{ "best_metric": null, "best_model_checkpoint": null, "epoch": 3.0, "global_step": 25191, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 6.613756613756614e-09, "loss": 1.3086, "step": 1 }, { "epoch": 0.0, "learning_rate": 1.3227513227513228e-08, "loss": 1.4039, "step": 2 }, { "epoch": 0.0, "learning_rate": 1.9841269841269843e-08, "loss": 1.4438, "step": 3 }, { "epoch": 0.0, "learning_rate": 2.6455026455026456e-08, "loss": 1.4044, "step": 4 }, { "epoch": 0.0, "learning_rate": 3.306878306878307e-08, "loss": 1.5134, "step": 5 }, { "epoch": 0.0, "learning_rate": 3.9682539682539686e-08, "loss": 2.2069, "step": 6 }, { "epoch": 0.0, "learning_rate": 4.6296296296296295e-08, "loss": 1.286, "step": 7 }, { "epoch": 0.0, "learning_rate": 5.291005291005291e-08, "loss": 1.4874, "step": 8 }, { "epoch": 0.0, "learning_rate": 5.952380952380953e-08, "loss": 1.6469, "step": 9 }, { "epoch": 0.0, "learning_rate": 6.613756613756614e-08, "loss": 1.4747, "step": 10 }, { "epoch": 0.0, "learning_rate": 7.275132275132275e-08, "loss": 1.6267, "step": 11 }, { "epoch": 0.0, "learning_rate": 7.936507936507937e-08, "loss": 1.8492, "step": 12 }, { "epoch": 0.0, "learning_rate": 8.597883597883599e-08, "loss": 1.239, "step": 13 }, { "epoch": 0.0, "learning_rate": 9.259259259259259e-08, "loss": 1.4817, "step": 14 }, { "epoch": 0.0, "learning_rate": 9.920634920634921e-08, "loss": 1.3354, "step": 15 }, { "epoch": 0.0, "learning_rate": 1.0582010582010582e-07, "loss": 1.2453, "step": 16 }, { "epoch": 0.0, "learning_rate": 1.1243386243386244e-07, "loss": 1.5055, "step": 17 }, { "epoch": 0.0, "learning_rate": 1.1904761904761906e-07, "loss": 1.843, "step": 18 }, { "epoch": 0.0, "learning_rate": 1.2566137566137567e-07, "loss": 1.4436, "step": 19 }, { "epoch": 0.0, "learning_rate": 1.3227513227513228e-07, "loss": 1.1088, "step": 20 }, { "epoch": 0.0, "learning_rate": 1.3888888888888888e-07, "loss": 1.2748, "step": 21 }, { "epoch": 0.0, "learning_rate": 1.455026455026455e-07, "loss": 1.3121, "step": 22 }, { "epoch": 0.0, "learning_rate": 1.521164021164021e-07, "loss": 1.3884, "step": 23 }, { "epoch": 0.0, "learning_rate": 1.5873015873015874e-07, "loss": 1.5117, "step": 24 }, { "epoch": 0.0, "learning_rate": 1.6534391534391535e-07, "loss": 2.3672, "step": 25 }, { "epoch": 0.0, "learning_rate": 1.7195767195767198e-07, "loss": 1.2078, "step": 26 }, { "epoch": 0.0, "learning_rate": 1.7857142857142858e-07, "loss": 1.4369, "step": 27 }, { "epoch": 0.0, "learning_rate": 1.8518518518518518e-07, "loss": 1.2445, "step": 28 }, { "epoch": 0.0, "learning_rate": 1.917989417989418e-07, "loss": 1.5161, "step": 29 }, { "epoch": 0.0, "learning_rate": 1.9841269841269841e-07, "loss": 1.5007, "step": 30 }, { "epoch": 0.0, "learning_rate": 2.0502645502645504e-07, "loss": 1.9113, "step": 31 }, { "epoch": 0.0, "learning_rate": 2.1164021164021165e-07, "loss": 0.8008, "step": 32 }, { "epoch": 0.0, "learning_rate": 2.1825396825396825e-07, "loss": 1.1016, "step": 33 }, { "epoch": 0.0, "learning_rate": 2.2486772486772488e-07, "loss": 1.4125, "step": 34 }, { "epoch": 0.0, "learning_rate": 2.3148148148148148e-07, "loss": 1.2676, "step": 35 }, { "epoch": 0.0, "learning_rate": 2.3809523809523811e-07, "loss": 1.2257, "step": 36 }, { "epoch": 0.0, "learning_rate": 2.447089947089947e-07, "loss": 1.494, "step": 37 }, { "epoch": 0.0, "learning_rate": 2.5132275132275135e-07, "loss": 1.0554, "step": 38 }, { "epoch": 0.0, "learning_rate": 2.579365079365079e-07, "loss": 1.1332, "step": 39 }, { "epoch": 0.0, "learning_rate": 2.6455026455026455e-07, "loss": 1.0489, "step": 40 }, { "epoch": 0.0, "learning_rate": 2.711640211640212e-07, "loss": 1.0754, "step": 41 }, { "epoch": 0.01, "learning_rate": 2.7777777777777776e-07, "loss": 1.2266, "step": 42 }, { "epoch": 0.01, "learning_rate": 2.843915343915344e-07, "loss": 1.3672, "step": 43 }, { "epoch": 0.01, "learning_rate": 2.91005291005291e-07, "loss": 1.0209, "step": 44 }, { "epoch": 0.01, "learning_rate": 2.9761904761904765e-07, "loss": 1.0242, "step": 45 }, { "epoch": 0.01, "learning_rate": 3.042328042328042e-07, "loss": 1.0129, "step": 46 }, { "epoch": 0.01, "learning_rate": 3.1084656084656085e-07, "loss": 1.0109, "step": 47 }, { "epoch": 0.01, "learning_rate": 3.174603174603175e-07, "loss": 1.1333, "step": 48 }, { "epoch": 0.01, "learning_rate": 3.2407407407407406e-07, "loss": 1.3307, "step": 49 }, { "epoch": 0.01, "learning_rate": 3.306878306878307e-07, "loss": 1.5486, "step": 50 }, { "epoch": 0.01, "learning_rate": 3.373015873015873e-07, "loss": 0.897, "step": 51 }, { "epoch": 0.01, "learning_rate": 3.4391534391534395e-07, "loss": 0.9816, "step": 52 }, { "epoch": 0.01, "learning_rate": 3.5052910052910053e-07, "loss": 0.9706, "step": 53 }, { "epoch": 0.01, "learning_rate": 3.5714285714285716e-07, "loss": 1.0027, "step": 54 }, { "epoch": 0.01, "learning_rate": 3.637566137566138e-07, "loss": 1.0247, "step": 55 }, { "epoch": 0.01, "learning_rate": 3.7037037037037036e-07, "loss": 1.5865, "step": 56 }, { "epoch": 0.01, "learning_rate": 3.76984126984127e-07, "loss": 0.9071, "step": 57 }, { "epoch": 0.01, "learning_rate": 3.835978835978836e-07, "loss": 0.7808, "step": 58 }, { "epoch": 0.01, "learning_rate": 3.902116402116402e-07, "loss": 0.8423, "step": 59 }, { "epoch": 0.01, "learning_rate": 3.9682539682539683e-07, "loss": 1.0032, "step": 60 }, { "epoch": 0.01, "learning_rate": 4.0343915343915346e-07, "loss": 1.2943, "step": 61 }, { "epoch": 0.01, "learning_rate": 4.100529100529101e-07, "loss": 1.298, "step": 62 }, { "epoch": 0.01, "learning_rate": 4.1666666666666667e-07, "loss": 0.7518, "step": 63 }, { "epoch": 0.01, "learning_rate": 4.232804232804233e-07, "loss": 0.7103, "step": 64 }, { "epoch": 0.01, "learning_rate": 4.298941798941799e-07, "loss": 1.0547, "step": 65 }, { "epoch": 0.01, "learning_rate": 4.365079365079365e-07, "loss": 0.9, "step": 66 }, { "epoch": 0.01, "learning_rate": 4.4312169312169313e-07, "loss": 0.9746, "step": 67 }, { "epoch": 0.01, "learning_rate": 4.4973544973544976e-07, "loss": 1.1663, "step": 68 }, { "epoch": 0.01, "learning_rate": 4.5634920634920634e-07, "loss": 0.8976, "step": 69 }, { "epoch": 0.01, "learning_rate": 4.6296296296296297e-07, "loss": 0.7256, "step": 70 }, { "epoch": 0.01, "learning_rate": 4.695767195767196e-07, "loss": 0.7661, "step": 71 }, { "epoch": 0.01, "learning_rate": 4.7619047619047623e-07, "loss": 0.8931, "step": 72 }, { "epoch": 0.01, "learning_rate": 4.828042328042329e-07, "loss": 0.8812, "step": 73 }, { "epoch": 0.01, "learning_rate": 4.894179894179894e-07, "loss": 1.2401, "step": 74 }, { "epoch": 0.01, "learning_rate": 4.96031746031746e-07, "loss": 1.6117, "step": 75 }, { "epoch": 0.01, "learning_rate": 5.026455026455027e-07, "loss": 0.8296, "step": 76 }, { "epoch": 0.01, "learning_rate": 5.092592592592593e-07, "loss": 0.868, "step": 77 }, { "epoch": 0.01, "learning_rate": 5.158730158730158e-07, "loss": 0.9933, "step": 78 }, { "epoch": 0.01, "learning_rate": 5.224867724867725e-07, "loss": 1.0289, "step": 79 }, { "epoch": 0.01, "learning_rate": 5.291005291005291e-07, "loss": 1.2158, "step": 80 }, { "epoch": 0.01, "learning_rate": 5.357142857142857e-07, "loss": 1.3981, "step": 81 }, { "epoch": 0.01, "learning_rate": 5.423280423280424e-07, "loss": 0.7398, "step": 82 }, { "epoch": 0.01, "learning_rate": 5.489417989417989e-07, "loss": 0.7808, "step": 83 }, { "epoch": 0.01, "learning_rate": 5.555555555555555e-07, "loss": 0.7844, "step": 84 }, { "epoch": 0.01, "learning_rate": 5.621693121693122e-07, "loss": 1.0536, "step": 85 }, { "epoch": 0.01, "learning_rate": 5.687830687830688e-07, "loss": 1.1084, "step": 86 }, { "epoch": 0.01, "learning_rate": 5.753968253968255e-07, "loss": 1.3779, "step": 87 }, { "epoch": 0.01, "learning_rate": 5.82010582010582e-07, "loss": 0.7865, "step": 88 }, { "epoch": 0.01, "learning_rate": 5.886243386243386e-07, "loss": 0.6959, "step": 89 }, { "epoch": 0.01, "learning_rate": 5.952380952380953e-07, "loss": 0.8174, "step": 90 }, { "epoch": 0.01, "learning_rate": 6.018518518518519e-07, "loss": 1.0207, "step": 91 }, { "epoch": 0.01, "learning_rate": 6.084656084656085e-07, "loss": 1.1578, "step": 92 }, { "epoch": 0.01, "learning_rate": 6.150793650793651e-07, "loss": 1.353, "step": 93 }, { "epoch": 0.01, "learning_rate": 6.216931216931217e-07, "loss": 1.0031, "step": 94 }, { "epoch": 0.01, "learning_rate": 6.283068783068784e-07, "loss": 0.6857, "step": 95 }, { "epoch": 0.01, "learning_rate": 6.34920634920635e-07, "loss": 0.8983, "step": 96 }, { "epoch": 0.01, "learning_rate": 6.415343915343917e-07, "loss": 1.0657, "step": 97 }, { "epoch": 0.01, "learning_rate": 6.481481481481481e-07, "loss": 1.0965, "step": 98 }, { "epoch": 0.01, "learning_rate": 6.547619047619048e-07, "loss": 1.25, "step": 99 }, { "epoch": 0.01, "learning_rate": 6.613756613756614e-07, "loss": 1.5225, "step": 100 }, { "epoch": 0.01, "learning_rate": 6.679894179894181e-07, "loss": 0.7686, "step": 101 }, { "epoch": 0.01, "learning_rate": 6.746031746031746e-07, "loss": 0.7784, "step": 102 }, { "epoch": 0.01, "learning_rate": 6.812169312169313e-07, "loss": 0.8859, "step": 103 }, { "epoch": 0.01, "learning_rate": 6.878306878306879e-07, "loss": 1.1345, "step": 104 }, { "epoch": 0.01, "learning_rate": 6.944444444444446e-07, "loss": 1.1334, "step": 105 }, { "epoch": 0.01, "learning_rate": 7.010582010582011e-07, "loss": 1.3607, "step": 106 }, { "epoch": 0.01, "learning_rate": 7.076719576719577e-07, "loss": 0.5116, "step": 107 }, { "epoch": 0.01, "learning_rate": 7.142857142857143e-07, "loss": 0.875, "step": 108 }, { "epoch": 0.01, "learning_rate": 7.20899470899471e-07, "loss": 0.7676, "step": 109 }, { "epoch": 0.01, "learning_rate": 7.275132275132276e-07, "loss": 0.9667, "step": 110 }, { "epoch": 0.01, "learning_rate": 7.341269841269843e-07, "loss": 1.1851, "step": 111 }, { "epoch": 0.01, "learning_rate": 7.407407407407407e-07, "loss": 1.3382, "step": 112 }, { "epoch": 0.01, "learning_rate": 7.473544973544974e-07, "loss": 0.6625, "step": 113 }, { "epoch": 0.01, "learning_rate": 7.53968253968254e-07, "loss": 0.9593, "step": 114 }, { "epoch": 0.01, "learning_rate": 7.605820105820107e-07, "loss": 0.8945, "step": 115 }, { "epoch": 0.01, "learning_rate": 7.671957671957672e-07, "loss": 0.8804, "step": 116 }, { "epoch": 0.01, "learning_rate": 7.738095238095239e-07, "loss": 1.1133, "step": 117 }, { "epoch": 0.01, "learning_rate": 7.804232804232804e-07, "loss": 1.2021, "step": 118 }, { "epoch": 0.01, "learning_rate": 7.870370370370371e-07, "loss": 0.8356, "step": 119 }, { "epoch": 0.01, "learning_rate": 7.936507936507937e-07, "loss": 0.6666, "step": 120 }, { "epoch": 0.01, "learning_rate": 8.002645502645503e-07, "loss": 0.9064, "step": 121 }, { "epoch": 0.01, "learning_rate": 8.068783068783069e-07, "loss": 0.9771, "step": 122 }, { "epoch": 0.01, "learning_rate": 8.134920634920636e-07, "loss": 0.9865, "step": 123 }, { "epoch": 0.01, "learning_rate": 8.201058201058202e-07, "loss": 1.4589, "step": 124 }, { "epoch": 0.01, "learning_rate": 8.267195767195769e-07, "loss": 1.5343, "step": 125 }, { "epoch": 0.02, "learning_rate": 8.333333333333333e-07, "loss": 0.7586, "step": 126 }, { "epoch": 0.02, "learning_rate": 8.3994708994709e-07, "loss": 0.8182, "step": 127 }, { "epoch": 0.02, "learning_rate": 8.465608465608466e-07, "loss": 0.8649, "step": 128 }, { "epoch": 0.02, "learning_rate": 8.531746031746033e-07, "loss": 1.0579, "step": 129 }, { "epoch": 0.02, "learning_rate": 8.597883597883598e-07, "loss": 1.2183, "step": 130 }, { "epoch": 0.02, "learning_rate": 8.664021164021165e-07, "loss": 1.3435, "step": 131 }, { "epoch": 0.02, "learning_rate": 8.73015873015873e-07, "loss": 0.5499, "step": 132 }, { "epoch": 0.02, "learning_rate": 8.796296296296297e-07, "loss": 0.9047, "step": 133 }, { "epoch": 0.02, "learning_rate": 8.862433862433863e-07, "loss": 0.8955, "step": 134 }, { "epoch": 0.02, "learning_rate": 8.928571428571429e-07, "loss": 0.9008, "step": 135 }, { "epoch": 0.02, "learning_rate": 8.994708994708995e-07, "loss": 1.1393, "step": 136 }, { "epoch": 0.02, "learning_rate": 9.060846560846562e-07, "loss": 1.267, "step": 137 }, { "epoch": 0.02, "learning_rate": 9.126984126984127e-07, "loss": 0.7103, "step": 138 }, { "epoch": 0.02, "learning_rate": 9.193121693121694e-07, "loss": 0.7194, "step": 139 }, { "epoch": 0.02, "learning_rate": 9.259259259259259e-07, "loss": 0.8021, "step": 140 }, { "epoch": 0.02, "learning_rate": 9.325396825396826e-07, "loss": 0.8434, "step": 141 }, { "epoch": 0.02, "learning_rate": 9.391534391534392e-07, "loss": 1.138, "step": 142 }, { "epoch": 0.02, "learning_rate": 9.457671957671959e-07, "loss": 1.2589, "step": 143 }, { "epoch": 0.02, "learning_rate": 9.523809523809525e-07, "loss": 0.8187, "step": 144 }, { "epoch": 0.02, "learning_rate": 9.58994708994709e-07, "loss": 0.6318, "step": 145 }, { "epoch": 0.02, "learning_rate": 9.656084656084657e-07, "loss": 0.7803, "step": 146 }, { "epoch": 0.02, "learning_rate": 9.722222222222224e-07, "loss": 0.8155, "step": 147 }, { "epoch": 0.02, "learning_rate": 9.788359788359789e-07, "loss": 0.9269, "step": 148 }, { "epoch": 0.02, "learning_rate": 9.854497354497355e-07, "loss": 1.2373, "step": 149 }, { "epoch": 0.02, "learning_rate": 9.92063492063492e-07, "loss": 1.3292, "step": 150 }, { "epoch": 0.02, "learning_rate": 9.986772486772487e-07, "loss": 0.5286, "step": 151 }, { "epoch": 0.02, "learning_rate": 1.0052910052910054e-06, "loss": 0.7296, "step": 152 }, { "epoch": 0.02, "learning_rate": 1.011904761904762e-06, "loss": 0.8096, "step": 153 }, { "epoch": 0.02, "learning_rate": 1.0185185185185185e-06, "loss": 0.9318, "step": 154 }, { "epoch": 0.02, "learning_rate": 1.0251322751322752e-06, "loss": 1.3248, "step": 155 }, { "epoch": 0.02, "learning_rate": 1.0317460317460317e-06, "loss": 1.3584, "step": 156 }, { "epoch": 0.02, "learning_rate": 1.0383597883597884e-06, "loss": 0.6577, "step": 157 }, { "epoch": 0.02, "learning_rate": 1.044973544973545e-06, "loss": 0.6933, "step": 158 }, { "epoch": 0.02, "learning_rate": 1.0515873015873017e-06, "loss": 0.7947, "step": 159 }, { "epoch": 0.02, "learning_rate": 1.0582010582010582e-06, "loss": 0.9689, "step": 160 }, { "epoch": 0.02, "learning_rate": 1.0648148148148149e-06, "loss": 1.0975, "step": 161 }, { "epoch": 0.02, "learning_rate": 1.0714285714285714e-06, "loss": 1.3164, "step": 162 }, { "epoch": 0.02, "learning_rate": 1.078042328042328e-06, "loss": 0.8447, "step": 163 }, { "epoch": 0.02, "learning_rate": 1.0846560846560847e-06, "loss": 0.7375, "step": 164 }, { "epoch": 0.02, "learning_rate": 1.0912698412698414e-06, "loss": 0.7682, "step": 165 }, { "epoch": 0.02, "learning_rate": 1.0978835978835979e-06, "loss": 0.934, "step": 166 }, { "epoch": 0.02, "learning_rate": 1.1044973544973546e-06, "loss": 1.0671, "step": 167 }, { "epoch": 0.02, "learning_rate": 1.111111111111111e-06, "loss": 1.1253, "step": 168 }, { "epoch": 0.02, "learning_rate": 1.1177248677248677e-06, "loss": 0.6514, "step": 169 }, { "epoch": 0.02, "learning_rate": 1.1243386243386244e-06, "loss": 0.8506, "step": 170 }, { "epoch": 0.02, "learning_rate": 1.130952380952381e-06, "loss": 0.782, "step": 171 }, { "epoch": 0.02, "learning_rate": 1.1375661375661376e-06, "loss": 1.058, "step": 172 }, { "epoch": 0.02, "learning_rate": 1.1441798941798942e-06, "loss": 1.0795, "step": 173 }, { "epoch": 0.02, "learning_rate": 1.150793650793651e-06, "loss": 1.0887, "step": 174 }, { "epoch": 0.02, "learning_rate": 1.1574074074074076e-06, "loss": 1.3044, "step": 175 }, { "epoch": 0.02, "learning_rate": 1.164021164021164e-06, "loss": 0.5947, "step": 176 }, { "epoch": 0.02, "learning_rate": 1.1706349206349208e-06, "loss": 0.7128, "step": 177 }, { "epoch": 0.02, "learning_rate": 1.1772486772486772e-06, "loss": 0.756, "step": 178 }, { "epoch": 0.02, "learning_rate": 1.183862433862434e-06, "loss": 0.9897, "step": 179 }, { "epoch": 0.02, "learning_rate": 1.1904761904761906e-06, "loss": 1.1821, "step": 180 }, { "epoch": 0.02, "learning_rate": 1.1970899470899473e-06, "loss": 1.2322, "step": 181 }, { "epoch": 0.02, "learning_rate": 1.2037037037037037e-06, "loss": 0.6343, "step": 182 }, { "epoch": 0.02, "learning_rate": 1.2103174603174604e-06, "loss": 0.7867, "step": 183 }, { "epoch": 0.02, "learning_rate": 1.216931216931217e-06, "loss": 0.8974, "step": 184 }, { "epoch": 0.02, "learning_rate": 1.2235449735449736e-06, "loss": 0.9382, "step": 185 }, { "epoch": 0.02, "learning_rate": 1.2301587301587303e-06, "loss": 1.1755, "step": 186 }, { "epoch": 0.02, "learning_rate": 1.236772486772487e-06, "loss": 1.2977, "step": 187 }, { "epoch": 0.02, "learning_rate": 1.2433862433862434e-06, "loss": 0.5866, "step": 188 }, { "epoch": 0.02, "learning_rate": 1.25e-06, "loss": 0.6663, "step": 189 }, { "epoch": 0.02, "learning_rate": 1.2566137566137568e-06, "loss": 0.8334, "step": 190 }, { "epoch": 0.02, "learning_rate": 1.2632275132275135e-06, "loss": 0.9017, "step": 191 }, { "epoch": 0.02, "learning_rate": 1.26984126984127e-06, "loss": 1.0393, "step": 192 }, { "epoch": 0.02, "learning_rate": 1.2764550264550266e-06, "loss": 1.1745, "step": 193 }, { "epoch": 0.02, "learning_rate": 1.2830687830687833e-06, "loss": 0.9574, "step": 194 }, { "epoch": 0.02, "learning_rate": 1.28968253968254e-06, "loss": 0.5904, "step": 195 }, { "epoch": 0.02, "learning_rate": 1.2962962962962962e-06, "loss": 0.9204, "step": 196 }, { "epoch": 0.02, "learning_rate": 1.302910052910053e-06, "loss": 0.926, "step": 197 }, { "epoch": 0.02, "learning_rate": 1.3095238095238096e-06, "loss": 1.0837, "step": 198 }, { "epoch": 0.02, "learning_rate": 1.3161375661375663e-06, "loss": 1.2081, "step": 199 }, { "epoch": 0.02, "learning_rate": 1.3227513227513228e-06, "loss": 1.3442, "step": 200 }, { "epoch": 0.02, "learning_rate": 1.3293650793650794e-06, "loss": 0.6534, "step": 201 }, { "epoch": 0.02, "learning_rate": 1.3359788359788361e-06, "loss": 0.686, "step": 202 }, { "epoch": 0.02, "learning_rate": 1.3425925925925928e-06, "loss": 0.9498, "step": 203 }, { "epoch": 0.02, "learning_rate": 1.3492063492063493e-06, "loss": 0.9054, "step": 204 }, { "epoch": 0.02, "learning_rate": 1.355820105820106e-06, "loss": 1.2251, "step": 205 }, { "epoch": 0.02, "learning_rate": 1.3624338624338626e-06, "loss": 1.4073, "step": 206 }, { "epoch": 0.02, "learning_rate": 1.3690476190476193e-06, "loss": 0.4999, "step": 207 }, { "epoch": 0.02, "learning_rate": 1.3756613756613758e-06, "loss": 0.7112, "step": 208 }, { "epoch": 0.02, "learning_rate": 1.3822751322751325e-06, "loss": 0.7972, "step": 209 }, { "epoch": 0.03, "learning_rate": 1.3888888888888892e-06, "loss": 0.8142, "step": 210 }, { "epoch": 0.03, "learning_rate": 1.3955026455026458e-06, "loss": 1.0131, "step": 211 }, { "epoch": 0.03, "learning_rate": 1.4021164021164021e-06, "loss": 1.2659, "step": 212 }, { "epoch": 0.03, "learning_rate": 1.4087301587301588e-06, "loss": 0.7356, "step": 213 }, { "epoch": 0.03, "learning_rate": 1.4153439153439155e-06, "loss": 0.7061, "step": 214 }, { "epoch": 0.03, "learning_rate": 1.4219576719576722e-06, "loss": 0.8694, "step": 215 }, { "epoch": 0.03, "learning_rate": 1.4285714285714286e-06, "loss": 0.9563, "step": 216 }, { "epoch": 0.03, "learning_rate": 1.4351851851851853e-06, "loss": 1.0591, "step": 217 }, { "epoch": 0.03, "learning_rate": 1.441798941798942e-06, "loss": 1.3339, "step": 218 }, { "epoch": 0.03, "learning_rate": 1.4484126984126987e-06, "loss": 0.9561, "step": 219 }, { "epoch": 0.03, "learning_rate": 1.4550264550264551e-06, "loss": 0.7409, "step": 220 }, { "epoch": 0.03, "learning_rate": 1.4616402116402118e-06, "loss": 0.8483, "step": 221 }, { "epoch": 0.03, "learning_rate": 1.4682539682539685e-06, "loss": 0.8664, "step": 222 }, { "epoch": 0.03, "learning_rate": 1.4748677248677252e-06, "loss": 0.9741, "step": 223 }, { "epoch": 0.03, "learning_rate": 1.4814814814814815e-06, "loss": 1.19, "step": 224 }, { "epoch": 0.03, "learning_rate": 1.4880952380952381e-06, "loss": 1.1793, "step": 225 }, { "epoch": 0.03, "learning_rate": 1.4947089947089948e-06, "loss": 0.5823, "step": 226 }, { "epoch": 0.03, "learning_rate": 1.5013227513227515e-06, "loss": 0.8461, "step": 227 }, { "epoch": 0.03, "learning_rate": 1.507936507936508e-06, "loss": 0.824, "step": 228 }, { "epoch": 0.03, "learning_rate": 1.5145502645502647e-06, "loss": 1.0762, "step": 229 }, { "epoch": 0.03, "learning_rate": 1.5211640211640213e-06, "loss": 1.2544, "step": 230 }, { "epoch": 0.03, "learning_rate": 1.527777777777778e-06, "loss": 1.2062, "step": 231 }, { "epoch": 0.03, "learning_rate": 1.5343915343915345e-06, "loss": 0.6839, "step": 232 }, { "epoch": 0.03, "learning_rate": 1.5410052910052912e-06, "loss": 0.7703, "step": 233 }, { "epoch": 0.03, "learning_rate": 1.5476190476190479e-06, "loss": 0.9899, "step": 234 }, { "epoch": 0.03, "learning_rate": 1.5542328042328045e-06, "loss": 0.7933, "step": 235 }, { "epoch": 0.03, "learning_rate": 1.5608465608465608e-06, "loss": 1.0571, "step": 236 }, { "epoch": 0.03, "learning_rate": 1.5674603174603175e-06, "loss": 1.3212, "step": 237 }, { "epoch": 0.03, "learning_rate": 1.5740740740740742e-06, "loss": 0.6733, "step": 238 }, { "epoch": 0.03, "learning_rate": 1.5806878306878308e-06, "loss": 0.7157, "step": 239 }, { "epoch": 0.03, "learning_rate": 1.5873015873015873e-06, "loss": 0.6963, "step": 240 }, { "epoch": 0.03, "learning_rate": 1.593915343915344e-06, "loss": 0.9681, "step": 241 }, { "epoch": 0.03, "learning_rate": 1.6005291005291007e-06, "loss": 0.9355, "step": 242 }, { "epoch": 0.03, "learning_rate": 1.6071428571428574e-06, "loss": 1.1028, "step": 243 }, { "epoch": 0.03, "learning_rate": 1.6137566137566138e-06, "loss": 1.046, "step": 244 }, { "epoch": 0.03, "learning_rate": 1.6203703703703705e-06, "loss": 0.5188, "step": 245 }, { "epoch": 0.03, "learning_rate": 1.6269841269841272e-06, "loss": 0.6788, "step": 246 }, { "epoch": 0.03, "learning_rate": 1.6335978835978839e-06, "loss": 0.9629, "step": 247 }, { "epoch": 0.03, "learning_rate": 1.6402116402116404e-06, "loss": 1.04, "step": 248 }, { "epoch": 0.03, "learning_rate": 1.646825396825397e-06, "loss": 1.1167, "step": 249 }, { "epoch": 0.03, "learning_rate": 1.6534391534391537e-06, "loss": 1.4464, "step": 250 }, { "epoch": 0.03, "learning_rate": 1.6600529100529104e-06, "loss": 0.7301, "step": 251 }, { "epoch": 0.03, "learning_rate": 1.6666666666666667e-06, "loss": 0.8103, "step": 252 }, { "epoch": 0.03, "learning_rate": 1.6732804232804233e-06, "loss": 0.7995, "step": 253 }, { "epoch": 0.03, "learning_rate": 1.67989417989418e-06, "loss": 0.9402, "step": 254 }, { "epoch": 0.03, "learning_rate": 1.6865079365079367e-06, "loss": 0.9996, "step": 255 }, { "epoch": 0.03, "learning_rate": 1.6931216931216932e-06, "loss": 1.4265, "step": 256 }, { "epoch": 0.03, "learning_rate": 1.6997354497354499e-06, "loss": 0.6853, "step": 257 }, { "epoch": 0.03, "learning_rate": 1.7063492063492065e-06, "loss": 0.6169, "step": 258 }, { "epoch": 0.03, "learning_rate": 1.7129629629629632e-06, "loss": 0.6927, "step": 259 }, { "epoch": 0.03, "learning_rate": 1.7195767195767197e-06, "loss": 0.9808, "step": 260 }, { "epoch": 0.03, "learning_rate": 1.7261904761904764e-06, "loss": 1.218, "step": 261 }, { "epoch": 0.03, "learning_rate": 1.732804232804233e-06, "loss": 1.3341, "step": 262 }, { "epoch": 0.03, "learning_rate": 1.7394179894179897e-06, "loss": 0.7073, "step": 263 }, { "epoch": 0.03, "learning_rate": 1.746031746031746e-06, "loss": 0.6801, "step": 264 }, { "epoch": 0.03, "learning_rate": 1.7526455026455027e-06, "loss": 0.7469, "step": 265 }, { "epoch": 0.03, "learning_rate": 1.7592592592592594e-06, "loss": 0.8754, "step": 266 }, { "epoch": 0.03, "learning_rate": 1.765873015873016e-06, "loss": 1.1025, "step": 267 }, { "epoch": 0.03, "learning_rate": 1.7724867724867725e-06, "loss": 1.2089, "step": 268 }, { "epoch": 0.03, "learning_rate": 1.7791005291005292e-06, "loss": 0.9923, "step": 269 }, { "epoch": 0.03, "learning_rate": 1.7857142857142859e-06, "loss": 0.5755, "step": 270 }, { "epoch": 0.03, "learning_rate": 1.7923280423280426e-06, "loss": 0.6801, "step": 271 }, { "epoch": 0.03, "learning_rate": 1.798941798941799e-06, "loss": 0.8605, "step": 272 }, { "epoch": 0.03, "learning_rate": 1.8055555555555557e-06, "loss": 1.0542, "step": 273 }, { "epoch": 0.03, "learning_rate": 1.8121693121693124e-06, "loss": 1.2363, "step": 274 }, { "epoch": 0.03, "learning_rate": 1.818783068783069e-06, "loss": 1.3785, "step": 275 }, { "epoch": 0.03, "learning_rate": 1.8253968253968254e-06, "loss": 0.8527, "step": 276 }, { "epoch": 0.03, "learning_rate": 1.832010582010582e-06, "loss": 0.864, "step": 277 }, { "epoch": 0.03, "learning_rate": 1.8386243386243387e-06, "loss": 0.9462, "step": 278 }, { "epoch": 0.03, "learning_rate": 1.8452380952380954e-06, "loss": 1.069, "step": 279 }, { "epoch": 0.03, "learning_rate": 1.8518518518518519e-06, "loss": 1.1281, "step": 280 }, { "epoch": 0.03, "learning_rate": 1.8584656084656086e-06, "loss": 1.3628, "step": 281 }, { "epoch": 0.03, "learning_rate": 1.8650793650793652e-06, "loss": 0.6513, "step": 282 }, { "epoch": 0.03, "learning_rate": 1.871693121693122e-06, "loss": 0.8503, "step": 283 }, { "epoch": 0.03, "learning_rate": 1.8783068783068784e-06, "loss": 0.8668, "step": 284 }, { "epoch": 0.03, "learning_rate": 1.884920634920635e-06, "loss": 0.9442, "step": 285 }, { "epoch": 0.03, "learning_rate": 1.8915343915343918e-06, "loss": 1.2191, "step": 286 }, { "epoch": 0.03, "learning_rate": 1.8981481481481484e-06, "loss": 1.4247, "step": 287 }, { "epoch": 0.03, "learning_rate": 1.904761904761905e-06, "loss": 0.8961, "step": 288 }, { "epoch": 0.03, "learning_rate": 1.9113756613756614e-06, "loss": 0.7441, "step": 289 }, { "epoch": 0.03, "learning_rate": 1.917989417989418e-06, "loss": 1.05, "step": 290 }, { "epoch": 0.03, "learning_rate": 1.9246031746031747e-06, "loss": 0.8515, "step": 291 }, { "epoch": 0.03, "learning_rate": 1.9312169312169314e-06, "loss": 0.9595, "step": 292 }, { "epoch": 0.03, "learning_rate": 1.937830687830688e-06, "loss": 1.1749, "step": 293 }, { "epoch": 0.04, "learning_rate": 1.944444444444445e-06, "loss": 0.7219, "step": 294 }, { "epoch": 0.04, "learning_rate": 1.9510582010582015e-06, "loss": 0.7764, "step": 295 }, { "epoch": 0.04, "learning_rate": 1.9576719576719577e-06, "loss": 0.7834, "step": 296 }, { "epoch": 0.04, "learning_rate": 1.9642857142857144e-06, "loss": 0.8802, "step": 297 }, { "epoch": 0.04, "learning_rate": 1.970899470899471e-06, "loss": 0.9821, "step": 298 }, { "epoch": 0.04, "learning_rate": 1.9775132275132278e-06, "loss": 1.3199, "step": 299 }, { "epoch": 0.04, "learning_rate": 1.984126984126984e-06, "loss": 1.2568, "step": 300 }, { "epoch": 0.04, "learning_rate": 1.9907407407407407e-06, "loss": 0.6248, "step": 301 }, { "epoch": 0.04, "learning_rate": 1.9973544973544974e-06, "loss": 0.8932, "step": 302 }, { "epoch": 0.04, "learning_rate": 2.003968253968254e-06, "loss": 0.9427, "step": 303 }, { "epoch": 0.04, "learning_rate": 2.0105820105820108e-06, "loss": 0.9426, "step": 304 }, { "epoch": 0.04, "learning_rate": 2.0171957671957675e-06, "loss": 1.144, "step": 305 }, { "epoch": 0.04, "learning_rate": 2.023809523809524e-06, "loss": 1.2833, "step": 306 }, { "epoch": 0.04, "learning_rate": 2.030423280423281e-06, "loss": 0.5651, "step": 307 }, { "epoch": 0.04, "learning_rate": 2.037037037037037e-06, "loss": 0.866, "step": 308 }, { "epoch": 0.04, "learning_rate": 2.0436507936507938e-06, "loss": 0.7631, "step": 309 }, { "epoch": 0.04, "learning_rate": 2.0502645502645504e-06, "loss": 0.9208, "step": 310 }, { "epoch": 0.04, "learning_rate": 2.056878306878307e-06, "loss": 0.9788, "step": 311 }, { "epoch": 0.04, "learning_rate": 2.0634920634920634e-06, "loss": 1.2521, "step": 312 }, { "epoch": 0.04, "learning_rate": 2.07010582010582e-06, "loss": 0.5952, "step": 313 }, { "epoch": 0.04, "learning_rate": 2.0767195767195768e-06, "loss": 0.8161, "step": 314 }, { "epoch": 0.04, "learning_rate": 2.0833333333333334e-06, "loss": 0.8612, "step": 315 }, { "epoch": 0.04, "learning_rate": 2.08994708994709e-06, "loss": 0.9609, "step": 316 }, { "epoch": 0.04, "learning_rate": 2.096560846560847e-06, "loss": 1.0099, "step": 317 }, { "epoch": 0.04, "learning_rate": 2.1031746031746035e-06, "loss": 1.2107, "step": 318 }, { "epoch": 0.04, "learning_rate": 2.10978835978836e-06, "loss": 0.9742, "step": 319 }, { "epoch": 0.04, "learning_rate": 2.1164021164021164e-06, "loss": 0.6676, "step": 320 }, { "epoch": 0.04, "learning_rate": 2.123015873015873e-06, "loss": 0.9579, "step": 321 }, { "epoch": 0.04, "learning_rate": 2.1296296296296298e-06, "loss": 0.7764, "step": 322 }, { "epoch": 0.04, "learning_rate": 2.1362433862433865e-06, "loss": 0.9472, "step": 323 }, { "epoch": 0.04, "learning_rate": 2.1428571428571427e-06, "loss": 1.1719, "step": 324 }, { "epoch": 0.04, "learning_rate": 2.1494708994708994e-06, "loss": 1.2726, "step": 325 }, { "epoch": 0.04, "learning_rate": 2.156084656084656e-06, "loss": 0.6882, "step": 326 }, { "epoch": 0.04, "learning_rate": 2.1626984126984128e-06, "loss": 0.7218, "step": 327 }, { "epoch": 0.04, "learning_rate": 2.1693121693121695e-06, "loss": 0.9272, "step": 328 }, { "epoch": 0.04, "learning_rate": 2.175925925925926e-06, "loss": 0.7887, "step": 329 }, { "epoch": 0.04, "learning_rate": 2.182539682539683e-06, "loss": 1.12, "step": 330 }, { "epoch": 0.04, "learning_rate": 2.1891534391534395e-06, "loss": 1.4636, "step": 331 }, { "epoch": 0.04, "learning_rate": 2.1957671957671958e-06, "loss": 0.5907, "step": 332 }, { "epoch": 0.04, "learning_rate": 2.2023809523809525e-06, "loss": 0.8202, "step": 333 }, { "epoch": 0.04, "learning_rate": 2.208994708994709e-06, "loss": 0.7427, "step": 334 }, { "epoch": 0.04, "learning_rate": 2.215608465608466e-06, "loss": 0.9754, "step": 335 }, { "epoch": 0.04, "learning_rate": 2.222222222222222e-06, "loss": 1.0842, "step": 336 }, { "epoch": 0.04, "learning_rate": 2.2288359788359788e-06, "loss": 1.4355, "step": 337 }, { "epoch": 0.04, "learning_rate": 2.2354497354497354e-06, "loss": 0.5849, "step": 338 }, { "epoch": 0.04, "learning_rate": 2.242063492063492e-06, "loss": 0.8389, "step": 339 }, { "epoch": 0.04, "learning_rate": 2.248677248677249e-06, "loss": 0.9048, "step": 340 }, { "epoch": 0.04, "learning_rate": 2.2552910052910055e-06, "loss": 0.8651, "step": 341 }, { "epoch": 0.04, "learning_rate": 2.261904761904762e-06, "loss": 1.1431, "step": 342 }, { "epoch": 0.04, "learning_rate": 2.268518518518519e-06, "loss": 1.1469, "step": 343 }, { "epoch": 0.04, "learning_rate": 2.275132275132275e-06, "loss": 1.0254, "step": 344 }, { "epoch": 0.04, "learning_rate": 2.281746031746032e-06, "loss": 0.6148, "step": 345 }, { "epoch": 0.04, "learning_rate": 2.2883597883597885e-06, "loss": 0.8769, "step": 346 }, { "epoch": 0.04, "learning_rate": 2.294973544973545e-06, "loss": 0.7907, "step": 347 }, { "epoch": 0.04, "learning_rate": 2.301587301587302e-06, "loss": 0.893, "step": 348 }, { "epoch": 0.04, "learning_rate": 2.3082010582010585e-06, "loss": 1.2274, "step": 349 }, { "epoch": 0.04, "learning_rate": 2.314814814814815e-06, "loss": 1.2374, "step": 350 }, { "epoch": 0.04, "learning_rate": 2.321428571428572e-06, "loss": 0.7605, "step": 351 }, { "epoch": 0.04, "learning_rate": 2.328042328042328e-06, "loss": 0.6963, "step": 352 }, { "epoch": 0.04, "learning_rate": 2.334656084656085e-06, "loss": 0.7174, "step": 353 }, { "epoch": 0.04, "learning_rate": 2.3412698412698415e-06, "loss": 0.9388, "step": 354 }, { "epoch": 0.04, "learning_rate": 2.347883597883598e-06, "loss": 1.1626, "step": 355 }, { "epoch": 0.04, "learning_rate": 2.3544973544973545e-06, "loss": 1.2371, "step": 356 }, { "epoch": 0.04, "learning_rate": 2.361111111111111e-06, "loss": 0.6292, "step": 357 }, { "epoch": 0.04, "learning_rate": 2.367724867724868e-06, "loss": 0.7321, "step": 358 }, { "epoch": 0.04, "learning_rate": 2.3743386243386245e-06, "loss": 0.7773, "step": 359 }, { "epoch": 0.04, "learning_rate": 2.380952380952381e-06, "loss": 0.9193, "step": 360 }, { "epoch": 0.04, "learning_rate": 2.387566137566138e-06, "loss": 0.9747, "step": 361 }, { "epoch": 0.04, "learning_rate": 2.3941798941798946e-06, "loss": 1.2629, "step": 362 }, { "epoch": 0.04, "learning_rate": 2.4007936507936512e-06, "loss": 0.6673, "step": 363 }, { "epoch": 0.04, "learning_rate": 2.4074074074074075e-06, "loss": 0.727, "step": 364 }, { "epoch": 0.04, "learning_rate": 2.414021164021164e-06, "loss": 0.7894, "step": 365 }, { "epoch": 0.04, "learning_rate": 2.420634920634921e-06, "loss": 0.9012, "step": 366 }, { "epoch": 0.04, "learning_rate": 2.4272486772486775e-06, "loss": 1.0351, "step": 367 }, { "epoch": 0.04, "learning_rate": 2.433862433862434e-06, "loss": 1.2742, "step": 368 }, { "epoch": 0.04, "learning_rate": 2.4404761904761905e-06, "loss": 0.8917, "step": 369 }, { "epoch": 0.04, "learning_rate": 2.447089947089947e-06, "loss": 0.7109, "step": 370 }, { "epoch": 0.04, "learning_rate": 2.453703703703704e-06, "loss": 0.7301, "step": 371 }, { "epoch": 0.04, "learning_rate": 2.4603174603174605e-06, "loss": 0.8201, "step": 372 }, { "epoch": 0.04, "learning_rate": 2.4669312169312172e-06, "loss": 1.0271, "step": 373 }, { "epoch": 0.04, "learning_rate": 2.473544973544974e-06, "loss": 1.087, "step": 374 }, { "epoch": 0.04, "learning_rate": 2.4801587301587306e-06, "loss": 1.2751, "step": 375 }, { "epoch": 0.04, "learning_rate": 2.486772486772487e-06, "loss": 0.5811, "step": 376 }, { "epoch": 0.04, "learning_rate": 2.4933862433862435e-06, "loss": 0.8555, "step": 377 }, { "epoch": 0.05, "learning_rate": 2.5e-06, "loss": 0.7955, "step": 378 }, { "epoch": 0.05, "learning_rate": 2.5066137566137565e-06, "loss": 0.9423, "step": 379 }, { "epoch": 0.05, "learning_rate": 2.5132275132275136e-06, "loss": 1.1555, "step": 380 }, { "epoch": 0.05, "learning_rate": 2.51984126984127e-06, "loss": 1.2766, "step": 381 }, { "epoch": 0.05, "learning_rate": 2.526455026455027e-06, "loss": 0.5354, "step": 382 }, { "epoch": 0.05, "learning_rate": 2.533068783068783e-06, "loss": 0.7489, "step": 383 }, { "epoch": 0.05, "learning_rate": 2.53968253968254e-06, "loss": 0.7859, "step": 384 }, { "epoch": 0.05, "learning_rate": 2.5462962962962966e-06, "loss": 0.8033, "step": 385 }, { "epoch": 0.05, "learning_rate": 2.5529100529100532e-06, "loss": 1.1963, "step": 386 }, { "epoch": 0.05, "learning_rate": 2.5595238095238095e-06, "loss": 1.3662, "step": 387 }, { "epoch": 0.05, "learning_rate": 2.5661375661375666e-06, "loss": 0.5267, "step": 388 }, { "epoch": 0.05, "learning_rate": 2.572751322751323e-06, "loss": 0.7312, "step": 389 }, { "epoch": 0.05, "learning_rate": 2.57936507936508e-06, "loss": 0.7994, "step": 390 }, { "epoch": 0.05, "learning_rate": 2.5859788359788362e-06, "loss": 0.7785, "step": 391 }, { "epoch": 0.05, "learning_rate": 2.5925925925925925e-06, "loss": 1.007, "step": 392 }, { "epoch": 0.05, "learning_rate": 2.5992063492063496e-06, "loss": 1.2466, "step": 393 }, { "epoch": 0.05, "learning_rate": 2.605820105820106e-06, "loss": 0.7769, "step": 394 }, { "epoch": 0.05, "learning_rate": 2.6124338624338625e-06, "loss": 0.6974, "step": 395 }, { "epoch": 0.05, "learning_rate": 2.6190476190476192e-06, "loss": 0.7756, "step": 396 }, { "epoch": 0.05, "learning_rate": 2.625661375661376e-06, "loss": 0.8962, "step": 397 }, { "epoch": 0.05, "learning_rate": 2.6322751322751326e-06, "loss": 1.0339, "step": 398 }, { "epoch": 0.05, "learning_rate": 2.6388888888888893e-06, "loss": 1.2208, "step": 399 }, { "epoch": 0.05, "learning_rate": 2.6455026455026455e-06, "loss": 1.2767, "step": 400 }, { "epoch": 0.05, "learning_rate": 2.6521164021164026e-06, "loss": 0.6136, "step": 401 }, { "epoch": 0.05, "learning_rate": 2.658730158730159e-06, "loss": 0.7764, "step": 402 }, { "epoch": 0.05, "learning_rate": 2.665343915343915e-06, "loss": 0.7423, "step": 403 }, { "epoch": 0.05, "learning_rate": 2.6719576719576723e-06, "loss": 0.907, "step": 404 }, { "epoch": 0.05, "learning_rate": 2.6785714285714285e-06, "loss": 1.1033, "step": 405 }, { "epoch": 0.05, "learning_rate": 2.6851851851851856e-06, "loss": 1.4498, "step": 406 }, { "epoch": 0.05, "learning_rate": 2.691798941798942e-06, "loss": 0.7806, "step": 407 }, { "epoch": 0.05, "learning_rate": 2.6984126984126986e-06, "loss": 0.7929, "step": 408 }, { "epoch": 0.05, "learning_rate": 2.7050264550264552e-06, "loss": 0.8618, "step": 409 }, { "epoch": 0.05, "learning_rate": 2.711640211640212e-06, "loss": 0.7509, "step": 410 }, { "epoch": 0.05, "learning_rate": 2.718253968253968e-06, "loss": 1.0909, "step": 411 }, { "epoch": 0.05, "learning_rate": 2.7248677248677253e-06, "loss": 1.2327, "step": 412 }, { "epoch": 0.05, "learning_rate": 2.7314814814814816e-06, "loss": 0.6099, "step": 413 }, { "epoch": 0.05, "learning_rate": 2.7380952380952387e-06, "loss": 0.7523, "step": 414 }, { "epoch": 0.05, "learning_rate": 2.744708994708995e-06, "loss": 0.7127, "step": 415 }, { "epoch": 0.05, "learning_rate": 2.7513227513227516e-06, "loss": 0.6982, "step": 416 }, { "epoch": 0.05, "learning_rate": 2.7579365079365083e-06, "loss": 0.9084, "step": 417 }, { "epoch": 0.05, "learning_rate": 2.764550264550265e-06, "loss": 1.2943, "step": 418 }, { "epoch": 0.05, "learning_rate": 2.7711640211640212e-06, "loss": 0.8036, "step": 419 }, { "epoch": 0.05, "learning_rate": 2.7777777777777783e-06, "loss": 0.7156, "step": 420 }, { "epoch": 0.05, "learning_rate": 2.7843915343915346e-06, "loss": 0.8509, "step": 421 }, { "epoch": 0.05, "learning_rate": 2.7910052910052917e-06, "loss": 0.8589, "step": 422 }, { "epoch": 0.05, "learning_rate": 2.797619047619048e-06, "loss": 0.9274, "step": 423 }, { "epoch": 0.05, "learning_rate": 2.8042328042328042e-06, "loss": 1.2012, "step": 424 }, { "epoch": 0.05, "learning_rate": 2.8108465608465613e-06, "loss": 1.1424, "step": 425 }, { "epoch": 0.05, "learning_rate": 2.8174603174603176e-06, "loss": 0.6775, "step": 426 }, { "epoch": 0.05, "learning_rate": 2.8240740740740743e-06, "loss": 0.793, "step": 427 }, { "epoch": 0.05, "learning_rate": 2.830687830687831e-06, "loss": 0.8286, "step": 428 }, { "epoch": 0.05, "learning_rate": 2.8373015873015876e-06, "loss": 0.9824, "step": 429 }, { "epoch": 0.05, "learning_rate": 2.8439153439153443e-06, "loss": 0.9916, "step": 430 }, { "epoch": 0.05, "learning_rate": 2.850529100529101e-06, "loss": 1.2114, "step": 431 }, { "epoch": 0.05, "learning_rate": 2.8571428571428573e-06, "loss": 0.5928, "step": 432 }, { "epoch": 0.05, "learning_rate": 2.8637566137566144e-06, "loss": 0.898, "step": 433 }, { "epoch": 0.05, "learning_rate": 2.8703703703703706e-06, "loss": 0.7552, "step": 434 }, { "epoch": 0.05, "learning_rate": 2.876984126984127e-06, "loss": 0.811, "step": 435 }, { "epoch": 0.05, "learning_rate": 2.883597883597884e-06, "loss": 1.0371, "step": 436 }, { "epoch": 0.05, "learning_rate": 2.8902116402116402e-06, "loss": 1.3533, "step": 437 }, { "epoch": 0.05, "learning_rate": 2.8968253968253974e-06, "loss": 0.8336, "step": 438 }, { "epoch": 0.05, "learning_rate": 2.9034391534391536e-06, "loss": 0.8073, "step": 439 }, { "epoch": 0.05, "learning_rate": 2.9100529100529103e-06, "loss": 0.726, "step": 440 }, { "epoch": 0.05, "learning_rate": 2.916666666666667e-06, "loss": 0.7798, "step": 441 }, { "epoch": 0.05, "learning_rate": 2.9232804232804237e-06, "loss": 1.0632, "step": 442 }, { "epoch": 0.05, "learning_rate": 2.92989417989418e-06, "loss": 1.2617, "step": 443 }, { "epoch": 0.05, "learning_rate": 2.936507936507937e-06, "loss": 0.6513, "step": 444 }, { "epoch": 0.05, "learning_rate": 2.9431216931216933e-06, "loss": 0.8278, "step": 445 }, { "epoch": 0.05, "learning_rate": 2.9497354497354504e-06, "loss": 0.7585, "step": 446 }, { "epoch": 0.05, "learning_rate": 2.9563492063492066e-06, "loss": 0.8964, "step": 447 }, { "epoch": 0.05, "learning_rate": 2.962962962962963e-06, "loss": 0.997, "step": 448 }, { "epoch": 0.05, "learning_rate": 2.96957671957672e-06, "loss": 1.1446, "step": 449 }, { "epoch": 0.05, "learning_rate": 2.9761904761904763e-06, "loss": 1.2881, "step": 450 }, { "epoch": 0.05, "learning_rate": 2.982804232804233e-06, "loss": 0.591, "step": 451 }, { "epoch": 0.05, "learning_rate": 2.9894179894179896e-06, "loss": 0.787, "step": 452 }, { "epoch": 0.05, "learning_rate": 2.9960317460317463e-06, "loss": 0.6917, "step": 453 }, { "epoch": 0.05, "learning_rate": 3.002645502645503e-06, "loss": 0.8376, "step": 454 }, { "epoch": 0.05, "learning_rate": 3.0092592592592597e-06, "loss": 1.1552, "step": 455 }, { "epoch": 0.05, "learning_rate": 3.015873015873016e-06, "loss": 1.299, "step": 456 }, { "epoch": 0.05, "learning_rate": 3.022486772486773e-06, "loss": 0.6683, "step": 457 }, { "epoch": 0.05, "learning_rate": 3.0291005291005293e-06, "loss": 0.72, "step": 458 }, { "epoch": 0.05, "learning_rate": 3.0357142857142856e-06, "loss": 1.0106, "step": 459 }, { "epoch": 0.05, "learning_rate": 3.0423280423280427e-06, "loss": 0.7916, "step": 460 }, { "epoch": 0.05, "learning_rate": 3.048941798941799e-06, "loss": 1.0896, "step": 461 }, { "epoch": 0.06, "learning_rate": 3.055555555555556e-06, "loss": 1.3276, "step": 462 }, { "epoch": 0.06, "learning_rate": 3.0621693121693123e-06, "loss": 0.7529, "step": 463 }, { "epoch": 0.06, "learning_rate": 3.068783068783069e-06, "loss": 0.7788, "step": 464 }, { "epoch": 0.06, "learning_rate": 3.0753968253968257e-06, "loss": 0.7, "step": 465 }, { "epoch": 0.06, "learning_rate": 3.0820105820105823e-06, "loss": 0.8147, "step": 466 }, { "epoch": 0.06, "learning_rate": 3.0886243386243386e-06, "loss": 0.9976, "step": 467 }, { "epoch": 0.06, "learning_rate": 3.0952380952380957e-06, "loss": 1.2769, "step": 468 }, { "epoch": 0.06, "learning_rate": 3.101851851851852e-06, "loss": 0.8912, "step": 469 }, { "epoch": 0.06, "learning_rate": 3.108465608465609e-06, "loss": 0.6753, "step": 470 }, { "epoch": 0.06, "learning_rate": 3.1150793650793653e-06, "loss": 0.7242, "step": 471 }, { "epoch": 0.06, "learning_rate": 3.1216931216931216e-06, "loss": 0.7346, "step": 472 }, { "epoch": 0.06, "learning_rate": 3.1283068783068787e-06, "loss": 0.9068, "step": 473 }, { "epoch": 0.06, "learning_rate": 3.134920634920635e-06, "loss": 1.2903, "step": 474 }, { "epoch": 0.06, "learning_rate": 3.1415343915343916e-06, "loss": 1.2606, "step": 475 }, { "epoch": 0.06, "learning_rate": 3.1481481481481483e-06, "loss": 0.5543, "step": 476 }, { "epoch": 0.06, "learning_rate": 3.154761904761905e-06, "loss": 0.7961, "step": 477 }, { "epoch": 0.06, "learning_rate": 3.1613756613756617e-06, "loss": 0.9419, "step": 478 }, { "epoch": 0.06, "learning_rate": 3.1679894179894184e-06, "loss": 0.9861, "step": 479 }, { "epoch": 0.06, "learning_rate": 3.1746031746031746e-06, "loss": 1.0844, "step": 480 }, { "epoch": 0.06, "learning_rate": 3.1812169312169317e-06, "loss": 1.3391, "step": 481 }, { "epoch": 0.06, "learning_rate": 3.187830687830688e-06, "loss": 0.6942, "step": 482 }, { "epoch": 0.06, "learning_rate": 3.1944444444444443e-06, "loss": 0.9461, "step": 483 }, { "epoch": 0.06, "learning_rate": 3.2010582010582014e-06, "loss": 0.6358, "step": 484 }, { "epoch": 0.06, "learning_rate": 3.2076719576719576e-06, "loss": 0.863, "step": 485 }, { "epoch": 0.06, "learning_rate": 3.2142857142857147e-06, "loss": 1.0994, "step": 486 }, { "epoch": 0.06, "learning_rate": 3.220899470899471e-06, "loss": 1.174, "step": 487 }, { "epoch": 0.06, "learning_rate": 3.2275132275132277e-06, "loss": 0.7418, "step": 488 }, { "epoch": 0.06, "learning_rate": 3.2341269841269844e-06, "loss": 0.5861, "step": 489 }, { "epoch": 0.06, "learning_rate": 3.240740740740741e-06, "loss": 0.8369, "step": 490 }, { "epoch": 0.06, "learning_rate": 3.2473544973544973e-06, "loss": 0.7725, "step": 491 }, { "epoch": 0.06, "learning_rate": 3.2539682539682544e-06, "loss": 1.0631, "step": 492 }, { "epoch": 0.06, "learning_rate": 3.2605820105820107e-06, "loss": 1.1386, "step": 493 }, { "epoch": 0.06, "learning_rate": 3.2671957671957678e-06, "loss": 0.74, "step": 494 }, { "epoch": 0.06, "learning_rate": 3.273809523809524e-06, "loss": 0.6156, "step": 495 }, { "epoch": 0.06, "learning_rate": 3.2804232804232807e-06, "loss": 0.8089, "step": 496 }, { "epoch": 0.06, "learning_rate": 3.2870370370370374e-06, "loss": 0.8671, "step": 497 }, { "epoch": 0.06, "learning_rate": 3.293650793650794e-06, "loss": 0.8911, "step": 498 }, { "epoch": 0.06, "learning_rate": 3.3002645502645503e-06, "loss": 1.1646, "step": 499 }, { "epoch": 0.06, "learning_rate": 3.3068783068783074e-06, "loss": 1.1494, "step": 500 }, { "epoch": 0.06, "learning_rate": 3.3134920634920637e-06, "loss": 0.6026, "step": 501 }, { "epoch": 0.06, "learning_rate": 3.320105820105821e-06, "loss": 0.8433, "step": 502 }, { "epoch": 0.06, "learning_rate": 3.326719576719577e-06, "loss": 0.8438, "step": 503 }, { "epoch": 0.06, "learning_rate": 3.3333333333333333e-06, "loss": 1.046, "step": 504 }, { "epoch": 0.06, "learning_rate": 3.3399470899470904e-06, "loss": 1.153, "step": 505 }, { "epoch": 0.06, "learning_rate": 3.3465608465608467e-06, "loss": 1.3383, "step": 506 }, { "epoch": 0.06, "learning_rate": 3.3531746031746034e-06, "loss": 0.79, "step": 507 }, { "epoch": 0.06, "learning_rate": 3.35978835978836e-06, "loss": 0.8764, "step": 508 }, { "epoch": 0.06, "learning_rate": 3.3664021164021167e-06, "loss": 0.8051, "step": 509 }, { "epoch": 0.06, "learning_rate": 3.3730158730158734e-06, "loss": 0.873, "step": 510 }, { "epoch": 0.06, "learning_rate": 3.37962962962963e-06, "loss": 1.1486, "step": 511 }, { "epoch": 0.06, "learning_rate": 3.3862433862433864e-06, "loss": 1.2235, "step": 512 }, { "epoch": 0.06, "learning_rate": 3.3928571428571435e-06, "loss": 0.7694, "step": 513 }, { "epoch": 0.06, "learning_rate": 3.3994708994708997e-06, "loss": 0.6143, "step": 514 }, { "epoch": 0.06, "learning_rate": 3.406084656084656e-06, "loss": 0.7252, "step": 515 }, { "epoch": 0.06, "learning_rate": 3.412698412698413e-06, "loss": 0.6915, "step": 516 }, { "epoch": 0.06, "learning_rate": 3.4193121693121694e-06, "loss": 1.0388, "step": 517 }, { "epoch": 0.06, "learning_rate": 3.4259259259259265e-06, "loss": 1.2185, "step": 518 }, { "epoch": 0.06, "learning_rate": 3.4325396825396827e-06, "loss": 0.8055, "step": 519 }, { "epoch": 0.06, "learning_rate": 3.4391534391534394e-06, "loss": 0.5847, "step": 520 }, { "epoch": 0.06, "learning_rate": 3.445767195767196e-06, "loss": 0.7213, "step": 521 }, { "epoch": 0.06, "learning_rate": 3.4523809523809528e-06, "loss": 0.8445, "step": 522 }, { "epoch": 0.06, "learning_rate": 3.458994708994709e-06, "loss": 0.8607, "step": 523 }, { "epoch": 0.06, "learning_rate": 3.465608465608466e-06, "loss": 1.1553, "step": 524 }, { "epoch": 0.06, "learning_rate": 3.4722222222222224e-06, "loss": 1.213, "step": 525 }, { "epoch": 0.06, "learning_rate": 3.4788359788359795e-06, "loss": 0.6516, "step": 526 }, { "epoch": 0.06, "learning_rate": 3.4854497354497358e-06, "loss": 0.6732, "step": 527 }, { "epoch": 0.06, "learning_rate": 3.492063492063492e-06, "loss": 0.7253, "step": 528 }, { "epoch": 0.06, "learning_rate": 3.498677248677249e-06, "loss": 0.9674, "step": 529 }, { "epoch": 0.06, "learning_rate": 3.5052910052910054e-06, "loss": 1.1552, "step": 530 }, { "epoch": 0.06, "learning_rate": 3.511904761904762e-06, "loss": 1.2574, "step": 531 }, { "epoch": 0.06, "learning_rate": 3.5185185185185187e-06, "loss": 0.5589, "step": 532 }, { "epoch": 0.06, "learning_rate": 3.5251322751322754e-06, "loss": 0.8324, "step": 533 }, { "epoch": 0.06, "learning_rate": 3.531746031746032e-06, "loss": 0.7639, "step": 534 }, { "epoch": 0.06, "learning_rate": 3.538359788359789e-06, "loss": 0.937, "step": 535 }, { "epoch": 0.06, "learning_rate": 3.544973544973545e-06, "loss": 1.0141, "step": 536 }, { "epoch": 0.06, "learning_rate": 3.551587301587302e-06, "loss": 1.1882, "step": 537 }, { "epoch": 0.06, "learning_rate": 3.5582010582010584e-06, "loss": 0.7623, "step": 538 }, { "epoch": 0.06, "learning_rate": 3.5648148148148147e-06, "loss": 0.6705, "step": 539 }, { "epoch": 0.06, "learning_rate": 3.5714285714285718e-06, "loss": 0.8517, "step": 540 }, { "epoch": 0.06, "learning_rate": 3.578042328042328e-06, "loss": 0.9467, "step": 541 }, { "epoch": 0.06, "learning_rate": 3.584656084656085e-06, "loss": 0.951, "step": 542 }, { "epoch": 0.06, "learning_rate": 3.5912698412698414e-06, "loss": 1.183, "step": 543 }, { "epoch": 0.06, "learning_rate": 3.597883597883598e-06, "loss": 0.9055, "step": 544 }, { "epoch": 0.06, "learning_rate": 3.6044973544973548e-06, "loss": 0.5807, "step": 545 }, { "epoch": 0.07, "learning_rate": 3.6111111111111115e-06, "loss": 0.7399, "step": 546 }, { "epoch": 0.07, "learning_rate": 3.6177248677248677e-06, "loss": 0.8683, "step": 547 }, { "epoch": 0.07, "learning_rate": 3.624338624338625e-06, "loss": 0.9711, "step": 548 }, { "epoch": 0.07, "learning_rate": 3.630952380952381e-06, "loss": 1.1151, "step": 549 }, { "epoch": 0.07, "learning_rate": 3.637566137566138e-06, "loss": 1.2936, "step": 550 }, { "epoch": 0.07, "learning_rate": 3.6441798941798944e-06, "loss": 0.8434, "step": 551 }, { "epoch": 0.07, "learning_rate": 3.6507936507936507e-06, "loss": 0.7618, "step": 552 }, { "epoch": 0.07, "learning_rate": 3.657407407407408e-06, "loss": 0.8002, "step": 553 }, { "epoch": 0.07, "learning_rate": 3.664021164021164e-06, "loss": 1.0427, "step": 554 }, { "epoch": 0.07, "learning_rate": 3.6706349206349208e-06, "loss": 1.092, "step": 555 }, { "epoch": 0.07, "learning_rate": 3.6772486772486774e-06, "loss": 1.2969, "step": 556 }, { "epoch": 0.07, "learning_rate": 3.683862433862434e-06, "loss": 0.734, "step": 557 }, { "epoch": 0.07, "learning_rate": 3.690476190476191e-06, "loss": 0.7414, "step": 558 }, { "epoch": 0.07, "learning_rate": 3.6970899470899475e-06, "loss": 0.5969, "step": 559 }, { "epoch": 0.07, "learning_rate": 3.7037037037037037e-06, "loss": 0.7771, "step": 560 }, { "epoch": 0.07, "learning_rate": 3.710317460317461e-06, "loss": 0.9702, "step": 561 }, { "epoch": 0.07, "learning_rate": 3.716931216931217e-06, "loss": 1.2158, "step": 562 }, { "epoch": 0.07, "learning_rate": 3.7235449735449734e-06, "loss": 0.581, "step": 563 }, { "epoch": 0.07, "learning_rate": 3.7301587301587305e-06, "loss": 0.6633, "step": 564 }, { "epoch": 0.07, "learning_rate": 3.7367724867724867e-06, "loss": 0.8743, "step": 565 }, { "epoch": 0.07, "learning_rate": 3.743386243386244e-06, "loss": 0.9763, "step": 566 }, { "epoch": 0.07, "learning_rate": 3.7500000000000005e-06, "loss": 0.8975, "step": 567 }, { "epoch": 0.07, "learning_rate": 3.7566137566137568e-06, "loss": 1.1936, "step": 568 }, { "epoch": 0.07, "learning_rate": 3.763227513227514e-06, "loss": 0.8271, "step": 569 }, { "epoch": 0.07, "learning_rate": 3.76984126984127e-06, "loss": 0.6668, "step": 570 }, { "epoch": 0.07, "learning_rate": 3.7764550264550272e-06, "loss": 0.921, "step": 571 }, { "epoch": 0.07, "learning_rate": 3.7830687830687835e-06, "loss": 0.8734, "step": 572 }, { "epoch": 0.07, "learning_rate": 3.7896825396825398e-06, "loss": 0.9106, "step": 573 }, { "epoch": 0.07, "learning_rate": 3.796296296296297e-06, "loss": 1.1083, "step": 574 }, { "epoch": 0.07, "learning_rate": 3.802910052910053e-06, "loss": 1.2137, "step": 575 }, { "epoch": 0.07, "learning_rate": 3.80952380952381e-06, "loss": 0.7753, "step": 576 }, { "epoch": 0.07, "learning_rate": 3.8161375661375665e-06, "loss": 0.7725, "step": 577 }, { "epoch": 0.07, "learning_rate": 3.822751322751323e-06, "loss": 0.6835, "step": 578 }, { "epoch": 0.07, "learning_rate": 3.82936507936508e-06, "loss": 0.8622, "step": 579 }, { "epoch": 0.07, "learning_rate": 3.835978835978836e-06, "loss": 1.0584, "step": 580 }, { "epoch": 0.07, "learning_rate": 3.842592592592592e-06, "loss": 1.2699, "step": 581 }, { "epoch": 0.07, "learning_rate": 3.8492063492063495e-06, "loss": 0.6119, "step": 582 }, { "epoch": 0.07, "learning_rate": 3.855820105820106e-06, "loss": 0.7579, "step": 583 }, { "epoch": 0.07, "learning_rate": 3.862433862433863e-06, "loss": 0.7685, "step": 584 }, { "epoch": 0.07, "learning_rate": 3.869047619047619e-06, "loss": 0.903, "step": 585 }, { "epoch": 0.07, "learning_rate": 3.875661375661376e-06, "loss": 1.0602, "step": 586 }, { "epoch": 0.07, "learning_rate": 3.8822751322751325e-06, "loss": 1.2032, "step": 587 }, { "epoch": 0.07, "learning_rate": 3.88888888888889e-06, "loss": 0.761, "step": 588 }, { "epoch": 0.07, "learning_rate": 3.895502645502646e-06, "loss": 0.7067, "step": 589 }, { "epoch": 0.07, "learning_rate": 3.902116402116403e-06, "loss": 0.9801, "step": 590 }, { "epoch": 0.07, "learning_rate": 3.908730158730159e-06, "loss": 0.9194, "step": 591 }, { "epoch": 0.07, "learning_rate": 3.9153439153439155e-06, "loss": 1.1058, "step": 592 }, { "epoch": 0.07, "learning_rate": 3.9219576719576726e-06, "loss": 1.1526, "step": 593 }, { "epoch": 0.07, "learning_rate": 3.928571428571429e-06, "loss": 0.6931, "step": 594 }, { "epoch": 0.07, "learning_rate": 3.935185185185186e-06, "loss": 0.5364, "step": 595 }, { "epoch": 0.07, "learning_rate": 3.941798941798942e-06, "loss": 0.7604, "step": 596 }, { "epoch": 0.07, "learning_rate": 3.9484126984126985e-06, "loss": 0.7549, "step": 597 }, { "epoch": 0.07, "learning_rate": 3.9550264550264556e-06, "loss": 0.9008, "step": 598 }, { "epoch": 0.07, "learning_rate": 3.961640211640212e-06, "loss": 1.2732, "step": 599 }, { "epoch": 0.07, "learning_rate": 3.968253968253968e-06, "loss": 1.3118, "step": 600 }, { "epoch": 0.07, "learning_rate": 3.974867724867725e-06, "loss": 0.5177, "step": 601 }, { "epoch": 0.07, "learning_rate": 3.9814814814814814e-06, "loss": 0.9288, "step": 602 }, { "epoch": 0.07, "learning_rate": 3.9880952380952386e-06, "loss": 0.889, "step": 603 }, { "epoch": 0.07, "learning_rate": 3.994708994708995e-06, "loss": 0.9097, "step": 604 }, { "epoch": 0.07, "learning_rate": 4.001322751322752e-06, "loss": 1.0258, "step": 605 }, { "epoch": 0.07, "learning_rate": 4.007936507936508e-06, "loss": 1.3937, "step": 606 }, { "epoch": 0.07, "learning_rate": 4.014550264550265e-06, "loss": 0.6629, "step": 607 }, { "epoch": 0.07, "learning_rate": 4.0211640211640215e-06, "loss": 0.6897, "step": 608 }, { "epoch": 0.07, "learning_rate": 4.027777777777779e-06, "loss": 0.7234, "step": 609 }, { "epoch": 0.07, "learning_rate": 4.034391534391535e-06, "loss": 0.8906, "step": 610 }, { "epoch": 0.07, "learning_rate": 4.041005291005292e-06, "loss": 1.0852, "step": 611 }, { "epoch": 0.07, "learning_rate": 4.047619047619048e-06, "loss": 1.326, "step": 612 }, { "epoch": 0.07, "learning_rate": 4.0542328042328045e-06, "loss": 0.8149, "step": 613 }, { "epoch": 0.07, "learning_rate": 4.060846560846562e-06, "loss": 0.6535, "step": 614 }, { "epoch": 0.07, "learning_rate": 4.067460317460318e-06, "loss": 0.7877, "step": 615 }, { "epoch": 0.07, "learning_rate": 4.074074074074074e-06, "loss": 0.8626, "step": 616 }, { "epoch": 0.07, "learning_rate": 4.080687830687831e-06, "loss": 0.9562, "step": 617 }, { "epoch": 0.07, "learning_rate": 4.0873015873015875e-06, "loss": 1.2418, "step": 618 }, { "epoch": 0.07, "learning_rate": 4.093915343915345e-06, "loss": 0.6763, "step": 619 }, { "epoch": 0.07, "learning_rate": 4.100529100529101e-06, "loss": 0.6025, "step": 620 }, { "epoch": 0.07, "learning_rate": 4.107142857142857e-06, "loss": 0.8231, "step": 621 }, { "epoch": 0.07, "learning_rate": 4.113756613756614e-06, "loss": 0.8341, "step": 622 }, { "epoch": 0.07, "learning_rate": 4.1203703703703705e-06, "loss": 0.8991, "step": 623 }, { "epoch": 0.07, "learning_rate": 4.126984126984127e-06, "loss": 1.0975, "step": 624 }, { "epoch": 0.07, "learning_rate": 4.133597883597884e-06, "loss": 1.375, "step": 625 }, { "epoch": 0.07, "learning_rate": 4.14021164021164e-06, "loss": 0.6696, "step": 626 }, { "epoch": 0.07, "learning_rate": 4.146825396825397e-06, "loss": 0.7348, "step": 627 }, { "epoch": 0.07, "learning_rate": 4.1534391534391535e-06, "loss": 0.8757, "step": 628 }, { "epoch": 0.07, "learning_rate": 4.160052910052911e-06, "loss": 0.7754, "step": 629 }, { "epoch": 0.08, "learning_rate": 4.166666666666667e-06, "loss": 1.0305, "step": 630 }, { "epoch": 0.08, "learning_rate": 4.173280423280424e-06, "loss": 1.3422, "step": 631 }, { "epoch": 0.08, "learning_rate": 4.17989417989418e-06, "loss": 0.5894, "step": 632 }, { "epoch": 0.08, "learning_rate": 4.186507936507937e-06, "loss": 0.7899, "step": 633 }, { "epoch": 0.08, "learning_rate": 4.193121693121694e-06, "loss": 0.7224, "step": 634 }, { "epoch": 0.08, "learning_rate": 4.199735449735451e-06, "loss": 0.9237, "step": 635 }, { "epoch": 0.08, "learning_rate": 4.206349206349207e-06, "loss": 1.0524, "step": 636 }, { "epoch": 0.08, "learning_rate": 4.212962962962963e-06, "loss": 1.3436, "step": 637 }, { "epoch": 0.08, "learning_rate": 4.21957671957672e-06, "loss": 0.6125, "step": 638 }, { "epoch": 0.08, "learning_rate": 4.226190476190477e-06, "loss": 0.7274, "step": 639 }, { "epoch": 0.08, "learning_rate": 4.232804232804233e-06, "loss": 0.7806, "step": 640 }, { "epoch": 0.08, "learning_rate": 4.23941798941799e-06, "loss": 0.8775, "step": 641 }, { "epoch": 0.08, "learning_rate": 4.246031746031746e-06, "loss": 1.03, "step": 642 }, { "epoch": 0.08, "learning_rate": 4.252645502645503e-06, "loss": 1.2089, "step": 643 }, { "epoch": 0.08, "learning_rate": 4.2592592592592596e-06, "loss": 0.9413, "step": 644 }, { "epoch": 0.08, "learning_rate": 4.265873015873016e-06, "loss": 0.6888, "step": 645 }, { "epoch": 0.08, "learning_rate": 4.272486772486773e-06, "loss": 0.8541, "step": 646 }, { "epoch": 0.08, "learning_rate": 4.279100529100529e-06, "loss": 1.0685, "step": 647 }, { "epoch": 0.08, "learning_rate": 4.2857142857142855e-06, "loss": 0.9454, "step": 648 }, { "epoch": 0.08, "learning_rate": 4.2923280423280426e-06, "loss": 1.0637, "step": 649 }, { "epoch": 0.08, "learning_rate": 4.298941798941799e-06, "loss": 1.1763, "step": 650 }, { "epoch": 0.08, "learning_rate": 4.305555555555556e-06, "loss": 0.5584, "step": 651 }, { "epoch": 0.08, "learning_rate": 4.312169312169312e-06, "loss": 0.925, "step": 652 }, { "epoch": 0.08, "learning_rate": 4.318783068783069e-06, "loss": 0.7245, "step": 653 }, { "epoch": 0.08, "learning_rate": 4.3253968253968256e-06, "loss": 0.9949, "step": 654 }, { "epoch": 0.08, "learning_rate": 4.332010582010583e-06, "loss": 1.173, "step": 655 }, { "epoch": 0.08, "learning_rate": 4.338624338624339e-06, "loss": 1.3098, "step": 656 }, { "epoch": 0.08, "learning_rate": 4.345238095238096e-06, "loss": 0.6284, "step": 657 }, { "epoch": 0.08, "learning_rate": 4.351851851851852e-06, "loss": 0.5641, "step": 658 }, { "epoch": 0.08, "learning_rate": 4.358465608465609e-06, "loss": 0.8459, "step": 659 }, { "epoch": 0.08, "learning_rate": 4.365079365079366e-06, "loss": 0.8015, "step": 660 }, { "epoch": 0.08, "learning_rate": 4.371693121693122e-06, "loss": 0.9128, "step": 661 }, { "epoch": 0.08, "learning_rate": 4.378306878306879e-06, "loss": 1.3464, "step": 662 }, { "epoch": 0.08, "learning_rate": 4.384920634920635e-06, "loss": 0.6997, "step": 663 }, { "epoch": 0.08, "learning_rate": 4.3915343915343915e-06, "loss": 0.7065, "step": 664 }, { "epoch": 0.08, "learning_rate": 4.398148148148149e-06, "loss": 0.6693, "step": 665 }, { "epoch": 0.08, "learning_rate": 4.404761904761905e-06, "loss": 0.8119, "step": 666 }, { "epoch": 0.08, "learning_rate": 4.411375661375662e-06, "loss": 1.1023, "step": 667 }, { "epoch": 0.08, "learning_rate": 4.417989417989418e-06, "loss": 1.3596, "step": 668 }, { "epoch": 0.08, "learning_rate": 4.4246031746031745e-06, "loss": 0.8378, "step": 669 }, { "epoch": 0.08, "learning_rate": 4.431216931216932e-06, "loss": 0.5354, "step": 670 }, { "epoch": 0.08, "learning_rate": 4.437830687830688e-06, "loss": 0.8381, "step": 671 }, { "epoch": 0.08, "learning_rate": 4.444444444444444e-06, "loss": 0.8724, "step": 672 }, { "epoch": 0.08, "learning_rate": 4.451058201058201e-06, "loss": 0.9596, "step": 673 }, { "epoch": 0.08, "learning_rate": 4.4576719576719575e-06, "loss": 1.2207, "step": 674 }, { "epoch": 0.08, "learning_rate": 4.464285714285715e-06, "loss": 1.3151, "step": 675 }, { "epoch": 0.08, "learning_rate": 4.470899470899471e-06, "loss": 0.717, "step": 676 }, { "epoch": 0.08, "learning_rate": 4.477513227513228e-06, "loss": 0.8546, "step": 677 }, { "epoch": 0.08, "learning_rate": 4.484126984126984e-06, "loss": 0.893, "step": 678 }, { "epoch": 0.08, "learning_rate": 4.490740740740741e-06, "loss": 0.792, "step": 679 }, { "epoch": 0.08, "learning_rate": 4.497354497354498e-06, "loss": 1.0485, "step": 680 }, { "epoch": 0.08, "learning_rate": 4.503968253968255e-06, "loss": 1.3645, "step": 681 }, { "epoch": 0.08, "learning_rate": 4.510582010582011e-06, "loss": 0.6582, "step": 682 }, { "epoch": 0.08, "learning_rate": 4.517195767195768e-06, "loss": 0.7728, "step": 683 }, { "epoch": 0.08, "learning_rate": 4.523809523809524e-06, "loss": 0.8137, "step": 684 }, { "epoch": 0.08, "learning_rate": 4.530423280423281e-06, "loss": 0.884, "step": 685 }, { "epoch": 0.08, "learning_rate": 4.537037037037038e-06, "loss": 1.1492, "step": 686 }, { "epoch": 0.08, "learning_rate": 4.543650793650794e-06, "loss": 1.3397, "step": 687 }, { "epoch": 0.08, "learning_rate": 4.55026455026455e-06, "loss": 0.8062, "step": 688 }, { "epoch": 0.08, "learning_rate": 4.556878306878307e-06, "loss": 0.7523, "step": 689 }, { "epoch": 0.08, "learning_rate": 4.563492063492064e-06, "loss": 0.7814, "step": 690 }, { "epoch": 0.08, "learning_rate": 4.570105820105821e-06, "loss": 0.7753, "step": 691 }, { "epoch": 0.08, "learning_rate": 4.576719576719577e-06, "loss": 0.9909, "step": 692 }, { "epoch": 0.08, "learning_rate": 4.583333333333333e-06, "loss": 1.1506, "step": 693 }, { "epoch": 0.08, "learning_rate": 4.58994708994709e-06, "loss": 0.8944, "step": 694 }, { "epoch": 0.08, "learning_rate": 4.596560846560847e-06, "loss": 0.5899, "step": 695 }, { "epoch": 0.08, "learning_rate": 4.603174603174604e-06, "loss": 0.8016, "step": 696 }, { "epoch": 0.08, "learning_rate": 4.60978835978836e-06, "loss": 0.8606, "step": 697 }, { "epoch": 0.08, "learning_rate": 4.616402116402117e-06, "loss": 0.9451, "step": 698 }, { "epoch": 0.08, "learning_rate": 4.623015873015873e-06, "loss": 1.1436, "step": 699 }, { "epoch": 0.08, "learning_rate": 4.62962962962963e-06, "loss": 1.1317, "step": 700 }, { "epoch": 0.08, "learning_rate": 4.636243386243387e-06, "loss": 0.641, "step": 701 }, { "epoch": 0.08, "learning_rate": 4.642857142857144e-06, "loss": 0.7295, "step": 702 }, { "epoch": 0.08, "learning_rate": 4.6494708994709e-06, "loss": 0.8403, "step": 703 }, { "epoch": 0.08, "learning_rate": 4.656084656084656e-06, "loss": 0.8971, "step": 704 }, { "epoch": 0.08, "learning_rate": 4.662698412698413e-06, "loss": 1.0157, "step": 705 }, { "epoch": 0.08, "learning_rate": 4.66931216931217e-06, "loss": 1.3307, "step": 706 }, { "epoch": 0.08, "learning_rate": 4.675925925925927e-06, "loss": 0.654, "step": 707 }, { "epoch": 0.08, "learning_rate": 4.682539682539683e-06, "loss": 0.7075, "step": 708 }, { "epoch": 0.08, "learning_rate": 4.689153439153439e-06, "loss": 0.8172, "step": 709 }, { "epoch": 0.08, "learning_rate": 4.695767195767196e-06, "loss": 0.7452, "step": 710 }, { "epoch": 0.08, "learning_rate": 4.702380952380953e-06, "loss": 0.9093, "step": 711 }, { "epoch": 0.08, "learning_rate": 4.708994708994709e-06, "loss": 1.3031, "step": 712 }, { "epoch": 0.08, "learning_rate": 4.715608465608466e-06, "loss": 0.5832, "step": 713 }, { "epoch": 0.09, "learning_rate": 4.722222222222222e-06, "loss": 0.7559, "step": 714 }, { "epoch": 0.09, "learning_rate": 4.728835978835979e-06, "loss": 0.8123, "step": 715 }, { "epoch": 0.09, "learning_rate": 4.735449735449736e-06, "loss": 0.9152, "step": 716 }, { "epoch": 0.09, "learning_rate": 4.742063492063492e-06, "loss": 1.0739, "step": 717 }, { "epoch": 0.09, "learning_rate": 4.748677248677249e-06, "loss": 1.29, "step": 718 }, { "epoch": 0.09, "learning_rate": 4.755291005291005e-06, "loss": 0.9351, "step": 719 }, { "epoch": 0.09, "learning_rate": 4.761904761904762e-06, "loss": 0.728, "step": 720 }, { "epoch": 0.09, "learning_rate": 4.768518518518519e-06, "loss": 0.7407, "step": 721 }, { "epoch": 0.09, "learning_rate": 4.775132275132276e-06, "loss": 0.8779, "step": 722 }, { "epoch": 0.09, "learning_rate": 4.781746031746032e-06, "loss": 1.0151, "step": 723 }, { "epoch": 0.09, "learning_rate": 4.788359788359789e-06, "loss": 1.1292, "step": 724 }, { "epoch": 0.09, "learning_rate": 4.794973544973545e-06, "loss": 1.162, "step": 725 }, { "epoch": 0.09, "learning_rate": 4.8015873015873025e-06, "loss": 0.5285, "step": 726 }, { "epoch": 0.09, "learning_rate": 4.808201058201059e-06, "loss": 0.8145, "step": 727 }, { "epoch": 0.09, "learning_rate": 4.814814814814815e-06, "loss": 0.7543, "step": 728 }, { "epoch": 0.09, "learning_rate": 4.821428571428572e-06, "loss": 0.8286, "step": 729 }, { "epoch": 0.09, "learning_rate": 4.828042328042328e-06, "loss": 1.0366, "step": 730 }, { "epoch": 0.09, "learning_rate": 4.8346560846560855e-06, "loss": 1.2541, "step": 731 }, { "epoch": 0.09, "learning_rate": 4.841269841269842e-06, "loss": 0.876, "step": 732 }, { "epoch": 0.09, "learning_rate": 4.847883597883598e-06, "loss": 0.7519, "step": 733 }, { "epoch": 0.09, "learning_rate": 4.854497354497355e-06, "loss": 0.8464, "step": 734 }, { "epoch": 0.09, "learning_rate": 4.861111111111111e-06, "loss": 0.859, "step": 735 }, { "epoch": 0.09, "learning_rate": 4.867724867724868e-06, "loss": 1.089, "step": 736 }, { "epoch": 0.09, "learning_rate": 4.874338624338625e-06, "loss": 1.2796, "step": 737 }, { "epoch": 0.09, "learning_rate": 4.880952380952381e-06, "loss": 0.9313, "step": 738 }, { "epoch": 0.09, "learning_rate": 4.887566137566138e-06, "loss": 0.6956, "step": 739 }, { "epoch": 0.09, "learning_rate": 4.894179894179894e-06, "loss": 0.7674, "step": 740 }, { "epoch": 0.09, "learning_rate": 4.900793650793651e-06, "loss": 0.9257, "step": 741 }, { "epoch": 0.09, "learning_rate": 4.907407407407408e-06, "loss": 1.0126, "step": 742 }, { "epoch": 0.09, "learning_rate": 4.914021164021164e-06, "loss": 1.1863, "step": 743 }, { "epoch": 0.09, "learning_rate": 4.920634920634921e-06, "loss": 0.6659, "step": 744 }, { "epoch": 0.09, "learning_rate": 4.927248677248677e-06, "loss": 0.7207, "step": 745 }, { "epoch": 0.09, "learning_rate": 4.9338624338624344e-06, "loss": 0.8567, "step": 746 }, { "epoch": 0.09, "learning_rate": 4.940476190476191e-06, "loss": 0.8634, "step": 747 }, { "epoch": 0.09, "learning_rate": 4.947089947089948e-06, "loss": 0.8951, "step": 748 }, { "epoch": 0.09, "learning_rate": 4.953703703703704e-06, "loss": 1.1134, "step": 749 }, { "epoch": 0.09, "learning_rate": 4.960317460317461e-06, "loss": 1.2572, "step": 750 }, { "epoch": 0.09, "learning_rate": 4.966931216931217e-06, "loss": 0.5885, "step": 751 }, { "epoch": 0.09, "learning_rate": 4.973544973544974e-06, "loss": 0.7863, "step": 752 }, { "epoch": 0.09, "learning_rate": 4.980158730158731e-06, "loss": 0.7435, "step": 753 }, { "epoch": 0.09, "learning_rate": 4.986772486772487e-06, "loss": 0.8483, "step": 754 }, { "epoch": 0.09, "learning_rate": 4.993386243386244e-06, "loss": 1.1165, "step": 755 }, { "epoch": 0.09, "learning_rate": 5e-06, "loss": 1.2216, "step": 756 }, { "epoch": 0.09, "learning_rate": 4.999999979337396e-06, "loss": 0.6127, "step": 757 }, { "epoch": 0.09, "learning_rate": 4.9999999173495815e-06, "loss": 0.6134, "step": 758 }, { "epoch": 0.09, "learning_rate": 4.999999814036559e-06, "loss": 0.9492, "step": 759 }, { "epoch": 0.09, "learning_rate": 4.999999669398329e-06, "loss": 0.9343, "step": 760 }, { "epoch": 0.09, "learning_rate": 4.999999483434895e-06, "loss": 1.1546, "step": 761 }, { "epoch": 0.09, "learning_rate": 4.999999256146261e-06, "loss": 1.2448, "step": 762 }, { "epoch": 0.09, "learning_rate": 4.999998987532428e-06, "loss": 0.5912, "step": 763 }, { "epoch": 0.09, "learning_rate": 4.999998677593403e-06, "loss": 0.6474, "step": 764 }, { "epoch": 0.09, "learning_rate": 4.99999832632919e-06, "loss": 0.7995, "step": 765 }, { "epoch": 0.09, "learning_rate": 4.999997933739794e-06, "loss": 0.8795, "step": 766 }, { "epoch": 0.09, "learning_rate": 4.999997499825223e-06, "loss": 1.0734, "step": 767 }, { "epoch": 0.09, "learning_rate": 4.999997024585483e-06, "loss": 1.2309, "step": 768 }, { "epoch": 0.09, "learning_rate": 4.999996508020583e-06, "loss": 0.5523, "step": 769 }, { "epoch": 0.09, "learning_rate": 4.9999959501305315e-06, "loss": 0.6344, "step": 770 }, { "epoch": 0.09, "learning_rate": 4.999995350915337e-06, "loss": 0.7922, "step": 771 }, { "epoch": 0.09, "learning_rate": 4.999994710375008e-06, "loss": 0.746, "step": 772 }, { "epoch": 0.09, "learning_rate": 4.9999940285095584e-06, "loss": 0.944, "step": 773 }, { "epoch": 0.09, "learning_rate": 4.999993305318996e-06, "loss": 1.076, "step": 774 }, { "epoch": 0.09, "learning_rate": 4.999992540803337e-06, "loss": 1.2994, "step": 775 }, { "epoch": 0.09, "learning_rate": 4.999991734962589e-06, "loss": 0.6816, "step": 776 }, { "epoch": 0.09, "learning_rate": 4.999990887796769e-06, "loss": 0.7232, "step": 777 }, { "epoch": 0.09, "learning_rate": 4.999989999305891e-06, "loss": 0.7333, "step": 778 }, { "epoch": 0.09, "learning_rate": 4.999989069489967e-06, "loss": 1.0078, "step": 779 }, { "epoch": 0.09, "learning_rate": 4.999988098349014e-06, "loss": 0.9816, "step": 780 }, { "epoch": 0.09, "learning_rate": 4.999987085883049e-06, "loss": 1.2744, "step": 781 }, { "epoch": 0.09, "learning_rate": 4.999986032092087e-06, "loss": 0.7098, "step": 782 }, { "epoch": 0.09, "learning_rate": 4.999984936976145e-06, "loss": 0.6626, "step": 783 }, { "epoch": 0.09, "learning_rate": 4.9999838005352444e-06, "loss": 0.825, "step": 784 }, { "epoch": 0.09, "learning_rate": 4.9999826227694e-06, "loss": 0.822, "step": 785 }, { "epoch": 0.09, "learning_rate": 4.999981403678634e-06, "loss": 0.9572, "step": 786 }, { "epoch": 0.09, "learning_rate": 4.999980143262966e-06, "loss": 1.1682, "step": 787 }, { "epoch": 0.09, "learning_rate": 4.999978841522416e-06, "loss": 0.5972, "step": 788 }, { "epoch": 0.09, "learning_rate": 4.999977498457006e-06, "loss": 0.7487, "step": 789 }, { "epoch": 0.09, "learning_rate": 4.999976114066758e-06, "loss": 0.7863, "step": 790 }, { "epoch": 0.09, "learning_rate": 4.999974688351695e-06, "loss": 0.6886, "step": 791 }, { "epoch": 0.09, "learning_rate": 4.999973221311842e-06, "loss": 0.962, "step": 792 }, { "epoch": 0.09, "learning_rate": 4.9999717129472194e-06, "loss": 1.1889, "step": 793 }, { "epoch": 0.09, "learning_rate": 4.999970163257856e-06, "loss": 0.9993, "step": 794 }, { "epoch": 0.09, "learning_rate": 4.999968572243776e-06, "loss": 0.6734, "step": 795 }, { "epoch": 0.09, "learning_rate": 4.9999669399050064e-06, "loss": 0.7914, "step": 796 }, { "epoch": 0.09, "learning_rate": 4.999965266241572e-06, "loss": 0.8148, "step": 797 }, { "epoch": 0.1, "learning_rate": 4.999963551253503e-06, "loss": 0.9412, "step": 798 }, { "epoch": 0.1, "learning_rate": 4.999961794940826e-06, "loss": 1.0443, "step": 799 }, { "epoch": 0.1, "learning_rate": 4.9999599973035716e-06, "loss": 1.2415, "step": 800 }, { "epoch": 0.1, "learning_rate": 4.999958158341769e-06, "loss": 0.532, "step": 801 }, { "epoch": 0.1, "learning_rate": 4.999956278055448e-06, "loss": 0.8003, "step": 802 }, { "epoch": 0.1, "learning_rate": 4.999954356444639e-06, "loss": 0.717, "step": 803 }, { "epoch": 0.1, "learning_rate": 4.999952393509375e-06, "loss": 0.934, "step": 804 }, { "epoch": 0.1, "learning_rate": 4.999950389249689e-06, "loss": 1.054, "step": 805 }, { "epoch": 0.1, "learning_rate": 4.999948343665613e-06, "loss": 1.3883, "step": 806 }, { "epoch": 0.1, "learning_rate": 4.999946256757181e-06, "loss": 0.6692, "step": 807 }, { "epoch": 0.1, "learning_rate": 4.999944128524428e-06, "loss": 0.6881, "step": 808 }, { "epoch": 0.1, "learning_rate": 4.999941958967388e-06, "loss": 0.9653, "step": 809 }, { "epoch": 0.1, "learning_rate": 4.9999397480860975e-06, "loss": 0.8037, "step": 810 }, { "epoch": 0.1, "learning_rate": 4.9999374958805944e-06, "loss": 1.0613, "step": 811 }, { "epoch": 0.1, "learning_rate": 4.999935202350914e-06, "loss": 1.2924, "step": 812 }, { "epoch": 0.1, "learning_rate": 4.999932867497095e-06, "loss": 0.9532, "step": 813 }, { "epoch": 0.1, "learning_rate": 4.9999304913191755e-06, "loss": 0.7249, "step": 814 }, { "epoch": 0.1, "learning_rate": 4.999928073817196e-06, "loss": 0.7932, "step": 815 }, { "epoch": 0.1, "learning_rate": 4.999925614991195e-06, "loss": 0.8578, "step": 816 }, { "epoch": 0.1, "learning_rate": 4.999923114841214e-06, "loss": 1.0411, "step": 817 }, { "epoch": 0.1, "learning_rate": 4.999920573367295e-06, "loss": 1.1913, "step": 818 }, { "epoch": 0.1, "learning_rate": 4.999917990569478e-06, "loss": 0.655, "step": 819 }, { "epoch": 0.1, "learning_rate": 4.999915366447807e-06, "loss": 0.5974, "step": 820 }, { "epoch": 0.1, "learning_rate": 4.9999127010023255e-06, "loss": 0.7926, "step": 821 }, { "epoch": 0.1, "learning_rate": 4.999909994233078e-06, "loss": 0.8997, "step": 822 }, { "epoch": 0.1, "learning_rate": 4.999907246140107e-06, "loss": 0.8611, "step": 823 }, { "epoch": 0.1, "learning_rate": 4.999904456723461e-06, "loss": 1.1777, "step": 824 }, { "epoch": 0.1, "learning_rate": 4.999901625983184e-06, "loss": 1.2001, "step": 825 }, { "epoch": 0.1, "learning_rate": 4.999898753919323e-06, "loss": 0.5459, "step": 826 }, { "epoch": 0.1, "learning_rate": 4.999895840531926e-06, "loss": 0.7918, "step": 827 }, { "epoch": 0.1, "learning_rate": 4.9998928858210414e-06, "loss": 0.9011, "step": 828 }, { "epoch": 0.1, "learning_rate": 4.999889889786717e-06, "loss": 0.9861, "step": 829 }, { "epoch": 0.1, "learning_rate": 4.999886852429003e-06, "loss": 1.2312, "step": 830 }, { "epoch": 0.1, "learning_rate": 4.99988377374795e-06, "loss": 1.0755, "step": 831 }, { "epoch": 0.1, "learning_rate": 4.999880653743609e-06, "loss": 0.6482, "step": 832 }, { "epoch": 0.1, "learning_rate": 4.99987749241603e-06, "loss": 0.6711, "step": 833 }, { "epoch": 0.1, "learning_rate": 4.999874289765267e-06, "loss": 0.785, "step": 834 }, { "epoch": 0.1, "learning_rate": 4.999871045791371e-06, "loss": 0.882, "step": 835 }, { "epoch": 0.1, "learning_rate": 4.999867760494399e-06, "loss": 1.1793, "step": 836 }, { "epoch": 0.1, "learning_rate": 4.999864433874401e-06, "loss": 1.1193, "step": 837 }, { "epoch": 0.1, "learning_rate": 4.9998610659314354e-06, "loss": 0.7876, "step": 838 }, { "epoch": 0.1, "learning_rate": 4.999857656665557e-06, "loss": 0.6193, "step": 839 }, { "epoch": 0.1, "learning_rate": 4.999854206076821e-06, "loss": 0.7379, "step": 840 }, { "epoch": 0.1, "learning_rate": 4.999850714165285e-06, "loss": 0.8792, "step": 841 }, { "epoch": 0.1, "learning_rate": 4.9998471809310065e-06, "loss": 0.9507, "step": 842 }, { "epoch": 0.1, "learning_rate": 4.9998436063740455e-06, "loss": 1.0549, "step": 843 }, { "epoch": 0.1, "learning_rate": 4.999839990494459e-06, "loss": 0.9504, "step": 844 }, { "epoch": 0.1, "learning_rate": 4.999836333292309e-06, "loss": 0.5938, "step": 845 }, { "epoch": 0.1, "learning_rate": 4.999832634767654e-06, "loss": 0.7899, "step": 846 }, { "epoch": 0.1, "learning_rate": 4.999828894920555e-06, "loss": 0.8511, "step": 847 }, { "epoch": 0.1, "learning_rate": 4.999825113751076e-06, "loss": 0.9274, "step": 848 }, { "epoch": 0.1, "learning_rate": 4.999821291259277e-06, "loss": 0.9846, "step": 849 }, { "epoch": 0.1, "learning_rate": 4.999817427445222e-06, "loss": 1.2659, "step": 850 }, { "epoch": 0.1, "learning_rate": 4.999813522308977e-06, "loss": 0.7488, "step": 851 }, { "epoch": 0.1, "learning_rate": 4.999809575850603e-06, "loss": 0.7787, "step": 852 }, { "epoch": 0.1, "learning_rate": 4.999805588070168e-06, "loss": 0.817, "step": 853 }, { "epoch": 0.1, "learning_rate": 4.999801558967736e-06, "loss": 1.0566, "step": 854 }, { "epoch": 0.1, "learning_rate": 4.999797488543375e-06, "loss": 1.1191, "step": 855 }, { "epoch": 0.1, "learning_rate": 4.999793376797152e-06, "loss": 1.3959, "step": 856 }, { "epoch": 0.1, "learning_rate": 4.999789223729135e-06, "loss": 0.7302, "step": 857 }, { "epoch": 0.1, "learning_rate": 4.9997850293393915e-06, "loss": 0.9539, "step": 858 }, { "epoch": 0.1, "learning_rate": 4.999780793627992e-06, "loss": 0.7952, "step": 859 }, { "epoch": 0.1, "learning_rate": 4.999776516595006e-06, "loss": 0.9849, "step": 860 }, { "epoch": 0.1, "learning_rate": 4.999772198240504e-06, "loss": 1.1094, "step": 861 }, { "epoch": 0.1, "learning_rate": 4.999767838564559e-06, "loss": 1.1298, "step": 862 }, { "epoch": 0.1, "learning_rate": 4.999763437567241e-06, "loss": 0.6961, "step": 863 }, { "epoch": 0.1, "learning_rate": 4.999758995248624e-06, "loss": 0.6935, "step": 864 }, { "epoch": 0.1, "learning_rate": 4.999754511608781e-06, "loss": 0.9776, "step": 865 }, { "epoch": 0.1, "learning_rate": 4.999749986647787e-06, "loss": 0.8979, "step": 866 }, { "epoch": 0.1, "learning_rate": 4.999745420365715e-06, "loss": 1.1551, "step": 867 }, { "epoch": 0.1, "learning_rate": 4.999740812762641e-06, "loss": 1.0905, "step": 868 }, { "epoch": 0.1, "learning_rate": 4.999736163838643e-06, "loss": 0.9499, "step": 869 }, { "epoch": 0.1, "learning_rate": 4.999731473593795e-06, "loss": 0.6115, "step": 870 }, { "epoch": 0.1, "learning_rate": 4.999726742028177e-06, "loss": 0.6646, "step": 871 }, { "epoch": 0.1, "learning_rate": 4.999721969141866e-06, "loss": 0.6958, "step": 872 }, { "epoch": 0.1, "learning_rate": 4.999717154934941e-06, "loss": 0.8004, "step": 873 }, { "epoch": 0.1, "learning_rate": 4.999712299407482e-06, "loss": 0.9758, "step": 874 }, { "epoch": 0.1, "learning_rate": 4.999707402559569e-06, "loss": 1.0038, "step": 875 }, { "epoch": 0.1, "learning_rate": 4.999702464391283e-06, "loss": 0.6298, "step": 876 }, { "epoch": 0.1, "learning_rate": 4.9996974849027046e-06, "loss": 0.589, "step": 877 }, { "epoch": 0.1, "learning_rate": 4.999692464093917e-06, "loss": 0.8774, "step": 878 }, { "epoch": 0.1, "learning_rate": 4.999687401965004e-06, "loss": 0.9124, "step": 879 }, { "epoch": 0.1, "learning_rate": 4.999682298516049e-06, "loss": 1.1527, "step": 880 }, { "epoch": 0.1, "learning_rate": 4.999677153747134e-06, "loss": 1.1258, "step": 881 }, { "epoch": 0.11, "learning_rate": 4.999671967658348e-06, "loss": 0.7131, "step": 882 }, { "epoch": 0.11, "learning_rate": 4.999666740249774e-06, "loss": 0.7366, "step": 883 }, { "epoch": 0.11, "learning_rate": 4.999661471521498e-06, "loss": 0.7928, "step": 884 }, { "epoch": 0.11, "learning_rate": 4.9996561614736094e-06, "loss": 0.957, "step": 885 }, { "epoch": 0.11, "learning_rate": 4.999650810106193e-06, "loss": 1.0848, "step": 886 }, { "epoch": 0.11, "learning_rate": 4.999645417419341e-06, "loss": 1.0711, "step": 887 }, { "epoch": 0.11, "learning_rate": 4.999639983413139e-06, "loss": 0.5824, "step": 888 }, { "epoch": 0.11, "learning_rate": 4.9996345080876795e-06, "loss": 0.6068, "step": 889 }, { "epoch": 0.11, "learning_rate": 4.999628991443052e-06, "loss": 0.7346, "step": 890 }, { "epoch": 0.11, "learning_rate": 4.999623433479346e-06, "loss": 0.8664, "step": 891 }, { "epoch": 0.11, "learning_rate": 4.999617834196656e-06, "loss": 0.919, "step": 892 }, { "epoch": 0.11, "learning_rate": 4.999612193595074e-06, "loss": 1.0568, "step": 893 }, { "epoch": 0.11, "learning_rate": 4.999606511674691e-06, "loss": 0.701, "step": 894 }, { "epoch": 0.11, "learning_rate": 4.999600788435604e-06, "loss": 0.4677, "step": 895 }, { "epoch": 0.11, "learning_rate": 4.999595023877905e-06, "loss": 0.6699, "step": 896 }, { "epoch": 0.11, "learning_rate": 4.999589218001692e-06, "loss": 0.9475, "step": 897 }, { "epoch": 0.11, "learning_rate": 4.999583370807059e-06, "loss": 0.9066, "step": 898 }, { "epoch": 0.11, "learning_rate": 4.999577482294103e-06, "loss": 1.1902, "step": 899 }, { "epoch": 0.11, "learning_rate": 4.999571552462922e-06, "loss": 1.2964, "step": 900 }, { "epoch": 0.11, "learning_rate": 4.999565581313612e-06, "loss": 0.6679, "step": 901 }, { "epoch": 0.11, "learning_rate": 4.999559568846274e-06, "loss": 0.7247, "step": 902 }, { "epoch": 0.11, "learning_rate": 4.999553515061007e-06, "loss": 0.8603, "step": 903 }, { "epoch": 0.11, "learning_rate": 4.999547419957911e-06, "loss": 0.8243, "step": 904 }, { "epoch": 0.11, "learning_rate": 4.999541283537086e-06, "loss": 1.0913, "step": 905 }, { "epoch": 0.11, "learning_rate": 4.999535105798633e-06, "loss": 1.3589, "step": 906 }, { "epoch": 0.11, "learning_rate": 4.999528886742656e-06, "loss": 0.6371, "step": 907 }, { "epoch": 0.11, "learning_rate": 4.999522626369256e-06, "loss": 0.7565, "step": 908 }, { "epoch": 0.11, "learning_rate": 4.999516324678538e-06, "loss": 0.8881, "step": 909 }, { "epoch": 0.11, "learning_rate": 4.999509981670605e-06, "loss": 0.8745, "step": 910 }, { "epoch": 0.11, "learning_rate": 4.999503597345563e-06, "loss": 0.9054, "step": 911 }, { "epoch": 0.11, "learning_rate": 4.999497171703516e-06, "loss": 1.1276, "step": 912 }, { "epoch": 0.11, "learning_rate": 4.999490704744572e-06, "loss": 0.6135, "step": 913 }, { "epoch": 0.11, "learning_rate": 4.999484196468835e-06, "loss": 0.649, "step": 914 }, { "epoch": 0.11, "learning_rate": 4.999477646876416e-06, "loss": 0.769, "step": 915 }, { "epoch": 0.11, "learning_rate": 4.999471055967422e-06, "loss": 0.9275, "step": 916 }, { "epoch": 0.11, "learning_rate": 4.999464423741961e-06, "loss": 0.9847, "step": 917 }, { "epoch": 0.11, "learning_rate": 4.999457750200143e-06, "loss": 1.1522, "step": 918 }, { "epoch": 0.11, "learning_rate": 4.99945103534208e-06, "loss": 0.7349, "step": 919 }, { "epoch": 0.11, "learning_rate": 4.999444279167881e-06, "loss": 0.5875, "step": 920 }, { "epoch": 0.11, "learning_rate": 4.999437481677658e-06, "loss": 0.7306, "step": 921 }, { "epoch": 0.11, "learning_rate": 4.999430642871524e-06, "loss": 0.8034, "step": 922 }, { "epoch": 0.11, "learning_rate": 4.9994237627495915e-06, "loss": 0.9687, "step": 923 }, { "epoch": 0.11, "learning_rate": 4.9994168413119745e-06, "loss": 1.1108, "step": 924 }, { "epoch": 0.11, "learning_rate": 4.999409878558787e-06, "loss": 1.1926, "step": 925 }, { "epoch": 0.11, "learning_rate": 4.999402874490146e-06, "loss": 0.5711, "step": 926 }, { "epoch": 0.11, "learning_rate": 4.999395829106165e-06, "loss": 0.6667, "step": 927 }, { "epoch": 0.11, "learning_rate": 4.999388742406961e-06, "loss": 0.7207, "step": 928 }, { "epoch": 0.11, "learning_rate": 4.999381614392651e-06, "loss": 0.7457, "step": 929 }, { "epoch": 0.11, "learning_rate": 4.999374445063354e-06, "loss": 1.0742, "step": 930 }, { "epoch": 0.11, "learning_rate": 4.999367234419187e-06, "loss": 1.236, "step": 931 }, { "epoch": 0.11, "learning_rate": 4.99935998246027e-06, "loss": 0.5624, "step": 932 }, { "epoch": 0.11, "learning_rate": 4.999352689186723e-06, "loss": 0.683, "step": 933 }, { "epoch": 0.11, "learning_rate": 4.999345354598667e-06, "loss": 0.8116, "step": 934 }, { "epoch": 0.11, "learning_rate": 4.999337978696221e-06, "loss": 0.9042, "step": 935 }, { "epoch": 0.11, "learning_rate": 4.99933056147951e-06, "loss": 1.0035, "step": 936 }, { "epoch": 0.11, "learning_rate": 4.999323102948655e-06, "loss": 1.3898, "step": 937 }, { "epoch": 0.11, "learning_rate": 4.999315603103778e-06, "loss": 0.7812, "step": 938 }, { "epoch": 0.11, "learning_rate": 4.999308061945006e-06, "loss": 0.6775, "step": 939 }, { "epoch": 0.11, "learning_rate": 4.999300479472461e-06, "loss": 0.7116, "step": 940 }, { "epoch": 0.11, "learning_rate": 4.99929285568627e-06, "loss": 0.8152, "step": 941 }, { "epoch": 0.11, "learning_rate": 4.999285190586558e-06, "loss": 0.9962, "step": 942 }, { "epoch": 0.11, "learning_rate": 4.999277484173452e-06, "loss": 1.2441, "step": 943 }, { "epoch": 0.11, "learning_rate": 4.99926973644708e-06, "loss": 0.6141, "step": 944 }, { "epoch": 0.11, "learning_rate": 4.999261947407568e-06, "loss": 0.7621, "step": 945 }, { "epoch": 0.11, "learning_rate": 4.999254117055049e-06, "loss": 0.68, "step": 946 }, { "epoch": 0.11, "learning_rate": 4.999246245389647e-06, "loss": 1.0553, "step": 947 }, { "epoch": 0.11, "learning_rate": 4.999238332411497e-06, "loss": 0.9097, "step": 948 }, { "epoch": 0.11, "learning_rate": 4.999230378120726e-06, "loss": 1.0996, "step": 949 }, { "epoch": 0.11, "learning_rate": 4.999222382517468e-06, "loss": 1.3025, "step": 950 }, { "epoch": 0.11, "learning_rate": 4.999214345601855e-06, "loss": 0.6139, "step": 951 }, { "epoch": 0.11, "learning_rate": 4.999206267374018e-06, "loss": 0.8279, "step": 952 }, { "epoch": 0.11, "learning_rate": 4.9991981478340925e-06, "loss": 0.7665, "step": 953 }, { "epoch": 0.11, "learning_rate": 4.999189986982211e-06, "loss": 0.8675, "step": 954 }, { "epoch": 0.11, "learning_rate": 4.9991817848185095e-06, "loss": 1.0778, "step": 955 }, { "epoch": 0.11, "learning_rate": 4.9991735413431246e-06, "loss": 1.3854, "step": 956 }, { "epoch": 0.11, "learning_rate": 4.99916525655619e-06, "loss": 0.5359, "step": 957 }, { "epoch": 0.11, "learning_rate": 4.999156930457845e-06, "loss": 0.6097, "step": 958 }, { "epoch": 0.11, "learning_rate": 4.999148563048225e-06, "loss": 0.6829, "step": 959 }, { "epoch": 0.11, "learning_rate": 4.999140154327471e-06, "loss": 0.8957, "step": 960 }, { "epoch": 0.11, "learning_rate": 4.99913170429572e-06, "loss": 0.9681, "step": 961 }, { "epoch": 0.11, "learning_rate": 4.999123212953112e-06, "loss": 1.0997, "step": 962 }, { "epoch": 0.11, "learning_rate": 4.999114680299787e-06, "loss": 0.76, "step": 963 }, { "epoch": 0.11, "learning_rate": 4.9991061063358876e-06, "loss": 0.5425, "step": 964 }, { "epoch": 0.11, "learning_rate": 4.999097491061554e-06, "loss": 0.7393, "step": 965 }, { "epoch": 0.12, "learning_rate": 4.99908883447693e-06, "loss": 0.8924, "step": 966 }, { "epoch": 0.12, "learning_rate": 4.999080136582157e-06, "loss": 1.0893, "step": 967 }, { "epoch": 0.12, "learning_rate": 4.99907139737738e-06, "loss": 1.2559, "step": 968 }, { "epoch": 0.12, "learning_rate": 4.999062616862743e-06, "loss": 0.9365, "step": 969 }, { "epoch": 0.12, "learning_rate": 4.999053795038391e-06, "loss": 0.7254, "step": 970 }, { "epoch": 0.12, "learning_rate": 4.999044931904471e-06, "loss": 0.7063, "step": 971 }, { "epoch": 0.12, "learning_rate": 4.999036027461128e-06, "loss": 0.8624, "step": 972 }, { "epoch": 0.12, "learning_rate": 4.999027081708509e-06, "loss": 0.9133, "step": 973 }, { "epoch": 0.12, "learning_rate": 4.999018094646764e-06, "loss": 1.0233, "step": 974 }, { "epoch": 0.12, "learning_rate": 4.999009066276041e-06, "loss": 1.3339, "step": 975 }, { "epoch": 0.12, "learning_rate": 4.998999996596486e-06, "loss": 0.461, "step": 976 }, { "epoch": 0.12, "learning_rate": 4.9989908856082524e-06, "loss": 0.6199, "step": 977 }, { "epoch": 0.12, "learning_rate": 4.99898173331149e-06, "loss": 0.7489, "step": 978 }, { "epoch": 0.12, "learning_rate": 4.9989725397063504e-06, "loss": 0.9199, "step": 979 }, { "epoch": 0.12, "learning_rate": 4.998963304792984e-06, "loss": 1.113, "step": 980 }, { "epoch": 0.12, "learning_rate": 4.998954028571545e-06, "loss": 1.2565, "step": 981 }, { "epoch": 0.12, "learning_rate": 4.998944711042186e-06, "loss": 0.6554, "step": 982 }, { "epoch": 0.12, "learning_rate": 4.998935352205061e-06, "loss": 0.7332, "step": 983 }, { "epoch": 0.12, "learning_rate": 4.9989259520603255e-06, "loss": 0.9026, "step": 984 }, { "epoch": 0.12, "learning_rate": 4.998916510608135e-06, "loss": 0.9092, "step": 985 }, { "epoch": 0.12, "learning_rate": 4.998907027848643e-06, "loss": 0.9689, "step": 986 }, { "epoch": 0.12, "learning_rate": 4.99889750378201e-06, "loss": 1.1832, "step": 987 }, { "epoch": 0.12, "learning_rate": 4.99888793840839e-06, "loss": 0.5037, "step": 988 }, { "epoch": 0.12, "learning_rate": 4.9988783317279435e-06, "loss": 0.7167, "step": 989 }, { "epoch": 0.12, "learning_rate": 4.998868683740828e-06, "loss": 0.7404, "step": 990 }, { "epoch": 0.12, "learning_rate": 4.998858994447205e-06, "loss": 1.004, "step": 991 }, { "epoch": 0.12, "learning_rate": 4.998849263847232e-06, "loss": 1.0215, "step": 992 }, { "epoch": 0.12, "learning_rate": 4.998839491941071e-06, "loss": 1.1967, "step": 993 }, { "epoch": 0.12, "learning_rate": 4.9988296787288835e-06, "loss": 0.9022, "step": 994 }, { "epoch": 0.12, "learning_rate": 4.998819824210832e-06, "loss": 0.6256, "step": 995 }, { "epoch": 0.12, "learning_rate": 4.9988099283870796e-06, "loss": 0.7501, "step": 996 }, { "epoch": 0.12, "learning_rate": 4.9987999912577885e-06, "loss": 0.9911, "step": 997 }, { "epoch": 0.12, "learning_rate": 4.998790012823124e-06, "loss": 1.0301, "step": 998 }, { "epoch": 0.12, "learning_rate": 4.9987799930832515e-06, "loss": 1.2141, "step": 999 }, { "epoch": 0.12, "learning_rate": 4.998769932038335e-06, "loss": 1.1174, "step": 1000 }, { "epoch": 0.12, "learning_rate": 4.998759829688543e-06, "loss": 0.5916, "step": 1001 }, { "epoch": 0.12, "learning_rate": 4.9987496860340415e-06, "loss": 0.6645, "step": 1002 }, { "epoch": 0.12, "learning_rate": 4.998739501074997e-06, "loss": 0.6315, "step": 1003 }, { "epoch": 0.12, "learning_rate": 4.998729274811579e-06, "loss": 0.7505, "step": 1004 }, { "epoch": 0.12, "learning_rate": 4.9987190072439575e-06, "loss": 1.045, "step": 1005 }, { "epoch": 0.12, "learning_rate": 4.9987086983723e-06, "loss": 1.2207, "step": 1006 }, { "epoch": 0.12, "learning_rate": 4.998698348196778e-06, "loss": 0.5487, "step": 1007 }, { "epoch": 0.12, "learning_rate": 4.998687956717563e-06, "loss": 0.7296, "step": 1008 }, { "epoch": 0.12, "learning_rate": 4.998677523934827e-06, "loss": 0.8494, "step": 1009 }, { "epoch": 0.12, "learning_rate": 4.998667049848741e-06, "loss": 0.8809, "step": 1010 }, { "epoch": 0.12, "learning_rate": 4.998656534459479e-06, "loss": 0.9637, "step": 1011 }, { "epoch": 0.12, "learning_rate": 4.9986459777672155e-06, "loss": 1.2391, "step": 1012 }, { "epoch": 0.12, "learning_rate": 4.998635379772124e-06, "loss": 0.7143, "step": 1013 }, { "epoch": 0.12, "learning_rate": 4.99862474047438e-06, "loss": 0.5076, "step": 1014 }, { "epoch": 0.12, "learning_rate": 4.99861405987416e-06, "loss": 0.7954, "step": 1015 }, { "epoch": 0.12, "learning_rate": 4.9986033379716396e-06, "loss": 0.8237, "step": 1016 }, { "epoch": 0.12, "learning_rate": 4.998592574766996e-06, "loss": 0.9663, "step": 1017 }, { "epoch": 0.12, "learning_rate": 4.998581770260408e-06, "loss": 1.1343, "step": 1018 }, { "epoch": 0.12, "learning_rate": 4.998570924452053e-06, "loss": 0.7448, "step": 1019 }, { "epoch": 0.12, "learning_rate": 4.998560037342112e-06, "loss": 0.636, "step": 1020 }, { "epoch": 0.12, "learning_rate": 4.9985491089307625e-06, "loss": 0.788, "step": 1021 }, { "epoch": 0.12, "learning_rate": 4.998538139218187e-06, "loss": 0.8372, "step": 1022 }, { "epoch": 0.12, "learning_rate": 4.998527128204568e-06, "loss": 0.8665, "step": 1023 }, { "epoch": 0.12, "learning_rate": 4.998516075890084e-06, "loss": 0.9912, "step": 1024 }, { "epoch": 0.12, "learning_rate": 4.998504982274921e-06, "loss": 1.4487, "step": 1025 }, { "epoch": 0.12, "learning_rate": 4.99849384735926e-06, "loss": 0.6384, "step": 1026 }, { "epoch": 0.12, "learning_rate": 4.998482671143287e-06, "loss": 0.6875, "step": 1027 }, { "epoch": 0.12, "learning_rate": 4.998471453627184e-06, "loss": 0.7269, "step": 1028 }, { "epoch": 0.12, "learning_rate": 4.99846019481114e-06, "loss": 0.8922, "step": 1029 }, { "epoch": 0.12, "learning_rate": 4.998448894695339e-06, "loss": 0.9849, "step": 1030 }, { "epoch": 0.12, "learning_rate": 4.998437553279968e-06, "loss": 1.2545, "step": 1031 }, { "epoch": 0.12, "learning_rate": 4.998426170565213e-06, "loss": 0.5759, "step": 1032 }, { "epoch": 0.12, "learning_rate": 4.998414746551266e-06, "loss": 0.769, "step": 1033 }, { "epoch": 0.12, "learning_rate": 4.998403281238313e-06, "loss": 0.7999, "step": 1034 }, { "epoch": 0.12, "learning_rate": 4.998391774626545e-06, "loss": 1.0019, "step": 1035 }, { "epoch": 0.12, "learning_rate": 4.99838022671615e-06, "loss": 1.0974, "step": 1036 }, { "epoch": 0.12, "learning_rate": 4.99836863750732e-06, "loss": 1.2756, "step": 1037 }, { "epoch": 0.12, "learning_rate": 4.998357007000247e-06, "loss": 0.824, "step": 1038 }, { "epoch": 0.12, "learning_rate": 4.998345335195124e-06, "loss": 0.7142, "step": 1039 }, { "epoch": 0.12, "learning_rate": 4.998333622092142e-06, "loss": 0.6845, "step": 1040 }, { "epoch": 0.12, "learning_rate": 4.998321867691496e-06, "loss": 0.8512, "step": 1041 }, { "epoch": 0.12, "learning_rate": 4.99831007199338e-06, "loss": 1.077, "step": 1042 }, { "epoch": 0.12, "learning_rate": 4.998298234997989e-06, "loss": 1.2394, "step": 1043 }, { "epoch": 0.12, "learning_rate": 4.9982863567055184e-06, "loss": 0.8383, "step": 1044 }, { "epoch": 0.12, "learning_rate": 4.998274437116164e-06, "loss": 0.6694, "step": 1045 }, { "epoch": 0.12, "learning_rate": 4.998262476230124e-06, "loss": 0.8515, "step": 1046 }, { "epoch": 0.12, "learning_rate": 4.998250474047596e-06, "loss": 0.6774, "step": 1047 }, { "epoch": 0.12, "learning_rate": 4.998238430568778e-06, "loss": 0.8694, "step": 1048 }, { "epoch": 0.12, "learning_rate": 4.998226345793869e-06, "loss": 1.135, "step": 1049 }, { "epoch": 0.13, "learning_rate": 4.998214219723069e-06, "loss": 1.1887, "step": 1050 }, { "epoch": 0.13, "learning_rate": 4.998202052356578e-06, "loss": 0.6328, "step": 1051 }, { "epoch": 0.13, "learning_rate": 4.998189843694598e-06, "loss": 0.6915, "step": 1052 }, { "epoch": 0.13, "learning_rate": 4.99817759373733e-06, "loss": 0.8031, "step": 1053 }, { "epoch": 0.13, "learning_rate": 4.998165302484977e-06, "loss": 0.8773, "step": 1054 }, { "epoch": 0.13, "learning_rate": 4.998152969937741e-06, "loss": 1.0197, "step": 1055 }, { "epoch": 0.13, "learning_rate": 4.998140596095827e-06, "loss": 1.1694, "step": 1056 }, { "epoch": 0.13, "learning_rate": 4.99812818095944e-06, "loss": 0.716, "step": 1057 }, { "epoch": 0.13, "learning_rate": 4.998115724528784e-06, "loss": 0.761, "step": 1058 }, { "epoch": 0.13, "learning_rate": 4.9981032268040665e-06, "loss": 0.8197, "step": 1059 }, { "epoch": 0.13, "learning_rate": 4.998090687785492e-06, "loss": 0.934, "step": 1060 }, { "epoch": 0.13, "learning_rate": 4.998078107473269e-06, "loss": 1.0667, "step": 1061 }, { "epoch": 0.13, "learning_rate": 4.998065485867605e-06, "loss": 1.2789, "step": 1062 }, { "epoch": 0.13, "learning_rate": 4.99805282296871e-06, "loss": 0.5526, "step": 1063 }, { "epoch": 0.13, "learning_rate": 4.998040118776791e-06, "loss": 0.6422, "step": 1064 }, { "epoch": 0.13, "learning_rate": 4.998027373292061e-06, "loss": 0.8306, "step": 1065 }, { "epoch": 0.13, "learning_rate": 4.998014586514727e-06, "loss": 0.7911, "step": 1066 }, { "epoch": 0.13, "learning_rate": 4.998001758445002e-06, "loss": 1.1498, "step": 1067 }, { "epoch": 0.13, "learning_rate": 4.9979888890831e-06, "loss": 1.253, "step": 1068 }, { "epoch": 0.13, "learning_rate": 4.9979759784292316e-06, "loss": 1.0152, "step": 1069 }, { "epoch": 0.13, "learning_rate": 4.997963026483611e-06, "loss": 0.5453, "step": 1070 }, { "epoch": 0.13, "learning_rate": 4.997950033246451e-06, "loss": 0.9041, "step": 1071 }, { "epoch": 0.13, "learning_rate": 4.997936998717968e-06, "loss": 0.7732, "step": 1072 }, { "epoch": 0.13, "learning_rate": 4.997923922898376e-06, "loss": 0.8784, "step": 1073 }, { "epoch": 0.13, "learning_rate": 4.9979108057878924e-06, "loss": 1.0691, "step": 1074 }, { "epoch": 0.13, "learning_rate": 4.997897647386734e-06, "loss": 1.1321, "step": 1075 }, { "epoch": 0.13, "learning_rate": 4.997884447695118e-06, "loss": 0.635, "step": 1076 }, { "epoch": 0.13, "learning_rate": 4.997871206713262e-06, "loss": 0.5515, "step": 1077 }, { "epoch": 0.13, "learning_rate": 4.997857924441386e-06, "loss": 0.8344, "step": 1078 }, { "epoch": 0.13, "learning_rate": 4.997844600879708e-06, "loss": 0.888, "step": 1079 }, { "epoch": 0.13, "learning_rate": 4.99783123602845e-06, "loss": 1.1334, "step": 1080 }, { "epoch": 0.13, "learning_rate": 4.997817829887831e-06, "loss": 1.2745, "step": 1081 }, { "epoch": 0.13, "learning_rate": 4.997804382458074e-06, "loss": 0.5398, "step": 1082 }, { "epoch": 0.13, "learning_rate": 4.997790893739402e-06, "loss": 0.837, "step": 1083 }, { "epoch": 0.13, "learning_rate": 4.997777363732036e-06, "loss": 0.8115, "step": 1084 }, { "epoch": 0.13, "learning_rate": 4.997763792436201e-06, "loss": 0.7825, "step": 1085 }, { "epoch": 0.13, "learning_rate": 4.997750179852121e-06, "loss": 0.9957, "step": 1086 }, { "epoch": 0.13, "learning_rate": 4.99773652598002e-06, "loss": 1.2649, "step": 1087 }, { "epoch": 0.13, "learning_rate": 4.9977228308201266e-06, "loss": 0.5632, "step": 1088 }, { "epoch": 0.13, "learning_rate": 4.997709094372664e-06, "loss": 0.5362, "step": 1089 }, { "epoch": 0.13, "learning_rate": 4.99769531663786e-06, "loss": 0.8364, "step": 1090 }, { "epoch": 0.13, "learning_rate": 4.9976814976159434e-06, "loss": 0.8486, "step": 1091 }, { "epoch": 0.13, "learning_rate": 4.997667637307142e-06, "loss": 0.9121, "step": 1092 }, { "epoch": 0.13, "learning_rate": 4.997653735711685e-06, "loss": 1.2681, "step": 1093 }, { "epoch": 0.13, "learning_rate": 4.997639792829802e-06, "loss": 0.7464, "step": 1094 }, { "epoch": 0.13, "learning_rate": 4.997625808661724e-06, "loss": 0.6746, "step": 1095 }, { "epoch": 0.13, "learning_rate": 4.997611783207682e-06, "loss": 0.7521, "step": 1096 }, { "epoch": 0.13, "learning_rate": 4.997597716467907e-06, "loss": 0.8268, "step": 1097 }, { "epoch": 0.13, "learning_rate": 4.997583608442632e-06, "loss": 0.9209, "step": 1098 }, { "epoch": 0.13, "learning_rate": 4.997569459132092e-06, "loss": 1.2273, "step": 1099 }, { "epoch": 0.13, "learning_rate": 4.9975552685365185e-06, "loss": 1.1861, "step": 1100 }, { "epoch": 0.13, "learning_rate": 4.997541036656146e-06, "loss": 0.4956, "step": 1101 }, { "epoch": 0.13, "learning_rate": 4.997526763491211e-06, "loss": 0.6743, "step": 1102 }, { "epoch": 0.13, "learning_rate": 4.997512449041949e-06, "loss": 0.7964, "step": 1103 }, { "epoch": 0.13, "learning_rate": 4.997498093308597e-06, "loss": 0.711, "step": 1104 }, { "epoch": 0.13, "learning_rate": 4.997483696291392e-06, "loss": 0.9311, "step": 1105 }, { "epoch": 0.13, "learning_rate": 4.997469257990572e-06, "loss": 1.2161, "step": 1106 }, { "epoch": 0.13, "learning_rate": 4.997454778406375e-06, "loss": 0.556, "step": 1107 }, { "epoch": 0.13, "learning_rate": 4.997440257539041e-06, "loss": 0.7422, "step": 1108 }, { "epoch": 0.13, "learning_rate": 4.9974256953888105e-06, "loss": 0.9067, "step": 1109 }, { "epoch": 0.13, "learning_rate": 4.997411091955924e-06, "loss": 0.8485, "step": 1110 }, { "epoch": 0.13, "learning_rate": 4.997396447240622e-06, "loss": 1.0279, "step": 1111 }, { "epoch": 0.13, "learning_rate": 4.997381761243147e-06, "loss": 1.246, "step": 1112 }, { "epoch": 0.13, "learning_rate": 4.997367033963742e-06, "loss": 0.7919, "step": 1113 }, { "epoch": 0.13, "learning_rate": 4.99735226540265e-06, "loss": 0.7786, "step": 1114 }, { "epoch": 0.13, "learning_rate": 4.997337455560117e-06, "loss": 0.6273, "step": 1115 }, { "epoch": 0.13, "learning_rate": 4.997322604436385e-06, "loss": 0.7901, "step": 1116 }, { "epoch": 0.13, "learning_rate": 4.9973077120317015e-06, "loss": 0.8798, "step": 1117 }, { "epoch": 0.13, "learning_rate": 4.997292778346312e-06, "loss": 1.1775, "step": 1118 }, { "epoch": 0.13, "learning_rate": 4.997277803380462e-06, "loss": 0.9794, "step": 1119 }, { "epoch": 0.13, "learning_rate": 4.997262787134402e-06, "loss": 0.736, "step": 1120 }, { "epoch": 0.13, "learning_rate": 4.997247729608378e-06, "loss": 0.6387, "step": 1121 }, { "epoch": 0.13, "learning_rate": 4.99723263080264e-06, "loss": 0.8116, "step": 1122 }, { "epoch": 0.13, "learning_rate": 4.997217490717437e-06, "loss": 0.895, "step": 1123 }, { "epoch": 0.13, "learning_rate": 4.997202309353018e-06, "loss": 0.9884, "step": 1124 }, { "epoch": 0.13, "learning_rate": 4.997187086709636e-06, "loss": 1.2635, "step": 1125 }, { "epoch": 0.13, "learning_rate": 4.997171822787542e-06, "loss": 0.7327, "step": 1126 }, { "epoch": 0.13, "learning_rate": 4.997156517586989e-06, "loss": 0.6925, "step": 1127 }, { "epoch": 0.13, "learning_rate": 4.997141171108228e-06, "loss": 0.8961, "step": 1128 }, { "epoch": 0.13, "learning_rate": 4.997125783351515e-06, "loss": 0.9345, "step": 1129 }, { "epoch": 0.13, "learning_rate": 4.997110354317103e-06, "loss": 1.1652, "step": 1130 }, { "epoch": 0.13, "learning_rate": 4.997094884005247e-06, "loss": 1.1567, "step": 1131 }, { "epoch": 0.13, "learning_rate": 4.997079372416203e-06, "loss": 0.7325, "step": 1132 }, { "epoch": 0.13, "learning_rate": 4.997063819550228e-06, "loss": 0.7345, "step": 1133 }, { "epoch": 0.14, "learning_rate": 4.997048225407578e-06, "loss": 0.7782, "step": 1134 }, { "epoch": 0.14, "learning_rate": 4.997032589988511e-06, "loss": 0.8147, "step": 1135 }, { "epoch": 0.14, "learning_rate": 4.9970169132932864e-06, "loss": 0.8359, "step": 1136 }, { "epoch": 0.14, "learning_rate": 4.997001195322163e-06, "loss": 1.1685, "step": 1137 }, { "epoch": 0.14, "learning_rate": 4.9969854360754e-06, "loss": 0.5924, "step": 1138 }, { "epoch": 0.14, "learning_rate": 4.996969635553258e-06, "loss": 0.6141, "step": 1139 }, { "epoch": 0.14, "learning_rate": 4.996953793755998e-06, "loss": 0.7781, "step": 1140 }, { "epoch": 0.14, "learning_rate": 4.9969379106838834e-06, "loss": 0.8743, "step": 1141 }, { "epoch": 0.14, "learning_rate": 4.9969219863371745e-06, "loss": 0.998, "step": 1142 }, { "epoch": 0.14, "learning_rate": 4.9969060207161365e-06, "loss": 1.0399, "step": 1143 }, { "epoch": 0.14, "learning_rate": 4.996890013821032e-06, "loss": 1.0696, "step": 1144 }, { "epoch": 0.14, "learning_rate": 4.996873965652126e-06, "loss": 0.5114, "step": 1145 }, { "epoch": 0.14, "learning_rate": 4.996857876209684e-06, "loss": 0.8862, "step": 1146 }, { "epoch": 0.14, "learning_rate": 4.996841745493972e-06, "loss": 0.9769, "step": 1147 }, { "epoch": 0.14, "learning_rate": 4.9968255735052565e-06, "loss": 1.0691, "step": 1148 }, { "epoch": 0.14, "learning_rate": 4.996809360243804e-06, "loss": 1.0714, "step": 1149 }, { "epoch": 0.14, "learning_rate": 4.996793105709885e-06, "loss": 1.2348, "step": 1150 }, { "epoch": 0.14, "learning_rate": 4.996776809903765e-06, "loss": 0.582, "step": 1151 }, { "epoch": 0.14, "learning_rate": 4.996760472825715e-06, "loss": 0.6732, "step": 1152 }, { "epoch": 0.14, "learning_rate": 4.996744094476005e-06, "loss": 0.7482, "step": 1153 }, { "epoch": 0.14, "learning_rate": 4.9967276748549054e-06, "loss": 0.8575, "step": 1154 }, { "epoch": 0.14, "learning_rate": 4.996711213962688e-06, "loss": 1.0477, "step": 1155 }, { "epoch": 0.14, "learning_rate": 4.996694711799625e-06, "loss": 1.4276, "step": 1156 }, { "epoch": 0.14, "learning_rate": 4.996678168365989e-06, "loss": 0.6389, "step": 1157 }, { "epoch": 0.14, "learning_rate": 4.996661583662053e-06, "loss": 0.7081, "step": 1158 }, { "epoch": 0.14, "learning_rate": 4.9966449576880915e-06, "loss": 0.7413, "step": 1159 }, { "epoch": 0.14, "learning_rate": 4.996628290444379e-06, "loss": 0.8838, "step": 1160 }, { "epoch": 0.14, "learning_rate": 4.9966115819311926e-06, "loss": 1.0761, "step": 1161 }, { "epoch": 0.14, "learning_rate": 4.996594832148806e-06, "loss": 1.2899, "step": 1162 }, { "epoch": 0.14, "learning_rate": 4.996578041097499e-06, "loss": 0.7416, "step": 1163 }, { "epoch": 0.14, "learning_rate": 4.996561208777546e-06, "loss": 0.6115, "step": 1164 }, { "epoch": 0.14, "learning_rate": 4.996544335189227e-06, "loss": 0.8391, "step": 1165 }, { "epoch": 0.14, "learning_rate": 4.9965274203328216e-06, "loss": 0.9283, "step": 1166 }, { "epoch": 0.14, "learning_rate": 4.9965104642086075e-06, "loss": 1.0988, "step": 1167 }, { "epoch": 0.14, "learning_rate": 4.996493466816866e-06, "loss": 1.0505, "step": 1168 }, { "epoch": 0.14, "learning_rate": 4.9964764281578784e-06, "loss": 0.7603, "step": 1169 }, { "epoch": 0.14, "learning_rate": 4.996459348231926e-06, "loss": 0.7509, "step": 1170 }, { "epoch": 0.14, "learning_rate": 4.996442227039291e-06, "loss": 0.785, "step": 1171 }, { "epoch": 0.14, "learning_rate": 4.996425064580256e-06, "loss": 0.8885, "step": 1172 }, { "epoch": 0.14, "learning_rate": 4.9964078608551054e-06, "loss": 0.9901, "step": 1173 }, { "epoch": 0.14, "learning_rate": 4.996390615864123e-06, "loss": 1.0763, "step": 1174 }, { "epoch": 0.14, "learning_rate": 4.996373329607596e-06, "loss": 1.1923, "step": 1175 }, { "epoch": 0.14, "learning_rate": 4.9963560020858065e-06, "loss": 0.5967, "step": 1176 }, { "epoch": 0.14, "learning_rate": 4.996338633299044e-06, "loss": 0.6938, "step": 1177 }, { "epoch": 0.14, "learning_rate": 4.996321223247593e-06, "loss": 0.9233, "step": 1178 }, { "epoch": 0.14, "learning_rate": 4.996303771931744e-06, "loss": 0.9077, "step": 1179 }, { "epoch": 0.14, "learning_rate": 4.9962862793517844e-06, "loss": 1.0839, "step": 1180 }, { "epoch": 0.14, "learning_rate": 4.9962687455080024e-06, "loss": 1.1658, "step": 1181 }, { "epoch": 0.14, "learning_rate": 4.9962511704006885e-06, "loss": 0.6026, "step": 1182 }, { "epoch": 0.14, "learning_rate": 4.996233554030134e-06, "loss": 0.7299, "step": 1183 }, { "epoch": 0.14, "learning_rate": 4.996215896396629e-06, "loss": 0.8218, "step": 1184 }, { "epoch": 0.14, "learning_rate": 4.996198197500466e-06, "loss": 0.8217, "step": 1185 }, { "epoch": 0.14, "learning_rate": 4.996180457341937e-06, "loss": 0.9702, "step": 1186 }, { "epoch": 0.14, "learning_rate": 4.9961626759213355e-06, "loss": 1.1988, "step": 1187 }, { "epoch": 0.14, "learning_rate": 4.996144853238956e-06, "loss": 0.8195, "step": 1188 }, { "epoch": 0.14, "learning_rate": 4.996126989295093e-06, "loss": 0.6442, "step": 1189 }, { "epoch": 0.14, "learning_rate": 4.99610908409004e-06, "loss": 0.8069, "step": 1190 }, { "epoch": 0.14, "learning_rate": 4.996091137624095e-06, "loss": 0.8997, "step": 1191 }, { "epoch": 0.14, "learning_rate": 4.9960731498975545e-06, "loss": 0.9124, "step": 1192 }, { "epoch": 0.14, "learning_rate": 4.996055120910716e-06, "loss": 1.1331, "step": 1193 }, { "epoch": 0.14, "learning_rate": 4.996037050663875e-06, "loss": 0.8692, "step": 1194 }, { "epoch": 0.14, "learning_rate": 4.996018939157334e-06, "loss": 0.6692, "step": 1195 }, { "epoch": 0.14, "learning_rate": 4.9960007863913896e-06, "loss": 0.679, "step": 1196 }, { "epoch": 0.14, "learning_rate": 4.995982592366343e-06, "loss": 0.8061, "step": 1197 }, { "epoch": 0.14, "learning_rate": 4.9959643570824945e-06, "loss": 0.9413, "step": 1198 }, { "epoch": 0.14, "learning_rate": 4.995946080540146e-06, "loss": 1.0625, "step": 1199 }, { "epoch": 0.14, "learning_rate": 4.995927762739599e-06, "loss": 1.2108, "step": 1200 }, { "epoch": 0.14, "learning_rate": 4.995909403681157e-06, "loss": 0.6838, "step": 1201 }, { "epoch": 0.14, "learning_rate": 4.995891003365123e-06, "loss": 0.6948, "step": 1202 }, { "epoch": 0.14, "learning_rate": 4.995872561791801e-06, "loss": 0.9304, "step": 1203 }, { "epoch": 0.14, "learning_rate": 4.995854078961497e-06, "loss": 0.9292, "step": 1204 }, { "epoch": 0.14, "learning_rate": 4.995835554874515e-06, "loss": 1.0355, "step": 1205 }, { "epoch": 0.14, "learning_rate": 4.995816989531162e-06, "loss": 1.2294, "step": 1206 }, { "epoch": 0.14, "learning_rate": 4.995798382931745e-06, "loss": 0.6859, "step": 1207 }, { "epoch": 0.14, "learning_rate": 4.9957797350765705e-06, "loss": 0.8073, "step": 1208 }, { "epoch": 0.14, "learning_rate": 4.995761045965948e-06, "loss": 0.792, "step": 1209 }, { "epoch": 0.14, "learning_rate": 4.995742315600186e-06, "loss": 0.7867, "step": 1210 }, { "epoch": 0.14, "learning_rate": 4.995723543979594e-06, "loss": 0.9457, "step": 1211 }, { "epoch": 0.14, "learning_rate": 4.995704731104483e-06, "loss": 1.3826, "step": 1212 }, { "epoch": 0.14, "learning_rate": 4.995685876975162e-06, "loss": 0.7195, "step": 1213 }, { "epoch": 0.14, "learning_rate": 4.9956669815919454e-06, "loss": 0.7087, "step": 1214 }, { "epoch": 0.14, "learning_rate": 4.995648044955144e-06, "loss": 0.7303, "step": 1215 }, { "epoch": 0.14, "learning_rate": 4.9956290670650705e-06, "loss": 0.7623, "step": 1216 }, { "epoch": 0.14, "learning_rate": 4.99561004792204e-06, "loss": 1.0175, "step": 1217 }, { "epoch": 0.15, "learning_rate": 4.9955909875263654e-06, "loss": 1.1846, "step": 1218 }, { "epoch": 0.15, "learning_rate": 4.995571885878363e-06, "loss": 0.9123, "step": 1219 }, { "epoch": 0.15, "learning_rate": 4.995552742978347e-06, "loss": 0.6274, "step": 1220 }, { "epoch": 0.15, "learning_rate": 4.995533558826635e-06, "loss": 0.9218, "step": 1221 }, { "epoch": 0.15, "learning_rate": 4.995514333423544e-06, "loss": 0.6736, "step": 1222 }, { "epoch": 0.15, "learning_rate": 4.995495066769392e-06, "loss": 0.9038, "step": 1223 }, { "epoch": 0.15, "learning_rate": 4.9954757588644965e-06, "loss": 1.1661, "step": 1224 }, { "epoch": 0.15, "learning_rate": 4.995456409709178e-06, "loss": 1.4006, "step": 1225 }, { "epoch": 0.15, "learning_rate": 4.995437019303755e-06, "loss": 0.4529, "step": 1226 }, { "epoch": 0.15, "learning_rate": 4.995417587648549e-06, "loss": 1.0062, "step": 1227 }, { "epoch": 0.15, "learning_rate": 4.9953981147438815e-06, "loss": 0.7712, "step": 1228 }, { "epoch": 0.15, "learning_rate": 4.995378600590073e-06, "loss": 0.94, "step": 1229 }, { "epoch": 0.15, "learning_rate": 4.995359045187447e-06, "loss": 1.0329, "step": 1230 }, { "epoch": 0.15, "learning_rate": 4.995339448536326e-06, "loss": 1.248, "step": 1231 }, { "epoch": 0.15, "learning_rate": 4.995319810637036e-06, "loss": 0.5397, "step": 1232 }, { "epoch": 0.15, "learning_rate": 4.995300131489899e-06, "loss": 0.819, "step": 1233 }, { "epoch": 0.15, "learning_rate": 4.995280411095242e-06, "loss": 0.7771, "step": 1234 }, { "epoch": 0.15, "learning_rate": 4.99526064945339e-06, "loss": 0.8945, "step": 1235 }, { "epoch": 0.15, "learning_rate": 4.9952408465646705e-06, "loss": 1.0191, "step": 1236 }, { "epoch": 0.15, "learning_rate": 4.99522100242941e-06, "loss": 1.2717, "step": 1237 }, { "epoch": 0.15, "learning_rate": 4.995201117047938e-06, "loss": 0.6588, "step": 1238 }, { "epoch": 0.15, "learning_rate": 4.9951811904205804e-06, "loss": 0.548, "step": 1239 }, { "epoch": 0.15, "learning_rate": 4.99516122254767e-06, "loss": 0.5589, "step": 1240 }, { "epoch": 0.15, "learning_rate": 4.995141213429534e-06, "loss": 1.0034, "step": 1241 }, { "epoch": 0.15, "learning_rate": 4.995121163066505e-06, "loss": 0.9609, "step": 1242 }, { "epoch": 0.15, "learning_rate": 4.995101071458914e-06, "loss": 1.2655, "step": 1243 }, { "epoch": 0.15, "learning_rate": 4.995080938607093e-06, "loss": 0.7803, "step": 1244 }, { "epoch": 0.15, "learning_rate": 4.995060764511374e-06, "loss": 0.7791, "step": 1245 }, { "epoch": 0.15, "learning_rate": 4.995040549172092e-06, "loss": 0.7411, "step": 1246 }, { "epoch": 0.15, "learning_rate": 4.99502029258958e-06, "loss": 0.8316, "step": 1247 }, { "epoch": 0.15, "learning_rate": 4.994999994764174e-06, "loss": 0.9215, "step": 1248 }, { "epoch": 0.15, "learning_rate": 4.994979655696207e-06, "loss": 1.1776, "step": 1249 }, { "epoch": 0.15, "learning_rate": 4.9949592753860185e-06, "loss": 1.1238, "step": 1250 }, { "epoch": 0.15, "learning_rate": 4.9949388538339425e-06, "loss": 0.7301, "step": 1251 }, { "epoch": 0.15, "learning_rate": 4.994918391040319e-06, "loss": 0.6729, "step": 1252 }, { "epoch": 0.15, "learning_rate": 4.994897887005485e-06, "loss": 0.9491, "step": 1253 }, { "epoch": 0.15, "learning_rate": 4.994877341729779e-06, "loss": 0.89, "step": 1254 }, { "epoch": 0.15, "learning_rate": 4.994856755213542e-06, "loss": 1.0406, "step": 1255 }, { "epoch": 0.15, "learning_rate": 4.9948361274571135e-06, "loss": 1.2448, "step": 1256 }, { "epoch": 0.15, "learning_rate": 4.994815458460834e-06, "loss": 0.6524, "step": 1257 }, { "epoch": 0.15, "learning_rate": 4.994794748225046e-06, "loss": 0.8069, "step": 1258 }, { "epoch": 0.15, "learning_rate": 4.994773996750091e-06, "loss": 0.866, "step": 1259 }, { "epoch": 0.15, "learning_rate": 4.994753204036312e-06, "loss": 0.9691, "step": 1260 }, { "epoch": 0.15, "learning_rate": 4.994732370084055e-06, "loss": 1.1285, "step": 1261 }, { "epoch": 0.15, "learning_rate": 4.994711494893661e-06, "loss": 1.2073, "step": 1262 }, { "epoch": 0.15, "learning_rate": 4.994690578465477e-06, "loss": 0.7791, "step": 1263 }, { "epoch": 0.15, "learning_rate": 4.994669620799849e-06, "loss": 0.6158, "step": 1264 }, { "epoch": 0.15, "learning_rate": 4.994648621897123e-06, "loss": 0.7253, "step": 1265 }, { "epoch": 0.15, "learning_rate": 4.994627581757644e-06, "loss": 0.6826, "step": 1266 }, { "epoch": 0.15, "learning_rate": 4.9946065003817635e-06, "loss": 0.857, "step": 1267 }, { "epoch": 0.15, "learning_rate": 4.994585377769829e-06, "loss": 0.9534, "step": 1268 }, { "epoch": 0.15, "learning_rate": 4.994564213922187e-06, "loss": 0.8153, "step": 1269 }, { "epoch": 0.15, "learning_rate": 4.99454300883919e-06, "loss": 0.7538, "step": 1270 }, { "epoch": 0.15, "learning_rate": 4.994521762521188e-06, "loss": 0.7379, "step": 1271 }, { "epoch": 0.15, "learning_rate": 4.994500474968531e-06, "loss": 0.6847, "step": 1272 }, { "epoch": 0.15, "learning_rate": 4.994479146181573e-06, "loss": 0.9579, "step": 1273 }, { "epoch": 0.15, "learning_rate": 4.994457776160665e-06, "loss": 1.191, "step": 1274 }, { "epoch": 0.15, "learning_rate": 4.994436364906161e-06, "loss": 1.2145, "step": 1275 }, { "epoch": 0.15, "learning_rate": 4.994414912418414e-06, "loss": 0.6237, "step": 1276 }, { "epoch": 0.15, "learning_rate": 4.9943934186977794e-06, "loss": 0.8755, "step": 1277 }, { "epoch": 0.15, "learning_rate": 4.994371883744613e-06, "loss": 0.7837, "step": 1278 }, { "epoch": 0.15, "learning_rate": 4.994350307559269e-06, "loss": 1.0143, "step": 1279 }, { "epoch": 0.15, "learning_rate": 4.994328690142105e-06, "loss": 1.1765, "step": 1280 }, { "epoch": 0.15, "learning_rate": 4.994307031493479e-06, "loss": 1.2198, "step": 1281 }, { "epoch": 0.15, "learning_rate": 4.994285331613748e-06, "loss": 0.6179, "step": 1282 }, { "epoch": 0.15, "learning_rate": 4.994263590503272e-06, "loss": 0.9074, "step": 1283 }, { "epoch": 0.15, "learning_rate": 4.99424180816241e-06, "loss": 0.8749, "step": 1284 }, { "epoch": 0.15, "learning_rate": 4.99421998459152e-06, "loss": 0.7822, "step": 1285 }, { "epoch": 0.15, "learning_rate": 4.994198119790966e-06, "loss": 1.1291, "step": 1286 }, { "epoch": 0.15, "learning_rate": 4.994176213761107e-06, "loss": 1.2932, "step": 1287 }, { "epoch": 0.15, "learning_rate": 4.994154266502306e-06, "loss": 0.6964, "step": 1288 }, { "epoch": 0.15, "learning_rate": 4.994132278014925e-06, "loss": 0.6725, "step": 1289 }, { "epoch": 0.15, "learning_rate": 4.994110248299329e-06, "loss": 0.7352, "step": 1290 }, { "epoch": 0.15, "learning_rate": 4.994088177355882e-06, "loss": 0.7727, "step": 1291 }, { "epoch": 0.15, "learning_rate": 4.9940660651849465e-06, "loss": 0.7991, "step": 1292 }, { "epoch": 0.15, "learning_rate": 4.994043911786891e-06, "loss": 1.1263, "step": 1293 }, { "epoch": 0.15, "learning_rate": 4.994021717162079e-06, "loss": 0.7335, "step": 1294 }, { "epoch": 0.15, "learning_rate": 4.99399948131088e-06, "loss": 0.663, "step": 1295 }, { "epoch": 0.15, "learning_rate": 4.99397720423366e-06, "loss": 0.6772, "step": 1296 }, { "epoch": 0.15, "learning_rate": 4.993954885930787e-06, "loss": 0.7849, "step": 1297 }, { "epoch": 0.15, "learning_rate": 4.993932526402631e-06, "loss": 0.8328, "step": 1298 }, { "epoch": 0.15, "learning_rate": 4.993910125649561e-06, "loss": 1.069, "step": 1299 }, { "epoch": 0.15, "learning_rate": 4.993887683671947e-06, "loss": 1.3547, "step": 1300 }, { "epoch": 0.15, "learning_rate": 4.993865200470161e-06, "loss": 0.5386, "step": 1301 }, { "epoch": 0.16, "learning_rate": 4.993842676044573e-06, "loss": 0.7769, "step": 1302 }, { "epoch": 0.16, "learning_rate": 4.9938201103955574e-06, "loss": 0.9903, "step": 1303 }, { "epoch": 0.16, "learning_rate": 4.993797503523486e-06, "loss": 0.6756, "step": 1304 }, { "epoch": 0.16, "learning_rate": 4.993774855428733e-06, "loss": 1.1007, "step": 1305 }, { "epoch": 0.16, "learning_rate": 4.993752166111671e-06, "loss": 1.2388, "step": 1306 }, { "epoch": 0.16, "learning_rate": 4.993729435572678e-06, "loss": 0.4477, "step": 1307 }, { "epoch": 0.16, "learning_rate": 4.993706663812127e-06, "loss": 0.7401, "step": 1308 }, { "epoch": 0.16, "learning_rate": 4.993683850830395e-06, "loss": 0.6879, "step": 1309 }, { "epoch": 0.16, "learning_rate": 4.993660996627861e-06, "loss": 0.782, "step": 1310 }, { "epoch": 0.16, "learning_rate": 4.9936381012049015e-06, "loss": 0.9929, "step": 1311 }, { "epoch": 0.16, "learning_rate": 4.993615164561895e-06, "loss": 1.1624, "step": 1312 }, { "epoch": 0.16, "learning_rate": 4.99359218669922e-06, "loss": 0.51, "step": 1313 }, { "epoch": 0.16, "learning_rate": 4.993569167617257e-06, "loss": 0.6319, "step": 1314 }, { "epoch": 0.16, "learning_rate": 4.9935461073163865e-06, "loss": 0.6623, "step": 1315 }, { "epoch": 0.16, "learning_rate": 4.9935230057969894e-06, "loss": 0.938, "step": 1316 }, { "epoch": 0.16, "learning_rate": 4.993499863059449e-06, "loss": 0.9533, "step": 1317 }, { "epoch": 0.16, "learning_rate": 4.993476679104145e-06, "loss": 0.981, "step": 1318 }, { "epoch": 0.16, "learning_rate": 4.993453453931463e-06, "loss": 0.7969, "step": 1319 }, { "epoch": 0.16, "learning_rate": 4.993430187541786e-06, "loss": 0.7069, "step": 1320 }, { "epoch": 0.16, "learning_rate": 4.993406879935499e-06, "loss": 0.6108, "step": 1321 }, { "epoch": 0.16, "learning_rate": 4.993383531112987e-06, "loss": 0.7646, "step": 1322 }, { "epoch": 0.16, "learning_rate": 4.993360141074636e-06, "loss": 0.9302, "step": 1323 }, { "epoch": 0.16, "learning_rate": 4.993336709820833e-06, "loss": 1.2549, "step": 1324 }, { "epoch": 0.16, "learning_rate": 4.993313237351964e-06, "loss": 1.1942, "step": 1325 }, { "epoch": 0.16, "learning_rate": 4.993289723668419e-06, "loss": 0.7486, "step": 1326 }, { "epoch": 0.16, "learning_rate": 4.993266168770586e-06, "loss": 0.7796, "step": 1327 }, { "epoch": 0.16, "learning_rate": 4.993242572658853e-06, "loss": 0.9099, "step": 1328 }, { "epoch": 0.16, "learning_rate": 4.993218935333611e-06, "loss": 0.9307, "step": 1329 }, { "epoch": 0.16, "learning_rate": 4.993195256795251e-06, "loss": 1.0562, "step": 1330 }, { "epoch": 0.16, "learning_rate": 4.9931715370441654e-06, "loss": 1.0952, "step": 1331 }, { "epoch": 0.16, "learning_rate": 4.993147776080744e-06, "loss": 0.6944, "step": 1332 }, { "epoch": 0.16, "learning_rate": 4.99312397390538e-06, "loss": 0.8353, "step": 1333 }, { "epoch": 0.16, "learning_rate": 4.993100130518469e-06, "loss": 0.8541, "step": 1334 }, { "epoch": 0.16, "learning_rate": 4.993076245920403e-06, "loss": 0.8128, "step": 1335 }, { "epoch": 0.16, "learning_rate": 4.9930523201115775e-06, "loss": 0.89, "step": 1336 }, { "epoch": 0.16, "learning_rate": 4.993028353092388e-06, "loss": 1.1188, "step": 1337 }, { "epoch": 0.16, "learning_rate": 4.993004344863231e-06, "loss": 0.6975, "step": 1338 }, { "epoch": 0.16, "learning_rate": 4.992980295424502e-06, "loss": 0.646, "step": 1339 }, { "epoch": 0.16, "learning_rate": 4.9929562047766e-06, "loss": 0.8557, "step": 1340 }, { "epoch": 0.16, "learning_rate": 4.992932072919923e-06, "loss": 0.7306, "step": 1341 }, { "epoch": 0.16, "learning_rate": 4.99290789985487e-06, "loss": 0.9195, "step": 1342 }, { "epoch": 0.16, "learning_rate": 4.992883685581839e-06, "loss": 1.103, "step": 1343 }, { "epoch": 0.16, "learning_rate": 4.992859430101232e-06, "loss": 0.8997, "step": 1344 }, { "epoch": 0.16, "learning_rate": 4.9928351334134505e-06, "loss": 0.6049, "step": 1345 }, { "epoch": 0.16, "learning_rate": 4.992810795518894e-06, "loss": 0.6922, "step": 1346 }, { "epoch": 0.16, "learning_rate": 4.9927864164179664e-06, "loss": 0.7708, "step": 1347 }, { "epoch": 0.16, "learning_rate": 4.992761996111069e-06, "loss": 0.8291, "step": 1348 }, { "epoch": 0.16, "learning_rate": 4.992737534598609e-06, "loss": 1.0695, "step": 1349 }, { "epoch": 0.16, "learning_rate": 4.992713031880986e-06, "loss": 1.2231, "step": 1350 }, { "epoch": 0.16, "learning_rate": 4.992688487958609e-06, "loss": 0.5374, "step": 1351 }, { "epoch": 0.16, "learning_rate": 4.9926639028318814e-06, "loss": 0.7996, "step": 1352 }, { "epoch": 0.16, "learning_rate": 4.992639276501211e-06, "loss": 0.9156, "step": 1353 }, { "epoch": 0.16, "learning_rate": 4.992614608967003e-06, "loss": 1.0687, "step": 1354 }, { "epoch": 0.16, "learning_rate": 4.9925899002296675e-06, "loss": 1.1367, "step": 1355 }, { "epoch": 0.16, "learning_rate": 4.992565150289611e-06, "loss": 1.1502, "step": 1356 }, { "epoch": 0.16, "learning_rate": 4.992540359147244e-06, "loss": 0.7061, "step": 1357 }, { "epoch": 0.16, "learning_rate": 4.992515526802976e-06, "loss": 0.6797, "step": 1358 }, { "epoch": 0.16, "learning_rate": 4.992490653257216e-06, "loss": 0.607, "step": 1359 }, { "epoch": 0.16, "learning_rate": 4.992465738510377e-06, "loss": 0.7901, "step": 1360 }, { "epoch": 0.16, "learning_rate": 4.992440782562871e-06, "loss": 1.0031, "step": 1361 }, { "epoch": 0.16, "learning_rate": 4.992415785415109e-06, "loss": 1.1336, "step": 1362 }, { "epoch": 0.16, "learning_rate": 4.9923907470675046e-06, "loss": 0.6336, "step": 1363 }, { "epoch": 0.16, "learning_rate": 4.992365667520472e-06, "loss": 0.5844, "step": 1364 }, { "epoch": 0.16, "learning_rate": 4.992340546774427e-06, "loss": 0.719, "step": 1365 }, { "epoch": 0.16, "learning_rate": 4.992315384829783e-06, "loss": 0.8087, "step": 1366 }, { "epoch": 0.16, "learning_rate": 4.992290181686956e-06, "loss": 1.0443, "step": 1367 }, { "epoch": 0.16, "learning_rate": 4.9922649373463634e-06, "loss": 1.2816, "step": 1368 }, { "epoch": 0.16, "learning_rate": 4.992239651808423e-06, "loss": 0.9691, "step": 1369 }, { "epoch": 0.16, "learning_rate": 4.992214325073551e-06, "loss": 0.5459, "step": 1370 }, { "epoch": 0.16, "learning_rate": 4.992188957142168e-06, "loss": 0.7373, "step": 1371 }, { "epoch": 0.16, "learning_rate": 4.9921635480146915e-06, "loss": 0.8127, "step": 1372 }, { "epoch": 0.16, "learning_rate": 4.992138097691543e-06, "loss": 0.9886, "step": 1373 }, { "epoch": 0.16, "learning_rate": 4.992112606173142e-06, "loss": 1.1368, "step": 1374 }, { "epoch": 0.16, "learning_rate": 4.992087073459911e-06, "loss": 1.2603, "step": 1375 }, { "epoch": 0.16, "learning_rate": 4.992061499552272e-06, "loss": 0.5622, "step": 1376 }, { "epoch": 0.16, "learning_rate": 4.992035884450647e-06, "loss": 0.8884, "step": 1377 }, { "epoch": 0.16, "learning_rate": 4.992010228155459e-06, "loss": 0.7605, "step": 1378 }, { "epoch": 0.16, "learning_rate": 4.991984530667134e-06, "loss": 0.8807, "step": 1379 }, { "epoch": 0.16, "learning_rate": 4.991958791986095e-06, "loss": 1.0217, "step": 1380 }, { "epoch": 0.16, "learning_rate": 4.991933012112768e-06, "loss": 1.3441, "step": 1381 }, { "epoch": 0.16, "learning_rate": 4.99190719104758e-06, "loss": 0.6416, "step": 1382 }, { "epoch": 0.16, "learning_rate": 4.9918813287909564e-06, "loss": 0.7195, "step": 1383 }, { "epoch": 0.16, "learning_rate": 4.991855425343325e-06, "loss": 0.8107, "step": 1384 }, { "epoch": 0.16, "learning_rate": 4.991829480705115e-06, "loss": 0.9647, "step": 1385 }, { "epoch": 0.17, "learning_rate": 4.991803494876755e-06, "loss": 1.0362, "step": 1386 }, { "epoch": 0.17, "learning_rate": 4.991777467858673e-06, "loss": 1.2006, "step": 1387 }, { "epoch": 0.17, "learning_rate": 4.991751399651301e-06, "loss": 0.7724, "step": 1388 }, { "epoch": 0.17, "learning_rate": 4.991725290255069e-06, "loss": 0.6151, "step": 1389 }, { "epoch": 0.17, "learning_rate": 4.991699139670409e-06, "loss": 0.7767, "step": 1390 }, { "epoch": 0.17, "learning_rate": 4.9916729478977535e-06, "loss": 0.7419, "step": 1391 }, { "epoch": 0.17, "learning_rate": 4.991646714937535e-06, "loss": 0.8771, "step": 1392 }, { "epoch": 0.17, "learning_rate": 4.991620440790187e-06, "loss": 1.0452, "step": 1393 }, { "epoch": 0.17, "learning_rate": 4.991594125456144e-06, "loss": 0.8411, "step": 1394 }, { "epoch": 0.17, "learning_rate": 4.991567768935841e-06, "loss": 0.6578, "step": 1395 }, { "epoch": 0.17, "learning_rate": 4.9915413712297135e-06, "loss": 0.6555, "step": 1396 }, { "epoch": 0.17, "learning_rate": 4.991514932338198e-06, "loss": 0.6777, "step": 1397 }, { "epoch": 0.17, "learning_rate": 4.991488452261733e-06, "loss": 0.8616, "step": 1398 }, { "epoch": 0.17, "learning_rate": 4.991461931000754e-06, "loss": 1.0671, "step": 1399 }, { "epoch": 0.17, "learning_rate": 4.991435368555699e-06, "loss": 1.3143, "step": 1400 }, { "epoch": 0.17, "learning_rate": 4.99140876492701e-06, "loss": 0.7174, "step": 1401 }, { "epoch": 0.17, "learning_rate": 4.991382120115125e-06, "loss": 0.604, "step": 1402 }, { "epoch": 0.17, "learning_rate": 4.991355434120484e-06, "loss": 0.8427, "step": 1403 }, { "epoch": 0.17, "learning_rate": 4.9913287069435275e-06, "loss": 0.9625, "step": 1404 }, { "epoch": 0.17, "learning_rate": 4.991301938584701e-06, "loss": 1.0538, "step": 1405 }, { "epoch": 0.17, "learning_rate": 4.991275129044442e-06, "loss": 1.2371, "step": 1406 }, { "epoch": 0.17, "learning_rate": 4.991248278323198e-06, "loss": 0.5698, "step": 1407 }, { "epoch": 0.17, "learning_rate": 4.991221386421409e-06, "loss": 0.7545, "step": 1408 }, { "epoch": 0.17, "learning_rate": 4.991194453339523e-06, "loss": 0.8531, "step": 1409 }, { "epoch": 0.17, "learning_rate": 4.991167479077983e-06, "loss": 0.7087, "step": 1410 }, { "epoch": 0.17, "learning_rate": 4.991140463637235e-06, "loss": 0.9258, "step": 1411 }, { "epoch": 0.17, "learning_rate": 4.991113407017727e-06, "loss": 1.3041, "step": 1412 }, { "epoch": 0.17, "learning_rate": 4.991086309219906e-06, "loss": 0.7869, "step": 1413 }, { "epoch": 0.17, "learning_rate": 4.991059170244218e-06, "loss": 0.6544, "step": 1414 }, { "epoch": 0.17, "learning_rate": 4.991031990091113e-06, "loss": 0.8492, "step": 1415 }, { "epoch": 0.17, "learning_rate": 4.991004768761041e-06, "loss": 0.8942, "step": 1416 }, { "epoch": 0.17, "learning_rate": 4.99097750625445e-06, "loss": 0.9943, "step": 1417 }, { "epoch": 0.17, "learning_rate": 4.990950202571793e-06, "loss": 1.0748, "step": 1418 }, { "epoch": 0.17, "learning_rate": 4.9909228577135195e-06, "loss": 0.6904, "step": 1419 }, { "epoch": 0.17, "learning_rate": 4.990895471680082e-06, "loss": 0.6387, "step": 1420 }, { "epoch": 0.17, "learning_rate": 4.990868044471933e-06, "loss": 0.7111, "step": 1421 }, { "epoch": 0.17, "learning_rate": 4.9908405760895275e-06, "loss": 0.826, "step": 1422 }, { "epoch": 0.17, "learning_rate": 4.990813066533318e-06, "loss": 0.9386, "step": 1423 }, { "epoch": 0.17, "learning_rate": 4.990785515803759e-06, "loss": 1.1308, "step": 1424 }, { "epoch": 0.17, "learning_rate": 4.990757923901306e-06, "loss": 1.147, "step": 1425 }, { "epoch": 0.17, "learning_rate": 4.990730290826417e-06, "loss": 0.5054, "step": 1426 }, { "epoch": 0.17, "learning_rate": 4.9907026165795455e-06, "loss": 0.8217, "step": 1427 }, { "epoch": 0.17, "learning_rate": 4.990674901161152e-06, "loss": 0.7706, "step": 1428 }, { "epoch": 0.17, "learning_rate": 4.990647144571693e-06, "loss": 0.7675, "step": 1429 }, { "epoch": 0.17, "learning_rate": 4.990619346811628e-06, "loss": 1.0863, "step": 1430 }, { "epoch": 0.17, "learning_rate": 4.990591507881416e-06, "loss": 1.1013, "step": 1431 }, { "epoch": 0.17, "learning_rate": 4.990563627781517e-06, "loss": 0.7048, "step": 1432 }, { "epoch": 0.17, "learning_rate": 4.990535706512393e-06, "loss": 0.6103, "step": 1433 }, { "epoch": 0.17, "learning_rate": 4.990507744074504e-06, "loss": 0.8112, "step": 1434 }, { "epoch": 0.17, "learning_rate": 4.990479740468314e-06, "loss": 0.7964, "step": 1435 }, { "epoch": 0.17, "learning_rate": 4.990451695694284e-06, "loss": 0.9946, "step": 1436 }, { "epoch": 0.17, "learning_rate": 4.990423609752879e-06, "loss": 1.2088, "step": 1437 }, { "epoch": 0.17, "learning_rate": 4.9903954826445625e-06, "loss": 0.7127, "step": 1438 }, { "epoch": 0.17, "learning_rate": 4.9903673143698e-06, "loss": 0.7899, "step": 1439 }, { "epoch": 0.17, "learning_rate": 4.990339104929056e-06, "loss": 0.784, "step": 1440 }, { "epoch": 0.17, "learning_rate": 4.990310854322798e-06, "loss": 0.8028, "step": 1441 }, { "epoch": 0.17, "learning_rate": 4.990282562551494e-06, "loss": 0.9258, "step": 1442 }, { "epoch": 0.17, "learning_rate": 4.9902542296156085e-06, "loss": 1.0984, "step": 1443 }, { "epoch": 0.17, "learning_rate": 4.990225855515613e-06, "loss": 0.8879, "step": 1444 }, { "epoch": 0.17, "learning_rate": 4.990197440251975e-06, "loss": 0.6047, "step": 1445 }, { "epoch": 0.17, "learning_rate": 4.990168983825163e-06, "loss": 0.7153, "step": 1446 }, { "epoch": 0.17, "learning_rate": 4.99014048623565e-06, "loss": 0.8404, "step": 1447 }, { "epoch": 0.17, "learning_rate": 4.990111947483906e-06, "loss": 0.9123, "step": 1448 }, { "epoch": 0.17, "learning_rate": 4.9900833675704015e-06, "loss": 1.0748, "step": 1449 }, { "epoch": 0.17, "learning_rate": 4.990054746495611e-06, "loss": 1.2278, "step": 1450 }, { "epoch": 0.17, "learning_rate": 4.990026084260006e-06, "loss": 0.6532, "step": 1451 }, { "epoch": 0.17, "learning_rate": 4.989997380864061e-06, "loss": 0.6609, "step": 1452 }, { "epoch": 0.17, "learning_rate": 4.98996863630825e-06, "loss": 0.7666, "step": 1453 }, { "epoch": 0.17, "learning_rate": 4.989939850593049e-06, "loss": 0.874, "step": 1454 }, { "epoch": 0.17, "learning_rate": 4.989911023718934e-06, "loss": 0.8311, "step": 1455 }, { "epoch": 0.17, "learning_rate": 4.98988215568638e-06, "loss": 1.1784, "step": 1456 }, { "epoch": 0.17, "learning_rate": 4.989853246495865e-06, "loss": 0.6139, "step": 1457 }, { "epoch": 0.17, "learning_rate": 4.9898242961478675e-06, "loss": 0.6084, "step": 1458 }, { "epoch": 0.17, "learning_rate": 4.989795304642865e-06, "loss": 0.794, "step": 1459 }, { "epoch": 0.17, "learning_rate": 4.989766271981337e-06, "loss": 0.9771, "step": 1460 }, { "epoch": 0.17, "learning_rate": 4.989737198163764e-06, "loss": 1.0758, "step": 1461 }, { "epoch": 0.17, "learning_rate": 4.989708083190626e-06, "loss": 1.1797, "step": 1462 }, { "epoch": 0.17, "learning_rate": 4.9896789270624045e-06, "loss": 0.7897, "step": 1463 }, { "epoch": 0.17, "learning_rate": 4.989649729779581e-06, "loss": 0.6568, "step": 1464 }, { "epoch": 0.17, "learning_rate": 4.98962049134264e-06, "loss": 0.6401, "step": 1465 }, { "epoch": 0.17, "learning_rate": 4.989591211752062e-06, "loss": 0.7467, "step": 1466 }, { "epoch": 0.17, "learning_rate": 4.9895618910083335e-06, "loss": 1.0157, "step": 1467 }, { "epoch": 0.17, "learning_rate": 4.989532529111937e-06, "loss": 1.2253, "step": 1468 }, { "epoch": 0.17, "learning_rate": 4.989503126063359e-06, "loss": 0.9949, "step": 1469 }, { "epoch": 0.18, "learning_rate": 4.989473681863086e-06, "loss": 0.8016, "step": 1470 }, { "epoch": 0.18, "learning_rate": 4.9894441965116035e-06, "loss": 0.6939, "step": 1471 }, { "epoch": 0.18, "learning_rate": 4.9894146700094005e-06, "loss": 0.7296, "step": 1472 }, { "epoch": 0.18, "learning_rate": 4.989385102356963e-06, "loss": 0.9413, "step": 1473 }, { "epoch": 0.18, "learning_rate": 4.9893554935547814e-06, "loss": 1.2047, "step": 1474 }, { "epoch": 0.18, "learning_rate": 4.989325843603345e-06, "loss": 1.2472, "step": 1475 }, { "epoch": 0.18, "learning_rate": 4.989296152503143e-06, "loss": 0.6793, "step": 1476 }, { "epoch": 0.18, "learning_rate": 4.989266420254667e-06, "loss": 0.7228, "step": 1477 }, { "epoch": 0.18, "learning_rate": 4.989236646858408e-06, "loss": 0.7632, "step": 1478 }, { "epoch": 0.18, "learning_rate": 4.989206832314859e-06, "loss": 1.0048, "step": 1479 }, { "epoch": 0.18, "learning_rate": 4.989176976624511e-06, "loss": 1.1353, "step": 1480 }, { "epoch": 0.18, "learning_rate": 4.98914707978786e-06, "loss": 1.3679, "step": 1481 }, { "epoch": 0.18, "learning_rate": 4.989117141805398e-06, "loss": 0.542, "step": 1482 }, { "epoch": 0.18, "learning_rate": 4.989087162677621e-06, "loss": 0.7082, "step": 1483 }, { "epoch": 0.18, "learning_rate": 4.989057142405025e-06, "loss": 0.7993, "step": 1484 }, { "epoch": 0.18, "learning_rate": 4.9890270809881046e-06, "loss": 1.0287, "step": 1485 }, { "epoch": 0.18, "learning_rate": 4.9889969784273585e-06, "loss": 1.0513, "step": 1486 }, { "epoch": 0.18, "learning_rate": 4.988966834723284e-06, "loss": 1.2357, "step": 1487 }, { "epoch": 0.18, "learning_rate": 4.988936649876377e-06, "loss": 0.707, "step": 1488 }, { "epoch": 0.18, "learning_rate": 4.98890642388714e-06, "loss": 0.516, "step": 1489 }, { "epoch": 0.18, "learning_rate": 4.98887615675607e-06, "loss": 0.6823, "step": 1490 }, { "epoch": 0.18, "learning_rate": 4.988845848483669e-06, "loss": 0.6639, "step": 1491 }, { "epoch": 0.18, "learning_rate": 4.988815499070436e-06, "loss": 0.9032, "step": 1492 }, { "epoch": 0.18, "learning_rate": 4.988785108516876e-06, "loss": 1.1788, "step": 1493 }, { "epoch": 0.18, "learning_rate": 4.988754676823487e-06, "loss": 0.8521, "step": 1494 }, { "epoch": 0.18, "learning_rate": 4.988724203990776e-06, "loss": 0.6561, "step": 1495 }, { "epoch": 0.18, "learning_rate": 4.988693690019245e-06, "loss": 0.6513, "step": 1496 }, { "epoch": 0.18, "learning_rate": 4.988663134909398e-06, "loss": 0.7625, "step": 1497 }, { "epoch": 0.18, "learning_rate": 4.98863253866174e-06, "loss": 0.8759, "step": 1498 }, { "epoch": 0.18, "learning_rate": 4.98860190127678e-06, "loss": 1.1012, "step": 1499 }, { "epoch": 0.18, "learning_rate": 4.988571222755019e-06, "loss": 1.3065, "step": 1500 }, { "epoch": 0.18, "learning_rate": 4.988540503096968e-06, "loss": 0.5296, "step": 1501 }, { "epoch": 0.18, "learning_rate": 4.988509742303133e-06, "loss": 0.6427, "step": 1502 }, { "epoch": 0.18, "learning_rate": 4.9884789403740235e-06, "loss": 0.7374, "step": 1503 }, { "epoch": 0.18, "learning_rate": 4.9884480973101486e-06, "loss": 0.9116, "step": 1504 }, { "epoch": 0.18, "learning_rate": 4.988417213112018e-06, "loss": 1.1785, "step": 1505 }, { "epoch": 0.18, "learning_rate": 4.988386287780141e-06, "loss": 1.1666, "step": 1506 }, { "epoch": 0.18, "learning_rate": 4.988355321315031e-06, "loss": 0.7606, "step": 1507 }, { "epoch": 0.18, "learning_rate": 4.988324313717199e-06, "loss": 0.7726, "step": 1508 }, { "epoch": 0.18, "learning_rate": 4.988293264987155e-06, "loss": 0.7798, "step": 1509 }, { "epoch": 0.18, "learning_rate": 4.988262175125417e-06, "loss": 0.8765, "step": 1510 }, { "epoch": 0.18, "learning_rate": 4.988231044132496e-06, "loss": 0.9547, "step": 1511 }, { "epoch": 0.18, "learning_rate": 4.988199872008906e-06, "loss": 1.135, "step": 1512 }, { "epoch": 0.18, "learning_rate": 4.988168658755164e-06, "loss": 0.7655, "step": 1513 }, { "epoch": 0.18, "learning_rate": 4.988137404371786e-06, "loss": 0.6831, "step": 1514 }, { "epoch": 0.18, "learning_rate": 4.988106108859287e-06, "loss": 0.8075, "step": 1515 }, { "epoch": 0.18, "learning_rate": 4.988074772218186e-06, "loss": 0.8932, "step": 1516 }, { "epoch": 0.18, "learning_rate": 4.9880433944489995e-06, "loss": 1.0068, "step": 1517 }, { "epoch": 0.18, "learning_rate": 4.988011975552247e-06, "loss": 1.275, "step": 1518 }, { "epoch": 0.18, "learning_rate": 4.987980515528448e-06, "loss": 0.7118, "step": 1519 }, { "epoch": 0.18, "learning_rate": 4.987949014378122e-06, "loss": 0.6107, "step": 1520 }, { "epoch": 0.18, "learning_rate": 4.9879174721017905e-06, "loss": 0.6756, "step": 1521 }, { "epoch": 0.18, "learning_rate": 4.987885888699974e-06, "loss": 0.7234, "step": 1522 }, { "epoch": 0.18, "learning_rate": 4.987854264173196e-06, "loss": 0.9714, "step": 1523 }, { "epoch": 0.18, "learning_rate": 4.987822598521977e-06, "loss": 1.0836, "step": 1524 }, { "epoch": 0.18, "learning_rate": 4.987790891746842e-06, "loss": 1.2979, "step": 1525 }, { "epoch": 0.18, "learning_rate": 4.987759143848315e-06, "loss": 0.5941, "step": 1526 }, { "epoch": 0.18, "learning_rate": 4.9877273548269215e-06, "loss": 0.7059, "step": 1527 }, { "epoch": 0.18, "learning_rate": 4.9876955246831846e-06, "loss": 0.8223, "step": 1528 }, { "epoch": 0.18, "learning_rate": 4.987663653417633e-06, "loss": 0.9351, "step": 1529 }, { "epoch": 0.18, "learning_rate": 4.9876317410307915e-06, "loss": 1.1298, "step": 1530 }, { "epoch": 0.18, "learning_rate": 4.98759978752319e-06, "loss": 1.1823, "step": 1531 }, { "epoch": 0.18, "learning_rate": 4.987567792895354e-06, "loss": 0.5988, "step": 1532 }, { "epoch": 0.18, "learning_rate": 4.987535757147814e-06, "loss": 0.5984, "step": 1533 }, { "epoch": 0.18, "learning_rate": 4.9875036802811e-06, "loss": 0.7908, "step": 1534 }, { "epoch": 0.18, "learning_rate": 4.987471562295741e-06, "loss": 0.7884, "step": 1535 }, { "epoch": 0.18, "learning_rate": 4.987439403192267e-06, "loss": 1.1124, "step": 1536 }, { "epoch": 0.18, "learning_rate": 4.987407202971213e-06, "loss": 1.1041, "step": 1537 }, { "epoch": 0.18, "learning_rate": 4.987374961633107e-06, "loss": 0.8779, "step": 1538 }, { "epoch": 0.18, "learning_rate": 4.987342679178486e-06, "loss": 0.6853, "step": 1539 }, { "epoch": 0.18, "learning_rate": 4.987310355607882e-06, "loss": 0.7932, "step": 1540 }, { "epoch": 0.18, "learning_rate": 4.987277990921828e-06, "loss": 0.8654, "step": 1541 }, { "epoch": 0.18, "learning_rate": 4.98724558512086e-06, "loss": 0.973, "step": 1542 }, { "epoch": 0.18, "learning_rate": 4.987213138205514e-06, "loss": 1.2211, "step": 1543 }, { "epoch": 0.18, "learning_rate": 4.987180650176327e-06, "loss": 0.777, "step": 1544 }, { "epoch": 0.18, "learning_rate": 4.987148121033834e-06, "loss": 0.7213, "step": 1545 }, { "epoch": 0.18, "learning_rate": 4.987115550778575e-06, "loss": 0.8888, "step": 1546 }, { "epoch": 0.18, "learning_rate": 4.987082939411086e-06, "loss": 0.8337, "step": 1547 }, { "epoch": 0.18, "learning_rate": 4.9870502869319086e-06, "loss": 0.8388, "step": 1548 }, { "epoch": 0.18, "learning_rate": 4.987017593341581e-06, "loss": 1.1209, "step": 1549 }, { "epoch": 0.18, "learning_rate": 4.986984858640643e-06, "loss": 1.2564, "step": 1550 }, { "epoch": 0.18, "learning_rate": 4.986952082829638e-06, "loss": 0.7535, "step": 1551 }, { "epoch": 0.18, "learning_rate": 4.986919265909106e-06, "loss": 0.8504, "step": 1552 }, { "epoch": 0.18, "learning_rate": 4.986886407879589e-06, "loss": 0.8275, "step": 1553 }, { "epoch": 0.19, "learning_rate": 4.9868535087416315e-06, "loss": 0.9612, "step": 1554 }, { "epoch": 0.19, "learning_rate": 4.986820568495776e-06, "loss": 1.1761, "step": 1555 }, { "epoch": 0.19, "learning_rate": 4.9867875871425694e-06, "loss": 1.1796, "step": 1556 }, { "epoch": 0.19, "learning_rate": 4.986754564682555e-06, "loss": 0.8477, "step": 1557 }, { "epoch": 0.19, "learning_rate": 4.986721501116278e-06, "loss": 0.6454, "step": 1558 }, { "epoch": 0.19, "learning_rate": 4.986688396444287e-06, "loss": 0.6772, "step": 1559 }, { "epoch": 0.19, "learning_rate": 4.9866552506671275e-06, "loss": 0.812, "step": 1560 }, { "epoch": 0.19, "learning_rate": 4.986622063785348e-06, "loss": 0.9604, "step": 1561 }, { "epoch": 0.19, "learning_rate": 4.9865888357994975e-06, "loss": 1.2356, "step": 1562 }, { "epoch": 0.19, "learning_rate": 4.986555566710125e-06, "loss": 0.6158, "step": 1563 }, { "epoch": 0.19, "learning_rate": 4.98652225651778e-06, "loss": 0.7076, "step": 1564 }, { "epoch": 0.19, "learning_rate": 4.986488905223014e-06, "loss": 0.6085, "step": 1565 }, { "epoch": 0.19, "learning_rate": 4.986455512826377e-06, "loss": 0.838, "step": 1566 }, { "epoch": 0.19, "learning_rate": 4.986422079328422e-06, "loss": 0.985, "step": 1567 }, { "epoch": 0.19, "learning_rate": 4.986388604729702e-06, "loss": 1.1489, "step": 1568 }, { "epoch": 0.19, "learning_rate": 4.986355089030769e-06, "loss": 0.8614, "step": 1569 }, { "epoch": 0.19, "learning_rate": 4.9863215322321774e-06, "loss": 0.5672, "step": 1570 }, { "epoch": 0.19, "learning_rate": 4.986287934334484e-06, "loss": 0.8925, "step": 1571 }, { "epoch": 0.19, "learning_rate": 4.9862542953382405e-06, "loss": 0.7885, "step": 1572 }, { "epoch": 0.19, "learning_rate": 4.986220615244005e-06, "loss": 0.9438, "step": 1573 }, { "epoch": 0.19, "learning_rate": 4.9861868940523355e-06, "loss": 1.1071, "step": 1574 }, { "epoch": 0.19, "learning_rate": 4.9861531317637875e-06, "loss": 1.1537, "step": 1575 }, { "epoch": 0.19, "learning_rate": 4.986119328378919e-06, "loss": 0.6012, "step": 1576 }, { "epoch": 0.19, "learning_rate": 4.98608548389829e-06, "loss": 0.6787, "step": 1577 }, { "epoch": 0.19, "learning_rate": 4.986051598322459e-06, "loss": 0.7843, "step": 1578 }, { "epoch": 0.19, "learning_rate": 4.986017671651987e-06, "loss": 0.765, "step": 1579 }, { "epoch": 0.19, "learning_rate": 4.985983703887434e-06, "loss": 0.9816, "step": 1580 }, { "epoch": 0.19, "learning_rate": 4.985949695029362e-06, "loss": 1.2195, "step": 1581 }, { "epoch": 0.19, "learning_rate": 4.985915645078333e-06, "loss": 0.633, "step": 1582 }, { "epoch": 0.19, "learning_rate": 4.985881554034909e-06, "loss": 0.6713, "step": 1583 }, { "epoch": 0.19, "learning_rate": 4.985847421899655e-06, "loss": 0.6766, "step": 1584 }, { "epoch": 0.19, "learning_rate": 4.985813248673134e-06, "loss": 0.829, "step": 1585 }, { "epoch": 0.19, "learning_rate": 4.985779034355912e-06, "loss": 1.0441, "step": 1586 }, { "epoch": 0.19, "learning_rate": 4.985744778948553e-06, "loss": 1.251, "step": 1587 }, { "epoch": 0.19, "learning_rate": 4.985710482451625e-06, "loss": 0.5994, "step": 1588 }, { "epoch": 0.19, "learning_rate": 4.985676144865695e-06, "loss": 0.5438, "step": 1589 }, { "epoch": 0.19, "learning_rate": 4.985641766191329e-06, "loss": 0.6674, "step": 1590 }, { "epoch": 0.19, "learning_rate": 4.985607346429095e-06, "loss": 0.9827, "step": 1591 }, { "epoch": 0.19, "learning_rate": 4.9855728855795645e-06, "loss": 1.0189, "step": 1592 }, { "epoch": 0.19, "learning_rate": 4.985538383643305e-06, "loss": 1.0906, "step": 1593 }, { "epoch": 0.19, "learning_rate": 4.985503840620888e-06, "loss": 0.6685, "step": 1594 }, { "epoch": 0.19, "learning_rate": 4.985469256512884e-06, "loss": 0.6148, "step": 1595 }, { "epoch": 0.19, "learning_rate": 4.985434631319864e-06, "loss": 0.6462, "step": 1596 }, { "epoch": 0.19, "learning_rate": 4.985399965042401e-06, "loss": 0.7664, "step": 1597 }, { "epoch": 0.19, "learning_rate": 4.985365257681068e-06, "loss": 0.8448, "step": 1598 }, { "epoch": 0.19, "learning_rate": 4.985330509236439e-06, "loss": 1.1302, "step": 1599 }, { "epoch": 0.19, "learning_rate": 4.9852957197090875e-06, "loss": 1.309, "step": 1600 }, { "epoch": 0.19, "learning_rate": 4.98526088909959e-06, "loss": 0.6452, "step": 1601 }, { "epoch": 0.19, "learning_rate": 4.985226017408522e-06, "loss": 0.6845, "step": 1602 }, { "epoch": 0.19, "learning_rate": 4.985191104636459e-06, "loss": 0.8794, "step": 1603 }, { "epoch": 0.19, "learning_rate": 4.985156150783978e-06, "loss": 0.962, "step": 1604 }, { "epoch": 0.19, "learning_rate": 4.9851211558516574e-06, "loss": 1.0886, "step": 1605 }, { "epoch": 0.19, "learning_rate": 4.9850861198400755e-06, "loss": 1.2073, "step": 1606 }, { "epoch": 0.19, "learning_rate": 4.985051042749812e-06, "loss": 0.5466, "step": 1607 }, { "epoch": 0.19, "learning_rate": 4.985015924581445e-06, "loss": 0.732, "step": 1608 }, { "epoch": 0.19, "learning_rate": 4.984980765335558e-06, "loss": 0.7329, "step": 1609 }, { "epoch": 0.19, "learning_rate": 4.9849455650127296e-06, "loss": 1.0174, "step": 1610 }, { "epoch": 0.19, "learning_rate": 4.984910323613543e-06, "loss": 1.0331, "step": 1611 }, { "epoch": 0.19, "learning_rate": 4.9848750411385796e-06, "loss": 1.1432, "step": 1612 }, { "epoch": 0.19, "learning_rate": 4.984839717588424e-06, "loss": 0.6209, "step": 1613 }, { "epoch": 0.19, "learning_rate": 4.984804352963659e-06, "loss": 0.5156, "step": 1614 }, { "epoch": 0.19, "learning_rate": 4.984768947264868e-06, "loss": 0.7184, "step": 1615 }, { "epoch": 0.19, "learning_rate": 4.98473350049264e-06, "loss": 0.9073, "step": 1616 }, { "epoch": 0.19, "learning_rate": 4.984698012647558e-06, "loss": 0.9877, "step": 1617 }, { "epoch": 0.19, "learning_rate": 4.98466248373021e-06, "loss": 1.1735, "step": 1618 }, { "epoch": 0.19, "learning_rate": 4.984626913741182e-06, "loss": 0.8167, "step": 1619 }, { "epoch": 0.19, "learning_rate": 4.9845913026810624e-06, "loss": 0.5926, "step": 1620 }, { "epoch": 0.19, "learning_rate": 4.98455565055044e-06, "loss": 0.6674, "step": 1621 }, { "epoch": 0.19, "learning_rate": 4.984519957349905e-06, "loss": 0.8162, "step": 1622 }, { "epoch": 0.19, "learning_rate": 4.984484223080046e-06, "loss": 0.8461, "step": 1623 }, { "epoch": 0.19, "learning_rate": 4.984448447741455e-06, "loss": 1.0282, "step": 1624 }, { "epoch": 0.19, "learning_rate": 4.984412631334722e-06, "loss": 1.085, "step": 1625 }, { "epoch": 0.19, "learning_rate": 4.98437677386044e-06, "loss": 0.5329, "step": 1626 }, { "epoch": 0.19, "learning_rate": 4.984340875319201e-06, "loss": 0.8065, "step": 1627 }, { "epoch": 0.19, "learning_rate": 4.9843049357116e-06, "loss": 0.8572, "step": 1628 }, { "epoch": 0.19, "learning_rate": 4.984268955038229e-06, "loss": 0.9555, "step": 1629 }, { "epoch": 0.19, "learning_rate": 4.984232933299684e-06, "loss": 1.0678, "step": 1630 }, { "epoch": 0.19, "learning_rate": 4.984196870496561e-06, "loss": 1.2722, "step": 1631 }, { "epoch": 0.19, "learning_rate": 4.984160766629454e-06, "loss": 0.8097, "step": 1632 }, { "epoch": 0.19, "learning_rate": 4.984124621698961e-06, "loss": 0.7115, "step": 1633 }, { "epoch": 0.19, "learning_rate": 4.984088435705681e-06, "loss": 0.8911, "step": 1634 }, { "epoch": 0.19, "learning_rate": 4.984052208650209e-06, "loss": 0.8757, "step": 1635 }, { "epoch": 0.19, "learning_rate": 4.984015940533147e-06, "loss": 0.8526, "step": 1636 }, { "epoch": 0.19, "learning_rate": 4.983979631355093e-06, "loss": 1.1285, "step": 1637 }, { "epoch": 0.2, "learning_rate": 4.983943281116646e-06, "loss": 0.684, "step": 1638 }, { "epoch": 0.2, "learning_rate": 4.983906889818409e-06, "loss": 0.6533, "step": 1639 }, { "epoch": 0.2, "learning_rate": 4.983870457460982e-06, "loss": 0.7376, "step": 1640 }, { "epoch": 0.2, "learning_rate": 4.983833984044969e-06, "loss": 0.7678, "step": 1641 }, { "epoch": 0.2, "learning_rate": 4.983797469570971e-06, "loss": 0.8503, "step": 1642 }, { "epoch": 0.2, "learning_rate": 4.983760914039594e-06, "loss": 1.0806, "step": 1643 }, { "epoch": 0.2, "learning_rate": 4.983724317451439e-06, "loss": 0.775, "step": 1644 }, { "epoch": 0.2, "learning_rate": 4.983687679807113e-06, "loss": 0.6717, "step": 1645 }, { "epoch": 0.2, "learning_rate": 4.9836510011072215e-06, "loss": 0.7126, "step": 1646 }, { "epoch": 0.2, "learning_rate": 4.98361428135237e-06, "loss": 0.9661, "step": 1647 }, { "epoch": 0.2, "learning_rate": 4.983577520543167e-06, "loss": 0.9023, "step": 1648 }, { "epoch": 0.2, "learning_rate": 4.983540718680219e-06, "loss": 1.2021, "step": 1649 }, { "epoch": 0.2, "learning_rate": 4.983503875764133e-06, "loss": 1.2332, "step": 1650 }, { "epoch": 0.2, "learning_rate": 4.9834669917955215e-06, "loss": 0.6279, "step": 1651 }, { "epoch": 0.2, "learning_rate": 4.983430066774991e-06, "loss": 0.761, "step": 1652 }, { "epoch": 0.2, "learning_rate": 4.983393100703154e-06, "loss": 0.7363, "step": 1653 }, { "epoch": 0.2, "learning_rate": 4.98335609358062e-06, "loss": 0.9186, "step": 1654 }, { "epoch": 0.2, "learning_rate": 4.983319045408002e-06, "loss": 1.0411, "step": 1655 }, { "epoch": 0.2, "learning_rate": 4.9832819561859114e-06, "loss": 1.179, "step": 1656 }, { "epoch": 0.2, "learning_rate": 4.983244825914962e-06, "loss": 0.4872, "step": 1657 }, { "epoch": 0.2, "learning_rate": 4.9832076545957666e-06, "loss": 0.5679, "step": 1658 }, { "epoch": 0.2, "learning_rate": 4.9831704422289404e-06, "loss": 0.8223, "step": 1659 }, { "epoch": 0.2, "learning_rate": 4.983133188815099e-06, "loss": 0.856, "step": 1660 }, { "epoch": 0.2, "learning_rate": 4.983095894354858e-06, "loss": 0.9482, "step": 1661 }, { "epoch": 0.2, "learning_rate": 4.9830585588488325e-06, "loss": 1.1441, "step": 1662 }, { "epoch": 0.2, "learning_rate": 4.983021182297642e-06, "loss": 0.7313, "step": 1663 }, { "epoch": 0.2, "learning_rate": 4.982983764701902e-06, "loss": 0.5116, "step": 1664 }, { "epoch": 0.2, "learning_rate": 4.982946306062232e-06, "loss": 0.838, "step": 1665 }, { "epoch": 0.2, "learning_rate": 4.9829088063792515e-06, "loss": 0.8015, "step": 1666 }, { "epoch": 0.2, "learning_rate": 4.9828712656535795e-06, "loss": 0.9193, "step": 1667 }, { "epoch": 0.2, "learning_rate": 4.982833683885838e-06, "loss": 1.2296, "step": 1668 }, { "epoch": 0.2, "learning_rate": 4.982796061076647e-06, "loss": 0.8184, "step": 1669 }, { "epoch": 0.2, "learning_rate": 4.982758397226629e-06, "loss": 0.6613, "step": 1670 }, { "epoch": 0.2, "learning_rate": 4.982720692336406e-06, "loss": 0.8235, "step": 1671 }, { "epoch": 0.2, "learning_rate": 4.9826829464066016e-06, "loss": 0.8181, "step": 1672 }, { "epoch": 0.2, "learning_rate": 4.98264515943784e-06, "loss": 0.9396, "step": 1673 }, { "epoch": 0.2, "learning_rate": 4.982607331430746e-06, "loss": 1.0961, "step": 1674 }, { "epoch": 0.2, "learning_rate": 4.982569462385944e-06, "loss": 1.1522, "step": 1675 }, { "epoch": 0.2, "learning_rate": 4.9825315523040605e-06, "loss": 0.6308, "step": 1676 }, { "epoch": 0.2, "learning_rate": 4.982493601185722e-06, "loss": 0.5938, "step": 1677 }, { "epoch": 0.2, "learning_rate": 4.982455609031557e-06, "loss": 0.7415, "step": 1678 }, { "epoch": 0.2, "learning_rate": 4.982417575842191e-06, "loss": 0.7914, "step": 1679 }, { "epoch": 0.2, "learning_rate": 4.982379501618255e-06, "loss": 1.0374, "step": 1680 }, { "epoch": 0.2, "learning_rate": 4.9823413863603775e-06, "loss": 1.2313, "step": 1681 }, { "epoch": 0.2, "learning_rate": 4.982303230069189e-06, "loss": 0.6807, "step": 1682 }, { "epoch": 0.2, "learning_rate": 4.982265032745319e-06, "loss": 0.7777, "step": 1683 }, { "epoch": 0.2, "learning_rate": 4.9822267943894e-06, "loss": 0.7659, "step": 1684 }, { "epoch": 0.2, "learning_rate": 4.982188515002064e-06, "loss": 0.9519, "step": 1685 }, { "epoch": 0.2, "learning_rate": 4.9821501945839435e-06, "loss": 1.0558, "step": 1686 }, { "epoch": 0.2, "learning_rate": 4.982111833135672e-06, "loss": 1.2535, "step": 1687 }, { "epoch": 0.2, "learning_rate": 4.982073430657883e-06, "loss": 0.6801, "step": 1688 }, { "epoch": 0.2, "learning_rate": 4.982034987151213e-06, "loss": 0.5591, "step": 1689 }, { "epoch": 0.2, "learning_rate": 4.981996502616296e-06, "loss": 0.7061, "step": 1690 }, { "epoch": 0.2, "learning_rate": 4.981957977053768e-06, "loss": 0.8212, "step": 1691 }, { "epoch": 0.2, "learning_rate": 4.981919410464267e-06, "loss": 0.8105, "step": 1692 }, { "epoch": 0.2, "learning_rate": 4.98188080284843e-06, "loss": 1.2249, "step": 1693 }, { "epoch": 0.2, "learning_rate": 4.981842154206895e-06, "loss": 0.9294, "step": 1694 }, { "epoch": 0.2, "learning_rate": 4.9818034645403005e-06, "loss": 0.6283, "step": 1695 }, { "epoch": 0.2, "learning_rate": 4.981764733849287e-06, "loss": 0.6639, "step": 1696 }, { "epoch": 0.2, "learning_rate": 4.981725962134494e-06, "loss": 0.7476, "step": 1697 }, { "epoch": 0.2, "learning_rate": 4.981687149396562e-06, "loss": 0.7786, "step": 1698 }, { "epoch": 0.2, "learning_rate": 4.981648295636135e-06, "loss": 1.1376, "step": 1699 }, { "epoch": 0.2, "learning_rate": 4.981609400853852e-06, "loss": 1.1501, "step": 1700 }, { "epoch": 0.2, "learning_rate": 4.981570465050357e-06, "loss": 0.651, "step": 1701 }, { "epoch": 0.2, "learning_rate": 4.9815314882262945e-06, "loss": 0.6466, "step": 1702 }, { "epoch": 0.2, "learning_rate": 4.981492470382309e-06, "loss": 0.8031, "step": 1703 }, { "epoch": 0.2, "learning_rate": 4.981453411519044e-06, "loss": 0.8811, "step": 1704 }, { "epoch": 0.2, "learning_rate": 4.981414311637146e-06, "loss": 1.0806, "step": 1705 }, { "epoch": 0.2, "learning_rate": 4.9813751707372615e-06, "loss": 1.2333, "step": 1706 }, { "epoch": 0.2, "learning_rate": 4.981335988820037e-06, "loss": 0.7871, "step": 1707 }, { "epoch": 0.2, "learning_rate": 4.98129676588612e-06, "loss": 0.7994, "step": 1708 }, { "epoch": 0.2, "learning_rate": 4.981257501936159e-06, "loss": 0.7936, "step": 1709 }, { "epoch": 0.2, "learning_rate": 4.981218196970805e-06, "loss": 0.9031, "step": 1710 }, { "epoch": 0.2, "learning_rate": 4.981178850990704e-06, "loss": 0.9847, "step": 1711 }, { "epoch": 0.2, "learning_rate": 4.981139463996511e-06, "loss": 1.1851, "step": 1712 }, { "epoch": 0.2, "learning_rate": 4.981100035988872e-06, "loss": 0.6431, "step": 1713 }, { "epoch": 0.2, "learning_rate": 4.981060566968442e-06, "loss": 0.6081, "step": 1714 }, { "epoch": 0.2, "learning_rate": 4.981021056935873e-06, "loss": 0.7763, "step": 1715 }, { "epoch": 0.2, "learning_rate": 4.980981505891818e-06, "loss": 0.8527, "step": 1716 }, { "epoch": 0.2, "learning_rate": 4.9809419138369295e-06, "loss": 0.9796, "step": 1717 }, { "epoch": 0.2, "learning_rate": 4.980902280771863e-06, "loss": 1.0619, "step": 1718 }, { "epoch": 0.2, "learning_rate": 4.980862606697274e-06, "loss": 0.8423, "step": 1719 }, { "epoch": 0.2, "learning_rate": 4.980822891613819e-06, "loss": 0.5513, "step": 1720 }, { "epoch": 0.2, "learning_rate": 4.980783135522153e-06, "loss": 0.669, "step": 1721 }, { "epoch": 0.21, "learning_rate": 4.980743338422932e-06, "loss": 0.7856, "step": 1722 }, { "epoch": 0.21, "learning_rate": 4.980703500316817e-06, "loss": 0.8971, "step": 1723 }, { "epoch": 0.21, "learning_rate": 4.980663621204465e-06, "loss": 1.171, "step": 1724 }, { "epoch": 0.21, "learning_rate": 4.980623701086534e-06, "loss": 1.1126, "step": 1725 }, { "epoch": 0.21, "learning_rate": 4.980583739963686e-06, "loss": 0.6365, "step": 1726 }, { "epoch": 0.21, "learning_rate": 4.98054373783658e-06, "loss": 0.7949, "step": 1727 }, { "epoch": 0.21, "learning_rate": 4.980503694705879e-06, "loss": 0.6891, "step": 1728 }, { "epoch": 0.21, "learning_rate": 4.9804636105722435e-06, "loss": 0.8514, "step": 1729 }, { "epoch": 0.21, "learning_rate": 4.980423485436336e-06, "loss": 1.0302, "step": 1730 }, { "epoch": 0.21, "learning_rate": 4.9803833192988205e-06, "loss": 1.2506, "step": 1731 }, { "epoch": 0.21, "learning_rate": 4.98034311216036e-06, "loss": 0.8295, "step": 1732 }, { "epoch": 0.21, "learning_rate": 4.98030286402162e-06, "loss": 0.6897, "step": 1733 }, { "epoch": 0.21, "learning_rate": 4.980262574883265e-06, "loss": 0.6954, "step": 1734 }, { "epoch": 0.21, "learning_rate": 4.980222244745963e-06, "loss": 0.7489, "step": 1735 }, { "epoch": 0.21, "learning_rate": 4.980181873610379e-06, "loss": 0.9369, "step": 1736 }, { "epoch": 0.21, "learning_rate": 4.980141461477179e-06, "loss": 1.2017, "step": 1737 }, { "epoch": 0.21, "learning_rate": 4.980101008347034e-06, "loss": 0.6444, "step": 1738 }, { "epoch": 0.21, "learning_rate": 4.98006051422061e-06, "loss": 0.4647, "step": 1739 }, { "epoch": 0.21, "learning_rate": 4.980019979098579e-06, "loss": 0.7288, "step": 1740 }, { "epoch": 0.21, "learning_rate": 4.979979402981608e-06, "loss": 0.8381, "step": 1741 }, { "epoch": 0.21, "learning_rate": 4.97993878587037e-06, "loss": 1.0635, "step": 1742 }, { "epoch": 0.21, "learning_rate": 4.9798981277655366e-06, "loss": 1.1661, "step": 1743 }, { "epoch": 0.21, "learning_rate": 4.979857428667779e-06, "loss": 0.9716, "step": 1744 }, { "epoch": 0.21, "learning_rate": 4.97981668857777e-06, "loss": 0.773, "step": 1745 }, { "epoch": 0.21, "learning_rate": 4.979775907496182e-06, "loss": 0.5557, "step": 1746 }, { "epoch": 0.21, "learning_rate": 4.979735085423691e-06, "loss": 0.7896, "step": 1747 }, { "epoch": 0.21, "learning_rate": 4.979694222360971e-06, "loss": 0.9283, "step": 1748 }, { "epoch": 0.21, "learning_rate": 4.979653318308697e-06, "loss": 1.2263, "step": 1749 }, { "epoch": 0.21, "learning_rate": 4.979612373267546e-06, "loss": 1.2791, "step": 1750 }, { "epoch": 0.21, "learning_rate": 4.979571387238194e-06, "loss": 0.575, "step": 1751 }, { "epoch": 0.21, "learning_rate": 4.97953036022132e-06, "loss": 0.7044, "step": 1752 }, { "epoch": 0.21, "learning_rate": 4.9794892922175994e-06, "loss": 0.7404, "step": 1753 }, { "epoch": 0.21, "learning_rate": 4.979448183227714e-06, "loss": 0.7915, "step": 1754 }, { "epoch": 0.21, "learning_rate": 4.979407033252342e-06, "loss": 1.1224, "step": 1755 }, { "epoch": 0.21, "learning_rate": 4.979365842292163e-06, "loss": 1.3356, "step": 1756 }, { "epoch": 0.21, "learning_rate": 4.979324610347859e-06, "loss": 0.6058, "step": 1757 }, { "epoch": 0.21, "learning_rate": 4.979283337420111e-06, "loss": 0.6853, "step": 1758 }, { "epoch": 0.21, "learning_rate": 4.9792420235096015e-06, "loss": 0.7806, "step": 1759 }, { "epoch": 0.21, "learning_rate": 4.9792006686170135e-06, "loss": 0.8459, "step": 1760 }, { "epoch": 0.21, "learning_rate": 4.97915927274303e-06, "loss": 0.999, "step": 1761 }, { "epoch": 0.21, "learning_rate": 4.979117835888337e-06, "loss": 1.4088, "step": 1762 }, { "epoch": 0.21, "learning_rate": 4.979076358053616e-06, "loss": 0.6527, "step": 1763 }, { "epoch": 0.21, "learning_rate": 4.979034839239556e-06, "loss": 0.5678, "step": 1764 }, { "epoch": 0.21, "learning_rate": 4.978993279446842e-06, "loss": 0.7416, "step": 1765 }, { "epoch": 0.21, "learning_rate": 4.978951678676161e-06, "loss": 0.8389, "step": 1766 }, { "epoch": 0.21, "learning_rate": 4.978910036928201e-06, "loss": 0.8727, "step": 1767 }, { "epoch": 0.21, "learning_rate": 4.9788683542036505e-06, "loss": 0.9894, "step": 1768 }, { "epoch": 0.21, "learning_rate": 4.978826630503198e-06, "loss": 0.7818, "step": 1769 }, { "epoch": 0.21, "learning_rate": 4.978784865827533e-06, "loss": 0.688, "step": 1770 }, { "epoch": 0.21, "learning_rate": 4.978743060177346e-06, "loss": 0.765, "step": 1771 }, { "epoch": 0.21, "learning_rate": 4.978701213553327e-06, "loss": 0.8708, "step": 1772 }, { "epoch": 0.21, "learning_rate": 4.9786593259561715e-06, "loss": 1.0123, "step": 1773 }, { "epoch": 0.21, "learning_rate": 4.978617397386568e-06, "loss": 1.1462, "step": 1774 }, { "epoch": 0.21, "learning_rate": 4.97857542784521e-06, "loss": 1.1458, "step": 1775 }, { "epoch": 0.21, "learning_rate": 4.978533417332794e-06, "loss": 0.5026, "step": 1776 }, { "epoch": 0.21, "learning_rate": 4.978491365850012e-06, "loss": 0.6688, "step": 1777 }, { "epoch": 0.21, "learning_rate": 4.97844927339756e-06, "loss": 0.9295, "step": 1778 }, { "epoch": 0.21, "learning_rate": 4.978407139976133e-06, "loss": 0.9823, "step": 1779 }, { "epoch": 0.21, "learning_rate": 4.978364965586428e-06, "loss": 1.1866, "step": 1780 }, { "epoch": 0.21, "learning_rate": 4.978322750229142e-06, "loss": 1.3119, "step": 1781 }, { "epoch": 0.21, "learning_rate": 4.978280493904974e-06, "loss": 0.5538, "step": 1782 }, { "epoch": 0.21, "learning_rate": 4.978238196614621e-06, "loss": 0.7604, "step": 1783 }, { "epoch": 0.21, "learning_rate": 4.978195858358782e-06, "loss": 0.7787, "step": 1784 }, { "epoch": 0.21, "learning_rate": 4.978153479138158e-06, "loss": 0.6689, "step": 1785 }, { "epoch": 0.21, "learning_rate": 4.97811105895345e-06, "loss": 1.0226, "step": 1786 }, { "epoch": 0.21, "learning_rate": 4.9780685978053565e-06, "loss": 1.215, "step": 1787 }, { "epoch": 0.21, "learning_rate": 4.978026095694582e-06, "loss": 0.4794, "step": 1788 }, { "epoch": 0.21, "learning_rate": 4.977983552621829e-06, "loss": 0.5989, "step": 1789 }, { "epoch": 0.21, "learning_rate": 4.977940968587799e-06, "loss": 0.7401, "step": 1790 }, { "epoch": 0.21, "learning_rate": 4.977898343593197e-06, "loss": 0.8755, "step": 1791 }, { "epoch": 0.21, "learning_rate": 4.9778556776387275e-06, "loss": 0.8745, "step": 1792 }, { "epoch": 0.21, "learning_rate": 4.977812970725096e-06, "loss": 1.224, "step": 1793 }, { "epoch": 0.21, "learning_rate": 4.977770222853008e-06, "loss": 0.8797, "step": 1794 }, { "epoch": 0.21, "learning_rate": 4.977727434023171e-06, "loss": 0.5956, "step": 1795 }, { "epoch": 0.21, "learning_rate": 4.97768460423629e-06, "loss": 0.804, "step": 1796 }, { "epoch": 0.21, "learning_rate": 4.977641733493076e-06, "loss": 0.8188, "step": 1797 }, { "epoch": 0.21, "learning_rate": 4.977598821794236e-06, "loss": 0.8919, "step": 1798 }, { "epoch": 0.21, "learning_rate": 4.977555869140479e-06, "loss": 1.2035, "step": 1799 }, { "epoch": 0.21, "learning_rate": 4.9775128755325165e-06, "loss": 1.1486, "step": 1800 }, { "epoch": 0.21, "learning_rate": 4.977469840971057e-06, "loss": 0.6195, "step": 1801 }, { "epoch": 0.21, "learning_rate": 4.977426765456815e-06, "loss": 0.7643, "step": 1802 }, { "epoch": 0.21, "learning_rate": 4.977383648990499e-06, "loss": 0.8327, "step": 1803 }, { "epoch": 0.21, "learning_rate": 4.977340491572824e-06, "loss": 1.0436, "step": 1804 }, { "epoch": 0.21, "learning_rate": 4.9772972932045026e-06, "loss": 1.0833, "step": 1805 }, { "epoch": 0.22, "learning_rate": 4.9772540538862496e-06, "loss": 1.2388, "step": 1806 }, { "epoch": 0.22, "learning_rate": 4.977210773618779e-06, "loss": 0.7076, "step": 1807 }, { "epoch": 0.22, "learning_rate": 4.977167452402806e-06, "loss": 0.8354, "step": 1808 }, { "epoch": 0.22, "learning_rate": 4.977124090239048e-06, "loss": 0.8718, "step": 1809 }, { "epoch": 0.22, "learning_rate": 4.9770806871282206e-06, "loss": 0.858, "step": 1810 }, { "epoch": 0.22, "learning_rate": 4.977037243071041e-06, "loss": 1.0722, "step": 1811 }, { "epoch": 0.22, "learning_rate": 4.976993758068229e-06, "loss": 1.1469, "step": 1812 }, { "epoch": 0.22, "learning_rate": 4.976950232120501e-06, "loss": 0.68, "step": 1813 }, { "epoch": 0.22, "learning_rate": 4.976906665228579e-06, "loss": 0.7198, "step": 1814 }, { "epoch": 0.22, "learning_rate": 4.976863057393182e-06, "loss": 0.8095, "step": 1815 }, { "epoch": 0.22, "learning_rate": 4.97681940861503e-06, "loss": 0.6833, "step": 1816 }, { "epoch": 0.22, "learning_rate": 4.976775718894846e-06, "loss": 1.0167, "step": 1817 }, { "epoch": 0.22, "learning_rate": 4.9767319882333505e-06, "loss": 1.2444, "step": 1818 }, { "epoch": 0.22, "learning_rate": 4.976688216631269e-06, "loss": 0.5692, "step": 1819 }, { "epoch": 0.22, "learning_rate": 4.976644404089323e-06, "loss": 0.6683, "step": 1820 }, { "epoch": 0.22, "learning_rate": 4.976600550608236e-06, "loss": 0.7256, "step": 1821 }, { "epoch": 0.22, "learning_rate": 4.976556656188736e-06, "loss": 0.6869, "step": 1822 }, { "epoch": 0.22, "learning_rate": 4.976512720831546e-06, "loss": 0.8458, "step": 1823 }, { "epoch": 0.22, "learning_rate": 4.976468744537391e-06, "loss": 1.102, "step": 1824 }, { "epoch": 0.22, "learning_rate": 4.976424727307002e-06, "loss": 1.2414, "step": 1825 }, { "epoch": 0.22, "learning_rate": 4.976380669141104e-06, "loss": 0.6088, "step": 1826 }, { "epoch": 0.22, "learning_rate": 4.976336570040426e-06, "loss": 0.7917, "step": 1827 }, { "epoch": 0.22, "learning_rate": 4.976292430005696e-06, "loss": 0.7032, "step": 1828 }, { "epoch": 0.22, "learning_rate": 4.9762482490376454e-06, "loss": 0.8544, "step": 1829 }, { "epoch": 0.22, "learning_rate": 4.976204027137003e-06, "loss": 1.0518, "step": 1830 }, { "epoch": 0.22, "learning_rate": 4.9761597643045e-06, "loss": 1.1568, "step": 1831 }, { "epoch": 0.22, "learning_rate": 4.976115460540868e-06, "loss": 0.6338, "step": 1832 }, { "epoch": 0.22, "learning_rate": 4.97607111584684e-06, "loss": 0.7607, "step": 1833 }, { "epoch": 0.22, "learning_rate": 4.976026730223149e-06, "loss": 0.6995, "step": 1834 }, { "epoch": 0.22, "learning_rate": 4.9759823036705275e-06, "loss": 0.8825, "step": 1835 }, { "epoch": 0.22, "learning_rate": 4.975937836189712e-06, "loss": 1.0978, "step": 1836 }, { "epoch": 0.22, "learning_rate": 4.9758933277814354e-06, "loss": 1.2501, "step": 1837 }, { "epoch": 0.22, "learning_rate": 4.975848778446435e-06, "loss": 0.587, "step": 1838 }, { "epoch": 0.22, "learning_rate": 4.975804188185447e-06, "loss": 0.6024, "step": 1839 }, { "epoch": 0.22, "learning_rate": 4.975759556999207e-06, "loss": 0.8532, "step": 1840 }, { "epoch": 0.22, "learning_rate": 4.975714884888454e-06, "loss": 0.9252, "step": 1841 }, { "epoch": 0.22, "learning_rate": 4.975670171853926e-06, "loss": 1.2354, "step": 1842 }, { "epoch": 0.22, "learning_rate": 4.975625417896363e-06, "loss": 1.0982, "step": 1843 }, { "epoch": 0.22, "learning_rate": 4.9755806230165036e-06, "loss": 0.9062, "step": 1844 }, { "epoch": 0.22, "learning_rate": 4.975535787215089e-06, "loss": 0.5541, "step": 1845 }, { "epoch": 0.22, "learning_rate": 4.97549091049286e-06, "loss": 0.7372, "step": 1846 }, { "epoch": 0.22, "learning_rate": 4.975445992850558e-06, "loss": 0.8127, "step": 1847 }, { "epoch": 0.22, "learning_rate": 4.975401034288927e-06, "loss": 0.9316, "step": 1848 }, { "epoch": 0.22, "learning_rate": 4.975356034808709e-06, "loss": 0.9893, "step": 1849 }, { "epoch": 0.22, "learning_rate": 4.975310994410648e-06, "loss": 1.2659, "step": 1850 }, { "epoch": 0.22, "learning_rate": 4.9752659130954885e-06, "loss": 0.5979, "step": 1851 }, { "epoch": 0.22, "learning_rate": 4.975220790863976e-06, "loss": 0.7172, "step": 1852 }, { "epoch": 0.22, "learning_rate": 4.975175627716856e-06, "loss": 0.7421, "step": 1853 }, { "epoch": 0.22, "learning_rate": 4.975130423654875e-06, "loss": 0.7478, "step": 1854 }, { "epoch": 0.22, "learning_rate": 4.97508517867878e-06, "loss": 1.0944, "step": 1855 }, { "epoch": 0.22, "learning_rate": 4.97503989278932e-06, "loss": 1.1772, "step": 1856 }, { "epoch": 0.22, "learning_rate": 4.974994565987243e-06, "loss": 0.601, "step": 1857 }, { "epoch": 0.22, "learning_rate": 4.974949198273298e-06, "loss": 0.5058, "step": 1858 }, { "epoch": 0.22, "learning_rate": 4.974903789648234e-06, "loss": 0.7466, "step": 1859 }, { "epoch": 0.22, "learning_rate": 4.974858340112803e-06, "loss": 0.8682, "step": 1860 }, { "epoch": 0.22, "learning_rate": 4.974812849667757e-06, "loss": 1.0173, "step": 1861 }, { "epoch": 0.22, "learning_rate": 4.9747673183138465e-06, "loss": 1.1284, "step": 1862 }, { "epoch": 0.22, "learning_rate": 4.974721746051825e-06, "loss": 0.6888, "step": 1863 }, { "epoch": 0.22, "learning_rate": 4.974676132882446e-06, "loss": 0.7443, "step": 1864 }, { "epoch": 0.22, "learning_rate": 4.974630478806461e-06, "loss": 0.7303, "step": 1865 }, { "epoch": 0.22, "learning_rate": 4.974584783824627e-06, "loss": 0.9673, "step": 1866 }, { "epoch": 0.22, "learning_rate": 4.9745390479377e-06, "loss": 0.8995, "step": 1867 }, { "epoch": 0.22, "learning_rate": 4.9744932711464344e-06, "loss": 1.1483, "step": 1868 }, { "epoch": 0.22, "learning_rate": 4.9744474534515866e-06, "loss": 0.8125, "step": 1869 }, { "epoch": 0.22, "learning_rate": 4.974401594853916e-06, "loss": 0.6168, "step": 1870 }, { "epoch": 0.22, "learning_rate": 4.9743556953541784e-06, "loss": 0.8098, "step": 1871 }, { "epoch": 0.22, "learning_rate": 4.9743097549531335e-06, "loss": 0.7339, "step": 1872 }, { "epoch": 0.22, "learning_rate": 4.974263773651542e-06, "loss": 0.9902, "step": 1873 }, { "epoch": 0.22, "learning_rate": 4.974217751450162e-06, "loss": 1.0598, "step": 1874 }, { "epoch": 0.22, "learning_rate": 4.974171688349755e-06, "loss": 1.0089, "step": 1875 }, { "epoch": 0.22, "learning_rate": 4.974125584351082e-06, "loss": 0.5772, "step": 1876 }, { "epoch": 0.22, "learning_rate": 4.974079439454906e-06, "loss": 0.5599, "step": 1877 }, { "epoch": 0.22, "learning_rate": 4.974033253661989e-06, "loss": 0.7428, "step": 1878 }, { "epoch": 0.22, "learning_rate": 4.973987026973095e-06, "loss": 0.7906, "step": 1879 }, { "epoch": 0.22, "learning_rate": 4.973940759388987e-06, "loss": 1.0161, "step": 1880 }, { "epoch": 0.22, "learning_rate": 4.973894450910432e-06, "loss": 1.1573, "step": 1881 }, { "epoch": 0.22, "learning_rate": 4.9738481015381925e-06, "loss": 0.6722, "step": 1882 }, { "epoch": 0.22, "learning_rate": 4.973801711273037e-06, "loss": 0.7292, "step": 1883 }, { "epoch": 0.22, "learning_rate": 4.973755280115733e-06, "loss": 0.8029, "step": 1884 }, { "epoch": 0.22, "learning_rate": 4.9737088080670445e-06, "loss": 0.9957, "step": 1885 }, { "epoch": 0.22, "learning_rate": 4.973662295127743e-06, "loss": 0.9885, "step": 1886 }, { "epoch": 0.22, "learning_rate": 4.973615741298596e-06, "loss": 1.2119, "step": 1887 }, { "epoch": 0.22, "learning_rate": 4.973569146580374e-06, "loss": 0.4943, "step": 1888 }, { "epoch": 0.22, "learning_rate": 4.973522510973846e-06, "loss": 0.6452, "step": 1889 }, { "epoch": 0.23, "learning_rate": 4.973475834479783e-06, "loss": 0.7117, "step": 1890 }, { "epoch": 0.23, "learning_rate": 4.9734291170989565e-06, "loss": 0.9523, "step": 1891 }, { "epoch": 0.23, "learning_rate": 4.97338235883214e-06, "loss": 0.9424, "step": 1892 }, { "epoch": 0.23, "learning_rate": 4.973335559680106e-06, "loss": 1.1714, "step": 1893 }, { "epoch": 0.23, "learning_rate": 4.973288719643626e-06, "loss": 0.7204, "step": 1894 }, { "epoch": 0.23, "learning_rate": 4.973241838723477e-06, "loss": 0.709, "step": 1895 }, { "epoch": 0.23, "learning_rate": 4.973194916920433e-06, "loss": 0.6883, "step": 1896 }, { "epoch": 0.23, "learning_rate": 4.973147954235269e-06, "loss": 0.911, "step": 1897 }, { "epoch": 0.23, "learning_rate": 4.973100950668762e-06, "loss": 0.8715, "step": 1898 }, { "epoch": 0.23, "learning_rate": 4.973053906221689e-06, "loss": 1.0031, "step": 1899 }, { "epoch": 0.23, "learning_rate": 4.973006820894828e-06, "loss": 1.136, "step": 1900 }, { "epoch": 0.23, "learning_rate": 4.972959694688956e-06, "loss": 0.6012, "step": 1901 }, { "epoch": 0.23, "learning_rate": 4.972912527604852e-06, "loss": 0.7016, "step": 1902 }, { "epoch": 0.23, "learning_rate": 4.972865319643297e-06, "loss": 0.79, "step": 1903 }, { "epoch": 0.23, "learning_rate": 4.9728180708050715e-06, "loss": 0.9024, "step": 1904 }, { "epoch": 0.23, "learning_rate": 4.972770781090955e-06, "loss": 0.9266, "step": 1905 }, { "epoch": 0.23, "learning_rate": 4.972723450501731e-06, "loss": 1.1663, "step": 1906 }, { "epoch": 0.23, "learning_rate": 4.97267607903818e-06, "loss": 0.5361, "step": 1907 }, { "epoch": 0.23, "learning_rate": 4.972628666701086e-06, "loss": 0.7868, "step": 1908 }, { "epoch": 0.23, "learning_rate": 4.9725812134912325e-06, "loss": 0.8772, "step": 1909 }, { "epoch": 0.23, "learning_rate": 4.972533719409405e-06, "loss": 0.8594, "step": 1910 }, { "epoch": 0.23, "learning_rate": 4.9724861844563865e-06, "loss": 1.0751, "step": 1911 }, { "epoch": 0.23, "learning_rate": 4.972438608632964e-06, "loss": 1.288, "step": 1912 }, { "epoch": 0.23, "learning_rate": 4.9723909919399246e-06, "loss": 0.7106, "step": 1913 }, { "epoch": 0.23, "learning_rate": 4.972343334378055e-06, "loss": 0.6228, "step": 1914 }, { "epoch": 0.23, "learning_rate": 4.972295635948142e-06, "loss": 0.679, "step": 1915 }, { "epoch": 0.23, "learning_rate": 4.972247896650975e-06, "loss": 0.6308, "step": 1916 }, { "epoch": 0.23, "learning_rate": 4.972200116487342e-06, "loss": 0.9052, "step": 1917 }, { "epoch": 0.23, "learning_rate": 4.972152295458034e-06, "loss": 1.0011, "step": 1918 }, { "epoch": 0.23, "learning_rate": 4.9721044335638416e-06, "loss": 0.8273, "step": 1919 }, { "epoch": 0.23, "learning_rate": 4.9720565308055545e-06, "loss": 0.5647, "step": 1920 }, { "epoch": 0.23, "learning_rate": 4.972008587183966e-06, "loss": 0.7611, "step": 1921 }, { "epoch": 0.23, "learning_rate": 4.971960602699869e-06, "loss": 0.8375, "step": 1922 }, { "epoch": 0.23, "learning_rate": 4.971912577354055e-06, "loss": 0.8243, "step": 1923 }, { "epoch": 0.23, "learning_rate": 4.971864511147319e-06, "loss": 1.0422, "step": 1924 }, { "epoch": 0.23, "learning_rate": 4.971816404080455e-06, "loss": 1.3354, "step": 1925 }, { "epoch": 0.23, "learning_rate": 4.971768256154258e-06, "loss": 0.5087, "step": 1926 }, { "epoch": 0.23, "learning_rate": 4.971720067369525e-06, "loss": 0.6754, "step": 1927 }, { "epoch": 0.23, "learning_rate": 4.971671837727051e-06, "loss": 0.7292, "step": 1928 }, { "epoch": 0.23, "learning_rate": 4.9716235672276355e-06, "loss": 0.8123, "step": 1929 }, { "epoch": 0.23, "learning_rate": 4.971575255872074e-06, "loss": 0.9537, "step": 1930 }, { "epoch": 0.23, "learning_rate": 4.971526903661167e-06, "loss": 1.1326, "step": 1931 }, { "epoch": 0.23, "learning_rate": 4.971478510595712e-06, "loss": 0.5493, "step": 1932 }, { "epoch": 0.23, "learning_rate": 4.97143007667651e-06, "loss": 0.5632, "step": 1933 }, { "epoch": 0.23, "learning_rate": 4.971381601904362e-06, "loss": 0.9263, "step": 1934 }, { "epoch": 0.23, "learning_rate": 4.971333086280068e-06, "loss": 0.8087, "step": 1935 }, { "epoch": 0.23, "learning_rate": 4.971284529804432e-06, "loss": 0.9507, "step": 1936 }, { "epoch": 0.23, "learning_rate": 4.9712359324782545e-06, "loss": 1.1005, "step": 1937 }, { "epoch": 