step": 95 }, { "epoch": 0.02, "learning_rate": 2.9739743358033617e-05, "loss": 3.2945, "step": 96 }, { "epoch": 0.02, "learning_rate": 2.973703235134647e-05, "loss": 2.2057, "step": 97 }, { "epoch": 0.02, "learning_rate": 2.9734321344659317e-05, "loss": 2.5681, "step": 98 }, { "epoch": 0.02, "learning_rate": 2.973161033797217e-05, "loss": 2.3096, "step": 99 }, { "epoch": 0.02, "learning_rate": 2.9728899331285018e-05, "loss": 2.7998, "step": 100 }, { "epoch": 0.02, "learning_rate": 2.9726188324597866e-05, "loss": 2.6053, "step": 101 }, { "epoch": 0.02, "learning_rate": 2.9723477317910718e-05, "loss": 3.1633, "step": 102 }, { "epoch": 0.02, "learning_rate": 2.972076631122357e-05, "loss": 2.2774, "step": 103 }, { "epoch": 0.02, "learning_rate": 2.9718055304536418e-05, "loss": 2.1969, "step": 104 }, { "epoch": 0.02, "learning_rate": 2.971534429784927e-05, "loss": 1.9002, "step": 105 }, { "epoch": 0.02, "learning_rate": 2.9712633291162122e-05, "loss": 2.1258, "step": 106 }, { "epoch": 0.02, "learning_rate": 2.970992228447497e-05, "loss": 2.0907, "step": 107 }, { "epoch": 0.02, "learning_rate": 2.970721127778782e-05, "loss": 3.2952, "step": 108 }, { "epoch": 0.02, "learning_rate": 2.9704500271100667e-05, "loss": 2.2971, "step": 109 }, { "epoch": 0.02, "learning_rate": 2.970178926441352e-05, "loss": 1.8442, "step": 110 }, { "epoch": 0.02, "learning_rate": 2.969907825772637e-05, "loss": 3.0905, "step": 111 }, { "epoch": 0.02, "learning_rate": 2.969636725103922e-05, "loss": 2.6187, "step": 112 }, { "epoch": 0.02, "learning_rate": 2.969365624435207e-05, "loss": 1.9357, "step": 113 }, { "epoch": 0.02, "learning_rate": 2.9690945237664923e-05, "loss": 2.9793, "step": 114 }, { "epoch": 0.02, "learning_rate": 2.968823423097777e-05, "loss": 2.9322, "step": 115 }, { "epoch": 0.02, "learning_rate": 2.968552322429062e-05, "loss": 2.4332, "step": 116 }, { "epoch": 0.02, "learning_rate": 2.968281221760347e-05, "loss": 2.7374, "step": 117 }, { "epoch": 0.02, "learning_rate": 2.968010121091632e-05, "loss": 3.457, "step": 118 }, { "epoch": 0.02, "learning_rate": 2.9677390204229172e-05, "loss": 2.7925, "step": 119 }, { "epoch": 0.02, "learning_rate": 2.967467919754202e-05, "loss": 3.1397, "step": 120 }, { "epoch": 0.02, "learning_rate": 2.9671968190854872e-05, "loss": 2.3759, "step": 121 }, { "epoch": 0.02, "learning_rate": 2.9669257184167724e-05, "loss": 2.5331, "step": 122 }, { "epoch": 0.02, "learning_rate": 2.9666546177480572e-05, "loss": 2.2555, "step": 123 }, { "epoch": 0.02, "learning_rate": 2.966383517079342e-05, "loss": 2.9601, "step": 124 }, { "epoch": 0.02, "learning_rate": 2.9661124164106273e-05, "loss": 2.4208, "step": 125 }, { "epoch": 0.02, "learning_rate": 2.965841315741912e-05, "loss": 2.5289, "step": 126 }, { "epoch": 0.02, "learning_rate": 2.9655702150731973e-05, "loss": 2.614, "step": 127 }, { "epoch": 0.02, "learning_rate": 2.9652991144044825e-05, "loss": 3.0021, "step": 128 }, { "epoch": 0.02, "learning_rate": 2.9650280137357673e-05, "loss": 2.6222, "step": 129 }, { "epoch": 0.02, "learning_rate": 2.9647569130670525e-05, "loss": 2.2244, "step": 130 }, { "epoch": 0.02, "learning_rate": 2.964485812398337e-05, "loss": 2.1159, "step": 131 }, { "epoch": 0.02, "learning_rate": 2.9642147117296222e-05, "loss": 2.8683, "step": 132 }, { "epoch": 0.02, "learning_rate": 2.9639436110609074e-05, "loss": 2.086, "step": 133 }, { "epoch": 0.02, "learning_rate": 2.9636725103921922e-05, "loss": 2.3683, "step": 134 }, { "epoch": 0.02, "learning_rate": 2.9634014097234774e-05, "loss": 2.9748, "step": 135 }, { "epoch": 0.02, "learning_rate": 2.9631303090547626e-05, "loss": 3.119, "step": 136 }, { "epoch": 0.02, "learning_rate": 2.9628592083860474e-05, "loss": 2.9806, "step": 137 }, { "epoch": 0.02, "learning_rate": 2.9625881077173326e-05, "loss": 2.4764, "step": 138 }, { "epoch": 0.03, "learning_rate": 2.9623170070486175e-05, "loss": 2.1445, "step": 139 }, { "epoch": 0.03, "learning_rate": 2.9620459063799023e-05, "loss": 2.7098, "step": 140 }, { "epoch": 0.03, "learning_rate": 2.9617748057111875e-05, "loss": 2.6619, "step": 141 }, { "epoch": 0.03, "learning_rate": 2.9615037050424723e-05, "loss": 2.3003, "step": 142 }, { "epoch": 0.03, "learning_rate": 2.9612326043737575e-05, "loss": 2.8239, "step": 143 }, { "epoch": 0.03, "learning_rate": 2.9609615037050427e-05, "loss": 2.0689, "step": 144 }, { "epoch": 0.03, "learning_rate": 2.9606904030363275e-05, "loss": 2.3302, "step": 145 }, { "epoch": 0.03, "learning_rate": 2.9604193023676127e-05, "loss": 2.141, "step": 146 }, { "epoch": 0.03, "learning_rate": 2.9601482016988976e-05, "loss": 2.69, "step": 147 }, { "epoch": 0.03, "learning_rate": 2.9598771010301824e-05, "loss": 2.0252, "step": 148 }, { "epoch": 0.03, "learning_rate": 2.9596060003614676e-05, "loss": 2.4255, "step": 149 }, { "epoch": 0.03, "learning_rate": 2.9593348996927528e-05, "loss": 1.9766, "step": 150 }, { "epoch": 0.03, "learning_rate": 2.9590637990240376e-05, "loss": 3.1633, "step": 151 }, { "epoch": 0.03, "learning_rate": 2.9587926983553228e-05, "loss": 2.6725, "step": 152 }, { "epoch": 0.03, "learning_rate": 2.9585215976866076e-05, "loss": 2.3386, "step": 153 }, { "epoch": 0.03, "learning_rate": 2.9582504970178928e-05, "loss": 2.2246, "step": 154 }, { "epoch": 0.03, "learning_rate": 2.9579793963491777e-05, "loss": 2.4671, "step": 155 }, { "epoch": 0.03, "learning_rate": 2.9577082956804625e-05, "loss": 2.2176, "step": 156 }, { "epoch": 0.03, "learning_rate": 2.9574371950117477e-05, "loss": 2.3291, "step": 157 }, { "epoch": 0.03, "learning_rate": 2.957166094343033e-05, "loss": 2.2117, "step": 158 }, { "epoch": 0.03, "learning_rate": 2.9568949936743177e-05, "loss": 2.4156, "step": 159 }, { "epoch": 0.03, "learning_rate": 2.956623893005603e-05, "loss": 2.6955, "step": 160 }, { "epoch": 0.03, "learning_rate": 2.956352792336888e-05, "loss": 2.2952, "step": 161 }, { "epoch": 0.03, "learning_rate": 2.956081691668173e-05, "loss": 2.639, "step": 162 }, { "epoch": 0.03, "learning_rate": 2.9558105909994578e-05, "loss": 2.5189, "step": 163 }, { "epoch": 0.03, "learning_rate": 2.9555394903307426e-05, "loss": 3.0183, "step": 164 }, { "epoch": 0.03, "learning_rate": 2.9552683896620278e-05, "loss": 1.5289, "step": 165 }, { "epoch": 0.03, "learning_rate": 2.954997288993313e-05, "loss": 2.2809, "step": 166 }, { "epoch": 0.03, "learning_rate": 2.9547261883245978e-05, "loss": 2.1165, "step": 167 }, { "epoch": 0.03, "learning_rate": 2.954455087655883e-05, "loss": 2.4877, "step": 168 }, { "epoch": 0.03, "learning_rate": 2.9541839869871682e-05, "loss": 2.4834, "step": 169 }, { "epoch": 0.03, "learning_rate": 2.953912886318453e-05, "loss": 2.4929, "step": 170 }, { "epoch": 0.03, "learning_rate": 2.953641785649738e-05, "loss": 1.7292, "step": 171 }, { "epoch": 0.03, "learning_rate": 2.953370684981023e-05, "loss": 2.0687, "step": 172 }, { "epoch": 0.03, "learning_rate": 2.953099584312308e-05, "loss": 1.7756, "step": 173 }, { "epoch": 0.03, "learning_rate": 2.952828483643593e-05, "loss": 1.9391, "step": 174 }, { "epoch": 0.03, "learning_rate": 2.952557382974878e-05, "loss": 2.465, "step": 175 }, { "epoch": 0.03, "learning_rate": 2.952286282306163e-05, "loss": 2.8734, "step": 176 }, { "epoch": 0.03, "learning_rate": 2.9520151816374483e-05, "loss": 2.5771, "step": 177 }, { "epoch": 0.03, "learning_rate": 2.951744080968733e-05, "loss": 1.8889, "step": 178 }, { "epoch": 0.03, "learning_rate": 2.951472980300018e-05, "loss": 2.1676, "step": 179 }, { "epoch": 0.03, "learning_rate": 2.9512018796313032e-05, "loss": 3.1156, "step": 180 }, { "epoch": 0.03, "learning_rate": 2.950930778962588e-05, "loss": 1.7913, "step": 181 }, { "epoch": 0.03, "learning_rate": 2.9506596782938732e-05, "loss": 2.6479, "step": 182 }, { "epoch": 0.03, "learning_rate": 2.9503885776251584e-05, "loss": 2.4518, "step": 183 }, { "epoch": 0.03, "learning_rate": 2.9501174769564432e-05, "loss": 2.0942, "step": 184 }, { "epoch": 0.03, "learning_rate": 2.9498463762877284e-05, "loss": 2.4306, "step": 185 }, { "epoch": 0.03, "learning_rate": 2.9495752756190133e-05, "loss": 2.5361, "step": 186 }, { "epoch": 0.03, "learning_rate": 2.949304174950298e-05, "loss": 2.5573, "step": 187 }, { "epoch": 0.03, "learning_rate": 2.9490330742815833e-05, "loss": 2.4955, "step": 188 }, { "epoch": 0.03, "learning_rate": 2.948761973612868e-05, "loss": 2.7852, "step": 189 }, { "epoch": 0.03, "learning_rate": 2.9484908729441533e-05, "loss": 2.1162, "step": 190 }, { "epoch": 0.03, "learning_rate": 2.9482197722754385e-05, "loss": 2.3774, "step": 191 }, { "epoch": 0.03, "learning_rate": 2.9479486716067233e-05, "loss": 2.148, "step": 192 }, { "epoch": 0.03, "learning_rate": 2.9476775709380085e-05, "loss": 2.5626, "step": 193 }, { "epoch": 0.04, "learning_rate": 2.9474064702692937e-05, "loss": 2.4154, "step": 194 }, { "epoch": 0.04, "learning_rate": 2.9471353696005782e-05, "loss": 2.1039, "step": 195 }, { "epoch": 0.04, "learning_rate": 2.9468642689318634e-05, "loss": 2.3193, "step": 196 }, { "epoch": 0.04, "learning_rate": 2.9465931682631482e-05, "loss": 2.2391, "step": 197 }, { "epoch": 0.04, "learning_rate": 2.9463220675944334e-05, "loss": 2.0979, "step": 198 }, { "epoch": 0.04, "learning_rate": 2.9460509669257186e-05, "loss": 2.2808, "step": 199 }, { "epoch": 0.04, "learning_rate": 2.9457798662570034e-05, "loss": 2.659, "step": 200 }, { "epoch": 0.04, "learning_rate": 2.9455087655882886e-05, "loss": 2.3003, "step": 201 }, { "epoch": 0.04, "learning_rate": 2.9452376649195738e-05, "loss": 2.4122, "step": 202 }, { "epoch": 0.04, "learning_rate": 2.9449665642508583e-05, "loss": 2.6408, "step": 203 }, { "epoch": 0.04, "learning_rate": 2.9446954635821435e-05, "loss": 2.2382, "step": 204 }, { "epoch": 0.04, "learning_rate": 2.9444243629134287e-05, "loss": 1.7609, "step": 205 }, { "epoch": 0.04, "learning_rate": 2.9441532622447135e-05, "loss": 2.5257, "step": 206 }, { "epoch": 0.04, "learning_rate": 2.9438821615759987e-05, "loss": 2.5992, "step": 207 }, { "epoch": 0.04, "learning_rate": 2.9436110609072835e-05, "loss": 2.3529, "step": 208 }, { "epoch": 0.04, "learning_rate": 2.9433399602385687e-05, "loss": 2.1197, "step": 209 }, { "epoch": 0.04, "learning_rate": 2.943068859569854e-05, "loss": 2.0271, "step": 210 }, { "epoch": 0.04, "learning_rate": 2.9427977589011384e-05, "loss": 2.099, "step": 211 }, { "epoch": 0.04, "learning_rate": 2.9425266582324236e-05, "loss": 2.555, "step": 212 }, { "epoch": 0.04, "learning_rate": 2.9422555575637088e-05, "loss": 2.0059, "step": 213 }, { "epoch": 0.04, "learning_rate": 2.9419844568949936e-05, "loss": 2.146, "step": 214 }, { "epoch": 0.04, "learning_rate": 2.9417133562262788e-05, "loss": 1.9306, "step": 215 }, { "epoch": 0.04, "learning_rate": 2.941442255557564e-05, "loss": 2.3847, "step": 216 }, { "epoch": 0.04, "learning_rate": 2.941171154888849e-05, "loss": 2.7286, "step": 217 }, { "epoch": 0.04, "learning_rate": 2.940900054220134e-05, "loss": 2.1696, "step": 218 }, { "epoch": 0.04, "learning_rate": 2.940628953551419e-05, "loss": 2.1023, "step": 219 }, { "epoch": 0.04, "learning_rate": 2.9403578528827037e-05, "loss": 1.9038, "step": 220 }, { "epoch": 0.04, "learning_rate": 2.940086752213989e-05, "loss": 2.4279, "step": 221 }, { "epoch": 0.04, "learning_rate": 2.9398156515452737e-05, "loss": 1.7805, "step": 222 }, { "epoch": 0.04, "learning_rate": 2.939544550876559e-05, "loss": 2.851, "step": 223 }, { "epoch": 0.04, "learning_rate": 2.939273450207844e-05, "loss": 1.854, "step": 224 }, { "epoch": 0.04, "learning_rate": 2.939002349539129e-05, "loss": 2.999, "step": 225 }, { "epoch": 0.04, "learning_rate": 2.938731248870414e-05, "loss": 2.2498, "step": 226 }, { "epoch": 0.04, "learning_rate": 2.938460148201699e-05, "loss": 1.4847, "step": 227 }, { "epoch": 0.04, "learning_rate": 2.9381890475329838e-05, "loss": 1.7983, "step": 228 }, { "epoch": 0.04, "learning_rate": 2.937917946864269e-05, "loss": 1.2853, "step": 229 }, { "epoch": 0.04, "learning_rate": 2.9376468461955542e-05, "loss": 1.609, "step": 230 }, { "epoch": 0.04, "learning_rate": 2.937375745526839e-05, "loss": 1.8714, "step": 231 }, { "epoch": 0.04, "learning_rate": 2.9371046448581242e-05, "loss": 1.6379, "step": 232 }, { "epoch": 0.04, "learning_rate": 2.936833544189409e-05, "loss": 2.4026, "step": 233 }, { "epoch": 0.04, "learning_rate": 2.9365624435206942e-05, "loss": 2.8235, "step": 234 }, { "epoch": 0.04, "learning_rate": 2.936291342851979e-05, "loss": 2.344, "step": 235 }, { "epoch": 0.04, "learning_rate": 2.936020242183264e-05, "loss": 2.1571, "step": 236 }, { "epoch": 0.04, "learning_rate": 2.935749141514549e-05, "loss": 2.164, "step": 237 }, { "epoch": 0.04, "learning_rate": 2.9354780408458343e-05, "loss": 2.3228, "step": 238 }, { "epoch": 0.04, "learning_rate": 2.935206940177119e-05, "loss": 1.7097, "step": 239 }, { "epoch": 0.04, "learning_rate": 2.9349358395084043e-05, "loss": 2.2812, "step": 240 }, { "epoch": 0.04, "learning_rate": 2.9346647388396895e-05, "loss": 2.5421, "step": 241 }, { "epoch": 0.04, "learning_rate": 2.9343936381709743e-05, "loss": 2.3694, "step": 242 }, { "epoch": 0.04, "learning_rate": 2.9341225375022592e-05, "loss": 2.2034, "step": 243 }, { "epoch": 0.04, "learning_rate": 2.933851436833544e-05, "loss": 2.028, "step": 244 }, { "epoch": 0.04, "learning_rate": 2.9335803361648292e-05, "loss": 1.5084, "step": 245 }, { "epoch": 0.04, "learning_rate": 2.9333092354961144e-05, "loss": 1.9911, "step": 246 }, { "epoch": 0.04, "learning_rate": 2.9330381348273992e-05, "loss": 2.4768, "step": 247 }, { "epoch": 0.04, "learning_rate": 2.9327670341586844e-05, "loss": 1.7299, "step": 248 }, { "epoch": 0.05, "learning_rate": 2.9324959334899696e-05, "loss": 2.146, "step": 249 }, { "epoch": 0.05, "learning_rate": 2.9322248328212544e-05, "loss": 2.3871, "step": 250 }, { "epoch": 0.05, "learning_rate": 2.9319537321525393e-05, "loss": 2.2222, "step": 251 }, { "epoch": 0.05, "learning_rate": 2.9316826314838245e-05, "loss": 2.1465, "step": 252 }, { "epoch": 0.05, "learning_rate": 2.9314115308151093e-05, "loss": 1.4889, "step": 253 }, { "epoch": 0.05, "learning_rate": 2.9311404301463945e-05, "loss": 2.7812, "step": 254 }, { "epoch": 0.05, "learning_rate": 2.9308693294776793e-05, "loss": 2.8301, "step": 255 }, { "epoch": 0.05, "learning_rate": 2.9305982288089645e-05, "loss": 2.0572, "step": 256 }, { "epoch": 0.05, "learning_rate": 2.9303271281402497e-05, "loss": 1.7446, "step": 257 }, { "epoch": 0.05, "learning_rate": 2.9300560274715342e-05, "loss": 2.1099, "step": 258 }, { "epoch": 0.05, "learning_rate": 2.9297849268028194e-05, "loss": 2.8554, "step": 259 }, { "epoch": 0.05, "learning_rate": 2.9295138261341046e-05, "loss": 2.2578, "step": 260 }, { "epoch": 0.05, "learning_rate": 2.9292427254653894e-05, "loss": 2.6806, "step": 261 }, { "epoch": 0.05, "learning_rate": 2.9289716247966746e-05, "loss": 1.7889, "step": 262 }, { "epoch": 0.05, "learning_rate": 2.9287005241279598e-05, "loss": 1.7681, "step": 263 }, { "epoch": 0.05, "learning_rate": 2.9284294234592446e-05, "loss": 1.5942, "step": 264 }, { "epoch": 0.05, "learning_rate": 2.9281583227905298e-05, "loss": 2.119, "step": 265 }, { "epoch": 0.05, "learning_rate": 2.9278872221218143e-05, "loss": 2.3912, "step": 266 }, { "epoch": 0.05, "learning_rate": 2.9276161214530995e-05, "loss": 2.0893, "step": 267 }, { "epoch": 0.05, "learning_rate": 2.9273450207843847e-05, "loss": 2.9637, "step": 268 }, { "epoch": 0.05, "learning_rate": 2.9270739201156695e-05, "loss": 2.2754, "step": 269 }, { "epoch": 0.05, "learning_rate": 2.9268028194469547e-05, "loss": 1.6581, "step": 270 }, { "epoch": 0.05, "learning_rate": 2.92653171877824e-05, "loss": 2.1254, "step": 271 }, { "epoch": 0.05, "learning_rate": 2.9262606181095247e-05, "loss": 2.1677, "step": 272 }, { "epoch": 0.05, "learning_rate": 2.92598951744081e-05, "loss": 2.2016, "step": 273 }, { "epoch": 0.05, "learning_rate": 2.9257184167720948e-05, "loss": 1.4915, "step": 274 }, { "epoch": 0.05, "learning_rate": 2.9254473161033796e-05, "loss": 1.3906, "step": 275 }, { "epoch": 0.05, "learning_rate": 2.9251762154346648e-05, "loss": 1.9389, "step": 276 }, { "epoch": 0.05, "learning_rate": 2.9249051147659496e-05, "loss": 1.6062, "step": 277 }, { "epoch": 0.05, "learning_rate": 2.9246340140972348e-05, "loss": 1.5853, "step": 278 }, { "epoch": 0.05, "learning_rate": 2.92436291342852e-05, "loss": 2.0449, "step": 279 }, { "epoch": 0.05, "learning_rate": 2.924091812759805e-05, "loss": 1.8977, "step": 280 }, { "epoch": 0.05, "learning_rate": 2.92382071209109e-05, "loss": 1.6778, "step": 281 }, { "epoch": 0.05, "learning_rate": 2.923549611422375e-05, "loss": 1.701, "step": 282 }, { "epoch": 0.05, "learning_rate": 2.9232785107536597e-05, "loss": 1.5615, "step": 283 }, { "epoch": 0.05, "learning_rate": 2.923007410084945e-05, "loss": 1.7431, "step": 284 }, { "epoch": 0.05, "learning_rate": 2.92273630941623e-05, "loss": 2.1237, "step": 285 }, { "epoch": 0.05, "learning_rate": 2.922465208747515e-05, "loss": 2.4777, "step": 286 }, { "epoch": 0.05, "learning_rate": 2.9221941080788e-05, "loss": 1.7966, "step": 287 }, { "epoch": 0.05, "learning_rate": 2.921923007410085e-05, "loss": 2.1663, "step": 288 }, { "epoch": 0.05, "learning_rate": 2.92165190674137e-05, "loss": 2.2379, "step": 289 }, { "epoch": 0.05, "learning_rate": 2.921380806072655e-05, "loss": 2.0654, "step": 290 }, { "epoch": 0.05, "learning_rate": 2.9211097054039398e-05, "loss": 2.3468, "step": 291 }, { "epoch": 0.05, "learning_rate": 2.920838604735225e-05, "loss": 2.0488, "step": 292 }, { "epoch": 0.05, "learning_rate": 2.9205675040665102e-05, "loss": 1.6951, "step": 293 }, { "epoch": 0.05, "learning_rate": 2.920296403397795e-05, "loss": 2.3308, "step": 294 }, { "epoch": 0.05, "learning_rate": 2.9200253027290802e-05, "loss": 1.1227, "step": 295 }, { "epoch": 0.05, "learning_rate": 2.9197542020603654e-05, "loss": 2.2433, "step": 296 }, { "epoch": 0.05, "learning_rate": 2.9194831013916502e-05, "loss": 1.2546, "step": 297 }, { "epoch": 0.05, "learning_rate": 2.919212000722935e-05, "loss": 1.9151, "step": 298 }, { "epoch": 0.05, "learning_rate": 2.91894090005422e-05, "loss": 1.8822, "step": 299 }, { "epoch": 0.05, "learning_rate": 2.918669799385505e-05, "loss": 1.9898, "step": 300 }, { "epoch": 0.05, "learning_rate": 2.9183986987167903e-05, "loss": 1.8095, "step": 301 }, { "epoch": 0.05, "learning_rate": 2.918127598048075e-05, "loss": 2.2742, "step": 302 }, { "epoch": 0.05, "learning_rate": 2.9178564973793603e-05, "loss": 2.5536, "step": 303 }, { "epoch": 0.05, "learning_rate": 2.9175853967106455e-05, "loss": 1.6182, "step": 304 }, { "epoch": 0.06, "learning_rate": 2.9173142960419304e-05, "loss": 2.1913, "step": 305 }, { "epoch": 0.06, "learning_rate": 2.9170431953732152e-05, "loss": 1.7617, "step": 306 }, { "epoch": 0.06, "learning_rate": 2.9167720947045004e-05, "loss": 2.4153, "step": 307 }, { "epoch": 0.06, "learning_rate": 2.9165009940357852e-05, "loss": 1.8653, "step": 308 }, { "epoch": 0.06, "learning_rate": 2.9162298933670704e-05, "loss": 1.9201, "step": 309 }, { "epoch": 0.06, "learning_rate": 2.9159587926983553e-05, "loss": 2.4575, "step": 310 }, { "epoch": 0.06, "learning_rate": 2.9156876920296404e-05, "loss": 1.5466, "step": 311 }, { "epoch": 0.06, "learning_rate": 2.9154165913609256e-05, "loss": 2.294, "step": 312 }, { "epoch": 0.06, "learning_rate": 2.9151454906922105e-05, "loss": 1.7047, "step": 313 }, { "epoch": 0.06, "learning_rate": 2.9148743900234953e-05, "loss": 1.9904, "step": 314 }, { "epoch": 0.06, "learning_rate": 2.9146032893547805e-05, "loss": 1.8362, "step": 315 }, { "epoch": 0.06, "learning_rate": 2.9143321886860653e-05, "loss": 2.3346, "step": 316 }, { "epoch": 0.06, "learning_rate": 2.9140610880173505e-05, "loss": 1.1031, "step": 317 }, { "epoch": 0.06, "learning_rate": 2.9137899873486357e-05, "loss": 1.9463, "step": 318 }, { "epoch": 0.06, "learning_rate": 2.9135188866799205e-05, "loss": 1.5553, "step": 319 }, { "epoch": 0.06, "learning_rate": 2.9132477860112057e-05, "loss": 1.2304, "step": 320 }, { "epoch": 0.06, "learning_rate": 2.9129766853424906e-05, "loss": 1.8553, "step": 321 }, { "epoch": 0.06, "learning_rate": 2.9127055846737754e-05, "loss": 1.5508, "step": 322 }, { "epoch": 0.06, "learning_rate": 2.9124344840050606e-05, "loss": 1.7433, "step": 323 }, { "epoch": 0.06, "learning_rate": 2.9121633833363454e-05, "loss": 2.0361, "step": 324 }, { "epoch": 0.06, "learning_rate": 2.9118922826676306e-05, "loss": 1.5907, "step": 325 }, { "epoch": 0.06, "learning_rate": 2.9116211819989158e-05, "loss": 1.7085, "step": 326 }, { "epoch": 0.06, "learning_rate": 2.9113500813302007e-05, "loss": 1.4822, "step": 327 }, { "epoch": 0.06, "learning_rate": 2.911078980661486e-05, "loss": 1.7785, "step": 328 }, { "epoch": 0.06, "learning_rate": 2.910807879992771e-05, "loss": 1.397, "step": 329 }, { "epoch": 0.06, "learning_rate": 2.9105367793240555e-05, "loss": 1.6978, "step": 330 }, { "epoch": 0.06, "learning_rate": 2.9102656786553407e-05, "loss": 2.1525, "step": 331 }, { "epoch": 0.06, "learning_rate": 2.9099945779866256e-05, "loss": 1.7079, "step": 332 }, { "epoch": 0.06, "learning_rate": 2.9097234773179107e-05, "loss": 1.6592, "step": 333 }, { "epoch": 0.06, "learning_rate": 2.909452376649196e-05, "loss": 1.8333, "step": 334 }, { "epoch": 0.06, "learning_rate": 2.9091812759804808e-05, "loss": 2.0608, "step": 335 }, { "epoch": 0.06, "learning_rate": 2.908910175311766e-05, "loss": 1.3938, "step": 336 }, { "epoch": 0.06, "learning_rate": 2.908639074643051e-05, "loss": 1.8675, "step": 337 }, { "epoch": 0.06, "learning_rate": 2.9083679739743356e-05, "loss": 1.9688, "step": 338 }, { "epoch": 0.06, "learning_rate": 2.9080968733056208e-05, "loss": 2.0622, "step": 339 }, { "epoch": 0.06, "learning_rate": 2.907825772636906e-05, "loss": 2.5251, "step": 340 }, { "epoch": 0.06, "learning_rate": 2.907554671968191e-05, "loss": 1.8319, "step": 341 }, { "epoch": 0.06, "learning_rate": 2.907283571299476e-05, "loss": 2.5061, "step": 342 }, { "epoch": 0.06, "learning_rate": 2.907012470630761e-05, "loss": 2.1739, "step": 343 }, { "epoch": 0.06, "learning_rate": 2.906741369962046e-05, "loss": 2.056, "step": 344 }, { "epoch": 0.06, "learning_rate": 2.9064702692933312e-05, "loss": 1.8152, "step": 345 }, { "epoch": 0.06, "learning_rate": 2.9061991686246157e-05, "loss": 1.326, "step": 346 }, { "epoch": 0.06, "learning_rate": 2.905928067955901e-05, "loss": 1.2062, "step": 347 }, { "epoch": 0.06, "learning_rate": 2.905656967287186e-05, "loss": 2.1967, "step": 348 }, { "epoch": 0.06, "learning_rate": 2.905385866618471e-05, "loss": 1.3308, "step": 349 }, { "epoch": 0.06, "learning_rate": 2.905114765949756e-05, "loss": 1.5806, "step": 350 }, { "epoch": 0.06, "learning_rate": 2.9048436652810413e-05, "loss": 2.0727, "step": 351 }, { "epoch": 0.06, "learning_rate": 2.904572564612326e-05, "loss": 1.5464, "step": 352 }, { "epoch": 0.06, "learning_rate": 2.9043014639436113e-05, "loss": 1.4285, "step": 353 }, { "epoch": 0.06, "learning_rate": 2.9040303632748962e-05, "loss": 1.7527, "step": 354 }, { "epoch": 0.06, "learning_rate": 2.903759262606181e-05, "loss": 1.3718, "step": 355 }, { "epoch": 0.06, "learning_rate": 2.9034881619374662e-05, "loss": 1.4609, "step": 356 }, { "epoch": 0.06, "learning_rate": 2.903217061268751e-05, "loss": 1.8429, "step": 357 }, { "epoch": 0.06, "learning_rate": 2.9029459606000362e-05, "loss": 1.5308, "step": 358 }, { "epoch": 0.06, "learning_rate": 2.9026748599313214e-05, "loss": 1.9681, "step": 359 }, { "epoch": 0.07, "learning_rate": 2.9024037592626063e-05, "loss": 2.3457, "step": 360 }, { "epoch": 0.07, "learning_rate": 2.9021326585938914e-05, "loss": 2.3322, "step": 361 }, { "epoch": 0.07, "learning_rate": 2.9018615579251763e-05, "loss": 2.255, "step": 362 }, { "epoch": 0.07, "learning_rate": 2.901590457256461e-05, "loss": 2.3572, "step": 363 }, { "epoch": 0.07, "learning_rate": 2.9013193565877463e-05, "loss": 2.0241, "step": 364 }, { "epoch": 0.07, "learning_rate": 2.9010482559190315e-05, "loss": 1.5013, "step": 365 }, { "epoch": 0.07, "learning_rate": 2.9007771552503163e-05, "loss": 1.7231, "step": 366 }, { "epoch": 0.07, "learning_rate": 2.9005060545816015e-05, "loss": 1.5714, "step": 367 }, { "epoch": 0.07, "learning_rate": 2.9002349539128864e-05, "loss": 2.3282, "step": 368 }, { "epoch": 0.07, "learning_rate": 2.8999638532441716e-05, "loss": 1.3672, "step": 369 }, { "epoch": 0.07, "learning_rate": 2.8996927525754564e-05, "loss": 2.4871, "step": 370 }, { "epoch": 0.07, "learning_rate": 2.8994216519067412e-05, "loss": 1.6951, "step": 371 }, { "epoch": 0.07, "learning_rate": 2.8991505512380264e-05, "loss": 1.8999, "step": 372 }, { "epoch": 0.07, "learning_rate": 2.8988794505693116e-05, "loss": 1.9182, "step": 373 }, { "epoch": 0.07, "learning_rate": 2.8986083499005965e-05, "loss": 1.81, "step": 374 }, { "epoch": 0.07, "learning_rate": 2.8983372492318816e-05, "loss": 1.4093, "step": 375 }, { "epoch": 0.07, "learning_rate": 2.8980661485631668e-05, "loss": 1.9885, "step": 376 }, { "epoch": 0.07, "learning_rate": 2.8977950478944517e-05, "loss": 1.758, "step": 377 }, { "epoch": 0.07, "learning_rate": 2.8975239472257365e-05, "loss": 1.2293, "step": 378 }, { "epoch": 0.07, "learning_rate": 2.8972528465570214e-05, "loss": 1.6351, "step": 379 }, { "epoch": 0.07, "learning_rate": 2.8969817458883065e-05, "loss": 1.2605, "step": 380 }, { "epoch": 0.07, "learning_rate": 2.8967106452195917e-05, "loss": 1.4166, "step": 381 }, { "epoch": 0.07, "learning_rate": 2.8964395445508766e-05, "loss": 1.7398, "step": 382 }, { "epoch": 0.07, "learning_rate": 2.8961684438821617e-05, "loss": 1.7061, "step": 383 }, { "epoch": 0.07, "learning_rate": 2.895897343213447e-05, "loss": 2.4242, "step": 384 }, { "epoch": 0.07, "learning_rate": 2.8956262425447318e-05, "loss": 1.5752, "step": 385 }, { "epoch": 0.07, "learning_rate": 2.8953551418760166e-05, "loss": 1.8631, "step": 386 }, { "epoch": 0.07, "learning_rate": 2.8950840412073018e-05, "loss": 2.061, "step": 387 }, { "epoch": 0.07, "learning_rate": 2.8948129405385866e-05, "loss": 1.7565, "step": 388 }, { "epoch": 0.07, "learning_rate": 2.8945418398698718e-05, "loss": 1.0947, "step": 389 }, { "epoch": 0.07, "learning_rate": 2.8942707392011567e-05, "loss": 1.8912, "step": 390 }, { "epoch": 0.07, "learning_rate": 2.893999638532442e-05, "loss": 1.8797, "step": 391 }, { "epoch": 0.07, "learning_rate": 2.893728537863727e-05, "loss": 1.5819, "step": 392 }, { "epoch": 0.07, "learning_rate": 2.8934574371950115e-05, "loss": 2.0879, "step": 393 }, { "epoch": 0.07, "learning_rate": 2.8931863365262967e-05, "loss": 1.4488, "step": 394 }, { "epoch": 0.07, "learning_rate": 2.892915235857582e-05, "loss": 1.4853, "step": 395 }, { "epoch": 0.07, "learning_rate": 2.8926441351888667e-05, "loss": 1.2138, "step": 396 }, { "epoch": 0.07, "learning_rate": 2.892373034520152e-05, "loss": 1.1239, "step": 397 }, { "epoch": 0.07, "learning_rate": 2.892101933851437e-05, "loss": 1.6596, "step": 398 }, { "epoch": 0.07, "learning_rate": 2.891830833182722e-05, "loss": 1.1356, "step": 399 }, { "epoch": 0.07, "learning_rate": 2.891559732514007e-05, "loss": 1.6265, "step": 400 }, { "epoch": 0.07, "learning_rate": 2.8912886318452916e-05, "loss": 1.2724, "step": 401 }, { "epoch": 0.07, "learning_rate": 2.8910175311765768e-05, "loss": 1.2583, "step": 402 }, { "epoch": 0.07, "learning_rate": 2.890746430507862e-05, "loss": 0.9703, "step": 403 }, { "epoch": 0.07, "learning_rate": 2.890475329839147e-05, "loss": 1.8156, "step": 404 }, { "epoch": 0.07, "learning_rate": 2.890204229170432e-05, "loss": 1.693, "step": 405 }, { "epoch": 0.07, "learning_rate": 2.8899331285017172e-05, "loss": 1.7402, "step": 406 }, { "epoch": 0.07, "learning_rate": 2.889662027833002e-05, "loss": 1.2778, "step": 407 }, { "epoch": 0.07, "learning_rate": 2.8893909271642872e-05, "loss": 1.2939, "step": 408 }, { "epoch": 0.07, "learning_rate": 2.889119826495572e-05, "loss": 2.0742, "step": 409 }, { "epoch": 0.07, "learning_rate": 2.888848725826857e-05, "loss": 1.941, "step": 410 }, { "epoch": 0.07, "learning_rate": 2.888577625158142e-05, "loss": 1.2712, "step": 411 }, { "epoch": 0.07, "learning_rate": 2.888306524489427e-05, "loss": 2.4306, "step": 412 }, { "epoch": 0.07, "learning_rate": 2.888035423820712e-05, "loss": 1.7812, "step": 413 }, { "epoch": 0.07, "learning_rate": 2.8877643231519973e-05, "loss": 1.4988, "step": 414 }, { "epoch": 0.08, "learning_rate": 2.8874932224832822e-05, "loss": 1.4322, "step": 415 }, { "epoch": 0.08, "learning_rate": 2.8872221218145674e-05, "loss": 1.5044, "step": 416 }, { "epoch": 0.08, "learning_rate": 2.8869510211458522e-05, "loss": 2.2452, "step": 417 }, { "epoch": 0.08, "learning_rate": 2.886679920477137e-05, "loss": 1.9538, "step": 418 }, { "epoch": 0.08, "learning_rate": 2.8864088198084222e-05, "loss": 1.0781, "step": 419 }, { "epoch": 0.08, "learning_rate": 2.8861377191397074e-05, "loss": 1.5358, "step": 420 }, { "epoch": 0.08, "learning_rate": 2.8858666184709923e-05, "loss": 1.9906, "step": 421 }, { "epoch": 0.08, "learning_rate": 2.8855955178022774e-05, "loss": 1.7883, "step": 422 }, { "epoch": 0.08, "learning_rate": 2.8853244171335623e-05, "loss": 1.5432, "step": 423 }, { "epoch": 0.08, "learning_rate": 2.8850533164648475e-05, "loss": 1.7593, "step": 424 }, { "epoch": 0.08, "learning_rate": 2.8847822157961323e-05, "loss": 1.4443, "step": 425 }, { "epoch": 0.08, "learning_rate": 2.884511115127417e-05, "loss": 1.2129, "step": 426 }, { "epoch": 0.08, "learning_rate": 2.8842400144587023e-05, "loss": 1.4344, "step": 427 }, { "epoch": 0.08, "learning_rate": 2.8839689137899875e-05, "loss": 2.0925, "step": 428 }, { "epoch": 0.08, "learning_rate": 2.8836978131212724e-05, "loss": 1.3983, "step": 429 }, { "epoch": 0.08, "learning_rate": 2.8834267124525575e-05, "loss": 2.1201, "step": 430 }, { "epoch": 0.08, "learning_rate": 2.8831556117838427e-05, "loss": 1.7714, "step": 431 }, { "epoch": 0.08, "learning_rate": 2.8828845111151276e-05, "loss": 2.0383, "step": 432 }, { "epoch": 0.08, "learning_rate": 2.8826134104464124e-05, "loss": 1.355, "step": 433 }, { "epoch": 0.08, "learning_rate": 2.8823423097776973e-05, "loss": 1.3772, "step": 434 }, { "epoch": 0.08, "learning_rate": 2.8820712091089824e-05, "loss": 1.3664, "step": 435 }, { "epoch": 0.08, "learning_rate": 2.8818001084402676e-05, "loss": 1.3119, "step": 436 }, { "epoch": 0.08, "learning_rate": 2.8815290077715525e-05, "loss": 2.7856, "step": 437 }, { "epoch": 0.08, "learning_rate": 2.8812579071028377e-05, "loss": 2.0245, "step": 438 }, { "epoch": 0.08, "learning_rate": 2.880986806434123e-05, "loss": 2.2244, "step": 439 }, { "epoch": 0.08, "learning_rate": 2.8807157057654077e-05, "loss": 1.4509, "step": 440 }, { "epoch": 0.08, "learning_rate": 2.8804446050966925e-05, "loss": 1.6023, "step": 441 }, { "epoch": 0.08, "learning_rate": 2.8801735044279777e-05, "loss": 1.123, "step": 442 }, { "epoch": 0.08, "learning_rate": 2.8799024037592625e-05, "loss": 1.7081, "step": 443 }, { "epoch": 0.08, "learning_rate": 2.8796313030905477e-05, "loss": 1.4622, "step": 444 }, { "epoch": 0.08, "learning_rate": 2.8793602024218326e-05, "loss": 1.9179, "step": 445 }, { "epoch": 0.08, "learning_rate": 2.8790891017531178e-05, "loss": 1.2908, "step": 446 }, { "epoch": 0.08, "learning_rate": 2.878818001084403e-05, "loss": 1.2704, "step": 447 }, { "epoch": 0.08, "learning_rate": 2.8785469004156878e-05, "loss": 1.8802, "step": 448 }, { "epoch": 0.08, "learning_rate": 2.8782757997469726e-05, "loss": 2.3381, "step": 449 }, { "epoch": 0.08, "learning_rate": 2.8780046990782578e-05, "loss": 1.3237, "step": 450 }, { "epoch": 0.08, "learning_rate": 2.8777335984095427e-05, "loss": 1.5885, "step": 451 }, { "epoch": 0.08, "learning_rate": 2.877462497740828e-05, "loss": 1.82, "step": 452 }, { "epoch": 0.08, "learning_rate": 2.877191397072113e-05, "loss": 1.8053, "step": 453 }, { "epoch": 0.08, "learning_rate": 2.876920296403398e-05, "loss": 1.1485, "step": 454 }, { "epoch": 0.08, "learning_rate": 2.876649195734683e-05, "loss": 1.7767, "step": 455 }, { "epoch": 0.08, "learning_rate": 2.876378095065968e-05, "loss": 1.5774, "step": 456 }, { "epoch": 0.08, "learning_rate": 2.8761069943972527e-05, "loss": 2.1925, "step": 457 }, { "epoch": 0.08, "learning_rate": 2.875835893728538e-05, "loss": 2.3281, "step": 458 }, { "epoch": 0.08, "learning_rate": 2.8755647930598228e-05, "loss": 1.8536, "step": 459 }, { "epoch": 0.08, "learning_rate": 2.875293692391108e-05, "loss": 1.7365, "step": 460 }, { "epoch": 0.08, "learning_rate": 2.875022591722393e-05, "loss": 1.8846, "step": 461 }, { "epoch": 0.08, "learning_rate": 2.874751491053678e-05, "loss": 1.3164, "step": 462 }, { "epoch": 0.08, "learning_rate": 2.874480390384963e-05, "loss": 1.8203, "step": 463 }, { "epoch": 0.08, "learning_rate": 2.8742092897162483e-05, "loss": 1.635, "step": 464 }, { "epoch": 0.08, "learning_rate": 2.873938189047533e-05, "loss": 2.1423, "step": 465 }, { "epoch": 0.08, "learning_rate": 2.873667088378818e-05, "loss": 1.8326, "step": 466 }, { "epoch": 0.08, "learning_rate": 2.873395987710103e-05, "loss": 1.98, "step": 467 }, { "epoch": 0.08, "learning_rate": 2.873124887041388e-05, "loss": 1.2633, "step": 468 }, { "epoch": 0.08, "learning_rate": 2.8728537863726732e-05, "loss": 1.5057, "step": 469 }, { "epoch": 0.08, "learning_rate": 2.872582685703958e-05, "loss": 1.2216, "step": 470 }, { "epoch": 0.09, "learning_rate": 2.8723115850352433e-05, "loss": 1.9135, "step": 471 }, { "epoch": 0.09, "learning_rate": 2.8720404843665284e-05, "loss": 1.1969, "step": 472 }, { "epoch": 0.09, "learning_rate": 2.871769383697813e-05, "loss": 1.1371, "step": 473 }, { "epoch": 0.09, "learning_rate": 2.871498283029098e-05, "loss": 2.0456, "step": 474 }, { "epoch": 0.09, "learning_rate": 2.8712271823603833e-05, "loss": 1.6056, "step": 475 }, { "epoch": 0.09, "learning_rate": 2.870956081691668e-05, "loss": 1.1252, "step": 476 }, { "epoch": 0.09, "learning_rate": 2.8706849810229533e-05, "loss": 1.7922, "step": 477 }, { "epoch": 0.09, "learning_rate": 2.8704138803542382e-05, "loss": 1.5013, "step": 478 }, { "epoch": 0.09, "learning_rate": 2.8701427796855234e-05, "loss": 1.5741, "step": 479 }, { "epoch": 0.09, "learning_rate": 2.8698716790168086e-05, "loss": 0.9703, "step": 480 }, { "epoch": 0.09, "learning_rate": 2.869600578348093e-05, "loss": 1.7088, "step": 481 }, { "epoch": 0.09, "learning_rate": 2.8693294776793782e-05, "loss": 1.6732, "step": 482 }, { "epoch": 0.09, "learning_rate": 2.8690583770106634e-05, "loss": 1.526, "step": 483 }, { "epoch": 0.09, "learning_rate": 2.8687872763419483e-05, "loss": 2.6089, "step": 484 }, { "epoch": 0.09, "learning_rate": 2.8685161756732335e-05, "loss": 1.694, "step": 485 }, { "epoch": 0.09, "learning_rate": 2.8682450750045186e-05, "loss": 1.6834, "step": 486 }, { "epoch": 0.09, "learning_rate": 2.8679739743358035e-05, "loss": 1.3771, "step": 487 }, { "epoch": 0.09, "learning_rate": 2.8677028736670887e-05, "loss": 0.9727, "step": 488 }, { "epoch": 0.09, "learning_rate": 2.867431772998373e-05, "loss": 1.3986, "step": 489 }, { "epoch": 0.09, "learning_rate": 2.8671606723296583e-05, "loss": 1.7924, "step": 490 }, { "epoch": 0.09, "learning_rate": 2.8668895716609435e-05, "loss": 0.802, "step": 491 }, { "epoch": 0.09, "learning_rate": 2.8666184709922284e-05, "loss": 1.4909, "step": 492 }, { "epoch": 0.09, "learning_rate": 2.8663473703235136e-05, "loss": 1.3867, "step": 493 }, { "epoch": 0.09, "learning_rate": 2.8660762696547987e-05, "loss": 1.6186, "step": 494 }, { "epoch": 0.09, "learning_rate": 2.8658051689860836e-05, "loss": 0.9674, "step": 495 }, { "epoch": 0.09, "learning_rate": 2.8655340683173688e-05, "loss": 1.3863, "step": 496 }, { "epoch": 0.09, "learning_rate": 2.8652629676486536e-05, "loss": 1.5791, "step": 497 }, { "epoch": 0.09, "learning_rate": 2.8649918669799385e-05, "loss": 2.6728, "step": 498 }, { "epoch": 0.09, "learning_rate": 2.8647207663112236e-05, "loss": 1.6087, "step": 499 }, { "epoch": 0.09, "learning_rate": 2.8644496656425085e-05, "loss": 1.6939, "step": 500 }, { "epoch": 0.09, "learning_rate": 2.8641785649737937e-05, "loss": 1.6359, "step": 501 }, { "epoch": 0.09, "learning_rate": 2.863907464305079e-05, "loss": 1.2352, "step": 502 }, { "epoch": 0.09, "learning_rate": 2.8636363636363637e-05, "loss": 1.6874, "step": 503 }, { "epoch": 0.09, "learning_rate": 2.863365262967649e-05, "loss": 1.5015, "step": 504 }, { "epoch": 0.09, "learning_rate": 2.8630941622989337e-05, "loss": 1.6769, "step": 505 }, { "epoch": 0.09, "learning_rate": 2.8628230616302186e-05, "loss": 2.0185, "step": 506 }, { "epoch": 0.09, "learning_rate": 2.8625519609615037e-05, "loss": 1.559, "step": 507 }, { "epoch": 0.09, "learning_rate": 2.862280860292789e-05, "loss": 1.2689, "step": 508 }, { "epoch": 0.09, "learning_rate": 2.8620097596240738e-05, "loss": 1.7635, "step": 509 }, { "epoch": 0.09, "learning_rate": 2.861738658955359e-05, "loss": 1.4442, "step": 510 }, { "epoch": 0.09, "learning_rate": 2.8614675582866438e-05, "loss": 1.7883, "step": 511 }, { "epoch": 0.09, "learning_rate": 2.861196457617929e-05, "loss": 1.459, "step": 512 }, { "epoch": 0.09, "learning_rate": 2.8609253569492138e-05, "loss": 1.7306, "step": 513 }, { "epoch": 0.09, "learning_rate": 2.8606542562804987e-05, "loss": 1.5112, "step": 514 }, { "epoch": 0.09, "learning_rate": 2.860383155611784e-05, "loss": 1.8925, "step": 515 }, { "epoch": 0.09, "learning_rate": 2.860112054943069e-05, "loss": 1.5551, "step": 516 }, { "epoch": 0.09, "learning_rate": 2.859840954274354e-05, "loss": 1.475, "step": 517 }, { "epoch": 0.09, "learning_rate": 2.859569853605639e-05, "loss": 1.1458, "step": 518 }, { "epoch": 0.09, "learning_rate": 2.8592987529369242e-05, "loss": 2.1878, "step": 519 }, { "epoch": 0.09, "learning_rate": 2.8590276522682088e-05, "loss": 1.6488, "step": 520 }, { "epoch": 0.09, "learning_rate": 2.858756551599494e-05, "loss": 2.6784, "step": 521 }, { "epoch": 0.09, "learning_rate": 2.858485450930779e-05, "loss": 0.9568, "step": 522 }, { "epoch": 0.09, "learning_rate": 2.858214350262064e-05, "loss": 1.0686, "step": 523 }, { "epoch": 0.09, "learning_rate": 2.857943249593349e-05, "loss": 1.8142, "step": 524 }, { "epoch": 0.09, "learning_rate": 2.857672148924634e-05, "loss": 2.5144, "step": 525 }, { "epoch": 0.1, "learning_rate": 2.8574010482559192e-05, "loss": 1.3901, "step": 526 }, { "epoch": 0.1, "learning_rate": 2.8571299475872044e-05, "loss": 1.2016, "step": 527 }, { "epoch": 0.1, "learning_rate": 2.856858846918489e-05, "loss": 2.4518, "step": 528 }, { "epoch": 0.1, "learning_rate": 2.856587746249774e-05, "loss": 1.9905, "step": 529 }, { "epoch": 0.1, "learning_rate": 2.8563166455810592e-05, "loss": 1.5354, "step": 530 }, { "epoch": 0.1, "learning_rate": 2.856045544912344e-05, "loss": 1.3906, "step": 531 }, { "epoch": 0.1, "learning_rate": 2.8557744442436293e-05, "loss": 1.5522, "step": 532 }, { "epoch": 0.1, "learning_rate": 2.8555033435749144e-05, "loss": 1.7235, "step": 533 }, { "epoch": 0.1, "learning_rate": 2.8552322429061993e-05, "loss": 1.9096, "step": 534 }, { "epoch": 0.1, "learning_rate": 2.8549611422374845e-05, "loss": 1.5327, "step": 535 }, { "epoch": 0.1, "learning_rate": 2.854690041568769e-05, "loss": 1.7822, "step": 536 }, { "epoch": 0.1, "learning_rate": 2.854418940900054e-05, "loss": 1.7668, "step": 537 }, { "epoch": 0.1, "learning_rate": 2.8541478402313393e-05, "loss": 1.503, "step": 538 }, { "epoch": 0.1, "learning_rate": 2.8538767395626242e-05, "loss": 1.9722, "step": 539 }, { "epoch": 0.1, "learning_rate": 2.8536056388939094e-05, "loss": 1.9234, "step": 540 }, { "epoch": 0.1, "learning_rate": 2.8533345382251945e-05, "loss": 1.2621, "step": 541 }, { "epoch": 0.1, "learning_rate": 2.8530634375564794e-05, "loss": 2.249, "step": 542 }, { "epoch": 0.1, "learning_rate": 2.8527923368877646e-05, "loss": 2.1252, "step": 543 }, { "epoch": 0.1, "learning_rate": 2.8525212362190494e-05, "loss": 2.4448, "step": 544 }, { "epoch": 0.1, "learning_rate": 2.8522501355503343e-05, "loss": 1.5071, "step": 545 }, { "epoch": 0.1, "learning_rate": 2.8519790348816194e-05, "loss": 2.082, "step": 546 }, { "epoch": 0.1, "learning_rate": 2.8517079342129043e-05, "loss": 2.3638, "step": 547 }, { "epoch": 0.1, "learning_rate": 2.8514368335441895e-05, "loss": 1.0422, "step": 548 }, { "epoch": 0.1, "learning_rate": 2.8511657328754746e-05, "loss": 0.9536, "step": 549 }, { "epoch": 0.1, "learning_rate": 2.8508946322067595e-05, "loss": 1.7131, "step": 550 }, { "epoch": 0.1, "learning_rate": 2.8506235315380447e-05, "loss": 1.532, "step": 551 }, { "epoch": 0.1, "learning_rate": 2.8503524308693295e-05, "loss": 1.5345, "step": 552 }, { "epoch": 0.1, "learning_rate": 2.8500813302006144e-05, "loss": 1.3215, "step": 553 }, { "epoch": 0.1, "learning_rate": 2.8498102295318995e-05, "loss": 0.9806, "step": 554 }, { "epoch": 0.1, "learning_rate": 2.8495391288631847e-05, "loss": 1.7096, "step": 555 }, { "epoch": 0.1, "learning_rate": 2.8492680281944696e-05, "loss": 1.3683, "step": 556 }, { "epoch": 0.1, "learning_rate": 2.8489969275257548e-05, "loss": 1.3177, "step": 557 }, { "epoch": 0.1, "learning_rate": 2.8487258268570396e-05, "loss": 1.2712, "step": 558 }, { "epoch": 0.1, "learning_rate": 2.8484547261883248e-05, "loss": 1.4483, "step": 559 }, { "epoch": 0.1, "learning_rate": 2.8481836255196096e-05, "loss": 1.0642, "step": 560 }, { "epoch": 0.1, "learning_rate": 2.8479125248508945e-05, "loss": 1.4139, "step": 561 }, { "epoch": 0.1, "learning_rate": 2.8476414241821797e-05, "loss": 1.2133, "step": 562 }, { "epoch": 0.1, "learning_rate": 2.847370323513465e-05, "loss": 0.8381, "step": 563 }, { "epoch": 0.1, "learning_rate": 2.8470992228447497e-05, "loss": 0.8278, "step": 564 }, { "epoch": 0.1, "learning_rate": 2.846828122176035e-05, "loss": 1.4284, "step": 565 }, { "epoch": 0.1, "learning_rate": 2.84655702150732e-05, "loss": 1.8055, "step": 566 }, { "epoch": 0.1, "learning_rate": 2.846285920838605e-05, "loss": 0.9977, "step": 567 }, { "epoch": 0.1, "learning_rate": 2.8460148201698897e-05, "loss": 1.7536, "step": 568 }, { "epoch": 0.1, "learning_rate": 2.8457437195011746e-05, "loss": 1.2484, "step": 569 }, { "epoch": 0.1, "learning_rate": 2.8454726188324598e-05, "loss": 1.3921, "step": 570 }, { "epoch": 0.1, "learning_rate": 2.845201518163745e-05, "loss": 1.4167, "step": 571 }, { "epoch": 0.1, "learning_rate": 2.8449304174950298e-05, "loss": 0.7402, "step": 572 }, { "epoch": 0.1, "learning_rate": 2.844659316826315e-05, "loss": 1.3316, "step": 573 }, { "epoch": 0.1, "learning_rate": 2.8443882161576e-05, "loss": 1.648, "step": 574 }, { "epoch": 0.1, "learning_rate": 2.844117115488885e-05, "loss": 1.9794, "step": 575 }, { "epoch": 0.1, "learning_rate": 2.84384601482017e-05, "loss": 0.8002, "step": 576 }, { "epoch": 0.1, "learning_rate": 2.843574914151455e-05, "loss": 0.8159, "step": 577 }, { "epoch": 0.1, "learning_rate": 2.84330381348274e-05, "loss": 1.617, "step": 578 }, { "epoch": 0.1, "learning_rate": 2.843032712814025e-05, "loss": 2.2053, "step": 579 }, { "epoch": 0.1, "learning_rate": 2.84276161214531e-05, "loss": 1.7646, "step": 580 }, { "epoch": 0.11, "learning_rate": 2.842490511476595e-05, "loss": 1.1045, "step": 581 }, { "epoch": 0.11, "learning_rate": 2.8422194108078803e-05, "loss": 1.8775, "step": 582 }, { "epoch": 0.11, "learning_rate": 2.841948310139165e-05, "loss": 1.2446, "step": 583 }, { "epoch": 0.11, "learning_rate": 2.84167720947045e-05, "loss": 1.7057, "step": 584 }, { "epoch": 0.11, "learning_rate": 2.841406108801735e-05, "loss": 1.3601, "step": 585 }, { "epoch": 0.11, "learning_rate": 2.84113500813302e-05, "loss": 1.2179, "step": 586 }, { "epoch": 0.11, "learning_rate": 2.840863907464305e-05, "loss": 1.318, "step": 587 }, { "epoch": 0.11, "learning_rate": 2.8405928067955903e-05, "loss": 1.2701, "step": 588 }, { "epoch": 0.11, "learning_rate": 2.8403217061268752e-05, "loss": 2.278, "step": 589 }, { "epoch": 0.11, "learning_rate": 2.8400506054581604e-05, "loss": 1.8509, "step": 590 }, { "epoch": 0.11, "learning_rate": 2.8397795047894452e-05, "loss": 1.8724, "step": 591 }, { "epoch": 0.11, "learning_rate": 2.83950840412073e-05, "loss": 1.2666, "step": 592 }, { "epoch": 0.11, "learning_rate": 2.8392373034520152e-05, "loss": 1.7754, "step": 593 }, { "epoch": 0.11, "learning_rate": 2.8389662027833e-05, "loss": 2.1415, "step": 594 }, { "epoch": 0.11, "learning_rate": 2.8386951021145853e-05, "loss": 1.8377, "step": 595 }, { "epoch": 0.11, "learning_rate": 2.8384240014458705e-05, "loss": 1.3448, "step": 596 }, { "epoch": 0.11, "learning_rate": 2.8381529007771553e-05, "loss": 1.4621, "step": 597 }, { "epoch": 0.11, "learning_rate": 2.8378818001084405e-05, "loss": 1.2806, "step": 598 }, { "epoch": 0.11, "learning_rate": 2.8376106994397257e-05, "loss": 1.6919, "step": 599 }, { "epoch": 0.11, "learning_rate": 2.83733959877101e-05, "loss": 1.883, "step": 600 }, { "epoch": 0.11, "learning_rate": 2.8370684981022953e-05, "loss": 1.1773, "step": 601 }, { "epoch": 0.11, "learning_rate": 2.8367973974335802e-05, "loss": 0.9346, "step": 602 }, { "epoch": 0.11, "learning_rate": 2.8365262967648654e-05, "loss": 1.3429, "step": 603 }, { "epoch": 0.11, "learning_rate": 2.8362551960961506e-05, "loss": 1.6346, "step": 604 }, { "epoch": 0.11, "learning_rate": 2.8359840954274354e-05, "loss": 2.0229, "step": 605 }, { "epoch": 0.11, "learning_rate": 2.8357129947587206e-05, "loss": 1.2839, "step": 606 }, { "epoch": 0.11, "learning_rate": 2.8354418940900058e-05, "loss": 1.7024, "step": 607 }, { "epoch": 0.11, "learning_rate": 2.8351707934212903e-05, "loss": 1.4734, "step": 608 }, { "epoch": 0.11, "learning_rate": 2.8348996927525755e-05, "loss": 1.76, "step": 609 }, { "epoch": 0.11, "learning_rate": 2.8346285920838606e-05, "loss": 1.5971, "step": 610 }, { "epoch": 0.11, "learning_rate": 2.8343574914151455e-05, "loss": 1.1585, "step": 611 }, { "epoch": 0.11, "learning_rate": 2.8340863907464307e-05, "loss": 0.9819, "step": 612 }, { "epoch": 0.11, "learning_rate": 2.8338152900777155e-05, "loss": 1.7558, "step": 613 }, { "epoch": 0.11, "learning_rate": 2.8335441894090007e-05, "loss": 1.6226, "step": 614 }, { "epoch": 0.11, "learning_rate": 2.833273088740286e-05, "loss": 1.2949, "step": 615 }, { "epoch": 0.11, "learning_rate": 2.8330019880715704e-05, "loss": 1.6631, "step": 616 }, { "epoch": 0.11, "learning_rate": 2.8327308874028556e-05, "loss": 1.4852, "step": 617 }, { "epoch": 0.11, "learning_rate": 2.8324597867341407e-05, "loss": 1.3654, "step": 618 }, { "epoch": 0.11, "learning_rate": 2.8321886860654256e-05, "loss": 1.9607, "step": 619 }, { "epoch": 0.11, "learning_rate": 2.8319175853967108e-05, "loss": 2.0572, "step": 620 }, { "epoch": 0.11, "learning_rate": 2.831646484727996e-05, "loss": 1.0707, "step": 621 }, { "epoch": 0.11, "learning_rate": 2.8313753840592808e-05, "loss": 1.229, "step": 622 }, { "epoch": 0.11, "learning_rate": 2.831104283390566e-05, "loss": 2.1652, "step": 623 }, { "epoch": 0.11, "learning_rate": 2.8308331827218505e-05, "loss": 1.3688, "step": 624 }, { "epoch": 0.11, "learning_rate": 2.8305620820531357e-05, "loss": 1.3316, "step": 625 }, { "epoch": 0.11, "learning_rate": 2.830290981384421e-05, "loss": 1.3555, "step": 626 }, { "epoch": 0.11, "learning_rate": 2.8300198807157057e-05, "loss": 1.165, "step": 627 }, { "epoch": 0.11, "learning_rate": 2.829748780046991e-05, "loss": 2.5704, "step": 628 }, { "epoch": 0.11, "learning_rate": 2.829477679378276e-05, "loss": 1.2057, "step": 629 }, { "epoch": 0.11, "learning_rate": 2.829206578709561e-05, "loss": 1.7036, "step": 630 }, { "epoch": 0.11, "learning_rate": 2.828935478040846e-05, "loss": 1.4412, "step": 631 }, { "epoch": 0.11, "learning_rate": 2.828664377372131e-05, "loss": 1.8601, "step": 632 }, { "epoch": 0.11, "learning_rate": 2.8283932767034158e-05, "loss": 1.4765, "step": 633 }, { "epoch": 0.11, "learning_rate": 2.828122176034701e-05, "loss": 1.5202, "step": 634 }, { "epoch": 0.11, "learning_rate": 2.8278510753659858e-05, "loss": 1.4951, "step": 635 }, { "epoch": 0.11, "learning_rate": 2.827579974697271e-05, "loss": 0.9366, "step": 636 }, { "epoch": 0.12, "learning_rate": 2.8273088740285562e-05, "loss": 1.5464, "step": 637 }, { "epoch": 0.12, "learning_rate": 2.827037773359841e-05, "loss": 1.4515, "step": 638 }, { "epoch": 0.12, "learning_rate": 2.8267666726911262e-05, "loss": 1.5152, "step": 639 }, { "epoch": 0.12, "learning_rate": 2.826495572022411e-05, "loss": 1.093, "step": 640 }, { "epoch": 0.12, "learning_rate": 2.826224471353696e-05, "loss": 1.2114, "step": 641 }, { "epoch": 0.12, "learning_rate": 2.825953370684981e-05, "loss": 1.9168, "step": 642 }, { "epoch": 0.12, "learning_rate": 2.8256822700162663e-05, "loss": 1.1559, "step": 643 }, { "epoch": 0.12, "learning_rate": 2.825411169347551e-05, "loss": 1.6528, "step": 644 }, { "epoch": 0.12, "learning_rate": 2.8251400686788363e-05, "loss": 1.1606, "step": 645 }, { "epoch": 0.12, "learning_rate": 2.824868968010121e-05, "loss": 1.0794, "step": 646 }, { "epoch": 0.12, "learning_rate": 2.824597867341406e-05, "loss": 1.0433, "step": 647 }, { "epoch": 0.12, "learning_rate": 2.824326766672691e-05, "loss": 1.023, "step": 648 }, { "epoch": 0.12, "learning_rate": 2.824055666003976e-05, "loss": 1.1549, "step": 649 }, { "epoch": 0.12, "learning_rate": 2.8237845653352612e-05, "loss": 0.9864, "step": 650 }, { "epoch": 0.12, "learning_rate": 2.8235134646665464e-05, "loss": 1.6893, "step": 651 }, { "epoch": 0.12, "learning_rate": 2.8232423639978312e-05, "loss": 0.8083, "step": 652 }, { "epoch": 0.12, "learning_rate": 2.8229712633291164e-05, "loss": 0.9089, "step": 653 }, { "epoch": 0.12, "learning_rate": 2.8227001626604016e-05, "loss": 1.8591, "step": 654 }, { "epoch": 0.12, "learning_rate": 2.822429061991686e-05, "loss": 1.3048, "step": 655 }, { "epoch": 0.12, "learning_rate": 2.8221579613229713e-05, "loss": 2.1456, "step": 656 }, { "epoch": 0.12, "learning_rate": 2.8218868606542564e-05, "loss": 1.6491, "step": 657 }, { "epoch": 0.12, "learning_rate": 2.8216157599855413e-05, "loss": 0.9065, "step": 658 }, { "epoch": 0.12, "learning_rate": 2.8213446593168265e-05, "loss": 1.9128, "step": 659 }, { "epoch": 0.12, "learning_rate": 2.8210735586481113e-05, "loss": 1.6907, "step": 660 }, { "epoch": 0.12, "learning_rate": 2.8208024579793965e-05, "loss": 1.0875, "step": 661 }, { "epoch": 0.12, "learning_rate": 2.8205313573106817e-05, "loss": 1.0132, "step": 662 }, { "epoch": 0.12, "learning_rate": 2.8202602566419662e-05, "loss": 1.4922, "step": 663 }, { "epoch": 0.12, "learning_rate": 2.8199891559732514e-05, "loss": 1.1859, "step": 664 }, { "epoch": 0.12, "learning_rate": 2.8197180553045365e-05, "loss": 1.6883, "step": 665 }, { "epoch": 0.12, "learning_rate": 2.8194469546358214e-05, "loss": 1.7627, "step": 666 }, { "epoch": 0.12, "learning_rate": 2.8191758539671066e-05, "loss": 1.625, "step": 667 }, { "epoch": 0.12, "learning_rate": 2.8189047532983918e-05, "loss": 0.9397, "step": 668 }, { "epoch": 0.12, "learning_rate": 2.8186336526296766e-05, "loss": 1.6626, "step": 669 }, { "epoch": 0.12, "learning_rate": 2.8183625519609618e-05, "loss": 1.2204, "step": 670 }, { "epoch": 0.12, "learning_rate": 2.8180914512922463e-05, "loss": 1.2684, "step": 671 }, { "epoch": 0.12, "learning_rate": 2.8178203506235315e-05, "loss": 1.4496, "step": 672 }, { "epoch": 0.12, "learning_rate": 2.8175492499548167e-05, "loss": 1.2482, "step": 673 }, { "epoch": 0.12, "learning_rate": 2.8172781492861015e-05, "loss": 1.2186, "step": 674 }, { "epoch": 0.12, "learning_rate": 2.8170070486173867e-05, "loss": 1.0006, "step": 675 }, { "epoch": 0.12, "learning_rate": 2.816735947948672e-05, "loss": 1.336, "step": 676 }, { "epoch": 0.12, "learning_rate": 2.8164648472799567e-05, "loss": 1.8145, "step": 677 }, { "epoch": 0.12, "learning_rate": 2.816193746611242e-05, "loss": 1.8692, "step": 678 }, { "epoch": 0.12, "learning_rate": 2.8159226459425267e-05, "loss": 2.5005, "step": 679 }, { "epoch": 0.12, "learning_rate": 2.8156515452738116e-05, "loss": 1.6582, "step": 680 }, { "epoch": 0.12, "learning_rate": 2.8153804446050968e-05, "loss": 2.3782, "step": 681 }, { "epoch": 0.12, "learning_rate": 2.8151093439363816e-05, "loss": 2.8293, "step": 682 }, { "epoch": 0.12, "learning_rate": 2.8148382432676668e-05, "loss": 1.4738, "step": 683 }, { "epoch": 0.12, "learning_rate": 2.814567142598952e-05, "loss": 1.7961, "step": 684 }, { "epoch": 0.12, "learning_rate": 2.8142960419302368e-05, "loss": 1.7145, "step": 685 }, { "epoch": 0.12, "learning_rate": 2.814024941261522e-05, "loss": 0.9449, "step": 686 }, { "epoch": 0.12, "learning_rate": 2.813753840592807e-05, "loss": 1.1767, "step": 687 }, { "epoch": 0.12, "learning_rate": 2.8134827399240917e-05, "loss": 1.4792, "step": 688 }, { "epoch": 0.12, "learning_rate": 2.813211639255377e-05, "loss": 1.2841, "step": 689 }, { "epoch": 0.12, "learning_rate": 2.812940538586662e-05, "loss": 1.864, "step": 690 }, { "epoch": 0.12, "learning_rate": 2.812669437917947e-05, "loss": 1.2993, "step": 691 }, { "epoch": 0.13, "learning_rate": 2.812398337249232e-05, "loss": 2.2131, "step": 692 }, { "epoch": 0.13, "learning_rate": 2.812127236580517e-05, "loss": 1.2374, "step": 693 }, { "epoch": 0.13, "learning_rate": 2.811856135911802e-05, "loss": 0.9613, "step": 694 }, { "epoch": 0.13, "learning_rate": 2.811585035243087e-05, "loss": 0.9574, "step": 695 }, { "epoch": 0.13, "learning_rate": 2.8113139345743718e-05, "loss": 1.208, "step": 696 }, { "epoch": 0.13, "learning_rate": 2.811042833905657e-05, "loss": 1.4253, "step": 697 }, { "epoch": 0.13, "learning_rate": 2.810771733236942e-05, "loss": 1.2199, "step": 698 }, { "epoch": 0.13, "learning_rate": 2.810500632568227e-05, "loss": 1.0694, "step": 699 }, { "epoch": 0.13, "learning_rate": 2.8102295318995122e-05, "loss": 2.1214, "step": 700 }, { "epoch": 0.13, "learning_rate": 2.8099584312307974e-05, "loss": 1.0465, "step": 701 }, { "epoch": 0.13, "learning_rate": 2.8096873305620822e-05, "loss": 1.4756, "step": 702 }, { "epoch": 0.13, "learning_rate": 2.809416229893367e-05, "loss": 1.7896, "step": 703 }, { "epoch": 0.13, "learning_rate": 2.809145129224652e-05, "loss": 1.118, "step": 704 }, { "epoch": 0.13, "learning_rate": 2.808874028555937e-05, "loss": 1.5795, "step": 705 }, { "epoch": 0.13, "learning_rate": 2.8086029278872223e-05, "loss": 1.8597, "step": 706 }, { "epoch": 0.13, "learning_rate": 2.808331827218507e-05, "loss": 1.3051, "step": 707 }, { "epoch": 0.13, "learning_rate": 2.8080607265497923e-05, "loss": 2.2246, "step": 708 }, { "epoch": 0.13, "learning_rate": 2.8077896258810775e-05, "loss": 1.3521, "step": 709 }, { "epoch": 0.13, "learning_rate": 2.8075185252123623e-05, "loss": 1.7955, "step": 710 }, { "epoch": 0.13, "learning_rate": 2.807247424543647e-05, "loss": 1.9576, "step": 711 }, { "epoch": 0.13, "learning_rate": 2.8069763238749323e-05, "loss": 0.7538, "step": 712 }, { "epoch": 0.13, "learning_rate": 2.8067052232062172e-05, "loss": 2.2223, "step": 713 }, { "epoch": 0.13, "learning_rate": 2.8064341225375024e-05, "loss": 2.1558, "step": 714 }, { "epoch": 0.13, "learning_rate": 2.8061630218687872e-05, "loss": 2.0614, "step": 715 }, { "epoch": 0.13, "learning_rate": 2.8058919212000724e-05, "loss": 1.6626, "step": 716 }, { "epoch": 0.13, "learning_rate": 2.8056208205313576e-05, "loss": 1.229, "step": 717 }, { "epoch": 0.13, "learning_rate": 2.8053497198626424e-05, "loss": 1.4634, "step": 718 }, { "epoch": 0.13, "learning_rate": 2.8050786191939273e-05, "loss": 1.4366, "step": 719 }, { "epoch": 0.13, "learning_rate": 2.8048075185252125e-05, "loss": 0.9298, "step": 720 }, { "epoch": 0.13, "learning_rate": 2.8045364178564973e-05, "loss": 1.1517, "step": 721 }, { "epoch": 0.13, "learning_rate": 2.8042653171877825e-05, "loss": 2.3786, "step": 722 }, { "epoch": 0.13, "learning_rate": 2.8039942165190677e-05, "loss": 1.9484, "step": 723 }, { "epoch": 0.13, "learning_rate": 2.8037231158503525e-05, "loss": 1.4541, "step": 724 }, { "epoch": 0.13, "learning_rate": 2.8034520151816377e-05, "loss": 1.6789, "step": 725 }, { "epoch": 0.13, "learning_rate": 2.8031809145129225e-05, "loss": 0.9742, "step": 726 }, { "epoch": 0.13, "learning_rate": 2.8029098138442074e-05, "loss": 1.3029, "step": 727 }, { "epoch": 0.13, "learning_rate": 2.8026387131754926e-05, "loss": 1.2607, "step": 728 }, { "epoch": 0.13, "learning_rate": 2.8023676125067774e-05, "loss": 1.0348, "step": 729 }, { "epoch": 0.13, "learning_rate": 2.8020965118380626e-05, "loss": 1.278, "step": 730 }, { "epoch": 0.13, "learning_rate": 2.8018254111693478e-05, "loss": 1.1018, "step": 731 }, { "epoch": 0.13, "learning_rate": 2.8015543105006326e-05, "loss": 1.7663, "step": 732 }, { "epoch": 0.13, "learning_rate": 2.8012832098319178e-05, "loss": 1.0289, "step": 733 }, { "epoch": 0.13, "learning_rate": 2.801012109163203e-05, "loss": 1.8278, "step": 734 }, { "epoch": 0.13, "learning_rate": 2.8007410084944875e-05, "loss": 1.8795, "step": 735 }, { "epoch": 0.13, "learning_rate": 2.8004699078257727e-05, "loss": 1.4956, "step": 736 }, { "epoch": 0.13, "learning_rate": 2.8001988071570575e-05, "loss": 1.201, "step": 737 }, { "epoch": 0.13, "learning_rate": 2.7999277064883427e-05, "loss": 1.3892, "step": 738 }, { "epoch": 0.13, "learning_rate": 2.799656605819628e-05, "loss": 0.8644, "step": 739 }, { "epoch": 0.13, "learning_rate": 2.7993855051509127e-05, "loss": 1.2239, "step": 740 }, { "epoch": 0.13, "learning_rate": 2.799114404482198e-05, "loss": 1.8824, "step": 741 }, { "epoch": 0.13, "learning_rate": 2.798843303813483e-05, "loss": 1.7849, "step": 742 }, { "epoch": 0.13, "learning_rate": 2.7985722031447676e-05, "loss": 1.6522, "step": 743 }, { "epoch": 0.13, "learning_rate": 2.7983011024760528e-05, "loss": 1.9213, "step": 744 }, { "epoch": 0.13, "learning_rate": 2.798030001807338e-05, "loss": 1.7799, "step": 745 }, { "epoch": 0.13, "learning_rate": 2.7977589011386228e-05, "loss": 1.5043, "step": 746 }, { "epoch": 0.14, "learning_rate": 2.797487800469908e-05, "loss": 0.9324, "step": 747 }, { "epoch": 0.14, "learning_rate": 2.7972166998011928e-05, "loss": 1.5638, "step": 748 }, { "epoch": 0.14, "learning_rate": 2.796945599132478e-05, "loss": 1.6166, "step": 749 }, { "epoch": 0.14, "learning_rate": 2.7966744984637632e-05, "loss": 1.0174, "step": 750 }, { "epoch": 0.14, "learning_rate": 2.7964033977950477e-05, "loss": 1.2929, "step": 751 }, { "epoch": 0.14, "learning_rate": 2.796132297126333e-05, "loss": 1.2656, "step": 752 }, { "epoch": 0.14, "learning_rate": 2.795861196457618e-05, "loss": 1.3442, "step": 753 }, { "epoch": 0.14, "learning_rate": 2.795590095788903e-05, "loss": 2.075, "step": 754 }, { "epoch": 0.14, "learning_rate": 2.795318995120188e-05, "loss": 1.6361, "step": 755 }, { "epoch": 0.14, "learning_rate": 2.7950478944514733e-05, "loss": 1.3289, "step": 756 }, { "epoch": 0.14, "learning_rate": 2.794776793782758e-05, "loss": 1.9188, "step": 757 }, { "epoch": 0.14, "learning_rate": 2.7945056931140433e-05, "loss": 1.4185, "step": 758 }, { "epoch": 0.14, "learning_rate": 2.7942345924453278e-05, "loss": 1.4485, "step": 759 }, { "epoch": 0.14, "learning_rate": 2.793963491776613e-05, "loss": 1.5796, "step": 760 }, { "epoch": 0.14, "learning_rate": 2.7936923911078982e-05, "loss": 1.5179, "step": 761 }, { "epoch": 0.14, "learning_rate": 2.793421290439183e-05, "loss": 1.1652, "step": 762 }, { "epoch": 0.14, "learning_rate": 2.7931501897704682e-05, "loss": 1.4621, "step": 763 }, { "epoch": 0.14, "learning_rate": 2.7928790891017534e-05, "loss": 2.2181, "step": 764 }, { "epoch": 0.14, "learning_rate": 2.7926079884330382e-05, "loss": 0.7592, "step": 765 }, { "epoch": 0.14, "learning_rate": 2.7923368877643234e-05, "loss": 0.9101, "step": 766 }, { "epoch": 0.14, "learning_rate": 2.7920657870956083e-05, "loss": 1.4595, "step": 767 }, { "epoch": 0.14, "learning_rate": 2.791794686426893e-05, "loss": 1.9067, "step": 768 }, { "epoch": 0.14, "learning_rate": 2.7915235857581783e-05, "loss": 1.1606, "step": 769 }, { "epoch": 0.14, "learning_rate": 2.791252485089463e-05, "loss": 0.9222, "step": 770 }, { "epoch": 0.14, "learning_rate": 2.7909813844207483e-05, "loss": 1.8671, "step": 771 }, { "epoch": 0.14, "learning_rate": 2.7907102837520335e-05, "loss": 0.886, "step": 772 }, { "epoch": 0.14, "learning_rate": 2.7904391830833183e-05, "loss": 1.6442, "step": 773 }, { "epoch": 0.14, "learning_rate": 2.7901680824146032e-05, "loss": 1.9321, "step": 774 }, { "epoch": 0.14, "learning_rate": 2.7898969817458884e-05, "loss": 1.2881, "step": 775 }, { "epoch": 0.14, "learning_rate": 2.7896258810771732e-05, "loss": 1.8495, "step": 776 }, { "epoch": 0.14, "learning_rate": 2.7893547804084584e-05, "loss": 2.0968, "step": 777 }, { "epoch": 0.14, "learning_rate": 2.7890836797397436e-05, "loss": 1.6106, "step": 778 }, { "epoch": 0.14, "learning_rate": 2.7888125790710284e-05, "loss": 1.8478, "step": 779 }, { "epoch": 0.14, "learning_rate": 2.7885414784023136e-05, "loss": 1.4958, "step": 780 }, { "epoch": 0.14, "learning_rate": 2.7882703777335984e-05, "loss": 1.6092, "step": 781 }, { "epoch": 0.14, "learning_rate": 2.7879992770648833e-05, "loss": 1.5398, "step": 782 }, { "epoch": 0.14, "learning_rate": 2.7877281763961685e-05, "loss": 1.2929, "step": 783 }, { "epoch": 0.14, "learning_rate": 2.7874570757274533e-05, "loss": 1.8968, "step": 784 }, { "epoch": 0.14, "learning_rate": 2.7871859750587385e-05, "loss": 1.5639, "step": 785 }, { "epoch": 0.14, "learning_rate": 2.7869148743900237e-05, "loss": 0.9741, "step": 786 }, { "epoch": 0.14, "learning_rate": 2.7866437737213085e-05, "loss": 1.7381, "step": 787 }, { "epoch": 0.14, "learning_rate": 2.7863726730525937e-05, "loss": 1.915, "step": 788 }, { "epoch": 0.14, "learning_rate": 2.786101572383879e-05, "loss": 1.27, "step": 789 }, { "epoch": 0.14, "learning_rate": 2.7858304717151634e-05, "loss": 1.1909, "step": 790 }, { "epoch": 0.14, "learning_rate": 2.7855593710464486e-05, "loss": 1.4762, "step": 791 }, { "epoch": 0.14, "learning_rate": 2.7852882703777338e-05, "loss": 1.362, "step": 792 }, { "epoch": 0.14, "learning_rate": 2.7850171697090186e-05, "loss": 1.1847, "step": 793 }, { "epoch": 0.14, "learning_rate": 2.7847460690403038e-05, "loss": 1.4034, "step": 794 }, { "epoch": 0.14, "learning_rate": 2.7844749683715886e-05, "loss": 1.607, "step": 795 }, { "epoch": 0.14, "learning_rate": 2.7842038677028738e-05, "loss": 2.0379, "step": 796 }, { "epoch": 0.14, "learning_rate": 2.783932767034159e-05, "loss": 1.362, "step": 797 }, { "epoch": 0.14, "learning_rate": 2.7836616663654435e-05, "loss": 1.0423, "step": 798 }, { "epoch": 0.14, "learning_rate": 2.7833905656967287e-05, "loss": 1.3649, "step": 799 }, { "epoch": 0.14, "learning_rate": 2.783119465028014e-05, "loss": 2.1617, "step": 800 }, { "epoch": 0.14, "learning_rate": 2.7828483643592987e-05, "loss": 1.2382, "step": 801 }, { "epoch": 0.14, "learning_rate": 2.782577263690584e-05, "loss": 1.6331, "step": 802 }, { "epoch": 0.15, "learning_rate": 2.782306163021869e-05, "loss": 1.3135, "step": 803 }, { "epoch": 0.15, "learning_rate": 2.782035062353154e-05, "loss": 1.9213, "step": 804 }, { "epoch": 0.15, "learning_rate": 2.781763961684439e-05, "loss": 1.0475, "step": 805 }, { "epoch": 0.15, "learning_rate": 2.7814928610157236e-05, "loss": 1.7022, "step": 806 }, { "epoch": 0.15, "learning_rate": 2.7812217603470088e-05, "loss": 1.0803, "step": 807 }, { "epoch": 0.15, "learning_rate": 2.780950659678294e-05, "loss": 1.3057, "step": 808 }, { "epoch": 0.15, "learning_rate": 2.7806795590095788e-05, "loss": 1.388, "step": 809 }, { "epoch": 0.15, "learning_rate": 2.780408458340864e-05, "loss": 1.3236, "step": 810 }, { "epoch": 0.15, "learning_rate": 2.7801373576721492e-05, "loss": 1.5372, "step": 811 }, { "epoch": 0.15, "learning_rate": 2.779866257003434e-05, "loss": 1.4807, "step": 812 }, { "epoch": 0.15, "learning_rate": 2.7795951563347192e-05, "loss": 1.4414, "step": 813 }, { "epoch": 0.15, "learning_rate": 2.779324055666004e-05, "loss": 1.1731, "step": 814 }, { "epoch": 0.15, "learning_rate": 2.779052954997289e-05, "loss": 1.4755, "step": 815 }, { "epoch": 0.15, "learning_rate": 2.778781854328574e-05, "loss": 1.3674, "step": 816 }, { "epoch": 0.15, "learning_rate": 2.778510753659859e-05, "loss": 1.2507, "step": 817 }, { "epoch": 0.15, "learning_rate": 2.778239652991144e-05, "loss": 0.4954, "step": 818 }, { "epoch": 0.15, "learning_rate": 2.7779685523224293e-05, "loss": 1.4325, "step": 819 }, { "epoch": 0.15, "learning_rate": 2.777697451653714e-05, "loss": 1.5546, "step": 820 }, { "epoch": 0.15, "learning_rate": 2.7774263509849993e-05, "loss": 1.7425, "step": 821 }, { "epoch": 0.15, "learning_rate": 2.777155250316284e-05, "loss": 1.7167, "step": 822 }, { "epoch": 0.15, "learning_rate": 2.776884149647569e-05, "loss": 1.5157, "step": 823 }, { "epoch": 0.15, "learning_rate": 2.7766130489788542e-05, "loss": 1.6686, "step": 824 }, { "epoch": 0.15, "learning_rate": 2.7763419483101394e-05, "loss": 1.0087, "step": 825 }, { "epoch": 0.15, "learning_rate": 2.7760708476414242e-05, "loss": 0.8828, "step": 826 }, { "epoch": 0.15, "learning_rate": 2.7757997469727094e-05, "loss": 1.3901, "step": 827 }, { "epoch": 0.15, "learning_rate": 2.7755286463039942e-05, "loss": 1.3027, "step": 828 }, { "epoch": 0.15, "learning_rate": 2.7752575456352794e-05, "loss": 1.6511, "step": 829 }, { "epoch": 0.15, "learning_rate": 2.7749864449665643e-05, "loss": 1.7985, "step": 830 }, { "epoch": 0.15, "learning_rate": 2.774715344297849e-05, "loss": 0.9074, "step": 831 }, { "epoch": 0.15, "learning_rate": 2.7744442436291343e-05, "loss": 1.0855, "step": 832 }, { "epoch": 0.15, "learning_rate": 2.7741731429604195e-05, "loss": 1.6934, "step": 833 }, { "epoch": 0.15, "learning_rate": 2.7739020422917043e-05, "loss": 1.1602, "step": 834 }, { "epoch": 0.15, "learning_rate": 2.7736309416229895e-05, "loss": 0.8767, "step": 835 }, { "epoch": 0.15, "learning_rate": 2.7733598409542747e-05, "loss": 0.8478, "step": 836 }, { "epoch": 0.15, "learning_rate": 2.7730887402855595e-05, "loss": 1.5823, "step": 837 }, { "epoch": 0.15, "learning_rate": 2.7728176396168444e-05, "loss": 0.9519, "step": 838 }, { "epoch": 0.15, "learning_rate": 2.7725465389481292e-05, "loss": 1.3653, "step": 839 }, { "epoch": 0.15, "learning_rate": 2.7722754382794144e-05, "loss": 1.621, "step": 840 }, { "epoch": 0.15, "learning_rate": 2.7720043376106996e-05, "loss": 1.3014, "step": 841 }, { "epoch": 0.15, "learning_rate": 2.7717332369419844e-05, "loss": 1.6295, "step": 842 }, { "epoch": 0.15, "learning_rate": 2.7714621362732696e-05, "loss": 1.3253, "step": 843 }, { "epoch": 0.15, "learning_rate": 2.7711910356045548e-05, "loss": 1.2103, "step": 844 }, { "epoch": 0.15, "learning_rate": 2.7709199349358396e-05, "loss": 0.6711, "step": 845 }, { "epoch": 0.15, "learning_rate": 2.7706488342671245e-05, "loss": 1.1278, "step": 846 }, { "epoch": 0.15, "learning_rate": 2.7703777335984097e-05, "loss": 0.9123, "step": 847 }, { "epoch": 0.15, "learning_rate": 2.7701066329296945e-05, "loss": 1.2692, "step": 848 }, { "epoch": 0.15, "learning_rate": 2.7698355322609797e-05, "loss": 1.6772, "step": 849 }, { "epoch": 0.15, "learning_rate": 2.7695644315922645e-05, "loss": 1.6089, "step": 850 }, { "epoch": 0.15, "learning_rate": 2.7692933309235497e-05, "loss": 1.5244, "step": 851 }, { "epoch": 0.15, "learning_rate": 2.769022230254835e-05, "loss": 1.1262, "step": 852 }, { "epoch": 0.15, "learning_rate": 2.7687511295861197e-05, "loss": 1.8361, "step": 853 }, { "epoch": 0.15, "learning_rate": 2.7684800289174046e-05, "loss": 1.6302, "step": 854 }, { "epoch": 0.15, "learning_rate": 2.7682089282486898e-05, "loss": 1.759, "step": 855 }, { "epoch": 0.15, "learning_rate": 2.7679378275799746e-05, "loss": 1.5601, "step": 856 }, { "epoch": 0.15, "learning_rate": 2.7676667269112598e-05, "loss": 1.811, "step": 857 }, { "epoch": 0.16, "learning_rate": 2.767395626242545e-05, "loss": 1.3286, "step": 858 }, { "epoch": 0.16, "learning_rate": 2.7671245255738298e-05, "loss": 1.4216, "step": 859 }, { "epoch": 0.16, "learning_rate": 2.766853424905115e-05, "loss": 1.4646, "step": 860 }, { "epoch": 0.16, "learning_rate": 2.7665823242364e-05, "loss": 0.564, "step": 861 }, { "epoch": 0.16, "learning_rate": 2.7663112235676847e-05, "loss": 2.3867, "step": 862 }, { "epoch": 0.16, "learning_rate": 2.76604012289897e-05, "loss": 2.031, "step": 863 }, { "epoch": 0.16, "learning_rate": 2.7657690222302547e-05, "loss": 1.5519, "step": 864 }, { "epoch": 0.16, "learning_rate": 2.76549792156154e-05, "loss": 1.0555, "step": 865 }, { "epoch": 0.16, "learning_rate": 2.765226820892825e-05, "loss": 2.2191, "step": 866 }, { "epoch": 0.16, "learning_rate": 2.76495572022411e-05, "loss": 1.0895, "step": 867 }, { "epoch": 0.16, "learning_rate": 2.764684619555395e-05, "loss": 1.3412, "step": 868 }, { "epoch": 0.16, "learning_rate": 2.7644135188866803e-05, "loss": 0.9356, "step": 869 }, { "epoch": 0.16, "learning_rate": 2.7641424182179648e-05, "loss": 1.011, "step": 870 }, { "epoch": 0.16, "learning_rate": 2.76387131754925e-05, "loss": 0.7509, "step": 871 }, { "epoch": 0.16, "learning_rate": 2.763600216880535e-05, "loss": 1.1129, "step": 872 }, { "epoch": 0.16, "learning_rate": 2.76332911621182e-05, "loss": 1.3762, "step": 873 }, { "epoch": 0.16, "learning_rate": 2.7630580155431052e-05, "loss": 1.3219, "step": 874 }, { "epoch": 0.16, "learning_rate": 2.76278691487439e-05, "loss": 1.0541, "step": 875 }, { "epoch": 0.16, "learning_rate": 2.7625158142056752e-05, "loss": 1.2152, "step": 876 }, { "epoch": 0.16, "learning_rate": 2.7622447135369604e-05, "loss": 2.121, "step": 877 }, { "epoch": 0.16, "learning_rate": 2.761973612868245e-05, "loss": 1.8951, "step": 878 }, { "epoch": 0.16, "learning_rate": 2.76170251219953e-05, "loss": 2.478, "step": 879 }, { "epoch": 0.16, "learning_rate": 2.7614314115308153e-05, "loss": 1.8836, "step": 880 }, { "epoch": 0.16, "learning_rate": 2.7611603108621e-05, "loss": 2.1591, "step": 881 }, { "epoch": 0.16, "learning_rate": 2.7608892101933853e-05, "loss": 2.227, "step": 882 }, { "epoch": 0.16, "learning_rate": 2.76061810952467e-05, "loss": 1.1518, "step": 883 }, { "epoch": 0.16, "learning_rate": 2.7603470088559553e-05, "loss": 1.3148, "step": 884 }, { "epoch": 0.16, "learning_rate": 2.7600759081872405e-05, "loss": 1.403, "step": 885 }, { "epoch": 0.16, "learning_rate": 2.759804807518525e-05, "loss": 0.9945, "step": 886 }, { "epoch": 0.16, "learning_rate": 2.7595337068498102e-05, "loss": 1.0527, "step": 887 }, { "epoch": 0.16, "learning_rate": 2.7592626061810954e-05, "loss": 0.9933, "step": 888 }, { "epoch": 0.16, "learning_rate": 2.7589915055123802e-05, "loss": 1.7834, "step": 889 }, { "epoch": 0.16, "learning_rate": 2.7587204048436654e-05, "loss": 1.2275, "step": 890 }, { "epoch": 0.16, "learning_rate": 2.7584493041749506e-05, "loss": 1.0062, "step": 891 }, { "epoch": 0.16, "learning_rate": 2.7581782035062354e-05, "loss": 0.9146, "step": 892 }, { "epoch": 0.16, "learning_rate": 2.7579071028375206e-05, "loss": 1.1434, "step": 893 }, { "epoch": 0.16, "learning_rate": 2.757636002168805e-05, "loss": 1.7508, "step": 894 }, { "epoch": 0.16, "learning_rate": 2.7573649015000903e-05, "loss": 1.5335, "step": 895 }, { "epoch": 0.16, "learning_rate": 2.7570938008313755e-05, "loss": 1.7775, "step": 896 }, { "epoch": 0.16, "learning_rate": 2.7568227001626603e-05, "loss": 1.1619, "step": 897 }, { "epoch": 0.16, "learning_rate": 2.7565515994939455e-05, "loss": 1.5116, "step": 898 }, { "epoch": 0.16, "learning_rate": 2.7562804988252307e-05, "loss": 0.8639, "step": 899 }, { "epoch": 0.16, "learning_rate": 2.7560093981565155e-05, "loss": 1.0809, "step": 900 }, { "epoch": 0.16, "learning_rate": 2.7557382974878004e-05, "loss": 1.5509, "step": 901 }, { "epoch": 0.16, "learning_rate": 2.7554671968190856e-05, "loss": 1.4252, "step": 902 }, { "epoch": 0.16, "learning_rate": 2.7551960961503704e-05, "loss": 1.8589, "step": 903 }, { "epoch": 0.16, "learning_rate": 2.7549249954816556e-05, "loss": 0.8532, "step": 904 }, { "epoch": 0.16, "learning_rate": 2.7546538948129404e-05, "loss": 1.4916, "step": 905 }, { "epoch": 0.16, "learning_rate": 2.7543827941442256e-05, "loss": 2.1607, "step": 906 }, { "epoch": 0.16, "learning_rate": 2.7541116934755108e-05, "loss": 1.5667, "step": 907 }, { "epoch": 0.16, "learning_rate": 2.7538405928067957e-05, "loss": 0.962, "step": 908 }, { "epoch": 0.16, "learning_rate": 2.7535694921380805e-05, "loss": 1.3283, "step": 909 }, { "epoch": 0.16, "learning_rate": 2.7532983914693657e-05, "loss": 1.3265, "step": 910 }, { "epoch": 0.16, "learning_rate": 2.7530272908006505e-05, "loss": 1.7632, "step": 911 }, { "epoch": 0.16, "learning_rate": 2.7527561901319357e-05, "loss": 1.6695, "step": 912 }, { "epoch": 0.17, "learning_rate": 2.752485089463221e-05, "loss": 1.242, "step": 913 }, { "epoch": 0.17, "learning_rate": 2.7522139887945057e-05, "loss": 1.2528, "step": 914 }, { "epoch": 0.17, "learning_rate": 2.751942888125791e-05, "loss": 1.5344, "step": 915 }, { "epoch": 0.17, "learning_rate": 2.7516717874570758e-05, "loss": 1.4, "step": 916 }, { "epoch": 0.17, "learning_rate": 2.7514006867883606e-05, "loss": 1.0778, "step": 917 }, { "epoch": 0.17, "learning_rate": 2.7511295861196458e-05, "loss": 1.1307, "step": 918 }, { "epoch": 0.17, "learning_rate": 2.7508584854509306e-05, "loss": 1.1655, "step": 919 }, { "epoch": 0.17, "learning_rate": 2.7505873847822158e-05, "loss": 1.6872, "step": 920 }, { "epoch": 0.17, "learning_rate": 2.750316284113501e-05, "loss": 2.2609, "step": 921 }, { "epoch": 0.17, "learning_rate": 2.750045183444786e-05, "loss": 1.7723, "step": 922 }, { "epoch": 0.17, "learning_rate": 2.749774082776071e-05, "loss": 1.8643, "step": 923 }, { "epoch": 0.17, "learning_rate": 2.7495029821073562e-05, "loss": 1.8209, "step": 924 }, { "epoch": 0.17, "learning_rate": 2.7492318814386407e-05, "loss": 1.4284, "step": 925 }, { "epoch": 0.17, "learning_rate": 2.748960780769926e-05, "loss": 0.9093, "step": 926 }, { "epoch": 0.17, "learning_rate": 2.7486896801012107e-05, "loss": 1.5051, "step": 927 }, { "epoch": 0.17, "learning_rate": 2.748418579432496e-05, "loss": 1.358, "step": 928 }, { "epoch": 0.17, "learning_rate": 2.748147478763781e-05, "loss": 2.0595, "step": 929 }, { "epoch": 0.17, "learning_rate": 2.747876378095066e-05, "loss": 1.9383, "step": 930 }, { "epoch": 0.17, "learning_rate": 2.747605277426351e-05, "loss": 1.0778, "step": 931 }, { "epoch": 0.17, "learning_rate": 2.7473341767576363e-05, "loss": 1.6838, "step": 932 }, { "epoch": 0.17, "learning_rate": 2.7470630760889208e-05, "loss": 2.0678, "step": 933 }, { "epoch": 0.17, "learning_rate": 2.746791975420206e-05, "loss": 1.017, "step": 934 }, { "epoch": 0.17, "learning_rate": 2.7465208747514912e-05, "loss": 1.3344, "step": 935 }, { "epoch": 0.17, "learning_rate": 2.746249774082776e-05, "loss": 1.4836, "step": 936 }, { "epoch": 0.17, "learning_rate": 2.7459786734140612e-05, "loss": 1.5601, "step": 937 }, { "epoch": 0.17, "learning_rate": 2.745707572745346e-05, "loss": 1.626, "step": 938 }, { "epoch": 0.17, "learning_rate": 2.7454364720766312e-05, "loss": 1.4621, "step": 939 }, { "epoch": 0.17, "learning_rate": 2.7451653714079164e-05, "loss": 1.5824, "step": 940 }, { "epoch": 0.17, "learning_rate": 2.744894270739201e-05, "loss": 1.3691, "step": 941 }, { "epoch": 0.17, "learning_rate": 2.744623170070486e-05, "loss": 1.1128, "step": 942 }, { "epoch": 0.17, "learning_rate": 2.7443520694017713e-05, "loss": 1.1357, "step": 943 }, { "epoch": 0.17, "learning_rate": 2.744080968733056e-05, "loss": 1.4141, "step": 944 }, { "epoch": 0.17, "learning_rate": 2.7438098680643413e-05, "loss": 1.6681, "step": 945 }, { "epoch": 0.17, "learning_rate": 2.7435387673956265e-05, "loss": 1.008, "step": 946 }, { "epoch": 0.17, "learning_rate": 2.7432676667269113e-05, "loss": 0.7155, "step": 947 }, { "epoch": 0.17, "learning_rate": 2.7429965660581965e-05, "loss": 0.9281, "step": 948 }, { "epoch": 0.17, "learning_rate": 2.7427254653894814e-05, "loss": 1.4412, "step": 949 }, { "epoch": 0.17, "learning_rate": 2.7424543647207662e-05, "loss": 2.068, "step": 950 }, { "epoch": 0.17, "learning_rate": 2.7421832640520514e-05, "loss": 1.26, "step": 951 }, { "epoch": 0.17, "learning_rate": 2.7419121633833362e-05, "loss": 1.5473, "step": 952 }, { "epoch": 0.17, "learning_rate": 2.7416410627146214e-05, "loss": 1.1881, "step": 953 }, { "epoch": 0.17, "learning_rate": 2.7413699620459066e-05, "loss": 0.9008, "step": 954 }, { "epoch": 0.17, "learning_rate": 2.7410988613771915e-05, "loss": 1.4195, "step": 955 }, { "epoch": 0.17, "learning_rate": 2.7408277607084766e-05, "loss": 1.7989, "step": 956 }, { "epoch": 0.17, "learning_rate": 2.7405566600397615e-05, "loss": 1.0972, "step": 957 }, { "epoch": 0.17, "learning_rate": 2.7402855593710463e-05, "loss": 1.6404, "step": 958 }, { "epoch": 0.17, "learning_rate": 2.7400144587023315e-05, "loss": 1.4384, "step": 959 }, { "epoch": 0.17, "learning_rate": 2.7397433580336167e-05, "loss": 0.884, "step": 960 }, { "epoch": 0.17, "learning_rate": 2.7394722573649015e-05, "loss": 1.9524, "step": 961 }, { "epoch": 0.17, "learning_rate": 2.7392011566961867e-05, "loss": 0.9372, "step": 962 }, { "epoch": 0.17, "learning_rate": 2.7389300560274716e-05, "loss": 1.6075, "step": 963 }, { "epoch": 0.17, "learning_rate": 2.7386589553587567e-05, "loss": 1.7458, "step": 964 }, { "epoch": 0.17, "learning_rate": 2.7383878546900416e-05, "loss": 1.55, "step": 965 }, { "epoch": 0.17, "learning_rate": 2.7381167540213264e-05, "loss": 1.1779, "step": 966 }, { "epoch": 0.17, "learning_rate": 2.7378456533526116e-05, "loss": 1.395, "step": 967 }, { "epoch": 0.17, "learning_rate": 2.7375745526838968e-05, "loss": 1.7095, "step": 968 }, { "epoch": 0.18, "learning_rate": 2.7373034520151816e-05, "loss": 1.291, "step": 969 }, { "epoch": 0.18, "learning_rate": 2.7370323513464668e-05, "loss": 1.8019, "step": 970 }, { "epoch": 0.18, "learning_rate": 2.736761250677752e-05, "loss": 1.2895, "step": 971 }, { "epoch": 0.18, "learning_rate": 2.736490150009037e-05, "loss": 1.5737, "step": 972 }, { "epoch": 0.18, "learning_rate": 2.7362190493403217e-05, "loss": 1.4497, "step": 973 }, { "epoch": 0.18, "learning_rate": 2.7359479486716065e-05, "loss": 1.139, "step": 974 }, { "epoch": 0.18, "learning_rate": 2.7356768480028917e-05, "loss": 1.3104, "step": 975 }, { "epoch": 0.18, "learning_rate": 2.735405747334177e-05, "loss": 1.507, "step": 976 }, { "epoch": 0.18, "learning_rate": 2.7351346466654618e-05, "loss": 1.7832, "step": 977 }, { "epoch": 0.18, "learning_rate": 2.734863545996747e-05, "loss": 1.4104, "step": 978 }, { "epoch": 0.18, "learning_rate": 2.734592445328032e-05, "loss": 0.7787, "step": 979 }, { "epoch": 0.18, "learning_rate": 2.734321344659317e-05, "loss": 1.614, "step": 980 }, { "epoch": 0.18, "learning_rate": 2.7340502439906018e-05, "loss": 0.8666, "step": 981 }, { "epoch": 0.18, "learning_rate": 2.733779143321887e-05, "loss": 2.1354, "step": 982 }, { "epoch": 0.18, "learning_rate": 2.733508042653172e-05, "loss": 0.776, "step": 983 }, { "epoch": 0.18, "learning_rate": 2.733236941984457e-05, "loss": 1.5464, "step": 984 }, { "epoch": 0.18, "learning_rate": 2.732965841315742e-05, "loss": 0.9871, "step": 985 }, { "epoch": 0.18, "learning_rate": 2.732694740647027e-05, "loss": 1.4985, "step": 986 }, { "epoch": 0.18, "learning_rate": 2.7324236399783122e-05, "loss": 0.8762, "step": 987 }, { "epoch": 0.18, "learning_rate": 2.732152539309597e-05, "loss": 1.4693, "step": 988 }, { "epoch": 0.18, "learning_rate": 2.731881438640882e-05, "loss": 1.27, "step": 989 }, { "epoch": 0.18, "learning_rate": 2.731610337972167e-05, "loss": 1.352, "step": 990 }, { "epoch": 0.18, "learning_rate": 2.731339237303452e-05, "loss": 1.769, "step": 991 }, { "epoch": 0.18, "learning_rate": 2.731068136634737e-05, "loss": 1.3887, "step": 992 }, { "epoch": 0.18, "learning_rate": 2.7307970359660223e-05, "loss": 1.2489, "step": 993 }, { "epoch": 0.18, "learning_rate": 2.730525935297307e-05, "loss": 1.4682, "step": 994 }, { "epoch": 0.18, "learning_rate": 2.7302548346285923e-05, "loss": 1.4084, "step": 995 }, { "epoch": 0.18, "learning_rate": 2.7299837339598772e-05, "loss": 1.3116, "step": 996 }, { "epoch": 0.18, "learning_rate": 2.729712633291162e-05, "loss": 0.8455, "step": 997 }, { "epoch": 0.18, "learning_rate": 2.7294415326224472e-05, "loss": 1.6154, "step": 998 }, { "epoch": 0.18, "learning_rate": 2.729170431953732e-05, "loss": 1.2562, "step": 999 }, { "epoch": 0.18, "learning_rate": 2.7288993312850172e-05, "loss": 0.7434, "step": 1000 }, { "epoch": 0.18, "learning_rate": 2.7286282306163024e-05, "loss": 1.3334, "step": 1001 }, { "epoch": 0.18, "learning_rate": 2.7283571299475873e-05, "loss": 0.8564, "step": 1002 }, { "epoch": 0.18, "learning_rate": 2.7280860292788724e-05, "loss": 1.2792, "step": 1003 }, { "epoch": 0.18, "learning_rate": 2.7278149286101576e-05, "loss": 1.4907, "step": 1004 }, { "epoch": 0.18, "learning_rate": 2.727543827941442e-05, "loss": 1.7123, "step": 1005 }, { "epoch": 0.18, "learning_rate": 2.7272727272727273e-05, "loss": 1.6822, "step": 1006 }, { "epoch": 0.18, "learning_rate": 2.727001626604012e-05, "loss": 1.5703, "step": 1007 }, { "epoch": 0.18, "learning_rate": 2.7267305259352973e-05, "loss": 0.9858, "step": 1008 }, { "epoch": 0.18, "learning_rate": 2.7264594252665825e-05, "loss": 1.3807, "step": 1009 }, { "epoch": 0.18, "learning_rate": 2.7261883245978674e-05, "loss": 2.101, "step": 1010 }, { "epoch": 0.18, "learning_rate": 2.7259172239291525e-05, "loss": 1.1209, "step": 1011 }, { "epoch": 0.18, "learning_rate": 2.7256461232604377e-05, "loss": 1.6005, "step": 1012 }, { "epoch": 0.18, "learning_rate": 2.7253750225917222e-05, "loss": 1.7607, "step": 1013 }, { "epoch": 0.18, "learning_rate": 2.7251039219230074e-05, "loss": 0.9062, "step": 1014 }, { "epoch": 0.18, "learning_rate": 2.7248328212542926e-05, "loss": 1.4628, "step": 1015 }, { "epoch": 0.18, "learning_rate": 2.7245617205855774e-05, "loss": 1.9137, "step": 1016 }, { "epoch": 0.18, "learning_rate": 2.7242906199168626e-05, "loss": 1.8804, "step": 1017 }, { "epoch": 0.18, "learning_rate": 2.7240195192481475e-05, "loss": 1.8524, "step": 1018 }, { "epoch": 0.18, "learning_rate": 2.7237484185794327e-05, "loss": 0.9643, "step": 1019 }, { "epoch": 0.18, "learning_rate": 2.723477317910718e-05, "loss": 1.5404, "step": 1020 }, { "epoch": 0.18, "learning_rate": 2.7232062172420023e-05, "loss": 1.7029, "step": 1021 }, { "epoch": 0.18, "learning_rate": 2.7229351165732875e-05, "loss": 1.8739, "step": 1022 }, { "epoch": 0.18, "learning_rate": 2.7226640159045727e-05, "loss": 1.2704, "step": 1023 }, { "epoch": 0.19, "learning_rate": 2.7223929152358576e-05, "loss": 1.7513, "step": 1024 }, { "epoch": 0.19, "learning_rate": 2.7221218145671427e-05, "loss": 1.112, "step": 1025 }, { "epoch": 0.19, "learning_rate": 2.721850713898428e-05, "loss": 1.3223, "step": 1026 }, { "epoch": 0.19, "learning_rate": 2.7215796132297128e-05, "loss": 1.0846, "step": 1027 }, { "epoch": 0.19, "learning_rate": 2.721308512560998e-05, "loss": 1.7419, "step": 1028 }, { "epoch": 0.19, "learning_rate": 2.7210374118922825e-05, "loss": 2.0355, "step": 1029 }, { "epoch": 0.19, "learning_rate": 2.7207663112235676e-05, "loss": 0.8841, "step": 1030 }, { "epoch": 0.19, "learning_rate": 2.7204952105548528e-05, "loss": 0.9158, "step": 1031 }, { "epoch": 0.19, "learning_rate": 2.7202241098861377e-05, "loss": 1.0803, "step": 1032 }, { "epoch": 0.19, "learning_rate": 2.719953009217423e-05, "loss": 1.3277, "step": 1033 }, { "epoch": 0.19, "learning_rate": 2.719681908548708e-05, "loss": 1.2473, "step": 1034 }, { "epoch": 0.19, "learning_rate": 2.719410807879993e-05, "loss": 0.7079, "step": 1035 }, { "epoch": 0.19, "learning_rate": 2.7191397072112777e-05, "loss": 2.4951, "step": 1036 }, { "epoch": 0.19, "learning_rate": 2.718868606542563e-05, "loss": 1.2336, "step": 1037 }, { "epoch": 0.19, "learning_rate": 2.7185975058738477e-05, "loss": 1.4885, "step": 1038 }, { "epoch": 0.19, "learning_rate": 2.718326405205133e-05, "loss": 2.0095, "step": 1039 }, { "epoch": 0.19, "learning_rate": 2.7180553045364178e-05, "loss": 1.6769, "step": 1040 }, { "epoch": 0.19, "learning_rate": 2.717784203867703e-05, "loss": 1.979, "step": 1041 }, { "epoch": 0.19, "learning_rate": 2.717513103198988e-05, "loss": 1.5445, "step": 1042 }, { "epoch": 0.19, "learning_rate": 2.717242002530273e-05, "loss": 1.7232, "step": 1043 }, { "epoch": 0.19, "learning_rate": 2.7169709018615578e-05, "loss": 1.6426, "step": 1044 }, { "epoch": 0.19, "learning_rate": 2.716699801192843e-05, "loss": 1.3445, "step": 1045 }, { "epoch": 0.19, "learning_rate": 2.716428700524128e-05, "loss": 1.1767, "step": 1046 }, { "epoch": 0.19, "learning_rate": 2.716157599855413e-05, "loss": 1.2197, "step": 1047 }, { "epoch": 0.19, "learning_rate": 2.7158864991866982e-05, "loss": 1.3449, "step": 1048 }, { "epoch": 0.19, "learning_rate": 2.715615398517983e-05, "loss": 1.5214, "step": 1049 }, { "epoch": 0.19, "learning_rate": 2.7153442978492682e-05, "loss": 1.0026, "step": 1050 }, { "epoch": 0.19, "learning_rate": 2.715073197180553e-05, "loss": 1.9728, "step": 1051 }, { "epoch": 0.19, "learning_rate": 2.714802096511838e-05, "loss": 0.8824, "step": 1052 }, { "epoch": 0.19, "learning_rate": 2.714530995843123e-05, "loss": 0.957, "step": 1053 }, { "epoch": 0.19, "learning_rate": 2.714259895174408e-05, "loss": 2.0404, "step": 1054 }, { "epoch": 0.19, "learning_rate": 2.713988794505693e-05, "loss": 1.3442, "step": 1055 }, { "epoch": 0.19, "learning_rate": 2.7137176938369783e-05, "loss": 1.8999, "step": 1056 }, { "epoch": 0.19, "learning_rate": 2.713446593168263e-05, "loss": 1.908, "step": 1057 }, { "epoch": 0.19, "learning_rate": 2.7131754924995483e-05, "loss": 1.6524, "step": 1058 }, { "epoch": 0.19, "learning_rate": 2.7129043918308335e-05, "loss": 0.7136, "step": 1059 }, { "epoch": 0.19, "learning_rate": 2.712633291162118e-05, "loss": 1.6288, "step": 1060 }, { "epoch": 0.19, "learning_rate": 2.7123621904934032e-05, "loss": 0.9408, "step": 1061 }, { "epoch": 0.19, "learning_rate": 2.712091089824688e-05, "loss": 1.3893, "step": 1062 }, { "epoch": 0.19, "learning_rate": 2.7118199891559732e-05, "loss": 1.1287, "step": 1063 }, { "epoch": 0.19, "learning_rate": 2.7115488884872584e-05, "loss": 0.9585, "step": 1064 }, { "epoch": 0.19, "learning_rate": 2.7112777878185433e-05, "loss": 1.5622, "step": 1065 }, { "epoch": 0.19, "learning_rate": 2.7110066871498285e-05, "loss": 1.2484, "step": 1066 }, { "epoch": 0.19, "learning_rate": 2.7107355864811136e-05, "loss": 1.4023, "step": 1067 }, { "epoch": 0.19, "learning_rate": 2.710464485812398e-05, "loss": 0.8818, "step": 1068 }, { "epoch": 0.19, "learning_rate": 2.7101933851436833e-05, "loss": 1.3145, "step": 1069 }, { "epoch": 0.19, "learning_rate": 2.7099222844749685e-05, "loss": 0.9571, "step": 1070 }, { "epoch": 0.19, "learning_rate": 2.7096511838062534e-05, "loss": 0.8563, "step": 1071 }, { "epoch": 0.19, "learning_rate": 2.7093800831375385e-05, "loss": 1.3125, "step": 1072 }, { "epoch": 0.19, "learning_rate": 2.7091089824688234e-05, "loss": 1.6771, "step": 1073 }, { "epoch": 0.19, "learning_rate": 2.7088378818001086e-05, "loss": 1.535, "step": 1074 }, { "epoch": 0.19, "learning_rate": 2.7085667811313937e-05, "loss": 1.9488, "step": 1075 }, { "epoch": 0.19, "learning_rate": 2.7082956804626783e-05, "loss": 0.7889, "step": 1076 }, { "epoch": 0.19, "learning_rate": 2.7080245797939634e-05, "loss": 1.361, "step": 1077 }, { "epoch": 0.19, "learning_rate": 2.7077534791252486e-05, "loss": 0.8496, "step": 1078 }, { "epoch": 0.2, "learning_rate": 2.7074823784565335e-05, "loss": 1.3736, "step": 1079 }, { "epoch": 0.2, "learning_rate": 2.7072112777878186e-05, "loss": 1.5377, "step": 1080 }, { "epoch": 0.2, "learning_rate": 2.7069401771191038e-05, "loss": 1.335, "step": 1081 }, { "epoch": 0.2, "learning_rate": 2.7066690764503887e-05, "loss": 1.3907, "step": 1082 }, { "epoch": 0.2, "learning_rate": 2.706397975781674e-05, "loss": 1.8112, "step": 1083 }, { "epoch": 0.2, "learning_rate": 2.7061268751129587e-05, "loss": 1.682, "step": 1084 }, { "epoch": 0.2, "learning_rate": 2.7058557744442435e-05, "loss": 0.5081, "step": 1085 }, { "epoch": 0.2, "learning_rate": 2.7055846737755287e-05, "loss": 1.7465, "step": 1086 }, { "epoch": 0.2, "learning_rate": 2.7053135731068136e-05, "loss": 1.4016, "step": 1087 }, { "epoch": 0.2, "learning_rate": 2.7050424724380988e-05, "loss": 0.9802, "step": 1088 }, { "epoch": 0.2, "learning_rate": 2.704771371769384e-05, "loss": 0.6116, "step": 1089 }, { "epoch": 0.2, "learning_rate": 2.7045002711006688e-05, "loss": 1.599, "step": 1090 }, { "epoch": 0.2, "learning_rate": 2.704229170431954e-05, "loss": 0.9053, "step": 1091 }, { "epoch": 0.2, "learning_rate": 2.7039580697632388e-05, "loss": 1.4653, "step": 1092 }, { "epoch": 0.2, "learning_rate": 2.7036869690945236e-05, "loss": 2.3477, "step": 1093 }, { "epoch": 0.2, "learning_rate": 2.703415868425809e-05, "loss": 1.8853, "step": 1094 }, { "epoch": 0.2, "learning_rate": 2.703144767757094e-05, "loss": 1.2223, "step": 1095 }, { "epoch": 0.2, "learning_rate": 2.702873667088379e-05, "loss": 1.5236, "step": 1096 }, { "epoch": 0.2, "learning_rate": 2.702602566419664e-05, "loss": 0.9324, "step": 1097 }, { "epoch": 0.2, "learning_rate": 2.702331465750949e-05, "loss": 1.582, "step": 1098 }, { "epoch": 0.2, "learning_rate": 2.702060365082234e-05, "loss": 1.0098, "step": 1099 }, { "epoch": 0.2, "learning_rate": 2.701789264413519e-05, "loss": 1.2031, "step": 1100 }, { "epoch": 0.2, "learning_rate": 2.7015181637448038e-05, "loss": 1.5803, "step": 1101 }, { "epoch": 0.2, "learning_rate": 2.701247063076089e-05, "loss": 1.923, "step": 1102 }, { "epoch": 0.2, "learning_rate": 2.700975962407374e-05, "loss": 0.943, "step": 1103 }, { "epoch": 0.2, "learning_rate": 2.700704861738659e-05, "loss": 1.3184, "step": 1104 }, { "epoch": 0.2, "learning_rate": 2.700433761069944e-05, "loss": 1.6709, "step": 1105 }, { "epoch": 0.2, "learning_rate": 2.7001626604012293e-05, "loss": 1.492, "step": 1106 }, { "epoch": 0.2, "learning_rate": 2.6998915597325142e-05, "loss": 1.1729, "step": 1107 }, { "epoch": 0.2, "learning_rate": 2.699620459063799e-05, "loss": 1.1529, "step": 1108 }, { "epoch": 0.2, "learning_rate": 2.699349358395084e-05, "loss": 1.0311, "step": 1109 }, { "epoch": 0.2, "learning_rate": 2.699078257726369e-05, "loss": 1.0418, "step": 1110 }, { "epoch": 0.2, "learning_rate": 2.6988071570576542e-05, "loss": 1.4124, "step": 1111 }, { "epoch": 0.2, "learning_rate": 2.698536056388939e-05, "loss": 1.0457, "step": 1112 }, { "epoch": 0.2, "learning_rate": 2.6982649557202243e-05, "loss": 1.3759, "step": 1113 }, { "epoch": 0.2, "learning_rate": 2.6979938550515094e-05, "loss": 2.1346, "step": 1114 }, { "epoch": 0.2, "learning_rate": 2.6977227543827943e-05, "loss": 1.847, "step": 1115 }, { "epoch": 0.2, "learning_rate": 2.697451653714079e-05, "loss": 1.2934, "step": 1116 }, { "epoch": 0.2, "learning_rate": 2.6971805530453643e-05, "loss": 1.9505, "step": 1117 }, { "epoch": 0.2, "learning_rate": 2.696909452376649e-05, "loss": 0.9823, "step": 1118 }, { "epoch": 0.2, "learning_rate": 2.6966383517079343e-05, "loss": 1.3221, "step": 1119 }, { "epoch": 0.2, "learning_rate": 2.6963672510392192e-05, "loss": 1.4005, "step": 1120 }, { "epoch": 0.2, "learning_rate": 2.6960961503705044e-05, "loss": 1.5384, "step": 1121 }, { "epoch": 0.2, "learning_rate": 2.6958250497017895e-05, "loss": 1.5855, "step": 1122 }, { "epoch": 0.2, "learning_rate": 2.6955539490330744e-05, "loss": 0.9163, "step": 1123 }, { "epoch": 0.2, "learning_rate": 2.6952828483643592e-05, "loss": 1.4976, "step": 1124 }, { "epoch": 0.2, "learning_rate": 2.6950117476956444e-05, "loss": 2.178, "step": 1125 }, { "epoch": 0.2, "learning_rate": 2.6947406470269293e-05, "loss": 1.6179, "step": 1126 }, { "epoch": 0.2, "learning_rate": 2.6944695463582144e-05, "loss": 1.2563, "step": 1127 }, { "epoch": 0.2, "learning_rate": 2.6941984456894996e-05, "loss": 1.0921, "step": 1128 }, { "epoch": 0.2, "learning_rate": 2.6939273450207845e-05, "loss": 1.6534, "step": 1129 }, { "epoch": 0.2, "learning_rate": 2.6936562443520697e-05, "loss": 1.0301, "step": 1130 }, { "epoch": 0.2, "learning_rate": 2.6933851436833545e-05, "loss": 1.6061, "step": 1131 }, { "epoch": 0.2, "learning_rate": 2.6931140430146393e-05, "loss": 1.697, "step": 1132 }, { "epoch": 0.2, "learning_rate": 2.6928429423459245e-05, "loss": 1.8328, "step": 1133 }, { "epoch": 0.2, "learning_rate": 2.6925718416772094e-05, "loss": 1.9209, "step": 1134 }, { "epoch": 0.21, "learning_rate": 2.6923007410084946e-05, "loss": 1.5088, "step": 1135 }, { "epoch": 0.21, "learning_rate": 2.6920296403397797e-05, "loss": 1.1921, "step": 1136 }, { "epoch": 0.21, "learning_rate": 2.6917585396710646e-05, "loss": 1.1036, "step": 1137 }, { "epoch": 0.21, "learning_rate": 2.6914874390023498e-05, "loss": 2.1663, "step": 1138 }, { "epoch": 0.21, "learning_rate": 2.691216338333635e-05, "loss": 1.516, "step": 1139 }, { "epoch": 0.21, "learning_rate": 2.6909452376649194e-05, "loss": 1.9421, "step": 1140 }, { "epoch": 0.21, "learning_rate": 2.6906741369962046e-05, "loss": 1.905, "step": 1141 }, { "epoch": 0.21, "learning_rate": 2.6904030363274895e-05, "loss": 1.1975, "step": 1142 }, { "epoch": 0.21, "learning_rate": 2.6901319356587747e-05, "loss": 1.6965, "step": 1143 }, { "epoch": 0.21, "learning_rate": 2.68986083499006e-05, "loss": 0.8957, "step": 1144 }, { "epoch": 0.21, "learning_rate": 2.6895897343213447e-05, "loss": 0.9554, "step": 1145 }, { "epoch": 0.21, "learning_rate": 2.68931863365263e-05, "loss": 0.8314, "step": 1146 }, { "epoch": 0.21, "learning_rate": 2.689047532983915e-05, "loss": 1.28, "step": 1147 }, { "epoch": 0.21, "learning_rate": 2.6887764323151996e-05, "loss": 1.0887, "step": 1148 }, { "epoch": 0.21, "learning_rate": 2.6885053316464847e-05, "loss": 1.9061, "step": 1149 }, { "epoch": 0.21, "learning_rate": 2.68823423097777e-05, "loss": 1.0683, "step": 1150 }, { "epoch": 0.21, "learning_rate": 2.6879631303090548e-05, "loss": 1.9752, "step": 1151 }, { "epoch": 0.21, "learning_rate": 2.68769202964034e-05, "loss": 1.1224, "step": 1152 }, { "epoch": 0.21, "learning_rate": 2.6874209289716248e-05, "loss": 0.9348, "step": 1153 }, { "epoch": 0.21, "learning_rate": 2.68714982830291e-05, "loss": 0.8522, "step": 1154 }, { "epoch": 0.21, "learning_rate": 2.686878727634195e-05, "loss": 1.6131, "step": 1155 }, { "epoch": 0.21, "learning_rate": 2.6866076269654797e-05, "loss": 0.9252, "step": 1156 }, { "epoch": 0.21, "learning_rate": 2.686336526296765e-05, "loss": 1.4503, "step": 1157 }, { "epoch": 0.21, "learning_rate": 2.68606542562805e-05, "loss": 1.8702, "step": 1158 }, { "epoch": 0.21, "learning_rate": 2.685794324959335e-05, "loss": 0.9522, "step": 1159 }, { "epoch": 0.21, "learning_rate": 2.68552322429062e-05, "loss": 1.861, "step": 1160 }, { "epoch": 0.21, "learning_rate": 2.6852521236219052e-05, "loss": 1.9098, "step": 1161 }, { "epoch": 0.21, "learning_rate": 2.68498102295319e-05, "loss": 2.2332, "step": 1162 }, { "epoch": 0.21, "learning_rate": 2.684709922284475e-05, "loss": 2.073, "step": 1163 }, { "epoch": 0.21, "learning_rate": 2.6844388216157598e-05, "loss": 1.5089, "step": 1164 }, { "epoch": 0.21, "learning_rate": 2.684167720947045e-05, "loss": 1.206, "step": 1165 }, { "epoch": 0.21, "learning_rate": 2.68389662027833e-05, "loss": 1.0134, "step": 1166 }, { "epoch": 0.21, "learning_rate": 2.683625519609615e-05, "loss": 1.2653, "step": 1167 }, { "epoch": 0.21, "learning_rate": 2.6833544189409e-05, "loss": 1.0223, "step": 1168 }, { "epoch": 0.21, "learning_rate": 2.6830833182721853e-05, "loss": 1.7447, "step": 1169 }, { "epoch": 0.21, "learning_rate": 2.6828122176034702e-05, "loss": 1.1116, "step": 1170 }, { "epoch": 0.21, "learning_rate": 2.682541116934755e-05, "loss": 0.7189, "step": 1171 }, { "epoch": 0.21, "learning_rate": 2.6822700162660402e-05, "loss": 1.3053, "step": 1172 }, { "epoch": 0.21, "learning_rate": 2.681998915597325e-05, "loss": 1.6461, "step": 1173 }, { "epoch": 0.21, "learning_rate": 2.6817278149286102e-05, "loss": 1.5068, "step": 1174 }, { "epoch": 0.21, "learning_rate": 2.681456714259895e-05, "loss": 1.1268, "step": 1175 }, { "epoch": 0.21, "learning_rate": 2.6811856135911803e-05, "loss": 1.3778, "step": 1176 }, { "epoch": 0.21, "learning_rate": 2.6809145129224655e-05, "loss": 1.1914, "step": 1177 }, { "epoch": 0.21, "learning_rate": 2.6806434122537503e-05, "loss": 1.0249, "step": 1178 }, { "epoch": 0.21, "learning_rate": 2.680372311585035e-05, "loss": 1.0466, "step": 1179 }, { "epoch": 0.21, "learning_rate": 2.6801012109163203e-05, "loss": 1.3174, "step": 1180 }, { "epoch": 0.21, "learning_rate": 2.679830110247605e-05, "loss": 1.4967, "step": 1181 }, { "epoch": 0.21, "learning_rate": 2.6795590095788904e-05, "loss": 0.7624, "step": 1182 }, { "epoch": 0.21, "learning_rate": 2.6792879089101755e-05, "loss": 1.4425, "step": 1183 }, { "epoch": 0.21, "learning_rate": 2.6790168082414604e-05, "loss": 1.2821, "step": 1184 }, { "epoch": 0.21, "learning_rate": 2.6787457075727456e-05, "loss": 1.7487, "step": 1185 }, { "epoch": 0.21, "learning_rate": 2.6784746069040304e-05, "loss": 1.9901, "step": 1186 }, { "epoch": 0.21, "learning_rate": 2.6782035062353152e-05, "loss": 1.4772, "step": 1187 }, { "epoch": 0.21, "learning_rate": 2.6779324055666004e-05, "loss": 1.4579, "step": 1188 }, { "epoch": 0.21, "learning_rate": 2.6776613048978853e-05, "loss": 1.5285, "step": 1189 }, { "epoch": 0.22, "learning_rate": 2.6773902042291705e-05, "loss": 1.1125, "step": 1190 }, { "epoch": 0.22, "learning_rate": 2.6771191035604556e-05, "loss": 1.4084, "step": 1191 }, { "epoch": 0.22, "learning_rate": 2.6768480028917405e-05, "loss": 1.3202, "step": 1192 }, { "epoch": 0.22, "learning_rate": 2.6765769022230257e-05, "loss": 1.1488, "step": 1193 }, { "epoch": 0.22, "learning_rate": 2.676305801554311e-05, "loss": 1.8258, "step": 1194 }, { "epoch": 0.22, "learning_rate": 2.6760347008855954e-05, "loss": 1.9122, "step": 1195 }, { "epoch": 0.22, "learning_rate": 2.6757636002168805e-05, "loss": 2.2394, "step": 1196 }, { "epoch": 0.22, "learning_rate": 2.6754924995481654e-05, "loss": 2.0075, "step": 1197 }, { "epoch": 0.22, "learning_rate": 2.6752213988794506e-05, "loss": 1.1875, "step": 1198 }, { "epoch": 0.22, "learning_rate": 2.6749502982107357e-05, "loss": 1.4799, "step": 1199 }, { "epoch": 0.22, "learning_rate": 2.6746791975420206e-05, "loss": 0.9844, "step": 1200 }, { "epoch": 0.22, "learning_rate": 2.6744080968733058e-05, "loss": 1.3212, "step": 1201 }, { "epoch": 0.22, "learning_rate": 2.674136996204591e-05, "loss": 1.3499, "step": 1202 }, { "epoch": 0.22, "learning_rate": 2.6738658955358755e-05, "loss": 1.0824, "step": 1203 }, { "epoch": 0.22, "learning_rate": 2.6735947948671606e-05, "loss": 1.8892, "step": 1204 }, { "epoch": 0.22, "learning_rate": 2.6733236941984458e-05, "loss": 0.8448, "step": 1205 }, { "epoch": 0.22, "learning_rate": 2.6730525935297307e-05, "loss": 1.2644, "step": 1206 }, { "epoch": 0.22, "learning_rate": 2.672781492861016e-05, "loss": 1.6783, "step": 1207 }, { "epoch": 0.22, "learning_rate": 2.6725103921923007e-05, "loss": 1.2407, "step": 1208 }, { "epoch": 0.22, "learning_rate": 2.672239291523586e-05, "loss": 1.6852, "step": 1209 }, { "epoch": 0.22, "learning_rate": 2.671968190854871e-05, "loss": 1.4024, "step": 1210 }, { "epoch": 0.22, "learning_rate": 2.6716970901861556e-05, "loss": 1.0861, "step": 1211 }, { "epoch": 0.22, "learning_rate": 2.6714259895174408e-05, "loss": 1.2643, "step": 1212 }, { "epoch": 0.22, "learning_rate": 2.671154888848726e-05, "loss": 1.5964, "step": 1213 }, { "epoch": 0.22, "learning_rate": 2.6708837881800108e-05, "loss": 1.2262, "step": 1214 }, { "epoch": 0.22, "learning_rate": 2.670612687511296e-05, "loss": 1.8245, "step": 1215 }, { "epoch": 0.22, "learning_rate": 2.670341586842581e-05, "loss": 1.4097, "step": 1216 }, { "epoch": 0.22, "learning_rate": 2.670070486173866e-05, "loss": 1.0879, "step": 1217 }, { "epoch": 0.22, "learning_rate": 2.6697993855051512e-05, "loss": 1.4111, "step": 1218 }, { "epoch": 0.22, "learning_rate": 2.669528284836436e-05, "loss": 1.3538, "step": 1219 }, { "epoch": 0.22, "learning_rate": 2.669257184167721e-05, "loss": 1.0038, "step": 1220 }, { "epoch": 0.22, "learning_rate": 2.668986083499006e-05, "loss": 1.5316, "step": 1221 }, { "epoch": 0.22, "learning_rate": 2.668714982830291e-05, "loss": 1.6324, "step": 1222 }, { "epoch": 0.22, "learning_rate": 2.668443882161576e-05, "loss": 1.3163, "step": 1223 }, { "epoch": 0.22, "learning_rate": 2.6681727814928613e-05, "loss": 1.0891, "step": 1224 }, { "epoch": 0.22, "learning_rate": 2.667901680824146e-05, "loss": 1.0024, "step": 1225 }, { "epoch": 0.22, "learning_rate": 2.6676305801554313e-05, "loss": 1.4381, "step": 1226 }, { "epoch": 0.22, "learning_rate": 2.667359479486716e-05, "loss": 1.3101, "step": 1227 }, { "epoch": 0.22, "learning_rate": 2.667088378818001e-05, "loss": 1.7932, "step": 1228 }, { "epoch": 0.22, "learning_rate": 2.666817278149286e-05, "loss": 1.0972, "step": 1229 }, { "epoch": 0.22, "learning_rate": 2.6665461774805713e-05, "loss": 1.2447, "step": 1230 }, { "epoch": 0.22, "learning_rate": 2.6662750768118562e-05, "loss": 0.8819, "step": 1231 }, { "epoch": 0.22, "learning_rate": 2.6660039761431414e-05, "loss": 1.0214, "step": 1232 }, { "epoch": 0.22, "learning_rate": 2.6657328754744262e-05, "loss": 0.9914, "step": 1233 }, { "epoch": 0.22, "learning_rate": 2.6654617748057114e-05, "loss": 1.1109, "step": 1234 }, { "epoch": 0.22, "learning_rate": 2.6651906741369962e-05, "loss": 1.5488, "step": 1235 }, { "epoch": 0.22, "learning_rate": 2.664919573468281e-05, "loss": 1.078, "step": 1236 }, { "epoch": 0.22, "learning_rate": 2.6646484727995663e-05, "loss": 0.8268, "step": 1237 }, { "epoch": 0.22, "learning_rate": 2.6643773721308514e-05, "loss": 1.0925, "step": 1238 }, { "epoch": 0.22, "learning_rate": 2.6641062714621363e-05, "loss": 0.9388, "step": 1239 }, { "epoch": 0.22, "learning_rate": 2.6638351707934215e-05, "loss": 1.371, "step": 1240 }, { "epoch": 0.22, "learning_rate": 2.6635640701247067e-05, "loss": 0.8691, "step": 1241 }, { "epoch": 0.22, "learning_rate": 2.6632929694559915e-05, "loss": 1.2595, "step": 1242 }, { "epoch": 0.22, "learning_rate": 2.6630218687872763e-05, "loss": 1.4102, "step": 1243 }, { "epoch": 0.22, "learning_rate": 2.6627507681185612e-05, "loss": 0.8254, "step": 1244 }, { "epoch": 0.23, "learning_rate": 2.6624796674498464e-05, "loss": 1.061, "step": 1245 }, { "epoch": 0.23, "learning_rate": 2.6622085667811316e-05, "loss": 1.4701, "step": 1246 }, { "epoch": 0.23, "learning_rate": 2.6619374661124164e-05, "loss": 1.245, "step": 1247 }, { "epoch": 0.23, "learning_rate": 2.6616663654437016e-05, "loss": 2.1855, "step": 1248 }, { "epoch": 0.23, "learning_rate": 2.6613952647749868e-05, "loss": 0.9895, "step": 1249 }, { "epoch": 0.23, "learning_rate": 2.6611241641062716e-05, "loss": 1.1953, "step": 1250 }, { "epoch": 0.23, "learning_rate": 2.6608530634375564e-05, "loss": 1.4532, "step": 1251 }, { "epoch": 0.23, "learning_rate": 2.6605819627688416e-05, "loss": 1.4985, "step": 1252 }, { "epoch": 0.23, "learning_rate": 2.6603108621001265e-05, "loss": 2.0284, "step": 1253 }, { "epoch": 0.23, "learning_rate": 2.6600397614314117e-05, "loss": 1.5249, "step": 1254 }, { "epoch": 0.23, "learning_rate": 2.6597686607626965e-05, "loss": 1.1197, "step": 1255 }, { "epoch": 0.23, "learning_rate": 2.6594975600939817e-05, "loss": 0.9072, "step": 1256 }, { "epoch": 0.23, "learning_rate": 2.659226459425267e-05, "loss": 0.8468, "step": 1257 }, { "epoch": 0.23, "learning_rate": 2.6589553587565517e-05, "loss": 1.1874, "step": 1258 }, { "epoch": 0.23, "learning_rate": 2.6586842580878366e-05, "loss": 1.0292, "step": 1259 }, { "epoch": 0.23, "learning_rate": 2.6584131574191217e-05, "loss": 1.2354, "step": 1260 }, { "epoch": 0.23, "learning_rate": 2.6581420567504066e-05, "loss": 1.3056, "step": 1261 }, { "epoch": 0.23, "learning_rate": 2.6578709560816918e-05, "loss": 2.0188, "step": 1262 }, { "epoch": 0.23, "learning_rate": 2.657599855412977e-05, "loss": 0.7933, "step": 1263 }, { "epoch": 0.23, "learning_rate": 2.6573287547442618e-05, "loss": 1.28, "step": 1264 }, { "epoch": 0.23, "learning_rate": 2.657057654075547e-05, "loss": 1.1875, "step": 1265 }, { "epoch": 0.23, "learning_rate": 2.6567865534068318e-05, "loss": 1.0247, "step": 1266 }, { "epoch": 0.23, "learning_rate": 2.6565154527381167e-05, "loss": 1.1614, "step": 1267 }, { "epoch": 0.23, "learning_rate": 2.656244352069402e-05, "loss": 0.9148, "step": 1268 }, { "epoch": 0.23, "learning_rate": 2.6559732514006867e-05, "loss": 1.2753, "step": 1269 }, { "epoch": 0.23, "learning_rate": 2.655702150731972e-05, "loss": 0.5297, "step": 1270 }, { "epoch": 0.23, "learning_rate": 2.655431050063257e-05, "loss": 1.4993, "step": 1271 }, { "epoch": 0.23, "learning_rate": 2.655159949394542e-05, "loss": 2.1229, "step": 1272 }, { "epoch": 0.23, "learning_rate": 2.654888848725827e-05, "loss": 1.0702, "step": 1273 }, { "epoch": 0.23, "learning_rate": 2.6546177480571123e-05, "loss": 1.2908, "step": 1274 }, { "epoch": 0.23, "learning_rate": 2.6543466473883968e-05, "loss": 1.9813, "step": 1275 }, { "epoch": 0.23, "learning_rate": 2.654075546719682e-05, "loss": 1.4393, "step": 1276 }, { "epoch": 0.23, "learning_rate": 2.6538044460509668e-05, "loss": 1.1724, "step": 1277 }, { "epoch": 0.23, "learning_rate": 2.653533345382252e-05, "loss": 1.5605, "step": 1278 }, { "epoch": 0.23, "learning_rate": 2.653262244713537e-05, "loss": 1.5062, "step": 1279 }, { "epoch": 0.23, "learning_rate": 2.652991144044822e-05, "loss": 1.3566, "step": 1280 }, { "epoch": 0.23, "learning_rate": 2.6527200433761072e-05, "loss": 1.2801, "step": 1281 }, { "epoch": 0.23, "learning_rate": 2.6524489427073924e-05, "loss": 0.8925, "step": 1282 }, { "epoch": 0.23, "learning_rate": 2.652177842038677e-05, "loss": 0.8406, "step": 1283 }, { "epoch": 0.23, "learning_rate": 2.651906741369962e-05, "loss": 1.0333, "step": 1284 }, { "epoch": 0.23, "learning_rate": 2.6516356407012472e-05, "loss": 1.152, "step": 1285 }, { "epoch": 0.23, "learning_rate": 2.651364540032532e-05, "loss": 1.0095, "step": 1286 }, { "epoch": 0.23, "learning_rate": 2.6510934393638173e-05, "loss": 1.8366, "step": 1287 }, { "epoch": 0.23, "learning_rate": 2.650822338695102e-05, "loss": 2.0164, "step": 1288 }, { "epoch": 0.23, "learning_rate": 2.6505512380263873e-05, "loss": 1.1124, "step": 1289 }, { "epoch": 0.23, "learning_rate": 2.650280137357672e-05, "loss": 0.7654, "step": 1290 }, { "epoch": 0.23, "learning_rate": 2.650009036688957e-05, "loss": 0.7203, "step": 1291 }, { "epoch": 0.23, "learning_rate": 2.649737936020242e-05, "loss": 2.5863, "step": 1292 }, { "epoch": 0.23, "learning_rate": 2.6494668353515274e-05, "loss": 1.2758, "step": 1293 }, { "epoch": 0.23, "learning_rate": 2.6491957346828122e-05, "loss": 1.7567, "step": 1294 }, { "epoch": 0.23, "learning_rate": 2.6489246340140974e-05, "loss": 1.4353, "step": 1295 }, { "epoch": 0.23, "learning_rate": 2.6486535333453826e-05, "loss": 0.8867, "step": 1296 }, { "epoch": 0.23, "learning_rate": 2.6483824326766674e-05, "loss": 1.6953, "step": 1297 }, { "epoch": 0.23, "learning_rate": 2.6481113320079522e-05, "loss": 0.8214, "step": 1298 }, { "epoch": 0.23, "learning_rate": 2.647840231339237e-05, "loss": 1.3896, "step": 1299 }, { "epoch": 0.23, "learning_rate": 2.6475691306705223e-05, "loss": 1.4595, "step": 1300 }, { "epoch": 0.24, "learning_rate": 2.6472980300018075e-05, "loss": 1.9674, "step": 1301 }, { "epoch": 0.24, "learning_rate": 2.6470269293330923e-05, "loss": 1.0965, "step": 1302 }, { "epoch": 0.24, "learning_rate": 2.6467558286643775e-05, "loss": 1.2373, "step": 1303 }, { "epoch": 0.24, "learning_rate": 2.6464847279956627e-05, "loss": 1.1647, "step": 1304 }, { "epoch": 0.24, "learning_rate": 2.6462136273269475e-05, "loss": 2.2518, "step": 1305 }, { "epoch": 0.24, "learning_rate": 2.6459425266582324e-05, "loss": 0.8376, "step": 1306 }, { "epoch": 0.24, "learning_rate": 2.6456714259895175e-05, "loss": 1.6597, "step": 1307 }, { "epoch": 0.24, "learning_rate": 2.6454003253208024e-05, "loss": 1.349, "step": 1308 }, { "epoch": 0.24, "learning_rate": 2.6451292246520876e-05, "loss": 1.9598, "step": 1309 }, { "epoch": 0.24, "learning_rate": 2.6448581239833724e-05, "loss": 1.052, "step": 1310 }, { "epoch": 0.24, "learning_rate": 2.6445870233146576e-05, "loss": 1.3689, "step": 1311 }, { "epoch": 0.24, "learning_rate": 2.6443159226459428e-05, "loss": 1.8317, "step": 1312 }, { "epoch": 0.24, "learning_rate": 2.6440448219772276e-05, "loss": 1.5849, "step": 1313 }, { "epoch": 0.24, "learning_rate": 2.6437737213085125e-05, "loss": 1.1807, "step": 1314 }, { "epoch": 0.24, "learning_rate": 2.6435026206397976e-05, "loss": 1.09, "step": 1315 }, { "epoch": 0.24, "learning_rate": 2.6432315199710825e-05, "loss": 1.5653, "step": 1316 }, { "epoch": 0.24, "learning_rate": 2.6429604193023677e-05, "loss": 1.5569, "step": 1317 }, { "epoch": 0.24, "learning_rate": 2.642689318633653e-05, "loss": 0.8231, "step": 1318 }, { "epoch": 0.24, "learning_rate": 2.6424182179649377e-05, "loss": 1.1209, "step": 1319 }, { "epoch": 0.24, "learning_rate": 2.642147117296223e-05, "loss": 1.4046, "step": 1320 }, { "epoch": 0.24, "learning_rate": 2.6418760166275077e-05, "loss": 1.4829, "step": 1321 }, { "epoch": 0.24, "learning_rate": 2.6416049159587926e-05, "loss": 1.1787, "step": 1322 }, { "epoch": 0.24, "learning_rate": 2.6413338152900778e-05, "loss": 1.0707, "step": 1323 }, { "epoch": 0.24, "learning_rate": 2.6410627146213626e-05, "loss": 1.1956, "step": 1324 }, { "epoch": 0.24, "learning_rate": 2.6407916139526478e-05, "loss": 0.9727, "step": 1325 }, { "epoch": 0.24, "learning_rate": 2.640520513283933e-05, "loss": 1.1307, "step": 1326 }, { "epoch": 0.24, "learning_rate": 2.6402494126152178e-05, "loss": 1.4066, "step": 1327 }, { "epoch": 0.24, "learning_rate": 2.639978311946503e-05, "loss": 1.4925, "step": 1328 }, { "epoch": 0.24, "learning_rate": 2.6397072112777882e-05, "loss": 1.9837, "step": 1329 }, { "epoch": 0.24, "learning_rate": 2.6394361106090727e-05, "loss": 1.4772, "step": 1330 }, { "epoch": 0.24, "learning_rate": 2.639165009940358e-05, "loss": 1.2627, "step": 1331 }, { "epoch": 0.24, "learning_rate": 2.6388939092716427e-05, "loss": 0.8261, "step": 1332 }, { "epoch": 0.24, "learning_rate": 2.638622808602928e-05, "loss": 1.6376, "step": 1333 }, { "epoch": 0.24, "learning_rate": 2.638351707934213e-05, "loss": 0.6412, "step": 1334 }, { "epoch": 0.24, "learning_rate": 2.638080607265498e-05, "loss": 1.3523, "step": 1335 }, { "epoch": 0.24, "learning_rate": 2.637809506596783e-05, "loss": 1.0674, "step": 1336 }, { "epoch": 0.24, "learning_rate": 2.6375384059280683e-05, "loss": 1.9176, "step": 1337 }, { "epoch": 0.24, "learning_rate": 2.6372673052593528e-05, "loss": 2.0115, "step": 1338 }, { "epoch": 0.24, "learning_rate": 2.636996204590638e-05, "loss": 1.4501, "step": 1339 }, { "epoch": 0.24, "learning_rate": 2.636725103921923e-05, "loss": 1.5736, "step": 1340 }, { "epoch": 0.24, "learning_rate": 2.636454003253208e-05, "loss": 1.3871, "step": 1341 }, { "epoch": 0.24, "learning_rate": 2.6361829025844932e-05, "loss": 1.9154, "step": 1342 }, { "epoch": 0.24, "learning_rate": 2.635911801915778e-05, "loss": 0.7863, "step": 1343 }, { "epoch": 0.24, "learning_rate": 2.6356407012470632e-05, "loss": 1.9283, "step": 1344 }, { "epoch": 0.24, "learning_rate": 2.6353696005783484e-05, "loss": 1.3438, "step": 1345 }, { "epoch": 0.24, "learning_rate": 2.635098499909633e-05, "loss": 0.5046, "step": 1346 }, { "epoch": 0.24, "learning_rate": 2.634827399240918e-05, "loss": 0.7495, "step": 1347 }, { "epoch": 0.24, "learning_rate": 2.6345562985722033e-05, "loss": 1.5145, "step": 1348 }, { "epoch": 0.24, "learning_rate": 2.634285197903488e-05, "loss": 0.4123, "step": 1349 }, { "epoch": 0.24, "learning_rate": 2.6340140972347733e-05, "loss": 0.8501, "step": 1350 }, { "epoch": 0.24, "learning_rate": 2.6337429965660585e-05, "loss": 0.6387, "step": 1351 }, { "epoch": 0.24, "learning_rate": 2.6334718958973433e-05, "loss": 0.7562, "step": 1352 }, { "epoch": 0.24, "learning_rate": 2.6332007952286285e-05, "loss": 1.0448, "step": 1353 }, { "epoch": 0.24, "learning_rate": 2.6329296945599133e-05, "loss": 1.9441, "step": 1354 }, { "epoch": 0.24, "learning_rate": 2.6326585938911982e-05, "loss": 1.8611, "step": 1355 }, { "epoch": 0.25, "learning_rate": 2.6323874932224834e-05, "loss": 1.6296, "step": 1356 }, { "epoch": 0.25, "learning_rate": 2.6321163925537682e-05, "loss": 1.5682, "step": 1357 }, { "epoch": 0.25, "learning_rate": 2.6318452918850534e-05, "loss": 1.6131, "step": 1358 }, { "epoch": 0.25, "learning_rate": 2.6315741912163386e-05, "loss": 1.8757, "step": 1359 }, { "epoch": 0.25, "learning_rate": 2.6313030905476234e-05, "loss": 1.0363, "step": 1360 }, { "epoch": 0.25, "learning_rate": 2.6310319898789086e-05, "loss": 1.3865, "step": 1361 }, { "epoch": 0.25, "learning_rate": 2.6307608892101934e-05, "loss": 1.8389, "step": 1362 }, { "epoch": 0.25, "learning_rate": 2.6304897885414783e-05, "loss": 1.6124, "step": 1363 }, { "epoch": 0.25, "learning_rate": 2.6302186878727635e-05, "loss": 1.4154, "step": 1364 }, { "epoch": 0.25, "learning_rate": 2.6299475872040487e-05, "loss": 2.0021, "step": 1365 }, { "epoch": 0.25, "learning_rate": 2.6296764865353335e-05, "loss": 1.7975, "step": 1366 }, { "epoch": 0.25, "learning_rate": 2.6294053858666187e-05, "loss": 1.0141, "step": 1367 }, { "epoch": 0.25, "learning_rate": 2.6291342851979035e-05, "loss": 1.1708, "step": 1368 }, { "epoch": 0.25, "learning_rate": 2.6288631845291887e-05, "loss": 1.3817, "step": 1369 }, { "epoch": 0.25, "learning_rate": 2.6285920838604736e-05, "loss": 1.4273, "step": 1370 }, { "epoch": 0.25, "learning_rate": 2.6283209831917584e-05, "loss": 1.7679, "step": 1371 }, { "epoch": 0.25, "learning_rate": 2.6280498825230436e-05, "loss": 0.9169, "step": 1372 }, { "epoch": 0.25, "learning_rate": 2.6277787818543288e-05, "loss": 1.2078, "step": 1373 }, { "epoch": 0.25, "learning_rate": 2.6275076811856136e-05, "loss": 2.0918, "step": 1374 }, { "epoch": 0.25, "learning_rate": 2.6272365805168988e-05, "loss": 1.2816, "step": 1375 }, { "epoch": 0.25, "learning_rate": 2.626965479848184e-05, "loss": 1.0514, "step": 1376 }, { "epoch": 0.25, "learning_rate": 2.6266943791794688e-05, "loss": 1.1679, "step": 1377 }, { "epoch": 0.25, "learning_rate": 2.6264232785107537e-05, "loss": 1.5958, "step": 1378 }, { "epoch": 0.25, "learning_rate": 2.6261521778420385e-05, "loss": 1.0676, "step": 1379 }, { "epoch": 0.25, "learning_rate": 2.6258810771733237e-05, "loss": 0.6821, "step": 1380 }, { "epoch": 0.25, "learning_rate": 2.625609976504609e-05, "loss": 1.8319, "step": 1381 }, { "epoch": 0.25, "learning_rate": 2.6253388758358937e-05, "loss": 0.8483, "step": 1382 }, { "epoch": 0.25, "learning_rate": 2.625067775167179e-05, "loss": 0.8749, "step": 1383 }, { "epoch": 0.25, "learning_rate": 2.624796674498464e-05, "loss": 2.2219, "step": 1384 }, { "epoch": 0.25, "learning_rate": 2.624525573829749e-05, "loss": 0.8713, "step": 1385 }, { "epoch": 0.25, "learning_rate": 2.6242544731610338e-05, "loss": 1.4106, "step": 1386 }, { "epoch": 0.25, "learning_rate": 2.623983372492319e-05, "loss": 0.9499, "step": 1387 }, { "epoch": 0.25, "learning_rate": 2.6237122718236038e-05, "loss": 1.3781, "step": 1388 }, { "epoch": 0.25, "learning_rate": 2.623441171154889e-05, "loss": 0.8493, "step": 1389 }, { "epoch": 0.25, "learning_rate": 2.6231700704861738e-05, "loss": 1.3811, "step": 1390 }, { "epoch": 0.25, "learning_rate": 2.622898969817459e-05, "loss": 1.2714, "step": 1391 }, { "epoch": 0.25, "learning_rate": 2.6226278691487442e-05, "loss": 2.0452, "step": 1392 }, { "epoch": 0.25, "learning_rate": 2.622356768480029e-05, "loss": 1.2902, "step": 1393 }, { "epoch": 0.25, "learning_rate": 2.622085667811314e-05, "loss": 1.3841, "step": 1394 }, { "epoch": 0.25, "learning_rate": 2.621814567142599e-05, "loss": 1.3565, "step": 1395 }, { "epoch": 0.25, "learning_rate": 2.621543466473884e-05, "loss": 1.8285, "step": 1396 }, { "epoch": 0.25, "learning_rate": 2.621272365805169e-05, "loss": 1.5335, "step": 1397 }, { "epoch": 0.25, "learning_rate": 2.6210012651364543e-05, "loss": 0.7227, "step": 1398 }, { "epoch": 0.25, "learning_rate": 2.620730164467739e-05, "loss": 0.9507, "step": 1399 }, { "epoch": 0.25, "learning_rate": 2.6204590637990243e-05, "loss": 0.7399, "step": 1400 }, { "epoch": 0.25, "learning_rate": 2.620187963130309e-05, "loss": 1.2909, "step": 1401 }, { "epoch": 0.25, "learning_rate": 2.619916862461594e-05, "loss": 1.5417, "step": 1402 }, { "epoch": 0.25, "learning_rate": 2.619645761792879e-05, "loss": 1.1104, "step": 1403 }, { "epoch": 0.25, "learning_rate": 2.619374661124164e-05, "loss": 1.9094, "step": 1404 }, { "epoch": 0.25, "learning_rate": 2.6191035604554492e-05, "loss": 1.2441, "step": 1405 }, { "epoch": 0.25, "learning_rate": 2.6188324597867344e-05, "loss": 1.2958, "step": 1406 }, { "epoch": 0.25, "learning_rate": 2.6185613591180192e-05, "loss": 0.9607, "step": 1407 }, { "epoch": 0.25, "learning_rate": 2.6182902584493044e-05, "loss": 1.4696, "step": 1408 }, { "epoch": 0.25, "learning_rate": 2.6180191577805896e-05, "loss": 1.0358, "step": 1409 }, { "epoch": 0.25, "learning_rate": 2.617748057111874e-05, "loss": 1.3348, "step": 1410 }, { "epoch": 0.26, "learning_rate": 2.6174769564431593e-05, "loss": 1.31, "step": 1411 }, { "epoch": 0.26, "learning_rate": 2.617205855774444e-05, "loss": 1.0976, "step": 1412 }, { "epoch": 0.26, "learning_rate": 2.6169347551057293e-05, "loss": 1.4541, "step": 1413 }, { "epoch": 0.26, "learning_rate": 2.6166636544370145e-05, "loss": 1.4559, "step": 1414 }, { "epoch": 0.26, "learning_rate": 2.6163925537682993e-05, "loss": 1.1864, "step": 1415 }, { "epoch": 0.26, "learning_rate": 2.6161214530995845e-05, "loss": 1.9255, "step": 1416 }, { "epoch": 0.26, "learning_rate": 2.6158503524308694e-05, "loss": 0.9856, "step": 1417 }, { "epoch": 0.26, "learning_rate": 2.6155792517621542e-05, "loss": 1.7671, "step": 1418 }, { "epoch": 0.26, "learning_rate": 2.6153081510934394e-05, "loss": 1.5877, "step": 1419 }, { "epoch": 0.26, "learning_rate": 2.6150370504247246e-05, "loss": 0.8024, "step": 1420 }, { "epoch": 0.26, "learning_rate": 2.6147659497560094e-05, "loss": 1.7569, "step": 1421 }, { "epoch": 0.26, "learning_rate": 2.6144948490872946e-05, "loss": 1.365, "step": 1422 }, { "epoch": 0.26, "learning_rate": 2.6142237484185794e-05, "loss": 2.0304, "step": 1423 }, { "epoch": 0.26, "learning_rate": 2.6139526477498646e-05, "loss": 1.2768, "step": 1424 }, { "epoch": 0.26, "learning_rate": 2.6136815470811495e-05, "loss": 0.9408, "step": 1425 }, { "epoch": 0.26, "learning_rate": 2.6134104464124343e-05, "loss": 0.7061, "step": 1426 }, { "epoch": 0.26, "learning_rate": 2.6131393457437195e-05, "loss": 1.2166, "step": 1427 }, { "epoch": 0.26, "learning_rate": 2.6128682450750047e-05, "loss": 1.5715, "step": 1428 }, { "epoch": 0.26, "learning_rate": 2.6125971444062895e-05, "loss": 2.7555, "step": 1429 }, { "epoch": 0.26, "learning_rate": 2.6123260437375747e-05, "loss": 0.9139, "step": 1430 }, { "epoch": 0.26, "learning_rate": 2.61205494306886e-05, "loss": 1.1008, "step": 1431 }, { "epoch": 0.26, "learning_rate": 2.6117838424001447e-05, "loss": 1.5912, "step": 1432 }, { "epoch": 0.26, "learning_rate": 2.6115127417314296e-05, "loss": 1.3287, "step": 1433 }, { "epoch": 0.26, "learning_rate": 2.6112416410627144e-05, "loss": 2.138, "step": 1434 }, { "epoch": 0.26, "learning_rate": 2.6109705403939996e-05, "loss": 0.8436, "step": 1435 }, { "epoch": 0.26, "learning_rate": 2.6106994397252848e-05, "loss": 1.3273, "step": 1436 }, { "epoch": 0.26, "learning_rate": 2.6104283390565696e-05, "loss": 1.1466, "step": 1437 }, { "epoch": 0.26, "learning_rate": 2.6101572383878548e-05, "loss": 1.1747, "step": 1438 }, { "epoch": 0.26, "learning_rate": 2.60988613771914e-05, "loss": 1.2587, "step": 1439 }, { "epoch": 0.26, "learning_rate": 2.609615037050425e-05, "loss": 1.3306, "step": 1440 }, { "epoch": 0.26, "learning_rate": 2.6093439363817097e-05, "loss": 1.1549, "step": 1441 }, { "epoch": 0.26, "learning_rate": 2.609072835712995e-05, "loss": 1.1906, "step": 1442 }, { "epoch": 0.26, "learning_rate": 2.6088017350442797e-05, "loss": 1.6197, "step": 1443 }, { "epoch": 0.26, "learning_rate": 2.608530634375565e-05, "loss": 0.8611, "step": 1444 }, { "epoch": 0.26, "learning_rate": 2.6082595337068497e-05, "loss": 1.6689, "step": 1445 }, { "epoch": 0.26, "learning_rate": 2.607988433038135e-05, "loss": 1.7255, "step": 1446 }, { "epoch": 0.26, "learning_rate": 2.60771733236942e-05, "loss": 1.3371, "step": 1447 }, { "epoch": 0.26, "learning_rate": 2.607446231700705e-05, "loss": 0.9383, "step": 1448 }, { "epoch": 0.26, "learning_rate": 2.6071751310319898e-05, "loss": 1.434, "step": 1449 }, { "epoch": 0.26, "learning_rate": 2.606904030363275e-05, "loss": 0.9466, "step": 1450 }, { "epoch": 0.26, "learning_rate": 2.6066329296945598e-05, "loss": 2.1412, "step": 1451 }, { "epoch": 0.26, "learning_rate": 2.606361829025845e-05, "loss": 1.3303, "step": 1452 }, { "epoch": 0.26, "learning_rate": 2.6060907283571302e-05, "loss": 0.8075, "step": 1453 }, { "epoch": 0.26, "learning_rate": 2.605819627688415e-05, "loss": 1.1731, "step": 1454 }, { "epoch": 0.26, "learning_rate": 2.6055485270197002e-05, "loss": 0.5022, "step": 1455 }, { "epoch": 0.26, "learning_rate": 2.605277426350985e-05, "loss": 0.9909, "step": 1456 }, { "epoch": 0.26, "learning_rate": 2.60500632568227e-05, "loss": 1.6177, "step": 1457 }, { "epoch": 0.26, "learning_rate": 2.604735225013555e-05, "loss": 1.1802, "step": 1458 }, { "epoch": 0.26, "learning_rate": 2.60446412434484e-05, "loss": 1.3656, "step": 1459 }, { "epoch": 0.26, "learning_rate": 2.604193023676125e-05, "loss": 1.8931, "step": 1460 }, { "epoch": 0.26, "learning_rate": 2.6039219230074103e-05, "loss": 0.672, "step": 1461 }, { "epoch": 0.26, "learning_rate": 2.603650822338695e-05, "loss": 1.729, "step": 1462 }, { "epoch": 0.26, "learning_rate": 2.6033797216699803e-05, "loss": 1.6297, "step": 1463 }, { "epoch": 0.26, "learning_rate": 2.6031086210012655e-05, "loss": 0.8893, "step": 1464 }, { "epoch": 0.26, "learning_rate": 2.60283752033255e-05, "loss": 0.9511, "step": 1465 }, { "epoch": 0.26, "learning_rate": 2.6025664196638352e-05, "loss": 0.4248, "step": 1466 }, { "epoch": 0.27, "learning_rate": 2.60229531899512e-05, "loss": 1.9539, "step": 1467 }, { "epoch": 0.27, "learning_rate": 2.6020242183264052e-05, "loss": 1.0707, "step": 1468 }, { "epoch": 0.27, "learning_rate": 2.6017531176576904e-05, "loss": 1.0558, "step": 1469 }, { "epoch": 0.27, "learning_rate": 2.6014820169889752e-05, "loss": 0.5778, "step": 1470 }, { "epoch": 0.27, "learning_rate": 2.6012109163202604e-05, "loss": 1.0383, "step": 1471 }, { "epoch": 0.27, "learning_rate": 2.6009398156515456e-05, "loss": 1.1809, "step": 1472 }, { "epoch": 0.27, "learning_rate": 2.60066871498283e-05, "loss": 1.4051, "step": 1473 }, { "epoch": 0.27, "learning_rate": 2.6003976143141153e-05, "loss": 1.8138, "step": 1474 }, { "epoch": 0.27, "learning_rate": 2.6001265136454005e-05, "loss": 1.853, "step": 1475 }, { "epoch": 0.27, "learning_rate": 2.5998554129766853e-05, "loss": 1.9002, "step": 1476 }, { "epoch": 0.27, "learning_rate": 2.5995843123079705e-05, "loss": 1.767, "step": 1477 }, { "epoch": 0.27, "learning_rate": 2.5993132116392553e-05, "loss": 1.3136, "step": 1478 }, { "epoch": 0.27, "learning_rate": 2.5990421109705405e-05, "loss": 1.0388, "step": 1479 }, { "epoch": 0.27, "learning_rate": 2.5987710103018257e-05, "loss": 1.7101, "step": 1480 }, { "epoch": 0.27, "learning_rate": 2.5984999096331102e-05, "loss": 1.7014, "step": 1481 }, { "epoch": 0.27, "learning_rate": 2.5982288089643954e-05, "loss": 1.1828, "step": 1482 }, { "epoch": 0.27, "learning_rate": 2.5979577082956806e-05, "loss": 1.4049, "step": 1483 }, { "epoch": 0.27, "learning_rate": 2.5976866076269654e-05, "loss": 1.4467, "step": 1484 }, { "epoch": 0.27, "learning_rate": 2.5974155069582506e-05, "loss": 1.7342, "step": 1485 }, { "epoch": 0.27, "learning_rate": 2.5971444062895358e-05, "loss": 0.7352, "step": 1486 }, { "epoch": 0.27, "learning_rate": 2.5968733056208206e-05, "loss": 1.4479, "step": 1487 }, { "epoch": 0.27, "learning_rate": 2.5966022049521058e-05, "loss": 1.5758, "step": 1488 }, { "epoch": 0.27, "learning_rate": 2.5963311042833903e-05, "loss": 1.0552, "step": 1489 }, { "epoch": 0.27, "learning_rate": 2.5960600036146755e-05, "loss": 0.7282, "step": 1490 }, { "epoch": 0.27, "learning_rate": 2.5957889029459607e-05, "loss": 1.3718, "step": 1491 }, { "epoch": 0.27, "learning_rate": 2.5955178022772455e-05, "loss": 1.5486, "step": 1492 }, { "epoch": 0.27, "learning_rate": 2.5952467016085307e-05, "loss": 0.9, "step": 1493 }, { "epoch": 0.27, "learning_rate": 2.594975600939816e-05, "loss": 1.4181, "step": 1494 }, { "epoch": 0.27, "learning_rate": 2.5947045002711007e-05, "loss": 1.9002, "step": 1495 }, { "epoch": 0.27, "learning_rate": 2.594433399602386e-05, "loss": 1.3293, "step": 1496 }, { "epoch": 0.27, "learning_rate": 2.5941622989336708e-05, "loss": 0.9657, "step": 1497 }, { "epoch": 0.27, "learning_rate": 2.5938911982649556e-05, "loss": 0.8766, "step": 1498 }, { "epoch": 0.27, "learning_rate": 2.5936200975962408e-05, "loss": 1.6083, "step": 1499 }, { "epoch": 0.27, "learning_rate": 2.5933489969275256e-05, "loss": 1.307, "step": 1500 }, { "epoch": 0.27, "learning_rate": 2.5930778962588108e-05, "loss": 2.0437, "step": 1501 }, { "epoch": 0.27, "learning_rate": 2.592806795590096e-05, "loss": 1.7398, "step": 1502 }, { "epoch": 0.27, "learning_rate": 2.592535694921381e-05, "loss": 1.4841, "step": 1503 }, { "epoch": 0.27, "learning_rate": 2.592264594252666e-05, "loss": 1.5207, "step": 1504 }, { "epoch": 0.27, "learning_rate": 2.591993493583951e-05, "loss": 1.4371, "step": 1505 }, { "epoch": 0.27, "learning_rate": 2.5917223929152357e-05, "loss": 2.246, "step": 1506 }, { "epoch": 0.27, "learning_rate": 2.591451292246521e-05, "loss": 1.0463, "step": 1507 }, { "epoch": 0.27, "learning_rate": 2.591180191577806e-05, "loss": 1.6371, "step": 1508 }, { "epoch": 0.27, "learning_rate": 2.590909090909091e-05, "loss": 0.7876, "step": 1509 }, { "epoch": 0.27, "learning_rate": 2.590637990240376e-05, "loss": 1.4542, "step": 1510 }, { "epoch": 0.27, "learning_rate": 2.590366889571661e-05, "loss": 0.9741, "step": 1511 }, { "epoch": 0.27, "learning_rate": 2.590095788902946e-05, "loss": 1.3681, "step": 1512 }, { "epoch": 0.27, "learning_rate": 2.589824688234231e-05, "loss": 1.1525, "step": 1513 }, { "epoch": 0.27, "learning_rate": 2.5895535875655158e-05, "loss": 1.5836, "step": 1514 }, { "epoch": 0.27, "learning_rate": 2.589282486896801e-05, "loss": 0.9936, "step": 1515 }, { "epoch": 0.27, "learning_rate": 2.5890113862280862e-05, "loss": 1.2944, "step": 1516 }, { "epoch": 0.27, "learning_rate": 2.588740285559371e-05, "loss": 1.1131, "step": 1517 }, { "epoch": 0.27, "learning_rate": 2.5884691848906562e-05, "loss": 1.1701, "step": 1518 }, { "epoch": 0.27, "learning_rate": 2.5881980842219414e-05, "loss": 0.5133, "step": 1519 }, { "epoch": 0.27, "learning_rate": 2.5879269835532262e-05, "loss": 1.4879, "step": 1520 }, { "epoch": 0.27, "learning_rate": 2.587655882884511e-05, "loss": 1.6739, "step": 1521 }, { "epoch": 0.28, "learning_rate": 2.5873847822157963e-05, "loss": 1.3926, "step": 1522 }, { "epoch": 0.28, "learning_rate": 2.587113681547081e-05, "loss": 1.2939, "step": 1523 }, { "epoch": 0.28, "learning_rate": 2.5868425808783663e-05, "loss": 1.4108, "step": 1524 }, { "epoch": 0.28, "learning_rate": 2.586571480209651e-05, "loss": 1.5889, "step": 1525 }, { "epoch": 0.28, "learning_rate": 2.5863003795409363e-05, "loss": 1.0803, "step": 1526 }, { "epoch": 0.28, "learning_rate": 2.5860292788722215e-05, "loss": 1.0482, "step": 1527 }, { "epoch": 0.28, "learning_rate": 2.5857581782035064e-05, "loss": 1.6213, "step": 1528 }, { "epoch": 0.28, "learning_rate": 2.5854870775347912e-05, "loss": 1.1227, "step": 1529 }, { "epoch": 0.28, "learning_rate": 2.5852159768660764e-05, "loss": 0.9129, "step": 1530 }, { "epoch": 0.28, "learning_rate": 2.5849448761973612e-05, "loss": 0.8621, "step": 1531 }, { "epoch": 0.28, "learning_rate": 2.5846737755286464e-05, "loss": 1.1604, "step": 1532 }, { "epoch": 0.28, "learning_rate": 2.5844026748599316e-05, "loss": 0.7889, "step": 1533 }, { "epoch": 0.28, "learning_rate": 2.5841315741912164e-05, "loss": 1.8673, "step": 1534 }, { "epoch": 0.28, "learning_rate": 2.5838604735225016e-05, "loss": 2.1102, "step": 1535 }, { "epoch": 0.28, "learning_rate": 2.5835893728537865e-05, "loss": 1.1475, "step": 1536 }, { "epoch": 0.28, "learning_rate": 2.5833182721850713e-05, "loss": 1.0998, "step": 1537 }, { "epoch": 0.28, "learning_rate": 2.5830471715163565e-05, "loss": 1.7855, "step": 1538 }, { "epoch": 0.28, "learning_rate": 2.5827760708476413e-05, "loss": 1.3715, "step": 1539 }, { "epoch": 0.28, "learning_rate": 2.5825049701789265e-05, "loss": 1.0533, "step": 1540 }, { "epoch": 0.28, "learning_rate": 2.5822338695102117e-05, "loss": 1.7408, "step": 1541 }, { "epoch": 0.28, "learning_rate": 2.5819627688414965e-05, "loss": 1.4924, "step": 1542 }, { "epoch": 0.28, "learning_rate": 2.5816916681727817e-05, "loss": 0.7911, "step": 1543 }, { "epoch": 0.28, "learning_rate": 2.581420567504067e-05, "loss": 1.1504, "step": 1544 }, { "epoch": 0.28, "learning_rate": 2.5811494668353514e-05, "loss": 1.1141, "step": 1545 }, { "epoch": 0.28, "learning_rate": 2.5808783661666366e-05, "loss": 2.1897, "step": 1546 }, { "epoch": 0.28, "learning_rate": 2.5806072654979214e-05, "loss": 1.5049, "step": 1547 }, { "epoch": 0.28, "learning_rate": 2.5803361648292066e-05, "loss": 1.0368, "step": 1548 }, { "epoch": 0.28, "learning_rate": 2.5800650641604918e-05, "loss": 0.7496, "step": 1549 }, { "epoch": 0.28, "learning_rate": 2.5797939634917766e-05, "loss": 1.7828, "step": 1550 }, { "epoch": 0.28, "learning_rate": 2.579522862823062e-05, "loss": 0.9959, "step": 1551 }, { "epoch": 0.28, "learning_rate": 2.5792517621543467e-05, "loss": 1.267, "step": 1552 }, { "epoch": 0.28, "learning_rate": 2.5789806614856315e-05, "loss": 1.0095, "step": 1553 }, { "epoch": 0.28, "learning_rate": 2.5787095608169167e-05, "loss": 1.2605, "step": 1554 }, { "epoch": 0.28, "learning_rate": 2.578438460148202e-05, "loss": 1.3202, "step": 1555 }, { "epoch": 0.28, "learning_rate": 2.5781673594794867e-05, "loss": 0.8702, "step": 1556 }, { "epoch": 0.28, "learning_rate": 2.577896258810772e-05, "loss": 0.6592, "step": 1557 }, { "epoch": 0.28, "learning_rate": 2.5776251581420568e-05, "loss": 1.8501, "step": 1558 }, { "epoch": 0.28, "learning_rate": 2.577354057473342e-05, "loss": 1.4974, "step": 1559 }, { "epoch": 0.28, "learning_rate": 2.5770829568046268e-05, "loss": 1.663, "step": 1560 }, { "epoch": 0.28, "learning_rate": 2.5768118561359116e-05, "loss": 2.0877, "step": 1561 }, { "epoch": 0.28, "learning_rate": 2.5765407554671968e-05, "loss": 2.8446, "step": 1562 }, { "epoch": 0.28, "learning_rate": 2.576269654798482e-05, "loss": 1.1022, "step": 1563 }, { "epoch": 0.28, "learning_rate": 2.575998554129767e-05, "loss": 1.0946, "step": 1564 }, { "epoch": 0.28, "learning_rate": 2.575727453461052e-05, "loss": 1.3637, "step": 1565 }, { "epoch": 0.28, "learning_rate": 2.5754563527923372e-05, "loss": 0.9731, "step": 1566 }, { "epoch": 0.28, "learning_rate": 2.575185252123622e-05, "loss": 1.2922, "step": 1567 }, { "epoch": 0.28, "learning_rate": 2.574914151454907e-05, "loss": 0.9852, "step": 1568 }, { "epoch": 0.28, "learning_rate": 2.5746430507861917e-05, "loss": 1.0115, "step": 1569 }, { "epoch": 0.28, "learning_rate": 2.574371950117477e-05, "loss": 1.4487, "step": 1570 }, { "epoch": 0.28, "learning_rate": 2.574100849448762e-05, "loss": 1.6086, "step": 1571 }, { "epoch": 0.28, "learning_rate": 2.573829748780047e-05, "loss": 1.045, "step": 1572 }, { "epoch": 0.28, "learning_rate": 2.573558648111332e-05, "loss": 1.1803, "step": 1573 }, { "epoch": 0.28, "learning_rate": 2.5732875474426173e-05, "loss": 1.5205, "step": 1574 }, { "epoch": 0.28, "learning_rate": 2.573016446773902e-05, "loss": 2.2361, "step": 1575 }, { "epoch": 0.28, "learning_rate": 2.572745346105187e-05, "loss": 1.1503, "step": 1576 }, { "epoch": 0.29, "learning_rate": 2.5724742454364722e-05, "loss": 0.7773, "step": 1577 }, { "epoch": 0.29, "learning_rate": 2.572203144767757e-05, "loss": 1.3313, "step": 1578 }, { "epoch": 0.29, "learning_rate": 2.5719320440990422e-05, "loss": 1.8346, "step": 1579 }, { "epoch": 0.29, "learning_rate": 2.571660943430327e-05, "loss": 1.4168, "step": 1580 }, { "epoch": 0.29, "learning_rate": 2.5713898427616122e-05, "loss": 0.9021, "step": 1581 }, { "epoch": 0.29, "learning_rate": 2.5711187420928974e-05, "loss": 1.5324, "step": 1582 }, { "epoch": 0.29, "learning_rate": 2.5708476414241823e-05, "loss": 1.3366, "step": 1583 }, { "epoch": 0.29, "learning_rate": 2.570576540755467e-05, "loss": 1.1465, "step": 1584 }, { "epoch": 0.29, "learning_rate": 2.5703054400867523e-05, "loss": 1.3531, "step": 1585 }, { "epoch": 0.29, "learning_rate": 2.570034339418037e-05, "loss": 1.4026, "step": 1586 }, { "epoch": 0.29, "learning_rate": 2.5697632387493223e-05, "loss": 1.4479, "step": 1587 }, { "epoch": 0.29, "learning_rate": 2.5694921380806075e-05, "loss": 1.3254, "step": 1588 }, { "epoch": 0.29, "learning_rate": 2.5692210374118923e-05, "loss": 2.1051, "step": 1589 }, { "epoch": 0.29, "learning_rate": 2.5689499367431775e-05, "loss": 0.8256, "step": 1590 }, { "epoch": 0.29, "learning_rate": 2.5686788360744624e-05, "loss": 1.4373, "step": 1591 }, { "epoch": 0.29, "learning_rate": 2.5684077354057472e-05, "loss": 1.4133, "step": 1592 }, { "epoch": 0.29, "learning_rate": 2.5681366347370324e-05, "loss": 0.8527, "step": 1593 }, { "epoch": 0.29, "learning_rate": 2.5678655340683172e-05, "loss": 1.702, "step": 1594 }, { "epoch": 0.29, "learning_rate": 2.5675944333996024e-05, "loss": 1.4186, "step": 1595 }, { "epoch": 0.29, "learning_rate": 2.5673233327308876e-05, "loss": 1.2115, "step": 1596 }, { "epoch": 0.29, "learning_rate": 2.5670522320621724e-05, "loss": 0.9229, "step": 1597 }, { "epoch": 0.29, "learning_rate": 2.5667811313934576e-05, "loss": 1.1023, "step": 1598 }, { "epoch": 0.29, "learning_rate": 2.5665100307247428e-05, "loss": 0.7672, "step": 1599 }, { "epoch": 0.29, "learning_rate": 2.5662389300560273e-05, "loss": 1.118, "step": 1600 }, { "epoch": 0.29, "learning_rate": 2.5659678293873125e-05, "loss": 1.4962, "step": 1601 }, { "epoch": 0.29, "learning_rate": 2.5656967287185973e-05, "loss": 1.2252, "step": 1602 }, { "epoch": 0.29, "learning_rate": 2.5654256280498825e-05, "loss": 0.8598, "step": 1603 }, { "epoch": 0.29, "learning_rate": 2.5651545273811677e-05, "loss": 0.945, "step": 1604 }, { "epoch": 0.29, "learning_rate": 2.5648834267124526e-05, "loss": 0.7467, "step": 1605 }, { "epoch": 0.29, "learning_rate": 2.5646123260437377e-05, "loss": 0.7556, "step": 1606 }, { "epoch": 0.29, "learning_rate": 2.564341225375023e-05, "loss": 1.7143, "step": 1607 }, { "epoch": 0.29, "learning_rate": 2.5640701247063074e-05, "loss": 2.3025, "step": 1608 }, { "epoch": 0.29, "learning_rate": 2.5637990240375926e-05, "loss": 1.1177, "step": 1609 }, { "epoch": 0.29, "learning_rate": 2.5635279233688778e-05, "loss": 1.1454, "step": 1610 }, { "epoch": 0.29, "learning_rate": 2.5632568227001626e-05, "loss": 1.7139, "step": 1611 }, { "epoch": 0.29, "learning_rate": 2.5629857220314478e-05, "loss": 0.9625, "step": 1612 }, { "epoch": 0.29, "learning_rate": 2.5627146213627327e-05, "loss": 1.3888, "step": 1613 }, { "epoch": 0.29, "learning_rate": 2.562443520694018e-05, "loss": 1.1835, "step": 1614 }, { "epoch": 0.29, "learning_rate": 2.562172420025303e-05, "loss": 0.7513, "step": 1615 }, { "epoch": 0.29, "learning_rate": 2.5619013193565875e-05, "loss": 1.4343, "step": 1616 }, { "epoch": 0.29, "learning_rate": 2.5616302186878727e-05, "loss": 0.7465, "step": 1617 }, { "epoch": 0.29, "learning_rate": 2.561359118019158e-05, "loss": 1.5541, "step": 1618 }, { "epoch": 0.29, "learning_rate": 2.5610880173504427e-05, "loss": 1.3451, "step": 1619 }, { "epoch": 0.29, "learning_rate": 2.560816916681728e-05, "loss": 1.7839, "step": 1620 }, { "epoch": 0.29, "learning_rate": 2.560545816013013e-05, "loss": 1.2889, "step": 1621 }, { "epoch": 0.29, "learning_rate": 2.560274715344298e-05, "loss": 1.6942, "step": 1622 }, { "epoch": 0.29, "learning_rate": 2.560003614675583e-05, "loss": 1.0189, "step": 1623 }, { "epoch": 0.29, "learning_rate": 2.5597325140068676e-05, "loss": 1.1684, "step": 1624 }, { "epoch": 0.29, "learning_rate": 2.5594614133381528e-05, "loss": 0.9459, "step": 1625 }, { "epoch": 0.29, "learning_rate": 2.559190312669438e-05, "loss": 0.7816, "step": 1626 }, { "epoch": 0.29, "learning_rate": 2.558919212000723e-05, "loss": 1.4363, "step": 1627 }, { "epoch": 0.29, "learning_rate": 2.558648111332008e-05, "loss": 1.203, "step": 1628 }, { "epoch": 0.29, "learning_rate": 2.5583770106632932e-05, "loss": 1.1942, "step": 1629 }, { "epoch": 0.29, "learning_rate": 2.558105909994578e-05, "loss": 1.3758, "step": 1630 }, { "epoch": 0.29, "learning_rate": 2.5578348093258632e-05, "loss": 1.5777, "step": 1631 }, { "epoch": 0.29, "learning_rate": 2.557563708657148e-05, "loss": 1.136, "step": 1632 }, { "epoch": 0.3, "learning_rate": 2.557292607988433e-05, "loss": 0.7505, "step": 1633 }, { "epoch": 0.3, "learning_rate": 2.557021507319718e-05, "loss": 0.903, "step": 1634 }, { "epoch": 0.3, "learning_rate": 2.556750406651003e-05, "loss": 0.463, "step": 1635 }, { "epoch": 0.3, "learning_rate": 2.556479305982288e-05, "loss": 1.1016, "step": 1636 }, { "epoch": 0.3, "learning_rate": 2.5562082053135733e-05, "loss": 1.791, "step": 1637 }, { "epoch": 0.3, "learning_rate": 2.555937104644858e-05, "loss": 1.457, "step": 1638 }, { "epoch": 0.3, "learning_rate": 2.5556660039761434e-05, "loss": 1.2548, "step": 1639 }, { "epoch": 0.3, "learning_rate": 2.5553949033074282e-05, "loss": 1.2888, "step": 1640 }, { "epoch": 0.3, "learning_rate": 2.555123802638713e-05, "loss": 1.6987, "step": 1641 }, { "epoch": 0.3, "learning_rate": 2.5548527019699982e-05, "loss": 1.0535, "step": 1642 }, { "epoch": 0.3, "learning_rate": 2.5545816013012834e-05, "loss": 0.8335, "step": 1643 }, { "epoch": 0.3, "learning_rate": 2.5543105006325682e-05, "loss": 1.862, "step": 1644 }, { "epoch": 0.3, "learning_rate": 2.5540393999638534e-05, "loss": 0.9572, "step": 1645 }, { "epoch": 0.3, "learning_rate": 2.5537682992951383e-05, "loss": 1.3104, "step": 1646 }, { "epoch": 0.3, "learning_rate": 2.5534971986264235e-05, "loss": 1.5589, "step": 1647 }, { "epoch": 0.3, "learning_rate": 2.5532260979577083e-05, "loss": 1.9529, "step": 1648 }, { "epoch": 0.3, "learning_rate": 2.552954997288993e-05, "loss": 1.2147, "step": 1649 }, { "epoch": 0.3, "learning_rate": 2.5526838966202783e-05, "loss": 1.0551, "step": 1650 }, { "epoch": 0.3, "learning_rate": 2.5524127959515635e-05, "loss": 1.9553, "step": 1651 }, { "epoch": 0.3, "learning_rate": 2.5521416952828484e-05, "loss": 1.006, "step": 1652 }, { "epoch": 0.3, "learning_rate": 2.5518705946141335e-05, "loss": 1.4315, "step": 1653 }, { "epoch": 0.3, "learning_rate": 2.5515994939454187e-05, "loss": 1.7222, "step": 1654 }, { "epoch": 0.3, "learning_rate": 2.5513283932767036e-05, "loss": 1.4023, "step": 1655 }, { "epoch": 0.3, "learning_rate": 2.5510572926079884e-05, "loss": 0.8351, "step": 1656 }, { "epoch": 0.3, "learning_rate": 2.5507861919392736e-05, "loss": 1.6844, "step": 1657 }, { "epoch": 0.3, "learning_rate": 2.5505150912705584e-05, "loss": 1.6053, "step": 1658 }, { "epoch": 0.3, "learning_rate": 2.5502439906018436e-05, "loss": 1.1303, "step": 1659 }, { "epoch": 0.3, "learning_rate": 2.5499728899331285e-05, "loss": 1.0493, "step": 1660 }, { "epoch": 0.3, "learning_rate": 2.5497017892644136e-05, "loss": 0.8867, "step": 1661 }, { "epoch": 0.3, "learning_rate": 2.5494306885956988e-05, "loss": 1.4348, "step": 1662 }, { "epoch": 0.3, "learning_rate": 2.5491595879269837e-05, "loss": 1.0324, "step": 1663 }, { "epoch": 0.3, "learning_rate": 2.5488884872582685e-05, "loss": 0.9931, "step": 1664 }, { "epoch": 0.3, "learning_rate": 2.5486173865895537e-05, "loss": 1.0841, "step": 1665 }, { "epoch": 0.3, "learning_rate": 2.5483462859208385e-05, "loss": 0.7782, "step": 1666 }, { "epoch": 0.3, "learning_rate": 2.5480751852521237e-05, "loss": 1.2662, "step": 1667 }, { "epoch": 0.3, "learning_rate": 2.547804084583409e-05, "loss": 1.0697, "step": 1668 }, { "epoch": 0.3, "learning_rate": 2.5475329839146938e-05, "loss": 1.1349, "step": 1669 }, { "epoch": 0.3, "learning_rate": 2.547261883245979e-05, "loss": 1.0086, "step": 1670 }, { "epoch": 0.3, "learning_rate": 2.5469907825772638e-05, "loss": 1.5121, "step": 1671 }, { "epoch": 0.3, "learning_rate": 2.5467196819085486e-05, "loss": 1.0966, "step": 1672 }, { "epoch": 0.3, "learning_rate": 2.5464485812398338e-05, "loss": 2.0329, "step": 1673 }, { "epoch": 0.3, "learning_rate": 2.5461774805711187e-05, "loss": 1.0444, "step": 1674 }, { "epoch": 0.3, "learning_rate": 2.545906379902404e-05, "loss": 0.9935, "step": 1675 }, { "epoch": 0.3, "learning_rate": 2.545635279233689e-05, "loss": 1.078, "step": 1676 }, { "epoch": 0.3, "learning_rate": 2.545364178564974e-05, "loss": 1.0357, "step": 1677 }, { "epoch": 0.3, "learning_rate": 2.545093077896259e-05, "loss": 1.2109, "step": 1678 }, { "epoch": 0.3, "learning_rate": 2.544821977227544e-05, "loss": 0.7156, "step": 1679 }, { "epoch": 0.3, "learning_rate": 2.5445508765588287e-05, "loss": 0.9496, "step": 1680 }, { "epoch": 0.3, "learning_rate": 2.544279775890114e-05, "loss": 0.8757, "step": 1681 }, { "epoch": 0.3, "learning_rate": 2.5440086752213988e-05, "loss": 1.7017, "step": 1682 }, { "epoch": 0.3, "learning_rate": 2.543737574552684e-05, "loss": 1.4691, "step": 1683 }, { "epoch": 0.3, "learning_rate": 2.543466473883969e-05, "loss": 1.0146, "step": 1684 }, { "epoch": 0.3, "learning_rate": 2.543195373215254e-05, "loss": 1.2964, "step": 1685 }, { "epoch": 0.3, "learning_rate": 2.542924272546539e-05, "loss": 1.0002, "step": 1686 }, { "epoch": 0.3, "learning_rate": 2.542653171877824e-05, "loss": 0.7505, "step": 1687 }, { "epoch": 0.31, "learning_rate": 2.542382071209109e-05, "loss": 1.5425, "step": 1688 }, { "epoch": 0.31, "learning_rate": 2.542110970540394e-05, "loss": 0.8198, "step": 1689 }, { "epoch": 0.31, "learning_rate": 2.5418398698716792e-05, "loss": 1.5075, "step": 1690 }, { "epoch": 0.31, "learning_rate": 2.541568769202964e-05, "loss": 1.1314, "step": 1691 }, { "epoch": 0.31, "learning_rate": 2.5412976685342492e-05, "loss": 1.1758, "step": 1692 }, { "epoch": 0.31, "learning_rate": 2.541026567865534e-05, "loss": 1.2981, "step": 1693 }, { "epoch": 0.31, "learning_rate": 2.5407554671968193e-05, "loss": 1.0044, "step": 1694 }, { "epoch": 0.31, "learning_rate": 2.540484366528104e-05, "loss": 1.2013, "step": 1695 }, { "epoch": 0.31, "learning_rate": 2.540213265859389e-05, "loss": 1.4362, "step": 1696 }, { "epoch": 0.31, "learning_rate": 2.539942165190674e-05, "loss": 0.8256, "step": 1697 }, { "epoch": 0.31, "learning_rate": 2.5396710645219593e-05, "loss": 0.9782, "step": 1698 }, { "epoch": 0.31, "learning_rate": 2.539399963853244e-05, "loss": 1.2352, "step": 1699 }, { "epoch": 0.31, "learning_rate": 2.5391288631845293e-05, "loss": 1.5273, "step": 1700 }, { "epoch": 0.31, "learning_rate": 2.5388577625158145e-05, "loss": 1.8071, "step": 1701 }, { "epoch": 0.31, "learning_rate": 2.5385866618470994e-05, "loss": 1.7243, "step": 1702 }, { "epoch": 0.31, "learning_rate": 2.5383155611783842e-05, "loss": 1.3206, "step": 1703 }, { "epoch": 0.31, "learning_rate": 2.538044460509669e-05, "loss": 0.9009, "step": 1704 }, { "epoch": 0.31, "learning_rate": 2.5377733598409542e-05, "loss": 1.4974, "step": 1705 }, { "epoch": 0.31, "learning_rate": 2.5375022591722394e-05, "loss": 0.9846, "step": 1706 }, { "epoch": 0.31, "learning_rate": 2.5372311585035243e-05, "loss": 0.9099, "step": 1707 }, { "epoch": 0.31, "learning_rate": 2.5369600578348094e-05, "loss": 1.3303, "step": 1708 }, { "epoch": 0.31, "learning_rate": 2.5366889571660946e-05, "loss": 2.2947, "step": 1709 }, { "epoch": 0.31, "learning_rate": 2.5364178564973795e-05, "loss": 0.7815, "step": 1710 }, { "epoch": 0.31, "learning_rate": 2.5361467558286643e-05, "loss": 0.9099, "step": 1711 }, { "epoch": 0.31, "learning_rate": 2.5358756551599495e-05, "loss": 1.1867, "step": 1712 }, { "epoch": 0.31, "learning_rate": 2.5356045544912343e-05, "loss": 1.3073, "step": 1713 }, { "epoch": 0.31, "learning_rate": 2.5353334538225195e-05, "loss": 1.4253, "step": 1714 }, { "epoch": 0.31, "learning_rate": 2.5350623531538044e-05, "loss": 1.2457, "step": 1715 }, { "epoch": 0.31, "learning_rate": 2.5347912524850896e-05, "loss": 1.3308, "step": 1716 }, { "epoch": 0.31, "learning_rate": 2.5345201518163747e-05, "loss": 0.9309, "step": 1717 }, { "epoch": 0.31, "learning_rate": 2.5342490511476596e-05, "loss": 0.9241, "step": 1718 }, { "epoch": 0.31, "learning_rate": 2.5339779504789444e-05, "loss": 0.8722, "step": 1719 }, { "epoch": 0.31, "learning_rate": 2.5337068498102296e-05, "loss": 1.0257, "step": 1720 }, { "epoch": 0.31, "learning_rate": 2.5334357491415145e-05, "loss": 0.8382, "step": 1721 }, { "epoch": 0.31, "learning_rate": 2.5331646484727996e-05, "loss": 1.2003, "step": 1722 }, { "epoch": 0.31, "learning_rate": 2.5328935478040848e-05, "loss": 1.2713, "step": 1723 }, { "epoch": 0.31, "learning_rate": 2.5326224471353697e-05, "loss": 1.3793, "step": 1724 }, { "epoch": 0.31, "learning_rate": 2.532351346466655e-05, "loss": 1.5769, "step": 1725 }, { "epoch": 0.31, "learning_rate": 2.5320802457979397e-05, "loss": 1.2718, "step": 1726 }, { "epoch": 0.31, "learning_rate": 2.5318091451292245e-05, "loss": 1.0315, "step": 1727 }, { "epoch": 0.31, "learning_rate": 2.5315380444605097e-05, "loss": 1.3639, "step": 1728 }, { "epoch": 0.31, "learning_rate": 2.5312669437917946e-05, "loss": 1.5833, "step": 1729 }, { "epoch": 0.31, "learning_rate": 2.5309958431230797e-05, "loss": 1.7803, "step": 1730 }, { "epoch": 0.31, "learning_rate": 2.530724742454365e-05, "loss": 1.1188, "step": 1731 }, { "epoch": 0.31, "learning_rate": 2.5304536417856498e-05, "loss": 1.0086, "step": 1732 }, { "epoch": 0.31, "learning_rate": 2.530182541116935e-05, "loss": 0.9217, "step": 1733 }, { "epoch": 0.31, "learning_rate": 2.52991144044822e-05, "loss": 1.2276, "step": 1734 }, { "epoch": 0.31, "learning_rate": 2.5296403397795046e-05, "loss": 0.9087, "step": 1735 }, { "epoch": 0.31, "learning_rate": 2.5293692391107898e-05, "loss": 0.9575, "step": 1736 }, { "epoch": 0.31, "learning_rate": 2.5290981384420747e-05, "loss": 0.974, "step": 1737 }, { "epoch": 0.31, "learning_rate": 2.52882703777336e-05, "loss": 1.4562, "step": 1738 }, { "epoch": 0.31, "learning_rate": 2.528555937104645e-05, "loss": 1.1581, "step": 1739 }, { "epoch": 0.31, "learning_rate": 2.52828483643593e-05, "loss": 1.4369, "step": 1740 }, { "epoch": 0.31, "learning_rate": 2.528013735767215e-05, "loss": 0.6451, "step": 1741 }, { "epoch": 0.31, "learning_rate": 2.5277426350985002e-05, "loss": 0.9616, "step": 1742 }, { "epoch": 0.32, "learning_rate": 2.5274715344297847e-05, "loss": 0.7368, "step": 1743 }, { "epoch": 0.32, "learning_rate": 2.52720043376107e-05, "loss": 1.0667, "step": 1744 }, { "epoch": 0.32, "learning_rate": 2.526929333092355e-05, "loss": 1.3296, "step": 1745 }, { "epoch": 0.32, "learning_rate": 2.52665823242364e-05, "loss": 1.0468, "step": 1746 }, { "epoch": 0.32, "learning_rate": 2.526387131754925e-05, "loss": 0.5652, "step": 1747 }, { "epoch": 0.32, "learning_rate": 2.52611603108621e-05, "loss": 0.957, "step": 1748 }, { "epoch": 0.32, "learning_rate": 2.525844930417495e-05, "loss": 0.9995, "step": 1749 }, { "epoch": 0.32, "learning_rate": 2.5255738297487804e-05, "loss": 2.2533, "step": 1750 }, { "epoch": 0.32, "learning_rate": 2.525302729080065e-05, "loss": 0.9014, "step": 1751 }, { "epoch": 0.32, "learning_rate": 2.52503162841135e-05, "loss": 1.4253, "step": 1752 }, { "epoch": 0.32, "learning_rate": 2.5247605277426352e-05, "loss": 1.7122, "step": 1753 }, { "epoch": 0.32, "learning_rate": 2.52448942707392e-05, "loss": 1.2387, "step": 1754 }, { "epoch": 0.32, "learning_rate": 2.5242183264052052e-05, "loss": 1.1144, "step": 1755 }, { "epoch": 0.32, "learning_rate": 2.5239472257364904e-05, "loss": 1.3889, "step": 1756 }, { "epoch": 0.32, "learning_rate": 2.5236761250677753e-05, "loss": 1.3719, "step": 1757 }, { "epoch": 0.32, "learning_rate": 2.5234050243990605e-05, "loss": 0.6333, "step": 1758 }, { "epoch": 0.32, "learning_rate": 2.523133923730345e-05, "loss": 1.4543, "step": 1759 }, { "epoch": 0.32, "learning_rate": 2.52286282306163e-05, "loss": 1.0149, "step": 1760 }, { "epoch": 0.32, "learning_rate": 2.5225917223929153e-05, "loss": 0.7018, "step": 1761 }, { "epoch": 0.32, "learning_rate": 2.5223206217242002e-05, "loss": 0.9371, "step": 1762 }, { "epoch": 0.32, "learning_rate": 2.5220495210554854e-05, "loss": 0.7286, "step": 1763 }, { "epoch": 0.32, "learning_rate": 2.5217784203867705e-05, "loss": 0.9994, "step": 1764 }, { "epoch": 0.32, "learning_rate": 2.5215073197180554e-05, "loss": 0.9026, "step": 1765 }, { "epoch": 0.32, "learning_rate": 2.5212362190493406e-05, "loss": 1.7586, "step": 1766 }, { "epoch": 0.32, "learning_rate": 2.5209651183806254e-05, "loss": 0.861, "step": 1767 }, { "epoch": 0.32, "learning_rate": 2.5206940177119103e-05, "loss": 1.1644, "step": 1768 }, { "epoch": 0.32, "learning_rate": 2.5204229170431954e-05, "loss": 1.5825, "step": 1769 }, { "epoch": 0.32, "learning_rate": 2.5201518163744803e-05, "loss": 1.3523, "step": 1770 }, { "epoch": 0.32, "learning_rate": 2.5198807157057655e-05, "loss": 0.9715, "step": 1771 }, { "epoch": 0.32, "learning_rate": 2.5196096150370506e-05, "loss": 1.0266, "step": 1772 }, { "epoch": 0.32, "learning_rate": 2.5193385143683355e-05, "loss": 1.2442, "step": 1773 }, { "epoch": 0.32, "learning_rate": 2.5190674136996207e-05, "loss": 2.3621, "step": 1774 }, { "epoch": 0.32, "learning_rate": 2.5187963130309055e-05, "loss": 0.8689, "step": 1775 }, { "epoch": 0.32, "learning_rate": 2.5185252123621904e-05, "loss": 1.8434, "step": 1776 }, { "epoch": 0.32, "learning_rate": 2.5182541116934755e-05, "loss": 0.9996, "step": 1777 }, { "epoch": 0.32, "learning_rate": 2.5179830110247607e-05, "loss": 1.288, "step": 1778 }, { "epoch": 0.32, "learning_rate": 2.5177119103560456e-05, "loss": 1.5558, "step": 1779 }, { "epoch": 0.32, "learning_rate": 2.5174408096873308e-05, "loss": 1.1983, "step": 1780 }, { "epoch": 0.32, "learning_rate": 2.5171697090186156e-05, "loss": 1.3826, "step": 1781 }, { "epoch": 0.32, "learning_rate": 2.5168986083499008e-05, "loss": 1.9589, "step": 1782 }, { "epoch": 0.32, "learning_rate": 2.5166275076811856e-05, "loss": 1.5072, "step": 1783 }, { "epoch": 0.32, "learning_rate": 2.5163564070124705e-05, "loss": 0.6025, "step": 1784 }, { "epoch": 0.32, "learning_rate": 2.5160853063437557e-05, "loss": 1.1238, "step": 1785 }, { "epoch": 0.32, "learning_rate": 2.515814205675041e-05, "loss": 1.1352, "step": 1786 }, { "epoch": 0.32, "learning_rate": 2.5155431050063257e-05, "loss": 1.4418, "step": 1787 }, { "epoch": 0.32, "learning_rate": 2.515272004337611e-05, "loss": 1.3233, "step": 1788 }, { "epoch": 0.32, "learning_rate": 2.515000903668896e-05, "loss": 1.716, "step": 1789 }, { "epoch": 0.32, "learning_rate": 2.514729803000181e-05, "loss": 0.7926, "step": 1790 }, { "epoch": 0.32, "learning_rate": 2.5144587023314657e-05, "loss": 1.8319, "step": 1791 }, { "epoch": 0.32, "learning_rate": 2.514187601662751e-05, "loss": 1.8259, "step": 1792 }, { "epoch": 0.32, "learning_rate": 2.5139165009940358e-05, "loss": 1.1833, "step": 1793 }, { "epoch": 0.32, "learning_rate": 2.513645400325321e-05, "loss": 1.5688, "step": 1794 }, { "epoch": 0.32, "learning_rate": 2.5133742996566058e-05, "loss": 0.8188, "step": 1795 }, { "epoch": 0.32, "learning_rate": 2.513103198987891e-05, "loss": 0.8866, "step": 1796 }, { "epoch": 0.32, "learning_rate": 2.512832098319176e-05, "loss": 0.9267, "step": 1797 }, { "epoch": 0.32, "learning_rate": 2.512560997650461e-05, "loss": 1.5288, "step": 1798 }, { "epoch": 0.33, "learning_rate": 2.512289896981746e-05, "loss": 1.4252, "step": 1799 }, { "epoch": 0.33, "learning_rate": 2.512018796313031e-05, "loss": 0.9417, "step": 1800 }, { "epoch": 0.33, "learning_rate": 2.511747695644316e-05, "loss": 1.5164, "step": 1801 }, { "epoch": 0.33, "learning_rate": 2.511476594975601e-05, "loss": 1.2536, "step": 1802 }, { "epoch": 0.33, "learning_rate": 2.5112054943068862e-05, "loss": 1.2395, "step": 1803 }, { "epoch": 0.33, "learning_rate": 2.510934393638171e-05, "loss": 1.0061, "step": 1804 }, { "epoch": 0.33, "learning_rate": 2.5106632929694563e-05, "loss": 0.8089, "step": 1805 }, { "epoch": 0.33, "learning_rate": 2.5103921923007408e-05, "loss": 1.0337, "step": 1806 }, { "epoch": 0.33, "learning_rate": 2.510121091632026e-05, "loss": 1.1836, "step": 1807 }, { "epoch": 0.33, "learning_rate": 2.509849990963311e-05, "loss": 1.5772, "step": 1808 }, { "epoch": 0.33, "learning_rate": 2.509578890294596e-05, "loss": 1.6714, "step": 1809 }, { "epoch": 0.33, "learning_rate": 2.509307789625881e-05, "loss": 0.7858, "step": 1810 }, { "epoch": 0.33, "learning_rate": 2.5090366889571663e-05, "loss": 1.3068, "step": 1811 }, { "epoch": 0.33, "learning_rate": 2.5087655882884512e-05, "loss": 1.9102, "step": 1812 }, { "epoch": 0.33, "learning_rate": 2.5084944876197364e-05, "loss": 1.5618, "step": 1813 }, { "epoch": 0.33, "learning_rate": 2.5082233869510212e-05, "loss": 1.3727, "step": 1814 }, { "epoch": 0.33, "learning_rate": 2.507952286282306e-05, "loss": 1.3484, "step": 1815 }, { "epoch": 0.33, "learning_rate": 2.5076811856135912e-05, "loss": 1.3635, "step": 1816 }, { "epoch": 0.33, "learning_rate": 2.507410084944876e-05, "loss": 1.1727, "step": 1817 }, { "epoch": 0.33, "learning_rate": 2.5071389842761613e-05, "loss": 0.8225, "step": 1818 }, { "epoch": 0.33, "learning_rate": 2.5068678836074464e-05, "loss": 1.4039, "step": 1819 }, { "epoch": 0.33, "learning_rate": 2.5065967829387313e-05, "loss": 0.9442, "step": 1820 }, { "epoch": 0.33, "learning_rate": 2.5063256822700165e-05, "loss": 1.4311, "step": 1821 }, { "epoch": 0.33, "learning_rate": 2.5060545816013013e-05, "loss": 1.4734, "step": 1822 }, { "epoch": 0.33, "learning_rate": 2.505783480932586e-05, "loss": 2.085, "step": 1823 }, { "epoch": 0.33, "learning_rate": 2.5055123802638713e-05, "loss": 1.4116, "step": 1824 }, { "epoch": 0.33, "learning_rate": 2.5052412795951565e-05, "loss": 1.4294, "step": 1825 }, { "epoch": 0.33, "learning_rate": 2.5049701789264414e-05, "loss": 1.35, "step": 1826 }, { "epoch": 0.33, "learning_rate": 2.5046990782577266e-05, "loss": 0.953, "step": 1827 }, { "epoch": 0.33, "learning_rate": 2.5044279775890114e-05, "loss": 0.7535, "step": 1828 }, { "epoch": 0.33, "learning_rate": 2.5041568769202966e-05, "loss": 0.7802, "step": 1829 }, { "epoch": 0.33, "learning_rate": 2.5038857762515814e-05, "loss": 1.0053, "step": 1830 }, { "epoch": 0.33, "learning_rate": 2.5036146755828663e-05, "loss": 1.2801, "step": 1831 }, { "epoch": 0.33, "learning_rate": 2.5033435749141515e-05, "loss": 0.8092, "step": 1832 }, { "epoch": 0.33, "learning_rate": 2.5030724742454366e-05, "loss": 1.3808, "step": 1833 }, { "epoch": 0.33, "learning_rate": 2.5028013735767215e-05, "loss": 1.0823, "step": 1834 }, { "epoch": 0.33, "learning_rate": 2.5025302729080067e-05, "loss": 1.6902, "step": 1835 }, { "epoch": 0.33, "learning_rate": 2.502259172239292e-05, "loss": 1.8357, "step": 1836 }, { "epoch": 0.33, "learning_rate": 2.5019880715705767e-05, "loss": 1.1301, "step": 1837 }, { "epoch": 0.33, "learning_rate": 2.5017169709018615e-05, "loss": 1.5305, "step": 1838 }, { "epoch": 0.33, "learning_rate": 2.5014458702331464e-05, "loss": 0.8192, "step": 1839 }, { "epoch": 0.33, "learning_rate": 2.5011747695644316e-05, "loss": 1.0925, "step": 1840 }, { "epoch": 0.33, "learning_rate": 2.5009036688957167e-05, "loss": 1.7527, "step": 1841 }, { "epoch": 0.33, "learning_rate": 2.5006325682270016e-05, "loss": 0.8384, "step": 1842 }, { "epoch": 0.33, "learning_rate": 2.5003614675582868e-05, "loss": 1.0831, "step": 1843 }, { "epoch": 0.33, "learning_rate": 2.500090366889572e-05, "loss": 1.1535, "step": 1844 }, { "epoch": 0.33, "learning_rate": 2.4998192662208568e-05, "loss": 1.7578, "step": 1845 }, { "epoch": 0.33, "learning_rate": 2.4995481655521416e-05, "loss": 0.8481, "step": 1846 }, { "epoch": 0.33, "learning_rate": 2.4992770648834268e-05, "loss": 0.944, "step": 1847 }, { "epoch": 0.33, "learning_rate": 2.4990059642147117e-05, "loss": 1.9263, "step": 1848 }, { "epoch": 0.33, "learning_rate": 2.498734863545997e-05, "loss": 1.0041, "step": 1849 }, { "epoch": 0.33, "learning_rate": 2.4984637628772817e-05, "loss": 0.831, "step": 1850 }, { "epoch": 0.33, "learning_rate": 2.498192662208567e-05, "loss": 1.4712, "step": 1851 }, { "epoch": 0.33, "learning_rate": 2.497921561539852e-05, "loss": 1.3355, "step": 1852 }, { "epoch": 0.33, "learning_rate": 2.497650460871137e-05, "loss": 1.2194, "step": 1853 }, { "epoch": 0.34, "learning_rate": 2.4973793602024217e-05, "loss": 1.2598, "step": 1854 }, { "epoch": 0.34, "learning_rate": 2.497108259533707e-05, "loss": 0.8445, "step": 1855 }, { "epoch": 0.34, "learning_rate": 2.4968371588649918e-05, "loss": 1.2749, "step": 1856 }, { "epoch": 0.34, "learning_rate": 2.496566058196277e-05, "loss": 1.2631, "step": 1857 }, { "epoch": 0.34, "learning_rate": 2.496294957527562e-05, "loss": 2.0738, "step": 1858 }, { "epoch": 0.34, "learning_rate": 2.496023856858847e-05, "loss": 0.7497, "step": 1859 }, { "epoch": 0.34, "learning_rate": 2.495752756190132e-05, "loss": 1.3969, "step": 1860 }, { "epoch": 0.34, "learning_rate": 2.495481655521417e-05, "loss": 1.1078, "step": 1861 }, { "epoch": 0.34, "learning_rate": 2.495210554852702e-05, "loss": 1.459, "step": 1862 }, { "epoch": 0.34, "learning_rate": 2.494939454183987e-05, "loss": 0.9606, "step": 1863 }, { "epoch": 0.34, "learning_rate": 2.494668353515272e-05, "loss": 0.9038, "step": 1864 }, { "epoch": 0.34, "learning_rate": 2.494397252846557e-05, "loss": 1.1814, "step": 1865 }, { "epoch": 0.34, "learning_rate": 2.4941261521778422e-05, "loss": 1.0077, "step": 1866 }, { "epoch": 0.34, "learning_rate": 2.493855051509127e-05, "loss": 0.9019, "step": 1867 }, { "epoch": 0.34, "learning_rate": 2.4935839508404123e-05, "loss": 1.2856, "step": 1868 }, { "epoch": 0.34, "learning_rate": 2.4933128501716975e-05, "loss": 1.8335, "step": 1869 }, { "epoch": 0.34, "learning_rate": 2.493041749502982e-05, "loss": 1.2818, "step": 1870 }, { "epoch": 0.34, "learning_rate": 2.492770648834267e-05, "loss": 1.4508, "step": 1871 }, { "epoch": 0.34, "learning_rate": 2.492499548165552e-05, "loss": 0.9881, "step": 1872 }, { "epoch": 0.34, "learning_rate": 2.4922284474968372e-05, "loss": 1.6283, "step": 1873 }, { "epoch": 0.34, "learning_rate": 2.4919573468281224e-05, "loss": 1.4064, "step": 1874 }, { "epoch": 0.34, "learning_rate": 2.4916862461594072e-05, "loss": 1.5154, "step": 1875 }, { "epoch": 0.34, "learning_rate": 2.4914151454906924e-05, "loss": 1.9165, "step": 1876 }, { "epoch": 0.34, "learning_rate": 2.4911440448219776e-05, "loss": 0.9501, "step": 1877 }, { "epoch": 0.34, "learning_rate": 2.490872944153262e-05, "loss": 1.3334, "step": 1878 }, { "epoch": 0.34, "learning_rate": 2.4906018434845473e-05, "loss": 1.188, "step": 1879 }, { "epoch": 0.34, "learning_rate": 2.4903307428158324e-05, "loss": 1.4408, "step": 1880 }, { "epoch": 0.34, "learning_rate": 2.4900596421471173e-05, "loss": 1.4223, "step": 1881 }, { "epoch": 0.34, "learning_rate": 2.4897885414784025e-05, "loss": 1.0629, "step": 1882 }, { "epoch": 0.34, "learning_rate": 2.4895174408096873e-05, "loss": 1.4311, "step": 1883 }, { "epoch": 0.34, "learning_rate": 2.4892463401409725e-05, "loss": 0.9365, "step": 1884 }, { "epoch": 0.34, "learning_rate": 2.4889752394722577e-05, "loss": 1.1324, "step": 1885 }, { "epoch": 0.34, "learning_rate": 2.4887041388035422e-05, "loss": 0.9397, "step": 1886 }, { "epoch": 0.34, "learning_rate": 2.4884330381348274e-05, "loss": 1.3931, "step": 1887 }, { "epoch": 0.34, "learning_rate": 2.4881619374661125e-05, "loss": 1.2352, "step": 1888 }, { "epoch": 0.34, "learning_rate": 2.4878908367973974e-05, "loss": 1.5986, "step": 1889 }, { "epoch": 0.34, "learning_rate": 2.4876197361286826e-05, "loss": 0.7665, "step": 1890 }, { "epoch": 0.34, "learning_rate": 2.4873486354599678e-05, "loss": 1.2098, "step": 1891 }, { "epoch": 0.34, "learning_rate": 2.4870775347912526e-05, "loss": 0.7148, "step": 1892 }, { "epoch": 0.34, "learning_rate": 2.4868064341225378e-05, "loss": 1.4751, "step": 1893 }, { "epoch": 0.34, "learning_rate": 2.4865353334538223e-05, "loss": 0.6546, "step": 1894 }, { "epoch": 0.34, "learning_rate": 2.4862642327851075e-05, "loss": 1.4889, "step": 1895 }, { "epoch": 0.34, "learning_rate": 2.4859931321163926e-05, "loss": 1.3491, "step": 1896 }, { "epoch": 0.34, "learning_rate": 2.4857220314476775e-05, "loss": 1.8591, "step": 1897 }, { "epoch": 0.34, "learning_rate": 2.4854509307789627e-05, "loss": 0.7064, "step": 1898 }, { "epoch": 0.34, "learning_rate": 2.485179830110248e-05, "loss": 1.6264, "step": 1899 }, { "epoch": 0.34, "learning_rate": 2.4849087294415327e-05, "loss": 0.8073, "step": 1900 }, { "epoch": 0.34, "learning_rate": 2.484637628772818e-05, "loss": 1.4424, "step": 1901 }, { "epoch": 0.34, "learning_rate": 2.4843665281041027e-05, "loss": 1.2692, "step": 1902 }, { "epoch": 0.34, "learning_rate": 2.4840954274353876e-05, "loss": 0.9922, "step": 1903 }, { "epoch": 0.34, "learning_rate": 2.4838243267666728e-05, "loss": 1.2554, "step": 1904 }, { "epoch": 0.34, "learning_rate": 2.4835532260979576e-05, "loss": 1.3235, "step": 1905 }, { "epoch": 0.34, "learning_rate": 2.4832821254292428e-05, "loss": 1.294, "step": 1906 }, { "epoch": 0.34, "learning_rate": 2.483011024760528e-05, "loss": 1.0946, "step": 1907 }, { "epoch": 0.34, "learning_rate": 2.4827399240918128e-05, "loss": 0.7807, "step": 1908 }, { "epoch": 0.35, "learning_rate": 2.482468823423098e-05, "loss": 1.0342, "step": 1909 }, { "epoch": 0.35, "learning_rate": 2.482197722754383e-05, "loss": 1.1477, "step": 1910 }, { "epoch": 0.35, "learning_rate": 2.4819266220856677e-05, "loss": 1.551, "step": 1911 }, { "epoch": 0.35, "learning_rate": 2.481655521416953e-05, "loss": 1.0227, "step": 1912 }, { "epoch": 0.35, "learning_rate": 2.481384420748238e-05, "loss": 1.0447, "step": 1913 }, { "epoch": 0.35, "learning_rate": 2.481113320079523e-05, "loss": 1.6264, "step": 1914 }, { "epoch": 0.35, "learning_rate": 2.480842219410808e-05, "loss": 1.4302, "step": 1915 }, { "epoch": 0.35, "learning_rate": 2.480571118742093e-05, "loss": 0.8095, "step": 1916 }, { "epoch": 0.35, "learning_rate": 2.480300018073378e-05, "loss": 1.5579, "step": 1917 }, { "epoch": 0.35, "learning_rate": 2.480028917404663e-05, "loss": 1.6211, "step": 1918 }, { "epoch": 0.35, "learning_rate": 2.4797578167359478e-05, "loss": 1.3138, "step": 1919 }, { "epoch": 0.35, "learning_rate": 2.479486716067233e-05, "loss": 1.6187, "step": 1920 }, { "epoch": 0.35, "learning_rate": 2.479215615398518e-05, "loss": 1.4486, "step": 1921 }, { "epoch": 0.35, "learning_rate": 2.478944514729803e-05, "loss": 0.8844, "step": 1922 }, { "epoch": 0.35, "learning_rate": 2.4786734140610882e-05, "loss": 1.2516, "step": 1923 }, { "epoch": 0.35, "learning_rate": 2.4784023133923734e-05, "loss": 1.3925, "step": 1924 }, { "epoch": 0.35, "learning_rate": 2.4781312127236582e-05, "loss": 1.6536, "step": 1925 }, { "epoch": 0.35, "learning_rate": 2.477860112054943e-05, "loss": 0.9204, "step": 1926 }, { "epoch": 0.35, "learning_rate": 2.477589011386228e-05, "loss": 1.4509, "step": 1927 }, { "epoch": 0.35, "learning_rate": 2.477317910717513e-05, "loss": 1.6333, "step": 1928 }, { "epoch": 0.35, "learning_rate": 2.4770468100487983e-05, "loss": 1.5115, "step": 1929 }, { "epoch": 0.35, "learning_rate": 2.476775709380083e-05, "loss": 1.1619, "step": 1930 }, { "epoch": 0.35, "learning_rate": 2.4765046087113683e-05, "loss": 1.5214, "step": 1931 }, { "epoch": 0.35, "learning_rate": 2.4762335080426535e-05, "loss": 1.0447, "step": 1932 }, { "epoch": 0.35, "learning_rate": 2.475962407373938e-05, "loss": 1.3267, "step": 1933 }, { "epoch": 0.35, "learning_rate": 2.475691306705223e-05, "loss": 1.6898, "step": 1934 }, { "epoch": 0.35, "learning_rate": 2.4754202060365083e-05, "loss": 1.087, "step": 1935 }, { "epoch": 0.35, "learning_rate": 2.4751491053677932e-05, "loss": 0.5982, "step": 1936 }, { "epoch": 0.35, "learning_rate": 2.4748780046990784e-05, "loss": 0.9983, "step": 1937 }, { "epoch": 0.35, "learning_rate": 2.4746069040303632e-05, "loss": 0.9715, "step": 1938 }, { "epoch": 0.35, "learning_rate": 2.4743358033616484e-05, "loss": 2.1069, "step": 1939 }, { "epoch": 0.35, "learning_rate": 2.4740647026929336e-05, "loss": 0.8109, "step": 1940 }, { "epoch": 0.35, "learning_rate": 2.473793602024218e-05, "loss": 0.8613, "step": 1941 }, { "epoch": 0.35, "learning_rate": 2.4735225013555033e-05, "loss": 1.0206, "step": 1942 }, { "epoch": 0.35, "learning_rate": 2.4732514006867885e-05, "loss": 1.0283, "step": 1943 }, { "epoch": 0.35, "learning_rate": 2.4729803000180733e-05, "loss": 1.4938, "step": 1944 }, { "epoch": 0.35, "learning_rate": 2.4727091993493585e-05, "loss": 1.3322, "step": 1945 }, { "epoch": 0.35, "learning_rate": 2.4724380986806437e-05, "loss": 1.1452, "step": 1946 }, { "epoch": 0.35, "learning_rate": 2.4721669980119285e-05, "loss": 0.8256, "step": 1947 }, { "epoch": 0.35, "learning_rate": 2.4718958973432137e-05, "loss": 0.4273, "step": 1948 }, { "epoch": 0.35, "learning_rate": 2.4716247966744985e-05, "loss": 1.3966, "step": 1949 }, { "epoch": 0.35, "learning_rate": 2.4713536960057834e-05, "loss": 0.9004, "step": 1950 }, { "epoch": 0.35, "learning_rate": 2.4710825953370686e-05, "loss": 1.1062, "step": 1951 }, { "epoch": 0.35, "learning_rate": 2.4708114946683534e-05, "loss": 0.8992, "step": 1952 }, { "epoch": 0.35, "learning_rate": 2.4705403939996386e-05, "loss": 2.0131, "step": 1953 }, { "epoch": 0.35, "learning_rate": 2.4702692933309238e-05, "loss": 1.9976, "step": 1954 }, { "epoch": 0.35, "learning_rate": 2.4699981926622086e-05, "loss": 1.1143, "step": 1955 }, { "epoch": 0.35, "learning_rate": 2.4697270919934938e-05, "loss": 1.171, "step": 1956 }, { "epoch": 0.35, "learning_rate": 2.4694559913247786e-05, "loss": 0.5717, "step": 1957 }, { "epoch": 0.35, "learning_rate": 2.4691848906560635e-05, "loss": 1.2602, "step": 1958 }, { "epoch": 0.35, "learning_rate": 2.4689137899873487e-05, "loss": 1.1918, "step": 1959 }, { "epoch": 0.35, "learning_rate": 2.468642689318634e-05, "loss": 1.4114, "step": 1960 }, { "epoch": 0.35, "learning_rate": 2.4683715886499187e-05, "loss": 1.4456, "step": 1961 }, { "epoch": 0.35, "learning_rate": 2.468100487981204e-05, "loss": 1.2482, "step": 1962 }, { "epoch": 0.35, "learning_rate": 2.4678293873124887e-05, "loss": 1.1974, "step": 1963 }, { "epoch": 0.35, "learning_rate": 2.467558286643774e-05, "loss": 0.7729, "step": 1964 }, { "epoch": 0.36, "learning_rate": 2.4672871859750587e-05, "loss": 1.1815, "step": 1965 }, { "epoch": 0.36, "learning_rate": 2.4670160853063436e-05, "loss": 0.621, "step": 1966 }, { "epoch": 0.36, "learning_rate": 2.4667449846376288e-05, "loss": 1.7501, "step": 1967 }, { "epoch": 0.36, "learning_rate": 2.466473883968914e-05, "loss": 1.8788, "step": 1968 }, { "epoch": 0.36, "learning_rate": 2.4662027833001988e-05, "loss": 1.2313, "step": 1969 }, { "epoch": 0.36, "learning_rate": 2.465931682631484e-05, "loss": 0.7143, "step": 1970 }, { "epoch": 0.36, "learning_rate": 2.465660581962769e-05, "loss": 1.354, "step": 1971 }, { "epoch": 0.36, "learning_rate": 2.465389481294054e-05, "loss": 0.9082, "step": 1972 }, { "epoch": 0.36, "learning_rate": 2.465118380625339e-05, "loss": 0.7762, "step": 1973 }, { "epoch": 0.36, "learning_rate": 2.4648472799566237e-05, "loss": 1.7169, "step": 1974 }, { "epoch": 0.36, "learning_rate": 2.464576179287909e-05, "loss": 1.0956, "step": 1975 }, { "epoch": 0.36, "learning_rate": 2.464305078619194e-05, "loss": 1.2026, "step": 1976 }, { "epoch": 0.36, "learning_rate": 2.464033977950479e-05, "loss": 0.9381, "step": 1977 }, { "epoch": 0.36, "learning_rate": 2.463762877281764e-05, "loss": 0.8769, "step": 1978 }, { "epoch": 0.36, "learning_rate": 2.4634917766130493e-05, "loss": 1.2861, "step": 1979 }, { "epoch": 0.36, "learning_rate": 2.463220675944334e-05, "loss": 0.968, "step": 1980 }, { "epoch": 0.36, "learning_rate": 2.462949575275619e-05, "loss": 1.3165, "step": 1981 }, { "epoch": 0.36, "learning_rate": 2.462678474606904e-05, "loss": 1.0459, "step": 1982 }, { "epoch": 0.36, "learning_rate": 2.462407373938189e-05, "loss": 0.776, "step": 1983 }, { "epoch": 0.36, "learning_rate": 2.4621362732694742e-05, "loss": 0.8852, "step": 1984 }, { "epoch": 0.36, "learning_rate": 2.461865172600759e-05, "loss": 1.4213, "step": 1985 }, { "epoch": 0.36, "learning_rate": 2.4615940719320442e-05, "loss": 0.5337, "step": 1986 }, { "epoch": 0.36, "learning_rate": 2.4613229712633294e-05, "loss": 1.825, "step": 1987 }, { "epoch": 0.36, "learning_rate": 2.4610518705946142e-05, "loss": 0.8373, "step": 1988 }, { "epoch": 0.36, "learning_rate": 2.460780769925899e-05, "loss": 1.1053, "step": 1989 }, { "epoch": 0.36, "learning_rate": 2.4605096692571843e-05, "loss": 0.6605, "step": 1990 }, { "epoch": 0.36, "learning_rate": 2.460238568588469e-05, "loss": 1.8019, "step": 1991 }, { "epoch": 0.36, "learning_rate": 2.4599674679197543e-05, "loss": 1.4201, "step": 1992 }, { "epoch": 0.36, "learning_rate": 2.4596963672510395e-05, "loss": 1.0032, "step": 1993 }, { "epoch": 0.36, "learning_rate": 2.4594252665823243e-05, "loss": 1.6718, "step": 1994 }, { "epoch": 0.36, "learning_rate": 2.4591541659136095e-05, "loss": 2.1751, "step": 1995 }, { "epoch": 0.36, "learning_rate": 2.4588830652448943e-05, "loss": 1.2718, "step": 1996 }, { "epoch": 0.36, "learning_rate": 2.4586119645761792e-05, "loss": 0.9445, "step": 1997 }, { "epoch": 0.36, "learning_rate": 2.4583408639074644e-05, "loss": 0.8237, "step": 1998 }, { "epoch": 0.36, "learning_rate": 2.4580697632387492e-05, "loss": 1.0303, "step": 1999 }, { "epoch": 0.36, "learning_rate": 2.4577986625700344e-05, "loss": 1.4472, "step": 2000 }, { "epoch": 0.36, "learning_rate": 2.4575275619013196e-05, "loss": 1.5977, "step": 2001 }, { "epoch": 0.36, "learning_rate": 2.4572564612326044e-05, "loss": 2.1449, "step": 2002 }, { "epoch": 0.36, "learning_rate": 2.4569853605638896e-05, "loss": 1.0558, "step": 2003 }, { "epoch": 0.36, "learning_rate": 2.4567142598951748e-05, "loss": 1.9011, "step": 2004 }, { "epoch": 0.36, "learning_rate": 2.4564431592264593e-05, "loss": 0.8356, "step": 2005 }, { "epoch": 0.36, "learning_rate": 2.4561720585577445e-05, "loss": 1.0818, "step": 2006 }, { "epoch": 0.36, "learning_rate": 2.4559009578890293e-05, "loss": 0.78, "step": 2007 }, { "epoch": 0.36, "learning_rate": 2.4556298572203145e-05, "loss": 1.157, "step": 2008 }, { "epoch": 0.36, "learning_rate": 2.4553587565515997e-05, "loss": 0.8392, "step": 2009 }, { "epoch": 0.36, "learning_rate": 2.4550876558828845e-05, "loss": 1.2216, "step": 2010 }, { "epoch": 0.36, "learning_rate": 2.4548165552141697e-05, "loss": 1.4739, "step": 2011 }, { "epoch": 0.36, "learning_rate": 2.454545454545455e-05, "loss": 1.0475, "step": 2012 }, { "epoch": 0.36, "learning_rate": 2.4542743538767394e-05, "loss": 1.1236, "step": 2013 }, { "epoch": 0.36, "learning_rate": 2.4540032532080246e-05, "loss": 1.2995, "step": 2014 }, { "epoch": 0.36, "learning_rate": 2.4537321525393098e-05, "loss": 1.1659, "step": 2015 }, { "epoch": 0.36, "learning_rate": 2.4534610518705946e-05, "loss": 1.1664, "step": 2016 }, { "epoch": 0.36, "learning_rate": 2.4531899512018798e-05, "loss": 1.1078, "step": 2017 }, { "epoch": 0.36, "learning_rate": 2.4529188505331646e-05, "loss": 0.9977, "step": 2018 }, { "epoch": 0.36, "learning_rate": 2.4526477498644498e-05, "loss": 1.8178, "step": 2019 }, { "epoch": 0.37, "learning_rate": 2.452376649195735e-05, "loss": 0.8057, "step": 2020 }, { "epoch": 0.37, "learning_rate": 2.4521055485270195e-05, "loss": 1.3083, "step": 2021 }, { "epoch": 0.37, "learning_rate": 2.4518344478583047e-05, "loss": 1.5139, "step": 2022 }, { "epoch": 0.37, "learning_rate": 2.45156334718959e-05, "loss": 1.5432, "step": 2023 }, { "epoch": 0.37, "learning_rate": 2.4512922465208747e-05, "loss": 1.1899, "step": 2024 }, { "epoch": 0.37, "learning_rate": 2.45102114585216e-05, "loss": 1.3049, "step": 2025 }, { "epoch": 0.37, "learning_rate": 2.450750045183445e-05, "loss": 1.4896, "step": 2026 }, { "epoch": 0.37, "learning_rate": 2.45047894451473e-05, "loss": 0.9359, "step": 2027 }, { "epoch": 0.37, "learning_rate": 2.450207843846015e-05, "loss": 0.8117, "step": 2028 }, { "epoch": 0.37, "learning_rate": 2.4499367431772996e-05, "loss": 0.6928, "step": 2029 }, { "epoch": 0.37, "learning_rate": 2.4496656425085848e-05, "loss": 1.2503, "step": 2030 }, { "epoch": 0.37, "learning_rate": 2.44939454183987e-05, "loss": 0.8414, "step": 2031 }, { "epoch": 0.37, "learning_rate": 2.4491234411711548e-05, "loss": 1.2217, "step": 2032 }, { "epoch": 0.37, "learning_rate": 2.44885234050244e-05, "loss": 1.1883, "step": 2033 }, { "epoch": 0.37, "learning_rate": 2.4485812398337252e-05, "loss": 1.3894, "step": 2034 }, { "epoch": 0.37, "learning_rate": 2.44831013916501e-05, "loss": 1.3068, "step": 2035 }, { "epoch": 0.37, "learning_rate": 2.4480390384962952e-05, "loss": 1.2053, "step": 2036 }, { "epoch": 0.37, "learning_rate": 2.44776793782758e-05, "loss": 0.9728, "step": 2037 }, { "epoch": 0.37, "learning_rate": 2.447496837158865e-05, "loss": 1.1762, "step": 2038 }, { "epoch": 0.37, "learning_rate": 2.44722573649015e-05, "loss": 1.3626, "step": 2039 }, { "epoch": 0.37, "learning_rate": 2.446954635821435e-05, "loss": 1.4571, "step": 2040 }, { "epoch": 0.37, "learning_rate": 2.44668353515272e-05, "loss": 1.3886, "step": 2041 }, { "epoch": 0.37, "learning_rate": 2.4464124344840053e-05, "loss": 0.7049, "step": 2042 }, { "epoch": 0.37, "learning_rate": 2.44614133381529e-05, "loss": 1.2077, "step": 2043 }, { "epoch": 0.37, "learning_rate": 2.4458702331465753e-05, "loss": 1.4998, "step": 2044 }, { "epoch": 0.37, "learning_rate": 2.44559913247786e-05, "loss": 1.0361, "step": 2045 }, { "epoch": 0.37, "learning_rate": 2.445328031809145e-05, "loss": 1.7911, "step": 2046 }, { "epoch": 0.37, "learning_rate": 2.4450569311404302e-05, "loss": 1.1731, "step": 2047 }, { "epoch": 0.37, "learning_rate": 2.4447858304717154e-05, "loss": 1.0849, "step": 2048 }, { "epoch": 0.37, "learning_rate": 2.4445147298030002e-05, "loss": 0.9568, "step": 2049 }, { "epoch": 0.37, "learning_rate": 2.4442436291342854e-05, "loss": 0.8268, "step": 2050 }, { "epoch": 0.37, "learning_rate": 2.4439725284655702e-05, "loss": 1.7054, "step": 2051 }, { "epoch": 0.37, "learning_rate": 2.4437014277968554e-05, "loss": 1.0002, "step": 2052 }, { "epoch": 0.37, "learning_rate": 2.4434303271281403e-05, "loss": 1.7002, "step": 2053 }, { "epoch": 0.37, "learning_rate": 2.443159226459425e-05, "loss": 1.0973, "step": 2054 }, { "epoch": 0.37, "learning_rate": 2.4428881257907103e-05, "loss": 0.9582, "step": 2055 }, { "epoch": 0.37, "learning_rate": 2.4426170251219955e-05, "loss": 0.7233, "step": 2056 }, { "epoch": 0.37, "learning_rate": 2.4423459244532803e-05, "loss": 1.6023, "step": 2057 }, { "epoch": 0.37, "learning_rate": 2.4420748237845655e-05, "loss": 0.9085, "step": 2058 }, { "epoch": 0.37, "learning_rate": 2.4418037231158507e-05, "loss": 1.3297, "step": 2059 }, { "epoch": 0.37, "learning_rate": 2.4415326224471352e-05, "loss": 0.9637, "step": 2060 }, { "epoch": 0.37, "learning_rate": 2.4412615217784204e-05, "loss": 1.2127, "step": 2061 }, { "epoch": 0.37, "learning_rate": 2.4409904211097052e-05, "loss": 1.1051, "step": 2062 }, { "epoch": 0.37, "learning_rate": 2.4407193204409904e-05, "loss": 1.7339, "step": 2063 }, { "epoch": 0.37, "learning_rate": 2.4404482197722756e-05, "loss": 0.8813, "step": 2064 }, { "epoch": 0.37, "learning_rate": 2.4401771191035604e-05, "loss": 1.0552, "step": 2065 }, { "epoch": 0.37, "learning_rate": 2.4399060184348456e-05, "loss": 0.9388, "step": 2066 }, { "epoch": 0.37, "learning_rate": 2.4396349177661308e-05, "loss": 0.6764, "step": 2067 }, { "epoch": 0.37, "learning_rate": 2.4393638170974153e-05, "loss": 1.6658, "step": 2068 }, { "epoch": 0.37, "learning_rate": 2.4390927164287005e-05, "loss": 0.9085, "step": 2069 }, { "epoch": 0.37, "learning_rate": 2.4388216157599857e-05, "loss": 1.3377, "step": 2070 }, { "epoch": 0.37, "learning_rate": 2.4385505150912705e-05, "loss": 2.0475, "step": 2071 }, { "epoch": 0.37, "learning_rate": 2.4382794144225557e-05, "loss": 1.2374, "step": 2072 }, { "epoch": 0.37, "learning_rate": 2.4380083137538405e-05, "loss": 1.5758, "step": 2073 }, { "epoch": 0.37, "learning_rate": 2.4377372130851257e-05, "loss": 1.375, "step": 2074 }, { "epoch": 0.38, "learning_rate": 2.437466112416411e-05, "loss": 1.3435, "step": 2075 }, { "epoch": 0.38, "learning_rate": 2.4371950117476954e-05, "loss": 1.4027, "step": 2076 }, { "epoch": 0.38, "learning_rate": 2.4369239110789806e-05, "loss": 0.7623, "step": 2077 }, { "epoch": 0.38, "learning_rate": 2.4366528104102658e-05, "loss": 1.023, "step": 2078 }, { "epoch": 0.38, "learning_rate": 2.4363817097415506e-05, "loss": 1.1767, "step": 2079 }, { "epoch": 0.38, "learning_rate": 2.4361106090728358e-05, "loss": 1.054, "step": 2080 }, { "epoch": 0.38, "learning_rate": 2.435839508404121e-05, "loss": 0.834, "step": 2081 }, { "epoch": 0.38, "learning_rate": 2.4355684077354058e-05, "loss": 0.8116, "step": 2082 }, { "epoch": 0.38, "learning_rate": 2.435297307066691e-05, "loss": 1.3115, "step": 2083 }, { "epoch": 0.38, "learning_rate": 2.435026206397976e-05, "loss": 1.5312, "step": 2084 }, { "epoch": 0.38, "learning_rate": 2.4347551057292607e-05, "loss": 0.8192, "step": 2085 }, { "epoch": 0.38, "learning_rate": 2.434484005060546e-05, "loss": 1.4844, "step": 2086 }, { "epoch": 0.38, "learning_rate": 2.4342129043918307e-05, "loss": 1.4448, "step": 2087 }, { "epoch": 0.38, "learning_rate": 2.433941803723116e-05, "loss": 0.7022, "step": 2088 }, { "epoch": 0.38, "learning_rate": 2.433670703054401e-05, "loss": 1.6559, "step": 2089 }, { "epoch": 0.38, "learning_rate": 2.433399602385686e-05, "loss": 1.5535, "step": 2090 }, { "epoch": 0.38, "learning_rate": 2.433128501716971e-05, "loss": 0.9177, "step": 2091 }, { "epoch": 0.38, "learning_rate": 2.432857401048256e-05, "loss": 1.4468, "step": 2092 }, { "epoch": 0.38, "learning_rate": 2.4325863003795408e-05, "loss": 0.775, "step": 2093 }, { "epoch": 0.38, "learning_rate": 2.432315199710826e-05, "loss": 1.4082, "step": 2094 }, { "epoch": 0.38, "learning_rate": 2.432044099042111e-05, "loss": 1.1325, "step": 2095 }, { "epoch": 0.38, "learning_rate": 2.431772998373396e-05, "loss": 1.365, "step": 2096 }, { "epoch": 0.38, "learning_rate": 2.4315018977046812e-05, "loss": 0.9558, "step": 2097 }, { "epoch": 0.38, "learning_rate": 2.431230797035966e-05, "loss": 1.0901, "step": 2098 }, { "epoch": 0.38, "learning_rate": 2.4309596963672512e-05, "loss": 1.1875, "step": 2099 }, { "epoch": 0.38, "learning_rate": 2.430688595698536e-05, "loss": 1.782, "step": 2100 }, { "epoch": 0.38, "learning_rate": 2.430417495029821e-05, "loss": 1.553, "step": 2101 }, { "epoch": 0.38, "learning_rate": 2.430146394361106e-05, "loss": 1.5362, "step": 2102 }, { "epoch": 0.38, "learning_rate": 2.4298752936923913e-05, "loss": 1.2558, "step": 2103 }, { "epoch": 0.38, "learning_rate": 2.429604193023676e-05, "loss": 0.7557, "step": 2104 }, { "epoch": 0.38, "learning_rate": 2.4293330923549613e-05, "loss": 1.2201, "step": 2105 }, { "epoch": 0.38, "learning_rate": 2.4290619916862465e-05, "loss": 1.0287, "step": 2106 }, { "epoch": 0.38, "learning_rate": 2.4287908910175313e-05, "loss": 1.6101, "step": 2107 }, { "epoch": 0.38, "learning_rate": 2.4285197903488162e-05, "loss": 1.1532, "step": 2108 }, { "epoch": 0.38, "learning_rate": 2.428248689680101e-05, "loss": 1.2248, "step": 2109 }, { "epoch": 0.38, "learning_rate": 2.4279775890113862e-05, "loss": 0.7936, "step": 2110 }, { "epoch": 0.38, "learning_rate": 2.4277064883426714e-05, "loss": 1.074, "step": 2111 }, { "epoch": 0.38, "learning_rate": 2.4274353876739562e-05, "loss": 1.0821, "step": 2112 }, { "epoch": 0.38, "learning_rate": 2.4271642870052414e-05, "loss": 1.0698, "step": 2113 }, { "epoch": 0.38, "learning_rate": 2.4268931863365266e-05, "loss": 1.1504, "step": 2114 }, { "epoch": 0.38, "learning_rate": 2.4266220856678114e-05, "loss": 0.6582, "step": 2115 }, { "epoch": 0.38, "learning_rate": 2.4263509849990963e-05, "loss": 1.4718, "step": 2116 }, { "epoch": 0.38, "learning_rate": 2.4260798843303815e-05, "loss": 1.2342, "step": 2117 }, { "epoch": 0.38, "learning_rate": 2.4258087836616663e-05, "loss": 1.0043, "step": 2118 }, { "epoch": 0.38, "learning_rate": 2.4255376829929515e-05, "loss": 1.3158, "step": 2119 }, { "epoch": 0.38, "learning_rate": 2.4252665823242363e-05, "loss": 1.0514, "step": 2120 }, { "epoch": 0.38, "learning_rate": 2.4249954816555215e-05, "loss": 1.1588, "step": 2121 }, { "epoch": 0.38, "learning_rate": 2.4247243809868067e-05, "loss": 0.5303, "step": 2122 }, { "epoch": 0.38, "learning_rate": 2.4244532803180915e-05, "loss": 2.1307, "step": 2123 }, { "epoch": 0.38, "learning_rate": 2.4241821796493764e-05, "loss": 1.509, "step": 2124 }, { "epoch": 0.38, "learning_rate": 2.4239110789806616e-05, "loss": 1.0963, "step": 2125 }, { "epoch": 0.38, "learning_rate": 2.4236399783119464e-05, "loss": 1.2594, "step": 2126 }, { "epoch": 0.38, "learning_rate": 2.4233688776432316e-05, "loss": 1.8318, "step": 2127 }, { "epoch": 0.38, "learning_rate": 2.4230977769745168e-05, "loss": 1.2823, "step": 2128 }, { "epoch": 0.38, "learning_rate": 2.4228266763058016e-05, "loss": 1.0145, "step": 2129 }, { "epoch": 0.38, "learning_rate": 2.4225555756370868e-05, "loss": 1.3054, "step": 2130 }, { "epoch": 0.39, "learning_rate": 2.4222844749683717e-05, "loss": 0.8296, "step": 2131 }, { "epoch": 0.39, "learning_rate": 2.4220133742996565e-05, "loss": 1.0242, "step": 2132 }, { "epoch": 0.39, "learning_rate": 2.4217422736309417e-05, "loss": 0.5926, "step": 2133 }, { "epoch": 0.39, "learning_rate": 2.4214711729622265e-05, "loss": 1.2796, "step": 2134 }, { "epoch": 0.39, "learning_rate": 2.4212000722935117e-05, "loss": 1.6548, "step": 2135 }, { "epoch": 0.39, "learning_rate": 2.420928971624797e-05, "loss": 1.6447, "step": 2136 }, { "epoch": 0.39, "learning_rate": 2.4206578709560817e-05, "loss": 0.7154, "step": 2137 }, { "epoch": 0.39, "learning_rate": 2.420386770287367e-05, "loss": 1.4078, "step": 2138 }, { "epoch": 0.39, "learning_rate": 2.420115669618652e-05, "loss": 1.1202, "step": 2139 }, { "epoch": 0.39, "learning_rate": 2.4198445689499366e-05, "loss": 0.5693, "step": 2140 }, { "epoch": 0.39, "learning_rate": 2.4195734682812218e-05, "loss": 1.2453, "step": 2141 }, { "epoch": 0.39, "learning_rate": 2.4193023676125066e-05, "loss": 1.2984, "step": 2142 }, { "epoch": 0.39, "learning_rate": 2.4190312669437918e-05, "loss": 1.5564, "step": 2143 }, { "epoch": 0.39, "learning_rate": 2.418760166275077e-05, "loss": 1.6073, "step": 2144 }, { "epoch": 0.39, "learning_rate": 2.418489065606362e-05, "loss": 1.4399, "step": 2145 }, { "epoch": 0.39, "learning_rate": 2.418217964937647e-05, "loss": 1.7798, "step": 2146 }, { "epoch": 0.39, "learning_rate": 2.4179468642689322e-05, "loss": 1.02, "step": 2147 }, { "epoch": 0.39, "learning_rate": 2.4176757636002167e-05, "loss": 1.1867, "step": 2148 }, { "epoch": 0.39, "learning_rate": 2.417404662931502e-05, "loss": 1.2824, "step": 2149 }, { "epoch": 0.39, "learning_rate": 2.417133562262787e-05, "loss": 1.2157, "step": 2150 }, { "epoch": 0.39, "learning_rate": 2.416862461594072e-05, "loss": 1.2818, "step": 2151 }, { "epoch": 0.39, "learning_rate": 2.416591360925357e-05, "loss": 1.0538, "step": 2152 }, { "epoch": 0.39, "learning_rate": 2.416320260256642e-05, "loss": 1.1994, "step": 2153 }, { "epoch": 0.39, "learning_rate": 2.416049159587927e-05, "loss": 1.2114, "step": 2154 }, { "epoch": 0.39, "learning_rate": 2.4157780589192123e-05, "loss": 0.8309, "step": 2155 }, { "epoch": 0.39, "learning_rate": 2.4155069582504968e-05, "loss": 1.9156, "step": 2156 }, { "epoch": 0.39, "learning_rate": 2.415235857581782e-05, "loss": 1.2494, "step": 2157 }, { "epoch": 0.39, "learning_rate": 2.4149647569130672e-05, "loss": 0.6158, "step": 2158 }, { "epoch": 0.39, "learning_rate": 2.414693656244352e-05, "loss": 1.2505, "step": 2159 }, { "epoch": 0.39, "learning_rate": 2.4144225555756372e-05, "loss": 0.623, "step": 2160 }, { "epoch": 0.39, "learning_rate": 2.4141514549069224e-05, "loss": 1.8813, "step": 2161 }, { "epoch": 0.39, "learning_rate": 2.4138803542382072e-05, "loss": 0.9891, "step": 2162 }, { "epoch": 0.39, "learning_rate": 2.4136092535694924e-05, "loss": 1.0912, "step": 2163 }, { "epoch": 0.39, "learning_rate": 2.413338152900777e-05, "loss": 1.2862, "step": 2164 }, { "epoch": 0.39, "learning_rate": 2.413067052232062e-05, "loss": 1.5618, "step": 2165 }, { "epoch": 0.39, "learning_rate": 2.4127959515633473e-05, "loss": 0.9929, "step": 2166 }, { "epoch": 0.39, "learning_rate": 2.412524850894632e-05, "loss": 1.3252, "step": 2167 }, { "epoch": 0.39, "learning_rate": 2.4122537502259173e-05, "loss": 1.7251, "step": 2168 }, { "epoch": 0.39, "learning_rate": 2.4119826495572025e-05, "loss": 1.6687, "step": 2169 }, { "epoch": 0.39, "learning_rate": 2.4117115488884873e-05, "loss": 1.9936, "step": 2170 }, { "epoch": 0.39, "learning_rate": 2.4114404482197725e-05, "loss": 1.2977, "step": 2171 }, { "epoch": 0.39, "learning_rate": 2.4111693475510574e-05, "loss": 1.3784, "step": 2172 }, { "epoch": 0.39, "learning_rate": 2.4108982468823422e-05, "loss": 0.8416, "step": 2173 }, { "epoch": 0.39, "learning_rate": 2.4106271462136274e-05, "loss": 0.4672, "step": 2174 }, { "epoch": 0.39, "learning_rate": 2.4103560455449122e-05, "loss": 1.4769, "step": 2175 }, { "epoch": 0.39, "learning_rate": 2.4100849448761974e-05, "loss": 0.7983, "step": 2176 }, { "epoch": 0.39, "learning_rate": 2.4098138442074826e-05, "loss": 1.4203, "step": 2177 }, { "epoch": 0.39, "learning_rate": 2.4095427435387675e-05, "loss": 1.2469, "step": 2178 }, { "epoch": 0.39, "learning_rate": 2.4092716428700526e-05, "loss": 1.2625, "step": 2179 }, { "epoch": 0.39, "learning_rate": 2.4090005422013375e-05, "loss": 1.5468, "step": 2180 }, { "epoch": 0.39, "learning_rate": 2.4087294415326223e-05, "loss": 0.8464, "step": 2181 }, { "epoch": 0.39, "learning_rate": 2.4084583408639075e-05, "loss": 1.0478, "step": 2182 }, { "epoch": 0.39, "learning_rate": 2.4081872401951927e-05, "loss": 1.3646, "step": 2183 }, { "epoch": 0.39, "learning_rate": 2.4079161395264775e-05, "loss": 0.9235, "step": 2184 }, { "epoch": 0.39, "learning_rate": 2.4076450388577627e-05, "loss": 1.1486, "step": 2185 }, { "epoch": 0.4, "learning_rate": 2.4073739381890476e-05, "loss": 0.5875, "step": 2186 }, { "epoch": 0.4, "learning_rate": 2.4071028375203327e-05, "loss": 0.8181, "step": 2187 }, { "epoch": 0.4, "learning_rate": 2.4068317368516176e-05, "loss": 1.5728, "step": 2188 }, { "epoch": 0.4, "learning_rate": 2.4065606361829024e-05, "loss": 0.9413, "step": 2189 }, { "epoch": 0.4, "learning_rate": 2.4062895355141876e-05, "loss": 1.3664, "step": 2190 }, { "epoch": 0.4, "learning_rate": 2.4060184348454728e-05, "loss": 2.0352, "step": 2191 }, { "epoch": 0.4, "learning_rate": 2.4057473341767576e-05, "loss": 1.5712, "step": 2192 }, { "epoch": 0.4, "learning_rate": 2.4054762335080428e-05, "loss": 1.2458, "step": 2193 }, { "epoch": 0.4, "learning_rate": 2.405205132839328e-05, "loss": 0.4059, "step": 2194 }, { "epoch": 0.4, "learning_rate": 2.4049340321706125e-05, "loss": 1.1423, "step": 2195 }, { "epoch": 0.4, "learning_rate": 2.4046629315018977e-05, "loss": 0.9882, "step": 2196 }, { "epoch": 0.4, "learning_rate": 2.4043918308331825e-05, "loss": 1.1364, "step": 2197 }, { "epoch": 0.4, "learning_rate": 2.4041207301644677e-05, "loss": 1.0111, "step": 2198 }, { "epoch": 0.4, "learning_rate": 2.403849629495753e-05, "loss": 0.8417, "step": 2199 }, { "epoch": 0.4, "learning_rate": 2.4035785288270377e-05, "loss": 0.5473, "step": 2200 }, { "epoch": 0.4, "learning_rate": 2.403307428158323e-05, "loss": 1.8869, "step": 2201 }, { "epoch": 0.4, "learning_rate": 2.403036327489608e-05, "loss": 0.8166, "step": 2202 }, { "epoch": 0.4, "learning_rate": 2.4027652268208926e-05, "loss": 1.572, "step": 2203 }, { "epoch": 0.4, "learning_rate": 2.4024941261521778e-05, "loss": 1.2963, "step": 2204 }, { "epoch": 0.4, "learning_rate": 2.402223025483463e-05, "loss": 0.806, "step": 2205 }, { "epoch": 0.4, "learning_rate": 2.4019519248147478e-05, "loss": 1.3091, "step": 2206 }, { "epoch": 0.4, "learning_rate": 2.401680824146033e-05, "loss": 0.7697, "step": 2207 }, { "epoch": 0.4, "learning_rate": 2.401409723477318e-05, "loss": 0.5964, "step": 2208 }, { "epoch": 0.4, "learning_rate": 2.401138622808603e-05, "loss": 1.8209, "step": 2209 }, { "epoch": 0.4, "learning_rate": 2.4008675221398882e-05, "loss": 0.5016, "step": 2210 }, { "epoch": 0.4, "learning_rate": 2.4005964214711727e-05, "loss": 0.6549, "step": 2211 }, { "epoch": 0.4, "learning_rate": 2.400325320802458e-05, "loss": 1.2623, "step": 2212 }, { "epoch": 0.4, "learning_rate": 2.400054220133743e-05, "loss": 1.748, "step": 2213 }, { "epoch": 0.4, "learning_rate": 2.399783119465028e-05, "loss": 1.0977, "step": 2214 }, { "epoch": 0.4, "learning_rate": 2.399512018796313e-05, "loss": 1.1477, "step": 2215 }, { "epoch": 0.4, "learning_rate": 2.3992409181275983e-05, "loss": 1.1651, "step": 2216 }, { "epoch": 0.4, "learning_rate": 2.398969817458883e-05, "loss": 0.5238, "step": 2217 }, { "epoch": 0.4, "learning_rate": 2.3986987167901683e-05, "loss": 0.9948, "step": 2218 }, { "epoch": 0.4, "learning_rate": 2.3984276161214532e-05, "loss": 1.0966, "step": 2219 }, { "epoch": 0.4, "learning_rate": 2.398156515452738e-05, "loss": 0.8436, "step": 2220 }, { "epoch": 0.4, "learning_rate": 2.3978854147840232e-05, "loss": 1.0425, "step": 2221 }, { "epoch": 0.4, "learning_rate": 2.397614314115308e-05, "loss": 0.9452, "step": 2222 }, { "epoch": 0.4, "learning_rate": 2.3973432134465932e-05, "loss": 0.8566, "step": 2223 }, { "epoch": 0.4, "learning_rate": 2.3970721127778784e-05, "loss": 1.2939, "step": 2224 }, { "epoch": 0.4, "learning_rate": 2.3968010121091633e-05, "loss": 0.7339, "step": 2225 }, { "epoch": 0.4, "learning_rate": 2.3965299114404484e-05, "loss": 1.3667, "step": 2226 }, { "epoch": 0.4, "learning_rate": 2.3962588107717333e-05, "loss": 1.5799, "step": 2227 }, { "epoch": 0.4, "learning_rate": 2.395987710103018e-05, "loss": 1.4564, "step": 2228 }, { "epoch": 0.4, "learning_rate": 2.3957166094343033e-05, "loss": 0.7972, "step": 2229 }, { "epoch": 0.4, "learning_rate": 2.3954455087655885e-05, "loss": 0.9817, "step": 2230 }, { "epoch": 0.4, "learning_rate": 2.3951744080968733e-05, "loss": 1.6074, "step": 2231 }, { "epoch": 0.4, "learning_rate": 2.3949033074281585e-05, "loss": 0.9172, "step": 2232 }, { "epoch": 0.4, "learning_rate": 2.3946322067594434e-05, "loss": 1.3537, "step": 2233 }, { "epoch": 0.4, "learning_rate": 2.3943611060907285e-05, "loss": 1.0103, "step": 2234 }, { "epoch": 0.4, "learning_rate": 2.3940900054220134e-05, "loss": 1.4013, "step": 2235 }, { "epoch": 0.4, "learning_rate": 2.3938189047532982e-05, "loss": 1.469, "step": 2236 }, { "epoch": 0.4, "learning_rate": 2.3935478040845834e-05, "loss": 0.4735, "step": 2237 }, { "epoch": 0.4, "learning_rate": 2.3932767034158686e-05, "loss": 2.148, "step": 2238 }, { "epoch": 0.4, "learning_rate": 2.3930056027471534e-05, "loss": 1.2625, "step": 2239 }, { "epoch": 0.4, "learning_rate": 2.3927345020784386e-05, "loss": 1.4482, "step": 2240 }, { "epoch": 0.41, "learning_rate": 2.3924634014097238e-05, "loss": 1.1109, "step": 2241 }, { "epoch": 0.41, "learning_rate": 2.3921923007410087e-05, "loss": 1.5921, "step": 2242 }, { "epoch": 0.41, "learning_rate": 2.3919212000722935e-05, "loss": 0.7212, "step": 2243 }, { "epoch": 0.41, "learning_rate": 2.3916500994035783e-05, "loss": 1.4481, "step": 2244 }, { "epoch": 0.41, "learning_rate": 2.3913789987348635e-05, "loss": 1.4787, "step": 2245 }, { "epoch": 0.41, "learning_rate": 2.3911078980661487e-05, "loss": 1.2271, "step": 2246 }, { "epoch": 0.41, "learning_rate": 2.3908367973974335e-05, "loss": 1.4152, "step": 2247 }, { "epoch": 0.41, "learning_rate": 2.3905656967287187e-05, "loss": 1.7165, "step": 2248 }, { "epoch": 0.41, "learning_rate": 2.390294596060004e-05, "loss": 1.0711, "step": 2249 }, { "epoch": 0.41, "learning_rate": 2.3900234953912888e-05, "loss": 0.9369, "step": 2250 }, { "epoch": 0.41, "learning_rate": 2.3897523947225736e-05, "loss": 1.4862, "step": 2251 }, { "epoch": 0.41, "learning_rate": 2.3894812940538588e-05, "loss": 1.3677, "step": 2252 }, { "epoch": 0.41, "learning_rate": 2.3892101933851436e-05, "loss": 1.4574, "step": 2253 }, { "epoch": 0.41, "learning_rate": 2.3889390927164288e-05, "loss": 1.0956, "step": 2254 }, { "epoch": 0.41, "learning_rate": 2.3886679920477137e-05, "loss": 1.7279, "step": 2255 }, { "epoch": 0.41, "learning_rate": 2.388396891378999e-05, "loss": 0.8231, "step": 2256 }, { "epoch": 0.41, "learning_rate": 2.388125790710284e-05, "loss": 1.5255, "step": 2257 }, { "epoch": 0.41, "learning_rate": 2.387854690041569e-05, "loss": 1.5622, "step": 2258 }, { "epoch": 0.41, "learning_rate": 2.3875835893728537e-05, "loss": 1.1532, "step": 2259 }, { "epoch": 0.41, "learning_rate": 2.387312488704139e-05, "loss": 1.3246, "step": 2260 }, { "epoch": 0.41, "learning_rate": 2.3870413880354237e-05, "loss": 2.099, "step": 2261 }, { "epoch": 0.41, "learning_rate": 2.386770287366709e-05, "loss": 0.9531, "step": 2262 }, { "epoch": 0.41, "learning_rate": 2.386499186697994e-05, "loss": 1.454, "step": 2263 }, { "epoch": 0.41, "learning_rate": 2.386228086029279e-05, "loss": 1.2043, "step": 2264 }, { "epoch": 0.41, "learning_rate": 2.385956985360564e-05, "loss": 0.9555, "step": 2265 }, { "epoch": 0.41, "learning_rate": 2.385685884691849e-05, "loss": 1.2246, "step": 2266 }, { "epoch": 0.41, "learning_rate": 2.3854147840231338e-05, "loss": 1.6236, "step": 2267 }, { "epoch": 0.41, "learning_rate": 2.385143683354419e-05, "loss": 1.2358, "step": 2268 }, { "epoch": 0.41, "learning_rate": 2.384872582685704e-05, "loss": 1.1675, "step": 2269 }, { "epoch": 0.41, "learning_rate": 2.384601482016989e-05, "loss": 1.1806, "step": 2270 }, { "epoch": 0.41, "learning_rate": 2.3843303813482742e-05, "loss": 0.9505, "step": 2271 }, { "epoch": 0.41, "learning_rate": 2.384059280679559e-05, "loss": 1.1672, "step": 2272 }, { "epoch": 0.41, "learning_rate": 2.3837881800108442e-05, "loss": 2.185, "step": 2273 }, { "epoch": 0.41, "learning_rate": 2.3835170793421294e-05, "loss": 1.0071, "step": 2274 }, { "epoch": 0.41, "learning_rate": 2.383245978673414e-05, "loss": 1.3398, "step": 2275 }, { "epoch": 0.41, "learning_rate": 2.382974878004699e-05, "loss": 1.3656, "step": 2276 }, { "epoch": 0.41, "learning_rate": 2.382703777335984e-05, "loss": 1.3683, "step": 2277 }, { "epoch": 0.41, "learning_rate": 2.382432676667269e-05, "loss": 1.0329, "step": 2278 }, { "epoch": 0.41, "learning_rate": 2.3821615759985543e-05, "loss": 1.182, "step": 2279 }, { "epoch": 0.41, "learning_rate": 2.381890475329839e-05, "loss": 1.9082, "step": 2280 }, { "epoch": 0.41, "learning_rate": 2.3816193746611243e-05, "loss": 1.591, "step": 2281 }, { "epoch": 0.41, "learning_rate": 2.3813482739924095e-05, "loss": 1.397, "step": 2282 }, { "epoch": 0.41, "learning_rate": 2.381077173323694e-05, "loss": 0.9524, "step": 2283 }, { "epoch": 0.41, "learning_rate": 2.3808060726549792e-05, "loss": 0.6558, "step": 2284 }, { "epoch": 0.41, "learning_rate": 2.3805349719862644e-05, "loss": 1.3888, "step": 2285 }, { "epoch": 0.41, "learning_rate": 2.3802638713175492e-05, "loss": 0.799, "step": 2286 }, { "epoch": 0.41, "learning_rate": 2.3799927706488344e-05, "loss": 1.2298, "step": 2287 }, { "epoch": 0.41, "learning_rate": 2.3797216699801193e-05, "loss": 1.437, "step": 2288 }, { "epoch": 0.41, "learning_rate": 2.3794505693114045e-05, "loss": 1.1332, "step": 2289 }, { "epoch": 0.41, "learning_rate": 2.3791794686426896e-05, "loss": 1.0847, "step": 2290 }, { "epoch": 0.41, "learning_rate": 2.378908367973974e-05, "loss": 1.5278, "step": 2291 }, { "epoch": 0.41, "learning_rate": 2.3786372673052593e-05, "loss": 1.1832, "step": 2292 }, { "epoch": 0.41, "learning_rate": 2.3783661666365445e-05, "loss": 1.2184, "step": 2293 }, { "epoch": 0.41, "learning_rate": 2.3780950659678293e-05, "loss": 1.2552, "step": 2294 }, { "epoch": 0.41, "learning_rate": 2.3778239652991145e-05, "loss": 1.9193, "step": 2295 }, { "epoch": 0.41, "learning_rate": 2.3775528646303997e-05, "loss": 0.5525, "step": 2296 }, { "epoch": 0.42, "learning_rate": 2.3772817639616846e-05, "loss": 1.1558, "step": 2297 }, { "epoch": 0.42, "learning_rate": 2.3770106632929697e-05, "loss": 0.8142, "step": 2298 }, { "epoch": 0.42, "learning_rate": 2.3767395626242542e-05, "loss": 0.6157, "step": 2299 }, { "epoch": 0.42, "learning_rate": 2.3764684619555394e-05, "loss": 0.6937, "step": 2300 }, { "epoch": 0.42, "learning_rate": 2.3761973612868246e-05, "loss": 1.5899, "step": 2301 }, { "epoch": 0.42, "learning_rate": 2.3759262606181095e-05, "loss": 1.4488, "step": 2302 }, { "epoch": 0.42, "learning_rate": 2.3756551599493946e-05, "loss": 1.0313, "step": 2303 }, { "epoch": 0.42, "learning_rate": 2.3753840592806798e-05, "loss": 1.4468, "step": 2304 }, { "epoch": 0.42, "learning_rate": 2.3751129586119647e-05, "loss": 1.7909, "step": 2305 }, { "epoch": 0.42, "learning_rate": 2.37484185794325e-05, "loss": 1.1526, "step": 2306 }, { "epoch": 0.42, "learning_rate": 2.3745707572745347e-05, "loss": 0.9367, "step": 2307 }, { "epoch": 0.42, "learning_rate": 2.3742996566058195e-05, "loss": 1.3155, "step": 2308 }, { "epoch": 0.42, "learning_rate": 2.3740285559371047e-05, "loss": 1.8858, "step": 2309 }, { "epoch": 0.42, "learning_rate": 2.3737574552683896e-05, "loss": 0.7563, "step": 2310 }, { "epoch": 0.42, "learning_rate": 2.3734863545996747e-05, "loss": 1.9904, "step": 2311 }, { "epoch": 0.42, "learning_rate": 2.37321525393096e-05, "loss": 1.2404, "step": 2312 }, { "epoch": 0.42, "learning_rate": 2.3729441532622448e-05, "loss": 1.5479, "step": 2313 }, { "epoch": 0.42, "learning_rate": 2.37267305259353e-05, "loss": 0.5804, "step": 2314 }, { "epoch": 0.42, "learning_rate": 2.3724019519248148e-05, "loss": 0.7561, "step": 2315 }, { "epoch": 0.42, "learning_rate": 2.3721308512560996e-05, "loss": 0.4602, "step": 2316 }, { "epoch": 0.42, "learning_rate": 2.3718597505873848e-05, "loss": 1.0947, "step": 2317 }, { "epoch": 0.42, "learning_rate": 2.37158864991867e-05, "loss": 1.8721, "step": 2318 }, { "epoch": 0.42, "learning_rate": 2.371317549249955e-05, "loss": 0.8899, "step": 2319 }, { "epoch": 0.42, "learning_rate": 2.37104644858124e-05, "loss": 1.2732, "step": 2320 }, { "epoch": 0.42, "learning_rate": 2.370775347912525e-05, "loss": 1.3942, "step": 2321 }, { "epoch": 0.42, "learning_rate": 2.3705042472438097e-05, "loss": 0.7586, "step": 2322 }, { "epoch": 0.42, "learning_rate": 2.370233146575095e-05, "loss": 1.1311, "step": 2323 }, { "epoch": 0.42, "learning_rate": 2.3699620459063798e-05, "loss": 0.996, "step": 2324 }, { "epoch": 0.42, "learning_rate": 2.369690945237665e-05, "loss": 0.6677, "step": 2325 }, { "epoch": 0.42, "learning_rate": 2.36941984456895e-05, "loss": 0.7442, "step": 2326 }, { "epoch": 0.42, "learning_rate": 2.369148743900235e-05, "loss": 1.1012, "step": 2327 }, { "epoch": 0.42, "learning_rate": 2.36887764323152e-05, "loss": 0.993, "step": 2328 }, { "epoch": 0.42, "learning_rate": 2.3686065425628053e-05, "loss": 1.9037, "step": 2329 }, { "epoch": 0.42, "learning_rate": 2.36833544189409e-05, "loss": 1.5857, "step": 2330 }, { "epoch": 0.42, "learning_rate": 2.368064341225375e-05, "loss": 1.3356, "step": 2331 }, { "epoch": 0.42, "learning_rate": 2.36779324055666e-05, "loss": 2.1125, "step": 2332 }, { "epoch": 0.42, "learning_rate": 2.367522139887945e-05, "loss": 1.1604, "step": 2333 }, { "epoch": 0.42, "learning_rate": 2.3672510392192302e-05, "loss": 1.1784, "step": 2334 }, { "epoch": 0.42, "learning_rate": 2.366979938550515e-05, "loss": 0.9955, "step": 2335 }, { "epoch": 0.42, "learning_rate": 2.3667088378818003e-05, "loss": 1.8163, "step": 2336 }, { "epoch": 0.42, "learning_rate": 2.3664377372130854e-05, "loss": 1.5226, "step": 2337 }, { "epoch": 0.42, "learning_rate": 2.36616663654437e-05, "loss": 1.2592, "step": 2338 }, { "epoch": 0.42, "learning_rate": 2.365895535875655e-05, "loss": 0.8219, "step": 2339 }, { "epoch": 0.42, "learning_rate": 2.3656244352069403e-05, "loss": 1.2998, "step": 2340 }, { "epoch": 0.42, "learning_rate": 2.365353334538225e-05, "loss": 0.5786, "step": 2341 }, { "epoch": 0.42, "learning_rate": 2.3650822338695103e-05, "loss": 0.7553, "step": 2342 }, { "epoch": 0.42, "learning_rate": 2.3648111332007952e-05, "loss": 0.8261, "step": 2343 }, { "epoch": 0.42, "learning_rate": 2.3645400325320804e-05, "loss": 0.8902, "step": 2344 }, { "epoch": 0.42, "learning_rate": 2.3642689318633655e-05, "loss": 1.4218, "step": 2345 }, { "epoch": 0.42, "learning_rate": 2.36399783119465e-05, "loss": 2.0535, "step": 2346 }, { "epoch": 0.42, "learning_rate": 2.3637267305259352e-05, "loss": 1.106, "step": 2347 }, { "epoch": 0.42, "learning_rate": 2.3634556298572204e-05, "loss": 0.9401, "step": 2348 }, { "epoch": 0.42, "learning_rate": 2.3631845291885053e-05, "loss": 1.1768, "step": 2349 }, { "epoch": 0.42, "learning_rate": 2.3629134285197904e-05, "loss": 1.1693, "step": 2350 }, { "epoch": 0.42, "learning_rate": 2.3626423278510756e-05, "loss": 1.5217, "step": 2351 }, { "epoch": 0.43, "learning_rate": 2.3623712271823605e-05, "loss": 1.3653, "step": 2352 }, { "epoch": 0.43, "learning_rate": 2.3621001265136456e-05, "loss": 1.1751, "step": 2353 }, { "epoch": 0.43, "learning_rate": 2.36182902584493e-05, "loss": 0.6026, "step": 2354 }, { "epoch": 0.43, "learning_rate": 2.3615579251762153e-05, "loss": 1.0704, "step": 2355 }, { "epoch": 0.43, "learning_rate": 2.3612868245075005e-05, "loss": 0.6952, "step": 2356 }, { "epoch": 0.43, "learning_rate": 2.3610157238387854e-05, "loss": 1.0285, "step": 2357 }, { "epoch": 0.43, "learning_rate": 2.3607446231700705e-05, "loss": 0.8762, "step": 2358 }, { "epoch": 0.43, "learning_rate": 2.3604735225013557e-05, "loss": 1.2487, "step": 2359 }, { "epoch": 0.43, "learning_rate": 2.3602024218326406e-05, "loss": 1.5772, "step": 2360 }, { "epoch": 0.43, "learning_rate": 2.3599313211639258e-05, "loss": 1.4174, "step": 2361 }, { "epoch": 0.43, "learning_rate": 2.3596602204952106e-05, "loss": 0.9224, "step": 2362 }, { "epoch": 0.43, "learning_rate": 2.3593891198264954e-05, "loss": 1.2349, "step": 2363 }, { "epoch": 0.43, "learning_rate": 2.3591180191577806e-05, "loss": 0.9831, "step": 2364 }, { "epoch": 0.43, "learning_rate": 2.3588469184890655e-05, "loss": 0.8056, "step": 2365 }, { "epoch": 0.43, "learning_rate": 2.3585758178203507e-05, "loss": 0.8357, "step": 2366 }, { "epoch": 0.43, "learning_rate": 2.358304717151636e-05, "loss": 1.8032, "step": 2367 }, { "epoch": 0.43, "learning_rate": 2.3580336164829207e-05, "loss": 1.4574, "step": 2368 }, { "epoch": 0.43, "learning_rate": 2.357762515814206e-05, "loss": 1.0022, "step": 2369 }, { "epoch": 0.43, "learning_rate": 2.3574914151454907e-05, "loss": 2.7011, "step": 2370 }, { "epoch": 0.43, "learning_rate": 2.3572203144767756e-05, "loss": 1.3247, "step": 2371 }, { "epoch": 0.43, "learning_rate": 2.3569492138080607e-05, "loss": 1.3479, "step": 2372 }, { "epoch": 0.43, "learning_rate": 2.356678113139346e-05, "loss": 1.1369, "step": 2373 }, { "epoch": 0.43, "learning_rate": 2.3564070124706308e-05, "loss": 0.927, "step": 2374 }, { "epoch": 0.43, "learning_rate": 2.356135911801916e-05, "loss": 1.6194, "step": 2375 }, { "epoch": 0.43, "learning_rate": 2.3558648111332008e-05, "loss": 1.6716, "step": 2376 }, { "epoch": 0.43, "learning_rate": 2.355593710464486e-05, "loss": 1.3711, "step": 2377 }, { "epoch": 0.43, "learning_rate": 2.3553226097957708e-05, "loss": 1.3161, "step": 2378 }, { "epoch": 0.43, "learning_rate": 2.3550515091270557e-05, "loss": 1.923, "step": 2379 }, { "epoch": 0.43, "learning_rate": 2.354780408458341e-05, "loss": 1.018, "step": 2380 }, { "epoch": 0.43, "learning_rate": 2.354509307789626e-05, "loss": 0.9533, "step": 2381 }, { "epoch": 0.43, "learning_rate": 2.354238207120911e-05, "loss": 1.6427, "step": 2382 }, { "epoch": 0.43, "learning_rate": 2.353967106452196e-05, "loss": 1.7822, "step": 2383 }, { "epoch": 0.43, "learning_rate": 2.3536960057834812e-05, "loss": 1.1497, "step": 2384 }, { "epoch": 0.43, "learning_rate": 2.353424905114766e-05, "loss": 1.1434, "step": 2385 }, { "epoch": 0.43, "learning_rate": 2.353153804446051e-05, "loss": 1.343, "step": 2386 }, { "epoch": 0.43, "learning_rate": 2.352882703777336e-05, "loss": 0.659, "step": 2387 }, { "epoch": 0.43, "learning_rate": 2.352611603108621e-05, "loss": 1.2358, "step": 2388 }, { "epoch": 0.43, "learning_rate": 2.352340502439906e-05, "loss": 1.0148, "step": 2389 }, { "epoch": 0.43, "learning_rate": 2.352069401771191e-05, "loss": 1.0435, "step": 2390 }, { "epoch": 0.43, "learning_rate": 2.351798301102476e-05, "loss": 0.6263, "step": 2391 }, { "epoch": 0.43, "learning_rate": 2.3515272004337613e-05, "loss": 1.5884, "step": 2392 }, { "epoch": 0.43, "learning_rate": 2.3512560997650462e-05, "loss": 0.7568, "step": 2393 }, { "epoch": 0.43, "learning_rate": 2.350984999096331e-05, "loss": 0.9756, "step": 2394 }, { "epoch": 0.43, "learning_rate": 2.3507138984276162e-05, "loss": 1.2026, "step": 2395 }, { "epoch": 0.43, "learning_rate": 2.350442797758901e-05, "loss": 1.0988, "step": 2396 }, { "epoch": 0.43, "learning_rate": 2.3501716970901862e-05, "loss": 0.9797, "step": 2397 }, { "epoch": 0.43, "learning_rate": 2.3499005964214714e-05, "loss": 1.2356, "step": 2398 }, { "epoch": 0.43, "learning_rate": 2.3496294957527563e-05, "loss": 0.9085, "step": 2399 }, { "epoch": 0.43, "learning_rate": 2.3493583950840414e-05, "loss": 1.3297, "step": 2400 }, { "epoch": 0.43, "learning_rate": 2.3490872944153263e-05, "loss": 0.8318, "step": 2401 }, { "epoch": 0.43, "learning_rate": 2.348816193746611e-05, "loss": 1.0876, "step": 2402 }, { "epoch": 0.43, "learning_rate": 2.3485450930778963e-05, "loss": 1.3799, "step": 2403 }, { "epoch": 0.43, "learning_rate": 2.348273992409181e-05, "loss": 1.83, "step": 2404 }, { "epoch": 0.43, "learning_rate": 2.3480028917404663e-05, "loss": 1.8694, "step": 2405 }, { "epoch": 0.43, "learning_rate": 2.3477317910717515e-05, "loss": 1.3406, "step": 2406 }, { "epoch": 0.44, "learning_rate": 2.3474606904030364e-05, "loss": 1.4933, "step": 2407 }, { "epoch": 0.44, "learning_rate": 2.3471895897343216e-05, "loss": 1.1008, "step": 2408 }, { "epoch": 0.44, "learning_rate": 2.3469184890656067e-05, "loss": 1.688, "step": 2409 }, { "epoch": 0.44, "learning_rate": 2.3466473883968912e-05, "loss": 0.7195, "step": 2410 }, { "epoch": 0.44, "learning_rate": 2.3463762877281764e-05, "loss": 1.2534, "step": 2411 }, { "epoch": 0.44, "learning_rate": 2.3461051870594613e-05, "loss": 1.1997, "step": 2412 }, { "epoch": 0.44, "learning_rate": 2.3458340863907465e-05, "loss": 1.089, "step": 2413 }, { "epoch": 0.44, "learning_rate": 2.3455629857220316e-05, "loss": 2.0519, "step": 2414 }, { "epoch": 0.44, "learning_rate": 2.3452918850533165e-05, "loss": 1.2082, "step": 2415 }, { "epoch": 0.44, "learning_rate": 2.3450207843846017e-05, "loss": 1.1277, "step": 2416 }, { "epoch": 0.44, "learning_rate": 2.344749683715887e-05, "loss": 0.9234, "step": 2417 }, { "epoch": 0.44, "learning_rate": 2.3444785830471714e-05, "loss": 1.5067, "step": 2418 }, { "epoch": 0.44, "learning_rate": 2.3442074823784565e-05, "loss": 1.1323, "step": 2419 }, { "epoch": 0.44, "learning_rate": 2.3439363817097417e-05, "loss": 1.0997, "step": 2420 }, { "epoch": 0.44, "learning_rate": 2.3436652810410266e-05, "loss": 1.2511, "step": 2421 }, { "epoch": 0.44, "learning_rate": 2.3433941803723117e-05, "loss": 1.1347, "step": 2422 }, { "epoch": 0.44, "learning_rate": 2.3431230797035966e-05, "loss": 1.1108, "step": 2423 }, { "epoch": 0.44, "learning_rate": 2.3428519790348818e-05, "loss": 1.4562, "step": 2424 }, { "epoch": 0.44, "learning_rate": 2.342580878366167e-05, "loss": 1.5261, "step": 2425 }, { "epoch": 0.44, "learning_rate": 2.3423097776974515e-05, "loss": 1.6823, "step": 2426 }, { "epoch": 0.44, "learning_rate": 2.3420386770287366e-05, "loss": 1.0256, "step": 2427 }, { "epoch": 0.44, "learning_rate": 2.3417675763600218e-05, "loss": 0.8976, "step": 2428 }, { "epoch": 0.44, "learning_rate": 2.3414964756913067e-05, "loss": 0.9809, "step": 2429 }, { "epoch": 0.44, "learning_rate": 2.341225375022592e-05, "loss": 1.3989, "step": 2430 }, { "epoch": 0.44, "learning_rate": 2.340954274353877e-05, "loss": 1.0022, "step": 2431 }, { "epoch": 0.44, "learning_rate": 2.340683173685162e-05, "loss": 0.9726, "step": 2432 }, { "epoch": 0.44, "learning_rate": 2.340412073016447e-05, "loss": 1.3563, "step": 2433 }, { "epoch": 0.44, "learning_rate": 2.3401409723477316e-05, "loss": 2.009, "step": 2434 }, { "epoch": 0.44, "learning_rate": 2.3398698716790168e-05, "loss": 1.0674, "step": 2435 }, { "epoch": 0.44, "learning_rate": 2.339598771010302e-05, "loss": 0.9326, "step": 2436 }, { "epoch": 0.44, "learning_rate": 2.3393276703415868e-05, "loss": 1.3494, "step": 2437 }, { "epoch": 0.44, "learning_rate": 2.339056569672872e-05, "loss": 0.5545, "step": 2438 }, { "epoch": 0.44, "learning_rate": 2.338785469004157e-05, "loss": 0.5226, "step": 2439 }, { "epoch": 0.44, "learning_rate": 2.338514368335442e-05, "loss": 0.9466, "step": 2440 }, { "epoch": 0.44, "learning_rate": 2.3382432676667272e-05, "loss": 1.41, "step": 2441 }, { "epoch": 0.44, "learning_rate": 2.337972166998012e-05, "loss": 1.3288, "step": 2442 }, { "epoch": 0.44, "learning_rate": 2.337701066329297e-05, "loss": 1.1152, "step": 2443 }, { "epoch": 0.44, "learning_rate": 2.337429965660582e-05, "loss": 1.3215, "step": 2444 }, { "epoch": 0.44, "learning_rate": 2.337158864991867e-05, "loss": 0.8241, "step": 2445 }, { "epoch": 0.44, "learning_rate": 2.336887764323152e-05, "loss": 0.8262, "step": 2446 }, { "epoch": 0.44, "learning_rate": 2.3366166636544373e-05, "loss": 1.5372, "step": 2447 }, { "epoch": 0.44, "learning_rate": 2.336345562985722e-05, "loss": 1.6053, "step": 2448 }, { "epoch": 0.44, "learning_rate": 2.336074462317007e-05, "loss": 1.5333, "step": 2449 }, { "epoch": 0.44, "learning_rate": 2.335803361648292e-05, "loss": 1.6691, "step": 2450 }, { "epoch": 0.44, "learning_rate": 2.335532260979577e-05, "loss": 1.4022, "step": 2451 }, { "epoch": 0.44, "learning_rate": 2.335261160310862e-05, "loss": 1.8385, "step": 2452 }, { "epoch": 0.44, "learning_rate": 2.3349900596421473e-05, "loss": 1.2039, "step": 2453 }, { "epoch": 0.44, "learning_rate": 2.3347189589734322e-05, "loss": 1.3524, "step": 2454 }, { "epoch": 0.44, "learning_rate": 2.3344478583047174e-05, "loss": 1.0112, "step": 2455 }, { "epoch": 0.44, "learning_rate": 2.3341767576360022e-05, "loss": 1.6196, "step": 2456 }, { "epoch": 0.44, "learning_rate": 2.333905656967287e-05, "loss": 1.1512, "step": 2457 }, { "epoch": 0.44, "learning_rate": 2.3336345562985722e-05, "loss": 1.3548, "step": 2458 }, { "epoch": 0.44, "learning_rate": 2.333363455629857e-05, "loss": 1.4898, "step": 2459 }, { "epoch": 0.44, "learning_rate": 2.3330923549611423e-05, "loss": 1.545, "step": 2460 }, { "epoch": 0.44, "learning_rate": 2.3328212542924274e-05, "loss": 0.7086, "step": 2461 }, { "epoch": 0.44, "learning_rate": 2.3325501536237123e-05, "loss": 1.124, "step": 2462 }, { "epoch": 0.45, "learning_rate": 2.3322790529549975e-05, "loss": 1.2361, "step": 2463 }, { "epoch": 0.45, "learning_rate": 2.3320079522862826e-05, "loss": 1.2221, "step": 2464 }, { "epoch": 0.45, "learning_rate": 2.331736851617567e-05, "loss": 1.0439, "step": 2465 }, { "epoch": 0.45, "learning_rate": 2.3314657509488523e-05, "loss": 1.0238, "step": 2466 }, { "epoch": 0.45, "learning_rate": 2.3311946502801372e-05, "loss": 1.368, "step": 2467 }, { "epoch": 0.45, "learning_rate": 2.3309235496114224e-05, "loss": 0.9674, "step": 2468 }, { "epoch": 0.45, "learning_rate": 2.3306524489427075e-05, "loss": 1.3677, "step": 2469 }, { "epoch": 0.45, "learning_rate": 2.3303813482739924e-05, "loss": 1.2976, "step": 2470 }, { "epoch": 0.45, "learning_rate": 2.3301102476052776e-05, "loss": 0.6421, "step": 2471 }, { "epoch": 0.45, "learning_rate": 2.3298391469365628e-05, "loss": 1.0993, "step": 2472 }, { "epoch": 0.45, "learning_rate": 2.3295680462678473e-05, "loss": 0.7451, "step": 2473 }, { "epoch": 0.45, "learning_rate": 2.3292969455991324e-05, "loss": 1.2589, "step": 2474 }, { "epoch": 0.45, "learning_rate": 2.3290258449304176e-05, "loss": 1.3327, "step": 2475 }, { "epoch": 0.45, "learning_rate": 2.3287547442617025e-05, "loss": 1.2164, "step": 2476 }, { "epoch": 0.45, "learning_rate": 2.3284836435929877e-05, "loss": 1.4598, "step": 2477 }, { "epoch": 0.45, "learning_rate": 2.3282125429242725e-05, "loss": 1.0097, "step": 2478 }, { "epoch": 0.45, "learning_rate": 2.3279414422555577e-05, "loss": 0.9791, "step": 2479 }, { "epoch": 0.45, "learning_rate": 2.327670341586843e-05, "loss": 1.3603, "step": 2480 }, { "epoch": 0.45, "learning_rate": 2.3273992409181274e-05, "loss": 1.0052, "step": 2481 }, { "epoch": 0.45, "learning_rate": 2.3271281402494126e-05, "loss": 1.3255, "step": 2482 }, { "epoch": 0.45, "learning_rate": 2.3268570395806977e-05, "loss": 0.8956, "step": 2483 }, { "epoch": 0.45, "learning_rate": 2.3265859389119826e-05, "loss": 0.9221, "step": 2484 }, { "epoch": 0.45, "learning_rate": 2.3263148382432678e-05, "loss": 1.1468, "step": 2485 }, { "epoch": 0.45, "learning_rate": 2.326043737574553e-05, "loss": 1.5179, "step": 2486 }, { "epoch": 0.45, "learning_rate": 2.3257726369058378e-05, "loss": 1.0728, "step": 2487 }, { "epoch": 0.45, "learning_rate": 2.325501536237123e-05, "loss": 0.9659, "step": 2488 }, { "epoch": 0.45, "learning_rate": 2.3252304355684075e-05, "loss": 1.2329, "step": 2489 }, { "epoch": 0.45, "learning_rate": 2.3249593348996927e-05, "loss": 1.0801, "step": 2490 }, { "epoch": 0.45, "learning_rate": 2.324688234230978e-05, "loss": 1.1115, "step": 2491 }, { "epoch": 0.45, "learning_rate": 2.3244171335622627e-05, "loss": 0.8596, "step": 2492 }, { "epoch": 0.45, "learning_rate": 2.324146032893548e-05, "loss": 0.969, "step": 2493 }, { "epoch": 0.45, "learning_rate": 2.323874932224833e-05, "loss": 2.0738, "step": 2494 }, { "epoch": 0.45, "learning_rate": 2.323603831556118e-05, "loss": 0.8447, "step": 2495 }, { "epoch": 0.45, "learning_rate": 2.323332730887403e-05, "loss": 1.284, "step": 2496 }, { "epoch": 0.45, "learning_rate": 2.323061630218688e-05, "loss": 1.0287, "step": 2497 }, { "epoch": 0.45, "learning_rate": 2.3227905295499728e-05, "loss": 2.33, "step": 2498 }, { "epoch": 0.45, "learning_rate": 2.322519428881258e-05, "loss": 0.7537, "step": 2499 }, { "epoch": 0.45, "learning_rate": 2.3222483282125428e-05, "loss": 0.9814, "step": 2500 }, { "epoch": 0.45, "learning_rate": 2.321977227543828e-05, "loss": 0.9191, "step": 2501 }, { "epoch": 0.45, "learning_rate": 2.321706126875113e-05, "loss": 1.7856, "step": 2502 }, { "epoch": 0.45, "learning_rate": 2.321435026206398e-05, "loss": 0.8673, "step": 2503 }, { "epoch": 0.45, "learning_rate": 2.3211639255376832e-05, "loss": 1.1204, "step": 2504 }, { "epoch": 0.45, "learning_rate": 2.320892824868968e-05, "loss": 1.7684, "step": 2505 }, { "epoch": 0.45, "learning_rate": 2.320621724200253e-05, "loss": 1.1761, "step": 2506 }, { "epoch": 0.45, "learning_rate": 2.320350623531538e-05, "loss": 1.063, "step": 2507 }, { "epoch": 0.45, "learning_rate": 2.3200795228628232e-05, "loss": 1.3832, "step": 2508 }, { "epoch": 0.45, "learning_rate": 2.319808422194108e-05, "loss": 1.3464, "step": 2509 }, { "epoch": 0.45, "learning_rate": 2.3195373215253933e-05, "loss": 1.2805, "step": 2510 }, { "epoch": 0.45, "learning_rate": 2.319266220856678e-05, "loss": 1.3078, "step": 2511 }, { "epoch": 0.45, "learning_rate": 2.3189951201879633e-05, "loss": 1.5248, "step": 2512 }, { "epoch": 0.45, "learning_rate": 2.318724019519248e-05, "loss": 0.7933, "step": 2513 }, { "epoch": 0.45, "learning_rate": 2.318452918850533e-05, "loss": 1.4914, "step": 2514 }, { "epoch": 0.45, "learning_rate": 2.318181818181818e-05, "loss": 0.7121, "step": 2515 }, { "epoch": 0.45, "learning_rate": 2.3179107175131033e-05, "loss": 1.1224, "step": 2516 }, { "epoch": 0.45, "learning_rate": 2.3176396168443882e-05, "loss": 0.9486, "step": 2517 }, { "epoch": 0.46, "learning_rate": 2.3173685161756734e-05, "loss": 1.4888, "step": 2518 }, { "epoch": 0.46, "learning_rate": 2.3170974155069586e-05, "loss": 1.2805, "step": 2519 }, { "epoch": 0.46, "learning_rate": 2.3168263148382434e-05, "loss": 0.7885, "step": 2520 }, { "epoch": 0.46, "learning_rate": 2.3165552141695282e-05, "loss": 1.5259, "step": 2521 }, { "epoch": 0.46, "learning_rate": 2.3162841135008134e-05, "loss": 1.2773, "step": 2522 }, { "epoch": 0.46, "learning_rate": 2.3160130128320983e-05, "loss": 1.0892, "step": 2523 }, { "epoch": 0.46, "learning_rate": 2.3157419121633835e-05, "loss": 1.1343, "step": 2524 }, { "epoch": 0.46, "learning_rate": 2.3154708114946683e-05, "loss": 1.2165, "step": 2525 }, { "epoch": 0.46, "learning_rate": 2.3151997108259535e-05, "loss": 1.2809, "step": 2526 }, { "epoch": 0.46, "learning_rate": 2.3149286101572387e-05, "loss": 1.1479, "step": 2527 }, { "epoch": 0.46, "learning_rate": 2.3146575094885235e-05, "loss": 0.8648, "step": 2528 }, { "epoch": 0.46, "learning_rate": 2.3143864088198084e-05, "loss": 1.487, "step": 2529 }, { "epoch": 0.46, "learning_rate": 2.3141153081510935e-05, "loss": 1.0891, "step": 2530 }, { "epoch": 0.46, "learning_rate": 2.3138442074823784e-05, "loss": 1.8268, "step": 2531 }, { "epoch": 0.46, "learning_rate": 2.3135731068136636e-05, "loss": 1.9328, "step": 2532 }, { "epoch": 0.46, "learning_rate": 2.3133020061449487e-05, "loss": 2.0673, "step": 2533 }, { "epoch": 0.46, "learning_rate": 2.3130309054762336e-05, "loss": 1.3955, "step": 2534 }, { "epoch": 0.46, "learning_rate": 2.3127598048075188e-05, "loss": 1.1761, "step": 2535 }, { "epoch": 0.46, "learning_rate": 2.3124887041388036e-05, "loss": 1.2405, "step": 2536 }, { "epoch": 0.46, "learning_rate": 2.3122176034700885e-05, "loss": 1.1013, "step": 2537 }, { "epoch": 0.46, "learning_rate": 2.3119465028013736e-05, "loss": 1.1936, "step": 2538 }, { "epoch": 0.46, "learning_rate": 2.3116754021326585e-05, "loss": 1.0171, "step": 2539 }, { "epoch": 0.46, "learning_rate": 2.3114043014639437e-05, "loss": 1.074, "step": 2540 }, { "epoch": 0.46, "learning_rate": 2.311133200795229e-05, "loss": 0.9175, "step": 2541 }, { "epoch": 0.46, "learning_rate": 2.3108621001265137e-05, "loss": 1.0123, "step": 2542 }, { "epoch": 0.46, "learning_rate": 2.310590999457799e-05, "loss": 0.8281, "step": 2543 }, { "epoch": 0.46, "learning_rate": 2.310319898789084e-05, "loss": 1.4192, "step": 2544 }, { "epoch": 0.46, "learning_rate": 2.3100487981203686e-05, "loss": 0.9169, "step": 2545 }, { "epoch": 0.46, "learning_rate": 2.3097776974516537e-05, "loss": 0.8404, "step": 2546 }, { "epoch": 0.46, "learning_rate": 2.3095065967829386e-05, "loss": 1.1723, "step": 2547 }, { "epoch": 0.46, "learning_rate": 2.3092354961142238e-05, "loss": 1.0385, "step": 2548 }, { "epoch": 0.46, "learning_rate": 2.308964395445509e-05, "loss": 1.6477, "step": 2549 }, { "epoch": 0.46, "learning_rate": 2.3086932947767938e-05, "loss": 1.2986, "step": 2550 }, { "epoch": 0.46, "learning_rate": 2.308422194108079e-05, "loss": 1.7164, "step": 2551 }, { "epoch": 0.46, "learning_rate": 2.308151093439364e-05, "loss": 1.5366, "step": 2552 }, { "epoch": 0.46, "learning_rate": 2.3078799927706487e-05, "loss": 1.1967, "step": 2553 }, { "epoch": 0.46, "learning_rate": 2.307608892101934e-05, "loss": 0.9932, "step": 2554 }, { "epoch": 0.46, "learning_rate": 2.307337791433219e-05, "loss": 1.1254, "step": 2555 }, { "epoch": 0.46, "learning_rate": 2.307066690764504e-05, "loss": 1.6888, "step": 2556 }, { "epoch": 0.46, "learning_rate": 2.306795590095789e-05, "loss": 1.6402, "step": 2557 }, { "epoch": 0.46, "learning_rate": 2.306524489427074e-05, "loss": 1.0731, "step": 2558 }, { "epoch": 0.46, "learning_rate": 2.306253388758359e-05, "loss": 1.1868, "step": 2559 }, { "epoch": 0.46, "learning_rate": 2.3059822880896443e-05, "loss": 0.8945, "step": 2560 }, { "epoch": 0.46, "learning_rate": 2.3057111874209288e-05, "loss": 1.0833, "step": 2561 }, { "epoch": 0.46, "learning_rate": 2.305440086752214e-05, "loss": 1.6174, "step": 2562 }, { "epoch": 0.46, "learning_rate": 2.305168986083499e-05, "loss": 1.4126, "step": 2563 }, { "epoch": 0.46, "learning_rate": 2.304897885414784e-05, "loss": 1.0266, "step": 2564 }, { "epoch": 0.46, "learning_rate": 2.3046267847460692e-05, "loss": 1.3593, "step": 2565 }, { "epoch": 0.46, "learning_rate": 2.3043556840773544e-05, "loss": 0.8148, "step": 2566 }, { "epoch": 0.46, "learning_rate": 2.3040845834086392e-05, "loss": 1.19, "step": 2567 }, { "epoch": 0.46, "learning_rate": 2.3038134827399244e-05, "loss": 1.1812, "step": 2568 }, { "epoch": 0.46, "learning_rate": 2.303542382071209e-05, "loss": 1.1793, "step": 2569 }, { "epoch": 0.46, "learning_rate": 2.303271281402494e-05, "loss": 0.8835, "step": 2570 }, { "epoch": 0.46, "learning_rate": 2.3030001807337793e-05, "loss": 0.8831, "step": 2571 }, { "epoch": 0.46, "learning_rate": 2.302729080065064e-05, "loss": 1.2524, "step": 2572 }, { "epoch": 0.47, "learning_rate": 2.3024579793963493e-05, "loss": 1.1729, "step": 2573 }, { "epoch": 0.47, "learning_rate": 2.3021868787276345e-05, "loss": 0.6382, "step": 2574 }, { "epoch": 0.47, "learning_rate": 2.3019157780589193e-05, "loss": 1.1642, "step": 2575 }, { "epoch": 0.47, "learning_rate": 2.301644677390204e-05, "loss": 1.2007, "step": 2576 }, { "epoch": 0.47, "learning_rate": 2.3013735767214893e-05, "loss": 1.1283, "step": 2577 }, { "epoch": 0.47, "learning_rate": 2.3011024760527742e-05, "loss": 0.4128, "step": 2578 }, { "epoch": 0.47, "learning_rate": 2.3008313753840594e-05, "loss": 0.7594, "step": 2579 }, { "epoch": 0.47, "learning_rate": 2.3005602747153442e-05, "loss": 0.8315, "step": 2580 }, { "epoch": 0.47, "learning_rate": 2.3002891740466294e-05, "loss": 0.3069, "step": 2581 }, { "epoch": 0.47, "learning_rate": 2.3000180733779146e-05, "loss": 1.2659, "step": 2582 }, { "epoch": 0.47, "learning_rate": 2.2997469727091994e-05, "loss": 0.9657, "step": 2583 }, { "epoch": 0.47, "learning_rate": 2.2994758720404843e-05, "loss": 1.5858, "step": 2584 }, { "epoch": 0.47, "learning_rate": 2.2992047713717694e-05, "loss": 1.5045, "step": 2585 }, { "epoch": 0.47, "learning_rate": 2.2989336707030543e-05, "loss": 1.2794, "step": 2586 }, { "epoch": 0.47, "learning_rate": 2.2986625700343395e-05, "loss": 0.7852, "step": 2587 }, { "epoch": 0.47, "learning_rate": 2.2983914693656247e-05, "loss": 1.0453, "step": 2588 }, { "epoch": 0.47, "learning_rate": 2.2981203686969095e-05, "loss": 0.7578, "step": 2589 }, { "epoch": 0.47, "learning_rate": 2.2978492680281947e-05, "loss": 0.8432, "step": 2590 }, { "epoch": 0.47, "learning_rate": 2.2975781673594795e-05, "loss": 1.0325, "step": 2591 }, { "epoch": 0.47, "learning_rate": 2.2973070666907644e-05, "loss": 1.4529, "step": 2592 }, { "epoch": 0.47, "learning_rate": 2.2970359660220495e-05, "loss": 0.9481, "step": 2593 }, { "epoch": 0.47, "learning_rate": 2.2967648653533344e-05, "loss": 0.7675, "step": 2594 }, { "epoch": 0.47, "learning_rate": 2.2964937646846196e-05, "loss": 0.8807, "step": 2595 }, { "epoch": 0.47, "learning_rate": 2.2962226640159048e-05, "loss": 0.8559, "step": 2596 }, { "epoch": 0.47, "learning_rate": 2.2959515633471896e-05, "loss": 0.5845, "step": 2597 }, { "epoch": 0.47, "learning_rate": 2.2956804626784748e-05, "loss": 1.7292, "step": 2598 }, { "epoch": 0.47, "learning_rate": 2.29540936200976e-05, "loss": 1.5098, "step": 2599 }, { "epoch": 0.47, "learning_rate": 2.2951382613410445e-05, "loss": 1.3042, "step": 2600 }, { "epoch": 0.47, "learning_rate": 2.2948671606723297e-05, "loss": 0.9294, "step": 2601 }, { "epoch": 0.47, "learning_rate": 2.2945960600036145e-05, "loss": 0.8091, "step": 2602 }, { "epoch": 0.47, "learning_rate": 2.2943249593348997e-05, "loss": 1.0391, "step": 2603 }, { "epoch": 0.47, "learning_rate": 2.294053858666185e-05, "loss": 1.5594, "step": 2604 }, { "epoch": 0.47, "learning_rate": 2.2937827579974697e-05, "loss": 1.0532, "step": 2605 }, { "epoch": 0.47, "learning_rate": 2.293511657328755e-05, "loss": 1.774, "step": 2606 }, { "epoch": 0.47, "learning_rate": 2.29324055666004e-05, "loss": 0.9467, "step": 2607 }, { "epoch": 0.47, "learning_rate": 2.2929694559913246e-05, "loss": 1.7616, "step": 2608 }, { "epoch": 0.47, "learning_rate": 2.2926983553226098e-05, "loss": 1.2328, "step": 2609 }, { "epoch": 0.47, "learning_rate": 2.292427254653895e-05, "loss": 1.2486, "step": 2610 }, { "epoch": 0.47, "learning_rate": 2.2921561539851798e-05, "loss": 0.7373, "step": 2611 }, { "epoch": 0.47, "learning_rate": 2.291885053316465e-05, "loss": 1.1546, "step": 2612 }, { "epoch": 0.47, "learning_rate": 2.2916139526477498e-05, "loss": 1.4478, "step": 2613 }, { "epoch": 0.47, "learning_rate": 2.291342851979035e-05, "loss": 0.7097, "step": 2614 }, { "epoch": 0.47, "learning_rate": 2.2910717513103202e-05, "loss": 1.1594, "step": 2615 }, { "epoch": 0.47, "learning_rate": 2.2908006506416047e-05, "loss": 1.4498, "step": 2616 }, { "epoch": 0.47, "learning_rate": 2.29052954997289e-05, "loss": 1.2365, "step": 2617 }, { "epoch": 0.47, "learning_rate": 2.290258449304175e-05, "loss": 1.2301, "step": 2618 }, { "epoch": 0.47, "learning_rate": 2.28998734863546e-05, "loss": 1.1628, "step": 2619 }, { "epoch": 0.47, "learning_rate": 2.289716247966745e-05, "loss": 1.8171, "step": 2620 }, { "epoch": 0.47, "learning_rate": 2.2894451472980303e-05, "loss": 1.7916, "step": 2621 }, { "epoch": 0.47, "learning_rate": 2.289174046629315e-05, "loss": 1.3967, "step": 2622 }, { "epoch": 0.47, "learning_rate": 2.2889029459606003e-05, "loss": 0.5457, "step": 2623 }, { "epoch": 0.47, "learning_rate": 2.2886318452918848e-05, "loss": 1.6054, "step": 2624 }, { "epoch": 0.47, "learning_rate": 2.28836074462317e-05, "loss": 1.5075, "step": 2625 }, { "epoch": 0.47, "learning_rate": 2.288089643954455e-05, "loss": 0.5689, "step": 2626 }, { "epoch": 0.47, "learning_rate": 2.28781854328574e-05, "loss": 1.1938, "step": 2627 }, { "epoch": 0.47, "learning_rate": 2.2875474426170252e-05, "loss": 0.6357, "step": 2628 }, { "epoch": 0.48, "learning_rate": 2.2872763419483104e-05, "loss": 0.9684, "step": 2629 }, { "epoch": 0.48, "learning_rate": 2.2870052412795952e-05, "loss": 0.581, "step": 2630 }, { "epoch": 0.48, "learning_rate": 2.2867341406108804e-05, "loss": 0.8185, "step": 2631 }, { "epoch": 0.48, "learning_rate": 2.2864630399421652e-05, "loss": 1.1678, "step": 2632 }, { "epoch": 0.48, "learning_rate": 2.28619193927345e-05, "loss": 1.1385, "step": 2633 }, { "epoch": 0.48, "learning_rate": 2.2859208386047353e-05, "loss": 1.7497, "step": 2634 }, { "epoch": 0.48, "learning_rate": 2.28564973793602e-05, "loss": 1.0074, "step": 2635 }, { "epoch": 0.48, "learning_rate": 2.2853786372673053e-05, "loss": 1.1658, "step": 2636 }, { "epoch": 0.48, "learning_rate": 2.2851075365985905e-05, "loss": 0.6113, "step": 2637 }, { "epoch": 0.48, "learning_rate": 2.2848364359298753e-05, "loss": 1.1769, "step": 2638 }, { "epoch": 0.48, "learning_rate": 2.2845653352611605e-05, "loss": 0.9724, "step": 2639 }, { "epoch": 0.48, "learning_rate": 2.2842942345924454e-05, "loss": 1.1804, "step": 2640 }, { "epoch": 0.48, "learning_rate": 2.2840231339237302e-05, "loss": 0.5754, "step": 2641 }, { "epoch": 0.48, "learning_rate": 2.2837520332550154e-05, "loss": 1.9807, "step": 2642 }, { "epoch": 0.48, "learning_rate": 2.2834809325863006e-05, "loss": 0.4841, "step": 2643 }, { "epoch": 0.48, "learning_rate": 2.2832098319175854e-05, "loss": 1.1221, "step": 2644 }, { "epoch": 0.48, "learning_rate": 2.2829387312488706e-05, "loss": 1.3306, "step": 2645 }, { "epoch": 0.48, "learning_rate": 2.2826676305801554e-05, "loss": 0.9189, "step": 2646 }, { "epoch": 0.48, "learning_rate": 2.2823965299114406e-05, "loss": 0.6892, "step": 2647 }, { "epoch": 0.48, "learning_rate": 2.2821254292427255e-05, "loss": 1.3046, "step": 2648 }, { "epoch": 0.48, "learning_rate": 2.2818543285740103e-05, "loss": 1.0688, "step": 2649 }, { "epoch": 0.48, "learning_rate": 2.2815832279052955e-05, "loss": 0.8203, "step": 2650 }, { "epoch": 0.48, "learning_rate": 2.2813121272365807e-05, "loss": 1.843, "step": 2651 }, { "epoch": 0.48, "learning_rate": 2.2810410265678655e-05, "loss": 1.1329, "step": 2652 }, { "epoch": 0.48, "learning_rate": 2.2807699258991507e-05, "loss": 2.1101, "step": 2653 }, { "epoch": 0.48, "learning_rate": 2.280498825230436e-05, "loss": 1.2624, "step": 2654 }, { "epoch": 0.48, "learning_rate": 2.2802277245617207e-05, "loss": 0.717, "step": 2655 }, { "epoch": 0.48, "learning_rate": 2.2799566238930056e-05, "loss": 2.0108, "step": 2656 }, { "epoch": 0.48, "learning_rate": 2.2796855232242907e-05, "loss": 0.7684, "step": 2657 }, { "epoch": 0.48, "learning_rate": 2.2794144225555756e-05, "loss": 1.6446, "step": 2658 }, { "epoch": 0.48, "learning_rate": 2.2791433218868608e-05, "loss": 1.0892, "step": 2659 }, { "epoch": 0.48, "learning_rate": 2.2788722212181456e-05, "loss": 1.5776, "step": 2660 }, { "epoch": 0.48, "learning_rate": 2.2786011205494308e-05, "loss": 0.5545, "step": 2661 }, { "epoch": 0.48, "learning_rate": 2.278330019880716e-05, "loss": 0.9295, "step": 2662 }, { "epoch": 0.48, "learning_rate": 2.2780589192120008e-05, "loss": 1.1234, "step": 2663 }, { "epoch": 0.48, "learning_rate": 2.2777878185432857e-05, "loss": 1.627, "step": 2664 }, { "epoch": 0.48, "learning_rate": 2.277516717874571e-05, "loss": 1.0762, "step": 2665 }, { "epoch": 0.48, "learning_rate": 2.2772456172058557e-05, "loss": 1.6484, "step": 2666 }, { "epoch": 0.48, "learning_rate": 2.276974516537141e-05, "loss": 1.0249, "step": 2667 }, { "epoch": 0.48, "learning_rate": 2.276703415868426e-05, "loss": 0.7531, "step": 2668 }, { "epoch": 0.48, "learning_rate": 2.276432315199711e-05, "loss": 1.3454, "step": 2669 }, { "epoch": 0.48, "learning_rate": 2.276161214530996e-05, "loss": 1.4199, "step": 2670 }, { "epoch": 0.48, "learning_rate": 2.275890113862281e-05, "loss": 0.9429, "step": 2671 }, { "epoch": 0.48, "learning_rate": 2.2756190131935658e-05, "loss": 1.3103, "step": 2672 }, { "epoch": 0.48, "learning_rate": 2.275347912524851e-05, "loss": 0.7981, "step": 2673 }, { "epoch": 0.48, "learning_rate": 2.2750768118561358e-05, "loss": 1.0778, "step": 2674 }, { "epoch": 0.48, "learning_rate": 2.274805711187421e-05, "loss": 0.7832, "step": 2675 }, { "epoch": 0.48, "learning_rate": 2.2745346105187062e-05, "loss": 1.2617, "step": 2676 }, { "epoch": 0.48, "learning_rate": 2.274263509849991e-05, "loss": 1.0401, "step": 2677 }, { "epoch": 0.48, "learning_rate": 2.2739924091812762e-05, "loss": 1.0455, "step": 2678 }, { "epoch": 0.48, "learning_rate": 2.2737213085125614e-05, "loss": 1.3311, "step": 2679 }, { "epoch": 0.48, "learning_rate": 2.273450207843846e-05, "loss": 0.9609, "step": 2680 }, { "epoch": 0.48, "learning_rate": 2.273179107175131e-05, "loss": 1.2156, "step": 2681 }, { "epoch": 0.48, "learning_rate": 2.272908006506416e-05, "loss": 1.498, "step": 2682 }, { "epoch": 0.48, "learning_rate": 2.272636905837701e-05, "loss": 1.5096, "step": 2683 }, { "epoch": 0.49, "learning_rate": 2.2723658051689863e-05, "loss": 1.4701, "step": 2684 }, { "epoch": 0.49, "learning_rate": 2.272094704500271e-05, "loss": 1.336, "step": 2685 }, { "epoch": 0.49, "learning_rate": 2.2718236038315563e-05, "loss": 0.9846, "step": 2686 }, { "epoch": 0.49, "learning_rate": 2.2715525031628415e-05, "loss": 1.1262, "step": 2687 }, { "epoch": 0.49, "learning_rate": 2.271281402494126e-05, "loss": 0.8643, "step": 2688 }, { "epoch": 0.49, "learning_rate": 2.2710103018254112e-05, "loss": 1.214, "step": 2689 }, { "epoch": 0.49, "learning_rate": 2.2707392011566964e-05, "loss": 0.8235, "step": 2690 }, { "epoch": 0.49, "learning_rate": 2.2704681004879812e-05, "loss": 1.8499, "step": 2691 }, { "epoch": 0.49, "learning_rate": 2.2701969998192664e-05, "loss": 1.1244, "step": 2692 }, { "epoch": 0.49, "learning_rate": 2.2699258991505512e-05, "loss": 2.2244, "step": 2693 }, { "epoch": 0.49, "learning_rate": 2.2696547984818364e-05, "loss": 1.3113, "step": 2694 }, { "epoch": 0.49, "learning_rate": 2.2693836978131216e-05, "loss": 1.0602, "step": 2695 }, { "epoch": 0.49, "learning_rate": 2.269112597144406e-05, "loss": 1.9117, "step": 2696 }, { "epoch": 0.49, "learning_rate": 2.2688414964756913e-05, "loss": 0.9877, "step": 2697 }, { "epoch": 0.49, "learning_rate": 2.2685703958069765e-05, "loss": 1.2872, "step": 2698 }, { "epoch": 0.49, "learning_rate": 2.2682992951382613e-05, "loss": 0.7006, "step": 2699 }, { "epoch": 0.49, "learning_rate": 2.2680281944695465e-05, "loss": 1.6356, "step": 2700 }, { "epoch": 0.49, "learning_rate": 2.2677570938008317e-05, "loss": 1.4323, "step": 2701 }, { "epoch": 0.49, "learning_rate": 2.2674859931321165e-05, "loss": 0.9458, "step": 2702 }, { "epoch": 0.49, "learning_rate": 2.2672148924634014e-05, "loss": 1.1537, "step": 2703 }, { "epoch": 0.49, "learning_rate": 2.2669437917946862e-05, "loss": 1.0647, "step": 2704 }, { "epoch": 0.49, "learning_rate": 2.2666726911259714e-05, "loss": 1.1344, "step": 2705 }, { "epoch": 0.49, "learning_rate": 2.2664015904572566e-05, "loss": 1.541, "step": 2706 }, { "epoch": 0.49, "learning_rate": 2.2661304897885414e-05, "loss": 1.5436, "step": 2707 }, { "epoch": 0.49, "learning_rate": 2.2658593891198266e-05, "loss": 1.2083, "step": 2708 }, { "epoch": 0.49, "learning_rate": 2.2655882884511118e-05, "loss": 1.0746, "step": 2709 }, { "epoch": 0.49, "learning_rate": 2.2653171877823966e-05, "loss": 0.5945, "step": 2710 }, { "epoch": 0.49, "learning_rate": 2.2650460871136815e-05, "loss": 0.9751, "step": 2711 }, { "epoch": 0.49, "learning_rate": 2.2647749864449667e-05, "loss": 1.4879, "step": 2712 }, { "epoch": 0.49, "learning_rate": 2.2645038857762515e-05, "loss": 1.4742, "step": 2713 }, { "epoch": 0.49, "learning_rate": 2.2642327851075367e-05, "loss": 1.1639, "step": 2714 }, { "epoch": 0.49, "learning_rate": 2.2639616844388215e-05, "loss": 2.2402, "step": 2715 }, { "epoch": 0.49, "learning_rate": 2.2636905837701067e-05, "loss": 1.844, "step": 2716 }, { "epoch": 0.49, "learning_rate": 2.263419483101392e-05, "loss": 1.3859, "step": 2717 }, { "epoch": 0.49, "learning_rate": 2.2631483824326767e-05, "loss": 0.8465, "step": 2718 }, { "epoch": 0.49, "learning_rate": 2.2628772817639616e-05, "loss": 1.3929, "step": 2719 }, { "epoch": 0.49, "learning_rate": 2.2626061810952468e-05, "loss": 0.6815, "step": 2720 }, { "epoch": 0.49, "learning_rate": 2.2623350804265316e-05, "loss": 0.8552, "step": 2721 }, { "epoch": 0.49, "learning_rate": 2.2620639797578168e-05, "loss": 0.7843, "step": 2722 }, { "epoch": 0.49, "learning_rate": 2.261792879089102e-05, "loss": 0.8342, "step": 2723 }, { "epoch": 0.49, "learning_rate": 2.2615217784203868e-05, "loss": 1.2083, "step": 2724 }, { "epoch": 0.49, "learning_rate": 2.261250677751672e-05, "loss": 1.3473, "step": 2725 }, { "epoch": 0.49, "learning_rate": 2.260979577082957e-05, "loss": 1.0745, "step": 2726 }, { "epoch": 0.49, "learning_rate": 2.2607084764142417e-05, "loss": 0.956, "step": 2727 }, { "epoch": 0.49, "learning_rate": 2.260437375745527e-05, "loss": 1.1845, "step": 2728 }, { "epoch": 0.49, "learning_rate": 2.2601662750768117e-05, "loss": 1.2661, "step": 2729 }, { "epoch": 0.49, "learning_rate": 2.259895174408097e-05, "loss": 0.7601, "step": 2730 }, { "epoch": 0.49, "learning_rate": 2.259624073739382e-05, "loss": 0.9305, "step": 2731 }, { "epoch": 0.49, "learning_rate": 2.259352973070667e-05, "loss": 1.1569, "step": 2732 }, { "epoch": 0.49, "learning_rate": 2.259081872401952e-05, "loss": 1.3495, "step": 2733 }, { "epoch": 0.49, "learning_rate": 2.2588107717332373e-05, "loss": 1.1909, "step": 2734 }, { "epoch": 0.49, "learning_rate": 2.2585396710645218e-05, "loss": 0.9159, "step": 2735 }, { "epoch": 0.49, "learning_rate": 2.258268570395807e-05, "loss": 0.6867, "step": 2736 }, { "epoch": 0.49, "learning_rate": 2.2579974697270918e-05, "loss": 1.101, "step": 2737 }, { "epoch": 0.49, "learning_rate": 2.257726369058377e-05, "loss": 0.947, "step": 2738 }, { "epoch": 0.5, "learning_rate": 2.2574552683896622e-05, "loss": 1.0144, "step": 2739 }, { "epoch": 0.5, "learning_rate": 2.257184167720947e-05, "loss": 0.7859, "step": 2740 }, { "epoch": 0.5, "learning_rate": 2.2569130670522322e-05, "loss": 1.5625, "step": 2741 }, { "epoch": 0.5, "learning_rate": 2.2566419663835174e-05, "loss": 0.9716, "step": 2742 }, { "epoch": 0.5, "learning_rate": 2.256370865714802e-05, "loss": 1.8594, "step": 2743 }, { "epoch": 0.5, "learning_rate": 2.256099765046087e-05, "loss": 1.1613, "step": 2744 }, { "epoch": 0.5, "learning_rate": 2.2558286643773723e-05, "loss": 1.4764, "step": 2745 }, { "epoch": 0.5, "learning_rate": 2.255557563708657e-05, "loss": 1.1649, "step": 2746 }, { "epoch": 0.5, "learning_rate": 2.2552864630399423e-05, "loss": 1.2833, "step": 2747 }, { "epoch": 0.5, "learning_rate": 2.255015362371227e-05, "loss": 0.6972, "step": 2748 }, { "epoch": 0.5, "learning_rate": 2.2547442617025123e-05, "loss": 1.0133, "step": 2749 }, { "epoch": 0.5, "learning_rate": 2.2544731610337975e-05, "loss": 1.1573, "step": 2750 }, { "epoch": 0.5, "learning_rate": 2.254202060365082e-05, "loss": 0.9214, "step": 2751 }, { "epoch": 0.5, "learning_rate": 2.2539309596963672e-05, "loss": 1.364, "step": 2752 }, { "epoch": 0.5, "learning_rate": 2.2536598590276524e-05, "loss": 0.8611, "step": 2753 }, { "epoch": 0.5, "learning_rate": 2.2533887583589372e-05, "loss": 1.9025, "step": 2754 }, { "epoch": 0.5, "learning_rate": 2.2531176576902224e-05, "loss": 1.6231, "step": 2755 }, { "epoch": 0.5, "learning_rate": 2.2528465570215076e-05, "loss": 0.5948, "step": 2756 }, { "epoch": 0.5, "learning_rate": 2.2525754563527924e-05, "loss": 1.6562, "step": 2757 }, { "epoch": 0.5, "learning_rate": 2.2523043556840776e-05, "loss": 0.9437, "step": 2758 }, { "epoch": 0.5, "learning_rate": 2.252033255015362e-05, "loss": 1.201, "step": 2759 }, { "epoch": 0.5, "learning_rate": 2.2517621543466473e-05, "loss": 1.2754, "step": 2760 }, { "epoch": 0.5, "learning_rate": 2.2514910536779325e-05, "loss": 1.1685, "step": 2761 }, { "epoch": 0.5, "learning_rate": 2.2512199530092173e-05, "loss": 1.0006, "step": 2762 }, { "epoch": 0.5, "learning_rate": 2.2509488523405025e-05, "loss": 1.07, "step": 2763 }, { "epoch": 0.5, "learning_rate": 2.2506777516717877e-05, "loss": 1.2641, "step": 2764 }, { "epoch": 0.5, "learning_rate": 2.2504066510030725e-05, "loss": 1.1196, "step": 2765 }, { "epoch": 0.5, "learning_rate": 2.2501355503343577e-05, "loss": 1.7994, "step": 2766 }, { "epoch": 0.5, "learning_rate": 2.2498644496656426e-05, "loss": 1.6871, "step": 2767 }, { "epoch": 0.5, "learning_rate": 2.2495933489969274e-05, "loss": 1.2453, "step": 2768 }, { "epoch": 0.5, "learning_rate": 2.2493222483282126e-05, "loss": 1.71, "step": 2769 }, { "epoch": 0.5, "learning_rate": 2.2490511476594974e-05, "loss": 0.7197, "step": 2770 }, { "epoch": 0.5, "learning_rate": 2.2487800469907826e-05, "loss": 1.0904, "step": 2771 }, { "epoch": 0.5, "learning_rate": 2.2485089463220678e-05, "loss": 1.1329, "step": 2772 }, { "epoch": 0.5, "learning_rate": 2.2482378456533526e-05, "loss": 0.7769, "step": 2773 }, { "epoch": 0.5, "learning_rate": 2.2479667449846378e-05, "loss": 0.9858, "step": 2774 }, { "epoch": 0.5, "learning_rate": 2.2476956443159227e-05, "loss": 0.8735, "step": 2775 }, { "epoch": 0.5, "learning_rate": 2.2474245436472075e-05, "loss": 0.9711, "step": 2776 }, { "epoch": 0.5, "learning_rate": 2.2471534429784927e-05, "loss": 1.0851, "step": 2777 }, { "epoch": 0.5, "learning_rate": 2.246882342309778e-05, "loss": 0.8491, "step": 2778 }, { "epoch": 0.5, "learning_rate": 2.2466112416410627e-05, "loss": 1.1438, "step": 2779 }, { "epoch": 0.5, "learning_rate": 2.246340140972348e-05, "loss": 0.9458, "step": 2780 }, { "epoch": 0.5, "learning_rate": 2.2460690403036328e-05, "loss": 0.9925, "step": 2781 }, { "epoch": 0.5, "learning_rate": 2.245797939634918e-05, "loss": 1.2869, "step": 2782 }, { "epoch": 0.5, "learning_rate": 2.2455268389662028e-05, "loss": 1.6554, "step": 2783 }, { "epoch": 0.5, "learning_rate": 2.2452557382974876e-05, "loss": 1.071, "step": 2784 }, { "epoch": 0.5, "learning_rate": 2.2449846376287728e-05, "loss": 1.3779, "step": 2785 }, { "epoch": 0.5, "learning_rate": 2.244713536960058e-05, "loss": 0.8727, "step": 2786 }, { "epoch": 0.5, "learning_rate": 2.244442436291343e-05, "loss": 1.0708, "step": 2787 }, { "epoch": 0.5, "learning_rate": 2.244171335622628e-05, "loss": 0.6878, "step": 2788 }, { "epoch": 0.5, "learning_rate": 2.2439002349539132e-05, "loss": 1.2369, "step": 2789 }, { "epoch": 0.5, "learning_rate": 2.243629134285198e-05, "loss": 0.8881, "step": 2790 }, { "epoch": 0.5, "learning_rate": 2.243358033616483e-05, "loss": 1.1966, "step": 2791 }, { "epoch": 0.5, "learning_rate": 2.2430869329477677e-05, "loss": 1.2627, "step": 2792 }, { "epoch": 0.5, "learning_rate": 2.242815832279053e-05, "loss": 2.1045, "step": 2793 }, { "epoch": 0.5, "learning_rate": 2.242544731610338e-05, "loss": 1.5591, "step": 2794 }, { "epoch": 0.51, "learning_rate": 2.242273630941623e-05, "loss": 1.2934, "step": 2795 }, { "epoch": 0.51, "learning_rate": 2.242002530272908e-05, "loss": 1.1859, "step": 2796 }, { "epoch": 0.51, "learning_rate": 2.2417314296041933e-05, "loss": 1.1705, "step": 2797 }, { "epoch": 0.51, "learning_rate": 2.241460328935478e-05, "loss": 1.0475, "step": 2798 }, { "epoch": 0.51, "learning_rate": 2.241189228266763e-05, "loss": 0.7045, "step": 2799 }, { "epoch": 0.51, "learning_rate": 2.2409181275980482e-05, "loss": 0.7889, "step": 2800 }, { "epoch": 0.51, "learning_rate": 2.240647026929333e-05, "loss": 0.8758, "step": 2801 }, { "epoch": 0.51, "learning_rate": 2.2403759262606182e-05, "loss": 0.523, "step": 2802 }, { "epoch": 0.51, "learning_rate": 2.240104825591903e-05, "loss": 1.5045, "step": 2803 }, { "epoch": 0.51, "learning_rate": 2.2398337249231882e-05, "loss": 0.8713, "step": 2804 }, { "epoch": 0.51, "learning_rate": 2.2395626242544734e-05, "loss": 0.9667, "step": 2805 }, { "epoch": 0.51, "learning_rate": 2.2392915235857583e-05, "loss": 1.2602, "step": 2806 }, { "epoch": 0.51, "learning_rate": 2.239020422917043e-05, "loss": 1.1728, "step": 2807 }, { "epoch": 0.51, "learning_rate": 2.2387493222483283e-05, "loss": 1.3302, "step": 2808 }, { "epoch": 0.51, "learning_rate": 2.238478221579613e-05, "loss": 0.5874, "step": 2809 }, { "epoch": 0.51, "learning_rate": 2.2382071209108983e-05, "loss": 1.0306, "step": 2810 }, { "epoch": 0.51, "learning_rate": 2.2379360202421835e-05, "loss": 1.5271, "step": 2811 }, { "epoch": 0.51, "learning_rate": 2.2376649195734683e-05, "loss": 0.9787, "step": 2812 }, { "epoch": 0.51, "learning_rate": 2.2373938189047535e-05, "loss": 0.777, "step": 2813 }, { "epoch": 0.51, "learning_rate": 2.2371227182360384e-05, "loss": 1.1083, "step": 2814 }, { "epoch": 0.51, "learning_rate": 2.2368516175673232e-05, "loss": 0.8653, "step": 2815 }, { "epoch": 0.51, "learning_rate": 2.2365805168986084e-05, "loss": 1.397, "step": 2816 }, { "epoch": 0.51, "learning_rate": 2.2363094162298932e-05, "loss": 1.0673, "step": 2817 }, { "epoch": 0.51, "learning_rate": 2.2360383155611784e-05, "loss": 1.037, "step": 2818 }, { "epoch": 0.51, "learning_rate": 2.2357672148924636e-05, "loss": 1.0714, "step": 2819 }, { "epoch": 0.51, "learning_rate": 2.2354961142237484e-05, "loss": 1.2925, "step": 2820 }, { "epoch": 0.51, "learning_rate": 2.2352250135550336e-05, "loss": 1.4657, "step": 2821 }, { "epoch": 0.51, "learning_rate": 2.2349539128863188e-05, "loss": 1.3895, "step": 2822 }, { "epoch": 0.51, "learning_rate": 2.2346828122176033e-05, "loss": 1.5961, "step": 2823 }, { "epoch": 0.51, "learning_rate": 2.2344117115488885e-05, "loss": 0.9422, "step": 2824 }, { "epoch": 0.51, "learning_rate": 2.2341406108801737e-05, "loss": 1.045, "step": 2825 }, { "epoch": 0.51, "learning_rate": 2.2338695102114585e-05, "loss": 1.0587, "step": 2826 }, { "epoch": 0.51, "learning_rate": 2.2335984095427437e-05, "loss": 1.314, "step": 2827 }, { "epoch": 0.51, "learning_rate": 2.2333273088740286e-05, "loss": 0.8567, "step": 2828 }, { "epoch": 0.51, "learning_rate": 2.2330562082053137e-05, "loss": 1.0555, "step": 2829 }, { "epoch": 0.51, "learning_rate": 2.232785107536599e-05, "loss": 1.2103, "step": 2830 }, { "epoch": 0.51, "learning_rate": 2.2325140068678834e-05, "loss": 0.8395, "step": 2831 }, { "epoch": 0.51, "learning_rate": 2.2322429061991686e-05, "loss": 0.7501, "step": 2832 }, { "epoch": 0.51, "learning_rate": 2.2319718055304538e-05, "loss": 1.7479, "step": 2833 }, { "epoch": 0.51, "learning_rate": 2.2317007048617386e-05, "loss": 0.8401, "step": 2834 }, { "epoch": 0.51, "learning_rate": 2.2314296041930238e-05, "loss": 0.4739, "step": 2835 }, { "epoch": 0.51, "learning_rate": 2.231158503524309e-05, "loss": 0.5916, "step": 2836 }, { "epoch": 0.51, "learning_rate": 2.230887402855594e-05, "loss": 1.0268, "step": 2837 }, { "epoch": 0.51, "learning_rate": 2.2306163021868787e-05, "loss": 1.0316, "step": 2838 }, { "epoch": 0.51, "learning_rate": 2.2303452015181635e-05, "loss": 2.02, "step": 2839 }, { "epoch": 0.51, "learning_rate": 2.2300741008494487e-05, "loss": 1.0642, "step": 2840 }, { "epoch": 0.51, "learning_rate": 2.229803000180734e-05, "loss": 1.0758, "step": 2841 }, { "epoch": 0.51, "learning_rate": 2.2295318995120187e-05, "loss": 1.1164, "step": 2842 }, { "epoch": 0.51, "learning_rate": 2.229260798843304e-05, "loss": 0.3823, "step": 2843 }, { "epoch": 0.51, "learning_rate": 2.228989698174589e-05, "loss": 1.7754, "step": 2844 }, { "epoch": 0.51, "learning_rate": 2.228718597505874e-05, "loss": 1.3162, "step": 2845 }, { "epoch": 0.51, "learning_rate": 2.2284474968371588e-05, "loss": 0.8049, "step": 2846 }, { "epoch": 0.51, "learning_rate": 2.228176396168444e-05, "loss": 1.0828, "step": 2847 }, { "epoch": 0.51, "learning_rate": 2.2279052954997288e-05, "loss": 1.1998, "step": 2848 }, { "epoch": 0.51, "learning_rate": 2.227634194831014e-05, "loss": 1.3491, "step": 2849 }, { "epoch": 0.52, "learning_rate": 2.227363094162299e-05, "loss": 0.7326, "step": 2850 }, { "epoch": 0.52, "learning_rate": 2.227091993493584e-05, "loss": 1.0348, "step": 2851 }, { "epoch": 0.52, "learning_rate": 2.2268208928248692e-05, "loss": 1.0651, "step": 2852 }, { "epoch": 0.52, "learning_rate": 2.226549792156154e-05, "loss": 1.5366, "step": 2853 }, { "epoch": 0.52, "learning_rate": 2.226278691487439e-05, "loss": 1.2118, "step": 2854 }, { "epoch": 0.52, "learning_rate": 2.226007590818724e-05, "loss": 1.2921, "step": 2855 }, { "epoch": 0.52, "learning_rate": 2.225736490150009e-05, "loss": 1.0864, "step": 2856 }, { "epoch": 0.52, "learning_rate": 2.225465389481294e-05, "loss": 1.2245, "step": 2857 }, { "epoch": 0.52, "learning_rate": 2.2251942888125793e-05, "loss": 1.4224, "step": 2858 }, { "epoch": 0.52, "learning_rate": 2.224923188143864e-05, "loss": 0.9809, "step": 2859 }, { "epoch": 0.52, "learning_rate": 2.2246520874751493e-05, "loss": 1.9378, "step": 2860 }, { "epoch": 0.52, "learning_rate": 2.224380986806434e-05, "loss": 1.1902, "step": 2861 }, { "epoch": 0.52, "learning_rate": 2.224109886137719e-05, "loss": 1.2968, "step": 2862 }, { "epoch": 0.52, "learning_rate": 2.2238387854690042e-05, "loss": 0.9618, "step": 2863 }, { "epoch": 0.52, "learning_rate": 2.223567684800289e-05, "loss": 1.0969, "step": 2864 }, { "epoch": 0.52, "learning_rate": 2.2232965841315742e-05, "loss": 1.1247, "step": 2865 }, { "epoch": 0.52, "learning_rate": 2.2230254834628594e-05, "loss": 0.7749, "step": 2866 }, { "epoch": 0.52, "learning_rate": 2.2227543827941442e-05, "loss": 1.2687, "step": 2867 }, { "epoch": 0.52, "learning_rate": 2.2224832821254294e-05, "loss": 0.8778, "step": 2868 }, { "epoch": 0.52, "learning_rate": 2.2222121814567146e-05, "loss": 1.0632, "step": 2869 }, { "epoch": 0.52, "learning_rate": 2.221941080787999e-05, "loss": 0.7937, "step": 2870 }, { "epoch": 0.52, "learning_rate": 2.2216699801192843e-05, "loss": 1.2775, "step": 2871 }, { "epoch": 0.52, "learning_rate": 2.221398879450569e-05, "loss": 1.7353, "step": 2872 }, { "epoch": 0.52, "learning_rate": 2.2211277787818543e-05, "loss": 1.0878, "step": 2873 }, { "epoch": 0.52, "learning_rate": 2.2208566781131395e-05, "loss": 1.258, "step": 2874 }, { "epoch": 0.52, "learning_rate": 2.2205855774444244e-05, "loss": 1.5433, "step": 2875 }, { "epoch": 0.52, "learning_rate": 2.2203144767757095e-05, "loss": 0.7939, "step": 2876 }, { "epoch": 0.52, "learning_rate": 2.2200433761069947e-05, "loss": 0.6046, "step": 2877 }, { "epoch": 0.52, "learning_rate": 2.2197722754382792e-05, "loss": 1.1023, "step": 2878 }, { "epoch": 0.52, "learning_rate": 2.2195011747695644e-05, "loss": 1.32, "step": 2879 }, { "epoch": 0.52, "learning_rate": 2.2192300741008496e-05, "loss": 1.2482, "step": 2880 }, { "epoch": 0.52, "learning_rate": 2.2189589734321344e-05, "loss": 0.5129, "step": 2881 }, { "epoch": 0.52, "learning_rate": 2.2186878727634196e-05, "loss": 1.5711, "step": 2882 }, { "epoch": 0.52, "learning_rate": 2.2184167720947045e-05, "loss": 1.4724, "step": 2883 }, { "epoch": 0.52, "learning_rate": 2.2181456714259896e-05, "loss": 0.8942, "step": 2884 }, { "epoch": 0.52, "learning_rate": 2.2178745707572748e-05, "loss": 1.5246, "step": 2885 }, { "epoch": 0.52, "learning_rate": 2.2176034700885593e-05, "loss": 0.7355, "step": 2886 }, { "epoch": 0.52, "learning_rate": 2.2173323694198445e-05, "loss": 0.5699, "step": 2887 }, { "epoch": 0.52, "learning_rate": 2.2170612687511297e-05, "loss": 0.9849, "step": 2888 }, { "epoch": 0.52, "learning_rate": 2.2167901680824145e-05, "loss": 1.6791, "step": 2889 }, { "epoch": 0.52, "learning_rate": 2.2165190674136997e-05, "loss": 0.8512, "step": 2890 }, { "epoch": 0.52, "learning_rate": 2.216247966744985e-05, "loss": 0.4472, "step": 2891 }, { "epoch": 0.52, "learning_rate": 2.2159768660762698e-05, "loss": 1.3965, "step": 2892 }, { "epoch": 0.52, "learning_rate": 2.215705765407555e-05, "loss": 1.684, "step": 2893 }, { "epoch": 0.52, "learning_rate": 2.2154346647388394e-05, "loss": 0.6206, "step": 2894 }, { "epoch": 0.52, "learning_rate": 2.2151635640701246e-05, "loss": 0.9334, "step": 2895 }, { "epoch": 0.52, "learning_rate": 2.2148924634014098e-05, "loss": 0.8117, "step": 2896 }, { "epoch": 0.52, "learning_rate": 2.2146213627326946e-05, "loss": 1.0192, "step": 2897 }, { "epoch": 0.52, "learning_rate": 2.21435026206398e-05, "loss": 1.2857, "step": 2898 }, { "epoch": 0.52, "learning_rate": 2.214079161395265e-05, "loss": 0.7117, "step": 2899 }, { "epoch": 0.52, "learning_rate": 2.21380806072655e-05, "loss": 0.8905, "step": 2900 }, { "epoch": 0.52, "learning_rate": 2.213536960057835e-05, "loss": 0.5292, "step": 2901 }, { "epoch": 0.52, "learning_rate": 2.21326585938912e-05, "loss": 1.1926, "step": 2902 }, { "epoch": 0.52, "learning_rate": 2.2129947587204047e-05, "loss": 1.3373, "step": 2903 }, { "epoch": 0.52, "learning_rate": 2.21272365805169e-05, "loss": 1.8868, "step": 2904 }, { "epoch": 0.53, "learning_rate": 2.2124525573829748e-05, "loss": 0.9239, "step": 2905 }, { "epoch": 0.53, "learning_rate": 2.21218145671426e-05, "loss": 1.1332, "step": 2906 }, { "epoch": 0.53, "learning_rate": 2.211910356045545e-05, "loss": 0.9563, "step": 2907 }, { "epoch": 0.53, "learning_rate": 2.21163925537683e-05, "loss": 1.2724, "step": 2908 }, { "epoch": 0.53, "learning_rate": 2.211368154708115e-05, "loss": 2.0656, "step": 2909 }, { "epoch": 0.53, "learning_rate": 2.2110970540394e-05, "loss": 1.0206, "step": 2910 }, { "epoch": 0.53, "learning_rate": 2.210825953370685e-05, "loss": 0.49, "step": 2911 }, { "epoch": 0.53, "learning_rate": 2.21055485270197e-05, "loss": 0.7558, "step": 2912 }, { "epoch": 0.53, "learning_rate": 2.2102837520332552e-05, "loss": 0.8732, "step": 2913 }, { "epoch": 0.53, "learning_rate": 2.21001265136454e-05, "loss": 0.9655, "step": 2914 }, { "epoch": 0.53, "learning_rate": 2.2097415506958252e-05, "loss": 1.4663, "step": 2915 }, { "epoch": 0.53, "learning_rate": 2.20947045002711e-05, "loss": 0.8722, "step": 2916 }, { "epoch": 0.53, "learning_rate": 2.2091993493583953e-05, "loss": 0.942, "step": 2917 }, { "epoch": 0.53, "learning_rate": 2.20892824868968e-05, "loss": 1.9171, "step": 2918 }, { "epoch": 0.53, "learning_rate": 2.208657148020965e-05, "loss": 1.7415, "step": 2919 }, { "epoch": 0.53, "learning_rate": 2.20838604735225e-05, "loss": 0.9103, "step": 2920 }, { "epoch": 0.53, "learning_rate": 2.2081149466835353e-05, "loss": 1.3583, "step": 2921 }, { "epoch": 0.53, "learning_rate": 2.20784384601482e-05, "loss": 1.2745, "step": 2922 }, { "epoch": 0.53, "learning_rate": 2.2075727453461053e-05, "loss": 0.921, "step": 2923 }, { "epoch": 0.53, "learning_rate": 2.2073016446773905e-05, "loss": 2.0519, "step": 2924 }, { "epoch": 0.53, "learning_rate": 2.2070305440086754e-05, "loss": 1.2331, "step": 2925 }, { "epoch": 0.53, "learning_rate": 2.2067594433399602e-05, "loss": 0.8287, "step": 2926 }, { "epoch": 0.53, "learning_rate": 2.206488342671245e-05, "loss": 2.0217, "step": 2927 }, { "epoch": 0.53, "learning_rate": 2.2062172420025302e-05, "loss": 1.1362, "step": 2928 }, { "epoch": 0.53, "learning_rate": 2.2059461413338154e-05, "loss": 1.2753, "step": 2929 }, { "epoch": 0.53, "learning_rate": 2.2056750406651003e-05, "loss": 0.706, "step": 2930 }, { "epoch": 0.53, "learning_rate": 2.2054039399963854e-05, "loss": 1.299, "step": 2931 }, { "epoch": 0.53, "learning_rate": 2.2051328393276706e-05, "loss": 1.2175, "step": 2932 }, { "epoch": 0.53, "learning_rate": 2.2048617386589555e-05, "loss": 0.7148, "step": 2933 }, { "epoch": 0.53, "learning_rate": 2.2045906379902403e-05, "loss": 0.7652, "step": 2934 }, { "epoch": 0.53, "learning_rate": 2.2043195373215255e-05, "loss": 1.0376, "step": 2935 }, { "epoch": 0.53, "learning_rate": 2.2040484366528103e-05, "loss": 1.1558, "step": 2936 }, { "epoch": 0.53, "learning_rate": 2.2037773359840955e-05, "loss": 1.3665, "step": 2937 }, { "epoch": 0.53, "learning_rate": 2.2035062353153804e-05, "loss": 0.8624, "step": 2938 }, { "epoch": 0.53, "learning_rate": 2.2032351346466656e-05, "loss": 0.8083, "step": 2939 }, { "epoch": 0.53, "learning_rate": 2.2029640339779507e-05, "loss": 1.4162, "step": 2940 }, { "epoch": 0.53, "learning_rate": 2.2026929333092356e-05, "loss": 1.4274, "step": 2941 }, { "epoch": 0.53, "learning_rate": 2.2024218326405204e-05, "loss": 0.7836, "step": 2942 }, { "epoch": 0.53, "learning_rate": 2.2021507319718056e-05, "loss": 0.5605, "step": 2943 }, { "epoch": 0.53, "learning_rate": 2.2018796313030904e-05, "loss": 1.0585, "step": 2944 }, { "epoch": 0.53, "learning_rate": 2.2016085306343756e-05, "loss": 1.1949, "step": 2945 }, { "epoch": 0.53, "learning_rate": 2.2013374299656608e-05, "loss": 0.6535, "step": 2946 }, { "epoch": 0.53, "learning_rate": 2.2010663292969457e-05, "loss": 0.6094, "step": 2947 }, { "epoch": 0.53, "learning_rate": 2.200795228628231e-05, "loss": 1.0031, "step": 2948 }, { "epoch": 0.53, "learning_rate": 2.2005241279595157e-05, "loss": 1.9643, "step": 2949 }, { "epoch": 0.53, "learning_rate": 2.2002530272908005e-05, "loss": 1.1095, "step": 2950 }, { "epoch": 0.53, "learning_rate": 2.1999819266220857e-05, "loss": 0.7009, "step": 2951 }, { "epoch": 0.53, "learning_rate": 2.1997108259533706e-05, "loss": 1.3671, "step": 2952 }, { "epoch": 0.53, "learning_rate": 2.1994397252846557e-05, "loss": 0.7613, "step": 2953 }, { "epoch": 0.53, "learning_rate": 2.199168624615941e-05, "loss": 1.1691, "step": 2954 }, { "epoch": 0.53, "learning_rate": 2.1988975239472258e-05, "loss": 1.2319, "step": 2955 }, { "epoch": 0.53, "learning_rate": 2.198626423278511e-05, "loss": 0.9527, "step": 2956 }, { "epoch": 0.53, "learning_rate": 2.198355322609796e-05, "loss": 1.3602, "step": 2957 }, { "epoch": 0.53, "learning_rate": 2.1980842219410806e-05, "loss": 1.7027, "step": 2958 }, { "epoch": 0.53, "learning_rate": 2.1978131212723658e-05, "loss": 1.0628, "step": 2959 }, { "epoch": 0.53, "learning_rate": 2.197542020603651e-05, "loss": 1.2119, "step": 2960 }, { "epoch": 0.54, "learning_rate": 2.197270919934936e-05, "loss": 1.4021, "step": 2961 }, { "epoch": 0.54, "learning_rate": 2.196999819266221e-05, "loss": 0.792, "step": 2962 }, { "epoch": 0.54, "learning_rate": 2.196728718597506e-05, "loss": 1.1217, "step": 2963 }, { "epoch": 0.54, "learning_rate": 2.196457617928791e-05, "loss": 1.0014, "step": 2964 }, { "epoch": 0.54, "learning_rate": 2.196186517260076e-05, "loss": 1.2547, "step": 2965 }, { "epoch": 0.54, "learning_rate": 2.1959154165913607e-05, "loss": 0.8004, "step": 2966 }, { "epoch": 0.54, "learning_rate": 2.195644315922646e-05, "loss": 0.8958, "step": 2967 }, { "epoch": 0.54, "learning_rate": 2.195373215253931e-05, "loss": 1.0376, "step": 2968 }, { "epoch": 0.54, "learning_rate": 2.195102114585216e-05, "loss": 1.6509, "step": 2969 }, { "epoch": 0.54, "learning_rate": 2.194831013916501e-05, "loss": 0.5054, "step": 2970 }, { "epoch": 0.54, "learning_rate": 2.1945599132477863e-05, "loss": 1.2597, "step": 2971 }, { "epoch": 0.54, "learning_rate": 2.194288812579071e-05, "loss": 0.9387, "step": 2972 }, { "epoch": 0.54, "learning_rate": 2.194017711910356e-05, "loss": 1.0279, "step": 2973 }, { "epoch": 0.54, "learning_rate": 2.193746611241641e-05, "loss": 1.2189, "step": 2974 }, { "epoch": 0.54, "learning_rate": 2.193475510572926e-05, "loss": 1.4721, "step": 2975 }, { "epoch": 0.54, "learning_rate": 2.1932044099042112e-05, "loss": 1.2638, "step": 2976 }, { "epoch": 0.54, "learning_rate": 2.192933309235496e-05, "loss": 1.0106, "step": 2977 }, { "epoch": 0.54, "learning_rate": 2.1926622085667812e-05, "loss": 1.5666, "step": 2978 }, { "epoch": 0.54, "learning_rate": 2.1923911078980664e-05, "loss": 0.709, "step": 2979 }, { "epoch": 0.54, "learning_rate": 2.1921200072293513e-05, "loss": 1.0123, "step": 2980 }, { "epoch": 0.54, "learning_rate": 2.191848906560636e-05, "loss": 1.4001, "step": 2981 }, { "epoch": 0.54, "learning_rate": 2.1915778058919213e-05, "loss": 1.5693, "step": 2982 }, { "epoch": 0.54, "learning_rate": 2.191306705223206e-05, "loss": 1.5753, "step": 2983 }, { "epoch": 0.54, "learning_rate": 2.1910356045544913e-05, "loss": 2.1441, "step": 2984 }, { "epoch": 0.54, "learning_rate": 2.190764503885776e-05, "loss": 1.3234, "step": 2985 }, { "epoch": 0.54, "learning_rate": 2.1904934032170614e-05, "loss": 1.3043, "step": 2986 }, { "epoch": 0.54, "learning_rate": 2.1902223025483465e-05, "loss": 0.7331, "step": 2987 }, { "epoch": 0.54, "learning_rate": 2.1899512018796314e-05, "loss": 1.3138, "step": 2988 }, { "epoch": 0.54, "learning_rate": 2.1896801012109162e-05, "loss": 0.6274, "step": 2989 }, { "epoch": 0.54, "learning_rate": 2.1894090005422014e-05, "loss": 0.716, "step": 2990 }, { "epoch": 0.54, "learning_rate": 2.1891378998734862e-05, "loss": 0.913, "step": 2991 }, { "epoch": 0.54, "learning_rate": 2.1888667992047714e-05, "loss": 0.8352, "step": 2992 }, { "epoch": 0.54, "learning_rate": 2.1885956985360566e-05, "loss": 0.7715, "step": 2993 }, { "epoch": 0.54, "learning_rate": 2.1883245978673415e-05, "loss": 1.582, "step": 2994 }, { "epoch": 0.54, "learning_rate": 2.1880534971986266e-05, "loss": 1.2387, "step": 2995 }, { "epoch": 0.54, "learning_rate": 2.1877823965299115e-05, "loss": 1.5785, "step": 2996 }, { "epoch": 0.54, "learning_rate": 2.1875112958611963e-05, "loss": 0.7255, "step": 2997 }, { "epoch": 0.54, "learning_rate": 2.1872401951924815e-05, "loss": 1.8531, "step": 2998 }, { "epoch": 0.54, "learning_rate": 2.1869690945237664e-05, "loss": 0.9536, "step": 2999 }, { "epoch": 0.54, "learning_rate": 2.1866979938550515e-05, "loss": 1.9245, "step": 3000 }, { "epoch": 0.54, "learning_rate": 2.1864268931863367e-05, "loss": 1.7439, "step": 3001 }, { "epoch": 0.54, "learning_rate": 2.1861557925176216e-05, "loss": 0.9687, "step": 3002 }, { "epoch": 0.54, "learning_rate": 2.1858846918489067e-05, "loss": 0.6568, "step": 3003 }, { "epoch": 0.54, "learning_rate": 2.185613591180192e-05, "loss": 1.4576, "step": 3004 }, { "epoch": 0.54, "learning_rate": 2.1853424905114764e-05, "loss": 1.1933, "step": 3005 }, { "epoch": 0.54, "learning_rate": 2.1850713898427616e-05, "loss": 1.1139, "step": 3006 }, { "epoch": 0.54, "learning_rate": 2.1848002891740465e-05, "loss": 1.6975, "step": 3007 }, { "epoch": 0.54, "learning_rate": 2.1845291885053316e-05, "loss": 0.8295, "step": 3008 }, { "epoch": 0.54, "learning_rate": 2.1842580878366168e-05, "loss": 1.2251, "step": 3009 }, { "epoch": 0.54, "learning_rate": 2.1839869871679017e-05, "loss": 1.1892, "step": 3010 }, { "epoch": 0.54, "learning_rate": 2.183715886499187e-05, "loss": 1.6769, "step": 3011 }, { "epoch": 0.54, "learning_rate": 2.183444785830472e-05, "loss": 1.3387, "step": 3012 }, { "epoch": 0.54, "learning_rate": 2.1831736851617565e-05, "loss": 1.1882, "step": 3013 }, { "epoch": 0.54, "learning_rate": 2.1829025844930417e-05, "loss": 0.8529, "step": 3014 }, { "epoch": 0.54, "learning_rate": 2.182631483824327e-05, "loss": 0.4764, "step": 3015 }, { "epoch": 0.55, "learning_rate": 2.1823603831556118e-05, "loss": 1.1067, "step": 3016 }, { "epoch": 0.55, "learning_rate": 2.182089282486897e-05, "loss": 0.8623, "step": 3017 }, { "epoch": 0.55, "learning_rate": 2.1818181818181818e-05, "loss": 1.6177, "step": 3018 }, { "epoch": 0.55, "learning_rate": 2.181547081149467e-05, "loss": 1.4498, "step": 3019 }, { "epoch": 0.55, "learning_rate": 2.181275980480752e-05, "loss": 1.397, "step": 3020 }, { "epoch": 0.55, "learning_rate": 2.1810048798120367e-05, "loss": 1.1543, "step": 3021 }, { "epoch": 0.55, "learning_rate": 2.180733779143322e-05, "loss": 0.7869, "step": 3022 }, { "epoch": 0.55, "learning_rate": 2.180462678474607e-05, "loss": 0.9198, "step": 3023 }, { "epoch": 0.55, "learning_rate": 2.180191577805892e-05, "loss": 1.2333, "step": 3024 }, { "epoch": 0.55, "learning_rate": 2.179920477137177e-05, "loss": 1.1389, "step": 3025 }, { "epoch": 0.55, "learning_rate": 2.1796493764684622e-05, "loss": 0.5293, "step": 3026 }, { "epoch": 0.55, "learning_rate": 2.179378275799747e-05, "loss": 1.0109, "step": 3027 }, { "epoch": 0.55, "learning_rate": 2.1791071751310323e-05, "loss": 0.816, "step": 3028 }, { "epoch": 0.55, "learning_rate": 2.1788360744623168e-05, "loss": 1.187, "step": 3029 }, { "epoch": 0.55, "learning_rate": 2.178564973793602e-05, "loss": 1.4009, "step": 3030 }, { "epoch": 0.55, "learning_rate": 2.178293873124887e-05, "loss": 0.751, "step": 3031 }, { "epoch": 0.55, "learning_rate": 2.178022772456172e-05, "loss": 0.989, "step": 3032 }, { "epoch": 0.55, "learning_rate": 2.177751671787457e-05, "loss": 1.5859, "step": 3033 }, { "epoch": 0.55, "learning_rate": 2.1774805711187423e-05, "loss": 1.178, "step": 3034 }, { "epoch": 0.55, "learning_rate": 2.1772094704500272e-05, "loss": 0.8491, "step": 3035 }, { "epoch": 0.55, "learning_rate": 2.1769383697813124e-05, "loss": 1.2169, "step": 3036 }, { "epoch": 0.55, "learning_rate": 2.1766672691125972e-05, "loss": 1.0945, "step": 3037 }, { "epoch": 0.55, "learning_rate": 2.176396168443882e-05, "loss": 1.6336, "step": 3038 }, { "epoch": 0.55, "learning_rate": 2.1761250677751672e-05, "loss": 1.2394, "step": 3039 }, { "epoch": 0.55, "learning_rate": 2.175853967106452e-05, "loss": 0.6624, "step": 3040 }, { "epoch": 0.55, "learning_rate": 2.1755828664377373e-05, "loss": 0.9287, "step": 3041 }, { "epoch": 0.55, "learning_rate": 2.1753117657690224e-05, "loss": 1.3069, "step": 3042 }, { "epoch": 0.55, "learning_rate": 2.1750406651003073e-05, "loss": 0.8948, "step": 3043 }, { "epoch": 0.55, "learning_rate": 2.1747695644315925e-05, "loss": 1.1573, "step": 3044 }, { "epoch": 0.55, "learning_rate": 2.1744984637628773e-05, "loss": 1.1402, "step": 3045 }, { "epoch": 0.55, "learning_rate": 2.174227363094162e-05, "loss": 0.7158, "step": 3046 }, { "epoch": 0.55, "learning_rate": 2.1739562624254473e-05, "loss": 0.7714, "step": 3047 }, { "epoch": 0.55, "learning_rate": 2.1736851617567325e-05, "loss": 1.2003, "step": 3048 }, { "epoch": 0.55, "learning_rate": 2.1734140610880174e-05, "loss": 0.8763, "step": 3049 }, { "epoch": 0.55, "learning_rate": 2.1731429604193025e-05, "loss": 0.6484, "step": 3050 }, { "epoch": 0.55, "learning_rate": 2.1728718597505874e-05, "loss": 0.4154, "step": 3051 }, { "epoch": 0.55, "learning_rate": 2.1726007590818726e-05, "loss": 1.2448, "step": 3052 }, { "epoch": 0.55, "learning_rate": 2.1723296584131574e-05, "loss": 0.9251, "step": 3053 }, { "epoch": 0.55, "learning_rate": 2.1720585577444423e-05, "loss": 1.5072, "step": 3054 }, { "epoch": 0.55, "learning_rate": 2.1717874570757274e-05, "loss": 1.1441, "step": 3055 }, { "epoch": 0.55, "learning_rate": 2.1715163564070126e-05, "loss": 1.2719, "step": 3056 }, { "epoch": 0.55, "learning_rate": 2.1712452557382975e-05, "loss": 0.9455, "step": 3057 }, { "epoch": 0.55, "learning_rate": 2.1709741550695827e-05, "loss": 1.2311, "step": 3058 }, { "epoch": 0.55, "learning_rate": 2.170703054400868e-05, "loss": 1.3236, "step": 3059 }, { "epoch": 0.55, "learning_rate": 2.1704319537321527e-05, "loss": 0.9978, "step": 3060 }, { "epoch": 0.55, "learning_rate": 2.1701608530634375e-05, "loss": 1.0554, "step": 3061 }, { "epoch": 0.55, "learning_rate": 2.1698897523947224e-05, "loss": 0.5062, "step": 3062 }, { "epoch": 0.55, "learning_rate": 2.1696186517260076e-05, "loss": 0.8602, "step": 3063 }, { "epoch": 0.55, "learning_rate": 2.1693475510572927e-05, "loss": 0.7589, "step": 3064 }, { "epoch": 0.55, "learning_rate": 2.1690764503885776e-05, "loss": 0.8269, "step": 3065 }, { "epoch": 0.55, "learning_rate": 2.1688053497198628e-05, "loss": 1.209, "step": 3066 }, { "epoch": 0.55, "learning_rate": 2.168534249051148e-05, "loss": 1.2394, "step": 3067 }, { "epoch": 0.55, "learning_rate": 2.1682631483824328e-05, "loss": 1.3134, "step": 3068 }, { "epoch": 0.55, "learning_rate": 2.1679920477137176e-05, "loss": 1.0344, "step": 3069 }, { "epoch": 0.55, "learning_rate": 2.1677209470450028e-05, "loss": 0.98, "step": 3070 }, { "epoch": 0.56, "learning_rate": 2.1674498463762877e-05, "loss": 1.1698, "step": 3071 }, { "epoch": 0.56, "learning_rate": 2.167178745707573e-05, "loss": 1.7238, "step": 3072 }, { "epoch": 0.56, "learning_rate": 2.1669076450388577e-05, "loss": 1.0837, "step": 3073 }, { "epoch": 0.56, "learning_rate": 2.166636544370143e-05, "loss": 0.8278, "step": 3074 }, { "epoch": 0.56, "learning_rate": 2.166365443701428e-05, "loss": 0.6957, "step": 3075 }, { "epoch": 0.56, "learning_rate": 2.166094343032713e-05, "loss": 0.4411, "step": 3076 }, { "epoch": 0.56, "learning_rate": 2.1658232423639977e-05, "loss": 1.1785, "step": 3077 }, { "epoch": 0.56, "learning_rate": 2.165552141695283e-05, "loss": 0.9394, "step": 3078 }, { "epoch": 0.56, "learning_rate": 2.1652810410265678e-05, "loss": 0.7096, "step": 3079 }, { "epoch": 0.56, "learning_rate": 2.165009940357853e-05, "loss": 0.7523, "step": 3080 }, { "epoch": 0.56, "learning_rate": 2.164738839689138e-05, "loss": 1.1642, "step": 3081 }, { "epoch": 0.56, "learning_rate": 2.164467739020423e-05, "loss": 0.7022, "step": 3082 }, { "epoch": 0.56, "learning_rate": 2.164196638351708e-05, "loss": 1.7923, "step": 3083 }, { "epoch": 0.56, "learning_rate": 2.163925537682993e-05, "loss": 0.9748, "step": 3084 }, { "epoch": 0.56, "learning_rate": 2.163654437014278e-05, "loss": 0.7188, "step": 3085 }, { "epoch": 0.56, "learning_rate": 2.163383336345563e-05, "loss": 0.6758, "step": 3086 }, { "epoch": 0.56, "learning_rate": 2.163112235676848e-05, "loss": 0.7185, "step": 3087 }, { "epoch": 0.56, "learning_rate": 2.162841135008133e-05, "loss": 1.7332, "step": 3088 }, { "epoch": 0.56, "learning_rate": 2.1625700343394182e-05, "loss": 1.3405, "step": 3089 }, { "epoch": 0.56, "learning_rate": 2.162298933670703e-05, "loss": 0.3282, "step": 3090 }, { "epoch": 0.56, "learning_rate": 2.1620278330019883e-05, "loss": 0.9142, "step": 3091 }, { "epoch": 0.56, "learning_rate": 2.161756732333273e-05, "loss": 1.4252, "step": 3092 }, { "epoch": 0.56, "learning_rate": 2.161485631664558e-05, "loss": 1.172, "step": 3093 }, { "epoch": 0.56, "learning_rate": 2.161214530995843e-05, "loss": 1.0124, "step": 3094 }, { "epoch": 0.56, "learning_rate": 2.1609434303271283e-05, "loss": 0.8948, "step": 3095 }, { "epoch": 0.56, "learning_rate": 2.160672329658413e-05, "loss": 1.6762, "step": 3096 }, { "epoch": 0.56, "learning_rate": 2.1604012289896984e-05, "loss": 0.8871, "step": 3097 }, { "epoch": 0.56, "learning_rate": 2.1601301283209832e-05, "loss": 1.4385, "step": 3098 }, { "epoch": 0.56, "learning_rate": 2.1598590276522684e-05, "loss": 1.0284, "step": 3099 }, { "epoch": 0.56, "learning_rate": 2.1595879269835532e-05, "loss": 1.1597, "step": 3100 }, { "epoch": 0.56, "learning_rate": 2.159316826314838e-05, "loss": 1.0411, "step": 3101 }, { "epoch": 0.56, "learning_rate": 2.1590457256461232e-05, "loss": 1.0185, "step": 3102 }, { "epoch": 0.56, "learning_rate": 2.1587746249774084e-05, "loss": 0.9167, "step": 3103 }, { "epoch": 0.56, "learning_rate": 2.1585035243086933e-05, "loss": 0.6924, "step": 3104 }, { "epoch": 0.56, "learning_rate": 2.1582324236399785e-05, "loss": 1.1227, "step": 3105 }, { "epoch": 0.56, "learning_rate": 2.1579613229712636e-05, "loss": 1.5394, "step": 3106 }, { "epoch": 0.56, "learning_rate": 2.1576902223025485e-05, "loss": 1.1203, "step": 3107 }, { "epoch": 0.56, "learning_rate": 2.1574191216338333e-05, "loss": 1.3883, "step": 3108 }, { "epoch": 0.56, "learning_rate": 2.1571480209651182e-05, "loss": 0.9308, "step": 3109 }, { "epoch": 0.56, "learning_rate": 2.1568769202964034e-05, "loss": 0.9264, "step": 3110 }, { "epoch": 0.56, "learning_rate": 2.1566058196276885e-05, "loss": 0.8648, "step": 3111 }, { "epoch": 0.56, "learning_rate": 2.1563347189589734e-05, "loss": 0.7224, "step": 3112 }, { "epoch": 0.56, "learning_rate": 2.1560636182902586e-05, "loss": 0.8179, "step": 3113 }, { "epoch": 0.56, "learning_rate": 2.1557925176215437e-05, "loss": 1.2663, "step": 3114 }, { "epoch": 0.56, "learning_rate": 2.1555214169528286e-05, "loss": 0.8658, "step": 3115 }, { "epoch": 0.56, "learning_rate": 2.1552503162841134e-05, "loss": 0.9581, "step": 3116 }, { "epoch": 0.56, "learning_rate": 2.1549792156153986e-05, "loss": 1.1434, "step": 3117 }, { "epoch": 0.56, "learning_rate": 2.1547081149466835e-05, "loss": 1.3513, "step": 3118 }, { "epoch": 0.56, "learning_rate": 2.1544370142779686e-05, "loss": 0.7452, "step": 3119 }, { "epoch": 0.56, "learning_rate": 2.1541659136092535e-05, "loss": 0.5634, "step": 3120 }, { "epoch": 0.56, "learning_rate": 2.1538948129405387e-05, "loss": 0.9787, "step": 3121 }, { "epoch": 0.56, "learning_rate": 2.153623712271824e-05, "loss": 1.2245, "step": 3122 }, { "epoch": 0.56, "learning_rate": 2.1533526116031087e-05, "loss": 1.8836, "step": 3123 }, { "epoch": 0.56, "learning_rate": 2.1530815109343935e-05, "loss": 0.7448, "step": 3124 }, { "epoch": 0.56, "learning_rate": 2.1528104102656787e-05, "loss": 1.7999, "step": 3125 }, { "epoch": 0.56, "learning_rate": 2.1525393095969636e-05, "loss": 0.9918, "step": 3126 }, { "epoch": 0.57, "learning_rate": 2.1522682089282488e-05, "loss": 1.0073, "step": 3127 }, { "epoch": 0.57, "learning_rate": 2.151997108259534e-05, "loss": 0.5542, "step": 3128 }, { "epoch": 0.57, "learning_rate": 2.1517260075908188e-05, "loss": 0.7866, "step": 3129 }, { "epoch": 0.57, "learning_rate": 2.151454906922104e-05, "loss": 0.9478, "step": 3130 }, { "epoch": 0.57, "learning_rate": 2.1511838062533888e-05, "loss": 1.4503, "step": 3131 }, { "epoch": 0.57, "learning_rate": 2.1509127055846737e-05, "loss": 0.959, "step": 3132 }, { "epoch": 0.57, "learning_rate": 2.150641604915959e-05, "loss": 1.4657, "step": 3133 }, { "epoch": 0.57, "learning_rate": 2.1503705042472437e-05, "loss": 1.0293, "step": 3134 }, { "epoch": 0.57, "learning_rate": 2.150099403578529e-05, "loss": 1.1509, "step": 3135 }, { "epoch": 0.57, "learning_rate": 2.149828302909814e-05, "loss": 0.8585, "step": 3136 }, { "epoch": 0.57, "learning_rate": 2.149557202241099e-05, "loss": 0.6931, "step": 3137 }, { "epoch": 0.57, "learning_rate": 2.149286101572384e-05, "loss": 1.4469, "step": 3138 }, { "epoch": 0.57, "learning_rate": 2.1490150009036693e-05, "loss": 0.6549, "step": 3139 }, { "epoch": 0.57, "learning_rate": 2.1487439002349538e-05, "loss": 0.7261, "step": 3140 }, { "epoch": 0.57, "learning_rate": 2.148472799566239e-05, "loss": 2.2247, "step": 3141 }, { "epoch": 0.57, "learning_rate": 2.1482016988975238e-05, "loss": 0.5169, "step": 3142 }, { "epoch": 0.57, "learning_rate": 2.147930598228809e-05, "loss": 1.0405, "step": 3143 }, { "epoch": 0.57, "learning_rate": 2.147659497560094e-05, "loss": 0.4712, "step": 3144 }, { "epoch": 0.57, "learning_rate": 2.147388396891379e-05, "loss": 1.2003, "step": 3145 }, { "epoch": 0.57, "learning_rate": 2.1471172962226642e-05, "loss": 1.2177, "step": 3146 }, { "epoch": 0.57, "learning_rate": 2.1468461955539494e-05, "loss": 1.6578, "step": 3147 }, { "epoch": 0.57, "learning_rate": 2.146575094885234e-05, "loss": 1.0064, "step": 3148 }, { "epoch": 0.57, "learning_rate": 2.146303994216519e-05, "loss": 0.8489, "step": 3149 }, { "epoch": 0.57, "learning_rate": 2.1460328935478042e-05, "loss": 1.0613, "step": 3150 }, { "epoch": 0.57, "learning_rate": 2.145761792879089e-05, "loss": 0.9721, "step": 3151 }, { "epoch": 0.57, "learning_rate": 2.1454906922103743e-05, "loss": 1.0083, "step": 3152 }, { "epoch": 0.57, "learning_rate": 2.145219591541659e-05, "loss": 1.1089, "step": 3153 }, { "epoch": 0.57, "learning_rate": 2.1449484908729443e-05, "loss": 1.4667, "step": 3154 }, { "epoch": 0.57, "learning_rate": 2.1446773902042295e-05, "loss": 1.093, "step": 3155 }, { "epoch": 0.57, "learning_rate": 2.144406289535514e-05, "loss": 2.34, "step": 3156 }, { "epoch": 0.57, "learning_rate": 2.144135188866799e-05, "loss": 1.7171, "step": 3157 }, { "epoch": 0.57, "learning_rate": 2.1438640881980843e-05, "loss": 1.5607, "step": 3158 }, { "epoch": 0.57, "learning_rate": 2.1435929875293692e-05, "loss": 0.7553, "step": 3159 }, { "epoch": 0.57, "learning_rate": 2.1433218868606544e-05, "loss": 0.9305, "step": 3160 }, { "epoch": 0.57, "learning_rate": 2.1430507861919395e-05, "loss": 1.1169, "step": 3161 }, { "epoch": 0.57, "learning_rate": 2.1427796855232244e-05, "loss": 1.3446, "step": 3162 }, { "epoch": 0.57, "learning_rate": 2.1425085848545096e-05, "loss": 1.7467, "step": 3163 }, { "epoch": 0.57, "learning_rate": 2.142237484185794e-05, "loss": 0.7248, "step": 3164 }, { "epoch": 0.57, "learning_rate": 2.1419663835170793e-05, "loss": 0.4572, "step": 3165 }, { "epoch": 0.57, "learning_rate": 2.1416952828483644e-05, "loss": 1.1811, "step": 3166 }, { "epoch": 0.57, "learning_rate": 2.1414241821796493e-05, "loss": 0.5321, "step": 3167 }, { "epoch": 0.57, "learning_rate": 2.1411530815109345e-05, "loss": 1.3098, "step": 3168 }, { "epoch": 0.57, "learning_rate": 2.1408819808422197e-05, "loss": 1.1535, "step": 3169 }, { "epoch": 0.57, "learning_rate": 2.1406108801735045e-05, "loss": 1.1779, "step": 3170 }, { "epoch": 0.57, "learning_rate": 2.1403397795047897e-05, "loss": 0.8619, "step": 3171 }, { "epoch": 0.57, "learning_rate": 2.1400686788360745e-05, "loss": 1.0569, "step": 3172 }, { "epoch": 0.57, "learning_rate": 2.1397975781673594e-05, "loss": 0.9686, "step": 3173 }, { "epoch": 0.57, "learning_rate": 2.1395264774986446e-05, "loss": 1.1947, "step": 3174 }, { "epoch": 0.57, "learning_rate": 2.1392553768299294e-05, "loss": 1.0008, "step": 3175 }, { "epoch": 0.57, "learning_rate": 2.1389842761612146e-05, "loss": 0.9364, "step": 3176 }, { "epoch": 0.57, "learning_rate": 2.1387131754924998e-05, "loss": 1.0304, "step": 3177 }, { "epoch": 0.57, "learning_rate": 2.1384420748237846e-05, "loss": 1.079, "step": 3178 }, { "epoch": 0.57, "learning_rate": 2.1381709741550698e-05, "loss": 0.8914, "step": 3179 }, { "epoch": 0.57, "learning_rate": 2.1378998734863546e-05, "loss": 1.0938, "step": 3180 }, { "epoch": 0.57, "learning_rate": 2.1376287728176395e-05, "loss": 0.813, "step": 3181 }, { "epoch": 0.58, "learning_rate": 2.1373576721489247e-05, "loss": 1.2219, "step": 3182 }, { "epoch": 0.58, "learning_rate": 2.13708657148021e-05, "loss": 1.5664, "step": 3183 }, { "epoch": 0.58, "learning_rate": 2.1368154708114947e-05, "loss": 1.481, "step": 3184 }, { "epoch": 0.58, "learning_rate": 2.13654437014278e-05, "loss": 1.4211, "step": 3185 }, { "epoch": 0.58, "learning_rate": 2.1362732694740647e-05, "loss": 1.6001, "step": 3186 }, { "epoch": 0.58, "learning_rate": 2.13600216880535e-05, "loss": 0.7866, "step": 3187 }, { "epoch": 0.58, "learning_rate": 2.1357310681366347e-05, "loss": 0.5483, "step": 3188 }, { "epoch": 0.58, "learning_rate": 2.1354599674679196e-05, "loss": 1.4259, "step": 3189 }, { "epoch": 0.58, "learning_rate": 2.1351888667992048e-05, "loss": 0.6448, "step": 3190 }, { "epoch": 0.58, "learning_rate": 2.13491776613049e-05, "loss": 0.6042, "step": 3191 }, { "epoch": 0.58, "learning_rate": 2.1346466654617748e-05, "loss": 1.1506, "step": 3192 }, { "epoch": 0.58, "learning_rate": 2.13437556479306e-05, "loss": 1.782, "step": 3193 }, { "epoch": 0.58, "learning_rate": 2.134104464124345e-05, "loss": 1.5897, "step": 3194 }, { "epoch": 0.58, "learning_rate": 2.13383336345563e-05, "loss": 1.5026, "step": 3195 }, { "epoch": 0.58, "learning_rate": 2.133562262786915e-05, "loss": 0.6999, "step": 3196 }, { "epoch": 0.58, "learning_rate": 2.1332911621181997e-05, "loss": 0.7414, "step": 3197 }, { "epoch": 0.58, "learning_rate": 2.133020061449485e-05, "loss": 0.7156, "step": 3198 }, { "epoch": 0.58, "learning_rate": 2.13274896078077e-05, "loss": 0.9523, "step": 3199 }, { "epoch": 0.58, "learning_rate": 2.132477860112055e-05, "loss": 0.839, "step": 3200 }, { "epoch": 0.58, "learning_rate": 2.13220675944334e-05, "loss": 0.6559, "step": 3201 }, { "epoch": 0.58, "learning_rate": 2.1319356587746253e-05, "loss": 0.7158, "step": 3202 }, { "epoch": 0.58, "learning_rate": 2.13166455810591e-05, "loss": 1.2256, "step": 3203 }, { "epoch": 0.58, "learning_rate": 2.131393457437195e-05, "loss": 1.1799, "step": 3204 }, { "epoch": 0.58, "learning_rate": 2.13112235676848e-05, "loss": 1.4566, "step": 3205 }, { "epoch": 0.58, "learning_rate": 2.130851256099765e-05, "loss": 1.11, "step": 3206 }, { "epoch": 0.58, "learning_rate": 2.13058015543105e-05, "loss": 0.9933, "step": 3207 }, { "epoch": 0.58, "learning_rate": 2.130309054762335e-05, "loss": 0.5365, "step": 3208 }, { "epoch": 0.58, "learning_rate": 2.1300379540936202e-05, "loss": 1.9083, "step": 3209 }, { "epoch": 0.58, "learning_rate": 2.1297668534249054e-05, "loss": 1.2911, "step": 3210 }, { "epoch": 0.58, "learning_rate": 2.1294957527561902e-05, "loss": 0.5806, "step": 3211 }, { "epoch": 0.58, "learning_rate": 2.129224652087475e-05, "loss": 1.1521, "step": 3212 }, { "epoch": 0.58, "learning_rate": 2.1289535514187602e-05, "loss": 0.7978, "step": 3213 }, { "epoch": 0.58, "learning_rate": 2.128682450750045e-05, "loss": 0.4929, "step": 3214 }, { "epoch": 0.58, "learning_rate": 2.1284113500813303e-05, "loss": 0.9789, "step": 3215 }, { "epoch": 0.58, "learning_rate": 2.1281402494126155e-05, "loss": 0.6801, "step": 3216 }, { "epoch": 0.58, "learning_rate": 2.1278691487439003e-05, "loss": 0.8488, "step": 3217 }, { "epoch": 0.58, "learning_rate": 2.1275980480751855e-05, "loss": 0.9293, "step": 3218 }, { "epoch": 0.58, "learning_rate": 2.12732694740647e-05, "loss": 2.012, "step": 3219 }, { "epoch": 0.58, "learning_rate": 2.1270558467377552e-05, "loss": 1.565, "step": 3220 }, { "epoch": 0.58, "learning_rate": 2.1267847460690404e-05, "loss": 0.9762, "step": 3221 }, { "epoch": 0.58, "learning_rate": 2.1265136454003252e-05, "loss": 0.6194, "step": 3222 }, { "epoch": 0.58, "learning_rate": 2.1262425447316104e-05, "loss": 0.8901, "step": 3223 }, { "epoch": 0.58, "learning_rate": 2.1259714440628956e-05, "loss": 1.4618, "step": 3224 }, { "epoch": 0.58, "learning_rate": 2.1257003433941804e-05, "loss": 1.0463, "step": 3225 }, { "epoch": 0.58, "learning_rate": 2.1254292427254656e-05, "loss": 1.1777, "step": 3226 }, { "epoch": 0.58, "learning_rate": 2.1251581420567504e-05, "loss": 1.0263, "step": 3227 }, { "epoch": 0.58, "learning_rate": 2.1248870413880353e-05, "loss": 1.8811, "step": 3228 }, { "epoch": 0.58, "learning_rate": 2.1246159407193205e-05, "loss": 1.9334, "step": 3229 }, { "epoch": 0.58, "learning_rate": 2.1243448400506053e-05, "loss": 1.0355, "step": 3230 }, { "epoch": 0.58, "learning_rate": 2.1240737393818905e-05, "loss": 0.6356, "step": 3231 }, { "epoch": 0.58, "learning_rate": 2.1238026387131757e-05, "loss": 0.9367, "step": 3232 }, { "epoch": 0.58, "learning_rate": 2.1235315380444605e-05, "loss": 1.1926, "step": 3233 }, { "epoch": 0.58, "learning_rate": 2.1232604373757457e-05, "loss": 1.218, "step": 3234 }, { "epoch": 0.58, "learning_rate": 2.1229893367070305e-05, "loss": 1.5409, "step": 3235 }, { "epoch": 0.58, "learning_rate": 2.1227182360383154e-05, "loss": 1.4854, "step": 3236 }, { "epoch": 0.59, "learning_rate": 2.1224471353696006e-05, "loss": 0.9461, "step": 3237 }, { "epoch": 0.59, "learning_rate": 2.1221760347008858e-05, "loss": 0.9321, "step": 3238 }, { "epoch": 0.59, "learning_rate": 2.1219049340321706e-05, "loss": 1.4536, "step": 3239 }, { "epoch": 0.59, "learning_rate": 2.1216338333634558e-05, "loss": 1.0371, "step": 3240 }, { "epoch": 0.59, "learning_rate": 2.121362732694741e-05, "loss": 1.493, "step": 3241 }, { "epoch": 0.59, "learning_rate": 2.1210916320260258e-05, "loss": 0.9621, "step": 3242 }, { "epoch": 0.59, "learning_rate": 2.1208205313573106e-05, "loss": 1.3227, "step": 3243 }, { "epoch": 0.59, "learning_rate": 2.1205494306885955e-05, "loss": 1.3034, "step": 3244 }, { "epoch": 0.59, "learning_rate": 2.1202783300198807e-05, "loss": 1.0614, "step": 3245 }, { "epoch": 0.59, "learning_rate": 2.120007229351166e-05, "loss": 1.4969, "step": 3246 }, { "epoch": 0.59, "learning_rate": 2.1197361286824507e-05, "loss": 1.2753, "step": 3247 }, { "epoch": 0.59, "learning_rate": 2.119465028013736e-05, "loss": 1.1021, "step": 3248 }, { "epoch": 0.59, "learning_rate": 2.119193927345021e-05, "loss": 1.0957, "step": 3249 }, { "epoch": 0.59, "learning_rate": 2.118922826676306e-05, "loss": 0.8463, "step": 3250 }, { "epoch": 0.59, "learning_rate": 2.1186517260075908e-05, "loss": 1.2886, "step": 3251 }, { "epoch": 0.59, "learning_rate": 2.118380625338876e-05, "loss": 1.596, "step": 3252 }, { "epoch": 0.59, "learning_rate": 2.1181095246701608e-05, "loss": 1.2522, "step": 3253 }, { "epoch": 0.59, "learning_rate": 2.117838424001446e-05, "loss": 1.7282, "step": 3254 }, { "epoch": 0.59, "learning_rate": 2.1175673233327308e-05, "loss": 0.9257, "step": 3255 }, { "epoch": 0.59, "learning_rate": 2.117296222664016e-05, "loss": 0.7881, "step": 3256 }, { "epoch": 0.59, "learning_rate": 2.1170251219953012e-05, "loss": 0.917, "step": 3257 }, { "epoch": 0.59, "learning_rate": 2.116754021326586e-05, "loss": 0.8564, "step": 3258 }, { "epoch": 0.59, "learning_rate": 2.116482920657871e-05, "loss": 0.9394, "step": 3259 }, { "epoch": 0.59, "learning_rate": 2.116211819989156e-05, "loss": 0.8649, "step": 3260 }, { "epoch": 0.59, "learning_rate": 2.115940719320441e-05, "loss": 1.209, "step": 3261 }, { "epoch": 0.59, "learning_rate": 2.115669618651726e-05, "loss": 0.9743, "step": 3262 }, { "epoch": 0.59, "learning_rate": 2.1153985179830113e-05, "loss": 1.2655, "step": 3263 }, { "epoch": 0.59, "learning_rate": 2.115127417314296e-05, "loss": 0.3595, "step": 3264 }, { "epoch": 0.59, "learning_rate": 2.1148563166455813e-05, "loss": 1.1327, "step": 3265 }, { "epoch": 0.59, "learning_rate": 2.114585215976866e-05, "loss": 1.075, "step": 3266 }, { "epoch": 0.59, "learning_rate": 2.114314115308151e-05, "loss": 0.7423, "step": 3267 }, { "epoch": 0.59, "learning_rate": 2.114043014639436e-05, "loss": 1.4075, "step": 3268 }, { "epoch": 0.59, "learning_rate": 2.113771913970721e-05, "loss": 1.0945, "step": 3269 }, { "epoch": 0.59, "learning_rate": 2.1135008133020062e-05, "loss": 1.0845, "step": 3270 }, { "epoch": 0.59, "learning_rate": 2.1132297126332914e-05, "loss": 1.0856, "step": 3271 }, { "epoch": 0.59, "learning_rate": 2.1129586119645762e-05, "loss": 1.4097, "step": 3272 }, { "epoch": 0.59, "learning_rate": 2.1126875112958614e-05, "loss": 1.7163, "step": 3273 }, { "epoch": 0.59, "learning_rate": 2.1124164106271466e-05, "loss": 1.1001, "step": 3274 }, { "epoch": 0.59, "learning_rate": 2.112145309958431e-05, "loss": 0.8026, "step": 3275 }, { "epoch": 0.59, "learning_rate": 2.1118742092897163e-05, "loss": 1.0456, "step": 3276 }, { "epoch": 0.59, "learning_rate": 2.111603108621001e-05, "loss": 0.7903, "step": 3277 }, { "epoch": 0.59, "learning_rate": 2.1113320079522863e-05, "loss": 0.748, "step": 3278 }, { "epoch": 0.59, "learning_rate": 2.1110609072835715e-05, "loss": 1.6543, "step": 3279 }, { "epoch": 0.59, "learning_rate": 2.1107898066148563e-05, "loss": 1.4999, "step": 3280 }, { "epoch": 0.59, "learning_rate": 2.1105187059461415e-05, "loss": 1.4714, "step": 3281 }, { "epoch": 0.59, "learning_rate": 2.1102476052774267e-05, "loss": 0.7203, "step": 3282 }, { "epoch": 0.59, "learning_rate": 2.1099765046087112e-05, "loss": 0.4269, "step": 3283 }, { "epoch": 0.59, "learning_rate": 2.1097054039399964e-05, "loss": 1.5728, "step": 3284 }, { "epoch": 0.59, "learning_rate": 2.1094343032712816e-05, "loss": 1.2004, "step": 3285 }, { "epoch": 0.59, "learning_rate": 2.1091632026025664e-05, "loss": 0.805, "step": 3286 }, { "epoch": 0.59, "learning_rate": 2.1088921019338516e-05, "loss": 1.1509, "step": 3287 }, { "epoch": 0.59, "learning_rate": 2.1086210012651364e-05, "loss": 1.0825, "step": 3288 }, { "epoch": 0.59, "learning_rate": 2.1083499005964216e-05, "loss": 1.6732, "step": 3289 }, { "epoch": 0.59, "learning_rate": 2.1080787999277068e-05, "loss": 0.9405, "step": 3290 }, { "epoch": 0.59, "learning_rate": 2.1078076992589913e-05, "loss": 0.7807, "step": 3291 }, { "epoch": 0.59, "learning_rate": 2.1075365985902765e-05, "loss": 0.6019, "step": 3292 }, { "epoch": 0.6, "learning_rate": 2.1072654979215617e-05, "loss": 0.9395, "step": 3293 }, { "epoch": 0.6, "learning_rate": 2.1069943972528465e-05, "loss": 0.9578, "step": 3294 }, { "epoch": 0.6, "learning_rate": 2.1067232965841317e-05, "loss": 1.1064, "step": 3295 }, { "epoch": 0.6, "learning_rate": 2.106452195915417e-05, "loss": 0.541, "step": 3296 }, { "epoch": 0.6, "learning_rate": 2.1061810952467017e-05, "loss": 1.5081, "step": 3297 }, { "epoch": 0.6, "learning_rate": 2.105909994577987e-05, "loss": 0.5629, "step": 3298 }, { "epoch": 0.6, "learning_rate": 2.1056388939092714e-05, "loss": 1.3447, "step": 3299 }, { "epoch": 0.6, "learning_rate": 2.1053677932405566e-05, "loss": 0.7994, "step": 3300 }, { "epoch": 0.6, "learning_rate": 2.1050966925718418e-05, "loss": 1.4431, "step": 3301 }, { "epoch": 0.6, "learning_rate": 2.1048255919031266e-05, "loss": 0.712, "step": 3302 }, { "epoch": 0.6, "learning_rate": 2.1045544912344118e-05, "loss": 1.2973, "step": 3303 }, { "epoch": 0.6, "learning_rate": 2.104283390565697e-05, "loss": 0.9633, "step": 3304 }, { "epoch": 0.6, "learning_rate": 2.1040122898969818e-05, "loss": 0.8663, "step": 3305 }, { "epoch": 0.6, "learning_rate": 2.103741189228267e-05, "loss": 0.8544, "step": 3306 }, { "epoch": 0.6, "learning_rate": 2.103470088559552e-05, "loss": 0.9926, "step": 3307 }, { "epoch": 0.6, "learning_rate": 2.1031989878908367e-05, "loss": 0.911, "step": 3308 }, { "epoch": 0.6, "learning_rate": 2.102927887222122e-05, "loss": 1.6048, "step": 3309 }, { "epoch": 0.6, "learning_rate": 2.1026567865534067e-05, "loss": 1.2664, "step": 3310 }, { "epoch": 0.6, "learning_rate": 2.102385685884692e-05, "loss": 0.7902, "step": 3311 }, { "epoch": 0.6, "learning_rate": 2.102114585215977e-05, "loss": 1.0517, "step": 3312 }, { "epoch": 0.6, "learning_rate": 2.101843484547262e-05, "loss": 1.2808, "step": 3313 }, { "epoch": 0.6, "learning_rate": 2.101572383878547e-05, "loss": 1.3929, "step": 3314 }, { "epoch": 0.6, "learning_rate": 2.101301283209832e-05, "loss": 0.6119, "step": 3315 }, { "epoch": 0.6, "learning_rate": 2.1010301825411168e-05, "loss": 0.6693, "step": 3316 }, { "epoch": 0.6, "learning_rate": 2.100759081872402e-05, "loss": 1.8007, "step": 3317 }, { "epoch": 0.6, "learning_rate": 2.100487981203687e-05, "loss": 0.9462, "step": 3318 }, { "epoch": 0.6, "learning_rate": 2.100216880534972e-05, "loss": 0.7892, "step": 3319 }, { "epoch": 0.6, "learning_rate": 2.0999457798662572e-05, "loss": 0.9151, "step": 3320 }, { "epoch": 0.6, "learning_rate": 2.099674679197542e-05, "loss": 1.3854, "step": 3321 }, { "epoch": 0.6, "learning_rate": 2.0994035785288272e-05, "loss": 0.6304, "step": 3322 }, { "epoch": 0.6, "learning_rate": 2.099132477860112e-05, "loss": 1.0213, "step": 3323 }, { "epoch": 0.6, "learning_rate": 2.098861377191397e-05, "loss": 1.0603, "step": 3324 }, { "epoch": 0.6, "learning_rate": 2.098590276522682e-05, "loss": 1.5936, "step": 3325 }, { "epoch": 0.6, "learning_rate": 2.0983191758539673e-05, "loss": 1.1177, "step": 3326 }, { "epoch": 0.6, "learning_rate": 2.098048075185252e-05, "loss": 0.9335, "step": 3327 }, { "epoch": 0.6, "learning_rate": 2.0977769745165373e-05, "loss": 0.9477, "step": 3328 }, { "epoch": 0.6, "learning_rate": 2.0975058738478225e-05, "loss": 0.7928, "step": 3329 }, { "epoch": 0.6, "learning_rate": 2.0972347731791073e-05, "loss": 1.4592, "step": 3330 }, { "epoch": 0.6, "learning_rate": 2.0969636725103922e-05, "loss": 0.5197, "step": 3331 }, { "epoch": 0.6, "learning_rate": 2.096692571841677e-05, "loss": 0.6507, "step": 3332 }, { "epoch": 0.6, "learning_rate": 2.0964214711729622e-05, "loss": 1.019, "step": 3333 }, { "epoch": 0.6, "learning_rate": 2.0961503705042474e-05, "loss": 0.745, "step": 3334 }, { "epoch": 0.6, "learning_rate": 2.0958792698355322e-05, "loss": 1.224, "step": 3335 }, { "epoch": 0.6, "learning_rate": 2.0956081691668174e-05, "loss": 1.4343, "step": 3336 }, { "epoch": 0.6, "learning_rate": 2.0953370684981026e-05, "loss": 1.378, "step": 3337 }, { "epoch": 0.6, "learning_rate": 2.0950659678293874e-05, "loss": 1.1295, "step": 3338 }, { "epoch": 0.6, "learning_rate": 2.0947948671606723e-05, "loss": 1.4274, "step": 3339 }, { "epoch": 0.6, "learning_rate": 2.0945237664919575e-05, "loss": 0.9234, "step": 3340 }, { "epoch": 0.6, "learning_rate": 2.0942526658232423e-05, "loss": 1.4343, "step": 3341 }, { "epoch": 0.6, "learning_rate": 2.0939815651545275e-05, "loss": 1.6079, "step": 3342 }, { "epoch": 0.6, "learning_rate": 2.0937104644858123e-05, "loss": 1.2017, "step": 3343 }, { "epoch": 0.6, "learning_rate": 2.0934393638170975e-05, "loss": 0.7711, "step": 3344 }, { "epoch": 0.6, "learning_rate": 2.0931682631483827e-05, "loss": 1.7588, "step": 3345 }, { "epoch": 0.6, "learning_rate": 2.0928971624796672e-05, "loss": 0.7621, "step": 3346 }, { "epoch": 0.6, "learning_rate": 2.0926260618109524e-05, "loss": 1.2343, "step": 3347 }, { "epoch": 0.61, "learning_rate": 2.0923549611422376e-05, "loss": 1.5275, "step": 3348 }, { "epoch": 0.61, "learning_rate": 2.0920838604735224e-05, "loss": 0.7572, "step": 3349 }, { "epoch": 0.61, "learning_rate": 2.0918127598048076e-05, "loss": 0.8793, "step": 3350 }, { "epoch": 0.61, "learning_rate": 2.0915416591360928e-05, "loss": 1.0444, "step": 3351 }, { "epoch": 0.61, "learning_rate": 2.0912705584673776e-05, "loss": 1.3012, "step": 3352 }, { "epoch": 0.61, "learning_rate": 2.0909994577986628e-05, "loss": 0.7841, "step": 3353 }, { "epoch": 0.61, "learning_rate": 2.0907283571299473e-05, "loss": 0.9415, "step": 3354 }, { "epoch": 0.61, "learning_rate": 2.0904572564612325e-05, "loss": 1.0005, "step": 3355 }, { "epoch": 0.61, "learning_rate": 2.0901861557925177e-05, "loss": 1.366, "step": 3356 }, { "epoch": 0.61, "learning_rate": 2.0899150551238025e-05, "loss": 0.9817, "step": 3357 }, { "epoch": 0.61, "learning_rate": 2.0896439544550877e-05, "loss": 1.0284, "step": 3358 }, { "epoch": 0.61, "learning_rate": 2.089372853786373e-05, "loss": 0.8626, "step": 3359 }, { "epoch": 0.61, "learning_rate": 2.0891017531176577e-05, "loss": 0.7846, "step": 3360 }, { "epoch": 0.61, "learning_rate": 2.088830652448943e-05, "loss": 1.0374, "step": 3361 }, { "epoch": 0.61, "learning_rate": 2.0885595517802278e-05, "loss": 1.3026, "step": 3362 }, { "epoch": 0.61, "learning_rate": 2.0882884511115126e-05, "loss": 1.4062, "step": 3363 }, { "epoch": 0.61, "learning_rate": 2.0880173504427978e-05, "loss": 1.1251, "step": 3364 }, { "epoch": 0.61, "learning_rate": 2.0877462497740826e-05, "loss": 1.151, "step": 3365 }, { "epoch": 0.61, "learning_rate": 2.0874751491053678e-05, "loss": 0.8404, "step": 3366 }, { "epoch": 0.61, "learning_rate": 2.087204048436653e-05, "loss": 1.1557, "step": 3367 }, { "epoch": 0.61, "learning_rate": 2.086932947767938e-05, "loss": 0.8615, "step": 3368 }, { "epoch": 0.61, "learning_rate": 2.086661847099223e-05, "loss": 1.0396, "step": 3369 }, { "epoch": 0.61, "learning_rate": 2.086390746430508e-05, "loss": 0.9555, "step": 3370 }, { "epoch": 0.61, "learning_rate": 2.0861196457617927e-05, "loss": 0.5523, "step": 3371 }, { "epoch": 0.61, "learning_rate": 2.085848545093078e-05, "loss": 1.1664, "step": 3372 }, { "epoch": 0.61, "learning_rate": 2.085577444424363e-05, "loss": 1.1253, "step": 3373 }, { "epoch": 0.61, "learning_rate": 2.085306343755648e-05, "loss": 1.7582, "step": 3374 }, { "epoch": 0.61, "learning_rate": 2.085035243086933e-05, "loss": 0.755, "step": 3375 }, { "epoch": 0.61, "learning_rate": 2.084764142418218e-05, "loss": 0.8542, "step": 3376 }, { "epoch": 0.61, "learning_rate": 2.084493041749503e-05, "loss": 0.3727, "step": 3377 }, { "epoch": 0.61, "learning_rate": 2.084221941080788e-05, "loss": 0.7129, "step": 3378 }, { "epoch": 0.61, "learning_rate": 2.0839508404120728e-05, "loss": 0.6944, "step": 3379 }, { "epoch": 0.61, "learning_rate": 2.083679739743358e-05, "loss": 0.5392, "step": 3380 }, { "epoch": 0.61, "learning_rate": 2.0834086390746432e-05, "loss": 1.4523, "step": 3381 }, { "epoch": 0.61, "learning_rate": 2.083137538405928e-05, "loss": 0.9858, "step": 3382 }, { "epoch": 0.61, "learning_rate": 2.0828664377372132e-05, "loss": 1.1177, "step": 3383 }, { "epoch": 0.61, "learning_rate": 2.0825953370684984e-05, "loss": 1.4077, "step": 3384 }, { "epoch": 0.61, "learning_rate": 2.0823242363997832e-05, "loss": 0.814, "step": 3385 }, { "epoch": 0.61, "learning_rate": 2.082053135731068e-05, "loss": 1.2457, "step": 3386 }, { "epoch": 0.61, "learning_rate": 2.0817820350623533e-05, "loss": 0.9667, "step": 3387 }, { "epoch": 0.61, "learning_rate": 2.081510934393638e-05, "loss": 1.5132, "step": 3388 }, { "epoch": 0.61, "learning_rate": 2.0812398337249233e-05, "loss": 1.1758, "step": 3389 }, { "epoch": 0.61, "learning_rate": 2.080968733056208e-05, "loss": 0.4334, "step": 3390 }, { "epoch": 0.61, "learning_rate": 2.0806976323874933e-05, "loss": 1.1875, "step": 3391 }, { "epoch": 0.61, "learning_rate": 2.0804265317187785e-05, "loss": 1.3782, "step": 3392 }, { "epoch": 0.61, "learning_rate": 2.0801554310500633e-05, "loss": 1.0704, "step": 3393 }, { "epoch": 0.61, "learning_rate": 2.0798843303813482e-05, "loss": 1.3556, "step": 3394 }, { "epoch": 0.61, "learning_rate": 2.0796132297126334e-05, "loss": 0.7563, "step": 3395 }, { "epoch": 0.61, "learning_rate": 2.0793421290439182e-05, "loss": 0.9992, "step": 3396 }, { "epoch": 0.61, "learning_rate": 2.0790710283752034e-05, "loss": 0.773, "step": 3397 }, { "epoch": 0.61, "learning_rate": 2.0787999277064886e-05, "loss": 0.6794, "step": 3398 }, { "epoch": 0.61, "learning_rate": 2.0785288270377734e-05, "loss": 1.0613, "step": 3399 }, { "epoch": 0.61, "learning_rate": 2.0782577263690586e-05, "loss": 0.6604, "step": 3400 }, { "epoch": 0.61, "learning_rate": 2.0779866257003434e-05, "loss": 1.1398, "step": 3401 }, { "epoch": 0.61, "learning_rate": 2.0777155250316283e-05, "loss": 1.3313, "step": 3402 }, { "epoch": 0.62, "learning_rate": 2.0774444243629135e-05, "loss": 1.0045, "step": 3403 }, { "epoch": 0.62, "learning_rate": 2.0771733236941983e-05, "loss": 1.4629, "step": 3404 }, { "epoch": 0.62, "learning_rate": 2.0769022230254835e-05, "loss": 1.0273, "step": 3405 }, { "epoch": 0.62, "learning_rate": 2.0766311223567687e-05, "loss": 0.6382, "step": 3406 }, { "epoch": 0.62, "learning_rate": 2.0763600216880535e-05, "loss": 0.9502, "step": 3407 }, { "epoch": 0.62, "learning_rate": 2.0760889210193387e-05, "loss": 1.0106, "step": 3408 }, { "epoch": 0.62, "learning_rate": 2.075817820350624e-05, "loss": 1.1508, "step": 3409 }, { "epoch": 0.62, "learning_rate": 2.0755467196819084e-05, "loss": 0.5907, "step": 3410 }, { "epoch": 0.62, "learning_rate": 2.0752756190131936e-05, "loss": 1.1414, "step": 3411 }, { "epoch": 0.62, "learning_rate": 2.0750045183444784e-05, "loss": 0.5278, "step": 3412 }, { "epoch": 0.62, "learning_rate": 2.0747334176757636e-05, "loss": 1.2906, "step": 3413 }, { "epoch": 0.62, "learning_rate": 2.0744623170070488e-05, "loss": 1.5898, "step": 3414 }, { "epoch": 0.62, "learning_rate": 2.0741912163383336e-05, "loss": 0.9212, "step": 3415 }, { "epoch": 0.62, "learning_rate": 2.0739201156696188e-05, "loss": 0.9523, "step": 3416 }, { "epoch": 0.62, "learning_rate": 2.073649015000904e-05, "loss": 1.0968, "step": 3417 }, { "epoch": 0.62, "learning_rate": 2.0733779143321885e-05, "loss": 1.2949, "step": 3418 }, { "epoch": 0.62, "learning_rate": 2.0731068136634737e-05, "loss": 0.8415, "step": 3419 }, { "epoch": 0.62, "learning_rate": 2.072835712994759e-05, "loss": 0.4212, "step": 3420 }, { "epoch": 0.62, "learning_rate": 2.0725646123260437e-05, "loss": 0.6548, "step": 3421 }, { "epoch": 0.62, "learning_rate": 2.072293511657329e-05, "loss": 1.1268, "step": 3422 }, { "epoch": 0.62, "learning_rate": 2.0720224109886137e-05, "loss": 0.8852, "step": 3423 }, { "epoch": 0.62, "learning_rate": 2.071751310319899e-05, "loss": 1.2319, "step": 3424 }, { "epoch": 0.62, "learning_rate": 2.071480209651184e-05, "loss": 0.9522, "step": 3425 }, { "epoch": 0.62, "learning_rate": 2.0712091089824686e-05, "loss": 0.8172, "step": 3426 }, { "epoch": 0.62, "learning_rate": 2.0709380083137538e-05, "loss": 1.1703, "step": 3427 }, { "epoch": 0.62, "learning_rate": 2.070666907645039e-05, "loss": 1.1332, "step": 3428 }, { "epoch": 0.62, "learning_rate": 2.0703958069763238e-05, "loss": 1.3144, "step": 3429 }, { "epoch": 0.62, "learning_rate": 2.070124706307609e-05, "loss": 1.261, "step": 3430 }, { "epoch": 0.62, "learning_rate": 2.0698536056388942e-05, "loss": 0.9735, "step": 3431 }, { "epoch": 0.62, "learning_rate": 2.069582504970179e-05, "loss": 0.6355, "step": 3432 }, { "epoch": 0.62, "learning_rate": 2.0693114043014642e-05, "loss": 1.225, "step": 3433 }, { "epoch": 0.62, "learning_rate": 2.0690403036327487e-05, "loss": 0.8805, "step": 3434 }, { "epoch": 0.62, "learning_rate": 2.068769202964034e-05, "loss": 1.6727, "step": 3435 }, { "epoch": 0.62, "learning_rate": 2.068498102295319e-05, "loss": 1.335, "step": 3436 }, { "epoch": 0.62, "learning_rate": 2.068227001626604e-05, "loss": 1.0886, "step": 3437 }, { "epoch": 0.62, "learning_rate": 2.067955900957889e-05, "loss": 0.6521, "step": 3438 }, { "epoch": 0.62, "learning_rate": 2.0676848002891743e-05, "loss": 0.9213, "step": 3439 }, { "epoch": 0.62, "learning_rate": 2.067413699620459e-05, "loss": 1.6397, "step": 3440 }, { "epoch": 0.62, "learning_rate": 2.0671425989517443e-05, "loss": 1.2038, "step": 3441 }, { "epoch": 0.62, "learning_rate": 2.066871498283029e-05, "loss": 0.7556, "step": 3442 }, { "epoch": 0.62, "learning_rate": 2.066600397614314e-05, "loss": 1.7865, "step": 3443 }, { "epoch": 0.62, "learning_rate": 2.0663292969455992e-05, "loss": 1.2091, "step": 3444 }, { "epoch": 0.62, "learning_rate": 2.066058196276884e-05, "loss": 0.6642, "step": 3445 }, { "epoch": 0.62, "learning_rate": 2.0657870956081692e-05, "loss": 0.8466, "step": 3446 }, { "epoch": 0.62, "learning_rate": 2.0655159949394544e-05, "loss": 0.9253, "step": 3447 }, { "epoch": 0.62, "learning_rate": 2.0652448942707392e-05, "loss": 0.6411, "step": 3448 }, { "epoch": 0.62, "learning_rate": 2.0649737936020244e-05, "loss": 1.3187, "step": 3449 }, { "epoch": 0.62, "learning_rate": 2.0647026929333093e-05, "loss": 0.8677, "step": 3450 }, { "epoch": 0.62, "learning_rate": 2.064431592264594e-05, "loss": 0.7702, "step": 3451 }, { "epoch": 0.62, "learning_rate": 2.0641604915958793e-05, "loss": 0.5456, "step": 3452 }, { "epoch": 0.62, "learning_rate": 2.0638893909271645e-05, "loss": 0.8948, "step": 3453 }, { "epoch": 0.62, "learning_rate": 2.0636182902584493e-05, "loss": 0.6605, "step": 3454 }, { "epoch": 0.62, "learning_rate": 2.0633471895897345e-05, "loss": 1.1267, "step": 3455 }, { "epoch": 0.62, "learning_rate": 2.0630760889210194e-05, "loss": 1.0637, "step": 3456 }, { "epoch": 0.62, "learning_rate": 2.0628049882523045e-05, "loss": 0.5561, "step": 3457 }, { "epoch": 0.62, "learning_rate": 2.0625338875835894e-05, "loss": 0.5923, "step": 3458 }, { "epoch": 0.63, "learning_rate": 2.0622627869148742e-05, "loss": 0.7312, "step": 3459 }, { "epoch": 0.63, "learning_rate": 2.0619916862461594e-05, "loss": 0.9911, "step": 3460 }, { "epoch": 0.63, "learning_rate": 2.0617205855774446e-05, "loss": 0.7742, "step": 3461 }, { "epoch": 0.63, "learning_rate": 2.0614494849087294e-05, "loss": 1.3984, "step": 3462 }, { "epoch": 0.63, "learning_rate": 2.0611783842400146e-05, "loss": 1.5618, "step": 3463 }, { "epoch": 0.63, "learning_rate": 2.0609072835712998e-05, "loss": 1.247, "step": 3464 }, { "epoch": 0.63, "learning_rate": 2.0606361829025846e-05, "loss": 1.4109, "step": 3465 }, { "epoch": 0.63, "learning_rate": 2.0603650822338695e-05, "loss": 1.5626, "step": 3466 }, { "epoch": 0.63, "learning_rate": 2.0600939815651543e-05, "loss": 0.7101, "step": 3467 }, { "epoch": 0.63, "learning_rate": 2.0598228808964395e-05, "loss": 0.9821, "step": 3468 }, { "epoch": 0.63, "learning_rate": 2.0595517802277247e-05, "loss": 1.9671, "step": 3469 }, { "epoch": 0.63, "learning_rate": 2.0592806795590095e-05, "loss": 0.7168, "step": 3470 }, { "epoch": 0.63, "learning_rate": 2.0590095788902947e-05, "loss": 0.9813, "step": 3471 }, { "epoch": 0.63, "learning_rate": 2.05873847822158e-05, "loss": 0.7714, "step": 3472 }, { "epoch": 0.63, "learning_rate": 2.0584673775528648e-05, "loss": 0.7837, "step": 3473 }, { "epoch": 0.63, "learning_rate": 2.0581962768841496e-05, "loss": 0.8505, "step": 3474 }, { "epoch": 0.63, "learning_rate": 2.0579251762154348e-05, "loss": 0.5281, "step": 3475 }, { "epoch": 0.63, "learning_rate": 2.0576540755467196e-05, "loss": 0.4116, "step": 3476 }, { "epoch": 0.63, "learning_rate": 2.0573829748780048e-05, "loss": 1.0766, "step": 3477 }, { "epoch": 0.63, "learning_rate": 2.0571118742092897e-05, "loss": 1.2112, "step": 3478 }, { "epoch": 0.63, "learning_rate": 2.056840773540575e-05, "loss": 0.7084, "step": 3479 }, { "epoch": 0.63, "learning_rate": 2.05656967287186e-05, "loss": 1.4651, "step": 3480 }, { "epoch": 0.63, "learning_rate": 2.0562985722031445e-05, "loss": 1.215, "step": 3481 }, { "epoch": 0.63, "learning_rate": 2.0560274715344297e-05, "loss": 1.0493, "step": 3482 }, { "epoch": 0.63, "learning_rate": 2.055756370865715e-05, "loss": 1.5412, "step": 3483 }, { "epoch": 0.63, "learning_rate": 2.0554852701969997e-05, "loss": 0.8539, "step": 3484 }, { "epoch": 0.63, "learning_rate": 2.055214169528285e-05, "loss": 0.8703, "step": 3485 }, { "epoch": 0.63, "learning_rate": 2.05494306885957e-05, "loss": 1.5589, "step": 3486 }, { "epoch": 0.63, "learning_rate": 2.054671968190855e-05, "loss": 0.9223, "step": 3487 }, { "epoch": 0.63, "learning_rate": 2.05440086752214e-05, "loss": 1.1912, "step": 3488 }, { "epoch": 0.63, "learning_rate": 2.0541297668534246e-05, "loss": 1.2405, "step": 3489 }, { "epoch": 0.63, "learning_rate": 2.0538586661847098e-05, "loss": 1.2682, "step": 3490 }, { "epoch": 0.63, "learning_rate": 2.053587565515995e-05, "loss": 0.8639, "step": 3491 }, { "epoch": 0.63, "learning_rate": 2.05331646484728e-05, "loss": 0.3881, "step": 3492 }, { "epoch": 0.63, "learning_rate": 2.053045364178565e-05, "loss": 1.1806, "step": 3493 }, { "epoch": 0.63, "learning_rate": 2.0527742635098502e-05, "loss": 1.0377, "step": 3494 }, { "epoch": 0.63, "learning_rate": 2.052503162841135e-05, "loss": 1.4751, "step": 3495 }, { "epoch": 0.63, "learning_rate": 2.0522320621724202e-05, "loss": 1.5216, "step": 3496 }, { "epoch": 0.63, "learning_rate": 2.051960961503705e-05, "loss": 0.7185, "step": 3497 }, { "epoch": 0.63, "learning_rate": 2.05168986083499e-05, "loss": 0.9607, "step": 3498 }, { "epoch": 0.63, "learning_rate": 2.051418760166275e-05, "loss": 1.0594, "step": 3499 }, { "epoch": 0.63, "learning_rate": 2.05114765949756e-05, "loss": 1.0368, "step": 3500 }, { "epoch": 0.63, "learning_rate": 2.050876558828845e-05, "loss": 1.4737, "step": 3501 }, { "epoch": 0.63, "learning_rate": 2.0506054581601303e-05, "loss": 0.8493, "step": 3502 }, { "epoch": 0.63, "learning_rate": 2.050334357491415e-05, "loss": 0.9163, "step": 3503 }, { "epoch": 0.63, "learning_rate": 2.0500632568227003e-05, "loss": 2.0134, "step": 3504 }, { "epoch": 0.63, "learning_rate": 2.0497921561539852e-05, "loss": 1.1078, "step": 3505 }, { "epoch": 0.63, "learning_rate": 2.04952105548527e-05, "loss": 0.768, "step": 3506 }, { "epoch": 0.63, "learning_rate": 2.0492499548165552e-05, "loss": 1.3386, "step": 3507 }, { "epoch": 0.63, "learning_rate": 2.0489788541478404e-05, "loss": 1.3482, "step": 3508 }, { "epoch": 0.63, "learning_rate": 2.0487077534791252e-05, "loss": 1.8138, "step": 3509 }, { "epoch": 0.63, "learning_rate": 2.0484366528104104e-05, "loss": 1.4171, "step": 3510 }, { "epoch": 0.63, "learning_rate": 2.0481655521416953e-05, "loss": 0.522, "step": 3511 }, { "epoch": 0.63, "learning_rate": 2.0478944514729804e-05, "loss": 1.1614, "step": 3512 }, { "epoch": 0.63, "learning_rate": 2.0476233508042653e-05, "loss": 1.0401, "step": 3513 }, { "epoch": 0.64, "learning_rate": 2.04735225013555e-05, "loss": 1.5474, "step": 3514 }, { "epoch": 0.64, "learning_rate": 2.0470811494668353e-05, "loss": 0.9819, "step": 3515 }, { "epoch": 0.64, "learning_rate": 2.0468100487981205e-05, "loss": 1.5093, "step": 3516 }, { "epoch": 0.64, "learning_rate": 2.0465389481294053e-05, "loss": 0.8226, "step": 3517 }, { "epoch": 0.64, "learning_rate": 2.0462678474606905e-05, "loss": 0.7087, "step": 3518 }, { "epoch": 0.64, "learning_rate": 2.0459967467919757e-05, "loss": 1.0719, "step": 3519 }, { "epoch": 0.64, "learning_rate": 2.0457256461232606e-05, "loss": 1.8979, "step": 3520 }, { "epoch": 0.64, "learning_rate": 2.0454545454545454e-05, "loss": 0.884, "step": 3521 }, { "epoch": 0.64, "learning_rate": 2.0451834447858306e-05, "loss": 1.2079, "step": 3522 }, { "epoch": 0.64, "learning_rate": 2.0449123441171154e-05, "loss": 1.4106, "step": 3523 }, { "epoch": 0.64, "learning_rate": 2.0446412434484006e-05, "loss": 1.3836, "step": 3524 }, { "epoch": 0.64, "learning_rate": 2.0443701427796855e-05, "loss": 0.3169, "step": 3525 }, { "epoch": 0.64, "learning_rate": 2.0440990421109706e-05, "loss": 1.3256, "step": 3526 }, { "epoch": 0.64, "learning_rate": 2.0438279414422558e-05, "loss": 1.4159, "step": 3527 }, { "epoch": 0.64, "learning_rate": 2.0435568407735407e-05, "loss": 1.5456, "step": 3528 }, { "epoch": 0.64, "learning_rate": 2.0432857401048255e-05, "loss": 0.818, "step": 3529 }, { "epoch": 0.64, "learning_rate": 2.0430146394361107e-05, "loss": 1.4478, "step": 3530 }, { "epoch": 0.64, "learning_rate": 2.0427435387673955e-05, "loss": 0.4572, "step": 3531 }, { "epoch": 0.64, "learning_rate": 2.0424724380986807e-05, "loss": 1.4433, "step": 3532 }, { "epoch": 0.64, "learning_rate": 2.042201337429966e-05, "loss": 0.8517, "step": 3533 }, { "epoch": 0.64, "learning_rate": 2.0419302367612507e-05, "loss": 0.7079, "step": 3534 }, { "epoch": 0.64, "learning_rate": 2.041659136092536e-05, "loss": 1.7292, "step": 3535 }, { "epoch": 0.64, "learning_rate": 2.0413880354238208e-05, "loss": 1.2975, "step": 3536 }, { "epoch": 0.64, "learning_rate": 2.0411169347551056e-05, "loss": 0.8975, "step": 3537 }, { "epoch": 0.64, "learning_rate": 2.0408458340863908e-05, "loss": 1.8126, "step": 3538 }, { "epoch": 0.64, "learning_rate": 2.0405747334176756e-05, "loss": 1.3394, "step": 3539 }, { "epoch": 0.64, "learning_rate": 2.0403036327489608e-05, "loss": 1.1765, "step": 3540 }, { "epoch": 0.64, "learning_rate": 2.040032532080246e-05, "loss": 1.161, "step": 3541 }, { "epoch": 0.64, "learning_rate": 2.039761431411531e-05, "loss": 1.0989, "step": 3542 }, { "epoch": 0.64, "learning_rate": 2.039490330742816e-05, "loss": 1.2807, "step": 3543 }, { "epoch": 0.64, "learning_rate": 2.0392192300741012e-05, "loss": 0.833, "step": 3544 }, { "epoch": 0.64, "learning_rate": 2.0389481294053857e-05, "loss": 0.7306, "step": 3545 }, { "epoch": 0.64, "learning_rate": 2.038677028736671e-05, "loss": 0.7282, "step": 3546 }, { "epoch": 0.64, "learning_rate": 2.0384059280679557e-05, "loss": 0.7799, "step": 3547 }, { "epoch": 0.64, "learning_rate": 2.038134827399241e-05, "loss": 0.7586, "step": 3548 }, { "epoch": 0.64, "learning_rate": 2.037863726730526e-05, "loss": 0.8173, "step": 3549 }, { "epoch": 0.64, "learning_rate": 2.037592626061811e-05, "loss": 1.2999, "step": 3550 }, { "epoch": 0.64, "learning_rate": 2.037321525393096e-05, "loss": 1.1455, "step": 3551 }, { "epoch": 0.64, "learning_rate": 2.0370504247243813e-05, "loss": 1.1488, "step": 3552 }, { "epoch": 0.64, "learning_rate": 2.0367793240556658e-05, "loss": 1.1661, "step": 3553 }, { "epoch": 0.64, "learning_rate": 2.036508223386951e-05, "loss": 0.8837, "step": 3554 }, { "epoch": 0.64, "learning_rate": 2.0362371227182362e-05, "loss": 0.9838, "step": 3555 }, { "epoch": 0.64, "learning_rate": 2.035966022049521e-05, "loss": 0.835, "step": 3556 }, { "epoch": 0.64, "learning_rate": 2.0356949213808062e-05, "loss": 1.5615, "step": 3557 }, { "epoch": 0.64, "learning_rate": 2.035423820712091e-05, "loss": 0.6445, "step": 3558 }, { "epoch": 0.64, "learning_rate": 2.0351527200433762e-05, "loss": 0.6797, "step": 3559 }, { "epoch": 0.64, "learning_rate": 2.0348816193746614e-05, "loss": 0.7464, "step": 3560 }, { "epoch": 0.64, "learning_rate": 2.034610518705946e-05, "loss": 1.3786, "step": 3561 }, { "epoch": 0.64, "learning_rate": 2.034339418037231e-05, "loss": 1.1244, "step": 3562 }, { "epoch": 0.64, "learning_rate": 2.0340683173685163e-05, "loss": 1.7654, "step": 3563 }, { "epoch": 0.64, "learning_rate": 2.033797216699801e-05, "loss": 0.6305, "step": 3564 }, { "epoch": 0.64, "learning_rate": 2.0335261160310863e-05, "loss": 1.0371, "step": 3565 }, { "epoch": 0.64, "learning_rate": 2.0332550153623715e-05, "loss": 1.3299, "step": 3566 }, { "epoch": 0.64, "learning_rate": 2.0329839146936564e-05, "loss": 1.1249, "step": 3567 }, { "epoch": 0.64, "learning_rate": 2.0327128140249415e-05, "loss": 1.0815, "step": 3568 }, { "epoch": 0.65, "learning_rate": 2.032441713356226e-05, "loss": 1.0021, "step": 3569 }, { "epoch": 0.65, "learning_rate": 2.0321706126875112e-05, "loss": 1.3505, "step": 3570 }, { "epoch": 0.65, "learning_rate": 2.0318995120187964e-05, "loss": 0.8733, "step": 3571 }, { "epoch": 0.65, "learning_rate": 2.0316284113500813e-05, "loss": 1.5493, "step": 3572 }, { "epoch": 0.65, "learning_rate": 2.0313573106813664e-05, "loss": 1.2247, "step": 3573 }, { "epoch": 0.65, "learning_rate": 2.0310862100126516e-05, "loss": 1.4848, "step": 3574 }, { "epoch": 0.65, "learning_rate": 2.0308151093439365e-05, "loss": 1.4827, "step": 3575 }, { "epoch": 0.65, "learning_rate": 2.0305440086752216e-05, "loss": 1.3206, "step": 3576 }, { "epoch": 0.65, "learning_rate": 2.0302729080065065e-05, "loss": 1.24, "step": 3577 }, { "epoch": 0.65, "learning_rate": 2.0300018073377913e-05, "loss": 1.4466, "step": 3578 }, { "epoch": 0.65, "learning_rate": 2.0297307066690765e-05, "loss": 1.1962, "step": 3579 }, { "epoch": 0.65, "learning_rate": 2.0294596060003614e-05, "loss": 1.4631, "step": 3580 }, { "epoch": 0.65, "learning_rate": 2.0291885053316465e-05, "loss": 0.65, "step": 3581 }, { "epoch": 0.65, "learning_rate": 2.0289174046629317e-05, "loss": 0.7068, "step": 3582 }, { "epoch": 0.65, "learning_rate": 2.0286463039942166e-05, "loss": 1.3252, "step": 3583 }, { "epoch": 0.65, "learning_rate": 2.0283752033255018e-05, "loss": 0.8861, "step": 3584 }, { "epoch": 0.65, "learning_rate": 2.0281041026567866e-05, "loss": 0.7568, "step": 3585 }, { "epoch": 0.65, "learning_rate": 2.0278330019880714e-05, "loss": 0.7004, "step": 3586 }, { "epoch": 0.65, "learning_rate": 2.0275619013193566e-05, "loss": 1.1659, "step": 3587 }, { "epoch": 0.65, "learning_rate": 2.0272908006506418e-05, "loss": 1.2604, "step": 3588 }, { "epoch": 0.65, "learning_rate": 2.0270196999819267e-05, "loss": 1.1533, "step": 3589 }, { "epoch": 0.65, "learning_rate": 2.026748599313212e-05, "loss": 1.6119, "step": 3590 }, { "epoch": 0.65, "learning_rate": 2.0264774986444967e-05, "loss": 1.2304, "step": 3591 }, { "epoch": 0.65, "learning_rate": 2.026206397975782e-05, "loss": 1.4198, "step": 3592 }, { "epoch": 0.65, "learning_rate": 2.0259352973070667e-05, "loss": 1.2144, "step": 3593 }, { "epoch": 0.65, "learning_rate": 2.0256641966383515e-05, "loss": 1.5157, "step": 3594 }, { "epoch": 0.65, "learning_rate": 2.0253930959696367e-05, "loss": 1.3676, "step": 3595 }, { "epoch": 0.65, "learning_rate": 2.025121995300922e-05, "loss": 1.3213, "step": 3596 }, { "epoch": 0.65, "learning_rate": 2.0248508946322068e-05, "loss": 1.2879, "step": 3597 }, { "epoch": 0.65, "learning_rate": 2.024579793963492e-05, "loss": 1.2519, "step": 3598 }, { "epoch": 0.65, "learning_rate": 2.024308693294777e-05, "loss": 1.2649, "step": 3599 }, { "epoch": 0.65, "learning_rate": 2.024037592626062e-05, "loss": 0.8982, "step": 3600 }, { "epoch": 0.65, "learning_rate": 2.0237664919573468e-05, "loss": 1.5512, "step": 3601 }, { "epoch": 0.65, "learning_rate": 2.0234953912886317e-05, "loss": 1.7909, "step": 3602 }, { "epoch": 0.65, "learning_rate": 2.023224290619917e-05, "loss": 1.0944, "step": 3603 }, { "epoch": 0.65, "learning_rate": 2.022953189951202e-05, "loss": 0.8527, "step": 3604 }, { "epoch": 0.65, "learning_rate": 2.022682089282487e-05, "loss": 1.0548, "step": 3605 }, { "epoch": 0.65, "learning_rate": 2.022410988613772e-05, "loss": 0.7332, "step": 3606 }, { "epoch": 0.65, "learning_rate": 2.0221398879450572e-05, "loss": 1.2988, "step": 3607 }, { "epoch": 0.65, "learning_rate": 2.0218687872763417e-05, "loss": 1.2426, "step": 3608 }, { "epoch": 0.65, "learning_rate": 2.021597686607627e-05, "loss": 1.1944, "step": 3609 }, { "epoch": 0.65, "learning_rate": 2.021326585938912e-05, "loss": 1.1323, "step": 3610 }, { "epoch": 0.65, "learning_rate": 2.021055485270197e-05, "loss": 1.7349, "step": 3611 }, { "epoch": 0.65, "learning_rate": 2.020784384601482e-05, "loss": 1.7982, "step": 3612 }, { "epoch": 0.65, "learning_rate": 2.020513283932767e-05, "loss": 1.3135, "step": 3613 }, { "epoch": 0.65, "learning_rate": 2.020242183264052e-05, "loss": 1.017, "step": 3614 }, { "epoch": 0.65, "learning_rate": 2.0199710825953373e-05, "loss": 1.0909, "step": 3615 }, { "epoch": 0.65, "learning_rate": 2.019699981926622e-05, "loss": 1.02, "step": 3616 }, { "epoch": 0.65, "learning_rate": 2.019428881257907e-05, "loss": 1.1153, "step": 3617 }, { "epoch": 0.65, "learning_rate": 2.0191577805891922e-05, "loss": 1.1785, "step": 3618 }, { "epoch": 0.65, "learning_rate": 2.018886679920477e-05, "loss": 1.7016, "step": 3619 }, { "epoch": 0.65, "learning_rate": 2.0186155792517622e-05, "loss": 1.8671, "step": 3620 }, { "epoch": 0.65, "learning_rate": 2.0183444785830474e-05, "loss": 2.3902, "step": 3621 }, { "epoch": 0.65, "learning_rate": 2.0180733779143323e-05, "loss": 0.9382, "step": 3622 }, { "epoch": 0.65, "learning_rate": 2.0178022772456174e-05, "loss": 0.993, "step": 3623 }, { "epoch": 0.65, "learning_rate": 2.017531176576902e-05, "loss": 0.9339, "step": 3624 }, { "epoch": 0.66, "learning_rate": 2.017260075908187e-05, "loss": 1.1157, "step": 3625 }, { "epoch": 0.66, "learning_rate": 2.0169889752394723e-05, "loss": 1.6316, "step": 3626 }, { "epoch": 0.66, "learning_rate": 2.016717874570757e-05, "loss": 0.8353, "step": 3627 }, { "epoch": 0.66, "learning_rate": 2.0164467739020423e-05, "loss": 1.2622, "step": 3628 }, { "epoch": 0.66, "learning_rate": 2.0161756732333275e-05, "loss": 1.1541, "step": 3629 }, { "epoch": 0.66, "learning_rate": 2.0159045725646124e-05, "loss": 0.8347, "step": 3630 }, { "epoch": 0.66, "learning_rate": 2.0156334718958976e-05, "loss": 0.716, "step": 3631 }, { "epoch": 0.66, "learning_rate": 2.0153623712271824e-05, "loss": 0.7632, "step": 3632 }, { "epoch": 0.66, "learning_rate": 2.0150912705584672e-05, "loss": 2.0456, "step": 3633 }, { "epoch": 0.66, "learning_rate": 2.0148201698897524e-05, "loss": 0.6377, "step": 3634 }, { "epoch": 0.66, "learning_rate": 2.0145490692210373e-05, "loss": 0.9711, "step": 3635 }, { "epoch": 0.66, "learning_rate": 2.0142779685523225e-05, "loss": 0.7942, "step": 3636 }, { "epoch": 0.66, "learning_rate": 2.0140068678836076e-05, "loss": 1.4861, "step": 3637 }, { "epoch": 0.66, "learning_rate": 2.0137357672148925e-05, "loss": 1.3111, "step": 3638 }, { "epoch": 0.66, "learning_rate": 2.0134646665461777e-05, "loss": 1.9785, "step": 3639 }, { "epoch": 0.66, "learning_rate": 2.0131935658774625e-05, "loss": 1.0455, "step": 3640 }, { "epoch": 0.66, "learning_rate": 2.0129224652087473e-05, "loss": 0.5749, "step": 3641 }, { "epoch": 0.66, "learning_rate": 2.0126513645400325e-05, "loss": 1.3891, "step": 3642 }, { "epoch": 0.66, "learning_rate": 2.0123802638713177e-05, "loss": 0.7461, "step": 3643 }, { "epoch": 0.66, "learning_rate": 2.0121091632026026e-05, "loss": 0.7936, "step": 3644 }, { "epoch": 0.66, "learning_rate": 2.0118380625338877e-05, "loss": 0.6865, "step": 3645 }, { "epoch": 0.66, "learning_rate": 2.0115669618651726e-05, "loss": 1.0683, "step": 3646 }, { "epoch": 0.66, "learning_rate": 2.0112958611964578e-05, "loss": 0.9281, "step": 3647 }, { "epoch": 0.66, "learning_rate": 2.0110247605277426e-05, "loss": 0.9059, "step": 3648 }, { "epoch": 0.66, "learning_rate": 2.0107536598590275e-05, "loss": 0.7708, "step": 3649 }, { "epoch": 0.66, "learning_rate": 2.0104825591903126e-05, "loss": 1.0112, "step": 3650 }, { "epoch": 0.66, "learning_rate": 2.0102114585215978e-05, "loss": 0.8516, "step": 3651 }, { "epoch": 0.66, "learning_rate": 2.0099403578528827e-05, "loss": 0.7029, "step": 3652 }, { "epoch": 0.66, "learning_rate": 2.009669257184168e-05, "loss": 0.7266, "step": 3653 }, { "epoch": 0.66, "learning_rate": 2.009398156515453e-05, "loss": 1.4128, "step": 3654 }, { "epoch": 0.66, "learning_rate": 2.009127055846738e-05, "loss": 0.7282, "step": 3655 }, { "epoch": 0.66, "learning_rate": 2.0088559551780227e-05, "loss": 1.7294, "step": 3656 }, { "epoch": 0.66, "learning_rate": 2.008584854509308e-05, "loss": 0.7275, "step": 3657 }, { "epoch": 0.66, "learning_rate": 2.0083137538405927e-05, "loss": 1.4397, "step": 3658 }, { "epoch": 0.66, "learning_rate": 2.008042653171878e-05, "loss": 1.4077, "step": 3659 }, { "epoch": 0.66, "learning_rate": 2.0077715525031628e-05, "loss": 1.05, "step": 3660 }, { "epoch": 0.66, "learning_rate": 2.007500451834448e-05, "loss": 1.3948, "step": 3661 }, { "epoch": 0.66, "learning_rate": 2.007229351165733e-05, "loss": 1.5731, "step": 3662 }, { "epoch": 0.66, "learning_rate": 2.006958250497018e-05, "loss": 0.8805, "step": 3663 }, { "epoch": 0.66, "learning_rate": 2.0066871498283028e-05, "loss": 1.622, "step": 3664 }, { "epoch": 0.66, "learning_rate": 2.006416049159588e-05, "loss": 0.9625, "step": 3665 }, { "epoch": 0.66, "learning_rate": 2.006144948490873e-05, "loss": 1.5329, "step": 3666 }, { "epoch": 0.66, "learning_rate": 2.005873847822158e-05, "loss": 1.1283, "step": 3667 }, { "epoch": 0.66, "learning_rate": 2.0056027471534432e-05, "loss": 1.0333, "step": 3668 }, { "epoch": 0.66, "learning_rate": 2.005331646484728e-05, "loss": 0.6907, "step": 3669 }, { "epoch": 0.66, "learning_rate": 2.0050605458160132e-05, "loss": 0.89, "step": 3670 }, { "epoch": 0.66, "learning_rate": 2.004789445147298e-05, "loss": 1.4768, "step": 3671 }, { "epoch": 0.66, "learning_rate": 2.004518344478583e-05, "loss": 1.1558, "step": 3672 }, { "epoch": 0.66, "learning_rate": 2.004247243809868e-05, "loss": 0.6401, "step": 3673 }, { "epoch": 0.66, "learning_rate": 2.003976143141153e-05, "loss": 1.0333, "step": 3674 }, { "epoch": 0.66, "learning_rate": 2.003705042472438e-05, "loss": 0.8987, "step": 3675 }, { "epoch": 0.66, "learning_rate": 2.0034339418037233e-05, "loss": 0.6736, "step": 3676 }, { "epoch": 0.66, "learning_rate": 2.0031628411350082e-05, "loss": 0.8723, "step": 3677 }, { "epoch": 0.66, "learning_rate": 2.0028917404662934e-05, "loss": 0.66, "step": 3678 }, { "epoch": 0.66, "learning_rate": 2.0026206397975785e-05, "loss": 0.9167, "step": 3679 }, { "epoch": 0.67, "learning_rate": 2.002349539128863e-05, "loss": 0.9237, "step": 3680 }, { "epoch": 0.67, "learning_rate": 2.0020784384601482e-05, "loss": 1.4225, "step": 3681 }, { "epoch": 0.67, "learning_rate": 2.001807337791433e-05, "loss": 0.8092, "step": 3682 }, { "epoch": 0.67, "learning_rate": 2.0015362371227183e-05, "loss": 0.9835, "step": 3683 }, { "epoch": 0.67, "learning_rate": 2.0012651364540034e-05, "loss": 0.7454, "step": 3684 }, { "epoch": 0.67, "learning_rate": 2.0009940357852883e-05, "loss": 0.6463, "step": 3685 }, { "epoch": 0.67, "learning_rate": 2.0007229351165735e-05, "loss": 1.7042, "step": 3686 }, { "epoch": 0.67, "learning_rate": 2.0004518344478586e-05, "loss": 0.8608, "step": 3687 }, { "epoch": 0.67, "learning_rate": 2.000180733779143e-05, "loss": 0.8677, "step": 3688 }, { "epoch": 0.67, "learning_rate": 1.9999096331104283e-05, "loss": 1.0433, "step": 3689 }, { "epoch": 0.67, "learning_rate": 1.9996385324417135e-05, "loss": 1.1645, "step": 3690 }, { "epoch": 0.67, "learning_rate": 1.9993674317729984e-05, "loss": 0.6941, "step": 3691 }, { "epoch": 0.67, "learning_rate": 1.9990963311042835e-05, "loss": 1.1219, "step": 3692 }, { "epoch": 0.67, "learning_rate": 1.9988252304355684e-05, "loss": 0.7059, "step": 3693 }, { "epoch": 0.67, "learning_rate": 1.9985541297668536e-05, "loss": 0.8843, "step": 3694 }, { "epoch": 0.67, "learning_rate": 1.9982830290981388e-05, "loss": 1.5204, "step": 3695 }, { "epoch": 0.67, "learning_rate": 1.9980119284294233e-05, "loss": 0.8513, "step": 3696 }, { "epoch": 0.67, "learning_rate": 1.9977408277607084e-05, "loss": 1.0107, "step": 3697 }, { "epoch": 0.67, "learning_rate": 1.9974697270919936e-05, "loss": 1.1234, "step": 3698 }, { "epoch": 0.67, "learning_rate": 1.9971986264232785e-05, "loss": 0.8458, "step": 3699 }, { "epoch": 0.67, "learning_rate": 1.9969275257545636e-05, "loss": 1.3555, "step": 3700 }, { "epoch": 0.67, "learning_rate": 1.996656425085849e-05, "loss": 2.0302, "step": 3701 }, { "epoch": 0.67, "learning_rate": 1.9963853244171337e-05, "loss": 1.2269, "step": 3702 }, { "epoch": 0.67, "learning_rate": 1.996114223748419e-05, "loss": 1.8074, "step": 3703 }, { "epoch": 0.67, "learning_rate": 1.9958431230797034e-05, "loss": 0.7036, "step": 3704 }, { "epoch": 0.67, "learning_rate": 1.9955720224109885e-05, "loss": 1.4765, "step": 3705 }, { "epoch": 0.67, "learning_rate": 1.9953009217422737e-05, "loss": 0.8694, "step": 3706 }, { "epoch": 0.67, "learning_rate": 1.9950298210735586e-05, "loss": 0.8117, "step": 3707 }, { "epoch": 0.67, "learning_rate": 1.9947587204048438e-05, "loss": 0.7407, "step": 3708 }, { "epoch": 0.67, "learning_rate": 1.994487619736129e-05, "loss": 1.18, "step": 3709 }, { "epoch": 0.67, "learning_rate": 1.9942165190674138e-05, "loss": 0.7556, "step": 3710 }, { "epoch": 0.67, "learning_rate": 1.993945418398699e-05, "loss": 0.8502, "step": 3711 }, { "epoch": 0.67, "learning_rate": 1.9936743177299838e-05, "loss": 0.5981, "step": 3712 }, { "epoch": 0.67, "learning_rate": 1.9934032170612687e-05, "loss": 1.5343, "step": 3713 }, { "epoch": 0.67, "learning_rate": 1.993132116392554e-05, "loss": 1.7154, "step": 3714 }, { "epoch": 0.67, "learning_rate": 1.9928610157238387e-05, "loss": 1.0356, "step": 3715 }, { "epoch": 0.67, "learning_rate": 1.992589915055124e-05, "loss": 0.5032, "step": 3716 }, { "epoch": 0.67, "learning_rate": 1.992318814386409e-05, "loss": 0.9331, "step": 3717 }, { "epoch": 0.67, "learning_rate": 1.992047713717694e-05, "loss": 1.1515, "step": 3718 }, { "epoch": 0.67, "learning_rate": 1.991776613048979e-05, "loss": 1.2783, "step": 3719 }, { "epoch": 0.67, "learning_rate": 1.991505512380264e-05, "loss": 1.1145, "step": 3720 }, { "epoch": 0.67, "learning_rate": 1.9912344117115488e-05, "loss": 0.9458, "step": 3721 }, { "epoch": 0.67, "learning_rate": 1.990963311042834e-05, "loss": 1.265, "step": 3722 }, { "epoch": 0.67, "learning_rate": 1.990692210374119e-05, "loss": 1.3359, "step": 3723 }, { "epoch": 0.67, "learning_rate": 1.990421109705404e-05, "loss": 1.1038, "step": 3724 }, { "epoch": 0.67, "learning_rate": 1.990150009036689e-05, "loss": 1.7429, "step": 3725 }, { "epoch": 0.67, "learning_rate": 1.989878908367974e-05, "loss": 0.5978, "step": 3726 }, { "epoch": 0.67, "learning_rate": 1.9896078076992592e-05, "loss": 1.4628, "step": 3727 }, { "epoch": 0.67, "learning_rate": 1.989336707030544e-05, "loss": 0.7541, "step": 3728 }, { "epoch": 0.67, "learning_rate": 1.989065606361829e-05, "loss": 0.8658, "step": 3729 }, { "epoch": 0.67, "learning_rate": 1.988794505693114e-05, "loss": 0.8625, "step": 3730 }, { "epoch": 0.67, "learning_rate": 1.9885234050243992e-05, "loss": 0.7895, "step": 3731 }, { "epoch": 0.67, "learning_rate": 1.988252304355684e-05, "loss": 1.194, "step": 3732 }, { "epoch": 0.67, "learning_rate": 1.9879812036869693e-05, "loss": 0.9372, "step": 3733 }, { "epoch": 0.67, "learning_rate": 1.9877101030182544e-05, "loss": 1.0075, "step": 3734 }, { "epoch": 0.68, "learning_rate": 1.987439002349539e-05, "loss": 1.344, "step": 3735 }, { "epoch": 0.68, "learning_rate": 1.987167901680824e-05, "loss": 0.6528, "step": 3736 }, { "epoch": 0.68, "learning_rate": 1.986896801012109e-05, "loss": 0.5607, "step": 3737 }, { "epoch": 0.68, "learning_rate": 1.986625700343394e-05, "loss": 1.2331, "step": 3738 }, { "epoch": 0.68, "learning_rate": 1.9863545996746793e-05, "loss": 0.9794, "step": 3739 }, { "epoch": 0.68, "learning_rate": 1.9860834990059642e-05, "loss": 1.0068, "step": 3740 }, { "epoch": 0.68, "learning_rate": 1.9858123983372494e-05, "loss": 1.5656, "step": 3741 }, { "epoch": 0.68, "learning_rate": 1.9855412976685346e-05, "loss": 1.5748, "step": 3742 }, { "epoch": 0.68, "learning_rate": 1.985270196999819e-05, "loss": 1.0757, "step": 3743 }, { "epoch": 0.68, "learning_rate": 1.9849990963311042e-05, "loss": 1.5369, "step": 3744 }, { "epoch": 0.68, "learning_rate": 1.9847279956623894e-05, "loss": 1.2089, "step": 3745 }, { "epoch": 0.68, "learning_rate": 1.9844568949936743e-05, "loss": 0.6986, "step": 3746 }, { "epoch": 0.68, "learning_rate": 1.9841857943249594e-05, "loss": 0.9283, "step": 3747 }, { "epoch": 0.68, "learning_rate": 1.9839146936562443e-05, "loss": 2.0622, "step": 3748 }, { "epoch": 0.68, "learning_rate": 1.9836435929875295e-05, "loss": 0.8607, "step": 3749 }, { "epoch": 0.68, "learning_rate": 1.9833724923188147e-05, "loss": 0.9245, "step": 3750 }, { "epoch": 0.68, "learning_rate": 1.983101391650099e-05, "loss": 0.9548, "step": 3751 }, { "epoch": 0.68, "learning_rate": 1.9828302909813843e-05, "loss": 0.7727, "step": 3752 }, { "epoch": 0.68, "learning_rate": 1.9825591903126695e-05, "loss": 1.1521, "step": 3753 }, { "epoch": 0.68, "learning_rate": 1.9822880896439544e-05, "loss": 1.1703, "step": 3754 }, { "epoch": 0.68, "learning_rate": 1.9820169889752396e-05, "loss": 1.644, "step": 3755 }, { "epoch": 0.68, "learning_rate": 1.9817458883065247e-05, "loss": 1.186, "step": 3756 }, { "epoch": 0.68, "learning_rate": 1.9814747876378096e-05, "loss": 0.6417, "step": 3757 }, { "epoch": 0.68, "learning_rate": 1.9812036869690948e-05, "loss": 0.8381, "step": 3758 }, { "epoch": 0.68, "learning_rate": 1.9809325863003793e-05, "loss": 0.4992, "step": 3759 }, { "epoch": 0.68, "learning_rate": 1.9806614856316645e-05, "loss": 1.2858, "step": 3760 }, { "epoch": 0.68, "learning_rate": 1.9803903849629496e-05, "loss": 1.3896, "step": 3761 }, { "epoch": 0.68, "learning_rate": 1.9801192842942345e-05, "loss": 1.6097, "step": 3762 }, { "epoch": 0.68, "learning_rate": 1.9798481836255197e-05, "loss": 1.1619, "step": 3763 }, { "epoch": 0.68, "learning_rate": 1.979577082956805e-05, "loss": 1.8957, "step": 3764 }, { "epoch": 0.68, "learning_rate": 1.9793059822880897e-05, "loss": 1.0039, "step": 3765 }, { "epoch": 0.68, "learning_rate": 1.979034881619375e-05, "loss": 0.6757, "step": 3766 }, { "epoch": 0.68, "learning_rate": 1.9787637809506597e-05, "loss": 1.832, "step": 3767 }, { "epoch": 0.68, "learning_rate": 1.9784926802819446e-05, "loss": 1.1423, "step": 3768 }, { "epoch": 0.68, "learning_rate": 1.9782215796132297e-05, "loss": 1.0356, "step": 3769 }, { "epoch": 0.68, "learning_rate": 1.9779504789445146e-05, "loss": 1.4206, "step": 3770 }, { "epoch": 0.68, "learning_rate": 1.9776793782757998e-05, "loss": 1.1189, "step": 3771 }, { "epoch": 0.68, "learning_rate": 1.977408277607085e-05, "loss": 0.8766, "step": 3772 }, { "epoch": 0.68, "learning_rate": 1.9771371769383698e-05, "loss": 0.4699, "step": 3773 }, { "epoch": 0.68, "learning_rate": 1.976866076269655e-05, "loss": 1.5267, "step": 3774 }, { "epoch": 0.68, "learning_rate": 1.9765949756009398e-05, "loss": 0.9998, "step": 3775 }, { "epoch": 0.68, "learning_rate": 1.9763238749322247e-05, "loss": 1.1473, "step": 3776 }, { "epoch": 0.68, "learning_rate": 1.97605277426351e-05, "loss": 1.0678, "step": 3777 }, { "epoch": 0.68, "learning_rate": 1.975781673594795e-05, "loss": 0.982, "step": 3778 }, { "epoch": 0.68, "learning_rate": 1.97551057292608e-05, "loss": 1.2674, "step": 3779 }, { "epoch": 0.68, "learning_rate": 1.975239472257365e-05, "loss": 0.814, "step": 3780 }, { "epoch": 0.68, "learning_rate": 1.97496837158865e-05, "loss": 1.7516, "step": 3781 }, { "epoch": 0.68, "learning_rate": 1.974697270919935e-05, "loss": 0.8262, "step": 3782 }, { "epoch": 0.68, "learning_rate": 1.97442617025122e-05, "loss": 0.8391, "step": 3783 }, { "epoch": 0.68, "learning_rate": 1.9741550695825048e-05, "loss": 0.7029, "step": 3784 }, { "epoch": 0.68, "learning_rate": 1.97388396891379e-05, "loss": 0.8976, "step": 3785 }, { "epoch": 0.68, "learning_rate": 1.973612868245075e-05, "loss": 1.0556, "step": 3786 }, { "epoch": 0.68, "learning_rate": 1.97334176757636e-05, "loss": 0.7749, "step": 3787 }, { "epoch": 0.68, "learning_rate": 1.9730706669076452e-05, "loss": 0.9961, "step": 3788 }, { "epoch": 0.68, "learning_rate": 1.9727995662389304e-05, "loss": 1.0159, "step": 3789 }, { "epoch": 0.68, "learning_rate": 1.9725284655702152e-05, "loss": 1.0764, "step": 3790 }, { "epoch": 0.69, "learning_rate": 1.9722573649015e-05, "loss": 1.3712, "step": 3791 }, { "epoch": 0.69, "learning_rate": 1.971986264232785e-05, "loss": 0.4004, "step": 3792 }, { "epoch": 0.69, "learning_rate": 1.97171516356407e-05, "loss": 1.3471, "step": 3793 }, { "epoch": 0.69, "learning_rate": 1.9714440628953553e-05, "loss": 1.0559, "step": 3794 }, { "epoch": 0.69, "learning_rate": 1.97117296222664e-05, "loss": 1.4032, "step": 3795 }, { "epoch": 0.69, "learning_rate": 1.9709018615579253e-05, "loss": 0.7082, "step": 3796 }, { "epoch": 0.69, "learning_rate": 1.9706307608892105e-05, "loss": 0.8234, "step": 3797 }, { "epoch": 0.69, "learning_rate": 1.9703596602204953e-05, "loss": 1.0415, "step": 3798 }, { "epoch": 0.69, "learning_rate": 1.97008855955178e-05, "loss": 1.0271, "step": 3799 }, { "epoch": 0.69, "learning_rate": 1.9698174588830653e-05, "loss": 0.95, "step": 3800 }, { "epoch": 0.69, "learning_rate": 1.9695463582143502e-05, "loss": 1.0266, "step": 3801 }, { "epoch": 0.69, "learning_rate": 1.9692752575456354e-05, "loss": 0.8429, "step": 3802 }, { "epoch": 0.69, "learning_rate": 1.9690041568769202e-05, "loss": 0.9637, "step": 3803 }, { "epoch": 0.69, "learning_rate": 1.9687330562082054e-05, "loss": 1.8121, "step": 3804 }, { "epoch": 0.69, "learning_rate": 1.9684619555394906e-05, "loss": 0.9268, "step": 3805 }, { "epoch": 0.69, "learning_rate": 1.9681908548707754e-05, "loss": 0.8463, "step": 3806 }, { "epoch": 0.69, "learning_rate": 1.9679197542020603e-05, "loss": 1.5033, "step": 3807 }, { "epoch": 0.69, "learning_rate": 1.9676486535333454e-05, "loss": 1.2861, "step": 3808 }, { "epoch": 0.69, "learning_rate": 1.9673775528646303e-05, "loss": 0.8522, "step": 3809 }, { "epoch": 0.69, "learning_rate": 1.9671064521959155e-05, "loss": 1.3782, "step": 3810 }, { "epoch": 0.69, "learning_rate": 1.9668353515272006e-05, "loss": 1.0024, "step": 3811 }, { "epoch": 0.69, "learning_rate": 1.9665642508584855e-05, "loss": 0.9942, "step": 3812 }, { "epoch": 0.69, "learning_rate": 1.9662931501897707e-05, "loss": 1.3379, "step": 3813 }, { "epoch": 0.69, "learning_rate": 1.9660220495210555e-05, "loss": 0.9784, "step": 3814 }, { "epoch": 0.69, "learning_rate": 1.9657509488523404e-05, "loss": 1.4604, "step": 3815 }, { "epoch": 0.69, "learning_rate": 1.9654798481836255e-05, "loss": 0.8767, "step": 3816 }, { "epoch": 0.69, "learning_rate": 1.9652087475149104e-05, "loss": 1.1462, "step": 3817 }, { "epoch": 0.69, "learning_rate": 1.9649376468461956e-05, "loss": 1.2224, "step": 3818 }, { "epoch": 0.69, "learning_rate": 1.9646665461774808e-05, "loss": 1.4932, "step": 3819 }, { "epoch": 0.69, "learning_rate": 1.9643954455087656e-05, "loss": 0.874, "step": 3820 }, { "epoch": 0.69, "learning_rate": 1.9641243448400508e-05, "loss": 1.2282, "step": 3821 }, { "epoch": 0.69, "learning_rate": 1.963853244171336e-05, "loss": 0.7201, "step": 3822 }, { "epoch": 0.69, "learning_rate": 1.9635821435026205e-05, "loss": 1.1135, "step": 3823 }, { "epoch": 0.69, "learning_rate": 1.9633110428339057e-05, "loss": 0.7634, "step": 3824 }, { "epoch": 0.69, "learning_rate": 1.963039942165191e-05, "loss": 0.9046, "step": 3825 }, { "epoch": 0.69, "learning_rate": 1.9627688414964757e-05, "loss": 1.3199, "step": 3826 }, { "epoch": 0.69, "learning_rate": 1.962497740827761e-05, "loss": 1.0768, "step": 3827 }, { "epoch": 0.69, "learning_rate": 1.9622266401590457e-05, "loss": 1.4359, "step": 3828 }, { "epoch": 0.69, "learning_rate": 1.961955539490331e-05, "loss": 0.7406, "step": 3829 }, { "epoch": 0.69, "learning_rate": 1.961684438821616e-05, "loss": 2.1055, "step": 3830 }, { "epoch": 0.69, "learning_rate": 1.9614133381529006e-05, "loss": 0.6902, "step": 3831 }, { "epoch": 0.69, "learning_rate": 1.9611422374841858e-05, "loss": 1.0175, "step": 3832 }, { "epoch": 0.69, "learning_rate": 1.960871136815471e-05, "loss": 1.3748, "step": 3833 }, { "epoch": 0.69, "learning_rate": 1.9606000361467558e-05, "loss": 1.0229, "step": 3834 }, { "epoch": 0.69, "learning_rate": 1.960328935478041e-05, "loss": 1.3676, "step": 3835 }, { "epoch": 0.69, "learning_rate": 1.960057834809326e-05, "loss": 1.3217, "step": 3836 }, { "epoch": 0.69, "learning_rate": 1.959786734140611e-05, "loss": 0.9154, "step": 3837 }, { "epoch": 0.69, "learning_rate": 1.9595156334718962e-05, "loss": 1.4735, "step": 3838 }, { "epoch": 0.69, "learning_rate": 1.9592445328031807e-05, "loss": 0.5877, "step": 3839 }, { "epoch": 0.69, "learning_rate": 1.958973432134466e-05, "loss": 1.041, "step": 3840 }, { "epoch": 0.69, "learning_rate": 1.958702331465751e-05, "loss": 1.6739, "step": 3841 }, { "epoch": 0.69, "learning_rate": 1.958431230797036e-05, "loss": 1.2322, "step": 3842 }, { "epoch": 0.69, "learning_rate": 1.958160130128321e-05, "loss": 1.5909, "step": 3843 }, { "epoch": 0.69, "learning_rate": 1.9578890294596063e-05, "loss": 1.1424, "step": 3844 }, { "epoch": 0.69, "learning_rate": 1.957617928790891e-05, "loss": 1.3239, "step": 3845 }, { "epoch": 0.7, "learning_rate": 1.9573468281221763e-05, "loss": 1.4752, "step": 3846 }, { "epoch": 0.7, "learning_rate": 1.957075727453461e-05, "loss": 1.3017, "step": 3847 }, { "epoch": 0.7, "learning_rate": 1.956804626784746e-05, "loss": 1.1166, "step": 3848 }, { "epoch": 0.7, "learning_rate": 1.956533526116031e-05, "loss": 0.6359, "step": 3849 }, { "epoch": 0.7, "learning_rate": 1.956262425447316e-05, "loss": 0.8282, "step": 3850 }, { "epoch": 0.7, "learning_rate": 1.9559913247786012e-05, "loss": 1.3681, "step": 3851 }, { "epoch": 0.7, "learning_rate": 1.9557202241098864e-05, "loss": 1.2307, "step": 3852 }, { "epoch": 0.7, "learning_rate": 1.9554491234411712e-05, "loss": 1.0592, "step": 3853 }, { "epoch": 0.7, "learning_rate": 1.9551780227724564e-05, "loss": 0.8498, "step": 3854 }, { "epoch": 0.7, "learning_rate": 1.9549069221037412e-05, "loss": 0.8158, "step": 3855 }, { "epoch": 0.7, "learning_rate": 1.954635821435026e-05, "loss": 0.9665, "step": 3856 }, { "epoch": 0.7, "learning_rate": 1.9543647207663113e-05, "loss": 1.0416, "step": 3857 }, { "epoch": 0.7, "learning_rate": 1.9540936200975964e-05, "loss": 1.0196, "step": 3858 }, { "epoch": 0.7, "learning_rate": 1.9538225194288813e-05, "loss": 0.9932, "step": 3859 }, { "epoch": 0.7, "learning_rate": 1.9535514187601665e-05, "loss": 1.0378, "step": 3860 }, { "epoch": 0.7, "learning_rate": 1.9532803180914513e-05, "loss": 0.921, "step": 3861 }, { "epoch": 0.7, "learning_rate": 1.953009217422736e-05, "loss": 0.8472, "step": 3862 }, { "epoch": 0.7, "learning_rate": 1.9527381167540213e-05, "loss": 0.6799, "step": 3863 }, { "epoch": 0.7, "learning_rate": 1.9524670160853062e-05, "loss": 1.5484, "step": 3864 }, { "epoch": 0.7, "learning_rate": 1.9521959154165914e-05, "loss": 0.8046, "step": 3865 }, { "epoch": 0.7, "learning_rate": 1.9519248147478766e-05, "loss": 0.803, "step": 3866 }, { "epoch": 0.7, "learning_rate": 1.9516537140791614e-05, "loss": 2.3273, "step": 3867 }, { "epoch": 0.7, "learning_rate": 1.9513826134104466e-05, "loss": 1.1658, "step": 3868 }, { "epoch": 0.7, "learning_rate": 1.9511115127417318e-05, "loss": 1.3727, "step": 3869 }, { "epoch": 0.7, "learning_rate": 1.9508404120730163e-05, "loss": 0.977, "step": 3870 }, { "epoch": 0.7, "learning_rate": 1.9505693114043015e-05, "loss": 0.7631, "step": 3871 }, { "epoch": 0.7, "learning_rate": 1.9502982107355863e-05, "loss": 0.9046, "step": 3872 }, { "epoch": 0.7, "learning_rate": 1.9500271100668715e-05, "loss": 0.9734, "step": 3873 }, { "epoch": 0.7, "learning_rate": 1.9497560093981567e-05, "loss": 1.6604, "step": 3874 }, { "epoch": 0.7, "learning_rate": 1.9494849087294415e-05, "loss": 0.7932, "step": 3875 }, { "epoch": 0.7, "learning_rate": 1.9492138080607267e-05, "loss": 0.7411, "step": 3876 }, { "epoch": 0.7, "learning_rate": 1.948942707392012e-05, "loss": 0.9883, "step": 3877 }, { "epoch": 0.7, "learning_rate": 1.9486716067232964e-05, "loss": 0.7718, "step": 3878 }, { "epoch": 0.7, "learning_rate": 1.9484005060545816e-05, "loss": 1.4541, "step": 3879 }, { "epoch": 0.7, "learning_rate": 1.9481294053858667e-05, "loss": 0.9717, "step": 3880 }, { "epoch": 0.7, "learning_rate": 1.9478583047171516e-05, "loss": 0.6504, "step": 3881 }, { "epoch": 0.7, "learning_rate": 1.9475872040484368e-05, "loss": 1.0141, "step": 3882 }, { "epoch": 0.7, "learning_rate": 1.9473161033797216e-05, "loss": 1.3702, "step": 3883 }, { "epoch": 0.7, "learning_rate": 1.9470450027110068e-05, "loss": 1.3121, "step": 3884 }, { "epoch": 0.7, "learning_rate": 1.946773902042292e-05, "loss": 1.5202, "step": 3885 }, { "epoch": 0.7, "learning_rate": 1.9465028013735765e-05, "loss": 1.2855, "step": 3886 }, { "epoch": 0.7, "learning_rate": 1.9462317007048617e-05, "loss": 1.7586, "step": 3887 }, { "epoch": 0.7, "learning_rate": 1.945960600036147e-05, "loss": 0.5615, "step": 3888 }, { "epoch": 0.7, "learning_rate": 1.9456894993674317e-05, "loss": 1.62, "step": 3889 }, { "epoch": 0.7, "learning_rate": 1.945418398698717e-05, "loss": 1.3363, "step": 3890 }, { "epoch": 0.7, "learning_rate": 1.945147298030002e-05, "loss": 0.7385, "step": 3891 }, { "epoch": 0.7, "learning_rate": 1.944876197361287e-05, "loss": 1.283, "step": 3892 }, { "epoch": 0.7, "learning_rate": 1.944605096692572e-05, "loss": 1.3374, "step": 3893 }, { "epoch": 0.7, "learning_rate": 1.9443339960238566e-05, "loss": 1.647, "step": 3894 }, { "epoch": 0.7, "learning_rate": 1.9440628953551418e-05, "loss": 1.2666, "step": 3895 }, { "epoch": 0.7, "learning_rate": 1.943791794686427e-05, "loss": 0.9315, "step": 3896 }, { "epoch": 0.7, "learning_rate": 1.9435206940177118e-05, "loss": 1.092, "step": 3897 }, { "epoch": 0.7, "learning_rate": 1.943249593348997e-05, "loss": 1.3366, "step": 3898 }, { "epoch": 0.7, "learning_rate": 1.942978492680282e-05, "loss": 1.0479, "step": 3899 }, { "epoch": 0.7, "learning_rate": 1.942707392011567e-05, "loss": 0.8157, "step": 3900 }, { "epoch": 0.71, "learning_rate": 1.9424362913428522e-05, "loss": 0.72, "step": 3901 }, { "epoch": 0.71, "learning_rate": 1.942165190674137e-05, "loss": 0.9575, "step": 3902 }, { "epoch": 0.71, "learning_rate": 1.941894090005422e-05, "loss": 0.7445, "step": 3903 }, { "epoch": 0.71, "learning_rate": 1.941622989336707e-05, "loss": 1.0547, "step": 3904 }, { "epoch": 0.71, "learning_rate": 1.941351888667992e-05, "loss": 0.9665, "step": 3905 }, { "epoch": 0.71, "learning_rate": 1.941080787999277e-05, "loss": 1.1723, "step": 3906 }, { "epoch": 0.71, "learning_rate": 1.9408096873305623e-05, "loss": 0.9559, "step": 3907 }, { "epoch": 0.71, "learning_rate": 1.940538586661847e-05, "loss": 1.1658, "step": 3908 }, { "epoch": 0.71, "learning_rate": 1.9402674859931323e-05, "loss": 1.465, "step": 3909 }, { "epoch": 0.71, "learning_rate": 1.939996385324417e-05, "loss": 1.1602, "step": 3910 }, { "epoch": 0.71, "learning_rate": 1.939725284655702e-05, "loss": 1.0362, "step": 3911 }, { "epoch": 0.71, "learning_rate": 1.9394541839869872e-05, "loss": 1.1434, "step": 3912 }, { "epoch": 0.71, "learning_rate": 1.9391830833182724e-05, "loss": 1.5449, "step": 3913 }, { "epoch": 0.71, "learning_rate": 1.9389119826495572e-05, "loss": 1.0499, "step": 3914 }, { "epoch": 0.71, "learning_rate": 1.9386408819808424e-05, "loss": 1.0711, "step": 3915 }, { "epoch": 0.71, "learning_rate": 1.9383697813121272e-05, "loss": 1.0556, "step": 3916 }, { "epoch": 0.71, "learning_rate": 1.9380986806434124e-05, "loss": 1.5451, "step": 3917 }, { "epoch": 0.71, "learning_rate": 1.9378275799746973e-05, "loss": 1.3343, "step": 3918 }, { "epoch": 0.71, "learning_rate": 1.937556479305982e-05, "loss": 0.6553, "step": 3919 }, { "epoch": 0.71, "learning_rate": 1.9372853786372673e-05, "loss": 1.2744, "step": 3920 }, { "epoch": 0.71, "learning_rate": 1.9370142779685525e-05, "loss": 1.4696, "step": 3921 }, { "epoch": 0.71, "learning_rate": 1.9367431772998373e-05, "loss": 0.597, "step": 3922 }, { "epoch": 0.71, "learning_rate": 1.9364720766311225e-05, "loss": 0.5027, "step": 3923 }, { "epoch": 0.71, "learning_rate": 1.9362009759624077e-05, "loss": 0.7699, "step": 3924 }, { "epoch": 0.71, "learning_rate": 1.9359298752936925e-05, "loss": 1.9725, "step": 3925 }, { "epoch": 0.71, "learning_rate": 1.9356587746249774e-05, "loss": 1.2395, "step": 3926 }, { "epoch": 0.71, "learning_rate": 1.9353876739562622e-05, "loss": 1.0793, "step": 3927 }, { "epoch": 0.71, "learning_rate": 1.9351165732875474e-05, "loss": 1.2189, "step": 3928 }, { "epoch": 0.71, "learning_rate": 1.9348454726188326e-05, "loss": 1.5831, "step": 3929 }, { "epoch": 0.71, "learning_rate": 1.9345743719501174e-05, "loss": 0.6298, "step": 3930 }, { "epoch": 0.71, "learning_rate": 1.9343032712814026e-05, "loss": 1.7994, "step": 3931 }, { "epoch": 0.71, "learning_rate": 1.9340321706126878e-05, "loss": 0.7811, "step": 3932 }, { "epoch": 0.71, "learning_rate": 1.9337610699439726e-05, "loss": 1.0983, "step": 3933 }, { "epoch": 0.71, "learning_rate": 1.9334899692752575e-05, "loss": 0.7896, "step": 3934 }, { "epoch": 0.71, "learning_rate": 1.9332188686065427e-05, "loss": 1.2125, "step": 3935 }, { "epoch": 0.71, "learning_rate": 1.9329477679378275e-05, "loss": 0.6568, "step": 3936 }, { "epoch": 0.71, "learning_rate": 1.9326766672691127e-05, "loss": 2.0823, "step": 3937 }, { "epoch": 0.71, "learning_rate": 1.9324055666003975e-05, "loss": 1.241, "step": 3938 }, { "epoch": 0.71, "learning_rate": 1.9321344659316827e-05, "loss": 1.0296, "step": 3939 }, { "epoch": 0.71, "learning_rate": 1.931863365262968e-05, "loss": 0.9837, "step": 3940 }, { "epoch": 0.71, "learning_rate": 1.9315922645942527e-05, "loss": 1.0052, "step": 3941 }, { "epoch": 0.71, "learning_rate": 1.9313211639255376e-05, "loss": 1.2343, "step": 3942 }, { "epoch": 0.71, "learning_rate": 1.9310500632568228e-05, "loss": 1.2475, "step": 3943 }, { "epoch": 0.71, "learning_rate": 1.9307789625881076e-05, "loss": 0.9537, "step": 3944 }, { "epoch": 0.71, "learning_rate": 1.9305078619193928e-05, "loss": 1.0209, "step": 3945 }, { "epoch": 0.71, "learning_rate": 1.930236761250678e-05, "loss": 0.4511, "step": 3946 }, { "epoch": 0.71, "learning_rate": 1.9299656605819628e-05, "loss": 1.5751, "step": 3947 }, { "epoch": 0.71, "learning_rate": 1.929694559913248e-05, "loss": 1.018, "step": 3948 }, { "epoch": 0.71, "learning_rate": 1.929423459244533e-05, "loss": 1.6488, "step": 3949 }, { "epoch": 0.71, "learning_rate": 1.9291523585758177e-05, "loss": 0.9212, "step": 3950 }, { "epoch": 0.71, "learning_rate": 1.928881257907103e-05, "loss": 0.7322, "step": 3951 }, { "epoch": 0.71, "learning_rate": 1.9286101572383877e-05, "loss": 1.5123, "step": 3952 }, { "epoch": 0.71, "learning_rate": 1.928339056569673e-05, "loss": 1.1912, "step": 3953 }, { "epoch": 0.71, "learning_rate": 1.928067955900958e-05, "loss": 0.9929, "step": 3954 }, { "epoch": 0.71, "learning_rate": 1.927796855232243e-05, "loss": 0.962, "step": 3955 }, { "epoch": 0.71, "learning_rate": 1.927525754563528e-05, "loss": 0.484, "step": 3956 }, { "epoch": 0.72, "learning_rate": 1.9272546538948133e-05, "loss": 0.694, "step": 3957 }, { "epoch": 0.72, "learning_rate": 1.9269835532260978e-05, "loss": 0.8535, "step": 3958 }, { "epoch": 0.72, "learning_rate": 1.926712452557383e-05, "loss": 0.7333, "step": 3959 }, { "epoch": 0.72, "learning_rate": 1.926441351888668e-05, "loss": 1.1163, "step": 3960 }, { "epoch": 0.72, "learning_rate": 1.926170251219953e-05, "loss": 1.1838, "step": 3961 }, { "epoch": 0.72, "learning_rate": 1.9258991505512382e-05, "loss": 0.9434, "step": 3962 }, { "epoch": 0.72, "learning_rate": 1.925628049882523e-05, "loss": 0.9635, "step": 3963 }, { "epoch": 0.72, "learning_rate": 1.9253569492138082e-05, "loss": 0.6636, "step": 3964 }, { "epoch": 0.72, "learning_rate": 1.9250858485450934e-05, "loss": 2.2022, "step": 3965 }, { "epoch": 0.72, "learning_rate": 1.924814747876378e-05, "loss": 1.1621, "step": 3966 }, { "epoch": 0.72, "learning_rate": 1.924543647207663e-05, "loss": 0.993, "step": 3967 }, { "epoch": 0.72, "learning_rate": 1.9242725465389483e-05, "loss": 1.3704, "step": 3968 }, { "epoch": 0.72, "learning_rate": 1.924001445870233e-05, "loss": 0.8394, "step": 3969 }, { "epoch": 0.72, "learning_rate": 1.9237303452015183e-05, "loss": 1.7982, "step": 3970 }, { "epoch": 0.72, "learning_rate": 1.9234592445328035e-05, "loss": 1.6057, "step": 3971 }, { "epoch": 0.72, "learning_rate": 1.9231881438640883e-05, "loss": 0.7783, "step": 3972 }, { "epoch": 0.72, "learning_rate": 1.9229170431953735e-05, "loss": 0.9191, "step": 3973 }, { "epoch": 0.72, "learning_rate": 1.922645942526658e-05, "loss": 0.707, "step": 3974 }, { "epoch": 0.72, "learning_rate": 1.9223748418579432e-05, "loss": 1.5708, "step": 3975 }, { "epoch": 0.72, "learning_rate": 1.9221037411892284e-05, "loss": 0.7478, "step": 3976 }, { "epoch": 0.72, "learning_rate": 1.9218326405205132e-05, "loss": 1.7565, "step": 3977 }, { "epoch": 0.72, "learning_rate": 1.9215615398517984e-05, "loss": 0.7325, "step": 3978 }, { "epoch": 0.72, "learning_rate": 1.9212904391830836e-05, "loss": 1.4454, "step": 3979 }, { "epoch": 0.72, "learning_rate": 1.9210193385143684e-05, "loss": 1.2218, "step": 3980 }, { "epoch": 0.72, "learning_rate": 1.9207482378456536e-05, "loss": 1.0243, "step": 3981 }, { "epoch": 0.72, "learning_rate": 1.9204771371769385e-05, "loss": 1.2788, "step": 3982 }, { "epoch": 0.72, "learning_rate": 1.9202060365082233e-05, "loss": 1.9337, "step": 3983 }, { "epoch": 0.72, "learning_rate": 1.9199349358395085e-05, "loss": 1.2044, "step": 3984 }, { "epoch": 0.72, "learning_rate": 1.9196638351707933e-05, "loss": 1.2931, "step": 3985 }, { "epoch": 0.72, "learning_rate": 1.9193927345020785e-05, "loss": 0.9468, "step": 3986 }, { "epoch": 0.72, "learning_rate": 1.9191216338333637e-05, "loss": 1.2431, "step": 3987 }, { "epoch": 0.72, "learning_rate": 1.9188505331646485e-05, "loss": 0.9403, "step": 3988 }, { "epoch": 0.72, "learning_rate": 1.9185794324959334e-05, "loss": 1.0679, "step": 3989 }, { "epoch": 0.72, "learning_rate": 1.9183083318272186e-05, "loss": 1.3353, "step": 3990 }, { "epoch": 0.72, "learning_rate": 1.9180372311585034e-05, "loss": 1.1181, "step": 3991 }, { "epoch": 0.72, "learning_rate": 1.9177661304897886e-05, "loss": 0.7054, "step": 3992 }, { "epoch": 0.72, "learning_rate": 1.9174950298210738e-05, "loss": 1.6698, "step": 3993 }, { "epoch": 0.72, "learning_rate": 1.9172239291523586e-05, "loss": 1.3985, "step": 3994 }, { "epoch": 0.72, "learning_rate": 1.9169528284836438e-05, "loss": 1.1807, "step": 3995 }, { "epoch": 0.72, "learning_rate": 1.9166817278149286e-05, "loss": 1.0632, "step": 3996 }, { "epoch": 0.72, "learning_rate": 1.9164106271462135e-05, "loss": 1.0608, "step": 3997 }, { "epoch": 0.72, "learning_rate": 1.9161395264774987e-05, "loss": 1.2629, "step": 3998 }, { "epoch": 0.72, "learning_rate": 1.9158684258087835e-05, "loss": 0.7905, "step": 3999 }, { "epoch": 0.72, "learning_rate": 1.9155973251400687e-05, "loss": 1.4931, "step": 4000 }, { "epoch": 0.72, "learning_rate": 1.915326224471354e-05, "loss": 0.9537, "step": 4001 }, { "epoch": 0.72, "learning_rate": 1.9150551238026387e-05, "loss": 1.1855, "step": 4002 }, { "epoch": 0.72, "learning_rate": 1.914784023133924e-05, "loss": 1.2077, "step": 4003 }, { "epoch": 0.72, "learning_rate": 1.914512922465209e-05, "loss": 1.3596, "step": 4004 }, { "epoch": 0.72, "learning_rate": 1.9142418217964936e-05, "loss": 0.9903, "step": 4005 }, { "epoch": 0.72, "learning_rate": 1.9139707211277788e-05, "loss": 0.8865, "step": 4006 }, { "epoch": 0.72, "learning_rate": 1.9136996204590636e-05, "loss": 1.1667, "step": 4007 }, { "epoch": 0.72, "learning_rate": 1.9134285197903488e-05, "loss": 0.6299, "step": 4008 }, { "epoch": 0.72, "learning_rate": 1.913157419121634e-05, "loss": 1.1428, "step": 4009 }, { "epoch": 0.72, "learning_rate": 1.9128863184529188e-05, "loss": 1.3083, "step": 4010 }, { "epoch": 0.72, "learning_rate": 1.912615217784204e-05, "loss": 1.0007, "step": 4011 }, { "epoch": 0.73, "learning_rate": 1.9123441171154892e-05, "loss": 0.7108, "step": 4012 }, { "epoch": 0.73, "learning_rate": 1.9120730164467737e-05, "loss": 0.8071, "step": 4013 }, { "epoch": 0.73, "learning_rate": 1.911801915778059e-05, "loss": 0.6903, "step": 4014 }, { "epoch": 0.73, "learning_rate": 1.911530815109344e-05, "loss": 1.0942, "step": 4015 }, { "epoch": 0.73, "learning_rate": 1.911259714440629e-05, "loss": 1.1092, "step": 4016 }, { "epoch": 0.73, "learning_rate": 1.910988613771914e-05, "loss": 1.6213, "step": 4017 }, { "epoch": 0.73, "learning_rate": 1.910717513103199e-05, "loss": 1.6607, "step": 4018 }, { "epoch": 0.73, "learning_rate": 1.910446412434484e-05, "loss": 1.0253, "step": 4019 }, { "epoch": 0.73, "learning_rate": 1.9101753117657693e-05, "loss": 1.3673, "step": 4020 }, { "epoch": 0.73, "learning_rate": 1.9099042110970538e-05, "loss": 1.3775, "step": 4021 }, { "epoch": 0.73, "learning_rate": 1.909633110428339e-05, "loss": 1.3057, "step": 4022 }, { "epoch": 0.73, "learning_rate": 1.9093620097596242e-05, "loss": 0.8943, "step": 4023 }, { "epoch": 0.73, "learning_rate": 1.909090909090909e-05, "loss": 1.4115, "step": 4024 }, { "epoch": 0.73, "learning_rate": 1.9088198084221942e-05, "loss": 0.9618, "step": 4025 }, { "epoch": 0.73, "learning_rate": 1.9085487077534794e-05, "loss": 1.5904, "step": 4026 }, { "epoch": 0.73, "learning_rate": 1.9082776070847642e-05, "loss": 1.3625, "step": 4027 }, { "epoch": 0.73, "learning_rate": 1.9080065064160494e-05, "loss": 1.0466, "step": 4028 }, { "epoch": 0.73, "learning_rate": 1.907735405747334e-05, "loss": 1.0959, "step": 4029 }, { "epoch": 0.73, "learning_rate": 1.907464305078619e-05, "loss": 0.8278, "step": 4030 }, { "epoch": 0.73, "learning_rate": 1.9071932044099043e-05, "loss": 0.9809, "step": 4031 }, { "epoch": 0.73, "learning_rate": 1.906922103741189e-05, "loss": 0.8774, "step": 4032 }, { "epoch": 0.73, "learning_rate": 1.9066510030724743e-05, "loss": 1.3126, "step": 4033 }, { "epoch": 0.73, "learning_rate": 1.9063799024037595e-05, "loss": 1.8258, "step": 4034 }, { "epoch": 0.73, "learning_rate": 1.9061088017350443e-05, "loss": 1.1314, "step": 4035 }, { "epoch": 0.73, "learning_rate": 1.9058377010663295e-05, "loss": 1.2836, "step": 4036 }, { "epoch": 0.73, "learning_rate": 1.9055666003976144e-05, "loss": 0.778, "step": 4037 }, { "epoch": 0.73, "learning_rate": 1.9052954997288992e-05, "loss": 0.8445, "step": 4038 }, { "epoch": 0.73, "learning_rate": 1.9050243990601844e-05, "loss": 0.8307, "step": 4039 }, { "epoch": 0.73, "learning_rate": 1.9047532983914692e-05, "loss": 0.8397, "step": 4040 }, { "epoch": 0.73, "learning_rate": 1.9044821977227544e-05, "loss": 0.8399, "step": 4041 }, { "epoch": 0.73, "learning_rate": 1.9042110970540396e-05, "loss": 0.8318, "step": 4042 }, { "epoch": 0.73, "learning_rate": 1.9039399963853244e-05, "loss": 1.4683, "step": 4043 }, { "epoch": 0.73, "learning_rate": 1.9036688957166096e-05, "loss": 0.6677, "step": 4044 }, { "epoch": 0.73, "learning_rate": 1.9033977950478945e-05, "loss": 0.8368, "step": 4045 }, { "epoch": 0.73, "learning_rate": 1.9031266943791793e-05, "loss": 1.7353, "step": 4046 }, { "epoch": 0.73, "learning_rate": 1.9028555937104645e-05, "loss": 0.9286, "step": 4047 }, { "epoch": 0.73, "learning_rate": 1.9025844930417497e-05, "loss": 1.3322, "step": 4048 }, { "epoch": 0.73, "learning_rate": 1.9023133923730345e-05, "loss": 1.0779, "step": 4049 }, { "epoch": 0.73, "learning_rate": 1.9020422917043197e-05, "loss": 0.987, "step": 4050 }, { "epoch": 0.73, "learning_rate": 1.9017711910356045e-05, "loss": 1.2969, "step": 4051 }, { "epoch": 0.73, "learning_rate": 1.9015000903668897e-05, "loss": 1.9701, "step": 4052 }, { "epoch": 0.73, "learning_rate": 1.9012289896981746e-05, "loss": 0.9566, "step": 4053 }, { "epoch": 0.73, "learning_rate": 1.9009578890294594e-05, "loss": 0.544, "step": 4054 }, { "epoch": 0.73, "learning_rate": 1.9006867883607446e-05, "loss": 1.1464, "step": 4055 }, { "epoch": 0.73, "learning_rate": 1.9004156876920298e-05, "loss": 1.4729, "step": 4056 }, { "epoch": 0.73, "learning_rate": 1.9001445870233146e-05, "loss": 1.131, "step": 4057 }, { "epoch": 0.73, "learning_rate": 1.8998734863545998e-05, "loss": 0.994, "step": 4058 }, { "epoch": 0.73, "learning_rate": 1.899602385685885e-05, "loss": 0.6982, "step": 4059 }, { "epoch": 0.73, "learning_rate": 1.89933128501717e-05, "loss": 1.6543, "step": 4060 }, { "epoch": 0.73, "learning_rate": 1.8990601843484547e-05, "loss": 1.5743, "step": 4061 }, { "epoch": 0.73, "learning_rate": 1.8987890836797395e-05, "loss": 1.7055, "step": 4062 }, { "epoch": 0.73, "learning_rate": 1.8985179830110247e-05, "loss": 1.1897, "step": 4063 }, { "epoch": 0.73, "learning_rate": 1.89824688234231e-05, "loss": 1.6874, "step": 4064 }, { "epoch": 0.73, "learning_rate": 1.8979757816735947e-05, "loss": 1.0816, "step": 4065 }, { "epoch": 0.73, "learning_rate": 1.89770468100488e-05, "loss": 0.8552, "step": 4066 }, { "epoch": 0.74, "learning_rate": 1.897433580336165e-05, "loss": 1.1093, "step": 4067 }, { "epoch": 0.74, "learning_rate": 1.89716247966745e-05, "loss": 1.2928, "step": 4068 }, { "epoch": 0.74, "learning_rate": 1.8968913789987348e-05, "loss": 1.0201, "step": 4069 }, { "epoch": 0.74, "learning_rate": 1.89662027833002e-05, "loss": 1.3667, "step": 4070 }, { "epoch": 0.74, "learning_rate": 1.8963491776613048e-05, "loss": 1.068, "step": 4071 }, { "epoch": 0.74, "learning_rate": 1.89607807699259e-05, "loss": 1.3521, "step": 4072 }, { "epoch": 0.74, "learning_rate": 1.895806976323875e-05, "loss": 0.8441, "step": 4073 }, { "epoch": 0.74, "learning_rate": 1.89553587565516e-05, "loss": 1.1148, "step": 4074 }, { "epoch": 0.74, "learning_rate": 1.8952647749864452e-05, "loss": 0.5434, "step": 4075 }, { "epoch": 0.74, "learning_rate": 1.89499367431773e-05, "loss": 1.5676, "step": 4076 }, { "epoch": 0.74, "learning_rate": 1.894722573649015e-05, "loss": 0.9449, "step": 4077 }, { "epoch": 0.74, "learning_rate": 1.8944514729803e-05, "loss": 1.0002, "step": 4078 }, { "epoch": 0.74, "learning_rate": 1.894180372311585e-05, "loss": 0.8795, "step": 4079 }, { "epoch": 0.74, "learning_rate": 1.89390927164287e-05, "loss": 0.8081, "step": 4080 }, { "epoch": 0.74, "learning_rate": 1.8936381709741553e-05, "loss": 0.7566, "step": 4081 }, { "epoch": 0.74, "learning_rate": 1.89336707030544e-05, "loss": 0.7675, "step": 4082 }, { "epoch": 0.74, "learning_rate": 1.8930959696367253e-05, "loss": 1.404, "step": 4083 }, { "epoch": 0.74, "learning_rate": 1.89282486896801e-05, "loss": 0.967, "step": 4084 }, { "epoch": 0.74, "learning_rate": 1.892553768299295e-05, "loss": 0.9925, "step": 4085 }, { "epoch": 0.74, "learning_rate": 1.8922826676305802e-05, "loss": 1.064, "step": 4086 }, { "epoch": 0.74, "learning_rate": 1.892011566961865e-05, "loss": 0.8018, "step": 4087 }, { "epoch": 0.74, "learning_rate": 1.8917404662931502e-05, "loss": 1.0773, "step": 4088 }, { "epoch": 0.74, "learning_rate": 1.8914693656244354e-05, "loss": 1.0949, "step": 4089 }, { "epoch": 0.74, "learning_rate": 1.8911982649557202e-05, "loss": 1.0855, "step": 4090 }, { "epoch": 0.74, "learning_rate": 1.8909271642870054e-05, "loss": 0.9051, "step": 4091 }, { "epoch": 0.74, "learning_rate": 1.8906560636182906e-05, "loss": 1.0342, "step": 4092 }, { "epoch": 0.74, "learning_rate": 1.890384962949575e-05, "loss": 1.246, "step": 4093 }, { "epoch": 0.74, "learning_rate": 1.8901138622808603e-05, "loss": 0.9919, "step": 4094 }, { "epoch": 0.74, "learning_rate": 1.8898427616121455e-05, "loss": 1.9156, "step": 4095 }, { "epoch": 0.74, "learning_rate": 1.8895716609434303e-05, "loss": 1.0148, "step": 4096 }, { "epoch": 0.74, "learning_rate": 1.8893005602747155e-05, "loss": 0.6305, "step": 4097 }, { "epoch": 0.74, "learning_rate": 1.8890294596060003e-05, "loss": 0.8766, "step": 4098 }, { "epoch": 0.74, "learning_rate": 1.8887583589372855e-05, "loss": 1.1614, "step": 4099 }, { "epoch": 0.74, "learning_rate": 1.8884872582685707e-05, "loss": 1.2272, "step": 4100 }, { "epoch": 0.74, "learning_rate": 1.8882161575998552e-05, "loss": 0.7604, "step": 4101 }, { "epoch": 0.74, "learning_rate": 1.8879450569311404e-05, "loss": 1.0257, "step": 4102 }, { "epoch": 0.74, "learning_rate": 1.8876739562624256e-05, "loss": 1.1084, "step": 4103 }, { "epoch": 0.74, "learning_rate": 1.8874028555937104e-05, "loss": 1.6394, "step": 4104 }, { "epoch": 0.74, "learning_rate": 1.8871317549249956e-05, "loss": 1.358, "step": 4105 }, { "epoch": 0.74, "learning_rate": 1.8868606542562808e-05, "loss": 1.3384, "step": 4106 }, { "epoch": 0.74, "learning_rate": 1.8865895535875656e-05, "loss": 0.9843, "step": 4107 }, { "epoch": 0.74, "learning_rate": 1.8863184529188508e-05, "loss": 1.0821, "step": 4108 }, { "epoch": 0.74, "learning_rate": 1.8860473522501353e-05, "loss": 0.9494, "step": 4109 }, { "epoch": 0.74, "learning_rate": 1.8857762515814205e-05, "loss": 1.5904, "step": 4110 }, { "epoch": 0.74, "learning_rate": 1.8855051509127057e-05, "loss": 0.7814, "step": 4111 }, { "epoch": 0.74, "learning_rate": 1.8852340502439905e-05, "loss": 1.3541, "step": 4112 }, { "epoch": 0.74, "learning_rate": 1.8849629495752757e-05, "loss": 1.4688, "step": 4113 }, { "epoch": 0.74, "learning_rate": 1.884691848906561e-05, "loss": 0.7477, "step": 4114 }, { "epoch": 0.74, "learning_rate": 1.8844207482378457e-05, "loss": 0.7241, "step": 4115 }, { "epoch": 0.74, "learning_rate": 1.884149647569131e-05, "loss": 1.0155, "step": 4116 }, { "epoch": 0.74, "learning_rate": 1.8838785469004158e-05, "loss": 1.2676, "step": 4117 }, { "epoch": 0.74, "learning_rate": 1.8836074462317006e-05, "loss": 0.6602, "step": 4118 }, { "epoch": 0.74, "learning_rate": 1.8833363455629858e-05, "loss": 0.8138, "step": 4119 }, { "epoch": 0.74, "learning_rate": 1.8830652448942706e-05, "loss": 0.7063, "step": 4120 }, { "epoch": 0.74, "learning_rate": 1.8827941442255558e-05, "loss": 0.9549, "step": 4121 }, { "epoch": 0.74, "learning_rate": 1.882523043556841e-05, "loss": 1.5397, "step": 4122 }, { "epoch": 0.75, "learning_rate": 1.882251942888126e-05, "loss": 1.9312, "step": 4123 }, { "epoch": 0.75, "learning_rate": 1.8819808422194107e-05, "loss": 1.7265, "step": 4124 }, { "epoch": 0.75, "learning_rate": 1.881709741550696e-05, "loss": 0.9665, "step": 4125 }, { "epoch": 0.75, "learning_rate": 1.8814386408819807e-05, "loss": 1.2353, "step": 4126 }, { "epoch": 0.75, "learning_rate": 1.881167540213266e-05, "loss": 1.7221, "step": 4127 }, { "epoch": 0.75, "learning_rate": 1.880896439544551e-05, "loss": 0.3966, "step": 4128 }, { "epoch": 0.75, "learning_rate": 1.880625338875836e-05, "loss": 0.8913, "step": 4129 }, { "epoch": 0.75, "learning_rate": 1.880354238207121e-05, "loss": 0.707, "step": 4130 }, { "epoch": 0.75, "learning_rate": 1.880083137538406e-05, "loss": 1.1635, "step": 4131 }, { "epoch": 0.75, "learning_rate": 1.8798120368696908e-05, "loss": 0.8502, "step": 4132 }, { "epoch": 0.75, "learning_rate": 1.879540936200976e-05, "loss": 1.124, "step": 4133 }, { "epoch": 0.75, "learning_rate": 1.879269835532261e-05, "loss": 0.9938, "step": 4134 }, { "epoch": 0.75, "learning_rate": 1.878998734863546e-05, "loss": 0.9543, "step": 4135 }, { "epoch": 0.75, "learning_rate": 1.8787276341948312e-05, "loss": 1.2232, "step": 4136 }, { "epoch": 0.75, "learning_rate": 1.878456533526116e-05, "loss": 1.5025, "step": 4137 }, { "epoch": 0.75, "learning_rate": 1.8781854328574012e-05, "loss": 0.5931, "step": 4138 }, { "epoch": 0.75, "learning_rate": 1.8779143321886864e-05, "loss": 1.4356, "step": 4139 }, { "epoch": 0.75, "learning_rate": 1.877643231519971e-05, "loss": 2.3294, "step": 4140 }, { "epoch": 0.75, "learning_rate": 1.877372130851256e-05, "loss": 1.0618, "step": 4141 }, { "epoch": 0.75, "learning_rate": 1.877101030182541e-05, "loss": 0.8148, "step": 4142 }, { "epoch": 0.75, "learning_rate": 1.876829929513826e-05, "loss": 1.1325, "step": 4143 }, { "epoch": 0.75, "learning_rate": 1.8765588288451113e-05, "loss": 1.0747, "step": 4144 }, { "epoch": 0.75, "learning_rate": 1.876287728176396e-05, "loss": 1.1443, "step": 4145 }, { "epoch": 0.75, "learning_rate": 1.8760166275076813e-05, "loss": 0.6908, "step": 4146 }, { "epoch": 0.75, "learning_rate": 1.8757455268389665e-05, "loss": 1.1958, "step": 4147 }, { "epoch": 0.75, "learning_rate": 1.875474426170251e-05, "loss": 2.0696, "step": 4148 }, { "epoch": 0.75, "learning_rate": 1.8752033255015362e-05, "loss": 1.4314, "step": 4149 }, { "epoch": 0.75, "learning_rate": 1.8749322248328214e-05, "loss": 1.0035, "step": 4150 }, { "epoch": 0.75, "learning_rate": 1.8746611241641062e-05, "loss": 0.7965, "step": 4151 }, { "epoch": 0.75, "learning_rate": 1.8743900234953914e-05, "loss": 1.6436, "step": 4152 }, { "epoch": 0.75, "learning_rate": 1.8741189228266763e-05, "loss": 1.4065, "step": 4153 }, { "epoch": 0.75, "learning_rate": 1.8738478221579614e-05, "loss": 1.3834, "step": 4154 }, { "epoch": 0.75, "learning_rate": 1.8735767214892466e-05, "loss": 1.2043, "step": 4155 }, { "epoch": 0.75, "learning_rate": 1.873305620820531e-05, "loss": 0.5839, "step": 4156 }, { "epoch": 0.75, "learning_rate": 1.8730345201518163e-05, "loss": 1.3701, "step": 4157 }, { "epoch": 0.75, "learning_rate": 1.8727634194831015e-05, "loss": 0.8201, "step": 4158 }, { "epoch": 0.75, "learning_rate": 1.8724923188143863e-05, "loss": 1.0926, "step": 4159 }, { "epoch": 0.75, "learning_rate": 1.8722212181456715e-05, "loss": 1.1272, "step": 4160 }, { "epoch": 0.75, "learning_rate": 1.8719501174769567e-05, "loss": 0.698, "step": 4161 }, { "epoch": 0.75, "learning_rate": 1.8716790168082415e-05, "loss": 1.2859, "step": 4162 }, { "epoch": 0.75, "learning_rate": 1.8714079161395267e-05, "loss": 0.8272, "step": 4163 }, { "epoch": 0.75, "learning_rate": 1.8711368154708112e-05, "loss": 0.8738, "step": 4164 }, { "epoch": 0.75, "learning_rate": 1.8708657148020964e-05, "loss": 1.0421, "step": 4165 }, { "epoch": 0.75, "learning_rate": 1.8705946141333816e-05, "loss": 0.9787, "step": 4166 }, { "epoch": 0.75, "learning_rate": 1.8703235134646664e-05, "loss": 1.2029, "step": 4167 }, { "epoch": 0.75, "learning_rate": 1.8700524127959516e-05, "loss": 1.0998, "step": 4168 }, { "epoch": 0.75, "learning_rate": 1.8697813121272368e-05, "loss": 1.1193, "step": 4169 }, { "epoch": 0.75, "learning_rate": 1.8695102114585217e-05, "loss": 1.3004, "step": 4170 }, { "epoch": 0.75, "learning_rate": 1.869239110789807e-05, "loss": 0.8194, "step": 4171 }, { "epoch": 0.75, "learning_rate": 1.8689680101210917e-05, "loss": 1.0888, "step": 4172 }, { "epoch": 0.75, "learning_rate": 1.8686969094523765e-05, "loss": 1.2701, "step": 4173 }, { "epoch": 0.75, "learning_rate": 1.8684258087836617e-05, "loss": 1.294, "step": 4174 }, { "epoch": 0.75, "learning_rate": 1.8681547081149466e-05, "loss": 0.9969, "step": 4175 }, { "epoch": 0.75, "learning_rate": 1.8678836074462317e-05, "loss": 1.3942, "step": 4176 }, { "epoch": 0.75, "learning_rate": 1.867612506777517e-05, "loss": 1.0515, "step": 4177 }, { "epoch": 0.76, "learning_rate": 1.8673414061088018e-05, "loss": 1.3677, "step": 4178 }, { "epoch": 0.76, "learning_rate": 1.867070305440087e-05, "loss": 0.7796, "step": 4179 }, { "epoch": 0.76, "learning_rate": 1.8667992047713718e-05, "loss": 0.7902, "step": 4180 }, { "epoch": 0.76, "learning_rate": 1.8665281041026566e-05, "loss": 0.7968, "step": 4181 }, { "epoch": 0.76, "learning_rate": 1.8662570034339418e-05, "loss": 0.6289, "step": 4182 }, { "epoch": 0.76, "learning_rate": 1.865985902765227e-05, "loss": 1.1148, "step": 4183 }, { "epoch": 0.76, "learning_rate": 1.865714802096512e-05, "loss": 1.8155, "step": 4184 }, { "epoch": 0.76, "learning_rate": 1.865443701427797e-05, "loss": 0.6225, "step": 4185 }, { "epoch": 0.76, "learning_rate": 1.865172600759082e-05, "loss": 1.1766, "step": 4186 }, { "epoch": 0.76, "learning_rate": 1.864901500090367e-05, "loss": 1.1404, "step": 4187 }, { "epoch": 0.76, "learning_rate": 1.864630399421652e-05, "loss": 1.1526, "step": 4188 }, { "epoch": 0.76, "learning_rate": 1.8643592987529367e-05, "loss": 1.5008, "step": 4189 }, { "epoch": 0.76, "learning_rate": 1.864088198084222e-05, "loss": 1.8709, "step": 4190 }, { "epoch": 0.76, "learning_rate": 1.863817097415507e-05, "loss": 1.2361, "step": 4191 }, { "epoch": 0.76, "learning_rate": 1.863545996746792e-05, "loss": 1.0664, "step": 4192 }, { "epoch": 0.76, "learning_rate": 1.863274896078077e-05, "loss": 0.638, "step": 4193 }, { "epoch": 0.76, "learning_rate": 1.8630037954093623e-05, "loss": 0.6465, "step": 4194 }, { "epoch": 0.76, "learning_rate": 1.862732694740647e-05, "loss": 1.2751, "step": 4195 }, { "epoch": 0.76, "learning_rate": 1.862461594071932e-05, "loss": 1.0229, "step": 4196 }, { "epoch": 0.76, "learning_rate": 1.862190493403217e-05, "loss": 0.9281, "step": 4197 }, { "epoch": 0.76, "learning_rate": 1.861919392734502e-05, "loss": 0.8675, "step": 4198 }, { "epoch": 0.76, "learning_rate": 1.8616482920657872e-05, "loss": 0.5485, "step": 4199 }, { "epoch": 0.76, "learning_rate": 1.861377191397072e-05, "loss": 1.3977, "step": 4200 }, { "epoch": 0.76, "learning_rate": 1.8611060907283572e-05, "loss": 0.7601, "step": 4201 }, { "epoch": 0.76, "learning_rate": 1.8608349900596424e-05, "loss": 0.5618, "step": 4202 }, { "epoch": 0.76, "learning_rate": 1.8605638893909273e-05, "loss": 0.5934, "step": 4203 }, { "epoch": 0.76, "learning_rate": 1.860292788722212e-05, "loss": 1.6908, "step": 4204 }, { "epoch": 0.76, "learning_rate": 1.8600216880534973e-05, "loss": 0.6517, "step": 4205 }, { "epoch": 0.76, "learning_rate": 1.859750587384782e-05, "loss": 1.4908, "step": 4206 }, { "epoch": 0.76, "learning_rate": 1.8594794867160673e-05, "loss": 0.851, "step": 4207 }, { "epoch": 0.76, "learning_rate": 1.859208386047352e-05, "loss": 0.72, "step": 4208 }, { "epoch": 0.76, "learning_rate": 1.8589372853786373e-05, "loss": 1.2055, "step": 4209 }, { "epoch": 0.76, "learning_rate": 1.8586661847099225e-05, "loss": 1.821, "step": 4210 }, { "epoch": 0.76, "learning_rate": 1.8583950840412074e-05, "loss": 1.1098, "step": 4211 }, { "epoch": 0.76, "learning_rate": 1.8581239833724922e-05, "loss": 1.4492, "step": 4212 }, { "epoch": 0.76, "learning_rate": 1.8578528827037774e-05, "loss": 1.4953, "step": 4213 }, { "epoch": 0.76, "learning_rate": 1.8575817820350622e-05, "loss": 1.3702, "step": 4214 }, { "epoch": 0.76, "learning_rate": 1.8573106813663474e-05, "loss": 1.5406, "step": 