0.23, "learning_rate": 4.97118729430234e-06, "loss": 0.5352, "step": 1938 }, { "epoch": 0.23, "learning_rate": 4.971138615277492e-06, "loss": 0.6468, "step": 1939 }, { "epoch": 0.23, "learning_rate": 4.971089895404515e-06, "loss": 0.8373, "step": 1940 }, { "epoch": 0.23, "learning_rate": 4.9710411346842144e-06, "loss": 0.9501, "step": 1941 }, { "epoch": 0.23, "learning_rate": 4.9709923331173975e-06, "loss": 1.0516, "step": 1942 }, { "epoch": 0.23, "learning_rate": 4.97094349070487e-06, "loss": 1.2673, "step": 1943 }, { "epoch": 0.23, "learning_rate": 4.970894607447438e-06, "loss": 0.9605, "step": 1944 }, { "epoch": 0.23, "learning_rate": 4.970845683345912e-06, "loss": 0.6585, "step": 1945 }, { "epoch": 0.23, "learning_rate": 4.970796718401099e-06, "loss": 0.7989, "step": 1946 }, { "epoch": 0.23, "learning_rate": 4.970747712613809e-06, "loss": 0.7585, "step": 1947 }, { "epoch": 0.23, "learning_rate": 4.970698665984852e-06, "loss": 0.9023, "step": 1948 }, { "epoch": 0.23, "learning_rate": 4.970649578515039e-06, "loss": 1.0497, "step": 1949 }, { "epoch": 0.23, "learning_rate": 4.970600450205181e-06, "loss": 1.121, "step": 1950 }, { "epoch": 0.23, "learning_rate": 4.970551281056091e-06, "loss": 0.5803, "step": 1951 }, { "epoch": 0.23, "learning_rate": 4.970502071068579e-06, "loss": 0.6273, "step": 1952 }, { "epoch": 0.23, "learning_rate": 4.970452820243462e-06, "loss": 0.8276, "step": 1953 }, { "epoch": 0.23, "learning_rate": 4.9704035285815525e-06, "loss": 0.8816, "step": 1954 }, { "epoch": 0.23, "learning_rate": 4.970354196083666e-06, "loss": 1.0073, "step": 1955 }, { "epoch": 0.23, "learning_rate": 4.970304822750616e-06, "loss": 1.1238, "step": 1956 }, { "epoch": 0.23, "learning_rate": 4.970255408583221e-06, "loss": 0.6472, "step": 1957 }, { "epoch": 0.23, "learning_rate": 4.970205953582296e-06, "loss": 0.6927, "step": 1958 }, { "epoch": 0.23, "learning_rate": 4.97015645774866e-06, "loss": 0.795, "step": 1959 }, { "epoch": 0.23, "learning_rate": 4.97010692108313e-06, "loss": 0.9686, "step": 1960 }, { "epoch": 0.23, "learning_rate": 4.9700573435865254e-06, "loss": 1.0021, "step": 1961 }, { "epoch": 0.23, "learning_rate": 4.970007725259665e-06, "loss": 1.1542, "step": 1962 }, { "epoch": 0.23, "learning_rate": 4.969958066103371e-06, "loss": 0.5932, "step": 1963 }, { "epoch": 0.23, "learning_rate": 4.969908366118462e-06, "loss": 0.5537, "step": 1964 }, { "epoch": 0.23, "learning_rate": 4.969858625305762e-06, "loss": 0.6378, "step": 1965 }, { "epoch": 0.23, "learning_rate": 4.96980884366609e-06, "loss": 0.8623, "step": 1966 }, { "epoch": 0.23, "learning_rate": 4.969759021200271e-06, "loss": 0.8222, "step": 1967 }, { "epoch": 0.23, "learning_rate": 4.969709157909128e-06, "loss": 1.0863, "step": 1968 }, { "epoch": 0.23, "learning_rate": 4.969659253793486e-06, "loss": 0.8043, "step": 1969 }, { "epoch": 0.23, "learning_rate": 4.969609308854169e-06, "loss": 0.6057, "step": 1970 }, { "epoch": 0.23, "learning_rate": 4.969559323092004e-06, "loss": 0.735, "step": 1971 }, { "epoch": 0.23, "learning_rate": 4.969509296507814e-06, "loss": 0.7139, "step": 1972 }, { "epoch": 0.23, "learning_rate": 4.96945922910243e-06, "loss": 0.8989, "step": 1973 }, { "epoch": 0.24, "learning_rate": 4.969409120876677e-06, "loss": 1.0823, "step": 1974 }, { "epoch": 0.24, "learning_rate": 4.969358971831384e-06, "loss": 1.1742, "step": 1975 }, { "epoch": 0.24, "learning_rate": 4.969308781967379e-06, "loss": 0.5391, "step": 1976 }, { "epoch": 0.24, "learning_rate": 4.969258551285494e-06, "loss": 0.8214, "step": 1977 }, { "epoch": 0.24, "learning_rate": 4.969208279786558e-06, "loss": 0.8599, "step": 1978 }, { "epoch": 0.24, "learning_rate": 4.9691579674714e-06, "loss": 0.8306, "step": 1979 }, { "epoch": 0.24, "learning_rate": 4.9691076143408565e-06, "loss": 0.9617, "step": 1980 }, { "epoch": 0.24, "learning_rate": 4.969057220395756e-06, "loss": 1.0851, "step": 1981 }, { "epoch": 0.24, "learning_rate": 4.969006785636931e-06, "loss": 0.6463, "step": 1982 }, { "epoch": 0.24, "learning_rate": 4.9689563100652185e-06, "loss": 0.6226, "step": 1983 }, { "epoch": 0.24, "learning_rate": 4.96890579368145e-06, "loss": 0.7287, "step": 1984 }, { "epoch": 0.24, "learning_rate": 4.968855236486462e-06, "loss": 0.9044, "step": 1985 }, { "epoch": 0.24, "learning_rate": 4.968804638481089e-06, "loss": 0.9879, "step": 1986 }, { "epoch": 0.24, "learning_rate": 4.96875399966617e-06, "loss": 1.1956, "step": 1987 }, { "epoch": 0.24, "learning_rate": 4.9687033200425386e-06, "loss": 0.6589, "step": 1988 }, { "epoch": 0.24, "learning_rate": 4.968652599611034e-06, "loss": 0.6365, "step": 1989 }, { "epoch": 0.24, "learning_rate": 4.968601838372496e-06, "loss": 0.8712, "step": 1990 }, { "epoch": 0.24, "learning_rate": 4.968551036327761e-06, "loss": 0.9012, "step": 1991 }, { "epoch": 0.24, "learning_rate": 4.968500193477672e-06, "loss": 0.9868, "step": 1992 }, { "epoch": 0.24, "learning_rate": 4.968449309823066e-06, "loss": 1.056, "step": 1993 }, { "epoch": 0.24, "learning_rate": 4.968398385364787e-06, "loss": 0.8386, "step": 1994 }, { "epoch": 0.24, "learning_rate": 4.968347420103675e-06, "loss": 0.7139, "step": 1995 }, { "epoch": 0.24, "learning_rate": 4.968296414040573e-06, "loss": 0.77, "step": 1996 }, { "epoch": 0.24, "learning_rate": 4.968245367176324e-06, "loss": 0.6925, "step": 1997 }, { "epoch": 0.24, "learning_rate": 4.968194279511773e-06, "loss": 0.9225, "step": 1998 }, { "epoch": 0.24, "learning_rate": 4.968143151047763e-06, "loss": 1.0093, "step": 1999 }, { "epoch": 0.24, "learning_rate": 4.968091981785139e-06, "loss": 1.2112, "step": 2000 }, { "epoch": 0.24, "learning_rate": 4.968040771724748e-06, "loss": 0.4355, "step": 2001 }, { "epoch": 0.24, "learning_rate": 4.967989520867436e-06, "loss": 0.7183, "step": 2002 }, { "epoch": 0.24, "learning_rate": 4.96793822921405e-06, "loss": 0.7179, "step": 2003 }, { "epoch": 0.24, "learning_rate": 4.9678868967654385e-06, "loss": 0.8893, "step": 2004 }, { "epoch": 0.24, "learning_rate": 4.9678355235224486e-06, "loss": 1.0, "step": 2005 }, { "epoch": 0.24, "learning_rate": 4.967784109485931e-06, "loss": 1.1433, "step": 2006 }, { "epoch": 0.24, "learning_rate": 4.967732654656734e-06, "loss": 0.5925, "step": 2007 }, { "epoch": 0.24, "learning_rate": 4.967681159035711e-06, "loss": 0.5527, "step": 2008 }, { "epoch": 0.24, "learning_rate": 4.96762962262371e-06, "loss": 0.8578, "step": 2009 }, { "epoch": 0.24, "learning_rate": 4.967578045421585e-06, "loss": 0.7823, "step": 2010 }, { "epoch": 0.24, "learning_rate": 4.967526427430187e-06, "loss": 0.9341, "step": 2011 }, { "epoch": 0.24, "learning_rate": 4.967474768650371e-06, "loss": 1.1735, "step": 2012 }, { "epoch": 0.24, "learning_rate": 4.9674230690829895e-06, "loss": 0.7438, "step": 2013 }, { "epoch": 0.24, "learning_rate": 4.9673713287288975e-06, "loss": 0.6676, "step": 2014 }, { "epoch": 0.24, "learning_rate": 4.967319547588951e-06, "loss": 0.8055, "step": 2015 }, { "epoch": 0.24, "learning_rate": 4.967267725664005e-06, "loss": 0.709, "step": 2016 }, { "epoch": 0.24, "learning_rate": 4.967215862954916e-06, "loss": 1.0275, "step": 2017 }, { "epoch": 0.24, "learning_rate": 4.967163959462543e-06, "loss": 1.1804, "step": 2018 }, { "epoch": 0.24, "learning_rate": 4.967112015187742e-06, "loss": 0.7082, "step": 2019 }, { "epoch": 0.24, "learning_rate": 4.967060030131372e-06, "loss": 0.5132, "step": 2020 }, { "epoch": 0.24, "learning_rate": 4.967008004294294e-06, "loss": 0.8976, "step": 2021 }, { "epoch": 0.24, "learning_rate": 4.966955937677365e-06, "loss": 0.7873, "step": 2022 }, { "epoch": 0.24, "learning_rate": 4.966903830281449e-06, "loss": 0.9478, "step": 2023 }, { "epoch": 0.24, "learning_rate": 4.966851682107404e-06, "loss": 0.9997, "step": 2024 }, { "epoch": 0.24, "learning_rate": 4.966799493156095e-06, "loss": 1.171, "step": 2025 }, { "epoch": 0.24, "learning_rate": 4.966747263428383e-06, "loss": 0.6635, "step": 2026 }, { "epoch": 0.24, "learning_rate": 4.966694992925132e-06, "loss": 0.6815, "step": 2027 }, { "epoch": 0.24, "learning_rate": 4.9666426816472045e-06, "loss": 0.8283, "step": 2028 }, { "epoch": 0.24, "learning_rate": 4.966590329595468e-06, "loss": 0.9668, "step": 2029 }, { "epoch": 0.24, "learning_rate": 4.966537936770786e-06, "loss": 1.1257, "step": 2030 }, { "epoch": 0.24, "learning_rate": 4.966485503174025e-06, "loss": 1.283, "step": 2031 }, { "epoch": 0.24, "learning_rate": 4.966433028806051e-06, "loss": 0.5403, "step": 2032 }, { "epoch": 0.24, "learning_rate": 4.9663805136677325e-06, "loss": 0.6607, "step": 2033 }, { "epoch": 0.24, "learning_rate": 4.9663279577599375e-06, "loss": 0.711, "step": 2034 }, { "epoch": 0.24, "learning_rate": 4.9662753610835345e-06, "loss": 0.8213, "step": 2035 }, { "epoch": 0.24, "learning_rate": 4.9662227236393925e-06, "loss": 0.9943, "step": 2036 }, { "epoch": 0.24, "learning_rate": 4.966170045428382e-06, "loss": 1.1933, "step": 2037 }, { "epoch": 0.24, "learning_rate": 4.966117326451374e-06, "loss": 0.5841, "step": 2038 }, { "epoch": 0.24, "learning_rate": 4.966064566709239e-06, "loss": 0.6797, "step": 2039 }, { "epoch": 0.24, "learning_rate": 4.966011766202851e-06, "loss": 0.6802, "step": 2040 }, { "epoch": 0.24, "learning_rate": 4.96595892493308e-06, "loss": 0.9213, "step": 2041 }, { "epoch": 0.24, "learning_rate": 4.9659060429008025e-06, "loss": 0.9323, "step": 2042 }, { "epoch": 0.24, "learning_rate": 4.96585312010689e-06, "loss": 1.0964, "step": 2043 }, { "epoch": 0.24, "learning_rate": 4.965800156552219e-06, "loss": 0.7559, "step": 2044 }, { "epoch": 0.24, "learning_rate": 4.965747152237664e-06, "loss": 0.5787, "step": 2045 }, { "epoch": 0.24, "learning_rate": 4.965694107164103e-06, "loss": 0.7095, "step": 2046 }, { "epoch": 0.24, "learning_rate": 4.9656410213324104e-06, "loss": 0.7954, "step": 2047 }, { "epoch": 0.24, "learning_rate": 4.965587894743465e-06, "loss": 0.8055, "step": 2048 }, { "epoch": 0.24, "learning_rate": 4.965534727398145e-06, "loss": 1.1335, "step": 2049 }, { "epoch": 0.24, "learning_rate": 4.9654815192973294e-06, "loss": 1.3026, "step": 2050 }, { "epoch": 0.24, "learning_rate": 4.965428270441897e-06, "loss": 0.663, "step": 2051 }, { "epoch": 0.24, "learning_rate": 4.965374980832729e-06, "loss": 0.681, "step": 2052 }, { "epoch": 0.24, "learning_rate": 4.965321650470705e-06, "loss": 0.5775, "step": 2053 }, { "epoch": 0.24, "learning_rate": 4.965268279356707e-06, "loss": 0.9841, "step": 2054 }, { "epoch": 0.24, "learning_rate": 4.965214867491618e-06, "loss": 1.0707, "step": 2055 }, { "epoch": 0.24, "learning_rate": 4.96516141487632e-06, "loss": 1.1645, "step": 2056 }, { "epoch": 0.24, "learning_rate": 4.965107921511698e-06, "loss": 0.8499, "step": 2057 }, { "epoch": 0.25, "learning_rate": 4.965054387398635e-06, "loss": 0.789, "step": 2058 }, { "epoch": 0.25, "learning_rate": 4.965000812538016e-06, "loss": 0.7135, "step": 2059 }, { "epoch": 0.25, "learning_rate": 4.964947196930726e-06, "loss": 0.7654, "step": 2060 }, { "epoch": 0.25, "learning_rate": 4.964893540577652e-06, "loss": 0.9149, "step": 2061 }, { "epoch": 0.25, "learning_rate": 4.964839843479682e-06, "loss": 1.0593, "step": 2062 }, { "epoch": 0.25, "learning_rate": 4.964786105637702e-06, "loss": 0.6974, "step": 2063 }, { "epoch": 0.25, "learning_rate": 4.9647323270526e-06, "loss": 0.6358, "step": 2064 }, { "epoch": 0.25, "learning_rate": 4.964678507725267e-06, "loss": 0.7969, "step": 2065 }, { "epoch": 0.25, "learning_rate": 4.964624647656591e-06, "loss": 0.7585, "step": 2066 }, { "epoch": 0.25, "learning_rate": 4.9645707468474635e-06, "loss": 0.8979, "step": 2067 }, { "epoch": 0.25, "learning_rate": 4.9645168052987734e-06, "loss": 1.0234, "step": 2068 }, { "epoch": 0.25, "learning_rate": 4.964462823011415e-06, "loss": 0.8254, "step": 2069 }, { "epoch": 0.25, "learning_rate": 4.964408799986279e-06, "loss": 0.7748, "step": 2070 }, { "epoch": 0.25, "learning_rate": 4.964354736224258e-06, "loss": 0.681, "step": 2071 }, { "epoch": 0.25, "learning_rate": 4.964300631726248e-06, "loss": 0.7857, "step": 2072 }, { "epoch": 0.25, "learning_rate": 4.96424648649314e-06, "loss": 0.8348, "step": 2073 }, { "epoch": 0.25, "learning_rate": 4.9641923005258316e-06, "loss": 0.9493, "step": 2074 }, { "epoch": 0.25, "learning_rate": 4.964138073825218e-06, "loss": 1.0524, "step": 2075 }, { "epoch": 0.25, "learning_rate": 4.964083806392194e-06, "loss": 0.741, "step": 2076 }, { "epoch": 0.25, "learning_rate": 4.96402949822766e-06, "loss": 0.6742, "step": 2077 }, { "epoch": 0.25, "learning_rate": 4.96397514933251e-06, "loss": 0.9023, "step": 2078 }, { "epoch": 0.25, "learning_rate": 4.963920759707645e-06, "loss": 0.8455, "step": 2079 }, { "epoch": 0.25, "learning_rate": 4.9638663293539625e-06, "loss": 1.0541, "step": 2080 }, { "epoch": 0.25, "learning_rate": 4.963811858272363e-06, "loss": 1.1199, "step": 2081 }, { "epoch": 0.25, "learning_rate": 4.963757346463747e-06, "loss": 0.7076, "step": 2082 }, { "epoch": 0.25, "learning_rate": 4.963702793929015e-06, "loss": 0.6024, "step": 2083 }, { "epoch": 0.25, "learning_rate": 4.963648200669069e-06, "loss": 0.774, "step": 2084 }, { "epoch": 0.25, "learning_rate": 4.9635935666848125e-06, "loss": 0.6421, "step": 2085 }, { "epoch": 0.25, "learning_rate": 4.963538891977146e-06, "loss": 0.8601, "step": 2086 }, { "epoch": 0.25, "learning_rate": 4.963484176546977e-06, "loss": 1.1873, "step": 2087 }, { "epoch": 0.25, "learning_rate": 4.963429420395206e-06, "loss": 0.7055, "step": 2088 }, { "epoch": 0.25, "learning_rate": 4.963374623522741e-06, "loss": 0.9306, "step": 2089 }, { "epoch": 0.25, "learning_rate": 4.963319785930486e-06, "loss": 0.808, "step": 2090 }, { "epoch": 0.25, "learning_rate": 4.963264907619349e-06, "loss": 0.6639, "step": 2091 }, { "epoch": 0.25, "learning_rate": 4.963209988590237e-06, "loss": 1.019, "step": 2092 }, { "epoch": 0.25, "learning_rate": 4.963155028844056e-06, "loss": 1.0684, "step": 2093 }, { "epoch": 0.25, "learning_rate": 4.963100028381715e-06, "loss": 0.7445, "step": 2094 }, { "epoch": 0.25, "learning_rate": 4.963044987204125e-06, "loss": 0.6685, "step": 2095 }, { "epoch": 0.25, "learning_rate": 4.962989905312194e-06, "loss": 0.7537, "step": 2096 }, { "epoch": 0.25, "learning_rate": 4.962934782706834e-06, "loss": 0.7061, "step": 2097 }, { "epoch": 0.25, "learning_rate": 4.962879619388955e-06, "loss": 0.8677, "step": 2098 }, { "epoch": 0.25, "learning_rate": 4.962824415359469e-06, "loss": 1.1206, "step": 2099 }, { "epoch": 0.25, "learning_rate": 4.96276917061929e-06, "loss": 1.267, "step": 2100 }, { "epoch": 0.25, "learning_rate": 4.962713885169329e-06, "loss": 0.5759, "step": 2101 }, { "epoch": 0.25, "learning_rate": 4.962658559010501e-06, "loss": 0.7584, "step": 2102 }, { "epoch": 0.25, "learning_rate": 4.96260319214372e-06, "loss": 0.7578, "step": 2103 }, { "epoch": 0.25, "learning_rate": 4.9625477845699024e-06, "loss": 0.7712, "step": 2104 }, { "epoch": 0.25, "learning_rate": 4.962492336289963e-06, "loss": 1.0602, "step": 2105 }, { "epoch": 0.25, "learning_rate": 4.962436847304818e-06, "loss": 1.2606, "step": 2106 }, { "epoch": 0.25, "learning_rate": 4.9623813176153855e-06, "loss": 0.6489, "step": 2107 }, { "epoch": 0.25, "learning_rate": 4.9623257472225836e-06, "loss": 0.6765, "step": 2108 }, { "epoch": 0.25, "learning_rate": 4.962270136127331e-06, "loss": 0.7937, "step": 2109 }, { "epoch": 0.25, "learning_rate": 4.9622144843305454e-06, "loss": 0.9044, "step": 2110 }, { "epoch": 0.25, "learning_rate": 4.962158791833148e-06, "loss": 0.9559, "step": 2111 }, { "epoch": 0.25, "learning_rate": 4.96210305863606e-06, "loss": 1.0677, "step": 2112 }, { "epoch": 0.25, "learning_rate": 4.962047284740201e-06, "loss": 0.7025, "step": 2113 }, { "epoch": 0.25, "learning_rate": 4.961991470146493e-06, "loss": 0.601, "step": 2114 }, { "epoch": 0.25, "learning_rate": 4.961935614855861e-06, "loss": 0.7734, "step": 2115 }, { "epoch": 0.25, "learning_rate": 4.961879718869226e-06, "loss": 0.9294, "step": 2116 }, { "epoch": 0.25, "learning_rate": 4.961823782187513e-06, "loss": 0.9947, "step": 2117 }, { "epoch": 0.25, "learning_rate": 4.961767804811647e-06, "loss": 1.029, "step": 2118 }, { "epoch": 0.25, "learning_rate": 4.961711786742551e-06, "loss": 1.0294, "step": 2119 }, { "epoch": 0.25, "learning_rate": 4.961655727981153e-06, "loss": 0.5762, "step": 2120 }, { "epoch": 0.25, "learning_rate": 4.96159962852838e-06, "loss": 0.7733, "step": 2121 }, { "epoch": 0.25, "learning_rate": 4.961543488385159e-06, "loss": 0.7422, "step": 2122 }, { "epoch": 0.25, "learning_rate": 4.961487307552416e-06, "loss": 0.9727, "step": 2123 }, { "epoch": 0.25, "learning_rate": 4.961431086031082e-06, "loss": 1.1586, "step": 2124 }, { "epoch": 0.25, "learning_rate": 4.9613748238220864e-06, "loss": 1.3157, "step": 2125 }, { "epoch": 0.25, "learning_rate": 4.961318520926357e-06, "loss": 0.6391, "step": 2126 }, { "epoch": 0.25, "learning_rate": 4.961262177344827e-06, "loss": 0.741, "step": 2127 }, { "epoch": 0.25, "learning_rate": 4.961205793078426e-06, "loss": 0.7385, "step": 2128 }, { "epoch": 0.25, "learning_rate": 4.961149368128088e-06, "loss": 0.9473, "step": 2129 }, { "epoch": 0.25, "learning_rate": 4.961092902494743e-06, "loss": 1.0667, "step": 2130 }, { "epoch": 0.25, "learning_rate": 4.961036396179327e-06, "loss": 1.2559, "step": 2131 }, { "epoch": 0.25, "learning_rate": 4.9609798491827724e-06, "loss": 0.6161, "step": 2132 }, { "epoch": 0.25, "learning_rate": 4.960923261506014e-06, "loss": 0.6469, "step": 2133 }, { "epoch": 0.25, "learning_rate": 4.960866633149988e-06, "loss": 0.8235, "step": 2134 }, { "epoch": 0.25, "learning_rate": 4.9608099641156295e-06, "loss": 0.8183, "step": 2135 }, { "epoch": 0.25, "learning_rate": 4.960753254403876e-06, "loss": 0.9857, "step": 2136 }, { "epoch": 0.25, "learning_rate": 4.960696504015665e-06, "loss": 1.1403, "step": 2137 }, { "epoch": 0.25, "learning_rate": 4.960639712951934e-06, "loss": 0.6204, "step": 2138 }, { "epoch": 0.25, "learning_rate": 4.960582881213622e-06, "loss": 0.7061, "step": 2139 }, { "epoch": 0.25, "learning_rate": 4.96052600880167e-06, "loss": 0.7131, "step": 2140 }, { "epoch": 0.25, "learning_rate": 4.9604690957170155e-06, "loss": 0.7035, "step": 2141 }, { "epoch": 0.26, "learning_rate": 4.9604121419606e-06, "loss": 0.9066, "step": 2142 }, { "epoch": 0.26, "learning_rate": 4.960355147533366e-06, "loss": 1.0964, "step": 2143 }, { "epoch": 0.26, "learning_rate": 4.960298112436255e-06, "loss": 0.6632, "step": 2144 }, { "epoch": 0.26, "learning_rate": 4.9602410366702095e-06, "loss": 0.5924, "step": 2145 }, { "epoch": 0.26, "learning_rate": 4.960183920236173e-06, "loss": 0.7336, "step": 2146 }, { "epoch": 0.26, "learning_rate": 4.96012676313509e-06, "loss": 0.8618, "step": 2147 }, { "epoch": 0.26, "learning_rate": 4.960069565367905e-06, "loss": 1.0037, "step": 2148 }, { "epoch": 0.26, "learning_rate": 4.960012326935564e-06, "loss": 1.1484, "step": 2149 }, { "epoch": 0.26, "learning_rate": 4.959955047839013e-06, "loss": 1.2364, "step": 2150 }, { "epoch": 0.26, "learning_rate": 4.959897728079198e-06, "loss": 0.6263, "step": 2151 }, { "epoch": 0.26, "learning_rate": 4.959840367657068e-06, "loss": 0.6499, "step": 2152 }, { "epoch": 0.26, "learning_rate": 4.95978296657357e-06, "loss": 0.882, "step": 2153 }, { "epoch": 0.26, "learning_rate": 4.959725524829653e-06, "loss": 0.9546, "step": 2154 }, { "epoch": 0.26, "learning_rate": 4.959668042426267e-06, "loss": 0.9713, "step": 2155 }, { "epoch": 0.26, "learning_rate": 4.959610519364361e-06, "loss": 1.2773, "step": 2156 }, { "epoch": 0.26, "learning_rate": 4.959552955644887e-06, "loss": 0.5172, "step": 2157 }, { "epoch": 0.26, "learning_rate": 4.959495351268797e-06, "loss": 0.776, "step": 2158 }, { "epoch": 0.26, "learning_rate": 4.959437706237043e-06, "loss": 0.6381, "step": 2159 }, { "epoch": 0.26, "learning_rate": 4.959380020550576e-06, "loss": 0.7557, "step": 2160 }, { "epoch": 0.26, "learning_rate": 4.959322294210351e-06, "loss": 1.0059, "step": 2161 }, { "epoch": 0.26, "learning_rate": 4.959264527217322e-06, "loss": 1.0933, "step": 2162 }, { "epoch": 0.26, "learning_rate": 4.9592067195724446e-06, "loss": 0.72, "step": 2163 }, { "epoch": 0.26, "learning_rate": 4.959148871276674e-06, "loss": 0.658, "step": 2164 }, { "epoch": 0.26, "learning_rate": 4.959090982330966e-06, "loss": 0.7431, "step": 2165 }, { "epoch": 0.26, "learning_rate": 4.959033052736277e-06, "loss": 0.8209, "step": 2166 }, { "epoch": 0.26, "learning_rate": 4.958975082493566e-06, "loss": 1.0637, "step": 2167 }, { "epoch": 0.26, "learning_rate": 4.958917071603791e-06, "loss": 1.1249, "step": 2168 }, { "epoch": 0.26, "learning_rate": 4.95885902006791e-06, "loss": 0.7982, "step": 2169 }, { "epoch": 0.26, "learning_rate": 4.958800927886883e-06, "loss": 0.6309, "step": 2170 }, { "epoch": 0.26, "learning_rate": 4.958742795061671e-06, "loss": 0.785, "step": 2171 }, { "epoch": 0.26, "learning_rate": 4.958684621593233e-06, "loss": 0.8201, "step": 2172 }, { "epoch": 0.26, "learning_rate": 4.958626407482533e-06, "loss": 1.065, "step": 2173 }, { "epoch": 0.26, "learning_rate": 4.958568152730532e-06, "loss": 1.1644, "step": 2174 }, { "epoch": 0.26, "learning_rate": 4.958509857338192e-06, "loss": 1.2242, "step": 2175 }, { "epoch": 0.26, "learning_rate": 4.958451521306478e-06, "loss": 0.7219, "step": 2176 }, { "epoch": 0.26, "learning_rate": 4.958393144636356e-06, "loss": 0.6503, "step": 2177 }, { "epoch": 0.26, "learning_rate": 4.958334727328788e-06, "loss": 0.8098, "step": 2178 }, { "epoch": 0.26, "learning_rate": 4.95827626938474e-06, "loss": 0.7474, "step": 2179 }, { "epoch": 0.26, "learning_rate": 4.9582177708051795e-06, "loss": 1.1242, "step": 2180 }, { "epoch": 0.26, "learning_rate": 4.958159231591073e-06, "loss": 1.0491, "step": 2181 }, { "epoch": 0.26, "learning_rate": 4.958100651743388e-06, "loss": 0.6186, "step": 2182 }, { "epoch": 0.26, "learning_rate": 4.958042031263093e-06, "loss": 0.6214, "step": 2183 }, { "epoch": 0.26, "learning_rate": 4.9579833701511575e-06, "loss": 0.7088, "step": 2184 }, { "epoch": 0.26, "learning_rate": 4.95792466840855e-06, "loss": 0.9839, "step": 2185 }, { "epoch": 0.26, "learning_rate": 4.957865926036242e-06, "loss": 1.1071, "step": 2186 }, { "epoch": 0.26, "learning_rate": 4.957807143035204e-06, "loss": 1.3241, "step": 2187 }, { "epoch": 0.26, "learning_rate": 4.957748319406407e-06, "loss": 0.5757, "step": 2188 }, { "epoch": 0.26, "learning_rate": 4.957689455150825e-06, "loss": 0.6738, "step": 2189 }, { "epoch": 0.26, "learning_rate": 4.957630550269429e-06, "loss": 0.7637, "step": 2190 }, { "epoch": 0.26, "learning_rate": 4.957571604763195e-06, "loss": 0.7378, "step": 2191 }, { "epoch": 0.26, "learning_rate": 4.9575126186330955e-06, "loss": 0.8703, "step": 2192 }, { "epoch": 0.26, "learning_rate": 4.957453591880106e-06, "loss": 1.2171, "step": 2193 }, { "epoch": 0.26, "learning_rate": 4.957394524505202e-06, "loss": 0.6628, "step": 2194 }, { "epoch": 0.26, "learning_rate": 4.957335416509362e-06, "loss": 0.575, "step": 2195 }, { "epoch": 0.26, "learning_rate": 4.957276267893561e-06, "loss": 0.6096, "step": 2196 }, { "epoch": 0.26, "learning_rate": 4.957217078658776e-06, "loss": 0.7608, "step": 2197 }, { "epoch": 0.26, "learning_rate": 4.957157848805987e-06, "loss": 1.0412, "step": 2198 }, { "epoch": 0.26, "learning_rate": 4.957098578336172e-06, "loss": 1.114, "step": 2199 }, { "epoch": 0.26, "learning_rate": 4.957039267250312e-06, "loss": 1.1648, "step": 2200 }, { "epoch": 0.26, "learning_rate": 4.956979915549387e-06, "loss": 0.6677, "step": 2201 }, { "epoch": 0.26, "learning_rate": 4.956920523234377e-06, "loss": 0.6707, "step": 2202 }, { "epoch": 0.26, "learning_rate": 4.956861090306265e-06, "loss": 0.7937, "step": 2203 }, { "epoch": 0.26, "learning_rate": 4.956801616766033e-06, "loss": 0.9758, "step": 2204 }, { "epoch": 0.26, "learning_rate": 4.956742102614665e-06, "loss": 1.0959, "step": 2205 }, { "epoch": 0.26, "learning_rate": 4.956682547853143e-06, "loss": 1.2205, "step": 2206 }, { "epoch": 0.26, "learning_rate": 4.9566229524824525e-06, "loss": 0.631, "step": 2207 }, { "epoch": 0.26, "learning_rate": 4.9565633165035785e-06, "loss": 0.6415, "step": 2208 }, { "epoch": 0.26, "learning_rate": 4.956503639917507e-06, "loss": 0.7881, "step": 2209 }, { "epoch": 0.26, "learning_rate": 4.956443922725224e-06, "loss": 0.8502, "step": 2210 }, { "epoch": 0.26, "learning_rate": 4.9563841649277165e-06, "loss": 1.0037, "step": 2211 }, { "epoch": 0.26, "learning_rate": 4.956324366525973e-06, "loss": 1.1251, "step": 2212 }, { "epoch": 0.26, "learning_rate": 4.956264527520982e-06, "loss": 0.8553, "step": 2213 }, { "epoch": 0.26, "learning_rate": 4.956204647913731e-06, "loss": 0.6772, "step": 2214 }, { "epoch": 0.26, "learning_rate": 4.9561447277052125e-06, "loss": 0.8212, "step": 2215 }, { "epoch": 0.26, "learning_rate": 4.956084766896415e-06, "loss": 0.8157, "step": 2216 }, { "epoch": 0.26, "learning_rate": 4.95602476548833e-06, "loss": 0.9643, "step": 2217 }, { "epoch": 0.26, "learning_rate": 4.9559647234819496e-06, "loss": 1.293, "step": 2218 }, { "epoch": 0.26, "learning_rate": 4.955904640878267e-06, "loss": 0.6765, "step": 2219 }, { "epoch": 0.26, "learning_rate": 4.955844517678274e-06, "loss": 0.6916, "step": 2220 }, { "epoch": 0.26, "learning_rate": 4.9557843538829644e-06, "loss": 0.718, "step": 2221 }, { "epoch": 0.26, "learning_rate": 4.955724149493334e-06, "loss": 0.8783, "step": 2222 }, { "epoch": 0.26, "learning_rate": 4.955663904510378e-06, "loss": 0.9524, "step": 2223 }, { "epoch": 0.26, "learning_rate": 4.955603618935091e-06, "loss": 1.0405, "step": 2224 }, { "epoch": 0.26, "learning_rate": 4.95554329276847e-06, "loss": 1.2506, "step": 2225 }, { "epoch": 0.27, "learning_rate": 4.955482926011512e-06, "loss": 0.6225, "step": 2226 }, { "epoch": 0.27, "learning_rate": 4.955422518665216e-06, "loss": 0.645, "step": 2227 }, { "epoch": 0.27, "learning_rate": 4.955362070730579e-06, "loss": 0.9911, "step": 2228 }, { "epoch": 0.27, "learning_rate": 4.955301582208602e-06, "loss": 1.0331, "step": 2229 }, { "epoch": 0.27, "learning_rate": 4.955241053100284e-06, "loss": 1.1055, "step": 2230 }, { "epoch": 0.27, "learning_rate": 4.955180483406624e-06, "loss": 1.2327, "step": 2231 }, { "epoch": 0.27, "learning_rate": 4.955119873128625e-06, "loss": 0.6156, "step": 2232 }, { "epoch": 0.27, "learning_rate": 4.955059222267289e-06, "loss": 0.7053, "step": 2233 }, { "epoch": 0.27, "learning_rate": 4.954998530823618e-06, "loss": 0.7447, "step": 2234 }, { "epoch": 0.27, "learning_rate": 4.954937798798615e-06, "loss": 0.905, "step": 2235 }, { "epoch": 0.27, "learning_rate": 4.954877026193284e-06, "loss": 1.0745, "step": 2236 }, { "epoch": 0.27, "learning_rate": 4.954816213008629e-06, "loss": 1.1779, "step": 2237 }, { "epoch": 0.27, "learning_rate": 4.954755359245657e-06, "loss": 0.6182, "step": 2238 }, { "epoch": 0.27, "learning_rate": 4.9546944649053745e-06, "loss": 0.6354, "step": 2239 }, { "epoch": 0.27, "learning_rate": 4.954633529988785e-06, "loss": 0.8158, "step": 2240 }, { "epoch": 0.27, "learning_rate": 4.954572554496897e-06, "loss": 0.8233, "step": 2241 }, { "epoch": 0.27, "learning_rate": 4.954511538430719e-06, "loss": 0.8757, "step": 2242 }, { "epoch": 0.27, "learning_rate": 4.954450481791259e-06, "loss": 1.2485, "step": 2243 }, { "epoch": 0.27, "learning_rate": 4.954389384579527e-06, "loss": 0.7748, "step": 2244 }, { "epoch": 0.27, "learning_rate": 4.954328246796532e-06, "loss": 0.6135, "step": 2245 }, { "epoch": 0.27, "learning_rate": 4.954267068443286e-06, "loss": 0.7397, "step": 2246 }, { "epoch": 0.27, "learning_rate": 4.954205849520799e-06, "loss": 0.8096, "step": 2247 }, { "epoch": 0.27, "learning_rate": 4.954144590030083e-06, "loss": 0.7984, "step": 2248 }, { "epoch": 0.27, "learning_rate": 4.954083289972152e-06, "loss": 1.1638, "step": 2249 }, { "epoch": 0.27, "learning_rate": 4.954021949348018e-06, "loss": 1.2409, "step": 2250 }, { "epoch": 0.27, "learning_rate": 4.953960568158695e-06, "loss": 0.4909, "step": 2251 }, { "epoch": 0.27, "learning_rate": 4.953899146405198e-06, "loss": 0.7408, "step": 2252 }, { "epoch": 0.27, "learning_rate": 4.953837684088542e-06, "loss": 0.8362, "step": 2253 }, { "epoch": 0.27, "learning_rate": 4.953776181209743e-06, "loss": 0.7921, "step": 2254 }, { "epoch": 0.27, "learning_rate": 4.9537146377698184e-06, "loss": 1.1614, "step": 2255 }, { "epoch": 0.27, "learning_rate": 4.953653053769785e-06, "loss": 1.1667, "step": 2256 }, { "epoch": 0.27, "learning_rate": 4.95359142921066e-06, "loss": 0.6945, "step": 2257 }, { "epoch": 0.27, "learning_rate": 4.953529764093463e-06, "loss": 0.6257, "step": 2258 }, { "epoch": 0.27, "learning_rate": 4.953468058419214e-06, "loss": 0.7283, "step": 2259 }, { "epoch": 0.27, "learning_rate": 4.953406312188931e-06, "loss": 0.8154, "step": 2260 }, { "epoch": 0.27, "learning_rate": 4.953344525403636e-06, "loss": 0.8915, "step": 2261 }, { "epoch": 0.27, "learning_rate": 4.953282698064351e-06, "loss": 1.1277, "step": 2262 }, { "epoch": 0.27, "learning_rate": 4.953220830172097e-06, "loss": 0.8164, "step": 2263 }, { "epoch": 0.27, "learning_rate": 4.953158921727896e-06, "loss": 0.6198, "step": 2264 }, { "epoch": 0.27, "learning_rate": 4.953096972732773e-06, "loss": 0.8116, "step": 2265 }, { "epoch": 0.27, "learning_rate": 4.95303498318775e-06, "loss": 0.9011, "step": 2266 }, { "epoch": 0.27, "learning_rate": 4.952972953093854e-06, "loss": 0.9302, "step": 2267 }, { "epoch": 0.27, "learning_rate": 4.952910882452109e-06, "loss": 1.0752, "step": 2268 }, { "epoch": 0.27, "learning_rate": 4.952848771263541e-06, "loss": 0.8142, "step": 2269 }, { "epoch": 0.27, "learning_rate": 4.952786619529177e-06, "loss": 0.5333, "step": 2270 }, { "epoch": 0.27, "learning_rate": 4.952724427250046e-06, "loss": 0.7292, "step": 2271 }, { "epoch": 0.27, "learning_rate": 4.952662194427172e-06, "loss": 0.9572, "step": 2272 }, { "epoch": 0.27, "learning_rate": 4.952599921061587e-06, "loss": 0.9664, "step": 2273 }, { "epoch": 0.27, "learning_rate": 4.95253760715432e-06, "loss": 1.2103, "step": 2274 }, { "epoch": 0.27, "learning_rate": 4.9524752527064e-06, "loss": 1.2708, "step": 2275 }, { "epoch": 0.27, "learning_rate": 4.952412857718859e-06, "loss": 0.5752, "step": 2276 }, { "epoch": 0.27, "learning_rate": 4.952350422192727e-06, "loss": 0.7552, "step": 2277 }, { "epoch": 0.27, "learning_rate": 4.952287946129037e-06, "loss": 0.8125, "step": 2278 }, { "epoch": 0.27, "learning_rate": 4.952225429528822e-06, "loss": 0.8258, "step": 2279 }, { "epoch": 0.27, "learning_rate": 4.9521628723931145e-06, "loss": 1.1151, "step": 2280 }, { "epoch": 0.27, "learning_rate": 4.952100274722949e-06, "loss": 1.265, "step": 2281 }, { "epoch": 0.27, "learning_rate": 4.95203763651936e-06, "loss": 0.7113, "step": 2282 }, { "epoch": 0.27, "learning_rate": 4.951974957783383e-06, "loss": 0.6252, "step": 2283 }, { "epoch": 0.27, "learning_rate": 4.9519122385160545e-06, "loss": 0.6935, "step": 2284 }, { "epoch": 0.27, "learning_rate": 4.951849478718411e-06, "loss": 0.798, "step": 2285 }, { "epoch": 0.27, "learning_rate": 4.951786678391489e-06, "loss": 0.9919, "step": 2286 }, { "epoch": 0.27, "learning_rate": 4.951723837536329e-06, "loss": 1.2069, "step": 2287 }, { "epoch": 0.27, "learning_rate": 4.9516609561539675e-06, "loss": 0.6319, "step": 2288 }, { "epoch": 0.27, "learning_rate": 4.951598034245444e-06, "loss": 0.7158, "step": 2289 }, { "epoch": 0.27, "learning_rate": 4.9515350718118e-06, "loss": 0.6903, "step": 2290 }, { "epoch": 0.27, "learning_rate": 4.951472068854076e-06, "loss": 0.9225, "step": 2291 }, { "epoch": 0.27, "learning_rate": 4.9514090253733126e-06, "loss": 1.0721, "step": 2292 }, { "epoch": 0.27, "learning_rate": 4.951345941370552e-06, "loss": 1.14, "step": 2293 }, { "epoch": 0.27, "learning_rate": 4.951282816846837e-06, "loss": 1.0873, "step": 2294 }, { "epoch": 0.27, "learning_rate": 4.951219651803212e-06, "loss": 0.669, "step": 2295 }, { "epoch": 0.27, "learning_rate": 4.951156446240722e-06, "loss": 0.7612, "step": 2296 }, { "epoch": 0.27, "learning_rate": 4.9510932001604075e-06, "loss": 0.8745, "step": 2297 }, { "epoch": 0.27, "learning_rate": 4.951029913563318e-06, "loss": 0.8677, "step": 2298 }, { "epoch": 0.27, "learning_rate": 4.950966586450499e-06, "loss": 1.1971, "step": 2299 }, { "epoch": 0.27, "learning_rate": 4.950903218822996e-06, "loss": 1.163, "step": 2300 }, { "epoch": 0.27, "learning_rate": 4.950839810681857e-06, "loss": 0.5777, "step": 2301 }, { "epoch": 0.27, "learning_rate": 4.9507763620281305e-06, "loss": 0.7592, "step": 2302 }, { "epoch": 0.27, "learning_rate": 4.950712872862864e-06, "loss": 0.7556, "step": 2303 }, { "epoch": 0.27, "learning_rate": 4.950649343187109e-06, "loss": 0.8337, "step": 2304 }, { "epoch": 0.27, "learning_rate": 4.950585773001916e-06, "loss": 1.0772, "step": 2305 }, { "epoch": 0.27, "learning_rate": 4.950522162308332e-06, "loss": 1.0429, "step": 2306 }, { "epoch": 0.27, "learning_rate": 4.950458511107413e-06, "loss": 0.7031, "step": 2307 }, { "epoch": 0.27, "learning_rate": 4.950394819400209e-06, "loss": 0.5497, "step": 2308 }, { "epoch": 0.27, "learning_rate": 4.950331087187772e-06, "loss": 0.7216, "step": 2309 }, { "epoch": 0.28, "learning_rate": 4.9502673144711575e-06, "loss": 0.9553, "step": 2310 }, { "epoch": 0.28, "learning_rate": 4.950203501251418e-06, "loss": 0.9769, "step": 2311 }, { "epoch": 0.28, "learning_rate": 4.95013964752961e-06, "loss": 0.9912, "step": 2312 }, { "epoch": 0.28, "learning_rate": 4.950075753306787e-06, "loss": 0.6229, "step": 2313 }, { "epoch": 0.28, "learning_rate": 4.950011818584006e-06, "loss": 0.6168, "step": 2314 }, { "epoch": 0.28, "learning_rate": 4.949947843362325e-06, "loss": 0.8111, "step": 2315 }, { "epoch": 0.28, "learning_rate": 4.949883827642801e-06, "loss": 0.8073, "step": 2316 }, { "epoch": 0.28, "learning_rate": 4.949819771426491e-06, "loss": 1.007, "step": 2317 }, { "epoch": 0.28, "learning_rate": 4.949755674714454e-06, "loss": 1.2477, "step": 2318 }, { "epoch": 0.28, "learning_rate": 4.949691537507752e-06, "loss": 1.0384, "step": 2319 }, { "epoch": 0.28, "learning_rate": 4.949627359807442e-06, "loss": 0.7155, "step": 2320 }, { "epoch": 0.28, "learning_rate": 4.949563141614586e-06, "loss": 0.7785, "step": 2321 }, { "epoch": 0.28, "learning_rate": 4.949498882930246e-06, "loss": 0.8768, "step": 2322 }, { "epoch": 0.28, "learning_rate": 4.949434583755483e-06, "loss": 1.0007, "step": 2323 }, { "epoch": 0.28, "learning_rate": 4.949370244091362e-06, "loss": 1.0754, "step": 2324 }, { "epoch": 0.28, "learning_rate": 4.949305863938946e-06, "loss": 1.2821, "step": 2325 }, { "epoch": 0.28, "learning_rate": 4.949241443299297e-06, "loss": 0.5118, "step": 2326 }, { "epoch": 0.28, "learning_rate": 4.949176982173482e-06, "loss": 0.8151, "step": 2327 }, { "epoch": 0.28, "learning_rate": 4.949112480562566e-06, "loss": 0.8133, "step": 2328 }, { "epoch": 0.28, "learning_rate": 4.949047938467616e-06, "loss": 0.9539, "step": 2329 }, { "epoch": 0.28, "learning_rate": 4.948983355889697e-06, "loss": 1.0492, "step": 2330 }, { "epoch": 0.28, "learning_rate": 4.948918732829879e-06, "loss": 1.1148, "step": 2331 }, { "epoch": 0.28, "learning_rate": 4.948854069289228e-06, "loss": 0.6313, "step": 2332 }, { "epoch": 0.28, "learning_rate": 4.948789365268814e-06, "loss": 0.7426, "step": 2333 }, { "epoch": 0.28, "learning_rate": 4.948724620769706e-06, "loss": 0.7899, "step": 2334 }, { "epoch": 0.28, "learning_rate": 4.9486598357929734e-06, "loss": 0.9056, "step": 2335 }, { "epoch": 0.28, "learning_rate": 4.9485950103396905e-06, "loss": 1.0602, "step": 2336 }, { "epoch": 0.28, "learning_rate": 4.9485301444109255e-06, "loss": 1.2278, "step": 2337 }, { "epoch": 0.28, "learning_rate": 4.948465238007752e-06, "loss": 0.5556, "step": 2338 }, { "epoch": 0.28, "learning_rate": 4.948400291131242e-06, "loss": 0.797, "step": 2339 }, { "epoch": 0.28, "learning_rate": 4.94833530378247e-06, "loss": 0.7437, "step": 2340 }, { "epoch": 0.28, "learning_rate": 4.948270275962511e-06, "loss": 0.7968, "step": 2341 }, { "epoch": 0.28, "learning_rate": 4.948205207672438e-06, "loss": 1.0074, "step": 2342 }, { "epoch": 0.28, "learning_rate": 4.948140098913327e-06, "loss": 1.0725, "step": 2343 }, { "epoch": 0.28, "learning_rate": 4.948074949686256e-06, "loss": 0.7872, "step": 2344 }, { "epoch": 0.28, "learning_rate": 4.9480097599923e-06, "loss": 0.7364, "step": 2345 }, { "epoch": 0.28, "learning_rate": 4.947944529832537e-06, "loss": 0.6349, "step": 2346 }, { "epoch": 0.28, "learning_rate": 4.947879259208046e-06, "loss": 0.892, "step": 2347 }, { "epoch": 0.28, "learning_rate": 4.947813948119906e-06, "loss": 0.9302, "step": 2348 }, { "epoch": 0.28, "learning_rate": 4.947748596569195e-06, "loss": 0.9566, "step": 2349 }, { "epoch": 0.28, "learning_rate": 4.947683204556994e-06, "loss": 1.2382, "step": 2350 }, { "epoch": 0.28, "learning_rate": 4.947617772084385e-06, "loss": 0.6026, "step": 2351 }, { "epoch": 0.28, "learning_rate": 4.9475522991524496e-06, "loss": 0.6855, "step": 2352 }, { "epoch": 0.28, "learning_rate": 4.947486785762268e-06, "loss": 0.6583, "step": 2353 }, { "epoch": 0.28, "learning_rate": 4.947421231914925e-06, "loss": 0.9089, "step": 2354 }, { "epoch": 0.28, "learning_rate": 4.947355637611504e-06, "loss": 1.1046, "step": 2355 }, { "epoch": 0.28, "learning_rate": 4.947290002853088e-06, "loss": 1.3025, "step": 2356 }, { "epoch": 0.28, "learning_rate": 4.947224327640764e-06, "loss": 0.7637, "step": 2357 }, { "epoch": 0.28, "learning_rate": 4.947158611975616e-06, "loss": 0.7787, "step": 2358 }, { "epoch": 0.28, "learning_rate": 4.947092855858731e-06, "loss": 0.8105, "step": 2359 }, { "epoch": 0.28, "learning_rate": 4.947027059291195e-06, "loss": 0.8831, "step": 2360 }, { "epoch": 0.28, "learning_rate": 4.946961222274098e-06, "loss": 1.1347, "step": 2361 }, { "epoch": 0.28, "learning_rate": 4.946895344808525e-06, "loss": 1.1132, "step": 2362 }, { "epoch": 0.28, "learning_rate": 4.9468294268955675e-06, "loss": 0.788, "step": 2363 }, { "epoch": 0.28, "learning_rate": 4.946763468536314e-06, "loss": 0.6767, "step": 2364 }, { "epoch": 0.28, "learning_rate": 4.9466974697318556e-06, "loss": 0.7383, "step": 2365 }, { "epoch": 0.28, "learning_rate": 4.9466314304832815e-06, "loss": 0.85, "step": 2366 }, { "epoch": 0.28, "learning_rate": 4.946565350791685e-06, "loss": 1.118, "step": 2367 }, { "epoch": 0.28, "learning_rate": 4.94649923065816e-06, "loss": 1.2773, "step": 2368 }, { "epoch": 0.28, "learning_rate": 4.946433070083795e-06, "loss": 0.5953, "step": 2369 }, { "epoch": 0.28, "learning_rate": 4.946366869069687e-06, "loss": 0.699, "step": 2370 }, { "epoch": 0.28, "learning_rate": 4.946300627616929e-06, "loss": 0.7559, "step": 2371 }, { "epoch": 0.28, "learning_rate": 4.946234345726617e-06, "loss": 0.7203, "step": 2372 }, { "epoch": 0.28, "learning_rate": 4.946168023399845e-06, "loss": 0.958, "step": 2373 }, { "epoch": 0.28, "learning_rate": 4.946101660637711e-06, "loss": 1.1291, "step": 2374 }, { "epoch": 0.28, "learning_rate": 4.946035257441312e-06, "loss": 1.2896, "step": 2375 }, { "epoch": 0.28, "learning_rate": 4.945968813811743e-06, "loss": 0.7094, "step": 2376 }, { "epoch": 0.28, "learning_rate": 4.945902329750106e-06, "loss": 0.6611, "step": 2377 }, { "epoch": 0.28, "learning_rate": 4.9458358052574975e-06, "loss": 0.6906, "step": 2378 }, { "epoch": 0.28, "learning_rate": 4.945769240335019e-06, "loss": 0.8604, "step": 2379 }, { "epoch": 0.28, "learning_rate": 4.945702634983769e-06, "loss": 0.8971, "step": 2380 }, { "epoch": 0.28, "learning_rate": 4.945635989204849e-06, "loss": 1.2343, "step": 2381 }, { "epoch": 0.28, "learning_rate": 4.94556930299936e-06, "loss": 0.6699, "step": 2382 }, { "epoch": 0.28, "learning_rate": 4.945502576368406e-06, "loss": 0.6738, "step": 2383 }, { "epoch": 0.28, "learning_rate": 4.94543580931309e-06, "loss": 0.7643, "step": 2384 }, { "epoch": 0.28, "learning_rate": 4.9453690018345144e-06, "loss": 0.959, "step": 2385 }, { "epoch": 0.28, "learning_rate": 4.945302153933784e-06, "loss": 1.0908, "step": 2386 }, { "epoch": 0.28, "learning_rate": 4.945235265612004e-06, "loss": 1.1307, "step": 2387 }, { "epoch": 0.28, "learning_rate": 4.94516833687028e-06, "loss": 0.7426, "step": 2388 }, { "epoch": 0.28, "learning_rate": 4.945101367709717e-06, "loss": 0.6943, "step": 2389 }, { "epoch": 0.28, "learning_rate": 4.945034358131424e-06, "loss": 0.8094, "step": 2390 }, { "epoch": 0.28, "learning_rate": 4.944967308136508e-06, "loss": 0.8947, "step": 2391 }, { "epoch": 0.28, "learning_rate": 4.944900217726078e-06, "loss": 0.946, "step": 2392 }, { "epoch": 0.28, "learning_rate": 4.944833086901241e-06, "loss": 1.2033, "step": 2393 }, { "epoch": 0.29, "learning_rate": 4.944765915663109e-06, "loss": 0.909, "step": 2394 }, { "epoch": 0.29, "learning_rate": 4.94469870401279e-06, "loss": 0.8103, "step": 2395 }, { "epoch": 0.29, "learning_rate": 4.9446314519513975e-06, "loss": 0.7951, "step": 2396 }, { "epoch": 0.29, "learning_rate": 4.944564159480042e-06, "loss": 0.8017, "step": 2397 }, { "epoch": 0.29, "learning_rate": 4.944496826599835e-06, "loss": 0.919, "step": 2398 }, { "epoch": 0.29, "learning_rate": 4.944429453311891e-06, "loss": 1.0285, "step": 2399 }, { "epoch": 0.29, "learning_rate": 4.944362039617322e-06, "loss": 1.2297, "step": 2400 }, { "epoch": 0.29, "learning_rate": 4.9442945855172445e-06, "loss": 0.5107, "step": 2401 }, { "epoch": 0.29, "learning_rate": 4.944227091012772e-06, "loss": 0.7814, "step": 2402 }, { "epoch": 0.29, "learning_rate": 4.9441595561050205e-06, "loss": 0.6932, "step": 2403 }, { "epoch": 0.29, "learning_rate": 4.944091980795106e-06, "loss": 0.8699, "step": 2404 }, { "epoch": 0.29, "learning_rate": 4.9440243650841466e-06, "loss": 1.0629, "step": 2405 }, { "epoch": 0.29, "learning_rate": 4.94395670897326e-06, "loss": 1.1212, "step": 2406 }, { "epoch": 0.29, "learning_rate": 4.943889012463563e-06, "loss": 0.6418, "step": 2407 }, { "epoch": 0.29, "learning_rate": 4.943821275556176e-06, "loss": 0.7061, "step": 2408 }, { "epoch": 0.29, "learning_rate": 4.943753498252218e-06, "loss": 0.7771, "step": 2409 }, { "epoch": 0.29, "learning_rate": 4.94368568055281e-06, "loss": 0.9138, "step": 2410 }, { "epoch": 0.29, "learning_rate": 4.9436178224590716e-06, "loss": 0.9938, "step": 2411 }, { "epoch": 0.29, "learning_rate": 4.943549923972126e-06, "loss": 1.2011, "step": 2412 }, { "epoch": 0.29, "learning_rate": 4.943481985093096e-06, "loss": 0.7185, "step": 2413 }, { "epoch": 0.29, "learning_rate": 4.943414005823103e-06, "loss": 0.6139, "step": 2414 }, { "epoch": 0.29, "learning_rate": 4.943345986163272e-06, "loss": 0.8826, "step": 2415 }, { "epoch": 0.29, "learning_rate": 4.943277926114727e-06, "loss": 0.9424, "step": 2416 }, { "epoch": 0.29, "learning_rate": 4.943209825678592e-06, "loss": 1.0247, "step": 2417 }, { "epoch": 0.29, "learning_rate": 4.943141684855994e-06, "loss": 1.0726, "step": 2418 }, { "epoch": 0.29, "learning_rate": 4.943073503648058e-06, "loss": 0.7733, "step": 2419 }, { "epoch": 0.29, "learning_rate": 4.943005282055914e-06, "loss": 0.5775, "step": 2420 }, { "epoch": 0.29, "learning_rate": 4.942937020080686e-06, "loss": 0.664, "step": 2421 }, { "epoch": 0.29, "learning_rate": 4.942868717723505e-06, "loss": 0.806, "step": 2422 }, { "epoch": 0.29, "learning_rate": 4.942800374985499e-06, "loss": 0.9738, "step": 2423 }, { "epoch": 0.29, "learning_rate": 4.942731991867798e-06, "loss": 1.1346, "step": 2424 }, { "epoch": 0.29, "learning_rate": 4.942663568371532e-06, "loss": 1.2616, "step": 2425 }, { "epoch": 0.29, "learning_rate": 4.942595104497832e-06, "loss": 0.4618, "step": 2426 }, { "epoch": 0.29, "learning_rate": 4.94252660024783e-06, "loss": 0.6714, "step": 2427 }, { "epoch": 0.29, "learning_rate": 4.942458055622658e-06, "loss": 0.8904, "step": 2428 }, { "epoch": 0.29, "learning_rate": 4.94238947062345e-06, "loss": 0.828, "step": 2429 }, { "epoch": 0.29, "learning_rate": 4.942320845251339e-06, "loss": 1.0413, "step": 2430 }, { "epoch": 0.29, "learning_rate": 4.9422521795074595e-06, "loss": 1.3041, "step": 2431 }, { "epoch": 0.29, "learning_rate": 4.942183473392947e-06, "loss": 0.7207, "step": 2432 }, { "epoch": 0.29, "learning_rate": 4.942114726908936e-06, "loss": 0.6964, "step": 2433 }, { "epoch": 0.29, "learning_rate": 4.942045940056564e-06, "loss": 0.7691, "step": 2434 }, { "epoch": 0.29, "learning_rate": 4.941977112836967e-06, "loss": 0.8795, "step": 2435 }, { "epoch": 0.29, "learning_rate": 4.941908245251284e-06, "loss": 1.0418, "step": 2436 }, { "epoch": 0.29, "learning_rate": 4.941839337300653e-06, "loss": 1.1679, "step": 2437 }, { "epoch": 0.29, "learning_rate": 4.941770388986213e-06, "loss": 0.5328, "step": 2438 }, { "epoch": 0.29, "learning_rate": 4.941701400309103e-06, "loss": 0.6343, "step": 2439 }, { "epoch": 0.29, "learning_rate": 4.9416323712704644e-06, "loss": 0.7919, "step": 2440 }, { "epoch": 0.29, "learning_rate": 4.9415633018714385e-06, "loss": 0.6914, "step": 2441 }, { "epoch": 0.29, "learning_rate": 4.941494192113165e-06, "loss": 0.9901, "step": 2442 }, { "epoch": 0.29, "learning_rate": 4.941425041996788e-06, "loss": 1.1091, "step": 2443 }, { "epoch": 0.29, "learning_rate": 4.94135585152345e-06, "loss": 0.9509, "step": 2444 }, { "epoch": 0.29, "learning_rate": 4.941286620694296e-06, "loss": 0.5852, "step": 2445 }, { "epoch": 0.29, "learning_rate": 4.941217349510469e-06, "loss": 0.7133, "step": 2446 }, { "epoch": 0.29, "learning_rate": 4.9411480379731135e-06, "loss": 0.7481, "step": 2447 }, { "epoch": 0.29, "learning_rate": 4.941078686083377e-06, "loss": 0.9186, "step": 2448 }, { "epoch": 0.29, "learning_rate": 4.941009293842405e-06, "loss": 1.0874, "step": 2449 }, { "epoch": 0.29, "learning_rate": 4.9409398612513435e-06, "loss": 1.3225, "step": 2450 }, { "epoch": 0.29, "learning_rate": 4.940870388311342e-06, "loss": 0.6135, "step": 2451 }, { "epoch": 0.29, "learning_rate": 4.9408008750235485e-06, "loss": 0.7871, "step": 2452 }, { "epoch": 0.29, "learning_rate": 4.940731321389112e-06, "loss": 0.6379, "step": 2453 }, { "epoch": 0.29, "learning_rate": 4.94066172740918e-06, "loss": 0.8857, "step": 2454 }, { "epoch": 0.29, "learning_rate": 4.940592093084907e-06, "loss": 1.08, "step": 2455 }, { "epoch": 0.29, "learning_rate": 4.940522418417442e-06, "loss": 1.2114, "step": 2456 }, { "epoch": 0.29, "learning_rate": 4.940452703407935e-06, "loss": 0.5777, "step": 2457 }, { "epoch": 0.29, "learning_rate": 4.94038294805754e-06, "loss": 0.8636, "step": 2458 }, { "epoch": 0.29, "learning_rate": 4.940313152367412e-06, "loss": 0.699, "step": 2459 }, { "epoch": 0.29, "learning_rate": 4.940243316338701e-06, "loss": 0.8509, "step": 2460 }, { "epoch": 0.29, "learning_rate": 4.9401734399725645e-06, "loss": 1.1357, "step": 2461 }, { "epoch": 0.29, "learning_rate": 4.940103523270156e-06, "loss": 1.1702, "step": 2462 }, { "epoch": 0.29, "learning_rate": 4.940033566232632e-06, "loss": 0.5967, "step": 2463 }, { "epoch": 0.29, "learning_rate": 4.939963568861147e-06, "loss": 0.744, "step": 2464 }, { "epoch": 0.29, "learning_rate": 4.939893531156861e-06, "loss": 0.635, "step": 2465 }, { "epoch": 0.29, "learning_rate": 4.939823453120929e-06, "loss": 0.8002, "step": 2466 }, { "epoch": 0.29, "learning_rate": 4.939753334754511e-06, "loss": 0.8386, "step": 2467 }, { "epoch": 0.29, "learning_rate": 4.939683176058765e-06, "loss": 1.2125, "step": 2468 }, { "epoch": 0.29, "learning_rate": 4.939612977034852e-06, "loss": 0.7208, "step": 2469 }, { "epoch": 0.29, "learning_rate": 4.939542737683932e-06, "loss": 0.7047, "step": 2470 }, { "epoch": 0.29, "learning_rate": 4.9394724580071665e-06, "loss": 0.6531, "step": 2471 }, { "epoch": 0.29, "learning_rate": 4.939402138005715e-06, "loss": 0.7523, "step": 2472 }, { "epoch": 0.29, "learning_rate": 4.9393317776807424e-06, "loss": 0.9406, "step": 2473 }, { "epoch": 0.29, "learning_rate": 4.939261377033411e-06, "loss": 1.1565, "step": 2474 }, { "epoch": 0.29, "learning_rate": 4.939190936064884e-06, "loss": 1.0567, "step": 2475 }, { "epoch": 0.29, "learning_rate": 4.939120454776326e-06, "loss": 0.6211, "step": 2476 }, { "epoch": 0.29, "learning_rate": 4.939049933168902e-06, "loss": 0.6347, "step": 2477 }, { "epoch": 0.3, "learning_rate": 4.93897937124378e-06, "loss": 0.7608, "step": 2478 }, { "epoch": 0.3, "learning_rate": 4.938908769002122e-06, "loss": 0.9667, "step": 2479 }, { "epoch": 0.3, "learning_rate": 4.938838126445098e-06, "loss": 0.9754, "step": 2480 }, { "epoch": 0.3, "learning_rate": 4.938767443573876e-06, "loss": 1.2918, "step": 2481 }, { "epoch": 0.3, "learning_rate": 4.938696720389622e-06, "loss": 0.5102, "step": 2482 }, { "epoch": 0.3, "learning_rate": 4.938625956893507e-06, "loss": 0.5972, "step": 2483 }, { "epoch": 0.3, "learning_rate": 4.938555153086701e-06, "loss": 0.8243, "step": 2484 }, { "epoch": 0.3, "learning_rate": 4.938484308970374e-06, "loss": 0.855, "step": 2485 }, { "epoch": 0.3, "learning_rate": 4.938413424545696e-06, "loss": 1.0073, "step": 2486 }, { "epoch": 0.3, "learning_rate": 4.938342499813839e-06, "loss": 1.2134, "step": 2487 }, { "epoch": 0.3, "learning_rate": 4.938271534775977e-06, "loss": 0.5612, "step": 2488 }, { "epoch": 0.3, "learning_rate": 4.938200529433282e-06, "loss": 0.5313, "step": 2489 }, { "epoch": 0.3, "learning_rate": 4.938129483786927e-06, "loss": 0.7212, "step": 2490 }, { "epoch": 0.3, "learning_rate": 4.9380583978380865e-06, "loss": 0.7652, "step": 2491 }, { "epoch": 0.3, "learning_rate": 4.9379872715879375e-06, "loss": 0.904, "step": 2492 }, { "epoch": 0.3, "learning_rate": 4.937916105037654e-06, "loss": 1.0625, "step": 2493 }, { "epoch": 0.3, "learning_rate": 4.937844898188412e-06, "loss": 0.583, "step": 2494 }, { "epoch": 0.3, "learning_rate": 4.93777365104139e-06, "loss": 0.6802, "step": 2495 }, { "epoch": 0.3, "learning_rate": 4.937702363597765e-06, "loss": 0.843, "step": 2496 }, { "epoch": 0.3, "learning_rate": 4.937631035858715e-06, "loss": 0.9251, "step": 2497 }, { "epoch": 0.3, "learning_rate": 4.937559667825419e-06, "loss": 0.8905, "step": 2498 }, { "epoch": 0.3, "learning_rate": 4.937488259499058e-06, "loss": 1.1176, "step": 2499 }, { "epoch": 0.3, "learning_rate": 4.937416810880811e-06, "loss": 1.2133, "step": 2500 }, { "epoch": 0.3, "learning_rate": 4.937345321971861e-06, "loss": 0.6135, "step": 2501 }, { "epoch": 0.3, "learning_rate": 4.937273792773386e-06, "loss": 0.8956, "step": 2502 }, { "epoch": 0.3, "learning_rate": 4.9372022232865725e-06, "loss": 0.7874, "step": 2503 }, { "epoch": 0.3, "learning_rate": 4.9371306135126e-06, "loss": 0.8179, "step": 2504 }, { "epoch": 0.3, "learning_rate": 4.937058963452655e-06, "loss": 1.1727, "step": 2505 }, { "epoch": 0.3, "learning_rate": 4.9369872731079215e-06, "loss": 1.1213, "step": 2506 }, { "epoch": 0.3, "learning_rate": 4.936915542479583e-06, "loss": 0.5458, "step": 2507 }, { "epoch": 0.3, "learning_rate": 4.936843771568827e-06, "loss": 0.6364, "step": 2508 }, { "epoch": 0.3, "learning_rate": 4.936771960376838e-06, "loss": 0.7003, "step": 2509 }, { "epoch": 0.3, "learning_rate": 4.936700108904805e-06, "loss": 0.7672, "step": 2510 }, { "epoch": 0.3, "learning_rate": 4.936628217153914e-06, "loss": 0.9584, "step": 2511 }, { "epoch": 0.3, "learning_rate": 4.936556285125355e-06, "loss": 1.1068, "step": 2512 }, { "epoch": 0.3, "learning_rate": 4.936484312820315e-06, "loss": 0.7997, "step": 2513 }, { "epoch": 0.3, "learning_rate": 4.936412300239986e-06, "loss": 0.6079, "step": 2514 }, { "epoch": 0.3, "learning_rate": 4.936340247385556e-06, "loss": 0.7293, "step": 2515 }, { "epoch": 0.3, "learning_rate": 4.936268154258218e-06, "loss": 0.7729, "step": 2516 }, { "epoch": 0.3, "learning_rate": 4.936196020859163e-06, "loss": 0.9158, "step": 2517 }, { "epoch": 0.3, "learning_rate": 4.936123847189584e-06, "loss": 1.1523, "step": 2518 }, { "epoch": 0.3, "learning_rate": 4.936051633250671e-06, "loss": 0.6144, "step": 2519 }, { "epoch": 0.3, "learning_rate": 4.935979379043623e-06, "loss": 0.6465, "step": 2520 }, { "epoch": 0.3, "learning_rate": 4.93590708456963e-06, "loss": 0.7505, "step": 2521 }, { "epoch": 0.3, "learning_rate": 4.935834749829889e-06, "loss": 0.8691, "step": 2522 }, { "epoch": 0.3, "learning_rate": 4.935762374825595e-06, "loss": 0.9377, "step": 2523 }, { "epoch": 0.3, "learning_rate": 4.935689959557944e-06, "loss": 1.1897, "step": 2524 }, { "epoch": 0.3, "learning_rate": 4.935617504028134e-06, "loss": 1.2456, "step": 2525 }, { "epoch": 0.3, "learning_rate": 4.9355450082373626e-06, "loss": 0.5602, "step": 2526 }, { "epoch": 0.3, "learning_rate": 4.935472472186827e-06, "loss": 0.8318, "step": 2527 }, { "epoch": 0.3, "learning_rate": 4.935399895877728e-06, "loss": 0.8767, "step": 2528 }, { "epoch": 0.3, "learning_rate": 4.935327279311264e-06, "loss": 0.9247, "step": 2529 }, { "epoch": 0.3, "learning_rate": 4.9352546224886354e-06, "loss": 0.8984, "step": 2530 }, { "epoch": 0.3, "learning_rate": 4.935181925411044e-06, "loss": 0.9058, "step": 2531 }, { "epoch": 0.3, "learning_rate": 4.935109188079692e-06, "loss": 0.4464, "step": 2532 }, { "epoch": 0.3, "learning_rate": 4.935036410495779e-06, "loss": 0.7471, "step": 2533 }, { "epoch": 0.3, "learning_rate": 4.9349635926605115e-06, "loss": 0.6288, "step": 2534 }, { "epoch": 0.3, "learning_rate": 4.93489073457509e-06, "loss": 0.7536, "step": 2535 }, { "epoch": 0.3, "learning_rate": 4.934817836240721e-06, "loss": 1.0649, "step": 2536 }, { "epoch": 0.3, "learning_rate": 4.934744897658609e-06, "loss": 1.122, "step": 2537 }, { "epoch": 0.3, "learning_rate": 4.93467191882996e-06, "loss": 0.6296, "step": 2538 }, { "epoch": 0.3, "learning_rate": 4.934598899755978e-06, "loss": 0.7195, "step": 2539 }, { "epoch": 0.3, "learning_rate": 4.934525840437873e-06, "loss": 0.7722, "step": 2540 }, { "epoch": 0.3, "learning_rate": 4.934452740876853e-06, "loss": 0.8709, "step": 2541 }, { "epoch": 0.3, "learning_rate": 4.934379601074124e-06, "loss": 0.929, "step": 2542 }, { "epoch": 0.3, "learning_rate": 4.934306421030895e-06, "loss": 0.9797, "step": 2543 }, { "epoch": 0.3, "learning_rate": 4.934233200748378e-06, "loss": 0.7606, "step": 2544 }, { "epoch": 0.3, "learning_rate": 4.934159940227781e-06, "loss": 0.6418, "step": 2545 }, { "epoch": 0.3, "learning_rate": 4.934086639470316e-06, "loss": 0.8067, "step": 2546 }, { "epoch": 0.3, "learning_rate": 4.934013298477194e-06, "loss": 0.9301, "step": 2547 }, { "epoch": 0.3, "learning_rate": 4.933939917249629e-06, "loss": 0.891, "step": 2548 }, { "epoch": 0.3, "learning_rate": 4.933866495788833e-06, "loss": 1.079, "step": 2549 }, { "epoch": 0.3, "learning_rate": 4.93379303409602e-06, "loss": 1.225, "step": 2550 }, { "epoch": 0.3, "learning_rate": 4.933719532172402e-06, "loss": 0.7872, "step": 2551 }, { "epoch": 0.3, "learning_rate": 4.933645990019198e-06, "loss": 0.7383, "step": 2552 }, { "epoch": 0.3, "learning_rate": 4.93357240763762e-06, "loss": 0.9291, "step": 2553 }, { "epoch": 0.3, "learning_rate": 4.933498785028887e-06, "loss": 0.7888, "step": 2554 }, { "epoch": 0.3, "learning_rate": 4.933425122194214e-06, "loss": 1.0325, "step": 2555 }, { "epoch": 0.3, "learning_rate": 4.933351419134821e-06, "loss": 1.2681, "step": 2556 }, { "epoch": 0.3, "learning_rate": 4.933277675851923e-06, "loss": 0.7083, "step": 2557 }, { "epoch": 0.3, "learning_rate": 4.933203892346742e-06, "loss": 0.877, "step": 2558 }, { "epoch": 0.3, "learning_rate": 4.933130068620496e-06, "loss": 0.8132, "step": 2559 }, { "epoch": 0.3, "learning_rate": 4.933056204674406e-06, "loss": 0.9301, "step": 2560 }, { "epoch": 0.3, "learning_rate": 4.932982300509693e-06, "loss": 0.9936, "step": 2561 }, { "epoch": 0.31, "learning_rate": 4.9329083561275785e-06, "loss": 1.2234, "step": 2562 }, { "epoch": 0.31, "learning_rate": 4.9328343715292845e-06, "loss": 0.65, "step": 2563 }, { "epoch": 0.31, "learning_rate": 4.932760346716035e-06, "loss": 0.6384, "step": 2564 }, { "epoch": 0.31, "learning_rate": 4.932686281689051e-06, "loss": 0.7199, "step": 2565 }, { "epoch": 0.31, "learning_rate": 4.93261217644956e-06, "loss": 0.6942, "step": 2566 }, { "epoch": 0.31, "learning_rate": 4.9325380309987856e-06, "loss": 0.9307, "step": 2567 }, { "epoch": 0.31, "learning_rate": 4.932463845337952e-06, "loss": 1.0665, "step": 2568 }, { "epoch": 0.31, "learning_rate": 4.932389619468289e-06, "loss": 0.7932, "step": 2569 }, { "epoch": 0.31, "learning_rate": 4.93231535339102e-06, "loss": 0.5847, "step": 2570 }, { "epoch": 0.31, "learning_rate": 4.932241047107375e-06, "loss": 0.6728, "step": 2571 }, { "epoch": 0.31, "learning_rate": 4.932166700618581e-06, "loss": 0.7596, "step": 2572 }, { "epoch": 0.31, "learning_rate": 4.932092313925868e-06, "loss": 0.8668, "step": 2573 }, { "epoch": 0.31, "learning_rate": 4.932017887030465e-06, "loss": 1.0533, "step": 2574 }, { "epoch": 0.31, "learning_rate": 4.931943419933602e-06, "loss": 1.1145, "step": 2575 }, { "epoch": 0.31, "learning_rate": 4.93186891263651e-06, "loss": 0.6083, "step": 2576 }, { "epoch": 0.31, "learning_rate": 4.93179436514042e-06, "loss": 0.6688, "step": 2577 }, { "epoch": 0.31, "learning_rate": 4.931719777446566e-06, "loss": 0.8293, "step": 2578 }, { "epoch": 0.31, "learning_rate": 4.93164514955618e-06, "loss": 0.8011, "step": 2579 }, { "epoch": 0.31, "learning_rate": 4.931570481470496e-06, "loss": 1.0377, "step": 2580 }, { "epoch": 0.31, "learning_rate": 4.931495773190747e-06, "loss": 1.2313, "step": 2581 }, { "epoch": 0.31, "learning_rate": 4.9314210247181695e-06, "loss": 0.6522, "step": 2582 }, { "epoch": 0.31, "learning_rate": 4.931346236053998e-06, "loss": 0.6485, "step": 2583 }, { "epoch": 0.31, "learning_rate": 4.931271407199469e-06, "loss": 0.7545, "step": 2584 }, { "epoch": 0.31, "learning_rate": 4.9311965381558204e-06, "loss": 0.7835, "step": 2585 }, { "epoch": 0.31, "learning_rate": 4.931121628924289e-06, "loss": 0.9636, "step": 2586 }, { "epoch": 0.31, "learning_rate": 4.931046679506113e-06, "loss": 1.142, "step": 2587 }, { "epoch": 0.31, "learning_rate": 4.9309716899025305e-06, "loss": 0.6378, "step": 2588 }, { "epoch": 0.31, "learning_rate": 4.930896660114784e-06, "loss": 0.6559, "step": 2589 }, { "epoch": 0.31, "learning_rate": 4.93082159014411e-06, "loss": 0.6965, "step": 2590 }, { "epoch": 0.31, "learning_rate": 4.930746479991752e-06, "loss": 0.7763, "step": 2591 }, { "epoch": 0.31, "learning_rate": 4.93067132965895e-06, "loss": 0.7722, "step": 2592 }, { "epoch": 0.31, "learning_rate": 4.930596139146947e-06, "loss": 1.2042, "step": 2593 }, { "epoch": 0.31, "learning_rate": 4.930520908456986e-06, "loss": 0.6476, "step": 2594 }, { "epoch": 0.31, "learning_rate": 4.9304456375903105e-06, "loss": 0.5836, "step": 2595 }, { "epoch": 0.31, "learning_rate": 4.930370326548165e-06, "loss": 0.8599, "step": 2596 }, { "epoch": 0.31, "learning_rate": 4.930294975331794e-06, "loss": 0.8421, "step": 2597 }, { "epoch": 0.31, "learning_rate": 4.930219583942443e-06, "loss": 0.9244, "step": 2598 }, { "epoch": 0.31, "learning_rate": 4.9301441523813585e-06, "loss": 1.052, "step": 2599 }, { "epoch": 0.31, "learning_rate": 4.930068680649788e-06, "loss": 1.2391, "step": 2600 }, { "epoch": 0.31, "learning_rate": 4.929993168748976e-06, "loss": 0.6225, "step": 2601 }, { "epoch": 0.31, "learning_rate": 4.9299176166801755e-06, "loss": 0.8046, "step": 2602 }, { "epoch": 0.31, "learning_rate": 4.929842024444632e-06, "loss": 0.7538, "step": 2603 }, { "epoch": 0.31, "learning_rate": 4.929766392043596e-06, "loss": 0.8442, "step": 2604 }, { "epoch": 0.31, "learning_rate": 4.929690719478317e-06, "loss": 1.0975, "step": 2605 }, { "epoch": 0.31, "learning_rate": 4.929615006750047e-06, "loss": 1.2108, "step": 2606 }, { "epoch": 0.31, "learning_rate": 4.929539253860037e-06, "loss": 0.6703, "step": 2607 }, { "epoch": 0.31, "learning_rate": 4.92946346080954e-06, "loss": 0.7254, "step": 2608 }, { "epoch": 0.31, "learning_rate": 4.929387627599808e-06, "loss": 0.7613, "step": 2609 }, { "epoch": 0.31, "learning_rate": 4.929311754232093e-06, "loss": 0.7777, "step": 2610 }, { "epoch": 0.31, "learning_rate": 4.9292358407076526e-06, "loss": 1.0671, "step": 2611 }, { "epoch": 0.31, "learning_rate": 4.9291598870277395e-06, "loss": 1.1594, "step": 2612 }, { "epoch": 0.31, "learning_rate": 4.9290838931936096e-06, "loss": 0.6961, "step": 2613 }, { "epoch": 0.31, "learning_rate": 4.92900785920652e-06, "loss": 0.5165, "step": 2614 }, { "epoch": 0.31, "learning_rate": 4.928931785067725e-06, "loss": 0.6976, "step": 2615 }, { "epoch": 0.31, "learning_rate": 4.928855670778485e-06, "loss": 0.8259, "step": 2616 }, { "epoch": 0.31, "learning_rate": 4.9287795163400566e-06, "loss": 1.0985, "step": 2617 }, { "epoch": 0.31, "learning_rate": 4.928703321753698e-06, "loss": 1.2559, "step": 2618 }, { "epoch": 0.31, "learning_rate": 4.928627087020671e-06, "loss": 0.7507, "step": 2619 }, { "epoch": 0.31, "learning_rate": 4.928550812142234e-06, "loss": 0.6375, "step": 2620 }, { "epoch": 0.31, "learning_rate": 4.928474497119649e-06, "loss": 0.7556, "step": 2621 }, { "epoch": 0.31, "learning_rate": 4.928398141954176e-06, "loss": 0.8031, "step": 2622 }, { "epoch": 0.31, "learning_rate": 4.928321746647078e-06, "loss": 0.7867, "step": 2623 }, { "epoch": 0.31, "learning_rate": 4.928245311199618e-06, "loss": 1.0482, "step": 2624 }, { "epoch": 0.31, "learning_rate": 4.928168835613059e-06, "loss": 1.3118, "step": 2625 }, { "epoch": 0.31, "learning_rate": 4.928092319888666e-06, "loss": 0.5849, "step": 2626 }, { "epoch": 0.31, "learning_rate": 4.928015764027702e-06, "loss": 0.891, "step": 2627 }, { "epoch": 0.31, "learning_rate": 4.927939168031436e-06, "loss": 0.8853, "step": 2628 }, { "epoch": 0.31, "learning_rate": 4.92786253190113e-06, "loss": 0.9519, "step": 2629 }, { "epoch": 0.31, "learning_rate": 4.927785855638053e-06, "loss": 1.0579, "step": 2630 }, { "epoch": 0.31, "learning_rate": 4.927709139243472e-06, "loss": 1.1857, "step": 2631 }, { "epoch": 0.31, "learning_rate": 4.927632382718655e-06, "loss": 0.6724, "step": 2632 }, { "epoch": 0.31, "learning_rate": 4.927555586064871e-06, "loss": 0.6844, "step": 2633 }, { "epoch": 0.31, "learning_rate": 4.927478749283391e-06, "loss": 0.8033, "step": 2634 }, { "epoch": 0.31, "learning_rate": 4.927401872375482e-06, "loss": 0.7654, "step": 2635 }, { "epoch": 0.31, "learning_rate": 4.927324955342417e-06, "loss": 0.9928, "step": 2636 }, { "epoch": 0.31, "learning_rate": 4.927247998185466e-06, "loss": 1.1896, "step": 2637 }, { "epoch": 0.31, "learning_rate": 4.9271710009059024e-06, "loss": 0.5879, "step": 2638 }, { "epoch": 0.31, "learning_rate": 4.9270939635049985e-06, "loss": 0.6805, "step": 2639 }, { "epoch": 0.31, "learning_rate": 4.927016885984028e-06, "loss": 0.8221, "step": 2640 }, { "epoch": 0.31, "learning_rate": 4.926939768344264e-06, "loss": 0.7748, "step": 2641 }, { "epoch": 0.31, "learning_rate": 4.9268626105869836e-06, "loss": 1.093, "step": 2642 }, { "epoch": 0.31, "learning_rate": 4.926785412713459e-06, "loss": 1.2499, "step": 2643 }, { "epoch": 0.31, "learning_rate": 4.926708174724969e-06, "loss": 0.7523, "step": 2644 }, { "epoch": 0.31, "learning_rate": 4.926630896622788e-06, "loss": 0.6256, "step": 2645 }, { "epoch": 0.32, "learning_rate": 4.9265535784081965e-06, "loss": 0.7784, "step": 2646 }, { "epoch": 0.32, "learning_rate": 4.926476220082469e-06, "loss": 0.7788, "step": 2647 }, { "epoch": 0.32, "learning_rate": 4.9263988216468865e-06, "loss": 0.9412, "step": 2648 }, { "epoch": 0.32, "learning_rate": 4.926321383102728e-06, "loss": 1.1037, "step": 2649 }, { "epoch": 0.32, "learning_rate": 4.926243904451274e-06, "loss": 1.1805, "step": 2650 }, { "epoch": 0.32, "learning_rate": 4.926166385693804e-06, "loss": 0.6299, "step": 2651 }, { "epoch": 0.32, "learning_rate": 4.9260888268316e-06, "loss": 0.6886, "step": 2652 }, { "epoch": 0.32, "learning_rate": 4.926011227865945e-06, "loss": 0.7658, "step": 2653 }, { "epoch": 0.32, "learning_rate": 4.925933588798119e-06, "loss": 1.0449, "step": 2654 }, { "epoch": 0.32, "learning_rate": 4.925855909629409e-06, "loss": 1.16, "step": 2655 }, { "epoch": 0.32, "learning_rate": 4.925778190361097e-06, "loss": 1.262, "step": 2656 }, { "epoch": 0.32, "learning_rate": 4.925700430994468e-06, "loss": 0.5615, "step": 2657 }, { "epoch": 0.32, "learning_rate": 4.925622631530806e-06, "loss": 0.6579, "step": 2658 }, { "epoch": 0.32, "learning_rate": 4.9255447919714e-06, "loss": 0.8024, "step": 2659 }, { "epoch": 0.32, "learning_rate": 4.925466912317535e-06, "loss": 0.8135, "step": 2660 }, { "epoch": 0.32, "learning_rate": 4.925388992570497e-06, "loss": 1.0268, "step": 2661 }, { "epoch": 0.32, "learning_rate": 4.925311032731577e-06, "loss": 1.1918, "step": 2662 }, { "epoch": 0.32, "learning_rate": 4.925233032802062e-06, "loss": 0.5604, "step": 2663 }, { "epoch": 0.32, "learning_rate": 4.92515499278324e-06, "loss": 0.4914, "step": 2664 }, { "epoch": 0.32, "learning_rate": 4.925076912676404e-06, "loss": 0.7233, "step": 2665 }, { "epoch": 0.32, "learning_rate": 4.924998792482843e-06, "loss": 0.766, "step": 2666 }, { "epoch": 0.32, "learning_rate": 4.924920632203847e-06, "loss": 0.8466, "step": 2667 }, { "epoch": 0.32, "learning_rate": 4.924842431840711e-06, "loss": 1.0371, "step": 2668 }, { "epoch": 0.32, "learning_rate": 4.924764191394725e-06, "loss": 0.6979, "step": 2669 }, { "epoch": 0.32, "learning_rate": 4.924685910867184e-06, "loss": 0.5578, "step": 2670 }, { "epoch": 0.32, "learning_rate": 4.924607590259382e-06, "loss": 0.823, "step": 2671 }, { "epoch": 0.32, "learning_rate": 4.924529229572613e-06, "loss": 0.8397, "step": 2672 }, { "epoch": 0.32, "learning_rate": 4.924450828808172e-06, "loss": 0.8704, "step": 2673 }, { "epoch": 0.32, "learning_rate": 4.924372387967355e-06, "loss": 1.1199, "step": 2674 }, { "epoch": 0.32, "learning_rate": 4.924293907051459e-06, "loss": 1.1724, "step": 2675 }, { "epoch": 0.32, "learning_rate": 4.9242153860617825e-06, "loss": 0.6191, "step": 2676 }, { "epoch": 0.32, "learning_rate": 4.9241368249996214e-06, "loss": 0.7279, "step": 2677 }, { "epoch": 0.32, "learning_rate": 4.924058223866276e-06, "loss": 0.8069, "step": 2678 }, { "epoch": 0.32, "learning_rate": 4.9239795826630435e-06, "loss": 0.9689, "step": 2679 }, { "epoch": 0.32, "learning_rate": 4.923900901391226e-06, "loss": 1.1015, "step": 2680 }, { "epoch": 0.32, "learning_rate": 4.923822180052123e-06, "loss": 1.0985, "step": 2681 }, { "epoch": 0.32, "learning_rate": 4.9237434186470355e-06, "loss": 0.6107, "step": 2682 }, { "epoch": 0.32, "learning_rate": 4.923664617177266e-06, "loss": 0.8242, "step": 2683 }, { "epoch": 0.32, "learning_rate": 4.923585775644117e-06, "loss": 0.8627, "step": 2684 }, { "epoch": 0.32, "learning_rate": 4.923506894048893e-06, "loss": 0.7687, "step": 2685 }, { "epoch": 0.32, "learning_rate": 4.923427972392896e-06, "loss": 1.0294, "step": 2686 }, { "epoch": 0.32, "learning_rate": 4.92334901067743e-06, "loss": 1.2703, "step": 2687 }, { "epoch": 0.32, "learning_rate": 4.923270008903803e-06, "loss": 0.5747, "step": 2688 }, { "epoch": 0.32, "learning_rate": 4.923190967073319e-06, "loss": 0.5859, "step": 2689 }, { "epoch": 0.32, "learning_rate": 4.923111885187285e-06, "loss": 0.7261, "step": 2690 }, { "epoch": 0.32, "learning_rate": 4.923032763247009e-06, "loss": 0.7106, "step": 2691 }, { "epoch": 0.32, "learning_rate": 4.922953601253797e-06, "loss": 0.9017, "step": 2692 }, { "epoch": 0.32, "learning_rate": 4.922874399208959e-06, "loss": 1.297, "step": 2693 }, { "epoch": 0.32, "learning_rate": 4.922795157113803e-06, "loss": 0.5456, "step": 2694 }, { "epoch": 0.32, "learning_rate": 4.922715874969641e-06, "loss": 0.6441, "step": 2695 }, { "epoch": 0.32, "learning_rate": 4.922636552777782e-06, "loss": 0.6924, "step": 2696 }, { "epoch": 0.32, "learning_rate": 4.922557190539537e-06, "loss": 0.7767, "step": 2697 }, { "epoch": 0.32, "learning_rate": 4.922477788256219e-06, "loss": 0.8791, "step": 2698 }, { "epoch": 0.32, "learning_rate": 4.92239834592914e-06, "loss": 1.0999, "step": 2699 }, { "epoch": 0.32, "learning_rate": 4.922318863559613e-06, "loss": 1.2937, "step": 2700 }, { "epoch": 0.32, "learning_rate": 4.922239341148952e-06, "loss": 0.6193, "step": 2701 }, { "epoch": 0.32, "learning_rate": 4.922159778698471e-06, "loss": 0.6458, "step": 2702 }, { "epoch": 0.32, "learning_rate": 4.9220801762094864e-06, "loss": 0.7717, "step": 2703 }, { "epoch": 0.32, "learning_rate": 4.922000533683312e-06, "loss": 0.9344, "step": 2704 }, { "epoch": 0.32, "learning_rate": 4.921920851121266e-06, "loss": 1.0836, "step": 2705 }, { "epoch": 0.32, "learning_rate": 4.921841128524666e-06, "loss": 1.2572, "step": 2706 }, { "epoch": 0.32, "learning_rate": 4.921761365894828e-06, "loss": 0.7162, "step": 2707 }, { "epoch": 0.32, "learning_rate": 4.921681563233072e-06, "loss": 0.7047, "step": 2708 }, { "epoch": 0.32, "learning_rate": 4.921601720540716e-06, "loss": 0.8005, "step": 2709 }, { "epoch": 0.32, "learning_rate": 4.9215218378190806e-06, "loss": 0.8798, "step": 2710 }, { "epoch": 0.32, "learning_rate": 4.921441915069486e-06, "loss": 1.1137, "step": 2711 }, { "epoch": 0.32, "learning_rate": 4.921361952293253e-06, "loss": 1.1366, "step": 2712 }, { "epoch": 0.32, "learning_rate": 4.921281949491705e-06, "loss": 0.7083, "step": 2713 }, { "epoch": 0.32, "learning_rate": 4.921201906666162e-06, "loss": 0.6727, "step": 2714 }, { "epoch": 0.32, "learning_rate": 4.921121823817949e-06, "loss": 0.7723, "step": 2715 }, { "epoch": 0.32, "learning_rate": 4.921041700948388e-06, "loss": 0.842, "step": 2716 }, { "epoch": 0.32, "learning_rate": 4.920961538058806e-06, "loss": 1.0531, "step": 2717 }, { "epoch": 0.32, "learning_rate": 4.920881335150526e-06, "loss": 1.0042, "step": 2718 }, { "epoch": 0.32, "learning_rate": 4.920801092224874e-06, "loss": 1.1038, "step": 2719 }, { "epoch": 0.32, "learning_rate": 4.920720809283177e-06, "loss": 0.5152, "step": 2720 }, { "epoch": 0.32, "learning_rate": 4.920640486326763e-06, "loss": 0.6825, "step": 2721 }, { "epoch": 0.32, "learning_rate": 4.920560123356958e-06, "loss": 0.6953, "step": 2722 }, { "epoch": 0.32, "learning_rate": 4.920479720375091e-06, "loss": 0.8015, "step": 2723 }, { "epoch": 0.32, "learning_rate": 4.920399277382491e-06, "loss": 1.1566, "step": 2724 }, { "epoch": 0.32, "learning_rate": 4.9203187943804875e-06, "loss": 1.1908, "step": 2725 }, { "epoch": 0.32, "learning_rate": 4.920238271370411e-06, "loss": 0.7959, "step": 2726 }, { "epoch": 0.32, "learning_rate": 4.920157708353594e-06, "loss": 0.8089, "step": 2727 }, { "epoch": 0.32, "learning_rate": 4.920077105331366e-06, "loss": 0.8699, "step": 2728 }, { "epoch": 0.32, "learning_rate": 4.919996462305061e-06, "loss": 0.8146, "step": 2729 }, { "epoch": 0.33, "learning_rate": 4.9199157792760115e-06, "loss": 1.1318, "step": 2730 }, { "epoch": 0.33, "learning_rate": 4.919835056245551e-06, "loss": 1.1772, "step": 2731 }, { "epoch": 0.33, "learning_rate": 4.919754293215013e-06, "loss": 0.7495, "step": 2732 }, { "epoch": 0.33, "learning_rate": 4.919673490185735e-06, "loss": 0.6739, "step": 2733 }, { "epoch": 0.33, "learning_rate": 4.91959264715905e-06, "loss": 0.7852, "step": 2734 }, { "epoch": 0.33, "learning_rate": 4.919511764136297e-06, "loss": 0.8159, "step": 2735 }, { "epoch": 0.33, "learning_rate": 4.919430841118809e-06, "loss": 0.9202, "step": 2736 }, { "epoch": 0.33, "learning_rate": 4.9193498781079285e-06, "loss": 1.1755, "step": 2737 }, { "epoch": 0.33, "learning_rate": 4.919268875104991e-06, "loss": 0.6794, "step": 2738 }, { "epoch": 0.33, "learning_rate": 4.919187832111336e-06, "loss": 0.6146, "step": 2739 }, { "epoch": 0.33, "learning_rate": 4.9191067491283025e-06, "loss": 0.7332, "step": 2740 }, { "epoch": 0.33, "learning_rate": 4.9190256261572324e-06, "loss": 0.7648, "step": 2741 }, { "epoch": 0.33, "learning_rate": 4.918944463199465e-06, "loss": 0.9862, "step": 2742 }, { "epoch": 0.33, "learning_rate": 4.918863260256342e-06, "loss": 1.2111, "step": 2743 }, { "epoch": 0.33, "learning_rate": 4.9187820173292075e-06, "loss": 0.7169, "step": 2744 }, { "epoch": 0.33, "learning_rate": 4.918700734419404e-06, "loss": 0.679, "step": 2745 }, { "epoch": 0.33, "learning_rate": 4.9186194115282724e-06, "loss": 0.8629, "step": 2746 }, { "epoch": 0.33, "learning_rate": 4.91853804865716e-06, "loss": 0.8291, "step": 2747 }, { "epoch": 0.33, "learning_rate": 4.918456645807411e-06, "loss": 0.9869, "step": 2748 }, { "epoch": 0.33, "learning_rate": 4.91837520298037e-06, "loss": 0.9572, "step": 2749 }, { "epoch": 0.33, "learning_rate": 4.918293720177385e-06, "loss": 1.0949, "step": 2750 }, { "epoch": 0.33, "learning_rate": 4.918212197399801e-06, "loss": 0.548, "step": 2751 }, { "epoch": 0.33, "learning_rate": 4.9181306346489665e-06, "loss": 0.587, "step": 2752 }, { "epoch": 0.33, "learning_rate": 4.91804903192623e-06, "loss": 0.8519, "step": 2753 }, { "epoch": 0.33, "learning_rate": 4.91796738923294e-06, "loss": 0.8488, "step": 2754 }, { "epoch": 0.33, "learning_rate": 4.917885706570447e-06, "loss": 1.0744, "step": 2755 }, { "epoch": 0.33, "learning_rate": 4.917803983940099e-06, "loss": 1.1411, "step": 2756 }, { "epoch": 0.33, "learning_rate": 4.9177222213432495e-06, "loss": 0.6419, "step": 2757 }, { "epoch": 0.33, "learning_rate": 4.917640418781248e-06, "loss": 0.7549, "step": 2758 }, { "epoch": 0.33, "learning_rate": 4.917558576255448e-06, "loss": 0.7659, "step": 2759 }, { "epoch": 0.33, "learning_rate": 4.917476693767202e-06, "loss": 0.8334, "step": 2760 }, { "epoch": 0.33, "learning_rate": 4.917394771317863e-06, "loss": 1.0215, "step": 2761 }, { "epoch": 0.33, "learning_rate": 4.917312808908785e-06, "loss": 1.2716, "step": 2762 }, { "epoch": 0.33, "learning_rate": 4.917230806541325e-06, "loss": 0.5656, "step": 2763 }, { "epoch": 0.33, "learning_rate": 4.9171487642168355e-06, "loss": 0.7528, "step": 2764 }, { "epoch": 0.33, "learning_rate": 4.917066681936675e-06, "loss": 0.6509, "step": 2765 }, { "epoch": 0.33, "learning_rate": 4.916984559702199e-06, "loss": 0.7694, "step": 2766 }, { "epoch": 0.33, "learning_rate": 4.916902397514765e-06, "loss": 0.8981, "step": 2767 }, { "epoch": 0.33, "learning_rate": 4.9168201953757326e-06, "loss": 1.1467, "step": 2768 }, { "epoch": 0.33, "learning_rate": 4.916737953286459e-06, "loss": 0.8993, "step": 2769 }, { "epoch": 0.33, "learning_rate": 4.916655671248304e-06, "loss": 0.6876, "step": 2770 }, { "epoch": 0.33, "learning_rate": 4.916573349262629e-06, "loss": 0.6992, "step": 2771 }, { "epoch": 0.33, "learning_rate": 4.9164909873307925e-06, "loss": 0.7618, "step": 2772 }, { "epoch": 0.33, "learning_rate": 4.916408585454158e-06, "loss": 0.8364, "step": 2773 }, { "epoch": 0.33, "learning_rate": 4.916326143634087e-06, "loss": 1.1863, "step": 2774 }, { "epoch": 0.33, "learning_rate": 4.916243661871941e-06, "loss": 1.3289, "step": 2775 }, { "epoch": 0.33, "learning_rate": 4.916161140169085e-06, "loss": 0.5551, "step": 2776 }, { "epoch": 0.33, "learning_rate": 4.9160785785268835e-06, "loss": 0.7612, "step": 2777 }, { "epoch": 0.33, "learning_rate": 4.9159959769467e-06, "loss": 0.8978, "step": 2778 }, { "epoch": 0.33, "learning_rate": 4.9159133354299e-06, "loss": 0.768, "step": 2779 }, { "epoch": 0.33, "learning_rate": 4.91583065397785e-06, "loss": 1.0844, "step": 2780 }, { "epoch": 0.33, "learning_rate": 4.915747932591916e-06, "loss": 1.2982, "step": 2781 }, { "epoch": 0.33, "learning_rate": 4.915665171273467e-06, "loss": 0.605, "step": 2782 }, { "epoch": 0.33, "learning_rate": 4.91558237002387e-06, "loss": 0.7772, "step": 2783 }, { "epoch": 0.33, "learning_rate": 4.9154995288444925e-06, "loss": 0.7025, "step": 2784 }, { "epoch": 0.33, "learning_rate": 4.915416647736706e-06, "loss": 0.9606, "step": 2785 }, { "epoch": 0.33, "learning_rate": 4.915333726701879e-06, "loss": 0.9804, "step": 2786 }, { "epoch": 0.33, "learning_rate": 4.915250765741384e-06, "loss": 1.1158, "step": 2787 }, { "epoch": 0.33, "learning_rate": 4.9151677648565895e-06, "loss": 0.8517, "step": 2788 }, { "epoch": 0.33, "learning_rate": 4.915084724048871e-06, "loss": 0.6169, "step": 2789 }, { "epoch": 0.33, "learning_rate": 4.915001643319598e-06, "loss": 0.6512, "step": 2790 }, { "epoch": 0.33, "learning_rate": 4.914918522670146e-06, "loss": 0.7052, "step": 2791 }, { "epoch": 0.33, "learning_rate": 4.914835362101888e-06, "loss": 0.9788, "step": 2792 }, { "epoch": 0.33, "learning_rate": 4.914752161616198e-06, "loss": 1.1114, "step": 2793 }, { "epoch": 0.33, "learning_rate": 4.9146689212144534e-06, "loss": 0.6338, "step": 2794 }, { "epoch": 0.33, "learning_rate": 4.9145856408980285e-06, "loss": 0.5468, "step": 2795 }, { "epoch": 0.33, "learning_rate": 4.914502320668301e-06, "loss": 0.8281, "step": 2796 }, { "epoch": 0.33, "learning_rate": 4.914418960526647e-06, "loss": 0.6891, "step": 2797 }, { "epoch": 0.33, "learning_rate": 4.914335560474445e-06, "loss": 0.9224, "step": 2798 }, { "epoch": 0.33, "learning_rate": 4.914252120513073e-06, "loss": 0.9748, "step": 2799 }, { "epoch": 0.33, "learning_rate": 4.914168640643911e-06, "loss": 1.4645, "step": 2800 }, { "epoch": 0.33, "learning_rate": 4.9140851208683395e-06, "loss": 0.6499, "step": 2801 }, { "epoch": 0.33, "learning_rate": 4.914001561187738e-06, "loss": 0.6451, "step": 2802 }, { "epoch": 0.33, "learning_rate": 4.913917961603488e-06, "loss": 0.6998, "step": 2803 }, { "epoch": 0.33, "learning_rate": 4.913834322116972e-06, "loss": 0.9276, "step": 2804 }, { "epoch": 0.33, "learning_rate": 4.913750642729572e-06, "loss": 0.9584, "step": 2805 }, { "epoch": 0.33, "learning_rate": 4.913666923442671e-06, "loss": 1.0212, "step": 2806 }, { "epoch": 0.33, "learning_rate": 4.913583164257654e-06, "loss": 0.6914, "step": 2807 }, { "epoch": 0.33, "learning_rate": 4.913499365175903e-06, "loss": 0.7136, "step": 2808 }, { "epoch": 0.33, "learning_rate": 4.913415526198807e-06, "loss": 0.7939, "step": 2809 }, { "epoch": 0.33, "learning_rate": 4.913331647327749e-06, "loss": 0.6966, "step": 2810 }, { "epoch": 0.33, "learning_rate": 4.9132477285641165e-06, "loss": 0.9725, "step": 2811 }, { "epoch": 0.33, "learning_rate": 4.913163769909296e-06, "loss": 1.2609, "step": 2812 }, { "epoch": 0.34, "learning_rate": 4.9130797713646765e-06, "loss": 0.5607, "step": 2813 }, { "epoch": 0.34, "learning_rate": 4.912995732931646e-06, "loss": 0.7832, "step": 2814 }, { "epoch": 0.34, "learning_rate": 4.9129116546115935e-06, "loss": 0.6264, "step": 2815 }, { "epoch": 0.34, "learning_rate": 4.912827536405909e-06, "loss": 0.7992, "step": 2816 }, { "epoch": 0.34, "learning_rate": 4.912743378315982e-06, "loss": 0.9448, "step": 2817 }, { "epoch": 0.34, "learning_rate": 4.912659180343205e-06, "loss": 1.037, "step": 2818 }, { "epoch": 0.34, "learning_rate": 4.91257494248897e-06, "loss": 0.7464, "step": 2819 }, { "epoch": 0.34, "learning_rate": 4.912490664754667e-06, "loss": 0.6739, "step": 2820 }, { "epoch": 0.34, "learning_rate": 4.9124063471416925e-06, "loss": 0.7838, "step": 2821 }, { "epoch": 0.34, "learning_rate": 4.912321989651438e-06, "loss": 0.9244, "step": 2822 }, { "epoch": 0.34, "learning_rate": 4.9122375922852995e-06, "loss": 0.927, "step": 2823 }, { "epoch": 0.34, "learning_rate": 4.912153155044671e-06, "loss": 1.0106, "step": 2824 }, { "epoch": 0.34, "learning_rate": 4.912068677930948e-06, "loss": 1.1797, "step": 2825 }, { "epoch": 0.34, "learning_rate": 4.9119841609455275e-06, "loss": 0.5519, "step": 2826 }, { "epoch": 0.34, "learning_rate": 4.911899604089806e-06, "loss": 0.6936, "step": 2827 }, { "epoch": 0.34, "learning_rate": 4.911815007365182e-06, "loss": 0.7838, "step": 2828 }, { "epoch": 0.34, "learning_rate": 4.911730370773054e-06, "loss": 0.8624, "step": 2829 }, { "epoch": 0.34, "learning_rate": 4.91164569431482e-06, "loss": 1.0731, "step": 2830 }, { "epoch": 0.34, "learning_rate": 4.911560977991881e-06, "loss": 1.1422, "step": 2831 }, { "epoch": 0.34, "learning_rate": 4.911476221805635e-06, "loss": 0.7126, "step": 2832 }, { "epoch": 0.34, "learning_rate": 4.911391425757486e-06, "loss": 0.8013, "step": 2833 }, { "epoch": 0.34, "learning_rate": 4.9113065898488346e-06, "loss": 0.7574, "step": 2834 }, { "epoch": 0.34, "learning_rate": 4.9112217140810825e-06, "loss": 0.8797, "step": 2835 }, { "epoch": 0.34, "learning_rate": 4.911136798455634e-06, "loss": 0.9867, "step": 2836 }, { "epoch": 0.34, "learning_rate": 4.91105184297389e-06, "loss": 1.0391, "step": 2837 }, { "epoch": 0.34, "learning_rate": 4.910966847637258e-06, "loss": 0.8569, "step": 2838 }, { "epoch": 0.34, "learning_rate": 4.910881812447142e-06, "loss": 0.7585, "step": 2839 }, { "epoch": 0.34, "learning_rate": 4.910796737404947e-06, "loss": 0.8766, "step": 2840 }, { "epoch": 0.34, "learning_rate": 4.91071162251208e-06, "loss": 0.8122, "step": 2841 }, { "epoch": 0.34, "learning_rate": 4.910626467769947e-06, "loss": 1.0159, "step": 2842 }, { "epoch": 0.34, "learning_rate": 4.910541273179956e-06, "loss": 1.1288, "step": 2843 }, { "epoch": 0.34, "learning_rate": 4.910456038743516e-06, "loss": 0.7306, "step": 2844 }, { "epoch": 0.34, "learning_rate": 4.910370764462036e-06, "loss": 0.5381, "step": 2845 }, { "epoch": 0.34, "learning_rate": 4.910285450336923e-06, "loss": 0.7851, "step": 2846 }, { "epoch": 0.34, "learning_rate": 4.910200096369591e-06, "loss": 0.7603, "step": 2847 }, { "epoch": 0.34, "learning_rate": 4.910114702561449e-06, "loss": 0.8848, "step": 2848 }, { "epoch": 0.34, "learning_rate": 4.910029268913909e-06, "loss": 1.0398, "step": 2849 }, { "epoch": 0.34, "learning_rate": 4.909943795428383e-06, "loss": 1.1293, "step": 2850 }, { "epoch": 0.34, "learning_rate": 4.909858282106283e-06, "loss": 0.6084, "step": 2851 }, { "epoch": 0.34, "learning_rate": 4.9097727289490235e-06, "loss": 0.6215, "step": 2852 }, { "epoch": 0.34, "learning_rate": 4.909687135958019e-06, "loss": 0.7261, "step": 2853 }, { "epoch": 0.34, "learning_rate": 4.909601503134684e-06, "loss": 1.0333, "step": 2854 }, { "epoch": 0.34, "learning_rate": 4.909515830480434e-06, "loss": 1.0416, "step": 2855 }, { "epoch": 0.34, "learning_rate": 4.909430117996684e-06, "loss": 1.1581, "step": 2856 }, { "epoch": 0.34, "learning_rate": 4.909344365684853e-06, "loss": 0.7031, "step": 2857 }, { "epoch": 0.34, "learning_rate": 4.9092585735463576e-06, "loss": 0.668, "step": 2858 }, { "epoch": 0.34, "learning_rate": 4.909172741582616e-06, "loss": 0.6863, "step": 2859 }, { "epoch": 0.34, "learning_rate": 4.909086869795047e-06, "loss": 0.8932, "step": 2860 }, { "epoch": 0.34, "learning_rate": 4.9090009581850685e-06, "loss": 0.9437, "step": 2861 }, { "epoch": 0.34, "learning_rate": 4.908915006754103e-06, "loss": 1.0753, "step": 2862 }, { "epoch": 0.34, "learning_rate": 4.90882901550357e-06, "loss": 0.51, "step": 2863 }, { "epoch": 0.34, "learning_rate": 4.9087429844348915e-06, "loss": 0.5885, "step": 2864 }, { "epoch": 0.34, "learning_rate": 4.90865691354949e-06, "loss": 0.7521, "step": 2865 }, { "epoch": 0.34, "learning_rate": 4.908570802848787e-06, "loss": 0.7366, "step": 2866 }, { "epoch": 0.34, "learning_rate": 4.908484652334206e-06, "loss": 0.9437, "step": 2867 }, { "epoch": 0.34, "learning_rate": 4.908398462007173e-06, "loss": 1.2172, "step": 2868 }, { "epoch": 0.34, "learning_rate": 4.908312231869109e-06, "loss": 0.6465, "step": 2869 }, { "epoch": 0.34, "learning_rate": 4.9082259619214434e-06, "loss": 0.6, "step": 2870 }, { "epoch": 0.34, "learning_rate": 4.908139652165601e-06, "loss": 0.7492, "step": 2871 }, { "epoch": 0.34, "learning_rate": 4.908053302603007e-06, "loss": 0.8167, "step": 2872 }, { "epoch": 0.34, "learning_rate": 4.907966913235091e-06, "loss": 0.882, "step": 2873 }, { "epoch": 0.34, "learning_rate": 4.9078804840632784e-06, "loss": 0.9653, "step": 2874 }, { "epoch": 0.34, "learning_rate": 4.907794015089e-06, "loss": 1.2894, "step": 2875 }, { "epoch": 0.34, "learning_rate": 4.907707506313685e-06, "loss": 0.613, "step": 2876 }, { "epoch": 0.34, "learning_rate": 4.907620957738763e-06, "loss": 0.5951, "step": 2877 }, { "epoch": 0.34, "learning_rate": 4.907534369365663e-06, "loss": 0.6025, "step": 2878 }, { "epoch": 0.34, "learning_rate": 4.907447741195819e-06, "loss": 0.8673, "step": 2879 }, { "epoch": 0.34, "learning_rate": 4.907361073230662e-06, "loss": 1.0388, "step": 2880 }, { "epoch": 0.34, "learning_rate": 4.907274365471623e-06, "loss": 1.1077, "step": 2881 }, { "epoch": 0.34, "learning_rate": 4.907187617920139e-06, "loss": 0.6617, "step": 2882 }, { "epoch": 0.34, "learning_rate": 4.907100830577641e-06, "loss": 0.7711, "step": 2883 }, { "epoch": 0.34, "learning_rate": 4.907014003445564e-06, "loss": 0.7881, "step": 2884 }, { "epoch": 0.34, "learning_rate": 4.906927136525343e-06, "loss": 0.9392, "step": 2885 }, { "epoch": 0.34, "learning_rate": 4.906840229818415e-06, "loss": 1.0636, "step": 2886 }, { "epoch": 0.34, "learning_rate": 4.906753283326216e-06, "loss": 1.1523, "step": 2887 }, { "epoch": 0.34, "learning_rate": 4.906666297050183e-06, "loss": 0.6369, "step": 2888 }, { "epoch": 0.34, "learning_rate": 4.906579270991755e-06, "loss": 0.7614, "step": 2889 }, { "epoch": 0.34, "learning_rate": 4.906492205152369e-06, "loss": 0.7676, "step": 2890 }, { "epoch": 0.34, "learning_rate": 4.906405099533466e-06, "loss": 0.7077, "step": 2891 }, { "epoch": 0.34, "learning_rate": 4.906317954136484e-06, "loss": 0.9564, "step": 2892 }, { "epoch": 0.34, "learning_rate": 4.906230768962864e-06, "loss": 1.1969, "step": 2893 }, { "epoch": 0.34, "learning_rate": 4.906143544014048e-06, "loss": 1.032, "step": 2894 }, { "epoch": 0.34, "learning_rate": 4.9060562792914766e-06, "loss": 0.6328, "step": 2895 }, { "epoch": 0.34, "learning_rate": 4.9059689747965934e-06, "loss": 0.724, "step": 2896 }, { "epoch": 0.35, "learning_rate": 4.905881630530841e-06, "loss": 0.7857, "step": 2897 }, { "epoch": 0.35, "learning_rate": 4.905794246495664e-06, "loss": 0.8929, "step": 2898 }, { "epoch": 0.35, "learning_rate": 4.905706822692506e-06, "loss": 0.9844, "step": 2899 }, { "epoch": 0.35, "learning_rate": 4.905619359122812e-06, "loss": 1.0496, "step": 2900 }, { "epoch": 0.35, "learning_rate": 4.9055318557880285e-06, "loss": 0.5723, "step": 2901 }, { "epoch": 0.35, "learning_rate": 4.905444312689601e-06, "loss": 0.7711, "step": 2902 }, { "epoch": 0.35, "learning_rate": 4.905356729828978e-06, "loss": 0.747, "step": 2903 }, { "epoch": 0.35, "learning_rate": 4.905269107207606e-06, "loss": 0.8947, "step": 2904 }, { "epoch": 0.35, "learning_rate": 4.905181444826934e-06, "loss": 1.0372, "step": 2905 }, { "epoch": 0.35, "learning_rate": 4.905093742688411e-06, "loss": 1.226, "step": 2906 }, { "epoch": 0.35, "learning_rate": 4.905006000793486e-06, "loss": 0.4613, "step": 2907 }, { "epoch": 0.35, "learning_rate": 4.904918219143611e-06, "loss": 0.5844, "step": 2908 }, { "epoch": 0.35, "learning_rate": 4.904830397740236e-06, "loss": 0.7249, "step": 2909 }, { "epoch": 0.35, "learning_rate": 4.904742536584813e-06, "loss": 0.788, "step": 2910 }, { "epoch": 0.35, "learning_rate": 4.904654635678792e-06, "loss": 0.9075, "step": 2911 }, { "epoch": 0.35, "learning_rate": 4.904566695023629e-06, "loss": 1.1142, "step": 2912 }, { "epoch": 0.35, "learning_rate": 4.904478714620778e-06, "loss": 0.7849, "step": 2913 }, { "epoch": 0.35, "learning_rate": 4.904390694471691e-06, "loss": 0.6373, "step": 2914 }, { "epoch": 0.35, "learning_rate": 4.9043026345778245e-06, "loss": 0.7706, "step": 2915 }, { "epoch": 0.35, "learning_rate": 4.9042145349406335e-06, "loss": 0.7768, "step": 2916 }, { "epoch": 0.35, "learning_rate": 4.904126395561575e-06, "loss": 0.9018, "step": 2917 }, { "epoch": 0.35, "learning_rate": 4.904038216442105e-06, "loss": 1.1276, "step": 2918 }, { "epoch": 0.35, "learning_rate": 4.903949997583682e-06, "loss": 1.0014, "step": 2919 }, { "epoch": 0.35, "learning_rate": 4.903861738987763e-06, "loss": 0.5497, "step": 2920 }, { "epoch": 0.35, "learning_rate": 4.903773440655808e-06, "loss": 0.7827, "step": 2921 }, { "epoch": 0.35, "learning_rate": 4.903685102589277e-06, "loss": 0.7644, "step": 2922 }, { "epoch": 0.35, "learning_rate": 4.903596724789629e-06, "loss": 0.8741, "step": 2923 }, { "epoch": 0.35, "learning_rate": 4.903508307258326e-06, "loss": 1.093, "step": 2924 }, { "epoch": 0.35, "learning_rate": 4.903419849996828e-06, "loss": 1.1772, "step": 2925 }, { "epoch": 0.35, "learning_rate": 4.9033313530066e-06, "loss": 0.6451, "step": 2926 }, { "epoch": 0.35, "learning_rate": 4.903242816289101e-06, "loss": 0.7286, "step": 2927 }, { "epoch": 0.35, "learning_rate": 4.903154239845798e-06, "loss": 0.8126, "step": 2928 }, { "epoch": 0.35, "learning_rate": 4.903065623678152e-06, "loss": 0.8325, "step": 2929 }, { "epoch": 0.35, "learning_rate": 4.902976967787632e-06, "loss": 1.0447, "step": 2930 }, { "epoch": 0.35, "learning_rate": 4.9028882721757e-06, "loss": 1.0688, "step": 2931 }, { "epoch": 0.35, "learning_rate": 4.9027995368438234e-06, "loss": 0.5714, "step": 2932 }, { "epoch": 0.35, "learning_rate": 4.902710761793468e-06, "loss": 0.6992, "step": 2933 }, { "epoch": 0.35, "learning_rate": 4.9026219470261026e-06, "loss": 0.7767, "step": 2934 }, { "epoch": 0.35, "learning_rate": 4.902533092543196e-06, "loss": 0.7749, "step": 2935 }, { "epoch": 0.35, "learning_rate": 4.902444198346215e-06, "loss": 0.8779, "step": 2936 }, { "epoch": 0.35, "learning_rate": 4.902355264436629e-06, "loss": 1.094, "step": 2937 }, { "epoch": 0.35, "learning_rate": 4.902266290815909e-06, "loss": 0.7882, "step": 2938 }, { "epoch": 0.35, "learning_rate": 4.902177277485527e-06, "loss": 0.5565, "step": 2939 }, { "epoch": 0.35, "learning_rate": 4.902088224446951e-06, "loss": 0.9407, "step": 2940 }, { "epoch": 0.35, "learning_rate": 4.901999131701657e-06, "loss": 0.8428, "step": 2941 }, { "epoch": 0.35, "learning_rate": 4.901909999251116e-06, "loss": 0.9373, "step": 2942 }, { "epoch": 0.35, "learning_rate": 4.9018208270967994e-06, "loss": 1.092, "step": 2943 }, { "epoch": 0.35, "learning_rate": 4.901731615240184e-06, "loss": 0.8612, "step": 2944 }, { "epoch": 0.35, "learning_rate": 4.9016423636827435e-06, "loss": 0.5692, "step": 2945 }, { "epoch": 0.35, "learning_rate": 4.901553072425953e-06, "loss": 0.6182, "step": 2946 }, { "epoch": 0.35, "learning_rate": 4.901463741471288e-06, "loss": 0.7533, "step": 2947 }, { "epoch": 0.35, "learning_rate": 4.901374370820227e-06, "loss": 0.8989, "step": 2948 }, { "epoch": 0.35, "learning_rate": 4.901284960474247e-06, "loss": 1.123, "step": 2949 }, { "epoch": 0.35, "learning_rate": 4.901195510434823e-06, "loss": 1.3676, "step": 2950 }, { "epoch": 0.35, "learning_rate": 4.901106020703437e-06, "loss": 0.575, "step": 2951 }, { "epoch": 0.35, "learning_rate": 4.901016491281566e-06, "loss": 0.7975, "step": 2952 }, { "epoch": 0.35, "learning_rate": 4.9009269221706915e-06, "loss": 0.7585, "step": 2953 }, { "epoch": 0.35, "learning_rate": 4.9008373133722946e-06, "loss": 0.9048, "step": 2954 }, { "epoch": 0.35, "learning_rate": 4.900747664887854e-06, "loss": 1.139, "step": 2955 }, { "epoch": 0.35, "learning_rate": 4.9006579767188535e-06, "loss": 1.1541, "step": 2956 }, { "epoch": 0.35, "learning_rate": 4.900568248866775e-06, "loss": 0.4654, "step": 2957 }, { "epoch": 0.35, "learning_rate": 4.900478481333102e-06, "loss": 0.7063, "step": 2958 }, { "epoch": 0.35, "learning_rate": 4.900388674119319e-06, "loss": 0.6067, "step": 2959 }, { "epoch": 0.35, "learning_rate": 4.9002988272269095e-06, "loss": 0.7205, "step": 2960 }, { "epoch": 0.35, "learning_rate": 4.9002089406573585e-06, "loss": 1.0519, "step": 2961 }, { "epoch": 0.35, "learning_rate": 4.900119014412152e-06, "loss": 1.2128, "step": 2962 }, { "epoch": 0.35, "learning_rate": 4.9000290484927775e-06, "loss": 0.6825, "step": 2963 }, { "epoch": 0.35, "learning_rate": 4.899939042900721e-06, "loss": 0.6332, "step": 2964 }, { "epoch": 0.35, "learning_rate": 4.899848997637472e-06, "loss": 0.9229, "step": 2965 }, { "epoch": 0.35, "learning_rate": 4.899758912704516e-06, "loss": 0.7105, "step": 2966 }, { "epoch": 0.35, "learning_rate": 4.8996687881033445e-06, "loss": 0.9245, "step": 2967 }, { "epoch": 0.35, "learning_rate": 4.899578623835447e-06, "loss": 1.0402, "step": 2968 }, { "epoch": 0.35, "learning_rate": 4.899488419902313e-06, "loss": 0.8157, "step": 2969 }, { "epoch": 0.35, "learning_rate": 4.899398176305434e-06, "loss": 0.5922, "step": 2970 }, { "epoch": 0.35, "learning_rate": 4.899307893046303e-06, "loss": 0.7924, "step": 2971 }, { "epoch": 0.35, "learning_rate": 4.89921757012641e-06, "loss": 0.741, "step": 2972 }, { "epoch": 0.35, "learning_rate": 4.89912720754725e-06, "loss": 0.8398, "step": 2973 }, { "epoch": 0.35, "learning_rate": 4.899036805310315e-06, "loss": 1.1019, "step": 2974 }, { "epoch": 0.35, "learning_rate": 4.8989463634171006e-06, "loss": 1.3372, "step": 2975 }, { "epoch": 0.35, "learning_rate": 4.898855881869102e-06, "loss": 0.575, "step": 2976 }, { "epoch": 0.35, "learning_rate": 4.8987653606678145e-06, "loss": 0.8018, "step": 2977 }, { "epoch": 0.35, "learning_rate": 4.898674799814733e-06, "loss": 0.8243, "step": 2978 }, { "epoch": 0.35, "learning_rate": 4.8985841993113566e-06, "loss": 0.8978, "step": 2979 }, { "epoch": 0.35, "learning_rate": 4.898493559159183e-06, "loss": 1.096, "step": 2980 }, { "epoch": 0.36, "learning_rate": 4.898402879359708e-06, "loss": 1.1219, "step": 2981 }, { "epoch": 0.36, "learning_rate": 4.898312159914434e-06, "loss": 0.5717, "step": 2982 }, { "epoch": 0.36, "learning_rate": 4.898221400824857e-06, "loss": 0.6399, "step": 2983 }, { "epoch": 0.36, "learning_rate": 4.89813060209248e-06, "loss": 0.7726, "step": 2984 }, { "epoch": 0.36, "learning_rate": 4.898039763718803e-06, "loss": 0.7892, "step": 2985 }, { "epoch": 0.36, "learning_rate": 4.897948885705327e-06, "loss": 0.9452, "step": 2986 }, { "epoch": 0.36, "learning_rate": 4.8978579680535555e-06, "loss": 1.092, "step": 2987 }, { "epoch": 0.36, "learning_rate": 4.89776701076499e-06, "loss": 0.5378, "step": 2988 }, { "epoch": 0.36, "learning_rate": 4.897676013841134e-06, "loss": 0.5218, "step": 2989 }, { "epoch": 0.36, "learning_rate": 4.897584977283494e-06, "loss": 0.701, "step": 2990 }, { "epoch": 0.36, "learning_rate": 4.897493901093572e-06, "loss": 0.8649, "step": 2991 }, { "epoch": 0.36, "learning_rate": 4.897402785272874e-06, "loss": 1.0237, "step": 2992 }, { "epoch": 0.36, "learning_rate": 4.897311629822908e-06, "loss": 1.074, "step": 2993 }, { "epoch": 0.36, "learning_rate": 4.8972204347451795e-06, "loss": 0.8935, "step": 2994 }, { "epoch": 0.36, "learning_rate": 4.897129200041196e-06, "loss": 0.5947, "step": 2995 }, { "epoch": 0.36, "learning_rate": 4.897037925712465e-06, "loss": 0.8088, "step": 2996 }, { "epoch": 0.36, "learning_rate": 4.896946611760496e-06, "loss": 0.8588, "step": 2997 }, { "epoch": 0.36, "learning_rate": 4.896855258186799e-06, "loss": 0.9548, "step": 2998 }, { "epoch": 0.36, "learning_rate": 4.896763864992883e-06, "loss": 1.1295, "step": 2999 }, { "epoch": 0.36, "learning_rate": 4.89667243218026e-06, "loss": 1.2805, "step": 3000 }, { "epoch": 0.36, "learning_rate": 4.89658095975044e-06, "loss": 0.6599, "step": 3001 }, { "epoch": 0.36, "learning_rate": 4.8964894477049355e-06, "loss": 0.7025, "step": 3002 }, { "epoch": 0.36, "learning_rate": 4.89639789604526e-06, "loss": 0.5833, "step": 3003 }, { "epoch": 0.36, "learning_rate": 4.896306304772925e-06, "loss": 0.7213, "step": 3004 }, { "epoch": 0.36, "learning_rate": 4.896214673889446e-06, "loss": 0.9707, "step": 3005 }, { "epoch": 0.36, "learning_rate": 4.8961230033963396e-06, "loss": 1.1792, "step": 3006 }, { "epoch": 0.36, "learning_rate": 4.8960312932951166e-06, "loss": 0.7209, "step": 3007 }, { "epoch": 0.36, "learning_rate": 4.895939543587296e-06, "loss": 0.8359, "step": 3008 }, { "epoch": 0.36, "learning_rate": 4.895847754274393e-06, "loss": 0.8574, "step": 3009 }, { "epoch": 0.36, "learning_rate": 4.895755925357926e-06, "loss": 0.85, "step": 3010 }, { "epoch": 0.36, "learning_rate": 4.8956640568394135e-06, "loss": 0.981, "step": 3011 }, { "epoch": 0.36, "learning_rate": 4.895572148720372e-06, "loss": 1.1606, "step": 3012 }, { "epoch": 0.36, "learning_rate": 4.895480201002322e-06, "loss": 0.6805, "step": 3013 }, { "epoch": 0.36, "learning_rate": 4.895388213686784e-06, "loss": 0.6245, "step": 3014 }, { "epoch": 0.36, "learning_rate": 4.8952961867752764e-06, "loss": 0.6494, "step": 3015 }, { "epoch": 0.36, "learning_rate": 4.895204120269323e-06, "loss": 0.7036, "step": 3016 }, { "epoch": 0.36, "learning_rate": 4.895112014170445e-06, "loss": 1.0486, "step": 3017 }, { "epoch": 0.36, "learning_rate": 4.895019868480163e-06, "loss": 1.1147, "step": 3018 }, { "epoch": 0.36, "learning_rate": 4.894927683200002e-06, "loss": 0.7987, "step": 3019 }, { "epoch": 0.36, "learning_rate": 4.894835458331486e-06, "loss": 0.7115, "step": 3020 }, { "epoch": 0.36, "learning_rate": 4.894743193876139e-06, "loss": 0.7745, "step": 3021 }, { "epoch": 0.36, "learning_rate": 4.894650889835486e-06, "loss": 0.7866, "step": 3022 }, { "epoch": 0.36, "learning_rate": 4.8945585462110525e-06, "loss": 0.7706, "step": 3023 }, { "epoch": 0.36, "learning_rate": 4.894466163004365e-06, "loss": 1.0565, "step": 3024 }, { "epoch": 0.36, "learning_rate": 4.8943737402169515e-06, "loss": 1.1268, "step": 3025 }, { "epoch": 0.36, "learning_rate": 4.894281277850339e-06, "loss": 0.6177, "step": 3026 }, { "epoch": 0.36, "learning_rate": 4.894188775906056e-06, "loss": 0.7631, "step": 3027 }, { "epoch": 0.36, "learning_rate": 4.894096234385632e-06, "loss": 0.8078, "step": 3028 }, { "epoch": 0.36, "learning_rate": 4.894003653290596e-06, "loss": 0.9168, "step": 3029 }, { "epoch": 0.36, "learning_rate": 4.8939110326224795e-06, "loss": 0.9889, "step": 3030 }, { "epoch": 0.36, "learning_rate": 4.893818372382811e-06, "loss": 1.3055, "step": 3031 }, { "epoch": 0.36, "learning_rate": 4.8937256725731256e-06, "loss": 0.6567, "step": 3032 }, { "epoch": 0.36, "learning_rate": 4.893632933194954e-06, "loss": 0.826, "step": 3033 }, { "epoch": 0.36, "learning_rate": 4.8935401542498286e-06, "loss": 0.7448, "step": 3034 }, { "epoch": 0.36, "learning_rate": 4.8934473357392835e-06, "loss": 0.8361, "step": 3035 }, { "epoch": 0.36, "learning_rate": 4.8933544776648535e-06, "loss": 0.9724, "step": 3036 }, { "epoch": 0.36, "learning_rate": 4.8932615800280726e-06, "loss": 1.0667, "step": 3037 }, { "epoch": 0.36, "learning_rate": 4.893168642830477e-06, "loss": 0.7094, "step": 3038 }, { "epoch": 0.36, "learning_rate": 4.893075666073603e-06, "loss": 0.7092, "step": 3039 }, { "epoch": 0.36, "learning_rate": 4.892982649758989e-06, "loss": 0.807, "step": 3040 }, { "epoch": 0.36, "learning_rate": 4.892889593888169e-06, "loss": 0.8697, "step": 3041 }, { "epoch": 0.36, "learning_rate": 4.892796498462684e-06, "loss": 0.8992, "step": 3042 }, { "epoch": 0.36, "learning_rate": 4.892703363484072e-06, "loss": 1.1846, "step": 3043 }, { "epoch": 0.36, "learning_rate": 4.892610188953872e-06, "loss": 1.0033, "step": 3044 }, { "epoch": 0.36, "learning_rate": 4.892516974873626e-06, "loss": 0.7317, "step": 3045 }, { "epoch": 0.36, "learning_rate": 4.892423721244873e-06, "loss": 0.6931, "step": 3046 }, { "epoch": 0.36, "learning_rate": 4.8923304280691556e-06, "loss": 0.773, "step": 3047 }, { "epoch": 0.36, "learning_rate": 4.892237095348014e-06, "loss": 0.8518, "step": 3048 }, { "epoch": 0.36, "learning_rate": 4.892143723082995e-06, "loss": 1.1801, "step": 3049 }, { "epoch": 0.36, "learning_rate": 4.892050311275638e-06, "loss": 1.2159, "step": 3050 }, { "epoch": 0.36, "learning_rate": 4.891956859927489e-06, "loss": 0.6246, "step": 3051 }, { "epoch": 0.36, "learning_rate": 4.891863369040092e-06, "loss": 0.7515, "step": 3052 }, { "epoch": 0.36, "learning_rate": 4.891769838614993e-06, "loss": 0.7971, "step": 3053 }, { "epoch": 0.36, "learning_rate": 4.8916762686537375e-06, "loss": 0.9308, "step": 3054 }, { "epoch": 0.36, "learning_rate": 4.891582659157873e-06, "loss": 1.0976, "step": 3055 }, { "epoch": 0.36, "learning_rate": 4.891489010128947e-06, "loss": 1.2822, "step": 3056 }, { "epoch": 0.36, "learning_rate": 4.891395321568506e-06, "loss": 0.6667, "step": 3057 }, { "epoch": 0.36, "learning_rate": 4.8913015934781e-06, "loss": 0.7166, "step": 3058 }, { "epoch": 0.36, "learning_rate": 4.891207825859279e-06, "loss": 0.751, "step": 3059 }, { "epoch": 0.36, "learning_rate": 4.891114018713591e-06, "loss": 0.831, "step": 3060 }, { "epoch": 0.36, "learning_rate": 4.891020172042587e-06, "loss": 0.9641, "step": 3061 }, { "epoch": 0.36, "learning_rate": 4.890926285847821e-06, "loss": 1.0994, "step": 3062 }, { "epoch": 0.36, "learning_rate": 4.8908323601308415e-06, "loss": 0.6443, "step": 3063 }, { "epoch": 0.36, "learning_rate": 4.890738394893203e-06, "loss": 0.7119, "step": 3064 }, { "epoch": 0.37, "learning_rate": 4.8906443901364575e-06, "loss": 0.6873, "step": 3065 }, { "epoch": 0.37, "learning_rate": 4.890550345862159e-06, "loss": 0.8515, "step": 3066 }, { "epoch": 0.37, "learning_rate": 4.890456262071864e-06, "loss": 1.0394, "step": 3067 }, { "epoch": 0.37, "learning_rate": 4.890362138767127e-06, "loss": 1.1689, "step": 3068 }, { "epoch": 0.37, "learning_rate": 4.890267975949502e-06, "loss": 0.8119, "step": 3069 }, { "epoch": 0.37, "learning_rate": 4.890173773620547e-06, "loss": 0.6697, "step": 3070 }, { "epoch": 0.37, "learning_rate": 4.890079531781819e-06, "loss": 0.7092, "step": 3071 }, { "epoch": 0.37, "learning_rate": 4.889985250434876e-06, "loss": 0.7712, "step": 3072 }, { "epoch": 0.37, "learning_rate": 4.889890929581277e-06, "loss": 0.9143, "step": 3073 }, { "epoch": 0.37, "learning_rate": 4.889796569222579e-06, "loss": 1.0325, "step": 3074 }, { "epoch": 0.37, "learning_rate": 4.8897021693603444e-06, "loss": 1.1315, "step": 3075 }, { "epoch": 0.37, "learning_rate": 4.889607729996132e-06, "loss": 0.5881, "step": 3076 }, { "epoch": 0.37, "learning_rate": 4.889513251131503e-06, "loss": 0.692, "step": 3077 }, { "epoch": 0.37, "learning_rate": 4.88941873276802e-06, "loss": 0.7596, "step": 3078 }, { "epoch": 0.37, "learning_rate": 4.889324174907244e-06, "loss": 0.8766, "step": 3079 }, { "epoch": 0.37, "learning_rate": 4.889229577550739e-06, "loss": 0.9281, "step": 3080 }, { "epoch": 0.37, "learning_rate": 4.889134940700069e-06, "loss": 1.0214, "step": 3081 }, { "epoch": 0.37, "learning_rate": 4.889040264356799e-06, "loss": 0.6438, "step": 3082 }, { "epoch": 0.37, "learning_rate": 4.888945548522491e-06, "loss": 0.7672, "step": 3083 }, { "epoch": 0.37, "learning_rate": 4.8888507931987135e-06, "loss": 0.8682, "step": 3084 }, { "epoch": 0.37, "learning_rate": 4.888755998387033e-06, "loss": 0.8085, "step": 3085 }, { "epoch": 0.37, "learning_rate": 4.888661164089014e-06, "loss": 1.0863, "step": 3086 }, { "epoch": 0.37, "learning_rate": 4.888566290306226e-06, "loss": 1.1303, "step": 3087 }, { "epoch": 0.37, "learning_rate": 4.888471377040237e-06, "loss": 0.6297, "step": 3088 }, { "epoch": 0.37, "learning_rate": 4.888376424292616e-06, "loss": 0.7841, "step": 3089 }, { "epoch": 0.37, "learning_rate": 4.888281432064931e-06, "loss": 0.6965, "step": 3090 }, { "epoch": 0.37, "learning_rate": 4.888186400358754e-06, "loss": 0.9282, "step": 3091 }, { "epoch": 0.37, "learning_rate": 4.888091329175656e-06, "loss": 0.9756, "step": 3092 }, { "epoch": 0.37, "learning_rate": 4.887996218517208e-06, "loss": 1.1267, "step": 3093 }, { "epoch": 0.37, "learning_rate": 4.887901068384982e-06, "loss": 0.5687, "step": 3094 }, { "epoch": 0.37, "learning_rate": 4.8878058787805506e-06, "loss": 0.6092, "step": 3095 }, { "epoch": 0.37, "learning_rate": 4.8877106497054884e-06, "loss": 0.6399, "step": 3096 }, { "epoch": 0.37, "learning_rate": 4.887615381161368e-06, "loss": 0.7335, "step": 3097 }, { "epoch": 0.37, "learning_rate": 4.887520073149765e-06, "loss": 0.8643, "step": 3098 }, { "epoch": 0.37, "learning_rate": 4.887424725672256e-06, "loss": 1.1061, "step": 3099 }, { "epoch": 0.37, "learning_rate": 4.887329338730414e-06, "loss": 1.2247, "step": 3100 }, { "epoch": 0.37, "learning_rate": 4.8872339123258194e-06, "loss": 0.6603, "step": 3101 }, { "epoch": 0.37, "learning_rate": 4.887138446460048e-06, "loss": 0.5793, "step": 3102 }, { "epoch": 0.37, "learning_rate": 4.887042941134676e-06, "loss": 0.8936, "step": 3103 }, { "epoch": 0.37, "learning_rate": 4.886947396351286e-06, "loss": 1.0369, "step": 3104 }, { "epoch": 0.37, "learning_rate": 4.886851812111454e-06, "loss": 1.1866, "step": 3105 }, { "epoch": 0.37, "learning_rate": 4.886756188416762e-06, "loss": 1.1368, "step": 3106 }, { "epoch": 0.37, "learning_rate": 4.886660525268789e-06, "loss": 0.5907, "step": 3107 }, { "epoch": 0.37, "learning_rate": 4.886564822669118e-06, "loss": 0.749, "step": 3108 }, { "epoch": 0.37, "learning_rate": 4.88646908061933e-06, "loss": 0.7027, "step": 3109 }, { "epoch": 0.37, "learning_rate": 4.886373299121008e-06, "loss": 0.9341, "step": 3110 }, { "epoch": 0.37, "learning_rate": 4.886277478175735e-06, "loss": 0.9734, "step": 3111 }, { "epoch": 0.37, "learning_rate": 4.886181617785095e-06, "loss": 1.2994, "step": 3112 }, { "epoch": 0.37, "learning_rate": 4.886085717950673e-06, "loss": 0.7697, "step": 3113 }, { "epoch": 0.37, "learning_rate": 4.885989778674054e-06, "loss": 0.5043, "step": 3114 }, { "epoch": 0.37, "learning_rate": 4.885893799956823e-06, "loss": 0.7766, "step": 3115 }, { "epoch": 0.37, "learning_rate": 4.8857977818005675e-06, "loss": 0.6798, "step": 3116 }, { "epoch": 0.37, "learning_rate": 4.885701724206875e-06, "loss": 0.8734, "step": 3117 }, { "epoch": 0.37, "learning_rate": 4.885605627177332e-06, "loss": 1.1812, "step": 3118 }, { "epoch": 0.37, "learning_rate": 4.885509490713528e-06, "loss": 1.1262, "step": 3119 }, { "epoch": 0.37, "learning_rate": 4.885413314817053e-06, "loss": 0.6146, "step": 3120 }, { "epoch": 0.37, "learning_rate": 4.885317099489495e-06, "loss": 0.7836, "step": 3121 }, { "epoch": 0.37, "learning_rate": 4.885220844732445e-06, "loss": 0.7622, "step": 3122 }, { "epoch": 0.37, "learning_rate": 4.8851245505474945e-06, "loss": 0.8376, "step": 3123 }, { "epoch": 0.37, "learning_rate": 4.885028216936235e-06, "loss": 1.0013, "step": 3124 }, { "epoch": 0.37, "learning_rate": 4.884931843900259e-06, "loss": 1.2391, "step": 3125 }, { "epoch": 0.37, "learning_rate": 4.88483543144116e-06, "loss": 0.5564, "step": 3126 }, { "epoch": 0.37, "learning_rate": 4.88473897956053e-06, "loss": 0.7474, "step": 3127 }, { "epoch": 0.37, "learning_rate": 4.884642488259965e-06, "loss": 0.7604, "step": 3128 }, { "epoch": 0.37, "learning_rate": 4.88454595754106e-06, "loss": 1.0668, "step": 3129 }, { "epoch": 0.37, "learning_rate": 4.88444938740541e-06, "loss": 1.0546, "step": 3130 }, { "epoch": 0.37, "learning_rate": 4.884352777854612e-06, "loss": 1.2197, "step": 3131 }, { "epoch": 0.37, "learning_rate": 4.884256128890262e-06, "loss": 0.4714, "step": 3132 }, { "epoch": 0.37, "learning_rate": 4.8841594405139584e-06, "loss": 0.6528, "step": 3133 }, { "epoch": 0.37, "learning_rate": 4.8840627127272996e-06, "loss": 0.7147, "step": 3134 }, { "epoch": 0.37, "learning_rate": 4.883965945531883e-06, "loss": 0.7533, "step": 3135 }, { "epoch": 0.37, "learning_rate": 4.88386913892931e-06, "loss": 1.0352, "step": 3136 }, { "epoch": 0.37, "learning_rate": 4.883772292921181e-06, "loss": 1.0804, "step": 3137 }, { "epoch": 0.37, "learning_rate": 4.8836754075090945e-06, "loss": 0.6971, "step": 3138 }, { "epoch": 0.37, "learning_rate": 4.883578482694654e-06, "loss": 0.6611, "step": 3139 }, { "epoch": 0.37, "learning_rate": 4.883481518479461e-06, "loss": 0.8544, "step": 3140 }, { "epoch": 0.37, "learning_rate": 4.883384514865119e-06, "loss": 0.7551, "step": 3141 }, { "epoch": 0.37, "learning_rate": 4.883287471853231e-06, "loss": 1.0074, "step": 3142 }, { "epoch": 0.37, "learning_rate": 4.8831903894454e-06, "loss": 1.1038, "step": 3143 }, { "epoch": 0.37, "learning_rate": 4.883093267643233e-06, "loss": 0.8861, "step": 3144 }, { "epoch": 0.37, "learning_rate": 4.882996106448334e-06, "loss": 0.5962, "step": 3145 }, { "epoch": 0.37, "learning_rate": 4.88289890586231e-06, "loss": 0.895, "step": 3146 }, { "epoch": 0.37, "learning_rate": 4.882801665886766e-06, "loss": 0.865, "step": 3147 }, { "epoch": 0.37, "learning_rate": 4.882704386523312e-06, "loss": 0.8748, "step": 3148 }, { "epoch": 0.38, "learning_rate": 4.882607067773553e-06, "loss": 1.1743, "step": 3149 }, { "epoch": 0.38, "learning_rate": 4.8825097096391e-06, "loss": 1.036, "step": 3150 }, { "epoch": 0.38, "learning_rate": 4.882412312121562e-06, "loss": 0.6996, "step": 3151 }, { "epoch": 0.38, "learning_rate": 4.882314875222548e-06, "loss": 0.765, "step": 3152 }, { "epoch": 0.38, "learning_rate": 4.8822173989436695e-06, "loss": 0.7743, "step": 3153 }, { "epoch": 0.38, "learning_rate": 4.8821198832865365e-06, "loss": 0.7254, "step": 3154 }, { "epoch": 0.38, "learning_rate": 4.882022328252764e-06, "loss": 1.0903, "step": 3155 }, { "epoch": 0.38, "learning_rate": 4.88192473384396e-06, "loss": 1.1188, "step": 3156 }, { "epoch": 0.38, "learning_rate": 4.881827100061742e-06, "loss": 0.5472, "step": 3157 }, { "epoch": 0.38, "learning_rate": 4.881729426907722e-06, "loss": 0.6946, "step": 3158 }, { "epoch": 0.38, "learning_rate": 4.881631714383515e-06, "loss": 0.7589, "step": 3159 }, { "epoch": 0.38, "learning_rate": 4.881533962490735e-06, "loss": 0.8531, "step": 3160 }, { "epoch": 0.38, "learning_rate": 4.881436171230999e-06, "loss": 0.8897, "step": 3161 }, { "epoch": 0.38, "learning_rate": 4.881338340605924e-06, "loss": 1.1292, "step": 3162 }, { "epoch": 0.38, "learning_rate": 4.881240470617126e-06, "loss": 0.6999, "step": 3163 }, { "epoch": 0.38, "learning_rate": 4.881142561266223e-06, "loss": 0.7072, "step": 3164 }, { "epoch": 0.38, "learning_rate": 4.881044612554834e-06, "loss": 0.8696, "step": 3165 }, { "epoch": 0.38, "learning_rate": 4.880946624484577e-06, "loss": 0.7762, "step": 3166 }, { "epoch": 0.38, "learning_rate": 4.880848597057074e-06, "loss": 0.9399, "step": 3167 }, { "epoch": 0.38, "learning_rate": 4.880750530273942e-06, "loss": 1.2681, "step": 3168 }, { "epoch": 0.38, "learning_rate": 4.880652424136806e-06, "loss": 0.7578, "step": 3169 }, { "epoch": 0.38, "learning_rate": 4.880554278647285e-06, "loss": 0.6569, "step": 3170 }, { "epoch": 0.38, "learning_rate": 4.880456093807001e-06, "loss": 0.688, "step": 3171 }, { "epoch": 0.38, "learning_rate": 4.880357869617579e-06, "loss": 0.7639, "step": 3172 }, { "epoch": 0.38, "learning_rate": 4.880259606080641e-06, "loss": 0.8615, "step": 3173 }, { "epoch": 0.38, "learning_rate": 4.880161303197812e-06, "loss": 1.0505, "step": 3174 }, { "epoch": 0.38, "learning_rate": 4.880062960970718e-06, "loss": 1.023, "step": 3175 }, { "epoch": 0.38, "learning_rate": 4.879964579400983e-06, "loss": 0.6687, "step": 3176 }, { "epoch": 0.38, "learning_rate": 4.879866158490234e-06, "loss": 0.8106, "step": 3177 }, { "epoch": 0.38, "learning_rate": 4.879767698240096e-06, "loss": 0.7868, "step": 3178 }, { "epoch": 0.38, "learning_rate": 4.8796691986522005e-06, "loss": 0.9131, "step": 3179 }, { "epoch": 0.38, "learning_rate": 4.8795706597281724e-06, "loss": 0.978, "step": 3180 }, { "epoch": 0.38, "learning_rate": 4.879472081469642e-06, "loss": 1.21, "step": 3181 }, { "epoch": 0.38, "learning_rate": 4.879373463878239e-06, "loss": 0.7314, "step": 3182 }, { "epoch": 0.38, "learning_rate": 4.8792748069555916e-06, "loss": 0.6681, "step": 3183 }, { "epoch": 0.38, "learning_rate": 4.879176110703333e-06, "loss": 0.8255, "step": 3184 }, { "epoch": 0.38, "learning_rate": 4.8790773751230936e-06, "loss": 0.9681, "step": 3185 }, { "epoch": 0.38, "learning_rate": 4.8789786002165055e-06, "loss": 1.072, "step": 3186 }, { "epoch": 0.38, "learning_rate": 4.878879785985201e-06, "loss": 1.1223, "step": 3187 }, { "epoch": 0.38, "learning_rate": 4.878780932430814e-06, "loss": 0.4315, "step": 3188 }, { "epoch": 0.38, "learning_rate": 4.8786820395549795e-06, "loss": 0.6476, "step": 3189 }, { "epoch": 0.38, "learning_rate": 4.878583107359332e-06, "loss": 0.8209, "step": 3190 }, { "epoch": 0.38, "learning_rate": 4.878484135845504e-06, "loss": 0.7334, "step": 3191 }, { "epoch": 0.38, "learning_rate": 4.878385125015135e-06, "loss": 0.9427, "step": 3192 }, { "epoch": 0.38, "learning_rate": 4.878286074869859e-06, "loss": 1.1688, "step": 3193 }, { "epoch": 0.38, "learning_rate": 4.8781869854113165e-06, "loss": 0.7447, "step": 3194 }, { "epoch": 0.38, "learning_rate": 4.878087856641143e-06, "loss": 0.5799, "step": 3195 }, { "epoch": 0.38, "learning_rate": 4.877988688560977e-06, "loss": 0.6727, "step": 3196 }, { "epoch": 0.38, "learning_rate": 4.877889481172458e-06, "loss": 0.7504, "step": 3197 }, { "epoch": 0.38, "learning_rate": 4.877790234477228e-06, "loss": 0.9225, "step": 3198 }, { "epoch": 0.38, "learning_rate": 4.877690948476925e-06, "loss": 1.1611, "step": 3199 }, { "epoch": 0.38, "learning_rate": 4.8775916231731915e-06, "loss": 1.1643, "step": 3200 }, { "epoch": 0.38, "learning_rate": 4.877492258567669e-06, "loss": 0.6742, "step": 3201 }, { "epoch": 0.38, "learning_rate": 4.877392854661998e-06, "loss": 0.8162, "step": 3202 }, { "epoch": 0.38, "learning_rate": 4.877293411457826e-06, "loss": 0.8534, "step": 3203 }, { "epoch": 0.38, "learning_rate": 4.877193928956794e-06, "loss": 0.9315, "step": 3204 }, { "epoch": 0.38, "learning_rate": 4.877094407160547e-06, "loss": 0.8834, "step": 3205 }, { "epoch": 0.38, "learning_rate": 4.876994846070729e-06, "loss": 1.1718, "step": 3206 }, { "epoch": 0.38, "learning_rate": 4.876895245688987e-06, "loss": 0.7525, "step": 3207 }, { "epoch": 0.38, "learning_rate": 4.876795606016968e-06, "loss": 0.5908, "step": 3208 }, { "epoch": 0.38, "learning_rate": 4.876695927056318e-06, "loss": 0.8033, "step": 3209 }, { "epoch": 0.38, "learning_rate": 4.876596208808685e-06, "loss": 0.7689, "step": 3210 }, { "epoch": 0.38, "learning_rate": 4.876496451275717e-06, "loss": 0.9476, "step": 3211 }, { "epoch": 0.38, "learning_rate": 4.876396654459063e-06, "loss": 1.151, "step": 3212 }, { "epoch": 0.38, "learning_rate": 4.876296818360372e-06, "loss": 0.5751, "step": 3213 }, { "epoch": 0.38, "learning_rate": 4.876196942981296e-06, "loss": 0.5322, "step": 3214 }, { "epoch": 0.38, "learning_rate": 4.8760970283234866e-06, "loss": 0.8322, "step": 3215 }, { "epoch": 0.38, "learning_rate": 4.875997074388592e-06, "loss": 0.742, "step": 3216 }, { "epoch": 0.38, "learning_rate": 4.8758970811782675e-06, "loss": 1.058, "step": 3217 }, { "epoch": 0.38, "learning_rate": 4.875797048694164e-06, "loss": 1.1997, "step": 3218 }, { "epoch": 0.38, "learning_rate": 4.875696976937937e-06, "loss": 0.8238, "step": 3219 }, { "epoch": 0.38, "learning_rate": 4.875596865911239e-06, "loss": 0.6282, "step": 3220 }, { "epoch": 0.38, "learning_rate": 4.875496715615725e-06, "loss": 0.8233, "step": 3221 }, { "epoch": 0.38, "learning_rate": 4.8753965260530525e-06, "loss": 0.819, "step": 3222 }, { "epoch": 0.38, "learning_rate": 4.8752962972248755e-06, "loss": 0.8989, "step": 3223 }, { "epoch": 0.38, "learning_rate": 4.875196029132851e-06, "loss": 1.0273, "step": 3224 }, { "epoch": 0.38, "learning_rate": 4.875095721778637e-06, "loss": 1.1208, "step": 3225 }, { "epoch": 0.38, "learning_rate": 4.8749953751638905e-06, "loss": 0.62, "step": 3226 }, { "epoch": 0.38, "learning_rate": 4.874894989290272e-06, "loss": 0.5931, "step": 3227 }, { "epoch": 0.38, "learning_rate": 4.87479456415944e-06, "loss": 0.7417, "step": 3228 }, { "epoch": 0.38, "learning_rate": 4.874694099773056e-06, "loss": 0.7235, "step": 3229 }, { "epoch": 0.38, "learning_rate": 4.8745935961327774e-06, "loss": 1.0293, "step": 3230 }, { "epoch": 0.38, "learning_rate": 4.874493053240268e-06, "loss": 1.15, "step": 3231 }, { "epoch": 0.38, "learning_rate": 4.874392471097189e-06, "loss": 0.3704, "step": 3232 }, { "epoch": 0.39, "learning_rate": 4.874291849705203e-06, "loss": 0.5848, "step": 3233 }, { "epoch": 0.39, "learning_rate": 4.8741911890659745e-06, "loss": 0.8118, "step": 3234 }, { "epoch": 0.39, "learning_rate": 4.874090489181165e-06, "loss": 0.8185, "step": 3235 }, { "epoch": 0.39, "learning_rate": 4.873989750052441e-06, "loss": 0.9588, "step": 3236 }, { "epoch": 0.39, "learning_rate": 4.873888971681468e-06, "loss": 1.2727, "step": 3237 }, { "epoch": 0.39, "learning_rate": 4.8737881540699105e-06, "loss": 0.7081, "step": 3238 }, { "epoch": 0.39, "learning_rate": 4.873687297219436e-06, "loss": 0.7717, "step": 3239 }, { "epoch": 0.39, "learning_rate": 4.8735864011317114e-06, "loss": 0.6964, "step": 3240 }, { "epoch": 0.39, "learning_rate": 4.873485465808404e-06, "loss": 0.7462, "step": 3241 }, { "epoch": 0.39, "learning_rate": 4.873384491251183e-06, "loss": 0.9904, "step": 3242 }, { "epoch": 0.39, "learning_rate": 4.8732834774617165e-06, "loss": 1.2266, "step": 3243 }, { "epoch": 0.39, "learning_rate": 4.873182424441676e-06, "loss": 0.9972, "step": 3244 }, { "epoch": 0.39, "learning_rate": 4.873081332192731e-06, "loss": 0.6367, "step": 3245 }, { "epoch": 0.39, "learning_rate": 4.872980200716552e-06, "loss": 0.7077, "step": 3246 }, { "epoch": 0.39, "learning_rate": 4.87287903001481e-06, "loss": 0.8412, "step": 3247 }, { "epoch": 0.39, "learning_rate": 4.8727778200891805e-06, "loss": 0.8851, "step": 3248 }, { "epoch": 0.39, "learning_rate": 4.872676570941333e-06, "loss": 1.1295, "step": 3249 }, { "epoch": 0.39, "learning_rate": 4.872575282572944e-06, "loss": 1.1501, "step": 3250 }, { "epoch": 0.39, "learning_rate": 4.872473954985685e-06, "loss": 0.5155, "step": 3251 }, { "epoch": 0.39, "learning_rate": 4.872372588181233e-06, "loss": 0.6624, "step": 3252 }, { "epoch": 0.39, "learning_rate": 4.872271182161264e-06, "loss": 0.7817, "step": 3253 }, { "epoch": 0.39, "learning_rate": 4.872169736927452e-06, "loss": 0.8583, "step": 3254 }, { "epoch": 0.39, "learning_rate": 4.872068252481476e-06, "loss": 0.9564, "step": 3255 }, { "epoch": 0.39, "learning_rate": 4.871966728825013e-06, "loss": 1.2951, "step": 3256 }, { "epoch": 0.39, "learning_rate": 4.8718651659597405e-06, "loss": 0.6242, "step": 3257 }, { "epoch": 0.39, "learning_rate": 4.871763563887339e-06, "loss": 0.7276, "step": 3258 }, { "epoch": 0.39, "learning_rate": 4.871661922609484e-06, "loss": 0.6928, "step": 3259 }, { "epoch": 0.39, "learning_rate": 4.871560242127861e-06, "loss": 0.8779, "step": 3260 }, { "epoch": 0.39, "learning_rate": 4.871458522444147e-06, "loss": 0.9922, "step": 3261 }, { "epoch": 0.39, "learning_rate": 4.871356763560025e-06, "loss": 1.1786, "step": 3262 }, { "epoch": 0.39, "learning_rate": 4.871254965477177e-06, "loss": 0.7061, "step": 3263 }, { "epoch": 0.39, "learning_rate": 4.871153128197284e-06, "loss": 0.6509, "step": 3264 }, { "epoch": 0.39, "learning_rate": 4.871051251722033e-06, "loss": 0.6519, "step": 3265 }, { "epoch": 0.39, "learning_rate": 4.870949336053104e-06, "loss": 0.9322, "step": 3266 }, { "epoch": 0.39, "learning_rate": 4.870847381192185e-06, "loss": 1.0408, "step": 3267 }, { "epoch": 0.39, "learning_rate": 4.870745387140959e-06, "loss": 1.0485, "step": 3268 }, { "epoch": 0.39, "learning_rate": 4.870643353901113e-06, "loss": 0.8544, "step": 3269 }, { "epoch": 0.39, "learning_rate": 4.870541281474334e-06, "loss": 0.6266, "step": 3270 }, { "epoch": 0.39, "learning_rate": 4.870439169862308e-06, "loss": 0.6455, "step": 3271 }, { "epoch": 0.39, "learning_rate": 4.870337019066724e-06, "loss": 0.8442, "step": 3272 }, { "epoch": 0.39, "learning_rate": 4.87023482908927e-06, "loss": 0.948, "step": 3273 }, { "epoch": 0.39, "learning_rate": 4.870132599931635e-06, "loss": 1.096, "step": 3274 }, { "epoch": 0.39, "learning_rate": 4.87003033159551e-06, "loss": 1.1026, "step": 3275 }, { "epoch": 0.39, "learning_rate": 4.869928024082584e-06, "loss": 0.6609, "step": 3276 }, { "epoch": 0.39, "learning_rate": 4.86982567739455e-06, "loss": 0.8346, "step": 3277 }, { "epoch": 0.39, "learning_rate": 4.869723291533097e-06, "loss": 0.874, "step": 3278 }, { "epoch": 0.39, "learning_rate": 4.8696208664999206e-06, "loss": 1.0365, "step": 3279 }, { "epoch": 0.39, "learning_rate": 4.8695184022967124e-06, "loss": 1.0511, "step": 3280 }, { "epoch": 0.39, "learning_rate": 4.869415898925165e-06, "loss": 1.1757, "step": 3281 }, { "epoch": 0.39, "learning_rate": 4.869313356386975e-06, "loss": 0.4586, "step": 3282 }, { "epoch": 0.39, "learning_rate": 4.869210774683837e-06, "loss": 0.6993, "step": 3283 }, { "epoch": 0.39, "learning_rate": 4.869108153817444e-06, "loss": 0.7181, "step": 3284 }, { "epoch": 0.39, "learning_rate": 4.869005493789496e-06, "loss": 0.7768, "step": 3285 }, { "epoch": 0.39, "learning_rate": 4.868902794601689e-06, "loss": 0.9448, "step": 3286 }, { "epoch": 0.39, "learning_rate": 4.868800056255719e-06, "loss": 1.1431, "step": 3287 }, { "epoch": 0.39, "learning_rate": 4.868697278753285e-06, "loss": 0.7197, "step": 3288 }, { "epoch": 0.39, "learning_rate": 4.8685944620960865e-06, "loss": 0.4414, "step": 3289 }, { "epoch": 0.39, "learning_rate": 4.868491606285823e-06, "loss": 0.7829, "step": 3290 }, { "epoch": 0.39, "learning_rate": 4.868388711324195e-06, "loss": 0.7792, "step": 3291 }, { "epoch": 0.39, "learning_rate": 4.868285777212903e-06, "loss": 0.8442, "step": 3292 }, { "epoch": 0.39, "learning_rate": 4.868182803953647e-06, "loss": 1.1962, "step": 3293 }, { "epoch": 0.39, "learning_rate": 4.868079791548131e-06, "loss": 0.9359, "step": 3294 }, { "epoch": 0.39, "learning_rate": 4.867976739998058e-06, "loss": 0.4755, "step": 3295 }, { "epoch": 0.39, "learning_rate": 4.867873649305131e-06, "loss": 0.7036, "step": 3296 }, { "epoch": 0.39, "learning_rate": 4.867770519471053e-06, "loss": 0.842, "step": 3297 }, { "epoch": 0.39, "learning_rate": 4.8676673504975306e-06, "loss": 0.8353, "step": 3298 }, { "epoch": 0.39, "learning_rate": 4.867564142386268e-06, "loss": 1.083, "step": 3299 }, { "epoch": 0.39, "learning_rate": 4.867460895138971e-06, "loss": 1.1482, "step": 3300 }, { "epoch": 0.39, "learning_rate": 4.867357608757346e-06, "loss": 0.558, "step": 3301 }, { "epoch": 0.39, "learning_rate": 4.867254283243102e-06, "loss": 0.6829, "step": 3302 }, { "epoch": 0.39, "learning_rate": 4.8671509185979475e-06, "loss": 0.8259, "step": 3303 }, { "epoch": 0.39, "learning_rate": 4.8670475148235874e-06, "loss": 0.7675, "step": 3304 }, { "epoch": 0.39, "learning_rate": 4.866944071921734e-06, "loss": 1.0072, "step": 3305 }, { "epoch": 0.39, "learning_rate": 4.866840589894097e-06, "loss": 1.079, "step": 3306 }, { "epoch": 0.39, "learning_rate": 4.866737068742386e-06, "loss": 0.5586, "step": 3307 }, { "epoch": 0.39, "learning_rate": 4.866633508468313e-06, "loss": 0.5901, "step": 3308 }, { "epoch": 0.39, "learning_rate": 4.866529909073589e-06, "loss": 0.7128, "step": 3309 }, { "epoch": 0.39, "learning_rate": 4.866426270559927e-06, "loss": 0.8706, "step": 3310 }, { "epoch": 0.39, "learning_rate": 4.86632259292904e-06, "loss": 0.9702, "step": 3311 }, { "epoch": 0.39, "learning_rate": 4.8662188761826436e-06, "loss": 1.2103, "step": 3312 }, { "epoch": 0.39, "learning_rate": 4.866115120322449e-06, "loss": 0.5456, "step": 3313 }, { "epoch": 0.39, "learning_rate": 4.8660113253501736e-06, "loss": 0.7189, "step": 3314 }, { "epoch": 0.39, "learning_rate": 4.865907491267532e-06, "loss": 0.7279, "step": 3315 }, { "epoch": 0.39, "learning_rate": 4.865803618076242e-06, "loss": 0.7927, "step": 3316 }, { "epoch": 0.4, "learning_rate": 4.865699705778018e-06, "loss": 0.8972, "step": 3317 }, { "epoch": 0.4, "learning_rate": 4.865595754374581e-06, "loss": 1.1027, "step": 3318 }, { "epoch": 0.4, "learning_rate": 4.8654917638676465e-06, "loss": 0.729, "step": 3319 }, { "epoch": 0.4, "learning_rate": 4.865387734258936e-06, "loss": 0.6348, "step": 3320 }, { "epoch": 0.4, "learning_rate": 4.865283665550167e-06, "loss": 0.6615, "step": 3321 }, { "epoch": 0.4, "learning_rate": 4.86517955774306e-06, "loss": 0.8086, "step": 3322 }, { "epoch": 0.4, "learning_rate": 4.865075410839337e-06, "loss": 0.8719, "step": 3323 }, { "epoch": 0.4, "learning_rate": 4.86497122484072e-06, "loss": 1.1117, "step": 3324 }, { "epoch": 0.4, "learning_rate": 4.864866999748929e-06, "loss": 1.2791, "step": 3325 }, { "epoch": 0.4, "learning_rate": 4.864762735565688e-06, "loss": 0.5352, "step": 3326 }, { "epoch": 0.4, "learning_rate": 4.864658432292721e-06, "loss": 0.5713, "step": 3327 }, { "epoch": 0.4, "learning_rate": 4.864554089931752e-06, "loss": 0.7283, "step": 3328 }, { "epoch": 0.4, "learning_rate": 4.864449708484505e-06, "loss": 0.8527, "step": 3329 }, { "epoch": 0.4, "learning_rate": 4.864345287952706e-06, "loss": 0.979, "step": 3330 }, { "epoch": 0.4, "learning_rate": 4.864240828338081e-06, "loss": 1.3196, "step": 3331 }, { "epoch": 0.4, "learning_rate": 4.864136329642357e-06, "loss": 0.6281, "step": 3332 }, { "epoch": 0.4, "learning_rate": 4.864031791867261e-06, "loss": 0.7944, "step": 3333 }, { "epoch": 0.4, "learning_rate": 4.863927215014521e-06, "loss": 0.7748, "step": 3334 }, { "epoch": 0.4, "learning_rate": 4.863822599085866e-06, "loss": 0.7282, "step": 3335 }, { "epoch": 0.4, "learning_rate": 4.863717944083024e-06, "loss": 1.0101, "step": 3336 }, { "epoch": 0.4, "learning_rate": 4.863613250007727e-06, "loss": 1.2582, "step": 3337 }, { "epoch": 0.4, "learning_rate": 4.8635085168617045e-06, "loss": 0.5963, "step": 3338 }, { "epoch": 0.4, "learning_rate": 4.863403744646688e-06, "loss": 0.6804, "step": 3339 }, { "epoch": 0.4, "learning_rate": 4.86329893336441e-06, "loss": 0.7593, "step": 3340 }, { "epoch": 0.4, "learning_rate": 4.863194083016601e-06, "loss": 0.6989, "step": 3341 }, { "epoch": 0.4, "learning_rate": 4.863089193604996e-06, "loss": 0.9444, "step": 3342 }, { "epoch": 0.4, "learning_rate": 4.862984265131329e-06, "loss": 1.0605, "step": 3343 }, { "epoch": 0.4, "learning_rate": 4.862879297597333e-06, "loss": 0.5153, "step": 3344 }, { "epoch": 0.4, "learning_rate": 4.8627742910047445e-06, "loss": 0.6762, "step": 3345 }, { "epoch": 0.4, "learning_rate": 4.862669245355298e-06, "loss": 0.6567, "step": 3346 }, { "epoch": 0.4, "learning_rate": 4.862564160650732e-06, "loss": 0.6751, "step": 3347 }, { "epoch": 0.4, "learning_rate": 4.86245903689278e-06, "loss": 0.9178, "step": 3348 }, { "epoch": 0.4, "learning_rate": 4.862353874083184e-06, "loss": 1.0249, "step": 3349 }, { "epoch": 0.4, "learning_rate": 4.8622486722236796e-06, "loss": 0.9838, "step": 3350 }, { "epoch": 0.4, "learning_rate": 4.862143431316005e-06, "loss": 0.6371, "step": 3351 }, { "epoch": 0.4, "learning_rate": 4.862038151361903e-06, "loss": 0.6668, "step": 3352 }, { "epoch": 0.4, "learning_rate": 4.861932832363112e-06, "loss": 0.7182, "step": 3353 }, { "epoch": 0.4, "learning_rate": 4.861827474321373e-06, "loss": 0.8192, "step": 3354 }, { "epoch": 0.4, "learning_rate": 4.861722077238428e-06, "loss": 1.131, "step": 3355 }, { "epoch": 0.4, "learning_rate": 4.861616641116018e-06, "loss": 1.0985, "step": 3356 }, { "epoch": 0.4, "learning_rate": 4.8615111659558865e-06, "loss": 0.5291, "step": 3357 }, { "epoch": 0.4, "learning_rate": 4.861405651759778e-06, "loss": 0.8085, "step": 3358 }, { "epoch": 0.4, "learning_rate": 4.861300098529436e-06, "loss": 0.6678, "step": 3359 }, { "epoch": 0.4, "learning_rate": 4.861194506266606e-06, "loss": 0.8726, "step": 3360 }, { "epoch": 0.4, "learning_rate": 4.861088874973031e-06, "loss": 1.0501, "step": 3361 }, { "epoch": 0.4, "learning_rate": 4.86098320465046e-06, "loss": 1.1269, "step": 3362 }, { "epoch": 0.4, "learning_rate": 4.860877495300638e-06, "loss": 0.6719, "step": 3363 }, { "epoch": 0.4, "learning_rate": 4.860771746925312e-06, "loss": 0.5803, "step": 3364 }, { "epoch": 0.4, "learning_rate": 4.8606659595262325e-06, "loss": 0.7776, "step": 3365 }, { "epoch": 0.4, "learning_rate": 4.860560133105146e-06, "loss": 0.8035, "step": 3366 }, { "epoch": 0.4, "learning_rate": 4.8604542676638025e-06, "loss": 0.8477, "step": 3367 }, { "epoch": 0.4, "learning_rate": 4.860348363203952e-06, "loss": 0.9783, "step": 3368 }, { "epoch": 0.4, "learning_rate": 4.860242419727345e-06, "loss": 0.5723, "step": 3369 }, { "epoch": 0.4, "learning_rate": 4.860136437235733e-06, "loss": 0.6696, "step": 3370 }, { "epoch": 0.4, "learning_rate": 4.8600304157308666e-06, "loss": 0.7538, "step": 3371 }, { "epoch": 0.4, "learning_rate": 4.8599243552145e-06, "loss": 0.8547, "step": 3372 }, { "epoch": 0.4, "learning_rate": 4.8598182556883865e-06, "loss": 0.7916, "step": 3373 }, { "epoch": 0.4, "learning_rate": 4.859712117154279e-06, "loss": 1.038, "step": 3374 }, { "epoch": 0.4, "learning_rate": 4.859605939613932e-06, "loss": 1.1794, "step": 3375 }, { "epoch": 0.4, "learning_rate": 4.8594997230691e-06, "loss": 0.4922, "step": 3376 }, { "epoch": 0.4, "learning_rate": 4.859393467521541e-06, "loss": 0.7693, "step": 3377 }, { "epoch": 0.4, "learning_rate": 4.859287172973009e-06, "loss": 0.7691, "step": 3378 }, { "epoch": 0.4, "learning_rate": 4.859180839425262e-06, "loss": 0.7701, "step": 3379 }, { "epoch": 0.4, "learning_rate": 4.859074466880058e-06, "loss": 1.0449, "step": 3380 }, { "epoch": 0.4, "learning_rate": 4.858968055339156e-06, "loss": 1.1199, "step": 3381 }, { "epoch": 0.4, "learning_rate": 4.858861604804312e-06, "loss": 0.6295, "step": 3382 }, { "epoch": 0.4, "learning_rate": 4.85875511527729e-06, "loss": 0.7396, "step": 3383 }, { "epoch": 0.4, "learning_rate": 4.858648586759847e-06, "loss": 0.7052, "step": 3384 }, { "epoch": 0.4, "learning_rate": 4.858542019253745e-06, "loss": 0.8111, "step": 3385 }, { "epoch": 0.4, "learning_rate": 4.858435412760746e-06, "loss": 0.9481, "step": 3386 }, { "epoch": 0.4, "learning_rate": 4.858328767282611e-06, "loss": 1.1735, "step": 3387 }, { "epoch": 0.4, "learning_rate": 4.858222082821104e-06, "loss": 0.5648, "step": 3388 }, { "epoch": 0.4, "learning_rate": 4.858115359377988e-06, "loss": 0.6969, "step": 3389 }, { "epoch": 0.4, "learning_rate": 4.858008596955027e-06, "loss": 0.8207, "step": 3390 }, { "epoch": 0.4, "learning_rate": 4.857901795553987e-06, "loss": 0.8593, "step": 3391 }, { "epoch": 0.4, "learning_rate": 4.857794955176632e-06, "loss": 0.9995, "step": 3392 }, { "epoch": 0.4, "learning_rate": 4.857688075824728e-06, "loss": 1.2078, "step": 3393 }, { "epoch": 0.4, "learning_rate": 4.857581157500043e-06, "loss": 0.7327, "step": 3394 }, { "epoch": 0.4, "learning_rate": 4.8574742002043426e-06, "loss": 0.8242, "step": 3395 }, { "epoch": 0.4, "learning_rate": 4.857367203939397e-06, "loss": 0.7146, "step": 3396 }, { "epoch": 0.4, "learning_rate": 4.857260168706973e-06, "loss": 0.8121, "step": 3397 }, { "epoch": 0.4, "learning_rate": 4.857153094508842e-06, "loss": 0.9003, "step": 3398 }, { "epoch": 0.4, "learning_rate": 4.857045981346771e-06, "loss": 1.0662, "step": 3399 }, { "epoch": 0.4, "learning_rate": 4.856938829222533e-06, "loss": 1.1275, "step": 3400 }, { "epoch": 0.41, "learning_rate": 4.856831638137897e-06, "loss": 0.619, "step": 3401 }, { "epoch": 0.41, "learning_rate": 4.856724408094637e-06, "loss": 0.6166, "step": 3402 }, { "epoch": 0.41, "learning_rate": 4.856617139094526e-06, "loss": 0.6946, "step": 3403 }, { "epoch": 0.41, "learning_rate": 4.856509831139334e-06, "loss": 0.9364, "step": 3404 }, { "epoch": 0.41, "learning_rate": 4.856402484230838e-06, "loss": 0.8266, "step": 3405 }, { "epoch": 0.41, "learning_rate": 4.856295098370811e-06, "loss": 1.2139, "step": 3406 }, { "epoch": 0.41, "learning_rate": 4.8561876735610296e-06, "loss": 0.6559, "step": 3407 }, { "epoch": 0.41, "learning_rate": 4.856080209803266e-06, "loss": 0.6773, "step": 3408 }, { "epoch": 0.41, "learning_rate": 4.8559727070993e-06, "loss": 0.8096, "step": 3409 }, { "epoch": 0.41, "learning_rate": 4.855865165450907e-06, "loss": 0.9825, "step": 3410 }, { "epoch": 0.41, "learning_rate": 4.855757584859866e-06, "loss": 1.0342, "step": 3411 }, { "epoch": 0.41, "learning_rate": 4.855649965327954e-06, "loss": 1.2651, "step": 3412 }, { "epoch": 0.41, "learning_rate": 4.85554230685695e-06, "loss": 0.5918, "step": 3413 }, { "epoch": 0.41, "learning_rate": 4.855434609448634e-06, "loss": 0.8152, "step": 3414 }, { "epoch": 0.41, "learning_rate": 4.855326873104787e-06, "loss": 0.7853, "step": 3415 }, { "epoch": 0.41, "learning_rate": 4.8552190978271885e-06, "loss": 0.9122, "step": 3416 }, { "epoch": 0.41, "learning_rate": 4.8551112836176215e-06, "loss": 0.9107, "step": 3417 }, { "epoch": 0.41, "learning_rate": 4.855003430477866e-06, "loss": 1.1666, "step": 3418 }, { "epoch": 0.41, "learning_rate": 4.854895538409708e-06, "loss": 0.7139, "step": 3419 }, { "epoch": 0.41, "learning_rate": 4.854787607414928e-06, "loss": 0.5775, "step": 3420 }, { "epoch": 0.41, "learning_rate": 4.854679637495311e-06, "loss": 0.6709, "step": 3421 }, { "epoch": 0.41, "learning_rate": 4.854571628652642e-06, "loss": 0.7595, "step": 3422 }, { "epoch": 0.41, "learning_rate": 4.854463580888707e-06, "loss": 0.8546, "step": 3423 }, { "epoch": 0.41, "learning_rate": 4.854355494205291e-06, "loss": 1.0026, "step": 3424 }, { "epoch": 0.41, "learning_rate": 4.854247368604183e-06, "loss": 0.9838, "step": 3425 }, { "epoch": 0.41, "learning_rate": 4.854139204087166e-06, "loss": 0.6632, "step": 3426 }, { "epoch": 0.41, "learning_rate": 4.8540310006560315e-06, "loss": 0.7616, "step": 3427 }, { "epoch": 0.41, "learning_rate": 4.853922758312567e-06, "loss": 0.8509, "step": 3428 }, { "epoch": 0.41, "learning_rate": 4.853814477058561e-06, "loss": 0.9688, "step": 3429 }, { "epoch": 0.41, "learning_rate": 4.853706156895806e-06, "loss": 1.0605, "step": 3430 }, { "epoch": 0.41, "learning_rate": 4.853597797826089e-06, "loss": 1.1926, "step": 3431 }, { "epoch": 0.41, "learning_rate": 4.853489399851203e-06, "loss": 0.5378, "step": 3432 }, { "epoch": 0.41, "learning_rate": 4.853380962972941e-06, "loss": 0.6853, "step": 3433 }, { "epoch": 0.41, "learning_rate": 4.8532724871930935e-06, "loss": 0.7759, "step": 3434 }, { "epoch": 0.41, "learning_rate": 4.853163972513454e-06, "loss": 0.962, "step": 3435 }, { "epoch": 0.41, "learning_rate": 4.853055418935817e-06, "loss": 1.0587, "step": 3436 }, { "epoch": 0.41, "learning_rate": 4.852946826461976e-06, "loss": 1.2656, "step": 3437 }, { "epoch": 0.41, "learning_rate": 4.852838195093728e-06, "loss": 0.6377, "step": 3438 }, { "epoch": 0.41, "learning_rate": 4.8527295248328656e-06, "loss": 0.6871, "step": 3439 }, { "epoch": 0.41, "learning_rate": 4.852620815681187e-06, "loss": 0.6944, "step": 3440 }, { "epoch": 0.41, "learning_rate": 4.8525120676404894e-06, "loss": 0.7876, "step": 3441 }, { "epoch": 0.41, "learning_rate": 4.85240328071257e-06, "loss": 0.9664, "step": 3442 }, { "epoch": 0.41, "learning_rate": 4.852294454899227e-06, "loss": 1.2405, "step": 3443 }, { "epoch": 0.41, "learning_rate": 4.852185590202258e-06, "loss": 0.8862, "step": 3444 }, { "epoch": 0.41, "learning_rate": 4.852076686623465e-06, "loss": 0.6631, "step": 3445 }, { "epoch": 0.41, "learning_rate": 4.851967744164647e-06, "loss": 0.6399, "step": 3446 }, { "epoch": 0.41, "learning_rate": 4.851858762827605e-06, "loss": 0.7742, "step": 3447 }, { "epoch": 0.41, "learning_rate": 4.85174974261414e-06, "loss": 0.9869, "step": 3448 }, { "epoch": 0.41, "learning_rate": 4.851640683526055e-06, "loss": 1.1252, "step": 3449 }, { "epoch": 0.41, "learning_rate": 4.851531585565151e-06, "loss": 1.1171, "step": 3450 }, { "epoch": 0.41, "learning_rate": 4.851422448733233e-06, "loss": 0.554, "step": 3451 }, { "epoch": 0.41, "learning_rate": 4.851313273032105e-06, "loss": 0.8114, "step": 3452 }, { "epoch": 0.41, "learning_rate": 4.851204058463571e-06, "loss": 0.7425, "step": 3453 }, { "epoch": 0.41, "learning_rate": 4.8510948050294375e-06, "loss": 0.8682, "step": 3454 }, { "epoch": 0.41, "learning_rate": 4.8509855127315084e-06, "loss": 1.0813, "step": 3455 }, { "epoch": 0.41, "learning_rate": 4.850876181571592e-06, "loss": 1.3116, "step": 3456 }, { "epoch": 0.41, "learning_rate": 4.850766811551495e-06, "loss": 0.6445, "step": 3457 }, { "epoch": 0.41, "learning_rate": 4.8506574026730255e-06, "loss": 0.6083, "step": 3458 }, { "epoch": 0.41, "learning_rate": 4.850547954937991e-06, "loss": 0.7698, "step": 3459 }, { "epoch": 0.41, "learning_rate": 4.850438468348203e-06, "loss": 0.9031, "step": 3460 }, { "epoch": 0.41, "learning_rate": 4.850328942905469e-06, "loss": 1.0428, "step": 3461 }, { "epoch": 0.41, "learning_rate": 4.8502193786116005e-06, "loss": 1.0413, "step": 3462 }, { "epoch": 0.41, "learning_rate": 4.850109775468409e-06, "loss": 0.6978, "step": 3463 }, { "epoch": 0.41, "learning_rate": 4.850000133477705e-06, "loss": 0.5949, "step": 3464 }, { "epoch": 0.41, "learning_rate": 4.849890452641302e-06, "loss": 0.6972, "step": 3465 }, { "epoch": 0.41, "learning_rate": 4.849780732961013e-06, "loss": 0.8054, "step": 3466 }, { "epoch": 0.41, "learning_rate": 4.849670974438651e-06, "loss": 0.9189, "step": 3467 }, { "epoch": 0.41, "learning_rate": 4.84956117707603e-06, "loss": 1.2153, "step": 3468 }, { "epoch": 0.41, "learning_rate": 4.849451340874966e-06, "loss": 0.8116, "step": 3469 }, { "epoch": 0.41, "learning_rate": 4.8493414658372746e-06, "loss": 0.6753, "step": 3470 }, { "epoch": 0.41, "learning_rate": 4.849231551964771e-06, "loss": 0.663, "step": 3471 }, { "epoch": 0.41, "learning_rate": 4.849121599259273e-06, "loss": 0.8819, "step": 3472 }, { "epoch": 0.41, "learning_rate": 4.849011607722599e-06, "loss": 0.9456, "step": 3473 }, { "epoch": 0.41, "learning_rate": 4.848901577356565e-06, "loss": 1.1144, "step": 3474 }, { "epoch": 0.41, "learning_rate": 4.84879150816299e-06, "loss": 1.2942, "step": 3475 }, { "epoch": 0.41, "learning_rate": 4.848681400143696e-06, "loss": 0.5576, "step": 3476 }, { "epoch": 0.41, "learning_rate": 4.8485712533005005e-06, "loss": 0.8297, "step": 3477 }, { "epoch": 0.41, "learning_rate": 4.8484610676352245e-06, "loss": 0.7137, "step": 3478 }, { "epoch": 0.41, "learning_rate": 4.848350843149691e-06, "loss": 1.0057, "step": 3479 }, { "epoch": 0.41, "learning_rate": 4.8482405798457196e-06, "loss": 1.0464, "step": 3480 }, { "epoch": 0.41, "learning_rate": 4.848130277725136e-06, "loss": 1.1429, "step": 3481 }, { "epoch": 0.41, "learning_rate": 4.848019936789762e-06, "loss": 0.4823, "step": 3482 }, { "epoch": 0.41, "learning_rate": 4.84790955704142e-06, "loss": 0.8312, "step": 3483 }, { "epoch": 0.41, "learning_rate": 4.847799138481937e-06, "loss": 0.6987, "step": 3484 }, { "epoch": 0.42, "learning_rate": 4.847688681113137e-06, "loss": 0.7271, "step": 3485 }, { "epoch": 0.42, "learning_rate": 4.847578184936846e-06, "loss": 1.0018, "step": 3486 }, { "epoch": 0.42, "learning_rate": 4.847467649954891e-06, "loss": 1.1842, "step": 3487 }, { "epoch": 0.42, "learning_rate": 4.847357076169098e-06, "loss": 0.6, "step": 3488 }, { "epoch": 0.42, "learning_rate": 4.847246463581296e-06, "loss": 0.6207, "step": 3489 }, { "epoch": 0.42, "learning_rate": 4.847135812193313e-06, "loss": 0.7974, "step": 3490 }, { "epoch": 0.42, "learning_rate": 4.8470251220069785e-06, "loss": 0.6221, "step": 3491 }, { "epoch": 0.42, "learning_rate": 4.846914393024121e-06, "loss": 0.9063, "step": 3492 }, { "epoch": 0.42, "learning_rate": 4.846803625246573e-06, "loss": 1.1021, "step": 3493 }, { "epoch": 0.42, "learning_rate": 4.846692818676163e-06, "loss": 0.7889, "step": 3494 }, { "epoch": 0.42, "learning_rate": 4.8465819733147245e-06, "loss": 0.5906, "step": 3495 }, { "epoch": 0.42, "learning_rate": 4.846471089164089e-06, "loss": 0.7846, "step": 3496 }, { "epoch": 0.42, "learning_rate": 4.846360166226089e-06, "loss": 0.789, "step": 3497 }, { "epoch": 0.42, "learning_rate": 4.846249204502559e-06, "loss": 0.8729, "step": 3498 }, { "epoch": 0.42, "learning_rate": 4.846138203995332e-06, "loss": 1.0502, "step": 3499 }, { "epoch": 0.42, "learning_rate": 4.846027164706244e-06, "loss": 1.2299, "step": 3500 }, { "epoch": 0.42, "learning_rate": 4.84591608663713e-06, "loss": 0.6472, "step": 3501 }, { "epoch": 0.42, "learning_rate": 4.845804969789827e-06, "loss": 0.6096, "step": 3502 }, { "epoch": 0.42, "learning_rate": 4.84569381416617e-06, "loss": 0.8249, "step": 3503 }, { "epoch": 0.42, "learning_rate": 4.845582619767999e-06, "loss": 0.8342, "step": 3504 }, { "epoch": 0.42, "learning_rate": 4.8454713865971495e-06, "loss": 0.9538, "step": 3505 }, { "epoch": 0.42, "learning_rate": 4.8453601146554605e-06, "loss": 1.1713, "step": 3506 }, { "epoch": 0.42, "learning_rate": 4.845248803944773e-06, "loss": 0.6929, "step": 3507 }, { "epoch": 0.42, "learning_rate": 4.845137454466926e-06, "loss": 0.7146, "step": 3508 }, { "epoch": 0.42, "learning_rate": 4.8450260662237595e-06, "loss": 0.7541, "step": 3509 }, { "epoch": 0.42, "learning_rate": 4.844914639217115e-06, "loss": 0.8486, "step": 3510 }, { "epoch": 0.42, "learning_rate": 4.844803173448836e-06, "loss": 1.0773, "step": 3511 }, { "epoch": 0.42, "learning_rate": 4.844691668920763e-06, "loss": 1.0594, "step": 3512 }, { "epoch": 0.42, "learning_rate": 4.84458012563474e-06, "loss": 0.6809, "step": 3513 }, { "epoch": 0.42, "learning_rate": 4.844468543592611e-06, "loss": 0.7355, "step": 3514 }, { "epoch": 0.42, "learning_rate": 4.84435692279622e-06, "loss": 0.8633, "step": 3515 }, { "epoch": 0.42, "learning_rate": 4.844245263247414e-06, "loss": 0.9111, "step": 3516 }, { "epoch": 0.42, "learning_rate": 4.844133564948035e-06, "loss": 0.908, "step": 3517 }, { "epoch": 0.42, "learning_rate": 4.844021827899932e-06, "loss": 1.0732, "step": 3518 }, { "epoch": 0.42, "learning_rate": 4.843910052104952e-06, "loss": 0.8651, "step": 3519 }, { "epoch": 0.42, "learning_rate": 4.843798237564942e-06, "loss": 0.6162, "step": 3520 }, { "epoch": 0.42, "learning_rate": 4.84368638428175e-06, "loss": 0.7826, "step": 3521 }, { "epoch": 0.42, "learning_rate": 4.843574492257226e-06, "loss": 0.7719, "step": 3522 }, { "epoch": 0.42, "learning_rate": 4.843462561493218e-06, "loss": 1.0146, "step": 3523 }, { "epoch": 0.42, "learning_rate": 4.843350591991578e-06, "loss": 0.9393, "step": 3524 }, { "epoch": 0.42, "learning_rate": 4.843238583754156e-06, "loss": 1.1913, "step": 3525 }, { "epoch": 0.42, "learning_rate": 4.843126536782804e-06, "loss": 0.6656, "step": 3526 }, { "epoch": 0.42, "learning_rate": 4.843014451079374e-06, "loss": 0.8082, "step": 3527 }, { "epoch": 0.42, "learning_rate": 4.842902326645718e-06, "loss": 0.7863, "step": 3528 }, { "epoch": 0.42, "learning_rate": 4.842790163483689e-06, "loss": 0.8964, "step": 3529 }, { "epoch": 0.42, "learning_rate": 4.842677961595143e-06, "loss": 1.0869, "step": 3530 }, { "epoch": 0.42, "learning_rate": 4.842565720981934e-06, "loss": 1.0785, "step": 3531 }, { "epoch": 0.42, "learning_rate": 4.842453441645917e-06, "loss": 0.5579, "step": 3532 }, { "epoch": 0.42, "learning_rate": 4.8423411235889475e-06, "loss": 0.7347, "step": 3533 }, { "epoch": 0.42, "learning_rate": 4.8422287668128835e-06, "loss": 0.7935, "step": 3534 }, { "epoch": 0.42, "learning_rate": 4.84211637131958e-06, "loss": 0.6624, "step": 3535 }, { "epoch": 0.42, "learning_rate": 4.8420039371108975e-06, "loss": 1.0118, "step": 3536 }, { "epoch": 0.42, "learning_rate": 4.841891464188692e-06, "loss": 1.3627, "step": 3537 }, { "epoch": 0.42, "learning_rate": 4.841778952554827e-06, "loss": 0.5925, "step": 3538 }, { "epoch": 0.42, "learning_rate": 4.841666402211157e-06, "loss": 0.7104, "step": 3539 }, { "epoch": 0.42, "learning_rate": 4.841553813159546e-06, "loss": 0.7436, "step": 3540 }, { "epoch": 0.42, "learning_rate": 4.8414411854018525e-06, "loss": 0.8306, "step": 3541 }, { "epoch": 0.42, "learning_rate": 4.8413285189399415e-06, "loss": 1.0201, "step": 3542 }, { "epoch": 0.42, "learning_rate": 4.841215813775673e-06, "loss": 1.1314, "step": 3543 }, { "epoch": 0.42, "learning_rate": 4.84110306991091e-06, "loss": 0.8155, "step": 3544 }, { "epoch": 0.42, "learning_rate": 4.840990287347518e-06, "loss": 0.7073, "step": 3545 }, { "epoch": 0.42, "learning_rate": 4.8408774660873595e-06, "loss": 0.6002, "step": 3546 }, { "epoch": 0.42, "learning_rate": 4.8407646061323e-06, "loss": 1.0374, "step": 3547 }, { "epoch": 0.42, "learning_rate": 4.840651707484205e-06, "loss": 0.74, "step": 3548 }, { "epoch": 0.42, "learning_rate": 4.840538770144941e-06, "loss": 1.1354, "step": 3549 }, { "epoch": 0.42, "learning_rate": 4.840425794116376e-06, "loss": 1.2257, "step": 3550 }, { "epoch": 0.42, "learning_rate": 4.840312779400375e-06, "loss": 0.5703, "step": 3551 }, { "epoch": 0.42, "learning_rate": 4.840199725998807e-06, "loss": 0.7844, "step": 3552 }, { "epoch": 0.42, "learning_rate": 4.8400866339135426e-06, "loss": 0.9882, "step": 3553 }, { "epoch": 0.42, "learning_rate": 4.839973503146448e-06, "loss": 0.7462, "step": 3554 }, { "epoch": 0.42, "learning_rate": 4.839860333699396e-06, "loss": 0.9943, "step": 3555 }, { "epoch": 0.42, "learning_rate": 4.839747125574256e-06, "loss": 1.1959, "step": 3556 }, { "epoch": 0.42, "learning_rate": 4.8396338787729e-06, "loss": 0.7811, "step": 3557 }, { "epoch": 0.42, "learning_rate": 4.8395205932972e-06, "loss": 0.6835, "step": 3558 }, { "epoch": 0.42, "learning_rate": 4.839407269149028e-06, "loss": 0.873, "step": 3559 }, { "epoch": 0.42, "learning_rate": 4.839293906330257e-06, "loss": 0.8513, "step": 3560 }, { "epoch": 0.42, "learning_rate": 4.839180504842763e-06, "loss": 1.0928, "step": 3561 }, { "epoch": 0.42, "learning_rate": 4.8390670646884175e-06, "loss": 1.2015, "step": 3562 }, { "epoch": 0.42, "learning_rate": 4.8389535858690976e-06, "loss": 0.7747, "step": 3563 }, { "epoch": 0.42, "learning_rate": 4.838840068386678e-06, "loss": 0.562, "step": 3564 }, { "epoch": 0.42, "learning_rate": 4.838726512243037e-06, "loss": 0.848, "step": 3565 }, { "epoch": 0.42, "learning_rate": 4.83861291744005e-06, "loss": 0.6619, "step": 3566 }, { "epoch": 0.42, "learning_rate": 4.838499283979595e-06, "loss": 0.9406, "step": 3567 }, { "epoch": 0.42, "learning_rate": 4.838385611863551e-06, "loss": 1.1951, "step": 3568 }, { "epoch": 0.43, "learning_rate": 4.8382719010937966e-06, "loss": 0.7731, "step": 3569 }, { "epoch": 0.43, "learning_rate": 4.838158151672211e-06, "loss": 0.5352, "step": 3570 }, { "epoch": 0.43, "learning_rate": 4.838044363600676e-06, "loss": 0.7085, "step": 3571 }, { "epoch": 0.43, "learning_rate": 4.837930536881071e-06, "loss": 0.6962, "step": 3572 }, { "epoch": 0.43, "learning_rate": 4.837816671515278e-06, "loss": 0.8768, "step": 3573 }, { "epoch": 0.43, "learning_rate": 4.837702767505179e-06, "loss": 1.1749, "step": 3574 }, { "epoch": 0.43, "learning_rate": 4.8375888248526574e-06, "loss": 1.0938, "step": 3575 }, { "epoch": 0.43, "learning_rate": 4.837474843559596e-06, "loss": 0.5314, "step": 3576 }, { "epoch": 0.43, "learning_rate": 4.837360823627879e-06, "loss": 0.6386, "step": 3577 }, { "epoch": 0.43, "learning_rate": 4.837246765059392e-06, "loss": 0.8394, "step": 3578 }, { "epoch": 0.43, "learning_rate": 4.837132667856021e-06, "loss": 0.7941, "step": 3579 }, { "epoch": 0.43, "learning_rate": 4.837018532019649e-06, "loss": 0.9886, "step": 3580 }, { "epoch": 0.43, "learning_rate": 4.836904357552165e-06, "loss": 1.0408, "step": 3581 }, { "epoch": 0.43, "learning_rate": 4.836790144455456e-06, "loss": 0.6164, "step": 3582 }, { "epoch": 0.43, "learning_rate": 4.83667589273141e-06, "loss": 0.7656, "step": 3583 }, { "epoch": 0.43, "learning_rate": 4.836561602381916e-06, "loss": 0.7715, "step": 3584 }, { "epoch": 0.43, "learning_rate": 4.836447273408862e-06, "loss": 0.9501, "step": 3585 }, { "epoch": 0.43, "learning_rate": 4.836332905814139e-06, "loss": 0.9746, "step": 3586 }, { "epoch": 0.43, "learning_rate": 4.8362184995996365e-06, "loss": 1.1727, "step": 3587 }, { "epoch": 0.43, "learning_rate": 4.8361040547672465e-06, "loss": 0.7655, "step": 3588 }, { "epoch": 0.43, "learning_rate": 4.83598957131886e-06, "loss": 0.6782, "step": 3589 }, { "epoch": 0.43, "learning_rate": 4.835875049256371e-06, "loss": 0.6918, "step": 3590 }, { "epoch": 0.43, "learning_rate": 4.83576048858167e-06, "loss": 0.7499, "step": 3591 }, { "epoch": 0.43, "learning_rate": 4.835645889296653e-06, "loss": 0.9419, "step": 3592 }, { "epoch": 0.43, "learning_rate": 4.835531251403214e-06, "loss": 0.9806, "step": 3593 }, { "epoch": 0.43, "learning_rate": 4.835416574903246e-06, "loss": 0.9007, "step": 3594 }, { "epoch": 0.43, "learning_rate": 4.835301859798646e-06, "loss": 0.6455, "step": 3595 }, { "epoch": 0.43, "learning_rate": 4.835187106091312e-06, "loss": 0.7299, "step": 3596 }, { "epoch": 0.43, "learning_rate": 4.835072313783138e-06, "loss": 0.7825, "step": 3597 }, { "epoch": 0.43, "learning_rate": 4.8349574828760235e-06, "loss": 0.856, "step": 3598 }, { "epoch": 0.43, "learning_rate": 4.834842613371865e-06, "loss": 1.089, "step": 3599 }, { "epoch": 0.43, "learning_rate": 4.834727705272563e-06, "loss": 1.2456, "step": 3600 }, { "epoch": 0.43, "learning_rate": 4.834612758580015e-06, "loss": 0.6425, "step": 3601 }, { "epoch": 0.43, "learning_rate": 4.834497773296123e-06, "loss": 0.79, "step": 3602 }, { "epoch": 0.43, "learning_rate": 4.834382749422787e-06, "loss": 0.7223, "step": 3603 }, { "epoch": 0.43, "learning_rate": 4.834267686961909e-06, "loss": 0.8521, "step": 3604 }, { "epoch": 0.43, "learning_rate": 4.834152585915389e-06, "loss": 1.0649, "step": 3605 }, { "epoch": 0.43, "learning_rate": 4.834037446285131e-06, "loss": 1.1112, "step": 3606 }, { "epoch": 0.43, "learning_rate": 4.833922268073039e-06, "loss": 0.6441, "step": 3607 }, { "epoch": 0.43, "learning_rate": 4.8338070512810155e-06, "loss": 0.6403, "step": 3608 }, { "epoch": 0.43, "learning_rate": 4.833691795910965e-06, "loss": 0.7244, "step": 3609 }, { "epoch": 0.43, "learning_rate": 4.833576501964794e-06, "loss": 0.8208, "step": 3610 }, { "epoch": 0.43, "learning_rate": 4.833461169444408e-06, "loss": 0.8951, "step": 3611 }, { "epoch": 0.43, "learning_rate": 4.833345798351713e-06, "loss": 1.2728, "step": 3612 }, { "epoch": 0.43, "learning_rate": 4.833230388688615e-06, "loss": 0.5308, "step": 3613 }, { "epoch": 0.43, "learning_rate": 4.833114940457024e-06, "loss": 0.6235, "step": 3614 }, { "epoch": 0.43, "learning_rate": 4.832999453658847e-06, "loss": 0.7118, "step": 3615 }, { "epoch": 0.43, "learning_rate": 4.832883928295994e-06, "loss": 0.8368, "step": 3616 }, { "epoch": 0.43, "learning_rate": 4.8327683643703725e-06, "loss": 0.9861, "step": 3617 }, { "epoch": 0.43, "learning_rate": 4.832652761883896e-06, "loss": 1.1038, "step": 3618 }, { "epoch": 0.43, "learning_rate": 4.832537120838472e-06, "loss": 0.7127, "step": 3619 }, { "epoch": 0.43, "learning_rate": 4.8324214412360146e-06, "loss": 0.6377, "step": 3620 }, { "epoch": 0.43, "learning_rate": 4.832305723078434e-06, "loss": 0.5999, "step": 3621 }, { "epoch": 0.43, "learning_rate": 4.832189966367646e-06, "loss": 0.7584, "step": 3622 }, { "epoch": 0.43, "learning_rate": 4.83207417110556e-06, "loss": 0.8759, "step": 3623 }, { "epoch": 0.43, "learning_rate": 4.831958337294094e-06, "loss": 1.0364, "step": 3624 }, { "epoch": 0.43, "learning_rate": 4.83184246493516e-06, "loss": 1.2248, "step": 3625 }, { "epoch": 0.43, "learning_rate": 4.831726554030675e-06, "loss": 0.64, "step": 3626 }, { "epoch": 0.43, "learning_rate": 4.831610604582553e-06, "loss": 0.7042, "step": 3627 }, { "epoch": 0.43, "learning_rate": 4.831494616592713e-06, "loss": 0.8192, "step": 3628 }, { "epoch": 0.43, "learning_rate": 4.831378590063071e-06, "loss": 0.9204, "step": 3629 }, { "epoch": 0.43, "learning_rate": 4.831262524995547e-06, "loss": 1.0476, "step": 3630 }, { "epoch": 0.43, "learning_rate": 4.831146421392057e-06, "loss": 1.2146, "step": 3631 }, { "epoch": 0.43, "learning_rate": 4.83103027925452e-06, "loss": 0.6449, "step": 3632 }, { "epoch": 0.43, "learning_rate": 4.830914098584858e-06, "loss": 0.619, "step": 3633 }, { "epoch": 0.43, "learning_rate": 4.83079787938499e-06, "loss": 0.835, "step": 3634 }, { "epoch": 0.43, "learning_rate": 4.830681621656837e-06, "loss": 0.796, "step": 3635 }, { "epoch": 0.43, "learning_rate": 4.8305653254023215e-06, "loss": 0.9095, "step": 3636 }, { "epoch": 0.43, "learning_rate": 4.830448990623366e-06, "loss": 1.2375, "step": 3637 }, { "epoch": 0.43, "learning_rate": 4.830332617321893e-06, "loss": 0.6657, "step": 3638 }, { "epoch": 0.43, "learning_rate": 4.830216205499826e-06, "loss": 0.6846, "step": 3639 }, { "epoch": 0.43, "learning_rate": 4.83009975515909e-06, "loss": 0.7305, "step": 3640 }, { "epoch": 0.43, "learning_rate": 4.829983266301609e-06, "loss": 0.7849, "step": 3641 }, { "epoch": 0.43, "learning_rate": 4.82986673892931e-06, "loss": 0.8853, "step": 3642 }, { "epoch": 0.43, "learning_rate": 4.829750173044118e-06, "loss": 1.1175, "step": 3643 }, { "epoch": 0.43, "learning_rate": 4.829633568647959e-06, "loss": 0.8446, "step": 3644 }, { "epoch": 0.43, "learning_rate": 4.829516925742764e-06, "loss": 0.6226, "step": 3645 }, { "epoch": 0.43, "learning_rate": 4.829400244330457e-06, "loss": 0.6639, "step": 3646 }, { "epoch": 0.43, "learning_rate": 4.829283524412969e-06, "loss": 0.719, "step": 3647 }, { "epoch": 0.43, "learning_rate": 4.829166765992228e-06, "loss": 1.0001, "step": 3648 }, { "epoch": 0.43, "learning_rate": 4.829049969070166e-06, "loss": 1.2556, "step": 3649 }, { "epoch": 0.43, "learning_rate": 4.828933133648712e-06, "loss": 1.1207, "step": 3650 }, { "epoch": 0.43, "learning_rate": 4.8288162597297986e-06, "loss": 0.5667, "step": 3651 }, { "epoch": 0.43, "learning_rate": 4.828699347315357e-06, "loss": 0.7621, "step": 3652 }, { "epoch": 0.44, "learning_rate": 4.8285823964073186e-06, "loss": 0.6834, "step": 3653 }, { "epoch": 0.44, "learning_rate": 4.828465407007619e-06, "loss": 0.8654, "step": 3654 }, { "epoch": 0.44, "learning_rate": 4.82834837911819e-06, "loss": 1.0318, "step": 3655 }, { "epoch": 0.44, "learning_rate": 4.828231312740967e-06, "loss": 1.1028, "step": 3656 }, { "epoch": 0.44, "learning_rate": 4.828114207877886e-06, "loss": 0.6156, "step": 3657 }, { "epoch": 0.44, "learning_rate": 4.827997064530881e-06, "loss": 0.6655, "step": 3658 }, { "epoch": 0.44, "learning_rate": 4.827879882701889e-06, "loss": 0.7634, "step": 3659 }, { "epoch": 0.44, "learning_rate": 4.827762662392848e-06, "loss": 0.7475, "step": 3660 }, { "epoch": 0.44, "learning_rate": 4.827645403605694e-06, "loss": 1.0707, "step": 3661 }, { "epoch": 0.44, "learning_rate": 4.827528106342367e-06, "loss": 1.1764, "step": 3662 }, { "epoch": 0.44, "learning_rate": 4.827410770604805e-06, "loss": 0.6921, "step": 3663 }, { "epoch": 0.44, "learning_rate": 4.8272933963949474e-06, "loss": 0.7209, "step": 3664 }, { "epoch": 0.44, "learning_rate": 4.827175983714735e-06, "loss": 0.733, "step": 3665 }, { "epoch": 0.44, "learning_rate": 4.8270585325661075e-06, "loss": 0.8495, "step": 3666 }, { "epoch": 0.44, "learning_rate": 4.826941042951008e-06, "loss": 0.9819, "step": 3667 }, { "epoch": 0.44, "learning_rate": 4.826823514871377e-06, "loss": 1.1256, "step": 3668 }, { "epoch": 0.44, "learning_rate": 4.8267059483291586e-06, "loss": 0.867, "step": 3669 }, { "epoch": 0.44, "learning_rate": 4.826588343326296e-06, "loss": 0.778, "step": 3670 }, { "epoch": 0.44, "learning_rate": 4.8264706998647325e-06, "loss": 0.8775, "step": 3671 }, { "epoch": 0.44, "learning_rate": 4.826353017946414e-06, "loss": 0.7485, "step": 3672 }, { "epoch": 0.44, "learning_rate": 4.826235297573284e-06, "loss": 0.9367, "step": 3673 }, { "epoch": 0.44, "learning_rate": 4.826117538747289e-06, "loss": 1.1037, "step": 3674 }, { "epoch": 0.44, "learning_rate": 4.825999741470377e-06, "loss": 1.2308, "step": 3675 }, { "epoch": 0.44, "learning_rate": 4.8258819057444926e-06, "loss": 0.6154, "step": 3676 }, { "epoch": 0.44, "learning_rate": 4.8257640315715856e-06, "loss": 0.7385, "step": 3677 }, { "epoch": 0.44, "learning_rate": 4.825646118953605e-06, "loss": 0.7407, "step": 3678 }, { "epoch": 0.44, "learning_rate": 4.825528167892498e-06, "loss": 0.9318, "step": 3679 }, { "epoch": 0.44, "learning_rate": 4.825410178390215e-06, "loss": 0.9061, "step": 3680 }, { "epoch": 0.44, "learning_rate": 4.825292150448708e-06, "loss": 1.2201, "step": 3681 }, { "epoch": 0.44, "learning_rate": 4.8251740840699255e-06, "loss": 0.6331, "step": 3682 }, { "epoch": 0.44, "learning_rate": 4.82505597925582e-06, "loss": 0.7808, "step": 3683 }, { "epoch": 0.44, "learning_rate": 4.824937836008346e-06, "loss": 0.6536, "step": 3684 }, { "epoch": 0.44, "learning_rate": 4.824819654329452e-06, "loss": 1.0396, "step": 3685 }, { "epoch": 0.44, "learning_rate": 4.8247014342210956e-06, "loss": 0.9227, "step": 3686 }, { "epoch": 0.44, "learning_rate": 4.824583175685229e-06, "loss": 1.1401, "step": 3687 }, { "epoch": 0.44, "learning_rate": 4.824464878723808e-06, "loss": 0.7312, "step": 3688 }, { "epoch": 0.44, "learning_rate": 4.8243465433387874e-06, "loss": 0.5641, "step": 3689 }, { "epoch": 0.44, "learning_rate": 4.8242281695321225e-06, "loss": 0.7684, "step": 3690 }, { "epoch": 0.44, "learning_rate": 4.824109757305772e-06, "loss": 0.9648, "step": 3691 }, { "epoch": 0.44, "learning_rate": 4.8239913066616915e-06, "loss": 1.0465, "step": 3692 }, { "epoch": 0.44, "learning_rate": 4.823872817601839e-06, "loss": 1.2303, "step": 3693 }, { "epoch": 0.44, "learning_rate": 4.823754290128175e-06, "loss": 0.7684, "step": 3694 }, { "epoch": 0.44, "learning_rate": 4.823635724242658e-06, "loss": 0.6758, "step": 3695 }, { "epoch": 0.44, "learning_rate": 4.823517119947247e-06, "loss": 0.7893, "step": 3696 }, { "epoch": 0.44, "learning_rate": 4.823398477243902e-06, "loss": 0.8334, "step": 3697 }, { "epoch": 0.44, "learning_rate": 4.823279796134587e-06, "loss": 0.8482, "step": 3698 }, { "epoch": 0.44, "learning_rate": 4.82316107662126e-06, "loss": 1.1896, "step": 3699 }, { "epoch": 0.44, "learning_rate": 4.823042318705887e-06, "loss": 1.3222, "step": 3700 }, { "epoch": 0.44, "learning_rate": 4.822923522390429e-06, "loss": 0.5499, "step": 3701 }, { "epoch": 0.44, "learning_rate": 4.822804687676851e-06, "loss": 0.6965, "step": 3702 }, { "epoch": 0.44, "learning_rate": 4.822685814567116e-06, "loss": 0.841, "step": 3703 }, { "epoch": 0.44, "learning_rate": 4.82256690306319e-06, "loss": 1.0364, "step": 3704 }, { "epoch": 0.44, "learning_rate": 4.822447953167038e-06, "loss": 0.9958, "step": 3705 }, { "epoch": 0.44, "learning_rate": 4.822328964880626e-06, "loss": 1.2241, "step": 3706 }, { "epoch": 0.44, "learning_rate": 4.822209938205921e-06, "loss": 0.5657, "step": 3707 }, { "epoch": 0.44, "learning_rate": 4.822090873144892e-06, "loss": 0.7548, "step": 3708 }, { "epoch": 0.44, "learning_rate": 4.821971769699505e-06, "loss": 0.6504, "step": 3709 }, { "epoch": 0.44, "learning_rate": 4.821852627871731e-06, "loss": 0.7724, "step": 3710 }, { "epoch": 0.44, "learning_rate": 4.8217334476635375e-06, "loss": 1.093, "step": 3711 }, { "epoch": 0.44, "learning_rate": 4.821614229076895e-06, "loss": 1.1905, "step": 3712 }, { "epoch": 0.44, "learning_rate": 4.821494972113775e-06, "loss": 0.6613, "step": 3713 }, { "epoch": 0.44, "learning_rate": 4.8213756767761485e-06, "loss": 0.5322, "step": 3714 }, { "epoch": 0.44, "learning_rate": 4.821256343065987e-06, "loss": 0.791, "step": 3715 }, { "epoch": 0.44, "learning_rate": 4.821136970985263e-06, "loss": 0.8989, "step": 3716 }, { "epoch": 0.44, "learning_rate": 4.821017560535951e-06, "loss": 0.9679, "step": 3717 }, { "epoch": 0.44, "learning_rate": 4.820898111720024e-06, "loss": 1.2863, "step": 3718 }, { "epoch": 0.44, "learning_rate": 4.820778624539456e-06, "loss": 0.6891, "step": 3719 }, { "epoch": 0.44, "learning_rate": 4.8206590989962225e-06, "loss": 0.5651, "step": 3720 }, { "epoch": 0.44, "learning_rate": 4.820539535092299e-06, "loss": 0.6129, "step": 3721 }, { "epoch": 0.44, "learning_rate": 4.820419932829663e-06, "loss": 0.6904, "step": 3722 }, { "epoch": 0.44, "learning_rate": 4.82030029221029e-06, "loss": 0.923, "step": 3723 }, { "epoch": 0.44, "learning_rate": 4.8201806132361595e-06, "loss": 1.0925, "step": 3724 }, { "epoch": 0.44, "learning_rate": 4.820060895909248e-06, "loss": 0.9559, "step": 3725 }, { "epoch": 0.44, "learning_rate": 4.8199411402315356e-06, "loss": 0.4868, "step": 3726 }, { "epoch": 0.44, "learning_rate": 4.819821346205002e-06, "loss": 0.8107, "step": 3727 }, { "epoch": 0.44, "learning_rate": 4.819701513831626e-06, "loss": 0.7663, "step": 3728 }, { "epoch": 0.44, "learning_rate": 4.819581643113389e-06, "loss": 0.8478, "step": 3729 }, { "epoch": 0.44, "learning_rate": 4.8194617340522734e-06, "loss": 1.1064, "step": 3730 }, { "epoch": 0.44, "learning_rate": 4.819341786650262e-06, "loss": 1.2731, "step": 3731 }, { "epoch": 0.44, "learning_rate": 4.819221800909334e-06, "loss": 0.5314, "step": 3732 }, { "epoch": 0.44, "learning_rate": 4.819101776831476e-06, "loss": 0.7421, "step": 3733 }, { "epoch": 0.44, "learning_rate": 4.818981714418671e-06, "loss": 0.7148, "step": 3734 }, { "epoch": 0.44, "learning_rate": 4.8188616136729034e-06, "loss": 0.8539, "step": 3735 }, { "epoch": 0.44, "learning_rate": 4.81874147459616e-06, "loss": 1.0686, "step": 3736 }, { "epoch": 0.45, "learning_rate": 4.818621297190424e-06, "loss": 1.1362, "step": 3737 }, { "epoch": 0.45, "learning_rate": 4.818501081457685e-06, "loss": 0.661, "step": 3738 }, { "epoch": 0.45, "learning_rate": 4.818380827399927e-06, "loss": 0.7323, "step": 3739 }, { "epoch": 0.45, "learning_rate": 4.81826053501914e-06, "loss": 0.7164, "step": 3740 }, { "epoch": 0.45, "learning_rate": 4.818140204317312e-06, "loss": 0.7481, "step": 3741 }, { "epoch": 0.45, "learning_rate": 4.818019835296432e-06, "loss": 0.9409, "step": 3742 }, { "epoch": 0.45, "learning_rate": 4.817899427958489e-06, "loss": 1.0173, "step": 3743 }, { "epoch": 0.45, "learning_rate": 4.817778982305474e-06, "loss": 0.6295, "step": 3744 }, { "epoch": 0.45, "learning_rate": 4.817658498339379e-06, "loss": 0.601, "step": 3745 }, { "epoch": 0.45, "learning_rate": 4.817537976062194e-06, "loss": 0.6571, "step": 3746 }, { "epoch": 0.45, "learning_rate": 4.817417415475911e-06, "loss": 0.7408, "step": 3747 }, { "epoch": 0.45, "learning_rate": 4.817296816582525e-06, "loss": 0.868, "step": 3748 }, { "epoch": 0.45, "learning_rate": 4.817176179384027e-06, "loss": 1.2729, "step": 3749 }, { "epoch": 0.45, "learning_rate": 4.817055503882413e-06, "loss": 1.0036, "step": 3750 }, { "epoch": 0.45, "learning_rate": 4.816934790079677e-06, "loss": 0.6844, "step": 3751 }, { "epoch": 0.45, "learning_rate": 4.816814037977814e-06, "loss": 0.7557, "step": 3752 }, { "epoch": 0.45, "learning_rate": 4.816693247578821e-06, "loss": 0.8548, "step": 3753 }, { "epoch": 0.45, "learning_rate": 4.816572418884695e-06, "loss": 0.9173, "step": 3754 }, { "epoch": 0.45, "learning_rate": 4.816451551897431e-06, "loss": 1.0801, "step": 3755 }, { "epoch": 0.45, "learning_rate": 4.81633064661903e-06, "loss": 1.114, "step": 3756 }, { "epoch": 0.45, "learning_rate": 4.816209703051488e-06, "loss": 0.4981, "step": 3757 }, { "epoch": 0.45, "learning_rate": 4.816088721196805e-06, "loss": 0.6462, "step": 3758 }, { "epoch": 0.45, "learning_rate": 4.815967701056982e-06, "loss": 0.6731, "step": 3759 }, { "epoch": 0.45, "learning_rate": 4.815846642634019e-06, "loss": 0.8989, "step": 3760 }, { "epoch": 0.45, "learning_rate": 4.815725545929916e-06, "loss": 1.1002, "step": 3761 }, { "epoch": 0.45, "learning_rate": 4.815604410946675e-06, "loss": 1.1872, "step": 3762 }, { "epoch": 0.45, "learning_rate": 4.815483237686299e-06, "loss": 0.497, "step": 3763 }, { "epoch": 0.45, "learning_rate": 4.81536202615079e-06, "loss": 0.7778, "step": 3764 }, { "epoch": 0.45, "learning_rate": 4.8152407763421545e-06, "loss": 0.8953, "step": 3765 }, { "epoch": 0.45, "learning_rate": 4.815119488262393e-06, "loss": 0.8446, "step": 3766 }, { "epoch": 0.45, "learning_rate": 4.814998161913513e-06, "loss": 0.9212, "step": 3767 }, { "epoch": 0.45, "learning_rate": 4.8148767972975185e-06, "loss": 1.1293, "step": 3768 }, { "epoch": 0.45, "learning_rate": 4.814755394416417e-06, "loss": 0.7769, "step": 3769 }, { "epoch": 0.45, "learning_rate": 4.814633953272214e-06, "loss": 0.5747, "step": 3770 }, { "epoch": 0.45, "learning_rate": 4.814512473866918e-06, "loss": 0.7214, "step": 3771 }, { "epoch": 0.45, "learning_rate": 4.8143909562025356e-06, "loss": 0.7739, "step": 3772 }, { "epoch": 0.45, "learning_rate": 4.814269400281078e-06, "loss": 0.8845, "step": 3773 }, { "epoch": 0.45, "learning_rate": 4.814147806104552e-06, "loss": 1.01, "step": 3774 }, { "epoch": 0.45, "learning_rate": 4.814026173674969e-06, "loss": 1.158, "step": 3775 }, { "epoch": 0.45, "learning_rate": 4.813904502994339e-06, "loss": 0.6462, "step": 3776 }, { "epoch": 0.45, "learning_rate": 4.813782794064674e-06, "loss": 0.7257, "step": 3777 }, { "epoch": 0.45, "learning_rate": 4.813661046887985e-06, "loss": 0.65, "step": 3778 }, { "epoch": 0.45, "learning_rate": 4.813539261466284e-06, "loss": 0.8497, "step": 3779 }, { "epoch": 0.45, "learning_rate": 4.813417437801585e-06, "loss": 1.0376, "step": 3780 }, { "epoch": 0.45, "learning_rate": 4.813295575895902e-06, "loss": 1.2643, "step": 3781 }, { "epoch": 0.45, "learning_rate": 4.813173675751249e-06, "loss": 0.6367, "step": 3782 }, { "epoch": 0.45, "learning_rate": 4.813051737369642e-06, "loss": 0.7586, "step": 3783 }, { "epoch": 0.45, "learning_rate": 4.812929760753094e-06, "loss": 0.6266, "step": 3784 }, { "epoch": 0.45, "learning_rate": 4.812807745903625e-06, "loss": 0.8297, "step": 3785 }, { "epoch": 0.45, "learning_rate": 4.812685692823248e-06, "loss": 1.0005, "step": 3786 }, { "epoch": 0.45, "learning_rate": 4.812563601513983e-06, "loss": 1.1116, "step": 3787 }, { "epoch": 0.45, "learning_rate": 4.812441471977848e-06, "loss": 0.5765, "step": 3788 }, { "epoch": 0.45, "learning_rate": 4.812319304216862e-06, "loss": 0.846, "step": 3789 }, { "epoch": 0.45, "learning_rate": 4.812197098233042e-06, "loss": 0.8259, "step": 3790 }, { "epoch": 0.45, "learning_rate": 4.812074854028411e-06, "loss": 0.741, "step": 3791 }, { "epoch": 0.45, "learning_rate": 4.811952571604989e-06, "loss": 0.9567, "step": 3792 }, { "epoch": 0.45, "learning_rate": 4.811830250964797e-06, "loss": 1.2483, "step": 3793 }, { "epoch": 0.45, "learning_rate": 4.811707892109857e-06, "loss": 0.8251, "step": 3794 }, { "epoch": 0.45, "learning_rate": 4.811585495042192e-06, "loss": 0.6324, "step": 3795 }, { "epoch": 0.45, "learning_rate": 4.811463059763823e-06, "loss": 0.8157, "step": 3796 }, { "epoch": 0.45, "learning_rate": 4.811340586276777e-06, "loss": 0.7631, "step": 3797 }, { "epoch": 0.45, "learning_rate": 4.811218074583077e-06, "loss": 0.8374, "step": 3798 }, { "epoch": 0.45, "learning_rate": 4.811095524684748e-06, "loss": 1.0977, "step": 3799 }, { "epoch": 0.45, "learning_rate": 4.810972936583817e-06, "loss": 1.1596, "step": 3800 }, { "epoch": 0.45, "learning_rate": 4.810850310282308e-06, "loss": 0.6201, "step": 3801 }, { "epoch": 0.45, "learning_rate": 4.810727645782251e-06, "loss": 0.7498, "step": 3802 }, { "epoch": 0.45, "learning_rate": 4.810604943085671e-06, "loss": 0.7644, "step": 3803 }, { "epoch": 0.45, "learning_rate": 4.810482202194598e-06, "loss": 0.8346, "step": 3804 }, { "epoch": 0.45, "learning_rate": 4.810359423111059e-06, "loss": 1.0628, "step": 3805 }, { "epoch": 0.45, "learning_rate": 4.810236605837086e-06, "loss": 1.2404, "step": 3806 }, { "epoch": 0.45, "learning_rate": 4.810113750374709e-06, "loss": 0.7141, "step": 3807 }, { "epoch": 0.45, "learning_rate": 4.809990856725957e-06, "loss": 0.5574, "step": 3808 }, { "epoch": 0.45, "learning_rate": 4.809867924892861e-06, "loss": 0.8259, "step": 3809 }, { "epoch": 0.45, "learning_rate": 4.809744954877457e-06, "loss": 0.9428, "step": 3810 }, { "epoch": 0.45, "learning_rate": 4.809621946681773e-06, "loss": 0.8954, "step": 3811 }, { "epoch": 0.45, "learning_rate": 4.809498900307845e-06, "loss": 1.0894, "step": 3812 }, { "epoch": 0.45, "learning_rate": 4.809375815757707e-06, "loss": 0.5108, "step": 3813 }, { "epoch": 0.45, "learning_rate": 4.809252693033392e-06, "loss": 0.713, "step": 3814 }, { "epoch": 0.45, "learning_rate": 4.809129532136937e-06, "loss": 0.7453, "step": 3815 }, { "epoch": 0.45, "learning_rate": 4.809006333070377e-06, "loss": 0.8442, "step": 3816 }, { "epoch": 0.45, "learning_rate": 4.808883095835748e-06, "loss": 0.8633, "step": 3817 }, { "epoch": 0.45, "learning_rate": 4.808759820435089e-06, "loss": 1.154, "step": 3818 }, { "epoch": 0.45, "learning_rate": 4.808636506870435e-06, "loss": 0.9221, "step": 3819 }, { "epoch": 0.45, "learning_rate": 4.808513155143826e-06, "loss": 0.4708, "step": 3820 }, { "epoch": 0.46, "learning_rate": 4.808389765257302e-06, "loss": 0.7328, "step": 3821 }, { "epoch": 0.46, "learning_rate": 4.8082663372129e-06, "loss": 0.7643, "step": 3822 }, { "epoch": 0.46, "learning_rate": 4.808142871012662e-06, "loss": 0.8382, "step": 3823 }, { "epoch": 0.46, "learning_rate": 4.80801936665863e-06, "loss": 1.1212, "step": 3824 }, { "epoch": 0.46, "learning_rate": 4.807895824152843e-06, "loss": 1.2327, "step": 3825 }, { "epoch": 0.46, "learning_rate": 4.807772243497345e-06, "loss": 0.5272, "step": 3826 }, { "epoch": 0.46, "learning_rate": 4.807648624694178e-06, "loss": 0.8101, "step": 3827 }, { "epoch": 0.46, "learning_rate": 4.807524967745385e-06, "loss": 0.8213, "step": 3828 }, { "epoch": 0.46, "learning_rate": 4.807401272653012e-06, "loss": 0.7282, "step": 3829 }, { "epoch": 0.46, "learning_rate": 4.807277539419101e-06, "loss": 0.9968, "step": 3830 }, { "epoch": 0.46, "learning_rate": 4.8071537680456995e-06, "loss": 1.1375, "step": 3831 }, { "epoch": 0.46, "learning_rate": 4.8070299585348525e-06, "loss": 0.5967, "step": 3832 }, { "epoch": 0.46, "learning_rate": 4.806906110888606e-06, "loss": 0.5448, "step": 3833 }, { "epoch": 0.46, "learning_rate": 4.806782225109009e-06, "loss": 0.6678, "step": 3834 }, { "epoch": 0.46, "learning_rate": 4.806658301198107e-06, "loss": 0.7438, "step": 3835 }, { "epoch": 0.46, "learning_rate": 4.806534339157951e-06, "loss": 0.9368, "step": 3836 }, { "epoch": 0.46, "learning_rate": 4.806410338990588e-06, "loss": 1.1705, "step": 3837 }, { "epoch": 0.46, "learning_rate": 4.806286300698069e-06, "loss": 0.7178, "step": 3838 }, { "epoch": 0.46, "learning_rate": 4.806162224282444e-06, "loss": 0.7641, "step": 3839 }, { "epoch": 0.46, "learning_rate": 4.806038109745764e-06, "loss": 0.821, "step": 3840 }, { "epoch": 0.46, "learning_rate": 4.80591395709008e-06, "loss": 0.8302, "step": 3841 }, { "epoch": 0.46, "learning_rate": 4.805789766317445e-06, "loss": 1.1298, "step": 3842 }, { "epoch": 0.46, "learning_rate": 4.805665537429912e-06, "loss": 1.1795, "step": 3843 }, { "epoch": 0.46, "learning_rate": 4.805541270429534e-06, "loss": 0.7894, "step": 3844 }, { "epoch": 0.46, "learning_rate": 4.805416965318366e-06, "loss": 0.6122, "step": 3845 }, { "epoch": 0.46, "learning_rate": 4.805292622098461e-06, "loss": 0.6443, "step": 3846 }, { "epoch": 0.46, "learning_rate": 4.805168240771876e-06, "loss": 0.6626, "step": 3847 }, { "epoch": 0.46, "learning_rate": 4.805043821340667e-06, "loss": 0.7868, "step": 3848 }, { "epoch": 0.46, "learning_rate": 4.804919363806889e-06, "loss": 1.1298, "step": 3849 }, { "epoch": 0.46, "learning_rate": 4.804794868172602e-06, "loss": 1.2004, "step": 3850 }, { "epoch": 0.46, "learning_rate": 4.804670334439862e-06, "loss": 0.6914, "step": 3851 }, { "epoch": 0.46, "learning_rate": 4.804545762610727e-06, "loss": 0.7328, "step": 3852 }, { "epoch": 0.46, "learning_rate": 4.804421152687258e-06, "loss": 0.8144, "step": 3853 }, { "epoch": 0.46, "learning_rate": 4.804296504671514e-06, "loss": 0.8633, "step": 3854 }, { "epoch": 0.46, "learning_rate": 4.804171818565555e-06, "loss": 1.0955, "step": 3855 }, { "epoch": 0.46, "learning_rate": 4.8040470943714435e-06, "loss": 1.4093, "step": 3856 }, { "epoch": 0.46, "learning_rate": 4.803922332091239e-06, "loss": 0.6231, "step": 3857 }, { "epoch": 0.46, "learning_rate": 4.803797531727006e-06, "loss": 0.7054, "step": 3858 }, { "epoch": 0.46, "learning_rate": 4.803672693280806e-06, "loss": 0.7702, "step": 3859 }, { "epoch": 0.46, "learning_rate": 4.8035478167547035e-06, "loss": 0.89, "step": 3860 }, { "epoch": 0.46, "learning_rate": 4.803422902150762e-06, "loss": 1.0033, "step": 3861 }, { "epoch": 0.46, "learning_rate": 4.803297949471046e-06, "loss": 1.1644, "step": 3862 }, { "epoch": 0.46, "learning_rate": 4.803172958717622e-06, "loss": 0.7417, "step": 3863 }, { "epoch": 0.46, "learning_rate": 4.803047929892556e-06, "loss": 0.6262, "step": 3864 }, { "epoch": 0.46, "learning_rate": 4.802922862997915e-06, "loss": 0.7188, "step": 3865 }, { "epoch": 0.46, "learning_rate": 4.802797758035766e-06, "loss": 0.6409, "step": 3866 }, { "epoch": 0.46, "learning_rate": 4.802672615008176e-06, "loss": 0.9857, "step": 3867 }, { "epoch": 0.46, "learning_rate": 4.802547433917214e-06, "loss": 1.2527, "step": 3868 }, { "epoch": 0.46, "learning_rate": 4.80242221476495e-06, "loss": 0.7995, "step": 3869 }, { "epoch": 0.46, "learning_rate": 4.802296957553454e-06, "loss": 0.6148, "step": 3870 }, { "epoch": 0.46, "learning_rate": 4.8021716622847965e-06, "loss": 0.7391, "step": 3871 }, { "epoch": 0.46, "learning_rate": 4.802046328961047e-06, "loss": 0.7156, "step": 3872 }, { "epoch": 0.46, "learning_rate": 4.80192095758428e-06, "loss": 1.0085, "step": 3873 }, { "epoch": 0.46, "learning_rate": 4.801795548156566e-06, "loss": 1.0471, "step": 3874 }, { "epoch": 0.46, "learning_rate": 4.801670100679978e-06, "loss": 0.9748, "step": 3875 }, { "epoch": 0.46, "learning_rate": 4.80154461515659e-06, "loss": 0.6412, "step": 3876 }, { "epoch": 0.46, "learning_rate": 4.801419091588477e-06, "loss": 0.7334, "step": 3877 }, { "epoch": 0.46, "learning_rate": 4.801293529977713e-06, "loss": 0.743, "step": 3878 }, { "epoch": 0.46, "learning_rate": 4.801167930326374e-06, "loss": 0.703, "step": 3879 }, { "epoch": 0.46, "learning_rate": 4.801042292636536e-06, "loss": 0.9045, "step": 3880 }, { "epoch": 0.46, "learning_rate": 4.800916616910276e-06, "loss": 1.1591, "step": 3881 }, { "epoch": 0.46, "learning_rate": 4.800790903149671e-06, "loss": 0.666, "step": 3882 }, { "epoch": 0.46, "learning_rate": 4.800665151356799e-06, "loss": 0.7076, "step": 3883 }, { "epoch": 0.46, "learning_rate": 4.800539361533739e-06, "loss": 0.7104, "step": 3884 }, { "epoch": 0.46, "learning_rate": 4.800413533682571e-06, "loss": 0.7579, "step": 3885 }, { "epoch": 0.46, "learning_rate": 4.800287667805374e-06, "loss": 0.9033, "step": 3886 }, { "epoch": 0.46, "learning_rate": 4.8001617639042285e-06, "loss": 1.1708, "step": 3887 }, { "epoch": 0.46, "learning_rate": 4.8000358219812165e-06, "loss": 0.672, "step": 3888 }, { "epoch": 0.46, "learning_rate": 4.799909842038419e-06, "loss": 0.7583, "step": 3889 }, { "epoch": 0.46, "learning_rate": 4.799783824077919e-06, "loss": 0.6889, "step": 3890 }, { "epoch": 0.46, "learning_rate": 4.7996577681018e-06, "loss": 0.8582, "step": 3891 }, { "epoch": 0.46, "learning_rate": 4.7995316741121445e-06, "loss": 0.9218, "step": 3892 }, { "epoch": 0.46, "learning_rate": 4.799405542111038e-06, "loss": 1.1349, "step": 3893 }, { "epoch": 0.46, "learning_rate": 4.799279372100564e-06, "loss": 0.7639, "step": 3894 }, { "epoch": 0.46, "learning_rate": 4.799153164082809e-06, "loss": 0.5934, "step": 3895 }, { "epoch": 0.46, "learning_rate": 4.79902691805986e-06, "loss": 0.7629, "step": 3896 }, { "epoch": 0.46, "learning_rate": 4.798900634033803e-06, "loss": 0.6749, "step": 3897 }, { "epoch": 0.46, "learning_rate": 4.798774312006725e-06, "loss": 0.8951, "step": 3898 }, { "epoch": 0.46, "learning_rate": 4.798647951980716e-06, "loss": 1.0399, "step": 3899 }, { "epoch": 0.46, "learning_rate": 4.798521553957863e-06, "loss": 1.0968, "step": 3900 }, { "epoch": 0.46, "learning_rate": 4.798395117940255e-06, "loss": 0.625, "step": 3901 }, { "epoch": 0.46, "learning_rate": 4.798268643929984e-06, "loss": 0.6279, "step": 3902 }, { "epoch": 0.46, "learning_rate": 4.798142131929139e-06, "loss": 0.8215, "step": 3903 }, { "epoch": 0.46, "learning_rate": 4.798015581939812e-06, "loss": 0.8751, "step": 3904 }, { "epoch": 0.47, "learning_rate": 4.797888993964094e-06, "loss": 1.0414, "step": 3905 }, { "epoch": 0.47, "learning_rate": 4.797762368004078e-06, "loss": 1.1177, "step": 3906 }, { "epoch": 0.47, "learning_rate": 4.797635704061858e-06, "loss": 0.6348, "step": 3907 }, { "epoch": 0.47, "learning_rate": 4.797509002139527e-06, "loss": 0.7304, "step": 3908 }, { "epoch": 0.47, "learning_rate": 4.7973822622391795e-06, "loss": 0.7547, "step": 3909 }, { "epoch": 0.47, "learning_rate": 4.79725548436291e-06, "loss": 0.7797, "step": 3910 }, { "epoch": 0.47, "learning_rate": 4.7971286685128145e-06, "loss": 1.0642, "step": 3911 }, { "epoch": 0.47, "learning_rate": 4.79700181469099e-06, "loss": 1.102, "step": 3912 }, { "epoch": 0.47, "learning_rate": 4.796874922899532e-06, "loss": 0.5549, "step": 3913 }, { "epoch": 0.47, "learning_rate": 4.796747993140539e-06, "loss": 0.6603, "step": 3914 }, { "epoch": 0.47, "learning_rate": 4.7966210254161095e-06, "loss": 0.7206, "step": 3915 }, { "epoch": 0.47, "learning_rate": 4.796494019728341e-06, "loss": 0.7331, "step": 3916 }, { "epoch": 0.47, "learning_rate": 4.796366976079334e-06, "loss": 0.8215, "step": 3917 }, { "epoch": 0.47, "learning_rate": 4.796239894471188e-06, "loss": 1.1381, "step": 3918 }, { "epoch": 0.47, "learning_rate": 4.796112774906005e-06, "loss": 0.5707, "step": 3919 }, { "epoch": 0.47, "learning_rate": 4.795985617385883e-06, "loss": 0.6112, "step": 3920 }, { "epoch": 0.47, "learning_rate": 4.795858421912928e-06, "loss": 0.7661, "step": 3921 }, { "epoch": 0.47, "learning_rate": 4.79573118848924e-06, "loss": 0.8534, "step": 3922 }, { "epoch": 0.47, "learning_rate": 4.795603917116922e-06, "loss": 0.9392, "step": 3923 }, { "epoch": 0.47, "learning_rate": 4.79547660779808e-06, "loss": 1.1183, "step": 3924 }, { "epoch": 0.47, "learning_rate": 4.795349260534816e-06, "loss": 1.2802, "step": 3925 }, { "epoch": 0.47, "learning_rate": 4.795221875329237e-06, "loss": 0.7553, "step": 3926 }, { "epoch": 0.47, "learning_rate": 4.795094452183448e-06, "loss": 0.6488, "step": 3927 }, { "epoch": 0.47, "learning_rate": 4.794966991099554e-06, "loss": 0.7913, "step": 3928 }, { "epoch": 0.47, "learning_rate": 4.794839492079663e-06, "loss": 0.8332, "step": 3929 }, { "epoch": 0.47, "learning_rate": 4.794711955125884e-06, "loss": 1.0266, "step": 3930 }, { "epoch": 0.47, "learning_rate": 4.794584380240324e-06, "loss": 1.2757, "step": 3931 }, { "epoch": 0.47, "learning_rate": 4.79445676742509e-06, "loss": 0.6796, "step": 3932 }, { "epoch": 0.47, "learning_rate": 4.794329116682294e-06, "loss": 0.7908, "step": 3933 }, { "epoch": 0.47, "learning_rate": 4.794201428014045e-06, "loss": 0.708, "step": 3934 }, { "epoch": 0.47, "learning_rate": 4.794073701422454e-06, "loss": 0.9406, "step": 3935 }, { "epoch": 0.47, "learning_rate": 4.793945936909633e-06, "loss": 0.9636, "step": 3936 }, { "epoch": 0.47, "learning_rate": 4.7938181344776915e-06, "loss": 1.2068, "step": 3937 }, { "epoch": 0.47, "learning_rate": 4.793690294128745e-06, "loss": 0.6183, "step": 3938 }, { "epoch": 0.47, "learning_rate": 4.793562415864905e-06, "loss": 0.6783, "step": 3939 }, { "epoch": 0.47, "learning_rate": 4.793434499688287e-06, "loss": 0.7139, "step": 3940 }, { "epoch": 0.47, "learning_rate": 4.793306545601002e-06, "loss": 0.7335, "step": 3941 }, { "epoch": 0.47, "learning_rate": 4.793178553605169e-06, "loss": 0.8466, "step": 3942 }, { "epoch": 0.47, "learning_rate": 4.793050523702902e-06, "loss": 0.9597, "step": 3943 }, { "epoch": 0.47, "learning_rate": 4.792922455896317e-06, "loss": 0.9058, "step": 3944 }, { "epoch": 0.47, "learning_rate": 4.7927943501875315e-06, "loss": 0.6684, "step": 3945 }, { "epoch": 0.47, "learning_rate": 4.792666206578664e-06, "loss": 0.6297, "step": 3946 }, { "epoch": 0.47, "learning_rate": 4.79253802507183e-06, "loss": 0.8602, "step": 3947 }, { "epoch": 0.47, "learning_rate": 4.792409805669151e-06, "loss": 0.88, "step": 3948 }, { "epoch": 0.47, "learning_rate": 4.7922815483727456e-06, "loss": 0.968, "step": 3949 }, { "epoch": 0.47, "learning_rate": 4.792153253184734e-06, "loss": 1.2447, "step": 3950 }, { "epoch": 0.47, "learning_rate": 4.792024920107236e-06, "loss": 0.5283, "step": 3951 }, { "epoch": 0.47, "learning_rate": 4.791896549142374e-06, "loss": 0.7375, "step": 3952 }, { "epoch": 0.47, "learning_rate": 4.791768140292271e-06, "loss": 0.7392, "step": 3953 }, { "epoch": 0.47, "learning_rate": 4.791639693559047e-06, "loss": 0.8772, "step": 3954 }, { "epoch": 0.47, "learning_rate": 4.791511208944827e-06, "loss": 1.0041, "step": 3955 }, { "epoch": 0.47, "learning_rate": 4.791382686451734e-06, "loss": 1.0818, "step": 3956 }, { "epoch": 0.47, "learning_rate": 4.791254126081894e-06, "loss": 0.5566, "step": 3957 }, { "epoch": 0.47, "learning_rate": 4.79112552783743e-06, "loss": 0.6492, "step": 3958 }, { "epoch": 0.47, "learning_rate": 4.7909968917204695e-06, "loss": 0.738, "step": 3959 }, { "epoch": 0.47, "learning_rate": 4.790868217733138e-06, "loss": 0.8504, "step": 3960 }, { "epoch": 0.47, "learning_rate": 4.790739505877562e-06, "loss": 0.9927, "step": 3961 }, { "epoch": 0.47, "learning_rate": 4.79061075615587e-06, "loss": 1.2295, "step": 3962 }, { "epoch": 0.47, "learning_rate": 4.79048196857019e-06, "loss": 0.6897, "step": 3963 }, { "epoch": 0.47, "learning_rate": 4.790353143122651e-06, "loss": 0.5715, "step": 3964 }, { "epoch": 0.47, "learning_rate": 4.790224279815383e-06, "loss": 0.7618, "step": 3965 }, { "epoch": 0.47, "learning_rate": 4.790095378650515e-06, "loss": 0.85, "step": 3966 }, { "epoch": 0.47, "learning_rate": 4.789966439630178e-06, "loss": 0.876, "step": 3967 }, { "epoch": 0.47, "learning_rate": 4.789837462756504e-06, "loss": 1.0861, "step": 3968 }, { "epoch": 0.47, "learning_rate": 4.789708448031625e-06, "loss": 0.792, "step": 3969 }, { "epoch": 0.47, "learning_rate": 4.789579395457672e-06, "loss": 0.647, "step": 3970 }, { "epoch": 0.47, "learning_rate": 4.78945030503678e-06, "loss": 0.6836, "step": 3971 }, { "epoch": 0.47, "learning_rate": 4.789321176771083e-06, "loss": 0.8652, "step": 3972 }, { "epoch": 0.47, "learning_rate": 4.789192010662714e-06, "loss": 0.894, "step": 3973 }, { "epoch": 0.47, "learning_rate": 4.789062806713809e-06, "loss": 1.0768, "step": 3974 }, { "epoch": 0.47, "learning_rate": 4.788933564926504e-06, "loss": 1.202, "step": 3975 }, { "epoch": 0.47, "learning_rate": 4.788804285302935e-06, "loss": 0.627, "step": 3976 }, { "epoch": 0.47, "learning_rate": 4.78867496784524e-06, "loss": 0.6691, "step": 3977 }, { "epoch": 0.47, "learning_rate": 4.788545612555555e-06, "loss": 0.7575, "step": 3978 }, { "epoch": 0.47, "learning_rate": 4.788416219436018e-06, "loss": 0.874, "step": 3979 }, { "epoch": 0.47, "learning_rate": 4.7882867884887705e-06, "loss": 0.9487, "step": 3980 }, { "epoch": 0.47, "learning_rate": 4.78815731971595e-06, "loss": 1.0869, "step": 3981 }, { "epoch": 0.47, "learning_rate": 4.7880278131196965e-06, "loss": 0.4616, "step": 3982 }, { "epoch": 0.47, "learning_rate": 4.787898268702152e-06, "loss": 0.6765, "step": 3983 }, { "epoch": 0.47, "learning_rate": 4.787768686465457e-06, "loss": 0.928, "step": 3984 }, { "epoch": 0.47, "learning_rate": 4.787639066411753e-06, "loss": 0.8416, "step": 3985 }, { "epoch": 0.47, "learning_rate": 4.787509408543184e-06, "loss": 0.9663, "step": 3986 }, { "epoch": 0.47, "learning_rate": 4.787379712861892e-06, "loss": 1.1968, "step": 3987 }, { "epoch": 0.47, "learning_rate": 4.787249979370021e-06, "loss": 0.7647, "step": 3988 }, { "epoch": 0.48, "learning_rate": 4.787120208069717e-06, "loss": 0.6562, "step": 3989 }, { "epoch": 0.48, "learning_rate": 4.786990398963124e-06, "loss": 0.7391, "step": 3990 }, { "epoch": 0.48, "learning_rate": 4.786860552052387e-06, "loss": 0.8333, "step": 3991 }, { "epoch": 0.48, "learning_rate": 4.786730667339653e-06, "loss": 1.0053, "step": 3992 }, { "epoch": 0.48, "learning_rate": 4.78660074482707e-06, "loss": 1.1066, "step": 3993 }, { "epoch": 0.48, "learning_rate": 4.7864707845167855e-06, "loss": 0.7335, "step": 3994 }, { "epoch": 0.48, "learning_rate": 4.786340786410945e-06, "loss": 0.6223, "step": 3995 }, { "epoch": 0.48, "learning_rate": 4.786210750511701e-06, "loss": 0.6543, "step": 3996 }, { "epoch": 0.48, "learning_rate": 4.786080676821201e-06, "loss": 0.8415, "step": 3997 }, { "epoch": 0.48, "learning_rate": 4.785950565341596e-06, "loss": 0.9072, "step": 3998 }, { "epoch": 0.48, "learning_rate": 4.7858204160750365e-06, "loss": 1.1049, "step": 3999 }, { "epoch": 0.48, "learning_rate": 4.785690229023673e-06, "loss": 1.2434, "step": 4000 }, { "epoch": 0.48, "learning_rate": 4.785560004189659e-06, "loss": 0.5216, "step": 4001 }, { "epoch": 0.48, "learning_rate": 4.785429741575146e-06, "loss": 0.6917, "step": 4002 }, { "epoch": 0.48, "learning_rate": 4.785299441182287e-06, "loss": 0.8069, "step": 4003 }, { "epoch": 0.48, "learning_rate": 4.785169103013237e-06, "loss": 1.0448, "step": 4004 }, { "epoch": 0.48, "learning_rate": 4.78503872707015e-06, "loss": 1.0486, "step": 4005 }, { "epoch": 0.48, "learning_rate": 4.784908313355181e-06, "loss": 1.309, "step": 4006 }, { "epoch": 0.48, "learning_rate": 4.784777861870486e-06, "loss": 0.6938, "step": 4007 }, { "epoch": 0.48, "learning_rate": 4.784647372618221e-06, "loss": 0.6102, "step": 4008 }, { "epoch": 0.48, "learning_rate": 4.7845168456005435e-06, "loss": 0.6292, "step": 4009 }, { "epoch": 0.48, "learning_rate": 4.784386280819611e-06, "loss": 0.7974, "step": 4010 }, { "epoch": 0.48, "learning_rate": 4.784255678277581e-06, "loss": 1.0023, "step": 4011 }, { "epoch": 0.48, "learning_rate": 4.7841250379766135e-06, "loss": 1.1693, "step": 4012 }, { "epoch": 0.48, "learning_rate": 4.783994359918868e-06, "loss": 0.6027, "step": 4013 }, { "epoch": 0.48, "learning_rate": 4.783863644106502e-06, "loss": 0.6812, "step": 4014 }, { "epoch": 0.48, "learning_rate": 4.78373289054168e-06, "loss": 0.6538, "step": 4015 }, { "epoch": 0.48, "learning_rate": 4.783602099226562e-06, "loss": 0.8951, "step": 4016 }, { "epoch": 0.48, "learning_rate": 4.783471270163309e-06, "loss": 1.1104, "step": 4017 }, { "epoch": 0.48, "learning_rate": 4.783340403354084e-06, "loss": 1.0669, "step": 4018 }, { "epoch": 0.48, "learning_rate": 4.7832094988010515e-06, "loss": 0.6559, "step": 4019 }, { "epoch": 0.48, "learning_rate": 4.783078556506374e-06, "loss": 0.5814, "step": 4020 }, { "epoch": 0.48, "learning_rate": 4.782947576472215e-06, "loss": 0.9186, "step": 4021 }, { "epoch": 0.48, "learning_rate": 4.782816558700743e-06, "loss": 0.8234, "step": 4022 }, { "epoch": 0.48, "learning_rate": 4.782685503194121e-06, "loss": 0.9138, "step": 4023 }, { "epoch": 0.48, "learning_rate": 4.782554409954516e-06, "loss": 1.0148, "step": 4024 }, { "epoch": 0.48, "learning_rate": 4.782423278984095e-06, "loss": 1.1987, "step": 4025 }, { "epoch": 0.48, "learning_rate": 4.782292110285026e-06, "loss": 0.5526, "step": 4026 }, { "epoch": 0.48, "learning_rate": 4.7821609038594774e-06, "loss": 0.7235, "step": 4027 }, { "epoch": 0.48, "learning_rate": 4.7820296597096165e-06, "loss": 0.8942, "step": 4028 }, { "epoch": 0.48, "learning_rate": 4.781898377837615e-06, "loss": 0.7963, "step": 4029 }, { "epoch": 0.48, "learning_rate": 4.781767058245641e-06, "loss": 0.9789, "step": 4030 }, { "epoch": 0.48, "learning_rate": 4.781635700935866e-06, "loss": 1.2221, "step": 4031 }, { "epoch": 0.48, "learning_rate": 4.781504305910462e-06, "loss": 0.6907, "step": 4032 }, { "epoch": 0.48, "learning_rate": 4.781372873171599e-06, "loss": 0.7111, "step": 4033 }, { "epoch": 0.48, "learning_rate": 4.781241402721453e-06, "loss": 0.7153, "step": 4034 }, { "epoch": 0.48, "learning_rate": 4.781109894562194e-06, "loss": 0.828, "step": 4035 }, { "epoch": 0.48, "learning_rate": 4.780978348695997e-06, "loss": 0.9186, "step": 4036 }, { "epoch": 0.48, "learning_rate": 4.780846765125037e-06, "loss": 1.2078, "step": 4037 }, { "epoch": 0.48, "learning_rate": 4.7807151438514886e-06, "loss": 0.506, "step": 4038 }, { "epoch": 0.48, "learning_rate": 4.780583484877526e-06, "loss": 0.6844, "step": 4039 }, { "epoch": 0.48, "learning_rate": 4.7804517882053295e-06, "loss": 0.6988, "step": 4040 }, { "epoch": 0.48, "learning_rate": 4.780320053837072e-06, "loss": 0.7256, "step": 4041 }, { "epoch": 0.48, "learning_rate": 4.780188281774933e-06, "loss": 0.8111, "step": 4042 }, { "epoch": 0.48, "learning_rate": 4.780056472021091e-06, "loss": 1.0438, "step": 4043 }, { "epoch": 0.48, "learning_rate": 4.779924624577723e-06, "loss": 0.6757, "step": 4044 }, { "epoch": 0.48, "learning_rate": 4.77979273944701e-06, "loss": 0.644, "step": 4045 }, { "epoch": 0.48, "learning_rate": 4.779660816631132e-06, "loss": 0.8717, "step": 4046 }, { "epoch": 0.48, "learning_rate": 4.77952885613227e-06, "loss": 0.8163, "step": 4047 }, { "epoch": 0.48, "learning_rate": 4.779396857952604e-06, "loss": 0.9487, "step": 4048 }, { "epoch": 0.48, "learning_rate": 4.779264822094317e-06, "loss": 1.1309, "step": 4049 }, { "epoch": 0.48, "learning_rate": 4.779132748559591e-06, "loss": 1.14, "step": 4050 }, { "epoch": 0.48, "learning_rate": 4.77900063735061e-06, "loss": 0.6611, "step": 4051 }, { "epoch": 0.48, "learning_rate": 4.778868488469557e-06, "loss": 0.7113, "step": 4052 }, { "epoch": 0.48, "learning_rate": 4.7787363019186165e-06, "loss": 0.7622, "step": 4053 }, { "epoch": 0.48, "learning_rate": 4.778604077699973e-06, "loss": 0.8983, "step": 4054 }, { "epoch": 0.48, "learning_rate": 4.778471815815814e-06, "loss": 1.0544, "step": 4055 }, { "epoch": 0.48, "learning_rate": 4.778339516268324e-06, "loss": 1.1225, "step": 4056 }, { "epoch": 0.48, "learning_rate": 4.778207179059691e-06, "loss": 0.6954, "step": 4057 }, { "epoch": 0.48, "learning_rate": 4.7780748041921025e-06, "loss": 0.5903, "step": 4058 }, { "epoch": 0.48, "learning_rate": 4.7779423916677455e-06, "loss": 0.7354, "step": 4059 }, { "epoch": 0.48, "learning_rate": 4.77780994148881e-06, "loss": 0.8782, "step": 4060 }, { "epoch": 0.48, "learning_rate": 4.777677453657486e-06, "loss": 0.9985, "step": 4061 }, { "epoch": 0.48, "learning_rate": 4.777544928175961e-06, "loss": 1.1807, "step": 4062 }, { "epoch": 0.48, "learning_rate": 4.777412365046428e-06, "loss": 0.6407, "step": 4063 }, { "epoch": 0.48, "learning_rate": 4.7772797642710775e-06, "loss": 0.7251, "step": 4064 }, { "epoch": 0.48, "learning_rate": 4.777147125852101e-06, "loss": 0.7026, "step": 4065 }, { "epoch": 0.48, "learning_rate": 4.7770144497916924e-06, "loss": 0.8723, "step": 4066 }, { "epoch": 0.48, "learning_rate": 4.776881736092043e-06, "loss": 0.9816, "step": 4067 }, { "epoch": 0.48, "learning_rate": 4.776748984755348e-06, "loss": 1.1398, "step": 4068 }, { "epoch": 0.48, "learning_rate": 4.776616195783801e-06, "loss": 1.0211, "step": 4069 }, { "epoch": 0.48, "learning_rate": 4.776483369179597e-06, "loss": 0.623, "step": 4070 }, { "epoch": 0.48, "learning_rate": 4.776350504944932e-06, "loss": 0.8618, "step": 4071 }, { "epoch": 0.48, "learning_rate": 4.776217603082003e-06, "loss": 0.706, "step": 4072 }, { "epoch": 0.49, "learning_rate": 4.776084663593005e-06, "loss": 0.7769, "step": 4073 }, { "epoch": 0.49, "learning_rate": 4.775951686480137e-06, "loss": 1.0548, "step": 4074 }, { "epoch": 0.49, "learning_rate": 4.775818671745596e-06, "loss": 1.1433, "step": 4075 }, { "epoch": 0.49, "learning_rate": 4.775685619391582e-06, "loss": 0.5916, "step": 4076 }, { "epoch": 0.49, "learning_rate": 4.775552529420294e-06, "loss": 0.7418, "step": 4077 }, { "epoch": 0.49, "learning_rate": 4.7754194018339315e-06, "loss": 0.7632, "step": 4078 }, { "epoch": 0.49, "learning_rate": 4.775286236634696e-06, "loss": 0.8776, "step": 4079 }, { "epoch": 0.49, "learning_rate": 4.775153033824786e-06, "loss": 0.9878, "step": 4080 }, { "epoch": 0.49, "learning_rate": 4.7750197934064065e-06, "loss": 1.1592, "step": 4081 }, { "epoch": 0.49, "learning_rate": 4.77488651538176e-06, "loss": 0.5415, "step": 4082 }, { "epoch": 0.49, "learning_rate": 4.774753199753048e-06, "loss": 0.7957, "step": 4083 }, { "epoch": 0.49, "learning_rate": 4.774619846522475e-06, "loss": 0.67, "step": 4084 }, { "epoch": 0.49, "learning_rate": 4.774486455692245e-06, "loss": 0.7883, "step": 4085 }, { "epoch": 0.49, "learning_rate": 4.774353027264562e-06, "loss": 0.9458, "step": 4086 }, { "epoch": 0.49, "learning_rate": 4.774219561241633e-06, "loss": 1.1296, "step": 4087 }, { "epoch": 0.49, "learning_rate": 4.774086057625665e-06, "loss": 0.6999, "step": 4088 }, { "epoch": 0.49, "learning_rate": 4.773952516418862e-06, "loss": 0.6937, "step": 4089 }, { "epoch": 0.49, "learning_rate": 4.7738189376234335e-06, "loss": 0.7458, "step": 4090 }, { "epoch": 0.49, "learning_rate": 4.7736853212415875e-06, "loss": 0.8848, "step": 4091 }, { "epoch": 0.49, "learning_rate": 4.773551667275533e-06, "loss": 1.0234, "step": 4092 }, { "epoch": 0.49, "learning_rate": 4.773417975727478e-06, "loss": 1.1552, "step": 4093 }, { "epoch": 0.49, "learning_rate": 4.773284246599634e-06, "loss": 1.0232, "step": 4094 }, { "epoch": 0.49, "learning_rate": 4.773150479894208e-06, "loss": 0.5685, "step": 4095 }, { "epoch": 0.49, "learning_rate": 4.7730166756134165e-06, "loss": 0.6599, "step": 4096 }, { "epoch": 0.49, "learning_rate": 4.772882833759468e-06, "loss": 0.8348, "step": 4097 }, { "epoch": 0.49, "learning_rate": 4.772748954334575e-06, "loss": 0.9402, "step": 4098 }, { "epoch": 0.49, "learning_rate": 4.772615037340952e-06, "loss": 1.0952, "step": 4099 }, { "epoch": 0.49, "learning_rate": 4.772481082780811e-06, "loss": 1.3501, "step": 4100 }, { "epoch": 0.49, "learning_rate": 4.772347090656367e-06, "loss": 0.4866, "step": 4101 }, { "epoch": 0.49, "learning_rate": 4.772213060969835e-06, "loss": 0.674, "step": 4102 }, { "epoch": 0.49, "learning_rate": 4.7720789937234305e-06, "loss": 0.9516, "step": 4103 }, { "epoch": 0.49, "learning_rate": 4.77194488891937e-06, "loss": 0.8626, "step": 4104 }, { "epoch": 0.49, "learning_rate": 4.77181074655987e-06, "loss": 1.0115, "step": 4105 }, { "epoch": 0.49, "learning_rate": 4.771676566647146e-06, "loss": 1.2285, "step": 4106 }, { "epoch": 0.49, "learning_rate": 4.77154234918342e-06, "loss": 0.6224, "step": 4107 }, { "epoch": 0.49, "learning_rate": 4.771408094170907e-06, "loss": 0.5969, "step": 4108 }, { "epoch": 0.49, "learning_rate": 4.7712738016118285e-06, "loss": 0.6967, "step": 4109 }, { "epoch": 0.49, "learning_rate": 4.771139471508402e-06, "loss": 0.7252, "step": 4110 }, { "epoch": 0.49, "learning_rate": 4.771005103862851e-06, "loss": 0.9582, "step": 4111 }, { "epoch": 0.49, "learning_rate": 4.770870698677395e-06, "loss": 1.1396, "step": 4112 }, { "epoch": 0.49, "learning_rate": 4.7707362559542544e-06, "loss": 0.6091, "step": 4113 }, { "epoch": 0.49, "learning_rate": 4.7706017756956545e-06, "loss": 0.6301, "step": 4114 }, { "epoch": 0.49, "learning_rate": 4.770467257903816e-06, "loss": 0.6303, "step": 4115 }, { "epoch": 0.49, "learning_rate": 4.770332702580963e-06, "loss": 0.738, "step": 4116 }, { "epoch": 0.49, "learning_rate": 4.77019810972932e-06, "loss": 0.8854, "step": 4117 }, { "epoch": 0.49, "learning_rate": 4.7700634793511126e-06, "loss": 0.9943, "step": 4118 }, { "epoch": 0.49, "learning_rate": 4.769928811448565e-06, "loss": 0.7985, "step": 4119 }, { "epoch": 0.49, "learning_rate": 4.769794106023904e-06, "loss": 0.575, "step": 4120 }, { "epoch": 0.49, "learning_rate": 4.7696593630793555e-06, "loss": 0.6772, "step": 4121 }, { "epoch": 0.49, "learning_rate": 4.769524582617147e-06, "loss": 0.8635, "step": 4122 }, { "epoch": 0.49, "learning_rate": 4.7693897646395075e-06, "loss": 0.8166, "step": 4123 }, { "epoch": 0.49, "learning_rate": 4.769254909148664e-06, "loss": 1.1204, "step": 4124 }, { "epoch": 0.49, "learning_rate": 4.769120016146846e-06, "loss": 1.2416, "step": 4125 }, { "epoch": 0.49, "learning_rate": 4.768985085636285e-06, "loss": 0.5199, "step": 4126 }, { "epoch": 0.49, "learning_rate": 4.7688501176192094e-06, "loss": 0.7633, "step": 4127 }, { "epoch": 0.49, "learning_rate": 4.768715112097851e-06, "loss": 0.6711, "step": 4128 }, { "epoch": 0.49, "learning_rate": 4.768580069074442e-06, "loss": 0.8603, "step": 4129 }, { "epoch": 0.49, "learning_rate": 4.768444988551214e-06, "loss": 1.0309, "step": 4130 }, { "epoch": 0.49, "learning_rate": 4.7683098705303995e-06, "loss": 1.2108, "step": 4131 }, { "epoch": 0.49, "learning_rate": 4.768174715014233e-06, "loss": 0.3971, "step": 4132 }, { "epoch": 0.49, "learning_rate": 4.768039522004948e-06, "loss": 0.6091, "step": 4133 }, { "epoch": 0.49, "learning_rate": 4.767904291504779e-06, "loss": 0.6766, "step": 4134 }, { "epoch": 0.49, "learning_rate": 4.767769023515963e-06, "loss": 0.7646, "step": 4135 }, { "epoch": 0.49, "learning_rate": 4.7676337180407326e-06, "loss": 0.918, "step": 4136 }, { "epoch": 0.49, "learning_rate": 4.767498375081328e-06, "loss": 1.2662, "step": 4137 }, { "epoch": 0.49, "learning_rate": 4.767362994639985e-06, "loss": 0.6201, "step": 4138 }, { "epoch": 0.49, "learning_rate": 4.767227576718941e-06, "loss": 0.5321, "step": 4139 }, { "epoch": 0.49, "learning_rate": 4.767092121320436e-06, "loss": 0.5784, "step": 4140 }, { "epoch": 0.49, "learning_rate": 4.766956628446707e-06, "loss": 0.7427, "step": 4141 }, { "epoch": 0.49, "learning_rate": 4.766821098099995e-06, "loss": 0.9689, "step": 4142 }, { "epoch": 0.49, "learning_rate": 4.76668553028254e-06, "loss": 1.1326, "step": 4143 }, { "epoch": 0.49, "learning_rate": 4.766549924996583e-06, "loss": 0.6729, "step": 4144 }, { "epoch": 0.49, "learning_rate": 4.7664142822443645e-06, "loss": 0.7047, "step": 4145 }, { "epoch": 0.49, "learning_rate": 4.76627860202813e-06, "loss": 0.7262, "step": 4146 }, { "epoch": 0.49, "learning_rate": 4.766142884350118e-06, "loss": 0.8752, "step": 4147 }, { "epoch": 0.49, "learning_rate": 4.7660071292125745e-06, "loss": 0.861, "step": 4148 }, { "epoch": 0.49, "learning_rate": 4.7658713366177434e-06, "loss": 1.1276, "step": 4149 }, { "epoch": 0.49, "learning_rate": 4.765735506567869e-06, "loss": 1.1536, "step": 4150 }, { "epoch": 0.49, "learning_rate": 4.765599639065197e-06, "loss": 0.6965, "step": 4151 }, { "epoch": 0.49, "learning_rate": 4.765463734111972e-06, "loss": 0.7203, "step": 4152 }, { "epoch": 0.49, "learning_rate": 4.765327791710442e-06, "loss": 0.7506, "step": 4153 }, { "epoch": 0.49, "learning_rate": 4.765191811862853e-06, "loss": 0.8942, "step": 4154 }, { "epoch": 0.49, "learning_rate": 4.765055794571454e-06, "loss": 1.2593, "step": 4155 }, { "epoch": 0.49, "learning_rate": 4.764919739838492e-06, "loss": 1.2319, "step": 4156 }, { "epoch": 0.5, "learning_rate": 4.764783647666217e-06, "loss": 0.5318, "step": 4157 }, { "epoch": 0.5, "learning_rate": 4.764647518056877e-06, "loss": 0.5645, "step": 4158 }, { "epoch": 0.5, "learning_rate": 4.764511351012725e-06, "loss": 0.7298, "step": 4159 }, { "epoch": 0.5, "learning_rate": 4.76437514653601e-06, "loss": 0.785, "step": 4160 }, { "epoch": 0.5, "learning_rate": 4.7642389046289835e-06, "loss": 1.0502, "step": 4161 }, { "epoch": 0.5, "learning_rate": 4.764102625293899e-06, "loss": 1.1343, "step": 4162 }, { "epoch": 0.5, "learning_rate": 4.763966308533007e-06, "loss": 0.5463, "step": 4163 }, { "epoch": 0.5, "learning_rate": 4.763829954348563e-06, "loss": 0.5526, "step": 4164 }, { "epoch": 0.5, "learning_rate": 4.763693562742818e-06, "loss": 0.6732, "step": 4165 }, { "epoch": 0.5, "learning_rate": 4.7635571337180305e-06, "loss": 0.9049, "step": 4166 }, { "epoch": 0.5, "learning_rate": 4.763420667276453e-06, "loss": 0.8898, "step": 4167 }, { "epoch": 0.5, "learning_rate": 4.763284163420342e-06, "loss": 1.2319, "step": 4168 }, { "epoch": 0.5, "learning_rate": 4.763147622151954e-06, "loss": 0.9448, "step": 4169 }, { "epoch": 0.5, "learning_rate": 4.763011043473545e-06, "loss": 0.6455, "step": 4170 }, { "epoch": 0.5, "learning_rate": 4.762874427387374e-06, "loss": 0.6561, "step": 4171 }, { "epoch": 0.5, "learning_rate": 4.762737773895699e-06, "loss": 0.7505, "step": 4172 }, { "epoch": 0.5, "learning_rate": 4.762601083000778e-06, "loss": 0.8512, "step": 4173 }, { "epoch": 0.5, "learning_rate": 4.762464354704873e-06, "loss": 1.0402, "step": 4174 }, { "epoch": 0.5, "learning_rate": 4.762327589010241e-06, "loss": 1.0781, "step": 4175 }, { "epoch": 0.5, "learning_rate": 4.762190785919145e-06, "loss": 0.6625, "step": 4176 }, { "epoch": 0.5, "learning_rate": 4.762053945433844e-06, "loss": 0.7049, "step": 4177 }, { "epoch": 0.5, "learning_rate": 4.7619170675566026e-06, "loss": 0.8209, "step": 4178 }, { "epoch": 0.5, "learning_rate": 4.761780152289683e-06, "loss": 0.8126, "step": 4179 }, { "epoch": 0.5, "learning_rate": 4.761643199635347e-06, "loss": 1.0392, "step": 4180 }, { "epoch": 0.5, "learning_rate": 4.761506209595859e-06, "loss": 1.1495, "step": 4181 }, { "epoch": 0.5, "learning_rate": 4.7613691821734834e-06, "loss": 0.6042, "step": 4182 }, { "epoch": 0.5, "learning_rate": 4.761232117370486e-06, "loss": 0.6471, "step": 4183 }, { "epoch": 0.5, "learning_rate": 4.761095015189131e-06, "loss": 0.7522, "step": 4184 }, { "epoch": 0.5, "learning_rate": 4.760957875631686e-06, "loss": 0.7566, "step": 4185 }, { "epoch": 0.5, "learning_rate": 4.760820698700418e-06, "loss": 1.0162, "step": 4186 }, { "epoch": 0.5, "learning_rate": 4.7606834843975945e-06, "loss": 1.1521, "step": 4187 }, { "epoch": 0.5, "learning_rate": 4.7605462327254824e-06, "loss": 0.666, "step": 4188 }, { "epoch": 0.5, "learning_rate": 4.760408943686352e-06, "loss": 0.577, "step": 4189 }, { "epoch": 0.5, "learning_rate": 4.760271617282473e-06, "loss": 0.7959, "step": 4190 }, { "epoch": 0.5, "learning_rate": 4.7601342535161124e-06, "loss": 0.9083, "step": 4191 }, { "epoch": 0.5, "learning_rate": 4.759996852389545e-06, "loss": 0.9991, "step": 4192 }, { "epoch": 0.5, "learning_rate": 4.759859413905038e-06, "loss": 1.0461, "step": 4193 }, { "epoch": 0.5, "learning_rate": 4.759721938064866e-06, "loss": 0.8255, "step": 4194 }, { "epoch": 0.5, "learning_rate": 4.759584424871302e-06, "loss": 0.5623, "step": 4195 }, { "epoch": 0.5, "learning_rate": 4.759446874326617e-06, "loss": 0.7698, "step": 4196 }, { "epoch": 0.5, "learning_rate": 4.759309286433085e-06, "loss": 0.8284, "step": 4197 }, { "epoch": 0.5, "learning_rate": 4.759171661192982e-06, "loss": 0.8955, "step": 4198 }, { "epoch": 0.5, "learning_rate": 4.759033998608581e-06, "loss": 1.0382, "step": 4199 }, { "epoch": 0.5, "learning_rate": 4.758896298682158e-06, "loss": 1.0657, "step": 4200 }, { "epoch": 0.5, "learning_rate": 4.7587585614159914e-06, "loss": 0.5811, "step": 4201 }, { "epoch": 0.5, "learning_rate": 4.758620786812355e-06, "loss": 0.7065, "step": 4202 }, { "epoch": 0.5, "learning_rate": 4.758482974873528e-06, "loss": 0.8237, "step": 4203 }, { "epoch": 0.5, "learning_rate": 4.758345125601788e-06, "loss": 0.8922, "step": 4204 }, { "epoch": 0.5, "learning_rate": 4.758207238999413e-06, "loss": 1.0637, "step": 4205 }, { "epoch": 0.5, "learning_rate": 4.758069315068683e-06, "loss": 1.0206, "step": 4206 }, { "epoch": 0.5, "learning_rate": 4.757931353811878e-06, "loss": 0.4895, "step": 4207 }, { "epoch": 0.5, "learning_rate": 4.757793355231278e-06, "loss": 0.6928, "step": 4208 }, { "epoch": 0.5, "learning_rate": 4.757655319329164e-06, "loss": 0.7575, "step": 4209 }, { "epoch": 0.5, "learning_rate": 4.757517246107819e-06, "loss": 0.8116, "step": 4210 }, { "epoch": 0.5, "learning_rate": 4.757379135569524e-06, "loss": 1.0857, "step": 4211 }, { "epoch": 0.5, "learning_rate": 4.7572409877165624e-06, "loss": 1.1064, "step": 4212 }, { "epoch": 0.5, "learning_rate": 4.757102802551219e-06, "loss": 0.6217, "step": 4213 }, { "epoch": 0.5, "learning_rate": 4.756964580075776e-06, "loss": 0.6788, "step": 4214 }, { "epoch": 0.5, "learning_rate": 4.756826320292519e-06, "loss": 0.6795, "step": 4215 }, { "epoch": 0.5, "learning_rate": 4.7566880232037334e-06, "loss": 0.6759, "step": 4216 }, { "epoch": 0.5, "learning_rate": 4.756549688811706e-06, "loss": 0.9112, "step": 4217 }, { "epoch": 0.5, "learning_rate": 4.756411317118723e-06, "loss": 1.3153, "step": 4218 }, { "epoch": 0.5, "learning_rate": 4.756272908127071e-06, "loss": 0.8364, "step": 4219 }, { "epoch": 0.5, "learning_rate": 4.756134461839038e-06, "loss": 0.7032, "step": 4220 }, { "epoch": 0.5, "learning_rate": 4.755995978256914e-06, "loss": 0.7038, "step": 4221 }, { "epoch": 0.5, "learning_rate": 4.7558574573829874e-06, "loss": 0.845, "step": 4222 }, { "epoch": 0.5, "learning_rate": 4.755718899219547e-06, "loss": 1.1115, "step": 4223 }, { "epoch": 0.5, "learning_rate": 4.755580303768884e-06, "loss": 1.2068, "step": 4224 }, { "epoch": 0.5, "learning_rate": 4.75544167103329e-06, "loss": 1.1252, "step": 4225 }, { "epoch": 0.5, "learning_rate": 4.755303001015056e-06, "loss": 0.659, "step": 4226 }, { "epoch": 0.5, "learning_rate": 4.755164293716474e-06, "loss": 0.7181, "step": 4227 }, { "epoch": 0.5, "learning_rate": 4.7550255491398354e-06, "loss": 0.7618, "step": 4228 }, { "epoch": 0.5, "learning_rate": 4.754886767287436e-06, "loss": 0.9525, "step": 4229 }, { "epoch": 0.5, "learning_rate": 4.75474794816157e-06, "loss": 1.106, "step": 4230 }, { "epoch": 0.5, "learning_rate": 4.754609091764531e-06, "loss": 1.0104, "step": 4231 }, { "epoch": 0.5, "learning_rate": 4.7544701980986155e-06, "loss": 0.5244, "step": 4232 }, { "epoch": 0.5, "learning_rate": 4.754331267166117e-06, "loss": 0.5594, "step": 4233 }, { "epoch": 0.5, "learning_rate": 4.7541922989693334e-06, "loss": 0.701, "step": 4234 }, { "epoch": 0.5, "learning_rate": 4.7540532935105625e-06, "loss": 0.7918, "step": 4235 }, { "epoch": 0.5, "learning_rate": 4.753914250792101e-06, "loss": 1.0901, "step": 4236 }, { "epoch": 0.5, "learning_rate": 4.753775170816249e-06, "loss": 1.1387, "step": 4237 }, { "epoch": 0.5, "learning_rate": 4.753636053585303e-06, "loss": 0.5913, "step": 4238 }, { "epoch": 0.5, "learning_rate": 4.753496899101565e-06, "loss": 0.7161, "step": 4239 }, { "epoch": 0.5, "learning_rate": 4.753357707367334e-06, "loss": 0.8124, "step": 4240 }, { "epoch": 0.51, "learning_rate": 4.75321847838491e-06, "loss": 0.8636, "step": 4241 }, { "epoch": 0.51, "learning_rate": 4.753079212156597e-06, "loss": 1.041, "step": 4242 }, { "epoch": 0.51, "learning_rate": 4.752939908684694e-06, "loss": 1.1563, "step": 4243 }, { "epoch": 0.51, "learning_rate": 4.752800567971506e-06, "loss": 0.7214, "step": 4244 }, { "epoch": 0.51, "learning_rate": 4.7526611900193355e-06, "loss": 0.5481, "step": 4245 }, { "epoch": 0.51, "learning_rate": 4.752521774830487e-06, "loss": 0.6724, "step": 4246 }, { "epoch": 0.51, "learning_rate": 4.752382322407265e-06, "loss": 0.6477, "step": 4247 }, { "epoch": 0.51, "learning_rate": 4.752242832751973e-06, "loss": 0.9312, "step": 4248 }, { "epoch": 0.51, "learning_rate": 4.752103305866919e-06, "loss": 1.0767, "step": 4249 }, { "epoch": 0.51, "learning_rate": 4.751963741754407e-06, "loss": 1.135, "step": 4250 }, { "epoch": 0.51, "learning_rate": 4.751824140416747e-06, "loss": 0.6463, "step": 4251 }, { "epoch": 0.51, "learning_rate": 4.751684501856244e-06, "loss": 0.7065, "step": 4252 }, { "epoch": 0.51, "learning_rate": 4.751544826075207e-06, "loss": 0.8517, "step": 4253 }, { "epoch": 0.51, "learning_rate": 4.751405113075945e-06, "loss": 0.9255, "step": 4254 }, { "epoch": 0.51, "learning_rate": 4.7512653628607685e-06, "loss": 1.0222, "step": 4255 }, { "epoch": 0.51, "learning_rate": 4.751125575431987e-06, "loss": 1.0013, "step": 4256 }, { "epoch": 0.51, "learning_rate": 4.75098575079191e-06, "loss": 0.6539, "step": 4257 }, { "epoch": 0.51, "learning_rate": 4.75084588894285e-06, "loss": 0.6616, "step": 4258 }, { "epoch": 0.51, "learning_rate": 4.750705989887118e-06, "loss": 0.8818, "step": 4259 }, { "epoch": 0.51, "learning_rate": 4.7505660536270275e-06, "loss": 0.7095, "step": 4260 }, { "epoch": 0.51, "learning_rate": 4.750426080164892e-06, "loss": 1.0336, "step": 4261 }, { "epoch": 0.51, "learning_rate": 4.750286069503024e-06, "loss": 1.1608, "step": 4262 }, { "epoch": 0.51, "learning_rate": 4.750146021643738e-06, "loss": 0.6277, "step": 4263 }, { "epoch": 0.51, "learning_rate": 4.75000593658935e-06, "loss": 0.7099, "step": 4264 }, { "epoch": 0.51, "learning_rate": 4.749865814342175e-06, "loss": 0.6666, "step": 4265 }, { "epoch": 0.51, "learning_rate": 4.749725654904529e-06, "loss": 0.8979, "step": 4266 }, { "epoch": 0.51, "learning_rate": 4.7495854582787295e-06, "loss": 0.9393, "step": 4267 }, { "epoch": 0.51, "learning_rate": 4.749445224467093e-06, "loss": 1.1456, "step": 4268 }, { "epoch": 0.51, "learning_rate": 4.74930495347194e-06, "loss": 0.687, "step": 4269 }, { "epoch": 0.51, "learning_rate": 4.7491646452955854e-06, "loss": 0.6837, "step": 4270 }, { "epoch": 0.51, "learning_rate": 4.749024299940351e-06, "loss": 0.7454, "step": 4271 }, { "epoch": 0.51, "learning_rate": 4.7488839174085566e-06, "loss": 0.764, "step": 4272 }, { "epoch": 0.51, "learning_rate": 4.748743497702521e-06, "loss": 0.9439, "step": 4273 }, { "epoch": 0.51, "learning_rate": 4.7486030408245685e-06, "loss": 1.1089, "step": 4274 }, { "epoch": 0.51, "learning_rate": 4.748462546777019e-06, "loss": 1.1711, "step": 4275 }, { "epoch": 0.51, "learning_rate": 4.748322015562194e-06, "loss": 0.6725, "step": 4276 }, { "epoch": 0.51, "learning_rate": 4.748181447182418e-06, "loss": 0.7135, "step": 4277 }, { "epoch": 0.51, "learning_rate": 4.748040841640013e-06, "loss": 0.968, "step": 4278 }, { "epoch": 0.51, "learning_rate": 4.7479001989373055e-06, "loss": 0.9661, "step": 4279 }, { "epoch": 0.51, "learning_rate": 4.747759519076619e-06, "loss": 1.0005, "step": 4280 }, { "epoch": 0.51, "learning_rate": 4.7476188020602786e-06, "loss": 1.0458, "step": 4281 }, { "epoch": 0.51, "learning_rate": 4.7474780478906115e-06, "loss": 0.5645, "step": 4282 }, { "epoch": 0.51, "learning_rate": 4.747337256569943e-06, "loss": 0.6513, "step": 4283 }, { "epoch": 0.51, "learning_rate": 4.7471964281006015e-06, "loss": 0.7008, "step": 4284 }, { "epoch": 0.51, "learning_rate": 4.747055562484914e-06, "loss": 0.805, "step": 4285 }, { "epoch": 0.51, "learning_rate": 4.7469146597252094e-06, "loss": 1.0709, "step": 4286 }, { "epoch": 0.51, "learning_rate": 4.7467737198238176e-06, "loss": 1.0842, "step": 4287 }, { "epoch": 0.51, "learning_rate": 4.746632742783067e-06, "loss": 0.554, "step": 4288 }, { "epoch": 0.51, "learning_rate": 4.74649172860529e-06, "loss": 0.5835, "step": 4289 }, { "epoch": 0.51, "learning_rate": 4.746350677292814e-06, "loss": 0.751, "step": 4290 }, { "epoch": 0.51, "learning_rate": 4.746209588847974e-06, "loss": 0.7872, "step": 4291 }, { "epoch": 0.51, "learning_rate": 4.746068463273101e-06, "loss": 0.8992, "step": 4292 }, { "epoch": 0.51, "learning_rate": 4.7459273005705285e-06, "loss": 1.1458, "step": 4293 }, { "epoch": 0.51, "learning_rate": 4.745786100742588e-06, "loss": 0.7243, "step": 4294 }, { "epoch": 0.51, "learning_rate": 4.7456448637916155e-06, "loss": 0.6801, "step": 4295 }, { "epoch": 0.51, "learning_rate": 4.745503589719945e-06, "loss": 0.7781, "step": 4296 }, { "epoch": 0.51, "learning_rate": 4.7453622785299115e-06, "loss": 0.7347, "step": 4297 }, { "epoch": 0.51, "learning_rate": 4.7452209302238516e-06, "loss": 0.886, "step": 4298 }, { "epoch": 0.51, "learning_rate": 4.745079544804101e-06, "loss": 1.1704, "step": 4299 }, { "epoch": 0.51, "learning_rate": 4.744938122272997e-06, "loss": 1.306, "step": 4300 }, { "epoch": 0.51, "learning_rate": 4.7447966626328774e-06, "loss": 0.7025, "step": 4301 }, { "epoch": 0.51, "learning_rate": 4.7446551658860816e-06, "loss": 0.7745, "step": 4302 }, { "epoch": 0.51, "learning_rate": 4.744513632034947e-06, "loss": 0.8024, "step": 4303 }, { "epoch": 0.51, "learning_rate": 4.744372061081813e-06, "loss": 0.9319, "step": 4304 }, { "epoch": 0.51, "learning_rate": 4.7442304530290215e-06, "loss": 0.9426, "step": 4305 }, { "epoch": 0.51, "learning_rate": 4.7440888078789125e-06, "loss": 1.0681, "step": 4306 }, { "epoch": 0.51, "learning_rate": 4.7439471256338265e-06, "loss": 0.5288, "step": 4307 }, { "epoch": 0.51, "learning_rate": 4.7438054062961055e-06, "loss": 0.6243, "step": 4308 }, { "epoch": 0.51, "learning_rate": 4.743663649868094e-06, "loss": 0.6732, "step": 4309 }, { "epoch": 0.51, "learning_rate": 4.7435218563521345e-06, "loss": 0.7736, "step": 4310 }, { "epoch": 0.51, "learning_rate": 4.7433800257505695e-06, "loss": 1.0208, "step": 4311 }, { "epoch": 0.51, "learning_rate": 4.743238158065745e-06, "loss": 1.1953, "step": 4312 }, { "epoch": 0.51, "learning_rate": 4.743096253300005e-06, "loss": 0.651, "step": 4313 }, { "epoch": 0.51, "learning_rate": 4.742954311455696e-06, "loss": 0.6969, "step": 4314 }, { "epoch": 0.51, "learning_rate": 4.742812332535165e-06, "loss": 0.8339, "step": 4315 }, { "epoch": 0.51, "learning_rate": 4.742670316540756e-06, "loss": 0.825, "step": 4316 }, { "epoch": 0.51, "learning_rate": 4.74252826347482e-06, "loss": 1.0143, "step": 4317 }, { "epoch": 0.51, "learning_rate": 4.742386173339703e-06, "loss": 1.2315, "step": 4318 }, { "epoch": 0.51, "learning_rate": 4.742244046137755e-06, "loss": 0.855, "step": 4319 }, { "epoch": 0.51, "learning_rate": 4.7421018818713236e-06, "loss": 0.6332, "step": 4320 }, { "epoch": 0.51, "learning_rate": 4.74195968054276e-06, "loss": 0.7363, "step": 4321 }, { "epoch": 0.51, "learning_rate": 4.741817442154416e-06, "loss": 0.7789, "step": 4322 }, { "epoch": 0.51, "learning_rate": 4.741675166708641e-06, "loss": 0.797, "step": 4323 }, { "epoch": 0.51, "learning_rate": 4.7415328542077876e-06, "loss": 1.0883, "step": 4324 }, { "epoch": 0.52, "learning_rate": 4.741390504654208e-06, "loss": 1.226, "step": 4325 }, { "epoch": 0.52, "learning_rate": 4.741248118050255e-06, "loss": 0.5689, "step": 4326 }, { "epoch": 0.52, "learning_rate": 4.7411056943982824e-06, "loss": 0.7479, "step": 4327 }, { "epoch": 0.52, "learning_rate": 4.740963233700645e-06, "loss": 0.7391, "step": 4328 }, { "epoch": 0.52, "learning_rate": 4.7408207359596976e-06, "loss": 0.9843, "step": 4329 }, { "epoch": 0.52, "learning_rate": 4.740678201177795e-06, "loss": 1.0119, "step": 4330 }, { "epoch": 0.52, "learning_rate": 4.740535629357294e-06, "loss": 1.0636, "step": 4331 }, { "epoch": 0.52, "learning_rate": 4.740393020500551e-06, "loss": 0.5685, "step": 4332 }, { "epoch": 0.52, "learning_rate": 4.740250374609923e-06, "loss": 0.594, "step": 4333 }, { "epoch": 0.52, "learning_rate": 4.7401076916877685e-06, "loss": 0.6561, "step": 4334 }, { "epoch": 0.52, "learning_rate": 4.739964971736446e-06, "loss": 0.7523, "step": 4335 }, { "epoch": 0.52, "learning_rate": 4.739822214758314e-06, "loss": 1.0299, "step": 4336 }, { "epoch": 0.52, "learning_rate": 4.739679420755734e-06, "loss": 1.1045, "step": 4337 }, { "epoch": 0.52, "learning_rate": 4.739536589731064e-06, "loss": 0.583, "step": 4338 }, { "epoch": 0.52, "learning_rate": 4.739393721686667e-06, "loss": 0.5516, "step": 4339 }, { "epoch": 0.52, "learning_rate": 4.7392508166249045e-06, "loss": 0.6963, "step": 4340 }, { "epoch": 0.52, "learning_rate": 4.739107874548137e-06, "loss": 0.929, "step": 4341 }, { "epoch": 0.52, "learning_rate": 4.738964895458729e-06, "loss": 0.8755, "step": 4342 }, { "epoch": 0.52, "learning_rate": 4.7388218793590426e-06, "loss": 1.2518, "step": 4343 }, { "epoch": 0.52, "learning_rate": 4.7386788262514435e-06, "loss": 0.8092, "step": 4344 }, { "epoch": 0.52, "learning_rate": 4.7385357361382956e-06, "loss": 0.5085, "step": 4345 }, { "epoch": 0.52, "learning_rate": 4.7383926090219636e-06, "loss": 0.7466, "step": 4346 }, { "epoch": 0.52, "learning_rate": 4.738249444904814e-06, "loss": 0.8531, "step": 4347 }, { "epoch": 0.52, "learning_rate": 4.738106243789215e-06, "loss": 0.8566, "step": 4348 }, { "epoch": 0.52, "learning_rate": 4.7379630056775295e-06, "loss": 1.0847, "step": 4349 }, { "epoch": 0.52, "learning_rate": 4.737819730572129e-06, "loss": 1.1655, "step": 4350 }, { "epoch": 0.52, "learning_rate": 4.7376764184753805e-06, "loss": 0.6811, "step": 4351 }, { "epoch": 0.52, "learning_rate": 4.7375330693896525e-06, "loss": 0.7446, "step": 4352 }, { "epoch": 0.52, "learning_rate": 4.737389683317316e-06, "loss": 0.7854, "step": 4353 }, { "epoch": 0.52, "learning_rate": 4.7372462602607394e-06, "loss": 0.9141, "step": 4354 }, { "epoch": 0.52, "learning_rate": 4.737102800222295e-06, "loss": 0.9323, "step": 4355 }, { "epoch": 0.52, "learning_rate": 4.7369593032043535e-06, "loss": 1.11, "step": 4356 }, { "epoch": 0.52, "learning_rate": 4.7368157692092875e-06, "loss": 0.7984, "step": 4357 }, { "epoch": 0.52, "learning_rate": 4.736672198239468e-06, "loss": 0.6901, "step": 4358 }, { "epoch": 0.52, "learning_rate": 4.7365285902972705e-06, "loss": 0.7753, "step": 4359 }, { "epoch": 0.52, "learning_rate": 4.736384945385067e-06, "loss": 0.769, "step": 4360 }, { "epoch": 0.52, "learning_rate": 4.736241263505233e-06, "loss": 1.0746, "step": 4361 }, { "epoch": 0.52, "learning_rate": 4.7360975446601434e-06, "loss": 1.1938, "step": 4362 }, { "epoch": 0.52, "learning_rate": 4.735953788852173e-06, "loss": 0.5291, "step": 4363 }, { "epoch": 0.52, "learning_rate": 4.7358099960837e-06, "loss": 0.705, "step": 4364 }, { "epoch": 0.52, "learning_rate": 4.7356661663570995e-06, "loss": 0.7399, "step": 4365 }, { "epoch": 0.52, "learning_rate": 4.735522299674749e-06, "loss": 0.7449, "step": 4366 }, { "epoch": 0.52, "learning_rate": 4.735378396039028e-06, "loss": 0.9584, "step": 4367 }, { "epoch": 0.52, "learning_rate": 4.735234455452316e-06, "loss": 1.1501, "step": 4368 }, { "epoch": 0.52, "learning_rate": 4.735090477916989e-06, "loss": 0.8415, "step": 4369 }, { "epoch": 0.52, "learning_rate": 4.734946463435429e-06, "loss": 0.6436, "step": 4370 }, { "epoch": 0.52, "learning_rate": 4.734802412010018e-06, "loss": 0.6798, "step": 4371 }, { "epoch": 0.52, "learning_rate": 4.734658323643135e-06, "loss": 0.7672, "step": 4372 }, { "epoch": 0.52, "learning_rate": 4.734514198337161e-06, "loss": 0.9817, "step": 4373 }, { "epoch": 0.52, "learning_rate": 4.734370036094481e-06, "loss": 1.0978, "step": 4374 }, { "epoch": 0.52, "learning_rate": 4.734225836917477e-06, "loss": 1.1871, "step": 4375 }, { "epoch": 0.52, "learning_rate": 4.734081600808531e-06, "loss": 0.6609, "step": 4376 }, { "epoch": 0.52, "learning_rate": 4.733937327770029e-06, "loss": 0.6501, "step": 4377 }, { "epoch": 0.52, "learning_rate": 4.733793017804357e-06, "loss": 0.879, "step": 4378 }, { "epoch": 0.52, "learning_rate": 4.733648670913898e-06, "loss": 0.9529, "step": 4379 }, { "epoch": 0.52, "learning_rate": 4.73350428710104e-06, "loss": 1.0264, "step": 4380 }, { "epoch": 0.52, "learning_rate": 4.7333598663681664e-06, "loss": 1.0218, "step": 4381 }, { "epoch": 0.52, "learning_rate": 4.733215408717669e-06, "loss": 0.7709, "step": 4382 }, { "epoch": 0.52, "learning_rate": 4.733070914151933e-06, "loss": 0.6742, "step": 4383 }, { "epoch": 0.52, "learning_rate": 4.732926382673347e-06, "loss": 0.7172, "step": 4384 }, { "epoch": 0.52, "learning_rate": 4.732781814284301e-06, "loss": 0.9518, "step": 4385 }, { "epoch": 0.52, "learning_rate": 4.732637208987184e-06, "loss": 0.9594, "step": 4386 }, { "epoch": 0.52, "learning_rate": 4.732492566784387e-06, "loss": 1.0811, "step": 4387 }, { "epoch": 0.52, "learning_rate": 4.7323478876783e-06, "loss": 0.5745, "step": 4388 }, { "epoch": 0.52, "learning_rate": 4.732203171671316e-06, "loss": 0.7453, "step": 4389 }, { "epoch": 0.52, "learning_rate": 4.732058418765825e-06, "loss": 0.766, "step": 4390 }, { "epoch": 0.52, "learning_rate": 4.731913628964222e-06, "loss": 0.8256, "step": 4391 }, { "epoch": 0.52, "learning_rate": 4.731768802268899e-06, "loss": 0.8403, "step": 4392 }, { "epoch": 0.52, "learning_rate": 4.73162393868225e-06, "loss": 1.1556, "step": 4393 }, { "epoch": 0.52, "learning_rate": 4.731479038206671e-06, "loss": 0.7793, "step": 4394 }, { "epoch": 0.52, "learning_rate": 4.731334100844556e-06, "loss": 0.5049, "step": 4395 }, { "epoch": 0.52, "learning_rate": 4.731189126598301e-06, "loss": 0.8242, "step": 4396 }, { "epoch": 0.52, "learning_rate": 4.7310441154703025e-06, "loss": 0.7117, "step": 4397 }, { "epoch": 0.52, "learning_rate": 4.730899067462958e-06, "loss": 0.9628, "step": 4398 }, { "epoch": 0.52, "learning_rate": 4.730753982578664e-06, "loss": 1.071, "step": 4399 }, { "epoch": 0.52, "learning_rate": 4.73060886081982e-06, "loss": 1.0128, "step": 4400 }, { "epoch": 0.52, "learning_rate": 4.730463702188824e-06, "loss": 0.4744, "step": 4401 }, { "epoch": 0.52, "learning_rate": 4.730318506688076e-06, "loss": 0.6618, "step": 4402 }, { "epoch": 0.52, "learning_rate": 4.7301732743199755e-06, "loss": 0.7344, "step": 4403 }, { "epoch": 0.52, "learning_rate": 4.730028005086924e-06, "loss": 0.8027, "step": 4404 }, { "epoch": 0.52, "learning_rate": 4.729882698991322e-06, "loss": 0.937, "step": 4405 }, { "epoch": 0.52, "learning_rate": 4.729737356035573e-06, "loss": 1.2997, "step": 4406 }, { "epoch": 0.52, "learning_rate": 4.729591976222078e-06, "loss": 0.6422, "step": 4407 }, { "epoch": 0.52, "learning_rate": 4.72944655955324e-06, "loss": 0.6968, "step": 4408 }, { "epoch": 0.53, "learning_rate": 4.729301106031464e-06, "loss": 0.7968, "step": 4409 }, { "epoch": 0.53, "learning_rate": 4.729155615659153e-06, "loss": 0.8132, "step": 4410 }, { "epoch": 0.53, "learning_rate": 4.729010088438713e-06, "loss": 0.9512, "step": 4411 }, { "epoch": 0.53, "learning_rate": 4.72886452437255e-06, "loss": 1.1336, "step": 4412 }, { "epoch": 0.53, "learning_rate": 4.728718923463068e-06, "loss": 0.7191, "step": 4413 }, { "epoch": 0.53, "learning_rate": 4.728573285712675e-06, "loss": 0.6822, "step": 4414 }, { "epoch": 0.53, "learning_rate": 4.72842761112378e-06, "loss": 0.681, "step": 4415 }, { "epoch": 0.53, "learning_rate": 4.728281899698789e-06, "loss": 0.7689, "step": 4416 }, { "epoch": 0.53, "learning_rate": 4.728136151440112e-06, "loss": 0.9353, "step": 4417 }, { "epoch": 0.53, "learning_rate": 4.727990366350156e-06, "loss": 1.1174, "step": 4418 }, { "epoch": 0.53, "learning_rate": 4.727844544431333e-06, "loss": 0.6487, "step": 4419 }, { "epoch": 0.53, "learning_rate": 4.727698685686053e-06, "loss": 0.6963, "step": 4420 }, { "epoch": 0.53, "learning_rate": 4.7275527901167266e-06, "loss": 0.7219, "step": 4421 }, { "epoch": 0.53, "learning_rate": 4.727406857725765e-06, "loss": 0.8127, "step": 4422 }, { "epoch": 0.53, "learning_rate": 4.727260888515582e-06, "loss": 0.8586, "step": 4423 }, { "epoch": 0.53, "learning_rate": 4.727114882488589e-06, "loss": 1.0379, "step": 4424 }, { "epoch": 0.53, "learning_rate": 4.726968839647202e-06, "loss": 1.186, "step": 4425 }, { "epoch": 0.53, "learning_rate": 4.726822759993831e-06, "loss": 0.6272, "step": 4426 }, { "epoch": 0.53, "learning_rate": 4.726676643530893e-06, "loss": 0.7717, "step": 4427 }, { "epoch": 0.53, "learning_rate": 4.726530490260805e-06, "loss": 0.8312, "step": 4428 }, { "epoch": 0.53, "learning_rate": 4.726384300185979e-06, "loss": 0.8863, "step": 4429 }, { "epoch": 0.53, "learning_rate": 4.726238073308835e-06, "loss": 0.9782, "step": 4430 }, { "epoch": 0.53, "learning_rate": 4.726091809631789e-06, "loss": 1.1414, "step": 4431 }, { "epoch": 0.53, "learning_rate": 4.725945509157258e-06, "loss": 0.5336, "step": 4432 }, { "epoch": 0.53, "learning_rate": 4.725799171887661e-06, "loss": 0.6003, "step": 4433 }, { "epoch": 0.53, "learning_rate": 4.725652797825418e-06, "loss": 0.7663, "step": 4434 }, { "epoch": 0.53, "learning_rate": 4.725506386972945e-06, "loss": 0.7534, "step": 4435 }, { "epoch": 0.53, "learning_rate": 4.725359939332667e-06, "loss": 0.9758, "step": 4436 }, { "epoch": 0.53, "learning_rate": 4.725213454907002e-06, "loss": 1.3157, "step": 4437 }, { "epoch": 0.53, "learning_rate": 4.725066933698372e-06, "loss": 0.7757, "step": 4438 }, { "epoch": 0.53, "learning_rate": 4.724920375709198e-06, "loss": 0.6064, "step": 4439 }, { "epoch": 0.53, "learning_rate": 4.724773780941905e-06, "loss": 0.7711, "step": 4440 }, { "epoch": 0.53, "learning_rate": 4.724627149398913e-06, "loss": 0.9109, "step": 4441 }, { "epoch": 0.53, "learning_rate": 4.7244804810826485e-06, "loss": 0.9946, "step": 4442 }, { "epoch": 0.53, "learning_rate": 4.724333775995535e-06, "loss": 1.1009, "step": 4443 }, { "epoch": 0.53, "learning_rate": 4.724187034139997e-06, "loss": 0.907, "step": 4444 }, { "epoch": 0.53, "learning_rate": 4.724040255518461e-06, "loss": 0.4723, "step": 4445 }, { "epoch": 0.53, "learning_rate": 4.723893440133353e-06, "loss": 0.8473, "step": 4446 }, { "epoch": 0.53, "learning_rate": 4.7237465879871e-06, "loss": 0.7036, "step": 4447 }, { "epoch": 0.53, "learning_rate": 4.72359969908213e-06, "loss": 0.9041, "step": 4448 }, { "epoch": 0.53, "learning_rate": 4.723452773420869e-06, "loss": 1.1531, "step": 4449 }, { "epoch": 0.53, "learning_rate": 4.723305811005748e-06, "loss": 1.2667, "step": 4450 }, { "epoch": 0.53, "learning_rate": 4.723158811839195e-06, "loss": 0.6584, "step": 4451 }, { "epoch": 0.53, "learning_rate": 4.723011775923641e-06, "loss": 0.7438, "step": 4452 }, { "epoch": 0.53, "learning_rate": 4.722864703261515e-06, "loss": 0.8616, "step": 4453 }, { "epoch": 0.53, "learning_rate": 4.722717593855249e-06, "loss": 0.8163, "step": 4454 }, { "epoch": 0.53, "learning_rate": 4.722570447707275e-06, "loss": 1.056, "step": 4455 }, { "epoch": 0.53, "learning_rate": 4.722423264820024e-06, "loss": 1.2097, "step": 4456 }, { "epoch": 0.53, "learning_rate": 4.722276045195931e-06, "loss": 0.5914, "step": 4457 }, { "epoch": 0.53, "learning_rate": 4.722128788837429e-06, "loss": 0.6808, "step": 4458 }, { "epoch": 0.53, "learning_rate": 4.72198149574695e-06, "loss": 0.7589, "step": 4459 }, { "epoch": 0.53, "learning_rate": 4.7218341659269314e-06, "loss": 0.933, "step": 4460 }, { "epoch": 0.53, "learning_rate": 4.721686799379808e-06, "loss": 1.0424, "step": 4461 }, { "epoch": 0.53, "learning_rate": 4.721539396108015e-06, "loss": 1.1464, "step": 4462 }, { "epoch": 0.53, "learning_rate": 4.72139195611399e-06, "loss": 0.5838, "step": 4463 }, { "epoch": 0.53, "learning_rate": 4.721244479400169e-06, "loss": 0.6194, "step": 4464 }, { "epoch": 0.53, "learning_rate": 4.72109696596899e-06, "loss": 0.5999, "step": 4465 }, { "epoch": 0.53, "learning_rate": 4.720949415822892e-06, "loss": 0.719, "step": 4466 }, { "epoch": 0.53, "learning_rate": 4.720801828964314e-06, "loss": 0.9712, "step": 4467 }, { "epoch": 0.53, "learning_rate": 4.720654205395695e-06, "loss": 1.1603, "step": 4468 }, { "epoch": 0.53, "learning_rate": 4.720506545119477e-06, "loss": 0.8265, "step": 4469 }, { "epoch": 0.53, "learning_rate": 4.720358848138098e-06, "loss": 0.5974, "step": 4470 }, { "epoch": 0.53, "learning_rate": 4.720211114454002e-06, "loss": 0.7397, "step": 4471 }, { "epoch": 0.53, "learning_rate": 4.720063344069629e-06, "loss": 0.83, "step": 4472 }, { "epoch": 0.53, "learning_rate": 4.719915536987423e-06, "loss": 0.7564, "step": 4473 }, { "epoch": 0.53, "learning_rate": 4.719767693209827e-06, "loss": 0.9618, "step": 4474 }, { "epoch": 0.53, "learning_rate": 4.719619812739284e-06, "loss": 1.2819, "step": 4475 }, { "epoch": 0.53, "learning_rate": 4.71947189557824e-06, "loss": 0.6396, "step": 4476 }, { "epoch": 0.53, "learning_rate": 4.7193239417291406e-06, "loss": 0.7332, "step": 4477 }, { "epoch": 0.53, "learning_rate": 4.719175951194428e-06, "loss": 0.7669, "step": 4478 }, { "epoch": 0.53, "learning_rate": 4.7190279239765526e-06, "loss": 0.9666, "step": 4479 }, { "epoch": 0.53, "learning_rate": 4.718879860077958e-06, "loss": 0.9338, "step": 4480 }, { "epoch": 0.53, "learning_rate": 4.718731759501094e-06, "loss": 0.9537, "step": 4481 }, { "epoch": 0.53, "learning_rate": 4.718583622248407e-06, "loss": 0.5085, "step": 4482 }, { "epoch": 0.53, "learning_rate": 4.718435448322347e-06, "loss": 0.5727, "step": 4483 }, { "epoch": 0.53, "learning_rate": 4.718287237725363e-06, "loss": 0.7741, "step": 4484 }, { "epoch": 0.53, "learning_rate": 4.718138990459905e-06, "loss": 0.7316, "step": 4485 }, { "epoch": 0.53, "learning_rate": 4.7179907065284236e-06, "loss": 0.8532, "step": 4486 }, { "epoch": 0.53, "learning_rate": 4.717842385933369e-06, "loss": 1.1194, "step": 4487 }, { "epoch": 0.53, "learning_rate": 4.717694028677194e-06, "loss": 0.791, "step": 4488 }, { "epoch": 0.53, "learning_rate": 4.71754563476235e-06, "loss": 0.7014, "step": 4489 }, { "epoch": 0.53, "learning_rate": 4.717397204191292e-06, "loss": 0.7102, "step": 4490 }, { "epoch": 0.53, "learning_rate": 4.717248736966471e-06, "loss": 0.6935, "step": 4491 }, { "epoch": 0.53, "learning_rate": 4.7171002330903424e-06, "loss": 0.9988, "step": 4492 }, { "epoch": 0.54, "learning_rate": 4.716951692565361e-06, "loss": 1.0872, "step": 4493 }, { "epoch": 0.54, "learning_rate": 4.716803115393983e-06, "loss": 0.7481, "step": 4494 }, { "epoch": 0.54, "learning_rate": 4.7166545015786625e-06, "loss": 0.5237, "step": 4495 }, { "epoch": 0.54, "learning_rate": 4.716505851121857e-06, "loss": 0.6038, "step": 4496 }, { "epoch": 0.54, "learning_rate": 4.7163571640260246e-06, "loss": 0.6688, "step": 4497 }, { "epoch": 0.54, "learning_rate": 4.716208440293622e-06, "loss": 0.9762, "step": 4498 }, { "epoch": 0.54, "learning_rate": 4.716059679927108e-06, "loss": 1.1415, "step": 4499 }, { "epoch": 0.54, "learning_rate": 4.715910882928941e-06, "loss": 1.1046, "step": 4500 }, { "epoch": 0.54, "learning_rate": 4.715762049301582e-06, "loss": 0.5758, "step": 4501 }, { "epoch": 0.54, "learning_rate": 4.7156131790474904e-06, "loss": 0.6634, "step": 4502 }, { "epoch": 0.54, "learning_rate": 4.715464272169125e-06, "loss": 0.8327, "step": 4503 }, { "epoch": 0.54, "learning_rate": 4.715315328668952e-06, "loss": 0.8698, "step": 4504 }, { "epoch": 0.54, "learning_rate": 4.715166348549429e-06, "loss": 1.1015, "step": 4505 }, { "epoch": 0.54, "learning_rate": 4.715017331813021e-06, "loss": 1.1828, "step": 4506 }, { "epoch": 0.54, "learning_rate": 4.714868278462191e-06, "loss": 0.5549, "step": 4507 }, { "epoch": 0.54, "learning_rate": 4.714719188499403e-06, "loss": 0.6937, "step": 4508 }, { "epoch": 0.54, "learning_rate": 4.714570061927119e-06, "loss": 0.7051, "step": 4509 }, { "epoch": 0.54, "learning_rate": 4.714420898747808e-06, "loss": 0.8266, "step": 4510 }, { "epoch": 0.54, "learning_rate": 4.714271698963932e-06, "loss": 1.0528, "step": 4511 }, { "epoch": 0.54, "learning_rate": 4.714122462577961e-06, "loss": 1.0869, "step": 4512 }, { "epoch": 0.54, "learning_rate": 4.713973189592358e-06, "loss": 0.6195, "step": 4513 }, { "epoch": 0.54, "learning_rate": 4.713823880009593e-06, "loss": 0.5913, "step": 4514 }, { "epoch": 0.54, "learning_rate": 4.713674533832135e-06, "loss": 0.7374, "step": 4515 }, { "epoch": 0.54, "learning_rate": 4.713525151062449e-06, "loss": 0.8616, "step": 4516 }, { "epoch": 0.54, "learning_rate": 4.713375731703008e-06, "loss": 0.8652, "step": 4517 }, { "epoch": 0.54, "learning_rate": 4.71322627575628e-06, "loss": 1.0475, "step": 4518 }, { "epoch": 0.54, "learning_rate": 4.713076783224736e-06, "loss": 0.853, "step": 4519 }, { "epoch": 0.54, "learning_rate": 4.712927254110847e-06, "loss": 0.6622, "step": 4520 }, { "epoch": 0.54, "learning_rate": 4.712777688417085e-06, "loss": 0.6213, "step": 4521 }, { "epoch": 0.54, "learning_rate": 4.712628086145923e-06, "loss": 0.7732, "step": 4522 }, { "epoch": 0.54, "learning_rate": 4.712478447299831e-06, "loss": 0.8885, "step": 4523 }, { "epoch": 0.54, "learning_rate": 4.712328771881286e-06, "loss": 1.1174, "step": 4524 }, { "epoch": 0.54, "learning_rate": 4.71217905989276e-06, "loss": 0.9649, "step": 4525 }, { "epoch": 0.54, "learning_rate": 4.7120293113367296e-06, "loss": 0.685, "step": 4526 }, { "epoch": 0.54, "learning_rate": 4.711879526215668e-06, "loss": 0.6037, "step": 4527 }, { "epoch": 0.54, "learning_rate": 4.711729704532053e-06, "loss": 0.7453, "step": 4528 }, { "epoch": 0.54, "learning_rate": 4.71157984628836e-06, "loss": 0.8501, "step": 4529 }, { "epoch": 0.54, "learning_rate": 4.711429951487066e-06, "loss": 1.0281, "step": 4530 }, { "epoch": 0.54, "learning_rate": 4.711280020130651e-06, "loss": 1.3248, "step": 4531 }, { "epoch": 0.54, "learning_rate": 4.7111300522215904e-06, "loss": 0.5124, "step": 4532 }, { "epoch": 0.54, "learning_rate": 4.710980047762365e-06, "loss": 0.6352, "step": 4533 }, { "epoch": 0.54, "learning_rate": 4.710830006755454e-06, "loss": 0.7104, "step": 4534 }, { "epoch": 0.54, "learning_rate": 4.710679929203337e-06, "loss": 0.8346, "step": 4535 }, { "epoch": 0.54, "learning_rate": 4.710529815108496e-06, "loss": 0.9916, "step": 4536 }, { "epoch": 0.54, "learning_rate": 4.710379664473411e-06, "loss": 1.2153, "step": 4537 }, { "epoch": 0.54, "learning_rate": 4.710229477300564e-06, "loss": 0.7354, "step": 4538 }, { "epoch": 0.54, "learning_rate": 4.71007925359244e-06, "loss": 0.7165, "step": 4539 }, { "epoch": 0.54, "learning_rate": 4.709928993351519e-06, "loss": 0.7727, "step": 4540 }, { "epoch": 0.54, "learning_rate": 4.709778696580287e-06, "loss": 0.8259, "step": 4541 }, { "epoch": 0.54, "learning_rate": 4.709628363281228e-06, "loss": 0.8823, "step": 4542 }, { "epoch": 0.54, "learning_rate": 4.7094779934568265e-06, "loss": 1.14, "step": 4543 }, { "epoch": 0.54, "learning_rate": 4.709327587109568e-06, "loss": 0.5741, "step": 4544 }, { "epoch": 0.54, "learning_rate": 4.7091771442419396e-06, "loss": 0.685, "step": 4545 }, { "epoch": 0.54, "learning_rate": 4.709026664856428e-06, "loss": 0.6938, "step": 4546 }, { "epoch": 0.54, "learning_rate": 4.708876148955519e-06, "loss": 0.7392, "step": 4547 }, { "epoch": 0.54, "learning_rate": 4.708725596541703e-06, "loss": 0.883, "step": 4548 }, { "epoch": 0.54, "learning_rate": 4.708575007617467e-06, "loss": 1.1136, "step": 4549 }, { "epoch": 0.54, "learning_rate": 4.708424382185301e-06, "loss": 1.1598, "step": 4550 }, { "epoch": 0.54, "learning_rate": 4.708273720247694e-06, "loss": 0.6353, "step": 4551 }, { "epoch": 0.54, "learning_rate": 4.708123021807139e-06, "loss": 0.7355, "step": 4552 }, { "epoch": 0.54, "learning_rate": 4.7079722868661235e-06, "loss": 0.9066, "step": 4553 }, { "epoch": 0.54, "learning_rate": 4.707821515427142e-06, "loss": 0.9625, "step": 4554 }, { "epoch": 0.54, "learning_rate": 4.707670707492684e-06, "loss": 0.9559, "step": 4555 }, { "epoch": 0.54, "learning_rate": 4.707519863065245e-06, "loss": 1.2261, "step": 4556 }, { "epoch": 0.54, "learning_rate": 4.707368982147318e-06, "loss": 0.6306, "step": 4557 }, { "epoch": 0.54, "learning_rate": 4.707218064741396e-06, "loss": 0.6796, "step": 4558 }, { "epoch": 0.54, "learning_rate": 4.7070671108499735e-06, "loss": 0.7321, "step": 4559 }, { "epoch": 0.54, "learning_rate": 4.706916120475547e-06, "loss": 0.8265, "step": 4560 }, { "epoch": 0.54, "learning_rate": 4.7067650936206125e-06, "loss": 1.1138, "step": 4561 }, { "epoch": 0.54, "learning_rate": 4.7066140302876645e-06, "loss": 1.161, "step": 4562 }, { "epoch": 0.54, "learning_rate": 4.706462930479202e-06, "loss": 0.7155, "step": 4563 }, { "epoch": 0.54, "learning_rate": 4.7063117941977224e-06, "loss": 0.7025, "step": 4564 }, { "epoch": 0.54, "learning_rate": 4.706160621445723e-06, "loss": 0.7157, "step": 4565 }, { "epoch": 0.54, "learning_rate": 4.706009412225705e-06, "loss": 0.8669, "step": 4566 }, { "epoch": 0.54, "learning_rate": 4.705858166540165e-06, "loss": 0.9243, "step": 4567 }, { "epoch": 0.54, "learning_rate": 4.705706884391604e-06, "loss": 1.0084, "step": 4568 }, { "epoch": 0.54, "learning_rate": 4.7055555657825246e-06, "loss": 0.7586, "step": 4569 }, { "epoch": 0.54, "learning_rate": 4.705404210715426e-06, "loss": 0.7245, "step": 4570 }, { "epoch": 0.54, "learning_rate": 4.705252819192811e-06, "loss": 0.763, "step": 4571 }, { "epoch": 0.54, "learning_rate": 4.705101391217181e-06, "loss": 0.7852, "step": 4572 }, { "epoch": 0.54, "learning_rate": 4.70494992679104e-06, "loss": 0.8058, "step": 4573 }, { "epoch": 0.54, "learning_rate": 4.704798425916892e-06, "loss": 0.9682, "step": 4574 }, { "epoch": 0.54, "learning_rate": 4.704646888597242e-06, "loss": 1.2778, "step": 4575 }, { "epoch": 0.54, "learning_rate": 4.704495314834593e-06, "loss": 0.5947, "step": 4576 }, { "epoch": 0.55, "learning_rate": 4.704343704631451e-06, "loss": 0.6303, "step": 4577 }, { "epoch": 0.55, "learning_rate": 4.704192057990323e-06, "loss": 0.7487, "step": 4578 }, { "epoch": 0.55, "learning_rate": 4.704040374913717e-06, "loss": 0.86, "step": 4579 }, { "epoch": 0.55, "learning_rate": 4.703888655404136e-06, "loss": 0.9691, "step": 4580 }, { "epoch": 0.55, "learning_rate": 4.7037368994640916e-06, "loss": 1.2056, "step": 4581 }, { "epoch": 0.55, "learning_rate": 4.703585107096092e-06, "loss": 0.5688, "step": 4582 }, { "epoch": 0.55, "learning_rate": 4.703433278302645e-06, "loss": 0.8564, "step": 4583 }, { "epoch": 0.55, "learning_rate": 4.7032814130862616e-06, "loss": 0.6874, "step": 4584 }, { "epoch": 0.55, "learning_rate": 4.703129511449452e-06, "loss": 0.8223, "step": 4585 }, { "epoch": 0.55, "learning_rate": 4.702977573394725e-06, "loss": 1.0145, "step": 4586 }, { "epoch": 0.55, "learning_rate": 4.702825598924595e-06, "loss": 1.1019, "step": 4587 }, { "epoch": 0.55, "learning_rate": 4.7026735880415734e-06, "loss": 0.7406, "step": 4588 }, { "epoch": 0.55, "learning_rate": 4.7025215407481716e-06, "loss": 0.6235, "step": 4589 }, { "epoch": 0.55, "learning_rate": 4.702369457046904e-06, "loss": 0.7779, "step": 4590 }, { "epoch": 0.55, "learning_rate": 4.702217336940285e-06, "loss": 0.7063, "step": 4591 }, { "epoch": 0.55, "learning_rate": 4.702065180430828e-06, "loss": 0.9635, "step": 4592 }, { "epoch": 0.55, "learning_rate": 4.70191298752105e-06, "loss": 1.0995, "step": 4593 }, { "epoch": 0.55, "learning_rate": 4.701760758213465e-06, "loss": 0.6155, "step": 4594 }, { "epoch": 0.55, "learning_rate": 4.70160849251059e-06, "loss": 0.696, "step": 4595 }, { "epoch": 0.55, "learning_rate": 4.7014561904149414e-06, "loss": 0.7403, "step": 4596 }, { "epoch": 0.55, "learning_rate": 4.7013038519290375e-06, "loss": 0.7516, "step": 4597 }, { "epoch": 0.55, "learning_rate": 4.701151477055396e-06, "loss": 0.8527, "step": 4598 }, { "epoch": 0.55, "learning_rate": 4.700999065796536e-06, "loss": 1.0493, "step": 4599 }, { "epoch": 0.55, "learning_rate": 4.700846618154977e-06, "loss": 1.1855, "step": 4600 }, { "epoch": 0.55, "learning_rate": 4.700694134133238e-06, "loss": 0.6291, "step": 4601 }, { "epoch": 0.55, "learning_rate": 4.700541613733841e-06, "loss": 0.7511, "step": 4602 }, { "epoch": 0.55, "learning_rate": 4.700389056959307e-06, "loss": 0.9086, "step": 4603 }, { "epoch": 0.55, "learning_rate": 4.7002364638121565e-06, "loss": 0.8477, "step": 4604 }, { "epoch": 0.55, "learning_rate": 4.700083834294913e-06, "loss": 1.0383, "step": 4605 }, { "epoch": 0.55, "learning_rate": 4.699931168410099e-06, "loss": 1.0692, "step": 4606 }, { "epoch": 0.55, "learning_rate": 4.699778466160237e-06, "loss": 0.7087, "step": 4607 }, { "epoch": 0.55, "learning_rate": 4.699625727547854e-06, "loss": 0.7246, "step": 4608 }, { "epoch": 0.55, "learning_rate": 4.699472952575472e-06, "loss": 0.6974, "step": 4609 }, { "epoch": 0.55, "learning_rate": 4.699320141245619e-06, "loss": 0.7996, "step": 4610 }, { "epoch": 0.55, "learning_rate": 4.6991672935608185e-06, "loss": 1.0336, "step": 4611 }, { "epoch": 0.55, "learning_rate": 4.699014409523599e-06, "loss": 1.018, "step": 4612 }, { "epoch": 0.55, "learning_rate": 4.698861489136486e-06, "loss": 0.5901, "step": 4613 }, { "epoch": 0.55, "learning_rate": 4.698708532402007e-06, "loss": 0.8172, "step": 4614 }, { "epoch": 0.55, "learning_rate": 4.698555539322693e-06, "loss": 0.8895, "step": 4615 }, { "epoch": 0.55, "learning_rate": 4.698402509901071e-06, "loss": 0.7956, "step": 4616 }, { "epoch": 0.55, "learning_rate": 4.698249444139672e-06, "loss": 1.0159, "step": 4617 }, { "epoch": 0.55, "learning_rate": 4.698096342041024e-06, "loss": 1.1814, "step": 4618 }, { "epoch": 0.55, "learning_rate": 4.69794320360766e-06, "loss": 0.9117, "step": 4619 }, { "epoch": 0.55, "learning_rate": 4.69779002884211e-06, "loss": 0.4471, "step": 4620 }, { "epoch": 0.55, "learning_rate": 4.697636817746906e-06, "loss": 0.8598, "step": 4621 }, { "epoch": 0.55, "learning_rate": 4.697483570324581e-06, "loss": 0.8992, "step": 4622 }, { "epoch": 0.55, "learning_rate": 4.697330286577668e-06, "loss": 0.8305, "step": 4623 }, { "epoch": 0.55, "learning_rate": 4.697176966508702e-06, "loss": 1.0579, "step": 4624 }, { "epoch": 0.55, "learning_rate": 4.697023610120216e-06, "loss": 1.2663, "step": 4625 }, { "epoch": 0.55, "learning_rate": 4.696870217414745e-06, "loss": 0.6326, "step": 4626 }, { "epoch": 0.55, "learning_rate": 4.696716788394825e-06, "loss": 0.7321, "step": 4627 }, { "epoch": 0.55, "learning_rate": 4.696563323062992e-06, "loss": 0.754, "step": 4628 }, { "epoch": 0.55, "learning_rate": 4.696409821421785e-06, "loss": 0.9327, "step": 4629 }, { "epoch": 0.55, "learning_rate": 4.696256283473737e-06, "loss": 1.0118, "step": 4630 }, { "epoch": 0.55, "learning_rate": 4.696102709221389e-06, "loss": 1.1372, "step": 4631 }, { "epoch": 0.55, "learning_rate": 4.695949098667279e-06, "loss": 0.6634, "step": 4632 }, { "epoch": 0.55, "learning_rate": 4.695795451813946e-06, "loss": 0.6613, "step": 4633 }, { "epoch": 0.55, "learning_rate": 4.69564176866393e-06, "loss": 0.8184, "step": 4634 }, { "epoch": 0.55, "learning_rate": 4.6954880492197715e-06, "loss": 0.7722, "step": 4635 }, { "epoch": 0.55, "learning_rate": 4.695334293484012e-06, "loss": 0.9285, "step": 4636 }, { "epoch": 0.55, "learning_rate": 4.695180501459191e-06, "loss": 1.1234, "step": 4637 }, { "epoch": 0.55, "learning_rate": 4.6950266731478534e-06, "loss": 0.5912, "step": 4638 }, { "epoch": 0.55, "learning_rate": 4.69487280855254e-06, "loss": 0.7178, "step": 4639 }, { "epoch": 0.55, "learning_rate": 4.6947189076757944e-06, "loss": 0.7141, "step": 4640 }, { "epoch": 0.55, "learning_rate": 4.694564970520163e-06, "loss": 0.8461, "step": 4641 }, { "epoch": 0.55, "learning_rate": 4.694410997088187e-06, "loss": 0.957, "step": 4642 }, { "epoch": 0.55, "learning_rate": 4.6942569873824125e-06, "loss": 1.0155, "step": 4643 }, { "epoch": 0.55, "learning_rate": 4.694102941405387e-06, "loss": 0.6419, "step": 4644 }, { "epoch": 0.55, "learning_rate": 4.693948859159655e-06, "loss": 0.5426, "step": 4645 }, { "epoch": 0.55, "learning_rate": 4.693794740647765e-06, "loss": 0.6242, "step": 4646 }, { "epoch": 0.55, "learning_rate": 4.693640585872263e-06, "loss": 0.8702, "step": 4647 }, { "epoch": 0.55, "learning_rate": 4.693486394835699e-06, "loss": 1.0558, "step": 4648 }, { "epoch": 0.55, "learning_rate": 4.69333216754062e-06, "loss": 1.0934, "step": 4649 }, { "epoch": 0.55, "learning_rate": 4.693177903989578e-06, "loss": 1.1777, "step": 4650 }, { "epoch": 0.55, "learning_rate": 4.693023604185119e-06, "loss": 0.6165, "step": 4651 }, { "epoch": 0.55, "learning_rate": 4.692869268129797e-06, "loss": 0.7798, "step": 4652 }, { "epoch": 0.55, "learning_rate": 4.692714895826162e-06, "loss": 0.7115, "step": 4653 }, { "epoch": 0.55, "learning_rate": 4.6925604872767664e-06, "loss": 0.915, "step": 4654 }, { "epoch": 0.55, "learning_rate": 4.6924060424841605e-06, "loss": 0.972, "step": 4655 }, { "epoch": 0.55, "learning_rate": 4.692251561450899e-06, "loss": 1.0744, "step": 4656 }, { "epoch": 0.55, "learning_rate": 4.692097044179536e-06, "loss": 0.6534, "step": 4657 }, { "epoch": 0.55, "learning_rate": 4.691942490672625e-06, "loss": 0.7248, "step": 4658 }, { "epoch": 0.55, "learning_rate": 4.691787900932721e-06, "loss": 0.7194, "step": 4659 }, { "epoch": 0.55, "learning_rate": 4.691633274962377e-06, "loss": 0.8223, "step": 4660 }, { "epoch": 0.56, "learning_rate": 4.691478612764153e-06, "loss": 0.8304, "step": 4661 }, { "epoch": 0.56, "learning_rate": 4.691323914340603e-06, "loss": 1.1798, "step": 4662 }, { "epoch": 0.56, "learning_rate": 4.691169179694285e-06, "loss": 0.6127, "step": 4663 }, { "epoch": 0.56, "learning_rate": 4.691014408827756e-06, "loss": 0.7139, "step": 4664 }, { "epoch": 0.56, "learning_rate": 4.690859601743576e-06, "loss": 0.8721, "step": 4665 }, { "epoch": 0.56, "learning_rate": 4.690704758444303e-06, "loss": 0.8706, "step": 4666 }, { "epoch": 0.56, "learning_rate": 4.690549878932495e-06, "loss": 0.9995, "step": 4667 }, { "epoch": 0.56, "learning_rate": 4.690394963210715e-06, "loss": 1.0836, "step": 4668 }, { "epoch": 0.56, "learning_rate": 4.690240011281522e-06, "loss": 0.7577, "step": 4669 }, { "epoch": 0.56, "learning_rate": 4.690085023147478e-06, "loss": 0.7844, "step": 4670 }, { "epoch": 0.56, "learning_rate": 4.689929998811145e-06, "loss": 0.7918, "step": 4671 }, { "epoch": 0.56, "learning_rate": 4.689774938275086e-06, "loss": 0.8079, "step": 4672 }, { "epoch": 0.56, "learning_rate": 4.6896198415418625e-06, "loss": 0.8942, "step": 4673 }, { "epoch": 0.56, "learning_rate": 4.68946470861404e-06, "loss": 1.1241, "step": 4674 }, { "epoch": 0.56, "learning_rate": 4.689309539494181e-06, "loss": 1.2001, "step": 4675 }, { "epoch": 0.56, "learning_rate": 4.689154334184852e-06, "loss": 0.471, "step": 4676 }, { "epoch": 0.56, "learning_rate": 4.688999092688618e-06, "loss": 0.7137, "step": 4677 }, { "epoch": 0.56, "learning_rate": 4.688843815008046e-06, "loss": 0.869, "step": 4678 }, { "epoch": 0.56, "learning_rate": 4.688688501145703e-06, "loss": 0.8461, "step": 4679 }, { "epoch": 0.56, "learning_rate": 4.6885331511041535e-06, "loss": 1.032, "step": 4680 }, { "epoch": 0.56, "learning_rate": 4.688377764885969e-06, "loss": 1.2781, "step": 4681 }, { "epoch": 0.56, "learning_rate": 4.6882223424937145e-06, "loss": 0.541, "step": 4682 }, { "epoch": 0.56, "learning_rate": 4.688066883929963e-06, "loss": 0.7805, "step": 4683 }, { "epoch": 0.56, "learning_rate": 4.687911389197281e-06, "loss": 0.7016, "step": 4684 }, { "epoch": 0.56, "learning_rate": 4.687755858298241e-06, "loss": 0.7831, "step": 4685 }, { "epoch": 0.56, "learning_rate": 4.687600291235413e-06, "loss": 0.9836, "step": 4686 }, { "epoch": 0.56, "learning_rate": 4.687444688011368e-06, "loss": 1.338, "step": 4687 }, { "epoch": 0.56, "learning_rate": 4.687289048628679e-06, "loss": 0.5979, "step": 4688 }, { "epoch": 0.56, "learning_rate": 4.687133373089918e-06, "loss": 0.6667, "step": 4689 }, { "epoch": 0.56, "learning_rate": 4.68697766139766e-06, "loss": 0.6419, "step": 4690 }, { "epoch": 0.56, "learning_rate": 4.686821913554477e-06, "loss": 0.737, "step": 4691 }, { "epoch": 0.56, "learning_rate": 4.686666129562945e-06, "loss": 0.9246, "step": 4692 }, { "epoch": 0.56, "learning_rate": 4.686510309425637e-06, "loss": 1.2116, "step": 4693 }, { "epoch": 0.56, "learning_rate": 4.686354453145132e-06, "loss": 0.7641, "step": 4694 }, { "epoch": 0.56, "learning_rate": 4.686198560724003e-06, "loss": 0.5671, "step": 4695 }, { "epoch": 0.56, "learning_rate": 4.6860426321648286e-06, "loss": 0.7959, "step": 4696 }, { "epoch": 0.56, "learning_rate": 4.685886667470187e-06, "loss": 0.665, "step": 4697 }, { "epoch": 0.56, "learning_rate": 4.685730666642654e-06, "loss": 1.0128, "step": 4698 }, { "epoch": 0.56, "learning_rate": 4.685574629684811e-06, "loss": 0.9909, "step": 4699 }, { "epoch": 0.56, "learning_rate": 4.685418556599235e-06, "loss": 1.1543, "step": 4700 }, { "epoch": 0.56, "learning_rate": 4.6852624473885074e-06, "loss": 0.5555, "step": 4701 }, { "epoch": 0.56, "learning_rate": 4.6851063020552075e-06, "loss": 0.7004, "step": 4702 }, { "epoch": 0.56, "learning_rate": 4.684950120601918e-06, "loss": 0.6401, "step": 4703 }, { "epoch": 0.56, "learning_rate": 4.684793903031219e-06, "loss": 0.7819, "step": 4704 }, { "epoch": 0.56, "learning_rate": 4.684637649345693e-06, "loss": 0.9496, "step": 4705 }, { "epoch": 0.56, "learning_rate": 4.684481359547924e-06, "loss": 1.2164, "step": 4706 }, { "epoch": 0.56, "learning_rate": 4.684325033640496e-06, "loss": 0.6441, "step": 4707 }, { "epoch": 0.56, "learning_rate": 4.684168671625991e-06, "loss": 0.7276, "step": 4708 }, { "epoch": 0.56, "learning_rate": 4.684012273506994e-06, "loss": 0.666, "step": 4709 }, { "epoch": 0.56, "learning_rate": 4.683855839286091e-06, "loss": 0.7415, "step": 4710 }, { "epoch": 0.56, "learning_rate": 4.683699368965868e-06, "loss": 0.9997, "step": 4711 }, { "epoch": 0.56, "learning_rate": 4.683542862548911e-06, "loss": 1.0874, "step": 4712 }, { "epoch": 0.56, "learning_rate": 4.683386320037808e-06, "loss": 0.6372, "step": 4713 }, { "epoch": 0.56, "learning_rate": 4.683229741435145e-06, "loss": 0.7505, "step": 4714 }, { "epoch": 0.56, "learning_rate": 4.683073126743511e-06, "loss": 0.896, "step": 4715 }, { "epoch": 0.56, "learning_rate": 4.682916475965497e-06, "loss": 0.7968, "step": 4716 }, { "epoch": 0.56, "learning_rate": 4.682759789103689e-06, "loss": 0.8842, "step": 4717 }, { "epoch": 0.56, "learning_rate": 4.682603066160678e-06, "loss": 1.2026, "step": 4718 }, { "epoch": 0.56, "learning_rate": 4.6824463071390555e-06, "loss": 0.9422, "step": 4719 }, { "epoch": 0.56, "learning_rate": 4.682289512041413e-06, "loss": 0.6745, "step": 4720 }, { "epoch": 0.56, "learning_rate": 4.682132680870341e-06, "loss": 0.6158, "step": 4721 }, { "epoch": 0.56, "learning_rate": 4.6819758136284346e-06, "loss": 0.7837, "step": 4722 }, { "epoch": 0.56, "learning_rate": 4.681818910318283e-06, "loss": 0.9321, "step": 4723 }, { "epoch": 0.56, "learning_rate": 4.681661970942482e-06, "loss": 1.0182, "step": 4724 }, { "epoch": 0.56, "learning_rate": 4.681504995503626e-06, "loss": 1.2065, "step": 4725 }, { "epoch": 0.56, "learning_rate": 4.68134798400431e-06, "loss": 0.6154, "step": 4726 }, { "epoch": 0.56, "learning_rate": 4.681190936447128e-06, "loss": 0.7573, "step": 4727 }, { "epoch": 0.56, "learning_rate": 4.681033852834676e-06, "loss": 0.7737, "step": 4728 }, { "epoch": 0.56, "learning_rate": 4.680876733169552e-06, "loss": 0.9562, "step": 4729 }, { "epoch": 0.56, "learning_rate": 4.680719577454353e-06, "loss": 1.0737, "step": 4730 }, { "epoch": 0.56, "learning_rate": 4.680562385691676e-06, "loss": 1.2374, "step": 4731 }, { "epoch": 0.56, "learning_rate": 4.68040515788412e-06, "loss": 0.7245, "step": 4732 }, { "epoch": 0.56, "learning_rate": 4.6802478940342845e-06, "loss": 0.7025, "step": 4733 }, { "epoch": 0.56, "learning_rate": 4.680090594144767e-06, "loss": 0.8708, "step": 4734 }, { "epoch": 0.56, "learning_rate": 4.67993325821817e-06, "loss": 0.7241, "step": 4735 }, { "epoch": 0.56, "learning_rate": 4.679775886257093e-06, "loss": 1.0739, "step": 4736 }, { "epoch": 0.56, "learning_rate": 4.679618478264138e-06, "loss": 1.2178, "step": 4737 }, { "epoch": 0.56, "learning_rate": 4.679461034241906e-06, "loss": 0.6375, "step": 4738 }, { "epoch": 0.56, "learning_rate": 4.679303554193001e-06, "loss": 0.7444, "step": 4739 }, { "epoch": 0.56, "learning_rate": 4.679146038120025e-06, "loss": 0.7127, "step": 4740 }, { "epoch": 0.56, "learning_rate": 4.6789884860255825e-06, "loss": 0.826, "step": 4741 }, { "epoch": 0.56, "learning_rate": 4.678830897912277e-06, "loss": 0.991, "step": 4742 }, { "epoch": 0.56, "learning_rate": 4.678673273782715e-06, "loss": 1.1588, "step": 4743 }, { "epoch": 0.56, "learning_rate": 4.6785156136395e-06, "loss": 0.5793, "step": 4744 }, { "epoch": 0.57, "learning_rate": 4.678357917485239e-06, "loss": 0.525, "step": 4745 }, { "epoch": 0.57, "learning_rate": 4.678200185322539e-06, "loss": 0.7034, "step": 4746 }, { "epoch": 0.57, "learning_rate": 4.678042417154007e-06, "loss": 0.6646, "step": 4747 }, { "epoch": 0.57, "learning_rate": 4.677884612982251e-06, "loss": 0.8528, "step": 4748 }, { "epoch": 0.57, "learning_rate": 4.67772677280988e-06, "loss": 1.028, "step": 4749 }, { "epoch": 0.57, "learning_rate": 4.677568896639503e-06, "loss": 1.2432, "step": 4750 }, { "epoch": 0.57, "learning_rate": 4.677410984473729e-06, "loss": 0.5359, "step": 4751 }, { "epoch": 0.57, "learning_rate": 4.677253036315169e-06, "loss": 0.7198, "step": 4752 }, { "epoch": 0.57, "learning_rate": 4.677095052166433e-06, "loss": 0.7919, "step": 4753 }, { "epoch": 0.57, "learning_rate": 4.676937032030133e-06, "loss": 0.8588, "step": 4754 }, { "epoch": 0.57, "learning_rate": 4.676778975908883e-06, "loss": 0.9029, "step": 4755 }, { "epoch": 0.57, "learning_rate": 4.676620883805293e-06, "loss": 1.2557, "step": 4756 }, { "epoch": 0.57, "learning_rate": 4.676462755721976e-06, "loss": 0.4726, "step": 4757 }, { "epoch": 0.57, "learning_rate": 4.676304591661548e-06, "loss": 0.6544, "step": 4758 }, { "epoch": 0.57, "learning_rate": 4.6761463916266225e-06, "loss": 0.7501, "step": 4759 }, { "epoch": 0.57, "learning_rate": 4.675988155619814e-06, "loss": 0.7269, "step": 4760 }, { "epoch": 0.57, "learning_rate": 4.6758298836437385e-06, "loss": 0.9265, "step": 4761 }, { "epoch": 0.57, "learning_rate": 4.675671575701014e-06, "loss": 1.1396, "step": 4762 }, { "epoch": 0.57, "learning_rate": 4.675513231794254e-06, "loss": 0.7808, "step": 4763 }, { "epoch": 0.57, "learning_rate": 4.67535485192608e-06, "loss": 0.556, "step": 4764 }, { "epoch": 0.57, "learning_rate": 4.675196436099105e-06, "loss": 0.7136, "step": 4765 }, { "epoch": 0.57, "learning_rate": 4.6750379843159525e-06, "loss": 0.7753, "step": 4766 }, { "epoch": 0.57, "learning_rate": 4.6748794965792385e-06, "loss": 0.9955, "step": 4767 }, { "epoch": 0.57, "learning_rate": 4.674720972891584e-06, "loss": 1.1008, "step": 4768 }, { "epoch": 0.57, "learning_rate": 4.67456241325561e-06, "loss": 0.6797, "step": 4769 }, { "epoch": 0.57, "learning_rate": 4.674403817673936e-06, "loss": 0.5904, "step": 4770 }, { "epoch": 0.57, "learning_rate": 4.674245186149186e-06, "loss": 0.7527, "step": 4771 }, { "epoch": 0.57, "learning_rate": 4.674086518683979e-06, "loss": 0.8019, "step": 4772 }, { "epoch": 0.57, "learning_rate": 4.67392781528094e-06, "loss": 0.7768, "step": 4773 }, { "epoch": 0.57, "learning_rate": 4.6737690759426915e-06, "loss": 1.0628, "step": 4774 }, { "epoch": 0.57, "learning_rate": 4.673610300671858e-06, "loss": 1.1266, "step": 4775 }, { "epoch": 0.57, "learning_rate": 4.673451489471065e-06, "loss": 0.7256, "step": 4776 }, { "epoch": 0.57, "learning_rate": 4.673292642342935e-06, "loss": 0.711, "step": 4777 }, { "epoch": 0.57, "learning_rate": 4.673133759290096e-06, "loss": 0.7076, "step": 4778 }, { "epoch": 0.57, "learning_rate": 4.6729748403151735e-06, "loss": 0.7442, "step": 4779 }, { "epoch": 0.57, "learning_rate": 4.6728158854207936e-06, "loss": 1.117, "step": 4780 }, { "epoch": 0.57, "learning_rate": 4.672656894609586e-06, "loss": 1.2559, "step": 4781 }, { "epoch": 0.57, "learning_rate": 4.672497867884177e-06, "loss": 0.6021, "step": 4782 }, { "epoch": 0.57, "learning_rate": 4.672338805247197e-06, "loss": 0.699, "step": 4783 }, { "epoch": 0.57, "learning_rate": 4.672179706701273e-06, "loss": 0.9437, "step": 4784 }, { "epoch": 0.57, "learning_rate": 4.672020572249036e-06, "loss": 0.7278, "step": 4785 }, { "epoch": 0.57, "learning_rate": 4.671861401893118e-06, "loss": 0.9158, "step": 4786 }, { "epoch": 0.57, "learning_rate": 4.671702195636147e-06, "loss": 1.118, "step": 4787 }, { "epoch": 0.57, "learning_rate": 4.6715429534807585e-06, "loss": 0.6444, "step": 4788 }, { "epoch": 0.57, "learning_rate": 4.671383675429581e-06, "loss": 0.6341, "step": 4789 }, { "epoch": 0.57, "learning_rate": 4.671224361485251e-06, "loss": 0.6375, "step": 4790 }, { "epoch": 0.57, "learning_rate": 4.671065011650399e-06, "loss": 0.7034, "step": 4791 }, { "epoch": 0.57, "learning_rate": 4.67090562592766e-06, "loss": 0.8395, "step": 4792 }, { "epoch": 0.57, "learning_rate": 4.67074620431967e-06, "loss": 1.087, "step": 4793 }, { "epoch": 0.57, "learning_rate": 4.670586746829061e-06, "loss": 0.7951, "step": 4794 }, { "epoch": 0.57, "learning_rate": 4.6704272534584725e-06, "loss": 0.5917, "step": 4795 }, { "epoch": 0.57, "learning_rate": 4.67026772421054e-06, "loss": 0.7181, "step": 4796 }, { "epoch": 0.57, "learning_rate": 4.670108159087898e-06, "loss": 0.7745, "step": 4797 }, { "epoch": 0.57, "learning_rate": 4.669948558093188e-06, "loss": 1.004, "step": 4798 }, { "epoch": 0.57, "learning_rate": 4.669788921229045e-06, "loss": 1.0463, "step": 4799 }, { "epoch": 0.57, "learning_rate": 4.66962924849811e-06, "loss": 1.2677, "step": 4800 }, { "epoch": 0.57, "learning_rate": 4.669469539903021e-06, "loss": 0.5765, "step": 4801 }, { "epoch": 0.57, "learning_rate": 4.669309795446417e-06, "loss": 0.7213, "step": 4802 }, { "epoch": 0.57, "learning_rate": 4.669150015130942e-06, "loss": 0.7711, "step": 4803 }, { "epoch": 0.57, "learning_rate": 4.668990198959234e-06, "loss": 0.9149, "step": 4804 }, { "epoch": 0.57, "learning_rate": 4.668830346933938e-06, "loss": 0.9547, "step": 4805 }, { "epoch": 0.57, "learning_rate": 4.668670459057693e-06, "loss": 1.1576, "step": 4806 }, { "epoch": 0.57, "learning_rate": 4.668510535333143e-06, "loss": 0.6539, "step": 4807 }, { "epoch": 0.57, "learning_rate": 4.668350575762932e-06, "loss": 0.6857, "step": 4808 }, { "epoch": 0.57, "learning_rate": 4.668190580349705e-06, "loss": 0.7243, "step": 4809 }, { "epoch": 0.57, "learning_rate": 4.668030549096105e-06, "loss": 0.6245, "step": 4810 }, { "epoch": 0.57, "learning_rate": 4.667870482004778e-06, "loss": 0.984, "step": 4811 }, { "epoch": 0.57, "learning_rate": 4.667710379078371e-06, "loss": 1.2251, "step": 4812 }, { "epoch": 0.57, "learning_rate": 4.667550240319529e-06, "loss": 0.6634, "step": 4813 }, { "epoch": 0.57, "learning_rate": 4.667390065730899e-06, "loss": 0.6573, "step": 4814 }, { "epoch": 0.57, "learning_rate": 4.66722985531513e-06, "loss": 0.8454, "step": 4815 }, { "epoch": 0.57, "learning_rate": 4.667069609074869e-06, "loss": 0.7888, "step": 4816 }, { "epoch": 0.57, "learning_rate": 4.666909327012766e-06, "loss": 0.975, "step": 4817 }, { "epoch": 0.57, "learning_rate": 4.6667490091314705e-06, "loss": 1.1935, "step": 4818 }, { "epoch": 0.57, "learning_rate": 4.666588655433631e-06, "loss": 0.6266, "step": 4819 }, { "epoch": 0.57, "learning_rate": 4.6664282659219e-06, "loss": 0.4594, "step": 4820 }, { "epoch": 0.57, "learning_rate": 4.6662678405989284e-06, "loss": 0.7226, "step": 4821 }, { "epoch": 0.57, "learning_rate": 4.666107379467366e-06, "loss": 0.7461, "step": 4822 }, { "epoch": 0.57, "learning_rate": 4.665946882529868e-06, "loss": 0.7908, "step": 4823 }, { "epoch": 0.57, "learning_rate": 4.665786349789086e-06, "loss": 1.0451, "step": 4824 }, { "epoch": 0.57, "learning_rate": 4.665625781247674e-06, "loss": 1.0558, "step": 4825 }, { "epoch": 0.57, "learning_rate": 4.665465176908287e-06, "loss": 0.7202, "step": 4826 }, { "epoch": 0.57, "learning_rate": 4.665304536773579e-06, "loss": 0.7363, "step": 4827 }, { "epoch": 0.57, "learning_rate": 4.6651438608462044e-06, "loss": 0.8231, "step": 4828 }, { "epoch": 0.58, "learning_rate": 4.66498314912882e-06, "loss": 0.7764, "step": 4829 }, { "epoch": 0.58, "learning_rate": 4.664822401624083e-06, "loss": 0.9638, "step": 4830 }, { "epoch": 0.58, "learning_rate": 4.664661618334649e-06, "loss": 1.0604, "step": 4831 }, { "epoch": 0.58, "learning_rate": 4.664500799263177e-06, "loss": 0.6937, "step": 4832 }, { "epoch": 0.58, "learning_rate": 4.664339944412327e-06, "loss": 0.6396, "step": 4833 }, { "epoch": 0.58, "learning_rate": 4.664179053784754e-06, "loss": 0.6476, "step": 4834 }, { "epoch": 0.58, "learning_rate": 4.6640181273831206e-06, "loss": 0.7979, "step": 4835 }, { "epoch": 0.58, "learning_rate": 4.663857165210086e-06, "loss": 0.9747, "step": 4836 }, { "epoch": 0.58, "learning_rate": 4.663696167268311e-06, "loss": 1.0461, "step": 4837 }, { "epoch": 0.58, "learning_rate": 4.663535133560457e-06, "loss": 0.8494, "step": 4838 }, { "epoch": 0.58, "learning_rate": 4.663374064089185e-06, "loss": 0.6447, "step": 4839 }, { "epoch": 0.58, "learning_rate": 4.663212958857159e-06, "loss": 0.9586, "step": 4840 }, { "epoch": 0.58, "learning_rate": 4.66305181786704e-06, "loss": 0.8168, "step": 4841 }, { "epoch": 0.58, "learning_rate": 4.662890641121494e-06, "loss": 1.0108, "step": 4842 }, { "epoch": 0.58, "learning_rate": 4.662729428623184e-06, "loss": 1.1141, "step": 4843 }, { "epoch": 0.58, "learning_rate": 4.662568180374775e-06, "loss": 0.6719, "step": 4844 }, { "epoch": 0.58, "learning_rate": 4.662406896378933e-06, "loss": 0.5742, "step": 4845 }, { "epoch": 0.58, "learning_rate": 4.6622455766383225e-06, "loss": 0.6097, "step": 4846 }, { "epoch": 0.58, "learning_rate": 4.662084221155612e-06, "loss": 0.7702, "step": 4847 }, { "epoch": 0.58, "learning_rate": 4.6619228299334675e-06, "loss": 0.8562, "step": 4848 }, { "epoch": 0.58, "learning_rate": 4.6617614029745575e-06, "loss": 1.0616, "step": 4849 }, { "epoch": 0.58, "learning_rate": 4.66159994028155e-06, "loss": 1.0507, "step": 4850 }, { "epoch": 0.58, "learning_rate": 4.661438441857114e-06, "loss": 0.5312, "step": 4851 }, { "epoch": 0.58, "learning_rate": 4.66127690770392e-06, "loss": 0.7852, "step": 4852 }, { "epoch": 0.58, "learning_rate": 4.661115337824637e-06, "loss": 0.841, "step": 4853 }, { "epoch": 0.58, "learning_rate": 4.6609537322219356e-06, "loss": 1.0201, "step": 4854 }, { "epoch": 0.58, "learning_rate": 4.660792090898488e-06, "loss": 0.9244, "step": 4855 }, { "epoch": 0.58, "learning_rate": 4.660630413856966e-06, "loss": 1.1698, "step": 4856 }, { "epoch": 0.58, "learning_rate": 4.660468701100042e-06, "loss": 0.5462, "step": 4857 }, { "epoch": 0.58, "learning_rate": 4.660306952630388e-06, "loss": 0.7669, "step": 4858 }, { "epoch": 0.58, "learning_rate": 4.66014516845068e-06, "loss": 0.7968, "step": 4859 }, { "epoch": 0.58, "learning_rate": 4.659983348563591e-06, "loss": 0.9554, "step": 4860 }, { "epoch": 0.58, "learning_rate": 4.659821492971795e-06, "loss": 1.0977, "step": 4861 }, { "epoch": 0.58, "learning_rate": 4.659659601677969e-06, "loss": 1.2079, "step": 4862 }, { "epoch": 0.58, "learning_rate": 4.6594976746847896e-06, "loss": 0.6467, "step": 4863 }, { "epoch": 0.58, "learning_rate": 4.659335711994932e-06, "loss": 0.6991, "step": 4864 }, { "epoch": 0.58, "learning_rate": 4.659173713611073e-06, "loss": 0.8102, "step": 4865 }, { "epoch": 0.58, "learning_rate": 4.659011679535893e-06, "loss": 0.7473, "step": 4866 }, { "epoch": 0.58, "learning_rate": 4.658849609772067e-06, "loss": 0.9189, "step": 4867 }, { "epoch": 0.58, "learning_rate": 4.658687504322277e-06, "loss": 1.1368, "step": 4868 }, { "epoch": 0.58, "learning_rate": 4.6585253631892e-06, "loss": 0.6787, "step": 4869 }, { "epoch": 0.58, "learning_rate": 4.658363186375519e-06, "loss": 0.6119, "step": 4870 }, { "epoch": 0.58, "learning_rate": 4.6582009738839125e-06, "loss": 0.724, "step": 4871 }, { "epoch": 0.58, "learning_rate": 4.658038725717064e-06, "loss": 0.6919, "step": 4872 }, { "epoch": 0.58, "learning_rate": 4.6578764418776535e-06, "loss": 0.8478, "step": 4873 }, { "epoch": 0.58, "learning_rate": 4.657714122368364e-06, "loss": 1.1396, "step": 4874 }, { "epoch": 0.58, "learning_rate": 4.65755176719188e-06, "loss": 1.2251, "step": 4875 }, { "epoch": 0.58, "learning_rate": 4.6573893763508835e-06, "loss": 0.5257, "step": 4876 }, { "epoch": 0.58, "learning_rate": 4.657226949848061e-06, "loss": 0.8073, "step": 4877 }, { "epoch": 0.58, "learning_rate": 4.657064487686094e-06, "loss": 0.764, "step": 4878 }, { "epoch": 0.58, "learning_rate": 4.656901989867671e-06, "loss": 0.907, "step": 4879 }, { "epoch": 0.58, "learning_rate": 4.656739456395477e-06, "loss": 0.9261, "step": 4880 }, { "epoch": 0.58, "learning_rate": 4.656576887272198e-06, "loss": 1.2406, "step": 4881 }, { "epoch": 0.58, "learning_rate": 4.6564142825005236e-06, "loss": 0.6024, "step": 4882 }, { "epoch": 0.58, "learning_rate": 4.656251642083139e-06, "loss": 0.7777, "step": 4883 }, { "epoch": 0.58, "learning_rate": 4.6560889660227346e-06, "loss": 0.6819, "step": 4884 }, { "epoch": 0.58, "learning_rate": 4.655926254321998e-06, "loss": 0.8489, "step": 4885 }, { "epoch": 0.58, "learning_rate": 4.65576350698362e-06, "loss": 0.9092, "step": 4886 }, { "epoch": 0.58, "learning_rate": 4.655600724010291e-06, "loss": 1.1244, "step": 4887 }, { "epoch": 0.58, "learning_rate": 4.655437905404701e-06, "loss": 0.5516, "step": 4888 }, { "epoch": 0.58, "learning_rate": 4.65527505116954e-06, "loss": 0.5968, "step": 4889 }, { "epoch": 0.58, "learning_rate": 4.655112161307503e-06, "loss": 0.7564, "step": 4890 }, { "epoch": 0.58, "learning_rate": 4.654949235821281e-06, "loss": 0.8386, "step": 4891 }, { "epoch": 0.58, "learning_rate": 4.654786274713567e-06, "loss": 0.8749, "step": 4892 }, { "epoch": 0.58, "learning_rate": 4.654623277987055e-06, "loss": 1.0938, "step": 4893 }, { "epoch": 0.58, "learning_rate": 4.654460245644439e-06, "loss": 0.6036, "step": 4894 }, { "epoch": 0.58, "learning_rate": 4.654297177688415e-06, "loss": 0.4687, "step": 4895 }, { "epoch": 0.58, "learning_rate": 4.654134074121678e-06, "loss": 0.7675, "step": 4896 }, { "epoch": 0.58, "learning_rate": 4.653970934946924e-06, "loss": 0.7724, "step": 4897 }, { "epoch": 0.58, "learning_rate": 4.65380776016685e-06, "loss": 0.9359, "step": 4898 }, { "epoch": 0.58, "learning_rate": 4.653644549784152e-06, "loss": 1.0935, "step": 4899 }, { "epoch": 0.58, "learning_rate": 4.6534813038015295e-06, "loss": 1.0349, "step": 4900 }, { "epoch": 0.58, "learning_rate": 4.65331802222168e-06, "loss": 0.4734, "step": 4901 }, { "epoch": 0.58, "learning_rate": 4.653154705047303e-06, "loss": 0.7293, "step": 4902 }, { "epoch": 0.58, "learning_rate": 4.652991352281099e-06, "loss": 0.8158, "step": 4903 }, { "epoch": 0.58, "learning_rate": 4.652827963925766e-06, "loss": 0.986, "step": 4904 }, { "epoch": 0.58, "learning_rate": 4.652664539984006e-06, "loss": 1.036, "step": 4905 }, { "epoch": 0.58, "learning_rate": 4.652501080458521e-06, "loss": 1.0977, "step": 4906 }, { "epoch": 0.58, "learning_rate": 4.652337585352014e-06, "loss": 0.4868, "step": 4907 }, { "epoch": 0.58, "learning_rate": 4.652174054667184e-06, "loss": 0.8306, "step": 4908 }, { "epoch": 0.58, "learning_rate": 4.652010488406737e-06, "loss": 0.7874, "step": 4909 }, { "epoch": 0.58, "learning_rate": 4.651846886573377e-06, "loss": 0.8143, "step": 4910 }, { "epoch": 0.58, "learning_rate": 4.651683249169806e-06, "loss": 0.9073, "step": 4911 }, { "epoch": 0.58, "learning_rate": 4.651519576198731e-06, "loss": 1.255, "step": 4912 }, { "epoch": 0.59, "learning_rate": 4.651355867662857e-06, "loss": 0.5544, "step": 4913 }, { "epoch": 0.59, "learning_rate": 4.65119212356489e-06, "loss": 0.5143, "step": 4914 }, { "epoch": 0.59, "learning_rate": 4.6510283439075365e-06, "loss": 0.795, "step": 4915 }, { "epoch": 0.59, "learning_rate": 4.650864528693504e-06, "loss": 0.8922, "step": 4916 }, { "epoch": 0.59, "learning_rate": 4.6507006779255e-06, "loss": 0.8875, "step": 4917 }, { "epoch": 0.59, "learning_rate": 4.650536791606234e-06, "loss": 1.1337, "step": 4918 }, { "epoch": 0.59, "learning_rate": 4.650372869738415e-06, "loss": 0.7687, "step": 4919 }, { "epoch": 0.59, "learning_rate": 4.65020891232475e-06, "loss": 0.6361, "step": 4920 }, { "epoch": 0.59, "learning_rate": 4.650044919367952e-06, "loss": 0.782, "step": 4921 }, { "epoch": 0.59, "learning_rate": 4.649880890870731e-06, "loss": 0.7188, "step": 4922 }, { "epoch": 0.59, "learning_rate": 4.649716826835799e-06, "loss": 0.9011, "step": 4923 }, { "epoch": 0.59, "learning_rate": 4.649552727265868e-06, "loss": 1.0432, "step": 4924 }, { "epoch": 0.59, "learning_rate": 4.649388592163649e-06, "loss": 1.1528, "step": 4925 }, { "epoch": 0.59, "learning_rate": 4.649224421531857e-06, "loss": 0.5477, "step": 4926 }, { "epoch": 0.59, "learning_rate": 4.649060215373205e-06, "loss": 0.7173, "step": 4927 }, { "epoch": 0.59, "learning_rate": 4.6488959736904074e-06, "loss": 0.7342, "step": 4928 }, { "epoch": 0.59, "learning_rate": 4.648731696486178e-06, "loss": 0.8683, "step": 4929 }, { "epoch": 0.59, "learning_rate": 4.648567383763234e-06, "loss": 0.9715, "step": 4930 }, { "epoch": 0.59, "learning_rate": 4.648403035524292e-06, "loss": 1.1708, "step": 4931 }, { "epoch": 0.59, "learning_rate": 4.648238651772066e-06, "loss": 0.4866, "step": 4932 }, { "epoch": 0.59, "learning_rate": 4.6480742325092745e-06, "loss": 0.7649, "step": 4933 }, { "epoch": 0.59, "learning_rate": 4.647909777738637e-06, "loss": 0.8788, "step": 4934 }, { "epoch": 0.59, "learning_rate": 4.6477452874628705e-06, "loss": 0.8133, "step": 4935 }, { "epoch": 0.59, "learning_rate": 4.647580761684693e-06, "loss": 0.9609, "step": 4936 }, { "epoch": 0.59, "learning_rate": 4.647416200406827e-06, "loss": 1.0992, "step": 4937 }, { "epoch": 0.59, "learning_rate": 4.647251603631989e-06, "loss": 0.6558, "step": 4938 }, { "epoch": 0.59, "learning_rate": 4.647086971362904e-06, "loss": 0.6236, "step": 4939 }, { "epoch": 0.59, "learning_rate": 4.646922303602289e-06, "loss": 0.7103, "step": 4940 }, { "epoch": 0.59, "learning_rate": 4.64675760035287e-06, "loss": 0.8568, "step": 4941 }, { "epoch": 0.59, "learning_rate": 4.6465928616173675e-06, "loss": 0.8566, "step": 4942 }, { "epoch": 0.59, "learning_rate": 4.646428087398504e-06, "loss": 0.9405, "step": 4943 }, { "epoch": 0.59, "learning_rate": 4.646263277699005e-06, "loss": 0.763, "step": 4944 }, { "epoch": 0.59, "learning_rate": 4.646098432521593e-06, "loss": 0.6158, "step": 4945 }, { "epoch": 0.59, "learning_rate": 4.645933551868995e-06, "loss": 0.6672, "step": 4946 }, { "epoch": 0.59, "learning_rate": 4.645768635743934e-06, "loss": 0.7053, "step": 4947 }, { "epoch": 0.59, "learning_rate": 4.645603684149138e-06, "loss": 0.888, "step": 4948 }, { "epoch": 0.59, "learning_rate": 4.645438697087333e-06, "loss": 1.0119, "step": 4949 }, { "epoch": 0.59, "learning_rate": 4.645273674561247e-06, "loss": 1.2556, "step": 4950 }, { "epoch": 0.59, "learning_rate": 4.645108616573606e-06, "loss": 0.6643, "step": 4951 }, { "epoch": 0.59, "learning_rate": 4.644943523127141e-06, "loss": 0.68, "step": 4952 }, { "epoch": 0.59, "learning_rate": 4.644778394224578e-06, "loss": 0.742, "step": 4953 }, { "epoch": 0.59, "learning_rate": 4.64461322986865e-06, "loss": 0.8505, "step": 4954 }, { "epoch": 0.59, "learning_rate": 4.644448030062084e-06, "loss": 1.0796, "step": 4955 }, { "epoch": 0.59, "learning_rate": 4.644282794807612e-06, "loss": 1.2031, "step": 4956 }, { "epoch": 0.59, "learning_rate": 4.644117524107966e-06, "loss": 0.5382, "step": 4957 }, { "epoch": 0.59, "learning_rate": 4.643952217965878e-06, "loss": 0.8233, "step": 4958 }, { "epoch": 0.59, "learning_rate": 4.643786876384078e-06, "loss": 0.7665, "step": 4959 }, { "epoch": 0.59, "learning_rate": 4.643621499365303e-06, "loss": 0.7371, "step": 4960 }, { "epoch": 0.59, "learning_rate": 4.643456086912285e-06, "loss": 1.0357, "step": 4961 }, { "epoch": 0.59, "learning_rate": 4.643290639027757e-06, "loss": 1.0684, "step": 4962 }, { "epoch": 0.59, "learning_rate": 4.643125155714455e-06, "loss": 0.6075, "step": 4963 }, { "epoch": 0.59, "learning_rate": 4.642959636975114e-06, "loss": 0.6535, "step": 4964 }, { "epoch": 0.59, "learning_rate": 4.642794082812471e-06, "loss": 0.6712, "step": 4965 }, { "epoch": 0.59, "learning_rate": 4.642628493229262e-06, "loss": 0.7999, "step": 4966 }, { "epoch": 0.59, "learning_rate": 4.642462868228223e-06, "loss": 0.909, "step": 4967 }, { "epoch": 0.59, "learning_rate": 4.642297207812095e-06, "loss": 1.0056, "step": 4968 }, { "epoch": 0.59, "learning_rate": 4.642131511983613e-06, "loss": 0.6473, "step": 4969 }, { "epoch": 0.59, "learning_rate": 4.641965780745518e-06, "loss": 0.5949, "step": 4970 }, { "epoch": 0.59, "learning_rate": 4.641800014100549e-06, "loss": 0.7949, "step": 4971 }, { "epoch": 0.59, "learning_rate": 4.641634212051446e-06, "loss": 0.7721, "step": 4972 }, { "epoch": 0.59, "learning_rate": 4.641468374600949e-06, "loss": 0.9136, "step": 4973 }, { "epoch": 0.59, "learning_rate": 4.641302501751801e-06, "loss": 1.0475, "step": 4974 }, { "epoch": 0.59, "learning_rate": 4.641136593506743e-06, "loss": 1.2648, "step": 4975 }, { "epoch": 0.59, "learning_rate": 4.640970649868517e-06, "loss": 0.6468, "step": 4976 }, { "epoch": 0.59, "learning_rate": 4.640804670839867e-06, "loss": 0.7027, "step": 4977 }, { "epoch": 0.59, "learning_rate": 4.640638656423535e-06, "loss": 0.8278, "step": 4978 }, { "epoch": 0.59, "learning_rate": 4.6404726066222675e-06, "loss": 0.8654, "step": 4979 }, { "epoch": 0.59, "learning_rate": 4.640306521438808e-06, "loss": 0.9852, "step": 4980 }, { "epoch": 0.59, "learning_rate": 4.640140400875902e-06, "loss": 1.1146, "step": 4981 }, { "epoch": 0.59, "learning_rate": 4.639974244936296e-06, "loss": 0.5523, "step": 4982 }, { "epoch": 0.59, "learning_rate": 4.6398080536227356e-06, "loss": 0.6627, "step": 4983 }, { "epoch": 0.59, "learning_rate": 4.639641826937968e-06, "loss": 0.781, "step": 4984 }, { "epoch": 0.59, "learning_rate": 4.639475564884742e-06, "loss": 0.7155, "step": 4985 }, { "epoch": 0.59, "learning_rate": 4.639309267465806e-06, "loss": 0.9763, "step": 4986 }, { "epoch": 0.59, "learning_rate": 4.639142934683908e-06, "loss": 1.0835, "step": 4987 }, { "epoch": 0.59, "learning_rate": 4.638976566541798e-06, "loss": 0.8427, "step": 4988 }, { "epoch": 0.59, "learning_rate": 4.638810163042225e-06, "loss": 0.5854, "step": 4989 }, { "epoch": 0.59, "learning_rate": 4.638643724187941e-06, "loss": 0.6819, "step": 4990 }, { "epoch": 0.59, "learning_rate": 4.638477249981697e-06, "loss": 0.7239, "step": 4991 }, { "epoch": 0.59, "learning_rate": 4.6383107404262445e-06, "loss": 0.952, "step": 4992 }, { "epoch": 0.59, "learning_rate": 4.638144195524336e-06, "loss": 1.1975, "step": 4993 }, { "epoch": 0.59, "learning_rate": 4.637977615278725e-06, "loss": 0.9852, "step": 4994 }, { "epoch": 0.59, "learning_rate": 4.637810999692164e-06, "loss": 0.7173, "step": 4995 }, { "epoch": 0.59, "learning_rate": 4.637644348767408e-06, "loss": 0.5804, "step": 4996 }, { "epoch": 0.6, "learning_rate": 4.6374776625072115e-06, "loss": 0.8321, "step": 4997 }, { "epoch": 0.6, "learning_rate": 4.6373109409143305e-06, "loss": 0.8824, "step": 4998 }, { "epoch": 0.6, "learning_rate": 4.63714418399152e-06, "loss": 0.9669, "step": 4999 }, { "epoch": 0.6, "learning_rate": 4.636977391741536e-06, "loss": 1.2135, "step": 5000 }, { "epoch": 0.6, "learning_rate": 4.6368105641671365e-06, "loss": 0.5746, "step": 5001 }, { "epoch": 0.6, "learning_rate": 4.63664370127108e-06, "loss": 0.6662, "step": 5002 }, { "epoch": 0.6, "learning_rate": 4.636476803056124e-06, "loss": 0.7424, "step": 5003 }, { "epoch": 0.6, "learning_rate": 4.636309869525027e-06, "loss": 0.9308, "step": 5004 }, { "epoch": 0.6, "learning_rate": 4.636142900680547e-06, "loss": 0.9384, "step": 5005 }, { "epoch": 0.6, "learning_rate": 4.635975896525447e-06, "loss": 1.3365, "step": 5006 }, { "epoch": 0.6, "learning_rate": 4.635808857062486e-06, "loss": 0.4858, "step": 5007 }, { "epoch": 0.6, "learning_rate": 4.6356417822944256e-06, "loss": 0.7663, "step": 5008 }, { "epoch": 0.6, "learning_rate": 4.635474672224027e-06, "loss": 0.7827, "step": 5009 }, { "epoch": 0.6, "learning_rate": 4.635307526854053e-06, "loss": 0.8944, "step": 5010 }, { "epoch": 0.6, "learning_rate": 4.635140346187267e-06, "loss": 0.8253, "step": 5011 }, { "epoch": 0.6, "learning_rate": 4.634973130226431e-06, "loss": 1.0917, "step": 5012 }, { "epoch": 0.6, "learning_rate": 4.63480587897431e-06, "loss": 0.4982, "step": 5013 }, { "epoch": 0.6, "learning_rate": 4.63463859243367e-06, "loss": 0.4955, "step": 5014 }, { "epoch": 0.6, "learning_rate": 4.634471270607274e-06, "loss": 0.8094, "step": 5015 }, { "epoch": 0.6, "learning_rate": 4.6343039134978884e-06, "loss": 0.7488, "step": 5016 }, { "epoch": 0.6, "learning_rate": 4.63413652110828e-06, "loss": 0.8613, "step": 5017 }, { "epoch": 0.6, "learning_rate": 4.6339690934412165e-06, "loss": 1.0189, "step": 5018 }, { "epoch": 0.6, "learning_rate": 4.633801630499465e-06, "loss": 0.816, "step": 5019 }, { "epoch": 0.6, "learning_rate": 4.633634132285793e-06, "loss": 0.4947, "step": 5020 }, { "epoch": 0.6, "learning_rate": 4.63346659880297e-06, "loss": 0.7025, "step": 5021 }, { "epoch": 0.6, "learning_rate": 4.633299030053765e-06, "loss": 0.7287, "step": 5022 }, { "epoch": 0.6, "learning_rate": 4.633131426040947e-06, "loss": 0.8094, "step": 5023 }, { "epoch": 0.6, "learning_rate": 4.632963786767289e-06, "loss": 1.0498, "step": 5024 }, { "epoch": 0.6, "learning_rate": 4.632796112235562e-06, "loss": 1.3438, "step": 5025 }, { "epoch": 0.6, "learning_rate": 4.632628402448533e-06, "loss": 0.5758, "step": 5026 }, { "epoch": 0.6, "learning_rate": 4.63246065740898e-06, "loss": 0.5854, "step": 5027 }, { "epoch": 0.6, "learning_rate": 4.632292877119672e-06, "loss": 0.8651, "step": 5028 }, { "epoch": 0.6, "learning_rate": 4.632125061583385e-06, "loss": 0.8225, "step": 5029 }, { "epoch": 0.6, "learning_rate": 4.631957210802891e-06, "loss": 1.1404, "step": 5030 }, { "epoch": 0.6, "learning_rate": 4.631789324780965e-06, "loss": 1.0517, "step": 5031 }, { "epoch": 0.6, "learning_rate": 4.631621403520384e-06, "loss": 0.6067, "step": 5032 }, { "epoch": 0.6, "learning_rate": 4.631453447023921e-06, "loss": 0.7812, "step": 5033 }, { "epoch": 0.6, "learning_rate": 4.631285455294354e-06, "loss": 0.7465, "step": 5034 }, { "epoch": 0.6, "learning_rate": 4.63111742833446e-06, "loss": 0.8526, "step": 5035 }, { "epoch": 0.6, "learning_rate": 4.630949366147016e-06, "loss": 1.0192, "step": 5036 }, { "epoch": 0.6, "learning_rate": 4.630781268734799e-06, "loss": 1.1506, "step": 5037 }, { "epoch": 0.6, "learning_rate": 4.630613136100591e-06, "loss": 0.6455, "step": 5038 }, { "epoch": 0.6, "learning_rate": 4.630444968247168e-06, "loss": 0.6048, "step": 5039 }, { "epoch": 0.6, "learning_rate": 4.630276765177311e-06, "loss": 0.7267, "step": 5040 }, { "epoch": 0.6, "learning_rate": 4.6301085268938e-06, "loss": 0.6657, "step": 5041 }, { "epoch": 0.6, "learning_rate": 4.629940253399416e-06, "loss": 0.8845, "step": 5042 }, { "epoch": 0.6, "learning_rate": 4.629771944696942e-06, "loss": 1.1598, "step": 5043 }, { "epoch": 0.6, "learning_rate": 4.629603600789159e-06, "loss": 0.805, "step": 5044 }, { "epoch": 0.6, "learning_rate": 4.6294352216788485e-06, "loss": 0.545, "step": 5045 }, { "epoch": 0.6, "learning_rate": 4.629266807368796e-06, "loss": 0.6857, "step": 5046 }, { "epoch": 0.6, "learning_rate": 4.629098357861785e-06, "loss": 0.7662, "step": 5047 }, { "epoch": 0.6, "learning_rate": 4.628929873160598e-06, "loss": 0.9031, "step": 5048 }, { "epoch": 0.6, "learning_rate": 4.628761353268023e-06, "loss": 1.0217, "step": 5049 }, { "epoch": 0.6, "learning_rate": 4.628592798186844e-06, "loss": 1.3464, "step": 5050 }, { "epoch": 0.6, "learning_rate": 4.628424207919847e-06, "loss": 0.6989, "step": 5051 }, { "epoch": 0.6, "learning_rate": 4.62825558246982e-06, "loss": 0.7259, "step": 5052 }, { "epoch": 0.6, "learning_rate": 4.628086921839549e-06, "loss": 0.7925, "step": 5053 }, { "epoch": 0.6, "learning_rate": 4.627918226031822e-06, "loss": 0.8531, "step": 5054 }, { "epoch": 0.6, "learning_rate": 4.627749495049429e-06, "loss": 0.9245, "step": 5055 }, { "epoch": 0.6, "learning_rate": 4.627580728895158e-06, "loss": 1.2225, "step": 5056 }, { "epoch": 0.6, "learning_rate": 4.6274119275717994e-06, "loss": 0.5659, "step": 5057 }, { "epoch": 0.6, "learning_rate": 4.627243091082143e-06, "loss": 0.5663, "step": 5058 }, { "epoch": 0.6, "learning_rate": 4.62707421942898e-06, "loss": 0.7582, "step": 5059 }, { "epoch": 0.6, "learning_rate": 4.626905312615101e-06, "loss": 0.7718, "step": 5060 }, { "epoch": 0.6, "learning_rate": 4.626736370643299e-06, "loss": 0.997, "step": 5061 }, { "epoch": 0.6, "learning_rate": 4.626567393516367e-06, "loss": 1.173, "step": 5062 }, { "epoch": 0.6, "learning_rate": 4.626398381237097e-06, "loss": 0.7081, "step": 5063 }, { "epoch": 0.6, "learning_rate": 4.626229333808283e-06, "loss": 0.7117, "step": 5064 }, { "epoch": 0.6, "learning_rate": 4.62606025123272e-06, "loss": 0.8038, "step": 5065 }, { "epoch": 0.6, "learning_rate": 4.625891133513203e-06, "loss": 0.8372, "step": 5066 }, { "epoch": 0.6, "learning_rate": 4.625721980652526e-06, "loss": 0.9609, "step": 5067 }, { "epoch": 0.6, "learning_rate": 4.6255527926534885e-06, "loss": 1.1067, "step": 5068 }, { "epoch": 0.6, "learning_rate": 4.625383569518883e-06, "loss": 0.8152, "step": 5069 }, { "epoch": 0.6, "learning_rate": 4.625214311251509e-06, "loss": 0.6837, "step": 5070 }, { "epoch": 0.6, "learning_rate": 4.625045017854165e-06, "loss": 0.6888, "step": 5071 }, { "epoch": 0.6, "learning_rate": 4.624875689329648e-06, "loss": 0.761, "step": 5072 }, { "epoch": 0.6, "learning_rate": 4.624706325680758e-06, "loss": 0.9041, "step": 5073 }, { "epoch": 0.6, "learning_rate": 4.6245369269102934e-06, "loss": 1.0453, "step": 5074 }, { "epoch": 0.6, "learning_rate": 4.624367493021055e-06, "loss": 1.0574, "step": 5075 }, { "epoch": 0.6, "learning_rate": 4.624198024015845e-06, "loss": 0.5758, "step": 5076 }, { "epoch": 0.6, "learning_rate": 4.6240285198974635e-06, "loss": 0.8394, "step": 5077 }, { "epoch": 0.6, "learning_rate": 4.623858980668711e-06, "loss": 0.8241, "step": 5078 }, { "epoch": 0.6, "learning_rate": 4.623689406332392e-06, "loss": 0.8759, "step": 5079 }, { "epoch": 0.6, "learning_rate": 4.623519796891308e-06, "loss": 0.9339, "step": 5080 }, { "epoch": 0.61, "learning_rate": 4.623350152348265e-06, "loss": 1.1411, "step": 5081 }, { "epoch": 0.61, "learning_rate": 4.623180472706065e-06, "loss": 0.5727, "step": 5082 }, { "epoch": 0.61, "learning_rate": 4.6230107579675145e-06, "loss": 0.5644, "step": 5083 }, { "epoch": 0.61, "learning_rate": 4.6228410081354166e-06, "loss": 0.6572, "step": 5084 }, { "epoch": 0.61, "learning_rate": 4.62267122321258e-06, "loss": 0.7549, "step": 5085 }, { "epoch": 0.61, "learning_rate": 4.622501403201809e-06, "loss": 0.8972, "step": 5086 }, { "epoch": 0.61, "learning_rate": 4.6223315481059124e-06, "loss": 1.1522, "step": 5087 }, { "epoch": 0.61, "learning_rate": 4.622161657927697e-06, "loss": 0.6842, "step": 5088 }, { "epoch": 0.61, "learning_rate": 4.621991732669972e-06, "loss": 0.73, "step": 5089 }, { "epoch": 0.61, "learning_rate": 4.621821772335544e-06, "loss": 0.7482, "step": 5090 }, { "epoch": 0.61, "learning_rate": 4.6216517769272255e-06, "loss": 0.8047, "step": 5091 }, { "epoch": 0.61, "learning_rate": 4.621481746447825e-06, "loss": 1.0259, "step": 5092 }, { "epoch": 0.61, "learning_rate": 4.621311680900152e-06, "loss": 1.1732, "step": 5093 }, { "epoch": 0.61, "learning_rate": 4.62114158028702e-06, "loss": 0.614, "step": 5094 }, { "epoch": 0.61, "learning_rate": 4.62097144461124e-06, "loss": 0.6127, "step": 5095 }, { "epoch": 0.61, "learning_rate": 4.620801273875623e-06, "loss": 0.755, "step": 5096 }, { "epoch": 0.61, "learning_rate": 4.620631068082984e-06, "loss": 0.6255, "step": 5097 }, { "epoch": 0.61, "learning_rate": 4.620460827236135e-06, "loss": 0.9103, "step": 5098 }, { "epoch": 0.61, "learning_rate": 4.6202905513378905e-06, "loss": 1.0585, "step": 5099 }, { "epoch": 0.61, "learning_rate": 4.620120240391065e-06, "loss": 1.068, "step": 5100 }, { "epoch": 0.61, "learning_rate": 4.619949894398474e-06, "loss": 0.6207, "step": 5101 }, { "epoch": 0.61, "learning_rate": 4.619779513362934e-06, "loss": 0.5503, "step": 5102 }, { "epoch": 0.61, "learning_rate": 4.619609097287261e-06, "loss": 0.8431, "step": 5103 }, { "epoch": 0.61, "learning_rate": 4.619438646174271e-06, "loss": 0.7997, "step": 5104 }, { "epoch": 0.61, "learning_rate": 4.619268160026783e-06, "loss": 1.1355, "step": 5105 }, { "epoch": 0.61, "learning_rate": 4.6190976388476145e-06, "loss": 1.2271, "step": 5106 }, { "epoch": 0.61, "learning_rate": 4.618927082639584e-06, "loss": 0.647, "step": 5107 }, { "epoch": 0.61, "learning_rate": 4.618756491405511e-06, "loss": 0.6565, "step": 5108 }, { "epoch": 0.61, "learning_rate": 4.6185858651482155e-06, "loss": 0.7435, "step": 5109 }, { "epoch": 0.61, "learning_rate": 4.618415203870518e-06, "loss": 0.8192, "step": 5110 }, { "epoch": 0.61, "learning_rate": 4.618244507575239e-06, "loss": 1.011, "step": 5111 }, { "epoch": 0.61, "learning_rate": 4.6180737762652006e-06, "loss": 1.1299, "step": 5112 }, { "epoch": 0.61, "learning_rate": 4.617903009943225e-06, "loss": 0.5251, "step": 5113 }, { "epoch": 0.61, "learning_rate": 4.617732208612135e-06, "loss": 0.572, "step": 5114 }, { "epoch": 0.61, "learning_rate": 4.6175613722747535e-06, "loss": 0.6087, "step": 5115 }, { "epoch": 0.61, "learning_rate": 4.617390500933906e-06, "loss": 0.9753, "step": 5116 }, { "epoch": 0.61, "learning_rate": 4.617219594592415e-06, "loss": 0.9149, "step": 5117 }, { "epoch": 0.61, "learning_rate": 4.617048653253106e-06, "loss": 1.0116, "step": 5118 }, { "epoch": 0.61, "learning_rate": 4.616877676918806e-06, "loss": 0.7473, "step": 5119 }, { "epoch": 0.61, "learning_rate": 4.6167066655923404e-06, "loss": 0.5397, "step": 5120 }, { "epoch": 0.61, "learning_rate": 4.616535619276535e-06, "loss": 0.5687, "step": 5121 }, { "epoch": 0.61, "learning_rate": 4.616364537974219e-06, "loss": 0.6976, "step": 5122 }, { "epoch": 0.61, "learning_rate": 4.61619342168822e-06, "loss": 0.7966, "step": 5123 }, { "epoch": 0.61, "learning_rate": 4.616022270421365e-06, "loss": 0.8878, "step": 5124 }, { "epoch": 0.61, "learning_rate": 4.615851084176486e-06, "loss": 1.1416, "step": 5125 }, { "epoch": 0.61, "learning_rate": 4.61567986295641e-06, "loss": 0.5372, "step": 5126 }, { "epoch": 0.61, "learning_rate": 4.615508606763968e-06, "loss": 0.6648, "step": 5127 }, { "epoch": 0.61, "learning_rate": 4.615337315601992e-06, "loss": 0.8316, "step": 5128 }, { "epoch": 0.61, "learning_rate": 4.615165989473312e-06, "loss": 0.8496, "step": 5129 }, { "epoch": 0.61, "learning_rate": 4.614994628380761e-06, "loss": 0.9926, "step": 5130 }, { "epoch": 0.61, "learning_rate": 4.614823232327171e-06, "loss": 1.2443, "step": 5131 }, { "epoch": 0.61, "learning_rate": 4.614651801315377e-06, "loss": 0.5096, "step": 5132 }, { "epoch": 0.61, "learning_rate": 4.614480335348209e-06, "loss": 0.6884, "step": 5133 }, { "epoch": 0.61, "learning_rate": 4.614308834428506e-06, "loss": 0.6512, "step": 5134 }, { "epoch": 0.61, "learning_rate": 4.614137298559099e-06, "loss": 0.647, "step": 5135 }, { "epoch": 0.61, "learning_rate": 4.613965727742825e-06, "loss": 1.0035, "step": 5136 }, { "epoch": 0.61, "learning_rate": 4.613794121982521e-06, "loss": 1.217, "step": 5137 }, { "epoch": 0.61, "learning_rate": 4.6136224812810225e-06, "loss": 0.555, "step": 5138 }, { "epoch": 0.61, "learning_rate": 4.613450805641167e-06, "loss": 0.801, "step": 5139 }, { "epoch": 0.61, "learning_rate": 4.613279095065792e-06, "loss": 0.8406, "step": 5140 }, { "epoch": 0.61, "learning_rate": 4.613107349557737e-06, "loss": 0.7193, "step": 5141 }, { "epoch": 0.61, "learning_rate": 4.61293556911984e-06, "loss": 1.0424, "step": 5142 }, { "epoch": 0.61, "learning_rate": 4.6127637537549395e-06, "loss": 1.1807, "step": 5143 }, { "epoch": 0.61, "learning_rate": 4.612591903465879e-06, "loss": 0.7659, "step": 5144 }, { "epoch": 0.61, "learning_rate": 4.612420018255495e-06, "loss": 0.4944, "step": 5145 }, { "epoch": 0.61, "learning_rate": 4.612248098126632e-06, "loss": 0.668, "step": 5146 }, { "epoch": 0.61, "learning_rate": 4.612076143082131e-06, "loss": 0.816, "step": 5147 }, { "epoch": 0.61, "learning_rate": 4.611904153124833e-06, "loss": 0.9812, "step": 5148 }, { "epoch": 0.61, "learning_rate": 4.611732128257583e-06, "loss": 1.0098, "step": 5149 }, { "epoch": 0.61, "learning_rate": 4.6115600684832226e-06, "loss": 1.2501, "step": 5150 }, { "epoch": 0.61, "learning_rate": 4.611387973804598e-06, "loss": 0.5587, "step": 5151 }, { "epoch": 0.61, "learning_rate": 4.611215844224554e-06, "loss": 0.7859, "step": 5152 }, { "epoch": 0.61, "learning_rate": 4.611043679745934e-06, "loss": 0.8018, "step": 5153 }, { "epoch": 0.61, "learning_rate": 4.610871480371585e-06, "loss": 0.9634, "step": 5154 }, { "epoch": 0.61, "learning_rate": 4.610699246104353e-06, "loss": 1.1238, "step": 5155 }, { "epoch": 0.61, "learning_rate": 4.610526976947085e-06, "loss": 1.1675, "step": 5156 }, { "epoch": 0.61, "learning_rate": 4.6103546729026295e-06, "loss": 0.5844, "step": 5157 }, { "epoch": 0.61, "learning_rate": 4.6101823339738335e-06, "loss": 0.7952, "step": 5158 }, { "epoch": 0.61, "learning_rate": 4.610009960163548e-06, "loss": 0.7475, "step": 5159 }, { "epoch": 0.61, "learning_rate": 4.6098375514746195e-06, "loss": 0.8542, "step": 5160 }, { "epoch": 0.61, "learning_rate": 4.6096651079098996e-06, "loss": 1.1024, "step": 5161 }, { "epoch": 0.61, "learning_rate": 4.6094926294722386e-06, "loss": 1.2305, "step": 5162 }, { "epoch": 0.61, "learning_rate": 4.609320116164488e-06, "loss": 0.711, "step": 5163 }, { "epoch": 0.61, "learning_rate": 4.609147567989498e-06, "loss": 0.616, "step": 5164 }, { "epoch": 0.62, "learning_rate": 4.608974984950122e-06, "loss": 0.7521, "step": 5165 }, { "epoch": 0.62, "learning_rate": 4.608802367049212e-06, "loss": 0.8326, "step": 5166 }, { "epoch": 0.62, "learning_rate": 4.608629714289623e-06, "loss": 0.9309, "step": 5167 }, { "epoch": 0.62, "learning_rate": 4.6084570266742075e-06, "loss": 1.0284, "step": 5168 }, { "epoch": 0.62, "learning_rate": 4.60828430420582e-06, "loss": 0.8394, "step": 5169 }, { "epoch": 0.62, "learning_rate": 4.608111546887316e-06, "loss": 0.4879, "step": 5170 }, { "epoch": 0.62, "learning_rate": 4.607938754721552e-06, "loss": 0.7154, "step": 5171 }, { "epoch": 0.62, "learning_rate": 4.607765927711383e-06, "loss": 0.7192, "step": 5172 }, { "epoch": 0.62, "learning_rate": 4.607593065859667e-06, "loss": 0.769, "step": 5173 }, { "epoch": 0.62, "learning_rate": 4.60742016916926e-06, "loss": 0.9637, "step": 5174 }, { "epoch": 0.62, "learning_rate": 4.607247237643021e-06, "loss": 1.2908, "step": 5175 }, { "epoch": 0.62, "learning_rate": 4.6070742712838076e-06, "loss": 0.6298, "step": 5176 }, { "epoch": 0.62, "learning_rate": 4.606901270094481e-06, "loss": 0.7177, "step": 5177 }, { "epoch": 0.62, "learning_rate": 4.606728234077899e-06, "loss": 0.8116, "step": 5178 }, { "epoch": 0.62, "learning_rate": 4.606555163236923e-06, "loss": 0.8317, "step": 5179 }, { "epoch": 0.62, "learning_rate": 4.606382057574413e-06, "loss": 1.006, "step": 5180 }, { "epoch": 0.62, "learning_rate": 4.60620891709323e-06, "loss": 1.0145, "step": 5181 }, { "epoch": 0.62, "learning_rate": 4.6060357417962385e-06, "loss": 0.6058, "step": 5182 }, { "epoch": 0.62, "learning_rate": 4.605862531686298e-06, "loss": 0.7036, "step": 5183 }, { "epoch": 0.62, "learning_rate": 4.605689286766274e-06, "loss": 0.8873, "step": 5184 }, { "epoch": 0.62, "learning_rate": 4.605516007039029e-06, "loss": 0.8525, "step": 5185 }, { "epoch": 0.62, "learning_rate": 4.605342692507429e-06, "loss": 1.0247, "step": 5186 }, { "epoch": 0.62, "learning_rate": 4.605169343174336e-06, "loss": 1.2142, "step": 5187 }, { "epoch": 0.62, "learning_rate": 4.604995959042618e-06, "loss": 0.6213, "step": 5188 }, { "epoch": 0.62, "learning_rate": 4.60482254011514e-06, "loss": 0.6341, "step": 5189 }, { "epoch": 0.62, "learning_rate": 4.6046490863947685e-06, "loss": 0.7548, "step": 5190 }, { "epoch": 0.62, "learning_rate": 4.604475597884371e-06, "loss": 0.7857, "step": 5191 }, { "epoch": 0.62, "learning_rate": 4.604302074586816e-06, "loss": 0.9786, "step": 5192 }, { "epoch": 0.62, "learning_rate": 4.604128516504971e-06, "loss": 1.1031, "step": 5193 }, { "epoch": 0.62, "learning_rate": 4.603954923641704e-06, "loss": 0.6622, "step": 5194 }, { "epoch": 0.62, "learning_rate": 4.603781295999886e-06, "loss": 0.6355, "step": 5195 }, { "epoch": 0.62, "learning_rate": 4.603607633582388e-06, "loss": 0.658, "step": 5196 }, { "epoch": 0.62, "learning_rate": 4.603433936392078e-06, "loss": 0.7226, "step": 5197 }, { "epoch": 0.62, "learning_rate": 4.603260204431829e-06, "loss": 0.8779, "step": 5198 }, { "epoch": 0.62, "learning_rate": 4.603086437704512e-06, "loss": 1.019, "step": 5199 }, { "epoch": 0.62, "learning_rate": 4.602912636212999e-06, "loss": 1.0804, "step": 5200 }, { "epoch": 0.62, "learning_rate": 4.602738799960165e-06, "loss": 0.4935, "step": 5201 }, { "epoch": 0.62, "learning_rate": 4.602564928948882e-06, "loss": 0.6577, "step": 5202 }, { "epoch": 0.62, "learning_rate": 4.602391023182024e-06, "loss": 0.7517, "step": 5203 }, { "epoch": 0.62, "learning_rate": 4.602217082662467e-06, "loss": 0.8231, "step": 5204 }, { "epoch": 0.62, "learning_rate": 4.602043107393084e-06, "loss": 0.9913, "step": 5205 }, { "epoch": 0.62, "learning_rate": 4.601869097376752e-06, "loss": 1.2126, "step": 5206 }, { "epoch": 0.62, "learning_rate": 4.6016950526163475e-06, "loss": 0.6646, "step": 5207 }, { "epoch": 0.62, "learning_rate": 4.601520973114748e-06, "loss": 0.7405, "step": 5208 }, { "epoch": 0.62, "learning_rate": 4.60134685887483e-06, "loss": 0.7449, "step": 5209 }, { "epoch": 0.62, "learning_rate": 4.601172709899472e-06, "loss": 0.7841, "step": 5210 }, { "epoch": 0.62, "learning_rate": 4.6009985261915536e-06, "loss": 1.0297, "step": 5211 }, { "epoch": 0.62, "learning_rate": 4.600824307753953e-06, "loss": 1.2222, "step": 5212 }, { "epoch": 0.62, "learning_rate": 4.6006500545895495e-06, "loss": 0.6928, "step": 5213 }, { "epoch": 0.62, "learning_rate": 4.600475766701225e-06, "loss": 0.5757, "step": 5214 }, { "epoch": 0.62, "learning_rate": 4.600301444091859e-06, "loss": 0.6942, "step": 5215 }, { "epoch": 0.62, "learning_rate": 4.600127086764336e-06, "loss": 0.8332, "step": 5216 }, { "epoch": 0.62, "learning_rate": 4.5999526947215346e-06, "loss": 0.8686, "step": 5217 }, { "epoch": 0.62, "learning_rate": 4.599778267966339e-06, "loss": 1.1174, "step": 5218 }, { "epoch": 0.62, "learning_rate": 4.599603806501632e-06, "loss": 0.7053, "step": 5219 }, { "epoch": 0.62, "learning_rate": 4.599429310330299e-06, "loss": 0.6787, "step": 5220 }, { "epoch": 0.62, "learning_rate": 4.599254779455222e-06, "loss": 0.8508, "step": 5221 }, { "epoch": 0.62, "learning_rate": 4.599080213879289e-06, "loss": 0.7966, "step": 5222 }, { "epoch": 0.62, "learning_rate": 4.598905613605383e-06, "loss": 0.8605, "step": 5223 }, { "epoch": 0.62, "learning_rate": 4.598730978636391e-06, "loss": 1.1887, "step": 5224 }, { "epoch": 0.62, "learning_rate": 4.5985563089752e-06, "loss": 1.2468, "step": 5225 }, { "epoch": 0.62, "learning_rate": 4.598381604624697e-06, "loss": 0.5827, "step": 5226 }, { "epoch": 0.62, "learning_rate": 4.59820686558777e-06, "loss": 0.799, "step": 5227 }, { "epoch": 0.62, "learning_rate": 4.598032091867308e-06, "loss": 0.6985, "step": 5228 }, { "epoch": 0.62, "learning_rate": 4.597857283466198e-06, "loss": 0.853, "step": 5229 }, { "epoch": 0.62, "learning_rate": 4.597682440387334e-06, "loss": 0.8336, "step": 5230 }, { "epoch": 0.62, "learning_rate": 4.5975075626336e-06, "loss": 1.1678, "step": 5231 }, { "epoch": 0.62, "learning_rate": 4.597332650207892e-06, "loss": 0.58, "step": 5232 }, { "epoch": 0.62, "learning_rate": 4.597157703113099e-06, "loss": 0.752, "step": 5233 }, { "epoch": 0.62, "learning_rate": 4.596982721352112e-06, "loss": 0.7916, "step": 5234 }, { "epoch": 0.62, "learning_rate": 4.596807704927826e-06, "loss": 0.7653, "step": 5235 }, { "epoch": 0.62, "learning_rate": 4.596632653843132e-06, "loss": 1.0431, "step": 5236 }, { "epoch": 0.62, "learning_rate": 4.596457568100925e-06, "loss": 1.1603, "step": 5237 }, { "epoch": 0.62, "learning_rate": 4.596282447704098e-06, "loss": 0.5568, "step": 5238 }, { "epoch": 0.62, "learning_rate": 4.5961072926555464e-06, "loss": 0.558, "step": 5239 }, { "epoch": 0.62, "learning_rate": 4.595932102958165e-06, "loss": 0.6319, "step": 5240 }, { "epoch": 0.62, "learning_rate": 4.595756878614851e-06, "loss": 0.8804, "step": 5241 }, { "epoch": 0.62, "learning_rate": 4.5955816196284985e-06, "loss": 0.9119, "step": 5242 }, { "epoch": 0.62, "learning_rate": 4.595406326002006e-06, "loss": 1.2582, "step": 5243 }, { "epoch": 0.62, "learning_rate": 4.595230997738272e-06, "loss": 0.8181, "step": 5244 }, { "epoch": 0.62, "learning_rate": 4.595055634840194e-06, "loss": 0.5987, "step": 5245 }, { "epoch": 0.62, "learning_rate": 4.5948802373106705e-06, "loss": 0.7334, "step": 5246 }, { "epoch": 0.62, "learning_rate": 4.594704805152601e-06, "loss": 0.7777, "step": 5247 }, { "epoch": 0.62, "learning_rate": 4.594529338368885e-06, "loss": 0.8229, "step": 5248 }, { "epoch": 0.63, "learning_rate": 4.594353836962423e-06, "loss": 1.1534, "step": 5249 }, { "epoch": 0.63, "learning_rate": 4.5941783009361176e-06, "loss": 1.1877, "step": 5250 }, { "epoch": 0.63, "learning_rate": 4.5940027302928675e-06, "loss": 0.6042, "step": 5251 }, { "epoch": 0.63, "learning_rate": 4.593827125035577e-06, "loss": 0.7881, "step": 5252 }, { "epoch": 0.63, "learning_rate": 4.593651485167149e-06, "loss": 0.6562, "step": 5253 }, { "epoch": 0.63, "learning_rate": 4.593475810690486e-06, "loss": 0.9125, "step": 5254 }, { "epoch": 0.63, "learning_rate": 4.593300101608492e-06, "loss": 1.1117, "step": 5255 }, { "epoch": 0.63, "learning_rate": 4.593124357924072e-06, "loss": 1.2953, "step": 5256 }, { "epoch": 0.63, "learning_rate": 4.5929485796401306e-06, "loss": 0.5113, "step": 5257 }, { "epoch": 0.63, "learning_rate": 4.592772766759573e-06, "loss": 0.6274, "step": 5258 }, { "epoch": 0.63, "learning_rate": 4.592596919285307e-06, "loss": 0.821, "step": 5259 }, { "epoch": 0.63, "learning_rate": 4.592421037220238e-06, "loss": 0.8574, "step": 5260 }, { "epoch": 0.63, "learning_rate": 4.592245120567273e-06, "loss": 1.0002, "step": 5261 }, { "epoch": 0.63, "learning_rate": 4.592069169329322e-06, "loss": 1.2119, "step": 5262 }, { "epoch": 0.63, "learning_rate": 4.591893183509291e-06, "loss": 0.6943, "step": 5263 }, { "epoch": 0.63, "learning_rate": 4.591717163110091e-06, "loss": 0.7603, "step": 5264 }, { "epoch": 0.63, "learning_rate": 4.5915411081346296e-06, "loss": 0.7302, "step": 5265 }, { "epoch": 0.63, "learning_rate": 4.591365018585819e-06, "loss": 0.7786, "step": 5266 }, { "epoch": 0.63, "learning_rate": 4.5911888944665685e-06, "loss": 0.9459, "step": 5267 }, { "epoch": 0.63, "learning_rate": 4.59101273577979e-06, "loss": 1.0513, "step": 5268 }, { "epoch": 0.63, "learning_rate": 4.590836542528396e-06, "loss": 0.6471, "step": 5269 }, { "epoch": 0.63, "learning_rate": 4.5906603147153e-06, "loss": 0.6093, "step": 5270 }, { "epoch": 0.63, "learning_rate": 4.590484052343411e-06, "loss": 0.6796, "step": 5271 }, { "epoch": 0.63, "learning_rate": 4.590307755415647e-06, "loss": 0.7669, "step": 5272 }, { "epoch": 0.63, "learning_rate": 4.590131423934919e-06, "loss": 0.8149, "step": 5273 }, { "epoch": 0.63, "learning_rate": 4.589955057904144e-06, "loss": 1.0554, "step": 5274 }, { "epoch": 0.63, "learning_rate": 4.589778657326237e-06, "loss": 1.1677, "step": 5275 }, { "epoch": 0.63, "learning_rate": 4.589602222204113e-06, "loss": 0.7378, "step": 5276 }, { "epoch": 0.63, "learning_rate": 4.589425752540689e-06, "loss": 0.7444, "step": 5277 }, { "epoch": 0.63, "learning_rate": 4.589249248338882e-06, "loss": 0.7377, "step": 5278 }, { "epoch": 0.63, "learning_rate": 4.58907270960161e-06, "loss": 1.0414, "step": 5279 }, { "epoch": 0.63, "learning_rate": 4.588896136331789e-06, "loss": 1.0774, "step": 5280 }, { "epoch": 0.63, "learning_rate": 4.588719528532342e-06, "loss": 1.2614, "step": 5281 }, { "epoch": 0.63, "learning_rate": 4.588542886206184e-06, "loss": 0.6138, "step": 5282 }, { "epoch": 0.63, "learning_rate": 4.588366209356238e-06, "loss": 0.5211, "step": 5283 }, { "epoch": 0.63, "learning_rate": 4.588189497985423e-06, "loss": 0.7799, "step": 5284 }, { "epoch": 0.63, "learning_rate": 4.5880127520966605e-06, "loss": 0.7308, "step": 5285 }, { "epoch": 0.63, "learning_rate": 4.587835971692873e-06, "loss": 0.9388, "step": 5286 }, { "epoch": 0.63, "learning_rate": 4.587659156776979e-06, "loss": 1.1456, "step": 5287 }, { "epoch": 0.63, "learning_rate": 4.587482307351906e-06, "loss": 0.6663, "step": 5288 }, { "epoch": 0.63, "learning_rate": 4.587305423420575e-06, "loss": 0.5649, "step": 5289 }, { "epoch": 0.63, "learning_rate": 4.5871285049859105e-06, "loss": 0.6273, "step": 5290 }, { "epoch": 0.63, "learning_rate": 4.586951552050835e-06, "loss": 0.7203, "step": 5291 }, { "epoch": 0.63, "learning_rate": 4.586774564618277e-06, "loss": 1.0287, "step": 5292 }, { "epoch": 0.63, "learning_rate": 4.58659754269116e-06, "loss": 1.2324, "step": 5293 }, { "epoch": 0.63, "learning_rate": 4.586420486272411e-06, "loss": 0.7613, "step": 5294 }, { "epoch": 0.63, "learning_rate": 4.586243395364955e-06, "loss": 0.6389, "step": 5295 }, { "epoch": 0.63, "learning_rate": 4.586066269971722e-06, "loss": 0.6889, "step": 5296 }, { "epoch": 0.63, "learning_rate": 4.585889110095638e-06, "loss": 0.8228, "step": 5297 }, { "epoch": 0.63, "learning_rate": 4.585711915739631e-06, "loss": 0.9243, "step": 5298 }, { "epoch": 0.63, "learning_rate": 4.585534686906632e-06, "loss": 0.917, "step": 5299 }, { "epoch": 0.63, "learning_rate": 4.58535742359957e-06, "loss": 1.0576, "step": 5300 }, { "epoch": 0.63, "learning_rate": 4.585180125821375e-06, "loss": 0.7544, "step": 5301 }, { "epoch": 0.63, "learning_rate": 4.5850027935749766e-06, "loss": 0.6919, "step": 5302 }, { "epoch": 0.63, "learning_rate": 4.584825426863307e-06, "loss": 0.7003, "step": 5303 }, { "epoch": 0.63, "learning_rate": 4.584648025689299e-06, "loss": 0.7747, "step": 5304 }, { "epoch": 0.63, "learning_rate": 4.584470590055883e-06, "loss": 1.0558, "step": 5305 }, { "epoch": 0.63, "learning_rate": 4.584293119965994e-06, "loss": 1.1522, "step": 5306 }, { "epoch": 0.63, "learning_rate": 4.5841156154225655e-06, "loss": 0.5262, "step": 5307 }, { "epoch": 0.63, "learning_rate": 4.58393807642853e-06, "loss": 0.6492, "step": 5308 }, { "epoch": 0.63, "learning_rate": 4.583760502986824e-06, "loss": 0.7125, "step": 5309 }, { "epoch": 0.63, "learning_rate": 4.583582895100381e-06, "loss": 0.8129, "step": 5310 }, { "epoch": 0.63, "learning_rate": 4.583405252772138e-06, "loss": 0.8748, "step": 5311 }, { "epoch": 0.63, "learning_rate": 4.583227576005033e-06, "loss": 1.0743, "step": 5312 }, { "epoch": 0.63, "learning_rate": 4.583049864801999e-06, "loss": 0.5827, "step": 5313 }, { "epoch": 0.63, "learning_rate": 4.582872119165978e-06, "loss": 0.7127, "step": 5314 }, { "epoch": 0.63, "learning_rate": 4.582694339099906e-06, "loss": 0.803, "step": 5315 }, { "epoch": 0.63, "learning_rate": 4.58251652460672e-06, "loss": 0.8235, "step": 5316 }, { "epoch": 0.63, "learning_rate": 4.582338675689363e-06, "loss": 1.0161, "step": 5317 }, { "epoch": 0.63, "learning_rate": 4.582160792350772e-06, "loss": 1.2803, "step": 5318 }, { "epoch": 0.63, "learning_rate": 4.5819828745938885e-06, "loss": 0.8122, "step": 5319 }, { "epoch": 0.63, "learning_rate": 4.581804922421654e-06, "loss": 0.6795, "step": 5320 }, { "epoch": 0.63, "learning_rate": 4.581626935837009e-06, "loss": 0.6628, "step": 5321 }, { "epoch": 0.63, "learning_rate": 4.581448914842895e-06, "loss": 0.7802, "step": 5322 }, { "epoch": 0.63, "learning_rate": 4.581270859442257e-06, "loss": 0.844, "step": 5323 }, { "epoch": 0.63, "learning_rate": 4.581092769638037e-06, "loss": 1.092, "step": 5324 }, { "epoch": 0.63, "learning_rate": 4.580914645433179e-06, "loss": 1.1373, "step": 5325 }, { "epoch": 0.63, "learning_rate": 4.580736486830627e-06, "loss": 0.5704, "step": 5326 }, { "epoch": 0.63, "learning_rate": 4.580558293833327e-06, "loss": 0.6637, "step": 5327 }, { "epoch": 0.63, "learning_rate": 4.580380066444224e-06, "loss": 0.6105, "step": 5328 }, { "epoch": 0.63, "learning_rate": 4.5802018046662625e-06, "loss": 0.8876, "step": 5329 }, { "epoch": 0.63, "learning_rate": 4.580023508502391e-06, "loss": 0.9879, "step": 5330 }, { "epoch": 0.63, "learning_rate": 4.579845177955556e-06, "loss": 1.1948, "step": 5331 }, { "epoch": 0.63, "learning_rate": 4.579666813028707e-06, "loss": 0.5508, "step": 5332 }, { "epoch": 0.64, "learning_rate": 4.57948841372479e-06, "loss": 0.8338, "step": 5333 }, { "epoch": 0.64, "learning_rate": 4.579309980046756e-06, "loss": 0.7136, "step": 5334 }, { "epoch": 0.64, "learning_rate": 4.579131511997553e-06, "loss": 0.7766, "step": 5335 }, { "epoch": 0.64, "learning_rate": 4.578953009580132e-06, "loss": 0.891, "step": 5336 }, { "epoch": 0.64, "learning_rate": 4.578774472797442e-06, "loss": 0.9718, "step": 5337 }, { "epoch": 0.64, "learning_rate": 4.578595901652437e-06, "loss": 0.8219, "step": 5338 }, { "epoch": 0.64, "learning_rate": 4.578417296148067e-06, "loss": 0.7625, "step": 5339 }, { "epoch": 0.64, "learning_rate": 4.578238656287285e-06, "loss": 0.7523, "step": 5340 }, { "epoch": 0.64, "learning_rate": 4.578059982073043e-06, "loss": 0.7069, "step": 5341 }, { "epoch": 0.64, "learning_rate": 4.577881273508295e-06, "loss": 0.9771, "step": 5342 }, { "epoch": 0.64, "learning_rate": 4.577702530595995e-06, "loss": 1.0726, "step": 5343 }, { "epoch": 0.64, "learning_rate": 4.577523753339099e-06, "loss": 0.8737, "step": 5344 }, { "epoch": 0.64, "learning_rate": 4.5773449417405605e-06, "loss": 0.6797, "step": 5345 }, { "epoch": 0.64, "learning_rate": 4.577166095803336e-06, "loss": 0.716, "step": 5346 }, { "epoch": 0.64, "learning_rate": 4.576987215530381e-06, "loss": 0.6479, "step": 5347 }, { "epoch": 0.64, "learning_rate": 4.576808300924653e-06, "loss": 0.9441, "step": 5348 }, { "epoch": 0.64, "learning_rate": 4.576629351989111e-06, "loss": 1.1143, "step": 5349 }, { "epoch": 0.64, "learning_rate": 4.576450368726711e-06, "loss": 1.2782, "step": 5350 }, { "epoch": 0.64, "learning_rate": 4.5762713511404115e-06, "loss": 0.4957, "step": 5351 }, { "epoch": 0.64, "learning_rate": 4.5760922992331715e-06, "loss": 0.9162, "step": 5352 }, { "epoch": 0.64, "learning_rate": 4.575913213007953e-06, "loss": 0.7073, "step": 5353 }, { "epoch": 0.64, "learning_rate": 4.575734092467714e-06, "loss": 0.814, "step": 5354 }, { "epoch": 0.64, "learning_rate": 4.575554937615417e-06, "loss": 1.0699, "step": 5355 }, { "epoch": 0.64, "learning_rate": 4.575375748454023e-06, "loss": 1.1423, "step": 5356 }, { "epoch": 0.64, "learning_rate": 4.575196524986494e-06, "loss": 0.4997, "step": 5357 }, { "epoch": 0.64, "learning_rate": 4.575017267215791e-06, "loss": 0.6239, "step": 5358 }, { "epoch": 0.64, "learning_rate": 4.574837975144879e-06, "loss": 0.7424, "step": 5359 }, { "epoch": 0.64, "learning_rate": 4.574658648776721e-06, "loss": 0.8281, "step": 5360 }, { "epoch": 0.64, "learning_rate": 4.574479288114282e-06, "loss": 0.9674, "step": 5361 }, { "epoch": 0.64, "learning_rate": 4.574299893160526e-06, "loss": 1.1244, "step": 5362 }, { "epoch": 0.64, "learning_rate": 4.5741204639184185e-06, "loss": 0.5233, "step": 5363 }, { "epoch": 0.64, "learning_rate": 4.573941000390927e-06, "loss": 0.7449, "step": 5364 }, { "epoch": 0.64, "learning_rate": 4.573761502581015e-06, "loss": 0.7666, "step": 5365 }, { "epoch": 0.64, "learning_rate": 4.573581970491652e-06, "loss": 0.832, "step": 5366 }, { "epoch": 0.64, "learning_rate": 4.5734024041258055e-06, "loss": 1.0508, "step": 5367 }, { "epoch": 0.64, "learning_rate": 4.573222803486442e-06, "loss": 1.0988, "step": 5368 }, { "epoch": 0.64, "learning_rate": 4.573043168576533e-06, "loss": 0.7874, "step": 5369 }, { "epoch": 0.64, "learning_rate": 4.572863499399046e-06, "loss": 0.6757, "step": 5370 }, { "epoch": 0.64, "learning_rate": 4.572683795956952e-06, "loss": 0.7426, "step": 5371 }, { "epoch": 0.64, "learning_rate": 4.57250405825322e-06, "loss": 0.7605, "step": 5372 }, { "epoch": 0.64, "learning_rate": 4.5723242862908225e-06, "loss": 0.9813, "step": 5373 }, { "epoch": 0.64, "learning_rate": 4.57214448007273e-06, "loss": 1.0131, "step": 5374 }, { "epoch": 0.64, "learning_rate": 4.571964639601917e-06, "loss": 1.0833, "step": 5375 }, { "epoch": 0.64, "learning_rate": 4.571784764881353e-06, "loss": 0.6656, "step": 5376 }, { "epoch": 0.64, "learning_rate": 4.571604855914014e-06, "loss": 0.7924, "step": 5377 }, { "epoch": 0.64, "learning_rate": 4.571424912702872e-06, "loss": 0.8095, "step": 5378 }, { "epoch": 0.64, "learning_rate": 4.571244935250903e-06, "loss": 0.9023, "step": 5379 }, { "epoch": 0.64, "learning_rate": 4.571064923561081e-06, "loss": 1.0569, "step": 5380 }, { "epoch": 0.64, "learning_rate": 4.5708848776363824e-06, "loss": 1.2153, "step": 5381 }, { "epoch": 0.64, "learning_rate": 4.570704797479783e-06, "loss": 0.4426, "step": 5382 }, { "epoch": 0.64, "learning_rate": 4.570524683094259e-06, "loss": 0.6685, "step": 5383 }, { "epoch": 0.64, "learning_rate": 4.570344534482789e-06, "loss": 0.7655, "step": 5384 }, { "epoch": 0.64, "learning_rate": 4.570164351648349e-06, "loss": 0.8218, "step": 5385 }, { "epoch": 0.64, "learning_rate": 4.56998413459392e-06, "loss": 1.0131, "step": 5386 }, { "epoch": 0.64, "learning_rate": 4.569803883322479e-06, "loss": 1.0476, "step": 5387 }, { "epoch": 0.64, "learning_rate": 4.569623597837006e-06, "loss": 0.6667, "step": 5388 }, { "epoch": 0.64, "learning_rate": 4.569443278140482e-06, "loss": 0.7321, "step": 5389 }, { "epoch": 0.64, "learning_rate": 4.569262924235886e-06, "loss": 0.6508, "step": 5390 }, { "epoch": 0.64, "learning_rate": 4.5690825361262005e-06, "loss": 0.7108, "step": 5391 }, { "epoch": 0.64, "learning_rate": 4.568902113814408e-06, "loss": 1.0516, "step": 5392 }, { "epoch": 0.64, "learning_rate": 4.568721657303489e-06, "loss": 1.2039, "step": 5393 }, { "epoch": 0.64, "learning_rate": 4.568541166596427e-06, "loss": 0.8587, "step": 5394 }, { "epoch": 0.64, "learning_rate": 4.568360641696207e-06, "loss": 0.6318, "step": 5395 }, { "epoch": 0.64, "learning_rate": 4.568180082605811e-06, "loss": 0.7113, "step": 5396 }, { "epoch": 0.64, "learning_rate": 4.5679994893282256e-06, "loss": 0.7501, "step": 5397 }, { "epoch": 0.64, "learning_rate": 4.5678188618664345e-06, "loss": 0.9438, "step": 5398 }, { "epoch": 0.64, "learning_rate": 4.567638200223424e-06, "loss": 1.1943, "step": 5399 }, { "epoch": 0.64, "learning_rate": 4.567457504402182e-06, "loss": 1.0693, "step": 5400 }, { "epoch": 0.64, "learning_rate": 4.5672767744056925e-06, "loss": 0.6222, "step": 5401 }, { "epoch": 0.64, "learning_rate": 4.567096010236945e-06, "loss": 0.8075, "step": 5402 }, { "epoch": 0.64, "learning_rate": 4.566915211898927e-06, "loss": 0.8884, "step": 5403 }, { "epoch": 0.64, "learning_rate": 4.566734379394627e-06, "loss": 0.9146, "step": 5404 }, { "epoch": 0.64, "learning_rate": 4.566553512727034e-06, "loss": 1.0397, "step": 5405 }, { "epoch": 0.64, "learning_rate": 4.566372611899138e-06, "loss": 1.3266, "step": 5406 }, { "epoch": 0.64, "learning_rate": 4.56619167691393e-06, "loss": 0.5822, "step": 5407 }, { "epoch": 0.64, "learning_rate": 4.5660107077744e-06, "loss": 0.7363, "step": 5408 }, { "epoch": 0.64, "learning_rate": 4.5658297044835395e-06, "loss": 0.7693, "step": 5409 }, { "epoch": 0.64, "learning_rate": 4.565648667044341e-06, "loss": 0.7617, "step": 5410 }, { "epoch": 0.64, "learning_rate": 4.565467595459796e-06, "loss": 0.9074, "step": 5411 }, { "epoch": 0.64, "learning_rate": 4.565286489732898e-06, "loss": 0.9899, "step": 5412 }, { "epoch": 0.64, "learning_rate": 4.565105349866643e-06, "loss": 0.5693, "step": 5413 }, { "epoch": 0.64, "learning_rate": 4.564924175864021e-06, "loss": 0.6281, "step": 5414 }, { "epoch": 0.64, "learning_rate": 4.564742967728031e-06, "loss": 0.5653, "step": 5415 }, { "epoch": 0.64, "learning_rate": 4.564561725461666e-06, "loss": 0.7485, "step": 5416 }, { "epoch": 0.65, "learning_rate": 4.564380449067922e-06, "loss": 0.9016, "step": 5417 }, { "epoch": 0.65, "learning_rate": 4.564199138549797e-06, "loss": 1.1556, "step": 5418 }, { "epoch": 0.65, "learning_rate": 4.564017793910285e-06, "loss": 0.6751, "step": 5419 }, { "epoch": 0.65, "learning_rate": 4.563836415152387e-06, "loss": 0.402, "step": 5420 }, { "epoch": 0.65, "learning_rate": 4.563655002279099e-06, "loss": 0.83, "step": 5421 }, { "epoch": 0.65, "learning_rate": 4.563473555293422e-06, "loss": 0.9472, "step": 5422 }, { "epoch": 0.65, "learning_rate": 4.563292074198353e-06, "loss": 0.8469, "step": 5423 }, { "epoch": 0.65, "learning_rate": 4.563110558996894e-06, "loss": 1.1151, "step": 5424 }, { "epoch": 0.65, "learning_rate": 4.562929009692043e-06, "loss": 1.0289, "step": 5425 }, { "epoch": 0.65, "learning_rate": 4.562747426286803e-06, "loss": 0.6404, "step": 5426 }, { "epoch": 0.65, "learning_rate": 4.562565808784175e-06, "loss": 0.7626, "step": 5427 }, { "epoch": 0.65, "learning_rate": 4.562384157187161e-06, "loss": 0.791, "step": 5428 }, { "epoch": 0.65, "learning_rate": 4.562202471498764e-06, "loss": 0.894, "step": 5429 }, { "epoch": 0.65, "learning_rate": 4.562020751721987e-06, "loss": 1.0399, "step": 5430 }, { "epoch": 0.65, "learning_rate": 4.561838997859834e-06, "loss": 1.1445, "step": 5431 }, { "epoch": 0.65, "learning_rate": 4.561657209915309e-06, "loss": 0.7023, "step": 5432 }, { "epoch": 0.65, "learning_rate": 4.5614753878914175e-06, "loss": 0.6812, "step": 5433 }, { "epoch": 0.65, "learning_rate": 4.561293531791165e-06, "loss": 0.7401, "step": 5434 }, { "epoch": 0.65, "learning_rate": 4.561111641617558e-06, "loss": 0.7781, "step": 5435 }, { "epoch": 0.65, "learning_rate": 4.5609297173736025e-06, "loss": 0.8646, "step": 5436 }, { "epoch": 0.65, "learning_rate": 4.560747759062305e-06, "loss": 1.0756, "step": 5437 }, { "epoch": 0.65, "learning_rate": 4.560565766686674e-06, "loss": 0.7891, "step": 5438 }, { "epoch": 0.65, "learning_rate": 4.560383740249719e-06, "loss": 0.7104, "step": 5439 }, { "epoch": 0.65, "learning_rate": 4.560201679754448e-06, "loss": 0.7955, "step": 5440 }, { "epoch": 0.65, "learning_rate": 4.5600195852038696e-06, "loss": 0.7991, "step": 5441 }, { "epoch": 0.65, "learning_rate": 4.559837456600995e-06, "loss": 0.9167, "step": 5442 }, { "epoch": 0.65, "learning_rate": 4.559655293948834e-06, "loss": 1.0787, "step": 5443 }, { "epoch": 0.65, "learning_rate": 4.5594730972503985e-06, "loss": 0.7325, "step": 5444 }, { "epoch": 0.65, "learning_rate": 4.5592908665087e-06, "loss": 0.5761, "step": 5445 }, { "epoch": 0.65, "learning_rate": 4.55910860172675e-06, "loss": 0.8262, "step": 5446 }, { "epoch": 0.65, "learning_rate": 4.558926302907562e-06, "loss": 0.7239, "step": 5447 }, { "epoch": 0.65, "learning_rate": 4.55874397005415e-06, "loss": 0.8629, "step": 5448 }, { "epoch": 0.65, "learning_rate": 4.558561603169528e-06, "loss": 1.022, "step": 5449 }, { "epoch": 0.65, "learning_rate": 4.5583792022567094e-06, "loss": 0.9434, "step": 5450 }, { "epoch": 0.65, "learning_rate": 4.55819676731871e-06, "loss": 0.5778, "step": 5451 }, { "epoch": 0.65, "learning_rate": 4.558014298358545e-06, "loss": 0.7061, "step": 5452 }, { "epoch": 0.65, "learning_rate": 4.557831795379231e-06, "loss": 0.675, "step": 5453 }, { "epoch": 0.65, "learning_rate": 4.557649258383785e-06, "loss": 0.9159, "step": 5454 }, { "epoch": 0.65, "learning_rate": 4.557466687375224e-06, "loss": 1.0358, "step": 5455 }, { "epoch": 0.65, "learning_rate": 4.557284082356566e-06, "loss": 1.3269, "step": 5456 }, { "epoch": 0.65, "learning_rate": 4.55710144333083e-06, "loss": 0.4571, "step": 5457 }, { "epoch": 0.65, "learning_rate": 4.556918770301034e-06, "loss": 0.6495, "step": 5458 }, { "epoch": 0.65, "learning_rate": 4.556736063270198e-06, "loss": 0.617, "step": 5459 }, { "epoch": 0.65, "learning_rate": 4.556553322241343e-06, "loss": 0.7794, "step": 5460 }, { "epoch": 0.65, "learning_rate": 4.556370547217488e-06, "loss": 1.0024, "step": 5461 }, { "epoch": 0.65, "learning_rate": 4.556187738201656e-06, "loss": 1.2484, "step": 5462 }, { "epoch": 0.65, "learning_rate": 4.556004895196868e-06, "loss": 0.8924, "step": 5463 }, { "epoch": 0.65, "learning_rate": 4.555822018206146e-06, "loss": 0.688, "step": 5464 }, { "epoch": 0.65, "learning_rate": 4.5556391072325145e-06, "loss": 0.7199, "step": 5465 }, { "epoch": 0.65, "learning_rate": 4.555456162278996e-06, "loss": 0.8165, "step": 5466 }, { "epoch": 0.65, "learning_rate": 4.555273183348614e-06, "loss": 0.9779, "step": 5467 }, { "epoch": 0.65, "learning_rate": 4.5550901704443945e-06, "loss": 1.1536, "step": 5468 }, { "epoch": 0.65, "learning_rate": 4.554907123569362e-06, "loss": 0.7057, "step": 5469 }, { "epoch": 0.65, "learning_rate": 4.554724042726542e-06, "loss": 0.7448, "step": 5470 }, { "epoch": 0.65, "learning_rate": 4.554540927918961e-06, "loss": 0.6281, "step": 5471 }, { "epoch": 0.65, "learning_rate": 4.554357779149646e-06, "loss": 0.7757, "step": 5472 }, { "epoch": 0.65, "learning_rate": 4.554174596421625e-06, "loss": 0.9635, "step": 5473 }, { "epoch": 0.65, "learning_rate": 4.553991379737925e-06, "loss": 0.8873, "step": 5474 }, { "epoch": 0.65, "learning_rate": 4.553808129101577e-06, "loss": 1.2994, "step": 5475 }, { "epoch": 0.65, "learning_rate": 4.5536248445156066e-06, "loss": 0.5506, "step": 5476 }, { "epoch": 0.65, "learning_rate": 4.553441525983045e-06, "loss": 0.7945, "step": 5477 }, { "epoch": 0.65, "learning_rate": 4.553258173506923e-06, "loss": 0.7871, "step": 5478 }, { "epoch": 0.65, "learning_rate": 4.5530747870902715e-06, "loss": 0.84, "step": 5479 }, { "epoch": 0.65, "learning_rate": 4.552891366736121e-06, "loss": 1.0148, "step": 5480 }, { "epoch": 0.65, "learning_rate": 4.552707912447504e-06, "loss": 1.165, "step": 5481 }, { "epoch": 0.65, "learning_rate": 4.5525244242274535e-06, "loss": 0.569, "step": 5482 }, { "epoch": 0.65, "learning_rate": 4.552340902079002e-06, "loss": 0.6751, "step": 5483 }, { "epoch": 0.65, "learning_rate": 4.5521573460051824e-06, "loss": 0.6802, "step": 5484 }, { "epoch": 0.65, "learning_rate": 4.55197375600903e-06, "loss": 0.7645, "step": 5485 }, { "epoch": 0.65, "learning_rate": 4.55179013209358e-06, "loss": 0.9907, "step": 5486 }, { "epoch": 0.65, "learning_rate": 4.551606474261866e-06, "loss": 1.1234, "step": 5487 }, { "epoch": 0.65, "learning_rate": 4.551422782516926e-06, "loss": 0.624, "step": 5488 }, { "epoch": 0.65, "learning_rate": 4.551239056861793e-06, "loss": 0.6163, "step": 5489 }, { "epoch": 0.65, "learning_rate": 4.5510552972995095e-06, "loss": 0.747, "step": 5490 }, { "epoch": 0.65, "learning_rate": 4.550871503833107e-06, "loss": 0.7097, "step": 5491 }, { "epoch": 0.65, "learning_rate": 4.5506876764656285e-06, "loss": 0.976, "step": 5492 }, { "epoch": 0.65, "learning_rate": 4.550503815200109e-06, "loss": 1.1423, "step": 5493 }, { "epoch": 0.65, "learning_rate": 4.55031992003959e-06, "loss": 0.8126, "step": 5494 }, { "epoch": 0.65, "learning_rate": 4.550135990987111e-06, "loss": 0.6314, "step": 5495 }, { "epoch": 0.65, "learning_rate": 4.549952028045712e-06, "loss": 0.7457, "step": 5496 }, { "epoch": 0.65, "learning_rate": 4.5497680312184335e-06, "loss": 0.936, "step": 5497 }, { "epoch": 0.65, "learning_rate": 4.549584000508318e-06, "loss": 0.933, "step": 5498 }, { "epoch": 0.65, "learning_rate": 4.549399935918407e-06, "loss": 1.1362, "step": 5499 }, { "epoch": 0.65, "learning_rate": 4.549215837451742e-06, "loss": 1.1634, "step": 5500 }, { "epoch": 0.66, "learning_rate": 4.549031705111369e-06, "loss": 0.6071, "step": 5501 }, { "epoch": 0.66, "learning_rate": 4.548847538900329e-06, "loss": 0.6299, "step": 5502 }, { "epoch": 0.66, "learning_rate": 4.548663338821667e-06, "loss": 0.723, "step": 5503 }, { "epoch": 0.66, "learning_rate": 4.548479104878428e-06, "loss": 0.7108, "step": 5504 }, { "epoch": 0.66, "learning_rate": 4.548294837073659e-06, "loss": 1.0643, "step": 5505 }, { "epoch": 0.66, "learning_rate": 4.548110535410403e-06, "loss": 1.1769, "step": 5506 }, { "epoch": 0.66, "learning_rate": 4.547926199891709e-06, "loss": 0.5813, "step": 5507 }, { "epoch": 0.66, "learning_rate": 4.547741830520622e-06, "loss": 0.7004, "step": 5508 }, { "epoch": 0.66, "learning_rate": 4.547557427300191e-06, "loss": 0.6899, "step": 5509 }, { "epoch": 0.66, "learning_rate": 4.5473729902334655e-06, "loss": 0.8652, "step": 5510 }, { "epoch": 0.66, "learning_rate": 4.547188519323491e-06, "loss": 0.8347, "step": 5511 }, { "epoch": 0.66, "learning_rate": 4.547004014573319e-06, "loss": 1.101, "step": 5512 }, { "epoch": 0.66, "learning_rate": 4.546819475985999e-06, "loss": 0.6295, "step": 5513 }, { "epoch": 0.66, "learning_rate": 4.54663490356458e-06, "loss": 0.6908, "step": 5514 }, { "epoch": 0.66, "learning_rate": 4.546450297312116e-06, "loss": 0.6797, "step": 5515 }, { "epoch": 0.66, "learning_rate": 4.546265657231657e-06, "loss": 0.675, "step": 5516 }, { "epoch": 0.66, "learning_rate": 4.546080983326253e-06, "loss": 0.8755, "step": 5517 }, { "epoch": 0.66, "learning_rate": 4.545896275598961e-06, "loss": 1.2243, "step": 5518 }, { "epoch": 0.66, "learning_rate": 4.54571153405283e-06, "loss": 0.9299, "step": 5519 }, { "epoch": 0.66, "learning_rate": 4.545526758690917e-06, "loss": 0.4847, "step": 5520 }, { "epoch": 0.66, "learning_rate": 4.545341949516274e-06, "loss": 0.7351, "step": 5521 }, { "epoch": 0.66, "learning_rate": 4.545157106531958e-06, "loss": 0.8987, "step": 5522 }, { "epoch": 0.66, "learning_rate": 4.544972229741022e-06, "loss": 1.0862, "step": 5523 }, { "epoch": 0.66, "learning_rate": 4.544787319146525e-06, "loss": 1.0432, "step": 5524 }, { "epoch": 0.66, "learning_rate": 4.5446023747515215e-06, "loss": 1.1569, "step": 5525 }, { "epoch": 0.66, "learning_rate": 4.544417396559069e-06, "loss": 0.4818, "step": 5526 }, { "epoch": 0.66, "learning_rate": 4.544232384572226e-06, "loss": 0.7835, "step": 5527 }, { "epoch": 0.66, "learning_rate": 4.54404733879405e-06, "loss": 0.71, "step": 5528 }, { "epoch": 0.66, "learning_rate": 4.5438622592276e-06, "loss": 0.7975, "step": 5529 }, { "epoch": 0.66, "learning_rate": 4.543677145875934e-06, "loss": 1.0444, "step": 5530 }, { "epoch": 0.66, "learning_rate": 4.543491998742114e-06, "loss": 1.2542, "step": 5531 }, { "epoch": 0.66, "learning_rate": 4.5433068178292e-06, "loss": 0.7928, "step": 5532 }, { "epoch": 0.66, "learning_rate": 4.543121603140254e-06, "loss": 0.7343, "step": 5533 }, { "epoch": 0.66, "learning_rate": 4.542936354678336e-06, "loss": 0.719, "step": 5534 }, { "epoch": 0.66, "learning_rate": 4.542751072446507e-06, "loss": 0.8365, "step": 5535 }, { "epoch": 0.66, "learning_rate": 4.542565756447833e-06, "loss": 1.0239, "step": 5536 }, { "epoch": 0.66, "learning_rate": 4.542380406685375e-06, "loss": 1.1903, "step": 5537 }, { "epoch": 0.66, "learning_rate": 4.542195023162197e-06, "loss": 0.6349, "step": 5538 }, { "epoch": 0.66, "learning_rate": 4.542009605881365e-06, "loss": 0.7761, "step": 5539 }, { "epoch": 0.66, "learning_rate": 4.541824154845942e-06, "loss": 0.7657, "step": 5540 }, { "epoch": 0.66, "learning_rate": 4.5416386700589956e-06, "loss": 0.8081, "step": 5541 }, { "epoch": 0.66, "learning_rate": 4.54145315152359e-06, "loss": 0.9976, "step": 5542 }, { "epoch": 0.66, "learning_rate": 4.541267599242793e-06, "loss": 1.29, "step": 5543 }, { "epoch": 0.66, "learning_rate": 4.54108201321967e-06, "loss": 0.8661, "step": 5544 }, { "epoch": 0.66, "learning_rate": 4.54089639345729e-06, "loss": 0.6629, "step": 5545 }, { "epoch": 0.66, "learning_rate": 4.540710739958722e-06, "loss": 0.8716, "step": 5546 }, { "epoch": 0.66, "learning_rate": 4.540525052727035e-06, "loss": 0.7798, "step": 5547 }, { "epoch": 0.66, "learning_rate": 4.540339331765297e-06, "loss": 0.7507, "step": 5548 }, { "epoch": 0.66, "learning_rate": 4.540153577076578e-06, "loss": 1.0547, "step": 5549 }, { "epoch": 0.66, "learning_rate": 4.53996778866395e-06, "loss": 1.0582, "step": 5550 }, { "epoch": 0.66, "learning_rate": 4.539781966530484e-06, "loss": 0.5342, "step": 5551 }, { "epoch": 0.66, "learning_rate": 4.53959611067925e-06, "loss": 0.6505, "step": 5552 }, { "epoch": 0.66, "learning_rate": 4.539410221113321e-06, "loss": 0.8262, "step": 5553 }, { "epoch": 0.66, "learning_rate": 4.53922429783577e-06, "loss": 0.7814, "step": 5554 }, { "epoch": 0.66, "learning_rate": 4.5390383408496704e-06, "loss": 1.0088, "step": 5555 }, { "epoch": 0.66, "learning_rate": 4.538852350158097e-06, "loss": 1.1686, "step": 5556 }, { "epoch": 0.66, "learning_rate": 4.538666325764122e-06, "loss": 0.4732, "step": 5557 }, { "epoch": 0.66, "learning_rate": 4.538480267670822e-06, "loss": 0.6932, "step": 5558 }, { "epoch": 0.66, "learning_rate": 4.538294175881272e-06, "loss": 0.7104, "step": 5559 }, { "epoch": 0.66, "learning_rate": 4.538108050398548e-06, "loss": 0.7418, "step": 5560 }, { "epoch": 0.66, "learning_rate": 4.537921891225726e-06, "loss": 1.0038, "step": 5561 }, { "epoch": 0.66, "learning_rate": 4.5377356983658855e-06, "loss": 1.2507, "step": 5562 }, { "epoch": 0.66, "learning_rate": 4.537549471822103e-06, "loss": 0.6994, "step": 5563 }, { "epoch": 0.66, "learning_rate": 4.537363211597457e-06, "loss": 0.5862, "step": 5564 }, { "epoch": 0.66, "learning_rate": 4.537176917695024e-06, "loss": 0.8962, "step": 5565 }, { "epoch": 0.66, "learning_rate": 4.536990590117888e-06, "loss": 0.8287, "step": 5566 }, { "epoch": 0.66, "learning_rate": 4.536804228869125e-06, "loss": 0.9523, "step": 5567 }, { "epoch": 0.66, "learning_rate": 4.536617833951818e-06, "loss": 1.0932, "step": 5568 }, { "epoch": 0.66, "learning_rate": 4.536431405369048e-06, "loss": 0.6414, "step": 5569 }, { "epoch": 0.66, "learning_rate": 4.5362449431238955e-06, "loss": 0.6124, "step": 5570 }, { "epoch": 0.66, "learning_rate": 4.536058447219443e-06, "loss": 0.81, "step": 5571 }, { "epoch": 0.66, "learning_rate": 4.535871917658773e-06, "loss": 0.7007, "step": 5572 }, { "epoch": 0.66, "learning_rate": 4.53568535444497e-06, "loss": 0.7971, "step": 5573 }, { "epoch": 0.66, "learning_rate": 4.535498757581117e-06, "loss": 1.1167, "step": 5574 }, { "epoch": 0.66, "learning_rate": 4.5353121270702996e-06, "loss": 1.1656, "step": 5575 }, { "epoch": 0.66, "learning_rate": 4.535125462915601e-06, "loss": 0.6004, "step": 5576 }, { "epoch": 0.66, "learning_rate": 4.534938765120108e-06, "loss": 0.6571, "step": 5577 }, { "epoch": 0.66, "learning_rate": 4.534752033686907e-06, "loss": 0.7978, "step": 5578 }, { "epoch": 0.66, "learning_rate": 4.534565268619083e-06, "loss": 0.9713, "step": 5579 }, { "epoch": 0.66, "learning_rate": 4.534378469919725e-06, "loss": 1.0473, "step": 5580 }, { "epoch": 0.66, "learning_rate": 4.534191637591921e-06, "loss": 1.3193, "step": 5581 }, { "epoch": 0.66, "learning_rate": 4.5340047716387575e-06, "loss": 0.6112, "step": 5582 }, { "epoch": 0.66, "learning_rate": 4.533817872063324e-06, "loss": 0.7195, "step": 5583 }, { "epoch": 0.66, "learning_rate": 4.533630938868711e-06, "loss": 0.8328, "step": 5584 }, { "epoch": 0.67, "learning_rate": 4.533443972058009e-06, "loss": 0.702, "step": 5585 }, { "epoch": 0.67, "learning_rate": 4.533256971634307e-06, "loss": 1.002, "step": 5586 }, { "epoch": 0.67, "learning_rate": 4.533069937600696e-06, "loss": 1.0815, "step": 5587 }, { "epoch": 0.67, "learning_rate": 4.532882869960269e-06, "loss": 0.6772, "step": 5588 }, { "epoch": 0.67, "learning_rate": 4.532695768716117e-06, "loss": 0.6353, "step": 5589 }, { "epoch": 0.67, "learning_rate": 4.532508633871334e-06, "loss": 0.6382, "step": 5590 }, { "epoch": 0.67, "learning_rate": 4.532321465429012e-06, "loss": 0.7204, "step": 5591 }, { "epoch": 0.67, "learning_rate": 4.532134263392246e-06, "loss": 0.9243, "step": 5592 }, { "epoch": 0.67, "learning_rate": 4.53194702776413e-06, "loss": 1.2368, "step": 5593 }, { "epoch": 0.67, "learning_rate": 4.531759758547759e-06, "loss": 0.6802, "step": 5594 }, { "epoch": 0.67, "learning_rate": 4.531572455746229e-06, "loss": 0.721, "step": 5595 }, { "epoch": 0.67, "learning_rate": 4.531385119362635e-06, "loss": 0.7267, "step": 5596 }, { "epoch": 0.67, "learning_rate": 4.531197749400077e-06, "loss": 0.762, "step": 5597 }, { "epoch": 0.67, "learning_rate": 4.531010345861648e-06, "loss": 0.722, "step": 5598 }, { "epoch": 0.67, "learning_rate": 4.530822908750447e-06, "loss": 0.9951, "step": 5599 }, { "epoch": 0.67, "learning_rate": 4.530635438069574e-06, "loss": 1.1486, "step": 5600 }, { "epoch": 0.67, "learning_rate": 4.530447933822126e-06, "loss": 0.6585, "step": 5601 }, { "epoch": 0.67, "learning_rate": 4.530260396011205e-06, "loss": 0.7217, "step": 5602 }, { "epoch": 0.67, "learning_rate": 4.530072824639907e-06, "loss": 0.7975, "step": 5603 }, { "epoch": 0.67, "learning_rate": 4.529885219711336e-06, "loss": 0.7869, "step": 5604 }, { "epoch": 0.67, "learning_rate": 4.529697581228592e-06, "loss": 1.0695, "step": 5605 }, { "epoch": 0.67, "learning_rate": 4.529509909194777e-06, "loss": 1.1848, "step": 5606 }, { "epoch": 0.67, "learning_rate": 4.529322203612992e-06, "loss": 0.6718, "step": 5607 }, { "epoch": 0.67, "learning_rate": 4.529134464486341e-06, "loss": 0.6841, "step": 5608 }, { "epoch": 0.67, "learning_rate": 4.528946691817927e-06, "loss": 0.8715, "step": 5609 }, { "epoch": 0.67, "learning_rate": 4.528758885610854e-06, "loss": 0.9285, "step": 5610 }, { "epoch": 0.67, "learning_rate": 4.5285710458682274e-06, "loss": 0.9644, "step": 5611 }, { "epoch": 0.67, "learning_rate": 4.52838317259315e-06, "loss": 1.2192, "step": 5612 }, { "epoch": 0.67, "learning_rate": 4.528195265788729e-06, "loss": 0.6136, "step": 5613 }, { "epoch": 0.67, "learning_rate": 4.52800732545807e-06, "loss": 0.5895, "step": 5614 }, { "epoch": 0.67, "learning_rate": 4.52781935160428e-06, "loss": 0.7141, "step": 5615 }, { "epoch": 0.67, "learning_rate": 4.527631344230466e-06, "loss": 0.8092, "step": 5616 }, { "epoch": 0.67, "learning_rate": 4.527443303339735e-06, "loss": 0.8825, "step": 5617 }, { "epoch": 0.67, "learning_rate": 4.527255228935197e-06, "loss": 1.0549, "step": 5618 }, { "epoch": 0.67, "learning_rate": 4.527067121019959e-06, "loss": 0.8265, "step": 5619 }, { "epoch": 0.67, "learning_rate": 4.526878979597133e-06, "loss": 0.7116, "step": 5620 }, { "epoch": 0.67, "learning_rate": 4.5266908046698255e-06, "loss": 0.6524, "step": 5621 }, { "epoch": 0.67, "learning_rate": 4.52650259624115e-06, "loss": 0.7607, "step": 5622 }, { "epoch": 0.67, "learning_rate": 4.526314354314216e-06, "loss": 1.0906, "step": 5623 }, { "epoch": 0.67, "learning_rate": 4.5261260788921354e-06, "loss": 1.2638, "step": 5624 }, { "epoch": 0.67, "learning_rate": 4.525937769978022e-06, "loss": 1.0609, "step": 5625 }, { "epoch": 0.67, "learning_rate": 4.525749427574987e-06, "loss": 0.503, "step": 5626 }, { "epoch": 0.67, "learning_rate": 4.525561051686142e-06, "loss": 0.7226, "step": 5627 }, { "epoch": 0.67, "learning_rate": 4.5253726423146045e-06, "loss": 0.8851, "step": 5628 }, { "epoch": 0.67, "learning_rate": 4.525184199463488e-06, "loss": 0.8441, "step": 5629 }, { "epoch": 0.67, "learning_rate": 4.524995723135906e-06, "loss": 1.0079, "step": 5630 }, { "epoch": 0.67, "learning_rate": 4.524807213334975e-06, "loss": 1.0535, "step": 5631 }, { "epoch": 0.67, "learning_rate": 4.52461867006381e-06, "loss": 0.5177, "step": 5632 }, { "epoch": 0.67, "learning_rate": 4.524430093325529e-06, "loss": 0.6554, "step": 5633 }, { "epoch": 0.67, "learning_rate": 4.52424148312325e-06, "loss": 0.8575, "step": 5634 }, { "epoch": 0.67, "learning_rate": 4.524052839460088e-06, "loss": 0.9162, "step": 5635 }, { "epoch": 0.67, "learning_rate": 4.523864162339162e-06, "loss": 0.894, "step": 5636 }, { "epoch": 0.67, "learning_rate": 4.523675451763593e-06, "loss": 1.1588, "step": 5637 }, { "epoch": 0.67, "learning_rate": 4.523486707736499e-06, "loss": 0.5834, "step": 5638 }, { "epoch": 0.67, "learning_rate": 4.523297930260999e-06, "loss": 0.7252, "step": 5639 }, { "epoch": 0.67, "learning_rate": 4.5231091193402155e-06, "loss": 0.6811, "step": 5640 }, { "epoch": 0.67, "learning_rate": 4.522920274977268e-06, "loss": 0.8509, "step": 5641 }, { "epoch": 0.67, "learning_rate": 4.522731397175278e-06, "loss": 0.7811, "step": 5642 }, { "epoch": 0.67, "learning_rate": 4.522542485937369e-06, "loss": 1.1521, "step": 5643 }, { "epoch": 0.67, "learning_rate": 4.522353541266663e-06, "loss": 0.6816, "step": 5644 }, { "epoch": 0.67, "learning_rate": 4.522164563166283e-06, "loss": 0.6336, "step": 5645 }, { "epoch": 0.67, "learning_rate": 4.521975551639353e-06, "loss": 0.7562, "step": 5646 }, { "epoch": 0.67, "learning_rate": 4.521786506688997e-06, "loss": 0.7134, "step": 5647 }, { "epoch": 0.67, "learning_rate": 4.521597428318342e-06, "loss": 0.9004, "step": 5648 }, { "epoch": 0.67, "learning_rate": 4.521408316530511e-06, "loss": 1.1266, "step": 5649 }, { "epoch": 0.67, "learning_rate": 4.52121917132863e-06, "loss": 1.0144, "step": 5650 }, { "epoch": 0.67, "learning_rate": 4.521029992715828e-06, "loss": 0.6213, "step": 5651 }, { "epoch": 0.67, "learning_rate": 4.52084078069523e-06, "loss": 0.7906, "step": 5652 }, { "epoch": 0.67, "learning_rate": 4.5206515352699645e-06, "loss": 0.8297, "step": 5653 }, { "epoch": 0.67, "learning_rate": 4.52046225644316e-06, "loss": 0.7873, "step": 5654 }, { "epoch": 0.67, "learning_rate": 4.520272944217944e-06, "loss": 0.9665, "step": 5655 }, { "epoch": 0.67, "learning_rate": 4.520083598597446e-06, "loss": 1.2101, "step": 5656 }, { "epoch": 0.67, "learning_rate": 4.519894219584798e-06, "loss": 0.5424, "step": 5657 }, { "epoch": 0.67, "learning_rate": 4.519704807183129e-06, "loss": 0.5911, "step": 5658 }, { "epoch": 0.67, "learning_rate": 4.51951536139557e-06, "loss": 0.7466, "step": 5659 }, { "epoch": 0.67, "learning_rate": 4.519325882225252e-06, "loss": 0.8676, "step": 5660 }, { "epoch": 0.67, "learning_rate": 4.519136369675309e-06, "loss": 1.0114, "step": 5661 }, { "epoch": 0.67, "learning_rate": 4.518946823748871e-06, "loss": 1.1876, "step": 5662 }, { "epoch": 0.67, "learning_rate": 4.518757244449073e-06, "loss": 0.5433, "step": 5663 }, { "epoch": 0.67, "learning_rate": 4.518567631779049e-06, "loss": 0.7633, "step": 5664 }, { "epoch": 0.67, "learning_rate": 4.518377985741932e-06, "loss": 0.7689, "step": 5665 }, { "epoch": 0.67, "learning_rate": 4.518188306340857e-06, "loss": 0.8015, "step": 5666 }, { "epoch": 0.67, "learning_rate": 4.5179985935789615e-06, "loss": 0.8757, "step": 5667 }, { "epoch": 0.68, "learning_rate": 4.517808847459379e-06, "loss": 1.1818, "step": 5668 }, { "epoch": 0.68, "learning_rate": 4.517619067985246e-06, "loss": 0.8443, "step": 5669 }, { "epoch": 0.68, "learning_rate": 4.517429255159702e-06, "loss": 0.6208, "step": 5670 }, { "epoch": 0.68, "learning_rate": 4.517239408985883e-06, "loss": 0.804, "step": 5671 }, { "epoch": 0.68, "learning_rate": 4.517049529466927e-06, "loss": 0.7433, "step": 5672 }, { "epoch": 0.68, "learning_rate": 4.516859616605973e-06, "loss": 0.9041, "step": 5673 }, { "epoch": 0.68, "learning_rate": 4.5166696704061605e-06, "loss": 1.1085, "step": 5674 }, { "epoch": 0.68, "learning_rate": 4.516479690870628e-06, "loss": 1.1392, "step": 5675 }, { "epoch": 0.68, "learning_rate": 4.5162896780025185e-06, "loss": 0.5576, "step": 5676 }, { "epoch": 0.68, "learning_rate": 4.516099631804971e-06, "loss": 0.8668, "step": 5677 }, { "epoch": 0.68, "learning_rate": 4.515909552281128e-06, "loss": 0.8399, "step": 5678 }, { "epoch": 0.68, "learning_rate": 4.5157194394341295e-06, "loss": 0.7902, "step": 5679 }, { "epoch": 0.68, "learning_rate": 4.515529293267121e-06, "loss": 0.8552, "step": 5680 }, { "epoch": 0.68, "learning_rate": 4.515339113783243e-06, "loss": 1.0942, "step": 5681 }, { "epoch": 0.68, "learning_rate": 4.515148900985641e-06, "loss": 0.6672, "step": 5682 }, { "epoch": 0.68, "learning_rate": 4.5149586548774585e-06, "loss": 0.8088, "step": 5683 }, { "epoch": 0.68, "learning_rate": 4.51476837546184e-06, "loss": 0.779, "step": 5684 }, { "epoch": 0.68, "learning_rate": 4.514578062741932e-06, "loss": 0.794, "step": 5685 }, { "epoch": 0.68, "learning_rate": 4.514387716720879e-06, "loss": 1.0216, "step": 5686 }, { "epoch": 0.68, "learning_rate": 4.514197337401828e-06, "loss": 1.1355, "step": 5687 }, { "epoch": 0.68, "learning_rate": 4.514006924787926e-06, "loss": 0.5884, "step": 5688 }, { "epoch": 0.68, "learning_rate": 4.51381647888232e-06, "loss": 0.6722, "step": 5689 }, { "epoch": 0.68, "learning_rate": 4.5136259996881595e-06, "loss": 0.8976, "step": 5690 }, { "epoch": 0.68, "learning_rate": 4.513435487208593e-06, "loss": 0.7089, "step": 5691 }, { "epoch": 0.68, "learning_rate": 4.513244941446767e-06, "loss": 0.9844, "step": 5692 }, { "epoch": 0.68, "learning_rate": 4.513054362405834e-06, "loss": 1.1016, "step": 5693 }, { "epoch": 0.68, "learning_rate": 4.512863750088944e-06, "loss": 0.8108, "step": 5694 }, { "epoch": 0.68, "learning_rate": 4.512673104499247e-06, "loss": 0.588, "step": 5695 }, { "epoch": 0.68, "learning_rate": 4.512482425639895e-06, "loss": 0.7352, "step": 5696 }, { "epoch": 0.68, "learning_rate": 4.51229171351404e-06, "loss": 0.8398, "step": 5697 }, { "epoch": 0.68, "learning_rate": 4.512100968124833e-06, "loss": 0.8685, "step": 5698 }, { "epoch": 0.68, "learning_rate": 4.511910189475428e-06, "loss": 1.0956, "step": 5699 }, { "epoch": 0.68, "learning_rate": 4.51171937756898e-06, "loss": 1.2144, "step": 5700 }, { "epoch": 0.68, "learning_rate": 4.511528532408641e-06, "loss": 0.597, "step": 5701 }, { "epoch": 0.68, "learning_rate": 4.511337653997567e-06, "loss": 0.6387, "step": 5702 }, { "epoch": 0.68, "learning_rate": 4.5111467423389125e-06, "loss": 0.8628, "step": 5703 }, { "epoch": 0.68, "learning_rate": 4.510955797435834e-06, "loss": 0.9944, "step": 5704 }, { "epoch": 0.68, "learning_rate": 4.510764819291487e-06, "loss": 0.9827, "step": 5705 }, { "epoch": 0.68, "learning_rate": 4.510573807909029e-06, "loss": 0.9881, "step": 5706 }, { "epoch": 0.68, "learning_rate": 4.510382763291618e-06, "loss": 0.6404, "step": 5707 }, { "epoch": 0.68, "learning_rate": 4.51019168544241e-06, "loss": 0.691, "step": 5708 }, { "epoch": 0.68, "learning_rate": 4.510000574364565e-06, "loss": 0.746, "step": 5709 }, { "epoch": 0.68, "learning_rate": 4.509809430061243e-06, "loss": 0.8985, "step": 5710 }, { "epoch": 0.68, "learning_rate": 4.5096182525356015e-06, "loss": 1.0183, "step": 5711 }, { "epoch": 0.68, "learning_rate": 4.5094270417908015e-06, "loss": 1.1643, "step": 5712 }, { "epoch": 0.68, "learning_rate": 4.5092357978300035e-06, "loss": 0.5468, "step": 5713 }, { "epoch": 0.68, "learning_rate": 4.50904452065637e-06, "loss": 0.7547, "step": 5714 }, { "epoch": 0.68, "learning_rate": 4.508853210273062e-06, "loss": 0.6965, "step": 5715 }, { "epoch": 0.68, "learning_rate": 4.508661866683241e-06, "loss": 0.711, "step": 5716 }, { "epoch": 0.68, "learning_rate": 4.5084704898900725e-06, "loss": 0.9114, "step": 5717 }, { "epoch": 0.68, "learning_rate": 4.508279079896717e-06, "loss": 1.0408, "step": 5718 }, { "epoch": 0.68, "learning_rate": 4.5080876367063405e-06, "loss": 0.846, "step": 5719 }, { "epoch": 0.68, "learning_rate": 4.507896160322106e-06, "loss": 0.5218, "step": 5720 }, { "epoch": 0.68, "learning_rate": 4.50770465074718e-06, "loss": 0.7486, "step": 5721 }, { "epoch": 0.68, "learning_rate": 4.507513107984728e-06, "loss": 0.8259, "step": 5722 }, { "epoch": 0.68, "learning_rate": 4.5073215320379155e-06, "loss": 0.9996, "step": 5723 }, { "epoch": 0.68, "learning_rate": 4.507129922909909e-06, "loss": 1.0076, "step": 5724 }, { "epoch": 0.68, "learning_rate": 4.506938280603878e-06, "loss": 1.2086, "step": 5725 }, { "epoch": 0.68, "learning_rate": 4.506746605122988e-06, "loss": 0.676, "step": 5726 }, { "epoch": 0.68, "learning_rate": 4.506554896470408e-06, "loss": 0.6079, "step": 5727 }, { "epoch": 0.68, "learning_rate": 4.506363154649307e-06, "loss": 0.7177, "step": 5728 }, { "epoch": 0.68, "learning_rate": 4.5061713796628556e-06, "loss": 0.7924, "step": 5729 }, { "epoch": 0.68, "learning_rate": 4.505979571514222e-06, "loss": 0.9575, "step": 5730 }, { "epoch": 0.68, "learning_rate": 4.505787730206579e-06, "loss": 1.0289, "step": 5731 }, { "epoch": 0.68, "learning_rate": 4.505595855743096e-06, "loss": 0.6205, "step": 5732 }, { "epoch": 0.68, "learning_rate": 4.505403948126944e-06, "loss": 0.7101, "step": 5733 }, { "epoch": 0.68, "learning_rate": 4.505212007361298e-06, "loss": 0.5782, "step": 5734 }, { "epoch": 0.68, "learning_rate": 4.505020033449329e-06, "loss": 0.8275, "step": 5735 }, { "epoch": 0.68, "learning_rate": 4.50482802639421e-06, "loss": 1.0073, "step": 5736 }, { "epoch": 0.68, "learning_rate": 4.504635986199115e-06, "loss": 1.0743, "step": 5737 }, { "epoch": 0.68, "learning_rate": 4.504443912867221e-06, "loss": 0.6023, "step": 5738 }, { "epoch": 0.68, "learning_rate": 4.5042518064016995e-06, "loss": 0.5823, "step": 5739 }, { "epoch": 0.68, "learning_rate": 4.504059666805728e-06, "loss": 0.7206, "step": 5740 }, { "epoch": 0.68, "learning_rate": 4.503867494082481e-06, "loss": 0.842, "step": 5741 }, { "epoch": 0.68, "learning_rate": 4.503675288235138e-06, "loss": 0.8727, "step": 5742 }, { "epoch": 0.68, "learning_rate": 4.503483049266872e-06, "loss": 1.0688, "step": 5743 }, { "epoch": 0.68, "learning_rate": 4.503290777180865e-06, "loss": 0.7924, "step": 5744 }, { "epoch": 0.68, "learning_rate": 4.503098471980293e-06, "loss": 0.4986, "step": 5745 }, { "epoch": 0.68, "learning_rate": 4.502906133668334e-06, "loss": 0.6954, "step": 5746 }, { "epoch": 0.68, "learning_rate": 4.5027137622481695e-06, "loss": 0.8176, "step": 5747 }, { "epoch": 0.68, "learning_rate": 4.502521357722979e-06, "loss": 0.9262, "step": 5748 }, { "epoch": 0.68, "learning_rate": 4.502328920095942e-06, "loss": 0.9672, "step": 5749 }, { "epoch": 0.68, "learning_rate": 4.502136449370239e-06, "loss": 1.1959, "step": 5750 }, { "epoch": 0.68, "learning_rate": 4.501943945549054e-06, "loss": 0.542, "step": 5751 }, { "epoch": 0.69, "learning_rate": 4.501751408635567e-06, "loss": 0.6594, "step": 5752 }, { "epoch": 0.69, "learning_rate": 4.501558838632961e-06, "loss": 0.8097, "step": 5753 }, { "epoch": 0.69, "learning_rate": 4.50136623554442e-06, "loss": 0.7889, "step": 5754 }, { "epoch": 0.69, "learning_rate": 4.501173599373126e-06, "loss": 0.9979, "step": 5755 }, { "epoch": 0.69, "learning_rate": 4.5009809301222665e-06, "loss": 1.3202, "step": 5756 }, { "epoch": 0.69, "learning_rate": 4.5007882277950225e-06, "loss": 0.7196, "step": 5757 }, { "epoch": 0.69, "learning_rate": 4.500595492394582e-06, "loss": 0.5687, "step": 5758 }, { "epoch": 0.69, "learning_rate": 4.50040272392413e-06, "loss": 0.9301, "step": 5759 }, { "epoch": 0.69, "learning_rate": 4.5002099223868526e-06, "loss": 0.6554, "step": 5760 }, { "epoch": 0.69, "learning_rate": 4.500017087785938e-06, "loss": 0.8925, "step": 5761 }, { "epoch": 0.69, "learning_rate": 4.499824220124574e-06, "loss": 1.0857, "step": 5762 }, { "epoch": 0.69, "learning_rate": 4.499631319405946e-06, "loss": 0.658, "step": 5763 }, { "epoch": 0.69, "learning_rate": 4.4994383856332455e-06, "loss": 0.6881, "step": 5764 }, { "epoch": 0.69, "learning_rate": 4.4992454188096604e-06, "loss": 0.8043, "step": 5765 }, { "epoch": 0.69, "learning_rate": 4.499052418938381e-06, "loss": 0.7017, "step": 5766 }, { "epoch": 0.69, "learning_rate": 4.498859386022597e-06, "loss": 0.8528, "step": 5767 }, { "epoch": 0.69, "learning_rate": 4.4986663200655e-06, "loss": 1.1277, "step": 5768 }, { "epoch": 0.69, "learning_rate": 4.498473221070281e-06, "loss": 0.7992, "step": 5769 }, { "epoch": 0.69, "learning_rate": 4.498280089040132e-06, "loss": 0.6155, "step": 5770 }, { "epoch": 0.69, "learning_rate": 4.498086923978245e-06, "loss": 0.686, "step": 5771 }, { "epoch": 0.69, "learning_rate": 4.497893725887815e-06, "loss": 0.6549, "step": 5772 }, { "epoch": 0.69, "learning_rate": 4.497700494772033e-06, "loss": 0.8318, "step": 5773 }, { "epoch": 0.69, "learning_rate": 4.497507230634094e-06, "loss": 1.1205, "step": 5774 }, { "epoch": 0.69, "learning_rate": 4.497313933477193e-06, "loss": 1.1512, "step": 5775 }, { "epoch": 0.69, "learning_rate": 4.497120603304525e-06, "loss": 0.6172, "step": 5776 }, { "epoch": 0.69, "learning_rate": 4.496927240119286e-06, "loss": 0.6707, "step": 5777 }, { "epoch": 0.69, "learning_rate": 4.4967338439246725e-06, "loss": 0.9102, "step": 5778 }, { "epoch": 0.69, "learning_rate": 4.49654041472388e-06, "loss": 0.8217, "step": 5779 }, { "epoch": 0.69, "learning_rate": 4.4963469525201076e-06, "loss": 0.9112, "step": 5780 }, { "epoch": 0.69, "learning_rate": 4.496153457316553e-06, "loss": 1.107, "step": 5781 }, { "epoch": 0.69, "learning_rate": 4.495959929116413e-06, "loss": 0.6026, "step": 5782 }, { "epoch": 0.69, "learning_rate": 4.495766367922889e-06, "loss": 0.5279, "step": 5783 }, { "epoch": 0.69, "learning_rate": 4.495572773739179e-06, "loss": 0.799, "step": 5784 }, { "epoch": 0.69, "learning_rate": 4.495379146568484e-06, "loss": 0.8589, "step": 5785 }, { "epoch": 0.69, "learning_rate": 4.495185486414004e-06, "loss": 0.9726, "step": 5786 }, { "epoch": 0.69, "learning_rate": 4.494991793278941e-06, "loss": 1.1976, "step": 5787 }, { "epoch": 0.69, "learning_rate": 4.494798067166496e-06, "loss": 0.7274, "step": 5788 }, { "epoch": 0.69, "learning_rate": 4.494604308079872e-06, "loss": 0.6759, "step": 5789 }, { "epoch": 0.69, "learning_rate": 4.49441051602227e-06, "loss": 0.6779, "step": 5790 }, { "epoch": 0.69, "learning_rate": 4.494216690996896e-06, "loss": 0.8298, "step": 5791 }, { "epoch": 0.69, "learning_rate": 4.494022833006952e-06, "loss": 0.9523, "step": 5792 }, { "epoch": 0.69, "learning_rate": 4.493828942055644e-06, "loss": 1.1243, "step": 5793 }, { "epoch": 0.69, "learning_rate": 4.493635018146176e-06, "loss": 0.7959, "step": 5794 }, { "epoch": 0.69, "learning_rate": 4.493441061281754e-06, "loss": 0.4818, "step": 5795 }, { "epoch": 0.69, "learning_rate": 4.493247071465584e-06, "loss": 0.7616, "step": 5796 }, { "epoch": 0.69, "learning_rate": 4.4930530487008725e-06, "loss": 0.8063, "step": 5797 }, { "epoch": 0.69, "learning_rate": 4.492858992990827e-06, "loss": 0.7628, "step": 5798 }, { "epoch": 0.69, "learning_rate": 4.492664904338655e-06, "loss": 1.0601, "step": 5799 }, { "epoch": 0.69, "learning_rate": 4.492470782747564e-06, "loss": 1.0308, "step": 5800 }, { "epoch": 0.69, "learning_rate": 4.492276628220765e-06, "loss": 0.4871, "step": 5801 }, { "epoch": 0.69, "learning_rate": 4.492082440761467e-06, "loss": 0.6307, "step": 5802 }, { "epoch": 0.69, "learning_rate": 4.491888220372878e-06, "loss": 0.7826, "step": 5803 }, { "epoch": 0.69, "learning_rate": 4.491693967058209e-06, "loss": 1.0693, "step": 5804 }, { "epoch": 0.69, "learning_rate": 4.491499680820672e-06, "loss": 0.9947, "step": 5805 }, { "epoch": 0.69, "learning_rate": 4.491305361663479e-06, "loss": 1.1946, "step": 5806 }, { "epoch": 0.69, "learning_rate": 4.49111100958984e-06, "loss": 0.5234, "step": 5807 }, { "epoch": 0.69, "learning_rate": 4.49091662460297e-06, "loss": 0.5822, "step": 5808 }, { "epoch": 0.69, "learning_rate": 4.490722206706081e-06, "loss": 0.8619, "step": 5809 }, { "epoch": 0.69, "learning_rate": 4.490527755902386e-06, "loss": 0.8066, "step": 5810 }, { "epoch": 0.69, "learning_rate": 4.490333272195101e-06, "loss": 1.0796, "step": 5811 }, { "epoch": 0.69, "learning_rate": 4.49013875558744e-06, "loss": 1.1574, "step": 5812 }, { "epoch": 0.69, "learning_rate": 4.489944206082618e-06, "loss": 0.7857, "step": 5813 }, { "epoch": 0.69, "learning_rate": 4.489749623683852e-06, "loss": 0.611, "step": 5814 }, { "epoch": 0.69, "learning_rate": 4.489555008394357e-06, "loss": 0.7497, "step": 5815 }, { "epoch": 0.69, "learning_rate": 4.489360360217351e-06, "loss": 0.8697, "step": 5816 }, { "epoch": 0.69, "learning_rate": 4.489165679156051e-06, "loss": 0.8646, "step": 5817 }, { "epoch": 0.69, "learning_rate": 4.488970965213676e-06, "loss": 1.1097, "step": 5818 }, { "epoch": 0.69, "learning_rate": 4.488776218393444e-06, "loss": 0.8241, "step": 5819 }, { "epoch": 0.69, "learning_rate": 4.488581438698575e-06, "loss": 0.6205, "step": 5820 }, { "epoch": 0.69, "learning_rate": 4.488386626132286e-06, "loss": 0.7004, "step": 5821 }, { "epoch": 0.69, "learning_rate": 4.488191780697801e-06, "loss": 0.8, "step": 5822 }, { "epoch": 0.69, "learning_rate": 4.4879969023983385e-06, "loss": 0.9453, "step": 5823 }, { "epoch": 0.69, "learning_rate": 4.48780199123712e-06, "loss": 1.1914, "step": 5824 }, { "epoch": 0.69, "learning_rate": 4.487607047217368e-06, "loss": 1.0735, "step": 5825 }, { "epoch": 0.69, "learning_rate": 4.487412070342305e-06, "loss": 0.6432, "step": 5826 }, { "epoch": 0.69, "learning_rate": 4.487217060615153e-06, "loss": 0.8563, "step": 5827 }, { "epoch": 0.69, "learning_rate": 4.487022018039137e-06, "loss": 0.685, "step": 5828 }, { "epoch": 0.69, "learning_rate": 4.48682694261748e-06, "loss": 0.8758, "step": 5829 }, { "epoch": 0.69, "learning_rate": 4.486631834353406e-06, "loss": 1.0374, "step": 5830 }, { "epoch": 0.69, "learning_rate": 4.486436693250143e-06, "loss": 1.046, "step": 5831 }, { "epoch": 0.69, "learning_rate": 4.4862415193109135e-06, "loss": 0.6931, "step": 5832 }, { "epoch": 0.69, "learning_rate": 4.486046312538945e-06, "loss": 0.715, "step": 5833 }, { "epoch": 0.69, "learning_rate": 4.485851072937466e-06, "loss": 0.643, "step": 5834 }, { "epoch": 0.69, "learning_rate": 4.4856558005097e-06, "loss": 0.7681, "step": 5835 }, { "epoch": 0.7, "learning_rate": 4.485460495258878e-06, "loss": 0.9736, "step": 5836 }, { "epoch": 0.7, "learning_rate": 4.4852651571882274e-06, "loss": 1.2021, "step": 5837 }, { "epoch": 0.7, "learning_rate": 4.485069786300977e-06, "loss": 0.6776, "step": 5838 }, { "epoch": 0.7, "learning_rate": 4.484874382600356e-06, "loss": 0.5363, "step": 5839 }, { "epoch": 0.7, "learning_rate": 4.484678946089596e-06, "loss": 0.8426, "step": 5840 }, { "epoch": 0.7, "learning_rate": 4.4844834767719255e-06, "loss": 0.7903, "step": 5841 }, { "epoch": 0.7, "learning_rate": 4.484287974650577e-06, "loss": 0.8654, "step": 5842 }, { "epoch": 0.7, "learning_rate": 4.484092439728782e-06, "loss": 1.246, "step": 5843 }, { "epoch": 0.7, "learning_rate": 4.483896872009772e-06, "loss": 0.878, "step": 5844 }, { "epoch": 0.7, "learning_rate": 4.483701271496781e-06, "loss": 0.5815, "step": 5845 }, { "epoch": 0.7, "learning_rate": 4.483505638193041e-06, "loss": 0.6721, "step": 5846 }, { "epoch": 0.7, "learning_rate": 4.483309972101786e-06, "loss": 0.8433, "step": 5847 }, { "epoch": 0.7, "learning_rate": 4.483114273226251e-06, "loss": 0.9178, "step": 5848 }, { "epoch": 0.7, "learning_rate": 4.482918541569671e-06, "loss": 1.0694, "step": 5849 }, { "epoch": 0.7, "learning_rate": 4.482722777135281e-06, "loss": 1.2499, "step": 5850 }, { "epoch": 0.7, "learning_rate": 4.482526979926317e-06, "loss": 0.6032, "step": 5851 }, { "epoch": 0.7, "learning_rate": 4.482331149946017e-06, "loss": 0.6912, "step": 5852 }, { "epoch": 0.7, "learning_rate": 4.482135287197615e-06, "loss": 0.7361, "step": 5853 }, { "epoch": 0.7, "learning_rate": 4.4819393916843515e-06, "loss": 0.8337, "step": 5854 }, { "epoch": 0.7, "learning_rate": 4.481743463409463e-06, "loss": 1.0129, "step": 5855 }, { "epoch": 0.7, "learning_rate": 4.481547502376188e-06, "loss": 1.3303, "step": 5856 }, { "epoch": 0.7, "learning_rate": 4.481351508587767e-06, "loss": 0.759, "step": 5857 }, { "epoch": 0.7, "learning_rate": 4.48115548204744e-06, "loss": 0.6847, "step": 5858 }, { "epoch": 0.7, "learning_rate": 4.480959422758446e-06, "loss": 0.6254, "step": 5859 }, { "epoch": 0.7, "learning_rate": 4.480763330724027e-06, "loss": 0.7075, "step": 5860 }, { "epoch": 0.7, "learning_rate": 4.480567205947424e-06, "loss": 1.0607, "step": 5861 }, { "epoch": 0.7, "learning_rate": 4.480371048431878e-06, "loss": 1.1538, "step": 5862 }, { "epoch": 0.7, "learning_rate": 4.480174858180633e-06, "loss": 0.6657, "step": 5863 }, { "epoch": 0.7, "learning_rate": 4.479978635196932e-06, "loss": 0.7275, "step": 5864 }, { "epoch": 0.7, "learning_rate": 4.479782379484017e-06, "loss": 0.708, "step": 5865 }, { "epoch": 0.7, "learning_rate": 4.479586091045134e-06, "loss": 0.8842, "step": 5866 }, { "epoch": 0.7, "learning_rate": 4.479389769883527e-06, "loss": 0.8942, "step": 5867 }, { "epoch": 0.7, "learning_rate": 4.4791934160024395e-06, "loss": 1.1504, "step": 5868 }, { "epoch": 0.7, "learning_rate": 4.47899702940512e-06, "loss": 0.7522, "step": 5869 }, { "epoch": 0.7, "learning_rate": 4.478800610094813e-06, "loss": 0.4785, "step": 5870 }, { "epoch": 0.7, "learning_rate": 4.4786041580747666e-06, "loss": 0.6436, "step": 5871 }, { "epoch": 0.7, "learning_rate": 4.478407673348227e-06, "loss": 0.8121, "step": 5872 }, { "epoch": 0.7, "learning_rate": 4.4782111559184426e-06, "loss": 0.8103, "step": 5873 }, { "epoch": 0.7, "learning_rate": 4.478014605788661e-06, "loss": 1.0264, "step": 5874 }, { "epoch": 0.7, "learning_rate": 4.477818022962133e-06, "loss": 1.3119, "step": 5875 }, { "epoch": 0.7, "learning_rate": 4.4776214074421075e-06, "loss": 0.627, "step": 5876 }, { "epoch": 0.7, "learning_rate": 4.477424759231833e-06, "loss": 0.8544, "step": 5877 }, { "epoch": 0.7, "learning_rate": 4.4772280783345624e-06, "loss": 0.7659, "step": 5878 }, { "epoch": 0.7, "learning_rate": 4.4770313647535454e-06, "loss": 0.8982, "step": 5879 }, { "epoch": 0.7, "learning_rate": 4.476834618492034e-06, "loss": 1.0944, "step": 5880 }, { "epoch": 0.7, "learning_rate": 4.4766378395532804e-06, "loss": 1.2886, "step": 5881 }, { "epoch": 0.7, "learning_rate": 4.476441027940537e-06, "loss": 0.4774, "step": 5882 }, { "epoch": 0.7, "learning_rate": 4.476244183657058e-06, "loss": 0.7392, "step": 5883 }, { "epoch": 0.7, "learning_rate": 4.476047306706097e-06, "loss": 0.8735, "step": 5884 }, { "epoch": 0.7, "learning_rate": 4.475850397090909e-06, "loss": 0.8897, "step": 5885 }, { "epoch": 0.7, "learning_rate": 4.475653454814746e-06, "loss": 1.0269, "step": 5886 }, { "epoch": 0.7, "learning_rate": 4.475456479880868e-06, "loss": 1.1836, "step": 5887 }, { "epoch": 0.7, "learning_rate": 4.475259472292527e-06, "loss": 0.6116, "step": 5888 }, { "epoch": 0.7, "learning_rate": 4.475062432052981e-06, "loss": 0.6747, "step": 5889 }, { "epoch": 0.7, "learning_rate": 4.474865359165489e-06, "loss": 0.6691, "step": 5890 }, { "epoch": 0.7, "learning_rate": 4.474668253633305e-06, "loss": 0.6987, "step": 5891 }, { "epoch": 0.7, "learning_rate": 4.4744711154596895e-06, "loss": 0.838, "step": 5892 }, { "epoch": 0.7, "learning_rate": 4.4742739446479014e-06, "loss": 1.173, "step": 5893 }, { "epoch": 0.7, "learning_rate": 4.4740767412011994e-06, "loss": 0.8498, "step": 5894 }, { "epoch": 0.7, "learning_rate": 4.473879505122844e-06, "loss": 0.6482, "step": 5895 }, { "epoch": 0.7, "learning_rate": 4.473682236416093e-06, "loss": 0.7054, "step": 5896 }, { "epoch": 0.7, "learning_rate": 4.47348493508421e-06, "loss": 0.8467, "step": 5897 }, { "epoch": 0.7, "learning_rate": 4.473287601130455e-06, "loss": 1.0428, "step": 5898 }, { "epoch": 0.7, "learning_rate": 4.473090234558091e-06, "loss": 1.0713, "step": 5899 }, { "epoch": 0.7, "learning_rate": 4.472892835370379e-06, "loss": 1.3126, "step": 5900 }, { "epoch": 0.7, "learning_rate": 4.472695403570583e-06, "loss": 0.6761, "step": 5901 }, { "epoch": 0.7, "learning_rate": 4.472497939161967e-06, "loss": 0.6875, "step": 5902 }, { "epoch": 0.7, "learning_rate": 4.472300442147795e-06, "loss": 0.7723, "step": 5903 }, { "epoch": 0.7, "learning_rate": 4.4721029125313305e-06, "loss": 0.861, "step": 5904 }, { "epoch": 0.7, "learning_rate": 4.4719053503158396e-06, "loss": 0.9988, "step": 5905 }, { "epoch": 0.7, "learning_rate": 4.471707755504587e-06, "loss": 1.1387, "step": 5906 }, { "epoch": 0.7, "learning_rate": 4.471510128100841e-06, "loss": 0.7941, "step": 5907 }, { "epoch": 0.7, "learning_rate": 4.471312468107866e-06, "loss": 0.6603, "step": 5908 }, { "epoch": 0.7, "learning_rate": 4.471114775528931e-06, "loss": 0.8224, "step": 5909 }, { "epoch": 0.7, "learning_rate": 4.470917050367304e-06, "loss": 0.8344, "step": 5910 }, { "epoch": 0.7, "learning_rate": 4.4707192926262525e-06, "loss": 0.8964, "step": 5911 }, { "epoch": 0.7, "learning_rate": 4.470521502309045e-06, "loss": 1.1966, "step": 5912 }, { "epoch": 0.7, "learning_rate": 4.470323679418952e-06, "loss": 0.5584, "step": 5913 }, { "epoch": 0.7, "learning_rate": 4.470125823959244e-06, "loss": 0.6518, "step": 5914 }, { "epoch": 0.7, "learning_rate": 4.46992793593319e-06, "loss": 0.7696, "step": 5915 }, { "epoch": 0.7, "learning_rate": 4.469730015344062e-06, "loss": 0.8727, "step": 5916 }, { "epoch": 0.7, "learning_rate": 4.4695320621951325e-06, "loss": 1.1443, "step": 5917 }, { "epoch": 0.7, "learning_rate": 4.469334076489671e-06, "loss": 1.1083, "step": 5918 }, { "epoch": 0.7, "learning_rate": 4.469136058230953e-06, "loss": 0.5648, "step": 5919 }, { "epoch": 0.71, "learning_rate": 4.468938007422251e-06, "loss": 0.5858, "step": 5920 }, { "epoch": 0.71, "learning_rate": 4.468739924066838e-06, "loss": 0.842, "step": 5921 }, { "epoch": 0.71, "learning_rate": 4.468541808167989e-06, "loss": 0.876, "step": 5922 }, { "epoch": 0.71, "learning_rate": 4.4683436597289796e-06, "loss": 0.8596, "step": 5923 }, { "epoch": 0.71, "learning_rate": 4.468145478753083e-06, "loss": 1.0642, "step": 5924 }, { "epoch": 0.71, "learning_rate": 4.467947265243578e-06, "loss": 1.1444, "step": 5925 }, { "epoch": 0.71, "learning_rate": 4.467749019203737e-06, "loss": 0.5911, "step": 5926 }, { "epoch": 0.71, "learning_rate": 4.467550740636842e-06, "loss": 0.7603, "step": 5927 }, { "epoch": 0.71, "learning_rate": 4.467352429546168e-06, "loss": 0.874, "step": 5928 }, { "epoch": 0.71, "learning_rate": 4.467154085934993e-06, "loss": 0.8608, "step": 5929 }, { "epoch": 0.71, "learning_rate": 4.4669557098065955e-06, "loss": 0.9908, "step": 5930 }, { "epoch": 0.71, "learning_rate": 4.466757301164255e-06, "loss": 1.1256, "step": 5931 }, { "epoch": 0.71, "learning_rate": 4.466558860011251e-06, "loss": 0.4684, "step": 5932 }, { "epoch": 0.71, "learning_rate": 4.466360386350865e-06, "loss": 0.6372, "step": 5933 }, { "epoch": 0.71, "learning_rate": 4.466161880186376e-06, "loss": 0.7738, "step": 5934 }, { "epoch": 0.71, "learning_rate": 4.465963341521067e-06, "loss": 0.8322, "step": 5935 }, { "epoch": 0.71, "learning_rate": 4.465764770358218e-06, "loss": 0.9178, "step": 5936 }, { "epoch": 0.71, "learning_rate": 4.465566166701113e-06, "loss": 1.2151, "step": 5937 }, { "epoch": 0.71, "learning_rate": 4.465367530553034e-06, "loss": 0.5711, "step": 5938 }, { "epoch": 0.71, "learning_rate": 4.465168861917265e-06, "loss": 0.5671, "step": 5939 }, { "epoch": 0.71, "learning_rate": 4.46497016079709e-06, "loss": 0.6007, "step": 5940 }, { "epoch": 0.71, "learning_rate": 4.464771427195793e-06, "loss": 0.8097, "step": 5941 }, { "epoch": 0.71, "learning_rate": 4.46457266111666e-06, "loss": 0.784, "step": 5942 }, { "epoch": 0.71, "learning_rate": 4.464373862562976e-06, "loss": 1.0408, "step": 5943 }, { "epoch": 0.71, "learning_rate": 4.464175031538027e-06, "loss": 0.7008, "step": 5944 }, { "epoch": 0.71, "learning_rate": 4.4639761680451e-06, "loss": 0.5755, "step": 5945 }, { "epoch": 0.71, "learning_rate": 4.463777272087482e-06, "loss": 0.8031, "step": 5946 }, { "epoch": 0.71, "learning_rate": 4.463578343668461e-06, "loss": 0.8195, "step": 5947 }, { "epoch": 0.71, "learning_rate": 4.4633793827913256e-06, "loss": 0.729, "step": 5948 }, { "epoch": 0.71, "learning_rate": 4.4631803894593636e-06, "loss": 1.1122, "step": 5949 }, { "epoch": 0.71, "learning_rate": 4.462981363675865e-06, "loss": 1.1887, "step": 5950 }, { "epoch": 0.71, "learning_rate": 4.46278230544412e-06, "loss": 0.5904, "step": 5951 }, { "epoch": 0.71, "learning_rate": 4.462583214767418e-06, "loss": 0.6249, "step": 5952 }, { "epoch": 0.71, "learning_rate": 4.462384091649052e-06, "loss": 0.7458, "step": 5953 }, { "epoch": 0.71, "learning_rate": 4.462184936092313e-06, "loss": 0.8209, "step": 5954 }, { "epoch": 0.71, "learning_rate": 4.4619857481004905e-06, "loss": 0.9534, "step": 5955 }, { "epoch": 0.71, "learning_rate": 4.46178652767688e-06, "loss": 1.0654, "step": 5956 }, { "epoch": 0.71, "learning_rate": 4.461587274824774e-06, "loss": 0.5753, "step": 5957 }, { "epoch": 0.71, "learning_rate": 4.461387989547464e-06, "loss": 0.8311, "step": 5958 }, { "epoch": 0.71, "learning_rate": 4.4611886718482474e-06, "loss": 0.8386, "step": 5959 }, { "epoch": 0.71, "learning_rate": 4.4609893217304175e-06, "loss": 0.741, "step": 5960 }, { "epoch": 0.71, "learning_rate": 4.4607899391972695e-06, "loss": 1.0422, "step": 5961 }, { "epoch": 0.71, "learning_rate": 4.460590524252098e-06, "loss": 1.2213, "step": 5962 }, { "epoch": 0.71, "learning_rate": 4.460391076898202e-06, "loss": 0.4617, "step": 5963 }, { "epoch": 0.71, "learning_rate": 4.460191597138877e-06, "loss": 0.7562, "step": 5964 }, { "epoch": 0.71, "learning_rate": 4.45999208497742e-06, "loss": 0.6824, "step": 5965 }, { "epoch": 0.71, "learning_rate": 4.4597925404171295e-06, "loss": 0.7532, "step": 5966 }, { "epoch": 0.71, "learning_rate": 4.459592963461304e-06, "loss": 1.0375, "step": 5967 }, { "epoch": 0.71, "learning_rate": 4.459393354113242e-06, "loss": 0.9365, "step": 5968 }, { "epoch": 0.71, "learning_rate": 4.459193712376244e-06, "loss": 0.6967, "step": 5969 }, { "epoch": 0.71, "learning_rate": 4.458994038253609e-06, "loss": 0.4398, "step": 5970 }, { "epoch": 0.71, "learning_rate": 4.4587943317486385e-06, "loss": 0.6908, "step": 5971 }, { "epoch": 0.71, "learning_rate": 4.458594592864633e-06, "loss": 0.7787, "step": 5972 }, { "epoch": 0.71, "learning_rate": 4.458394821604894e-06, "loss": 0.8671, "step": 5973 }, { "epoch": 0.71, "learning_rate": 4.458195017972726e-06, "loss": 0.9955, "step": 5974 }, { "epoch": 0.71, "learning_rate": 4.457995181971429e-06, "loss": 1.0948, "step": 5975 }, { "epoch": 0.71, "learning_rate": 4.457795313604307e-06, "loss": 0.7073, "step": 5976 }, { "epoch": 0.71, "learning_rate": 4.457595412874665e-06, "loss": 0.6598, "step": 5977 }, { "epoch": 0.71, "learning_rate": 4.457395479785806e-06, "loss": 0.7747, "step": 5978 }, { "epoch": 0.71, "learning_rate": 4.4571955143410365e-06, "loss": 0.7867, "step": 5979 }, { "epoch": 0.71, "learning_rate": 4.45699551654366e-06, "loss": 0.9673, "step": 5980 }, { "epoch": 0.71, "learning_rate": 4.456795486396983e-06, "loss": 1.0984, "step": 5981 }, { "epoch": 0.71, "learning_rate": 4.456595423904313e-06, "loss": 0.5611, "step": 5982 }, { "epoch": 0.71, "learning_rate": 4.456395329068957e-06, "loss": 0.7314, "step": 5983 }, { "epoch": 0.71, "learning_rate": 4.456195201894221e-06, "loss": 0.8926, "step": 5984 }, { "epoch": 0.71, "learning_rate": 4.455995042383414e-06, "loss": 0.8384, "step": 5985 }, { "epoch": 0.71, "learning_rate": 4.455794850539847e-06, "loss": 0.9442, "step": 5986 }, { "epoch": 0.71, "learning_rate": 4.4555946263668245e-06, "loss": 1.0292, "step": 5987 }, { "epoch": 0.71, "learning_rate": 4.455394369867659e-06, "loss": 0.6186, "step": 5988 }, { "epoch": 0.71, "learning_rate": 4.455194081045662e-06, "loss": 0.6917, "step": 5989 }, { "epoch": 0.71, "learning_rate": 4.454993759904141e-06, "loss": 0.7336, "step": 5990 }, { "epoch": 0.71, "learning_rate": 4.454793406446411e-06, "loss": 0.8164, "step": 5991 }, { "epoch": 0.71, "learning_rate": 4.454593020675782e-06, "loss": 0.9844, "step": 5992 }, { "epoch": 0.71, "learning_rate": 4.454392602595564e-06, "loss": 1.0938, "step": 5993 }, { "epoch": 0.71, "learning_rate": 4.454192152209074e-06, "loss": 0.782, "step": 5994 }, { "epoch": 0.71, "learning_rate": 4.453991669519624e-06, "loss": 0.6353, "step": 5995 }, { "epoch": 0.71, "learning_rate": 4.453791154530527e-06, "loss": 0.7118, "step": 5996 }, { "epoch": 0.71, "learning_rate": 4.453590607245099e-06, "loss": 0.8065, "step": 5997 }, { "epoch": 0.71, "learning_rate": 4.453390027666653e-06, "loss": 0.8646, "step": 5998 }, { "epoch": 0.71, "learning_rate": 4.4531894157985075e-06, "loss": 1.0301, "step": 5999 }, { "epoch": 0.71, "learning_rate": 4.452988771643976e-06, "loss": 1.0984, "step": 6000 }, { "epoch": 0.71, "learning_rate": 4.452788095206377e-06, "loss": 0.588, "step": 6001 }, { "epoch": 0.71, "learning_rate": 4.452587386489026e-06, "loss": 0.6574, "step": 6002 }, { "epoch": 0.71, "learning_rate": 4.452386645495243e-06, "loss": 0.8624, "step": 6003 }, { "epoch": 0.72, "learning_rate": 4.452185872228344e-06, "loss": 0.8325, "step": 6004 }, { "epoch": 0.72, "learning_rate": 4.451985066691649e-06, "loss": 0.9456, "step": 6005 }, { "epoch": 0.72, "learning_rate": 4.451784228888476e-06, "loss": 1.0186, "step": 6006 }, { "epoch": 0.72, "learning_rate": 4.451583358822147e-06, "loss": 0.471, "step": 6007 }, { "epoch": 0.72, "learning_rate": 4.451382456495982e-06, "loss": 0.5967, "step": 6008 }, { "epoch": 0.72, "learning_rate": 4.4511815219133e-06, "loss": 0.7813, "step": 6009 }, { "epoch": 0.72, "learning_rate": 4.450980555077424e-06, "loss": 0.903, "step": 6010 }, { "epoch": 0.72, "learning_rate": 4.450779555991676e-06, "loss": 1.0537, "step": 6011 }, { "epoch": 0.72, "learning_rate": 4.450578524659378e-06, "loss": 1.0592, "step": 6012 }, { "epoch": 0.72, "learning_rate": 4.450377461083854e-06, "loss": 0.5471, "step": 6013 }, { "epoch": 0.72, "learning_rate": 4.450176365268426e-06, "loss": 0.5448, "step": 6014 }, { "epoch": 0.72, "learning_rate": 4.44997523721642e-06, "loss": 0.7571, "step": 6015 }, { "epoch": 0.72, "learning_rate": 4.449774076931159e-06, "loss": 0.7025, "step": 6016 }, { "epoch": 0.72, "learning_rate": 4.4495728844159694e-06, "loss": 0.9861, "step": 6017 }, { "epoch": 0.72, "learning_rate": 4.4493716596741755e-06, "loss": 1.1349, "step": 6018 }, { "epoch": 0.72, "learning_rate": 4.449170402709105e-06, "loss": 0.7231, "step": 6019 }, { "epoch": 0.72, "learning_rate": 4.448969113524084e-06, "loss": 0.6561, "step": 6020 }, { "epoch": 0.72, "learning_rate": 4.44876779212244e-06, "loss": 0.7997, "step": 6021 }, { "epoch": 0.72, "learning_rate": 4.448566438507502e-06, "loss": 0.7986, "step": 6022 }, { "epoch": 0.72, "learning_rate": 4.448365052682596e-06, "loss": 0.8766, "step": 6023 }, { "epoch": 0.72, "learning_rate": 4.4481636346510524e-06, "loss": 1.1808, "step": 6024 }, { "epoch": 0.72, "learning_rate": 4.447962184416201e-06, "loss": 1.1873, "step": 6025 }, { "epoch": 0.72, "learning_rate": 4.447760701981371e-06, "loss": 0.4316, "step": 6026 }, { "epoch": 0.72, "learning_rate": 4.447559187349894e-06, "loss": 0.6285, "step": 6027 }, { "epoch": 0.72, "learning_rate": 4.447357640525098e-06, "loss": 0.64, "step": 6028 }, { "epoch": 0.72, "learning_rate": 4.447156061510319e-06, "loss": 0.8893, "step": 6029 }, { "epoch": 0.72, "learning_rate": 4.4469544503088855e-06, "loss": 0.9604, "step": 6030 }, { "epoch": 0.72, "learning_rate": 4.446752806924133e-06, "loss": 1.3772, "step": 6031 }, { "epoch": 0.72, "learning_rate": 4.4465511313593925e-06, "loss": 0.4875, "step": 6032 }, { "epoch": 0.72, "learning_rate": 4.446349423617998e-06, "loss": 0.8408, "step": 6033 }, { "epoch": 0.72, "learning_rate": 4.446147683703285e-06, "loss": 0.8603, "step": 6034 }, { "epoch": 0.72, "learning_rate": 4.445945911618587e-06, "loss": 0.8053, "step": 6035 }, { "epoch": 0.72, "learning_rate": 4.445744107367239e-06, "loss": 1.0132, "step": 6036 }, { "epoch": 0.72, "learning_rate": 4.445542270952579e-06, "loss": 1.0734, "step": 6037 }, { "epoch": 0.72, "learning_rate": 4.445340402377942e-06, "loss": 0.5731, "step": 6038 }, { "epoch": 0.72, "learning_rate": 4.4451385016466646e-06, "loss": 0.6476, "step": 6039 }, { "epoch": 0.72, "learning_rate": 4.444936568762084e-06, "loss": 0.6508, "step": 6040 }, { "epoch": 0.72, "learning_rate": 4.444734603727539e-06, "loss": 0.7905, "step": 6041 }, { "epoch": 0.72, "learning_rate": 4.444532606546368e-06, "loss": 0.9659, "step": 6042 }, { "epoch": 0.72, "learning_rate": 4.44433057722191e-06, "loss": 1.027, "step": 6043 }, { "epoch": 0.72, "learning_rate": 4.444128515757505e-06, "loss": 0.7306, "step": 6044 }, { "epoch": 0.72, "learning_rate": 4.443926422156491e-06, "loss": 0.6822, "step": 6045 }, { "epoch": 0.72, "learning_rate": 4.443724296422211e-06, "loss": 0.7328, "step": 6046 }, { "epoch": 0.72, "learning_rate": 4.443522138558005e-06, "loss": 0.8017, "step": 6047 }, { "epoch": 0.72, "learning_rate": 4.443319948567215e-06, "loss": 0.848, "step": 6048 }, { "epoch": 0.72, "learning_rate": 4.443117726453183e-06, "loss": 1.0734, "step": 6049 }, { "epoch": 0.72, "learning_rate": 4.4429154722192524e-06, "loss": 1.2198, "step": 6050 }, { "epoch": 0.72, "learning_rate": 4.442713185868767e-06, "loss": 0.5589, "step": 6051 }, { "epoch": 0.72, "learning_rate": 4.442510867405068e-06, "loss": 0.7151, "step": 6052 }, { "epoch": 0.72, "learning_rate": 4.4423085168315015e-06, "loss": 0.6755, "step": 6053 }, { "epoch": 0.72, "learning_rate": 4.442106134151413e-06, "loss": 0.8324, "step": 6054 }, { "epoch": 0.72, "learning_rate": 4.441903719368147e-06, "loss": 1.1058, "step": 6055 }, { "epoch": 0.72, "learning_rate": 4.441701272485049e-06, "loss": 1.235, "step": 6056 }, { "epoch": 0.72, "learning_rate": 4.441498793505467e-06, "loss": 0.5536, "step": 6057 }, { "epoch": 0.72, "learning_rate": 4.441296282432746e-06, "loss": 0.6647, "step": 6058 }, { "epoch": 0.72, "learning_rate": 4.441093739270235e-06, "loss": 0.8114, "step": 6059 }, { "epoch": 0.72, "learning_rate": 4.4408911640212815e-06, "loss": 0.8969, "step": 6060 }, { "epoch": 0.72, "learning_rate": 4.440688556689234e-06, "loss": 1.011, "step": 6061 }, { "epoch": 0.72, "learning_rate": 4.440485917277443e-06, "loss": 1.189, "step": 6062 }, { "epoch": 0.72, "learning_rate": 4.440283245789255e-06, "loss": 0.6333, "step": 6063 }, { "epoch": 0.72, "learning_rate": 4.440080542228024e-06, "loss": 0.7005, "step": 6064 }, { "epoch": 0.72, "learning_rate": 4.439877806597098e-06, "loss": 0.7918, "step": 6065 }, { "epoch": 0.72, "learning_rate": 4.43967503889983e-06, "loss": 0.8429, "step": 6066 }, { "epoch": 0.72, "learning_rate": 4.439472239139569e-06, "loss": 1.0099, "step": 6067 }, { "epoch": 0.72, "learning_rate": 4.439269407319671e-06, "loss": 1.1944, "step": 6068 }, { "epoch": 0.72, "learning_rate": 4.439066543443485e-06, "loss": 0.7673, "step": 6069 }, { "epoch": 0.72, "learning_rate": 4.438863647514368e-06, "loss": 0.6261, "step": 6070 }, { "epoch": 0.72, "learning_rate": 4.438660719535672e-06, "loss": 0.5462, "step": 6071 }, { "epoch": 0.72, "learning_rate": 4.438457759510751e-06, "loss": 0.813, "step": 6072 }, { "epoch": 0.72, "learning_rate": 4.438254767442961e-06, "loss": 0.8597, "step": 6073 }, { "epoch": 0.72, "learning_rate": 4.438051743335657e-06, "loss": 1.157, "step": 6074 }, { "epoch": 0.72, "learning_rate": 4.437848687192195e-06, "loss": 1.1863, "step": 6075 }, { "epoch": 0.72, "learning_rate": 4.437645599015933e-06, "loss": 0.6254, "step": 6076 }, { "epoch": 0.72, "learning_rate": 4.4374424788102245e-06, "loss": 0.7642, "step": 6077 }, { "epoch": 0.72, "learning_rate": 4.4372393265784305e-06, "loss": 0.7366, "step": 6078 }, { "epoch": 0.72, "learning_rate": 4.437036142323907e-06, "loss": 0.86, "step": 6079 }, { "epoch": 0.72, "learning_rate": 4.4368329260500146e-06, "loss": 0.929, "step": 6080 }, { "epoch": 0.72, "learning_rate": 4.436629677760111e-06, "loss": 1.191, "step": 6081 }, { "epoch": 0.72, "learning_rate": 4.436426397457556e-06, "loss": 0.6014, "step": 6082 }, { "epoch": 0.72, "learning_rate": 4.436223085145711e-06, "loss": 0.6363, "step": 6083 }, { "epoch": 0.72, "learning_rate": 4.4360197408279345e-06, "loss": 0.8282, "step": 6084 }, { "epoch": 0.72, "learning_rate": 4.43581636450759e-06, "loss": 0.677, "step": 6085 }, { "epoch": 0.72, "learning_rate": 4.435612956188039e-06, "loss": 0.903, "step": 6086 }, { "epoch": 0.72, "learning_rate": 4.435409515872643e-06, "loss": 0.9637, "step": 6087 }, { "epoch": 0.73, "learning_rate": 4.4352060435647655e-06, "loss": 0.568, "step": 6088 }, { "epoch": 0.73, "learning_rate": 4.435002539267769e-06, "loss": 0.7031, "step": 6089 }, { "epoch": 0.73, "learning_rate": 4.4347990029850196e-06, "loss": 0.711, "step": 6090 }, { "epoch": 0.73, "learning_rate": 4.43459543471988e-06, "loss": 0.8687, "step": 6091 }, { "epoch": 0.73, "learning_rate": 4.434391834475714e-06, "loss": 0.9743, "step": 6092 }, { "epoch": 0.73, "learning_rate": 4.4341882022558895e-06, "loss": 1.2622, "step": 6093 }, { "epoch": 0.73, "learning_rate": 4.433984538063772e-06, "loss": 0.7772, "step": 6094 }, { "epoch": 0.73, "learning_rate": 4.433780841902728e-06, "loss": 0.5721, "step": 6095 }, { "epoch": 0.73, "learning_rate": 4.433577113776124e-06, "loss": 0.8015, "step": 6096 }, { "epoch": 0.73, "learning_rate": 4.433373353687328e-06, "loss": 0.7162, "step": 6097 }, { "epoch": 0.73, "learning_rate": 4.433169561639708e-06, "loss": 0.847, "step": 6098 }, { "epoch": 0.73, "learning_rate": 4.4329657376366345e-06, "loss": 0.9431, "step": 6099 }, { "epoch": 0.73, "learning_rate": 4.432761881681473e-06, "loss": 1.2186, "step": 6100 }, { "epoch": 0.73, "learning_rate": 4.432557993777597e-06, "loss": 0.4924, "step": 6101 }, { "epoch": 0.73, "learning_rate": 4.432354073928374e-06, "loss": 0.6655, "step": 6102 }, { "epoch": 0.73, "learning_rate": 4.432150122137178e-06, "loss": 0.7657, "step": 6103 }, { "epoch": 0.73, "learning_rate": 4.431946138407377e-06, "loss": 0.9335, "step": 6104 }, { "epoch": 0.73, "learning_rate": 4.431742122742344e-06, "loss": 1.0482, "step": 6105 }, { "epoch": 0.73, "learning_rate": 4.431538075145452e-06, "loss": 1.1923, "step": 6106 }, { "epoch": 0.73, "learning_rate": 4.431333995620074e-06, "loss": 0.5036, "step": 6107 }, { "epoch": 0.73, "learning_rate": 4.431129884169581e-06, "loss": 0.6675, "step": 6108 }, { "epoch": 0.73, "learning_rate": 4.430925740797352e-06, "loss": 0.7679, "step": 6109 }, { "epoch": 0.73, "learning_rate": 4.430721565506756e-06, "loss": 0.8586, "step": 6110 }, { "epoch": 0.73, "learning_rate": 4.430517358301172e-06, "loss": 1.0141, "step": 6111 }, { "epoch": 0.73, "learning_rate": 4.4303131191839734e-06, "loss": 1.1874, "step": 6112 }, { "epoch": 0.73, "learning_rate": 4.430108848158537e-06, "loss": 0.7382, "step": 6113 }, { "epoch": 0.73, "learning_rate": 4.429904545228239e-06, "loss": 0.4928, "step": 6114 }, { "epoch": 0.73, "learning_rate": 4.429700210396458e-06, "loss": 0.7753, "step": 6115 }, { "epoch": 0.73, "learning_rate": 4.42949584366657e-06, "loss": 0.7894, "step": 6116 }, { "epoch": 0.73, "learning_rate": 4.429291445041954e-06, "loss": 0.9386, "step": 6117 }, { "epoch": 0.73, "learning_rate": 4.429087014525988e-06, "loss": 1.0718, "step": 6118 }, { "epoch": 0.73, "learning_rate": 4.428882552122053e-06, "loss": 0.8113, "step": 6119 }, { "epoch": 0.73, "learning_rate": 4.428678057833526e-06, "loss": 0.6591, "step": 6120 }, { "epoch": 0.73, "learning_rate": 4.428473531663789e-06, "loss": 0.7642, "step": 6121 }, { "epoch": 0.73, "learning_rate": 4.4282689736162245e-06, "loss": 0.8186, "step": 6122 }, { "epoch": 0.73, "learning_rate": 4.4280643836942105e-06, "loss": 0.903, "step": 6123 }, { "epoch": 0.73, "learning_rate": 4.427859761901131e-06, "loss": 0.9794, "step": 6124 }, { "epoch": 0.73, "learning_rate": 4.427655108240367e-06, "loss": 1.3392, "step": 6125 }, { "epoch": 0.73, "learning_rate": 4.4274504227153036e-06, "loss": 0.6485, "step": 6126 }, { "epoch": 0.73, "learning_rate": 4.427245705329323e-06, "loss": 0.7639, "step": 6127 }, { "epoch": 0.73, "learning_rate": 4.427040956085809e-06, "loss": 0.7297, "step": 6128 }, { "epoch": 0.73, "learning_rate": 4.426836174988145e-06, "loss": 0.8127, "step": 6129 }, { "epoch": 0.73, "learning_rate": 4.426631362039718e-06, "loss": 0.9178, "step": 6130 }, { "epoch": 0.73, "learning_rate": 4.426426517243913e-06, "loss": 1.1876, "step": 6131 }, { "epoch": 0.73, "learning_rate": 4.426221640604117e-06, "loss": 0.5543, "step": 6132 }, { "epoch": 0.73, "learning_rate": 4.426016732123715e-06, "loss": 0.6818, "step": 6133 }, { "epoch": 0.73, "learning_rate": 4.425811791806094e-06, "loss": 0.6327, "step": 6134 }, { "epoch": 0.73, "learning_rate": 4.425606819654644e-06, "loss": 0.8889, "step": 6135 }, { "epoch": 0.73, "learning_rate": 4.42540181567275e-06, "loss": 0.9316, "step": 6136 }, { "epoch": 0.73, "learning_rate": 4.425196779863804e-06, "loss": 1.1558, "step": 6137 }, { "epoch": 0.73, "learning_rate": 4.424991712231192e-06, "loss": 0.5494, "step": 6138 }, { "epoch": 0.73, "learning_rate": 4.424786612778307e-06, "loss": 0.5952, "step": 6139 }, { "epoch": 0.73, "learning_rate": 4.424581481508537e-06, "loss": 0.8031, "step": 6140 }, { "epoch": 0.73, "learning_rate": 4.424376318425274e-06, "loss": 0.6582, "step": 6141 }, { "epoch": 0.73, "learning_rate": 4.42417112353191e-06, "loss": 0.8374, "step": 6142 }, { "epoch": 0.73, "learning_rate": 4.423965896831834e-06, "loss": 1.1, "step": 6143 }, { "epoch": 0.73, "learning_rate": 4.423760638328442e-06, "loss": 0.8786, "step": 6144 }, { "epoch": 0.73, "learning_rate": 4.4235553480251235e-06, "loss": 0.6491, "step": 6145 }, { "epoch": 0.73, "learning_rate": 4.423350025925275e-06, "loss": 0.6884, "step": 6146 }, { "epoch": 0.73, "learning_rate": 4.4231446720322875e-06, "loss": 0.7536, "step": 6147 }, { "epoch": 0.73, "learning_rate": 4.422939286349559e-06, "loss": 0.9675, "step": 6148 }, { "epoch": 0.73, "learning_rate": 4.422733868880482e-06, "loss": 0.8228, "step": 6149 }, { "epoch": 0.73, "learning_rate": 4.422528419628454e-06, "loss": 1.173, "step": 6150 }, { "epoch": 0.73, "learning_rate": 4.422322938596868e-06, "loss": 0.6599, "step": 6151 }, { "epoch": 0.73, "learning_rate": 4.422117425789123e-06, "loss": 0.6508, "step": 6152 }, { "epoch": 0.73, "learning_rate": 4.421911881208617e-06, "loss": 0.7592, "step": 6153 }, { "epoch": 0.73, "learning_rate": 4.421706304858745e-06, "loss": 0.8228, "step": 6154 }, { "epoch": 0.73, "learning_rate": 4.421500696742907e-06, "loss": 1.1271, "step": 6155 }, { "epoch": 0.73, "learning_rate": 4.421295056864501e-06, "loss": 1.1472, "step": 6156 }, { "epoch": 0.73, "learning_rate": 4.421089385226927e-06, "loss": 0.6527, "step": 6157 }, { "epoch": 0.73, "learning_rate": 4.4208836818335845e-06, "loss": 0.7872, "step": 6158 }, { "epoch": 0.73, "learning_rate": 4.420677946687872e-06, "loss": 0.8299, "step": 6159 }, { "epoch": 0.73, "learning_rate": 4.420472179793193e-06, "loss": 0.7939, "step": 6160 }, { "epoch": 0.73, "learning_rate": 4.420266381152948e-06, "loss": 0.9689, "step": 6161 }, { "epoch": 0.73, "learning_rate": 4.420060550770539e-06, "loss": 1.1222, "step": 6162 }, { "epoch": 0.73, "learning_rate": 4.4198546886493666e-06, "loss": 0.8007, "step": 6163 }, { "epoch": 0.73, "learning_rate": 4.419648794792835e-06, "loss": 0.5829, "step": 6164 }, { "epoch": 0.73, "learning_rate": 4.419442869204349e-06, "loss": 0.7608, "step": 6165 }, { "epoch": 0.73, "learning_rate": 4.419236911887311e-06, "loss": 0.7864, "step": 6166 }, { "epoch": 0.73, "learning_rate": 4.419030922845126e-06, "loss": 0.9631, "step": 6167 }, { "epoch": 0.73, "learning_rate": 4.418824902081198e-06, "loss": 1.0319, "step": 6168 }, { "epoch": 0.73, "learning_rate": 4.418618849598933e-06, "loss": 0.9979, "step": 6169 }, { "epoch": 0.73, "learning_rate": 4.418412765401739e-06, "loss": 0.6603, "step": 6170 }, { "epoch": 0.73, "learning_rate": 4.41820664949302e-06, "loss": 0.6997, "step": 6171 }, { "epoch": 0.74, "learning_rate": 4.418000501876184e-06, "loss": 0.8242, "step": 6172 }, { "epoch": 0.74, "learning_rate": 4.4177943225546395e-06, "loss": 1.0148, "step": 6173 }, { "epoch": 0.74, "learning_rate": 4.417588111531794e-06, "loss": 1.0897, "step": 6174 }, { "epoch": 0.74, "learning_rate": 4.417381868811056e-06, "loss": 1.2049, "step": 6175 }, { "epoch": 0.74, "learning_rate": 4.417175594395834e-06, "loss": 0.5726, "step": 6176 }, { "epoch": 0.74, "learning_rate": 4.4169692882895385e-06, "loss": 0.843, "step": 6177 }, { "epoch": 0.74, "learning_rate": 4.4167629504955815e-06, "loss": 0.722, "step": 6178 }, { "epoch": 0.74, "learning_rate": 4.41655658101737e-06, "loss": 0.8644, "step": 6179 }, { "epoch": 0.74, "learning_rate": 4.416350179858318e-06, "loss": 0.881, "step": 6180 }, { "epoch": 0.74, "learning_rate": 4.4161437470218386e-06, "loss": 1.4104, "step": 6181 }, { "epoch": 0.74, "learning_rate": 4.41593728251134e-06, "loss": 0.4957, "step": 6182 }, { "epoch": 0.74, "learning_rate": 4.415730786330238e-06, "loss": 0.7516, "step": 6183 }, { "epoch": 0.74, "learning_rate": 4.415524258481946e-06, "loss": 0.8123, "step": 6184 }, { "epoch": 0.74, "learning_rate": 4.415317698969877e-06, "loss": 0.736, "step": 6185 }, { "epoch": 0.74, "learning_rate": 4.415111107797445e-06, "loss": 0.9565, "step": 6186 }, { "epoch": 0.74, "learning_rate": 4.414904484968067e-06, "loss": 1.1348, "step": 6187 }, { "epoch": 0.74, "learning_rate": 4.414697830485156e-06, "loss": 0.6982, "step": 6188 }, { "epoch": 0.74, "learning_rate": 4.41449114435213e-06, "loss": 0.6093, "step": 6189 }, { "epoch": 0.74, "learning_rate": 4.414284426572405e-06, "loss": 0.7701, "step": 6190 }, { "epoch": 0.74, "learning_rate": 4.414077677149396e-06, "loss": 0.7616, "step": 6191 }, { "epoch": 0.74, "learning_rate": 4.413870896086524e-06, "loss": 0.9691, "step": 6192 }, { "epoch": 0.74, "learning_rate": 4.413664083387205e-06, "loss": 1.0577, "step": 6193 }, { "epoch": 0.74, "learning_rate": 4.413457239054858e-06, "loss": 0.9167, "step": 6194 }, { "epoch": 0.74, "learning_rate": 4.413250363092903e-06, "loss": 0.5737, "step": 6195 }, { "epoch": 0.74, "learning_rate": 4.4130434555047585e-06, "loss": 0.6086, "step": 6196 }, { "epoch": 0.74, "learning_rate": 4.4128365162938444e-06, "loss": 0.7396, "step": 6197 }, { "epoch": 0.74, "learning_rate": 4.4126295454635835e-06, "loss": 0.8574, "step": 6198 }, { "epoch": 0.74, "learning_rate": 4.412422543017395e-06, "loss": 1.0401, "step": 6199 }, { "epoch": 0.74, "learning_rate": 4.412215508958701e-06, "loss": 1.0977, "step": 6200 }, { "epoch": 0.74, "learning_rate": 4.412008443290926e-06, "loss": 0.6279, "step": 6201 }, { "epoch": 0.74, "learning_rate": 4.411801346017489e-06, "loss": 0.6428, "step": 6202 }, { "epoch": 0.74, "learning_rate": 4.411594217141816e-06, "loss": 0.7553, "step": 6203 }, { "epoch": 0.74, "learning_rate": 4.411387056667331e-06, "loss": 0.8997, "step": 6204 }, { "epoch": 0.74, "learning_rate": 4.411179864597457e-06, "loss": 0.96, "step": 6205 }, { "epoch": 0.74, "learning_rate": 4.410972640935619e-06, "loss": 1.2073, "step": 6206 }, { "epoch": 0.74, "learning_rate": 4.410765385685243e-06, "loss": 0.7106, "step": 6207 }, { "epoch": 0.74, "learning_rate": 4.410558098849755e-06, "loss": 0.7576, "step": 6208 }, { "epoch": 0.74, "learning_rate": 4.410350780432582e-06, "loss": 0.765, "step": 6209 }, { "epoch": 0.74, "learning_rate": 4.410143430437149e-06, "loss": 0.8785, "step": 6210 }, { "epoch": 0.74, "learning_rate": 4.409936048866886e-06, "loss": 1.0903, "step": 6211 }, { "epoch": 0.74, "learning_rate": 4.409728635725219e-06, "loss": 1.0982, "step": 6212 }, { "epoch": 0.74, "learning_rate": 4.409521191015578e-06, "loss": 0.7711, "step": 6213 }, { "epoch": 0.74, "learning_rate": 4.409313714741391e-06, "loss": 0.6137, "step": 6214 }, { "epoch": 0.74, "learning_rate": 4.409106206906089e-06, "loss": 0.6307, "step": 6215 }, { "epoch": 0.74, "learning_rate": 4.4088986675131e-06, "loss": 0.8274, "step": 6216 }, { "epoch": 0.74, "learning_rate": 4.408691096565857e-06, "loss": 0.9642, "step": 6217 }, { "epoch": 0.74, "learning_rate": 4.408483494067789e-06, "loss": 1.1622, "step": 6218 }, { "epoch": 0.74, "learning_rate": 4.408275860022329e-06, "loss": 0.6013, "step": 6219 }, { "epoch": 0.74, "learning_rate": 4.40806819443291e-06, "loss": 0.5896, "step": 6220 }, { "epoch": 0.74, "learning_rate": 4.4078604973029624e-06, "loss": 0.751, "step": 6221 }, { "epoch": 0.74, "learning_rate": 4.407652768635921e-06, "loss": 0.7229, "step": 6222 }, { "epoch": 0.74, "learning_rate": 4.407445008435219e-06, "loss": 0.8145, "step": 6223 }, { "epoch": 0.74, "learning_rate": 4.407237216704291e-06, "loss": 0.9908, "step": 6224 }, { "epoch": 0.74, "learning_rate": 4.407029393446572e-06, "loss": 0.948, "step": 6225 }, { "epoch": 0.74, "learning_rate": 4.4068215386654976e-06, "loss": 0.6332, "step": 6226 }, { "epoch": 0.74, "learning_rate": 4.406613652364502e-06, "loss": 0.769, "step": 6227 }, { "epoch": 0.74, "learning_rate": 4.406405734547024e-06, "loss": 0.7092, "step": 6228 }, { "epoch": 0.74, "learning_rate": 4.406197785216498e-06, "loss": 0.7726, "step": 6229 }, { "epoch": 0.74, "learning_rate": 4.4059898043763635e-06, "loss": 0.8994, "step": 6230 }, { "epoch": 0.74, "learning_rate": 4.405781792030058e-06, "loss": 1.077, "step": 6231 }, { "epoch": 0.74, "learning_rate": 4.405573748181019e-06, "loss": 0.661, "step": 6232 }, { "epoch": 0.74, "learning_rate": 4.405365672832687e-06, "loss": 0.6546, "step": 6233 }, { "epoch": 0.74, "learning_rate": 4.405157565988499e-06, "loss": 0.6556, "step": 6234 }, { "epoch": 0.74, "learning_rate": 4.4049494276518975e-06, "loss": 0.8041, "step": 6235 }, { "epoch": 0.74, "learning_rate": 4.404741257826322e-06, "loss": 1.0032, "step": 6236 }, { "epoch": 0.74, "learning_rate": 4.404533056515213e-06, "loss": 1.1442, "step": 6237 }, { "epoch": 0.74, "learning_rate": 4.404324823722014e-06, "loss": 0.8081, "step": 6238 }, { "epoch": 0.74, "learning_rate": 4.4041165594501645e-06, "loss": 0.6435, "step": 6239 }, { "epoch": 0.74, "learning_rate": 4.40390826370311e-06, "loss": 0.7396, "step": 6240 }, { "epoch": 0.74, "learning_rate": 4.403699936484291e-06, "loss": 0.8409, "step": 6241 }, { "epoch": 0.74, "learning_rate": 4.403491577797152e-06, "loss": 1.0379, "step": 6242 }, { "epoch": 0.74, "learning_rate": 4.403283187645138e-06, "loss": 1.1581, "step": 6243 }, { "epoch": 0.74, "learning_rate": 4.403074766031694e-06, "loss": 0.6801, "step": 6244 }, { "epoch": 0.74, "learning_rate": 4.402866312960263e-06, "loss": 0.6075, "step": 6245 }, { "epoch": 0.74, "learning_rate": 4.402657828434293e-06, "loss": 0.7372, "step": 6246 }, { "epoch": 0.74, "learning_rate": 4.402449312457231e-06, "loss": 0.7812, "step": 6247 }, { "epoch": 0.74, "learning_rate": 4.40224076503252e-06, "loss": 0.8751, "step": 6248 }, { "epoch": 0.74, "learning_rate": 4.40203218616361e-06, "loss": 0.9441, "step": 6249 }, { "epoch": 0.74, "learning_rate": 4.4018235758539486e-06, "loss": 1.131, "step": 6250 }, { "epoch": 0.74, "learning_rate": 4.401614934106984e-06, "loss": 0.5453, "step": 6251 }, { "epoch": 0.74, "learning_rate": 4.4014062609261655e-06, "loss": 0.7309, "step": 6252 }, { "epoch": 0.74, "learning_rate": 4.401197556314941e-06, "loss": 0.7855, "step": 6253 }, { "epoch": 0.74, "learning_rate": 4.400988820276762e-06, "loss": 0.8327, "step": 6254 }, { "epoch": 0.74, "learning_rate": 4.400780052815078e-06, "loss": 0.9512, "step": 6255 }, { "epoch": 0.75, "learning_rate": 4.40057125393334e-06, "loss": 1.2062, "step": 6256 }, { "epoch": 0.75, "learning_rate": 4.4003624236349994e-06, "loss": 0.4991, "step": 6257 }, { "epoch": 0.75, "learning_rate": 4.40015356192351e-06, "loss": 0.806, "step": 6258 }, { "epoch": 0.75, "learning_rate": 4.399944668802322e-06, "loss": 0.8994, "step": 6259 }, { "epoch": 0.75, "learning_rate": 4.399735744274889e-06, "loss": 0.8503, "step": 6260 }, { "epoch": 0.75, "learning_rate": 4.399526788344664e-06, "loss": 0.9467, "step": 6261 }, { "epoch": 0.75, "learning_rate": 4.399317801015103e-06, "loss": 1.1881, "step": 6262 }, { "epoch": 0.75, "learning_rate": 4.399108782289658e-06, "loss": 0.6168, "step": 6263 }, { "epoch": 0.75, "learning_rate": 4.398899732171787e-06, "loss": 0.6732, "step": 6264 }, { "epoch": 0.75, "learning_rate": 4.398690650664944e-06, "loss": 0.7314, "step": 6265 }, { "epoch": 0.75, "learning_rate": 4.398481537772584e-06, "loss": 0.9152, "step": 6266 }, { "epoch": 0.75, "learning_rate": 4.398272393498166e-06, "loss": 0.9788, "step": 6267 }, { "epoch": 0.75, "learning_rate": 4.398063217845146e-06, "loss": 1.19, "step": 6268 }, { "epoch": 0.75, "learning_rate": 4.3978540108169824e-06, "loss": 0.4566, "step": 6269 }, { "epoch": 0.75, "learning_rate": 4.397644772417131e-06, "loss": 0.6423, "step": 6270 }, { "epoch": 0.75, "learning_rate": 4.397435502649053e-06, "loss": 0.8105, "step": 6271 }, { "epoch": 0.75, "learning_rate": 4.397226201516207e-06, "loss": 0.7133, "step": 6272 }, { "epoch": 0.75, "learning_rate": 4.397016869022053e-06, "loss": 0.9279, "step": 6273 }, { "epoch": 0.75, "learning_rate": 4.396807505170051e-06, "loss": 1.0872, "step": 6274 }, { "epoch": 0.75, "learning_rate": 4.396598109963661e-06, "loss": 1.1143, "step": 6275 }, { "epoch": 0.75, "learning_rate": 4.396388683406346e-06, "loss": 0.6959, "step": 6276 }, { "epoch": 0.75, "learning_rate": 4.396179225501567e-06, "loss": 0.6558, "step": 6277 }, { "epoch": 0.75, "learning_rate": 4.395969736252785e-06, "loss": 0.7188, "step": 6278 }, { "epoch": 0.75, "learning_rate": 4.3957602156634645e-06, "loss": 0.9503, "step": 6279 }, { "epoch": 0.75, "learning_rate": 4.39555066373707e-06, "loss": 1.1643, "step": 6280 }, { "epoch": 0.75, "learning_rate": 4.395341080477063e-06, "loss": 1.0365, "step": 6281 }, { "epoch": 0.75, "learning_rate": 4.395131465886909e-06, "loss": 0.6583, "step": 6282 }, { "epoch": 0.75, "learning_rate": 4.394921819970073e-06, "loss": 0.7398, "step": 6283 }, { "epoch": 0.75, "learning_rate": 4.39471214273002e-06, "loss": 0.7343, "step": 6284 }, { "epoch": 0.75, "learning_rate": 4.3945024341702165e-06, "loss": 0.7908, "step": 6285 }, { "epoch": 0.75, "learning_rate": 4.394292694294129e-06, "loss": 0.992, "step": 6286 }, { "epoch": 0.75, "learning_rate": 4.394082923105224e-06, "loss": 1.3034, "step": 6287 }, { "epoch": 0.75, "learning_rate": 4.39387312060697e-06, "loss": 0.6005, "step": 6288 }, { "epoch": 0.75, "learning_rate": 4.393663286802833e-06, "loss": 0.5542, "step": 6289 }, { "epoch": 0.75, "learning_rate": 4.393453421696285e-06, "loss": 0.7861, "step": 6290 }, { "epoch": 0.75, "learning_rate": 4.3932435252907914e-06, "loss": 0.7277, "step": 6291 }, { "epoch": 0.75, "learning_rate": 4.393033597589824e-06, "loss": 0.944, "step": 6292 }, { "epoch": 0.75, "learning_rate": 4.392823638596852e-06, "loss": 1.1843, "step": 6293 }, { "epoch": 0.75, "learning_rate": 4.392613648315347e-06, "loss": 0.8492, "step": 6294 }, { "epoch": 0.75, "learning_rate": 4.392403626748779e-06, "loss": 0.6437, "step": 6295 }, { "epoch": 0.75, "learning_rate": 4.392193573900621e-06, "loss": 0.7703, "step": 6296 }, { "epoch": 0.75, "learning_rate": 4.391983489774344e-06, "loss": 0.6228, "step": 6297 }, { "epoch": 0.75, "learning_rate": 4.3917733743734215e-06, "loss": 0.8534, "step": 6298 }, { "epoch": 0.75, "learning_rate": 4.391563227701326e-06, "loss": 1.062, "step": 6299 }, { "epoch": 0.75, "learning_rate": 4.391353049761532e-06, "loss": 1.0986, "step": 6300 }, { "epoch": 0.75, "learning_rate": 4.391142840557513e-06, "loss": 0.6104, "step": 6301 }, { "epoch": 0.75, "learning_rate": 4.3909326000927444e-06, "loss": 0.8114, "step": 6302 }, { "epoch": 0.75, "learning_rate": 4.390722328370701e-06, "loss": 0.7231, "step": 6303 }, { "epoch": 0.75, "learning_rate": 4.390512025394858e-06, "loss": 0.8407, "step": 6304 }, { "epoch": 0.75, "learning_rate": 4.390301691168695e-06, "loss": 0.9892, "step": 6305 }, { "epoch": 0.75, "learning_rate": 4.3900913256956846e-06, "loss": 1.1021, "step": 6306 }, { "epoch": 0.75, "learning_rate": 4.389880928979307e-06, "loss": 0.5677, "step": 6307 }, { "epoch": 0.75, "learning_rate": 4.3896705010230386e-06, "loss": 0.6422, "step": 6308 }, { "epoch": 0.75, "learning_rate": 4.389460041830358e-06, "loss": 0.7829, "step": 6309 }, { "epoch": 0.75, "learning_rate": 4.389249551404745e-06, "loss": 0.8477, "step": 6310 }, { "epoch": 0.75, "learning_rate": 4.389039029749677e-06, "loss": 0.9521, "step": 6311 }, { "epoch": 0.75, "learning_rate": 4.388828476868637e-06, "loss": 1.1179, "step": 6312 }, { "epoch": 0.75, "learning_rate": 4.388617892765103e-06, "loss": 0.6678, "step": 6313 }, { "epoch": 0.75, "learning_rate": 4.388407277442557e-06, "loss": 0.6198, "step": 6314 }, { "epoch": 0.75, "learning_rate": 4.38819663090448e-06, "loss": 0.7858, "step": 6315 }, { "epoch": 0.75, "learning_rate": 4.387985953154354e-06, "loss": 0.7786, "step": 6316 }, { "epoch": 0.75, "learning_rate": 4.387775244195662e-06, "loss": 0.9894, "step": 6317 }, { "epoch": 0.75, "learning_rate": 4.387564504031887e-06, "loss": 1.1506, "step": 6318 }, { "epoch": 0.75, "learning_rate": 4.3873537326665124e-06, "loss": 0.9097, "step": 6319 }, { "epoch": 0.75, "learning_rate": 4.3871429301030225e-06, "loss": 0.6272, "step": 6320 }, { "epoch": 0.75, "learning_rate": 4.3869320963449015e-06, "loss": 0.7184, "step": 6321 }, { "epoch": 0.75, "learning_rate": 4.386721231395634e-06, "loss": 0.7718, "step": 6322 }, { "epoch": 0.75, "learning_rate": 4.386510335258707e-06, "loss": 0.8643, "step": 6323 }, { "epoch": 0.75, "learning_rate": 4.386299407937605e-06, "loss": 1.0891, "step": 6324 }, { "epoch": 0.75, "learning_rate": 4.386088449435817e-06, "loss": 1.1608, "step": 6325 }, { "epoch": 0.75, "learning_rate": 4.385877459756828e-06, "loss": 0.5229, "step": 6326 }, { "epoch": 0.75, "learning_rate": 4.385666438904125e-06, "loss": 0.604, "step": 6327 }, { "epoch": 0.75, "learning_rate": 4.385455386881199e-06, "loss": 0.7542, "step": 6328 }, { "epoch": 0.75, "learning_rate": 4.3852443036915374e-06, "loss": 0.7372, "step": 6329 }, { "epoch": 0.75, "learning_rate": 4.385033189338628e-06, "loss": 1.0051, "step": 6330 }, { "epoch": 0.75, "learning_rate": 4.3848220438259625e-06, "loss": 1.0344, "step": 6331 }, { "epoch": 0.75, "learning_rate": 4.384610867157031e-06, "loss": 0.4884, "step": 6332 }, { "epoch": 0.75, "learning_rate": 4.384399659335323e-06, "loss": 0.6236, "step": 6333 }, { "epoch": 0.75, "learning_rate": 4.384188420364331e-06, "loss": 0.5714, "step": 6334 }, { "epoch": 0.75, "learning_rate": 4.383977150247546e-06, "loss": 0.7104, "step": 6335 }, { "epoch": 0.75, "learning_rate": 4.38376584898846e-06, "loss": 0.8658, "step": 6336 }, { "epoch": 0.75, "learning_rate": 4.383554516590568e-06, "loss": 1.1566, "step": 6337 }, { "epoch": 0.75, "learning_rate": 4.383343153057361e-06, "loss": 0.7996, "step": 6338 }, { "epoch": 0.75, "learning_rate": 4.383131758392334e-06, "loss": 0.7223, "step": 6339 }, { "epoch": 0.76, "learning_rate": 4.38292033259898e-06, "loss": 0.7127, "step": 6340 }, { "epoch": 0.76, "learning_rate": 4.382708875680796e-06, "loss": 0.7018, "step": 6341 }, { "epoch": 0.76, "learning_rate": 4.382497387641276e-06, "loss": 0.9263, "step": 6342 }, { "epoch": 0.76, "learning_rate": 4.382285868483915e-06, "loss": 1.148, "step": 6343 }, { "epoch": 0.76, "learning_rate": 4.382074318212212e-06, "loss": 0.5654, "step": 6344 }, { "epoch": 0.76, "learning_rate": 4.381862736829663e-06, "loss": 0.6021, "step": 6345 }, { "epoch": 0.76, "learning_rate": 4.3816511243397645e-06, "loss": 0.8099, "step": 6346 }, { "epoch": 0.76, "learning_rate": 4.381439480746015e-06, "loss": 0.7984, "step": 6347 }, { "epoch": 0.76, "learning_rate": 4.381227806051913e-06, "loss": 0.887, "step": 6348 }, { "epoch": 0.76, "learning_rate": 4.381016100260958e-06, "loss": 0.9567, "step": 6349 }, { "epoch": 0.76, "learning_rate": 4.380804363376647e-06, "loss": 1.1255, "step": 6350 }, { "epoch": 0.76, "learning_rate": 4.380592595402484e-06, "loss": 0.4876, "step": 6351 }, { "epoch": 0.76, "learning_rate": 4.380380796341967e-06, "loss": 0.6002, "step": 6352 }, { "epoch": 0.76, "learning_rate": 4.380168966198598e-06, "loss": 0.8516, "step": 6353 }, { "epoch": 0.76, "learning_rate": 4.379957104975878e-06, "loss": 0.8785, "step": 6354 }, { "epoch": 0.76, "learning_rate": 4.37974521267731e-06, "loss": 1.1107, "step": 6355 }, { "epoch": 0.76, "learning_rate": 4.379533289306395e-06, "loss": 1.1561, "step": 6356 }, { "epoch": 0.76, "learning_rate": 4.379321334866638e-06, "loss": 0.6539, "step": 6357 }, { "epoch": 0.76, "learning_rate": 4.37910934936154e-06, "loss": 0.6956, "step": 6358 }, { "epoch": 0.76, "learning_rate": 4.378897332794608e-06, "loss": 0.8457, "step": 6359 }, { "epoch": 0.76, "learning_rate": 4.378685285169345e-06, "loss": 0.924, "step": 6360 }, { "epoch": 0.76, "learning_rate": 4.378473206489257e-06, "loss": 1.0164, "step": 6361 }, { "epoch": 0.76, "learning_rate": 4.378261096757849e-06, "loss": 1.0994, "step": 6362 }, { "epoch": 0.76, "learning_rate": 4.378048955978628e-06, "loss": 0.5548, "step": 6363 }, { "epoch": 0.76, "learning_rate": 4.3778367841551e-06, "loss": 0.5164, "step": 6364 }, { "epoch": 0.76, "learning_rate": 4.3776245812907715e-06, "loss": 0.7911, "step": 6365 }, { "epoch": 0.76, "learning_rate": 4.377412347389151e-06, "loss": 0.7641, "step": 6366 }, { "epoch": 0.76, "learning_rate": 4.377200082453748e-06, "loss": 0.9522, "step": 6367 }, { "epoch": 0.76, "learning_rate": 4.37698778648807e-06, "loss": 1.076, "step": 6368 }, { "epoch": 0.76, "learning_rate": 4.376775459495626e-06, "loss": 0.6611, "step": 6369 }, { "epoch": 0.76, "learning_rate": 4.376563101479927e-06, "loss": 0.7798, "step": 6370 }, { "epoch": 0.76, "learning_rate": 4.3763507124444815e-06, "loss": 0.7842, "step": 6371 }, { "epoch": 0.76, "learning_rate": 4.376138292392801e-06, "loss": 0.7546, "step": 6372 }, { "epoch": 0.76, "learning_rate": 4.375925841328398e-06, "loss": 0.9444, "step": 6373 }, { "epoch": 0.76, "learning_rate": 4.375713359254783e-06, "loss": 1.0101, "step": 6374 }, { "epoch": 0.76, "learning_rate": 4.3755008461754686e-06, "loss": 1.0884, "step": 6375 }, { "epoch": 0.76, "learning_rate": 4.3752883020939685e-06, "loss": 0.5256, "step": 6376 }, { "epoch": 0.76, "learning_rate": 4.375075727013795e-06, "loss": 0.661, "step": 6377 }, { "epoch": 0.76, "learning_rate": 4.374863120938463e-06, "loss": 0.7281, "step": 6378 }, { "epoch": 0.76, "learning_rate": 4.374650483871486e-06, "loss": 0.8354, "step": 6379 }, { "epoch": 0.76, "learning_rate": 4.3744378158163785e-06, "loss": 0.9932, "step": 6380 }, { "epoch": 0.76, "learning_rate": 4.374225116776657e-06, "loss": 1.0593, "step": 6381 }, { "epoch": 0.76, "learning_rate": 4.374012386755837e-06, "loss": 0.5963, "step": 6382 }, { "epoch": 0.76, "learning_rate": 4.373799625757436e-06, "loss": 0.6492, "step": 6383 }, { "epoch": 0.76, "learning_rate": 4.373586833784968e-06, "loss": 0.5894, "step": 6384 }, { "epoch": 0.76, "learning_rate": 4.373374010841954e-06, "loss": 0.8876, "step": 6385 }, { "epoch": 0.76, "learning_rate": 4.37316115693191e-06, "loss": 1.0902, "step": 6386 }, { "epoch": 0.76, "learning_rate": 4.3729482720583545e-06, "loss": 1.0737, "step": 6387 }, { "epoch": 0.76, "learning_rate": 4.372735356224807e-06, "loss": 0.5818, "step": 6388 }, { "epoch": 0.76, "learning_rate": 4.372522409434788e-06, "loss": 0.581, "step": 6389 }, { "epoch": 0.76, "learning_rate": 4.3723094316918155e-06, "loss": 0.7495, "step": 6390 }, { "epoch": 0.76, "learning_rate": 4.372096422999411e-06, "loss": 0.8274, "step": 6391 }, { "epoch": 0.76, "learning_rate": 4.371883383361096e-06, "loss": 0.9178, "step": 6392 }, { "epoch": 0.76, "learning_rate": 4.371670312780391e-06, "loss": 1.0352, "step": 6393 }, { "epoch": 0.76, "learning_rate": 4.371457211260819e-06, "loss": 0.8777, "step": 6394 }, { "epoch": 0.76, "learning_rate": 4.3712440788059025e-06, "loss": 0.6465, "step": 6395 }, { "epoch": 0.76, "learning_rate": 4.371030915419164e-06, "loss": 0.9385, "step": 6396 }, { "epoch": 0.76, "learning_rate": 4.370817721104127e-06, "loss": 0.8626, "step": 6397 }, { "epoch": 0.76, "learning_rate": 4.3706044958643175e-06, "loss": 0.9451, "step": 6398 }, { "epoch": 0.76, "learning_rate": 4.370391239703257e-06, "loss": 1.1382, "step": 6399 }, { "epoch": 0.76, "learning_rate": 4.370177952624474e-06, "loss": 1.2597, "step": 6400 }, { "epoch": 0.76, "learning_rate": 4.369964634631492e-06, "loss": 0.5149, "step": 6401 }, { "epoch": 0.76, "learning_rate": 4.369751285727837e-06, "loss": 0.6609, "step": 6402 }, { "epoch": 0.76, "learning_rate": 4.369537905917038e-06, "loss": 0.7886, "step": 6403 }, { "epoch": 0.76, "learning_rate": 4.369324495202618e-06, "loss": 0.9289, "step": 6404 }, { "epoch": 0.76, "learning_rate": 4.369111053588109e-06, "loss": 0.9767, "step": 6405 }, { "epoch": 0.76, "learning_rate": 4.368897581077037e-06, "loss": 1.1016, "step": 6406 }, { "epoch": 0.76, "learning_rate": 4.368684077672932e-06, "loss": 0.6676, "step": 6407 }, { "epoch": 0.76, "learning_rate": 4.368470543379322e-06, "loss": 0.7009, "step": 6408 }, { "epoch": 0.76, "learning_rate": 4.368256978199736e-06, "loss": 0.6681, "step": 6409 }, { "epoch": 0.76, "learning_rate": 4.368043382137707e-06, "loss": 0.7394, "step": 6410 }, { "epoch": 0.76, "learning_rate": 4.367829755196763e-06, "loss": 0.9286, "step": 6411 }, { "epoch": 0.76, "learning_rate": 4.367616097380437e-06, "loss": 1.2596, "step": 6412 }, { "epoch": 0.76, "learning_rate": 4.36740240869226e-06, "loss": 0.7239, "step": 6413 }, { "epoch": 0.76, "learning_rate": 4.367188689135764e-06, "loss": 0.5818, "step": 6414 }, { "epoch": 0.76, "learning_rate": 4.366974938714483e-06, "loss": 0.8641, "step": 6415 }, { "epoch": 0.76, "learning_rate": 4.366761157431949e-06, "loss": 0.7534, "step": 6416 }, { "epoch": 0.76, "learning_rate": 4.3665473452916975e-06, "loss": 0.9117, "step": 6417 }, { "epoch": 0.76, "learning_rate": 4.366333502297261e-06, "loss": 1.1547, "step": 6418 }, { "epoch": 0.76, "learning_rate": 4.366119628452174e-06, "loss": 0.7136, "step": 6419 }, { "epoch": 0.76, "learning_rate": 4.3659057237599754e-06, "loss": 0.5963, "step": 6420 }, { "epoch": 0.76, "learning_rate": 4.365691788224197e-06, "loss": 0.9187, "step": 6421 }, { "epoch": 0.76, "learning_rate": 4.365477821848377e-06, "loss": 0.7991, "step": 6422 }, { "epoch": 0.76, "learning_rate": 4.365263824636053e-06, "loss": 0.8319, "step": 6423 }, { "epoch": 0.77, "learning_rate": 4.36504979659076e-06, "loss": 1.0699, "step": 6424 }, { "epoch": 0.77, "learning_rate": 4.364835737716038e-06, "loss": 1.0542, "step": 6425 }, { "epoch": 0.77, "learning_rate": 4.364621648015426e-06, "loss": 0.4779, "step": 6426 }, { "epoch": 0.77, "learning_rate": 4.36440752749246e-06, "loss": 0.6569, "step": 6427 }, { "epoch": 0.77, "learning_rate": 4.364193376150681e-06, "loss": 0.7695, "step": 6428 }, { "epoch": 0.77, "learning_rate": 4.36397919399363e-06, "loss": 0.9731, "step": 6429 }, { "epoch": 0.77, "learning_rate": 4.363764981024846e-06, "loss": 1.0718, "step": 6430 }, { "epoch": 0.77, "learning_rate": 4.363550737247871e-06, "loss": 1.1064, "step": 6431 }, { "epoch": 0.77, "learning_rate": 4.363336462666246e-06, "loss": 0.5294, "step": 6432 }, { "epoch": 0.77, "learning_rate": 4.363122157283513e-06, "loss": 0.6295, "step": 6433 }, { "epoch": 0.77, "learning_rate": 4.362907821103214e-06, "loss": 0.8047, "step": 6434 }, { "epoch": 0.77, "learning_rate": 4.362693454128892e-06, "loss": 0.8368, "step": 6435 }, { "epoch": 0.77, "learning_rate": 4.362479056364092e-06, "loss": 0.9213, "step": 6436 }, { "epoch": 0.77, "learning_rate": 4.362264627812356e-06, "loss": 1.11, "step": 6437 }, { "epoch": 0.77, "learning_rate": 4.36205016847723e-06, "loss": 0.5586, "step": 6438 }, { "epoch": 0.77, "learning_rate": 4.3618356783622586e-06, "loss": 0.6834, "step": 6439 }, { "epoch": 0.77, "learning_rate": 4.361621157470986e-06, "loss": 0.6269, "step": 6440 }, { "epoch": 0.77, "learning_rate": 4.361406605806961e-06, "loss": 0.8362, "step": 6441 }, { "epoch": 0.77, "learning_rate": 4.361192023373728e-06, "loss": 1.0289, "step": 6442 }, { "epoch": 0.77, "learning_rate": 4.360977410174835e-06, "loss": 1.223, "step": 6443 }, { "epoch": 0.77, "learning_rate": 4.360762766213829e-06, "loss": 0.69, "step": 6444 }, { "epoch": 0.77, "learning_rate": 4.360548091494257e-06, "loss": 0.5259, "step": 6445 }, { "epoch": 0.77, "learning_rate": 4.360333386019671e-06, "loss": 0.6692, "step": 6446 }, { "epoch": 0.77, "learning_rate": 4.360118649793617e-06, "loss": 0.8213, "step": 6447 }, { "epoch": 0.77, "learning_rate": 4.359903882819646e-06, "loss": 0.9233, "step": 6448 }, { "epoch": 0.77, "learning_rate": 4.359689085101308e-06, "loss": 1.1087, "step": 6449 }, { "epoch": 0.77, "learning_rate": 4.359474256642152e-06, "loss": 1.1199, "step": 6450 }, { "epoch": 0.77, "learning_rate": 4.359259397445732e-06, "loss": 0.5846, "step": 6451 }, { "epoch": 0.77, "learning_rate": 4.359044507515597e-06, "loss": 0.8175, "step": 6452 }, { "epoch": 0.77, "learning_rate": 4.358829586855302e-06, "loss": 0.6059, "step": 6453 }, { "epoch": 0.77, "learning_rate": 4.3586146354683964e-06, "loss": 0.6732, "step": 6454 }, { "epoch": 0.77, "learning_rate": 4.3583996533584355e-06, "loss": 0.9354, "step": 6455 }, { "epoch": 0.77, "learning_rate": 4.358184640528972e-06, "loss": 1.0676, "step": 6456 }, { "epoch": 0.77, "learning_rate": 4.357969596983561e-06, "loss": 0.6638, "step": 6457 }, { "epoch": 0.77, "learning_rate": 4.3577545227257565e-06, "loss": 0.8428, "step": 6458 }, { "epoch": 0.77, "learning_rate": 4.357539417759114e-06, "loss": 0.7956, "step": 6459 }, { "epoch": 0.77, "learning_rate": 4.357324282087189e-06, "loss": 0.7856, "step": 6460 }, { "epoch": 0.77, "learning_rate": 4.357109115713538e-06, "loss": 0.9109, "step": 6461 }, { "epoch": 0.77, "learning_rate": 4.356893918641717e-06, "loss": 1.2835, "step": 6462 }, { "epoch": 0.77, "learning_rate": 4.356678690875284e-06, "loss": 0.6923, "step": 6463 }, { "epoch": 0.77, "learning_rate": 4.3564634324177965e-06, "loss": 0.6282, "step": 6464 }, { "epoch": 0.77, "learning_rate": 4.356248143272812e-06, "loss": 0.7487, "step": 6465 }, { "epoch": 0.77, "learning_rate": 4.356032823443891e-06, "loss": 0.787, "step": 6466 }, { "epoch": 0.77, "learning_rate": 4.355817472934591e-06, "loss": 0.9251, "step": 6467 }, { "epoch": 0.77, "learning_rate": 4.3556020917484725e-06, "loss": 1.0471, "step": 6468 }, { "epoch": 0.77, "learning_rate": 4.355386679889096e-06, "loss": 0.7316, "step": 6469 }, { "epoch": 0.77, "learning_rate": 4.355171237360022e-06, "loss": 0.7641, "step": 6470 }, { "epoch": 0.77, "learning_rate": 4.354955764164811e-06, "loss": 0.7046, "step": 6471 }, { "epoch": 0.77, "learning_rate": 4.3547402603070265e-06, "loss": 0.765, "step": 6472 }, { "epoch": 0.77, "learning_rate": 4.35452472579023e-06, "loss": 0.8159, "step": 6473 }, { "epoch": 0.77, "learning_rate": 4.3543091606179835e-06, "loss": 1.0561, "step": 6474 }, { "epoch": 0.77, "learning_rate": 4.3540935647938515e-06, "loss": 0.9709, "step": 6475 }, { "epoch": 0.77, "learning_rate": 4.353877938321397e-06, "loss": 0.7343, "step": 6476 }, { "epoch": 0.77, "learning_rate": 4.353662281204184e-06, "loss": 0.7061, "step": 6477 }, { "epoch": 0.77, "learning_rate": 4.3534465934457795e-06, "loss": 0.7335, "step": 6478 }, { "epoch": 0.77, "learning_rate": 4.353230875049747e-06, "loss": 0.9147, "step": 6479 }, { "epoch": 0.77, "learning_rate": 4.353015126019651e-06, "loss": 1.0192, "step": 6480 }, { "epoch": 0.77, "learning_rate": 4.352799346359061e-06, "loss": 1.2386, "step": 6481 }, { "epoch": 0.77, "learning_rate": 4.352583536071542e-06, "loss": 0.6838, "step": 6482 }, { "epoch": 0.77, "learning_rate": 4.352367695160661e-06, "loss": 0.557, "step": 6483 }, { "epoch": 0.77, "learning_rate": 4.352151823629987e-06, "loss": 0.7297, "step": 6484 }, { "epoch": 0.77, "learning_rate": 4.351935921483088e-06, "loss": 0.9091, "step": 6485 }, { "epoch": 0.77, "learning_rate": 4.351719988723533e-06, "loss": 1.0331, "step": 6486 }, { "epoch": 0.77, "learning_rate": 4.351504025354892e-06, "loss": 1.1446, "step": 6487 }, { "epoch": 0.77, "learning_rate": 4.3512880313807325e-06, "loss": 0.7885, "step": 6488 }, { "epoch": 0.77, "learning_rate": 4.351072006804627e-06, "loss": 0.5414, "step": 6489 }, { "epoch": 0.77, "learning_rate": 4.350855951630147e-06, "loss": 0.6742, "step": 6490 }, { "epoch": 0.77, "learning_rate": 4.350639865860862e-06, "loss": 0.7298, "step": 6491 }, { "epoch": 0.77, "learning_rate": 4.3504237495003444e-06, "loss": 0.8806, "step": 6492 }, { "epoch": 0.77, "learning_rate": 4.350207602552168e-06, "loss": 1.1708, "step": 6493 }, { "epoch": 0.77, "learning_rate": 4.349991425019904e-06, "loss": 0.7906, "step": 6494 }, { "epoch": 0.77, "learning_rate": 4.349775216907126e-06, "loss": 0.7493, "step": 6495 }, { "epoch": 0.77, "learning_rate": 4.3495589782174095e-06, "loss": 0.6281, "step": 6496 }, { "epoch": 0.77, "learning_rate": 4.349342708954327e-06, "loss": 0.8061, "step": 6497 }, { "epoch": 0.77, "learning_rate": 4.349126409121454e-06, "loss": 0.8427, "step": 6498 }, { "epoch": 0.77, "learning_rate": 4.348910078722368e-06, "loss": 1.0181, "step": 6499 }, { "epoch": 0.77, "learning_rate": 4.348693717760642e-06, "loss": 1.0753, "step": 6500 }, { "epoch": 0.77, "learning_rate": 4.3484773262398536e-06, "loss": 0.6356, "step": 6501 }, { "epoch": 0.77, "learning_rate": 4.348260904163581e-06, "loss": 0.8126, "step": 6502 }, { "epoch": 0.77, "learning_rate": 4.3480444515353995e-06, "loss": 0.8039, "step": 6503 }, { "epoch": 0.77, "learning_rate": 4.347827968358889e-06, "loss": 0.8398, "step": 6504 }, { "epoch": 0.77, "learning_rate": 4.347611454637626e-06, "loss": 0.9506, "step": 6505 }, { "epoch": 0.77, "learning_rate": 4.347394910375191e-06, "loss": 1.2196, "step": 6506 }, { "epoch": 0.77, "learning_rate": 4.347178335575164e-06, "loss": 0.5335, "step": 6507 }, { "epoch": 0.78, "learning_rate": 4.346961730241124e-06, "loss": 0.8273, "step": 6508 }, { "epoch": 0.78, "learning_rate": 4.346745094376651e-06, "loss": 0.7741, "step": 6509 }, { "epoch": 0.78, "learning_rate": 4.346528427985327e-06, "loss": 0.8237, "step": 6510 }, { "epoch": 0.78, "learning_rate": 4.346311731070733e-06, "loss": 1.0814, "step": 6511 }, { "epoch": 0.78, "learning_rate": 4.34609500363645e-06, "loss": 1.1732, "step": 6512 }, { "epoch": 0.78, "learning_rate": 4.345878245686064e-06, "loss": 0.5655, "step": 6513 }, { "epoch": 0.78, "learning_rate": 4.345661457223153e-06, "loss": 0.6566, "step": 6514 }, { "epoch": 0.78, "learning_rate": 4.345444638251306e-06, "loss": 0.6234, "step": 6515 }, { "epoch": 0.78, "learning_rate": 4.3452277887741025e-06, "loss": 0.7474, "step": 6516 }, { "epoch": 0.78, "learning_rate": 4.345010908795129e-06, "loss": 0.9557, "step": 6517 }, { "epoch": 0.78, "learning_rate": 4.344793998317971e-06, "loss": 1.1913, "step": 6518 }, { "epoch": 0.78, "learning_rate": 4.344577057346213e-06, "loss": 0.6703, "step": 6519 }, { "epoch": 0.78, "learning_rate": 4.344360085883442e-06, "loss": 0.5446, "step": 6520 }, { "epoch": 0.78, "learning_rate": 4.344143083933244e-06, "loss": 0.5302, "step": 6521 }, { "epoch": 0.78, "learning_rate": 4.343926051499205e-06, "loss": 0.7901, "step": 6522 }, { "epoch": 0.78, "learning_rate": 4.343708988584915e-06, "loss": 0.8785, "step": 6523 }, { "epoch": 0.78, "learning_rate": 4.3434918951939595e-06, "loss": 1.0147, "step": 6524 }, { "epoch": 0.78, "learning_rate": 4.343274771329929e-06, "loss": 1.2278, "step": 6525 }, { "epoch": 0.78, "learning_rate": 4.343057616996413e-06, "loss": 0.4672, "step": 6526 }, { "epoch": 0.78, "learning_rate": 4.342840432196997e-06, "loss": 0.8086, "step": 6527 }, { "epoch": 0.78, "learning_rate": 4.342623216935277e-06, "loss": 0.7527, "step": 6528 }, { "epoch": 0.78, "learning_rate": 4.34240597121484e-06, "loss": 0.7704, "step": 6529 }, { "epoch": 0.78, "learning_rate": 4.342188695039276e-06, "loss": 0.9358, "step": 6530 }, { "epoch": 0.78, "learning_rate": 4.34197138841218e-06, "loss": 1.1628, "step": 6531 }, { "epoch": 0.78, "learning_rate": 4.341754051337141e-06, "loss": 0.5277, "step": 6532 }, { "epoch": 0.78, "learning_rate": 4.3415366838177544e-06, "loss": 0.6547, "step": 6533 }, { "epoch": 0.78, "learning_rate": 4.3413192858576105e-06, "loss": 0.6866, "step": 6534 }, { "epoch": 0.78, "learning_rate": 4.341101857460305e-06, "loss": 0.853, "step": 6535 }, { "epoch": 0.78, "learning_rate": 4.340884398629432e-06, "loss": 0.9987, "step": 6536 }, { "epoch": 0.78, "learning_rate": 4.340666909368585e-06, "loss": 1.0687, "step": 6537 }, { "epoch": 0.78, "learning_rate": 4.340449389681359e-06, "loss": 0.6367, "step": 6538 }, { "epoch": 0.78, "learning_rate": 4.34023183957135e-06, "loss": 0.5271, "step": 6539 }, { "epoch": 0.78, "learning_rate": 4.340014259042154e-06, "loss": 0.8287, "step": 6540 }, { "epoch": 0.78, "learning_rate": 4.33979664809737e-06, "loss": 0.7567, "step": 6541 }, { "epoch": 0.78, "learning_rate": 4.339579006740592e-06, "loss": 1.0399, "step": 6542 }, { "epoch": 0.78, "learning_rate": 4.339361334975419e-06, "loss": 1.1031, "step": 6543 }, { "epoch": 0.78, "learning_rate": 4.339143632805447e-06, "loss": 0.7037, "step": 6544 }, { "epoch": 0.78, "learning_rate": 4.338925900234278e-06, "loss": 0.5958, "step": 6545 }, { "epoch": 0.78, "learning_rate": 4.338708137265509e-06, "loss": 0.7241, "step": 6546 }, { "epoch": 0.78, "learning_rate": 4.33849034390274e-06, "loss": 0.744, "step": 6547 }, { "epoch": 0.78, "learning_rate": 4.338272520149572e-06, "loss": 0.9385, "step": 6548 }, { "epoch": 0.78, "learning_rate": 4.338054666009604e-06, "loss": 1.046, "step": 6549 }, { "epoch": 0.78, "learning_rate": 4.33783678148644e-06, "loss": 1.0684, "step": 6550 }, { "epoch": 0.78, "learning_rate": 4.337618866583678e-06, "loss": 0.5213, "step": 6551 }, { "epoch": 0.78, "learning_rate": 4.337400921304923e-06, "loss": 0.5636, "step": 6552 }, { "epoch": 0.78, "learning_rate": 4.3371829456537765e-06, "loss": 0.648, "step": 6553 }, { "epoch": 0.78, "learning_rate": 4.336964939633841e-06, "loss": 0.8714, "step": 6554 }, { "epoch": 0.78, "learning_rate": 4.336746903248722e-06, "loss": 1.0087, "step": 6555 }, { "epoch": 0.78, "learning_rate": 4.336528836502022e-06, "loss": 1.2073, "step": 6556 }, { "epoch": 0.78, "learning_rate": 4.336310739397346e-06, "loss": 0.6251, "step": 6557 }, { "epoch": 0.78, "learning_rate": 4.336092611938299e-06, "loss": 0.605, "step": 6558 }, { "epoch": 0.78, "learning_rate": 4.335874454128488e-06, "loss": 0.6989, "step": 6559 }, { "epoch": 0.78, "learning_rate": 4.335656265971519e-06, "loss": 0.8112, "step": 6560 }, { "epoch": 0.78, "learning_rate": 4.335438047470996e-06, "loss": 1.0061, "step": 6561 }, { "epoch": 0.78, "learning_rate": 4.335219798630529e-06, "loss": 1.1349, "step": 6562 }, { "epoch": 0.78, "learning_rate": 4.335001519453724e-06, "loss": 0.5464, "step": 6563 }, { "epoch": 0.78, "learning_rate": 4.334783209944191e-06, "loss": 0.6457, "step": 6564 }, { "epoch": 0.78, "learning_rate": 4.334564870105537e-06, "loss": 0.6899, "step": 6565 }, { "epoch": 0.78, "learning_rate": 4.334346499941372e-06, "loss": 0.7962, "step": 6566 }, { "epoch": 0.78, "learning_rate": 4.334128099455306e-06, "loss": 0.8926, "step": 6567 }, { "epoch": 0.78, "learning_rate": 4.333909668650947e-06, "loss": 1.1114, "step": 6568 }, { "epoch": 0.78, "learning_rate": 4.333691207531908e-06, "loss": 0.847, "step": 6569 }, { "epoch": 0.78, "learning_rate": 4.3334727161018005e-06, "loss": 0.6174, "step": 6570 }, { "epoch": 0.78, "learning_rate": 4.3332541943642345e-06, "loss": 0.6714, "step": 6571 }, { "epoch": 0.78, "learning_rate": 4.333035642322822e-06, "loss": 0.7583, "step": 6572 }, { "epoch": 0.78, "learning_rate": 4.332817059981178e-06, "loss": 1.0056, "step": 6573 }, { "epoch": 0.78, "learning_rate": 4.332598447342913e-06, "loss": 1.2078, "step": 6574 }, { "epoch": 0.78, "learning_rate": 4.332379804411643e-06, "loss": 1.1771, "step": 6575 }, { "epoch": 0.78, "learning_rate": 4.33216113119098e-06, "loss": 0.4382, "step": 6576 }, { "epoch": 0.78, "learning_rate": 4.3319424276845405e-06, "loss": 0.7545, "step": 6577 }, { "epoch": 0.78, "learning_rate": 4.331723693895939e-06, "loss": 0.7935, "step": 6578 }, { "epoch": 0.78, "learning_rate": 4.331504929828791e-06, "loss": 0.9697, "step": 6579 }, { "epoch": 0.78, "learning_rate": 4.331286135486712e-06, "loss": 0.929, "step": 6580 }, { "epoch": 0.78, "learning_rate": 4.331067310873321e-06, "loss": 1.1226, "step": 6581 }, { "epoch": 0.78, "learning_rate": 4.330848455992232e-06, "loss": 0.6977, "step": 6582 }, { "epoch": 0.78, "learning_rate": 4.330629570847065e-06, "loss": 0.6487, "step": 6583 }, { "epoch": 0.78, "learning_rate": 4.330410655441438e-06, "loss": 0.7159, "step": 6584 }, { "epoch": 0.78, "learning_rate": 4.3301917097789696e-06, "loss": 0.7676, "step": 6585 }, { "epoch": 0.78, "learning_rate": 4.329972733863278e-06, "loss": 1.04, "step": 6586 }, { "epoch": 0.78, "learning_rate": 4.329753727697984e-06, "loss": 1.1771, "step": 6587 }, { "epoch": 0.78, "learning_rate": 4.329534691286707e-06, "loss": 0.6765, "step": 6588 }, { "epoch": 0.78, "learning_rate": 4.329315624633068e-06, "loss": 0.715, "step": 6589 }, { "epoch": 0.78, "learning_rate": 4.329096527740688e-06, "loss": 0.7323, "step": 6590 }, { "epoch": 0.78, "learning_rate": 4.32887740061319e-06, "loss": 0.7207, "step": 6591 }, { "epoch": 0.79, "learning_rate": 4.328658243254195e-06, "loss": 0.8806, "step": 6592 }, { "epoch": 0.79, "learning_rate": 4.328439055667325e-06, "loss": 1.1792, "step": 6593 }, { "epoch": 0.79, "learning_rate": 4.328219837856204e-06, "loss": 0.835, "step": 6594 }, { "epoch": 0.79, "learning_rate": 4.3280005898244555e-06, "loss": 0.608, "step": 6595 }, { "epoch": 0.79, "learning_rate": 4.327781311575705e-06, "loss": 0.8155, "step": 6596 }, { "epoch": 0.79, "learning_rate": 4.3275620031135755e-06, "loss": 0.7636, "step": 6597 }, { "epoch": 0.79, "learning_rate": 4.327342664441692e-06, "loss": 0.957, "step": 6598 }, { "epoch": 0.79, "learning_rate": 4.327123295563682e-06, "loss": 1.2081, "step": 6599 }, { "epoch": 0.79, "learning_rate": 4.32690389648317e-06, "loss": 1.1689, "step": 6600 }, { "epoch": 0.79, "learning_rate": 4.3266844672037835e-06, "loss": 0.6346, "step": 6601 }, { "epoch": 0.79, "learning_rate": 4.3264650077291495e-06, "loss": 0.786, "step": 6602 }, { "epoch": 0.79, "learning_rate": 4.326245518062896e-06, "loss": 0.7689, "step": 6603 }, { "epoch": 0.79, "learning_rate": 4.32602599820865e-06, "loss": 0.9451, "step": 6604 }, { "epoch": 0.79, "learning_rate": 4.325806448170041e-06, "loss": 1.0413, "step": 6605 }, { "epoch": 0.79, "learning_rate": 4.325586867950699e-06, "loss": 1.0952, "step": 6606 }, { "epoch": 0.79, "learning_rate": 4.325367257554253e-06, "loss": 0.5824, "step": 6607 }, { "epoch": 0.79, "learning_rate": 4.325147616984332e-06, "loss": 0.5863, "step": 6608 }, { "epoch": 0.79, "learning_rate": 4.324927946244569e-06, "loss": 0.7886, "step": 6609 }, { "epoch": 0.79, "learning_rate": 4.324708245338593e-06, "loss": 0.7893, "step": 6610 }, { "epoch": 0.79, "learning_rate": 4.324488514270037e-06, "loss": 0.8929, "step": 6611 }, { "epoch": 0.79, "learning_rate": 4.324268753042533e-06, "loss": 1.1255, "step": 6612 }, { "epoch": 0.79, "learning_rate": 4.324048961659713e-06, "loss": 0.6528, "step": 6613 }, { "epoch": 0.79, "learning_rate": 4.323829140125211e-06, "loss": 0.7248, "step": 6614 }, { "epoch": 0.79, "learning_rate": 4.32360928844266e-06, "loss": 0.7822, "step": 6615 }, { "epoch": 0.79, "learning_rate": 4.3233894066156946e-06, "loss": 0.7956, "step": 6616 }, { "epoch": 0.79, "learning_rate": 4.323169494647949e-06, "loss": 0.8466, "step": 6617 }, { "epoch": 0.79, "learning_rate": 4.322949552543059e-06, "loss": 1.0551, "step": 6618 }, { "epoch": 0.79, "learning_rate": 4.32272958030466e-06, "loss": 0.8636, "step": 6619 }, { "epoch": 0.79, "learning_rate": 4.322509577936388e-06, "loss": 0.5568, "step": 6620 }, { "epoch": 0.79, "learning_rate": 4.32228954544188e-06, "loss": 0.5664, "step": 6621 }, { "epoch": 0.79, "learning_rate": 4.322069482824773e-06, "loss": 0.9176, "step": 6622 }, { "epoch": 0.79, "learning_rate": 4.3218493900887035e-06, "loss": 0.9223, "step": 6623 }, { "epoch": 0.79, "learning_rate": 4.321629267237311e-06, "loss": 1.016, "step": 6624 }, { "epoch": 0.79, "learning_rate": 4.321409114274234e-06, "loss": 1.0027, "step": 6625 }, { "epoch": 0.79, "learning_rate": 4.321188931203112e-06, "loss": 0.7003, "step": 6626 }, { "epoch": 0.79, "learning_rate": 4.320968718027584e-06, "loss": 0.6141, "step": 6627 }, { "epoch": 0.79, "learning_rate": 4.3207484747512905e-06, "loss": 0.6732, "step": 6628 }, { "epoch": 0.79, "learning_rate": 4.320528201377871e-06, "loss": 0.8237, "step": 6629 }, { "epoch": 0.79, "learning_rate": 4.320307897910969e-06, "loss": 1.1089, "step": 6630 }, { "epoch": 0.79, "learning_rate": 4.320087564354224e-06, "loss": 1.0814, "step": 6631 }, { "epoch": 0.79, "learning_rate": 4.3198672007112785e-06, "loss": 0.5471, "step": 6632 }, { "epoch": 0.79, "learning_rate": 4.319646806985776e-06, "loss": 0.7909, "step": 6633 }, { "epoch": 0.79, "learning_rate": 4.319426383181358e-06, "loss": 0.7687, "step": 6634 }, { "epoch": 0.79, "learning_rate": 4.319205929301671e-06, "loss": 0.7646, "step": 6635 }, { "epoch": 0.79, "learning_rate": 4.318985445350357e-06, "loss": 0.9301, "step": 6636 }, { "epoch": 0.79, "learning_rate": 4.3187649313310594e-06, "loss": 1.1027, "step": 6637 }, { "epoch": 0.79, "learning_rate": 4.3185443872474255e-06, "loss": 0.6456, "step": 6638 }, { "epoch": 0.79, "learning_rate": 4.318323813103101e-06, "loss": 0.6611, "step": 6639 }, { "epoch": 0.79, "learning_rate": 4.318103208901731e-06, "loss": 0.7368, "step": 6640 }, { "epoch": 0.79, "learning_rate": 4.317882574646962e-06, "loss": 0.7448, "step": 6641 }, { "epoch": 0.79, "learning_rate": 4.317661910342442e-06, "loss": 1.0024, "step": 6642 }, { "epoch": 0.79, "learning_rate": 4.3174412159918185e-06, "loss": 1.0871, "step": 6643 }, { "epoch": 0.79, "learning_rate": 4.317220491598739e-06, "loss": 0.8144, "step": 6644 }, { "epoch": 0.79, "learning_rate": 4.316999737166852e-06, "loss": 0.7119, "step": 6645 }, { "epoch": 0.79, "learning_rate": 4.316778952699806e-06, "loss": 0.7822, "step": 6646 }, { "epoch": 0.79, "learning_rate": 4.316558138201253e-06, "loss": 0.7395, "step": 6647 }, { "epoch": 0.79, "learning_rate": 4.316337293674841e-06, "loss": 0.8926, "step": 6648 }, { "epoch": 0.79, "learning_rate": 4.316116419124221e-06, "loss": 1.0695, "step": 6649 }, { "epoch": 0.79, "learning_rate": 4.315895514553044e-06, "loss": 1.0617, "step": 6650 }, { "epoch": 0.79, "learning_rate": 4.315674579964962e-06, "loss": 0.6856, "step": 6651 }, { "epoch": 0.79, "learning_rate": 4.315453615363627e-06, "loss": 0.7694, "step": 6652 }, { "epoch": 0.79, "learning_rate": 4.315232620752691e-06, "loss": 0.6636, "step": 6653 }, { "epoch": 0.79, "learning_rate": 4.315011596135808e-06, "loss": 0.7544, "step": 6654 }, { "epoch": 0.79, "learning_rate": 4.31479054151663e-06, "loss": 1.0141, "step": 6655 }, { "epoch": 0.79, "learning_rate": 4.314569456898813e-06, "loss": 1.184, "step": 6656 }, { "epoch": 0.79, "learning_rate": 4.3143483422860114e-06, "loss": 0.5044, "step": 6657 }, { "epoch": 0.79, "learning_rate": 4.314127197681878e-06, "loss": 0.6146, "step": 6658 }, { "epoch": 0.79, "learning_rate": 4.31390602309007e-06, "loss": 0.719, "step": 6659 }, { "epoch": 0.79, "learning_rate": 4.3136848185142435e-06, "loss": 0.7718, "step": 6660 }, { "epoch": 0.79, "learning_rate": 4.313463583958055e-06, "loss": 1.0331, "step": 6661 }, { "epoch": 0.79, "learning_rate": 4.313242319425162e-06, "loss": 1.1517, "step": 6662 }, { "epoch": 0.79, "learning_rate": 4.31302102491922e-06, "loss": 0.6006, "step": 6663 }, { "epoch": 0.79, "learning_rate": 4.312799700443889e-06, "loss": 0.5828, "step": 6664 }, { "epoch": 0.79, "learning_rate": 4.312578346002826e-06, "loss": 0.6753, "step": 6665 }, { "epoch": 0.79, "learning_rate": 4.312356961599692e-06, "loss": 0.7781, "step": 6666 }, { "epoch": 0.79, "learning_rate": 4.312135547238144e-06, "loss": 0.7943, "step": 6667 }, { "epoch": 0.79, "learning_rate": 4.3119141029218445e-06, "loss": 1.0295, "step": 6668 }, { "epoch": 0.79, "learning_rate": 4.311692628654453e-06, "loss": 0.7595, "step": 6669 }, { "epoch": 0.79, "learning_rate": 4.31147112443963e-06, "loss": 0.6349, "step": 6670 }, { "epoch": 0.79, "learning_rate": 4.311249590281037e-06, "loss": 0.6976, "step": 6671 }, { "epoch": 0.79, "learning_rate": 4.311028026182336e-06, "loss": 0.679, "step": 6672 }, { "epoch": 0.79, "learning_rate": 4.3108064321471905e-06, "loss": 0.8102, "step": 6673 }, { "epoch": 0.79, "learning_rate": 4.3105848081792625e-06, "loss": 1.0493, "step": 6674 }, { "epoch": 0.79, "learning_rate": 4.3103631542822155e-06, "loss": 1.3481, "step": 6675 }, { "epoch": 0.8, "learning_rate": 4.310141470459715e-06, "loss": 0.5701, "step": 6676 }, { "epoch": 0.8, "learning_rate": 4.309919756715422e-06, "loss": 0.7371, "step": 6677 }, { "epoch": 0.8, "learning_rate": 4.309698013053004e-06, "loss": 0.7725, "step": 6678 }, { "epoch": 0.8, "learning_rate": 4.309476239476127e-06, "loss": 0.9293, "step": 6679 }, { "epoch": 0.8, "learning_rate": 4.309254435988456e-06, "loss": 0.9494, "step": 6680 }, { "epoch": 0.8, "learning_rate": 4.309032602593657e-06, "loss": 1.1735, "step": 6681 }, { "epoch": 0.8, "learning_rate": 4.308810739295396e-06, "loss": 0.5179, "step": 6682 }, { "epoch": 0.8, "learning_rate": 4.308588846097343e-06, "loss": 0.8007, "step": 6683 }, { "epoch": 0.8, "learning_rate": 4.308366923003164e-06, "loss": 0.7057, "step": 6684 }, { "epoch": 0.8, "learning_rate": 4.308144970016529e-06, "loss": 0.889, "step": 6685 }, { "epoch": 0.8, "learning_rate": 4.307922987141106e-06, "loss": 1.0669, "step": 6686 }, { "epoch": 0.8, "learning_rate": 4.307700974380563e-06, "loss": 1.1738, "step": 6687 }, { "epoch": 0.8, "learning_rate": 4.307478931738573e-06, "loss": 0.6222, "step": 6688 }, { "epoch": 0.8, "learning_rate": 4.307256859218803e-06, "loss": 0.5554, "step": 6689 }, { "epoch": 0.8, "learning_rate": 4.307034756824925e-06, "loss": 0.7391, "step": 6690 }, { "epoch": 0.8, "learning_rate": 4.306812624560612e-06, "loss": 0.7345, "step": 6691 }, { "epoch": 0.8, "learning_rate": 4.306590462429535e-06, "loss": 0.8207, "step": 6692 }, { "epoch": 0.8, "learning_rate": 4.306368270435365e-06, "loss": 1.0651, "step": 6693 }, { "epoch": 0.8, "learning_rate": 4.306146048581777e-06, "loss": 0.8551, "step": 6694 }, { "epoch": 0.8, "learning_rate": 4.305923796872443e-06, "loss": 0.7641, "step": 6695 }, { "epoch": 0.8, "learning_rate": 4.305701515311037e-06, "loss": 0.6754, "step": 6696 }, { "epoch": 0.8, "learning_rate": 4.3054792039012335e-06, "loss": 0.7685, "step": 6697 }, { "epoch": 0.8, "learning_rate": 4.305256862646708e-06, "loss": 0.9597, "step": 6698 }, { "epoch": 0.8, "learning_rate": 4.305034491551133e-06, "loss": 1.1108, "step": 6699 }, { "epoch": 0.8, "learning_rate": 4.304812090618188e-06, "loss": 0.9427, "step": 6700 }, { "epoch": 0.8, "learning_rate": 4.304589659851547e-06, "loss": 0.6527, "step": 6701 }, { "epoch": 0.8, "learning_rate": 4.304367199254888e-06, "loss": 0.6354, "step": 6702 }, { "epoch": 0.8, "learning_rate": 4.304144708831888e-06, "loss": 0.8301, "step": 6703 }, { "epoch": 0.8, "learning_rate": 4.303922188586224e-06, "loss": 0.867, "step": 6704 }, { "epoch": 0.8, "learning_rate": 4.303699638521575e-06, "loss": 1.0379, "step": 6705 }, { "epoch": 0.8, "learning_rate": 4.30347705864162e-06, "loss": 1.2329, "step": 6706 }, { "epoch": 0.8, "learning_rate": 4.303254448950037e-06, "loss": 0.6249, "step": 6707 }, { "epoch": 0.8, "learning_rate": 4.303031809450508e-06, "loss": 0.7415, "step": 6708 }, { "epoch": 0.8, "learning_rate": 4.302809140146711e-06, "loss": 0.8375, "step": 6709 }, { "epoch": 0.8, "learning_rate": 4.3025864410423275e-06, "loss": 0.9201, "step": 6710 }, { "epoch": 0.8, "learning_rate": 4.302363712141039e-06, "loss": 0.971, "step": 6711 }, { "epoch": 0.8, "learning_rate": 4.3021409534465275e-06, "loss": 1.1758, "step": 6712 }, { "epoch": 0.8, "learning_rate": 4.301918164962474e-06, "loss": 0.6625, "step": 6713 }, { "epoch": 0.8, "learning_rate": 4.3016953466925635e-06, "loss": 0.7071, "step": 6714 }, { "epoch": 0.8, "learning_rate": 4.3014724986404766e-06, "loss": 0.7577, "step": 6715 }, { "epoch": 0.8, "learning_rate": 4.301249620809899e-06, "loss": 0.8036, "step": 6716 }, { "epoch": 0.8, "learning_rate": 4.3010267132045125e-06, "loss": 0.9365, "step": 6717 }, { "epoch": 0.8, "learning_rate": 4.3008037758280046e-06, "loss": 1.2125, "step": 6718 }, { "epoch": 0.8, "learning_rate": 4.3005808086840585e-06, "loss": 0.8345, "step": 6719 }, { "epoch": 0.8, "learning_rate": 4.3003578117763615e-06, "loss": 0.6095, "step": 6720 }, { "epoch": 0.8, "learning_rate": 4.300134785108598e-06, "loss": 0.7062, "step": 6721 }, { "epoch": 0.8, "learning_rate": 4.299911728684456e-06, "loss": 0.7606, "step": 6722 }, { "epoch": 0.8, "learning_rate": 4.2996886425076215e-06, "loss": 0.8945, "step": 6723 }, { "epoch": 0.8, "learning_rate": 4.2994655265817835e-06, "loss": 1.0066, "step": 6724 }, { "epoch": 0.8, "learning_rate": 4.2992423809106285e-06, "loss": 1.2658, "step": 6725 }, { "epoch": 0.8, "learning_rate": 4.299019205497846e-06, "loss": 0.6824, "step": 6726 }, { "epoch": 0.8, "learning_rate": 4.298796000347126e-06, "loss": 0.717, "step": 6727 }, { "epoch": 0.8, "learning_rate": 4.2985727654621564e-06, "loss": 0.8539, "step": 6728 }, { "epoch": 0.8, "learning_rate": 4.2983495008466285e-06, "loss": 0.9457, "step": 6729 }, { "epoch": 0.8, "learning_rate": 4.298126206504232e-06, "loss": 1.1549, "step": 6730 }, { "epoch": 0.8, "learning_rate": 4.297902882438659e-06, "loss": 1.0947, "step": 6731 }, { "epoch": 0.8, "learning_rate": 4.2976795286536e-06, "loss": 0.6387, "step": 6732 }, { "epoch": 0.8, "learning_rate": 4.2974561451527476e-06, "loss": 0.7143, "step": 6733 }, { "epoch": 0.8, "learning_rate": 4.2972327319397944e-06, "loss": 0.9361, "step": 6734 }, { "epoch": 0.8, "learning_rate": 4.297009289018434e-06, "loss": 0.7829, "step": 6735 }, { "epoch": 0.8, "learning_rate": 4.296785816392358e-06, "loss": 0.8969, "step": 6736 }, { "epoch": 0.8, "learning_rate": 4.296562314065262e-06, "loss": 1.1154, "step": 6737 }, { "epoch": 0.8, "learning_rate": 4.296338782040841e-06, "loss": 0.6412, "step": 6738 }, { "epoch": 0.8, "learning_rate": 4.296115220322788e-06, "loss": 0.5145, "step": 6739 }, { "epoch": 0.8, "learning_rate": 4.295891628914801e-06, "loss": 0.7599, "step": 6740 }, { "epoch": 0.8, "learning_rate": 4.295668007820574e-06, "loss": 0.6887, "step": 6741 }, { "epoch": 0.8, "learning_rate": 4.295444357043805e-06, "loss": 0.8641, "step": 6742 }, { "epoch": 0.8, "learning_rate": 4.295220676588189e-06, "loss": 1.1998, "step": 6743 }, { "epoch": 0.8, "learning_rate": 4.2949969664574255e-06, "loss": 0.8582, "step": 6744 }, { "epoch": 0.8, "learning_rate": 4.2947732266552105e-06, "loss": 0.8291, "step": 6745 }, { "epoch": 0.8, "learning_rate": 4.294549457185244e-06, "loss": 0.7014, "step": 6746 }, { "epoch": 0.8, "learning_rate": 4.2943256580512245e-06, "loss": 0.7499, "step": 6747 }, { "epoch": 0.8, "learning_rate": 4.294101829256852e-06, "loss": 0.9335, "step": 6748 }, { "epoch": 0.8, "learning_rate": 4.293877970805824e-06, "loss": 1.0967, "step": 6749 }, { "epoch": 0.8, "learning_rate": 4.293654082701843e-06, "loss": 1.1298, "step": 6750 }, { "epoch": 0.8, "learning_rate": 4.293430164948611e-06, "loss": 0.4785, "step": 6751 }, { "epoch": 0.8, "learning_rate": 4.293206217549827e-06, "loss": 0.7979, "step": 6752 }, { "epoch": 0.8, "learning_rate": 4.292982240509194e-06, "loss": 0.6339, "step": 6753 }, { "epoch": 0.8, "learning_rate": 4.2927582338304135e-06, "loss": 0.8209, "step": 6754 }, { "epoch": 0.8, "learning_rate": 4.29253419751719e-06, "loss": 1.0956, "step": 6755 }, { "epoch": 0.8, "learning_rate": 4.2923101315732244e-06, "loss": 1.0503, "step": 6756 }, { "epoch": 0.8, "learning_rate": 4.292086036002222e-06, "loss": 0.441, "step": 6757 }, { "epoch": 0.8, "learning_rate": 4.291861910807889e-06, "loss": 0.6553, "step": 6758 }, { "epoch": 0.8, "learning_rate": 4.291637755993927e-06, "loss": 0.7497, "step": 6759 }, { "epoch": 0.81, "learning_rate": 4.291413571564042e-06, "loss": 0.8225, "step": 6760 }, { "epoch": 0.81, "learning_rate": 4.291189357521941e-06, "loss": 0.9625, "step": 6761 }, { "epoch": 0.81, "learning_rate": 4.290965113871329e-06, "loss": 1.0508, "step": 6762 }, { "epoch": 0.81, "learning_rate": 4.290740840615914e-06, "loss": 0.6252, "step": 6763 }, { "epoch": 0.81, "learning_rate": 4.290516537759403e-06, "loss": 0.6141, "step": 6764 }, { "epoch": 0.81, "learning_rate": 4.2902922053055026e-06, "loss": 0.6853, "step": 6765 }, { "epoch": 0.81, "learning_rate": 4.2900678432579225e-06, "loss": 0.787, "step": 6766 }, { "epoch": 0.81, "learning_rate": 4.2898434516203705e-06, "loss": 0.9293, "step": 6767 }, { "epoch": 0.81, "learning_rate": 4.2896190303965564e-06, "loss": 1.1193, "step": 6768 }, { "epoch": 0.81, "learning_rate": 4.2893945795901896e-06, "loss": 0.563, "step": 6769 }, { "epoch": 0.81, "learning_rate": 4.289170099204979e-06, "loss": 0.6007, "step": 6770 }, { "epoch": 0.81, "learning_rate": 4.288945589244637e-06, "loss": 0.7391, "step": 6771 }, { "epoch": 0.81, "learning_rate": 4.288721049712875e-06, "loss": 0.7263, "step": 6772 }, { "epoch": 0.81, "learning_rate": 4.288496480613404e-06, "loss": 0.8466, "step": 6773 }, { "epoch": 0.81, "learning_rate": 4.288271881949936e-06, "loss": 1.0721, "step": 6774 }, { "epoch": 0.81, "learning_rate": 4.288047253726183e-06, "loss": 0.9766, "step": 6775 }, { "epoch": 0.81, "learning_rate": 4.287822595945859e-06, "loss": 0.4848, "step": 6776 }, { "epoch": 0.81, "learning_rate": 4.287597908612678e-06, "loss": 0.659, "step": 6777 }, { "epoch": 0.81, "learning_rate": 4.287373191730354e-06, "loss": 0.7177, "step": 6778 }, { "epoch": 0.81, "learning_rate": 4.2871484453026e-06, "loss": 0.8481, "step": 6779 }, { "epoch": 0.81, "learning_rate": 4.286923669333134e-06, "loss": 0.9365, "step": 6780 }, { "epoch": 0.81, "learning_rate": 4.286698863825668e-06, "loss": 1.2365, "step": 6781 }, { "epoch": 0.81, "learning_rate": 4.28647402878392e-06, "loss": 0.6336, "step": 6782 }, { "epoch": 0.81, "learning_rate": 4.286249164211607e-06, "loss": 0.5326, "step": 6783 }, { "epoch": 0.81, "learning_rate": 4.286024270112446e-06, "loss": 0.7355, "step": 6784 }, { "epoch": 0.81, "learning_rate": 4.285799346490152e-06, "loss": 0.827, "step": 6785 }, { "epoch": 0.81, "learning_rate": 4.285574393348447e-06, "loss": 1.0781, "step": 6786 }, { "epoch": 0.81, "learning_rate": 4.285349410691047e-06, "loss": 1.1783, "step": 6787 }, { "epoch": 0.81, "learning_rate": 4.285124398521671e-06, "loss": 0.5145, "step": 6788 }, { "epoch": 0.81, "learning_rate": 4.284899356844039e-06, "loss": 0.7639, "step": 6789 }, { "epoch": 0.81, "learning_rate": 4.284674285661871e-06, "loss": 0.6256, "step": 6790 }, { "epoch": 0.81, "learning_rate": 4.284449184978888e-06, "loss": 0.8102, "step": 6791 }, { "epoch": 0.81, "learning_rate": 4.284224054798809e-06, "loss": 0.8908, "step": 6792 }, { "epoch": 0.81, "learning_rate": 4.283998895125357e-06, "loss": 1.0854, "step": 6793 }, { "epoch": 0.81, "learning_rate": 4.283773705962254e-06, "loss": 0.8221, "step": 6794 }, { "epoch": 0.81, "learning_rate": 4.283548487313222e-06, "loss": 0.5783, "step": 6795 }, { "epoch": 0.81, "learning_rate": 4.283323239181984e-06, "loss": 0.855, "step": 6796 }, { "epoch": 0.81, "learning_rate": 4.2830979615722625e-06, "loss": 0.8584, "step": 6797 }, { "epoch": 0.81, "learning_rate": 4.282872654487782e-06, "loss": 0.947, "step": 6798 }, { "epoch": 0.81, "learning_rate": 4.282647317932267e-06, "loss": 0.9918, "step": 6799 }, { "epoch": 0.81, "learning_rate": 4.282421951909443e-06, "loss": 1.146, "step": 6800 }, { "epoch": 0.81, "learning_rate": 4.282196556423034e-06, "loss": 0.5928, "step": 6801 }, { "epoch": 0.81, "learning_rate": 4.281971131476766e-06, "loss": 0.7269, "step": 6802 }, { "epoch": 0.81, "learning_rate": 4.2817456770743665e-06, "loss": 0.8916, "step": 6803 }, { "epoch": 0.81, "learning_rate": 4.281520193219561e-06, "loss": 0.875, "step": 6804 }, { "epoch": 0.81, "learning_rate": 4.281294679916077e-06, "loss": 1.0726, "step": 6805 }, { "epoch": 0.81, "learning_rate": 4.281069137167644e-06, "loss": 1.1769, "step": 6806 }, { "epoch": 0.81, "learning_rate": 4.280843564977987e-06, "loss": 0.599, "step": 6807 }, { "epoch": 0.81, "learning_rate": 4.280617963350837e-06, "loss": 0.6354, "step": 6808 }, { "epoch": 0.81, "learning_rate": 4.280392332289923e-06, "loss": 0.7825, "step": 6809 }, { "epoch": 0.81, "learning_rate": 4.280166671798973e-06, "loss": 0.8446, "step": 6810 }, { "epoch": 0.81, "learning_rate": 4.27994098188172e-06, "loss": 1.0146, "step": 6811 }, { "epoch": 0.81, "learning_rate": 4.279715262541892e-06, "loss": 1.1295, "step": 6812 }, { "epoch": 0.81, "learning_rate": 4.279489513783223e-06, "loss": 0.5126, "step": 6813 }, { "epoch": 0.81, "learning_rate": 4.279263735609442e-06, "loss": 0.6233, "step": 6814 }, { "epoch": 0.81, "learning_rate": 4.279037928024283e-06, "loss": 0.7129, "step": 6815 }, { "epoch": 0.81, "learning_rate": 4.278812091031477e-06, "loss": 0.7578, "step": 6816 }, { "epoch": 0.81, "learning_rate": 4.278586224634757e-06, "loss": 0.8146, "step": 6817 }, { "epoch": 0.81, "learning_rate": 4.278360328837859e-06, "loss": 1.0866, "step": 6818 }, { "epoch": 0.81, "learning_rate": 4.278134403644515e-06, "loss": 0.7297, "step": 6819 }, { "epoch": 0.81, "learning_rate": 4.27790844905846e-06, "loss": 0.5298, "step": 6820 }, { "epoch": 0.81, "learning_rate": 4.27768246508343e-06, "loss": 0.7187, "step": 6821 }, { "epoch": 0.81, "learning_rate": 4.277456451723158e-06, "loss": 0.7515, "step": 6822 }, { "epoch": 0.81, "learning_rate": 4.277230408981384e-06, "loss": 0.993, "step": 6823 }, { "epoch": 0.81, "learning_rate": 4.277004336861841e-06, "loss": 0.9476, "step": 6824 }, { "epoch": 0.81, "learning_rate": 4.276778235368266e-06, "loss": 1.1486, "step": 6825 }, { "epoch": 0.81, "learning_rate": 4.276552104504399e-06, "loss": 0.5158, "step": 6826 }, { "epoch": 0.81, "learning_rate": 4.276325944273977e-06, "loss": 0.6292, "step": 6827 }, { "epoch": 0.81, "learning_rate": 4.276099754680737e-06, "loss": 0.7183, "step": 6828 }, { "epoch": 0.81, "learning_rate": 4.27587353572842e-06, "loss": 0.7801, "step": 6829 }, { "epoch": 0.81, "learning_rate": 4.275647287420765e-06, "loss": 1.0485, "step": 6830 }, { "epoch": 0.81, "learning_rate": 4.27542100976151e-06, "loss": 1.0333, "step": 6831 }, { "epoch": 0.81, "learning_rate": 4.275194702754397e-06, "loss": 0.7489, "step": 6832 }, { "epoch": 0.81, "learning_rate": 4.274968366403166e-06, "loss": 0.7788, "step": 6833 }, { "epoch": 0.81, "learning_rate": 4.27474200071156e-06, "loss": 0.6174, "step": 6834 }, { "epoch": 0.81, "learning_rate": 4.274515605683319e-06, "loss": 0.8412, "step": 6835 }, { "epoch": 0.81, "learning_rate": 4.274289181322187e-06, "loss": 0.9916, "step": 6836 }, { "epoch": 0.81, "learning_rate": 4.2740627276319055e-06, "loss": 1.1884, "step": 6837 }, { "epoch": 0.81, "learning_rate": 4.273836244616218e-06, "loss": 0.683, "step": 6838 }, { "epoch": 0.81, "learning_rate": 4.27360973227887e-06, "loss": 0.6342, "step": 6839 }, { "epoch": 0.81, "learning_rate": 4.273383190623603e-06, "loss": 0.6768, "step": 6840 }, { "epoch": 0.81, "learning_rate": 4.273156619654163e-06, "loss": 0.7894, "step": 6841 }, { "epoch": 0.81, "learning_rate": 4.272930019374295e-06, "loss": 0.843, "step": 6842 }, { "epoch": 0.81, "learning_rate": 4.272703389787747e-06, "loss": 1.0356, "step": 6843 }, { "epoch": 0.82, "learning_rate": 4.272476730898261e-06, "loss": 0.9065, "step": 6844 }, { "epoch": 0.82, "learning_rate": 4.272250042709587e-06, "loss": 0.5736, "step": 6845 }, { "epoch": 0.82, "learning_rate": 4.272023325225471e-06, "loss": 0.6941, "step": 6846 }, { "epoch": 0.82, "learning_rate": 4.2717965784496606e-06, "loss": 0.7484, "step": 6847 }, { "epoch": 0.82, "learning_rate": 4.271569802385904e-06, "loss": 0.825, "step": 6848 }, { "epoch": 0.82, "learning_rate": 4.27134299703795e-06, "loss": 1.0894, "step": 6849 }, { "epoch": 0.82, "learning_rate": 4.271116162409548e-06, "loss": 0.995, "step": 6850 }, { "epoch": 0.82, "learning_rate": 4.270889298504446e-06, "loss": 0.6473, "step": 6851 }, { "epoch": 0.82, "learning_rate": 4.270662405326397e-06, "loss": 0.6754, "step": 6852 }, { "epoch": 0.82, "learning_rate": 4.270435482879149e-06, "loss": 0.6882, "step": 6853 }, { "epoch": 0.82, "learning_rate": 4.270208531166453e-06, "loss": 1.0335, "step": 6854 }, { "epoch": 0.82, "learning_rate": 4.269981550192063e-06, "loss": 1.0043, "step": 6855 }, { "epoch": 0.82, "learning_rate": 4.269754539959729e-06, "loss": 1.2223, "step": 6856 }, { "epoch": 0.82, "learning_rate": 4.269527500473204e-06, "loss": 0.6673, "step": 6857 }, { "epoch": 0.82, "learning_rate": 4.2693004317362405e-06, "loss": 0.7147, "step": 6858 }, { "epoch": 0.82, "learning_rate": 4.269073333752592e-06, "loss": 0.7185, "step": 6859 }, { "epoch": 0.82, "learning_rate": 4.268846206526014e-06, "loss": 0.8205, "step": 6860 }, { "epoch": 0.82, "learning_rate": 4.268619050060259e-06, "loss": 0.9791, "step": 6861 }, { "epoch": 0.82, "learning_rate": 4.2683918643590825e-06, "loss": 1.1773, "step": 6862 }, { "epoch": 0.82, "learning_rate": 4.2681646494262414e-06, "loss": 0.5539, "step": 6863 }, { "epoch": 0.82, "learning_rate": 4.26793740526549e-06, "loss": 0.6029, "step": 6864 }, { "epoch": 0.82, "learning_rate": 4.267710131880583e-06, "loss": 0.7282, "step": 6865 }, { "epoch": 0.82, "learning_rate": 4.267482829275281e-06, "loss": 0.6721, "step": 6866 }, { "epoch": 0.82, "learning_rate": 4.26725549745334e-06, "loss": 0.9241, "step": 6867 }, { "epoch": 0.82, "learning_rate": 4.267028136418516e-06, "loss": 1.0176, "step": 6868 }, { "epoch": 0.82, "learning_rate": 4.26680074617457e-06, "loss": 0.8701, "step": 6869 }, { "epoch": 0.82, "learning_rate": 4.2665733267252575e-06, "loss": 0.6017, "step": 6870 }, { "epoch": 0.82, "learning_rate": 4.2663458780743414e-06, "loss": 0.9164, "step": 6871 }, { "epoch": 0.82, "learning_rate": 4.266118400225579e-06, "loss": 0.7526, "step": 6872 }, { "epoch": 0.82, "learning_rate": 4.265890893182731e-06, "loss": 0.9025, "step": 6873 }, { "epoch": 0.82, "learning_rate": 4.265663356949559e-06, "loss": 1.1512, "step": 6874 }, { "epoch": 0.82, "learning_rate": 4.265435791529823e-06, "loss": 1.1669, "step": 6875 }, { "epoch": 0.82, "learning_rate": 4.265208196927285e-06, "loss": 0.6346, "step": 6876 }, { "epoch": 0.82, "learning_rate": 4.264980573145709e-06, "loss": 0.8811, "step": 6877 }, { "epoch": 0.82, "learning_rate": 4.264752920188854e-06, "loss": 0.7628, "step": 6878 }, { "epoch": 0.82, "learning_rate": 4.2645252380604865e-06, "loss": 0.9666, "step": 6879 }, { "epoch": 0.82, "learning_rate": 4.264297526764368e-06, "loss": 1.0464, "step": 6880 }, { "epoch": 0.82, "learning_rate": 4.264069786304263e-06, "loss": 1.046, "step": 6881 }, { "epoch": 0.82, "learning_rate": 4.263842016683938e-06, "loss": 0.6478, "step": 6882 }, { "epoch": 0.82, "learning_rate": 4.263614217907156e-06, "loss": 0.5954, "step": 6883 }, { "epoch": 0.82, "learning_rate": 4.263386389977682e-06, "loss": 0.5735, "step": 6884 }, { "epoch": 0.82, "learning_rate": 4.263158532899284e-06, "loss": 0.7913, "step": 6885 }, { "epoch": 0.82, "learning_rate": 4.262930646675727e-06, "loss": 0.9848, "step": 6886 }, { "epoch": 0.82, "learning_rate": 4.262702731310779e-06, "loss": 1.1462, "step": 6887 }, { "epoch": 0.82, "learning_rate": 4.262474786808206e-06, "loss": 0.6125, "step": 6888 }, { "epoch": 0.82, "learning_rate": 4.262246813171777e-06, "loss": 0.4948, "step": 6889 }, { "epoch": 0.82, "learning_rate": 4.262018810405261e-06, "loss": 0.5919, "step": 6890 }, { "epoch": 0.82, "learning_rate": 4.261790778512426e-06, "loss": 0.7788, "step": 6891 }, { "epoch": 0.82, "learning_rate": 4.261562717497042e-06, "loss": 0.9362, "step": 6892 }, { "epoch": 0.82, "learning_rate": 4.261334627362878e-06, "loss": 0.9702, "step": 6893 }, { "epoch": 0.82, "learning_rate": 4.261106508113704e-06, "loss": 0.7582, "step": 6894 }, { "epoch": 0.82, "learning_rate": 4.260878359753294e-06, "loss": 0.6386, "step": 6895 }, { "epoch": 0.82, "learning_rate": 4.260650182285414e-06, "loss": 0.7433, "step": 6896 }, { "epoch": 0.82, "learning_rate": 4.26042197571384e-06, "loss": 0.8325, "step": 6897 }, { "epoch": 0.82, "learning_rate": 4.260193740042343e-06, "loss": 0.9687, "step": 6898 }, { "epoch": 0.82, "learning_rate": 4.259965475274696e-06, "loss": 1.1583, "step": 6899 }, { "epoch": 0.82, "learning_rate": 4.259737181414673e-06, "loss": 1.2081, "step": 6900 }, { "epoch": 0.82, "learning_rate": 4.259508858466045e-06, "loss": 0.5804, "step": 6901 }, { "epoch": 0.82, "learning_rate": 4.259280506432589e-06, "loss": 0.7574, "step": 6902 }, { "epoch": 0.82, "learning_rate": 4.259052125318078e-06, "loss": 0.6521, "step": 6903 }, { "epoch": 0.82, "learning_rate": 4.258823715126289e-06, "loss": 0.8326, "step": 6904 }, { "epoch": 0.82, "learning_rate": 4.258595275860995e-06, "loss": 0.9958, "step": 6905 }, { "epoch": 0.82, "learning_rate": 4.258366807525974e-06, "loss": 1.4019, "step": 6906 }, { "epoch": 0.82, "learning_rate": 4.258138310125002e-06, "loss": 0.6223, "step": 6907 }, { "epoch": 0.82, "learning_rate": 4.257909783661856e-06, "loss": 0.6721, "step": 6908 }, { "epoch": 0.82, "learning_rate": 4.257681228140314e-06, "loss": 0.8351, "step": 6909 }, { "epoch": 0.82, "learning_rate": 4.257452643564155e-06, "loss": 0.7521, "step": 6910 }, { "epoch": 0.82, "learning_rate": 4.257224029937155e-06, "loss": 1.0045, "step": 6911 }, { "epoch": 0.82, "learning_rate": 4.256995387263095e-06, "loss": 1.0537, "step": 6912 }, { "epoch": 0.82, "learning_rate": 4.256766715545753e-06, "loss": 0.6358, "step": 6913 }, { "epoch": 0.82, "learning_rate": 4.25653801478891e-06, "loss": 0.7247, "step": 6914 }, { "epoch": 0.82, "learning_rate": 4.256309284996347e-06, "loss": 0.7426, "step": 6915 }, { "epoch": 0.82, "learning_rate": 4.256080526171844e-06, "loss": 0.7107, "step": 6916 }, { "epoch": 0.82, "learning_rate": 4.255851738319182e-06, "loss": 0.8879, "step": 6917 }, { "epoch": 0.82, "learning_rate": 4.2556229214421436e-06, "loss": 1.0022, "step": 6918 }, { "epoch": 0.82, "learning_rate": 4.255394075544511e-06, "loss": 0.6257, "step": 6919 }, { "epoch": 0.82, "learning_rate": 4.255165200630067e-06, "loss": 0.6945, "step": 6920 }, { "epoch": 0.82, "learning_rate": 4.254936296702594e-06, "loss": 0.7123, "step": 6921 }, { "epoch": 0.82, "learning_rate": 4.2547073637658785e-06, "loss": 0.9008, "step": 6922 }, { "epoch": 0.82, "learning_rate": 4.254478401823703e-06, "loss": 0.9233, "step": 6923 }, { "epoch": 0.82, "learning_rate": 4.254249410879851e-06, "loss": 0.9926, "step": 6924 }, { "epoch": 0.82, "learning_rate": 4.25402039093811e-06, "loss": 1.2064, "step": 6925 }, { "epoch": 0.82, "learning_rate": 4.253791342002265e-06, "loss": 0.7716, "step": 6926 }, { "epoch": 0.82, "learning_rate": 4.2535622640761006e-06, "loss": 0.7589, "step": 6927 }, { "epoch": 0.83, "learning_rate": 4.2533331571634065e-06, "loss": 0.7835, "step": 6928 }, { "epoch": 0.83, "learning_rate": 4.2531040212679666e-06, "loss": 0.9174, "step": 6929 }, { "epoch": 0.83, "learning_rate": 4.252874856393571e-06, "loss": 1.1103, "step": 6930 }, { "epoch": 0.83, "learning_rate": 4.252645662544007e-06, "loss": 1.1265, "step": 6931 }, { "epoch": 0.83, "learning_rate": 4.252416439723063e-06, "loss": 0.5184, "step": 6932 }, { "epoch": 0.83, "learning_rate": 4.252187187934528e-06, "loss": 0.7159, "step": 6933 }, { "epoch": 0.83, "learning_rate": 4.251957907182192e-06, "loss": 0.6956, "step": 6934 }, { "epoch": 0.83, "learning_rate": 4.251728597469845e-06, "loss": 0.8711, "step": 6935 }, { "epoch": 0.83, "learning_rate": 4.2514992588012774e-06, "loss": 1.0696, "step": 6936 }, { "epoch": 0.83, "learning_rate": 4.251269891180279e-06, "loss": 1.2189, "step": 6937 }, { "epoch": 0.83, "learning_rate": 4.251040494610643e-06, "loss": 0.6499, "step": 6938 }, { "epoch": 0.83, "learning_rate": 4.250811069096161e-06, "loss": 0.7657, "step": 6939 }, { "epoch": 0.83, "learning_rate": 4.250581614640625e-06, "loss": 0.6836, "step": 6940 }, { "epoch": 0.83, "learning_rate": 4.250352131247828e-06, "loss": 0.6692, "step": 6941 }, { "epoch": 0.83, "learning_rate": 4.250122618921563e-06, "loss": 1.0245, "step": 6942 }, { "epoch": 0.83, "learning_rate": 4.249893077665625e-06, "loss": 1.0959, "step": 6943 }, { "epoch": 0.83, "learning_rate": 4.2496635074838075e-06, "loss": 0.7984, "step": 6944 }, { "epoch": 0.83, "learning_rate": 4.2494339083799045e-06, "loss": 0.6526, "step": 6945 }, { "epoch": 0.83, "learning_rate": 4.249204280357712e-06, "loss": 0.7545, "step": 6946 }, { "epoch": 0.83, "learning_rate": 4.248974623421027e-06, "loss": 0.6799, "step": 6947 }, { "epoch": 0.83, "learning_rate": 4.248744937573646e-06, "loss": 0.8972, "step": 6948 }, { "epoch": 0.83, "learning_rate": 4.2485152228193625e-06, "loss": 0.9141, "step": 6949 }, { "epoch": 0.83, "learning_rate": 4.2482854791619756e-06, "loss": 1.2445, "step": 6950 }, { "epoch": 0.83, "learning_rate": 4.248055706605284e-06, "loss": 0.679, "step": 6951 }, { "epoch": 0.83, "learning_rate": 4.247825905153085e-06, "loss": 0.7109, "step": 6952 }, { "epoch": 0.83, "learning_rate": 4.247596074809176e-06, "loss": 0.6508, "step": 6953 }, { "epoch": 0.83, "learning_rate": 4.247366215577357e-06, "loss": 0.873, "step": 6954 }, { "epoch": 0.83, "learning_rate": 4.247136327461429e-06, "loss": 1.121, "step": 6955 }, { "epoch": 0.83, "learning_rate": 4.2469064104651915e-06, "loss": 1.0901, "step": 6956 }, { "epoch": 0.83, "learning_rate": 4.246676464592443e-06, "loss": 0.5223, "step": 6957 }, { "epoch": 0.83, "learning_rate": 4.2464464898469865e-06, "loss": 0.6953, "step": 6958 }, { "epoch": 0.83, "learning_rate": 4.246216486232623e-06, "loss": 0.7235, "step": 6959 }, { "epoch": 0.83, "learning_rate": 4.245986453753155e-06, "loss": 0.9032, "step": 6960 }, { "epoch": 0.83, "learning_rate": 4.245756392412383e-06, "loss": 1.0978, "step": 6961 }, { "epoch": 0.83, "learning_rate": 4.245526302214112e-06, "loss": 1.0832, "step": 6962 }, { "epoch": 0.83, "learning_rate": 4.245296183162146e-06, "loss": 0.6422, "step": 6963 }, { "epoch": 0.83, "learning_rate": 4.245066035260286e-06, "loss": 0.599, "step": 6964 }, { "epoch": 0.83, "learning_rate": 4.244835858512339e-06, "loss": 0.7042, "step": 6965 }, { "epoch": 0.83, "learning_rate": 4.244605652922108e-06, "loss": 0.7363, "step": 6966 }, { "epoch": 0.83, "learning_rate": 4.244375418493401e-06, "loss": 0.9694, "step": 6967 }, { "epoch": 0.83, "learning_rate": 4.24414515523002e-06, "loss": 1.0959, "step": 6968 }, { "epoch": 0.83, "learning_rate": 4.243914863135774e-06, "loss": 0.8777, "step": 6969 }, { "epoch": 0.83, "learning_rate": 4.2436845422144685e-06, "loss": 0.7082, "step": 6970 }, { "epoch": 0.83, "learning_rate": 4.243454192469911e-06, "loss": 0.7848, "step": 6971 }, { "epoch": 0.83, "learning_rate": 4.24322381390591e-06, "loss": 0.7845, "step": 6972 }, { "epoch": 0.83, "learning_rate": 4.242993406526274e-06, "loss": 0.7989, "step": 6973 }, { "epoch": 0.83, "learning_rate": 4.24276297033481e-06, "loss": 1.0426, "step": 6974 }, { "epoch": 0.83, "learning_rate": 4.242532505335327e-06, "loss": 1.1562, "step": 6975 }, { "epoch": 0.83, "learning_rate": 4.242302011531637e-06, "loss": 0.6618, "step": 6976 }, { "epoch": 0.83, "learning_rate": 4.242071488927547e-06, "loss": 0.6386, "step": 6977 }, { "epoch": 0.83, "learning_rate": 4.24184093752687e-06, "loss": 0.8469, "step": 6978 }, { "epoch": 0.83, "learning_rate": 4.241610357333416e-06, "loss": 0.8633, "step": 6979 }, { "epoch": 0.83, "learning_rate": 4.2413797483509964e-06, "loss": 1.0152, "step": 6980 }, { "epoch": 0.83, "learning_rate": 4.241149110583424e-06, "loss": 1.0977, "step": 6981 }, { "epoch": 0.83, "learning_rate": 4.2409184440345106e-06, "loss": 0.7292, "step": 6982 }, { "epoch": 0.83, "learning_rate": 4.240687748708069e-06, "loss": 0.6759, "step": 6983 }, { "epoch": 0.83, "learning_rate": 4.240457024607913e-06, "loss": 0.8425, "step": 6984 }, { "epoch": 0.83, "learning_rate": 4.240226271737856e-06, "loss": 0.7441, "step": 6985 }, { "epoch": 0.83, "learning_rate": 4.239995490101713e-06, "loss": 0.8965, "step": 6986 }, { "epoch": 0.83, "learning_rate": 4.239764679703299e-06, "loss": 1.1672, "step": 6987 }, { "epoch": 0.83, "learning_rate": 4.239533840546428e-06, "loss": 0.6256, "step": 6988 }, { "epoch": 0.83, "learning_rate": 4.239302972634917e-06, "loss": 0.5779, "step": 6989 }, { "epoch": 0.83, "learning_rate": 4.239072075972582e-06, "loss": 0.7551, "step": 6990 }, { "epoch": 0.83, "learning_rate": 4.23884115056324e-06, "loss": 0.7495, "step": 6991 }, { "epoch": 0.83, "learning_rate": 4.238610196410707e-06, "loss": 0.8735, "step": 6992 }, { "epoch": 0.83, "learning_rate": 4.238379213518803e-06, "loss": 0.9989, "step": 6993 }, { "epoch": 0.83, "learning_rate": 4.238148201891344e-06, "loss": 0.5756, "step": 6994 }, { "epoch": 0.83, "learning_rate": 4.237917161532149e-06, "loss": 0.548, "step": 6995 }, { "epoch": 0.83, "learning_rate": 4.237686092445037e-06, "loss": 0.8288, "step": 6996 }, { "epoch": 0.83, "learning_rate": 4.237454994633829e-06, "loss": 0.8521, "step": 6997 }, { "epoch": 0.83, "learning_rate": 4.237223868102344e-06, "loss": 0.8644, "step": 6998 }, { "epoch": 0.83, "learning_rate": 4.236992712854403e-06, "loss": 1.1175, "step": 6999 }, { "epoch": 0.83, "learning_rate": 4.236761528893826e-06, "loss": 1.1101, "step": 7000 }, { "epoch": 0.83, "learning_rate": 4.236530316224435e-06, "loss": 0.4969, "step": 7001 }, { "epoch": 0.83, "learning_rate": 4.236299074850053e-06, "loss": 0.6324, "step": 7002 }, { "epoch": 0.83, "learning_rate": 4.236067804774502e-06, "loss": 0.8826, "step": 7003 }, { "epoch": 0.83, "learning_rate": 4.235836506001604e-06, "loss": 0.8651, "step": 7004 }, { "epoch": 0.83, "learning_rate": 4.235605178535182e-06, "loss": 0.9796, "step": 7005 }, { "epoch": 0.83, "learning_rate": 4.235373822379062e-06, "loss": 1.1622, "step": 7006 }, { "epoch": 0.83, "learning_rate": 4.235142437537066e-06, "loss": 0.6544, "step": 7007 }, { "epoch": 0.83, "learning_rate": 4.234911024013021e-06, "loss": 0.7073, "step": 7008 }, { "epoch": 0.83, "learning_rate": 4.234679581810751e-06, "loss": 0.8874, "step": 7009 }, { "epoch": 0.83, "learning_rate": 4.2344481109340816e-06, "loss": 0.8244, "step": 7010 }, { "epoch": 0.83, "learning_rate": 4.2342166113868385e-06, "loss": 0.9739, "step": 7011 }, { "epoch": 0.84, "learning_rate": 4.233985083172851e-06, "loss": 0.9967, "step": 7012 }, { "epoch": 0.84, "learning_rate": 4.233753526295944e-06, "loss": 0.5876, "step": 7013 }, { "epoch": 0.84, "learning_rate": 4.233521940759946e-06, "loss": 0.6346, "step": 7014 }, { "epoch": 0.84, "learning_rate": 4.233290326568684e-06, "loss": 0.7378, "step": 7015 }, { "epoch": 0.84, "learning_rate": 4.233058683725988e-06, "loss": 0.8348, "step": 7016 }, { "epoch": 0.84, "learning_rate": 4.232827012235687e-06, "loss": 0.8809, "step": 7017 }, { "epoch": 0.84, "learning_rate": 4.232595312101609e-06, "loss": 1.0805, "step": 7018 }, { "epoch": 0.84, "learning_rate": 4.232363583327586e-06, "loss": 0.7941, "step": 7019 }, { "epoch": 0.84, "learning_rate": 4.232131825917447e-06, "loss": 0.5345, "step": 7020 }, { "epoch": 0.84, "learning_rate": 4.231900039875024e-06, "loss": 0.8005, "step": 7021 }, { "epoch": 0.84, "learning_rate": 4.231668225204148e-06, "loss": 0.8328, "step": 7022 }, { "epoch": 0.84, "learning_rate": 4.231436381908652e-06, "loss": 0.9749, "step": 7023 }, { "epoch": 0.84, "learning_rate": 4.231204509992365e-06, "loss": 1.0655, "step": 7024 }, { "epoch": 0.84, "learning_rate": 4.230972609459123e-06, "loss": 1.0898, "step": 7025 }, { "epoch": 0.84, "learning_rate": 4.2307406803127595e-06, "loss": 0.5444, "step": 7026 }, { "epoch": 0.84, "learning_rate": 4.230508722557107e-06, "loss": 0.755, "step": 7027 }, { "epoch": 0.84, "learning_rate": 4.2302767361959995e-06, "loss": 0.7991, "step": 7028 }, { "epoch": 0.84, "learning_rate": 4.230044721233273e-06, "loss": 0.8957, "step": 7029 }, { "epoch": 0.84, "learning_rate": 4.229812677672761e-06, "loss": 1.0477, "step": 7030 }, { "epoch": 0.84, "learning_rate": 4.229580605518302e-06, "loss": 1.0796, "step": 7031 }, { "epoch": 0.84, "learning_rate": 4.229348504773729e-06, "loss": 0.5674, "step": 7032 }, { "epoch": 0.84, "learning_rate": 4.229116375442881e-06, "loss": 0.7312, "step": 7033 }, { "epoch": 0.84, "learning_rate": 4.228884217529594e-06, "loss": 0.6918, "step": 7034 }, { "epoch": 0.84, "learning_rate": 4.228652031037706e-06, "loss": 0.8556, "step": 7035 }, { "epoch": 0.84, "learning_rate": 4.228419815971054e-06, "loss": 1.0836, "step": 7036 }, { "epoch": 0.84, "learning_rate": 4.228187572333478e-06, "loss": 1.1047, "step": 7037 }, { "epoch": 0.84, "learning_rate": 4.227955300128817e-06, "loss": 0.6487, "step": 7038 }, { "epoch": 0.84, "learning_rate": 4.227722999360909e-06, "loss": 0.655, "step": 7039 }, { "epoch": 0.84, "learning_rate": 4.227490670033595e-06, "loss": 0.7519, "step": 7040 }, { "epoch": 0.84, "learning_rate": 4.227258312150716e-06, "loss": 0.8053, "step": 7041 }, { "epoch": 0.84, "learning_rate": 4.2270259257161116e-06, "loss": 0.858, "step": 7042 }, { "epoch": 0.84, "learning_rate": 4.226793510733623e-06, "loss": 1.1176, "step": 7043 }, { "epoch": 0.84, "learning_rate": 4.226561067207094e-06, "loss": 0.7357, "step": 7044 }, { "epoch": 0.84, "learning_rate": 4.226328595140365e-06, "loss": 0.6388, "step": 7045 }, { "epoch": 0.84, "learning_rate": 4.2260960945372794e-06, "loss": 0.685, "step": 7046 }, { "epoch": 0.84, "learning_rate": 4.225863565401681e-06, "loss": 0.7831, "step": 7047 }, { "epoch": 0.84, "learning_rate": 4.225631007737413e-06, "loss": 0.8235, "step": 7048 }, { "epoch": 0.84, "learning_rate": 4.2253984215483195e-06, "loss": 0.882, "step": 7049 }, { "epoch": 0.84, "learning_rate": 4.225165806838246e-06, "loss": 1.136, "step": 7050 }, { "epoch": 0.84, "learning_rate": 4.224933163611036e-06, "loss": 0.5839, "step": 7051 }, { "epoch": 0.84, "learning_rate": 4.224700491870536e-06, "loss": 0.7036, "step": 7052 }, { "epoch": 0.84, "learning_rate": 4.224467791620594e-06, "loss": 0.7441, "step": 7053 }, { "epoch": 0.84, "learning_rate": 4.2242350628650515e-06, "loss": 0.9072, "step": 7054 }, { "epoch": 0.84, "learning_rate": 4.224002305607761e-06, "loss": 0.8786, "step": 7055 }, { "epoch": 0.84, "learning_rate": 4.223769519852568e-06, "loss": 1.1497, "step": 7056 }, { "epoch": 0.84, "learning_rate": 4.2235367056033195e-06, "loss": 0.637, "step": 7057 }, { "epoch": 0.84, "learning_rate": 4.223303862863864e-06, "loss": 0.6291, "step": 7058 }, { "epoch": 0.84, "learning_rate": 4.223070991638052e-06, "loss": 0.7394, "step": 7059 }, { "epoch": 0.84, "learning_rate": 4.22283809192973e-06, "loss": 0.767, "step": 7060 }, { "epoch": 0.84, "learning_rate": 4.222605163742753e-06, "loss": 0.9698, "step": 7061 }, { "epoch": 0.84, "learning_rate": 4.2223722070809654e-06, "loss": 1.1307, "step": 7062 }, { "epoch": 0.84, "learning_rate": 4.2221392219482225e-06, "loss": 0.6148, "step": 7063 }, { "epoch": 0.84, "learning_rate": 4.2219062083483735e-06, "loss": 0.5966, "step": 7064 }, { "epoch": 0.84, "learning_rate": 4.22167316628527e-06, "loss": 0.6942, "step": 7065 }, { "epoch": 0.84, "learning_rate": 4.2214400957627636e-06, "loss": 0.7169, "step": 7066 }, { "epoch": 0.84, "learning_rate": 4.221206996784709e-06, "loss": 0.8681, "step": 7067 }, { "epoch": 0.84, "learning_rate": 4.220973869354958e-06, "loss": 1.0638, "step": 7068 }, { "epoch": 0.84, "learning_rate": 4.220740713477365e-06, "loss": 0.6742, "step": 7069 }, { "epoch": 0.84, "learning_rate": 4.220507529155784e-06, "loss": 0.6311, "step": 7070 }, { "epoch": 0.84, "learning_rate": 4.220274316394068e-06, "loss": 0.7456, "step": 7071 }, { "epoch": 0.84, "learning_rate": 4.220041075196074e-06, "loss": 0.7643, "step": 7072 }, { "epoch": 0.84, "learning_rate": 4.219807805565656e-06, "loss": 0.8878, "step": 7073 }, { "epoch": 0.84, "learning_rate": 4.2195745075066716e-06, "loss": 1.07, "step": 7074 }, { "epoch": 0.84, "learning_rate": 4.2193411810229755e-06, "loss": 1.2606, "step": 7075 }, { "epoch": 0.84, "learning_rate": 4.2191078261184256e-06, "loss": 0.5686, "step": 7076 }, { "epoch": 0.84, "learning_rate": 4.218874442796879e-06, "loss": 0.8107, "step": 7077 }, { "epoch": 0.84, "learning_rate": 4.218641031062194e-06, "loss": 0.782, "step": 7078 }, { "epoch": 0.84, "learning_rate": 4.2184075909182285e-06, "loss": 0.9251, "step": 7079 }, { "epoch": 0.84, "learning_rate": 4.218174122368841e-06, "loss": 1.0441, "step": 7080 }, { "epoch": 0.84, "learning_rate": 4.2179406254178915e-06, "loss": 1.0623, "step": 7081 }, { "epoch": 0.84, "learning_rate": 4.217707100069239e-06, "loss": 0.6913, "step": 7082 }, { "epoch": 0.84, "learning_rate": 4.217473546326745e-06, "loss": 0.8434, "step": 7083 }, { "epoch": 0.84, "learning_rate": 4.217239964194268e-06, "loss": 0.8802, "step": 7084 }, { "epoch": 0.84, "learning_rate": 4.2170063536756715e-06, "loss": 0.8485, "step": 7085 }, { "epoch": 0.84, "learning_rate": 4.216772714774814e-06, "loss": 0.9908, "step": 7086 }, { "epoch": 0.84, "learning_rate": 4.216539047495561e-06, "loss": 1.2302, "step": 7087 }, { "epoch": 0.84, "learning_rate": 4.216305351841773e-06, "loss": 0.7078, "step": 7088 }, { "epoch": 0.84, "learning_rate": 4.216071627817313e-06, "loss": 0.6512, "step": 7089 }, { "epoch": 0.84, "learning_rate": 4.2158378754260455e-06, "loss": 0.7169, "step": 7090 }, { "epoch": 0.84, "learning_rate": 4.215604094671835e-06, "loss": 0.743, "step": 7091 }, { "epoch": 0.84, "learning_rate": 4.215370285558543e-06, "loss": 0.8836, "step": 7092 }, { "epoch": 0.84, "learning_rate": 4.215136448090038e-06, "loss": 1.2698, "step": 7093 }, { "epoch": 0.84, "learning_rate": 4.214902582270182e-06, "loss": 0.5608, "step": 7094 }, { "epoch": 0.84, "learning_rate": 4.214668688102843e-06, "loss": 0.6964, "step": 7095 }, { "epoch": 0.85, "learning_rate": 4.214434765591887e-06, "loss": 0.7505, "step": 7096 }, { "epoch": 0.85, "learning_rate": 4.214200814741181e-06, "loss": 0.8292, "step": 7097 }, { "epoch": 0.85, "learning_rate": 4.2139668355545905e-06, "loss": 1.0214, "step": 7098 }, { "epoch": 0.85, "learning_rate": 4.2137328280359846e-06, "loss": 1.1054, "step": 7099 }, { "epoch": 0.85, "learning_rate": 4.2134987921892325e-06, "loss": 1.2103, "step": 7100 }, { "epoch": 0.85, "learning_rate": 4.2132647280182e-06, "loss": 0.5622, "step": 7101 }, { "epoch": 0.85, "learning_rate": 4.213030635526759e-06, "loss": 0.7539, "step": 7102 }, { "epoch": 0.85, "learning_rate": 4.2127965147187765e-06, "loss": 0.6509, "step": 7103 }, { "epoch": 0.85, "learning_rate": 4.212562365598125e-06, "loss": 0.9363, "step": 7104 }, { "epoch": 0.85, "learning_rate": 4.2123281881686736e-06, "loss": 0.9586, "step": 7105 }, { "epoch": 0.85, "learning_rate": 4.2120939824342936e-06, "loss": 1.1924, "step": 7106 }, { "epoch": 0.85, "learning_rate": 4.211859748398856e-06, "loss": 0.5121, "step": 7107 }, { "epoch": 0.85, "learning_rate": 4.2116254860662345e-06, "loss": 0.6625, "step": 7108 }, { "epoch": 0.85, "learning_rate": 4.211391195440298e-06, "loss": 0.7496, "step": 7109 }, { "epoch": 0.85, "learning_rate": 4.2111568765249235e-06, "loss": 0.8557, "step": 7110 }, { "epoch": 0.85, "learning_rate": 4.2109225293239804e-06, "loss": 1.0196, "step": 7111 }, { "epoch": 0.85, "learning_rate": 4.210688153841346e-06, "loss": 1.2703, "step": 7112 }, { "epoch": 0.85, "learning_rate": 4.2104537500808916e-06, "loss": 0.8044, "step": 7113 }, { "epoch": 0.85, "learning_rate": 4.210219318046494e-06, "loss": 0.6038, "step": 7114 }, { "epoch": 0.85, "learning_rate": 4.209984857742028e-06, "loss": 0.7947, "step": 7115 }, { "epoch": 0.85, "learning_rate": 4.209750369171368e-06, "loss": 0.7291, "step": 7116 }, { "epoch": 0.85, "learning_rate": 4.209515852338391e-06, "loss": 0.8934, "step": 7117 }, { "epoch": 0.85, "learning_rate": 4.209281307246973e-06, "loss": 1.1348, "step": 7118 }, { "epoch": 0.85, "learning_rate": 4.209046733900993e-06, "loss": 0.599, "step": 7119 }, { "epoch": 0.85, "learning_rate": 4.208812132304326e-06, "loss": 0.5912, "step": 7120 }, { "epoch": 0.85, "learning_rate": 4.208577502460851e-06, "loss": 0.852, "step": 7121 }, { "epoch": 0.85, "learning_rate": 4.208342844374447e-06, "loss": 0.8442, "step": 7122 }, { "epoch": 0.85, "learning_rate": 4.208108158048992e-06, "loss": 0.9493, "step": 7123 }, { "epoch": 0.85, "learning_rate": 4.2078734434883665e-06, "loss": 1.1153, "step": 7124 }, { "epoch": 0.85, "learning_rate": 4.2076387006964495e-06, "loss": 1.1537, "step": 7125 }, { "epoch": 0.85, "learning_rate": 4.207403929677121e-06, "loss": 0.6057, "step": 7126 }, { "epoch": 0.85, "learning_rate": 4.2071691304342625e-06, "loss": 0.8181, "step": 7127 }, { "epoch": 0.85, "learning_rate": 4.206934302971755e-06, "loss": 0.7501, "step": 7128 }, { "epoch": 0.85, "learning_rate": 4.206699447293481e-06, "loss": 0.7441, "step": 7129 }, { "epoch": 0.85, "learning_rate": 4.2064645634033204e-06, "loss": 0.9583, "step": 7130 }, { "epoch": 0.85, "learning_rate": 4.206229651305158e-06, "loss": 1.1494, "step": 7131 }, { "epoch": 0.85, "learning_rate": 4.205994711002877e-06, "loss": 0.5139, "step": 7132 }, { "epoch": 0.85, "learning_rate": 4.20575974250036e-06, "loss": 0.688, "step": 7133 }, { "epoch": 0.85, "learning_rate": 4.2055247458014915e-06, "loss": 0.7877, "step": 7134 }, { "epoch": 0.85, "learning_rate": 4.205289720910155e-06, "loss": 0.9684, "step": 7135 }, { "epoch": 0.85, "learning_rate": 4.205054667830237e-06, "loss": 1.0933, "step": 7136 }, { "epoch": 0.85, "learning_rate": 4.204819586565622e-06, "loss": 1.1132, "step": 7137 }, { "epoch": 0.85, "learning_rate": 4.204584477120196e-06, "loss": 0.7042, "step": 7138 }, { "epoch": 0.85, "learning_rate": 4.204349339497846e-06, "loss": 0.7148, "step": 7139 }, { "epoch": 0.85, "learning_rate": 4.204114173702459e-06, "loss": 0.6912, "step": 7140 }, { "epoch": 0.85, "learning_rate": 4.2038789797379205e-06, "loss": 0.7744, "step": 7141 }, { "epoch": 0.85, "learning_rate": 4.20364375760812e-06, "loss": 0.9207, "step": 7142 }, { "epoch": 0.85, "learning_rate": 4.203408507316945e-06, "loss": 1.0846, "step": 7143 }, { "epoch": 0.85, "learning_rate": 4.203173228868285e-06, "loss": 0.9198, "step": 7144 }, { "epoch": 0.85, "learning_rate": 4.202937922266029e-06, "loss": 0.4843, "step": 7145 }, { "epoch": 0.85, "learning_rate": 4.202702587514065e-06, "loss": 0.6862, "step": 7146 }, { "epoch": 0.85, "learning_rate": 4.202467224616285e-06, "loss": 0.6689, "step": 7147 }, { "epoch": 0.85, "learning_rate": 4.202231833576579e-06, "loss": 0.9896, "step": 7148 }, { "epoch": 0.85, "learning_rate": 4.201996414398837e-06, "loss": 1.0706, "step": 7149 }, { "epoch": 0.85, "learning_rate": 4.201760967086953e-06, "loss": 1.053, "step": 7150 }, { "epoch": 0.85, "learning_rate": 4.201525491644817e-06, "loss": 0.5465, "step": 7151 }, { "epoch": 0.85, "learning_rate": 4.201289988076321e-06, "loss": 0.709, "step": 7152 }, { "epoch": 0.85, "learning_rate": 4.20105445638536e-06, "loss": 0.789, "step": 7153 }, { "epoch": 0.85, "learning_rate": 4.2008188965758244e-06, "loss": 0.7395, "step": 7154 }, { "epoch": 0.85, "learning_rate": 4.200583308651611e-06, "loss": 1.1064, "step": 7155 }, { "epoch": 0.85, "learning_rate": 4.200347692616612e-06, "loss": 1.1292, "step": 7156 }, { "epoch": 0.85, "learning_rate": 4.200112048474724e-06, "loss": 0.4881, "step": 7157 }, { "epoch": 0.85, "learning_rate": 4.19987637622984e-06, "loss": 0.6357, "step": 7158 }, { "epoch": 0.85, "learning_rate": 4.199640675885858e-06, "loss": 0.7471, "step": 7159 }, { "epoch": 0.85, "learning_rate": 4.199404947446673e-06, "loss": 0.7399, "step": 7160 }, { "epoch": 0.85, "learning_rate": 4.199169190916181e-06, "loss": 1.0961, "step": 7161 }, { "epoch": 0.85, "learning_rate": 4.198933406298279e-06, "loss": 1.0853, "step": 7162 }, { "epoch": 0.85, "learning_rate": 4.198697593596867e-06, "loss": 0.5836, "step": 7163 }, { "epoch": 0.85, "learning_rate": 4.198461752815841e-06, "loss": 0.6571, "step": 7164 }, { "epoch": 0.85, "learning_rate": 4.198225883959099e-06, "loss": 0.7677, "step": 7165 }, { "epoch": 0.85, "learning_rate": 4.197989987030541e-06, "loss": 0.8381, "step": 7166 }, { "epoch": 0.85, "learning_rate": 4.197754062034065e-06, "loss": 0.9798, "step": 7167 }, { "epoch": 0.85, "learning_rate": 4.197518108973573e-06, "loss": 1.0261, "step": 7168 }, { "epoch": 0.85, "learning_rate": 4.1972821278529635e-06, "loss": 0.7634, "step": 7169 }, { "epoch": 0.85, "learning_rate": 4.197046118676139e-06, "loss": 0.5687, "step": 7170 }, { "epoch": 0.85, "learning_rate": 4.196810081446999e-06, "loss": 0.7908, "step": 7171 }, { "epoch": 0.85, "learning_rate": 4.196574016169446e-06, "loss": 0.7135, "step": 7172 }, { "epoch": 0.85, "learning_rate": 4.196337922847382e-06, "loss": 0.8986, "step": 7173 }, { "epoch": 0.85, "learning_rate": 4.196101801484711e-06, "loss": 0.9976, "step": 7174 }, { "epoch": 0.85, "learning_rate": 4.195865652085334e-06, "loss": 1.1169, "step": 7175 }, { "epoch": 0.85, "learning_rate": 4.195629474653155e-06, "loss": 0.5829, "step": 7176 }, { "epoch": 0.85, "learning_rate": 4.195393269192079e-06, "loss": 0.7663, "step": 7177 }, { "epoch": 0.85, "learning_rate": 4.19515703570601e-06, "loss": 0.6951, "step": 7178 }, { "epoch": 0.85, "learning_rate": 4.194920774198853e-06, "loss": 0.8954, "step": 7179 }, { "epoch": 0.86, "learning_rate": 4.194684484674514e-06, "loss": 0.9714, "step": 7180 }, { "epoch": 0.86, "learning_rate": 4.194448167136897e-06, "loss": 1.084, "step": 7181 }, { "epoch": 0.86, "learning_rate": 4.194211821589911e-06, "loss": 0.6289, "step": 7182 }, { "epoch": 0.86, "learning_rate": 4.193975448037461e-06, "loss": 0.755, "step": 7183 }, { "epoch": 0.86, "learning_rate": 4.193739046483454e-06, "loss": 0.6619, "step": 7184 }, { "epoch": 0.86, "learning_rate": 4.193502616931799e-06, "loss": 0.8414, "step": 7185 }, { "epoch": 0.86, "learning_rate": 4.193266159386404e-06, "loss": 0.9775, "step": 7186 }, { "epoch": 0.86, "learning_rate": 4.1930296738511765e-06, "loss": 1.2035, "step": 7187 }, { "epoch": 0.86, "learning_rate": 4.1927931603300266e-06, "loss": 0.7705, "step": 7188 }, { "epoch": 0.86, "learning_rate": 4.192556618826864e-06, "loss": 0.4731, "step": 7189 }, { "epoch": 0.86, "learning_rate": 4.192320049345598e-06, "loss": 0.6789, "step": 7190 }, { "epoch": 0.86, "learning_rate": 4.19208345189014e-06, "loss": 0.8726, "step": 7191 }, { "epoch": 0.86, "learning_rate": 4.1918468264644e-06, "loss": 1.0736, "step": 7192 }, { "epoch": 0.86, "learning_rate": 4.191610173072291e-06, "loss": 1.101, "step": 7193 }, { "epoch": 0.86, "learning_rate": 4.191373491717723e-06, "loss": 0.7133, "step": 7194 }, { "epoch": 0.86, "learning_rate": 4.19113678240461e-06, "loss": 0.6503, "step": 7195 }, { "epoch": 0.86, "learning_rate": 4.1909000451368645e-06, "loss": 0.6679, "step": 7196 }, { "epoch": 0.86, "learning_rate": 4.190663279918398e-06, "loss": 0.7521, "step": 7197 }, { "epoch": 0.86, "learning_rate": 4.190426486753126e-06, "loss": 0.8584, "step": 7198 }, { "epoch": 0.86, "learning_rate": 4.1901896656449625e-06, "loss": 1.0102, "step": 7199 }, { "epoch": 0.86, "learning_rate": 4.189952816597822e-06, "loss": 1.358, "step": 7200 }, { "epoch": 0.86, "learning_rate": 4.18971593961562e-06, "loss": 0.5829, "step": 7201 }, { "epoch": 0.86, "learning_rate": 4.189479034702272e-06, "loss": 0.7463, "step": 7202 }, { "epoch": 0.86, "learning_rate": 4.189242101861692e-06, "loss": 0.7336, "step": 7203 }, { "epoch": 0.86, "learning_rate": 4.189005141097801e-06, "loss": 0.8243, "step": 7204 }, { "epoch": 0.86, "learning_rate": 4.188768152414511e-06, "loss": 0.9689, "step": 7205 }, { "epoch": 0.86, "learning_rate": 4.188531135815743e-06, "loss": 1.2422, "step": 7206 }, { "epoch": 0.86, "learning_rate": 4.188294091305413e-06, "loss": 0.6207, "step": 7207 }, { "epoch": 0.86, "learning_rate": 4.18805701888744e-06, "loss": 0.6204, "step": 7208 }, { "epoch": 0.86, "learning_rate": 4.187819918565744e-06, "loss": 0.7434, "step": 7209 }, { "epoch": 0.86, "learning_rate": 4.187582790344243e-06, "loss": 0.7904, "step": 7210 }, { "epoch": 0.86, "learning_rate": 4.187345634226856e-06, "loss": 0.9862, "step": 7211 }, { "epoch": 0.86, "learning_rate": 4.187108450217503e-06, "loss": 1.0944, "step": 7212 }, { "epoch": 0.86, "learning_rate": 4.186871238320108e-06, "loss": 0.5237, "step": 7213 }, { "epoch": 0.86, "learning_rate": 4.186633998538588e-06, "loss": 0.6383, "step": 7214 }, { "epoch": 0.86, "learning_rate": 4.186396730876868e-06, "loss": 0.7516, "step": 7215 }, { "epoch": 0.86, "learning_rate": 4.1861594353388676e-06, "loss": 0.7998, "step": 7216 }, { "epoch": 0.86, "learning_rate": 4.185922111928511e-06, "loss": 0.9113, "step": 7217 }, { "epoch": 0.86, "learning_rate": 4.185684760649719e-06, "loss": 1.2545, "step": 7218 }, { "epoch": 0.86, "learning_rate": 4.185447381506418e-06, "loss": 0.7523, "step": 7219 }, { "epoch": 0.86, "learning_rate": 4.1852099745025296e-06, "loss": 0.5693, "step": 7220 }, { "epoch": 0.86, "learning_rate": 4.184972539641979e-06, "loss": 0.757, "step": 7221 }, { "epoch": 0.86, "learning_rate": 4.184735076928691e-06, "loss": 0.6838, "step": 7222 }, { "epoch": 0.86, "learning_rate": 4.1844975863665915e-06, "loss": 0.8801, "step": 7223 }, { "epoch": 0.86, "learning_rate": 4.1842600679596045e-06, "loss": 1.0143, "step": 7224 }, { "epoch": 0.86, "learning_rate": 4.184022521711658e-06, "loss": 1.0294, "step": 7225 }, { "epoch": 0.86, "learning_rate": 4.1837849476266775e-06, "loss": 0.5982, "step": 7226 }, { "epoch": 0.86, "learning_rate": 4.1835473457085904e-06, "loss": 0.7388, "step": 7227 }, { "epoch": 0.86, "learning_rate": 4.183309715961325e-06, "loss": 0.6941, "step": 7228 }, { "epoch": 0.86, "learning_rate": 4.183072058388808e-06, "loss": 0.8464, "step": 7229 }, { "epoch": 0.86, "learning_rate": 4.1828343729949695e-06, "loss": 1.0359, "step": 7230 }, { "epoch": 0.86, "learning_rate": 4.182596659783737e-06, "loss": 1.0933, "step": 7231 }, { "epoch": 0.86, "learning_rate": 4.182358918759041e-06, "loss": 0.5638, "step": 7232 }, { "epoch": 0.86, "learning_rate": 4.182121149924811e-06, "loss": 0.7628, "step": 7233 }, { "epoch": 0.86, "learning_rate": 4.181883353284976e-06, "loss": 0.7182, "step": 7234 }, { "epoch": 0.86, "learning_rate": 4.1816455288434685e-06, "loss": 0.8384, "step": 7235 }, { "epoch": 0.86, "learning_rate": 4.1814076766042206e-06, "loss": 0.9344, "step": 7236 }, { "epoch": 0.86, "learning_rate": 4.181169796571162e-06, "loss": 1.1712, "step": 7237 }, { "epoch": 0.86, "learning_rate": 4.180931888748225e-06, "loss": 0.497, "step": 7238 }, { "epoch": 0.86, "learning_rate": 4.180693953139343e-06, "loss": 0.5833, "step": 7239 }, { "epoch": 0.86, "learning_rate": 4.18045598974845e-06, "loss": 0.7892, "step": 7240 }, { "epoch": 0.86, "learning_rate": 4.180217998579477e-06, "loss": 0.8646, "step": 7241 }, { "epoch": 0.86, "learning_rate": 4.179979979636361e-06, "loss": 0.8643, "step": 7242 }, { "epoch": 0.86, "learning_rate": 4.179741932923034e-06, "loss": 1.0197, "step": 7243 }, { "epoch": 0.86, "learning_rate": 4.1795038584434325e-06, "loss": 0.7654, "step": 7244 }, { "epoch": 0.86, "learning_rate": 4.17926575620149e-06, "loss": 0.5819, "step": 7245 }, { "epoch": 0.86, "learning_rate": 4.179027626201145e-06, "loss": 0.7919, "step": 7246 }, { "epoch": 0.86, "learning_rate": 4.178789468446332e-06, "loss": 0.6945, "step": 7247 }, { "epoch": 0.86, "learning_rate": 4.178551282940988e-06, "loss": 0.9219, "step": 7248 }, { "epoch": 0.86, "learning_rate": 4.1783130696890505e-06, "loss": 0.9319, "step": 7249 }, { "epoch": 0.86, "learning_rate": 4.178074828694457e-06, "loss": 1.2268, "step": 7250 }, { "epoch": 0.86, "learning_rate": 4.177836559961147e-06, "loss": 0.5535, "step": 7251 }, { "epoch": 0.86, "learning_rate": 4.177598263493057e-06, "loss": 0.6301, "step": 7252 }, { "epoch": 0.86, "learning_rate": 4.177359939294127e-06, "loss": 0.8097, "step": 7253 }, { "epoch": 0.86, "learning_rate": 4.177121587368296e-06, "loss": 0.782, "step": 7254 }, { "epoch": 0.86, "learning_rate": 4.176883207719504e-06, "loss": 0.9323, "step": 7255 }, { "epoch": 0.86, "learning_rate": 4.176644800351694e-06, "loss": 1.0393, "step": 7256 }, { "epoch": 0.86, "learning_rate": 4.176406365268802e-06, "loss": 0.6233, "step": 7257 }, { "epoch": 0.86, "learning_rate": 4.176167902474774e-06, "loss": 0.7326, "step": 7258 }, { "epoch": 0.86, "learning_rate": 4.17592941197355e-06, "loss": 0.5587, "step": 7259 }, { "epoch": 0.86, "learning_rate": 4.1756908937690725e-06, "loss": 0.6289, "step": 7260 }, { "epoch": 0.86, "learning_rate": 4.1754523478652816e-06, "loss": 0.9072, "step": 7261 }, { "epoch": 0.86, "learning_rate": 4.175213774266125e-06, "loss": 1.1295, "step": 7262 }, { "epoch": 0.86, "learning_rate": 4.174975172975543e-06, "loss": 0.7003, "step": 7263 }, { "epoch": 0.87, "learning_rate": 4.174736543997482e-06, "loss": 0.6109, "step": 7264 }, { "epoch": 0.87, "learning_rate": 4.174497887335884e-06, "loss": 0.8014, "step": 7265 }, { "epoch": 0.87, "learning_rate": 4.174259202994696e-06, "loss": 0.7126, "step": 7266 }, { "epoch": 0.87, "learning_rate": 4.174020490977864e-06, "loss": 0.8822, "step": 7267 }, { "epoch": 0.87, "learning_rate": 4.173781751289332e-06, "loss": 1.1029, "step": 7268 }, { "epoch": 0.87, "learning_rate": 4.173542983933046e-06, "loss": 0.6414, "step": 7269 }, { "epoch": 0.87, "learning_rate": 4.173304188912956e-06, "loss": 0.5228, "step": 7270 }, { "epoch": 0.87, "learning_rate": 4.173065366233006e-06, "loss": 0.7251, "step": 7271 }, { "epoch": 0.87, "learning_rate": 4.172826515897146e-06, "loss": 0.7186, "step": 7272 }, { "epoch": 0.87, "learning_rate": 4.1725876379093225e-06, "loss": 0.8027, "step": 7273 }, { "epoch": 0.87, "learning_rate": 4.172348732273486e-06, "loss": 1.163, "step": 7274 }, { "epoch": 0.87, "learning_rate": 4.1721097989935845e-06, "loss": 1.3757, "step": 7275 }, { "epoch": 0.87, "learning_rate": 4.171870838073567e-06, "loss": 0.6767, "step": 7276 }, { "epoch": 0.87, "learning_rate": 4.1716318495173845e-06, "loss": 0.5923, "step": 7277 }, { "epoch": 0.87, "learning_rate": 4.171392833328987e-06, "loss": 0.7256, "step": 7278 }, { "epoch": 0.87, "learning_rate": 4.171153789512327e-06, "loss": 0.7579, "step": 7279 }, { "epoch": 0.87, "learning_rate": 4.170914718071353e-06, "loss": 0.9656, "step": 7280 }, { "epoch": 0.87, "learning_rate": 4.17067561901002e-06, "loss": 1.0573, "step": 7281 }, { "epoch": 0.87, "learning_rate": 4.170436492332277e-06, "loss": 0.6446, "step": 7282 }, { "epoch": 0.87, "learning_rate": 4.17019733804208e-06, "loss": 0.7052, "step": 7283 }, { "epoch": 0.87, "learning_rate": 4.16995815614338e-06, "loss": 0.7839, "step": 7284 }, { "epoch": 0.87, "learning_rate": 4.1697189466401324e-06, "loss": 0.7862, "step": 7285 }, { "epoch": 0.87, "learning_rate": 4.16947970953629e-06, "loss": 0.9712, "step": 7286 }, { "epoch": 0.87, "learning_rate": 4.1692404448358085e-06, "loss": 1.0937, "step": 7287 }, { "epoch": 0.87, "learning_rate": 4.169001152542642e-06, "loss": 0.6682, "step": 7288 }, { "epoch": 0.87, "learning_rate": 4.168761832660746e-06, "loss": 0.6009, "step": 7289 }, { "epoch": 0.87, "learning_rate": 4.168522485194077e-06, "loss": 0.6793, "step": 7290 }, { "epoch": 0.87, "learning_rate": 4.168283110146591e-06, "loss": 0.7806, "step": 7291 }, { "epoch": 0.87, "learning_rate": 4.168043707522247e-06, "loss": 0.8519, "step": 7292 }, { "epoch": 0.87, "learning_rate": 4.167804277324999e-06, "loss": 1.0978, "step": 7293 }, { "epoch": 0.87, "learning_rate": 4.167564819558807e-06, "loss": 0.8307, "step": 7294 }, { "epoch": 0.87, "learning_rate": 4.167325334227628e-06, "loss": 0.497, "step": 7295 }, { "epoch": 0.87, "learning_rate": 4.167085821335422e-06, "loss": 0.6092, "step": 7296 }, { "epoch": 0.87, "learning_rate": 4.166846280886147e-06, "loss": 0.8367, "step": 7297 }, { "epoch": 0.87, "learning_rate": 4.1666067128837635e-06, "loss": 0.9105, "step": 7298 }, { "epoch": 0.87, "learning_rate": 4.1663671173322305e-06, "loss": 1.0829, "step": 7299 }, { "epoch": 0.87, "learning_rate": 4.16612749423551e-06, "loss": 1.0839, "step": 7300 }, { "epoch": 0.87, "learning_rate": 4.165887843597562e-06, "loss": 0.5813, "step": 7301 }, { "epoch": 0.87, "learning_rate": 4.165648165422348e-06, "loss": 0.6852, "step": 7302 }, { "epoch": 0.87, "learning_rate": 4.16540845971383e-06, "loss": 0.7961, "step": 7303 }, { "epoch": 0.87, "learning_rate": 4.165168726475971e-06, "loss": 0.9528, "step": 7304 }, { "epoch": 0.87, "learning_rate": 4.164928965712732e-06, "loss": 0.9638, "step": 7305 }, { "epoch": 0.87, "learning_rate": 4.164689177428079e-06, "loss": 1.1339, "step": 7306 }, { "epoch": 0.87, "learning_rate": 4.1644493616259736e-06, "loss": 0.6017, "step": 7307 }, { "epoch": 0.87, "learning_rate": 4.164209518310381e-06, "loss": 0.6932, "step": 7308 }, { "epoch": 0.87, "learning_rate": 4.163969647485264e-06, "loss": 0.6388, "step": 7309 }, { "epoch": 0.87, "learning_rate": 4.16372974915459e-06, "loss": 0.8557, "step": 7310 }, { "epoch": 0.87, "learning_rate": 4.163489823322324e-06, "loss": 0.9825, "step": 7311 }, { "epoch": 0.87, "learning_rate": 4.163249869992431e-06, "loss": 1.1594, "step": 7312 }, { "epoch": 0.87, "learning_rate": 4.1630098891688784e-06, "loss": 0.6898, "step": 7313 }, { "epoch": 0.87, "learning_rate": 4.162769880855633e-06, "loss": 0.6758, "step": 7314 }, { "epoch": 0.87, "learning_rate": 4.162529845056662e-06, "loss": 0.6644, "step": 7315 }, { "epoch": 0.87, "learning_rate": 4.162289781775933e-06, "loss": 0.8192, "step": 7316 }, { "epoch": 0.87, "learning_rate": 4.162049691017414e-06, "loss": 0.7686, "step": 7317 }, { "epoch": 0.87, "learning_rate": 4.161809572785075e-06, "loss": 1.0596, "step": 7318 }, { "epoch": 0.87, "learning_rate": 4.161569427082883e-06, "loss": 0.9164, "step": 7319 }, { "epoch": 0.87, "learning_rate": 4.16132925391481e-06, "loss": 0.5238, "step": 7320 }, { "epoch": 0.87, "learning_rate": 4.161089053284825e-06, "loss": 0.7622, "step": 7321 }, { "epoch": 0.87, "learning_rate": 4.160848825196898e-06, "loss": 0.7521, "step": 7322 }, { "epoch": 0.87, "learning_rate": 4.160608569655002e-06, "loss": 0.9004, "step": 7323 }, { "epoch": 0.87, "learning_rate": 4.160368286663106e-06, "loss": 1.0993, "step": 7324 }, { "epoch": 0.87, "learning_rate": 4.160127976225182e-06, "loss": 1.1616, "step": 7325 }, { "epoch": 0.87, "learning_rate": 4.159887638345205e-06, "loss": 0.567, "step": 7326 }, { "epoch": 0.87, "learning_rate": 4.159647273027145e-06, "loss": 0.6533, "step": 7327 }, { "epoch": 0.87, "learning_rate": 4.1594068802749764e-06, "loss": 0.7806, "step": 7328 }, { "epoch": 0.87, "learning_rate": 4.159166460092673e-06, "loss": 0.9044, "step": 7329 }, { "epoch": 0.87, "learning_rate": 4.158926012484208e-06, "loss": 1.0295, "step": 7330 }, { "epoch": 0.87, "learning_rate": 4.158685537453558e-06, "loss": 1.2091, "step": 7331 }, { "epoch": 0.87, "learning_rate": 4.158445035004697e-06, "loss": 0.4783, "step": 7332 }, { "epoch": 0.87, "learning_rate": 4.158204505141599e-06, "loss": 0.6876, "step": 7333 }, { "epoch": 0.87, "learning_rate": 4.157963947868242e-06, "loss": 0.7133, "step": 7334 }, { "epoch": 0.87, "learning_rate": 4.157723363188603e-06, "loss": 0.8195, "step": 7335 }, { "epoch": 0.87, "learning_rate": 4.157482751106657e-06, "loss": 0.9371, "step": 7336 }, { "epoch": 0.87, "learning_rate": 4.1572421116263815e-06, "loss": 1.1188, "step": 7337 }, { "epoch": 0.87, "learning_rate": 4.157001444751755e-06, "loss": 0.6992, "step": 7338 }, { "epoch": 0.87, "learning_rate": 4.156760750486756e-06, "loss": 0.6259, "step": 7339 }, { "epoch": 0.87, "learning_rate": 4.156520028835363e-06, "loss": 0.8048, "step": 7340 }, { "epoch": 0.87, "learning_rate": 4.156279279801554e-06, "loss": 0.8519, "step": 7341 }, { "epoch": 0.87, "learning_rate": 4.15603850338931e-06, "loss": 0.8254, "step": 7342 }, { "epoch": 0.87, "learning_rate": 4.155797699602611e-06, "loss": 1.0246, "step": 7343 }, { "epoch": 0.87, "learning_rate": 4.155556868445436e-06, "loss": 0.8801, "step": 7344 }, { "epoch": 0.87, "learning_rate": 4.155316009921768e-06, "loss": 0.6715, "step": 7345 }, { "epoch": 0.87, "learning_rate": 4.155075124035587e-06, "loss": 0.7901, "step": 7346 }, { "epoch": 0.87, "learning_rate": 4.154834210790875e-06, "loss": 0.8359, "step": 7347 }, { "epoch": 0.88, "learning_rate": 4.154593270191616e-06, "loss": 0.7837, "step": 7348 }, { "epoch": 0.88, "learning_rate": 4.154352302241791e-06, "loss": 1.0223, "step": 7349 }, { "epoch": 0.88, "learning_rate": 4.1541113069453835e-06, "loss": 1.1681, "step": 7350 }, { "epoch": 0.88, "learning_rate": 4.153870284306376e-06, "loss": 0.6847, "step": 7351 }, { "epoch": 0.88, "learning_rate": 4.153629234328756e-06, "loss": 0.7256, "step": 7352 }, { "epoch": 0.88, "learning_rate": 4.1533881570165045e-06, "loss": 0.8258, "step": 7353 }, { "epoch": 0.88, "learning_rate": 4.1531470523736085e-06, "loss": 0.7866, "step": 7354 }, { "epoch": 0.88, "learning_rate": 4.1529059204040525e-06, "loss": 0.9877, "step": 7355 }, { "epoch": 0.88, "learning_rate": 4.152664761111824e-06, "loss": 0.9755, "step": 7356 }, { "epoch": 0.88, "learning_rate": 4.152423574500908e-06, "loss": 0.646, "step": 7357 }, { "epoch": 0.88, "learning_rate": 4.152182360575291e-06, "loss": 0.7057, "step": 7358 }, { "epoch": 0.88, "learning_rate": 4.151941119338961e-06, "loss": 0.6491, "step": 7359 }, { "epoch": 0.88, "learning_rate": 4.1516998507959065e-06, "loss": 0.8377, "step": 7360 }, { "epoch": 0.88, "learning_rate": 4.1514585549501145e-06, "loss": 0.8065, "step": 7361 }, { "epoch": 0.88, "learning_rate": 4.151217231805575e-06, "loss": 1.2143, "step": 7362 }, { "epoch": 0.88, "learning_rate": 4.150975881366274e-06, "loss": 0.614, "step": 7363 }, { "epoch": 0.88, "learning_rate": 4.150734503636204e-06, "loss": 0.5314, "step": 7364 }, { "epoch": 0.88, "learning_rate": 4.150493098619354e-06, "loss": 0.6606, "step": 7365 }, { "epoch": 0.88, "learning_rate": 4.150251666319716e-06, "loss": 0.7203, "step": 7366 }, { "epoch": 0.88, "learning_rate": 4.150010206741278e-06, "loss": 0.9323, "step": 7367 }, { "epoch": 0.88, "learning_rate": 4.149768719888033e-06, "loss": 1.0681, "step": 7368 }, { "epoch": 0.88, "learning_rate": 4.149527205763974e-06, "loss": 0.7893, "step": 7369 }, { "epoch": 0.88, "learning_rate": 4.149285664373091e-06, "loss": 0.595, "step": 7370 }, { "epoch": 0.88, "learning_rate": 4.149044095719377e-06, "loss": 0.652, "step": 7371 }, { "epoch": 0.88, "learning_rate": 4.148802499806826e-06, "loss": 0.8307, "step": 7372 }, { "epoch": 0.88, "learning_rate": 4.1485608766394315e-06, "loss": 0.913, "step": 7373 }, { "epoch": 0.88, "learning_rate": 4.148319226221187e-06, "loss": 1.1088, "step": 7374 }, { "epoch": 0.88, "learning_rate": 4.148077548556088e-06, "loss": 1.1145, "step": 7375 }, { "epoch": 0.88, "learning_rate": 4.1478358436481285e-06, "loss": 0.5427, "step": 7376 }, { "epoch": 0.88, "learning_rate": 4.147594111501304e-06, "loss": 0.6289, "step": 7377 }, { "epoch": 0.88, "learning_rate": 4.1473523521196115e-06, "loss": 0.6976, "step": 7378 }, { "epoch": 0.88, "learning_rate": 4.1471105655070455e-06, "loss": 0.9385, "step": 7379 }, { "epoch": 0.88, "learning_rate": 4.146868751667604e-06, "loss": 0.9059, "step": 7380 }, { "epoch": 0.88, "learning_rate": 4.146626910605284e-06, "loss": 1.027, "step": 7381 }, { "epoch": 0.88, "learning_rate": 4.146385042324083e-06, "loss": 0.4951, "step": 7382 }, { "epoch": 0.88, "learning_rate": 4.146143146827999e-06, "loss": 0.605, "step": 7383 }, { "epoch": 0.88, "learning_rate": 4.145901224121031e-06, "loss": 0.7142, "step": 7384 }, { "epoch": 0.88, "learning_rate": 4.145659274207177e-06, "loss": 0.7661, "step": 7385 }, { "epoch": 0.88, "learning_rate": 4.145417297090437e-06, "loss": 1.0106, "step": 7386 }, { "epoch": 0.88, "learning_rate": 4.145175292774811e-06, "loss": 1.2902, "step": 7387 }, { "epoch": 0.88, "learning_rate": 4.1449332612643e-06, "loss": 0.6733, "step": 7388 }, { "epoch": 0.88, "learning_rate": 4.144691202562904e-06, "loss": 0.7151, "step": 7389 }, { "epoch": 0.88, "learning_rate": 4.144449116674625e-06, "loss": 0.6412, "step": 7390 }, { "epoch": 0.88, "learning_rate": 4.144207003603463e-06, "loss": 0.807, "step": 7391 }, { "epoch": 0.88, "learning_rate": 4.143964863353422e-06, "loss": 0.9757, "step": 7392 }, { "epoch": 0.88, "learning_rate": 4.143722695928503e-06, "loss": 1.1646, "step": 7393 }, { "epoch": 0.88, "learning_rate": 4.143480501332711e-06, "loss": 0.7507, "step": 7394 }, { "epoch": 0.88, "learning_rate": 4.143238279570047e-06, "loss": 0.5189, "step": 7395 }, { "epoch": 0.88, "learning_rate": 4.142996030644518e-06, "loss": 0.738, "step": 7396 }, { "epoch": 0.88, "learning_rate": 4.142753754560126e-06, "loss": 0.7454, "step": 7397 }, { "epoch": 0.88, "learning_rate": 4.142511451320876e-06, "loss": 0.8247, "step": 7398 }, { "epoch": 0.88, "learning_rate": 4.142269120930775e-06, "loss": 0.8886, "step": 7399 }, { "epoch": 0.88, "learning_rate": 4.142026763393827e-06, "loss": 1.1802, "step": 7400 }, { "epoch": 0.88, "learning_rate": 4.141784378714039e-06, "loss": 0.5688, "step": 7401 }, { "epoch": 0.88, "learning_rate": 4.141541966895418e-06, "loss": 0.5546, "step": 7402 }, { "epoch": 0.88, "learning_rate": 4.14129952794197e-06, "loss": 0.5733, "step": 7403 }, { "epoch": 0.88, "learning_rate": 4.141057061857703e-06, "loss": 0.9108, "step": 7404 }, { "epoch": 0.88, "learning_rate": 4.140814568646625e-06, "loss": 1.0319, "step": 7405 }, { "epoch": 0.88, "learning_rate": 4.140572048312744e-06, "loss": 1.172, "step": 7406 }, { "epoch": 0.88, "learning_rate": 4.140329500860071e-06, "loss": 0.7192, "step": 7407 }, { "epoch": 0.88, "learning_rate": 4.140086926292612e-06, "loss": 0.7664, "step": 7408 }, { "epoch": 0.88, "learning_rate": 4.139844324614379e-06, "loss": 0.5546, "step": 7409 }, { "epoch": 0.88, "learning_rate": 4.1396016958293825e-06, "loss": 0.8075, "step": 7410 }, { "epoch": 0.88, "learning_rate": 4.139359039941632e-06, "loss": 1.0184, "step": 7411 }, { "epoch": 0.88, "learning_rate": 4.1391163569551384e-06, "loss": 1.0953, "step": 7412 }, { "epoch": 0.88, "learning_rate": 4.138873646873915e-06, "loss": 0.6674, "step": 7413 }, { "epoch": 0.88, "learning_rate": 4.138630909701972e-06, "loss": 0.7207, "step": 7414 }, { "epoch": 0.88, "learning_rate": 4.138388145443323e-06, "loss": 0.6776, "step": 7415 }, { "epoch": 0.88, "learning_rate": 4.13814535410198e-06, "loss": 0.7247, "step": 7416 }, { "epoch": 0.88, "learning_rate": 4.137902535681957e-06, "loss": 0.9265, "step": 7417 }, { "epoch": 0.88, "learning_rate": 4.1376596901872674e-06, "loss": 1.0381, "step": 7418 }, { "epoch": 0.88, "learning_rate": 4.137416817621925e-06, "loss": 0.8307, "step": 7419 }, { "epoch": 0.88, "learning_rate": 4.137173917989947e-06, "loss": 0.5825, "step": 7420 }, { "epoch": 0.88, "learning_rate": 4.136930991295346e-06, "loss": 0.6204, "step": 7421 }, { "epoch": 0.88, "learning_rate": 4.136688037542138e-06, "loss": 0.7377, "step": 7422 }, { "epoch": 0.88, "learning_rate": 4.136445056734339e-06, "loss": 0.9382, "step": 7423 }, { "epoch": 0.88, "learning_rate": 4.136202048875968e-06, "loss": 0.9879, "step": 7424 }, { "epoch": 0.88, "learning_rate": 4.135959013971038e-06, "loss": 1.1292, "step": 7425 }, { "epoch": 0.88, "learning_rate": 4.135715952023569e-06, "loss": 0.6667, "step": 7426 }, { "epoch": 0.88, "learning_rate": 4.135472863037578e-06, "loss": 0.7434, "step": 7427 }, { "epoch": 0.88, "learning_rate": 4.135229747017082e-06, "loss": 0.6452, "step": 7428 }, { "epoch": 0.88, "learning_rate": 4.134986603966102e-06, "loss": 0.7371, "step": 7429 }, { "epoch": 0.88, "learning_rate": 4.134743433888657e-06, "loss": 0.9767, "step": 7430 }, { "epoch": 0.88, "learning_rate": 4.134500236788765e-06, "loss": 1.1578, "step": 7431 }, { "epoch": 0.89, "learning_rate": 4.134257012670448e-06, "loss": 0.5576, "step": 7432 }, { "epoch": 0.89, "learning_rate": 4.1340137615377244e-06, "loss": 0.5123, "step": 7433 }, { "epoch": 0.89, "learning_rate": 4.133770483394616e-06, "loss": 0.7577, "step": 7434 }, { "epoch": 0.89, "learning_rate": 4.133527178245145e-06, "loss": 0.6942, "step": 7435 }, { "epoch": 0.89, "learning_rate": 4.133283846093333e-06, "loss": 0.8702, "step": 7436 }, { "epoch": 0.89, "learning_rate": 4.1330404869432014e-06, "loss": 1.0833, "step": 7437 }, { "epoch": 0.89, "learning_rate": 4.132797100798773e-06, "loss": 0.6299, "step": 7438 }, { "epoch": 0.89, "learning_rate": 4.132553687664072e-06, "loss": 0.6861, "step": 7439 }, { "epoch": 0.89, "learning_rate": 4.132310247543122e-06, "loss": 0.7803, "step": 7440 }, { "epoch": 0.89, "learning_rate": 4.132066780439947e-06, "loss": 0.7496, "step": 7441 }, { "epoch": 0.89, "learning_rate": 4.13182328635857e-06, "loss": 0.8882, "step": 7442 }, { "epoch": 0.89, "learning_rate": 4.131579765303019e-06, "loss": 1.0557, "step": 7443 }, { "epoch": 0.89, "learning_rate": 4.131336217277315e-06, "loss": 0.7084, "step": 7444 }, { "epoch": 0.89, "learning_rate": 4.131092642285489e-06, "loss": 0.637, "step": 7445 }, { "epoch": 0.89, "learning_rate": 4.130849040331563e-06, "loss": 0.7518, "step": 7446 }, { "epoch": 0.89, "learning_rate": 4.130605411419566e-06, "loss": 0.7617, "step": 7447 }, { "epoch": 0.89, "learning_rate": 4.130361755553525e-06, "loss": 0.8555, "step": 7448 }, { "epoch": 0.89, "learning_rate": 4.1301180727374675e-06, "loss": 1.0285, "step": 7449 }, { "epoch": 0.89, "learning_rate": 4.129874362975421e-06, "loss": 1.1605, "step": 7450 }, { "epoch": 0.89, "learning_rate": 4.129630626271415e-06, "loss": 0.6655, "step": 7451 }, { "epoch": 0.89, "learning_rate": 4.129386862629477e-06, "loss": 0.6387, "step": 7452 }, { "epoch": 0.89, "learning_rate": 4.129143072053639e-06, "loss": 0.7202, "step": 7453 }, { "epoch": 0.89, "learning_rate": 4.128899254547927e-06, "loss": 0.8146, "step": 7454 }, { "epoch": 0.89, "learning_rate": 4.128655410116375e-06, "loss": 1.045, "step": 7455 }, { "epoch": 0.89, "learning_rate": 4.1284115387630135e-06, "loss": 1.101, "step": 7456 }, { "epoch": 0.89, "learning_rate": 4.128167640491871e-06, "loss": 0.5364, "step": 7457 }, { "epoch": 0.89, "learning_rate": 4.127923715306982e-06, "loss": 0.8671, "step": 7458 }, { "epoch": 0.89, "learning_rate": 4.127679763212376e-06, "loss": 0.7925, "step": 7459 }, { "epoch": 0.89, "learning_rate": 4.1274357842120884e-06, "loss": 0.9316, "step": 7460 }, { "epoch": 0.89, "learning_rate": 4.12719177831015e-06, "loss": 0.9898, "step": 7461 }, { "epoch": 0.89, "learning_rate": 4.126947745510596e-06, "loss": 1.1357, "step": 7462 }, { "epoch": 0.89, "learning_rate": 4.126703685817458e-06, "loss": 0.6408, "step": 7463 }, { "epoch": 0.89, "learning_rate": 4.126459599234772e-06, "loss": 0.5682, "step": 7464 }, { "epoch": 0.89, "learning_rate": 4.126215485766573e-06, "loss": 0.6359, "step": 7465 }, { "epoch": 0.89, "learning_rate": 4.125971345416894e-06, "loss": 0.6583, "step": 7466 }, { "epoch": 0.89, "learning_rate": 4.125727178189774e-06, "loss": 0.9744, "step": 7467 }, { "epoch": 0.89, "learning_rate": 4.125482984089245e-06, "loss": 1.009, "step": 7468 }, { "epoch": 0.89, "learning_rate": 4.125238763119348e-06, "loss": 0.6463, "step": 7469 }, { "epoch": 0.89, "learning_rate": 4.124994515284117e-06, "loss": 0.6483, "step": 7470 }, { "epoch": 0.89, "learning_rate": 4.124750240587591e-06, "loss": 0.6151, "step": 7471 }, { "epoch": 0.89, "learning_rate": 4.124505939033806e-06, "loss": 0.7521, "step": 7472 }, { "epoch": 0.89, "learning_rate": 4.124261610626803e-06, "loss": 0.9077, "step": 7473 }, { "epoch": 0.89, "learning_rate": 4.124017255370618e-06, "loss": 1.1223, "step": 7474 }, { "epoch": 0.89, "learning_rate": 4.123772873269292e-06, "loss": 1.1122, "step": 7475 }, { "epoch": 0.89, "learning_rate": 4.1235284643268636e-06, "loss": 0.5962, "step": 7476 }, { "epoch": 0.89, "learning_rate": 4.123284028547374e-06, "loss": 0.6831, "step": 7477 }, { "epoch": 0.89, "learning_rate": 4.123039565934863e-06, "loss": 0.691, "step": 7478 }, { "epoch": 0.89, "learning_rate": 4.122795076493373e-06, "loss": 0.8726, "step": 7479 }, { "epoch": 0.89, "learning_rate": 4.1225505602269425e-06, "loss": 1.0529, "step": 7480 }, { "epoch": 0.89, "learning_rate": 4.122306017139616e-06, "loss": 1.0375, "step": 7481 }, { "epoch": 0.89, "learning_rate": 4.122061447235435e-06, "loss": 0.4772, "step": 7482 }, { "epoch": 0.89, "learning_rate": 4.121816850518442e-06, "loss": 0.6483, "step": 7483 }, { "epoch": 0.89, "learning_rate": 4.12157222699268e-06, "loss": 0.7993, "step": 7484 }, { "epoch": 0.89, "learning_rate": 4.121327576662193e-06, "loss": 0.9163, "step": 7485 }, { "epoch": 0.89, "learning_rate": 4.121082899531026e-06, "loss": 1.0074, "step": 7486 }, { "epoch": 0.89, "learning_rate": 4.1208381956032215e-06, "loss": 1.1228, "step": 7487 }, { "epoch": 0.89, "learning_rate": 4.120593464882827e-06, "loss": 0.4699, "step": 7488 }, { "epoch": 0.89, "learning_rate": 4.120348707373885e-06, "loss": 0.6436, "step": 7489 }, { "epoch": 0.89, "learning_rate": 4.120103923080444e-06, "loss": 0.8015, "step": 7490 }, { "epoch": 0.89, "learning_rate": 4.119859112006549e-06, "loss": 0.6058, "step": 7491 }, { "epoch": 0.89, "learning_rate": 4.1196142741562465e-06, "loss": 0.863, "step": 7492 }, { "epoch": 0.89, "learning_rate": 4.119369409533584e-06, "loss": 1.0562, "step": 7493 }, { "epoch": 0.89, "learning_rate": 4.119124518142611e-06, "loss": 0.7708, "step": 7494 }, { "epoch": 0.89, "learning_rate": 4.118879599987372e-06, "loss": 0.6649, "step": 7495 }, { "epoch": 0.89, "learning_rate": 4.118634655071918e-06, "loss": 0.7429, "step": 7496 }, { "epoch": 0.89, "learning_rate": 4.118389683400298e-06, "loss": 0.7564, "step": 7497 }, { "epoch": 0.89, "learning_rate": 4.118144684976559e-06, "loss": 0.8815, "step": 7498 }, { "epoch": 0.89, "learning_rate": 4.117899659804755e-06, "loss": 1.0561, "step": 7499 }, { "epoch": 0.89, "learning_rate": 4.117654607888932e-06, "loss": 1.0839, "step": 7500 }, { "epoch": 0.89, "learning_rate": 4.117409529233143e-06, "loss": 0.7176, "step": 7501 }, { "epoch": 0.89, "learning_rate": 4.117164423841439e-06, "loss": 0.7083, "step": 7502 }, { "epoch": 0.89, "learning_rate": 4.1169192917178716e-06, "loss": 0.7498, "step": 7503 }, { "epoch": 0.89, "learning_rate": 4.1166741328664915e-06, "loss": 0.7691, "step": 7504 }, { "epoch": 0.89, "learning_rate": 4.116428947291354e-06, "loss": 1.0036, "step": 7505 }, { "epoch": 0.89, "learning_rate": 4.11618373499651e-06, "loss": 1.282, "step": 7506 }, { "epoch": 0.89, "learning_rate": 4.115938495986012e-06, "loss": 0.6319, "step": 7507 }, { "epoch": 0.89, "learning_rate": 4.115693230263916e-06, "loss": 0.7419, "step": 7508 }, { "epoch": 0.89, "learning_rate": 4.115447937834275e-06, "loss": 0.7387, "step": 7509 }, { "epoch": 0.89, "learning_rate": 4.115202618701144e-06, "loss": 0.8643, "step": 7510 }, { "epoch": 0.89, "learning_rate": 4.1149572728685785e-06, "loss": 0.9409, "step": 7511 }, { "epoch": 0.89, "learning_rate": 4.114711900340634e-06, "loss": 1.2389, "step": 7512 }, { "epoch": 0.89, "learning_rate": 4.114466501121366e-06, "loss": 0.7415, "step": 7513 }, { "epoch": 0.89, "learning_rate": 4.114221075214831e-06, "loss": 0.6161, "step": 7514 }, { "epoch": 0.89, "learning_rate": 4.113975622625086e-06, "loss": 0.8217, "step": 7515 }, { "epoch": 0.9, "learning_rate": 4.113730143356189e-06, "loss": 0.7673, "step": 7516 }, { "epoch": 0.9, "learning_rate": 4.113484637412198e-06, "loss": 0.9288, "step": 7517 }, { "epoch": 0.9, "learning_rate": 4.11323910479717e-06, "loss": 1.1273, "step": 7518 }, { "epoch": 0.9, "learning_rate": 4.112993545515164e-06, "loss": 0.8146, "step": 7519 }, { "epoch": 0.9, "learning_rate": 4.112747959570239e-06, "loss": 0.4409, "step": 7520 }, { "epoch": 0.9, "learning_rate": 4.112502346966455e-06, "loss": 0.6881, "step": 7521 }, { "epoch": 0.9, "learning_rate": 4.112256707707873e-06, "loss": 0.6965, "step": 7522 }, { "epoch": 0.9, "learning_rate": 4.112011041798552e-06, "loss": 0.7458, "step": 7523 }, { "epoch": 0.9, "learning_rate": 4.111765349242553e-06, "loss": 0.9948, "step": 7524 }, { "epoch": 0.9, "learning_rate": 4.111519630043936e-06, "loss": 0.9755, "step": 7525 }, { "epoch": 0.9, "learning_rate": 4.111273884206767e-06, "loss": 0.5989, "step": 7526 }, { "epoch": 0.9, "learning_rate": 4.111028111735104e-06, "loss": 0.7734, "step": 7527 }, { "epoch": 0.9, "learning_rate": 4.110782312633011e-06, "loss": 0.7797, "step": 7528 }, { "epoch": 0.9, "learning_rate": 4.110536486904552e-06, "loss": 0.7255, "step": 7529 }, { "epoch": 0.9, "learning_rate": 4.110290634553788e-06, "loss": 1.0646, "step": 7530 }, { "epoch": 0.9, "learning_rate": 4.110044755584787e-06, "loss": 1.0347, "step": 7531 }, { "epoch": 0.9, "learning_rate": 4.109798850001609e-06, "loss": 0.5121, "step": 7532 }, { "epoch": 0.9, "learning_rate": 4.1095529178083225e-06, "loss": 0.652, "step": 7533 }, { "epoch": 0.9, "learning_rate": 4.1093069590089905e-06, "loss": 0.6975, "step": 7534 }, { "epoch": 0.9, "learning_rate": 4.10906097360768e-06, "loss": 0.8423, "step": 7535 }, { "epoch": 0.9, "learning_rate": 4.108814961608456e-06, "loss": 1.0851, "step": 7536 }, { "epoch": 0.9, "learning_rate": 4.108568923015386e-06, "loss": 1.1281, "step": 7537 }, { "epoch": 0.9, "learning_rate": 4.108322857832537e-06, "loss": 0.494, "step": 7538 }, { "epoch": 0.9, "learning_rate": 4.108076766063976e-06, "loss": 0.8158, "step": 7539 }, { "epoch": 0.9, "learning_rate": 4.107830647713771e-06, "loss": 0.6682, "step": 7540 }, { "epoch": 0.9, "learning_rate": 4.107584502785991e-06, "loss": 0.7645, "step": 7541 }, { "epoch": 0.9, "learning_rate": 4.107338331284703e-06, "loss": 1.0505, "step": 7542 }, { "epoch": 0.9, "learning_rate": 4.107092133213978e-06, "loss": 1.2812, "step": 7543 }, { "epoch": 0.9, "learning_rate": 4.106845908577885e-06, "loss": 0.5745, "step": 7544 }, { "epoch": 0.9, "learning_rate": 4.106599657380495e-06, "loss": 0.7063, "step": 7545 }, { "epoch": 0.9, "learning_rate": 4.106353379625878e-06, "loss": 0.7175, "step": 7546 }, { "epoch": 0.9, "learning_rate": 4.106107075318105e-06, "loss": 0.9175, "step": 7547 }, { "epoch": 0.9, "learning_rate": 4.1058607444612465e-06, "loss": 0.7729, "step": 7548 }, { "epoch": 0.9, "learning_rate": 4.105614387059376e-06, "loss": 1.0716, "step": 7549 }, { "epoch": 0.9, "learning_rate": 4.105368003116564e-06, "loss": 1.1698, "step": 7550 }, { "epoch": 0.9, "learning_rate": 4.105121592636884e-06, "loss": 0.5123, "step": 7551 }, { "epoch": 0.9, "learning_rate": 4.104875155624411e-06, "loss": 0.6757, "step": 7552 }, { "epoch": 0.9, "learning_rate": 4.104628692083216e-06, "loss": 0.7803, "step": 7553 }, { "epoch": 0.9, "learning_rate": 4.104382202017373e-06, "loss": 0.7812, "step": 7554 }, { "epoch": 0.9, "learning_rate": 4.1041356854309585e-06, "loss": 0.9944, "step": 7555 }, { "epoch": 0.9, "learning_rate": 4.103889142328048e-06, "loss": 1.0112, "step": 7556 }, { "epoch": 0.9, "learning_rate": 4.103642572712714e-06, "loss": 0.5964, "step": 7557 }, { "epoch": 0.9, "learning_rate": 4.103395976589033e-06, "loss": 0.6998, "step": 7558 }, { "epoch": 0.9, "learning_rate": 4.103149353961084e-06, "loss": 0.7991, "step": 7559 }, { "epoch": 0.9, "learning_rate": 4.10290270483294e-06, "loss": 0.7814, "step": 7560 }, { "epoch": 0.9, "learning_rate": 4.1026560292086795e-06, "loss": 1.0609, "step": 7561 }, { "epoch": 0.9, "learning_rate": 4.102409327092381e-06, "loss": 1.1301, "step": 7562 }, { "epoch": 0.9, "learning_rate": 4.102162598488122e-06, "loss": 0.6711, "step": 7563 }, { "epoch": 0.9, "learning_rate": 4.1019158433999804e-06, "loss": 0.4834, "step": 7564 }, { "epoch": 0.9, "learning_rate": 4.101669061832037e-06, "loss": 0.6555, "step": 7565 }, { "epoch": 0.9, "learning_rate": 4.101422253788368e-06, "loss": 0.7155, "step": 7566 }, { "epoch": 0.9, "learning_rate": 4.101175419273056e-06, "loss": 0.911, "step": 7567 }, { "epoch": 0.9, "learning_rate": 4.100928558290178e-06, "loss": 0.9877, "step": 7568 }, { "epoch": 0.9, "learning_rate": 4.100681670843819e-06, "loss": 0.6319, "step": 7569 }, { "epoch": 0.9, "learning_rate": 4.100434756938056e-06, "loss": 0.6975, "step": 7570 }, { "epoch": 0.9, "learning_rate": 4.100187816576973e-06, "loss": 0.6619, "step": 7571 }, { "epoch": 0.9, "learning_rate": 4.09994084976465e-06, "loss": 0.754, "step": 7572 }, { "epoch": 0.9, "learning_rate": 4.099693856505172e-06, "loss": 0.9459, "step": 7573 }, { "epoch": 0.9, "learning_rate": 4.099446836802619e-06, "loss": 1.0495, "step": 7574 }, { "epoch": 0.9, "learning_rate": 4.099199790661076e-06, "loss": 1.1516, "step": 7575 }, { "epoch": 0.9, "learning_rate": 4.098952718084627e-06, "loss": 0.4811, "step": 7576 }, { "epoch": 0.9, "learning_rate": 4.098705619077355e-06, "loss": 0.6352, "step": 7577 }, { "epoch": 0.9, "learning_rate": 4.098458493643344e-06, "loss": 0.6992, "step": 7578 }, { "epoch": 0.9, "learning_rate": 4.098211341786682e-06, "loss": 0.9197, "step": 7579 }, { "epoch": 0.9, "learning_rate": 4.0979641635114505e-06, "loss": 1.0332, "step": 7580 }, { "epoch": 0.9, "learning_rate": 4.0977169588217385e-06, "loss": 1.1463, "step": 7581 }, { "epoch": 0.9, "learning_rate": 4.0974697277216316e-06, "loss": 0.4425, "step": 7582 }, { "epoch": 0.9, "learning_rate": 4.097222470215214e-06, "loss": 0.6196, "step": 7583 }, { "epoch": 0.9, "learning_rate": 4.096975186306576e-06, "loss": 0.7408, "step": 7584 }, { "epoch": 0.9, "learning_rate": 4.096727875999804e-06, "loss": 0.869, "step": 7585 }, { "epoch": 0.9, "learning_rate": 4.096480539298987e-06, "loss": 0.9266, "step": 7586 }, { "epoch": 0.9, "learning_rate": 4.0962331762082125e-06, "loss": 1.0692, "step": 7587 }, { "epoch": 0.9, "learning_rate": 4.09598578673157e-06, "loss": 0.6724, "step": 7588 }, { "epoch": 0.9, "learning_rate": 4.095738370873147e-06, "loss": 0.519, "step": 7589 }, { "epoch": 0.9, "learning_rate": 4.095490928637036e-06, "loss": 0.7398, "step": 7590 }, { "epoch": 0.9, "learning_rate": 4.095243460027326e-06, "loss": 0.6829, "step": 7591 }, { "epoch": 0.9, "learning_rate": 4.0949959650481075e-06, "loss": 0.9255, "step": 7592 }, { "epoch": 0.9, "learning_rate": 4.094748443703472e-06, "loss": 1.1036, "step": 7593 }, { "epoch": 0.9, "learning_rate": 4.094500895997511e-06, "loss": 0.7124, "step": 7594 }, { "epoch": 0.9, "learning_rate": 4.094253321934316e-06, "loss": 0.7186, "step": 7595 }, { "epoch": 0.9, "learning_rate": 4.094005721517981e-06, "loss": 0.735, "step": 7596 }, { "epoch": 0.9, "learning_rate": 4.093758094752596e-06, "loss": 0.7333, "step": 7597 }, { "epoch": 0.9, "learning_rate": 4.093510441642257e-06, "loss": 0.9217, "step": 7598 }, { "epoch": 0.9, "learning_rate": 4.093262762191056e-06, "loss": 1.1214, "step": 7599 }, { "epoch": 0.91, "learning_rate": 4.093015056403089e-06, "loss": 1.1277, "step": 7600 }, { "epoch": 0.91, "learning_rate": 4.092767324282449e-06, "loss": 0.5207, "step": 7601 }, { "epoch": 0.91, "learning_rate": 4.0925195658332305e-06, "loss": 0.7467, "step": 7602 }, { "epoch": 0.91, "learning_rate": 4.092271781059531e-06, "loss": 0.7816, "step": 7603 }, { "epoch": 0.91, "learning_rate": 4.092023969965445e-06, "loss": 0.8435, "step": 7604 }, { "epoch": 0.91, "learning_rate": 4.091776132555069e-06, "loss": 0.965, "step": 7605 }, { "epoch": 0.91, "learning_rate": 4.091528268832501e-06, "loss": 1.1208, "step": 7606 }, { "epoch": 0.91, "learning_rate": 4.091280378801836e-06, "loss": 0.4805, "step": 7607 }, { "epoch": 0.91, "learning_rate": 4.091032462467173e-06, "loss": 0.6639, "step": 7608 }, { "epoch": 0.91, "learning_rate": 4.09078451983261e-06, "loss": 0.8068, "step": 7609 }, { "epoch": 0.91, "learning_rate": 4.090536550902246e-06, "loss": 0.8166, "step": 7610 }, { "epoch": 0.91, "learning_rate": 4.090288555680178e-06, "loss": 1.1055, "step": 7611 }, { "epoch": 0.91, "learning_rate": 4.090040534170508e-06, "loss": 1.09, "step": 7612 }, { "epoch": 0.91, "learning_rate": 4.089792486377334e-06, "loss": 0.59, "step": 7613 }, { "epoch": 0.91, "learning_rate": 4.0895444123047575e-06, "loss": 0.6591, "step": 7614 }, { "epoch": 0.91, "learning_rate": 4.089296311956878e-06, "loss": 0.7289, "step": 7615 }, { "epoch": 0.91, "learning_rate": 4.089048185337797e-06, "loss": 0.6992, "step": 7616 }, { "epoch": 0.91, "learning_rate": 4.088800032451616e-06, "loss": 0.8701, "step": 7617 }, { "epoch": 0.91, "learning_rate": 4.088551853302437e-06, "loss": 1.0433, "step": 7618 }, { "epoch": 0.91, "learning_rate": 4.088303647894363e-06, "loss": 0.7949, "step": 7619 }, { "epoch": 0.91, "learning_rate": 4.0880554162314966e-06, "loss": 0.6316, "step": 7620 }, { "epoch": 0.91, "learning_rate": 4.0878071583179406e-06, "loss": 0.658, "step": 7621 }, { "epoch": 0.91, "learning_rate": 4.087558874157799e-06, "loss": 0.7175, "step": 7622 }, { "epoch": 0.91, "learning_rate": 4.087310563755176e-06, "loss": 0.9404, "step": 7623 }, { "epoch": 0.91, "learning_rate": 4.087062227114177e-06, "loss": 1.1541, "step": 7624 }, { "epoch": 0.91, "learning_rate": 4.0868138642389055e-06, "loss": 1.0182, "step": 7625 }, { "epoch": 0.91, "learning_rate": 4.086565475133467e-06, "loss": 0.4513, "step": 7626 }, { "epoch": 0.91, "learning_rate": 4.086317059801969e-06, "loss": 0.6591, "step": 7627 }, { "epoch": 0.91, "learning_rate": 4.086068618248517e-06, "loss": 0.7757, "step": 7628 }, { "epoch": 0.91, "learning_rate": 4.085820150477218e-06, "loss": 0.7839, "step": 7629 }, { "epoch": 0.91, "learning_rate": 4.085571656492178e-06, "loss": 0.9773, "step": 7630 }, { "epoch": 0.91, "learning_rate": 4.085323136297506e-06, "loss": 1.1522, "step": 7631 }, { "epoch": 0.91, "learning_rate": 4.085074589897309e-06, "loss": 0.5439, "step": 7632 }, { "epoch": 0.91, "learning_rate": 4.084826017295696e-06, "loss": 0.7518, "step": 7633 }, { "epoch": 0.91, "learning_rate": 4.084577418496775e-06, "loss": 0.6263, "step": 7634 }, { "epoch": 0.91, "learning_rate": 4.084328793504659e-06, "loss": 0.8381, "step": 7635 }, { "epoch": 0.91, "learning_rate": 4.084080142323454e-06, "loss": 0.9481, "step": 7636 }, { "epoch": 0.91, "learning_rate": 4.0838314649572704e-06, "loss": 1.1547, "step": 7637 }, { "epoch": 0.91, "learning_rate": 4.08358276141022e-06, "loss": 0.5982, "step": 7638 }, { "epoch": 0.91, "learning_rate": 4.083334031686414e-06, "loss": 0.5338, "step": 7639 }, { "epoch": 0.91, "learning_rate": 4.083085275789964e-06, "loss": 0.8196, "step": 7640 }, { "epoch": 0.91, "learning_rate": 4.082836493724981e-06, "loss": 0.801, "step": 7641 }, { "epoch": 0.91, "learning_rate": 4.0825876854955784e-06, "loss": 0.9081, "step": 7642 }, { "epoch": 0.91, "learning_rate": 4.082338851105868e-06, "loss": 1.1313, "step": 7643 }, { "epoch": 0.91, "learning_rate": 4.082089990559964e-06, "loss": 0.765, "step": 7644 }, { "epoch": 0.91, "learning_rate": 4.08184110386198e-06, "loss": 0.6381, "step": 7645 }, { "epoch": 0.91, "learning_rate": 4.0815921910160294e-06, "loss": 0.6839, "step": 7646 }, { "epoch": 0.91, "learning_rate": 4.0813432520262275e-06, "loss": 0.7066, "step": 7647 }, { "epoch": 0.91, "learning_rate": 4.081094286896688e-06, "loss": 0.8225, "step": 7648 }, { "epoch": 0.91, "learning_rate": 4.080845295631528e-06, "loss": 1.0474, "step": 7649 }, { "epoch": 0.91, "learning_rate": 4.0805962782348625e-06, "loss": 1.2429, "step": 7650 }, { "epoch": 0.91, "learning_rate": 4.080347234710809e-06, "loss": 0.7083, "step": 7651 }, { "epoch": 0.91, "learning_rate": 4.080098165063482e-06, "loss": 0.5905, "step": 7652 }, { "epoch": 0.91, "learning_rate": 4.079849069297e-06, "loss": 0.7083, "step": 7653 }, { "epoch": 0.91, "learning_rate": 4.07959994741548e-06, "loss": 1.0009, "step": 7654 }, { "epoch": 0.91, "learning_rate": 4.079350799423041e-06, "loss": 0.979, "step": 7655 }, { "epoch": 0.91, "learning_rate": 4.0791016253237994e-06, "loss": 1.1791, "step": 7656 }, { "epoch": 0.91, "learning_rate": 4.078852425121877e-06, "loss": 0.5387, "step": 7657 }, { "epoch": 0.91, "learning_rate": 4.07860319882139e-06, "loss": 0.6489, "step": 7658 }, { "epoch": 0.91, "learning_rate": 4.078353946426461e-06, "loss": 0.6609, "step": 7659 }, { "epoch": 0.91, "learning_rate": 4.078104667941209e-06, "loss": 0.8615, "step": 7660 }, { "epoch": 0.91, "learning_rate": 4.077855363369753e-06, "loss": 1.0304, "step": 7661 }, { "epoch": 0.91, "learning_rate": 4.077606032716216e-06, "loss": 0.9066, "step": 7662 }, { "epoch": 0.91, "learning_rate": 4.07735667598472e-06, "loss": 0.6719, "step": 7663 }, { "epoch": 0.91, "learning_rate": 4.077107293179384e-06, "loss": 0.602, "step": 7664 }, { "epoch": 0.91, "learning_rate": 4.076857884304334e-06, "loss": 0.7336, "step": 7665 }, { "epoch": 0.91, "learning_rate": 4.076608449363689e-06, "loss": 0.8481, "step": 7666 }, { "epoch": 0.91, "learning_rate": 4.076358988361576e-06, "loss": 0.9031, "step": 7667 }, { "epoch": 0.91, "learning_rate": 4.076109501302115e-06, "loss": 1.2042, "step": 7668 }, { "epoch": 0.91, "learning_rate": 4.0758599881894325e-06, "loss": 0.7031, "step": 7669 }, { "epoch": 0.91, "learning_rate": 4.075610449027652e-06, "loss": 0.5966, "step": 7670 }, { "epoch": 0.91, "learning_rate": 4.075360883820898e-06, "loss": 0.745, "step": 7671 }, { "epoch": 0.91, "learning_rate": 4.075111292573297e-06, "loss": 0.7531, "step": 7672 }, { "epoch": 0.91, "learning_rate": 4.074861675288974e-06, "loss": 0.9239, "step": 7673 }, { "epoch": 0.91, "learning_rate": 4.074612031972056e-06, "loss": 1.0902, "step": 7674 }, { "epoch": 0.91, "learning_rate": 4.074362362626668e-06, "loss": 1.2255, "step": 7675 }, { "epoch": 0.91, "learning_rate": 4.074112667256939e-06, "loss": 0.4512, "step": 7676 }, { "epoch": 0.91, "learning_rate": 4.073862945866995e-06, "loss": 0.7739, "step": 7677 }, { "epoch": 0.91, "learning_rate": 4.073613198460965e-06, "loss": 0.6685, "step": 7678 }, { "epoch": 0.91, "learning_rate": 4.073363425042976e-06, "loss": 0.8231, "step": 7679 }, { "epoch": 0.91, "learning_rate": 4.073113625617157e-06, "loss": 1.0833, "step": 7680 }, { "epoch": 0.91, "learning_rate": 4.072863800187639e-06, "loss": 1.1325, "step": 7681 }, { "epoch": 0.91, "learning_rate": 4.07261394875855e-06, "loss": 0.6207, "step": 7682 }, { "epoch": 0.91, "learning_rate": 4.072364071334021e-06, "loss": 0.6891, "step": 7683 }, { "epoch": 0.92, "learning_rate": 4.072114167918182e-06, "loss": 0.8663, "step": 7684 }, { "epoch": 0.92, "learning_rate": 4.071864238515163e-06, "loss": 0.8461, "step": 7685 }, { "epoch": 0.92, "learning_rate": 4.071614283129096e-06, "loss": 0.9406, "step": 7686 }, { "epoch": 0.92, "learning_rate": 4.071364301764113e-06, "loss": 1.1465, "step": 7687 }, { "epoch": 0.92, "learning_rate": 4.0711142944243485e-06, "loss": 0.692, "step": 7688 }, { "epoch": 0.92, "learning_rate": 4.070864261113931e-06, "loss": 0.8611, "step": 7689 }, { "epoch": 0.92, "learning_rate": 4.070614201836995e-06, "loss": 0.723, "step": 7690 }, { "epoch": 0.92, "learning_rate": 4.070364116597676e-06, "loss": 0.8431, "step": 7691 }, { "epoch": 0.92, "learning_rate": 4.070114005400105e-06, "loss": 0.9984, "step": 7692 }, { "epoch": 0.92, "learning_rate": 4.069863868248419e-06, "loss": 1.0543, "step": 7693 }, { "epoch": 0.92, "learning_rate": 4.069613705146751e-06, "loss": 0.823, "step": 7694 }, { "epoch": 0.92, "learning_rate": 4.0693635160992364e-06, "loss": 0.5746, "step": 7695 }, { "epoch": 0.92, "learning_rate": 4.069113301110011e-06, "loss": 0.7682, "step": 7696 }, { "epoch": 0.92, "learning_rate": 4.068863060183212e-06, "loss": 0.7681, "step": 7697 }, { "epoch": 0.92, "learning_rate": 4.068612793322974e-06, "loss": 0.8351, "step": 7698 }, { "epoch": 0.92, "learning_rate": 4.068362500533436e-06, "loss": 1.1354, "step": 7699 }, { "epoch": 0.92, "learning_rate": 4.068112181818733e-06, "loss": 1.0803, "step": 7700 }, { "epoch": 0.92, "learning_rate": 4.067861837183006e-06, "loss": 0.6471, "step": 7701 }, { "epoch": 0.92, "learning_rate": 4.067611466630389e-06, "loss": 0.6305, "step": 7702 }, { "epoch": 0.92, "learning_rate": 4.067361070165024e-06, "loss": 0.773, "step": 7703 }, { "epoch": 0.92, "learning_rate": 4.06711064779105e-06, "loss": 0.7857, "step": 7704 }, { "epoch": 0.92, "learning_rate": 4.066860199512603e-06, "loss": 1.0402, "step": 7705 }, { "epoch": 0.92, "learning_rate": 4.066609725333827e-06, "loss": 1.0582, "step": 7706 }, { "epoch": 0.92, "learning_rate": 4.06635922525886e-06, "loss": 0.5213, "step": 7707 }, { "epoch": 0.92, "learning_rate": 4.0661086992918454e-06, "loss": 0.6299, "step": 7708 }, { "epoch": 0.92, "learning_rate": 4.065858147436921e-06, "loss": 0.8354, "step": 7709 }, { "epoch": 0.92, "learning_rate": 4.065607569698231e-06, "loss": 0.7382, "step": 7710 }, { "epoch": 0.92, "learning_rate": 4.065356966079915e-06, "loss": 1.0218, "step": 7711 }, { "epoch": 0.92, "learning_rate": 4.065106336586118e-06, "loss": 1.0891, "step": 7712 }, { "epoch": 0.92, "learning_rate": 4.064855681220983e-06, "loss": 0.6724, "step": 7713 }, { "epoch": 0.92, "learning_rate": 4.064604999988651e-06, "loss": 0.8057, "step": 7714 }, { "epoch": 0.92, "learning_rate": 4.064354292893267e-06, "loss": 0.763, "step": 7715 }, { "epoch": 0.92, "learning_rate": 4.0641035599389755e-06, "loss": 0.6879, "step": 7716 }, { "epoch": 0.92, "learning_rate": 4.063852801129921e-06, "loss": 1.0325, "step": 7717 }, { "epoch": 0.92, "learning_rate": 4.063602016470248e-06, "loss": 1.1243, "step": 7718 }, { "epoch": 0.92, "learning_rate": 4.063351205964104e-06, "loss": 0.6728, "step": 7719 }, { "epoch": 0.92, "learning_rate": 4.063100369615631e-06, "loss": 0.6524, "step": 7720 }, { "epoch": 0.92, "learning_rate": 4.062849507428979e-06, "loss": 0.6596, "step": 7721 }, { "epoch": 