4215 }, { "epoch": 0.76, "learning_rate": 1.8570395806976326e-05, "loss": 0.8064, "step": 4216 }, { "epoch": 0.76, "learning_rate": 1.8567684800289175e-05, "loss": 0.891, "step": 4217 }, { "epoch": 0.76, "learning_rate": 1.8564973793602026e-05, "loss": 0.5757, "step": 4218 }, { "epoch": 0.76, "learning_rate": 1.8562262786914875e-05, "loss": 0.9051, "step": 4219 }, { "epoch": 0.76, "learning_rate": 1.8559551780227723e-05, "loss": 1.3779, "step": 4220 }, { "epoch": 0.76, "learning_rate": 1.8556840773540575e-05, "loss": 1.2026, "step": 4221 }, { "epoch": 0.76, "learning_rate": 1.8554129766853424e-05, "loss": 0.9599, "step": 4222 }, { "epoch": 0.76, "learning_rate": 1.8551418760166275e-05, "loss": 0.9706, "step": 4223 }, { "epoch": 0.76, "learning_rate": 1.8548707753479127e-05, "loss": 1.2327, "step": 4224 }, { "epoch": 0.76, "learning_rate": 1.8545996746791976e-05, "loss": 2.0394, "step": 4225 }, { "epoch": 0.76, "learning_rate": 1.8543285740104827e-05, "loss": 1.3096, "step": 4226 }, { "epoch": 0.76, "learning_rate": 1.854057473341768e-05, "loss": 0.6357, "step": 4227 }, { "epoch": 0.76, "learning_rate": 1.8537863726730524e-05, "loss": 1.4146, "step": 4228 }, { "epoch": 0.76, "learning_rate": 1.8535152720043376e-05, "loss": 1.1033, "step": 4229 }, { "epoch": 0.76, "learning_rate": 1.8532441713356225e-05, "loss": 1.1678, "step": 4230 }, { "epoch": 0.76, "learning_rate": 1.8529730706669076e-05, "loss": 1.0214, "step": 4231 }, { "epoch": 0.76, "learning_rate": 1.8527019699981928e-05, "loss": 1.1439, "step": 4232 }, { "epoch": 0.77, "learning_rate": 1.8524308693294777e-05, "loss": 1.2791, "step": 4233 }, { "epoch": 0.77, "learning_rate": 1.852159768660763e-05, "loss": 0.8565, "step": 4234 }, { "epoch": 0.77, "learning_rate": 1.851888667992048e-05, "loss": 1.0165, "step": 4235 }, { "epoch": 0.77, "learning_rate": 1.8516175673233325e-05, "loss": 1.4325, "step": 4236 }, { "epoch": 0.77, "learning_rate": 1.8513464666546177e-05, "loss": 0.8627, "step": 4237 }, { "epoch": 0.77, "learning_rate": 1.851075365985903e-05, "loss": 1.4789, "step": 4238 }, { "epoch": 0.77, "learning_rate": 1.8508042653171878e-05, "loss": 1.2126, "step": 4239 }, { "epoch": 0.77, "learning_rate": 1.850533164648473e-05, "loss": 1.7906, "step": 4240 }, { "epoch": 0.77, "learning_rate": 1.8502620639797578e-05, "loss": 0.9936, "step": 4241 }, { "epoch": 0.77, "learning_rate": 1.849990963311043e-05, "loss": 1.0173, "step": 4242 }, { "epoch": 0.77, "learning_rate": 1.849719862642328e-05, "loss": 0.964, "step": 4243 }, { "epoch": 0.77, "learning_rate": 1.8494487619736126e-05, "loss": 0.6267, "step": 4244 }, { "epoch": 0.77, "learning_rate": 1.8491776613048978e-05, "loss": 0.5763, "step": 4245 }, { "epoch": 0.77, "learning_rate": 1.848906560636183e-05, "loss": 1.2983, "step": 4246 }, { "epoch": 0.77, "learning_rate": 1.848635459967468e-05, "loss": 0.5374, "step": 4247 }, { "epoch": 0.77, "learning_rate": 1.848364359298753e-05, "loss": 0.6969, "step": 4248 }, { "epoch": 0.77, "learning_rate": 1.8480932586300382e-05, "loss": 1.2343, "step": 4249 }, { "epoch": 0.77, "learning_rate": 1.847822157961323e-05, "loss": 1.276, "step": 4250 }, { "epoch": 0.77, "learning_rate": 1.847551057292608e-05, "loss": 1.4366, "step": 4251 }, { "epoch": 0.77, "learning_rate": 1.847279956623893e-05, "loss": 0.4237, "step": 4252 }, { "epoch": 0.77, "learning_rate": 1.847008855955178e-05, "loss": 1.362, "step": 4253 }, { "epoch": 0.77, "learning_rate": 1.846737755286463e-05, "loss": 1.6428, "step": 4254 }, { "epoch": 0.77, "learning_rate": 1.846466654617748e-05, "loss": 0.9694, "step": 4255 }, { "epoch": 0.77, "learning_rate": 1.846195553949033e-05, "loss": 1.0922, "step": 4256 }, { "epoch": 0.77, "learning_rate": 1.8459244532803183e-05, "loss": 0.5934, "step": 4257 }, { "epoch": 0.77, "learning_rate": 1.8456533526116032e-05, "loss": 0.5442, "step": 4258 }, { "epoch": 0.77, "learning_rate": 1.845382251942888e-05, "loss": 1.1124, "step": 4259 }, { "epoch": 0.77, "learning_rate": 1.8451111512741732e-05, "loss": 1.4212, "step": 4260 }, { "epoch": 0.77, "learning_rate": 1.844840050605458e-05, "loss": 0.8321, "step": 4261 }, { "epoch": 0.77, "learning_rate": 1.8445689499367432e-05, "loss": 1.4919, "step": 4262 }, { "epoch": 0.77, "learning_rate": 1.8442978492680284e-05, "loss": 0.5961, "step": 4263 }, { "epoch": 0.77, "learning_rate": 1.8440267485993133e-05, "loss": 1.4431, "step": 4264 }, { "epoch": 0.77, "learning_rate": 1.8437556479305984e-05, "loss": 0.8581, "step": 4265 }, { "epoch": 0.77, "learning_rate": 1.8434845472618833e-05, "loss": 1.1713, "step": 4266 }, { "epoch": 0.77, "learning_rate": 1.843213446593168e-05, "loss": 0.7302, "step": 4267 }, { "epoch": 0.77, "learning_rate": 1.8429423459244533e-05, "loss": 0.7104, "step": 4268 }, { "epoch": 0.77, "learning_rate": 1.842671245255738e-05, "loss": 0.739, "step": 4269 }, { "epoch": 0.77, "learning_rate": 1.8424001445870233e-05, "loss": 1.0571, "step": 4270 }, { "epoch": 0.77, "learning_rate": 1.8421290439183085e-05, "loss": 1.122, "step": 4271 }, { "epoch": 0.77, "learning_rate": 1.8418579432495934e-05, "loss": 1.2705, "step": 4272 }, { "epoch": 0.77, "learning_rate": 1.8415868425808785e-05, "loss": 0.7409, "step": 4273 }, { "epoch": 0.77, "learning_rate": 1.8413157419121637e-05, "loss": 1.6146, "step": 4274 }, { "epoch": 0.77, "learning_rate": 1.8410446412434482e-05, "loss": 1.2217, "step": 4275 }, { "epoch": 0.77, "learning_rate": 1.8407735405747334e-05, "loss": 1.4253, "step": 4276 }, { "epoch": 0.77, "learning_rate": 1.8405024399060183e-05, "loss": 1.1224, "step": 4277 }, { "epoch": 0.77, "learning_rate": 1.8402313392373034e-05, "loss": 0.7487, "step": 4278 }, { "epoch": 0.77, "learning_rate": 1.8399602385685886e-05, "loss": 1.4264, "step": 4279 }, { "epoch": 0.77, "learning_rate": 1.8396891378998735e-05, "loss": 1.1208, "step": 4280 }, { "epoch": 0.77, "learning_rate": 1.8394180372311587e-05, "loss": 1.3596, "step": 4281 }, { "epoch": 0.77, "learning_rate": 1.839146936562444e-05, "loss": 1.268, "step": 4282 }, { "epoch": 0.77, "learning_rate": 1.8388758358937283e-05, "loss": 0.5843, "step": 4283 }, { "epoch": 0.77, "learning_rate": 1.8386047352250135e-05, "loss": 1.3829, "step": 4284 }, { "epoch": 0.77, "learning_rate": 1.8383336345562987e-05, "loss": 1.2514, "step": 4285 }, { "epoch": 0.77, "learning_rate": 1.8380625338875836e-05, "loss": 1.3679, "step": 4286 }, { "epoch": 0.77, "learning_rate": 1.8377914332188687e-05, "loss": 1.0779, "step": 4287 }, { "epoch": 0.77, "learning_rate": 1.8375203325501536e-05, "loss": 1.1515, "step": 4288 }, { "epoch": 0.78, "learning_rate": 1.8372492318814388e-05, "loss": 1.5945, "step": 4289 }, { "epoch": 0.78, "learning_rate": 1.836978131212724e-05, "loss": 1.1297, "step": 4290 }, { "epoch": 0.78, "learning_rate": 1.8367070305440084e-05, "loss": 0.8146, "step": 4291 }, { "epoch": 0.78, "learning_rate": 1.8364359298752936e-05, "loss": 0.5189, "step": 4292 }, { "epoch": 0.78, "learning_rate": 1.8361648292065788e-05, "loss": 1.027, "step": 4293 }, { "epoch": 0.78, "learning_rate": 1.8358937285378637e-05, "loss": 1.3715, "step": 4294 }, { "epoch": 0.78, "learning_rate": 1.835622627869149e-05, "loss": 0.6385, "step": 4295 }, { "epoch": 0.78, "learning_rate": 1.835351527200434e-05, "loss": 1.1151, "step": 4296 }, { "epoch": 0.78, "learning_rate": 1.835080426531719e-05, "loss": 0.9086, "step": 4297 }, { "epoch": 0.78, "learning_rate": 1.834809325863004e-05, "loss": 0.9816, "step": 4298 }, { "epoch": 0.78, "learning_rate": 1.8345382251942886e-05, "loss": 0.6504, "step": 4299 }, { "epoch": 0.78, "learning_rate": 1.8342671245255737e-05, "loss": 0.8519, "step": 4300 }, { "epoch": 0.78, "learning_rate": 1.833996023856859e-05, "loss": 0.6331, "step": 4301 }, { "epoch": 0.78, "learning_rate": 1.8337249231881438e-05, "loss": 1.629, "step": 4302 }, { "epoch": 0.78, "learning_rate": 1.833453822519429e-05, "loss": 1.1911, "step": 4303 }, { "epoch": 0.78, "learning_rate": 1.833182721850714e-05, "loss": 1.9639, "step": 4304 }, { "epoch": 0.78, "learning_rate": 1.832911621181999e-05, "loss": 1.0587, "step": 4305 }, { "epoch": 0.78, "learning_rate": 1.832640520513284e-05, "loss": 0.8219, "step": 4306 }, { "epoch": 0.78, "learning_rate": 1.832369419844569e-05, "loss": 1.5406, "step": 4307 }, { "epoch": 0.78, "learning_rate": 1.832098319175854e-05, "loss": 1.138, "step": 4308 }, { "epoch": 0.78, "learning_rate": 1.831827218507139e-05, "loss": 1.0084, "step": 4309 }, { "epoch": 0.78, "learning_rate": 1.831556117838424e-05, "loss": 1.0953, "step": 4310 }, { "epoch": 0.78, "learning_rate": 1.831285017169709e-05, "loss": 0.8653, "step": 4311 }, { "epoch": 0.78, "learning_rate": 1.8310139165009942e-05, "loss": 1.1015, "step": 4312 }, { "epoch": 0.78, "learning_rate": 1.830742815832279e-05, "loss": 0.7257, "step": 4313 }, { "epoch": 0.78, "learning_rate": 1.8304717151635643e-05, "loss": 1.2965, "step": 4314 }, { "epoch": 0.78, "learning_rate": 1.830200614494849e-05, "loss": 1.2147, "step": 4315 }, { "epoch": 0.78, "learning_rate": 1.829929513826134e-05, "loss": 0.5823, "step": 4316 }, { "epoch": 0.78, "learning_rate": 1.829658413157419e-05, "loss": 1.0227, "step": 4317 }, { "epoch": 0.78, "learning_rate": 1.8293873124887043e-05, "loss": 0.9248, "step": 4318 }, { "epoch": 0.78, "learning_rate": 1.829116211819989e-05, "loss": 0.8707, "step": 4319 }, { "epoch": 0.78, "learning_rate": 1.8288451111512743e-05, "loss": 2.2306, "step": 4320 }, { "epoch": 0.78, "learning_rate": 1.8285740104825592e-05, "loss": 0.5393, "step": 4321 }, { "epoch": 0.78, "learning_rate": 1.8283029098138444e-05, "loss": 0.9625, "step": 4322 }, { "epoch": 0.78, "learning_rate": 1.8280318091451292e-05, "loss": 0.8994, "step": 4323 }, { "epoch": 0.78, "learning_rate": 1.827760708476414e-05, "loss": 0.6064, "step": 4324 }, { "epoch": 0.78, "learning_rate": 1.8274896078076992e-05, "loss": 2.0444, "step": 4325 }, { "epoch": 0.78, "learning_rate": 1.8272185071389844e-05, "loss": 0.9139, "step": 4326 }, { "epoch": 0.78, "learning_rate": 1.8269474064702693e-05, "loss": 0.8183, "step": 4327 }, { "epoch": 0.78, "learning_rate": 1.8266763058015545e-05, "loss": 0.8431, "step": 4328 }, { "epoch": 0.78, "learning_rate": 1.8264052051328396e-05, "loss": 1.4326, "step": 4329 }, { "epoch": 0.78, "learning_rate": 1.8261341044641245e-05, "loss": 0.8422, "step": 4330 }, { "epoch": 0.78, "learning_rate": 1.8258630037954093e-05, "loss": 1.0096, "step": 4331 }, { "epoch": 0.78, "learning_rate": 1.825591903126694e-05, "loss": 0.6654, "step": 4332 }, { "epoch": 0.78, "learning_rate": 1.8253208024579794e-05, "loss": 1.3655, "step": 4333 }, { "epoch": 0.78, "learning_rate": 1.8250497017892645e-05, "loss": 0.6697, "step": 4334 }, { "epoch": 0.78, "learning_rate": 1.8247786011205494e-05, "loss": 1.186, "step": 4335 }, { "epoch": 0.78, "learning_rate": 1.8245075004518346e-05, "loss": 0.909, "step": 4336 }, { "epoch": 0.78, "learning_rate": 1.8242363997831197e-05, "loss": 0.9414, "step": 4337 }, { "epoch": 0.78, "learning_rate": 1.8239652991144046e-05, "loss": 0.77, "step": 4338 }, { "epoch": 0.78, "learning_rate": 1.8236941984456894e-05, "loss": 1.0151, "step": 4339 }, { "epoch": 0.78, "learning_rate": 1.8234230977769746e-05, "loss": 0.8142, "step": 4340 }, { "epoch": 0.78, "learning_rate": 1.8231519971082595e-05, "loss": 2.0465, "step": 4341 }, { "epoch": 0.78, "learning_rate": 1.8228808964395446e-05, "loss": 1.122, "step": 4342 }, { "epoch": 0.78, "learning_rate": 1.8226097957708295e-05, "loss": 1.3653, "step": 4343 }, { "epoch": 0.79, "learning_rate": 1.8223386951021147e-05, "loss": 1.2649, "step": 4344 }, { "epoch": 0.79, "learning_rate": 1.8220675944334e-05, "loss": 0.5998, "step": 4345 }, { "epoch": 0.79, "learning_rate": 1.8217964937646847e-05, "loss": 1.5289, "step": 4346 }, { "epoch": 0.79, "learning_rate": 1.8215253930959695e-05, "loss": 0.8517, "step": 4347 }, { "epoch": 0.79, "learning_rate": 1.8212542924272547e-05, "loss": 0.8784, "step": 4348 }, { "epoch": 0.79, "learning_rate": 1.8209831917585396e-05, "loss": 0.565, "step": 4349 }, { "epoch": 0.79, "learning_rate": 1.8207120910898247e-05, "loss": 0.6475, "step": 4350 }, { "epoch": 0.79, "learning_rate": 1.82044099042111e-05, "loss": 1.147, "step": 4351 }, { "epoch": 0.79, "learning_rate": 1.8201698897523948e-05, "loss": 0.5739, "step": 4352 }, { "epoch": 0.79, "learning_rate": 1.81989878908368e-05, "loss": 0.9452, "step": 4353 }, { "epoch": 0.79, "learning_rate": 1.8196276884149648e-05, "loss": 0.9651, "step": 4354 }, { "epoch": 0.79, "learning_rate": 1.8193565877462496e-05, "loss": 0.8367, "step": 4355 }, { "epoch": 0.79, "learning_rate": 1.8190854870775348e-05, "loss": 0.547, "step": 4356 }, { "epoch": 0.79, "learning_rate": 1.8188143864088197e-05, "loss": 1.0891, "step": 4357 }, { "epoch": 0.79, "learning_rate": 1.818543285740105e-05, "loss": 1.2991, "step": 4358 }, { "epoch": 0.79, "learning_rate": 1.81827218507139e-05, "loss": 1.1247, "step": 4359 }, { "epoch": 0.79, "learning_rate": 1.818001084402675e-05, "loss": 0.9976, "step": 4360 }, { "epoch": 0.79, "learning_rate": 1.81772998373396e-05, "loss": 1.6211, "step": 4361 }, { "epoch": 0.79, "learning_rate": 1.8174588830652452e-05, "loss": 0.76, "step": 4362 }, { "epoch": 0.79, "learning_rate": 1.8171877823965298e-05, "loss": 0.8031, "step": 4363 }, { "epoch": 0.79, "learning_rate": 1.816916681727815e-05, "loss": 1.2614, "step": 4364 }, { "epoch": 0.79, "learning_rate": 1.8166455810590998e-05, "loss": 1.4244, "step": 4365 }, { "epoch": 0.79, "learning_rate": 1.816374480390385e-05, "loss": 1.2054, "step": 4366 }, { "epoch": 0.79, "learning_rate": 1.81610337972167e-05, "loss": 0.8935, "step": 4367 }, { "epoch": 0.79, "learning_rate": 1.815832279052955e-05, "loss": 0.2763, "step": 4368 }, { "epoch": 0.79, "learning_rate": 1.8155611783842402e-05, "loss": 0.6614, "step": 4369 }, { "epoch": 0.79, "learning_rate": 1.8152900777155254e-05, "loss": 1.2757, "step": 4370 }, { "epoch": 0.79, "learning_rate": 1.81501897704681e-05, "loss": 0.9526, "step": 4371 }, { "epoch": 0.79, "learning_rate": 1.814747876378095e-05, "loss": 1.472, "step": 4372 }, { "epoch": 0.79, "learning_rate": 1.8144767757093802e-05, "loss": 0.6655, "step": 4373 }, { "epoch": 0.79, "learning_rate": 1.814205675040665e-05, "loss": 0.6632, "step": 4374 }, { "epoch": 0.79, "learning_rate": 1.8139345743719503e-05, "loss": 0.4186, "step": 4375 }, { "epoch": 0.79, "learning_rate": 1.813663473703235e-05, "loss": 2.4365, "step": 4376 }, { "epoch": 0.79, "learning_rate": 1.8133923730345203e-05, "loss": 0.9592, "step": 4377 }, { "epoch": 0.79, "learning_rate": 1.813121272365805e-05, "loss": 1.6827, "step": 4378 }, { "epoch": 0.79, "learning_rate": 1.81285017169709e-05, "loss": 0.96, "step": 4379 }, { "epoch": 0.79, "learning_rate": 1.812579071028375e-05, "loss": 0.8843, "step": 4380 }, { "epoch": 0.79, "learning_rate": 1.8123079703596603e-05, "loss": 1.1583, "step": 4381 }, { "epoch": 0.79, "learning_rate": 1.8120368696909452e-05, "loss": 0.617, "step": 4382 }, { "epoch": 0.79, "learning_rate": 1.8117657690222304e-05, "loss": 0.5268, "step": 4383 }, { "epoch": 0.79, "learning_rate": 1.8114946683535155e-05, "loss": 1.2809, "step": 4384 }, { "epoch": 0.79, "learning_rate": 1.8112235676848004e-05, "loss": 0.8942, "step": 4385 }, { "epoch": 0.79, "learning_rate": 1.8109524670160852e-05, "loss": 0.9922, "step": 4386 }, { "epoch": 0.79, "learning_rate": 1.8106813663473704e-05, "loss": 1.1683, "step": 4387 }, { "epoch": 0.79, "learning_rate": 1.8104102656786553e-05, "loss": 0.86, "step": 4388 }, { "epoch": 0.79, "learning_rate": 1.8101391650099404e-05, "loss": 0.7063, "step": 4389 }, { "epoch": 0.79, "learning_rate": 1.8098680643412253e-05, "loss": 0.9559, "step": 4390 }, { "epoch": 0.79, "learning_rate": 1.8095969636725105e-05, "loss": 1.4254, "step": 4391 }, { "epoch": 0.79, "learning_rate": 1.8093258630037957e-05, "loss": 1.2093, "step": 4392 }, { "epoch": 0.79, "learning_rate": 1.8090547623350805e-05, "loss": 0.8174, "step": 4393 }, { "epoch": 0.79, "learning_rate": 1.8087836616663653e-05, "loss": 1.2157, "step": 4394 }, { "epoch": 0.79, "learning_rate": 1.8085125609976505e-05, "loss": 1.2371, "step": 4395 }, { "epoch": 0.79, "learning_rate": 1.8082414603289354e-05, "loss": 1.7108, "step": 4396 }, { "epoch": 0.79, "learning_rate": 1.8079703596602205e-05, "loss": 0.8066, "step": 4397 }, { "epoch": 0.79, "learning_rate": 1.8076992589915057e-05, "loss": 1.1828, "step": 4398 }, { "epoch": 0.8, "learning_rate": 1.8074281583227906e-05, "loss": 1.27, "step": 4399 }, { "epoch": 0.8, "learning_rate": 1.8071570576540758e-05, "loss": 0.6907, "step": 4400 }, { "epoch": 0.8, "learning_rate": 1.8068859569853606e-05, "loss": 0.9107, "step": 4401 }, { "epoch": 0.8, "learning_rate": 1.8066148563166454e-05, "loss": 0.7045, "step": 4402 }, { "epoch": 0.8, "learning_rate": 1.8063437556479306e-05, "loss": 1.0304, "step": 4403 }, { "epoch": 0.8, "learning_rate": 1.8060726549792155e-05, "loss": 1.2366, "step": 4404 }, { "epoch": 0.8, "learning_rate": 1.8058015543105007e-05, "loss": 0.8061, "step": 4405 }, { "epoch": 0.8, "learning_rate": 1.805530453641786e-05, "loss": 1.1401, "step": 4406 }, { "epoch": 0.8, "learning_rate": 1.8052593529730707e-05, "loss": 1.1435, "step": 4407 }, { "epoch": 0.8, "learning_rate": 1.804988252304356e-05, "loss": 1.3847, "step": 4408 }, { "epoch": 0.8, "learning_rate": 1.804717151635641e-05, "loss": 1.6655, "step": 4409 }, { "epoch": 0.8, "learning_rate": 1.8044460509669256e-05, "loss": 1.1093, "step": 4410 }, { "epoch": 0.8, "learning_rate": 1.8041749502982107e-05, "loss": 0.9479, "step": 4411 }, { "epoch": 0.8, "learning_rate": 1.8039038496294956e-05, "loss": 1.7696, "step": 4412 }, { "epoch": 0.8, "learning_rate": 1.8036327489607808e-05, "loss": 0.7989, "step": 4413 }, { "epoch": 0.8, "learning_rate": 1.803361648292066e-05, "loss": 0.9169, "step": 4414 }, { "epoch": 0.8, "learning_rate": 1.8030905476233508e-05, "loss": 1.7174, "step": 4415 }, { "epoch": 0.8, "learning_rate": 1.802819446954636e-05, "loss": 1.2347, "step": 4416 }, { "epoch": 0.8, "learning_rate": 1.802548346285921e-05, "loss": 0.7493, "step": 4417 }, { "epoch": 0.8, "learning_rate": 1.8022772456172057e-05, "loss": 0.7018, "step": 4418 }, { "epoch": 0.8, "learning_rate": 1.802006144948491e-05, "loss": 1.2641, "step": 4419 }, { "epoch": 0.8, "learning_rate": 1.801735044279776e-05, "loss": 0.8615, "step": 4420 }, { "epoch": 0.8, "learning_rate": 1.801463943611061e-05, "loss": 0.7981, "step": 4421 }, { "epoch": 0.8, "learning_rate": 1.801192842942346e-05, "loss": 1.2107, "step": 4422 }, { "epoch": 0.8, "learning_rate": 1.800921742273631e-05, "loss": 1.381, "step": 4423 }, { "epoch": 0.8, "learning_rate": 1.800650641604916e-05, "loss": 1.0947, "step": 4424 }, { "epoch": 0.8, "learning_rate": 1.8003795409362013e-05, "loss": 1.1983, "step": 4425 }, { "epoch": 0.8, "learning_rate": 1.8001084402674858e-05, "loss": 0.8163, "step": 4426 }, { "epoch": 0.8, "learning_rate": 1.799837339598771e-05, "loss": 1.188, "step": 4427 }, { "epoch": 0.8, "learning_rate": 1.799566238930056e-05, "loss": 0.3761, "step": 4428 }, { "epoch": 0.8, "learning_rate": 1.799295138261341e-05, "loss": 0.9463, "step": 4429 }, { "epoch": 0.8, "learning_rate": 1.799024037592626e-05, "loss": 0.8172, "step": 4430 }, { "epoch": 0.8, "learning_rate": 1.7987529369239113e-05, "loss": 0.6889, "step": 4431 }, { "epoch": 0.8, "learning_rate": 1.7984818362551962e-05, "loss": 1.0304, "step": 4432 }, { "epoch": 0.8, "learning_rate": 1.7982107355864814e-05, "loss": 0.591, "step": 4433 }, { "epoch": 0.8, "learning_rate": 1.797939634917766e-05, "loss": 1.4121, "step": 4434 }, { "epoch": 0.8, "learning_rate": 1.797668534249051e-05, "loss": 0.7765, "step": 4435 }, { "epoch": 0.8, "learning_rate": 1.7973974335803362e-05, "loss": 0.6924, "step": 4436 }, { "epoch": 0.8, "learning_rate": 1.797126332911621e-05, "loss": 1.1338, "step": 4437 }, { "epoch": 0.8, "learning_rate": 1.7968552322429063e-05, "loss": 1.4003, "step": 4438 }, { "epoch": 0.8, "learning_rate": 1.7965841315741915e-05, "loss": 1.3175, "step": 4439 }, { "epoch": 0.8, "learning_rate": 1.7963130309054763e-05, "loss": 0.718, "step": 4440 }, { "epoch": 0.8, "learning_rate": 1.7960419302367615e-05, "loss": 2.0737, "step": 4441 }, { "epoch": 0.8, "learning_rate": 1.7957708295680463e-05, "loss": 0.526, "step": 4442 }, { "epoch": 0.8, "learning_rate": 1.795499728899331e-05, "loss": 1.2347, "step": 4443 }, { "epoch": 0.8, "learning_rate": 1.7952286282306163e-05, "loss": 0.8449, "step": 4444 }, { "epoch": 0.8, "learning_rate": 1.7949575275619012e-05, "loss": 0.6274, "step": 4445 }, { "epoch": 0.8, "learning_rate": 1.7946864268931864e-05, "loss": 0.9191, "step": 4446 }, { "epoch": 0.8, "learning_rate": 1.7944153262244716e-05, "loss": 1.2571, "step": 4447 }, { "epoch": 0.8, "learning_rate": 1.7941442255557564e-05, "loss": 0.9457, "step": 4448 }, { "epoch": 0.8, "learning_rate": 1.7938731248870416e-05, "loss": 0.7726, "step": 4449 }, { "epoch": 0.8, "learning_rate": 1.7936020242183264e-05, "loss": 1.1323, "step": 4450 }, { "epoch": 0.8, "learning_rate": 1.7933309235496113e-05, "loss": 1.0855, "step": 4451 }, { "epoch": 0.8, "learning_rate": 1.7930598228808965e-05, "loss": 0.4221, "step": 4452 }, { "epoch": 0.8, "learning_rate": 1.7927887222121816e-05, "loss": 0.7875, "step": 4453 }, { "epoch": 0.8, "learning_rate": 1.7925176215434665e-05, "loss": 0.7797, "step": 4454 }, { "epoch": 0.81, "learning_rate": 1.7922465208747517e-05, "loss": 1.2405, "step": 4455 }, { "epoch": 0.81, "learning_rate": 1.7919754202060365e-05, "loss": 1.2233, "step": 4456 }, { "epoch": 0.81, "learning_rate": 1.7917043195373217e-05, "loss": 1.1371, "step": 4457 }, { "epoch": 0.81, "learning_rate": 1.7914332188686065e-05, "loss": 1.5776, "step": 4458 }, { "epoch": 0.81, "learning_rate": 1.7911621181998914e-05, "loss": 1.263, "step": 4459 }, { "epoch": 0.81, "learning_rate": 1.7908910175311766e-05, "loss": 1.3746, "step": 4460 }, { "epoch": 0.81, "learning_rate": 1.7906199168624617e-05, "loss": 0.8386, "step": 4461 }, { "epoch": 0.81, "learning_rate": 1.7903488161937466e-05, "loss": 1.2925, "step": 4462 }, { "epoch": 0.81, "learning_rate": 1.7900777155250318e-05, "loss": 1.5016, "step": 4463 }, { "epoch": 0.81, "learning_rate": 1.789806614856317e-05, "loss": 0.9947, "step": 4464 }, { "epoch": 0.81, "learning_rate": 1.7895355141876018e-05, "loss": 1.0871, "step": 4465 }, { "epoch": 0.81, "learning_rate": 1.7892644135188866e-05, "loss": 0.8877, "step": 4466 }, { "epoch": 0.81, "learning_rate": 1.7889933128501715e-05, "loss": 0.537, "step": 4467 }, { "epoch": 0.81, "learning_rate": 1.7887222121814567e-05, "loss": 1.2409, "step": 4468 }, { "epoch": 0.81, "learning_rate": 1.788451111512742e-05, "loss": 0.7719, "step": 4469 }, { "epoch": 0.81, "learning_rate": 1.7881800108440267e-05, "loss": 1.5367, "step": 4470 }, { "epoch": 0.81, "learning_rate": 1.787908910175312e-05, "loss": 0.8905, "step": 4471 }, { "epoch": 0.81, "learning_rate": 1.787637809506597e-05, "loss": 0.5374, "step": 4472 }, { "epoch": 0.81, "learning_rate": 1.787366708837882e-05, "loss": 0.4875, "step": 4473 }, { "epoch": 0.81, "learning_rate": 1.7870956081691668e-05, "loss": 0.6507, "step": 4474 }, { "epoch": 0.81, "learning_rate": 1.786824507500452e-05, "loss": 1.0189, "step": 4475 }, { "epoch": 0.81, "learning_rate": 1.7865534068317368e-05, "loss": 1.069, "step": 4476 }, { "epoch": 0.81, "learning_rate": 1.786282306163022e-05, "loss": 1.1652, "step": 4477 }, { "epoch": 0.81, "learning_rate": 1.7860112054943068e-05, "loss": 1.7269, "step": 4478 }, { "epoch": 0.81, "learning_rate": 1.785740104825592e-05, "loss": 0.9838, "step": 4479 }, { "epoch": 0.81, "learning_rate": 1.7854690041568772e-05, "loss": 1.3521, "step": 4480 }, { "epoch": 0.81, "learning_rate": 1.785197903488162e-05, "loss": 1.1756, "step": 4481 }, { "epoch": 0.81, "learning_rate": 1.784926802819447e-05, "loss": 1.7456, "step": 4482 }, { "epoch": 0.81, "learning_rate": 1.784655702150732e-05, "loss": 1.0554, "step": 4483 }, { "epoch": 0.81, "learning_rate": 1.784384601482017e-05, "loss": 1.8781, "step": 4484 }, { "epoch": 0.81, "learning_rate": 1.784113500813302e-05, "loss": 1.0892, "step": 4485 }, { "epoch": 0.81, "learning_rate": 1.7838424001445873e-05, "loss": 1.1143, "step": 4486 }, { "epoch": 0.81, "learning_rate": 1.783571299475872e-05, "loss": 1.8474, "step": 4487 }, { "epoch": 0.81, "learning_rate": 1.7833001988071573e-05, "loss": 1.4579, "step": 4488 }, { "epoch": 0.81, "learning_rate": 1.783029098138442e-05, "loss": 1.4139, "step": 4489 }, { "epoch": 0.81, "learning_rate": 1.782757997469727e-05, "loss": 1.4583, "step": 4490 }, { "epoch": 0.81, "learning_rate": 1.782486896801012e-05, "loss": 1.4062, "step": 4491 }, { "epoch": 0.81, "learning_rate": 1.782215796132297e-05, "loss": 0.8195, "step": 4492 }, { "epoch": 0.81, "learning_rate": 1.7819446954635822e-05, "loss": 1.3321, "step": 4493 }, { "epoch": 0.81, "learning_rate": 1.7816735947948674e-05, "loss": 0.825, "step": 4494 }, { "epoch": 0.81, "learning_rate": 1.7814024941261522e-05, "loss": 0.9267, "step": 4495 }, { "epoch": 0.81, "learning_rate": 1.7811313934574374e-05, "loss": 0.9442, "step": 4496 }, { "epoch": 0.81, "learning_rate": 1.7808602927887226e-05, "loss": 0.6047, "step": 4497 }, { "epoch": 0.81, "learning_rate": 1.780589192120007e-05, "loss": 1.7276, "step": 4498 }, { "epoch": 0.81, "learning_rate": 1.7803180914512923e-05, "loss": 1.2077, "step": 4499 }, { "epoch": 0.81, "learning_rate": 1.780046990782577e-05, "loss": 1.3754, "step": 4500 }, { "epoch": 0.81, "learning_rate": 1.7797758901138623e-05, "loss": 0.8203, "step": 4501 }, { "epoch": 0.81, "learning_rate": 1.7795047894451475e-05, "loss": 1.5838, "step": 4502 }, { "epoch": 0.81, "learning_rate": 1.7792336887764323e-05, "loss": 1.3027, "step": 4503 }, { "epoch": 0.81, "learning_rate": 1.7789625881077175e-05, "loss": 1.1864, "step": 4504 }, { "epoch": 0.81, "learning_rate": 1.7786914874390023e-05, "loss": 0.6006, "step": 4505 }, { "epoch": 0.81, "learning_rate": 1.7784203867702872e-05, "loss": 1.6358, "step": 4506 }, { "epoch": 0.81, "learning_rate": 1.7781492861015724e-05, "loss": 1.0484, "step": 4507 }, { "epoch": 0.81, "learning_rate": 1.7778781854328575e-05, "loss": 1.4896, "step": 4508 }, { "epoch": 0.81, "learning_rate": 1.7776070847641424e-05, "loss": 0.9935, "step": 4509 }, { "epoch": 0.82, "learning_rate": 1.7773359840954276e-05, "loss": 1.0999, "step": 4510 }, { "epoch": 0.82, "learning_rate": 1.7770648834267124e-05, "loss": 0.7235, "step": 4511 }, { "epoch": 0.82, "learning_rate": 1.7767937827579976e-05, "loss": 0.9316, "step": 4512 }, { "epoch": 0.82, "learning_rate": 1.7765226820892824e-05, "loss": 0.9655, "step": 4513 }, { "epoch": 0.82, "learning_rate": 1.7762515814205673e-05, "loss": 1.194, "step": 4514 }, { "epoch": 0.82, "learning_rate": 1.7759804807518525e-05, "loss": 0.7268, "step": 4515 }, { "epoch": 0.82, "learning_rate": 1.7757093800831377e-05, "loss": 1.3662, "step": 4516 }, { "epoch": 0.82, "learning_rate": 1.7754382794144225e-05, "loss": 0.7376, "step": 4517 }, { "epoch": 0.82, "learning_rate": 1.7751671787457077e-05, "loss": 1.0793, "step": 4518 }, { "epoch": 0.82, "learning_rate": 1.774896078076993e-05, "loss": 1.1982, "step": 4519 }, { "epoch": 0.82, "learning_rate": 1.7746249774082777e-05, "loss": 0.4049, "step": 4520 }, { "epoch": 0.82, "learning_rate": 1.7743538767395626e-05, "loss": 1.0116, "step": 4521 }, { "epoch": 0.82, "learning_rate": 1.7740827760708477e-05, "loss": 0.9908, "step": 4522 }, { "epoch": 0.82, "learning_rate": 1.7738116754021326e-05, "loss": 1.0823, "step": 4523 }, { "epoch": 0.82, "learning_rate": 1.7735405747334178e-05, "loss": 1.2418, "step": 4524 }, { "epoch": 0.82, "learning_rate": 1.7732694740647026e-05, "loss": 0.4781, "step": 4525 }, { "epoch": 0.82, "learning_rate": 1.7729983733959878e-05, "loss": 0.8269, "step": 4526 }, { "epoch": 0.82, "learning_rate": 1.772727272727273e-05, "loss": 0.9127, "step": 4527 }, { "epoch": 0.82, "learning_rate": 1.7724561720585578e-05, "loss": 0.9925, "step": 4528 }, { "epoch": 0.82, "learning_rate": 1.7721850713898427e-05, "loss": 1.4721, "step": 4529 }, { "epoch": 0.82, "learning_rate": 1.771913970721128e-05, "loss": 0.7638, "step": 4530 }, { "epoch": 0.82, "learning_rate": 1.7716428700524127e-05, "loss": 0.9276, "step": 4531 }, { "epoch": 0.82, "learning_rate": 1.771371769383698e-05, "loss": 1.2598, "step": 4532 }, { "epoch": 0.82, "learning_rate": 1.771100668714983e-05, "loss": 0.4888, "step": 4533 }, { "epoch": 0.82, "learning_rate": 1.770829568046268e-05, "loss": 1.009, "step": 4534 }, { "epoch": 0.82, "learning_rate": 1.770558467377553e-05, "loss": 0.9235, "step": 4535 }, { "epoch": 0.82, "learning_rate": 1.770287366708838e-05, "loss": 1.4992, "step": 4536 }, { "epoch": 0.82, "learning_rate": 1.7700162660401228e-05, "loss": 0.4267, "step": 4537 }, { "epoch": 0.82, "learning_rate": 1.769745165371408e-05, "loss": 0.7024, "step": 4538 }, { "epoch": 0.82, "learning_rate": 1.7694740647026928e-05, "loss": 1.2894, "step": 4539 }, { "epoch": 0.82, "learning_rate": 1.769202964033978e-05, "loss": 0.9925, "step": 4540 }, { "epoch": 0.82, "learning_rate": 1.768931863365263e-05, "loss": 0.8499, "step": 4541 }, { "epoch": 0.82, "learning_rate": 1.768660762696548e-05, "loss": 1.0042, "step": 4542 }, { "epoch": 0.82, "learning_rate": 1.7683896620278332e-05, "loss": 1.618, "step": 4543 }, { "epoch": 0.82, "learning_rate": 1.7681185613591184e-05, "loss": 1.1203, "step": 4544 }, { "epoch": 0.82, "learning_rate": 1.767847460690403e-05, "loss": 0.9237, "step": 4545 }, { "epoch": 0.82, "learning_rate": 1.767576360021688e-05, "loss": 0.8653, "step": 4546 }, { "epoch": 0.82, "learning_rate": 1.767305259352973e-05, "loss": 1.5441, "step": 4547 }, { "epoch": 0.82, "learning_rate": 1.767034158684258e-05, "loss": 1.0454, "step": 4548 }, { "epoch": 0.82, "learning_rate": 1.7667630580155433e-05, "loss": 1.2437, "step": 4549 }, { "epoch": 0.82, "learning_rate": 1.766491957346828e-05, "loss": 0.6711, "step": 4550 }, { "epoch": 0.82, "learning_rate": 1.7662208566781133e-05, "loss": 1.0274, "step": 4551 }, { "epoch": 0.82, "learning_rate": 1.7659497560093985e-05, "loss": 1.3314, "step": 4552 }, { "epoch": 0.82, "learning_rate": 1.765678655340683e-05, "loss": 1.0322, "step": 4553 }, { "epoch": 0.82, "learning_rate": 1.765407554671968e-05, "loss": 0.9242, "step": 4554 }, { "epoch": 0.82, "learning_rate": 1.7651364540032533e-05, "loss": 0.9441, "step": 4555 }, { "epoch": 0.82, "learning_rate": 1.7648653533345382e-05, "loss": 0.8559, "step": 4556 }, { "epoch": 0.82, "learning_rate": 1.7645942526658234e-05, "loss": 1.3913, "step": 4557 }, { "epoch": 0.82, "learning_rate": 1.7643231519971082e-05, "loss": 1.2264, "step": 4558 }, { "epoch": 0.82, "learning_rate": 1.7640520513283934e-05, "loss": 0.8557, "step": 4559 }, { "epoch": 0.82, "learning_rate": 1.7637809506596786e-05, "loss": 1.5583, "step": 4560 }, { "epoch": 0.82, "learning_rate": 1.763509849990963e-05, "loss": 0.6354, "step": 4561 }, { "epoch": 0.82, "learning_rate": 1.7632387493222483e-05, "loss": 0.7293, "step": 4562 }, { "epoch": 0.82, "learning_rate": 1.7629676486535335e-05, "loss": 1.3512, "step": 4563 }, { "epoch": 0.82, "learning_rate": 1.7626965479848183e-05, "loss": 0.6288, "step": 4564 }, { "epoch": 0.83, "learning_rate": 1.7624254473161035e-05, "loss": 1.0428, "step": 4565 }, { "epoch": 0.83, "learning_rate": 1.7621543466473887e-05, "loss": 0.6989, "step": 4566 }, { "epoch": 0.83, "learning_rate": 1.7618832459786735e-05, "loss": 1.0504, "step": 4567 }, { "epoch": 0.83, "learning_rate": 1.7616121453099587e-05, "loss": 0.4123, "step": 4568 }, { "epoch": 0.83, "learning_rate": 1.7613410446412432e-05, "loss": 1.0068, "step": 4569 }, { "epoch": 0.83, "learning_rate": 1.7610699439725284e-05, "loss": 1.1343, "step": 4570 }, { "epoch": 0.83, "learning_rate": 1.7607988433038136e-05, "loss": 0.9047, "step": 4571 }, { "epoch": 0.83, "learning_rate": 1.7605277426350984e-05, "loss": 1.1231, "step": 4572 }, { "epoch": 0.83, "learning_rate": 1.7602566419663836e-05, "loss": 1.2359, "step": 4573 }, { "epoch": 0.83, "learning_rate": 1.7599855412976688e-05, "loss": 0.9265, "step": 4574 }, { "epoch": 0.83, "learning_rate": 1.7597144406289536e-05, "loss": 1.4307, "step": 4575 }, { "epoch": 0.83, "learning_rate": 1.7594433399602388e-05, "loss": 1.0356, "step": 4576 }, { "epoch": 0.83, "learning_rate": 1.7591722392915236e-05, "loss": 0.7226, "step": 4577 }, { "epoch": 0.83, "learning_rate": 1.7589011386228085e-05, "loss": 0.8735, "step": 4578 }, { "epoch": 0.83, "learning_rate": 1.7586300379540937e-05, "loss": 1.1502, "step": 4579 }, { "epoch": 0.83, "learning_rate": 1.7583589372853785e-05, "loss": 1.097, "step": 4580 }, { "epoch": 0.83, "learning_rate": 1.7580878366166637e-05, "loss": 1.1921, "step": 4581 }, { "epoch": 0.83, "learning_rate": 1.757816735947949e-05, "loss": 0.4772, "step": 4582 }, { "epoch": 0.83, "learning_rate": 1.7575456352792337e-05, "loss": 1.1642, "step": 4583 }, { "epoch": 0.83, "learning_rate": 1.757274534610519e-05, "loss": 0.8066, "step": 4584 }, { "epoch": 0.83, "learning_rate": 1.7570034339418038e-05, "loss": 0.8469, "step": 4585 }, { "epoch": 0.83, "learning_rate": 1.7567323332730886e-05, "loss": 0.8983, "step": 4586 }, { "epoch": 0.83, "learning_rate": 1.7564612326043738e-05, "loss": 0.9336, "step": 4587 }, { "epoch": 0.83, "learning_rate": 1.756190131935659e-05, "loss": 1.0043, "step": 4588 }, { "epoch": 0.83, "learning_rate": 1.7559190312669438e-05, "loss": 0.8821, "step": 4589 }, { "epoch": 0.83, "learning_rate": 1.755647930598229e-05, "loss": 1.2149, "step": 4590 }, { "epoch": 0.83, "learning_rate": 1.755376829929514e-05, "loss": 1.0645, "step": 4591 }, { "epoch": 0.83, "learning_rate": 1.755105729260799e-05, "loss": 0.708, "step": 4592 }, { "epoch": 0.83, "learning_rate": 1.754834628592084e-05, "loss": 1.005, "step": 4593 }, { "epoch": 0.83, "learning_rate": 1.7545635279233687e-05, "loss": 1.0898, "step": 4594 }, { "epoch": 0.83, "learning_rate": 1.754292427254654e-05, "loss": 1.6099, "step": 4595 }, { "epoch": 0.83, "learning_rate": 1.754021326585939e-05, "loss": 0.9139, "step": 4596 }, { "epoch": 0.83, "learning_rate": 1.753750225917224e-05, "loss": 0.7761, "step": 4597 }, { "epoch": 0.83, "learning_rate": 1.753479125248509e-05, "loss": 0.9851, "step": 4598 }, { "epoch": 0.83, "learning_rate": 1.7532080245797943e-05, "loss": 0.6468, "step": 4599 }, { "epoch": 0.83, "learning_rate": 1.752936923911079e-05, "loss": 0.6085, "step": 4600 }, { "epoch": 0.83, "learning_rate": 1.752665823242364e-05, "loss": 1.5047, "step": 4601 }, { "epoch": 0.83, "learning_rate": 1.7523947225736488e-05, "loss": 0.636, "step": 4602 }, { "epoch": 0.83, "learning_rate": 1.752123621904934e-05, "loss": 1.3126, "step": 4603 }, { "epoch": 0.83, "learning_rate": 1.7518525212362192e-05, "loss": 1.3003, "step": 4604 }, { "epoch": 0.83, "learning_rate": 1.751581420567504e-05, "loss": 1.1036, "step": 4605 }, { "epoch": 0.83, "learning_rate": 1.7513103198987892e-05, "loss": 0.7883, "step": 4606 }, { "epoch": 0.83, "learning_rate": 1.7510392192300744e-05, "loss": 0.8368, "step": 4607 }, { "epoch": 0.83, "learning_rate": 1.7507681185613592e-05, "loss": 1.2756, "step": 4608 }, { "epoch": 0.83, "learning_rate": 1.750497017892644e-05, "loss": 0.6788, "step": 4609 }, { "epoch": 0.83, "learning_rate": 1.7502259172239293e-05, "loss": 1.4481, "step": 4610 }, { "epoch": 0.83, "learning_rate": 1.749954816555214e-05, "loss": 1.2022, "step": 4611 }, { "epoch": 0.83, "learning_rate": 1.7496837158864993e-05, "loss": 0.7114, "step": 4612 }, { "epoch": 0.83, "learning_rate": 1.749412615217784e-05, "loss": 0.9125, "step": 4613 }, { "epoch": 0.83, "learning_rate": 1.7491415145490693e-05, "loss": 0.8512, "step": 4614 }, { "epoch": 0.83, "learning_rate": 1.7488704138803545e-05, "loss": 1.3416, "step": 4615 }, { "epoch": 0.83, "learning_rate": 1.7485993132116393e-05, "loss": 1.4536, "step": 4616 }, { "epoch": 0.83, "learning_rate": 1.7483282125429242e-05, "loss": 0.9174, "step": 4617 }, { "epoch": 0.83, "learning_rate": 1.7480571118742094e-05, "loss": 0.9589, "step": 4618 }, { "epoch": 0.83, "learning_rate": 1.7477860112054942e-05, "loss": 1.4604, "step": 4619 }, { "epoch": 0.83, "learning_rate": 1.7475149105367794e-05, "loss": 1.9218, "step": 4620 }, { "epoch": 0.84, "learning_rate": 1.7472438098680646e-05, "loss": 1.0021, "step": 4621 }, { "epoch": 0.84, "learning_rate": 1.7469727091993494e-05, "loss": 1.3787, "step": 4622 }, { "epoch": 0.84, "learning_rate": 1.7467016085306346e-05, "loss": 0.5256, "step": 4623 }, { "epoch": 0.84, "learning_rate": 1.7464305078619194e-05, "loss": 1.5687, "step": 4624 }, { "epoch": 0.84, "learning_rate": 1.7461594071932043e-05, "loss": 1.1185, "step": 4625 }, { "epoch": 0.84, "learning_rate": 1.7458883065244895e-05, "loss": 1.241, "step": 4626 }, { "epoch": 0.84, "learning_rate": 1.7456172058557743e-05, "loss": 0.848, "step": 4627 }, { "epoch": 0.84, "learning_rate": 1.7453461051870595e-05, "loss": 0.9042, "step": 4628 }, { "epoch": 0.84, "learning_rate": 1.7450750045183447e-05, "loss": 1.4105, "step": 4629 }, { "epoch": 0.84, "learning_rate": 1.7448039038496295e-05, "loss": 0.9124, "step": 4630 }, { "epoch": 0.84, "learning_rate": 1.7445328031809147e-05, "loss": 1.1391, "step": 4631 }, { "epoch": 0.84, "learning_rate": 1.7442617025122e-05, "loss": 1.213, "step": 4632 }, { "epoch": 0.84, "learning_rate": 1.7439906018434844e-05, "loss": 0.7524, "step": 4633 }, { "epoch": 0.84, "learning_rate": 1.7437195011747696e-05, "loss": 1.1746, "step": 4634 }, { "epoch": 0.84, "learning_rate": 1.7434484005060544e-05, "loss": 0.4376, "step": 4635 }, { "epoch": 0.84, "learning_rate": 1.7431772998373396e-05, "loss": 1.1607, "step": 4636 }, { "epoch": 0.84, "learning_rate": 1.7429061991686248e-05, "loss": 0.6289, "step": 4637 }, { "epoch": 0.84, "learning_rate": 1.7426350984999096e-05, "loss": 1.1885, "step": 4638 }, { "epoch": 0.84, "learning_rate": 1.7423639978311948e-05, "loss": 0.8244, "step": 4639 }, { "epoch": 0.84, "learning_rate": 1.7420928971624797e-05, "loss": 1.3661, "step": 4640 }, { "epoch": 0.84, "learning_rate": 1.7418217964937645e-05, "loss": 1.1229, "step": 4641 }, { "epoch": 0.84, "learning_rate": 1.7415506958250497e-05, "loss": 1.6224, "step": 4642 }, { "epoch": 0.84, "learning_rate": 1.741279595156335e-05, "loss": 0.9307, "step": 4643 }, { "epoch": 0.84, "learning_rate": 1.7410084944876197e-05, "loss": 1.1467, "step": 4644 }, { "epoch": 0.84, "learning_rate": 1.740737393818905e-05, "loss": 1.0854, "step": 4645 }, { "epoch": 0.84, "learning_rate": 1.7404662931501897e-05, "loss": 0.84, "step": 4646 }, { "epoch": 0.84, "learning_rate": 1.740195192481475e-05, "loss": 0.9103, "step": 4647 }, { "epoch": 0.84, "learning_rate": 1.7399240918127598e-05, "loss": 0.9581, "step": 4648 }, { "epoch": 0.84, "learning_rate": 1.7396529911440446e-05, "loss": 1.0502, "step": 4649 }, { "epoch": 0.84, "learning_rate": 1.7393818904753298e-05, "loss": 0.8749, "step": 4650 }, { "epoch": 0.84, "learning_rate": 1.739110789806615e-05, "loss": 1.5861, "step": 4651 }, { "epoch": 0.84, "learning_rate": 1.7388396891378998e-05, "loss": 1.4079, "step": 4652 }, { "epoch": 0.84, "learning_rate": 1.738568588469185e-05, "loss": 1.3503, "step": 4653 }, { "epoch": 0.84, "learning_rate": 1.7382974878004702e-05, "loss": 0.5297, "step": 4654 }, { "epoch": 0.84, "learning_rate": 1.738026387131755e-05, "loss": 1.0633, "step": 4655 }, { "epoch": 0.84, "learning_rate": 1.73775528646304e-05, "loss": 1.0931, "step": 4656 }, { "epoch": 0.84, "learning_rate": 1.7374841857943247e-05, "loss": 0.8989, "step": 4657 }, { "epoch": 0.84, "learning_rate": 1.73721308512561e-05, "loss": 0.5541, "step": 4658 }, { "epoch": 0.84, "learning_rate": 1.736941984456895e-05, "loss": 1.65, "step": 4659 }, { "epoch": 0.84, "learning_rate": 1.73667088378818e-05, "loss": 0.7601, "step": 4660 }, { "epoch": 0.84, "learning_rate": 1.736399783119465e-05, "loss": 1.0323, "step": 4661 }, { "epoch": 0.84, "learning_rate": 1.7361286824507503e-05, "loss": 1.4239, "step": 4662 }, { "epoch": 0.84, "learning_rate": 1.735857581782035e-05, "loss": 1.0541, "step": 4663 }, { "epoch": 0.84, "learning_rate": 1.73558648111332e-05, "loss": 1.6723, "step": 4664 }, { "epoch": 0.84, "learning_rate": 1.735315380444605e-05, "loss": 1.0245, "step": 4665 }, { "epoch": 0.84, "learning_rate": 1.73504427977589e-05, "loss": 0.6822, "step": 4666 }, { "epoch": 0.84, "learning_rate": 1.7347731791071752e-05, "loss": 1.3602, "step": 4667 }, { "epoch": 0.84, "learning_rate": 1.73450207843846e-05, "loss": 1.0967, "step": 4668 }, { "epoch": 0.84, "learning_rate": 1.7342309777697452e-05, "loss": 0.7576, "step": 4669 }, { "epoch": 0.84, "learning_rate": 1.7339598771010304e-05, "loss": 1.3642, "step": 4670 }, { "epoch": 0.84, "learning_rate": 1.7336887764323152e-05, "loss": 1.0377, "step": 4671 }, { "epoch": 0.84, "learning_rate": 1.7334176757636e-05, "loss": 0.6459, "step": 4672 }, { "epoch": 0.84, "learning_rate": 1.7331465750948853e-05, "loss": 0.6122, "step": 4673 }, { "epoch": 0.84, "learning_rate": 1.73287547442617e-05, "loss": 1.1022, "step": 4674 }, { "epoch": 0.84, "learning_rate": 1.7326043737574553e-05, "loss": 1.2358, "step": 4675 }, { "epoch": 0.85, "learning_rate": 1.7323332730887405e-05, "loss": 0.6083, "step": 4676 }, { "epoch": 0.85, "learning_rate": 1.7320621724200253e-05, "loss": 0.3541, "step": 4677 }, { "epoch": 0.85, "learning_rate": 1.7317910717513105e-05, "loss": 0.7101, "step": 4678 }, { "epoch": 0.85, "learning_rate": 1.7315199710825954e-05, "loss": 1.2578, "step": 4679 }, { "epoch": 0.85, "learning_rate": 1.7312488704138802e-05, "loss": 0.7378, "step": 4680 }, { "epoch": 0.85, "learning_rate": 1.7309777697451654e-05, "loss": 1.7323, "step": 4681 }, { "epoch": 0.85, "learning_rate": 1.7307066690764502e-05, "loss": 1.116, "step": 4682 }, { "epoch": 0.85, "learning_rate": 1.7304355684077354e-05, "loss": 1.0687, "step": 4683 }, { "epoch": 0.85, "learning_rate": 1.7301644677390206e-05, "loss": 0.6903, "step": 4684 }, { "epoch": 0.85, "learning_rate": 1.7298933670703054e-05, "loss": 1.7352, "step": 4685 }, { "epoch": 0.85, "learning_rate": 1.7296222664015906e-05, "loss": 1.1499, "step": 4686 }, { "epoch": 0.85, "learning_rate": 1.7293511657328758e-05, "loss": 1.494, "step": 4687 }, { "epoch": 0.85, "learning_rate": 1.7290800650641603e-05, "loss": 1.1009, "step": 4688 }, { "epoch": 0.85, "learning_rate": 1.7288089643954455e-05, "loss": 1.0635, "step": 4689 }, { "epoch": 0.85, "learning_rate": 1.7285378637267307e-05, "loss": 1.5619, "step": 4690 }, { "epoch": 0.85, "learning_rate": 1.7282667630580155e-05, "loss": 1.3291, "step": 4691 }, { "epoch": 0.85, "learning_rate": 1.7279956623893007e-05, "loss": 0.6858, "step": 4692 }, { "epoch": 0.85, "learning_rate": 1.7277245617205855e-05, "loss": 1.765, "step": 4693 }, { "epoch": 0.85, "learning_rate": 1.7274534610518707e-05, "loss": 1.8316, "step": 4694 }, { "epoch": 0.85, "learning_rate": 1.727182360383156e-05, "loss": 0.8618, "step": 4695 }, { "epoch": 0.85, "learning_rate": 1.7269112597144404e-05, "loss": 0.8818, "step": 4696 }, { "epoch": 0.85, "learning_rate": 1.7266401590457256e-05, "loss": 0.7021, "step": 4697 }, { "epoch": 0.85, "learning_rate": 1.7263690583770108e-05, "loss": 0.7633, "step": 4698 }, { "epoch": 0.85, "learning_rate": 1.7260979577082956e-05, "loss": 0.7513, "step": 4699 }, { "epoch": 0.85, "learning_rate": 1.7258268570395808e-05, "loss": 0.5514, "step": 4700 }, { "epoch": 0.85, "learning_rate": 1.725555756370866e-05, "loss": 0.8506, "step": 4701 }, { "epoch": 0.85, "learning_rate": 1.7252846557021508e-05, "loss": 0.9183, "step": 4702 }, { "epoch": 0.85, "learning_rate": 1.725013555033436e-05, "loss": 0.6893, "step": 4703 }, { "epoch": 0.85, "learning_rate": 1.7247424543647205e-05, "loss": 1.3491, "step": 4704 }, { "epoch": 0.85, "learning_rate": 1.7244713536960057e-05, "loss": 0.8916, "step": 4705 }, { "epoch": 0.85, "learning_rate": 1.724200253027291e-05, "loss": 0.8347, "step": 4706 }, { "epoch": 0.85, "learning_rate": 1.7239291523585757e-05, "loss": 1.0367, "step": 4707 }, { "epoch": 0.85, "learning_rate": 1.723658051689861e-05, "loss": 1.1317, "step": 4708 }, { "epoch": 0.85, "learning_rate": 1.723386951021146e-05, "loss": 0.9048, "step": 4709 }, { "epoch": 0.85, "learning_rate": 1.723115850352431e-05, "loss": 1.8, "step": 4710 }, { "epoch": 0.85, "learning_rate": 1.722844749683716e-05, "loss": 1.6862, "step": 4711 }, { "epoch": 0.85, "learning_rate": 1.722573649015001e-05, "loss": 0.9499, "step": 4712 }, { "epoch": 0.85, "learning_rate": 1.7223025483462858e-05, "loss": 1.1415, "step": 4713 }, { "epoch": 0.85, "learning_rate": 1.722031447677571e-05, "loss": 1.0226, "step": 4714 }, { "epoch": 0.85, "learning_rate": 1.721760347008856e-05, "loss": 0.8851, "step": 4715 }, { "epoch": 0.85, "learning_rate": 1.721489246340141e-05, "loss": 0.7617, "step": 4716 }, { "epoch": 0.85, "learning_rate": 1.7212181456714262e-05, "loss": 1.0254, "step": 4717 }, { "epoch": 0.85, "learning_rate": 1.720947045002711e-05, "loss": 1.219, "step": 4718 }, { "epoch": 0.85, "learning_rate": 1.7206759443339962e-05, "loss": 0.8698, "step": 4719 }, { "epoch": 0.85, "learning_rate": 1.720404843665281e-05, "loss": 0.8, "step": 4720 }, { "epoch": 0.85, "learning_rate": 1.720133742996566e-05, "loss": 0.9874, "step": 4721 }, { "epoch": 0.85, "learning_rate": 1.719862642327851e-05, "loss": 0.6884, "step": 4722 }, { "epoch": 0.85, "learning_rate": 1.7195915416591363e-05, "loss": 1.1494, "step": 4723 }, { "epoch": 0.85, "learning_rate": 1.719320440990421e-05, "loss": 0.6089, "step": 4724 }, { "epoch": 0.85, "learning_rate": 1.7190493403217063e-05, "loss": 1.2596, "step": 4725 }, { "epoch": 0.85, "learning_rate": 1.718778239652991e-05, "loss": 0.8194, "step": 4726 }, { "epoch": 0.85, "learning_rate": 1.7185071389842763e-05, "loss": 0.9576, "step": 4727 }, { "epoch": 0.85, "learning_rate": 1.7182360383155612e-05, "loss": 0.7751, "step": 4728 }, { "epoch": 0.85, "learning_rate": 1.717964937646846e-05, "loss": 0.8206, "step": 4729 }, { "epoch": 0.85, "learning_rate": 1.7176938369781312e-05, "loss": 0.6249, "step": 4730 }, { "epoch": 0.86, "learning_rate": 1.7174227363094164e-05, "loss": 0.5834, "step": 4731 }, { "epoch": 0.86, "learning_rate": 1.7171516356407012e-05, "loss": 1.035, "step": 4732 }, { "epoch": 0.86, "learning_rate": 1.7168805349719864e-05, "loss": 0.4106, "step": 4733 }, { "epoch": 0.86, "learning_rate": 1.7166094343032716e-05, "loss": 1.2875, "step": 4734 }, { "epoch": 0.86, "learning_rate": 1.7163383336345564e-05, "loss": 0.3775, "step": 4735 }, { "epoch": 0.86, "learning_rate": 1.7160672329658413e-05, "loss": 0.5558, "step": 4736 }, { "epoch": 0.86, "learning_rate": 1.715796132297126e-05, "loss": 0.8299, "step": 4737 }, { "epoch": 0.86, "learning_rate": 1.7155250316284113e-05, "loss": 1.3059, "step": 4738 }, { "epoch": 0.86, "learning_rate": 1.7152539309596965e-05, "loss": 0.7861, "step": 4739 }, { "epoch": 0.86, "learning_rate": 1.7149828302909813e-05, "loss": 0.9918, "step": 4740 }, { "epoch": 0.86, "learning_rate": 1.7147117296222665e-05, "loss": 1.0345, "step": 4741 }, { "epoch": 0.86, "learning_rate": 1.7144406289535517e-05, "loss": 1.25, "step": 4742 }, { "epoch": 0.86, "learning_rate": 1.7141695282848366e-05, "loss": 0.7712, "step": 4743 }, { "epoch": 0.86, "learning_rate": 1.7138984276161214e-05, "loss": 0.7474, "step": 4744 }, { "epoch": 0.86, "learning_rate": 1.7136273269474066e-05, "loss": 1.4456, "step": 4745 }, { "epoch": 0.86, "learning_rate": 1.7133562262786914e-05, "loss": 1.1951, "step": 4746 }, { "epoch": 0.86, "learning_rate": 1.7130851256099766e-05, "loss": 1.1473, "step": 4747 }, { "epoch": 0.86, "learning_rate": 1.7128140249412614e-05, "loss": 1.1567, "step": 4748 }, { "epoch": 0.86, "learning_rate": 1.7125429242725466e-05, "loss": 0.645, "step": 4749 }, { "epoch": 0.86, "learning_rate": 1.7122718236038318e-05, "loss": 1.1651, "step": 4750 }, { "epoch": 0.86, "learning_rate": 1.7120007229351167e-05, "loss": 0.778, "step": 4751 }, { "epoch": 0.86, "learning_rate": 1.7117296222664015e-05, "loss": 0.7623, "step": 4752 }, { "epoch": 0.86, "learning_rate": 1.7114585215976867e-05, "loss": 0.8656, "step": 4753 }, { "epoch": 0.86, "learning_rate": 1.7111874209289715e-05, "loss": 0.7073, "step": 4754 }, { "epoch": 0.86, "learning_rate": 1.7109163202602567e-05, "loss": 0.8097, "step": 4755 }, { "epoch": 0.86, "learning_rate": 1.710645219591542e-05, "loss": 1.5955, "step": 4756 }, { "epoch": 0.86, "learning_rate": 1.7103741189228267e-05, "loss": 1.4534, "step": 4757 }, { "epoch": 0.86, "learning_rate": 1.710103018254112e-05, "loss": 0.6405, "step": 4758 }, { "epoch": 0.86, "learning_rate": 1.7098319175853968e-05, "loss": 1.123, "step": 4759 }, { "epoch": 0.86, "learning_rate": 1.7095608169166816e-05, "loss": 1.2666, "step": 4760 }, { "epoch": 0.86, "learning_rate": 1.7092897162479668e-05, "loss": 1.297, "step": 4761 }, { "epoch": 0.86, "learning_rate": 1.7090186155792516e-05, "loss": 1.1549, "step": 4762 }, { "epoch": 0.86, "learning_rate": 1.7087475149105368e-05, "loss": 1.7002, "step": 4763 }, { "epoch": 0.86, "learning_rate": 1.708476414241822e-05, "loss": 0.818, "step": 4764 }, { "epoch": 0.86, "learning_rate": 1.708205313573107e-05, "loss": 1.1864, "step": 4765 }, { "epoch": 0.86, "learning_rate": 1.707934212904392e-05, "loss": 0.7844, "step": 4766 }, { "epoch": 0.86, "learning_rate": 1.707663112235677e-05, "loss": 0.9548, "step": 4767 }, { "epoch": 0.86, "learning_rate": 1.7073920115669617e-05, "loss": 0.5466, "step": 4768 }, { "epoch": 0.86, "learning_rate": 1.707120910898247e-05, "loss": 0.9263, "step": 4769 }, { "epoch": 0.86, "learning_rate": 1.7068498102295317e-05, "loss": 1.1059, "step": 4770 }, { "epoch": 0.86, "learning_rate": 1.706578709560817e-05, "loss": 0.8236, "step": 4771 }, { "epoch": 0.86, "learning_rate": 1.706307608892102e-05, "loss": 1.3797, "step": 4772 }, { "epoch": 0.86, "learning_rate": 1.706036508223387e-05, "loss": 0.6193, "step": 4773 }, { "epoch": 0.86, "learning_rate": 1.705765407554672e-05, "loss": 0.9856, "step": 4774 }, { "epoch": 0.86, "learning_rate": 1.705494306885957e-05, "loss": 0.97, "step": 4775 }, { "epoch": 0.86, "learning_rate": 1.7052232062172418e-05, "loss": 0.845, "step": 4776 }, { "epoch": 0.86, "learning_rate": 1.704952105548527e-05, "loss": 0.907, "step": 4777 }, { "epoch": 0.86, "learning_rate": 1.7046810048798122e-05, "loss": 1.0505, "step": 4778 }, { "epoch": 0.86, "learning_rate": 1.704409904211097e-05, "loss": 0.3676, "step": 4779 }, { "epoch": 0.86, "learning_rate": 1.7041388035423822e-05, "loss": 1.0944, "step": 4780 }, { "epoch": 0.86, "learning_rate": 1.703867702873667e-05, "loss": 1.0597, "step": 4781 }, { "epoch": 0.86, "learning_rate": 1.7035966022049522e-05, "loss": 0.7349, "step": 4782 }, { "epoch": 0.86, "learning_rate": 1.703325501536237e-05, "loss": 0.9958, "step": 4783 }, { "epoch": 0.86, "learning_rate": 1.703054400867522e-05, "loss": 0.8659, "step": 4784 }, { "epoch": 0.86, "learning_rate": 1.702783300198807e-05, "loss": 1.5973, "step": 4785 }, { "epoch": 0.86, "learning_rate": 1.7025121995300923e-05, "loss": 0.8172, "step": 4786 }, { "epoch": 0.87, "learning_rate": 1.702241098861377e-05, "loss": 0.5419, "step": 4787 }, { "epoch": 0.87, "learning_rate": 1.7019699981926623e-05, "loss": 1.3502, "step": 4788 }, { "epoch": 0.87, "learning_rate": 1.7016988975239475e-05, "loss": 1.4353, "step": 4789 }, { "epoch": 0.87, "learning_rate": 1.7014277968552324e-05, "loss": 1.5014, "step": 4790 }, { "epoch": 0.87, "learning_rate": 1.7011566961865172e-05, "loss": 0.9939, "step": 4791 }, { "epoch": 0.87, "learning_rate": 1.700885595517802e-05, "loss": 0.9973, "step": 4792 }, { "epoch": 0.87, "learning_rate": 1.7006144948490872e-05, "loss": 1.0742, "step": 4793 }, { "epoch": 0.87, "learning_rate": 1.7003433941803724e-05, "loss": 1.3112, "step": 4794 }, { "epoch": 0.87, "learning_rate": 1.7000722935116572e-05, "loss": 1.0476, "step": 4795 }, { "epoch": 0.87, "learning_rate": 1.6998011928429424e-05, "loss": 0.5298, "step": 4796 }, { "epoch": 0.87, "learning_rate": 1.6995300921742276e-05, "loss": 0.7054, "step": 4797 }, { "epoch": 0.87, "learning_rate": 1.6992589915055125e-05, "loss": 0.6539, "step": 4798 }, { "epoch": 0.87, "learning_rate": 1.6989878908367973e-05, "loss": 1.2498, "step": 4799 }, { "epoch": 0.87, "learning_rate": 1.6987167901680825e-05, "loss": 0.9033, "step": 4800 }, { "epoch": 0.87, "learning_rate": 1.6984456894993673e-05, "loss": 1.3983, "step": 4801 }, { "epoch": 0.87, "learning_rate": 1.6981745888306525e-05, "loss": 0.6962, "step": 4802 }, { "epoch": 0.87, "learning_rate": 1.6979034881619374e-05, "loss": 1.2198, "step": 4803 }, { "epoch": 0.87, "learning_rate": 1.6976323874932225e-05, "loss": 1.0254, "step": 4804 }, { "epoch": 0.87, "learning_rate": 1.6973612868245077e-05, "loss": 1.1439, "step": 4805 }, { "epoch": 0.87, "learning_rate": 1.6970901861557926e-05, "loss": 0.9676, "step": 4806 }, { "epoch": 0.87, "learning_rate": 1.6968190854870774e-05, "loss": 1.1937, "step": 4807 }, { "epoch": 0.87, "learning_rate": 1.6965479848183626e-05, "loss": 1.0055, "step": 4808 }, { "epoch": 0.87, "learning_rate": 1.6962768841496474e-05, "loss": 1.0342, "step": 4809 }, { "epoch": 0.87, "learning_rate": 1.6960057834809326e-05, "loss": 0.858, "step": 4810 }, { "epoch": 0.87, "learning_rate": 1.6957346828122178e-05, "loss": 1.0247, "step": 4811 }, { "epoch": 0.87, "learning_rate": 1.6954635821435026e-05, "loss": 0.9037, "step": 4812 }, { "epoch": 0.87, "learning_rate": 1.6951924814747878e-05, "loss": 0.4737, "step": 4813 }, { "epoch": 0.87, "learning_rate": 1.6949213808060727e-05, "loss": 1.062, "step": 4814 }, { "epoch": 0.87, "learning_rate": 1.6946502801373575e-05, "loss": 0.8617, "step": 4815 }, { "epoch": 0.87, "learning_rate": 1.6943791794686427e-05, "loss": 1.0945, "step": 4816 }, { "epoch": 0.87, "learning_rate": 1.6941080787999275e-05, "loss": 1.1956, "step": 4817 }, { "epoch": 0.87, "learning_rate": 1.6938369781312127e-05, "loss": 1.3248, "step": 4818 }, { "epoch": 0.87, "learning_rate": 1.693565877462498e-05, "loss": 0.9391, "step": 4819 }, { "epoch": 0.87, "learning_rate": 1.6932947767937828e-05, "loss": 1.0622, "step": 4820 }, { "epoch": 0.87, "learning_rate": 1.693023676125068e-05, "loss": 2.2351, "step": 4821 }, { "epoch": 0.87, "learning_rate": 1.692752575456353e-05, "loss": 1.6395, "step": 4822 }, { "epoch": 0.87, "learning_rate": 1.6924814747876376e-05, "loss": 1.3244, "step": 4823 }, { "epoch": 0.87, "learning_rate": 1.6922103741189228e-05, "loss": 1.2545, "step": 4824 }, { "epoch": 0.87, "learning_rate": 1.691939273450208e-05, "loss": 1.1751, "step": 4825 }, { "epoch": 0.87, "learning_rate": 1.691668172781493e-05, "loss": 1.2472, "step": 4826 }, { "epoch": 0.87, "learning_rate": 1.691397072112778e-05, "loss": 1.0651, "step": 4827 }, { "epoch": 0.87, "learning_rate": 1.691125971444063e-05, "loss": 0.7773, "step": 4828 }, { "epoch": 0.87, "learning_rate": 1.690854870775348e-05, "loss": 1.1124, "step": 4829 }, { "epoch": 0.87, "learning_rate": 1.6905837701066332e-05, "loss": 1.1527, "step": 4830 }, { "epoch": 0.87, "learning_rate": 1.6903126694379177e-05, "loss": 1.4963, "step": 4831 }, { "epoch": 0.87, "learning_rate": 1.690041568769203e-05, "loss": 1.6689, "step": 4832 }, { "epoch": 0.87, "learning_rate": 1.689770468100488e-05, "loss": 0.8068, "step": 4833 }, { "epoch": 0.87, "learning_rate": 1.689499367431773e-05, "loss": 0.7083, "step": 4834 }, { "epoch": 0.87, "learning_rate": 1.689228266763058e-05, "loss": 1.1572, "step": 4835 }, { "epoch": 0.87, "learning_rate": 1.6889571660943433e-05, "loss": 1.0501, "step": 4836 }, { "epoch": 0.87, "learning_rate": 1.688686065425628e-05, "loss": 0.7177, "step": 4837 }, { "epoch": 0.87, "learning_rate": 1.6884149647569133e-05, "loss": 0.9744, "step": 4838 }, { "epoch": 0.87, "learning_rate": 1.688143864088198e-05, "loss": 1.1837, "step": 4839 }, { "epoch": 0.87, "learning_rate": 1.687872763419483e-05, "loss": 0.7404, "step": 4840 }, { "epoch": 0.87, "learning_rate": 1.6876016627507682e-05, "loss": 1.4238, "step": 4841 }, { "epoch": 0.88, "learning_rate": 1.687330562082053e-05, "loss": 0.6617, "step": 4842 }, { "epoch": 0.88, "learning_rate": 1.6870594614133382e-05, "loss": 0.6831, "step": 4843 }, { "epoch": 0.88, "learning_rate": 1.6867883607446234e-05, "loss": 1.6539, "step": 4844 }, { "epoch": 0.88, "learning_rate": 1.6865172600759083e-05, "loss": 0.7519, "step": 4845 }, { "epoch": 0.88, "learning_rate": 1.6862461594071934e-05, "loss": 1.1228, "step": 4846 }, { "epoch": 0.88, "learning_rate": 1.6859750587384783e-05, "loss": 1.448, "step": 4847 }, { "epoch": 0.88, "learning_rate": 1.685703958069763e-05, "loss": 0.9032, "step": 4848 }, { "epoch": 0.88, "learning_rate": 1.6854328574010483e-05, "loss": 0.9694, "step": 4849 }, { "epoch": 0.88, "learning_rate": 1.685161756732333e-05, "loss": 1.6121, "step": 4850 }, { "epoch": 0.88, "learning_rate": 1.6848906560636183e-05, "loss": 1.4552, "step": 4851 }, { "epoch": 0.88, "learning_rate": 1.6846195553949035e-05, "loss": 0.4529, "step": 4852 }, { "epoch": 0.88, "learning_rate": 1.6843484547261884e-05, "loss": 1.1369, "step": 4853 }, { "epoch": 0.88, "learning_rate": 1.6840773540574735e-05, "loss": 0.8789, "step": 4854 }, { "epoch": 0.88, "learning_rate": 1.6838062533887584e-05, "loss": 1.2186, "step": 4855 }, { "epoch": 0.88, "learning_rate": 1.6835351527200432e-05, "loss": 0.8443, "step": 4856 }, { "epoch": 0.88, "learning_rate": 1.6832640520513284e-05, "loss": 0.9262, "step": 4857 }, { "epoch": 0.88, "learning_rate": 1.6829929513826136e-05, "loss": 1.1893, "step": 4858 }, { "epoch": 0.88, "learning_rate": 1.6827218507138984e-05, "loss": 1.7251, "step": 4859 }, { "epoch": 0.88, "learning_rate": 1.6824507500451836e-05, "loss": 0.8216, "step": 4860 }, { "epoch": 0.88, "learning_rate": 1.6821796493764685e-05, "loss": 1.2642, "step": 4861 }, { "epoch": 0.88, "learning_rate": 1.6819085487077537e-05, "loss": 1.8864, "step": 4862 }, { "epoch": 0.88, "learning_rate": 1.6816374480390385e-05, "loss": 1.183, "step": 4863 }, { "epoch": 0.88, "learning_rate": 1.6813663473703233e-05, "loss": 0.8349, "step": 4864 }, { "epoch": 0.88, "learning_rate": 1.6810952467016085e-05, "loss": 1.3434, "step": 4865 }, { "epoch": 0.88, "learning_rate": 1.6808241460328937e-05, "loss": 1.0468, "step": 4866 }, { "epoch": 0.88, "learning_rate": 1.6805530453641786e-05, "loss": 0.8324, "step": 4867 }, { "epoch": 0.88, "learning_rate": 1.6802819446954637e-05, "loss": 0.8893, "step": 4868 }, { "epoch": 0.88, "learning_rate": 1.680010844026749e-05, "loss": 0.792, "step": 4869 }, { "epoch": 0.88, "learning_rate": 1.6797397433580338e-05, "loss": 1.5437, "step": 4870 }, { "epoch": 0.88, "learning_rate": 1.6794686426893186e-05, "loss": 0.9923, "step": 4871 }, { "epoch": 0.88, "learning_rate": 1.6791975420206035e-05, "loss": 1.0018, "step": 4872 }, { "epoch": 0.88, "learning_rate": 1.6789264413518886e-05, "loss": 0.8635, "step": 4873 }, { "epoch": 0.88, "learning_rate": 1.6786553406831738e-05, "loss": 0.7745, "step": 4874 }, { "epoch": 0.88, "learning_rate": 1.6783842400144587e-05, "loss": 0.7958, "step": 4875 }, { "epoch": 0.88, "learning_rate": 1.678113139345744e-05, "loss": 1.5573, "step": 4876 }, { "epoch": 0.88, "learning_rate": 1.677842038677029e-05, "loss": 1.0298, "step": 4877 }, { "epoch": 0.88, "learning_rate": 1.677570938008314e-05, "loss": 0.5684, "step": 4878 }, { "epoch": 0.88, "learning_rate": 1.6772998373395987e-05, "loss": 0.655, "step": 4879 }, { "epoch": 0.88, "learning_rate": 1.677028736670884e-05, "loss": 1.2193, "step": 4880 }, { "epoch": 0.88, "learning_rate": 1.6767576360021687e-05, "loss": 1.072, "step": 4881 }, { "epoch": 0.88, "learning_rate": 1.676486535333454e-05, "loss": 0.956, "step": 4882 }, { "epoch": 0.88, "learning_rate": 1.6762154346647388e-05, "loss": 0.6732, "step": 4883 }, { "epoch": 0.88, "learning_rate": 1.675944333996024e-05, "loss": 0.6357, "step": 4884 }, { "epoch": 0.88, "learning_rate": 1.675673233327309e-05, "loss": 0.6086, "step": 4885 }, { "epoch": 0.88, "learning_rate": 1.675402132658594e-05, "loss": 1.7686, "step": 4886 }, { "epoch": 0.88, "learning_rate": 1.6751310319898788e-05, "loss": 1.3086, "step": 4887 }, { "epoch": 0.88, "learning_rate": 1.674859931321164e-05, "loss": 1.0996, "step": 4888 }, { "epoch": 0.88, "learning_rate": 1.674588830652449e-05, "loss": 0.7122, "step": 4889 }, { "epoch": 0.88, "learning_rate": 1.674317729983734e-05, "loss": 0.5803, "step": 4890 }, { "epoch": 0.88, "learning_rate": 1.6740466293150192e-05, "loss": 0.8058, "step": 4891 }, { "epoch": 0.88, "learning_rate": 1.673775528646304e-05, "loss": 1.0548, "step": 4892 }, { "epoch": 0.88, "learning_rate": 1.6735044279775892e-05, "loss": 1.0645, "step": 4893 }, { "epoch": 0.88, "learning_rate": 1.6732333273088737e-05, "loss": 1.301, "step": 4894 }, { "epoch": 0.88, "learning_rate": 1.672962226640159e-05, "loss": 1.3286, "step": 4895 }, { "epoch": 0.88, "learning_rate": 1.672691125971444e-05, "loss": 0.4104, "step": 4896 }, { "epoch": 0.89, "learning_rate": 1.672420025302729e-05, "loss": 1.0109, "step": 4897 }, { "epoch": 0.89, "learning_rate": 1.672148924634014e-05, "loss": 0.8175, "step": 4898 }, { "epoch": 0.89, "learning_rate": 1.6718778239652993e-05, "loss": 1.7588, "step": 4899 }, { "epoch": 0.89, "learning_rate": 1.671606723296584e-05, "loss": 1.1818, "step": 4900 }, { "epoch": 0.89, "learning_rate": 1.6713356226278693e-05, "loss": 1.2349, "step": 4901 }, { "epoch": 0.89, "learning_rate": 1.6710645219591542e-05, "loss": 1.3404, "step": 4902 }, { "epoch": 0.89, "learning_rate": 1.670793421290439e-05, "loss": 2.1308, "step": 4903 }, { "epoch": 0.89, "learning_rate": 1.6705223206217242e-05, "loss": 1.1734, "step": 4904 }, { "epoch": 0.89, "learning_rate": 1.670251219953009e-05, "loss": 1.0889, "step": 4905 }, { "epoch": 0.89, "learning_rate": 1.6699801192842942e-05, "loss": 1.2443, "step": 4906 }, { "epoch": 0.89, "learning_rate": 1.6697090186155794e-05, "loss": 0.7884, "step": 4907 }, { "epoch": 0.89, "learning_rate": 1.6694379179468643e-05, "loss": 1.0721, "step": 4908 }, { "epoch": 0.89, "learning_rate": 1.6691668172781495e-05, "loss": 1.0752, "step": 4909 }, { "epoch": 0.89, "learning_rate": 1.6688957166094343e-05, "loss": 1.5995, "step": 4910 }, { "epoch": 0.89, "learning_rate": 1.668624615940719e-05, "loss": 0.7656, "step": 4911 }, { "epoch": 0.89, "learning_rate": 1.6683535152720043e-05, "loss": 1.2299, "step": 4912 }, { "epoch": 0.89, "learning_rate": 1.6680824146032895e-05, "loss": 1.0593, "step": 4913 }, { "epoch": 0.89, "learning_rate": 1.6678113139345744e-05, "loss": 0.9947, "step": 4914 }, { "epoch": 0.89, "learning_rate": 1.6675402132658595e-05, "loss": 1.0913, "step": 4915 }, { "epoch": 0.89, "learning_rate": 1.6672691125971444e-05, "loss": 0.7264, "step": 4916 }, { "epoch": 0.89, "learning_rate": 1.6669980119284296e-05, "loss": 0.5847, "step": 4917 }, { "epoch": 0.89, "learning_rate": 1.6667269112597144e-05, "loss": 1.0365, "step": 4918 }, { "epoch": 0.89, "learning_rate": 1.6664558105909993e-05, "loss": 0.8294, "step": 4919 }, { "epoch": 0.89, "learning_rate": 1.6661847099222844e-05, "loss": 1.7234, "step": 4920 }, { "epoch": 0.89, "learning_rate": 1.6659136092535696e-05, "loss": 1.7913, "step": 4921 }, { "epoch": 0.89, "learning_rate": 1.6656425085848545e-05, "loss": 1.1199, "step": 4922 }, { "epoch": 0.89, "learning_rate": 1.6653714079161396e-05, "loss": 1.2265, "step": 4923 }, { "epoch": 0.89, "learning_rate": 1.6651003072474248e-05, "loss": 1.6522, "step": 4924 }, { "epoch": 0.89, "learning_rate": 1.6648292065787097e-05, "loss": 1.6103, "step": 4925 }, { "epoch": 0.89, "learning_rate": 1.6645581059099945e-05, "loss": 0.956, "step": 4926 }, { "epoch": 0.89, "learning_rate": 1.6642870052412794e-05, "loss": 1.449, "step": 4927 }, { "epoch": 0.89, "learning_rate": 1.6640159045725645e-05, "loss": 0.8971, "step": 4928 }, { "epoch": 0.89, "learning_rate": 1.6637448039038497e-05, "loss": 0.9813, "step": 4929 }, { "epoch": 0.89, "learning_rate": 1.6634737032351346e-05, "loss": 1.2775, "step": 4930 }, { "epoch": 0.89, "learning_rate": 1.6632026025664198e-05, "loss": 1.0043, "step": 4931 }, { "epoch": 0.89, "learning_rate": 1.662931501897705e-05, "loss": 0.9735, "step": 4932 }, { "epoch": 0.89, "learning_rate": 1.6626604012289898e-05, "loss": 0.6799, "step": 4933 }, { "epoch": 0.89, "learning_rate": 1.6623893005602746e-05, "loss": 0.8231, "step": 4934 }, { "epoch": 0.89, "learning_rate": 1.6621181998915598e-05, "loss": 1.2358, "step": 4935 }, { "epoch": 0.89, "learning_rate": 1.6618470992228447e-05, "loss": 1.1393, "step": 4936 }, { "epoch": 0.89, "learning_rate": 1.66157599855413e-05, "loss": 0.7146, "step": 4937 }, { "epoch": 0.89, "learning_rate": 1.6613048978854147e-05, "loss": 1.1669, "step": 4938 }, { "epoch": 0.89, "learning_rate": 1.6610337972167e-05, "loss": 1.9243, "step": 4939 }, { "epoch": 0.89, "learning_rate": 1.660762696547985e-05, "loss": 0.6006, "step": 4940 }, { "epoch": 0.89, "learning_rate": 1.66049159587927e-05, "loss": 1.1236, "step": 4941 }, { "epoch": 0.89, "learning_rate": 1.6602204952105547e-05, "loss": 0.8888, "step": 4942 }, { "epoch": 0.89, "learning_rate": 1.65994939454184e-05, "loss": 1.1098, "step": 4943 }, { "epoch": 0.89, "learning_rate": 1.6596782938731248e-05, "loss": 1.25, "step": 4944 }, { "epoch": 0.89, "learning_rate": 1.65940719320441e-05, "loss": 0.6377, "step": 4945 }, { "epoch": 0.89, "learning_rate": 1.659136092535695e-05, "loss": 1.1992, "step": 4946 }, { "epoch": 0.89, "learning_rate": 1.65886499186698e-05, "loss": 1.8272, "step": 4947 }, { "epoch": 0.89, "learning_rate": 1.658593891198265e-05, "loss": 1.3303, "step": 4948 }, { "epoch": 0.89, "learning_rate": 1.65832279052955e-05, "loss": 0.8509, "step": 4949 }, { "epoch": 0.89, "learning_rate": 1.658051689860835e-05, "loss": 0.6771, "step": 4950 }, { "epoch": 0.89, "learning_rate": 1.65778058919212e-05, "loss": 1.1429, "step": 4951 }, { "epoch": 0.89, "learning_rate": 1.657509488523405e-05, "loss": 1.0372, "step": 4952 }, { "epoch": 0.9, "learning_rate": 1.65723838785469e-05, "loss": 0.9179, "step": 4953 }, { "epoch": 0.9, "learning_rate": 1.6569672871859752e-05, "loss": 1.2971, "step": 4954 }, { "epoch": 0.9, "learning_rate": 1.65669618651726e-05, "loss": 0.9167, "step": 4955 }, { "epoch": 0.9, "learning_rate": 1.6564250858485453e-05, "loss": 1.9013, "step": 4956 }, { "epoch": 0.9, "learning_rate": 1.6561539851798304e-05, "loss": 0.7445, "step": 4957 }, { "epoch": 0.9, "learning_rate": 1.655882884511115e-05, "loss": 0.8801, "step": 4958 }, { "epoch": 0.9, "learning_rate": 1.6556117838424e-05, "loss": 0.8752, "step": 4959 }, { "epoch": 0.9, "learning_rate": 1.6553406831736853e-05, "loss": 1.3146, "step": 4960 }, { "epoch": 0.9, "learning_rate": 1.65506958250497e-05, "loss": 1.8209, "step": 4961 }, { "epoch": 0.9, "learning_rate": 1.6547984818362553e-05, "loss": 0.7834, "step": 4962 }, { "epoch": 0.9, "learning_rate": 1.6545273811675402e-05, "loss": 0.7302, "step": 4963 }, { "epoch": 0.9, "learning_rate": 1.6542562804988254e-05, "loss": 1.3716, "step": 4964 }, { "epoch": 0.9, "learning_rate": 1.6539851798301105e-05, "loss": 1.6315, "step": 4965 }, { "epoch": 0.9, "learning_rate": 1.653714079161395e-05, "loss": 0.9814, "step": 4966 }, { "epoch": 0.9, "learning_rate": 1.6534429784926802e-05, "loss": 1.6803, "step": 4967 }, { "epoch": 0.9, "learning_rate": 1.6531718778239654e-05, "loss": 0.7457, "step": 4968 }, { "epoch": 0.9, "learning_rate": 1.6529007771552503e-05, "loss": 0.9625, "step": 4969 }, { "epoch": 0.9, "learning_rate": 1.6526296764865354e-05, "loss": 0.5657, "step": 4970 }, { "epoch": 0.9, "learning_rate": 1.6523585758178206e-05, "loss": 1.4864, "step": 4971 }, { "epoch": 0.9, "learning_rate": 1.6520874751491055e-05, "loss": 0.7145, "step": 4972 }, { "epoch": 0.9, "learning_rate": 1.6518163744803907e-05, "loss": 1.0613, "step": 4973 }, { "epoch": 0.9, "learning_rate": 1.651545273811675e-05, "loss": 0.8687, "step": 4974 }, { "epoch": 0.9, "learning_rate": 1.6512741731429603e-05, "loss": 0.5594, "step": 4975 }, { "epoch": 0.9, "learning_rate": 1.6510030724742455e-05, "loss": 1.3705, "step": 4976 }, { "epoch": 0.9, "learning_rate": 1.6507319718055304e-05, "loss": 0.9064, "step": 4977 }, { "epoch": 0.9, "learning_rate": 1.6504608711368156e-05, "loss": 0.6735, "step": 4978 }, { "epoch": 0.9, "learning_rate": 1.6501897704681007e-05, "loss": 0.9884, "step": 4979 }, { "epoch": 0.9, "learning_rate": 1.6499186697993856e-05, "loss": 0.7746, "step": 4980 }, { "epoch": 0.9, "learning_rate": 1.6496475691306708e-05, "loss": 0.7133, "step": 4981 }, { "epoch": 0.9, "learning_rate": 1.6493764684619556e-05, "loss": 1.3132, "step": 4982 }, { "epoch": 0.9, "learning_rate": 1.6491053677932405e-05, "loss": 1.2309, "step": 4983 }, { "epoch": 0.9, "learning_rate": 1.6488342671245256e-05, "loss": 1.8962, "step": 4984 }, { "epoch": 0.9, "learning_rate": 1.6485631664558105e-05, "loss": 0.7989, "step": 4985 }, { "epoch": 0.9, "learning_rate": 1.6482920657870957e-05, "loss": 0.5087, "step": 4986 }, { "epoch": 0.9, "learning_rate": 1.648020965118381e-05, "loss": 1.439, "step": 4987 }, { "epoch": 0.9, "learning_rate": 1.6477498644496657e-05, "loss": 1.0163, "step": 4988 }, { "epoch": 0.9, "learning_rate": 1.647478763780951e-05, "loss": 0.9161, "step": 4989 }, { "epoch": 0.9, "learning_rate": 1.6472076631122357e-05, "loss": 1.2906, "step": 4990 }, { "epoch": 0.9, "learning_rate": 1.6469365624435206e-05, "loss": 0.6974, "step": 4991 }, { "epoch": 0.9, "learning_rate": 1.6466654617748057e-05, "loss": 0.694, "step": 4992 }, { "epoch": 0.9, "learning_rate": 1.646394361106091e-05, "loss": 0.9936, "step": 4993 }, { "epoch": 0.9, "learning_rate": 1.6461232604373758e-05, "loss": 0.714, "step": 4994 }, { "epoch": 0.9, "learning_rate": 1.645852159768661e-05, "loss": 1.102, "step": 4995 }, { "epoch": 0.9, "learning_rate": 1.6455810590999458e-05, "loss": 1.1824, "step": 4996 }, { "epoch": 0.9, "learning_rate": 1.645309958431231e-05, "loss": 0.6751, "step": 4997 }, { "epoch": 0.9, "learning_rate": 1.6450388577625158e-05, "loss": 1.1895, "step": 4998 }, { "epoch": 0.9, "learning_rate": 1.6447677570938007e-05, "loss": 0.8975, "step": 4999 }, { "epoch": 0.9, "learning_rate": 1.644496656425086e-05, "loss": 0.9645, "step": 5000 }, { "epoch": 0.9, "learning_rate": 1.644225555756371e-05, "loss": 0.8344, "step": 5001 }, { "epoch": 0.9, "learning_rate": 1.643954455087656e-05, "loss": 0.7746, "step": 5002 }, { "epoch": 0.9, "learning_rate": 1.643683354418941e-05, "loss": 0.5767, "step": 5003 }, { "epoch": 0.9, "learning_rate": 1.6434122537502262e-05, "loss": 0.8043, "step": 5004 }, { "epoch": 0.9, "learning_rate": 1.643141153081511e-05, "loss": 1.6544, "step": 5005 }, { "epoch": 0.9, "learning_rate": 1.642870052412796e-05, "loss": 0.981, "step": 5006 }, { "epoch": 0.9, "learning_rate": 1.6425989517440808e-05, "loss": 0.8566, "step": 5007 }, { "epoch": 0.91, "learning_rate": 1.642327851075366e-05, "loss": 1.5994, "step": 5008 }, { "epoch": 0.91, "learning_rate": 1.642056750406651e-05, "loss": 0.8646, "step": 5009 }, { "epoch": 0.91, "learning_rate": 1.641785649737936e-05, "loss": 0.7152, "step": 5010 }, { "epoch": 0.91, "learning_rate": 1.641514549069221e-05, "loss": 1.4722, "step": 5011 }, { "epoch": 0.91, "learning_rate": 1.6412434484005063e-05, "loss": 1.016, "step": 5012 }, { "epoch": 0.91, "learning_rate": 1.6409723477317912e-05, "loss": 0.8141, "step": 5013 }, { "epoch": 0.91, "learning_rate": 1.640701247063076e-05, "loss": 0.4782, "step": 5014 }, { "epoch": 0.91, "learning_rate": 1.6404301463943612e-05, "loss": 0.9168, "step": 5015 }, { "epoch": 0.91, "learning_rate": 1.640159045725646e-05, "loss": 0.7802, "step": 5016 }, { "epoch": 0.91, "learning_rate": 1.6398879450569312e-05, "loss": 1.4356, "step": 5017 }, { "epoch": 0.91, "learning_rate": 1.639616844388216e-05, "loss": 1.0088, "step": 5018 }, { "epoch": 0.91, "learning_rate": 1.6393457437195013e-05, "loss": 0.7062, "step": 5019 }, { "epoch": 0.91, "learning_rate": 1.6390746430507865e-05, "loss": 1.3257, "step": 5020 }, { "epoch": 0.91, "learning_rate": 1.638803542382071e-05, "loss": 0.5383, "step": 5021 }, { "epoch": 0.91, "learning_rate": 1.638532441713356e-05, "loss": 1.2583, "step": 5022 }, { "epoch": 0.91, "learning_rate": 1.6382613410446413e-05, "loss": 0.9119, "step": 5023 }, { "epoch": 0.91, "learning_rate": 1.6379902403759262e-05, "loss": 1.3315, "step": 5024 }, { "epoch": 0.91, "learning_rate": 1.6377191397072114e-05, "loss": 1.0579, "step": 5025 }, { "epoch": 0.91, "learning_rate": 1.6374480390384965e-05, "loss": 0.9688, "step": 5026 }, { "epoch": 0.91, "learning_rate": 1.6371769383697814e-05, "loss": 1.3893, "step": 5027 }, { "epoch": 0.91, "learning_rate": 1.6369058377010666e-05, "loss": 1.5397, "step": 5028 }, { "epoch": 0.91, "learning_rate": 1.636634737032351e-05, "loss": 1.2042, "step": 5029 }, { "epoch": 0.91, "learning_rate": 1.6363636363636363e-05, "loss": 1.382, "step": 5030 }, { "epoch": 0.91, "learning_rate": 1.6360925356949214e-05, "loss": 1.0245, "step": 5031 }, { "epoch": 0.91, "learning_rate": 1.6358214350262063e-05, "loss": 0.729, "step": 5032 }, { "epoch": 0.91, "learning_rate": 1.6355503343574915e-05, "loss": 1.3579, "step": 5033 }, { "epoch": 0.91, "learning_rate": 1.6352792336887766e-05, "loss": 1.5212, "step": 5034 }, { "epoch": 0.91, "learning_rate": 1.6350081330200615e-05, "loss": 1.0578, "step": 5035 }, { "epoch": 0.91, "learning_rate": 1.6347370323513467e-05, "loss": 0.9893, "step": 5036 }, { "epoch": 0.91, "learning_rate": 1.6344659316826315e-05, "loss": 1.279, "step": 5037 }, { "epoch": 0.91, "learning_rate": 1.6341948310139164e-05, "loss": 1.0578, "step": 5038 }, { "epoch": 0.91, "learning_rate": 1.6339237303452015e-05, "loss": 0.5792, "step": 5039 }, { "epoch": 0.91, "learning_rate": 1.6336526296764864e-05, "loss": 0.4096, "step": 5040 }, { "epoch": 0.91, "learning_rate": 1.6333815290077716e-05, "loss": 1.0492, "step": 5041 }, { "epoch": 0.91, "learning_rate": 1.6331104283390568e-05, "loss": 1.1527, "step": 5042 }, { "epoch": 0.91, "learning_rate": 1.6328393276703416e-05, "loss": 1.4784, "step": 5043 }, { "epoch": 0.91, "learning_rate": 1.6325682270016268e-05, "loss": 1.103, "step": 5044 }, { "epoch": 0.91, "learning_rate": 1.6322971263329116e-05, "loss": 0.946, "step": 5045 }, { "epoch": 0.91, "learning_rate": 1.6320260256641965e-05, "loss": 1.3014, "step": 5046 }, { "epoch": 0.91, "learning_rate": 1.6317549249954816e-05, "loss": 1.0123, "step": 5047 }, { "epoch": 0.91, "learning_rate": 1.631483824326767e-05, "loss": 0.5164, "step": 5048 }, { "epoch": 0.91, "learning_rate": 1.6312127236580517e-05, "loss": 1.5717, "step": 5049 }, { "epoch": 0.91, "learning_rate": 1.630941622989337e-05, "loss": 1.4067, "step": 5050 }, { "epoch": 0.91, "learning_rate": 1.6306705223206217e-05, "loss": 0.8694, "step": 5051 }, { "epoch": 0.91, "learning_rate": 1.630399421651907e-05, "loss": 1.0442, "step": 5052 }, { "epoch": 0.91, "learning_rate": 1.6301283209831917e-05, "loss": 1.8876, "step": 5053 }, { "epoch": 0.91, "learning_rate": 1.6298572203144766e-05, "loss": 1.8832, "step": 5054 }, { "epoch": 0.91, "learning_rate": 1.6295861196457618e-05, "loss": 1.0942, "step": 5055 }, { "epoch": 0.91, "learning_rate": 1.629315018977047e-05, "loss": 1.4258, "step": 5056 }, { "epoch": 0.91, "learning_rate": 1.6290439183083318e-05, "loss": 0.5786, "step": 5057 }, { "epoch": 0.91, "learning_rate": 1.628772817639617e-05, "loss": 1.4255, "step": 5058 }, { "epoch": 0.91, "learning_rate": 1.628501716970902e-05, "loss": 1.1229, "step": 5059 }, { "epoch": 0.91, "learning_rate": 1.628230616302187e-05, "loss": 0.515, "step": 5060 }, { "epoch": 0.91, "learning_rate": 1.627959515633472e-05, "loss": 1.2637, "step": 5061 }, { "epoch": 0.91, "learning_rate": 1.6276884149647567e-05, "loss": 0.9214, "step": 5062 }, { "epoch": 0.92, "learning_rate": 1.627417314296042e-05, "loss": 0.9832, "step": 5063 }, { "epoch": 0.92, "learning_rate": 1.627146213627327e-05, "loss": 1.4539, "step": 5064 }, { "epoch": 0.92, "learning_rate": 1.626875112958612e-05, "loss": 1.4174, "step": 5065 }, { "epoch": 0.92, "learning_rate": 1.626604012289897e-05, "loss": 0.8352, "step": 5066 }, { "epoch": 0.92, "learning_rate": 1.6263329116211823e-05, "loss": 1.2933, "step": 5067 }, { "epoch": 0.92, "learning_rate": 1.626061810952467e-05, "loss": 0.8276, "step": 5068 }, { "epoch": 0.92, "learning_rate": 1.625790710283752e-05, "loss": 1.0929, "step": 5069 }, { "epoch": 0.92, "learning_rate": 1.625519609615037e-05, "loss": 1.559, "step": 5070 }, { "epoch": 0.92, "learning_rate": 1.625248508946322e-05, "loss": 1.2441, "step": 5071 }, { "epoch": 0.92, "learning_rate": 1.624977408277607e-05, "loss": 0.6635, "step": 5072 }, { "epoch": 0.92, "learning_rate": 1.624706307608892e-05, "loss": 1.0901, "step": 5073 }, { "epoch": 0.92, "learning_rate": 1.6244352069401772e-05, "loss": 1.4035, "step": 5074 }, { "epoch": 0.92, "learning_rate": 1.6241641062714624e-05, "loss": 1.0948, "step": 5075 }, { "epoch": 0.92, "learning_rate": 1.6238930056027472e-05, "loss": 1.6721, "step": 5076 }, { "epoch": 0.92, "learning_rate": 1.623621904934032e-05, "loss": 1.4091, "step": 5077 }, { "epoch": 0.92, "learning_rate": 1.6233508042653172e-05, "loss": 1.1267, "step": 5078 }, { "epoch": 0.92, "learning_rate": 1.623079703596602e-05, "loss": 0.6203, "step": 5079 }, { "epoch": 0.92, "learning_rate": 1.6228086029278873e-05, "loss": 1.1724, "step": 5080 }, { "epoch": 0.92, "learning_rate": 1.6225375022591724e-05, "loss": 1.0458, "step": 5081 }, { "epoch": 0.92, "learning_rate": 1.6222664015904573e-05, "loss": 1.0854, "step": 5082 }, { "epoch": 0.92, "learning_rate": 1.6219953009217425e-05, "loss": 1.5141, "step": 5083 }, { "epoch": 0.92, "learning_rate": 1.6217242002530273e-05, "loss": 1.0637, "step": 5084 }, { "epoch": 0.92, "learning_rate": 1.621453099584312e-05, "loss": 0.6852, "step": 5085 }, { "epoch": 0.92, "learning_rate": 1.6211819989155973e-05, "loss": 1.3279, "step": 5086 }, { "epoch": 0.92, "learning_rate": 1.6209108982468822e-05, "loss": 1.0476, "step": 5087 }, { "epoch": 0.92, "learning_rate": 1.6206397975781674e-05, "loss": 1.3343, "step": 5088 }, { "epoch": 0.92, "learning_rate": 1.6203686969094526e-05, "loss": 0.848, "step": 5089 }, { "epoch": 0.92, "learning_rate": 1.6200975962407374e-05, "loss": 1.0707, "step": 5090 }, { "epoch": 0.92, "learning_rate": 1.6198264955720226e-05, "loss": 1.6098, "step": 5091 }, { "epoch": 0.92, "learning_rate": 1.6195553949033078e-05, "loss": 1.2567, "step": 5092 }, { "epoch": 0.92, "learning_rate": 1.6192842942345923e-05, "loss": 0.7811, "step": 5093 }, { "epoch": 0.92, "learning_rate": 1.6190131935658774e-05, "loss": 0.3659, "step": 5094 }, { "epoch": 0.92, "learning_rate": 1.6187420928971623e-05, "loss": 0.9641, "step": 5095 }, { "epoch": 0.92, "learning_rate": 1.6184709922284475e-05, "loss": 0.6798, "step": 5096 }, { "epoch": 0.92, "learning_rate": 1.6181998915597327e-05, "loss": 1.1859, "step": 5097 }, { "epoch": 0.92, "learning_rate": 1.6179287908910175e-05, "loss": 1.0985, "step": 5098 }, { "epoch": 0.92, "learning_rate": 1.6176576902223027e-05, "loss": 1.4012, "step": 5099 }, { "epoch": 0.92, "learning_rate": 1.617386589553588e-05, "loss": 1.0076, "step": 5100 }, { "epoch": 0.92, "learning_rate": 1.6171154888848724e-05, "loss": 1.6947, "step": 5101 }, { "epoch": 0.92, "learning_rate": 1.6168443882161576e-05, "loss": 0.5921, "step": 5102 }, { "epoch": 0.92, "learning_rate": 1.6165732875474427e-05, "loss": 0.5874, "step": 5103 }, { "epoch": 0.92, "learning_rate": 1.6163021868787276e-05, "loss": 0.7131, "step": 5104 }, { "epoch": 0.92, "learning_rate": 1.6160310862100128e-05, "loss": 0.6165, "step": 5105 }, { "epoch": 0.92, "learning_rate": 1.6157599855412976e-05, "loss": 0.7594, "step": 5106 }, { "epoch": 0.92, "learning_rate": 1.6154888848725828e-05, "loss": 0.516, "step": 5107 }, { "epoch": 0.92, "learning_rate": 1.615217784203868e-05, "loss": 1.2701, "step": 5108 }, { "epoch": 0.92, "learning_rate": 1.6149466835351525e-05, "loss": 0.7541, "step": 5109 }, { "epoch": 0.92, "learning_rate": 1.6146755828664377e-05, "loss": 1.1418, "step": 5110 }, { "epoch": 0.92, "learning_rate": 1.614404482197723e-05, "loss": 2.3949, "step": 5111 }, { "epoch": 0.92, "learning_rate": 1.6141333815290077e-05, "loss": 1.1317, "step": 5112 }, { "epoch": 0.92, "learning_rate": 1.613862280860293e-05, "loss": 1.0987, "step": 5113 }, { "epoch": 0.92, "learning_rate": 1.613591180191578e-05, "loss": 1.2393, "step": 5114 }, { "epoch": 0.92, "learning_rate": 1.613320079522863e-05, "loss": 1.2883, "step": 5115 }, { "epoch": 0.92, "learning_rate": 1.613048978854148e-05, "loss": 1.0248, "step": 5116 }, { "epoch": 0.92, "learning_rate": 1.612777878185433e-05, "loss": 0.4514, "step": 5117 }, { "epoch": 0.92, "learning_rate": 1.6125067775167178e-05, "loss": 0.7938, "step": 5118 }, { "epoch": 0.93, "learning_rate": 1.612235676848003e-05, "loss": 1.0703, "step": 5119 }, { "epoch": 0.93, "learning_rate": 1.6119645761792878e-05, "loss": 1.34, "step": 5120 }, { "epoch": 0.93, "learning_rate": 1.611693475510573e-05, "loss": 0.9139, "step": 5121 }, { "epoch": 0.93, "learning_rate": 1.611422374841858e-05, "loss": 0.6142, "step": 5122 }, { "epoch": 0.93, "learning_rate": 1.611151274173143e-05, "loss": 0.7845, "step": 5123 }, { "epoch": 0.93, "learning_rate": 1.6108801735044282e-05, "loss": 0.5285, "step": 5124 }, { "epoch": 0.93, "learning_rate": 1.610609072835713e-05, "loss": 1.0733, "step": 5125 }, { "epoch": 0.93, "learning_rate": 1.610337972166998e-05, "loss": 1.4742, "step": 5126 }, { "epoch": 0.93, "learning_rate": 1.610066871498283e-05, "loss": 0.8093, "step": 5127 }, { "epoch": 0.93, "learning_rate": 1.6097957708295682e-05, "loss": 0.9224, "step": 5128 }, { "epoch": 0.93, "learning_rate": 1.609524670160853e-05, "loss": 1.3502, "step": 5129 }, { "epoch": 0.93, "learning_rate": 1.6092535694921383e-05, "loss": 1.2521, "step": 5130 }, { "epoch": 0.93, "learning_rate": 1.608982468823423e-05, "loss": 1.2548, "step": 5131 }, { "epoch": 0.93, "learning_rate": 1.6087113681547083e-05, "loss": 1.1075, "step": 5132 }, { "epoch": 0.93, "learning_rate": 1.608440267485993e-05, "loss": 1.1076, "step": 5133 }, { "epoch": 0.93, "learning_rate": 1.608169166817278e-05, "loss": 1.2556, "step": 5134 }, { "epoch": 0.93, "learning_rate": 1.607898066148563e-05, "loss": 1.0867, "step": 5135 }, { "epoch": 0.93, "learning_rate": 1.6076269654798484e-05, "loss": 1.4116, "step": 5136 }, { "epoch": 0.93, "learning_rate": 1.6073558648111332e-05, "loss": 0.589, "step": 5137 }, { "epoch": 0.93, "learning_rate": 1.6070847641424184e-05, "loss": 0.8783, "step": 5138 }, { "epoch": 0.93, "learning_rate": 1.6068136634737036e-05, "loss": 0.6979, "step": 5139 }, { "epoch": 0.93, "learning_rate": 1.6065425628049884e-05, "loss": 1.5155, "step": 5140 }, { "epoch": 0.93, "learning_rate": 1.6062714621362732e-05, "loss": 1.4528, "step": 5141 }, { "epoch": 0.93, "learning_rate": 1.606000361467558e-05, "loss": 1.8536, "step": 5142 }, { "epoch": 0.93, "learning_rate": 1.6057292607988433e-05, "loss": 0.6027, "step": 5143 }, { "epoch": 0.93, "learning_rate": 1.6054581601301285e-05, "loss": 1.295, "step": 5144 }, { "epoch": 0.93, "learning_rate": 1.6051870594614133e-05, "loss": 0.5293, "step": 5145 }, { "epoch": 0.93, "learning_rate": 1.6049159587926985e-05, "loss": 0.6493, "step": 5146 }, { "epoch": 0.93, "learning_rate": 1.6046448581239837e-05, "loss": 1.3251, "step": 5147 }, { "epoch": 0.93, "learning_rate": 1.6043737574552682e-05, "loss": 1.3884, "step": 5148 }, { "epoch": 0.93, "learning_rate": 1.6041026567865534e-05, "loss": 1.2661, "step": 5149 }, { "epoch": 0.93, "learning_rate": 1.6038315561178385e-05, "loss": 1.082, "step": 5150 }, { "epoch": 0.93, "learning_rate": 1.6035604554491234e-05, "loss": 1.6191, "step": 5151 }, { "epoch": 0.93, "learning_rate": 1.6032893547804086e-05, "loss": 1.2637, "step": 5152 }, { "epoch": 0.93, "learning_rate": 1.6030182541116934e-05, "loss": 0.678, "step": 5153 }, { "epoch": 0.93, "learning_rate": 1.6027471534429786e-05, "loss": 1.0601, "step": 5154 }, { "epoch": 0.93, "learning_rate": 1.6024760527742638e-05, "loss": 1.2245, "step": 5155 }, { "epoch": 0.93, "learning_rate": 1.6022049521055483e-05, "loss": 2.2108, "step": 5156 }, { "epoch": 0.93, "learning_rate": 1.6019338514368335e-05, "loss": 1.008, "step": 5157 }, { "epoch": 0.93, "learning_rate": 1.6016627507681186e-05, "loss": 0.7527, "step": 5158 }, { "epoch": 0.93, "learning_rate": 1.6013916500994035e-05, "loss": 1.4168, "step": 5159 }, { "epoch": 0.93, "learning_rate": 1.6011205494306887e-05, "loss": 1.2546, "step": 5160 }, { "epoch": 0.93, "learning_rate": 1.600849448761974e-05, "loss": 0.9305, "step": 5161 }, { "epoch": 0.93, "learning_rate": 1.6005783480932587e-05, "loss": 1.1232, "step": 5162 }, { "epoch": 0.93, "learning_rate": 1.600307247424544e-05, "loss": 0.8626, "step": 5163 }, { "epoch": 0.93, "learning_rate": 1.6000361467558284e-05, "loss": 0.843, "step": 5164 }, { "epoch": 0.93, "learning_rate": 1.5997650460871136e-05, "loss": 0.5371, "step": 5165 }, { "epoch": 0.93, "learning_rate": 1.5994939454183988e-05, "loss": 1.3516, "step": 5166 }, { "epoch": 0.93, "learning_rate": 1.5992228447496836e-05, "loss": 1.3652, "step": 5167 }, { "epoch": 0.93, "learning_rate": 1.5989517440809688e-05, "loss": 1.1057, "step": 5168 }, { "epoch": 0.93, "learning_rate": 1.598680643412254e-05, "loss": 0.5487, "step": 5169 }, { "epoch": 0.93, "learning_rate": 1.5984095427435388e-05, "loss": 1.0788, "step": 5170 }, { "epoch": 0.93, "learning_rate": 1.598138442074824e-05, "loss": 0.3918, "step": 5171 }, { "epoch": 0.93, "learning_rate": 1.597867341406109e-05, "loss": 0.9202, "step": 5172 }, { "epoch": 0.93, "learning_rate": 1.5975962407373937e-05, "loss": 0.6084, "step": 5173 }, { "epoch": 0.94, "learning_rate": 1.597325140068679e-05, "loss": 0.7975, "step": 5174 }, { "epoch": 0.94, "learning_rate": 1.5970540393999637e-05, "loss": 0.7279, "step": 5175 }, { "epoch": 0.94, "learning_rate": 1.596782938731249e-05, "loss": 1.3108, "step": 5176 }, { "epoch": 0.94, "learning_rate": 1.596511838062534e-05, "loss": 0.9894, "step": 5177 }, { "epoch": 0.94, "learning_rate": 1.596240737393819e-05, "loss": 0.8208, "step": 5178 }, { "epoch": 0.94, "learning_rate": 1.595969636725104e-05, "loss": 1.0175, "step": 5179 }, { "epoch": 0.94, "learning_rate": 1.595698536056389e-05, "loss": 0.8803, "step": 5180 }, { "epoch": 0.94, "learning_rate": 1.5954274353876738e-05, "loss": 0.7571, "step": 5181 }, { "epoch": 0.94, "learning_rate": 1.595156334718959e-05, "loss": 1.0003, "step": 5182 }, { "epoch": 0.94, "learning_rate": 1.594885234050244e-05, "loss": 1.3807, "step": 5183 }, { "epoch": 0.94, "learning_rate": 1.594614133381529e-05, "loss": 1.1672, "step": 5184 }, { "epoch": 0.94, "learning_rate": 1.5943430327128142e-05, "loss": 0.8342, "step": 5185 }, { "epoch": 0.94, "learning_rate": 1.594071932044099e-05, "loss": 1.871, "step": 5186 }, { "epoch": 0.94, "learning_rate": 1.5938008313753842e-05, "loss": 0.7603, "step": 5187 }, { "epoch": 0.94, "learning_rate": 1.593529730706669e-05, "loss": 1.0958, "step": 5188 }, { "epoch": 0.94, "learning_rate": 1.593258630037954e-05, "loss": 1.0077, "step": 5189 }, { "epoch": 0.94, "learning_rate": 1.592987529369239e-05, "loss": 0.6608, "step": 5190 }, { "epoch": 0.94, "learning_rate": 1.5927164287005243e-05, "loss": 0.8082, "step": 5191 }, { "epoch": 0.94, "learning_rate": 1.592445328031809e-05, "loss": 1.5015, "step": 5192 }, { "epoch": 0.94, "learning_rate": 1.5921742273630943e-05, "loss": 0.993, "step": 5193 }, { "epoch": 0.94, "learning_rate": 1.5919031266943795e-05, "loss": 0.955, "step": 5194 }, { "epoch": 0.94, "learning_rate": 1.5916320260256643e-05, "loss": 1.0259, "step": 5195 }, { "epoch": 0.94, "learning_rate": 1.591360925356949e-05, "loss": 0.8913, "step": 5196 }, { "epoch": 0.94, "learning_rate": 1.591089824688234e-05, "loss": 1.2243, "step": 5197 }, { "epoch": 0.94, "learning_rate": 1.5908187240195192e-05, "loss": 0.6481, "step": 5198 }, { "epoch": 0.94, "learning_rate": 1.5905476233508044e-05, "loss": 0.861, "step": 5199 }, { "epoch": 0.94, "learning_rate": 1.5902765226820892e-05, "loss": 1.3674, "step": 5200 }, { "epoch": 0.94, "learning_rate": 1.5900054220133744e-05, "loss": 0.834, "step": 5201 }, { "epoch": 0.94, "learning_rate": 1.5897343213446596e-05, "loss": 1.2469, "step": 5202 }, { "epoch": 0.94, "learning_rate": 1.5894632206759444e-05, "loss": 1.5884, "step": 5203 }, { "epoch": 0.94, "learning_rate": 1.5891921200072293e-05, "loss": 1.1917, "step": 5204 }, { "epoch": 0.94, "learning_rate": 1.5889210193385144e-05, "loss": 1.4111, "step": 5205 }, { "epoch": 0.94, "learning_rate": 1.5886499186697993e-05, "loss": 1.1558, "step": 5206 }, { "epoch": 0.94, "learning_rate": 1.5883788180010845e-05, "loss": 0.9956, "step": 5207 }, { "epoch": 0.94, "learning_rate": 1.5881077173323693e-05, "loss": 0.8855, "step": 5208 }, { "epoch": 0.94, "learning_rate": 1.5878366166636545e-05, "loss": 1.1025, "step": 5209 }, { "epoch": 0.94, "learning_rate": 1.5875655159949397e-05, "loss": 0.7603, "step": 5210 }, { "epoch": 0.94, "learning_rate": 1.5872944153262245e-05, "loss": 1.1254, "step": 5211 }, { "epoch": 0.94, "learning_rate": 1.5870233146575094e-05, "loss": 0.4683, "step": 5212 }, { "epoch": 0.94, "learning_rate": 1.5867522139887946e-05, "loss": 0.5453, "step": 5213 }, { "epoch": 0.94, "learning_rate": 1.5864811133200794e-05, "loss": 0.8735, "step": 5214 }, { "epoch": 0.94, "learning_rate": 1.5862100126513646e-05, "loss": 1.1454, "step": 5215 }, { "epoch": 0.94, "learning_rate": 1.5859389119826498e-05, "loss": 0.8239, "step": 5216 }, { "epoch": 0.94, "learning_rate": 1.5856678113139346e-05, "loss": 0.8279, "step": 5217 }, { "epoch": 0.94, "learning_rate": 1.5853967106452198e-05, "loss": 1.2636, "step": 5218 }, { "epoch": 0.94, "learning_rate": 1.5851256099765046e-05, "loss": 0.662, "step": 5219 }, { "epoch": 0.94, "learning_rate": 1.5848545093077895e-05, "loss": 0.9017, "step": 5220 }, { "epoch": 0.94, "learning_rate": 1.5845834086390747e-05, "loss": 0.682, "step": 5221 }, { "epoch": 0.94, "learning_rate": 1.5843123079703595e-05, "loss": 0.8926, "step": 5222 }, { "epoch": 0.94, "learning_rate": 1.5840412073016447e-05, "loss": 0.9322, "step": 5223 }, { "epoch": 0.94, "learning_rate": 1.58377010663293e-05, "loss": 0.8888, "step": 5224 }, { "epoch": 0.94, "learning_rate": 1.5834990059642147e-05, "loss": 1.0141, "step": 5225 }, { "epoch": 0.94, "learning_rate": 1.5832279052955e-05, "loss": 1.6533, "step": 5226 }, { "epoch": 0.94, "learning_rate": 1.582956804626785e-05, "loss": 1.4176, "step": 5227 }, { "epoch": 0.94, "learning_rate": 1.5826857039580696e-05, "loss": 1.4934, "step": 5228 }, { "epoch": 0.95, "learning_rate": 1.5824146032893548e-05, "loss": 1.3431, "step": 5229 }, { "epoch": 0.95, "learning_rate": 1.5821435026206396e-05, "loss": 1.3557, "step": 5230 }, { "epoch": 0.95, "learning_rate": 1.5818724019519248e-05, "loss": 1.0059, "step": 5231 }, { "epoch": 0.95, "learning_rate": 1.58160130128321e-05, "loss": 0.9429, "step": 5232 }, { "epoch": 0.95, "learning_rate": 1.5813302006144948e-05, "loss": 0.5794, "step": 5233 }, { "epoch": 0.95, "learning_rate": 1.58105909994578e-05, "loss": 1.4431, "step": 5234 }, { "epoch": 0.95, "learning_rate": 1.5807879992770652e-05, "loss": 1.45, "step": 5235 }, { "epoch": 0.95, "learning_rate": 1.5805168986083497e-05, "loss": 0.6255, "step": 5236 }, { "epoch": 0.95, "learning_rate": 1.580245797939635e-05, "loss": 1.3879, "step": 5237 }, { "epoch": 0.95, "learning_rate": 1.57997469727092e-05, "loss": 1.1289, "step": 5238 }, { "epoch": 0.95, "learning_rate": 1.579703596602205e-05, "loss": 0.7663, "step": 5239 }, { "epoch": 0.95, "learning_rate": 1.57943249593349e-05, "loss": 1.2135, "step": 5240 }, { "epoch": 0.95, "learning_rate": 1.579161395264775e-05, "loss": 0.6009, "step": 5241 }, { "epoch": 0.95, "learning_rate": 1.57889029459606e-05, "loss": 0.6908, "step": 5242 }, { "epoch": 0.95, "learning_rate": 1.5786191939273453e-05, "loss": 0.8083, "step": 5243 }, { "epoch": 0.95, "learning_rate": 1.5783480932586298e-05, "loss": 0.9709, "step": 5244 }, { "epoch": 0.95, "learning_rate": 1.578076992589915e-05, "loss": 1.0638, "step": 5245 }, { "epoch": 0.95, "learning_rate": 1.5778058919212e-05, "loss": 0.5968, "step": 5246 }, { "epoch": 0.95, "learning_rate": 1.577534791252485e-05, "loss": 1.1618, "step": 5247 }, { "epoch": 0.95, "learning_rate": 1.5772636905837702e-05, "loss": 0.8083, "step": 5248 }, { "epoch": 0.95, "learning_rate": 1.5769925899150554e-05, "loss": 1.008, "step": 5249 }, { "epoch": 0.95, "learning_rate": 1.5767214892463402e-05, "loss": 1.0994, "step": 5250 }, { "epoch": 0.95, "learning_rate": 1.5764503885776254e-05, "loss": 1.0966, "step": 5251 }, { "epoch": 0.95, "learning_rate": 1.5761792879089102e-05, "loss": 1.9729, "step": 5252 }, { "epoch": 0.95, "learning_rate": 1.575908187240195e-05, "loss": 1.145, "step": 5253 }, { "epoch": 0.95, "learning_rate": 1.5756370865714803e-05, "loss": 1.1135, "step": 5254 }, { "epoch": 0.95, "learning_rate": 1.575365985902765e-05, "loss": 0.7284, "step": 5255 }, { "epoch": 0.95, "learning_rate": 1.5750948852340503e-05, "loss": 0.6484, "step": 5256 }, { "epoch": 0.95, "learning_rate": 1.5748237845653355e-05, "loss": 0.7728, "step": 5257 }, { "epoch": 0.95, "learning_rate": 1.5745526838966203e-05, "loss": 1.3438, "step": 5258 }, { "epoch": 0.95, "learning_rate": 1.5742815832279055e-05, "loss": 0.643, "step": 5259 }, { "epoch": 0.95, "learning_rate": 1.5740104825591904e-05, "loss": 0.6385, "step": 5260 }, { "epoch": 0.95, "learning_rate": 1.5737393818904752e-05, "loss": 0.9823, "step": 5261 }, { "epoch": 0.95, "learning_rate": 1.5734682812217604e-05, "loss": 0.9718, "step": 5262 }, { "epoch": 0.95, "learning_rate": 1.5731971805530456e-05, "loss": 1.5641, "step": 5263 }, { "epoch": 0.95, "learning_rate": 1.5729260798843304e-05, "loss": 1.2556, "step": 5264 }, { "epoch": 0.95, "learning_rate": 1.5726549792156156e-05, "loss": 1.664, "step": 5265 }, { "epoch": 0.95, "learning_rate": 1.5723838785469004e-05, "loss": 0.774, "step": 5266 }, { "epoch": 0.95, "learning_rate": 1.5721127778781856e-05, "loss": 0.9493, "step": 5267 }, { "epoch": 0.95, "learning_rate": 1.5718416772094705e-05, "loss": 0.699, "step": 5268 }, { "epoch": 0.95, "learning_rate": 1.5715705765407553e-05, "loss": 1.0463, "step": 5269 }, { "epoch": 0.95, "learning_rate": 1.5712994758720405e-05, "loss": 1.5507, "step": 5270 }, { "epoch": 0.95, "learning_rate": 1.5710283752033257e-05, "loss": 1.4393, "step": 5271 }, { "epoch": 0.95, "learning_rate": 1.5707572745346105e-05, "loss": 1.4551, "step": 5272 }, { "epoch": 0.95, "learning_rate": 1.5704861738658957e-05, "loss": 0.7697, "step": 5273 }, { "epoch": 0.95, "learning_rate": 1.570215073197181e-05, "loss": 0.7313, "step": 5274 }, { "epoch": 0.95, "learning_rate": 1.5699439725284657e-05, "loss": 1.7024, "step": 5275 }, { "epoch": 0.95, "learning_rate": 1.5696728718597506e-05, "loss": 0.7787, "step": 5276 }, { "epoch": 0.95, "learning_rate": 1.5694017711910354e-05, "loss": 0.7197, "step": 5277 }, { "epoch": 0.95, "learning_rate": 1.5691306705223206e-05, "loss": 1.4819, "step": 5278 }, { "epoch": 0.95, "learning_rate": 1.5688595698536058e-05, "loss": 1.0557, "step": 5279 }, { "epoch": 0.95, "learning_rate": 1.5685884691848906e-05, "loss": 1.2803, "step": 5280 }, { "epoch": 0.95, "learning_rate": 1.5683173685161758e-05, "loss": 1.2151, "step": 5281 }, { "epoch": 0.95, "learning_rate": 1.568046267847461e-05, "loss": 0.5734, "step": 5282 }, { "epoch": 0.95, "learning_rate": 1.5677751671787455e-05, "loss": 0.7704, "step": 5283 }, { "epoch": 0.95, "learning_rate": 1.5675040665100307e-05, "loss": 1.0279, "step": 5284 }, { "epoch": 0.96, "learning_rate": 1.567232965841316e-05, "loss": 0.7368, "step": 5285 }, { "epoch": 0.96, "learning_rate": 1.5669618651726007e-05, "loss": 0.9594, "step": 5286 }, { "epoch": 0.96, "learning_rate": 1.566690764503886e-05, "loss": 1.3171, "step": 5287 }, { "epoch": 0.96, "learning_rate": 1.5664196638351707e-05, "loss": 1.0, "step": 5288 }, { "epoch": 0.96, "learning_rate": 1.566148563166456e-05, "loss": 1.0762, "step": 5289 }, { "epoch": 0.96, "learning_rate": 1.565877462497741e-05, "loss": 1.7014, "step": 5290 }, { "epoch": 0.96, "learning_rate": 1.5656063618290256e-05, "loss": 1.0496, "step": 5291 }, { "epoch": 0.96, "learning_rate": 1.5653352611603108e-05, "loss": 0.9807, "step": 5292 }, { "epoch": 0.96, "learning_rate": 1.565064160491596e-05, "loss": 1.4817, "step": 5293 }, { "epoch": 0.96, "learning_rate": 1.5647930598228808e-05, "loss": 0.6919, "step": 5294 }, { "epoch": 0.96, "learning_rate": 1.564521959154166e-05, "loss": 1.7316, "step": 5295 }, { "epoch": 0.96, "learning_rate": 1.5642508584854512e-05, "loss": 1.4623, "step": 5296 }, { "epoch": 0.96, "learning_rate": 1.563979757816736e-05, "loss": 0.8061, "step": 5297 }, { "epoch": 0.96, "learning_rate": 1.5637086571480212e-05, "loss": 1.277, "step": 5298 }, { "epoch": 0.96, "learning_rate": 1.5634375564793057e-05, "loss": 0.6344, "step": 5299 }, { "epoch": 0.96, "learning_rate": 1.563166455810591e-05, "loss": 1.0242, "step": 5300 }, { "epoch": 0.96, "learning_rate": 1.562895355141876e-05, "loss": 1.6488, "step": 5301 }, { "epoch": 0.96, "learning_rate": 1.562624254473161e-05, "loss": 0.6259, "step": 5302 }, { "epoch": 0.96, "learning_rate": 1.562353153804446e-05, "loss": 1.5117, "step": 5303 }, { "epoch": 0.96, "learning_rate": 1.5620820531357313e-05, "loss": 1.708, "step": 5304 }, { "epoch": 0.96, "learning_rate": 1.561810952467016e-05, "loss": 0.6163, "step": 5305 }, { "epoch": 0.96, "learning_rate": 1.5615398517983013e-05, "loss": 0.6516, "step": 5306 }, { "epoch": 0.96, "learning_rate": 1.561268751129586e-05, "loss": 0.5647, "step": 5307 }, { "epoch": 0.96, "learning_rate": 1.560997650460871e-05, "loss": 1.1277, "step": 5308 }, { "epoch": 0.96, "learning_rate": 1.5607265497921562e-05, "loss": 1.3995, "step": 5309 }, { "epoch": 0.96, "learning_rate": 1.560455449123441e-05, "loss": 0.7204, "step": 5310 }, { "epoch": 0.96, "learning_rate": 1.5601843484547262e-05, "loss": 0.9041, "step": 5311 }, { "epoch": 0.96, "learning_rate": 1.5599132477860114e-05, "loss": 1.0791, "step": 5312 }, { "epoch": 0.96, "learning_rate": 1.5596421471172962e-05, "loss": 0.6667, "step": 5313 }, { "epoch": 0.96, "learning_rate": 1.5593710464485814e-05, "loss": 1.2845, "step": 5314 }, { "epoch": 0.96, "learning_rate": 1.5590999457798663e-05, "loss": 0.8687, "step": 5315 }, { "epoch": 0.96, "learning_rate": 1.558828845111151e-05, "loss": 0.8094, "step": 5316 }, { "epoch": 0.96, "learning_rate": 1.5585577444424363e-05, "loss": 0.7386, "step": 5317 }, { "epoch": 0.96, "learning_rate": 1.5582866437737215e-05, "loss": 0.6185, "step": 5318 }, { "epoch": 0.96, "learning_rate": 1.5580155431050063e-05, "loss": 0.92, "step": 5319 }, { "epoch": 0.96, "learning_rate": 1.5577444424362915e-05, "loss": 1.412, "step": 5320 }, { "epoch": 0.96, "learning_rate": 1.5574733417675763e-05, "loss": 0.6173, "step": 5321 }, { "epoch": 0.96, "learning_rate": 1.5572022410988615e-05, "loss": 0.593, "step": 5322 }, { "epoch": 0.96, "learning_rate": 1.5569311404301464e-05, "loss": 0.6673, "step": 5323 }, { "epoch": 0.96, "learning_rate": 1.5566600397614312e-05, "loss": 1.0351, "step": 5324 }, { "epoch": 0.96, "learning_rate": 1.5563889390927164e-05, "loss": 1.5973, "step": 5325 }, { "epoch": 0.96, "learning_rate": 1.5561178384240016e-05, "loss": 1.0423, "step": 5326 }, { "epoch": 0.96, "learning_rate": 1.5558467377552864e-05, "loss": 0.4582, "step": 5327 }, { "epoch": 0.96, "learning_rate": 1.5555756370865716e-05, "loss": 1.0236, "step": 5328 }, { "epoch": 0.96, "learning_rate": 1.5553045364178568e-05, "loss": 1.0174, "step": 5329 }, { "epoch": 0.96, "learning_rate": 1.5550334357491416e-05, "loss": 0.9554, "step": 5330 }, { "epoch": 0.96, "learning_rate": 1.5547623350804265e-05, "loss": 1.1962, "step": 5331 }, { "epoch": 0.96, "learning_rate": 1.5544912344117113e-05, "loss": 0.8394, "step": 5332 }, { "epoch": 0.96, "learning_rate": 1.5542201337429965e-05, "loss": 1.0186, "step": 5333 }, { "epoch": 0.96, "learning_rate": 1.5539490330742817e-05, "loss": 0.7298, "step": 5334 }, { "epoch": 0.96, "learning_rate": 1.5536779324055665e-05, "loss": 0.9612, "step": 5335 }, { "epoch": 0.96, "learning_rate": 1.5534068317368517e-05, "loss": 0.6788, "step": 5336 }, { "epoch": 0.96, "learning_rate": 1.553135731068137e-05, "loss": 1.0405, "step": 5337 }, { "epoch": 0.96, "learning_rate": 1.5528646303994217e-05, "loss": 1.1887, "step": 5338 }, { "epoch": 0.96, "learning_rate": 1.5525935297307066e-05, "loss": 1.4203, "step": 5339 }, { "epoch": 0.97, "learning_rate": 1.5523224290619918e-05, "loss": 1.46, "step": 5340 }, { "epoch": 0.97, "learning_rate": 1.5520513283932766e-05, "loss": 0.7211, "step": 5341 }, { "epoch": 0.97, "learning_rate": 1.5517802277245618e-05, "loss": 1.9449, "step": 5342 }, { "epoch": 0.97, "learning_rate": 1.5515091270558466e-05, "loss": 1.0376, "step": 5343 }, { "epoch": 0.97, "learning_rate": 1.5512380263871318e-05, "loss": 0.7292, "step": 5344 }, { "epoch": 0.97, "learning_rate": 1.550966925718417e-05, "loss": 0.8157, "step": 5345 }, { "epoch": 0.97, "learning_rate": 1.550695825049702e-05, "loss": 0.5302, "step": 5346 }, { "epoch": 0.97, "learning_rate": 1.5504247243809867e-05, "loss": 0.4047, "step": 5347 }, { "epoch": 0.97, "learning_rate": 1.550153623712272e-05, "loss": 1.1466, "step": 5348 }, { "epoch": 0.97, "learning_rate": 1.5498825230435567e-05, "loss": 1.1015, "step": 5349 }, { "epoch": 0.97, "learning_rate": 1.549611422374842e-05, "loss": 1.0982, "step": 5350 }, { "epoch": 0.97, "learning_rate": 1.549340321706127e-05, "loss": 0.4668, "step": 5351 }, { "epoch": 0.97, "learning_rate": 1.549069221037412e-05, "loss": 1.2088, "step": 5352 }, { "epoch": 0.97, "learning_rate": 1.548798120368697e-05, "loss": 0.8883, "step": 5353 }, { "epoch": 0.97, "learning_rate": 1.548527019699982e-05, "loss": 0.7752, "step": 5354 }, { "epoch": 0.97, "learning_rate": 1.5482559190312668e-05, "loss": 0.5851, "step": 5355 }, { "epoch": 0.97, "learning_rate": 1.547984818362552e-05, "loss": 0.8657, "step": 5356 }, { "epoch": 0.97, "learning_rate": 1.5477137176938368e-05, "loss": 0.9864, "step": 5357 }, { "epoch": 0.97, "learning_rate": 1.547442617025122e-05, "loss": 1.3521, "step": 5358 }, { "epoch": 0.97, "learning_rate": 1.5471715163564072e-05, "loss": 1.0191, "step": 5359 }, { "epoch": 0.97, "learning_rate": 1.546900415687692e-05, "loss": 2.7395, "step": 5360 }, { "epoch": 0.97, "learning_rate": 1.5466293150189772e-05, "loss": 0.8937, "step": 5361 }, { "epoch": 0.97, "learning_rate": 1.5463582143502624e-05, "loss": 0.5669, "step": 5362 }, { "epoch": 0.97, "learning_rate": 1.546087113681547e-05, "loss": 1.2129, "step": 5363 }, { "epoch": 0.97, "learning_rate": 1.545816013012832e-05, "loss": 1.0425, "step": 5364 }, { "epoch": 0.97, "learning_rate": 1.545544912344117e-05, "loss": 0.8021, "step": 5365 }, { "epoch": 0.97, "learning_rate": 1.545273811675402e-05, "loss": 0.6912, "step": 5366 }, { "epoch": 0.97, "learning_rate": 1.5450027110066873e-05, "loss": 0.4094, "step": 5367 }, { "epoch": 0.97, "learning_rate": 1.544731610337972e-05, "loss": 2.218, "step": 5368 }, { "epoch": 0.97, "learning_rate": 1.5444605096692573e-05, "loss": 1.079, "step": 5369 }, { "epoch": 0.97, "learning_rate": 1.5441894090005425e-05, "loss": 1.1251, "step": 5370 }, { "epoch": 0.97, "learning_rate": 1.543918308331827e-05, "loss": 0.4502, "step": 5371 }, { "epoch": 0.97, "learning_rate": 1.5436472076631122e-05, "loss": 0.8557, "step": 5372 }, { "epoch": 0.97, "learning_rate": 1.5433761069943974e-05, "loss": 0.471, "step": 5373 }, { "epoch": 0.97, "learning_rate": 1.5431050063256822e-05, "loss": 0.6695, "step": 5374 }, { "epoch": 0.97, "learning_rate": 1.5428339056569674e-05, "loss": 1.0676, "step": 5375 }, { "epoch": 0.97, "learning_rate": 1.5425628049882523e-05, "loss": 0.4958, "step": 5376 }, { "epoch": 0.97, "learning_rate": 1.5422917043195374e-05, "loss": 0.7117, "step": 5377 }, { "epoch": 0.97, "learning_rate": 1.5420206036508226e-05, "loss": 0.9823, "step": 5378 }, { "epoch": 0.97, "learning_rate": 1.541749502982107e-05, "loss": 0.8855, "step": 5379 }, { "epoch": 0.97, "learning_rate": 1.5414784023133923e-05, "loss": 0.9955, "step": 5380 }, { "epoch": 0.97, "learning_rate": 1.5412073016446775e-05, "loss": 1.1487, "step": 5381 }, { "epoch": 0.97, "learning_rate": 1.5409362009759623e-05, "loss": 1.26, "step": 5382 }, { "epoch": 0.97, "learning_rate": 1.5406651003072475e-05, "loss": 1.321, "step": 5383 }, { "epoch": 0.97, "learning_rate": 1.5403939996385327e-05, "loss": 1.6027, "step": 5384 }, { "epoch": 0.97, "learning_rate": 1.5401228989698175e-05, "loss": 1.0719, "step": 5385 }, { "epoch": 0.97, "learning_rate": 1.5398517983011027e-05, "loss": 0.885, "step": 5386 }, { "epoch": 0.97, "learning_rate": 1.5395806976323876e-05, "loss": 0.9898, "step": 5387 }, { "epoch": 0.97, "learning_rate": 1.5393095969636724e-05, "loss": 1.4872, "step": 5388 }, { "epoch": 0.97, "learning_rate": 1.5390384962949576e-05, "loss": 0.8129, "step": 5389 }, { "epoch": 0.97, "learning_rate": 1.5387673956262424e-05, "loss": 0.9075, "step": 5390 }, { "epoch": 0.97, "learning_rate": 1.5384962949575276e-05, "loss": 0.8017, "step": 5391 }, { "epoch": 0.97, "learning_rate": 1.5382251942888128e-05, "loss": 0.9366, "step": 5392 }, { "epoch": 0.97, "learning_rate": 1.5379540936200976e-05, "loss": 0.775, "step": 5393 }, { "epoch": 0.97, "learning_rate": 1.537682992951383e-05, "loss": 1.0466, "step": 5394 }, { "epoch": 0.98, "learning_rate": 1.5374118922826677e-05, "loss": 0.8573, "step": 5395 }, { "epoch": 0.98, "learning_rate": 1.5371407916139525e-05, "loss": 1.2464, "step": 5396 }, { "epoch": 0.98, "learning_rate": 1.5368696909452377e-05, "loss": 1.4308, "step": 5397 }, { "epoch": 0.98, "learning_rate": 1.536598590276523e-05, "loss": 0.7171, "step": 5398 }, { "epoch": 0.98, "learning_rate": 1.5363274896078077e-05, "loss": 0.5372, "step": 5399 }, { "epoch": 0.98, "learning_rate": 1.536056388939093e-05, "loss": 1.8043, "step": 5400 }, { "epoch": 0.98, "learning_rate": 1.5357852882703778e-05, "loss": 1.1253, "step": 5401 }, { "epoch": 0.98, "learning_rate": 1.535514187601663e-05, "loss": 0.6049, "step": 5402 }, { "epoch": 0.98, "learning_rate": 1.5352430869329478e-05, "loss": 1.4478, "step": 5403 }, { "epoch": 0.98, "learning_rate": 1.5349719862642326e-05, "loss": 0.533, "step": 5404 }, { "epoch": 0.98, "learning_rate": 1.5347008855955178e-05, "loss": 1.0407, "step": 5405 }, { "epoch": 0.98, "learning_rate": 1.534429784926803e-05, "loss": 0.9305, "step": 5406 }, { "epoch": 0.98, "learning_rate": 1.534158684258088e-05, "loss": 1.0803, "step": 5407 }, { "epoch": 0.98, "learning_rate": 1.533887583589373e-05, "loss": 1.1562, "step": 5408 }, { "epoch": 0.98, "learning_rate": 1.5336164829206582e-05, "loss": 1.1156, "step": 5409 }, { "epoch": 0.98, "learning_rate": 1.5333453822519427e-05, "loss": 1.5203, "step": 5410 }, { "epoch": 0.98, "learning_rate": 1.533074281583228e-05, "loss": 1.5108, "step": 5411 }, { "epoch": 0.98, "learning_rate": 1.5328031809145127e-05, "loss": 1.0622, "step": 5412 }, { "epoch": 0.98, "learning_rate": 1.532532080245798e-05, "loss": 1.1097, "step": 5413 }, { "epoch": 0.98, "learning_rate": 1.532260979577083e-05, "loss": 0.5671, "step": 5414 }, { "epoch": 0.98, "learning_rate": 1.531989878908368e-05, "loss": 0.7148, "step": 5415 }, { "epoch": 0.98, "learning_rate": 1.531718778239653e-05, "loss": 0.7708, "step": 5416 }, { "epoch": 0.98, "learning_rate": 1.5314476775709383e-05, "loss": 1.1256, "step": 5417 }, { "epoch": 0.98, "learning_rate": 1.5311765769022228e-05, "loss": 1.6284, "step": 5418 }, { "epoch": 0.98, "learning_rate": 1.530905476233508e-05, "loss": 0.876, "step": 5419 }, { "epoch": 0.98, "learning_rate": 1.5306343755647932e-05, "loss": 0.8988, "step": 5420 }, { "epoch": 0.98, "learning_rate": 1.530363274896078e-05, "loss": 0.7363, "step": 5421 }, { "epoch": 0.98, "learning_rate": 1.5300921742273632e-05, "loss": 0.5934, "step": 5422 }, { "epoch": 0.98, "learning_rate": 1.529821073558648e-05, "loss": 0.7768, "step": 5423 }, { "epoch": 0.98, "learning_rate": 1.5295499728899332e-05, "loss": 1.8234, "step": 5424 }, { "epoch": 0.98, "learning_rate": 1.5292788722212184e-05, "loss": 0.9062, "step": 5425 }, { "epoch": 0.98, "learning_rate": 1.529007771552503e-05, "loss": 0.8457, "step": 5426 }, { "epoch": 0.98, "learning_rate": 1.528736670883788e-05, "loss": 0.6737, "step": 5427 }, { "epoch": 0.98, "learning_rate": 1.5284655702150733e-05, "loss": 1.9399, "step": 5428 }, { "epoch": 0.98, "learning_rate": 1.528194469546358e-05, "loss": 0.4403, "step": 5429 }, { "epoch": 0.98, "learning_rate": 1.5279233688776433e-05, "loss": 1.6649, "step": 5430 }, { "epoch": 0.98, "learning_rate": 1.5276522682089285e-05, "loss": 1.3973, "step": 5431 }, { "epoch": 0.98, "learning_rate": 1.5273811675402133e-05, "loss": 2.2588, "step": 5432 }, { "epoch": 0.98, "learning_rate": 1.5271100668714985e-05, "loss": 1.1527, "step": 5433 }, { "epoch": 0.98, "learning_rate": 1.526838966202783e-05, "loss": 0.6989, "step": 5434 }, { "epoch": 0.98, "learning_rate": 1.5265678655340682e-05, "loss": 1.8136, "step": 5435 }, { "epoch": 0.98, "learning_rate": 1.5262967648653534e-05, "loss": 1.8599, "step": 5436 }, { "epoch": 0.98, "learning_rate": 1.5260256641966382e-05, "loss": 1.0017, "step": 5437 }, { "epoch": 0.98, "learning_rate": 1.5257545635279234e-05, "loss": 1.0155, "step": 5438 }, { "epoch": 0.98, "learning_rate": 1.5254834628592084e-05, "loss": 0.8558, "step": 5439 }, { "epoch": 0.98, "learning_rate": 1.5252123621904936e-05, "loss": 1.1033, "step": 5440 }, { "epoch": 0.98, "learning_rate": 1.5249412615217786e-05, "loss": 0.5001, "step": 5441 }, { "epoch": 0.98, "learning_rate": 1.5246701608530633e-05, "loss": 0.9721, "step": 5442 }, { "epoch": 0.98, "learning_rate": 1.5243990601843483e-05, "loss": 0.6246, "step": 5443 }, { "epoch": 0.98, "learning_rate": 1.5241279595156335e-05, "loss": 0.8891, "step": 5444 }, { "epoch": 0.98, "learning_rate": 1.5238568588469185e-05, "loss": 1.2004, "step": 5445 }, { "epoch": 0.98, "learning_rate": 1.5235857581782035e-05, "loss": 0.5398, "step": 5446 }, { "epoch": 0.98, "learning_rate": 1.5233146575094885e-05, "loss": 0.7947, "step": 5447 }, { "epoch": 0.98, "learning_rate": 1.5230435568407737e-05, "loss": 1.2791, "step": 5448 }, { "epoch": 0.98, "learning_rate": 1.5227724561720587e-05, "loss": 1.4077, "step": 5449 }, { "epoch": 0.98, "learning_rate": 1.5225013555033434e-05, "loss": 0.6004, "step": 5450 }, { "epoch": 0.99, "learning_rate": 1.5222302548346286e-05, "loss": 0.9329, "step": 5451 }, { "epoch": 0.99, "learning_rate": 1.5219591541659136e-05, "loss": 1.0459, "step": 5452 }, { "epoch": 0.99, "learning_rate": 1.5216880534971986e-05, "loss": 0.7665, "step": 5453 }, { "epoch": 0.99, "learning_rate": 1.5214169528284836e-05, "loss": 1.1639, "step": 5454 }, { "epoch": 0.99, "learning_rate": 1.5211458521597688e-05, "loss": 1.2977, "step": 5455 }, { "epoch": 0.99, "learning_rate": 1.5208747514910538e-05, "loss": 1.0295, "step": 5456 }, { "epoch": 0.99, "learning_rate": 1.5206036508223388e-05, "loss": 1.9028, "step": 5457 }, { "epoch": 0.99, "learning_rate": 1.5203325501536235e-05, "loss": 0.9741, "step": 5458 }, { "epoch": 0.99, "learning_rate": 1.5200614494849087e-05, "loss": 0.8317, "step": 5459 }, { "epoch": 0.99, "learning_rate": 1.5197903488161937e-05, "loss": 1.0806, "step": 5460 }, { "epoch": 0.99, "learning_rate": 1.5195192481474787e-05, "loss": 0.9996, "step": 5461 }, { "epoch": 0.99, "learning_rate": 1.5192481474787639e-05, "loss": 0.6801, "step": 5462 }, { "epoch": 0.99, "learning_rate": 1.518977046810049e-05, "loss": 0.8916, "step": 5463 }, { "epoch": 0.99, "learning_rate": 1.518705946141334e-05, "loss": 0.6978, "step": 5464 }, { "epoch": 0.99, "learning_rate": 1.518434845472619e-05, "loss": 1.0579, "step": 5465 }, { "epoch": 0.99, "learning_rate": 1.5181637448039038e-05, "loss": 1.2561, "step": 5466 }, { "epoch": 0.99, "learning_rate": 1.5178926441351888e-05, "loss": 0.8293, "step": 5467 }, { "epoch": 0.99, "learning_rate": 1.5176215434664738e-05, "loss": 1.2578, "step": 5468 }, { "epoch": 0.99, "learning_rate": 1.5173504427977588e-05, "loss": 1.601, "step": 5469 }, { "epoch": 0.99, "learning_rate": 1.517079342129044e-05, "loss": 1.1146, "step": 5470 }, { "epoch": 0.99, "learning_rate": 1.516808241460329e-05, "loss": 1.0887, "step": 5471 }, { "epoch": 0.99, "learning_rate": 1.516537140791614e-05, "loss": 0.7432, "step": 5472 }, { "epoch": 0.99, "learning_rate": 1.5162660401228992e-05, "loss": 0.8498, "step": 5473 }, { "epoch": 0.99, "learning_rate": 1.5159949394541839e-05, "loss": 0.6755, "step": 5474 }, { "epoch": 0.99, "learning_rate": 1.515723838785469e-05, "loss": 0.8192, "step": 5475 }, { "epoch": 0.99, "learning_rate": 1.515452738116754e-05, "loss": 0.9048, "step": 5476 }, { "epoch": 0.99, "learning_rate": 1.5151816374480391e-05, "loss": 0.9831, "step": 5477 }, { "epoch": 0.99, "learning_rate": 1.5149105367793241e-05, "loss": 0.7021, "step": 5478 }, { "epoch": 0.99, "learning_rate": 1.5146394361106091e-05, "loss": 0.625, "step": 5479 }, { "epoch": 0.99, "learning_rate": 1.5143683354418942e-05, "loss": 0.758, "step": 5480 }, { "epoch": 0.99, "learning_rate": 1.5140972347731793e-05, "loss": 0.6891, "step": 5481 }, { "epoch": 0.99, "learning_rate": 1.513826134104464e-05, "loss": 1.2584, "step": 5482 }, { "epoch": 0.99, "learning_rate": 1.513555033435749e-05, "loss": 0.7613, "step": 5483 }, { "epoch": 0.99, "learning_rate": 1.5132839327670342e-05, "loss": 1.3557, "step": 5484 }, { "epoch": 0.99, "learning_rate": 1.5130128320983192e-05, "loss": 1.1059, "step": 5485 }, { "epoch": 0.99, "learning_rate": 1.5127417314296042e-05, "loss": 1.1114, "step": 5486 }, { "epoch": 0.99, "learning_rate": 1.5124706307608893e-05, "loss": 1.1719, "step": 5487 }, { "epoch": 0.99, "learning_rate": 1.5121995300921744e-05, "loss": 0.9474, "step": 5488 }, { "epoch": 0.99, "learning_rate": 1.5119284294234594e-05, "loss": 1.5497, "step": 5489 }, { "epoch": 0.99, "learning_rate": 1.5116573287547441e-05, "loss": 1.2741, "step": 5490 }, { "epoch": 0.99, "learning_rate": 1.5113862280860293e-05, "loss": 0.9976, "step": 5491 }, { "epoch": 0.99, "learning_rate": 1.5111151274173143e-05, "loss": 0.6589, "step": 5492 }, { "epoch": 0.99, "learning_rate": 1.5108440267485993e-05, "loss": 0.5756, "step": 5493 }, { "epoch": 0.99, "learning_rate": 1.5105729260798843e-05, "loss": 1.1783, "step": 5494 }, { "epoch": 0.99, "learning_rate": 1.5103018254111695e-05, "loss": 0.7142, "step": 5495 }, { "epoch": 0.99, "learning_rate": 1.5100307247424545e-05, "loss": 1.3783, "step": 5496 }, { "epoch": 0.99, "learning_rate": 1.5097596240737396e-05, "loss": 0.805, "step": 5497 }, { "epoch": 0.99, "learning_rate": 1.5094885234050242e-05, "loss": 1.2242, "step": 5498 }, { "epoch": 0.99, "learning_rate": 1.5092174227363094e-05, "loss": 1.2519, "step": 5499 }, { "epoch": 0.99, "learning_rate": 1.5089463220675944e-05, "loss": 1.1809, "step": 5500 }, { "epoch": 0.99, "learning_rate": 1.5086752213988794e-05, "loss": 1.1696, "step": 5501 }, { "epoch": 0.99, "learning_rate": 1.5084041207301646e-05, "loss": 1.3307, "step": 5502 }, { "epoch": 0.99, "learning_rate": 1.5081330200614496e-05, "loss": 1.555, "step": 5503 }, { "epoch": 0.99, "learning_rate": 1.5078619193927346e-05, "loss": 1.0898, "step": 5504 }, { "epoch": 0.99, "learning_rate": 1.5075908187240197e-05, "loss": 1.5367, "step": 5505 }, { "epoch": 1.0, "learning_rate": 1.5073197180553045e-05, "loss": 0.9625, "step": 5506 }, { "epoch": 1.0, "learning_rate": 1.5070486173865895e-05, "loss": 0.9752, "step": 5507 }, { "epoch": 1.0, "learning_rate": 1.5067775167178745e-05, "loss": 1.1187, "step": 5508 }, { "epoch": 1.0, "learning_rate": 1.5065064160491595e-05, "loss": 1.2868, "step": 5509 }, { "epoch": 1.0, "learning_rate": 1.5062353153804447e-05, "loss": 1.4928, "step": 5510 }, { "epoch": 1.0, "learning_rate": 1.5059642147117297e-05, "loss": 1.0847, "step": 5511 }, { "epoch": 1.0, "learning_rate": 1.5056931140430148e-05, "loss": 0.489, "step": 5512 }, { "epoch": 1.0, "learning_rate": 1.5054220133743e-05, "loss": 1.0393, "step": 5513 }, { "epoch": 1.0, "learning_rate": 1.5051509127055846e-05, "loss": 0.6189, "step": 5514 }, { "epoch": 1.0, "learning_rate": 1.5048798120368696e-05, "loss": 0.4827, "step": 5515 }, { "epoch": 1.0, "learning_rate": 1.5046087113681546e-05, "loss": 1.3404, "step": 5516 }, { "epoch": 1.0, "learning_rate": 1.5043376106994398e-05, "loss": 1.2958, "step": 5517 }, { "epoch": 1.0, "learning_rate": 1.5040665100307248e-05, "loss": 1.3013, "step": 5518 }, { "epoch": 1.0, "learning_rate": 1.5037954093620099e-05, "loss": 0.6884, "step": 5519 }, { "epoch": 1.0, "learning_rate": 1.5035243086932949e-05, "loss": 1.1592, "step": 5520 }, { "epoch": 1.0, "learning_rate": 1.50325320802458e-05, "loss": 0.8021, "step": 5521 }, { "epoch": 1.0, "learning_rate": 1.5029821073558647e-05, "loss": 1.1366, "step": 5522 }, { "epoch": 1.0, "learning_rate": 1.5027110066871497e-05, "loss": 0.7576, "step": 5523 }, { "epoch": 1.0, "learning_rate": 1.502439906018435e-05, "loss": 1.3876, "step": 5524 }, { "epoch": 1.0, "learning_rate": 1.50216880534972e-05, "loss": 0.6201, "step": 5525 }, { "epoch": 1.0, "learning_rate": 1.501897704681005e-05, "loss": 0.8799, "step": 5526 }, { "epoch": 1.0, "learning_rate": 1.50162660401229e-05, "loss": 1.0425, "step": 5527 }, { "epoch": 1.0, "learning_rate": 1.5013555033435751e-05, "loss": 0.8548, "step": 5528 }, { "epoch": 1.0, "learning_rate": 1.5010844026748602e-05, "loss": 1.0247, "step": 5529 }, { "epoch": 1.0, "learning_rate": 1.5008133020061448e-05, "loss": 1.4346, "step": 5530 }, { "epoch": 1.0, "learning_rate": 1.5005422013374298e-05, "loss": 1.0356, "step": 5531 }, { "epoch": 1.0, "learning_rate": 1.500271100668715e-05, "loss": 0.926, "step": 5532 }, { "epoch": 1.0, "learning_rate": 1.5e-05, "loss": 0.5837, "step": 5533 }, { "epoch": 1.0, "learning_rate": 1.499728899331285e-05, "loss": 0.5663, "step": 5534 }, { "epoch": 1.0, "learning_rate": 1.49945779866257e-05, "loss": 0.9003, "step": 5535 }, { "epoch": 1.0, "learning_rate": 1.499186697993855e-05, "loss": 0.4541, "step": 5536 }, { "epoch": 1.0, "learning_rate": 1.4989155973251401e-05, "loss": 0.6022, "step": 5537 }, { "epoch": 1.0, "learning_rate": 1.4986444966564251e-05, "loss": 0.4398, "step": 5538 }, { "epoch": 1.0, "learning_rate": 1.4983733959877101e-05, "loss": 0.8599, "step": 5539 }, { "epoch": 1.0, "learning_rate": 1.4981022953189951e-05, "loss": 1.0225, "step": 5540 }, { "epoch": 1.0, "learning_rate": 1.4978311946502801e-05, "loss": 0.8979, "step": 5541 }, { "epoch": 1.0, "learning_rate": 1.4975600939815652e-05, "loss": 0.9195, "step": 5542 }, { "epoch": 1.0, "learning_rate": 1.4972889933128502e-05, "loss": 0.9469, "step": 5543 }, { "epoch": 1.0, "learning_rate": 1.4970178926441352e-05, "loss": 0.5126, "step": 5544 }, { "epoch": 1.0, "learning_rate": 1.4967467919754202e-05, "loss": 0.5765, "step": 5545 }, { "epoch": 1.0, "learning_rate": 1.4964756913067054e-05, "loss": 0.6599, "step": 5546 }, { "epoch": 1.0, "learning_rate": 1.4962045906379902e-05, "loss": 1.1221, "step": 5547 }, { "epoch": 1.0, "learning_rate": 1.4959334899692752e-05, "loss": 0.9048, "step": 5548 }, { "epoch": 1.0, "learning_rate": 1.4956623893005603e-05, "loss": 1.2028, "step": 5549 }, { "epoch": 1.0, "learning_rate": 1.4953912886318454e-05, "loss": 0.9839, "step": 5550 }, { "epoch": 1.0, "learning_rate": 1.4951201879631303e-05, "loss": 0.6324, "step": 5551 }, { "epoch": 1.0, "learning_rate": 1.4948490872944153e-05, "loss": 0.3832, "step": 5552 }, { "epoch": 1.0, "learning_rate": 1.4945779866257003e-05, "loss": 0.5698, "step": 5553 }, { "epoch": 1.0, "learning_rate": 1.4943068859569855e-05, "loss": 0.9444, "step": 5554 }, { "epoch": 1.0, "learning_rate": 1.4940357852882703e-05, "loss": 0.5652, "step": 5555 }, { "epoch": 1.0, "learning_rate": 1.4937646846195553e-05, "loss": 1.2299, "step": 5556 }, { "epoch": 1.0, "learning_rate": 1.4934935839508405e-05, "loss": 1.1567, "step": 5557 }, { "epoch": 1.0, "learning_rate": 1.4932224832821255e-05, "loss": 0.723, "step": 5558 }, { "epoch": 1.0, "learning_rate": 1.4929513826134104e-05, "loss": 0.3911, "step": 5559 }, { "epoch": 1.0, "learning_rate": 1.4926802819446954e-05, "loss": 0.7779, "step": 5560 }, { "epoch": 1.01, "learning_rate": 1.4924091812759806e-05, "loss": 0.7027, "step": 5561 }, { "epoch": 1.01, "learning_rate": 1.4921380806072656e-05, "loss": 0.7149, "step": 5562 }, { "epoch": 1.01, "learning_rate": 1.4918669799385504e-05, "loss": 0.6883, "step": 5563 }, { "epoch": 1.01, "learning_rate": 1.4915958792698355e-05, "loss": 0.617, "step": 5564 }, { "epoch": 1.01, "learning_rate": 1.4913247786011206e-05, "loss": 0.6131, "step": 5565 }, { "epoch": 1.01, "learning_rate": 1.4910536779324057e-05, "loss": 0.8333, "step": 5566 }, { "epoch": 1.01, "learning_rate": 1.4907825772636905e-05, "loss": 0.4004, "step": 5567 }, { "epoch": 1.01, "learning_rate": 1.4905114765949757e-05, "loss": 0.9322, "step": 5568 }, { "epoch": 1.01, "learning_rate": 1.4902403759262607e-05, "loss": 0.6913, "step": 5569 }, { "epoch": 1.01, "learning_rate": 1.4899692752575457e-05, "loss": 0.9208, "step": 5570 }, { "epoch": 1.01, "learning_rate": 1.4896981745888305e-05, "loss": 1.3039, "step": 5571 }, { "epoch": 1.01, "learning_rate": 1.4894270739201157e-05, "loss": 0.3245, "step": 5572 }, { "epoch": 1.01, "learning_rate": 1.4891559732514007e-05, "loss": 0.9616, "step": 5573 }, { "epoch": 1.01, "learning_rate": 1.4888848725826858e-05, "loss": 0.3705, "step": 5574 }, { "epoch": 1.01, "learning_rate": 1.4886137719139708e-05, "loss": 0.7389, "step": 5575 }, { "epoch": 1.01, "learning_rate": 1.4883426712452558e-05, "loss": 1.1486, "step": 5576 }, { "epoch": 1.01, "learning_rate": 1.4880715705765408e-05, "loss": 0.5826, "step": 5577 }, { "epoch": 1.01, "learning_rate": 1.4878004699078258e-05, "loss": 0.7287, "step": 5578 }, { "epoch": 1.01, "learning_rate": 1.4875293692391108e-05, "loss": 0.7756, "step": 5579 }, { "epoch": 1.01, "learning_rate": 1.4872582685703958e-05, "loss": 1.1944, "step": 5580 }, { "epoch": 1.01, "learning_rate": 1.4869871679016809e-05, "loss": 1.0138, "step": 5581 }, { "epoch": 1.01, "learning_rate": 1.4867160672329659e-05, "loss": 0.4776, "step": 5582 }, { "epoch": 1.01, "learning_rate": 1.4864449665642509e-05, "loss": 0.6291, "step": 5583 }, { "epoch": 1.01, "learning_rate": 1.4861738658955359e-05, "loss": 0.6152, "step": 5584 }, { "epoch": 1.01, "learning_rate": 1.4859027652268209e-05, "loss": 0.9985, "step": 5585 }, { "epoch": 