0.92, "learning_rate": 4.0625986194082945e-06, "loss": 0.7816, "step": 7722 }, { "epoch": 0.92, "learning_rate": 4.062347705557722e-06, "loss": 0.9855, "step": 7723 }, { "epoch": 0.92, "learning_rate": 4.062096765881411e-06, "loss": 1.0998, "step": 7724 }, { "epoch": 0.92, "learning_rate": 4.061845800383511e-06, "loss": 1.2595, "step": 7725 }, { "epoch": 0.92, "learning_rate": 4.061594809068168e-06, "loss": 0.7657, "step": 7726 }, { "epoch": 0.92, "learning_rate": 4.061343791939531e-06, "loss": 0.6563, "step": 7727 }, { "epoch": 0.92, "learning_rate": 4.061092749001751e-06, "loss": 0.7671, "step": 7728 }, { "epoch": 0.92, "learning_rate": 4.060841680258977e-06, "loss": 0.7541, "step": 7729 }, { "epoch": 0.92, "learning_rate": 4.060590585715358e-06, "loss": 0.9607, "step": 7730 }, { "epoch": 0.92, "learning_rate": 4.060339465375046e-06, "loss": 1.1115, "step": 7731 }, { "epoch": 0.92, "learning_rate": 4.060088319242192e-06, "loss": 0.5534, "step": 7732 }, { "epoch": 0.92, "learning_rate": 4.059837147320948e-06, "loss": 0.7203, "step": 7733 }, { "epoch": 0.92, "learning_rate": 4.059585949615464e-06, "loss": 0.7239, "step": 7734 }, { "epoch": 0.92, "learning_rate": 4.0593347261298936e-06, "loss": 0.6967, "step": 7735 }, { "epoch": 0.92, "learning_rate": 4.05908347686839e-06, "loss": 0.9683, "step": 7736 }, { "epoch": 0.92, "learning_rate": 4.058832201835104e-06, "loss": 1.0576, "step": 7737 }, { "epoch": 0.92, "learning_rate": 4.058580901034193e-06, "loss": 0.679, "step": 7738 }, { "epoch": 0.92, "learning_rate": 4.058329574469808e-06, "loss": 0.5042, "step": 7739 }, { "epoch": 0.92, "learning_rate": 4.058078222146105e-06, "loss": 0.7319, "step": 7740 }, { "epoch": 0.92, "learning_rate": 4.0578268440672365e-06, "loss": 0.8071, "step": 7741 }, { "epoch": 0.92, "learning_rate": 4.057575440237361e-06, "loss": 0.7961, "step": 7742 }, { "epoch": 0.92, "learning_rate": 4.057324010660632e-06, "loss": 1.0328, "step": 7743 }, { "epoch": 0.92, "learning_rate": 4.057072555341207e-06, "loss": 0.9936, "step": 7744 }, { "epoch": 0.92, "learning_rate": 4.056821074283242e-06, "loss": 0.6285, "step": 7745 }, { "epoch": 0.92, "learning_rate": 4.056569567490894e-06, "loss": 0.8237, "step": 7746 }, { "epoch": 0.92, "learning_rate": 4.056318034968321e-06, "loss": 0.74, "step": 7747 }, { "epoch": 0.92, "learning_rate": 4.056066476719679e-06, "loss": 0.9649, "step": 7748 }, { "epoch": 0.92, "learning_rate": 4.055814892749128e-06, "loss": 0.9869, "step": 7749 }, { "epoch": 0.92, "learning_rate": 4.055563283060827e-06, "loss": 1.0445, "step": 7750 }, { "epoch": 0.92, "learning_rate": 4.0553116476589336e-06, "loss": 0.6194, "step": 7751 }, { "epoch": 0.92, "learning_rate": 4.055059986547608e-06, "loss": 0.7609, "step": 7752 }, { "epoch": 0.92, "learning_rate": 4.054808299731011e-06, "loss": 0.731, "step": 7753 }, { "epoch": 0.92, "learning_rate": 4.054556587213303e-06, "loss": 0.7086, "step": 7754 }, { "epoch": 0.92, "learning_rate": 4.0543048489986435e-06, "loss": 1.0148, "step": 7755 }, { "epoch": 0.92, "learning_rate": 4.054053085091194e-06, "loss": 1.2289, "step": 7756 }, { "epoch": 0.92, "learning_rate": 4.053801295495116e-06, "loss": 0.5026, "step": 7757 }, { "epoch": 0.92, "learning_rate": 4.053549480214574e-06, "loss": 0.6434, "step": 7758 }, { "epoch": 0.92, "learning_rate": 4.053297639253727e-06, "loss": 0.8113, "step": 7759 }, { "epoch": 0.92, "learning_rate": 4.05304577261674e-06, "loss": 0.8981, "step": 7760 }, { "epoch": 0.92, "learning_rate": 4.052793880307777e-06, "loss": 1.0164, "step": 7761 }, { "epoch": 0.92, "learning_rate": 4.052541962331001e-06, "loss": 1.2384, "step": 7762 }, { "epoch": 0.92, "learning_rate": 4.052290018690576e-06, "loss": 0.6121, "step": 7763 }, { "epoch": 0.92, "learning_rate": 4.052038049390665e-06, "loss": 0.6909, "step": 7764 }, { "epoch": 0.92, "learning_rate": 4.051786054435437e-06, "loss": 0.7592, "step": 7765 }, { "epoch": 0.92, "learning_rate": 4.051534033829054e-06, "loss": 0.7712, "step": 7766 }, { "epoch": 0.92, "learning_rate": 4.051281987575684e-06, "loss": 1.047, "step": 7767 }, { "epoch": 0.93, "learning_rate": 4.0510299156794916e-06, "loss": 1.1168, "step": 7768 }, { "epoch": 0.93, "learning_rate": 4.050777818144645e-06, "loss": 0.8456, "step": 7769 }, { "epoch": 0.93, "learning_rate": 4.050525694975312e-06, "loss": 0.6082, "step": 7770 }, { "epoch": 0.93, "learning_rate": 4.050273546175658e-06, "loss": 0.5903, "step": 7771 }, { "epoch": 0.93, "learning_rate": 4.050021371749853e-06, "loss": 0.6685, "step": 7772 }, { "epoch": 0.93, "learning_rate": 4.049769171702064e-06, "loss": 0.8373, "step": 7773 }, { "epoch": 0.93, "learning_rate": 4.04951694603646e-06, "loss": 0.9263, "step": 7774 }, { "epoch": 0.93, "learning_rate": 4.049264694757212e-06, "loss": 1.2262, "step": 7775 }, { "epoch": 0.93, "learning_rate": 4.0490124178684884e-06, "loss": 0.6265, "step": 7776 }, { "epoch": 0.93, "learning_rate": 4.048760115374459e-06, "loss": 0.7428, "step": 7777 }, { "epoch": 0.93, "learning_rate": 4.048507787279295e-06, "loss": 0.7859, "step": 7778 }, { "epoch": 0.93, "learning_rate": 4.048255433587169e-06, "loss": 0.7783, "step": 7779 }, { "epoch": 0.93, "learning_rate": 4.048003054302249e-06, "loss": 1.0604, "step": 7780 }, { "epoch": 0.93, "learning_rate": 4.047750649428708e-06, "loss": 1.1187, "step": 7781 }, { "epoch": 0.93, "learning_rate": 4.04749821897072e-06, "loss": 0.7238, "step": 7782 }, { "epoch": 0.93, "learning_rate": 4.047245762932457e-06, "loss": 0.6826, "step": 7783 }, { "epoch": 0.93, "learning_rate": 4.046993281318091e-06, "loss": 0.7752, "step": 7784 }, { "epoch": 0.93, "learning_rate": 4.046740774131796e-06, "loss": 0.74, "step": 7785 }, { "epoch": 0.93, "learning_rate": 4.046488241377747e-06, "loss": 0.8591, "step": 7786 }, { "epoch": 0.93, "learning_rate": 4.046235683060118e-06, "loss": 1.1658, "step": 7787 }, { "epoch": 0.93, "learning_rate": 4.0459830991830825e-06, "loss": 0.6231, "step": 7788 }, { "epoch": 0.93, "learning_rate": 4.045730489750817e-06, "loss": 0.5505, "step": 7789 }, { "epoch": 0.93, "learning_rate": 4.0454778547674965e-06, "loss": 0.7671, "step": 7790 }, { "epoch": 0.93, "learning_rate": 4.0452251942372975e-06, "loss": 0.8124, "step": 7791 }, { "epoch": 0.93, "learning_rate": 4.0449725081643964e-06, "loss": 0.9063, "step": 7792 }, { "epoch": 0.93, "learning_rate": 4.044719796552971e-06, "loss": 1.2302, "step": 7793 }, { "epoch": 0.93, "learning_rate": 4.044467059407197e-06, "loss": 0.7933, "step": 7794 }, { "epoch": 0.93, "learning_rate": 4.044214296731252e-06, "loss": 0.5101, "step": 7795 }, { "epoch": 0.93, "learning_rate": 4.043961508529316e-06, "loss": 0.7089, "step": 7796 }, { "epoch": 0.93, "learning_rate": 4.043708694805567e-06, "loss": 0.6986, "step": 7797 }, { "epoch": 0.93, "learning_rate": 4.043455855564183e-06, "loss": 0.9489, "step": 7798 }, { "epoch": 0.93, "learning_rate": 4.043202990809345e-06, "loss": 1.0753, "step": 7799 }, { "epoch": 0.93, "learning_rate": 4.0429501005452325e-06, "loss": 1.1316, "step": 7800 }, { "epoch": 0.93, "learning_rate": 4.042697184776023e-06, "loss": 0.588, "step": 7801 }, { "epoch": 0.93, "learning_rate": 4.0424442435059016e-06, "loss": 0.6247, "step": 7802 }, { "epoch": 0.93, "learning_rate": 4.042191276739047e-06, "loss": 0.7307, "step": 7803 }, { "epoch": 0.93, "learning_rate": 4.0419382844796415e-06, "loss": 0.9195, "step": 7804 }, { "epoch": 0.93, "learning_rate": 4.0416852667318664e-06, "loss": 1.0271, "step": 7805 }, { "epoch": 0.93, "learning_rate": 4.041432223499904e-06, "loss": 1.0802, "step": 7806 }, { "epoch": 0.93, "learning_rate": 4.041179154787939e-06, "loss": 0.6385, "step": 7807 }, { "epoch": 0.93, "learning_rate": 4.040926060600152e-06, "loss": 0.6934, "step": 7808 }, { "epoch": 0.93, "learning_rate": 4.040672940940728e-06, "loss": 0.8547, "step": 7809 }, { "epoch": 0.93, "learning_rate": 4.0404197958138515e-06, "loss": 0.6593, "step": 7810 }, { "epoch": 0.93, "learning_rate": 4.040166625223706e-06, "loss": 0.9645, "step": 7811 }, { "epoch": 0.93, "learning_rate": 4.039913429174478e-06, "loss": 1.1121, "step": 7812 }, { "epoch": 0.93, "learning_rate": 4.039660207670351e-06, "loss": 0.5292, "step": 7813 }, { "epoch": 0.93, "learning_rate": 4.039406960715512e-06, "loss": 0.5758, "step": 7814 }, { "epoch": 0.93, "learning_rate": 4.039153688314146e-06, "loss": 0.634, "step": 7815 }, { "epoch": 0.93, "learning_rate": 4.038900390470441e-06, "loss": 0.8119, "step": 7816 }, { "epoch": 0.93, "learning_rate": 4.038647067188584e-06, "loss": 0.8949, "step": 7817 }, { "epoch": 0.93, "learning_rate": 4.038393718472761e-06, "loss": 1.1613, "step": 7818 }, { "epoch": 0.93, "learning_rate": 4.038140344327161e-06, "loss": 0.9591, "step": 7819 }, { "epoch": 0.93, "learning_rate": 4.037886944755971e-06, "loss": 0.5941, "step": 7820 }, { "epoch": 0.93, "learning_rate": 4.037633519763382e-06, "loss": 0.8192, "step": 7821 }, { "epoch": 0.93, "learning_rate": 4.037380069353581e-06, "loss": 0.7651, "step": 7822 }, { "epoch": 0.93, "learning_rate": 4.03712659353076e-06, "loss": 0.8506, "step": 7823 }, { "epoch": 0.93, "learning_rate": 4.0368730922991065e-06, "loss": 0.947, "step": 7824 }, { "epoch": 0.93, "learning_rate": 4.036619565662812e-06, "loss": 1.361, "step": 7825 }, { "epoch": 0.93, "learning_rate": 4.036366013626066e-06, "loss": 0.5887, "step": 7826 }, { "epoch": 0.93, "learning_rate": 4.0361124361930625e-06, "loss": 0.6306, "step": 7827 }, { "epoch": 0.93, "learning_rate": 4.035858833367991e-06, "loss": 0.6082, "step": 7828 }, { "epoch": 0.93, "learning_rate": 4.0356052051550435e-06, "loss": 0.8643, "step": 7829 }, { "epoch": 0.93, "learning_rate": 4.035351551558414e-06, "loss": 0.9399, "step": 7830 }, { "epoch": 0.93, "learning_rate": 4.035097872582294e-06, "loss": 1.0629, "step": 7831 }, { "epoch": 0.93, "learning_rate": 4.034844168230878e-06, "loss": 0.5327, "step": 7832 }, { "epoch": 0.93, "learning_rate": 4.034590438508358e-06, "loss": 0.6378, "step": 7833 }, { "epoch": 0.93, "learning_rate": 4.0343366834189305e-06, "loss": 0.6741, "step": 7834 }, { "epoch": 0.93, "learning_rate": 4.034082902966788e-06, "loss": 0.7922, "step": 7835 }, { "epoch": 0.93, "learning_rate": 4.033829097156127e-06, "loss": 0.9742, "step": 7836 }, { "epoch": 0.93, "learning_rate": 4.033575265991142e-06, "loss": 1.1209, "step": 7837 }, { "epoch": 0.93, "learning_rate": 4.033321409476029e-06, "loss": 0.5663, "step": 7838 }, { "epoch": 0.93, "learning_rate": 4.033067527614984e-06, "loss": 0.6224, "step": 7839 }, { "epoch": 0.93, "learning_rate": 4.032813620412206e-06, "loss": 0.6264, "step": 7840 }, { "epoch": 0.93, "learning_rate": 4.032559687871888e-06, "loss": 0.7523, "step": 7841 }, { "epoch": 0.93, "learning_rate": 4.032305729998231e-06, "loss": 0.9875, "step": 7842 }, { "epoch": 0.93, "learning_rate": 4.032051746795431e-06, "loss": 1.1171, "step": 7843 }, { "epoch": 0.93, "learning_rate": 4.031797738267689e-06, "loss": 0.6664, "step": 7844 }, { "epoch": 0.93, "learning_rate": 4.0315437044191996e-06, "loss": 0.6491, "step": 7845 }, { "epoch": 0.93, "learning_rate": 4.031289645254165e-06, "loss": 0.6227, "step": 7846 }, { "epoch": 0.93, "learning_rate": 4.0310355607767845e-06, "loss": 0.79, "step": 7847 }, { "epoch": 0.93, "learning_rate": 4.030781450991257e-06, "loss": 0.9284, "step": 7848 }, { "epoch": 0.93, "learning_rate": 4.030527315901784e-06, "loss": 1.077, "step": 7849 }, { "epoch": 0.93, "learning_rate": 4.030273155512566e-06, "loss": 1.1589, "step": 7850 }, { "epoch": 0.93, "learning_rate": 4.030018969827805e-06, "loss": 0.6167, "step": 7851 }, { "epoch": 0.94, "learning_rate": 4.029764758851701e-06, "loss": 0.5953, "step": 7852 }, { "epoch": 0.94, "learning_rate": 4.029510522588458e-06, "loss": 0.7818, "step": 7853 }, { "epoch": 0.94, "learning_rate": 4.029256261042277e-06, "loss": 0.8386, "step": 7854 }, { "epoch": 0.94, "learning_rate": 4.029001974217363e-06, "loss": 1.0362, "step": 7855 }, { "epoch": 0.94, "learning_rate": 4.028747662117917e-06, "loss": 1.2306, "step": 7856 }, { "epoch": 0.94, "learning_rate": 4.028493324748144e-06, "loss": 0.6927, "step": 7857 }, { "epoch": 0.94, "learning_rate": 4.028238962112248e-06, "loss": 0.7632, "step": 7858 }, { "epoch": 0.94, "learning_rate": 4.027984574214434e-06, "loss": 0.6008, "step": 7859 }, { "epoch": 0.94, "learning_rate": 4.027730161058906e-06, "loss": 0.8267, "step": 7860 }, { "epoch": 0.94, "learning_rate": 4.027475722649871e-06, "loss": 1.0759, "step": 7861 }, { "epoch": 0.94, "learning_rate": 4.027221258991534e-06, "loss": 1.0732, "step": 7862 }, { "epoch": 0.94, "learning_rate": 4.026966770088102e-06, "loss": 0.5282, "step": 7863 }, { "epoch": 0.94, "learning_rate": 4.026712255943781e-06, "loss": 0.5604, "step": 7864 }, { "epoch": 0.94, "learning_rate": 4.026457716562777e-06, "loss": 0.6216, "step": 7865 }, { "epoch": 0.94, "learning_rate": 4.026203151949301e-06, "loss": 0.8477, "step": 7866 }, { "epoch": 0.94, "learning_rate": 4.025948562107557e-06, "loss": 0.8065, "step": 7867 }, { "epoch": 0.94, "learning_rate": 4.025693947041756e-06, "loss": 1.0908, "step": 7868 }, { "epoch": 0.94, "learning_rate": 4.025439306756106e-06, "loss": 0.7498, "step": 7869 }, { "epoch": 0.94, "learning_rate": 4.025184641254816e-06, "loss": 0.662, "step": 7870 }, { "epoch": 0.94, "learning_rate": 4.024929950542096e-06, "loss": 0.7018, "step": 7871 }, { "epoch": 0.94, "learning_rate": 4.024675234622156e-06, "loss": 0.7705, "step": 7872 }, { "epoch": 0.94, "learning_rate": 4.024420493499206e-06, "loss": 0.9223, "step": 7873 }, { "epoch": 0.94, "learning_rate": 4.024165727177458e-06, "loss": 1.0602, "step": 7874 }, { "epoch": 0.94, "learning_rate": 4.023910935661123e-06, "loss": 1.1957, "step": 7875 }, { "epoch": 0.94, "learning_rate": 4.023656118954412e-06, "loss": 0.5322, "step": 7876 }, { "epoch": 0.94, "learning_rate": 4.0234012770615375e-06, "loss": 0.7308, "step": 7877 }, { "epoch": 0.94, "learning_rate": 4.023146409986711e-06, "loss": 0.728, "step": 7878 }, { "epoch": 0.94, "learning_rate": 4.022891517734148e-06, "loss": 0.8116, "step": 7879 }, { "epoch": 0.94, "learning_rate": 4.02263660030806e-06, "loss": 0.9282, "step": 7880 }, { "epoch": 0.94, "learning_rate": 4.022381657712661e-06, "loss": 1.1876, "step": 7881 }, { "epoch": 0.94, "learning_rate": 4.0221266899521655e-06, "loss": 0.6734, "step": 7882 }, { "epoch": 0.94, "learning_rate": 4.021871697030788e-06, "loss": 0.6745, "step": 7883 }, { "epoch": 0.94, "learning_rate": 4.021616678952744e-06, "loss": 0.8853, "step": 7884 }, { "epoch": 0.94, "learning_rate": 4.021361635722248e-06, "loss": 0.808, "step": 7885 }, { "epoch": 0.94, "learning_rate": 4.021106567343518e-06, "loss": 0.9862, "step": 7886 }, { "epoch": 0.94, "learning_rate": 4.020851473820767e-06, "loss": 1.2584, "step": 7887 }, { "epoch": 0.94, "learning_rate": 4.020596355158214e-06, "loss": 0.6673, "step": 7888 }, { "epoch": 0.94, "learning_rate": 4.020341211360076e-06, "loss": 0.7968, "step": 7889 }, { "epoch": 0.94, "learning_rate": 4.02008604243057e-06, "loss": 0.7657, "step": 7890 }, { "epoch": 0.94, "learning_rate": 4.019830848373914e-06, "loss": 0.6653, "step": 7891 }, { "epoch": 0.94, "learning_rate": 4.019575629194328e-06, "loss": 0.859, "step": 7892 }, { "epoch": 0.94, "learning_rate": 4.019320384896028e-06, "loss": 1.094, "step": 7893 }, { "epoch": 0.94, "learning_rate": 4.019065115483235e-06, "loss": 0.8168, "step": 7894 }, { "epoch": 0.94, "learning_rate": 4.018809820960168e-06, "loss": 0.5162, "step": 7895 }, { "epoch": 0.94, "learning_rate": 4.0185545013310475e-06, "loss": 0.6632, "step": 7896 }, { "epoch": 0.94, "learning_rate": 4.018299156600093e-06, "loss": 0.7375, "step": 7897 }, { "epoch": 0.94, "learning_rate": 4.0180437867715264e-06, "loss": 1.0183, "step": 7898 }, { "epoch": 0.94, "learning_rate": 4.017788391849569e-06, "loss": 1.0763, "step": 7899 }, { "epoch": 0.94, "learning_rate": 4.017532971838442e-06, "loss": 1.1046, "step": 7900 }, { "epoch": 0.94, "learning_rate": 4.017277526742367e-06, "loss": 0.7325, "step": 7901 }, { "epoch": 0.94, "learning_rate": 4.0170220565655675e-06, "loss": 0.7708, "step": 7902 }, { "epoch": 0.94, "learning_rate": 4.016766561312267e-06, "loss": 0.7601, "step": 7903 }, { "epoch": 0.94, "learning_rate": 4.0165110409866865e-06, "loss": 0.8691, "step": 7904 }, { "epoch": 0.94, "learning_rate": 4.016255495593052e-06, "loss": 1.0234, "step": 7905 }, { "epoch": 0.94, "learning_rate": 4.015999925135586e-06, "loss": 1.0686, "step": 7906 }, { "epoch": 0.94, "learning_rate": 4.015744329618515e-06, "loss": 0.4876, "step": 7907 }, { "epoch": 0.94, "learning_rate": 4.015488709046064e-06, "loss": 0.625, "step": 7908 }, { "epoch": 0.94, "learning_rate": 4.0152330634224554e-06, "loss": 0.7599, "step": 7909 }, { "epoch": 0.94, "learning_rate": 4.014977392751919e-06, "loss": 0.7994, "step": 7910 }, { "epoch": 0.94, "learning_rate": 4.014721697038678e-06, "loss": 0.8871, "step": 7911 }, { "epoch": 0.94, "learning_rate": 4.014465976286961e-06, "loss": 1.103, "step": 7912 }, { "epoch": 0.94, "learning_rate": 4.014210230500993e-06, "loss": 0.55, "step": 7913 }, { "epoch": 0.94, "learning_rate": 4.013954459685004e-06, "loss": 0.6598, "step": 7914 }, { "epoch": 0.94, "learning_rate": 4.013698663843221e-06, "loss": 0.7032, "step": 7915 }, { "epoch": 0.94, "learning_rate": 4.013442842979871e-06, "loss": 0.7932, "step": 7916 }, { "epoch": 0.94, "learning_rate": 4.013186997099184e-06, "loss": 0.9385, "step": 7917 }, { "epoch": 0.94, "learning_rate": 4.01293112620539e-06, "loss": 1.1365, "step": 7918 }, { "epoch": 0.94, "learning_rate": 4.0126752303027176e-06, "loss": 0.7638, "step": 7919 }, { "epoch": 0.94, "learning_rate": 4.012419309395395e-06, "loss": 0.549, "step": 7920 }, { "epoch": 0.94, "learning_rate": 4.012163363487655e-06, "loss": 0.5539, "step": 7921 }, { "epoch": 0.94, "learning_rate": 4.011907392583729e-06, "loss": 0.7334, "step": 7922 }, { "epoch": 0.94, "learning_rate": 4.011651396687845e-06, "loss": 0.9633, "step": 7923 }, { "epoch": 0.94, "learning_rate": 4.011395375804238e-06, "loss": 1.1091, "step": 7924 }, { "epoch": 0.94, "learning_rate": 4.011139329937139e-06, "loss": 1.0645, "step": 7925 }, { "epoch": 0.94, "learning_rate": 4.010883259090779e-06, "loss": 0.5551, "step": 7926 }, { "epoch": 0.94, "learning_rate": 4.010627163269392e-06, "loss": 0.7471, "step": 7927 }, { "epoch": 0.94, "learning_rate": 4.010371042477212e-06, "loss": 0.8082, "step": 7928 }, { "epoch": 0.94, "learning_rate": 4.010114896718471e-06, "loss": 0.9218, "step": 7929 }, { "epoch": 0.94, "learning_rate": 4.009858725997405e-06, "loss": 0.9691, "step": 7930 }, { "epoch": 0.94, "learning_rate": 4.0096025303182466e-06, "loss": 1.1213, "step": 7931 }, { "epoch": 0.94, "learning_rate": 4.009346309685233e-06, "loss": 0.582, "step": 7932 }, { "epoch": 0.94, "learning_rate": 4.009090064102597e-06, "loss": 0.674, "step": 7933 }, { "epoch": 0.94, "learning_rate": 4.008833793574576e-06, "loss": 0.6575, "step": 7934 }, { "epoch": 0.94, "learning_rate": 4.008577498105405e-06, "loss": 0.7657, "step": 7935 }, { "epoch": 0.95, "learning_rate": 4.0083211776993225e-06, "loss": 0.9689, "step": 7936 }, { "epoch": 0.95, "learning_rate": 4.008064832360564e-06, "loss": 1.1618, "step": 7937 }, { "epoch": 0.95, "learning_rate": 4.007808462093368e-06, "loss": 0.6717, "step": 7938 }, { "epoch": 0.95, "learning_rate": 4.007552066901971e-06, "loss": 0.6439, "step": 7939 }, { "epoch": 0.95, "learning_rate": 4.007295646790611e-06, "loss": 0.7214, "step": 7940 }, { "epoch": 0.95, "learning_rate": 4.007039201763528e-06, "loss": 0.6637, "step": 7941 }, { "epoch": 0.95, "learning_rate": 4.0067827318249614e-06, "loss": 1.0128, "step": 7942 }, { "epoch": 0.95, "learning_rate": 4.006526236979149e-06, "loss": 1.0641, "step": 7943 }, { "epoch": 0.95, "learning_rate": 4.006269717230332e-06, "loss": 0.7938, "step": 7944 }, { "epoch": 0.95, "learning_rate": 4.00601317258275e-06, "loss": 0.6744, "step": 7945 }, { "epoch": 0.95, "learning_rate": 4.005756603040643e-06, "loss": 0.7728, "step": 7946 }, { "epoch": 0.95, "learning_rate": 4.005500008608254e-06, "loss": 0.7437, "step": 7947 }, { "epoch": 0.95, "learning_rate": 4.005243389289823e-06, "loss": 0.9413, "step": 7948 }, { "epoch": 0.95, "learning_rate": 4.0049867450895935e-06, "loss": 1.1068, "step": 7949 }, { "epoch": 0.95, "learning_rate": 4.004730076011806e-06, "loss": 1.1182, "step": 7950 }, { "epoch": 0.95, "learning_rate": 4.004473382060705e-06, "loss": 0.5758, "step": 7951 }, { "epoch": 0.95, "learning_rate": 4.004216663240531e-06, "loss": 0.5754, "step": 7952 }, { "epoch": 0.95, "learning_rate": 4.003959919555531e-06, "loss": 0.7516, "step": 7953 }, { "epoch": 0.95, "learning_rate": 4.003703151009946e-06, "loss": 0.8095, "step": 7954 }, { "epoch": 0.95, "learning_rate": 4.003446357608023e-06, "loss": 1.0477, "step": 7955 }, { "epoch": 0.95, "learning_rate": 4.003189539354005e-06, "loss": 1.087, "step": 7956 }, { "epoch": 0.95, "learning_rate": 4.002932696252137e-06, "loss": 0.506, "step": 7957 }, { "epoch": 0.95, "learning_rate": 4.002675828306667e-06, "loss": 0.558, "step": 7958 }, { "epoch": 0.95, "learning_rate": 4.002418935521838e-06, "loss": 0.7987, "step": 7959 }, { "epoch": 0.95, "learning_rate": 4.0021620179019e-06, "loss": 0.8303, "step": 7960 }, { "epoch": 0.95, "learning_rate": 4.0019050754510955e-06, "loss": 0.9982, "step": 7961 }, { "epoch": 0.95, "learning_rate": 4.001648108173675e-06, "loss": 1.0563, "step": 7962 }, { "epoch": 0.95, "learning_rate": 4.001391116073885e-06, "loss": 0.7259, "step": 7963 }, { "epoch": 0.95, "learning_rate": 4.001134099155974e-06, "loss": 0.5648, "step": 7964 }, { "epoch": 0.95, "learning_rate": 4.000877057424191e-06, "loss": 0.6748, "step": 7965 }, { "epoch": 0.95, "learning_rate": 4.000619990882784e-06, "loss": 0.8192, "step": 7966 }, { "epoch": 0.95, "learning_rate": 4.0003628995360025e-06, "loss": 0.8283, "step": 7967 }, { "epoch": 0.95, "learning_rate": 4.0001057833880965e-06, "loss": 0.9896, "step": 7968 }, { "epoch": 0.95, "learning_rate": 3.999848642443315e-06, "loss": 0.9326, "step": 7969 }, { "epoch": 0.95, "learning_rate": 3.999591476705911e-06, "loss": 0.5878, "step": 7970 }, { "epoch": 0.95, "learning_rate": 3.9993342861801335e-06, "loss": 0.822, "step": 7971 }, { "epoch": 0.95, "learning_rate": 3.9990770708702345e-06, "loss": 0.7967, "step": 7972 }, { "epoch": 0.95, "learning_rate": 3.998819830780465e-06, "loss": 0.8164, "step": 7973 }, { "epoch": 0.95, "learning_rate": 3.998562565915079e-06, "loss": 1.1311, "step": 7974 }, { "epoch": 0.95, "learning_rate": 3.998305276278328e-06, "loss": 1.1072, "step": 7975 }, { "epoch": 0.95, "learning_rate": 3.998047961874465e-06, "loss": 0.4602, "step": 7976 }, { "epoch": 0.95, "learning_rate": 3.997790622707743e-06, "loss": 0.811, "step": 7977 }, { "epoch": 0.95, "learning_rate": 3.997533258782417e-06, "loss": 0.6877, "step": 7978 }, { "epoch": 0.95, "learning_rate": 3.99727587010274e-06, "loss": 0.9269, "step": 7979 }, { "epoch": 0.95, "learning_rate": 3.997018456672967e-06, "loss": 1.0923, "step": 7980 }, { "epoch": 0.95, "learning_rate": 3.9967610184973534e-06, "loss": 1.0951, "step": 7981 }, { "epoch": 0.95, "learning_rate": 3.996503555580155e-06, "loss": 0.583, "step": 7982 }, { "epoch": 0.95, "learning_rate": 3.996246067925627e-06, "loss": 0.6507, "step": 7983 }, { "epoch": 0.95, "learning_rate": 3.9959885555380265e-06, "loss": 0.7997, "step": 7984 }, { "epoch": 0.95, "learning_rate": 3.995731018421609e-06, "loss": 0.7231, "step": 7985 }, { "epoch": 0.95, "learning_rate": 3.995473456580632e-06, "loss": 1.0627, "step": 7986 }, { "epoch": 0.95, "learning_rate": 3.995215870019354e-06, "loss": 1.1868, "step": 7987 }, { "epoch": 0.95, "learning_rate": 3.994958258742032e-06, "loss": 0.6755, "step": 7988 }, { "epoch": 0.95, "learning_rate": 3.994700622752925e-06, "loss": 0.724, "step": 7989 }, { "epoch": 0.95, "learning_rate": 3.9944429620562904e-06, "loss": 0.554, "step": 7990 }, { "epoch": 0.95, "learning_rate": 3.994185276656389e-06, "loss": 0.9002, "step": 7991 }, { "epoch": 0.95, "learning_rate": 3.993927566557479e-06, "loss": 0.8557, "step": 7992 }, { "epoch": 0.95, "learning_rate": 3.993669831763821e-06, "loss": 1.0566, "step": 7993 }, { "epoch": 0.95, "learning_rate": 3.993412072279676e-06, "loss": 0.7506, "step": 7994 }, { "epoch": 0.95, "learning_rate": 3.993154288109303e-06, "loss": 0.5494, "step": 7995 }, { "epoch": 0.95, "learning_rate": 3.992896479256966e-06, "loss": 0.6835, "step": 7996 }, { "epoch": 0.95, "learning_rate": 3.992638645726924e-06, "loss": 0.7054, "step": 7997 }, { "epoch": 0.95, "learning_rate": 3.99238078752344e-06, "loss": 0.819, "step": 7998 }, { "epoch": 0.95, "learning_rate": 3.992122904650776e-06, "loss": 0.9577, "step": 7999 }, { "epoch": 0.95, "learning_rate": 3.991864997113196e-06, "loss": 1.1131, "step": 8000 }, { "epoch": 0.95, "learning_rate": 3.991607064914961e-06, "loss": 0.5557, "step": 8001 }, { "epoch": 0.95, "learning_rate": 3.991349108060338e-06, "loss": 0.7762, "step": 8002 }, { "epoch": 0.95, "learning_rate": 3.991091126553588e-06, "loss": 0.775, "step": 8003 }, { "epoch": 0.95, "learning_rate": 3.990833120398976e-06, "loss": 1.0393, "step": 8004 }, { "epoch": 0.95, "learning_rate": 3.9905750896007676e-06, "loss": 1.0193, "step": 8005 }, { "epoch": 0.95, "learning_rate": 3.9903170341632295e-06, "loss": 1.2026, "step": 8006 }, { "epoch": 0.95, "learning_rate": 3.990058954090624e-06, "loss": 0.642, "step": 8007 }, { "epoch": 0.95, "learning_rate": 3.98980084938722e-06, "loss": 0.5684, "step": 8008 }, { "epoch": 0.95, "learning_rate": 3.989542720057282e-06, "loss": 0.7199, "step": 8009 }, { "epoch": 0.95, "learning_rate": 3.989284566105079e-06, "loss": 0.7617, "step": 8010 }, { "epoch": 0.95, "learning_rate": 3.989026387534877e-06, "loss": 0.9842, "step": 8011 }, { "epoch": 0.95, "learning_rate": 3.988768184350943e-06, "loss": 1.0975, "step": 8012 }, { "epoch": 0.95, "learning_rate": 3.988509956557546e-06, "loss": 0.4801, "step": 8013 }, { "epoch": 0.95, "learning_rate": 3.988251704158956e-06, "loss": 0.7189, "step": 8014 }, { "epoch": 0.95, "learning_rate": 3.987993427159439e-06, "loss": 0.7265, "step": 8015 }, { "epoch": 0.95, "learning_rate": 3.987735125563266e-06, "loss": 0.7118, "step": 8016 }, { "epoch": 0.95, "learning_rate": 3.987476799374707e-06, "loss": 0.8771, "step": 8017 }, { "epoch": 0.95, "learning_rate": 3.987218448598031e-06, "loss": 1.2106, "step": 8018 }, { "epoch": 0.95, "learning_rate": 3.98696007323751e-06, "loss": 0.8259, "step": 8019 }, { "epoch": 0.96, "learning_rate": 3.9867016732974135e-06, "loss": 0.4519, "step": 8020 }, { "epoch": 0.96, "learning_rate": 3.9864432487820145e-06, "loss": 0.7949, "step": 8021 }, { "epoch": 0.96, "learning_rate": 3.986184799695583e-06, "loss": 0.7206, "step": 8022 }, { "epoch": 0.96, "learning_rate": 3.985926326042393e-06, "loss": 0.8745, "step": 8023 }, { "epoch": 0.96, "learning_rate": 3.985667827826716e-06, "loss": 1.0499, "step": 8024 }, { "epoch": 0.96, "learning_rate": 3.9854093050528245e-06, "loss": 1.0607, "step": 8025 }, { "epoch": 0.96, "learning_rate": 3.985150757724993e-06, "loss": 0.5417, "step": 8026 }, { "epoch": 0.96, "learning_rate": 3.984892185847494e-06, "loss": 0.6283, "step": 8027 }, { "epoch": 0.96, "learning_rate": 3.984633589424604e-06, "loss": 0.7249, "step": 8028 }, { "epoch": 0.96, "learning_rate": 3.984374968460595e-06, "loss": 0.8469, "step": 8029 }, { "epoch": 0.96, "learning_rate": 3.984116322959744e-06, "loss": 1.0352, "step": 8030 }, { "epoch": 0.96, "learning_rate": 3.983857652926325e-06, "loss": 1.3213, "step": 8031 }, { "epoch": 0.96, "learning_rate": 3.983598958364615e-06, "loss": 0.4592, "step": 8032 }, { "epoch": 0.96, "learning_rate": 3.9833402392788886e-06, "loss": 0.5522, "step": 8033 }, { "epoch": 0.96, "learning_rate": 3.9830814956734254e-06, "loss": 0.6737, "step": 8034 }, { "epoch": 0.96, "learning_rate": 3.9828227275524994e-06, "loss": 0.7708, "step": 8035 }, { "epoch": 0.96, "learning_rate": 3.98256393492039e-06, "loss": 1.0041, "step": 8036 }, { "epoch": 0.96, "learning_rate": 3.982305117781374e-06, "loss": 1.1451, "step": 8037 }, { "epoch": 0.96, "learning_rate": 3.98204627613973e-06, "loss": 0.5516, "step": 8038 }, { "epoch": 0.96, "learning_rate": 3.981787409999737e-06, "loss": 0.6167, "step": 8039 }, { "epoch": 0.96, "learning_rate": 3.981528519365673e-06, "loss": 0.6959, "step": 8040 }, { "epoch": 0.96, "learning_rate": 3.981269604241819e-06, "loss": 0.7684, "step": 8041 }, { "epoch": 0.96, "learning_rate": 3.981010664632454e-06, "loss": 0.8094, "step": 8042 }, { "epoch": 0.96, "learning_rate": 3.980751700541859e-06, "loss": 1.0596, "step": 8043 }, { "epoch": 0.96, "learning_rate": 3.980492711974313e-06, "loss": 0.744, "step": 8044 }, { "epoch": 0.96, "learning_rate": 3.980233698934098e-06, "loss": 0.6589, "step": 8045 }, { "epoch": 0.96, "learning_rate": 3.979974661425497e-06, "loss": 0.7808, "step": 8046 }, { "epoch": 0.96, "learning_rate": 3.9797155994527905e-06, "loss": 0.7598, "step": 8047 }, { "epoch": 0.96, "learning_rate": 3.979456513020261e-06, "loss": 0.7982, "step": 8048 }, { "epoch": 0.96, "learning_rate": 3.97919740213219e-06, "loss": 1.1082, "step": 8049 }, { "epoch": 0.96, "learning_rate": 3.978938266792863e-06, "loss": 1.1424, "step": 8050 }, { "epoch": 0.96, "learning_rate": 3.978679107006562e-06, "loss": 0.643, "step": 8051 }, { "epoch": 0.96, "learning_rate": 3.978419922777572e-06, "loss": 0.6711, "step": 8052 }, { "epoch": 0.96, "learning_rate": 3.978160714110176e-06, "loss": 0.7046, "step": 8053 }, { "epoch": 0.96, "learning_rate": 3.9779014810086595e-06, "loss": 0.8325, "step": 8054 }, { "epoch": 0.96, "learning_rate": 3.9776422234773075e-06, "loss": 1.0291, "step": 8055 }, { "epoch": 0.96, "learning_rate": 3.977382941520405e-06, "loss": 1.1445, "step": 8056 }, { "epoch": 0.96, "learning_rate": 3.977123635142239e-06, "loss": 0.5784, "step": 8057 }, { "epoch": 0.96, "learning_rate": 3.9768643043470956e-06, "loss": 0.5153, "step": 8058 }, { "epoch": 0.96, "learning_rate": 3.976604949139261e-06, "loss": 0.7537, "step": 8059 }, { "epoch": 0.96, "learning_rate": 3.9763455695230225e-06, "loss": 0.7598, "step": 8060 }, { "epoch": 0.96, "learning_rate": 3.976086165502668e-06, "loss": 0.9953, "step": 8061 }, { "epoch": 0.96, "learning_rate": 3.975826737082486e-06, "loss": 1.1125, "step": 8062 }, { "epoch": 0.96, "learning_rate": 3.975567284266763e-06, "loss": 0.6007, "step": 8063 }, { "epoch": 0.96, "learning_rate": 3.97530780705979e-06, "loss": 0.6152, "step": 8064 }, { "epoch": 0.96, "learning_rate": 3.975048305465855e-06, "loss": 0.7319, "step": 8065 }, { "epoch": 0.96, "learning_rate": 3.974788779489247e-06, "loss": 0.6711, "step": 8066 }, { "epoch": 0.96, "learning_rate": 3.974529229134257e-06, "loss": 0.8224, "step": 8067 }, { "epoch": 0.96, "learning_rate": 3.974269654405175e-06, "loss": 1.1298, "step": 8068 }, { "epoch": 0.96, "learning_rate": 3.974010055306293e-06, "loss": 0.6003, "step": 8069 }, { "epoch": 0.96, "learning_rate": 3.973750431841901e-06, "loss": 0.7902, "step": 8070 }, { "epoch": 0.96, "learning_rate": 3.97349078401629e-06, "loss": 0.8666, "step": 8071 }, { "epoch": 0.96, "learning_rate": 3.973231111833752e-06, "loss": 0.743, "step": 8072 }, { "epoch": 0.96, "learning_rate": 3.972971415298582e-06, "loss": 0.8162, "step": 8073 }, { "epoch": 0.96, "learning_rate": 3.972711694415069e-06, "loss": 1.1128, "step": 8074 }, { "epoch": 0.96, "learning_rate": 3.972451949187509e-06, "loss": 1.1995, "step": 8075 }, { "epoch": 0.96, "learning_rate": 3.9721921796201945e-06, "loss": 0.5338, "step": 8076 }, { "epoch": 0.96, "learning_rate": 3.97193238571742e-06, "loss": 0.6191, "step": 8077 }, { "epoch": 0.96, "learning_rate": 3.97167256748348e-06, "loss": 0.6989, "step": 8078 }, { "epoch": 0.96, "learning_rate": 3.9714127249226685e-06, "loss": 0.8826, "step": 8079 }, { "epoch": 0.96, "learning_rate": 3.971152858039281e-06, "loss": 0.8877, "step": 8080 }, { "epoch": 0.96, "learning_rate": 3.970892966837614e-06, "loss": 1.1553, "step": 8081 }, { "epoch": 0.96, "learning_rate": 3.970633051321962e-06, "loss": 0.5706, "step": 8082 }, { "epoch": 0.96, "learning_rate": 3.970373111496623e-06, "loss": 0.6362, "step": 8083 }, { "epoch": 0.96, "learning_rate": 3.970113147365892e-06, "loss": 0.8798, "step": 8084 }, { "epoch": 0.96, "learning_rate": 3.9698531589340686e-06, "loss": 0.9244, "step": 8085 }, { "epoch": 0.96, "learning_rate": 3.969593146205449e-06, "loss": 1.0114, "step": 8086 }, { "epoch": 0.96, "learning_rate": 3.9693331091843304e-06, "loss": 1.1706, "step": 8087 }, { "epoch": 0.96, "learning_rate": 3.969073047875013e-06, "loss": 0.6037, "step": 8088 }, { "epoch": 0.96, "learning_rate": 3.968812962281795e-06, "loss": 0.7892, "step": 8089 }, { "epoch": 0.96, "learning_rate": 3.968552852408975e-06, "loss": 0.7998, "step": 8090 }, { "epoch": 0.96, "learning_rate": 3.968292718260853e-06, "loss": 0.7755, "step": 8091 }, { "epoch": 0.96, "learning_rate": 3.96803255984173e-06, "loss": 0.9621, "step": 8092 }, { "epoch": 0.96, "learning_rate": 3.9677723771559045e-06, "loss": 1.1617, "step": 8093 }, { "epoch": 0.96, "learning_rate": 3.967512170207679e-06, "loss": 0.8918, "step": 8094 }, { "epoch": 0.96, "learning_rate": 3.967251939001354e-06, "loss": 0.5994, "step": 8095 }, { "epoch": 0.96, "learning_rate": 3.966991683541233e-06, "loss": 0.6665, "step": 8096 }, { "epoch": 0.96, "learning_rate": 3.966731403831614e-06, "loss": 0.7239, "step": 8097 }, { "epoch": 0.96, "learning_rate": 3.966471099876803e-06, "loss": 0.8782, "step": 8098 }, { "epoch": 0.96, "learning_rate": 3.966210771681101e-06, "loss": 1.0558, "step": 8099 }, { "epoch": 0.96, "learning_rate": 3.965950419248813e-06, "loss": 1.2418, "step": 8100 }, { "epoch": 0.96, "learning_rate": 3.96569004258424e-06, "loss": 0.5518, "step": 8101 }, { "epoch": 0.96, "learning_rate": 3.965429641691689e-06, "loss": 0.6741, "step": 8102 }, { "epoch": 0.96, "learning_rate": 3.965169216575463e-06, "loss": 0.7219, "step": 8103 }, { "epoch": 0.97, "learning_rate": 3.964908767239866e-06, "loss": 0.7908, "step": 8104 }, { "epoch": 0.97, "learning_rate": 3.964648293689205e-06, "loss": 1.0432, "step": 8105 }, { "epoch": 0.97, "learning_rate": 3.964387795927784e-06, "loss": 0.9421, "step": 8106 }, { "epoch": 0.97, "learning_rate": 3.964127273959911e-06, "loss": 0.5743, "step": 8107 }, { "epoch": 0.97, "learning_rate": 3.96386672778989e-06, "loss": 0.6527, "step": 8108 }, { "epoch": 0.97, "learning_rate": 3.963606157422031e-06, "loss": 0.8341, "step": 8109 }, { "epoch": 0.97, "learning_rate": 3.963345562860638e-06, "loss": 0.73, "step": 8110 }, { "epoch": 0.97, "learning_rate": 3.96308494411002e-06, "loss": 0.9117, "step": 8111 }, { "epoch": 0.97, "learning_rate": 3.962824301174485e-06, "loss": 1.1648, "step": 8112 }, { "epoch": 0.97, "learning_rate": 3.962563634058342e-06, "loss": 0.701, "step": 8113 }, { "epoch": 0.97, "learning_rate": 3.962302942765899e-06, "loss": 0.612, "step": 8114 }, { "epoch": 0.97, "learning_rate": 3.962042227301467e-06, "loss": 0.6093, "step": 8115 }, { "epoch": 0.97, "learning_rate": 3.961781487669353e-06, "loss": 0.7362, "step": 8116 }, { "epoch": 0.97, "learning_rate": 3.961520723873869e-06, "loss": 0.9453, "step": 8117 }, { "epoch": 0.97, "learning_rate": 3.961259935919324e-06, "loss": 1.1496, "step": 8118 }, { "epoch": 0.97, "learning_rate": 3.9609991238100305e-06, "loss": 0.8949, "step": 8119 }, { "epoch": 0.97, "learning_rate": 3.960738287550298e-06, "loss": 0.5613, "step": 8120 }, { "epoch": 0.97, "learning_rate": 3.96047742714444e-06, "loss": 0.7289, "step": 8121 }, { "epoch": 0.97, "learning_rate": 3.960216542596768e-06, "loss": 0.6958, "step": 8122 }, { "epoch": 0.97, "learning_rate": 3.959955633911592e-06, "loss": 0.8143, "step": 8123 }, { "epoch": 0.97, "learning_rate": 3.959694701093228e-06, "loss": 1.0623, "step": 8124 }, { "epoch": 0.97, "learning_rate": 3.959433744145989e-06, "loss": 1.2359, "step": 8125 }, { "epoch": 0.97, "learning_rate": 3.959172763074186e-06, "loss": 0.7224, "step": 8126 }, { "epoch": 0.97, "learning_rate": 3.958911757882136e-06, "loss": 0.6476, "step": 8127 }, { "epoch": 0.97, "learning_rate": 3.958650728574153e-06, "loss": 0.6532, "step": 8128 }, { "epoch": 0.97, "learning_rate": 3.958389675154549e-06, "loss": 0.8267, "step": 8129 }, { "epoch": 0.97, "learning_rate": 3.958128597627643e-06, "loss": 0.9588, "step": 8130 }, { "epoch": 0.97, "learning_rate": 3.957867495997747e-06, "loss": 1.1082, "step": 8131 }, { "epoch": 0.97, "learning_rate": 3.957606370269181e-06, "loss": 0.5977, "step": 8132 }, { "epoch": 0.97, "learning_rate": 3.957345220446259e-06, "loss": 0.6982, "step": 8133 }, { "epoch": 0.97, "learning_rate": 3.957084046533297e-06, "loss": 0.6465, "step": 8134 }, { "epoch": 0.97, "learning_rate": 3.956822848534615e-06, "loss": 0.746, "step": 8135 }, { "epoch": 0.97, "learning_rate": 3.956561626454528e-06, "loss": 1.045, "step": 8136 }, { "epoch": 0.97, "learning_rate": 3.956300380297356e-06, "loss": 1.0396, "step": 8137 }, { "epoch": 0.97, "learning_rate": 3.956039110067416e-06, "loss": 0.5461, "step": 8138 }, { "epoch": 0.97, "learning_rate": 3.955777815769027e-06, "loss": 0.5887, "step": 8139 }, { "epoch": 0.97, "learning_rate": 3.9555164974065086e-06, "loss": 0.6561, "step": 8140 }, { "epoch": 0.97, "learning_rate": 3.955255154984181e-06, "loss": 0.7977, "step": 8141 }, { "epoch": 0.97, "learning_rate": 3.954993788506362e-06, "loss": 0.8403, "step": 8142 }, { "epoch": 0.97, "learning_rate": 3.954732397977376e-06, "loss": 0.9448, "step": 8143 }, { "epoch": 0.97, "learning_rate": 3.9544709834015405e-06, "loss": 0.6411, "step": 8144 }, { "epoch": 0.97, "learning_rate": 3.954209544783178e-06, "loss": 0.4858, "step": 8145 }, { "epoch": 0.97, "learning_rate": 3.953948082126608e-06, "loss": 0.7302, "step": 8146 }, { "epoch": 0.97, "learning_rate": 3.953686595436156e-06, "loss": 0.9052, "step": 8147 }, { "epoch": 0.97, "learning_rate": 3.953425084716141e-06, "loss": 0.8486, "step": 8148 }, { "epoch": 0.97, "learning_rate": 3.953163549970889e-06, "loss": 0.9937, "step": 8149 }, { "epoch": 0.97, "learning_rate": 3.95290199120472e-06, "loss": 1.0283, "step": 8150 }, { "epoch": 0.97, "learning_rate": 3.952640408421961e-06, "loss": 0.4832, "step": 8151 }, { "epoch": 0.97, "learning_rate": 3.952378801626932e-06, "loss": 0.6496, "step": 8152 }, { "epoch": 0.97, "learning_rate": 3.9521171708239615e-06, "loss": 0.8539, "step": 8153 }, { "epoch": 0.97, "learning_rate": 3.951855516017371e-06, "loss": 0.7203, "step": 8154 }, { "epoch": 0.97, "learning_rate": 3.951593837211488e-06, "loss": 1.0828, "step": 8155 }, { "epoch": 0.97, "learning_rate": 3.951332134410636e-06, "loss": 1.1959, "step": 8156 }, { "epoch": 0.97, "learning_rate": 3.951070407619143e-06, "loss": 0.6338, "step": 8157 }, { "epoch": 0.97, "learning_rate": 3.950808656841334e-06, "loss": 0.7174, "step": 8158 }, { "epoch": 0.97, "learning_rate": 3.950546882081536e-06, "loss": 0.6285, "step": 8159 }, { "epoch": 0.97, "learning_rate": 3.950285083344076e-06, "loss": 0.8058, "step": 8160 }, { "epoch": 0.97, "learning_rate": 3.950023260633284e-06, "loss": 0.9805, "step": 8161 }, { "epoch": 0.97, "learning_rate": 3.949761413953483e-06, "loss": 1.0964, "step": 8162 }, { "epoch": 0.97, "learning_rate": 3.949499543309006e-06, "loss": 0.5656, "step": 8163 }, { "epoch": 0.97, "learning_rate": 3.94923764870418e-06, "loss": 0.6102, "step": 8164 }, { "epoch": 0.97, "learning_rate": 3.948975730143333e-06, "loss": 0.646, "step": 8165 }, { "epoch": 0.97, "learning_rate": 3.948713787630796e-06, "loss": 0.6685, "step": 8166 }, { "epoch": 0.97, "learning_rate": 3.948451821170899e-06, "loss": 0.985, "step": 8167 }, { "epoch": 0.97, "learning_rate": 3.948189830767971e-06, "loss": 1.1495, "step": 8168 }, { "epoch": 0.97, "learning_rate": 3.947927816426344e-06, "loss": 0.7728, "step": 8169 }, { "epoch": 0.97, "learning_rate": 3.947665778150349e-06, "loss": 0.6245, "step": 8170 }, { "epoch": 0.97, "learning_rate": 3.947403715944317e-06, "loss": 0.6535, "step": 8171 }, { "epoch": 0.97, "learning_rate": 3.94714162981258e-06, "loss": 0.7036, "step": 8172 }, { "epoch": 0.97, "learning_rate": 3.94687951975947e-06, "loss": 0.8427, "step": 8173 }, { "epoch": 0.97, "learning_rate": 3.946617385789321e-06, "loss": 1.0641, "step": 8174 }, { "epoch": 0.97, "learning_rate": 3.9463552279064645e-06, "loss": 1.1672, "step": 8175 }, { "epoch": 0.97, "learning_rate": 3.946093046115236e-06, "loss": 0.5676, "step": 8176 }, { "epoch": 0.97, "learning_rate": 3.945830840419966e-06, "loss": 0.6684, "step": 8177 }, { "epoch": 0.97, "learning_rate": 3.9455686108249925e-06, "loss": 0.7033, "step": 8178 }, { "epoch": 0.97, "learning_rate": 3.9453063573346486e-06, "loss": 0.8364, "step": 8179 }, { "epoch": 0.97, "learning_rate": 3.9450440799532685e-06, "loss": 0.9966, "step": 8180 }, { "epoch": 0.97, "learning_rate": 3.944781778685189e-06, "loss": 1.2303, "step": 8181 }, { "epoch": 0.97, "learning_rate": 3.944519453534746e-06, "loss": 0.6095, "step": 8182 }, { "epoch": 0.97, "learning_rate": 3.944257104506274e-06, "loss": 0.7481, "step": 8183 }, { "epoch": 0.97, "learning_rate": 3.943994731604112e-06, "loss": 0.7974, "step": 8184 }, { "epoch": 0.97, "learning_rate": 3.943732334832595e-06, "loss": 0.7145, "step": 8185 }, { "epoch": 0.97, "learning_rate": 3.943469914196062e-06, "loss": 0.9825, "step": 8186 }, { "epoch": 0.97, "learning_rate": 3.94320746969885e-06, "loss": 1.0881, "step": 8187 }, { "epoch": 0.98, "learning_rate": 3.942945001345297e-06, "loss": 0.5479, "step": 8188 }, { "epoch": 0.98, "learning_rate": 3.942682509139742e-06, "loss": 0.5594, "step": 8189 }, { "epoch": 0.98, "learning_rate": 3.942419993086526e-06, "loss": 0.772, "step": 8190 }, { "epoch": 0.98, "learning_rate": 3.942157453189984e-06, "loss": 0.8136, "step": 8191 }, { "epoch": 0.98, "learning_rate": 3.941894889454459e-06, "loss": 0.8599, "step": 8192 }, { "epoch": 0.98, "learning_rate": 3.941632301884292e-06, "loss": 0.9928, "step": 8193 }, { "epoch": 0.98, "learning_rate": 3.94136969048382e-06, "loss": 0.7273, "step": 8194 }, { "epoch": 0.98, "learning_rate": 3.941107055257387e-06, "loss": 0.6718, "step": 8195 }, { "epoch": 0.98, "learning_rate": 3.940844396209334e-06, "loss": 0.6675, "step": 8196 }, { "epoch": 0.98, "learning_rate": 3.940581713344001e-06, "loss": 0.732, "step": 8197 }, { "epoch": 0.98, "learning_rate": 3.940319006665732e-06, "loss": 0.7963, "step": 8198 }, { "epoch": 0.98, "learning_rate": 3.94005627617887e-06, "loss": 0.982, "step": 8199 }, { "epoch": 0.98, "learning_rate": 3.939793521887756e-06, "loss": 1.2019, "step": 8200 }, { "epoch": 0.98, "learning_rate": 3.939530743796735e-06, "loss": 0.7651, "step": 8201 }, { "epoch": 0.98, "learning_rate": 3.939267941910148e-06, "loss": 0.8197, "step": 8202 }, { "epoch": 0.98, "learning_rate": 3.939005116232343e-06, "loss": 0.8101, "step": 8203 }, { "epoch": 0.98, "learning_rate": 3.938742266767662e-06, "loss": 0.9241, "step": 8204 }, { "epoch": 0.98, "learning_rate": 3.938479393520451e-06, "loss": 1.0321, "step": 8205 }, { "epoch": 0.98, "learning_rate": 3.938216496495054e-06, "loss": 1.0891, "step": 8206 }, { "epoch": 0.98, "learning_rate": 3.937953575695819e-06, "loss": 0.7868, "step": 8207 }, { "epoch": 0.98, "learning_rate": 3.93769063112709e-06, "loss": 0.731, "step": 8208 }, { "epoch": 0.98, "learning_rate": 3.937427662793214e-06, "loss": 0.8092, "step": 8209 }, { "epoch": 0.98, "learning_rate": 3.937164670698538e-06, "loss": 0.826, "step": 8210 }, { "epoch": 0.98, "learning_rate": 3.9369016548474105e-06, "loss": 0.9535, "step": 8211 }, { "epoch": 0.98, "learning_rate": 3.936638615244177e-06, "loss": 1.199, "step": 8212 }, { "epoch": 0.98, "learning_rate": 3.936375551893187e-06, "loss": 0.3984, "step": 8213 }, { "epoch": 0.98, "learning_rate": 3.936112464798789e-06, "loss": 0.5604, "step": 8214 }, { "epoch": 0.98, "learning_rate": 3.935849353965331e-06, "loss": 0.7112, "step": 8215 }, { "epoch": 0.98, "learning_rate": 3.935586219397163e-06, "loss": 0.8298, "step": 8216 }, { "epoch": 0.98, "learning_rate": 3.9353230610986335e-06, "loss": 1.0152, "step": 8217 }, { "epoch": 0.98, "learning_rate": 3.935059879074094e-06, "loss": 1.0912, "step": 8218 }, { "epoch": 0.98, "learning_rate": 3.934796673327894e-06, "loss": 0.5724, "step": 8219 }, { "epoch": 0.98, "learning_rate": 3.934533443864386e-06, "loss": 0.5539, "step": 8220 }, { "epoch": 0.98, "learning_rate": 3.934270190687918e-06, "loss": 0.7182, "step": 8221 }, { "epoch": 0.98, "learning_rate": 3.934006913802844e-06, "loss": 0.8035, "step": 8222 }, { "epoch": 0.98, "learning_rate": 3.933743613213516e-06, "loss": 0.9002, "step": 8223 }, { "epoch": 0.98, "learning_rate": 3.933480288924284e-06, "loss": 0.9172, "step": 8224 }, { "epoch": 0.98, "learning_rate": 3.933216940939504e-06, "loss": 0.9562, "step": 8225 }, { "epoch": 0.98, "learning_rate": 3.932953569263528e-06, "loss": 0.5633, "step": 8226 }, { "epoch": 0.98, "learning_rate": 3.932690173900708e-06, "loss": 0.621, "step": 8227 }, { "epoch": 0.98, "learning_rate": 3.932426754855401e-06, "loss": 0.7855, "step": 8228 }, { "epoch": 0.98, "learning_rate": 3.932163312131958e-06, "loss": 0.9217, "step": 8229 }, { "epoch": 0.98, "learning_rate": 3.931899845734736e-06, "loss": 1.0491, "step": 8230 }, { "epoch": 0.98, "learning_rate": 3.931636355668089e-06, "loss": 1.1649, "step": 8231 }, { "epoch": 0.98, "learning_rate": 3.931372841936373e-06, "loss": 0.6588, "step": 8232 }, { "epoch": 0.98, "learning_rate": 3.9311093045439445e-06, "loss": 0.6801, "step": 8233 }, { "epoch": 0.98, "learning_rate": 3.930845743495159e-06, "loss": 0.7278, "step": 8234 }, { "epoch": 0.98, "learning_rate": 3.930582158794373e-06, "loss": 0.7186, "step": 8235 }, { "epoch": 0.98, "learning_rate": 3.9303185504459426e-06, "loss": 0.9481, "step": 8236 }, { "epoch": 0.98, "learning_rate": 3.930054918454228e-06, "loss": 1.1715, "step": 8237 }, { "epoch": 0.98, "learning_rate": 3.929791262823585e-06, "loss": 0.5388, "step": 8238 }, { "epoch": 0.98, "learning_rate": 3.9295275835583735e-06, "loss": 0.7153, "step": 8239 }, { "epoch": 0.98, "learning_rate": 3.92926388066295e-06, "loss": 0.7156, "step": 8240 }, { "epoch": 0.98, "learning_rate": 3.929000154141675e-06, "loss": 0.713, "step": 8241 }, { "epoch": 0.98, "learning_rate": 3.928736403998907e-06, "loss": 0.8361, "step": 8242 }, { "epoch": 0.98, "learning_rate": 3.928472630239008e-06, "loss": 0.9948, "step": 8243 }, { "epoch": 0.98, "learning_rate": 3.928208832866335e-06, "loss": 0.8767, "step": 8244 }, { "epoch": 0.98, "learning_rate": 3.92794501188525e-06, "loss": 0.6157, "step": 8245 }, { "epoch": 0.98, "learning_rate": 3.927681167300115e-06, "loss": 0.7764, "step": 8246 }, { "epoch": 0.98, "learning_rate": 3.9274172991152905e-06, "loss": 0.7208, "step": 8247 }, { "epoch": 0.98, "learning_rate": 3.927153407335138e-06, "loss": 0.8562, "step": 8248 }, { "epoch": 0.98, "learning_rate": 3.92688949196402e-06, "loss": 0.941, "step": 8249 }, { "epoch": 0.98, "learning_rate": 3.926625553006299e-06, "loss": 1.3032, "step": 8250 }, { "epoch": 0.98, "learning_rate": 3.926361590466337e-06, "loss": 0.6261, "step": 8251 }, { "epoch": 0.98, "learning_rate": 3.9260976043485e-06, "loss": 0.6523, "step": 8252 }, { "epoch": 0.98, "learning_rate": 3.925833594657149e-06, "loss": 0.5844, "step": 8253 }, { "epoch": 0.98, "learning_rate": 3.92556956139665e-06, "loss": 0.821, "step": 8254 }, { "epoch": 0.98, "learning_rate": 3.925305504571364e-06, "loss": 1.011, "step": 8255 }, { "epoch": 0.98, "learning_rate": 3.92504142418566e-06, "loss": 1.2334, "step": 8256 }, { "epoch": 0.98, "learning_rate": 3.924777320243902e-06, "loss": 0.5365, "step": 8257 }, { "epoch": 0.98, "learning_rate": 3.924513192750456e-06, "loss": 0.6503, "step": 8258 }, { "epoch": 0.98, "learning_rate": 3.924249041709686e-06, "loss": 0.867, "step": 8259 }, { "epoch": 0.98, "learning_rate": 3.923984867125959e-06, "loss": 0.8127, "step": 8260 }, { "epoch": 0.98, "learning_rate": 3.923720669003645e-06, "loss": 1.0377, "step": 8261 }, { "epoch": 0.98, "learning_rate": 3.923456447347108e-06, "loss": 1.1592, "step": 8262 }, { "epoch": 0.98, "learning_rate": 3.923192202160716e-06, "loss": 0.5586, "step": 8263 }, { "epoch": 0.98, "learning_rate": 3.922927933448838e-06, "loss": 0.6169, "step": 8264 }, { "epoch": 0.98, "learning_rate": 3.922663641215841e-06, "loss": 0.6464, "step": 8265 }, { "epoch": 0.98, "learning_rate": 3.922399325466096e-06, "loss": 0.7725, "step": 8266 }, { "epoch": 0.98, "learning_rate": 3.9221349862039695e-06, "loss": 0.8953, "step": 8267 }, { "epoch": 0.98, "learning_rate": 3.921870623433834e-06, "loss": 1.1086, "step": 8268 }, { "epoch": 0.98, "learning_rate": 3.921606237160056e-06, "loss": 0.7231, "step": 8269 }, { "epoch": 0.98, "learning_rate": 3.921341827387009e-06, "loss": 0.4399, "step": 8270 }, { "epoch": 0.98, "learning_rate": 3.921077394119062e-06, "loss": 0.6543, "step": 8271 }, { "epoch": 0.99, "learning_rate": 3.920812937360585e-06, "loss": 0.7375, "step": 8272 }, { "epoch": 0.99, "learning_rate": 3.920548457115953e-06, "loss": 0.7564, "step": 8273 }, { "epoch": 0.99, "learning_rate": 3.9202839533895345e-06, "loss": 0.9471, "step": 8274 }, { "epoch": 0.99, "learning_rate": 3.920019426185703e-06, "loss": 1.1071, "step": 8275 }, { "epoch": 0.99, "learning_rate": 3.919754875508832e-06, "loss": 0.6877, "step": 8276 }, { "epoch": 0.99, "learning_rate": 3.919490301363293e-06, "loss": 0.6494, "step": 8277 }, { "epoch": 0.99, "learning_rate": 3.919225703753461e-06, "loss": 0.6555, "step": 8278 }, { "epoch": 0.99, "learning_rate": 3.918961082683709e-06, "loss": 0.7397, "step": 8279 }, { "epoch": 0.99, "learning_rate": 3.918696438158411e-06, "loss": 0.9586, "step": 8280 }, { "epoch": 0.99, "learning_rate": 3.918431770181941e-06, "loss": 1.125, "step": 8281 }, { "epoch": 0.99, "learning_rate": 3.918167078758676e-06, "loss": 0.4887, "step": 8282 }, { "epoch": 0.99, "learning_rate": 3.91790236389299e-06, "loss": 0.8211, "step": 8283 }, { "epoch": 0.99, "learning_rate": 3.917637625589259e-06, "loss": 0.8241, "step": 8284 }, { "epoch": 0.99, "learning_rate": 3.917372863851858e-06, "loss": 0.8336, "step": 8285 }, { "epoch": 0.99, "learning_rate": 3.917108078685166e-06, "loss": 1.0485, "step": 8286 }, { "epoch": 0.99, "learning_rate": 3.916843270093557e-06, "loss": 1.1744, "step": 8287 }, { "epoch": 0.99, "learning_rate": 3.916578438081412e-06, "loss": 0.6048, "step": 8288 }, { "epoch": 0.99, "learning_rate": 3.916313582653105e-06, "loss": 0.7031, "step": 8289 }, { "epoch": 0.99, "learning_rate": 3.9160487038130166e-06, "loss": 0.7996, "step": 8290 }, { "epoch": 0.99, "learning_rate": 3.915783801565524e-06, "loss": 0.6357, "step": 8291 }, { "epoch": 0.99, "learning_rate": 3.915518875915006e-06, "loss": 0.8487, "step": 8292 }, { "epoch": 0.99, "learning_rate": 3.9152539268658425e-06, "loss": 1.1406, "step": 8293 }, { "epoch": 0.99, "learning_rate": 3.914988954422413e-06, "loss": 0.8883, "step": 8294 }, { "epoch": 0.99, "learning_rate": 3.914723958589097e-06, "loss": 0.597, "step": 8295 }, { "epoch": 0.99, "learning_rate": 3.914458939370275e-06, "loss": 0.6719, "step": 8296 }, { "epoch": 0.99, "learning_rate": 3.914193896770329e-06, "loss": 0.7905, "step": 8297 }, { "epoch": 0.99, "learning_rate": 3.913928830793638e-06, "loss": 0.7988, "step": 8298 }, { "epoch": 0.99, "learning_rate": 3.9136637414445855e-06, "loss": 1.0562, "step": 8299 }, { "epoch": 0.99, "learning_rate": 3.913398628727552e-06, "loss": 1.1013, "step": 8300 }, { "epoch": 0.99, "learning_rate": 3.913133492646922e-06, "loss": 0.6392, "step": 8301 }, { "epoch": 0.99, "learning_rate": 3.9128683332070755e-06, "loss": 0.7435, "step": 8302 }, { "epoch": 0.99, "learning_rate": 3.912603150412398e-06, "loss": 0.7118, "step": 8303 }, { "epoch": 0.99, "learning_rate": 3.9123379442672705e-06, "loss": 0.7776, "step": 8304 }, { "epoch": 0.99, "learning_rate": 3.91207271477608e-06, "loss": 1.0381, "step": 8305 }, { "epoch": 0.99, "learning_rate": 3.911807461943208e-06, "loss": 1.1913, "step": 8306 }, { "epoch": 0.99, "learning_rate": 3.91154218577304e-06, "loss": 0.3767, "step": 8307 }, { "epoch": 0.99, "learning_rate": 3.911276886269962e-06, "loss": 0.6437, "step": 8308 }, { "epoch": 0.99, "learning_rate": 3.911011563438357e-06, "loss": 0.6838, "step": 8309 }, { "epoch": 0.99, "learning_rate": 3.910746217282615e-06, "loss": 0.8021, "step": 8310 }, { "epoch": 0.99, "learning_rate": 3.910480847807118e-06, "loss": 1.0175, "step": 8311 }, { "epoch": 0.99, "learning_rate": 3.910215455016254e-06, "loss": 1.1653, "step": 8312 }, { "epoch": 0.99, "learning_rate": 3.909950038914411e-06, "loss": 0.6672, "step": 8313 }, { "epoch": 0.99, "learning_rate": 3.909684599505976e-06, "loss": 0.7323, "step": 8314 }, { "epoch": 0.99, "learning_rate": 3.909419136795335e-06, "loss": 0.666, "step": 8315 }, { "epoch": 0.99, "learning_rate": 3.909153650786878e-06, "loss": 0.7287, "step": 8316 }, { "epoch": 0.99, "learning_rate": 3.9088881414849925e-06, "loss": 1.0025, "step": 8317 }, { "epoch": 0.99, "learning_rate": 3.908622608894069e-06, "loss": 1.1208, "step": 8318 }, { "epoch": 0.99, "learning_rate": 3.908357053018495e-06, "loss": 0.8216, "step": 8319 }, { "epoch": 0.99, "learning_rate": 3.9080914738626605e-06, "loss": 0.5982, "step": 8320 }, { "epoch": 0.99, "learning_rate": 3.9078258714309556e-06, "loss": 0.6957, "step": 8321 }, { "epoch": 0.99, "learning_rate": 3.907560245727771e-06, "loss": 0.832, "step": 8322 }, { "epoch": 0.99, "learning_rate": 3.9072945967574975e-06, "loss": 0.9232, "step": 8323 }, { "epoch": 0.99, "learning_rate": 3.907028924524526e-06, "loss": 1.2015, "step": 8324 }, { "epoch": 0.99, "learning_rate": 3.906763229033249e-06, "loss": 1.1324, "step": 8325 }, { "epoch": 0.99, "learning_rate": 3.9064975102880575e-06, "loss": 0.6274, "step": 8326 }, { "epoch": 0.99, "learning_rate": 3.9062317682933445e-06, "loss": 0.697, "step": 8327 }, { "epoch": 0.99, "learning_rate": 3.905966003053502e-06, "loss": 0.8585, "step": 8328 }, { "epoch": 0.99, "learning_rate": 3.905700214572924e-06, "loss": 0.9487, "step": 8329 }, { "epoch": 0.99, "learning_rate": 3.905434402856002e-06, "loss": 0.8734, "step": 8330 }, { "epoch": 0.99, "learning_rate": 3.905168567907133e-06, "loss": 1.1145, "step": 8331 }, { "epoch": 0.99, "learning_rate": 3.904902709730708e-06, "loss": 0.717, "step": 8332 }, { "epoch": 0.99, "learning_rate": 3.904636828331125e-06, "loss": 0.7352, "step": 8333 }, { "epoch": 0.99, "learning_rate": 3.9043709237127765e-06, "loss": 0.8888, "step": 8334 }, { "epoch": 0.99, "learning_rate": 3.904104995880059e-06, "loss": 0.8015, "step": 8335 }, { "epoch": 0.99, "learning_rate": 3.903839044837368e-06, "loss": 0.8178, "step": 8336 }, { "epoch": 0.99, "learning_rate": 3.9035730705891e-06, "loss": 0.9888, "step": 8337 }, { "epoch": 0.99, "learning_rate": 3.903307073139651e-06, "loss": 0.5537, "step": 8338 }, { "epoch": 0.99, "learning_rate": 3.9030410524934196e-06, "loss": 0.6082, "step": 8339 }, { "epoch": 0.99, "learning_rate": 3.902775008654801e-06, "loss": 0.6606, "step": 8340 }, { "epoch": 0.99, "learning_rate": 3.902508941628193e-06, "loss": 0.79, "step": 8341 }, { "epoch": 0.99, "learning_rate": 3.902242851417996e-06, "loss": 0.7978, "step": 8342 }, { "epoch": 0.99, "learning_rate": 3.901976738028605e-06, "loss": 1.0959, "step": 8343 }, { "epoch": 0.99, "learning_rate": 3.901710601464422e-06, "loss": 0.6136, "step": 8344 }, { "epoch": 0.99, "learning_rate": 3.901444441729846e-06, "loss": 0.5025, "step": 8345 }, { "epoch": 0.99, "learning_rate": 3.901178258829275e-06, "loss": 0.6978, "step": 8346 }, { "epoch": 0.99, "learning_rate": 3.90091205276711e-06, "loss": 0.7799, "step": 8347 }, { "epoch": 0.99, "learning_rate": 3.900645823547751e-06, "loss": 0.7658, "step": 8348 }, { "epoch": 0.99, "learning_rate": 3.900379571175599e-06, "loss": 1.0771, "step": 8349 }, { "epoch": 0.99, "learning_rate": 3.900113295655056e-06, "loss": 1.0976, "step": 8350 }, { "epoch": 0.99, "learning_rate": 3.899846996990521e-06, "loss": 0.5335, "step": 8351 }, { "epoch": 0.99, "learning_rate": 3.899580675186399e-06, "loss": 0.7292, "step": 8352 }, { "epoch": 0.99, "learning_rate": 3.899314330247091e-06, "loss": 0.8143, "step": 8353 }, { "epoch": 0.99, "learning_rate": 3.899047962177e-06, "loss": 0.8154, "step": 8354 }, { "epoch": 0.99, "learning_rate": 3.8987815709805276e-06, "loss": 0.9271, "step": 8355 }, { "epoch": 1.0, "learning_rate": 3.898515156662079e-06, "loss": 1.0474, "step": 8356 }, { "epoch": 1.0, "learning_rate": 3.8982487192260574e-06, "loss": 0.7803, "step": 8357 }, { "epoch": 1.0, "learning_rate": 3.897982258676867e-06, "loss": 0.6317, "step": 8358 }, { "epoch": 1.0, "learning_rate": 3.8977157750189135e-06, "loss": 0.7823, "step": 8359 }, { "epoch": 1.0, "learning_rate": 3.8974492682566e-06, "loss": 0.7811, "step": 8360 }, { "epoch": 1.0, "learning_rate": 3.897182738394334e-06, "loss": 0.9457, "step": 8361 }, { "epoch": 1.0, "learning_rate": 3.896916185436519e-06, "loss": 1.1165, "step": 8362 }, { "epoch": 1.0, "learning_rate": 3.896649609387562e-06, "loss": 0.452, "step": 8363 }, { "epoch": 1.0, "learning_rate": 3.896383010251871e-06, "loss": 0.706, "step": 8364 }, { "epoch": 1.0, "learning_rate": 3.89611638803385e-06, "loss": 0.7354, "step": 8365 }, { "epoch": 1.0, "learning_rate": 3.895849742737909e-06, "loss": 0.7836, "step": 8366 }, { "epoch": 1.0, "learning_rate": 3.895583074368455e-06, "loss": 0.7994, "step": 8367 }, { "epoch": 1.0, "learning_rate": 3.895316382929896e-06, "loss": 1.0659, "step": 8368 }, { "epoch": 1.0, "learning_rate": 3.89504966842664e-06, "loss": 0.8085, "step": 8369 }, { "epoch": 1.0, "learning_rate": 3.8947829308630945e-06, "loss": 0.5924, "step": 8370 }, { "epoch": 1.0, "learning_rate": 3.894516170243671e-06, "loss": 0.7686, "step": 8371 }, { "epoch": 1.0, "learning_rate": 3.894249386572778e-06, "loss": 0.7361, "step": 8372 }, { "epoch": 1.0, "learning_rate": 3.8939825798548256e-06, "loss": 0.7758, "step": 8373 }, { "epoch": 1.0, "learning_rate": 3.893715750094225e-06, "loss": 1.0977, "step": 8374 }, { "epoch": 1.0, "learning_rate": 3.893448897295385e-06, "loss": 1.0829, "step": 8375 }, { "epoch": 1.0, "learning_rate": 3.893182021462718e-06, "loss": 0.5486, "step": 8376 }, { "epoch": 1.0, "learning_rate": 3.892915122600635e-06, "loss": 0.6915, "step": 8377 }, { "epoch": 1.0, "learning_rate": 3.892648200713549e-06, "loss": 0.8365, "step": 8378 }, { "epoch": 1.0, "learning_rate": 3.892381255805872e-06, "loss": 0.9946, "step": 8379 }, { "epoch": 1.0, "learning_rate": 3.892114287882015e-06, "loss": 1.0254, "step": 8380 }, { "epoch": 1.0, "learning_rate": 3.891847296946392e-06, "loss": 1.2361, "step": 8381 }, { "epoch": 1.0, "learning_rate": 3.891580283003417e-06, "loss": 0.5163, "step": 8382 }, { "epoch": 1.0, "learning_rate": 3.891313246057502e-06, "loss": 0.5428, "step": 8383 }, { "epoch": 1.0, "learning_rate": 3.891046186113063e-06, "loss": 0.716, "step": 8384 }, { "epoch": 1.0, "learning_rate": 3.8907791031745135e-06, "loss": 0.9549, "step": 8385 }, { "epoch": 1.0, "learning_rate": 3.890511997246269e-06, "loss": 1.0635, "step": 8386 }, { "epoch": 1.0, "learning_rate": 3.890244868332745e-06, "loss": 1.2096, "step": 8387 }, { "epoch": 1.0, "learning_rate": 3.8899777164383566e-06, "loss": 0.6429, "step": 8388 }, { "epoch": 1.0, "learning_rate": 3.88971054156752e-06, "loss": 0.6601, "step": 8389 }, { "epoch": 1.0, "learning_rate": 3.88944334372465e-06, "loss": 0.7559, "step": 8390 }, { "epoch": 1.0, "learning_rate": 3.889176122914167e-06, "loss": 0.7706, "step": 8391 }, { "epoch": 1.0, "learning_rate": 3.8889088791404845e-06, "loss": 0.9923, "step": 8392 }, { "epoch": 1.0, "learning_rate": 3.888641612408023e-06, "loss": 1.0498, "step": 8393 }, { "epoch": 1.0, "learning_rate": 3.8883743227211975e-06, "loss": 0.7701, "step": 8394 }, { "epoch": 1.0, "learning_rate": 3.88810701008443e-06, "loss": 0.6053, "step": 8395 }, { "epoch": 1.0, "learning_rate": 3.8878396745021355e-06, "loss": 0.9841, "step": 8396 }, { "epoch": 1.0, "learning_rate": 3.887572315978735e-06, "loss": 1.1664, "step": 8397 }, { "epoch": 1.0, "learning_rate": 3.887304934518648e-06, "loss": 0.4154, "step": 8398 }, { "epoch": 1.0, "learning_rate": 3.887037530126294e-06, "loss": 0.5098, "step": 8399 }, { "epoch": 1.0, "learning_rate": 3.886770102806093e-06, "loss": 0.5768, "step": 8400 }, { "epoch": 1.0, "learning_rate": 3.886502652562467e-06, "loss": 0.6013, "step": 8401 }, { "epoch": 1.0, "learning_rate": 3.886235179399834e-06, "loss": 0.801, "step": 8402 }, { "epoch": 1.0, "learning_rate": 3.885967683322619e-06, "loss": 0.8752, "step": 8403 }, { "epoch": 1.0, "learning_rate": 3.885700164335241e-06, "loss": 0.4471, "step": 8404 }, { "epoch": 1.0, "learning_rate": 3.885432622442123e-06, "loss": 0.4466, "step": 8405 }, { "epoch": 1.0, "learning_rate": 3.885165057647687e-06, "loss": 0.5255, "step": 8406 }, { "epoch": 1.0, "learning_rate": 3.884897469956357e-06, "loss": 0.6827, "step": 8407 }, { "epoch": 1.0, "learning_rate": 3.8846298593725565e-06, "loss": 0.8232, "step": 8408 }, { "epoch": 1.0, "learning_rate": 3.884362225900706e-06, "loss": 0.9198, "step": 8409 }, { "epoch": 1.0, "learning_rate": 3.884094569545233e-06, "loss": 0.5314, "step": 8410 }, { "epoch": 1.0, "learning_rate": 3.883826890310561e-06, "loss": 0.4578, "step": 8411 }, { "epoch": 1.0, "learning_rate": 3.8835591882011146e-06, "loss": 0.5366, "step": 8412 }, { "epoch": 1.0, "learning_rate": 3.883291463221318e-06, "loss": 0.5563, "step": 8413 }, { "epoch": 1.0, "learning_rate": 3.883023715375598e-06, "loss": 0.8241, "step": 8414 }, { "epoch": 1.0, "learning_rate": 3.882755944668379e-06, "loss": 0.8541, "step": 8415 }, { "epoch": 1.0, "learning_rate": 3.882488151104089e-06, "loss": 0.6877, "step": 8416 }, { "epoch": 1.0, "learning_rate": 3.882220334687153e-06, "loss": 0.4631, "step": 8417 }, { "epoch": 1.0, "learning_rate": 3.881952495422e-06, "loss": 0.541, "step": 8418 }, { "epoch": 1.0, "learning_rate": 3.881684633313056e-06, "loss": 0.5454, "step": 8419 }, { "epoch": 1.0, "learning_rate": 3.8814167483647485e-06, "loss": 0.6429, "step": 8420 }, { "epoch": 1.0, "learning_rate": 3.881148840581507e-06, "loss": 0.6879, "step": 8421 }, { "epoch": 1.0, "learning_rate": 3.8808809099677585e-06, "loss": 0.8817, "step": 8422 }, { "epoch": 1.0, "learning_rate": 3.880612956527933e-06, "loss": 0.5547, "step": 8423 }, { "epoch": 1.0, "learning_rate": 3.88034498026646e-06, "loss": 0.5652, "step": 8424 }, { "epoch": 1.0, "learning_rate": 3.880076981187768e-06, "loss": 0.582, "step": 8425 }, { "epoch": 1.0, "learning_rate": 3.8798089592962876e-06, "loss": 0.6782, "step": 8426 }, { "epoch": 1.0, "learning_rate": 3.87954091459645e-06, "loss": 0.6739, "step": 8427 }, { "epoch": 1.0, "learning_rate": 3.879272847092685e-06, "loss": 0.794, "step": 8428 }, { "epoch": 1.0, "learning_rate": 3.879004756789424e-06, "loss": 0.48, "step": 8429 }, { "epoch": 1.0, "learning_rate": 3.878736643691099e-06, "loss": 0.5086, "step": 8430 }, { "epoch": 1.0, "learning_rate": 3.8784685078021415e-06, "loss": 0.5905, "step": 8431 }, { "epoch": 1.0, "learning_rate": 3.878200349126984e-06, "loss": 0.5716, "step": 8432 }, { "epoch": 1.0, "learning_rate": 3.877932167670058e-06, "loss": 0.7832, "step": 8433 }, { "epoch": 1.0, "learning_rate": 3.8776639634357984e-06, "loss": 0.915, "step": 8434 }, { "epoch": 1.0, "learning_rate": 3.877395736428639e-06, "loss": 0.4129, "step": 8435 }, { "epoch": 1.0, "learning_rate": 3.877127486653011e-06, "loss": 0.4329, "step": 8436 }, { "epoch": 1.0, "learning_rate": 3.8768592141133505e-06, "loss": 0.4908, "step": 8437 }, { "epoch": 1.0, "learning_rate": 3.876590918814092e-06, "loss": 0.6809, "step": 8438 }, { "epoch": 1.01, "learning_rate": 3.87632260075967e-06, "loss": 0.7781, "step": 8439 }, { "epoch": 1.01, "learning_rate": 3.876054259954519e-06, "loss": 0.8303, "step": 8440 }, { "epoch": 1.01, "learning_rate": 3.875785896403077e-06, "loss": 0.6066, "step": 8441 }, { "epoch": 1.01, "learning_rate": 3.875517510109778e-06, "loss": 0.4092, "step": 8442 }, { "epoch": 1.01, "learning_rate": 3.875249101079059e-06, "loss": 0.5218, "step": 8443 }, { "epoch": 1.01, "learning_rate": 3.874980669315357e-06, "loss": 0.5931, "step": 8444 }, { "epoch": 1.01, "learning_rate": 3.87471221482311e-06, "loss": 0.6832, "step": 8445 }, { "epoch": 1.01, "learning_rate": 3.874443737606754e-06, "loss": 0.7986, "step": 8446 }, { "epoch": 1.01, "learning_rate": 3.874175237670727e-06, "loss": 0.7451, "step": 8447 }, { "epoch": 1.01, "learning_rate": 3.87390671501947e-06, "loss": 0.3338, "step": 8448 }, { "epoch": 1.01, "learning_rate": 3.873638169657417e-06, "loss": 0.6017, "step": 8449 }, { "epoch": 1.01, "learning_rate": 3.873369601589012e-06, "loss": 0.5865, "step": 8450 }, { "epoch": 1.01, "learning_rate": 3.873101010818692e-06, "loss": 0.7054, "step": 8451 }, { "epoch": 1.01, "learning_rate": 3.872832397350896e-06, "loss": 0.7661, "step": 8452 }, { "epoch": 1.01, "learning_rate": 3.872563761190067e-06, "loss": 0.9332, "step": 8453 }, { "epoch": 1.01, "learning_rate": 3.872295102340642e-06, "loss": 0.4504, "step": 8454 }, { "epoch": 1.01, "learning_rate": 3.872026420807065e-06, "loss": 0.4865, "step": 8455 }, { "epoch": 1.01, "learning_rate": 3.8717577165937755e-06, "loss": 0.5847, "step": 8456 }, { "epoch": 1.01, "learning_rate": 3.871488989705217e-06, "loss": 0.4786, "step": 8457 }, { "epoch": 1.01, "learning_rate": 3.871220240145829e-06, "loss": 0.8617, "step": 8458 }, { "epoch": 1.01, "learning_rate": 3.870951467920056e-06, "loss": 0.8851, "step": 8459 }, { "epoch": 1.01, "learning_rate": 3.87068267303234e-06, "loss": 0.369, "step": 8460 }, { "epoch": 1.01, "learning_rate": 3.870413855487125e-06, "loss": 0.6058, "step": 8461 }, { "epoch": 1.01, "learning_rate": 3.870145015288853e-06, "loss": 0.6008, "step": 8462 }, { "epoch": 1.01, "learning_rate": 3.869876152441969e-06, "loss": 0.5258, "step": 8463 }, { "epoch": 1.01, "learning_rate": 3.869607266950917e-06, "loss": 0.692, "step": 8464 }, { "epoch": 1.01, "learning_rate": 3.869338358820144e-06, "loss": 0.8179, "step": 8465 }, { "epoch": 1.01, "learning_rate": 3.869069428054091e-06, "loss": 0.5527, "step": 8466 }, { "epoch": 1.01, "learning_rate": 3.868800474657207e-06, "loss": 0.4373, "step": 8467 }, { "epoch": 1.01, "learning_rate": 3.868531498633935e-06, "loss": 0.5572, "step": 8468 }, { "epoch": 1.01, "learning_rate": 3.868262499988723e-06, "loss": 0.501, "step": 8469 }, { "epoch": 1.01, "learning_rate": 3.867993478726018e-06, "loss": 0.6103, "step": 8470 }, { "epoch": 1.01, "learning_rate": 3.867724434850264e-06, "loss": 0.8067, "step": 8471 }, { "epoch": 1.01, "learning_rate": 3.867455368365912e-06, "loss": 0.877, "step": 8472 }, { "epoch": 1.01, "learning_rate": 3.867186279277407e-06, "loss": 0.4972, "step": 8473 }, { "epoch": 1.01, "learning_rate": 3.866917167589199e-06, "loss": 0.5357, "step": 8474 }, { "epoch": 1.01, "learning_rate": 3.8666480333057345e-06, "loss": 0.6423, "step": 8475 }, { "epoch": 1.01, "learning_rate": 3.866378876431463e-06, "loss": 0.707, "step": 8476 }, { "epoch": 1.01, "learning_rate": 3.866109696970836e-06, "loss": 0.8532, "step": 8477 }, { "epoch": 1.01, "learning_rate": 3.8658404949283e-06, "loss": 0.8443, "step": 8478 }, { "epoch": 1.01, "learning_rate": 3.865571270308306e-06, "loss": 0.5146, "step": 8479 }, { "epoch": 1.01, "learning_rate": 3.8653020231153035e-06, "loss": 0.6837, "step": 8480 }, { "epoch": 1.01, "learning_rate": 3.865032753353745e-06, "loss": 0.5756, "step": 8481 }, { "epoch": 1.01, "learning_rate": 3.86476346102808e-06, "loss": 0.6271, "step": 8482 }, { "epoch": 1.01, "learning_rate": 3.864494146142761e-06, "loss": 0.7723, "step": 8483 }, { "epoch": 1.01, "learning_rate": 3.864224808702239e-06, "loss": 0.8673, "step": 8484 }, { "epoch": 1.01, "learning_rate": 3.863955448710967e-06, "loss": 0.4644, "step": 8485 }, { "epoch": 1.01, "learning_rate": 3.863686066173397e-06, "loss": 0.568, "step": 8486 }, { "epoch": 1.01, "learning_rate": 3.863416661093981e-06, "loss": 0.4396, "step": 8487 }, { "epoch": 1.01, "learning_rate": 3.8631472334771735e-06, "loss": 0.627, "step": 8488 }, { "epoch": 1.01, "learning_rate": 3.862877783327428e-06, "loss": 0.7554, "step": 8489 }, { "epoch": 1.01, "learning_rate": 3.862608310649198e-06, "loss": 0.8271, "step": 8490 }, { "epoch": 1.01, "learning_rate": 3.86233881544694e-06, "loss": 0.4288, "step": 8491 }, { "epoch": 1.01, "learning_rate": 3.862069297725105e-06, "loss": 0.5643, "step": 8492 }, { "epoch": 1.01, "learning_rate": 3.861799757488151e-06, "loss": 0.5183, "step": 8493 }, { "epoch": 1.01, "learning_rate": 3.861530194740532e-06, "loss": 0.579, "step": 8494 }, { "epoch": 1.01, "learning_rate": 3.8612606094867055e-06, "loss": 0.6942, "step": 8495 }, { "epoch": 1.01, "learning_rate": 3.860991001731127e-06, "loss": 0.8005, "step": 8496 }, { "epoch": 1.01, "learning_rate": 3.8607213714782524e-06, "loss": 0.8004, "step": 8497 }, { "epoch": 1.01, "learning_rate": 3.86045171873254e-06, "loss": 0.472, "step": 8498 }, { "epoch": 1.01, "learning_rate": 3.860182043498446e-06, "loss": 0.5835, "step": 8499 }, { "epoch": 1.01, "learning_rate": 3.85991234578043e-06, "loss": 0.6244, "step": 8500 }, { "epoch": 1.01, "learning_rate": 3.859642625582947e-06, "loss": 0.6941, "step": 8501 }, { "epoch": 1.01, "learning_rate": 3.859372882910458e-06, "loss": 0.846, "step": 8502 }, { "epoch": 1.01, "learning_rate": 3.859103117767421e-06, "loss": 0.9241, "step": 8503 }, { "epoch": 1.01, "learning_rate": 3.858833330158297e-06, "loss": 0.47, "step": 8504 }, { "epoch": 1.01, "learning_rate": 3.858563520087542e-06, "loss": 0.6109, "step": 8505 }, { "epoch": 1.01, "learning_rate": 3.858293687559619e-06, "loss": 0.6906, "step": 8506 }, { "epoch": 1.01, "learning_rate": 3.858023832578986e-06, "loss": 0.6003, "step": 8507 }, { "epoch": 1.01, "learning_rate": 3.8577539551501075e-06, "loss": 0.7035, "step": 8508 }, { "epoch": 1.01, "learning_rate": 3.857484055277441e-06, "loss": 0.9017, "step": 8509 }, { "epoch": 1.01, "learning_rate": 3.857214132965449e-06, "loss": 0.5545, "step": 8510 }, { "epoch": 1.01, "learning_rate": 3.856944188218593e-06, "loss": 0.5104, "step": 8511 }, { "epoch": 1.01, "learning_rate": 3.856674221041336e-06, "loss": 0.6641, "step": 8512 }, { "epoch": 1.01, "learning_rate": 3.856404231438141e-06, "loss": 0.5693, "step": 8513 }, { "epoch": 1.01, "learning_rate": 3.85613421941347e-06, "loss": 0.6243, "step": 8514 }, { "epoch": 1.01, "learning_rate": 3.855864184971787e-06, "loss": 0.7577, "step": 8515 }, { "epoch": 1.01, "learning_rate": 3.8555941281175545e-06, "loss": 0.451, "step": 8516 }, { "epoch": 1.01, "learning_rate": 3.8553240488552375e-06, "loss": 0.4186, "step": 8517 }, { "epoch": 1.01, "learning_rate": 3.8550539471893e-06, "loss": 0.5225, "step": 8518 }, { "epoch": 1.01, "learning_rate": 3.8547838231242074e-06, "loss": 0.5324, "step": 8519 }, { "epoch": 1.01, "learning_rate": 3.854513676664424e-06, "loss": 0.8171, "step": 8520 }, { "epoch": 1.01, "learning_rate": 3.854243507814416e-06, "loss": 0.8942, "step": 8521 }, { "epoch": 1.01, "learning_rate": 3.85397331657865e-06, "loss": 0.718, "step": 8522 }, { "epoch": 1.02, "learning_rate": 3.853703102961591e-06, "loss": 0.4765, "step": 8523 }, { "epoch": 1.02, "learning_rate": 3.853432866967705e-06, "loss": 0.6399, "step": 8524 }, { "epoch": 1.02, "learning_rate": 3.8531626086014625e-06, "loss": 0.5439, "step": 8525 }, { "epoch": 1.02, "learning_rate": 3.852892327867327e-06, "loss": 0.7711, "step": 8526 }, { "epoch": 1.02, "learning_rate": 3.852622024769768e-06, "loss": 0.8548, "step": 8527 }, { "epoch": 1.02, "learning_rate": 3.852351699313254e-06, "loss": 0.8644, "step": 8528 }, { "epoch": 1.02, "learning_rate": 3.852081351502252e-06, "loss": 0.4361, "step": 8529 }, { "epoch": 1.02, "learning_rate": 3.8518109813412315e-06, "loss": 0.6197, "step": 8530 }, { "epoch": 1.02, "learning_rate": 3.8515405888346635e-06, "loss": 0.6127, "step": 8531 }, { "epoch": 1.02, "learning_rate": 3.851270173987015e-06, "loss": 0.5805, "step": 8532 }, { "epoch": 1.02, "learning_rate": 3.850999736802758e-06, "loss": 0.7487, "step": 8533 }, { "epoch": 1.02, "learning_rate": 3.850729277286361e-06, "loss": 0.7845, "step": 8534 }, { "epoch": 1.02, "learning_rate": 3.850458795442296e-06, "loss": 0.576, "step": 8535 }, { "epoch": 1.02, "learning_rate": 3.850188291275036e-06, "loss": 0.4356, "step": 8536 }, { "epoch": 1.02, "learning_rate": 3.8499177647890474e-06, "loss": 0.6342, "step": 8537 }, { "epoch": 1.02, "learning_rate": 3.849647215988805e-06, "loss": 0.6219, "step": 8538 }, { "epoch": 1.02, "learning_rate": 3.849376644878783e-06, "loss": 0.6622, "step": 8539 }, { "epoch": 1.02, "learning_rate": 3.84910605146345e-06, "loss": 0.8339, "step": 8540 }, { "epoch": 1.02, "learning_rate": 3.8488354357472815e-06, "loss": 0.6378, "step": 8541 }, { "epoch": 1.02, "learning_rate": 3.84856479773475e-06, "loss": 0.4897, "step": 8542 }, { "epoch": 1.02, "learning_rate": 3.848294137430329e-06, "loss": 0.5669, "step": 8543 }, { "epoch": 1.02, "learning_rate": 3.848023454838493e-06, "loss": 0.5651, "step": 8544 }, { "epoch": 1.02, "learning_rate": 3.847752749963715e-06, "loss": 0.8363, "step": 8545 }, { "epoch": 1.02, "learning_rate": 3.8474820228104735e-06, "loss": 0.8584, "step": 8546 }, { "epoch": 1.02, "learning_rate": 3.847211273383239e-06, "loss": 0.8365, "step": 8547 }, { "epoch": 1.02, "learning_rate": 3.8469405016864895e-06, "loss": 0.3777, "step": 8548 }, { "epoch": 1.02, "learning_rate": 3.846669707724701e-06, "loss": 0.4988, "step": 8549 }, { "epoch": 1.02, "learning_rate": 3.8463988915023494e-06, "loss": 0.5956, "step": 8550 }, { "epoch": 1.02, "learning_rate": 3.846128053023911e-06, "loss": 0.621, "step": 8551 }, { "epoch": 1.02, "learning_rate": 3.845857192293862e-06, "loss": 0.7096, "step": 8552 }, { "epoch": 1.02, "learning_rate": 3.845586309316681e-06, "loss": 0.7781, "step": 8553 }, { "epoch": 1.02, "learning_rate": 3.845315404096846e-06, "loss": 0.3797, "step": 8554 }, { "epoch": 1.02, "learning_rate": 3.845044476638834e-06, "loss": 0.5147, "step": 8555 }, { "epoch": 1.02, "learning_rate": 3.844773526947125e-06, "loss": 0.5552, "step": 8556 }, { "epoch": 1.02, "learning_rate": 3.8445025550261965e-06, "loss": 0.5541, "step": 8557 }, { "epoch": 1.02, "learning_rate": 3.844231560880526e-06, "loss": 0.7375, "step": 8558 }, { "epoch": 1.02, "learning_rate": 3.843960544514597e-06, "loss": 0.8344, "step": 8559 }, { "epoch": 1.02, "learning_rate": 3.843689505932887e-06, "loss": 0.4516, "step": 8560 }, { "epoch": 1.02, "learning_rate": 3.843418445139877e-06, "loss": 0.5267, "step": 8561 }, { "epoch": 1.02, "learning_rate": 3.843147362140047e-06, "loss": 0.4578, "step": 8562 }, { "epoch": 1.02, "learning_rate": 3.842876256937878e-06, "loss": 0.6, "step": 8563 }, { "epoch": 1.02, "learning_rate": 3.842605129537852e-06, "loss": 0.7094, "step": 8564 }, { "epoch": 1.02, "learning_rate": 3.84233397994445e-06, "loss": 0.8138, "step": 8565 }, { "epoch": 1.02, "learning_rate": 3.842062808162156e-06, "loss": 0.4674, "step": 8566 }, { "epoch": 1.02, "learning_rate": 3.841791614195451e-06, "loss": 0.5048, "step": 8567 }, { "epoch": 1.02, "learning_rate": 3.841520398048817e-06, "loss": 0.5787, "step": 8568 }, { "epoch": 1.02, "learning_rate": 3.841249159726739e-06, "loss": 0.6723, "step": 8569 }, { "epoch": 1.02, "learning_rate": 3.8409778992336995e-06, "loss": 0.7382, "step": 8570 }, { "epoch": 1.02, "learning_rate": 3.840706616574184e-06, "loss": 0.8865, "step": 8571 }, { "epoch": 1.02, "learning_rate": 3.840435311752674e-06, "loss": 0.8187, "step": 8572 }, { "epoch": 1.02, "learning_rate": 3.840163984773656e-06, "loss": 0.4305, "step": 8573 }, { "epoch": 1.02, "learning_rate": 3.839892635641616e-06, "loss": 0.5554, "step": 8574 }, { "epoch": 1.02, "learning_rate": 3.839621264361037e-06, "loss": 0.5243, "step": 8575 }, { "epoch": 1.02, "learning_rate": 3.839349870936407e-06, "loss": 0.6671, "step": 8576 }, { "epoch": 1.02, "learning_rate": 3.8390784553722105e-06, "loss": 0.9263, "step": 8577 }, { "epoch": 1.02, "learning_rate": 3.838807017672936e-06, "loss": 0.9484, "step": 8578 }, { "epoch": 1.02, "learning_rate": 3.838535557843069e-06, "loss": 0.4114, "step": 8579 }, { "epoch": 1.02, "learning_rate": 3.838264075887096e-06, "loss": 0.5726, "step": 8580 }, { "epoch": 1.02, "learning_rate": 3.837992571809506e-06, "loss": 0.6418, "step": 8581 }, { "epoch": 1.02, "learning_rate": 3.837721045614786e-06, "loss": 0.7156, "step": 8582 }, { "epoch": 1.02, "learning_rate": 3.837449497307426e-06, "loss": 0.7533, "step": 8583 }, { "epoch": 1.02, "learning_rate": 3.837177926891912e-06, "loss": 0.7455, "step": 8584 }, { "epoch": 1.02, "learning_rate": 3.836906334372737e-06, "loss": 0.4445, "step": 8585 }, { "epoch": 1.02, "learning_rate": 3.836634719754385e-06, "loss": 0.5375, "step": 8586 }, { "epoch": 1.02, "learning_rate": 3.8363630830413516e-06, "loss": 0.4898, "step": 8587 }, { "epoch": 1.02, "learning_rate": 3.836091424238123e-06, "loss": 0.6005, "step": 8588 }, { "epoch": 1.02, "learning_rate": 3.835819743349191e-06, "loss": 0.7539, "step": 8589 }, { "epoch": 1.02, "learning_rate": 3.835548040379047e-06, "loss": 0.9317, "step": 8590 }, { "epoch": 1.02, "learning_rate": 3.835276315332182e-06, "loss": 0.6104, "step": 8591 }, { "epoch": 1.02, "learning_rate": 3.8350045682130866e-06, "loss": 0.4348, "step": 8592 }, { "epoch": 1.02, "learning_rate": 3.8347327990262555e-06, "loss": 0.4412, "step": 8593 }, { "epoch": 1.02, "learning_rate": 3.834461007776177e-06, "loss": 0.7478, "step": 8594 }, { "epoch": 1.02, "learning_rate": 3.834189194467347e-06, "loss": 0.671, "step": 8595 }, { "epoch": 1.02, "learning_rate": 3.8339173591042585e-06, "loss": 0.7719, "step": 8596 }, { "epoch": 1.02, "learning_rate": 3.833645501691403e-06, "loss": 0.8067, "step": 8597 }, { "epoch": 1.02, "learning_rate": 3.833373622233276e-06, "loss": 0.5259, "step": 8598 }, { "epoch": 1.02, "learning_rate": 3.833101720734371e-06, "loss": 0.5302, "step": 8599 }, { "epoch": 1.02, "learning_rate": 3.832829797199183e-06, "loss": 0.5128, "step": 8600 }, { "epoch": 1.02, "learning_rate": 3.832557851632205e-06, "loss": 0.5581, "step": 8601 }, { "epoch": 1.02, "learning_rate": 3.832285884037936e-06, "loss": 0.8264, "step": 8602 }, { "epoch": 1.02, "learning_rate": 3.832013894420868e-06, "loss": 0.8571, "step": 8603 }, { "epoch": 1.02, "learning_rate": 3.8317418827855004e-06, "loss": 0.4517, "step": 8604 }, { "epoch": 1.02, "learning_rate": 3.831469849136326e-06, "loss": 0.428, "step": 8605 }, { "epoch": 1.02, "learning_rate": 3.831197793477844e-06, "loss": 0.5215, "step": 8606 }, { "epoch": 1.03, "learning_rate": 3.830925715814551e-06, "loss": 0.7368, "step": 8607 }, { "epoch": 1.03, "learning_rate": 3.830653616150944e-06, "loss": 0.8428, "step": 8608 }, { "epoch": 1.03, "learning_rate": 3.830381494491521e-06, "loss": 0.7504, "step": 8609 }, { "epoch": 1.03, "learning_rate": 3.83010935084078e-06, "loss": 0.4246, "step": 8610 }, { "epoch": 1.03, "learning_rate": 3.82983718520322e-06, "loss": 0.5637, "step": 8611 }, { "epoch": 1.03, "learning_rate": 3.82956499758334e-06, "loss": 0.5671, "step": 8612 }, { "epoch": 1.03, "learning_rate": 3.829292787985639e-06, "loss": 0.6207, "step": 8613 }, { "epoch": 1.03, "learning_rate": 3.829020556414616e-06, "loss": 0.7634, "step": 8614 }, { "epoch": 1.03, "learning_rate": 3.828748302874773e-06, "loss": 0.913, "step": 8615 }, { "epoch": 1.03, "learning_rate": 3.828476027370608e-06, "loss": 0.7633, "step": 8616 }, { "epoch": 1.03, "learning_rate": 3.828203729906623e-06, "loss": 0.3552, "step": 8617 }, { "epoch": 1.03, "learning_rate": 3.827931410487319e-06, "loss": 0.6248, "step": 8618 }, { "epoch": 1.03, "learning_rate": 3.8276590691171975e-06, "loss": 0.6042, "step": 8619 }, { "epoch": 1.03, "learning_rate": 3.82738670580076e-06, "loss": 0.6921, "step": 8620 }, { "epoch": 1.03, "learning_rate": 3.827114320542509e-06, "loss": 0.8552, "step": 8621 }, { "epoch": 1.03, "learning_rate": 3.826841913346947e-06, "loss": 0.7715, "step": 8622 }, { "epoch": 1.03, "learning_rate": 3.826569484218577e-06, "loss": 0.4747, "step": 8623 }, { "epoch": 1.03, "learning_rate": 3.826297033161902e-06, "loss": 0.6085, "step": 8624 }, { "epoch": 1.03, "learning_rate": 3.826024560181425e-06, "loss": 0.5749, "step": 8625 }, { "epoch": 1.03, "learning_rate": 3.825752065281652e-06, "loss": 0.7343, "step": 8626 }, { "epoch": 1.03, "learning_rate": 3.8254795484670844e-06, "loss": 0.801, "step": 8627 }, { "epoch": 1.03, "learning_rate": 3.825207009742231e-06, "loss": 0.7521, "step": 8628 }, { "epoch": 1.03, "learning_rate": 3.824934449111593e-06, "loss": 0.406, "step": 8629 }, { "epoch": 1.03, "learning_rate": 3.824661866579677e-06, "loss": 0.4511, "step": 8630 }, { "epoch": 1.03, "learning_rate": 3.824389262150991e-06, "loss": 0.5372, "step": 8631 }, { "epoch": 1.03, "learning_rate": 3.824116635830037e-06, "loss": 0.61, "step": 8632 }, { "epoch": 1.03, "learning_rate": 3.823843987621325e-06, "loss": 0.7049, "step": 8633 }, { "epoch": 1.03, "learning_rate": 3.823571317529361e-06, "loss": 0.8854, "step": 8634 }, { "epoch": 1.03, "learning_rate": 3.823298625558652e-06, "loss": 0.2924, "step": 8635 }, { "epoch": 1.03, "learning_rate": 3.823025911713706e-06, "loss": 0.4999, "step": 8636 }, { "epoch": 1.03, "learning_rate": 3.82275317599903e-06, "loss": 0.6428, "step": 8637 }, { "epoch": 1.03, "learning_rate": 3.822480418419133e-06, "loss": 0.5797, "step": 8638 }, { "epoch": 1.03, "learning_rate": 3.822207638978524e-06, "loss": 0.6655, "step": 8639 }, { "epoch": 1.03, "learning_rate": 3.821934837681712e-06, "loss": 0.9128, "step": 8640 }, { "epoch": 1.03, "learning_rate": 3.821662014533206e-06, "loss": 0.5661, "step": 8641 }, { "epoch": 1.03, "learning_rate": 3.8213891695375165e-06, "loss": 0.3702, "step": 8642 }, { "epoch": 1.03, "learning_rate": 3.821116302699153e-06, "loss": 0.5646, "step": 8643 }, { "epoch": 1.03, "learning_rate": 3.820843414022625e-06, "loss": 0.6319, "step": 8644 }, { "epoch": 1.03, "learning_rate": 3.820570503512445e-06, "loss": 0.7384, "step": 8645 }, { "epoch": 1.03, "learning_rate": 3.820297571173125e-06, "loss": 0.7733, "step": 8646 }, { "epoch": 1.03, "learning_rate": 3.820024617009175e-06, "loss": 0.8761, "step": 8647 }, { "epoch": 1.03, "learning_rate": 3.8197516410251065e-06, "loss": 0.4721, "step": 8648 }, { "epoch": 1.03, "learning_rate": 3.819478643225433e-06, "loss": 0.4247, "step": 8649 }, { "epoch": 1.03, "learning_rate": 3.819205623614667e-06, "loss": 0.5773, "step": 8650 }, { "epoch": 1.03, "learning_rate": 3.818932582197321e-06, "loss": 0.6084, "step": 8651 }, { "epoch": 1.03, "learning_rate": 3.818659518977908e-06, "loss": 0.7149, "step": 8652 }, { "epoch": 1.03, "learning_rate": 3.8183864339609444e-06, "loss": 0.8342, "step": 8653 }, { "epoch": 1.03, "learning_rate": 3.818113327150941e-06, "loss": 0.3521, "step": 8654 }, { "epoch": 1.03, "learning_rate": 3.817840198552415e-06, "loss": 0.6097, "step": 8655 }, { "epoch": 1.03, "learning_rate": 3.817567048169879e-06, "loss": 0.5707, "step": 8656 }, { "epoch": 1.03, "learning_rate": 3.8172938760078495e-06, "loss": 0.6256, "step": 8657 }, { "epoch": 1.03, "learning_rate": 3.817020682070841e-06, "loss": 0.6857, "step": 8658 }, { "epoch": 1.03, "learning_rate": 3.816747466363371e-06, "loss": 0.8906, "step": 8659 }, { "epoch": 1.03, "learning_rate": 3.8164742288899546e-06, "loss": 0.5082, "step": 8660 }, { "epoch": 1.03, "learning_rate": 3.8162009696551076e-06, "loss": 0.4292, "step": 8661 }, { "epoch": 1.03, "learning_rate": 3.8159276886633494e-06, "loss": 0.5064, "step": 8662 }, { "epoch": 1.03, "learning_rate": 3.815654385919195e-06, "loss": 0.6023, "step": 8663 }, { "epoch": 1.03, "learning_rate": 3.815381061427165e-06, "loss": 0.7458, "step": 8664 }, { "epoch": 1.03, "learning_rate": 3.8151077151917744e-06, "loss": 0.8065, "step": 8665 }, { "epoch": 1.03, "learning_rate": 3.814834347217543e-06, "loss": 0.5333, "step": 8666 }, { "epoch": 1.03, "learning_rate": 3.81456095750899e-06, "loss": 0.4585, "step": 8667 }, { "epoch": 1.03, "learning_rate": 3.814287546070634e-06, "loss": 0.5609, "step": 8668 }, { "epoch": 1.03, "learning_rate": 3.814014112906994e-06, "loss": 0.6635, "step": 8669 }, { "epoch": 1.03, "learning_rate": 3.813740658022591e-06, "loss": 0.7623, "step": 8670 }, { "epoch": 1.03, "learning_rate": 3.8134671814219447e-06, "loss": 0.7491, "step": 8671 }, { "epoch": 1.03, "learning_rate": 3.813193683109576e-06, "loss": 0.8905, "step": 8672 }, { "epoch": 1.03, "learning_rate": 3.812920163090005e-06, "loss": 0.4964, "step": 8673 }, { "epoch": 1.03, "learning_rate": 3.812646621367754e-06, "loss": 0.518, "step": 8674 }, { "epoch": 1.03, "learning_rate": 3.8123730579473438e-06, "loss": 0.637, "step": 8675 }, { "epoch": 1.03, "learning_rate": 3.812099472833296e-06, "loss": 0.5796, "step": 8676 }, { "epoch": 1.03, "learning_rate": 3.811825866030135e-06, "loss": 0.7704, "step": 8677 }, { "epoch": 1.03, "learning_rate": 3.811552237542382e-06, "loss": 0.9446, "step": 8678 }, { "epoch": 1.03, "learning_rate": 3.8112785873745607e-06, "loss": 0.4474, "step": 8679 }, { "epoch": 1.03, "learning_rate": 3.8110049155311936e-06, "loss": 0.4767, "step": 8680 }, { "epoch": 1.03, "learning_rate": 3.8107312220168057e-06, "loss": 0.5186, "step": 8681 }, { "epoch": 1.03, "learning_rate": 3.810457506835921e-06, "loss": 0.6477, "step": 8682 }, { "epoch": 1.03, "learning_rate": 3.810183769993064e-06, "loss": 0.7169, "step": 8683 }, { "epoch": 1.03, "learning_rate": 3.8099100114927586e-06, "loss": 0.7833, "step": 8684 }, { "epoch": 1.03, "learning_rate": 3.809636231339531e-06, "loss": 0.6055, "step": 8685 }, { "epoch": 1.03, "learning_rate": 3.809362429537907e-06, "loss": 0.4244, "step": 8686 }, { "epoch": 1.03, "learning_rate": 3.8090886060924113e-06, "loss": 0.4795, "step": 8687 }, { "epoch": 1.03, "learning_rate": 3.808814761007572e-06, "loss": 0.6874, "step": 8688 }, { "epoch": 1.03, "learning_rate": 3.808540894287915e-06, "loss": 0.6805, "step": 8689 }, { "epoch": 1.03, "learning_rate": 3.8082670059379667e-06, "loss": 0.6947, "step": 8690 }, { "epoch": 1.04, "learning_rate": 3.8079930959622547e-06, "loss": 0.5908, "step": 8691 }, { "epoch": 1.04, "learning_rate": 3.8077191643653076e-06, "loss": 0.478, "step": 8692 }, { "epoch": 1.04, "learning_rate": 3.807445211151653e-06, "loss": 0.5454, "step": 8693 }, { "epoch": 1.04, "learning_rate": 3.8071712363258197e-06, "loss": 0.6081, "step": 8694 }, { "epoch": 1.04, "learning_rate": 3.806897239892335e-06, "loss": 0.608, "step": 8695 }, { "epoch": 1.04, "learning_rate": 3.8066232218557302e-06, "loss": 0.7086, "step": 8696 }, { "epoch": 1.04, "learning_rate": 3.8063491822205333e-06, "loss": 0.8622, "step": 8697 }, { "epoch": 1.04, "learning_rate": 3.8060751209912757e-06, "loss": 0.3945, "step": 8698 }, { "epoch": 1.04, "learning_rate": 3.805801038172486e-06, "loss": 0.5512, "step": 8699 }, { "epoch": 1.04, "learning_rate": 3.8055269337686964e-06, "loss": 0.4933, "step": 8700 }, { "epoch": 1.04, "learning_rate": 3.8052528077844353e-06, "loss": 0.6765, "step": 8701 }, { "epoch": 1.04, "learning_rate": 3.8049786602242377e-06, "loss": 0.7798, "step": 8702 }, { "epoch": 1.04, "learning_rate": 3.8047044910926324e-06, "loss": 0.8878, "step": 8703 }, { "epoch": 1.04, "learning_rate": 3.804430300394153e-06, "loss": 0.4586, "step": 8704 }, { "epoch": 1.04, "learning_rate": 3.8041560881333305e-06, "loss": 0.6097, "step": 8705 }, { "epoch": 1.04, "learning_rate": 3.8038818543146983e-06, "loss": 0.4609, "step": 8706 }, { "epoch": 1.04, "learning_rate": 3.803607598942791e-06, "loss": 0.7216, "step": 8707 }, { "epoch": 1.04, "learning_rate": 3.803333322022139e-06, "loss": 0.7641, "step": 8708 }, { "epoch": 1.04, "learning_rate": 3.803059023557279e-06, "loss": 0.9106, "step": 8709 }, { "epoch": 1.04, "learning_rate": 3.8027847035527436e-06, "loss": 0.6273, "step": 8710 }, { "epoch": 1.04, "learning_rate": 3.802510362013068e-06, "loss": 0.5453, "step": 8711 }, { "epoch": 1.04, "learning_rate": 3.8022359989427863e-06, "loss": 0.4799, "step": 8712 }, { "epoch": 1.04, "learning_rate": 3.8019616143464344e-06, "loss": 0.5914, "step": 8713 }, { "epoch": 1.04, "learning_rate": 3.8016872082285483e-06, "loss": 0.7153, "step": 8714 }, { "epoch": 1.04, "learning_rate": 3.801412780593663e-06, "loss": 0.8495, "step": 8715 }, { "epoch": 1.04, "learning_rate": 3.8011383314463154e-06, "loss": 0.5395, "step": 8716 }, { "epoch": 1.04, "learning_rate": 3.8008638607910425e-06, "loss": 0.4459, "step": 8717 }, { "epoch": 1.04, "learning_rate": 3.8005893686323796e-06, "loss": 0.415, "step": 8718 }, { "epoch": 1.04, "learning_rate": 3.8003148549748665e-06, "loss": 0.5663, "step": 8719 }, { "epoch": 1.04, "learning_rate": 3.8000403198230385e-06, "loss": 0.7065, "step": 8720 }, { "epoch": 1.04, "learning_rate": 3.7997657631814366e-06, "loss": 0.7787, "step": 8721 }, { "epoch": 1.04, "learning_rate": 3.7994911850545967e-06, "loss": 0.9027, "step": 8722 }, { "epoch": 1.04, "learning_rate": 3.7992165854470575e-06, "loss": 0.4588, "step": 8723 }, { "epoch": 1.04, "learning_rate": 3.7989419643633605e-06, "loss": 0.4421, "step": 8724 }, { "epoch": 1.04, "learning_rate": 3.7986673218080433e-06, "loss": 0.5109, "step": 8725 }, { "epoch": 1.04, "learning_rate": 3.7983926577856466e-06, "loss": 0.6346, "step": 8726 }, { "epoch": 1.04, "learning_rate": 3.79811797230071e-06, "loss": 0.7889, "step": 8727 }, { "epoch": 1.04, "learning_rate": 3.7978432653577744e-06, "loss": 0.8787, "step": 8728 }, { "epoch": 1.04, "learning_rate": 3.7975685369613808e-06, "loss": 0.4622, "step": 8729 }, { "epoch": 1.04, "learning_rate": 3.7972937871160707e-06, "loss": 0.6861, "step": 8730 }, { "epoch": 1.04, "learning_rate": 3.797019015826385e-06, "loss": 0.4872, "step": 8731 }, { "epoch": 1.04, "learning_rate": 3.7967442230968667e-06, "loss": 0.6453, "step": 8732 }, { "epoch": 1.04, "learning_rate": 3.7964694089320564e-06, "loss": 0.6705, "step": 8733 }, { "epoch": 1.04, "learning_rate": 3.796194573336499e-06, "loss": 0.9178, "step": 8734 }, { "epoch": 1.04, "learning_rate": 3.7959197163147366e-06, "loss": 0.561, "step": 8735 }, { "epoch": 1.04, "learning_rate": 3.7956448378713123e-06, "loss": 0.5676, "step": 8736 }, { "epoch": 1.04, "learning_rate": 3.79536993801077e-06, "loss": 0.5231, "step": 8737 }, { "epoch": 1.04, "learning_rate": 3.795095016737654e-06, "loss": 0.6897, "step": 8738 }, { "epoch": 1.04, "learning_rate": 3.794820074056509e-06, "loss": 0.7327, "step": 8739 }, { "epoch": 1.04, "learning_rate": 3.7945451099718787e-06, "loss": 0.8766, "step": 8740 }, { "epoch": 1.04, "learning_rate": 3.79427012448831e-06, "loss": 0.572, "step": 8741 }, { "epoch": 1.04, "learning_rate": 3.7939951176103474e-06, "loss": 0.4149, "step": 8742 }, { "epoch": 1.04, "learning_rate": 3.793720089342537e-06, "loss": 0.5628, "step": 8743 }, { "epoch": 1.04, "learning_rate": 3.793445039689424e-06, "loss": 0.6578, "step": 8744 }, { "epoch": 1.04, "learning_rate": 3.7931699686555568e-06, "loss": 0.6816, "step": 8745 }, { "epoch": 1.04, "learning_rate": 3.792894876245482e-06, "loss": 0.8363, "step": 8746 }, { "epoch": 1.04, "learning_rate": 3.7926197624637457e-06, "loss": 0.8717, "step": 8747 }, { "epoch": 1.04, "learning_rate": 3.7923446273148962e-06, "loss": 0.4508, "step": 8748 }, { "epoch": 1.04, "learning_rate": 3.7920694708034816e-06, "loss": 0.6642, "step": 8749 }, { "epoch": 1.04, "learning_rate": 3.7917942929340507e-06, "loss": 0.5497, "step": 8750 }, { "epoch": 1.04, "learning_rate": 3.7915190937111512e-06, "loss": 0.7006, "step": 8751 }, { "epoch": 1.04, "learning_rate": 3.791243873139333e-06, "loss": 0.8035, "step": 8752 }, { "epoch": 1.04, "learning_rate": 3.7909686312231454e-06, "loss": 1.0386, "step": 8753 }, { "epoch": 1.04, "learning_rate": 3.790693367967138e-06, "loss": 0.4731, "step": 8754 }, { "epoch": 1.04, "learning_rate": 3.790418083375861e-06, "loss": 0.5415, "step": 8755 }, { "epoch": 1.04, "learning_rate": 3.790142777453864e-06, "loss": 0.568, "step": 8756 }, { "epoch": 1.04, "learning_rate": 3.789867450205699e-06, "loss": 0.6437, "step": 8757 }, { "epoch": 1.04, "learning_rate": 3.7895921016359173e-06, "loss": 0.7178, "step": 8758 }, { "epoch": 1.04, "learning_rate": 3.789316731749069e-06, "loss": 0.934, "step": 8759 }, { "epoch": 1.04, "learning_rate": 3.7890413405497084e-06, "loss": 0.3953, "step": 8760 }, { "epoch": 1.04, "learning_rate": 3.788765928042385e-06, "loss": 0.5751, "step": 8761 }, { "epoch": 1.04, "learning_rate": 3.7884904942316536e-06, "loss": 0.5632, "step": 8762 }, { "epoch": 1.04, "learning_rate": 3.788215039122065e-06, "loss": 0.6836, "step": 8763 }, { "epoch": 1.04, "learning_rate": 3.7879395627181754e-06, "loss": 0.72, "step": 8764 }, { "epoch": 1.04, "learning_rate": 3.7876640650245357e-06, "loss": 0.8262, "step": 8765 }, { "epoch": 1.04, "learning_rate": 3.7873885460457005e-06, "loss": 0.5034, "step": 8766 }, { "epoch": 1.04, "learning_rate": 3.7871130057862255e-06, "loss": 0.5, "step": 8767 }, { "epoch": 1.04, "learning_rate": 3.7868374442506644e-06, "loss": 0.6944, "step": 8768 }, { "epoch": 1.04, "learning_rate": 3.7865618614435724e-06, "loss": 0.5974, "step": 8769 }, { "epoch": 1.04, "learning_rate": 3.7862862573695042e-06, "loss": 0.6205, "step": 8770 }, { "epoch": 1.04, "learning_rate": 3.786010632033017e-06, "loss": 0.6818, "step": 8771 }, { "epoch": 1.04, "learning_rate": 3.785734985438666e-06, "loss": 0.8843, "step": 8772 }, { "epoch": 1.04, "learning_rate": 3.785459317591007e-06, "loss": 0.4359, "step": 8773 }, { "epoch": 1.04, "learning_rate": 3.7851836284945987e-06, "loss": 0.4967, "step": 8774 }, { "epoch": 1.05, "learning_rate": 3.784907918153997e-06, "loss": 0.5286, "step": 8775 }, { "epoch": 1.05, "learning_rate": 3.784632186573759e-06, "loss": 0.8107, "step": 8776 }, { "epoch": 1.05, "learning_rate": 3.784356433758444e-06, "loss": 0.7331, "step": 8777 }, { "epoch": 1.05, "learning_rate": 3.784080659712609e-06, "loss": 0.8453, "step": 8778 }, { "epoch": 1.05, "learning_rate": 3.783804864440813e-06, "loss": 0.368, "step": 8779 }, { "epoch": 1.05, "learning_rate": 3.7835290479476138e-06, "loss": 0.5076, "step": 8780 }, { "epoch": 1.05, "learning_rate": 3.783253210237573e-06, "loss": 0.5678, "step": 8781 }, { "epoch": 1.05, "learning_rate": 3.782977351315249e-06, "loss": 0.5332, "step": 8782 }, { "epoch": 1.05, "learning_rate": 3.782701471185201e-06, "loss": 0.5946, "step": 8783 }, { "epoch": 1.05, "learning_rate": 3.78242556985199e-06, "loss": 0.7982, "step": 8784 }, { "epoch": 1.05, "learning_rate": 3.782149647320178e-06, "loss": 0.4778, "step": 8785 }, { "epoch": 1.05, "learning_rate": 3.7818737035943242e-06, "loss": 0.4787, "step": 8786 }, { "epoch": 1.05, "learning_rate": 3.7815977386789892e-06, "loss": 0.5309, "step": 8787 }, { "epoch": 1.05, "learning_rate": 3.7813217525787376e-06, "loss": 0.6066, "step": 8788 }, { "epoch": 1.05, "learning_rate": 3.78104574529813e-06, "loss": 0.7376, "step": 8789 }, { "epoch": 1.05, "learning_rate": 3.7807697168417277e-06, "loss": 0.878, "step": 8790 }, { "epoch": 1.05, "learning_rate": 3.7804936672140944e-06, "loss": 0.687, "step": 8791 }, { "epoch": 1.05, "learning_rate": 3.780217596419794e-06, "loss": 0.4568, "step": 8792 }, { "epoch": 1.05, "learning_rate": 3.779941504463389e-06, "loss": 0.4251, "step": 8793 }, { "epoch": 1.05, "learning_rate": 3.779665391349444e-06, "loss": 0.7025, "step": 8794 }, { "epoch": 1.05, "learning_rate": 3.779389257082522e-06, "loss": 0.7397, "step": 8795 }, { "epoch": 1.05, "learning_rate": 3.7791131016671896e-06, "loss": 0.6713, "step": 8796 }, { "epoch": 1.05, "learning_rate": 3.778836925108009e-06, "loss": 0.8552, "step": 8797 }, { "epoch": 1.05, "learning_rate": 3.778560727409547e-06, "loss": 0.4831, "step": 8798 }, { "epoch": 1.05, "learning_rate": 3.778284508576369e-06, "loss": 0.5191, "step": 8799 }, { "epoch": 1.05, "learning_rate": 3.7780082686130414e-06, "loss": 0.6522, "step": 8800 }, { "epoch": 1.05, "learning_rate": 3.777732007524129e-06, "loss": 0.6501, "step": 8801 }, { "epoch": 1.05, "learning_rate": 3.7774557253141997e-06, "loss": 0.7281, "step": 8802 }, { "epoch": 1.05, "learning_rate": 3.77717942198782e-06, "loss": 0.7619, "step": 8803 }, { "epoch": 1.05, "learning_rate": 3.776903097549557e-06, "loss": 0.4895, "step": 8804 }, { "epoch": 1.05, "learning_rate": 3.776626752003979e-06, "loss": 0.5221, "step": 8805 }, { "epoch": 1.05, "learning_rate": 3.7763503853556533e-06, "loss": 0.525, "step": 8806 }, { "epoch": 1.05, "learning_rate": 3.77607399760915e-06, "loss": 0.5669, "step": 8807 }, { "epoch": 1.05, "learning_rate": 3.7757975887690344e-06, "loss": 0.7697, "step": 8808 }, { "epoch": 1.05, "learning_rate": 3.775521158839879e-06, "loss": 0.8973, "step": 8809 }, { "epoch": 1.05, "learning_rate": 3.775244707826251e-06, "loss": 0.3534, "step": 8810 }, { "epoch": 1.05, "learning_rate": 3.7749682357327215e-06, "loss": 0.4956, "step": 8811 }, { "epoch": 1.05, "learning_rate": 3.7746917425638596e-06, "loss": 0.4775, "step": 8812 }, { "epoch": 1.05, "learning_rate": 3.774415228324237e-06, "loss": 0.6138, "step": 8813 }, { "epoch": 1.05, "learning_rate": 3.774138693018423e-06, "loss": 0.7476, "step": 8814 }, { "epoch": 1.05, "learning_rate": 3.7738621366509904e-06, "loss": 0.753, "step": 8815 }, { "epoch": 1.05, "learning_rate": 3.7735855592265093e-06, "loss": 0.3751, "step": 8816 }, { "epoch": 1.05, "learning_rate": 3.773308960749552e-06, "loss": 0.5138, "step": 8817 }, { "epoch": 1.05, "learning_rate": 3.7730323412246903e-06, "loss": 0.5168, "step": 8818 }, { "epoch": 1.05, "learning_rate": 3.7727557006564975e-06, "loss": 0.4709, "step": 8819 }, { "epoch": 1.05, "learning_rate": 3.772479039049547e-06, "loss": 0.5874, "step": 8820 }, { "epoch": 1.05, "learning_rate": 3.77220235640841e-06, "loss": 0.8553, "step": 8821 }, { "epoch": 1.05, "learning_rate": 3.7719256527376625e-06, "loss": 0.8189, "step": 8822 }, { "epoch": 1.05, "learning_rate": 3.7716489280418757e-06, "loss": 0.4657, "step": 8823 }, { "epoch": 1.05, "learning_rate": 3.7713721823256267e-06, "loss": 0.4652, "step": 8824 }, { "epoch": 1.05, "learning_rate": 3.7710954155934882e-06, "loss": 0.4884, "step": 8825 }, { "epoch": 1.05, "learning_rate": 3.770818627850036e-06, "loss": 0.575, "step": 8826 }, { "epoch": 1.05, "learning_rate": 3.7705418190998448e-06, "loss": 0.6519, "step": 8827 }, { "epoch": 1.05, "learning_rate": 3.770264989347492e-06, "loss": 0.8521, "step": 8828 }, { "epoch": 1.05, "learning_rate": 3.769988138597551e-06, "loss": 0.5398, "step": 8829 }, { "epoch": 1.05, "learning_rate": 3.7697112668546004e-06, "loss": 0.4957, "step": 8830 }, { "epoch": 1.05, "learning_rate": 3.769434374123215e-06, "loss": 0.616, "step": 8831 }, { "epoch": 1.05, "learning_rate": 3.7691574604079744e-06, "loss": 0.6315, "step": 8832 }, { "epoch": 1.05, "learning_rate": 3.7688805257134537e-06, "loss": 0.7333, "step": 8833 }, { "epoch": 1.05, "learning_rate": 3.7686035700442315e-06, "loss": 0.7971, "step": 8834 }, { "epoch": 1.05, "learning_rate": 3.7683265934048853e-06, "loss": 0.4996, "step": 8835 }, { "epoch": 1.05, "learning_rate": 3.7680495957999944e-06, "loss": 0.4808, "step": 8836 }, { "epoch": 1.05, "learning_rate": 3.767772577234138e-06, "loss": 0.5634, "step": 8837 }, { "epoch": 1.05, "learning_rate": 3.7674955377118932e-06, "loss": 0.6755, "step": 8838 }, { "epoch": 1.05, "learning_rate": 3.7672184772378423e-06, "loss": 0.6936, "step": 8839 }, { "epoch": 1.05, "learning_rate": 3.7669413958165624e-06, "loss": 0.8257, "step": 8840 }, { "epoch": 1.05, "learning_rate": 3.7666642934526356e-06, "loss": 0.4621, "step": 8841 }, { "epoch": 1.05, "learning_rate": 3.7663871701506417e-06, "loss": 0.3293, "step": 8842 }, { "epoch": 1.05, "learning_rate": 3.7661100259151616e-06, "loss": 0.5074, "step": 8843 }, { "epoch": 1.05, "learning_rate": 3.7658328607507755e-06, "loss": 0.5818, "step": 8844 }, { "epoch": 1.05, "learning_rate": 3.7655556746620676e-06, "loss": 0.689, "step": 8845 }, { "epoch": 1.05, "learning_rate": 3.7652784676536173e-06, "loss": 0.8238, "step": 8846 }, { "epoch": 1.05, "learning_rate": 3.7650012397300085e-06, "loss": 0.6559, "step": 8847 }, { "epoch": 1.05, "learning_rate": 3.764723990895822e-06, "loss": 0.4376, "step": 8848 }, { "epoch": 1.05, "learning_rate": 3.764446721155642e-06, "loss": 0.57, "step": 8849 }, { "epoch": 1.05, "learning_rate": 3.764169430514053e-06, "loss": 0.607, "step": 8850 }, { "epoch": 1.05, "learning_rate": 3.7638921189756355e-06, "loss": 0.6653, "step": 8851 }, { "epoch": 1.05, "learning_rate": 3.7636147865449762e-06, "loss": 0.8225, "step": 8852 }, { "epoch": 1.05, "learning_rate": 3.763337433226658e-06, "loss": 1.0198, "step": 8853 }, { "epoch": 1.05, "learning_rate": 3.7630600590252662e-06, "loss": 0.386, "step": 8854 }, { "epoch": 1.05, "learning_rate": 3.762782663945386e-06, "loss": 0.5201, "step": 8855 }, { "epoch": 1.05, "learning_rate": 3.762505247991601e-06, "loss": 0.6368, "step": 8856 }, { "epoch": 1.05, "learning_rate": 3.7622278111685e-06, "loss": 0.6305, "step": 8857 }, { "epoch": 1.05, "learning_rate": 3.7619503534806666e-06, "loss": 0.7411, "step": 8858 }, { "epoch": 1.06, "learning_rate": 3.761672874932688e-06, "loss": 0.8337, "step": 8859 }, { "epoch": 1.06, "learning_rate": 3.761395375529152e-06, "loss": 0.5128, "step": 8860 }, { "epoch": 1.06, "learning_rate": 3.761117855274643e-06, "loss": 0.4656, "step": 8861 }, { "epoch": 1.06, "learning_rate": 3.7608403141737506e-06, "loss": 0.5168, "step": 8862 }, { "epoch": 1.06, "learning_rate": 3.7605627522310617e-06, "loss": 0.6211, "step": 8863 }, { "epoch": 1.06, "learning_rate": 3.7602851694511654e-06, "loss": 0.6104, "step": 8864 }, { "epoch": 1.06, "learning_rate": 3.7600075658386497e-06, "loss": 0.7979, "step": 8865 }, { "epoch": 1.06, "learning_rate": 3.7597299413981024e-06, "loss": 0.654, "step": 8866 }, { "epoch": 1.06, "learning_rate": 3.759452296134114e-06, "loss": 0.4867, "step": 8867 }, { "epoch": 1.06, "learning_rate": 3.759174630051273e-06, "loss": 0.561, "step": 8868 }, { "epoch": 1.06, "learning_rate": 3.75889694315417e-06, "loss": 0.7177, "step": 8869 }, { "epoch": 1.06, "learning_rate": 3.758619235447395e-06, "loss": 0.7015, "step": 8870 }, { "epoch": 1.06, "learning_rate": 3.758341506935539e-06, "loss": 0.7687, "step": 8871 }, { "epoch": 1.06, "learning_rate": 3.758063757623191e-06, "loss": 0.7979, "step": 8872 }, { "epoch": 1.06, "learning_rate": 3.7577859875149443e-06, "loss": 0.4117, "step": 8873 }, { "epoch": 1.06, "learning_rate": 3.757508196615389e-06, "loss": 0.5197, "step": 8874 }, { "epoch": 1.06, "learning_rate": 3.7572303849291192e-06, "loss": 0.6309, "step": 8875 }, { "epoch": 1.06, "learning_rate": 3.756952552460724e-06, "loss": 0.6131, "step": 8876 }, { "epoch": 1.06, "learning_rate": 3.7566746992147984e-06, "loss": 0.9041, "step": 8877 }, { "epoch": 1.06, "learning_rate": 3.7563968251959348e-06, "loss": 0.7682, "step": 8878 }, { "epoch": 1.06, "learning_rate": 3.7561189304087258e-06, "loss": 0.4395, "step": 8879 }, { "epoch": 1.06, "learning_rate": 3.7558410148577657e-06, "loss": 0.5128, "step": 8880 }, { "epoch": 1.06, "learning_rate": 3.755563078547648e-06, "loss": 0.5344, "step": 8881 }, { "epoch": 1.06, "learning_rate": 3.7552851214829676e-06, "loss": 0.6133, "step": 8882 }, { "epoch": 1.06, "learning_rate": 3.755007143668319e-06, "loss": 0.7393, "step": 8883 }, { "epoch": 1.06, "learning_rate": 3.754729145108296e-06, "loss": 0.7007, "step": 8884 }, { "epoch": 1.06, "learning_rate": 3.7544511258074963e-06, "loss": 0.3795, "step": 8885 }, { "epoch": 1.06, "learning_rate": 3.7541730857705133e-06, "loss": 0.5222, "step": 8886 }, { "epoch": 1.06, "learning_rate": 3.753895025001944e-06, "loss": 0.5768, "step": 8887 }, { "epoch": 1.06, "learning_rate": 3.753616943506385e-06, "loss": 0.6758, "step": 8888 }, { "epoch": 1.06, "learning_rate": 3.7533388412884332e-06, "loss": 0.6049, "step": 8889 }, { "epoch": 1.06, "learning_rate": 3.7530607183526847e-06, "loss": 0.7562, "step": 8890 }, { "epoch": 1.06, "learning_rate": 3.7527825747037366e-06, "loss": 0.6602, "step": 8891 }, { "epoch": 1.06, "learning_rate": 3.7525044103461885e-06, "loss": 0.4536, "step": 8892 }, { "epoch": 1.06, "learning_rate": 3.752226225284637e-06, "loss": 0.5427, "step": 8893 }, { "epoch": 1.06, "learning_rate": 3.7519480195236813e-06, "loss": 0.6506, "step": 8894 }, { "epoch": 1.06, "learning_rate": 3.751669793067919e-06, "loss": 0.7175, "step": 8895 }, { "epoch": 1.06, "learning_rate": 3.751391545921951e-06, "loss": 0.7641, "step": 8896 }, { "epoch": 1.06, "learning_rate": 3.751113278090375e-06, "loss": 0.7664, "step": 8897 }, { "epoch": 1.06, "learning_rate": 3.750834989577792e-06, "loss": 0.4649, "step": 8898 }, { "epoch": 1.06, "learning_rate": 3.750556680388802e-06, "loss": 0.4848, "step": 8899 }, { "epoch": 1.06, "learning_rate": 3.750278350528004e-06, "loss": 0.6421, "step": 8900 }, { "epoch": 1.06, "learning_rate": 3.7500000000000005e-06, "loss": 0.683, "step": 8901 }, { "epoch": 1.06, "learning_rate": 3.749721628809392e-06, "loss": 0.8178, "step": 8902 }, { "epoch": 1.06, "learning_rate": 3.7494432369607808e-06, "loss": 0.9294, "step": 8903 }, { "epoch": 1.06, "learning_rate": 3.749164824458767e-06, "loss": 0.4225, "step": 8904 }, { "epoch": 1.06, "learning_rate": 3.748886391307954e-06, "loss": 0.6084, "step": 8905 }, { "epoch": 1.06, "learning_rate": 3.748607937512945e-06, "loss": 0.6345, "step": 8906 }, { "epoch": 1.06, "learning_rate": 3.7483294630783417e-06, "loss": 0.5557, "step": 8907 }, { "epoch": 1.06, "learning_rate": 3.7480509680087475e-06, "loss": 0.8682, "step": 8908 }, { "epoch": 1.06, "learning_rate": 3.747772452308766e-06, "loss": 0.832, "step": 8909 }, { "epoch": 1.06, "learning_rate": 3.7474939159830014e-06, "loss": 0.5389, "step": 8910 }, { "epoch": 1.06, "learning_rate": 3.747215359036057e-06, "loss": 0.4934, "step": 8911 }, { "epoch": 1.06, "learning_rate": 3.746936781472539e-06, "loss": 0.549, "step": 8912 }, { "epoch": 1.06, "learning_rate": 3.7466581832970507e-06, "loss": 0.5267, "step": 8913 }, { "epoch": 1.06, "learning_rate": 3.7463795645141988e-06, "loss": 0.6845, "step": 8914 }, { "epoch": 1.06, "learning_rate": 3.7461009251285875e-06, "loss": 0.7948, "step": 8915 }, { "epoch": 1.06, "learning_rate": 3.7458222651448235e-06, "loss": 0.7028, "step": 8916 }, { "epoch": 1.06, "learning_rate": 3.7455435845675126e-06, "loss": 0.4717, "step": 8917 }, { "epoch": 1.06, "learning_rate": 3.745264883401262e-06, "loss": 0.5746, "step": 8918 }, { "epoch": 1.06, "learning_rate": 3.7449861616506782e-06, "loss": 0.644, "step": 8919 }, { "epoch": 1.06, "learning_rate": 3.7447074193203693e-06, "loss": 0.6907, "step": 8920 }, { "epoch": 1.06, "learning_rate": 3.744428656414942e-06, "loss": 0.7971, "step": 8921 }, { "epoch": 1.06, "learning_rate": 3.7441498729390046e-06, "loss": 0.7284, "step": 8922 }, { "epoch": 1.06, "learning_rate": 3.7438710688971645e-06, "loss": 0.38, "step": 8923 }, { "epoch": 1.06, "learning_rate": 3.7435922442940325e-06, "loss": 0.5171, "step": 8924 }, { "epoch": 1.06, "learning_rate": 3.7433133991342164e-06, "loss": 0.5775, "step": 8925 }, { "epoch": 1.06, "learning_rate": 3.743034533422325e-06, "loss": 0.5866, "step": 8926 }, { "epoch": 1.06, "learning_rate": 3.742755647162968e-06, "loss": 0.6826, "step": 8927 }, { "epoch": 1.06, "learning_rate": 3.742476740360757e-06, "loss": 0.8482, "step": 8928 }, { "epoch": 1.06, "learning_rate": 3.7421978130203008e-06, "loss": 0.3871, "step": 8929 }, { "epoch": 1.06, "learning_rate": 3.7419188651462096e-06, "loss": 0.5303, "step": 8930 }, { "epoch": 1.06, "learning_rate": 3.7416398967430967e-06, "loss": 0.5612, "step": 8931 }, { "epoch": 1.06, "learning_rate": 3.7413609078155712e-06, "loss": 0.673, "step": 8932 }, { "epoch": 1.06, "learning_rate": 3.741081898368246e-06, "loss": 0.7963, "step": 8933 }, { "epoch": 1.06, "learning_rate": 3.740802868405733e-06, "loss": 0.8118, "step": 8934 }, { "epoch": 1.06, "learning_rate": 3.740523817932644e-06, "loss": 0.5337, "step": 8935 }, { "epoch": 1.06, "learning_rate": 3.7402447469535928e-06, "loss": 0.5793, "step": 8936 }, { "epoch": 1.06, "learning_rate": 3.7399656554731922e-06, "loss": 0.5538, "step": 8937 }, { "epoch": 1.06, "learning_rate": 3.7396865434960538e-06, "loss": 0.5243, "step": 8938 }, { "epoch": 1.06, "learning_rate": 3.7394074110267946e-06, "loss": 0.6962, "step": 8939 }, { "epoch": 1.06, "learning_rate": 3.7391282580700256e-06, "loss": 0.7847, "step": 8940 }, { "epoch": 1.06, "learning_rate": 3.738849084630363e-06, "loss": 0.5073, "step": 8941 }, { "epoch": 1.06, "learning_rate": 3.738569890712421e-06, "loss": 0.4391, "step": 8942 }, { "epoch": 1.07, "learning_rate": 3.7382906763208153e-06, "loss": 0.5747, "step": 8943 }, { "epoch": 1.07, "learning_rate": 3.7380114414601603e-06, "loss": 0.5855, "step": 8944 }, { "epoch": 1.07, "learning_rate": 3.7377321861350724e-06, "loss": 0.6933, "step": 8945 }, { "epoch": 1.07, "learning_rate": 3.737452910350167e-06, "loss": 0.7935, "step": 8946 }, { "epoch": 1.07, "learning_rate": 3.7371736141100624e-06, "loss": 0.8819, "step": 8947 }, { "epoch": 1.07, "learning_rate": 3.7368942974193733e-06, "loss": 0.6396, "step": 8948 }, { "epoch": 1.07, "learning_rate": 3.736614960282717e-06, "loss": 0.5221, "step": 8949 }, { "epoch": 1.07, "learning_rate": 3.736335602704714e-06, "loss": 0.5835, "step": 8950 }, { "epoch": 1.07, "learning_rate": 3.7360562246899775e-06, "loss": 0.6548, "step": 8951 }, { "epoch": 1.07, "learning_rate": 3.735776826243129e-06, "loss": 0.7759, "step": 8952 }, { "epoch": 1.07, "learning_rate": 3.7354974073687855e-06, "loss": 0.9961, "step": 8953 }, { "epoch": 1.07, "learning_rate": 3.735217968071566e-06, "loss": 0.5589, "step": 8954 }, { "epoch": 1.07, "learning_rate": 3.7349385083560896e-06, "loss": 0.4656, "step": 8955 }, { "epoch": 1.07, "learning_rate": 3.734659028226977e-06, "loss": 0.5896, "step": 8956 }, { "epoch": 1.07, "learning_rate": 3.7343795276888463e-06, "loss": 0.632, "step": 8957 }, { "epoch": 1.07, "learning_rate": 3.734100006746319e-06, "loss": 0.8013, "step": 8958 }, { "epoch": 1.07, "learning_rate": 3.7338204654040146e-06, "loss": 0.7646, "step": 8959 }, { "epoch": 1.07, "learning_rate": 3.733540903666555e-06, "loss": 0.3568, "step": 8960 }, { "epoch": 1.07, "learning_rate": 3.7332613215385606e-06, "loss": 0.6089, "step": 8961 }, { "epoch": 1.07, "learning_rate": 3.732981719024652e-06, "loss": 0.4884, "step": 8962 }, { "epoch": 1.07, "learning_rate": 3.732702096129453e-06, "loss": 0.6056, "step": 8963 }, { "epoch": 1.07, "learning_rate": 3.7324224528575846e-06, "loss": 0.7339, "step": 8964 }, { "epoch": 1.07, "learning_rate": 3.7321427892136703e-06, "loss": 0.8228, "step": 8965 }, { "epoch": 1.07, "learning_rate": 3.7318631052023313e-06, "loss": 0.6483, "step": 8966 }, { "epoch": 1.07, "learning_rate": 3.7315834008281927e-06, "loss": 0.5129, "step": 8967 }, { "epoch": 1.07, "learning_rate": 3.7313036760958765e-06, "loss": 0.5587, "step": 8968 }, { "epoch": 1.07, "learning_rate": 3.7310239310100077e-06, "loss": 0.5613, "step": 8969 }, { "epoch": 1.07, "learning_rate": 3.7307441655752096e-06, "loss": 0.6126, "step": 8970 }, { "epoch": 1.07, "learning_rate": 3.7304643797961082e-06, "loss": 0.8525, "step": 8971 }, { "epoch": 1.07, "learning_rate": 3.730184573677326e-06, "loss": 0.8666, "step": 8972 }, { "epoch": 1.07, "learning_rate": 3.72990474722349e-06, "loss": 0.4862, "step": 8973 }, { "epoch": 1.07, "learning_rate": 3.729624900439226e-06, "loss": 0.5116, "step": 8974 }, { "epoch": 1.07, "learning_rate": 3.7293450333291585e-06, "loss": 0.6306, "step": 8975 }, { "epoch": 1.07, "learning_rate": 3.7290651458979147e-06, "loss": 0.699, "step": 8976 }, { "epoch": 1.07, "learning_rate": 3.7287852381501203e-06, "loss": 0.8421, "step": 8977 }, { "epoch": 1.07, "learning_rate": 3.7285053100904038e-06, "loss": 0.802, "step": 8978 }, { "epoch": 1.07, "learning_rate": 3.728225361723391e-06, "loss": 0.3667, "step": 8979 }, { "epoch": 1.07, "learning_rate": 3.72794539305371e-06, "loss": 0.5605, "step": 8980 }, { "epoch": 1.07, "learning_rate": 3.7276654040859884e-06, "loss": 0.5376, "step": 8981 }, { "epoch": 1.07, "learning_rate": 3.727385394824855e-06, "loss": 0.6506, "step": 8982 }, { "epoch": 1.07, "learning_rate": 3.7271053652749377e-06, "loss": 0.742, "step": 8983 }, { "epoch": 1.07, "learning_rate": 3.7268253154408663e-06, "loss": 1.0288, "step": 8984 }, { "epoch": 1.07, "learning_rate": 3.726545245327269e-06, "loss": 0.4089, "step": 8985 }, { "epoch": 1.07, "learning_rate": 3.726265154938776e-06, "loss": 0.5658, "step": 8986 }, { "epoch": 1.07, "learning_rate": 3.725985044280017e-06, "loss": 0.5283, "step": 8987 }, { "epoch": 1.07, "learning_rate": 3.7257049133556222e-06, "loss": 0.566, "step": 8988 }, { "epoch": 1.07, "learning_rate": 3.725424762170223e-06, "loss": 0.8341, "step": 8989 }, { "epoch": 1.07, "learning_rate": 3.725144590728449e-06, "loss": 0.7739, "step": 8990 }, { "epoch": 1.07, "learning_rate": 3.724864399034932e-06, "loss": 0.599, "step": 8991 }, { "epoch": 1.07, "learning_rate": 3.7245841870943046e-06, "loss": 0.4019, "step": 8992 }, { "epoch": 1.07, "learning_rate": 3.7243039549111976e-06, "loss": 0.5532, "step": 8993 }, { "epoch": 1.07, "learning_rate": 3.7240237024902424e-06, "loss": 0.6476, "step": 8994 }, { "epoch": 1.07, "learning_rate": 3.7237434298360744e-06, "loss": 0.7137, "step": 8995 }, { "epoch": 1.07, "learning_rate": 3.7234631369533234e-06, "loss": 0.9098, "step": 8996 }, { "epoch": 1.07, "learning_rate": 3.723182823846625e-06, "loss": 0.8358, "step": 8997 }, { "epoch": 1.07, "learning_rate": 3.7229024905206113e-06, "loss": 0.4267, "step": 8998 }, { "epoch": 1.07, "learning_rate": 3.7226221369799163e-06, "loss": 0.5558, "step": 8999 }, { "epoch": 1.07, "learning_rate": 3.7223417632291757e-06, "loss": 0.6515, "step": 9000 }, { "epoch": 1.07, "learning_rate": 3.7220613692730236e-06, "loss": 0.6773, "step": 9001 }, { "epoch": 1.07, "learning_rate": 3.7217809551160932e-06, "loss": 0.7997, "step": 9002 }, { "epoch": 1.07, "learning_rate": 3.721500520763023e-06, "loss": 0.8013, "step": 9003 }, { "epoch": 1.07, "learning_rate": 3.7212200662184445e-06, "loss": 0.4177, "step": 9004 }, { "epoch": 1.07, "learning_rate": 3.7209395914869973e-06, "loss": 0.4891, "step": 9005 }, { "epoch": 1.07, "learning_rate": 3.720659096573315e-06, "loss": 0.6304, "step": 9006 }, { "epoch": 1.07, "learning_rate": 3.720378581482037e-06, "loss": 0.5778, "step": 9007 }, { "epoch": 1.07, "learning_rate": 3.7200980462177977e-06, "loss": 0.7063, "step": 9008 }, { "epoch": 1.07, "learning_rate": 3.719817490785235e-06, "loss": 0.8047, "step": 9009 }, { "epoch": 1.07, "learning_rate": 3.7195369151889878e-06, "loss": 0.4975, "step": 9010 }, { "epoch": 1.07, "learning_rate": 3.719256319433692e-06, "loss": 0.6683, "step": 9011 }, { "epoch": 1.07, "learning_rate": 3.718975703523988e-06, "loss": 0.6506, "step": 9012 }, { "epoch": 1.07, "learning_rate": 3.718695067464512e-06, "loss": 0.6577, "step": 9013 }, { "epoch": 1.07, "learning_rate": 3.718414411259906e-06, "loss": 0.7051, "step": 9014 }, { "epoch": 1.07, "learning_rate": 3.718133734914806e-06, "loss": 0.7928, "step": 9015 }, { "epoch": 1.07, "learning_rate": 3.717853038433854e-06, "loss": 0.4566, "step": 9016 }, { "epoch": 1.07, "learning_rate": 3.717572321821689e-06, "loss": 0.5009, "step": 9017 }, { "epoch": 1.07, "learning_rate": 3.7172915850829515e-06, "loss": 0.6252, "step": 9018 }, { "epoch": 1.07, "learning_rate": 3.717010828222281e-06, "loss": 0.6281, "step": 9019 }, { "epoch": 1.07, "learning_rate": 3.7167300512443205e-06, "loss": 0.5862, "step": 9020 }, { "epoch": 1.07, "learning_rate": 3.7164492541537096e-06, "loss": 0.7645, "step": 9021 }, { "epoch": 1.07, "learning_rate": 3.716168436955091e-06, "loss": 0.9146, "step": 9022 }, { "epoch": 1.07, "learning_rate": 3.715887599653106e-06, "loss": 0.4458, "step": 9023 }, { "epoch": 1.07, "learning_rate": 3.7156067422523956e-06, "loss": 0.6278, "step": 9024 }, { "epoch": 1.07, "learning_rate": 3.715325864757606e-06, "loss": 0.5193, "step": 9025 }, { "epoch": 1.07, "learning_rate": 3.7150449671733755e-06, "loss": 0.6054, "step": 9026 }, { "epoch": 1.08, "learning_rate": 3.7147640495043513e-06, "loss": 0.7415, "step": 9027 }, { "epoch": 1.08, "learning_rate": 3.7144831117551748e-06, "loss": 0.9224, "step": 9028 }, { "epoch": 1.08, "learning_rate": 3.714202153930491e-06, "loss": 0.3547, "step": 9029 }, { "epoch": 1.08, "learning_rate": 3.713921176034944e-06, "loss": 0.5752, "step": 9030 }, { "epoch": 1.08, "learning_rate": 3.7136401780731775e-06, "loss": 0.5763, "step": 9031 }, { "epoch": 1.08, "learning_rate": 3.713359160049837e-06, "loss": 0.5901, "step": 9032 }, { "epoch": 1.08, "learning_rate": 3.713078121969568e-06, "loss": 0.7722, "step": 9033 }, { "epoch": 1.08, "learning_rate": 3.7127970638370154e-06, "loss": 0.8414, "step": 9034 }, { "epoch": 1.08, "learning_rate": 3.712515985656826e-06, "loss": 0.4658, "step": 9035 }, { "epoch": 1.08, "learning_rate": 3.7122348874336455e-06, "loss": 0.5307, "step": 9036 }, { "epoch": 1.08, "learning_rate": 3.7119537691721206e-06, "loss": 0.643, "step": 9037 }, { "epoch": 1.08, "learning_rate": 3.7116726308768985e-06, "loss": 0.6314, "step": 9038 }, { "epoch": 1.08, "learning_rate": 3.711391472552626e-06, "loss": 0.7266, "step": 9039 }, { "epoch": 1.08, "learning_rate": 3.7111102942039513e-06, "loss": 0.9103, "step": 9040 }, { "epoch": 1.08, "learning_rate": 3.710829095835521e-06, "loss": 0.5948, "step": 9041 }, { "epoch": 1.08, "learning_rate": 3.710547877451984e-06, "loss": 0.5163, "step": 9042 }, { "epoch": 1.08, "learning_rate": 3.71026663905799e-06, "loss": 0.5437, "step": 9043 }, { "epoch": 1.08, "learning_rate": 3.7099853806581865e-06, "loss": 0.5703, "step": 9044 }, { "epoch": 1.08, "learning_rate": 3.7097041022572224e-06, "loss": 0.7006, "step": 9045 }, { "epoch": 1.08, "learning_rate": 3.7094228038597484e-06, "loss": 0.8358, "step": 9046 }, { "epoch": 1.08, "learning_rate": 3.7091414854704144e-06, "loss": 0.7923, "step": 9047 }, { "epoch": 1.08, "learning_rate": 3.70886014709387e-06, "loss": 0.4019, "step": 9048 }, { "epoch": 1.08, "learning_rate": 3.7085787887347656e-06, "loss": 0.5906, "step": 9049 }, { "epoch": 1.08, "learning_rate": 3.708297410397753e-06, "loss": 0.6628, "step": 9050 }, { "epoch": 1.08, "learning_rate": 3.708016012087482e-06, "loss": 0.6765, "step": 9051 }, { "epoch": 1.08, "learning_rate": 3.7077345938086057e-06, "loss": 0.9036, "step": 9052 }, { "epoch": 1.08, "learning_rate": 3.707453155565775e-06, "loss": 0.8237, "step": 9053 }, { "epoch": 1.08, "learning_rate": 3.7071716973636422e-06, "loss": 0.3377, "step": 9054 }, { "epoch": 1.08, "learning_rate": 3.7068902192068605e-06, "loss": 0.4331, "step": 9055 }, { "epoch": 1.08, "learning_rate": 3.706608721100081e-06, "loss": 0.5482, "step": 9056 }, { "epoch": 1.08, "learning_rate": 3.7063272030479595e-06, "loss": 0.6154, "step": 9057 }, { "epoch": 1.08, "learning_rate": 3.7060456650551475e-06, "loss": 0.7438, "step": 9058 }, { "epoch": 1.08, "learning_rate": 3.7057641071262995e-06, "loss": 0.7805, "step": 9059 }, { "epoch": 1.08, "learning_rate": 3.7054825292660695e-06, "loss": 0.4665, "step": 9060 }, { "epoch": 1.08, "learning_rate": 3.7052009314791125e-06, "loss": 0.4267, "step": 9061 }, { "epoch": 1.08, "learning_rate": 3.7049193137700825e-06, "loss": 0.4976, "step": 9062 }, { "epoch": 1.08, "learning_rate": 3.704637676143635e-06, "loss": 0.6374, "step": 9063 }, { "epoch": 1.08, "learning_rate": 3.7043560186044257e-06, "loss": 0.6985, "step": 9064 }, { "epoch": 1.08, "learning_rate": 3.7040743411571113e-06, "loss": 0.8743, "step": 9065 }, { "epoch": 1.08, "learning_rate": 3.703792643806346e-06, "loss": 0.5951, "step": 9066 }, { "epoch": 1.08, "learning_rate": 3.7035109265567876e-06, "loss": 0.4408, "step": 9067 }, { "epoch": 1.08, "learning_rate": 3.7032291894130928e-06, "loss": 0.5181, "step": 9068 }, { "epoch": 1.08, "learning_rate": 3.7029474323799187e-06, "loss": 0.5344, "step": 9069 }, { "epoch": 1.08, "learning_rate": 3.7026656554619215e-06, "loss": 0.5865, "step": 9070 }, { "epoch": 1.08, "learning_rate": 3.7023838586637617e-06, "loss": 0.7912, "step": 9071 }, { "epoch": 1.08, "learning_rate": 3.702102041990095e-06, "loss": 0.7841, "step": 9072 }, { "epoch": 1.08, "learning_rate": 3.7018202054455798e-06, "loss": 0.4137, "step": 9073 }, { "epoch": 1.08, "learning_rate": 3.701538349034877e-06, "loss": 0.6313, "step": 9074 }, { "epoch": 1.08, "learning_rate": 3.7012564727626444e-06, "loss": 0.6088, "step": 9075 }, { "epoch": 1.08, "learning_rate": 3.7009745766335416e-06, "loss": 0.6978, "step": 9076 }, { "epoch": 1.08, "learning_rate": 3.700692660652227e-06, "loss": 0.8547, "step": 9077 }, { "epoch": 1.08, "learning_rate": 3.7004107248233636e-06, "loss": 0.8643, "step": 9078 }, { "epoch": 1.08, "learning_rate": 3.700128769151609e-06, "loss": 0.4848, "step": 9079 }, { "epoch": 1.08, "learning_rate": 3.699846793641626e-06, "loss": 0.4632, "step": 9080 }, { "epoch": 1.08, "learning_rate": 3.6995647982980744e-06, "loss": 0.606, "step": 9081 }, { "epoch": 1.08, "learning_rate": 3.699282783125616e-06, "loss": 0.6109, "step": 9082 }, { "epoch": 1.08, "learning_rate": 3.6990007481289127e-06, "loss": 0.7303, "step": 9083 }, { "epoch": 1.08, "learning_rate": 3.698718693312626e-06, "loss": 0.8604, "step": 9084 }, { "epoch": 1.08, "learning_rate": 3.698436618681419e-06, "loss": 0.4509, "step": 9085 }, { "epoch": 1.08, "learning_rate": 3.6981545242399545e-06, "loss": 0.6135, "step": 9086 }, { "epoch": 1.08, "learning_rate": 3.6978724099928954e-06, "loss": 0.5981, "step": 9087 }, { "epoch": 1.08, "learning_rate": 3.6975902759449037e-06, "loss": 0.713, "step": 9088 }, { "epoch": 1.08, "learning_rate": 3.697308122100645e-06, "loss": 0.5706, "step": 9089 }, { "epoch": 1.08, "learning_rate": 3.697025948464783e-06, "loss": 0.897, "step": 9090 }, { "epoch": 1.08, "learning_rate": 3.696743755041981e-06, "loss": 0.6527, "step": 9091 }, { "epoch": 1.08, "learning_rate": 3.6964615418369044e-06, "loss": 0.447, "step": 9092 }, { "epoch": 1.08, "learning_rate": 3.6961793088542185e-06, "loss": 0.5731, "step": 9093 }, { "epoch": 1.08, "learning_rate": 3.6958970560985874e-06, "loss": 0.6217, "step": 9094 }, { "epoch": 1.08, "learning_rate": 3.695614783574678e-06, "loss": 0.6462, "step": 9095 }, { "epoch": 1.08, "learning_rate": 3.6953324912871563e-06, "loss": 0.8574, "step": 9096 }, { "epoch": 1.08, "learning_rate": 3.6950501792406884e-06, "loss": 0.8688, "step": 9097 }, { "epoch": 1.08, "learning_rate": 3.6947678474399397e-06, "loss": 0.4816, "step": 9098 }, { "epoch": 1.08, "learning_rate": 3.6944854958895794e-06, "loss": 0.5686, "step": 9099 }, { "epoch": 1.08, "learning_rate": 3.694203124594273e-06, "loss": 0.5185, "step": 9100 }, { "epoch": 1.08, "learning_rate": 3.693920733558689e-06, "loss": 0.6397, "step": 9101 }, { "epoch": 1.08, "learning_rate": 3.6936383227874946e-06, "loss": 0.7393, "step": 9102 }, { "epoch": 1.08, "learning_rate": 3.6933558922853584e-06, "loss": 0.8804, "step": 9103 }, { "epoch": 1.08, "learning_rate": 3.693073442056951e-06, "loss": 0.4778, "step": 9104 }, { "epoch": 1.08, "learning_rate": 3.692790972106938e-06, "loss": 0.5465, "step": 9105 }, { "epoch": 1.08, "learning_rate": 3.69250848243999e-06, "loss": 0.6124, "step": 9106 }, { "epoch": 1.08, "learning_rate": 3.692225973060777e-06, "loss": 0.7399, "step": 9107 }, { "epoch": 1.08, "learning_rate": 3.6919434439739688e-06, "loss": 0.7761, "step": 9108 }, { "epoch": 1.08, "learning_rate": 3.6916608951842346e-06, "loss": 0.9412, "step": 9109 }, { "epoch": 1.08, "learning_rate": 3.691378326696247e-06, "loss": 0.3877, "step": 9110 }, { "epoch": 1.09, "learning_rate": 3.691095738514675e-06, "loss": 0.5686, "step": 9111 }, { "epoch": 1.09, "learning_rate": 3.6908131306441913e-06, "loss": 0.4548, "step": 9112 }, { "epoch": 1.09, "learning_rate": 3.6905305030894656e-06, "loss": 0.7126, "step": 9113 }, { "epoch": 1.09, "learning_rate": 3.690247855855172e-06, "loss": 0.6759, "step": 9114 }, { "epoch": 1.09, "learning_rate": 3.68996518894598e-06, "loss": 0.7037, "step": 9115 }, { "epoch": 1.09, "learning_rate": 3.689682502366565e-06, "loss": 0.623, "step": 9116 }, { "epoch": 1.09, "learning_rate": 3.689399796121598e-06, "loss": 0.4857, "step": 9117 }, { "epoch": 1.09, "learning_rate": 3.689117070215752e-06, "loss": 0.5201, "step": 9118 }, { "epoch": 1.09, "learning_rate": 3.6888343246537012e-06, "loss": 0.7281, "step": 9119 }, { "epoch": 1.09, "learning_rate": 3.688551559440119e-06, "loss": 0.7468, "step": 9120 }, { "epoch": 1.09, "learning_rate": 3.688268774579681e-06, "loss": 0.9052, "step": 9121 }, { "epoch": 1.09, "learning_rate": 3.6879859700770595e-06, "loss": 0.8833, "step": 9122 }, { "epoch": 1.09, "learning_rate": 3.6877031459369304e-06, "loss": 0.3846, "step": 9123 }, { "epoch": 1.09, "learning_rate": 3.687420302163968e-06, "loss": 0.5869, "step": 9124 }, { "epoch": 1.09, "learning_rate": 3.68713743876285e-06, "loss": 0.6118, "step": 9125 }, { "epoch": 1.09, "learning_rate": 3.686854555738249e-06, "loss": 0.6384, "step": 9126 }, { "epoch": 1.09, "learning_rate": 3.686571653094843e-06, "loss": 0.7382, "step": 9127 }, { "epoch": 1.09, "learning_rate": 3.6862887308373083e-06, "loss": 0.9407, "step": 9128 }, { "epoch": 1.09, "learning_rate": 3.6860057889703217e-06, "loss": 0.3617, "step": 9129 }, { "epoch": 1.09, "learning_rate": 3.685722827498559e-06, "loss": 0.5147, "step": 9130 }, { "epoch": 1.09, "learning_rate": 3.685439846426699e-06, "loss": 0.6196, "step": 9131 }, { "epoch": 1.09, "learning_rate": 3.6851568457594188e-06, "loss": 0.6076, "step": 9132 }, { "epoch": 1.09, "learning_rate": 3.684873825501396e-06, "loss": 0.6699, "step": 9133 }, { "epoch": 1.09, "learning_rate": 3.68459078565731e-06, "loss": 0.8737, "step": 9134 }, { "epoch": 1.09, "learning_rate": 3.6843077262318387e-06, "loss": 0.4684, "step": 9135 }, { "epoch": 1.09, "learning_rate": 3.684024647229662e-06, "loss": 0.5663, "step": 9136 }, { "epoch": 1.09, "learning_rate": 3.6837415486554577e-06, "loss": 0.5894, "step": 9137 }, { "epoch": 1.09, "learning_rate": 3.6834584305139065e-06, "loss": 0.5738, "step": 9138 }, { "epoch": 1.09, "learning_rate": 3.683175292809689e-06, "loss": 0.7, "step": 9139 }, { "epoch": 1.09, "learning_rate": 3.6828921355474834e-06, "loss": 0.8521, "step": 9140 }, { "epoch": 1.09, "learning_rate": 3.6826089587319714e-06, "loss": 0.5244, "step": 9141 }, { "epoch": 1.09, "learning_rate": 3.682325762367835e-06, "loss": 0.4051, "step": 9142 }, { "epoch": 1.09, "learning_rate": 3.6820425464597543e-06, "loss": 0.6434, "step": 9143 }, { "epoch": 1.09, "learning_rate": 3.6817593110124107e-06, "loss": 0.5677, "step": 9144 }, { "epoch": 1.09, "learning_rate": 3.681476056030486e-06, "loss": 0.7154, "step": 9145 }, { "epoch": 1.09, "learning_rate": 3.6811927815186645e-06, "loss": 0.7823, "step": 9146 }, { "epoch": 1.09, "learning_rate": 3.6809094874816263e-06, "loss": 0.7852, "step": 9147 }, { "epoch": 1.09, "learning_rate": 3.6806261739240557e-06, "loss": 0.3585, "step": 9148 }, { "epoch": 1.09, "learning_rate": 3.6803428408506347e-06, "loss": 0.4819, "step": 9149 }, { "epoch": 1.09, "learning_rate": 3.6800594882660485e-06, "loss": 0.571, "step": 9150 }, { "epoch": 1.09, "learning_rate": 3.6797761161749797e-06, "loss": 0.5959, "step": 9151 }, { "epoch": 1.09, "learning_rate": 3.6794927245821115e-06, "loss": 0.8099, "step": 9152 }, { "epoch": 1.09, "learning_rate": 3.679209313492131e-06, "loss": 0.8658, "step": 9153 }, { "epoch": 1.09, "learning_rate": 3.6789258829097207e-06, "loss": 0.504, "step": 9154 }, { "epoch": 1.09, "learning_rate": 3.6786424328395676e-06, "loss": 0.624, "step": 9155 }, { "epoch": 1.09, "learning_rate": 3.678358963286355e-06, "loss": 0.5334, "step": 9156 }, { "epoch": 1.09, "learning_rate": 3.6780754742547707e-06, "loss": 0.5274, "step": 9157 }, { "epoch": 1.09, "learning_rate": 3.6777919657494994e-06, "loss": 0.7844, "step": 9158 }, { "epoch": 1.09, "learning_rate": 3.6775084377752286e-06, "loss": 0.7608, "step": 9159 }, { "epoch": 1.09, "learning_rate": 3.6772248903366443e-06, "loss": 0.3694, "step": 9160 }, { "epoch": 1.09, "learning_rate": 3.676941323438434e-06, "loss": 0.3848, "step": 9161 }, { "epoch": 1.09, "learning_rate": 3.676657737085284e-06, "loss": 0.4924, "step": 9162 }, { "epoch": 1.09, "learning_rate": 3.6763741312818836e-06, "loss": 0.6642, "step": 9163 }, { "epoch": 1.09, "learning_rate": 3.67609050603292e-06, "loss": 0.6956, "step": 9164 }, { "epoch": 1.09, "learning_rate": 3.675806861343082e-06, "loss": 0.7284, "step": 9165 }, { "epoch": 1.09, "learning_rate": 3.675523197217058e-06, "loss": 0.4756, "step": 9166 }, { "epoch": 1.09, "learning_rate": 3.675239513659536e-06, "loss": 0.4478, "step": 9167 }, { "epoch": 1.09, "learning_rate": 3.674955810675207e-06, "loss": 0.5446, "step": 9168 }, { "epoch": 1.09, "learning_rate": 3.674672088268759e-06, "loss": 0.6232, "step": 9169 }, { "epoch": 1.09, "learning_rate": 3.674388346444884e-06, "loss": 0.6208, "step": 9170 }, { "epoch": 1.09, "learning_rate": 3.6741045852082703e-06, "loss": 0.7405, "step": 9171 }, { "epoch": 1.09, "learning_rate": 3.6738208045636098e-06, "loss": 0.85, "step": 9172 }, { "epoch": 1.09, "learning_rate": 3.673537004515592e-06, "loss": 0.3741, "step": 9173 }, { "epoch": 1.09, "learning_rate": 3.6732531850689103e-06, "loss": 0.5204, "step": 9174 }, { "epoch": 1.09, "learning_rate": 3.6729693462282544e-06, "loss": 0.5721, "step": 9175 }, { "epoch": 1.09, "learning_rate": 3.6726854879983164e-06, "loss": 0.629, "step": 9176 }, { "epoch": 1.09, "learning_rate": 3.672401610383789e-06, "loss": 0.8225, "step": 9177 }, { "epoch": 1.09, "learning_rate": 3.6721177133893647e-06, "loss": 0.9111, "step": 9178 }, { "epoch": 1.09, "learning_rate": 3.671833797019737e-06, "loss": 0.4409, "step": 9179 }, { "epoch": 1.09, "learning_rate": 3.6715498612795974e-06, "loss": 0.5906, "step": 9180 }, { "epoch": 1.09, "learning_rate": 3.6712659061736406e-06, "loss": 0.5167, "step": 9181 }, { "epoch": 1.09, "learning_rate": 3.6709819317065605e-06, "loss": 0.6243, "step": 9182 }, { "epoch": 1.09, "learning_rate": 3.6706979378830504e-06, "loss": 0.7432, "step": 9183 }, { "epoch": 1.09, "learning_rate": 3.670413924707805e-06, "loss": 0.73, "step": 9184 }, { "epoch": 1.09, "learning_rate": 3.6701298921855195e-06, "loss": 0.3752, "step": 9185 }, { "epoch": 1.09, "learning_rate": 3.6698458403208885e-06, "loss": 0.4686, "step": 9186 }, { "epoch": 1.09, "learning_rate": 3.669561769118608e-06, "loss": 0.5518, "step": 9187 }, { "epoch": 1.09, "learning_rate": 3.6692776785833726e-06, "loss": 0.6236, "step": 9188 }, { "epoch": 1.09, "learning_rate": 3.6689935687198796e-06, "loss": 0.6437, "step": 9189 }, { "epoch": 1.09, "learning_rate": 3.668709439532825e-06, "loss": 0.8199, "step": 9190 }, { "epoch": 1.09, "learning_rate": 3.6684252910269046e-06, "loss": 0.5182, "step": 9191 }, { "epoch": 1.09, "learning_rate": 3.6681411232068165e-06, "loss": 0.556, "step": 9192 }, { "epoch": 1.09, "learning_rate": 3.667856936077258e-06, "loss": 0.5568, "step": 9193 }, { "epoch": 1.09, "learning_rate": 3.6675727296429254e-06, "loss": 0.5988, "step": 9194 }, { "epoch": 1.1, "learning_rate": 3.667288503908518e-06, "loss": 0.786, "step": 9195 }, { "epoch": 1.1, "learning_rate": 3.6670042588787346e-06, "loss": 0.7534, "step": 9196 }, { "epoch": 1.1, "learning_rate": 3.666719994558272e-06, "loss": 0.8295, "step": 9197 }, { "epoch": 1.1, "learning_rate": 3.66643571095183e-06, "loss": 0.5296, "step": 9198 }, { "epoch": 1.1, "learning_rate": 3.666151408064108e-06, "loss": 0.4554, "step": 9199 }, { "epoch": 1.1, "learning_rate": 3.665867085899806e-06, "loss": 0.6567, "step": 9200 }, { "epoch": 1.1, "learning_rate": 3.665582744463622e-06, "loss": 0.7071, "step": 9201 }, { "epoch": 1.1, "learning_rate": 3.6652983837602583e-06, "loss": 0.7465, "step": 9202 }, { "epoch": 1.1, "learning_rate": 3.6650140037944137e-06, "loss": 0.7238, "step": 9203 }, { "epoch": 1.1, "learning_rate": 3.6647296045707914e-06, "loss": 0.4719, "step": 9204 }, { "epoch": 1.1, "learning_rate": 3.6644451860940895e-06, "loss": 0.6363, "step": 9205 }, { "epoch": 1.1, "learning_rate": 3.664160748369011e-06, "loss": 0.5833, "step": 9206 }, { "epoch": 1.1, "learning_rate": 3.663876291400259e-06, "loss": 0.6342, "step": 9207 }, { "epoch": 1.1, "learning_rate": 3.663591815192533e-06, "loss": 0.7911, "step": 9208 }, { "epoch": 1.1, "learning_rate": 3.663307319750537e-06, "loss": 0.8048, "step": 9209 }, { "epoch": 1.1, "learning_rate": 3.6630228050789735e-06, "loss": 0.4145, "step": 9210 }, { "epoch": 1.1, "learning_rate": 3.6627382711825455e-06, "loss": 0.5064, "step": 9211 }, { "epoch": 1.1, "learning_rate": 3.6624537180659565e-06, "loss": 0.6782, "step": 9212 }, { "epoch": 1.1, "learning_rate": 3.662169145733909e-06, "loss": 0.6629, "step": 9213 }, { "epoch": 1.1, "learning_rate": 3.66188455419111e-06, "loss": 0.6791, "step": 9214 }, { "epoch": 1.1, "learning_rate": 3.66159994344226e-06, "loss": 0.8432, "step": 9215 }, { "epoch": 1.1, "learning_rate": 3.6613153134920655e-06, "loss": 0.727, "step": 9216 }, { "epoch": 1.1, "learning_rate": 3.661030664345232e-06, "loss": 0.3688, "step": 9217 }, { "epoch": 1.1, "learning_rate": 3.6607459960064645e-06, "loss": 0.6122, "step": 9218 }, { "epoch": 1.1, "learning_rate": 3.660461308480468e-06, "loss": 0.6944, "step": 9219 }, { "epoch": 1.1, "learning_rate": 3.660176601771948e-06, "loss": 0.6669, "step": 9220 }, { "epoch": 1.1, "learning_rate": 3.659891875885613e-06, "loss": 0.8452, "step": 9221 }, { "epoch": 1.1, "learning_rate": 3.6596071308261666e-06, "loss": 0.784, "step": 9222 }, { "epoch": 1.1, "learning_rate": 3.6593223665983173e-06, "loss": 0.5756, "step": 9223 }, { "epoch": 1.1, "learning_rate": 3.659037583206772e-06, "loss": 0.6015, "step": 9224 }, { "epoch": 1.1, "learning_rate": 3.658752780656239e-06, "loss": 0.6122, "step": 9225 }, { "epoch": 1.1, "learning_rate": 3.6584679589514245e-06, "loss": 0.6414, "step": 9226 }, { "epoch": 1.1, "learning_rate": 3.6581831180970373e-06, "loss": 0.7051, "step": 9227 }, { "epoch": 1.1, "learning_rate": 3.6578982580977866e-06, "loss": 0.9273, "step": 9228 }, { "epoch": 1.1, "learning_rate": 3.65761337895838e-06, "loss": 0.4331, "step": 9229 }, { "epoch": 1.1, "learning_rate": 3.6573284806835273e-06, "loss": 0.5403, "step": 9230 }, { "epoch": 1.1, "learning_rate": 3.657043563277938e-06, "loss": 0.6633, "step": 9231 }, { "epoch": 1.1, "learning_rate": 3.656758626746321e-06, "loss": 0.5945, "step": 9232 }, { "epoch": 1.1, "learning_rate": 3.6564736710933874e-06, "loss": 0.7559, "step": 9233 }, { "epoch": 1.1, "learning_rate": 3.656188696323847e-06, "loss": 0.8126, "step": 9234 }, { "epoch": 1.1, "learning_rate": 3.6559037024424092e-06, "loss": 0.4363, "step": 9235 }, { "epoch": 1.1, "learning_rate": 3.6556186894537883e-06, "loss": 0.5206, "step": 9236 }, { "epoch": 1.1, "learning_rate": 3.655333657362692e-06, "loss": 0.5859, "step": 9237 }, { "epoch": 1.1, "learning_rate": 3.6550486061738333e-06, "loss": 0.62, "step": 9238 }, { "epoch": 1.1, "learning_rate": 3.654763535891925e-06, "loss": 0.6087, "step": 9239 }, { "epoch": 1.1, "learning_rate": 3.654478446521678e-06, "loss": 0.8014, "step": 9240 }, { "epoch": 1.1, "learning_rate": 3.6541933380678053e-06, "loss": 0.4249, "step": 9241 }, { "epoch": 1.1, "learning_rate": 3.6539082105350205e-06, "loss": 0.5558, "step": 9242 }, { "epoch": 1.1, "learning_rate": 3.6536230639280355e-06, "loss": 0.5843, "step": 9243 }, { "epoch": 1.1, "learning_rate": 3.6533378982515657e-06, "loss": 0.5597, "step": 9244 }, { "epoch": 1.1, "learning_rate": 3.6530527135103222e-06, "loss": 0.6375, "step": 9245 }, { "epoch": 1.1, "learning_rate": 3.652767509709021e-06, "loss": 0.8242, "step": 9246 }, { "epoch": 1.1, "learning_rate": 3.652482286852377e-06, "loss": 0.8146, "step": 9247 }, { "epoch": 1.1, "learning_rate": 3.6521970449451035e-06, "loss": 0.4811, "step": 9248 }, { "epoch": 1.1, "learning_rate": 3.651911783991916e-06, "loss": 0.6817, "step": 9249 }, { "epoch": 1.1, "learning_rate": 3.65162650399753e-06, "loss": 0.5277, "step": 9250 }, { "epoch": 1.1, "learning_rate": 3.651341204966662e-06, "loss": 0.6589, "step": 9251 }, { "epoch": 1.1, "learning_rate": 3.651055886904026e-06, "loss": 0.795, "step": 9252 }, { "epoch": 1.1, "learning_rate": 3.6507705498143405e-06, "loss": 0.8657, "step": 9253 }, { "epoch": 1.1, "learning_rate": 3.650485193702321e-06, "loss": 0.456, "step": 9254 }, { "epoch": 1.1, "learning_rate": 3.650199818572685e-06, "loss": 0.5225, "step": 9255 }, { "epoch": 1.1, "learning_rate": 3.649914424430149e-06, "loss": 0.604, "step": 9256 }, { "epoch": 1.1, "learning_rate": 3.6496290112794316e-06, "loss": 0.7228, "step": 9257 }, { "epoch": 1.1, "learning_rate": 3.6493435791252496e-06, "loss": 0.7636, "step": 9258 }, { "epoch": 1.1, "learning_rate": 3.649058127972322e-06, "loss": 0.8865, "step": 9259 }, { "epoch": 1.1, "learning_rate": 3.6487726578253675e-06, "loss": 0.3909, "step": 9260 }, { "epoch": 1.1, "learning_rate": 3.6484871686891044e-06, "loss": 0.6465, "step": 9261 }, { "epoch": 1.1, "learning_rate": 3.6482016605682517e-06, "loss": 0.5826, "step": 9262 }, { "epoch": 1.1, "learning_rate": 3.6479161334675294e-06, "loss": 0.662, "step": 9263 }, { "epoch": 1.1, "learning_rate": 3.647630587391657e-06, "loss": 0.6861, "step": 9264 }, { "epoch": 1.1, "learning_rate": 3.6473450223453555e-06, "loss": 0.7739, "step": 9265 }, { "epoch": 1.1, "learning_rate": 3.6470594383333437e-06, "loss": 0.5228, "step": 9266 }, { "epoch": 1.1, "learning_rate": 3.6467738353603427e-06, "loss": 0.5247, "step": 9267 }, { "epoch": 1.1, "learning_rate": 3.6464882134310755e-06, "loss": 0.5466, "step": 9268 }, { "epoch": 1.1, "learning_rate": 3.6462025725502605e-06, "loss": 0.5703, "step": 9269 }, { "epoch": 1.1, "learning_rate": 3.6459169127226212e-06, "loss": 0.6704, "step": 9270 }, { "epoch": 1.1, "learning_rate": 3.64563123395288e-06, "loss": 0.7184, "step": 9271 }, { "epoch": 1.1, "learning_rate": 3.6453455362457574e-06, "loss": 0.8549, "step": 9272 }, { "epoch": 1.1, "learning_rate": 3.6450598196059768e-06, "loss": 0.4455, "step": 9273 }, { "epoch": 1.1, "learning_rate": 3.6447740840382616e-06, "loss": 0.5044, "step": 9274 }, { "epoch": 1.1, "learning_rate": 3.6444883295473352e-06, "loss": 0.5615, "step": 9275 }, { "epoch": 1.1, "learning_rate": 3.644202556137921e-06, "loss": 0.6956, "step": 9276 }, { "epoch": 1.1, "learning_rate": 3.643916763814741e-06, "loss": 0.7186, "step": 9277 }, { "epoch": 1.1, "learning_rate": 3.643630952582522e-06, "loss": 0.817, "step": 9278 }, { "epoch": 1.11, "learning_rate": 3.643345122445988e-06, "loss": 0.448, "step": 9279 }, { "epoch": 1.11, "learning_rate": 3.6430592734098617e-06, "loss": 0.5748, "step": 9280 }, { "epoch": 1.11, "learning_rate": 3.6427734054788706e-06, "loss": 0.6323, "step": 9281 }, { "epoch": 1.11, "learning_rate": 3.642487518657739e-06, "loss": 0.6049, "step": 9282 }, { "epoch": 1.11, "learning_rate": 3.6422016129511934e-06, "loss": 0.8617, "step": 9283 }, { "epoch": 1.11, "learning_rate": 3.641915688363958e-06, "loss": 0.8414, "step": 9284 }, { "epoch": 1.11, "learning_rate": 3.6416297449007614e-06, "loss": 0.4518, "step": 9285 }, { "epoch": 1.11, "learning_rate": 3.6413437825663296e-06, "loss": 0.4195, "step": 9286 }, { "epoch": 1.11, "learning_rate": 3.6410578013653887e-06, "loss": 0.5954, "step": 9287 }, { "epoch": 1.11, "learning_rate": 3.6407718013026665e-06, "loss": 0.514, "step": 9288 }, { "epoch": 1.11, "learning_rate": 3.640485782382891e-06, "loss": 0.7226, "step": 9289 }, { "epoch": 1.11, "learning_rate": 3.64019974461079e-06, "loss": 0.9258, "step": 9290 }, { "epoch": 1.11, "learning_rate": 3.639913687991091e-06, "loss": 0.6643, "step": 9291 }, { "epoch": 1.11, "learning_rate": 3.6396276125285236e-06, "loss": 0.3888, "step": 9292 }, { "epoch": 1.11, "learning_rate": 3.6393415182278164e-06, "loss": 0.5928, "step": 9293 }, { "epoch": 1.11, "learning_rate": 3.6390554050936983e-06, "loss": 0.567, "step": 9294 }, { "epoch": 1.11, "learning_rate": 3.6387692731308975e-06, "loss": 0.6694, "step": 9295 }, { "epoch": 1.11, "learning_rate": 3.6384831223441465e-06, "loss": 0.8159, "step": 9296 }, { "epoch": 1.11, "learning_rate": 3.6381969527381732e-06, "loss": 0.9109, "step": 9297 }, { "epoch": 1.11, "learning_rate": 3.6379107643177094e-06, "loss": 0.3403, "step": 9298 }, { "epoch": 1.11, "learning_rate": 3.6376245570874843e-06, "loss": 0.5007, "step": 9299 }, { "epoch": 1.11, "learning_rate": 3.6373383310522307e-06, "loss": 0.5148, "step": 9300 }, { "epoch": 1.11, "learning_rate": 3.637052086216678e-06, "loss": 0.6129, "step": 9301 }, { "epoch": 1.11, "learning_rate": 3.63676582258556e-06, "loss": 0.8455, "step": 9302 }, { "epoch": 1.11, "learning_rate": 3.6364795401636067e-06, "loss": 0.8605, "step": 9303 }, { "epoch": 1.11, "learning_rate": 3.636193238955552e-06, "loss": 0.3955, "step": 9304 }, { "epoch": 1.11, "learning_rate": 3.635906918966127e-06, "loss": 0.4541, "step": 9305 }, { "epoch": 1.11, "learning_rate": 3.635620580200066e-06, "loss": 0.6089, "step": 9306 }, { "epoch": 1.11, "learning_rate": 3.6353342226621015e-06, "loss": 0.6771, "step": 9307 }, { "epoch": 1.11, "learning_rate": 3.6350478463569664e-06, "loss": 0.5488, "step": 9308 }, { "epoch": 1.11, "learning_rate": 3.634761451289395e-06, "loss": 0.9551, "step": 9309 }, { "epoch": 1.11, "learning_rate": 3.6344750374641225e-06, "loss": 0.4388, "step": 9310 }, { "epoch": 1.11, "learning_rate": 3.6341886048858826e-06, "loss": 0.509, "step": 9311 }, { "epoch": 1.11, "learning_rate": 3.6339021535594095e-06, "loss": 0.6296, "step": 9312 }, { "epoch": 1.11, "learning_rate": 3.633615683489438e-06, "loss": 0.5556, "step": 9313 }, { "epoch": 1.11, "learning_rate": 3.6333291946807047e-06, "loss": 0.7548, "step": 9314 }, { "epoch": 1.11, "learning_rate": 3.6330426871379454e-06, "loss": 0.9031, "step": 9315 }, { "epoch": 1.11, "learning_rate": 3.632756160865894e-06, "loss": 0.5211, "step": 9316 }, { "epoch": 1.11, "learning_rate": 3.632469615869289e-06, "loss": 0.5074, "step": 9317 }, { "epoch": 1.11, "learning_rate": 3.6321830521528662e-06, "loss": 0.4859, "step": 9318 }, { "epoch": 1.11, "learning_rate": 3.6318964697213626e-06, "loss": 0.6472, "step": 9319 }, { "epoch": 1.11, "learning_rate": 3.631609868579515e-06, "loss": 0.6736, "step": 9320 }, { "epoch": 1.11, "learning_rate": 3.6313232487320617e-06, "loss": 0.7857, "step": 9321 }, { "epoch": 1.11, "learning_rate": 3.63103661018374e-06, "loss": 0.8322, "step": 9322 }, { "epoch": 1.11, "learning_rate": 3.6307499529392886e-06, "loss": 0.4249, "step": 9323 }, { "epoch": 1.11, "learning_rate": 3.6304632770034448e-06, "loss": 0.5206, "step": 9324 }, { "epoch": 1.11, "learning_rate": 3.6301765823809488e-06, "loss": 0.5463, "step": 9325 }, { "epoch": 1.11, "learning_rate": 3.6298898690765393e-06, "loss": 0.7061, "step": 9326 }, { "epoch": 1.11, "learning_rate": 3.6296031370949547e-06, "loss": 0.7632, "step": 9327 }, { "epoch": 1.11, "learning_rate": 3.6293163864409358e-06, "loss": 0.9142, "step": 9328 }, { "epoch": 1.11, "learning_rate": 3.6290296171192225e-06, "loss": 0.35, "step": 9329 }, { "epoch": 1.11, "learning_rate": 3.628742829134555e-06, "loss": 0.461, "step": 9330 }, { "epoch": 1.11, "learning_rate": 3.628456022491673e-06, "loss": 0.5576, "step": 9331 }, { "epoch": 1.11, "learning_rate": 3.628169197195319e-06, "loss": 0.5565, "step": 9332 }, { "epoch": 1.11, "learning_rate": 3.6278823532502333e-06, "loss": 0.8126, "step": 9333 }, { "epoch": 1.11, "learning_rate": 3.6275954906611577e-06, "loss": 0.8421, "step": 9334 }, { "epoch": 1.11, "learning_rate": 3.6273086094328336e-06, "loss": 0.4886, "step": 9335 }, { "epoch": 1.11, "learning_rate": 3.6270217095700045e-06, "loss": 0.3866, "step": 9336 }, { "epoch": 1.11, "learning_rate": 3.626734791077411e-06, "loss": 0.5774, "step": 9337 }, { "epoch": 1.11, "learning_rate": 3.6264478539597976e-06, "loss": 0.5942, "step": 9338 }, { "epoch": 1.11, "learning_rate": 3.6261608982219065e-06, "loss": 0.6566, "step": 9339 }, { "epoch": 1.11, "learning_rate": 3.6258739238684813e-06, "loss": 0.8569, "step": 9340 }, { "epoch": 1.11, "learning_rate": 3.6255869309042656e-06, "loss": 0.4974, "step": 9341 }, { "epoch": 1.11, "learning_rate": 3.6252999193340034e-06, "loss": 0.5035, "step": 9342 }, { "epoch": 1.11, "learning_rate": 3.6250128891624394e-06, "loss": 0.4879, "step": 9343 }, { "epoch": 1.11, "learning_rate": 3.6247258403943175e-06, "loss": 0.5307, "step": 9344 }, { "epoch": 1.11, "learning_rate": 3.624438773034383e-06, "loss": 0.7361, "step": 9345 }, { "epoch": 1.11, "learning_rate": 3.624151687087381e-06, "loss": 0.9071, "step": 9346 }, { "epoch": 1.11, "learning_rate": 3.6238645825580585e-06, "loss": 0.8984, "step": 9347 }, { "epoch": 1.11, "learning_rate": 3.623577459451159e-06, "loss": 0.3355, "step": 9348 }, { "epoch": 1.11, "learning_rate": 3.6232903177714306e-06, "loss": 0.4148, "step": 9349 }, { "epoch": 1.11, "learning_rate": 3.6230031575236186e-06, "loss": 0.5438, "step": 9350 }, { "epoch": 1.11, "learning_rate": 3.62271597871247e-06, "loss": 0.6412, "step": 9351 }, { "epoch": 1.11, "learning_rate": 3.6224287813427323e-06, "loss": 0.9917, "step": 9352 }, { "epoch": 1.11, "learning_rate": 3.622141565419153e-06, "loss": 0.8518, "step": 9353 }, { "epoch": 1.11, "learning_rate": 3.6218543309464786e-06, "loss": 0.4252, "step": 9354 }, { "epoch": 1.11, "learning_rate": 3.6215670779294587e-06, "loss": 0.615, "step": 9355 }, { "epoch": 1.11, "learning_rate": 3.6212798063728404e-06, "loss": 0.6197, "step": 9356 }, { "epoch": 1.11, "learning_rate": 3.6209925162813734e-06, "loss": 0.6378, "step": 9357 }, { "epoch": 1.11, "learning_rate": 3.620705207659806e-06, "loss": 0.6988, "step": 9358 }, { "epoch": 1.11, "learning_rate": 3.620417880512887e-06, "loss": 0.8425, "step": 9359 }, { "epoch": 1.11, "learning_rate": 3.6201305348453667e-06, "loss": 0.4608, "step": 9360 }, { "epoch": 1.11, "learning_rate": 3.6198431706619945e-06, "loss": 0.5663, "step": 9361 }, { "epoch": 1.11, "learning_rate": 3.6195557879675215e-06, "loss": 0.571, "step": 9362 }, { "epoch": 1.12, "learning_rate": 3.6192683867666962e-06, "loss": 0.7296, "step": 9363 }, { "epoch": 1.12, "learning_rate": 3.6189809670642713e-06, "loss": 0.7751, "step": 9364 }, { "epoch": 1.12, "learning_rate": 3.6186935288649972e-06, "loss": 0.9976, "step": 9365 }, { "epoch": 1.12, "learning_rate": 3.6184060721736248e-06, "loss": 0.4698, "step": 9366 }, { "epoch": 1.12, "learning_rate": 3.6181185969949057e-06, "loss": 0.3463, "step": 9367 }, { "epoch": 1.12, "learning_rate": 3.6178311033335935e-06, "loss": 0.6629, "step": 9368 }, { "epoch": 1.12, "learning_rate": 3.6175435911944385e-06, "loss": 0.5903, "step": 9369 }, { "epoch": 1.12, "learning_rate": 3.617256060582195e-06, "loss": 0.6788, "step": 9370 }, { "epoch": 1.12, "learning_rate": 3.6169685115016147e-06, "loss": 0.7953, "step": 9371 }, { "epoch": 1.12, "learning_rate": 3.616680943957451e-06, "loss": 0.7528, "step": 9372 }, { "epoch": 1.12, "learning_rate": 3.6163933579544576e-06, "loss": 0.3677, "step": 9373 }, { "epoch": 1.12, "learning_rate": 3.6161057534973883e-06, "loss": 0.5656, "step": 9374 }, { "epoch": 1.12, "learning_rate": 3.6158181305909974e-06, "loss": 0.5722, "step": 9375 }, { "epoch": 1.12, "learning_rate": 3.615530489240039e-06, "loss": 0.6266, "step": 9376 }, { "epoch": 1.12, "learning_rate": 3.615242829449268e-06, "loss": 0.745, "step": 9377 }, { "epoch": 1.12, "learning_rate": 3.6149551512234394e-06, "loss": 0.8137, "step": 9378 }, { "epoch": 1.12, "learning_rate": 3.6146674545673094e-06, "loss": 0.4953, "step": 9379 }, { "epoch": 1.12, "learning_rate": 3.614379739485632e-06, "loss": 0.4158, "step": 9380 }, { "epoch": 1.12, "learning_rate": 3.6140920059831642e-06, "loss": 0.5478, "step": 9381 }, { "epoch": 1.12, "learning_rate": 3.6138042540646625e-06, "loss": 0.6788, "step": 9382 }, { "epoch": 1.12, "learning_rate": 3.6135164837348824e-06, "loss": 0.7302, "step": 9383 }, { "epoch": 1.12, "learning_rate": 3.6132286949985812e-06, "loss": 0.8097, "step": 9384 }, { "epoch": 1.12, "learning_rate": 3.6129408878605164e-06, "loss": 0.4671, "step": 9385 }, { "epoch": 1.12, "learning_rate": 3.6126530623254465e-06, "loss": 0.5194, "step": 9386 }, { "epoch": 1.12, "learning_rate": 3.6123652183981273e-06, "loss": 0.5423, "step": 9387 }, { "epoch": 1.12, "learning_rate": 3.612077356083317e-06, "loss": 0.636, "step": 9388 }, { "epoch": 1.12, "learning_rate": 3.611789475385775e-06, "loss": 0.6873, "step": 9389 }, { "epoch": 1.12, "learning_rate": 3.6115015763102612e-06, "loss": 0.7545, "step": 9390 }, { "epoch": 1.12, "learning_rate": 3.611213658861531e-06, "loss": 0.4491, "step": 9391 }, { "epoch": 1.12, "learning_rate": 3.610925723044347e-06, "loss": 0.4227, "step": 9392 }, { "epoch": 1.12, "learning_rate": 3.6106377688634675e-06, "loss": 0.5233, "step": 9393 }, { "epoch": 1.12, "learning_rate": 3.6103497963236523e-06, "loss": 0.6808, "step": 9394 }, { "epoch": 1.12, "learning_rate": 3.610061805429661e-06, "loss": 0.6395, "step": 9395 }, { "epoch": 1.12, "learning_rate": 3.609773796186256e-06, "loss": 0.8414, "step": 9396 }, { "epoch": 1.12, "learning_rate": 3.6094857685981966e-06, "loss": 0.7336, "step": 9397 }, { "epoch": 1.12, "learning_rate": 3.6091977226702446e-06, "loss": 0.4067, "step": 9398 }, { "epoch": 1.12, "learning_rate": 3.608909658407161e-06, "loss": 0.5706, "step": 9399 }, { "epoch": 1.12, "learning_rate": 3.608621575813708e-06, "loss": 0.5508, "step": 9400 }, { "epoch": 1.12, "learning_rate": 3.608333474894647e-06, "loss": 0.624, "step": 9401 }, { "epoch": 1.12, "learning_rate": 3.6080453556547407e-06, "loss": 0.8934, "step": 9402 }, { "epoch": 1.12, "learning_rate": 3.6077572180987514e-06, "loss": 0.8532, "step": 9403 }, { "epoch": 1.12, "learning_rate": 3.607469062231443e-06, "loss": 0.4047, "step": 9404 }, { "epoch": 1.12, "learning_rate": 3.6071808880575776e-06, "loss": 0.4617, "step": 9405 }, { "epoch": 1.12, "learning_rate": 3.6068926955819193e-06, "loss": 0.5843, "step": 9406 }, { "epoch": 1.12, "learning_rate": 3.606604484809232e-06, "loss": 0.6363, "step": 9407 }, { "epoch": 1.12, "learning_rate": 3.60631625574428e-06, "loss": 0.7975, "step": 9408 }, { "epoch": 1.12, "learning_rate": 3.606028008391827e-06, "loss": 0.8868, "step": 9409 }, { "epoch": 1.12, "learning_rate": 3.6057397427566376e-06, "loss": 0.5262, "step": 9410 }, { "epoch": 1.12, "learning_rate": 3.605451458843479e-06, "loss": 0.5144, "step": 9411 }, { "epoch": 1.12, "learning_rate": 3.6051631566571137e-06, "loss": 0.5083, "step": 9412 }, { "epoch": 1.12, "learning_rate": 3.6048748362023094e-06, "loss": 0.5675, "step": 9413 }, { "epoch": 1.12, "learning_rate": 3.6045864974838314e-06, "loss": 0.6169, "step": 9414 }, { "epoch": 1.12, "learning_rate": 3.604298140506446e-06, "loss": 0.9314, "step": 9415 }, { "epoch": 1.12, "learning_rate": 3.6040097652749184e-06, "loss": 0.6889, "step": 9416 }, { "epoch": 1.12, "learning_rate": 3.603721371794018e-06, "loss": 0.4683, "step": 9417 }, { "epoch": 1.12, "learning_rate": 3.6034329600685104e-06, "loss": 0.558, "step": 9418 }, { "epoch": 1.12, "learning_rate": 3.6031445301031638e-06, "loss": 0.682, "step": 9419 }, { "epoch": 1.12, "learning_rate": 3.6028560819027445e-06, "loss": 0.6306, "step": 9420 }, { "epoch": 1.12, "learning_rate": 3.6025676154720222e-06, "loss": 0.7156, "step": 9421 }, { "epoch": 1.12, "learning_rate": 3.6022791308157657e-06, "loss": 0.985, "step": 9422 }, { "epoch": 1.12, "learning_rate": 3.601990627938741e-06, "loss": 0.4493, "step": 9423 }, { "epoch": 1.12, "learning_rate": 3.6017021068457196e-06, "loss": 0.5538, "step": 9424 }, { "epoch": 1.12, "learning_rate": 3.60141356754147e-06, "loss": 0.5422, "step": 9425 }, { "epoch": 1.12, "learning_rate": 3.601125010030762e-06, "loss": 0.7337, "step": 9426 }, { "epoch": 1.12, "learning_rate": 3.6008364343183642e-06, "loss": 0.9322, "step": 9427 }, { "epoch": 1.12, "learning_rate": 3.6005478404090487e-06, "loss": 0.8812, "step": 9428 }, { "epoch": 1.12, "learning_rate": 3.6002592283075843e-06, "loss": 0.4753, "step": 9429 }, { "epoch": 1.12, "learning_rate": 3.5999705980187433e-06, "loss": 0.5294, "step": 9430 }, { "epoch": 1.12, "learning_rate": 3.599681949547295e-06, "loss": 0.5699, "step": 9431 }, { "epoch": 1.12, "learning_rate": 3.5993932828980127e-06, "loss": 0.5634, "step": 9432 }, { "epoch": 1.12, "learning_rate": 3.599104598075667e-06, "loss": 0.7787, "step": 9433 }, { "epoch": 1.12, "learning_rate": 3.5988158950850306e-06, "loss": 0.8895, "step": 9434 }, { "epoch": 1.12, "learning_rate": 3.5985271739308737e-06, "loss": 0.5695, "step": 9435 }, { "epoch": 1.12, "learning_rate": 3.5982384346179724e-06, "loss": 0.6075, "step": 9436 }, { "epoch": 1.12, "learning_rate": 3.5979496771510963e-06, "loss": 0.5484, "step": 9437 }, { "epoch": 1.12, "learning_rate": 3.59766090153502e-06, "loss": 0.6376, "step": 9438 }, { "epoch": 1.12, "learning_rate": 3.597372107774517e-06, "loss": 0.6854, "step": 9439 }, { "epoch": 1.12, "learning_rate": 3.597083295874362e-06, "loss": 0.8646, "step": 9440 }, { "epoch": 1.12, "learning_rate": 3.596794465839327e-06, "loss": 0.6003, "step": 9441 }, { "epoch": 1.12, "learning_rate": 3.596505617674187e-06, "loss": 0.4137, "step": 9442 }, { "epoch": 1.12, "learning_rate": 3.596216751383719e-06, "loss": 0.5016, "step": 9443 }, { "epoch": 1.12, "learning_rate": 3.595927866972694e-06, "loss": 0.6093, "step": 9444 }, { "epoch": 1.12, "learning_rate": 3.59563896444589e-06, "loss": 0.7001, "step": 9445 }, { "epoch": 1.12, "learning_rate": 3.595350043808082e-06, "loss": 0.7652, "step": 9446 }, { "epoch": 1.13, "learning_rate": 3.595061105064047e-06, "loss": 1.0274, "step": 9447 }, { "epoch": 1.13, "learning_rate": 3.5947721482185587e-06, "loss": 0.4669, "step": 9448 }, { "epoch": 1.13, "learning_rate": 3.5944831732763952e-06, "loss": 0.5929, "step": 9449 }, { "epoch": 1.13, "learning_rate": 3.5941941802423335e-06, "loss": 0.4907, "step": 9450 }, { "epoch": 1.13, "learning_rate": 3.5939051691211503e-06, "loss": 0.5805, "step": 9451 }, { "epoch": 1.13, "learning_rate": 3.5936161399176217e-06, "loss": 0.801, "step": 9452 }, { "epoch": 1.13, "learning_rate": 3.593327092636528e-06, "loss": 0.8124, "step": 9453 }, { "epoch": 1.13, "learning_rate": 3.5930380272826453e-06, "loss": 0.3763, "step": 9454 }, { "epoch": 1.13, "learning_rate": 3.5927489438607516e-06, "loss": 0.5037, "step": 9455 }, { "epoch": 1.13, "learning_rate": 3.5924598423756266e-06, "loss": 0.5862, "step": 9456 }, { "epoch": 1.13, "learning_rate": 3.592170722832049e-06, "loss": 0.6685, "step": 9457 }, { "epoch": 1.13, "learning_rate": 3.591881585234798e-06, "loss": 0.8604, "step": 9458 }, { "epoch": 1.13, "learning_rate": 3.591592429588652e-06, "loss": 0.8109, "step": 9459 }, { "epoch": 1.13, "learning_rate": 3.5913032558983914e-06, "loss": 0.4443, "step": 9460 }, { "epoch": 1.13, "learning_rate": 3.591014064168797e-06, "loss": 0.4492, "step": 9461 }, { "epoch": 1.13, "learning_rate": 3.5907248544046487e-06, "loss": 0.5333, "step": 9462 }, { "epoch": 1.13, "learning_rate": 3.590435626610726e-06, "loss": 0.6834, "step": 9463 }, { "epoch": 1.13, "learning_rate": 3.5901463807918117e-06, "loss": 0.7747, "step": 9464 }, { "epoch": 1.13, "learning_rate": 3.5898571169526863e-06, "loss": 0.8748, "step": 9465 }, { "epoch": 1.13, "learning_rate": 3.589567835098131e-06, "loss": 0.5495, "step": 9466 }, { "epoch": 1.13, "learning_rate": 3.589278535232928e-06, "loss": 0.4485, "step": 9467 }, { "epoch": 1.13, "learning_rate": 3.58898921736186e-06, "loss": 0.5792, "step": 9468 }, { "epoch": 1.13, "learning_rate": 3.588699881489709e-06, "loss": 0.6726, "step": 9469 }, { "epoch": 1.13, "learning_rate": 3.5884105276212567e-06, "loss": 0.6668, "step": 9470 }, { "epoch": 1.13, "learning_rate": 3.588121155761287e-06, "loss": 0.7922, "step": 9471 }, { "epoch": 1.13, "learning_rate": 3.587831765914584e-06, "loss": 0.8537, "step": 9472 }, { "epoch": 1.13, "learning_rate": 3.5875423580859303e-06, "loss": 0.4027, "step": 9473 }, { "epoch": 1.13, "learning_rate": 3.58725293228011e-06, "loss": 0.4956, "step": 9474 }, { "epoch": 1.13, "learning_rate": 3.586963488501908e-06, "loss": 0.6031, "step": 9475 }, { "epoch": 1.13, "learning_rate": 3.5866740267561077e-06, "loss": 0.5595, "step": 9476 }, { "epoch": 1.13, "learning_rate": 3.586384547047495e-06, "loss": 0.6169, "step": 9477 }, { "epoch": 1.13, "learning_rate": 3.5860950493808544e-06, "loss": 0.7535, "step": 9478 }, { "epoch": 1.13, "learning_rate": 3.585805533760972e-06, "loss": 0.4521, "step": 9479 }, { "epoch": 1.13, "learning_rate": 3.585516000192632e-06, "loss": 0.4374, "step": 9480 }, { "epoch": 1.13, "learning_rate": 3.585226448680622e-06, "loss": 0.5787, "step": 9481 }, { "epoch": 1.13, "learning_rate": 3.584936879229728e-06, "loss": 0.7014, "step": 9482 }, { "epoch": 1.13, "learning_rate": 3.5846472918447363e-06, "loss": 0.8036, "step": 9483 }, { "epoch": 1.13, "learning_rate": 3.584357686530434e-06, "loss": 0.8453, "step": 9484 }, { "epoch": 1.13, "learning_rate": 3.5840680632916068e-06, "loss": 0.5343, "step": 9485 }, { "epoch": 1.13, "learning_rate": 3.5837784221330458e-06, "loss": 0.5049, "step": 9486 }, { "epoch": 1.13, "learning_rate": 3.583488763059535e-06, "loss": 0.5328, "step": 9487 }, { "epoch": 1.13, "learning_rate": 3.5831990860758642e-06, "loss": 0.5974, "step": 9488 }, { "epoch": 1.13, "learning_rate": 3.582909391186821e-06, "loss": 0.7244, "step": 9489 }, { "epoch": 1.13, "learning_rate": 3.582619678397197e-06, "loss": 0.9255, "step": 9490 }, { "epoch": 1.13, "learning_rate": 3.582329947711777e-06, "loss": 0.7013, "step": 9491 }, { "epoch": 1.13, "learning_rate": 3.582040199135352e-06, "loss": 0.3711, "step": 9492 }, { "epoch": 1.13, "learning_rate": 3.5817504326727126e-06, "loss": 0.5547, "step": 9493 }, { "epoch": 1.13, "learning_rate": 3.5814606483286474e-06, "loss": 0.7851, "step": 9494 }, { "epoch": 1.13, "learning_rate": 3.5811708461079463e-06, "loss": 0.6883, "step": 9495 }, { "epoch": 1.13, "learning_rate": 3.5808810260154015e-06, "loss": 0.8022, "step": 9496 }, { "epoch": 1.13, "learning_rate": 3.580591188055802e-06, "loss": 1.0528, "step": 9497 }, { "epoch": 1.13, "learning_rate": 3.58030133223394e-06, "loss": 0.3504, "step": 9498 }, { "epoch": 1.13, "learning_rate": 3.5800114585546053e-06, "loss": 0.5876, "step": 9499 }, { "epoch": 1.13, "learning_rate": 3.5797215670225923e-06, "loss": 0.4995, "step": 9500 }, { "epoch": 1.13, "learning_rate": 3.57943165764269e-06, "loss": 0.6352, "step": 9501 }, { "epoch": 1.13, "learning_rate": 3.579141730419692e-06, "loss": 0.7377, "step": 9502 }, { "epoch": 1.13, "learning_rate": 3.5788517853583903e-06, "loss": 0.8998, "step": 9503 }, { "epoch": 1.13, "learning_rate": 3.5785618224635787e-06, "loss": 0.4683, "step": 9504 }, { "epoch": 1.13, "learning_rate": 3.5782718417400496e-06, "loss": 0.5504, "step": 9505 }, { "epoch": 1.13, "learning_rate": 3.5779818431925956e-06, "loss": 0.6555, "step": 9506 }, { "epoch": 1.13, "learning_rate": 3.577691826826012e-06, "loss": 0.7388, "step": 9507 }, { "epoch": 1.13, "learning_rate": 3.5774017926450922e-06, "loss": 0.8533, "step": 9508 }, { "epoch": 1.13, "learning_rate": 3.577111740654631e-06, "loss": 0.8654, "step": 9509 }, { "epoch": 1.13, "learning_rate": 3.5768216708594206e-06, "loss": 0.5078, "step": 9510 }, { "epoch": 1.13, "learning_rate": 3.5765315832642593e-06, "loss": 0.4092, "step": 9511 }, { "epoch": 1.13, "learning_rate": 3.5762414778739395e-06, "loss": 0.5831, "step": 9512 }, { "epoch": 1.13, "learning_rate": 3.5759513546932584e-06, "loss": 0.5974, "step": 9513 }, { "epoch": 1.13, "learning_rate": 3.5756612137270106e-06, "loss": 0.6927, "step": 9514 }, { "epoch": 1.13, "learning_rate": 3.575371054979993e-06, "loss": 0.9152, "step": 9515 }, { "epoch": 1.13, "learning_rate": 3.575080878457002e-06, "loss": 0.6244, "step": 9516 }, { "epoch": 1.13, "learning_rate": 3.5747906841628333e-06, "loss": 0.3911, "step": 9517 }, { "epoch": 1.13, "learning_rate": 3.5745004721022846e-06, "loss": 0.5117, "step": 9518 }, { "epoch": 1.13, "learning_rate": 3.574210242280153e-06, "loss": 0.4667, "step": 9519 }, { "epoch": 1.13, "learning_rate": 3.573919994701236e-06, "loss": 0.643, "step": 9520 }, { "epoch": 1.13, "learning_rate": 3.5736297293703306e-06, "loss": 0.7752, "step": 9521 }, { "epoch": 1.13, "learning_rate": 3.5733394462922364e-06, "loss": 0.6321, "step": 9522 }, { "epoch": 1.13, "learning_rate": 3.573049145471751e-06, "loss": 0.4823, "step": 9523 }, { "epoch": 1.13, "learning_rate": 3.5727588269136735e-06, "loss": 0.4722, "step": 9524 }, { "epoch": 1.13, "learning_rate": 3.5724684906228025e-06, "loss": 0.5809, "step": 9525 }, { "epoch": 1.13, "learning_rate": 3.572178136603937e-06, "loss": 0.6054, "step": 9526 }, { "epoch": 1.13, "learning_rate": 3.5718877648618765e-06, "loss": 0.7289, "step": 9527 }, { "epoch": 1.13, "learning_rate": 3.571597375401422e-06, "loss": 0.8677, "step": 9528 }, { "epoch": 1.13, "learning_rate": 3.571306968227373e-06, "loss": 0.58, "step": 9529 }, { "epoch": 1.13, "learning_rate": 3.5710165433445304e-06, "loss": 0.5641, "step": 9530 }, { "epoch": 1.14, "learning_rate": 3.570726100757693e-06, "loss": 0.5742, "step": 9531 }, { "epoch": 1.14, "learning_rate": 3.5704356404716646e-06, "loss": 0.6718, "step": 9532 }, { "epoch": 1.14, "learning_rate": 3.5701451624912453e-06, "loss": 0.7598, "step": 9533 }, { "epoch": 1.14, "learning_rate": 3.5698546668212357e-06, "loss": 0.8584, "step": 9534 }, { "epoch": 1.14, "learning_rate": 3.5695641534664394e-06, "loss": 0.5126, "step": 9535 }, { "epoch": 1.14, "learning_rate": 3.5692736224316583e-06, "loss": 0.6106, "step": 9536 }, { "epoch": 1.14, "learning_rate": 3.5689830737216936e-06, "loss": 0.5678, "step": 9537 }, { "epoch": 1.14, "learning_rate": 3.568692507341349e-06, "loss": 0.7571, "step": 9538 }, { "epoch": 1.14, "learning_rate": 3.568401923295428e-06, "loss": 0.8123, "step": 9539 }, { "epoch": 1.14, "learning_rate": 3.5681113215887337e-06, "loss": 0.8932, "step": 9540 }, { "epoch": 1.14, "learning_rate": 3.567820702226069e-06, "loss": 0.5948, "step": 9541 }, { "epoch": 1.14, "learning_rate": 3.567530065212239e-06, "loss": 0.4293, "step": 9542 }, { "epoch": 1.14, "learning_rate": 3.567239410552048e-06, "loss": 0.5818, "step": 9543 }, { "epoch": 1.14, "learning_rate": 3.5669487382502987e-06, "loss": 0.6035, "step": 9544 }, { "epoch": 1.14, "learning_rate": 3.566658048311798e-06, "loss": 0.7638, "step": 9545 }, { "epoch": 1.14, "learning_rate": 3.5663673407413495e-06, "loss": 0.8742, "step": 9546 }, { "epoch": 1.14, "learning_rate": 3.566076615543761e-06, "loss": 0.7817, "step": 9547 }, { "epoch": 1.14, "learning_rate": 3.565785872723835e-06, "loss": 0.6403, "step": 9548 }, { "epoch": 1.14, "learning_rate": 3.5654951122863798e-06, "loss": 0.6353, "step": 9549 }, { "epoch": 1.14, "learning_rate": 3.5652043342362007e-06, "loss": 0.5914, "step": 9550 }, { "epoch": 1.14, "learning_rate": 3.564913538578105e-06, "loss": 0.4787, "step": 9551 }, { "epoch": 1.14, "learning_rate": 3.5646227253168986e-06, "loss": 0.7896, "step": 9552 }, { "epoch": 1.14, "learning_rate": 3.5643318944573895e-06, "loss": 0.7432, "step": 9553 }, { "epoch": 1.14, "learning_rate": 3.5640410460043856e-06, "loss": 0.5475, "step": 9554 }, { "epoch": 1.14, "learning_rate": 3.5637501799626926e-06, "loss": 0.6393, "step": 9555 }, { "epoch": 1.14, "learning_rate": 3.563459296337121e-06, "loss": 0.5053, "step": 9556 }, { "epoch": 1.14, "learning_rate": 3.5631683951324775e-06, "loss": 0.65, "step": 9557 }, { "epoch": 1.14, "learning_rate": 3.5628774763535713e-06, "loss": 0.7607, "step": 9558 }, { "epoch": 1.14, "learning_rate": 3.562586540005211e-06, "loss": 0.8207, "step": 9559 }, { "epoch": 1.14, "learning_rate": 3.562295586092206e-06, "loss": 0.5072, "step": 9560 }, { "epoch": 1.14, "learning_rate": 3.562004614619366e-06, "loss": 0.5371, "step": 9561 }, { "epoch": 1.14, "learning_rate": 3.561713625591501e-06, "loss": 0.5366, "step": 9562 }, { "epoch": 1.14, "learning_rate": 3.5614226190134194e-06, "loss": 0.592, "step": 9563 }, { "epoch": 1.14, "learning_rate": 3.561131594889934e-06, "loss": 0.7369, "step": 9564 }, { "epoch": 1.14, "learning_rate": 3.560840553225855e-06, "loss": 0.8057, "step": 9565 }, { "epoch": 1.14, "learning_rate": 3.560549494025991e-06, "loss": 0.5752, "step": 9566 }, { "epoch": 1.14, "learning_rate": 3.560258417295156e-06, "loss": 0.5197, "step": 9567 }, { "epoch": 1.14, "learning_rate": 3.5599673230381596e-06, "loss": 0.653, "step": 9568 }, { "epoch": 1.14, "learning_rate": 3.559676211259815e-06, "loss": 0.5602, "step": 9569 }, { "epoch": 1.14, "learning_rate": 3.5593850819649333e-06, "loss": 0.7198, "step": 9570 }, { "epoch": 1.14, "learning_rate": 3.5590939351583275e-06, "loss": 0.7431, "step": 9571 }, { "epoch": 1.14, "learning_rate": 3.55880277084481e-06, "loss": 0.653, "step": 9572 }, { "epoch": 1.14, "learning_rate": 3.558511589029194e-06, "loss": 0.414, "step": 9573 }, { "epoch": 1.14, "learning_rate": 3.5582203897162924e-06, "loss": 0.5695, "step": 9574 }, { "epoch": 1.14, "learning_rate": 3.5579291729109194e-06, "loss": 0.5192, "step": 9575 }, { "epoch": 1.14, "learning_rate": 3.5576379386178883e-06, "loss": 0.6262, "step": 9576 }, { "epoch": 1.14, "learning_rate": 3.557346686842013e-06, "loss": 0.6987, "step": 9577 }, { "epoch": 1.14, "learning_rate": 3.557055417588108e-06, "loss": 0.8768, "step": 9578 }, { "epoch": 1.14, "learning_rate": 3.556764130860989e-06, "loss": 0.6017, "step": 9579 }, { "epoch": 1.14, "learning_rate": 3.5564728266654702e-06, "loss": 0.5359, "step": 9580 }, { "epoch": 1.14, "learning_rate": 3.5561815050063664e-06, "loss": 0.501, "step": 9581 }, { "epoch": 1.14, "learning_rate": 3.5558901658884943e-06, "loss": 0.5588, "step": 9582 }, { "epoch": 1.14, "learning_rate": 3.555598809316669e-06, "loss": 0.8019, "step": 9583 }, { "epoch": 1.14, "learning_rate": 3.555307435295706e-06, "loss": 0.9003, "step": 9584 }, { "epoch": 1.14, "learning_rate": 3.555016043830423e-06, "loss": 0.54, "step": 9585 }, { "epoch": 1.14, "learning_rate": 3.554724634925637e-06, "loss": 0.3977, "step": 9586 }, { "epoch": 1.14, "learning_rate": 3.5544332085861633e-06, "loss": 0.5451, "step": 9587 }, { "epoch": 1.14, "learning_rate": 3.554141764816821e-06, "loss": 0.6275, "step": 9588 }, { "epoch": 1.14, "learning_rate": 3.553850303622426e-06, "loss": 0.6481, "step": 9589 }, { "epoch": 1.14, "learning_rate": 3.553558825007798e-06, "loss": 0.8705, "step": 9590 }, { "epoch": 1.14, "learning_rate": 3.5532673289777534e-06, "loss": 0.6027, "step": 9591 }, { "epoch": 1.14, "learning_rate": 3.552975815537112e-06, "loss": 0.3567, "step": 9592 }, { "epoch": 1.14, "learning_rate": 3.552684284690692e-06, "loss": 0.4392, "step": 9593 }, { "epoch": 1.14, "learning_rate": 3.5523927364433125e-06, "loss": 0.4929, "step": 9594 }, { "epoch": 1.14, "learning_rate": 3.5521011707997923e-06, "loss": 0.5785, "step": 9595 }, { "epoch": 1.14, "learning_rate": 3.551809587764952e-06, "loss": 0.8262, "step": 9596 }, { "epoch": 1.14, "learning_rate": 3.551517987343611e-06, "loss": 0.8228, "step": 9597 }, { "epoch": 1.14, "learning_rate": 3.5512263695405885e-06, "loss": 0.3518, "step": 9598 }, { "epoch": 1.14, "learning_rate": 3.550934734360707e-06, "loss": 0.4954, "step": 9599 }, { "epoch": 1.14, "learning_rate": 3.5506430818087854e-06, "loss": 0.5609, "step": 9600 }, { "epoch": 1.14, "learning_rate": 3.5503514118896463e-06, "loss": 0.6487, "step": 9601 }, { "epoch": 1.14, "learning_rate": 3.5500597246081092e-06, "loss": 0.7396, "step": 9602 }, { "epoch": 1.14, "learning_rate": 3.5497680199689977e-06, "loss": 0.8961, "step": 9603 }, { "epoch": 1.14, "learning_rate": 3.5494762979771325e-06, "loss": 0.3943, "step": 9604 }, { "epoch": 1.14, "learning_rate": 3.549184558637336e-06, "loss": 0.4228, "step": 9605 }, { "epoch": 1.14, "learning_rate": 3.54889280195443e-06, "loss": 0.5533, "step": 9606 }, { "epoch": 1.14, "learning_rate": 3.5486010279332383e-06, "loss": 0.6153, "step": 9607 }, { "epoch": 1.14, "learning_rate": 3.548309236578584e-06, "loss": 0.7589, "step": 9608 }, { "epoch": 1.14, "learning_rate": 3.54801742789529e-06, "loss": 0.7573, "step": 9609 }, { "epoch": 1.14, "learning_rate": 3.5477256018881796e-06, "loss": 0.5287, "step": 9610 }, { "epoch": 1.14, "learning_rate": 3.547433758562078e-06, "loss": 0.4755, "step": 9611 }, { "epoch": 1.14, "learning_rate": 3.547141897921808e-06, "loss": 0.6299, "step": 9612 }, { "epoch": 1.14, "learning_rate": 3.546850019972194e-06, "loss": 0.6375, "step": 9613 }, { "epoch": 1.14, "learning_rate": 3.546558124718062e-06, "loss": 0.6735, "step": 9614 }, { "epoch": 1.15, "learning_rate": 3.546266212164237e-06, "loss": 0.8591, "step": 9615 }, { "epoch": 1.15, "learning_rate": 3.5459742823155423e-06, "loss": 0.5856, "step": 9616 }, { "epoch": 1.15, "learning_rate": 3.5456823351768048e-06, "loss": 0.444, "step": 9617 }, { "epoch": 1.15, "learning_rate": 3.5453903707528513e-06, "loss": 0.5992, "step": 9618 }, { "epoch": 1.15, "learning_rate": 3.545098389048508e-06, "loss": 0.5936, "step": 9619 }, { "epoch": 1.15, "learning_rate": 3.5448063900685995e-06, "loss": 0.7105, "step": 9620 }, { "epoch": 1.15, "learning_rate": 3.544514373817954e-06, "loss": 0.8293, "step": 9621 }, { "epoch": 1.15, "learning_rate": 3.5442223403013985e-06, "loss": 0.8784, "step": 9622 }, { "epoch": 1.15, "learning_rate": 3.54393028952376e-06, "loss": 0.5193, "step": 9623 }, { "epoch": 1.15, "learning_rate": 3.543638221489866e-06, "loss": 0.685, "step": 9624 }, { "epoch": 1.15, "learning_rate": 3.543346136204545e-06, "loss": 0.4835, "step": 9625 }, { "epoch": 1.15, "learning_rate": 3.543054033672625e-06, "loss": 0.6121, "step": 9626 }, { "epoch": 1.15, "learning_rate": 3.542761913898933e-06, "loss": 0.864, "step": 9627 }, { "epoch": 1.15, "learning_rate": 3.5424697768883004e-06, "loss": 0.7522, "step": 9628 }, { "epoch": 1.15, "learning_rate": 3.5421776226455552e-06, "loss": 0.4646, "step": 9629 }, { "epoch": 1.15, "learning_rate": 3.5418854511755253e-06, "loss": 0.5898, "step": 9630 }, { "epoch": 1.15, "learning_rate": 3.541593262483042e-06, "loss": 0.5175, "step": 9631 }, { "epoch": 1.15, "learning_rate": 3.5413010565729342e-06, "loss": 0.7173, "step": 9632 }, { "epoch": 1.15, "learning_rate": 3.5410088334500335e-06, "loss": 0.8083, "step": 9633 }, { "epoch": 1.15, "learning_rate": 3.540716593119169e-06, "loss": 0.749, "step": 9634 }, { "epoch": 1.15, "learning_rate": 3.5404243355851717e-06, "loss": 0.4802, "step": 9635 }, { "epoch": 1.15, "learning_rate": 3.5401320608528733e-06, "loss": 0.5689, "step": 9636 }, { "epoch": 1.15, "learning_rate": 3.539839768927104e-06, "loss": 0.5093, "step": 9637 }, { "epoch": 1.15, "learning_rate": 3.539547459812696e-06, "loss": 0.5591, "step": 9638 }, { "epoch": 1.15, "learning_rate": 3.5392551335144822e-06, "loss": 0.7043, "step": 9639 }, { "epoch": 1.15, "learning_rate": 3.538962790037293e-06, "loss": 0.8287, "step": 9640 }, { "epoch": 1.15, "learning_rate": 3.538670429385962e-06, "loss": 0.4672, "step": 9641 }, { "epoch": 1.15, "learning_rate": 3.5383780515653216e-06, "loss": 0.4546, "step": 9642 }, { "epoch": 1.15, "learning_rate": 3.5380856565802054e-06, "loss": 0.4457, "step": 9643 }, { "epoch": 1.15, "learning_rate": 3.5377932444354455e-06, "loss": 