1.01, "learning_rate": 1.4856316645581061e-05, "loss": 1.308, "step": 5586 }, { "epoch": 1.01, "learning_rate": 1.485360563889391e-05, "loss": 1.4361, "step": 5587 }, { "epoch": 1.01, "learning_rate": 1.485089463220676e-05, "loss": 0.4825, "step": 5588 }, { "epoch": 1.01, "learning_rate": 1.484818362551961e-05, "loss": 1.4488, "step": 5589 }, { "epoch": 1.01, "learning_rate": 1.4845472618832461e-05, "loss": 0.6456, "step": 5590 }, { "epoch": 1.01, "learning_rate": 1.484276161214531e-05, "loss": 0.3343, "step": 5591 }, { "epoch": 1.01, "learning_rate": 1.484005060545816e-05, "loss": 0.5991, "step": 5592 }, { "epoch": 1.01, "learning_rate": 1.483733959877101e-05, "loss": 0.6022, "step": 5593 }, { "epoch": 1.01, "learning_rate": 1.4834628592083862e-05, "loss": 0.8182, "step": 5594 }, { "epoch": 1.01, "learning_rate": 1.483191758539671e-05, "loss": 0.8771, "step": 5595 }, { "epoch": 1.01, "learning_rate": 1.482920657870956e-05, "loss": 0.9489, "step": 5596 }, { "epoch": 1.01, "learning_rate": 1.4826495572022412e-05, "loss": 0.6271, "step": 5597 }, { "epoch": 1.01, "learning_rate": 1.4823784565335262e-05, "loss": 0.9622, "step": 5598 }, { "epoch": 1.01, "learning_rate": 1.4821073558648111e-05, "loss": 0.4204, "step": 5599 }, { "epoch": 1.01, "learning_rate": 1.4818362551960961e-05, "loss": 0.6196, "step": 5600 }, { "epoch": 1.01, "learning_rate": 1.4815651545273813e-05, "loss": 0.5093, "step": 5601 }, { "epoch": 1.01, "learning_rate": 1.4812940538586663e-05, "loss": 0.6176, "step": 5602 }, { "epoch": 1.01, "learning_rate": 1.4810229531899511e-05, "loss": 0.312, "step": 5603 }, { "epoch": 1.01, "learning_rate": 1.4807518525212362e-05, "loss": 0.7161, "step": 5604 }, { "epoch": 1.01, "learning_rate": 1.4804807518525213e-05, "loss": 0.5342, "step": 5605 }, { "epoch": 1.01, "learning_rate": 1.4802096511838064e-05, "loss": 0.397, "step": 5606 }, { "epoch": 1.01, "learning_rate": 1.4799385505150912e-05, "loss": 1.6012, "step": 5607 }, { "epoch": 1.01, "learning_rate": 1.4796674498463764e-05, "loss": 0.6011, "step": 5608 }, { "epoch": 1.01, "learning_rate": 1.4793963491776614e-05, "loss": 0.4135, "step": 5609 }, { "epoch": 1.01, "learning_rate": 1.4791252485089464e-05, "loss": 1.0275, "step": 5610 }, { "epoch": 1.01, "learning_rate": 1.4788541478402313e-05, "loss": 1.3031, "step": 5611 }, { "epoch": 1.01, "learning_rate": 1.4785830471715164e-05, "loss": 0.8895, "step": 5612 }, { "epoch": 1.01, "learning_rate": 1.4783119465028015e-05, "loss": 0.4339, "step": 5613 }, { "epoch": 1.01, "learning_rate": 1.4780408458340865e-05, "loss": 1.2779, "step": 5614 }, { "epoch": 1.01, "learning_rate": 1.4777697451653713e-05, "loss": 0.7632, "step": 5615 }, { "epoch": 1.02, "learning_rate": 1.4774986444966565e-05, "loss": 1.0145, "step": 5616 }, { "epoch": 1.02, "learning_rate": 1.4772275438279415e-05, "loss": 0.6004, "step": 5617 }, { "epoch": 1.02, "learning_rate": 1.4769564431592265e-05, "loss": 0.9721, "step": 5618 }, { "epoch": 1.02, "learning_rate": 1.4766853424905115e-05, "loss": 0.8531, "step": 5619 }, { "epoch": 1.02, "learning_rate": 1.4764142418217965e-05, "loss": 1.1264, "step": 5620 }, { "epoch": 1.02, "learning_rate": 1.4761431411530816e-05, "loss": 1.2396, "step": 5621 }, { "epoch": 1.02, "learning_rate": 1.4758720404843666e-05, "loss": 0.6151, "step": 5622 }, { "epoch": 1.02, "learning_rate": 1.4756009398156516e-05, "loss": 0.3889, "step": 5623 }, { "epoch": 1.02, "learning_rate": 1.4753298391469366e-05, "loss": 0.4879, "step": 5624 }, { "epoch": 1.02, "learning_rate": 1.4750587384782216e-05, "loss": 0.9232, "step": 5625 }, { "epoch": 1.02, "learning_rate": 1.4747876378095066e-05, "loss": 0.7024, "step": 5626 }, { "epoch": 1.02, "learning_rate": 1.4745165371407916e-05, "loss": 1.251, "step": 5627 }, { "epoch": 1.02, "learning_rate": 1.4742454364720767e-05, "loss": 0.744, "step": 5628 }, { "epoch": 1.02, "learning_rate": 1.4739743358033617e-05, "loss": 0.5651, "step": 5629 }, { "epoch": 1.02, "learning_rate": 1.4737032351346468e-05, "loss": 0.8896, "step": 5630 }, { "epoch": 1.02, "learning_rate": 1.4734321344659317e-05, "loss": 0.5959, "step": 5631 }, { "epoch": 1.02, "learning_rate": 1.4731610337972167e-05, "loss": 1.5154, "step": 5632 }, { "epoch": 1.02, "learning_rate": 1.4728899331285017e-05, "loss": 0.4811, "step": 5633 }, { "epoch": 1.02, "learning_rate": 1.4726188324597869e-05, "loss": 1.0965, "step": 5634 }, { "epoch": 1.02, "learning_rate": 1.4723477317910717e-05, "loss": 0.8403, "step": 5635 }, { "epoch": 1.02, "learning_rate": 1.4720766311223568e-05, "loss": 0.7401, "step": 5636 }, { "epoch": 1.02, "learning_rate": 1.4718055304536418e-05, "loss": 0.4894, "step": 5637 }, { "epoch": 1.02, "learning_rate": 1.471534429784927e-05, "loss": 0.5438, "step": 5638 }, { "epoch": 1.02, "learning_rate": 1.4712633291162118e-05, "loss": 0.8055, "step": 5639 }, { "epoch": 1.02, "learning_rate": 1.4709922284474968e-05, "loss": 0.7423, "step": 5640 }, { "epoch": 1.02, "learning_rate": 1.470721127778782e-05, "loss": 0.8079, "step": 5641 }, { "epoch": 1.02, "learning_rate": 1.470450027110067e-05, "loss": 0.5526, "step": 5642 }, { "epoch": 1.02, "learning_rate": 1.4701789264413519e-05, "loss": 0.6293, "step": 5643 }, { "epoch": 1.02, "learning_rate": 1.4699078257726369e-05, "loss": 0.8374, "step": 5644 }, { "epoch": 1.02, "learning_rate": 1.469636725103922e-05, "loss": 0.6737, "step": 5645 }, { "epoch": 1.02, "learning_rate": 1.469365624435207e-05, "loss": 0.695, "step": 5646 }, { "epoch": 1.02, "learning_rate": 1.4690945237664919e-05, "loss": 1.1417, "step": 5647 }, { "epoch": 1.02, "learning_rate": 1.4688234230977771e-05, "loss": 1.6744, "step": 5648 }, { "epoch": 1.02, "learning_rate": 1.4685523224290621e-05, "loss": 0.7383, "step": 5649 }, { "epoch": 1.02, "learning_rate": 1.4682812217603471e-05, "loss": 0.6321, "step": 5650 }, { "epoch": 1.02, "learning_rate": 1.468010121091632e-05, "loss": 0.2684, "step": 5651 }, { "epoch": 1.02, "learning_rate": 1.4677390204229171e-05, "loss": 0.485, "step": 5652 }, { "epoch": 1.02, "learning_rate": 1.4674679197542022e-05, "loss": 0.1926, "step": 5653 }, { "epoch": 1.02, "learning_rate": 1.4671968190854872e-05, "loss": 0.6469, "step": 5654 }, { "epoch": 1.02, "learning_rate": 1.466925718416772e-05, "loss": 0.5135, "step": 5655 }, { "epoch": 1.02, "learning_rate": 1.4666546177480572e-05, "loss": 0.936, "step": 5656 }, { "epoch": 1.02, "learning_rate": 1.4663835170793422e-05, "loss": 0.6915, "step": 5657 }, { "epoch": 1.02, "learning_rate": 1.4661124164106272e-05, "loss": 0.7359, "step": 5658 }, { "epoch": 1.02, "learning_rate": 1.4658413157419122e-05, "loss": 0.8522, "step": 5659 }, { "epoch": 1.02, "learning_rate": 1.4655702150731973e-05, "loss": 0.5363, "step": 5660 }, { "epoch": 1.02, "learning_rate": 1.4652991144044823e-05, "loss": 1.0846, "step": 5661 }, { "epoch": 1.02, "learning_rate": 1.4650280137357671e-05, "loss": 0.8446, "step": 5662 }, { "epoch": 1.02, "learning_rate": 1.4647569130670523e-05, "loss": 0.8633, "step": 5663 }, { "epoch": 1.02, "learning_rate": 1.4644858123983373e-05, "loss": 0.6174, "step": 5664 }, { "epoch": 1.02, "learning_rate": 1.4642147117296223e-05, "loss": 0.9205, "step": 5665 }, { "epoch": 1.02, "learning_rate": 1.4639436110609072e-05, "loss": 0.6606, "step": 5666 }, { "epoch": 1.02, "learning_rate": 1.4636725103921923e-05, "loss": 0.7247, "step": 5667 }, { "epoch": 1.02, "learning_rate": 1.4634014097234774e-05, "loss": 1.6146, "step": 5668 }, { "epoch": 1.02, "learning_rate": 1.4631303090547624e-05, "loss": 0.5345, "step": 5669 }, { "epoch": 1.02, "learning_rate": 1.4628592083860474e-05, "loss": 0.8331, "step": 5670 }, { "epoch": 1.02, "learning_rate": 1.4625881077173324e-05, "loss": 0.923, "step": 5671 }, { "epoch": 1.03, "learning_rate": 1.4623170070486174e-05, "loss": 0.8851, "step": 5672 }, { "epoch": 1.03, "learning_rate": 1.4620459063799024e-05, "loss": 0.9294, "step": 5673 }, { "epoch": 1.03, "learning_rate": 1.4617748057111874e-05, "loss": 0.6119, "step": 5674 }, { "epoch": 1.03, "learning_rate": 1.4615037050424725e-05, "loss": 1.1553, "step": 5675 }, { "epoch": 1.03, "learning_rate": 1.4612326043737575e-05, "loss": 0.6947, "step": 5676 }, { "epoch": 1.03, "learning_rate": 1.4609615037050425e-05, "loss": 1.3493, "step": 5677 }, { "epoch": 1.03, "learning_rate": 1.4606904030363275e-05, "loss": 0.6865, "step": 5678 }, { "epoch": 1.03, "learning_rate": 1.4604193023676125e-05, "loss": 0.7651, "step": 5679 }, { "epoch": 1.03, "learning_rate": 1.4601482016988975e-05, "loss": 0.5318, "step": 5680 }, { "epoch": 1.03, "learning_rate": 1.4598771010301827e-05, "loss": 0.5395, "step": 5681 }, { "epoch": 1.03, "learning_rate": 1.4596060003614675e-05, "loss": 0.7239, "step": 5682 }, { "epoch": 1.03, "learning_rate": 1.4593348996927526e-05, "loss": 1.0899, "step": 5683 }, { "epoch": 1.03, "learning_rate": 1.4590637990240376e-05, "loss": 1.1303, "step": 5684 }, { "epoch": 1.03, "learning_rate": 1.4587926983553228e-05, "loss": 1.0575, "step": 5685 }, { "epoch": 1.03, "learning_rate": 1.4585215976866076e-05, "loss": 0.6633, "step": 5686 }, { "epoch": 1.03, "learning_rate": 1.4582504970178926e-05, "loss": 0.2632, "step": 5687 }, { "epoch": 1.03, "learning_rate": 1.4579793963491776e-05, "loss": 0.641, "step": 5688 }, { "epoch": 1.03, "learning_rate": 1.4577082956804628e-05, "loss": 0.653, "step": 5689 }, { "epoch": 1.03, "learning_rate": 1.4574371950117477e-05, "loss": 0.5775, "step": 5690 }, { "epoch": 1.03, "learning_rate": 1.4571660943430327e-05, "loss": 0.8516, "step": 5691 }, { "epoch": 1.03, "learning_rate": 1.4568949936743179e-05, "loss": 0.8436, "step": 5692 }, { "epoch": 1.03, "learning_rate": 1.4566238930056029e-05, "loss": 0.9707, "step": 5693 }, { "epoch": 1.03, "learning_rate": 1.4563527923368877e-05, "loss": 0.3721, "step": 5694 }, { "epoch": 1.03, "learning_rate": 1.4560816916681727e-05, "loss": 0.4385, "step": 5695 }, { "epoch": 1.03, "learning_rate": 1.4558105909994579e-05, "loss": 0.5403, "step": 5696 }, { "epoch": 1.03, "learning_rate": 1.455539490330743e-05, "loss": 0.3315, "step": 5697 }, { "epoch": 1.03, "learning_rate": 1.4552683896620278e-05, "loss": 1.0782, "step": 5698 }, { "epoch": 1.03, "learning_rate": 1.4549972889933128e-05, "loss": 0.7411, "step": 5699 }, { "epoch": 1.03, "learning_rate": 1.454726188324598e-05, "loss": 0.7978, "step": 5700 }, { "epoch": 1.03, "learning_rate": 1.454455087655883e-05, "loss": 0.6505, "step": 5701 }, { "epoch": 1.03, "learning_rate": 1.4541839869871678e-05, "loss": 0.349, "step": 5702 }, { "epoch": 1.03, "learning_rate": 1.453912886318453e-05, "loss": 0.4873, "step": 5703 }, { "epoch": 1.03, "learning_rate": 1.453641785649738e-05, "loss": 0.9697, "step": 5704 }, { "epoch": 1.03, "learning_rate": 1.453370684981023e-05, "loss": 1.4095, "step": 5705 }, { "epoch": 1.03, "learning_rate": 1.4530995843123079e-05, "loss": 1.2429, "step": 5706 }, { "epoch": 1.03, "learning_rate": 1.452828483643593e-05, "loss": 1.251, "step": 5707 }, { "epoch": 1.03, "learning_rate": 1.452557382974878e-05, "loss": 0.8815, "step": 5708 }, { "epoch": 1.03, "learning_rate": 1.452286282306163e-05, "loss": 0.6302, "step": 5709 }, { "epoch": 1.03, "learning_rate": 1.4520151816374481e-05, "loss": 0.9274, "step": 5710 }, { "epoch": 1.03, "learning_rate": 1.4517440809687331e-05, "loss": 0.8147, "step": 5711 }, { "epoch": 1.03, "learning_rate": 1.4514729803000181e-05, "loss": 0.5193, "step": 5712 }, { "epoch": 1.03, "learning_rate": 1.4512018796313031e-05, "loss": 0.9526, "step": 5713 }, { "epoch": 1.03, "learning_rate": 1.4509307789625881e-05, "loss": 1.651, "step": 5714 }, { "epoch": 1.03, "learning_rate": 1.4506596782938732e-05, "loss": 0.563, "step": 5715 }, { "epoch": 1.03, "learning_rate": 1.4503885776251582e-05, "loss": 1.012, "step": 5716 }, { "epoch": 1.03, "learning_rate": 1.4501174769564432e-05, "loss": 0.6023, "step": 5717 }, { "epoch": 1.03, "learning_rate": 1.4498463762877282e-05, "loss": 1.0398, "step": 5718 }, { "epoch": 1.03, "learning_rate": 1.4495752756190132e-05, "loss": 0.9023, "step": 5719 }, { "epoch": 1.03, "learning_rate": 1.4493041749502982e-05, "loss": 0.3622, "step": 5720 }, { "epoch": 1.03, "learning_rate": 1.4490330742815834e-05, "loss": 0.5426, "step": 5721 }, { "epoch": 1.03, "learning_rate": 1.4487619736128683e-05, "loss": 0.6314, "step": 5722 }, { "epoch": 1.03, "learning_rate": 1.4484908729441533e-05, "loss": 0.3036, "step": 5723 }, { "epoch": 1.03, "learning_rate": 1.4482197722754383e-05, "loss": 0.3499, "step": 5724 }, { "epoch": 1.03, "learning_rate": 1.4479486716067235e-05, "loss": 0.9772, "step": 5725 }, { "epoch": 1.03, "learning_rate": 1.4476775709380083e-05, "loss": 0.4542, "step": 5726 }, { "epoch": 1.04, "learning_rate": 1.4474064702692933e-05, "loss": 0.6556, "step": 5727 }, { "epoch": 1.04, "learning_rate": 1.4471353696005783e-05, "loss": 0.5315, "step": 5728 }, { "epoch": 1.04, "learning_rate": 1.4468642689318635e-05, "loss": 0.5378, "step": 5729 }, { "epoch": 1.04, "learning_rate": 1.4465931682631484e-05, "loss": 0.8806, "step": 5730 }, { "epoch": 1.04, "learning_rate": 1.4463220675944334e-05, "loss": 0.5878, "step": 5731 }, { "epoch": 1.04, "learning_rate": 1.4460509669257186e-05, "loss": 0.824, "step": 5732 }, { "epoch": 1.04, "learning_rate": 1.4457798662570036e-05, "loss": 0.5159, "step": 5733 }, { "epoch": 1.04, "learning_rate": 1.4455087655882884e-05, "loss": 0.6551, "step": 5734 }, { "epoch": 1.04, "learning_rate": 1.4452376649195734e-05, "loss": 0.5943, "step": 5735 }, { "epoch": 1.04, "learning_rate": 1.4449665642508586e-05, "loss": 0.7423, "step": 5736 }, { "epoch": 1.04, "learning_rate": 1.4446954635821436e-05, "loss": 0.4333, "step": 5737 }, { "epoch": 1.04, "learning_rate": 1.4444243629134285e-05, "loss": 0.5298, "step": 5738 }, { "epoch": 1.04, "learning_rate": 1.4441532622447135e-05, "loss": 1.4624, "step": 5739 }, { "epoch": 1.04, "learning_rate": 1.4438821615759987e-05, "loss": 0.6598, "step": 5740 }, { "epoch": 1.04, "learning_rate": 1.4436110609072837e-05, "loss": 0.9028, "step": 5741 }, { "epoch": 1.04, "learning_rate": 1.4433399602385685e-05, "loss": 0.4932, "step": 5742 }, { "epoch": 1.04, "learning_rate": 1.4430688595698537e-05, "loss": 0.4483, "step": 5743 }, { "epoch": 1.04, "learning_rate": 1.4427977589011387e-05, "loss": 0.3577, "step": 5744 }, { "epoch": 1.04, "learning_rate": 1.4425266582324237e-05, "loss": 0.9098, "step": 5745 }, { "epoch": 1.04, "learning_rate": 1.4422555575637086e-05, "loss": 0.6337, "step": 5746 }, { "epoch": 1.04, "learning_rate": 1.4419844568949938e-05, "loss": 1.2091, "step": 5747 }, { "epoch": 1.04, "learning_rate": 1.4417133562262788e-05, "loss": 0.7867, "step": 5748 }, { "epoch": 1.04, "learning_rate": 1.4414422555575638e-05, "loss": 0.9405, "step": 5749 }, { "epoch": 1.04, "learning_rate": 1.4411711548888486e-05, "loss": 0.5504, "step": 5750 }, { "epoch": 1.04, "learning_rate": 1.4409000542201338e-05, "loss": 1.1522, "step": 5751 }, { "epoch": 1.04, "learning_rate": 1.4406289535514188e-05, "loss": 0.7034, "step": 5752 }, { "epoch": 1.04, "learning_rate": 1.4403578528827038e-05, "loss": 1.3517, "step": 5753 }, { "epoch": 1.04, "learning_rate": 1.4400867522139889e-05, "loss": 0.8245, "step": 5754 }, { "epoch": 1.04, "learning_rate": 1.4398156515452739e-05, "loss": 0.9378, "step": 5755 }, { "epoch": 1.04, "learning_rate": 1.4395445508765589e-05, "loss": 0.553, "step": 5756 }, { "epoch": 1.04, "learning_rate": 1.4392734502078439e-05, "loss": 0.3249, "step": 5757 }, { "epoch": 1.04, "learning_rate": 1.4390023495391289e-05, "loss": 0.9385, "step": 5758 }, { "epoch": 1.04, "learning_rate": 1.438731248870414e-05, "loss": 0.8901, "step": 5759 }, { "epoch": 1.04, "learning_rate": 1.438460148201699e-05, "loss": 0.5505, "step": 5760 }, { "epoch": 1.04, "learning_rate": 1.438189047532984e-05, "loss": 0.5537, "step": 5761 }, { "epoch": 1.04, "learning_rate": 1.437917946864269e-05, "loss": 0.5389, "step": 5762 }, { "epoch": 1.04, "learning_rate": 1.437646846195554e-05, "loss": 1.0437, "step": 5763 }, { "epoch": 1.04, "learning_rate": 1.437375745526839e-05, "loss": 0.7571, "step": 5764 }, { "epoch": 1.04, "learning_rate": 1.4371046448581242e-05, "loss": 0.2544, "step": 5765 }, { "epoch": 1.04, "learning_rate": 1.436833544189409e-05, "loss": 1.3678, "step": 5766 }, { "epoch": 1.04, "learning_rate": 1.436562443520694e-05, "loss": 0.6217, "step": 5767 }, { "epoch": 1.04, "learning_rate": 1.436291342851979e-05, "loss": 0.563, "step": 5768 }, { "epoch": 1.04, "learning_rate": 1.4360202421832642e-05, "loss": 0.6501, "step": 5769 }, { "epoch": 1.04, "learning_rate": 1.435749141514549e-05, "loss": 0.6962, "step": 5770 }, { "epoch": 1.04, "learning_rate": 1.435478040845834e-05, "loss": 0.463, "step": 5771 }, { "epoch": 1.04, "learning_rate": 1.4352069401771191e-05, "loss": 0.4408, "step": 5772 }, { "epoch": 1.04, "learning_rate": 1.4349358395084043e-05, "loss": 1.2917, "step": 5773 }, { "epoch": 1.04, "learning_rate": 1.4346647388396891e-05, "loss": 0.6168, "step": 5774 }, { "epoch": 1.04, "learning_rate": 1.4343936381709741e-05, "loss": 0.6093, "step": 5775 }, { "epoch": 1.04, "learning_rate": 1.4341225375022593e-05, "loss": 0.9205, "step": 5776 }, { "epoch": 1.04, "learning_rate": 1.4338514368335443e-05, "loss": 1.1308, "step": 5777 }, { "epoch": 1.04, "learning_rate": 1.4335803361648292e-05, "loss": 0.6457, "step": 5778 }, { "epoch": 1.04, "learning_rate": 1.4333092354961142e-05, "loss": 1.4777, "step": 5779 }, { "epoch": 1.04, "learning_rate": 1.4330381348273994e-05, "loss": 1.0482, "step": 5780 }, { "epoch": 1.04, "learning_rate": 1.4327670341586844e-05, "loss": 0.2784, "step": 5781 }, { "epoch": 1.05, "learning_rate": 1.4324959334899692e-05, "loss": 0.4392, "step": 5782 }, { "epoch": 1.05, "learning_rate": 1.4322248328212542e-05, "loss": 0.7412, "step": 5783 }, { "epoch": 1.05, "learning_rate": 1.4319537321525394e-05, "loss": 1.0361, "step": 5784 }, { "epoch": 1.05, "learning_rate": 1.4316826314838244e-05, "loss": 0.6282, "step": 5785 }, { "epoch": 1.05, "learning_rate": 1.4314115308151093e-05, "loss": 0.6589, "step": 5786 }, { "epoch": 1.05, "learning_rate": 1.4311404301463945e-05, "loss": 0.3509, "step": 5787 }, { "epoch": 1.05, "learning_rate": 1.4308693294776795e-05, "loss": 1.2844, "step": 5788 }, { "epoch": 1.05, "learning_rate": 1.4305982288089645e-05, "loss": 1.0125, "step": 5789 }, { "epoch": 1.05, "learning_rate": 1.4303271281402493e-05, "loss": 0.514, "step": 5790 }, { "epoch": 1.05, "learning_rate": 1.4300560274715345e-05, "loss": 0.3558, "step": 5791 }, { "epoch": 1.05, "learning_rate": 1.4297849268028195e-05, "loss": 0.8019, "step": 5792 }, { "epoch": 1.05, "learning_rate": 1.4295138261341044e-05, "loss": 1.8817, "step": 5793 }, { "epoch": 1.05, "learning_rate": 1.4292427254653896e-05, "loss": 0.81, "step": 5794 }, { "epoch": 1.05, "learning_rate": 1.4289716247966746e-05, "loss": 0.8137, "step": 5795 }, { "epoch": 1.05, "learning_rate": 1.4287005241279596e-05, "loss": 0.5107, "step": 5796 }, { "epoch": 1.05, "learning_rate": 1.4284294234592444e-05, "loss": 0.5051, "step": 5797 }, { "epoch": 1.05, "learning_rate": 1.4281583227905296e-05, "loss": 0.6331, "step": 5798 }, { "epoch": 1.05, "learning_rate": 1.4278872221218146e-05, "loss": 1.6405, "step": 5799 }, { "epoch": 1.05, "learning_rate": 1.4276161214530996e-05, "loss": 1.3661, "step": 5800 }, { "epoch": 1.05, "learning_rate": 1.4273450207843845e-05, "loss": 0.8658, "step": 5801 }, { "epoch": 1.05, "learning_rate": 1.4270739201156697e-05, "loss": 0.5994, "step": 5802 }, { "epoch": 1.05, "learning_rate": 1.4268028194469547e-05, "loss": 0.6257, "step": 5803 }, { "epoch": 1.05, "learning_rate": 1.4265317187782397e-05, "loss": 1.0357, "step": 5804 }, { "epoch": 1.05, "learning_rate": 1.4262606181095247e-05, "loss": 0.4254, "step": 5805 }, { "epoch": 1.05, "learning_rate": 1.4259895174408097e-05, "loss": 1.0541, "step": 5806 }, { "epoch": 1.05, "learning_rate": 1.4257184167720947e-05, "loss": 0.7983, "step": 5807 }, { "epoch": 1.05, "learning_rate": 1.4254473161033797e-05, "loss": 0.6099, "step": 5808 }, { "epoch": 1.05, "learning_rate": 1.4251762154346648e-05, "loss": 0.6707, "step": 5809 }, { "epoch": 1.05, "learning_rate": 1.4249051147659498e-05, "loss": 0.3487, "step": 5810 }, { "epoch": 1.05, "learning_rate": 1.4246340140972348e-05, "loss": 0.4183, "step": 5811 }, { "epoch": 1.05, "learning_rate": 1.4243629134285198e-05, "loss": 0.5595, "step": 5812 }, { "epoch": 1.05, "learning_rate": 1.4240918127598048e-05, "loss": 0.665, "step": 5813 }, { "epoch": 1.05, "learning_rate": 1.4238207120910898e-05, "loss": 0.6353, "step": 5814 }, { "epoch": 1.05, "learning_rate": 1.4235496114223748e-05, "loss": 0.7528, "step": 5815 }, { "epoch": 1.05, "learning_rate": 1.42327851075366e-05, "loss": 0.7185, "step": 5816 }, { "epoch": 1.05, "learning_rate": 1.4230074100849449e-05, "loss": 0.5758, "step": 5817 }, { "epoch": 1.05, "learning_rate": 1.4227363094162299e-05, "loss": 1.0978, "step": 5818 }, { "epoch": 1.05, "learning_rate": 1.4224652087475149e-05, "loss": 0.6903, "step": 5819 }, { "epoch": 1.05, "learning_rate": 1.4221941080788e-05, "loss": 0.6113, "step": 5820 }, { "epoch": 1.05, "learning_rate": 1.421923007410085e-05, "loss": 0.9427, "step": 5821 }, { "epoch": 1.05, "learning_rate": 1.42165190674137e-05, "loss": 1.1351, "step": 5822 }, { "epoch": 1.05, "learning_rate": 1.421380806072655e-05, "loss": 1.2412, "step": 5823 }, { "epoch": 1.05, "learning_rate": 1.4211097054039401e-05, "loss": 1.0727, "step": 5824 }, { "epoch": 1.05, "learning_rate": 1.420838604735225e-05, "loss": 1.3172, "step": 5825 }, { "epoch": 1.05, "learning_rate": 1.42056750406651e-05, "loss": 1.2707, "step": 5826 }, { "epoch": 1.05, "learning_rate": 1.4202964033977952e-05, "loss": 0.6921, "step": 5827 }, { "epoch": 1.05, "learning_rate": 1.4200253027290802e-05, "loss": 0.771, "step": 5828 }, { "epoch": 1.05, "learning_rate": 1.419754202060365e-05, "loss": 0.777, "step": 5829 }, { "epoch": 1.05, "learning_rate": 1.41948310139165e-05, "loss": 0.8933, "step": 5830 }, { "epoch": 1.05, "learning_rate": 1.4192120007229352e-05, "loss": 0.9732, "step": 5831 }, { "epoch": 1.05, "learning_rate": 1.4189409000542202e-05, "loss": 1.4409, "step": 5832 }, { "epoch": 1.05, "learning_rate": 1.418669799385505e-05, "loss": 0.5356, "step": 5833 }, { "epoch": 1.05, "learning_rate": 1.4183986987167901e-05, "loss": 0.7628, "step": 5834 }, { "epoch": 1.05, "learning_rate": 1.4181275980480753e-05, "loss": 0.4097, "step": 5835 }, { "epoch": 1.05, "learning_rate": 1.4178564973793603e-05, "loss": 0.5645, "step": 5836 }, { "epoch": 1.05, "learning_rate": 1.4175853967106451e-05, "loss": 1.6319, "step": 5837 }, { "epoch": 1.06, "learning_rate": 1.4173142960419303e-05, "loss": 0.4153, "step": 5838 }, { "epoch": 1.06, "learning_rate": 1.4170431953732153e-05, "loss": 0.3204, "step": 5839 }, { "epoch": 1.06, "learning_rate": 1.4167720947045003e-05, "loss": 1.0146, "step": 5840 }, { "epoch": 1.06, "learning_rate": 1.4165009940357852e-05, "loss": 0.4845, "step": 5841 }, { "epoch": 1.06, "learning_rate": 1.4162298933670704e-05, "loss": 0.8445, "step": 5842 }, { "epoch": 1.06, "learning_rate": 1.4159587926983554e-05, "loss": 0.9649, "step": 5843 }, { "epoch": 1.06, "learning_rate": 1.4156876920296404e-05, "loss": 0.6378, "step": 5844 }, { "epoch": 1.06, "learning_rate": 1.4154165913609252e-05, "loss": 0.7075, "step": 5845 }, { "epoch": 1.06, "learning_rate": 1.4151454906922104e-05, "loss": 1.0163, "step": 5846 }, { "epoch": 1.06, "learning_rate": 1.4148743900234954e-05, "loss": 0.9241, "step": 5847 }, { "epoch": 1.06, "learning_rate": 1.4146032893547805e-05, "loss": 0.5373, "step": 5848 }, { "epoch": 1.06, "learning_rate": 1.4143321886860655e-05, "loss": 0.3693, "step": 5849 }, { "epoch": 1.06, "learning_rate": 1.4140610880173505e-05, "loss": 0.6257, "step": 5850 }, { "epoch": 1.06, "learning_rate": 1.4137899873486355e-05, "loss": 1.7402, "step": 5851 }, { "epoch": 1.06, "learning_rate": 1.4135188866799205e-05, "loss": 0.6471, "step": 5852 }, { "epoch": 1.06, "learning_rate": 1.4132477860112055e-05, "loss": 0.6731, "step": 5853 }, { "epoch": 1.06, "learning_rate": 1.4129766853424905e-05, "loss": 0.2536, "step": 5854 }, { "epoch": 1.06, "learning_rate": 1.4127055846737755e-05, "loss": 0.5165, "step": 5855 }, { "epoch": 1.06, "learning_rate": 1.4124344840050606e-05, "loss": 0.944, "step": 5856 }, { "epoch": 1.06, "learning_rate": 1.4121633833363456e-05, "loss": 0.6208, "step": 5857 }, { "epoch": 1.06, "learning_rate": 1.4118922826676306e-05, "loss": 0.9141, "step": 5858 }, { "epoch": 1.06, "learning_rate": 1.4116211819989156e-05, "loss": 0.5705, "step": 5859 }, { "epoch": 1.06, "learning_rate": 1.4113500813302008e-05, "loss": 0.769, "step": 5860 }, { "epoch": 1.06, "learning_rate": 1.4110789806614856e-05, "loss": 0.9631, "step": 5861 }, { "epoch": 1.06, "learning_rate": 1.4108078799927706e-05, "loss": 0.8447, "step": 5862 }, { "epoch": 1.06, "learning_rate": 1.4105367793240557e-05, "loss": 0.7453, "step": 5863 }, { "epoch": 1.06, "learning_rate": 1.4102656786553408e-05, "loss": 0.8252, "step": 5864 }, { "epoch": 1.06, "learning_rate": 1.4099945779866257e-05, "loss": 0.5605, "step": 5865 }, { "epoch": 1.06, "learning_rate": 1.4097234773179107e-05, "loss": 1.1823, "step": 5866 }, { "epoch": 1.06, "learning_rate": 1.4094523766491959e-05, "loss": 0.9855, "step": 5867 }, { "epoch": 1.06, "learning_rate": 1.4091812759804809e-05, "loss": 0.7961, "step": 5868 }, { "epoch": 1.06, "learning_rate": 1.4089101753117657e-05, "loss": 1.0346, "step": 5869 }, { "epoch": 1.06, "learning_rate": 1.4086390746430507e-05, "loss": 0.3844, "step": 5870 }, { "epoch": 1.06, "learning_rate": 1.408367973974336e-05, "loss": 0.7874, "step": 5871 }, { "epoch": 1.06, "learning_rate": 1.408096873305621e-05, "loss": 0.6795, "step": 5872 }, { "epoch": 1.06, "learning_rate": 1.4078257726369058e-05, "loss": 1.0123, "step": 5873 }, { "epoch": 1.06, "learning_rate": 1.4075546719681908e-05, "loss": 0.8188, "step": 5874 }, { "epoch": 1.06, "learning_rate": 1.407283571299476e-05, "loss": 1.1488, "step": 5875 }, { "epoch": 1.06, "learning_rate": 1.407012470630761e-05, "loss": 0.8594, "step": 5876 }, { "epoch": 1.06, "learning_rate": 1.4067413699620458e-05, "loss": 1.2148, "step": 5877 }, { "epoch": 1.06, "learning_rate": 1.406470269293331e-05, "loss": 1.2651, "step": 5878 }, { "epoch": 1.06, "learning_rate": 1.406199168624616e-05, "loss": 0.3839, "step": 5879 }, { "epoch": 1.06, "learning_rate": 1.405928067955901e-05, "loss": 0.9483, "step": 5880 }, { "epoch": 1.06, "learning_rate": 1.4056569672871859e-05, "loss": 0.873, "step": 5881 }, { "epoch": 1.06, "learning_rate": 1.405385866618471e-05, "loss": 1.1389, "step": 5882 }, { "epoch": 1.06, "learning_rate": 1.4051147659497561e-05, "loss": 0.9083, "step": 5883 }, { "epoch": 1.06, "learning_rate": 1.4048436652810411e-05, "loss": 0.6906, "step": 5884 }, { "epoch": 1.06, "learning_rate": 1.404572564612326e-05, "loss": 0.47, "step": 5885 }, { "epoch": 1.06, "learning_rate": 1.4043014639436111e-05, "loss": 1.4145, "step": 5886 }, { "epoch": 1.06, "learning_rate": 1.4040303632748961e-05, "loss": 0.4298, "step": 5887 }, { "epoch": 1.06, "learning_rate": 1.4037592626061812e-05, "loss": 0.4465, "step": 5888 }, { "epoch": 1.06, "learning_rate": 1.4034881619374662e-05, "loss": 0.7088, "step": 5889 }, { "epoch": 1.06, "learning_rate": 1.4032170612687512e-05, "loss": 0.7924, "step": 5890 }, { "epoch": 1.06, "learning_rate": 1.4029459606000362e-05, "loss": 0.9601, "step": 5891 }, { "epoch": 1.06, "learning_rate": 1.4026748599313212e-05, "loss": 0.4331, "step": 5892 }, { "epoch": 1.07, "learning_rate": 1.4024037592626062e-05, "loss": 1.0758, "step": 5893 }, { "epoch": 1.07, "learning_rate": 1.4021326585938912e-05, "loss": 0.8251, "step": 5894 }, { "epoch": 1.07, "learning_rate": 1.4018615579251763e-05, "loss": 0.4596, "step": 5895 }, { "epoch": 1.07, "learning_rate": 1.4015904572564613e-05, "loss": 0.9647, "step": 5896 }, { "epoch": 1.07, "learning_rate": 1.4013193565877463e-05, "loss": 0.7133, "step": 5897 }, { "epoch": 1.07, "learning_rate": 1.4010482559190313e-05, "loss": 0.5316, "step": 5898 }, { "epoch": 1.07, "learning_rate": 1.4007771552503163e-05, "loss": 0.2294, "step": 5899 }, { "epoch": 1.07, "learning_rate": 1.4005060545816015e-05, "loss": 0.3849, "step": 5900 }, { "epoch": 1.07, "learning_rate": 1.4002349539128863e-05, "loss": 0.6596, "step": 5901 }, { "epoch": 1.07, "learning_rate": 1.3999638532441713e-05, "loss": 1.3988, "step": 5902 }, { "epoch": 1.07, "learning_rate": 1.3996927525754564e-05, "loss": 1.125, "step": 5903 }, { "epoch": 1.07, "learning_rate": 1.3994216519067415e-05, "loss": 0.6586, "step": 5904 }, { "epoch": 1.07, "learning_rate": 1.3991505512380264e-05, "loss": 0.8022, "step": 5905 }, { "epoch": 1.07, "learning_rate": 1.3988794505693114e-05, "loss": 0.7646, "step": 5906 }, { "epoch": 1.07, "learning_rate": 1.3986083499005964e-05, "loss": 1.4023, "step": 5907 }, { "epoch": 1.07, "learning_rate": 1.3983372492318816e-05, "loss": 0.7168, "step": 5908 }, { "epoch": 1.07, "learning_rate": 1.3980661485631664e-05, "loss": 1.1186, "step": 5909 }, { "epoch": 1.07, "learning_rate": 1.3977950478944515e-05, "loss": 0.976, "step": 5910 }, { "epoch": 1.07, "learning_rate": 1.3975239472257366e-05, "loss": 0.5543, "step": 5911 }, { "epoch": 1.07, "learning_rate": 1.3972528465570217e-05, "loss": 0.8459, "step": 5912 }, { "epoch": 1.07, "learning_rate": 1.3969817458883065e-05, "loss": 0.9324, "step": 5913 }, { "epoch": 1.07, "learning_rate": 1.3967106452195915e-05, "loss": 0.7864, "step": 5914 }, { "epoch": 1.07, "learning_rate": 1.3964395445508767e-05, "loss": 0.8522, "step": 5915 }, { "epoch": 1.07, "learning_rate": 1.3961684438821617e-05, "loss": 0.3541, "step": 5916 }, { "epoch": 1.07, "learning_rate": 1.3958973432134465e-05, "loss": 0.5696, "step": 5917 }, { "epoch": 1.07, "learning_rate": 1.3956262425447316e-05, "loss": 0.7721, "step": 5918 }, { "epoch": 1.07, "learning_rate": 1.3953551418760167e-05, "loss": 0.566, "step": 5919 }, { "epoch": 1.07, "learning_rate": 1.3950840412073016e-05, "loss": 0.4833, "step": 5920 }, { "epoch": 1.07, "learning_rate": 1.3948129405385866e-05, "loss": 1.631, "step": 5921 }, { "epoch": 1.07, "learning_rate": 1.3945418398698718e-05, "loss": 0.9078, "step": 5922 }, { "epoch": 1.07, "learning_rate": 1.3942707392011568e-05, "loss": 0.8633, "step": 5923 }, { "epoch": 1.07, "learning_rate": 1.3939996385324416e-05, "loss": 0.8362, "step": 5924 }, { "epoch": 1.07, "learning_rate": 1.3937285378637267e-05, "loss": 0.4952, "step": 5925 }, { "epoch": 1.07, "learning_rate": 1.3934574371950118e-05, "loss": 1.0697, "step": 5926 }, { "epoch": 1.07, "learning_rate": 1.3931863365262969e-05, "loss": 0.4137, "step": 5927 }, { "epoch": 1.07, "learning_rate": 1.3929152358575817e-05, "loss": 0.9626, "step": 5928 }, { "epoch": 1.07, "learning_rate": 1.3926441351888669e-05, "loss": 1.3431, "step": 5929 }, { "epoch": 1.07, "learning_rate": 1.3923730345201519e-05, "loss": 0.7368, "step": 5930 }, { "epoch": 1.07, "learning_rate": 1.3921019338514369e-05, "loss": 0.7624, "step": 5931 }, { "epoch": 1.07, "learning_rate": 1.3918308331827218e-05, "loss": 0.9743, "step": 5932 }, { "epoch": 1.07, "learning_rate": 1.391559732514007e-05, "loss": 0.5314, "step": 5933 }, { "epoch": 1.07, "learning_rate": 1.391288631845292e-05, "loss": 0.6364, "step": 5934 }, { "epoch": 1.07, "learning_rate": 1.391017531176577e-05, "loss": 2.0612, "step": 5935 }, { "epoch": 1.07, "learning_rate": 1.3907464305078618e-05, "loss": 0.4816, "step": 5936 }, { "epoch": 1.07, "learning_rate": 1.390475329839147e-05, "loss": 0.5359, "step": 5937 }, { "epoch": 1.07, "learning_rate": 1.390204229170432e-05, "loss": 0.3881, "step": 5938 }, { "epoch": 1.07, "learning_rate": 1.389933128501717e-05, "loss": 0.6049, "step": 5939 }, { "epoch": 1.07, "learning_rate": 1.389662027833002e-05, "loss": 0.5488, "step": 5940 }, { "epoch": 1.07, "learning_rate": 1.389390927164287e-05, "loss": 0.6886, "step": 5941 }, { "epoch": 1.07, "learning_rate": 1.389119826495572e-05, "loss": 1.0777, "step": 5942 }, { "epoch": 1.07, "learning_rate": 1.388848725826857e-05, "loss": 1.042, "step": 5943 }, { "epoch": 1.07, "learning_rate": 1.388577625158142e-05, "loss": 0.635, "step": 5944 }, { "epoch": 1.07, "learning_rate": 1.3883065244894271e-05, "loss": 0.5828, "step": 5945 }, { "epoch": 1.07, "learning_rate": 1.3880354238207121e-05, "loss": 0.6372, "step": 5946 }, { "epoch": 1.07, "learning_rate": 1.3877643231519971e-05, "loss": 0.8708, "step": 5947 }, { "epoch": 1.08, "learning_rate": 1.3874932224832821e-05, "loss": 0.7267, "step": 5948 }, { "epoch": 1.08, "learning_rate": 1.3872221218145671e-05, "loss": 0.961, "step": 5949 }, { "epoch": 1.08, "learning_rate": 1.3869510211458522e-05, "loss": 0.5725, "step": 5950 }, { "epoch": 1.08, "learning_rate": 1.3866799204771373e-05, "loss": 0.8499, "step": 5951 }, { "epoch": 1.08, "learning_rate": 1.3864088198084222e-05, "loss": 0.7066, "step": 5952 }, { "epoch": 1.08, "learning_rate": 1.3861377191397072e-05, "loss": 0.5437, "step": 5953 }, { "epoch": 1.08, "learning_rate": 1.3858666184709922e-05, "loss": 0.5363, "step": 5954 }, { "epoch": 1.08, "learning_rate": 1.3855955178022774e-05, "loss": 1.2216, "step": 5955 }, { "epoch": 1.08, "learning_rate": 1.3853244171335622e-05, "loss": 1.1036, "step": 5956 }, { "epoch": 1.08, "learning_rate": 1.3850533164648473e-05, "loss": 0.5927, "step": 5957 }, { "epoch": 1.08, "learning_rate": 1.3847822157961323e-05, "loss": 1.3341, "step": 5958 }, { "epoch": 1.08, "learning_rate": 1.3845111151274175e-05, "loss": 0.36, "step": 5959 }, { "epoch": 1.08, "learning_rate": 1.3842400144587023e-05, "loss": 0.4835, "step": 5960 }, { "epoch": 1.08, "learning_rate": 1.3839689137899873e-05, "loss": 0.299, "step": 5961 }, { "epoch": 1.08, "learning_rate": 1.3836978131212725e-05, "loss": 0.7922, "step": 5962 }, { "epoch": 1.08, "learning_rate": 1.3834267124525575e-05, "loss": 0.6063, "step": 5963 }, { "epoch": 1.08, "learning_rate": 1.3831556117838423e-05, "loss": 0.9014, "step": 5964 }, { "epoch": 1.08, "learning_rate": 1.3828845111151274e-05, "loss": 0.5459, "step": 5965 }, { "epoch": 1.08, "learning_rate": 1.3826134104464125e-05, "loss": 0.1886, "step": 5966 }, { "epoch": 1.08, "learning_rate": 1.3823423097776976e-05, "loss": 0.8285, "step": 5967 }, { "epoch": 1.08, "learning_rate": 1.3820712091089824e-05, "loss": 0.5889, "step": 5968 }, { "epoch": 1.08, "learning_rate": 1.3818001084402674e-05, "loss": 0.6608, "step": 5969 }, { "epoch": 1.08, "learning_rate": 1.3815290077715526e-05, "loss": 1.0414, "step": 5970 }, { "epoch": 1.08, "learning_rate": 1.3812579071028376e-05, "loss": 0.9816, "step": 5971 }, { "epoch": 1.08, "learning_rate": 1.3809868064341225e-05, "loss": 0.7627, "step": 5972 }, { "epoch": 1.08, "learning_rate": 1.3807157057654076e-05, "loss": 0.3823, "step": 5973 }, { "epoch": 1.08, "learning_rate": 1.3804446050966927e-05, "loss": 0.213, "step": 5974 }, { "epoch": 1.08, "learning_rate": 1.3801735044279777e-05, "loss": 0.9144, "step": 5975 }, { "epoch": 1.08, "learning_rate": 1.3799024037592625e-05, "loss": 1.1956, "step": 5976 }, { "epoch": 1.08, "learning_rate": 1.3796313030905477e-05, "loss": 0.4528, "step": 5977 }, { "epoch": 1.08, "learning_rate": 1.3793602024218327e-05, "loss": 1.2194, "step": 5978 }, { "epoch": 1.08, "learning_rate": 1.3790891017531177e-05, "loss": 0.7514, "step": 5979 }, { "epoch": 1.08, "learning_rate": 1.3788180010844026e-05, "loss": 0.3891, "step": 5980 }, { "epoch": 1.08, "learning_rate": 1.3785469004156877e-05, "loss": 0.7241, "step": 5981 }, { "epoch": 1.08, "learning_rate": 1.3782757997469728e-05, "loss": 0.3104, "step": 5982 }, { "epoch": 1.08, "learning_rate": 1.3780046990782578e-05, "loss": 0.7944, "step": 5983 }, { "epoch": 1.08, "learning_rate": 1.3777335984095428e-05, "loss": 0.7143, "step": 5984 }, { "epoch": 1.08, "learning_rate": 1.3774624977408278e-05, "loss": 1.0681, "step": 5985 }, { "epoch": 1.08, "learning_rate": 1.3771913970721128e-05, "loss": 0.7574, "step": 5986 }, { "epoch": 1.08, "learning_rate": 1.3769202964033978e-05, "loss": 0.7397, "step": 5987 }, { "epoch": 1.08, "learning_rate": 1.3766491957346828e-05, "loss": 1.4618, "step": 5988 }, { "epoch": 1.08, "learning_rate": 1.3763780950659679e-05, "loss": 0.4512, "step": 5989 }, { "epoch": 1.08, "learning_rate": 1.3761069943972529e-05, "loss": 0.8855, "step": 5990 }, { "epoch": 1.08, "learning_rate": 1.3758358937285379e-05, "loss": 0.8474, "step": 5991 }, { "epoch": 1.08, "learning_rate": 1.3755647930598229e-05, "loss": 0.4525, "step": 5992 }, { "epoch": 1.08, "learning_rate": 1.3752936923911079e-05, "loss": 0.4292, "step": 5993 }, { "epoch": 1.08, "learning_rate": 1.375022591722393e-05, "loss": 1.7782, "step": 5994 }, { "epoch": 1.08, "learning_rate": 1.3747514910536781e-05, "loss": 0.3203, "step": 5995 }, { "epoch": 1.08, "learning_rate": 1.374480390384963e-05, "loss": 0.6068, "step": 5996 }, { "epoch": 1.08, "learning_rate": 1.374209289716248e-05, "loss": 0.9309, "step": 5997 }, { "epoch": 1.08, "learning_rate": 1.373938189047533e-05, "loss": 0.9432, "step": 5998 }, { "epoch": 1.08, "learning_rate": 1.3736670883788182e-05, "loss": 0.8318, "step": 5999 }, { "epoch": 1.08, "learning_rate": 1.373395987710103e-05, "loss": 1.006, "step": 6000 }, { "epoch": 1.08, "learning_rate": 1.373124887041388e-05, "loss": 0.8278, "step": 6001 }, { "epoch": 1.08, "learning_rate": 1.372853786372673e-05, "loss": 0.8997, "step": 6002 }, { "epoch": 1.08, "learning_rate": 1.3725826857039582e-05, "loss": 0.872, "step": 6003 }, { "epoch": 1.09, "learning_rate": 1.372311585035243e-05, "loss": 0.6971, "step": 6004 }, { "epoch": 1.09, "learning_rate": 1.372040484366528e-05, "loss": 0.2731, "step": 6005 }, { "epoch": 1.09, "learning_rate": 1.3717693836978133e-05, "loss": 0.5502, "step": 6006 }, { "epoch": 1.09, "learning_rate": 1.3714982830290983e-05, "loss": 0.5804, "step": 6007 }, { "epoch": 1.09, "learning_rate": 1.3712271823603831e-05, "loss": 0.8271, "step": 6008 }, { "epoch": 1.09, "learning_rate": 1.3709560816916681e-05, "loss": 0.507, "step": 6009 }, { "epoch": 1.09, "learning_rate": 1.3706849810229533e-05, "loss": 0.5841, "step": 6010 }, { "epoch": 1.09, "learning_rate": 1.3704138803542383e-05, "loss": 0.6081, "step": 6011 }, { "epoch": 1.09, "learning_rate": 1.3701427796855232e-05, "loss": 0.6044, "step": 6012 }, { "epoch": 1.09, "learning_rate": 1.3698716790168083e-05, "loss": 1.3378, "step": 6013 }, { "epoch": 1.09, "learning_rate": 1.3696005783480934e-05, "loss": 0.7957, "step": 6014 }, { "epoch": 1.09, "learning_rate": 1.3693294776793784e-05, "loss": 0.5404, "step": 6015 }, { "epoch": 1.09, "learning_rate": 1.3690583770106632e-05, "loss": 0.3431, "step": 6016 }, { "epoch": 1.09, "learning_rate": 1.3687872763419484e-05, "loss": 0.3355, "step": 6017 }, { "epoch": 1.09, "learning_rate": 1.3685161756732334e-05, "loss": 0.5256, "step": 6018 }, { "epoch": 1.09, "learning_rate": 1.3682450750045184e-05, "loss": 0.9899, "step": 6019 }, { "epoch": 1.09, "learning_rate": 1.3679739743358033e-05, "loss": 0.7846, "step": 6020 }, { "epoch": 1.09, "learning_rate": 1.3677028736670885e-05, "loss": 0.389, "step": 6021 }, { "epoch": 1.09, "learning_rate": 1.3674317729983735e-05, "loss": 0.365, "step": 6022 }, { "epoch": 1.09, "learning_rate": 1.3671606723296585e-05, "loss": 0.6241, "step": 6023 }, { "epoch": 1.09, "learning_rate": 1.3668895716609435e-05, "loss": 1.0476, "step": 6024 }, { "epoch": 1.09, "learning_rate": 1.3666184709922285e-05, "loss": 0.5208, "step": 6025 }, { "epoch": 1.09, "learning_rate": 1.3663473703235135e-05, "loss": 0.7417, "step": 6026 }, { "epoch": 1.09, "learning_rate": 1.3660762696547985e-05, "loss": 1.0138, "step": 6027 }, { "epoch": 1.09, "learning_rate": 1.3658051689860835e-05, "loss": 0.7379, "step": 6028 }, { "epoch": 1.09, "learning_rate": 1.3655340683173686e-05, "loss": 0.3691, "step": 6029 }, { "epoch": 1.09, "learning_rate": 1.3652629676486536e-05, "loss": 1.104, "step": 6030 }, { "epoch": 1.09, "learning_rate": 1.3649918669799386e-05, "loss": 0.5152, "step": 6031 }, { "epoch": 1.09, "learning_rate": 1.3647207663112236e-05, "loss": 0.5506, "step": 6032 }, { "epoch": 1.09, "learning_rate": 1.3644496656425086e-05, "loss": 0.275, "step": 6033 }, { "epoch": 1.09, "learning_rate": 1.3641785649737936e-05, "loss": 0.6083, "step": 6034 }, { "epoch": 1.09, "learning_rate": 1.3639074643050788e-05, "loss": 0.9019, "step": 6035 }, { "epoch": 1.09, "learning_rate": 1.3636363636363637e-05, "loss": 0.5112, "step": 6036 }, { "epoch": 1.09, "learning_rate": 1.3633652629676487e-05, "loss": 1.2197, "step": 6037 }, { "epoch": 1.09, "learning_rate": 1.3630941622989337e-05, "loss": 0.3263, "step": 6038 }, { "epoch": 1.09, "learning_rate": 1.3628230616302189e-05, "loss": 0.6089, "step": 6039 }, { "epoch": 1.09, "learning_rate": 1.3625519609615037e-05, "loss": 0.6528, "step": 6040 }, { "epoch": 1.09, "learning_rate": 1.3622808602927887e-05, "loss": 0.6726, "step": 6041 }, { "epoch": 1.09, "learning_rate": 1.3620097596240737e-05, "loss": 0.2664, "step": 6042 }, { "epoch": 1.09, "learning_rate": 1.361738658955359e-05, "loss": 0.8235, "step": 6043 }, { "epoch": 1.09, "learning_rate": 1.3614675582866438e-05, "loss": 1.0574, "step": 6044 }, { "epoch": 1.09, "learning_rate": 1.3611964576179288e-05, "loss": 0.6102, "step": 6045 }, { "epoch": 1.09, "learning_rate": 1.360925356949214e-05, "loss": 0.5476, "step": 6046 }, { "epoch": 1.09, "learning_rate": 1.360654256280499e-05, "loss": 0.7671, "step": 6047 }, { "epoch": 1.09, "learning_rate": 1.3603831556117838e-05, "loss": 1.4105, "step": 6048 }, { "epoch": 1.09, "learning_rate": 1.3601120549430688e-05, "loss": 1.2462, "step": 6049 }, { "epoch": 1.09, "learning_rate": 1.359840954274354e-05, "loss": 0.5274, "step": 6050 }, { "epoch": 1.09, "learning_rate": 1.3595698536056389e-05, "loss": 1.1666, "step": 6051 }, { "epoch": 1.09, "learning_rate": 1.3592987529369239e-05, "loss": 0.8861, "step": 6052 }, { "epoch": 1.09, "learning_rate": 1.3590276522682089e-05, "loss": 0.875, "step": 6053 }, { "epoch": 1.09, "learning_rate": 1.358756551599494e-05, "loss": 0.7631, "step": 6054 }, { "epoch": 1.09, "learning_rate": 1.3584854509307789e-05, "loss": 1.0639, "step": 6055 }, { "epoch": 1.09, "learning_rate": 1.358214350262064e-05, "loss": 0.5986, "step": 6056 }, { "epoch": 1.09, "learning_rate": 1.3579432495933491e-05, "loss": 0.3236, "step": 6057 }, { "epoch": 1.09, "learning_rate": 1.3576721489246341e-05, "loss": 0.7314, "step": 6058 }, { "epoch": 1.1, "learning_rate": 1.357401048255919e-05, "loss": 0.3029, "step": 6059 }, { "epoch": 1.1, "learning_rate": 1.357129947587204e-05, "loss": 0.9787, "step": 6060 }, { "epoch": 1.1, "learning_rate": 1.3568588469184892e-05, "loss": 1.2568, "step": 6061 }, { "epoch": 1.1, "learning_rate": 1.3565877462497742e-05, "loss": 0.5095, "step": 6062 }, { "epoch": 1.1, "learning_rate": 1.356316645581059e-05, "loss": 0.7301, "step": 6063 }, { "epoch": 1.1, "learning_rate": 1.356045544912344e-05, "loss": 1.3568, "step": 6064 }, { "epoch": 1.1, "learning_rate": 1.3557744442436292e-05, "loss": 0.7776, "step": 6065 }, { "epoch": 1.1, "learning_rate": 1.3555033435749142e-05, "loss": 0.4777, "step": 6066 }, { "epoch": 1.1, "learning_rate": 1.355232242906199e-05, "loss": 1.2122, "step": 6067 }, { "epoch": 1.1, "learning_rate": 1.3549611422374843e-05, "loss": 0.6927, "step": 6068 }, { "epoch": 1.1, "learning_rate": 1.3546900415687693e-05, "loss": 0.7312, "step": 6069 }, { "epoch": 1.1, "learning_rate": 1.3544189409000543e-05, "loss": 0.8637, "step": 6070 }, { "epoch": 1.1, "learning_rate": 1.3541478402313391e-05, "loss": 0.5769, "step": 6071 }, { "epoch": 1.1, "learning_rate": 1.3538767395626243e-05, "loss": 0.6722, "step": 6072 }, { "epoch": 1.1, "learning_rate": 1.3536056388939093e-05, "loss": 1.2725, "step": 6073 }, { "epoch": 1.1, "learning_rate": 1.3533345382251943e-05, "loss": 1.1408, "step": 6074 }, { "epoch": 1.1, "learning_rate": 1.3530634375564793e-05, "loss": 0.7475, "step": 6075 }, { "epoch": 1.1, "learning_rate": 1.3527923368877644e-05, "loss": 0.7868, "step": 6076 }, { "epoch": 1.1, "learning_rate": 1.3525212362190494e-05, "loss": 0.5833, "step": 6077 }, { "epoch": 1.1, "learning_rate": 1.3522501355503344e-05, "loss": 0.8061, "step": 6078 }, { "epoch": 1.1, "learning_rate": 1.3519790348816194e-05, "loss": 1.0358, "step": 6079 }, { "epoch": 1.1, "learning_rate": 1.3517079342129044e-05, "loss": 0.3987, "step": 6080 }, { "epoch": 1.1, "learning_rate": 1.3514368335441894e-05, "loss": 1.1641, "step": 6081 }, { "epoch": 1.1, "learning_rate": 1.3511657328754744e-05, "loss": 0.7834, "step": 6082 }, { "epoch": 1.1, "learning_rate": 1.3508946322067595e-05, "loss": 0.7323, "step": 6083 }, { "epoch": 1.1, "learning_rate": 1.3506235315380445e-05, "loss": 0.449, "step": 6084 }, { "epoch": 1.1, "learning_rate": 1.3503524308693295e-05, "loss": 0.5398, "step": 6085 }, { "epoch": 1.1, "learning_rate": 1.3500813302006147e-05, "loss": 0.7275, "step": 6086 }, { "epoch": 1.1, "learning_rate": 1.3498102295318995e-05, "loss": 0.6031, "step": 6087 }, { "epoch": 1.1, "learning_rate": 1.3495391288631845e-05, "loss": 0.5165, "step": 6088 }, { "epoch": 1.1, "learning_rate": 1.3492680281944695e-05, "loss": 0.5194, "step": 6089 }, { "epoch": 1.1, "learning_rate": 1.3489969275257547e-05, "loss": 1.018, "step": 6090 }, { "epoch": 1.1, "learning_rate": 1.3487258268570396e-05, "loss": 0.5559, "step": 6091 }, { "epoch": 1.1, "learning_rate": 1.3484547261883246e-05, "loss": 0.6703, "step": 6092 }, { "epoch": 1.1, "learning_rate": 1.3481836255196096e-05, "loss": 0.564, "step": 6093 }, { "epoch": 1.1, "learning_rate": 1.3479125248508948e-05, "loss": 0.6148, "step": 6094 }, { "epoch": 1.1, "learning_rate": 1.3476414241821796e-05, "loss": 0.3827, "step": 6095 }, { "epoch": 1.1, "learning_rate": 1.3473703235134646e-05, "loss": 0.6711, "step": 6096 }, { "epoch": 1.1, "learning_rate": 1.3470992228447498e-05, "loss": 1.246, "step": 6097 }, { "epoch": 1.1, "learning_rate": 1.3468281221760348e-05, "loss": 1.0122, "step": 6098 }, { "epoch": 1.1, "learning_rate": 1.3465570215073197e-05, "loss": 0.8021, "step": 6099 }, { "epoch": 1.1, "learning_rate": 1.3462859208386047e-05, "loss": 1.2986, "step": 6100 }, { "epoch": 1.1, "learning_rate": 1.3460148201698899e-05, "loss": 1.0375, "step": 6101 }, { "epoch": 1.1, "learning_rate": 1.3457437195011749e-05, "loss": 0.7123, "step": 6102 }, { "epoch": 1.1, "learning_rate": 1.3454726188324597e-05, "loss": 0.7062, "step": 6103 }, { "epoch": 1.1, "learning_rate": 1.3452015181637447e-05, "loss": 0.6073, "step": 6104 }, { "epoch": 1.1, "learning_rate": 1.34493041749503e-05, "loss": 0.5038, "step": 6105 }, { "epoch": 1.1, "learning_rate": 1.344659316826315e-05, "loss": 0.4975, "step": 6106 }, { "epoch": 1.1, "learning_rate": 1.3443882161575998e-05, "loss": 0.6583, "step": 6107 }, { "epoch": 1.1, "learning_rate": 1.344117115488885e-05, "loss": 0.5616, "step": 6108 }, { "epoch": 1.1, "learning_rate": 1.34384601482017e-05, "loss": 0.4696, "step": 6109 }, { "epoch": 1.1, "learning_rate": 1.343574914151455e-05, "loss": 0.5032, "step": 6110 }, { "epoch": 1.1, "learning_rate": 1.3433038134827398e-05, "loss": 0.5866, "step": 6111 }, { "epoch": 1.1, "learning_rate": 1.343032712814025e-05, "loss": 0.8136, "step": 6112 }, { "epoch": 1.1, "learning_rate": 1.34276161214531e-05, "loss": 0.9989, "step": 6113 }, { "epoch": 1.11, "learning_rate": 1.342490511476595e-05, "loss": 0.5957, "step": 6114 }, { "epoch": 1.11, "learning_rate": 1.3422194108078799e-05, "loss": 0.5001, "step": 6115 }, { "epoch": 1.11, "learning_rate": 1.341948310139165e-05, "loss": 0.7725, "step": 6116 }, { "epoch": 1.11, "learning_rate": 1.34167720947045e-05, "loss": 0.4996, "step": 6117 }, { "epoch": 1.11, "learning_rate": 1.3414061088017351e-05, "loss": 0.367, "step": 6118 }, { "epoch": 1.11, "learning_rate": 1.3411350081330201e-05, "loss": 0.5885, "step": 6119 }, { "epoch": 1.11, "learning_rate": 1.3408639074643051e-05, "loss": 1.3497, "step": 6120 }, { "epoch": 1.11, "learning_rate": 1.3405928067955901e-05, "loss": 0.6595, "step": 6121 }, { "epoch": 1.11, "learning_rate": 1.3403217061268751e-05, "loss": 0.7056, "step": 6122 }, { "epoch": 1.11, "learning_rate": 1.3400506054581602e-05, "loss": 0.518, "step": 6123 }, { "epoch": 1.11, "learning_rate": 1.3397795047894452e-05, "loss": 0.9093, "step": 6124 }, { "epoch": 1.11, "learning_rate": 1.3395084041207302e-05, "loss": 1.0142, "step": 6125 }, { "epoch": 1.11, "learning_rate": 1.3392373034520152e-05, "loss": 1.2031, "step": 6126 }, { "epoch": 1.11, "learning_rate": 1.3389662027833002e-05, "loss": 0.2102, "step": 6127 }, { "epoch": 1.11, "learning_rate": 1.3386951021145852e-05, "loss": 0.4649, "step": 6128 }, { "epoch": 1.11, "learning_rate": 1.3384240014458702e-05, "loss": 0.6684, "step": 6129 }, { "epoch": 1.11, "learning_rate": 1.3381529007771554e-05, "loss": 1.0927, "step": 6130 }, { "epoch": 1.11, "learning_rate": 1.3378818001084403e-05, "loss": 0.3045, "step": 6131 }, { "epoch": 1.11, "learning_rate": 1.3376106994397253e-05, "loss": 0.6557, "step": 6132 }, { "epoch": 1.11, "learning_rate": 1.3373395987710103e-05, "loss": 0.8452, "step": 6133 }, { "epoch": 1.11, "learning_rate": 1.3370684981022955e-05, "loss": 0.7044, "step": 6134 }, { "epoch": 1.11, "learning_rate": 1.3367973974335803e-05, "loss": 0.865, "step": 6135 }, { "epoch": 1.11, "learning_rate": 1.3365262967648653e-05, "loss": 0.8789, "step": 6136 }, { "epoch": 1.11, "learning_rate": 1.3362551960961504e-05, "loss": 0.6229, "step": 6137 }, { "epoch": 1.11, "learning_rate": 1.3359840954274355e-05, "loss": 0.5162, "step": 6138 }, { "epoch": 1.11, "learning_rate": 1.3357129947587204e-05, "loss": 1.3168, "step": 6139 }, { "epoch": 1.11, "learning_rate": 1.3354418940900054e-05, "loss": 0.528, "step": 6140 }, { "epoch": 1.11, "learning_rate": 1.3351707934212906e-05, "loss": 0.7332, "step": 6141 }, { "epoch": 1.11, "learning_rate": 1.3348996927525756e-05, "loss": 0.6342, "step": 6142 }, { "epoch": 1.11, "learning_rate": 1.3346285920838604e-05, "loss": 0.9201, "step": 6143 }, { "epoch": 1.11, "learning_rate": 1.3343574914151454e-05, "loss": 0.4709, "step": 6144 }, { "epoch": 1.11, "learning_rate": 1.3340863907464306e-05, "loss": 0.912, "step": 6145 }, { "epoch": 1.11, "learning_rate": 1.3338152900777156e-05, "loss": 1.1978, "step": 6146 }, { "epoch": 1.11, "learning_rate": 1.3335441894090005e-05, "loss": 0.6843, "step": 6147 }, { "epoch": 1.11, "learning_rate": 1.3332730887402857e-05, "loss": 0.8842, "step": 6148 }, { "epoch": 1.11, "learning_rate": 1.3330019880715707e-05, "loss": 0.5563, "step": 6149 }, { "epoch": 1.11, "learning_rate": 1.3327308874028557e-05, "loss": 0.6073, "step": 6150 }, { "epoch": 1.11, "learning_rate": 1.3324597867341405e-05, "loss": 0.9835, "step": 6151 }, { "epoch": 1.11, "learning_rate": 1.3321886860654257e-05, "loss": 0.3019, "step": 6152 }, { "epoch": 1.11, "learning_rate": 1.3319175853967107e-05, "loss": 0.5619, "step": 6153 }, { "epoch": 1.11, "learning_rate": 1.3316464847279957e-05, "loss": 0.4364, "step": 6154 }, { "epoch": 1.11, "learning_rate": 1.3313753840592806e-05, "loss": 0.4416, "step": 6155 }, { "epoch": 1.11, "learning_rate": 1.3311042833905658e-05, "loss": 0.7417, "step": 6156 }, { "epoch": 1.11, "learning_rate": 1.3308331827218508e-05, "loss": 0.7508, "step": 6157 }, { "epoch": 1.11, "learning_rate": 1.3305620820531358e-05, "loss": 0.8621, "step": 6158 }, { "epoch": 1.11, "learning_rate": 1.3302909813844208e-05, "loss": 1.6416, "step": 6159 }, { "epoch": 1.11, "learning_rate": 1.3300198807157058e-05, "loss": 0.8251, "step": 6160 }, { "epoch": 1.11, "learning_rate": 1.3297487800469908e-05, "loss": 0.5112, "step": 6161 }, { "epoch": 1.11, "learning_rate": 1.3294776793782759e-05, "loss": 1.049, "step": 6162 }, { "epoch": 1.11, "learning_rate": 1.3292065787095609e-05, "loss": 1.0598, "step": 6163 }, { "epoch": 1.11, "learning_rate": 1.3289354780408459e-05, "loss": 0.6682, "step": 6164 }, { "epoch": 1.11, "learning_rate": 1.3286643773721309e-05, "loss": 0.7128, "step": 6165 }, { "epoch": 1.11, "learning_rate": 1.3283932767034159e-05, "loss": 0.5696, "step": 6166 }, { "epoch": 1.11, "learning_rate": 1.328122176034701e-05, "loss": 0.6331, "step": 6167 }, { "epoch": 1.11, "learning_rate": 1.327851075365986e-05, "loss": 0.9614, "step": 6168 }, { "epoch": 1.11, "learning_rate": 1.327579974697271e-05, "loss": 0.443, "step": 6169 }, { "epoch": 1.12, "learning_rate": 1.3273088740285561e-05, "loss": 1.0588, "step": 6170 }, { "epoch": 1.12, "learning_rate": 1.327037773359841e-05, "loss": 0.6571, "step": 6171 }, { "epoch": 1.12, "learning_rate": 1.326766672691126e-05, "loss": 0.9747, "step": 6172 }, { "epoch": 1.12, "learning_rate": 1.326495572022411e-05, "loss": 0.5156, "step": 6173 }, { "epoch": 1.12, "learning_rate": 1.3262244713536962e-05, "loss": 0.7507, "step": 6174 }, { "epoch": 1.12, "learning_rate": 1.325953370684981e-05, "loss": 0.7796, "step": 6175 }, { "epoch": 1.12, "learning_rate": 1.325682270016266e-05, "loss": 0.4679, "step": 6176 }, { "epoch": 1.12, "learning_rate": 1.325411169347551e-05, "loss": 0.5842, "step": 6177 }, { "epoch": 1.12, "learning_rate": 1.325140068678836e-05, "loss": 0.5943, "step": 6178 }, { "epoch": 1.12, "learning_rate": 1.324868968010121e-05, "loss": 0.6149, "step": 6179 }, { "epoch": 1.12, "learning_rate": 1.3245978673414061e-05, "loss": 0.5097, "step": 6180 }, { "epoch": 1.12, "learning_rate": 1.3243267666726913e-05, "loss": 0.8437, "step": 6181 }, { "epoch": 1.12, "learning_rate": 1.3240556660039761e-05, "loss": 0.693, "step": 6182 }, { "epoch": 1.12, "learning_rate": 1.3237845653352611e-05, "loss": 0.3009, "step": 6183 }, { "epoch": 1.12, "learning_rate": 1.3235134646665462e-05, "loss": 0.3705, "step": 6184 }, { "epoch": 1.12, "learning_rate": 1.3232423639978313e-05, "loss": 0.8361, "step": 6185 }, { "epoch": 1.12, "learning_rate": 1.3229712633291162e-05, "loss": 0.6604, "step": 6186 }, { "epoch": 1.12, "learning_rate": 1.3227001626604012e-05, "loss": 0.9009, "step": 6187 }, { "epoch": 1.12, "learning_rate": 1.3224290619916862e-05, "loss": 1.1071, "step": 6188 }, { "epoch": 1.12, "learning_rate": 1.3221579613229714e-05, "loss": 0.9466, "step": 6189 }, { "epoch": 1.12, "learning_rate": 1.3218868606542562e-05, "loss": 0.5926, "step": 6190 }, { "epoch": 1.12, "learning_rate": 1.3216157599855412e-05, "loss": 0.4299, "step": 6191 }, { "epoch": 1.12, "learning_rate": 1.3213446593168264e-05, "loss": 0.6417, "step": 6192 }, { "epoch": 1.12, "learning_rate": 1.3210735586481114e-05, "loss": 1.1849, "step": 6193 }, { "epoch": 1.12, "learning_rate": 1.3208024579793963e-05, "loss": 0.8757, "step": 6194 }, { "epoch": 1.12, "learning_rate": 1.3205313573106813e-05, "loss": 0.64, "step": 6195 }, { "epoch": 1.12, "learning_rate": 1.3202602566419665e-05, "loss": 0.8668, "step": 6196 }, { "epoch": 1.12, "learning_rate": 1.3199891559732515e-05, "loss": 0.737, "step": 6197 }, { "epoch": 1.12, "learning_rate": 1.3197180553045363e-05, "loss": 0.9014, "step": 6198 }, { "epoch": 1.12, "learning_rate": 1.3194469546358214e-05, "loss": 0.8172, "step": 6199 }, { "epoch": 1.12, "learning_rate": 1.3191758539671065e-05, "loss": 0.5132, "step": 6200 }, { "epoch": 1.12, "learning_rate": 1.3189047532983915e-05, "loss": 0.8469, "step": 6201 }, { "epoch": 1.12, "learning_rate": 1.3186336526296764e-05, "loss": 0.4232, "step": 6202 }, { "epoch": 1.12, "learning_rate": 1.3183625519609616e-05, "loss": 0.9239, "step": 6203 }, { "epoch": 1.12, "learning_rate": 1.3180914512922466e-05, "loss": 0.89, "step": 6204 }, { "epoch": 1.12, "learning_rate": 1.3178203506235316e-05, "loss": 0.7265, "step": 6205 }, { "epoch": 1.12, "learning_rate": 1.3175492499548164e-05, "loss": 0.3803, "step": 6206 }, { "epoch": 1.12, "learning_rate": 1.3172781492861016e-05, "loss": 0.5901, "step": 6207 }, { "epoch": 1.12, "learning_rate": 1.3170070486173866e-05, "loss": 1.5071, "step": 6208 }, { "epoch": 1.12, "learning_rate": 1.3167359479486717e-05, "loss": 0.3956, "step": 6209 }, { "epoch": 1.12, "learning_rate": 1.3164648472799567e-05, "loss": 0.8102, "step": 6210 }, { "epoch": 1.12, "learning_rate": 1.3161937466112417e-05, "loss": 1.1202, "step": 6211 }, { "epoch": 1.12, "learning_rate": 1.3159226459425267e-05, "loss": 0.997, "step": 6212 }, { "epoch": 1.12, "learning_rate": 1.3156515452738117e-05, "loss": 0.48, "step": 6213 }, { "epoch": 1.12, "learning_rate": 1.3153804446050967e-05, "loss": 1.0016, "step": 6214 }, { "epoch": 1.12, "learning_rate": 1.3151093439363817e-05, "loss": 0.8513, "step": 6215 }, { "epoch": 1.12, "learning_rate": 1.3148382432676668e-05, "loss": 1.1349, "step": 6216 }, { "epoch": 1.12, "learning_rate": 1.3145671425989518e-05, "loss": 0.3783, "step": 6217 }, { "epoch": 1.12, "learning_rate": 1.3142960419302368e-05, "loss": 0.7984, "step": 6218 }, { "epoch": 1.12, "learning_rate": 1.3140249412615218e-05, "loss": 1.0478, "step": 6219 }, { "epoch": 1.12, "learning_rate": 1.3137538405928068e-05, "loss": 1.3222, "step": 6220 }, { "epoch": 1.12, "learning_rate": 1.313482739924092e-05, "loss": 0.928, "step": 6221 }, { "epoch": 1.12, "learning_rate": 1.3132116392553768e-05, "loss": 0.7064, "step": 6222 }, { "epoch": 1.12, "learning_rate": 1.3129405385866618e-05, "loss": 0.7003, "step": 6223 }, { "epoch": 1.12, "learning_rate": 1.3126694379179469e-05, "loss": 0.7898, "step": 6224 }, { "epoch": 1.13, "learning_rate": 1.312398337249232e-05, "loss": 0.7026, "step": 6225 }, { "epoch": 1.13, "learning_rate": 1.3121272365805169e-05, "loss": 0.3694, "step": 6226 }, { "epoch": 1.13, "learning_rate": 1.3118561359118019e-05, "loss": 0.8476, "step": 6227 }, { "epoch": 1.13, "learning_rate": 1.3115850352430869e-05, "loss": 0.3128, "step": 6228 }, { "epoch": 1.13, "learning_rate": 1.3113139345743721e-05, "loss": 1.4827, "step": 6229 }, { "epoch": 1.13, "learning_rate": 1.311042833905657e-05, "loss": 1.1308, "step": 6230 }, { "epoch": 1.13, "learning_rate": 1.310771733236942e-05, "loss": 0.5603, "step": 6231 }, { "epoch": 1.13, "learning_rate": 1.3105006325682271e-05, "loss": 1.3545, "step": 6232 }, { "epoch": 1.13, "learning_rate": 1.3102295318995121e-05, "loss": 0.6076, "step": 6233 }, { "epoch": 1.13, "learning_rate": 1.309958431230797e-05, "loss": 0.7324, "step": 6234 }, { "epoch": 1.13, "learning_rate": 1.309687330562082e-05, "loss": 0.6979, "step": 6235 }, { "epoch": 1.13, "learning_rate": 1.3094162298933672e-05, "loss": 0.8522, "step": 6236 }, { "epoch": 1.13, "learning_rate": 1.3091451292246522e-05, "loss": 1.0198, "step": 6237 }, { "epoch": 1.13, "learning_rate": 1.308874028555937e-05, "loss": 0.6362, "step": 6238 }, { "epoch": 1.13, "learning_rate": 1.308602927887222e-05, "loss": 1.034, "step": 6239 }, { "epoch": 1.13, "learning_rate": 1.3083318272185072e-05, "loss": 0.3182, "step": 6240 }, { "epoch": 1.13, "learning_rate": 1.3080607265497923e-05, "loss": 0.5005, "step": 6241 }, { "epoch": 1.13, "learning_rate": 1.3077896258810771e-05, "loss": 0.6326, "step": 6242 }, { "epoch": 1.13, "learning_rate": 1.3075185252123623e-05, "loss": 0.9357, "step": 6243 }, { "epoch": 1.13, "learning_rate": 1.3072474245436473e-05, "loss": 1.2534, "step": 6244 }, { "epoch": 1.13, "learning_rate": 1.3069763238749323e-05, "loss": 0.3894, "step": 6245 }, { "epoch": 1.13, "learning_rate": 1.3067052232062172e-05, "loss": 0.6693, "step": 6246 }, { "epoch": 1.13, "learning_rate": 1.3064341225375023e-05, "loss": 0.678, "step": 6247 }, { "epoch": 1.13, "learning_rate": 1.3061630218687873e-05, "loss": 0.8112, "step": 6248 }, { "epoch": 1.13, "learning_rate": 1.3058919212000724e-05, "loss": 0.5743, "step": 6249 }, { "epoch": 1.13, "learning_rate": 1.3056208205313572e-05, "loss": 1.0898, "step": 6250 }, { "epoch": 1.13, "learning_rate": 1.3053497198626424e-05, "loss": 1.0544, "step": 6251 }, { "epoch": 1.13, "learning_rate": 1.3050786191939274e-05, "loss": 0.706, "step": 6252 }, { "epoch": 1.13, "learning_rate": 1.3048075185252124e-05, "loss": 0.5533, "step": 6253 }, { "epoch": 1.13, "learning_rate": 1.3045364178564974e-05, "loss": 0.9592, "step": 6254 }, { "epoch": 1.13, "learning_rate": 1.3042653171877824e-05, "loss": 0.9352, "step": 6255 }, { "epoch": 1.13, "learning_rate": 1.3039942165190675e-05, "loss": 0.6091, "step": 6256 }, { "epoch": 1.13, "learning_rate": 1.3037231158503525e-05, "loss": 0.7162, "step": 6257 }, { "epoch": 1.13, "learning_rate": 1.3034520151816375e-05, "loss": 0.3892, "step": 6258 }, { "epoch": 1.13, "learning_rate": 1.3031809145129225e-05, "loss": 0.74, "step": 6259 }, { "epoch": 1.13, "learning_rate": 1.3029098138442075e-05, "loss": 0.8499, "step": 6260 }, { "epoch": 1.13, "learning_rate": 1.3026387131754925e-05, "loss": 0.7854, "step": 6261 }, { "epoch": 1.13, "learning_rate": 1.3023676125067775e-05, "loss": 0.9418, "step": 6262 }, { "epoch": 1.13, "learning_rate": 1.3020965118380626e-05, "loss": 0.6984, "step": 6263 }, { "epoch": 1.13, "learning_rate": 1.3018254111693476e-05, "loss": 0.9513, "step": 6264 }, { "epoch": 1.13, "learning_rate": 1.3015543105006327e-05, "loss": 0.796, "step": 6265 }, { "epoch": 1.13, "learning_rate": 1.3012832098319176e-05, "loss": 0.7072, "step": 6266 }, { "epoch": 1.13, "learning_rate": 1.3010121091632026e-05, "loss": 0.7162, "step": 6267 }, { "epoch": 1.13, "learning_rate": 1.3007410084944876e-05, "loss": 1.3401, "step": 6268 }, { "epoch": 1.13, "learning_rate": 1.3004699078257728e-05, "loss": 0.4969, "step": 6269 }, { "epoch": 1.13, "learning_rate": 1.3001988071570576e-05, "loss": 0.7597, "step": 6270 }, { "epoch": 1.13, "learning_rate": 1.2999277064883427e-05, "loss": 0.6423, "step": 6271 }, { "epoch": 1.13, "learning_rate": 1.2996566058196277e-05, "loss": 0.8652, "step": 6272 }, { "epoch": 1.13, "learning_rate": 1.2993855051509129e-05, "loss": 0.6046, "step": 6273 }, { "epoch": 1.13, "learning_rate": 1.2991144044821977e-05, "loss": 0.7569, "step": 6274 }, { "epoch": 1.13, "learning_rate": 1.2988433038134827e-05, "loss": 0.6642, "step": 6275 }, { "epoch": 1.13, "learning_rate": 1.2985722031447679e-05, "loss": 0.891, "step": 6276 }, { "epoch": 1.13, "learning_rate": 1.2983011024760529e-05, "loss": 1.1957, "step": 6277 }, { "epoch": 1.13, "learning_rate": 1.2980300018073378e-05, "loss": 1.2995, "step": 6278 }, { "epoch": 1.13, "learning_rate": 1.2977589011386228e-05, "loss": 0.5969, "step": 6279 }, { "epoch": 1.14, "learning_rate": 1.297487800469908e-05, "loss": 1.0949, "step": 6280 }, { "epoch": 1.14, "learning_rate": 1.297216699801193e-05, "loss": 0.6796, "step": 6281 }, { "epoch": 1.14, "learning_rate": 1.2969455991324778e-05, "loss": 0.7334, "step": 6282 }, { "epoch": 1.14, "learning_rate": 1.2966744984637628e-05, "loss": 0.6799, "step": 6283 }, { "epoch": 1.14, "learning_rate": 1.296403397795048e-05, "loss": 0.4949, "step": 6284 }, { "epoch": 1.14, "learning_rate": 1.296132297126333e-05, "loss": 0.9708, "step": 6285 }, { "epoch": 1.14, "learning_rate": 1.2958611964576179e-05, "loss": 1.0362, "step": 6286 }, { "epoch": 1.14, "learning_rate": 1.295590095788903e-05, "loss": 0.7686, "step": 6287 }, { "epoch": 1.14, "learning_rate": 1.295318995120188e-05, "loss": 1.1834, "step": 6288 }, { "epoch": 1.14, "learning_rate": 1.295047894451473e-05, "loss": 0.8694, "step": 6289 }, { "epoch": 1.14, "learning_rate": 1.2947767937827579e-05, "loss": 0.6905, "step": 6290 }, { "epoch": 1.14, "learning_rate": 1.2945056931140431e-05, "loss": 0.835, "step": 6291 }, { "epoch": 1.14, "learning_rate": 1.2942345924453281e-05, "loss": 0.4318, "step": 6292 }, { "epoch": 1.14, "learning_rate": 1.2939634917766131e-05, "loss": 0.4647, "step": 6293 }, { "epoch": 1.14, "learning_rate": 1.2936923911078981e-05, "loss": 0.5753, "step": 6294 }, { "epoch": 1.14, "learning_rate": 1.2934212904391831e-05, "loss": 0.7272, "step": 6295 }, { "epoch": 1.14, "learning_rate": 1.2931501897704682e-05, "loss": 1.1073, "step": 6296 }, { "epoch": 1.14, "learning_rate": 1.2928790891017532e-05, "loss": 0.6738, "step": 6297 }, { "epoch": 1.14, "learning_rate": 1.2926079884330382e-05, "loss": 0.6608, "step": 6298 }, { "epoch": 1.14, "learning_rate": 1.2923368877643232e-05, "loss": 0.7425, "step": 6299 }, { "epoch": 1.14, "learning_rate": 1.2920657870956082e-05, "loss": 1.1392, "step": 6300 }, { "epoch": 1.14, "learning_rate": 1.2917946864268932e-05, "loss": 0.8794, "step": 6301 }, { "epoch": 1.14, "learning_rate": 1.2915235857581782e-05, "loss": 0.8214, "step": 6302 }, { "epoch": 1.14, "learning_rate": 1.2912524850894633e-05, "loss": 1.2687, "step": 6303 }, { "epoch": 1.14, "learning_rate": 1.2909813844207483e-05, "loss": 0.3708, "step": 6304 }, { "epoch": 1.14, "learning_rate": 1.2907102837520335e-05, "loss": 0.6483, "step": 6305 }, { "epoch": 1.14, "learning_rate": 1.2904391830833183e-05, "loss": 1.102, "step": 6306 }, { "epoch": 1.14, "learning_rate": 1.2901680824146033e-05, "loss": 0.4869, "step": 6307 }, { "epoch": 1.14, "learning_rate": 1.2898969817458883e-05, "loss": 0.4796, "step": 6308 }, { "epoch": 1.14, "learning_rate": 1.2896258810771733e-05, "loss": 0.9457, "step": 6309 }, { "epoch": 1.14, "learning_rate": 1.2893547804084584e-05, "loss": 0.8779, "step": 6310 }, { "epoch": 1.14, "learning_rate": 1.2890836797397434e-05, "loss": 0.8772, "step": 6311 }, { "epoch": 1.14, "learning_rate": 1.2888125790710284e-05, "loss": 1.5246, "step": 6312 }, { "epoch": 1.14, "learning_rate": 1.2885414784023134e-05, "loss": 0.7619, "step": 6313 }, { "epoch": 1.14, "learning_rate": 1.2882703777335984e-05, "loss": 0.5698, "step": 6314 }, { "epoch": 1.14, "learning_rate": 1.2879992770648834e-05, "loss": 0.7231, "step": 6315 }, { "epoch": 1.14, "learning_rate": 1.2877281763961686e-05, "loss": 0.3488, "step": 6316 }, { "epoch": 1.14, "learning_rate": 1.2874570757274534e-05, "loss": 0.8592, "step": 6317 }, { "epoch": 1.14, "learning_rate": 1.2871859750587385e-05, "loss": 0.4722, "step": 6318 }, { "epoch": 1.14, "learning_rate": 1.2869148743900235e-05, "loss": 1.2459, "step": 6319 }, { "epoch": 1.14, "learning_rate": 1.2866437737213087e-05, "loss": 1.3073, "step": 6320 }, { "epoch": 1.14, "learning_rate": 1.2863726730525935e-05, "loss": 0.5651, "step": 6321 }, { "epoch": 1.14, "learning_rate": 1.2861015723838785e-05, "loss": 0.7705, "step": 6322 }, { "epoch": 1.14, "learning_rate": 1.2858304717151635e-05, "loss": 1.2908, "step": 6323 }, { "epoch": 1.14, "learning_rate": 1.2855593710464487e-05, "loss": 0.5805, "step": 6324 }, { "epoch": 1.14, "learning_rate": 1.2852882703777336e-05, "loss": 0.9386, "step": 6325 }, { "epoch": 1.14, "learning_rate": 1.2850171697090186e-05, "loss": 0.9373, "step": 6326 }, { "epoch": 1.14, "learning_rate": 1.2847460690403037e-05, "loss": 0.759, "step": 6327 }, { "epoch": 1.14, "learning_rate": 1.2844749683715888e-05, "loss": 0.7463, "step": 6328 }, { "epoch": 1.14, "learning_rate": 1.2842038677028736e-05, "loss": 0.9098, "step": 6329 }, { "epoch": 1.14, "learning_rate": 1.2839327670341586e-05, "loss": 0.5634, "step": 6330 }, { "epoch": 1.14, "learning_rate": 1.2836616663654438e-05, "loss": 0.9679, "step": 6331 }, { "epoch": 1.14, "learning_rate": 1.2833905656967288e-05, "loss": 0.7109, "step": 6332 }, { "epoch": 1.14, "learning_rate": 1.2831194650280137e-05, "loss": 0.6662, "step": 6333 }, { "epoch": 1.14, "learning_rate": 1.2828483643592987e-05, "loss": 1.0887, "step": 6334 }, { "epoch": 1.14, "learning_rate": 1.2825772636905839e-05, "loss": 0.8073, "step": 6335 }, { "epoch": 1.15, "learning_rate": 1.2823061630218689e-05, "loss": 0.9684, "step": 6336 }, { "epoch": 1.15, "learning_rate": 1.2820350623531537e-05, "loss": 1.3566, "step": 6337 }, { "epoch": 1.15, "learning_rate": 1.2817639616844389e-05, "loss": 0.3994, "step": 6338 }, { "epoch": 1.15, "learning_rate": 1.2814928610157239e-05, "loss": 0.6194, "step": 6339 }, { "epoch": 1.15, "learning_rate": 1.281221760347009e-05, "loss": 0.2282, "step": 6340 }, { "epoch": 1.15, "learning_rate": 1.2809506596782938e-05, "loss": 0.3164, "step": 6341 }, { "epoch": 1.15, "learning_rate": 1.280679559009579e-05, "loss": 0.8155, "step": 6342 }, { "epoch": 1.15, "learning_rate": 1.280408458340864e-05, "loss": 0.8004, "step": 6343 }, { "epoch": 1.15, "learning_rate": 1.280137357672149e-05, "loss": 0.4303, "step": 6344 }, { "epoch": 1.15, "learning_rate": 1.2798662570034338e-05, "loss": 0.4617, "step": 6345 }, { "epoch": 1.15, "learning_rate": 1.279595156334719e-05, "loss": 0.413, "step": 6346 }, { "epoch": 1.15, "learning_rate": 1.279324055666004e-05, "loss": 0.7791, "step": 6347 }, { "epoch": 1.15, "learning_rate": 1.279052954997289e-05, "loss": 1.0579, "step": 6348 }, { "epoch": 1.15, "learning_rate": 1.278781854328574e-05, "loss": 0.7867, "step": 6349 }, { "epoch": 1.15, "learning_rate": 1.278510753659859e-05, "loss": 0.7576, "step": 6350 }, { "epoch": 1.15, "learning_rate": 1.278239652991144e-05, "loss": 0.5292, "step": 6351 }, { "epoch": 1.15, "learning_rate": 1.277968552322429e-05, "loss": 0.8325, "step": 6352 }, { "epoch": 1.15, "learning_rate": 1.2776974516537141e-05, "loss": 0.4815, "step": 6353 }, { "epoch": 1.15, "learning_rate": 1.2774263509849991e-05, "loss": 0.8654, "step": 6354 }, { "epoch": 1.15, "learning_rate": 1.2771552503162841e-05, "loss": 0.3588, "step": 6355 }, { "epoch": 1.15, "learning_rate": 1.2768841496475691e-05, "loss": 1.1332, "step": 6356 }, { "epoch": 1.15, "learning_rate": 1.2766130489788542e-05, "loss": 1.0681, "step": 6357 }, { "epoch": 1.15, "learning_rate": 1.2763419483101392e-05, "loss": 0.7456, "step": 6358 }, { "epoch": 1.15, "learning_rate": 1.2760708476414242e-05, "loss": 1.0204, "step": 6359 }, { "epoch": 1.15, "learning_rate": 1.2757997469727094e-05, "loss": 0.7675, "step": 6360 }, { "epoch": 1.15, "learning_rate": 1.2755286463039942e-05, "loss": 0.9255, "step": 6361 }, { "epoch": 1.15, "learning_rate": 1.2752575456352792e-05, "loss": 0.2724, "step": 6362 }, { "epoch": 1.15, "learning_rate": 1.2749864449665642e-05, "loss": 0.7007, "step": 6363 }, { "epoch": 1.15, "learning_rate": 1.2747153442978494e-05, "loss": 0.3575, "step": 6364 }, { "epoch": 1.15, "learning_rate": 1.2744442436291343e-05, "loss": 0.3988, "step": 6365 }, { "epoch": 1.15, "learning_rate": 1.2741731429604193e-05, "loss": 0.566, "step": 6366 }, { "epoch": 1.15, "learning_rate": 1.2739020422917045e-05, "loss": 0.7877, "step": 6367 }, { "epoch": 1.15, "learning_rate": 1.2736309416229895e-05, "loss": 0.4182, "step": 6368 }, { "epoch": 1.15, "learning_rate": 1.2733598409542743e-05, "loss": 0.3386, "step": 6369 }, { "epoch": 1.15, "learning_rate": 1.2730887402855593e-05, "loss": 0.5951, "step": 6370 }, { "epoch": 1.15, "learning_rate": 1.2728176396168445e-05, "loss": 0.3599, "step": 6371 }, { "epoch": 1.15, "learning_rate": 1.2725465389481295e-05, "loss": 0.8627, "step": 6372 }, { "epoch": 1.15, "learning_rate": 1.2722754382794144e-05, "loss": 0.6682, "step": 6373 }, { "epoch": 1.15, "learning_rate": 1.2720043376106994e-05, "loss": 0.8527, "step": 6374 }, { "epoch": 1.15, "learning_rate": 1.2717332369419846e-05, "loss": 0.8002, "step": 6375 }, { "epoch": 1.15, "learning_rate": 1.2714621362732696e-05, "loss": 0.3303, "step": 6376 }, { "epoch": 1.15, "learning_rate": 1.2711910356045544e-05, "loss": 0.3699, "step": 6377 }, { "epoch": 1.15, "learning_rate": 1.2709199349358396e-05, "loss": 1.0323, "step": 6378 }, { "epoch": 1.15, "learning_rate": 1.2706488342671246e-05, "loss": 1.1121, "step": 6379 }, { "epoch": 1.15, "learning_rate": 1.2703777335984096e-05, "loss": 0.6671, "step": 6380 }, { "epoch": 1.15, "learning_rate": 1.2701066329296945e-05, "loss": 0.4811, "step": 6381 }, { "epoch": 1.15, "learning_rate": 1.2698355322609797e-05, "loss": 0.3926, "step": 6382 }, { "epoch": 1.15, "learning_rate": 1.2695644315922647e-05, "loss": 0.6474, "step": 6383 }, { "epoch": 1.15, "learning_rate": 1.2692933309235497e-05, "loss": 0.5972, "step": 6384 }, { "epoch": 1.15, "learning_rate": 1.2690222302548345e-05, "loss": 0.5509, "step": 6385 }, { "epoch": 1.15, "learning_rate": 1.2687511295861197e-05, "loss": 0.4739, "step": 6386 }, { "epoch": 1.15, "learning_rate": 1.2684800289174047e-05, "loss": 0.4934, "step": 6387 }, { "epoch": 1.15, "learning_rate": 1.2682089282486897e-05, "loss": 1.1021, "step": 6388 }, { "epoch": 1.15, "learning_rate": 1.2679378275799748e-05, "loss": 0.3672, "step": 6389 }, { "epoch": 1.15, "learning_rate": 1.2676667269112598e-05, "loss": 0.8824, "step": 6390 }, { "epoch": 1.16, "learning_rate": 1.2673956262425448e-05, "loss": 0.8742, "step": 6391 }, { "epoch": 1.16, "learning_rate": 1.2671245255738298e-05, "loss": 1.605, "step": 6392 }, { "epoch": 1.16, "learning_rate": 1.2668534249051148e-05, "loss": 0.8753, "step": 6393 }, { "epoch": 1.16, "learning_rate": 1.2665823242363998e-05, "loss": 0.6917, "step": 6394 }, { "epoch": 1.16, "learning_rate": 1.2663112235676848e-05, "loss": 0.5042, "step": 6395 }, { "epoch": 1.16, "learning_rate": 1.2660401228989698e-05, "loss": 0.6785, "step": 6396 }, { "epoch": 1.16, "learning_rate": 1.2657690222302549e-05, "loss": 0.7277, "step": 6397 }, { "epoch": 1.16, "learning_rate": 1.2654979215615399e-05, "loss": 0.7969, "step": 6398 }, { "epoch": 1.16, "learning_rate": 1.2652268208928249e-05, "loss": 1.2634, "step": 6399 }, { "epoch": 1.16, "learning_rate": 1.26495572022411e-05, "loss": 1.1243, "step": 6400 }, { "epoch": 1.16, "learning_rate": 1.2646846195553949e-05, "loss": 1.0245, "step": 6401 }, { "epoch": 1.16, "learning_rate": 1.26441351888668e-05, "loss": 0.6678, "step": 6402 }, { "epoch": 1.16, "learning_rate": 1.264142418217965e-05, "loss": 0.8244, "step": 6403 }, { "epoch": 1.16, "learning_rate": 1.2638713175492501e-05, "loss": 0.5072, "step": 6404 }, { "epoch": 1.16, "learning_rate": 1.263600216880535e-05, "loss": 1.0529, "step": 6405 }, { "epoch": 1.16, "learning_rate": 1.26332911621182e-05, "loss": 0.885, "step": 6406 }, { "epoch": 1.16, "learning_rate": 1.263058015543105e-05, "loss": 0.9148, "step": 6407 }, { "epoch": 1.16, "learning_rate": 1.2627869148743902e-05, "loss": 0.9444, "step": 6408 }, { "epoch": 1.16, "learning_rate": 1.262515814205675e-05, "loss": 0.4614, "step": 6409 }, { "epoch": 1.16, "learning_rate": 1.26224471353696e-05, "loss": 1.3502, "step": 6410 }, { "epoch": 1.16, "learning_rate": 1.2619736128682452e-05, "loss": 0.6912, "step": 6411 }, { "epoch": 1.16, "learning_rate": 1.2617025121995302e-05, "loss": 0.5585, "step": 6412 }, { "epoch": 1.16, "learning_rate": 1.261431411530815e-05, "loss": 0.6058, "step": 6413 }, { "epoch": 1.16, "learning_rate": 1.2611603108621001e-05, "loss": 1.058, "step": 6414 }, { "epoch": 1.16, "learning_rate": 1.2608892101933853e-05, "loss": 0.4257, "step": 6415 }, { "epoch": 1.16, "learning_rate": 1.2606181095246703e-05, "loss": 1.2023, "step": 6416 }, { "epoch": 1.16, "learning_rate": 1.2603470088559551e-05, "loss": 1.1217, "step": 6417 }, { "epoch": 1.16, "learning_rate": 1.2600759081872401e-05, "loss": 0.6235, "step": 6418 }, { "epoch": 1.16, "learning_rate": 1.2598048075185253e-05, "loss": 1.2784, "step": 6419 }, { "epoch": 1.16, "learning_rate": 1.2595337068498103e-05, "loss": 0.7181, "step": 6420 }, { "epoch": 1.16, "learning_rate": 1.2592626061810952e-05, "loss": 0.9669, "step": 6421 }, { "epoch": 1.16, "learning_rate": 1.2589915055123804e-05, "loss": 0.5905, "step": 6422 }, { "epoch": 1.16, "learning_rate": 1.2587204048436654e-05, "loss": 0.7687, "step": 6423 }, { "epoch": 1.16, "learning_rate": 1.2584493041749504e-05, "loss": 1.1354, "step": 6424 }, { "epoch": 1.16, "learning_rate": 1.2581782035062352e-05, "loss": 0.3847, "step": 6425 }, { "epoch": 1.16, "learning_rate": 1.2579071028375204e-05, "loss": 0.6026, "step": 6426 }, { "epoch": 1.16, "learning_rate": 1.2576360021688054e-05, "loss": 1.0003, "step": 6427 }, { "epoch": 1.16, "learning_rate": 1.2573649015000904e-05, "loss": 1.0054, "step": 6428 }, { "epoch": 1.16, "learning_rate": 1.2570938008313755e-05, "loss": 0.6941, "step": 6429 }, { "epoch": 1.16, "learning_rate": 1.2568227001626605e-05, "loss": 0.8265, "step": 6430 }, { "epoch": 1.16, "learning_rate": 1.2565515994939455e-05, "loss": 0.8575, "step": 6431 }, { "epoch": 1.16, "learning_rate": 1.2562804988252305e-05, "loss": 1.19, "step": 6432 }, { "epoch": 1.16, "learning_rate": 1.2560093981565155e-05, "loss": 0.6002, "step": 6433 }, { "epoch": 1.16, "learning_rate": 1.2557382974878005e-05, "loss": 0.7103, "step": 6434 }, { "epoch": 1.16, "learning_rate": 1.2554671968190855e-05, "loss": 0.7917, "step": 6435 }, { "epoch": 1.16, "learning_rate": 1.2551960961503704e-05, "loss": 1.3591, "step": 6436 }, { "epoch": 1.16, "learning_rate": 1.2549249954816556e-05, "loss": 0.6386, "step": 6437 }, { "epoch": 1.16, "learning_rate": 1.2546538948129406e-05, "loss": 0.9402, "step": 6438 }, { "epoch": 1.16, "learning_rate": 1.2543827941442256e-05, "loss": 0.8583, "step": 6439 }, { "epoch": 1.16, "learning_rate": 1.2541116934755106e-05, "loss": 0.4781, "step": 6440 }, { "epoch": 1.16, "learning_rate": 1.2538405928067956e-05, "loss": 0.3711, "step": 6441 }, { "epoch": 1.16, "learning_rate": 1.2535694921380806e-05, "loss": 0.6488, "step": 6442 }, { "epoch": 1.16, "learning_rate": 1.2532983914693656e-05, "loss": 1.5025, "step": 6443 }, { "epoch": 1.16, "learning_rate": 1.2530272908006507e-05, "loss": 0.4595, "step": 6444 }, { "epoch": 1.16, "learning_rate": 1.2527561901319357e-05, "loss": 0.6928, "step": 6445 }, { "epoch": 1.17, "learning_rate": 1.2524850894632207e-05, "loss": 0.6008, "step": 6446 }, { "epoch": 1.17, "learning_rate": 1.2522139887945057e-05, "loss": 0.3659, "step": 6447 }, { "epoch": 1.17, "learning_rate": 1.2519428881257907e-05, "loss": 0.5121, "step": 6448 }, { "epoch": 1.17, "learning_rate": 1.2516717874570757e-05, "loss": 0.887, "step": 6449 }, { "epoch": 1.17, "learning_rate": 1.2514006867883607e-05, "loss": 0.8846, "step": 6450 }, { "epoch": 1.17, "learning_rate": 1.251129586119646e-05, "loss": 0.3213, "step": 6451 }, { "epoch": 1.17, "learning_rate": 1.2508584854509308e-05, "loss": 0.7707, "step": 6452 }, { "epoch": 1.17, "learning_rate": 1.2505873847822158e-05, "loss": 0.5519, "step": 6453 }, { "epoch": 1.17, "learning_rate": 1.2503162841135008e-05, "loss": 0.3421, "step": 6454 }, { "epoch": 1.17, "learning_rate": 1.250045183444786e-05, "loss": 0.4982, "step": 6455 }, { "epoch": 1.17, "learning_rate": 1.2497740827760708e-05, "loss": 0.7351, "step": 6456 }, { "epoch": 1.17, "learning_rate": 1.2495029821073558e-05, "loss": 0.6069, "step": 6457 }, { "epoch": 1.17, "learning_rate": 1.2492318814386408e-05, "loss": 0.3631, "step": 6458 }, { "epoch": 1.17, "learning_rate": 1.248960780769926e-05, "loss": 0.4433, "step": 6459 }, { "epoch": 1.17, "learning_rate": 1.2486896801012109e-05, "loss": 0.7834, "step": 6460 }, { "epoch": 1.17, "learning_rate": 1.2484185794324959e-05, "loss": 0.408, "step": 6461 }, { "epoch": 1.17, "learning_rate": 1.248147478763781e-05, "loss": 1.7454, "step": 6462 }, { "epoch": 1.17, "learning_rate": 1.247876378095066e-05, "loss": 0.7318, "step": 6463 }, { "epoch": 1.17, "learning_rate": 1.247605277426351e-05, "loss": 0.7207, "step": 6464 }, { "epoch": 1.17, "learning_rate": 1.247334176757636e-05, "loss": 0.7997, "step": 6465 }, { "epoch": 1.17, "learning_rate": 1.2470630760889211e-05, "loss": 1.0833, "step": 6466 }, { "epoch": 1.17, "learning_rate": 1.2467919754202061e-05, "loss": 0.9509, "step": 6467 }, { "epoch": 1.17, "learning_rate": 1.246520874751491e-05, "loss": 1.2502, "step": 6468 }, { "epoch": 1.17, "learning_rate": 1.246249774082776e-05, "loss": 0.889, "step": 6469 }, { "epoch": 1.17, "learning_rate": 1.2459786734140612e-05, "loss": 0.9816, "step": 6470 }, { "epoch": 1.17, "learning_rate": 1.2457075727453462e-05, "loss": 0.714, "step": 6471 }, { "epoch": 1.17, "learning_rate": 1.245436472076631e-05, "loss": 0.7084, "step": 6472 }, { "epoch": 1.17, "learning_rate": 1.2451653714079162e-05, "loss": 1.0743, "step": 6473 }, { "epoch": 1.17, "learning_rate": 1.2448942707392012e-05, "loss": 0.8788, "step": 6474 }, { "epoch": 1.17, "learning_rate": 1.2446231700704862e-05, "loss": 1.7487, "step": 6475 }, { "epoch": 1.17, "learning_rate": 1.2443520694017711e-05, "loss": 0.9687, "step": 6476 }, { "epoch": 1.17, "learning_rate": 1.2440809687330563e-05, "loss": 0.3445, "step": 6477 }, { "epoch": 1.17, "learning_rate": 1.2438098680643413e-05, "loss": 0.7704, "step": 6478 }, { "epoch": 1.17, "learning_rate": 1.2435387673956263e-05, "loss": 0.6117, "step": 6479 }, { "epoch": 1.17, "learning_rate": 1.2432676667269111e-05, "loss": 0.8468, "step": 6480 }, { "epoch": 1.17, "learning_rate": 1.2429965660581963e-05, "loss": 0.5808, "step": 6481 }, { "epoch": 1.17, "learning_rate": 1.2427254653894813e-05, "loss": 0.6651, "step": 6482 }, { "epoch": 1.17, "learning_rate": 1.2424543647207664e-05, "loss": 0.6929, "step": 6483 }, { "epoch": 1.17, "learning_rate": 1.2421832640520514e-05, "loss": 1.379, "step": 6484 }, { "epoch": 1.17, "learning_rate": 1.2419121633833364e-05, "loss": 0.8245, "step": 6485 }, { "epoch": 1.17, "learning_rate": 1.2416410627146214e-05, "loss": 0.6173, "step": 6486 }, { "epoch": 1.17, "learning_rate": 1.2413699620459064e-05, "loss": 0.9656, "step": 6487 }, { "epoch": 1.17, "learning_rate": 1.2410988613771914e-05, "loss": 0.3949, "step": 6488 }, { "epoch": 1.17, "learning_rate": 1.2408277607084764e-05, "loss": 1.2557, "step": 6489 }, { "epoch": 1.17, "learning_rate": 1.2405566600397614e-05, "loss": 0.5234, "step": 6490 }, { "epoch": 1.17, "learning_rate": 1.2402855593710465e-05, "loss": 0.7472, "step": 6491 }, { "epoch": 1.17, "learning_rate": 1.2400144587023315e-05, "loss": 0.7477, "step": 6492 }, { "epoch": 1.17, "learning_rate": 1.2397433580336165e-05, "loss": 0.5979, "step": 6493 }, { "epoch": 1.17, "learning_rate": 1.2394722573649015e-05, "loss": 0.7197, "step": 6494 }, { "epoch": 1.17, "learning_rate": 1.2392011566961867e-05, "loss": 0.7786, "step": 6495 }, { "epoch": 1.17, "learning_rate": 1.2389300560274715e-05, "loss": 0.5717, "step": 6496 }, { "epoch": 1.17, "learning_rate": 1.2386589553587565e-05, "loss": 0.399, "step": 6497 }, { "epoch": 1.17, "learning_rate": 1.2383878546900416e-05, "loss": 0.9171, "step": 6498 }, { "epoch": 1.17, "learning_rate": 1.2381167540213267e-05, "loss": 0.2632, "step": 6499 }, { "epoch": 1.17, "learning_rate": 1.2378456533526116e-05, "loss": 0.8072, "step": 6500 }, { "epoch": 1.17, "learning_rate": 1.2375745526838966e-05, "loss": 0.6546, "step": 6501 }, { "epoch": 1.18, "learning_rate": 1.2373034520151816e-05, "loss": 0.3771, "step": 6502 }, { "epoch": 1.18, "learning_rate": 1.2370323513464668e-05, "loss": 0.4381, "step": 6503 }, { "epoch": 1.18, "learning_rate": 1.2367612506777516e-05, "loss": 0.6862, "step": 6504 }, { "epoch": 1.18, "learning_rate": 1.2364901500090366e-05, "loss": 0.4519, "step": 6505 }, { "epoch": 1.18, "learning_rate": 1.2362190493403218e-05, "loss": 0.5725, "step": 6506 }, { "epoch": 1.18, "learning_rate": 1.2359479486716068e-05, "loss": 0.2837, "step": 6507 }, { "epoch": 1.18, "learning_rate": 1.2356768480028917e-05, "loss": 0.5084, "step": 6508 }, { "epoch": 1.18, "learning_rate": 1.2354057473341767e-05, "loss": 1.1361, "step": 6509 }, { "epoch": 1.18, "learning_rate": 1.2351346466654619e-05, "loss": 0.3455, "step": 6510 }, { "epoch": 1.18, "learning_rate": 1.2348635459967469e-05, "loss": 0.853, "step": 6511 }, { "epoch": 1.18, "learning_rate": 1.2345924453280317e-05, "loss": 1.0885, "step": 6512 }, { "epoch": 1.18, "learning_rate": 1.234321344659317e-05, "loss": 0.5123, "step": 6513 }, { "epoch": 1.18, "learning_rate": 1.234050243990602e-05, "loss": 0.1509, "step": 6514 }, { "epoch": 1.18, "learning_rate": 1.233779143321887e-05, "loss": 0.3887, "step": 6515 }, { "epoch": 1.18, "learning_rate": 1.2335080426531718e-05, "loss": 0.5246, "step": 6516 }, { "epoch": 1.18, "learning_rate": 1.233236941984457e-05, "loss": 0.8463, "step": 6517 }, { "epoch": 1.18, "learning_rate": 1.232965841315742e-05, "loss": 0.3708, "step": 6518 }, { "epoch": 1.18, "learning_rate": 1.232694740647027e-05, "loss": 1.5485, "step": 6519 }, { "epoch": 1.18, "learning_rate": 1.2324236399783118e-05, "loss": 0.8813, "step": 6520 }, { "epoch": 1.18, "learning_rate": 1.232152539309597e-05, "loss": 0.7185, "step": 6521 }, { "epoch": 1.18, "learning_rate": 1.231881438640882e-05, "loss": 1.0883, "step": 6522 }, { "epoch": 1.18, "learning_rate": 1.231610337972167e-05, "loss": 1.0791, "step": 6523 }, { "epoch": 1.18, "learning_rate": 1.231339237303452e-05, "loss": 0.5397, "step": 6524 }, { "epoch": 1.18, "learning_rate": 1.2310681366347371e-05, "loss": 0.237, "step": 6525 }, { "epoch": 1.18, "learning_rate": 1.2307970359660221e-05, "loss": 1.1271, "step": 6526 }, { "epoch": 1.18, "learning_rate": 1.2305259352973071e-05, "loss": 0.3513, "step": 6527 }, { "epoch": 1.18, "learning_rate": 1.2302548346285921e-05, "loss": 1.3603, "step": 6528 }, { "epoch": 1.18, "learning_rate": 1.2299837339598771e-05, "loss": 0.7032, "step": 6529 }, { "epoch": 1.18, "learning_rate": 1.2297126332911622e-05, "loss": 0.8888, "step": 6530 }, { "epoch": 1.18, "learning_rate": 1.2294415326224472e-05, "loss": 0.3433, "step": 6531 }, { "epoch": 1.18, "learning_rate": 1.2291704319537322e-05, "loss": 0.615, "step": 6532 }, { "epoch": 1.18, "learning_rate": 1.2288993312850172e-05, "loss": 0.4064, "step": 6533 }, { "epoch": 1.18, "learning_rate": 1.2286282306163022e-05, "loss": 0.694, "step": 6534 }, { "epoch": 1.18, "learning_rate": 1.2283571299475874e-05, "loss": 0.4127, "step": 6535 }, { "epoch": 1.18, "learning_rate": 1.2280860292788722e-05, "loss": 1.0799, "step": 6536 }, { "epoch": 1.18, "learning_rate": 1.2278149286101572e-05, "loss": 0.5359, "step": 6537 }, { "epoch": 1.18, "learning_rate": 1.2275438279414423e-05, "loss": 0.6206, "step": 6538 }, { "epoch": 1.18, "learning_rate": 1.2272727272727274e-05, "loss": 0.9596, "step": 6539 }, { "epoch": 1.18, "learning_rate": 1.2270016266040123e-05, "loss": 0.6268, "step": 6540 }, { "epoch": 1.18, "learning_rate": 1.2267305259352973e-05, "loss": 0.6175, "step": 6541 }, { "epoch": 1.18, "learning_rate": 1.2264594252665823e-05, "loss": 0.6118, "step": 6542 }, { "epoch": 1.18, "learning_rate": 1.2261883245978675e-05, "loss": 0.6777, "step": 6543 }, { "epoch": 1.18, "learning_rate": 1.2259172239291523e-05, "loss": 0.6634, "step": 6544 }, { "epoch": 1.18, "learning_rate": 1.2256461232604374e-05, "loss": 0.7132, "step": 6545 }, { "epoch": 1.18, "learning_rate": 1.2253750225917225e-05, "loss": 1.5854, "step": 6546 }, { "epoch": 1.18, "learning_rate": 1.2251039219230075e-05, "loss": 0.903, "step": 6547 }, { "epoch": 1.18, "learning_rate": 1.2248328212542924e-05, "loss": 0.7507, "step": 6548 }, { "epoch": 1.18, "learning_rate": 1.2245617205855774e-05, "loss": 0.4926, "step": 6549 }, { "epoch": 1.18, "learning_rate": 1.2242906199168626e-05, "loss": 0.3488, "step": 6550 }, { "epoch": 1.18, "learning_rate": 1.2240195192481476e-05, "loss": 0.7626, "step": 6551 }, { "epoch": 1.18, "learning_rate": 1.2237484185794324e-05, "loss": 0.8279, "step": 6552 }, { "epoch": 1.18, "learning_rate": 1.2234773179107175e-05, "loss": 0.6161, "step": 6553 }, { "epoch": 1.18, "learning_rate": 1.2232062172420026e-05, "loss": 0.5911, "step": 6554 }, { "epoch": 1.18, "learning_rate": 1.2229351165732877e-05, "loss": 0.922, "step": 6555 }, { "epoch": 1.18, "learning_rate": 1.2226640159045725e-05, "loss": 0.597, "step": 6556 }, { "epoch": 1.19, "learning_rate": 1.2223929152358577e-05, "loss": 0.7614, "step": 6557 }, { "epoch": 1.19, "learning_rate": 1.2221218145671427e-05, "loss": 1.1001, "step": 6558 }, { "epoch": 1.19, "learning_rate": 1.2218507138984277e-05, "loss": 0.7458, "step": 6559 }, { "epoch": 1.19, "learning_rate": 1.2215796132297126e-05, "loss": 1.6541, "step": 6560 }, { "epoch": 1.19, "learning_rate": 1.2213085125609977e-05, "loss": 0.7054, "step": 6561 }, { "epoch": 1.19, "learning_rate": 1.2210374118922828e-05, "loss": 0.5629, "step": 6562 }, { "epoch": 1.19, "learning_rate": 1.2207663112235676e-05, "loss": 0.4375, "step": 6563 }, { "epoch": 1.19, "learning_rate": 1.2204952105548526e-05, "loss": 1.1596, "step": 6564 }, { "epoch": 1.19, "learning_rate": 1.2202241098861378e-05, "loss": 0.5133, "step": 6565 }, { "epoch": 1.19, "learning_rate": 1.2199530092174228e-05, "loss": 0.9913, "step": 6566 }, { "epoch": 1.19, "learning_rate": 1.2196819085487076e-05, "loss": 0.4922, "step": 6567 }, { "epoch": 1.19, "learning_rate": 1.2194108078799928e-05, "loss": 0.5061, "step": 6568 }, { "epoch": 1.19, "learning_rate": 1.2191397072112778e-05, "loss": 1.0862, "step": 6569 }, { "epoch": 1.19, "learning_rate": 1.2188686065425629e-05, "loss": 1.0112, "step": 6570 }, { "epoch": 1.19, "learning_rate": 1.2185975058738477e-05, "loss": 0.3718, "step": 6571 }, { "epoch": 1.19, "learning_rate": 1.2183264052051329e-05, "loss": 0.6588, "step": 6572 }, { "epoch": 1.19, "learning_rate": 1.2180553045364179e-05, "loss": 1.1649, "step": 6573 }, { "epoch": 1.19, "learning_rate": 1.2177842038677029e-05, "loss": 0.7089, "step": 6574 }, { "epoch": 1.19, "learning_rate": 1.217513103198988e-05, "loss": 1.0986, "step": 6575 }, { "epoch": 1.19, "learning_rate": 1.217242002530273e-05, "loss": 0.7297, "step": 6576 }, { "epoch": 1.19, "learning_rate": 1.216970901861558e-05, "loss": 0.7003, "step": 6577 }, { "epoch": 1.19, "learning_rate": 1.216699801192843e-05, "loss": 0.7548, "step": 6578 }, { "epoch": 1.19, "learning_rate": 1.216428700524128e-05, "loss": 1.2692, "step": 6579 }, { "epoch": 1.19, "learning_rate": 1.216157599855413e-05, "loss": 0.3116, "step": 6580 }, { "epoch": 1.19, "learning_rate": 1.215886499186698e-05, "loss": 0.4447, "step": 6581 }, { "epoch": 1.19, "learning_rate": 1.215615398517983e-05, "loss": 0.7836, "step": 6582 }, { "epoch": 1.19, "learning_rate": 1.215344297849268e-05, "loss": 0.5022, "step": 6583 }, { "epoch": 1.19, "learning_rate": 1.215073197180553e-05, "loss": 0.7238, "step": 6584 }, { "epoch": 1.19, "learning_rate": 1.214802096511838e-05, "loss": 0.4024, "step": 6585 }, { "epoch": 1.19, "learning_rate": 1.2145309958431232e-05, "loss": 0.8071, "step": 6586 }, { "epoch": 1.19, "learning_rate": 1.2142598951744081e-05, "loss": 0.8077, "step": 6587 }, { "epoch": 1.19, "learning_rate": 1.2139887945056931e-05, "loss": 0.6015, "step": 6588 }, { "epoch": 1.19, "learning_rate": 1.2137176938369781e-05, "loss": 0.5271, "step": 6589 }, { "epoch": 1.19, "learning_rate": 1.2134465931682633e-05, "loss": 0.3076, "step": 6590 }, { "epoch": 1.19, "learning_rate": 1.2131754924995481e-05, "loss": 0.525, "step": 6591 }, { "epoch": 1.19, "learning_rate": 1.2129043918308332e-05, "loss": 0.7324, "step": 6592 }, { "epoch": 1.19, "learning_rate": 1.2126332911621182e-05, "loss": 0.7709, "step": 6593 }, { "epoch": 1.19, "learning_rate": 1.2123621904934034e-05, "loss": 0.7182, "step": 6594 }, { "epoch": 1.19, "learning_rate": 1.2120910898246882e-05, "loss": 0.5859, "step": 6595 }, { "epoch": 1.19, "learning_rate": 1.2118199891559732e-05, "loss": 0.7984, "step": 6596 }, { "epoch": 1.19, "learning_rate": 1.2115488884872584e-05, "loss": 0.8507, "step": 6597 }, { "epoch": 1.19, "learning_rate": 1.2112777878185434e-05, "loss": 1.3842, "step": 6598 }, { "epoch": 1.19, "learning_rate": 1.2110066871498282e-05, "loss": 0.5764, "step": 6599 }, { "epoch": 1.19, "learning_rate": 1.2107355864811133e-05, "loss": 0.4751, "step": 6600 }, { "epoch": 1.19, "learning_rate": 1.2104644858123984e-05, "loss": 1.0302, "step": 6601 }, { "epoch": 1.19, "learning_rate": 1.2101933851436835e-05, "loss": 0.6489, "step": 6602 }, { "epoch": 1.19, "learning_rate": 1.2099222844749683e-05, "loss": 0.7988, "step": 6603 }, { "epoch": 1.19, "learning_rate": 1.2096511838062533e-05, "loss": 0.8447, "step": 6604 }, { "epoch": 1.19, "learning_rate": 1.2093800831375385e-05, "loss": 0.8654, "step": 6605 }, { "epoch": 1.19, "learning_rate": 1.2091089824688235e-05, "loss": 0.9138, "step": 6606 }, { "epoch": 1.19, "learning_rate": 1.2088378818001084e-05, "loss": 0.7611, "step": 6607 }, { "epoch": 1.19, "learning_rate": 1.2085667811313935e-05, "loss": 0.6682, "step": 6608 }, { "epoch": 1.19, "learning_rate": 1.2082956804626786e-05, "loss": 0.1835, "step": 6609 }, { "epoch": 1.19, "learning_rate": 1.2080245797939636e-05, "loss": 0.7497, "step": 6610 }, { "epoch": 1.19, "learning_rate": 1.2077534791252484e-05, "loss": 0.6658, "step": 6611 }, { "epoch": 1.2, "learning_rate": 1.2074823784565336e-05, "loss": 0.4575, "step": 6612 }, { "epoch": 1.2, "learning_rate": 1.2072112777878186e-05, "loss": 0.6721, "step": 6613 }, { "epoch": 1.2, "learning_rate": 1.2069401771191036e-05, "loss": 0.5622, "step": 6614 }, { "epoch": 1.2, "learning_rate": 1.2066690764503885e-05, "loss": 0.5794, "step": 6615 }, { "epoch": 1.2, "learning_rate": 1.2063979757816736e-05, "loss": 1.1448, "step": 6616 }, { "epoch": 1.2, "learning_rate": 1.2061268751129587e-05, "loss": 0.891, "step": 6617 }, { "epoch": 1.2, "learning_rate": 1.2058557744442437e-05, "loss": 1.1077, "step": 6618 }, { "epoch": 1.2, "learning_rate": 1.2055846737755287e-05, "loss": 0.6911, "step": 6619 }, { "epoch": 1.2, "learning_rate": 1.2053135731068137e-05, "loss": 0.3951, "step": 6620 }, { "epoch": 1.2, "learning_rate": 1.2050424724380987e-05, "loss": 0.5944, "step": 6621 }, { "epoch": 1.2, "learning_rate": 1.2047713717693837e-05, "loss": 0.5774, "step": 6622 }, { "epoch": 1.2, "learning_rate": 1.2045002711006687e-05, "loss": 0.8888, "step": 6623 }, { "epoch": 1.2, "learning_rate": 1.2042291704319538e-05, "loss": 0.6002, "step": 6624 }, { "epoch": 1.2, "learning_rate": 1.2039580697632388e-05, "loss": 0.6224, "step": 6625 }, { "epoch": 1.2, "learning_rate": 1.2036869690945238e-05, "loss": 0.9889, "step": 6626 }, { "epoch": 1.2, "learning_rate": 1.2034158684258088e-05, "loss": 0.8903, "step": 6627 }, { "epoch": 1.2, "learning_rate": 1.2031447677570938e-05, "loss": 0.8149, "step": 6628 }, { "epoch": 1.2, "learning_rate": 1.2028736670883788e-05, "loss": 0.7644, "step": 6629 }, { "epoch": 1.2, "learning_rate": 1.202602566419664e-05, "loss": 0.4954, "step": 6630 }, { "epoch": 1.2, "learning_rate": 1.2023314657509488e-05, "loss": 0.7142, "step": 6631 }, { "epoch": 1.2, "learning_rate": 1.2020603650822339e-05, "loss": 1.5383, "step": 6632 }, { "epoch": 1.2, "learning_rate": 1.2017892644135189e-05, "loss": 0.6527, "step": 6633 }, { "epoch": 1.2, "learning_rate": 1.201518163744804e-05, "loss": 0.7238, "step": 6634 }, { "epoch": 1.2, "learning_rate": 1.2012470630760889e-05, "loss": 0.2843, "step": 6635 }, { "epoch": 1.2, "learning_rate": 1.2009759624073739e-05, "loss": 0.4686, "step": 6636 }, { "epoch": 1.2, "learning_rate": 1.200704861738659e-05, "loss": 0.4224, "step": 6637 }, { "epoch": 1.2, "learning_rate": 1.2004337610699441e-05, "loss": 0.6891, "step": 6638 }, { "epoch": 1.2, "learning_rate": 1.200162660401229e-05, "loss": 0.9522, "step": 6639 }, { "epoch": 1.2, "learning_rate": 1.199891559732514e-05, "loss": 0.726, "step": 6640 }, { "epoch": 1.2, "learning_rate": 1.1996204590637992e-05, "loss": 0.8303, "step": 6641 }, { "epoch": 1.2, "learning_rate": 1.1993493583950842e-05, "loss": 1.0954, "step": 6642 }, { "epoch": 1.2, "learning_rate": 1.199078257726369e-05, "loss": 0.9115, "step": 6643 }, { "epoch": 1.2, "learning_rate": 1.198807157057654e-05, "loss": 0.6196, "step": 6644 }, { "epoch": 1.2, "learning_rate": 1.1985360563889392e-05, "loss": 0.8324, "step": 6645 }, { "epoch": 1.2, "learning_rate": 1.1982649557202242e-05, "loss": 0.9563, "step": 6646 }, { "epoch": 1.2, "learning_rate": 1.197993855051509e-05, "loss": 0.7134, "step": 6647 }, { "epoch": 1.2, "learning_rate": 1.1977227543827942e-05, "loss": 0.1742, "step": 6648 }, { "epoch": 1.2, "learning_rate": 1.1974516537140793e-05, "loss": 0.6277, "step": 6649 }, { "epoch": 1.2, "learning_rate": 1.1971805530453643e-05, "loss": 0.9776, "step": 6650 }, { "epoch": 1.2, "learning_rate": 1.1969094523766491e-05, "loss": 0.6442, "step": 6651 }, { "epoch": 1.2, "learning_rate": 1.1966383517079343e-05, "loss": 0.7641, "step": 6652 }, { "epoch": 1.2, "learning_rate": 1.1963672510392193e-05, "loss": 0.8408, "step": 6653 }, { "epoch": 1.2, "learning_rate": 1.1960961503705043e-05, "loss": 0.7479, "step": 6654 }, { "epoch": 1.2, "learning_rate": 1.1958250497017892e-05, "loss": 0.7017, "step": 6655 }, { "epoch": 1.2, "learning_rate": 1.1955539490330744e-05, "loss": 0.605, "step": 6656 }, { "epoch": 1.2, "learning_rate": 1.1952828483643594e-05, "loss": 0.6104, "step": 6657 }, { "epoch": 1.2, "learning_rate": 1.1950117476956444e-05, "loss": 0.8632, "step": 6658 }, { "epoch": 1.2, "learning_rate": 1.1947406470269294e-05, "loss": 0.9794, "step": 6659 }, { "epoch": 1.2, "learning_rate": 1.1944695463582144e-05, "loss": 0.799, "step": 6660 }, { "epoch": 1.2, "learning_rate": 1.1941984456894994e-05, "loss": 0.8647, "step": 6661 }, { "epoch": 1.2, "learning_rate": 1.1939273450207844e-05, "loss": 1.048, "step": 6662 }, { "epoch": 1.2, "learning_rate": 1.1936562443520694e-05, "loss": 0.7605, "step": 6663 }, { "epoch": 1.2, "learning_rate": 1.1933851436833545e-05, "loss": 0.874, "step": 6664 }, { "epoch": 1.2, "learning_rate": 1.1931140430146395e-05, "loss": 0.9914, "step": 6665 }, { "epoch": 1.2, "learning_rate": 1.1928429423459245e-05, "loss": 0.347, "step": 6666 }, { "epoch": 1.2, "learning_rate": 1.1925718416772095e-05, "loss": 0.7608, "step": 6667 }, { "epoch": 1.21, "learning_rate": 1.1923007410084945e-05, "loss": 0.9772, "step": 6668 }, { "epoch": 1.21, "learning_rate": 1.1920296403397795e-05, "loss": 1.116, "step": 6669 }, { "epoch": 1.21, "learning_rate": 1.1917585396710647e-05, "loss": 0.7087, "step": 6670 }, { "epoch": 1.21, "learning_rate": 1.1914874390023496e-05, "loss": 0.3347, "step": 6671 }, { "epoch": 1.21, "learning_rate": 1.1912163383336346e-05, "loss": 0.7175, "step": 6672 }, { "epoch": 1.21, "learning_rate": 1.1909452376649196e-05, "loss": 0.8057, "step": 6673 }, { "epoch": 1.21, "learning_rate": 1.1906741369962048e-05, "loss": 0.892, "step": 6674 }, { "epoch": 1.21, "learning_rate": 1.1904030363274896e-05, "loss": 1.088, "step": 6675 }, { "epoch": 1.21, "learning_rate": 1.1901319356587746e-05, "loss": 1.1839, "step": 6676 }, { "epoch": 1.21, "learning_rate": 1.1898608349900596e-05, "loss": 0.6793, "step": 6677 }, { "epoch": 1.21, "learning_rate": 1.1895897343213448e-05, "loss": 1.0384, "step": 6678 }, { "epoch": 1.21, "learning_rate": 1.1893186336526297e-05, "loss": 0.4471, "step": 6679 }, { "epoch": 1.21, "learning_rate": 1.1890475329839147e-05, "loss": 0.8522, "step": 6680 }, { "epoch": 1.21, "learning_rate": 1.1887764323151999e-05, "loss": 0.7664, "step": 6681 }, { "epoch": 1.21, "learning_rate": 1.1885053316464849e-05, "loss": 0.5078, "step": 6682 }, { "epoch": 1.21, "learning_rate": 1.1882342309777697e-05, "loss": 0.4592, "step": 6683 }, { "epoch": 1.21, "learning_rate": 1.1879631303090547e-05, "loss": 0.6838, "step": 6684 }, { "epoch": 1.21, "learning_rate": 1.1876920296403399e-05, "loss": 0.5781, "step": 6685 }, { "epoch": 1.21, "learning_rate": 1.187420928971625e-05, "loss": 0.565, "step": 6686 }, { "epoch": 1.21, "learning_rate": 1.1871498283029098e-05, "loss": 0.2057, "step": 6687 }, { "epoch": 1.21, "learning_rate": 1.1868787276341948e-05, "loss": 0.9536, "step": 6688 }, { "epoch": 1.21, "learning_rate": 1.18660762696548e-05, "loss": 0.8723, "step": 6689 }, { "epoch": 1.21, "learning_rate": 1.186336526296765e-05, "loss": 1.5269, "step": 6690 }, { "epoch": 1.21, "learning_rate": 1.1860654256280498e-05, "loss": 0.8074, "step": 6691 }, { "epoch": 1.21, "learning_rate": 1.185794324959335e-05, "loss": 1.4261, "step": 6692 }, { "epoch": 1.21, "learning_rate": 1.18552322429062e-05, "loss": 0.3773, "step": 6693 }, { "epoch": 1.21, "learning_rate": 1.1852521236219049e-05, "loss": 0.4411, "step": 6694 }, { "epoch": 1.21, "learning_rate": 1.1849810229531899e-05, "loss": 0.6681, "step": 6695 }, { "epoch": 1.21, "learning_rate": 1.184709922284475e-05, "loss": 0.6053, "step": 6696 }, { "epoch": 1.21, "learning_rate": 1.18443882161576e-05, "loss": 0.4049, "step": 6697 }, { "epoch": 1.21, "learning_rate": 1.184167720947045e-05, "loss": 0.4753, "step": 6698 }, { "epoch": 1.21, "learning_rate": 1.18389662027833e-05, "loss": 0.7444, "step": 6699 }, { "epoch": 1.21, "learning_rate": 1.1836255196096151e-05, "loss": 0.9939, "step": 6700 }, { "epoch": 1.21, "learning_rate": 1.1833544189409001e-05, "loss": 0.7566, "step": 6701 }, { "epoch": 1.21, "learning_rate": 1.183083318272185e-05, "loss": 0.9627, "step": 6702 }, { "epoch": 1.21, "learning_rate": 1.1828122176034702e-05, "loss": 0.471, "step": 6703 }, { "epoch": 1.21, "learning_rate": 1.1825411169347552e-05, "loss": 1.1481, "step": 6704 }, { "epoch": 1.21, "learning_rate": 1.1822700162660402e-05, "loss": 0.8515, "step": 6705 }, { "epoch": 1.21, "learning_rate": 1.181998915597325e-05, "loss": 0.916, "step": 6706 }, { "epoch": 1.21, "learning_rate": 1.1817278149286102e-05, "loss": 0.3349, "step": 6707 }, { "epoch": 1.21, "learning_rate": 1.1814567142598952e-05, "loss": 0.719, "step": 6708 }, { "epoch": 1.21, "learning_rate": 1.1811856135911802e-05, "loss": 0.4969, "step": 6709 }, { "epoch": 1.21, "learning_rate": 1.180914512922465e-05, "loss": 1.8959, "step": 6710 }, { "epoch": 1.21, "learning_rate": 1.1806434122537503e-05, "loss": 0.7535, "step": 6711 }, { "epoch": 1.21, "learning_rate": 1.1803723115850353e-05, "loss": 0.8415, "step": 6712 }, { "epoch": 1.21, "learning_rate": 1.1801012109163203e-05, "loss": 0.316, "step": 6713 }, { "epoch": 1.21, "learning_rate": 1.1798301102476053e-05, "loss": 0.2819, "step": 6714 }, { "epoch": 1.21, "learning_rate": 1.1795590095788903e-05, "loss": 0.4256, "step": 6715 }, { "epoch": 1.21, "learning_rate": 1.1792879089101753e-05, "loss": 0.4168, "step": 6716 }, { "epoch": 1.21, "learning_rate": 1.1790168082414603e-05, "loss": 0.5197, "step": 6717 }, { "epoch": 1.21, "learning_rate": 1.1787457075727454e-05, "loss": 0.9711, "step": 6718 }, { "epoch": 1.21, "learning_rate": 1.1784746069040304e-05, "loss": 0.5571, "step": 6719 }, { "epoch": 1.21, "learning_rate": 1.1782035062353154e-05, "loss": 0.4059, "step": 6720 }, { "epoch": 1.21, "learning_rate": 1.1779324055666004e-05, "loss": 0.561, "step": 6721 }, { "epoch": 1.21, "learning_rate": 1.1776613048978854e-05, "loss": 0.7873, "step": 6722 }, { "epoch": 1.22, "learning_rate": 1.1773902042291704e-05, "loss": 0.888, "step": 6723 }, { "epoch": 1.22, "learning_rate": 1.1771191035604554e-05, "loss": 0.8873, "step": 6724 }, { "epoch": 1.22, "learning_rate": 1.1768480028917406e-05, "loss": 0.7664, "step": 6725 }, { "epoch": 1.22, "learning_rate": 1.1765769022230255e-05, "loss": 0.4909, "step": 6726 }, { "epoch": 1.22, "learning_rate": 1.1763058015543105e-05, "loss": 0.3441, "step": 6727 }, { "epoch": 1.22, "learning_rate": 1.1760347008855955e-05, "loss": 0.8432, "step": 6728 }, { "epoch": 1.22, "learning_rate": 1.1757636002168807e-05, "loss": 0.6074, "step": 6729 }, { "epoch": 1.22, "learning_rate": 1.1754924995481655e-05, "loss": 0.7976, "step": 6730 }, { "epoch": 1.22, "learning_rate": 1.1752213988794505e-05, "loss": 0.4777, "step": 6731 }, { "epoch": 1.22, "learning_rate": 1.1749502982107357e-05, "loss": 0.5164, "step": 6732 }, { "epoch": 1.22, "learning_rate": 1.1746791975420207e-05, "loss": 0.8684, "step": 6733 }, { "epoch": 1.22, "learning_rate": 1.1744080968733056e-05, "loss": 0.4811, "step": 6734 }, { "epoch": 1.22, "learning_rate": 1.1741369962045906e-05, "loss": 0.8935, "step": 6735 }, { "epoch": 1.22, "learning_rate": 1.1738658955358758e-05, "loss": 0.2007, "step": 6736 }, { "epoch": 1.22, "learning_rate": 1.1735947948671608e-05, "loss": 1.2846, "step": 6737 }, { "epoch": 1.22, "learning_rate": 1.1733236941984456e-05, "loss": 0.5166, "step": 6738 }, { "epoch": 1.22, "learning_rate": 1.1730525935297306e-05, "loss": 0.3892, "step": 6739 }, { "epoch": 1.22, "learning_rate": 1.1727814928610158e-05, "loss": 0.6265, "step": 6740 }, { "epoch": 1.22, "learning_rate": 1.1725103921923008e-05, "loss": 0.9424, "step": 6741 }, { "epoch": 1.22, "learning_rate": 1.1722392915235857e-05, "loss": 0.572, "step": 6742 }, { "epoch": 1.22, "learning_rate": 1.1719681908548709e-05, "loss": 0.6178, "step": 6743 }, { "epoch": 1.22, "learning_rate": 1.1716970901861559e-05, "loss": 1.0315, "step": 6744 }, { "epoch": 1.22, "learning_rate": 1.1714259895174409e-05, "loss": 0.8011, "step": 6745 }, { "epoch": 1.22, "learning_rate": 1.1711548888487257e-05, "loss": 0.9047, "step": 6746 }, { "epoch": 1.22, "learning_rate": 1.1708837881800109e-05, "loss": 0.581, "step": 6747 }, { "epoch": 1.22, "learning_rate": 1.170612687511296e-05, "loss": 0.7391, "step": 6748 }, { "epoch": 1.22, "learning_rate": 1.170341586842581e-05, "loss": 0.3121, "step": 6749 }, { "epoch": 1.22, "learning_rate": 1.1700704861738658e-05, "loss": 0.913, "step": 6750 }, { "epoch": 1.22, "learning_rate": 1.169799385505151e-05, "loss": 0.4636, "step": 6751 }, { "epoch": 1.22, "learning_rate": 1.169528284836436e-05, "loss": 0.9814, "step": 6752 }, { "epoch": 1.22, "learning_rate": 1.169257184167721e-05, "loss": 0.4006, "step": 6753 }, { "epoch": 1.22, "learning_rate": 1.168986083499006e-05, "loss": 0.6563, "step": 6754 }, { "epoch": 1.22, "learning_rate": 1.168714982830291e-05, "loss": 0.3115, "step": 6755 }, { "epoch": 1.22, "learning_rate": 1.168443882161576e-05, "loss": 0.6904, "step": 6756 }, { "epoch": 1.22, "learning_rate": 1.168172781492861e-05, "loss": 0.4753, "step": 6757 }, { "epoch": 1.22, "learning_rate": 1.167901680824146e-05, "loss": 1.0419, "step": 6758 }, { "epoch": 1.22, "learning_rate": 1.167630580155431e-05, "loss": 0.453, "step": 6759 }, { "epoch": 1.22, "learning_rate": 1.1673594794867161e-05, "loss": 0.4925, "step": 6760 }, { "epoch": 1.22, "learning_rate": 1.1670883788180011e-05, "loss": 0.6016, "step": 6761 }, { "epoch": 1.22, "learning_rate": 1.1668172781492861e-05, "loss": 0.9623, "step": 6762 }, { "epoch": 1.22, "learning_rate": 1.1665461774805711e-05, "loss": 0.5758, "step": 6763 }, { "epoch": 1.22, "learning_rate": 1.1662750768118561e-05, "loss": 0.6336, "step": 6764 }, { "epoch": 1.22, "learning_rate": 1.1660039761431413e-05, "loss": 0.7075, "step": 6765 }, { "epoch": 1.22, "learning_rate": 1.1657328754744262e-05, "loss": 0.8526, "step": 6766 }, { "epoch": 1.22, "learning_rate": 1.1654617748057112e-05, "loss": 1.1127, "step": 6767 }, { "epoch": 1.22, "learning_rate": 1.1651906741369962e-05, "loss": 0.5839, "step": 6768 }, { "epoch": 1.22, "learning_rate": 1.1649195734682814e-05, "loss": 0.4878, "step": 6769 }, { "epoch": 1.22, "learning_rate": 1.1646484727995662e-05, "loss": 0.6818, "step": 6770 }, { "epoch": 1.22, "learning_rate": 1.1643773721308512e-05, "loss": 0.3835, "step": 6771 }, { "epoch": 1.22, "learning_rate": 1.1641062714621362e-05, "loss": 1.0338, "step": 6772 }, { "epoch": 1.22, "learning_rate": 1.1638351707934214e-05, "loss": 1.1369, "step": 6773 }, { "epoch": 1.22, "learning_rate": 1.1635640701247063e-05, "loss": 1.2145, "step": 6774 }, { "epoch": 1.22, "learning_rate": 1.1632929694559913e-05, "loss": 0.8426, "step": 6775 }, { "epoch": 1.22, "learning_rate": 1.1630218687872765e-05, "loss": 0.4741, "step": 6776 }, { "epoch": 1.22, "learning_rate": 1.1627507681185615e-05, "loss": 0.8344, "step": 6777 }, { "epoch": 1.23, "learning_rate": 1.1624796674498463e-05, "loss": 0.8859, "step": 6778 }, { "epoch": 1.23, "learning_rate": 1.1622085667811313e-05, "loss": 1.0766, "step": 6779 }, { "epoch": 1.23, "learning_rate": 1.1619374661124165e-05, "loss": 0.4363, "step": 6780 }, { "epoch": 1.23, "learning_rate": 1.1616663654437015e-05, "loss": 0.9984, "step": 6781 }, { "epoch": 1.23, "learning_rate": 1.1613952647749864e-05, "loss": 0.3309, "step": 6782 }, { "epoch": 1.23, "learning_rate": 1.1611241641062714e-05, "loss": 0.9075, "step": 6783 }, { "epoch": 1.23, "learning_rate": 1.1608530634375566e-05, "loss": 1.0762, "step": 6784 }, { "epoch": 1.23, "learning_rate": 1.1605819627688416e-05, "loss": 0.6518, "step": 6785 }, { "epoch": 1.23, "learning_rate": 1.1603108621001264e-05, "loss": 0.8786, "step": 6786 }, { "epoch": 1.23, "learning_rate": 1.1600397614314116e-05, "loss": 1.5716, "step": 6787 }, { "epoch": 1.23, "learning_rate": 1.1597686607626966e-05, "loss": 0.6714, "step": 6788 }, { "epoch": 1.23, "learning_rate": 1.1594975600939816e-05, "loss": 0.8334, "step": 6789 }, { "epoch": 1.23, "learning_rate": 1.1592264594252665e-05, "loss": 0.6749, "step": 6790 }, { "epoch": 1.23, "learning_rate": 1.1589553587565517e-05, "loss": 0.3843, "step": 6791 }, { "epoch": 1.23, "learning_rate": 1.1586842580878367e-05, "loss": 1.4001, "step": 6792 }, { "epoch": 1.23, "learning_rate": 1.1584131574191217e-05, "loss": 0.694, "step": 6793 }, { "epoch": 1.23, "learning_rate": 1.1581420567504067e-05, "loss": 0.929, "step": 6794 }, { "epoch": 1.23, "learning_rate": 1.1578709560816917e-05, "loss": 1.4753, "step": 6795 }, { "epoch": 1.23, "learning_rate": 1.1575998554129767e-05, "loss": 0.6092, "step": 6796 }, { "epoch": 1.23, "learning_rate": 1.1573287547442618e-05, "loss": 0.2423, "step": 6797 }, { "epoch": 1.23, "learning_rate": 1.1570576540755468e-05, "loss": 0.3878, "step": 6798 }, { "epoch": 1.23, "learning_rate": 1.1567865534068318e-05, "loss": 0.6807, "step": 6799 }, { "epoch": 1.23, "learning_rate": 1.1565154527381168e-05, "loss": 0.9788, "step": 6800 }, { "epoch": 1.23, "learning_rate": 1.1562443520694018e-05, "loss": 0.7937, "step": 6801 }, { "epoch": 1.23, "learning_rate": 1.1559732514006868e-05, "loss": 0.6321, "step": 6802 }, { "epoch": 1.23, "learning_rate": 1.1557021507319718e-05, "loss": 0.5548, "step": 6803 }, { "epoch": 1.23, "learning_rate": 1.1554310500632568e-05, "loss": 0.9091, "step": 6804 }, { "epoch": 1.23, "learning_rate": 1.155159949394542e-05, "loss": 1.6249, "step": 6805 }, { "epoch": 1.23, "learning_rate": 1.1548888487258269e-05, "loss": 0.5825, "step": 6806 }, { "epoch": 1.23, "learning_rate": 1.1546177480571119e-05, "loss": 0.6571, "step": 6807 }, { "epoch": 1.23, "learning_rate": 1.1543466473883969e-05, "loss": 1.3844, "step": 6808 }, { "epoch": 1.23, "learning_rate": 1.154075546719682e-05, "loss": 0.7596, "step": 6809 }, { "epoch": 1.23, "learning_rate": 1.153804446050967e-05, "loss": 1.3036, "step": 6810 }, { "epoch": 1.23, "learning_rate": 1.153533345382252e-05, "loss": 0.4896, "step": 6811 }, { "epoch": 1.23, "learning_rate": 1.153262244713537e-05, "loss": 0.4819, "step": 6812 }, { "epoch": 1.23, "learning_rate": 1.1529911440448221e-05, "loss": 0.9002, "step": 6813 }, { "epoch": 1.23, "learning_rate": 1.152720043376107e-05, "loss": 1.1818, "step": 6814 }, { "epoch": 1.23, "learning_rate": 1.152448942707392e-05, "loss": 0.9201, "step": 6815 }, { "epoch": 1.23, "learning_rate": 1.1521778420386772e-05, "loss": 0.5027, "step": 6816 }, { "epoch": 1.23, "learning_rate": 1.1519067413699622e-05, "loss": 0.6419, "step": 6817 }, { "epoch": 1.23, "learning_rate": 1.151635640701247e-05, "loss": 0.901, "step": 6818 }, { "epoch": 1.23, "learning_rate": 1.151364540032532e-05, "loss": 0.8599, "step": 6819 }, { "epoch": 1.23, "learning_rate": 1.1510934393638172e-05, "loss": 0.8811, "step": 6820 }, { "epoch": 1.23, "learning_rate": 1.150822338695102e-05, "loss": 1.1609, "step": 6821 }, { "epoch": 1.23, "learning_rate": 1.1505512380263871e-05, "loss": 0.9348, "step": 6822 }, { "epoch": 1.23, "learning_rate": 1.1502801373576721e-05, "loss": 0.7219, "step": 6823 }, { "epoch": 1.23, "learning_rate": 1.1500090366889573e-05, "loss": 0.3646, "step": 6824 }, { "epoch": 1.23, "learning_rate": 1.1497379360202421e-05, "loss": 0.7937, "step": 6825 }, { "epoch": 1.23, "learning_rate": 1.1494668353515271e-05, "loss": 0.988, "step": 6826 }, { "epoch": 1.23, "learning_rate": 1.1491957346828123e-05, "loss": 0.5307, "step": 6827 }, { "epoch": 1.23, "learning_rate": 1.1489246340140973e-05, "loss": 0.4513, "step": 6828 }, { "epoch": 1.23, "learning_rate": 1.1486535333453822e-05, "loss": 0.2617, "step": 6829 }, { "epoch": 1.23, "learning_rate": 1.1483824326766672e-05, "loss": 0.7294, "step": 6830 }, { "epoch": 1.23, "learning_rate": 1.1481113320079524e-05, "loss": 0.9037, "step": 6831 }, { "epoch": 1.23, "learning_rate": 1.1478402313392374e-05, "loss": 0.4138, "step": 6832 }, { "epoch": 1.23, "learning_rate": 1.1475691306705222e-05, "loss": 1.1285, "step": 6833 }, { "epoch": 1.24, "learning_rate": 1.1472980300018073e-05, "loss": 0.5668, "step": 6834 }, { "epoch": 1.24, "learning_rate": 1.1470269293330924e-05, "loss": 1.2138, "step": 6835 }, { "epoch": 1.24, "learning_rate": 1.1467558286643774e-05, "loss": 1.5432, "step": 6836 }, { "epoch": 1.24, "learning_rate": 1.1464847279956623e-05, "loss": 0.8182, "step": 6837 }, { "epoch": 1.24, "learning_rate": 1.1462136273269475e-05, "loss": 1.3202, "step": 6838 }, { "epoch": 1.24, "learning_rate": 1.1459425266582325e-05, "loss": 0.2133, "step": 6839 }, { "epoch": 1.24, "learning_rate": 1.1456714259895175e-05, "loss": 0.8568, "step": 6840 }, { "epoch": 1.24, "learning_rate": 1.1454003253208023e-05, "loss": 0.6373, "step": 6841 }, { "epoch": 1.24, "learning_rate": 1.1451292246520875e-05, "loss": 1.0704, "step": 6842 }, { "epoch": 1.24, "learning_rate": 1.1448581239833725e-05, "loss": 0.7791, "step": 6843 }, { "epoch": 1.24, "learning_rate": 1.1445870233146576e-05, "loss": 0.299, "step": 6844 }, { "epoch": 1.24, "learning_rate": 1.1443159226459424e-05, "loss": 0.9682, "step": 6845 }, { "epoch": 1.24, "learning_rate": 1.1440448219772276e-05, "loss": 0.85, "step": 6846 }, { "epoch": 1.24, "learning_rate": 1.1437737213085126e-05, "loss": 0.558, "step": 6847 }, { "epoch": 1.24, "learning_rate": 1.1435026206397976e-05, "loss": 0.739, "step": 6848 }, { "epoch": 1.24, "learning_rate": 1.1432315199710826e-05, "loss": 0.6641, "step": 6849 }, { "epoch": 1.24, "learning_rate": 1.1429604193023676e-05, "loss": 0.9615, "step": 6850 }, { "epoch": 1.24, "learning_rate": 1.1426893186336526e-05, "loss": 0.923, "step": 6851 }, { "epoch": 1.24, "learning_rate": 1.1424182179649377e-05, "loss": 1.1064, "step": 6852 }, { "epoch": 1.24, "learning_rate": 1.1421471172962227e-05, "loss": 0.7199, "step": 6853 }, { "epoch": 1.24, "learning_rate": 1.1418760166275077e-05, "loss": 1.3492, "step": 6854 }, { "epoch": 1.24, "learning_rate": 1.1416049159587927e-05, "loss": 0.6984, "step": 6855 }, { "epoch": 1.24, "learning_rate": 1.1413338152900777e-05, "loss": 0.5894, "step": 6856 }, { "epoch": 1.24, "learning_rate": 1.1410627146213627e-05, "loss": 0.8874, "step": 6857 }, { "epoch": 1.24, "learning_rate": 1.1407916139526477e-05, "loss": 1.0943, "step": 6858 }, { "epoch": 1.24, "learning_rate": 1.1405205132839328e-05, "loss": 0.5359, "step": 6859 }, { "epoch": 1.24, "learning_rate": 1.140249412615218e-05, "loss": 0.4131, "step": 6860 }, { "epoch": 1.24, "learning_rate": 1.1399783119465028e-05, "loss": 1.1564, "step": 6861 }, { "epoch": 1.24, "learning_rate": 1.1397072112777878e-05, "loss": 0.9742, "step": 6862 }, { "epoch": 1.24, "learning_rate": 1.1394361106090728e-05, "loss": 0.6532, "step": 6863 }, { "epoch": 1.24, "learning_rate": 1.139165009940358e-05, "loss": 1.0793, "step": 6864 }, { "epoch": 1.24, "learning_rate": 1.1388939092716428e-05, "loss": 1.0352, "step": 6865 }, { "epoch": 1.24, "learning_rate": 1.1386228086029278e-05, "loss": 1.1557, "step": 6866 }, { "epoch": 1.24, "learning_rate": 1.138351707934213e-05, "loss": 0.4315, "step": 6867 }, { "epoch": 1.24, "learning_rate": 1.138080607265498e-05, "loss": 0.962, "step": 6868 }, { "epoch": 1.24, "learning_rate": 1.1378095065967829e-05, "loss": 0.7513, "step": 6869 }, { "epoch": 1.24, "learning_rate": 1.1375384059280679e-05, "loss": 0.8683, "step": 6870 }, { "epoch": 1.24, "learning_rate": 1.1372673052593531e-05, "loss": 0.442, "step": 6871 }, { "epoch": 1.24, "learning_rate": 1.1369962045906381e-05, "loss": 1.2248, "step": 6872 }, { "epoch": 1.24, "learning_rate": 1.136725103921923e-05, "loss": 0.9796, "step": 6873 }, { "epoch": 1.24, "learning_rate": 1.136454003253208e-05, "loss": 0.6348, "step": 6874 }, { "epoch": 1.24, "learning_rate": 1.1361829025844931e-05, "loss": 0.8473, "step": 6875 }, { "epoch": 1.24, "learning_rate": 1.1359118019157782e-05, "loss": 0.5202, "step": 6876 }, { "epoch": 1.24, "learning_rate": 1.135640701247063e-05, "loss": 0.5747, "step": 6877 }, { "epoch": 1.24, "learning_rate": 1.1353696005783482e-05, "loss": 0.7409, "step": 6878 }, { "epoch": 1.24, "learning_rate": 1.1350984999096332e-05, "loss": 0.9418, "step": 6879 }, { "epoch": 1.24, "learning_rate": 1.1348273992409182e-05, "loss": 0.6307, "step": 6880 }, { "epoch": 1.24, "learning_rate": 1.134556298572203e-05, "loss": 0.3327, "step": 6881 }, { "epoch": 1.24, "learning_rate": 1.1342851979034882e-05, "loss": 0.4767, "step": 6882 }, { "epoch": 1.24, "learning_rate": 1.1340140972347732e-05, "loss": 1.0905, "step": 6883 }, { "epoch": 1.24, "learning_rate": 1.1337429965660583e-05, "loss": 0.8052, "step": 6884 }, { "epoch": 1.24, "learning_rate": 1.1334718958973431e-05, "loss": 0.5208, "step": 6885 }, { "epoch": 1.24, "learning_rate": 1.1332007952286283e-05, "loss": 1.1702, "step": 6886 }, { "epoch": 1.24, "learning_rate": 1.1329296945599133e-05, "loss": 0.5225, "step": 6887 }, { "epoch": 1.24, "learning_rate": 1.1326585938911983e-05, "loss": 1.0993, "step": 6888 }, { "epoch": 1.25, "learning_rate": 1.1323874932224833e-05, "loss": 0.41, "step": 6889 }, { "epoch": 1.25, "learning_rate": 1.1321163925537683e-05, "loss": 0.632, "step": 6890 }, { "epoch": 1.25, "learning_rate": 1.1318452918850534e-05, "loss": 0.5734, "step": 6891 }, { "epoch": 1.25, "learning_rate": 1.1315741912163384e-05, "loss": 0.7813, "step": 6892 }, { "epoch": 1.25, "learning_rate": 1.1313030905476234e-05, "loss": 0.7105, "step": 6893 }, { "epoch": 1.25, "learning_rate": 1.1310319898789084e-05, "loss": 0.9137, "step": 6894 }, { "epoch": 1.25, "learning_rate": 1.1307608892101934e-05, "loss": 1.3251, "step": 6895 }, { "epoch": 1.25, "learning_rate": 1.1304897885414784e-05, "loss": 0.3388, "step": 6896 }, { "epoch": 1.25, "learning_rate": 1.1302186878727634e-05, "loss": 0.3183, "step": 6897 }, { "epoch": 1.25, "learning_rate": 1.1299475872040484e-05, "loss": 0.7065, "step": 6898 }, { "epoch": 1.25, "learning_rate": 1.1296764865353335e-05, "loss": 0.4162, "step": 6899 }, { "epoch": 1.25, "learning_rate": 1.1294053858666186e-05, "loss": 0.6559, "step": 6900 }, { "epoch": 1.25, "learning_rate": 1.1291342851979035e-05, "loss": 0.5197, "step": 6901 }, { "epoch": 1.25, "learning_rate": 1.1288631845291885e-05, "loss": 0.8723, "step": 6902 }, { "epoch": 1.25, "learning_rate": 1.1285920838604735e-05, "loss": 0.2543, "step": 6903 }, { "epoch": 1.25, "learning_rate": 1.1283209831917587e-05, "loss": 0.2812, "step": 6904 }, { "epoch": 1.25, "learning_rate": 1.1280498825230435e-05, "loss": 0.5815, "step": 6905 }, { "epoch": 1.25, "learning_rate": 1.1277787818543286e-05, "loss": 0.9922, "step": 6906 }, { "epoch": 1.25, "learning_rate": 1.1275076811856136e-05, "loss": 0.6065, "step": 6907 }, { "epoch": 1.25, "learning_rate": 1.1272365805168988e-05, "loss": 0.8615, "step": 6908 }, { "epoch": 1.25, "learning_rate": 1.1269654798481836e-05, "loss": 2.3779, "step": 6909 }, { "epoch": 1.25, "learning_rate": 1.1266943791794686e-05, "loss": 1.1846, "step": 6910 }, { "epoch": 1.25, "learning_rate": 1.1264232785107538e-05, "loss": 1.0128, "step": 6911 }, { "epoch": 1.25, "learning_rate": 1.1261521778420388e-05, "loss": 1.4916, "step": 6912 }, { "epoch": 1.25, "learning_rate": 1.1258810771733237e-05, "loss": 0.8098, "step": 6913 }, { "epoch": 1.25, "learning_rate": 1.1256099765046087e-05, "loss": 1.6556, "step": 6914 }, { "epoch": 1.25, "learning_rate": 1.1253388758358938e-05, "loss": 0.4126, "step": 6915 }, { "epoch": 1.25, "learning_rate": 1.1250677751671789e-05, "loss": 0.3735, "step": 6916 }, { "epoch": 1.25, "learning_rate": 1.1247966744984637e-05, "loss": 0.8498, "step": 6917 }, { "epoch": 1.25, "learning_rate": 1.1245255738297487e-05, "loss": 0.9097, "step": 6918 }, { "epoch": 1.25, "learning_rate": 1.1242544731610339e-05, "loss": 0.5394, "step": 6919 }, { "epoch": 1.25, "learning_rate": 1.1239833724923189e-05, "loss": 1.3781, "step": 6920 }, { "epoch": 1.25, "learning_rate": 1.1237122718236038e-05, "loss": 0.6467, "step": 6921 }, { "epoch": 1.25, "learning_rate": 1.123441171154889e-05, "loss": 1.0262, "step": 6922 }, { "epoch": 1.25, "learning_rate": 1.123170070486174e-05, "loss": 0.8958, "step": 6923 }, { "epoch": 1.25, "learning_rate": 1.122898969817459e-05, "loss": 0.5348, "step": 6924 }, { "epoch": 1.25, "learning_rate": 1.1226278691487438e-05, "loss": 0.4264, "step": 6925 }, { "epoch": 1.25, "learning_rate": 1.122356768480029e-05, "loss": 0.7974, "step": 6926 }, { "epoch": 1.25, "learning_rate": 1.122085667811314e-05, "loss": 1.2929, "step": 6927 }, { "epoch": 1.25, "learning_rate": 1.121814567142599e-05, "loss": 1.274, "step": 6928 }, { "epoch": 1.25, "learning_rate": 1.1215434664738839e-05, "loss": 0.4915, "step": 6929 }, { "epoch": 1.25, "learning_rate": 1.121272365805169e-05, "loss": 1.0302, "step": 6930 }, { "epoch": 1.25, "learning_rate": 1.121001265136454e-05, "loss": 0.64, "step": 6931 }, { "epoch": 1.25, "learning_rate": 1.120730164467739e-05, "loss": 0.6552, "step": 6932 }, { "epoch": 1.25, "learning_rate": 1.1204590637990241e-05, "loss": 1.0045, "step": 6933 }, { "epoch": 1.25, "learning_rate": 1.1201879631303091e-05, "loss": 0.4965, "step": 6934 }, { "epoch": 1.25, "learning_rate": 1.1199168624615941e-05, "loss": 0.5878, "step": 6935 }, { "epoch": 1.25, "learning_rate": 1.1196457617928791e-05, "loss": 1.0789, "step": 6936 }, { "epoch": 1.25, "learning_rate": 1.1193746611241641e-05, "loss": 0.4876, "step": 6937 }, { "epoch": 1.25, "learning_rate": 1.1191035604554492e-05, "loss": 1.4158, "step": 6938 }, { "epoch": 1.25, "learning_rate": 1.1188324597867342e-05, "loss": 1.0386, "step": 6939 }, { "epoch": 1.25, "learning_rate": 1.1185613591180192e-05, "loss": 0.3571, "step": 6940 }, { "epoch": 1.25, "learning_rate": 1.1182902584493042e-05, "loss": 1.2793, "step": 6941 }, { "epoch": 1.25, "learning_rate": 1.1180191577805892e-05, "loss": 0.6268, "step": 6942 }, { "epoch": 1.25, "learning_rate": 1.1177480571118742e-05, "loss": 1.1256, "step": 6943 }, { "epoch": 1.26, "learning_rate": 1.1174769564431594e-05, "loss": 0.5232, "step": 6944 }, { "epoch": 1.26, "learning_rate": 1.1172058557744442e-05, "loss": 0.7091, "step": 6945 }, { "epoch": 1.26, "learning_rate": 1.1169347551057293e-05, "loss": 0.842, "step": 6946 }, { "epoch": 1.26, "learning_rate": 1.1166636544370143e-05, "loss": 0.5706, "step": 6947 }, { "epoch": 1.26, "learning_rate": 1.1163925537682995e-05, "loss": 0.4592, "step": 6948 }, { "epoch": 1.26, "learning_rate": 1.1161214530995843e-05, "loss": 0.4989, "step": 6949 }, { "epoch": 1.26, "learning_rate": 1.1158503524308693e-05, "loss": 0.4811, "step": 6950 }, { "epoch": 1.26, "learning_rate": 1.1155792517621545e-05, "loss": 0.5002, "step": 6951 }, { "epoch": 1.26, "learning_rate": 1.1153081510934393e-05, "loss": 0.7966, "step": 6952 }, { "epoch": 1.26, "learning_rate": 1.1150370504247244e-05, "loss": 0.9187, "step": 6953 }, { "epoch": 1.26, "learning_rate": 1.1147659497560094e-05, "loss": 0.5668, "step": 6954 }, { "epoch": 1.26, "learning_rate": 1.1144948490872946e-05, "loss": 1.1276, "step": 6955 }, { "epoch": 1.26, "learning_rate": 1.1142237484185794e-05, "loss": 0.6042, "step": 6956 }, { "epoch": 1.26, "learning_rate": 1.1139526477498644e-05, "loss": 0.5531, "step": 6957 }, { "epoch": 1.26, "learning_rate": 1.1136815470811494e-05, "loss": 0.4058, "step": 6958 }, { "epoch": 1.26, "learning_rate": 1.1134104464124346e-05, "loss": 0.5792, "step": 6959 }, { "epoch": 1.26, "learning_rate": 1.1131393457437195e-05, "loss": 0.3526, "step": 6960 }, { "epoch": 1.26, "learning_rate": 1.1128682450750045e-05, "loss": 1.2807, "step": 6961 }, { "epoch": 1.26, "learning_rate": 1.1125971444062896e-05, "loss": 0.6256, "step": 6962 }, { "epoch": 1.26, "learning_rate": 1.1123260437375747e-05, "loss": 1.0049, "step": 6963 }, { "epoch": 1.26, "learning_rate": 1.1120549430688595e-05, "loss": 0.6497, "step": 6964 }, { "epoch": 1.26, "learning_rate": 1.1117838424001445e-05, "loss": 0.2009, "step": 6965 }, { "epoch": 1.26, "learning_rate": 1.1115127417314297e-05, "loss": 0.8671, "step": 6966 }, { "epoch": 1.26, "learning_rate": 1.1112416410627147e-05, "loss": 0.7391, "step": 6967 }, { "epoch": 1.26, "learning_rate": 1.1109705403939996e-05, "loss": 0.7062, "step": 6968 }, { "epoch": 1.26, "learning_rate": 1.1106994397252846e-05, "loss": 0.4605, "step": 6969 }, { "epoch": 1.26, "learning_rate": 1.1104283390565698e-05, "loss": 0.9617, "step": 6970 }, { "epoch": 1.26, "learning_rate": 1.1101572383878548e-05, "loss": 0.3355, "step": 6971 }, { "epoch": 1.26, "learning_rate": 1.1098861377191396e-05, "loss": 1.045, "step": 6972 }, { "epoch": 1.26, "learning_rate": 1.1096150370504248e-05, "loss": 0.5992, "step": 6973 }, { "epoch": 1.26, "learning_rate": 1.1093439363817098e-05, "loss": 0.9269, "step": 6974 }, { "epoch": 1.26, "learning_rate": 1.1090728357129948e-05, "loss": 0.654, "step": 6975 }, { "epoch": 1.26, "learning_rate": 1.1088017350442797e-05, "loss": 0.9317, "step": 6976 }, { "epoch": 1.26, "learning_rate": 1.1085306343755648e-05, "loss": 0.4813, "step": 6977 }, { "epoch": 1.26, "learning_rate": 1.1082595337068499e-05, "loss": 0.6321, "step": 6978 }, { "epoch": 1.26, "learning_rate": 1.1079884330381349e-05, "loss": 0.3644, "step": 6979 }, { "epoch": 1.26, "learning_rate": 1.1077173323694197e-05, "loss": 0.9885, "step": 6980 }, { "epoch": 1.26, "learning_rate": 1.1074462317007049e-05, "loss": 1.282, "step": 6981 }, { "epoch": 1.26, "learning_rate": 1.10717513103199e-05, "loss": 0.9609, "step": 6982 }, { "epoch": 1.26, "learning_rate": 1.106904030363275e-05, "loss": 0.8972, "step": 6983 }, { "epoch": 1.26, "learning_rate": 1.10663292969456e-05, "loss": 0.6032, "step": 6984 }, { "epoch": 1.26, "learning_rate": 1.106361829025845e-05, "loss": 0.636, "step": 6985 }, { "epoch": 1.26, "learning_rate": 1.10609072835713e-05, "loss": 0.6255, "step": 6986 }, { "epoch": 1.26, "learning_rate": 1.105819627688415e-05, "loss": 0.4946, "step": 6987 }, { "epoch": 1.26, "learning_rate": 1.1055485270197e-05, "loss": 0.2652, "step": 6988 }, { "epoch": 1.26, "learning_rate": 1.105277426350985e-05, "loss": 0.6025, "step": 6989 }, { "epoch": 1.26, "learning_rate": 1.10500632568227e-05, "loss": 0.7848, "step": 6990 }, { "epoch": 1.26, "learning_rate": 1.104735225013555e-05, "loss": 0.6187, "step": 6991 }, { "epoch": 1.26, "learning_rate": 1.10446412434484e-05, "loss": 0.9246, "step": 6992 }, { "epoch": 1.26, "learning_rate": 1.104193023676125e-05, "loss": 0.4462, "step": 6993 }, { "epoch": 1.26, "learning_rate": 1.10392192300741e-05, "loss": 0.3091, "step": 6994 }, { "epoch": 1.26, "learning_rate": 1.1036508223386953e-05, "loss": 0.5771, "step": 6995 }, { "epoch": 1.26, "learning_rate": 1.1033797216699801e-05, "loss": 0.6854, "step": 6996 }, { "epoch": 1.26, "learning_rate": 1.1031086210012651e-05, "loss": 1.0573, "step": 6997 }, { "epoch": 1.26, "learning_rate": 1.1028375203325501e-05, "loss": 0.6211, "step": 6998 }, { "epoch": 1.26, "learning_rate": 1.1025664196638353e-05, "loss": 0.7094, "step": 6999 }, { "epoch": 1.27, "learning_rate": 1.1022953189951202e-05, "loss": 0.5543, "step": 7000 }, { "epoch": 1.27, "learning_rate": 1.1020242183264052e-05, "loss": 0.8649, "step": 7001 }, { "epoch": 1.27, "learning_rate": 1.1017531176576902e-05, "loss": 0.5473, "step": 7002 }, { "epoch": 1.27, "learning_rate": 1.1014820169889754e-05, "loss": 0.6957, "step": 7003 }, { "epoch": 1.27, "learning_rate": 1.1012109163202602e-05, "loss": 0.5622, "step": 7004 }, { "epoch": 1.27, "learning_rate": 1.1009398156515452e-05, "loss": 1.0004, "step": 7005 }, { "epoch": 1.27, "learning_rate": 1.1006687149828304e-05, "loss": 0.9419, "step": 7006 }, { "epoch": 1.27, "learning_rate": 1.1003976143141154e-05, "loss": 0.8258, "step": 7007 }, { "epoch": 1.27, "learning_rate": 1.1001265136454003e-05, "loss": 0.7511, "step": 7008 }, { "epoch": 1.27, "learning_rate": 1.0998554129766853e-05, "loss": 0.5556, "step": 7009 }, { "epoch": 1.27, "learning_rate": 1.0995843123079705e-05, "loss": 0.6924, "step": 7010 }, { "epoch": 1.27, "learning_rate": 1.0993132116392555e-05, "loss": 1.0198, "step": 7011 }, { "epoch": 1.27, "learning_rate": 1.0990421109705403e-05, "loss": 1.2882, "step": 7012 }, { "epoch": 1.27, "learning_rate": 1.0987710103018255e-05, "loss": 0.6564, "step": 7013 }, { "epoch": 1.27, "learning_rate": 1.0984999096331105e-05, "loss": 0.7301, "step": 7014 }, { "epoch": 1.27, "learning_rate": 1.0982288089643955e-05, "loss": 0.5352, "step": 7015 }, { "epoch": 1.27, "learning_rate": 1.0979577082956804e-05, "loss": 1.3387, "step": 7016 }, { "epoch": 1.27, "learning_rate": 1.0976866076269656e-05, "loss": 0.6634, "step": 7017 }, { "epoch": 1.27, "learning_rate": 1.0974155069582506e-05, "loss": 0.7781, "step": 7018 }, { "epoch": 1.27, "learning_rate": 1.0971444062895356e-05, "loss": 0.4235, "step": 7019 }, { "epoch": 1.27, "learning_rate": 1.0968733056208204e-05, "loss": 0.4303, "step": 7020 }, { "epoch": 1.27, "learning_rate": 1.0966022049521056e-05, "loss": 0.5759, "step": 7021 }, { "epoch": 1.27, "learning_rate": 1.0963311042833906e-05, "loss": 1.3737, "step": 7022 }, { "epoch": 1.27, "learning_rate": 1.0960600036146756e-05, "loss": 1.1332, "step": 7023 }, { "epoch": 1.27, "learning_rate": 1.0957889029459606e-05, "loss": 0.7133, "step": 7024 }, { "epoch": 1.27, "learning_rate": 1.0955178022772457e-05, "loss": 0.5309, "step": 7025 }, { "epoch": 1.27, "learning_rate": 1.0952467016085307e-05, "loss": 1.0056, "step": 7026 }, { "epoch": 1.27, "learning_rate": 1.0949756009398157e-05, "loss": 0.4239, "step": 7027 }, { "epoch": 1.27, "learning_rate": 1.0947045002711007e-05, "loss": 0.6574, "step": 7028 }, { "epoch": 1.27, "learning_rate": 1.0944333996023857e-05, "loss": 0.6253, "step": 7029 }, { "epoch": 1.27, "learning_rate": 1.0941622989336707e-05, "loss": 0.8318, "step": 7030 }, { "epoch": 1.27, "learning_rate": 1.0938911982649557e-05, "loss": 1.311, "step": 7031 }, { "epoch": 1.27, "learning_rate": 1.0936200975962408e-05, "loss": 0.9185, "step": 7032 }, { "epoch": 1.27, "learning_rate": 1.0933489969275258e-05, "loss": 1.021, "step": 7033 }, { "epoch": 1.27, "learning_rate": 1.0930778962588108e-05, "loss": 0.2973, "step": 7034 }, { "epoch": 1.27, "learning_rate": 1.092806795590096e-05, "loss": 0.5277, "step": 7035 }, { "epoch": 1.27, "learning_rate": 1.0925356949213808e-05, "loss": 0.9622, "step": 7036 }, { "epoch": 1.27, "learning_rate": 1.0922645942526658e-05, "loss": 1.233, "step": 7037 }, { "epoch": 1.27, "learning_rate": 1.0919934935839508e-05, "loss": 0.3025, "step": 7038 }, { "epoch": 1.27, "learning_rate": 1.091722392915236e-05, "loss": 1.1745, "step": 7039 }, { "epoch": 1.27, "learning_rate": 1.0914512922465209e-05, "loss": 1.3969, "step": 7040 }, { "epoch": 1.27, "learning_rate": 1.0911801915778059e-05, "loss": 0.5934, "step": 7041 }, { "epoch": 1.27, "learning_rate": 1.0909090909090909e-05, "loss": 0.9006, "step": 7042 }, { "epoch": 1.27, "learning_rate": 1.090637990240376e-05, "loss": 0.9446, "step": 7043 }, { "epoch": 1.27, "learning_rate": 1.090366889571661e-05, "loss": 1.112, "step": 7044 }, { "epoch": 1.27, "learning_rate": 1.090095788902946e-05, "loss": 0.9792, "step": 7045 }, { "epoch": 1.27, "learning_rate": 1.0898246882342311e-05, "loss": 0.7942, "step": 7046 }, { "epoch": 1.27, "learning_rate": 1.0895535875655161e-05, "loss": 0.8267, "step": 7047 }, { "epoch": 1.27, "learning_rate": 1.089282486896801e-05, "loss": 0.6433, "step": 7048 }, { "epoch": 1.27, "learning_rate": 1.089011386228086e-05, "loss": 0.634, "step": 7049 }, { "epoch": 1.27, "learning_rate": 1.0887402855593712e-05, "loss": 1.6257, "step": 7050 }, { "epoch": 1.27, "learning_rate": 1.0884691848906562e-05, "loss": 0.1894, "step": 7051 }, { "epoch": 1.27, "learning_rate": 1.088198084221941e-05, "loss": 0.386, "step": 7052 }, { "epoch": 1.27, "learning_rate": 1.087926983553226e-05, "loss": 0.5396, "step": 7053 }, { "epoch": 1.27, "learning_rate": 1.0876558828845112e-05, "loss": 0.8079, "step": 7054 }, { "epoch": 1.28, "learning_rate": 1.0873847822157962e-05, "loss": 0.5782, "step": 7055 }, { "epoch": 1.28, "learning_rate": 1.087113681547081e-05, "loss": 0.5505, "step": 7056 }, { "epoch": 1.28, "learning_rate": 1.0868425808783663e-05, "loss": 0.6073, "step": 7057 }, { "epoch": 1.28, "learning_rate": 1.0865714802096513e-05, "loss": 1.1413, "step": 7058 }, { "epoch": 1.28, "learning_rate": 1.0863003795409363e-05, "loss": 0.6535, "step": 7059 }, { "epoch": 1.28, "learning_rate": 1.0860292788722211e-05, "loss": 0.6832, "step": 7060 }, { "epoch": 1.28, "learning_rate": 1.0857581782035063e-05, "loss": 1.1889, "step": 7061 }, { "epoch": 1.28, "learning_rate": 1.0854870775347913e-05, "loss": 0.9343, "step": 7062 }, { "epoch": 1.28, "learning_rate": 1.0852159768660763e-05, "loss": 1.3548, "step": 7063 }, { "epoch": 1.28, "learning_rate": 1.0849448761973612e-05, "loss": 0.4971, "step": 7064 }, { "epoch": 1.28, "learning_rate": 1.0846737755286464e-05, "loss": 0.3967, "step": 7065 }, { "epoch": 1.28, "learning_rate": 1.0844026748599314e-05, "loss": 0.936, "step": 7066 }, { "epoch": 1.28, "learning_rate": 1.0841315741912164e-05, "loss": 0.6097, "step": 7067 }, { "epoch": 1.28, "learning_rate": 1.0838604735225014e-05, "loss": 0.6898, "step": 7068 }, { "epoch": 1.28, "learning_rate": 1.0835893728537864e-05, "loss": 0.8378, "step": 7069 }, { "epoch": 1.28, "learning_rate": 1.0833182721850714e-05, "loss": 1.0784, "step": 7070 }, { "epoch": 1.28, "learning_rate": 1.0830471715163564e-05, "loss": 0.7271, "step": 7071 }, { "epoch": 1.28, "learning_rate": 1.0827760708476415e-05, "loss": 0.9828, "step": 7072 }, { "epoch": 1.28, "learning_rate": 1.0825049701789265e-05, "loss": 1.2882, "step": 7073 }, { "epoch": 1.28, "learning_rate": 1.0822338695102115e-05, "loss": 0.6431, "step": 7074 }, { "epoch": 1.28, "learning_rate": 1.0819627688414965e-05, "loss": 0.2794, "step": 7075 }, { "epoch": 1.28, "learning_rate": 1.0816916681727815e-05, "loss": 0.6779, "step": 7076 }, { "epoch": 1.28, "learning_rate": 1.0814205675040665e-05, "loss": 0.7869, "step": 7077 }, { "epoch": 1.28, "learning_rate": 1.0811494668353515e-05, "loss": 0.7716, "step": 7078 }, { "epoch": 1.28, "learning_rate": 1.0808783661666366e-05, "loss": 0.738, "step": 7079 }, { "epoch": 1.28, "learning_rate": 1.0806072654979216e-05, "loss": 1.0807, "step": 7080 }, { "epoch": 1.28, "learning_rate": 1.0803361648292066e-05, "loss": 0.7477, "step": 7081 }, { "epoch": 1.28, "learning_rate": 1.0800650641604916e-05, "loss": 0.3269, "step": 7082 }, { "epoch": 1.28, "learning_rate": 1.0797939634917766e-05, "loss": 0.8454, "step": 7083 }, { "epoch": 1.28, "learning_rate": 1.0795228628230616e-05, "loss": 0.8828, "step": 7084 }, { "epoch": 1.28, "learning_rate": 1.0792517621543466e-05, "loss": 0.7937, "step": 7085 }, { "epoch": 1.28, "learning_rate": 1.0789806614856318e-05, "loss": 0.8288, "step": 7086 }, { "epoch": 1.28, "learning_rate": 1.0787095608169167e-05, "loss": 1.0285, "step": 7087 }, { "epoch": 1.28, "learning_rate": 1.0784384601482017e-05, "loss": 0.9279, "step": 7088 }, { "epoch": 1.28, "learning_rate": 1.0781673594794867e-05, "loss": 0.5269, "step": 7089 }, { "epoch": 1.28, "learning_rate": 1.0778962588107719e-05, "loss": 0.5685, "step": 7090 }, { "epoch": 1.28, "learning_rate": 1.0776251581420567e-05, "loss": 0.7239, "step": 7091 }, { "epoch": 1.28, "learning_rate": 1.0773540574733417e-05, "loss": 0.8828, "step": 7092 }, { "epoch": 1.28, "learning_rate": 1.0770829568046267e-05, "loss": 0.4329, "step": 7093 }, { "epoch": 1.28, "learning_rate": 1.076811856135912e-05, "loss": 0.7346, "step": 7094 }, { "epoch": 1.28, "learning_rate": 1.0765407554671968e-05, "loss": 0.6491, "step": 7095 }, { "epoch": 1.28, "learning_rate": 1.0762696547984818e-05, "loss": 0.5734, "step": 7096 }, { "epoch": 1.28, "learning_rate": 1.075998554129767e-05, "loss": 0.8371, "step": 7097 }, { "epoch": 1.28, "learning_rate": 1.075727453461052e-05, "loss": 1.6556, "step": 7098 }, { "epoch": 1.28, "learning_rate": 1.0754563527923368e-05, "loss": 1.1206, "step": 7099 }, { "epoch": 1.28, "learning_rate": 1.0751852521236218e-05, "loss": 0.6181, "step": 7100 }, { "epoch": 1.28, "learning_rate": 1.074914151454907e-05, "loss": 0.9211, "step": 7101 }, { "epoch": 1.28, "learning_rate": 1.074643050786192e-05, "loss": 0.621, "step": 7102 }, { "epoch": 1.28, "learning_rate": 1.0743719501174769e-05, "loss": 0.6225, "step": 7103 }, { "epoch": 1.28, "learning_rate": 1.0741008494487619e-05, "loss": 0.5728, "step": 7104 }, { "epoch": 1.28, "learning_rate": 1.073829748780047e-05, "loss": 0.4669, "step": 7105 }, { "epoch": 1.28, "learning_rate": 1.0735586481113321e-05, "loss": 0.4916, "step": 7106 }, { "epoch": 1.28, "learning_rate": 1.073287547442617e-05, "loss": 0.6737, "step": 7107 }, { "epoch": 1.28, "learning_rate": 1.0730164467739021e-05, "loss": 0.8336, "step": 7108 }, { "epoch": 1.28, "learning_rate": 1.0727453461051871e-05, "loss": 0.521, "step": 7109 }, { "epoch": 1.29, "learning_rate": 1.0724742454364721e-05, "loss": 0.4678, "step": 7110 }, { "epoch": 1.29, "learning_rate": 1.072203144767757e-05, "loss": 0.6854, "step": 7111 }, { "epoch": 1.29, "learning_rate": 1.0719320440990422e-05, "loss": 0.4818, "step": 7112 }, { "epoch": 1.29, "learning_rate": 1.0716609434303272e-05, "loss": 0.7278, "step": 7113 }, { "epoch": 1.29, "learning_rate": 1.0713898427616122e-05, "loss": 0.4774, "step": 7114 }, { "epoch": 1.29, "learning_rate": 1.071118742092897e-05, "loss": 0.6599, "step": 7115 }, { "epoch": 1.29, "learning_rate": 1.0708476414241822e-05, "loss": 0.9295, "step": 7116 }, { "epoch": 1.29, "learning_rate": 1.0705765407554672e-05, "loss": 0.4477, "step": 7117 }, { "epoch": 1.29, "learning_rate": 1.0703054400867522e-05, "loss": 0.6161, "step": 7118 }, { "epoch": 1.29, "learning_rate": 1.0700343394180373e-05, "loss": 1.0749, "step": 7119 }, { "epoch": 1.29, "learning_rate": 1.0697632387493223e-05, "loss": 0.5939, "step": 7120 }, { "epoch": 1.29, "learning_rate": 1.0694921380806073e-05, "loss": 0.9207, "step": 7121 }, { "epoch": 1.29, "learning_rate": 1.0692210374118923e-05, "loss": 0.5064, "step": 7122 }, { "epoch": 1.29, "learning_rate": 1.0689499367431773e-05, "loss": 0.5255, "step": 7123 }, { "epoch": 1.29, "learning_rate": 1.0686788360744623e-05, "loss": 1.2325, "step": 7124 }, { "epoch": 1.29, "learning_rate": 1.0684077354057473e-05, "loss": 1.1526, "step": 7125 }, { "epoch": 1.29, "learning_rate": 1.0681366347370324e-05, "loss": 0.6871, "step": 7126 }, { "epoch": 1.29, "learning_rate": 1.0678655340683174e-05, "loss": 0.3412, "step": 7127 }, { "epoch": 1.29, "learning_rate": 1.0675944333996024e-05, "loss": 1.4971, "step": 7128 }, { "epoch": 1.29, "learning_rate": 1.0673233327308874e-05, "loss": 0.5714, "step": 7129 }, { "epoch": 1.29, "learning_rate": 1.0670522320621726e-05, "loss": 0.9512, "step": 7130 }, { "epoch": 1.29, "learning_rate": 1.0667811313934574e-05, "loss": 1.2546, "step": 7131 }, { "epoch": 1.29, "learning_rate": 1.0665100307247424e-05, "loss": 0.6855, "step": 7132 }, { "epoch": 1.29, "learning_rate": 1.0662389300560275e-05, "loss": 0.6911, "step": 7133 }, { "epoch": 1.29, "learning_rate": 1.0659678293873126e-05, "loss": 0.2957, "step": 7134 }, { "epoch": 1.29, "learning_rate": 1.0656967287185975e-05, "loss": 0.7959, "step": 7135 }, { "epoch": 1.29, "learning_rate": 1.0654256280498825e-05, "loss": 0.5716, "step": 7136 }, { "epoch": 1.29, "learning_rate": 1.0651545273811675e-05, "loss": 0.6212, "step": 7137 }, { "epoch": 1.29, "learning_rate": 1.0648834267124527e-05, "loss": 0.7058, "step": 7138 }, { "epoch": 1.29, "learning_rate": 1.0646123260437375e-05, "loss": 1.1381, "step": 7139 }, { "epoch": 1.29, "learning_rate": 1.0643412253750225e-05, "loss": 0.9161, "step": 7140 }, { "epoch": 1.29, "learning_rate": 1.0640701247063077e-05, "loss": 0.7285, "step": 7141 }, { "epoch": 1.29, "learning_rate": 1.0637990240375927e-05, "loss": 0.8909, "step": 7142 }, { "epoch": 1.29, "learning_rate": 1.0635279233688776e-05, "loss": 0.7889, "step": 7143 }, { "epoch": 1.29, "learning_rate": 1.0632568227001626e-05, "loss": 1.3969, "step": 7144 }, { "epoch": 1.29, "learning_rate": 1.0629857220314478e-05, "loss": 0.4555, "step": 7145 }, { "epoch": 1.29, "learning_rate": 1.0627146213627328e-05, "loss": 0.8253, "step": 7146 }, { "epoch": 1.29, "learning_rate": 1.0624435206940176e-05, "loss": 0.8738, "step": 7147 }, { "epoch": 1.29, "learning_rate": 1.0621724200253027e-05, "loss": 0.4559, "step": 7148 }, { "epoch": 1.29, "learning_rate": 1.0619013193565878e-05, "loss": 0.6937, "step": 7149 }, { "epoch": 1.29, "learning_rate": 1.0616302186878728e-05, "loss": 0.7368, "step": 7150 }, { "epoch": 1.29, "learning_rate": 1.0613591180191577e-05, "loss": 0.3157, "step": 7151 }, { "epoch": 1.29, "learning_rate": 1.0610880173504429e-05, "loss": 1.023, "step": 7152 }, { "epoch": 1.29, "learning_rate": 1.0608169166817279e-05, "loss": 0.7385, "step": 7153 }, { "epoch": 1.29, "learning_rate": 1.0605458160130129e-05, "loss": 0.731, "step": 7154 }, { "epoch": 1.29, "learning_rate": 1.0602747153442977e-05, "loss": 0.9162, "step": 7155 }, { "epoch": 1.29, "learning_rate": 1.060003614675583e-05, "loss": 0.4214, "step": 7156 }, { "epoch": 1.29, "learning_rate": 1.059732514006868e-05, "loss": 0.7421, "step": 7157 }, { "epoch": 1.29, "learning_rate": 1.059461413338153e-05, "loss": 1.092, "step": 7158 }, { "epoch": 1.29, "learning_rate": 1.059190312669438e-05, "loss": 0.9875, "step": 7159 }, { "epoch": 1.29, "learning_rate": 1.058919212000723e-05, "loss": 0.9953, "step": 7160 }, { "epoch": 1.29, "learning_rate": 1.058648111332008e-05, "loss": 0.7342, "step": 7161 }, { "epoch": 1.29, "learning_rate": 1.058377010663293e-05, "loss": 0.727, "step": 7162 }, { "epoch": 1.29, "learning_rate": 1.058105909994578e-05, "loss": 0.787, "step": 7163 }, { "epoch": 1.29, "learning_rate": 1.057834809325863e-05, "loss": 0.7953, "step": 7164 }, { "epoch": 1.29, "learning_rate": 1.057563708657148e-05, "loss": 0.5149, "step": 7165 }, { "epoch": 1.3, "learning_rate": 1.057292607988433e-05, "loss": 0.7082, "step": 7166 }, { "epoch": 1.3, "learning_rate": 1.057021507319718e-05, "loss": 0.6657, "step": 7167 }, { "epoch": 1.3, "learning_rate": 1.0567504066510031e-05, "loss": 0.4633, "step": 7168 }, { "epoch": 1.3, "learning_rate": 1.0564793059822881e-05, "loss": 0.426, "step": 7169 }, { "epoch": 1.3, "learning_rate": 1.0562082053135733e-05, "loss": 1.5247, "step": 7170 }, { "epoch": 1.3, "learning_rate": 1.0559371046448581e-05, "loss": 0.2347, "step": 7171 }, { "epoch": 1.3, "learning_rate": 1.0556660039761431e-05, "loss": 0.6417, "step": 7172 }, { "epoch": 1.3, "learning_rate": 1.0553949033074282e-05, "loss": 0.4872, "step": 7173 }, { "epoch": 1.3, "learning_rate": 1.0551238026387133e-05, "loss": 0.8514, "step": 7174 }, { "epoch": 1.3, "learning_rate": 1.0548527019699982e-05, "loss": 0.8878, "step": 7175 }, { "epoch": 1.3, "learning_rate": 1.0545816013012832e-05, "loss": 0.8238, "step": 7176 }, { "epoch": 1.3, "learning_rate": 1.0543105006325682e-05, "loss": 0.7042, "step": 7177 }, { "epoch": 1.3, "learning_rate": 1.0540393999638534e-05, "loss": 0.8976, "step": 7178 }, { "epoch": 1.3, "learning_rate": 1.0537682992951382e-05, "loss": 0.5031, "step": 7179 }, { "epoch": 1.3, "learning_rate": 1.0534971986264233e-05, "loss": 0.6858, "step": 7180 }, { "epoch": 1.3, "learning_rate": 1.0532260979577084e-05, "loss": 0.764, "step": 7181 }, { "epoch": 1.3, "learning_rate": 1.0529549972889934e-05, "loss": 0.6781, "step": 7182 }, { "epoch": 1.3, "learning_rate": 1.0526838966202783e-05, "loss": 1.0712, "step": 7183 }, { "epoch": 1.3, "learning_rate": 1.0524127959515633e-05, "loss": 1.2349, "step": 7184 }, { "epoch": 1.3, "learning_rate": 1.0521416952828485e-05, "loss": 0.41, "step": 7185 }, { "epoch": 1.3, "learning_rate": 1.0518705946141335e-05, "loss": 0.4693, "step": 7186 }, { "epoch": 1.3, "learning_rate": 1.0515994939454183e-05, "loss": 0.7404, "step": 7187 }, { "epoch": 1.3, "learning_rate": 1.0513283932767034e-05, "loss": 0.5827, "step": 7188 }, { "epoch": 1.3, "learning_rate": 1.0510572926079885e-05, "loss": 0.561, "step": 7189 }, { "epoch": 1.3, "learning_rate": 1.0507861919392736e-05, "loss": 0.6129, "step": 7190 }, { "epoch": 1.3, "learning_rate": 1.0505150912705584e-05, "loss": 0.3873, "step": 7191 }, { "epoch": 1.3, "learning_rate": 1.0502439906018436e-05, "loss": 0.6434, "step": 7192 }, { "epoch": 1.3, "learning_rate": 1.0499728899331286e-05, "loss": 0.3498, "step": 7193 }, { "epoch": 1.3, "learning_rate": 1.0497017892644136e-05, "loss": 1.4526, "step": 7194 }, { "epoch": 1.3, "learning_rate": 1.0494306885956985e-05, "loss": 0.4942, "step": 7195 }, { "epoch": 1.3, "learning_rate": 1.0491595879269836e-05, "loss": 0.7435, "step": 7196 }, { "epoch": 1.3, "learning_rate": 1.0488884872582686e-05, "loss": 0.4472, "step": 7197 }, { "epoch": 1.3, "learning_rate": 1.0486173865895537e-05, "loss": 0.7496, "step": 7198 }, { "epoch": 1.3, "learning_rate": 1.0483462859208385e-05, "loss": 0.5203, "step": 7199 }, { "epoch": 1.3, "learning_rate": 1.0480751852521237e-05, "loss": 0.668, "step": 7200 }, { "epoch": 1.3, "learning_rate": 1.0478040845834087e-05, "loss": 0.7517, "step": 7201 }, { "epoch": 1.3, "learning_rate": 1.0475329839146937e-05, "loss": 0.4458, "step": 7202 }, { "epoch": 1.3, "learning_rate": 1.0472618832459787e-05, "loss": 0.9633, "step": 7203 }, { "epoch": 1.3, "learning_rate": 1.0469907825772637e-05, "loss": 0.9121, "step": 7204 }, { "epoch": 1.3, "learning_rate": 1.0467196819085488e-05, "loss": 0.3104, "step": 7205 }, { "epoch": 1.3, "learning_rate": 1.0464485812398336e-05, "loss": 0.422, "step": 7206 }, { "epoch": 1.3, "learning_rate": 1.0461774805711188e-05, "loss": 0.9605, "step": 7207 }, { "epoch": 1.3, "learning_rate": 1.0459063799024038e-05, "loss": 0.5814, "step": 7208 }, { "epoch": 1.3, "learning_rate": 1.0456352792336888e-05, "loss": 0.4729, "step": 7209 }, { "epoch": 1.3, "learning_rate": 1.0453641785649737e-05, "loss": 0.902, "step": 7210 }, { "epoch": 1.3, "learning_rate": 1.0450930778962588e-05, "loss": 0.975, "step": 7211 }, { "epoch": 1.3, "learning_rate": 1.0448219772275439e-05, "loss": 0.535, "step": 7212 }, { "epoch": 1.3, "learning_rate": 1.0445508765588289e-05, "loss": 1.2676, "step": 7213 }, { "epoch": 1.3, "learning_rate": 1.0442797758901139e-05, "loss": 0.3711, "step": 7214 }, { "epoch": 1.3, "learning_rate": 1.0440086752213989e-05, "loss": 1.6249, "step": 7215 }, { "epoch": 1.3, "learning_rate": 1.0437375745526839e-05, "loss": 0.4161, "step": 7216 }, { "epoch": 1.3, "learning_rate": 1.043466473883969e-05, "loss": 1.4169, "step": 7217 }, { "epoch": 1.3, "learning_rate": 1.043195373215254e-05, "loss": 1.2506, "step": 7218 }, { "epoch": 1.3, "learning_rate": 1.042924272546539e-05, "loss": 0.5756, "step": 7219 }, { "epoch": 1.3, "learning_rate": 1.042653171877824e-05, "loss": 1.0062, "step": 7220 }, { "epoch": 1.31, "learning_rate": 1.042382071209109e-05, "loss": 1.1382, "step": 7221 }, { "epoch": 1.31, "learning_rate": 1.042110970540394e-05, "loss": 0.8369, "step": 7222 }, { "epoch": 1.31, "learning_rate": 1.041839869871679e-05, "loss": 0.7244, "step": 7223 }, { "epoch": 1.31, "learning_rate": 1.041568769202964e-05, "loss": 1.3378, "step": 7224 }, { "epoch": 1.31, "learning_rate": 1.0412976685342492e-05, "loss": 0.7405, "step": 7225 }, { "epoch": 1.31, "learning_rate": 1.041026567865534e-05, "loss": 0.7756, "step": 7226 }, { "epoch": 1.31, "learning_rate": 1.040755467196819e-05, "loss": 0.8793, "step": 7227 }, { "epoch": 1.31, "learning_rate": 1.040484366528104e-05, "loss": 0.8128, "step": 7228 }, { "epoch": 1.31, "learning_rate": 1.0402132658593892e-05, "loss": 0.6797, "step": 7229 }, { "epoch": 1.31, "learning_rate": 1.0399421651906741e-05, "loss": 0.6518, "step": 7230 }, { "epoch": 1.31, "learning_rate": 1.0396710645219591e-05, "loss": 1.3884, "step": 7231 }, { "epoch": 1.31, "learning_rate": 1.0393999638532443e-05, "loss": 0.6175, "step": 7232 }, { "epoch": 1.31, "learning_rate": 1.0391288631845293e-05, "loss": 0.4584, "step": 7233 }, { "epoch": 1.31, "learning_rate": 1.0388577625158141e-05, "loss": 0.8309, "step": 7234 }, { "epoch": 1.31, "learning_rate": 1.0385866618470992e-05, "loss": 0.6802, "step": 7235 }, { "epoch": 1.31, "learning_rate": 1.0383155611783843e-05, "loss": 1.3064, "step": 7236 }, { "epoch": 1.31, "learning_rate": 1.0380444605096694e-05, "loss": 0.6262, "step": 7237 }, { "epoch": 1.31, "learning_rate": 1.0377733598409542e-05, "loss": 0.5949, "step": 7238 }, { "epoch": 1.31, "learning_rate": 1.0375022591722392e-05, "loss": 0.715, "step": 7239 }, { "epoch": 1.31, "learning_rate": 1.0372311585035244e-05, "loss": 1.1859, "step": 7240 }, { "epoch": 1.31, "learning_rate": 1.0369600578348094e-05, "loss": 1.1094, "step": 7241 }, { "epoch": 1.31, "learning_rate": 1.0366889571660943e-05, "loss": 0.8369, "step": 7242 }, { "epoch": 1.31, "learning_rate": 1.0364178564973794e-05, "loss": 0.3584, "step": 7243 }, { "epoch": 1.31, "learning_rate": 1.0361467558286644e-05, "loss": 0.7603, "step": 7244 }, { "epoch": 1.31, "learning_rate": 1.0358756551599495e-05, "loss": 0.4338, "step": 7245 }, { "epoch": 1.31, "learning_rate": 1.0356045544912343e-05, "loss": 1.4111, "step": 7246 }, { "epoch": 1.31, "learning_rate": 1.0353334538225195e-05, "loss": 0.9076, "step": 7247 }, { "epoch": 1.31, "learning_rate": 1.0350623531538045e-05, "loss": 0.4934, "step": 7248 }, { "epoch": 1.31, "learning_rate": 1.0347912524850895e-05, "loss": 0.5025, "step": 7249 }, { "epoch": 1.31, "learning_rate": 1.0345201518163744e-05, "loss": 0.489, "step": 7250 }, { "epoch": 1.31, "learning_rate": 1.0342490511476595e-05, "loss": 0.4028, "step": 7251 }, { "epoch": 1.31, "learning_rate": 1.0339779504789446e-05, "loss": 0.7569, "step": 7252 }, { "epoch": 1.31, "learning_rate": 1.0337068498102296e-05, "loss": 0.5704, "step": 7253 }, { "epoch": 1.31, "learning_rate": 1.0334357491415146e-05, "loss": 0.7647, "step": 7254 }, { "epoch": 1.31, "learning_rate": 1.0331646484727996e-05, "loss": 0.4931, "step": 7255 }, { "epoch": 1.31, "learning_rate": 1.0328935478040846e-05, "loss": 0.3332, "step": 7256 }, { "epoch": 1.31, "learning_rate": 1.0326224471353696e-05, "loss": 0.7672, "step": 7257 }, { "epoch": 1.31, "learning_rate": 1.0323513464666546e-05, "loss": 0.916, "step": 7258 }, { "epoch": 1.31, "learning_rate": 1.0320802457979397e-05, "loss": 0.5823, "step": 7259 }, { "epoch": 1.31, "learning_rate": 1.0318091451292247e-05, "loss": 0.654, "step": 7260 }, { "epoch": 1.31, "learning_rate": 1.0315380444605097e-05, "loss": 0.2991, "step": 7261 }, { "epoch": 1.31, "learning_rate": 1.0312669437917947e-05, "loss": 0.6088, "step": 7262 }, { "epoch": 1.31, "learning_rate": 1.0309958431230797e-05, "loss": 0.3768, "step": 7263 }, { "epoch": 1.31, "learning_rate": 1.0307247424543647e-05, "loss": 0.2819, "step": 7264 }, { "epoch": 1.31, "learning_rate": 1.0304536417856499e-05, "loss": 1.3167, "step": 7265 }, { "epoch": 1.31, "learning_rate": 1.0301825411169347e-05, "loss": 0.4969, "step": 7266 }, { "epoch": 1.31, "learning_rate": 1.0299114404482198e-05, "loss": 1.289, "step": 7267 }, { "epoch": 1.31, "learning_rate": 1.0296403397795048e-05, "loss": 1.3074, "step": 7268 }, { "epoch": 1.31, "learning_rate": 1.02936923911079e-05, "loss": 1.2198, "step": 7269 }, { "epoch": 1.31, "learning_rate": 1.0290981384420748e-05, "loss": 0.4957, "step": 7270 }, { "epoch": 1.31, "learning_rate": 1.0288270377733598e-05, "loss": 0.9931, "step": 7271 }, { "epoch": 1.31, "learning_rate": 1.0285559371046448e-05, "loss": 0.6172, "step": 7272 }, { "epoch": 1.31, "learning_rate": 1.02828483643593e-05, "loss": 0.4901, "step": 7273 }, { "epoch": 1.31, "learning_rate": 1.0280137357672149e-05, "loss": 0.4769, "step": 7274 }, { "epoch": 1.31, "learning_rate": 1.0277426350984999e-05, "loss": 1.1693, "step": 7275 }, { "epoch": 1.32, "learning_rate": 1.027471534429785e-05, "loss": 0.6612, "step": 7276 }, { "epoch": 1.32, "learning_rate": 1.02720043376107e-05, "loss": 0.7568, "step": 7277 }, { "epoch": 1.32, "learning_rate": 1.0269293330923549e-05, "loss": 0.4356, "step": 7278 }, { "epoch": 1.32, "learning_rate": 1.02665823242364e-05, "loss": 0.8485, "step": 7279 }, { "epoch": 1.32, "learning_rate": 1.0263871317549251e-05, "loss": 0.7524, "step": 7280 }, { "epoch": 1.32, "learning_rate": 1.0261160310862101e-05, "loss": 0.6768, "step": 7281 }, { "epoch": 1.32, "learning_rate": 1.025844930417495e-05, "loss": 0.6888, "step": 7282 }, { "epoch": 1.32, "learning_rate": 1.02557382974878e-05, "loss": 0.8601, "step": 7283 }, { "epoch": 1.32, "learning_rate": 1.0253027290800652e-05, "loss": 0.403, "step": 7284 }, { "epoch": 1.32, "learning_rate": 1.0250316284113502e-05, "loss": 0.7395, "step": 7285 }, { "epoch": 1.32, "learning_rate": 1.024760527742635e-05, "loss": 1.0422, "step": 7286 }, { "epoch": 1.32, "learning_rate": 1.0244894270739202e-05, "loss": 0.3811, "step": 7287 }, { "epoch": 1.32, "learning_rate": 1.0242183264052052e-05, "loss": 1.5316, "step": 7288 }, { "epoch": 1.32, "learning_rate": 1.0239472257364902e-05, "loss": 0.7933, "step": 7289 }, { "epoch": 1.32, "learning_rate": 1.023676125067775e-05, "loss": 0.3903, "step": 7290 }, { "epoch": 1.32, "learning_rate": 1.0234050243990603e-05, "loss": 0.8577, "step": 7291 }, { "epoch": 1.32, "learning_rate": 1.0231339237303453e-05, "loss": 0.3097, "step": 7292 }, { "epoch": 1.32, "learning_rate": 1.0228628230616303e-05, "loss": 0.8924, "step": 7293 }, { "epoch": 1.32, "learning_rate": 1.0225917223929153e-05, "loss": 1.1151, "step": 7294 }, { "epoch": 1.32, "learning_rate": 1.0223206217242003e-05, "loss": 0.893, "step": 7295 }, { "epoch": 1.32, "learning_rate": 1.0220495210554853e-05, "loss": 0.9173, "step": 7296 }, { "epoch": 1.32, "learning_rate": 1.0217784203867703e-05, "loss": 0.9617, "step": 7297 }, { "epoch": 1.32, "learning_rate": 1.0215073197180553e-05, "loss": 1.1645, "step": 7298 }, { "epoch": 1.32, "learning_rate": 1.0212362190493404e-05, "loss": 0.6054, "step": 7299 }, { "epoch": 1.32, "learning_rate": 1.0209651183806254e-05, "loss": 0.6049, "step": 7300 }, { "epoch": 1.32, "learning_rate": 1.0206940177119104e-05, "loss": 0.8186, "step": 7301 }, { "epoch": 1.32, "learning_rate": 1.0204229170431954e-05, "loss": 0.7171, "step": 7302 }, { "epoch": 1.32, "learning_rate": 1.0201518163744804e-05, "loss": 0.997, "step": 7303 }, { "epoch": 1.32, "learning_rate": 1.0198807157057654e-05, "loss": 0.5754, "step": 7304 }, { "epoch": 1.32, "learning_rate": 1.0196096150370506e-05, "loss": 0.6729, "step": 7305 }, { "epoch": 1.32, "learning_rate": 1.0193385143683355e-05, "loss": 0.6995, "step": 7306 }, { "epoch": 1.32, "learning_rate": 1.0190674136996205e-05, "loss": 0.4577, "step": 7307 }, { "epoch": 1.32, "learning_rate": 1.0187963130309055e-05, "loss": 0.569, "step": 7308 }, { "epoch": 1.32, "learning_rate": 1.0185252123621907e-05, "loss": 0.7172, "step": 7309 }, { "epoch": 1.32, "learning_rate": 1.0182541116934755e-05, "loss": 0.5611, "step": 7310 }, { "epoch": 1.32, "learning_rate": 1.0179830110247605e-05, "loss": 0.8308, "step": 7311 }, { "epoch": 1.32, "learning_rate": 1.0177119103560455e-05, "loss": 0.9325, "step": 7312 }, { "epoch": 1.32, "learning_rate": 1.0174408096873307e-05, "loss": 0.8739, "step": 7313 }, { "epoch": 1.32, "learning_rate": 1.0171697090186156e-05, "loss": 0.5199, "step": 7314 }, { "epoch": 1.32, "learning_rate": 1.0168986083499006e-05, "loss": 0.8739, "step": 7315 }, { "epoch": 1.32, "learning_rate": 1.0166275076811858e-05, "loss": 0.8731, "step": 7316 }, { "epoch": 1.32, "learning_rate": 1.0163564070124708e-05, "loss": 0.7899, "step": 7317 }, { "epoch": 1.32, "learning_rate": 1.0160853063437556e-05, "loss": 0.4418, "step": 7318 }, { "epoch": 1.32, "learning_rate": 1.0158142056750406e-05, "loss": 1.1247, "step": 7319 }, { "epoch": 1.32, "learning_rate": 1.0155431050063258e-05, "loss": 0.8157, "step": 7320 }, { "epoch": 1.32, "learning_rate": 1.0152720043376108e-05, "loss": 0.7374, "step": 7321 }, { "epoch": 1.32, "learning_rate": 1.0150009036688957e-05, "loss": 0.5971, "step": 7322 }, { "epoch": 1.32, "learning_rate": 1.0147298030001807e-05, "loss": 0.421, "step": 7323 }, { "epoch": 1.32, "learning_rate": 1.0144587023314659e-05, "loss": 1.0179, "step": 7324 }, { "epoch": 1.32, "learning_rate": 1.0141876016627509e-05, "loss": 0.2138, "step": 7325 }, { "epoch": 1.32, "learning_rate": 1.0139165009940357e-05, "loss": 0.596, "step": 7326 }, { "epoch": 1.32, "learning_rate": 1.0136454003253209e-05, "loss": 0.521, "step": 7327 }, { "epoch": 1.32, "learning_rate": 1.013374299656606e-05, "loss": 0.8187, "step": 7328 }, { "epoch": 1.32, "learning_rate": 1.013103198987891e-05, "loss": 1.1435, "step": 7329 }, { "epoch": 1.32, "learning_rate": 1.0128320983191758e-05, "loss": 0.5564, "step": 7330 }, { "epoch": 1.32, "learning_rate": 1.012560997650461e-05, "loss": 0.9373, "step": 7331 }, { "epoch": 1.33, "learning_rate": 1.012289896981746e-05, "loss": 0.5774, "step": 7332 }, { "epoch": 1.33, "learning_rate": 1.012018796313031e-05, "loss": 1.6148, "step": 7333 }, { "epoch": 1.33, "learning_rate": 1.0117476956443158e-05, "loss": 1.063, "step": 7334 }, { "epoch": 1.33, "learning_rate": 1.011476594975601e-05, "loss": 0.5739, "step": 7335 }, { "epoch": 1.33, "learning_rate": 1.011205494306886e-05, "loss": 0.6914, "step": 7336 }, { "epoch": 1.33, "learning_rate": 1.0109343936381709e-05, "loss": 1.2101, "step": 7337 }, { "epoch": 1.33, "learning_rate": 1.010663292969456e-05, "loss": 0.8411, "step": 7338 }, { "epoch": 1.33, "learning_rate": 1.010392192300741e-05, "loss": 0.474, "step": 7339 }, { "epoch": 1.33, "learning_rate": 1.010121091632026e-05, "loss": 0.9923, "step": 7340 }, { "epoch": 1.33, "learning_rate": 1.009849990963311e-05, "loss": 0.5416, "step": 7341 }, { "epoch": 1.33, "learning_rate": 1.0095788902945961e-05, "loss": 1.1217, "step": 7342 }, { "epoch": 1.33, "learning_rate": 1.0093077896258811e-05, "loss": 1.1523, "step": 7343 }, { "epoch": 1.33, "learning_rate": 1.0090366889571661e-05, "loss": 0.2338, "step": 7344 }, { "epoch": 1.33, "learning_rate": 1.008765588288451e-05, "loss": 0.822, "step": 7345 }, { "epoch": 1.33, "learning_rate": 1.0084944876197362e-05, "loss": 0.6711, "step": 7346 }, { "epoch": 1.33, "learning_rate": 1.0082233869510212e-05, "loss": 1.211, "step": 7347 }, { "epoch": 1.33, "learning_rate": 1.0079522862823062e-05, "loss": 1.115, "step": 7348 }, { "epoch": 1.33, "learning_rate": 1.0076811856135912e-05, "loss": 0.636, "step": 7349 }, { "epoch": 1.33, "learning_rate": 1.0074100849448762e-05, "loss": 0.5756, "step": 7350 }, { "epoch": 1.33, "learning_rate": 1.0071389842761612e-05, "loss": 0.5914, "step": 7351 }, { "epoch": 1.33, "learning_rate": 1.0068678836074462e-05, "loss": 0.9724, "step": 7352 }, { "epoch": 1.33, "learning_rate": 1.0065967829387313e-05, "loss": 1.0997, "step": 7353 }, { "epoch": 1.33, "learning_rate": 1.0063256822700163e-05, "loss": 0.6568, "step": 7354 }, { "epoch": 1.33, "learning_rate": 1.0060545816013013e-05, "loss": 0.8923, "step": 7355 }, { "epoch": 1.33, "learning_rate": 1.0057834809325863e-05, "loss": 0.9068, "step": 7356 }, { "epoch": 1.33, "learning_rate": 1.0055123802638713e-05, "loss": 0.735, "step": 7357 }, { "epoch": 1.33, "learning_rate": 1.0052412795951563e-05, "loss": 0.9011, "step": 7358 }, { "epoch": 1.33, "learning_rate": 1.0049701789264413e-05, "loss": 0.7454, "step": 7359 }, { "epoch": 1.33, "learning_rate": 1.0046990782577265e-05, "loss": 0.9642, "step": 7360 }, { "epoch": 1.33, "learning_rate": 1.0044279775890114e-05, "loss": 1.0498, "step": 7361 }, { "epoch": 1.33, "learning_rate": 1.0041568769202964e-05, "loss": 0.9302, "step": 7362 }, { "epoch": 1.33, "learning_rate": 1.0038857762515814e-05, "loss": 0.8513, "step": 7363 }, { "epoch": 1.33, "learning_rate": 1.0036146755828666e-05, "loss": 1.8154, "step": 7364 }, { "epoch": 1.33, "learning_rate": 1.0033435749141514e-05, "loss": 0.6327, "step": 7365 }, { "epoch": 1.33, "learning_rate": 1.0030724742454364e-05, "loss": 0.4356, "step": 7366 }, { "epoch": 1.33, "learning_rate": 1.0028013735767216e-05, "loss": 0.6875, "step": 7367 }, { "epoch": 1.33, "learning_rate": 1.0025302729080066e-05, "loss": 0.747, "step": 7368 }, { "epoch": 1.33, "learning_rate": 1.0022591722392915e-05, "loss": 0.6727, "step": 7369 }, { "epoch": 1.33, "learning_rate": 1.0019880715705765e-05, "loss": 0.7854, "step": 7370 }, { "epoch": 1.33, "learning_rate": 1.0017169709018617e-05, "loss": 0.5905, "step": 7371 }, { "epoch": 1.33, "learning_rate": 1.0014458702331467e-05, "loss": 0.4973, "step": 7372 }, { "epoch": 1.33, "learning_rate": 1.0011747695644315e-05, "loss": 0.7463, "step": 7373 }, { "epoch": 1.33, "learning_rate": 1.0009036688957165e-05, "loss": 0.5761, "step": 7374 }, { "epoch": 1.33, "learning_rate": 1.0006325682270017e-05, "loss": 0.2975, "step": 7375 }, { "epoch": 1.33, "learning_rate": 1.0003614675582867e-05, "loss": 0.4248, "step": 7376 }, { "epoch": 1.33, "learning_rate": 1.0000903668895716e-05, "loss": 0.6652, "step": 7377 }, { "epoch": 1.33, "learning_rate": 9.998192662208568e-06, "loss": 0.7463, "step": 7378 }, { "epoch": 1.33, "learning_rate": 9.995481655521418e-06, "loss": 1.0752, "step": 7379 }, { "epoch": 1.33, "learning_rate": 9.992770648834268e-06, "loss": 1.4347, "step": 7380 }, { "epoch": 1.33, "learning_rate": 9.990059642147116e-06, "loss": 0.7374, "step": 7381 }, { "epoch": 1.33, "learning_rate": 9.987348635459968e-06, "loss": 0.8555, "step": 7382 }, { "epoch": 1.33, "learning_rate": 9.984637628772818e-06, "loss": 0.8902, "step": 7383 }, { "epoch": 1.33, "learning_rate": 9.981926622085668e-06, "loss": 0.4673, "step": 7384 }, { "epoch": 1.33, "learning_rate": 9.979215615398517e-06, "loss": 0.7997, "step": 7385 }, { "epoch": 1.33, "learning_rate": 9.976504608711369e-06, "loss": 0.2471, "step": 7386 }, { "epoch": 1.34, "learning_rate": 9.973793602024219e-06, "loss": 0.4077, "step": 7387 }, { "epoch": 1.34, "learning_rate": 9.971082595337069e-06, "loss": 0.6114, "step": 7388 }, { "epoch": 1.34, "learning_rate": 9.968371588649919e-06, "loss": 0.7286, "step": 7389 }, { "epoch": 1.34, "learning_rate": 9.96566058196277e-06, "loss": 0.4368, "step": 7390 }, { "epoch": 1.34, "learning_rate": 9.96294957527562e-06, "loss": 0.8838, "step": 7391 }, { "epoch": 1.34, "learning_rate": 9.96023856858847e-06, "loss": 0.4206, "step": 7392 }, { "epoch": 1.34, "learning_rate": 9.95752756190132e-06, "loss": 0.5621, "step": 7393 }, { "epoch": 1.34, "learning_rate": 9.95481655521417e-06, "loss": 0.5712, "step": 7394 }, { "epoch": 1.34, "learning_rate": 9.95210554852702e-06, "loss": 0.2881, "step": 7395 }, { "epoch": 1.34, "learning_rate": 9.94939454183987e-06, "loss": 0.8783, "step": 7396 }, { "epoch": 1.34, "learning_rate": 9.94668353515272e-06, "loss": 0.9139, "step": 7397 }, { "epoch": 1.34, "learning_rate": 9.94397252846557e-06, "loss": 0.8623, "step": 7398 }, { "epoch": 1.34, "learning_rate": 9.94126152177842e-06, "loss": 0.478, "step": 7399 }, { "epoch": 1.34, "learning_rate": 9.938550515091272e-06, "loss": 1.453, "step": 7400 }, { "epoch": 1.34, "learning_rate": 9.93583950840412e-06, "loss": 0.458, "step": 7401 }, { "epoch": 1.34, "learning_rate": 9.93312850171697e-06, "loss": 1.082, "step": 7402 }, { "epoch": 1.34, "learning_rate": 9.930417495029821e-06, "loss": 0.739, "step": 7403 }, { "epoch": 1.34, "learning_rate": 9.927706488342673e-06, "loss": 0.3713, "step": 7404 }, { "epoch": 1.34, "learning_rate": 9.924995481655521e-06, "loss": 0.4289, "step": 7405 }, { "epoch": 1.34, "learning_rate": 9.922284474968371e-06, "loss": 0.5732, "step": 7406 }, { "epoch": 1.34, "learning_rate": 9.919573468281221e-06, "loss": 0.5232, "step": 7407 }, { "epoch": 1.34, "learning_rate": 9.916862461594073e-06, "loss": 0.4859, "step": 7408 }, { "epoch": 1.34, "learning_rate": 9.914151454906922e-06, "loss": 1.0036, "step": 7409 }, { "epoch": 1.34, "learning_rate": 9.911440448219772e-06, "loss": 0.4063, "step": 7410 }, { "epoch": 1.34, "learning_rate": 9.908729441532624e-06, "loss": 1.1261, "step": 7411 }, { "epoch": 1.34, "learning_rate": 9.906018434845474e-06, "loss": 0.7025, "step": 7412 }, { "epoch": 1.34, "learning_rate": 9.903307428158322e-06, "loss": 0.9827, "step": 7413 }, { "epoch": 1.34, "learning_rate": 9.900596421471172e-06, "loss": 0.6172, "step": 7414 }, { "epoch": 1.34, "learning_rate": 9.897885414784024e-06, "loss": 0.5881, "step": 7415 }, { "epoch": 1.34, "learning_rate": 9.895174408096874e-06, "loss": 0.5181, "step": 7416 }, { "epoch": 1.34, "learning_rate": 9.892463401409723e-06, "loss": 0.3689, "step": 7417 }, { "epoch": 1.34, "learning_rate": 9.889752394722573e-06, "loss": 1.5051, "step": 7418 }, { "epoch": 1.34, "learning_rate": 9.887041388035425e-06, "loss": 0.6273, "step": 7419 }, { "epoch": 1.34, "learning_rate": 9.884330381348275e-06, "loss": 0.9165, "step": 7420 }, { "epoch": 1.34, "learning_rate": 9.881619374661123e-06, "loss": 0.4755, "step": 7421 }, { "epoch": 1.34, "learning_rate": 9.878908367973975e-06, "loss": 0.4918, "step": 7422 }, { "epoch": 1.34, "learning_rate": 9.876197361286825e-06, "loss": 1.0813, "step": 7423 }, { "epoch": 1.34, "learning_rate": 9.873486354599675e-06, "loss": 1.2054, "step": 7424 }, { "epoch": 1.34, "learning_rate": 9.870775347912524e-06, "loss": 0.8199, "step": 7425 }, { "epoch": 1.34, "learning_rate": 9.868064341225376e-06, "loss": 0.5002, "step": 7426 }, { "epoch": 1.34, "learning_rate": 9.865353334538226e-06, "loss": 0.6508, "step": 7427 }, { "epoch": 1.34, "learning_rate": 9.862642327851076e-06, "loss": 0.773, "step": 7428 }, { "epoch": 1.34, "learning_rate": 9.859931321163924e-06, "loss": 0.7562, "step": 7429 }, { "epoch": 1.34, "learning_rate": 9.857220314476776e-06, "loss": 1.0973, "step": 7430 }, { "epoch": 1.34, "learning_rate": 9.854509307789626e-06, "loss": 0.1034, "step": 7431 }, { "epoch": 1.34, "learning_rate": 9.851798301102477e-06, "loss": 0.4629, "step": 7432 }, { "epoch": 1.34, "learning_rate": 9.849087294415327e-06, "loss": 0.3854, "step": 7433 }, { "epoch": 1.34, "learning_rate": 9.846376287728177e-06, "loss": 1.4522, "step": 7434 }, { "epoch": 1.34, "learning_rate": 9.843665281041027e-06, "loss": 0.5309, "step": 7435 }, { "epoch": 1.34, "learning_rate": 9.840954274353877e-06, "loss": 0.6278, "step": 7436 }, { "epoch": 1.34, "learning_rate": 9.838243267666727e-06, "loss": 0.7585, "step": 7437 }, { "epoch": 1.34, "learning_rate": 9.835532260979577e-06, "loss": 0.604, "step": 7438 }, { "epoch": 1.34, "learning_rate": 9.832821254292427e-06, "loss": 0.8087, "step": 7439 }, { "epoch": 1.34, "learning_rate": 9.830110247605278e-06, "loss": 0.76, "step": 7440 }, { "epoch": 1.34, "learning_rate": 9.827399240918128e-06, "loss": 0.6892, "step": 7441 }, { "epoch": 1.35, "learning_rate": 9.824688234230978e-06, "loss": 0.7042, "step": 7442 }, { "epoch": 1.35, "learning_rate": 9.821977227543828e-06, "loss": 0.1889, "step": 7443 }, { "epoch": 1.35, "learning_rate": 9.81926622085668e-06, "loss": 0.4579, "step": 7444 }, { "epoch": 1.35, "learning_rate": 9.816555214169528e-06, "loss": 0.5406, "step": 7445 }, { "epoch": 1.35, "learning_rate": 9.813844207482378e-06, "loss": 1.4598, "step": 7446 }, { "epoch": 1.35, "learning_rate": 9.811133200795229e-06, "loss": 0.6361, "step": 7447 }, { "epoch": 1.35, "learning_rate": 9.80842219410808e-06, "loss": 0.4549, "step": 7448 }, { "epoch": 1.35, "learning_rate": 9.805711187420929e-06, "loss": 0.9091, "step": 7449 }, { "epoch": 1.35, "learning_rate": 9.803000180733779e-06, "loss": 0.9014, "step": 7450 }, { "epoch": 1.35, "learning_rate": 9.80028917404663e-06, "loss": 0.6989, "step": 7451 }, { "epoch": 1.35, "learning_rate": 9.797578167359481e-06, "loss": 0.6969, "step": 7452 }, { "epoch": 1.35, "learning_rate": 9.79486716067233e-06, "loss": 0.4463, "step": 7453 }, { "epoch": 1.35, "learning_rate": 9.79215615398518e-06, "loss": 0.7364, "step": 7454 }, { "epoch": 1.35, "learning_rate": 9.789445147298031e-06, "loss": 0.7181, "step": 7455 }, { "epoch": 1.35, "learning_rate": 9.786734140610881e-06, "loss": 0.4627, "step": 7456 }, { "epoch": 1.35, "learning_rate": 9.78402313392373e-06, "loss": 1.0914, "step": 7457 }, { "epoch": 1.35, "learning_rate": 9.78131212723658e-06, "loss": 0.9664, "step": 7458 }, { "epoch": 1.35, "learning_rate": 9.778601120549432e-06, "loss": 0.6553, "step": 7459 }, { "epoch": 1.35, "learning_rate": 9.775890113862282e-06, "loss": 0.291, "step": 7460 }, { "epoch": 1.35, "learning_rate": 9.77317910717513e-06, "loss": 0.2971, "step": 7461 }, { "epoch": 1.35, "learning_rate": 9.770468100487982e-06, "loss": 0.3771, "step": 7462 }, { "epoch": 1.35, "learning_rate": 9.767757093800832e-06, "loss": 0.7451, "step": 7463 }, { "epoch": 1.35, "learning_rate": 9.76504608711368e-06, "loss": 1.0545, "step": 7464 }, { "epoch": 1.35, "learning_rate": 9.762335080426531e-06, "loss": 0.353, "step": 7465 }, { "epoch": 1.35, "learning_rate": 9.759624073739383e-06, "loss": 0.4132, "step": 7466 }, { "epoch": 1.35, "learning_rate": 9.756913067052233e-06, "loss": 0.9501, "step": 7467 }, { "epoch": 1.35, "learning_rate": 9.754202060365081e-06, "loss": 0.5622, "step": 7468 }, { "epoch": 1.35, "learning_rate": 9.751491053677931e-06, "loss": 1.0204, "step": 7469 }, { "epoch": 1.35, "learning_rate": 9.748780046990783e-06, "loss": 0.6659, "step": 7470 }, { "epoch": 1.35, "learning_rate": 9.746069040303633e-06, "loss": 0.7444, "step": 7471 }, { "epoch": 1.35, "learning_rate": 9.743358033616482e-06, "loss": 0.5521, "step": 7472 }, { "epoch": 1.35, "learning_rate": 9.740647026929334e-06, "loss": 0.5753, "step": 7473 }, { "epoch": 1.35, "learning_rate": 9.737936020242184e-06, "loss": 1.2226, "step": 7474 }, { "epoch": 1.35, "learning_rate": 9.735225013555034e-06, "loss": 0.58, "step": 7475 }, { "epoch": 1.35, "learning_rate": 9.732514006867882e-06, "loss": 1.3297, "step": 7476 }, { "epoch": 1.35, "learning_rate": 9.729803000180734e-06, "loss": 0.2925, "step": 7477 }, { "epoch": 1.35, "learning_rate": 9.727091993493584e-06, "loss": 0.1994, "step": 7478 }, { "epoch": 1.35, "learning_rate": 9.724380986806435e-06, "loss": 1.1062, "step": 7479 }, { "epoch": 1.35, "learning_rate": 9.721669980119283e-06, "loss": 0.9868, "step": 7480 }, { "epoch": 1.35, "learning_rate": 9.718958973432135e-06, "loss": 0.4553, "step": 7481 }, { "epoch": 1.35, "learning_rate": 9.716247966744985e-06, "loss": 1.4773, "step": 7482 }, { "epoch": 1.35, "learning_rate": 9.713536960057835e-06, "loss": 1.219, "step": 7483 }, { "epoch": 1.35, "learning_rate": 9.710825953370685e-06, "loss": 1.2576, "step": 7484 }, { "epoch": 1.35, "learning_rate": 9.708114946683535e-06, "loss": 0.5212, "step": 7485 }, { "epoch": 1.35, "learning_rate": 9.705403939996385e-06, "loss": 0.4008, "step": 7486 }, { "epoch": 1.35, "learning_rate": 9.702692933309236e-06, "loss": 0.3816, "step": 7487 }, { "epoch": 1.35, "learning_rate": 9.699981926622086e-06, "loss": 0.4191, "step": 7488 }, { "epoch": 1.35, "learning_rate": 9.697270919934936e-06, "loss": 0.7299, "step": 7489 }, { "epoch": 1.35, "learning_rate": 9.694559913247786e-06, "loss": 0.3733, "step": 7490 }, { "epoch": 1.35, "learning_rate": 9.691848906560636e-06, "loss": 1.4099, "step": 7491 }, { "epoch": 1.35, "learning_rate": 9.689137899873486e-06, "loss": 0.9364, "step": 7492 }, { "epoch": 1.35, "learning_rate": 9.686426893186336e-06, "loss": 0.4753, "step": 7493 }, { "epoch": 1.35, "learning_rate": 9.683715886499187e-06, "loss": 0.9496, "step": 7494 }, { "epoch": 1.35, "learning_rate": 9.681004879812038e-06, "loss": 0.5614, "step": 7495 }, { "epoch": 1.35, "learning_rate": 9.678293873124887e-06, "loss": 0.5883, "step": 7496 }, { "epoch": 1.35, "learning_rate": 9.675582866437737e-06, "loss": 0.4759, "step": 7497 }, { "epoch": 1.36, "learning_rate": 9.672871859750587e-06, "loss": 0.6042, "step": 7498 }, { "epoch": 1.36, "learning_rate": 9.670160853063439e-06, "loss": 0.5148, "step": 7499 }, { "epoch": 1.36, "learning_rate": 9.667449846376287e-06, "loss": 0.8909, "step": 7500 }, { "epoch": 1.36, "learning_rate": 9.664738839689137e-06, "loss": 1.0063, "step": 7501 }, { "epoch": 1.36, "learning_rate": 9.662027833001988e-06, "loss": 0.7363, "step": 7502 }, { "epoch": 1.36, "learning_rate": 9.65931682631484e-06, "loss": 0.8027, "step": 7503 }, { "epoch": 1.36, "learning_rate": 9.656605819627688e-06, "loss": 1.073, "step": 7504 }, { "epoch": 1.36, "learning_rate": 9.653894812940538e-06, "loss": 0.8916, "step": 7505 }, { "epoch": 1.36, "learning_rate": 9.65118380625339e-06, "loss": 0.5188, "step": 7506 }, { "epoch": 1.36, "learning_rate": 9.64847279956624e-06, "loss": 0.7982, "step": 7507 }, { "epoch": 1.36, "learning_rate": 9.645761792879088e-06, "loss": 0.6558, "step": 7508 }, { "epoch": 1.36, "learning_rate": 9.643050786191939e-06, "loss": 0.272, "step": 7509 }, { "epoch": 1.36, "learning_rate": 9.64033977950479e-06, "loss": 0.9294, "step": 7510 }, { "epoch": 1.36, "learning_rate": 9.63762877281764e-06, "loss": 0.3927, "step": 7511 }, { "epoch": 1.36, "learning_rate": 9.634917766130489e-06, "loss": 0.6394, "step": 7512 }, { "epoch": 1.36, "learning_rate": 9.63220675944334e-06, "loss": 0.5851, "step": 7513 }, { "epoch": 1.36, "learning_rate": 9.629495752756191e-06, "loss": 0.4671, "step": 7514 }, { "epoch": 1.36, "learning_rate": 9.626784746069041e-06, "loss": 0.5094, "step": 7515 }, { "epoch": 1.36, "learning_rate": 9.62407373938189e-06, "loss": 0.7112, "step": 7516 }, { "epoch": 1.36, "learning_rate": 9.621362732694741e-06, "loss": 1.1395, "step": 7517 }, { "epoch": 1.36, "learning_rate": 9.618651726007591e-06, "loss": 0.5003, "step": 7518 }, { "epoch": 1.36, "learning_rate": 9.615940719320442e-06, "loss": 0.7251, "step": 7519 }, { "epoch": 1.36, "learning_rate": 9.61322971263329e-06, "loss": 0.609, "step": 7520 }, { "epoch": 1.36, "learning_rate": 9.610518705946142e-06, "loss": 0.7326, "step": 7521 }, { "epoch": 1.36, "learning_rate": 9.607807699258992e-06, "loss": 0.8627, "step": 7522 }, { "epoch": 1.36, "learning_rate": 9.605096692571842e-06, "loss": 0.951, "step": 7523 }, { "epoch": 1.36, "learning_rate": 9.602385685884692e-06, "loss": 0.7346, "step": 7524 }, { "epoch": 1.36, "learning_rate": 9.599674679197542e-06, "loss": 1.1065, "step": 7525 }, { "epoch": 1.36, "learning_rate": 9.596963672510393e-06, "loss": 1.5888, "step": 7526 }, { "epoch": 1.36, "learning_rate": 9.594252665823243e-06, "loss": 0.7644, "step": 7527 }, { "epoch": 1.36, "learning_rate": 9.591541659136093e-06, "loss": 0.45, "step": 7528 }, { "epoch": 1.36, "learning_rate": 9.588830652448943e-06, "loss": 0.984, "step": 7529 }, { "epoch": 1.36, "learning_rate": 9.586119645761793e-06, "loss": 0.7921, "step": 7530 }, { "epoch": 1.36, "learning_rate": 9.583408639074643e-06, "loss": 0.7304, "step": 7531 }, { "epoch": 1.36, "learning_rate": 9.580697632387493e-06, "loss": 0.7409, "step": 7532 }, { "epoch": 1.36, "learning_rate": 9.577986625700343e-06, "loss": 0.7294, "step": 7533 }, { "epoch": 1.36, "learning_rate": 9.575275619013194e-06, "loss": 0.7038, "step": 7534 }, { "epoch": 1.36, "learning_rate": 9.572564612326045e-06, "loss": 0.5271, "step": 7535 }, { "epoch": 1.36, "learning_rate": 9.569853605638894e-06, "loss": 0.7786, "step": 7536 }, { "epoch": 1.36, "learning_rate": 9.567142598951744e-06, "loss": 0.5628, "step": 7537 }, { "epoch": 1.36, "learning_rate": 9.564431592264594e-06, "loss": 0.9424, "step": 7538 }, { "epoch": 1.36, "learning_rate": 9.561720585577446e-06, "loss": 0.6744, "step": 7539 }, { "epoch": 1.36, "learning_rate": 9.559009578890294e-06, "loss": 0.802, "step": 7540 }, { "epoch": 1.36, "learning_rate": 9.556298572203145e-06, "loss": 0.2983, "step": 7541 }, { "epoch": 1.36, "learning_rate": 9.553587565515995e-06, "loss": 0.516, "step": 7542 }, { "epoch": 1.36, "learning_rate": 9.550876558828847e-06, "loss": 0.4636, "step": 7543 }, { "epoch": 1.36, "learning_rate": 9.548165552141695e-06, "loss": 0.8475, "step": 7544 }, { "epoch": 1.36, "learning_rate": 9.545454545454545e-06, "loss": 1.0413, "step": 7545 }, { "epoch": 1.36, "learning_rate": 9.542743538767397e-06, "loss": 0.9055, "step": 7546 }, { "epoch": 1.36, "learning_rate": 9.540032532080247e-06, "loss": 0.7543, "step": 7547 }, { "epoch": 1.36, "learning_rate": 9.537321525393095e-06, "loss": 0.9586, "step": 7548 }, { "epoch": 1.36, "learning_rate": 9.534610518705946e-06, "loss": 1.301, "step": 7549 }, { "epoch": 1.36, "learning_rate": 9.531899512018797e-06, "loss": 1.1415, "step": 7550 }, { "epoch": 1.36, "learning_rate": 9.529188505331648e-06, "loss": 0.7625, "step": 7551 }, { "epoch": 1.36, "learning_rate": 9.526477498644496e-06, "loss": 0.4903, "step": 7552 }, { "epoch": 1.37, "learning_rate": 9.523766491957346e-06, "loss": 0.9156, "step": 7553 }, { "epoch": 1.37, "learning_rate": 9.521055485270198e-06, "loss": 1.025, "step": 7554 }, { "epoch": 1.37, "learning_rate": 9.518344478583048e-06, "loss": 1.054, "step": 7555 }, { "epoch": 1.37, "learning_rate": 9.515633471895897e-06, "loss": 0.5766, "step": 7556 }, { "epoch": 1.37, "learning_rate": 9.512922465208748e-06, "loss": 0.5836, "step": 7557 }, { "epoch": 1.37, "learning_rate": 9.510211458521599e-06, "loss": 0.7723, "step": 7558 }, { "epoch": 1.37, "learning_rate": 9.507500451834449e-06, "loss": 0.892, "step": 7559 }, { "epoch": 1.37, "learning_rate": 9.504789445147297e-06, "loss": 0.5856, "step": 7560 }, { "epoch": 1.37, "learning_rate": 9.502078438460149e-06, "loss": 1.3629, "step": 7561 }, { "epoch": 1.37, "learning_rate": 9.499367431772999e-06, "loss": 0.7752, "step": 7562 }, { "epoch": 1.37, "learning_rate": 9.49665642508585e-06, "loss": 0.5183, "step": 7563 }, { "epoch": 1.37, "learning_rate": 9.493945418398698e-06, "loss": 1.2523, "step": 7564 }, { "epoch": 1.37, "learning_rate": 9.49123441171155e-06, "loss": 0.9689, "step": 7565 }, { "epoch": 1.37, "learning_rate": 9.4885234050244e-06, "loss": 1.1885, "step": 7566 }, { "epoch": 1.37, "learning_rate": 9.48581239833725e-06, "loss": 0.4217, "step": 7567 }, { "epoch": 1.37, "learning_rate": 9.4831013916501e-06, "loss": 0.7725, "step": 7568 }, { "epoch": 1.37, "learning_rate": 9.48039038496295e-06, "loss": 0.7745, "step": 7569 }, { "epoch": 1.37, "learning_rate": 9.4776793782758e-06, "loss": 1.0021, "step": 7570 }, { "epoch": 1.37, "learning_rate": 9.47496837158865e-06, "loss": 0.6021, "step": 7571 }, { "epoch": 1.37, "learning_rate": 9.4722573649015e-06, "loss": 1.2583, "step": 7572 }, { "epoch": 1.37, "learning_rate": 9.46954635821435e-06, "loss": 0.4528, "step": 7573 }, { "epoch": 1.37, "learning_rate": 9.4668353515272e-06, "loss": 0.7105, "step": 7574 }, { "epoch": 1.37, "learning_rate": 9.46412434484005e-06, "loss": 0.5873, "step": 7575 }, { "epoch": 1.37, "learning_rate": 9.461413338152901e-06, "loss": 1.2718, "step": 7576 }, { "epoch": 1.37, "learning_rate": 9.458702331465751e-06, "loss": 1.682, "step": 7577 }, { "epoch": 1.37, "learning_rate": 9.455991324778601e-06, "loss": 0.7744, "step": 7578 }, { "epoch": 1.37, "learning_rate": 9.453280318091453e-06, "loss": 0.7547, "step": 7579 }, { "epoch": 1.37, "learning_rate": 9.450569311404301e-06, "loss": 0.65, "step": 7580 }, { "epoch": 1.37, "learning_rate": 9.447858304717152e-06, "loss": 0.3903, "step": 7581 }, { "epoch": 1.37, "learning_rate": 9.445147298030002e-06, "loss": 0.95, "step": 7582 }, { "epoch": 1.37, "learning_rate": 9.442436291342854e-06, "loss": 1.0885, "step": 7583 }, { "epoch": 1.37, "learning_rate": 9.439725284655702e-06, "loss": 0.4694, "step": 7584 }, { "epoch": 1.37, "learning_rate": 9.437014277968552e-06, "loss": 0.8483, "step": 7585 }, { "epoch": 1.37, "learning_rate": 9.434303271281404e-06, "loss": 0.6125, "step": 7586 }, { "epoch": 1.37, "learning_rate": 9.431592264594254e-06, "loss": 1.1123, "step": 7587 }, { "epoch": 1.37, "learning_rate": 9.428881257907103e-06, "loss": 0.6633, "step": 7588 }, { "epoch": 1.37, "learning_rate": 9.426170251219953e-06, "loss": 0.9778, "step": 7589 }, { "epoch": 1.37, "learning_rate": 9.423459244532805e-06, "loss": 1.1371, "step": 7590 }, { "epoch": 1.37, "learning_rate": 9.420748237845655e-06, "loss": 0.8013, "step": 7591 }, { "epoch": 1.37, "learning_rate": 9.418037231158503e-06, "loss": 0.4071, "step": 7592 }, { "epoch": 1.37, "learning_rate": 9.415326224471353e-06, "loss": 0.6538, "step": 7593 }, { "epoch": 1.37, "learning_rate": 9.412615217784205e-06, "loss": 0.898, "step": 7594 }, { "epoch": 1.37, "learning_rate": 9.409904211097053e-06, "loss": 0.8677, "step": 7595 }, { "epoch": 1.37, "learning_rate": 9.407193204409904e-06, "loss": 0.9222, "step": 7596 }, { "epoch": 1.37, "learning_rate": 9.404482197722755e-06, "loss": 0.5672, "step": 7597 }, { "epoch": 1.37, "learning_rate": 9.401771191035606e-06, "loss": 0.3884, "step": 7598 }, { "epoch": 1.37, "learning_rate": 9.399060184348454e-06, "loss": 0.7195, "step": 7599 }, { "epoch": 1.37, "learning_rate": 9.396349177661304e-06, "loss": 0.6682, "step": 7600 }, { "epoch": 1.37, "learning_rate": 9.393638170974156e-06, "loss": 0.9906, "step": 7601 }, { "epoch": 1.37, "learning_rate": 9.390927164287006e-06, "loss": 0.4357, "step": 7602 }, { "epoch": 1.37, "learning_rate": 9.388216157599855e-06, "loss": 0.4259, "step": 7603 }, { "epoch": 1.37, "learning_rate": 9.385505150912705e-06, "loss": 0.3952, "step": 7604 }, { "epoch": 1.37, "learning_rate": 9.382794144225557e-06, "loss": 0.2353, "step": 7605 }, { "epoch": 1.37, "learning_rate": 9.380083137538407e-06, "loss": 1.4491, "step": 7606 }, { "epoch": 1.37, "learning_rate": 9.377372130851255e-06, "loss": 0.4996, "step": 7607 }, { "epoch": 1.38, "learning_rate": 9.374661124164107e-06, "loss": 1.1753, "step": 7608 }, { "epoch": 1.38, "learning_rate": 9.371950117476957e-06, "loss": 0.6262, "step": 7609 }, { "epoch": 1.38, "learning_rate": 9.369239110789807e-06, "loss": 0.9721, "step": 7610 }, { "epoch": 1.38, "learning_rate": 9.366528104102656e-06, "loss": 1.2643, "step": 7611 }, { "epoch": 1.38, "learning_rate": 9.363817097415507e-06, "loss": 0.4976, "step": 7612 }, { "epoch": 1.38, "learning_rate": 9.361106090728358e-06, "loss": 0.736, "step": 7613 }, { "epoch": 1.38, "learning_rate": 9.358395084041208e-06, "loss": 0.8335, "step": 7614 }, { "epoch": 1.38, "learning_rate": 9.355684077354056e-06, "loss": 0.8562, "step": 7615 }, { "epoch": 1.38, "learning_rate": 9.352973070666908e-06, "loss": 1.0362, "step": 7616 }, { "epoch": 1.38, "learning_rate": 9.350262063979758e-06, "loss": 0.8527, "step": 7617 }, { "epoch": 1.38, "learning_rate": 9.347551057292608e-06, "loss": 0.7404, "step": 7618 }, { "epoch": 1.38, "learning_rate": 9.344840050605458e-06, "loss": 0.8901, "step": 7619 }, { "epoch": 1.38, "learning_rate": 9.342129043918309e-06, "loss": 1.1493, "step": 7620 }, { "epoch": 1.38, "learning_rate": 9.339418037231159e-06, "loss": 0.4282, "step": 7621 }, { "epoch": 1.38, "learning_rate": 9.336707030544009e-06, "loss": 1.5558, "step": 7622 }, { "epoch": 1.38, "learning_rate": 9.333996023856859e-06, "loss": 0.9636, "step": 7623 }, { "epoch": 1.38, "learning_rate": 9.331285017169709e-06, "loss": 0.6224, "step": 7624 }, { "epoch": 1.38, "learning_rate": 9.32857401048256e-06, "loss": 0.5464, "step": 7625 }, { "epoch": 1.38, "learning_rate": 9.32586300379541e-06, "loss": 0.4813, "step": 7626 }, { "epoch": 1.38, "learning_rate": 9.32315199710826e-06, "loss": 0.9267, "step": 7627 }, { "epoch": 1.38, "learning_rate": 9.32044099042111e-06, "loss": 0.7691, "step": 7628 }, { "epoch": 1.38, "learning_rate": 9.31772998373396e-06, "loss": 0.6783, "step": 7629 }, { "epoch": 1.38, "learning_rate": 9.315018977046812e-06, "loss": 0.8425, "step": 7630 }, { "epoch": 1.38, "learning_rate": 9.31230797035966e-06, "loss": 0.7734, "step": 7631 }, { "epoch": 1.38, "learning_rate": 9.30959696367251e-06, "loss": 1.0502, "step": 7632 }, { "epoch": 1.38, "learning_rate": 9.30688595698536e-06, "loss": 0.6539, "step": 7633 }, { "epoch": 1.38, "learning_rate": 9.304174950298212e-06, "loss": 1.0851, "step": 7634 }, { "epoch": 1.38, "learning_rate": 9.30146394361106e-06, "loss": 0.7151, "step": 7635 }, { "epoch": 1.38, "learning_rate": 9.29875293692391e-06, "loss": 0.5287, "step": 7636 }, { "epoch": 1.38, "learning_rate": 9.29604193023676e-06, "loss": 0.6745, "step": 7637 }, { "epoch": 1.38, "learning_rate": 9.293330923549613e-06, "loss": 0.8131, "step": 7638 }, { "epoch": 1.38, "learning_rate": 9.290619916862461e-06, "loss": 0.6377, "step": 7639 }, { "epoch": 1.38, "learning_rate": 9.287908910175311e-06, "loss": 0.3432, "step": 7640 }, { "epoch": 1.38, "learning_rate": 9.285197903488163e-06, "loss": 0.5982, "step": 7641 }, { "epoch": 1.38, "learning_rate": 9.282486896801013e-06, "loss": 0.6758, "step": 7642 }, { "epoch": 1.38, "learning_rate": 9.279775890113862e-06, "loss": 0.8152, "step": 7643 }, { "epoch": 1.38, "learning_rate": 9.277064883426712e-06, "loss": 0.5692, "step": 7644 }, { "epoch": 1.38, "learning_rate": 9.274353876739564e-06, "loss": 1.3164, "step": 7645 }, { "epoch": 1.38, "learning_rate": 9.271642870052414e-06, "loss": 1.3846, "step": 7646 }, { "epoch": 1.38, "learning_rate": 9.268931863365262e-06, "loss": 0.422, "step": 7647 }, { "epoch": 1.38, "learning_rate": 9.266220856678112e-06, "loss": 0.693, "step": 7648 }, { "epoch": 1.38, "learning_rate": 9.263509849990964e-06, "loss": 0.6107, "step": 7649 }, { "epoch": 1.38, "learning_rate": 9.260798843303814e-06, "loss": 0.5498, "step": 7650 }, { "epoch": 1.38, "learning_rate": 9.258087836616663e-06, "loss": 0.6853, "step": 7651 }, { "epoch": 1.38, "learning_rate": 9.255376829929515e-06, "loss": 0.4098, "step": 7652 }, { "epoch": 1.38, "learning_rate": 9.252665823242365e-06, "loss": 0.5124, "step": 7653 }, { "epoch": 1.38, "learning_rate": 9.249954816555215e-06, "loss": 0.8236, "step": 7654 }, { "epoch": 1.38, "learning_rate": 9.247243809868063e-06, "loss": 0.4374, "step": 7655 }, { "epoch": 1.38, "learning_rate": 9.244532803180915e-06, "loss": 0.6336, "step": 7656 }, { "epoch": 1.38, "learning_rate": 9.241821796493765e-06, "loss": 1.0219, "step": 7657 }, { "epoch": 1.38, "learning_rate": 9.239110789806615e-06, "loss": 0.6245, "step": 7658 }, { "epoch": 1.38, "learning_rate": 9.236399783119465e-06, "loss": 0.5964, "step": 7659 }, { "epoch": 1.38, "learning_rate": 9.233688776432316e-06, "loss": 0.756, "step": 7660 }, { "epoch": 1.38, "learning_rate": 9.230977769745166e-06, "loss": 0.6138, "step": 7661 }, { "epoch": 1.38, "learning_rate": 9.228266763058016e-06, "loss": 0.9962, "step": 7662 }, { "epoch": 1.38, "learning_rate": 9.225555756370866e-06, "loss": 1.1396, "step": 7663 }, { "epoch": 1.39, "learning_rate": 9.222844749683716e-06, "loss": 0.2669, "step": 7664 }, { "epoch": 1.39, "learning_rate": 9.220133742996566e-06, "loss": 1.0401, "step": 7665 }, { "epoch": 1.39, "learning_rate": 9.217422736309416e-06, "loss": 0.5011, "step": 7666 }, { "epoch": 1.39, "learning_rate": 9.214711729622267e-06, "loss": 0.6607, "step": 7667 }, { "epoch": 1.39, "learning_rate": 9.212000722935117e-06, "loss": 0.8345, "step": 7668 }, { "epoch": 1.39, "learning_rate": 9.209289716247967e-06, "loss": 1.1276, "step": 7669 }, { "epoch": 1.39, "learning_rate": 9.206578709560819e-06, "loss": 0.4786, "step": 7670 }, { "epoch": 1.39, "learning_rate": 9.203867702873667e-06, "loss": 1.7071, "step": 7671 }, { "epoch": 1.39, "learning_rate": 9.201156696186517e-06, "loss": 1.0664, "step": 7672 }, { "epoch": 1.39, "learning_rate": 9.198445689499367e-06, "loss": 0.365, "step": 7673 }, { "epoch": 1.39, "learning_rate": 9.19573468281222e-06, "loss": 0.8267, "step": 7674 }, { "epoch": 1.39, "learning_rate": 9.193023676125068e-06, "loss": 0.8004, "step": 7675 }, { "epoch": 1.39, "learning_rate": 9.190312669437918e-06, "loss": 0.4534, "step": 7676 }, { "epoch": 1.39, "learning_rate": 9.187601662750768e-06, "loss": 0.538, "step": 7677 }, { "epoch": 1.39, "learning_rate": 9.18489065606362e-06, "loss": 1.1195, "step": 7678 }, { "epoch": 1.39, "learning_rate": 9.182179649376468e-06, "loss": 1.044, "step": 7679 }, { "epoch": 1.39, "learning_rate": 9.179468642689318e-06, "loss": 0.4678, "step": 7680 }, { "epoch": 1.39, "learning_rate": 9.17675763600217e-06, "loss": 0.5111, "step": 7681 }, { "epoch": 1.39, "learning_rate": 9.17404662931502e-06, "loss": 0.3776, "step": 7682 }, { "epoch": 1.39, "learning_rate": 9.171335622627869e-06, "loss": 1.3588, "step": 7683 }, { "epoch": 1.39, "learning_rate": 9.168624615940719e-06, "loss": 0.624, "step": 7684 }, { "epoch": 1.39, "learning_rate": 9.16591360925357e-06, "loss": 0.6591, "step": 7685 }, { "epoch": 1.39, "learning_rate": 9.16320260256642e-06, "loss": 1.5658, "step": 7686 }, { "epoch": 1.39, "learning_rate": 9.16049159587927e-06, "loss": 0.5702, "step": 7687 }, { "epoch": 1.39, "learning_rate": 9.15778058919212e-06, "loss": 0.7457, "step": 7688 }, { "epoch": 1.39, "learning_rate": 9.155069582504971e-06, "loss": 0.312, "step": 7689 }, { "epoch": 1.39, "learning_rate": 9.152358575817821e-06, "loss": 0.4547, "step": 7690 }, { "epoch": 1.39, "learning_rate": 9.14964756913067e-06, "loss": 0.5604, "step": 7691 }, { "epoch": 1.39, "learning_rate": 9.146936562443522e-06, "loss": 0.7341, "step": 7692 }, { "epoch": 1.39, "learning_rate": 9.144225555756372e-06, "loss": 0.8627, "step": 7693 }, { "epoch": 1.39, "learning_rate": 9.141514549069222e-06, "loss": 0.7394, "step": 7694 }, { "epoch": 1.39, "learning_rate": 9.13880354238207e-06, "loss": 0.577, "step": 7695 }, { "epoch": 1.39, "learning_rate": 9.136092535694922e-06, "loss": 0.5579, "step": 7696 }, { "epoch": 1.39, "learning_rate": 9.133381529007772e-06, "loss": 1.2796, "step": 7697 }, { "epoch": 1.39, "learning_rate": 9.130670522320622e-06, "loss": 0.5049, "step": 7698 }, { "epoch": 1.39, "learning_rate": 9.12795951563347e-06, "loss": 0.3448, "step": 7699 }, { "epoch": 1.39, "learning_rate": 9.125248508946323e-06, "loss": 0.5303, "step": 7700 }, { "epoch": 1.39, "learning_rate": 9.122537502259173e-06, "loss": 0.8504, "step": 7701 }, { "epoch": 1.39, "learning_rate": 9.119826495572023e-06, "loss": 0.8693, "step": 7702 }, { "epoch": 1.39, "learning_rate": 9.117115488884873e-06, "loss": 0.7966, "step": 7703 }, { "epoch": 1.39, "learning_rate": 9.114404482197723e-06, "loss": 0.7876, "step": 7704 }, { "epoch": 1.39, "learning_rate": 9.111693475510573e-06, "loss": 0.9455, "step": 7705 }, { "epoch": 1.39, "learning_rate": 9.108982468823423e-06, "loss": 0.46, "step": 7706 }, { "epoch": 1.39, "learning_rate": 9.106271462136274e-06, "loss": 1.2737, "step": 7707 }, { "epoch": 1.39, "learning_rate": 9.103560455449124e-06, "loss": 0.5929, "step": 7708 }, { "epoch": 1.39, "learning_rate": 9.100849448761974e-06, "loss": 0.9292, "step": 7709 }, { "epoch": 1.39, "learning_rate": 9.098138442074824e-06, "loss": 0.7327, "step": 7710 }, { "epoch": 1.39, "learning_rate": 9.095427435387674e-06, "loss": 0.3377, "step": 7711 }, { "epoch": 1.39, "learning_rate": 9.092716428700524e-06, "loss": 0.2392, "step": 7712 }, { "epoch": 1.39, "learning_rate": 9.090005422013374e-06, "loss": 1.267, "step": 7713 }, { "epoch": 1.39, "learning_rate": 9.087294415326226e-06, "loss": 0.9004, "step": 7714 }, { "epoch": 1.39, "learning_rate": 9.084583408639075e-06, "loss": 1.0837, "step": 7715 }, { "epoch": 1.39, "learning_rate": 9.081872401951925e-06, "loss": 0.6504, "step": 7716 }, { "epoch": 1.39, "learning_rate": 9.079161395264775e-06, "loss": 0.5802, "step": 7717 }, { "epoch": 1.39, "learning_rate": 9.076450388577627e-06, "loss": 0.8435, "step": 7718 }, { "epoch": 1.4, "learning_rate": 9.073739381890475e-06, "loss": 0.73, "step": 7719 }, { "epoch": 1.4, "learning_rate": 9.071028375203325e-06, "loss": 0.6607, "step": 7720 }, { "epoch": 1.4, "learning_rate": 9.068317368516175e-06, "loss": 0.3868, "step": 7721 }, { "epoch": 1.4, "learning_rate": 9.065606361829026e-06, "loss": 1.2318, "step": 7722 }, { "epoch": 1.4, "learning_rate": 9.062895355141876e-06, "loss": 0.4026, "step": 7723 }, { "epoch": 1.4, "learning_rate": 9.060184348454726e-06, "loss": 0.8977, "step": 7724 }, { "epoch": 1.4, "learning_rate": 9.057473341767578e-06, "loss": 1.2593, "step": 7725 }, { "epoch": 1.4, "learning_rate": 9.054762335080426e-06, "loss": 0.7869, "step": 7726 }, { "epoch": 1.4, "learning_rate": 9.052051328393276e-06, "loss": 0.5969, "step": 7727 }, { "epoch": 1.4, "learning_rate": 9.049340321706126e-06, "loss": 1.0934, "step": 7728 }, { "epoch": 1.4, "learning_rate": 9.046629315018978e-06, "loss": 0.7693, "step": 7729 }, { "epoch": 1.4, "learning_rate": 9.043918308331827e-06, "loss": 1.1636, "step": 7730 }, { "epoch": 1.4, "learning_rate": 9.041207301644677e-06, "loss": 0.6967, "step": 7731 }, { "epoch": 1.4, "learning_rate": 9.038496294957529e-06, "loss": 0.6274, "step": 7732 }, { "epoch": 1.4, "learning_rate": 9.035785288270379e-06, "loss": 0.6225, "step": 7733 }, { "epoch": 1.4, "learning_rate": 9.033074281583227e-06, "loss": 0.5823, "step": 7734 }, { "epoch": 1.4, "learning_rate": 9.030363274896077e-06, "loss": 0.4436, "step": 7735 }, { "epoch": 1.4, "learning_rate": 9.02765226820893e-06, "loss": 1.1646, "step": 7736 }, { "epoch": 1.4, "learning_rate": 9.02494126152178e-06, "loss": 0.5113, "step": 7737 }, { "epoch": 1.4, "learning_rate": 9.022230254834628e-06, "loss": 0.9221, "step": 7738 }, { "epoch": 1.4, "learning_rate": 9.019519248147478e-06, "loss": 1.4932, "step": 7739 }, { "epoch": 1.4, "learning_rate": 9.01680824146033e-06, "loss": 0.4017, "step": 7740 }, { "epoch": 1.4, "learning_rate": 9.01409723477318e-06, "loss": 0.7083, "step": 7741 }, { "epoch": 1.4, "learning_rate": 9.011386228086028e-06, "loss": 0.5244, "step": 7742 }, { "epoch": 1.4, "learning_rate": 9.00867522139888e-06, "loss": 0.5264, "step": 7743 }, { "epoch": 1.4, "learning_rate": 9.00596421471173e-06, "loss": 0.5224, "step": 7744 }, { "epoch": 1.4, "learning_rate": 9.00325320802458e-06, "loss": 0.5479, "step": 7745 }, { "epoch": 1.4, "learning_rate": 9.000542201337429e-06, "loss": 0.403, "step": 7746 }, { "epoch": 1.4, "learning_rate": 8.99783119465028e-06, "loss": 0.5372, "step": 7747 }, { "epoch": 1.4, "learning_rate": 8.99512018796313e-06, "loss": 0.5283, "step": 7748 }, { "epoch": 1.4, "learning_rate": 8.992409181275981e-06, "loss": 0.5517, "step": 7749 }, { "epoch": 1.4, "learning_rate": 8.98969817458883e-06, "loss": 0.4129, "step": 7750 }, { "epoch": 1.4, "learning_rate": 8.986987167901681e-06, "loss": 1.1471, "step": 7751 }, { "epoch": 1.4, "learning_rate": 8.984276161214531e-06, "loss": 0.6911, "step": 7752 }, { "epoch": 1.4, "learning_rate": 8.981565154527381e-06, "loss": 0.5543, "step": 7753 }, { "epoch": 1.4, "learning_rate": 8.978854147840232e-06, "loss": 0.9668, "step": 7754 }, { "epoch": 1.4, "learning_rate": 8.976143141153082e-06, "loss": 0.6657, "step": 7755 }, { "epoch": 1.4, "learning_rate": 8.973432134465932e-06, "loss": 0.4236, "step": 7756 }, { "epoch": 1.4, "learning_rate": 8.970721127778782e-06, "loss": 0.6293, "step": 7757 }, { "epoch": 1.4, "learning_rate": 8.968010121091632e-06, "loss": 0.8967, "step": 7758 }, { "epoch": 1.4, "learning_rate": 8.965299114404482e-06, "loss": 0.8678, "step": 7759 }, { "epoch": 1.4, "learning_rate": 8.962588107717332e-06, "loss": 0.3617, "step": 7760 }, { "epoch": 1.4, "learning_rate": 8.959877101030183e-06, "loss": 0.9478, "step": 7761 }, { "epoch": 1.4, "learning_rate": 8.957166094343033e-06, "loss": 0.876, "step": 7762 }, { "epoch": 1.4, "learning_rate": 8.954455087655883e-06, "loss": 1.255, "step": 7763 }, { "epoch": 1.4, "learning_rate": 8.951744080968733e-06, "loss": 0.3458, "step": 7764 }, { "epoch": 1.4, "learning_rate": 8.949033074281585e-06, "loss": 0.9163, "step": 7765 }, { "epoch": 1.4, "learning_rate": 8.946322067594433e-06, "loss": 1.0375, "step": 7766 }, { "epoch": 1.4, "learning_rate": 8.943611060907283e-06, "loss": 0.8946, "step": 7767 }, { "epoch": 1.4, "learning_rate": 8.940900054220133e-06, "loss": 0.8386, "step": 7768 }, { "epoch": 1.4, "learning_rate": 8.938189047532985e-06, "loss": 0.9836, "step": 7769 }, { "epoch": 1.4, "learning_rate": 8.935478040845834e-06, "loss": 0.4832, "step": 7770 }, { "epoch": 1.4, "learning_rate": 8.932767034158684e-06, "loss": 0.7125, "step": 7771 }, { "epoch": 1.4, "learning_rate": 8.930056027471534e-06, "loss": 1.3593, "step": 7772 }, { "epoch": 1.4, "learning_rate": 8.927345020784386e-06, "loss": 1.0481, "step": 7773 }, { "epoch": 1.41, "learning_rate": 8.924634014097234e-06, "loss": 1.456, "step": 7774 }, { "epoch": 1.41, "learning_rate": 8.921923007410084e-06, "loss": 0.9012, "step": 7775 }, { "epoch": 1.41, "learning_rate": 8.919212000722936e-06, "loss": 0.3924, "step": 7776 }, { "epoch": 1.41, "learning_rate": 8.916500994035786e-06, "loss": 0.7306, "step": 7777 }, { "epoch": 1.41, "learning_rate": 8.913789987348635e-06, "loss": 0.432, "step": 7778 }, { "epoch": 1.41, "learning_rate": 8.911078980661485e-06, "loss": 0.363, "step": 7779 }, { "epoch": 1.41, "learning_rate": 8.908367973974337e-06, "loss": 0.5297, "step": 7780 }, { "epoch": 1.41, "learning_rate": 8.905656967287187e-06, "loss": 0.8151, "step": 7781 }, { "epoch": 1.41, "learning_rate": 8.902945960600035e-06, "loss": 0.5638, "step": 7782 }, { "epoch": 1.41, "learning_rate": 8.900234953912886e-06, "loss": 1.2914, "step": 7783 }, { "epoch": 1.41, "learning_rate": 8.897523947225737e-06, "loss": 1.2291, "step": 7784 }, { "epoch": 1.41, "learning_rate": 8.894812940538587e-06, "loss": 0.8737, "step": 7785 }, { "epoch": 1.41, "learning_rate": 8.892101933851436e-06, "loss": 0.5395, "step": 7786 }, { "epoch": 1.41, "learning_rate": 8.889390927164288e-06, "loss": 0.954, "step": 7787 }, { "epoch": 1.41, "learning_rate": 8.886679920477138e-06, "loss": 0.562, "step": 7788 }, { "epoch": 1.41, "learning_rate": 8.883968913789988e-06, "loss": 0.6874, "step": 7789 }, { "epoch": 1.41, "learning_rate": 8.881257907102836e-06, "loss": 0.5032, "step": 7790 }, { "epoch": 1.41, "learning_rate": 8.878546900415688e-06, "loss": 1.0768, "step": 7791 }, { "epoch": 1.41, "learning_rate": 8.875835893728538e-06, "loss": 0.8397, "step": 7792 }, { "epoch": 1.41, "learning_rate": 8.873124887041389e-06, "loss": 1.0808, "step": 7793 }, { "epoch": 1.41, "learning_rate": 8.870413880354239e-06, "loss": 0.5349, "step": 7794 }, { "epoch": 1.41, "learning_rate": 8.867702873667089e-06, "loss": 1.1304, "step": 7795 }, { "epoch": 1.41, "learning_rate": 8.864991866979939e-06, "loss": 0.9149, "step": 7796 }, { "epoch": 1.41, "learning_rate": 8.862280860292789e-06, "loss": 0.7405, "step": 7797 }, { "epoch": 1.41, "learning_rate": 8.85956985360564e-06, "loss": 0.6289, "step": 7798 }, { "epoch": 1.41, "learning_rate": 8.85685884691849e-06, "loss": 0.9836, "step": 7799 }, { "epoch": 1.41, "learning_rate": 8.85414784023134e-06, "loss": 0.7687, "step": 7800 }, { "epoch": 1.41, "learning_rate": 8.85143683354419e-06, "loss": 0.8629, "step": 7801 }, { "epoch": 1.41, "learning_rate": 8.84872582685704e-06, "loss": 1.2765, "step": 7802 }, { "epoch": 1.41, "learning_rate": 8.84601482016989e-06, "loss": 0.7252, "step": 7803 }, { "epoch": 1.41, "learning_rate": 8.84330381348274e-06, "loss": 0.925, "step": 7804 }, { "epoch": 1.41, "learning_rate": 8.840592806795592e-06, "loss": 0.9136, "step": 7805 }, { "epoch": 1.41, "learning_rate": 8.83788180010844e-06, "loss": 0.7345, "step": 7806 }, { "epoch": 1.41, "learning_rate": 8.83517079342129e-06, "loss": 0.9217, "step": 7807 }, { "epoch": 1.41, "learning_rate": 8.83245978673414e-06, "loss": 0.828, "step": 7808 }, { "epoch": 1.41, "learning_rate": 8.829748780046992e-06, "loss": 0.9363, "step": 7809 }, { "epoch": 1.41, "learning_rate": 8.82703777335984e-06, "loss": 0.5302, "step": 7810 }, { "epoch": 1.41, "learning_rate": 8.824326766672691e-06, "loss": 0.7068, "step": 7811 }, { "epoch": 1.41, "learning_rate": 8.821615759985541e-06, "loss": 0.9144, "step": 7812 }, { "epoch": 1.41, "learning_rate": 8.818904753298393e-06, "loss": 0.6725, "step": 7813 }, { "epoch": 1.41, "learning_rate": 8.816193746611241e-06, "loss": 1.1948, "step": 7814 }, { "epoch": 1.41, "learning_rate": 8.813482739924091e-06, "loss": 0.67, "step": 7815 }, { "epoch": 1.41, "learning_rate": 8.810771733236943e-06, "loss": 1.1126, "step": 7816 }, { "epoch": 1.41, "learning_rate": 8.808060726549793e-06, "loss": 1.1886, "step": 7817 }, { "epoch": 1.41, "learning_rate": 8.805349719862642e-06, "loss": 0.4952, "step": 7818 }, { "epoch": 1.41, "learning_rate": 8.802638713175492e-06, "loss": 1.0208, "step": 7819 }, { "epoch": 1.41, "learning_rate": 8.799927706488344e-06, "loss": 0.2873, "step": 7820 }, { "epoch": 1.41, "learning_rate": 8.797216699801194e-06, "loss": 0.5679, "step": 7821 }, { "epoch": 1.41, "learning_rate": 8.794505693114042e-06, "loss": 0.5376, "step": 7822 }, { "epoch": 1.41, "learning_rate": 8.791794686426893e-06, "loss": 0.8961, "step": 7823 }, { "epoch": 1.41, "learning_rate": 8.789083679739744e-06, "loss": 0.5476, "step": 7824 }, { "epoch": 1.41, "learning_rate": 8.786372673052595e-06, "loss": 0.8635, "step": 7825 }, { "epoch": 1.41, "learning_rate": 8.783661666365443e-06, "loss": 0.7549, "step": 7826 }, { "epoch": 1.41, "learning_rate": 8.780950659678295e-06, "loss": 0.7098, "step": 7827 }, { "epoch": 1.41, "learning_rate": 8.778239652991145e-06, "loss": 0.7573, "step": 7828 }, { "epoch": 1.41, "learning_rate": 8.775528646303995e-06, "loss": 0.6885, "step": 7829 }, { "epoch": 1.42, "learning_rate": 8.772817639616844e-06, "loss": 0.4197, "step": 7830 }, { "epoch": 1.42, "learning_rate": 8.770106632929695e-06, "loss": 0.6567, "step": 7831 }, { "epoch": 1.42, "learning_rate": 8.767395626242545e-06, "loss": 1.0112, "step": 7832 }, { "epoch": 1.42, "learning_rate": 8.764684619555396e-06, "loss": 0.6069, "step": 7833 }, { "epoch": 1.42, "learning_rate": 8.761973612868244e-06, "loss": 1.1824, "step": 7834 }, { "epoch": 1.42, "learning_rate": 8.759262606181096e-06, "loss": 0.4564, "step": 7835 }, { "epoch": 1.42, "learning_rate": 8.756551599493946e-06, "loss": 0.8652, "step": 7836 }, { "epoch": 1.42, "learning_rate": 8.753840592806796e-06, "loss": 1.0299, "step": 7837 }, { "epoch": 1.42, "learning_rate": 8.751129586119646e-06, "loss": 0.9172, "step": 7838 }, { "epoch": 1.42, "learning_rate": 8.748418579432496e-06, "loss": 0.6413, "step": 7839 }, { "epoch": 1.42, "learning_rate": 8.745707572745347e-06, "loss": 0.6046, "step": 7840 }, { "epoch": 1.42, "learning_rate": 8.742996566058197e-06, "loss": 0.8188, "step": 7841 }, { "epoch": 1.42, "learning_rate": 8.740285559371047e-06, "loss": 0.989, "step": 7842 }, { "epoch": 1.42, "learning_rate": 8.737574552683897e-06, "loss": 0.5213, "step": 7843 }, { "epoch": 1.42, "learning_rate": 8.734863545996747e-06, "loss": 0.983, "step": 7844 }, { "epoch": 1.42, "learning_rate": 8.732152539309597e-06, "loss": 0.6833, "step": 7845 }, { "epoch": 1.42, "learning_rate": 8.729441532622447e-06, "loss": 0.9982, "step": 7846 }, { "epoch": 1.42, "learning_rate": 8.726730525935297e-06, "loss": 0.6018, "step": 7847 }, { "epoch": 1.42, "learning_rate": 8.724019519248148e-06, "loss": 0.5743, "step": 7848 }, { "epoch": 1.42, "learning_rate": 8.721308512561e-06, "loss": 1.1385, "step": 7849 }, { "epoch": 1.42, "learning_rate": 8.718597505873848e-06, "loss": 0.3802, "step": 7850 }, { "epoch": 1.42, "learning_rate": 8.715886499186698e-06, "loss": 0.78, "step": 7851 }, { "epoch": 1.42, "learning_rate": 8.713175492499548e-06, "loss": 0.5517, "step": 7852 }, { "epoch": 1.42, "learning_rate": 8.710464485812398e-06, "loss": 0.8821, "step": 7853 }, { "epoch": 1.42, "learning_rate": 8.707753479125248e-06, "loss": 0.867, "step": 7854 }, { "epoch": 1.42, "learning_rate": 8.705042472438099e-06, "loss": 0.6641, "step": 7855 }, { "epoch": 1.42, "learning_rate": 8.702331465750949e-06, "loss": 0.6496, "step": 7856 }, { "epoch": 1.42, "learning_rate": 8.699620459063799e-06, "loss": 0.494, "step": 7857 }, { "epoch": 1.42, "learning_rate": 8.696909452376649e-06, "loss": 0.3421, "step": 7858 }, { "epoch": 1.42, "learning_rate": 8.694198445689499e-06, "loss": 0.5273, "step": 7859 }, { "epoch": 1.42, "learning_rate": 8.691487439002351e-06, "loss": 0.512, "step": 7860 }, { "epoch": 1.42, "learning_rate": 8.6887764323152e-06, "loss": 0.8923, "step": 7861 }, { "epoch": 1.42, "learning_rate": 8.68606542562805e-06, "loss": 0.4999, "step": 7862 }, { "epoch": 1.42, "learning_rate": 8.6833544189409e-06, "loss": 0.5448, "step": 7863 }, { "epoch": 1.42, "learning_rate": 8.680643412253751e-06, "loss": 0.6772, "step": 7864 }, { "epoch": 1.42, "learning_rate": 8.6779324055666e-06, "loss": 0.8443, "step": 7865 }, { "epoch": 1.42, "learning_rate": 8.67522139887945e-06, "loss": 1.1802, "step": 7866 }, { "epoch": 1.42, "learning_rate": 8.6725103921923e-06, "loss": 1.0471, "step": 7867 }, { "epoch": 1.42, "learning_rate": 8.669799385505152e-06, "loss": 0.9145, "step": 7868 }, { "epoch": 1.42, "learning_rate": 8.667088378818e-06, "loss": 0.4264, "step": 7869 }, { "epoch": 1.42, "learning_rate": 8.66437737213085e-06, "loss": 0.7708, "step": 7870 }, { "epoch": 1.42, "learning_rate": 8.661666365443702e-06, "loss": 0.7625, "step": 7871 }, { "epoch": 1.42, "learning_rate": 8.658955358756553e-06, "loss": 0.6051, "step": 7872 }, { "epoch": 1.42, "learning_rate": 8.656244352069401e-06, "loss": 1.2892, "step": 7873 }, { "epoch": 1.42, "learning_rate": 8.653533345382251e-06, "loss": 0.717, "step": 7874 }, { "epoch": 1.42, "learning_rate": 8.650822338695103e-06, "loss": 0.6504, "step": 7875 }, { "epoch": 1.42, "learning_rate": 8.648111332007953e-06, "loss": 0.6169, "step": 7876 }, { "epoch": 1.42, "learning_rate": 8.645400325320802e-06, "loss": 0.7496, "step": 7877 }, { "epoch": 1.42, "learning_rate": 8.642689318633653e-06, "loss": 0.7396, "step": 7878 }, { "epoch": 1.42, "learning_rate": 8.639978311946503e-06, "loss": 1.1832, "step": 7879 }, { "epoch": 1.42, "learning_rate": 8.637267305259354e-06, "loss": 0.7687, "step": 7880 }, { "epoch": 1.42, "learning_rate": 8.634556298572202e-06, "loss": 0.5185, "step": 7881 }, { "epoch": 1.42, "learning_rate": 8.631845291885054e-06, "loss": 0.4892, "step": 7882 }, { "epoch": 1.42, "learning_rate": 8.629134285197904e-06, "loss": 0.9252, "step": 7883 }, { "epoch": 1.42, "learning_rate": 8.626423278510754e-06, "loss": 1.0747, "step": 7884 }, { "epoch": 1.43, "learning_rate": 8.623712271823603e-06, "loss": 0.4029, "step": 7885 }, { "epoch": 1.43, "learning_rate": 8.621001265136454e-06, "loss": 1.1343, "step": 7886 }, { "epoch": 1.43, "learning_rate": 8.618290258449305e-06, "loss": 0.882, "step": 7887 }, { "epoch": 1.43, "learning_rate": 8.615579251762155e-06, "loss": 0.7702, "step": 7888 }, { "epoch": 1.43, "learning_rate": 8.612868245075005e-06, "loss": 0.7618, "step": 7889 }, { "epoch": 1.43, "learning_rate": 8.610157238387855e-06, "loss": 0.7589, "step": 7890 }, { "epoch": 1.43, "learning_rate": 8.607446231700705e-06, "loss": 0.6345, "step": 7891 }, { "epoch": 1.43, "learning_rate": 8.604735225013555e-06, "loss": 1.9347, "step": 7892 }, { "epoch": 1.43, "learning_rate": 8.602024218326405e-06, "loss": 0.5055, "step": 7893 }, { "epoch": 1.43, "learning_rate": 8.599313211639255e-06, "loss": 0.5183, "step": 7894 }, { "epoch": 1.43, "learning_rate": 8.596602204952106e-06, "loss": 0.8486, "step": 7895 }, { "epoch": 1.43, "learning_rate": 8.593891198264956e-06, "loss": 0.7643, "step": 7896 }, { "epoch": 1.43, "learning_rate": 8.591180191577806e-06, "loss": 0.9265, "step": 7897 }, { "epoch": 1.43, "learning_rate": 8.588469184890656e-06, "loss": 0.7024, "step": 7898 }, { "epoch": 1.43, "learning_rate": 8.585758178203506e-06, "loss": 1.0878, "step": 7899 }, { "epoch": 1.43, "learning_rate": 8.583047171516358e-06, "loss": 0.4413, "step": 7900 }, { "epoch": 1.43, "learning_rate": 8.580336164829206e-06, "loss": 0.6243, "step": 7901 }, { "epoch": 1.43, "learning_rate": 8.577625158142057e-06, "loss": 1.1599, "step": 7902 }, { "epoch": 1.43, "learning_rate": 8.574914151454907e-06, "loss": 0.5533, "step": 7903 }, { "epoch": 1.43, "learning_rate": 8.572203144767759e-06, "loss": 0.9817, "step": 7904 }, { "epoch": 1.43, "learning_rate": 8.569492138080607e-06, "loss": 0.4128, "step": 7905 }, { "epoch": 1.43, "learning_rate": 8.566781131393457e-06, "loss": 0.5666, "step": 7906 }, { "epoch": 1.43, "learning_rate": 8.564070124706307e-06, "loss": 0.5893, "step": 7907 }, { "epoch": 1.43, "learning_rate": 8.561359118019159e-06, "loss": 0.6555, "step": 7908 }, { "epoch": 1.43, "learning_rate": 8.558648111332008e-06, "loss": 0.9207, "step": 7909 }, { "epoch": 1.43, "learning_rate": 8.555937104644858e-06, "loss": 0.7565, "step": 7910 }, { "epoch": 1.43, "learning_rate": 8.55322609795771e-06, "loss": 0.9737, "step": 7911 }, { "epoch": 1.43, "learning_rate": 8.55051509127056e-06, "loss": 0.7928, "step": 7912 }, { "epoch": 1.43, "learning_rate": 8.547804084583408e-06, "loss": 0.5413, "step": 7913 }, { "epoch": 1.43, "learning_rate": 8.545093077896258e-06, "loss": 0.5051, "step": 7914 }, { "epoch": 1.43, "learning_rate": 8.54238207120911e-06, "loss": 1.0011, "step": 7915 }, { "epoch": 1.43, "learning_rate": 8.53967106452196e-06, "loss": 0.4157, "step": 7916 }, { "epoch": 1.43, "learning_rate": 8.536960057834809e-06, "loss": 0.883, "step": 7917 }, { "epoch": 1.43, "learning_rate": 8.534249051147659e-06, "loss": 0.3563, "step": 7918 }, { "epoch": 1.43, "learning_rate": 8.53153804446051e-06, "loss": 0.8108, "step": 7919 }, { "epoch": 1.43, "learning_rate": 8.52882703777336e-06, "loss": 0.6694, "step": 7920 }, { "epoch": 1.43, "learning_rate": 8.526116031086209e-06, "loss": 0.7477, "step": 7921 }, { "epoch": 1.43, "learning_rate": 8.523405024399061e-06, "loss": 0.7262, "step": 7922 }, { "epoch": 1.43, "learning_rate": 8.520694017711911e-06, "loss": 1.0015, "step": 7923 }, { "epoch": 1.43, "learning_rate": 8.517983011024761e-06, "loss": 0.7528, "step": 7924 }, { "epoch": 1.43, "learning_rate": 8.51527200433761e-06, "loss": 0.5466, "step": 7925 }, { "epoch": 1.43, "learning_rate": 8.512560997650461e-06, "loss": 0.6371, "step": 7926 }, { "epoch": 1.43, "learning_rate": 8.509849990963312e-06, "loss": 0.8153, "step": 7927 }, { "epoch": 1.43, "learning_rate": 8.507138984276162e-06, "loss": 0.3913, "step": 7928 }, { "epoch": 1.43, "learning_rate": 8.50442797758901e-06, "loss": 0.2268, "step": 7929 }, { "epoch": 1.43, "learning_rate": 8.501716970901862e-06, "loss": 0.9791, "step": 7930 }, { "epoch": 1.43, "learning_rate": 8.499005964214712e-06, "loss": 0.5287, "step": 7931 }, { "epoch": 1.43, "learning_rate": 8.496294957527562e-06, "loss": 0.383, "step": 7932 }, { "epoch": 1.43, "learning_rate": 8.493583950840412e-06, "loss": 1.348, "step": 7933 }, { "epoch": 1.43, "learning_rate": 8.490872944153263e-06, "loss": 0.4886, "step": 7934 }, { "epoch": 1.43, "learning_rate": 8.488161937466113e-06, "loss": 1.3195, "step": 7935 }, { "epoch": 1.43, "learning_rate": 8.485450930778963e-06, "loss": 0.6721, "step": 7936 }, { "epoch": 1.43, "learning_rate": 8.482739924091813e-06, "loss": 0.9531, "step": 7937 }, { "epoch": 1.43, "learning_rate": 8.480028917404663e-06, "loss": 0.5794, "step": 7938 }, { "epoch": 1.43, "learning_rate": 8.477317910717513e-06, "loss": 0.5709, "step": 7939 }, { "epoch": 1.44, "learning_rate": 8.474606904030363e-06, "loss": 0.2856, "step": 7940 }, { "epoch": 1.44, "learning_rate": 8.471895897343213e-06, "loss": 0.4421, "step": 7941 }, { "epoch": 1.44, "learning_rate": 8.469184890656064e-06, "loss": 0.9459, "step": 7942 }, { "epoch": 1.44, "learning_rate": 8.466473883968914e-06, "loss": 0.5614, "step": 7943 }, { "epoch": 1.44, "learning_rate": 8.463762877281766e-06, "loss": 0.4854, "step": 7944 }, { "epoch": 1.44, "learning_rate": 8.461051870594614e-06, "loss": 0.9023, "step": 7945 }, { "epoch": 1.44, "learning_rate": 8.458340863907464e-06, "loss": 1.1129, "step": 7946 }, { "epoch": 1.44, "learning_rate": 8.455629857220314e-06, "loss": 0.7425, "step": 7947 }, { "epoch": 1.44, "learning_rate": 8.452918850533166e-06, "loss": 1.661, "step": 7948 }, { "epoch": 1.44, "learning_rate": 8.450207843846015e-06, "loss": 0.7432, "step": 7949 }, { "epoch": 1.44, "learning_rate": 8.447496837158865e-06, "loss": 0.3599, "step": 7950 }, { "epoch": 1.44, "learning_rate": 8.444785830471717e-06, "loss": 0.3183, "step": 7951 }, { "epoch": 1.44, "learning_rate": 8.442074823784567e-06, "loss": 1.0078, "step": 7952 }, { "epoch": 1.44, "learning_rate": 8.439363817097415e-06, "loss": 0.4438, "step": 7953 }, { "epoch": 1.44, "learning_rate": 8.436652810410265e-06, "loss": 0.7385, "step": 7954 }, { "epoch": 1.44, "learning_rate": 8.433941803723117e-06, "loss": 0.5553, "step": 7955 }, { "epoch": 1.44, "learning_rate": 8.431230797035967e-06, "loss": 0.8928, "step": 7956 }, { "epoch": 1.44, "learning_rate": 8.428519790348816e-06, "loss": 1.4796, "step": 7957 }, { "epoch": 1.44, "learning_rate": 8.425808783661666e-06, "loss": 0.6722, "step": 7958 }, { "epoch": 1.44, "learning_rate": 8.423097776974518e-06, "loss": 0.9029, "step": 7959 }, { "epoch": 1.44, "learning_rate": 8.420386770287368e-06, "loss": 1.0676, "step": 7960 }, { "epoch": 1.44, "learning_rate": 8.417675763600216e-06, "loss": 1.1192, "step": 7961 }, { "epoch": 1.44, "learning_rate": 8.414964756913068e-06, "loss": 0.7194, "step": 7962 }, { "epoch": 1.44, "learning_rate": 8.412253750225918e-06, "loss": 0.7491, "step": 7963 }, { "epoch": 1.44, "learning_rate": 8.409542743538768e-06, "loss": 0.2499, "step": 7964 }, { "epoch": 1.44, "learning_rate": 8.406831736851617e-06, "loss": 0.6886, "step": 7965 }, { "epoch": 1.44, "learning_rate": 8.404120730164469e-06, "loss": 0.4873, "step": 7966 }, { "epoch": 1.44, "learning_rate": 8.401409723477319e-06, "loss": 0.903, "step": 7967 }, { "epoch": 1.44, "learning_rate": 8.398698716790169e-06, "loss": 0.25, "step": 7968 }, { "epoch": 1.44, "learning_rate": 8.395987710103017e-06, "loss": 0.7368, "step": 7969 }, { "epoch": 1.44, "learning_rate": 8.393276703415869e-06, "loss": 0.8572, "step": 7970 }, { "epoch": 1.44, "learning_rate": 8.39056569672872e-06, "loss": 0.6121, "step": 7971 }, { "epoch": 1.44, "learning_rate": 8.38785469004157e-06, "loss": 0.9512, "step": 7972 }, { "epoch": 1.44, "learning_rate": 8.38514368335442e-06, "loss": 0.7775, "step": 7973 }, { "epoch": 1.44, "learning_rate": 8.38243267666727e-06, "loss": 0.6394, "step": 7974 }, { "epoch": 1.44, "learning_rate": 8.37972166998012e-06, "loss": 0.4587, "step": 7975 }, { "epoch": 1.44, "learning_rate": 8.37701066329297e-06, "loss": 0.6746, "step": 7976 }, { "epoch": 1.44, "learning_rate": 8.37429965660582e-06, "loss": 1.1504, "step": 7977 }, { "epoch": 1.44, "learning_rate": 8.37158864991867e-06, "loss": 0.5547, "step": 7978 }, { "epoch": 1.44, "learning_rate": 8.36887764323152e-06, "loss": 1.2446, "step": 7979 }, { "epoch": 1.44, "learning_rate": 8.366166636544369e-06, "loss": 0.8101, "step": 7980 }, { "epoch": 1.44, "learning_rate": 8.36345562985722e-06, "loss": 0.9884, "step": 7981 }, { "epoch": 1.44, "learning_rate": 8.36074462317007e-06, "loss": 0.2551, "step": 7982 }, { "epoch": 1.44, "learning_rate": 8.35803361648292e-06, "loss": 1.1404, "step": 7983 }, { "epoch": 1.44, "learning_rate": 8.355322609795771e-06, "loss": 0.9342, "step": 7984 }, { "epoch": 1.44, "learning_rate": 8.352611603108621e-06, "loss": 0.8623, "step": 7985 }, { "epoch": 1.44, "learning_rate": 8.349900596421471e-06, "loss": 0.8029, "step": 7986 }, { "epoch": 1.44, "learning_rate": 8.347189589734321e-06, "loss": 0.6882, "step": 7987 }, { "epoch": 1.44, "learning_rate": 8.344478583047172e-06, "loss": 1.022, "step": 7988 }, { "epoch": 1.44, "learning_rate": 8.341767576360022e-06, "loss": 0.6312, "step": 7989 }, { "epoch": 1.44, "learning_rate": 8.339056569672872e-06, "loss": 0.9081, "step": 7990 }, { "epoch": 1.44, "learning_rate": 8.336345562985722e-06, "loss": 1.1878, "step": 7991 }, { "epoch": 1.44, "learning_rate": 8.333634556298572e-06, "loss": 0.7175, "step": 7992 }, { "epoch": 1.44, "learning_rate": 8.330923549611422e-06, "loss": 0.6466, "step": 7993 }, { "epoch": 1.44, "learning_rate": 8.328212542924272e-06, "loss": 0.7424, "step": 7994 }, { "epoch": 1.44, "learning_rate": 8.325501536237124e-06, "loss": 0.2923, "step": 7995 }, { "epoch": 1.45, "learning_rate": 8.322790529549973e-06, "loss": 0.9585, "step": 7996 }, { "epoch": 1.45, "learning_rate": 8.320079522862823e-06, "loss": 1.4892, "step": 7997 }, { "epoch": 1.45, "learning_rate": 8.317368516175673e-06, "loss": 0.9758, "step": 7998 }, { "epoch": 1.45, "learning_rate": 8.314657509488525e-06, "loss": 0.9613, "step": 7999 }, { "epoch": 1.45, "learning_rate": 8.311946502801373e-06, "loss": 0.6733, "step": 8000 }, { "epoch": 1.45, "learning_rate": 8.309235496114223e-06, "loss": 0.5166, "step": 8001 }, { "epoch": 1.45, "learning_rate": 8.306524489427073e-06, "loss": 0.5047, "step": 8002 }, { "epoch": 1.45, "learning_rate": 8.303813482739925e-06, "loss": 0.5987, "step": 8003 }, { "epoch": 1.45, "learning_rate": 8.301102476052774e-06, "loss": 0.69, "step": 8004 }, { "epoch": 1.45, "learning_rate": 8.298391469365624e-06, "loss": 0.3259, "step": 8005 }, { "epoch": 1.45, "learning_rate": 8.295680462678476e-06, "loss": 0.829, "step": 8006 }, { "epoch": 1.45, "learning_rate": 8.292969455991326e-06, "loss": 1.4828, "step": 8007 }, { "epoch": 1.45, "learning_rate": 8.290258449304174e-06, "loss": 0.4466, "step": 8008 }, { "epoch": 1.45, "learning_rate": 8.287547442617024e-06, "loss": 0.8714, "step": 8009 }, { "epoch": 1.45, "learning_rate": 8.284836435929876e-06, "loss": 0.6558, "step": 8010 }, { "epoch": 1.45, "learning_rate": 8.282125429242726e-06, "loss": 0.7123, "step": 8011 }, { "epoch": 1.45, "learning_rate": 8.279414422555575e-06, "loss": 0.6129, "step": 8012 }, { "epoch": 1.45, "learning_rate": 8.276703415868427e-06, "loss": 0.5505, "step": 8013 }, { "epoch": 1.45, "learning_rate": 8.273992409181277e-06, "loss": 0.7377, "step": 8014 }, { "epoch": 1.45, "learning_rate": 8.271281402494127e-06, "loss": 0.708, "step": 8015 }, { "epoch": 1.45, "learning_rate": 8.268570395806975e-06, "loss": 1.2659, "step": 8016 }, { "epoch": 1.45, "learning_rate": 8.265859389119827e-06, "loss": 0.4903, "step": 8017 }, { "epoch": 1.45, "learning_rate": 8.263148382432677e-06, "loss": 0.9393, "step": 8018 }, { "epoch": 1.45, "learning_rate": 8.260437375745527e-06, "loss": 0.5652, "step": 8019 }, { "epoch": 1.45, "learning_rate": 8.257726369058376e-06, "loss": 0.5657, "step": 8020 }, { "epoch": 1.45, "learning_rate": 8.255015362371228e-06, "loss": 1.3215, "step": 8021 }, { "epoch": 1.45, "learning_rate": 8.252304355684078e-06, "loss": 0.5898, "step": 8022 }, { "epoch": 1.45, "learning_rate": 8.249593348996928e-06, "loss": 0.554, "step": 8023 }, { "epoch": 1.45, "learning_rate": 8.246882342309778e-06, "loss": 0.8496, "step": 8024 }, { "epoch": 1.45, "learning_rate": 8.244171335622628e-06, "loss": 0.4834, "step": 8025 }, { "epoch": 1.45, "learning_rate": 8.241460328935478e-06, "loss": 0.4574, "step": 8026 }, { "epoch": 1.45, "learning_rate": 8.238749322248328e-06, "loss": 0.7754, "step": 8027 }, { "epoch": 1.45, "learning_rate": 8.236038315561179e-06, "loss": 0.6865, "step": 8028 }, { "epoch": 1.45, "learning_rate": 8.233327308874029e-06, "loss": 0.4726, "step": 8029 }, { "epoch": 1.45, "learning_rate": 8.230616302186879e-06, "loss": 0.7753, "step": 8030 }, { "epoch": 1.45, "learning_rate": 8.227905295499729e-06, "loss": 0.6418, "step": 8031 }, { "epoch": 1.45, "learning_rate": 8.225194288812579e-06, "loss": 1.2401, "step": 8032 }, { "epoch": 1.45, "learning_rate": 8.22248328212543e-06, "loss": 1.252, "step": 8033 }, { "epoch": 1.45, "learning_rate": 8.21977227543828e-06, "loss": 0.744, "step": 8034 }, { "epoch": 1.45, "learning_rate": 8.217061268751131e-06, "loss": 0.9983, "step": 8035 }, { "epoch": 1.45, "learning_rate": 8.21435026206398e-06, "loss": 0.3009, "step": 8036 }, { "epoch": 1.45, "learning_rate": 8.21163925537683e-06, "loss": 0.9089, "step": 8037 }, { "epoch": 1.45, "learning_rate": 8.20892824868968e-06, "loss": 0.5436, "step": 8038 }, { "epoch": 1.45, "learning_rate": 8.206217242002532e-06, "loss": 0.5979, "step": 8039 }, { "epoch": 1.45, "learning_rate": 8.20350623531538e-06, "loss": 0.6548, "step": 8040 }, { "epoch": 1.45, "learning_rate": 8.20079522862823e-06, "loss": 0.8132, "step": 8041 }, { "epoch": 1.45, "learning_rate": 8.19808422194108e-06, "loss": 1.0955, "step": 8042 }, { "epoch": 1.45, "learning_rate": 8.195373215253932e-06, "loss": 0.7672, "step": 8043 }, { "epoch": 1.45, "learning_rate": 8.19266220856678e-06, "loss": 0.5935, "step": 8044 }, { "epoch": 1.45, "learning_rate": 8.189951201879631e-06, "loss": 0.5704, "step": 8045 }, { "epoch": 1.45, "learning_rate": 8.187240195192483e-06, "loss": 0.9381, "step": 8046 }, { "epoch": 1.45, "learning_rate": 8.184529188505333e-06, "loss": 0.7598, "step": 8047 }, { "epoch": 1.45, "learning_rate": 8.181818181818181e-06, "loss": 0.2614, "step": 8048 }, { "epoch": 1.45, "learning_rate": 8.179107175131031e-06, "loss": 0.6536, "step": 8049 }, { "epoch": 1.45, "learning_rate": 8.176396168443883e-06, "loss": 0.3815, "step": 8050 }, { "epoch": 1.46, "learning_rate": 8.173685161756733e-06, "loss": 0.6303, "step": 8051 }, { "epoch": 1.46, "learning_rate": 8.170974155069582e-06, "loss": 0.4688, "step": 8052 }, { "epoch": 1.46, "learning_rate": 8.168263148382432e-06, "loss": 0.5326, "step": 8053 }, { "epoch": 1.46, "learning_rate": 8.165552141695284e-06, "loss": 0.6305, "step": 8054 }, { "epoch": 1.46, "learning_rate": 8.162841135008134e-06, "loss": 0.3807, "step": 8055 }, { "epoch": 1.46, "learning_rate": 8.160130128320982e-06, "loss": 1.2247, "step": 8056 }, { "epoch": 1.46, "learning_rate": 8.157419121633834e-06, "loss": 0.4463, "step": 8057 }, { "epoch": 1.46, "learning_rate": 8.154708114946684e-06, "loss": 0.3788, "step": 8058 }, { "epoch": 1.46, "learning_rate": 8.151997108259534e-06, "loss": 0.4599, "step": 8059 }, { "epoch": 1.46, "learning_rate": 8.149286101572383e-06, "loss": 1.2463, "step": 8060 }, { "epoch": 1.46, "learning_rate": 8.146575094885235e-06, "loss": 0.4884, "step": 8061 }, { "epoch": 1.46, "learning_rate": 8.143864088198085e-06, "loss": 0.8776, "step": 8062 }, { "epoch": 1.46, "learning_rate": 8.141153081510935e-06, "loss": 0.7688, "step": 8063 }, { "epoch": 1.46, "learning_rate": 8.138442074823783e-06, "loss": 0.9937, "step": 8064 }, { "epoch": 1.46, "learning_rate": 8.135731068136635e-06, "loss": 0.74, "step": 8065 }, { "epoch": 1.46, "learning_rate": 8.133020061449485e-06, "loss": 0.5054, "step": 8066 }, { "epoch": 1.46, "learning_rate": 8.130309054762336e-06, "loss": 0.5043, "step": 8067 }, { "epoch": 1.46, "learning_rate": 8.127598048075186e-06, "loss": 0.3956, "step": 8068 }, { "epoch": 1.46, "learning_rate": 8.124887041388036e-06, "loss": 0.7983, "step": 8069 }, { "epoch": 1.46, "learning_rate": 8.122176034700886e-06, "loss": 0.6048, "step": 8070 }, { "epoch": 1.46, "learning_rate": 8.119465028013736e-06, "loss": 0.5847, "step": 8071 }, { "epoch": 1.46, "learning_rate": 8.116754021326586e-06, "loss": 0.4008, "step": 8072 }, { "epoch": 1.46, "learning_rate": 8.114043014639436e-06, "loss": 0.4078, "step": 8073 }, { "epoch": 1.46, "learning_rate": 8.111332007952286e-06, "loss": 0.5409, "step": 8074 }, { "epoch": 1.46, "learning_rate": 8.108621001265137e-06, "loss": 0.5566, "step": 8075 }, { "epoch": 1.46, "learning_rate": 8.105909994577987e-06, "loss": 0.8103, "step": 8076 }, { "epoch": 1.46, "learning_rate": 8.103198987890837e-06, "loss": 1.1829, "step": 8077 }, { "epoch": 1.46, "learning_rate": 8.100487981203687e-06, "loss": 0.7847, "step": 8078 }, { "epoch": 1.46, "learning_rate": 8.097776974516539e-06, "loss": 0.5455, "step": 8079 }, { "epoch": 1.46, "learning_rate": 8.095065967829387e-06, "loss": 0.7883, "step": 8080 }, { "epoch": 1.46, "learning_rate": 8.092354961142237e-06, "loss": 0.3697, "step": 8081 }, { "epoch": 1.46, "learning_rate": 8.089643954455088e-06, "loss": 0.6079, "step": 8082 }, { "epoch": 1.46, "learning_rate": 8.08693294776794e-06, "loss": 1.0136, "step": 8083 }, { "epoch": 1.46, "learning_rate": 8.084221941080788e-06, "loss": 0.6293, "step": 8084 }, { "epoch": 1.46, "learning_rate": 8.081510934393638e-06, "loss": 0.4942, "step": 8085 }, { "epoch": 1.46, "learning_rate": 8.078799927706488e-06, "loss": 0.8216, "step": 8086 }, { "epoch": 1.46, "learning_rate": 8.07608892101934e-06, "loss": 0.5227, "step": 8087 }, { "epoch": 1.46, "learning_rate": 8.073377914332188e-06, "loss": 0.8849, "step": 8088 }, { "epoch": 1.46, "learning_rate": 8.070666907645038e-06, "loss": 0.4059, "step": 8089 }, { "epoch": 1.46, "learning_rate": 8.06795590095789e-06, "loss": 1.0278, "step": 8090 }, { "epoch": 1.46, "learning_rate": 8.06524489427074e-06, "loss": 1.0014, "step": 8091 }, { "epoch": 1.46, "learning_rate": 8.062533887583589e-06, "loss": 0.3208, "step": 8092 }, { "epoch": 1.46, "learning_rate": 8.059822880896439e-06, "loss": 0.8024, "step": 8093 }, { "epoch": 1.46, "learning_rate": 8.05711187420929e-06, "loss": 0.6556, "step": 8094 }, { "epoch": 1.46, "learning_rate": 8.054400867522141e-06, "loss": 0.487, "step": 8095 }, { "epoch": 1.46, "learning_rate": 8.05168986083499e-06, "loss": 0.8235, "step": 8096 }, { "epoch": 1.46, "learning_rate": 8.048978854147841e-06, "loss": 0.7265, "step": 8097 }, { "epoch": 1.46, "learning_rate": 8.046267847460691e-06, "loss": 1.5138, "step": 8098 }, { "epoch": 1.46, "learning_rate": 8.043556840773541e-06, "loss": 0.478, "step": 8099 }, { "epoch": 1.46, "learning_rate": 8.04084583408639e-06, "loss": 0.3258, "step": 8100 }, { "epoch": 1.46, "learning_rate": 8.038134827399242e-06, "loss": 0.452, "step": 8101 }, { "epoch": 1.46, "learning_rate": 8.035423820712092e-06, "loss": 0.7487, "step": 8102 }, { "epoch": 1.46, "learning_rate": 8.032712814024942e-06, "loss": 0.4042, "step": 8103 }, { "epoch": 1.46, "learning_rate": 8.03000180733779e-06, "loss": 0.8215, "step": 8104 }, { "epoch": 1.46, "learning_rate": 8.027290800650642e-06, "loss": 0.9311, "step": 8105 }, { "epoch": 1.47, "learning_rate": 8.024579793963492e-06, "loss": 0.3339, "step": 8106 }, { "epoch": 1.47, "learning_rate": 8.021868787276341e-06, "loss": 0.672, "step": 8107 }, { "epoch": 1.47, "learning_rate": 8.019157780589193e-06, "loss": 0.5144, "step": 8108 }, { "epoch": 1.47, "learning_rate": 8.016446773902043e-06, "loss": 1.6358, "step": 8109 }, { "epoch": 1.47, "learning_rate": 8.013735767214893e-06, "loss": 1.0063, "step": 8110 }, { "epoch": 1.47, "learning_rate": 8.011024760527741e-06, "loss": 1.2398, "step": 8111 }, { "epoch": 1.47, "learning_rate": 8.008313753840593e-06, "loss": 0.1627, "step": 8112 }, { "epoch": 1.47, "learning_rate": 8.005602747153443e-06, "loss": 0.6331, "step": 8113 }, { "epoch": 1.47, "learning_rate": 8.002891740466294e-06, "loss": 1.0265, "step": 8114 }, { "epoch": 1.47, "learning_rate": 8.000180733779142e-06, "loss": 0.4923, "step": 8115 }, { "epoch": 1.47, "learning_rate": 7.997469727091994e-06, "loss": 0.467, "step": 8116 }, { "epoch": 1.47, "learning_rate": 7.994758720404844e-06, "loss": 0.4376, "step": 8117 }, { "epoch": 1.47, "learning_rate": 7.992047713717694e-06, "loss": 0.542, "step": 8118 }, { "epoch": 1.47, "learning_rate": 7.989336707030544e-06, "loss": 0.9454, "step": 8119 }, { "epoch": 1.47, "learning_rate": 7.986625700343394e-06, "loss": 0.3885, "step": 8120 }, { "epoch": 1.47, "learning_rate": 7.983914693656244e-06, "loss": 0.5236, "step": 8121 }, { "epoch": 1.47, "learning_rate": 7.981203686969095e-06, "loss": 0.7626, "step": 8122 }, { "epoch": 1.47, "learning_rate": 7.978492680281945e-06, "loss": 0.4572, "step": 8123 }, { "epoch": 1.47, "learning_rate": 7.975781673594795e-06, "loss": 0.397, "step": 8124 }, { "epoch": 1.47, "learning_rate": 7.973070666907645e-06, "loss": 1.16, "step": 8125 }, { "epoch": 1.47, "learning_rate": 7.970359660220495e-06, "loss": 0.46, "step": 8126 }, { "epoch": 1.47, "learning_rate": 7.967648653533345e-06, "loss": 0.4757, "step": 8127 }, { "epoch": 1.47, "learning_rate": 7.964937646846195e-06, "loss": 0.6473, "step": 8128 }, { "epoch": 1.47, "learning_rate": 7.962226640159046e-06, "loss": 0.377, "step": 8129 }, { "epoch": 1.47, "learning_rate": 7.959515633471897e-06, "loss": 0.5781, "step": 8130 }, { "epoch": 1.47, "learning_rate": 7.956804626784746e-06, "loss": 0.9883, "step": 8131 }, { "epoch": 1.47, "learning_rate": 7.954093620097596e-06, "loss": 1.5251, "step": 8132 }, { "epoch": 1.47, "learning_rate": 7.951382613410446e-06, "loss": 1.0381, "step": 8133 }, { "epoch": 1.47, "learning_rate": 7.948671606723298e-06, "loss": 0.4876, "step": 8134 }, { "epoch": 1.47, "learning_rate": 7.945960600036146e-06, "loss": 0.5931, "step": 8135 }, { "epoch": 1.47, "learning_rate": 7.943249593348996e-06, "loss": 0.4648, "step": 8136 }, { "epoch": 1.47, "learning_rate": 7.940538586661847e-06, "loss": 0.6156, "step": 8137 }, { "epoch": 1.47, "learning_rate": 7.937827579974698e-06, "loss": 0.2568, "step": 8138 }, { "epoch": 1.47, "learning_rate": 7.935116573287547e-06, "loss": 1.0615, "step": 8139 }, { "epoch": 1.47, "learning_rate": 7.932405566600397e-06, "loss": 0.457, "step": 8140 }, { "epoch": 1.47, "learning_rate": 7.929694559913249e-06, "loss": 0.8284, "step": 8141 }, { "epoch": 1.47, "learning_rate": 7.926983553226099e-06, "loss": 0.9705, "step": 8142 }, { "epoch": 1.47, "learning_rate": 7.924272546538947e-06, "loss": 0.8207, "step": 8143 }, { "epoch": 1.47, "learning_rate": 7.921561539851798e-06, "loss": 1.1405, "step": 8144 }, { "epoch": 1.47, "learning_rate": 7.91885053316465e-06, "loss": 0.478, "step": 8145 }, { "epoch": 1.47, "learning_rate": 7.9161395264775e-06, "loss": 0.3899, "step": 8146 }, { "epoch": 1.47, "learning_rate": 7.913428519790348e-06, "loss": 0.3261, "step": 8147 }, { "epoch": 1.47, "learning_rate": 7.910717513103198e-06, "loss": 1.0203, "step": 8148 }, { "epoch": 1.47, "learning_rate": 7.90800650641605e-06, "loss": 0.3252, "step": 8149 }, { "epoch": 1.47, "learning_rate": 7.9052954997289e-06, "loss": 0.5061, "step": 8150 }, { "epoch": 1.47, "learning_rate": 7.902584493041748e-06, "loss": 0.5262, "step": 8151 }, { "epoch": 1.47, "learning_rate": 7.8998734863546e-06, "loss": 0.9248, "step": 8152 }, { "epoch": 1.47, "learning_rate": 7.89716247966745e-06, "loss": 0.8459, "step": 8153 }, { "epoch": 1.47, "learning_rate": 7.8944514729803e-06, "loss": 0.6797, "step": 8154 }, { "epoch": 1.47, "learning_rate": 7.891740466293149e-06, "loss": 0.763, "step": 8155 }, { "epoch": 1.47, "learning_rate": 7.889029459606e-06, "loss": 0.7339, "step": 8156 }, { "epoch": 1.47, "learning_rate": 7.886318452918851e-06, "loss": 0.7194, "step": 8157 }, { "epoch": 1.47, "learning_rate": 7.883607446231701e-06, "loss": 0.5753, "step": 8158 }, { "epoch": 1.47, "learning_rate": 7.880896439544551e-06, "loss": 0.2518, "step": 8159 }, { "epoch": 1.47, "learning_rate": 7.878185432857401e-06, "loss": 0.4833, "step": 8160 }, { "epoch": 1.47, "learning_rate": 7.875474426170252e-06, "loss": 0.4611, "step": 8161 }, { "epoch": 1.48, "learning_rate": 7.872763419483102e-06, "loss": 0.4131, "step": 8162 }, { "epoch": 1.48, "learning_rate": 7.870052412795952e-06, "loss": 1.3127, "step": 8163 }, { "epoch": 1.48, "learning_rate": 7.867341406108802e-06, "loss": 0.5046, "step": 8164 }, { "epoch": 1.48, "learning_rate": 7.864630399421652e-06, "loss": 0.796, "step": 8165 }, { "epoch": 1.48, "learning_rate": 7.861919392734502e-06, "loss": 0.6103, "step": 8166 }, { "epoch": 1.48, "learning_rate": 7.859208386047352e-06, "loss": 0.6812, "step": 8167 }, { "epoch": 1.48, "learning_rate": 7.856497379360202e-06, "loss": 1.124, "step": 8168 }, { "epoch": 1.48, "learning_rate": 7.853786372673053e-06, "loss": 0.5555, "step": 8169 }, { "epoch": 1.48, "learning_rate": 7.851075365985904e-06, "loss": 0.8996, "step": 8170 }, { "epoch": 1.48, "learning_rate": 7.848364359298753e-06, "loss": 1.0582, "step": 8171 }, { "epoch": 1.48, "learning_rate": 7.845653352611603e-06, "loss": 0.6904, "step": 8172 }, { "epoch": 1.48, "learning_rate": 7.842942345924453e-06, "loss": 0.5264, "step": 8173 }, { "epoch": 1.48, "learning_rate": 7.840231339237305e-06, "loss": 0.879, "step": 8174 }, { "epoch": 1.48, "learning_rate": 7.837520332550153e-06, "loss": 0.5707, "step": 8175 }, { "epoch": 1.48, "learning_rate": 7.834809325863004e-06, "loss": 0.2619, "step": 8176 }, { "epoch": 1.48, "learning_rate": 7.832098319175854e-06, "loss": 1.1625, "step": 8177 }, { "epoch": 1.48, "learning_rate": 7.829387312488705e-06, "loss": 0.7555, "step": 8178 }, { "epoch": 1.48, "learning_rate": 7.826676305801554e-06, "loss": 0.3682, "step": 8179 }, { "epoch": 1.48, "learning_rate": 7.823965299114404e-06, "loss": 0.3796, "step": 8180 }, { "epoch": 1.48, "learning_rate": 7.821254292427256e-06, "loss": 0.6894, "step": 8181 }, { "epoch": 1.48, "learning_rate": 7.818543285740106e-06, "loss": 0.8114, "step": 8182 }, { "epoch": 1.48, "learning_rate": 7.815832279052954e-06, "loss": 0.6023, "step": 8183 }, { "epoch": 1.48, "learning_rate": 7.813121272365805e-06, "loss": 0.5902, "step": 8184 }, { "epoch": 1.48, "learning_rate": 7.810410265678656e-06, "loss": 0.5486, "step": 8185 }, { "epoch": 1.48, "learning_rate": 7.807699258991507e-06, "loss": 0.5045, "step": 8186 }, { "epoch": 1.48, "learning_rate": 7.804988252304355e-06, "loss": 0.696, "step": 8187 }, { "epoch": 1.48, "learning_rate": 7.802277245617205e-06, "loss": 0.8677, "step": 8188 }, { "epoch": 1.48, "learning_rate": 7.799566238930057e-06, "loss": 0.7395, "step": 8189 }, { "epoch": 1.48, "learning_rate": 7.796855232242907e-06, "loss": 0.4621, "step": 8190 }, { "epoch": 1.48, "learning_rate": 7.794144225555756e-06, "loss": 0.5944, "step": 8191 }, { "epoch": 1.48, "learning_rate": 7.791433218868607e-06, "loss": 0.3721, "step": 8192 }, { "epoch": 1.48, "learning_rate": 7.788722212181458e-06, "loss": 0.8495, "step": 8193 }, { "epoch": 1.48, "learning_rate": 7.786011205494308e-06, "loss": 0.9928, "step": 8194 }, { "epoch": 1.48, "learning_rate": 7.783300198807156e-06, "loss": 0.6327, "step": 8195 }, { "epoch": 1.48, "learning_rate": 7.780589192120008e-06, "loss": 0.6868, "step": 8196 }, { "epoch": 1.48, "learning_rate": 7.777878185432858e-06, "loss": 0.5273, "step": 8197 }, { "epoch": 1.48, "learning_rate": 7.775167178745708e-06, "loss": 1.0884, "step": 8198 }, { "epoch": 1.48, "learning_rate": 7.772456172058557e-06, "loss": 0.5793, "step": 8199 }, { "epoch": 1.48, "learning_rate": 7.769745165371408e-06, "loss": 1.0678, "step": 8200 }, { "epoch": 1.48, "learning_rate": 7.767034158684259e-06, "loss": 0.8874, "step": 8201 }, { "epoch": 1.48, "learning_rate": 7.764323151997109e-06, "loss": 0.5946, "step": 8202 }, { "epoch": 1.48, "learning_rate": 7.761612145309959e-06, "loss": 0.6136, "step": 8203 }, { "epoch": 1.48, "learning_rate": 7.758901138622809e-06, "loss": 0.8007, "step": 8204 }, { "epoch": 1.48, "learning_rate": 7.756190131935659e-06, "loss": 0.6009, "step": 8205 }, { "epoch": 1.48, "learning_rate": 7.75347912524851e-06, "loss": 0.3995, "step": 8206 }, { "epoch": 1.48, "learning_rate": 7.75076811856136e-06, "loss": 0.8846, "step": 8207 }, { "epoch": 1.48, "learning_rate": 7.74805711187421e-06, "loss": 1.2433, "step": 8208 }, { "epoch": 1.48, "learning_rate": 7.74534610518706e-06, "loss": 0.8688, "step": 8209 }, { "epoch": 1.48, "learning_rate": 7.74263509849991e-06, "loss": 0.895, "step": 8210 }, { "epoch": 1.48, "learning_rate": 7.73992409181276e-06, "loss": 1.0233, "step": 8211 }, { "epoch": 1.48, "learning_rate": 7.73721308512561e-06, "loss": 0.2502, "step": 8212 }, { "epoch": 1.48, "learning_rate": 7.73450207843846e-06, "loss": 0.9848, "step": 8213 }, { "epoch": 1.48, "learning_rate": 7.731791071751312e-06, "loss": 0.4661, "step": 8214 }, { "epoch": 1.48, "learning_rate": 7.72908006506416e-06, "loss": 0.2313, "step": 8215 }, { "epoch": 1.48, "learning_rate": 7.72636905837701e-06, "loss": 0.7625, "step": 8216 }, { "epoch": 1.49, "learning_rate": 7.72365805168986e-06, "loss": 0.4381, "step": 8217 }, { "epoch": 1.49, "learning_rate": 7.720947045002713e-06, "loss": 1.0334, "step": 8218 }, { "epoch": 1.49, "learning_rate": 7.718236038315561e-06, "loss": 0.4756, "step": 8219 }, { "epoch": 1.49, "learning_rate": 7.715525031628411e-06, "loss": 0.6965, "step": 8220 }, { "epoch": 1.49, "learning_rate": 7.712814024941261e-06, "loss": 0.5987, "step": 8221 }, { "epoch": 1.49, "learning_rate": 7.710103018254113e-06, "loss": 0.7566, "step": 8222 }, { "epoch": 1.49, "learning_rate": 7.707392011566962e-06, "loss": 0.6, "step": 8223 }, { "epoch": 1.49, "learning_rate": 7.704681004879812e-06, "loss": 0.8949, "step": 8224 }, { "epoch": 1.49, "learning_rate": 7.701969998192663e-06, "loss": 0.922, "step": 8225 }, { "epoch": 1.49, "learning_rate": 7.699258991505514e-06, "loss": 0.4133, "step": 8226 }, { "epoch": 1.49, "learning_rate": 7.696547984818362e-06, "loss": 0.9684, "step": 8227 }, { "epoch": 1.49, "learning_rate": 7.693836978131212e-06, "loss": 1.3043, "step": 8228 }, { "epoch": 1.49, "learning_rate": 7.691125971444064e-06, "loss": 0.6065, "step": 8229 }, { "epoch": 1.49, "learning_rate": 7.688414964756914e-06, "loss": 0.55, "step": 8230 }, { "epoch": 1.49, "learning_rate": 7.685703958069763e-06, "loss": 0.9427, "step": 8231 }, { "epoch": 1.49, "learning_rate": 7.682992951382614e-06, "loss": 0.6407, "step": 8232 }, { "epoch": 1.49, "learning_rate": 7.680281944695465e-06, "loss": 0.5488, "step": 8233 }, { "epoch": 1.49, "learning_rate": 7.677570938008315e-06, "loss": 0.6753, "step": 8234 }, { "epoch": 1.49, "learning_rate": 7.674859931321163e-06, "loss": 0.5487, "step": 8235 }, { "epoch": 1.49, "learning_rate": 7.672148924634015e-06, "loss": 0.9619, "step": 8236 }, { "epoch": 1.49, "learning_rate": 7.669437917946865e-06, "loss": 1.2972, "step": 8237 }, { "epoch": 1.49, "learning_rate": 7.666726911259714e-06, "loss": 0.6472, "step": 8238 }, { "epoch": 1.49, "learning_rate": 7.664015904572564e-06, "loss": 0.3026, "step": 8239 }, { "epoch": 1.49, "learning_rate": 7.661304897885416e-06, "loss": 0.5636, "step": 8240 }, { "epoch": 1.49, "learning_rate": 7.658593891198266e-06, "loss": 1.1247, "step": 8241 }, { "epoch": 1.49, "learning_rate": 7.655882884511114e-06, "loss": 0.8533, "step": 8242 }, { "epoch": 1.49, "learning_rate": 7.653171877823966e-06, "loss": 0.59, "step": 8243 }, { "epoch": 1.49, "learning_rate": 7.650460871136816e-06, "loss": 0.3753, "step": 8244 }, { "epoch": 1.49, "learning_rate": 7.647749864449666e-06, "loss": 1.2643, "step": 8245 }, { "epoch": 1.49, "learning_rate": 7.645038857762515e-06, "loss": 1.4287, "step": 8246 }, { "epoch": 1.49, "learning_rate": 7.642327851075366e-06, "loss": 0.2773, "step": 8247 }, { "epoch": 1.49, "learning_rate": 7.639616844388217e-06, "loss": 1.1414, "step": 8248 }, { "epoch": 1.49, "learning_rate": 7.636905837701067e-06, "loss": 1.0673, "step": 8249 }, { "epoch": 1.49, "learning_rate": 7.634194831013915e-06, "loss": 0.7876, "step": 8250 }, { "epoch": 1.49, "learning_rate": 7.631483824326767e-06, "loss": 0.3465, "step": 8251 }, { "epoch": 1.49, "learning_rate": 7.628772817639617e-06, "loss": 0.501, "step": 8252 }, { "epoch": 1.49, "learning_rate": 7.626061810952468e-06, "loss": 1.028, "step": 8253 }, { "epoch": 1.49, "learning_rate": 7.6233508042653165e-06, "loss": 0.5837, "step": 8254 }, { "epoch": 1.49, "learning_rate": 7.6206397975781675e-06, "loss": 0.456, "step": 8255 }, { "epoch": 1.49, "learning_rate": 7.617928790891018e-06, "loss": 0.9573, "step": 8256 }, { "epoch": 1.49, "learning_rate": 7.615217784203869e-06, "loss": 0.7422, "step": 8257 }, { "epoch": 1.49, "learning_rate": 7.612506777516717e-06, "loss": 0.2222, "step": 8258 }, { "epoch": 1.49, "learning_rate": 7.609795770829568e-06, "loss": 0.6923, "step": 8259 }, { "epoch": 1.49, "learning_rate": 7.607084764142418e-06, "loss": 0.8448, "step": 8260 }, { "epoch": 1.49, "learning_rate": 7.604373757455269e-06, "loss": 1.0488, "step": 8261 }, { "epoch": 1.49, "learning_rate": 7.601662750768118e-06, "loss": 0.5789, "step": 8262 }, { "epoch": 1.49, "learning_rate": 7.598951744080969e-06, "loss": 1.1445, "step": 8263 }, { "epoch": 1.49, "learning_rate": 7.5962407373938196e-06, "loss": 0.7405, "step": 8264 }, { "epoch": 1.49, "learning_rate": 7.59352973070667e-06, "loss": 0.8426, "step": 8265 }, { "epoch": 1.49, "learning_rate": 7.590818724019519e-06, "loss": 1.1669, "step": 8266 }, { "epoch": 1.49, "learning_rate": 7.588107717332369e-06, "loss": 0.5056, "step": 8267 }, { "epoch": 1.49, "learning_rate": 7.58539671064522e-06, "loss": 0.9213, "step": 8268 }, { "epoch": 1.49, "learning_rate": 7.58268570395807e-06, "loss": 0.7409, "step": 8269 }, { "epoch": 1.49, "learning_rate": 7.5799746972709195e-06, "loss": 0.867, "step": 8270 }, { "epoch": 1.49, "learning_rate": 7.57726369058377e-06, "loss": 0.4947, "step": 8271 }, { "epoch": 1.5, "learning_rate": 7.574552683896621e-06, "loss": 0.1851, "step": 8272 }, { "epoch": 1.5, "learning_rate": 7.571841677209471e-06, "loss": 0.5097, "step": 8273 }, { "epoch": 1.5, "learning_rate": 7.56913067052232e-06, "loss": 0.8268, "step": 8274 }, { "epoch": 1.5, "learning_rate": 7.566419663835171e-06, "loss": 1.1039, "step": 8275 }, { "epoch": 1.5, "learning_rate": 7.563708657148021e-06, "loss": 0.7724, "step": 8276 }, { "epoch": 1.5, "learning_rate": 7.560997650460872e-06, "loss": 0.7055, "step": 8277 }, { "epoch": 1.5, "learning_rate": 7.558286643773721e-06, "loss": 0.7112, "step": 8278 }, { "epoch": 1.5, "learning_rate": 7.555575637086572e-06, "loss": 0.6859, "step": 8279 }, { "epoch": 1.5, "learning_rate": 7.552864630399422e-06, "loss": 0.9108, "step": 8280 }, { "epoch": 1.5, "learning_rate": 7.550153623712273e-06, "loss": 0.6967, "step": 8281 }, { "epoch": 1.5, "learning_rate": 7.547442617025121e-06, "loss": 0.669, "step": 8282 }, { "epoch": 1.5, "learning_rate": 7.544731610337972e-06, "loss": 0.4187, "step": 8283 }, { "epoch": 1.5, "learning_rate": 7.542020603650823e-06, "loss": 0.7563, "step": 8284 }, { "epoch": 1.5, "learning_rate": 7.539309596963673e-06, "loss": 0.6327, "step": 8285 }, { "epoch": 1.5, "learning_rate": 7.5365985902765225e-06, "loss": 0.3952, "step": 8286 }, { "epoch": 1.5, "learning_rate": 7.533887583589373e-06, "loss": 0.5436, "step": 8287 }, { "epoch": 1.5, "learning_rate": 7.531176576902224e-06, "loss": 0.7845, "step": 8288 }, { "epoch": 1.5, "learning_rate": 7.528465570215074e-06, "loss": 1.043, "step": 8289 }, { "epoch": 1.5, "learning_rate": 7.525754563527923e-06, "loss": 0.9082, "step": 8290 }, { "epoch": 1.5, "learning_rate": 7.523043556840773e-06, "loss": 0.802, "step": 8291 }, { "epoch": 1.5, "learning_rate": 7.520332550153624e-06, "loss": 0.3468, "step": 8292 }, { "epoch": 1.5, "learning_rate": 7.517621543466474e-06, "loss": 0.5884, "step": 8293 }, { "epoch": 1.5, "learning_rate": 7.514910536779324e-06, "loss": 0.4866, "step": 8294 }, { "epoch": 1.5, "learning_rate": 7.512199530092175e-06, "loss": 0.3712, "step": 8295 }, { "epoch": 1.5, "learning_rate": 7.509488523405025e-06, "loss": 0.6535, "step": 8296 }, { "epoch": 1.5, "learning_rate": 7.506777516717876e-06, "loss": 0.4255, "step": 8297 }, { "epoch": 1.5, "learning_rate": 7.504066510030724e-06, "loss": 0.4746, "step": 8298 }, { "epoch": 1.5, "learning_rate": 7.501355503343575e-06, "loss": 0.6884, "step": 8299 }, { "epoch": 1.5, "learning_rate": 7.498644496656425e-06, "loss": 1.0978, "step": 8300 }, { "epoch": 1.5, "learning_rate": 7.495933489969275e-06, "loss": 0.6193, "step": 8301 }, { "epoch": 1.5, "learning_rate": 7.4932224832821255e-06, "loss": 0.586, "step": 8302 }, { "epoch": 1.5, "learning_rate": 7.490511476594976e-06, "loss": 0.939, "step": 8303 }, { "epoch": 1.5, "learning_rate": 7.487800469907826e-06, "loss": 0.6739, "step": 8304 }, { "epoch": 1.5, "learning_rate": 7.485089463220676e-06, "loss": 1.0132, "step": 8305 }, { "epoch": 1.5, "learning_rate": 7.482378456533527e-06, "loss": 0.552, "step": 8306 }, { "epoch": 1.5, "learning_rate": 7.479667449846376e-06, "loss": 0.5083, "step": 8307 }, { "epoch": 1.5, "learning_rate": 7.476956443159227e-06, "loss": 0.3576, "step": 8308 }, { "epoch": 1.5, "learning_rate": 7.4742454364720765e-06, "loss": 0.4419, "step": 8309 }, { "epoch": 1.5, "learning_rate": 7.4715344297849274e-06, "loss": 0.796, "step": 8310 }, { "epoch": 1.5, "learning_rate": 7.468823423097777e-06, "loss": 0.9608, "step": 8311 }, { "epoch": 1.5, "learning_rate": 7.466112416410628e-06, "loss": 0.3453, "step": 8312 }, { "epoch": 1.5, "learning_rate": 7.463401409723477e-06, "loss": 0.3931, "step": 8313 }, { "epoch": 1.5, "learning_rate": 7.460690403036328e-06, "loss": 0.9507, "step": 8314 }, { "epoch": 1.5, "learning_rate": 7.457979396349177e-06, "loss": 0.5254, "step": 8315 }, { "epoch": 1.5, "learning_rate": 7.455268389662028e-06, "loss": 0.6834, "step": 8316 }, { "epoch": 1.5, "learning_rate": 7.452557382974878e-06, "loss": 0.6183, "step": 8317 }, { "epoch": 1.5, "learning_rate": 7.4498463762877285e-06, "loss": 0.5268, "step": 8318 }, { "epoch": 1.5, "learning_rate": 7.447135369600579e-06, "loss": 0.723, "step": 8319 }, { "epoch": 1.5, "learning_rate": 7.444424362913429e-06, "loss": 0.8636, "step": 8320 }, { "epoch": 1.5, "learning_rate": 7.441713356226279e-06, "loss": 1.0019, "step": 8321 }, { "epoch": 1.5, "learning_rate": 7.439002349539129e-06, "loss": 0.4219, "step": 8322 }, { "epoch": 1.5, "learning_rate": 7.436291342851979e-06, "loss": 0.7839, "step": 8323 }, { "epoch": 1.5, "learning_rate": 7.433580336164829e-06, "loss": 0.7367, "step": 8324 }, { "epoch": 1.5, "learning_rate": 7.4308693294776795e-06, "loss": 0.3546, "step": 8325 }, { "epoch": 1.5, "learning_rate": 7.4281583227905304e-06, "loss": 0.7369, "step": 8326 }, { "epoch": 1.5, "learning_rate": 7.42544731610338e-06, "loss": 1.0476, "step": 8327 }, { "epoch": 1.51, "learning_rate": 7.422736309416231e-06, "loss": 0.8574, "step": 8328 }, { "epoch": 1.51, "learning_rate": 7.42002530272908e-06, "loss": 0.785, "step": 8329 }, { "epoch": 1.51, "learning_rate": 7.417314296041931e-06, "loss": 0.3442, "step": 8330 }, { "epoch": 1.51, "learning_rate": 