0.6448, "step": 9644 }, { "epoch": 1.15, "learning_rate": 3.5375008151358765e-06, "loss": 0.6649, "step": 9645 }, { "epoch": 1.15, "learning_rate": 3.537208368686332e-06, "loss": 0.8337, "step": 9646 }, { "epoch": 1.15, "learning_rate": 3.5369159050916455e-06, "loss": 0.9589, "step": 9647 }, { "epoch": 1.15, "learning_rate": 3.5366234243566534e-06, "loss": 0.413, "step": 9648 }, { "epoch": 1.15, "learning_rate": 3.5363309264861877e-06, "loss": 0.5847, "step": 9649 }, { "epoch": 1.15, "learning_rate": 3.5360384114850854e-06, "loss": 0.5753, "step": 9650 }, { "epoch": 1.15, "learning_rate": 3.535745879358181e-06, "loss": 0.666, "step": 9651 }, { "epoch": 1.15, "learning_rate": 3.535453330110311e-06, "loss": 0.7278, "step": 9652 }, { "epoch": 1.15, "learning_rate": 3.53516076374631e-06, "loss": 0.8672, "step": 9653 }, { "epoch": 1.15, "learning_rate": 3.534868180271015e-06, "loss": 0.2772, "step": 9654 }, { "epoch": 1.15, "learning_rate": 3.5345755796892617e-06, "loss": 0.5979, "step": 9655 }, { "epoch": 1.15, "learning_rate": 3.5342829620058877e-06, "loss": 0.4398, "step": 9656 }, { "epoch": 1.15, "learning_rate": 3.5339903272257298e-06, "loss": 0.6602, "step": 9657 }, { "epoch": 1.15, "learning_rate": 3.5336976753536247e-06, "loss": 0.7935, "step": 9658 }, { "epoch": 1.15, "learning_rate": 3.5334050063944104e-06, "loss": 0.946, "step": 9659 }, { "epoch": 1.15, "learning_rate": 3.533112320352924e-06, "loss": 0.5594, "step": 9660 }, { "epoch": 1.15, "learning_rate": 3.5328196172340045e-06, "loss": 0.4897, "step": 9661 }, { "epoch": 1.15, "learning_rate": 3.5325268970424907e-06, "loss": 0.5795, "step": 9662 }, { "epoch": 1.15, "learning_rate": 3.5322341597832206e-06, "loss": 0.6708, "step": 9663 }, { "epoch": 1.15, "learning_rate": 3.531941405461032e-06, "loss": 0.6899, "step": 9664 }, { "epoch": 1.15, "learning_rate": 3.531648634080767e-06, "loss": 0.8276, "step": 9665 }, { "epoch": 1.15, "learning_rate": 3.5313558456472623e-06, "loss": 0.6007, "step": 9666 }, { "epoch": 1.15, "learning_rate": 3.53106304016536e-06, "loss": 0.4347, "step": 9667 }, { "epoch": 1.15, "learning_rate": 3.5307702176398982e-06, "loss": 0.4733, "step": 9668 }, { "epoch": 1.15, "learning_rate": 3.530477378075719e-06, "loss": 0.5297, "step": 9669 }, { "epoch": 1.15, "learning_rate": 3.5301845214776615e-06, "loss": 0.6898, "step": 9670 }, { "epoch": 1.15, "learning_rate": 3.5298916478505678e-06, "loss": 0.7731, "step": 9671 }, { "epoch": 1.15, "learning_rate": 3.5295987571992786e-06, "loss": 0.7273, "step": 9672 }, { "epoch": 1.15, "learning_rate": 3.529305849528636e-06, "loss": 0.3982, "step": 9673 }, { "epoch": 1.15, "learning_rate": 3.529012924843481e-06, "loss": 0.6427, "step": 9674 }, { "epoch": 1.15, "learning_rate": 3.528719983148656e-06, "loss": 0.6138, "step": 9675 }, { "epoch": 1.15, "learning_rate": 3.528427024449004e-06, "loss": 0.7206, "step": 9676 }, { "epoch": 1.15, "learning_rate": 3.5281340487493654e-06, "loss": 0.7576, "step": 9677 }, { "epoch": 1.15, "learning_rate": 3.527841056054586e-06, "loss": 0.8265, "step": 9678 }, { "epoch": 1.15, "learning_rate": 3.5275480463695077e-06, "loss": 0.3479, "step": 9679 }, { "epoch": 1.15, "learning_rate": 3.5272550196989734e-06, "loss": 0.5727, "step": 9680 }, { "epoch": 1.15, "learning_rate": 3.5269619760478273e-06, "loss": 0.6089, "step": 9681 }, { "epoch": 1.15, "learning_rate": 3.5266689154209144e-06, "loss": 0.6135, "step": 9682 }, { "epoch": 1.15, "learning_rate": 3.5263758378230777e-06, "loss": 0.7954, "step": 9683 }, { "epoch": 1.15, "learning_rate": 3.5260827432591626e-06, "loss": 0.8754, "step": 9684 }, { "epoch": 1.15, "learning_rate": 3.525789631734013e-06, "loss": 0.4871, "step": 9685 }, { "epoch": 1.15, "learning_rate": 3.5254965032524756e-06, "loss": 0.6313, "step": 9686 }, { "epoch": 1.15, "learning_rate": 3.5252033578193938e-06, "loss": 0.6101, "step": 9687 }, { "epoch": 1.15, "learning_rate": 3.5249101954396153e-06, "loss": 0.5409, "step": 9688 }, { "epoch": 1.15, "learning_rate": 3.524617016117985e-06, "loss": 0.7776, "step": 9689 }, { "epoch": 1.15, "learning_rate": 3.52432381985935e-06, "loss": 0.7959, "step": 9690 }, { "epoch": 1.15, "learning_rate": 3.524030606668556e-06, "loss": 0.6585, "step": 9691 }, { "epoch": 1.15, "learning_rate": 3.5237373765504497e-06, "loss": 0.4693, "step": 9692 }, { "epoch": 1.15, "learning_rate": 3.5234441295098786e-06, "loss": 0.5138, "step": 9693 }, { "epoch": 1.15, "learning_rate": 3.523150865551691e-06, "loss": 0.569, "step": 9694 }, { "epoch": 1.15, "learning_rate": 3.5228575846807333e-06, "loss": 0.6805, "step": 9695 }, { "epoch": 1.15, "learning_rate": 3.5225642869018535e-06, "loss": 0.7938, "step": 9696 }, { "epoch": 1.15, "learning_rate": 3.522270972219901e-06, "loss": 0.8214, "step": 9697 }, { "epoch": 1.15, "learning_rate": 3.5219776406397226e-06, "loss": 0.506, "step": 9698 }, { "epoch": 1.16, "learning_rate": 3.5216842921661687e-06, "loss": 0.5931, "step": 9699 }, { "epoch": 1.16, "learning_rate": 3.521390926804088e-06, "loss": 0.6089, "step": 9700 }, { "epoch": 1.16, "learning_rate": 3.5210975445583294e-06, "loss": 0.6224, "step": 9701 }, { "epoch": 1.16, "learning_rate": 3.520804145433742e-06, "loss": 0.8572, "step": 9702 }, { "epoch": 1.16, "learning_rate": 3.520510729435177e-06, "loss": 0.7405, "step": 9703 }, { "epoch": 1.16, "learning_rate": 3.520217296567484e-06, "loss": 0.4704, "step": 9704 }, { "epoch": 1.16, "learning_rate": 3.5199238468355136e-06, "loss": 0.4937, "step": 9705 }, { "epoch": 1.16, "learning_rate": 3.519630380244116e-06, "loss": 0.521, "step": 9706 }, { "epoch": 1.16, "learning_rate": 3.5193368967981434e-06, "loss": 0.5852, "step": 9707 }, { "epoch": 1.16, "learning_rate": 3.5190433965024464e-06, "loss": 0.7972, "step": 9708 }, { "epoch": 1.16, "learning_rate": 3.518749879361876e-06, "loss": 0.8918, "step": 9709 }, { "epoch": 1.16, "learning_rate": 3.518456345381284e-06, "loss": 0.4651, "step": 9710 }, { "epoch": 1.16, "learning_rate": 3.518162794565524e-06, "loss": 0.5718, "step": 9711 }, { "epoch": 1.16, "learning_rate": 3.5178692269194477e-06, "loss": 0.4955, "step": 9712 }, { "epoch": 1.16, "learning_rate": 3.5175756424479067e-06, "loss": 0.5575, "step": 9713 }, { "epoch": 1.16, "learning_rate": 3.5172820411557564e-06, "loss": 0.6617, "step": 9714 }, { "epoch": 1.16, "learning_rate": 3.516988423047847e-06, "loss": 0.769, "step": 9715 }, { "epoch": 1.16, "learning_rate": 3.516694788129035e-06, "loss": 0.4644, "step": 9716 }, { "epoch": 1.16, "learning_rate": 3.516401136404172e-06, "loss": 0.588, "step": 9717 }, { "epoch": 1.16, "learning_rate": 3.5161074678781137e-06, "loss": 0.5215, "step": 9718 }, { "epoch": 1.16, "learning_rate": 3.515813782555713e-06, "loss": 0.6595, "step": 9719 }, { "epoch": 1.16, "learning_rate": 3.5155200804418256e-06, "loss": 0.7112, "step": 9720 }, { "epoch": 1.16, "learning_rate": 3.5152263615413058e-06, "loss": 0.8282, "step": 9721 }, { "epoch": 1.16, "learning_rate": 3.5149326258590092e-06, "loss": 0.7566, "step": 9722 }, { "epoch": 1.16, "learning_rate": 3.5146388733997916e-06, "loss": 0.4757, "step": 9723 }, { "epoch": 1.16, "learning_rate": 3.514345104168507e-06, "loss": 0.458, "step": 9724 }, { "epoch": 1.16, "learning_rate": 3.514051318170014e-06, "loss": 0.6799, "step": 9725 }, { "epoch": 1.16, "learning_rate": 3.5137575154091676e-06, "loss": 0.642, "step": 9726 }, { "epoch": 1.16, "learning_rate": 3.513463695890824e-06, "loss": 0.7492, "step": 9727 }, { "epoch": 1.16, "learning_rate": 3.5131698596198398e-06, "loss": 0.9949, "step": 9728 }, { "epoch": 1.16, "learning_rate": 3.5128760066010743e-06, "loss": 0.3405, "step": 9729 }, { "epoch": 1.16, "learning_rate": 3.512582136839382e-06, "loss": 0.5531, "step": 9730 }, { "epoch": 1.16, "learning_rate": 3.5122882503396227e-06, "loss": 0.6134, "step": 9731 }, { "epoch": 1.16, "learning_rate": 3.5119943471066532e-06, "loss": 0.6053, "step": 9732 }, { "epoch": 1.16, "learning_rate": 3.5117004271453332e-06, "loss": 0.6913, "step": 9733 }, { "epoch": 1.16, "learning_rate": 3.5114064904605195e-06, "loss": 0.7693, "step": 9734 }, { "epoch": 1.16, "learning_rate": 3.511112537057072e-06, "loss": 0.3938, "step": 9735 }, { "epoch": 1.16, "learning_rate": 3.510818566939849e-06, "loss": 0.6133, "step": 9736 }, { "epoch": 1.16, "learning_rate": 3.5105245801137107e-06, "loss": 0.5353, "step": 9737 }, { "epoch": 1.16, "learning_rate": 3.5102305765835153e-06, "loss": 0.6452, "step": 9738 }, { "epoch": 1.16, "learning_rate": 3.509936556354125e-06, "loss": 0.7581, "step": 9739 }, { "epoch": 1.16, "learning_rate": 3.5096425194303986e-06, "loss": 0.8756, "step": 9740 }, { "epoch": 1.16, "learning_rate": 3.509348465817196e-06, "loss": 0.4439, "step": 9741 }, { "epoch": 1.16, "learning_rate": 3.509054395519379e-06, "loss": 0.4337, "step": 9742 }, { "epoch": 1.16, "learning_rate": 3.508760308541808e-06, "loss": 0.5993, "step": 9743 }, { "epoch": 1.16, "learning_rate": 3.508466204889345e-06, "loss": 0.5403, "step": 9744 }, { "epoch": 1.16, "learning_rate": 3.5081720845668494e-06, "loss": 0.6522, "step": 9745 }, { "epoch": 1.16, "learning_rate": 3.507877947579186e-06, "loss": 0.7687, "step": 9746 }, { "epoch": 1.16, "learning_rate": 3.5075837939312154e-06, "loss": 0.7689, "step": 9747 }, { "epoch": 1.16, "learning_rate": 3.5072896236277997e-06, "loss": 0.422, "step": 9748 }, { "epoch": 1.16, "learning_rate": 3.5069954366738024e-06, "loss": 0.5133, "step": 9749 }, { "epoch": 1.16, "learning_rate": 3.5067012330740857e-06, "loss": 0.6081, "step": 9750 }, { "epoch": 1.16, "learning_rate": 3.506407012833514e-06, "loss": 0.7033, "step": 9751 }, { "epoch": 1.16, "learning_rate": 3.506112775956949e-06, "loss": 0.7582, "step": 9752 }, { "epoch": 1.16, "learning_rate": 3.5058185224492553e-06, "loss": 0.844, "step": 9753 }, { "epoch": 1.16, "learning_rate": 3.5055242523152983e-06, "loss": 0.5921, "step": 9754 }, { "epoch": 1.16, "learning_rate": 3.5052299655599397e-06, "loss": 0.3653, "step": 9755 }, { "epoch": 1.16, "learning_rate": 3.504935662188046e-06, "loss": 0.5284, "step": 9756 }, { "epoch": 1.16, "learning_rate": 3.504641342204481e-06, "loss": 0.6759, "step": 9757 }, { "epoch": 1.16, "learning_rate": 3.504347005614111e-06, "loss": 0.6886, "step": 9758 }, { "epoch": 1.16, "learning_rate": 3.504052652421801e-06, "loss": 0.9344, "step": 9759 }, { "epoch": 1.16, "learning_rate": 3.503758282632415e-06, "loss": 0.5049, "step": 9760 }, { "epoch": 1.16, "learning_rate": 3.5034638962508216e-06, "loss": 0.4856, "step": 9761 }, { "epoch": 1.16, "learning_rate": 3.5031694932818853e-06, "loss": 0.5505, "step": 9762 }, { "epoch": 1.16, "learning_rate": 3.5028750737304732e-06, "loss": 0.5986, "step": 9763 }, { "epoch": 1.16, "learning_rate": 3.5025806376014515e-06, "loss": 0.7243, "step": 9764 }, { "epoch": 1.16, "learning_rate": 3.5022861848996894e-06, "loss": 0.853, "step": 9765 }, { "epoch": 1.16, "learning_rate": 3.501991715630051e-06, "loss": 0.4754, "step": 9766 }, { "epoch": 1.16, "learning_rate": 3.5016972297974056e-06, "loss": 0.4185, "step": 9767 }, { "epoch": 1.16, "learning_rate": 3.5014027274066214e-06, "loss": 0.5723, "step": 9768 }, { "epoch": 1.16, "learning_rate": 3.501108208462566e-06, "loss": 0.5746, "step": 9769 }, { "epoch": 1.16, "learning_rate": 3.5008136729701077e-06, "loss": 0.6633, "step": 9770 }, { "epoch": 1.16, "learning_rate": 3.500519120934115e-06, "loss": 0.9286, "step": 9771 }, { "epoch": 1.16, "learning_rate": 3.5002245523594584e-06, "loss": 0.8211, "step": 9772 }, { "epoch": 1.16, "learning_rate": 3.499929967251005e-06, "loss": 0.442, "step": 9773 }, { "epoch": 1.16, "learning_rate": 3.499635365613626e-06, "loss": 0.5895, "step": 9774 }, { "epoch": 1.16, "learning_rate": 3.49934074745219e-06, "loss": 0.6993, "step": 9775 }, { "epoch": 1.16, "learning_rate": 3.4990461127715685e-06, "loss": 0.6768, "step": 9776 }, { "epoch": 1.16, "learning_rate": 3.49875146157663e-06, "loss": 0.6918, "step": 9777 }, { "epoch": 1.16, "learning_rate": 3.4984567938722465e-06, "loss": 0.8113, "step": 9778 }, { "epoch": 1.16, "learning_rate": 3.4981621096632885e-06, "loss": 0.47, "step": 9779 }, { "epoch": 1.16, "learning_rate": 3.497867408954627e-06, "loss": 0.5415, "step": 9780 }, { "epoch": 1.16, "learning_rate": 3.4975726917511326e-06, "loss": 0.5169, "step": 9781 }, { "epoch": 1.16, "learning_rate": 3.497277958057679e-06, "loss": 0.5739, "step": 9782 }, { "epoch": 1.17, "learning_rate": 3.4969832078791367e-06, "loss": 0.778, "step": 9783 }, { "epoch": 1.17, "learning_rate": 3.4966884412203783e-06, "loss": 0.7824, "step": 9784 }, { "epoch": 1.17, "learning_rate": 3.496393658086276e-06, "loss": 0.4376, "step": 9785 }, { "epoch": 1.17, "learning_rate": 3.496098858481704e-06, "loss": 0.5614, "step": 9786 }, { "epoch": 1.17, "learning_rate": 3.4958040424115336e-06, "loss": 0.5181, "step": 9787 }, { "epoch": 1.17, "learning_rate": 3.4955092098806386e-06, "loss": 0.6205, "step": 9788 }, { "epoch": 1.17, "learning_rate": 3.4952143608938934e-06, "loss": 0.7559, "step": 9789 }, { "epoch": 1.17, "learning_rate": 3.4949194954561703e-06, "loss": 0.7613, "step": 9790 }, { "epoch": 1.17, "learning_rate": 3.4946246135723454e-06, "loss": 0.563, "step": 9791 }, { "epoch": 1.17, "learning_rate": 3.4943297152472917e-06, "loss": 0.3942, "step": 9792 }, { "epoch": 1.17, "learning_rate": 3.4940348004858847e-06, "loss": 0.5492, "step": 9793 }, { "epoch": 1.17, "learning_rate": 3.493739869292999e-06, "loss": 0.6846, "step": 9794 }, { "epoch": 1.17, "learning_rate": 3.49344492167351e-06, "loss": 0.6689, "step": 9795 }, { "epoch": 1.17, "learning_rate": 3.4931499576322926e-06, "loss": 0.8615, "step": 9796 }, { "epoch": 1.17, "learning_rate": 3.4928549771742238e-06, "loss": 0.8277, "step": 9797 }, { "epoch": 1.17, "learning_rate": 3.4925599803041783e-06, "loss": 0.3801, "step": 9798 }, { "epoch": 1.17, "learning_rate": 3.492264967027033e-06, "loss": 0.505, "step": 9799 }, { "epoch": 1.17, "learning_rate": 3.491969937347665e-06, "loss": 0.6373, "step": 9800 }, { "epoch": 1.17, "learning_rate": 3.4916748912709506e-06, "loss": 0.7035, "step": 9801 }, { "epoch": 1.17, "learning_rate": 3.491379828801767e-06, "loss": 0.7988, "step": 9802 }, { "epoch": 1.17, "learning_rate": 3.4910847499449906e-06, "loss": 0.8284, "step": 9803 }, { "epoch": 1.17, "learning_rate": 3.4907896547055014e-06, "loss": 0.5006, "step": 9804 }, { "epoch": 1.17, "learning_rate": 3.490494543088176e-06, "loss": 0.5577, "step": 9805 }, { "epoch": 1.17, "learning_rate": 3.4901994150978926e-06, "loss": 0.5448, "step": 9806 }, { "epoch": 1.17, "learning_rate": 3.489904270739529e-06, "loss": 0.607, "step": 9807 }, { "epoch": 1.17, "learning_rate": 3.4896091100179662e-06, "loss": 0.8661, "step": 9808 }, { "epoch": 1.17, "learning_rate": 3.4893139329380804e-06, "loss": 0.7519, "step": 9809 }, { "epoch": 1.17, "learning_rate": 3.489018739504753e-06, "loss": 0.5154, "step": 9810 }, { "epoch": 1.17, "learning_rate": 3.488723529722863e-06, "loss": 0.6359, "step": 9811 }, { "epoch": 1.17, "learning_rate": 3.4884283035972897e-06, "loss": 0.5685, "step": 9812 }, { "epoch": 1.17, "learning_rate": 3.4881330611329135e-06, "loss": 0.6541, "step": 9813 }, { "epoch": 1.17, "learning_rate": 3.487837802334615e-06, "loss": 0.6945, "step": 9814 }, { "epoch": 1.17, "learning_rate": 3.4875425272072747e-06, "loss": 0.7761, "step": 9815 }, { "epoch": 1.17, "learning_rate": 3.4872472357557742e-06, "loss": 0.5058, "step": 9816 }, { "epoch": 1.17, "learning_rate": 3.4869519279849926e-06, "loss": 0.4885, "step": 9817 }, { "epoch": 1.17, "learning_rate": 3.486656603899814e-06, "loss": 0.5828, "step": 9818 }, { "epoch": 1.17, "learning_rate": 3.486361263505119e-06, "loss": 0.6355, "step": 9819 }, { "epoch": 1.17, "learning_rate": 3.486065906805789e-06, "loss": 0.8288, "step": 9820 }, { "epoch": 1.17, "learning_rate": 3.4857705338067073e-06, "loss": 0.9208, "step": 9821 }, { "epoch": 1.17, "learning_rate": 3.4854751445127556e-06, "loss": 0.8264, "step": 9822 }, { "epoch": 1.17, "learning_rate": 3.485179738928818e-06, "loss": 0.5344, "step": 9823 }, { "epoch": 1.17, "learning_rate": 3.4848843170597747e-06, "loss": 0.4952, "step": 9824 }, { "epoch": 1.17, "learning_rate": 3.4845888789105127e-06, "loss": 0.6012, "step": 9825 }, { "epoch": 1.17, "learning_rate": 3.4842934244859133e-06, "loss": 0.656, "step": 9826 }, { "epoch": 1.17, "learning_rate": 3.483997953790861e-06, "loss": 0.7429, "step": 9827 }, { "epoch": 1.17, "learning_rate": 3.48370246683024e-06, "loss": 0.9512, "step": 9828 }, { "epoch": 1.17, "learning_rate": 3.4834069636089353e-06, "loss": 0.5932, "step": 9829 }, { "epoch": 1.17, "learning_rate": 3.48311144413183e-06, "loss": 0.4369, "step": 9830 }, { "epoch": 1.17, "learning_rate": 3.4828159084038103e-06, "loss": 0.4837, "step": 9831 }, { "epoch": 1.17, "learning_rate": 3.482520356429761e-06, "loss": 0.6013, "step": 9832 }, { "epoch": 1.17, "learning_rate": 3.4822247882145687e-06, "loss": 0.8076, "step": 9833 }, { "epoch": 1.17, "learning_rate": 3.4819292037631176e-06, "loss": 0.9106, "step": 9834 }, { "epoch": 1.17, "learning_rate": 3.481633603080294e-06, "loss": 0.4543, "step": 9835 }, { "epoch": 1.17, "learning_rate": 3.481337986170985e-06, "loss": 0.5303, "step": 9836 }, { "epoch": 1.17, "learning_rate": 3.4810423530400765e-06, "loss": 0.5506, "step": 9837 }, { "epoch": 1.17, "learning_rate": 3.480746703692456e-06, "loss": 0.5831, "step": 9838 }, { "epoch": 1.17, "learning_rate": 3.4804510381330096e-06, "loss": 0.7812, "step": 9839 }, { "epoch": 1.17, "learning_rate": 3.4801553563666263e-06, "loss": 0.9039, "step": 9840 }, { "epoch": 1.17, "learning_rate": 3.4798596583981918e-06, "loss": 0.6048, "step": 9841 }, { "epoch": 1.17, "learning_rate": 3.479563944232595e-06, "loss": 0.4536, "step": 9842 }, { "epoch": 1.17, "learning_rate": 3.4792682138747247e-06, "loss": 0.597, "step": 9843 }, { "epoch": 1.17, "learning_rate": 3.4789724673294677e-06, "loss": 0.6148, "step": 9844 }, { "epoch": 1.17, "learning_rate": 3.4786767046017144e-06, "loss": 0.7369, "step": 9845 }, { "epoch": 1.17, "learning_rate": 3.478380925696353e-06, "loss": 0.8549, "step": 9846 }, { "epoch": 1.17, "learning_rate": 3.478085130618272e-06, "loss": 0.8047, "step": 9847 }, { "epoch": 1.17, "learning_rate": 3.4777893193723628e-06, "loss": 0.4889, "step": 9848 }, { "epoch": 1.17, "learning_rate": 3.4774934919635135e-06, "loss": 0.5269, "step": 9849 }, { "epoch": 1.17, "learning_rate": 3.4771976483966146e-06, "loss": 0.5685, "step": 9850 }, { "epoch": 1.17, "learning_rate": 3.4769017886765577e-06, "loss": 0.6747, "step": 9851 }, { "epoch": 1.17, "learning_rate": 3.476605912808231e-06, "loss": 0.7285, "step": 9852 }, { "epoch": 1.17, "learning_rate": 3.4763100207965273e-06, "loss": 0.85, "step": 9853 }, { "epoch": 1.17, "learning_rate": 3.4760141126463363e-06, "loss": 0.5011, "step": 9854 }, { "epoch": 1.17, "learning_rate": 3.4757181883625512e-06, "loss": 0.6373, "step": 9855 }, { "epoch": 1.17, "learning_rate": 3.4754222479500615e-06, "loss": 0.6149, "step": 9856 }, { "epoch": 1.17, "learning_rate": 3.4751262914137605e-06, "loss": 0.613, "step": 9857 }, { "epoch": 1.17, "learning_rate": 3.4748303187585407e-06, "loss": 0.6676, "step": 9858 }, { "epoch": 1.17, "learning_rate": 3.4745343299892937e-06, "loss": 0.7415, "step": 9859 }, { "epoch": 1.17, "learning_rate": 3.474238325110912e-06, "loss": 0.4315, "step": 9860 }, { "epoch": 1.17, "learning_rate": 3.4739423041282895e-06, "loss": 0.4657, "step": 9861 }, { "epoch": 1.17, "learning_rate": 3.473646267046318e-06, "loss": 0.5753, "step": 9862 }, { "epoch": 1.17, "learning_rate": 3.4733502138698933e-06, "loss": 0.6214, "step": 9863 }, { "epoch": 1.17, "learning_rate": 3.473054144603907e-06, "loss": 0.7885, "step": 9864 }, { "epoch": 1.17, "learning_rate": 3.472758059253255e-06, "loss": 0.8273, "step": 9865 }, { "epoch": 1.17, "learning_rate": 3.4724619578228304e-06, "loss": 0.5365, "step": 9866 }, { "epoch": 1.18, "learning_rate": 3.4721658403175274e-06, "loss": 0.5013, "step": 9867 }, { "epoch": 1.18, "learning_rate": 3.471869706742242e-06, "loss": 0.4825, "step": 9868 }, { "epoch": 1.18, "learning_rate": 3.4715735571018685e-06, "loss": 0.5514, "step": 9869 }, { "epoch": 1.18, "learning_rate": 3.4712773914013026e-06, "loss": 0.6821, "step": 9870 }, { "epoch": 1.18, "learning_rate": 3.4709812096454398e-06, "loss": 0.8484, "step": 9871 }, { "epoch": 1.18, "learning_rate": 3.470685011839177e-06, "loss": 0.9142, "step": 9872 }, { "epoch": 1.18, "learning_rate": 3.4703887979874085e-06, "loss": 0.4837, "step": 9873 }, { "epoch": 1.18, "learning_rate": 3.470092568095032e-06, "loss": 0.5441, "step": 9874 }, { "epoch": 1.18, "learning_rate": 3.469796322166944e-06, "loss": 0.5189, "step": 9875 }, { "epoch": 1.18, "learning_rate": 3.469500060208042e-06, "loss": 0.6397, "step": 9876 }, { "epoch": 1.18, "learning_rate": 3.4692037822232217e-06, "loss": 0.7458, "step": 9877 }, { "epoch": 1.18, "learning_rate": 3.4689074882173825e-06, "loss": 0.745, "step": 9878 }, { "epoch": 1.18, "learning_rate": 3.46861117819542e-06, "loss": 0.4649, "step": 9879 }, { "epoch": 1.18, "learning_rate": 3.4683148521622343e-06, "loss": 0.4689, "step": 9880 }, { "epoch": 1.18, "learning_rate": 3.4680185101227225e-06, "loss": 0.6186, "step": 9881 }, { "epoch": 1.18, "learning_rate": 3.4677221520817837e-06, "loss": 0.6229, "step": 9882 }, { "epoch": 1.18, "learning_rate": 3.4674257780443178e-06, "loss": 0.6986, "step": 9883 }, { "epoch": 1.18, "learning_rate": 3.467129388015221e-06, "loss": 0.8467, "step": 9884 }, { "epoch": 1.18, "learning_rate": 3.4668329819993944e-06, "loss": 0.4724, "step": 9885 }, { "epoch": 1.18, "learning_rate": 3.4665365600017375e-06, "loss": 0.4833, "step": 9886 }, { "epoch": 1.18, "learning_rate": 3.466240122027151e-06, "loss": 0.5161, "step": 9887 }, { "epoch": 1.18, "learning_rate": 3.465943668080533e-06, "loss": 0.6488, "step": 9888 }, { "epoch": 1.18, "learning_rate": 3.465647198166786e-06, "loss": 0.7217, "step": 9889 }, { "epoch": 1.18, "learning_rate": 3.465350712290809e-06, "loss": 0.9363, "step": 9890 }, { "epoch": 1.18, "learning_rate": 3.4650542104575047e-06, "loss": 0.4259, "step": 9891 }, { "epoch": 1.18, "learning_rate": 3.4647576926717726e-06, "loss": 0.4853, "step": 9892 }, { "epoch": 1.18, "learning_rate": 3.464461158938515e-06, "loss": 0.4665, "step": 9893 }, { "epoch": 1.18, "learning_rate": 3.464164609262634e-06, "loss": 0.5144, "step": 9894 }, { "epoch": 1.18, "learning_rate": 3.4638680436490307e-06, "loss": 0.6305, "step": 9895 }, { "epoch": 1.18, "learning_rate": 3.4635714621026072e-06, "loss": 0.7184, "step": 9896 }, { "epoch": 1.18, "learning_rate": 3.4632748646282683e-06, "loss": 0.8107, "step": 9897 }, { "epoch": 1.18, "learning_rate": 3.462978251230914e-06, "loss": 0.3927, "step": 9898 }, { "epoch": 1.18, "learning_rate": 3.462681621915448e-06, "loss": 0.4122, "step": 9899 }, { "epoch": 1.18, "learning_rate": 3.4623849766867745e-06, "loss": 0.5774, "step": 9900 }, { "epoch": 1.18, "learning_rate": 3.4620883155497963e-06, "loss": 0.5597, "step": 9901 }, { "epoch": 1.18, "learning_rate": 3.461791638509418e-06, "loss": 0.7883, "step": 9902 }, { "epoch": 1.18, "learning_rate": 3.4614949455705428e-06, "loss": 0.79, "step": 9903 }, { "epoch": 1.18, "learning_rate": 3.4611982367380752e-06, "loss": 0.4984, "step": 9904 }, { "epoch": 1.18, "learning_rate": 3.460901512016921e-06, "loss": 0.5076, "step": 9905 }, { "epoch": 1.18, "learning_rate": 3.4606047714119833e-06, "loss": 0.5656, "step": 9906 }, { "epoch": 1.18, "learning_rate": 3.4603080149281683e-06, "loss": 0.6698, "step": 9907 }, { "epoch": 1.18, "learning_rate": 3.460011242570382e-06, "loss": 0.7613, "step": 9908 }, { "epoch": 1.18, "learning_rate": 3.4597144543435284e-06, "loss": 0.7456, "step": 9909 }, { "epoch": 1.18, "learning_rate": 3.4594176502525146e-06, "loss": 0.5595, "step": 9910 }, { "epoch": 1.18, "learning_rate": 3.459120830302247e-06, "loss": 0.4196, "step": 9911 }, { "epoch": 1.18, "learning_rate": 3.458823994497631e-06, "loss": 0.5157, "step": 9912 }, { "epoch": 1.18, "learning_rate": 3.458527142843574e-06, "loss": 0.6314, "step": 9913 }, { "epoch": 1.18, "learning_rate": 3.458230275344983e-06, "loss": 0.7266, "step": 9914 }, { "epoch": 1.18, "learning_rate": 3.457933392006766e-06, "loss": 0.8918, "step": 9915 }, { "epoch": 1.18, "learning_rate": 3.457636492833828e-06, "loss": 0.582, "step": 9916 }, { "epoch": 1.18, "learning_rate": 3.4573395778310797e-06, "loss": 0.4986, "step": 9917 }, { "epoch": 1.18, "learning_rate": 3.457042647003427e-06, "loss": 0.6216, "step": 9918 }, { "epoch": 1.18, "learning_rate": 3.45674570035578e-06, "loss": 0.5447, "step": 9919 }, { "epoch": 1.18, "learning_rate": 3.4564487378930456e-06, "loss": 0.7759, "step": 9920 }, { "epoch": 1.18, "learning_rate": 3.4561517596201332e-06, "loss": 0.7662, "step": 9921 }, { "epoch": 1.18, "learning_rate": 3.455854765541953e-06, "loss": 0.813, "step": 9922 }, { "epoch": 1.18, "learning_rate": 3.4555577556634125e-06, "loss": 0.4164, "step": 9923 }, { "epoch": 1.18, "learning_rate": 3.4552607299894212e-06, "loss": 0.6569, "step": 9924 }, { "epoch": 1.18, "learning_rate": 3.454963688524891e-06, "loss": 0.6411, "step": 9925 }, { "epoch": 1.18, "learning_rate": 3.4546666312747312e-06, "loss": 0.5655, "step": 9926 }, { "epoch": 1.18, "learning_rate": 3.4543695582438513e-06, "loss": 0.6632, "step": 9927 }, { "epoch": 1.18, "learning_rate": 3.4540724694371622e-06, "loss": 0.8432, "step": 9928 }, { "epoch": 1.18, "learning_rate": 3.453775364859576e-06, "loss": 0.4472, "step": 9929 }, { "epoch": 1.18, "learning_rate": 3.453478244516003e-06, "loss": 0.4866, "step": 9930 }, { "epoch": 1.18, "learning_rate": 3.453181108411354e-06, "loss": 0.5546, "step": 9931 }, { "epoch": 1.18, "learning_rate": 3.452883956550541e-06, "loss": 0.5296, "step": 9932 }, { "epoch": 1.18, "learning_rate": 3.4525867889384775e-06, "loss": 0.7018, "step": 9933 }, { "epoch": 1.18, "learning_rate": 3.452289605580074e-06, "loss": 0.9292, "step": 9934 }, { "epoch": 1.18, "learning_rate": 3.451992406480242e-06, "loss": 0.5669, "step": 9935 }, { "epoch": 1.18, "learning_rate": 3.4516951916438974e-06, "loss": 0.4376, "step": 9936 }, { "epoch": 1.18, "learning_rate": 3.4513979610759507e-06, "loss": 0.5436, "step": 9937 }, { "epoch": 1.18, "learning_rate": 3.4511007147813156e-06, "loss": 0.5705, "step": 9938 }, { "epoch": 1.18, "learning_rate": 3.4508034527649066e-06, "loss": 0.7357, "step": 9939 }, { "epoch": 1.18, "learning_rate": 3.4505061750316366e-06, "loss": 0.7736, "step": 9940 }, { "epoch": 1.18, "learning_rate": 3.4502088815864195e-06, "loss": 0.5891, "step": 9941 }, { "epoch": 1.18, "learning_rate": 3.44991157243417e-06, "loss": 0.4876, "step": 9942 }, { "epoch": 1.18, "learning_rate": 3.449614247579802e-06, "loss": 0.6436, "step": 9943 }, { "epoch": 1.18, "learning_rate": 3.449316907028232e-06, "loss": 0.5895, "step": 9944 }, { "epoch": 1.18, "learning_rate": 3.4490195507843727e-06, "loss": 0.6809, "step": 9945 }, { "epoch": 1.18, "learning_rate": 3.4487221788531404e-06, "loss": 0.9119, "step": 9946 }, { "epoch": 1.18, "learning_rate": 3.448424791239452e-06, "loss": 0.7847, "step": 9947 }, { "epoch": 1.18, "learning_rate": 3.448127387948222e-06, "loss": 0.5182, "step": 9948 }, { "epoch": 1.18, "learning_rate": 3.4478299689843663e-06, "loss": 0.6022, "step": 9949 }, { "epoch": 1.18, "learning_rate": 3.447532534352801e-06, "loss": 0.5657, "step": 9950 }, { "epoch": 1.19, "learning_rate": 3.4472350840584453e-06, "loss": 0.6113, "step": 9951 }, { "epoch": 1.19, "learning_rate": 3.4469376181062127e-06, "loss": 0.7313, "step": 9952 }, { "epoch": 1.19, "learning_rate": 3.446640136501022e-06, "loss": 0.9338, "step": 9953 }, { "epoch": 1.19, "learning_rate": 3.4463426392477907e-06, "loss": 0.5467, "step": 9954 }, { "epoch": 1.19, "learning_rate": 3.4460451263514362e-06, "loss": 0.5745, "step": 9955 }, { "epoch": 1.19, "learning_rate": 3.4457475978168757e-06, "loss": 0.5482, "step": 9956 }, { "epoch": 1.19, "learning_rate": 3.4454500536490287e-06, "loss": 0.6448, "step": 9957 }, { "epoch": 1.19, "learning_rate": 3.4451524938528135e-06, "loss": 0.6273, "step": 9958 }, { "epoch": 1.19, "learning_rate": 3.4448549184331476e-06, "loss": 0.8309, "step": 9959 }, { "epoch": 1.19, "learning_rate": 3.4445573273949506e-06, "loss": 0.4787, "step": 9960 }, { "epoch": 1.19, "learning_rate": 3.4442597207431416e-06, "loss": 0.5408, "step": 9961 }, { "epoch": 1.19, "learning_rate": 3.443962098482641e-06, "loss": 0.6817, "step": 9962 }, { "epoch": 1.19, "learning_rate": 3.4436644606183667e-06, "loss": 0.6516, "step": 9963 }, { "epoch": 1.19, "learning_rate": 3.4433668071552407e-06, "loss": 0.7955, "step": 9964 }, { "epoch": 1.19, "learning_rate": 3.443069138098181e-06, "loss": 0.8929, "step": 9965 }, { "epoch": 1.19, "learning_rate": 3.4427714534521106e-06, "loss": 0.4674, "step": 9966 }, { "epoch": 1.19, "learning_rate": 3.4424737532219477e-06, "loss": 0.4851, "step": 9967 }, { "epoch": 1.19, "learning_rate": 3.4421760374126152e-06, "loss": 0.5684, "step": 9968 }, { "epoch": 1.19, "learning_rate": 3.441878306029034e-06, "loss": 0.5705, "step": 9969 }, { "epoch": 1.19, "learning_rate": 3.4415805590761255e-06, "loss": 0.7823, "step": 9970 }, { "epoch": 1.19, "learning_rate": 3.4412827965588103e-06, "loss": 0.8333, "step": 9971 }, { "epoch": 1.19, "learning_rate": 3.440985018482013e-06, "loss": 0.8046, "step": 9972 }, { "epoch": 1.19, "learning_rate": 3.440687224850653e-06, "loss": 0.459, "step": 9973 }, { "epoch": 1.19, "learning_rate": 3.440389415669655e-06, "loss": 0.564, "step": 9974 }, { "epoch": 1.19, "learning_rate": 3.4400915909439404e-06, "loss": 0.5358, "step": 9975 }, { "epoch": 1.19, "learning_rate": 3.4397937506784336e-06, "loss": 0.6206, "step": 9976 }, { "epoch": 1.19, "learning_rate": 3.4394958948780572e-06, "loss": 0.7382, "step": 9977 }, { "epoch": 1.19, "learning_rate": 3.439198023547734e-06, "loss": 0.8227, "step": 9978 }, { "epoch": 1.19, "learning_rate": 3.4389001366923896e-06, "loss": 0.3608, "step": 9979 }, { "epoch": 1.19, "learning_rate": 3.438602234316947e-06, "loss": 0.5008, "step": 9980 }, { "epoch": 1.19, "learning_rate": 3.438304316426331e-06, "loss": 0.6025, "step": 9981 }, { "epoch": 1.19, "learning_rate": 3.4380063830254647e-06, "loss": 0.6342, "step": 9982 }, { "epoch": 1.19, "learning_rate": 3.437708434119276e-06, "loss": 0.7682, "step": 9983 }, { "epoch": 1.19, "learning_rate": 3.4374104697126864e-06, "loss": 0.9807, "step": 9984 }, { "epoch": 1.19, "learning_rate": 3.437112489810624e-06, "loss": 0.4363, "step": 9985 }, { "epoch": 1.19, "learning_rate": 3.436814494418014e-06, "loss": 0.5223, "step": 9986 }, { "epoch": 1.19, "learning_rate": 3.436516483539781e-06, "loss": 0.4799, "step": 9987 }, { "epoch": 1.19, "learning_rate": 3.4362184571808514e-06, "loss": 0.5868, "step": 9988 }, { "epoch": 1.19, "learning_rate": 3.4359204153461528e-06, "loss": 0.6973, "step": 9989 }, { "epoch": 1.19, "learning_rate": 3.4356223580406116e-06, "loss": 0.9097, "step": 9990 }, { "epoch": 1.19, "learning_rate": 3.435324285269154e-06, "loss": 0.575, "step": 9991 }, { "epoch": 1.19, "learning_rate": 3.4350261970367072e-06, "loss": 0.507, "step": 9992 }, { "epoch": 1.19, "learning_rate": 3.4347280933481992e-06, "loss": 0.6068, "step": 9993 }, { "epoch": 1.19, "learning_rate": 3.434429974208558e-06, "loss": 0.5912, "step": 9994 }, { "epoch": 1.19, "learning_rate": 3.4341318396227097e-06, "loss": 0.7764, "step": 9995 }, { "epoch": 1.19, "learning_rate": 3.4338336895955846e-06, "loss": 0.7867, "step": 9996 }, { "epoch": 1.19, "learning_rate": 3.43353552413211e-06, "loss": 0.9821, "step": 9997 }, { "epoch": 1.19, "learning_rate": 3.433237343237215e-06, "loss": 0.4034, "step": 9998 }, { "epoch": 1.19, "learning_rate": 3.432939146915828e-06, "loss": 0.5957, "step": 9999 }, { "epoch": 1.19, "learning_rate": 3.4326409351728785e-06, "loss": 0.6091, "step": 10000 }, { "epoch": 1.19, "learning_rate": 3.4323427080132964e-06, "loss": 0.6507, "step": 10001 }, { "epoch": 1.19, "learning_rate": 3.4320444654420116e-06, "loss": 0.7783, "step": 10002 }, { "epoch": 1.19, "learning_rate": 3.431746207463953e-06, "loss": 0.83, "step": 10003 }, { "epoch": 1.19, "learning_rate": 3.431447934084051e-06, "loss": 0.5158, "step": 10004 }, { "epoch": 1.19, "learning_rate": 3.4311496453072375e-06, "loss": 0.5636, "step": 10005 }, { "epoch": 1.19, "learning_rate": 3.430851341138442e-06, "loss": 0.5756, "step": 10006 }, { "epoch": 1.19, "learning_rate": 3.430553021582595e-06, "loss": 0.5851, "step": 10007 }, { "epoch": 1.19, "learning_rate": 3.4302546866446295e-06, "loss": 0.726, "step": 10008 }, { "epoch": 1.19, "learning_rate": 3.4299563363294753e-06, "loss": 0.8472, "step": 10009 }, { "epoch": 1.19, "learning_rate": 3.4296579706420647e-06, "loss": 0.5811, "step": 10010 }, { "epoch": 1.19, "learning_rate": 3.42935958958733e-06, "loss": 0.4367, "step": 10011 }, { "epoch": 1.19, "learning_rate": 3.4290611931702037e-06, "loss": 0.622, "step": 10012 }, { "epoch": 1.19, "learning_rate": 3.4287627813956183e-06, "loss": 0.5276, "step": 10013 }, { "epoch": 1.19, "learning_rate": 3.4284643542685047e-06, "loss": 0.7551, "step": 10014 }, { "epoch": 1.19, "learning_rate": 3.428165911793799e-06, "loss": 0.7884, "step": 10015 }, { "epoch": 1.19, "learning_rate": 3.427867453976431e-06, "loss": 0.6183, "step": 10016 }, { "epoch": 1.19, "learning_rate": 3.427568980821338e-06, "loss": 0.5023, "step": 10017 }, { "epoch": 1.19, "learning_rate": 3.427270492333451e-06, "loss": 0.4805, "step": 10018 }, { "epoch": 1.19, "learning_rate": 3.4269719885177057e-06, "loss": 0.6417, "step": 10019 }, { "epoch": 1.19, "learning_rate": 3.4266734693790345e-06, "loss": 0.7855, "step": 10020 }, { "epoch": 1.19, "learning_rate": 3.4263749349223743e-06, "loss": 0.7665, "step": 10021 }, { "epoch": 1.19, "learning_rate": 3.426076385152658e-06, "loss": 0.809, "step": 10022 }, { "epoch": 1.19, "learning_rate": 3.425777820074821e-06, "loss": 0.4496, "step": 10023 }, { "epoch": 1.19, "learning_rate": 3.425479239693799e-06, "loss": 0.5549, "step": 10024 }, { "epoch": 1.19, "learning_rate": 3.425180644014528e-06, "loss": 0.622, "step": 10025 }, { "epoch": 1.19, "learning_rate": 3.424882033041944e-06, "loss": 0.6852, "step": 10026 }, { "epoch": 1.19, "learning_rate": 3.4245834067809813e-06, "loss": 0.8505, "step": 10027 }, { "epoch": 1.19, "learning_rate": 3.4242847652365772e-06, "loss": 0.8686, "step": 10028 }, { "epoch": 1.19, "learning_rate": 3.4239861084136695e-06, "loss": 0.4399, "step": 10029 }, { "epoch": 1.19, "learning_rate": 3.4236874363171936e-06, "loss": 0.578, "step": 10030 }, { "epoch": 1.19, "learning_rate": 3.4233887489520857e-06, "loss": 0.5813, "step": 10031 }, { "epoch": 1.19, "learning_rate": 3.4230900463232854e-06, "loss": 0.6642, "step": 10032 }, { "epoch": 1.19, "learning_rate": 3.4227913284357285e-06, "loss": 0.8365, "step": 10033 }, { "epoch": 1.19, "learning_rate": 3.4224925952943543e-06, "loss": 0.8895, "step": 10034 }, { "epoch": 1.2, "learning_rate": 3.422193846904099e-06, "loss": 0.2953, "step": 10035 }, { "epoch": 1.2, "learning_rate": 3.4218950832699027e-06, "loss": 0.4405, "step": 10036 }, { "epoch": 1.2, "learning_rate": 3.4215963043967038e-06, "loss": 0.6032, "step": 10037 }, { "epoch": 1.2, "learning_rate": 3.4212975102894403e-06, "loss": 0.664, "step": 10038 }, { "epoch": 1.2, "learning_rate": 3.420998700953051e-06, "loss": 0.6594, "step": 10039 }, { "epoch": 1.2, "learning_rate": 3.420699876392477e-06, "loss": 0.7667, "step": 10040 }, { "epoch": 1.2, "learning_rate": 3.4204010366126562e-06, "loss": 0.5422, "step": 10041 }, { "epoch": 1.2, "learning_rate": 3.4201021816185287e-06, "loss": 0.4607, "step": 10042 }, { "epoch": 1.2, "learning_rate": 3.4198033114150352e-06, "loss": 0.5051, "step": 10043 }, { "epoch": 1.2, "learning_rate": 3.4195044260071166e-06, "loss": 0.6373, "step": 10044 }, { "epoch": 1.2, "learning_rate": 3.419205525399712e-06, "loss": 0.7024, "step": 10045 }, { "epoch": 1.2, "learning_rate": 3.4189066095977623e-06, "loss": 0.8112, "step": 10046 }, { "epoch": 1.2, "learning_rate": 3.41860767860621e-06, "loss": 0.8634, "step": 10047 }, { "epoch": 1.2, "learning_rate": 3.4183087324299957e-06, "loss": 0.4621, "step": 10048 }, { "epoch": 1.2, "learning_rate": 3.4180097710740604e-06, "loss": 0.6213, "step": 10049 }, { "epoch": 1.2, "learning_rate": 3.417710794543347e-06, "loss": 0.5608, "step": 10050 }, { "epoch": 1.2, "learning_rate": 3.4174118028427977e-06, "loss": 0.6072, "step": 10051 }, { "epoch": 1.2, "learning_rate": 3.417112795977353e-06, "loss": 0.8038, "step": 10052 }, { "epoch": 1.2, "learning_rate": 3.416813773951958e-06, "loss": 0.7343, "step": 10053 }, { "epoch": 1.2, "learning_rate": 3.4165147367715546e-06, "loss": 0.4974, "step": 10054 }, { "epoch": 1.2, "learning_rate": 3.416215684441085e-06, "loss": 0.5578, "step": 10055 }, { "epoch": 1.2, "learning_rate": 3.415916616965493e-06, "loss": 0.5157, "step": 10056 }, { "epoch": 1.2, "learning_rate": 3.4156175343497228e-06, "loss": 0.6255, "step": 10057 }, { "epoch": 1.2, "learning_rate": 3.415318436598718e-06, "loss": 0.796, "step": 10058 }, { "epoch": 1.2, "learning_rate": 3.4150193237174227e-06, "loss": 0.8308, "step": 10059 }, { "epoch": 1.2, "learning_rate": 3.414720195710781e-06, "loss": 0.3224, "step": 10060 }, { "epoch": 1.2, "learning_rate": 3.4144210525837374e-06, "loss": 0.4408, "step": 10061 }, { "epoch": 1.2, "learning_rate": 3.414121894341238e-06, "loss": 0.67, "step": 10062 }, { "epoch": 1.2, "learning_rate": 3.4138227209882262e-06, "loss": 0.5731, "step": 10063 }, { "epoch": 1.2, "learning_rate": 3.413523532529649e-06, "loss": 0.6665, "step": 10064 }, { "epoch": 1.2, "learning_rate": 3.4132243289704502e-06, "loss": 0.8362, "step": 10065 }, { "epoch": 1.2, "learning_rate": 3.4129251103155767e-06, "loss": 0.5871, "step": 10066 }, { "epoch": 1.2, "learning_rate": 3.412625876569975e-06, "loss": 0.4668, "step": 10067 }, { "epoch": 1.2, "learning_rate": 3.4123266277385908e-06, "loss": 0.6149, "step": 10068 }, { "epoch": 1.2, "learning_rate": 3.4120273638263712e-06, "loss": 0.649, "step": 10069 }, { "epoch": 1.2, "learning_rate": 3.411728084838262e-06, "loss": 0.588, "step": 10070 }, { "epoch": 1.2, "learning_rate": 3.4114287907792115e-06, "loss": 0.7357, "step": 10071 }, { "epoch": 1.2, "learning_rate": 3.4111294816541673e-06, "loss": 0.8718, "step": 10072 }, { "epoch": 1.2, "learning_rate": 3.410830157468076e-06, "loss": 0.4521, "step": 10073 }, { "epoch": 1.2, "learning_rate": 3.4105308182258846e-06, "loss": 0.5286, "step": 10074 }, { "epoch": 1.2, "learning_rate": 3.4102314639325433e-06, "loss": 0.6052, "step": 10075 }, { "epoch": 1.2, "learning_rate": 3.409932094592999e-06, "loss": 0.6942, "step": 10076 }, { "epoch": 1.2, "learning_rate": 3.409632710212202e-06, "loss": 0.7321, "step": 10077 }, { "epoch": 1.2, "learning_rate": 3.409333310795099e-06, "loss": 0.8632, "step": 10078 }, { "epoch": 1.2, "learning_rate": 3.40903389634664e-06, "loss": 0.3906, "step": 10079 }, { "epoch": 1.2, "learning_rate": 3.4087344668717747e-06, "loss": 0.562, "step": 10080 }, { "epoch": 1.2, "learning_rate": 3.4084350223754525e-06, "loss": 0.578, "step": 10081 }, { "epoch": 1.2, "learning_rate": 3.4081355628626227e-06, "loss": 0.6467, "step": 10082 }, { "epoch": 1.2, "learning_rate": 3.407836088338237e-06, "loss": 0.8588, "step": 10083 }, { "epoch": 1.2, "learning_rate": 3.4075365988072435e-06, "loss": 0.8767, "step": 10084 }, { "epoch": 1.2, "learning_rate": 3.407237094274594e-06, "loss": 0.5641, "step": 10085 }, { "epoch": 1.2, "learning_rate": 3.40693757474524e-06, "loss": 0.5745, "step": 10086 }, { "epoch": 1.2, "learning_rate": 3.4066380402241313e-06, "loss": 0.4842, "step": 10087 }, { "epoch": 1.2, "learning_rate": 3.406338490716219e-06, "loss": 0.4483, "step": 10088 }, { "epoch": 1.2, "learning_rate": 3.406038926226456e-06, "loss": 0.662, "step": 10089 }, { "epoch": 1.2, "learning_rate": 3.4057393467597937e-06, "loss": 0.7701, "step": 10090 }, { "epoch": 1.2, "learning_rate": 3.4054397523211837e-06, "loss": 0.569, "step": 10091 }, { "epoch": 1.2, "learning_rate": 3.405140142915579e-06, "loss": 0.505, "step": 10092 }, { "epoch": 1.2, "learning_rate": 3.404840518547931e-06, "loss": 0.5889, "step": 10093 }, { "epoch": 1.2, "learning_rate": 3.404540879223195e-06, "loss": 0.5781, "step": 10094 }, { "epoch": 1.2, "learning_rate": 3.4042412249463204e-06, "loss": 0.6131, "step": 10095 }, { "epoch": 1.2, "learning_rate": 3.403941555722263e-06, "loss": 0.9105, "step": 10096 }, { "epoch": 1.2, "learning_rate": 3.4036418715559766e-06, "loss": 0.9573, "step": 10097 }, { "epoch": 1.2, "learning_rate": 3.403342172452414e-06, "loss": 0.4232, "step": 10098 }, { "epoch": 1.2, "learning_rate": 3.403042458416529e-06, "loss": 0.5627, "step": 10099 }, { "epoch": 1.2, "learning_rate": 3.4027427294532765e-06, "loss": 0.6157, "step": 10100 }, { "epoch": 1.2, "learning_rate": 3.402442985567611e-06, "loss": 0.7148, "step": 10101 }, { "epoch": 1.2, "learning_rate": 3.4021432267644878e-06, "loss": 0.7709, "step": 10102 }, { "epoch": 1.2, "learning_rate": 3.4018434530488604e-06, "loss": 0.8224, "step": 10103 }, { "epoch": 1.2, "learning_rate": 3.4015436644256857e-06, "loss": 0.4049, "step": 10104 }, { "epoch": 1.2, "learning_rate": 3.4012438608999187e-06, "loss": 0.6142, "step": 10105 }, { "epoch": 1.2, "learning_rate": 3.4009440424765143e-06, "loss": 0.5582, "step": 10106 }, { "epoch": 1.2, "learning_rate": 3.4006442091604304e-06, "loss": 0.673, "step": 10107 }, { "epoch": 1.2, "learning_rate": 3.400344360956621e-06, "loss": 0.7869, "step": 10108 }, { "epoch": 1.2, "learning_rate": 3.4000444978700447e-06, "loss": 0.8342, "step": 10109 }, { "epoch": 1.2, "learning_rate": 3.3997446199056565e-06, "loss": 0.4845, "step": 10110 }, { "epoch": 1.2, "learning_rate": 3.399444727068415e-06, "loss": 0.5398, "step": 10111 }, { "epoch": 1.2, "learning_rate": 3.3991448193632757e-06, "loss": 0.5626, "step": 10112 }, { "epoch": 1.2, "learning_rate": 3.398844896795198e-06, "loss": 0.5421, "step": 10113 }, { "epoch": 1.2, "learning_rate": 3.398544959369138e-06, "loss": 0.7617, "step": 10114 }, { "epoch": 1.2, "learning_rate": 3.398245007090055e-06, "loss": 0.7437, "step": 10115 }, { "epoch": 1.2, "learning_rate": 3.397945039962907e-06, "loss": 0.5812, "step": 10116 }, { "epoch": 1.2, "learning_rate": 3.3976450579926516e-06, "loss": 0.5563, "step": 10117 }, { "epoch": 1.2, "learning_rate": 3.397345061184248e-06, "loss": 0.4924, "step": 10118 }, { "epoch": 1.21, "learning_rate": 3.397045049542656e-06, "loss": 0.6033, "step": 10119 }, { "epoch": 1.21, "learning_rate": 3.396745023072834e-06, "loss": 0.7747, "step": 10120 }, { "epoch": 1.21, "learning_rate": 3.39644498177974e-06, "loss": 0.784, "step": 10121 }, { "epoch": 1.21, "learning_rate": 3.3961449256683367e-06, "loss": 0.7586, "step": 10122 }, { "epoch": 1.21, "learning_rate": 3.395844854743582e-06, "loss": 0.4042, "step": 10123 }, { "epoch": 1.21, "learning_rate": 3.395544769010437e-06, "loss": 0.6014, "step": 10124 }, { "epoch": 1.21, "learning_rate": 3.3952446684738615e-06, "loss": 0.5574, "step": 10125 }, { "epoch": 1.21, "learning_rate": 3.394944553138817e-06, "loss": 0.6447, "step": 10126 }, { "epoch": 1.21, "learning_rate": 3.394644423010263e-06, "loss": 0.6898, "step": 10127 }, { "epoch": 1.21, "learning_rate": 3.3943442780931623e-06, "loss": 0.8118, "step": 10128 }, { "epoch": 1.21, "learning_rate": 3.394044118392476e-06, "loss": 0.4713, "step": 10129 }, { "epoch": 1.21, "learning_rate": 3.3937439439131647e-06, "loss": 0.616, "step": 10130 }, { "epoch": 1.21, "learning_rate": 3.393443754660191e-06, "loss": 0.5474, "step": 10131 }, { "epoch": 1.21, "learning_rate": 3.3931435506385175e-06, "loss": 0.5906, "step": 10132 }, { "epoch": 1.21, "learning_rate": 3.3928433318531063e-06, "loss": 0.7466, "step": 10133 }, { "epoch": 1.21, "learning_rate": 3.3925430983089197e-06, "loss": 0.8401, "step": 10134 }, { "epoch": 1.21, "learning_rate": 3.3922428500109205e-06, "loss": 0.457, "step": 10135 }, { "epoch": 1.21, "learning_rate": 3.391942586964072e-06, "loss": 0.4364, "step": 10136 }, { "epoch": 1.21, "learning_rate": 3.3916423091733384e-06, "loss": 0.5684, "step": 10137 }, { "epoch": 1.21, "learning_rate": 3.3913420166436823e-06, "loss": 0.6183, "step": 10138 }, { "epoch": 1.21, "learning_rate": 3.391041709380068e-06, "loss": 0.7733, "step": 10139 }, { "epoch": 1.21, "learning_rate": 3.3907413873874594e-06, "loss": 0.7876, "step": 10140 }, { "epoch": 1.21, "learning_rate": 3.3904410506708207e-06, "loss": 0.576, "step": 10141 }, { "epoch": 1.21, "learning_rate": 3.3901406992351167e-06, "loss": 0.474, "step": 10142 }, { "epoch": 1.21, "learning_rate": 3.3898403330853125e-06, "loss": 0.5884, "step": 10143 }, { "epoch": 1.21, "learning_rate": 3.389539952226373e-06, "loss": 0.6039, "step": 10144 }, { "epoch": 1.21, "learning_rate": 3.389239556663263e-06, "loss": 0.747, "step": 10145 }, { "epoch": 1.21, "learning_rate": 3.3889391464009487e-06, "loss": 0.7939, "step": 10146 }, { "epoch": 1.21, "learning_rate": 3.388638721444396e-06, "loss": 0.9853, "step": 10147 }, { "epoch": 1.21, "learning_rate": 3.3883382817985705e-06, "loss": 0.4974, "step": 10148 }, { "epoch": 1.21, "learning_rate": 3.388037827468439e-06, "loss": 0.5949, "step": 10149 }, { "epoch": 1.21, "learning_rate": 3.387737358458967e-06, "loss": 0.6658, "step": 10150 }, { "epoch": 1.21, "learning_rate": 3.387436874775123e-06, "loss": 0.6241, "step": 10151 }, { "epoch": 1.21, "learning_rate": 3.3871363764218723e-06, "loss": 0.7805, "step": 10152 }, { "epoch": 1.21, "learning_rate": 3.386835863404182e-06, "loss": 0.8627, "step": 10153 }, { "epoch": 1.21, "learning_rate": 3.3865353357270216e-06, "loss": 0.459, "step": 10154 }, { "epoch": 1.21, "learning_rate": 3.3862347933953575e-06, "loss": 0.4686, "step": 10155 }, { "epoch": 1.21, "learning_rate": 3.3859342364141586e-06, "loss": 0.5846, "step": 10156 }, { "epoch": 1.21, "learning_rate": 3.3856336647883907e-06, "loss": 0.5588, "step": 10157 }, { "epoch": 1.21, "learning_rate": 3.3853330785230254e-06, "loss": 0.7621, "step": 10158 }, { "epoch": 1.21, "learning_rate": 3.3850324776230297e-06, "loss": 0.9273, "step": 10159 }, { "epoch": 1.21, "learning_rate": 3.384731862093373e-06, "loss": 0.5242, "step": 10160 }, { "epoch": 1.21, "learning_rate": 3.3844312319390237e-06, "loss": 0.4317, "step": 10161 }, { "epoch": 1.21, "learning_rate": 3.3841305871649533e-06, "loss": 0.6766, "step": 10162 }, { "epoch": 1.21, "learning_rate": 3.383829927776129e-06, "loss": 0.6359, "step": 10163 }, { "epoch": 1.21, "learning_rate": 3.383529253777522e-06, "loss": 0.767, "step": 10164 }, { "epoch": 1.21, "learning_rate": 3.3832285651741028e-06, "loss": 0.8278, "step": 10165 }, { "epoch": 1.21, "learning_rate": 3.3829278619708404e-06, "loss": 0.4824, "step": 10166 }, { "epoch": 1.21, "learning_rate": 3.3826271441727064e-06, "loss": 0.4846, "step": 10167 }, { "epoch": 1.21, "learning_rate": 3.3823264117846722e-06, "loss": 0.6189, "step": 10168 }, { "epoch": 1.21, "learning_rate": 3.382025664811709e-06, "loss": 0.5611, "step": 10169 }, { "epoch": 1.21, "learning_rate": 3.3817249032587868e-06, "loss": 0.7001, "step": 10170 }, { "epoch": 1.21, "learning_rate": 3.381424127130878e-06, "loss": 0.8199, "step": 10171 }, { "epoch": 1.21, "learning_rate": 3.381123336432954e-06, "loss": 0.7602, "step": 10172 }, { "epoch": 1.21, "learning_rate": 3.380822531169988e-06, "loss": 0.4774, "step": 10173 }, { "epoch": 1.21, "learning_rate": 3.380521711346951e-06, "loss": 0.5683, "step": 10174 }, { "epoch": 1.21, "learning_rate": 3.380220876968817e-06, "loss": 0.6252, "step": 10175 }, { "epoch": 1.21, "learning_rate": 3.3799200280405575e-06, "loss": 0.6578, "step": 10176 }, { "epoch": 1.21, "learning_rate": 3.379619164567147e-06, "loss": 0.8893, "step": 10177 }, { "epoch": 1.21, "learning_rate": 3.379318286553557e-06, "loss": 0.8226, "step": 10178 }, { "epoch": 1.21, "learning_rate": 3.3790173940047616e-06, "loss": 0.5203, "step": 10179 }, { "epoch": 1.21, "learning_rate": 3.3787164869257357e-06, "loss": 0.566, "step": 10180 }, { "epoch": 1.21, "learning_rate": 3.378415565321452e-06, "loss": 0.669, "step": 10181 }, { "epoch": 1.21, "learning_rate": 3.378114629196886e-06, "loss": 0.62, "step": 10182 }, { "epoch": 1.21, "learning_rate": 3.3778136785570113e-06, "loss": 0.8384, "step": 10183 }, { "epoch": 1.21, "learning_rate": 3.3775127134068025e-06, "loss": 0.7407, "step": 10184 }, { "epoch": 1.21, "learning_rate": 3.3772117337512345e-06, "loss": 0.6504, "step": 10185 }, { "epoch": 1.21, "learning_rate": 3.3769107395952836e-06, "loss": 0.4525, "step": 10186 }, { "epoch": 1.21, "learning_rate": 3.376609730943925e-06, "loss": 0.591, "step": 10187 }, { "epoch": 1.21, "learning_rate": 3.3763087078021335e-06, "loss": 0.631, "step": 10188 }, { "epoch": 1.21, "learning_rate": 3.3760076701748852e-06, "loss": 0.7159, "step": 10189 }, { "epoch": 1.21, "learning_rate": 3.375706618067157e-06, "loss": 0.8818, "step": 10190 }, { "epoch": 1.21, "learning_rate": 3.3754055514839246e-06, "loss": 0.5663, "step": 10191 }, { "epoch": 1.21, "learning_rate": 3.3751044704301657e-06, "loss": 0.5277, "step": 10192 }, { "epoch": 1.21, "learning_rate": 3.3748033749108556e-06, "loss": 0.5267, "step": 10193 }, { "epoch": 1.21, "learning_rate": 3.3745022649309733e-06, "loss": 0.6728, "step": 10194 }, { "epoch": 1.21, "learning_rate": 3.3742011404954937e-06, "loss": 0.6242, "step": 10195 }, { "epoch": 1.21, "learning_rate": 3.373900001609397e-06, "loss": 0.8702, "step": 10196 }, { "epoch": 1.21, "learning_rate": 3.3735988482776598e-06, "loss": 0.8083, "step": 10197 }, { "epoch": 1.21, "learning_rate": 3.3732976805052607e-06, "loss": 0.4381, "step": 10198 }, { "epoch": 1.21, "learning_rate": 3.3729964982971764e-06, "loss": 0.4825, "step": 10199 }, { "epoch": 1.21, "learning_rate": 3.372695301658388e-06, "loss": 0.4972, "step": 10200 }, { "epoch": 1.21, "learning_rate": 3.3723940905938722e-06, "loss": 0.6451, "step": 10201 }, { "epoch": 1.21, "learning_rate": 3.37209286510861e-06, "loss": 0.8082, "step": 10202 }, { "epoch": 1.22, "learning_rate": 3.371791625207579e-06, "loss": 1.1108, "step": 10203 }, { "epoch": 1.22, "learning_rate": 3.3714903708957588e-06, "loss": 0.4074, "step": 10204 }, { "epoch": 1.22, "learning_rate": 3.3711891021781307e-06, "loss": 0.6155, "step": 10205 }, { "epoch": 1.22, "learning_rate": 3.3708878190596724e-06, "loss": 0.5839, "step": 10206 }, { "epoch": 1.22, "learning_rate": 3.370586521545366e-06, "loss": 0.673, "step": 10207 }, { "epoch": 1.22, "learning_rate": 3.370285209640191e-06, "loss": 0.7516, "step": 10208 }, { "epoch": 1.22, "learning_rate": 3.3699838833491288e-06, "loss": 0.8482, "step": 10209 }, { "epoch": 1.22, "learning_rate": 3.3696825426771597e-06, "loss": 0.3798, "step": 10210 }, { "epoch": 1.22, "learning_rate": 3.3693811876292654e-06, "loss": 0.5143, "step": 10211 }, { "epoch": 1.22, "learning_rate": 3.3690798182104275e-06, "loss": 0.4994, "step": 10212 }, { "epoch": 1.22, "learning_rate": 3.368778434425627e-06, "loss": 0.6098, "step": 10213 }, { "epoch": 1.22, "learning_rate": 3.368477036279846e-06, "loss": 0.8005, "step": 10214 }, { "epoch": 1.22, "learning_rate": 3.3681756237780662e-06, "loss": 0.7754, "step": 10215 }, { "epoch": 1.22, "learning_rate": 3.3678741969252717e-06, "loss": 0.6532, "step": 10216 }, { "epoch": 1.22, "learning_rate": 3.3675727557264427e-06, "loss": 0.4504, "step": 10217 }, { "epoch": 1.22, "learning_rate": 3.367271300186564e-06, "loss": 0.6133, "step": 10218 }, { "epoch": 1.22, "learning_rate": 3.366969830310618e-06, "loss": 0.5191, "step": 10219 }, { "epoch": 1.22, "learning_rate": 3.3666683461035875e-06, "loss": 0.7164, "step": 10220 }, { "epoch": 1.22, "learning_rate": 3.3663668475704566e-06, "loss": 0.9593, "step": 10221 }, { "epoch": 1.22, "learning_rate": 3.3660653347162092e-06, "loss": 0.8683, "step": 10222 }, { "epoch": 1.22, "learning_rate": 3.3657638075458286e-06, "loss": 0.4239, "step": 10223 }, { "epoch": 1.22, "learning_rate": 3.3654622660643003e-06, "loss": 0.5884, "step": 10224 }, { "epoch": 1.22, "learning_rate": 3.3651607102766075e-06, "loss": 0.5047, "step": 10225 }, { "epoch": 1.22, "learning_rate": 3.3648591401877366e-06, "loss": 0.5895, "step": 10226 }, { "epoch": 1.22, "learning_rate": 3.364557555802671e-06, "loss": 0.7464, "step": 10227 }, { "epoch": 1.22, "learning_rate": 3.364255957126396e-06, "loss": 0.7891, "step": 10228 }, { "epoch": 1.22, "learning_rate": 3.363954344163898e-06, "loss": 0.4252, "step": 10229 }, { "epoch": 1.22, "learning_rate": 3.363652716920163e-06, "loss": 0.4084, "step": 10230 }, { "epoch": 1.22, "learning_rate": 3.3633510754001743e-06, "loss": 0.5396, "step": 10231 }, { "epoch": 1.22, "learning_rate": 3.363049419608921e-06, "loss": 0.7615, "step": 10232 }, { "epoch": 1.22, "learning_rate": 3.3627477495513883e-06, "loss": 0.8306, "step": 10233 }, { "epoch": 1.22, "learning_rate": 3.362446065232563e-06, "loss": 0.8124, "step": 10234 }, { "epoch": 1.22, "learning_rate": 3.3621443666574316e-06, "loss": 0.5333, "step": 10235 }, { "epoch": 1.22, "learning_rate": 3.361842653830981e-06, "loss": 0.5123, "step": 10236 }, { "epoch": 1.22, "learning_rate": 3.3615409267582e-06, "loss": 0.5762, "step": 10237 }, { "epoch": 1.22, "learning_rate": 3.3612391854440747e-06, "loss": 0.4381, "step": 10238 }, { "epoch": 1.22, "learning_rate": 3.360937429893593e-06, "loss": 0.7719, "step": 10239 }, { "epoch": 1.22, "learning_rate": 3.3606356601117442e-06, "loss": 0.7657, "step": 10240 }, { "epoch": 1.22, "learning_rate": 3.3603338761035154e-06, "loss": 0.5904, "step": 10241 }, { "epoch": 1.22, "learning_rate": 3.3600320778738947e-06, "loss": 0.4139, "step": 10242 }, { "epoch": 1.22, "learning_rate": 3.3597302654278725e-06, "loss": 0.6609, "step": 10243 }, { "epoch": 1.22, "learning_rate": 3.3594284387704366e-06, "loss": 0.6114, "step": 10244 }, { "epoch": 1.22, "learning_rate": 3.359126597906577e-06, "loss": 0.5576, "step": 10245 }, { "epoch": 1.22, "learning_rate": 3.3588247428412817e-06, "loss": 0.7695, "step": 10246 }, { "epoch": 1.22, "learning_rate": 3.3585228735795416e-06, "loss": 0.6754, "step": 10247 }, { "epoch": 1.22, "learning_rate": 3.3582209901263472e-06, "loss": 0.5072, "step": 10248 }, { "epoch": 1.22, "learning_rate": 3.357919092486687e-06, "loss": 0.3866, "step": 10249 }, { "epoch": 1.22, "learning_rate": 3.3576171806655522e-06, "loss": 0.5864, "step": 10250 }, { "epoch": 1.22, "learning_rate": 3.357315254667934e-06, "loss": 0.7049, "step": 10251 }, { "epoch": 1.22, "learning_rate": 3.357013314498823e-06, "loss": 0.7583, "step": 10252 }, { "epoch": 1.22, "learning_rate": 3.3567113601632085e-06, "loss": 0.7549, "step": 10253 }, { "epoch": 1.22, "learning_rate": 3.3564093916660843e-06, "loss": 0.5007, "step": 10254 }, { "epoch": 1.22, "learning_rate": 3.3561074090124412e-06, "loss": 0.5942, "step": 10255 }, { "epoch": 1.22, "learning_rate": 3.3558054122072705e-06, "loss": 0.6507, "step": 10256 }, { "epoch": 1.22, "learning_rate": 3.3555034012555637e-06, "loss": 0.8039, "step": 10257 }, { "epoch": 1.22, "learning_rate": 3.355201376162315e-06, "loss": 0.8265, "step": 10258 }, { "epoch": 1.22, "learning_rate": 3.3548993369325155e-06, "loss": 0.9883, "step": 10259 }, { "epoch": 1.22, "learning_rate": 3.354597283571158e-06, "loss": 0.4239, "step": 10260 }, { "epoch": 1.22, "learning_rate": 3.354295216083236e-06, "loss": 0.442, "step": 10261 }, { "epoch": 1.22, "learning_rate": 3.3539931344737433e-06, "loss": 0.5865, "step": 10262 }, { "epoch": 1.22, "learning_rate": 3.3536910387476714e-06, "loss": 0.6358, "step": 10263 }, { "epoch": 1.22, "learning_rate": 3.353388928910014e-06, "loss": 0.6493, "step": 10264 }, { "epoch": 1.22, "learning_rate": 3.353086804965767e-06, "loss": 0.8394, "step": 10265 }, { "epoch": 1.22, "learning_rate": 3.3527846669199243e-06, "loss": 0.4559, "step": 10266 }, { "epoch": 1.22, "learning_rate": 3.352482514777479e-06, "loss": 0.4391, "step": 10267 }, { "epoch": 1.22, "learning_rate": 3.3521803485434256e-06, "loss": 0.5139, "step": 10268 }, { "epoch": 1.22, "learning_rate": 3.3518781682227604e-06, "loss": 0.5595, "step": 10269 }, { "epoch": 1.22, "learning_rate": 3.3515759738204766e-06, "loss": 0.7176, "step": 10270 }, { "epoch": 1.22, "learning_rate": 3.351273765341571e-06, "loss": 0.83, "step": 10271 }, { "epoch": 1.22, "learning_rate": 3.350971542791039e-06, "loss": 0.8392, "step": 10272 }, { "epoch": 1.22, "learning_rate": 3.350669306173876e-06, "loss": 0.3495, "step": 10273 }, { "epoch": 1.22, "learning_rate": 3.350367055495077e-06, "loss": 0.5552, "step": 10274 }, { "epoch": 1.22, "learning_rate": 3.3500647907596394e-06, "loss": 0.639, "step": 10275 }, { "epoch": 1.22, "learning_rate": 3.349762511972561e-06, "loss": 0.5735, "step": 10276 }, { "epoch": 1.22, "learning_rate": 3.349460219138835e-06, "loss": 0.9422, "step": 10277 }, { "epoch": 1.22, "learning_rate": 3.3491579122634614e-06, "loss": 0.7713, "step": 10278 }, { "epoch": 1.22, "learning_rate": 3.3488555913514363e-06, "loss": 0.4924, "step": 10279 }, { "epoch": 1.22, "learning_rate": 3.3485532564077573e-06, "loss": 0.5216, "step": 10280 }, { "epoch": 1.22, "learning_rate": 3.348250907437421e-06, "loss": 0.5556, "step": 10281 }, { "epoch": 1.22, "learning_rate": 3.3479485444454263e-06, "loss": 0.5781, "step": 10282 }, { "epoch": 1.22, "learning_rate": 3.347646167436771e-06, "loss": 0.7585, "step": 10283 }, { "epoch": 1.22, "learning_rate": 3.347343776416454e-06, "loss": 0.8918, "step": 10284 }, { "epoch": 1.22, "learning_rate": 3.3470413713894722e-06, "loss": 0.4383, "step": 10285 }, { "epoch": 1.22, "learning_rate": 3.3467389523608262e-06, "loss": 0.3535, "step": 10286 }, { "epoch": 1.23, "learning_rate": 3.3464365193355143e-06, "loss": 0.5692, "step": 10287 }, { "epoch": 1.23, "learning_rate": 3.346134072318535e-06, "loss": 0.608, "step": 10288 }, { "epoch": 1.23, "learning_rate": 3.3458316113148885e-06, "loss": 0.6335, "step": 10289 }, { "epoch": 1.23, "learning_rate": 3.3455291363295757e-06, "loss": 0.7763, "step": 10290 }, { "epoch": 1.23, "learning_rate": 3.3452266473675945e-06, "loss": 0.568, "step": 10291 }, { "epoch": 1.23, "learning_rate": 3.344924144433946e-06, "loss": 0.465, "step": 10292 }, { "epoch": 1.23, "learning_rate": 3.3446216275336302e-06, "loss": 0.447, "step": 10293 }, { "epoch": 1.23, "learning_rate": 3.344319096671649e-06, "loss": 0.5642, "step": 10294 }, { "epoch": 1.23, "learning_rate": 3.3440165518530017e-06, "loss": 0.7002, "step": 10295 }, { "epoch": 1.23, "learning_rate": 3.3437139930826896e-06, "loss": 0.8209, "step": 10296 }, { "epoch": 1.23, "learning_rate": 3.343411420365715e-06, "loss": 0.8908, "step": 10297 }, { "epoch": 1.23, "learning_rate": 3.343108833707078e-06, "loss": 0.4429, "step": 10298 }, { "epoch": 1.23, "learning_rate": 3.342806233111782e-06, "loss": 0.5261, "step": 10299 }, { "epoch": 1.23, "learning_rate": 3.3425036185848275e-06, "loss": 0.6183, "step": 10300 }, { "epoch": 1.23, "learning_rate": 3.342200990131218e-06, "loss": 0.755, "step": 10301 }, { "epoch": 1.23, "learning_rate": 3.341898347755956e-06, "loss": 0.7938, "step": 10302 }, { "epoch": 1.23, "learning_rate": 3.3415956914640437e-06, "loss": 0.7902, "step": 10303 }, { "epoch": 1.23, "learning_rate": 3.3412930212604826e-06, "loss": 0.5087, "step": 10304 }, { "epoch": 1.23, "learning_rate": 3.3409903371502787e-06, "loss": 0.5173, "step": 10305 }, { "epoch": 1.23, "learning_rate": 3.340687639138433e-06, "loss": 0.5189, "step": 10306 }, { "epoch": 1.23, "learning_rate": 3.3403849272299503e-06, "loss": 0.5881, "step": 10307 }, { "epoch": 1.23, "learning_rate": 3.340082201429835e-06, "loss": 0.7835, "step": 10308 }, { "epoch": 1.23, "learning_rate": 3.3397794617430903e-06, "loss": 0.8348, "step": 10309 }, { "epoch": 1.23, "learning_rate": 3.3394767081747194e-06, "loss": 0.5448, "step": 10310 }, { "epoch": 1.23, "learning_rate": 3.3391739407297287e-06, "loss": 0.481, "step": 10311 }, { "epoch": 1.23, "learning_rate": 3.338871159413123e-06, "loss": 0.5839, "step": 10312 }, { "epoch": 1.23, "learning_rate": 3.338568364229906e-06, "loss": 0.5995, "step": 10313 }, { "epoch": 1.23, "learning_rate": 3.338265555185084e-06, "loss": 0.6804, "step": 10314 }, { "epoch": 1.23, "learning_rate": 3.337962732283661e-06, "loss": 0.8278, "step": 10315 }, { "epoch": 1.23, "learning_rate": 3.337659895530645e-06, "loss": 0.6653, "step": 10316 }, { "epoch": 1.23, "learning_rate": 3.33735704493104e-06, "loss": 0.4541, "step": 10317 }, { "epoch": 1.23, "learning_rate": 3.3370541804898527e-06, "loss": 0.4938, "step": 10318 }, { "epoch": 1.23, "learning_rate": 3.33675130221209e-06, "loss": 0.5575, "step": 10319 }, { "epoch": 1.23, "learning_rate": 3.3364484101027582e-06, "loss": 0.7361, "step": 10320 }, { "epoch": 1.23, "learning_rate": 3.3361455041668634e-06, "loss": 0.8317, "step": 10321 }, { "epoch": 1.23, "learning_rate": 3.335842584409413e-06, "loss": 0.8347, "step": 10322 }, { "epoch": 1.23, "learning_rate": 3.3355396508354156e-06, "loss": 0.4504, "step": 10323 }, { "epoch": 1.23, "learning_rate": 3.335236703449877e-06, "loss": 0.4949, "step": 10324 }, { "epoch": 1.23, "learning_rate": 3.3349337422578053e-06, "loss": 0.5211, "step": 10325 }, { "epoch": 1.23, "learning_rate": 3.3346307672642096e-06, "loss": 0.7279, "step": 10326 }, { "epoch": 1.23, "learning_rate": 3.3343277784740972e-06, "loss": 0.7859, "step": 10327 }, { "epoch": 1.23, "learning_rate": 3.334024775892476e-06, "loss": 0.8337, "step": 10328 }, { "epoch": 1.23, "learning_rate": 3.3337217595243555e-06, "loss": 0.4363, "step": 10329 }, { "epoch": 1.23, "learning_rate": 3.3334187293747445e-06, "loss": 0.4722, "step": 10330 }, { "epoch": 1.23, "learning_rate": 3.3331156854486525e-06, "loss": 0.5692, "step": 10331 }, { "epoch": 1.23, "learning_rate": 3.332812627751087e-06, "loss": 0.6973, "step": 10332 }, { "epoch": 1.23, "learning_rate": 3.3325095562870597e-06, "loss": 0.825, "step": 10333 }, { "epoch": 1.23, "learning_rate": 3.3322064710615796e-06, "loss": 0.8686, "step": 10334 }, { "epoch": 1.23, "learning_rate": 3.331903372079657e-06, "loss": 0.3674, "step": 10335 }, { "epoch": 1.23, "learning_rate": 3.3316002593463007e-06, "loss": 0.5074, "step": 10336 }, { "epoch": 1.23, "learning_rate": 3.331297132866524e-06, "loss": 0.5901, "step": 10337 }, { "epoch": 1.23, "learning_rate": 3.3309939926453343e-06, "loss": 0.5374, "step": 10338 }, { "epoch": 1.23, "learning_rate": 3.330690838687745e-06, "loss": 0.6849, "step": 10339 }, { "epoch": 1.23, "learning_rate": 3.3303876709987665e-06, "loss": 0.7514, "step": 10340 }, { "epoch": 1.23, "learning_rate": 3.33008448958341e-06, "loss": 0.5069, "step": 10341 }, { "epoch": 1.23, "learning_rate": 3.3297812944466866e-06, "loss": 0.4724, "step": 10342 }, { "epoch": 1.23, "learning_rate": 3.3294780855936097e-06, "loss": 0.575, "step": 10343 }, { "epoch": 1.23, "learning_rate": 3.3291748630291902e-06, "loss": 0.5119, "step": 10344 }, { "epoch": 1.23, "learning_rate": 3.3288716267584407e-06, "loss": 0.6702, "step": 10345 }, { "epoch": 1.23, "learning_rate": 3.328568376786374e-06, "loss": 0.7357, "step": 10346 }, { "epoch": 1.23, "learning_rate": 3.328265113118002e-06, "loss": 0.891, "step": 10347 }, { "epoch": 1.23, "learning_rate": 3.3279618357583383e-06, "loss": 0.3972, "step": 10348 }, { "epoch": 1.23, "learning_rate": 3.3276585447123957e-06, "loss": 0.5796, "step": 10349 }, { "epoch": 1.23, "learning_rate": 3.3273552399851883e-06, "loss": 0.6143, "step": 10350 }, { "epoch": 1.23, "learning_rate": 3.3270519215817285e-06, "loss": 0.5345, "step": 10351 }, { "epoch": 1.23, "learning_rate": 3.3267485895070325e-06, "loss": 0.7705, "step": 10352 }, { "epoch": 1.23, "learning_rate": 3.3264452437661116e-06, "loss": 0.8843, "step": 10353 }, { "epoch": 1.23, "learning_rate": 3.326141884363982e-06, "loss": 0.4207, "step": 10354 }, { "epoch": 1.23, "learning_rate": 3.325838511305658e-06, "loss": 0.5362, "step": 10355 }, { "epoch": 1.23, "learning_rate": 3.325535124596154e-06, "loss": 0.5152, "step": 10356 }, { "epoch": 1.23, "learning_rate": 3.3252317242404848e-06, "loss": 0.5236, "step": 10357 }, { "epoch": 1.23, "learning_rate": 3.3249283102436657e-06, "loss": 0.8015, "step": 10358 }, { "epoch": 1.23, "learning_rate": 3.3246248826107143e-06, "loss": 0.9064, "step": 10359 }, { "epoch": 1.23, "learning_rate": 3.3243214413466417e-06, "loss": 0.549, "step": 10360 }, { "epoch": 1.23, "learning_rate": 3.3240179864564683e-06, "loss": 0.4946, "step": 10361 }, { "epoch": 1.23, "learning_rate": 3.323714517945208e-06, "loss": 0.5746, "step": 10362 }, { "epoch": 1.23, "learning_rate": 3.3234110358178778e-06, "loss": 0.6692, "step": 10363 }, { "epoch": 1.23, "learning_rate": 3.323107540079493e-06, "loss": 0.795, "step": 10364 }, { "epoch": 1.23, "learning_rate": 3.3228040307350728e-06, "loss": 0.8881, "step": 10365 }, { "epoch": 1.23, "learning_rate": 3.3225005077896323e-06, "loss": 0.6183, "step": 10366 }, { "epoch": 1.23, "learning_rate": 3.32219697124819e-06, "loss": 0.3968, "step": 10367 }, { "epoch": 1.23, "learning_rate": 3.3218934211157615e-06, "loss": 0.6438, "step": 10368 }, { "epoch": 1.23, "learning_rate": 3.3215898573973678e-06, "loss": 0.5307, "step": 10369 }, { "epoch": 1.23, "learning_rate": 3.321286280098023e-06, "loss": 0.6367, "step": 10370 }, { "epoch": 1.24, "learning_rate": 3.3209826892227475e-06, "loss": 0.8925, "step": 10371 }, { "epoch": 1.24, "learning_rate": 3.3206790847765596e-06, "loss": 0.7784, "step": 10372 }, { "epoch": 1.24, "learning_rate": 3.320375466764478e-06, "loss": 0.4809, "step": 10373 }, { "epoch": 1.24, "learning_rate": 3.3200718351915194e-06, "loss": 0.6086, "step": 10374 }, { "epoch": 1.24, "learning_rate": 3.3197681900627067e-06, "loss": 0.5781, "step": 10375 }, { "epoch": 1.24, "learning_rate": 3.3194645313830553e-06, "loss": 0.5684, "step": 10376 }, { "epoch": 1.24, "learning_rate": 3.3191608591575874e-06, "loss": 0.8066, "step": 10377 }, { "epoch": 1.24, "learning_rate": 3.318857173391322e-06, "loss": 0.6601, "step": 10378 }, { "epoch": 1.24, "learning_rate": 3.3185534740892776e-06, "loss": 0.4095, "step": 10379 }, { "epoch": 1.24, "learning_rate": 3.318249761256477e-06, "loss": 0.6022, "step": 10380 }, { "epoch": 1.24, "learning_rate": 3.3179460348979383e-06, "loss": 0.5735, "step": 10381 }, { "epoch": 1.24, "learning_rate": 3.317642295018683e-06, "loss": 0.685, "step": 10382 }, { "epoch": 1.24, "learning_rate": 3.3173385416237325e-06, "loss": 0.7546, "step": 10383 }, { "epoch": 1.24, "learning_rate": 3.3170347747181077e-06, "loss": 0.8657, "step": 10384 }, { "epoch": 1.24, "learning_rate": 3.3167309943068283e-06, "loss": 0.4948, "step": 10385 }, { "epoch": 1.24, "learning_rate": 3.3164272003949174e-06, "loss": 0.4761, "step": 10386 }, { "epoch": 1.24, "learning_rate": 3.3161233929873965e-06, "loss": 0.5225, "step": 10387 }, { "epoch": 1.24, "learning_rate": 3.315819572089288e-06, "loss": 0.56, "step": 10388 }, { "epoch": 1.24, "learning_rate": 3.3155157377056124e-06, "loss": 0.6839, "step": 10389 }, { "epoch": 1.24, "learning_rate": 3.3152118898413937e-06, "loss": 0.8535, "step": 10390 }, { "epoch": 1.24, "learning_rate": 3.314908028501655e-06, "loss": 0.5748, "step": 10391 }, { "epoch": 1.24, "learning_rate": 3.314604153691417e-06, "loss": 0.4827, "step": 10392 }, { "epoch": 1.24, "learning_rate": 3.314300265415704e-06, "loss": 0.6469, "step": 10393 }, { "epoch": 1.24, "learning_rate": 3.3139963636795402e-06, "loss": 0.6297, "step": 10394 }, { "epoch": 1.24, "learning_rate": 3.313692448487948e-06, "loss": 0.6323, "step": 10395 }, { "epoch": 1.24, "learning_rate": 3.3133885198459503e-06, "loss": 0.7885, "step": 10396 }, { "epoch": 1.24, "learning_rate": 3.313084577758573e-06, "loss": 0.8985, "step": 10397 }, { "epoch": 1.24, "learning_rate": 3.3127806222308394e-06, "loss": 0.3648, "step": 10398 }, { "epoch": 1.24, "learning_rate": 3.3124766532677738e-06, "loss": 0.4544, "step": 10399 }, { "epoch": 1.24, "learning_rate": 3.3121726708744006e-06, "loss": 0.5954, "step": 10400 }, { "epoch": 1.24, "learning_rate": 3.3118686750557453e-06, "loss": 0.6946, "step": 10401 }, { "epoch": 1.24, "learning_rate": 3.311564665816833e-06, "loss": 0.7798, "step": 10402 }, { "epoch": 1.24, "learning_rate": 3.311260643162689e-06, "loss": 0.735, "step": 10403 }, { "epoch": 1.24, "learning_rate": 3.3109566070983367e-06, "loss": 0.4141, "step": 10404 }, { "epoch": 1.24, "learning_rate": 3.3106525576288056e-06, "loss": 0.5921, "step": 10405 }, { "epoch": 1.24, "learning_rate": 3.3103484947591192e-06, "loss": 0.5211, "step": 10406 }, { "epoch": 1.24, "learning_rate": 3.3100444184943035e-06, "loss": 0.6061, "step": 10407 }, { "epoch": 1.24, "learning_rate": 3.309740328839386e-06, "loss": 0.7403, "step": 10408 }, { "epoch": 1.24, "learning_rate": 3.309436225799393e-06, "loss": 0.8667, "step": 10409 }, { "epoch": 1.24, "learning_rate": 3.3091321093793514e-06, "loss": 0.343, "step": 10410 }, { "epoch": 1.24, "learning_rate": 3.3088279795842875e-06, "loss": 0.5392, "step": 10411 }, { "epoch": 1.24, "learning_rate": 3.3085238364192307e-06, "loss": 0.6377, "step": 10412 }, { "epoch": 1.24, "learning_rate": 3.3082196798892053e-06, "loss": 0.719, "step": 10413 }, { "epoch": 1.24, "learning_rate": 3.3079155099992418e-06, "loss": 0.8284, "step": 10414 }, { "epoch": 1.24, "learning_rate": 3.3076113267543664e-06, "loss": 0.7935, "step": 10415 }, { "epoch": 1.24, "learning_rate": 3.307307130159609e-06, "loss": 0.5409, "step": 10416 }, { "epoch": 1.24, "learning_rate": 3.307002920219996e-06, "loss": 0.5009, "step": 10417 }, { "epoch": 1.24, "learning_rate": 3.306698696940557e-06, "loss": 0.4968, "step": 10418 }, { "epoch": 1.24, "learning_rate": 3.306394460326322e-06, "loss": 0.5621, "step": 10419 }, { "epoch": 1.24, "learning_rate": 3.306090210382318e-06, "loss": 0.6609, "step": 10420 }, { "epoch": 1.24, "learning_rate": 3.305785947113575e-06, "loss": 0.7819, "step": 10421 }, { "epoch": 1.24, "learning_rate": 3.3054816705251237e-06, "loss": 0.746, "step": 10422 }, { "epoch": 1.24, "learning_rate": 3.305177380621992e-06, "loss": 0.4933, "step": 10423 }, { "epoch": 1.24, "learning_rate": 3.304873077409211e-06, "loss": 0.5031, "step": 10424 }, { "epoch": 1.24, "learning_rate": 3.3045687608918103e-06, "loss": 0.6219, "step": 10425 }, { "epoch": 1.24, "learning_rate": 3.3042644310748206e-06, "loss": 0.6066, "step": 10426 }, { "epoch": 1.24, "learning_rate": 3.3039600879632723e-06, "loss": 0.7436, "step": 10427 }, { "epoch": 1.24, "learning_rate": 3.303655731562196e-06, "loss": 0.8939, "step": 10428 }, { "epoch": 1.24, "learning_rate": 3.3033513618766233e-06, "loss": 0.3847, "step": 10429 }, { "epoch": 1.24, "learning_rate": 3.303046978911585e-06, "loss": 0.5169, "step": 10430 }, { "epoch": 1.24, "learning_rate": 3.3027425826721133e-06, "loss": 0.5298, "step": 10431 }, { "epoch": 1.24, "learning_rate": 3.3024381731632383e-06, "loss": 0.5757, "step": 10432 }, { "epoch": 1.24, "learning_rate": 3.302133750389994e-06, "loss": 0.7278, "step": 10433 }, { "epoch": 1.24, "learning_rate": 3.3018293143574108e-06, "loss": 0.827, "step": 10434 }, { "epoch": 1.24, "learning_rate": 3.3015248650705222e-06, "loss": 0.4085, "step": 10435 }, { "epoch": 1.24, "learning_rate": 3.3012204025343596e-06, "loss": 0.6073, "step": 10436 }, { "epoch": 1.24, "learning_rate": 3.300915926753958e-06, "loss": 0.6371, "step": 10437 }, { "epoch": 1.24, "learning_rate": 3.3006114377343484e-06, "loss": 0.6706, "step": 10438 }, { "epoch": 1.24, "learning_rate": 3.300306935480564e-06, "loss": 0.7537, "step": 10439 }, { "epoch": 1.24, "learning_rate": 3.300002419997639e-06, "loss": 0.7767, "step": 10440 }, { "epoch": 1.24, "learning_rate": 3.299697891290607e-06, "loss": 0.6205, "step": 10441 }, { "epoch": 1.24, "learning_rate": 3.2993933493645016e-06, "loss": 0.4798, "step": 10442 }, { "epoch": 1.24, "learning_rate": 3.2990887942243567e-06, "loss": 0.5507, "step": 10443 }, { "epoch": 1.24, "learning_rate": 3.2987842258752073e-06, "loss": 0.6178, "step": 10444 }, { "epoch": 1.24, "learning_rate": 3.2984796443220878e-06, "loss": 0.6461, "step": 10445 }, { "epoch": 1.24, "learning_rate": 3.2981750495700325e-06, "loss": 0.7585, "step": 10446 }, { "epoch": 1.24, "learning_rate": 3.297870441624077e-06, "loss": 0.861, "step": 10447 }, { "epoch": 1.24, "learning_rate": 3.297565820489257e-06, "loss": 0.4981, "step": 10448 }, { "epoch": 1.24, "learning_rate": 3.297261186170605e-06, "loss": 0.5081, "step": 10449 }, { "epoch": 1.24, "learning_rate": 3.29695653867316e-06, "loss": 0.5064, "step": 10450 }, { "epoch": 1.24, "learning_rate": 3.2966518780019565e-06, "loss": 0.5791, "step": 10451 }, { "epoch": 1.24, "learning_rate": 3.2963472041620304e-06, "loss": 0.7687, "step": 10452 }, { "epoch": 1.24, "learning_rate": 3.2960425171584174e-06, "loss": 0.8409, "step": 10453 }, { "epoch": 1.24, "learning_rate": 3.295737816996156e-06, "loss": 0.4531, "step": 10454 }, { "epoch": 1.25, "learning_rate": 3.2954331036802824e-06, "loss": 0.5692, "step": 10455 }, { "epoch": 1.25, "learning_rate": 3.2951283772158315e-06, "loss": 0.5695, "step": 10456 }, { "epoch": 1.25, "learning_rate": 3.2948236376078417e-06, "loss": 0.6315, "step": 10457 }, { "epoch": 1.25, "learning_rate": 3.2945188848613506e-06, "loss": 0.7203, "step": 10458 }, { "epoch": 1.25, "learning_rate": 3.294214118981397e-06, "loss": 0.7736, "step": 10459 }, { "epoch": 1.25, "learning_rate": 3.2939093399730156e-06, "loss": 0.549, "step": 10460 }, { "epoch": 1.25, "learning_rate": 3.293604547841247e-06, "loss": 0.4774, "step": 10461 }, { "epoch": 1.25, "learning_rate": 3.293299742591128e-06, "loss": 0.4312, "step": 10462 }, { "epoch": 1.25, "learning_rate": 3.2929949242276993e-06, "loss": 0.644, "step": 10463 }, { "epoch": 1.25, "learning_rate": 3.2926900927559963e-06, "loss": 0.5826, "step": 10464 }, { "epoch": 1.25, "learning_rate": 3.2923852481810604e-06, "loss": 0.7487, "step": 10465 }, { "epoch": 1.25, "learning_rate": 3.2920803905079297e-06, "loss": 0.4623, "step": 10466 }, { "epoch": 1.25, "learning_rate": 3.2917755197416436e-06, "loss": 0.4731, "step": 10467 }, { "epoch": 1.25, "learning_rate": 3.2914706358872412e-06, "loss": 0.5248, "step": 10468 }, { "epoch": 1.25, "learning_rate": 3.291165738949764e-06, "loss": 0.6854, "step": 10469 }, { "epoch": 1.25, "learning_rate": 3.29086082893425e-06, "loss": 0.7173, "step": 10470 }, { "epoch": 1.25, "learning_rate": 3.29055590584574e-06, "loss": 0.7247, "step": 10471 }, { "epoch": 1.25, "learning_rate": 3.2902509696892747e-06, "loss": 0.8196, "step": 10472 }, { "epoch": 1.25, "learning_rate": 3.2899460204698946e-06, "loss": 0.356, "step": 10473 }, { "epoch": 1.25, "learning_rate": 3.2896410581926406e-06, "loss": 0.4807, "step": 10474 }, { "epoch": 1.25, "learning_rate": 3.289336082862553e-06, "loss": 0.6371, "step": 10475 }, { "epoch": 1.25, "learning_rate": 3.289031094484675e-06, "loss": 0.648, "step": 10476 }, { "epoch": 1.25, "learning_rate": 3.288726093064046e-06, "loss": 0.7778, "step": 10477 }, { "epoch": 1.25, "learning_rate": 3.2884210786057088e-06, "loss": 0.8678, "step": 10478 }, { "epoch": 1.25, "learning_rate": 3.288116051114704e-06, "loss": 0.4816, "step": 10479 }, { "epoch": 1.25, "learning_rate": 3.2878110105960766e-06, "loss": 0.6941, "step": 10480 }, { "epoch": 1.25, "learning_rate": 3.287505957054865e-06, "loss": 0.555, "step": 10481 }, { "epoch": 1.25, "learning_rate": 3.287200890496115e-06, "loss": 0.6273, "step": 10482 }, { "epoch": 1.25, "learning_rate": 3.286895810924868e-06, "loss": 0.7835, "step": 10483 }, { "epoch": 1.25, "learning_rate": 3.286590718346168e-06, "loss": 0.8488, "step": 10484 }, { "epoch": 1.25, "learning_rate": 3.2862856127650568e-06, "loss": 0.5409, "step": 10485 }, { "epoch": 1.25, "learning_rate": 3.285980494186578e-06, "loss": 0.5413, "step": 10486 }, { "epoch": 1.25, "learning_rate": 3.285675362615776e-06, "loss": 0.6603, "step": 10487 }, { "epoch": 1.25, "learning_rate": 3.2853702180576945e-06, "loss": 0.7122, "step": 10488 }, { "epoch": 1.25, "learning_rate": 3.285065060517378e-06, "loss": 0.7557, "step": 10489 }, { "epoch": 1.25, "learning_rate": 3.284759889999869e-06, "loss": 0.855, "step": 10490 }, { "epoch": 1.25, "learning_rate": 3.284454706510214e-06, "loss": 0.4008, "step": 10491 }, { "epoch": 1.25, "learning_rate": 3.284149510053456e-06, "loss": 0.4609, "step": 10492 }, { "epoch": 1.25, "learning_rate": 3.2838443006346415e-06, "loss": 0.5652, "step": 10493 }, { "epoch": 1.25, "learning_rate": 3.283539078258815e-06, "loss": 0.6742, "step": 10494 }, { "epoch": 1.25, "learning_rate": 3.2832338429310217e-06, "loss": 0.5843, "step": 10495 }, { "epoch": 1.25, "learning_rate": 3.2829285946563067e-06, "loss": 0.7437, "step": 10496 }, { "epoch": 1.25, "learning_rate": 3.2826233334397172e-06, "loss": 0.9261, "step": 10497 }, { "epoch": 1.25, "learning_rate": 3.282318059286298e-06, "loss": 0.4598, "step": 10498 }, { "epoch": 1.25, "learning_rate": 3.282012772201095e-06, "loss": 0.6059, "step": 10499 }, { "epoch": 1.25, "learning_rate": 3.2817074721891554e-06, "loss": 0.603, "step": 10500 }, { "epoch": 1.25, "learning_rate": 3.281402159255527e-06, "loss": 0.6109, "step": 10501 }, { "epoch": 1.25, "learning_rate": 3.2810968334052545e-06, "loss": 0.6585, "step": 10502 }, { "epoch": 1.25, "learning_rate": 3.2807914946433857e-06, "loss": 0.8082, "step": 10503 }, { "epoch": 1.25, "learning_rate": 3.280486142974968e-06, "loss": 0.5418, "step": 10504 }, { "epoch": 1.25, "learning_rate": 3.280180778405049e-06, "loss": 0.4579, "step": 10505 }, { "epoch": 1.25, "learning_rate": 3.2798754009386762e-06, "loss": 0.5088, "step": 10506 }, { "epoch": 1.25, "learning_rate": 3.279570010580897e-06, "loss": 0.756, "step": 10507 }, { "epoch": 1.25, "learning_rate": 3.2792646073367605e-06, "loss": 0.752, "step": 10508 }, { "epoch": 1.25, "learning_rate": 3.2789591912113154e-06, "loss": 0.8171, "step": 10509 }, { "epoch": 1.25, "learning_rate": 3.278653762209609e-06, "loss": 0.3771, "step": 10510 }, { "epoch": 1.25, "learning_rate": 3.27834832033669e-06, "loss": 0.4479, "step": 10511 }, { "epoch": 1.25, "learning_rate": 3.2780428655976094e-06, "loss": 0.4977, "step": 10512 }, { "epoch": 1.25, "learning_rate": 3.2777373979974135e-06, "loss": 0.5988, "step": 10513 }, { "epoch": 1.25, "learning_rate": 3.277431917541154e-06, "loss": 0.6838, "step": 10514 }, { "epoch": 1.25, "learning_rate": 3.2771264242338786e-06, "loss": 0.8277, "step": 10515 }, { "epoch": 1.25, "learning_rate": 3.27682091808064e-06, "loss": 0.4108, "step": 10516 }, { "epoch": 1.25, "learning_rate": 3.2765153990864846e-06, "loss": 0.4784, "step": 10517 }, { "epoch": 1.25, "learning_rate": 3.276209867256465e-06, "loss": 0.4601, "step": 10518 }, { "epoch": 1.25, "learning_rate": 3.2759043225956315e-06, "loss": 0.6259, "step": 10519 }, { "epoch": 1.25, "learning_rate": 3.275598765109035e-06, "loss": 0.7286, "step": 10520 }, { "epoch": 1.25, "learning_rate": 3.2752931948017255e-06, "loss": 0.8598, "step": 10521 }, { "epoch": 1.25, "learning_rate": 3.2749876116787534e-06, "loss": 0.9589, "step": 10522 }, { "epoch": 1.25, "learning_rate": 3.2746820157451724e-06, "loss": 0.4088, "step": 10523 }, { "epoch": 1.25, "learning_rate": 3.2743764070060317e-06, "loss": 0.4587, "step": 10524 }, { "epoch": 1.25, "learning_rate": 3.2740707854663843e-06, "loss": 0.6449, "step": 10525 }, { "epoch": 1.25, "learning_rate": 3.2737651511312823e-06, "loss": 0.6434, "step": 10526 }, { "epoch": 1.25, "learning_rate": 3.273459504005777e-06, "loss": 0.7521, "step": 10527 }, { "epoch": 1.25, "learning_rate": 3.273153844094921e-06, "loss": 0.823, "step": 10528 }, { "epoch": 1.25, "learning_rate": 3.2728481714037672e-06, "loss": 0.4758, "step": 10529 }, { "epoch": 1.25, "learning_rate": 3.272542485937369e-06, "loss": 0.5728, "step": 10530 }, { "epoch": 1.25, "learning_rate": 3.2722367877007782e-06, "loss": 0.4264, "step": 10531 }, { "epoch": 1.25, "learning_rate": 3.2719310766990483e-06, "loss": 0.5778, "step": 10532 }, { "epoch": 1.25, "learning_rate": 3.2716253529372333e-06, "loss": 0.7421, "step": 10533 }, { "epoch": 1.25, "learning_rate": 3.2713196164203866e-06, "loss": 0.95, "step": 10534 }, { "epoch": 1.25, "learning_rate": 3.271013867153561e-06, "loss": 0.4582, "step": 10535 }, { "epoch": 1.25, "learning_rate": 3.2707081051418123e-06, "loss": 0.5347, "step": 10536 }, { "epoch": 1.25, "learning_rate": 3.270402330390194e-06, "loss": 0.5165, "step": 10537 }, { "epoch": 1.25, "learning_rate": 3.2700965429037602e-06, "loss": 0.5826, "step": 10538 }, { "epoch": 1.26, "learning_rate": 3.269790742687565e-06, "loss": 0.7761, "step": 10539 }, { "epoch": 1.26, "learning_rate": 3.269484929746666e-06, "loss": 0.938, "step": 10540 }, { "epoch": 1.26, "learning_rate": 3.2691791040861153e-06, "loss": 0.4326, "step": 10541 }, { "epoch": 1.26, "learning_rate": 3.26887326571097e-06, "loss": 0.4369, "step": 10542 }, { "epoch": 1.26, "learning_rate": 3.268567414626285e-06, "loss": 0.5171, "step": 10543 }, { "epoch": 1.26, "learning_rate": 3.268261550837116e-06, "loss": 0.5312, "step": 10544 }, { "epoch": 1.26, "learning_rate": 3.2679556743485196e-06, "loss": 0.7447, "step": 10545 }, { "epoch": 1.26, "learning_rate": 3.2676497851655512e-06, "loss": 0.7263, "step": 10546 }, { "epoch": 1.26, "learning_rate": 3.267343883293267e-06, "loss": 0.9212, "step": 10547 }, { "epoch": 1.26, "learning_rate": 3.267037968736725e-06, "loss": 0.5199, "step": 10548 }, { "epoch": 1.26, "learning_rate": 3.2667320415009803e-06, "loss": 0.6537, "step": 10549 }, { "epoch": 1.26, "learning_rate": 3.266426101591091e-06, "loss": 0.5663, "step": 10550 }, { "epoch": 1.26, "learning_rate": 3.2661201490121135e-06, "loss": 0.7739, "step": 10551 }, { "epoch": 1.26, "learning_rate": 3.265814183769106e-06, "loss": 0.7418, "step": 10552 }, { "epoch": 1.26, "learning_rate": 3.2655082058671258e-06, "loss": 0.7728, "step": 10553 }, { "epoch": 1.26, "learning_rate": 3.26520221531123e-06, "loss": 0.4741, "step": 10554 }, { "epoch": 1.26, "learning_rate": 3.2648962121064785e-06, "loss": 0.5602, "step": 10555 }, { "epoch": 1.26, "learning_rate": 3.2645901962579275e-06, "loss": 0.5896, "step": 10556 }, { "epoch": 1.26, "learning_rate": 3.264284167770637e-06, "loss": 0.6146, "step": 10557 }, { "epoch": 1.26, "learning_rate": 3.2639781266496637e-06, "loss": 0.7367, "step": 10558 }, { "epoch": 1.26, "learning_rate": 3.2636720729000704e-06, "loss": 0.858, "step": 10559 }, { "epoch": 1.26, "learning_rate": 3.2633660065269118e-06, "loss": 0.8126, "step": 10560 }, { "epoch": 1.26, "learning_rate": 3.263059927535249e-06, "loss": 0.4957, "step": 10561 }, { "epoch": 1.26, "learning_rate": 3.2627538359301424e-06, "loss": 0.5114, "step": 10562 }, { "epoch": 1.26, "learning_rate": 3.2624477317166508e-06, "loss": 0.5377, "step": 10563 }, { "epoch": 1.26, "learning_rate": 3.2621416148998336e-06, "loss": 0.8013, "step": 10564 }, { "epoch": 1.26, "learning_rate": 3.2618354854847524e-06, "loss": 0.8012, "step": 10565 }, { "epoch": 1.26, "learning_rate": 3.2615293434764665e-06, "loss": 0.4577, "step": 10566 }, { "epoch": 1.26, "learning_rate": 3.261223188880036e-06, "loss": 0.4789, "step": 10567 }, { "epoch": 1.26, "learning_rate": 3.260917021700523e-06, "loss": 0.5084, "step": 10568 }, { "epoch": 1.26, "learning_rate": 3.260610841942988e-06, "loss": 0.5425, "step": 10569 }, { "epoch": 1.26, "learning_rate": 3.2603046496124914e-06, "loss": 0.7473, "step": 10570 }, { "epoch": 1.26, "learning_rate": 3.259998444714095e-06, "loss": 0.8312, "step": 10571 }, { "epoch": 1.26, "learning_rate": 3.2596922272528607e-06, "loss": 0.7433, "step": 10572 }, { "epoch": 1.26, "learning_rate": 3.2593859972338506e-06, "loss": 0.4961, "step": 10573 }, { "epoch": 1.26, "learning_rate": 3.2590797546621255e-06, "loss": 0.5826, "step": 10574 }, { "epoch": 1.26, "learning_rate": 3.2587734995427484e-06, "loss": 0.5002, "step": 10575 }, { "epoch": 1.26, "learning_rate": 3.2584672318807818e-06, "loss": 0.6097, "step": 10576 }, { "epoch": 1.26, "learning_rate": 3.2581609516812884e-06, "loss": 0.6619, "step": 10577 }, { "epoch": 1.26, "learning_rate": 3.257854658949331e-06, "loss": 0.8765, "step": 10578 }, { "epoch": 1.26, "learning_rate": 3.257548353689972e-06, "loss": 0.4569, "step": 10579 }, { "epoch": 1.26, "learning_rate": 3.2572420359082764e-06, "loss": 0.5825, "step": 10580 }, { "epoch": 1.26, "learning_rate": 3.2569357056093055e-06, "loss": 0.65, "step": 10581 }, { "epoch": 1.26, "learning_rate": 3.256629362798124e-06, "loss": 0.5686, "step": 10582 }, { "epoch": 1.26, "learning_rate": 3.2563230074797952e-06, "loss": 0.6682, "step": 10583 }, { "epoch": 1.26, "learning_rate": 3.256016639659384e-06, "loss": 0.8419, "step": 10584 }, { "epoch": 1.26, "learning_rate": 3.2557102593419545e-06, "loss": 0.546, "step": 10585 }, { "epoch": 1.26, "learning_rate": 3.2554038665325703e-06, "loss": 0.5184, "step": 10586 }, { "epoch": 1.26, "learning_rate": 3.255097461236297e-06, "loss": 0.5731, "step": 10587 }, { "epoch": 1.26, "learning_rate": 3.2547910434582003e-06, "loss": 0.5, "step": 10588 }, { "epoch": 1.26, "learning_rate": 3.2544846132033437e-06, "loss": 0.6791, "step": 10589 }, { "epoch": 1.26, "learning_rate": 3.2541781704767926e-06, "loss": 0.7855, "step": 10590 }, { "epoch": 1.26, "learning_rate": 3.253871715283614e-06, "loss": 0.4825, "step": 10591 }, { "epoch": 1.26, "learning_rate": 3.253565247628872e-06, "loss": 0.4039, "step": 10592 }, { "epoch": 1.26, "learning_rate": 3.2532587675176336e-06, "loss": 0.588, "step": 10593 }, { "epoch": 1.26, "learning_rate": 3.2529522749549648e-06, "loss": 0.6387, "step": 10594 }, { "epoch": 1.26, "learning_rate": 3.252645769945931e-06, "loss": 0.6602, "step": 10595 }, { "epoch": 1.26, "learning_rate": 3.2523392524955992e-06, "loss": 0.6996, "step": 10596 }, { "epoch": 1.26, "learning_rate": 3.2520327226090377e-06, "loss": 0.7087, "step": 10597 }, { "epoch": 1.26, "learning_rate": 3.251726180291312e-06, "loss": 0.428, "step": 10598 }, { "epoch": 1.26, "learning_rate": 3.2514196255474885e-06, "loss": 0.5684, "step": 10599 }, { "epoch": 1.26, "learning_rate": 3.251113058382636e-06, "loss": 0.5362, "step": 10600 }, { "epoch": 1.26, "learning_rate": 3.250806478801821e-06, "loss": 0.6964, "step": 10601 }, { "epoch": 1.26, "learning_rate": 3.2504998868101136e-06, "loss": 0.8608, "step": 10602 }, { "epoch": 1.26, "learning_rate": 3.250193282412578e-06, "loss": 0.8346, "step": 10603 }, { "epoch": 1.26, "learning_rate": 3.249886665614286e-06, "loss": 0.3727, "step": 10604 }, { "epoch": 1.26, "learning_rate": 3.2495800364203043e-06, "loss": 0.6651, "step": 10605 }, { "epoch": 1.26, "learning_rate": 3.2492733948357016e-06, "loss": 0.5868, "step": 10606 }, { "epoch": 1.26, "learning_rate": 3.2489667408655463e-06, "loss": 0.6468, "step": 10607 }, { "epoch": 1.26, "learning_rate": 3.248660074514908e-06, "loss": 0.6262, "step": 10608 }, { "epoch": 1.26, "learning_rate": 3.2483533957888563e-06, "loss": 0.8699, "step": 10609 }, { "epoch": 1.26, "learning_rate": 3.24804670469246e-06, "loss": 0.4325, "step": 10610 }, { "epoch": 1.26, "learning_rate": 3.2477400012307885e-06, "loss": 0.5169, "step": 10611 }, { "epoch": 1.26, "learning_rate": 3.2474332854089136e-06, "loss": 0.594, "step": 10612 }, { "epoch": 1.26, "learning_rate": 3.2471265572319023e-06, "loss": 0.5881, "step": 10613 }, { "epoch": 1.26, "learning_rate": 3.2468198167048266e-06, "loss": 0.6595, "step": 10614 }, { "epoch": 1.26, "learning_rate": 3.246513063832757e-06, "loss": 0.8228, "step": 10615 }, { "epoch": 1.26, "learning_rate": 3.246206298620763e-06, "loss": 0.5471, "step": 10616 }, { "epoch": 1.26, "learning_rate": 3.2458995210739176e-06, "loss": 0.4929, "step": 10617 }, { "epoch": 1.26, "learning_rate": 3.245592731197289e-06, "loss": 0.5742, "step": 10618 }, { "epoch": 1.26, "learning_rate": 3.245285928995951e-06, "loss": 0.7167, "step": 10619 }, { "epoch": 1.26, "learning_rate": 3.244979114474974e-06, "loss": 0.7231, "step": 10620 }, { "epoch": 1.26, "learning_rate": 3.2446722876394294e-06, "loss": 0.7748, "step": 10621 }, { "epoch": 1.26, "learning_rate": 3.244365448494389e-06, "loss": 0.8397, "step": 10622 }, { "epoch": 1.27, "learning_rate": 3.2440585970449267e-06, "loss": 0.5249, "step": 10623 }, { "epoch": 1.27, "learning_rate": 3.243751733296112e-06, "loss": 0.5991, "step": 10624 }, { "epoch": 1.27, "learning_rate": 3.243444857253019e-06, "loss": 0.5701, "step": 10625 }, { "epoch": 1.27, "learning_rate": 3.2431379689207206e-06, "loss": 0.621, "step": 10626 }, { "epoch": 1.27, "learning_rate": 3.242831068304289e-06, "loss": 0.7562, "step": 10627 }, { "epoch": 1.27, "learning_rate": 3.242524155408797e-06, "loss": 0.7688, "step": 10628 }, { "epoch": 1.27, "learning_rate": 3.242217230239319e-06, "loss": 0.4179, "step": 10629 }, { "epoch": 1.27, "learning_rate": 3.2419102928009284e-06, "loss": 0.4735, "step": 10630 }, { "epoch": 1.27, "learning_rate": 3.2416033430986975e-06, "loss": 0.6523, "step": 10631 }, { "epoch": 1.27, "learning_rate": 3.241296381137702e-06, "loss": 0.5979, "step": 10632 }, { "epoch": 1.27, "learning_rate": 3.2409894069230134e-06, "loss": 0.7352, "step": 10633 }, { "epoch": 1.27, "learning_rate": 3.2406824204597097e-06, "loss": 1.0271, "step": 10634 }, { "epoch": 1.27, "learning_rate": 3.240375421752862e-06, "loss": 0.3983, "step": 10635 }, { "epoch": 1.27, "learning_rate": 3.2400684108075473e-06, "loss": 0.4917, "step": 10636 }, { "epoch": 1.27, "learning_rate": 3.23976138762884e-06, "loss": 0.54, "step": 10637 }, { "epoch": 1.27, "learning_rate": 3.2394543522218143e-06, "loss": 0.5744, "step": 10638 }, { "epoch": 1.27, "learning_rate": 3.2391473045915455e-06, "loss": 0.671, "step": 10639 }, { "epoch": 1.27, "learning_rate": 3.2388402447431106e-06, "loss": 0.8242, "step": 10640 }, { "epoch": 1.27, "learning_rate": 3.2385331726815837e-06, "loss": 0.6651, "step": 10641 }, { "epoch": 1.27, "learning_rate": 3.2382260884120425e-06, "loss": 0.506, "step": 10642 }, { "epoch": 1.27, "learning_rate": 3.2379189919395615e-06, "loss": 0.5907, "step": 10643 }, { "epoch": 1.27, "learning_rate": 3.2376118832692177e-06, "loss": 0.5504, "step": 10644 }, { "epoch": 1.27, "learning_rate": 3.2373047624060884e-06, "loss": 0.6784, "step": 10645 }, { "epoch": 1.27, "learning_rate": 3.236997629355248e-06, "loss": 0.7315, "step": 10646 }, { "epoch": 1.27, "learning_rate": 3.2366904841217763e-06, "loss": 0.8946, "step": 10647 }, { "epoch": 1.27, "learning_rate": 3.2363833267107482e-06, "loss": 0.4281, "step": 10648 }, { "epoch": 1.27, "learning_rate": 3.2360761571272426e-06, "loss": 0.5285, "step": 10649 }, { "epoch": 1.27, "learning_rate": 3.2357689753763357e-06, "loss": 0.5616, "step": 10650 }, { "epoch": 1.27, "learning_rate": 3.2354617814631063e-06, "loss": 0.6743, "step": 10651 }, { "epoch": 1.27, "learning_rate": 3.2351545753926316e-06, "loss": 0.7719, "step": 10652 }, { "epoch": 1.27, "learning_rate": 3.2348473571699903e-06, "loss": 0.8364, "step": 10653 }, { "epoch": 1.27, "learning_rate": 3.23454012680026e-06, "loss": 0.3704, "step": 10654 }, { "epoch": 1.27, "learning_rate": 3.234232884288521e-06, "loss": 0.529, "step": 10655 }, { "epoch": 1.27, "learning_rate": 3.2339256296398497e-06, "loss": 0.544, "step": 10656 }, { "epoch": 1.27, "learning_rate": 3.2336183628593265e-06, "loss": 0.6021, "step": 10657 }, { "epoch": 1.27, "learning_rate": 3.23331108395203e-06, "loss": 0.7035, "step": 10658 }, { "epoch": 1.27, "learning_rate": 3.2330037929230406e-06, "loss": 0.8552, "step": 10659 }, { "epoch": 1.27, "learning_rate": 3.232696489777436e-06, "loss": 0.5013, "step": 10660 }, { "epoch": 1.27, "learning_rate": 3.2323891745202977e-06, "loss": 0.4622, "step": 10661 }, { "epoch": 1.27, "learning_rate": 3.232081847156705e-06, "loss": 0.5508, "step": 10662 }, { "epoch": 1.27, "learning_rate": 3.231774507691737e-06, "loss": 0.5581, "step": 10663 }, { "epoch": 1.27, "learning_rate": 3.2314671561304763e-06, "loss": 0.6593, "step": 10664 }, { "epoch": 1.27, "learning_rate": 3.231159792478e-06, "loss": 0.8139, "step": 10665 }, { "epoch": 1.27, "learning_rate": 3.2308524167393933e-06, "loss": 0.7078, "step": 10666 }, { "epoch": 1.27, "learning_rate": 3.230545028919734e-06, "loss": 0.4537, "step": 10667 }, { "epoch": 1.27, "learning_rate": 3.230237629024104e-06, "loss": 0.4944, "step": 10668 }, { "epoch": 1.27, "learning_rate": 3.2299302170575843e-06, "loss": 0.5214, "step": 10669 }, { "epoch": 1.27, "learning_rate": 3.2296227930252576e-06, "loss": 0.617, "step": 10670 }, { "epoch": 1.27, "learning_rate": 3.2293153569322038e-06, "loss": 0.8276, "step": 10671 }, { "epoch": 1.27, "learning_rate": 3.229007908783507e-06, "loss": 0.9752, "step": 10672 }, { "epoch": 1.27, "learning_rate": 3.228700448584248e-06, "loss": 0.6739, "step": 10673 }, { "epoch": 1.27, "learning_rate": 3.2283929763395092e-06, "loss": 0.6847, "step": 10674 }, { "epoch": 1.27, "learning_rate": 3.2280854920543735e-06, "loss": 0.5837, "step": 10675 }, { "epoch": 1.27, "learning_rate": 3.2277779957339237e-06, "loss": 0.6462, "step": 10676 }, { "epoch": 1.27, "learning_rate": 3.227470487383243e-06, "loss": 0.886, "step": 10677 }, { "epoch": 1.27, "learning_rate": 3.2271629670074135e-06, "loss": 0.8135, "step": 10678 }, { "epoch": 1.27, "learning_rate": 3.2268554346115196e-06, "loss": 0.4256, "step": 10679 }, { "epoch": 1.27, "learning_rate": 3.2265478902006437e-06, "loss": 0.5714, "step": 10680 }, { "epoch": 1.27, "learning_rate": 3.226240333779871e-06, "loss": 0.553, "step": 10681 }, { "epoch": 1.27, "learning_rate": 3.225932765354283e-06, "loss": 0.6317, "step": 10682 }, { "epoch": 1.27, "learning_rate": 3.2256251849289673e-06, "loss": 0.7911, "step": 10683 }, { "epoch": 1.27, "learning_rate": 3.225317592509006e-06, "loss": 0.8591, "step": 10684 }, { "epoch": 1.27, "learning_rate": 3.2250099880994834e-06, "loss": 0.5082, "step": 10685 }, { "epoch": 1.27, "learning_rate": 3.224702371705485e-06, "loss": 0.5864, "step": 10686 }, { "epoch": 1.27, "learning_rate": 3.224394743332096e-06, "loss": 0.5919, "step": 10687 }, { "epoch": 1.27, "learning_rate": 3.224087102984401e-06, "loss": 0.5676, "step": 10688 }, { "epoch": 1.27, "learning_rate": 3.223779450667486e-06, "loss": 0.7213, "step": 10689 }, { "epoch": 1.27, "learning_rate": 3.2234717863864347e-06, "loss": 0.7802, "step": 10690 }, { "epoch": 1.27, "learning_rate": 3.223164110146335e-06, "loss": 0.6217, "step": 10691 }, { "epoch": 1.27, "learning_rate": 3.222856421952271e-06, "loss": 0.4247, "step": 10692 }, { "epoch": 1.27, "learning_rate": 3.222548721809331e-06, "loss": 0.5595, "step": 10693 }, { "epoch": 1.27, "learning_rate": 3.2222410097225994e-06, "loss": 0.6115, "step": 10694 }, { "epoch": 1.27, "learning_rate": 3.221933285697163e-06, "loss": 0.6219, "step": 10695 }, { "epoch": 1.27, "learning_rate": 3.2216255497381088e-06, "loss": 0.9022, "step": 10696 }, { "epoch": 1.27, "learning_rate": 3.2213178018505236e-06, "loss": 0.7646, "step": 10697 }, { "epoch": 1.27, "learning_rate": 3.221010042039496e-06, "loss": 0.4907, "step": 10698 }, { "epoch": 1.27, "learning_rate": 3.2207022703101097e-06, "loss": 0.5299, "step": 10699 }, { "epoch": 1.27, "learning_rate": 3.2203944866674563e-06, "loss": 0.5369, "step": 10700 }, { "epoch": 1.27, "learning_rate": 3.22008669111662e-06, "loss": 0.7342, "step": 10701 }, { "epoch": 1.27, "learning_rate": 3.219778883662692e-06, "loss": 0.8051, "step": 10702 }, { "epoch": 1.27, "learning_rate": 3.2194710643107572e-06, "loss": 0.999, "step": 10703 }, { "epoch": 1.27, "learning_rate": 3.2191632330659063e-06, "loss": 0.4678, "step": 10704 }, { "epoch": 1.27, "learning_rate": 3.2188553899332262e-06, "loss": 0.5667, "step": 10705 }, { "epoch": 1.27, "learning_rate": 3.2185475349178064e-06, "loss": 0.5823, "step": 10706 }, { "epoch": 1.28, "learning_rate": 3.2182396680247353e-06, "loss": 0.6045, "step": 10707 }, { "epoch": 1.28, "learning_rate": 3.217931789259102e-06, "loss": 0.7463, "step": 10708 }, { "epoch": 1.28, "learning_rate": 3.2176238986259976e-06, "loss": 1.006, "step": 10709 }, { "epoch": 1.28, "learning_rate": 3.2173159961305084e-06, "loss": 0.4994, "step": 10710 }, { "epoch": 1.28, "learning_rate": 3.217008081777726e-06, "loss": 0.5623, "step": 10711 }, { "epoch": 1.28, "learning_rate": 3.2167001555727403e-06, "loss": 0.4623, "step": 10712 }, { "epoch": 1.28, "learning_rate": 3.21639221752064e-06, "loss": 0.5325, "step": 10713 }, { "epoch": 1.28, "learning_rate": 3.216084267626516e-06, "loss": 0.7574, "step": 10714 }, { "epoch": 1.28, "learning_rate": 3.2157763058954605e-06, "loss": 0.8328, "step": 10715 }, { "epoch": 1.28, "learning_rate": 3.215468332332562e-06, "loss": 0.6337, "step": 10716 }, { "epoch": 1.28, "learning_rate": 3.2151603469429117e-06, "loss": 0.4379, "step": 10717 }, { "epoch": 1.28, "learning_rate": 3.2148523497316003e-06, "loss": 0.6726, "step": 10718 }, { "epoch": 1.28, "learning_rate": 3.2145443407037203e-06, "loss": 0.6011, "step": 10719 }, { "epoch": 1.28, "learning_rate": 3.214236319864362e-06, "loss": 0.8045, "step": 10720 }, { "epoch": 1.28, "learning_rate": 3.213928287218618e-06, "loss": 0.8041, "step": 10721 }, { "epoch": 1.28, "learning_rate": 3.2136202427715784e-06, "loss": 0.8766, "step": 10722 }, { "epoch": 1.28, "learning_rate": 3.213312186528338e-06, "loss": 0.4535, "step": 10723 }, { "epoch": 1.28, "learning_rate": 3.2130041184939865e-06, "loss": 0.4829, "step": 10724 }, { "epoch": 1.28, "learning_rate": 3.212696038673616e-06, "loss": 0.4767, "step": 10725 }, { "epoch": 1.28, "learning_rate": 3.2123879470723213e-06, "loss": 0.4634, "step": 10726 }, { "epoch": 1.28, "learning_rate": 3.2120798436951938e-06, "loss": 0.8027, "step": 10727 }, { "epoch": 1.28, "learning_rate": 3.211771728547327e-06, "loss": 0.7726, "step": 10728 }, { "epoch": 1.28, "learning_rate": 3.2114636016338137e-06, "loss": 0.3852, "step": 10729 }, { "epoch": 1.28, "learning_rate": 3.2111554629597475e-06, "loss": 0.5792, "step": 10730 }, { "epoch": 1.28, "learning_rate": 3.2108473125302214e-06, "loss": 0.4107, "step": 10731 }, { "epoch": 1.28, "learning_rate": 3.21053915035033e-06, "loss": 0.5453, "step": 10732 }, { "epoch": 1.28, "learning_rate": 3.210230976425166e-06, "loss": 0.7882, "step": 10733 }, { "epoch": 1.28, "learning_rate": 3.2099227907598257e-06, "loss": 0.8632, "step": 10734 }, { "epoch": 1.28, "learning_rate": 3.2096145933594008e-06, "loss": 0.5101, "step": 10735 }, { "epoch": 1.28, "learning_rate": 3.209306384228988e-06, "loss": 0.3905, "step": 10736 }, { "epoch": 1.28, "learning_rate": 3.208998163373681e-06, "loss": 0.5809, "step": 10737 }, { "epoch": 1.28, "learning_rate": 3.2086899307985746e-06, "loss": 0.4879, "step": 10738 }, { "epoch": 1.28, "learning_rate": 3.2083816865087643e-06, "loss": 0.7499, "step": 10739 }, { "epoch": 1.28, "learning_rate": 3.2080734305093453e-06, "loss": 0.8781, "step": 10740 }, { "epoch": 1.28, "learning_rate": 3.2077651628054136e-06, "loss": 0.4609, "step": 10741 }, { "epoch": 1.28, "learning_rate": 3.2074568834020634e-06, "loss": 0.4275, "step": 10742 }, { "epoch": 1.28, "learning_rate": 3.207148592304392e-06, "loss": 0.5043, "step": 10743 }, { "epoch": 1.28, "learning_rate": 3.2068402895174942e-06, "loss": 0.5366, "step": 10744 }, { "epoch": 1.28, "learning_rate": 3.2065319750464693e-06, "loss": 0.7288, "step": 10745 }, { "epoch": 1.28, "learning_rate": 3.206223648896409e-06, "loss": 0.8257, "step": 10746 }, { "epoch": 1.28, "learning_rate": 3.2059153110724144e-06, "loss": 0.7715, "step": 10747 }, { "epoch": 1.28, "learning_rate": 3.2056069615795798e-06, "loss": 0.4716, "step": 10748 }, { "epoch": 1.28, "learning_rate": 3.205298600423003e-06, "loss": 0.6093, "step": 10749 }, { "epoch": 1.28, "learning_rate": 3.204990227607781e-06, "loss": 0.4565, "step": 10750 }, { "epoch": 1.28, "learning_rate": 3.2046818431390115e-06, "loss": 0.7128, "step": 10751 }, { "epoch": 1.28, "learning_rate": 3.2043734470217917e-06, "loss": 0.7698, "step": 10752 }, { "epoch": 1.28, "learning_rate": 3.2040650392612206e-06, "loss": 0.9675, "step": 10753 }, { "epoch": 1.28, "learning_rate": 3.2037566198623943e-06, "loss": 0.3674, "step": 10754 }, { "epoch": 1.28, "learning_rate": 3.203448188830413e-06, "loss": 0.4823, "step": 10755 }, { "epoch": 1.28, "learning_rate": 3.2031397461703733e-06, "loss": 0.4931, "step": 10756 }, { "epoch": 1.28, "learning_rate": 3.2028312918873748e-06, "loss": 0.647, "step": 10757 }, { "epoch": 1.28, "learning_rate": 3.202522825986516e-06, "loss": 0.7863, "step": 10758 }, { "epoch": 1.28, "learning_rate": 3.2022143484728964e-06, "loss": 0.8239, "step": 10759 }, { "epoch": 1.28, "learning_rate": 3.2019058593516144e-06, "loss": 0.4632, "step": 10760 }, { "epoch": 1.28, "learning_rate": 3.2015973586277695e-06, "loss": 0.4947, "step": 10761 }, { "epoch": 1.28, "learning_rate": 3.2012888463064613e-06, "loss": 0.5452, "step": 10762 }, { "epoch": 1.28, "learning_rate": 3.20098032239279e-06, "loss": 0.6514, "step": 10763 }, { "epoch": 1.28, "learning_rate": 3.200671786891855e-06, "loss": 0.7273, "step": 10764 }, { "epoch": 1.28, "learning_rate": 3.2003632398087564e-06, "loss": 0.8193, "step": 10765 }, { "epoch": 1.28, "learning_rate": 3.2000546811485953e-06, "loss": 0.516, "step": 10766 }, { "epoch": 1.28, "learning_rate": 3.199746110916471e-06, "loss": 0.39, "step": 10767 }, { "epoch": 1.28, "learning_rate": 3.199437529117485e-06, "loss": 0.5735, "step": 10768 }, { "epoch": 1.28, "learning_rate": 3.199128935756738e-06, "loss": 0.516, "step": 10769 }, { "epoch": 1.28, "learning_rate": 3.1988203308393305e-06, "loss": 0.6312, "step": 10770 }, { "epoch": 1.28, "learning_rate": 3.1985117143703646e-06, "loss": 0.8435, "step": 10771 }, { "epoch": 1.28, "learning_rate": 3.1982030863549416e-06, "loss": 0.6737, "step": 10772 }, { "epoch": 1.28, "learning_rate": 3.1978944467981633e-06, "loss": 0.4308, "step": 10773 }, { "epoch": 1.28, "learning_rate": 3.1975857957051303e-06, "loss": 0.6334, "step": 10774 }, { "epoch": 1.28, "learning_rate": 3.1972771330809465e-06, "loss": 0.6503, "step": 10775 }, { "epoch": 1.28, "learning_rate": 3.196968458930713e-06, "loss": 0.6783, "step": 10776 }, { "epoch": 1.28, "learning_rate": 3.196659773259532e-06, "loss": 0.8922, "step": 10777 }, { "epoch": 1.28, "learning_rate": 3.196351076072507e-06, "loss": 0.6849, "step": 10778 }, { "epoch": 1.28, "learning_rate": 3.1960423673747397e-06, "loss": 0.4401, "step": 10779 }, { "epoch": 1.28, "learning_rate": 3.195733647171334e-06, "loss": 0.5608, "step": 10780 }, { "epoch": 1.28, "learning_rate": 3.1954249154673926e-06, "loss": 0.5928, "step": 10781 }, { "epoch": 1.28, "learning_rate": 3.195116172268019e-06, "loss": 0.798, "step": 10782 }, { "epoch": 1.28, "learning_rate": 3.1948074175783173e-06, "loss": 0.7912, "step": 10783 }, { "epoch": 1.28, "learning_rate": 3.19449865140339e-06, "loss": 0.7544, "step": 10784 }, { "epoch": 1.28, "learning_rate": 3.1941898737483424e-06, "loss": 0.5213, "step": 10785 }, { "epoch": 1.28, "learning_rate": 3.1938810846182765e-06, "loss": 0.4946, "step": 10786 }, { "epoch": 1.28, "learning_rate": 3.1935722840182994e-06, "loss": 0.5459, "step": 10787 }, { "epoch": 1.28, "learning_rate": 3.1932634719535145e-06, "loss": 0.5779, "step": 10788 }, { "epoch": 1.28, "learning_rate": 3.192954648429025e-06, "loss": 0.7141, "step": 10789 }, { "epoch": 1.28, "learning_rate": 3.1926458134499378e-06, "loss": 0.7571, "step": 10790 }, { "epoch": 1.29, "learning_rate": 3.1923369670213574e-06, "loss": 0.524, "step": 10791 }, { "epoch": 1.29, "learning_rate": 3.1920281091483886e-06, "loss": 0.3729, "step": 10792 }, { "epoch": 1.29, "learning_rate": 3.191719239836136e-06, "loss": 0.5724, "step": 10793 }, { "epoch": 1.29, "learning_rate": 3.1914103590897076e-06, "loss": 0.5351, "step": 10794 }, { "epoch": 1.29, "learning_rate": 3.1911014669142076e-06, "loss": 0.6724, "step": 10795 }, { "epoch": 1.29, "learning_rate": 3.190792563314743e-06, "loss": 0.7515, "step": 10796 }, { "epoch": 1.29, "learning_rate": 3.190483648296418e-06, "loss": 0.7771, "step": 10797 }, { "epoch": 1.29, "learning_rate": 3.190174721864342e-06, "loss": 0.4209, "step": 10798 }, { "epoch": 1.29, "learning_rate": 3.1898657840236184e-06, "loss": 0.5865, "step": 10799 }, { "epoch": 1.29, "learning_rate": 3.1895568347793565e-06, "loss": 0.7102, "step": 10800 }, { "epoch": 1.29, "learning_rate": 3.1892478741366613e-06, "loss": 0.7726, "step": 10801 }, { "epoch": 1.29, "learning_rate": 3.1889389021006423e-06, "loss": 0.9155, "step": 10802 }, { "epoch": 1.29, "learning_rate": 3.1886299186764043e-06, "loss": 0.6879, "step": 10803 }, { "epoch": 1.29, "learning_rate": 3.188320923869056e-06, "loss": 0.3975, "step": 10804 }, { "epoch": 1.29, "learning_rate": 3.188011917683706e-06, "loss": 0.5097, "step": 10805 }, { "epoch": 1.29, "learning_rate": 3.1877029001254607e-06, "loss": 0.5742, "step": 10806 }, { "epoch": 1.29, "learning_rate": 3.187393871199429e-06, "loss": 0.5984, "step": 10807 }, { "epoch": 1.29, "learning_rate": 3.187084830910718e-06, "loss": 0.7499, "step": 10808 }, { "epoch": 1.29, "learning_rate": 3.1867757792644382e-06, "loss": 0.8641, "step": 10809 }, { "epoch": 1.29, "learning_rate": 3.1864667162656963e-06, "loss": 0.5367, "step": 10810 }, { "epoch": 1.29, "learning_rate": 3.1861576419196026e-06, "loss": 0.4868, "step": 10811 }, { "epoch": 1.29, "learning_rate": 3.185848556231266e-06, "loss": 0.5312, "step": 10812 }, { "epoch": 1.29, "learning_rate": 3.185539459205794e-06, "loss": 0.5486, "step": 10813 }, { "epoch": 1.29, "learning_rate": 3.1852303508482973e-06, "loss": 0.6945, "step": 10814 }, { "epoch": 1.29, "learning_rate": 3.184921231163886e-06, "loss": 0.8725, "step": 10815 }, { "epoch": 1.29, "learning_rate": 3.184612100157669e-06, "loss": 0.5909, "step": 10816 }, { "epoch": 1.29, "learning_rate": 3.184302957834757e-06, "loss": 0.4315, "step": 10817 }, { "epoch": 1.29, "learning_rate": 3.183993804200259e-06, "loss": 0.6181, "step": 10818 }, { "epoch": 1.29, "learning_rate": 3.183684639259287e-06, "loss": 0.6921, "step": 10819 }, { "epoch": 1.29, "learning_rate": 3.18337546301695e-06, "loss": 0.688, "step": 10820 }, { "epoch": 1.29, "learning_rate": 3.1830662754783587e-06, "loss": 0.9605, "step": 10821 }, { "epoch": 1.29, "learning_rate": 3.1827570766486256e-06, "loss": 0.7144, "step": 10822 }, { "epoch": 1.29, "learning_rate": 3.18244786653286e-06, "loss": 0.4644, "step": 10823 }, { "epoch": 1.29, "learning_rate": 3.182138645136174e-06, "loss": 0.4889, "step": 10824 }, { "epoch": 1.29, "learning_rate": 3.1818294124636783e-06, "loss": 0.7332, "step": 10825 }, { "epoch": 1.29, "learning_rate": 3.1815201685204868e-06, "loss": 0.5897, "step": 10826 }, { "epoch": 1.29, "learning_rate": 3.1812109133117088e-06, "loss": 0.847, "step": 10827 }, { "epoch": 1.29, "learning_rate": 3.180901646842457e-06, "loss": 0.8785, "step": 10828 }, { "epoch": 1.29, "learning_rate": 3.180592369117844e-06, "loss": 0.3696, "step": 10829 }, { "epoch": 1.29, "learning_rate": 3.1802830801429824e-06, "loss": 0.5731, "step": 10830 }, { "epoch": 1.29, "learning_rate": 3.179973779922984e-06, "loss": 0.4754, "step": 10831 }, { "epoch": 1.29, "learning_rate": 3.1796644684629625e-06, "loss": 0.6492, "step": 10832 }, { "epoch": 1.29, "learning_rate": 3.1793551457680296e-06, "loss": 0.6983, "step": 10833 }, { "epoch": 1.29, "learning_rate": 3.1790458118433e-06, "loss": 0.8732, "step": 10834 }, { "epoch": 1.29, "learning_rate": 3.1787364666938852e-06, "loss": 0.4899, "step": 10835 }, { "epoch": 1.29, "learning_rate": 3.1784271103249005e-06, "loss": 0.4397, "step": 10836 }, { "epoch": 1.29, "learning_rate": 3.178117742741458e-06, "loss": 0.4749, "step": 10837 }, { "epoch": 1.29, "learning_rate": 3.177808363948673e-06, "loss": 0.6588, "step": 10838 }, { "epoch": 1.29, "learning_rate": 3.177498973951658e-06, "loss": 0.6672, "step": 10839 }, { "epoch": 1.29, "learning_rate": 3.177189572755529e-06, "loss": 0.8623, "step": 10840 }, { "epoch": 1.29, "learning_rate": 3.1768801603654e-06, "loss": 0.6478, "step": 10841 }, { "epoch": 1.29, "learning_rate": 3.1765707367863836e-06, "loss": 0.5111, "step": 10842 }, { "epoch": 1.29, "learning_rate": 3.1762613020235966e-06, "loss": 0.4749, "step": 10843 }, { "epoch": 1.29, "learning_rate": 3.1759518560821533e-06, "loss": 0.5706, "step": 10844 }, { "epoch": 1.29, "learning_rate": 3.175642398967171e-06, "loss": 0.7009, "step": 10845 }, { "epoch": 1.29, "learning_rate": 3.175332930683761e-06, "loss": 0.7846, "step": 10846 }, { "epoch": 1.29, "learning_rate": 3.175023451237042e-06, "loss": 0.8285, "step": 10847 }, { "epoch": 1.29, "learning_rate": 3.1747139606321285e-06, "loss": 0.496, "step": 10848 }, { "epoch": 1.29, "learning_rate": 3.1744044588741366e-06, "loss": 0.5341, "step": 10849 }, { "epoch": 1.29, "learning_rate": 3.174094945968182e-06, "loss": 0.5656, "step": 10850 }, { "epoch": 1.29, "learning_rate": 3.173785421919382e-06, "loss": 0.6194, "step": 10851 }, { "epoch": 1.29, "learning_rate": 3.173475886732853e-06, "loss": 0.8071, "step": 10852 }, { "epoch": 1.29, "learning_rate": 3.1731663404137098e-06, "loss": 0.8372, "step": 10853 }, { "epoch": 1.29, "learning_rate": 3.172856782967071e-06, "loss": 0.4123, "step": 10854 }, { "epoch": 1.29, "learning_rate": 3.1725472143980535e-06, "loss": 0.5796, "step": 10855 }, { "epoch": 1.29, "learning_rate": 3.1722376347117735e-06, "loss": 0.5293, "step": 10856 }, { "epoch": 1.29, "learning_rate": 3.171928043913349e-06, "loss": 0.573, "step": 10857 }, { "epoch": 1.29, "learning_rate": 3.1716184420078978e-06, "loss": 0.7686, "step": 10858 }, { "epoch": 1.29, "learning_rate": 3.1713088290005376e-06, "loss": 0.7291, "step": 10859 }, { "epoch": 1.29, "learning_rate": 3.170999204896386e-06, "loss": 0.5103, "step": 10860 }, { "epoch": 1.29, "learning_rate": 3.1706895697005608e-06, "loss": 0.5308, "step": 10861 }, { "epoch": 1.29, "learning_rate": 3.1703799234181813e-06, "loss": 0.4919, "step": 10862 }, { "epoch": 1.29, "learning_rate": 3.1700702660543654e-06, "loss": 0.5117, "step": 10863 }, { "epoch": 1.29, "learning_rate": 3.1697605976142317e-06, "loss": 0.629, "step": 10864 }, { "epoch": 1.29, "learning_rate": 3.1694509181028988e-06, "loss": 0.8312, "step": 10865 }, { "epoch": 1.29, "learning_rate": 3.169141227525487e-06, "loss": 0.4407, "step": 10866 }, { "epoch": 1.29, "learning_rate": 3.168831525887113e-06, "loss": 0.5726, "step": 10867 }, { "epoch": 1.29, "learning_rate": 3.1685218131928987e-06, "loss": 0.5426, "step": 10868 }, { "epoch": 1.29, "learning_rate": 3.168212089447963e-06, "loss": 0.5899, "step": 10869 }, { "epoch": 1.29, "learning_rate": 3.167902354657425e-06, "loss": 0.6931, "step": 10870 }, { "epoch": 1.29, "learning_rate": 3.1675926088264054e-06, "loss": 0.7834, "step": 10871 }, { "epoch": 1.29, "learning_rate": 3.167282851960023e-06, "loss": 0.7963, "step": 10872 }, { "epoch": 1.29, "learning_rate": 3.1669730840633993e-06, "loss": 0.4169, "step": 10873 }, { "epoch": 1.29, "learning_rate": 3.1666633051416555e-06, "loss": 0.5012, "step": 10874 }, { "epoch": 1.3, "learning_rate": 3.1663535151999107e-06, "loss": 0.5713, "step": 10875 }, { "epoch": 1.3, "learning_rate": 3.1660437142432863e-06, "loss": 0.6332, "step": 10876 }, { "epoch": 1.3, "learning_rate": 3.165733902276904e-06, "loss": 0.7518, "step": 10877 }, { "epoch": 1.3, "learning_rate": 3.165424079305884e-06, "loss": 0.916, "step": 10878 }, { "epoch": 1.3, "learning_rate": 3.165114245335348e-06, "loss": 0.4835, "step": 10879 }, { "epoch": 1.3, "learning_rate": 3.164804400370418e-06, "loss": 0.6024, "step": 10880 }, { "epoch": 1.3, "learning_rate": 3.1644945444162155e-06, "loss": 0.4983, "step": 10881 }, { "epoch": 1.3, "learning_rate": 3.1641846774778616e-06, "loss": 0.6348, "step": 10882 }, { "epoch": 1.3, "learning_rate": 3.16387479956048e-06, "loss": 0.7704, "step": 10883 }, { "epoch": 1.3, "learning_rate": 3.163564910669193e-06, "loss": 0.8174, "step": 10884 }, { "epoch": 1.3, "learning_rate": 3.163255010809121e-06, "loss": 0.5558, "step": 10885 }, { "epoch": 1.3, "learning_rate": 3.162945099985389e-06, "loss": 0.4852, "step": 10886 }, { "epoch": 1.3, "learning_rate": 3.162635178203118e-06, "loss": 0.5076, "step": 10887 }, { "epoch": 1.3, "learning_rate": 3.162325245467433e-06, "loss": 0.6388, "step": 10888 }, { "epoch": 1.3, "learning_rate": 3.162015301783455e-06, "loss": 0.6952, "step": 10889 }, { "epoch": 1.3, "learning_rate": 3.1617053471563087e-06, "loss": 0.8531, "step": 10890 }, { "epoch": 1.3, "learning_rate": 3.161395381591118e-06, "loss": 0.688, "step": 10891 }, { "epoch": 1.3, "learning_rate": 3.1610854050930063e-06, "loss": 0.4425, "step": 10892 }, { "epoch": 1.3, "learning_rate": 3.1607754176670964e-06, "loss": 0.5211, "step": 10893 }, { "epoch": 1.3, "learning_rate": 3.160465419318514e-06, "loss": 0.5652, "step": 10894 }, { "epoch": 1.3, "learning_rate": 3.1601554100523827e-06, "loss": 0.7061, "step": 10895 }, { "epoch": 1.3, "learning_rate": 3.1598453898738276e-06, "loss": 0.764, "step": 10896 }, { "epoch": 1.3, "learning_rate": 3.1595353587879716e-06, "loss": 0.8081, "step": 10897 }, { "epoch": 1.3, "learning_rate": 3.1592253167999427e-06, "loss": 0.4287, "step": 10898 }, { "epoch": 1.3, "learning_rate": 3.158915263914863e-06, "loss": 0.5673, "step": 10899 }, { "epoch": 1.3, "learning_rate": 3.158605200137858e-06, "loss": 0.6923, "step": 10900 }, { "epoch": 1.3, "learning_rate": 3.1582951254740546e-06, "loss": 0.6218, "step": 10901 }, { "epoch": 1.3, "learning_rate": 3.1579850399285774e-06, "loss": 0.8217, "step": 10902 }, { "epoch": 1.3, "learning_rate": 3.1576749435065525e-06, "loss": 0.8926, "step": 10903 }, { "epoch": 1.3, "learning_rate": 3.157364836213105e-06, "loss": 0.3279, "step": 10904 }, { "epoch": 1.3, "learning_rate": 3.157054718053362e-06, "loss": 0.5371, "step": 10905 }, { "epoch": 1.3, "learning_rate": 3.156744589032449e-06, "loss": 0.5052, "step": 10906 }, { "epoch": 1.3, "learning_rate": 3.156434449155493e-06, "loss": 0.7037, "step": 10907 }, { "epoch": 1.3, "learning_rate": 3.1561242984276203e-06, "loss": 0.7252, "step": 10908 }, { "epoch": 1.3, "learning_rate": 3.1558141368539587e-06, "loss": 0.8007, "step": 10909 }, { "epoch": 1.3, "learning_rate": 3.1555039644396335e-06, "loss": 0.5121, "step": 10910 }, { "epoch": 1.3, "learning_rate": 3.155193781189773e-06, "loss": 0.4931, "step": 10911 }, { "epoch": 1.3, "learning_rate": 3.154883587109505e-06, "loss": 0.647, "step": 10912 }, { "epoch": 1.3, "learning_rate": 3.154573382203956e-06, "loss": 0.6431, "step": 10913 }, { "epoch": 1.3, "learning_rate": 3.154263166478253e-06, "loss": 0.6469, "step": 10914 }, { "epoch": 1.3, "learning_rate": 3.1539529399375264e-06, "loss": 0.8835, "step": 10915 }, { "epoch": 1.3, "learning_rate": 3.1536427025869022e-06, "loss": 0.4582, "step": 10916 }, { "epoch": 1.3, "learning_rate": 3.1533324544315092e-06, "loss": 0.5081, "step": 10917 }, { "epoch": 1.3, "learning_rate": 3.1530221954764762e-06, "loss": 0.5357, "step": 10918 }, { "epoch": 1.3, "learning_rate": 3.152711925726931e-06, "loss": 0.534, "step": 10919 }, { "epoch": 1.3, "learning_rate": 3.152401645188004e-06, "loss": 0.6776, "step": 10920 }, { "epoch": 1.3, "learning_rate": 3.1520913538648225e-06, "loss": 0.8707, "step": 10921 }, { "epoch": 1.3, "learning_rate": 3.1517810517625156e-06, "loss": 0.814, "step": 10922 }, { "epoch": 1.3, "learning_rate": 3.1514707388862135e-06, "loss": 0.5462, "step": 10923 }, { "epoch": 1.3, "learning_rate": 3.1511604152410458e-06, "loss": 0.4742, "step": 10924 }, { "epoch": 1.3, "learning_rate": 3.150850080832141e-06, "loss": 0.5291, "step": 10925 }, { "epoch": 1.3, "learning_rate": 3.1505397356646307e-06, "loss": 0.6814, "step": 10926 }, { "epoch": 1.3, "learning_rate": 3.150229379743644e-06, "loss": 0.8011, "step": 10927 }, { "epoch": 1.3, "learning_rate": 3.1499190130743106e-06, "loss": 0.8629, "step": 10928 }, { "epoch": 1.3, "learning_rate": 3.149608635661761e-06, "loss": 0.5271, "step": 10929 }, { "epoch": 1.3, "learning_rate": 3.1492982475111274e-06, "loss": 0.6617, "step": 10930 }, { "epoch": 1.3, "learning_rate": 3.148987848627539e-06, "loss": 0.4789, "step": 10931 }, { "epoch": 1.3, "learning_rate": 3.148677439016126e-06, "loss": 0.6375, "step": 10932 }, { "epoch": 1.3, "learning_rate": 3.1483670186820214e-06, "loss": 0.7444, "step": 10933 }, { "epoch": 1.3, "learning_rate": 3.148056587630355e-06, "loss": 0.8179, "step": 10934 }, { "epoch": 1.3, "learning_rate": 3.1477461458662597e-06, "loss": 0.474, "step": 10935 }, { "epoch": 1.3, "learning_rate": 3.1474356933948646e-06, "loss": 0.3778, "step": 10936 }, { "epoch": 1.3, "learning_rate": 3.1471252302213045e-06, "loss": 0.6475, "step": 10937 }, { "epoch": 1.3, "learning_rate": 3.1468147563507096e-06, "loss": 0.569, "step": 10938 }, { "epoch": 1.3, "learning_rate": 3.1465042717882126e-06, "loss": 0.6873, "step": 10939 }, { "epoch": 1.3, "learning_rate": 3.1461937765389454e-06, "loss": 0.8629, "step": 10940 }, { "epoch": 1.3, "learning_rate": 3.145883270608042e-06, "loss": 0.6406, "step": 10941 }, { "epoch": 1.3, "learning_rate": 3.1455727540006332e-06, "loss": 0.5852, "step": 10942 }, { "epoch": 1.3, "learning_rate": 3.1452622267218522e-06, "loss": 0.5485, "step": 10943 }, { "epoch": 1.3, "learning_rate": 3.1449516887768326e-06, "loss": 0.5789, "step": 10944 }, { "epoch": 1.3, "learning_rate": 3.144641140170708e-06, "loss": 0.5739, "step": 10945 }, { "epoch": 1.3, "learning_rate": 3.144330580908611e-06, "loss": 0.829, "step": 10946 }, { "epoch": 1.3, "learning_rate": 3.144020010995675e-06, "loss": 0.7701, "step": 10947 }, { "epoch": 1.3, "learning_rate": 3.1437094304370353e-06, "loss": 0.4643, "step": 10948 }, { "epoch": 1.3, "learning_rate": 3.143398839237824e-06, "loss": 0.5395, "step": 10949 }, { "epoch": 1.3, "learning_rate": 3.143088237403176e-06, "loss": 0.5436, "step": 10950 }, { "epoch": 1.3, "learning_rate": 3.142777624938225e-06, "loss": 0.7125, "step": 10951 }, { "epoch": 1.3, "learning_rate": 3.142467001848107e-06, "loss": 0.8264, "step": 10952 }, { "epoch": 1.3, "learning_rate": 3.142156368137954e-06, "loss": 1.0292, "step": 10953 }, { "epoch": 1.3, "learning_rate": 3.1418457238129035e-06, "loss": 0.453, "step": 10954 }, { "epoch": 1.3, "learning_rate": 3.1415350688780886e-06, "loss": 0.489, "step": 10955 }, { "epoch": 1.3, "learning_rate": 3.1412244033386465e-06, "loss": 0.6597, "step": 10956 }, { "epoch": 1.3, "learning_rate": 3.1409137271997094e-06, "loss": 0.6011, "step": 10957 }, { "epoch": 1.3, "learning_rate": 3.140603040466416e-06, "loss": 0.7498, "step": 10958 }, { "epoch": 1.31, "learning_rate": 3.1402923431439004e-06, "loss": 0.7922, "step": 10959 }, { "epoch": 1.31, "learning_rate": 3.139981635237298e-06, "loss": 0.4784, "step": 10960 }, { "epoch": 1.31, "learning_rate": 3.1396709167517453e-06, "loss": 0.3871, "step": 10961 }, { "epoch": 1.31, "learning_rate": 3.1393601876923796e-06, "loss": 0.4828, "step": 10962 }, { "epoch": 1.31, "learning_rate": 3.139049448064336e-06, "loss": 0.654, "step": 10963 }, { "epoch": 1.31, "learning_rate": 3.138738697872751e-06, "loss": 0.7057, "step": 10964 }, { "epoch": 1.31, "learning_rate": 3.138427937122763e-06, "loss": 0.8596, "step": 10965 }, { "epoch": 1.31, "learning_rate": 3.1381171658195065e-06, "loss": 0.5031, "step": 10966 }, { "epoch": 1.31, "learning_rate": 3.13780638396812e-06, "loss": 0.5542, "step": 10967 }, { "epoch": 1.31, "learning_rate": 3.1374955915737397e-06, "loss": 0.4666, "step": 10968 }, { "epoch": 1.31, "learning_rate": 3.1371847886415047e-06, "loss": 0.6266, "step": 10969 }, { "epoch": 1.31, "learning_rate": 3.1368739751765514e-06, "loss": 0.5472, "step": 10970 }, { "epoch": 1.31, "learning_rate": 3.1365631511840183e-06, "loss": 0.7587, "step": 10971 }, { "epoch": 1.31, "learning_rate": 3.136252316669042e-06, "loss": 0.8357, "step": 10972 }, { "epoch": 1.31, "learning_rate": 3.1359414716367618e-06, "loss": 0.4055, "step": 10973 }, { "epoch": 1.31, "learning_rate": 3.1356306160923166e-06, "loss": 0.5479, "step": 10974 }, { "epoch": 1.31, "learning_rate": 3.1353197500408432e-06, "loss": 0.5651, "step": 10975 }, { "epoch": 1.31, "learning_rate": 3.1350088734874805e-06, "loss": 0.6985, "step": 10976 }, { "epoch": 1.31, "learning_rate": 3.1346979864373696e-06, "loss": 0.8138, "step": 10977 }, { "epoch": 1.31, "learning_rate": 3.1343870888956456e-06, "loss": 0.8489, "step": 10978 }, { "epoch": 1.31, "learning_rate": 3.1340761808674513e-06, "loss": 0.39, "step": 10979 }, { "epoch": 1.31, "learning_rate": 3.133765262357924e-06, "loss": 0.5441, "step": 10980 }, { "epoch": 1.31, "learning_rate": 3.133454333372204e-06, "loss": 0.6399, "step": 10981 }, { "epoch": 1.31, "learning_rate": 3.1331433939154303e-06, "loss": 0.6421, "step": 10982 }, { "epoch": 1.31, "learning_rate": 3.1328324439927433e-06, "loss": 0.7141, "step": 10983 }, { "epoch": 1.31, "learning_rate": 3.1325214836092833e-06, "loss": 0.7258, "step": 10984 }, { "epoch": 1.31, "learning_rate": 3.1322105127701896e-06, "loss": 0.476, "step": 10985 }, { "epoch": 1.31, "learning_rate": 3.1318995314806032e-06, "loss": 0.5415, "step": 10986 }, { "epoch": 1.31, "learning_rate": 3.131588539745664e-06, "loss": 0.6141, "step": 10987 }, { "epoch": 1.31, "learning_rate": 3.1312775375705145e-06, "loss": 0.5634, "step": 10988 }, { "epoch": 1.31, "learning_rate": 3.1309665249602934e-06, "loss": 0.7538, "step": 10989 }, { "epoch": 1.31, "learning_rate": 3.130655501920143e-06, "loss": 0.7799, "step": 10990 }, { "epoch": 1.31, "learning_rate": 3.1303444684552043e-06, "loss": 0.5703, "step": 10991 }, { "epoch": 1.31, "learning_rate": 3.130033424570619e-06, "loss": 0.4339, "step": 10992 }, { "epoch": 1.31, "learning_rate": 3.129722370271527e-06, "loss": 0.6158, "step": 10993 }, { "epoch": 1.31, "learning_rate": 3.1294113055630722e-06, "loss": 0.5829, "step": 10994 }, { "epoch": 1.31, "learning_rate": 3.1291002304503967e-06, "loss": 0.6703, "step": 10995 }, { "epoch": 1.31, "learning_rate": 3.12878914493864e-06, "loss": 0.8638, "step": 10996 }, { "epoch": 1.31, "learning_rate": 3.1284780490329463e-06, "loss": 0.8304, "step": 10997 }, { "epoch": 1.31, "learning_rate": 3.1281669427384587e-06, "loss": 0.3537, "step": 10998 }, { "epoch": 1.31, "learning_rate": 3.1278558260603183e-06, "loss": 0.4609, "step": 10999 }, { "epoch": 1.31, "learning_rate": 3.1275446990036674e-06, "loss": 0.5345, "step": 11000 }, { "epoch": 1.31, "learning_rate": 3.127233561573651e-06, "loss": 0.6876, "step": 11001 }, { "epoch": 1.31, "learning_rate": 3.126922413775411e-06, "loss": 0.71, "step": 11002 }, { "epoch": 1.31, "learning_rate": 3.1266112556140914e-06, "loss": 0.7354, "step": 11003 }, { "epoch": 1.31, "learning_rate": 3.126300087094834e-06, "loss": 0.3965, "step": 11004 }, { "epoch": 1.31, "learning_rate": 3.1259889082227847e-06, "loss": 0.4784, "step": 11005 }, { "epoch": 1.31, "learning_rate": 3.125677719003086e-06, "loss": 0.5665, "step": 11006 }, { "epoch": 1.31, "learning_rate": 3.1253665194408815e-06, "loss": 0.5681, "step": 11007 }, { "epoch": 1.31, "learning_rate": 3.1250553095413162e-06, "loss": 0.7748, "step": 11008 }, { "epoch": 1.31, "learning_rate": 3.1247440893095347e-06, "loss": 0.7847, "step": 11009 }, { "epoch": 1.31, "learning_rate": 3.1244328587506805e-06, "loss": 0.4563, "step": 11010 }, { "epoch": 1.31, "learning_rate": 3.1241216178698993e-06, "loss": 0.4987, "step": 11011 }, { "epoch": 1.31, "learning_rate": 3.123810366672335e-06, "loss": 0.7015, "step": 11012 }, { "epoch": 1.31, "learning_rate": 3.123499105163133e-06, "loss": 0.5862, "step": 11013 }, { "epoch": 1.31, "learning_rate": 3.1231878333474388e-06, "loss": 0.661, "step": 11014 }, { "epoch": 1.31, "learning_rate": 3.1228765512303967e-06, "loss": 0.8237, "step": 11015 }, { "epoch": 1.31, "learning_rate": 3.1225652588171534e-06, "loss": 0.5554, "step": 11016 }, { "epoch": 1.31, "learning_rate": 3.122253956112854e-06, "loss": 0.5302, "step": 11017 }, { "epoch": 1.31, "learning_rate": 3.1219426431226448e-06, "loss": 0.5231, "step": 11018 }, { "epoch": 1.31, "learning_rate": 3.1216313198516714e-06, "loss": 0.5712, "step": 11019 }, { "epoch": 1.31, "learning_rate": 3.1213199863050805e-06, "loss": 0.6877, "step": 11020 }, { "epoch": 1.31, "learning_rate": 3.1210086424880173e-06, "loss": 0.7936, "step": 11021 }, { "epoch": 1.31, "learning_rate": 3.120697288405629e-06, "loss": 0.8572, "step": 11022 }, { "epoch": 1.31, "learning_rate": 3.120385924063063e-06, "loss": 0.4118, "step": 11023 }, { "epoch": 1.31, "learning_rate": 3.120074549465466e-06, "loss": 0.4918, "step": 11024 }, { "epoch": 1.31, "learning_rate": 3.1197631646179838e-06, "loss": 0.4641, "step": 11025 }, { "epoch": 1.31, "learning_rate": 3.1194517695257646e-06, "loss": 0.6498, "step": 11026 }, { "epoch": 1.31, "learning_rate": 3.119140364193957e-06, "loss": 0.83, "step": 11027 }, { "epoch": 1.31, "learning_rate": 3.1188289486277057e-06, "loss": 0.8947, "step": 11028 }, { "epoch": 1.31, "learning_rate": 3.11851752283216e-06, "loss": 0.5394, "step": 11029 }, { "epoch": 1.31, "learning_rate": 3.118206086812468e-06, "loss": 0.4684, "step": 11030 }, { "epoch": 1.31, "learning_rate": 3.1178946405737787e-06, "loss": 0.4865, "step": 11031 }, { "epoch": 1.31, "learning_rate": 3.117583184121237e-06, "loss": 0.6779, "step": 11032 }, { "epoch": 1.31, "learning_rate": 3.1172717174599952e-06, "loss": 0.6719, "step": 11033 }, { "epoch": 1.31, "learning_rate": 3.116960240595199e-06, "loss": 0.8029, "step": 11034 }, { "epoch": 1.31, "learning_rate": 3.1166487535319995e-06, "loss": 0.5878, "step": 11035 }, { "epoch": 1.31, "learning_rate": 3.116337256275543e-06, "loss": 0.4228, "step": 11036 }, { "epoch": 1.31, "learning_rate": 3.116025748830981e-06, "loss": 0.5608, "step": 11037 }, { "epoch": 1.31, "learning_rate": 3.115714231203461e-06, "loss": 0.6657, "step": 11038 }, { "epoch": 1.31, "learning_rate": 3.115402703398134e-06, "loss": 0.6967, "step": 11039 }, { "epoch": 1.31, "learning_rate": 3.115091165420147e-06, "loss": 0.8669, "step": 11040 }, { "epoch": 1.31, "learning_rate": 3.114779617274653e-06, "loss": 0.4831, "step": 11041 }, { "epoch": 1.31, "learning_rate": 3.1144680589668004e-06, "loss": 0.5428, "step": 11042 }, { "epoch": 1.32, "learning_rate": 3.114156490501738e-06, "loss": 0.5314, "step": 11043 }, { "epoch": 1.32, "learning_rate": 3.113844911884618e-06, "loss": 0.6143, "step": 11044 }, { "epoch": 1.32, "learning_rate": 3.11353332312059e-06, "loss": 0.6364, "step": 11045 }, { "epoch": 1.32, "learning_rate": 3.113221724214805e-06, "loss": 0.8097, "step": 11046 }, { "epoch": 1.32, "learning_rate": 3.112910115172412e-06, "loss": 0.875, "step": 11047 }, { "epoch": 1.32, "learning_rate": 3.112598495998565e-06, "loss": 0.4294, "step": 11048 }, { "epoch": 1.32, "learning_rate": 3.112286866698413e-06, "loss": 0.524, "step": 11049 }, { "epoch": 1.32, "learning_rate": 3.111975227277108e-06, "loss": 0.608, "step": 11050 }, { "epoch": 1.32, "learning_rate": 3.1116635777398004e-06, "loss": 0.6979, "step": 11051 }, { "epoch": 1.32, "learning_rate": 3.1113519180916433e-06, "loss": 0.8362, "step": 11052 }, { "epoch": 1.32, "learning_rate": 3.1110402483377873e-06, "loss": 0.773, "step": 11053 }, { "epoch": 1.32, "learning_rate": 3.1107285684833843e-06, "loss": 0.4924, "step": 11054 }, { "epoch": 1.32, "learning_rate": 3.1104168785335875e-06, "loss": 0.6172, "step": 11055 }, { "epoch": 1.32, "learning_rate": 3.1101051784935488e-06, "loss": 0.5687, "step": 11056 }, { "epoch": 1.32, "learning_rate": 3.1097934683684194e-06, "loss": 0.6343, "step": 11057 }, { "epoch": 1.32, "learning_rate": 3.109481748163353e-06, "loss": 0.8212, "step": 11058 }, { "epoch": 1.32, "learning_rate": 3.1091700178835026e-06, "loss": 0.8411, "step": 11059 }, { "epoch": 1.32, "learning_rate": 3.1088582775340206e-06, "loss": 0.528, "step": 11060 }, { "epoch": 1.32, "learning_rate": 3.10854652712006e-06, "loss": 0.4803, "step": 11061 }, { "epoch": 1.32, "learning_rate": 3.108234766646774e-06, "loss": 0.554, "step": 11062 }, { "epoch": 1.32, "learning_rate": 3.1079229961193168e-06, "loss": 0.6237, "step": 11063 }, { "epoch": 1.32, "learning_rate": 3.1076112155428412e-06, "loss": 0.673, "step": 11064 }, { "epoch": 1.32, "learning_rate": 3.1072994249225014e-06, "loss": 0.8346, "step": 11065 }, { "epoch": 1.32, "learning_rate": 3.1069876242634506e-06, "loss": 0.6568, "step": 11066 }, { "epoch": 1.32, "learning_rate": 3.106675813570844e-06, "loss": 0.4613, "step": 11067 }, { "epoch": 1.32, "learning_rate": 3.1063639928498348e-06, "loss": 0.5568, "step": 11068 }, { "epoch": 1.32, "learning_rate": 3.1060521621055785e-06, "loss": 0.6512, "step": 11069 }, { "epoch": 1.32, "learning_rate": 3.105740321343229e-06, "loss": 0.5652, "step": 11070 }, { "epoch": 1.32, "learning_rate": 3.105428470567941e-06, "loss": 0.7034, "step": 11071 }, { "epoch": 1.32, "learning_rate": 3.105116609784869e-06, "loss": 0.8002, "step": 11072 }, { "epoch": 1.32, "learning_rate": 3.1048047389991693e-06, "loss": 0.4907, "step": 11073 }, { "epoch": 1.32, "learning_rate": 3.104492858215997e-06, "loss": 0.6187, "step": 11074 }, { "epoch": 1.32, "learning_rate": 3.1041809674405066e-06, "loss": 0.5726, "step": 11075 }, { "epoch": 1.32, "learning_rate": 3.1038690666778536e-06, "loss": 0.6133, "step": 11076 }, { "epoch": 1.32, "learning_rate": 3.103557155933195e-06, "loss": 0.8404, "step": 11077 }, { "epoch": 1.32, "learning_rate": 3.103245235211686e-06, "loss": 0.9206, "step": 11078 }, { "epoch": 1.32, "learning_rate": 3.1029333045184817e-06, "loss": 0.383, "step": 11079 }, { "epoch": 1.32, "learning_rate": 3.10262136385874e-06, "loss": 0.5454, "step": 11080 }, { "epoch": 1.32, "learning_rate": 3.1023094132376165e-06, "loss": 0.5353, "step": 11081 }, { "epoch": 1.32, "learning_rate": 3.101997452660268e-06, "loss": 0.6074, "step": 11082 }, { "epoch": 1.32, "learning_rate": 3.1016854821318504e-06, "loss": 0.7474, "step": 11083 }, { "epoch": 1.32, "learning_rate": 3.1013735016575224e-06, "loss": 0.8152, "step": 11084 }, { "epoch": 1.32, "learning_rate": 3.101061511242439e-06, "loss": 0.3604, "step": 11085 }, { "epoch": 1.32, "learning_rate": 3.100749510891759e-06, "loss": 0.4878, "step": 11086 }, { "epoch": 1.32, "learning_rate": 3.1004375006106384e-06, "loss": 0.5478, "step": 11087 }, { "epoch": 1.32, "learning_rate": 3.100125480404237e-06, "loss": 0.7187, "step": 11088 }, { "epoch": 1.32, "learning_rate": 3.0998134502777093e-06, "loss": 0.7249, "step": 11089 }, { "epoch": 1.32, "learning_rate": 3.099501410236216e-06, "loss": 0.864, "step": 11090 }, { "epoch": 1.32, "learning_rate": 3.0991893602849138e-06, "loss": 0.6769, "step": 11091 }, { "epoch": 1.32, "learning_rate": 3.098877300428962e-06, "loss": 0.6191, "step": 11092 }, { "epoch": 1.32, "learning_rate": 3.098565230673517e-06, "loss": 0.4853, "step": 11093 }, { "epoch": 1.32, "learning_rate": 3.0982531510237386e-06, "loss": 0.5641, "step": 11094 }, { "epoch": 1.32, "learning_rate": 3.0979410614847864e-06, "loss": 0.6956, "step": 11095 }, { "epoch": 1.32, "learning_rate": 3.0976289620618173e-06, "loss": 0.7997, "step": 11096 }, { "epoch": 1.32, "learning_rate": 3.0973168527599915e-06, "loss": 0.8331, "step": 11097 }, { "epoch": 1.32, "learning_rate": 3.0970047335844683e-06, "loss": 0.5334, "step": 11098 }, { "epoch": 1.32, "learning_rate": 3.096692604540407e-06, "loss": 0.4736, "step": 11099 }, { "epoch": 1.32, "learning_rate": 3.096380465632966e-06, "loss": 0.6699, "step": 11100 }, { "epoch": 1.32, "learning_rate": 3.096068316867307e-06, "loss": 0.6516, "step": 11101 }, { "epoch": 1.32, "learning_rate": 3.0957561582485873e-06, "loss": 0.8137, "step": 11102 }, { "epoch": 1.32, "learning_rate": 3.0954439897819692e-06, "loss": 0.8001, "step": 11103 }, { "epoch": 1.32, "learning_rate": 3.0951318114726115e-06, "loss": 0.4953, "step": 11104 }, { "epoch": 1.32, "learning_rate": 3.0948196233256754e-06, "loss": 0.5109, "step": 11105 }, { "epoch": 1.32, "learning_rate": 3.094507425346322e-06, "loss": 0.6048, "step": 11106 }, { "epoch": 1.32, "learning_rate": 3.0941952175397094e-06, "loss": 0.545, "step": 11107 }, { "epoch": 1.32, "learning_rate": 3.093882999911001e-06, "loss": 0.7779, "step": 11108 }, { "epoch": 1.32, "learning_rate": 3.0935707724653563e-06, "loss": 0.8701, "step": 11109 }, { "epoch": 1.32, "learning_rate": 3.093258535207937e-06, "loss": 0.4523, "step": 11110 }, { "epoch": 1.32, "learning_rate": 3.0929462881439044e-06, "loss": 0.4837, "step": 11111 }, { "epoch": 1.32, "learning_rate": 3.09263403127842e-06, "loss": 0.5514, "step": 11112 }, { "epoch": 1.32, "learning_rate": 3.092321764616645e-06, "loss": 0.6251, "step": 11113 }, { "epoch": 1.32, "learning_rate": 3.092009488163742e-06, "loss": 0.7247, "step": 11114 }, { "epoch": 1.32, "learning_rate": 3.0916972019248713e-06, "loss": 0.8821, "step": 11115 }, { "epoch": 1.32, "learning_rate": 3.091384905905197e-06, "loss": 0.6109, "step": 11116 }, { "epoch": 1.32, "learning_rate": 3.091072600109881e-06, "loss": 0.5478, "step": 11117 }, { "epoch": 1.32, "learning_rate": 3.0907602845440853e-06, "loss": 0.5173, "step": 11118 }, { "epoch": 1.32, "learning_rate": 3.0904479592129717e-06, "loss": 0.6018, "step": 11119 }, { "epoch": 1.32, "learning_rate": 3.090135624121705e-06, "loss": 0.7738, "step": 11120 }, { "epoch": 1.32, "learning_rate": 3.0898232792754457e-06, "loss": 0.7896, "step": 11121 }, { "epoch": 1.32, "learning_rate": 3.0895109246793586e-06, "loss": 0.8776, "step": 11122 }, { "epoch": 1.32, "learning_rate": 3.0891985603386064e-06, "loss": 0.4727, "step": 11123 }, { "epoch": 1.32, "learning_rate": 3.088886186258353e-06, "loss": 0.5079, "step": 11124 }, { "epoch": 1.32, "learning_rate": 3.088573802443762e-06, "loss": 0.4681, "step": 11125 }, { "epoch": 1.32, "learning_rate": 3.0882614088999944e-06, "loss": 0.6107, "step": 11126 }, { "epoch": 1.33, "learning_rate": 3.087949005632218e-06, "loss": 0.7651, "step": 11127 }, { "epoch": 1.33, "learning_rate": 3.0876365926455954e-06, "loss": 0.8783, "step": 11128 }, { "epoch": 1.33, "learning_rate": 3.0873241699452904e-06, "loss": 0.455, "step": 11129 }, { "epoch": 1.33, "learning_rate": 3.0870117375364666e-06, "loss": 0.4973, "step": 11130 }, { "epoch": 1.33, "learning_rate": 3.086699295424291e-06, "loss": 0.5382, "step": 11131 }, { "epoch": 1.33, "learning_rate": 3.086386843613926e-06, "loss": 0.5015, "step": 11132 }, { "epoch": 1.33, "learning_rate": 3.086074382110538e-06, "loss": 0.6591, "step": 11133 }, { "epoch": 1.33, "learning_rate": 3.0857619109192906e-06, "loss": 0.7927, "step": 11134 }, { "epoch": 1.33, "learning_rate": 3.08544943004535e-06, "loss": 0.5385, "step": 11135 }, { "epoch": 1.33, "learning_rate": 3.08513693949388e-06, "loss": 0.4808, "step": 11136 }, { "epoch": 1.33, "learning_rate": 3.0848244392700484e-06, "loss": 0.6061, "step": 11137 }, { "epoch": 1.33, "learning_rate": 3.0845119293790204e-06, "loss": 0.5678, "step": 11138 }, { "epoch": 1.33, "learning_rate": 3.08419940982596e-06, "loss": 0.6767, "step": 11139 }, { "epoch": 1.33, "learning_rate": 3.083886880616035e-06, "loss": 0.8619, "step": 11140 }, { "epoch": 1.33, "learning_rate": 3.0835743417544107e-06, "loss": 0.5654, "step": 11141 }, { "epoch": 1.33, "learning_rate": 3.083261793246253e-06, "loss": 0.4719, "step": 11142 }, { "epoch": 1.33, "learning_rate": 3.082949235096729e-06, "loss": 0.4435, "step": 11143 }, { "epoch": 1.33, "learning_rate": 3.082636667311006e-06, "loss": 0.6205, "step": 11144 }, { "epoch": 1.33, "learning_rate": 3.0823240898942497e-06, "loss": 0.7108, "step": 11145 }, { "epoch": 1.33, "learning_rate": 3.082011502851627e-06, "loss": 0.9098, "step": 11146 }, { "epoch": 1.33, "learning_rate": 3.0816989061883053e-06, "loss": 0.9427, "step": 11147 }, { "epoch": 1.33, "learning_rate": 3.0813862999094523e-06, "loss": 0.4127, "step": 11148 }, { "epoch": 1.33, "learning_rate": 3.081073684020235e-06, "loss": 0.6846, "step": 11149 }, { "epoch": 1.33, "learning_rate": 3.080761058525821e-06, "loss": 0.626, "step": 11150 }, { "epoch": 1.33, "learning_rate": 3.080448423431377e-06, "loss": 0.5927, "step": 11151 }, { "epoch": 1.33, "learning_rate": 3.0801357787420732e-06, "loss": 0.7396, "step": 11152 }, { "epoch": 1.33, "learning_rate": 3.0798231244630748e-06, "loss": 0.7861, "step": 11153 }, { "epoch": 1.33, "learning_rate": 3.079510460599553e-06, "loss": 0.4796, "step": 11154 }, { "epoch": 1.33, "learning_rate": 3.0791977871566742e-06, "loss": 0.4955, "step": 11155 }, { "epoch": 1.33, "learning_rate": 3.078885104139607e-06, "loss": 0.5696, "step": 11156 }, { "epoch": 1.33, "learning_rate": 3.078572411553521e-06, "loss": 0.7241, "step": 11157 }, { "epoch": 1.33, "learning_rate": 3.0782597094035838e-06, "loss": 0.7798, "step": 11158 }, { "epoch": 1.33, "learning_rate": 3.0779469976949655e-06, "loss": 0.9255, "step": 11159 }, { "epoch": 1.33, "learning_rate": 3.077634276432835e-06, "loss": 0.3352, "step": 11160 }, { "epoch": 1.33, "learning_rate": 3.077321545622361e-06, "loss": 0.5515, "step": 11161 }, { "epoch": 1.33, "learning_rate": 3.0770088052687136e-06, "loss": 0.6491, "step": 11162 }, { "epoch": 1.33, "learning_rate": 3.0766960553770625e-06, "loss": 0.7183, "step": 11163 }, { "epoch": 1.33, "learning_rate": 3.076383295952577e-06, "loss": 0.8231, "step": 11164 }, { "epoch": 1.33, "learning_rate": 3.076070527000427e-06, "loss": 0.8735, "step": 11165 }, { "epoch": 1.33, "learning_rate": 3.0757577485257833e-06, "loss": 0.4404, "step": 11166 }, { "epoch": 1.33, "learning_rate": 3.0754449605338153e-06, "loss": 0.4456, "step": 11167 }, { "epoch": 1.33, "learning_rate": 3.075132163029694e-06, "loss": 0.5272, "step": 11168 }, { "epoch": 1.33, "learning_rate": 3.0748193560185895e-06, "loss": 0.5675, "step": 11169 }, { "epoch": 1.33, "learning_rate": 3.0745065395056732e-06, "loss": 0.7245, "step": 11170 }, { "epoch": 1.33, "learning_rate": 3.0741937134961158e-06, "loss": 0.8022, "step": 11171 }, { "epoch": 1.33, "learning_rate": 3.073880877995088e-06, "loss": 0.8849, "step": 11172 }, { "epoch": 1.33, "learning_rate": 3.0735680330077606e-06, "loss": 0.3621, "step": 11173 }, { "epoch": 1.33, "learning_rate": 3.073255178539307e-06, "loss": 0.6105, "step": 11174 }, { "epoch": 1.33, "learning_rate": 3.0729423145948956e-06, "loss": 0.6055, "step": 11175 }, { "epoch": 1.33, "learning_rate": 3.0726294411796997e-06, "loss": 0.6233, "step": 11176 }, { "epoch": 1.33, "learning_rate": 3.0723165582988916e-06, "loss": 0.7588, "step": 11177 }, { "epoch": 1.33, "learning_rate": 3.0720036659576428e-06, "loss": 0.8076, "step": 11178 }, { "epoch": 1.33, "learning_rate": 3.071690764161125e-06, "loss": 0.5226, "step": 11179 }, { "epoch": 1.33, "learning_rate": 3.0713778529145115e-06, "loss": 0.5441, "step": 11180 }, { "epoch": 1.33, "learning_rate": 3.071064932222974e-06, "loss": 0.6294, "step": 11181 }, { "epoch": 1.33, "learning_rate": 3.0707520020916854e-06, "loss": 0.6003, "step": 11182 }, { "epoch": 1.33, "learning_rate": 3.070439062525817e-06, "loss": 0.8105, "step": 11183 }, { "epoch": 1.33, "learning_rate": 3.070126113530545e-06, "loss": 0.9913, "step": 11184 }, { "epoch": 1.33, "learning_rate": 3.06981315511104e-06, "loss": 0.5178, "step": 11185 }, { "epoch": 1.33, "learning_rate": 3.0695001872724743e-06, "loss": 0.4231, "step": 11186 }, { "epoch": 1.33, "learning_rate": 3.0691872100200245e-06, "loss": 0.5533, "step": 11187 }, { "epoch": 1.33, "learning_rate": 3.0688742233588613e-06, "loss": 0.5321, "step": 11188 }, { "epoch": 1.33, "learning_rate": 3.06856122729416e-06, "loss": 0.5408, "step": 11189 }, { "epoch": 1.33, "learning_rate": 3.068248221831093e-06, "loss": 0.7893, "step": 11190 }, { "epoch": 1.33, "learning_rate": 3.0679352069748357e-06, "loss": 0.5147, "step": 11191 }, { "epoch": 1.33, "learning_rate": 3.0676221827305624e-06, "loss": 0.4884, "step": 11192 }, { "epoch": 1.33, "learning_rate": 3.0673091491034463e-06, "loss": 0.4861, "step": 11193 }, { "epoch": 1.33, "learning_rate": 3.066996106098662e-06, "loss": 0.6579, "step": 11194 }, { "epoch": 1.33, "learning_rate": 3.0666830537213854e-06, "loss": 0.6083, "step": 11195 }, { "epoch": 1.33, "learning_rate": 3.0663699919767896e-06, "loss": 0.7265, "step": 11196 }, { "epoch": 1.33, "learning_rate": 3.066056920870051e-06, "loss": 0.7666, "step": 11197 }, { "epoch": 1.33, "learning_rate": 3.0657438404063432e-06, "loss": 0.4884, "step": 11198 }, { "epoch": 1.33, "learning_rate": 3.0654307505908426e-06, "loss": 0.4667, "step": 11199 }, { "epoch": 1.33, "learning_rate": 3.065117651428724e-06, "loss": 0.5967, "step": 11200 }, { "epoch": 1.33, "learning_rate": 3.0648045429251637e-06, "loss": 0.7049, "step": 11201 }, { "epoch": 1.33, "learning_rate": 3.064491425085337e-06, "loss": 0.8434, "step": 11202 }, { "epoch": 1.33, "learning_rate": 3.0641782979144193e-06, "loss": 0.9116, "step": 11203 }, { "epoch": 1.33, "learning_rate": 3.063865161417587e-06, "loss": 0.4788, "step": 11204 }, { "epoch": 1.33, "learning_rate": 3.063552015600016e-06, "loss": 0.4724, "step": 11205 }, { "epoch": 1.33, "learning_rate": 3.063238860466884e-06, "loss": 0.6696, "step": 11206 }, { "epoch": 1.33, "learning_rate": 3.062925696023365e-06, "loss": 0.6678, "step": 11207 }, { "epoch": 1.33, "learning_rate": 3.0626125222746377e-06, "loss": 0.667, "step": 11208 }, { "epoch": 1.33, "learning_rate": 3.062299339225878e-06, "loss": 0.812, "step": 11209 }, { "epoch": 1.34, "learning_rate": 3.0619861468822636e-06, "loss": 0.5594, "step": 11210 }, { "epoch": 1.34, "learning_rate": 3.0616729452489704e-06, "loss": 0.4259, "step": 11211 }, { "epoch": 1.34, "learning_rate": 3.061359734331176e-06, "loss": 0.6297, "step": 11212 }, { "epoch": 1.34, "learning_rate": 3.061046514134059e-06, "loss": 0.584, "step": 11213 }, { "epoch": 1.34, "learning_rate": 3.060733284662796e-06, "loss": 0.7798, "step": 11214 }, { "epoch": 1.34, "learning_rate": 3.0604200459225634e-06, "loss": 0.8239, "step": 11215 }, { "epoch": 1.34, "learning_rate": 3.0601067979185417e-06, "loss": 0.3989, "step": 11216 }, { "epoch": 1.34, "learning_rate": 3.0597935406559078e-06, "loss": 0.4326, "step": 11217 }, { "epoch": 1.34, "learning_rate": 3.0594802741398387e-06, "loss": 0.4893, "step": 11218 }, { "epoch": 1.34, "learning_rate": 3.0591669983755145e-06, "loss": 0.5404, "step": 11219 }, { "epoch": 1.34, "learning_rate": 3.058853713368113e-06, "loss": 0.6078, "step": 11220 }, { "epoch": 1.34, "learning_rate": 3.0585404191228124e-06, "loss": 0.6833, "step": 11221 }, { "epoch": 1.34, "learning_rate": 3.058227115644791e-06, "loss": 0.8455, "step": 11222 }, { "epoch": 1.34, "learning_rate": 3.05791380293923e-06, "loss": 0.4899, "step": 11223 }, { "epoch": 1.34, "learning_rate": 3.057600481011306e-06, "loss": 0.5752, "step": 11224 }, { "epoch": 1.34, "learning_rate": 3.0572871498662e-06, "loss": 0.5564, "step": 11225 }, { "epoch": 1.34, "learning_rate": 3.0569738095090895e-06, "loss": 0.6931, "step": 11226 }, { "epoch": 1.34, "learning_rate": 3.056660459945156e-06, "loss": 0.8024, "step": 11227 }, { "epoch": 1.34, "learning_rate": 3.0563471011795785e-06, "loss": 0.8266, "step": 11228 }, { "epoch": 1.34, "learning_rate": 3.0560337332175366e-06, "loss": 0.5154, "step": 11229 }, { "epoch": 1.34, "learning_rate": 3.05572035606421e-06, "loss": 0.6729, "step": 11230 }, { "epoch": 1.34, "learning_rate": 3.055406969724781e-06, "loss": 0.4577, "step": 11231 }, { "epoch": 1.34, "learning_rate": 3.055093574204426e-06, "loss": 0.6156, "step": 11232 }, { "epoch": 1.34, "learning_rate": 3.0547801695083285e-06, "loss": 0.8235, "step": 11233 }, { "epoch": 1.34, "learning_rate": 3.0544667556416685e-06, "loss": 0.7781, "step": 11234 }, { "epoch": 1.34, "learning_rate": 3.054153332609626e-06, "loss": 0.4573, "step": 11235 }, { "epoch": 1.34, "learning_rate": 3.053839900417383e-06, "loss": 0.5723, "step": 11236 }, { "epoch": 1.34, "learning_rate": 3.053526459070119e-06, "loss": 0.5466, "step": 11237 }, { "epoch": 1.34, "learning_rate": 3.053213008573018e-06, "loss": 0.5889, "step": 11238 }, { "epoch": 1.34, "learning_rate": 3.0528995489312577e-06, "loss": 0.6541, "step": 11239 }, { "epoch": 1.34, "learning_rate": 3.0525860801500224e-06, "loss": 0.8689, "step": 11240 }, { "epoch": 1.34, "learning_rate": 3.052272602234493e-06, "loss": 0.5737, "step": 11241 }, { "epoch": 1.34, "learning_rate": 3.0519591151898504e-06, "loss": 0.4808, "step": 11242 }, { "epoch": 1.34, "learning_rate": 3.051645619021277e-06, "loss": 0.5288, "step": 11243 }, { "epoch": 1.34, "learning_rate": 3.0513321137339564e-06, "loss": 0.5813, "step": 11244 }, { "epoch": 1.34, "learning_rate": 3.0510185993330694e-06, "loss": 0.667, "step": 11245 }, { "epoch": 1.34, "learning_rate": 3.050705075823798e-06, "loss": 0.9539, "step": 11246 }, { "epoch": 1.34, "learning_rate": 3.0503915432113262e-06, "loss": 0.9285, "step": 11247 }, { "epoch": 1.34, "learning_rate": 3.0500780015008357e-06, "loss": 0.467, "step": 11248 }, { "epoch": 1.34, "learning_rate": 3.0497644506975105e-06, "loss": 0.4932, "step": 11249 }, { "epoch": 1.34, "learning_rate": 3.0494508908065318e-06, "loss": 0.5992, "step": 11250 }, { "epoch": 1.34, "learning_rate": 3.049137321833084e-06, "loss": 0.7295, "step": 11251 }, { "epoch": 1.34, "learning_rate": 3.0488237437823503e-06, "loss": 0.7688, "step": 11252 }, { "epoch": 1.34, "learning_rate": 3.0485101566595144e-06, "loss": 0.8193, "step": 11253 }, { "epoch": 1.34, "learning_rate": 3.0481965604697582e-06, "loss": 0.4267, "step": 11254 }, { "epoch": 1.34, "learning_rate": 3.0478829552182683e-06, "loss": 0.4607, "step": 11255 }, { "epoch": 1.34, "learning_rate": 3.047569340910227e-06, "loss": 0.595, "step": 11256 }, { "epoch": 1.34, "learning_rate": 3.047255717550818e-06, "loss": 0.6002, "step": 11257 }, { "epoch": 1.34, "learning_rate": 3.046942085145226e-06, "loss": 0.725, "step": 11258 }, { "epoch": 1.34, "learning_rate": 3.046628443698636e-06, "loss": 0.8876, "step": 11259 }, { "epoch": 1.34, "learning_rate": 3.046314793216232e-06, "loss": 0.4, "step": 11260 }, { "epoch": 1.34, "learning_rate": 3.0460011337031986e-06, "loss": 0.4107, "step": 11261 }, { "epoch": 1.34, "learning_rate": 3.04568746516472e-06, "loss": 0.6093, "step": 11262 }, { "epoch": 1.34, "learning_rate": 3.0453737876059826e-06, "loss": 0.679, "step": 11263 }, { "epoch": 1.34, "learning_rate": 3.04506010103217e-06, "loss": 0.6387, "step": 11264 }, { "epoch": 1.34, "learning_rate": 3.0447464054484687e-06, "loss": 0.8008, "step": 11265 }, { "epoch": 1.34, "learning_rate": 3.0444327008600638e-06, "loss": 0.6652, "step": 11266 }, { "epoch": 1.34, "learning_rate": 3.0441189872721404e-06, "loss": 0.4185, "step": 11267 }, { "epoch": 1.34, "learning_rate": 3.043805264689884e-06, "loss": 0.6065, "step": 11268 }, { "epoch": 1.34, "learning_rate": 3.043491533118481e-06, "loss": 0.6073, "step": 11269 }, { "epoch": 1.34, "learning_rate": 3.0431777925631184e-06, "loss": 0.791, "step": 11270 }, { "epoch": 1.34, "learning_rate": 3.0428640430289806e-06, "loss": 0.8139, "step": 11271 }, { "epoch": 1.34, "learning_rate": 3.0425502845212547e-06, "loss": 0.9142, "step": 11272 }, { "epoch": 1.34, "learning_rate": 3.0422365170451266e-06, "loss": 0.5058, "step": 11273 }, { "epoch": 1.34, "learning_rate": 3.0419227406057844e-06, "loss": 0.4473, "step": 11274 }, { "epoch": 1.34, "learning_rate": 3.0416089552084127e-06, "loss": 0.5228, "step": 11275 }, { "epoch": 1.34, "learning_rate": 3.0412951608582005e-06, "loss": 0.6971, "step": 11276 }, { "epoch": 1.34, "learning_rate": 3.0409813575603337e-06, "loss": 0.7842, "step": 11277 }, { "epoch": 1.34, "learning_rate": 3.040667545319999e-06, "loss": 0.9492, "step": 11278 }, { "epoch": 1.34, "learning_rate": 3.0403537241423845e-06, "loss": 0.4206, "step": 11279 }, { "epoch": 1.34, "learning_rate": 3.040039894032678e-06, "loss": 0.5574, "step": 11280 }, { "epoch": 1.34, "learning_rate": 3.0397260549960677e-06, "loss": 0.611, "step": 11281 }, { "epoch": 1.34, "learning_rate": 3.0394122070377387e-06, "loss": 0.6654, "step": 11282 }, { "epoch": 1.34, "learning_rate": 3.039098350162881e-06, "loss": 0.7081, "step": 11283 }, { "epoch": 1.34, "learning_rate": 3.0387844843766832e-06, "loss": 0.9007, "step": 11284 }, { "epoch": 1.34, "learning_rate": 3.0384706096843325e-06, "loss": 0.4075, "step": 11285 }, { "epoch": 1.34, "learning_rate": 3.0381567260910166e-06, "loss": 0.4953, "step": 11286 }, { "epoch": 1.34, "learning_rate": 3.0378428336019255e-06, "loss": 0.5932, "step": 11287 }, { "epoch": 1.34, "learning_rate": 3.0375289322222475e-06, "loss": 0.7331, "step": 11288 }, { "epoch": 1.34, "learning_rate": 3.0372150219571706e-06, "loss": 0.7211, "step": 11289 }, { "epoch": 1.34, "learning_rate": 3.0369011028118837e-06, "loss": 0.8937, "step": 11290 }, { "epoch": 1.34, "learning_rate": 3.0365871747915774e-06, "loss": 0.7247, "step": 11291 }, { "epoch": 1.34, "learning_rate": 3.0362732379014403e-06, "loss": 0.5062, "step": 11292 }, { "epoch": 1.34, "learning_rate": 3.035959292146661e-06, "loss": 0.5798, "step": 11293 }, { "epoch": 1.35, "learning_rate": 3.0356453375324295e-06, "loss": 0.6941, "step": 11294 }, { "epoch": 1.35, "learning_rate": 3.0353313740639367e-06, "loss": 0.696, "step": 11295 }, { "epoch": 1.35, "learning_rate": 3.0350174017463704e-06, "loss": 0.7244, "step": 11296 }, { "epoch": 1.35, "learning_rate": 3.034703420584922e-06, "loss": 0.9075, "step": 11297 }, { "epoch": 1.35, "learning_rate": 3.034389430584781e-06, "loss": 0.4058, "step": 11298 }, { "epoch": 1.35, "learning_rate": 3.034075431751138e-06, "loss": 0.5106, "step": 11299 }, { "epoch": 1.35, "learning_rate": 3.033761424089183e-06, "loss": 0.6263, "step": 11300 }, { "epoch": 1.35, "learning_rate": 3.0334474076041066e-06, "loss": 0.5911, "step": 11301 }, { "epoch": 1.35, "learning_rate": 3.0331333823011007e-06, "loss": 0.7699, "step": 11302 }, { "epoch": 1.35, "learning_rate": 3.0328193481853544e-06, "loss": 0.9609, "step": 11303 }, { "epoch": 1.35, "learning_rate": 3.03250530526206e-06, "loss": 0.4914, "step": 11304 }, { "epoch": 1.35, "learning_rate": 3.032191253536408e-06, "loss": 0.5203, "step": 11305 }, { "epoch": 1.35, "learning_rate": 3.031877193013591e-06, "loss": 0.5594, "step": 11306 }, { "epoch": 1.35, "learning_rate": 3.031563123698798e-06, "loss": 0.6002, "step": 11307 }, { "epoch": 1.35, "learning_rate": 3.031249045597222e-06, "loss": 0.7705, "step": 11308 }, { "epoch": 1.35, "learning_rate": 3.0309349587140556e-06, "loss": 1.0187, "step": 11309 }, { "epoch": 1.35, "learning_rate": 3.03062086305449e-06, "loss": 0.4437, "step": 11310 }, { "epoch": 1.35, "learning_rate": 3.0303067586237154e-06, "loss": 0.4926, "step": 11311 }, { "epoch": 1.35, "learning_rate": 3.0299926454269267e-06, "loss": 0.4943, "step": 11312 }, { "epoch": 1.35, "learning_rate": 3.029678523469315e-06, "loss": 0.6741, "step": 11313 }, { "epoch": 1.35, "learning_rate": 3.029364392756073e-06, "loss": 0.8242, "step": 11314 }, { "epoch": 1.35, "learning_rate": 3.0290502532923926e-06, "loss": 0.9167, "step": 11315 }, { "epoch": 1.35, "learning_rate": 3.0287361050834674e-06, "loss": 0.5975, "step": 11316 }, { "epoch": 1.35, "learning_rate": 3.0284219481344907e-06, "loss": 0.363, "step": 11317 }, { "epoch": 1.35, "learning_rate": 3.0281077824506536e-06, "loss": 0.5519, "step": 11318 }, { "epoch": 1.35, "learning_rate": 3.027793608037152e-06, "loss": 0.5602, "step": 11319 }, { "epoch": 1.35, "learning_rate": 3.027479424899177e-06, "loss": 0.6166, "step": 11320 }, { "epoch": 1.35, "learning_rate": 3.0271652330419226e-06, "loss": 0.8834, "step": 11321 }, { "epoch": 1.35, "learning_rate": 3.0268510324705823e-06, "loss": 0.8785, "step": 11322 }, { "epoch": 1.35, "learning_rate": 3.0265368231903508e-06, "loss": 0.447, "step": 11323 }, { "epoch": 1.35, "learning_rate": 3.026222605206422e-06, "loss": 0.4299, "step": 11324 }, { "epoch": 1.35, "learning_rate": 3.0259083785239884e-06, "loss": 0.5307, "step": 11325 }, { "epoch": 1.35, "learning_rate": 3.0255941431482455e-06, "loss": 0.5924, "step": 11326 }, { "epoch": 1.35, "learning_rate": 3.0252798990843875e-06, "loss": 0.82, "step": 11327 }, { "epoch": 1.35, "learning_rate": 3.0249656463376086e-06, "loss": 0.7506, "step": 11328 }, { "epoch": 1.35, "learning_rate": 3.024651384913104e-06, "loss": 0.4186, "step": 11329 }, { "epoch": 1.35, "learning_rate": 3.0243371148160673e-06, "loss": 0.5245, "step": 11330 }, { "epoch": 1.35, "learning_rate": 3.0240228360516944e-06, "loss": 0.542, "step": 11331 }, { "epoch": 1.35, "learning_rate": 3.02370854862518e-06, "loss": 0.5403, "step": 11332 }, { "epoch": 1.35, "learning_rate": 3.023394252541719e-06, "loss": 0.7229, "step": 11333 }, { "epoch": 1.35, "learning_rate": 3.0230799478065076e-06, "loss": 0.9247, "step": 11334 }, { "epoch": 1.35, "learning_rate": 3.0227656344247406e-06, "loss": 0.4936, "step": 11335 }, { "epoch": 1.35, "learning_rate": 3.022451312401614e-06, "loss": 0.5985, "step": 11336 }, { "epoch": 1.35, "learning_rate": 3.0221369817423224e-06, "loss": 0.5178, "step": 11337 }, { "epoch": 1.35, "learning_rate": 3.0218226424520642e-06, "loss": 0.5654, "step": 11338 }, { "epoch": 1.35, "learning_rate": 3.021508294536033e-06, "loss": 0.7637, "step": 11339 }, { "epoch": 1.35, "learning_rate": 3.021193937999426e-06, "loss": 0.8569, "step": 11340 }, { "epoch": 1.35, "learning_rate": 3.0208795728474393e-06, "loss": 0.7255, "step": 11341 }, { "epoch": 1.35, "learning_rate": 3.0205651990852704e-06, "loss": 0.5336, "step": 11342 }, { "epoch": 1.35, "learning_rate": 3.0202508167181137e-06, "loss": 0.6991, "step": 11343 }, { "epoch": 1.35, "learning_rate": 3.0199364257511687e-06, "loss": 0.6685, "step": 11344 }, { "epoch": 1.35, "learning_rate": 3.0196220261896303e-06, "loss": 0.6409, "step": 11345 }, { "epoch": 1.35, "learning_rate": 3.0193076180386966e-06, "loss": 0.8107, "step": 11346 }, { "epoch": 1.35, "learning_rate": 3.018993201303564e-06, "loss": 0.8772, "step": 11347 }, { "epoch": 1.35, "learning_rate": 3.01867877598943e-06, "loss": 0.344, "step": 11348 }, { "epoch": 1.35, "learning_rate": 3.018364342101493e-06, "loss": 0.49, "step": 11349 }, { "epoch": 1.35, "learning_rate": 3.018049899644949e-06, "loss": 0.5866, "step": 11350 }, { "epoch": 1.35, "learning_rate": 3.017735448624998e-06, "loss": 0.5925, "step": 11351 }, { "epoch": 1.35, "learning_rate": 3.0174209890468358e-06, "loss": 0.735, "step": 11352 }, { "epoch": 1.35, "learning_rate": 3.0171065209156615e-06, "loss": 0.8135, "step": 11353 }, { "epoch": 1.35, "learning_rate": 3.016792044236672e-06, "loss": 0.4072, "step": 11354 }, { "epoch": 1.35, "learning_rate": 3.016477559015068e-06, "loss": 0.5475, "step": 11355 }, { "epoch": 1.35, "learning_rate": 3.016163065256046e-06, "loss": 0.5207, "step": 11356 }, { "epoch": 1.35, "learning_rate": 3.0158485629648054e-06, "loss": 0.7261, "step": 11357 }, { "epoch": 1.35, "learning_rate": 3.0155340521465444e-06, "loss": 0.7719, "step": 11358 }, { "epoch": 1.35, "learning_rate": 3.0152195328064627e-06, "loss": 0.8396, "step": 11359 }, { "epoch": 1.35, "learning_rate": 3.014905004949759e-06, "loss": 0.4328, "step": 11360 }, { "epoch": 1.35, "learning_rate": 3.0145904685816317e-06, "loss": 0.45, "step": 11361 }, { "epoch": 1.35, "learning_rate": 3.0142759237072815e-06, "loss": 0.5585, "step": 11362 }, { "epoch": 1.35, "learning_rate": 3.013961370331907e-06, "loss": 0.7045, "step": 11363 }, { "epoch": 1.35, "learning_rate": 3.0136468084607074e-06, "loss": 0.8154, "step": 11364 }, { "epoch": 1.35, "learning_rate": 3.0133322380988833e-06, "loss": 0.8242, "step": 11365 }, { "epoch": 1.35, "learning_rate": 3.013017659251634e-06, "loss": 0.5986, "step": 11366 }, { "epoch": 1.35, "learning_rate": 3.01270307192416e-06, "loss": 0.5277, "step": 11367 }, { "epoch": 1.35, "learning_rate": 3.0123884761216613e-06, "loss": 0.5909, "step": 11368 }, { "epoch": 1.35, "learning_rate": 3.0120738718493374e-06, "loss": 0.6745, "step": 11369 }, { "epoch": 1.35, "learning_rate": 3.01175925911239e-06, "loss": 0.6428, "step": 11370 }, { "epoch": 1.35, "learning_rate": 3.0114446379160197e-06, "loss": 0.7262, "step": 11371 }, { "epoch": 1.35, "learning_rate": 3.011130008265426e-06, "loss": 0.8734, "step": 11372 }, { "epoch": 1.35, "learning_rate": 3.0108153701658098e-06, "loss": 0.4608, "step": 11373 }, { "epoch": 1.35, "learning_rate": 3.010500723622374e-06, "loss": 0.5512, "step": 11374 }, { "epoch": 1.35, "learning_rate": 3.0101860686403174e-06, "loss": 0.5732, "step": 11375 }, { "epoch": 1.35, "learning_rate": 3.0098714052248425e-06, "loss": 0.5971, "step": 11376 }, { "epoch": 1.35, "learning_rate": 3.0095567333811504e-06, "loss": 0.688, "step": 11377 }, { "epoch": 1.36, "learning_rate": 3.0092420531144434e-06, "loss": 0.7117, "step": 11378 }, { "epoch": 1.36, "learning_rate": 3.008927364429922e-06, "loss": 0.4748, "step": 11379 }, { "epoch": 1.36, "learning_rate": 3.008612667332788e-06, "loss": 0.7304, "step": 11380 }, { "epoch": 1.36, "learning_rate": 3.0082979618282455e-06, "loss": 0.4943, "step": 11381 }, { "epoch": 1.36, "learning_rate": 3.0079832479214932e-06, "loss": 0.6974, "step": 11382 }, { "epoch": 1.36, "learning_rate": 3.007668525617736e-06, "loss": 0.743, "step": 11383 }, { "epoch": 1.36, "learning_rate": 3.007353794922176e-06, "loss": 0.8779, "step": 11384 }, { "epoch": 1.36, "learning_rate": 3.007039055840015e-06, "loss": 0.5064, "step": 11385 }, { "epoch": 1.36, "learning_rate": 3.0067243083764554e-06, "loss": 0.465, "step": 11386 }, { "epoch": 1.36, "learning_rate": 3.0064095525367005e-06, "loss": 0.6098, "step": 11387 }, { "epoch": 1.36, "learning_rate": 3.006094788325953e-06, "loss": 0.6127, "step": 11388 }, { "epoch": 1.36, "learning_rate": 3.0057800157494168e-06, "loss": 0.7421, "step": 11389 }, { "epoch": 1.36, "learning_rate": 3.0054652348122937e-06, "loss": 0.9641, "step": 11390 }, { "epoch": 1.36, "learning_rate": 3.0051504455197887e-06, "loss": 0.653, "step": 11391 }, { "epoch": 1.36, "learning_rate": 3.0048356478771046e-06, "loss": 0.4771, "step": 11392 }, { "epoch": 1.36, "learning_rate": 3.0045208418894438e-06, "loss": 0.6434, "step": 11393 }, { "epoch": 1.36, "learning_rate": 3.004206027562012e-06, "loss": 0.6025, "step": 11394 }, { "epoch": 1.36, "learning_rate": 3.0038912049000123e-06, "loss": 0.7486, "step": 11395 }, { "epoch": 1.36, "learning_rate": 3.0035763739086487e-06, "loss": 0.8586, "step": 11396 }, { "epoch": 1.36, "learning_rate": 3.0032615345931248e-06, "loss": 0.8927, "step": 11397 }, { "epoch": 1.36, "learning_rate": 3.0029466869586454e-06, "loss": 0.5405, "step": 11398 }, { "epoch": 1.36, "learning_rate": 3.0026318310104158e-06, "loss": 0.5466, "step": 11399 }, { "epoch": 1.36, "learning_rate": 3.0023169667536395e-06, "loss": 0.5644, "step": 11400 }, { "epoch": 1.36, "learning_rate": 3.002002094193521e-06, "loss": 0.6866, "step": 11401 }, { "epoch": 1.36, "learning_rate": 3.001687213335267e-06, "loss": 0.7646, "step": 11402 }, { "epoch": 1.36, "learning_rate": 3.0013723241840807e-06, "loss": 0.8763, "step": 11403 }, { "epoch": 1.36, "learning_rate": 3.0010574267451682e-06, "loss": 0.5385, "step": 11404 }, { "epoch": 1.36, "learning_rate": 3.0007425210237333e-06, "loss": 0.4651, "step": 11405 }, { "epoch": 1.36, "learning_rate": 3.0004276070249836e-06, "loss": 0.569, "step": 11406 }, { "epoch": 1.36, "learning_rate": 3.000112684754123e-06, "loss": 0.6904, "step": 11407 }, { "epoch": 1.36, "learning_rate": 2.9997977542163582e-06, "loss": 0.7722, "step": 11408 }, { "epoch": 1.36, "learning_rate": 2.9994828154168942e-06, "loss": 0.8024, "step": 11409 }, { "epoch": 1.36, "learning_rate": 2.9991678683609376e-06, "loss": 0.5348, "step": 11410 }, { "epoch": 1.36, "learning_rate": 2.9988529130536942e-06, "loss": 0.5473, "step": 11411 }, { "epoch": 1.36, "learning_rate": 2.9985379495003696e-06, "loss": 0.5749, "step": 11412 }, { "epoch": 1.36, "learning_rate": 2.9982229777061715e-06, "loss": 0.6326, "step": 11413 }, { "epoch": 1.36, "learning_rate": 2.997907997676306e-06, "loss": 0.785, "step": 11414 }, { "epoch": 1.36, "learning_rate": 2.9975930094159794e-06, "loss": 0.8753, "step": 11415 }, { "epoch": 1.36, "learning_rate": 2.997278012930398e-06, "loss": 0.541, "step": 11416 }, { "epoch": 1.36, "learning_rate": 2.9969630082247703e-06, "loss": 0.5231, "step": 11417 }, { "epoch": 1.36, "learning_rate": 2.9966479953043016e-06, "loss": 0.5057, "step": 11418 }, { "epoch": 1.36, "learning_rate": 2.9963329741742e-06, "loss": 0.6001, "step": 11419 }, { "epoch": 1.36, "learning_rate": 2.996017944839673e-06, "loss": 0.6726, "step": 11420 }, { "epoch": 1.36, "learning_rate": 2.9957029073059276e-06, "loss": 0.8021, "step": 11421 }, { "epoch": 1.36, "learning_rate": 2.995387861578171e-06, "loss": 0.7661, "step": 11422 }, { "epoch": 1.36, "learning_rate": 2.995072807661612e-06, "loss": 0.4106, "step": 11423 }, { "epoch": 1.36, "learning_rate": 2.9947577455614584e-06, "loss": 0.6753, "step": 11424 }, { "epoch": 1.36, "learning_rate": 2.994442675282917e-06, "loss": 0.5611, "step": 11425 }, { "epoch": 1.36, "learning_rate": 2.9941275968311972e-06, "loss": 0.6634, "step": 11426 }, { "epoch": 1.36, "learning_rate": 2.993812510211506e-06, "loss": 0.7143, "step": 11427 }, { "epoch": 1.36, "learning_rate": 2.9934974154290542e-06, "loss": 0.8145, "step": 11428 }, { "epoch": 1.36, "learning_rate": 2.993182312489047e-06, "loss": 0.4632, "step": 11429 }, { "epoch": 1.36, "learning_rate": 2.9928672013966953e-06, "loss": 0.3569, "step": 11430 }, { "epoch": 1.36, "learning_rate": 2.992552082157208e-06, "loss": 0.6191, "step": 11431 }, { "epoch": 1.36, "learning_rate": 2.9922369547757927e-06, "loss": 0.5761, "step": 11432 }, { "epoch": 1.36, "learning_rate": 2.9919218192576595e-06, "loss": 0.7071, "step": 11433 }, { "epoch": 1.36, "learning_rate": 2.9916066756080177e-06, "loss": 0.8474, "step": 11434 }, { "epoch": 1.36, "learning_rate": 2.9912915238320755e-06, "loss": 0.6433, "step": 11435 }, { "epoch": 1.36, "learning_rate": 2.990976363935044e-06, "loss": 0.4668, "step": 11436 }, { "epoch": 1.36, "learning_rate": 2.990661195922131e-06, "loss": 0.5613, "step": 11437 }, { "epoch": 1.36, "learning_rate": 2.990346019798549e-06, "loss": 0.5272, "step": 11438 }, { "epoch": 1.36, "learning_rate": 2.9900308355695046e-06, "loss": 0.66, "step": 11439 }, { "epoch": 1.36, "learning_rate": 2.9897156432402097e-06, "loss": 0.8652, "step": 11440 }, { "epoch": 1.36, "learning_rate": 2.9894004428158747e-06, "loss": 0.3183, "step": 11441 }, { "epoch": 1.36, "learning_rate": 2.9890852343017084e-06, "loss": 0.4792, "step": 11442 }, { "epoch": 1.36, "learning_rate": 2.9887700177029235e-06, "loss": 0.4521, "step": 11443 }, { "epoch": 1.36, "learning_rate": 2.9884547930247275e-06, "loss": 0.6115, "step": 11444 }, { "epoch": 1.36, "learning_rate": 2.988139560272334e-06, "loss": 0.5857, "step": 11445 }, { "epoch": 1.36, "learning_rate": 2.987824319450952e-06, "loss": 0.8908, "step": 11446 }, { "epoch": 1.36, "learning_rate": 2.9875090705657933e-06, "loss": 0.8464, "step": 11447 }, { "epoch": 1.36, "learning_rate": 2.9871938136220685e-06, "loss": 0.346, "step": 11448 }, { "epoch": 1.36, "learning_rate": 2.9868785486249895e-06, "loss": 0.6687, "step": 11449 }, { "epoch": 1.36, "learning_rate": 2.9865632755797667e-06, "loss": 0.5527, "step": 11450 }, { "epoch": 1.36, "learning_rate": 2.986247994491613e-06, "loss": 0.5706, "step": 11451 }, { "epoch": 1.36, "learning_rate": 2.9859327053657387e-06, "loss": 0.7353, "step": 11452 }, { "epoch": 1.36, "learning_rate": 2.9856174082073565e-06, "loss": 0.8811, "step": 11453 }, { "epoch": 1.36, "learning_rate": 2.9853021030216766e-06, "loss": 0.4512, "step": 11454 }, { "epoch": 1.36, "learning_rate": 2.9849867898139127e-06, "loss": 0.5105, "step": 11455 }, { "epoch": 1.36, "learning_rate": 2.984671468589277e-06, "loss": 0.591, "step": 11456 }, { "epoch": 1.36, "learning_rate": 2.9843561393529806e-06, "loss": 0.5578, "step": 11457 }, { "epoch": 1.36, "learning_rate": 2.9840408021102372e-06, "loss": 0.7979, "step": 11458 }, { "epoch": 1.36, "learning_rate": 2.9837254568662576e-06, "loss": 0.8357, "step": 11459 }, { "epoch": 1.36, "learning_rate": 2.9834101036262576e-06, "loss": 0.4096, "step": 11460 }, { "epoch": 1.36, "learning_rate": 2.983094742395446e-06, "loss": 0.5224, "step": 11461 }, { "epoch": 1.37, "learning_rate": 2.9827793731790387e-06, "loss": 0.5374, "step": 11462 }, { "epoch": 1.37, "learning_rate": 2.9824639959822484e-06, "loss": 0.5266, "step": 11463 }, { "epoch": 1.37, "learning_rate": 2.982148610810287e-06, "loss": 0.5814, "step": 11464 }, { "epoch": 1.37, "learning_rate": 2.9818332176683684e-06, "loss": 0.7898, "step": 11465 }, { "epoch": 1.37, "learning_rate": 2.9815178165617065e-06, "loss": 0.6154, "step": 11466 }, { "epoch": 1.37, "learning_rate": 2.981202407495515e-06, "loss": 0.5504, "step": 11467 }, { "epoch": 1.37, "learning_rate": 2.980886990475007e-06, "loss": 0.4871, "step": 11468 }, { "epoch": 1.37, "learning_rate": 2.980571565505397e-06, "loss": 0.5936, "step": 11469 }, { "epoch": 1.37, "learning_rate": 2.9802561325918983e-06, "loss": 0.6835, "step": 11470 }, { "epoch": 1.37, "learning_rate": 2.9799406917397257e-06, "loss": 0.7053, "step": 11471 }, { "epoch": 1.37, "learning_rate": 2.9796252429540934e-06, "loss": 0.9488, "step": 11472 }, { "epoch": 1.37, "learning_rate": 2.9793097862402153e-06, "loss": 0.4521, "step": 11473 }, { "epoch": 1.37, "learning_rate": 2.9789943216033062e-06, "loss": 0.4849, "step": 11474 }, { "epoch": 1.37, "learning_rate": 2.9786788490485814e-06, "loss": 0.598, "step": 11475 }, { "epoch": 1.37, "learning_rate": 2.978363368581254e-06, "loss": 0.6863, "step": 11476 }, { "epoch": 1.37, "learning_rate": 2.9780478802065403e-06, "loss": 0.7917, "step": 11477 }, { "epoch": 1.37, "learning_rate": 2.9777323839296556e-06, "loss": 0.7151, "step": 11478 }, { "epoch": 1.37, "learning_rate": 2.977416879755814e-06, "loss": 0.4396, "step": 11479 }, { "epoch": 1.37, "learning_rate": 2.977101367690231e-06, "loss": 0.5103, "step": 11480 }, { "epoch": 1.37, "learning_rate": 2.9767858477381236e-06, "loss": 0.5139, "step": 11481 }, { "epoch": 1.37, "learning_rate": 2.976470319904705e-06, "loss": 0.5961, "step": 11482 }, { "epoch": 1.37, "learning_rate": 2.9761547841951923e-06, "loss": 0.7094, "step": 11483 }, { "epoch": 1.37, "learning_rate": 2.975839240614801e-06, "loss": 0.9257, "step": 11484 }, { "epoch": 1.37, "learning_rate": 2.9755236891687476e-06, "loss": 0.4001, "step": 11485 }, { "epoch": 1.37, "learning_rate": 2.975208129862247e-06, "loss": 0.395, "step": 11486 }, { "epoch": 1.37, "learning_rate": 2.9748925627005166e-06, "loss": 0.5304, "step": 11487 }, { "epoch": 1.37, "learning_rate": 2.974576987688772e-06, "loss": 0.5717, "step": 11488 }, { "epoch": 1.37, "learning_rate": 2.9742614048322304e-06, "loss": 0.7306, "step": 11489 }, { "epoch": 1.37, "learning_rate": 2.9739458141361076e-06, "loss": 0.7464, "step": 11490 }, { "epoch": 1.37, "learning_rate": 2.9736302156056207e-06, "loss": 0.4527, "step": 11491 }, { "epoch": 1.37, "learning_rate": 2.9733146092459876e-06, "loss": 0.3987, "step": 11492 }, { "epoch": 1.37, "learning_rate": 2.9729989950624225e-06, "loss": 0.5092, "step": 11493 }, { "epoch": 1.37, "learning_rate": 2.9726833730601458e-06, "loss": 0.5529, "step": 11494 }, { "epoch": 1.37, "learning_rate": 2.9723677432443733e-06, "loss": 0.6782, "step": 11495 }, { "epoch": 1.37, "learning_rate": 2.9720521056203217e-06, "loss": 0.7588, "step": 11496 }, { "epoch": 1.37, "learning_rate": 2.9717364601932086e-06, "loss": 0.8452, "step": 11497 }, { "epoch": 1.37, "learning_rate": 2.971420806968253e-06, "loss": 0.4885, "step": 11498 }, { "epoch": 1.37, "learning_rate": 2.9711051459506722e-06, "loss": 0.4792, "step": 11499 }, { "epoch": 1.37, "learning_rate": 2.9707894771456836e-06, "loss": 0.5681, "step": 11500 }, { "epoch": 1.37, "learning_rate": 2.9704738005585047e-06, "loss": 0.5898, "step": 11501 }, { "epoch": 1.37, "learning_rate": 2.9701581161943556e-06, "loss": 0.8925, "step": 11502 }, { "epoch": 1.37, "learning_rate": 2.9698424240584534e-06, "loss": 0.7337, "step": 11503 }, { "epoch": 1.37, "learning_rate": 2.9695267241560155e-06, "loss": 0.4204, "step": 11504 }, { "epoch": 1.37, "learning_rate": 2.9692110164922617e-06, "loss": 0.51, "step": 11505 }, { "epoch": 1.37, "learning_rate": 2.968895301072411e-06, "loss": 0.5299, "step": 11506 }, { "epoch": 1.37, "learning_rate": 2.968579577901681e-06, "loss": 0.6055, "step": 11507 }, { "epoch": 1.37, "learning_rate": 2.9682638469852907e-06, "loss": 0.7762, "step": 11508 }, { "epoch": 1.37, "learning_rate": 2.967948108328461e-06, "loss": 0.8943, "step": 11509 }, { "epoch": 1.37, "learning_rate": 2.9676323619364095e-06, "loss": 0.4893, "step": 11510 }, { "epoch": 1.37, "learning_rate": 2.9673166078143554e-06, "loss": 0.5431, "step": 11511 }, { "epoch": 1.37, "learning_rate": 2.967000845967519e-06, "loss": 0.6763, "step": 11512 }, { "epoch": 1.37, "learning_rate": 2.966685076401119e-06, "loss": 0.7162, "step": 11513 }, { "epoch": 1.37, "learning_rate": 2.966369299120376e-06, "loss": 0.7339, "step": 11514 }, { "epoch": 1.37, "learning_rate": 2.966053514130509e-06, "loss": 0.7593, "step": 11515 }, { "epoch": 1.37, "learning_rate": 2.9657377214367384e-06, "loss": 0.6381, "step": 11516 }, { "epoch": 1.37, "learning_rate": 2.965421921044285e-06, "loss": 0.5341, "step": 11517 }, { "epoch": 1.37, "learning_rate": 2.9651061129583665e-06, "loss": 0.4359, "step": 11518 }, { "epoch": 1.37, "learning_rate": 2.9647902971842058e-06, "loss": 0.4605, "step": 11519 }, { "epoch": 1.37, "learning_rate": 2.9644744737270226e-06, "loss": 0.6691, "step": 11520 }, { "epoch": 1.37, "learning_rate": 2.9641586425920375e-06, "loss": 0.8061, "step": 11521 }, { "epoch": 1.37, "learning_rate": 2.9638428037844714e-06, "loss": 0.9689, "step": 11522 }, { "epoch": 1.37, "learning_rate": 2.963526957309544e-06, "loss": 0.3458, "step": 11523 }, { "epoch": 1.37, "learning_rate": 2.9632111031724775e-06, "loss": 0.588, "step": 11524 }, { "epoch": 1.37, "learning_rate": 2.9628952413784925e-06, "loss": 0.6375, "step": 11525 }, { "epoch": 1.37, "learning_rate": 2.96257937193281e-06, "loss": 0.6871, "step": 11526 }, { "epoch": 1.37, "learning_rate": 2.9622634948406513e-06, "loss": 0.6641, "step": 11527 }, { "epoch": 1.37, "learning_rate": 2.9619476101072395e-06, "loss": 0.8367, "step": 11528 }, { "epoch": 1.37, "learning_rate": 2.9616317177377944e-06, "loss": 0.4913, "step": 11529 }, { "epoch": 1.37, "learning_rate": 2.9613158177375383e-06, "loss": 0.5326, "step": 11530 }, { "epoch": 1.37, "learning_rate": 2.960999910111693e-06, "loss": 0.5386, "step": 11531 }, { "epoch": 1.37, "learning_rate": 2.960683994865481e-06, "loss": 0.7275, "step": 11532 }, { "epoch": 1.37, "learning_rate": 2.960368072004123e-06, "loss": 0.7061, "step": 11533 }, { "epoch": 1.37, "learning_rate": 2.960052141532842e-06, "loss": 0.8354, "step": 11534 }, { "epoch": 1.37, "learning_rate": 2.959736203456862e-06, "loss": 0.5179, "step": 11535 }, { "epoch": 1.37, "learning_rate": 2.9594202577814024e-06, "loss": 0.5302, "step": 11536 }, { "epoch": 1.37, "learning_rate": 2.9591043045116885e-06, "loss": 0.6561, "step": 11537 }, { "epoch": 1.37, "learning_rate": 2.9587883436529414e-06, "loss": 0.5917, "step": 11538 }, { "epoch": 1.37, "learning_rate": 2.9584723752103843e-06, "loss": 0.6827, "step": 11539 }, { "epoch": 1.37, "learning_rate": 2.9581563991892405e-06, "loss": 0.8107, "step": 11540 }, { "epoch": 1.37, "learning_rate": 2.9578404155947332e-06, "loss": 0.6651, "step": 11541 }, { "epoch": 1.37, "learning_rate": 2.9575244244320856e-06, "loss": 0.4228, "step": 11542 }, { "epoch": 1.37, "learning_rate": 2.95720842570652e-06, "loss": 0.4692, "step": 11543 }, { "epoch": 1.37, "learning_rate": 2.9568924194232612e-06, "loss": 0.5885, "step": 11544 }, { "epoch": 