7.41460328935478e-06, "loss": 0.4353, "step": 8331 }, { "epoch": 1.51, "learning_rate": 7.411892282667631e-06, "loss": 1.0711, "step": 8332 }, { "epoch": 1.51, "learning_rate": 7.4091812759804805e-06, "loss": 0.8475, "step": 8333 }, { "epoch": 1.51, "learning_rate": 7.4064702692933315e-06, "loss": 0.6347, "step": 8334 }, { "epoch": 1.51, "learning_rate": 7.403759262606181e-06, "loss": 0.3825, "step": 8335 }, { "epoch": 1.51, "learning_rate": 7.401048255919032e-06, "loss": 0.544, "step": 8336 }, { "epoch": 1.51, "learning_rate": 7.398337249231882e-06, "loss": 0.9881, "step": 8337 }, { "epoch": 1.51, "learning_rate": 7.395626242544732e-06, "loss": 0.8659, "step": 8338 }, { "epoch": 1.51, "learning_rate": 7.392915235857582e-06, "loss": 1.0478, "step": 8339 }, { "epoch": 1.51, "learning_rate": 7.390204229170432e-06, "loss": 0.3521, "step": 8340 }, { "epoch": 1.51, "learning_rate": 7.3874932224832825e-06, "loss": 1.3313, "step": 8341 }, { "epoch": 1.51, "learning_rate": 7.384782215796133e-06, "loss": 0.8493, "step": 8342 }, { "epoch": 1.51, "learning_rate": 7.382071209108983e-06, "loss": 1.0239, "step": 8343 }, { "epoch": 1.51, "learning_rate": 7.379360202421833e-06, "loss": 0.5812, "step": 8344 }, { "epoch": 1.51, "learning_rate": 7.376649195734683e-06, "loss": 0.6302, "step": 8345 }, { "epoch": 1.51, "learning_rate": 7.373938189047533e-06, "loss": 0.8121, "step": 8346 }, { "epoch": 1.51, "learning_rate": 7.371227182360383e-06, "loss": 0.5729, "step": 8347 }, { "epoch": 1.51, "learning_rate": 7.368516175673234e-06, "loss": 0.9677, "step": 8348 }, { "epoch": 1.51, "learning_rate": 7.3658051689860835e-06, "loss": 0.863, "step": 8349 }, { "epoch": 1.51, "learning_rate": 7.3630941622989345e-06, "loss": 0.901, "step": 8350 }, { "epoch": 1.51, "learning_rate": 7.360383155611784e-06, "loss": 0.5833, "step": 8351 }, { "epoch": 1.51, "learning_rate": 7.357672148924635e-06, "loss": 0.941, "step": 8352 }, { "epoch": 1.51, "learning_rate": 7.354961142237484e-06, "loss": 1.1317, "step": 8353 }, { "epoch": 1.51, "learning_rate": 7.352250135550335e-06, "loss": 0.5051, "step": 8354 }, { "epoch": 1.51, "learning_rate": 7.349539128863184e-06, "loss": 0.8127, "step": 8355 }, { "epoch": 1.51, "learning_rate": 7.346828122176035e-06, "loss": 1.2771, "step": 8356 }, { "epoch": 1.51, "learning_rate": 7.3441171154888855e-06, "loss": 1.1447, "step": 8357 }, { "epoch": 1.51, "learning_rate": 7.341406108801736e-06, "loss": 0.5847, "step": 8358 }, { "epoch": 1.51, "learning_rate": 7.338695102114586e-06, "loss": 0.4691, "step": 8359 }, { "epoch": 1.51, "learning_rate": 7.335984095427436e-06, "loss": 0.3858, "step": 8360 }, { "epoch": 1.51, "learning_rate": 7.333273088740286e-06, "loss": 0.7011, "step": 8361 }, { "epoch": 1.51, "learning_rate": 7.330562082053136e-06, "loss": 0.4833, "step": 8362 }, { "epoch": 1.51, "learning_rate": 7.327851075365986e-06, "loss": 0.596, "step": 8363 }, { "epoch": 1.51, "learning_rate": 7.3251400686788355e-06, "loss": 0.8733, "step": 8364 }, { "epoch": 1.51, "learning_rate": 7.3224290619916865e-06, "loss": 0.7442, "step": 8365 }, { "epoch": 1.51, "learning_rate": 7.319718055304536e-06, "loss": 0.5954, "step": 8366 }, { "epoch": 1.51, "learning_rate": 7.317007048617387e-06, "loss": 0.2545, "step": 8367 }, { "epoch": 1.51, "learning_rate": 7.314296041930237e-06, "loss": 0.7818, "step": 8368 }, { "epoch": 1.51, "learning_rate": 7.311585035243087e-06, "loss": 1.0704, "step": 8369 }, { "epoch": 1.51, "learning_rate": 7.308874028555937e-06, "loss": 0.3638, "step": 8370 }, { "epoch": 1.51, "learning_rate": 7.306163021868787e-06, "loss": 0.9443, "step": 8371 }, { "epoch": 1.51, "learning_rate": 7.3034520151816375e-06, "loss": 0.5985, "step": 8372 }, { "epoch": 1.51, "learning_rate": 7.300741008494488e-06, "loss": 0.6278, "step": 8373 }, { "epoch": 1.51, "learning_rate": 7.298030001807338e-06, "loss": 0.8745, "step": 8374 }, { "epoch": 1.51, "learning_rate": 7.295318995120188e-06, "loss": 0.3822, "step": 8375 }, { "epoch": 1.51, "learning_rate": 7.292607988433038e-06, "loss": 0.4006, "step": 8376 }, { "epoch": 1.51, "learning_rate": 7.289896981745888e-06, "loss": 0.5727, "step": 8377 }, { "epoch": 1.51, "learning_rate": 7.287185975058738e-06, "loss": 1.0015, "step": 8378 }, { "epoch": 1.51, "learning_rate": 7.284474968371589e-06, "loss": 0.6898, "step": 8379 }, { "epoch": 1.51, "learning_rate": 7.2817639616844385e-06, "loss": 0.5028, "step": 8380 }, { "epoch": 1.51, "learning_rate": 7.2790529549972895e-06, "loss": 0.7508, "step": 8381 }, { "epoch": 1.51, "learning_rate": 7.276341948310139e-06, "loss": 1.0986, "step": 8382 }, { "epoch": 1.52, "learning_rate": 7.27363094162299e-06, "loss": 0.6106, "step": 8383 }, { "epoch": 1.52, "learning_rate": 7.270919934935839e-06, "loss": 0.4873, "step": 8384 }, { "epoch": 1.52, "learning_rate": 7.26820892824869e-06, "loss": 0.5998, "step": 8385 }, { "epoch": 1.52, "learning_rate": 7.265497921561539e-06, "loss": 0.7228, "step": 8386 }, { "epoch": 1.52, "learning_rate": 7.26278691487439e-06, "loss": 0.9796, "step": 8387 }, { "epoch": 1.52, "learning_rate": 7.2600759081872405e-06, "loss": 0.5247, "step": 8388 }, { "epoch": 1.52, "learning_rate": 7.257364901500091e-06, "loss": 0.4997, "step": 8389 }, { "epoch": 1.52, "learning_rate": 7.254653894812941e-06, "loss": 0.359, "step": 8390 }, { "epoch": 1.52, "learning_rate": 7.251942888125791e-06, "loss": 1.0076, "step": 8391 }, { "epoch": 1.52, "learning_rate": 7.249231881438641e-06, "loss": 0.5428, "step": 8392 }, { "epoch": 1.52, "learning_rate": 7.246520874751491e-06, "loss": 0.8909, "step": 8393 }, { "epoch": 1.52, "learning_rate": 7.243809868064341e-06, "loss": 0.3545, "step": 8394 }, { "epoch": 1.52, "learning_rate": 7.241098861377191e-06, "loss": 0.4079, "step": 8395 }, { "epoch": 1.52, "learning_rate": 7.2383878546900415e-06, "loss": 0.8445, "step": 8396 }, { "epoch": 1.52, "learning_rate": 7.235676848002892e-06, "loss": 0.3583, "step": 8397 }, { "epoch": 1.52, "learning_rate": 7.232965841315742e-06, "loss": 0.5734, "step": 8398 }, { "epoch": 1.52, "learning_rate": 7.230254834628593e-06, "loss": 0.6661, "step": 8399 }, { "epoch": 1.52, "learning_rate": 7.227543827941442e-06, "loss": 0.4528, "step": 8400 }, { "epoch": 1.52, "learning_rate": 7.224832821254293e-06, "loss": 0.9728, "step": 8401 }, { "epoch": 1.52, "learning_rate": 7.222121814567142e-06, "loss": 0.8578, "step": 8402 }, { "epoch": 1.52, "learning_rate": 7.219410807879993e-06, "loss": 0.7047, "step": 8403 }, { "epoch": 1.52, "learning_rate": 7.216699801192843e-06, "loss": 0.7709, "step": 8404 }, { "epoch": 1.52, "learning_rate": 7.213988794505694e-06, "loss": 0.614, "step": 8405 }, { "epoch": 1.52, "learning_rate": 7.211277787818543e-06, "loss": 0.5528, "step": 8406 }, { "epoch": 1.52, "learning_rate": 7.208566781131394e-06, "loss": 0.4512, "step": 8407 }, { "epoch": 1.52, "learning_rate": 7.205855774444243e-06, "loss": 0.2948, "step": 8408 }, { "epoch": 1.52, "learning_rate": 7.203144767757094e-06, "loss": 1.4246, "step": 8409 }, { "epoch": 1.52, "learning_rate": 7.200433761069944e-06, "loss": 0.4908, "step": 8410 }, { "epoch": 1.52, "learning_rate": 7.197722754382794e-06, "loss": 0.6584, "step": 8411 }, { "epoch": 1.52, "learning_rate": 7.1950117476956445e-06, "loss": 0.7397, "step": 8412 }, { "epoch": 1.52, "learning_rate": 7.192300741008495e-06, "loss": 0.7471, "step": 8413 }, { "epoch": 1.52, "learning_rate": 7.189589734321345e-06, "loss": 0.6738, "step": 8414 }, { "epoch": 1.52, "learning_rate": 7.186878727634195e-06, "loss": 1.0142, "step": 8415 }, { "epoch": 1.52, "learning_rate": 7.184167720947045e-06, "loss": 0.5617, "step": 8416 }, { "epoch": 1.52, "learning_rate": 7.181456714259895e-06, "loss": 0.8871, "step": 8417 }, { "epoch": 1.52, "learning_rate": 7.178745707572745e-06, "loss": 0.753, "step": 8418 }, { "epoch": 1.52, "learning_rate": 7.1760347008855955e-06, "loss": 0.774, "step": 8419 }, { "epoch": 1.52, "learning_rate": 7.173323694198446e-06, "loss": 0.8231, "step": 8420 }, { "epoch": 1.52, "learning_rate": 7.170612687511297e-06, "loss": 0.7353, "step": 8421 }, { "epoch": 1.52, "learning_rate": 7.167901680824146e-06, "loss": 0.2894, "step": 8422 }, { "epoch": 1.52, "learning_rate": 7.165190674136997e-06, "loss": 0.869, "step": 8423 }, { "epoch": 1.52, "learning_rate": 7.162479667449846e-06, "loss": 0.9516, "step": 8424 }, { "epoch": 1.52, "learning_rate": 7.159768660762697e-06, "loss": 0.9238, "step": 8425 }, { "epoch": 1.52, "learning_rate": 7.157057654075546e-06, "loss": 1.003, "step": 8426 }, { "epoch": 1.52, "learning_rate": 7.154346647388397e-06, "loss": 0.7044, "step": 8427 }, { "epoch": 1.52, "learning_rate": 7.151635640701247e-06, "loss": 0.7875, "step": 8428 }, { "epoch": 1.52, "learning_rate": 7.148924634014098e-06, "loss": 1.1253, "step": 8429 }, { "epoch": 1.52, "learning_rate": 7.146213627326948e-06, "loss": 0.8505, "step": 8430 }, { "epoch": 1.52, "learning_rate": 7.143502620639798e-06, "loss": 1.2931, "step": 8431 }, { "epoch": 1.52, "learning_rate": 7.140791613952648e-06, "loss": 1.0084, "step": 8432 }, { "epoch": 1.52, "learning_rate": 7.138080607265498e-06, "loss": 0.1755, "step": 8433 }, { "epoch": 1.52, "learning_rate": 7.135369600578348e-06, "loss": 0.5311, "step": 8434 }, { "epoch": 1.52, "learning_rate": 7.1326585938911985e-06, "loss": 0.8491, "step": 8435 }, { "epoch": 1.52, "learning_rate": 7.129947587204049e-06, "loss": 0.834, "step": 8436 }, { "epoch": 1.52, "learning_rate": 7.127236580516899e-06, "loss": 0.688, "step": 8437 }, { "epoch": 1.53, "learning_rate": 7.124525573829749e-06, "loss": 0.6879, "step": 8438 }, { "epoch": 1.53, "learning_rate": 7.121814567142599e-06, "loss": 0.4357, "step": 8439 }, { "epoch": 1.53, "learning_rate": 7.119103560455449e-06, "loss": 0.52, "step": 8440 }, { "epoch": 1.53, "learning_rate": 7.1163925537683e-06, "loss": 1.3684, "step": 8441 }, { "epoch": 1.53, "learning_rate": 7.113681547081149e-06, "loss": 0.8686, "step": 8442 }, { "epoch": 1.53, "learning_rate": 7.110970540394e-06, "loss": 1.1816, "step": 8443 }, { "epoch": 1.53, "learning_rate": 7.10825953370685e-06, "loss": 0.5505, "step": 8444 }, { "epoch": 1.53, "learning_rate": 7.105548527019701e-06, "loss": 0.9164, "step": 8445 }, { "epoch": 1.53, "learning_rate": 7.10283752033255e-06, "loss": 0.7786, "step": 8446 }, { "epoch": 1.53, "learning_rate": 7.100126513645401e-06, "loss": 0.3617, "step": 8447 }, { "epoch": 1.53, "learning_rate": 7.09741550695825e-06, "loss": 1.1011, "step": 8448 }, { "epoch": 1.53, "learning_rate": 7.094704500271101e-06, "loss": 0.8007, "step": 8449 }, { "epoch": 1.53, "learning_rate": 7.0919934935839505e-06, "loss": 0.8459, "step": 8450 }, { "epoch": 1.53, "learning_rate": 7.0892824868968015e-06, "loss": 0.4651, "step": 8451 }, { "epoch": 1.53, "learning_rate": 7.086571480209652e-06, "loss": 0.5697, "step": 8452 }, { "epoch": 1.53, "learning_rate": 7.083860473522502e-06, "loss": 0.6735, "step": 8453 }, { "epoch": 1.53, "learning_rate": 7.081149466835352e-06, "loss": 0.4451, "step": 8454 }, { "epoch": 1.53, "learning_rate": 7.078438460148202e-06, "loss": 0.444, "step": 8455 }, { "epoch": 1.53, "learning_rate": 7.075727453461052e-06, "loss": 0.9431, "step": 8456 }, { "epoch": 1.53, "learning_rate": 7.073016446773902e-06, "loss": 0.8925, "step": 8457 }, { "epoch": 1.53, "learning_rate": 7.070305440086752e-06, "loss": 0.6176, "step": 8458 }, { "epoch": 1.53, "learning_rate": 7.0675944333996025e-06, "loss": 0.6532, "step": 8459 }, { "epoch": 1.53, "learning_rate": 7.064883426712453e-06, "loss": 0.6747, "step": 8460 }, { "epoch": 1.53, "learning_rate": 7.062172420025303e-06, "loss": 0.3023, "step": 8461 }, { "epoch": 1.53, "learning_rate": 7.059461413338153e-06, "loss": 0.5022, "step": 8462 }, { "epoch": 1.53, "learning_rate": 7.056750406651004e-06, "loss": 1.1917, "step": 8463 }, { "epoch": 1.53, "learning_rate": 7.054039399963853e-06, "loss": 0.4532, "step": 8464 }, { "epoch": 1.53, "learning_rate": 7.051328393276704e-06, "loss": 1.0085, "step": 8465 }, { "epoch": 1.53, "learning_rate": 7.0486173865895535e-06, "loss": 0.7288, "step": 8466 }, { "epoch": 1.53, "learning_rate": 7.0459063799024045e-06, "loss": 0.5697, "step": 8467 }, { "epoch": 1.53, "learning_rate": 7.043195373215254e-06, "loss": 0.9053, "step": 8468 }, { "epoch": 1.53, "learning_rate": 7.040484366528105e-06, "loss": 0.671, "step": 8469 }, { "epoch": 1.53, "learning_rate": 7.037773359840954e-06, "loss": 1.6087, "step": 8470 }, { "epoch": 1.53, "learning_rate": 7.035062353153805e-06, "loss": 1.4068, "step": 8471 }, { "epoch": 1.53, "learning_rate": 7.032351346466655e-06, "loss": 0.209, "step": 8472 }, { "epoch": 1.53, "learning_rate": 7.029640339779505e-06, "loss": 0.7848, "step": 8473 }, { "epoch": 1.53, "learning_rate": 7.026929333092355e-06, "loss": 0.4969, "step": 8474 }, { "epoch": 1.53, "learning_rate": 7.0242183264052055e-06, "loss": 0.3174, "step": 8475 }, { "epoch": 1.53, "learning_rate": 7.021507319718056e-06, "loss": 0.5952, "step": 8476 }, { "epoch": 1.53, "learning_rate": 7.018796313030906e-06, "loss": 0.7228, "step": 8477 }, { "epoch": 1.53, "learning_rate": 7.016085306343756e-06, "loss": 0.7072, "step": 8478 }, { "epoch": 1.53, "learning_rate": 7.013374299656606e-06, "loss": 0.7552, "step": 8479 }, { "epoch": 1.53, "learning_rate": 7.010663292969456e-06, "loss": 0.6156, "step": 8480 }, { "epoch": 1.53, "learning_rate": 7.007952286282306e-06, "loss": 0.6532, "step": 8481 }, { "epoch": 1.53, "learning_rate": 7.0052412795951565e-06, "loss": 0.7468, "step": 8482 }, { "epoch": 1.53, "learning_rate": 7.0025302729080075e-06, "loss": 1.342, "step": 8483 }, { "epoch": 1.53, "learning_rate": 6.999819266220857e-06, "loss": 0.791, "step": 8484 }, { "epoch": 1.53, "learning_rate": 6.997108259533708e-06, "loss": 1.6015, "step": 8485 }, { "epoch": 1.53, "learning_rate": 6.994397252846557e-06, "loss": 0.7344, "step": 8486 }, { "epoch": 1.53, "learning_rate": 6.991686246159408e-06, "loss": 0.5428, "step": 8487 }, { "epoch": 1.53, "learning_rate": 6.988975239472257e-06, "loss": 0.4249, "step": 8488 }, { "epoch": 1.53, "learning_rate": 6.986264232785108e-06, "loss": 0.355, "step": 8489 }, { "epoch": 1.53, "learning_rate": 6.9835532260979575e-06, "loss": 0.6635, "step": 8490 }, { "epoch": 1.53, "learning_rate": 6.9808422194108085e-06, "loss": 0.8606, "step": 8491 }, { "epoch": 1.53, "learning_rate": 6.978131212723658e-06, "loss": 0.9154, "step": 8492 }, { "epoch": 1.53, "learning_rate": 6.975420206036508e-06, "loss": 0.6457, "step": 8493 }, { "epoch": 1.54, "learning_rate": 6.972709199349359e-06, "loss": 0.5231, "step": 8494 }, { "epoch": 1.54, "learning_rate": 6.969998192662208e-06, "loss": 0.6569, "step": 8495 }, { "epoch": 1.54, "learning_rate": 6.967287185975059e-06, "loss": 0.343, "step": 8496 }, { "epoch": 1.54, "learning_rate": 6.9645761792879085e-06, "loss": 0.8034, "step": 8497 }, { "epoch": 1.54, "learning_rate": 6.9618651726007595e-06, "loss": 1.2258, "step": 8498 }, { "epoch": 1.54, "learning_rate": 6.959154165913609e-06, "loss": 0.8054, "step": 8499 }, { "epoch": 1.54, "learning_rate": 6.95644315922646e-06, "loss": 0.9699, "step": 8500 }, { "epoch": 1.54, "learning_rate": 6.953732152539309e-06, "loss": 1.1778, "step": 8501 }, { "epoch": 1.54, "learning_rate": 6.95102114585216e-06, "loss": 0.7753, "step": 8502 }, { "epoch": 1.54, "learning_rate": 6.94831013916501e-06, "loss": 0.4886, "step": 8503 }, { "epoch": 1.54, "learning_rate": 6.94559913247786e-06, "loss": 0.6477, "step": 8504 }, { "epoch": 1.54, "learning_rate": 6.94288812579071e-06, "loss": 0.3436, "step": 8505 }, { "epoch": 1.54, "learning_rate": 6.9401771191035605e-06, "loss": 0.7044, "step": 8506 }, { "epoch": 1.54, "learning_rate": 6.937466112416411e-06, "loss": 0.8918, "step": 8507 }, { "epoch": 1.54, "learning_rate": 6.934755105729261e-06, "loss": 0.3289, "step": 8508 }, { "epoch": 1.54, "learning_rate": 6.932044099042111e-06, "loss": 0.6832, "step": 8509 }, { "epoch": 1.54, "learning_rate": 6.929333092354961e-06, "loss": 0.8814, "step": 8510 }, { "epoch": 1.54, "learning_rate": 6.926622085667811e-06, "loss": 0.7411, "step": 8511 }, { "epoch": 1.54, "learning_rate": 6.923911078980661e-06, "loss": 0.4909, "step": 8512 }, { "epoch": 1.54, "learning_rate": 6.9212000722935115e-06, "loss": 0.7341, "step": 8513 }, { "epoch": 1.54, "learning_rate": 6.9184890656063625e-06, "loss": 0.7814, "step": 8514 }, { "epoch": 1.54, "learning_rate": 6.915778058919212e-06, "loss": 1.6679, "step": 8515 }, { "epoch": 1.54, "learning_rate": 6.913067052232063e-06, "loss": 0.616, "step": 8516 }, { "epoch": 1.54, "learning_rate": 6.910356045544912e-06, "loss": 0.7942, "step": 8517 }, { "epoch": 1.54, "learning_rate": 6.907645038857763e-06, "loss": 0.5756, "step": 8518 }, { "epoch": 1.54, "learning_rate": 6.904934032170612e-06, "loss": 0.7993, "step": 8519 }, { "epoch": 1.54, "learning_rate": 6.902223025483463e-06, "loss": 0.5581, "step": 8520 }, { "epoch": 1.54, "learning_rate": 6.8995120187963126e-06, "loss": 1.9331, "step": 8521 }, { "epoch": 1.54, "learning_rate": 6.8968010121091635e-06, "loss": 1.5581, "step": 8522 }, { "epoch": 1.54, "learning_rate": 6.894090005422013e-06, "loss": 1.1671, "step": 8523 }, { "epoch": 1.54, "learning_rate": 6.891378998734864e-06, "loss": 0.4594, "step": 8524 }, { "epoch": 1.54, "learning_rate": 6.888667992047714e-06, "loss": 0.7586, "step": 8525 }, { "epoch": 1.54, "learning_rate": 6.885956985360564e-06, "loss": 1.313, "step": 8526 }, { "epoch": 1.54, "learning_rate": 6.883245978673414e-06, "loss": 0.5281, "step": 8527 }, { "epoch": 1.54, "learning_rate": 6.880534971986264e-06, "loss": 0.5902, "step": 8528 }, { "epoch": 1.54, "learning_rate": 6.8778239652991145e-06, "loss": 1.5184, "step": 8529 }, { "epoch": 1.54, "learning_rate": 6.875112958611965e-06, "loss": 0.6719, "step": 8530 }, { "epoch": 1.54, "learning_rate": 6.872401951924815e-06, "loss": 1.6231, "step": 8531 }, { "epoch": 1.54, "learning_rate": 6.869690945237665e-06, "loss": 1.2459, "step": 8532 }, { "epoch": 1.54, "learning_rate": 6.866979938550515e-06, "loss": 0.8367, "step": 8533 }, { "epoch": 1.54, "learning_rate": 6.864268931863365e-06, "loss": 0.569, "step": 8534 }, { "epoch": 1.54, "learning_rate": 6.861557925176215e-06, "loss": 0.9468, "step": 8535 }, { "epoch": 1.54, "learning_rate": 6.858846918489066e-06, "loss": 0.4397, "step": 8536 }, { "epoch": 1.54, "learning_rate": 6.8561359118019156e-06, "loss": 0.5984, "step": 8537 }, { "epoch": 1.54, "learning_rate": 6.8534249051147665e-06, "loss": 0.5644, "step": 8538 }, { "epoch": 1.54, "learning_rate": 6.850713898427616e-06, "loss": 0.863, "step": 8539 }, { "epoch": 1.54, "learning_rate": 6.848002891740467e-06, "loss": 1.091, "step": 8540 }, { "epoch": 1.54, "learning_rate": 6.845291885053316e-06, "loss": 0.6102, "step": 8541 }, { "epoch": 1.54, "learning_rate": 6.842580878366167e-06, "loss": 0.4846, "step": 8542 }, { "epoch": 1.54, "learning_rate": 6.839869871679016e-06, "loss": 0.5137, "step": 8543 }, { "epoch": 1.54, "learning_rate": 6.837158864991867e-06, "loss": 0.8453, "step": 8544 }, { "epoch": 1.54, "learning_rate": 6.8344478583047175e-06, "loss": 0.9524, "step": 8545 }, { "epoch": 1.54, "learning_rate": 6.831736851617568e-06, "loss": 0.7491, "step": 8546 }, { "epoch": 1.54, "learning_rate": 6.829025844930418e-06, "loss": 0.7067, "step": 8547 }, { "epoch": 1.54, "learning_rate": 6.826314838243268e-06, "loss": 1.1253, "step": 8548 }, { "epoch": 1.55, "learning_rate": 6.823603831556118e-06, "loss": 0.6747, "step": 8549 }, { "epoch": 1.55, "learning_rate": 6.820892824868968e-06, "loss": 1.2118, "step": 8550 }, { "epoch": 1.55, "learning_rate": 6.818181818181818e-06, "loss": 1.0043, "step": 8551 }, { "epoch": 1.55, "learning_rate": 6.815470811494668e-06, "loss": 0.3632, "step": 8552 }, { "epoch": 1.55, "learning_rate": 6.8127598048075185e-06, "loss": 0.559, "step": 8553 }, { "epoch": 1.55, "learning_rate": 6.810048798120369e-06, "loss": 0.9913, "step": 8554 }, { "epoch": 1.55, "learning_rate": 6.807337791433219e-06, "loss": 1.0954, "step": 8555 }, { "epoch": 1.55, "learning_rate": 6.80462678474607e-06, "loss": 1.1664, "step": 8556 }, { "epoch": 1.55, "learning_rate": 6.801915778058919e-06, "loss": 1.1034, "step": 8557 }, { "epoch": 1.55, "learning_rate": 6.79920477137177e-06, "loss": 0.693, "step": 8558 }, { "epoch": 1.55, "learning_rate": 6.796493764684619e-06, "loss": 0.6172, "step": 8559 }, { "epoch": 1.55, "learning_rate": 6.79378275799747e-06, "loss": 0.5212, "step": 8560 }, { "epoch": 1.55, "learning_rate": 6.79107175131032e-06, "loss": 0.8275, "step": 8561 }, { "epoch": 1.55, "learning_rate": 6.788360744623171e-06, "loss": 0.5996, "step": 8562 }, { "epoch": 1.55, "learning_rate": 6.78564973793602e-06, "loss": 0.7489, "step": 8563 }, { "epoch": 1.55, "learning_rate": 6.782938731248871e-06, "loss": 0.777, "step": 8564 }, { "epoch": 1.55, "learning_rate": 6.78022772456172e-06, "loss": 0.4875, "step": 8565 }, { "epoch": 1.55, "learning_rate": 6.777516717874571e-06, "loss": 0.6299, "step": 8566 }, { "epoch": 1.55, "learning_rate": 6.774805711187421e-06, "loss": 0.4112, "step": 8567 }, { "epoch": 1.55, "learning_rate": 6.772094704500271e-06, "loss": 0.4625, "step": 8568 }, { "epoch": 1.55, "learning_rate": 6.7693836978131215e-06, "loss": 0.3211, "step": 8569 }, { "epoch": 1.55, "learning_rate": 6.766672691125972e-06, "loss": 0.6223, "step": 8570 }, { "epoch": 1.55, "learning_rate": 6.763961684438822e-06, "loss": 1.1397, "step": 8571 }, { "epoch": 1.55, "learning_rate": 6.761250677751672e-06, "loss": 0.6395, "step": 8572 }, { "epoch": 1.55, "learning_rate": 6.758539671064522e-06, "loss": 1.2524, "step": 8573 }, { "epoch": 1.55, "learning_rate": 6.755828664377372e-06, "loss": 0.7462, "step": 8574 }, { "epoch": 1.55, "learning_rate": 6.753117657690222e-06, "loss": 0.7295, "step": 8575 }, { "epoch": 1.55, "learning_rate": 6.750406651003073e-06, "loss": 0.4818, "step": 8576 }, { "epoch": 1.55, "learning_rate": 6.747695644315923e-06, "loss": 0.836, "step": 8577 }, { "epoch": 1.55, "learning_rate": 6.744984637628774e-06, "loss": 1.8775, "step": 8578 }, { "epoch": 1.55, "learning_rate": 6.742273630941623e-06, "loss": 0.4003, "step": 8579 }, { "epoch": 1.55, "learning_rate": 6.739562624254474e-06, "loss": 0.7783, "step": 8580 }, { "epoch": 1.55, "learning_rate": 6.736851617567323e-06, "loss": 0.8992, "step": 8581 }, { "epoch": 1.55, "learning_rate": 6.734140610880174e-06, "loss": 0.6447, "step": 8582 }, { "epoch": 1.55, "learning_rate": 6.731429604193023e-06, "loss": 0.657, "step": 8583 }, { "epoch": 1.55, "learning_rate": 6.728718597505874e-06, "loss": 1.1998, "step": 8584 }, { "epoch": 1.55, "learning_rate": 6.726007590818724e-06, "loss": 0.4485, "step": 8585 }, { "epoch": 1.55, "learning_rate": 6.723296584131575e-06, "loss": 0.4304, "step": 8586 }, { "epoch": 1.55, "learning_rate": 6.720585577444425e-06, "loss": 0.784, "step": 8587 }, { "epoch": 1.55, "learning_rate": 6.717874570757275e-06, "loss": 0.4407, "step": 8588 }, { "epoch": 1.55, "learning_rate": 6.715163564070125e-06, "loss": 0.9332, "step": 8589 }, { "epoch": 1.55, "learning_rate": 6.712452557382975e-06, "loss": 0.673, "step": 8590 }, { "epoch": 1.55, "learning_rate": 6.709741550695825e-06, "loss": 0.8703, "step": 8591 }, { "epoch": 1.55, "learning_rate": 6.7070305440086755e-06, "loss": 0.4873, "step": 8592 }, { "epoch": 1.55, "learning_rate": 6.704319537321526e-06, "loss": 0.6134, "step": 8593 }, { "epoch": 1.55, "learning_rate": 6.701608530634376e-06, "loss": 0.8319, "step": 8594 }, { "epoch": 1.55, "learning_rate": 6.698897523947226e-06, "loss": 0.7386, "step": 8595 }, { "epoch": 1.55, "learning_rate": 6.696186517260076e-06, "loss": 0.5986, "step": 8596 }, { "epoch": 1.55, "learning_rate": 6.693475510572926e-06, "loss": 1.0678, "step": 8597 }, { "epoch": 1.55, "learning_rate": 6.690764503885777e-06, "loss": 0.9933, "step": 8598 }, { "epoch": 1.55, "learning_rate": 6.688053497198626e-06, "loss": 0.7352, "step": 8599 }, { "epoch": 1.55, "learning_rate": 6.685342490511477e-06, "loss": 0.6493, "step": 8600 }, { "epoch": 1.55, "learning_rate": 6.682631483824327e-06, "loss": 1.0013, "step": 8601 }, { "epoch": 1.55, "learning_rate": 6.679920477137178e-06, "loss": 0.7824, "step": 8602 }, { "epoch": 1.55, "learning_rate": 6.677209470450027e-06, "loss": 0.9829, "step": 8603 }, { "epoch": 1.56, "learning_rate": 6.674498463762878e-06, "loss": 0.6249, "step": 8604 }, { "epoch": 1.56, "learning_rate": 6.671787457075727e-06, "loss": 1.1069, "step": 8605 }, { "epoch": 1.56, "learning_rate": 6.669076450388578e-06, "loss": 0.7548, "step": 8606 }, { "epoch": 1.56, "learning_rate": 6.666365443701428e-06, "loss": 0.8153, "step": 8607 }, { "epoch": 1.56, "learning_rate": 6.6636544370142785e-06, "loss": 0.5669, "step": 8608 }, { "epoch": 1.56, "learning_rate": 6.660943430327129e-06, "loss": 0.7197, "step": 8609 }, { "epoch": 1.56, "learning_rate": 6.658232423639979e-06, "loss": 1.1128, "step": 8610 }, { "epoch": 1.56, "learning_rate": 6.655521416952829e-06, "loss": 0.7135, "step": 8611 }, { "epoch": 1.56, "learning_rate": 6.652810410265679e-06, "loss": 0.9771, "step": 8612 }, { "epoch": 1.56, "learning_rate": 6.650099403578529e-06, "loss": 0.7621, "step": 8613 }, { "epoch": 1.56, "learning_rate": 6.647388396891379e-06, "loss": 0.3773, "step": 8614 }, { "epoch": 1.56, "learning_rate": 6.644677390204229e-06, "loss": 0.9806, "step": 8615 }, { "epoch": 1.56, "learning_rate": 6.6419663835170795e-06, "loss": 0.592, "step": 8616 }, { "epoch": 1.56, "learning_rate": 6.63925537682993e-06, "loss": 0.8536, "step": 8617 }, { "epoch": 1.56, "learning_rate": 6.636544370142781e-06, "loss": 0.9502, "step": 8618 }, { "epoch": 1.56, "learning_rate": 6.63383336345563e-06, "loss": 0.8085, "step": 8619 }, { "epoch": 1.56, "learning_rate": 6.631122356768481e-06, "loss": 0.853, "step": 8620 }, { "epoch": 1.56, "learning_rate": 6.62841135008133e-06, "loss": 0.648, "step": 8621 }, { "epoch": 1.56, "learning_rate": 6.62570034339418e-06, "loss": 0.3217, "step": 8622 }, { "epoch": 1.56, "learning_rate": 6.6229893367070305e-06, "loss": 0.8712, "step": 8623 }, { "epoch": 1.56, "learning_rate": 6.620278330019881e-06, "loss": 0.5077, "step": 8624 }, { "epoch": 1.56, "learning_rate": 6.617567323332731e-06, "loss": 1.1509, "step": 8625 }, { "epoch": 1.56, "learning_rate": 6.614856316645581e-06, "loss": 1.1332, "step": 8626 }, { "epoch": 1.56, "learning_rate": 6.612145309958431e-06, "loss": 0.6361, "step": 8627 }, { "epoch": 1.56, "learning_rate": 6.609434303271281e-06, "loss": 0.6616, "step": 8628 }, { "epoch": 1.56, "learning_rate": 6.606723296584132e-06, "loss": 1.0817, "step": 8629 }, { "epoch": 1.56, "learning_rate": 6.6040122898969814e-06, "loss": 0.7077, "step": 8630 }, { "epoch": 1.56, "learning_rate": 6.601301283209832e-06, "loss": 0.8557, "step": 8631 }, { "epoch": 1.56, "learning_rate": 6.598590276522682e-06, "loss": 0.5243, "step": 8632 }, { "epoch": 1.56, "learning_rate": 6.595879269835533e-06, "loss": 0.5715, "step": 8633 }, { "epoch": 1.56, "learning_rate": 6.593168263148382e-06, "loss": 0.5524, "step": 8634 }, { "epoch": 1.56, "learning_rate": 6.590457256461233e-06, "loss": 0.5164, "step": 8635 }, { "epoch": 1.56, "learning_rate": 6.587746249774082e-06, "loss": 0.7603, "step": 8636 }, { "epoch": 1.56, "learning_rate": 6.585035243086933e-06, "loss": 0.4789, "step": 8637 }, { "epoch": 1.56, "learning_rate": 6.582324236399783e-06, "loss": 0.4926, "step": 8638 }, { "epoch": 1.56, "learning_rate": 6.5796132297126335e-06, "loss": 0.7025, "step": 8639 }, { "epoch": 1.56, "learning_rate": 6.576902223025484e-06, "loss": 0.9108, "step": 8640 }, { "epoch": 1.56, "learning_rate": 6.574191216338334e-06, "loss": 0.8879, "step": 8641 }, { "epoch": 1.56, "learning_rate": 6.571480209651184e-06, "loss": 0.8828, "step": 8642 }, { "epoch": 1.56, "learning_rate": 6.568769202964034e-06, "loss": 1.3115, "step": 8643 }, { "epoch": 1.56, "learning_rate": 6.566058196276884e-06, "loss": 0.9246, "step": 8644 }, { "epoch": 1.56, "learning_rate": 6.563347189589734e-06, "loss": 0.3729, "step": 8645 }, { "epoch": 1.56, "learning_rate": 6.560636182902584e-06, "loss": 0.5538, "step": 8646 }, { "epoch": 1.56, "learning_rate": 6.5579251762154346e-06, "loss": 0.4809, "step": 8647 }, { "epoch": 1.56, "learning_rate": 6.555214169528285e-06, "loss": 1.13, "step": 8648 }, { "epoch": 1.56, "learning_rate": 6.552503162841136e-06, "loss": 0.6287, "step": 8649 }, { "epoch": 1.56, "learning_rate": 6.549792156153985e-06, "loss": 0.7703, "step": 8650 }, { "epoch": 1.56, "learning_rate": 6.547081149466836e-06, "loss": 0.6717, "step": 8651 }, { "epoch": 1.56, "learning_rate": 6.544370142779685e-06, "loss": 1.2531, "step": 8652 }, { "epoch": 1.56, "learning_rate": 6.541659136092536e-06, "loss": 0.7819, "step": 8653 }, { "epoch": 1.56, "learning_rate": 6.5389481294053855e-06, "loss": 0.5916, "step": 8654 }, { "epoch": 1.56, "learning_rate": 6.5362371227182365e-06, "loss": 0.4528, "step": 8655 }, { "epoch": 1.56, "learning_rate": 6.533526116031086e-06, "loss": 0.6232, "step": 8656 }, { "epoch": 1.56, "learning_rate": 6.530815109343937e-06, "loss": 0.6878, "step": 8657 }, { "epoch": 1.56, "learning_rate": 6.528104102656786e-06, "loss": 1.8885, "step": 8658 }, { "epoch": 1.56, "learning_rate": 6.525393095969637e-06, "loss": 0.8755, "step": 8659 }, { "epoch": 1.57, "learning_rate": 6.522682089282487e-06, "loss": 1.0469, "step": 8660 }, { "epoch": 1.57, "learning_rate": 6.519971082595337e-06, "loss": 1.356, "step": 8661 }, { "epoch": 1.57, "learning_rate": 6.517260075908187e-06, "loss": 0.471, "step": 8662 }, { "epoch": 1.57, "learning_rate": 6.5145490692210376e-06, "loss": 0.691, "step": 8663 }, { "epoch": 1.57, "learning_rate": 6.511838062533888e-06, "loss": 1.0728, "step": 8664 }, { "epoch": 1.57, "learning_rate": 6.509127055846738e-06, "loss": 0.6957, "step": 8665 }, { "epoch": 1.57, "learning_rate": 6.506416049159588e-06, "loss": 0.7217, "step": 8666 }, { "epoch": 1.57, "learning_rate": 6.503705042472438e-06, "loss": 0.8706, "step": 8667 }, { "epoch": 1.57, "learning_rate": 6.500994035785288e-06, "loss": 0.6522, "step": 8668 }, { "epoch": 1.57, "learning_rate": 6.498283029098138e-06, "loss": 0.4959, "step": 8669 }, { "epoch": 1.57, "learning_rate": 6.4955720224109885e-06, "loss": 0.6057, "step": 8670 }, { "epoch": 1.57, "learning_rate": 6.4928610157238395e-06, "loss": 0.781, "step": 8671 }, { "epoch": 1.57, "learning_rate": 6.490150009036689e-06, "loss": 0.5593, "step": 8672 }, { "epoch": 1.57, "learning_rate": 6.48743900234954e-06, "loss": 0.9888, "step": 8673 }, { "epoch": 1.57, "learning_rate": 6.484727995662389e-06, "loss": 0.2368, "step": 8674 }, { "epoch": 1.57, "learning_rate": 6.48201698897524e-06, "loss": 0.42, "step": 8675 }, { "epoch": 1.57, "learning_rate": 6.479305982288089e-06, "loss": 0.7071, "step": 8676 }, { "epoch": 1.57, "learning_rate": 6.47659497560094e-06, "loss": 1.1776, "step": 8677 }, { "epoch": 1.57, "learning_rate": 6.4738839689137896e-06, "loss": 0.7175, "step": 8678 }, { "epoch": 1.57, "learning_rate": 6.4711729622266405e-06, "loss": 1.2072, "step": 8679 }, { "epoch": 1.57, "learning_rate": 6.468461955539491e-06, "loss": 0.624, "step": 8680 }, { "epoch": 1.57, "learning_rate": 6.465750948852341e-06, "loss": 1.0989, "step": 8681 }, { "epoch": 1.57, "learning_rate": 6.463039942165191e-06, "loss": 0.7008, "step": 8682 }, { "epoch": 1.57, "learning_rate": 6.460328935478041e-06, "loss": 0.3796, "step": 8683 }, { "epoch": 1.57, "learning_rate": 6.457617928790891e-06, "loss": 0.5342, "step": 8684 }, { "epoch": 1.57, "learning_rate": 6.454906922103741e-06, "loss": 0.6983, "step": 8685 }, { "epoch": 1.57, "learning_rate": 6.4521959154165915e-06, "loss": 0.5084, "step": 8686 }, { "epoch": 1.57, "learning_rate": 6.449484908729442e-06, "loss": 0.533, "step": 8687 }, { "epoch": 1.57, "learning_rate": 6.446773902042292e-06, "loss": 0.9049, "step": 8688 }, { "epoch": 1.57, "learning_rate": 6.444062895355142e-06, "loss": 0.9739, "step": 8689 }, { "epoch": 1.57, "learning_rate": 6.441351888667992e-06, "loss": 0.6287, "step": 8690 }, { "epoch": 1.57, "learning_rate": 6.438640881980843e-06, "loss": 0.7055, "step": 8691 }, { "epoch": 1.57, "learning_rate": 6.435929875293692e-06, "loss": 0.8758, "step": 8692 }, { "epoch": 1.57, "learning_rate": 6.433218868606543e-06, "loss": 0.8909, "step": 8693 }, { "epoch": 1.57, "learning_rate": 6.4305078619193926e-06, "loss": 0.6485, "step": 8694 }, { "epoch": 1.57, "learning_rate": 6.4277968552322435e-06, "loss": 0.8112, "step": 8695 }, { "epoch": 1.57, "learning_rate": 6.425085848545093e-06, "loss": 0.4544, "step": 8696 }, { "epoch": 1.57, "learning_rate": 6.422374841857944e-06, "loss": 0.6546, "step": 8697 }, { "epoch": 1.57, "learning_rate": 6.419663835170793e-06, "loss": 0.6754, "step": 8698 }, { "epoch": 1.57, "learning_rate": 6.416952828483644e-06, "loss": 0.5298, "step": 8699 }, { "epoch": 1.57, "learning_rate": 6.414241821796493e-06, "loss": 0.2845, "step": 8700 }, { "epoch": 1.57, "learning_rate": 6.411530815109344e-06, "loss": 0.7181, "step": 8701 }, { "epoch": 1.57, "learning_rate": 6.4088198084221945e-06, "loss": 0.509, "step": 8702 }, { "epoch": 1.57, "learning_rate": 6.406108801735045e-06, "loss": 0.6829, "step": 8703 }, { "epoch": 1.57, "learning_rate": 6.403397795047895e-06, "loss": 0.3071, "step": 8704 }, { "epoch": 1.57, "learning_rate": 6.400686788360745e-06, "loss": 0.7668, "step": 8705 }, { "epoch": 1.57, "learning_rate": 6.397975781673595e-06, "loss": 0.7801, "step": 8706 }, { "epoch": 1.57, "learning_rate": 6.395264774986445e-06, "loss": 0.9597, "step": 8707 }, { "epoch": 1.57, "learning_rate": 6.392553768299295e-06, "loss": 0.6261, "step": 8708 }, { "epoch": 1.57, "learning_rate": 6.389842761612145e-06, "loss": 0.4166, "step": 8709 }, { "epoch": 1.57, "learning_rate": 6.3871317549249956e-06, "loss": 1.147, "step": 8710 }, { "epoch": 1.57, "learning_rate": 6.384420748237846e-06, "loss": 1.2534, "step": 8711 }, { "epoch": 1.57, "learning_rate": 6.381709741550696e-06, "loss": 0.9894, "step": 8712 }, { "epoch": 1.57, "learning_rate": 6.378998734863547e-06, "loss": 0.9123, "step": 8713 }, { "epoch": 1.57, "learning_rate": 6.376287728176396e-06, "loss": 0.7348, "step": 8714 }, { "epoch": 1.58, "learning_rate": 6.373576721489247e-06, "loss": 1.1816, "step": 8715 }, { "epoch": 1.58, "learning_rate": 6.370865714802096e-06, "loss": 0.8072, "step": 8716 }, { "epoch": 1.58, "learning_rate": 6.368154708114947e-06, "loss": 0.5074, "step": 8717 }, { "epoch": 1.58, "learning_rate": 6.365443701427797e-06, "loss": 0.9935, "step": 8718 }, { "epoch": 1.58, "learning_rate": 6.362732694740648e-06, "loss": 1.0297, "step": 8719 }, { "epoch": 1.58, "learning_rate": 6.360021688053497e-06, "loss": 0.798, "step": 8720 }, { "epoch": 1.58, "learning_rate": 6.357310681366348e-06, "loss": 0.9234, "step": 8721 }, { "epoch": 1.58, "learning_rate": 6.354599674679198e-06, "loss": 0.7363, "step": 8722 }, { "epoch": 1.58, "learning_rate": 6.351888667992048e-06, "loss": 0.6379, "step": 8723 }, { "epoch": 1.58, "learning_rate": 6.349177661304898e-06, "loss": 0.5436, "step": 8724 }, { "epoch": 1.58, "learning_rate": 6.346466654617748e-06, "loss": 0.617, "step": 8725 }, { "epoch": 1.58, "learning_rate": 6.3437556479305986e-06, "loss": 0.5502, "step": 8726 }, { "epoch": 1.58, "learning_rate": 6.341044641243449e-06, "loss": 1.0478, "step": 8727 }, { "epoch": 1.58, "learning_rate": 6.338333634556299e-06, "loss": 0.8771, "step": 8728 }, { "epoch": 1.58, "learning_rate": 6.335622627869149e-06, "loss": 0.6438, "step": 8729 }, { "epoch": 1.58, "learning_rate": 6.332911621181999e-06, "loss": 0.5523, "step": 8730 }, { "epoch": 1.58, "learning_rate": 6.330200614494849e-06, "loss": 0.9454, "step": 8731 }, { "epoch": 1.58, "learning_rate": 6.327489607807699e-06, "loss": 0.7091, "step": 8732 }, { "epoch": 1.58, "learning_rate": 6.32477860112055e-06, "loss": 0.5218, "step": 8733 }, { "epoch": 1.58, "learning_rate": 6.3220675944334e-06, "loss": 0.8287, "step": 8734 }, { "epoch": 1.58, "learning_rate": 6.319356587746251e-06, "loss": 0.4317, "step": 8735 }, { "epoch": 1.58, "learning_rate": 6.3166455810591e-06, "loss": 0.4041, "step": 8736 }, { "epoch": 1.58, "learning_rate": 6.313934574371951e-06, "loss": 0.7773, "step": 8737 }, { "epoch": 1.58, "learning_rate": 6.3112235676848e-06, "loss": 0.6891, "step": 8738 }, { "epoch": 1.58, "learning_rate": 6.308512560997651e-06, "loss": 0.4886, "step": 8739 }, { "epoch": 1.58, "learning_rate": 6.3058015543105004e-06, "loss": 0.5094, "step": 8740 }, { "epoch": 1.58, "learning_rate": 6.303090547623351e-06, "loss": 0.1469, "step": 8741 }, { "epoch": 1.58, "learning_rate": 6.300379540936201e-06, "loss": 1.0671, "step": 8742 }, { "epoch": 1.58, "learning_rate": 6.297668534249052e-06, "loss": 0.5897, "step": 8743 }, { "epoch": 1.58, "learning_rate": 6.294957527561902e-06, "loss": 1.0532, "step": 8744 }, { "epoch": 1.58, "learning_rate": 6.292246520874752e-06, "loss": 0.9879, "step": 8745 }, { "epoch": 1.58, "learning_rate": 6.289535514187602e-06, "loss": 0.9693, "step": 8746 }, { "epoch": 1.58, "learning_rate": 6.286824507500452e-06, "loss": 0.5236, "step": 8747 }, { "epoch": 1.58, "learning_rate": 6.284113500813302e-06, "loss": 0.7199, "step": 8748 }, { "epoch": 1.58, "learning_rate": 6.2814024941261525e-06, "loss": 0.3578, "step": 8749 }, { "epoch": 1.58, "learning_rate": 6.278691487439003e-06, "loss": 0.7723, "step": 8750 }, { "epoch": 1.58, "learning_rate": 6.275980480751852e-06, "loss": 0.7054, "step": 8751 }, { "epoch": 1.58, "learning_rate": 6.273269474064703e-06, "loss": 0.7483, "step": 8752 }, { "epoch": 1.58, "learning_rate": 6.270558467377553e-06, "loss": 0.6075, "step": 8753 }, { "epoch": 1.58, "learning_rate": 6.267847460690403e-06, "loss": 0.6399, "step": 8754 }, { "epoch": 1.58, "learning_rate": 6.265136454003253e-06, "loss": 0.5812, "step": 8755 }, { "epoch": 1.58, "learning_rate": 6.2624254473161034e-06, "loss": 0.5053, "step": 8756 }, { "epoch": 1.58, "learning_rate": 6.2597144406289536e-06, "loss": 0.5182, "step": 8757 }, { "epoch": 1.58, "learning_rate": 6.257003433941804e-06, "loss": 0.5506, "step": 8758 }, { "epoch": 1.58, "learning_rate": 6.254292427254654e-06, "loss": 1.297, "step": 8759 }, { "epoch": 1.58, "learning_rate": 6.251581420567504e-06, "loss": 0.8934, "step": 8760 }, { "epoch": 1.58, "learning_rate": 6.248870413880354e-06, "loss": 0.5544, "step": 8761 }, { "epoch": 1.58, "learning_rate": 6.246159407193204e-06, "loss": 1.0436, "step": 8762 }, { "epoch": 1.58, "learning_rate": 6.243448400506054e-06, "loss": 0.2801, "step": 8763 }, { "epoch": 1.58, "learning_rate": 6.240737393818905e-06, "loss": 0.5361, "step": 8764 }, { "epoch": 1.58, "learning_rate": 6.238026387131755e-06, "loss": 0.9106, "step": 8765 }, { "epoch": 1.58, "learning_rate": 6.235315380444606e-06, "loss": 0.4074, "step": 8766 }, { "epoch": 1.58, "learning_rate": 6.232604373757455e-06, "loss": 0.2991, "step": 8767 }, { "epoch": 1.58, "learning_rate": 6.229893367070306e-06, "loss": 0.4592, "step": 8768 }, { "epoch": 1.58, "learning_rate": 6.227182360383155e-06, "loss": 0.5734, "step": 8769 }, { "epoch": 1.59, "learning_rate": 6.224471353696006e-06, "loss": 0.9326, "step": 8770 }, { "epoch": 1.59, "learning_rate": 6.2217603470088554e-06, "loss": 0.546, "step": 8771 }, { "epoch": 1.59, "learning_rate": 6.219049340321706e-06, "loss": 0.5449, "step": 8772 }, { "epoch": 1.59, "learning_rate": 6.216338333634556e-06, "loss": 0.5751, "step": 8773 }, { "epoch": 1.59, "learning_rate": 6.213627326947407e-06, "loss": 1.0262, "step": 8774 }, { "epoch": 1.59, "learning_rate": 6.210916320260257e-06, "loss": 0.6826, "step": 8775 }, { "epoch": 1.59, "learning_rate": 6.208205313573107e-06, "loss": 0.718, "step": 8776 }, { "epoch": 1.59, "learning_rate": 6.205494306885957e-06, "loss": 1.0161, "step": 8777 }, { "epoch": 1.59, "learning_rate": 6.202783300198807e-06, "loss": 0.8099, "step": 8778 }, { "epoch": 1.59, "learning_rate": 6.200072293511657e-06, "loss": 0.6308, "step": 8779 }, { "epoch": 1.59, "learning_rate": 6.1973612868245075e-06, "loss": 0.5657, "step": 8780 }, { "epoch": 1.59, "learning_rate": 6.194650280137358e-06, "loss": 0.4828, "step": 8781 }, { "epoch": 1.59, "learning_rate": 6.191939273450208e-06, "loss": 1.3101, "step": 8782 }, { "epoch": 1.59, "learning_rate": 6.189228266763058e-06, "loss": 0.9839, "step": 8783 }, { "epoch": 1.59, "learning_rate": 6.186517260075908e-06, "loss": 0.8069, "step": 8784 }, { "epoch": 1.59, "learning_rate": 6.183806253388758e-06, "loss": 0.757, "step": 8785 }, { "epoch": 1.59, "learning_rate": 6.181095246701609e-06, "loss": 0.6693, "step": 8786 }, { "epoch": 1.59, "learning_rate": 6.1783842400144584e-06, "loss": 0.7423, "step": 8787 }, { "epoch": 1.59, "learning_rate": 6.175673233327309e-06, "loss": 1.0856, "step": 8788 }, { "epoch": 1.59, "learning_rate": 6.172962226640159e-06, "loss": 0.6278, "step": 8789 }, { "epoch": 1.59, "learning_rate": 6.17025121995301e-06, "loss": 0.4967, "step": 8790 }, { "epoch": 1.59, "learning_rate": 6.167540213265859e-06, "loss": 1.2782, "step": 8791 }, { "epoch": 1.59, "learning_rate": 6.16482920657871e-06, "loss": 0.6197, "step": 8792 }, { "epoch": 1.59, "learning_rate": 6.162118199891559e-06, "loss": 0.6058, "step": 8793 }, { "epoch": 1.59, "learning_rate": 6.15940719320441e-06, "loss": 0.4606, "step": 8794 }, { "epoch": 1.59, "learning_rate": 6.15669618651726e-06, "loss": 0.8129, "step": 8795 }, { "epoch": 1.59, "learning_rate": 6.1539851798301105e-06, "loss": 0.3179, "step": 8796 }, { "epoch": 1.59, "learning_rate": 6.151274173142961e-06, "loss": 1.268, "step": 8797 }, { "epoch": 1.59, "learning_rate": 6.148563166455811e-06, "loss": 0.816, "step": 8798 }, { "epoch": 1.59, "learning_rate": 6.145852159768661e-06, "loss": 0.4757, "step": 8799 }, { "epoch": 1.59, "learning_rate": 6.143141153081511e-06, "loss": 0.6696, "step": 8800 }, { "epoch": 1.59, "learning_rate": 6.140430146394361e-06, "loss": 0.3763, "step": 8801 }, { "epoch": 1.59, "learning_rate": 6.137719139707211e-06, "loss": 1.0589, "step": 8802 }, { "epoch": 1.59, "learning_rate": 6.1350081330200614e-06, "loss": 0.531, "step": 8803 }, { "epoch": 1.59, "learning_rate": 6.1322971263329116e-06, "loss": 0.5036, "step": 8804 }, { "epoch": 1.59, "learning_rate": 6.129586119645762e-06, "loss": 0.7865, "step": 8805 }, { "epoch": 1.59, "learning_rate": 6.126875112958613e-06, "loss": 0.5736, "step": 8806 }, { "epoch": 1.59, "learning_rate": 6.124164106271462e-06, "loss": 0.5102, "step": 8807 }, { "epoch": 1.59, "learning_rate": 6.121453099584313e-06, "loss": 0.7812, "step": 8808 }, { "epoch": 1.59, "learning_rate": 6.118742092897162e-06, "loss": 0.6978, "step": 8809 }, { "epoch": 1.59, "learning_rate": 6.116031086210013e-06, "loss": 0.6072, "step": 8810 }, { "epoch": 1.59, "learning_rate": 6.1133200795228625e-06, "loss": 1.086, "step": 8811 }, { "epoch": 1.59, "learning_rate": 6.1106090728357135e-06, "loss": 0.9041, "step": 8812 }, { "epoch": 1.59, "learning_rate": 6.107898066148563e-06, "loss": 0.7278, "step": 8813 }, { "epoch": 1.59, "learning_rate": 6.105187059461414e-06, "loss": 0.4973, "step": 8814 }, { "epoch": 1.59, "learning_rate": 6.102476052774263e-06, "loss": 0.5898, "step": 8815 }, { "epoch": 1.59, "learning_rate": 6.099765046087114e-06, "loss": 0.6626, "step": 8816 }, { "epoch": 1.59, "learning_rate": 6.097054039399964e-06, "loss": 0.5354, "step": 8817 }, { "epoch": 1.59, "learning_rate": 6.094343032712814e-06, "loss": 0.6949, "step": 8818 }, { "epoch": 1.59, "learning_rate": 6.0916320260256644e-06, "loss": 0.5048, "step": 8819 }, { "epoch": 1.59, "learning_rate": 6.0889210193385146e-06, "loss": 0.4692, "step": 8820 }, { "epoch": 1.59, "learning_rate": 6.086210012651365e-06, "loss": 0.8737, "step": 8821 }, { "epoch": 1.59, "learning_rate": 6.083499005964215e-06, "loss": 0.1897, "step": 8822 }, { "epoch": 1.59, "learning_rate": 6.080787999277065e-06, "loss": 1.5147, "step": 8823 }, { "epoch": 1.59, "learning_rate": 6.078076992589915e-06, "loss": 0.3551, "step": 8824 }, { "epoch": 1.59, "learning_rate": 6.075365985902765e-06, "loss": 1.0718, "step": 8825 }, { "epoch": 1.6, "learning_rate": 6.072654979215616e-06, "loss": 0.7521, "step": 8826 }, { "epoch": 1.6, "learning_rate": 6.0699439725284655e-06, "loss": 0.8257, "step": 8827 }, { "epoch": 1.6, "learning_rate": 6.0672329658413165e-06, "loss": 0.7786, "step": 8828 }, { "epoch": 1.6, "learning_rate": 6.064521959154166e-06, "loss": 0.5013, "step": 8829 }, { "epoch": 1.6, "learning_rate": 6.061810952467017e-06, "loss": 0.5438, "step": 8830 }, { "epoch": 1.6, "learning_rate": 6.059099945779866e-06, "loss": 0.9802, "step": 8831 }, { "epoch": 1.6, "learning_rate": 6.056388939092717e-06, "loss": 0.7648, "step": 8832 }, { "epoch": 1.6, "learning_rate": 6.053677932405566e-06, "loss": 0.5853, "step": 8833 }, { "epoch": 1.6, "learning_rate": 6.050966925718417e-06, "loss": 0.4936, "step": 8834 }, { "epoch": 1.6, "learning_rate": 6.0482559190312666e-06, "loss": 0.8174, "step": 8835 }, { "epoch": 1.6, "learning_rate": 6.0455449123441176e-06, "loss": 1.0442, "step": 8836 }, { "epoch": 1.6, "learning_rate": 6.042833905656968e-06, "loss": 0.7055, "step": 8837 }, { "epoch": 1.6, "learning_rate": 6.040122898969818e-06, "loss": 0.4862, "step": 8838 }, { "epoch": 1.6, "learning_rate": 6.037411892282668e-06, "loss": 1.451, "step": 8839 }, { "epoch": 1.6, "learning_rate": 6.034700885595518e-06, "loss": 0.3204, "step": 8840 }, { "epoch": 1.6, "learning_rate": 6.031989878908368e-06, "loss": 1.0529, "step": 8841 }, { "epoch": 1.6, "learning_rate": 6.029278872221218e-06, "loss": 0.8512, "step": 8842 }, { "epoch": 1.6, "learning_rate": 6.0265678655340685e-06, "loss": 0.7907, "step": 8843 }, { "epoch": 1.6, "learning_rate": 6.023856858846919e-06, "loss": 0.6741, "step": 8844 }, { "epoch": 1.6, "learning_rate": 6.021145852159769e-06, "loss": 0.4705, "step": 8845 }, { "epoch": 1.6, "learning_rate": 6.018434845472619e-06, "loss": 0.7518, "step": 8846 }, { "epoch": 1.6, "learning_rate": 6.015723838785469e-06, "loss": 1.0986, "step": 8847 }, { "epoch": 1.6, "learning_rate": 6.01301283209832e-06, "loss": 0.6649, "step": 8848 }, { "epoch": 1.6, "learning_rate": 6.010301825411169e-06, "loss": 0.7486, "step": 8849 }, { "epoch": 1.6, "learning_rate": 6.00759081872402e-06, "loss": 0.5806, "step": 8850 }, { "epoch": 1.6, "learning_rate": 6.0048798120368696e-06, "loss": 1.1544, "step": 8851 }, { "epoch": 1.6, "learning_rate": 6.0021688053497206e-06, "loss": 0.9651, "step": 8852 }, { "epoch": 1.6, "learning_rate": 5.99945779866257e-06, "loss": 0.6697, "step": 8853 }, { "epoch": 1.6, "learning_rate": 5.996746791975421e-06, "loss": 1.5782, "step": 8854 }, { "epoch": 1.6, "learning_rate": 5.99403578528827e-06, "loss": 1.5109, "step": 8855 }, { "epoch": 1.6, "learning_rate": 5.991324778601121e-06, "loss": 0.7533, "step": 8856 }, { "epoch": 1.6, "learning_rate": 5.988613771913971e-06, "loss": 0.629, "step": 8857 }, { "epoch": 1.6, "learning_rate": 5.985902765226821e-06, "loss": 0.632, "step": 8858 }, { "epoch": 1.6, "learning_rate": 5.9831917585396715e-06, "loss": 0.3751, "step": 8859 }, { "epoch": 1.6, "learning_rate": 5.980480751852522e-06, "loss": 0.4738, "step": 8860 }, { "epoch": 1.6, "learning_rate": 5.977769745165372e-06, "loss": 0.631, "step": 8861 }, { "epoch": 1.6, "learning_rate": 5.975058738478222e-06, "loss": 1.4597, "step": 8862 }, { "epoch": 1.6, "learning_rate": 5.972347731791072e-06, "loss": 0.4903, "step": 8863 }, { "epoch": 1.6, "learning_rate": 5.969636725103922e-06, "loss": 1.0206, "step": 8864 }, { "epoch": 1.6, "learning_rate": 5.966925718416772e-06, "loss": 0.9624, "step": 8865 }, { "epoch": 1.6, "learning_rate": 5.9642147117296224e-06, "loss": 0.9267, "step": 8866 }, { "epoch": 1.6, "learning_rate": 5.9615037050424726e-06, "loss": 0.3251, "step": 8867 }, { "epoch": 1.6, "learning_rate": 5.9587926983553235e-06, "loss": 0.654, "step": 8868 }, { "epoch": 1.6, "learning_rate": 5.956081691668173e-06, "loss": 0.5509, "step": 8869 }, { "epoch": 1.6, "learning_rate": 5.953370684981024e-06, "loss": 1.1019, "step": 8870 }, { "epoch": 1.6, "learning_rate": 5.950659678293873e-06, "loss": 0.9747, "step": 8871 }, { "epoch": 1.6, "learning_rate": 5.947948671606724e-06, "loss": 0.527, "step": 8872 }, { "epoch": 1.6, "learning_rate": 5.945237664919573e-06, "loss": 0.7298, "step": 8873 }, { "epoch": 1.6, "learning_rate": 5.942526658232424e-06, "loss": 0.7865, "step": 8874 }, { "epoch": 1.6, "learning_rate": 5.939815651545274e-06, "loss": 0.8983, "step": 8875 }, { "epoch": 1.6, "learning_rate": 5.937104644858125e-06, "loss": 0.9977, "step": 8876 }, { "epoch": 1.6, "learning_rate": 5.934393638170974e-06, "loss": 0.7884, "step": 8877 }, { "epoch": 1.6, "learning_rate": 5.931682631483825e-06, "loss": 1.0159, "step": 8878 }, { "epoch": 1.6, "learning_rate": 5.928971624796675e-06, "loss": 0.8029, "step": 8879 }, { "epoch": 1.6, "learning_rate": 5.926260618109524e-06, "loss": 0.6148, "step": 8880 }, { "epoch": 1.61, "learning_rate": 5.923549611422375e-06, "loss": 0.4329, "step": 8881 }, { "epoch": 1.61, "learning_rate": 5.920838604735225e-06, "loss": 0.2583, "step": 8882 }, { "epoch": 1.61, "learning_rate": 5.9181275980480756e-06, "loss": 0.4888, "step": 8883 }, { "epoch": 1.61, "learning_rate": 5.915416591360925e-06, "loss": 1.0132, "step": 8884 }, { "epoch": 1.61, "learning_rate": 5.912705584673776e-06, "loss": 0.6623, "step": 8885 }, { "epoch": 1.61, "learning_rate": 5.909994577986625e-06, "loss": 0.2688, "step": 8886 }, { "epoch": 1.61, "learning_rate": 5.907283571299476e-06, "loss": 0.6678, "step": 8887 }, { "epoch": 1.61, "learning_rate": 5.904572564612325e-06, "loss": 0.7663, "step": 8888 }, { "epoch": 1.61, "learning_rate": 5.901861557925176e-06, "loss": 1.2094, "step": 8889 }, { "epoch": 1.61, "learning_rate": 5.8991505512380265e-06, "loss": 0.6238, "step": 8890 }, { "epoch": 1.61, "learning_rate": 5.896439544550877e-06, "loss": 0.9928, "step": 8891 }, { "epoch": 1.61, "learning_rate": 5.893728537863727e-06, "loss": 0.8429, "step": 8892 }, { "epoch": 1.61, "learning_rate": 5.891017531176577e-06, "loss": 0.9018, "step": 8893 }, { "epoch": 1.61, "learning_rate": 5.888306524489427e-06, "loss": 0.3631, "step": 8894 }, { "epoch": 1.61, "learning_rate": 5.885595517802277e-06, "loss": 0.7923, "step": 8895 }, { "epoch": 1.61, "learning_rate": 5.882884511115127e-06, "loss": 0.8326, "step": 8896 }, { "epoch": 1.61, "learning_rate": 5.8801735044279774e-06, "loss": 0.3312, "step": 8897 }, { "epoch": 1.61, "learning_rate": 5.877462497740828e-06, "loss": 0.809, "step": 8898 }, { "epoch": 1.61, "learning_rate": 5.8747514910536786e-06, "loss": 1.03, "step": 8899 }, { "epoch": 1.61, "learning_rate": 5.872040484366528e-06, "loss": 0.468, "step": 8900 }, { "epoch": 1.61, "learning_rate": 5.869329477679379e-06, "loss": 0.6469, "step": 8901 }, { "epoch": 1.61, "learning_rate": 5.866618470992228e-06, "loss": 0.4146, "step": 8902 }, { "epoch": 1.61, "learning_rate": 5.863907464305079e-06, "loss": 0.9671, "step": 8903 }, { "epoch": 1.61, "learning_rate": 5.861196457617928e-06, "loss": 1.0443, "step": 8904 }, { "epoch": 1.61, "learning_rate": 5.858485450930779e-06, "loss": 0.6905, "step": 8905 }, { "epoch": 1.61, "learning_rate": 5.855774444243629e-06, "loss": 0.6864, "step": 8906 }, { "epoch": 1.61, "learning_rate": 5.85306343755648e-06, "loss": 0.6882, "step": 8907 }, { "epoch": 1.61, "learning_rate": 5.850352430869329e-06, "loss": 0.6852, "step": 8908 }, { "epoch": 1.61, "learning_rate": 5.84764142418218e-06, "loss": 0.4812, "step": 8909 }, { "epoch": 1.61, "learning_rate": 5.84493041749503e-06, "loss": 0.4285, "step": 8910 }, { "epoch": 1.61, "learning_rate": 5.84221941080788e-06, "loss": 0.4574, "step": 8911 }, { "epoch": 1.61, "learning_rate": 5.83950840412073e-06, "loss": 0.9009, "step": 8912 }, { "epoch": 1.61, "learning_rate": 5.8367973974335804e-06, "loss": 0.909, "step": 8913 }, { "epoch": 1.61, "learning_rate": 5.8340863907464306e-06, "loss": 1.2897, "step": 8914 }, { "epoch": 1.61, "learning_rate": 5.831375384059281e-06, "loss": 0.495, "step": 8915 }, { "epoch": 1.61, "learning_rate": 5.828664377372131e-06, "loss": 0.5055, "step": 8916 }, { "epoch": 1.61, "learning_rate": 5.825953370684981e-06, "loss": 0.6995, "step": 8917 }, { "epoch": 1.61, "learning_rate": 5.823242363997831e-06, "loss": 0.5062, "step": 8918 }, { "epoch": 1.61, "learning_rate": 5.820531357310681e-06, "loss": 0.5657, "step": 8919 }, { "epoch": 1.61, "learning_rate": 5.817820350623531e-06, "loss": 0.8868, "step": 8920 }, { "epoch": 1.61, "learning_rate": 5.815109343936382e-06, "loss": 1.5595, "step": 8921 }, { "epoch": 1.61, "learning_rate": 5.812398337249232e-06, "loss": 0.8119, "step": 8922 }, { "epoch": 1.61, "learning_rate": 5.809687330562083e-06, "loss": 0.5477, "step": 8923 }, { "epoch": 1.61, "learning_rate": 5.806976323874932e-06, "loss": 1.5898, "step": 8924 }, { "epoch": 1.61, "learning_rate": 5.804265317187783e-06, "loss": 0.6719, "step": 8925 }, { "epoch": 1.61, "learning_rate": 5.801554310500632e-06, "loss": 0.9601, "step": 8926 }, { "epoch": 1.61, "learning_rate": 5.798843303813483e-06, "loss": 0.4697, "step": 8927 }, { "epoch": 1.61, "learning_rate": 5.7961322971263325e-06, "loss": 1.0032, "step": 8928 }, { "epoch": 1.61, "learning_rate": 5.7934212904391834e-06, "loss": 0.4334, "step": 8929 }, { "epoch": 1.61, "learning_rate": 5.7907102837520336e-06, "loss": 0.9229, "step": 8930 }, { "epoch": 1.61, "learning_rate": 5.787999277064884e-06, "loss": 0.3579, "step": 8931 }, { "epoch": 1.61, "learning_rate": 5.785288270377734e-06, "loss": 1.1131, "step": 8932 }, { "epoch": 1.61, "learning_rate": 5.782577263690584e-06, "loss": 1.2466, "step": 8933 }, { "epoch": 1.61, "learning_rate": 5.779866257003434e-06, "loss": 0.8599, "step": 8934 }, { "epoch": 1.61, "learning_rate": 5.777155250316284e-06, "loss": 0.9328, "step": 8935 }, { "epoch": 1.62, "learning_rate": 5.774444243629134e-06, "loss": 0.3379, "step": 8936 }, { "epoch": 1.62, "learning_rate": 5.7717332369419845e-06, "loss": 0.8112, "step": 8937 }, { "epoch": 1.62, "learning_rate": 5.769022230254835e-06, "loss": 0.8241, "step": 8938 }, { "epoch": 1.62, "learning_rate": 5.766311223567685e-06, "loss": 0.9322, "step": 8939 }, { "epoch": 1.62, "learning_rate": 5.763600216880535e-06, "loss": 1.472, "step": 8940 }, { "epoch": 1.62, "learning_rate": 5.760889210193386e-06, "loss": 0.3924, "step": 8941 }, { "epoch": 1.62, "learning_rate": 5.758178203506235e-06, "loss": 1.0491, "step": 8942 }, { "epoch": 1.62, "learning_rate": 5.755467196819086e-06, "loss": 1.0881, "step": 8943 }, { "epoch": 1.62, "learning_rate": 5.7527561901319354e-06, "loss": 0.4402, "step": 8944 }, { "epoch": 1.62, "learning_rate": 5.7500451834447864e-06, "loss": 0.8205, "step": 8945 }, { "epoch": 1.62, "learning_rate": 5.747334176757636e-06, "loss": 0.6428, "step": 8946 }, { "epoch": 1.62, "learning_rate": 5.744623170070487e-06, "loss": 0.7527, "step": 8947 }, { "epoch": 1.62, "learning_rate": 5.741912163383336e-06, "loss": 0.577, "step": 8948 }, { "epoch": 1.62, "learning_rate": 5.739201156696187e-06, "loss": 1.0935, "step": 8949 }, { "epoch": 1.62, "learning_rate": 5.736490150009036e-06, "loss": 0.5463, "step": 8950 }, { "epoch": 1.62, "learning_rate": 5.733779143321887e-06, "loss": 0.6127, "step": 8951 }, { "epoch": 1.62, "learning_rate": 5.731068136634737e-06, "loss": 0.2697, "step": 8952 }, { "epoch": 1.62, "learning_rate": 5.7283571299475875e-06, "loss": 0.9077, "step": 8953 }, { "epoch": 1.62, "learning_rate": 5.725646123260438e-06, "loss": 0.7328, "step": 8954 }, { "epoch": 1.62, "learning_rate": 5.722935116573288e-06, "loss": 0.6086, "step": 8955 }, { "epoch": 1.62, "learning_rate": 5.720224109886138e-06, "loss": 0.4658, "step": 8956 }, { "epoch": 1.62, "learning_rate": 5.717513103198988e-06, "loss": 0.3283, "step": 8957 }, { "epoch": 1.62, "learning_rate": 5.714802096511838e-06, "loss": 0.9804, "step": 8958 }, { "epoch": 1.62, "learning_rate": 5.712091089824688e-06, "loss": 0.7795, "step": 8959 }, { "epoch": 1.62, "learning_rate": 5.7093800831375384e-06, "loss": 1.1836, "step": 8960 }, { "epoch": 1.62, "learning_rate": 5.706669076450389e-06, "loss": 0.3803, "step": 8961 }, { "epoch": 1.62, "learning_rate": 5.703958069763239e-06, "loss": 0.9068, "step": 8962 }, { "epoch": 1.62, "learning_rate": 5.70124706307609e-06, "loss": 0.524, "step": 8963 }, { "epoch": 1.62, "learning_rate": 5.698536056388939e-06, "loss": 1.0942, "step": 8964 }, { "epoch": 1.62, "learning_rate": 5.69582504970179e-06, "loss": 1.4476, "step": 8965 }, { "epoch": 1.62, "learning_rate": 5.693114043014639e-06, "loss": 0.6086, "step": 8966 }, { "epoch": 1.62, "learning_rate": 5.69040303632749e-06, "loss": 0.8255, "step": 8967 }, { "epoch": 1.62, "learning_rate": 5.6876920296403395e-06, "loss": 0.3942, "step": 8968 }, { "epoch": 1.62, "learning_rate": 5.6849810229531905e-06, "loss": 0.7527, "step": 8969 }, { "epoch": 1.62, "learning_rate": 5.68227001626604e-06, "loss": 0.4681, "step": 8970 }, { "epoch": 1.62, "learning_rate": 5.679559009578891e-06, "loss": 0.3638, "step": 8971 }, { "epoch": 1.62, "learning_rate": 5.676848002891741e-06, "loss": 0.8736, "step": 8972 }, { "epoch": 1.62, "learning_rate": 5.674136996204591e-06, "loss": 0.532, "step": 8973 }, { "epoch": 1.62, "learning_rate": 5.671425989517441e-06, "loss": 0.7048, "step": 8974 }, { "epoch": 1.62, "learning_rate": 5.668714982830291e-06, "loss": 0.7523, "step": 8975 }, { "epoch": 1.62, "learning_rate": 5.6660039761431414e-06, "loss": 1.3392, "step": 8976 }, { "epoch": 1.62, "learning_rate": 5.6632929694559916e-06, "loss": 0.8277, "step": 8977 }, { "epoch": 1.62, "learning_rate": 5.660581962768842e-06, "loss": 0.5555, "step": 8978 }, { "epoch": 1.62, "learning_rate": 5.657870956081692e-06, "loss": 0.3566, "step": 8979 }, { "epoch": 1.62, "learning_rate": 5.655159949394542e-06, "loss": 0.2832, "step": 8980 }, { "epoch": 1.62, "learning_rate": 5.652448942707392e-06, "loss": 0.9535, "step": 8981 }, { "epoch": 1.62, "learning_rate": 5.649737936020242e-06, "loss": 0.7351, "step": 8982 }, { "epoch": 1.62, "learning_rate": 5.647026929333093e-06, "loss": 0.4079, "step": 8983 }, { "epoch": 1.62, "learning_rate": 5.6443159226459425e-06, "loss": 0.6665, "step": 8984 }, { "epoch": 1.62, "learning_rate": 5.6416049159587935e-06, "loss": 0.8965, "step": 8985 }, { "epoch": 1.62, "learning_rate": 5.638893909271643e-06, "loss": 0.415, "step": 8986 }, { "epoch": 1.62, "learning_rate": 5.636182902584494e-06, "loss": 0.523, "step": 8987 }, { "epoch": 1.62, "learning_rate": 5.633471895897343e-06, "loss": 1.0581, "step": 8988 }, { "epoch": 1.62, "learning_rate": 5.630760889210194e-06, "loss": 0.3374, "step": 8989 }, { "epoch": 1.62, "learning_rate": 5.628049882523043e-06, "loss": 0.9544, "step": 8990 }, { "epoch": 1.62, "learning_rate": 5.625338875835894e-06, "loss": 0.4447, "step": 8991 }, { "epoch": 1.63, "learning_rate": 5.622627869148744e-06, "loss": 0.5051, "step": 8992 }, { "epoch": 1.63, "learning_rate": 5.6199168624615946e-06, "loss": 0.785, "step": 8993 }, { "epoch": 1.63, "learning_rate": 5.617205855774445e-06, "loss": 0.8372, "step": 8994 }, { "epoch": 1.63, "learning_rate": 5.614494849087295e-06, "loss": 0.4577, "step": 8995 }, { "epoch": 1.63, "learning_rate": 5.611783842400145e-06, "loss": 0.5649, "step": 8996 }, { "epoch": 1.63, "learning_rate": 5.609072835712995e-06, "loss": 0.8379, "step": 8997 }, { "epoch": 1.63, "learning_rate": 5.606361829025845e-06, "loss": 0.5767, "step": 8998 }, { "epoch": 1.63, "learning_rate": 5.603650822338695e-06, "loss": 0.9685, "step": 8999 }, { "epoch": 1.63, "learning_rate": 5.6009398156515455e-06, "loss": 0.1959, "step": 9000 }, { "epoch": 1.63, "learning_rate": 5.598228808964396e-06, "loss": 0.3945, "step": 9001 }, { "epoch": 1.63, "learning_rate": 5.595517802277246e-06, "loss": 1.0678, "step": 9002 }, { "epoch": 1.63, "learning_rate": 5.592806795590096e-06, "loss": 0.9341, "step": 9003 }, { "epoch": 1.63, "learning_rate": 5.590095788902946e-06, "loss": 0.9933, "step": 9004 }, { "epoch": 1.63, "learning_rate": 5.587384782215797e-06, "loss": 0.288, "step": 9005 }, { "epoch": 1.63, "learning_rate": 5.584673775528646e-06, "loss": 0.7586, "step": 9006 }, { "epoch": 1.63, "learning_rate": 5.581962768841497e-06, "loss": 0.5916, "step": 9007 }, { "epoch": 1.63, "learning_rate": 5.579251762154347e-06, "loss": 0.3848, "step": 9008 }, { "epoch": 1.63, "learning_rate": 5.576540755467197e-06, "loss": 0.5213, "step": 9009 }, { "epoch": 1.63, "learning_rate": 5.573829748780047e-06, "loss": 0.676, "step": 9010 }, { "epoch": 1.63, "learning_rate": 5.571118742092897e-06, "loss": 0.2758, "step": 9011 }, { "epoch": 1.63, "learning_rate": 5.568407735405747e-06, "loss": 0.4299, "step": 9012 }, { "epoch": 1.63, "learning_rate": 5.565696728718597e-06, "loss": 0.5613, "step": 9013 }, { "epoch": 1.63, "learning_rate": 5.562985722031448e-06, "loss": 0.5301, "step": 9014 }, { "epoch": 1.63, "learning_rate": 5.5602747153442975e-06, "loss": 0.5746, "step": 9015 }, { "epoch": 1.63, "learning_rate": 5.5575637086571485e-06, "loss": 0.9516, "step": 9016 }, { "epoch": 1.63, "learning_rate": 5.554852701969998e-06, "loss": 0.5654, "step": 9017 }, { "epoch": 1.63, "learning_rate": 5.552141695282849e-06, "loss": 0.8581, "step": 9018 }, { "epoch": 1.63, "learning_rate": 5.549430688595698e-06, "loss": 0.613, "step": 9019 }, { "epoch": 1.63, "learning_rate": 5.546719681908549e-06, "loss": 0.2718, "step": 9020 }, { "epoch": 1.63, "learning_rate": 5.544008675221398e-06, "loss": 0.6314, "step": 9021 }, { "epoch": 1.63, "learning_rate": 5.541297668534249e-06, "loss": 0.8215, "step": 9022 }, { "epoch": 1.63, "learning_rate": 5.538586661847099e-06, "loss": 0.526, "step": 9023 }, { "epoch": 1.63, "learning_rate": 5.53587565515995e-06, "loss": 0.5649, "step": 9024 }, { "epoch": 1.63, "learning_rate": 5.5331646484728e-06, "loss": 0.384, "step": 9025 }, { "epoch": 1.63, "learning_rate": 5.53045364178565e-06, "loss": 0.6154, "step": 9026 }, { "epoch": 1.63, "learning_rate": 5.5277426350985e-06, "loss": 0.5309, "step": 9027 }, { "epoch": 1.63, "learning_rate": 5.52503162841135e-06, "loss": 0.9858, "step": 9028 }, { "epoch": 1.63, "learning_rate": 5.5223206217242e-06, "loss": 0.5709, "step": 9029 }, { "epoch": 1.63, "learning_rate": 5.51960961503705e-06, "loss": 0.5436, "step": 9030 }, { "epoch": 1.63, "learning_rate": 5.5168986083499005e-06, "loss": 0.6554, "step": 9031 }, { "epoch": 1.63, "learning_rate": 5.514187601662751e-06, "loss": 0.9473, "step": 9032 }, { "epoch": 1.63, "learning_rate": 5.511476594975601e-06, "loss": 1.3622, "step": 9033 }, { "epoch": 1.63, "learning_rate": 5.508765588288451e-06, "loss": 0.5045, "step": 9034 }, { "epoch": 1.63, "learning_rate": 5.506054581601301e-06, "loss": 0.7349, "step": 9035 }, { "epoch": 1.63, "learning_rate": 5.503343574914152e-06, "loss": 0.5615, "step": 9036 }, { "epoch": 1.63, "learning_rate": 5.500632568227001e-06, "loss": 0.5963, "step": 9037 }, { "epoch": 1.63, "learning_rate": 5.497921561539852e-06, "loss": 0.8136, "step": 9038 }, { "epoch": 1.63, "learning_rate": 5.495210554852702e-06, "loss": 1.2939, "step": 9039 }, { "epoch": 1.63, "learning_rate": 5.4924995481655526e-06, "loss": 0.657, "step": 9040 }, { "epoch": 1.63, "learning_rate": 5.489788541478402e-06, "loss": 0.5236, "step": 9041 }, { "epoch": 1.63, "learning_rate": 5.487077534791253e-06, "loss": 0.7494, "step": 9042 }, { "epoch": 1.63, "learning_rate": 5.484366528104102e-06, "loss": 0.7784, "step": 9043 }, { "epoch": 1.63, "learning_rate": 5.481655521416953e-06, "loss": 0.4958, "step": 9044 }, { "epoch": 1.63, "learning_rate": 5.478944514729803e-06, "loss": 0.3821, "step": 9045 }, { "epoch": 1.63, "learning_rate": 5.476233508042653e-06, "loss": 0.7203, "step": 9046 }, { "epoch": 1.64, "learning_rate": 5.4735225013555035e-06, "loss": 0.8321, "step": 9047 }, { "epoch": 1.64, "learning_rate": 5.470811494668354e-06, "loss": 1.0463, "step": 9048 }, { "epoch": 1.64, "learning_rate": 5.468100487981204e-06, "loss": 0.5348, "step": 9049 }, { "epoch": 1.64, "learning_rate": 5.465389481294054e-06, "loss": 0.575, "step": 9050 }, { "epoch": 1.64, "learning_rate": 5.462678474606904e-06, "loss": 1.0249, "step": 9051 }, { "epoch": 1.64, "learning_rate": 5.459967467919754e-06, "loss": 0.9024, "step": 9052 }, { "epoch": 1.64, "learning_rate": 5.457256461232604e-06, "loss": 1.1388, "step": 9053 }, { "epoch": 1.64, "learning_rate": 5.4545454545454545e-06, "loss": 0.3164, "step": 9054 }, { "epoch": 1.64, "learning_rate": 5.451834447858305e-06, "loss": 0.4322, "step": 9055 }, { "epoch": 1.64, "learning_rate": 5.4491234411711556e-06, "loss": 0.5523, "step": 9056 }, { "epoch": 1.64, "learning_rate": 5.446412434484005e-06, "loss": 0.5881, "step": 9057 }, { "epoch": 1.64, "learning_rate": 5.443701427796856e-06, "loss": 0.7437, "step": 9058 }, { "epoch": 1.64, "learning_rate": 5.440990421109705e-06, "loss": 0.6966, "step": 9059 }, { "epoch": 1.64, "learning_rate": 5.438279414422556e-06, "loss": 0.9957, "step": 9060 }, { "epoch": 1.64, "learning_rate": 5.435568407735405e-06, "loss": 0.7646, "step": 9061 }, { "epoch": 1.64, "learning_rate": 5.432857401048256e-06, "loss": 0.8973, "step": 9062 }, { "epoch": 1.64, "learning_rate": 5.430146394361106e-06, "loss": 1.2159, "step": 9063 }, { "epoch": 1.64, "learning_rate": 5.427435387673957e-06, "loss": 0.4425, "step": 9064 }, { "epoch": 1.64, "learning_rate": 5.424724380986806e-06, "loss": 0.6749, "step": 9065 }, { "epoch": 1.64, "learning_rate": 5.422013374299657e-06, "loss": 0.9409, "step": 9066 }, { "epoch": 1.64, "learning_rate": 5.419302367612507e-06, "loss": 0.4682, "step": 9067 }, { "epoch": 1.64, "learning_rate": 5.416591360925357e-06, "loss": 0.8415, "step": 9068 }, { "epoch": 1.64, "learning_rate": 5.413880354238207e-06, "loss": 0.9882, "step": 9069 }, { "epoch": 1.64, "learning_rate": 5.4111693475510574e-06, "loss": 0.6476, "step": 9070 }, { "epoch": 1.64, "learning_rate": 5.408458340863908e-06, "loss": 0.5787, "step": 9071 }, { "epoch": 1.64, "learning_rate": 5.405747334176758e-06, "loss": 0.516, "step": 9072 }, { "epoch": 1.64, "learning_rate": 5.403036327489608e-06, "loss": 0.3166, "step": 9073 }, { "epoch": 1.64, "learning_rate": 5.400325320802458e-06, "loss": 0.5086, "step": 9074 }, { "epoch": 1.64, "learning_rate": 5.397614314115308e-06, "loss": 0.6484, "step": 9075 }, { "epoch": 1.64, "learning_rate": 5.394903307428159e-06, "loss": 0.8418, "step": 9076 }, { "epoch": 1.64, "learning_rate": 5.392192300741008e-06, "loss": 0.3318, "step": 9077 }, { "epoch": 1.64, "learning_rate": 5.389481294053859e-06, "loss": 0.5553, "step": 9078 }, { "epoch": 1.64, "learning_rate": 5.386770287366709e-06, "loss": 0.6631, "step": 9079 }, { "epoch": 1.64, "learning_rate": 5.38405928067956e-06, "loss": 0.4087, "step": 9080 }, { "epoch": 1.64, "learning_rate": 5.381348273992409e-06, "loss": 0.437, "step": 9081 }, { "epoch": 1.64, "learning_rate": 5.37863726730526e-06, "loss": 0.1442, "step": 9082 }, { "epoch": 1.64, "learning_rate": 5.375926260618109e-06, "loss": 0.3968, "step": 9083 }, { "epoch": 1.64, "learning_rate": 5.37321525393096e-06, "loss": 1.3358, "step": 9084 }, { "epoch": 1.64, "learning_rate": 5.3705042472438095e-06, "loss": 0.8383, "step": 9085 }, { "epoch": 1.64, "learning_rate": 5.3677932405566604e-06, "loss": 1.0276, "step": 9086 }, { "epoch": 1.64, "learning_rate": 5.365082233869511e-06, "loss": 0.5964, "step": 9087 }, { "epoch": 1.64, "learning_rate": 5.362371227182361e-06, "loss": 0.5379, "step": 9088 }, { "epoch": 1.64, "learning_rate": 5.359660220495211e-06, "loss": 0.5839, "step": 9089 }, { "epoch": 1.64, "learning_rate": 5.356949213808061e-06, "loss": 0.3382, "step": 9090 }, { "epoch": 1.64, "learning_rate": 5.354238207120911e-06, "loss": 0.5306, "step": 9091 }, { "epoch": 1.64, "learning_rate": 5.351527200433761e-06, "loss": 0.4372, "step": 9092 }, { "epoch": 1.64, "learning_rate": 5.348816193746611e-06, "loss": 0.9238, "step": 9093 }, { "epoch": 1.64, "learning_rate": 5.3461051870594615e-06, "loss": 0.6682, "step": 9094 }, { "epoch": 1.64, "learning_rate": 5.343394180372312e-06, "loss": 0.5171, "step": 9095 }, { "epoch": 1.64, "learning_rate": 5.340683173685162e-06, "loss": 0.7511, "step": 9096 }, { "epoch": 1.64, "learning_rate": 5.337972166998012e-06, "loss": 0.4449, "step": 9097 }, { "epoch": 1.64, "learning_rate": 5.335261160310863e-06, "loss": 0.5947, "step": 9098 }, { "epoch": 1.64, "learning_rate": 5.332550153623712e-06, "loss": 0.5406, "step": 9099 }, { "epoch": 1.64, "learning_rate": 5.329839146936563e-06, "loss": 1.1446, "step": 9100 }, { "epoch": 1.64, "learning_rate": 5.3271281402494125e-06, "loss": 0.818, "step": 9101 }, { "epoch": 1.65, "learning_rate": 5.3244171335622634e-06, "loss": 0.6685, "step": 9102 }, { "epoch": 1.65, "learning_rate": 5.321706126875113e-06, "loss": 0.7771, "step": 9103 }, { "epoch": 1.65, "learning_rate": 5.318995120187964e-06, "loss": 0.5684, "step": 9104 }, { "epoch": 1.65, "learning_rate": 5.316284113500813e-06, "loss": 0.8451, "step": 9105 }, { "epoch": 1.65, "learning_rate": 5.313573106813664e-06, "loss": 0.9076, "step": 9106 }, { "epoch": 1.65, "learning_rate": 5.310862100126513e-06, "loss": 0.4679, "step": 9107 }, { "epoch": 1.65, "learning_rate": 5.308151093439364e-06, "loss": 0.964, "step": 9108 }, { "epoch": 1.65, "learning_rate": 5.305440086752214e-06, "loss": 0.9337, "step": 9109 }, { "epoch": 1.65, "learning_rate": 5.3027290800650645e-06, "loss": 0.4526, "step": 9110 }, { "epoch": 1.65, "learning_rate": 5.300018073377915e-06, "loss": 0.7673, "step": 9111 }, { "epoch": 1.65, "learning_rate": 5.297307066690765e-06, "loss": 0.7421, "step": 9112 }, { "epoch": 1.65, "learning_rate": 5.294596060003615e-06, "loss": 0.6718, "step": 9113 }, { "epoch": 1.65, "learning_rate": 5.291885053316465e-06, "loss": 1.5562, "step": 9114 }, { "epoch": 1.65, "learning_rate": 5.289174046629315e-06, "loss": 0.5488, "step": 9115 }, { "epoch": 1.65, "learning_rate": 5.286463039942165e-06, "loss": 0.3437, "step": 9116 }, { "epoch": 1.65, "learning_rate": 5.2837520332550155e-06, "loss": 0.8671, "step": 9117 }, { "epoch": 1.65, "learning_rate": 5.2810410265678664e-06, "loss": 0.6606, "step": 9118 }, { "epoch": 1.65, "learning_rate": 5.278330019880716e-06, "loss": 0.6802, "step": 9119 }, { "epoch": 1.65, "learning_rate": 5.275619013193567e-06, "loss": 1.1018, "step": 9120 }, { "epoch": 1.65, "learning_rate": 5.272908006506416e-06, "loss": 1.1308, "step": 9121 }, { "epoch": 1.65, "learning_rate": 5.270196999819267e-06, "loss": 0.6085, "step": 9122 }, { "epoch": 1.65, "learning_rate": 5.267485993132116e-06, "loss": 0.9535, "step": 9123 }, { "epoch": 1.65, "learning_rate": 5.264774986444967e-06, "loss": 1.3685, "step": 9124 }, { "epoch": 1.65, "learning_rate": 5.2620639797578165e-06, "loss": 0.7666, "step": 9125 }, { "epoch": 1.65, "learning_rate": 5.2593529730706675e-06, "loss": 0.3552, "step": 9126 }, { "epoch": 1.65, "learning_rate": 5.256641966383517e-06, "loss": 1.4132, "step": 9127 }, { "epoch": 1.65, "learning_rate": 5.253930959696368e-06, "loss": 0.9901, "step": 9128 }, { "epoch": 1.65, "learning_rate": 5.251219953009218e-06, "loss": 0.9036, "step": 9129 }, { "epoch": 1.65, "learning_rate": 5.248508946322068e-06, "loss": 1.131, "step": 9130 }, { "epoch": 1.65, "learning_rate": 5.245797939634918e-06, "loss": 0.841, "step": 9131 }, { "epoch": 1.65, "learning_rate": 5.243086932947768e-06, "loss": 0.6469, "step": 9132 }, { "epoch": 1.65, "learning_rate": 5.2403759262606184e-06, "loss": 1.1074, "step": 9133 }, { "epoch": 1.65, "learning_rate": 5.237664919573469e-06, "loss": 1.2591, "step": 9134 }, { "epoch": 1.65, "learning_rate": 5.234953912886319e-06, "loss": 0.6127, "step": 9135 }, { "epoch": 1.65, "learning_rate": 5.232242906199168e-06, "loss": 0.6614, "step": 9136 }, { "epoch": 1.65, "learning_rate": 5.229531899512019e-06, "loss": 0.4573, "step": 9137 }, { "epoch": 1.65, "learning_rate": 5.226820892824868e-06, "loss": 0.7872, "step": 9138 }, { "epoch": 1.65, "learning_rate": 5.224109886137719e-06, "loss": 0.8656, "step": 9139 }, { "epoch": 1.65, "learning_rate": 5.221398879450569e-06, "loss": 0.7673, "step": 9140 }, { "epoch": 1.65, "learning_rate": 5.2186878727634195e-06, "loss": 0.9613, "step": 9141 }, { "epoch": 1.65, "learning_rate": 5.21597686607627e-06, "loss": 0.6804, "step": 9142 }, { "epoch": 1.65, "learning_rate": 5.21326585938912e-06, "loss": 0.6187, "step": 9143 }, { "epoch": 1.65, "learning_rate": 5.21055485270197e-06, "loss": 0.5984, "step": 9144 }, { "epoch": 1.65, "learning_rate": 5.20784384601482e-06, "loss": 0.56, "step": 9145 }, { "epoch": 1.65, "learning_rate": 5.20513283932767e-06, "loss": 0.5804, "step": 9146 }, { "epoch": 1.65, "learning_rate": 5.20242183264052e-06, "loss": 0.8158, "step": 9147 }, { "epoch": 1.65, "learning_rate": 5.1997108259533705e-06, "loss": 0.6699, "step": 9148 }, { "epoch": 1.65, "learning_rate": 5.1969998192662214e-06, "loss": 0.4738, "step": 9149 }, { "epoch": 1.65, "learning_rate": 5.194288812579071e-06, "loss": 1.2853, "step": 9150 }, { "epoch": 1.65, "learning_rate": 5.191577805891922e-06, "loss": 0.6053, "step": 9151 }, { "epoch": 1.65, "learning_rate": 5.188866799204771e-06, "loss": 0.4979, "step": 9152 }, { "epoch": 1.65, "learning_rate": 5.186155792517622e-06, "loss": 0.2864, "step": 9153 }, { "epoch": 1.65, "learning_rate": 5.183444785830471e-06, "loss": 0.7486, "step": 9154 }, { "epoch": 1.65, "learning_rate": 5.180733779143322e-06, "loss": 0.8892, "step": 9155 }, { "epoch": 1.65, "learning_rate": 5.1780227724561715e-06, "loss": 0.729, "step": 9156 }, { "epoch": 1.65, "learning_rate": 5.1753117657690225e-06, "loss": 0.5572, "step": 9157 }, { "epoch": 1.66, "learning_rate": 5.172600759081872e-06, "loss": 0.3771, "step": 9158 }, { "epoch": 1.66, "learning_rate": 5.169889752394723e-06, "loss": 0.7284, "step": 9159 }, { "epoch": 1.66, "learning_rate": 5.167178745707573e-06, "loss": 0.452, "step": 9160 }, { "epoch": 1.66, "learning_rate": 5.164467739020423e-06, "loss": 0.9758, "step": 9161 }, { "epoch": 1.66, "learning_rate": 5.161756732333273e-06, "loss": 0.4953, "step": 9162 }, { "epoch": 1.66, "learning_rate": 5.159045725646123e-06, "loss": 0.7589, "step": 9163 }, { "epoch": 1.66, "learning_rate": 5.1563347189589735e-06, "loss": 0.4634, "step": 9164 }, { "epoch": 1.66, "learning_rate": 5.153623712271824e-06, "loss": 0.4688, "step": 9165 }, { "epoch": 1.66, "learning_rate": 5.150912705584674e-06, "loss": 0.7659, "step": 9166 }, { "epoch": 1.66, "learning_rate": 5.148201698897524e-06, "loss": 0.9163, "step": 9167 }, { "epoch": 1.66, "learning_rate": 5.145490692210374e-06, "loss": 0.4319, "step": 9168 }, { "epoch": 1.66, "learning_rate": 5.142779685523224e-06, "loss": 0.8272, "step": 9169 }, { "epoch": 1.66, "learning_rate": 5.140068678836074e-06, "loss": 0.6712, "step": 9170 }, { "epoch": 1.66, "learning_rate": 5.137357672148925e-06, "loss": 0.8178, "step": 9171 }, { "epoch": 1.66, "learning_rate": 5.1346466654617745e-06, "loss": 0.6627, "step": 9172 }, { "epoch": 1.66, "learning_rate": 5.1319356587746255e-06, "loss": 0.5641, "step": 9173 }, { "epoch": 1.66, "learning_rate": 5.129224652087475e-06, "loss": 0.4089, "step": 9174 }, { "epoch": 1.66, "learning_rate": 5.126513645400326e-06, "loss": 0.7003, "step": 9175 }, { "epoch": 1.66, "learning_rate": 5.123802638713175e-06, "loss": 0.4074, "step": 9176 }, { "epoch": 1.66, "learning_rate": 5.121091632026026e-06, "loss": 0.7352, "step": 9177 }, { "epoch": 1.66, "learning_rate": 5.118380625338875e-06, "loss": 0.4527, "step": 9178 }, { "epoch": 1.66, "learning_rate": 5.115669618651726e-06, "loss": 0.502, "step": 9179 }, { "epoch": 1.66, "learning_rate": 5.1129586119645765e-06, "loss": 0.5731, "step": 9180 }, { "epoch": 1.66, "learning_rate": 5.110247605277427e-06, "loss": 0.6958, "step": 9181 }, { "epoch": 1.66, "learning_rate": 5.107536598590277e-06, "loss": 1.0391, "step": 9182 }, { "epoch": 1.66, "learning_rate": 5.104825591903127e-06, "loss": 1.1297, "step": 9183 }, { "epoch": 1.66, "learning_rate": 5.102114585215977e-06, "loss": 0.2758, "step": 9184 }, { "epoch": 1.66, "learning_rate": 5.099403578528827e-06, "loss": 0.5251, "step": 9185 }, { "epoch": 1.66, "learning_rate": 5.096692571841677e-06, "loss": 0.6624, "step": 9186 }, { "epoch": 1.66, "learning_rate": 5.093981565154527e-06, "loss": 0.8752, "step": 9187 }, { "epoch": 1.66, "learning_rate": 5.0912705584673775e-06, "loss": 0.703, "step": 9188 }, { "epoch": 1.66, "learning_rate": 5.088559551780228e-06, "loss": 0.7894, "step": 9189 }, { "epoch": 1.66, "learning_rate": 5.085848545093078e-06, "loss": 0.9821, "step": 9190 }, { "epoch": 1.66, "learning_rate": 5.083137538405929e-06, "loss": 0.8873, "step": 9191 }, { "epoch": 1.66, "learning_rate": 5.080426531718778e-06, "loss": 0.7287, "step": 9192 }, { "epoch": 1.66, "learning_rate": 5.077715525031629e-06, "loss": 0.8343, "step": 9193 }, { "epoch": 1.66, "learning_rate": 5.075004518344478e-06, "loss": 1.0999, "step": 9194 }, { "epoch": 1.66, "learning_rate": 5.072293511657329e-06, "loss": 1.0766, "step": 9195 }, { "epoch": 1.66, "learning_rate": 5.069582504970179e-06, "loss": 0.6023, "step": 9196 }, { "epoch": 1.66, "learning_rate": 5.06687149828303e-06, "loss": 0.7197, "step": 9197 }, { "epoch": 1.66, "learning_rate": 5.064160491595879e-06, "loss": 0.6274, "step": 9198 }, { "epoch": 1.66, "learning_rate": 5.06144948490873e-06, "loss": 0.9422, "step": 9199 }, { "epoch": 1.66, "learning_rate": 5.058738478221579e-06, "loss": 0.2697, "step": 9200 }, { "epoch": 1.66, "learning_rate": 5.05602747153443e-06, "loss": 0.323, "step": 9201 }, { "epoch": 1.66, "learning_rate": 5.05331646484728e-06, "loss": 0.3066, "step": 9202 }, { "epoch": 1.66, "learning_rate": 5.05060545816013e-06, "loss": 0.613, "step": 9203 }, { "epoch": 1.66, "learning_rate": 5.0478944514729805e-06, "loss": 0.1395, "step": 9204 }, { "epoch": 1.66, "learning_rate": 5.045183444785831e-06, "loss": 0.5995, "step": 9205 }, { "epoch": 1.66, "learning_rate": 5.042472438098681e-06, "loss": 0.3757, "step": 9206 }, { "epoch": 1.66, "learning_rate": 5.039761431411531e-06, "loss": 0.8189, "step": 9207 }, { "epoch": 1.66, "learning_rate": 5.037050424724381e-06, "loss": 0.9125, "step": 9208 }, { "epoch": 1.66, "learning_rate": 5.034339418037231e-06, "loss": 0.5833, "step": 9209 }, { "epoch": 1.66, "learning_rate": 5.031628411350081e-06, "loss": 1.0466, "step": 9210 }, { "epoch": 1.66, "learning_rate": 5.0289174046629315e-06, "loss": 0.8143, "step": 9211 }, { "epoch": 1.66, "learning_rate": 5.026206397975782e-06, "loss": 0.4819, "step": 9212 }, { "epoch": 1.67, "learning_rate": 5.023495391288633e-06, "loss": 0.4987, "step": 9213 }, { "epoch": 1.67, "learning_rate": 5.020784384601482e-06, "loss": 0.4315, "step": 9214 }, { "epoch": 1.67, "learning_rate": 5.018073377914333e-06, "loss": 0.9148, "step": 9215 }, { "epoch": 1.67, "learning_rate": 5.015362371227182e-06, "loss": 0.9529, "step": 9216 }, { "epoch": 1.67, "learning_rate": 5.012651364540033e-06, "loss": 0.3067, "step": 9217 }, { "epoch": 1.67, "learning_rate": 5.009940357852882e-06, "loss": 0.5968, "step": 9218 }, { "epoch": 1.67, "learning_rate": 5.007229351165733e-06, "loss": 0.887, "step": 9219 }, { "epoch": 1.67, "learning_rate": 5.004518344478583e-06, "loss": 0.5938, "step": 9220 }, { "epoch": 1.67, "learning_rate": 5.001807337791434e-06, "loss": 0.666, "step": 9221 }, { "epoch": 1.67, "learning_rate": 4.999096331104284e-06, "loss": 0.3989, "step": 9222 }, { "epoch": 1.67, "learning_rate": 4.996385324417134e-06, "loss": 0.9631, "step": 9223 }, { "epoch": 1.67, "learning_rate": 4.993674317729984e-06, "loss": 0.6709, "step": 9224 }, { "epoch": 1.67, "learning_rate": 4.990963311042834e-06, "loss": 0.9908, "step": 9225 }, { "epoch": 1.67, "learning_rate": 4.988252304355684e-06, "loss": 0.3785, "step": 9226 }, { "epoch": 1.67, "learning_rate": 4.9855412976685345e-06, "loss": 1.1736, "step": 9227 }, { "epoch": 1.67, "learning_rate": 4.982830290981385e-06, "loss": 0.5567, "step": 9228 }, { "epoch": 1.67, "learning_rate": 4.980119284294235e-06, "loss": 1.0523, "step": 9229 }, { "epoch": 1.67, "learning_rate": 4.977408277607085e-06, "loss": 0.6438, "step": 9230 }, { "epoch": 1.67, "learning_rate": 4.974697270919935e-06, "loss": 1.0762, "step": 9231 }, { "epoch": 1.67, "learning_rate": 4.971986264232785e-06, "loss": 0.4437, "step": 9232 }, { "epoch": 1.67, "learning_rate": 4.969275257545636e-06, "loss": 1.0203, "step": 9233 }, { "epoch": 1.67, "learning_rate": 4.966564250858485e-06, "loss": 0.3124, "step": 9234 }, { "epoch": 1.67, "learning_rate": 4.963853244171336e-06, "loss": 1.0323, "step": 9235 }, { "epoch": 1.67, "learning_rate": 4.961142237484186e-06, "loss": 0.6006, "step": 9236 }, { "epoch": 1.67, "learning_rate": 4.958431230797037e-06, "loss": 0.5913, "step": 9237 }, { "epoch": 1.67, "learning_rate": 4.955720224109886e-06, "loss": 0.5179, "step": 9238 }, { "epoch": 1.67, "learning_rate": 4.953009217422737e-06, "loss": 0.2668, "step": 9239 }, { "epoch": 1.67, "learning_rate": 4.950298210735586e-06, "loss": 0.5129, "step": 9240 }, { "epoch": 1.67, "learning_rate": 4.947587204048437e-06, "loss": 1.3762, "step": 9241 }, { "epoch": 1.67, "learning_rate": 4.9448761973612865e-06, "loss": 0.4117, "step": 9242 }, { "epoch": 1.67, "learning_rate": 4.9421651906741375e-06, "loss": 0.8037, "step": 9243 }, { "epoch": 1.67, "learning_rate": 4.939454183986988e-06, "loss": 0.4175, "step": 9244 }, { "epoch": 1.67, "learning_rate": 4.936743177299838e-06, "loss": 0.7665, "step": 9245 }, { "epoch": 1.67, "learning_rate": 4.934032170612688e-06, "loss": 0.5787, "step": 9246 }, { "epoch": 1.67, "learning_rate": 4.931321163925538e-06, "loss": 0.7871, "step": 9247 }, { "epoch": 1.67, "learning_rate": 4.928610157238388e-06, "loss": 0.8295, "step": 9248 }, { "epoch": 1.67, "learning_rate": 4.925899150551238e-06, "loss": 1.0227, "step": 9249 }, { "epoch": 1.67, "learning_rate": 4.923188143864088e-06, "loss": 0.4973, "step": 9250 }, { "epoch": 1.67, "learning_rate": 4.9204771371769385e-06, "loss": 1.1122, "step": 9251 }, { "epoch": 1.67, "learning_rate": 4.917766130489789e-06, "loss": 0.246, "step": 9252 }, { "epoch": 1.67, "learning_rate": 4.915055123802639e-06, "loss": 0.8189, "step": 9253 }, { "epoch": 1.67, "learning_rate": 4.912344117115489e-06, "loss": 0.6513, "step": 9254 }, { "epoch": 1.67, "learning_rate": 4.90963311042834e-06, "loss": 0.917, "step": 9255 }, { "epoch": 1.67, "learning_rate": 4.906922103741189e-06, "loss": 1.0399, "step": 9256 }, { "epoch": 1.67, "learning_rate": 4.90421109705404e-06, "loss": 0.2941, "step": 9257 }, { "epoch": 1.67, "learning_rate": 4.9015000903668895e-06, "loss": 0.7664, "step": 9258 }, { "epoch": 1.67, "learning_rate": 4.8987890836797404e-06, "loss": 0.9145, "step": 9259 }, { "epoch": 1.67, "learning_rate": 4.89607807699259e-06, "loss": 0.671, "step": 9260 }, { "epoch": 1.67, "learning_rate": 4.893367070305441e-06, "loss": 1.2319, "step": 9261 }, { "epoch": 1.67, "learning_rate": 4.89065606361829e-06, "loss": 0.6252, "step": 9262 }, { "epoch": 1.67, "learning_rate": 4.887945056931141e-06, "loss": 0.8278, "step": 9263 }, { "epoch": 1.67, "learning_rate": 4.885234050243991e-06, "loss": 0.6311, "step": 9264 }, { "epoch": 1.67, "learning_rate": 4.88252304355684e-06, "loss": 0.655, "step": 9265 }, { "epoch": 1.67, "learning_rate": 4.879812036869691e-06, "loss": 0.8095, "step": 9266 }, { "epoch": 1.67, "learning_rate": 4.877101030182541e-06, "loss": 1.2631, "step": 9267 }, { "epoch": 1.68, "learning_rate": 4.874390023495392e-06, "loss": 0.7002, "step": 9268 }, { "epoch": 1.68, "learning_rate": 4.871679016808241e-06, "loss": 0.3073, "step": 9269 }, { "epoch": 1.68, "learning_rate": 4.868968010121092e-06, "loss": 0.5114, "step": 9270 }, { "epoch": 1.68, "learning_rate": 4.866257003433941e-06, "loss": 0.6907, "step": 9271 }, { "epoch": 1.68, "learning_rate": 4.863545996746792e-06, "loss": 1.1049, "step": 9272 }, { "epoch": 1.68, "learning_rate": 4.8608349900596415e-06, "loss": 0.6588, "step": 9273 }, { "epoch": 1.68, "learning_rate": 4.8581239833724925e-06, "loss": 0.9416, "step": 9274 }, { "epoch": 1.68, "learning_rate": 4.855412976685343e-06, "loss": 0.5644, "step": 9275 }, { "epoch": 1.68, "learning_rate": 4.852701969998193e-06, "loss": 0.8426, "step": 9276 }, { "epoch": 1.68, "learning_rate": 4.849990963311043e-06, "loss": 0.47, "step": 9277 }, { "epoch": 1.68, "learning_rate": 4.847279956623893e-06, "loss": 0.4525, "step": 9278 }, { "epoch": 1.68, "learning_rate": 4.844568949936743e-06, "loss": 0.6251, "step": 9279 }, { "epoch": 1.68, "learning_rate": 4.841857943249593e-06, "loss": 0.5869, "step": 9280 }, { "epoch": 1.68, "learning_rate": 4.839146936562443e-06, "loss": 0.5483, "step": 9281 }, { "epoch": 1.68, "learning_rate": 4.8364359298752935e-06, "loss": 1.0304, "step": 9282 }, { "epoch": 1.68, "learning_rate": 4.833724923188144e-06, "loss": 0.703, "step": 9283 }, { "epoch": 1.68, "learning_rate": 4.831013916500994e-06, "loss": 0.7922, "step": 9284 }, { "epoch": 1.68, "learning_rate": 4.828302909813844e-06, "loss": 0.7747, "step": 9285 }, { "epoch": 1.68, "learning_rate": 4.825591903126695e-06, "loss": 1.3807, "step": 9286 }, { "epoch": 1.68, "learning_rate": 4.822880896439544e-06, "loss": 0.2207, "step": 9287 }, { "epoch": 1.68, "learning_rate": 4.820169889752395e-06, "loss": 0.9782, "step": 9288 }, { "epoch": 1.68, "learning_rate": 4.8174588830652445e-06, "loss": 0.6662, "step": 9289 }, { "epoch": 1.68, "learning_rate": 4.8147478763780955e-06, "loss": 0.3307, "step": 9290 }, { "epoch": 1.68, "learning_rate": 4.812036869690945e-06, "loss": 0.3111, "step": 9291 }, { "epoch": 1.68, "learning_rate": 4.809325863003796e-06, "loss": 1.012, "step": 9292 }, { "epoch": 1.68, "learning_rate": 4.806614856316645e-06, "loss": 0.6079, "step": 9293 }, { "epoch": 1.68, "learning_rate": 4.803903849629496e-06, "loss": 0.2598, "step": 9294 }, { "epoch": 1.68, "learning_rate": 4.801192842942346e-06, "loss": 0.4644, "step": 9295 }, { "epoch": 1.68, "learning_rate": 4.798481836255196e-06, "loss": 0.5839, "step": 9296 }, { "epoch": 1.68, "learning_rate": 4.795770829568046e-06, "loss": 0.409, "step": 9297 }, { "epoch": 1.68, "learning_rate": 4.7930598228808965e-06, "loss": 0.5416, "step": 9298 }, { "epoch": 1.68, "learning_rate": 4.790348816193747e-06, "loss": 0.9995, "step": 9299 }, { "epoch": 1.68, "learning_rate": 4.787637809506597e-06, "loss": 0.619, "step": 9300 }, { "epoch": 1.68, "learning_rate": 4.784926802819447e-06, "loss": 0.6852, "step": 9301 }, { "epoch": 1.68, "learning_rate": 4.782215796132297e-06, "loss": 0.5877, "step": 9302 }, { "epoch": 1.68, "learning_rate": 4.779504789445147e-06, "loss": 0.5699, "step": 9303 }, { "epoch": 1.68, "learning_rate": 4.776793782757997e-06, "loss": 1.0485, "step": 9304 }, { "epoch": 1.68, "learning_rate": 4.7740827760708475e-06, "loss": 0.6142, "step": 9305 }, { "epoch": 1.68, "learning_rate": 4.7713717693836985e-06, "loss": 0.4487, "step": 9306 }, { "epoch": 1.68, "learning_rate": 4.768660762696548e-06, "loss": 1.0232, "step": 9307 }, { "epoch": 1.68, "learning_rate": 4.765949756009399e-06, "loss": 0.6001, "step": 9308 }, { "epoch": 1.68, "learning_rate": 4.763238749322248e-06, "loss": 0.8298, "step": 9309 }, { "epoch": 1.68, "learning_rate": 4.760527742635099e-06, "loss": 0.5678, "step": 9310 }, { "epoch": 1.68, "learning_rate": 4.757816735947948e-06, "loss": 0.5596, "step": 9311 }, { "epoch": 1.68, "learning_rate": 4.755105729260799e-06, "loss": 1.1565, "step": 9312 }, { "epoch": 1.68, "learning_rate": 4.7523947225736485e-06, "loss": 0.9954, "step": 9313 }, { "epoch": 1.68, "learning_rate": 4.7496837158864995e-06, "loss": 0.5722, "step": 9314 }, { "epoch": 1.68, "learning_rate": 4.746972709199349e-06, "loss": 0.5954, "step": 9315 }, { "epoch": 1.68, "learning_rate": 4.7442617025122e-06, "loss": 0.2294, "step": 9316 }, { "epoch": 1.68, "learning_rate": 4.74155069582505e-06, "loss": 0.71, "step": 9317 }, { "epoch": 1.68, "learning_rate": 4.7388396891379e-06, "loss": 0.3191, "step": 9318 }, { "epoch": 1.68, "learning_rate": 4.73612868245075e-06, "loss": 0.4105, "step": 9319 }, { "epoch": 1.68, "learning_rate": 4.7334176757636e-06, "loss": 0.6767, "step": 9320 }, { "epoch": 1.68, "learning_rate": 4.7307066690764505e-06, "loss": 0.8348, "step": 9321 }, { "epoch": 1.68, "learning_rate": 4.727995662389301e-06, "loss": 0.4041, "step": 9322 }, { "epoch": 1.68, "learning_rate": 4.725284655702151e-06, "loss": 0.3117, "step": 9323 }, { "epoch": 1.69, "learning_rate": 4.722573649015001e-06, "loss": 0.5348, "step": 9324 }, { "epoch": 1.69, "learning_rate": 4.719862642327851e-06, "loss": 0.7442, "step": 9325 }, { "epoch": 1.69, "learning_rate": 4.717151635640702e-06, "loss": 1.0833, "step": 9326 }, { "epoch": 1.69, "learning_rate": 4.714440628953551e-06, "loss": 1.1485, "step": 9327 }, { "epoch": 1.69, "learning_rate": 4.711729622266402e-06, "loss": 1.9149, "step": 9328 }, { "epoch": 1.69, "learning_rate": 4.7090186155792515e-06, "loss": 0.8764, "step": 9329 }, { "epoch": 1.69, "learning_rate": 4.7063076088921025e-06, "loss": 0.5323, "step": 9330 }, { "epoch": 1.69, "learning_rate": 4.703596602204952e-06, "loss": 0.9726, "step": 9331 }, { "epoch": 1.69, "learning_rate": 4.700885595517803e-06, "loss": 0.5225, "step": 9332 }, { "epoch": 1.69, "learning_rate": 4.698174588830652e-06, "loss": 0.3363, "step": 9333 }, { "epoch": 1.69, "learning_rate": 4.695463582143503e-06, "loss": 0.7363, "step": 9334 }, { "epoch": 1.69, "learning_rate": 4.692752575456352e-06, "loss": 0.584, "step": 9335 }, { "epoch": 1.69, "learning_rate": 4.690041568769203e-06, "loss": 0.74, "step": 9336 }, { "epoch": 1.69, "learning_rate": 4.6873305620820535e-06, "loss": 1.2045, "step": 9337 }, { "epoch": 1.69, "learning_rate": 4.684619555394904e-06, "loss": 0.6439, "step": 9338 }, { "epoch": 1.69, "learning_rate": 4.681908548707754e-06, "loss": 0.3403, "step": 9339 }, { "epoch": 1.69, "learning_rate": 4.679197542020604e-06, "loss": 0.4441, "step": 9340 }, { "epoch": 1.69, "learning_rate": 4.676486535333454e-06, "loss": 1.1893, "step": 9341 }, { "epoch": 1.69, "learning_rate": 4.673775528646304e-06, "loss": 0.4219, "step": 9342 }, { "epoch": 1.69, "learning_rate": 4.671064521959154e-06, "loss": 0.8455, "step": 9343 }, { "epoch": 1.69, "learning_rate": 4.668353515272004e-06, "loss": 0.698, "step": 9344 }, { "epoch": 1.69, "learning_rate": 4.6656425085848545e-06, "loss": 1.3332, "step": 9345 }, { "epoch": 1.69, "learning_rate": 4.662931501897705e-06, "loss": 0.8969, "step": 9346 }, { "epoch": 1.69, "learning_rate": 4.660220495210555e-06, "loss": 0.5274, "step": 9347 }, { "epoch": 1.69, "learning_rate": 4.657509488523406e-06, "loss": 0.5607, "step": 9348 }, { "epoch": 1.69, "learning_rate": 4.654798481836255e-06, "loss": 0.5054, "step": 9349 }, { "epoch": 1.69, "learning_rate": 4.652087475149106e-06, "loss": 0.5613, "step": 9350 }, { "epoch": 1.69, "learning_rate": 4.649376468461955e-06, "loss": 0.6604, "step": 9351 }, { "epoch": 1.69, "learning_rate": 4.646665461774806e-06, "loss": 0.4759, "step": 9352 }, { "epoch": 1.69, "learning_rate": 4.643954455087656e-06, "loss": 0.8697, "step": 9353 }, { "epoch": 1.69, "learning_rate": 4.641243448400507e-06, "loss": 0.7276, "step": 9354 }, { "epoch": 1.69, "learning_rate": 4.638532441713356e-06, "loss": 0.4265, "step": 9355 }, { "epoch": 1.69, "learning_rate": 4.635821435026207e-06, "loss": 0.8218, "step": 9356 }, { "epoch": 1.69, "learning_rate": 4.633110428339056e-06, "loss": 0.4626, "step": 9357 }, { "epoch": 1.69, "learning_rate": 4.630399421651907e-06, "loss": 1.165, "step": 9358 }, { "epoch": 1.69, "learning_rate": 4.627688414964757e-06, "loss": 0.5941, "step": 9359 }, { "epoch": 1.69, "learning_rate": 4.624977408277607e-06, "loss": 0.7853, "step": 9360 }, { "epoch": 1.69, "learning_rate": 4.6222664015904575e-06, "loss": 0.4446, "step": 9361 }, { "epoch": 1.69, "learning_rate": 4.619555394903308e-06, "loss": 0.933, "step": 9362 }, { "epoch": 1.69, "learning_rate": 4.616844388216158e-06, "loss": 0.5116, "step": 9363 }, { "epoch": 1.69, "learning_rate": 4.614133381529008e-06, "loss": 0.4605, "step": 9364 }, { "epoch": 1.69, "learning_rate": 4.611422374841858e-06, "loss": 0.5189, "step": 9365 }, { "epoch": 1.69, "learning_rate": 4.608711368154708e-06, "loss": 1.1852, "step": 9366 }, { "epoch": 1.69, "learning_rate": 4.606000361467558e-06, "loss": 1.4361, "step": 9367 }, { "epoch": 1.69, "learning_rate": 4.603289354780409e-06, "loss": 0.8671, "step": 9368 }, { "epoch": 1.69, "learning_rate": 4.600578348093259e-06, "loss": 1.4349, "step": 9369 }, { "epoch": 1.69, "learning_rate": 4.59786734140611e-06, "loss": 0.9449, "step": 9370 }, { "epoch": 1.69, "learning_rate": 4.595156334718959e-06, "loss": 0.3471, "step": 9371 }, { "epoch": 1.69, "learning_rate": 4.59244532803181e-06, "loss": 0.3512, "step": 9372 }, { "epoch": 1.69, "learning_rate": 4.589734321344659e-06, "loss": 0.8349, "step": 9373 }, { "epoch": 1.69, "learning_rate": 4.58702331465751e-06, "loss": 0.2901, "step": 9374 }, { "epoch": 1.69, "learning_rate": 4.584312307970359e-06, "loss": 1.9117, "step": 9375 }, { "epoch": 1.69, "learning_rate": 4.58160130128321e-06, "loss": 0.4059, "step": 9376 }, { "epoch": 1.69, "learning_rate": 4.57889029459606e-06, "loss": 1.1586, "step": 9377 }, { "epoch": 1.69, "learning_rate": 4.576179287908911e-06, "loss": 0.7952, "step": 9378 }, { "epoch": 1.7, "learning_rate": 4.573468281221761e-06, "loss": 1.3387, "step": 9379 }, { "epoch": 1.7, "learning_rate": 4.570757274534611e-06, "loss": 0.7304, "step": 9380 }, { "epoch": 1.7, "learning_rate": 4.568046267847461e-06, "loss": 0.6082, "step": 9381 }, { "epoch": 1.7, "learning_rate": 4.565335261160311e-06, "loss": 1.2664, "step": 9382 }, { "epoch": 1.7, "learning_rate": 4.562624254473161e-06, "loss": 0.5841, "step": 9383 }, { "epoch": 1.7, "learning_rate": 4.5599132477860115e-06, "loss": 0.6167, "step": 9384 }, { "epoch": 1.7, "learning_rate": 4.557202241098862e-06, "loss": 0.5878, "step": 9385 }, { "epoch": 1.7, "learning_rate": 4.554491234411712e-06, "loss": 0.775, "step": 9386 }, { "epoch": 1.7, "learning_rate": 4.551780227724562e-06, "loss": 0.6783, "step": 9387 }, { "epoch": 1.7, "learning_rate": 4.549069221037412e-06, "loss": 1.2785, "step": 9388 }, { "epoch": 1.7, "learning_rate": 4.546358214350262e-06, "loss": 0.3626, "step": 9389 }, { "epoch": 1.7, "learning_rate": 4.543647207663113e-06, "loss": 0.5547, "step": 9390 }, { "epoch": 1.7, "learning_rate": 4.540936200975962e-06, "loss": 0.829, "step": 9391 }, { "epoch": 1.7, "learning_rate": 4.538225194288813e-06, "loss": 0.4338, "step": 9392 }, { "epoch": 1.7, "learning_rate": 4.535514187601663e-06, "loss": 0.7174, "step": 9393 }, { "epoch": 1.7, "learning_rate": 4.532803180914513e-06, "loss": 0.4823, "step": 9394 }, { "epoch": 1.7, "learning_rate": 4.530092174227363e-06, "loss": 0.5559, "step": 9395 }, { "epoch": 1.7, "learning_rate": 4.527381167540213e-06, "loss": 0.3606, "step": 9396 }, { "epoch": 1.7, "learning_rate": 4.524670160853063e-06, "loss": 0.659, "step": 9397 }, { "epoch": 1.7, "learning_rate": 4.521959154165913e-06, "loss": 0.9755, "step": 9398 }, { "epoch": 1.7, "learning_rate": 4.519248147478764e-06, "loss": 0.914, "step": 9399 }, { "epoch": 1.7, "learning_rate": 4.516537140791614e-06, "loss": 0.4922, "step": 9400 }, { "epoch": 1.7, "learning_rate": 4.513826134104465e-06, "loss": 0.4213, "step": 9401 }, { "epoch": 1.7, "learning_rate": 4.511115127417314e-06, "loss": 0.6501, "step": 9402 }, { "epoch": 1.7, "learning_rate": 4.508404120730165e-06, "loss": 0.3864, "step": 9403 }, { "epoch": 1.7, "learning_rate": 4.505693114043014e-06, "loss": 0.3071, "step": 9404 }, { "epoch": 1.7, "learning_rate": 4.502982107355865e-06, "loss": 0.5461, "step": 9405 }, { "epoch": 1.7, "learning_rate": 4.500271100668714e-06, "loss": 0.3507, "step": 9406 }, { "epoch": 1.7, "learning_rate": 4.497560093981565e-06, "loss": 1.2569, "step": 9407 }, { "epoch": 1.7, "learning_rate": 4.494849087294415e-06, "loss": 0.3066, "step": 9408 }, { "epoch": 1.7, "learning_rate": 4.492138080607266e-06, "loss": 0.5371, "step": 9409 }, { "epoch": 1.7, "learning_rate": 4.489427073920116e-06, "loss": 1.3385, "step": 9410 }, { "epoch": 1.7, "learning_rate": 4.486716067232966e-06, "loss": 1.0904, "step": 9411 }, { "epoch": 1.7, "learning_rate": 4.484005060545816e-06, "loss": 0.3749, "step": 9412 }, { "epoch": 1.7, "learning_rate": 4.481294053858666e-06, "loss": 1.3792, "step": 9413 }, { "epoch": 1.7, "learning_rate": 4.478583047171516e-06, "loss": 0.9067, "step": 9414 }, { "epoch": 1.7, "learning_rate": 4.4758720404843665e-06, "loss": 0.7861, "step": 9415 }, { "epoch": 1.7, "learning_rate": 4.473161033797217e-06, "loss": 0.4449, "step": 9416 }, { "epoch": 1.7, "learning_rate": 4.470450027110067e-06, "loss": 0.9723, "step": 9417 }, { "epoch": 1.7, "learning_rate": 4.467739020422917e-06, "loss": 0.7365, "step": 9418 }, { "epoch": 1.7, "learning_rate": 4.465028013735767e-06, "loss": 0.8308, "step": 9419 }, { "epoch": 1.7, "learning_rate": 4.462317007048617e-06, "loss": 1.1629, "step": 9420 }, { "epoch": 1.7, "learning_rate": 4.459606000361468e-06, "loss": 0.7567, "step": 9421 }, { "epoch": 1.7, "learning_rate": 4.456894993674317e-06, "loss": 0.4649, "step": 9422 }, { "epoch": 1.7, "learning_rate": 4.454183986987168e-06, "loss": 0.6058, "step": 9423 }, { "epoch": 1.7, "learning_rate": 4.451472980300018e-06, "loss": 0.3447, "step": 9424 }, { "epoch": 1.7, "learning_rate": 4.448761973612869e-06, "loss": 0.1818, "step": 9425 }, { "epoch": 1.7, "learning_rate": 4.446050966925718e-06, "loss": 0.7973, "step": 9426 }, { "epoch": 1.7, "learning_rate": 4.443339960238569e-06, "loss": 0.8365, "step": 9427 }, { "epoch": 1.7, "learning_rate": 4.440628953551418e-06, "loss": 0.4018, "step": 9428 }, { "epoch": 1.7, "learning_rate": 4.437917946864269e-06, "loss": 0.5036, "step": 9429 }, { "epoch": 1.7, "learning_rate": 4.435206940177119e-06, "loss": 0.9588, "step": 9430 }, { "epoch": 1.7, "learning_rate": 4.4324959334899695e-06, "loss": 0.4034, "step": 9431 }, { "epoch": 1.7, "learning_rate": 4.42978492680282e-06, "loss": 0.3665, "step": 9432 }, { "epoch": 1.7, "learning_rate": 4.42707392011567e-06, "loss": 0.7348, "step": 9433 }, { "epoch": 1.71, "learning_rate": 4.42436291342852e-06, "loss": 0.6643, "step": 9434 }, { "epoch": 1.71, "learning_rate": 4.42165190674137e-06, "loss": 1.2816, "step": 9435 }, { "epoch": 1.71, "learning_rate": 4.41894090005422e-06, "loss": 0.5374, "step": 9436 }, { "epoch": 1.71, "learning_rate": 4.41622989336707e-06, "loss": 0.336, "step": 9437 }, { "epoch": 1.71, "learning_rate": 4.41351888667992e-06, "loss": 0.7675, "step": 9438 }, { "epoch": 1.71, "learning_rate": 4.4108078799927705e-06, "loss": 0.7727, "step": 9439 }, { "epoch": 1.71, "learning_rate": 4.408096873305621e-06, "loss": 0.4359, "step": 9440 }, { "epoch": 1.71, "learning_rate": 4.405385866618472e-06, "loss": 0.6603, "step": 9441 }, { "epoch": 1.71, "learning_rate": 4.402674859931321e-06, "loss": 0.7595, "step": 9442 }, { "epoch": 1.71, "learning_rate": 4.399963853244172e-06, "loss": 0.9059, "step": 9443 }, { "epoch": 1.71, "learning_rate": 4.397252846557021e-06, "loss": 0.4436, "step": 9444 }, { "epoch": 1.71, "learning_rate": 4.394541839869872e-06, "loss": 0.6836, "step": 9445 }, { "epoch": 1.71, "learning_rate": 4.3918308331827215e-06, "loss": 0.8947, "step": 9446 }, { "epoch": 1.71, "learning_rate": 4.3891198264955725e-06, "loss": 0.4465, "step": 9447 }, { "epoch": 1.71, "learning_rate": 4.386408819808422e-06, "loss": 0.8442, "step": 9448 }, { "epoch": 1.71, "learning_rate": 4.383697813121273e-06, "loss": 0.9494, "step": 9449 }, { "epoch": 1.71, "learning_rate": 4.380986806434122e-06, "loss": 0.377, "step": 9450 }, { "epoch": 1.71, "learning_rate": 4.378275799746973e-06, "loss": 0.4394, "step": 9451 }, { "epoch": 1.71, "learning_rate": 4.375564793059823e-06, "loss": 0.5691, "step": 9452 }, { "epoch": 1.71, "learning_rate": 4.372853786372673e-06, "loss": 0.7425, "step": 9453 }, { "epoch": 1.71, "learning_rate": 4.370142779685523e-06, "loss": 0.449, "step": 9454 }, { "epoch": 1.71, "learning_rate": 4.3674317729983735e-06, "loss": 0.8619, "step": 9455 }, { "epoch": 1.71, "learning_rate": 4.364720766311224e-06, "loss": 0.4276, "step": 9456 }, { "epoch": 1.71, "learning_rate": 4.362009759624074e-06, "loss": 0.9992, "step": 9457 }, { "epoch": 1.71, "learning_rate": 4.359298752936924e-06, "loss": 0.5713, "step": 9458 }, { "epoch": 1.71, "learning_rate": 4.356587746249774e-06, "loss": 0.5813, "step": 9459 }, { "epoch": 1.71, "learning_rate": 4.353876739562624e-06, "loss": 0.8345, "step": 9460 }, { "epoch": 1.71, "learning_rate": 4.351165732875474e-06, "loss": 0.883, "step": 9461 }, { "epoch": 1.71, "learning_rate": 4.3484547261883245e-06, "loss": 0.6273, "step": 9462 }, { "epoch": 1.71, "learning_rate": 4.3457437195011755e-06, "loss": 0.5962, "step": 9463 }, { "epoch": 1.71, "learning_rate": 4.343032712814025e-06, "loss": 0.4972, "step": 9464 }, { "epoch": 1.71, "learning_rate": 4.340321706126876e-06, "loss": 0.3704, "step": 9465 }, { "epoch": 1.71, "learning_rate": 4.337610699439725e-06, "loss": 0.6251, "step": 9466 }, { "epoch": 1.71, "learning_rate": 4.334899692752576e-06, "loss": 0.9535, "step": 9467 }, { "epoch": 1.71, "learning_rate": 4.332188686065425e-06, "loss": 0.7574, "step": 9468 }, { "epoch": 1.71, "learning_rate": 4.329477679378276e-06, "loss": 1.0524, "step": 9469 }, { "epoch": 1.71, "learning_rate": 4.3267666726911256e-06, "loss": 0.621, "step": 9470 }, { "epoch": 1.71, "learning_rate": 4.3240556660039765e-06, "loss": 0.7422, "step": 9471 }, { "epoch": 1.71, "learning_rate": 4.321344659316827e-06, "loss": 0.7913, "step": 9472 }, { "epoch": 1.71, "learning_rate": 4.318633652629677e-06, "loss": 0.6737, "step": 9473 }, { "epoch": 1.71, "learning_rate": 4.315922645942527e-06, "loss": 0.6045, "step": 9474 }, { "epoch": 1.71, "learning_rate": 4.313211639255377e-06, "loss": 1.1083, "step": 9475 }, { "epoch": 1.71, "learning_rate": 4.310500632568227e-06, "loss": 0.9874, "step": 9476 }, { "epoch": 1.71, "learning_rate": 4.307789625881077e-06, "loss": 0.8202, "step": 9477 }, { "epoch": 1.71, "learning_rate": 4.3050786191939275e-06, "loss": 0.7764, "step": 9478 }, { "epoch": 1.71, "learning_rate": 4.302367612506778e-06, "loss": 0.4286, "step": 9479 }, { "epoch": 1.71, "learning_rate": 4.299656605819628e-06, "loss": 0.9113, "step": 9480 }, { "epoch": 1.71, "learning_rate": 4.296945599132478e-06, "loss": 0.5784, "step": 9481 }, { "epoch": 1.71, "learning_rate": 4.294234592445328e-06, "loss": 0.4473, "step": 9482 }, { "epoch": 1.71, "learning_rate": 4.291523585758179e-06, "loss": 0.8275, "step": 9483 }, { "epoch": 1.71, "learning_rate": 4.288812579071028e-06, "loss": 0.7862, "step": 9484 }, { "epoch": 1.71, "learning_rate": 4.286101572383879e-06, "loss": 0.8375, "step": 9485 }, { "epoch": 1.71, "learning_rate": 4.2833905656967286e-06, "loss": 0.6301, "step": 9486 }, { "epoch": 1.71, "learning_rate": 4.2806795590095795e-06, "loss": 0.6484, "step": 9487 }, { "epoch": 1.71, "learning_rate": 4.277968552322429e-06, "loss": 0.1563, "step": 9488 }, { "epoch": 1.71, "learning_rate": 4.27525754563528e-06, "loss": 0.6369, "step": 9489 }, { "epoch": 1.72, "learning_rate": 4.272546538948129e-06, "loss": 1.2608, "step": 9490 }, { "epoch": 1.72, "learning_rate": 4.26983553226098e-06, "loss": 0.366, "step": 9491 }, { "epoch": 1.72, "learning_rate": 4.267124525573829e-06, "loss": 1.3092, "step": 9492 }, { "epoch": 1.72, "learning_rate": 4.26441351888668e-06, "loss": 0.459, "step": 9493 }, { "epoch": 1.72, "learning_rate": 4.2617025121995305e-06, "loss": 0.8614, "step": 9494 }, { "epoch": 1.72, "learning_rate": 4.258991505512381e-06, "loss": 0.7238, "step": 9495 }, { "epoch": 1.72, "learning_rate": 4.256280498825231e-06, "loss": 0.9773, "step": 9496 }, { "epoch": 1.72, "learning_rate": 4.253569492138081e-06, "loss": 0.7692, "step": 9497 }, { "epoch": 1.72, "learning_rate": 4.250858485450931e-06, "loss": 1.1718, "step": 9498 }, { "epoch": 1.72, "learning_rate": 4.248147478763781e-06, "loss": 0.5024, "step": 9499 }, { "epoch": 1.72, "learning_rate": 4.245436472076631e-06, "loss": 0.6558, "step": 9500 }, { "epoch": 1.72, "learning_rate": 4.242725465389481e-06, "loss": 1.1175, "step": 9501 }, { "epoch": 1.72, "learning_rate": 4.2400144587023315e-06, "loss": 0.4521, "step": 9502 }, { "epoch": 1.72, "learning_rate": 4.237303452015182e-06, "loss": 1.1415, "step": 9503 }, { "epoch": 1.72, "learning_rate": 4.234592445328032e-06, "loss": 0.4923, "step": 9504 }, { "epoch": 1.72, "learning_rate": 4.231881438640883e-06, "loss": 0.8468, "step": 9505 }, { "epoch": 1.72, "learning_rate": 4.229170431953732e-06, "loss": 0.7066, "step": 9506 }, { "epoch": 1.72, "learning_rate": 4.226459425266583e-06, "loss": 0.774, "step": 9507 }, { "epoch": 1.72, "learning_rate": 4.223748418579432e-06, "loss": 0.2755, "step": 9508 }, { "epoch": 1.72, "learning_rate": 4.221037411892283e-06, "loss": 1.411, "step": 9509 }, { "epoch": 1.72, "learning_rate": 4.218326405205133e-06, "loss": 0.7229, "step": 9510 }, { "epoch": 1.72, "learning_rate": 4.215615398517984e-06, "loss": 0.7627, "step": 9511 }, { "epoch": 1.72, "learning_rate": 4.212904391830833e-06, "loss": 0.3823, "step": 9512 }, { "epoch": 1.72, "learning_rate": 4.210193385143684e-06, "loss": 0.3822, "step": 9513 }, { "epoch": 1.72, "learning_rate": 4.207482378456534e-06, "loss": 0.6873, "step": 9514 }, { "epoch": 1.72, "learning_rate": 4.204771371769384e-06, "loss": 0.8223, "step": 9515 }, { "epoch": 1.72, "learning_rate": 4.202060365082234e-06, "loss": 1.5011, "step": 9516 }, { "epoch": 1.72, "learning_rate": 4.199349358395084e-06, "loss": 0.6902, "step": 9517 }, { "epoch": 1.72, "learning_rate": 4.1966383517079345e-06, "loss": 0.4249, "step": 9518 }, { "epoch": 1.72, "learning_rate": 4.193927345020785e-06, "loss": 0.3499, "step": 9519 }, { "epoch": 1.72, "learning_rate": 4.191216338333635e-06, "loss": 0.4396, "step": 9520 }, { "epoch": 1.72, "learning_rate": 4.188505331646485e-06, "loss": 0.6095, "step": 9521 }, { "epoch": 1.72, "learning_rate": 4.185794324959335e-06, "loss": 0.4936, "step": 9522 }, { "epoch": 1.72, "learning_rate": 4.183083318272184e-06, "loss": 0.7856, "step": 9523 }, { "epoch": 1.72, "learning_rate": 4.180372311585035e-06, "loss": 0.6175, "step": 9524 }, { "epoch": 1.72, "learning_rate": 4.1776613048978855e-06, "loss": 0.6013, "step": 9525 }, { "epoch": 1.72, "learning_rate": 4.174950298210736e-06, "loss": 0.528, "step": 9526 }, { "epoch": 1.72, "learning_rate": 4.172239291523586e-06, "loss": 0.6692, "step": 9527 }, { "epoch": 1.72, "learning_rate": 4.169528284836436e-06, "loss": 0.9641, "step": 9528 }, { "epoch": 1.72, "learning_rate": 4.166817278149286e-06, "loss": 0.3313, "step": 9529 }, { "epoch": 1.72, "learning_rate": 4.164106271462136e-06, "loss": 0.2831, "step": 9530 }, { "epoch": 1.72, "learning_rate": 4.161395264774986e-06, "loss": 0.8702, "step": 9531 }, { "epoch": 1.72, "learning_rate": 4.1586842580878364e-06, "loss": 0.9746, "step": 9532 }, { "epoch": 1.72, "learning_rate": 4.1559732514006866e-06, "loss": 1.1049, "step": 9533 }, { "epoch": 1.72, "learning_rate": 4.153262244713537e-06, "loss": 0.4998, "step": 9534 }, { "epoch": 1.72, "learning_rate": 4.150551238026387e-06, "loss": 0.8555, "step": 9535 }, { "epoch": 1.72, "learning_rate": 4.147840231339238e-06, "loss": 0.8808, "step": 9536 }, { "epoch": 1.72, "learning_rate": 4.145129224652087e-06, "loss": 1.2132, "step": 9537 }, { "epoch": 1.72, "learning_rate": 4.142418217964938e-06, "loss": 0.4083, "step": 9538 }, { "epoch": 1.72, "learning_rate": 4.139707211277787e-06, "loss": 0.5109, "step": 9539 }, { "epoch": 1.72, "learning_rate": 4.136996204590638e-06, "loss": 0.9875, "step": 9540 }, { "epoch": 1.72, "learning_rate": 4.134285197903488e-06, "loss": 0.8492, "step": 9541 }, { "epoch": 1.72, "learning_rate": 4.131574191216339e-06, "loss": 0.2111, "step": 9542 }, { "epoch": 1.72, "learning_rate": 4.128863184529188e-06, "loss": 0.759, "step": 9543 }, { "epoch": 1.72, "learning_rate": 4.126152177842039e-06, "loss": 1.254, "step": 9544 }, { "epoch": 1.73, "learning_rate": 4.123441171154889e-06, "loss": 0.6364, "step": 9545 }, { "epoch": 1.73, "learning_rate": 4.120730164467739e-06, "loss": 0.4625, "step": 9546 }, { "epoch": 1.73, "learning_rate": 4.118019157780589e-06, "loss": 0.3963, "step": 9547 }, { "epoch": 1.73, "learning_rate": 4.115308151093439e-06, "loss": 0.4284, "step": 9548 }, { "epoch": 1.73, "learning_rate": 4.1125971444062896e-06, "loss": 0.4909, "step": 9549 }, { "epoch": 1.73, "learning_rate": 4.10988613771914e-06, "loss": 0.6381, "step": 9550 }, { "epoch": 1.73, "learning_rate": 4.10717513103199e-06, "loss": 1.1513, "step": 9551 }, { "epoch": 1.73, "learning_rate": 4.10446412434484e-06, "loss": 0.8098, "step": 9552 }, { "epoch": 1.73, "learning_rate": 4.10175311765769e-06, "loss": 0.8029, "step": 9553 }, { "epoch": 1.73, "learning_rate": 4.09904211097054e-06, "loss": 1.1326, "step": 9554 }, { "epoch": 1.73, "learning_rate": 4.09633110428339e-06, "loss": 1.2485, "step": 9555 }, { "epoch": 1.73, "learning_rate": 4.093620097596241e-06, "loss": 1.1415, "step": 9556 }, { "epoch": 1.73, "learning_rate": 4.090909090909091e-06, "loss": 0.5008, "step": 9557 }, { "epoch": 1.73, "learning_rate": 4.088198084221942e-06, "loss": 0.523, "step": 9558 }, { "epoch": 1.73, "learning_rate": 4.085487077534791e-06, "loss": 0.4192, "step": 9559 }, { "epoch": 1.73, "learning_rate": 4.082776070847642e-06, "loss": 0.7489, "step": 9560 }, { "epoch": 1.73, "learning_rate": 4.080065064160491e-06, "loss": 0.7571, "step": 9561 }, { "epoch": 1.73, "learning_rate": 4.077354057473342e-06, "loss": 0.8337, "step": 9562 }, { "epoch": 1.73, "learning_rate": 4.0746430507861914e-06, "loss": 0.445, "step": 9563 }, { "epoch": 1.73, "learning_rate": 4.071932044099042e-06, "loss": 0.902, "step": 9564 }, { "epoch": 1.73, "learning_rate": 4.069221037411892e-06, "loss": 1.0017, "step": 9565 }, { "epoch": 1.73, "learning_rate": 4.066510030724743e-06, "loss": 0.5746, "step": 9566 }, { "epoch": 1.73, "learning_rate": 4.063799024037593e-06, "loss": 1.11, "step": 9567 }, { "epoch": 1.73, "learning_rate": 4.061088017350443e-06, "loss": 0.6818, "step": 9568 }, { "epoch": 1.73, "learning_rate": 4.058377010663293e-06, "loss": 0.8883, "step": 9569 }, { "epoch": 1.73, "learning_rate": 4.055666003976143e-06, "loss": 1.2243, "step": 9570 }, { "epoch": 1.73, "learning_rate": 4.052954997288993e-06, "loss": 0.3715, "step": 9571 }, { "epoch": 1.73, "learning_rate": 4.0502439906018435e-06, "loss": 1.948, "step": 9572 }, { "epoch": 1.73, "learning_rate": 4.047532983914694e-06, "loss": 0.7942, "step": 9573 }, { "epoch": 1.73, "learning_rate": 4.044821977227544e-06, "loss": 0.8005, "step": 9574 }, { "epoch": 1.73, "learning_rate": 4.042110970540394e-06, "loss": 0.3224, "step": 9575 }, { "epoch": 1.73, "learning_rate": 4.039399963853244e-06, "loss": 0.4293, "step": 9576 }, { "epoch": 1.73, "learning_rate": 4.036688957166094e-06, "loss": 0.5868, "step": 9577 }, { "epoch": 1.73, "learning_rate": 4.033977950478945e-06, "loss": 0.3092, "step": 9578 }, { "epoch": 1.73, "learning_rate": 4.0312669437917944e-06, "loss": 0.4705, "step": 9579 }, { "epoch": 1.73, "learning_rate": 4.028555937104645e-06, "loss": 0.4655, "step": 9580 }, { "epoch": 1.73, "learning_rate": 4.025844930417495e-06, "loss": 0.5963, "step": 9581 }, { "epoch": 1.73, "learning_rate": 4.023133923730346e-06, "loss": 0.5361, "step": 9582 }, { "epoch": 1.73, "learning_rate": 4.020422917043195e-06, "loss": 0.4996, "step": 9583 }, { "epoch": 1.73, "learning_rate": 4.017711910356046e-06, "loss": 1.1415, "step": 9584 }, { "epoch": 1.73, "learning_rate": 4.015000903668895e-06, "loss": 0.9153, "step": 9585 }, { "epoch": 1.73, "learning_rate": 4.012289896981746e-06, "loss": 0.7583, "step": 9586 }, { "epoch": 1.73, "learning_rate": 4.009578890294596e-06, "loss": 1.1842, "step": 9587 }, { "epoch": 1.73, "learning_rate": 4.0068678836074465e-06, "loss": 0.6722, "step": 9588 }, { "epoch": 1.73, "learning_rate": 4.004156876920297e-06, "loss": 0.5988, "step": 9589 }, { "epoch": 1.73, "learning_rate": 4.001445870233147e-06, "loss": 1.2633, "step": 9590 }, { "epoch": 1.73, "learning_rate": 3.998734863545997e-06, "loss": 1.0151, "step": 9591 }, { "epoch": 1.73, "learning_rate": 3.996023856858847e-06, "loss": 0.4393, "step": 9592 }, { "epoch": 1.73, "learning_rate": 3.993312850171697e-06, "loss": 1.2709, "step": 9593 }, { "epoch": 1.73, "learning_rate": 3.990601843484547e-06, "loss": 0.3904, "step": 9594 }, { "epoch": 1.73, "learning_rate": 3.9878908367973974e-06, "loss": 0.6969, "step": 9595 }, { "epoch": 1.73, "learning_rate": 3.9851798301102476e-06, "loss": 0.939, "step": 9596 }, { "epoch": 1.73, "learning_rate": 3.982468823423098e-06, "loss": 0.7608, "step": 9597 }, { "epoch": 1.73, "learning_rate": 3.979757816735949e-06, "loss": 0.6325, "step": 9598 }, { "epoch": 1.73, "learning_rate": 3.977046810048798e-06, "loss": 0.6731, "step": 9599 }, { "epoch": 1.74, "learning_rate": 3.974335803361649e-06, "loss": 0.4779, "step": 9600 }, { "epoch": 1.74, "learning_rate": 3.971624796674498e-06, "loss": 0.4265, "step": 9601 }, { "epoch": 1.74, "learning_rate": 3.968913789987349e-06, "loss": 0.6696, "step": 9602 }, { "epoch": 1.74, "learning_rate": 3.9662027833001985e-06, "loss": 0.8213, "step": 9603 }, { "epoch": 1.74, "learning_rate": 3.9634917766130495e-06, "loss": 0.7875, "step": 9604 }, { "epoch": 1.74, "learning_rate": 3.960780769925899e-06, "loss": 1.2326, "step": 9605 }, { "epoch": 1.74, "learning_rate": 3.95806976323875e-06, "loss": 0.6739, "step": 9606 }, { "epoch": 1.74, "learning_rate": 3.955358756551599e-06, "loss": 0.9987, "step": 9607 }, { "epoch": 1.74, "learning_rate": 3.95264774986445e-06, "loss": 0.4666, "step": 9608 }, { "epoch": 1.74, "learning_rate": 3.9499367431773e-06, "loss": 0.6388, "step": 9609 }, { "epoch": 1.74, "learning_rate": 3.94722573649015e-06, "loss": 0.4243, "step": 9610 }, { "epoch": 1.74, "learning_rate": 3.944514729803e-06, "loss": 0.9779, "step": 9611 }, { "epoch": 1.74, "learning_rate": 3.9418037231158506e-06, "loss": 0.5404, "step": 9612 }, { "epoch": 1.74, "learning_rate": 3.939092716428701e-06, "loss": 0.9921, "step": 9613 }, { "epoch": 1.74, "learning_rate": 3.936381709741551e-06, "loss": 0.7023, "step": 9614 }, { "epoch": 1.74, "learning_rate": 3.933670703054401e-06, "loss": 0.3834, "step": 9615 }, { "epoch": 1.74, "learning_rate": 3.930959696367251e-06, "loss": 0.39, "step": 9616 }, { "epoch": 1.74, "learning_rate": 3.928248689680101e-06, "loss": 0.2781, "step": 9617 }, { "epoch": 1.74, "learning_rate": 3.925537682992952e-06, "loss": 0.4108, "step": 9618 }, { "epoch": 1.74, "learning_rate": 3.9228266763058015e-06, "loss": 0.8558, "step": 9619 }, { "epoch": 1.74, "learning_rate": 3.9201156696186525e-06, "loss": 0.2474, "step": 9620 }, { "epoch": 1.74, "learning_rate": 3.917404662931502e-06, "loss": 1.2344, "step": 9621 }, { "epoch": 1.74, "learning_rate": 3.914693656244353e-06, "loss": 1.1337, "step": 9622 }, { "epoch": 1.74, "learning_rate": 3.911982649557202e-06, "loss": 1.0948, "step": 9623 }, { "epoch": 1.74, "learning_rate": 3.909271642870053e-06, "loss": 0.2201, "step": 9624 }, { "epoch": 1.74, "learning_rate": 3.906560636182902e-06, "loss": 0.7572, "step": 9625 }, { "epoch": 1.74, "learning_rate": 3.903849629495753e-06, "loss": 0.9082, "step": 9626 }, { "epoch": 1.74, "learning_rate": 3.9011386228086026e-06, "loss": 0.4791, "step": 9627 }, { "epoch": 1.74, "learning_rate": 3.8984276161214535e-06, "loss": 0.3645, "step": 9628 }, { "epoch": 1.74, "learning_rate": 3.895716609434304e-06, "loss": 0.6059, "step": 9629 }, { "epoch": 1.74, "learning_rate": 3.893005602747154e-06, "loss": 0.7769, "step": 9630 }, { "epoch": 1.74, "learning_rate": 3.890294596060004e-06, "loss": 0.761, "step": 9631 }, { "epoch": 1.74, "learning_rate": 3.887583589372854e-06, "loss": 1.1946, "step": 9632 }, { "epoch": 1.74, "learning_rate": 3.884872582685704e-06, "loss": 1.0597, "step": 9633 }, { "epoch": 1.74, "learning_rate": 3.882161575998554e-06, "loss": 0.332, "step": 9634 }, { "epoch": 1.74, "learning_rate": 3.8794505693114045e-06, "loss": 0.7806, "step": 9635 }, { "epoch": 1.74, "learning_rate": 3.876739562624255e-06, "loss": 0.9545, "step": 9636 }, { "epoch": 1.74, "learning_rate": 3.874028555937105e-06, "loss": 0.3832, "step": 9637 }, { "epoch": 1.74, "learning_rate": 3.871317549249955e-06, "loss": 0.4418, "step": 9638 }, { "epoch": 1.74, "learning_rate": 3.868606542562805e-06, "loss": 0.4998, "step": 9639 }, { "epoch": 1.74, "learning_rate": 3.865895535875656e-06, "loss": 0.6852, "step": 9640 }, { "epoch": 1.74, "learning_rate": 3.863184529188505e-06, "loss": 1.0239, "step": 9641 }, { "epoch": 1.74, "learning_rate": 3.860473522501356e-06, "loss": 2.0141, "step": 9642 }, { "epoch": 1.74, "learning_rate": 3.8577625158142056e-06, "loss": 1.022, "step": 9643 }, { "epoch": 1.74, "learning_rate": 3.8550515091270565e-06, "loss": 0.8086, "step": 9644 }, { "epoch": 1.74, "learning_rate": 3.852340502439906e-06, "loss": 0.8456, "step": 9645 }, { "epoch": 1.74, "learning_rate": 3.849629495752757e-06, "loss": 1.0007, "step": 9646 }, { "epoch": 1.74, "learning_rate": 3.846918489065606e-06, "loss": 1.0984, "step": 9647 }, { "epoch": 1.74, "learning_rate": 3.844207482378457e-06, "loss": 1.0914, "step": 9648 }, { "epoch": 1.74, "learning_rate": 3.841496475691307e-06, "loss": 0.8604, "step": 9649 }, { "epoch": 1.74, "learning_rate": 3.838785469004157e-06, "loss": 0.8404, "step": 9650 }, { "epoch": 1.74, "learning_rate": 3.8360744623170075e-06, "loss": 0.3993, "step": 9651 }, { "epoch": 1.74, "learning_rate": 3.833363455629857e-06, "loss": 0.9564, "step": 9652 }, { "epoch": 1.74, "learning_rate": 3.830652448942708e-06, "loss": 0.8051, "step": 9653 }, { "epoch": 1.74, "learning_rate": 3.827941442255557e-06, "loss": 0.6171, "step": 9654 }, { "epoch": 1.74, "learning_rate": 3.825230435568408e-06, "loss": 0.3182, "step": 9655 }, { "epoch": 1.75, "learning_rate": 3.822519428881257e-06, "loss": 0.4127, "step": 9656 }, { "epoch": 1.75, "learning_rate": 3.819808422194108e-06, "loss": 0.6127, "step": 9657 }, { "epoch": 1.75, "learning_rate": 3.817097415506958e-06, "loss": 0.7271, "step": 9658 }, { "epoch": 1.75, "learning_rate": 3.8143864088198086e-06, "loss": 1.0005, "step": 9659 }, { "epoch": 1.75, "learning_rate": 3.8116754021326583e-06, "loss": 0.5633, "step": 9660 }, { "epoch": 1.75, "learning_rate": 3.808964395445509e-06, "loss": 0.7104, "step": 9661 }, { "epoch": 1.75, "learning_rate": 3.8062533887583585e-06, "loss": 0.5262, "step": 9662 }, { "epoch": 1.75, "learning_rate": 3.803542382071209e-06, "loss": 0.7421, "step": 9663 }, { "epoch": 1.75, "learning_rate": 3.800831375384059e-06, "loss": 0.511, "step": 9664 }, { "epoch": 1.75, "learning_rate": 3.7981203686969098e-06, "loss": 0.8114, "step": 9665 }, { "epoch": 1.75, "learning_rate": 3.7954093620097595e-06, "loss": 0.3933, "step": 9666 }, { "epoch": 1.75, "learning_rate": 3.79269835532261e-06, "loss": 0.8046, "step": 9667 }, { "epoch": 1.75, "learning_rate": 3.7899873486354598e-06, "loss": 0.9466, "step": 9668 }, { "epoch": 1.75, "learning_rate": 3.7872763419483103e-06, "loss": 0.8342, "step": 9669 }, { "epoch": 1.75, "learning_rate": 3.78456533526116e-06, "loss": 0.3141, "step": 9670 }, { "epoch": 1.75, "learning_rate": 3.7818543285740106e-06, "loss": 1.3991, "step": 9671 }, { "epoch": 1.75, "learning_rate": 3.7791433218868603e-06, "loss": 0.4658, "step": 9672 }, { "epoch": 1.75, "learning_rate": 3.776432315199711e-06, "loss": 0.9234, "step": 9673 }, { "epoch": 1.75, "learning_rate": 3.7737213085125606e-06, "loss": 1.1364, "step": 9674 }, { "epoch": 1.75, "learning_rate": 3.7710103018254116e-06, "loss": 0.2972, "step": 9675 }, { "epoch": 1.75, "learning_rate": 3.7682992951382613e-06, "loss": 0.7504, "step": 9676 }, { "epoch": 1.75, "learning_rate": 3.765588288451112e-06, "loss": 0.859, "step": 9677 }, { "epoch": 1.75, "learning_rate": 3.7628772817639615e-06, "loss": 0.2717, "step": 9678 }, { "epoch": 1.75, "learning_rate": 3.760166275076812e-06, "loss": 0.5946, "step": 9679 }, { "epoch": 1.75, "learning_rate": 3.757455268389662e-06, "loss": 0.8904, "step": 9680 }, { "epoch": 1.75, "learning_rate": 3.7547442617025124e-06, "loss": 1.1015, "step": 9681 }, { "epoch": 1.75, "learning_rate": 3.752033255015362e-06, "loss": 0.3232, "step": 9682 }, { "epoch": 1.75, "learning_rate": 3.7493222483282126e-06, "loss": 1.6057, "step": 9683 }, { "epoch": 1.75, "learning_rate": 3.7466112416410628e-06, "loss": 0.7313, "step": 9684 }, { "epoch": 1.75, "learning_rate": 3.743900234953913e-06, "loss": 0.4386, "step": 9685 }, { "epoch": 1.75, "learning_rate": 3.7411892282667635e-06, "loss": 0.8, "step": 9686 }, { "epoch": 1.75, "learning_rate": 3.7384782215796136e-06, "loss": 0.6676, "step": 9687 }, { "epoch": 1.75, "learning_rate": 3.7357672148924637e-06, "loss": 0.4684, "step": 9688 }, { "epoch": 1.75, "learning_rate": 3.733056208205314e-06, "loss": 0.8062, "step": 9689 }, { "epoch": 1.75, "learning_rate": 3.730345201518164e-06, "loss": 0.5703, "step": 9690 }, { "epoch": 1.75, "learning_rate": 3.727634194831014e-06, "loss": 0.6991, "step": 9691 }, { "epoch": 1.75, "learning_rate": 3.7249231881438643e-06, "loss": 0.525, "step": 9692 }, { "epoch": 1.75, "learning_rate": 3.7222121814567144e-06, "loss": 0.988, "step": 9693 }, { "epoch": 1.75, "learning_rate": 3.7195011747695645e-06, "loss": 0.816, "step": 9694 }, { "epoch": 1.75, "learning_rate": 3.7167901680824147e-06, "loss": 0.5835, "step": 9695 }, { "epoch": 1.75, "learning_rate": 3.7140791613952652e-06, "loss": 0.6724, "step": 9696 }, { "epoch": 1.75, "learning_rate": 3.7113681547081154e-06, "loss": 0.6507, "step": 9697 }, { "epoch": 1.75, "learning_rate": 3.7086571480209655e-06, "loss": 1.057, "step": 9698 }, { "epoch": 1.75, "learning_rate": 3.7059461413338156e-06, "loss": 0.5659, "step": 9699 }, { "epoch": 1.75, "learning_rate": 3.7032351346466658e-06, "loss": 1.067, "step": 9700 }, { "epoch": 1.75, "learning_rate": 3.700524127959516e-06, "loss": 0.541, "step": 9701 }, { "epoch": 1.75, "learning_rate": 3.697813121272366e-06, "loss": 1.313, "step": 9702 }, { "epoch": 1.75, "learning_rate": 3.695102114585216e-06, "loss": 0.5392, "step": 9703 }, { "epoch": 1.75, "learning_rate": 3.6923911078980663e-06, "loss": 1.1528, "step": 9704 }, { "epoch": 1.75, "learning_rate": 3.6896801012109164e-06, "loss": 0.3625, "step": 9705 }, { "epoch": 1.75, "learning_rate": 3.6869690945237666e-06, "loss": 1.1705, "step": 9706 }, { "epoch": 1.75, "learning_rate": 3.684258087836617e-06, "loss": 0.8562, "step": 9707 }, { "epoch": 1.75, "learning_rate": 3.6815470811494673e-06, "loss": 0.6913, "step": 9708 }, { "epoch": 1.75, "learning_rate": 3.6788360744623174e-06, "loss": 0.6998, "step": 9709 }, { "epoch": 1.75, "learning_rate": 3.6761250677751675e-06, "loss": 0.3408, "step": 9710 }, { "epoch": 1.76, "learning_rate": 3.6734140610880177e-06, "loss": 0.28, "step": 9711 }, { "epoch": 1.76, "learning_rate": 3.670703054400868e-06, "loss": 0.9793, "step": 9712 }, { "epoch": 1.76, "learning_rate": 3.667992047713718e-06, "loss": 0.6835, "step": 9713 }, { "epoch": 1.76, "learning_rate": 3.665281041026568e-06, "loss": 0.8273, "step": 9714 }, { "epoch": 1.76, "learning_rate": 3.6625700343394178e-06, "loss": 0.6126, "step": 9715 }, { "epoch": 1.76, "learning_rate": 3.659859027652268e-06, "loss": 1.2518, "step": 9716 }, { "epoch": 1.76, "learning_rate": 3.6571480209651185e-06, "loss": 0.7079, "step": 9717 }, { "epoch": 1.76, "learning_rate": 3.6544370142779686e-06, "loss": 0.4783, "step": 9718 }, { "epoch": 1.76, "learning_rate": 3.6517260075908187e-06, "loss": 0.8771, "step": 9719 }, { "epoch": 1.76, "learning_rate": 3.649015000903669e-06, "loss": 0.9096, "step": 9720 }, { "epoch": 1.76, "learning_rate": 3.646303994216519e-06, "loss": 1.0454, "step": 9721 }, { "epoch": 1.76, "learning_rate": 3.643592987529369e-06, "loss": 0.4047, "step": 9722 }, { "epoch": 1.76, "learning_rate": 3.6408819808422193e-06, "loss": 0.7325, "step": 9723 }, { "epoch": 1.76, "learning_rate": 3.6381709741550694e-06, "loss": 0.431, "step": 9724 }, { "epoch": 1.76, "learning_rate": 3.6354599674679195e-06, "loss": 0.605, "step": 9725 }, { "epoch": 1.76, "learning_rate": 3.6327489607807697e-06, "loss": 0.9432, "step": 9726 }, { "epoch": 1.76, "learning_rate": 3.6300379540936202e-06, "loss": 1.3015, "step": 9727 }, { "epoch": 1.76, "learning_rate": 3.6273269474064704e-06, "loss": 0.6279, "step": 9728 }, { "epoch": 1.76, "learning_rate": 3.6246159407193205e-06, "loss": 0.8901, "step": 9729 }, { "epoch": 1.76, "learning_rate": 3.6219049340321706e-06, "loss": 0.553, "step": 9730 }, { "epoch": 1.76, "learning_rate": 3.6191939273450208e-06, "loss": 1.6206, "step": 9731 }, { "epoch": 1.76, "learning_rate": 3.616482920657871e-06, "loss": 0.4881, "step": 9732 }, { "epoch": 1.76, "learning_rate": 3.613771913970721e-06, "loss": 0.9155, "step": 9733 }, { "epoch": 1.76, "learning_rate": 3.611060907283571e-06, "loss": 0.471, "step": 9734 }, { "epoch": 1.76, "learning_rate": 3.6083499005964213e-06, "loss": 0.5447, "step": 9735 }, { "epoch": 1.76, "learning_rate": 3.6056388939092714e-06, "loss": 0.471, "step": 9736 }, { "epoch": 1.76, "learning_rate": 3.6029278872221216e-06, "loss": 0.3651, "step": 9737 }, { "epoch": 1.76, "learning_rate": 3.600216880534972e-06, "loss": 0.6613, "step": 9738 }, { "epoch": 1.76, "learning_rate": 3.5975058738478223e-06, "loss": 0.5009, "step": 9739 }, { "epoch": 1.76, "learning_rate": 3.5947948671606724e-06, "loss": 0.8997, "step": 9740 }, { "epoch": 1.76, "learning_rate": 3.5920838604735225e-06, "loss": 0.5292, "step": 9741 }, { "epoch": 1.76, "learning_rate": 3.5893728537863727e-06, "loss": 0.5906, "step": 9742 }, { "epoch": 1.76, "learning_rate": 3.586661847099223e-06, "loss": 0.6701, "step": 9743 }, { "epoch": 1.76, "learning_rate": 3.583950840412073e-06, "loss": 0.8378, "step": 9744 }, { "epoch": 1.76, "learning_rate": 3.581239833724923e-06, "loss": 0.5185, "step": 9745 }, { "epoch": 1.76, "learning_rate": 3.578528827037773e-06, "loss": 0.476, "step": 9746 }, { "epoch": 1.76, "learning_rate": 3.5758178203506233e-06, "loss": 0.6434, "step": 9747 }, { "epoch": 1.76, "learning_rate": 3.573106813663474e-06, "loss": 0.875, "step": 9748 }, { "epoch": 1.76, "learning_rate": 3.570395806976324e-06, "loss": 0.8034, "step": 9749 }, { "epoch": 1.76, "learning_rate": 3.567684800289174e-06, "loss": 0.6604, "step": 9750 }, { "epoch": 1.76, "learning_rate": 3.5649737936020243e-06, "loss": 0.576, "step": 9751 }, { "epoch": 1.76, "learning_rate": 3.5622627869148744e-06, "loss": 0.4867, "step": 9752 }, { "epoch": 1.76, "learning_rate": 3.5595517802277246e-06, "loss": 0.6011, "step": 9753 }, { "epoch": 1.76, "learning_rate": 3.5568407735405747e-06, "loss": 1.1433, "step": 9754 }, { "epoch": 1.76, "learning_rate": 3.554129766853425e-06, "loss": 0.8277, "step": 9755 }, { "epoch": 1.76, "learning_rate": 3.551418760166275e-06, "loss": 0.6959, "step": 9756 }, { "epoch": 1.76, "learning_rate": 3.548707753479125e-06, "loss": 0.9101, "step": 9757 }, { "epoch": 1.76, "learning_rate": 3.5459967467919752e-06, "loss": 1.0671, "step": 9758 }, { "epoch": 1.76, "learning_rate": 3.543285740104826e-06, "loss": 0.3417, "step": 9759 }, { "epoch": 1.76, "learning_rate": 3.540574733417676e-06, "loss": 0.3381, "step": 9760 }, { "epoch": 1.76, "learning_rate": 3.537863726730526e-06, "loss": 0.6561, "step": 9761 }, { "epoch": 1.76, "learning_rate": 3.535152720043376e-06, "loss": 1.0139, "step": 9762 }, { "epoch": 1.76, "learning_rate": 3.5324417133562263e-06, "loss": 1.0067, "step": 9763 }, { "epoch": 1.76, "learning_rate": 3.5297307066690765e-06, "loss": 0.5368, "step": 9764 }, { "epoch": 1.76, "learning_rate": 3.5270196999819266e-06, "loss": 1.6751, "step": 9765 }, { "epoch": 1.77, "learning_rate": 3.5243086932947767e-06, "loss": 0.554, "step": 9766 }, { "epoch": 1.77, "learning_rate": 3.521597686607627e-06, "loss": 0.7174, "step": 9767 }, { "epoch": 1.77, "learning_rate": 3.518886679920477e-06, "loss": 1.5647, "step": 9768 }, { "epoch": 1.77, "learning_rate": 3.5161756732333276e-06, "loss": 0.6134, "step": 9769 }, { "epoch": 1.77, "learning_rate": 3.5134646665461777e-06, "loss": 0.5974, "step": 9770 }, { "epoch": 1.77, "learning_rate": 3.510753659859028e-06, "loss": 1.3361, "step": 9771 }, { "epoch": 1.77, "learning_rate": 3.508042653171878e-06, "loss": 1.0205, "step": 9772 }, { "epoch": 1.77, "learning_rate": 3.505331646484728e-06, "loss": 0.9668, "step": 9773 }, { "epoch": 1.77, "learning_rate": 3.5026206397975782e-06, "loss": 0.7642, "step": 9774 }, { "epoch": 1.77, "learning_rate": 3.4999096331104284e-06, "loss": 1.0051, "step": 9775 }, { "epoch": 1.77, "learning_rate": 3.4971986264232785e-06, "loss": 0.4828, "step": 9776 }, { "epoch": 1.77, "learning_rate": 3.4944876197361286e-06, "loss": 0.8115, "step": 9777 }, { "epoch": 1.77, "learning_rate": 3.4917766130489788e-06, "loss": 0.8454, "step": 9778 }, { "epoch": 1.77, "learning_rate": 3.489065606361829e-06, "loss": 0.7556, "step": 9779 }, { "epoch": 1.77, "learning_rate": 3.4863545996746795e-06, "loss": 0.7336, "step": 9780 }, { "epoch": 1.77, "learning_rate": 3.4836435929875296e-06, "loss": 0.5414, "step": 9781 }, { "epoch": 1.77, "learning_rate": 3.4809325863003797e-06, "loss": 0.9604, "step": 9782 }, { "epoch": 1.77, "learning_rate": 3.47822157961323e-06, "loss": 0.8943, "step": 9783 }, { "epoch": 1.77, "learning_rate": 3.47551057292608e-06, "loss": 0.6449, "step": 9784 }, { "epoch": 1.77, "learning_rate": 3.47279956623893e-06, "loss": 0.9081, "step": 9785 }, { "epoch": 1.77, "learning_rate": 3.4700885595517803e-06, "loss": 0.4128, "step": 9786 }, { "epoch": 1.77, "learning_rate": 3.4673775528646304e-06, "loss": 0.8146, "step": 9787 }, { "epoch": 1.77, "learning_rate": 3.4646665461774805e-06, "loss": 0.9921, "step": 9788 }, { "epoch": 1.77, "learning_rate": 3.4619555394903307e-06, "loss": 0.4494, "step": 9789 }, { "epoch": 1.77, "learning_rate": 3.4592445328031812e-06, "loss": 0.7012, "step": 9790 }, { "epoch": 1.77, "learning_rate": 3.4565335261160314e-06, "loss": 0.952, "step": 9791 }, { "epoch": 1.77, "learning_rate": 3.4538225194288815e-06, "loss": 0.456, "step": 9792 }, { "epoch": 1.77, "learning_rate": 3.4511115127417316e-06, "loss": 1.187, "step": 9793 }, { "epoch": 1.77, "learning_rate": 3.4484005060545818e-06, "loss": 0.6563, "step": 9794 }, { "epoch": 1.77, "learning_rate": 3.445689499367432e-06, "loss": 1.1464, "step": 9795 }, { "epoch": 1.77, "learning_rate": 3.442978492680282e-06, "loss": 0.616, "step": 9796 }, { "epoch": 1.77, "learning_rate": 3.440267485993132e-06, "loss": 0.4379, "step": 9797 }, { "epoch": 1.77, "learning_rate": 3.4375564793059823e-06, "loss": 0.7548, "step": 9798 }, { "epoch": 1.77, "learning_rate": 3.4348454726188324e-06, "loss": 0.482, "step": 9799 }, { "epoch": 1.77, "learning_rate": 3.4321344659316826e-06, "loss": 1.0727, "step": 9800 }, { "epoch": 1.77, "learning_rate": 3.429423459244533e-06, "loss": 0.3801, "step": 9801 }, { "epoch": 1.77, "learning_rate": 3.4267124525573833e-06, "loss": 0.6252, "step": 9802 }, { "epoch": 1.77, "learning_rate": 3.4240014458702334e-06, "loss": 0.8532, "step": 9803 }, { "epoch": 1.77, "learning_rate": 3.4212904391830835e-06, "loss": 0.647, "step": 9804 }, { "epoch": 1.77, "learning_rate": 3.4185794324959337e-06, "loss": 0.9253, "step": 9805 }, { "epoch": 1.77, "learning_rate": 3.415868425808784e-06, "loss": 0.62, "step": 9806 }, { "epoch": 1.77, "learning_rate": 3.413157419121634e-06, "loss": 0.4528, "step": 9807 }, { "epoch": 1.77, "learning_rate": 3.410446412434484e-06, "loss": 0.6543, "step": 9808 }, { "epoch": 1.77, "learning_rate": 3.407735405747334e-06, "loss": 0.8096, "step": 9809 }, { "epoch": 1.77, "learning_rate": 3.4050243990601843e-06, "loss": 0.6247, "step": 9810 }, { "epoch": 1.77, "learning_rate": 3.402313392373035e-06, "loss": 0.4869, "step": 9811 }, { "epoch": 1.77, "learning_rate": 3.399602385685885e-06, "loss": 0.6954, "step": 9812 }, { "epoch": 1.77, "learning_rate": 3.396891378998735e-06, "loss": 0.4137, "step": 9813 }, { "epoch": 1.77, "learning_rate": 3.3941803723115853e-06, "loss": 0.8669, "step": 9814 }, { "epoch": 1.77, "learning_rate": 3.3914693656244354e-06, "loss": 0.8351, "step": 9815 }, { "epoch": 1.77, "learning_rate": 3.3887583589372856e-06, "loss": 1.2858, "step": 9816 }, { "epoch": 1.77, "learning_rate": 3.3860473522501357e-06, "loss": 0.6756, "step": 9817 }, { "epoch": 1.77, "learning_rate": 3.383336345562986e-06, "loss": 0.3924, "step": 9818 }, { "epoch": 1.77, "learning_rate": 3.380625338875836e-06, "loss": 0.6874, "step": 9819 }, { "epoch": 1.77, "learning_rate": 3.377914332188686e-06, "loss": 0.7277, "step": 9820 }, { "epoch": 1.77, "learning_rate": 3.3752033255015367e-06, "loss": 0.812, "step": 9821 }, { "epoch": 1.78, "learning_rate": 3.372492318814387e-06, "loss": 0.6922, "step": 9822 }, { "epoch": 1.78, "learning_rate": 3.369781312127237e-06, "loss": 0.4951, "step": 9823 }, { "epoch": 1.78, "learning_rate": 3.367070305440087e-06, "loss": 0.7613, "step": 9824 }, { "epoch": 1.78, "learning_rate": 3.364359298752937e-06, "loss": 0.7569, "step": 9825 }, { "epoch": 1.78, "learning_rate": 3.3616482920657873e-06, "loss": 0.4706, "step": 9826 }, { "epoch": 1.78, "learning_rate": 3.3589372853786375e-06, "loss": 0.4986, "step": 9827 }, { "epoch": 1.78, "learning_rate": 3.3562262786914876e-06, "loss": 0.7534, "step": 9828 }, { "epoch": 1.78, "learning_rate": 3.3535152720043377e-06, "loss": 0.9283, "step": 9829 }, { "epoch": 1.78, "learning_rate": 3.350804265317188e-06, "loss": 0.7925, "step": 9830 }, { "epoch": 1.78, "learning_rate": 3.348093258630038e-06, "loss": 0.7321, "step": 9831 }, { "epoch": 1.78, "learning_rate": 3.3453822519428886e-06, "loss": 1.1766, "step": 9832 }, { "epoch": 1.78, "learning_rate": 3.3426712452557387e-06, "loss": 0.5265, "step": 9833 }, { "epoch": 1.78, "learning_rate": 3.339960238568589e-06, "loss": 1.0538, "step": 9834 }, { "epoch": 1.78, "learning_rate": 3.337249231881439e-06, "loss": 1.3031, "step": 9835 }, { "epoch": 1.78, "learning_rate": 3.334538225194289e-06, "loss": 0.5133, "step": 9836 }, { "epoch": 1.78, "learning_rate": 3.3318272185071392e-06, "loss": 0.9593, "step": 9837 }, { "epoch": 1.78, "learning_rate": 3.3291162118199894e-06, "loss": 0.8543, "step": 9838 }, { "epoch": 1.78, "learning_rate": 3.3264052051328395e-06, "loss": 0.6113, "step": 9839 }, { "epoch": 1.78, "learning_rate": 3.3236941984456896e-06, "loss": 0.3548, "step": 9840 }, { "epoch": 1.78, "learning_rate": 3.3209831917585398e-06, "loss": 0.4825, "step": 9841 }, { "epoch": 1.78, "learning_rate": 3.3182721850713903e-06, "loss": 0.3474, "step": 9842 }, { "epoch": 1.78, "learning_rate": 3.3155611783842405e-06, "loss": 0.6888, "step": 9843 }, { "epoch": 1.78, "learning_rate": 3.31285017169709e-06, "loss": 0.3539, "step": 9844 }, { "epoch": 1.78, "learning_rate": 3.3101391650099403e-06, "loss": 0.8553, "step": 9845 }, { "epoch": 1.78, "learning_rate": 3.3074281583227904e-06, "loss": 0.9283, "step": 9846 }, { "epoch": 1.78, "learning_rate": 3.3047171516356406e-06, "loss": 0.4808, "step": 9847 }, { "epoch": 1.78, "learning_rate": 3.3020061449484907e-06, "loss": 0.7909, "step": 9848 }, { "epoch": 1.78, "learning_rate": 3.299295138261341e-06, "loss": 0.4704, "step": 9849 }, { "epoch": 1.78, "learning_rate": 3.296584131574191e-06, "loss": 0.4734, "step": 9850 }, { "epoch": 1.78, "learning_rate": 3.293873124887041e-06, "loss": 0.6067, "step": 9851 }, { "epoch": 1.78, "learning_rate": 3.2911621181998917e-06, "loss": 0.2858, "step": 9852 }, { "epoch": 1.78, "learning_rate": 3.288451111512742e-06, "loss": 0.8828, "step": 9853 }, { "epoch": 1.78, "learning_rate": 3.285740104825592e-06, "loss": 0.9451, "step": 9854 }, { "epoch": 1.78, "learning_rate": 3.283029098138442e-06, "loss": 0.82, "step": 9855 }, { "epoch": 1.78, "learning_rate": 3.280318091451292e-06, "loss": 1.4849, "step": 9856 }, { "epoch": 1.78, "learning_rate": 3.2776070847641423e-06, "loss": 0.8042, "step": 9857 }, { "epoch": 1.78, "learning_rate": 3.2748960780769925e-06, "loss": 0.857, "step": 9858 }, { "epoch": 1.78, "learning_rate": 3.2721850713898426e-06, "loss": 0.4337, "step": 9859 }, { "epoch": 1.78, "learning_rate": 3.2694740647026927e-06, "loss": 0.516, "step": 9860 }, { "epoch": 1.78, "learning_rate": 3.266763058015543e-06, "loss": 1.2103, "step": 9861 }, { "epoch": 1.78, "learning_rate": 3.264052051328393e-06, "loss": 0.6992, "step": 9862 }, { "epoch": 1.78, "learning_rate": 3.2613410446412436e-06, "loss": 1.2413, "step": 9863 }, { "epoch": 1.78, "learning_rate": 3.2586300379540937e-06, "loss": 0.6091, "step": 9864 }, { "epoch": 1.78, "learning_rate": 3.255919031266944e-06, "loss": 1.7611, "step": 9865 }, { "epoch": 1.78, "learning_rate": 3.253208024579794e-06, "loss": 0.3433, "step": 9866 }, { "epoch": 1.78, "learning_rate": 3.250497017892644e-06, "loss": 0.5347, "step": 9867 }, { "epoch": 1.78, "learning_rate": 3.2477860112054942e-06, "loss": 0.663, "step": 9868 }, { "epoch": 1.78, "learning_rate": 3.2450750045183444e-06, "loss": 1.308, "step": 9869 }, { "epoch": 1.78, "learning_rate": 3.2423639978311945e-06, "loss": 0.4179, "step": 9870 }, { "epoch": 1.78, "learning_rate": 3.2396529911440446e-06, "loss": 0.937, "step": 9871 }, { "epoch": 1.78, "learning_rate": 3.2369419844568948e-06, "loss": 0.9298, "step": 9872 }, { "epoch": 1.78, "learning_rate": 3.2342309777697453e-06, "loss": 1.0607, "step": 9873 }, { "epoch": 1.78, "learning_rate": 3.2315199710825955e-06, "loss": 0.6899, "step": 9874 }, { "epoch": 1.78, "learning_rate": 3.2288089643954456e-06, "loss": 1.1725, "step": 9875 }, { "epoch": 1.78, "learning_rate": 3.2260979577082957e-06, "loss": 0.4809, "step": 9876 }, { "epoch": 1.79, "learning_rate": 3.223386951021146e-06, "loss": 1.179, "step": 9877 }, { "epoch": 1.79, "learning_rate": 3.220675944333996e-06, "loss": 0.7439, "step": 9878 }, { "epoch": 1.79, "learning_rate": 3.217964937646846e-06, "loss": 0.6429, "step": 9879 }, { "epoch": 1.79, "learning_rate": 3.2152539309596963e-06, "loss": 1.1772, "step": 9880 }, { "epoch": 1.79, "learning_rate": 3.2125429242725464e-06, "loss": 0.4364, "step": 9881 }, { "epoch": 1.79, "learning_rate": 3.2098319175853965e-06, "loss": 1.1531, "step": 9882 }, { "epoch": 1.79, "learning_rate": 3.2071209108982467e-06, "loss": 0.4993, "step": 9883 }, { "epoch": 1.79, "learning_rate": 3.2044099042110972e-06, "loss": 1.3863, "step": 9884 }, { "epoch": 1.79, "learning_rate": 3.2016988975239474e-06, "loss": 1.0983, "step": 9885 }, { "epoch": 1.79, "learning_rate": 3.1989878908367975e-06, "loss": 0.7523, "step": 9886 }, { "epoch": 1.79, "learning_rate": 3.1962768841496476e-06, "loss": 0.3715, "step": 9887 }, { "epoch": 1.79, "learning_rate": 3.1935658774624978e-06, "loss": 0.8535, "step": 9888 }, { "epoch": 1.79, "learning_rate": 3.190854870775348e-06, "loss": 0.614, "step": 9889 }, { "epoch": 1.79, "learning_rate": 3.188143864088198e-06, "loss": 0.644, "step": 9890 }, { "epoch": 1.79, "learning_rate": 3.185432857401048e-06, "loss": 0.1579, "step": 9891 }, { "epoch": 1.79, "learning_rate": 3.1827218507138983e-06, "loss": 0.445, "step": 9892 }, { "epoch": 1.79, "learning_rate": 3.1800108440267484e-06, "loss": 0.8848, "step": 9893 }, { "epoch": 1.79, "learning_rate": 3.177299837339599e-06, "loss": 0.8593, "step": 9894 }, { "epoch": 1.79, "learning_rate": 3.174588830652449e-06, "loss": 0.3292, "step": 9895 }, { "epoch": 1.79, "learning_rate": 3.1718778239652993e-06, "loss": 0.6752, "step": 9896 }, { "epoch": 1.79, "learning_rate": 3.1691668172781494e-06, "loss": 0.7168, "step": 9897 }, { "epoch": 1.79, "learning_rate": 3.1664558105909995e-06, "loss": 0.5621, "step": 9898 }, { "epoch": 1.79, "learning_rate": 3.1637448039038497e-06, "loss": 0.7576, "step": 9899 }, { "epoch": 1.79, "learning_rate": 3.1610337972167e-06, "loss": 0.4026, "step": 9900 }, { "epoch": 1.79, "learning_rate": 3.15832279052955e-06, "loss": 0.4536, "step": 9901 }, { "epoch": 1.79, "learning_rate": 3.1556117838424e-06, "loss": 0.7091, "step": 9902 }, { "epoch": 1.79, "learning_rate": 3.1529007771552502e-06, "loss": 0.4421, "step": 9903 }, { "epoch": 1.79, "learning_rate": 3.1501897704681004e-06, "loss": 0.4083, "step": 9904 }, { "epoch": 1.79, "learning_rate": 3.147478763780951e-06, "loss": 1.158, "step": 9905 }, { "epoch": 1.79, "learning_rate": 3.144767757093801e-06, "loss": 0.546, "step": 9906 }, { "epoch": 1.79, "learning_rate": 3.142056750406651e-06, "loss": 0.3709, "step": 9907 }, { "epoch": 1.79, "learning_rate": 3.1393457437195013e-06, "loss": 0.6616, "step": 9908 }, { "epoch": 1.79, "learning_rate": 3.1366347370323514e-06, "loss": 0.5863, "step": 9909 }, { "epoch": 1.79, "learning_rate": 3.1339237303452016e-06, "loss": 0.7143, "step": 9910 }, { "epoch": 1.79, "learning_rate": 3.1312127236580517e-06, "loss": 0.7278, "step": 9911 }, { "epoch": 1.79, "learning_rate": 3.128501716970902e-06, "loss": 0.7436, "step": 9912 }, { "epoch": 1.79, "learning_rate": 3.125790710283752e-06, "loss": 0.2662, "step": 9913 }, { "epoch": 1.79, "learning_rate": 3.123079703596602e-06, "loss": 0.6911, "step": 9914 }, { "epoch": 1.79, "learning_rate": 3.1203686969094527e-06, "loss": 1.2497, "step": 9915 }, { "epoch": 1.79, "learning_rate": 3.117657690222303e-06, "loss": 0.8006, "step": 9916 }, { "epoch": 1.79, "learning_rate": 3.114946683535153e-06, "loss": 1.1769, "step": 9917 }, { "epoch": 1.79, "learning_rate": 3.112235676848003e-06, "loss": 1.1592, "step": 9918 }, { "epoch": 1.79, "learning_rate": 3.109524670160853e-06, "loss": 0.78, "step": 9919 }, { "epoch": 1.79, "learning_rate": 3.1068136634737033e-06, "loss": 1.2233, "step": 9920 }, { "epoch": 1.79, "learning_rate": 3.1041026567865535e-06, "loss": 0.4786, "step": 9921 }, { "epoch": 1.79, "learning_rate": 3.1013916500994036e-06, "loss": 1.0258, "step": 9922 }, { "epoch": 1.79, "learning_rate": 3.0986806434122537e-06, "loss": 0.3603, "step": 9923 }, { "epoch": 1.79, "learning_rate": 3.095969636725104e-06, "loss": 0.8764, "step": 9924 }, { "epoch": 1.79, "learning_rate": 3.093258630037954e-06, "loss": 1.071, "step": 9925 }, { "epoch": 1.79, "learning_rate": 3.0905476233508046e-06, "loss": 1.3105, "step": 9926 }, { "epoch": 1.79, "learning_rate": 3.0878366166636547e-06, "loss": 0.5938, "step": 9927 }, { "epoch": 1.79, "learning_rate": 3.085125609976505e-06, "loss": 0.8896, "step": 9928 }, { "epoch": 1.79, "learning_rate": 3.082414603289355e-06, "loss": 0.4319, "step": 9929 }, { "epoch": 1.79, "learning_rate": 3.079703596602205e-06, "loss": 0.7746, "step": 9930 }, { "epoch": 1.79, "learning_rate": 3.0769925899150552e-06, "loss": 0.5577, "step": 9931 }, { "epoch": 1.8, "learning_rate": 3.0742815832279054e-06, "loss": 0.8535, "step": 9932 }, { "epoch": 1.8, "learning_rate": 3.0715705765407555e-06, "loss": 0.4326, "step": 9933 }, { "epoch": 1.8, "learning_rate": 3.0688595698536056e-06, "loss": 0.3985, "step": 9934 }, { "epoch": 1.8, "learning_rate": 3.0661485631664558e-06, "loss": 0.9847, "step": 9935 }, { "epoch": 1.8, "learning_rate": 3.0634375564793063e-06, "loss": 1.1321, "step": 9936 }, { "epoch": 1.8, "learning_rate": 3.0607265497921565e-06, "loss": 0.7124, "step": 9937 }, { "epoch": 1.8, "learning_rate": 3.0580155431050066e-06, "loss": 0.7556, "step": 9938 }, { "epoch": 1.8, "learning_rate": 3.0553045364178567e-06, "loss": 0.6134, "step": 9939 }, { "epoch": 1.8, "learning_rate": 3.052593529730707e-06, "loss": 0.8479, "step": 9940 }, { "epoch": 1.8, "learning_rate": 3.049882523043557e-06, "loss": 0.5303, "step": 9941 }, { "epoch": 1.8, "learning_rate": 3.047171516356407e-06, "loss": 1.388, "step": 9942 }, { "epoch": 1.8, "learning_rate": 3.0444605096692573e-06, "loss": 0.5986, "step": 9943 }, { "epoch": 1.8, "learning_rate": 3.0417495029821074e-06, "loss": 0.3123, "step": 9944 }, { "epoch": 1.8, "learning_rate": 3.0390384962949575e-06, "loss": 0.9021, "step": 9945 }, { "epoch": 1.8, "learning_rate": 3.036327489607808e-06, "loss": 0.4803, "step": 9946 }, { "epoch": 1.8, "learning_rate": 3.0336164829206582e-06, "loss": 0.5428, "step": 9947 }, { "epoch": 1.8, "learning_rate": 3.0309054762335084e-06, "loss": 0.5711, "step": 9948 }, { "epoch": 1.8, "learning_rate": 3.0281944695463585e-06, "loss": 0.3289, "step": 9949 }, { "epoch": 1.8, "learning_rate": 3.0254834628592086e-06, "loss": 0.737, "step": 9950 }, { "epoch": 1.8, "learning_rate": 3.0227724561720588e-06, "loss": 0.9093, "step": 9951 }, { "epoch": 1.8, "learning_rate": 3.020061449484909e-06, "loss": 1.4915, "step": 9952 }, { "epoch": 1.8, "learning_rate": 3.017350442797759e-06, "loss": 0.7855, "step": 9953 }, { "epoch": 1.8, "learning_rate": 3.014639436110609e-06, "loss": 0.5284, "step": 9954 }, { "epoch": 1.8, "learning_rate": 3.0119284294234593e-06, "loss": 0.6052, "step": 9955 }, { "epoch": 1.8, "learning_rate": 3.0092174227363095e-06, "loss": 1.1309, "step": 9956 }, { "epoch": 1.8, "learning_rate": 3.00650641604916e-06, "loss": 0.5737, "step": 9957 }, { "epoch": 1.8, "learning_rate": 3.00379540936201e-06, "loss": 0.6817, "step": 9958 }, { "epoch": 1.8, "learning_rate": 3.0010844026748603e-06, "loss": 0.6194, "step": 9959 }, { "epoch": 1.8, "learning_rate": 2.9983733959877104e-06, "loss": 0.381, "step": 9960 }, { "epoch": 1.8, "learning_rate": 2.9956623893005605e-06, "loss": 1.6147, "step": 9961 }, { "epoch": 1.8, "learning_rate": 2.9929513826134107e-06, "loss": 0.6841, "step": 9962 }, { "epoch": 1.8, "learning_rate": 2.990240375926261e-06, "loss": 0.6506, "step": 9963 }, { "epoch": 1.8, "learning_rate": 2.987529369239111e-06, "loss": 0.5511, "step": 9964 }, { "epoch": 1.8, "learning_rate": 2.984818362551961e-06, "loss": 0.6983, "step": 9965 }, { "epoch": 1.8, "learning_rate": 2.9821073558648112e-06, "loss": 0.9454, "step": 9966 }, { "epoch": 1.8, "learning_rate": 2.9793963491776618e-06, "loss": 0.6412, "step": 9967 }, { "epoch": 1.8, "learning_rate": 2.976685342490512e-06, "loss": 0.6684, "step": 9968 }, { "epoch": 1.8, "learning_rate": 2.973974335803362e-06, "loss": 0.7774, "step": 9969 }, { "epoch": 1.8, "learning_rate": 2.971263329116212e-06, "loss": 0.3701, "step": 9970 }, { "epoch": 1.8, "learning_rate": 2.9685523224290623e-06, "loss": 0.5967, "step": 9971 }, { "epoch": 1.8, "learning_rate": 2.9658413157419124e-06, "loss": 0.7521, "step": 9972 }, { "epoch": 1.8, "learning_rate": 2.963130309054762e-06, "loss": 0.3783, "step": 9973 }, { "epoch": 1.8, "learning_rate": 2.9604193023676123e-06, "loss": 0.7262, "step": 9974 }, { "epoch": 1.8, "learning_rate": 2.9577082956804624e-06, "loss": 0.3602, "step": 9975 }, { "epoch": 1.8, "learning_rate": 2.9549972889933126e-06, "loss": 1.8659, "step": 9976 }, { "epoch": 1.8, "learning_rate": 2.9522862823061627e-06, "loss": 0.7844, "step": 9977 }, { "epoch": 1.8, "learning_rate": 2.9495752756190133e-06, "loss": 0.8235, "step": 9978 }, { "epoch": 1.8, "learning_rate": 2.9468642689318634e-06, "loss": 0.5405, "step": 9979 }, { "epoch": 1.8, "learning_rate": 2.9441532622447135e-06, "loss": 0.6533, "step": 9980 }, { "epoch": 1.8, "learning_rate": 2.9414422555575637e-06, "loss": 0.7635, "step": 9981 }, { "epoch": 1.8, "learning_rate": 2.938731248870414e-06, "loss": 0.9399, "step": 9982 }, { "epoch": 1.8, "learning_rate": 2.936020242183264e-06, "loss": 0.7491, "step": 9983 }, { "epoch": 1.8, "learning_rate": 2.933309235496114e-06, "loss": 0.8449, "step": 9984 }, { "epoch": 1.8, "learning_rate": 2.930598228808964e-06, "loss": 0.6107, "step": 9985 }, { "epoch": 1.8, "learning_rate": 2.9278872221218143e-06, "loss": 0.9974, "step": 9986 }, { "epoch": 1.8, "learning_rate": 2.9251762154346645e-06, "loss": 1.059, "step": 9987 }, { "epoch": 1.81, "learning_rate": 2.922465208747515e-06, "loss": 0.3379, "step": 9988 }, { "epoch": 1.81, "learning_rate": 2.919754202060365e-06, "loss": 1.103, "step": 9989 }, { "epoch": 1.81, "learning_rate": 2.9170431953732153e-06, "loss": 0.4525, "step": 9990 }, { "epoch": 1.81, "learning_rate": 2.9143321886860654e-06, "loss": 0.5262, "step": 9991 }, { "epoch": 1.81, "learning_rate": 2.9116211819989156e-06, "loss": 1.4366, "step": 9992 }, { "epoch": 1.81, "learning_rate": 2.9089101753117657e-06, "loss": 0.4769, "step": 9993 }, { "epoch": 1.81, "learning_rate": 2.906199168624616e-06, "loss": 0.9024, "step": 9994 }, { "epoch": 1.81, "learning_rate": 2.903488161937466e-06, "loss": 0.3784, "step": 9995 }, { "epoch": 1.81, "learning_rate": 2.900777155250316e-06, "loss": 0.9575, "step": 9996 }, { "epoch": 1.81, "learning_rate": 2.8980661485631662e-06, "loss": 0.5496, "step": 9997 }, { "epoch": 1.81, "learning_rate": 2.8953551418760168e-06, "loss": 0.7908, "step": 9998 }, { "epoch": 1.81, "learning_rate": 2.892644135188867e-06, "loss": 0.6308, "step": 9999 }, { "epoch": 1.81, "learning_rate": 2.889933128501717e-06, "loss": 0.5062, "step": 10000 }, { "epoch": 1.81, "learning_rate": 2.887222121814567e-06, "loss": 0.755, "step": 10001 }, { "epoch": 1.81, "learning_rate": 2.8845111151274173e-06, "loss": 0.8384, "step": 10002 }, { "epoch": 1.81, "learning_rate": 2.8818001084402675e-06, "loss": 0.5119, "step": 10003 }, { "epoch": 1.81, "learning_rate": 2.8790891017531176e-06, "loss": 0.4633, "step": 10004 }, { "epoch": 1.81, "learning_rate": 2.8763780950659677e-06, "loss": 0.6086, "step": 10005 }, { "epoch": 1.81, "learning_rate": 2.873667088378818e-06, "loss": 0.5182, "step": 10006 }, { "epoch": 1.81, "learning_rate": 2.870956081691668e-06, "loss": 1.4092, "step": 10007 }, { "epoch": 1.81, "learning_rate": 2.868245075004518e-06, "loss": 0.773, "step": 10008 }, { "epoch": 1.81, "learning_rate": 2.8655340683173687e-06, "loss": 0.6108, "step": 10009 }, { "epoch": 1.81, "learning_rate": 2.862823061630219e-06, "loss": 0.5841, "step": 10010 }, { "epoch": 1.81, "learning_rate": 2.860112054943069e-06, "loss": 0.5781, "step": 10011 }, { "epoch": 1.81, "learning_rate": 2.857401048255919e-06, "loss": 0.9404, "step": 10012 }, { "epoch": 1.81, "learning_rate": 2.8546900415687692e-06, "loss": 0.614, "step": 10013 }, { "epoch": 1.81, "learning_rate": 2.8519790348816194e-06, "loss": 0.6607, "step": 10014 }, { "epoch": 1.81, "learning_rate": 2.8492680281944695e-06, "loss": 0.5134, "step": 10015 }, { "epoch": 1.81, "learning_rate": 2.8465570215073196e-06, "loss": 0.7918, "step": 10016 }, { "epoch": 1.81, "learning_rate": 2.8438460148201698e-06, "loss": 0.4677, "step": 10017 }, { "epoch": 1.81, "learning_rate": 2.84113500813302e-06, "loss": 0.4002, "step": 10018 }, { "epoch": 1.81, "learning_rate": 2.8384240014458705e-06, "loss": 0.3726, "step": 10019 }, { "epoch": 1.81, "learning_rate": 2.8357129947587206e-06, "loss": 0.3862, "step": 10020 }, { "epoch": 1.81, "learning_rate": 2.8330019880715707e-06, "loss": 0.4107, "step": 10021 }, { "epoch": 1.81, "learning_rate": 2.830290981384421e-06, "loss": 0.513, "step": 10022 }, { "epoch": 1.81, "learning_rate": 2.827579974697271e-06, "loss": 0.686, "step": 10023 }, { "epoch": 1.81, "learning_rate": 2.824868968010121e-06, "loss": 0.4302, "step": 10024 }, { "epoch": 1.81, "learning_rate": 2.8221579613229713e-06, "loss": 0.9332, "step": 10025 }, { "epoch": 1.81, "learning_rate": 2.8194469546358214e-06, "loss": 0.7861, "step": 10026 }, { "epoch": 1.81, "learning_rate": 2.8167359479486715e-06, "loss": 0.6194, "step": 10027 }, { "epoch": 1.81, "learning_rate": 2.8140249412615217e-06, "loss": 0.6695, "step": 10028 }, { "epoch": 1.81, "learning_rate": 2.811313934574372e-06, "loss": 0.3674, "step": 10029 }, { "epoch": 1.81, "learning_rate": 2.8086029278872224e-06, "loss": 0.5039, "step": 10030 }, { "epoch": 1.81, "learning_rate": 2.8058919212000725e-06, "loss": 0.4844, "step": 10031 }, { "epoch": 1.81, "learning_rate": 2.8031809145129226e-06, "loss": 0.9962, "step": 10032 }, { "epoch": 1.81, "learning_rate": 2.8004699078257728e-06, "loss": 0.8146, "step": 10033 }, { "epoch": 1.81, "learning_rate": 2.797758901138623e-06, "loss": 0.48, "step": 10034 }, { "epoch": 1.81, "learning_rate": 2.795047894451473e-06, "loss": 1.0203, "step": 10035 }, { "epoch": 1.81, "learning_rate": 2.792336887764323e-06, "loss": 0.6507, "step": 10036 }, { "epoch": 1.81, "learning_rate": 2.7896258810771733e-06, "loss": 0.6475, "step": 10037 }, { "epoch": 1.81, "learning_rate": 2.7869148743900234e-06, "loss": 0.3103, "step": 10038 }, { "epoch": 1.81, "learning_rate": 2.7842038677028736e-06, "loss": 1.0561, "step": 10039 }, { "epoch": 1.81, "learning_rate": 2.781492861015724e-06, "loss": 0.3481, "step": 10040 }, { "epoch": 1.81, "learning_rate": 2.7787818543285743e-06, "loss": 0.6997, "step": 10041 }, { "epoch": 1.81, "learning_rate": 2.7760708476414244e-06, "loss": 1.0446, "step": 10042 }, { "epoch": 1.82, "learning_rate": 2.7733598409542745e-06, "loss": 1.1141, "step": 10043 }, { "epoch": 1.82, "learning_rate": 2.7706488342671247e-06, "loss": 0.1958, "step": 10044 }, { "epoch": 1.82, "learning_rate": 2.767937827579975e-06, "loss": 0.6018, "step": 10045 }, { "epoch": 1.82, "learning_rate": 2.765226820892825e-06, "loss": 0.244, "step": 10046 }, { "epoch": 1.82, "learning_rate": 2.762515814205675e-06, "loss": 0.5705, "step": 10047 }, { "epoch": 1.82, "learning_rate": 2.759804807518525e-06, "loss": 0.5253, "step": 10048 }, { "epoch": 1.82, "learning_rate": 2.7570938008313753e-06, "loss": 1.0956, "step": 10049 }, { "epoch": 1.82, "learning_rate": 2.7543827941442255e-06, "loss": 0.8115, "step": 10050 }, { "epoch": 1.82, "learning_rate": 2.751671787457076e-06, "loss": 1.2263, "step": 10051 }, { "epoch": 1.82, "learning_rate": 2.748960780769926e-06, "loss": 0.5767, "step": 10052 }, { "epoch": 1.82, "learning_rate": 2.7462497740827763e-06, "loss": 0.3492, "step": 10053 }, { "epoch": 1.82, "learning_rate": 2.7435387673956264e-06, "loss": 0.4542, "step": 10054 }, { "epoch": 1.82, "learning_rate": 2.7408277607084766e-06, "loss": 0.1409, "step": 10055 }, { "epoch": 1.82, "learning_rate": 2.7381167540213267e-06, "loss": 0.2844, "step": 10056 }, { "epoch": 1.82, "learning_rate": 2.735405747334177e-06, "loss": 0.4808, "step": 10057 }, { "epoch": 1.82, "learning_rate": 2.732694740647027e-06, "loss": 0.3602, "step": 10058 }, { "epoch": 1.82, "learning_rate": 2.729983733959877e-06, "loss": 1.0641, "step": 10059 }, { "epoch": 1.82, "learning_rate": 2.7272727272727272e-06, "loss": 1.067, "step": 10060 }, { "epoch": 1.82, "learning_rate": 2.7245617205855778e-06, "loss": 1.0052, "step": 10061 }, { "epoch": 1.82, "learning_rate": 2.721850713898428e-06, "loss": 0.4045, "step": 10062 }, { "epoch": 1.82, "learning_rate": 2.719139707211278e-06, "loss": 0.6802, "step": 10063 }, { "epoch": 1.82, "learning_rate": 2.716428700524128e-06, "loss": 0.7177, "step": 10064 }, { "epoch": 1.82, "learning_rate": 2.7137176938369783e-06, "loss": 0.7057, "step": 10065 }, { "epoch": 1.82, "learning_rate": 2.7110066871498285e-06, "loss": 0.4504, "step": 10066 }, { "epoch": 1.82, "learning_rate": 2.7082956804626786e-06, "loss": 1.2565, "step": 10067 }, { "epoch": 1.82, "learning_rate": 2.7055846737755287e-06, "loss": 0.9051, "step": 10068 }, { "epoch": 1.82, "learning_rate": 2.702873667088379e-06, "loss": 0.6989, "step": 10069 }, { "epoch": 1.82, "learning_rate": 2.700162660401229e-06, "loss": 0.9877, "step": 10070 }, { "epoch": 1.82, "learning_rate": 2.6974516537140796e-06, "loss": 0.851, "step": 10071 }, { "epoch": 1.82, "learning_rate": 2.6947406470269297e-06, "loss": 0.6756, "step": 10072 }, { "epoch": 1.82, "learning_rate": 2.69202964033978e-06, "loss": 0.4935, "step": 10073 }, { "epoch": 1.82, "learning_rate": 2.68931863365263e-06, "loss": 0.5466, "step": 10074 }, { "epoch": 1.82, "learning_rate": 2.68660762696548e-06, "loss": 0.4573, "step": 10075 }, { "epoch": 1.82, "learning_rate": 2.6838966202783302e-06, "loss": 0.7219, "step": 10076 }, { "epoch": 1.82, "learning_rate": 2.6811856135911804e-06, "loss": 0.9674, "step": 10077 }, { "epoch": 1.82, "learning_rate": 2.6784746069040305e-06, "loss": 0.6887, "step": 10078 }, { "epoch": 1.82, "learning_rate": 2.6757636002168806e-06, "loss": 0.7309, "step": 10079 }, { "epoch": 1.82, "learning_rate": 2.6730525935297308e-06, "loss": 0.6239, "step": 10080 }, { "epoch": 1.82, "learning_rate": 2.670341586842581e-06, "loss": 0.6339, "step": 10081 }, { "epoch": 1.82, "learning_rate": 2.6676305801554315e-06, "loss": 0.6514, "step": 10082 }, { "epoch": 1.82, "learning_rate": 2.6649195734682816e-06, "loss": 0.9033, "step": 10083 }, { "epoch": 1.82, "learning_rate": 2.6622085667811317e-06, "loss": 0.8428, "step": 10084 }, { "epoch": 1.82, "learning_rate": 2.659497560093982e-06, "loss": 0.853, "step": 10085 }, { "epoch": 1.82, "learning_rate": 2.656786553406832e-06, "loss": 0.5933, "step": 10086 }, { "epoch": 1.82, "learning_rate": 2.654075546719682e-06, "loss": 0.4804, "step": 10087 }, { "epoch": 1.82, "learning_rate": 2.6513645400325323e-06, "loss": 1.5426, "step": 10088 }, { "epoch": 1.82, "learning_rate": 2.6486535333453824e-06, "loss": 0.4933, "step": 10089 }, { "epoch": 1.82, "learning_rate": 2.6459425266582325e-06, "loss": 0.4915, "step": 10090 }, { "epoch": 1.82, "learning_rate": 2.6432315199710827e-06, "loss": 0.9157, "step": 10091 }, { "epoch": 1.82, "learning_rate": 2.6405205132839332e-06, "loss": 0.4462, "step": 10092 }, { "epoch": 1.82, "learning_rate": 2.6378095065967834e-06, "loss": 0.5005, "step": 10093 }, { "epoch": 1.82, "learning_rate": 2.6350984999096335e-06, "loss": 0.7214, "step": 10094 }, { "epoch": 1.82, "learning_rate": 2.6323874932224836e-06, "loss": 0.428, "step": 10095 }, { "epoch": 1.82, "learning_rate": 2.6296764865353338e-06, "loss": 0.5308, "step": 10096 }, { "epoch": 1.82, "learning_rate": 2.626965479848184e-06, "loss": 0.2224, "step": 10097 }, { "epoch": 1.83, "learning_rate": 2.624254473161034e-06, "loss": 1.0586, "step": 10098 }, { "epoch": 1.83, "learning_rate": 2.621543466473884e-06, "loss": 0.4583, "step": 10099 }, { "epoch": 1.83, "learning_rate": 2.6188324597867343e-06, "loss": 0.4739, "step": 10100 }, { "epoch": 1.83, "learning_rate": 2.616121453099584e-06, "loss": 1.1552, "step": 10101 }, { "epoch": 1.83, "learning_rate": 2.613410446412434e-06, "loss": 0.7765, "step": 10102 }, { "epoch": 1.83, "learning_rate": 2.6106994397252847e-06, "loss": 0.9269, "step": 10103 }, { "epoch": 1.83, "learning_rate": 2.607988433038135e-06, "loss": 0.4049, "step": 10104 }, { "epoch": 1.83, "learning_rate": 2.605277426350985e-06, "loss": 0.6263, "step": 10105 }, { "epoch": 1.83, "learning_rate": 2.602566419663835e-06, "loss": 0.6983, "step": 10106 }, { "epoch": 1.83, "learning_rate": 2.5998554129766852e-06, "loss": 0.7949, "step": 10107 }, { "epoch": 1.83, "learning_rate": 2.5971444062895354e-06, "loss": 0.6087, "step": 10108 }, { "epoch": 1.83, "learning_rate": 2.5944333996023855e-06, "loss": 0.4151, "step": 10109 }, { "epoch": 1.83, "learning_rate": 2.5917223929152356e-06, "loss": 0.4637, "step": 10110 }, { "epoch": 1.83, "learning_rate": 2.5890113862280858e-06, "loss": 0.8294, "step": 10111 }, { "epoch": 1.83, "learning_rate": 2.586300379540936e-06, "loss": 0.5237, "step": 10112 }, { "epoch": 1.83, "learning_rate": 2.5835893728537865e-06, "loss": 0.7152, "step": 10113 }, { "epoch": 1.83, "learning_rate": 2.5808783661666366e-06, "loss": 0.2683, "step": 10114 }, { "epoch": 1.83, "learning_rate": 2.5781673594794867e-06, "loss": 0.4688, "step": 10115 }, { "epoch": 1.83, "learning_rate": 2.575456352792337e-06, "loss": 0.9487, "step": 10116 }, { "epoch": 1.83, "learning_rate": 2.572745346105187e-06, "loss": 0.7854, "step": 10117 }, { "epoch": 1.83, "learning_rate": 2.570034339418037e-06, "loss": 0.5271, "step": 10118 }, { "epoch": 1.83, "learning_rate": 2.5673233327308873e-06, "loss": 0.3126, "step": 10119 }, { "epoch": 1.83, "learning_rate": 2.5646123260437374e-06, "loss": 0.8192, "step": 10120 }, { "epoch": 1.83, "learning_rate": 2.5619013193565875e-06, "loss": 1.3328, "step": 10121 }, { "epoch": 1.83, "learning_rate": 2.5591903126694377e-06, "loss": 0.6665, "step": 10122 }, { "epoch": 1.83, "learning_rate": 2.5564793059822882e-06, "loss": 0.5691, "step": 10123 }, { "epoch": 1.83, "learning_rate": 2.5537682992951384e-06, "loss": 1.0675, "step": 10124 }, { "epoch": 1.83, "learning_rate": 2.5510572926079885e-06, "loss": 0.4912, "step": 10125 }, { "epoch": 1.83, "learning_rate": 2.5483462859208386e-06, "loss": 0.7314, "step": 10126 }, { "epoch": 1.83, "learning_rate": 2.5456352792336888e-06, "loss": 0.9095, "step": 10127 }, { "epoch": 1.83, "learning_rate": 2.542924272546539e-06, "loss": 0.5296, "step": 10128 }, { "epoch": 1.83, "learning_rate": 2.540213265859389e-06, "loss": 0.4498, "step": 10129 }, { "epoch": 1.83, "learning_rate": 2.537502259172239e-06, "loss": 0.542, "step": 10130 }, { "epoch": 1.83, "learning_rate": 2.5347912524850893e-06, "loss": 0.5898, "step": 10131 }, { "epoch": 1.83, "learning_rate": 2.5320802457979394e-06, "loss": 0.4028, "step": 10132 }, { "epoch": 1.83, "learning_rate": 2.5293692391107896e-06, "loss": 1.0224, "step": 10133 }, { "epoch": 1.83, "learning_rate": 2.52665823242364e-06, "loss": 0.884, "step": 10134 }, { "epoch": 1.83, "learning_rate": 2.5239472257364903e-06, "loss": 0.634, "step": 10135 }, { "epoch": 1.83, "learning_rate": 2.5212362190493404e-06, "loss": 1.1597, "step": 10136 }, { "epoch": 1.83, "learning_rate": 2.5185252123621905e-06, "loss": 0.52, "step": 10137 }, { "epoch": 1.83, "learning_rate": 2.5158142056750407e-06, "loss": 0.4592, "step": 10138 }, { "epoch": 1.83, "learning_rate": 2.513103198987891e-06, "loss": 0.3364, "step": 10139 }, { "epoch": 1.83, "learning_rate": 2.510392192300741e-06, "loss": 0.4622, "step": 10140 }, { "epoch": 1.83, "learning_rate": 2.507681185613591e-06, "loss": 0.3739, "step": 10141 }, { "epoch": 1.83, "learning_rate": 2.504970178926441e-06, "loss": 1.4028, "step": 10142 }, { "epoch": 1.83, "learning_rate": 2.5022591722392913e-06, "loss": 0.3801, "step": 10143 }, { "epoch": 1.83, "learning_rate": 2.499548165552142e-06, "loss": 0.6635, "step": 10144 }, { "epoch": 1.83, "learning_rate": 2.496837158864992e-06, "loss": 0.3824, "step": 10145 }, { "epoch": 1.83, "learning_rate": 2.494126152177842e-06, "loss": 0.5158, "step": 10146 }, { "epoch": 1.83, "learning_rate": 2.4914151454906923e-06, "loss": 0.2677, "step": 10147 }, { "epoch": 1.83, "learning_rate": 2.4887041388035424e-06, "loss": 1.0068, "step": 10148 }, { "epoch": 1.83, "learning_rate": 2.4859931321163926e-06, "loss": 0.583, "step": 10149 }, { "epoch": 1.83, "learning_rate": 2.4832821254292427e-06, "loss": 0.3757, "step": 10150 }, { "epoch": 1.83, "learning_rate": 2.480571118742093e-06, "loss": 0.3284, "step": 10151 }, { "epoch": 1.83, "learning_rate": 2.477860112054943e-06, "loss": 0.4737, "step": 10152 }, { "epoch": 1.83, "learning_rate": 2.475149105367793e-06, "loss": 0.536, "step": 10153 }, { "epoch": 1.84, "learning_rate": 2.4724380986806432e-06, "loss": 0.8751, "step": 10154 }, { "epoch": 1.84, "learning_rate": 2.469727091993494e-06, "loss": 0.3968, "step": 10155 }, { "epoch": 1.84, "learning_rate": 2.467016085306344e-06, "loss": 0.4115, "step": 10156 }, { "epoch": 1.84, "learning_rate": 2.464305078619194e-06, "loss": 0.4183, "step": 10157 }, { "epoch": 1.84, "learning_rate": 2.461594071932044e-06, "loss": 0.6838, "step": 10158 }, { "epoch": 1.84, "learning_rate": 2.4588830652448943e-06, "loss": 0.6603, "step": 10159 }, { "epoch": 1.84, "learning_rate": 2.4561720585577445e-06, "loss": 1.2613, "step": 10160 }, { "epoch": 1.84, "learning_rate": 2.4534610518705946e-06, "loss": 0.7936, "step": 10161 }, { "epoch": 1.84, "learning_rate": 2.4507500451834447e-06, "loss": 0.7797, "step": 10162 }, { "epoch": 1.84, "learning_rate": 2.448039038496295e-06, "loss": 0.8824, "step": 10163 }, { "epoch": 1.84, "learning_rate": 2.445328031809145e-06, "loss": 0.3982, "step": 10164 }, { "epoch": 1.84, "learning_rate": 2.4426170251219956e-06, "loss": 0.4417, "step": 10165 }, { "epoch": 1.84, "learning_rate": 2.4399060184348457e-06, "loss": 1.0556, "step": 10166 }, { "epoch": 1.84, "learning_rate": 2.437195011747696e-06, "loss": 0.6784, "step": 10167 }, { "epoch": 1.84, "learning_rate": 2.434484005060546e-06, "loss": 0.7968, "step": 10168 }, { "epoch": 1.84, "learning_rate": 2.431772998373396e-06, "loss": 0.6599, "step": 10169 }, { "epoch": 1.84, "learning_rate": 2.4290619916862462e-06, "loss": 0.6369, "step": 10170 }, { "epoch": 1.84, "learning_rate": 2.4263509849990964e-06, "loss": 0.7848, "step": 10171 }, { "epoch": 1.84, "learning_rate": 2.4236399783119465e-06, "loss": 0.3059, "step": 10172 }, { "epoch": 1.84, "learning_rate": 2.4209289716247966e-06, "loss": 0.8157, "step": 10173 }, { "epoch": 1.84, "learning_rate": 2.4182179649376468e-06, "loss": 0.4453, "step": 10174 }, { "epoch": 1.84, "learning_rate": 2.415506958250497e-06, "loss": 0.6609, "step": 10175 }, { "epoch": 1.84, "learning_rate": 2.4127959515633475e-06, "loss": 1.077, "step": 10176 }, { "epoch": 1.84, "learning_rate": 2.4100849448761976e-06, "loss": 1.32, "step": 10177 }, { "epoch": 1.84, "learning_rate": 2.4073739381890477e-06, "loss": 0.5393, "step": 10178 }, { "epoch": 1.84, "learning_rate": 2.404662931501898e-06, "loss": 0.3613, "step": 10179 }, { "epoch": 1.84, "learning_rate": 2.401951924814748e-06, "loss": 0.5274, "step": 10180 }, { "epoch": 1.84, "learning_rate": 2.399240918127598e-06, "loss": 0.4968, "step": 10181 }, { "epoch": 1.84, "learning_rate": 2.3965299114404483e-06, "loss": 0.8655, "step": 10182 }, { "epoch": 1.84, "learning_rate": 2.3938189047532984e-06, "loss": 0.5843, "step": 10183 }, { "epoch": 1.84, "learning_rate": 2.3911078980661485e-06, "loss": 1.7945, "step": 10184 }, { "epoch": 1.84, "learning_rate": 2.3883968913789987e-06, "loss": 0.5788, "step": 10185 }, { "epoch": 1.84, "learning_rate": 2.3856858846918492e-06, "loss": 1.0592, "step": 10186 }, { "epoch": 1.84, "learning_rate": 2.3829748780046994e-06, "loss": 0.8104, "step": 10187 }, { "epoch": 1.84, "learning_rate": 2.3802638713175495e-06, "loss": 0.4367, "step": 10188 }, { "epoch": 1.84, "learning_rate": 2.3775528646303996e-06, "loss": 0.5262, "step": 10189 }, { "epoch": 1.84, "learning_rate": 2.3748418579432498e-06, "loss": 0.6155, "step": 10190 }, { "epoch": 1.84, "learning_rate": 2.3721308512561e-06, "loss": 0.4759, "step": 10191 }, { "epoch": 1.84, "learning_rate": 2.36941984456895e-06, "loss": 0.8325, "step": 10192 }, { "epoch": 1.84, "learning_rate": 2.3667088378818e-06, "loss": 0.8629, "step": 10193 }, { "epoch": 1.84, "learning_rate": 2.3639978311946503e-06, "loss": 0.2424, "step": 10194 }, { "epoch": 1.84, "learning_rate": 2.3612868245075004e-06, "loss": 0.3772, "step": 10195 }, { "epoch": 1.84, "learning_rate": 2.358575817820351e-06, "loss": 1.3057, "step": 10196 }, { "epoch": 1.84, "learning_rate": 2.355864811133201e-06, "loss": 0.6996, "step": 10197 }, { "epoch": 1.84, "learning_rate": 2.3531538044460513e-06, "loss": 0.4544, "step": 10198 }, { "epoch": 1.84, "learning_rate": 2.3504427977589014e-06, "loss": 0.7745, "step": 10199 }, { "epoch": 1.84, "learning_rate": 2.3477317910717515e-06, "loss": 0.9951, "step": 10200 }, { "epoch": 1.84, "learning_rate": 2.3450207843846017e-06, "loss": 0.4146, "step": 10201 }, { "epoch": 1.84, "learning_rate": 2.342309777697452e-06, "loss": 1.0268, "step": 10202 }, { "epoch": 1.84, "learning_rate": 2.339598771010302e-06, "loss": 0.3592, "step": 10203 }, { "epoch": 1.84, "learning_rate": 2.336887764323152e-06, "loss": 0.8202, "step": 10204 }, { "epoch": 1.84, "learning_rate": 2.334176757636002e-06, "loss": 0.5084, "step": 10205 }, { "epoch": 1.84, "learning_rate": 2.3314657509488523e-06, "loss": 0.9406, "step": 10206 }, { "epoch": 1.84, "learning_rate": 2.328754744261703e-06, "loss": 0.4275, "step": 10207 }, { "epoch": 1.84, "learning_rate": 2.326043737574553e-06, "loss": 1.1973, "step": 10208 }, { "epoch": 1.85, "learning_rate": 2.323332730887403e-06, "loss": 0.6664, "step": 10209 }, { "epoch": 1.85, "learning_rate": 2.3206217242002533e-06, "loss": 0.3191, "step": 10210 }, { "epoch": 1.85, "learning_rate": 2.3179107175131034e-06, "loss": 0.6007, "step": 10211 }, { "epoch": 1.85, "learning_rate": 2.3151997108259536e-06, "loss": 0.5316, "step": 10212 }, { "epoch": 1.85, "learning_rate": 2.3124887041388037e-06, "loss": 0.5509, "step": 10213 }, { "epoch": 1.85, "learning_rate": 2.309777697451654e-06, "loss": 0.7059, "step": 10214 }, { "epoch": 1.85, "learning_rate": 2.307066690764504e-06, "loss": 0.8001, "step": 10215 }, { "epoch": 1.85, "learning_rate": 2.304355684077354e-06, "loss": 0.3329, "step": 10216 }, { "epoch": 1.85, "learning_rate": 2.3016446773902047e-06, "loss": 0.784, "step": 10217 }, { "epoch": 1.85, "learning_rate": 2.298933670703055e-06, "loss": 0.6883, "step": 10218 }, { "epoch": 1.85, "learning_rate": 2.296222664015905e-06, "loss": 0.5989, "step": 10219 }, { "epoch": 1.85, "learning_rate": 2.293511657328755e-06, "loss": 0.5002, "step": 10220 }, { "epoch": 1.85, "learning_rate": 2.290800650641605e-06, "loss": 1.1605, "step": 10221 }, { "epoch": 1.85, "learning_rate": 2.2880896439544553e-06, "loss": 0.5619, "step": 10222 }, { "epoch": 1.85, "learning_rate": 2.2853786372673055e-06, "loss": 0.4224, "step": 10223 }, { "epoch": 1.85, "learning_rate": 2.2826676305801556e-06, "loss": 0.731, "step": 10224 }, { "epoch": 1.85, "learning_rate": 2.2799566238930057e-06, "loss": 0.5458, "step": 10225 }, { "epoch": 1.85, "learning_rate": 2.277245617205856e-06, "loss": 0.7749, "step": 10226 }, { "epoch": 1.85, "learning_rate": 2.274534610518706e-06, "loss": 0.9102, "step": 10227 }, { "epoch": 1.85, "learning_rate": 2.2718236038315566e-06, "loss": 0.7312, "step": 10228 }, { "epoch": 1.85, "learning_rate": 2.2691125971444067e-06, "loss": 1.3478, "step": 10229 }, { "epoch": 1.85, "learning_rate": 2.2664015904572564e-06, "loss": 0.5408, "step": 10230 }, { "epoch": 1.85, "learning_rate": 2.2636905837701065e-06, "loss": 0.8135, "step": 10231 }, { "epoch": 1.85, "learning_rate": 2.2609795770829567e-06, "loss": 0.4361, "step": 10232 }, { "epoch": 1.85, "learning_rate": 2.258268570395807e-06, "loss": 0.49, "step": 10233 }, { "epoch": 1.85, "learning_rate": 2.255557563708657e-06, "loss": 1.2824, "step": 10234 }, { "epoch": 1.85, "learning_rate": 2.252846557021507e-06, "loss": 0.4797, "step": 10235 }, { "epoch": 1.85, "learning_rate": 2.250135550334357e-06, "loss": 0.9736, "step": 10236 }, { "epoch": 1.85, "learning_rate": 2.2474245436472073e-06, "loss": 0.4713, "step": 10237 }, { "epoch": 1.85, "learning_rate": 2.244713536960058e-06, "loss": 1.0805, "step": 10238 }, { "epoch": 1.85, "learning_rate": 2.242002530272908e-06, "loss": 1.7266, "step": 10239 }, { "epoch": 1.85, "learning_rate": 2.239291523585758e-06, "loss": 0.2307, "step": 10240 }, { "epoch": 1.85, "learning_rate": 2.2365805168986083e-06, "loss": 0.3809, "step": 10241 }, { "epoch": 1.85, "learning_rate": 2.2338695102114584e-06, "loss": 0.4033, "step": 10242 }, { "epoch": 1.85, "learning_rate": 2.2311585035243086e-06, "loss": 0.4056, "step": 10243 }, { "epoch": 1.85, "learning_rate": 2.2284474968371587e-06, "loss": 1.0608, "step": 10244 }, { "epoch": 1.85, "learning_rate": 2.225736490150009e-06, "loss": 0.494, "step": 10245 }, { "epoch": 1.85, "learning_rate": 2.223025483462859e-06, "loss": 1.5838, "step": 10246 }, { "epoch": 1.85, "learning_rate": 2.220314476775709e-06, "loss": 0.7595, "step": 10247 }, { "epoch": 1.85, "learning_rate": 2.2176034700885597e-06, "loss": 0.4973, "step": 10248 }, { "epoch": 1.85, "learning_rate": 2.21489246340141e-06, "loss": 0.9759, "step": 10249 }, { "epoch": 1.85, "learning_rate": 2.21218145671426e-06, "loss": 0.6064, "step": 10250 }, { "epoch": 1.85, "learning_rate": 2.20947045002711e-06, "loss": 0.528, "step": 10251 }, { "epoch": 1.85, "learning_rate": 2.20675944333996e-06, "loss": 0.836, "step": 10252 }, { "epoch": 1.85, "learning_rate": 2.2040484366528103e-06, "loss": 0.8167, "step": 10253 }, { "epoch": 1.85, "learning_rate": 2.2013374299656605e-06, "loss": 0.5798, "step": 10254 }, { "epoch": 1.85, "learning_rate": 2.1986264232785106e-06, "loss": 0.8744, "step": 10255 }, { "epoch": 1.85, "learning_rate": 2.1959154165913607e-06, "loss": 0.778, "step": 10256 }, { "epoch": 1.85, "learning_rate": 2.193204409904211e-06, "loss": 0.7563, "step": 10257 }, { "epoch": 1.85, "learning_rate": 2.190493403217061e-06, "loss": 0.9175, "step": 10258 }, { "epoch": 1.85, "learning_rate": 2.1877823965299116e-06, "loss": 1.2403, "step": 10259 }, { "epoch": 1.85, "learning_rate": 2.1850713898427617e-06, "loss": 0.867, "step": 10260 }, { "epoch": 1.85, "learning_rate": 2.182360383155612e-06, "loss": 0.9073, "step": 10261 }, { "epoch": 1.85, "learning_rate": 2.179649376468462e-06, "loss": 0.7223, "step": 10262 }, { "epoch": 1.85, "learning_rate": 2.176938369781312e-06, "loss": 0.958, "step": 10263 }, { "epoch": 1.86, "learning_rate": 2.1742273630941622e-06, "loss": 0.4673, "step": 10264 }, { "epoch": 1.86, "learning_rate": 2.1715163564070124e-06, "loss": 0.6479, "step": 10265 }, { "epoch": 1.86, "learning_rate": 2.1688053497198625e-06, "loss": 1.4853, "step": 10266 }, { "epoch": 1.86, "learning_rate": 2.1660943430327126e-06, "loss": 0.8926, "step": 10267 }, { "epoch": 1.86, "learning_rate": 2.1633833363455628e-06, "loss": 0.9025, "step": 10268 }, { "epoch": 1.86, "learning_rate": 2.1606723296584133e-06, "loss": 0.7326, "step": 10269 }, { "epoch": 1.86, "learning_rate": 2.1579613229712635e-06, "loss": 0.6986, "step": 10270 }, { "epoch": 1.86, "learning_rate": 2.1552503162841136e-06, "loss": 0.4974, "step": 10271 }, { "epoch": 1.86, "learning_rate": 2.1525393095969637e-06, "loss": 0.7623, "step": 10272 }, { "epoch": 1.86, "learning_rate": 2.149828302909814e-06, "loss": 0.4255, "step": 10273 }, { "epoch": 1.86, "learning_rate": 2.147117296222664e-06, "loss": 0.5093, "step": 10274 }, { "epoch": 1.86, "learning_rate": 2.144406289535514e-06, "loss": 0.7983, "step": 10275 }, { "epoch": 1.86, "learning_rate": 2.1416952828483643e-06, "loss": 0.4706, "step": 10276 }, { "epoch": 1.86, "learning_rate": 2.1389842761612144e-06, "loss": 1.1749, "step": 10277 }, { "epoch": 1.86, "learning_rate": 2.1362732694740645e-06, "loss": 0.4182, "step": 10278 }, { "epoch": 1.86, "learning_rate": 2.1335622627869147e-06, "loss": 0.9088, "step": 10279 }, { "epoch": 1.86, "learning_rate": 2.1308512560997652e-06, "loss": 0.5313, "step": 10280 }, { "epoch": 1.86, "learning_rate": 2.1281402494126154e-06, "loss": 1.0063, "step": 10281 }, { "epoch": 1.86, "learning_rate": 2.1254292427254655e-06, "loss": 0.7883, "step": 10282 }, { "epoch": 1.86, "learning_rate": 2.1227182360383156e-06, "loss": 0.5137, "step": 10283 }, { "epoch": 1.86, "learning_rate": 2.1200072293511658e-06, "loss": 0.6751, "step": 10284 }, { "epoch": 1.86, "learning_rate": 2.117296222664016e-06, "loss": 0.719, "step": 10285 }, { "epoch": 1.86, "learning_rate": 2.114585215976866e-06, "loss": 1.3688, "step": 10286 }, { "epoch": 1.86, "learning_rate": 2.111874209289716e-06, "loss": 0.5674, "step": 10287 }, { "epoch": 1.86, "learning_rate": 2.1091632026025663e-06, "loss": 1.1287, "step": 10288 }, { "epoch": 1.86, "learning_rate": 2.1064521959154164e-06, "loss": 0.1908, "step": 10289 }, { "epoch": 1.86, "learning_rate": 2.103741189228267e-06, "loss": 0.5055, "step": 10290 }, { "epoch": 1.86, "learning_rate": 2.101030182541117e-06, "loss": 0.5807, "step": 10291 }, { "epoch": 1.86, "learning_rate": 2.0983191758539673e-06, "loss": 0.8923, "step": 10292 }, { "epoch": 1.86, "learning_rate": 2.0956081691668174e-06, "loss": 0.9181, "step": 10293 }, { "epoch": 1.86, "learning_rate": 2.0928971624796675e-06, "loss": 0.9167, "step": 10294 }, { "epoch": 1.86, "learning_rate": 2.0901861557925177e-06, "loss": 0.3454, "step": 10295 }, { "epoch": 1.86, "learning_rate": 2.087475149105368e-06, "loss": 1.0884, "step": 10296 }, { "epoch": 1.86, "learning_rate": 2.084764142418218e-06, "loss": 0.8546, "step": 10297 }, { "epoch": 1.86, "learning_rate": 2.082053135731068e-06, "loss": 0.6183, "step": 10298 }, { "epoch": 1.86, "learning_rate": 2.0793421290439182e-06, "loss": 0.3466, "step": 10299 }, { "epoch": 1.86, "learning_rate": 2.0766311223567683e-06, "loss": 0.2458, "step": 10300 }, { "epoch": 1.86, "learning_rate": 2.073920115669619e-06, "loss": 1.3436, "step": 10301 }, { "epoch": 1.86, "learning_rate": 2.071209108982469e-06, "loss": 0.2736, "step": 10302 }, { "epoch": 1.86, "learning_rate": 2.068498102295319e-06, "loss": 1.3985, "step": 10303 }, { "epoch": 1.86, "learning_rate": 2.0657870956081693e-06, "loss": 0.9754, "step": 10304 }, { "epoch": 1.86, "learning_rate": 2.0630760889210194e-06, "loss": 0.8038, "step": 10305 }, { "epoch": 1.86, "learning_rate": 2.0603650822338696e-06, "loss": 1.1964, "step": 10306 }, { "epoch": 1.86, "learning_rate": 2.0576540755467197e-06, "loss": 0.8602, "step": 10307 }, { "epoch": 1.86, "learning_rate": 2.05494306885957e-06, "loss": 0.8117, "step": 10308 }, { "epoch": 1.86, "learning_rate": 2.05223206217242e-06, "loss": 0.6771, "step": 10309 }, { "epoch": 1.86, "learning_rate": 2.04952105548527e-06, "loss": 0.6393, "step": 10310 }, { "epoch": 1.86, "learning_rate": 2.0468100487981207e-06, "loss": 1.1508, "step": 10311 }, { "epoch": 1.86, "learning_rate": 2.044099042110971e-06, "loss": 1.0455, "step": 10312 }, { "epoch": 1.86, "learning_rate": 2.041388035423821e-06, "loss": 0.8888, "step": 10313 }, { "epoch": 1.86, "learning_rate": 2.038677028736671e-06, "loss": 0.6354, "step": 10314 }, { "epoch": 1.86, "learning_rate": 2.035966022049521e-06, "loss": 0.8028, "step": 10315 }, { "epoch": 1.86, "learning_rate": 2.0332550153623713e-06, "loss": 0.5321, "step": 10316 }, { "epoch": 1.86, "learning_rate": 2.0305440086752215e-06, "loss": 0.2143, "step": 10317 }, { "epoch": 1.86, "learning_rate": 2.0278330019880716e-06, "loss": 1.0121, "step": 10318 }, { "epoch": 1.86, "learning_rate": 2.0251219953009217e-06, "loss": 0.8453, "step": 10319 }, { "epoch": 1.87, "learning_rate": 2.022410988613772e-06, "loss": 0.5743, "step": 10320 }, { "epoch": 1.87, "learning_rate": 2.019699981926622e-06, "loss": 0.6393, "step": 10321 }, { "epoch": 1.87, "learning_rate": 2.0169889752394726e-06, "loss": 1.0955, "step": 10322 }, { "epoch": 1.87, "learning_rate": 2.0142779685523227e-06, "loss": 0.7443, "step": 10323 }, { "epoch": 1.87, "learning_rate": 2.011566961865173e-06, "loss": 1.1907, "step": 10324 }, { "epoch": 1.87, "learning_rate": 2.008855955178023e-06, "loss": 0.5677, "step": 10325 }, { "epoch": 1.87, "learning_rate": 2.006144948490873e-06, "loss": 0.814, "step": 10326 }, { "epoch": 1.87, "learning_rate": 2.0034339418037232e-06, "loss": 0.6515, "step": 10327 }, { "epoch": 1.87, "learning_rate": 2.0007229351165734e-06, "loss": 0.2841, "step": 10328 }, { "epoch": 1.87, "learning_rate": 1.9980119284294235e-06, "loss": 0.3454, "step": 10329 }, { "epoch": 1.87, "learning_rate": 1.9953009217422736e-06, "loss": 0.9424, "step": 10330 }, { "epoch": 1.87, "learning_rate": 1.9925899150551238e-06, "loss": 0.8536, "step": 10331 }, { "epoch": 1.87, "learning_rate": 1.9898789083679743e-06, "loss": 0.519, "step": 10332 }, { "epoch": 1.87, "learning_rate": 1.9871679016808245e-06, "loss": 0.4993, "step": 10333 }, { "epoch": 1.87, "learning_rate": 1.9844568949936746e-06, "loss": 1.0661, "step": 10334 }, { "epoch": 1.87, "learning_rate": 1.9817458883065247e-06, "loss": 0.5924, "step": 10335 }, { "epoch": 1.87, "learning_rate": 1.979034881619375e-06, "loss": 0.8777, "step": 10336 }, { "epoch": 1.87, "learning_rate": 1.976323874932225e-06, "loss": 1.345, "step": 10337 }, { "epoch": 1.87, "learning_rate": 1.973612868245075e-06, "loss": 1.0124, "step": 10338 }, { "epoch": 1.87, "learning_rate": 1.9709018615579253e-06, "loss": 0.5424, "step": 10339 }, { "epoch": 1.87, "learning_rate": 1.9681908548707754e-06, "loss": 0.3028, "step": 10340 }, { "epoch": 1.87, "learning_rate": 1.9654798481836255e-06, "loss": 0.8006, "step": 10341 }, { "epoch": 1.87, "learning_rate": 1.962768841496476e-06, "loss": 0.4343, "step": 10342 }, { "epoch": 1.87, "learning_rate": 1.9600578348093262e-06, "loss": 0.4887, "step": 10343 }, { "epoch": 1.87, "learning_rate": 1.9573468281221764e-06, "loss": 0.6385, "step": 10344 }, { "epoch": 1.87, "learning_rate": 1.9546358214350265e-06, "loss": 0.382, "step": 10345 }, { "epoch": 1.87, "learning_rate": 1.9519248147478766e-06, "loss": 0.5754, "step": 10346 }, { "epoch": 1.87, "learning_rate": 1.9492138080607268e-06, "loss": 0.841, "step": 10347 }, { "epoch": 1.87, "learning_rate": 1.946502801373577e-06, "loss": 0.3559, "step": 10348 }, { "epoch": 1.87, "learning_rate": 1.943791794686427e-06, "loss": 0.5123, "step": 10349 }, { "epoch": 1.87, "learning_rate": 1.941080787999277e-06, "loss": 1.1674, "step": 10350 }, { "epoch": 1.87, "learning_rate": 1.9383697813121273e-06, "loss": 1.1815, "step": 10351 }, { "epoch": 1.87, "learning_rate": 1.9356587746249774e-06, "loss": 0.5617, "step": 10352 }, { "epoch": 1.87, "learning_rate": 1.932947767937828e-06, "loss": 1.193, "step": 10353 }, { "epoch": 1.87, "learning_rate": 1.930236761250678e-06, "loss": 0.7035, "step": 10354 }, { "epoch": 1.87, "learning_rate": 1.9275257545635283e-06, "loss": 0.6803, "step": 10355 }, { "epoch": 1.87, "learning_rate": 1.9248147478763784e-06, "loss": 0.836, "step": 10356 }, { "epoch": 1.87, "learning_rate": 1.9221037411892285e-06, "loss": 0.4997, "step": 10357 }, { "epoch": 1.87, "learning_rate": 1.9193927345020787e-06, "loss": 0.3195, "step": 10358 }, { "epoch": 1.87, "learning_rate": 1.9166817278149284e-06, "loss": 0.6498, "step": 10359 }, { "epoch": 1.87, "learning_rate": 1.9139707211277785e-06, "loss": 0.6987, "step": 10360 }, { "epoch": 1.87, "learning_rate": 1.9112597144406287e-06, "loss": 0.4904, "step": 10361 }, { "epoch": 1.87, "learning_rate": 1.908548707753479e-06, "loss": 0.8545, "step": 10362 }, { "epoch": 1.87, "learning_rate": 1.9058377010663291e-06, "loss": 0.571, "step": 10363 }, { "epoch": 1.87, "learning_rate": 1.9031266943791793e-06, "loss": 1.5356, "step": 10364 }, { "epoch": 1.87, "learning_rate": 1.9004156876920294e-06, "loss": 0.5474, "step": 10365 }, { "epoch": 1.87, "learning_rate": 1.8977046810048797e-06, "loss": 0.574, "step": 10366 }, { "epoch": 1.87, "learning_rate": 1.8949936743177299e-06, "loss": 0.2306, "step": 10367 }, { "epoch": 1.87, "learning_rate": 1.89228266763058e-06, "loss": 0.5692, "step": 10368 }, { "epoch": 1.87, "learning_rate": 1.8895716609434302e-06, "loss": 0.9575, "step": 10369 }, { "epoch": 1.87, "learning_rate": 1.8868606542562803e-06, "loss": 0.95, "step": 10370 }, { "epoch": 1.87, "learning_rate": 1.8841496475691306e-06, "loss": 0.3809, "step": 10371 }, { "epoch": 1.87, "learning_rate": 1.8814386408819808e-06, "loss": 0.4244, "step": 10372 }, { "epoch": 1.87, "learning_rate": 1.878727634194831e-06, "loss": 0.9214, "step": 10373 }, { "epoch": 1.87, "learning_rate": 1.876016627507681e-06, "loss": 0.9124, "step": 10374 }, { "epoch": 1.88, "learning_rate": 1.8733056208205314e-06, "loss": 0.5002, "step": 10375 }, { "epoch": 1.88, "learning_rate": 1.8705946141333817e-06, "loss": 0.5621, "step": 10376 }, { "epoch": 1.88, "learning_rate": 1.8678836074462319e-06, "loss": 0.4531, "step": 10377 }, { "epoch": 1.88, "learning_rate": 1.865172600759082e-06, "loss": 0.9359, "step": 10378 }, { "epoch": 1.88, "learning_rate": 1.8624615940719321e-06, "loss": 0.835, "step": 10379 }, { "epoch": 1.88, "learning_rate": 1.8597505873847823e-06, "loss": 0.7674, "step": 10380 }, { "epoch": 1.88, "learning_rate": 1.8570395806976326e-06, "loss": 0.7059, "step": 10381 }, { "epoch": 1.88, "learning_rate": 1.8543285740104827e-06, "loss": 0.7693, "step": 10382 }, { "epoch": 1.88, "learning_rate": 1.8516175673233329e-06, "loss": 1.2803, "step": 10383 }, { "epoch": 1.88, "learning_rate": 1.848906560636183e-06, "loss": 0.6517, "step": 10384 }, { "epoch": 1.88, "learning_rate": 1.8461955539490331e-06, "loss": 0.8029, "step": 10385 }, { "epoch": 1.88, "learning_rate": 1.8434845472618833e-06, "loss": 0.8352, "step": 10386 }, { "epoch": 1.88, "learning_rate": 1.8407735405747336e-06, "loss": 0.4797, "step": 10387 }, { "epoch": 1.88, "learning_rate": 1.8380625338875838e-06, "loss": 0.6228, "step": 10388 }, { "epoch": 1.88, "learning_rate": 1.835351527200434e-06, "loss": 0.8097, "step": 10389 }, { "epoch": 1.88, "learning_rate": 1.832640520513284e-06, "loss": 1.0476, "step": 10390 }, { "epoch": 1.88, "learning_rate": 1.829929513826134e-06, "loss": 1.2135, "step": 10391 }, { "epoch": 1.88, "learning_rate": 1.8272185071389843e-06, "loss": 0.3708, "step": 10392 }, { "epoch": 1.88, "learning_rate": 1.8245075004518344e-06, "loss": 1.1144, "step": 10393 }, { "epoch": 1.88, "learning_rate": 1.8217964937646846e-06, "loss": 0.5043, "step": 10394 }, { "epoch": 1.88, "learning_rate": 1.8190854870775347e-06, "loss": 1.399, "step": 10395 }, { "epoch": 1.88, "learning_rate": 1.8163744803903848e-06, "loss": 1.2086, "step": 10396 }, { "epoch": 1.88, "learning_rate": 1.8136634737032352e-06, "loss": 0.5265, "step": 10397 }, { "epoch": 1.88, "learning_rate": 1.8109524670160853e-06, "loss": 0.5771, "step": 10398 }, { "epoch": 1.88, "learning_rate": 1.8082414603289355e-06, "loss": 0.8652, "step": 10399 }, { "epoch": 1.88, "learning_rate": 1.8055304536417856e-06, "loss": 0.5774, "step": 10400 }, { "epoch": 1.88, "learning_rate": 1.8028194469546357e-06, "loss": 0.51, "step": 10401 }, { "epoch": 1.88, "learning_rate": 1.800108440267486e-06, "loss": 0.9326, "step": 10402 }, { "epoch": 1.88, "learning_rate": 1.7973974335803362e-06, "loss": 0.5035, "step": 10403 }, { "epoch": 1.88, "learning_rate": 1.7946864268931863e-06, "loss": 1.045, "step": 10404 }, { "epoch": 1.88, "learning_rate": 1.7919754202060365e-06, "loss": 0.7104, "step": 10405 }, { "epoch": 1.88, "learning_rate": 1.7892644135188866e-06, "loss": 0.9099, "step": 10406 }, { "epoch": 1.88, "learning_rate": 1.786553406831737e-06, "loss": 0.742, "step": 10407 }, { "epoch": 1.88, "learning_rate": 1.783842400144587e-06, "loss": 0.7695, "step": 10408 }, { "epoch": 1.88, "learning_rate": 1.7811313934574372e-06, "loss": 0.6477, "step": 10409 }, { "epoch": 1.88, "learning_rate": 1.7784203867702874e-06, "loss": 0.377, "step": 10410 }, { "epoch": 1.88, "learning_rate": 1.7757093800831375e-06, "loss": 0.3724, "step": 10411 }, { "epoch": 1.88, "learning_rate": 1.7729983733959876e-06, "loss": 0.8532, "step": 10412 }, { "epoch": 1.88, "learning_rate": 1.770287366708838e-06, "loss": 0.773, "step": 10413 }, { "epoch": 1.88, "learning_rate": 1.767576360021688e-06, "loss": 1.1095, "step": 10414 }, { "epoch": 1.88, "learning_rate": 1.7648653533345382e-06, "loss": 0.7127, "step": 10415 }, { "epoch": 1.88, "learning_rate": 1.7621543466473884e-06, "loss": 0.2894, "step": 10416 }, { "epoch": 1.88, "learning_rate": 1.7594433399602385e-06, "loss": 0.5705, "step": 10417 }, { "epoch": 1.88, "learning_rate": 1.7567323332730888e-06, "loss": 0.4046, "step": 10418 }, { "epoch": 1.88, "learning_rate": 1.754021326585939e-06, "loss": 0.8602, "step": 10419 }, { "epoch": 1.88, "learning_rate": 1.7513103198987891e-06, "loss": 0.8996, "step": 10420 }, { "epoch": 1.88, "learning_rate": 1.7485993132116393e-06, "loss": 0.8209, "step": 10421 }, { "epoch": 1.88, "learning_rate": 1.7458883065244894e-06, "loss": 1.0621, "step": 10422 }, { "epoch": 1.88, "learning_rate": 1.7431772998373397e-06, "loss": 0.5109, "step": 10423 }, { "epoch": 1.88, "learning_rate": 1.7404662931501899e-06, "loss": 0.4577, "step": 10424 }, { "epoch": 1.88, "learning_rate": 1.73775528646304e-06, "loss": 0.7108, "step": 10425 }, { "epoch": 1.88, "learning_rate": 1.7350442797758901e-06, "loss": 0.6516, "step": 10426 }, { "epoch": 1.88, "learning_rate": 1.7323332730887403e-06, "loss": 0.288, "step": 10427 }, { "epoch": 1.88, "learning_rate": 1.7296222664015906e-06, "loss": 0.7292, "step": 10428 }, { "epoch": 1.88, "learning_rate": 1.7269112597144407e-06, "loss": 0.886, "step": 10429 }, { "epoch": 1.89, "learning_rate": 1.7242002530272909e-06, "loss": 0.6064, "step": 10430 }, { "epoch": 1.89, "learning_rate": 1.721489246340141e-06, "loss": 0.8024, "step": 10431 }, { "epoch": 1.89, "learning_rate": 1.7187782396529912e-06, "loss": 1.0677, "step": 10432 }, { "epoch": 1.89, "learning_rate": 1.7160672329658413e-06, "loss": 0.8465, "step": 10433 }, { "epoch": 1.89, "learning_rate": 1.7133562262786916e-06, "loss": 0.7998, "step": 10434 }, { "epoch": 1.89, "learning_rate": 1.7106452195915418e-06, "loss": 0.7365, "step": 10435 }, { "epoch": 1.89, "learning_rate": 1.707934212904392e-06, "loss": 1.3211, "step": 10436 }, { "epoch": 1.89, "learning_rate": 1.705223206217242e-06, "loss": 0.3939, "step": 10437 }, { "epoch": 1.89, "learning_rate": 1.7025121995300922e-06, "loss": 1.2495, "step": 10438 }, { "epoch": 1.89, "learning_rate": 1.6998011928429425e-06, "loss": 0.2617, "step": 10439 }, { "epoch": 1.89, "learning_rate": 1.6970901861557927e-06, "loss": 0.5975, "step": 10440 }, { "epoch": 1.89, "learning_rate": 1.6943791794686428e-06, "loss": 0.8337, "step": 10441 }, { "epoch": 1.89, "learning_rate": 1.691668172781493e-06, "loss": 0.5721, "step": 10442 }, { "epoch": 1.89, "learning_rate": 1.688957166094343e-06, "loss": 0.969, "step": 10443 }, { "epoch": 1.89, "learning_rate": 1.6862461594071934e-06, "loss": 1.0905, "step": 10444 }, { "epoch": 1.89, "learning_rate": 1.6835351527200435e-06, "loss": 1.1851, "step": 10445 }, { "epoch": 1.89, "learning_rate": 1.6808241460328937e-06, "loss": 1.6319, "step": 10446 }, { "epoch": 1.89, "learning_rate": 1.6781131393457438e-06, "loss": 0.9048, "step": 10447 }, { "epoch": 1.89, "learning_rate": 1.675402132658594e-06, "loss": 0.8106, "step": 10448 }, { "epoch": 1.89, "learning_rate": 1.6726911259714443e-06, "loss": 0.513, "step": 10449 }, { "epoch": 1.89, "learning_rate": 1.6699801192842944e-06, "loss": 1.0035, "step": 10450 }, { "epoch": 1.89, "learning_rate": 1.6672691125971446e-06, "loss": 0.8802, "step": 10451 }, { "epoch": 1.89, "learning_rate": 1.6645581059099947e-06, "loss": 0.5418, "step": 10452 }, { "epoch": 1.89, "learning_rate": 1.6618470992228448e-06, "loss": 0.497, "step": 10453 }, { "epoch": 1.89, "learning_rate": 1.6591360925356952e-06, "loss": 0.4757, "step": 10454 }, { "epoch": 1.89, "learning_rate": 1.656425085848545e-06, "loss": 0.3434, "step": 10455 }, { "epoch": 1.89, "learning_rate": 1.6537140791613952e-06, "loss": 0.9826, "step": 10456 }, { "epoch": 1.89, "learning_rate": 1.6510030724742454e-06, "loss": 0.8279, "step": 10457 }, { "epoch": 1.89, "learning_rate": 1.6482920657870955e-06, "loss": 0.5949, "step": 10458 }, { "epoch": 1.89, "learning_rate": 1.6455810590999458e-06, "loss": 0.6174, "step": 10459 }, { "epoch": 1.89, "learning_rate": 1.642870052412796e-06, "loss": 0.7034, "step": 10460 }, { "epoch": 1.89, "learning_rate": 1.640159045725646e-06, "loss": 0.2829, "step": 10461 }, { "epoch": 1.89, "learning_rate": 1.6374480390384962e-06, "loss": 1.0014, "step": 10462 }, { "epoch": 1.89, "learning_rate": 1.6347370323513464e-06, "loss": 0.6543, "step": 10463 }, { "epoch": 1.89, "learning_rate": 1.6320260256641965e-06, "loss": 0.454, "step": 10464 }, { "epoch": 1.89, "learning_rate": 1.6293150189770469e-06, "loss": 1.005, "step": 10465 }, { "epoch": 1.89, "learning_rate": 1.626604012289897e-06, "loss": 1.0408, "step": 10466 }, { "epoch": 1.89, "learning_rate": 1.6238930056027471e-06, "loss": 0.6006, "step": 10467 }, { "epoch": 1.89, "learning_rate": 1.6211819989155973e-06, "loss": 0.6424, "step": 10468 }, { "epoch": 1.89, "learning_rate": 1.6184709922284474e-06, "loss": 0.9381, "step": 10469 }, { "epoch": 1.89, "learning_rate": 1.6157599855412977e-06, "loss": 0.9991, "step": 10470 }, { "epoch": 1.89, "learning_rate": 1.6130489788541479e-06, "loss": 0.9822, "step": 10471 }, { "epoch": 1.89, "learning_rate": 1.610337972166998e-06, "loss": 0.3622, "step": 10472 }, { "epoch": 1.89, "learning_rate": 1.6076269654798481e-06, "loss": 0.7503, "step": 10473 }, { "epoch": 1.89, "learning_rate": 1.6049159587926983e-06, "loss": 0.5281, "step": 10474 }, { "epoch": 1.89, "learning_rate": 1.6022049521055486e-06, "loss": 0.4735, "step": 10475 }, { "epoch": 1.89, "learning_rate": 1.5994939454183988e-06, "loss": 0.6558, "step": 10476 }, { "epoch": 1.89, "learning_rate": 1.5967829387312489e-06, "loss": 0.6747, "step": 10477 }, { "epoch": 1.89, "learning_rate": 1.594071932044099e-06, "loss": 0.6185, "step": 10478 }, { "epoch": 1.89, "learning_rate": 1.5913609253569492e-06, "loss": 1.2186, "step": 10479 }, { "epoch": 1.89, "learning_rate": 1.5886499186697995e-06, "loss": 0.611, "step": 10480 }, { "epoch": 1.89, "learning_rate": 1.5859389119826496e-06, "loss": 0.5137, "step": 10481 }, { "epoch": 1.89, "learning_rate": 1.5832279052954998e-06, "loss": 1.1754, "step": 10482 }, { "epoch": 1.89, "learning_rate": 1.58051689860835e-06, "loss": 0.91, "step": 10483 }, { "epoch": 1.89, "learning_rate": 1.5778058919212e-06, "loss": 0.9331, "step": 10484 }, { "epoch": 1.89, "learning_rate": 1.5750948852340502e-06, "loss": 0.7206, "step": 10485 }, { "epoch": 1.9, "learning_rate": 1.5723838785469005e-06, "loss": 0.5098, "step": 10486 }, { "epoch": 1.9, "learning_rate": 1.5696728718597507e-06, "loss": 0.8802, "step": 10487 }, { "epoch": 1.9, "learning_rate": 1.5669618651726008e-06, "loss": 0.4244, "step": 10488 }, { "epoch": 1.9, "learning_rate": 1.564250858485451e-06, "loss": 0.4698, "step": 10489 }, { "epoch": 1.9, "learning_rate": 1.561539851798301e-06, "loss": 0.7666, "step": 10490 }, { "epoch": 1.9, "learning_rate": 1.5588288451111514e-06, "loss": 0.5374, "step": 10491 }, { "epoch": 1.9, "learning_rate": 1.5561178384240015e-06, "loss": 0.5666, "step": 10492 }, { "epoch": 1.9, "learning_rate": 1.5534068317368517e-06, "loss": 0.9331, "step": 10493 }, { "epoch": 1.9, "learning_rate": 1.5506958250497018e-06, "loss": 1.0123, "step": 10494 }, { "epoch": 1.9, "learning_rate": 1.547984818362552e-06, "loss": 0.5551, "step": 10495 }, { "epoch": 1.9, "learning_rate": 1.5452738116754023e-06, "loss": 1.0967, "step": 10496 }, { "epoch": 1.9, "learning_rate": 1.5425628049882524e-06, "loss": 0.6481, "step": 10497 }, { "epoch": 1.9, "learning_rate": 1.5398517983011026e-06, "loss": 0.479, "step": 10498 }, { "epoch": 1.9, "learning_rate": 1.5371407916139527e-06, "loss": 0.9286, "step": 10499 }, { "epoch": 1.9, "learning_rate": 1.5344297849268028e-06, "loss": 0.6357, "step": 10500 }, { "epoch": 1.9, "learning_rate": 1.5317187782396532e-06, "loss": 1.0608, "step": 10501 }, { "epoch": 1.9, "learning_rate": 1.5290077715525033e-06, "loss": 1.2448, "step": 10502 }, { "epoch": 1.9, "learning_rate": 1.5262967648653534e-06, "loss": 0.2996, "step": 10503 }, { "epoch": 1.9, "learning_rate": 1.5235857581782036e-06, "loss": 0.8536, "step": 10504 }, { "epoch": 1.9, "learning_rate": 1.5208747514910537e-06, "loss": 0.6186, "step": 10505 }, { "epoch": 1.9, "learning_rate": 1.518163744803904e-06, "loss": 0.8747, "step": 10506 }, { "epoch": 1.9, "learning_rate": 1.5154527381167542e-06, "loss": 0.5219, "step": 10507 }, { "epoch": 1.9, "learning_rate": 1.5127417314296043e-06, "loss": 0.8289, "step": 10508 }, { "epoch": 1.9, "learning_rate": 1.5100307247424545e-06, "loss": 0.5836, "step": 10509 }, { "epoch": 1.9, "learning_rate": 1.5073197180553046e-06, "loss": 0.2553, "step": 10510 }, { "epoch": 1.9, "learning_rate": 1.5046087113681547e-06, "loss": 0.544, "step": 10511 }, { "epoch": 1.9, "learning_rate": 1.501897704681005e-06, "loss": 0.2638, "step": 10512 }, { "epoch": 1.9, "learning_rate": 1.4991866979938552e-06, "loss": 0.5012, "step": 10513 }, { "epoch": 1.9, "learning_rate": 1.4964756913067053e-06, "loss": 1.104, "step": 10514 }, { "epoch": 1.9, "learning_rate": 1.4937646846195555e-06, "loss": 0.4988, "step": 10515 }, { "epoch": 1.9, "learning_rate": 1.4910536779324056e-06, "loss": 0.8388, "step": 10516 }, { "epoch": 1.9, "learning_rate": 1.488342671245256e-06, "loss": 0.9396, "step": 10517 }, { "epoch": 1.9, "learning_rate": 1.485631664558106e-06, "loss": 0.9373, "step": 10518 }, { "epoch": 1.9, "learning_rate": 1.4829206578709562e-06, "loss": 0.8816, "step": 10519 }, { "epoch": 1.9, "learning_rate": 1.4802096511838061e-06, "loss": 0.4015, "step": 10520 }, { "epoch": 1.9, "learning_rate": 1.4774986444966563e-06, "loss": 0.6419, "step": 10521 }, { "epoch": 1.9, "learning_rate": 1.4747876378095066e-06, "loss": 0.6144, "step": 10522 }, { "epoch": 1.9, "learning_rate": 1.4720766311223568e-06, "loss": 0.4473, "step": 10523 }, { "epoch": 1.9, "learning_rate": 1.469365624435207e-06, "loss": 0.7284, "step": 10524 }, { "epoch": 1.9, "learning_rate": 1.466654617748057e-06, "loss": 0.3594, "step": 10525 }, { "epoch": 1.9, "learning_rate": 1.4639436110609072e-06, "loss": 0.9663, "step": 10526 }, { "epoch": 1.9, "learning_rate": 1.4612326043737575e-06, "loss": 0.6046, "step": 10527 }, { "epoch": 1.9, "learning_rate": 1.4585215976866076e-06, "loss": 0.7802, "step": 10528 }, { "epoch": 1.9, "learning_rate": 1.4558105909994578e-06, "loss": 0.4117, "step": 10529 }, { "epoch": 1.9, "learning_rate": 1.453099584312308e-06, "loss": 0.9009, "step": 10530 }, { "epoch": 1.9, "learning_rate": 1.450388577625158e-06, "loss": 0.5827, "step": 10531 }, { "epoch": 1.9, "learning_rate": 1.4476775709380084e-06, "loss": 0.8169, "step": 10532 }, { "epoch": 1.9, "learning_rate": 1.4449665642508585e-06, "loss": 0.975, "step": 10533 }, { "epoch": 1.9, "learning_rate": 1.4422555575637087e-06, "loss": 0.3795, "step": 10534 }, { "epoch": 1.9, "learning_rate": 1.4395445508765588e-06, "loss": 0.8887, "step": 10535 }, { "epoch": 1.9, "learning_rate": 1.436833544189409e-06, "loss": 1.1393, "step": 10536 }, { "epoch": 1.9, "learning_rate": 1.434122537502259e-06, "loss": 0.6201, "step": 10537 }, { "epoch": 1.9, "learning_rate": 1.4314115308151094e-06, "loss": 0.8955, "step": 10538 }, { "epoch": 1.9, "learning_rate": 1.4287005241279595e-06, "loss": 0.8106, "step": 10539 }, { "epoch": 1.9, "learning_rate": 1.4259895174408097e-06, "loss": 0.4805, "step": 10540 }, { "epoch": 1.91, "learning_rate": 1.4232785107536598e-06, "loss": 0.5858, "step": 10541 }, { "epoch": 1.91, "learning_rate": 1.42056750406651e-06, "loss": 0.533, "step": 10542 }, { "epoch": 1.91, "learning_rate": 1.4178564973793603e-06, "loss": 0.5374, "step": 10543 }, { "epoch": 1.91, "learning_rate": 1.4151454906922104e-06, "loss": 0.8659, "step": 10544 }, { "epoch": 1.91, "learning_rate": 1.4124344840050606e-06, "loss": 0.302, "step": 10545 }, { "epoch": 1.91, "learning_rate": 1.4097234773179107e-06, "loss": 0.9551, "step": 10546 }, { "epoch": 1.91, "learning_rate": 1.4070124706307608e-06, "loss": 0.5668, "step": 10547 }, { "epoch": 1.91, "learning_rate": 1.4043014639436112e-06, "loss": 0.5539, "step": 10548 }, { "epoch": 1.91, "learning_rate": 1.4015904572564613e-06, "loss": 1.0328, "step": 10549 }, { "epoch": 1.91, "learning_rate": 1.3988794505693114e-06, "loss": 0.5221, "step": 10550 }, { "epoch": 1.91, "learning_rate": 1.3961684438821616e-06, "loss": 1.0415, "step": 10551 }, { "epoch": 1.91, "learning_rate": 1.3934574371950117e-06, "loss": 0.8335, "step": 10552 }, { "epoch": 1.91, "learning_rate": 1.390746430507862e-06, "loss": 0.6166, "step": 10553 }, { "epoch": 1.91, "learning_rate": 1.3880354238207122e-06, "loss": 0.6832, "step": 10554 }, { "epoch": 1.91, "learning_rate": 1.3853244171335623e-06, "loss": 0.5536, "step": 10555 }, { "epoch": 1.91, "learning_rate": 1.3826134104464125e-06, "loss": 1.0655, "step": 10556 }, { "epoch": 1.91, "learning_rate": 1.3799024037592626e-06, "loss": 0.8088, "step": 10557 }, { "epoch": 1.91, "learning_rate": 1.3771913970721127e-06, "loss": 0.9028, "step": 10558 }, { "epoch": 1.91, "learning_rate": 1.374480390384963e-06, "loss": 0.6991, "step": 10559 }, { "epoch": 1.91, "learning_rate": 1.3717693836978132e-06, "loss": 0.8549, "step": 10560 }, { "epoch": 1.91, "learning_rate": 1.3690583770106633e-06, "loss": 0.4908, "step": 10561 }, { "epoch": 1.91, "learning_rate": 1.3663473703235135e-06, "loss": 0.6372, "step": 10562 }, { "epoch": 1.91, "learning_rate": 1.3636363636363636e-06, "loss": 0.797, "step": 10563 }, { "epoch": 1.91, "learning_rate": 1.360925356949214e-06, "loss": 0.5636, "step": 10564 }, { "epoch": 1.91, "learning_rate": 1.358214350262064e-06, "loss": 0.7699, "step": 10565 }, { "epoch": 1.91, "learning_rate": 1.3555033435749142e-06, "loss": 0.6549, "step": 10566 }, { "epoch": 1.91, "learning_rate": 1.3527923368877644e-06, "loss": 0.7668, "step": 10567 }, { "epoch": 1.91, "learning_rate": 1.3500813302006145e-06, "loss": 0.4243, "step": 10568 }, { "epoch": 1.91, "learning_rate": 1.3473703235134648e-06, "loss": 0.3843, "step": 10569 }, { "epoch": 1.91, "learning_rate": 1.344659316826315e-06, "loss": 0.6633, "step": 10570 }, { "epoch": 1.91, "learning_rate": 1.3419483101391651e-06, "loss": 0.4826, "step": 10571 }, { "epoch": 1.91, "learning_rate": 1.3392373034520152e-06, "loss": 0.9356, "step": 10572 }, { "epoch": 1.91, "learning_rate": 1.3365262967648654e-06, "loss": 0.7392, "step": 10573 }, { "epoch": 1.91, "learning_rate": 1.3338152900777157e-06, "loss": 0.4416, "step": 10574 }, { "epoch": 1.91, "learning_rate": 1.3311042833905659e-06, "loss": 0.8884, "step": 10575 }, { "epoch": 1.91, "learning_rate": 1.328393276703416e-06, "loss": 0.5443, "step": 10576 }, { "epoch": 1.91, "learning_rate": 1.3256822700162661e-06, "loss": 0.4771, "step": 10577 }, { "epoch": 1.91, "learning_rate": 1.3229712633291163e-06, "loss": 0.4661, "step": 10578 }, { "epoch": 1.91, "learning_rate": 1.3202602566419666e-06, "loss": 0.623, "step": 10579 }, { "epoch": 1.91, "learning_rate": 1.3175492499548167e-06, "loss": 0.5801, "step": 10580 }, { "epoch": 1.91, "learning_rate": 1.3148382432676669e-06, "loss": 1.3424, "step": 10581 }, { "epoch": 1.91, "learning_rate": 1.312127236580517e-06, "loss": 0.4303, "step": 10582 }, { "epoch": 1.91, "learning_rate": 1.3094162298933671e-06, "loss": 1.2358, "step": 10583 }, { "epoch": 1.91, "learning_rate": 1.306705223206217e-06, "loss": 1.0214, "step": 10584 }, { "epoch": 1.91, "learning_rate": 1.3039942165190674e-06, "loss": 0.7718, "step": 10585 }, { "epoch": 1.91, "learning_rate": 1.3012832098319175e-06, "loss": 1.1999, "step": 10586 }, { "epoch": 1.91, "learning_rate": 1.2985722031447677e-06, "loss": 0.7123, "step": 10587 }, { "epoch": 1.91, "learning_rate": 1.2958611964576178e-06, "loss": 0.7926, "step": 10588 }, { "epoch": 1.91, "learning_rate": 1.293150189770468e-06, "loss": 1.3456, "step": 10589 }, { "epoch": 1.91, "learning_rate": 1.2904391830833183e-06, "loss": 0.2802, "step": 10590 }, { "epoch": 1.91, "learning_rate": 1.2877281763961684e-06, "loss": 1.2769, "step": 10591 }, { "epoch": 1.91, "learning_rate": 1.2850171697090186e-06, "loss": 0.559, "step": 10592 }, { "epoch": 1.91, "learning_rate": 1.2823061630218687e-06, "loss": 0.7939, "step": 10593 }, { "epoch": 1.91, "learning_rate": 1.2795951563347188e-06, "loss": 0.8965, "step": 10594 }, { "epoch": 1.91, "learning_rate": 1.2768841496475692e-06, "loss": 0.3305, "step": 10595 }, { "epoch": 1.92, "learning_rate": 1.2741731429604193e-06, "loss": 1.4532, "step": 10596 }, { "epoch": 1.92, "learning_rate": 1.2714621362732694e-06, "loss": 1.1457, "step": 10597 }, { "epoch": 1.92, "learning_rate": 1.2687511295861196e-06, "loss": 0.6433, "step": 10598 }, { "epoch": 1.92, "learning_rate": 1.2660401228989697e-06, "loss": 0.5456, "step": 10599 }, { "epoch": 1.92, "learning_rate": 1.26332911621182e-06, "loss": 1.2682, "step": 10600 }, { "epoch": 1.92, "learning_rate": 1.2606181095246702e-06, "loss": 0.661, "step": 10601 }, { "epoch": 1.92, "learning_rate": 1.2579071028375203e-06, "loss": 0.8911, "step": 10602 }, { "epoch": 1.92, "learning_rate": 1.2551960961503705e-06, "loss": 0.3676, "step": 10603 }, { "epoch": 1.92, "learning_rate": 1.2524850894632206e-06, "loss": 0.3337, "step": 10604 }, { "epoch": 1.92, "learning_rate": 1.249774082776071e-06, "loss": 0.9288, "step": 10605 }, { "epoch": 1.92, "learning_rate": 1.247063076088921e-06, "loss": 0.4938, "step": 10606 }, { "epoch": 1.92, "learning_rate": 1.2443520694017712e-06, "loss": 0.8888, "step": 10607 }, { "epoch": 1.92, "learning_rate": 1.2416410627146213e-06, "loss": 1.0016, "step": 10608 }, { "epoch": 1.92, "learning_rate": 1.2389300560274715e-06, "loss": 0.91, "step": 10609 }, { "epoch": 1.92, "learning_rate": 1.2362190493403216e-06, "loss": 0.4438, "step": 10610 }, { "epoch": 1.92, "learning_rate": 1.233508042653172e-06, "loss": 0.5128, "step": 10611 }, { "epoch": 1.92, "learning_rate": 1.230797035966022e-06, "loss": 0.914, "step": 10612 }, { "epoch": 1.92, "learning_rate": 1.2280860292788722e-06, "loss": 0.9649, "step": 10613 }, { "epoch": 1.92, "learning_rate": 1.2253750225917224e-06, "loss": 0.612, "step": 10614 }, { "epoch": 1.92, "learning_rate": 1.2226640159045725e-06, "loss": 0.4757, "step": 10615 }, { "epoch": 1.92, "learning_rate": 1.2199530092174228e-06, "loss": 0.6444, "step": 10616 }, { "epoch": 1.92, "learning_rate": 1.217242002530273e-06, "loss": 0.6658, "step": 10617 }, { "epoch": 1.92, "learning_rate": 1.2145309958431231e-06, "loss": 0.3485, "step": 10618 }, { "epoch": 1.92, "learning_rate": 1.2118199891559733e-06, "loss": 0.8744, "step": 10619 }, { "epoch": 1.92, "learning_rate": 1.2091089824688234e-06, "loss": 0.9037, "step": 10620 }, { "epoch": 1.92, "learning_rate": 1.2063979757816737e-06, "loss": 0.9285, "step": 10621 }, { "epoch": 1.92, "learning_rate": 1.2036869690945239e-06, "loss": 0.2973, "step": 10622 }, { "epoch": 1.92, "learning_rate": 1.200975962407374e-06, "loss": 0.9977, "step": 10623 }, { "epoch": 1.92, "learning_rate": 1.1982649557202241e-06, "loss": 0.8466, "step": 10624 }, { "epoch": 1.92, "learning_rate": 1.1955539490330743e-06, "loss": 0.5699, "step": 10625 }, { "epoch": 1.92, "learning_rate": 1.1928429423459246e-06, "loss": 0.3273, "step": 10626 }, { "epoch": 1.92, "learning_rate": 1.1901319356587747e-06, "loss": 0.9475, "step": 10627 }, { "epoch": 1.92, "learning_rate": 1.1874209289716249e-06, "loss": 0.694, "step": 10628 }, { "epoch": 1.92, "learning_rate": 1.184709922284475e-06, "loss": 1.0804, "step": 10629 }, { "epoch": 1.92, "learning_rate": 1.1819989155973252e-06, "loss": 0.5092, "step": 10630 }, { "epoch": 1.92, "learning_rate": 1.1792879089101755e-06, "loss": 1.2004, "step": 10631 }, { "epoch": 1.92, "learning_rate": 1.1765769022230256e-06, "loss": 0.3413, "step": 10632 }, { "epoch": 1.92, "learning_rate": 1.1738658955358758e-06, "loss": 0.6452, "step": 10633 }, { "epoch": 1.92, "learning_rate": 1.171154888848726e-06, "loss": 0.8279, "step": 10634 }, { "epoch": 1.92, "learning_rate": 1.168443882161576e-06, "loss": 0.3293, "step": 10635 }, { "epoch": 1.92, "learning_rate": 1.1657328754744262e-06, "loss": 0.5474, "step": 10636 }, { "epoch": 1.92, "learning_rate": 1.1630218687872765e-06, "loss": 0.1642, "step": 10637 }, { "epoch": 1.92, "learning_rate": 1.1603108621001266e-06, "loss": 0.2176, "step": 10638 }, { "epoch": 1.92, "learning_rate": 1.1575998554129768e-06, "loss": 0.2949, "step": 10639 }, { "epoch": 1.92, "learning_rate": 1.154888848725827e-06, "loss": 1.0473, "step": 10640 }, { "epoch": 1.92, "learning_rate": 1.152177842038677e-06, "loss": 0.6519, "step": 10641 }, { "epoch": 1.92, "learning_rate": 1.1494668353515274e-06, "loss": 0.3889, "step": 10642 }, { "epoch": 1.92, "learning_rate": 1.1467558286643775e-06, "loss": 1.0133, "step": 10643 }, { "epoch": 1.92, "learning_rate": 1.1440448219772277e-06, "loss": 0.8063, "step": 10644 }, { "epoch": 1.92, "learning_rate": 1.1413338152900778e-06, "loss": 0.3005, "step": 10645 }, { "epoch": 1.92, "learning_rate": 1.138622808602928e-06, "loss": 0.7134, "step": 10646 }, { "epoch": 1.92, "learning_rate": 1.1359118019157783e-06, "loss": 0.8426, "step": 10647 }, { "epoch": 1.92, "learning_rate": 1.1332007952286282e-06, "loss": 0.6039, "step": 10648 }, { "epoch": 1.92, "learning_rate": 1.1304897885414783e-06, "loss": 0.939, "step": 10649 }, { "epoch": 1.92, "learning_rate": 1.1277787818543285e-06, "loss": 0.7691, "step": 10650 }, { "epoch": 1.92, "learning_rate": 1.1250677751671786e-06, "loss": 0.8365, "step": 10651 }, { "epoch": 1.93, "learning_rate": 1.122356768480029e-06, "loss": 0.4931, "step": 10652 }, { "epoch": 1.93, "learning_rate": 1.119645761792879e-06, "loss": 0.4145, "step": 10653 }, { "epoch": 1.93, "learning_rate": 1.1169347551057292e-06, "loss": 1.2955, "step": 10654 }, { "epoch": 1.93, "learning_rate": 1.1142237484185794e-06, "loss": 1.2013, "step": 10655 }, { "epoch": 1.93, "learning_rate": 1.1115127417314295e-06, "loss": 0.7726, "step": 10656 }, { "epoch": 1.93, "learning_rate": 1.1088017350442798e-06, "loss": 0.967, "step": 10657 }, { "epoch": 1.93, "learning_rate": 1.10609072835713e-06, "loss": 0.7377, "step": 10658 }, { "epoch": 1.93, "learning_rate": 1.10337972166998e-06, "loss": 0.4264, "step": 10659 }, { "epoch": 1.93, "learning_rate": 1.1006687149828302e-06, "loss": 0.9541, "step": 10660 }, { "epoch": 1.93, "learning_rate": 1.0979577082956804e-06, "loss": 0.5297, "step": 10661 }, { "epoch": 1.93, "learning_rate": 1.0952467016085305e-06, "loss": 0.5817, "step": 10662 }, { "epoch": 1.93, "learning_rate": 1.0925356949213809e-06, "loss": 1.2802, "step": 10663 }, { "epoch": 1.93, "learning_rate": 1.089824688234231e-06, "loss": 0.6664, "step": 10664 }, { "epoch": 1.93, "learning_rate": 1.0871136815470811e-06, "loss": 0.5913, "step": 10665 }, { "epoch": 1.93, "learning_rate": 1.0844026748599313e-06, "loss": 0.444, "step": 10666 }, { "epoch": 1.93, "learning_rate": 1.0816916681727814e-06, "loss": 0.655, "step": 10667 }, { "epoch": 1.93, "learning_rate": 1.0789806614856317e-06, "loss": 0.7119, "step": 10668 }, { "epoch": 1.93, "learning_rate": 1.0762696547984819e-06, "loss": 0.2152, "step": 10669 }, { "epoch": 1.93, "learning_rate": 1.073558648111332e-06, "loss": 0.5829, "step": 10670 }, { "epoch": 1.93, "learning_rate": 1.0708476414241821e-06, "loss": 0.6183, "step": 10671 }, { "epoch": 1.93, "learning_rate": 1.0681366347370323e-06, "loss": 0.8399, "step": 10672 }, { "epoch": 1.93, "learning_rate": 1.0654256280498826e-06, "loss": 1.498, "step": 10673 }, { "epoch": 1.93, "learning_rate": 1.0627146213627328e-06, "loss": 0.5784, "step": 10674 }, { "epoch": 1.93, "learning_rate": 1.0600036146755829e-06, "loss": 0.9685, "step": 10675 }, { "epoch": 1.93, "learning_rate": 1.057292607988433e-06, "loss": 0.2872, "step": 10676 }, { "epoch": 1.93, "learning_rate": 1.0545816013012832e-06, "loss": 0.9061, "step": 10677 }, { "epoch": 1.93, "learning_rate": 1.0518705946141335e-06, "loss": 0.2961, "step": 10678 }, { "epoch": 1.93, "learning_rate": 1.0491595879269836e-06, "loss": 1.3998, "step": 10679 }, { "epoch": 1.93, "learning_rate": 1.0464485812398338e-06, "loss": 0.9921, "step": 10680 }, { "epoch": 1.93, "learning_rate": 1.043737574552684e-06, "loss": 0.6809, "step": 10681 }, { "epoch": 1.93, "learning_rate": 1.041026567865534e-06, "loss": 0.9405, "step": 10682 }, { "epoch": 1.93, "learning_rate": 1.0383155611783842e-06, "loss": 0.4087, "step": 10683 }, { "epoch": 1.93, "learning_rate": 1.0356045544912345e-06, "loss": 0.7111, "step": 10684 }, { "epoch": 1.93, "learning_rate": 1.0328935478040847e-06, "loss": 0.3934, "step": 10685 }, { "epoch": 1.93, "learning_rate": 1.0301825411169348e-06, "loss": 0.4026, "step": 10686 }, { "epoch": 1.93, "learning_rate": 1.027471534429785e-06, "loss": 0.973, "step": 10687 }, { "epoch": 1.93, "learning_rate": 1.024760527742635e-06, "loss": 0.4613, "step": 10688 }, { "epoch": 1.93, "learning_rate": 1.0220495210554854e-06, "loss": 0.7476, "step": 10689 }, { "epoch": 1.93, "learning_rate": 1.0193385143683355e-06, "loss": 0.2904, "step": 10690 }, { "epoch": 1.93, "learning_rate": 1.0166275076811857e-06, "loss": 0.7323, "step": 10691 }, { "epoch": 1.93, "learning_rate": 1.0139165009940358e-06, "loss": 0.6783, "step": 10692 }, { "epoch": 1.93, "learning_rate": 1.011205494306886e-06, "loss": 0.6938, "step": 10693 }, { "epoch": 1.93, "learning_rate": 1.0084944876197363e-06, "loss": 0.7717, "step": 10694 }, { "epoch": 1.93, "learning_rate": 1.0057834809325864e-06, "loss": 0.8241, "step": 10695 }, { "epoch": 1.93, "learning_rate": 1.0030724742454366e-06, "loss": 1.4782, "step": 10696 }, { "epoch": 1.93, "learning_rate": 1.0003614675582867e-06, "loss": 1.1774, "step": 10697 }, { "epoch": 1.93, "learning_rate": 9.976504608711368e-07, "loss": 0.6376, "step": 10698 }, { "epoch": 1.93, "learning_rate": 9.949394541839872e-07, "loss": 0.4407, "step": 10699 }, { "epoch": 1.93, "learning_rate": 9.922284474968373e-07, "loss": 0.6535, "step": 10700 }, { "epoch": 1.93, "learning_rate": 9.895174408096874e-07, "loss": 1.0971, "step": 10701 }, { "epoch": 1.93, "learning_rate": 9.868064341225376e-07, "loss": 0.3355, "step": 10702 }, { "epoch": 1.93, "learning_rate": 9.840954274353877e-07, "loss": 0.8386, "step": 10703 }, { "epoch": 1.93, "learning_rate": 9.81384420748238e-07, "loss": 1.2594, "step": 10704 }, { "epoch": 1.93, "learning_rate": 9.786734140610882e-07, "loss": 1.3227, "step": 10705 }, { "epoch": 1.93, "learning_rate": 9.759624073739383e-07, "loss": 0.7263, "step": 10706 }, { "epoch": 1.94, "learning_rate": 9.732514006867885e-07, "loss": 0.4048, "step": 10707 }, { "epoch": 1.94, "learning_rate": 9.705403939996386e-07, "loss": 0.5995, "step": 10708 }, { "epoch": 1.94, "learning_rate": 9.678293873124887e-07, "loss": 0.5831, "step": 10709 }, { "epoch": 1.94, "learning_rate": 9.65118380625339e-07, "loss": 0.6266, "step": 10710 }, { "epoch": 1.94, "learning_rate": 9.624073739381892e-07, "loss": 1.2054, "step": 10711 }, { "epoch": 1.94, "learning_rate": 9.596963672510393e-07, "loss": 0.8757, "step": 10712 }, { "epoch": 1.94, "learning_rate": 9.569853605638893e-07, "loss": 0.6794, "step": 10713 }, { "epoch": 1.94, "learning_rate": 9.542743538767394e-07, "loss": 0.5199, "step": 10714 }, { "epoch": 1.94, "learning_rate": 9.515633471895896e-07, "loss": 0.9414, "step": 10715 }, { "epoch": 1.94, "learning_rate": 9.488523405024399e-07, "loss": 1.0303, "step": 10716 }, { "epoch": 1.94, "learning_rate": 9.4614133381529e-07, "loss": 0.7973, "step": 10717 }, { "epoch": 1.94, "learning_rate": 9.434303271281401e-07, "loss": 0.5439, "step": 10718 }, { "epoch": 1.94, "learning_rate": 9.407193204409904e-07, "loss": 0.2546, "step": 10719 }, { "epoch": 1.94, "learning_rate": 9.380083137538405e-07, "loss": 0.3948, "step": 10720 }, { "epoch": 1.94, "learning_rate": 9.352973070666909e-07, "loss": 0.8121, "step": 10721 }, { "epoch": 1.94, "learning_rate": 9.32586300379541e-07, "loss": 0.4386, "step": 10722 }, { "epoch": 1.94, "learning_rate": 9.298752936923911e-07, "loss": 0.5275, "step": 10723 }, { "epoch": 1.94, "learning_rate": 9.271642870052414e-07, "loss": 0.7608, "step": 10724 }, { "epoch": 1.94, "learning_rate": 9.244532803180915e-07, "loss": 1.1615, "step": 10725 }, { "epoch": 1.94, "learning_rate": 9.217422736309416e-07, "loss": 0.5794, "step": 10726 }, { "epoch": 1.94, "learning_rate": 9.190312669437919e-07, "loss": 0.4866, "step": 10727 }, { "epoch": 1.94, "learning_rate": 9.16320260256642e-07, "loss": 0.9027, "step": 10728 }, { "epoch": 1.94, "learning_rate": 9.136092535694921e-07, "loss": 1.0761, "step": 10729 }, { "epoch": 1.94, "learning_rate": 9.108982468823423e-07, "loss": 0.2938, "step": 10730 }, { "epoch": 1.94, "learning_rate": 9.081872401951924e-07, "loss": 0.7183, "step": 10731 }, { "epoch": 1.94, "learning_rate": 9.054762335080427e-07, "loss": 0.9061, "step": 10732 }, { "epoch": 1.94, "learning_rate": 9.027652268208928e-07, "loss": 0.7161, "step": 10733 }, { "epoch": 1.94, "learning_rate": 9.00054220133743e-07, "loss": 0.9015, "step": 10734 }, { "epoch": 1.94, "learning_rate": 8.973432134465932e-07, "loss": 1.1441, "step": 10735 }, { "epoch": 1.94, "learning_rate": 8.946322067594433e-07, "loss": 0.4616, "step": 10736 }, { "epoch": 1.94, "learning_rate": 8.919212000722935e-07, "loss": 0.39, "step": 10737 }, { "epoch": 1.94, "learning_rate": 8.892101933851437e-07, "loss": 0.6444, "step": 10738 }, { "epoch": 1.94, "learning_rate": 8.864991866979938e-07, "loss": 0.6283, "step": 10739 }, { "epoch": 1.94, "learning_rate": 8.83788180010844e-07, "loss": 0.8373, "step": 10740 }, { "epoch": 1.94, "learning_rate": 8.810771733236942e-07, "loss": 0.5074, "step": 10741 }, { "epoch": 1.94, "learning_rate": 8.783661666365444e-07, "loss": 0.9349, "step": 10742 }, { "epoch": 1.94, "learning_rate": 8.756551599493946e-07, "loss": 1.1557, "step": 10743 }, { "epoch": 1.94, "learning_rate": 8.729441532622447e-07, "loss": 0.8553, "step": 10744 }, { "epoch": 1.94, "learning_rate": 8.702331465750949e-07, "loss": 0.9448, "step": 10745 }, { "epoch": 1.94, "learning_rate": 8.675221398879451e-07, "loss": 0.4933, "step": 10746 }, { "epoch": 1.94, "learning_rate": 8.648111332007953e-07, "loss": 0.3823, "step": 10747 }, { "epoch": 1.94, "learning_rate": 8.621001265136454e-07, "loss": 0.7609, "step": 10748 }, { "epoch": 1.94, "learning_rate": 8.593891198264956e-07, "loss": 1.0833, "step": 10749 }, { "epoch": 1.94, "learning_rate": 8.566781131393458e-07, "loss": 0.5999, "step": 10750 }, { "epoch": 1.94, "learning_rate": 8.53967106452196e-07, "loss": 0.7522, "step": 10751 }, { "epoch": 1.94, "learning_rate": 8.512560997650461e-07, "loss": 0.7804, "step": 10752 }, { "epoch": 1.94, "learning_rate": 8.485450930778963e-07, "loss": 0.9872, "step": 10753 }, { "epoch": 1.94, "learning_rate": 8.458340863907465e-07, "loss": 0.7538, "step": 10754 }, { "epoch": 1.94, "learning_rate": 8.431230797035967e-07, "loss": 0.9187, "step": 10755 }, { "epoch": 1.94, "learning_rate": 8.404120730164468e-07, "loss": 0.3302, "step": 10756 }, { "epoch": 1.94, "learning_rate": 8.37701066329297e-07, "loss": 1.022, "step": 10757 }, { "epoch": 1.94, "learning_rate": 8.349900596421472e-07, "loss": 0.425, "step": 10758 }, { "epoch": 1.94, "learning_rate": 8.322790529549973e-07, "loss": 0.3351, "step": 10759 }, { "epoch": 1.94, "learning_rate": 8.295680462678476e-07, "loss": 0.7003, "step": 10760 }, { "epoch": 1.94, "learning_rate": 8.268570395806976e-07, "loss": 0.4509, "step": 10761 }, { "epoch": 1.95, "learning_rate": 8.241460328935477e-07, "loss": 0.4867, "step": 10762 }, { "epoch": 1.95, "learning_rate": 8.21435026206398e-07, "loss": 0.9975, "step": 10763 }, { "epoch": 1.95, "learning_rate": 8.187240195192481e-07, "loss": 0.9777, "step": 10764 }, { "epoch": 1.95, "learning_rate": 8.160130128320983e-07, "loss": 0.2929, "step": 10765 }, { "epoch": 1.95, "learning_rate": 8.133020061449485e-07, "loss": 0.8419, "step": 10766 }, { "epoch": 1.95, "learning_rate": 8.105909994577986e-07, "loss": 0.6973, "step": 10767 }, { "epoch": 1.95, "learning_rate": 8.078799927706489e-07, "loss": 0.5077, "step": 10768 }, { "epoch": 1.95, "learning_rate": 8.05168986083499e-07, "loss": 1.0246, "step": 10769 }, { "epoch": 1.95, "learning_rate": 8.024579793963491e-07, "loss": 0.9928, "step": 10770 }, { "epoch": 1.95, "learning_rate": 7.997469727091994e-07, "loss": 1.2263, "step": 10771 }, { "epoch": 1.95, "learning_rate": 7.970359660220495e-07, "loss": 0.468, "step": 10772 }, { "epoch": 1.95, "learning_rate": 7.943249593348998e-07, "loss": 0.634, "step": 10773 }, { "epoch": 1.95, "learning_rate": 7.916139526477499e-07, "loss": 1.1721, "step": 10774 }, { "epoch": 1.95, "learning_rate": 7.889029459606e-07, "loss": 0.9969, "step": 10775 }, { "epoch": 1.95, "learning_rate": 7.861919392734503e-07, "loss": 0.4106, "step": 10776 }, { "epoch": 1.95, "learning_rate": 7.834809325863004e-07, "loss": 0.7497, "step": 10777 }, { "epoch": 1.95, "learning_rate": 7.807699258991505e-07, "loss": 0.544, "step": 10778 }, { "epoch": 1.95, "learning_rate": 7.780589192120008e-07, "loss": 1.0381, "step": 10779 }, { "epoch": 1.95, "learning_rate": 7.753479125248509e-07, "loss": 0.5169, "step": 10780 }, { "epoch": 1.95, "learning_rate": 7.726369058377011e-07, "loss": 1.0806, "step": 10781 }, { "epoch": 1.95, "learning_rate": 7.699258991505513e-07, "loss": 0.4083, "step": 10782 }, { "epoch": 1.95, "learning_rate": 7.672148924634014e-07, "loss": 0.5238, "step": 10783 }, { "epoch": 1.95, "learning_rate": 7.645038857762517e-07, "loss": 0.7303, "step": 10784 }, { "epoch": 1.95, "learning_rate": 7.617928790891018e-07, "loss": 0.3904, "step": 10785 }, { "epoch": 1.95, "learning_rate": 7.59081872401952e-07, "loss": 0.7974, "step": 10786 }, { "epoch": 1.95, "learning_rate": 7.563708657148022e-07, "loss": 0.3397, "step": 10787 }, { "epoch": 1.95, "learning_rate": 7.536598590276523e-07, "loss": 1.1605, "step": 10788 }, { "epoch": 1.95, "learning_rate": 7.509488523405025e-07, "loss": 1.1229, "step": 10789 }, { "epoch": 1.95, "learning_rate": 7.482378456533527e-07, "loss": 0.2954, "step": 10790 }, { "epoch": 1.95, "learning_rate": 7.455268389662028e-07, "loss": 0.5715, "step": 10791 }, { "epoch": 1.95, "learning_rate": 7.42815832279053e-07, "loss": 0.6336, "step": 10792 }, { "epoch": 1.95, "learning_rate": 7.401048255919031e-07, "loss": 0.3827, "step": 10793 }, { "epoch": 1.95, "learning_rate": 7.373938189047533e-07, "loss": 0.6668, "step": 10794 }, { "epoch": 1.95, "learning_rate": 7.346828122176034e-07, "loss": 0.7688, "step": 10795 }, { "epoch": 1.95, "learning_rate": 7.319718055304536e-07, "loss": 0.6978, "step": 10796 }, { "epoch": 1.95, "learning_rate": 7.292607988433038e-07, "loss": 0.6805, "step": 10797 }, { "epoch": 1.95, "learning_rate": 7.26549792156154e-07, "loss": 0.5377, "step": 10798 }, { "epoch": 1.95, "learning_rate": 7.238387854690042e-07, "loss": 0.5236, "step": 10799 }, { "epoch": 1.95, "learning_rate": 7.211277787818543e-07, "loss": 1.3712, "step": 10800 }, { "epoch": 1.95, "learning_rate": 7.184167720947045e-07, "loss": 1.4037, "step": 10801 }, { "epoch": 1.95, "learning_rate": 7.157057654075547e-07, "loss": 0.4185, "step": 10802 }, { "epoch": 1.95, "learning_rate": 7.129947587204048e-07, "loss": 0.4329, "step": 10803 }, { "epoch": 1.95, "learning_rate": 7.10283752033255e-07, "loss": 0.9129, "step": 10804 }, { "epoch": 1.95, "learning_rate": 7.075727453461052e-07, "loss": 0.8111, "step": 10805 }, { "epoch": 1.95, "learning_rate": 7.048617386589553e-07, "loss": 0.5249, "step": 10806 }, { "epoch": 1.95, "learning_rate": 7.021507319718056e-07, "loss": 0.7029, "step": 10807 }, { "epoch": 1.95, "learning_rate": 6.994397252846557e-07, "loss": 1.2362, "step": 10808 }, { "epoch": 1.95, "learning_rate": 6.967287185975059e-07, "loss": 0.5102, "step": 10809 }, { "epoch": 1.95, "learning_rate": 6.940177119103561e-07, "loss": 0.9694, "step": 10810 }, { "epoch": 1.95, "learning_rate": 6.913067052232062e-07, "loss": 0.3643, "step": 10811 }, { "epoch": 1.95, "learning_rate": 6.885956985360564e-07, "loss": 0.3445, "step": 10812 }, { "epoch": 1.95, "learning_rate": 6.858846918489066e-07, "loss": 1.1187, "step": 10813 }, { "epoch": 1.95, "learning_rate": 6.831736851617567e-07, "loss": 1.2952, "step": 10814 }, { "epoch": 1.95, "learning_rate": 6.80462678474607e-07, "loss": 0.5129, "step": 10815 }, { "epoch": 1.95, "learning_rate": 6.777516717874571e-07, "loss": 1.026, "step": 10816 }, { "epoch": 1.95, "learning_rate": 6.750406651003072e-07, "loss": 0.5381, "step": 10817 }, { "epoch": 1.96, "learning_rate": 6.723296584131575e-07, "loss": 0.7493, "step": 10818 }, { "epoch": 1.96, "learning_rate": 6.696186517260076e-07, "loss": 0.6278, "step": 10819 }, { "epoch": 1.96, "learning_rate": 6.669076450388579e-07, "loss": 0.6154, "step": 10820 }, { "epoch": 1.96, "learning_rate": 6.64196638351708e-07, "loss": 0.481, "step": 10821 }, { "epoch": 1.96, "learning_rate": 6.614856316645581e-07, "loss": 0.4896, "step": 10822 }, { "epoch": 1.96, "learning_rate": 6.587746249774084e-07, "loss": 1.3428, "step": 10823 }, { "epoch": 1.96, "learning_rate": 6.560636182902585e-07, "loss": 1.2545, "step": 10824 }, { "epoch": 1.96, "learning_rate": 6.533526116031085e-07, "loss": 0.8438, "step": 10825 }, { "epoch": 1.96, "learning_rate": 6.506416049159588e-07, "loss": 0.7747, "step": 10826 }, { "epoch": 1.96, "learning_rate": 6.479305982288089e-07, "loss": 0.8123, "step": 10827 }, { "epoch": 1.96, "learning_rate": 6.452195915416591e-07, "loss": 0.6486, "step": 10828 }, { "epoch": 1.96, "learning_rate": 6.425085848545093e-07, "loss": 0.8435, "step": 10829 }, { "epoch": 1.96, "learning_rate": 6.397975781673594e-07, "loss": 0.8698, "step": 10830 }, { "epoch": 1.96, "learning_rate": 6.370865714802097e-07, "loss": 0.4283, "step": 10831 }, { "epoch": 1.96, "learning_rate": 6.343755647930598e-07, "loss": 0.5766, "step": 10832 }, { "epoch": 1.96, "learning_rate": 6.3166455810591e-07, "loss": 0.6586, "step": 10833 }, { "epoch": 1.96, "learning_rate": 6.289535514187602e-07, "loss": 0.4745, "step": 10834 }, { "epoch": 1.96, "learning_rate": 6.262425447316103e-07, "loss": 0.4739, "step": 10835 }, { "epoch": 1.96, "learning_rate": 6.235315380444605e-07, "loss": 0.5451, "step": 10836 }, { "epoch": 1.96, "learning_rate": 6.208205313573107e-07, "loss": 0.6707, "step": 10837 }, { "epoch": 1.96, "learning_rate": 6.181095246701608e-07, "loss": 0.3946, "step": 10838 }, { "epoch": 1.96, "learning_rate": 6.15398517983011e-07, "loss": 0.8898, "step": 10839 }, { "epoch": 1.96, "learning_rate": 6.126875112958612e-07, "loss": 0.7063, "step": 10840 }, { "epoch": 1.96, "learning_rate": 6.099765046087114e-07, "loss": 0.252, "step": 10841 }, { "epoch": 1.96, "learning_rate": 6.072654979215616e-07, "loss": 0.8446, "step": 10842 }, { "epoch": 1.96, "learning_rate": 6.045544912344117e-07, "loss": 0.6551, "step": 10843 }, { "epoch": 1.96, "learning_rate": 6.018434845472619e-07, "loss": 0.6257, "step": 10844 }, { "epoch": 1.96, "learning_rate": 5.991324778601121e-07, "loss": 0.6599, "step": 10845 }, { "epoch": 1.96, "learning_rate": 5.964214711729623e-07, "loss": 1.0033, "step": 10846 }, { "epoch": 1.96, "learning_rate": 5.937104644858124e-07, "loss": 0.748, "step": 10847 }, { "epoch": 1.96, "learning_rate": 5.909994577986626e-07, "loss": 0.4513, "step": 10848 }, { "epoch": 1.96, "learning_rate": 5.882884511115128e-07, "loss": 1.0466, "step": 10849 }, { "epoch": 1.96, "learning_rate": 5.85577444424363e-07, "loss": 0.4233, "step": 10850 }, { "epoch": 1.96, "learning_rate": 5.828664377372131e-07, "loss": 0.3063, "step": 10851 }, { "epoch": 1.96, "learning_rate": 5.801554310500633e-07, "loss": 0.9016, "step": 10852 }, { "epoch": 1.96, "learning_rate": 5.774444243629135e-07, "loss": 0.7211, "step": 10853 }, { "epoch": 1.96, "learning_rate": 5.747334176757637e-07, "loss": 0.7258, "step": 10854 }, { "epoch": 1.96, "learning_rate": 5.720224109886138e-07, "loss": 1.3917, "step": 10855 }, { "epoch": 1.96, "learning_rate": 5.69311404301464e-07, "loss": 0.546, "step": 10856 }, { "epoch": 1.96, "learning_rate": 5.666003976143141e-07, "loss": 1.1886, "step": 10857 }, { "epoch": 1.96, "learning_rate": 5.638893909271642e-07, "loss": 0.6637, "step": 10858 }, { "epoch": 1.96, "learning_rate": 5.611783842400145e-07, "loss": 0.7202, "step": 10859 }, { "epoch": 1.96, "learning_rate": 5.584673775528646e-07, "loss": 0.3564, "step": 10860 }, { "epoch": 1.96, "learning_rate": 5.557563708657147e-07, "loss": 0.5877, "step": 10861 }, { "epoch": 1.96, "learning_rate": 5.53045364178565e-07, "loss": 0.6769, "step": 10862 }, { "epoch": 1.96, "learning_rate": 5.503343574914151e-07, "loss": 0.8955, "step": 10863 }, { "epoch": 1.96, "learning_rate": 5.476233508042653e-07, "loss": 1.0589, "step": 10864 }, { "epoch": 1.96, "learning_rate": 5.449123441171155e-07, "loss": 0.8337, "step": 10865 }, { "epoch": 1.96, "learning_rate": 5.422013374299656e-07, "loss": 0.6903, "step": 10866 }, { "epoch": 1.96, "learning_rate": 5.394903307428159e-07, "loss": 0.3602, "step": 10867 }, { "epoch": 1.96, "learning_rate": 5.36779324055666e-07, "loss": 0.522, "step": 10868 }, { "epoch": 1.96, "learning_rate": 5.340683173685161e-07, "loss": 0.5718, "step": 10869 }, { "epoch": 1.96, "learning_rate": 5.313573106813664e-07, "loss": 0.6319, "step": 10870 }, { "epoch": 1.96, "learning_rate": 5.286463039942165e-07, "loss": 1.0768, "step": 10871 }, { "epoch": 1.96, "learning_rate": 5.259352973070668e-07, "loss": 0.4094, "step": 10872 }, { "epoch": 1.97, "learning_rate": 5.232242906199169e-07, "loss": 1.0955, "step": 10873 }, { "epoch": 1.97, "learning_rate": 5.20513283932767e-07, "loss": 0.4571, "step": 10874 }, { "epoch": 1.97, "learning_rate": 5.178022772456173e-07, "loss": 0.4865, "step": 10875 }, { "epoch": 1.97, "learning_rate": 5.150912705584674e-07, "loss": 0.6409, "step": 10876 }, { "epoch": 1.97, "learning_rate": 5.123802638713175e-07, "loss": 0.8702, "step": 10877 }, { "epoch": 1.97, "learning_rate": 5.096692571841678e-07, "loss": 0.6902, "step": 10878 }, { "epoch": 1.97, "learning_rate": 5.069582504970179e-07, "loss": 0.4931, "step": 10879 }, { "epoch": 1.97, "learning_rate": 5.042472438098681e-07, "loss": 0.3086, "step": 10880 }, { "epoch": 1.97, "learning_rate": 5.015362371227183e-07, "loss": 0.839, "step": 10881 }, { "epoch": 1.97, "learning_rate": 4.988252304355684e-07, "loss": 0.4937, "step": 10882 }, { "epoch": 1.97, "learning_rate": 4.961142237484187e-07, "loss": 0.6506, "step": 10883 }, { "epoch": 1.97, "learning_rate": 4.934032170612688e-07, "loss": 1.1031, "step": 10884 }, { "epoch": 1.97, "learning_rate": 4.90692210374119e-07, "loss": 0.645, "step": 10885 }, { "epoch": 1.97, "learning_rate": 4.879812036869692e-07, "loss": 0.5629, "step": 10886 }, { "epoch": 1.97, "learning_rate": 4.852701969998193e-07, "loss": 0.3914, "step": 10887 }, { "epoch": 1.97, "learning_rate": 4.825591903126695e-07, "loss": 0.7342, "step": 10888 }, { "epoch": 1.97, "learning_rate": 4.798481836255197e-07, "loss": 1.1886, "step": 10889 }, { "epoch": 1.97, "learning_rate": 4.771371769383697e-07, "loss": 0.4077, "step": 10890 }, { "epoch": 1.97, "learning_rate": 4.7442617025121994e-07, "loss": 1.1107, "step": 10891 }, { "epoch": 1.97, "learning_rate": 4.7171516356407007e-07, "loss": 0.6388, "step": 10892 }, { "epoch": 1.97, "learning_rate": 4.6900415687692026e-07, "loss": 0.4615, "step": 10893 }, { "epoch": 1.97, "learning_rate": 4.662931501897705e-07, "loss": 0.4432, "step": 10894 }, { "epoch": 1.97, "learning_rate": 4.635821435026207e-07, "loss": 0.7286, "step": 10895 }, { "epoch": 1.97, "learning_rate": 4.608711368154708e-07, "loss": 0.464, "step": 10896 }, { "epoch": 1.97, "learning_rate": 4.58160130128321e-07, "loss": 0.5851, "step": 10897 }, { "epoch": 1.97, "learning_rate": 4.5544912344117114e-07, "loss": 0.6786, "step": 10898 }, { "epoch": 1.97, "learning_rate": 4.5273811675402133e-07, "loss": 0.503, "step": 10899 }, { "epoch": 1.97, "learning_rate": 4.500271100668715e-07, "loss": 0.5892, "step": 10900 }, { "epoch": 1.97, "learning_rate": 4.4731610337972165e-07, "loss": 0.7091, "step": 10901 }, { "epoch": 1.97, "learning_rate": 4.4460509669257184e-07, "loss": 0.7288, "step": 10902 }, { "epoch": 1.97, "learning_rate": 4.41894090005422e-07, "loss": 1.4084, "step": 10903 }, { "epoch": 1.97, "learning_rate": 4.391830833182722e-07, "loss": 0.4794, "step": 10904 }, { "epoch": 1.97, "learning_rate": 4.3647207663112235e-07, "loss": 0.8416, "step": 10905 }, { "epoch": 1.97, "learning_rate": 4.3376106994397253e-07, "loss": 0.9282, "step": 10906 }, { "epoch": 1.97, "learning_rate": 4.310500632568227e-07, "loss": 0.6303, "step": 10907 }, { "epoch": 1.97, "learning_rate": 4.283390565696729e-07, "loss": 0.5357, "step": 10908 }, { "epoch": 1.97, "learning_rate": 4.2562804988252304e-07, "loss": 0.6424, "step": 10909 }, { "epoch": 1.97, "learning_rate": 4.2291704319537323e-07, "loss": 1.3173, "step": 10910 }, { "epoch": 1.97, "learning_rate": 4.202060365082234e-07, "loss": 0.5124, "step": 10911 }, { "epoch": 1.97, "learning_rate": 4.174950298210736e-07, "loss": 0.5378, "step": 10912 }, { "epoch": 1.97, "learning_rate": 4.147840231339238e-07, "loss": 1.0757, "step": 10913 }, { "epoch": 1.97, "learning_rate": 4.1207301644677387e-07, "loss": 0.5583, "step": 10914 }, { "epoch": 1.97, "learning_rate": 4.0936200975962406e-07, "loss": 1.1384, "step": 10915 }, { "epoch": 1.97, "learning_rate": 4.0665100307247425e-07, "loss": 0.5838, "step": 10916 }, { "epoch": 1.97, "learning_rate": 4.0393999638532443e-07, "loss": 0.548, "step": 10917 }, { "epoch": 1.97, "learning_rate": 4.0122898969817457e-07, "loss": 0.7062, "step": 10918 }, { "epoch": 1.97, "learning_rate": 3.9851798301102476e-07, "loss": 0.7535, "step": 10919 }, { "epoch": 1.97, "learning_rate": 3.9580697632387494e-07, "loss": 1.2658, "step": 10920 }, { "epoch": 1.97, "learning_rate": 3.9309596963672513e-07, "loss": 0.4446, "step": 10921 }, { "epoch": 1.97, "learning_rate": 3.9038496294957526e-07, "loss": 0.7511, "step": 10922 }, { "epoch": 1.97, "learning_rate": 3.8767395626242545e-07, "loss": 0.1788, "step": 10923 }, { "epoch": 1.97, "learning_rate": 3.8496294957527564e-07, "loss": 0.4641, "step": 10924 }, { "epoch": 1.97, "learning_rate": 3.822519428881258e-07, "loss": 0.4811, "step": 10925 }, { "epoch": 1.97, "learning_rate": 3.79540936200976e-07, "loss": 0.647, "step": 10926 }, { "epoch": 1.97, "learning_rate": 3.7682992951382615e-07, "loss": 0.3885, "step": 10927 }, { "epoch": 1.98, "learning_rate": 3.7411892282667633e-07, "loss": 0.5655, "step": 10928 }, { "epoch": 1.98, "learning_rate": 3.714079161395265e-07, "loss": 0.7734, "step": 10929 }, { "epoch": 1.98, "learning_rate": 3.6869690945237666e-07, "loss": 0.8253, "step": 10930 }, { "epoch": 1.98, "learning_rate": 3.659859027652268e-07, "loss": 0.9247, "step": 10931 }, { "epoch": 1.98, "learning_rate": 3.63274896078077e-07, "loss": 1.1011, "step": 10932 }, { "epoch": 1.98, "learning_rate": 3.6056388939092717e-07, "loss": 0.6545, "step": 10933 }, { "epoch": 1.98, "learning_rate": 3.5785288270377735e-07, "loss": 0.2961, "step": 10934 }, { "epoch": 1.98, "learning_rate": 3.551418760166275e-07, "loss": 0.5024, "step": 10935 }, { "epoch": 1.98, "learning_rate": 3.524308693294777e-07, "loss": 1.1048, "step": 10936 }, { "epoch": 1.98, "learning_rate": 3.4971986264232786e-07, "loss": 0.6476, "step": 10937 }, { "epoch": 1.98, "learning_rate": 3.4700885595517805e-07, "loss": 0.2974, "step": 10938 }, { "epoch": 1.98, "learning_rate": 3.442978492680282e-07, "loss": 0.7423, "step": 10939 }, { "epoch": 1.98, "learning_rate": 3.4158684258087837e-07, "loss": 0.7019, "step": 10940 }, { "epoch": 1.98, "learning_rate": 3.3887583589372856e-07, "loss": 0.654, "step": 10941 }, { "epoch": 1.98, "learning_rate": 3.3616482920657874e-07, "loss": 1.0116, "step": 10942 }, { "epoch": 1.98, "learning_rate": 3.3345382251942893e-07, "loss": 0.8012, "step": 10943 }, { "epoch": 1.98, "learning_rate": 3.3074281583227907e-07, "loss": 0.6982, "step": 10944 }, { "epoch": 1.98, "learning_rate": 3.2803180914512925e-07, "loss": 0.8199, "step": 10945 }, { "epoch": 1.98, "learning_rate": 3.253208024579794e-07, "loss": 0.7294, "step": 10946 }, { "epoch": 1.98, "learning_rate": 3.226097957708296e-07, "loss": 0.3557, "step": 10947 }, { "epoch": 1.98, "learning_rate": 3.198987890836797e-07, "loss": 0.413, "step": 10948 }, { "epoch": 1.98, "learning_rate": 3.171877823965299e-07, "loss": 0.3363, "step": 10949 }, { "epoch": 1.98, "learning_rate": 3.144767757093801e-07, "loss": 0.6272, "step": 10950 }, { "epoch": 1.98, "learning_rate": 3.1176576902223027e-07, "loss": 0.8527, "step": 10951 }, { "epoch": 1.98, "learning_rate": 3.090547623350804e-07, "loss": 0.235, "step": 10952 }, { "epoch": 1.98, "learning_rate": 3.063437556479306e-07, "loss": 0.5767, "step": 10953 }, { "epoch": 1.98, "learning_rate": 3.036327489607808e-07, "loss": 0.5585, "step": 10954 }, { "epoch": 1.98, "learning_rate": 3.0092174227363097e-07, "loss": 0.4596, "step": 10955 }, { "epoch": 1.98, "learning_rate": 2.9821073558648115e-07, "loss": 0.9205, "step": 10956 }, { "epoch": 1.98, "learning_rate": 2.954997288993313e-07, "loss": 0.9336, "step": 10957 }, { "epoch": 1.98, "learning_rate": 2.927887222121815e-07, "loss": 0.8773, "step": 10958 }, { "epoch": 1.98, "learning_rate": 2.9007771552503166e-07, "loss": 0.3223, "step": 10959 }, { "epoch": 1.98, "learning_rate": 2.8736670883788185e-07, "loss": 0.719, "step": 10960 }, { "epoch": 1.98, "learning_rate": 2.84655702150732e-07, "loss": 0.7092, "step": 10961 }, { "epoch": 1.98, "learning_rate": 2.819446954635821e-07, "loss": 0.8337, "step": 10962 }, { "epoch": 1.98, "learning_rate": 2.792336887764323e-07, "loss": 0.6949, "step": 10963 }, { "epoch": 1.98, "learning_rate": 2.765226820892825e-07, "loss": 0.851, "step": 10964 }, { "epoch": 1.98, "learning_rate": 2.7381167540213263e-07, "loss": 0.8821, "step": 10965 }, { "epoch": 1.98, "learning_rate": 2.711006687149828e-07, "loss": 0.9523, "step": 10966 }, { "epoch": 1.98, "learning_rate": 2.68389662027833e-07, "loss": 0.7278, "step": 10967 }, { "epoch": 1.98, "learning_rate": 2.656786553406832e-07, "loss": 0.6208, "step": 10968 }, { "epoch": 1.98, "learning_rate": 2.629676486535334e-07, "loss": 0.5151, "step": 10969 }, { "epoch": 1.98, "learning_rate": 2.602566419663835e-07, "loss": 1.0197, "step": 10970 }, { "epoch": 1.98, "learning_rate": 2.575456352792337e-07, "loss": 0.6817, "step": 10971 }, { "epoch": 1.98, "learning_rate": 2.548346285920839e-07, "loss": 0.6522, "step": 10972 }, { "epoch": 1.98, "learning_rate": 2.5212362190493407e-07, "loss": 0.5543, "step": 10973 }, { "epoch": 1.98, "learning_rate": 2.494126152177842e-07, "loss": 0.6465, "step": 10974 }, { "epoch": 1.98, "learning_rate": 2.467016085306344e-07, "loss": 0.7342, "step": 10975 }, { "epoch": 1.98, "learning_rate": 2.439906018434846e-07, "loss": 0.9579, "step": 10976 }, { "epoch": 1.98, "learning_rate": 2.4127959515633477e-07, "loss": 0.5579, "step": 10977 }, { "epoch": 1.98, "learning_rate": 2.3856858846918485e-07, "loss": 1.0729, "step": 10978 }, { "epoch": 1.98, "learning_rate": 2.3585758178203504e-07, "loss": 1.0067, "step": 10979 }, { "epoch": 1.98, "learning_rate": 2.3314657509488525e-07, "loss": 0.795, "step": 10980 }, { "epoch": 1.98, "learning_rate": 2.304355684077354e-07, "loss": 0.8886, "step": 10981 }, { "epoch": 1.98, "learning_rate": 2.2772456172058557e-07, "loss": 0.7426, "step": 10982 }, { "epoch": 1.98, "learning_rate": 2.2501355503343576e-07, "loss": 0.6107, "step": 10983 }, { "epoch": 1.99, "learning_rate": 2.2230254834628592e-07, "loss": 0.5422, "step": 10984 }, { "epoch": 1.99, "learning_rate": 2.195915416591361e-07, "loss": 0.6236, "step": 10985 }, { "epoch": 1.99, "learning_rate": 2.1688053497198627e-07, "loss": 0.6775, "step": 10986 }, { "epoch": 1.99, "learning_rate": 2.1416952828483645e-07, "loss": 0.3561, "step": 10987 }, { "epoch": 1.99, "learning_rate": 2.1145852159768661e-07, "loss": 0.3834, "step": 10988 }, { "epoch": 1.99, "learning_rate": 2.087475149105368e-07, "loss": 0.5268, "step": 10989 }, { "epoch": 1.99, "learning_rate": 2.0603650822338694e-07, "loss": 0.4498, "step": 10990 }, { "epoch": 1.99, "learning_rate": 2.0332550153623712e-07, "loss": 1.0319, "step": 10991 }, { "epoch": 1.99, "learning_rate": 2.0061449484908728e-07, "loss": 0.4949, "step": 10992 }, { "epoch": 1.99, "learning_rate": 1.9790348816193747e-07, "loss": 0.7002, "step": 10993 }, { "epoch": 1.99, "learning_rate": 1.9519248147478763e-07, "loss": 0.3424, "step": 10994 }, { "epoch": 1.99, "learning_rate": 1.9248147478763782e-07, "loss": 0.4705, "step": 10995 }, { "epoch": 1.99, "learning_rate": 1.89770468100488e-07, "loss": 0.256, "step": 10996 }, { "epoch": 1.99, "learning_rate": 1.8705946141333817e-07, "loss": 0.5019, "step": 10997 }, { "epoch": 1.99, "learning_rate": 1.8434845472618833e-07, "loss": 1.2005, "step": 10998 }, { "epoch": 1.99, "learning_rate": 1.816374480390385e-07, "loss": 0.9312, "step": 10999 }, { "epoch": 1.99, "learning_rate": 1.7892644135188868e-07, "loss": 0.731, "step": 11000 }, { "epoch": 1.99, "learning_rate": 1.7621543466473884e-07, "loss": 1.029, "step": 11001 }, { "epoch": 1.99, "learning_rate": 1.7350442797758902e-07, "loss": 0.6349, "step": 11002 }, { "epoch": 1.99, "learning_rate": 1.7079342129043918e-07, "loss": 0.5103, "step": 11003 }, { "epoch": 1.99, "learning_rate": 1.6808241460328937e-07, "loss": 0.9348, "step": 11004 }, { "epoch": 1.99, "learning_rate": 1.6537140791613953e-07, "loss": 0.7165, "step": 11005 }, { "epoch": 1.99, "learning_rate": 1.626604012289897e-07, "loss": 0.7424, "step": 11006 }, { "epoch": 1.99, "learning_rate": 1.5994939454183985e-07, "loss": 0.3513, "step": 11007 }, { "epoch": 1.99, "learning_rate": 1.5723838785469004e-07, "loss": 0.8015, "step": 11008 }, { "epoch": 1.99, "learning_rate": 1.545273811675402e-07, "loss": 0.905, "step": 11009 }, { "epoch": 1.99, "learning_rate": 1.518163744803904e-07, "loss": 1.167, "step": 11010 }, { "epoch": 1.99, "learning_rate": 1.4910536779324058e-07, "loss": 0.6511, "step": 11011 }, { "epoch": 1.99, "learning_rate": 1.4639436110609074e-07, "loss": 0.5468, "step": 11012 }, { "epoch": 1.99, "learning_rate": 1.4368335441894092e-07, "loss": 0.4404, "step": 11013 }, { "epoch": 1.99, "learning_rate": 1.4097234773179106e-07, "loss": 0.9051, "step": 11014 }, { "epoch": 1.99, "learning_rate": 1.3826134104464125e-07, "loss": 0.5647, "step": 11015 }, { "epoch": 1.99, "learning_rate": 1.355503343574914e-07, "loss": 0.6354, "step": 11016 }, { "epoch": 1.99, "learning_rate": 1.328393276703416e-07, "loss": 0.6661, "step": 11017 }, { "epoch": 1.99, "learning_rate": 1.3012832098319175e-07, "loss": 1.0375, "step": 11018 }, { "epoch": 1.99, "learning_rate": 1.2741731429604194e-07, "loss": 0.6944, "step": 11019 }, { "epoch": 1.99, "learning_rate": 1.247063076088921e-07, "loss": 0.38, "step": 11020 }, { "epoch": 1.99, "learning_rate": 1.219953009217423e-07, "loss": 0.6036, "step": 11021 }, { "epoch": 1.99, "learning_rate": 1.1928429423459242e-07, "loss": 1.2028, "step": 11022 }, { "epoch": 1.99, "learning_rate": 1.1657328754744262e-07, "loss": 1.0183, "step": 11023 }, { "epoch": 1.99, "learning_rate": 1.1386228086029279e-07, "loss": 0.6078, "step": 11024 }, { "epoch": 1.99, "learning_rate": 1.1115127417314296e-07, "loss": 0.474, "step": 11025 }, { "epoch": 1.99, "learning_rate": 1.0844026748599313e-07, "loss": 0.9146, "step": 11026 }, { "epoch": 1.99, "learning_rate": 1.0572926079884331e-07, "loss": 0.9846, "step": 11027 }, { "epoch": 1.99, "learning_rate": 1.0301825411169347e-07, "loss": 0.5561, "step": 11028 }, { "epoch": 1.99, "learning_rate": 1.0030724742454364e-07, "loss": 0.7157, "step": 11029 }, { "epoch": 1.99, "learning_rate": 9.759624073739382e-08, "loss": 0.7976, "step": 11030 }, { "epoch": 1.99, "learning_rate": 9.4885234050244e-08, "loss": 0.6179, "step": 11031 }, { "epoch": 1.99, "learning_rate": 9.217422736309416e-08, "loss": 0.7891, "step": 11032 }, { "epoch": 1.99, "learning_rate": 8.946322067594434e-08, "loss": 0.5762, "step": 11033 }, { "epoch": 1.99, "learning_rate": 8.675221398879451e-08, "loss": 0.471, "step": 11034 }, { "epoch": 1.99, "learning_rate": 8.404120730164469e-08, "loss": 0.4826, "step": 11035 }, { "epoch": 1.99, "learning_rate": 8.133020061449485e-08, "loss": 0.7229, "step": 11036 }, { "epoch": 1.99, "learning_rate": 7.861919392734502e-08, "loss": 0.3919, "step": 11037 }, { "epoch": 1.99, "learning_rate": 7.59081872401952e-08, "loss": 0.2105, "step": 11038 }, { "epoch": 2.0, "learning_rate": 7.319718055304537e-08, "loss": 0.6759, "step": 11039 }, { "epoch": 2.0, "learning_rate": 7.048617386589553e-08, "loss": 0.5047, "step": 11040 }, { "epoch": 2.0, "learning_rate": 6.77751671787457e-08, "loss": 0.8142, "step": 11041 }, { "epoch": 2.0, "learning_rate": 6.506416049159588e-08, "loss": 0.4116, "step": 11042 }, { "epoch": 2.0, "learning_rate": 6.235315380444605e-08, "loss": 0.9319, "step": 11043 }, { "epoch": 2.0, "learning_rate": 5.964214711729621e-08, "loss": 0.5246, "step": 11044 }, { "epoch": 2.0, "learning_rate": 5.693114043014639e-08, "loss": 0.9541, "step": 11045 }, { "epoch": 2.0, "learning_rate": 5.422013374299657e-08, "loss": 0.8759, "step": 11046 }, { "epoch": 2.0, "learning_rate": 5.1509127055846734e-08, "loss": 0.6786, "step": 11047 }, { "epoch": 2.0, "learning_rate": 4.879812036869691e-08, "loss": 0.7232, "step": 11048 }, { "epoch": 2.0, "learning_rate": 4.608711368154708e-08, "loss": 0.1013, "step": 11049 }, { "epoch": 2.0, "learning_rate": 4.3376106994397256e-08, "loss": 1.202, "step": 11050 }, { "epoch": 2.0, "learning_rate": 4.0665100307247423e-08, "loss": 0.686, "step": 11051 }, { "epoch": 2.0, "learning_rate": 3.79540936200976e-08, "loss": 1.0953, "step": 11052 }, { "epoch": 2.0, "learning_rate": 3.5243086932947765e-08, "loss": 0.4561, "step": 11053 }, { "epoch": 2.0, "learning_rate": 3.253208024579794e-08, "loss": 0.4277, "step": 11054 }, { "epoch": 2.0, "learning_rate": 2.9821073558648106e-08, "loss": 0.1749, "step": 11055 }, { "epoch": 2.0, "learning_rate": 2.7110066871498283e-08, "loss": 0.6099, "step": 11056 }, { "epoch": 2.0, "learning_rate": 2.4399060184348454e-08, "loss": 0.5117, "step": 11057 }, { "epoch": 2.0, "learning_rate": 2.1688053497198628e-08, "loss": 0.5277, "step": 11058 }, { "epoch": 2.0, "learning_rate": 1.89770468100488e-08, "loss": 1.204, "step": 11059 }, { "epoch": 2.0, "learning_rate": 1.626604012289897e-08, "loss": 0.4059, "step": 11060 }, { "epoch": 2.0, "learning_rate": 1.3555033435749142e-08, "loss": 0.8129, "step": 11061 }, { "epoch": 2.0, "learning_rate": 1.0844026748599314e-08, "loss": 0.474, "step": 11062 }, { "epoch": 2.0, "learning_rate": 8.133020061449485e-09, "loss": 0.7607, "step": 11063 }, { "epoch": 2.0, "learning_rate": 5.422013374299657e-09, "loss": 0.5483, "step": 11064 }, { "epoch": 2.0, "learning_rate": 2.7110066871498285e-09, "loss": 0.3523, "step": 11065 }, { "epoch": 2.0, "learning_rate": 0.0, "loss": 0.6686, "step": 11066 }, { "epoch": 2.0, "step": 11066, "total_flos": 3.4696551139946496e+16, "train_loss": 1.0129275434533833, "train_runtime": 2434.5113, "train_samples_per_second": 72.724, "train_steps_per_second": 4.545 } ], "max_steps": 11066, "num_train_epochs": 2, "total_flos": 3.4696551139946496e+16, "trial_name": null, "trial_params": null }