1.37, "learning_rate": 2.956576405587532e-06, "loss": 0.6031, "step": 11545 }, { "epoch": 1.38, "learning_rate": 2.956260384204558e-06, "loss": 0.7709, "step": 11546 }, { "epoch": 1.38, "learning_rate": 2.9559443552795603e-06, "loss": 0.8311, "step": 11547 }, { "epoch": 1.38, "learning_rate": 2.9556283188177643e-06, "loss": 0.5443, "step": 11548 }, { "epoch": 1.38, "learning_rate": 2.9553122748243945e-06, "loss": 0.5335, "step": 11549 }, { "epoch": 1.38, "learning_rate": 2.9549962233046736e-06, "loss": 0.6381, "step": 11550 }, { "epoch": 1.38, "learning_rate": 2.9546801642638278e-06, "loss": 0.7426, "step": 11551 }, { "epoch": 1.38, "learning_rate": 2.9543640977070804e-06, "loss": 0.8322, "step": 11552 }, { "epoch": 1.38, "learning_rate": 2.9540480236396562e-06, "loss": 0.7516, "step": 11553 }, { "epoch": 1.38, "learning_rate": 2.9537319420667807e-06, "loss": 0.4638, "step": 11554 }, { "epoch": 1.38, "learning_rate": 2.9534158529936775e-06, "loss": 0.6102, "step": 11555 }, { "epoch": 1.38, "learning_rate": 2.9530997564255728e-06, "loss": 0.5034, "step": 11556 }, { "epoch": 1.38, "learning_rate": 2.95278365236769e-06, "loss": 0.5987, "step": 11557 }, { "epoch": 1.38, "learning_rate": 2.9524675408252567e-06, "loss": 0.6754, "step": 11558 }, { "epoch": 1.38, "learning_rate": 2.9521514218034956e-06, "loss": 0.8897, "step": 11559 }, { "epoch": 1.38, "learning_rate": 2.9518352953076352e-06, "loss": 0.3647, "step": 11560 }, { "epoch": 1.38, "learning_rate": 2.9515191613428984e-06, "loss": 0.5522, "step": 11561 }, { "epoch": 1.38, "learning_rate": 2.951203019914512e-06, "loss": 0.6984, "step": 11562 }, { "epoch": 1.38, "learning_rate": 2.950886871027702e-06, "loss": 0.6162, "step": 11563 }, { "epoch": 1.38, "learning_rate": 2.9505707146876942e-06, "loss": 0.7551, "step": 11564 }, { "epoch": 1.38, "learning_rate": 2.950254550899715e-06, "loss": 0.8356, "step": 11565 }, { "epoch": 1.38, "learning_rate": 2.9499383796689895e-06, "loss": 0.7575, "step": 11566 }, { "epoch": 1.38, "learning_rate": 2.9496222010007463e-06, "loss": 0.347, "step": 11567 }, { "epoch": 1.38, "learning_rate": 2.949306014900209e-06, "loss": 0.619, "step": 11568 }, { "epoch": 1.38, "learning_rate": 2.948989821372606e-06, "loss": 0.5746, "step": 11569 }, { "epoch": 1.38, "learning_rate": 2.948673620423163e-06, "loss": 0.5686, "step": 11570 }, { "epoch": 1.38, "learning_rate": 2.948357412057109e-06, "loss": 0.8853, "step": 11571 }, { "epoch": 1.38, "learning_rate": 2.9480411962796678e-06, "loss": 0.8163, "step": 11572 }, { "epoch": 1.38, "learning_rate": 2.947724973096069e-06, "loss": 0.4435, "step": 11573 }, { "epoch": 1.38, "learning_rate": 2.947408742511538e-06, "loss": 0.5461, "step": 11574 }, { "epoch": 1.38, "learning_rate": 2.9470925045313036e-06, "loss": 0.6194, "step": 11575 }, { "epoch": 1.38, "learning_rate": 2.946776259160592e-06, "loss": 0.691, "step": 11576 }, { "epoch": 1.38, "learning_rate": 2.9464600064046315e-06, "loss": 0.7853, "step": 11577 }, { "epoch": 1.38, "learning_rate": 2.9461437462686498e-06, "loss": 1.043, "step": 11578 }, { "epoch": 1.38, "learning_rate": 2.945827478757875e-06, "loss": 0.4172, "step": 11579 }, { "epoch": 1.38, "learning_rate": 2.9455112038775334e-06, "loss": 0.532, "step": 11580 }, { "epoch": 1.38, "learning_rate": 2.945194921632856e-06, "loss": 0.6109, "step": 11581 }, { "epoch": 1.38, "learning_rate": 2.944878632029067e-06, "loss": 0.6006, "step": 11582 }, { "epoch": 1.38, "learning_rate": 2.9445623350713985e-06, "loss": 0.7576, "step": 11583 }, { "epoch": 1.38, "learning_rate": 2.944246030765077e-06, "loss": 0.7892, "step": 11584 }, { "epoch": 1.38, "learning_rate": 2.9439297191153314e-06, "loss": 0.4568, "step": 11585 }, { "epoch": 1.38, "learning_rate": 2.94361340012739e-06, "loss": 0.5104, "step": 11586 }, { "epoch": 1.38, "learning_rate": 2.9432970738064814e-06, "loss": 0.4839, "step": 11587 }, { "epoch": 1.38, "learning_rate": 2.9429807401578357e-06, "loss": 0.6318, "step": 11588 }, { "epoch": 1.38, "learning_rate": 2.942664399186681e-06, "loss": 0.7887, "step": 11589 }, { "epoch": 1.38, "learning_rate": 2.9423480508982465e-06, "loss": 0.8326, "step": 11590 }, { "epoch": 1.38, "learning_rate": 2.942031695297762e-06, "loss": 0.448, "step": 11591 }, { "epoch": 1.38, "learning_rate": 2.941715332390456e-06, "loss": 0.5273, "step": 11592 }, { "epoch": 1.38, "learning_rate": 2.9413989621815587e-06, "loss": 0.5221, "step": 11593 }, { "epoch": 1.38, "learning_rate": 2.9410825846762995e-06, "loss": 0.5958, "step": 11594 }, { "epoch": 1.38, "learning_rate": 2.9407661998799086e-06, "loss": 0.7142, "step": 11595 }, { "epoch": 1.38, "learning_rate": 2.940449807797615e-06, "loss": 0.7763, "step": 11596 }, { "epoch": 1.38, "learning_rate": 2.9401334084346487e-06, "loss": 0.9996, "step": 11597 }, { "epoch": 1.38, "learning_rate": 2.9398170017962406e-06, "loss": 0.4757, "step": 11598 }, { "epoch": 1.38, "learning_rate": 2.9395005878876207e-06, "loss": 0.5573, "step": 11599 }, { "epoch": 1.38, "learning_rate": 2.9391841667140192e-06, "loss": 0.5067, "step": 11600 }, { "epoch": 1.38, "learning_rate": 2.938867738280667e-06, "loss": 0.7123, "step": 11601 }, { "epoch": 1.38, "learning_rate": 2.938551302592793e-06, "loss": 0.7584, "step": 11602 }, { "epoch": 1.38, "learning_rate": 2.938234859655631e-06, "loss": 0.7655, "step": 11603 }, { "epoch": 1.38, "learning_rate": 2.9379184094744084e-06, "loss": 0.5762, "step": 11604 }, { "epoch": 1.38, "learning_rate": 2.937601952054358e-06, "loss": 0.5194, "step": 11605 }, { "epoch": 1.38, "learning_rate": 2.937285487400711e-06, "loss": 0.5358, "step": 11606 }, { "epoch": 1.38, "learning_rate": 2.9369690155186988e-06, "loss": 0.6248, "step": 11607 }, { "epoch": 1.38, "learning_rate": 2.9366525364135507e-06, "loss": 0.7824, "step": 11608 }, { "epoch": 1.38, "learning_rate": 2.9363360500905002e-06, "loss": 0.7884, "step": 11609 }, { "epoch": 1.38, "learning_rate": 2.9360195565547783e-06, "loss": 0.4758, "step": 11610 }, { "epoch": 1.38, "learning_rate": 2.9357030558116167e-06, "loss": 0.5784, "step": 11611 }, { "epoch": 1.38, "learning_rate": 2.9353865478662465e-06, "loss": 0.5616, "step": 11612 }, { "epoch": 1.38, "learning_rate": 2.9350700327239e-06, "loss": 0.5235, "step": 11613 }, { "epoch": 1.38, "learning_rate": 2.93475351038981e-06, "loss": 0.7352, "step": 11614 }, { "epoch": 1.38, "learning_rate": 2.934436980869208e-06, "loss": 0.8149, "step": 11615 }, { "epoch": 1.38, "learning_rate": 2.9341204441673267e-06, "loss": 0.6219, "step": 11616 }, { "epoch": 1.38, "learning_rate": 2.9338039002893968e-06, "loss": 0.4159, "step": 11617 }, { "epoch": 1.38, "learning_rate": 2.9334873492406528e-06, "loss": 0.563, "step": 11618 }, { "epoch": 1.38, "learning_rate": 2.9331707910263264e-06, "loss": 0.6522, "step": 11619 }, { "epoch": 1.38, "learning_rate": 2.93285422565165e-06, "loss": 0.6806, "step": 11620 }, { "epoch": 1.38, "learning_rate": 2.9325376531218575e-06, "loss": 0.712, "step": 11621 }, { "epoch": 1.38, "learning_rate": 2.9322210734421814e-06, "loss": 0.9896, "step": 11622 }, { "epoch": 1.38, "learning_rate": 2.931904486617854e-06, "loss": 0.4585, "step": 11623 }, { "epoch": 1.38, "learning_rate": 2.9315878926541103e-06, "loss": 0.5187, "step": 11624 }, { "epoch": 1.38, "learning_rate": 2.9312712915561812e-06, "loss": 0.5487, "step": 11625 }, { "epoch": 1.38, "learning_rate": 2.9309546833293023e-06, "loss": 0.6279, "step": 11626 }, { "epoch": 1.38, "learning_rate": 2.930638067978706e-06, "loss": 0.8376, "step": 11627 }, { "epoch": 1.38, "learning_rate": 2.930321445509627e-06, "loss": 0.7591, "step": 11628 }, { "epoch": 1.38, "learning_rate": 2.9300048159272975e-06, "loss": 0.4026, "step": 11629 }, { "epoch": 1.39, "learning_rate": 2.9296881792369525e-06, "loss": 0.6066, "step": 11630 }, { "epoch": 1.39, "learning_rate": 2.9293715354438262e-06, "loss": 0.5375, "step": 11631 }, { "epoch": 1.39, "learning_rate": 2.9290548845531526e-06, "loss": 0.6992, "step": 11632 }, { "epoch": 1.39, "learning_rate": 2.9287382265701654e-06, "loss": 0.7656, "step": 11633 }, { "epoch": 1.39, "learning_rate": 2.928421561500099e-06, "loss": 0.8458, "step": 11634 }, { "epoch": 1.39, "learning_rate": 2.928104889348189e-06, "loss": 0.4748, "step": 11635 }, { "epoch": 1.39, "learning_rate": 2.927788210119668e-06, "loss": 0.4514, "step": 11636 }, { "epoch": 1.39, "learning_rate": 2.9274715238197734e-06, "loss": 0.5985, "step": 11637 }, { "epoch": 1.39, "learning_rate": 2.9271548304537377e-06, "loss": 0.6603, "step": 11638 }, { "epoch": 1.39, "learning_rate": 2.9268381300267973e-06, "loss": 0.7243, "step": 11639 }, { "epoch": 1.39, "learning_rate": 2.9265214225441863e-06, "loss": 0.8222, "step": 11640 }, { "epoch": 1.39, "learning_rate": 2.926204708011141e-06, "loss": 0.5378, "step": 11641 }, { "epoch": 1.39, "learning_rate": 2.925887986432896e-06, "loss": 0.4313, "step": 11642 }, { "epoch": 1.39, "learning_rate": 2.9255712578146867e-06, "loss": 0.5909, "step": 11643 }, { "epoch": 1.39, "learning_rate": 2.9252545221617483e-06, "loss": 0.6191, "step": 11644 }, { "epoch": 1.39, "learning_rate": 2.924937779479318e-06, "loss": 0.5812, "step": 11645 }, { "epoch": 1.39, "learning_rate": 2.9246210297726294e-06, "loss": 0.7937, "step": 11646 }, { "epoch": 1.39, "learning_rate": 2.924304273046921e-06, "loss": 0.9279, "step": 11647 }, { "epoch": 1.39, "learning_rate": 2.923987509307426e-06, "loss": 0.4082, "step": 11648 }, { "epoch": 1.39, "learning_rate": 2.9236707385593827e-06, "loss": 0.4851, "step": 11649 }, { "epoch": 1.39, "learning_rate": 2.9233539608080263e-06, "loss": 0.6291, "step": 11650 }, { "epoch": 1.39, "learning_rate": 2.923037176058593e-06, "loss": 0.6729, "step": 11651 }, { "epoch": 1.39, "learning_rate": 2.92272038431632e-06, "loss": 0.7188, "step": 11652 }, { "epoch": 1.39, "learning_rate": 2.922403585586444e-06, "loss": 0.8572, "step": 11653 }, { "epoch": 1.39, "learning_rate": 2.922086779874201e-06, "loss": 0.4829, "step": 11654 }, { "epoch": 1.39, "learning_rate": 2.921769967184828e-06, "loss": 0.5104, "step": 11655 }, { "epoch": 1.39, "learning_rate": 2.921453147523562e-06, "loss": 0.6059, "step": 11656 }, { "epoch": 1.39, "learning_rate": 2.92113632089564e-06, "loss": 0.5589, "step": 11657 }, { "epoch": 1.39, "learning_rate": 2.9208194873062995e-06, "loss": 0.6679, "step": 11658 }, { "epoch": 1.39, "learning_rate": 2.9205026467607774e-06, "loss": 0.9056, "step": 11659 }, { "epoch": 1.39, "learning_rate": 2.920185799264312e-06, "loss": 0.374, "step": 11660 }, { "epoch": 1.39, "learning_rate": 2.919868944822139e-06, "loss": 0.3896, "step": 11661 }, { "epoch": 1.39, "learning_rate": 2.9195520834394973e-06, "loss": 0.462, "step": 11662 }, { "epoch": 1.39, "learning_rate": 2.9192352151216253e-06, "loss": 0.7441, "step": 11663 }, { "epoch": 1.39, "learning_rate": 2.9189183398737592e-06, "loss": 0.7143, "step": 11664 }, { "epoch": 1.39, "learning_rate": 2.9186014577011383e-06, "loss": 0.8418, "step": 11665 }, { "epoch": 1.39, "learning_rate": 2.9182845686089993e-06, "loss": 0.5735, "step": 11666 }, { "epoch": 1.39, "learning_rate": 2.917967672602583e-06, "loss": 0.4094, "step": 11667 }, { "epoch": 1.39, "learning_rate": 2.917650769687124e-06, "loss": 0.5827, "step": 11668 }, { "epoch": 1.39, "learning_rate": 2.917333859867864e-06, "loss": 0.6869, "step": 11669 }, { "epoch": 1.39, "learning_rate": 2.9170169431500395e-06, "loss": 0.7189, "step": 11670 }, { "epoch": 1.39, "learning_rate": 2.916700019538891e-06, "loss": 0.8444, "step": 11671 }, { "epoch": 1.39, "learning_rate": 2.9163830890396556e-06, "loss": 0.9127, "step": 11672 }, { "epoch": 1.39, "learning_rate": 2.916066151657573e-06, "loss": 0.4677, "step": 11673 }, { "epoch": 1.39, "learning_rate": 2.9157492073978816e-06, "loss": 0.5756, "step": 11674 }, { "epoch": 1.39, "learning_rate": 2.915432256265821e-06, "loss": 0.5487, "step": 11675 }, { "epoch": 1.39, "learning_rate": 2.9151152982666304e-06, "loss": 0.6924, "step": 11676 }, { "epoch": 1.39, "learning_rate": 2.914798333405549e-06, "loss": 0.9194, "step": 11677 }, { "epoch": 1.39, "learning_rate": 2.914481361687817e-06, "loss": 0.8357, "step": 11678 }, { "epoch": 1.39, "learning_rate": 2.9141643831186728e-06, "loss": 0.4517, "step": 11679 }, { "epoch": 1.39, "learning_rate": 2.9138473977033565e-06, "loss": 0.4988, "step": 11680 }, { "epoch": 1.39, "learning_rate": 2.9135304054471083e-06, "loss": 0.5974, "step": 11681 }, { "epoch": 1.39, "learning_rate": 2.9132134063551677e-06, "loss": 0.6185, "step": 11682 }, { "epoch": 1.39, "learning_rate": 2.9128964004327743e-06, "loss": 0.8401, "step": 11683 }, { "epoch": 1.39, "learning_rate": 2.9125793876851693e-06, "loss": 0.8156, "step": 11684 }, { "epoch": 1.39, "learning_rate": 2.9122623681175922e-06, "loss": 0.4446, "step": 11685 }, { "epoch": 1.39, "learning_rate": 2.911945341735284e-06, "loss": 0.4326, "step": 11686 }, { "epoch": 1.39, "learning_rate": 2.911628308543484e-06, "loss": 0.4537, "step": 11687 }, { "epoch": 1.39, "learning_rate": 2.9113112685474338e-06, "loss": 0.5481, "step": 11688 }, { "epoch": 1.39, "learning_rate": 2.910994221752374e-06, "loss": 0.6828, "step": 11689 }, { "epoch": 1.39, "learning_rate": 2.9106771681635452e-06, "loss": 0.9121, "step": 11690 }, { "epoch": 1.39, "learning_rate": 2.9103601077861875e-06, "loss": 0.6354, "step": 11691 }, { "epoch": 1.39, "learning_rate": 2.9100430406255448e-06, "loss": 0.464, "step": 11692 }, { "epoch": 1.39, "learning_rate": 2.909725966686854e-06, "loss": 0.5009, "step": 11693 }, { "epoch": 1.39, "learning_rate": 2.90940888597536e-06, "loss": 0.5019, "step": 11694 }, { "epoch": 1.39, "learning_rate": 2.9090917984963024e-06, "loss": 0.6992, "step": 11695 }, { "epoch": 1.39, "learning_rate": 2.908774704254923e-06, "loss": 0.6727, "step": 11696 }, { "epoch": 1.39, "learning_rate": 2.9084576032564636e-06, "loss": 0.8343, "step": 11697 }, { "epoch": 1.39, "learning_rate": 2.9081404955061655e-06, "loss": 0.437, "step": 11698 }, { "epoch": 1.39, "learning_rate": 2.907823381009271e-06, "loss": 0.6628, "step": 11699 }, { "epoch": 1.39, "learning_rate": 2.907506259771022e-06, "loss": 0.5359, "step": 11700 }, { "epoch": 1.39, "learning_rate": 2.90718913179666e-06, "loss": 0.5802, "step": 11701 }, { "epoch": 1.39, "learning_rate": 2.9068719970914266e-06, "loss": 0.7997, "step": 11702 }, { "epoch": 1.39, "learning_rate": 2.9065548556605668e-06, "loss": 0.7783, "step": 11703 }, { "epoch": 1.39, "learning_rate": 2.9062377075093195e-06, "loss": 0.406, "step": 11704 }, { "epoch": 1.39, "learning_rate": 2.9059205526429298e-06, "loss": 0.592, "step": 11705 }, { "epoch": 1.39, "learning_rate": 2.9056033910666385e-06, "loss": 0.7066, "step": 11706 }, { "epoch": 1.39, "learning_rate": 2.9052862227856906e-06, "loss": 0.5924, "step": 11707 }, { "epoch": 1.39, "learning_rate": 2.9049690478053257e-06, "loss": 0.7599, "step": 11708 }, { "epoch": 1.39, "learning_rate": 2.9046518661307887e-06, "loss": 0.8308, "step": 11709 }, { "epoch": 1.39, "learning_rate": 2.9043346777673235e-06, "loss": 0.4344, "step": 11710 }, { "epoch": 1.39, "learning_rate": 2.9040174827201717e-06, "loss": 0.3867, "step": 11711 }, { "epoch": 1.39, "learning_rate": 2.9037002809945768e-06, "loss": 0.5203, "step": 11712 }, { "epoch": 1.39, "learning_rate": 2.903383072595782e-06, "loss": 0.5786, "step": 11713 }, { "epoch": 1.4, "learning_rate": 2.903065857529032e-06, "loss": 0.6873, "step": 11714 }, { "epoch": 1.4, "learning_rate": 2.9027486357995687e-06, "loss": 0.8057, "step": 11715 }, { "epoch": 1.4, "learning_rate": 2.902431407412637e-06, "loss": 0.4919, "step": 11716 }, { "epoch": 1.4, "learning_rate": 2.902114172373481e-06, "loss": 0.5205, "step": 11717 }, { "epoch": 1.4, "learning_rate": 2.9017969306873435e-06, "loss": 0.4063, "step": 11718 }, { "epoch": 1.4, "learning_rate": 2.9014796823594684e-06, "loss": 0.6177, "step": 11719 }, { "epoch": 1.4, "learning_rate": 2.901162427395101e-06, "loss": 0.569, "step": 11720 }, { "epoch": 1.4, "learning_rate": 2.9008451657994853e-06, "loss": 0.77, "step": 11721 }, { "epoch": 1.4, "learning_rate": 2.900527897577865e-06, "loss": 0.8448, "step": 11722 }, { "epoch": 1.4, "learning_rate": 2.9002106227354846e-06, "loss": 0.4232, "step": 11723 }, { "epoch": 1.4, "learning_rate": 2.8998933412775903e-06, "loss": 0.6061, "step": 11724 }, { "epoch": 1.4, "learning_rate": 2.8995760532094242e-06, "loss": 0.5858, "step": 11725 }, { "epoch": 1.4, "learning_rate": 2.899258758536233e-06, "loss": 0.6873, "step": 11726 }, { "epoch": 1.4, "learning_rate": 2.898941457263261e-06, "loss": 0.7983, "step": 11727 }, { "epoch": 1.4, "learning_rate": 2.8986241493957536e-06, "loss": 0.9158, "step": 11728 }, { "epoch": 1.4, "learning_rate": 2.8983068349389554e-06, "loss": 0.4391, "step": 11729 }, { "epoch": 1.4, "learning_rate": 2.8979895138981107e-06, "loss": 0.4044, "step": 11730 }, { "epoch": 1.4, "learning_rate": 2.8976721862784667e-06, "loss": 0.6237, "step": 11731 }, { "epoch": 1.4, "learning_rate": 2.8973548520852684e-06, "loss": 0.6734, "step": 11732 }, { "epoch": 1.4, "learning_rate": 2.8970375113237616e-06, "loss": 0.8489, "step": 11733 }, { "epoch": 1.4, "learning_rate": 2.89672016399919e-06, "loss": 0.7233, "step": 11734 }, { "epoch": 1.4, "learning_rate": 2.8964028101168018e-06, "loss": 0.5636, "step": 11735 }, { "epoch": 1.4, "learning_rate": 2.8960854496818413e-06, "loss": 0.4636, "step": 11736 }, { "epoch": 1.4, "learning_rate": 2.8957680826995555e-06, "loss": 0.5715, "step": 11737 }, { "epoch": 1.4, "learning_rate": 2.89545070917519e-06, "loss": 0.4588, "step": 11738 }, { "epoch": 1.4, "learning_rate": 2.895133329113991e-06, "loss": 0.6275, "step": 11739 }, { "epoch": 1.4, "learning_rate": 2.894815942521204e-06, "loss": 0.8777, "step": 11740 }, { "epoch": 1.4, "learning_rate": 2.8944985494020778e-06, "loss": 0.4738, "step": 11741 }, { "epoch": 1.4, "learning_rate": 2.8941811497618573e-06, "loss": 0.4269, "step": 11742 }, { "epoch": 1.4, "learning_rate": 2.893863743605789e-06, "loss": 0.4727, "step": 11743 }, { "epoch": 1.4, "learning_rate": 2.89354633093912e-06, "loss": 0.6543, "step": 11744 }, { "epoch": 1.4, "learning_rate": 2.8932289117670964e-06, "loss": 0.757, "step": 11745 }, { "epoch": 1.4, "learning_rate": 2.892911486094967e-06, "loss": 0.7311, "step": 11746 }, { "epoch": 1.4, "learning_rate": 2.8925940539279775e-06, "loss": 0.8668, "step": 11747 }, { "epoch": 1.4, "learning_rate": 2.8922766152713748e-06, "loss": 0.3229, "step": 11748 }, { "epoch": 1.4, "learning_rate": 2.8919591701304074e-06, "loss": 0.7169, "step": 11749 }, { "epoch": 1.4, "learning_rate": 2.891641718510322e-06, "loss": 0.7458, "step": 11750 }, { "epoch": 1.4, "learning_rate": 2.891324260416365e-06, "loss": 0.5808, "step": 11751 }, { "epoch": 1.4, "learning_rate": 2.8910067958537865e-06, "loss": 0.8922, "step": 11752 }, { "epoch": 1.4, "learning_rate": 2.8906893248278323e-06, "loss": 0.8029, "step": 11753 }, { "epoch": 1.4, "learning_rate": 2.890371847343751e-06, "loss": 0.5048, "step": 11754 }, { "epoch": 1.4, "learning_rate": 2.8900543634067895e-06, "loss": 0.4746, "step": 11755 }, { "epoch": 1.4, "learning_rate": 2.8897368730221976e-06, "loss": 0.4464, "step": 11756 }, { "epoch": 1.4, "learning_rate": 2.8894193761952223e-06, "loss": 0.6132, "step": 11757 }, { "epoch": 1.4, "learning_rate": 2.889101872931112e-06, "loss": 0.6917, "step": 11758 }, { "epoch": 1.4, "learning_rate": 2.8887843632351146e-06, "loss": 0.732, "step": 11759 }, { "epoch": 1.4, "learning_rate": 2.88846684711248e-06, "loss": 0.5078, "step": 11760 }, { "epoch": 1.4, "learning_rate": 2.8881493245684554e-06, "loss": 0.4593, "step": 11761 }, { "epoch": 1.4, "learning_rate": 2.887831795608289e-06, "loss": 0.5776, "step": 11762 }, { "epoch": 1.4, "learning_rate": 2.887514260237232e-06, "loss": 0.6957, "step": 11763 }, { "epoch": 1.4, "learning_rate": 2.8871967184605308e-06, "loss": 0.7547, "step": 11764 }, { "epoch": 1.4, "learning_rate": 2.8868791702834357e-06, "loss": 0.864, "step": 11765 }, { "epoch": 1.4, "learning_rate": 2.886561615711195e-06, "loss": 0.5581, "step": 11766 }, { "epoch": 1.4, "learning_rate": 2.8862440547490595e-06, "loss": 0.3931, "step": 11767 }, { "epoch": 1.4, "learning_rate": 2.8859264874022757e-06, "loss": 0.5971, "step": 11768 }, { "epoch": 1.4, "learning_rate": 2.8856089136760958e-06, "loss": 0.6186, "step": 11769 }, { "epoch": 1.4, "learning_rate": 2.8852913335757676e-06, "loss": 0.5899, "step": 11770 }, { "epoch": 1.4, "learning_rate": 2.884973747106542e-06, "loss": 0.7199, "step": 11771 }, { "epoch": 1.4, "learning_rate": 2.884656154273667e-06, "loss": 0.8347, "step": 11772 }, { "epoch": 1.4, "learning_rate": 2.884338555082395e-06, "loss": 0.4314, "step": 11773 }, { "epoch": 1.4, "learning_rate": 2.8840209495379734e-06, "loss": 0.646, "step": 11774 }, { "epoch": 1.4, "learning_rate": 2.8837033376456536e-06, "loss": 0.5382, "step": 11775 }, { "epoch": 1.4, "learning_rate": 2.883385719410686e-06, "loss": 0.6004, "step": 11776 }, { "epoch": 1.4, "learning_rate": 2.8830680948383187e-06, "loss": 0.837, "step": 11777 }, { "epoch": 1.4, "learning_rate": 2.8827504639338057e-06, "loss": 0.8475, "step": 11778 }, { "epoch": 1.4, "learning_rate": 2.8824328267023936e-06, "loss": 0.5267, "step": 11779 }, { "epoch": 1.4, "learning_rate": 2.8821151831493353e-06, "loss": 0.5066, "step": 11780 }, { "epoch": 1.4, "learning_rate": 2.8817975332798816e-06, "loss": 0.5583, "step": 11781 }, { "epoch": 1.4, "learning_rate": 2.8814798770992824e-06, "loss": 0.5684, "step": 11782 }, { "epoch": 1.4, "learning_rate": 2.881162214612788e-06, "loss": 0.8494, "step": 11783 }, { "epoch": 1.4, "learning_rate": 2.880844545825651e-06, "loss": 0.9321, "step": 11784 }, { "epoch": 1.4, "learning_rate": 2.8805268707431215e-06, "loss": 0.4093, "step": 11785 }, { "epoch": 1.4, "learning_rate": 2.8802091893704513e-06, "loss": 0.4201, "step": 11786 }, { "epoch": 1.4, "learning_rate": 2.87989150171289e-06, "loss": 0.5395, "step": 11787 }, { "epoch": 1.4, "learning_rate": 2.879573807775692e-06, "loss": 0.5811, "step": 11788 }, { "epoch": 1.4, "learning_rate": 2.8792561075641064e-06, "loss": 0.6969, "step": 11789 }, { "epoch": 1.4, "learning_rate": 2.8789384010833853e-06, "loss": 0.8815, "step": 11790 }, { "epoch": 1.4, "learning_rate": 2.8786206883387813e-06, "loss": 0.5002, "step": 11791 }, { "epoch": 1.4, "learning_rate": 2.8783029693355446e-06, "loss": 0.4546, "step": 11792 }, { "epoch": 1.4, "learning_rate": 2.877985244078929e-06, "loss": 0.6315, "step": 11793 }, { "epoch": 1.4, "learning_rate": 2.877667512574185e-06, "loss": 0.5667, "step": 11794 }, { "epoch": 1.4, "learning_rate": 2.877349774826565e-06, "loss": 0.7972, "step": 11795 }, { "epoch": 1.4, "learning_rate": 2.8770320308413223e-06, "loss": 0.7753, "step": 11796 }, { "epoch": 1.4, "learning_rate": 2.876714280623708e-06, "loss": 0.8962, "step": 11797 }, { "epoch": 1.41, "learning_rate": 2.876396524178975e-06, "loss": 0.433, "step": 11798 }, { "epoch": 1.41, "learning_rate": 2.8760787615123764e-06, "loss": 0.6001, "step": 11799 }, { "epoch": 1.41, "learning_rate": 2.8757609926291642e-06, "loss": 0.6326, "step": 11800 }, { "epoch": 1.41, "learning_rate": 2.875443217534591e-06, "loss": 0.6927, "step": 11801 }, { "epoch": 1.41, "learning_rate": 2.87512543623391e-06, "loss": 0.8117, "step": 11802 }, { "epoch": 1.41, "learning_rate": 2.8748076487323745e-06, "loss": 0.834, "step": 11803 }, { "epoch": 1.41, "learning_rate": 2.874489855035236e-06, "loss": 0.508, "step": 11804 }, { "epoch": 1.41, "learning_rate": 2.87417205514775e-06, "loss": 0.5063, "step": 11805 }, { "epoch": 1.41, "learning_rate": 2.8738542490751675e-06, "loss": 0.608, "step": 11806 }, { "epoch": 1.41, "learning_rate": 2.8735364368227436e-06, "loss": 0.6593, "step": 11807 }, { "epoch": 1.41, "learning_rate": 2.873218618395731e-06, "loss": 0.7607, "step": 11808 }, { "epoch": 1.41, "learning_rate": 2.872900793799383e-06, "loss": 0.8205, "step": 11809 }, { "epoch": 1.41, "learning_rate": 2.8725829630389546e-06, "loss": 0.4245, "step": 11810 }, { "epoch": 1.41, "learning_rate": 2.872265126119697e-06, "loss": 0.4544, "step": 11811 }, { "epoch": 1.41, "learning_rate": 2.8719472830468664e-06, "loss": 0.5237, "step": 11812 }, { "epoch": 1.41, "learning_rate": 2.8716294338257158e-06, "loss": 0.6098, "step": 11813 }, { "epoch": 1.41, "learning_rate": 2.8713115784614998e-06, "loss": 0.6608, "step": 11814 }, { "epoch": 1.41, "learning_rate": 2.870993716959472e-06, "loss": 0.8848, "step": 11815 }, { "epoch": 1.41, "learning_rate": 2.870675849324887e-06, "loss": 0.5323, "step": 11816 }, { "epoch": 1.41, "learning_rate": 2.870357975562999e-06, "loss": 0.5068, "step": 11817 }, { "epoch": 1.41, "learning_rate": 2.8700400956790633e-06, "loss": 0.6402, "step": 11818 }, { "epoch": 1.41, "learning_rate": 2.869722209678332e-06, "loss": 0.5182, "step": 11819 }, { "epoch": 1.41, "learning_rate": 2.8694043175660635e-06, "loss": 0.7028, "step": 11820 }, { "epoch": 1.41, "learning_rate": 2.8690864193475103e-06, "loss": 0.7723, "step": 11821 }, { "epoch": 1.41, "learning_rate": 2.8687685150279265e-06, "loss": 0.9777, "step": 11822 }, { "epoch": 1.41, "learning_rate": 2.8684506046125693e-06, "loss": 0.6029, "step": 11823 }, { "epoch": 1.41, "learning_rate": 2.8681326881066924e-06, "loss": 0.6318, "step": 11824 }, { "epoch": 1.41, "learning_rate": 2.8678147655155512e-06, "loss": 0.6538, "step": 11825 }, { "epoch": 1.41, "learning_rate": 2.8674968368444004e-06, "loss": 0.6693, "step": 11826 }, { "epoch": 1.41, "learning_rate": 2.867178902098497e-06, "loss": 0.788, "step": 11827 }, { "epoch": 1.41, "learning_rate": 2.8668609612830955e-06, "loss": 0.9126, "step": 11828 }, { "epoch": 1.41, "learning_rate": 2.866543014403451e-06, "loss": 0.472, "step": 11829 }, { "epoch": 1.41, "learning_rate": 2.8662250614648194e-06, "loss": 0.581, "step": 11830 }, { "epoch": 1.41, "learning_rate": 2.8659071024724577e-06, "loss": 0.4908, "step": 11831 }, { "epoch": 1.41, "learning_rate": 2.865589137431621e-06, "loss": 0.6156, "step": 11832 }, { "epoch": 1.41, "learning_rate": 2.8652711663475644e-06, "loss": 0.7324, "step": 11833 }, { "epoch": 1.41, "learning_rate": 2.8649531892255446e-06, "loss": 0.9096, "step": 11834 }, { "epoch": 1.41, "learning_rate": 2.8646352060708192e-06, "loss": 0.4556, "step": 11835 }, { "epoch": 1.41, "learning_rate": 2.864317216888642e-06, "loss": 0.4529, "step": 11836 }, { "epoch": 1.41, "learning_rate": 2.8639992216842714e-06, "loss": 0.6178, "step": 11837 }, { "epoch": 1.41, "learning_rate": 2.8636812204629626e-06, "loss": 0.6787, "step": 11838 }, { "epoch": 1.41, "learning_rate": 2.863363213229973e-06, "loss": 0.7724, "step": 11839 }, { "epoch": 1.41, "learning_rate": 2.8630451999905594e-06, "loss": 0.886, "step": 11840 }, { "epoch": 1.41, "learning_rate": 2.8627271807499774e-06, "loss": 0.4541, "step": 11841 }, { "epoch": 1.41, "learning_rate": 2.862409155513485e-06, "loss": 0.3863, "step": 11842 }, { "epoch": 1.41, "learning_rate": 2.862091124286339e-06, "loss": 0.5049, "step": 11843 }, { "epoch": 1.41, "learning_rate": 2.8617730870737962e-06, "loss": 0.5748, "step": 11844 }, { "epoch": 1.41, "learning_rate": 2.861455043881114e-06, "loss": 0.6822, "step": 11845 }, { "epoch": 1.41, "learning_rate": 2.8611369947135496e-06, "loss": 0.8579, "step": 11846 }, { "epoch": 1.41, "learning_rate": 2.86081893957636e-06, "loss": 0.8909, "step": 11847 }, { "epoch": 1.41, "learning_rate": 2.8605008784748035e-06, "loss": 0.3826, "step": 11848 }, { "epoch": 1.41, "learning_rate": 2.860182811414137e-06, "loss": 0.5987, "step": 11849 }, { "epoch": 1.41, "learning_rate": 2.8598647383996193e-06, "loss": 0.7029, "step": 11850 }, { "epoch": 1.41, "learning_rate": 2.8595466594365056e-06, "loss": 0.6254, "step": 11851 }, { "epoch": 1.41, "learning_rate": 2.8592285745300565e-06, "loss": 0.8068, "step": 11852 }, { "epoch": 1.41, "learning_rate": 2.85891048368553e-06, "loss": 0.8007, "step": 11853 }, { "epoch": 1.41, "learning_rate": 2.8585923869081813e-06, "loss": 0.3832, "step": 11854 }, { "epoch": 1.41, "learning_rate": 2.8582742842032707e-06, "loss": 0.4664, "step": 11855 }, { "epoch": 1.41, "learning_rate": 2.8579561755760556e-06, "loss": 0.5596, "step": 11856 }, { "epoch": 1.41, "learning_rate": 2.857638061031796e-06, "loss": 0.6595, "step": 11857 }, { "epoch": 1.41, "learning_rate": 2.857319940575749e-06, "loss": 0.7972, "step": 11858 }, { "epoch": 1.41, "learning_rate": 2.8570018142131727e-06, "loss": 0.8674, "step": 11859 }, { "epoch": 1.41, "learning_rate": 2.8566836819493275e-06, "loss": 0.4626, "step": 11860 }, { "epoch": 1.41, "learning_rate": 2.8563655437894706e-06, "loss": 0.3993, "step": 11861 }, { "epoch": 1.41, "learning_rate": 2.856047399738861e-06, "loss": 0.4652, "step": 11862 }, { "epoch": 1.41, "learning_rate": 2.8557292498027583e-06, "loss": 0.5337, "step": 11863 }, { "epoch": 1.41, "learning_rate": 2.8554110939864217e-06, "loss": 0.7109, "step": 11864 }, { "epoch": 1.41, "learning_rate": 2.8550929322951094e-06, "loss": 0.8009, "step": 11865 }, { "epoch": 1.41, "learning_rate": 2.854774764734081e-06, "loss": 0.678, "step": 11866 }, { "epoch": 1.41, "learning_rate": 2.854456591308597e-06, "loss": 0.5318, "step": 11867 }, { "epoch": 1.41, "learning_rate": 2.8541384120239146e-06, "loss": 0.6184, "step": 11868 }, { "epoch": 1.41, "learning_rate": 2.853820226885295e-06, "loss": 0.5067, "step": 11869 }, { "epoch": 1.41, "learning_rate": 2.853502035897997e-06, "loss": 0.5938, "step": 11870 }, { "epoch": 1.41, "learning_rate": 2.853183839067281e-06, "loss": 0.6624, "step": 11871 }, { "epoch": 1.41, "learning_rate": 2.852865636398407e-06, "loss": 0.9286, "step": 11872 }, { "epoch": 1.41, "learning_rate": 2.8525474278966335e-06, "loss": 0.5214, "step": 11873 }, { "epoch": 1.41, "learning_rate": 2.852229213567222e-06, "loss": 0.5821, "step": 11874 }, { "epoch": 1.41, "learning_rate": 2.851910993415432e-06, "loss": 0.606, "step": 11875 }, { "epoch": 1.41, "learning_rate": 2.8515927674465236e-06, "loss": 0.6362, "step": 11876 }, { "epoch": 1.41, "learning_rate": 2.8512745356657566e-06, "loss": 0.8015, "step": 11877 }, { "epoch": 1.41, "learning_rate": 2.8509562980783937e-06, "loss": 0.8189, "step": 11878 }, { "epoch": 1.41, "learning_rate": 2.850638054689693e-06, "loss": 0.3907, "step": 11879 }, { "epoch": 1.41, "learning_rate": 2.8503198055049153e-06, "loss": 0.5483, "step": 11880 }, { "epoch": 1.41, "learning_rate": 2.850001550529322e-06, "loss": 0.5811, "step": 11881 }, { "epoch": 1.42, "learning_rate": 2.849683289768174e-06, "loss": 0.668, "step": 11882 }, { "epoch": 1.42, "learning_rate": 2.8493650232267317e-06, "loss": 0.7291, "step": 11883 }, { "epoch": 1.42, "learning_rate": 2.849046750910256e-06, "loss": 0.7415, "step": 11884 }, { "epoch": 1.42, "learning_rate": 2.8487284728240088e-06, "loss": 0.445, "step": 11885 }, { "epoch": 1.42, "learning_rate": 2.848410188973251e-06, "loss": 0.4935, "step": 11886 }, { "epoch": 1.42, "learning_rate": 2.8480918993632427e-06, "loss": 0.5258, "step": 11887 }, { "epoch": 1.42, "learning_rate": 2.8477736039992458e-06, "loss": 0.6195, "step": 11888 }, { "epoch": 1.42, "learning_rate": 2.847455302886523e-06, "loss": 0.6708, "step": 11889 }, { "epoch": 1.42, "learning_rate": 2.847136996030334e-06, "loss": 0.851, "step": 11890 }, { "epoch": 1.42, "learning_rate": 2.8468186834359426e-06, "loss": 0.4838, "step": 11891 }, { "epoch": 1.42, "learning_rate": 2.846500365108608e-06, "loss": 0.4582, "step": 11892 }, { "epoch": 1.42, "learning_rate": 2.846182041053594e-06, "loss": 0.5883, "step": 11893 }, { "epoch": 1.42, "learning_rate": 2.845863711276161e-06, "loss": 0.6506, "step": 11894 }, { "epoch": 1.42, "learning_rate": 2.8455453757815726e-06, "loss": 0.73, "step": 11895 }, { "epoch": 1.42, "learning_rate": 2.84522703457509e-06, "loss": 0.8238, "step": 11896 }, { "epoch": 1.42, "learning_rate": 2.8449086876619757e-06, "loss": 0.8234, "step": 11897 }, { "epoch": 1.42, "learning_rate": 2.8445903350474915e-06, "loss": 0.3698, "step": 11898 }, { "epoch": 1.42, "learning_rate": 2.8442719767369005e-06, "loss": 0.5378, "step": 11899 }, { "epoch": 1.42, "learning_rate": 2.8439536127354644e-06, "loss": 0.5407, "step": 11900 }, { "epoch": 1.42, "learning_rate": 2.8436352430484467e-06, "loss": 0.6602, "step": 11901 }, { "epoch": 1.42, "learning_rate": 2.8433168676811096e-06, "loss": 0.743, "step": 11902 }, { "epoch": 1.42, "learning_rate": 2.8429984866387157e-06, "loss": 0.8126, "step": 11903 }, { "epoch": 1.42, "learning_rate": 2.842680099926528e-06, "loss": 0.4547, "step": 11904 }, { "epoch": 1.42, "learning_rate": 2.842361707549809e-06, "loss": 0.5506, "step": 11905 }, { "epoch": 1.42, "learning_rate": 2.842043309513823e-06, "loss": 0.5179, "step": 11906 }, { "epoch": 1.42, "learning_rate": 2.841724905823832e-06, "loss": 0.6344, "step": 11907 }, { "epoch": 1.42, "learning_rate": 2.8414064964850994e-06, "loss": 0.6732, "step": 11908 }, { "epoch": 1.42, "learning_rate": 2.841088081502889e-06, "loss": 0.8854, "step": 11909 }, { "epoch": 1.42, "learning_rate": 2.840769660882465e-06, "loss": 0.3899, "step": 11910 }, { "epoch": 1.42, "learning_rate": 2.8404512346290885e-06, "loss": 0.4739, "step": 11911 }, { "epoch": 1.42, "learning_rate": 2.8401328027480253e-06, "loss": 0.5123, "step": 11912 }, { "epoch": 1.42, "learning_rate": 2.8398143652445374e-06, "loss": 0.5606, "step": 11913 }, { "epoch": 1.42, "learning_rate": 2.839495922123891e-06, "loss": 0.6591, "step": 11914 }, { "epoch": 1.42, "learning_rate": 2.8391774733913468e-06, "loss": 0.7801, "step": 11915 }, { "epoch": 1.42, "learning_rate": 2.838859019052172e-06, "loss": 0.5593, "step": 11916 }, { "epoch": 1.42, "learning_rate": 2.8385405591116286e-06, "loss": 0.544, "step": 11917 }, { "epoch": 1.42, "learning_rate": 2.8382220935749814e-06, "loss": 0.5282, "step": 11918 }, { "epoch": 1.42, "learning_rate": 2.837903622447495e-06, "loss": 0.684, "step": 11919 }, { "epoch": 1.42, "learning_rate": 2.837585145734432e-06, "loss": 0.6552, "step": 11920 }, { "epoch": 1.42, "learning_rate": 2.83726666344106e-06, "loss": 0.8043, "step": 11921 }, { "epoch": 1.42, "learning_rate": 2.8369481755726403e-06, "loss": 0.8802, "step": 11922 }, { "epoch": 1.42, "learning_rate": 2.8366296821344398e-06, "loss": 0.5137, "step": 11923 }, { "epoch": 1.42, "learning_rate": 2.8363111831317224e-06, "loss": 0.6423, "step": 11924 }, { "epoch": 1.42, "learning_rate": 2.835992678569753e-06, "loss": 0.5831, "step": 11925 }, { "epoch": 1.42, "learning_rate": 2.835674168453796e-06, "loss": 0.6985, "step": 11926 }, { "epoch": 1.42, "learning_rate": 2.8353556527891167e-06, "loss": 0.7781, "step": 11927 }, { "epoch": 1.42, "learning_rate": 2.8350371315809804e-06, "loss": 0.8508, "step": 11928 }, { "epoch": 1.42, "learning_rate": 2.8347186048346526e-06, "loss": 0.4435, "step": 11929 }, { "epoch": 1.42, "learning_rate": 2.8344000725553976e-06, "loss": 0.587, "step": 11930 }, { "epoch": 1.42, "learning_rate": 2.8340815347484818e-06, "loss": 0.5665, "step": 11931 }, { "epoch": 1.42, "learning_rate": 2.8337629914191708e-06, "loss": 0.565, "step": 11932 }, { "epoch": 1.42, "learning_rate": 2.833444442572729e-06, "loss": 0.8002, "step": 11933 }, { "epoch": 1.42, "learning_rate": 2.833125888214422e-06, "loss": 0.9145, "step": 11934 }, { "epoch": 1.42, "learning_rate": 2.832807328349517e-06, "loss": 0.5447, "step": 11935 }, { "epoch": 1.42, "learning_rate": 2.8324887629832787e-06, "loss": 0.5404, "step": 11936 }, { "epoch": 1.42, "learning_rate": 2.832170192120973e-06, "loss": 0.4331, "step": 11937 }, { "epoch": 1.42, "learning_rate": 2.8318516157678653e-06, "loss": 0.6044, "step": 11938 }, { "epoch": 1.42, "learning_rate": 2.831533033929224e-06, "loss": 0.7859, "step": 11939 }, { "epoch": 1.42, "learning_rate": 2.8312144466103137e-06, "loss": 0.7537, "step": 11940 }, { "epoch": 1.42, "learning_rate": 2.8308958538164e-06, "loss": 0.5887, "step": 11941 }, { "epoch": 1.42, "learning_rate": 2.83057725555275e-06, "loss": 0.4716, "step": 11942 }, { "epoch": 1.42, "learning_rate": 2.830258651824631e-06, "loss": 0.5584, "step": 11943 }, { "epoch": 1.42, "learning_rate": 2.8299400426373084e-06, "loss": 0.5367, "step": 11944 }, { "epoch": 1.42, "learning_rate": 2.8296214279960492e-06, "loss": 0.6868, "step": 11945 }, { "epoch": 1.42, "learning_rate": 2.8293028079061206e-06, "loss": 0.7495, "step": 11946 }, { "epoch": 1.42, "learning_rate": 2.828984182372788e-06, "loss": 0.8422, "step": 11947 }, { "epoch": 1.42, "learning_rate": 2.8286655514013206e-06, "loss": 0.5172, "step": 11948 }, { "epoch": 1.42, "learning_rate": 2.8283469149969834e-06, "loss": 0.5935, "step": 11949 }, { "epoch": 1.42, "learning_rate": 2.8280282731650438e-06, "loss": 0.5577, "step": 11950 }, { "epoch": 1.42, "learning_rate": 2.8277096259107693e-06, "loss": 0.7285, "step": 11951 }, { "epoch": 1.42, "learning_rate": 2.8273909732394274e-06, "loss": 0.7636, "step": 11952 }, { "epoch": 1.42, "learning_rate": 2.827072315156286e-06, "loss": 0.8035, "step": 11953 }, { "epoch": 1.42, "learning_rate": 2.8267536516666106e-06, "loss": 0.3632, "step": 11954 }, { "epoch": 1.42, "learning_rate": 2.8264349827756702e-06, "loss": 0.5285, "step": 11955 }, { "epoch": 1.42, "learning_rate": 2.8261163084887323e-06, "loss": 0.5816, "step": 11956 }, { "epoch": 1.42, "learning_rate": 2.825797628811065e-06, "loss": 0.6004, "step": 11957 }, { "epoch": 1.42, "learning_rate": 2.8254789437479346e-06, "loss": 0.8035, "step": 11958 }, { "epoch": 1.42, "learning_rate": 2.8251602533046106e-06, "loss": 0.8402, "step": 11959 }, { "epoch": 1.42, "learning_rate": 2.8248415574863603e-06, "loss": 0.4314, "step": 11960 }, { "epoch": 1.42, "learning_rate": 2.8245228562984518e-06, "loss": 0.5147, "step": 11961 }, { "epoch": 1.42, "learning_rate": 2.8242041497461527e-06, "loss": 0.5482, "step": 11962 }, { "epoch": 1.42, "learning_rate": 2.8238854378347325e-06, "loss": 0.6326, "step": 11963 }, { "epoch": 1.42, "learning_rate": 2.823566720569459e-06, "loss": 0.7805, "step": 11964 }, { "epoch": 1.42, "learning_rate": 2.823247997955599e-06, "loss": 0.7787, "step": 11965 }, { "epoch": 1.43, "learning_rate": 2.822929269998424e-06, "loss": 0.5139, "step": 11966 }, { "epoch": 1.43, "learning_rate": 2.822610536703201e-06, "loss": 0.5664, "step": 11967 }, { "epoch": 1.43, "learning_rate": 2.822291798075198e-06, "loss": 0.522, "step": 11968 }, { "epoch": 1.43, "learning_rate": 2.821973054119684e-06, "loss": 0.6283, "step": 11969 }, { "epoch": 1.43, "learning_rate": 2.82165430484193e-06, "loss": 0.6826, "step": 11970 }, { "epoch": 1.43, "learning_rate": 2.8213355502472027e-06, "loss": 0.8749, "step": 11971 }, { "epoch": 1.43, "learning_rate": 2.8210167903407714e-06, "loss": 0.825, "step": 11972 }, { "epoch": 1.43, "learning_rate": 2.820698025127906e-06, "loss": 0.3906, "step": 11973 }, { "epoch": 1.43, "learning_rate": 2.8203792546138753e-06, "loss": 0.5274, "step": 11974 }, { "epoch": 1.43, "learning_rate": 2.820060478803948e-06, "loss": 0.5614, "step": 11975 }, { "epoch": 1.43, "learning_rate": 2.8197416977033955e-06, "loss": 0.7232, "step": 11976 }, { "epoch": 1.43, "learning_rate": 2.8194229113174843e-06, "loss": 0.9179, "step": 11977 }, { "epoch": 1.43, "learning_rate": 2.8191041196514874e-06, "loss": 0.8114, "step": 11978 }, { "epoch": 1.43, "learning_rate": 2.818785322710671e-06, "loss": 0.3931, "step": 11979 }, { "epoch": 1.43, "learning_rate": 2.818466520500307e-06, "loss": 0.5522, "step": 11980 }, { "epoch": 1.43, "learning_rate": 2.818147713025665e-06, "loss": 0.6809, "step": 11981 }, { "epoch": 1.43, "learning_rate": 2.8178289002920144e-06, "loss": 0.6515, "step": 11982 }, { "epoch": 1.43, "learning_rate": 2.8175100823046253e-06, "loss": 0.7813, "step": 11983 }, { "epoch": 1.43, "learning_rate": 2.8171912590687677e-06, "loss": 0.8881, "step": 11984 }, { "epoch": 1.43, "learning_rate": 2.816872430589712e-06, "loss": 0.4285, "step": 11985 }, { "epoch": 1.43, "learning_rate": 2.816553596872729e-06, "loss": 0.4437, "step": 11986 }, { "epoch": 1.43, "learning_rate": 2.8162347579230885e-06, "loss": 0.6263, "step": 11987 }, { "epoch": 1.43, "learning_rate": 2.8159159137460603e-06, "loss": 0.6069, "step": 11988 }, { "epoch": 1.43, "learning_rate": 2.8155970643469165e-06, "loss": 0.8325, "step": 11989 }, { "epoch": 1.43, "learning_rate": 2.8152782097309255e-06, "loss": 0.844, "step": 11990 }, { "epoch": 1.43, "learning_rate": 2.81495934990336e-06, "loss": 0.6141, "step": 11991 }, { "epoch": 1.43, "learning_rate": 2.8146404848694895e-06, "loss": 0.4105, "step": 11992 }, { "epoch": 1.43, "learning_rate": 2.8143216146345862e-06, "loss": 0.5305, "step": 11993 }, { "epoch": 1.43, "learning_rate": 2.8140027392039193e-06, "loss": 0.5089, "step": 11994 }, { "epoch": 1.43, "learning_rate": 2.8136838585827616e-06, "loss": 0.6664, "step": 11995 }, { "epoch": 1.43, "learning_rate": 2.8133649727763832e-06, "loss": 0.8473, "step": 11996 }, { "epoch": 1.43, "learning_rate": 2.813046081790055e-06, "loss": 0.8556, "step": 11997 }, { "epoch": 1.43, "learning_rate": 2.8127271856290495e-06, "loss": 0.3704, "step": 11998 }, { "epoch": 1.43, "learning_rate": 2.8124082842986367e-06, "loss": 0.5305, "step": 11999 }, { "epoch": 1.43, "learning_rate": 2.8120893778040897e-06, "loss": 0.5821, "step": 12000 }, { "epoch": 1.43, "learning_rate": 2.811770466150678e-06, "loss": 0.6719, "step": 12001 }, { "epoch": 1.43, "learning_rate": 2.811451549343675e-06, "loss": 0.7781, "step": 12002 }, { "epoch": 1.43, "learning_rate": 2.8111326273883514e-06, "loss": 0.7961, "step": 12003 }, { "epoch": 1.43, "learning_rate": 2.81081370028998e-06, "loss": 0.4911, "step": 12004 }, { "epoch": 1.43, "learning_rate": 2.8104947680538307e-06, "loss": 0.427, "step": 12005 }, { "epoch": 1.43, "learning_rate": 2.8101758306851777e-06, "loss": 0.6201, "step": 12006 }, { "epoch": 1.43, "learning_rate": 2.8098568881892923e-06, "loss": 0.6453, "step": 12007 }, { "epoch": 1.43, "learning_rate": 2.809537940571447e-06, "loss": 0.8317, "step": 12008 }, { "epoch": 1.43, "learning_rate": 2.8092189878369127e-06, "loss": 0.8185, "step": 12009 }, { "epoch": 1.43, "learning_rate": 2.808900029990963e-06, "loss": 0.4048, "step": 12010 }, { "epoch": 1.43, "learning_rate": 2.8085810670388696e-06, "loss": 0.3708, "step": 12011 }, { "epoch": 1.43, "learning_rate": 2.8082620989859055e-06, "loss": 0.6059, "step": 12012 }, { "epoch": 1.43, "learning_rate": 2.8079431258373428e-06, "loss": 0.587, "step": 12013 }, { "epoch": 1.43, "learning_rate": 2.807624147598455e-06, "loss": 0.7527, "step": 12014 }, { "epoch": 1.43, "learning_rate": 2.8073051642745144e-06, "loss": 0.8134, "step": 12015 }, { "epoch": 1.43, "learning_rate": 2.806986175870792e-06, "loss": 0.5443, "step": 12016 }, { "epoch": 1.43, "learning_rate": 2.8066671823925633e-06, "loss": 0.497, "step": 12017 }, { "epoch": 1.43, "learning_rate": 2.8063481838451006e-06, "loss": 0.5507, "step": 12018 }, { "epoch": 1.43, "learning_rate": 2.8060291802336768e-06, "loss": 0.6762, "step": 12019 }, { "epoch": 1.43, "learning_rate": 2.8057101715635643e-06, "loss": 0.7285, "step": 12020 }, { "epoch": 1.43, "learning_rate": 2.8053911578400384e-06, "loss": 0.72, "step": 12021 }, { "epoch": 1.43, "learning_rate": 2.8050721390683694e-06, "loss": 0.899, "step": 12022 }, { "epoch": 1.43, "learning_rate": 2.804753115253833e-06, "loss": 0.4518, "step": 12023 }, { "epoch": 1.43, "learning_rate": 2.8044340864017024e-06, "loss": 0.5887, "step": 12024 }, { "epoch": 1.43, "learning_rate": 2.8041150525172508e-06, "loss": 0.5537, "step": 12025 }, { "epoch": 1.43, "learning_rate": 2.8037960136057505e-06, "loss": 0.5892, "step": 12026 }, { "epoch": 1.43, "learning_rate": 2.8034769696724782e-06, "loss": 0.6613, "step": 12027 }, { "epoch": 1.43, "learning_rate": 2.8031579207227054e-06, "loss": 0.9277, "step": 12028 }, { "epoch": 1.43, "learning_rate": 2.8028388667617073e-06, "loss": 0.3919, "step": 12029 }, { "epoch": 1.43, "learning_rate": 2.802519807794757e-06, "loss": 0.5778, "step": 12030 }, { "epoch": 1.43, "learning_rate": 2.802200743827129e-06, "loss": 0.4288, "step": 12031 }, { "epoch": 1.43, "learning_rate": 2.8018816748640977e-06, "loss": 0.8015, "step": 12032 }, { "epoch": 1.43, "learning_rate": 2.8015626009109355e-06, "loss": 0.6624, "step": 12033 }, { "epoch": 1.43, "learning_rate": 2.8012435219729196e-06, "loss": 0.888, "step": 12034 }, { "epoch": 1.43, "learning_rate": 2.8009244380553223e-06, "loss": 0.508, "step": 12035 }, { "epoch": 1.43, "learning_rate": 2.8006053491634195e-06, "loss": 0.4424, "step": 12036 }, { "epoch": 1.43, "learning_rate": 2.800286255302484e-06, "loss": 0.5865, "step": 12037 }, { "epoch": 1.43, "learning_rate": 2.799967156477792e-06, "loss": 0.593, "step": 12038 }, { "epoch": 1.43, "learning_rate": 2.799648052694618e-06, "loss": 0.7419, "step": 12039 }, { "epoch": 1.43, "learning_rate": 2.7993289439582366e-06, "loss": 0.9566, "step": 12040 }, { "epoch": 1.43, "learning_rate": 2.799009830273922e-06, "loss": 0.4259, "step": 12041 }, { "epoch": 1.43, "learning_rate": 2.79869071164695e-06, "loss": 0.4294, "step": 12042 }, { "epoch": 1.43, "learning_rate": 2.7983715880825963e-06, "loss": 0.504, "step": 12043 }, { "epoch": 1.43, "learning_rate": 2.798052459586135e-06, "loss": 0.5981, "step": 12044 }, { "epoch": 1.43, "learning_rate": 2.7977333261628407e-06, "loss": 0.6887, "step": 12045 }, { "epoch": 1.43, "learning_rate": 2.7974141878179907e-06, "loss": 0.9511, "step": 12046 }, { "epoch": 1.43, "learning_rate": 2.7970950445568586e-06, "loss": 0.8196, "step": 12047 }, { "epoch": 1.43, "learning_rate": 2.7967758963847197e-06, "loss": 0.4737, "step": 12048 }, { "epoch": 1.43, "learning_rate": 2.7964567433068514e-06, "loss": 0.4063, "step": 12049 }, { "epoch": 1.44, "learning_rate": 2.7961375853285277e-06, "loss": 0.6961, "step": 12050 }, { "epoch": 1.44, "learning_rate": 2.795818422455025e-06, "loss": 0.6571, "step": 12051 }, { "epoch": 1.44, "learning_rate": 2.7954992546916187e-06, "loss": 0.7278, "step": 12052 }, { "epoch": 1.44, "learning_rate": 2.795180082043586e-06, "loss": 0.8549, "step": 12053 }, { "epoch": 1.44, "learning_rate": 2.7948609045162007e-06, "loss": 0.519, "step": 12054 }, { "epoch": 1.44, "learning_rate": 2.7945417221147404e-06, "loss": 0.5774, "step": 12055 }, { "epoch": 1.44, "learning_rate": 2.79422253484448e-06, "loss": 0.4732, "step": 12056 }, { "epoch": 1.44, "learning_rate": 2.7939033427106976e-06, "loss": 0.6431, "step": 12057 }, { "epoch": 1.44, "learning_rate": 2.7935841457186674e-06, "loss": 0.6546, "step": 12058 }, { "epoch": 1.44, "learning_rate": 2.793264943873667e-06, "loss": 0.7902, "step": 12059 }, { "epoch": 1.44, "learning_rate": 2.792945737180972e-06, "loss": 0.3644, "step": 12060 }, { "epoch": 1.44, "learning_rate": 2.7926265256458602e-06, "loss": 0.3599, "step": 12061 }, { "epoch": 1.44, "learning_rate": 2.792307309273607e-06, "loss": 0.5116, "step": 12062 }, { "epoch": 1.44, "learning_rate": 2.791988088069489e-06, "loss": 0.5679, "step": 12063 }, { "epoch": 1.44, "learning_rate": 2.7916688620387846e-06, "loss": 0.7171, "step": 12064 }, { "epoch": 1.44, "learning_rate": 2.791349631186768e-06, "loss": 0.7176, "step": 12065 }, { "epoch": 1.44, "learning_rate": 2.7910303955187186e-06, "loss": 0.4881, "step": 12066 }, { "epoch": 1.44, "learning_rate": 2.790711155039912e-06, "loss": 0.4898, "step": 12067 }, { "epoch": 1.44, "learning_rate": 2.790391909755625e-06, "loss": 0.639, "step": 12068 }, { "epoch": 1.44, "learning_rate": 2.7900726596711358e-06, "loss": 0.5026, "step": 12069 }, { "epoch": 1.44, "learning_rate": 2.789753404791721e-06, "loss": 0.7315, "step": 12070 }, { "epoch": 1.44, "learning_rate": 2.789434145122659e-06, "loss": 0.8232, "step": 12071 }, { "epoch": 1.44, "learning_rate": 2.789114880669226e-06, "loss": 0.7159, "step": 12072 }, { "epoch": 1.44, "learning_rate": 2.7887956114366986e-06, "loss": 0.3989, "step": 12073 }, { "epoch": 1.44, "learning_rate": 2.7884763374303564e-06, "loss": 0.6255, "step": 12074 }, { "epoch": 1.44, "learning_rate": 2.788157058655476e-06, "loss": 0.5045, "step": 12075 }, { "epoch": 1.44, "learning_rate": 2.787837775117336e-06, "loss": 0.7044, "step": 12076 }, { "epoch": 1.44, "learning_rate": 2.7875184868212123e-06, "loss": 0.6758, "step": 12077 }, { "epoch": 1.44, "learning_rate": 2.7871991937723845e-06, "loss": 0.8182, "step": 12078 }, { "epoch": 1.44, "learning_rate": 2.78687989597613e-06, "loss": 0.4118, "step": 12079 }, { "epoch": 1.44, "learning_rate": 2.786560593437726e-06, "loss": 0.586, "step": 12080 }, { "epoch": 1.44, "learning_rate": 2.786241286162452e-06, "loss": 0.5965, "step": 12081 }, { "epoch": 1.44, "learning_rate": 2.785921974155586e-06, "loss": 0.7078, "step": 12082 }, { "epoch": 1.44, "learning_rate": 2.785602657422405e-06, "loss": 0.7282, "step": 12083 }, { "epoch": 1.44, "learning_rate": 2.785283335968188e-06, "loss": 0.8223, "step": 12084 }, { "epoch": 1.44, "learning_rate": 2.784964009798214e-06, "loss": 0.3967, "step": 12085 }, { "epoch": 1.44, "learning_rate": 2.7846446789177607e-06, "loss": 0.4754, "step": 12086 }, { "epoch": 1.44, "learning_rate": 2.784325343332107e-06, "loss": 0.5752, "step": 12087 }, { "epoch": 1.44, "learning_rate": 2.7840060030465316e-06, "loss": 0.6527, "step": 12088 }, { "epoch": 1.44, "learning_rate": 2.7836866580663136e-06, "loss": 0.7673, "step": 12089 }, { "epoch": 1.44, "learning_rate": 2.7833673083967306e-06, "loss": 0.7927, "step": 12090 }, { "epoch": 1.44, "learning_rate": 2.7830479540430623e-06, "loss": 0.5623, "step": 12091 }, { "epoch": 1.44, "learning_rate": 2.782728595010588e-06, "loss": 0.3918, "step": 12092 }, { "epoch": 1.44, "learning_rate": 2.7824092313045862e-06, "loss": 0.4668, "step": 12093 }, { "epoch": 1.44, "learning_rate": 2.782089862930336e-06, "loss": 0.5035, "step": 12094 }, { "epoch": 1.44, "learning_rate": 2.7817704898931164e-06, "loss": 0.601, "step": 12095 }, { "epoch": 1.44, "learning_rate": 2.781451112198208e-06, "loss": 0.7841, "step": 12096 }, { "epoch": 1.44, "learning_rate": 2.781131729850888e-06, "loss": 0.8594, "step": 12097 }, { "epoch": 1.44, "learning_rate": 2.780812342856437e-06, "loss": 0.4627, "step": 12098 }, { "epoch": 1.44, "learning_rate": 2.7804929512201344e-06, "loss": 0.5092, "step": 12099 }, { "epoch": 1.44, "learning_rate": 2.7801735549472607e-06, "loss": 0.5478, "step": 12100 }, { "epoch": 1.44, "learning_rate": 2.7798541540430934e-06, "loss": 0.6719, "step": 12101 }, { "epoch": 1.44, "learning_rate": 2.779534748512914e-06, "loss": 0.8212, "step": 12102 }, { "epoch": 1.44, "learning_rate": 2.779215338362002e-06, "loss": 0.947, "step": 12103 }, { "epoch": 1.44, "learning_rate": 2.778895923595637e-06, "loss": 0.4388, "step": 12104 }, { "epoch": 1.44, "learning_rate": 2.778576504219098e-06, "loss": 0.5139, "step": 12105 }, { "epoch": 1.44, "learning_rate": 2.7782570802376676e-06, "loss": 0.5137, "step": 12106 }, { "epoch": 1.44, "learning_rate": 2.777937651656623e-06, "loss": 0.7009, "step": 12107 }, { "epoch": 1.44, "learning_rate": 2.7776182184812465e-06, "loss": 0.7749, "step": 12108 }, { "epoch": 1.44, "learning_rate": 2.777298780716817e-06, "loss": 0.7979, "step": 12109 }, { "epoch": 1.44, "learning_rate": 2.776979338368616e-06, "loss": 0.4894, "step": 12110 }, { "epoch": 1.44, "learning_rate": 2.776659891441923e-06, "loss": 0.5409, "step": 12111 }, { "epoch": 1.44, "learning_rate": 2.7763404399420186e-06, "loss": 0.5611, "step": 12112 }, { "epoch": 1.44, "learning_rate": 2.7760209838741835e-06, "loss": 0.6799, "step": 12113 }, { "epoch": 1.44, "learning_rate": 2.775701523243699e-06, "loss": 0.7494, "step": 12114 }, { "epoch": 1.44, "learning_rate": 2.775382058055845e-06, "loss": 0.7978, "step": 12115 }, { "epoch": 1.44, "learning_rate": 2.7750625883159015e-06, "loss": 0.5336, "step": 12116 }, { "epoch": 1.44, "learning_rate": 2.774743114029151e-06, "loss": 0.4643, "step": 12117 }, { "epoch": 1.44, "learning_rate": 2.7744236352008742e-06, "loss": 0.582, "step": 12118 }, { "epoch": 1.44, "learning_rate": 2.7741041518363508e-06, "loss": 0.5667, "step": 12119 }, { "epoch": 1.44, "learning_rate": 2.7737846639408634e-06, "loss": 0.6426, "step": 12120 }, { "epoch": 1.44, "learning_rate": 2.773465171519693e-06, "loss": 0.8292, "step": 12121 }, { "epoch": 1.44, "learning_rate": 2.7731456745781195e-06, "loss": 0.9644, "step": 12122 }, { "epoch": 1.44, "learning_rate": 2.772826173121425e-06, "loss": 0.4291, "step": 12123 }, { "epoch": 1.44, "learning_rate": 2.7725066671548917e-06, "loss": 0.6272, "step": 12124 }, { "epoch": 1.44, "learning_rate": 2.7721871566838e-06, "loss": 0.5229, "step": 12125 }, { "epoch": 1.44, "learning_rate": 2.7718676417134316e-06, "loss": 0.6549, "step": 12126 }, { "epoch": 1.44, "learning_rate": 2.7715481222490683e-06, "loss": 0.7759, "step": 12127 }, { "epoch": 1.44, "learning_rate": 2.771228598295992e-06, "loss": 0.8641, "step": 12128 }, { "epoch": 1.44, "learning_rate": 2.7709090698594837e-06, "loss": 0.4803, "step": 12129 }, { "epoch": 1.44, "learning_rate": 2.7705895369448267e-06, "loss": 0.516, "step": 12130 }, { "epoch": 1.44, "learning_rate": 2.7702699995573008e-06, "loss": 0.4975, "step": 12131 }, { "epoch": 1.44, "learning_rate": 2.7699504577021902e-06, "loss": 0.6063, "step": 12132 }, { "epoch": 1.44, "learning_rate": 2.7696309113847746e-06, "loss": 0.6867, "step": 12133 }, { "epoch": 1.45, "learning_rate": 2.7693113606103385e-06, "loss": 0.8166, "step": 12134 }, { "epoch": 1.45, "learning_rate": 2.7689918053841632e-06, "loss": 0.4191, "step": 12135 }, { "epoch": 1.45, "learning_rate": 2.76867224571153e-06, "loss": 0.4399, "step": 12136 }, { "epoch": 1.45, "learning_rate": 2.7683526815977217e-06, "loss": 0.5164, "step": 12137 }, { "epoch": 1.45, "learning_rate": 2.768033113048022e-06, "loss": 0.5866, "step": 12138 }, { "epoch": 1.45, "learning_rate": 2.767713540067713e-06, "loss": 0.6692, "step": 12139 }, { "epoch": 1.45, "learning_rate": 2.767393962662075e-06, "loss": 0.7749, "step": 12140 }, { "epoch": 1.45, "learning_rate": 2.7670743808363933e-06, "loss": 0.5112, "step": 12141 }, { "epoch": 1.45, "learning_rate": 2.7667547945959496e-06, "loss": 0.429, "step": 12142 }, { "epoch": 1.45, "learning_rate": 2.7664352039460268e-06, "loss": 0.5195, "step": 12143 }, { "epoch": 1.45, "learning_rate": 2.7661156088919074e-06, "loss": 0.6052, "step": 12144 }, { "epoch": 1.45, "learning_rate": 2.7657960094388754e-06, "loss": 0.623, "step": 12145 }, { "epoch": 1.45, "learning_rate": 2.7654764055922123e-06, "loss": 0.7235, "step": 12146 }, { "epoch": 1.45, "learning_rate": 2.7651567973572024e-06, "loss": 0.8107, "step": 12147 }, { "epoch": 1.45, "learning_rate": 2.7648371847391277e-06, "loss": 0.5362, "step": 12148 }, { "epoch": 1.45, "learning_rate": 2.764517567743273e-06, "loss": 0.4411, "step": 12149 }, { "epoch": 1.45, "learning_rate": 2.76419794637492e-06, "loss": 0.636, "step": 12150 }, { "epoch": 1.45, "learning_rate": 2.7638783206393532e-06, "loss": 0.5808, "step": 12151 }, { "epoch": 1.45, "learning_rate": 2.7635586905418554e-06, "loss": 0.7537, "step": 12152 }, { "epoch": 1.45, "learning_rate": 2.763239056087711e-06, "loss": 0.8552, "step": 12153 }, { "epoch": 1.45, "learning_rate": 2.7629194172822016e-06, "loss": 0.3745, "step": 12154 }, { "epoch": 1.45, "learning_rate": 2.762599774130613e-06, "loss": 0.4424, "step": 12155 }, { "epoch": 1.45, "learning_rate": 2.7622801266382282e-06, "loss": 0.6255, "step": 12156 }, { "epoch": 1.45, "learning_rate": 2.7619604748103314e-06, "loss": 0.5513, "step": 12157 }, { "epoch": 1.45, "learning_rate": 2.7616408186522048e-06, "loss": 0.7919, "step": 12158 }, { "epoch": 1.45, "learning_rate": 2.761321158169134e-06, "loss": 0.7738, "step": 12159 }, { "epoch": 1.45, "learning_rate": 2.7610014933664024e-06, "loss": 0.3672, "step": 12160 }, { "epoch": 1.45, "learning_rate": 2.7606818242492943e-06, "loss": 0.4264, "step": 12161 }, { "epoch": 1.45, "learning_rate": 2.7603621508230945e-06, "loss": 0.6005, "step": 12162 }, { "epoch": 1.45, "learning_rate": 2.760042473093085e-06, "loss": 0.649, "step": 12163 }, { "epoch": 1.45, "learning_rate": 2.759722791064553e-06, "loss": 0.619, "step": 12164 }, { "epoch": 1.45, "learning_rate": 2.7594031047427806e-06, "loss": 0.8347, "step": 12165 }, { "epoch": 1.45, "learning_rate": 2.759083414133054e-06, "loss": 0.4014, "step": 12166 }, { "epoch": 1.45, "learning_rate": 2.758763719240656e-06, "loss": 0.4152, "step": 12167 }, { "epoch": 1.45, "learning_rate": 2.7584440200708722e-06, "loss": 0.5783, "step": 12168 }, { "epoch": 1.45, "learning_rate": 2.7581243166289868e-06, "loss": 0.7439, "step": 12169 }, { "epoch": 1.45, "learning_rate": 2.7578046089202855e-06, "loss": 0.6291, "step": 12170 }, { "epoch": 1.45, "learning_rate": 2.7574848969500518e-06, "loss": 0.7306, "step": 12171 }, { "epoch": 1.45, "learning_rate": 2.7571651807235715e-06, "loss": 0.6978, "step": 12172 }, { "epoch": 1.45, "learning_rate": 2.7568454602461293e-06, "loss": 0.4885, "step": 12173 }, { "epoch": 1.45, "learning_rate": 2.756525735523009e-06, "loss": 0.5491, "step": 12174 }, { "epoch": 1.45, "learning_rate": 2.7562060065594987e-06, "loss": 0.664, "step": 12175 }, { "epoch": 1.45, "learning_rate": 2.75588627336088e-06, "loss": 0.6996, "step": 12176 }, { "epoch": 1.45, "learning_rate": 2.7555665359324403e-06, "loss": 0.7396, "step": 12177 }, { "epoch": 1.45, "learning_rate": 2.755246794279464e-06, "loss": 0.9632, "step": 12178 }, { "epoch": 1.45, "learning_rate": 2.7549270484072373e-06, "loss": 0.5511, "step": 12179 }, { "epoch": 1.45, "learning_rate": 2.7546072983210444e-06, "loss": 0.4281, "step": 12180 }, { "epoch": 1.45, "learning_rate": 2.754287544026172e-06, "loss": 0.6426, "step": 12181 }, { "epoch": 1.45, "learning_rate": 2.753967785527905e-06, "loss": 0.6335, "step": 12182 }, { "epoch": 1.45, "learning_rate": 2.7536480228315294e-06, "loss": 0.7115, "step": 12183 }, { "epoch": 1.45, "learning_rate": 2.75332825594233e-06, "loss": 0.863, "step": 12184 }, { "epoch": 1.45, "learning_rate": 2.753008484865594e-06, "loss": 0.4794, "step": 12185 }, { "epoch": 1.45, "learning_rate": 2.752688709606607e-06, "loss": 0.5797, "step": 12186 }, { "epoch": 1.45, "learning_rate": 2.7523689301706537e-06, "loss": 0.5921, "step": 12187 }, { "epoch": 1.45, "learning_rate": 2.752049146563021e-06, "loss": 0.7186, "step": 12188 }, { "epoch": 1.45, "learning_rate": 2.751729358788995e-06, "loss": 0.6998, "step": 12189 }, { "epoch": 1.45, "learning_rate": 2.751409566853861e-06, "loss": 0.8415, "step": 12190 }, { "epoch": 1.45, "learning_rate": 2.751089770762906e-06, "loss": 0.4586, "step": 12191 }, { "epoch": 1.45, "learning_rate": 2.750769970521416e-06, "loss": 0.4696, "step": 12192 }, { "epoch": 1.45, "learning_rate": 2.750450166134677e-06, "loss": 0.5815, "step": 12193 }, { "epoch": 1.45, "learning_rate": 2.750130357607976e-06, "loss": 0.6166, "step": 12194 }, { "epoch": 1.45, "learning_rate": 2.7498105449465996e-06, "loss": 0.6496, "step": 12195 }, { "epoch": 1.45, "learning_rate": 2.749490728155834e-06, "loss": 0.8101, "step": 12196 }, { "epoch": 1.45, "learning_rate": 2.749170907240965e-06, "loss": 0.8005, "step": 12197 }, { "epoch": 1.45, "learning_rate": 2.74885108220728e-06, "loss": 0.3734, "step": 12198 }, { "epoch": 1.45, "learning_rate": 2.7485312530600663e-06, "loss": 0.5786, "step": 12199 }, { "epoch": 1.45, "learning_rate": 2.74821141980461e-06, "loss": 0.5794, "step": 12200 }, { "epoch": 1.45, "learning_rate": 2.7478915824461976e-06, "loss": 0.6701, "step": 12201 }, { "epoch": 1.45, "learning_rate": 2.747571740990117e-06, "loss": 0.7971, "step": 12202 }, { "epoch": 1.45, "learning_rate": 2.747251895441655e-06, "loss": 0.9059, "step": 12203 }, { "epoch": 1.45, "learning_rate": 2.7469320458060984e-06, "loss": 0.5005, "step": 12204 }, { "epoch": 1.45, "learning_rate": 2.746612192088734e-06, "loss": 0.4853, "step": 12205 }, { "epoch": 1.45, "learning_rate": 2.7462923342948494e-06, "loss": 0.5622, "step": 12206 }, { "epoch": 1.45, "learning_rate": 2.7459724724297326e-06, "loss": 0.5732, "step": 12207 }, { "epoch": 1.45, "learning_rate": 2.7456526064986687e-06, "loss": 0.7321, "step": 12208 }, { "epoch": 1.45, "learning_rate": 2.7453327365069475e-06, "loss": 0.8955, "step": 12209 }, { "epoch": 1.45, "learning_rate": 2.745012862459856e-06, "loss": 0.4222, "step": 12210 }, { "epoch": 1.45, "learning_rate": 2.7446929843626806e-06, "loss": 0.3947, "step": 12211 }, { "epoch": 1.45, "learning_rate": 2.7443731022207097e-06, "loss": 0.5603, "step": 12212 }, { "epoch": 1.45, "learning_rate": 2.744053216039231e-06, "loss": 0.628, "step": 12213 }, { "epoch": 1.45, "learning_rate": 2.7437333258235326e-06, "loss": 0.6409, "step": 12214 }, { "epoch": 1.45, "learning_rate": 2.7434134315789013e-06, "loss": 0.7646, "step": 12215 }, { "epoch": 1.45, "learning_rate": 2.743093533310626e-06, "loss": 0.6575, "step": 12216 }, { "epoch": 1.45, "learning_rate": 2.7427736310239935e-06, "loss": 0.4344, "step": 12217 }, { "epoch": 1.46, "learning_rate": 2.7424537247242935e-06, "loss": 0.6218, "step": 12218 }, { "epoch": 1.46, "learning_rate": 2.7421338144168127e-06, "loss": 0.6088, "step": 12219 }, { "epoch": 1.46, "learning_rate": 2.741813900106839e-06, "loss": 0.6805, "step": 12220 }, { "epoch": 1.46, "learning_rate": 2.7414939817996622e-06, "loss": 0.791, "step": 12221 }, { "epoch": 1.46, "learning_rate": 2.741174059500569e-06, "loss": 0.8408, "step": 12222 }, { "epoch": 1.46, "learning_rate": 2.740854133214848e-06, "loss": 0.4532, "step": 12223 }, { "epoch": 1.46, "learning_rate": 2.740534202947789e-06, "loss": 0.5481, "step": 12224 }, { "epoch": 1.46, "learning_rate": 2.740214268704679e-06, "loss": 0.5244, "step": 12225 }, { "epoch": 1.46, "learning_rate": 2.7398943304908067e-06, "loss": 0.6965, "step": 12226 }, { "epoch": 1.46, "learning_rate": 2.7395743883114606e-06, "loss": 0.7824, "step": 12227 }, { "epoch": 1.46, "learning_rate": 2.7392544421719306e-06, "loss": 0.9316, "step": 12228 }, { "epoch": 1.46, "learning_rate": 2.7389344920775047e-06, "loss": 0.5803, "step": 12229 }, { "epoch": 1.46, "learning_rate": 2.7386145380334713e-06, "loss": 0.5407, "step": 12230 }, { "epoch": 1.46, "learning_rate": 2.738294580045119e-06, "loss": 0.6248, "step": 12231 }, { "epoch": 1.46, "learning_rate": 2.7379746181177388e-06, "loss": 0.6187, "step": 12232 }, { "epoch": 1.46, "learning_rate": 2.737654652256616e-06, "loss": 0.7803, "step": 12233 }, { "epoch": 1.46, "learning_rate": 2.7373346824670432e-06, "loss": 0.8145, "step": 12234 }, { "epoch": 1.46, "learning_rate": 2.7370147087543085e-06, "loss": 0.368, "step": 12235 }, { "epoch": 1.46, "learning_rate": 2.7366947311237003e-06, "loss": 0.4902, "step": 12236 }, { "epoch": 1.46, "learning_rate": 2.736374749580509e-06, "loss": 0.5103, "step": 12237 }, { "epoch": 1.46, "learning_rate": 2.736054764130022e-06, "loss": 0.6042, "step": 12238 }, { "epoch": 1.46, "learning_rate": 2.7357347747775306e-06, "loss": 0.7102, "step": 12239 }, { "epoch": 1.46, "learning_rate": 2.7354147815283235e-06, "loss": 0.8624, "step": 12240 }, { "epoch": 1.46, "learning_rate": 2.735094784387691e-06, "loss": 0.5662, "step": 12241 }, { "epoch": 1.46, "learning_rate": 2.7347747833609205e-06, "loss": 0.4893, "step": 12242 }, { "epoch": 1.46, "learning_rate": 2.7344547784533044e-06, "loss": 0.4573, "step": 12243 }, { "epoch": 1.46, "learning_rate": 2.7341347696701306e-06, "loss": 0.5569, "step": 12244 }, { "epoch": 1.46, "learning_rate": 2.7338147570166896e-06, "loss": 0.7265, "step": 12245 }, { "epoch": 1.46, "learning_rate": 2.733494740498271e-06, "loss": 0.8901, "step": 12246 }, { "epoch": 1.46, "learning_rate": 2.733174720120165e-06, "loss": 0.8617, "step": 12247 }, { "epoch": 1.46, "learning_rate": 2.7328546958876602e-06, "loss": 0.3984, "step": 12248 }, { "epoch": 1.46, "learning_rate": 2.732534667806049e-06, "loss": 0.6188, "step": 12249 }, { "epoch": 1.46, "learning_rate": 2.7322146358806196e-06, "loss": 0.6353, "step": 12250 }, { "epoch": 1.46, "learning_rate": 2.731894600116663e-06, "loss": 0.67, "step": 12251 }, { "epoch": 1.46, "learning_rate": 2.7315745605194695e-06, "loss": 0.8486, "step": 12252 }, { "epoch": 1.46, "learning_rate": 2.731254517094329e-06, "loss": 0.9524, "step": 12253 }, { "epoch": 1.46, "learning_rate": 2.730934469846532e-06, "loss": 0.5808, "step": 12254 }, { "epoch": 1.46, "learning_rate": 2.7306144187813677e-06, "loss": 0.5905, "step": 12255 }, { "epoch": 1.46, "learning_rate": 2.7302943639041284e-06, "loss": 0.5771, "step": 12256 }, { "epoch": 1.46, "learning_rate": 2.7299743052201045e-06, "loss": 0.6288, "step": 12257 }, { "epoch": 1.46, "learning_rate": 2.7296542427345856e-06, "loss": 0.7077, "step": 12258 }, { "epoch": 1.46, "learning_rate": 2.7293341764528625e-06, "loss": 0.7466, "step": 12259 }, { "epoch": 1.46, "learning_rate": 2.7290141063802265e-06, "loss": 0.6893, "step": 12260 }, { "epoch": 1.46, "learning_rate": 2.7286940325219686e-06, "loss": 0.3912, "step": 12261 }, { "epoch": 1.46, "learning_rate": 2.728373954883379e-06, "loss": 0.5134, "step": 12262 }, { "epoch": 1.46, "learning_rate": 2.7280538734697476e-06, "loss": 0.575, "step": 12263 }, { "epoch": 1.46, "learning_rate": 2.727733788286368e-06, "loss": 0.7905, "step": 12264 }, { "epoch": 1.46, "learning_rate": 2.7274136993385287e-06, "loss": 0.8013, "step": 12265 }, { "epoch": 1.46, "learning_rate": 2.7270936066315224e-06, "loss": 0.5716, "step": 12266 }, { "epoch": 1.46, "learning_rate": 2.72677351017064e-06, "loss": 0.3608, "step": 12267 }, { "epoch": 1.46, "learning_rate": 2.7264534099611723e-06, "loss": 0.4899, "step": 12268 }, { "epoch": 1.46, "learning_rate": 2.7261333060084106e-06, "loss": 0.5925, "step": 12269 }, { "epoch": 1.46, "learning_rate": 2.7258131983176454e-06, "loss": 0.7487, "step": 12270 }, { "epoch": 1.46, "learning_rate": 2.7254930868941704e-06, "loss": 0.936, "step": 12271 }, { "epoch": 1.46, "learning_rate": 2.7251729717432757e-06, "loss": 0.7401, "step": 12272 }, { "epoch": 1.46, "learning_rate": 2.7248528528702527e-06, "loss": 0.4386, "step": 12273 }, { "epoch": 1.46, "learning_rate": 2.7245327302803927e-06, "loss": 0.532, "step": 12274 }, { "epoch": 1.46, "learning_rate": 2.7242126039789886e-06, "loss": 0.6379, "step": 12275 }, { "epoch": 1.46, "learning_rate": 2.7238924739713304e-06, "loss": 0.734, "step": 12276 }, { "epoch": 1.46, "learning_rate": 2.723572340262711e-06, "loss": 0.713, "step": 12277 }, { "epoch": 1.46, "learning_rate": 2.723252202858423e-06, "loss": 0.8365, "step": 12278 }, { "epoch": 1.46, "learning_rate": 2.722932061763757e-06, "loss": 0.344, "step": 12279 }, { "epoch": 1.46, "learning_rate": 2.722611916984004e-06, "loss": 0.5161, "step": 12280 }, { "epoch": 1.46, "learning_rate": 2.7222917685244586e-06, "loss": 0.5544, "step": 12281 }, { "epoch": 1.46, "learning_rate": 2.7219716163904118e-06, "loss": 0.6671, "step": 12282 }, { "epoch": 1.46, "learning_rate": 2.7216514605871554e-06, "loss": 0.6895, "step": 12283 }, { "epoch": 1.46, "learning_rate": 2.7213313011199814e-06, "loss": 0.8962, "step": 12284 }, { "epoch": 1.46, "learning_rate": 2.7210111379941824e-06, "loss": 0.4372, "step": 12285 }, { "epoch": 1.46, "learning_rate": 2.7206909712150507e-06, "loss": 0.4183, "step": 12286 }, { "epoch": 1.46, "learning_rate": 2.720370800787879e-06, "loss": 0.4827, "step": 12287 }, { "epoch": 1.46, "learning_rate": 2.7200506267179594e-06, "loss": 0.6753, "step": 12288 }, { "epoch": 1.46, "learning_rate": 2.7197304490105847e-06, "loss": 0.6474, "step": 12289 }, { "epoch": 1.46, "learning_rate": 2.7194102676710473e-06, "loss": 0.9274, "step": 12290 }, { "epoch": 1.46, "learning_rate": 2.7190900827046384e-06, "loss": 0.5138, "step": 12291 }, { "epoch": 1.46, "learning_rate": 2.7187698941166533e-06, "loss": 0.5614, "step": 12292 }, { "epoch": 1.46, "learning_rate": 2.7184497019123835e-06, "loss": 0.6314, "step": 12293 }, { "epoch": 1.46, "learning_rate": 2.718129506097122e-06, "loss": 0.5848, "step": 12294 }, { "epoch": 1.46, "learning_rate": 2.7178093066761603e-06, "loss": 0.6863, "step": 12295 }, { "epoch": 1.46, "learning_rate": 2.717489103654794e-06, "loss": 0.7493, "step": 12296 }, { "epoch": 1.46, "learning_rate": 2.717168897038313e-06, "loss": 0.8137, "step": 12297 }, { "epoch": 1.46, "learning_rate": 2.716848686832013e-06, "loss": 0.438, "step": 12298 }, { "epoch": 1.46, "learning_rate": 2.716528473041186e-06, "loss": 0.5307, "step": 12299 }, { "epoch": 1.46, "learning_rate": 2.716208255671125e-06, "loss": 0.6146, "step": 12300 }, { "epoch": 1.46, "learning_rate": 2.7158880347271233e-06, "loss": 0.6822, "step": 12301 }, { "epoch": 1.47, "learning_rate": 2.715567810214474e-06, "loss": 0.7014, "step": 12302 }, { "epoch": 1.47, "learning_rate": 2.715247582138471e-06, "loss": 0.8281, "step": 12303 }, { "epoch": 1.47, "learning_rate": 2.7149273505044076e-06, "loss": 0.4031, "step": 12304 }, { "epoch": 1.47, "learning_rate": 2.7146071153175767e-06, "loss": 0.6559, "step": 12305 }, { "epoch": 1.47, "learning_rate": 2.714286876583272e-06, "loss": 0.6108, "step": 12306 }, { "epoch": 1.47, "learning_rate": 2.713966634306788e-06, "loss": 0.5193, "step": 12307 }, { "epoch": 1.47, "learning_rate": 2.7136463884934173e-06, "loss": 0.7666, "step": 12308 }, { "epoch": 1.47, "learning_rate": 2.7133261391484535e-06, "loss": 0.8635, "step": 12309 }, { "epoch": 1.47, "learning_rate": 2.713005886277191e-06, "loss": 0.4387, "step": 12310 }, { "epoch": 1.47, "learning_rate": 2.7126856298849234e-06, "loss": 0.5017, "step": 12311 }, { "epoch": 1.47, "learning_rate": 2.7123653699769438e-06, "loss": 0.5395, "step": 12312 }, { "epoch": 1.47, "learning_rate": 2.7120451065585475e-06, "loss": 0.6371, "step": 12313 }, { "epoch": 1.47, "learning_rate": 2.711724839635028e-06, "loss": 0.5452, "step": 12314 }, { "epoch": 1.47, "learning_rate": 2.7114045692116786e-06, "loss": 0.8365, "step": 12315 }, { "epoch": 1.47, "learning_rate": 2.7110842952937943e-06, "loss": 0.6054, "step": 12316 }, { "epoch": 1.47, "learning_rate": 2.7107640178866683e-06, "loss": 0.5094, "step": 12317 }, { "epoch": 1.47, "learning_rate": 2.7104437369955968e-06, "loss": 0.5793, "step": 12318 }, { "epoch": 1.47, "learning_rate": 2.7101234526258714e-06, "loss": 0.5779, "step": 12319 }, { "epoch": 1.47, "learning_rate": 2.7098031647827875e-06, "loss": 0.6389, "step": 12320 }, { "epoch": 1.47, "learning_rate": 2.7094828734716403e-06, "loss": 0.7407, "step": 12321 }, { "epoch": 1.47, "learning_rate": 2.7091625786977237e-06, "loss": 0.9446, "step": 12322 }, { "epoch": 1.47, "learning_rate": 2.7088422804663317e-06, "loss": 0.3761, "step": 12323 }, { "epoch": 1.47, "learning_rate": 2.70852197878276e-06, "loss": 0.4644, "step": 12324 }, { "epoch": 1.47, "learning_rate": 2.708201673652302e-06, "loss": 0.5864, "step": 12325 }, { "epoch": 1.47, "learning_rate": 2.7078813650802526e-06, "loss": 0.6361, "step": 12326 }, { "epoch": 1.47, "learning_rate": 2.7075610530719065e-06, "loss": 0.8229, "step": 12327 }, { "epoch": 1.47, "learning_rate": 2.70724073763256e-06, "loss": 0.7197, "step": 12328 }, { "epoch": 1.47, "learning_rate": 2.706920418767506e-06, "loss": 0.4867, "step": 12329 }, { "epoch": 1.47, "learning_rate": 2.70660009648204e-06, "loss": 0.5603, "step": 12330 }, { "epoch": 1.47, "learning_rate": 2.7062797707814577e-06, "loss": 0.5652, "step": 12331 }, { "epoch": 1.47, "learning_rate": 2.7059594416710527e-06, "loss": 0.615, "step": 12332 }, { "epoch": 1.47, "learning_rate": 2.705639109156121e-06, "loss": 0.6595, "step": 12333 }, { "epoch": 1.47, "learning_rate": 2.705318773241958e-06, "loss": 0.8021, "step": 12334 }, { "epoch": 1.47, "learning_rate": 2.704998433933858e-06, "loss": 0.569, "step": 12335 }, { "epoch": 1.47, "learning_rate": 2.7046780912371175e-06, "loss": 0.4451, "step": 12336 }, { "epoch": 1.47, "learning_rate": 2.7043577451570304e-06, "loss": 0.6458, "step": 12337 }, { "epoch": 1.47, "learning_rate": 2.704037395698892e-06, "loss": 0.5626, "step": 12338 }, { "epoch": 1.47, "learning_rate": 2.7037170428679985e-06, "loss": 0.7962, "step": 12339 }, { "epoch": 1.47, "learning_rate": 2.703396686669646e-06, "loss": 0.8112, "step": 12340 }, { "epoch": 1.47, "learning_rate": 2.7030763271091285e-06, "loss": 0.6333, "step": 12341 }, { "epoch": 1.47, "learning_rate": 2.702755964191742e-06, "loss": 0.5226, "step": 12342 }, { "epoch": 1.47, "learning_rate": 2.702435597922784e-06, "loss": 0.5287, "step": 12343 }, { "epoch": 1.47, "learning_rate": 2.7021152283075476e-06, "loss": 0.4899, "step": 12344 }, { "epoch": 1.47, "learning_rate": 2.701794855351329e-06, "loss": 0.8141, "step": 12345 }, { "epoch": 1.47, "learning_rate": 2.7014744790594255e-06, "loss": 0.8145, "step": 12346 }, { "epoch": 1.47, "learning_rate": 2.7011540994371314e-06, "loss": 0.8786, "step": 12347 }, { "epoch": 1.47, "learning_rate": 2.7008337164897434e-06, "loss": 0.4058, "step": 12348 }, { "epoch": 1.47, "learning_rate": 2.7005133302225567e-06, "loss": 0.4536, "step": 12349 }, { "epoch": 1.47, "learning_rate": 2.7001929406408683e-06, "loss": 0.5691, "step": 12350 }, { "epoch": 1.47, "learning_rate": 2.6998725477499734e-06, "loss": 0.618, "step": 12351 }, { "epoch": 1.47, "learning_rate": 2.699552151555169e-06, "loss": 0.7586, "step": 12352 }, { "epoch": 1.47, "learning_rate": 2.6992317520617507e-06, "loss": 1.087, "step": 12353 }, { "epoch": 1.47, "learning_rate": 2.6989113492750148e-06, "loss": 0.4621, "step": 12354 }, { "epoch": 1.47, "learning_rate": 2.6985909432002565e-06, "loss": 0.5764, "step": 12355 }, { "epoch": 1.47, "learning_rate": 2.6982705338427743e-06, "loss": 0.4814, "step": 12356 }, { "epoch": 1.47, "learning_rate": 2.697950121207864e-06, "loss": 0.5777, "step": 12357 }, { "epoch": 1.47, "learning_rate": 2.697629705300821e-06, "loss": 0.7502, "step": 12358 }, { "epoch": 1.47, "learning_rate": 2.697309286126942e-06, "loss": 0.8766, "step": 12359 }, { "epoch": 1.47, "learning_rate": 2.696988863691524e-06, "loss": 0.4163, "step": 12360 }, { "epoch": 1.47, "learning_rate": 2.696668437999865e-06, "loss": 0.4662, "step": 12361 }, { "epoch": 1.47, "learning_rate": 2.6963480090572585e-06, "loss": 0.6434, "step": 12362 }, { "epoch": 1.47, "learning_rate": 2.6960275768690043e-06, "loss": 0.5956, "step": 12363 }, { "epoch": 1.47, "learning_rate": 2.695707141440397e-06, "loss": 0.7199, "step": 12364 }, { "epoch": 1.47, "learning_rate": 2.6953867027767346e-06, "loss": 0.8177, "step": 12365 }, { "epoch": 1.47, "learning_rate": 2.695066260883313e-06, "loss": 0.5065, "step": 12366 }, { "epoch": 1.47, "learning_rate": 2.69474581576543e-06, "loss": 0.501, "step": 12367 }, { "epoch": 1.47, "learning_rate": 2.694425367428382e-06, "loss": 0.6177, "step": 12368 }, { "epoch": 1.47, "learning_rate": 2.6941049158774667e-06, "loss": 0.5798, "step": 12369 }, { "epoch": 1.47, "learning_rate": 2.693784461117981e-06, "loss": 0.6662, "step": 12370 }, { "epoch": 1.47, "learning_rate": 2.693464003155222e-06, "loss": 0.9136, "step": 12371 }, { "epoch": 1.47, "learning_rate": 2.693143541994486e-06, "loss": 0.7169, "step": 12372 }, { "epoch": 1.47, "learning_rate": 2.6928230776410723e-06, "loss": 0.3271, "step": 12373 }, { "epoch": 1.47, "learning_rate": 2.692502610100276e-06, "loss": 0.6021, "step": 12374 }, { "epoch": 1.47, "learning_rate": 2.6921821393773963e-06, "loss": 0.5938, "step": 12375 }, { "epoch": 1.47, "learning_rate": 2.6918616654777285e-06, "loss": 0.6932, "step": 12376 }, { "epoch": 1.47, "learning_rate": 2.691541188406572e-06, "loss": 0.8692, "step": 12377 }, { "epoch": 1.47, "learning_rate": 2.691220708169224e-06, "loss": 0.867, "step": 12378 }, { "epoch": 1.47, "learning_rate": 2.6909002247709807e-06, "loss": 0.4263, "step": 12379 }, { "epoch": 1.47, "learning_rate": 2.6905797382171412e-06, "loss": 0.5716, "step": 12380 }, { "epoch": 1.47, "learning_rate": 2.690259248513002e-06, "loss": 0.6016, "step": 12381 }, { "epoch": 1.47, "learning_rate": 2.6899387556638624e-06, "loss": 0.7658, "step": 12382 }, { "epoch": 1.47, "learning_rate": 2.6896182596750187e-06, "loss": 0.7355, "step": 12383 }, { "epoch": 1.47, "learning_rate": 2.6892977605517697e-06, "loss": 0.8965, "step": 12384 }, { "epoch": 1.47, "learning_rate": 2.6889772582994116e-06, "loss": 0.6216, "step": 12385 }, { "epoch": 1.48, "learning_rate": 2.6886567529232456e-06, "loss": 0.3769, "step": 12386 }, { "epoch": 1.48, "learning_rate": 2.688336244428566e-06, "loss": 0.5393, "step": 12387 }, { "epoch": 1.48, "learning_rate": 2.688015732820673e-06, "loss": 0.6972, "step": 12388 }, { "epoch": 1.48, "learning_rate": 2.6876952181048643e-06, "loss": 0.6924, "step": 12389 }, { "epoch": 1.48, "learning_rate": 2.6873747002864373e-06, "loss": 0.7912, "step": 12390 }, { "epoch": 1.48, "learning_rate": 2.6870541793706912e-06, "loss": 0.5978, "step": 12391 }, { "epoch": 1.48, "learning_rate": 2.686733655362924e-06, "loss": 0.5114, "step": 12392 }, { "epoch": 1.48, "learning_rate": 2.6864131282684336e-06, "loss": 0.6304, "step": 12393 }, { "epoch": 1.48, "learning_rate": 2.686092598092519e-06, "loss": 0.6288, "step": 12394 }, { "epoch": 1.48, "learning_rate": 2.6857720648404774e-06, "loss": 0.6559, "step": 12395 }, { "epoch": 1.48, "learning_rate": 2.6854515285176086e-06, "loss": 0.8014, "step": 12396 }, { "epoch": 1.48, "learning_rate": 2.6851309891292106e-06, "loss": 0.7809, "step": 12397 }, { "epoch": 1.48, "learning_rate": 2.684810446680581e-06, "loss": 0.4807, "step": 12398 }, { "epoch": 1.48, "learning_rate": 2.6844899011770193e-06, "loss": 0.442, "step": 12399 }, { "epoch": 1.48, "learning_rate": 2.6841693526238244e-06, "loss": 0.6577, "step": 12400 }, { "epoch": 1.48, "learning_rate": 2.683848801026295e-06, "loss": 0.5903, "step": 12401 }, { "epoch": 1.48, "learning_rate": 2.6835282463897283e-06, "loss": 0.7993, "step": 12402 }, { "epoch": 1.48, "learning_rate": 2.6832076887194252e-06, "loss": 0.9341, "step": 12403 }, { "epoch": 1.48, "learning_rate": 2.6828871280206838e-06, "loss": 0.4158, "step": 12404 }, { "epoch": 1.48, "learning_rate": 2.6825665642988025e-06, "loss": 0.5158, "step": 12405 }, { "epoch": 1.48, "learning_rate": 2.6822459975590805e-06, "loss": 0.5834, "step": 12406 }, { "epoch": 1.48, "learning_rate": 2.6819254278068172e-06, "loss": 0.5484, "step": 12407 }, { "epoch": 1.48, "learning_rate": 2.6816048550473113e-06, "loss": 0.7986, "step": 12408 }, { "epoch": 1.48, "learning_rate": 2.6812842792858617e-06, "loss": 0.8472, "step": 12409 }, { "epoch": 1.48, "learning_rate": 2.680963700527768e-06, "loss": 0.441, "step": 12410 }, { "epoch": 1.48, "learning_rate": 2.680643118778329e-06, "loss": 0.4866, "step": 12411 }, { "epoch": 1.48, "learning_rate": 2.680322534042844e-06, "loss": 0.5757, "step": 12412 }, { "epoch": 1.48, "learning_rate": 2.680001946326612e-06, "loss": 0.6564, "step": 12413 }, { "epoch": 1.48, "learning_rate": 2.679681355634933e-06, "loss": 0.7667, "step": 12414 }, { "epoch": 1.48, "learning_rate": 2.6793607619731066e-06, "loss": 0.8815, "step": 12415 }, { "epoch": 1.48, "learning_rate": 2.6790401653464314e-06, "loss": 0.4316, "step": 12416 }, { "epoch": 1.48, "learning_rate": 2.6787195657602066e-06, "loss": 0.5493, "step": 12417 }, { "epoch": 1.48, "learning_rate": 2.6783989632197337e-06, "loss": 0.5676, "step": 12418 }, { "epoch": 1.48, "learning_rate": 2.67807835773031e-06, "loss": 0.567, "step": 12419 }, { "epoch": 1.48, "learning_rate": 2.677757749297236e-06, "loss": 0.6544, "step": 12420 }, { "epoch": 1.48, "learning_rate": 2.6774371379258124e-06, "loss": 0.8807, "step": 12421 }, { "epoch": 1.48, "learning_rate": 2.6771165236213377e-06, "loss": 0.7847, "step": 12422 }, { "epoch": 1.48, "learning_rate": 2.6767959063891114e-06, "loss": 0.5591, "step": 12423 }, { "epoch": 1.48, "learning_rate": 2.6764752862344338e-06, "loss": 0.5222, "step": 12424 }, { "epoch": 1.48, "learning_rate": 2.676154663162606e-06, "loss": 0.532, "step": 12425 }, { "epoch": 1.48, "learning_rate": 2.6758340371789264e-06, "loss": 0.5905, "step": 12426 }, { "epoch": 1.48, "learning_rate": 2.675513408288695e-06, "loss": 0.6605, "step": 12427 }, { "epoch": 1.48, "learning_rate": 2.6751927764972124e-06, "loss": 0.7496, "step": 12428 }, { "epoch": 1.48, "learning_rate": 2.6748721418097797e-06, "loss": 0.4409, "step": 12429 }, { "epoch": 1.48, "learning_rate": 2.674551504231694e-06, "loss": 0.5005, "step": 12430 }, { "epoch": 1.48, "learning_rate": 2.674230863768258e-06, "loss": 0.497, "step": 12431 }, { "epoch": 1.48, "learning_rate": 2.6739102204247715e-06, "loss": 0.7513, "step": 12432 }, { "epoch": 1.48, "learning_rate": 2.6735895742065344e-06, "loss": 0.8373, "step": 12433 }, { "epoch": 1.48, "learning_rate": 2.6732689251188465e-06, "loss": 0.8293, "step": 12434 }, { "epoch": 1.48, "learning_rate": 2.6729482731670094e-06, "loss": 0.5278, "step": 12435 }, { "epoch": 1.48, "learning_rate": 2.6726276183563223e-06, "loss": 0.4151, "step": 12436 }, { "epoch": 1.48, "learning_rate": 2.6723069606920865e-06, "loss": 0.6344, "step": 12437 }, { "epoch": 1.48, "learning_rate": 2.6719863001796013e-06, "loss": 0.5443, "step": 12438 }, { "epoch": 1.48, "learning_rate": 2.6716656368241695e-06, "loss": 0.6459, "step": 12439 }, { "epoch": 1.48, "learning_rate": 2.6713449706310895e-06, "loss": 0.774, "step": 12440 }, { "epoch": 1.48, "learning_rate": 2.6710243016056636e-06, "loss": 0.696, "step": 12441 }, { "epoch": 1.48, "learning_rate": 2.6707036297531906e-06, "loss": 0.5001, "step": 12442 }, { "epoch": 1.48, "learning_rate": 2.6703829550789735e-06, "loss": 0.519, "step": 12443 }, { "epoch": 1.48, "learning_rate": 2.6700622775883105e-06, "loss": 0.5365, "step": 12444 }, { "epoch": 1.48, "learning_rate": 2.6697415972865043e-06, "loss": 0.6656, "step": 12445 }, { "epoch": 1.48, "learning_rate": 2.6694209141788553e-06, "loss": 0.8278, "step": 12446 }, { "epoch": 1.48, "learning_rate": 2.6691002282706647e-06, "loss": 1.0274, "step": 12447 }, { "epoch": 1.48, "learning_rate": 2.6687795395672327e-06, "loss": 0.374, "step": 12448 }, { "epoch": 1.48, "learning_rate": 2.6684588480738606e-06, "loss": 0.588, "step": 12449 }, { "epoch": 1.48, "learning_rate": 2.6681381537958507e-06, "loss": 0.4857, "step": 12450 }, { "epoch": 1.48, "learning_rate": 2.6678174567385017e-06, "loss": 0.619, "step": 12451 }, { "epoch": 1.48, "learning_rate": 2.6674967569071163e-06, "loss": 0.8938, "step": 12452 }, { "epoch": 1.48, "learning_rate": 2.6671760543069963e-06, "loss": 0.8874, "step": 12453 }, { "epoch": 1.48, "learning_rate": 2.666855348943441e-06, "loss": 0.417, "step": 12454 }, { "epoch": 1.48, "learning_rate": 2.6665346408217533e-06, "loss": 0.527, "step": 12455 }, { "epoch": 1.48, "learning_rate": 2.666213929947234e-06, "loss": 0.5693, "step": 12456 }, { "epoch": 1.48, "learning_rate": 2.665893216325185e-06, "loss": 0.5401, "step": 12457 }, { "epoch": 1.48, "learning_rate": 2.6655724999609066e-06, "loss": 0.7104, "step": 12458 }, { "epoch": 1.48, "learning_rate": 2.6652517808597013e-06, "loss": 0.8915, "step": 12459 }, { "epoch": 1.48, "learning_rate": 2.664931059026869e-06, "loss": 0.4755, "step": 12460 }, { "epoch": 1.48, "learning_rate": 2.6646103344677145e-06, "loss": 0.5706, "step": 12461 }, { "epoch": 1.48, "learning_rate": 2.664289607187536e-06, "loss": 0.5088, "step": 12462 }, { "epoch": 1.48, "learning_rate": 2.663968877191636e-06, "loss": 0.6728, "step": 12463 }, { "epoch": 1.48, "learning_rate": 2.663648144485318e-06, "loss": 0.7217, "step": 12464 }, { "epoch": 1.48, "learning_rate": 2.663327409073882e-06, "loss": 0.8414, "step": 12465 }, { "epoch": 1.48, "learning_rate": 2.6630066709626297e-06, "loss": 0.5673, "step": 12466 }, { "epoch": 1.48, "learning_rate": 2.6626859301568637e-06, "loss": 0.4473, "step": 12467 }, { "epoch": 1.48, "learning_rate": 2.662365186661886e-06, "loss": 0.5144, "step": 12468 }, { "epoch": 1.48, "learning_rate": 2.6620444404829976e-06, "loss": 0.6662, "step": 12469 }, { "epoch": 1.49, "learning_rate": 2.661723691625501e-06, "loss": 0.6851, "step": 12470 }, { "epoch": 1.49, "learning_rate": 2.661402940094699e-06, "loss": 0.71, "step": 12471 }, { "epoch": 1.49, "learning_rate": 2.6610821858958922e-06, "loss": 0.8802, "step": 12472 }, { "epoch": 1.49, "learning_rate": 2.6607614290343838e-06, "loss": 0.4735, "step": 12473 }, { "epoch": 1.49, "learning_rate": 2.660440669515475e-06, "loss": 0.5824, "step": 12474 }, { "epoch": 1.49, "learning_rate": 2.660119907344469e-06, "loss": 0.5253, "step": 12475 }, { "epoch": 1.49, "learning_rate": 2.6597991425266673e-06, "loss": 0.6898, "step": 12476 }, { "epoch": 1.49, "learning_rate": 2.659478375067372e-06, "loss": 0.7897, "step": 12477 }, { "epoch": 1.49, "learning_rate": 2.659157604971886e-06, "loss": 0.9527, "step": 12478 }, { "epoch": 1.49, "learning_rate": 2.6588368322455117e-06, "loss": 0.39, "step": 12479 }, { "epoch": 1.49, "learning_rate": 2.658516056893551e-06, "loss": 0.5839, "step": 12480 }, { "epoch": 1.49, "learning_rate": 2.658195278921306e-06, "loss": 0.5767, "step": 12481 }, { "epoch": 1.49, "learning_rate": 2.6578744983340804e-06, "loss": 0.5943, "step": 12482 }, { "epoch": 1.49, "learning_rate": 2.6575537151371762e-06, "loss": 0.8292, "step": 12483 }, { "epoch": 1.49, "learning_rate": 2.6572329293358956e-06, "loss": 0.7278, "step": 12484 }, { "epoch": 1.49, "learning_rate": 2.6569121409355415e-06, "loss": 0.3381, "step": 12485 }, { "epoch": 1.49, "learning_rate": 2.656591349941417e-06, "loss": 0.5367, "step": 12486 }, { "epoch": 1.49, "learning_rate": 2.6562705563588237e-06, "loss": 0.5583, "step": 12487 }, { "epoch": 1.49, "learning_rate": 2.655949760193065e-06, "loss": 0.5353, "step": 12488 }, { "epoch": 1.49, "learning_rate": 2.655628961449444e-06, "loss": 0.6938, "step": 12489 }, { "epoch": 1.49, "learning_rate": 2.6553081601332634e-06, "loss": 0.8989, "step": 12490 }, { "epoch": 1.49, "learning_rate": 2.654987356249826e-06, "loss": 0.5779, "step": 12491 }, { "epoch": 1.49, "learning_rate": 2.654666549804434e-06, "loss": 0.4445, "step": 12492 }, { "epoch": 1.49, "learning_rate": 2.6543457408023914e-06, "loss": 0.6411, "step": 12493 }, { "epoch": 1.49, "learning_rate": 2.654024929249e-06, "loss": 0.5079, "step": 12494 }, { "epoch": 1.49, "learning_rate": 2.6537041151495647e-06, "loss": 0.5998, "step": 12495 }, { "epoch": 1.49, "learning_rate": 2.6533832985093866e-06, "loss": 0.8156, "step": 12496 }, { "epoch": 1.49, "learning_rate": 2.65306247933377e-06, "loss": 0.7826, "step": 12497 }, { "epoch": 1.49, "learning_rate": 2.6527416576280176e-06, "loss": 0.3528, "step": 12498 }, { "epoch": 1.49, "learning_rate": 2.6524208333974332e-06, "loss": 0.5628, "step": 12499 }, { "epoch": 1.49, "learning_rate": 2.6521000066473197e-06, "loss": 0.4327, "step": 12500 }, { "epoch": 1.49, "learning_rate": 2.65177917738298e-06, "loss": 0.5961, "step": 12501 }, { "epoch": 1.49, "learning_rate": 2.651458345609717e-06, "loss": 0.7326, "step": 12502 }, { "epoch": 1.49, "learning_rate": 2.651137511332836e-06, "loss": 0.8373, "step": 12503 }, { "epoch": 1.49, "learning_rate": 2.6508166745576393e-06, "loss": 0.5209, "step": 12504 }, { "epoch": 1.49, "learning_rate": 2.65049583528943e-06, "loss": 0.5945, "step": 12505 }, { "epoch": 1.49, "learning_rate": 2.650174993533512e-06, "loss": 0.6308, "step": 12506 }, { "epoch": 1.49, "learning_rate": 2.6498541492951883e-06, "loss": 0.6655, "step": 12507 }, { "epoch": 1.49, "learning_rate": 2.6495333025797632e-06, "loss": 0.7969, "step": 12508 }, { "epoch": 1.49, "learning_rate": 2.64921245339254e-06, "loss": 0.9068, "step": 12509 }, { "epoch": 1.49, "learning_rate": 2.648891601738822e-06, "loss": 0.4852, "step": 12510 }, { "epoch": 1.49, "learning_rate": 2.648570747623914e-06, "loss": 0.431, "step": 12511 }, { "epoch": 1.49, "learning_rate": 2.6482498910531197e-06, "loss": 0.6135, "step": 12512 }, { "epoch": 1.49, "learning_rate": 2.647929032031741e-06, "loss": 0.609, "step": 12513 }, { "epoch": 1.49, "learning_rate": 2.6476081705650837e-06, "loss": 0.7488, "step": 12514 }, { "epoch": 1.49, "learning_rate": 2.6472873066584503e-06, "loss": 0.7219, "step": 12515 }, { "epoch": 1.49, "learning_rate": 2.6469664403171457e-06, "loss": 0.8404, "step": 12516 }, { "epoch": 1.49, "learning_rate": 2.6466455715464735e-06, "loss": 0.5606, "step": 12517 }, { "epoch": 1.49, "learning_rate": 2.6463247003517383e-06, "loss": 0.5174, "step": 12518 }, { "epoch": 1.49, "learning_rate": 2.6460038267382424e-06, "loss": 0.5592, "step": 12519 }, { "epoch": 1.49, "learning_rate": 2.645682950711292e-06, "loss": 0.6505, "step": 12520 }, { "epoch": 1.49, "learning_rate": 2.6453620722761897e-06, "loss": 0.8458, "step": 12521 }, { "epoch": 1.49, "learning_rate": 2.6450411914382406e-06, "loss": 0.9356, "step": 12522 }, { "epoch": 1.49, "learning_rate": 2.6447203082027486e-06, "loss": 0.4292, "step": 12523 }, { "epoch": 1.49, "learning_rate": 2.6443994225750162e-06, "loss": 0.61, "step": 12524 }, { "epoch": 1.49, "learning_rate": 2.6440785345603505e-06, "loss": 0.5616, "step": 12525 }, { "epoch": 1.49, "learning_rate": 2.6437576441640543e-06, "loss": 0.717, "step": 12526 }, { "epoch": 1.49, "learning_rate": 2.6434367513914325e-06, "loss": 0.7277, "step": 12527 }, { "epoch": 1.49, "learning_rate": 2.643115856247789e-06, "loss": 0.9649, "step": 12528 }, { "epoch": 1.49, "learning_rate": 2.642794958738428e-06, "loss": 0.3294, "step": 12529 }, { "epoch": 1.49, "learning_rate": 2.642474058868655e-06, "loss": 0.3426, "step": 12530 }, { "epoch": 1.49, "learning_rate": 2.642153156643773e-06, "loss": 0.5913, "step": 12531 }, { "epoch": 1.49, "learning_rate": 2.641832252069088e-06, "loss": 0.6674, "step": 12532 }, { "epoch": 1.49, "learning_rate": 2.641511345149904e-06, "loss": 0.7173, "step": 12533 }, { "epoch": 1.49, "learning_rate": 2.6411904358915254e-06, "loss": 0.8331, "step": 12534 }, { "epoch": 1.49, "learning_rate": 2.6408695242992567e-06, "loss": 0.5217, "step": 12535 }, { "epoch": 1.49, "learning_rate": 2.6405486103784034e-06, "loss": 0.5208, "step": 12536 }, { "epoch": 1.49, "learning_rate": 2.6402276941342696e-06, "loss": 0.5069, "step": 12537 }, { "epoch": 1.49, "learning_rate": 2.6399067755721606e-06, "loss": 0.7248, "step": 12538 }, { "epoch": 1.49, "learning_rate": 2.6395858546973803e-06, "loss": 0.6938, "step": 12539 }, { "epoch": 1.49, "learning_rate": 2.639264931515235e-06, "loss": 0.9301, "step": 12540 }, { "epoch": 1.49, "learning_rate": 2.6389440060310277e-06, "loss": 0.6234, "step": 12541 }, { "epoch": 1.49, "learning_rate": 2.638623078250065e-06, "loss": 0.5054, "step": 12542 }, { "epoch": 1.49, "learning_rate": 2.6383021481776503e-06, "loss": 0.5783, "step": 12543 }, { "epoch": 1.49, "learning_rate": 2.6379812158190907e-06, "loss": 0.5883, "step": 12544 }, { "epoch": 1.49, "learning_rate": 2.637660281179689e-06, "loss": 0.6836, "step": 12545 }, { "epoch": 1.49, "learning_rate": 2.6373393442647517e-06, "loss": 0.8981, "step": 12546 }, { "epoch": 1.49, "learning_rate": 2.637018405079584e-06, "loss": 0.7968, "step": 12547 }, { "epoch": 1.49, "learning_rate": 2.6366974636294902e-06, "loss": 0.3938, "step": 12548 }, { "epoch": 1.49, "learning_rate": 2.6363765199197748e-06, "loss": 0.6342, "step": 12549 }, { "epoch": 1.49, "learning_rate": 2.636055573955746e-06, "loss": 0.5927, "step": 12550 }, { "epoch": 1.49, "learning_rate": 2.6357346257427058e-06, "loss": 0.7052, "step": 12551 }, { "epoch": 1.49, "learning_rate": 2.6354136752859612e-06, "loss": 0.8017, "step": 12552 }, { "epoch": 1.49, "learning_rate": 2.6350927225908174e-06, "loss": 0.7421, "step": 12553 }, { "epoch": 1.5, "learning_rate": 2.6347717676625796e-06, "loss": 0.4239, "step": 12554 }, { "epoch": 1.5, "learning_rate": 2.6344508105065532e-06, "loss": 0.5827, "step": 12555 }, { "epoch": 1.5, "learning_rate": 2.634129851128043e-06, "loss": 0.5707, "step": 12556 }, { "epoch": 1.5, "learning_rate": 2.6338088895323547e-06, "loss": 0.5701, "step": 12557 }, { "epoch": 1.5, "learning_rate": 2.633487925724796e-06, "loss": 0.7678, "step": 12558 }, { "epoch": 1.5, "learning_rate": 2.6331669597106697e-06, "loss": 0.8213, "step": 12559 }, { "epoch": 1.5, "learning_rate": 2.6328459914952814e-06, "loss": 0.41, "step": 12560 }, { "epoch": 1.5, "learning_rate": 2.63252502108394e-06, "loss": 0.4977, "step": 12561 }, { "epoch": 1.5, "learning_rate": 2.6322040484819468e-06, "loss": 0.5506, "step": 12562 }, { "epoch": 1.5, "learning_rate": 2.6318830736946107e-06, "loss": 0.4883, "step": 12563 }, { "epoch": 1.5, "learning_rate": 2.631562096727236e-06, "loss": 0.7626, "step": 12564 }, { "epoch": 1.5, "learning_rate": 2.631241117585129e-06, "loss": 0.9553, "step": 12565 }, { "epoch": 1.5, "learning_rate": 2.630920136273594e-06, "loss": 0.6867, "step": 12566 }, { "epoch": 1.5, "learning_rate": 2.6305991527979395e-06, "loss": 0.4651, "step": 12567 }, { "epoch": 1.5, "learning_rate": 2.6302781671634703e-06, "loss": 0.5167, "step": 12568 }, { "epoch": 1.5, "learning_rate": 2.6299571793754914e-06, "loss": 0.6344, "step": 12569 }, { "epoch": 1.5, "learning_rate": 2.6296361894393095e-06, "loss": 0.6859, "step": 12570 }, { "epoch": 1.5, "learning_rate": 2.6293151973602303e-06, "loss": 0.7432, "step": 12571 }, { "epoch": 1.5, "learning_rate": 2.628994203143561e-06, "loss": 0.7544, "step": 12572 }, { "epoch": 1.5, "learning_rate": 2.6286732067946058e-06, "loss": 0.382, "step": 12573 }, { "epoch": 1.5, "learning_rate": 2.628352208318672e-06, "loss": 0.5527, "step": 12574 }, { "epoch": 1.5, "learning_rate": 2.6280312077210655e-06, "loss": 0.6031, "step": 12575 }, { "epoch": 1.5, "learning_rate": 2.6277102050070925e-06, "loss": 0.6449, "step": 12576 }, { "epoch": 1.5, "learning_rate": 2.6273892001820585e-06, "loss": 0.9357, "step": 12577 }, { "epoch": 1.5, "learning_rate": 2.6270681932512706e-06, "loss": 0.9821, "step": 12578 }, { "epoch": 1.5, "learning_rate": 2.626747184220035e-06, "loss": 0.2826, "step": 12579 }, { "epoch": 1.5, "learning_rate": 2.626426173093658e-06, "loss": 0.4323, "step": 12580 }, { "epoch": 1.5, "learning_rate": 2.6261051598774447e-06, "loss": 0.6112, "step": 12581 }, { "epoch": 1.5, "learning_rate": 2.6257841445767036e-06, "loss": 0.638, "step": 12582 }, { "epoch": 1.5, "learning_rate": 2.6254631271967398e-06, "loss": 0.8026, "step": 12583 }, { "epoch": 1.5, "learning_rate": 2.6251421077428598e-06, "loss": 0.8793, "step": 12584 }, { "epoch": 1.5, "learning_rate": 2.6248210862203703e-06, "loss": 0.4839, "step": 12585 }, { "epoch": 1.5, "learning_rate": 2.624500062634578e-06, "loss": 0.5114, "step": 12586 }, { "epoch": 1.5, "learning_rate": 2.6241790369907892e-06, "loss": 0.556, "step": 12587 }, { "epoch": 1.5, "learning_rate": 2.62385800929431e-06, "loss": 0.6128, "step": 12588 }, { "epoch": 1.5, "learning_rate": 2.6235369795504477e-06, "loss": 0.6585, "step": 12589 }, { "epoch": 1.5, "learning_rate": 2.623215947764509e-06, "loss": 0.8581, "step": 12590 }, { "epoch": 1.5, "learning_rate": 2.6228949139418007e-06, "loss": 0.5777, "step": 12591 }, { "epoch": 1.5, "learning_rate": 2.622573878087628e-06, "loss": 0.5111, "step": 12592 }, { "epoch": 1.5, "learning_rate": 2.6222528402073006e-06, "loss": 0.5428, "step": 12593 }, { "epoch": 1.5, "learning_rate": 2.6219318003061215e-06, "loss": 0.5837, "step": 12594 }, { "epoch": 1.5, "learning_rate": 2.621610758389401e-06, "loss": 0.689, "step": 12595 }, { "epoch": 1.5, "learning_rate": 2.621289714462444e-06, "loss": 0.8214, "step": 12596 }, { "epoch": 1.5, "learning_rate": 2.6209686685305575e-06, "loss": 0.868, "step": 12597 }, { "epoch": 1.5, "learning_rate": 2.6206476205990487e-06, "loss": 0.3742, "step": 12598 }, { "epoch": 1.5, "learning_rate": 2.620326570673225e-06, "loss": 0.539, "step": 12599 }, { "epoch": 1.5, "learning_rate": 2.620005518758393e-06, "loss": 0.5038, "step": 12600 }, { "epoch": 1.5, "learning_rate": 2.6196844648598596e-06, "loss": 0.6214, "step": 12601 }, { "epoch": 1.5, "learning_rate": 2.6193634089829323e-06, "loss": 0.7582, "step": 12602 }, { "epoch": 1.5, "learning_rate": 2.619042351132917e-06, "loss": 0.928, "step": 12603 }, { "epoch": 1.5, "learning_rate": 2.618721291315123e-06, "loss": 0.493, "step": 12604 }, { "epoch": 1.5, "learning_rate": 2.6184002295348544e-06, "loss": 0.5868, "step": 12605 }, { "epoch": 1.5, "learning_rate": 2.6180791657974206e-06, "loss": 0.502, "step": 12606 }, { "epoch": 1.5, "learning_rate": 2.617758100108129e-06, "loss": 0.8212, "step": 12607 }, { "epoch": 1.5, "learning_rate": 2.617437032472286e-06, "loss": 0.6787, "step": 12608 }, { "epoch": 1.5, "learning_rate": 2.617115962895198e-06, "loss": 0.8171, "step": 12609 }, { "epoch": 1.5, "learning_rate": 2.6167948913821744e-06, "loss": 0.4607, "step": 12610 }, { "epoch": 1.5, "learning_rate": 2.6164738179385215e-06, "loss": 0.4455, "step": 12611 }, { "epoch": 1.5, "learning_rate": 2.616152742569546e-06, "loss": 0.4864, "step": 12612 }, { "epoch": 1.5, "learning_rate": 2.6158316652805555e-06, "loss": 0.5689, "step": 12613 }, { "epoch": 1.5, "learning_rate": 2.6155105860768588e-06, "loss": 0.7151, "step": 12614 }, { "epoch": 1.5, "learning_rate": 2.615189504963762e-06, "loss": 0.8416, "step": 12615 }, { "epoch": 1.5, "learning_rate": 2.614868421946573e-06, "loss": 0.6089, "step": 12616 }, { "epoch": 1.5, "learning_rate": 2.6145473370305997e-06, "loss": 0.4499, "step": 12617 }, { "epoch": 1.5, "learning_rate": 2.614226250221149e-06, "loss": 0.592, "step": 12618 }, { "epoch": 1.5, "learning_rate": 2.613905161523529e-06, "loss": 0.5782, "step": 12619 }, { "epoch": 1.5, "learning_rate": 2.6135840709430465e-06, "loss": 0.6933, "step": 12620 }, { "epoch": 1.5, "learning_rate": 2.6132629784850104e-06, "loss": 0.8043, "step": 12621 }, { "epoch": 1.5, "learning_rate": 2.612941884154728e-06, "loss": 0.7922, "step": 12622 }, { "epoch": 1.5, "learning_rate": 2.6126207879575063e-06, "loss": 0.3947, "step": 12623 }, { "epoch": 1.5, "learning_rate": 2.612299689898653e-06, "loss": 0.6037, "step": 12624 }, { "epoch": 1.5, "learning_rate": 2.611978589983477e-06, "loss": 0.5619, "step": 12625 }, { "epoch": 1.5, "learning_rate": 2.611657488217286e-06, "loss": 0.7307, "step": 12626 }, { "epoch": 1.5, "learning_rate": 2.611336384605387e-06, "loss": 0.8945, "step": 12627 }, { "epoch": 1.5, "learning_rate": 2.6110152791530874e-06, "loss": 0.6786, "step": 12628 }, { "epoch": 1.5, "learning_rate": 2.6106941718656977e-06, "loss": 0.2873, "step": 12629 }, { "epoch": 1.5, "learning_rate": 2.6103730627485225e-06, "loss": 0.4308, "step": 12630 }, { "epoch": 1.5, "learning_rate": 2.610051951806872e-06, "loss": 0.5433, "step": 12631 }, { "epoch": 1.5, "learning_rate": 2.609730839046053e-06, "loss": 0.6036, "step": 12632 }, { "epoch": 1.5, "learning_rate": 2.6094097244713743e-06, "loss": 0.778, "step": 12633 }, { "epoch": 1.5, "learning_rate": 2.609088608088144e-06, "loss": 0.9725, "step": 12634 }, { "epoch": 1.5, "learning_rate": 2.608767489901669e-06, "loss": 0.3701, "step": 12635 }, { "epoch": 1.5, "learning_rate": 2.6084463699172594e-06, "loss": 0.5245, "step": 12636 }, { "epoch": 1.5, "learning_rate": 2.6081252481402214e-06, "loss": 0.5059, "step": 12637 }, { "epoch": 1.51, "learning_rate": 2.607804124575864e-06, "loss": 0.6678, "step": 12638 }, { "epoch": 1.51, "learning_rate": 2.6074829992294954e-06, "loss": 0.7399, "step": 12639 }, { "epoch": 1.51, "learning_rate": 2.6071618721064246e-06, "loss": 0.8419, "step": 12640 }, { "epoch": 1.51, "learning_rate": 2.6068407432119576e-06, "loss": 0.643, "step": 12641 }, { "epoch": 1.51, "learning_rate": 2.6065196125514053e-06, "loss": 0.4784, "step": 12642 }, { "epoch": 1.51, "learning_rate": 2.6061984801300743e-06, "loss": 0.491, "step": 12643 }, { "epoch": 1.51, "learning_rate": 2.6058773459532744e-06, "loss": 0.5771, "step": 12644 }, { "epoch": 1.51, "learning_rate": 2.6055562100263117e-06, "loss": 0.7221, "step": 12645 }, { "epoch": 1.51, "learning_rate": 2.6052350723544966e-06, "loss": 0.8995, "step": 12646 }, { "epoch": 1.51, "learning_rate": 2.6049139329431373e-06, "loss": 0.9199, "step": 12647 }, { "epoch": 1.51, "learning_rate": 2.6045927917975415e-06, "loss": 0.5109, "step": 12648 }, { "epoch": 1.51, "learning_rate": 2.6042716489230184e-06, "loss": 0.5746, "step": 12649 }, { "epoch": 1.51, "learning_rate": 2.6039505043248765e-06, "loss": 0.6072, "step": 12650 }, { "epoch": 1.51, "learning_rate": 2.6036293580084237e-06, "loss": 0.7584, "step": 12651 }, { "epoch": 1.51, "learning_rate": 2.6033082099789682e-06, "loss": 0.8436, "step": 12652 }, { "epoch": 1.51, "learning_rate": 2.60298706024182e-06, "loss": 0.8972, "step": 12653 }, { "epoch": 1.51, "learning_rate": 2.6026659088022872e-06, "loss": 0.4245, "step": 12654 }, { "epoch": 1.51, "learning_rate": 2.602344755665678e-06, "loss": 0.5889, "step": 12655 }, { "epoch": 1.51, "learning_rate": 2.602023600837301e-06, "loss": 0.5714, "step": 12656 }, { "epoch": 1.51, "learning_rate": 2.601702444322466e-06, "loss": 0.6043, "step": 12657 }, { "epoch": 1.51, "learning_rate": 2.601381286126481e-06, "loss": 0.7107, "step": 12658 }, { "epoch": 1.51, "learning_rate": 2.601060126254655e-06, "loss": 0.8864, "step": 12659 }, { "epoch": 1.51, "learning_rate": 2.6007389647122957e-06, "loss": 0.4539, "step": 12660 }, { "epoch": 1.51, "learning_rate": 2.6004178015047137e-06, "loss": 0.5861, "step": 12661 }, { "epoch": 1.51, "learning_rate": 2.6000966366372165e-06, "loss": 0.5693, "step": 12662 }, { "epoch": 1.51, "learning_rate": 2.599775470115114e-06, "loss": 0.6122, "step": 12663 }, { "epoch": 1.51, "learning_rate": 2.599454301943714e-06, "loss": 0.728, "step": 12664 }, { "epoch": 1.51, "learning_rate": 2.599133132128327e-06, "loss": 0.7784, "step": 12665 }, { "epoch": 1.51, "learning_rate": 2.59881196067426e-06, "loss": 0.587, "step": 12666 }, { "epoch": 1.51, "learning_rate": 2.598490787586823e-06, "loss": 0.4034, "step": 12667 }, { "epoch": 1.51, "learning_rate": 2.5981696128713257e-06, "loss": 0.4882, "step": 12668 }, { "epoch": 1.51, "learning_rate": 2.5978484365330762e-06, "loss": 0.5949, "step": 12669 }, { "epoch": 1.51, "learning_rate": 2.5975272585773843e-06, "loss": 0.6371, "step": 12670 }, { "epoch": 1.51, "learning_rate": 2.5972060790095575e-06, "loss": 0.8256, "step": 12671 }, { "epoch": 1.51, "learning_rate": 2.5968848978349075e-06, "loss": 0.7953, "step": 12672 }, { "epoch": 1.51, "learning_rate": 2.596563715058741e-06, "loss": 0.4678, "step": 12673 }, { "epoch": 1.51, "learning_rate": 2.5962425306863687e-06, "loss": 0.6157, "step": 12674 }, { "epoch": 1.51, "learning_rate": 2.5959213447230992e-06, "loss": 0.5676, "step": 12675 }, { "epoch": 1.51, "learning_rate": 2.5956001571742413e-06, "loss": 0.7446, "step": 12676 }, { "epoch": 1.51, "learning_rate": 2.5952789680451044e-06, "loss": 0.7158, "step": 12677 }, { "epoch": 1.51, "learning_rate": 2.5949577773409996e-06, "loss": 0.8549, "step": 12678 }, { "epoch": 1.51, "learning_rate": 2.594636585067234e-06, "loss": 0.3999, "step": 12679 }, { "epoch": 1.51, "learning_rate": 2.594315391229118e-06, "loss": 0.5779, "step": 12680 }, { "epoch": 1.51, "learning_rate": 2.5939941958319603e-06, "loss": 0.5708, "step": 12681 }, { "epoch": 1.51, "learning_rate": 2.5936729988810705e-06, "loss": 0.6504, "step": 12682 }, { "epoch": 1.51, "learning_rate": 2.593351800381759e-06, "loss": 0.7253, "step": 12683 }, { "epoch": 1.51, "learning_rate": 2.5930306003393337e-06, "loss": 0.7043, "step": 12684 }, { "epoch": 1.51, "learning_rate": 2.592709398759105e-06, "loss": 0.5067, "step": 12685 }, { "epoch": 1.51, "learning_rate": 2.592388195646382e-06, "loss": 0.5216, "step": 12686 }, { "epoch": 1.51, "learning_rate": 2.5920669910064742e-06, "loss": 0.6146, "step": 12687 }, { "epoch": 1.51, "learning_rate": 2.5917457848446915e-06, "loss": 0.6476, "step": 12688 }, { "epoch": 1.51, "learning_rate": 2.5914245771663437e-06, "loss": 0.684, "step": 12689 }, { "epoch": 1.51, "learning_rate": 2.5911033679767396e-06, "loss": 0.7779, "step": 12690 }, { "epoch": 1.51, "learning_rate": 2.5907821572811893e-06, "loss": 0.3835, "step": 12691 }, { "epoch": 1.51, "learning_rate": 2.5904609450850016e-06, "loss": 0.3712, "step": 12692 }, { "epoch": 1.51, "learning_rate": 2.5901397313934885e-06, "loss": 0.5651, "step": 12693 }, { "epoch": 1.51, "learning_rate": 2.5898185162119564e-06, "loss": 0.6628, "step": 12694 }, { "epoch": 1.51, "learning_rate": 2.589497299545718e-06, "loss": 0.6122, "step": 12695 }, { "epoch": 1.51, "learning_rate": 2.589176081400081e-06, "loss": 0.7324, "step": 12696 }, { "epoch": 1.51, "learning_rate": 2.588854861780356e-06, "loss": 0.8192, "step": 12697 }, { "epoch": 1.51, "learning_rate": 2.5885336406918526e-06, "loss": 0.5865, "step": 12698 }, { "epoch": 1.51, "learning_rate": 2.5882124181398803e-06, "loss": 0.5565, "step": 12699 }, { "epoch": 1.51, "learning_rate": 2.58789119412975e-06, "loss": 0.5627, "step": 12700 }, { "epoch": 1.51, "learning_rate": 2.587569968666771e-06, "loss": 0.6431, "step": 12701 }, { "epoch": 1.51, "learning_rate": 2.587248741756253e-06, "loss": 0.8017, "step": 12702 }, { "epoch": 1.51, "learning_rate": 2.5869275134035057e-06, "loss": 0.9535, "step": 12703 }, { "epoch": 1.51, "learning_rate": 2.5866062836138397e-06, "loss": 0.3754, "step": 12704 }, { "epoch": 1.51, "learning_rate": 2.5862850523925635e-06, "loss": 0.4511, "step": 12705 }, { "epoch": 1.51, "learning_rate": 2.5859638197449892e-06, "loss": 0.5032, "step": 12706 }, { "epoch": 1.51, "learning_rate": 2.5856425856764258e-06, "loss": 0.7121, "step": 12707 }, { "epoch": 1.51, "learning_rate": 2.5853213501921827e-06, "loss": 0.7642, "step": 12708 }, { "epoch": 1.51, "learning_rate": 2.58500011329757e-06, "loss": 0.7617, "step": 12709 }, { "epoch": 1.51, "learning_rate": 2.584678874997899e-06, "loss": 0.5238, "step": 12710 }, { "epoch": 1.51, "learning_rate": 2.5843576352984795e-06, "loss": 0.5597, "step": 12711 }, { "epoch": 1.51, "learning_rate": 2.584036394204621e-06, "loss": 0.5349, "step": 12712 }, { "epoch": 1.51, "learning_rate": 2.5837151517216335e-06, "loss": 0.6673, "step": 12713 }, { "epoch": 1.51, "learning_rate": 2.5833939078548275e-06, "loss": 0.7504, "step": 12714 }, { "epoch": 1.51, "learning_rate": 2.583072662609514e-06, "loss": 0.8714, "step": 12715 }, { "epoch": 1.51, "learning_rate": 2.582751415991001e-06, "loss": 0.4312, "step": 12716 }, { "epoch": 1.51, "learning_rate": 2.5824301680046017e-06, "loss": 0.5181, "step": 12717 }, { "epoch": 1.51, "learning_rate": 2.5821089186556236e-06, "loss": 0.5138, "step": 12718 }, { "epoch": 1.51, "learning_rate": 2.5817876679493793e-06, "loss": 0.6163, "step": 12719 }, { "epoch": 1.51, "learning_rate": 2.581466415891177e-06, "loss": 0.6075, "step": 12720 }, { "epoch": 1.51, "learning_rate": 2.581145162486328e-06, "loss": 0.9115, "step": 12721 }, { "epoch": 1.52, "learning_rate": 2.5808239077401432e-06, "loss": 0.8061, "step": 12722 }, { "epoch": 1.52, "learning_rate": 2.5805026516579325e-06, "loss": 0.3867, "step": 12723 }, { "epoch": 1.52, "learning_rate": 2.5801813942450055e-06, "loss": 0.4872, "step": 12724 }, { "epoch": 1.52, "learning_rate": 2.579860135506674e-06, "loss": 0.6169, "step": 12725 }, { "epoch": 1.52, "learning_rate": 2.579538875448247e-06, "loss": 0.6624, "step": 12726 }, { "epoch": 1.52, "learning_rate": 2.5792176140750365e-06, "loss": 0.8176, "step": 12727 }, { "epoch": 1.52, "learning_rate": 2.578896351392352e-06, "loss": 0.806, "step": 12728 }, { "epoch": 1.52, "learning_rate": 2.5785750874055044e-06, "loss": 0.3048, "step": 12729 }, { "epoch": 1.52, "learning_rate": 2.578253822119804e-06, "loss": 0.5407, "step": 12730 }, { "epoch": 1.52, "learning_rate": 2.5779325555405603e-06, "loss": 0.4907, "step": 12731 }, { "epoch": 1.52, "learning_rate": 2.577611287673086e-06, "loss": 0.596, "step": 12732 }, { "epoch": 1.52, "learning_rate": 2.57729001852269e-06, "loss": 0.7563, "step": 12733 }, { "epoch": 1.52, "learning_rate": 2.576968748094684e-06, "loss": 0.742, "step": 12734 }, { "epoch": 1.52, "learning_rate": 2.5766474763943776e-06, "loss": 0.4002, "step": 12735 }, { "epoch": 1.52, "learning_rate": 2.576326203427083e-06, "loss": 0.5401, "step": 12736 }, { "epoch": 1.52, "learning_rate": 2.5760049291981087e-06, "loss": 0.5995, "step": 12737 }, { "epoch": 1.52, "learning_rate": 2.5756836537127673e-06, "loss": 0.5817, "step": 12738 }, { "epoch": 1.52, "learning_rate": 2.575362376976368e-06, "loss": 0.6271, "step": 12739 }, { "epoch": 1.52, "learning_rate": 2.5750410989942233e-06, "loss": 0.9255, "step": 12740 }, { "epoch": 1.52, "learning_rate": 2.574719819771642e-06, "loss": 0.7367, "step": 12741 }, { "epoch": 1.52, "learning_rate": 2.5743985393139364e-06, "loss": 0.4821, "step": 12742 }, { "epoch": 1.52, "learning_rate": 2.574077257626417e-06, "loss": 0.5287, "step": 12743 }, { "epoch": 1.52, "learning_rate": 2.5737559747143935e-06, "loss": 0.6062, "step": 12744 }, { "epoch": 1.52, "learning_rate": 2.573434690583178e-06, "loss": 0.6093, "step": 12745 }, { "epoch": 1.52, "learning_rate": 2.57311340523808e-06, "loss": 0.9005, "step": 12746 }, { "epoch": 1.52, "learning_rate": 2.572792118684413e-06, "loss": 0.7782, "step": 12747 }, { "epoch": 1.52, "learning_rate": 2.5724708309274848e-06, "loss": 0.4573, "step": 12748 }, { "epoch": 1.52, "learning_rate": 2.5721495419726085e-06, "loss": 0.4841, "step": 12749 }, { "epoch": 1.52, "learning_rate": 2.5718282518250937e-06, "loss": 0.5504, "step": 12750 }, { "epoch": 1.52, "learning_rate": 2.571506960490252e-06, "loss": 0.549, "step": 12751 }, { "epoch": 1.52, "learning_rate": 2.5711856679733936e-06, "loss": 0.7539, "step": 12752 }, { "epoch": 1.52, "learning_rate": 2.5708643742798313e-06, "loss": 0.8745, "step": 12753 }, { "epoch": 1.52, "learning_rate": 2.5705430794148745e-06, "loss": 0.3905, "step": 12754 }, { "epoch": 1.52, "learning_rate": 2.5702217833838346e-06, "loss": 0.545, "step": 12755 }, { "epoch": 1.52, "learning_rate": 2.5699004861920224e-06, "loss": 0.5536, "step": 12756 }, { "epoch": 1.52, "learning_rate": 2.5695791878447497e-06, "loss": 0.7124, "step": 12757 }, { "epoch": 1.52, "learning_rate": 2.5692578883473274e-06, "loss": 0.7389, "step": 12758 }, { "epoch": 1.52, "learning_rate": 2.568936587705066e-06, "loss": 0.7865, "step": 12759 }, { "epoch": 1.52, "learning_rate": 2.568615285923277e-06, "loss": 0.5681, "step": 12760 }, { "epoch": 1.52, "learning_rate": 2.568293983007272e-06, "loss": 0.486, "step": 12761 }, { "epoch": 1.52, "learning_rate": 2.5679726789623616e-06, "loss": 0.5983, "step": 12762 }, { "epoch": 1.52, "learning_rate": 2.5676513737938563e-06, "loss": 0.5932, "step": 12763 }, { "epoch": 1.52, "learning_rate": 2.567330067507069e-06, "loss": 0.7062, "step": 12764 }, { "epoch": 1.52, "learning_rate": 2.5670087601073103e-06, "loss": 0.8521, "step": 12765 }, { "epoch": 1.52, "learning_rate": 2.56668745159989e-06, "loss": 0.5845, "step": 12766 }, { "epoch": 1.52, "learning_rate": 2.566366141990121e-06, "loss": 0.46, "step": 12767 }, { "epoch": 1.52, "learning_rate": 2.566044831283315e-06, "loss": 0.5004, "step": 12768 }, { "epoch": 1.52, "learning_rate": 2.5657235194847812e-06, "loss": 0.5877, "step": 12769 }, { "epoch": 1.52, "learning_rate": 2.565402206599833e-06, "loss": 0.5818, "step": 12770 }, { "epoch": 1.52, "learning_rate": 2.56508089263378e-06, "loss": 0.763, "step": 12771 }, { "epoch": 1.52, "learning_rate": 2.564759577591935e-06, "loss": 0.8478, "step": 12772 }, { "epoch": 1.52, "learning_rate": 2.564438261479609e-06, "loss": 0.4152, "step": 12773 }, { "epoch": 1.52, "learning_rate": 2.5641169443021123e-06, "loss": 0.4661, "step": 12774 }, { "epoch": 1.52, "learning_rate": 2.5637956260647577e-06, "loss": 0.6376, "step": 12775 }, { "epoch": 1.52, "learning_rate": 2.5634743067728564e-06, "loss": 0.6318, "step": 12776 }, { "epoch": 1.52, "learning_rate": 2.5631529864317185e-06, "loss": 0.7443, "step": 12777 }, { "epoch": 1.52, "learning_rate": 2.5628316650466565e-06, "loss": 0.8549, "step": 12778 }, { "epoch": 1.52, "learning_rate": 2.562510342622983e-06, "loss": 0.4199, "step": 12779 }, { "epoch": 1.52, "learning_rate": 2.562189019166007e-06, "loss": 0.525, "step": 12780 }, { "epoch": 1.52, "learning_rate": 2.5618676946810415e-06, "loss": 0.5401, "step": 12781 }, { "epoch": 1.52, "learning_rate": 2.561546369173398e-06, "loss": 0.6318, "step": 12782 }, { "epoch": 1.52, "learning_rate": 2.5612250426483886e-06, "loss": 0.735, "step": 12783 }, { "epoch": 1.52, "learning_rate": 2.560903715111323e-06, "loss": 0.7103, "step": 12784 }, { "epoch": 1.52, "learning_rate": 2.5605823865675146e-06, "loss": 0.4383, "step": 12785 }, { "epoch": 1.52, "learning_rate": 2.560261057022274e-06, "loss": 0.5597, "step": 12786 }, { "epoch": 1.52, "learning_rate": 2.5599397264809133e-06, "loss": 0.5406, "step": 12787 }, { "epoch": 1.52, "learning_rate": 2.559618394948743e-06, "loss": 0.6234, "step": 12788 }, { "epoch": 1.52, "learning_rate": 2.5592970624310766e-06, "loss": 0.6988, "step": 12789 }, { "epoch": 1.52, "learning_rate": 2.5589757289332245e-06, "loss": 0.9103, "step": 12790 }, { "epoch": 1.52, "learning_rate": 2.5586543944604993e-06, "loss": 0.4029, "step": 12791 }, { "epoch": 1.52, "learning_rate": 2.558333059018211e-06, "loss": 0.507, "step": 12792 }, { "epoch": 1.52, "learning_rate": 2.558011722611673e-06, "loss": 0.4853, "step": 12793 }, { "epoch": 1.52, "learning_rate": 2.557690385246196e-06, "loss": 0.5571, "step": 12794 }, { "epoch": 1.52, "learning_rate": 2.5573690469270916e-06, "loss": 0.5948, "step": 12795 }, { "epoch": 1.52, "learning_rate": 2.557047707659673e-06, "loss": 0.8083, "step": 12796 }, { "epoch": 1.52, "learning_rate": 2.556726367449251e-06, "loss": 0.8196, "step": 12797 }, { "epoch": 1.52, "learning_rate": 2.556405026301137e-06, "loss": 0.4785, "step": 12798 }, { "epoch": 1.52, "learning_rate": 2.556083684220643e-06, "loss": 0.5412, "step": 12799 }, { "epoch": 1.52, "learning_rate": 2.555762341213081e-06, "loss": 0.5397, "step": 12800 }, { "epoch": 1.52, "learning_rate": 2.555440997283763e-06, "loss": 0.6931, "step": 12801 }, { "epoch": 1.52, "learning_rate": 2.5551196524380008e-06, "loss": 0.8451, "step": 12802 }, { "epoch": 1.52, "learning_rate": 2.5547983066811056e-06, "loss": 0.8613, "step": 12803 }, { "epoch": 1.52, "learning_rate": 2.5544769600183905e-06, "loss": 0.4594, "step": 12804 }, { "epoch": 1.52, "learning_rate": 2.554155612455166e-06, "loss": 0.5273, "step": 12805 }, { "epoch": 1.53, "learning_rate": 2.553834263996745e-06, "loss": 0.5637, "step": 12806 }, { "epoch": 1.53, "learning_rate": 2.5535129146484395e-06, "loss": 0.6609, "step": 12807 }, { "epoch": 1.53, "learning_rate": 2.553191564415561e-06, "loss": 0.7526, "step": 12808 }, { "epoch": 1.53, "learning_rate": 2.552870213303421e-06, "loss": 0.8838, "step": 12809 }, { "epoch": 1.53, "learning_rate": 2.5525488613173314e-06, "loss": 0.3496, "step": 12810 }, { "epoch": 1.53, "learning_rate": 2.552227508462606e-06, "loss": 0.4801, "step": 12811 }, { "epoch": 1.53, "learning_rate": 2.5519061547445544e-06, "loss": 0.6094, "step": 12812 }, { "epoch": 1.53, "learning_rate": 2.5515848001684907e-06, "loss": 0.5924, "step": 12813 }, { "epoch": 1.53, "learning_rate": 2.551263444739725e-06, "loss": 0.6578, "step": 12814 }, { "epoch": 1.53, "learning_rate": 2.5509420884635716e-06, "loss": 0.9178, "step": 12815 }, { "epoch": 1.53, "learning_rate": 2.5506207313453397e-06, "loss": 0.4759, "step": 12816 }, { "epoch": 1.53, "learning_rate": 2.550299373390344e-06, "loss": 0.6738, "step": 12817 }, { "epoch": 1.53, "learning_rate": 2.5499780146038945e-06, "loss": 0.5493, "step": 12818 }, { "epoch": 1.53, "learning_rate": 2.5496566549913047e-06, "loss": 0.644, "step": 12819 }, { "epoch": 1.53, "learning_rate": 2.5493352945578858e-06, "loss": 0.7574, "step": 12820 }, { "epoch": 1.53, "learning_rate": 2.5490139333089507e-06, "loss": 0.7578, "step": 12821 }, { "epoch": 1.53, "learning_rate": 2.5486925712498113e-06, "loss": 0.8873, "step": 12822 }, { "epoch": 1.53, "learning_rate": 2.5483712083857792e-06, "loss": 0.4669, "step": 12823 }, { "epoch": 1.53, "learning_rate": 2.5480498447221676e-06, "loss": 0.6104, "step": 12824 }, { "epoch": 1.53, "learning_rate": 2.547728480264287e-06, "loss": 0.6545, "step": 12825 }, { "epoch": 1.53, "learning_rate": 2.5474071150174523e-06, "loss": 0.6973, "step": 12826 }, { "epoch": 1.53, "learning_rate": 2.5470857489869726e-06, "loss": 0.8057, "step": 12827 }, { "epoch": 1.53, "learning_rate": 2.5467643821781616e-06, "loss": 0.8401, "step": 12828 }, { "epoch": 1.53, "learning_rate": 2.5464430145963322e-06, "loss": 0.4937, "step": 12829 }, { "epoch": 1.53, "learning_rate": 2.5461216462467954e-06, "loss": 0.4349, "step": 12830 }, { "epoch": 1.53, "learning_rate": 2.545800277134863e-06, "loss": 0.5479, "step": 12831 }, { "epoch": 1.53, "learning_rate": 2.5454789072658496e-06, "loss": 0.5784, "step": 12832 }, { "epoch": 1.53, "learning_rate": 2.545157536645066e-06, "loss": 0.744, "step": 12833 }, { "epoch": 1.53, "learning_rate": 2.5448361652778242e-06, "loss": 0.8346, "step": 12834 }, { "epoch": 1.53, "learning_rate": 2.5445147931694364e-06, "loss": 0.4997, "step": 12835 }, { "epoch": 1.53, "learning_rate": 2.544193420325216e-06, "loss": 0.5076, "step": 12836 }, { "epoch": 1.53, "learning_rate": 2.543872046750474e-06, "loss": 0.5653, "step": 12837 }, { "epoch": 1.53, "learning_rate": 2.5435506724505236e-06, "loss": 0.6182, "step": 12838 }, { "epoch": 1.53, "learning_rate": 2.5432292974306772e-06, "loss": 0.7524, "step": 12839 }, { "epoch": 1.53, "learning_rate": 2.542907921696247e-06, "loss": 0.9103, "step": 12840 }, { "epoch": 1.53, "learning_rate": 2.542586545252545e-06, "loss": 0.5522, "step": 12841 }, { "epoch": 1.53, "learning_rate": 2.542265168104883e-06, "loss": 0.4116, "step": 12842 }, { "epoch": 1.53, "learning_rate": 2.5419437902585754e-06, "loss": 0.6095, "step": 12843 }, { "epoch": 1.53, "learning_rate": 2.5416224117189332e-06, "loss": 0.6878, "step": 12844 }, { "epoch": 1.53, "learning_rate": 2.541301032491269e-06, "loss": 0.7496, "step": 12845 }, { "epoch": 1.53, "learning_rate": 2.540979652580895e-06, "loss": 0.7094, "step": 12846 }, { "epoch": 1.53, "learning_rate": 2.5406582719931245e-06, "loss": 0.8308, "step": 12847 }, { "epoch": 1.53, "learning_rate": 2.540336890733268e-06, "loss": 0.3553, "step": 12848 }, { "epoch": 1.53, "learning_rate": 2.5400155088066407e-06, "loss": 0.4904, "step": 12849 }, { "epoch": 1.53, "learning_rate": 2.5396941262185532e-06, "loss": 0.5876, "step": 12850 }, { "epoch": 1.53, "learning_rate": 2.5393727429743185e-06, "loss": 0.6618, "step": 12851 }, { "epoch": 1.53, "learning_rate": 2.5390513590792486e-06, "loss": 0.8167, "step": 12852 }, { "epoch": 1.53, "learning_rate": 2.5387299745386575e-06, "loss": 0.8018, "step": 12853 }, { "epoch": 1.53, "learning_rate": 2.538408589357856e-06, "loss": 0.4217, "step": 12854 }, { "epoch": 1.53, "learning_rate": 2.538087203542157e-06, "loss": 0.4774, "step": 12855 }, { "epoch": 1.53, "learning_rate": 2.5377658170968744e-06, "loss": 0.5689, "step": 12856 }, { "epoch": 1.53, "learning_rate": 2.537444430027318e-06, "loss": 0.7063, "step": 12857 }, { "epoch": 1.53, "learning_rate": 2.5371230423388036e-06, "loss": 0.6328, "step": 12858 }, { "epoch": 1.53, "learning_rate": 2.5368016540366412e-06, "loss": 0.7562, "step": 12859 }, { "epoch": 1.53, "learning_rate": 2.536480265126145e-06, "loss": 0.5223, "step": 12860 }, { "epoch": 1.53, "learning_rate": 2.536158875612627e-06, "loss": 0.518, "step": 12861 }, { "epoch": 1.53, "learning_rate": 2.5358374855013993e-06, "loss": 0.5971, "step": 12862 }, { "epoch": 1.53, "learning_rate": 2.535516094797775e-06, "loss": 0.6573, "step": 12863 }, { "epoch": 1.53, "learning_rate": 2.535194703507067e-06, "loss": 0.7252, "step": 12864 }, { "epoch": 1.53, "learning_rate": 2.5348733116345877e-06, "loss": 0.8077, "step": 12865 }, { "epoch": 1.53, "learning_rate": 2.534551919185649e-06, "loss": 0.5119, "step": 12866 }, { "epoch": 1.53, "learning_rate": 2.5342305261655642e-06, "loss": 0.4146, "step": 12867 }, { "epoch": 1.53, "learning_rate": 2.533909132579646e-06, "loss": 0.6621, "step": 12868 }, { "epoch": 1.53, "learning_rate": 2.5335877384332073e-06, "loss": 0.7462, "step": 12869 }, { "epoch": 1.53, "learning_rate": 2.5332663437315607e-06, "loss": 0.711, "step": 12870 }, { "epoch": 1.53, "learning_rate": 2.5329449484800186e-06, "loss": 0.8085, "step": 12871 }, { "epoch": 1.53, "learning_rate": 2.532623552683893e-06, "loss": 1.0035, "step": 12872 }, { "epoch": 1.53, "learning_rate": 2.5323021563484982e-06, "loss": 0.5347, "step": 12873 }, { "epoch": 1.53, "learning_rate": 2.531980759479145e-06, "loss": 0.4635, "step": 12874 }, { "epoch": 1.53, "learning_rate": 2.531659362081148e-06, "loss": 0.6113, "step": 12875 }, { "epoch": 1.53, "learning_rate": 2.5313379641598197e-06, "loss": 0.6914, "step": 12876 }, { "epoch": 1.53, "learning_rate": 2.5310165657204712e-06, "loss": 0.8065, "step": 12877 }, { "epoch": 1.53, "learning_rate": 2.530695166768416e-06, "loss": 0.8554, "step": 12878 }, { "epoch": 1.53, "learning_rate": 2.5303737673089686e-06, "loss": 0.5712, "step": 12879 }, { "epoch": 1.53, "learning_rate": 2.5300523673474388e-06, "loss": 0.5307, "step": 12880 }, { "epoch": 1.53, "learning_rate": 2.529730966889141e-06, "loss": 0.5862, "step": 12881 }, { "epoch": 1.53, "learning_rate": 2.529409565939388e-06, "loss": 0.5155, "step": 12882 }, { "epoch": 1.53, "learning_rate": 2.5290881645034932e-06, "loss": 0.6949, "step": 12883 }, { "epoch": 1.53, "learning_rate": 2.5287667625867674e-06, "loss": 0.8469, "step": 12884 }, { "epoch": 1.53, "learning_rate": 2.5284453601945254e-06, "loss": 0.7411, "step": 12885 }, { "epoch": 1.53, "learning_rate": 2.5281239573320792e-06, "loss": 0.506, "step": 12886 }, { "epoch": 1.53, "learning_rate": 2.527802554004741e-06, "loss": 0.5448, "step": 12887 }, { "epoch": 1.53, "learning_rate": 2.5274811502178246e-06, "loss": 0.5785, "step": 12888 }, { "epoch": 1.53, "learning_rate": 2.5271597459766427e-06, "loss": 0.6797, "step": 12889 }, { "epoch": 1.54, "learning_rate": 2.526838341286508e-06, "loss": 0.7695, "step": 12890 }, { "epoch": 1.54, "learning_rate": 2.526516936152733e-06, "loss": 0.6736, "step": 12891 }, { "epoch": 1.54, "learning_rate": 2.5261955305806306e-06, "loss": 0.4547, "step": 12892 }, { "epoch": 1.54, "learning_rate": 2.525874124575514e-06, "loss": 0.5428, "step": 12893 }, { "epoch": 1.54, "learning_rate": 2.5255527181426965e-06, "loss": 0.545, "step": 12894 }, { "epoch": 1.54, "learning_rate": 2.52523131128749e-06, "loss": 0.6849, "step": 12895 }, { "epoch": 1.54, "learning_rate": 2.5249099040152077e-06, "loss": 0.7477, "step": 12896 }, { "epoch": 1.54, "learning_rate": 2.524588496331163e-06, "loss": 0.8855, "step": 12897 }, { "epoch": 1.54, "learning_rate": 2.5242670882406676e-06, "loss": 0.6286, "step": 12898 }, { "epoch": 1.54, "learning_rate": 2.5239456797490354e-06, "loss": 0.6249, "step": 12899 }, { "epoch": 1.54, "learning_rate": 2.5236242708615795e-06, "loss": 0.5861, "step": 12900 }, { "epoch": 1.54, "learning_rate": 2.5233028615836126e-06, "loss": 0.6961, "step": 12901 }, { "epoch": 1.54, "learning_rate": 2.522981451920447e-06, "loss": 0.8859, "step": 12902 }, { "epoch": 1.54, "learning_rate": 2.5226600418773968e-06, "loss": 0.7143, "step": 12903 }, { "epoch": 1.54, "learning_rate": 2.5223386314597737e-06, "loss": 0.4983, "step": 12904 }, { "epoch": 1.54, "learning_rate": 2.522017220672891e-06, "loss": 0.5188, "step": 12905 }, { "epoch": 1.54, "learning_rate": 2.521695809522061e-06, "loss": 0.5713, "step": 12906 }, { "epoch": 1.54, "learning_rate": 2.5213743980125986e-06, "loss": 0.6423, "step": 12907 }, { "epoch": 1.54, "learning_rate": 2.521052986149816e-06, "loss": 0.7118, "step": 12908 }, { "epoch": 1.54, "learning_rate": 2.520731573939025e-06, "loss": 0.8481, "step": 12909 }, { "epoch": 1.54, "learning_rate": 2.5204101613855384e-06, "loss": 0.4619, "step": 12910 }, { "epoch": 1.54, "learning_rate": 2.520088748494672e-06, "loss": 0.5539, "step": 12911 }, { "epoch": 1.54, "learning_rate": 2.519767335271736e-06, "loss": 0.5439, "step": 12912 }, { "epoch": 1.54, "learning_rate": 2.5194459217220448e-06, "loss": 0.6979, "step": 12913 }, { "epoch": 1.54, "learning_rate": 2.5191245078509096e-06, "loss": 0.7273, "step": 12914 }, { "epoch": 1.54, "learning_rate": 2.518803093663646e-06, "loss": 0.8809, "step": 12915 }, { "epoch": 1.54, "learning_rate": 2.5184816791655647e-06, "loss": 0.7094, "step": 12916 }, { "epoch": 1.54, "learning_rate": 2.5181602643619805e-06, "loss": 0.5312, "step": 12917 }, { "epoch": 1.54, "learning_rate": 2.5178388492582058e-06, "loss": 0.5612, "step": 12918 }, { "epoch": 1.54, "learning_rate": 2.5175174338595526e-06, "loss": 0.6118, "step": 12919 }, { "epoch": 1.54, "learning_rate": 2.5171960181713353e-06, "loss": 0.6266, "step": 12920 }, { "epoch": 1.54, "learning_rate": 2.5168746021988655e-06, "loss": 0.7829, "step": 12921 }, { "epoch": 1.54, "learning_rate": 2.5165531859474585e-06, "loss": 0.7952, "step": 12922 }, { "epoch": 1.54, "learning_rate": 2.5162317694224247e-06, "loss": 0.4845, "step": 12923 }, { "epoch": 1.54, "learning_rate": 2.515910352629079e-06, "loss": 0.4496, "step": 12924 }, { "epoch": 1.54, "learning_rate": 2.5155889355727335e-06, "loss": 0.5907, "step": 12925 }, { "epoch": 1.54, "learning_rate": 2.515267518258703e-06, "loss": 0.6449, "step": 12926 }, { "epoch": 1.54, "learning_rate": 2.514946100692297e-06, "loss": 0.7217, "step": 12927 }, { "epoch": 1.54, "learning_rate": 2.514624682878832e-06, "loss": 0.9064, "step": 12928 }, { "epoch": 1.54, "learning_rate": 2.5143032648236195e-06, "loss": 0.4032, "step": 12929 }, { "epoch": 1.54, "learning_rate": 2.5139818465319727e-06, "loss": 0.5493, "step": 12930 }, { "epoch": 1.54, "learning_rate": 2.513660428009204e-06, "loss": 0.5952, "step": 12931 }, { "epoch": 1.54, "learning_rate": 2.5133390092606287e-06, "loss": 0.5835, "step": 12932 }, { "epoch": 1.54, "learning_rate": 2.513017590291558e-06, "loss": 0.6851, "step": 12933 }, { "epoch": 1.54, "learning_rate": 2.5126961711073054e-06, "loss": 0.8701, "step": 12934 }, { "epoch": 1.54, "learning_rate": 2.5123747517131842e-06, "loss": 0.4568, "step": 12935 }, { "epoch": 1.54, "learning_rate": 2.5120533321145067e-06, "loss": 0.5565, "step": 12936 }, { "epoch": 1.54, "learning_rate": 2.511731912316587e-06, "loss": 0.5359, "step": 12937 }, { "epoch": 1.54, "learning_rate": 2.5114104923247378e-06, "loss": 0.6165, "step": 12938 }, { "epoch": 1.54, "learning_rate": 2.5110890721442723e-06, "loss": 0.7083, "step": 12939 }, { "epoch": 1.54, "learning_rate": 2.5107676517805034e-06, "loss": 0.8677, "step": 12940 }, { "epoch": 1.54, "learning_rate": 2.5104462312387445e-06, "loss": 0.575, "step": 12941 }, { "epoch": 1.54, "learning_rate": 2.5101248105243082e-06, "loss": 0.4527, "step": 12942 }, { "epoch": 1.54, "learning_rate": 2.509803389642508e-06, "loss": 0.5203, "step": 12943 }, { "epoch": 1.54, "learning_rate": 2.509481968598657e-06, "loss": 0.5621, "step": 12944 }, { "epoch": 1.54, "learning_rate": 2.5091605473980684e-06, "loss": 0.7312, "step": 12945 }, { "epoch": 1.54, "learning_rate": 2.5088391260460547e-06, "loss": 0.8507, "step": 12946 }, { "epoch": 1.54, "learning_rate": 2.508517704547931e-06, "loss": 0.8817, "step": 12947 }, { "epoch": 1.54, "learning_rate": 2.5081962829090068e-06, "loss": 0.479, "step": 12948 }, { "epoch": 1.54, "learning_rate": 2.507874861134599e-06, "loss": 0.7234, "step": 12949 }, { "epoch": 1.54, "learning_rate": 2.5075534392300183e-06, "loss": 0.5562, "step": 12950 }, { "epoch": 1.54, "learning_rate": 2.5072320172005786e-06, "loss": 0.596, "step": 12951 }, { "epoch": 1.54, "learning_rate": 2.5069105950515937e-06, "loss": 0.7979, "step": 12952 }, { "epoch": 1.54, "learning_rate": 2.506589172788375e-06, "loss": 0.8172, "step": 12953 }, { "epoch": 1.54, "learning_rate": 2.5062677504162377e-06, "loss": 0.4761, "step": 12954 }, { "epoch": 1.54, "learning_rate": 2.5059463279404933e-06, "loss": 0.5706, "step": 12955 }, { "epoch": 1.54, "learning_rate": 2.505624905366456e-06, "loss": 0.5886, "step": 12956 }, { "epoch": 1.54, "learning_rate": 2.5053034826994384e-06, "loss": 0.644, "step": 12957 }, { "epoch": 1.54, "learning_rate": 2.504982059944754e-06, "loss": 0.7112, "step": 12958 }, { "epoch": 1.54, "learning_rate": 2.5046606371077152e-06, "loss": 0.8357, "step": 12959 }, { "epoch": 1.54, "learning_rate": 2.5043392141936364e-06, "loss": 0.4199, "step": 12960 }, { "epoch": 1.54, "learning_rate": 2.5040177912078296e-06, "loss": 0.4832, "step": 12961 }, { "epoch": 1.54, "learning_rate": 2.5036963681556082e-06, "loss": 0.4961, "step": 12962 }, { "epoch": 1.54, "learning_rate": 2.5033749450422857e-06, "loss": 0.5701, "step": 12963 }, { "epoch": 1.54, "learning_rate": 2.503053521873175e-06, "loss": 0.7625, "step": 12964 }, { "epoch": 1.54, "learning_rate": 2.5027320986535896e-06, "loss": 0.8749, "step": 12965 }, { "epoch": 1.54, "learning_rate": 2.502410675388842e-06, "loss": 0.6468, "step": 12966 }, { "epoch": 1.54, "learning_rate": 2.502089252084246e-06, "loss": 0.4389, "step": 12967 }, { "epoch": 1.54, "learning_rate": 2.5017678287451142e-06, "loss": 0.5123, "step": 12968 }, { "epoch": 1.54, "learning_rate": 2.501446405376761e-06, "loss": 0.6327, "step": 12969 }, { "epoch": 1.54, "learning_rate": 2.5011249819844975e-06, "loss": 0.6289, "step": 12970 }, { "epoch": 1.54, "learning_rate": 2.5008035585736385e-06, "loss": 0.8525, "step": 12971 }, { "epoch": 1.54, "learning_rate": 2.5004821351494964e-06, "loss": 0.9141, "step": 12972 }, { "epoch": 1.54, "learning_rate": 2.500160711717385e-06, "loss": 0.455, "step": 12973 }, { "epoch": 1.55, "learning_rate": 2.4998392882826163e-06, "loss": 0.5419, "step": 12974 }, { "epoch": 1.55, "learning_rate": 2.499517864850504e-06, "loss": 0.5876, "step": 12975 }, { "epoch": 1.55, "learning_rate": 2.4991964414263623e-06, "loss": 0.6973, "step": 12976 }, { "epoch": 1.55, "learning_rate": 2.498875018015503e-06, "loss": 0.7883, "step": 12977 }, { "epoch": 1.55, "learning_rate": 2.49855359462324e-06, "loss": 0.8973, "step": 12978 }, { "epoch": 1.55, "learning_rate": 2.4982321712548866e-06, "loss": 0.3667, "step": 12979 }, { "epoch": 1.55, "learning_rate": 2.4979107479157544e-06, "loss": 0.4939, "step": 12980 }, { "epoch": 1.55, "learning_rate": 2.497589324611158e-06, "loss": 0.6267, "step": 12981 }, { "epoch": 1.55, "learning_rate": 2.497267901346411e-06, "loss": 0.623, "step": 12982 }, { "epoch": 1.55, "learning_rate": 2.496946478126826e-06, "loss": 0.7128, "step": 12983 }, { "epoch": 1.55, "learning_rate": 2.4966250549577147e-06, "loss": 0.8655, "step": 12984 }, { "epoch": 1.55, "learning_rate": 2.496303631844393e-06, "loss": 0.4648, "step": 12985 }, { "epoch": 1.55, "learning_rate": 2.495982208792171e-06, "loss": 0.4704, "step": 12986 }, { "epoch": 1.55, "learning_rate": 2.495660785806365e-06, "loss": 0.5468, "step": 12987 }, { "epoch": 1.55, "learning_rate": 2.495339362892285e-06, "loss": 0.6552, "step": 12988 }, { "epoch": 1.55, "learning_rate": 2.495017940055247e-06, "loss": 0.8261, "step": 12989 }, { "epoch": 1.55, "learning_rate": 2.494696517300563e-06, "loss": 0.8545, "step": 12990 }, { "epoch": 1.55, "learning_rate": 2.4943750946335445e-06, "loss": 0.6636, "step": 12991 }, { "epoch": 1.55, "learning_rate": 2.4940536720595067e-06, "loss": 0.4197, "step": 12992 }, { "epoch": 1.55, "learning_rate": 2.4937322495837636e-06, "loss": 0.553, "step": 12993 }, { "epoch": 1.55, "learning_rate": 2.493410827211626e-06, "loss": 0.6274, "step": 12994 }, { "epoch": 1.55, "learning_rate": 2.493089404948407e-06, "loss": 0.6532, "step": 12995 }, { "epoch": 1.55, "learning_rate": 2.4927679827994214e-06, "loss": 0.9119, "step": 12996 }, { "epoch": 1.55, "learning_rate": 2.4924465607699826e-06, "loss": 0.8203, "step": 12997 }, { "epoch": 1.55, "learning_rate": 2.4921251388654023e-06, "loss": 0.3786, "step": 12998 }, { "epoch": 1.55, "learning_rate": 2.491803717090993e-06, "loss": 0.6576, "step": 12999 }, { "epoch": 1.55, "learning_rate": 2.491482295452071e-06, "loss": 0.5508, "step": 13000 }, { "epoch": 1.55, "learning_rate": 2.4911608739539457e-06, "loss": 0.5656, "step": 13001 }, { "epoch": 1.55, "learning_rate": 2.490839452601932e-06, "loss": 0.7731, "step": 13002 }, { "epoch": 1.55, "learning_rate": 2.490518031401343e-06, "loss": 0.8304, "step": 13003 }, { "epoch": 1.55, "learning_rate": 2.4901966103574925e-06, "loss": 0.4257, "step": 13004 }, { "epoch": 1.55, "learning_rate": 2.4898751894756926e-06, "loss": 0.5455, "step": 13005 }, { "epoch": 1.55, "learning_rate": 2.4895537687612563e-06, "loss": 0.5476, "step": 13006 }, { "epoch": 1.55, "learning_rate": 2.489232348219497e-06, "loss": 0.7048, "step": 13007 }, { "epoch": 1.55, "learning_rate": 2.488910927855729e-06, "loss": 0.7505, "step": 13008 }, { "epoch": 1.55, "learning_rate": 2.4885895076752626e-06, "loss": 0.9106, "step": 13009 }, { "epoch": 1.55, "learning_rate": 2.488268087683413e-06, "loss": 0.665, "step": 13010 }, { "epoch": 1.55, "learning_rate": 2.4879466678854937e-06, "loss": 0.5029, "step": 13011 }, { "epoch": 1.55, "learning_rate": 2.487625248286817e-06, "loss": 0.5653, "step": 13012 }, { "epoch": 1.55, "learning_rate": 2.4873038288926954e-06, "loss": 0.5908, "step": 13013 }, { "epoch": 1.55, "learning_rate": 2.4869824097084424e-06, "loss": 0.7595, "step": 13014 }, { "epoch": 1.55, "learning_rate": 2.486660990739372e-06, "loss": 0.7967, "step": 13015 }, { "epoch": 1.55, "learning_rate": 2.4863395719907963e-06, "loss": 0.5702, "step": 13016 }, { "epoch": 1.55, "learning_rate": 2.486018153468028e-06, "loss": 0.4058, "step": 13017 }, { "epoch": 1.55, "learning_rate": 2.4856967351763814e-06, "loss": 0.5663, "step": 13018 }, { "epoch": 1.55, "learning_rate": 2.4853753171211694e-06, "loss": 0.6566, "step": 13019 }, { "epoch": 1.55, "learning_rate": 2.4850538993077037e-06, "loss": 0.6922, "step": 13020 }, { "epoch": 1.55, "learning_rate": 2.4847324817412983e-06, "loss": 0.7405, "step": 13021 }, { "epoch": 1.55, "learning_rate": 2.484411064427267e-06, "loss": 0.7921, "step": 13022 }, { "epoch": 1.55, "learning_rate": 2.484089647370922e-06, "loss": 0.5451, "step": 13023 }, { "epoch": 1.55, "learning_rate": 2.483768230577576e-06, "loss": 0.6412, "step": 13024 }, { "epoch": 1.55, "learning_rate": 2.483446814052542e-06, "loss": 0.5981, "step": 13025 }, { "epoch": 1.55, "learning_rate": 2.4831253978011345e-06, "loss": 0.6456, "step": 13026 }, { "epoch": 1.55, "learning_rate": 2.482803981828666e-06, "loss": 0.7974, "step": 13027 }, { "epoch": 1.55, "learning_rate": 2.4824825661404482e-06, "loss": 0.9341, "step": 13028 }, { "epoch": 1.55, "learning_rate": 2.4821611507417955e-06, "loss": 0.4952, "step": 13029 }, { "epoch": 1.55, "learning_rate": 2.4818397356380207e-06, "loss": 0.4558, "step": 13030 }, { "epoch": 1.55, "learning_rate": 2.4815183208344357e-06, "loss": 0.4914, "step": 13031 }, { "epoch": 1.55, "learning_rate": 2.4811969063363543e-06, "loss": 0.6045, "step": 13032 }, { "epoch": 1.55, "learning_rate": 2.4808754921490904e-06, "loss": 0.6593, "step": 13033 }, { "epoch": 1.55, "learning_rate": 2.480554078277957e-06, "loss": 0.7874, "step": 13034 }, { "epoch": 1.55, "learning_rate": 2.480232664728265e-06, "loss": 0.6743, "step": 13035 }, { "epoch": 1.55, "learning_rate": 2.479911251505329e-06, "loss": 0.4403, "step": 13036 }, { "epoch": 1.55, "learning_rate": 2.479589838614461e-06, "loss": 0.6136, "step": 13037 }, { "epoch": 1.55, "learning_rate": 2.4792684260609764e-06, "loss": 0.5526, "step": 13038 }, { "epoch": 1.55, "learning_rate": 2.478947013850185e-06, "loss": 0.5893, "step": 13039 }, { "epoch": 1.55, "learning_rate": 2.478625601987402e-06, "loss": 0.8059, "step": 13040 }, { "epoch": 1.55, "learning_rate": 2.4783041904779386e-06, "loss": 0.4198, "step": 13041 }, { "epoch": 1.55, "learning_rate": 2.4779827793271103e-06, "loss": 0.414, "step": 13042 }, { "epoch": 1.55, "learning_rate": 2.4776613685402275e-06, "loss": 0.5456, "step": 13043 }, { "epoch": 1.55, "learning_rate": 2.4773399581226045e-06, "loss": 0.5756, "step": 13044 }, { "epoch": 1.55, "learning_rate": 2.477018548079553e-06, "loss": 0.4913, "step": 13045 }, { "epoch": 1.55, "learning_rate": 2.476697138416388e-06, "loss": 0.7477, "step": 13046 }, { "epoch": 1.55, "learning_rate": 2.476375729138421e-06, "loss": 0.8601, "step": 13047 }, { "epoch": 1.55, "learning_rate": 2.4760543202509646e-06, "loss": 0.4301, "step": 13048 }, { "epoch": 1.55, "learning_rate": 2.4757329117593336e-06, "loss": 0.5231, "step": 13049 }, { "epoch": 1.55, "learning_rate": 2.475411503668838e-06, "loss": 0.6524, "step": 13050 }, { "epoch": 1.55, "learning_rate": 2.475090095984793e-06, "loss": 0.6052, "step": 13051 }, { "epoch": 1.55, "learning_rate": 2.4747686887125105e-06, "loss": 0.7836, "step": 13052 }, { "epoch": 1.55, "learning_rate": 2.4744472818573043e-06, "loss": 0.7533, "step": 13053 }, { "epoch": 1.55, "learning_rate": 2.4741258754244864e-06, "loss": 0.4509, "step": 13054 }, { "epoch": 1.55, "learning_rate": 2.4738044694193698e-06, "loss": 0.4511, "step": 13055 }, { "epoch": 1.55, "learning_rate": 2.4734830638472673e-06, "loss": 0.4573, "step": 13056 }, { "epoch": 1.55, "learning_rate": 2.473161658713493e-06, "loss": 0.6276, "step": 13057 }, { "epoch": 1.56, "learning_rate": 2.472840254023358e-06, "loss": 0.811, "step": 13058 }, { "epoch": 1.56, "learning_rate": 2.4725188497821758e-06, "loss": 0.8934, "step": 13059 }, { "epoch": 1.56, "learning_rate": 2.472197445995259e-06, "loss": 0.4136, "step": 13060 }, { "epoch": 1.56, "learning_rate": 2.4718760426679216e-06, "loss": 0.45, "step": 13061 }, { "epoch": 1.56, "learning_rate": 2.4715546398054755e-06, "loss": 0.5164, "step": 13062 }, { "epoch": 1.56, "learning_rate": 2.471233237413233e-06, "loss": 0.6039, "step": 13063 }, { "epoch": 1.56, "learning_rate": 2.470911835496508e-06, "loss": 0.66, "step": 13064 }, { "epoch": 1.56, "learning_rate": 2.4705904340606128e-06, "loss": 0.8538, "step": 13065 }, { "epoch": 1.56, "learning_rate": 2.4702690331108593e-06, "loss": 0.5121, "step": 13066 }, { "epoch": 1.56, "learning_rate": 2.4699476326525616e-06, "loss": 0.4377, "step": 13067 }, { "epoch": 1.56, "learning_rate": 2.469626232691033e-06, "loss": 0.5349, "step": 13068 }, { "epoch": 1.56, "learning_rate": 2.4693048332315846e-06, "loss": 0.5427, "step": 13069 }, { "epoch": 1.56, "learning_rate": 2.4689834342795296e-06, "loss": 0.6967, "step": 13070 }, { "epoch": 1.56, "learning_rate": 2.4686620358401807e-06, "loss": 0.7738, "step": 13071 }, { "epoch": 1.56, "learning_rate": 2.4683406379188525e-06, "loss": 0.9256, "step": 13072 }, { "epoch": 1.56, "learning_rate": 2.4680192405208552e-06, "loss": 0.426, "step": 13073 }, { "epoch": 1.56, "learning_rate": 2.467697843651502e-06, "loss": 0.5626, "step": 13074 }, { "epoch": 1.56, "learning_rate": 2.4673764473161073e-06, "loss": 0.5693, "step": 13075 }, { "epoch": 1.56, "learning_rate": 2.4670550515199827e-06, "loss": 0.6214, "step": 13076 }, { "epoch": 1.56, "learning_rate": 2.4667336562684397e-06, "loss": 0.8284, "step": 13077 }, { "epoch": 1.56, "learning_rate": 2.4664122615667926e-06, "loss": 0.8084, "step": 13078 }, { "epoch": 1.56, "learning_rate": 2.4660908674203545e-06, "loss": 0.4093, "step": 13079 }, { "epoch": 1.56, "learning_rate": 2.4657694738344366e-06, "loss": 0.5228, "step": 13080 }, { "epoch": 1.56, "learning_rate": 2.4654480808143515e-06, "loss": 0.518, "step": 13081 }, { "epoch": 1.56, "learning_rate": 2.4651266883654127e-06, "loss": 0.5334, "step": 13082 }, { "epoch": 1.56, "learning_rate": 2.464805296492934e-06, "loss": 0.7714, "step": 13083 }, { "epoch": 1.56, "learning_rate": 2.464483905202226e-06, "loss": 0.9548, "step": 13084 }, { "epoch": 1.56, "learning_rate": 2.464162514498601e-06, "loss": 0.3867, "step": 13085 }, { "epoch": 1.56, "learning_rate": 2.4638411243873737e-06, "loss": 0.4644, "step": 13086 }, { "epoch": 1.56, "learning_rate": 2.4635197348738564e-06, "loss": 0.5529, "step": 13087 }, { "epoch": 1.56, "learning_rate": 2.463198345963359e-06, "loss": 0.5633, "step": 13088 }, { "epoch": 1.56, "learning_rate": 2.462876957661197e-06, "loss": 0.7014, "step": 13089 }, { "epoch": 1.56, "learning_rate": 2.462555569972682e-06, "loss": 0.9241, "step": 13090 }, { "epoch": 1.56, "learning_rate": 2.4622341829031273e-06, "loss": 0.4949, "step": 13091 }, { "epoch": 1.56, "learning_rate": 2.4619127964578437e-06, "loss": 0.4651, "step": 13092 }, { "epoch": 1.56, "learning_rate": 2.461591410642145e-06, "loss": 0.5142, "step": 13093 }, { "epoch": 1.56, "learning_rate": 2.4612700254613433e-06, "loss": 0.6537, "step": 13094 }, { "epoch": 1.56, "learning_rate": 2.4609486409207518e-06, "loss": 0.7005, "step": 13095 }, { "epoch": 1.56, "learning_rate": 2.460627257025682e-06, "loss": 0.7699, "step": 13096 }, { "epoch": 1.56, "learning_rate": 2.460305873781447e-06, "loss": 0.851, "step": 13097 }, { "epoch": 1.56, "learning_rate": 2.4599844911933606e-06, "loss": 0.4447, "step": 13098 }, { "epoch": 1.56, "learning_rate": 2.459663109266732e-06, "loss": 0.5576, "step": 13099 }, { "epoch": 1.56, "learning_rate": 2.4593417280068763e-06, "loss": 0.4862, "step": 13100 }, { "epoch": 1.56, "learning_rate": 2.4590203474191054e-06, "loss": 0.666, "step": 13101 }, { "epoch": 1.56, "learning_rate": 2.4586989675087322e-06, "loss": 0.7941, "step": 13102 }, { "epoch": 1.56, "learning_rate": 2.4583775882810676e-06, "loss": 0.8547, "step": 13103 }, { "epoch": 1.56, "learning_rate": 2.4580562097414254e-06, "loss": 0.4258, "step": 13104 }, { "epoch": 1.56, "learning_rate": 2.457734831895117e-06, "loss": 0.5603, "step": 13105 }, { "epoch": 1.56, "learning_rate": 2.457413454747456e-06, "loss": 0.6549, "step": 13106 }, { "epoch": 1.56, "learning_rate": 2.4570920783037536e-06, "loss": 0.5562, "step": 13107 }, { "epoch": 1.56, "learning_rate": 2.4567707025693236e-06, "loss": 0.668, "step": 13108 }, { "epoch": 1.56, "learning_rate": 2.4564493275494773e-06, "loss": 0.8391, "step": 13109 }, { "epoch": 1.56, "learning_rate": 2.4561279532495267e-06, "loss": 0.4353, "step": 13110 }, { "epoch": 1.56, "learning_rate": 2.4558065796747847e-06, "loss": 0.3798, "step": 13111 }, { "epoch": 1.56, "learning_rate": 2.455485206830564e-06, "loss": 0.5855, "step": 13112 }, { "epoch": 1.56, "learning_rate": 2.4551638347221774e-06, "loss": 0.6619, "step": 13113 }, { "epoch": 1.56, "learning_rate": 2.454842463354935e-06, "loss": 0.7158, "step": 13114 }, { "epoch": 1.56, "learning_rate": 2.454521092734151e-06, "loss": 0.8126, "step": 13115 }, { "epoch": 1.56, "learning_rate": 2.4541997228651364e-06, "loss": 0.531, "step": 13116 }, { "epoch": 1.56, "learning_rate": 2.453878353753206e-06, "loss": 0.4673, "step": 13117 }, { "epoch": 1.56, "learning_rate": 2.4535569854036686e-06, "loss": 0.5598, "step": 13118 }, { "epoch": 1.56, "learning_rate": 2.453235617821839e-06, "loss": 0.6043, "step": 13119 }, { "epoch": 1.56, "learning_rate": 2.4529142510130278e-06, "loss": 0.7163, "step": 13120 }, { "epoch": 1.56, "learning_rate": 2.452592884982549e-06, "loss": 0.7999, "step": 13121 }, { "epoch": 1.56, "learning_rate": 2.4522715197357137e-06, "loss": 0.9139, "step": 13122 }, { "epoch": 1.56, "learning_rate": 2.451950155277833e-06, "loss": 0.4215, "step": 13123 }, { "epoch": 1.56, "learning_rate": 2.4516287916142208e-06, "loss": 0.5008, "step": 13124 }, { "epoch": 1.56, "learning_rate": 2.4513074287501895e-06, "loss": 0.7069, "step": 13125 }, { "epoch": 1.56, "learning_rate": 2.45098606669105e-06, "loss": 0.6934, "step": 13126 }, { "epoch": 1.56, "learning_rate": 2.4506647054421146e-06, "loss": 0.8495, "step": 13127 }, { "epoch": 1.56, "learning_rate": 2.4503433450086966e-06, "loss": 0.8846, "step": 13128 }, { "epoch": 1.56, "learning_rate": 2.4500219853961063e-06, "loss": 0.4448, "step": 13129 }, { "epoch": 1.56, "learning_rate": 2.4497006266096574e-06, "loss": 0.4611, "step": 13130 }, { "epoch": 1.56, "learning_rate": 2.4493792686546607e-06, "loss": 0.5814, "step": 13131 }, { "epoch": 1.56, "learning_rate": 2.44905791153643e-06, "loss": 0.5659, "step": 13132 }, { "epoch": 1.56, "learning_rate": 2.4487365552602755e-06, "loss": 0.685, "step": 13133 }, { "epoch": 1.56, "learning_rate": 2.44841519983151e-06, "loss": 0.8632, "step": 13134 }, { "epoch": 1.56, "learning_rate": 2.448093845255445e-06, "loss": 0.3999, "step": 13135 }, { "epoch": 1.56, "learning_rate": 2.447772491537395e-06, "loss": 0.4846, "step": 13136 }, { "epoch": 1.56, "learning_rate": 2.447451138682669e-06, "loss": 0.5705, "step": 13137 }, { "epoch": 1.56, "learning_rate": 2.4471297866965797e-06, "loss": 0.649, "step": 13138 }, { "epoch": 1.56, "learning_rate": 2.4468084355844394e-06, "loss": 0.7834, "step": 13139 }, { "epoch": 1.56, "learning_rate": 2.4464870853515613e-06, "loss": 0.8069, "step": 13140 }, { "epoch": 1.56, "learning_rate": 2.4461657360032553e-06, "loss": 0.5053, "step": 13141 }, { "epoch": 1.57, "learning_rate": 2.445844387544834e-06, "loss": 0.5687, "step": 13142 }, { "epoch": 1.57, "learning_rate": 2.4455230399816103e-06, "loss": 0.5537, "step": 13143 }, { "epoch": 1.57, "learning_rate": 2.445201693318895e-06, "loss": 0.6017, "step": 13144 }, { "epoch": 1.57, "learning_rate": 2.444880347562e-06, "loss": 0.6831, "step": 13145 }, { "epoch": 1.57, "learning_rate": 2.444559002716237e-06, "loss": 0.7639, "step": 13146 }, { "epoch": 1.57, "learning_rate": 2.4442376587869197e-06, "loss": 0.8465, "step": 13147 }, { "epoch": 1.57, "learning_rate": 2.4439163157793584e-06, "loss": 0.3634, "step": 13148 }, { "epoch": 1.57, "learning_rate": 2.443594973698864e-06, "loss": 0.5643, "step": 13149 }, { "epoch": 1.57, "learning_rate": 2.44327363255075e-06, "loss": 0.6436, "step": 13150 }, { "epoch": 1.57, "learning_rate": 2.442952292340328e-06, "loss": 0.6341, "step": 13151 }, { "epoch": 1.57, "learning_rate": 2.442630953072909e-06, "loss": 0.7637, "step": 13152 }, { "epoch": 1.57, "learning_rate": 2.4423096147538045e-06, "loss": 0.8061, "step": 13153 }, { "epoch": 1.57, "learning_rate": 2.4419882773883276e-06, "loss": 0.4646, "step": 13154 }, { "epoch": 1.57, "learning_rate": 2.44166694098179e-06, "loss": 0.7045, "step": 13155 }, { "epoch": 1.57, "learning_rate": 2.441345605539502e-06, "loss": 0.5292, "step": 13156 }, { "epoch": 1.57, "learning_rate": 2.4410242710667755e-06, "loss": 0.6508, "step": 13157 }, { "epoch": 1.57, "learning_rate": 2.440702937568924e-06, "loss": 0.7673, "step": 13158 }, { "epoch": 1.57, "learning_rate": 2.4403816050512574e-06, "loss": 0.8602, "step": 13159 }, { "epoch": 1.57, "learning_rate": 2.440060273519087e-06, "loss": 0.4639, "step": 13160 }, { "epoch": 1.57, "learning_rate": 2.4397389429777265e-06, "loss": 0.4376, "step": 13161 }, { "epoch": 1.57, "learning_rate": 2.4394176134324867e-06, "loss": 0.5345, "step": 13162 }, { "epoch": 1.57, "learning_rate": 2.4390962848886775e-06, "loss": 0.5489, "step": 13163 }, { "epoch": 1.57, "learning_rate": 2.4387749573516122e-06, "loss": 0.8111, "step": 13164 }, { "epoch": 1.57, "learning_rate": 2.4384536308266023e-06, "loss": 0.8881, "step": 13165 }, { "epoch": 1.57, "learning_rate": 2.4381323053189593e-06, "loss": 0.4872, "step": 13166 }, { "epoch": 1.57, "learning_rate": 2.4378109808339935e-06, "loss": 0.4816, "step": 13167 }, { "epoch": 1.57, "learning_rate": 2.4374896573770178e-06, "loss": 0.5036, "step": 13168 }, { "epoch": 1.57, "learning_rate": 2.4371683349533435e-06, "loss": 0.5416, "step": 13169 }, { "epoch": 1.57, "learning_rate": 2.4368470135682827e-06, "loss": 0.6623, "step": 13170 }, { "epoch": 1.57, "learning_rate": 2.436525693227145e-06, "loss": 0.9219, "step": 13171 }, { "epoch": 1.57, "learning_rate": 2.436204373935243e-06, "loss": 0.771, "step": 13172 }, { "epoch": 1.57, "learning_rate": 2.435883055697889e-06, "loss": 0.3892, "step": 13173 }, { "epoch": 1.57, "learning_rate": 2.4355617385203924e-06, "loss": 0.4037, "step": 13174 }, { "epoch": 1.57, "learning_rate": 2.435240422408065e-06, "loss": 0.567, "step": 13175 }, { "epoch": 1.57, "learning_rate": 2.4349191073662203e-06, "loss": 0.6511, "step": 13176 }, { "epoch": 1.57, "learning_rate": 2.4345977934001683e-06, "loss": 0.7574, "step": 13177 }, { "epoch": 1.57, "learning_rate": 2.4342764805152196e-06, "loss": 0.881, "step": 13178 }, { "epoch": 1.57, "learning_rate": 2.4339551687166863e-06, "loss": 0.2946, "step": 13179 }, { "epoch": 1.57, "learning_rate": 2.433633858009879e-06, "loss": 0.4368, "step": 13180 }, { "epoch": 1.57, "learning_rate": 2.4333125484001107e-06, "loss": 0.4876, "step": 13181 }, { "epoch": 1.57, "learning_rate": 2.432991239892691e-06, "loss": 0.6679, "step": 13182 }, { "epoch": 1.57, "learning_rate": 2.432669932492932e-06, "loss": 0.7324, "step": 13183 }, { "epoch": 1.57, "learning_rate": 2.4323486262061437e-06, "loss": 0.8027, "step": 13184 }, { "epoch": 1.57, "learning_rate": 2.4320273210376397e-06, "loss": 0.4509, "step": 13185 }, { "epoch": 1.57, "learning_rate": 2.4317060169927287e-06, "loss": 0.4829, "step": 13186 }, { "epoch": 1.57, "learning_rate": 2.4313847140767233e-06, "loss": 0.4931, "step": 13187 }, { "epoch": 1.57, "learning_rate": 2.431063412294934e-06, "loss": 0.6177, "step": 13188 }, { "epoch": 1.57, "learning_rate": 2.4307421116526735e-06, "loss": 0.7011, "step": 13189 }, { "epoch": 1.57, "learning_rate": 2.430420812155251e-06, "loss": 0.84, "step": 13190 }, { "epoch": 1.57, "learning_rate": 2.4300995138079776e-06, "loss": 0.6043, "step": 13191 }, { "epoch": 1.57, "learning_rate": 2.4297782166161667e-06, "loss": 0.459, "step": 13192 }, { "epoch": 1.57, "learning_rate": 2.4294569205851263e-06, "loss": 0.5548, "step": 13193 }, { "epoch": 1.57, "learning_rate": 2.4291356257201696e-06, "loss": 0.574, "step": 13194 }, { "epoch": 1.57, "learning_rate": 2.4288143320266064e-06, "loss": 0.6122, "step": 13195 }, { "epoch": 1.57, "learning_rate": 2.428493039509749e-06, "loss": 0.7719, "step": 13196 }, { "epoch": 1.57, "learning_rate": 2.428171748174907e-06, "loss": 0.8276, "step": 13197 }, { "epoch": 1.57, "learning_rate": 2.4278504580273923e-06, "loss": 0.3872, "step": 13198 }, { "epoch": 1.57, "learning_rate": 2.4275291690725152e-06, "loss": 0.4736, "step": 13199 }, { "epoch": 1.57, "learning_rate": 2.4272078813155883e-06, "loss": 0.5521, "step": 13200 }, { "epoch": 1.57, "learning_rate": 2.4268865947619204e-06, "loss": 0.5877, "step": 13201 }, { "epoch": 1.57, "learning_rate": 2.4265653094168224e-06, "loss": 0.755, "step": 13202 }, { "epoch": 1.57, "learning_rate": 2.426244025285607e-06, "loss": 0.9688, "step": 13203 }, { "epoch": 1.57, "learning_rate": 2.425922742373584e-06, "loss": 0.317, "step": 13204 }, { "epoch": 1.57, "learning_rate": 2.4256014606860644e-06, "loss": 0.4442, "step": 13205 }, { "epoch": 1.57, "learning_rate": 2.4252801802283584e-06, "loss": 0.6415, "step": 13206 }, { "epoch": 1.57, "learning_rate": 2.424958901005778e-06, "loss": 0.7092, "step": 13207 }, { "epoch": 1.57, "learning_rate": 2.4246376230236325e-06, "loss": 0.7644, "step": 13208 }, { "epoch": 1.57, "learning_rate": 2.4243163462872336e-06, "loss": 0.9237, "step": 13209 }, { "epoch": 1.57, "learning_rate": 2.4239950708018913e-06, "loss": 0.4126, "step": 13210 }, { "epoch": 1.57, "learning_rate": 2.4236737965729183e-06, "loss": 0.4259, "step": 13211 }, { "epoch": 1.57, "learning_rate": 2.423352523605623e-06, "loss": 0.4018, "step": 13212 }, { "epoch": 1.57, "learning_rate": 2.4230312519053163e-06, "loss": 0.6067, "step": 13213 }, { "epoch": 1.57, "learning_rate": 2.42270998147731e-06, "loss": 0.6577, "step": 13214 }, { "epoch": 1.57, "learning_rate": 2.422388712326915e-06, "loss": 0.8886, "step": 13215 }, { "epoch": 1.57, "learning_rate": 2.42206744445944e-06, "loss": 0.6302, "step": 13216 }, { "epoch": 1.57, "learning_rate": 2.421746177880197e-06, "loss": 0.418, "step": 13217 }, { "epoch": 1.57, "learning_rate": 2.4214249125944964e-06, "loss": 0.5725, "step": 13218 }, { "epoch": 1.57, "learning_rate": 2.4211036486076493e-06, "loss": 0.5439, "step": 13219 }, { "epoch": 1.57, "learning_rate": 2.420782385924964e-06, "loss": 0.6419, "step": 13220 }, { "epoch": 1.57, "learning_rate": 2.420461124551753e-06, "loss": 0.8701, "step": 13221 }, { "epoch": 1.57, "learning_rate": 2.4201398644933273e-06, "loss": 0.9213, "step": 13222 }, { "epoch": 1.57, "learning_rate": 2.4198186057549953e-06, "loss": 0.4568, "step": 13223 }, { "epoch": 1.57, "learning_rate": 2.4194973483420687e-06, "loss": 0.6499, "step": 13224 }, { "epoch": 1.57, "learning_rate": 2.419176092259857e-06, "loss": 0.4619, "step": 13225 }, { "epoch": 1.58, "learning_rate": 2.4188548375136728e-06, "loss": 0.5787, "step": 13226 }, { "epoch": 1.58, "learning_rate": 2.4185335841088243e-06, "loss": 0.7406, "step": 13227 }, { "epoch": 1.58, "learning_rate": 2.418212332050622e-06, "loss": 0.7889, "step": 13228 }, { "epoch": 1.58, "learning_rate": 2.417891081344377e-06, "loss": 0.547, "step": 13229 }, { "epoch": 1.58, "learning_rate": 2.4175698319954e-06, "loss": 0.4932, "step": 13230 }, { "epoch": 1.58, "learning_rate": 2.4172485840089993e-06, "loss": 0.6012, "step": 13231 }, { "epoch": 1.58, "learning_rate": 2.4169273373904865e-06, "loss": 0.5769, "step": 13232 }, { "epoch": 1.58, "learning_rate": 2.416606092145173e-06, "loss": 0.7108, "step": 13233 }, { "epoch": 1.58, "learning_rate": 2.4162848482783678e-06, "loss": 0.736, "step": 13234 }, { "epoch": 1.58, "learning_rate": 2.41596360579538e-06, "loss": 0.3672, "step": 13235 }, { "epoch": 1.58, "learning_rate": 2.4156423647015213e-06, "loss": 0.4955, "step": 13236 }, { "epoch": 1.58, "learning_rate": 2.4153211250021013e-06, "loss": 0.6015, "step": 13237 }, { "epoch": 1.58, "learning_rate": 2.4149998867024304e-06, "loss": 0.5126, "step": 13238 }, { "epoch": 1.58, "learning_rate": 2.414678649807818e-06, "loss": 0.6458, "step": 13239 }, { "epoch": 1.58, "learning_rate": 2.4143574143235755e-06, "loss": 0.938, "step": 13240 }, { "epoch": 1.58, "learning_rate": 2.414036180255012e-06, "loss": 0.5869, "step": 13241 }, { "epoch": 1.58, "learning_rate": 2.413714947607437e-06, "loss": 0.5181, "step": 13242 }, { "epoch": 1.58, "learning_rate": 2.413393716386161e-06, "loss": 0.5023, "step": 13243 }, { "epoch": 1.58, "learning_rate": 2.4130724865964947e-06, "loss": 0.4907, "step": 13244 }, { "epoch": 1.58, "learning_rate": 2.4127512582437486e-06, "loss": 0.7259, "step": 13245 }, { "epoch": 1.58, "learning_rate": 2.4124300313332297e-06, "loss": 0.8945, "step": 13246 }, { "epoch": 1.58, "learning_rate": 2.4121088058702504e-06, "loss": 0.8117, "step": 13247 }, { "epoch": 1.58, "learning_rate": 2.4117875818601197e-06, "loss": 0.4911, "step": 13248 }, { "epoch": 1.58, "learning_rate": 2.411466359308148e-06, "loss": 0.5395, "step": 13249 }, { "epoch": 1.58, "learning_rate": 2.4111451382196447e-06, "loss": 0.6017, "step": 13250 }, { "epoch": 1.58, "learning_rate": 2.41082391859992e-06, "loss": 0.5421, "step": 13251 }, { "epoch": 1.58, "learning_rate": 2.4105027004542838e-06, "loss": 0.6632, "step": 13252 }, { "epoch": 1.58, "learning_rate": 2.410181483788044e-06, "loss": 0.864, "step": 13253 }, { "epoch": 1.58, "learning_rate": 2.4098602686065123e-06, "loss": 0.51, "step": 13254 }, { "epoch": 1.58, "learning_rate": 2.4095390549149984e-06, "loss": 0.4693, "step": 13255 }, { "epoch": 1.58, "learning_rate": 2.409217842718812e-06, "loss": 0.5267, "step": 13256 }, { "epoch": 1.58, "learning_rate": 2.4088966320232608e-06, "loss": 0.5818, "step": 13257 }, { "epoch": 1.58, "learning_rate": 2.408575422833657e-06, "loss": 0.7925, "step": 13258 }, { "epoch": 1.58, "learning_rate": 2.4082542151553085e-06, "loss": 0.8322, "step": 13259 }, { "epoch": 1.58, "learning_rate": 2.4079330089935266e-06, "loss": 0.4283, "step": 13260 }, { "epoch": 1.58, "learning_rate": 2.4076118043536186e-06, "loss": 0.4358, "step": 13261 }, { "epoch": 1.58, "learning_rate": 2.4072906012408957e-06, "loss": 0.592, "step": 13262 }, { "epoch": 1.58, "learning_rate": 2.4069693996606667e-06, "loss": 0.6452, "step": 13263 }, { "epoch": 1.58, "learning_rate": 2.406648199618242e-06, "loss": 0.7849, "step": 13264 }, { "epoch": 1.58, "learning_rate": 2.4063270011189304e-06, "loss": 0.7372, "step": 13265 }, { "epoch": 1.58, "learning_rate": 2.4060058041680405e-06, "loss": 0.4638, "step": 13266 }, { "epoch": 1.58, "learning_rate": 2.4056846087708823e-06, "loss": 0.4778, "step": 13267 }, { "epoch": 1.58, "learning_rate": 2.4053634149327666e-06, "loss": 0.5842, "step": 13268 }, { "epoch": 1.58, "learning_rate": 2.4050422226590013e-06, "loss": 0.6403, "step": 13269 }, { "epoch": 1.58, "learning_rate": 2.404721031954895e-06, "loss": 0.6078, "step": 13270 }, { "epoch": 1.58, "learning_rate": 2.4043998428257596e-06, "loss": 0.7127, "step": 13271 }, { "epoch": 1.58, "learning_rate": 2.404078655276902e-06, "loss": 0.8009, "step": 13272 }, { "epoch": 1.58, "learning_rate": 2.403757469313632e-06, "loss": 0.405, "step": 13273 }, { "epoch": 1.58, "learning_rate": 2.4034362849412595e-06, "loss": 0.4971, "step": 13274 }, { "epoch": 1.58, "learning_rate": 2.4031151021650938e-06, "loss": 0.5813, "step": 13275 }, { "epoch": 1.58, "learning_rate": 2.402793920990443e-06, "loss": 0.6094, "step": 13276 }, { "epoch": 1.58, "learning_rate": 2.4024727414226165e-06, "loss": 0.8542, "step": 13277 }, { "epoch": 1.58, "learning_rate": 2.4021515634669237e-06, "loss": 0.7913, "step": 13278 }, { "epoch": 1.58, "learning_rate": 2.401830387128675e-06, "loss": 0.4609, "step": 13279 }, { "epoch": 1.58, "learning_rate": 2.4015092124131773e-06, "loss": 0.5439, "step": 13280 }, { "epoch": 1.58, "learning_rate": 2.4011880393257404e-06, "loss": 0.5721, "step": 13281 }, { "epoch": 1.58, "learning_rate": 2.4008668678716735e-06, "loss": 0.6329, "step": 13282 }, { "epoch": 1.58, "learning_rate": 2.4005456980562863e-06, "loss": 0.77, "step": 13283 }, { "epoch": 1.58, "learning_rate": 2.400224529884887e-06, "loss": 0.8637, "step": 13284 }, { "epoch": 1.58, "learning_rate": 2.399903363362784e-06, "loss": 0.4315, "step": 13285 }, { "epoch": 1.58, "learning_rate": 2.3995821984952876e-06, "loss": 0.4121, "step": 13286 }, { "epoch": 1.58, "learning_rate": 2.3992610352877056e-06, "loss": 0.6501, "step": 13287 }, { "epoch": 1.58, "learning_rate": 2.3989398737453464e-06, "loss": 0.6363, "step": 13288 }, { "epoch": 1.58, "learning_rate": 2.3986187138735195e-06, "loss": 0.7067, "step": 13289 }, { "epoch": 1.58, "learning_rate": 2.398297555677535e-06, "loss": 0.7593, "step": 13290 }, { "epoch": 1.58, "learning_rate": 2.3979763991626996e-06, "loss": 0.8065, "step": 13291 }, { "epoch": 1.58, "learning_rate": 2.397655244334323e-06, "loss": 0.459, "step": 13292 }, { "epoch": 1.58, "learning_rate": 2.397334091197714e-06, "loss": 0.5438, "step": 13293 }, { "epoch": 1.58, "learning_rate": 2.397012939758181e-06, "loss": 0.558, "step": 13294 }, { "epoch": 1.58, "learning_rate": 2.3966917900210326e-06, "loss": 0.6961, "step": 13295 }, { "epoch": 1.58, "learning_rate": 2.396370641991577e-06, "loss": 0.8972, "step": 13296 }, { "epoch": 1.58, "learning_rate": 2.3960494956751243e-06, "loss": 0.7441, "step": 13297 }, { "epoch": 1.58, "learning_rate": 2.395728351076983e-06, "loss": 0.5081, "step": 13298 }, { "epoch": 1.58, "learning_rate": 2.395407208202459e-06, "loss": 0.4937, "step": 13299 }, { "epoch": 1.58, "learning_rate": 2.3950860670568627e-06, "loss": 0.5715, "step": 13300 }, { "epoch": 1.58, "learning_rate": 2.394764927645504e-06, "loss": 0.6159, "step": 13301 }, { "epoch": 1.58, "learning_rate": 2.394443789973689e-06, "loss": 0.7591, "step": 13302 }, { "epoch": 1.58, "learning_rate": 2.3941226540467265e-06, "loss": 0.8451, "step": 13303 }, { "epoch": 1.58, "learning_rate": 2.393801519869926e-06, "loss": 0.4175, "step": 13304 }, { "epoch": 1.58, "learning_rate": 2.393480387448596e-06, "loss": 0.5684, "step": 13305 }, { "epoch": 1.58, "learning_rate": 2.393159256788043e-06, "loss": 0.5878, "step": 13306 }, { "epoch": 1.58, "learning_rate": 2.3928381278935763e-06, "loss": 0.7281, "step": 13307 }, { "epoch": 1.58, "learning_rate": 2.392517000770505e-06, "loss": 0.7768, "step": 13308 }, { "epoch": 1.58, "learning_rate": 2.3921958754241375e-06, "loss": 0.9046, "step": 13309 }, { "epoch": 1.59, "learning_rate": 2.3918747518597794e-06, "loss": 0.4717, "step": 13310 }, { "epoch": 1.59, "learning_rate": 2.3915536300827414e-06, "loss": 0.4838, "step": 13311 }, { "epoch": 1.59, "learning_rate": 2.3912325100983313e-06, "loss": 0.5604, "step": 13312 }, { "epoch": 1.59, "learning_rate": 2.3909113919118574e-06, "loss": 0.6272, "step": 13313 }, { "epoch": 1.59, "learning_rate": 2.3905902755286265e-06, "loss": 0.6954, "step": 13314 }, { "epoch": 1.59, "learning_rate": 2.3902691609539477e-06, "loss": 0.752, "step": 13315 }, { "epoch": 1.59, "learning_rate": 2.38994804819313e-06, "loss": 0.6942, "step": 13316 }, { "epoch": 1.59, "learning_rate": 2.3896269372514787e-06, "loss": 0.4662, "step": 13317 }, { "epoch": 1.59, "learning_rate": 2.3893058281343035e-06, "loss": 0.4807, "step": 13318 }, { "epoch": 1.59, "learning_rate": 2.3889847208469126e-06, "loss": 0.5123, "step": 13319 }, { "epoch": 1.59, "learning_rate": 2.3886636153946144e-06, "loss": 0.6355, "step": 13320 }, { "epoch": 1.59, "learning_rate": 2.3883425117827146e-06, "loss": 0.8484, "step": 13321 }, { "epoch": 1.59, "learning_rate": 2.3880214100165233e-06, "loss": 0.7935, "step": 13322 }, { "epoch": 1.59, "learning_rate": 2.387700310101347e-06, "loss": 0.4673, "step": 13323 }, { "epoch": 1.59, "learning_rate": 2.387379212042495e-06, "loss": 0.5471, "step": 13324 }, { "epoch": 1.59, "learning_rate": 2.387058115845273e-06, "loss": 0.6074, "step": 13325 }, { "epoch": 1.59, "learning_rate": 2.38673702151499e-06, "loss": 0.702, "step": 13326 }, { "epoch": 1.59, "learning_rate": 2.3864159290569534e-06, "loss": 0.7271, "step": 13327 }, { "epoch": 1.59, "learning_rate": 2.386094838476472e-06, "loss": 0.9229, "step": 13328 }, { "epoch": 1.59, "learning_rate": 2.3857737497788515e-06, "loss": 0.5374, "step": 13329 }, { "epoch": 1.59, "learning_rate": 2.385452662969401e-06, "loss": 0.5403, "step": 13330 }, { "epoch": 1.59, "learning_rate": 2.385131578053427e-06, "loss": 0.5161, "step": 13331 }, { "epoch": 1.59, "learning_rate": 2.3848104950362387e-06, "loss": 0.6007, "step": 13332 }, { "epoch": 1.59, "learning_rate": 2.384489413923142e-06, "loss": 0.7864, "step": 13333 }, { "epoch": 1.59, "learning_rate": 2.384168334719445e-06, "loss": 0.8575, "step": 13334 }, { "epoch": 1.59, "learning_rate": 2.383847257430455e-06, "loss": 0.3569, "step": 13335 }, { "epoch": 1.59, "learning_rate": 2.38352618206148e-06, "loss": 0.4488, "step": 13336 }, { "epoch": 1.59, "learning_rate": 2.383205108617826e-06, "loss": 0.4966, "step": 13337 }, { "epoch": 1.59, "learning_rate": 2.3828840371048022e-06, "loss": 0.568, "step": 13338 }, { "epoch": 1.59, "learning_rate": 2.3825629675277154e-06, "loss": 0.7925, "step": 13339 }, { "epoch": 1.59, "learning_rate": 2.3822418998918714e-06, "loss": 1.0059, "step": 13340 }, { "epoch": 1.59, "learning_rate": 2.3819208342025798e-06, "loss": 0.521, "step": 13341 }, { "epoch": 1.59, "learning_rate": 2.3815997704651456e-06, "loss": 0.4723, "step": 13342 }, { "epoch": 1.59, "learning_rate": 2.3812787086848785e-06, "loss": 0.5799, "step": 13343 }, { "epoch": 1.59, "learning_rate": 2.3809576488670838e-06, "loss": 0.5158, "step": 13344 }, { "epoch": 1.59, "learning_rate": 2.3806365910170685e-06, "loss": 0.6766, "step": 13345 }, { "epoch": 1.59, "learning_rate": 2.3803155351401404e-06, "loss": 0.8018, "step": 13346 }, { "epoch": 1.59, "learning_rate": 2.3799944812416075e-06, "loss": 0.8502, "step": 13347 }, { "epoch": 1.59, "learning_rate": 2.3796734293267755e-06, "loss": 0.4327, "step": 13348 }, { "epoch": 1.59, "learning_rate": 2.3793523794009517e-06, "loss": 0.5675, "step": 13349 }, { "epoch": 1.59, "learning_rate": 2.3790313314694437e-06, "loss": 0.5811, "step": 13350 }, { "epoch": 1.59, "learning_rate": 2.378710285537557e-06, "loss": 0.6194, "step": 13351 }, { "epoch": 1.59, "learning_rate": 2.3783892416106e-06, "loss": 0.8765, "step": 13352 }, { "epoch": 1.59, "learning_rate": 2.3780681996938785e-06, "loss": 0.7466, "step": 13353 }, { "epoch": 1.59, "learning_rate": 2.377747159792701e-06, "loss": 0.373, "step": 13354 }, { "epoch": 1.59, "learning_rate": 2.3774261219123727e-06, "loss": 0.5146, "step": 13355 }, { "epoch": 1.59, "learning_rate": 2.3771050860582e-06, "loss": 0.5174, "step": 13356 }, { "epoch": 1.59, "learning_rate": 2.376784052235491e-06, "loss": 0.5567, "step": 13357 }, { "epoch": 1.59, "learning_rate": 2.3764630204495527e-06, "loss": 0.6656, "step": 13358 }, { "epoch": 1.59, "learning_rate": 2.3761419907056907e-06, "loss": 0.8503, "step": 13359 }, { "epoch": 1.59, "learning_rate": 2.3758209630092116e-06, "loss": 0.4997, "step": 13360 }, { "epoch": 1.59, "learning_rate": 2.3754999373654223e-06, "loss": 0.5387, "step": 13361 }, { "epoch": 1.59, "learning_rate": 2.375178913779631e-06, "loss": 0.4464, "step": 13362 }, { "epoch": 1.59, "learning_rate": 2.3748578922571406e-06, "loss": 0.5589, "step": 13363 }, { "epoch": 1.59, "learning_rate": 2.3745368728032606e-06, "loss": 0.6615, "step": 13364 }, { "epoch": 1.59, "learning_rate": 2.374215855423297e-06, "loss": 0.9005, "step": 13365 }, { "epoch": 1.59, "learning_rate": 2.3738948401225557e-06, "loss": 0.5289, "step": 13366 }, { "epoch": 1.59, "learning_rate": 2.3735738269063428e-06, "loss": 0.4338, "step": 13367 }, { "epoch": 1.59, "learning_rate": 2.373252815779965e-06, "loss": 0.4739, "step": 13368 }, { "epoch": 1.59, "learning_rate": 2.3729318067487303e-06, "loss": 0.6158, "step": 13369 }, { "epoch": 1.59, "learning_rate": 2.3726107998179427e-06, "loss": 0.7876, "step": 13370 }, { "epoch": 1.59, "learning_rate": 2.3722897949929083e-06, "loss": 0.8223, "step": 13371 }, { "epoch": 1.59, "learning_rate": 2.3719687922789354e-06, "loss": 0.7882, "step": 13372 }, { "epoch": 1.59, "learning_rate": 2.3716477916813293e-06, "loss": 0.4324, "step": 13373 }, { "epoch": 1.59, "learning_rate": 2.371326793205395e-06, "loss": 0.6129, "step": 13374 }, { "epoch": 1.59, "learning_rate": 2.3710057968564394e-06, "loss": 0.5649, "step": 13375 }, { "epoch": 1.59, "learning_rate": 2.3706848026397696e-06, "loss": 0.6929, "step": 13376 }, { "epoch": 1.59, "learning_rate": 2.3703638105606918e-06, "loss": 0.6973, "step": 13377 }, { "epoch": 1.59, "learning_rate": 2.3700428206245094e-06, "loss": 0.8091, "step": 13378 }, { "epoch": 1.59, "learning_rate": 2.3697218328365305e-06, "loss": 0.4965, "step": 13379 }, { "epoch": 1.59, "learning_rate": 2.369400847202061e-06, "loss": 0.4417, "step": 13380 }, { "epoch": 1.59, "learning_rate": 2.3690798637264064e-06, "loss": 0.6509, "step": 13381 }, { "epoch": 1.59, "learning_rate": 2.3687588824148723e-06, "loss": 0.6436, "step": 13382 }, { "epoch": 1.59, "learning_rate": 2.368437903272765e-06, "loss": 0.6798, "step": 13383 }, { "epoch": 1.59, "learning_rate": 2.368116926305391e-06, "loss": 0.8317, "step": 13384 }, { "epoch": 1.59, "learning_rate": 2.367795951518054e-06, "loss": 0.6713, "step": 13385 }, { "epoch": 1.59, "learning_rate": 2.3674749789160613e-06, "loss": 0.4809, "step": 13386 }, { "epoch": 1.59, "learning_rate": 2.3671540085047186e-06, "loss": 0.4934, "step": 13387 }, { "epoch": 1.59, "learning_rate": 2.366833040289332e-06, "loss": 0.6529, "step": 13388 }, { "epoch": 1.59, "learning_rate": 2.3665120742752053e-06, "loss": 0.6629, "step": 13389 }, { "epoch": 1.59, "learning_rate": 2.3661911104676457e-06, "loss": 0.9331, "step": 13390 }, { "epoch": 1.59, "learning_rate": 2.3658701488719575e-06, "loss": 0.486, "step": 13391 }, { "epoch": 1.59, "learning_rate": 2.3655491894934484e-06, "loss": 0.4968, "step": 13392 }, { "epoch": 1.59, "learning_rate": 2.3652282323374212e-06, "loss": 0.5598, "step": 13393 }, { "epoch": 1.6, "learning_rate": 2.3649072774091834e-06, "loss": 0.5906, "step": 13394 }, { "epoch": 1.6, "learning_rate": 2.36458632471404e-06, "loss": 0.78, "step": 13395 }, { "epoch": 1.6, "learning_rate": 2.364265374257295e-06, "loss": 0.8513, "step": 13396 }, { "epoch": 1.6, "learning_rate": 2.363944426044255e-06, "loss": 0.7281, "step": 13397 }, { "epoch": 1.6, "learning_rate": 2.363623480080225e-06, "loss": 0.4081, "step": 13398 }, { "epoch": 1.6, "learning_rate": 2.3633025363705115e-06, "loss": 0.5339, "step": 13399 }, { "epoch": 1.6, "learning_rate": 2.362981594920417e-06, "loss": 0.5997, "step": 13400 }, { "epoch": 1.6, "learning_rate": 2.3626606557352487e-06, "loss": 0.5672, "step": 13401 }, { "epoch": 1.6, "learning_rate": 2.3623397188203113e-06, "loss": 0.8402, "step": 13402 }, { "epoch": 1.6, "learning_rate": 2.3620187841809105e-06, "loss": 0.809, "step": 13403 }, { "epoch": 1.6, "learning_rate": 2.36169785182235e-06, "loss": 0.3937, "step": 13404 }, { "epoch": 1.6, "learning_rate": 2.361376921749936e-06, "loss": 0.5698, "step": 13405 }, { "epoch": 1.6, "learning_rate": 2.3610559939689727e-06, "loss": 0.6425, "step": 13406 }, { "epoch": 1.6, "learning_rate": 2.360735068484766e-06, "loss": 0.6272, "step": 13407 }, { "epoch": 1.6, "learning_rate": 2.3604141453026206e-06, "loss": 0.6823, "step": 13408 }, { "epoch": 1.6, "learning_rate": 2.36009322442784e-06, "loss": 0.8167, "step": 13409 }, { "epoch": 1.6, "learning_rate": 2.3597723058657304e-06, "loss": 0.455, "step": 13410 }, { "epoch": 1.6, "learning_rate": 2.3594513896215974e-06, "loss": 0.4721, "step": 13411 }, { "epoch": 1.6, "learning_rate": 2.359130475700744e-06, "loss": 0.5232, "step": 13412 }, { "epoch": 1.6, "learning_rate": 2.358809564108475e-06, "loss": 0.5921, "step": 13413 }, { "epoch": 1.6, "learning_rate": 2.358488654850097e-06, "loss": 0.6248, "step": 13414 }, { "epoch": 1.6, "learning_rate": 2.3581677479309126e-06, "loss": 0.8015, "step": 13415 }, { "epoch": 1.6, "learning_rate": 2.3578468433562276e-06, "loss": 0.6059, "step": 13416 }, { "epoch": 1.6, "learning_rate": 2.3575259411313454e-06, "loss": 0.5178, "step": 13417 }, { "epoch": 1.6, "learning_rate": 2.357205041261573e-06, "loss": 0.5688, "step": 13418 }, { "epoch": 1.6, "learning_rate": 2.3568841437522123e-06, "loss": 0.6687, "step": 13419 }, { "epoch": 1.6, "learning_rate": 2.3565632486085683e-06, "loss": 0.6465, "step": 13420 }, { "epoch": 1.6, "learning_rate": 2.3562423558359457e-06, "loss": 0.8701, "step": 13421 }, { "epoch": 1.6, "learning_rate": 2.35592146543965e-06, "loss": 0.8471, "step": 13422 }, { "epoch": 1.6, "learning_rate": 2.355600577424984e-06, "loss": 0.3908, "step": 13423 }, { "epoch": 1.6, "learning_rate": 2.3552796917972527e-06, "loss": 0.4664, "step": 13424 }, { "epoch": 1.6, "learning_rate": 2.3549588085617594e-06, "loss": 0.5542, "step": 13425 }, { "epoch": 1.6, "learning_rate": 2.3546379277238107e-06, "loss": 0.6664, "step": 13426 }, { "epoch": 1.6, "learning_rate": 2.3543170492887086e-06, "loss": 0.8264, "step": 13427 }, { "epoch": 1.6, "learning_rate": 2.3539961732617576e-06, "loss": 0.8228, "step": 13428 }, { "epoch": 1.6, "learning_rate": 2.353675299648263e-06, "loss": 0.4375, "step": 13429 }, { "epoch": 1.6, "learning_rate": 2.3533544284535273e-06, "loss": 0.4584, "step": 13430 }, { "epoch": 1.6, "learning_rate": 2.3530335596828547e-06, "loss": 0.6385, "step": 13431 }, { "epoch": 1.6, "learning_rate": 2.3527126933415497e-06, "loss": 0.5959, "step": 13432 }, { "epoch": 1.6, "learning_rate": 2.3523918294349176e-06, "loss": 0.7372, "step": 13433 }, { "epoch": 1.6, "learning_rate": 2.35207096796826e-06, "loss": 0.9008, "step": 13434 }, { "epoch": 1.6, "learning_rate": 2.3517501089468816e-06, "loss": 0.3633, "step": 13435 }, { "epoch": 1.6, "learning_rate": 2.3514292523760863e-06, "loss": 0.5169, "step": 13436 }, { "epoch": 1.6, "learning_rate": 2.3511083982611784e-06, "loss": 0.5846, "step": 13437 }, { "epoch": 1.6, "learning_rate": 2.350787546607461e-06, "loss": 0.6744, "step": 13438 }, { "epoch": 1.6, "learning_rate": 2.350466697420237e-06, "loss": 0.7682, "step": 13439 }, { "epoch": 1.6, "learning_rate": 2.3501458507048126e-06, "loss": 0.8445, "step": 13440 }, { "epoch": 1.6, "learning_rate": 2.3498250064664896e-06, "loss": 0.5332, "step": 13441 }, { "epoch": 1.6, "learning_rate": 2.349504164710571e-06, "loss": 0.3551, "step": 13442 }, { "epoch": 1.6, "learning_rate": 2.349183325442361e-06, "loss": 0.6126, "step": 13443 }, { "epoch": 1.6, "learning_rate": 2.3488624886671646e-06, "loss": 0.6251, "step": 13444 }, { "epoch": 1.6, "learning_rate": 2.3485416543902833e-06, "loss": 0.662, "step": 13445 }, { "epoch": 1.6, "learning_rate": 2.348220822617021e-06, "loss": 0.8569, "step": 13446 }, { "epoch": 1.6, "learning_rate": 2.3478999933526815e-06, "loss": 0.7705, "step": 13447 }, { "epoch": 1.6, "learning_rate": 2.3475791666025684e-06, "loss": 0.4976, "step": 13448 }, { "epoch": 1.6, "learning_rate": 2.347258342371983e-06, "loss": 0.5773, "step": 13449 }, { "epoch": 1.6, "learning_rate": 2.3469375206662305e-06, "loss": 0.7151, "step": 13450 }, { "epoch": 1.6, "learning_rate": 2.346616701490614e-06, "loss": 0.7084, "step": 13451 }, { "epoch": 1.6, "learning_rate": 2.346295884850437e-06, "loss": 0.7315, "step": 13452 }, { "epoch": 1.6, "learning_rate": 2.3459750707510002e-06, "loss": 0.8356, "step": 13453 }, { "epoch": 1.6, "learning_rate": 2.345654259197609e-06, "loss": 0.3712, "step": 13454 }, { "epoch": 1.6, "learning_rate": 2.3453334501955666e-06, "loss": 0.5204, "step": 13455 }, { "epoch": 1.6, "learning_rate": 2.3450126437501754e-06, "loss": 0.5153, "step": 13456 }, { "epoch": 1.6, "learning_rate": 2.3446918398667374e-06, "loss": 0.5695, "step": 13457 }, { "epoch": 1.6, "learning_rate": 2.3443710385505563e-06, "loss": 0.7925, "step": 13458 }, { "epoch": 1.6, "learning_rate": 2.344050239806936e-06, "loss": 0.8228, "step": 13459 }, { "epoch": 1.6, "learning_rate": 2.343729443641177e-06, "loss": 0.5027, "step": 13460 }, { "epoch": 1.6, "learning_rate": 2.343408650058584e-06, "loss": 0.5812, "step": 13461 }, { "epoch": 1.6, "learning_rate": 2.343087859064459e-06, "loss": 0.5687, "step": 13462 }, { "epoch": 1.6, "learning_rate": 2.3427670706641056e-06, "loss": 0.6172, "step": 13463 }, { "epoch": 1.6, "learning_rate": 2.3424462848628246e-06, "loss": 0.5385, "step": 13464 }, { "epoch": 1.6, "learning_rate": 2.3421255016659204e-06, "loss": 0.8181, "step": 13465 }, { "epoch": 1.6, "learning_rate": 2.3418047210786943e-06, "loss": 0.5298, "step": 13466 }, { "epoch": 1.6, "learning_rate": 2.3414839431064503e-06, "loss": 0.5923, "step": 13467 }, { "epoch": 1.6, "learning_rate": 2.3411631677544896e-06, "loss": 0.5177, "step": 13468 }, { "epoch": 1.6, "learning_rate": 2.340842395028115e-06, "loss": 0.5338, "step": 13469 }, { "epoch": 1.6, "learning_rate": 2.3405216249326285e-06, "loss": 0.6146, "step": 13470 }, { "epoch": 1.6, "learning_rate": 2.340200857473334e-06, "loss": 0.7045, "step": 13471 }, { "epoch": 1.6, "learning_rate": 2.339880092655532e-06, "loss": 0.7873, "step": 13472 }, { "epoch": 1.6, "learning_rate": 2.3395593304845254e-06, "loss": 0.4064, "step": 13473 }, { "epoch": 1.6, "learning_rate": 2.3392385709656166e-06, "loss": 0.5213, "step": 13474 }, { "epoch": 1.6, "learning_rate": 2.3389178141041086e-06, "loss": 0.6066, "step": 13475 }, { "epoch": 1.6, "learning_rate": 2.338597059905302e-06, "loss": 0.7234, "step": 13476 }, { "epoch": 1.6, "learning_rate": 2.3382763083744993e-06, "loss": 0.8549, "step": 13477 }, { "epoch": 1.61, "learning_rate": 2.3379555595170032e-06, "loss": 0.8066, "step": 13478 }, { "epoch": 1.61, "learning_rate": 2.337634813338115e-06, "loss": 0.431, "step": 13479 }, { "epoch": 1.61, "learning_rate": 2.337314069843137e-06, "loss": 0.5274, "step": 13480 }, { "epoch": 1.61, "learning_rate": 2.3369933290373707e-06, "loss": 0.6356, "step": 13481 }, { "epoch": 1.61, "learning_rate": 2.3366725909261195e-06, "loss": 0.643, "step": 13482 }, { "epoch": 1.61, "learning_rate": 2.336351855514683e-06, "loss": 0.7372, "step": 13483 }, { "epoch": 1.61, "learning_rate": 2.3360311228083644e-06, "loss": 0.7703, "step": 13484 }, { "epoch": 1.61, "learning_rate": 2.3357103928124646e-06, "loss": 0.5444, "step": 13485 }, { "epoch": 1.61, "learning_rate": 2.335389665532287e-06, "loss": 0.5781, "step": 13486 }, { "epoch": 1.61, "learning_rate": 2.3350689409731315e-06, "loss": 0.5349, "step": 13487 }, { "epoch": 1.61, "learning_rate": 2.3347482191403e-06, "loss": 0.5644, "step": 13488 }, { "epoch": 1.61, "learning_rate": 2.334427500039094e-06, "loss": 0.6984, "step": 13489 }, { "epoch": 1.61, "learning_rate": 2.334106783674816e-06, "loss": 0.8751, "step": 13490 }, { "epoch": 1.61, "learning_rate": 2.3337860700527664e-06, "loss": 0.7324, "step": 13491 }, { "epoch": 1.61, "learning_rate": 2.333465359178247e-06, "loss": 0.4868, "step": 13492 }, { "epoch": 1.61, "learning_rate": 2.33314465105656e-06, "loss": 0.5119, "step": 13493 }, { "epoch": 1.61, "learning_rate": 2.3328239456930045e-06, "loss": 0.685, "step": 13494 }, { "epoch": 1.61, "learning_rate": 2.332503243092884e-06, "loss": 0.7708, "step": 13495 }, { "epoch": 1.61, "learning_rate": 2.3321825432614987e-06, "loss": 0.9019, "step": 13496 }, { "epoch": 1.61, "learning_rate": 2.3318618462041506e-06, "loss": 0.6511, "step": 13497 }, { "epoch": 1.61, "learning_rate": 2.33154115192614e-06, "loss": 0.5057, "step": 13498 }, { "epoch": 1.61, "learning_rate": 2.331220460432768e-06, "loss": 0.534, "step": 13499 }, { "epoch": 1.61, "learning_rate": 2.3308997717293357e-06, "loss": 0.4836, "step": 13500 }, { "epoch": 1.61, "learning_rate": 2.330579085821145e-06, "loss": 0.6208, "step": 13501 }, { "epoch": 1.61, "learning_rate": 2.3302584027134965e-06, "loss": 0.8251, "step": 13502 }, { "epoch": 1.61, "learning_rate": 2.32993772241169e-06, "loss": 0.7835, "step": 13503 }, { "epoch": 1.61, "learning_rate": 2.3296170449210277e-06, "loss": 0.3972, "step": 13504 }, { "epoch": 1.61, "learning_rate": 2.32929637024681e-06, "loss": 0.4964, "step": 13505 }, { "epoch": 1.61, "learning_rate": 2.3289756983943372e-06, "loss": 0.4797, "step": 13506 }, { "epoch": 1.61, "learning_rate": 2.3286550293689104e-06, "loss": 0.6007, "step": 13507 }, { "epoch": 1.61, "learning_rate": 2.3283343631758314e-06, "loss": 0.7263, "step": 13508 }, { "epoch": 1.61, "learning_rate": 2.328013699820399e-06, "loss": 0.8264, "step": 13509 }, { "epoch": 1.61, "learning_rate": 2.3276930393079143e-06, "loss": 0.5176, "step": 13510 }, { "epoch": 1.61, "learning_rate": 2.327372381643678e-06, "loss": 0.6119, "step": 13511 }, { "epoch": 1.61, "learning_rate": 2.327051726832992e-06, "loss": 0.6268, "step": 13512 }, { "epoch": 1.61, "learning_rate": 2.3267310748811544e-06, "loss": 0.6336, "step": 13513 }, { "epoch": 1.61, "learning_rate": 2.3264104257934665e-06, "loss": 0.5981, "step": 13514 }, { "epoch": 1.61, "learning_rate": 2.3260897795752293e-06, "loss": 0.8444, "step": 13515 }, { "epoch": 1.61, "learning_rate": 2.3257691362317432e-06, "loss": 0.5353, "step": 13516 }, { "epoch": 1.61, "learning_rate": 2.325448495768307e-06, "loss": 0.553, "step": 13517 }, { "epoch": 1.61, "learning_rate": 2.3251278581902216e-06, "loss": 0.4827, "step": 13518 }, { "epoch": 1.61, "learning_rate": 2.3248072235027875e-06, "loss": 0.5916, "step": 13519 }, { "epoch": 1.61, "learning_rate": 2.324486591711306e-06, "loss": 0.6718, "step": 13520 }, { "epoch": 1.61, "learning_rate": 2.3241659628210745e-06, "loss": 0.795, "step": 13521 }, { "epoch": 1.61, "learning_rate": 2.3238453368373948e-06, "loss": 0.7607, "step": 13522 }, { "epoch": 1.61, "learning_rate": 2.3235247137655666e-06, "loss": 0.4261, "step": 13523 }, { "epoch": 1.61, "learning_rate": 2.3232040936108894e-06, "loss": 0.6277, "step": 13524 }, { "epoch": 1.61, "learning_rate": 2.322883476378663e-06, "loss": 0.52, "step": 13525 }, { "epoch": 1.61, "learning_rate": 2.3225628620741884e-06, "loss": 0.6742, "step": 13526 }, { "epoch": 1.61, "learning_rate": 2.322242250702765e-06, "loss": 0.659, "step": 13527 }, { "epoch": 1.61, "learning_rate": 2.3219216422696907e-06, "loss": 0.8242, "step": 13528 }, { "epoch": 1.61, "learning_rate": 2.321601036780267e-06, "loss": 0.3337, "step": 13529 }, { "epoch": 1.61, "learning_rate": 2.3212804342397934e-06, "loss": 0.5095, "step": 13530 }, { "epoch": 1.61, "learning_rate": 2.32095983465357e-06, "loss": 0.6222, "step": 13531 }, { "epoch": 1.61, "learning_rate": 2.3206392380268942e-06, "loss": 0.5897, "step": 13532 }, { "epoch": 1.61, "learning_rate": 2.3203186443650677e-06, "loss": 0.6808, "step": 13533 }, { "epoch": 1.61, "learning_rate": 2.3199980536733885e-06, "loss": 0.8195, "step": 13534 }, { "epoch": 1.61, "learning_rate": 2.3196774659571573e-06, "loss": 0.4367, "step": 13535 }, { "epoch": 1.61, "learning_rate": 2.3193568812216723e-06, "loss": 0.5793, "step": 13536 }, { "epoch": 1.61, "learning_rate": 2.319036299472233e-06, "loss": 0.5929, "step": 13537 }, { "epoch": 1.61, "learning_rate": 2.3187157207141396e-06, "loss": 0.6581, "step": 13538 }, { "epoch": 1.61, "learning_rate": 2.31839514495269e-06, "loss": 0.7043, "step": 13539 }, { "epoch": 1.61, "learning_rate": 2.318074572193183e-06, "loss": 0.9634, "step": 13540 }, { "epoch": 1.61, "learning_rate": 2.31775400244092e-06, "loss": 0.5587, "step": 13541 }, { "epoch": 1.61, "learning_rate": 2.3174334357011983e-06, "loss": 0.5146, "step": 13542 }, { "epoch": 1.61, "learning_rate": 2.3171128719793166e-06, "loss": 0.5213, "step": 13543 }, { "epoch": 1.61, "learning_rate": 2.3167923112805756e-06, "loss": 0.59, "step": 13544 }, { "epoch": 1.61, "learning_rate": 2.3164717536102717e-06, "loss": 0.5721, "step": 13545 }, { "epoch": 1.61, "learning_rate": 2.3161511989737065e-06, "loss": 0.8373, "step": 13546 }, { "epoch": 1.61, "learning_rate": 2.3158306473761765e-06, "loss": 0.7776, "step": 13547 }, { "epoch": 1.61, "learning_rate": 2.3155100988229816e-06, "loss": 0.4729, "step": 13548 }, { "epoch": 1.61, "learning_rate": 2.3151895533194195e-06, "loss": 0.5736, "step": 13549 }, { "epoch": 1.61, "learning_rate": 2.314869010870791e-06, "loss": 0.62, "step": 13550 }, { "epoch": 1.61, "learning_rate": 2.3145484714823922e-06, "loss": 0.6173, "step": 13551 }, { "epoch": 1.61, "learning_rate": 2.314227935159523e-06, "loss": 0.7787, "step": 13552 }, { "epoch": 1.61, "learning_rate": 2.3139074019074815e-06, "loss": 0.8051, "step": 13553 }, { "epoch": 1.61, "learning_rate": 2.313586871731567e-06, "loss": 0.3683, "step": 13554 }, { "epoch": 1.61, "learning_rate": 2.3132663446370766e-06, "loss": 0.4605, "step": 13555 }, { "epoch": 1.61, "learning_rate": 2.312945820629309e-06, "loss": 0.5598, "step": 13556 }, { "epoch": 1.61, "learning_rate": 2.3126252997135635e-06, "loss": 0.6899, "step": 13557 }, { "epoch": 1.61, "learning_rate": 2.3123047818951365e-06, "loss": 0.7512, "step": 13558 }, { "epoch": 1.61, "learning_rate": 2.311984267179328e-06, "loss": 0.7882, "step": 13559 }, { "epoch": 1.61, "learning_rate": 2.3116637555714344e-06, "loss": 0.7089, "step": 13560 }, { "epoch": 1.61, "learning_rate": 2.311343247076756e-06, "loss": 0.521, "step": 13561 }, { "epoch": 1.62, "learning_rate": 2.3110227417005888e-06, "loss": 0.653, "step": 13562 }, { "epoch": 1.62, "learning_rate": 2.310702239448231e-06, "loss": 0.6559, "step": 13563 }, { "epoch": 1.62, "learning_rate": 2.3103817403249813e-06, "loss": 0.7247, "step": 13564 }, { "epoch": 1.62, "learning_rate": 2.310061244336139e-06, "loss": 0.8469, "step": 13565 }, { "epoch": 1.62, "learning_rate": 2.3097407514869984e-06, "loss": 0.368, "step": 13566 }, { "epoch": 1.62, "learning_rate": 2.3094202617828596e-06, "loss": 0.3956, "step": 13567 }, { "epoch": 1.62, "learning_rate": 2.3090997752290197e-06, "loss": 0.5969, "step": 13568 }, { "epoch": 1.62, "learning_rate": 2.3087792918307773e-06, "loss": 0.5919, "step": 13569 }, { "epoch": 1.62, "learning_rate": 2.3084588115934285e-06, "loss": 0.7498, "step": 13570 }, { "epoch": 1.62, "learning_rate": 2.3081383345222715e-06, "loss": 0.8017, "step": 13571 }, { "epoch": 1.62, "learning_rate": 2.307817860622605e-06, "loss": 0.8737, "step": 13572 }, { "epoch": 1.62, "learning_rate": 2.307497389899725e-06, "loss": 0.4276, "step": 13573 }, { "epoch": 1.62, "learning_rate": 2.3071769223589286e-06, "loss": 0.5608, "step": 13574 }, { "epoch": 1.62, "learning_rate": 2.3068564580055135e-06, "loss": 0.5812, "step": 13575 }, { "epoch": 1.62, "learning_rate": 2.306535996844779e-06, "loss": 0.6261, "step": 13576 }, { "epoch": 1.62, "learning_rate": 2.30621553888202e-06, "loss": 0.7422, "step": 13577 }, { "epoch": 1.62, "learning_rate": 2.3058950841225337e-06, "loss": 0.8643, "step": 13578 }, { "epoch": 1.62, "learning_rate": 2.3055746325716184e-06, "loss": 0.4374, "step": 13579 }, { "epoch": 1.62, "learning_rate": 2.305254184234571e-06, "loss": 0.5712, "step": 13580 }, { "epoch": 1.62, "learning_rate": 2.3049337391166884e-06, "loss": 0.5032, "step": 13581 }, { "epoch": 1.62, "learning_rate": 2.3046132972232666e-06, "loss": 0.6498, "step": 13582 }, { "epoch": 1.62, "learning_rate": 2.304292858559604e-06, "loss": 0.6609, "step": 13583 }, { "epoch": 1.62, "learning_rate": 2.3039724231309974e-06, "loss": 0.7925, "step": 13584 }, { "epoch": 1.62, "learning_rate": 2.303651990942742e-06, "loss": 0.5096, "step": 13585 }, { "epoch": 1.62, "learning_rate": 2.3033315620001355e-06, "loss": 0.5682, "step": 13586 }, { "epoch": 1.62, "learning_rate": 2.3030111363084762e-06, "loss": 0.7248, "step": 13587 }, { "epoch": 1.62, "learning_rate": 2.3026907138730585e-06, "loss": 0.6753, "step": 13588 }, { "epoch": 1.62, "learning_rate": 2.3023702946991797e-06, "loss": 0.7667, "step": 13589 }, { "epoch": 1.62, "learning_rate": 2.3020498787921368e-06, "loss": 0.9916, "step": 13590 }, { "epoch": 1.62, "learning_rate": 2.3017294661572265e-06, "loss": 0.5097, "step": 13591 }, { "epoch": 1.62, "learning_rate": 2.301409056799744e-06, "loss": 0.4373, "step": 13592 }, { "epoch": 1.62, "learning_rate": 2.301088650724986e-06, "loss": 0.6412, "step": 13593 }, { "epoch": 1.62, "learning_rate": 2.30076824793825e-06, "loss": 0.6677, "step": 13594 }, { "epoch": 1.62, "learning_rate": 2.3004478484448324e-06, "loss": 0.7169, "step": 13595 }, { "epoch": 1.62, "learning_rate": 2.300127452250027e-06, "loss": 0.808, "step": 13596 }, { "epoch": 1.62, "learning_rate": 2.299807059359132e-06, "loss": 0.7727, "step": 13597 }, { "epoch": 1.62, "learning_rate": 2.2994866697774437e-06, "loss": 0.4085, "step": 13598 }, { "epoch": 1.62, "learning_rate": 2.299166283510258e-06, "loss": 0.5707, "step": 13599 }, { "epoch": 1.62, "learning_rate": 2.2988459005628695e-06, "loss": 0.5357, "step": 13600 }, { "epoch": 1.62, "learning_rate": 2.2985255209405754e-06, "loss": 0.6799, "step": 13601 }, { "epoch": 1.62, "learning_rate": 2.298205144648672e-06, "loss": 0.6529, "step": 13602 }, { "epoch": 1.62, "learning_rate": 2.2978847716924536e-06, "loss": 0.7429, "step": 13603 }, { "epoch": 1.62, "learning_rate": 2.2975644020772164e-06, "loss": 0.4354, "step": 13604 }, { "epoch": 1.62, "learning_rate": 2.2972440358082574e-06, "loss": 0.4868, "step": 13605 }, { "epoch": 1.62, "learning_rate": 2.2969236728908723e-06, "loss": 0.5241, "step": 13606 }, { "epoch": 1.62, "learning_rate": 2.296603313330355e-06, "loss": 0.6329, "step": 13607 }, { "epoch": 1.62, "learning_rate": 2.296282957132002e-06, "loss": 0.6954, "step": 13608 }, { "epoch": 1.62, "learning_rate": 2.2959626043011084e-06, "loss": 0.7806, "step": 13609 }, { "epoch": 1.62, "learning_rate": 2.2956422548429713e-06, "loss": 0.5891, "step": 13610 }, { "epoch": 1.62, "learning_rate": 2.2953219087628837e-06, "loss": 0.4331, "step": 13611 }, { "epoch": 1.62, "learning_rate": 2.2950015660661427e-06, "loss": 0.5045, "step": 13612 }, { "epoch": 1.62, "learning_rate": 2.2946812267580425e-06, "loss": 0.6829, "step": 13613 }, { "epoch": 1.62, "learning_rate": 2.2943608908438797e-06, "loss": 0.6704, "step": 13614 }, { "epoch": 1.62, "learning_rate": 2.2940405583289477e-06, "loss": 0.9143, "step": 13615 }, { "epoch": 1.62, "learning_rate": 2.293720229218543e-06, "loss": 0.4951, "step": 13616 }, { "epoch": 1.62, "learning_rate": 2.29339990351796e-06, "loss": 0.441, "step": 13617 }, { "epoch": 1.62, "learning_rate": 2.293079581232495e-06, "loss": 0.6314, "step": 13618 }, { "epoch": 1.62, "learning_rate": 2.292759262367441e-06, "loss": 0.5185, "step": 13619 }, { "epoch": 1.62, "learning_rate": 2.2924389469280935e-06, "loss": 0.6848, "step": 13620 }, { "epoch": 1.62, "learning_rate": 2.2921186349197487e-06, "loss": 0.8282, "step": 13621 }, { "epoch": 1.62, "learning_rate": 2.291798326347699e-06, "loss": 0.9265, "step": 13622 }, { "epoch": 1.62, "learning_rate": 2.291478021217241e-06, "loss": 0.4402, "step": 13623 }, { "epoch": 1.62, "learning_rate": 2.2911577195336683e-06, "loss": 0.5186, "step": 13624 }, { "epoch": 1.62, "learning_rate": 2.2908374213022775e-06, "loss": 0.6336, "step": 13625 }, { "epoch": 1.62, "learning_rate": 2.29051712652836e-06, "loss": 0.5866, "step": 13626 }, { "epoch": 1.62, "learning_rate": 2.290196835217213e-06, "loss": 0.6412, "step": 13627 }, { "epoch": 1.62, "learning_rate": 2.289876547374129e-06, "loss": 0.8635, "step": 13628 }, { "epoch": 1.62, "learning_rate": 2.2895562630044045e-06, "loss": 0.4732, "step": 13629 }, { "epoch": 1.62, "learning_rate": 2.289235982113332e-06, "loss": 0.4665, "step": 13630 }, { "epoch": 1.62, "learning_rate": 2.2889157047062065e-06, "loss": 0.6257, "step": 13631 }, { "epoch": 1.62, "learning_rate": 2.2885954307883214e-06, "loss": 0.6855, "step": 13632 }, { "epoch": 1.62, "learning_rate": 2.288275160364973e-06, "loss": 0.7512, "step": 13633 }, { "epoch": 1.62, "learning_rate": 2.287954893441453e-06, "loss": 0.8293, "step": 13634 }, { "epoch": 1.62, "learning_rate": 2.287634630023056e-06, "loss": 0.6667, "step": 13635 }, { "epoch": 1.62, "learning_rate": 2.287314370115078e-06, "loss": 0.4755, "step": 13636 }, { "epoch": 1.62, "learning_rate": 2.2869941137228096e-06, "loss": 0.5652, "step": 13637 }, { "epoch": 1.62, "learning_rate": 2.286673860851547e-06, "loss": 0.6448, "step": 13638 }, { "epoch": 1.62, "learning_rate": 2.286353611506583e-06, "loss": 0.7524, "step": 13639 }, { "epoch": 1.62, "learning_rate": 2.286033365693213e-06, "loss": 0.7812, "step": 13640 }, { "epoch": 1.62, "learning_rate": 2.2857131234167286e-06, "loss": 0.5312, "step": 13641 }, { "epoch": 1.62, "learning_rate": 2.2853928846824237e-06, "loss": 0.497, "step": 13642 }, { "epoch": 1.62, "learning_rate": 2.2850726494955928e-06, "loss": 0.6179, "step": 13643 }, { "epoch": 1.62, "learning_rate": 2.2847524178615297e-06, "loss": 0.5098, "step": 13644 }, { "epoch": 1.62, "learning_rate": 2.284432189785527e-06, "loss": 0.7043, "step": 13645 }, { "epoch": 1.63, "learning_rate": 2.284111965272877e-06, "loss": 0.817, "step": 13646 }, { "epoch": 1.63, "learning_rate": 2.2837917443288755e-06, "loss": 0.7794, "step": 13647 }, { "epoch": 1.63, "learning_rate": 2.2834715269588155e-06, "loss": 0.433, "step": 13648 }, { "epoch": 1.63, "learning_rate": 2.2831513131679873e-06, "loss": 0.5413, "step": 13649 }, { "epoch": 1.63, "learning_rate": 2.282831102961687e-06, "loss": 0.5559, "step": 13650 }, { "epoch": 1.63, "learning_rate": 2.2825108963452073e-06, "loss": 0.7303, "step": 13651 }, { "epoch": 1.63, "learning_rate": 2.28219069332384e-06, "loss": 0.9338, "step": 13652 }, { "epoch": 1.63, "learning_rate": 2.2818704939028792e-06, "loss": 0.8109, "step": 13653 }, { "epoch": 1.63, "learning_rate": 2.281550298087617e-06, "loss": 0.4521, "step": 13654 }, { "epoch": 1.63, "learning_rate": 2.281230105883347e-06, "loss": 0.5562, "step": 13655 }, { "epoch": 1.63, "learning_rate": 2.280909917295362e-06, "loss": 0.6183, "step": 13656 }, { "epoch": 1.63, "learning_rate": 2.280589732328954e-06, "loss": 0.605, "step": 13657 }, { "epoch": 1.63, "learning_rate": 2.280269550989416e-06, "loss": 0.7112, "step": 13658 }, { "epoch": 1.63, "learning_rate": 2.279949373282042e-06, "loss": 0.8126, "step": 13659 }, { "epoch": 1.63, "learning_rate": 2.2796291992121215e-06, "loss": 0.4243, "step": 13660 }, { "epoch": 1.63, "learning_rate": 2.2793090287849497e-06, "loss": 0.5265, "step": 13661 }, { "epoch": 1.63, "learning_rate": 2.278988862005818e-06, "loss": 0.5158, "step": 13662 }, { "epoch": 1.63, "learning_rate": 2.27866869888002e-06, "loss": 0.6118, "step": 13663 }, { "epoch": 1.63, "learning_rate": 2.2783485394128454e-06, "loss": 0.6266, "step": 13664 }, { "epoch": 1.63, "learning_rate": 2.278028383609589e-06, "loss": 0.6821, "step": 13665 }, { "epoch": 1.63, "learning_rate": 2.2777082314755418e-06, "loss": 0.5653, "step": 13666 }, { "epoch": 1.63, "learning_rate": 2.2773880830159963e-06, "loss": 0.5821, "step": 13667 }, { "epoch": 1.63, "learning_rate": 2.277067938236244e-06, "loss": 0.6201, "step": 13668 }, { "epoch": 1.63, "learning_rate": 2.2767477971415775e-06, "loss": 0.5421, "step": 13669 }, { "epoch": 1.63, "learning_rate": 2.2764276597372897e-06, "loss": 0.8326, "step": 13670 }, { "epoch": 1.63, "learning_rate": 2.27610752602867e-06, "loss": 0.946, "step": 13671 }, { "epoch": 1.63, "learning_rate": 2.2757873960210122e-06, "loss": 0.7808, "step": 13672 }, { "epoch": 1.63, "learning_rate": 2.2754672697196077e-06, "loss": 0.434, "step": 13673 }, { "epoch": 1.63, "learning_rate": 2.2751471471297486e-06, "loss": 0.5411, "step": 13674 }, { "epoch": 1.63, "learning_rate": 2.274827028256725e-06, "loss": 0.4619, "step": 13675 }, { "epoch": 1.63, "learning_rate": 2.27450691310583e-06, "loss": 0.8212, "step": 13676 }, { "epoch": 1.63, "learning_rate": 2.274186801682354e-06, "loss": 0.7369, "step": 13677 }, { "epoch": 1.63, "learning_rate": 2.2738666939915907e-06, "loss": 0.8423, "step": 13678 }, { "epoch": 1.63, "learning_rate": 2.2735465900388286e-06, "loss": 0.7096, "step": 13679 }, { "epoch": 1.63, "learning_rate": 2.273226489829361e-06, "loss": 0.5914, "step": 13680 }, { "epoch": 1.63, "learning_rate": 2.2729063933684785e-06, "loss": 0.5342, "step": 13681 }, { "epoch": 1.63, "learning_rate": 2.2725863006614717e-06, "loss": 0.6472, "step": 13682 }, { "epoch": 1.63, "learning_rate": 2.2722662117136326e-06, "loss": 0.6912, "step": 13683 }, { "epoch": 1.63, "learning_rate": 2.2719461265302524e-06, "loss": 0.8698, "step": 13684 }, { "epoch": 1.63, "learning_rate": 2.271626045116623e-06, "loss": 0.469, "step": 13685 }, { "epoch": 1.63, "learning_rate": 2.2713059674780323e-06, "loss": 0.5157, "step": 13686 }, { "epoch": 1.63, "learning_rate": 2.270985893619774e-06, "loss": 0.4969, "step": 13687 }, { "epoch": 1.63, "learning_rate": 2.2706658235471375e-06, "loss": 0.6242, "step": 13688 }, { "epoch": 1.63, "learning_rate": 2.2703457572654156e-06, "loss": 0.7048, "step": 13689 }, { "epoch": 1.63, "learning_rate": 2.2700256947798964e-06, "loss": 0.8049, "step": 13690 }, { "epoch": 1.63, "learning_rate": 2.269705636095872e-06, "loss": 0.5393, "step": 13691 }, { "epoch": 1.63, "learning_rate": 2.2693855812186323e-06, "loss": 0.5995, "step": 13692 }, { "epoch": 1.63, "learning_rate": 2.2690655301534698e-06, "loss": 0.6107, "step": 13693 }, { "epoch": 1.63, "learning_rate": 2.2687454829056718e-06, "loss": 0.5904, "step": 13694 }, { "epoch": 1.63, "learning_rate": 2.2684254394805313e-06, "loss": 0.7246, "step": 13695 }, { "epoch": 1.63, "learning_rate": 2.268105399883337e-06, "loss": 0.9045, "step": 13696 }, { "epoch": 1.63, "learning_rate": 2.2677853641193812e-06, "loss": 0.7248, "step": 13697 }, { "epoch": 1.63, "learning_rate": 2.267465332193952e-06, "loss": 0.5025, "step": 13698 }, { "epoch": 1.63, "learning_rate": 2.2671453041123397e-06, "loss": 0.5337, "step": 13699 }, { "epoch": 1.63, "learning_rate": 2.2668252798798364e-06, "loss": 0.5457, "step": 13700 }, { "epoch": 1.63, "learning_rate": 2.2665052595017297e-06, "loss": 0.6358, "step": 13701 }, { "epoch": 1.63, "learning_rate": 2.2661852429833113e-06, "loss": 0.7565, "step": 13702 }, { "epoch": 1.63, "learning_rate": 2.26586523032987e-06, "loss": 0.722, "step": 13703 }, { "epoch": 1.63, "learning_rate": 2.265545221546697e-06, "loss": 0.4271, "step": 13704 }, { "epoch": 1.63, "learning_rate": 2.2652252166390803e-06, "loss": 0.5974, "step": 13705 }, { "epoch": 1.63, "learning_rate": 2.2649052156123105e-06, "loss": 0.555, "step": 13706 }, { "epoch": 1.63, "learning_rate": 2.2645852184716765e-06, "loss": 0.5915, "step": 13707 }, { "epoch": 1.63, "learning_rate": 2.2642652252224703e-06, "loss": 0.7781, "step": 13708 }, { "epoch": 1.63, "learning_rate": 2.263945235869979e-06, "loss": 0.8103, "step": 13709 }, { "epoch": 1.63, "learning_rate": 2.2636252504194924e-06, "loss": 0.4391, "step": 13710 }, { "epoch": 1.63, "learning_rate": 2.2633052688762997e-06, "loss": 0.527, "step": 13711 }, { "epoch": 1.63, "learning_rate": 2.2629852912456924e-06, "loss": 0.545, "step": 13712 }, { "epoch": 1.63, "learning_rate": 2.262665317532957e-06, "loss": 0.6938, "step": 13713 }, { "epoch": 1.63, "learning_rate": 2.262345347743384e-06, "loss": 0.6476, "step": 13714 }, { "epoch": 1.63, "learning_rate": 2.262025381882263e-06, "loss": 0.8627, "step": 13715 }, { "epoch": 1.63, "learning_rate": 2.261705419954882e-06, "loss": 0.5995, "step": 13716 }, { "epoch": 1.63, "learning_rate": 2.2613854619665296e-06, "loss": 0.5181, "step": 13717 }, { "epoch": 1.63, "learning_rate": 2.2610655079224957e-06, "loss": 0.592, "step": 13718 }, { "epoch": 1.63, "learning_rate": 2.26074555782807e-06, "loss": 0.5686, "step": 13719 }, { "epoch": 1.63, "learning_rate": 2.2604256116885398e-06, "loss": 0.6373, "step": 13720 }, { "epoch": 1.63, "learning_rate": 2.2601056695091937e-06, "loss": 0.9478, "step": 13721 }, { "epoch": 1.63, "learning_rate": 2.259785731295322e-06, "loss": 0.6991, "step": 13722 }, { "epoch": 1.63, "learning_rate": 2.259465797052212e-06, "loss": 0.4566, "step": 13723 }, { "epoch": 1.63, "learning_rate": 2.2591458667851528e-06, "loss": 0.5122, "step": 13724 }, { "epoch": 1.63, "learning_rate": 2.2588259404994317e-06, "loss": 0.6349, "step": 13725 }, { "epoch": 1.63, "learning_rate": 2.2585060182003386e-06, "loss": 0.5828, "step": 13726 }, { "epoch": 1.63, "learning_rate": 2.2581860998931623e-06, "loss": 0.843, "step": 13727 }, { "epoch": 1.63, "learning_rate": 2.2578661855831886e-06, "loss": 0.8844, "step": 13728 }, { "epoch": 1.63, "learning_rate": 2.257546275275707e-06, "loss": 0.5327, "step": 13729 }, { "epoch": 1.64, "learning_rate": 2.257226368976007e-06, "loss": 0.4789, "step": 13730 }, { "epoch": 1.64, "learning_rate": 2.2569064666893754e-06, "loss": 0.553, "step": 13731 }, { "epoch": 1.64, "learning_rate": 2.256586568421099e-06, "loss": 0.7226, "step": 13732 }, { "epoch": 1.64, "learning_rate": 2.2562666741764682e-06, "loss": 0.7726, "step": 13733 }, { "epoch": 1.64, "learning_rate": 2.2559467839607703e-06, "loss": 0.9047, "step": 13734 }, { "epoch": 1.64, "learning_rate": 2.255626897779291e-06, "loss": 0.5656, "step": 13735 }, { "epoch": 1.64, "learning_rate": 2.25530701563732e-06, "loss": 0.5373, "step": 13736 }, { "epoch": 1.64, "learning_rate": 2.254987137540145e-06, "loss": 0.5158, "step": 13737 }, { "epoch": 1.64, "learning_rate": 2.2546672634930538e-06, "loss": 0.6531, "step": 13738 }, { "epoch": 1.64, "learning_rate": 2.2543473935013317e-06, "loss": 0.7216, "step": 13739 }, { "epoch": 1.64, "learning_rate": 2.2540275275702683e-06, "loss": 0.7652, "step": 13740 }, { "epoch": 1.64, "learning_rate": 2.253707665705151e-06, "loss": 0.6327, "step": 13741 }, { "epoch": 1.64, "learning_rate": 2.253387807911267e-06, "loss": 0.5309, "step": 13742 }, { "epoch": 1.64, "learning_rate": 2.2530679541939024e-06, "loss": 0.5912, "step": 13743 }, { "epoch": 1.64, "learning_rate": 2.2527481045583455e-06, "loss": 0.4993, "step": 13744 }, { "epoch": 1.64, "learning_rate": 2.252428259009884e-06, "loss": 0.5506, "step": 13745 }, { "epoch": 1.64, "learning_rate": 2.2521084175538028e-06, "loss": 0.7712, "step": 13746 }, { "epoch": 1.64, "learning_rate": 2.2517885801953904e-06, "loss": 0.7523, "step": 13747 }, { "epoch": 1.64, "learning_rate": 2.251468746939934e-06, "loss": 0.4828, "step": 13748 }, { "epoch": 1.64, "learning_rate": 2.251148917792721e-06, "loss": 0.7036, "step": 13749 }, { "epoch": 1.64, "learning_rate": 2.2508290927590357e-06, "loss": 0.5578, "step": 13750 }, { "epoch": 1.64, "learning_rate": 2.2505092718441667e-06, "loss": 0.6626, "step": 13751 }, { "epoch": 1.64, "learning_rate": 2.250189455053401e-06, "loss": 0.7689, "step": 13752 }, { "epoch": 1.64, "learning_rate": 2.249869642392025e-06, "loss": 0.8307, "step": 13753 }, { "epoch": 1.64, "learning_rate": 2.2495498338653234e-06, "loss": 0.4905, "step": 13754 }, { "epoch": 1.64, "learning_rate": 2.249230029478585e-06, "loss": 0.4996, "step": 13755 }, { "epoch": 1.64, "learning_rate": 2.2489102292370946e-06, "loss": 0.6311, "step": 13756 }, { "epoch": 1.64, "learning_rate": 2.24859043314614e-06, "loss": 0.5959, "step": 13757 }, { "epoch": 1.64, "learning_rate": 2.2482706412110063e-06, "loss": 0.7539, "step": 13758 }, { "epoch": 1.64, "learning_rate": 2.24795085343698e-06, "loss": 0.8086, "step": 13759 }, { "epoch": 1.64, "learning_rate": 2.2476310698293467e-06, "loss": 0.5413, "step": 13760 }, { "epoch": 1.64, "learning_rate": 2.247311290393394e-06, "loss": 0.5162, "step": 13761 }, { "epoch": 1.64, "learning_rate": 2.2469915151344064e-06, "loss": 0.6093, "step": 13762 }, { "epoch": 1.64, "learning_rate": 2.24667174405767e-06, "loss": 0.5239, "step": 13763 }, { "epoch": 1.64, "learning_rate": 2.2463519771684714e-06, "loss": 0.5803, "step": 13764 }, { "epoch": 1.64, "learning_rate": 2.2460322144720954e-06, "loss": 0.8393, "step": 13765 }, { "epoch": 1.64, "learning_rate": 2.2457124559738286e-06, "loss": 0.5387, "step": 13766 }, { "epoch": 1.64, "learning_rate": 2.245392701678956e-06, "loss": 0.4875, "step": 13767 }, { "epoch": 1.64, "learning_rate": 2.245072951592764e-06, "loss": 0.5679, "step": 13768 }, { "epoch": 1.64, "learning_rate": 2.2447532057205367e-06, "loss": 0.6664, "step": 13769 }, { "epoch": 1.64, "learning_rate": 2.2444334640675605e-06, "loss": 0.654, "step": 13770 }, { "epoch": 1.64, "learning_rate": 2.24411372663912e-06, "loss": 0.8777, "step": 13771 }, { "epoch": 1.64, "learning_rate": 2.243793993440503e-06, "loss": 0.9497, "step": 13772 }, { "epoch": 1.64, "learning_rate": 2.243474264476992e-06, "loss": 0.4081, "step": 13773 }, { "epoch": 1.64, "learning_rate": 2.2431545397538715e-06, "loss": 0.4919, "step": 13774 }, { "epoch": 1.64, "learning_rate": 2.242834819276429e-06, "loss": 0.5266, "step": 13775 }, { "epoch": 1.64, "learning_rate": 2.242515103049949e-06, "loss": 0.6566, "step": 13776 }, { "epoch": 1.64, "learning_rate": 2.2421953910797158e-06, "loss": 0.8538, "step": 13777 }, { "epoch": 1.64, "learning_rate": 2.241875683371013e-06, "loss": 0.9607, "step": 13778 }, { "epoch": 1.64, "learning_rate": 2.241555979929129e-06, "loss": 0.3858, "step": 13779 }, { "epoch": 1.64, "learning_rate": 2.2412362807593444e-06, "loss": 0.5543, "step": 13780 }, { "epoch": 1.64, "learning_rate": 2.240916585866947e-06, "loss": 0.5432, "step": 13781 }, { "epoch": 1.64, "learning_rate": 2.2405968952572194e-06, "loss": 0.6906, "step": 13782 }, { "epoch": 1.64, "learning_rate": 2.240277208935448e-06, "loss": 0.7936, "step": 13783 }, { "epoch": 1.64, "learning_rate": 2.2399575269069157e-06, "loss": 0.7872, "step": 13784 }, { "epoch": 1.64, "learning_rate": 2.2396378491769063e-06, "loss": 0.5334, "step": 13785 }, { "epoch": 1.64, "learning_rate": 2.2393181757507053e-06, "loss": 0.5447, "step": 13786 }, { "epoch": 1.64, "learning_rate": 2.238998506633598e-06, "loss": 0.6386, "step": 13787 }, { "epoch": 1.64, "learning_rate": 2.238678841830867e-06, "loss": 0.5867, "step": 13788 }, { "epoch": 1.64, "learning_rate": 2.2383591813477956e-06, "loss": 0.7041, "step": 13789 }, { "epoch": 1.64, "learning_rate": 2.23803952518967e-06, "loss": 0.8582, "step": 13790 }, { "epoch": 1.64, "learning_rate": 2.2377198733617726e-06, "loss": 0.5653, "step": 13791 }, { "epoch": 1.64, "learning_rate": 2.2374002258693877e-06, "loss": 0.6086, "step": 13792 }, { "epoch": 1.64, "learning_rate": 2.2370805827177984e-06, "loss": 0.7065, "step": 13793 }, { "epoch": 1.64, "learning_rate": 2.2367609439122903e-06, "loss": 0.502, "step": 13794 }, { "epoch": 1.64, "learning_rate": 2.2364413094581454e-06, "loss": 0.6687, "step": 13795 }, { "epoch": 1.64, "learning_rate": 2.2361216793606476e-06, "loss": 0.7526, "step": 13796 }, { "epoch": 1.64, "learning_rate": 2.2358020536250802e-06, "loss": 0.7626, "step": 13797 }, { "epoch": 1.64, "learning_rate": 2.2354824322567283e-06, "loss": 0.3677, "step": 13798 }, { "epoch": 1.64, "learning_rate": 2.235162815260873e-06, "loss": 0.4984, "step": 13799 }, { "epoch": 1.64, "learning_rate": 2.2348432026427985e-06, "loss": 0.5298, "step": 13800 }, { "epoch": 1.64, "learning_rate": 2.2345235944077885e-06, "loss": 0.6299, "step": 13801 }, { "epoch": 1.64, "learning_rate": 2.2342039905611263e-06, "loss": 0.7504, "step": 13802 }, { "epoch": 1.64, "learning_rate": 2.233884391108093e-06, "loss": 0.8006, "step": 13803 }, { "epoch": 1.64, "learning_rate": 2.2335647960539737e-06, "loss": 0.4191, "step": 13804 }, { "epoch": 1.64, "learning_rate": 2.233245205404051e-06, "loss": 0.5058, "step": 13805 }, { "epoch": 1.64, "learning_rate": 2.232925619163608e-06, "loss": 0.692, "step": 13806 }, { "epoch": 1.64, "learning_rate": 2.2326060373379253e-06, "loss": 0.5955, "step": 13807 }, { "epoch": 1.64, "learning_rate": 2.232286459932288e-06, "loss": 0.6818, "step": 13808 }, { "epoch": 1.64, "learning_rate": 2.2319668869519786e-06, "loss": 0.9026, "step": 13809 }, { "epoch": 1.64, "learning_rate": 2.2316473184022787e-06, "loss": 0.5214, "step": 13810 }, { "epoch": 1.64, "learning_rate": 2.231327754288471e-06, "loss": 0.4885, "step": 13811 }, { "epoch": 1.64, "learning_rate": 2.231008194615838e-06, "loss": 0.5344, "step": 13812 }, { "epoch": 1.64, "learning_rate": 2.2306886393896627e-06, "loss": 0.6285, "step": 13813 }, { "epoch": 1.65, "learning_rate": 2.230369088615226e-06, "loss": 0.6519, "step": 13814 }, { "epoch": 1.65, "learning_rate": 2.2300495422978106e-06, "loss": 0.8595, "step": 13815 }, { "epoch": 1.65, "learning_rate": 2.2297300004426996e-06, "loss": 0.5081, "step": 13816 }, { "epoch": 1.65, "learning_rate": 2.229410463055175e-06, "loss": 0.4012, "step": 13817 }, { "epoch": 1.65, "learning_rate": 2.2290909301405167e-06, "loss": 0.5091, "step": 13818 }, { "epoch": 1.65, "learning_rate": 2.228771401704009e-06, "loss": 0.5893, "step": 13819 }, { "epoch": 1.65, "learning_rate": 2.228451877750932e-06, "loss": 0.6461, "step": 13820 }, { "epoch": 1.65, "learning_rate": 2.228132358286569e-06, "loss": 0.7145, "step": 13821 }, { "epoch": 1.65, "learning_rate": 2.227812843316201e-06, "loss": 0.8143, "step": 13822 }, { "epoch": 1.65, "learning_rate": 2.227493332845109e-06, "loss": 0.5069, "step": 13823 }, { "epoch": 1.65, "learning_rate": 2.227173826878575e-06, "loss": 0.5587, "step": 13824 }, { "epoch": 1.65, "learning_rate": 2.2268543254218813e-06, "loss": 0.6007, "step": 13825 }, { "epoch": 1.65, "learning_rate": 2.226534828480308e-06, "loss": 0.6602, "step": 13826 }, { "epoch": 1.65, "learning_rate": 2.226215336059137e-06, "loss": 0.7535, "step": 13827 }, { "epoch": 1.65, "learning_rate": 2.22589584816365e-06, "loss": 0.7576, "step": 13828 }, { "epoch": 1.65, "learning_rate": 2.2255763647991266e-06, "loss": 0.4951, "step": 13829 }, { "epoch": 1.65, "learning_rate": 2.2252568859708496e-06, "loss": 0.457, "step": 13830 }, { "epoch": 1.65, "learning_rate": 2.224937411684099e-06, "loss": 0.6764, "step": 13831 }, { "epoch": 1.65, "learning_rate": 2.2246179419441567e-06, "loss": 0.6033, "step": 13832 }, { "epoch": 1.65, "learning_rate": 2.2242984767563018e-06, "loss": 0.6898, "step": 13833 }, { "epoch": 1.65, "learning_rate": 2.223979016125817e-06, "loss": 0.8516, "step": 13834 }, { "epoch": 1.65, "learning_rate": 2.2236595600579818e-06, "loss": 0.3215, "step": 13835 }, { "epoch": 1.65, "learning_rate": 2.223340108558078e-06, "loss": 0.4336, "step": 13836 }, { "epoch": 1.65, "learning_rate": 2.2230206616313847e-06, "loss": 0.5685, "step": 13837 }, { "epoch": 1.65, "learning_rate": 2.2227012192831832e-06, "loss": 0.654, "step": 13838 }, { "epoch": 1.65, "learning_rate": 2.222381781518754e-06, "loss": 0.7579, "step": 13839 }, { "epoch": 1.65, "learning_rate": 2.2220623483433775e-06, "loss": 0.8801, "step": 13840 }, { "epoch": 1.65, "learning_rate": 2.2217429197623337e-06, "loss": 0.5528, "step": 13841 }, { "epoch": 1.65, "learning_rate": 2.221423495780902e-06, "loss": 0.4125, "step": 13842 }, { "epoch": 1.65, "learning_rate": 2.2211040764043644e-06, "loss": 0.5163, "step": 13843 }, { "epoch": 1.65, "learning_rate": 2.2207846616379985e-06, "loss": 0.535, "step": 13844 }, { "epoch": 1.65, "learning_rate": 2.2204652514870867e-06, "loss": 0.622, "step": 13845 }, { "epoch": 1.65, "learning_rate": 2.2201458459569066e-06, "loss": 0.7621, "step": 13846 }, { "epoch": 1.65, "learning_rate": 2.2198264450527405e-06, "loss": 0.9011, "step": 13847 }, { "epoch": 1.65, "learning_rate": 2.2195070487798664e-06, "loss": 0.4152, "step": 13848 }, { "epoch": 1.65, "learning_rate": 2.2191876571435635e-06, "loss": 0.5228, "step": 13849 }, { "epoch": 1.65, "learning_rate": 2.2188682701491125e-06, "loss": 0.5317, "step": 13850 }, { "epoch": 1.65, "learning_rate": 2.2185488878017934e-06, "loss": 0.5963, "step": 13851 }, { "epoch": 1.65, "learning_rate": 2.2182295101068844e-06, "loss": 0.7129, "step": 13852 }, { "epoch": 1.65, "learning_rate": 2.2179101370696644e-06, "loss": 0.8446, "step": 13853 }, { "epoch": 1.65, "learning_rate": 2.217590768695414e-06, "loss": 0.4239, "step": 13854 }, { "epoch": 1.65, "learning_rate": 2.217271404989413e-06, "loss": 0.5454, "step": 13855 }, { "epoch": 1.65, "learning_rate": 2.216952045956938e-06, "loss": 0.6031, "step": 13856 }, { "epoch": 1.65, "learning_rate": 2.2166326916032698e-06, "loss": 0.5377, "step": 13857 }, { "epoch": 1.65, "learning_rate": 2.2163133419336877e-06, "loss": 0.8132, "step": 13858 }, { "epoch": 1.65, "learning_rate": 2.2159939969534693e-06, "loss": 0.8973, "step": 13859 }, { "epoch": 1.65, "learning_rate": 2.2156746566678937e-06, "loss": 0.5254, "step": 13860 }, { "epoch": 1.65, "learning_rate": 2.2153553210822397e-06, "loss": 0.5304, "step": 13861 }, { "epoch": 1.65, "learning_rate": 2.2150359902017873e-06, "loss": 0.5238, "step": 13862 }, { "epoch": 1.65, "learning_rate": 2.214716664031813e-06, "loss": 0.6487, "step": 13863 }, { "epoch": 1.65, "learning_rate": 2.2143973425775954e-06, "loss": 0.7943, "step": 13864 }, { "epoch": 1.65, "learning_rate": 2.214078025844415e-06, "loss": 0.7682, "step": 13865 }, { "epoch": 1.65, "learning_rate": 2.2137587138375486e-06, "loss": 0.4809, "step": 13866 }, { "epoch": 1.65, "learning_rate": 2.2134394065622743e-06, "loss": 0.4396, "step": 13867 }, { "epoch": 1.65, "learning_rate": 2.2131201040238708e-06, "loss": 0.6353, "step": 13868 }, { "epoch": 1.65, "learning_rate": 2.212800806227616e-06, "loss": 0.4643, "step": 13869 }, { "epoch": 1.65, "learning_rate": 2.2124815131787885e-06, "loss": 0.6546, "step": 13870 }, { "epoch": 1.65, "learning_rate": 2.212162224882665e-06, "loss": 0.7877, "step": 13871 }, { "epoch": 1.65, "learning_rate": 2.211842941344524e-06, "loss": 0.8367, "step": 13872 }, { "epoch": 1.65, "learning_rate": 2.2115236625696445e-06, "loss": 0.4408, "step": 13873 }, { "epoch": 1.65, "learning_rate": 2.2112043885633023e-06, "loss": 0.596, "step": 13874 }, { "epoch": 1.65, "learning_rate": 2.210885119330775e-06, "loss": 0.5768, "step": 13875 }, { "epoch": 1.65, "learning_rate": 2.2105658548773416e-06, "loss": 0.6082, "step": 13876 }, { "epoch": 1.65, "learning_rate": 2.21024659520828e-06, "loss": 0.7571, "step": 13877 }, { "epoch": 1.65, "learning_rate": 2.209927340328865e-06, "loss": 0.9388, "step": 13878 }, { "epoch": 1.65, "learning_rate": 2.2096080902443753e-06, "loss": 0.4821, "step": 13879 }, { "epoch": 1.65, "learning_rate": 2.209288844960089e-06, "loss": 0.5357, "step": 13880 }, { "epoch": 1.65, "learning_rate": 2.208969604481283e-06, "loss": 0.6252, "step": 13881 }, { "epoch": 1.65, "learning_rate": 2.2086503688132326e-06, "loss": 0.6975, "step": 13882 }, { "epoch": 1.65, "learning_rate": 2.2083311379612162e-06, "loss": 0.846, "step": 13883 }, { "epoch": 1.65, "learning_rate": 2.2080119119305114e-06, "loss": 0.871, "step": 13884 }, { "epoch": 1.65, "learning_rate": 2.2076926907263943e-06, "loss": 0.4337, "step": 13885 }, { "epoch": 1.65, "learning_rate": 2.2073734743541406e-06, "loss": 0.6111, "step": 13886 }, { "epoch": 1.65, "learning_rate": 2.2070542628190282e-06, "loss": 0.5878, "step": 13887 }, { "epoch": 1.65, "learning_rate": 2.2067350561263346e-06, "loss": 0.638, "step": 13888 }, { "epoch": 1.65, "learning_rate": 2.2064158542813334e-06, "loss": 0.6344, "step": 13889 }, { "epoch": 1.65, "learning_rate": 2.2060966572893032e-06, "loss": 0.8651, "step": 13890 }, { "epoch": 1.65, "learning_rate": 2.2057774651555203e-06, "loss": 0.5154, "step": 13891 }, { "epoch": 1.65, "learning_rate": 2.205458277885261e-06, "loss": 0.4481, "step": 13892 }, { "epoch": 1.65, "learning_rate": 2.2051390954838e-06, "loss": 0.591, "step": 13893 }, { "epoch": 1.65, "learning_rate": 2.2048199179564147e-06, "loss": 0.5399, "step": 13894 }, { "epoch": 1.65, "learning_rate": 2.2045007453083813e-06, "loss": 0.8867, "step": 13895 }, { "epoch": 1.65, "learning_rate": 2.204181577544976e-06, "loss": 0.7876, "step": 13896 }, { "epoch": 1.65, "learning_rate": 2.203862414671473e-06, "loss": 0.8518, "step": 13897 }, { "epoch": 1.66, "learning_rate": 2.2035432566931495e-06, "loss": 0.4222, "step": 13898 }, { "epoch": 1.66, "learning_rate": 2.2032241036152803e-06, "loss": 0.4144, "step": 13899 }, { "epoch": 1.66, "learning_rate": 2.202904955443143e-06, "loss": 0.6409, "step": 13900 }, { "epoch": 1.66, "learning_rate": 2.2025858121820105e-06, "loss": 0.5889, "step": 13901 }, { "epoch": 1.66, "learning_rate": 2.2022666738371597e-06, "loss": 0.86, "step": 13902 }, { "epoch": 1.66, "learning_rate": 2.2019475404138655e-06, "loss": 0.9633, "step": 13903 }, { "epoch": 1.66, "learning_rate": 2.2016284119174045e-06, "loss": 0.3874, "step": 13904 }, { "epoch": 1.66, "learning_rate": 2.2013092883530503e-06, "loss": 0.5157, "step": 13905 }, { "epoch": 1.66, "learning_rate": 2.2009901697260783e-06, "loss": 0.5591, "step": 13906 }, { "epoch": 1.66, "learning_rate": 2.2006710560417646e-06, "loss": 0.668, "step": 13907 }, { "epoch": 1.66, "learning_rate": 2.2003519473053825e-06, "loss": 0.6922, "step": 13908 }, { "epoch": 1.66, "learning_rate": 2.2000328435222084e-06, "loss": 0.9122, "step": 13909 }, { "epoch": 1.66, "learning_rate": 2.199713744697516e-06, "loss": 0.5046, "step": 13910 }, { "epoch": 1.66, "learning_rate": 2.199394650836582e-06, "loss": 0.5157, "step": 13911 }, { "epoch": 1.66, "learning_rate": 2.199075561944678e-06, "loss": 0.6734, "step": 13912 }, { "epoch": 1.66, "learning_rate": 2.1987564780270817e-06, "loss": 0.7592, "step": 13913 }, { "epoch": 1.66, "learning_rate": 2.1984373990890645e-06, "loss": 0.8064, "step": 13914 }, { "epoch": 1.66, "learning_rate": 2.198118325135904e-06, "loss": 0.9202, "step": 13915 }, { "epoch": 1.66, "learning_rate": 2.1977992561728724e-06, "loss": 0.4321, "step": 13916 }, { "epoch": 1.66, "learning_rate": 2.1974801922052436e-06, "loss": 0.544, "step": 13917 }, { "epoch": 1.66, "learning_rate": 2.1971611332382927e-06, "loss": 0.7447, "step": 13918 }, { "epoch": 1.66, "learning_rate": 2.196842079277295e-06, "loss": 0.5108, "step": 13919 }, { "epoch": 1.66, "learning_rate": 2.1965230303275226e-06, "loss": 0.6649, "step": 13920 }, { "epoch": 1.66, "learning_rate": 2.196203986394249e-06, "loss": 0.9266, "step": 13921 }, { "epoch": 1.66, "learning_rate": 2.1958849474827505e-06, "loss": 0.7684, "step": 13922 }, { "epoch": 1.66, "learning_rate": 2.195565913598298e-06, "loss": 0.3788, "step": 13923 }, { "epoch": 1.66, "learning_rate": 2.1952468847461674e-06, "loss": 0.4561, "step": 13924 }, { "epoch": 1.66, "learning_rate": 2.1949278609316306e-06, "loss": 0.5585, "step": 13925 }, { "epoch": 1.66, "learning_rate": 2.194608842159963e-06, "loss": 0.6768, "step": 13926 }, { "epoch": 1.66, "learning_rate": 2.194289828436436e-06, "loss": 0.72, "step": 13927 }, { "epoch": 1.66, "learning_rate": 2.1939708197663236e-06, "loss": 0.8707, "step": 13928 }, { "epoch": 1.66, "learning_rate": 2.193651816154899e-06, "loss": 0.5301, "step": 13929 }, { "epoch": 1.66, "learning_rate": 2.193332817607437e-06, "loss": 0.3569, "step": 13930 }, { "epoch": 1.66, "learning_rate": 2.1930138241292087e-06, "loss": 0.595, "step": 13931 }, { "epoch": 1.66, "learning_rate": 2.192694835725487e-06, "loss": 0.6232, "step": 13932 }, { "epoch": 1.66, "learning_rate": 2.1923758524015457e-06, "loss": 0.684, "step": 13933 }, { "epoch": 1.66, "learning_rate": 2.1920568741626585e-06, "loss": 0.8766, "step": 13934 }, { "epoch": 1.66, "learning_rate": 2.1917379010140953e-06, "loss": 0.4592, "step": 13935 }, { "epoch": 1.66, "learning_rate": 2.1914189329611308e-06, "loss": 0.5523, "step": 13936 }, { "epoch": 1.66, "learning_rate": 2.191099970009038e-06, "loss": 0.5592, "step": 13937 }, { "epoch": 1.66, "learning_rate": 2.1907810121630886e-06, "loss": 0.6486, "step": 13938 }, { "epoch": 1.66, "learning_rate": 2.190462059428554e-06, "loss": 0.6749, "step": 13939 }, { "epoch": 1.66, "learning_rate": 2.1901431118107073e-06, "loss": 0.7388, "step": 13940 }, { "epoch": 1.66, "learning_rate": 2.1898241693148228e-06, "loss": 0.5953, "step": 13941 }, { "epoch": 1.66, "learning_rate": 2.1895052319461697e-06, "loss": 0.5643, "step": 13942 }, { "epoch": 1.66, "learning_rate": 2.1891862997100212e-06, "loss": 0.6634, "step": 13943 }, { "epoch": 1.66, "learning_rate": 2.1888673726116494e-06, "loss": 0.5591, "step": 13944 }, { "epoch": 1.66, "learning_rate": 2.1885484506563265e-06, "loss": 0.6147, "step": 13945 }, { "epoch": 1.66, "learning_rate": 2.188229533849323e-06, "loss": 0.8427, "step": 13946 }, { "epoch": 1.66, "learning_rate": 2.187910622195911e-06, "loss": 0.7938, "step": 13947 }, { "epoch": 1.66, "learning_rate": 2.1875917157013638e-06, "loss": 0.4888, "step": 13948 }, { "epoch": 1.66, "learning_rate": 2.187272814370952e-06, "loss": 0.4748, "step": 13949 }, { "epoch": 1.66, "learning_rate": 2.1869539182099455e-06, "loss": 0.7178, "step": 13950 }, { "epoch": 1.66, "learning_rate": 2.186635027223617e-06, "loss": 0.6924, "step": 13951 }, { "epoch": 1.66, "learning_rate": 2.1863161414172393e-06, "loss": 0.8957, "step": 13952 }, { "epoch": 1.66, "learning_rate": 2.1859972607960815e-06, "loss": 0.8184, "step": 13953 }, { "epoch": 1.66, "learning_rate": 2.1856783853654146e-06, "loss": 0.3802, "step": 13954 }, { "epoch": 1.66, "learning_rate": 2.185359515130511e-06, "loss": 0.5226, "step": 13955 }, { "epoch": 1.66, "learning_rate": 2.185040650096641e-06, "loss": 0.5705, "step": 13956 }, { "epoch": 1.66, "learning_rate": 2.184721790269075e-06, "loss": 0.5579, "step": 13957 }, { "epoch": 1.66, "learning_rate": 2.1844029356530844e-06, "loss": 0.6985, "step": 13958 }, { "epoch": 1.66, "learning_rate": 2.18408408625394e-06, "loss": 0.8503, "step": 13959 }, { "epoch": 1.66, "learning_rate": 2.1837652420769128e-06, "loss": 0.4681, "step": 13960 }, { "epoch": 1.66, "learning_rate": 2.1834464031272714e-06, "loss": 0.5152, "step": 13961 }, { "epoch": 1.66, "learning_rate": 2.1831275694102882e-06, "loss": 0.5709, "step": 13962 }, { "epoch": 1.66, "learning_rate": 2.1828087409312323e-06, "loss": 0.6297, "step": 13963 }, { "epoch": 1.66, "learning_rate": 2.1824899176953755e-06, "loss": 0.7867, "step": 13964 }, { "epoch": 1.66, "learning_rate": 2.182171099707986e-06, "loss": 0.8887, "step": 13965 }, { "epoch": 1.66, "learning_rate": 2.181852286974336e-06, "loss": 0.5085, "step": 13966 }, { "epoch": 1.66, "learning_rate": 2.1815334794996933e-06, "loss": 0.3675, "step": 13967 }, { "epoch": 1.66, "learning_rate": 2.1812146772893297e-06, "loss": 0.572, "step": 13968 }, { "epoch": 1.66, "learning_rate": 2.1808958803485134e-06, "loss": 0.6401, "step": 13969 }, { "epoch": 1.66, "learning_rate": 2.1805770886825157e-06, "loss": 0.664, "step": 13970 }, { "epoch": 1.66, "learning_rate": 2.1802583022966062e-06, "loss": 0.8651, "step": 13971 }, { "epoch": 1.66, "learning_rate": 2.1799395211960523e-06, "loss": 0.7899, "step": 13972 }, { "epoch": 1.66, "learning_rate": 2.1796207453861255e-06, "loss": 0.4222, "step": 13973 }, { "epoch": 1.66, "learning_rate": 2.1793019748720943e-06, "loss": 0.5438, "step": 13974 }, { "epoch": 1.66, "learning_rate": 2.1789832096592294e-06, "loss": 0.5246, "step": 13975 }, { "epoch": 1.66, "learning_rate": 2.178664449752798e-06, "loss": 0.597, "step": 13976 }, { "epoch": 1.66, "learning_rate": 2.178345695158071e-06, "loss": 0.8371, "step": 13977 }, { "epoch": 1.66, "learning_rate": 2.1780269458803154e-06, "loss": 0.7599, "step": 13978 }, { "epoch": 1.66, "learning_rate": 2.177708201924803e-06, "loss": 0.4468, "step": 13979 }, { "epoch": 1.66, "learning_rate": 2.1773894632968e-06, "loss": 0.4672, "step": 13980 }, { "epoch": 1.66, "learning_rate": 2.1770707300015765e-06, "loss": 0.528, "step": 13981 }, { "epoch": 1.67, "learning_rate": 2.1767520020444004e-06, "loss": 0.5594, "step": 13982 }, { "epoch": 1.67, "learning_rate": 2.1764332794305423e-06, "loss": 0.697, "step": 13983 }, { "epoch": 1.67, "learning_rate": 2.1761145621652683e-06, "loss": 0.9682, "step": 13984 }, { "epoch": 1.67, "learning_rate": 2.1757958502538477e-06, "loss": 0.4268, "step": 13985 }, { "epoch": 1.67, "learning_rate": 2.1754771437015495e-06, "loss": 0.4503, "step": 13986 }, { "epoch": 1.67, "learning_rate": 2.1751584425136405e-06, "loss": 0.5857, "step": 13987 }, { "epoch": 1.67, "learning_rate": 2.17483974669539e-06, "loss": 0.6069, "step": 13988 }, { "epoch": 1.67, "learning_rate": 2.1745210562520654e-06, "loss": 0.8137, "step": 13989 }, { "epoch": 1.67, "learning_rate": 2.174202371188936e-06, "loss": 0.8224, "step": 13990 }, { "epoch": 1.67, "learning_rate": 2.1738836915112685e-06, "loss": 0.4025, "step": 13991 }, { "epoch": 1.67, "learning_rate": 2.17356501722433e-06, "loss": 0.4061, "step": 13992 }, { "epoch": 1.67, "learning_rate": 2.1732463483333894e-06, "loss": 0.4128, "step": 13993 }, { "epoch": 1.67, "learning_rate": 2.172927684843715e-06, "loss": 0.6714, "step": 13994 }, { "epoch": 1.67, "learning_rate": 2.172609026760573e-06, "loss": 0.6953, "step": 13995 }, { "epoch": 1.67, "learning_rate": 2.172290374089231e-06, "loss": 0.864, "step": 13996 }, { "epoch": 1.67, "learning_rate": 2.1719717268349562e-06, "loss": 0.8128, "step": 13997 }, { "epoch": 1.67, "learning_rate": 2.171653085003018e-06, "loss": 0.4301, "step": 13998 }, { "epoch": 1.67, "learning_rate": 2.1713344485986802e-06, "loss": 0.6035, "step": 13999 }, { "epoch": 1.67, "learning_rate": 2.1710158176272115e-06, "loss": 0.4674, "step": 14000 }, { "epoch": 1.67, "learning_rate": 2.1706971920938803e-06, "loss": 0.6766, "step": 14001 }, { "epoch": 1.67, "learning_rate": 2.1703785720039516e-06, "loss": 0.7462, "step": 14002 }, { "epoch": 1.67, "learning_rate": 2.170059957362692e-06, "loss": 0.7946, "step": 14003 }, { "epoch": 1.67, "learning_rate": 2.169741348175369e-06, "loss": 0.4687, "step": 14004 }, { "epoch": 1.67, "learning_rate": 2.1694227444472506e-06, "loss": 0.4487, "step": 14005 }, { "epoch": 1.67, "learning_rate": 2.169104146183601e-06, "loss": 0.5608, "step": 14006 }, { "epoch": 1.67, "learning_rate": 2.168785553389687e-06, "loss": 0.5206, "step": 14007 }, { "epoch": 1.67, "learning_rate": 2.1684669660707763e-06, "loss": 0.6867, "step": 14008 }, { "epoch": 1.67, "learning_rate": 2.168148384232135e-06, "loss": 0.8129, "step": 14009 }, { "epoch": 1.67, "learning_rate": 2.1678298078790284e-06, "loss": 0.4, "step": 14010 }, { "epoch": 1.67, "learning_rate": 2.167511237016722e-06, "loss": 0.582, "step": 14011 }, { "epoch": 1.67, "learning_rate": 2.1671926716504836e-06, "loss": 0.5479, "step": 14012 }, { "epoch": 1.67, "learning_rate": 2.166874111785579e-06, "loss": 0.6001, "step": 14013 }, { "epoch": 1.67, "learning_rate": 2.166555557427272e-06, "loss": 0.8678, "step": 14014 }, { "epoch": 1.67, "learning_rate": 2.1662370085808297e-06, "loss": 0.7922, "step": 14015 }, { "epoch": 1.67, "learning_rate": 2.1659184652515187e-06, "loss": 0.5237, "step": 14016 }, { "epoch": 1.67, "learning_rate": 2.165599927444603e-06, "loss": 0.411, "step": 14017 }, { "epoch": 1.67, "learning_rate": 2.165281395165348e-06, "loss": 0.5566, "step": 14018 }, { "epoch": 1.67, "learning_rate": 2.16496286841902e-06, "loss": 0.572, "step": 14019 }, { "epoch": 1.67, "learning_rate": 2.1646443472108846e-06, "loss": 0.6033, "step": 14020 }, { "epoch": 1.67, "learning_rate": 2.164325831546205e-06, "loss": 0.7852, "step": 14021 }, { "epoch": 1.67, "learning_rate": 2.164007321430248e-06, "loss": 0.801, "step": 14022 }, { "epoch": 1.67, "learning_rate": 2.1636888168682785e-06, "loss": 0.4275, "step": 14023 }, { "epoch": 1.67, "learning_rate": 2.1633703178655615e-06, "loss": 0.5667, "step": 14024 }, { "epoch": 1.67, "learning_rate": 2.16305182442736e-06, "loss": 0.5328, "step": 14025 }, { "epoch": 1.67, "learning_rate": 2.162733336558941e-06, "loss": 0.6447, "step": 14026 }, { "epoch": 1.67, "learning_rate": 2.162414854265568e-06, "loss": 0.7795, "step": 14027 }, { "epoch": 1.67, "learning_rate": 2.1620963775525068e-06, "loss": 0.8869, "step": 14028 }, { "epoch": 1.67, "learning_rate": 2.1617779064250194e-06, "loss": 0.4145, "step": 14029 }, { "epoch": 1.67, "learning_rate": 2.161459440888372e-06, "loss": 0.4983, "step": 14030 }, { "epoch": 1.67, "learning_rate": 2.1611409809478294e-06, "loss": 0.5404, "step": 14031 }, { "epoch": 1.67, "learning_rate": 2.1608225266086536e-06, "loss": 0.5574, "step": 14032 }, { "epoch": 1.67, "learning_rate": 2.16050407787611e-06, "loss": 0.6847, "step": 14033 }, { "epoch": 1.67, "learning_rate": 2.160185634755463e-06, "loss": 0.7975, "step": 14034 }, { "epoch": 1.67, "learning_rate": 2.1598671972519764e-06, "loss": 0.5144, "step": 14035 }, { "epoch": 1.67, "learning_rate": 2.1595487653709123e-06, "loss": 0.5195, "step": 14036 }, { "epoch": 1.67, "learning_rate": 2.159230339117536e-06, "loss": 0.5135, "step": 14037 }, { "epoch": 1.67, "learning_rate": 2.158911918497111e-06, "loss": 0.5789, "step": 14038 }, { "epoch": 1.67, "learning_rate": 2.1585935035149014e-06, "loss": 0.6028, "step": 14039 }, { "epoch": 1.67, "learning_rate": 2.1582750941761687e-06, "loss": 0.8716, "step": 14040 }, { "epoch": 1.67, "learning_rate": 2.157956690486178e-06, "loss": 0.6287, "step": 14041 }, { "epoch": 1.67, "learning_rate": 2.157638292450191e-06, "loss": 0.4091, "step": 14042 }, { "epoch": 1.67, "learning_rate": 2.1573199000734736e-06, "loss": 0.558, "step": 14043 }, { "epoch": 1.67, "learning_rate": 2.157001513361285e-06, "loss": 0.5735, "step": 14044 }, { "epoch": 1.67, "learning_rate": 2.1566831323188912e-06, "loss": 0.7482, "step": 14045 }, { "epoch": 1.67, "learning_rate": 2.1563647569515537e-06, "loss": 0.7213, "step": 14046 }, { "epoch": 1.67, "learning_rate": 2.1560463872645364e-06, "loss": 0.8363, "step": 14047 }, { "epoch": 1.67, "learning_rate": 2.155728023263101e-06, "loss": 0.6096, "step": 14048 }, { "epoch": 1.67, "learning_rate": 2.155409664952509e-06, "loss": 0.5021, "step": 14049 }, { "epoch": 1.67, "learning_rate": 2.1550913123380256e-06, "loss": 0.5387, "step": 14050 }, { "epoch": 1.67, "learning_rate": 2.1547729654249105e-06, "loss": 0.7116, "step": 14051 }, { "epoch": 1.67, "learning_rate": 2.1544546242184282e-06, "loss": 0.8187, "step": 14052 }, { "epoch": 1.67, "learning_rate": 2.154136288723839e-06, "loss": 0.8807, "step": 14053 }, { "epoch": 1.67, "learning_rate": 2.1538179589464074e-06, "loss": 0.4382, "step": 14054 }, { "epoch": 1.67, "learning_rate": 2.1534996348913924e-06, "loss": 0.5129, "step": 14055 }, { "epoch": 1.67, "learning_rate": 2.1531813165640587e-06, "loss": 0.4928, "step": 14056 }, { "epoch": 1.67, "learning_rate": 2.152863003969666e-06, "loss": 0.5005, "step": 14057 }, { "epoch": 1.67, "learning_rate": 2.1525446971134782e-06, "loss": 0.6238, "step": 14058 }, { "epoch": 1.67, "learning_rate": 2.152226396000755e-06, "loss": 0.9363, "step": 14059 }, { "epoch": 1.67, "learning_rate": 2.151908100636758e-06, "loss": 0.4593, "step": 14060 }, { "epoch": 1.67, "learning_rate": 2.1515898110267496e-06, "loss": 0.5014, "step": 14061 }, { "epoch": 1.67, "learning_rate": 2.1512715271759916e-06, "loss": 0.4975, "step": 14062 }, { "epoch": 1.67, "learning_rate": 2.1509532490897444e-06, "loss": 0.5177, "step": 14063 }, { "epoch": 1.67, "learning_rate": 2.1506349767732687e-06, "loss": 0.807, "step": 14064 }, { "epoch": 1.68, "learning_rate": 2.1503167102318272e-06, "loss": 0.853, "step": 14065 }, { "epoch": 1.68, "learning_rate": 2.1499984494706784e-06, "loss": 0.505, "step": 14066 }, { "epoch": 1.68, "learning_rate": 2.1496801944950855e-06, "loss": 0.4022, "step": 14067 }, { "epoch": 1.68, "learning_rate": 2.1493619453103076e-06, "loss": 0.545, "step": 14068 }, { "epoch": 1.68, "learning_rate": 2.1490437019216075e-06, "loss": 0.6372, "step": 14069 }, { "epoch": 1.68, "learning_rate": 2.1487254643342438e-06, "loss": 0.8144, "step": 14070 }, { "epoch": 1.68, "learning_rate": 2.148407232553477e-06, "loss": 0.8406, "step": 14071 }, { "epoch": 1.68, "learning_rate": 2.1480890065845683e-06, "loss": 0.8112, "step": 14072 }, { "epoch": 1.68, "learning_rate": 2.147770786432779e-06, "loss": 0.3139, "step": 14073 }, { "epoch": 1.68, "learning_rate": 2.1474525721033673e-06, "loss": 0.5571, "step": 14074 }, { "epoch": 1.68, "learning_rate": 2.1471343636015937e-06, "loss": 0.6198, "step": 14075 }, { "epoch": 1.68, "learning_rate": 2.1468161609327193e-06, "loss": 0.6406, "step": 14076 }, { "epoch": 1.68, "learning_rate": 2.146497964102004e-06, "loss": 0.762, "step": 14077 }, { "epoch": 1.68, "learning_rate": 2.146179773114706e-06, "loss": 0.7423, "step": 14078 }, { "epoch": 1.68, "learning_rate": 2.1458615879760858e-06, "loss": 0.418, "step": 14079 }, { "epoch": 1.68, "learning_rate": 2.1455434086914046e-06, "loss": 0.5593, "step": 14080 }, { "epoch": 1.68, "learning_rate": 2.1452252352659196e-06, "loss": 0.5766, "step": 14081 }, { "epoch": 1.68, "learning_rate": 2.144907067704891e-06, "loss": 0.6581, "step": 14082 }, { "epoch": 1.68, "learning_rate": 2.1445889060135787e-06, "loss": 0.7293, "step": 14083 }, { "epoch": 1.68, "learning_rate": 2.144270750197242e-06, "loss": 0.8282, "step": 14084 }, { "epoch": 1.68, "learning_rate": 2.1439526002611396e-06, "loss": 0.4146, "step": 14085 }, { "epoch": 1.68, "learning_rate": 2.1436344562105298e-06, "loss": 0.4692, "step": 14086 }, { "epoch": 1.68, "learning_rate": 2.1433163180506733e-06, "loss": 0.4236, "step": 14087 }, { "epoch": 1.68, "learning_rate": 2.142998185786828e-06, "loss": 0.6458, "step": 14088 }, { "epoch": 1.68, "learning_rate": 2.142680059424252e-06, "loss": 0.722, "step": 14089 }, { "epoch": 1.68, "learning_rate": 2.142361938968204e-06, "loss": 0.8373, "step": 14090 }, { "epoch": 1.68, "learning_rate": 2.1420438244239444e-06, "loss": 0.5179, "step": 14091 }, { "epoch": 1.68, "learning_rate": 2.141725715796731e-06, "loss": 0.5878, "step": 14092 }, { "epoch": 1.68, "learning_rate": 2.14140761309182e-06, "loss": 0.5959, "step": 14093 }, { "epoch": 1.68, "learning_rate": 2.1410895163144714e-06, "loss": 0.5369, "step": 14094 }, { "epoch": 1.68, "learning_rate": 2.140771425469944e-06, "loss": 0.7224, "step": 14095 }, { "epoch": 1.68, "learning_rate": 2.140453340563495e-06, "loss": 0.9417, "step": 14096 }, { "epoch": 1.68, "learning_rate": 2.1401352616003816e-06, "loss": 0.8429, "step": 14097 }, { "epoch": 1.68, "learning_rate": 2.1398171885858634e-06, "loss": 0.3416, "step": 14098 }, { "epoch": 1.68, "learning_rate": 2.1394991215251974e-06, "loss": 0.5399, "step": 14099 }, { "epoch": 1.68, "learning_rate": 2.1391810604236402e-06, "loss": 0.6799, "step": 14100 }, { "epoch": 1.68, "learning_rate": 2.138863005286451e-06, "loss": 0.635, "step": 14101 }, { "epoch": 1.68, "learning_rate": 2.1385449561188865e-06, "loss": 0.8498, "step": 14102 }, { "epoch": 1.68, "learning_rate": 2.1382269129262046e-06, "loss": 0.893, "step": 14103 }, { "epoch": 1.68, "learning_rate": 2.1379088757136617e-06, "loss": 0.5106, "step": 14104 }, { "epoch": 1.68, "learning_rate": 2.1375908444865157e-06, "loss": 0.5153, "step": 14105 }, { "epoch": 1.68, "learning_rate": 2.137272819250023e-06, "loss": 0.4981, "step": 14106 }, { "epoch": 1.68, "learning_rate": 2.136954800009442e-06, "loss": 0.5813, "step": 14107 }, { "epoch": 1.68, "learning_rate": 2.1366367867700273e-06, "loss": 0.7379, "step": 14108 }, { "epoch": 1.68, "learning_rate": 2.1363187795370378e-06, "loss": 0.8192, "step": 14109 }, { "epoch": 1.68, "learning_rate": 2.136000778315729e-06, "loss": 0.532, "step": 14110 }, { "epoch": 1.68, "learning_rate": 2.1356827831113586e-06, "loss": 0.4752, "step": 14111 }, { "epoch": 1.68, "learning_rate": 2.1353647939291816e-06, "loss": 0.4682, "step": 14112 }, { "epoch": 1.68, "learning_rate": 2.1350468107744554e-06, "loss": 0.5636, "step": 14113 }, { "epoch": 1.68, "learning_rate": 2.134728833652437e-06, "loss": 0.7167, "step": 14114 }, { "epoch": 1.68, "learning_rate": 2.1344108625683804e-06, "loss": 0.9178, "step": 14115 }, { "epoch": 1.68, "learning_rate": 2.134092897527543e-06, "loss": 0.5944, "step": 14116 }, { "epoch": 1.68, "learning_rate": 2.1337749385351806e-06, "loss": 0.412, "step": 14117 }, { "epoch": 1.68, "learning_rate": 2.13345698559655e-06, "loss": 0.5981, "step": 14118 }, { "epoch": 1.68, "learning_rate": 2.1331390387169053e-06, "loss": 0.6092, "step": 14119 }, { "epoch": 1.68, "learning_rate": 2.132821097901504e-06, "loss": 0.7181, "step": 14120 }, { "epoch": 1.68, "learning_rate": 2.1325031631555996e-06, "loss": 0.7859, "step": 14121 }, { "epoch": 1.68, "learning_rate": 2.13218523448445e-06, "loss": 0.7637, "step": 14122 }, { "epoch": 1.68, "learning_rate": 2.1318673118933085e-06, "loss": 0.5198, "step": 14123 }, { "epoch": 1.68, "learning_rate": 2.1315493953874316e-06, "loss": 0.5551, "step": 14124 }, { "epoch": 1.68, "learning_rate": 2.1312314849720735e-06, "loss": 0.5995, "step": 14125 }, { "epoch": 1.68, "learning_rate": 2.130913580652491e-06, "loss": 0.6998, "step": 14126 }, { "epoch": 1.68, "learning_rate": 2.1305956824339373e-06, "loss": 0.7017, "step": 14127 }, { "epoch": 1.68, "learning_rate": 2.1302777903216674e-06, "loss": 0.7856, "step": 14128 }, { "epoch": 1.68, "learning_rate": 2.129959904320938e-06, "loss": 0.4534, "step": 14129 }, { "epoch": 1.68, "learning_rate": 2.1296420244370015e-06, "loss": 0.527, "step": 14130 }, { "epoch": 1.68, "learning_rate": 2.1293241506751135e-06, "loss": 0.5263, "step": 14131 }, { "epoch": 1.68, "learning_rate": 2.1290062830405283e-06, "loss": 0.6313, "step": 14132 }, { "epoch": 1.68, "learning_rate": 2.128688421538501e-06, "loss": 0.845, "step": 14133 }, { "epoch": 1.68, "learning_rate": 2.128370566174285e-06, "loss": 0.8766, "step": 14134 }, { "epoch": 1.68, "learning_rate": 2.1280527169531344e-06, "loss": 0.4774, "step": 14135 }, { "epoch": 1.68, "learning_rate": 2.1277348738803033e-06, "loss": 0.454, "step": 14136 }, { "epoch": 1.68, "learning_rate": 2.127417036961047e-06, "loss": 0.5249, "step": 14137 }, { "epoch": 1.68, "learning_rate": 2.127099206200618e-06, "loss": 0.5597, "step": 14138 }, { "epoch": 1.68, "learning_rate": 2.1267813816042697e-06, "loss": 0.6476, "step": 14139 }, { "epoch": 1.68, "learning_rate": 2.1264635631772564e-06, "loss": 0.7432, "step": 14140 }, { "epoch": 1.68, "learning_rate": 2.126145750924833e-06, "loss": 0.5234, "step": 14141 }, { "epoch": 1.68, "learning_rate": 2.125827944852251e-06, "loss": 0.449, "step": 14142 }, { "epoch": 1.68, "learning_rate": 2.1255101449647644e-06, "loss": 0.5586, "step": 14143 }, { "epoch": 1.68, "learning_rate": 2.125192351267627e-06, "loss": 0.6165, "step": 14144 }, { "epoch": 1.68, "learning_rate": 2.1248745637660912e-06, "loss": 0.7726, "step": 14145 }, { "epoch": 1.68, "learning_rate": 2.1245567824654095e-06, "loss": 0.7965, "step": 14146 }, { "epoch": 1.68, "learning_rate": 2.124239007370836e-06, "loss": 0.7975, "step": 14147 }, { "epoch": 1.68, "learning_rate": 2.1239212384876244e-06, "loss": 0.4474, "step": 14148 }, { "epoch": 1.69, "learning_rate": 2.1236034758210255e-06, "loss": 0.4655, "step": 14149 }, { "epoch": 1.69, "learning_rate": 2.1232857193762923e-06, "loss": 0.6275, "step": 14150 }, { "epoch": 1.69, "learning_rate": 2.122967969158678e-06, "loss": 0.6701, "step": 14151 }, { "epoch": 1.69, "learning_rate": 2.1226502251734353e-06, "loss": 0.8727, "step": 14152 }, { "epoch": 1.69, "learning_rate": 2.122332487425816e-06, "loss": 0.947, "step": 14153 }, { "epoch": 1.69, "learning_rate": 2.1220147559210717e-06, "loss": 0.4732, "step": 14154 }, { "epoch": 1.69, "learning_rate": 2.121697030664456e-06, "loss": 0.5343, "step": 14155 }, { "epoch": 1.69, "learning_rate": 2.1213793116612203e-06, "loss": 0.5675, "step": 14156 }, { "epoch": 1.69, "learning_rate": 2.121061598916615e-06, "loss": 0.6291, "step": 14157 }, { "epoch": 1.69, "learning_rate": 2.120743892435894e-06, "loss": 0.8022, "step": 14158 }, { "epoch": 1.69, "learning_rate": 2.120426192224309e-06, "loss": 0.8648, "step": 14159 }, { "epoch": 1.69, "learning_rate": 2.1201084982871102e-06, "loss": 0.45, "step": 14160 }, { "epoch": 1.69, "learning_rate": 2.1197908106295495e-06, "loss": 0.4989, "step": 14161 }, { "epoch": 1.69, "learning_rate": 2.119473129256879e-06, "loss": 0.5815, "step": 14162 }, { "epoch": 1.69, "learning_rate": 2.1191554541743502e-06, "loss": 0.5898, "step": 14163 }, { "epoch": 1.69, "learning_rate": 2.1188377853872127e-06, "loss": 0.6653, "step": 14164 }, { "epoch": 1.69, "learning_rate": 2.1185201229007184e-06, "loss": 0.8308, "step": 14165 }, { "epoch": 1.69, "learning_rate": 2.1182024667201188e-06, "loss": 0.5275, "step": 14166 }, { "epoch": 1.69, "learning_rate": 2.1178848168506655e-06, "loss": 0.3421, "step": 14167 }, { "epoch": 1.69, "learning_rate": 2.1175671732976072e-06, "loss": 0.5496, "step": 14168 }, { "epoch": 1.69, "learning_rate": 2.1172495360661956e-06, "loss": 0.5683, "step": 14169 }, { "epoch": 1.69, "learning_rate": 2.1169319051616812e-06, "loss": 0.6525, "step": 14170 }, { "epoch": 1.69, "learning_rate": 2.1166142805893155e-06, "loss": 0.7802, "step": 14171 }, { "epoch": 1.69, "learning_rate": 2.116296662354347e-06, "loss": 0.7316, "step": 14172 }, { "epoch": 1.69, "learning_rate": 2.115979050462027e-06, "loss": 0.4998, "step": 14173 }, { "epoch": 1.69, "learning_rate": 2.1156614449176067e-06, "loss": 0.5343, "step": 14174 }, { "epoch": 1.69, "learning_rate": 2.1153438457263333e-06, "loss": 0.6238, "step": 14175 }, { "epoch": 1.69, "learning_rate": 2.1150262528934584e-06, "loss": 0.7245, "step": 14176 }, { "epoch": 1.69, "learning_rate": 2.114708666424233e-06, "loss": 0.8017, "step": 14177 }, { "epoch": 1.69, "learning_rate": 2.1143910863239055e-06, "loss": 0.7549, "step": 14178 }, { "epoch": 1.69, "learning_rate": 2.1140735125977247e-06, "loss": 0.4031, "step": 14179 }, { "epoch": 1.69, "learning_rate": 2.1137559452509417e-06, "loss": 0.4747, "step": 14180 }, { "epoch": 1.69, "learning_rate": 2.1134383842888055e-06, "loss": 0.6078, "step": 14181 }, { "epoch": 1.69, "learning_rate": 2.1131208297165655e-06, "loss": 0.5363, "step": 14182 }, { "epoch": 1.69, "learning_rate": 2.11280328153947e-06, "loss": 0.718, "step": 14183 }, { "epoch": 1.69, "learning_rate": 2.112485739762769e-06, "loss": 0.715, "step": 14184 }, { "epoch": 1.69, "learning_rate": 2.112168204391711e-06, "loss": 0.3592, "step": 14185 }, { "epoch": 1.69, "learning_rate": 2.111850675431546e-06, "loss": 0.3965, "step": 14186 }, { "epoch": 1.69, "learning_rate": 2.1115331528875207e-06, "loss": 0.6126, "step": 14187 }, { "epoch": 1.69, "learning_rate": 2.1112156367648858e-06, "loss": 0.6317, "step": 14188 }, { "epoch": 1.69, "learning_rate": 2.1108981270688884e-06, "loss": 0.7195, "step": 14189 }, { "epoch": 1.69, "learning_rate": 2.1105806238047786e-06, "loss": 0.8146, "step": 14190 }, { "epoch": 1.69, "learning_rate": 2.110263126977803e-06, "loss": 0.657, "step": 14191 }, { "epoch": 1.69, "learning_rate": 2.1099456365932105e-06, "loss": 0.4526, "step": 14192 }, { "epoch": 1.69, "learning_rate": 2.1096281526562504e-06, "loss": 0.4613, "step": 14193 }, { "epoch": 1.69, "learning_rate": 2.1093106751721685e-06, "loss": 0.6976, "step": 14194 }, { "epoch": 1.69, "learning_rate": 2.1089932041462144e-06, "loss": 0.7158, "step": 14195 }, { "epoch": 1.69, "learning_rate": 2.108675739583635e-06, "loss": 0.9081, "step": 14196 }, { "epoch": 1.69, "learning_rate": 2.1083582814896792e-06, "loss": 0.7796, "step": 14197 }, { "epoch": 1.69, "learning_rate": 2.1080408298695934e-06, "loss": 0.4761, "step": 14198 }, { "epoch": 1.69, "learning_rate": 2.1077233847286257e-06, "loss": 0.5543, "step": 14199 }, { "epoch": 1.69, "learning_rate": 2.107405946072023e-06, "loss": 0.6626, "step": 14200 }, { "epoch": 1.69, "learning_rate": 2.107088513905034e-06, "loss": 0.7362, "step": 14201 }, { "epoch": 1.69, "learning_rate": 2.1067710882329045e-06, "loss": 0.7333, "step": 14202 }, { "epoch": 1.69, "learning_rate": 2.106453669060881e-06, "loss": 0.8477, "step": 14203 }, { "epoch": 1.69, "learning_rate": 2.1061362563942115e-06, "loss": 0.4671, "step": 14204 }, { "epoch": 1.69, "learning_rate": 2.105818850238144e-06, "loss": 0.4887, "step": 14205 }, { "epoch": 1.69, "learning_rate": 2.105501450597923e-06, "loss": 0.5513, "step": 14206 }, { "epoch": 1.69, "learning_rate": 2.105184057478796e-06, "loss": 0.6299, "step": 14207 }, { "epoch": 1.69, "learning_rate": 2.1048666708860104e-06, "loss": 0.7125, "step": 14208 }, { "epoch": 1.69, "learning_rate": 2.1045492908248113e-06, "loss": 0.8373, "step": 14209 }, { "epoch": 1.69, "learning_rate": 2.1042319173004457e-06, "loss": 0.5205, "step": 14210 }, { "epoch": 1.69, "learning_rate": 2.103914550318159e-06, "loss": 0.4693, "step": 14211 }, { "epoch": 1.69, "learning_rate": 2.1035971898831995e-06, "loss": 0.5903, "step": 14212 }, { "epoch": 1.69, "learning_rate": 2.103279836000811e-06, "loss": 0.5522, "step": 14213 }, { "epoch": 1.69, "learning_rate": 2.1029624886762397e-06, "loss": 0.6957, "step": 14214 }, { "epoch": 1.69, "learning_rate": 2.1026451479147316e-06, "loss": 0.934, "step": 14215 }, { "epoch": 1.69, "learning_rate": 2.1023278137215338e-06, "loss": 0.6186, "step": 14216 }, { "epoch": 1.69, "learning_rate": 2.1020104861018897e-06, "loss": 0.5116, "step": 14217 }, { "epoch": 1.69, "learning_rate": 2.1016931650610455e-06, "loss": 0.6958, "step": 14218 }, { "epoch": 1.69, "learning_rate": 2.1013758506042472e-06, "loss": 0.5719, "step": 14219 }, { "epoch": 1.69, "learning_rate": 2.1010585427367398e-06, "loss": 0.6215, "step": 14220 }, { "epoch": 1.69, "learning_rate": 2.1007412414637675e-06, "loss": 0.8293, "step": 14221 }, { "epoch": 1.69, "learning_rate": 2.1004239467905757e-06, "loss": 0.8323, "step": 14222 }, { "epoch": 1.69, "learning_rate": 2.100106658722411e-06, "loss": 0.4473, "step": 14223 }, { "epoch": 1.69, "learning_rate": 2.099789377264516e-06, "loss": 0.6323, "step": 14224 }, { "epoch": 1.69, "learning_rate": 2.0994721024221356e-06, "loss": 0.5208, "step": 14225 }, { "epoch": 1.69, "learning_rate": 2.099154834200515e-06, "loss": 0.6661, "step": 14226 }, { "epoch": 1.69, "learning_rate": 2.0988375726048997e-06, "loss": 0.8578, "step": 14227 }, { "epoch": 1.69, "learning_rate": 2.098520317640532e-06, "loss": 0.7354, "step": 14228 }, { "epoch": 1.69, "learning_rate": 2.0982030693126573e-06, "loss": 0.4532, "step": 14229 }, { "epoch": 1.69, "learning_rate": 2.09788582762652e-06, "loss": 0.5771, "step": 14230 }, { "epoch": 1.69, "learning_rate": 2.097568592587364e-06, "loss": 0.5979, "step": 14231 }, { "epoch": 1.69, "learning_rate": 2.0972513642004317e-06, "loss": 0.6863, "step": 14232 }, { "epoch": 1.7, "learning_rate": 2.0969341424709687e-06, "loss": 0.6544, "step": 14233 }, { "epoch": 1.7, "learning_rate": 2.096616927404218e-06, "loss": 0.9274, "step": 14234 }, { "epoch": 1.7, "learning_rate": 2.0962997190054245e-06, "loss": 0.4451, "step": 14235 }, { "epoch": 1.7, "learning_rate": 2.095982517279829e-06, "loss": 0.3861, "step": 14236 }, { "epoch": 1.7, "learning_rate": 2.095665322232677e-06, "loss": 0.508, "step": 14237 }, { "epoch": 1.7, "learning_rate": 2.0953481338692117e-06, "loss": 0.652, "step": 14238 }, { "epoch": 1.7, "learning_rate": 2.095030952194675e-06, "loss": 0.7651, "step": 14239 }, { "epoch": 1.7, "learning_rate": 2.0947137772143107e-06, "loss": 0.7754, "step": 14240 }, { "epoch": 1.7, "learning_rate": 2.094396608933362e-06, "loss": 0.6087, "step": 14241 }, { "epoch": 1.7, "learning_rate": 2.0940794473570715e-06, "loss": 0.4388, "step": 14242 }, { "epoch": 1.7, "learning_rate": 2.0937622924906813e-06, "loss": 0.5638, "step": 14243 }, { "epoch": 1.7, "learning_rate": 2.093445144339434e-06, "loss": 0.5863, "step": 14244 }, { "epoch": 1.7, "learning_rate": 2.0931280029085733e-06, "loss": 0.7707, "step": 14245 }, { "epoch": 1.7, "learning_rate": 2.0928108682033414e-06, "loss": 0.7717, "step": 14246 }, { "epoch": 1.7, "learning_rate": 2.0924937402289793e-06, "loss": 0.7441, "step": 14247 }, { "epoch": 1.7, "learning_rate": 2.09217661899073e-06, "loss": 0.3375, "step": 14248 }, { "epoch": 1.7, "learning_rate": 2.091859504493835e-06, "loss": 0.5425, "step": 14249 }, { "epoch": 1.7, "learning_rate": 2.0915423967435377e-06, "loss": 0.5152, "step": 14250 }, { "epoch": 1.7, "learning_rate": 2.0912252957450772e-06, "loss": 0.5817, "step": 14251 }, { "epoch": 1.7, "learning_rate": 2.090908201503698e-06, "loss": 0.8358, "step": 14252 }, { "epoch": 1.7, "learning_rate": 2.0905911140246403e-06, "loss": 0.8878, "step": 14253 }, { "epoch": 1.7, "learning_rate": 2.0902740333131462e-06, "loss": 0.4617, "step": 14254 }, { "epoch": 1.7, "learning_rate": 2.089956959374456e-06, "loss": 0.4308, "step": 14255 }, { "epoch": 1.7, "learning_rate": 2.0896398922138124e-06, "loss": 0.5448, "step": 14256 }, { "epoch": 1.7, "learning_rate": 2.089322831836456e-06, "loss": 0.5451, "step": 14257 }, { "epoch": 1.7, "learning_rate": 2.0890057782476265e-06, "loss": 0.8179, "step": 14258 }, { "epoch": 1.7, "learning_rate": 2.088688731452567e-06, "loss": 0.8793, "step": 14259 }, { "epoch": 1.7, "learning_rate": 2.0883716914565166e-06, "loss": 0.5413, "step": 14260 }, { "epoch": 1.7, "learning_rate": 2.0880546582647177e-06, "loss": 0.5138, "step": 14261 }, { "epoch": 1.7, "learning_rate": 2.0877376318824086e-06, "loss": 0.5448, "step": 14262 }, { "epoch": 1.7, "learning_rate": 2.0874206123148315e-06, "loss": 0.4463, "step": 14263 }, { "epoch": 1.7, "learning_rate": 2.087103599567226e-06, "loss": 0.669, "step": 14264 }, { "epoch": 1.7, "learning_rate": 2.0867865936448335e-06, "loss": 0.8193, "step": 14265 }, { "epoch": 1.7, "learning_rate": 2.0864695945528926e-06, "loss": 0.4727, "step": 14266 }, { "epoch": 1.7, "learning_rate": 2.0861526022966443e-06, "loss": 0.4058, "step": 14267 }, { "epoch": 1.7, "learning_rate": 2.0858356168813276e-06, "loss": 0.6045, "step": 14268 }, { "epoch": 1.7, "learning_rate": 2.085518638312184e-06, "loss": 0.5888, "step": 14269 }, { "epoch": 1.7, "learning_rate": 2.085201666594452e-06, "loss": 0.725, "step": 14270 }, { "epoch": 1.7, "learning_rate": 2.08488470173337e-06, "loss": 0.7551, "step": 14271 }, { "epoch": 1.7, "learning_rate": 2.08456774373418e-06, "loss": 0.9043, "step": 14272 }, { "epoch": 1.7, "learning_rate": 2.084250792602119e-06, "loss": 0.4745, "step": 14273 }, { "epoch": 1.7, "learning_rate": 2.0839338483424284e-06, "loss": 0.5539, "step": 14274 }, { "epoch": 1.7, "learning_rate": 2.0836169109603448e-06, "loss": 0.6391, "step": 14275 }, { "epoch": 1.7, "learning_rate": 2.0832999804611103e-06, "loss": 0.5653, "step": 14276 }, { "epoch": 1.7, "learning_rate": 2.082983056849961e-06, "loss": 0.7188, "step": 14277 }, { "epoch": 1.7, "learning_rate": 2.0826661401321367e-06, "loss": 0.8764, "step": 14278 }, { "epoch": 1.7, "learning_rate": 2.0823492303128758e-06, "loss": 0.4571, "step": 14279 }, { "epoch": 1.7, "learning_rate": 2.0820323273974183e-06, "loss": 0.4245, "step": 14280 }, { "epoch": 1.7, "learning_rate": 2.081715431391001e-06, "loss": 0.5203, "step": 14281 }, { "epoch": 1.7, "learning_rate": 2.081398542298862e-06, "loss": 0.6667, "step": 14282 }, { "epoch": 1.7, "learning_rate": 2.0810816601262408e-06, "loss": 0.6733, "step": 14283 }, { "epoch": 1.7, "learning_rate": 2.080764784878376e-06, "loss": 0.8284, "step": 14284 }, { "epoch": 1.7, "learning_rate": 2.080447916560503e-06, "loss": 0.4433, "step": 14285 }, { "epoch": 1.7, "learning_rate": 2.0801310551778616e-06, "loss": 0.5561, "step": 14286 }, { "epoch": 1.7, "learning_rate": 2.0798142007356894e-06, "loss": 0.5461, "step": 14287 }, { "epoch": 1.7, "learning_rate": 2.0794973532392235e-06, "loss": 0.5587, "step": 14288 }, { "epoch": 1.7, "learning_rate": 2.0791805126937013e-06, "loss": 0.6831, "step": 14289 }, { "epoch": 1.7, "learning_rate": 2.07886367910436e-06, "loss": 0.8555, "step": 14290 }, { "epoch": 1.7, "learning_rate": 2.078546852476439e-06, "loss": 0.4193, "step": 14291 }, { "epoch": 1.7, "learning_rate": 2.0782300328151734e-06, "loss": 0.4294, "step": 14292 }, { "epoch": 1.7, "learning_rate": 2.0779132201257998e-06, "loss": 0.5899, "step": 14293 }, { "epoch": 1.7, "learning_rate": 2.0775964144135564e-06, "loss": 0.5264, "step": 14294 }, { "epoch": 1.7, "learning_rate": 2.0772796156836804e-06, "loss": 0.7378, "step": 14295 }, { "epoch": 1.7, "learning_rate": 2.0769628239414078e-06, "loss": 0.8284, "step": 14296 }, { "epoch": 1.7, "learning_rate": 2.0766460391919745e-06, "loss": 0.8675, "step": 14297 }, { "epoch": 1.7, "learning_rate": 2.076329261440618e-06, "loss": 0.4296, "step": 14298 }, { "epoch": 1.7, "learning_rate": 2.076012490692575e-06, "loss": 0.6244, "step": 14299 }, { "epoch": 1.7, "learning_rate": 2.0756957269530805e-06, "loss": 0.5729, "step": 14300 }, { "epoch": 1.7, "learning_rate": 2.0753789702273706e-06, "loss": 0.6993, "step": 14301 }, { "epoch": 1.7, "learning_rate": 2.0750622205206834e-06, "loss": 0.8538, "step": 14302 }, { "epoch": 1.7, "learning_rate": 2.0747454778382525e-06, "loss": 0.7224, "step": 14303 }, { "epoch": 1.7, "learning_rate": 2.074428742185314e-06, "loss": 0.526, "step": 14304 }, { "epoch": 1.7, "learning_rate": 2.074112013567105e-06, "loss": 0.4884, "step": 14305 }, { "epoch": 1.7, "learning_rate": 2.07379529198886e-06, "loss": 0.5541, "step": 14306 }, { "epoch": 1.7, "learning_rate": 2.0734785774558145e-06, "loss": 0.5269, "step": 14307 }, { "epoch": 1.7, "learning_rate": 2.0731618699732036e-06, "loss": 0.7134, "step": 14308 }, { "epoch": 1.7, "learning_rate": 2.0728451695462627e-06, "loss": 0.8075, "step": 14309 }, { "epoch": 1.7, "learning_rate": 2.072528476180228e-06, "loss": 0.5891, "step": 14310 }, { "epoch": 1.7, "learning_rate": 2.0722117898803325e-06, "loss": 0.5885, "step": 14311 }, { "epoch": 1.7, "learning_rate": 2.071895110651812e-06, "loss": 0.5504, "step": 14312 }, { "epoch": 1.7, "learning_rate": 2.0715784384999016e-06, "loss": 0.5973, "step": 14313 }, { "epoch": 1.7, "learning_rate": 2.071261773429836e-06, "loss": 0.7589, "step": 14314 }, { "epoch": 1.7, "learning_rate": 2.0709451154468486e-06, "loss": 0.863, "step": 14315 }, { "epoch": 1.7, "learning_rate": 2.0706284645561746e-06, "loss": 0.7067, "step": 14316 }, { "epoch": 1.71, "learning_rate": 2.0703118207630484e-06, "loss": 0.446, "step": 14317 }, { "epoch": 1.71, "learning_rate": 2.0699951840727034e-06, "loss": 0.5386, "step": 14318 }, { "epoch": 1.71, "learning_rate": 2.0696785544903734e-06, "loss": 0.6341, "step": 14319 }, { "epoch": 1.71, "learning_rate": 2.0693619320212944e-06, "loss": 0.5946, "step": 14320 }, { "epoch": 1.71, "learning_rate": 2.0690453166706985e-06, "loss": 0.8223, "step": 14321 }, { "epoch": 1.71, "learning_rate": 2.068728708443819e-06, "loss": 0.6716, "step": 14322 }, { "epoch": 1.71, "learning_rate": 2.0684121073458905e-06, "loss": 0.5285, "step": 14323 }, { "epoch": 1.71, "learning_rate": 2.0680955133821464e-06, "loss": 0.6552, "step": 14324 }, { "epoch": 1.71, "learning_rate": 2.06777892655782e-06, "loss": 0.6466, "step": 14325 }, { "epoch": 1.71, "learning_rate": 2.067462346878143e-06, "loss": 0.6496, "step": 14326 }, { "epoch": 1.71, "learning_rate": 2.0671457743483507e-06, "loss": 0.7252, "step": 14327 }, { "epoch": 1.71, "learning_rate": 2.066829208973674e-06, "loss": 0.9171, "step": 14328 }, { "epoch": 1.71, "learning_rate": 2.0665126507593485e-06, "loss": 0.4115, "step": 14329 }, { "epoch": 1.71, "learning_rate": 2.0661960997106036e-06, "loss": 0.429, "step": 14330 }, { "epoch": 1.71, "learning_rate": 2.0658795558326745e-06, "loss": 0.484, "step": 14331 }, { "epoch": 1.71, "learning_rate": 2.065563019130792e-06, "loss": 0.6201, "step": 14332 }, { "epoch": 1.71, "learning_rate": 2.0652464896101904e-06, "loss": 0.8396, "step": 14333 }, { "epoch": 1.71, "learning_rate": 2.0649299672761004e-06, "loss": 0.8536, "step": 14334 }, { "epoch": 1.71, "learning_rate": 2.064613452133754e-06, "loss": 0.4367, "step": 14335 }, { "epoch": 1.71, "learning_rate": 2.0642969441883846e-06, "loss": 0.5028, "step": 14336 }, { "epoch": 1.71, "learning_rate": 2.0639804434452225e-06, "loss": 0.6685, "step": 14337 }, { "epoch": 1.71, "learning_rate": 2.0636639499095006e-06, "loss": 0.5824, "step": 14338 }, { "epoch": 1.71, "learning_rate": 2.0633474635864493e-06, "loss": 0.848, "step": 14339 }, { "epoch": 1.71, "learning_rate": 2.063030984481303e-06, "loss": 0.8184, "step": 14340 }, { "epoch": 1.71, "learning_rate": 2.0627145125992897e-06, "loss": 0.6703, "step": 14341 }, { "epoch": 1.71, "learning_rate": 2.0623980479456426e-06, "loss": 0.5559, "step": 14342 }, { "epoch": 1.71, "learning_rate": 2.062081590525592e-06, "loss": 0.6161, "step": 14343 }, { "epoch": 1.71, "learning_rate": 2.0617651403443706e-06, "loss": 0.5782, "step": 14344 }, { "epoch": 1.71, "learning_rate": 2.0614486974072075e-06, "loss": 0.8195, "step": 14345 }, { "epoch": 1.71, "learning_rate": 2.061132261719334e-06, "loss": 0.7415, "step": 14346 }, { "epoch": 1.71, "learning_rate": 2.0608158332859807e-06, "loss": 0.7794, "step": 14347 }, { "epoch": 1.71, "learning_rate": 2.0604994121123797e-06, "loss": 0.5616, "step": 14348 }, { "epoch": 1.71, "learning_rate": 2.06018299820376e-06, "loss": 0.4871, "step": 14349 }, { "epoch": 1.71, "learning_rate": 2.0598665915653517e-06, "loss": 0.5778, "step": 14350 }, { "epoch": 1.71, "learning_rate": 2.0595501922023866e-06, "loss": 0.6869, "step": 14351 }, { "epoch": 1.71, "learning_rate": 2.0592338001200922e-06, "loss": 0.7383, "step": 14352 }, { "epoch": 1.71, "learning_rate": 2.058917415323701e-06, "loss": 0.8, "step": 14353 }, { "epoch": 1.71, "learning_rate": 2.0586010378184417e-06, "loss": 0.448, "step": 14354 }, { "epoch": 1.71, "learning_rate": 2.0582846676095446e-06, "loss": 0.5493, "step": 14355 }, { "epoch": 1.71, "learning_rate": 2.0579683047022393e-06, "loss": 0.589, "step": 14356 }, { "epoch": 1.71, "learning_rate": 2.057651949101754e-06, "loss": 0.747, "step": 14357 }, { "epoch": 1.71, "learning_rate": 2.057335600813319e-06, "loss": 0.772, "step": 14358 }, { "epoch": 1.71, "learning_rate": 2.057019259842165e-06, "loss": 0.814, "step": 14359 }, { "epoch": 1.71, "learning_rate": 2.056702926193519e-06, "loss": 0.6248, "step": 14360 }, { "epoch": 1.71, "learning_rate": 2.0563865998726105e-06, "loss": 0.4563, "step": 14361 }, { "epoch": 1.71, "learning_rate": 2.0560702808846695e-06, "loss": 0.5493, "step": 14362 }, { "epoch": 1.71, "learning_rate": 2.055753969234924e-06, "loss": 0.5441, "step": 14363 }, { "epoch": 1.71, "learning_rate": 2.0554376649286023e-06, "loss": 0.6847, "step": 14364 }, { "epoch": 1.71, "learning_rate": 2.055121367970933e-06, "loss": 0.9258, "step": 14365 }, { "epoch": 1.71, "learning_rate": 2.0548050783671454e-06, "loss": 0.5041, "step": 14366 }, { "epoch": 1.71, "learning_rate": 2.054488796122467e-06, "loss": 0.3749, "step": 14367 }, { "epoch": 1.71, "learning_rate": 2.054172521242126e-06, "loss": 0.5136, "step": 14368 }, { "epoch": 1.71, "learning_rate": 2.05385625373135e-06, "loss": 0.6622, "step": 14369 }, { "epoch": 1.71, "learning_rate": 2.0535399935953693e-06, "loss": 0.6604, "step": 14370 }, { "epoch": 1.71, "learning_rate": 2.053223740839409e-06, "loss": 0.6927, "step": 14371 }, { "epoch": 1.71, "learning_rate": 2.0529074954686973e-06, "loss": 0.9073, "step": 14372 }, { "epoch": 1.71, "learning_rate": 2.0525912574884627e-06, "loss": 0.3459, "step": 14373 }, { "epoch": 1.71, "learning_rate": 2.0522750269039325e-06, "loss": 0.5225, "step": 14374 }, { "epoch": 1.71, "learning_rate": 2.051958803720333e-06, "loss": 0.6501, "step": 14375 }, { "epoch": 1.71, "learning_rate": 2.051642587942892e-06, "loss": 0.7474, "step": 14376 }, { "epoch": 1.71, "learning_rate": 2.051326379576837e-06, "loss": 0.835, "step": 14377 }, { "epoch": 1.71, "learning_rate": 2.051010178627395e-06, "loss": 0.8553, "step": 14378 }, { "epoch": 1.71, "learning_rate": 2.0506939850997917e-06, "loss": 0.3012, "step": 14379 }, { "epoch": 1.71, "learning_rate": 2.0503777989992545e-06, "loss": 0.4591, "step": 14380 }, { "epoch": 1.71, "learning_rate": 2.050061620331011e-06, "loss": 0.539, "step": 14381 }, { "epoch": 1.71, "learning_rate": 2.0497454491002862e-06, "loss": 0.6041, "step": 14382 }, { "epoch": 1.71, "learning_rate": 2.049429285312306e-06, "loss": 0.7307, "step": 14383 }, { "epoch": 1.71, "learning_rate": 2.0491131289722983e-06, "loss": 0.8701, "step": 14384 }, { "epoch": 1.71, "learning_rate": 2.048796980085489e-06, "loss": 0.4869, "step": 14385 }, { "epoch": 1.71, "learning_rate": 2.0484808386571024e-06, "loss": 0.5705, "step": 14386 }, { "epoch": 1.71, "learning_rate": 2.0481647046923656e-06, "loss": 0.5069, "step": 14387 }, { "epoch": 1.71, "learning_rate": 2.0478485781965044e-06, "loss": 0.538, "step": 14388 }, { "epoch": 1.71, "learning_rate": 2.0475324591747446e-06, "loss": 0.6857, "step": 14389 }, { "epoch": 1.71, "learning_rate": 2.0472163476323103e-06, "loss": 0.8979, "step": 14390 }, { "epoch": 1.71, "learning_rate": 2.0469002435744285e-06, "loss": 0.6802, "step": 14391 }, { "epoch": 1.71, "learning_rate": 2.046584147006323e-06, "loss": 0.4405, "step": 14392 }, { "epoch": 1.71, "learning_rate": 2.0462680579332205e-06, "loss": 0.6429, "step": 14393 }, { "epoch": 1.71, "learning_rate": 2.045951976360344e-06, "loss": 0.6757, "step": 14394 }, { "epoch": 1.71, "learning_rate": 2.04563590229292e-06, "loss": 0.731, "step": 14395 }, { "epoch": 1.71, "learning_rate": 2.0453198357361726e-06, "loss": 0.8087, "step": 14396 }, { "epoch": 1.71, "learning_rate": 2.0450037766953273e-06, "loss": 0.8441, "step": 14397 }, { "epoch": 1.71, "learning_rate": 2.044687725175606e-06, "loss": 0.4012, "step": 14398 }, { "epoch": 1.71, "learning_rate": 2.044371681182236e-06, "loss": 0.5378, "step": 14399 }, { "epoch": 1.71, "learning_rate": 2.0440556447204406e-06, "loss": 0.6813, "step": 14400 }, { "epoch": 1.72, "learning_rate": 2.043739615795443e-06, "loss": 0.7432, "step": 14401 }, { "epoch": 1.72, "learning_rate": 2.0434235944124682e-06, "loss": 0.7546, "step": 14402 }, { "epoch": 1.72, "learning_rate": 2.0431075805767387e-06, "loss": 0.7825, "step": 14403 }, { "epoch": 1.72, "learning_rate": 2.0427915742934807e-06, "loss": 0.4717, "step": 14404 }, { "epoch": 1.72, "learning_rate": 2.0424755755679156e-06, "loss": 0.429, "step": 14405 }, { "epoch": 1.72, "learning_rate": 2.0421595844052676e-06, "loss": 0.5261, "step": 14406 }, { "epoch": 1.72, "learning_rate": 2.04184360081076e-06, "loss": 0.6363, "step": 14407 }, { "epoch": 1.72, "learning_rate": 2.0415276247896165e-06, "loss": 0.7112, "step": 14408 }, { "epoch": 1.72, "learning_rate": 2.0412116563470594e-06, "loss": 0.7816, "step": 14409 }, { "epoch": 1.72, "learning_rate": 2.0408956954883123e-06, "loss": 0.4738, "step": 14410 }, { "epoch": 1.72, "learning_rate": 2.0405797422185976e-06, "loss": 0.4083, "step": 14411 }, { "epoch": 1.72, "learning_rate": 2.0402637965431393e-06, "loss": 0.5884, "step": 14412 }, { "epoch": 1.72, "learning_rate": 2.0399478584671585e-06, "loss": 0.7011, "step": 14413 }, { "epoch": 1.72, "learning_rate": 2.0396319279958776e-06, "loss": 0.7668, "step": 14414 }, { "epoch": 1.72, "learning_rate": 2.0393160051345207e-06, "loss": 0.8251, "step": 14415 }, { "epoch": 1.72, "learning_rate": 2.0390000898883077e-06, "loss": 0.5904, "step": 14416 }, { "epoch": 1.72, "learning_rate": 2.038684182262462e-06, "loss": 0.4086, "step": 14417 }, { "epoch": 1.72, "learning_rate": 2.038368282262206e-06, "loss": 0.5473, "step": 14418 }, { "epoch": 1.72, "learning_rate": 2.0380523898927613e-06, "loss": 0.6053, "step": 14419 }, { "epoch": 1.72, "learning_rate": 2.037736505159349e-06, "loss": 0.6805, "step": 14420 }, { "epoch": 1.72, "learning_rate": 2.0374206280671905e-06, "loss": 0.8858, "step": 14421 }, { "epoch": 1.72, "learning_rate": 2.037104758621508e-06, "loss": 0.8293, "step": 14422 }, { "epoch": 1.72, "learning_rate": 2.0367888968275237e-06, "loss": 0.5537, "step": 14423 }, { "epoch": 1.72, "learning_rate": 2.0364730426904574e-06, "loss": 0.5359, "step": 14424 }, { "epoch": 1.72, "learning_rate": 2.0361571962155295e-06, "loss": 0.6013, "step": 14425 }, { "epoch": 1.72, "learning_rate": 2.035841357407963e-06, "loss": 0.6478, "step": 14426 }, { "epoch": 1.72, "learning_rate": 2.035525526272978e-06, "loss": 0.7719, "step": 14427 }, { "epoch": 1.72, "learning_rate": 2.0352097028157946e-06, "loss": 0.8674, "step": 14428 }, { "epoch": 1.72, "learning_rate": 2.034893887041634e-06, "loss": 0.3741, "step": 14429 }, { "epoch": 1.72, "learning_rate": 2.034578078955717e-06, "loss": 0.4577, "step": 14430 }, { "epoch": 1.72, "learning_rate": 2.034262278563263e-06, "loss": 0.5121, "step": 14431 }, { "epoch": 1.72, "learning_rate": 2.0339464858694917e-06, "loss": 0.701, "step": 14432 }, { "epoch": 1.72, "learning_rate": 2.033630700879624e-06, "loss": 0.7499, "step": 14433 }, { "epoch": 1.72, "learning_rate": 2.0333149235988816e-06, "loss": 0.8469, "step": 14434 }, { "epoch": 1.72, "learning_rate": 2.032999154032482e-06, "loss": 0.3881, "step": 14435 }, { "epoch": 1.72, "learning_rate": 2.032683392185645e-06, "loss": 0.4765, "step": 14436 }, { "epoch": 1.72, "learning_rate": 2.0323676380635905e-06, "loss": 0.5679, "step": 14437 }, { "epoch": 1.72, "learning_rate": 2.0320518916715397e-06, "loss": 0.7135, "step": 14438 }, { "epoch": 1.72, "learning_rate": 2.0317361530147097e-06, "loss": 0.7314, "step": 14439 }, { "epoch": 1.72, "learning_rate": 2.0314204220983197e-06, "loss": 0.8694, "step": 14440 }, { "epoch": 1.72, "learning_rate": 2.03110469892759e-06, "loss": 0.6898, "step": 14441 }, { "epoch": 1.72, "learning_rate": 2.030788983507739e-06, "loss": 0.4078, "step": 14442 }, { "epoch": 1.72, "learning_rate": 2.0304732758439853e-06, "loss": 0.4725, "step": 14443 }, { "epoch": 1.72, "learning_rate": 2.0301575759415474e-06, "loss": 0.5869, "step": 14444 }, { "epoch": 1.72, "learning_rate": 2.0298418838056452e-06, "loss": 0.6905, "step": 14445 }, { "epoch": 1.72, "learning_rate": 2.0295261994414957e-06, "loss": 0.8272, "step": 14446 }, { "epoch": 1.72, "learning_rate": 2.0292105228543173e-06, "loss": 1.0007, "step": 14447 }, { "epoch": 1.72, "learning_rate": 2.028894854049328e-06, "loss": 0.4623, "step": 14448 }, { "epoch": 1.72, "learning_rate": 2.0285791930317474e-06, "loss": 0.4753, "step": 14449 }, { "epoch": 1.72, "learning_rate": 2.028263539806792e-06, "loss": 0.548, "step": 14450 }, { "epoch": 1.72, "learning_rate": 2.027947894379679e-06, "loss": 0.6376, "step": 14451 }, { "epoch": 1.72, "learning_rate": 2.027632256755628e-06, "loss": 0.8275, "step": 14452 }, { "epoch": 1.72, "learning_rate": 2.0273166269398555e-06, "loss": 0.8428, "step": 14453 }, { "epoch": 1.72, "learning_rate": 2.027001004937578e-06, "loss": 0.5455, "step": 14454 }, { "epoch": 1.72, "learning_rate": 2.0266853907540132e-06, "loss": 0.5227, "step": 14455 }, { "epoch": 1.72, "learning_rate": 2.0263697843943792e-06, "loss": 0.606, "step": 14456 }, { "epoch": 1.72, "learning_rate": 2.0260541858638932e-06, "loss": 0.6253, "step": 14457 }, { "epoch": 1.72, "learning_rate": 2.0257385951677704e-06, "loss": 0.7496, "step": 14458 }, { "epoch": 1.72, "learning_rate": 2.0254230123112284e-06, "loss": 0.8609, "step": 14459 }, { "epoch": 1.72, "learning_rate": 2.0251074372994847e-06, "loss": 0.3729, "step": 14460 }, { "epoch": 1.72, "learning_rate": 2.024791870137754e-06, "loss": 0.488, "step": 14461 }, { "epoch": 1.72, "learning_rate": 2.0244763108312532e-06, "loss": 0.6172, "step": 14462 }, { "epoch": 1.72, "learning_rate": 2.0241607593851993e-06, "loss": 0.6333, "step": 14463 }, { "epoch": 1.72, "learning_rate": 2.023845215804809e-06, "loss": 0.6561, "step": 14464 }, { "epoch": 1.72, "learning_rate": 2.0235296800952957e-06, "loss": 0.8061, "step": 14465 }, { "epoch": 1.72, "learning_rate": 2.023214152261877e-06, "loss": 0.7086, "step": 14466 }, { "epoch": 1.72, "learning_rate": 2.022898632309769e-06, "loss": 0.4592, "step": 14467 }, { "epoch": 1.72, "learning_rate": 2.022583120244187e-06, "loss": 0.6505, "step": 14468 }, { "epoch": 1.72, "learning_rate": 2.022267616070345e-06, "loss": 0.5924, "step": 14469 }, { "epoch": 1.72, "learning_rate": 2.02195211979346e-06, "loss": 0.5849, "step": 14470 }, { "epoch": 1.72, "learning_rate": 2.0216366314187462e-06, "loss": 0.821, "step": 14471 }, { "epoch": 1.72, "learning_rate": 2.02132115095142e-06, "loss": 0.8011, "step": 14472 }, { "epoch": 1.72, "learning_rate": 2.021005678396694e-06, "loss": 0.3696, "step": 14473 }, { "epoch": 1.72, "learning_rate": 2.020690213759785e-06, "loss": 0.5021, "step": 14474 }, { "epoch": 1.72, "learning_rate": 2.0203747570459066e-06, "loss": 0.5642, "step": 14475 }, { "epoch": 1.72, "learning_rate": 2.0200593082602747e-06, "loss": 0.6201, "step": 14476 }, { "epoch": 1.72, "learning_rate": 2.0197438674081025e-06, "loss": 0.7618, "step": 14477 }, { "epoch": 1.72, "learning_rate": 2.0194284344946034e-06, "loss": 0.7674, "step": 14478 }, { "epoch": 1.72, "learning_rate": 2.019113009524994e-06, "loss": 0.5311, "step": 14479 }, { "epoch": 1.72, "learning_rate": 2.0187975925044854e-06, "loss": 0.6026, "step": 14480 }, { "epoch": 1.72, "learning_rate": 2.018482183438294e-06, "loss": 0.6088, "step": 14481 }, { "epoch": 1.72, "learning_rate": 2.018166782331632e-06, "loss": 0.666, "step": 14482 }, { "epoch": 1.72, "learning_rate": 2.017851389189714e-06, "loss": 0.851, "step": 14483 }, { "epoch": 1.72, "learning_rate": 2.017536004017753e-06, "loss": 0.7645, "step": 14484 }, { "epoch": 1.73, "learning_rate": 2.0172206268209617e-06, "loss": 0.4546, "step": 14485 }, { "epoch": 1.73, "learning_rate": 2.016905257604554e-06, "loss": 0.4857, "step": 14486 }, { "epoch": 1.73, "learning_rate": 2.0165898963737436e-06, "loss": 0.5376, "step": 14487 }, { "epoch": 1.73, "learning_rate": 2.016274543133743e-06, "loss": 0.6304, "step": 14488 }, { "epoch": 1.73, "learning_rate": 2.0159591978897636e-06, "loss": 0.7443, "step": 14489 }, { "epoch": 1.73, "learning_rate": 2.0156438606470193e-06, "loss": 0.7832, "step": 14490 }, { "epoch": 1.73, "learning_rate": 2.015328531410724e-06, "loss": 0.576, "step": 14491 }, { "epoch": 1.73, "learning_rate": 2.015013210186088e-06, "loss": 0.3937, "step": 14492 }, { "epoch": 1.73, "learning_rate": 2.014697896978324e-06, "loss": 0.485, "step": 14493 }, { "epoch": 1.73, "learning_rate": 2.014382591792645e-06, "loss": 0.6771, "step": 14494 }, { "epoch": 1.73, "learning_rate": 2.014067294634262e-06, "loss": 0.8001, "step": 14495 }, { "epoch": 1.73, "learning_rate": 2.013752005508388e-06, "loss": 0.8166, "step": 14496 }, { "epoch": 1.73, "learning_rate": 2.013436724420233e-06, "loss": 0.9468, "step": 14497 }, { "epoch": 1.73, "learning_rate": 2.0131214513750114e-06, "loss": 0.4145, "step": 14498 }, { "epoch": 1.73, "learning_rate": 2.0128061863779324e-06, "loss": 0.5371, "step": 14499 }, { "epoch": 1.73, "learning_rate": 2.0124909294342075e-06, "loss": 0.6526, "step": 14500 }, { "epoch": 1.73, "learning_rate": 2.0121756805490483e-06, "loss": 0.649, "step": 14501 }, { "epoch": 1.73, "learning_rate": 2.011860439727667e-06, "loss": 0.6917, "step": 14502 }, { "epoch": 1.73, "learning_rate": 2.0115452069752733e-06, "loss": 0.9487, "step": 14503 }, { "epoch": 1.73, "learning_rate": 2.0112299822970778e-06, "loss": 0.575, "step": 14504 }, { "epoch": 1.73, "learning_rate": 2.010914765698292e-06, "loss": 0.5256, "step": 14505 }, { "epoch": 1.73, "learning_rate": 2.010599557184126e-06, "loss": 0.5326, "step": 14506 }, { "epoch": 1.73, "learning_rate": 2.010284356759791e-06, "loss": 0.8137, "step": 14507 }, { "epoch": 1.73, "learning_rate": 2.009969164430496e-06, "loss": 0.7758, "step": 14508 }, { "epoch": 1.73, "learning_rate": 2.0096539802014523e-06, "loss": 0.8055, "step": 14509 }, { "epoch": 1.73, "learning_rate": 2.0093388040778696e-06, "loss": 0.4626, "step": 14510 }, { "epoch": 1.73, "learning_rate": 2.0090236360649568e-06, "loss": 0.495, "step": 14511 }, { "epoch": 1.73, "learning_rate": 2.0087084761679245e-06, "loss": 0.5987, "step": 14512 }, { "epoch": 1.73, "learning_rate": 2.0083933243919836e-06, "loss": 0.6849, "step": 14513 }, { "epoch": 1.73, "learning_rate": 2.0080781807423413e-06, "loss": 0.6972, "step": 14514 }, { "epoch": 1.73, "learning_rate": 2.0077630452242077e-06, "loss": 0.8421, "step": 14515 }, { "epoch": 1.73, "learning_rate": 2.0074479178427932e-06, "loss": 0.6645, "step": 14516 }, { "epoch": 1.73, "learning_rate": 2.0071327986033056e-06, "loss": 0.546, "step": 14517 }, { "epoch": 1.73, "learning_rate": 2.006817687510954e-06, "loss": 0.5074, "step": 14518 }, { "epoch": 1.73, "learning_rate": 2.0065025845709466e-06, "loss": 0.6489, "step": 14519 }, { "epoch": 1.73, "learning_rate": 2.006187489788494e-06, "loss": 0.7678, "step": 14520 }, { "epoch": 1.73, "learning_rate": 2.005872403168804e-06, "loss": 0.8134, "step": 14521 }, { "epoch": 1.73, "learning_rate": 2.0055573247170834e-06, "loss": 0.8744, "step": 14522 }, { "epoch": 1.73, "learning_rate": 2.005242254438542e-06, "loss": 0.5718, "step": 14523 }, { "epoch": 1.73, "learning_rate": 2.0049271923383884e-06, "loss": 0.5106, "step": 14524 }, { "epoch": 1.73, "learning_rate": 2.0046121384218296e-06, "loss": 0.5344, "step": 14525 }, { "epoch": 1.73, "learning_rate": 2.004297092694073e-06, "loss": 0.6771, "step": 14526 }, { "epoch": 1.73, "learning_rate": 2.0039820551603277e-06, "loss": 0.7936, "step": 14527 }, { "epoch": 1.73, "learning_rate": 2.0036670258258013e-06, "loss": 0.8304, "step": 14528 }, { "epoch": 1.73, "learning_rate": 2.0033520046956996e-06, "loss": 0.4405, "step": 14529 }, { "epoch": 1.73, "learning_rate": 2.0030369917752306e-06, "loss": 0.5709, "step": 14530 }, { "epoch": 1.73, "learning_rate": 2.0027219870696025e-06, "loss": 0.6216, "step": 14531 }, { "epoch": 1.73, "learning_rate": 2.0024069905840223e-06, "loss": 0.5897, "step": 14532 }, { "epoch": 1.73, "learning_rate": 2.0020920023236948e-06, "loss": 0.7199, "step": 14533 }, { "epoch": 1.73, "learning_rate": 2.0017770222938293e-06, "loss": 0.7391, "step": 14534 }, { "epoch": 1.73, "learning_rate": 2.001462050499631e-06, "loss": 0.4367, "step": 14535 }, { "epoch": 1.73, "learning_rate": 2.0011470869463074e-06, "loss": 0.5544, "step": 14536 }, { "epoch": 1.73, "learning_rate": 2.0008321316390632e-06, "loss": 0.6832, "step": 14537 }, { "epoch": 1.73, "learning_rate": 2.0005171845831066e-06, "loss": 0.4806, "step": 14538 }, { "epoch": 1.73, "learning_rate": 2.000202245783642e-06, "loss": 0.7266, "step": 14539 }, { "epoch": 1.73, "learning_rate": 1.999887315245878e-06, "loss": 0.8229, "step": 14540 }, { "epoch": 1.73, "learning_rate": 1.9995723929750168e-06, "loss": 0.5551, "step": 14541 }, { "epoch": 1.73, "learning_rate": 1.999257478976267e-06, "loss": 0.5227, "step": 14542 }, { "epoch": 1.73, "learning_rate": 1.9989425732548335e-06, "loss": 0.6133, "step": 14543 }, { "epoch": 1.73, "learning_rate": 1.99862767581592e-06, "loss": 0.6133, "step": 14544 }, { "epoch": 1.73, "learning_rate": 1.998312786664734e-06, "loss": 0.629, "step": 14545 }, { "epoch": 1.73, "learning_rate": 1.9979979058064784e-06, "loss": 0.87, "step": 14546 }, { "epoch": 1.73, "learning_rate": 1.9976830332463614e-06, "loss": 0.717, "step": 14547 }, { "epoch": 1.73, "learning_rate": 1.997368168989585e-06, "loss": 0.3252, "step": 14548 }, { "epoch": 1.73, "learning_rate": 1.997053313041355e-06, "loss": 0.5351, "step": 14549 }, { "epoch": 1.73, "learning_rate": 1.9967384654068756e-06, "loss": 0.4991, "step": 14550 }, { "epoch": 1.73, "learning_rate": 1.996423626091353e-06, "loss": 0.7515, "step": 14551 }, { "epoch": 1.73, "learning_rate": 1.996108795099988e-06, "loss": 0.8321, "step": 14552 }, { "epoch": 1.73, "learning_rate": 1.9957939724379886e-06, "loss": 0.8676, "step": 14553 }, { "epoch": 1.73, "learning_rate": 1.9954791581105558e-06, "loss": 0.6051, "step": 14554 }, { "epoch": 1.73, "learning_rate": 1.9951643521228963e-06, "loss": 0.6192, "step": 14555 }, { "epoch": 1.73, "learning_rate": 1.994849554480212e-06, "loss": 0.5827, "step": 14556 }, { "epoch": 1.73, "learning_rate": 1.9945347651877062e-06, "loss": 0.5255, "step": 14557 }, { "epoch": 1.73, "learning_rate": 1.9942199842505845e-06, "loss": 0.6983, "step": 14558 }, { "epoch": 1.73, "learning_rate": 1.9939052116740477e-06, "loss": 0.9115, "step": 14559 }, { "epoch": 1.73, "learning_rate": 1.9935904474633004e-06, "loss": 0.5703, "step": 14560 }, { "epoch": 1.73, "learning_rate": 1.993275691623545e-06, "loss": 0.4326, "step": 14561 }, { "epoch": 1.73, "learning_rate": 1.9929609441599863e-06, "loss": 0.5729, "step": 14562 }, { "epoch": 1.73, "learning_rate": 1.992646205077825e-06, "loss": 0.6772, "step": 14563 }, { "epoch": 1.73, "learning_rate": 1.9923314743822643e-06, "loss": 0.8442, "step": 14564 }, { "epoch": 1.73, "learning_rate": 1.9920167520785068e-06, "loss": 0.8486, "step": 14565 }, { "epoch": 1.73, "learning_rate": 1.991702038171756e-06, "loss": 0.4458, "step": 14566 }, { "epoch": 1.73, "learning_rate": 1.9913873326672124e-06, "loss": 0.4748, "step": 14567 }, { "epoch": 1.73, "learning_rate": 1.991072635570079e-06, "loss": 0.6177, "step": 14568 }, { "epoch": 1.74, "learning_rate": 1.990757946885557e-06, "loss": 0.5397, "step": 14569 }, { "epoch": 1.74, "learning_rate": 1.99044326661885e-06, "loss": 0.7631, "step": 14570 }, { "epoch": 1.74, "learning_rate": 1.990128594775158e-06, "loss": 0.806, "step": 14571 }, { "epoch": 1.74, "learning_rate": 1.989813931359683e-06, "loss": 0.618, "step": 14572 }, { "epoch": 1.74, "learning_rate": 1.9894992763776273e-06, "loss": 0.4315, "step": 14573 }, { "epoch": 1.74, "learning_rate": 1.9891846298341906e-06, "loss": 0.6316, "step": 14574 }, { "epoch": 1.74, "learning_rate": 1.9888699917345747e-06, "loss": 0.6222, "step": 14575 }, { "epoch": 1.74, "learning_rate": 1.9885553620839807e-06, "loss": 0.5703, "step": 14576 }, { "epoch": 1.74, "learning_rate": 1.9882407408876103e-06, "loss": 0.7476, "step": 14577 }, { "epoch": 1.74, "learning_rate": 1.987926128150663e-06, "loss": 0.844, "step": 14578 }, { "epoch": 1.74, "learning_rate": 1.9876115238783395e-06, "loss": 0.494, "step": 14579 }, { "epoch": 1.74, "learning_rate": 1.98729692807584e-06, "loss": 0.6718, "step": 14580 }, { "epoch": 1.74, "learning_rate": 1.9869823407483664e-06, "loss": 0.7459, "step": 14581 }, { "epoch": 1.74, "learning_rate": 1.9866677619011175e-06, "loss": 0.6804, "step": 14582 }, { "epoch": 1.74, "learning_rate": 1.986353191539293e-06, "loss": 0.8799, "step": 14583 }, { "epoch": 1.74, "learning_rate": 1.986038629668094e-06, "loss": 0.8463, "step": 14584 }, { "epoch": 1.74, "learning_rate": 1.9857240762927197e-06, "loss": 0.5995, "step": 14585 }, { "epoch": 1.74, "learning_rate": 1.9854095314183687e-06, "loss": 0.6143, "step": 14586 }, { "epoch": 1.74, "learning_rate": 1.9850949950502413e-06, "loss": 0.5513, "step": 14587 }, { "epoch": 1.74, "learning_rate": 1.984780467193538e-06, "loss": 0.648, "step": 14588 }, { "epoch": 1.74, "learning_rate": 1.9844659478534565e-06, "loss": 0.7335, "step": 14589 }, { "epoch": 1.74, "learning_rate": 1.9841514370351954e-06, "loss": 0.8521, "step": 14590 }, { "epoch": 1.74, "learning_rate": 1.983836934743955e-06, "loss": 0.509, "step": 14591 }, { "epoch": 1.74, "learning_rate": 1.983522440984933e-06, "loss": 0.5562, "step": 14592 }, { "epoch": 1.74, "learning_rate": 1.9832079557633284e-06, "loss": 0.5299, "step": 14593 }, { "epoch": 1.74, "learning_rate": 1.9828934790843393e-06, "loss": 0.5446, "step": 14594 }, { "epoch": 1.74, "learning_rate": 1.982579010953165e-06, "loss": 0.6485, "step": 14595 }, { "epoch": 1.74, "learning_rate": 1.9822645513750034e-06, "loss": 0.7778, "step": 14596 }, { "epoch": 1.74, "learning_rate": 1.9819501003550513e-06, "loss": 0.825, "step": 14597 }, { "epoch": 1.74, "learning_rate": 1.9816356578985074e-06, "loss": 0.413, "step": 14598 }, { "epoch": 1.74, "learning_rate": 1.9813212240105704e-06, "loss": 0.4662, "step": 14599 }, { "epoch": 1.74, "learning_rate": 1.9810067986964373e-06, "loss": 0.7425, "step": 14600 }, { "epoch": 1.74, "learning_rate": 1.9806923819613046e-06, "loss": 0.6613, "step": 14601 }, { "epoch": 1.74, "learning_rate": 1.9803779738103706e-06, "loss": 0.8185, "step": 14602 }, { "epoch": 1.74, "learning_rate": 1.9800635742488317e-06, "loss": 0.7855, "step": 14603 }, { "epoch": 1.74, "learning_rate": 1.9797491832818867e-06, "loss": 0.477, "step": 14604 }, { "epoch": 1.74, "learning_rate": 1.9794348009147304e-06, "loss": 0.447, "step": 14605 }, { "epoch": 1.74, "learning_rate": 1.979120427152561e-06, "loss": 0.5812, "step": 14606 }, { "epoch": 1.74, "learning_rate": 1.9788060620005753e-06, "loss": 0.6398, "step": 14607 }, { "epoch": 1.74, "learning_rate": 1.978491705463968e-06, "loss": 0.7252, "step": 14608 }, { "epoch": 1.74, "learning_rate": 1.9781773575479366e-06, "loss": 0.8342, "step": 14609 }, { "epoch": 1.74, "learning_rate": 1.9778630182576776e-06, "loss": 0.4497, "step": 14610 }, { "epoch": 1.74, "learning_rate": 1.9775486875983875e-06, "loss": 0.4267, "step": 14611 }, { "epoch": 1.74, "learning_rate": 1.9772343655752602e-06, "loss": 0.5057, "step": 14612 }, { "epoch": 1.74, "learning_rate": 1.9769200521934932e-06, "loss": 0.6338, "step": 14613 }, { "epoch": 1.74, "learning_rate": 1.9766057474582813e-06, "loss": 0.7154, "step": 14614 }, { "epoch": 1.74, "learning_rate": 1.976291451374821e-06, "loss": 0.8977, "step": 14615 }, { "epoch": 1.74, "learning_rate": 1.975977163948306e-06, "loss": 0.4326, "step": 14616 }, { "epoch": 1.74, "learning_rate": 1.975662885183933e-06, "loss": 0.4716, "step": 14617 }, { "epoch": 1.74, "learning_rate": 1.9753486150868965e-06, "loss": 0.541, "step": 14618 }, { "epoch": 1.74, "learning_rate": 1.975034353662392e-06, "loss": 0.5692, "step": 14619 }, { "epoch": 1.74, "learning_rate": 1.974720100915613e-06, "loss": 0.6278, "step": 14620 }, { "epoch": 1.74, "learning_rate": 1.9744058568517545e-06, "loss": 0.7009, "step": 14621 }, { "epoch": 1.74, "learning_rate": 1.9740916214760125e-06, "loss": 0.8412, "step": 14622 }, { "epoch": 1.74, "learning_rate": 1.973777394793579e-06, "loss": 0.4684, "step": 14623 }, { "epoch": 1.74, "learning_rate": 1.9734631768096496e-06, "loss": 0.5126, "step": 14624 }, { "epoch": 1.74, "learning_rate": 1.9731489675294177e-06, "loss": 0.5219, "step": 14625 }, { "epoch": 1.74, "learning_rate": 1.9728347669580786e-06, "loss": 0.6295, "step": 14626 }, { "epoch": 1.74, "learning_rate": 1.972520575100824e-06, "loss": 0.7298, "step": 14627 }, { "epoch": 1.74, "learning_rate": 1.9722063919628494e-06, "loss": 0.8816, "step": 14628 }, { "epoch": 1.74, "learning_rate": 1.9718922175493463e-06, "loss": 0.4143, "step": 14629 }, { "epoch": 1.74, "learning_rate": 1.9715780518655105e-06, "loss": 0.4977, "step": 14630 }, { "epoch": 1.74, "learning_rate": 1.9712638949165335e-06, "loss": 0.5815, "step": 14631 }, { "epoch": 1.74, "learning_rate": 1.970949746707608e-06, "loss": 0.5577, "step": 14632 }, { "epoch": 1.74, "learning_rate": 1.970635607243927e-06, "loss": 0.706, "step": 14633 }, { "epoch": 1.74, "learning_rate": 1.9703214765306856e-06, "loss": 0.9138, "step": 14634 }, { "epoch": 1.74, "learning_rate": 1.970007354573074e-06, "loss": 0.5056, "step": 14635 }, { "epoch": 1.74, "learning_rate": 1.969693241376285e-06, "loss": 0.4252, "step": 14636 }, { "epoch": 1.74, "learning_rate": 1.969379136945512e-06, "loss": 0.5074, "step": 14637 }, { "epoch": 1.74, "learning_rate": 1.969065041285945e-06, "loss": 0.5925, "step": 14638 }, { "epoch": 1.74, "learning_rate": 1.9687509544027784e-06, "loss": 0.6606, "step": 14639 }, { "epoch": 1.74, "learning_rate": 1.9684368763012024e-06, "loss": 0.8567, "step": 14640 }, { "epoch": 1.74, "learning_rate": 1.9681228069864107e-06, "loss": 0.6175, "step": 14641 }, { "epoch": 1.74, "learning_rate": 1.9678087464635923e-06, "loss": 0.3508, "step": 14642 }, { "epoch": 1.74, "learning_rate": 1.9674946947379404e-06, "loss": 0.5478, "step": 14643 }, { "epoch": 1.74, "learning_rate": 1.9671806518146456e-06, "loss": 0.6059, "step": 14644 }, { "epoch": 1.74, "learning_rate": 1.9668666176989005e-06, "loss": 0.6275, "step": 14645 }, { "epoch": 1.74, "learning_rate": 1.966552592395894e-06, "loss": 0.7211, "step": 14646 }, { "epoch": 1.74, "learning_rate": 1.9662385759108177e-06, "loss": 0.791, "step": 14647 }, { "epoch": 1.74, "learning_rate": 1.965924568248863e-06, "loss": 0.4032, "step": 14648 }, { "epoch": 1.74, "learning_rate": 1.96561056941522e-06, "loss": 0.5367, "step": 14649 }, { "epoch": 1.74, "learning_rate": 1.9652965794150792e-06, "loss": 0.6387, "step": 14650 }, { "epoch": 1.74, "learning_rate": 1.96498259825363e-06, "loss": 0.6246, "step": 14651 }, { "epoch": 1.74, "learning_rate": 1.9646686259360646e-06, "loss": 0.7283, "step": 14652 }, { "epoch": 1.75, "learning_rate": 1.964354662467571e-06, "loss": 0.8416, "step": 14653 }, { "epoch": 1.75, "learning_rate": 1.9640407078533397e-06, "loss": 0.5051, "step": 14654 }, { "epoch": 1.75, "learning_rate": 1.96372676209856e-06, "loss": 0.4815, "step": 14655 }, { "epoch": 1.75, "learning_rate": 1.9634128252084234e-06, "loss": 0.5626, "step": 14656 }, { "epoch": 1.75, "learning_rate": 1.9630988971881167e-06, "loss": 0.5885, "step": 14657 }, { "epoch": 1.75, "learning_rate": 1.9627849780428302e-06, "loss": 0.6437, "step": 14658 }, { "epoch": 1.75, "learning_rate": 1.9624710677777537e-06, "loss": 0.8042, "step": 14659 }, { "epoch": 1.75, "learning_rate": 1.962157166398076e-06, "loss": 0.4545, "step": 14660 }, { "epoch": 1.75, "learning_rate": 1.9618432739089843e-06, "loss": 0.3866, "step": 14661 }, { "epoch": 1.75, "learning_rate": 1.9615293903156683e-06, "loss": 0.6715, "step": 14662 }, { "epoch": 1.75, "learning_rate": 1.961215515623317e-06, "loss": 0.5674, "step": 14663 }, { "epoch": 1.75, "learning_rate": 1.96090164983712e-06, "loss": 0.6998, "step": 14664 }, { "epoch": 1.75, "learning_rate": 1.9605877929622617e-06, "loss": 0.7987, "step": 14665 }, { "epoch": 1.75, "learning_rate": 1.960273945003933e-06, "loss": 0.5987, "step": 14666 }, { "epoch": 1.75, "learning_rate": 1.9599601059673225e-06, "loss": 0.4412, "step": 14667 }, { "epoch": 1.75, "learning_rate": 1.959646275857616e-06, "loss": 0.5769, "step": 14668 }, { "epoch": 1.75, "learning_rate": 1.9593324546800012e-06, "loss": 0.5566, "step": 14669 }, { "epoch": 1.75, "learning_rate": 1.959018642439667e-06, "loss": 0.7949, "step": 14670 }, { "epoch": 1.75, "learning_rate": 1.958704839141801e-06, "loss": 0.8294, "step": 14671 }, { "epoch": 1.75, "learning_rate": 1.9583910447915877e-06, "loss": 0.7422, "step": 14672 }, { "epoch": 1.75, "learning_rate": 1.958077259394216e-06, "loss": 0.4595, "step": 14673 }, { "epoch": 1.75, "learning_rate": 1.957763482954874e-06, "loss": 0.5918, "step": 14674 }, { "epoch": 1.75, "learning_rate": 1.9574497154787465e-06, "loss": 0.5826, "step": 14675 }, { "epoch": 1.75, "learning_rate": 1.95713595697102e-06, "loss": 0.7119, "step": 14676 }, { "epoch": 1.75, "learning_rate": 1.9568222074368825e-06, "loss": 0.822, "step": 14677 }, { "epoch": 1.75, "learning_rate": 1.956508466881519e-06, "loss": 0.7423, "step": 14678 }, { "epoch": 1.75, "learning_rate": 1.9561947353101166e-06, "loss": 0.4442, "step": 14679 }, { "epoch": 1.75, "learning_rate": 1.9558810127278604e-06, "loss": 0.4389, "step": 14680 }, { "epoch": 1.75, "learning_rate": 1.955567299139937e-06, "loss": 0.6499, "step": 14681 }, { "epoch": 1.75, "learning_rate": 1.9552535945515313e-06, "loss": 0.5931, "step": 14682 }, { "epoch": 1.75, "learning_rate": 1.9549398989678307e-06, "loss": 0.7633, "step": 14683 }, { "epoch": 1.75, "learning_rate": 1.954626212394018e-06, "loss": 0.8397, "step": 14684 }, { "epoch": 1.75, "learning_rate": 1.9543125348352804e-06, "loss": 0.6614, "step": 14685 }, { "epoch": 1.75, "learning_rate": 1.953998866296803e-06, "loss": 0.4167, "step": 14686 }, { "epoch": 1.75, "learning_rate": 1.953685206783769e-06, "loss": 0.5391, "step": 14687 }, { "epoch": 1.75, "learning_rate": 1.9533715563013645e-06, "loss": 0.6132, "step": 14688 }, { "epoch": 1.75, "learning_rate": 1.953057914854774e-06, "loss": 0.8055, "step": 14689 }, { "epoch": 1.75, "learning_rate": 1.952744282449183e-06, "loss": 0.8492, "step": 14690 }, { "epoch": 1.75, "learning_rate": 1.952430659089774e-06, "loss": 0.5853, "step": 14691 }, { "epoch": 1.75, "learning_rate": 1.9521170447817325e-06, "loss": 0.4413, "step": 14692 }, { "epoch": 1.75, "learning_rate": 1.9518034395302413e-06, "loss": 0.5221, "step": 14693 }, { "epoch": 1.75, "learning_rate": 1.9514898433404873e-06, "loss": 0.7136, "step": 14694 }, { "epoch": 1.75, "learning_rate": 1.9511762562176505e-06, "loss": 0.6704, "step": 14695 }, { "epoch": 1.75, "learning_rate": 1.950862678166917e-06, "loss": 0.6915, "step": 14696 }, { "epoch": 1.75, "learning_rate": 1.9505491091934686e-06, "loss": 0.8803, "step": 14697 }, { "epoch": 1.75, "learning_rate": 1.950235549302491e-06, "loss": 0.418, "step": 14698 }, { "epoch": 1.75, "learning_rate": 1.949921998499165e-06, "loss": 0.5617, "step": 14699 }, { "epoch": 1.75, "learning_rate": 1.949608456788674e-06, "loss": 0.4998, "step": 14700 }, { "epoch": 1.75, "learning_rate": 1.9492949241762027e-06, "loss": 0.6977, "step": 14701 }, { "epoch": 1.75, "learning_rate": 1.9489814006669315e-06, "loss": 0.7683, "step": 14702 }, { "epoch": 1.75, "learning_rate": 1.9486678862660444e-06, "loss": 0.7628, "step": 14703 }, { "epoch": 1.75, "learning_rate": 1.948354380978723e-06, "loss": 0.4549, "step": 14704 }, { "epoch": 1.75, "learning_rate": 1.9480408848101505e-06, "loss": 0.4777, "step": 14705 }, { "epoch": 1.75, "learning_rate": 1.9477273977655083e-06, "loss": 0.6011, "step": 14706 }, { "epoch": 1.75, "learning_rate": 1.9474139198499785e-06, "loss": 0.6298, "step": 14707 }, { "epoch": 1.75, "learning_rate": 1.9471004510687427e-06, "loss": 0.7169, "step": 14708 }, { "epoch": 1.75, "learning_rate": 1.9467869914269835e-06, "loss": 0.8557, "step": 14709 }, { "epoch": 1.75, "learning_rate": 1.9464735409298813e-06, "loss": 0.3983, "step": 14710 }, { "epoch": 1.75, "learning_rate": 1.9461600995826175e-06, "loss": 0.5374, "step": 14711 }, { "epoch": 1.75, "learning_rate": 1.945846667390374e-06, "loss": 0.5914, "step": 14712 }, { "epoch": 1.75, "learning_rate": 1.9455332443583324e-06, "loss": 0.5982, "step": 14713 }, { "epoch": 1.75, "learning_rate": 1.9452198304916723e-06, "loss": 0.7198, "step": 14714 }, { "epoch": 1.75, "learning_rate": 1.9449064257955745e-06, "loss": 0.996, "step": 14715 }, { "epoch": 1.75, "learning_rate": 1.944593030275221e-06, "loss": 0.516, "step": 14716 }, { "epoch": 1.75, "learning_rate": 1.9442796439357907e-06, "loss": 0.4674, "step": 14717 }, { "epoch": 1.75, "learning_rate": 1.943966266782464e-06, "loss": 0.5073, "step": 14718 }, { "epoch": 1.75, "learning_rate": 1.943652898820422e-06, "loss": 0.6177, "step": 14719 }, { "epoch": 1.75, "learning_rate": 1.9433395400548445e-06, "loss": 0.7557, "step": 14720 }, { "epoch": 1.75, "learning_rate": 1.9430261904909113e-06, "loss": 0.8256, "step": 14721 }, { "epoch": 1.75, "learning_rate": 1.942712850133801e-06, "loss": 0.7356, "step": 14722 }, { "epoch": 1.75, "learning_rate": 1.942399518988694e-06, "loss": 0.4214, "step": 14723 }, { "epoch": 1.75, "learning_rate": 1.9420861970607715e-06, "loss": 0.5406, "step": 14724 }, { "epoch": 1.75, "learning_rate": 1.9417728843552093e-06, "loss": 0.5592, "step": 14725 }, { "epoch": 1.75, "learning_rate": 1.9414595808771885e-06, "loss": 0.6429, "step": 14726 }, { "epoch": 1.75, "learning_rate": 1.941146286631888e-06, "loss": 0.8218, "step": 14727 }, { "epoch": 1.75, "learning_rate": 1.9408330016244868e-06, "loss": 0.894, "step": 14728 }, { "epoch": 1.75, "learning_rate": 1.9405197258601617e-06, "loss": 0.4621, "step": 14729 }, { "epoch": 1.75, "learning_rate": 1.9402064593440926e-06, "loss": 0.5643, "step": 14730 }, { "epoch": 1.75, "learning_rate": 1.939893202081459e-06, "loss": 0.5403, "step": 14731 }, { "epoch": 1.75, "learning_rate": 1.939579954077437e-06, "loss": 0.5596, "step": 14732 }, { "epoch": 1.75, "learning_rate": 1.9392667153372054e-06, "loss": 0.7466, "step": 14733 }, { "epoch": 1.75, "learning_rate": 1.938953485865942e-06, "loss": 0.8715, "step": 14734 }, { "epoch": 1.75, "learning_rate": 1.9386402656688246e-06, "loss": 0.5446, "step": 14735 }, { "epoch": 1.75, "learning_rate": 1.938327054751031e-06, "loss": 0.5009, "step": 14736 }, { "epoch": 1.76, "learning_rate": 1.9380138531177372e-06, "loss": 0.5137, "step": 14737 }, { "epoch": 1.76, "learning_rate": 1.9377006607741224e-06, "loss": 0.6929, "step": 14738 }, { "epoch": 1.76, "learning_rate": 1.937387477725364e-06, "loss": 0.7631, "step": 14739 }, { "epoch": 1.76, "learning_rate": 1.937074303976636e-06, "loss": 0.9316, "step": 14740 }, { "epoch": 1.76, "learning_rate": 1.936761139533117e-06, "loss": 0.5068, "step": 14741 }, { "epoch": 1.76, "learning_rate": 1.9364479843999844e-06, "loss": 0.5106, "step": 14742 }, { "epoch": 1.76, "learning_rate": 1.9361348385824143e-06, "loss": 0.5807, "step": 14743 }, { "epoch": 1.76, "learning_rate": 1.9358217020855816e-06, "loss": 0.5636, "step": 14744 }, { "epoch": 1.76, "learning_rate": 1.935508574914664e-06, "loss": 0.7823, "step": 14745 }, { "epoch": 1.76, "learning_rate": 1.9351954570748367e-06, "loss": 0.8276, "step": 14746 }, { "epoch": 1.76, "learning_rate": 1.9348823485712765e-06, "loss": 0.8229, "step": 14747 }, { "epoch": 1.76, "learning_rate": 1.934569249409158e-06, "loss": 0.3903, "step": 14748 }, { "epoch": 1.76, "learning_rate": 1.934256159593657e-06, "loss": 0.4722, "step": 14749 }, { "epoch": 1.76, "learning_rate": 1.933943079129951e-06, "loss": 0.4695, "step": 14750 }, { "epoch": 1.76, "learning_rate": 1.9336300080232113e-06, "loss": 0.5034, "step": 14751 }, { "epoch": 1.76, "learning_rate": 1.933316946278615e-06, "loss": 0.7358, "step": 14752 }, { "epoch": 1.76, "learning_rate": 1.9330038939013383e-06, "loss": 0.9425, "step": 14753 }, { "epoch": 1.76, "learning_rate": 1.932690850896555e-06, "loss": 0.5447, "step": 14754 }, { "epoch": 1.76, "learning_rate": 1.9323778172694384e-06, "loss": 0.4968, "step": 14755 }, { "epoch": 1.76, "learning_rate": 1.9320647930251647e-06, "loss": 0.4814, "step": 14756 }, { "epoch": 1.76, "learning_rate": 1.9317517781689073e-06, "loss": 0.6122, "step": 14757 }, { "epoch": 1.76, "learning_rate": 1.9314387727058413e-06, "loss": 0.7976, "step": 14758 }, { "epoch": 1.76, "learning_rate": 1.9311257766411395e-06, "loss": 0.9514, "step": 14759 }, { "epoch": 1.76, "learning_rate": 1.930812789979977e-06, "loss": 0.4827, "step": 14760 }, { "epoch": 1.76, "learning_rate": 1.9304998127275252e-06, "loss": 0.4713, "step": 14761 }, { "epoch": 1.76, "learning_rate": 1.9301868448889614e-06, "loss": 0.5279, "step": 14762 }, { "epoch": 1.76, "learning_rate": 1.9298738864694556e-06, "loss": 0.579, "step": 14763 }, { "epoch": 1.76, "learning_rate": 1.929560937474183e-06, "loss": 0.783, "step": 14764 }, { "epoch": 1.76, "learning_rate": 1.9292479979083162e-06, "loss": 0.7785, "step": 14765 }, { "epoch": 1.76, "learning_rate": 1.928935067777027e-06, "loss": 0.5496, "step": 14766 }, { "epoch": 1.76, "learning_rate": 1.9286221470854894e-06, "loss": 0.5054, "step": 14767 }, { "epoch": 1.76, "learning_rate": 1.928309235838875e-06, "loss": 0.5515, "step": 14768 }, { "epoch": 1.76, "learning_rate": 1.927996334042358e-06, "loss": 0.6605, "step": 14769 }, { "epoch": 1.76, "learning_rate": 1.927683441701109e-06, "loss": 0.739, "step": 14770 }, { "epoch": 1.76, "learning_rate": 1.9273705588203007e-06, "loss": 0.8112, "step": 14771 }, { "epoch": 1.76, "learning_rate": 1.927057685405105e-06, "loss": 0.8366, "step": 14772 }, { "epoch": 1.76, "learning_rate": 1.9267448214606944e-06, "loss": 0.4303, "step": 14773 }, { "epoch": 1.76, "learning_rate": 1.92643196699224e-06, "loss": 0.5585, "step": 14774 }, { "epoch": 1.76, "learning_rate": 1.9261191220049126e-06, "loss": 0.5575, "step": 14775 }, { "epoch": 1.76, "learning_rate": 1.9258062865038842e-06, "loss": 0.6068, "step": 14776 }, { "epoch": 1.76, "learning_rate": 1.925493460494327e-06, "loss": 0.7273, "step": 14777 }, { "epoch": 1.76, "learning_rate": 1.925180643981411e-06, "loss": 0.8726, "step": 14778 }, { "epoch": 1.76, "learning_rate": 1.924867836970306e-06, "loss": 0.3865, "step": 14779 }, { "epoch": 1.76, "learning_rate": 1.9245550394661856e-06, "loss": 0.6003, "step": 14780 }, { "epoch": 1.76, "learning_rate": 1.9242422514742175e-06, "loss": 0.4891, "step": 14781 }, { "epoch": 1.76, "learning_rate": 1.9239294729995735e-06, "loss": 0.7484, "step": 14782 }, { "epoch": 1.76, "learning_rate": 1.9236167040474234e-06, "loss": 0.7636, "step": 14783 }, { "epoch": 1.76, "learning_rate": 1.9233039446229383e-06, "loss": 0.8146, "step": 14784 }, { "epoch": 1.76, "learning_rate": 1.9229911947312873e-06, "loss": 0.4119, "step": 14785 }, { "epoch": 1.76, "learning_rate": 1.9226784543776393e-06, "loss": 0.4434, "step": 14786 }, { "epoch": 1.76, "learning_rate": 1.922365723567165e-06, "loss": 0.4563, "step": 14787 }, { "epoch": 1.76, "learning_rate": 1.922053002305035e-06, "loss": 0.4838, "step": 14788 }, { "epoch": 1.76, "learning_rate": 1.921740290596417e-06, "loss": 0.7381, "step": 14789 }, { "epoch": 1.76, "learning_rate": 1.9214275884464796e-06, "loss": 0.7845, "step": 14790 }, { "epoch": 1.76, "learning_rate": 1.9211148958603937e-06, "loss": 0.7487, "step": 14791 }, { "epoch": 1.76, "learning_rate": 1.9208022128433266e-06, "loss": 0.3767, "step": 14792 }, { "epoch": 1.76, "learning_rate": 1.920489539400448e-06, "loss": 0.5448, "step": 14793 }, { "epoch": 1.76, "learning_rate": 1.9201768755369248e-06, "loss": 0.6524, "step": 14794 }, { "epoch": 1.76, "learning_rate": 1.919864221257928e-06, "loss": 0.6982, "step": 14795 }, { "epoch": 1.76, "learning_rate": 1.9195515765686237e-06, "loss": 0.77, "step": 14796 }, { "epoch": 1.76, "learning_rate": 1.91923894147418e-06, "loss": 0.7943, "step": 14797 }, { "epoch": 1.76, "learning_rate": 1.9189263159797654e-06, "loss": 0.476, "step": 14798 }, { "epoch": 1.76, "learning_rate": 1.9186137000905485e-06, "loss": 0.4968, "step": 14799 }, { "epoch": 1.76, "learning_rate": 1.918301093811695e-06, "loss": 0.5341, "step": 14800 }, { "epoch": 1.76, "learning_rate": 1.9179884971483737e-06, "loss": 0.6696, "step": 14801 }, { "epoch": 1.76, "learning_rate": 1.917675910105751e-06, "loss": 0.6771, "step": 14802 }, { "epoch": 1.76, "learning_rate": 1.9173633326889953e-06, "loss": 0.8593, "step": 14803 }, { "epoch": 1.76, "learning_rate": 1.9170507649032716e-06, "loss": 0.5553, "step": 14804 }, { "epoch": 1.76, "learning_rate": 1.9167382067537476e-06, "loss": 0.5125, "step": 14805 }, { "epoch": 1.76, "learning_rate": 1.9164256582455905e-06, "loss": 0.6044, "step": 14806 }, { "epoch": 1.76, "learning_rate": 1.9161131193839664e-06, "loss": 0.5485, "step": 14807 }, { "epoch": 1.76, "learning_rate": 1.915800590174041e-06, "loss": 0.6975, "step": 14808 }, { "epoch": 1.76, "learning_rate": 1.9154880706209804e-06, "loss": 0.9397, "step": 14809 }, { "epoch": 1.76, "learning_rate": 1.915175560729952e-06, "loss": 0.5966, "step": 14810 }, { "epoch": 1.76, "learning_rate": 1.9148630605061207e-06, "loss": 0.4419, "step": 14811 }, { "epoch": 1.76, "learning_rate": 1.9145505699546514e-06, "loss": 0.637, "step": 14812 }, { "epoch": 1.76, "learning_rate": 1.9142380890807106e-06, "loss": 0.6603, "step": 14813 }, { "epoch": 1.76, "learning_rate": 1.913925617889464e-06, "loss": 0.7029, "step": 14814 }, { "epoch": 1.76, "learning_rate": 1.9136131563860747e-06, "loss": 0.8499, "step": 14815 }, { "epoch": 1.76, "learning_rate": 1.9133007045757096e-06, "loss": 0.5666, "step": 14816 }, { "epoch": 1.76, "learning_rate": 1.9129882624635334e-06, "loss": 0.5562, "step": 14817 }, { "epoch": 1.76, "learning_rate": 1.912675830054711e-06, "loss": 0.5538, "step": 14818 }, { "epoch": 1.76, "learning_rate": 1.9123634073544054e-06, "loss": 0.5646, "step": 14819 }, { "epoch": 1.76, "learning_rate": 1.9120509943677823e-06, "loss": 0.701, "step": 14820 }, { "epoch": 1.77, "learning_rate": 1.911738591100005e-06, "loss": 0.9713, "step": 14821 }, { "epoch": 1.77, "learning_rate": 1.91142619755624e-06, "loss": 0.8777, "step": 14822 }, { "epoch": 1.77, "learning_rate": 1.9111138137416475e-06, "loss": 0.4399, "step": 14823 }, { "epoch": 1.77, "learning_rate": 1.910801439661394e-06, "loss": 0.4747, "step": 14824 }, { "epoch": 1.77, "learning_rate": 1.9104890753206414e-06, "loss": 0.459, "step": 14825 }, { "epoch": 1.77, "learning_rate": 1.910176720724555e-06, "loss": 0.8598, "step": 14826 }, { "epoch": 1.77, "learning_rate": 1.909864375878296e-06, "loss": 0.7977, "step": 14827 }, { "epoch": 1.77, "learning_rate": 1.9095520407870287e-06, "loss": 0.8408, "step": 14828 }, { "epoch": 1.77, "learning_rate": 1.909239715455916e-06, "loss": 0.4276, "step": 14829 }, { "epoch": 1.77, "learning_rate": 1.90892739989012e-06, "loss": 0.4357, "step": 14830 }, { "epoch": 1.77, "learning_rate": 1.9086150940948035e-06, "loss": 0.5495, "step": 14831 }, { "epoch": 1.77, "learning_rate": 1.9083027980751286e-06, "loss": 0.5128, "step": 14832 }, { "epoch": 1.77, "learning_rate": 1.9079905118362597e-06, "loss": 0.7961, "step": 14833 }, { "epoch": 1.77, "learning_rate": 1.907678235383356e-06, "loss": 0.8159, "step": 14834 }, { "epoch": 1.77, "learning_rate": 1.9073659687215814e-06, "loss": 0.3853, "step": 14835 }, { "epoch": 1.77, "learning_rate": 1.907053711856096e-06, "loss": 0.4458, "step": 14836 }, { "epoch": 1.77, "learning_rate": 1.906741464792064e-06, "loss": 0.4793, "step": 14837 }, { "epoch": 1.77, "learning_rate": 1.9064292275346443e-06, "loss": 0.6677, "step": 14838 }, { "epoch": 1.77, "learning_rate": 1.906117000089e-06, "loss": 0.8167, "step": 14839 }, { "epoch": 1.77, "learning_rate": 1.9058047824602904e-06, "loss": 0.8776, "step": 14840 }, { "epoch": 1.77, "learning_rate": 1.9054925746536792e-06, "loss": 0.6046, "step": 14841 }, { "epoch": 1.77, "learning_rate": 1.9051803766743248e-06, "loss": 0.332, "step": 14842 }, { "epoch": 1.77, "learning_rate": 1.9048681885273885e-06, "loss": 0.5572, "step": 14843 }, { "epoch": 1.77, "learning_rate": 1.9045560102180316e-06, "loss": 0.5284, "step": 14844 }, { "epoch": 1.77, "learning_rate": 1.9042438417514131e-06, "loss": 0.6564, "step": 14845 }, { "epoch": 1.77, "learning_rate": 1.9039316831326942e-06, "loss": 0.8302, "step": 14846 }, { "epoch": 1.77, "learning_rate": 1.9036195343670342e-06, "loss": 0.9004, "step": 14847 }, { "epoch": 1.77, "learning_rate": 1.9033073954595942e-06, "loss": 0.3719, "step": 14848 }, { "epoch": 1.77, "learning_rate": 1.9029952664155321e-06, "loss": 0.5311, "step": 14849 }, { "epoch": 1.77, "learning_rate": 1.9026831472400089e-06, "loss": 0.6231, "step": 14850 }, { "epoch": 1.77, "learning_rate": 1.9023710379381829e-06, "loss": 0.5669, "step": 14851 }, { "epoch": 1.77, "learning_rate": 1.9020589385152149e-06, "loss": 0.8834, "step": 14852 }, { "epoch": 1.77, "learning_rate": 1.9017468489762618e-06, "loss": 0.8944, "step": 14853 }, { "epoch": 1.77, "learning_rate": 1.9014347693264835e-06, "loss": 0.421, "step": 14854 }, { "epoch": 1.77, "learning_rate": 1.9011226995710387e-06, "loss": 0.5309, "step": 14855 }, { "epoch": 1.77, "learning_rate": 1.9008106397150866e-06, "loss": 0.6133, "step": 14856 }, { "epoch": 1.77, "learning_rate": 1.9004985897637846e-06, "loss": 0.6268, "step": 14857 }, { "epoch": 1.77, "learning_rate": 1.9001865497222907e-06, "loss": 0.7729, "step": 14858 }, { "epoch": 1.77, "learning_rate": 1.8998745195957647e-06, "loss": 0.7943, "step": 14859 }, { "epoch": 1.77, "learning_rate": 1.8995624993893622e-06, "loss": 0.478, "step": 14860 }, { "epoch": 1.77, "learning_rate": 1.8992504891082417e-06, "loss": 0.5135, "step": 14861 }, { "epoch": 1.77, "learning_rate": 1.8989384887575613e-06, "loss": 0.629, "step": 14862 }, { "epoch": 1.77, "learning_rate": 1.8986264983424788e-06, "loss": 0.5825, "step": 14863 }, { "epoch": 1.77, "learning_rate": 1.8983145178681503e-06, "loss": 0.7039, "step": 14864 }, { "epoch": 1.77, "learning_rate": 1.898002547339733e-06, "loss": 0.9037, "step": 14865 }, { "epoch": 1.77, "learning_rate": 1.8976905867623834e-06, "loss": 0.6268, "step": 14866 }, { "epoch": 1.77, "learning_rate": 1.8973786361412608e-06, "loss": 0.4263, "step": 14867 }, { "epoch": 1.77, "learning_rate": 1.897066695481519e-06, "loss": 0.5005, "step": 14868 }, { "epoch": 1.77, "learning_rate": 1.896754764788315e-06, "loss": 0.5713, "step": 14869 }, { "epoch": 1.77, "learning_rate": 1.8964428440668056e-06, "loss": 0.6103, "step": 14870 }, { "epoch": 1.77, "learning_rate": 1.8961309333221472e-06, "loss": 0.8357, "step": 14871 }, { "epoch": 1.77, "learning_rate": 1.8958190325594944e-06, "loss": 0.8538, "step": 14872 }, { "epoch": 1.77, "learning_rate": 1.8955071417840034e-06, "loss": 0.5219, "step": 14873 }, { "epoch": 1.77, "learning_rate": 1.895195261000831e-06, "loss": 0.4614, "step": 14874 }, { "epoch": 1.77, "learning_rate": 1.8948833902151315e-06, "loss": 0.5697, "step": 14875 }, { "epoch": 1.77, "learning_rate": 1.8945715294320599e-06, "loss": 0.7101, "step": 14876 }, { "epoch": 1.77, "learning_rate": 1.8942596786567718e-06, "loss": 0.8262, "step": 14877 }, { "epoch": 1.77, "learning_rate": 1.8939478378944223e-06, "loss": 0.8033, "step": 14878 }, { "epoch": 1.77, "learning_rate": 1.8936360071501658e-06, "loss": 0.4785, "step": 14879 }, { "epoch": 1.77, "learning_rate": 1.8933241864291564e-06, "loss": 0.5605, "step": 14880 }, { "epoch": 1.77, "learning_rate": 1.8930123757365498e-06, "loss": 0.4916, "step": 14881 }, { "epoch": 1.77, "learning_rate": 1.8927005750775e-06, "loss": 0.6234, "step": 14882 }, { "epoch": 1.77, "learning_rate": 1.8923887844571598e-06, "loss": 0.7233, "step": 14883 }, { "epoch": 1.77, "learning_rate": 1.8920770038806838e-06, "loss": 0.7796, "step": 14884 }, { "epoch": 1.77, "learning_rate": 1.8917652333532263e-06, "loss": 0.4408, "step": 14885 }, { "epoch": 1.77, "learning_rate": 1.8914534728799415e-06, "loss": 0.3918, "step": 14886 }, { "epoch": 1.77, "learning_rate": 1.8911417224659803e-06, "loss": 0.5523, "step": 14887 }, { "epoch": 1.77, "learning_rate": 1.8908299821164983e-06, "loss": 0.5098, "step": 14888 }, { "epoch": 1.77, "learning_rate": 1.8905182518366472e-06, "loss": 0.5943, "step": 14889 }, { "epoch": 1.77, "learning_rate": 1.8902065316315814e-06, "loss": 0.8788, "step": 14890 }, { "epoch": 1.77, "learning_rate": 1.8898948215064523e-06, "loss": 0.4918, "step": 14891 }, { "epoch": 1.77, "learning_rate": 1.889583121466413e-06, "loss": 0.4368, "step": 14892 }, { "epoch": 1.77, "learning_rate": 1.8892714315166167e-06, "loss": 0.5891, "step": 14893 }, { "epoch": 1.77, "learning_rate": 1.8889597516622138e-06, "loss": 0.5716, "step": 14894 }, { "epoch": 1.77, "learning_rate": 1.8886480819083571e-06, "loss": 0.6601, "step": 14895 }, { "epoch": 1.77, "learning_rate": 1.8883364222602e-06, "loss": 0.9165, "step": 14896 }, { "epoch": 1.77, "learning_rate": 1.8880247727228933e-06, "loss": 0.7498, "step": 14897 }, { "epoch": 1.77, "learning_rate": 1.8877131333015874e-06, "loss": 0.478, "step": 14898 }, { "epoch": 1.77, "learning_rate": 1.8874015040014356e-06, "loss": 0.4825, "step": 14899 }, { "epoch": 1.77, "learning_rate": 1.8870898848275877e-06, "loss": 0.5856, "step": 14900 }, { "epoch": 1.77, "learning_rate": 1.8867782757851965e-06, "loss": 0.6408, "step": 14901 }, { "epoch": 1.77, "learning_rate": 1.8864666768794107e-06, "loss": 0.928, "step": 14902 }, { "epoch": 1.77, "learning_rate": 1.8861550881153828e-06, "loss": 0.7712, "step": 14903 }, { "epoch": 1.77, "learning_rate": 1.8858435094982624e-06, "loss": 0.521, "step": 14904 }, { "epoch": 1.78, "learning_rate": 1.8855319410332013e-06, "loss": 0.6112, "step": 14905 }, { "epoch": 1.78, "learning_rate": 1.8852203827253476e-06, "loss": 0.5363, "step": 14906 }, { "epoch": 1.78, "learning_rate": 1.884908834579853e-06, "loss": 0.5961, "step": 14907 }, { "epoch": 1.78, "learning_rate": 1.8845972966018678e-06, "loss": 0.7729, "step": 14908 }, { "epoch": 1.78, "learning_rate": 1.8842857687965396e-06, "loss": 0.8139, "step": 14909 }, { "epoch": 1.78, "learning_rate": 1.88397425116902e-06, "loss": 0.4995, "step": 14910 }, { "epoch": 1.78, "learning_rate": 1.883662743724457e-06, "loss": 0.515, "step": 14911 }, { "epoch": 1.78, "learning_rate": 1.883351246468002e-06, "loss": 0.6462, "step": 14912 }, { "epoch": 1.78, "learning_rate": 1.8830397594048012e-06, "loss": 0.5628, "step": 14913 }, { "epoch": 1.78, "learning_rate": 1.8827282825400056e-06, "loss": 0.6884, "step": 14914 }, { "epoch": 1.78, "learning_rate": 1.8824168158787629e-06, "loss": 0.8447, "step": 14915 }, { "epoch": 1.78, "learning_rate": 1.8821053594262228e-06, "loss": 0.6298, "step": 14916 }, { "epoch": 1.78, "learning_rate": 1.8817939131875326e-06, "loss": 0.4325, "step": 14917 }, { "epoch": 1.78, "learning_rate": 1.8814824771678402e-06, "loss": 0.588, "step": 14918 }, { "epoch": 1.78, "learning_rate": 1.8811710513722947e-06, "loss": 0.6712, "step": 14919 }, { "epoch": 1.78, "learning_rate": 1.8808596358060447e-06, "loss": 0.5889, "step": 14920 }, { "epoch": 1.78, "learning_rate": 1.8805482304742358e-06, "loss": 0.7731, "step": 14921 }, { "epoch": 1.78, "learning_rate": 1.8802368353820166e-06, "loss": 0.8519, "step": 14922 }, { "epoch": 1.78, "learning_rate": 1.8799254505345354e-06, "loss": 0.4713, "step": 14923 }, { "epoch": 1.78, "learning_rate": 1.8796140759369375e-06, "loss": 0.5932, "step": 14924 }, { "epoch": 1.78, "learning_rate": 1.8793027115943716e-06, "loss": 0.5713, "step": 14925 }, { "epoch": 1.78, "learning_rate": 1.8789913575119834e-06, "loss": 0.7451, "step": 14926 }, { "epoch": 1.78, "learning_rate": 1.878680013694921e-06, "loss": 0.8221, "step": 14927 }, { "epoch": 1.78, "learning_rate": 1.8783686801483296e-06, "loss": 0.902, "step": 14928 }, { "epoch": 1.78, "learning_rate": 1.8780573568773558e-06, "loss": 0.4444, "step": 14929 }, { "epoch": 1.78, "learning_rate": 1.877746043887146e-06, "loss": 0.6024, "step": 14930 }, { "epoch": 1.78, "learning_rate": 1.8774347411828472e-06, "loss": 0.6065, "step": 14931 }, { "epoch": 1.78, "learning_rate": 1.8771234487696041e-06, "loss": 0.5485, "step": 14932 }, { "epoch": 1.78, "learning_rate": 1.8768121666525619e-06, "loss": 0.7908, "step": 14933 }, { "epoch": 1.78, "learning_rate": 1.8765008948368677e-06, "loss": 0.848, "step": 14934 }, { "epoch": 1.78, "learning_rate": 1.8761896333276657e-06, "loss": 0.4143, "step": 14935 }, { "epoch": 1.78, "learning_rate": 1.8758783821301016e-06, "loss": 0.4884, "step": 14936 }, { "epoch": 1.78, "learning_rate": 1.8755671412493195e-06, "loss": 0.6054, "step": 14937 }, { "epoch": 1.78, "learning_rate": 1.8752559106904661e-06, "loss": 0.591, "step": 14938 }, { "epoch": 1.78, "learning_rate": 1.8749446904586844e-06, "loss": 0.7158, "step": 14939 }, { "epoch": 1.78, "learning_rate": 1.8746334805591187e-06, "loss": 0.8917, "step": 14940 }, { "epoch": 1.78, "learning_rate": 1.8743222809969145e-06, "loss": 0.589, "step": 14941 }, { "epoch": 1.78, "learning_rate": 1.8740110917772161e-06, "loss": 0.3602, "step": 14942 }, { "epoch": 1.78, "learning_rate": 1.8736999129051664e-06, "loss": 0.459, "step": 14943 }, { "epoch": 1.78, "learning_rate": 1.8733887443859094e-06, "loss": 0.5672, "step": 14944 }, { "epoch": 1.78, "learning_rate": 1.8730775862245894e-06, "loss": 0.6719, "step": 14945 }, { "epoch": 1.78, "learning_rate": 1.87276643842635e-06, "loss": 0.8109, "step": 14946 }, { "epoch": 1.78, "learning_rate": 1.872455300996333e-06, "loss": 0.8448, "step": 14947 }, { "epoch": 1.78, "learning_rate": 1.8721441739396828e-06, "loss": 0.4446, "step": 14948 }, { "epoch": 1.78, "learning_rate": 1.8718330572615424e-06, "loss": 0.6205, "step": 14949 }, { "epoch": 1.78, "learning_rate": 1.8715219509670545e-06, "loss": 0.5775, "step": 14950 }, { "epoch": 1.78, "learning_rate": 1.8712108550613608e-06, "loss": 0.7087, "step": 14951 }, { "epoch": 1.78, "learning_rate": 1.8708997695496046e-06, "loss": 0.7071, "step": 14952 }, { "epoch": 1.78, "learning_rate": 1.8705886944369284e-06, "loss": 0.9141, "step": 14953 }, { "epoch": 1.78, "learning_rate": 1.870277629728474e-06, "loss": 0.4292, "step": 14954 }, { "epoch": 1.78, "learning_rate": 1.8699665754293823e-06, "loss": 0.6343, "step": 14955 }, { "epoch": 1.78, "learning_rate": 1.8696555315447968e-06, "loss": 0.5737, "step": 14956 }, { "epoch": 1.78, "learning_rate": 1.8693444980798584e-06, "loss": 0.7331, "step": 14957 }, { "epoch": 1.78, "learning_rate": 1.8690334750397076e-06, "loss": 0.7581, "step": 14958 }, { "epoch": 1.78, "learning_rate": 1.8687224624294863e-06, "loss": 0.7826, "step": 14959 }, { "epoch": 1.78, "learning_rate": 1.8684114602543363e-06, "loss": 0.4281, "step": 14960 }, { "epoch": 1.78, "learning_rate": 1.8681004685193982e-06, "loss": 0.4786, "step": 14961 }, { "epoch": 1.78, "learning_rate": 1.8677894872298113e-06, "loss": 0.4893, "step": 14962 }, { "epoch": 1.78, "learning_rate": 1.8674785163907171e-06, "loss": 0.6609, "step": 14963 }, { "epoch": 1.78, "learning_rate": 1.867167556007257e-06, "loss": 0.7148, "step": 14964 }, { "epoch": 1.78, "learning_rate": 1.8668566060845707e-06, "loss": 0.8755, "step": 14965 }, { "epoch": 1.78, "learning_rate": 1.866545666627797e-06, "loss": 0.5148, "step": 14966 }, { "epoch": 1.78, "learning_rate": 1.8662347376420767e-06, "loss": 0.379, "step": 14967 }, { "epoch": 1.78, "learning_rate": 1.8659238191325491e-06, "loss": 0.478, "step": 14968 }, { "epoch": 1.78, "learning_rate": 1.8656129111043548e-06, "loss": 0.5586, "step": 14969 }, { "epoch": 1.78, "learning_rate": 1.8653020135626314e-06, "loss": 0.7067, "step": 14970 }, { "epoch": 1.78, "learning_rate": 1.8649911265125194e-06, "loss": 0.8707, "step": 14971 }, { "epoch": 1.78, "learning_rate": 1.8646802499591582e-06, "loss": 0.865, "step": 14972 }, { "epoch": 1.78, "learning_rate": 1.8643693839076842e-06, "loss": 0.5192, "step": 14973 }, { "epoch": 1.78, "learning_rate": 1.8640585283632384e-06, "loss": 0.6294, "step": 14974 }, { "epoch": 1.78, "learning_rate": 1.8637476833309581e-06, "loss": 0.6381, "step": 14975 }, { "epoch": 1.78, "learning_rate": 1.863436848815983e-06, "loss": 0.6709, "step": 14976 }, { "epoch": 1.78, "learning_rate": 1.863126024823449e-06, "loss": 0.7896, "step": 14977 }, { "epoch": 1.78, "learning_rate": 1.8628152113584957e-06, "loss": 0.7671, "step": 14978 }, { "epoch": 1.78, "learning_rate": 1.8625044084262603e-06, "loss": 0.4851, "step": 14979 }, { "epoch": 1.78, "learning_rate": 1.8621936160318813e-06, "loss": 0.5168, "step": 14980 }, { "epoch": 1.78, "learning_rate": 1.8618828341804944e-06, "loss": 0.5711, "step": 14981 }, { "epoch": 1.78, "learning_rate": 1.8615720628772383e-06, "loss": 0.6465, "step": 14982 }, { "epoch": 1.78, "learning_rate": 1.8612613021272486e-06, "loss": 0.7864, "step": 14983 }, { "epoch": 1.78, "learning_rate": 1.8609505519356647e-06, "loss": 1.0552, "step": 14984 }, { "epoch": 1.78, "learning_rate": 1.8606398123076214e-06, "loss": 0.4748, "step": 14985 }, { "epoch": 1.78, "learning_rate": 1.8603290832482545e-06, "loss": 0.53, "step": 14986 }, { "epoch": 1.78, "learning_rate": 1.8600183647627033e-06, "loss": 0.4678, "step": 14987 }, { "epoch": 1.78, "learning_rate": 1.8597076568561006e-06, "loss": 0.6441, "step": 14988 }, { "epoch": 1.79, "learning_rate": 1.8593969595335848e-06, "loss": 0.7397, "step": 14989 }, { "epoch": 1.79, "learning_rate": 1.8590862728002906e-06, "loss": 0.8563, "step": 14990 }, { "epoch": 1.79, "learning_rate": 1.8587755966613552e-06, "loss": 0.5791, "step": 14991 }, { "epoch": 1.79, "learning_rate": 1.8584649311219116e-06, "loss": 0.4719, "step": 14992 }, { "epoch": 1.79, "learning_rate": 1.8581542761870975e-06, "loss": 0.5259, "step": 14993 }, { "epoch": 1.79, "learning_rate": 1.857843631862046e-06, "loss": 0.7035, "step": 14994 }, { "epoch": 1.79, "learning_rate": 1.8575329981518947e-06, "loss": 0.733, "step": 14995 }, { "epoch": 1.79, "learning_rate": 1.857222375061776e-06, "loss": 0.7706, "step": 14996 }, { "epoch": 1.79, "learning_rate": 1.8569117625968248e-06, "loss": 0.8372, "step": 14997 }, { "epoch": 1.79, "learning_rate": 1.8566011607621765e-06, "loss": 0.4942, "step": 14998 }, { "epoch": 1.79, "learning_rate": 1.8562905695629657e-06, "loss": 0.5175, "step": 14999 }, { "epoch": 1.79, "learning_rate": 1.8559799890043254e-06, "loss": 0.6359, "step": 15000 }, { "epoch": 1.79, "learning_rate": 1.8556694190913894e-06, "loss": 0.7357, "step": 15001 }, { "epoch": 1.79, "learning_rate": 1.855358859829293e-06, "loss": 0.7502, "step": 15002 }, { "epoch": 1.79, "learning_rate": 1.8550483112231682e-06, "loss": 0.7707, "step": 15003 }, { "epoch": 1.79, "learning_rate": 1.8547377732781486e-06, "loss": 0.3936, "step": 15004 }, { "epoch": 1.79, "learning_rate": 1.8544272459993674e-06, "loss": 0.4584, "step": 15005 }, { "epoch": 1.79, "learning_rate": 1.8541167293919594e-06, "loss": 0.5796, "step": 15006 }, { "epoch": 1.79, "learning_rate": 1.8538062234610548e-06, "loss": 0.5807, "step": 15007 }, { "epoch": 1.79, "learning_rate": 1.8534957282117878e-06, "loss": 0.8084, "step": 15008 }, { "epoch": 1.79, "learning_rate": 1.8531852436492904e-06, "loss": 0.8661, "step": 15009 }, { "epoch": 1.79, "learning_rate": 1.8528747697786964e-06, "loss": 0.4397, "step": 15010 }, { "epoch": 1.79, "learning_rate": 1.852564306605136e-06, "loss": 0.455, "step": 15011 }, { "epoch": 1.79, "learning_rate": 1.8522538541337416e-06, "loss": 0.4407, "step": 15012 }, { "epoch": 1.79, "learning_rate": 1.8519434123696455e-06, "loss": 0.575, "step": 15013 }, { "epoch": 1.79, "learning_rate": 1.85163298131798e-06, "loss": 0.6771, "step": 15014 }, { "epoch": 1.79, "learning_rate": 1.8513225609838748e-06, "loss": 0.7318, "step": 15015 }, { "epoch": 1.79, "learning_rate": 1.8510121513724619e-06, "loss": 0.5048, "step": 15016 }, { "epoch": 1.79, "learning_rate": 1.8507017524888737e-06, "loss": 0.3933, "step": 15017 }, { "epoch": 1.79, "learning_rate": 1.8503913643382392e-06, "loss": 0.6334, "step": 15018 }, { "epoch": 1.79, "learning_rate": 1.85008098692569e-06, "loss": 0.4557, "step": 15019 }, { "epoch": 1.79, "learning_rate": 1.8497706202563562e-06, "loss": 0.6728, "step": 15020 }, { "epoch": 1.79, "learning_rate": 1.8494602643353699e-06, "loss": 0.8737, "step": 15021 }, { "epoch": 1.79, "learning_rate": 1.8491499191678592e-06, "loss": 0.7285, "step": 15022 }, { "epoch": 1.79, "learning_rate": 1.8488395847589547e-06, "loss": 0.4594, "step": 15023 }, { "epoch": 1.79, "learning_rate": 1.8485292611137867e-06, "loss": 0.5502, "step": 15024 }, { "epoch": 1.79, "learning_rate": 1.8482189482374857e-06, "loss": 0.5388, "step": 15025 }, { "epoch": 1.79, "learning_rate": 1.8479086461351786e-06, "loss": 0.5778, "step": 15026 }, { "epoch": 1.79, "learning_rate": 1.8475983548119966e-06, "loss": 0.6994, "step": 15027 }, { "epoch": 1.79, "learning_rate": 1.847288074273069e-06, "loss": 0.853, "step": 15028 }, { "epoch": 1.79, "learning_rate": 1.8469778045235248e-06, "loss": 0.3589, "step": 15029 }, { "epoch": 1.79, "learning_rate": 1.8466675455684912e-06, "loss": 0.5052, "step": 15030 }, { "epoch": 1.79, "learning_rate": 1.8463572974130984e-06, "loss": 0.6221, "step": 15031 }, { "epoch": 1.79, "learning_rate": 1.8460470600624742e-06, "loss": 0.646, "step": 15032 }, { "epoch": 1.79, "learning_rate": 1.8457368335217474e-06, "loss": 0.7973, "step": 15033 }, { "epoch": 1.79, "learning_rate": 1.845426617796045e-06, "loss": 0.8871, "step": 15034 }, { "epoch": 1.79, "learning_rate": 1.8451164128904959e-06, "loss": 0.5572, "step": 15035 }, { "epoch": 1.79, "learning_rate": 1.8448062188102279e-06, "loss": 0.4013, "step": 15036 }, { "epoch": 1.79, "learning_rate": 1.844496035560367e-06, "loss": 0.5458, "step": 15037 }, { "epoch": 1.79, "learning_rate": 1.844185863146042e-06, "loss": 0.5334, "step": 15038 }, { "epoch": 1.79, "learning_rate": 1.8438757015723799e-06, "loss": 0.6976, "step": 15039 }, { "epoch": 1.79, "learning_rate": 1.843565550844508e-06, "loss": 0.9214, "step": 15040 }, { "epoch": 1.79, "learning_rate": 1.8432554109675516e-06, "loss": 0.4831, "step": 15041 }, { "epoch": 1.79, "learning_rate": 1.8429452819466392e-06, "loss": 0.4617, "step": 15042 }, { "epoch": 1.79, "learning_rate": 1.8426351637868956e-06, "loss": 0.5797, "step": 15043 }, { "epoch": 1.79, "learning_rate": 1.8423250564934492e-06, "loss": 0.5343, "step": 15044 }, { "epoch": 1.79, "learning_rate": 1.8420149600714233e-06, "loss": 0.6957, "step": 15045 }, { "epoch": 1.79, "learning_rate": 1.8417048745259465e-06, "loss": 0.9554, "step": 15046 }, { "epoch": 1.79, "learning_rate": 1.8413947998621423e-06, "loss": 0.9816, "step": 15047 }, { "epoch": 1.79, "learning_rate": 1.8410847360851384e-06, "loss": 0.42, "step": 15048 }, { "epoch": 1.79, "learning_rate": 1.8407746832000581e-06, "loss": 0.5284, "step": 15049 }, { "epoch": 1.79, "learning_rate": 1.840464641212028e-06, "loss": 0.562, "step": 15050 }, { "epoch": 1.79, "learning_rate": 1.8401546101261737e-06, "loss": 0.5882, "step": 15051 }, { "epoch": 1.79, "learning_rate": 1.8398445899476177e-06, "loss": 0.6979, "step": 15052 }, { "epoch": 1.79, "learning_rate": 1.839534580681487e-06, "loss": 0.761, "step": 15053 }, { "epoch": 1.79, "learning_rate": 1.8392245823329038e-06, "loss": 0.5249, "step": 15054 }, { "epoch": 1.79, "learning_rate": 1.8389145949069953e-06, "loss": 0.5717, "step": 15055 }, { "epoch": 1.79, "learning_rate": 1.8386046184088824e-06, "loss": 0.5544, "step": 15056 }, { "epoch": 1.79, "learning_rate": 1.8382946528436917e-06, "loss": 0.603, "step": 15057 }, { "epoch": 1.79, "learning_rate": 1.8379846982165452e-06, "loss": 0.8342, "step": 15058 }, { "epoch": 1.79, "learning_rate": 1.8376747545325685e-06, "loss": 0.8636, "step": 15059 }, { "epoch": 1.79, "learning_rate": 1.8373648217968828e-06, "loss": 0.4999, "step": 15060 }, { "epoch": 1.79, "learning_rate": 1.8370549000146118e-06, "loss": 0.4621, "step": 15061 }, { "epoch": 1.79, "learning_rate": 1.836744989190879e-06, "loss": 0.5975, "step": 15062 }, { "epoch": 1.79, "learning_rate": 1.8364350893308081e-06, "loss": 0.5787, "step": 15063 }, { "epoch": 1.79, "learning_rate": 1.8361252004395203e-06, "loss": 0.6591, "step": 15064 }, { "epoch": 1.79, "learning_rate": 1.8358153225221384e-06, "loss": 0.7618, "step": 15065 }, { "epoch": 1.79, "learning_rate": 1.835505455583786e-06, "loss": 0.5175, "step": 15066 }, { "epoch": 1.79, "learning_rate": 1.835195599629583e-06, "loss": 0.487, "step": 15067 }, { "epoch": 1.79, "learning_rate": 1.8348857546646526e-06, "loss": 0.509, "step": 15068 }, { "epoch": 1.79, "learning_rate": 1.8345759206941163e-06, "loss": 0.618, "step": 15069 }, { "epoch": 1.79, "learning_rate": 1.8342660977230969e-06, "loss": 0.6792, "step": 15070 }, { "epoch": 1.79, "learning_rate": 1.8339562857567145e-06, "loss": 0.8789, "step": 15071 }, { "epoch": 1.79, "learning_rate": 1.83364648480009e-06, "loss": 0.8618, "step": 15072 }, { "epoch": 1.8, "learning_rate": 1.8333366948583447e-06, "loss": 0.4584, "step": 15073 }, { "epoch": 1.8, "learning_rate": 1.833026915936601e-06, "loss": 0.5604, "step": 15074 }, { "epoch": 1.8, "learning_rate": 1.8327171480399778e-06, "loss": 0.6225, "step": 15075 }, { "epoch": 1.8, "learning_rate": 1.8324073911735956e-06, "loss": 0.6483, "step": 15076 }, { "epoch": 1.8, "learning_rate": 1.8320976453425757e-06, "loss": 0.8765, "step": 15077 }, { "epoch": 1.8, "learning_rate": 1.8317879105520378e-06, "loss": 0.7657, "step": 15078 }, { "epoch": 1.8, "learning_rate": 1.831478186807102e-06, "loss": 0.4727, "step": 15079 }, { "epoch": 1.8, "learning_rate": 1.8311684741128872e-06, "loss": 0.5432, "step": 15080 }, { "epoch": 1.8, "learning_rate": 1.8308587724745147e-06, "loss": 0.627, "step": 15081 }, { "epoch": 1.8, "learning_rate": 1.8305490818971023e-06, "loss": 0.6572, "step": 15082 }, { "epoch": 1.8, "learning_rate": 1.830239402385769e-06, "loss": 0.7688, "step": 15083 }, { "epoch": 1.8, "learning_rate": 1.8299297339456348e-06, "loss": 0.7285, "step": 15084 }, { "epoch": 1.8, "learning_rate": 1.8296200765818195e-06, "loss": 0.4571, "step": 15085 }, { "epoch": 1.8, "learning_rate": 1.8293104302994396e-06, "loss": 0.4817, "step": 15086 }, { "epoch": 1.8, "learning_rate": 1.8290007951036146e-06, "loss": 0.5174, "step": 15087 }, { "epoch": 1.8, "learning_rate": 1.8286911709994632e-06, "loss": 0.5712, "step": 15088 }, { "epoch": 1.8, "learning_rate": 1.8283815579921033e-06, "loss": 0.86, "step": 15089 }, { "epoch": 1.8, "learning_rate": 1.8280719560866517e-06, "loss": 0.8914, "step": 15090 }, { "epoch": 1.8, "learning_rate": 1.827762365288227e-06, "loss": 0.5199, "step": 15091 }, { "epoch": 1.8, "learning_rate": 1.8274527856019476e-06, "loss": 0.3783, "step": 15092 }, { "epoch": 1.8, "learning_rate": 1.8271432170329301e-06, "loss": 0.5106, "step": 15093 }, { "epoch": 1.8, "learning_rate": 1.8268336595862908e-06, "loss": 0.5092, "step": 15094 }, { "epoch": 1.8, "learning_rate": 1.826524113267148e-06, "loss": 0.661, "step": 15095 }, { "epoch": 1.8, "learning_rate": 1.8262145780806186e-06, "loss": 0.7351, "step": 15096 }, { "epoch": 1.8, "learning_rate": 1.8259050540318184e-06, "loss": 0.7609, "step": 15097 }, { "epoch": 1.8, "learning_rate": 1.8255955411258642e-06, "loss": 0.372, "step": 15098 }, { "epoch": 1.8, "learning_rate": 1.8252860393678723e-06, "loss": 0.5903, "step": 15099 }, { "epoch": 1.8, "learning_rate": 1.8249765487629596e-06, "loss": 0.5124, "step": 15100 }, { "epoch": 1.8, "learning_rate": 1.8246670693162397e-06, "loss": 0.7132, "step": 15101 }, { "epoch": 1.8, "learning_rate": 1.82435760103283e-06, "loss": 0.8051, "step": 15102 }, { "epoch": 1.8, "learning_rate": 1.8240481439178465e-06, "loss": 0.8157, "step": 15103 }, { "epoch": 1.8, "learning_rate": 1.8237386979764044e-06, "loss": 0.4715, "step": 15104 }, { "epoch": 1.8, "learning_rate": 1.8234292632136172e-06, "loss": 0.5724, "step": 15105 }, { "epoch": 1.8, "learning_rate": 1.8231198396346011e-06, "loss": 0.6091, "step": 15106 }, { "epoch": 1.8, "learning_rate": 1.8228104272444712e-06, "loss": 0.6585, "step": 15107 }, { "epoch": 1.8, "learning_rate": 1.8225010260483427e-06, "loss": 0.7311, "step": 15108 }, { "epoch": 1.8, "learning_rate": 1.822191636051328e-06, "loss": 0.9212, "step": 15109 }, { "epoch": 1.8, "learning_rate": 1.8218822572585426e-06, "loss": 0.4232, "step": 15110 }, { "epoch": 1.8, "learning_rate": 1.8215728896751e-06, "loss": 0.4507, "step": 15111 }, { "epoch": 1.8, "learning_rate": 1.8212635333061158e-06, "loss": 0.5677, "step": 15112 }, { "epoch": 1.8, "learning_rate": 1.8209541881567007e-06, "loss": 0.5498, "step": 15113 }, { "epoch": 1.8, "learning_rate": 1.8206448542319708e-06, "loss": 0.6252, "step": 15114 }, { "epoch": 1.8, "learning_rate": 1.8203355315370388e-06, "loss": 0.8848, "step": 15115 }, { "epoch": 1.8, "learning_rate": 1.8200262200770167e-06, "loss": 0.559, "step": 15116 }, { "epoch": 1.8, "learning_rate": 1.8197169198570186e-06, "loss": 0.4372, "step": 15117 }, { "epoch": 1.8, "learning_rate": 1.8194076308821563e-06, "loss": 0.5885, "step": 15118 }, { "epoch": 1.8, "learning_rate": 1.819098353157544e-06, "loss": 0.6154, "step": 15119 }, { "epoch": 1.8, "learning_rate": 1.818789086688292e-06, "loss": 0.6683, "step": 15120 }, { "epoch": 1.8, "learning_rate": 1.8184798314795143e-06, "loss": 0.8161, "step": 15121 }, { "epoch": 1.8, "learning_rate": 1.8181705875363215e-06, "loss": 0.8951, "step": 15122 }, { "epoch": 1.8, "learning_rate": 1.817861354863827e-06, "loss": 0.4536, "step": 15123 }, { "epoch": 1.8, "learning_rate": 1.8175521334671409e-06, "loss": 0.4728, "step": 15124 }, { "epoch": 1.8, "learning_rate": 1.8172429233513756e-06, "loss": 0.6261, "step": 15125 }, { "epoch": 1.8, "learning_rate": 1.8169337245216413e-06, "loss": 0.5974, "step": 15126 }, { "epoch": 1.8, "learning_rate": 1.8166245369830512e-06, "loss": 0.7845, "step": 15127 }, { "epoch": 1.8, "learning_rate": 1.8163153607407141e-06, "loss": 0.8498, "step": 15128 }, { "epoch": 1.8, "learning_rate": 1.816006195799741e-06, "loss": 0.4028, "step": 15129 }, { "epoch": 1.8, "learning_rate": 1.8156970421652442e-06, "loss": 0.349, "step": 15130 }, { "epoch": 1.8, "learning_rate": 1.8153878998423313e-06, "loss": 0.5316, "step": 15131 }, { "epoch": 1.8, "learning_rate": 1.8150787688361145e-06, "loss": 0.6102, "step": 15132 }, { "epoch": 1.8, "learning_rate": 1.8147696491517024e-06, "loss": 0.7533, "step": 15133 }, { "epoch": 1.8, "learning_rate": 1.814460540794207e-06, "loss": 0.8786, "step": 15134 }, { "epoch": 1.8, "learning_rate": 1.8141514437687352e-06, "loss": 0.5233, "step": 15135 }, { "epoch": 1.8, "learning_rate": 1.8138423580803979e-06, "loss": 0.4469, "step": 15136 }, { "epoch": 1.8, "learning_rate": 1.8135332837343037e-06, "loss": 0.5562, "step": 15137 }, { "epoch": 1.8, "learning_rate": 1.8132242207355626e-06, "loss": 0.5617, "step": 15138 }, { "epoch": 1.8, "learning_rate": 1.8129151690892827e-06, "loss": 0.6885, "step": 15139 }, { "epoch": 1.8, "learning_rate": 1.812606128800572e-06, "loss": 0.8115, "step": 15140 }, { "epoch": 1.8, "learning_rate": 1.8122970998745397e-06, "loss": 0.4189, "step": 15141 }, { "epoch": 1.8, "learning_rate": 1.811988082316295e-06, "loss": 0.4199, "step": 15142 }, { "epoch": 1.8, "learning_rate": 1.8116790761309444e-06, "loss": 0.5374, "step": 15143 }, { "epoch": 1.8, "learning_rate": 1.811370081323596e-06, "loss": 0.5623, "step": 15144 }, { "epoch": 1.8, "learning_rate": 1.811061097899359e-06, "loss": 0.7273, "step": 15145 }, { "epoch": 1.8, "learning_rate": 1.8107521258633394e-06, "loss": 0.7006, "step": 15146 }, { "epoch": 1.8, "learning_rate": 1.8104431652206445e-06, "loss": 0.7889, "step": 15147 }, { "epoch": 1.8, "learning_rate": 1.8101342159763818e-06, "loss": 0.3642, "step": 15148 }, { "epoch": 1.8, "learning_rate": 1.8098252781356596e-06, "loss": 0.552, "step": 15149 }, { "epoch": 1.8, "learning_rate": 1.8095163517035823e-06, "loss": 0.5522, "step": 15150 }, { "epoch": 1.8, "learning_rate": 1.809207436685258e-06, "loss": 0.7018, "step": 15151 }, { "epoch": 1.8, "learning_rate": 1.8088985330857922e-06, "loss": 0.7731, "step": 15152 }, { "epoch": 1.8, "learning_rate": 1.8085896409102932e-06, "loss": 0.8365, "step": 15153 }, { "epoch": 1.8, "learning_rate": 1.8082807601638644e-06, "loss": 0.3704, "step": 15154 }, { "epoch": 1.8, "learning_rate": 1.8079718908516124e-06, "loss": 0.4844, "step": 15155 }, { "epoch": 1.8, "learning_rate": 1.8076630329786435e-06, "loss": 0.5844, "step": 15156 }, { "epoch": 1.81, "learning_rate": 1.8073541865500633e-06, "loss": 0.624, "step": 15157 }, { "epoch": 1.81, "learning_rate": 1.8070453515709757e-06, "loss": 0.6039, "step": 15158 }, { "epoch": 1.81, "learning_rate": 1.8067365280464864e-06, "loss": 0.9695, "step": 15159 }, { "epoch": 1.81, "learning_rate": 1.8064277159817015e-06, "loss": 0.4369, "step": 15160 }, { "epoch": 1.81, "learning_rate": 1.8061189153817241e-06, "loss": 0.5551, "step": 15161 }, { "epoch": 1.81, "learning_rate": 1.8058101262516586e-06, "loss": 0.5405, "step": 15162 }, { "epoch": 1.81, "learning_rate": 1.8055013485966103e-06, "loss": 0.6805, "step": 15163 }, { "epoch": 1.81, "learning_rate": 1.8051925824216838e-06, "loss": 0.757, "step": 15164 }, { "epoch": 1.81, "learning_rate": 1.804883827731982e-06, "loss": 0.6625, "step": 15165 }, { "epoch": 1.81, "learning_rate": 1.8045750845326076e-06, "loss": 0.5939, "step": 15166 }, { "epoch": 1.81, "learning_rate": 1.8042663528286667e-06, "loss": 0.5193, "step": 15167 }, { "epoch": 1.81, "learning_rate": 1.8039576326252618e-06, "loss": 0.4978, "step": 15168 }, { "epoch": 1.81, "learning_rate": 1.803648923927494e-06, "loss": 0.6042, "step": 15169 }, { "epoch": 1.81, "learning_rate": 1.8033402267404681e-06, "loss": 0.6497, "step": 15170 }, { "epoch": 1.81, "learning_rate": 1.8030315410692878e-06, "loss": 0.7801, "step": 15171 }, { "epoch": 1.81, "learning_rate": 1.8027228669190547e-06, "loss": 0.6749, "step": 15172 }, { "epoch": 1.81, "learning_rate": 1.8024142042948701e-06, "loss": 0.4323, "step": 15173 }, { "epoch": 1.81, "learning_rate": 1.8021055532018378e-06, "loss": 0.5962, "step": 15174 }, { "epoch": 1.81, "learning_rate": 1.8017969136450585e-06, "loss": 0.4888, "step": 15175 }, { "epoch": 1.81, "learning_rate": 1.8014882856296362e-06, "loss": 0.703, "step": 15176 }, { "epoch": 1.81, "learning_rate": 1.8011796691606697e-06, "loss": 0.825, "step": 15177 }, { "epoch": 1.81, "learning_rate": 1.800871064243263e-06, "loss": 0.7051, "step": 15178 }, { "epoch": 1.81, "learning_rate": 1.8005624708825164e-06, "loss": 0.5635, "step": 15179 }, { "epoch": 1.81, "learning_rate": 1.80025388908353e-06, "loss": 0.5607, "step": 15180 }, { "epoch": 1.81, "learning_rate": 1.7999453188514055e-06, "loss": 0.6266, "step": 15181 }, { "epoch": 1.81, "learning_rate": 1.799636760191244e-06, "loss": 0.6612, "step": 15182 }, { "epoch": 1.81, "learning_rate": 1.7993282131081462e-06, "loss": 0.8067, "step": 15183 }, { "epoch": 1.81, "learning_rate": 1.7990196776072106e-06, "loss": 0.7806, "step": 15184 }, { "epoch": 1.81, "learning_rate": 1.7987111536935393e-06, "loss": 0.4953, "step": 15185 }, { "epoch": 1.81, "learning_rate": 1.7984026413722311e-06, "loss": 0.6631, "step": 15186 }, { "epoch": 1.81, "learning_rate": 1.7980941406483867e-06, "loss": 0.4773, "step": 15187 }, { "epoch": 1.81, "learning_rate": 1.797785651527104e-06, "loss": 0.6277, "step": 15188 }, { "epoch": 1.81, "learning_rate": 1.7974771740134845e-06, "loss": 0.7291, "step": 15189 }, { "epoch": 1.81, "learning_rate": 1.7971687081126254e-06, "loss": 0.948, "step": 15190 }, { "epoch": 1.81, "learning_rate": 1.7968602538296275e-06, "loss": 0.4225, "step": 15191 }, { "epoch": 1.81, "learning_rate": 1.796551811169588e-06, "loss": 0.5346, "step": 15192 }, { "epoch": 1.81, "learning_rate": 1.796243380137606e-06, "loss": 0.519, "step": 15193 }, { "epoch": 1.81, "learning_rate": 1.7959349607387807e-06, "loss": 0.555, "step": 15194 }, { "epoch": 1.81, "learning_rate": 1.7956265529782085e-06, "loss": 0.7067, "step": 15195 }, { "epoch": 1.81, "learning_rate": 1.7953181568609893e-06, "loss": 0.8339, "step": 15196 }, { "epoch": 1.81, "learning_rate": 1.7950097723922195e-06, "loss": 0.8611, "step": 15197 }, { "epoch": 1.81, "learning_rate": 1.794701399576998e-06, "loss": 0.4622, "step": 15198 }, { "epoch": 1.81, "learning_rate": 1.7943930384204209e-06, "loss": 0.5209, "step": 15199 }, { "epoch": 1.81, "learning_rate": 1.7940846889275865e-06, "loss": 0.5243, "step": 15200 }, { "epoch": 1.81, "learning_rate": 1.7937763511035904e-06, "loss": 0.7305, "step": 15201 }, { "epoch": 1.81, "learning_rate": 1.7934680249535322e-06, "loss": 0.7661, "step": 15202 }, { "epoch": 1.81, "learning_rate": 1.793159710482506e-06, "loss": 0.8802, "step": 15203 }, { "epoch": 1.81, "learning_rate": 1.792851407695609e-06, "loss": 0.4578, "step": 15204 }, { "epoch": 1.81, "learning_rate": 1.7925431165979368e-06, "loss": 0.4671, "step": 15205 }, { "epoch": 1.81, "learning_rate": 1.7922348371945878e-06, "loss": 0.5788, "step": 15206 }, { "epoch": 1.81, "learning_rate": 1.7919265694906558e-06, "loss": 0.6882, "step": 15207 }, { "epoch": 1.81, "learning_rate": 1.7916183134912363e-06, "loss": 0.7017, "step": 15208 }, { "epoch": 1.81, "learning_rate": 1.7913100692014264e-06, "loss": 0.8491, "step": 15209 }, { "epoch": 1.81, "learning_rate": 1.7910018366263199e-06, "loss": 0.5966, "step": 15210 }, { "epoch": 1.81, "learning_rate": 1.7906936157710129e-06, "loss": 0.4666, "step": 15211 }, { "epoch": 1.81, "learning_rate": 1.7903854066405992e-06, "loss": 0.5859, "step": 15212 }, { "epoch": 1.81, "learning_rate": 1.7900772092401758e-06, "loss": 0.6306, "step": 15213 }, { "epoch": 1.81, "learning_rate": 1.7897690235748346e-06, "loss": 0.8553, "step": 15214 }, { "epoch": 1.81, "learning_rate": 1.7894608496496706e-06, "loss": 0.8325, "step": 15215 }, { "epoch": 1.81, "learning_rate": 1.789152687469779e-06, "loss": 0.6469, "step": 15216 }, { "epoch": 1.81, "learning_rate": 1.7888445370402536e-06, "loss": 0.4444, "step": 15217 }, { "epoch": 1.81, "learning_rate": 1.7885363983661874e-06, "loss": 0.5569, "step": 15218 }, { "epoch": 1.81, "learning_rate": 1.7882282714526733e-06, "loss": 0.5398, "step": 15219 }, { "epoch": 1.81, "learning_rate": 1.787920156304806e-06, "loss": 0.7065, "step": 15220 }, { "epoch": 1.81, "learning_rate": 1.787612052927679e-06, "loss": 0.856, "step": 15221 }, { "epoch": 1.81, "learning_rate": 1.7873039613263843e-06, "loss": 0.7824, "step": 15222 }, { "epoch": 1.81, "learning_rate": 1.7869958815060145e-06, "loss": 0.4982, "step": 15223 }, { "epoch": 1.81, "learning_rate": 1.7866878134716637e-06, "loss": 0.6036, "step": 15224 }, { "epoch": 1.81, "learning_rate": 1.7863797572284222e-06, "loss": 0.6546, "step": 15225 }, { "epoch": 1.81, "learning_rate": 1.7860717127813828e-06, "loss": 0.8244, "step": 15226 }, { "epoch": 1.81, "learning_rate": 1.7857636801356382e-06, "loss": 0.7752, "step": 15227 }, { "epoch": 1.81, "learning_rate": 1.7854556592962808e-06, "loss": 0.8629, "step": 15228 }, { "epoch": 1.81, "learning_rate": 1.7851476502684007e-06, "loss": 0.421, "step": 15229 }, { "epoch": 1.81, "learning_rate": 1.7848396530570894e-06, "loss": 0.5328, "step": 15230 }, { "epoch": 1.81, "learning_rate": 1.7845316676674391e-06, "loss": 0.5486, "step": 15231 }, { "epoch": 1.81, "learning_rate": 1.784223694104541e-06, "loss": 0.6043, "step": 15232 }, { "epoch": 1.81, "learning_rate": 1.783915732373484e-06, "loss": 0.8271, "step": 15233 }, { "epoch": 1.81, "learning_rate": 1.7836077824793605e-06, "loss": 0.8854, "step": 15234 }, { "epoch": 1.81, "learning_rate": 1.7832998444272606e-06, "loss": 0.4048, "step": 15235 }, { "epoch": 1.81, "learning_rate": 1.7829919182222752e-06, "loss": 0.6292, "step": 15236 }, { "epoch": 1.81, "learning_rate": 1.7826840038694922e-06, "loss": 0.4842, "step": 15237 }, { "epoch": 1.81, "learning_rate": 1.7823761013740033e-06, "loss": 0.6528, "step": 15238 }, { "epoch": 1.81, "learning_rate": 1.7820682107408982e-06, "loss": 0.7551, "step": 15239 }, { "epoch": 1.81, "learning_rate": 1.7817603319752655e-06, "loss": 0.8321, "step": 15240 }, { "epoch": 1.82, "learning_rate": 1.781452465082194e-06, "loss": 0.4933, "step": 15241 }, { "epoch": 1.82, "learning_rate": 1.7811446100667746e-06, "loss": 0.516, "step": 15242 }, { "epoch": 1.82, "learning_rate": 1.7808367669340954e-06, "loss": 0.5799, "step": 15243 }, { "epoch": 1.82, "learning_rate": 1.7805289356892436e-06, "loss": 0.5518, "step": 15244 }, { "epoch": 1.82, "learning_rate": 1.780221116337309e-06, "loss": 0.6172, "step": 15245 }, { "epoch": 1.82, "learning_rate": 1.77991330888338e-06, "loss": 0.8516, "step": 15246 }, { "epoch": 1.82, "learning_rate": 1.7796055133325452e-06, "loss": 0.7541, "step": 15247 }, { "epoch": 1.82, "learning_rate": 1.7792977296898905e-06, "loss": 0.4128, "step": 15248 }, { "epoch": 1.82, "learning_rate": 1.778989957960505e-06, "loss": 0.5206, "step": 15249 }, { "epoch": 1.82, "learning_rate": 1.7786821981494766e-06, "loss": 0.519, "step": 15250 }, { "epoch": 1.82, "learning_rate": 1.778374450261892e-06, "loss": 0.5948, "step": 15251 }, { "epoch": 1.82, "learning_rate": 1.778066714302838e-06, "loss": 0.7952, "step": 15252 }, { "epoch": 1.82, "learning_rate": 1.7777589902774016e-06, "loss": 0.7775, "step": 15253 }, { "epoch": 1.82, "learning_rate": 1.7774512781906694e-06, "loss": 0.4534, "step": 15254 }, { "epoch": 1.82, "learning_rate": 1.7771435780477295e-06, "loss": 0.5178, "step": 15255 }, { "epoch": 1.82, "learning_rate": 1.7768358898536656e-06, "loss": 0.5307, "step": 15256 }, { "epoch": 1.82, "learning_rate": 1.776528213613566e-06, "loss": 0.8556, "step": 15257 }, { "epoch": 1.82, "learning_rate": 1.7762205493325158e-06, "loss": 0.7128, "step": 15258 }, { "epoch": 1.82, "learning_rate": 1.7759128970155997e-06, "loss": 0.8795, "step": 15259 }, { "epoch": 1.82, "learning_rate": 1.7756052566679047e-06, "loss": 0.437, "step": 15260 }, { "epoch": 1.82, "learning_rate": 1.775297628294515e-06, "loss": 0.5437, "step": 15261 }, { "epoch": 1.82, "learning_rate": 1.7749900119005176e-06, "loss": 0.5069, "step": 15262 }, { "epoch": 1.82, "learning_rate": 1.774682407490995e-06, "loss": 0.5976, "step": 15263 }, { "epoch": 1.82, "learning_rate": 1.7743748150710336e-06, "loss": 0.7999, "step": 15264 }, { "epoch": 1.82, "learning_rate": 1.7740672346457165e-06, "loss": 0.8613, "step": 15265 }, { "epoch": 1.82, "learning_rate": 1.7737596662201306e-06, "loss": 0.6176, "step": 15266 }, { "epoch": 1.82, "learning_rate": 1.7734521097993567e-06, "loss": 0.4104, "step": 15267 }, { "epoch": 1.82, "learning_rate": 1.7731445653884815e-06, "loss": 0.6176, "step": 15268 }, { "epoch": 1.82, "learning_rate": 1.7728370329925867e-06, "loss": 0.6729, "step": 15269 }, { "epoch": 1.82, "learning_rate": 1.772529512616758e-06, "loss": 0.7231, "step": 15270 }, { "epoch": 1.82, "learning_rate": 1.7722220042660771e-06, "loss": 0.8223, "step": 15271 }, { "epoch": 1.82, "learning_rate": 1.7719145079456269e-06, "loss": 0.8407, "step": 15272 }, { "epoch": 1.82, "learning_rate": 1.7716070236604916e-06, "loss": 0.4001, "step": 15273 }, { "epoch": 1.82, "learning_rate": 1.7712995514157526e-06, "loss": 0.5724, "step": 15274 }, { "epoch": 1.82, "learning_rate": 1.7709920912164939e-06, "loss": 0.5647, "step": 15275 }, { "epoch": 1.82, "learning_rate": 1.7706846430677962e-06, "loss": 0.6536, "step": 15276 }, { "epoch": 1.82, "learning_rate": 1.7703772069747438e-06, "loss": 0.8067, "step": 15277 }, { "epoch": 1.82, "learning_rate": 1.7700697829424163e-06, "loss": 0.8864, "step": 15278 }, { "epoch": 1.82, "learning_rate": 1.769762370975897e-06, "loss": 0.4889, "step": 15279 }, { "epoch": 1.82, "learning_rate": 1.7694549710802667e-06, "loss": 0.486, "step": 15280 }, { "epoch": 1.82, "learning_rate": 1.7691475832606075e-06, "loss": 0.5934, "step": 15281 }, { "epoch": 1.82, "learning_rate": 1.7688402075220001e-06, "loss": 0.7546, "step": 15282 }, { "epoch": 1.82, "learning_rate": 1.7685328438695248e-06, "loss": 0.8073, "step": 15283 }, { "epoch": 1.82, "learning_rate": 1.7682254923082628e-06, "loss": 0.8335, "step": 15284 }, { "epoch": 1.82, "learning_rate": 1.7679181528432959e-06, "loss": 0.462, "step": 15285 }, { "epoch": 1.82, "learning_rate": 1.7676108254797031e-06, "loss": 0.3794, "step": 15286 }, { "epoch": 1.82, "learning_rate": 1.7673035102225639e-06, "loss": 0.5372, "step": 15287 }, { "epoch": 1.82, "learning_rate": 1.7669962070769602e-06, "loss": 0.567, "step": 15288 }, { "epoch": 1.82, "learning_rate": 1.7666889160479706e-06, "loss": 0.7362, "step": 15289 }, { "epoch": 1.82, "learning_rate": 1.7663816371406739e-06, "loss": 0.8291, "step": 15290 }, { "epoch": 1.82, "learning_rate": 1.7660743703601503e-06, "loss": 0.6664, "step": 15291 }, { "epoch": 1.82, "learning_rate": 1.7657671157114803e-06, "loss": 0.4633, "step": 15292 }, { "epoch": 1.82, "learning_rate": 1.7654598731997404e-06, "loss": 0.5267, "step": 15293 }, { "epoch": 1.82, "learning_rate": 1.7651526428300103e-06, "loss": 0.614, "step": 15294 }, { "epoch": 1.82, "learning_rate": 1.7648454246073688e-06, "loss": 0.5891, "step": 15295 }, { "epoch": 1.82, "learning_rate": 1.7645382185368947e-06, "loss": 0.8705, "step": 15296 }, { "epoch": 1.82, "learning_rate": 1.7642310246236656e-06, "loss": 0.6684, "step": 15297 }, { "epoch": 1.82, "learning_rate": 1.7639238428727584e-06, "loss": 0.4359, "step": 15298 }, { "epoch": 1.82, "learning_rate": 1.7636166732892524e-06, "loss": 0.5708, "step": 15299 }, { "epoch": 1.82, "learning_rate": 1.7633095158782254e-06, "loss": 0.6464, "step": 15300 }, { "epoch": 1.82, "learning_rate": 1.7630023706447524e-06, "loss": 0.7014, "step": 15301 }, { "epoch": 1.82, "learning_rate": 1.7626952375939127e-06, "loss": 0.7775, "step": 15302 }, { "epoch": 1.82, "learning_rate": 1.7623881167307833e-06, "loss": 0.8494, "step": 15303 }, { "epoch": 1.82, "learning_rate": 1.7620810080604394e-06, "loss": 0.3705, "step": 15304 }, { "epoch": 1.82, "learning_rate": 1.761773911587958e-06, "loss": 0.4912, "step": 15305 }, { "epoch": 1.82, "learning_rate": 1.761466827318416e-06, "loss": 0.5548, "step": 15306 }, { "epoch": 1.82, "learning_rate": 1.7611597552568905e-06, "loss": 0.6303, "step": 15307 }, { "epoch": 1.82, "learning_rate": 1.7608526954084554e-06, "loss": 0.7544, "step": 15308 }, { "epoch": 1.82, "learning_rate": 1.7605456477781868e-06, "loss": 0.8093, "step": 15309 }, { "epoch": 1.82, "learning_rate": 1.7602386123711611e-06, "loss": 0.4473, "step": 15310 }, { "epoch": 1.82, "learning_rate": 1.759931589192454e-06, "loss": 0.4399, "step": 15311 }, { "epoch": 1.82, "learning_rate": 1.7596245782471387e-06, "loss": 0.4511, "step": 15312 }, { "epoch": 1.82, "learning_rate": 1.7593175795402911e-06, "loss": 0.5654, "step": 15313 }, { "epoch": 1.82, "learning_rate": 1.7590105930769866e-06, "loss": 0.7065, "step": 15314 }, { "epoch": 1.82, "learning_rate": 1.7587036188622996e-06, "loss": 0.8448, "step": 15315 }, { "epoch": 1.82, "learning_rate": 1.7583966569013033e-06, "loss": 0.7023, "step": 15316 }, { "epoch": 1.82, "learning_rate": 1.7580897071990726e-06, "loss": 0.4612, "step": 15317 }, { "epoch": 1.82, "learning_rate": 1.7577827697606811e-06, "loss": 0.63, "step": 15318 }, { "epoch": 1.82, "learning_rate": 1.7574758445912037e-06, "loss": 0.569, "step": 15319 }, { "epoch": 1.82, "learning_rate": 1.7571689316957117e-06, "loss": 0.8076, "step": 15320 }, { "epoch": 1.82, "learning_rate": 1.75686203107928e-06, "loss": 0.7758, "step": 15321 }, { "epoch": 1.82, "learning_rate": 1.756555142746982e-06, "loss": 0.9449, "step": 15322 }, { "epoch": 1.82, "learning_rate": 1.7562482667038888e-06, "loss": 0.3643, "step": 15323 }, { "epoch": 1.82, "learning_rate": 1.7559414029550742e-06, "loss": 0.5135, "step": 15324 }, { "epoch": 1.83, "learning_rate": 1.7556345515056112e-06, "loss": 0.6275, "step": 15325 }, { "epoch": 1.83, "learning_rate": 1.7553277123605719e-06, "loss": 0.6505, "step": 15326 }, { "epoch": 1.83, "learning_rate": 1.755020885525027e-06, "loss": 0.7202, "step": 15327 }, { "epoch": 1.83, "learning_rate": 1.7547140710040501e-06, "loss": 0.8408, "step": 15328 }, { "epoch": 1.83, "learning_rate": 1.7544072688027113e-06, "loss": 0.4738, "step": 15329 }, { "epoch": 1.83, "learning_rate": 1.754100478926084e-06, "loss": 0.5535, "step": 15330 }, { "epoch": 1.83, "learning_rate": 1.7537937013792372e-06, "loss": 0.5898, "step": 15331 }, { "epoch": 1.83, "learning_rate": 1.7534869361672442e-06, "loss": 0.7815, "step": 15332 }, { "epoch": 1.83, "learning_rate": 1.7531801832951738e-06, "loss": 0.7128, "step": 15333 }, { "epoch": 1.83, "learning_rate": 1.7528734427680988e-06, "loss": 0.8585, "step": 15334 }, { "epoch": 1.83, "learning_rate": 1.7525667145910875e-06, "loss": 0.5299, "step": 15335 }, { "epoch": 1.83, "learning_rate": 1.7522599987692113e-06, "loss": 0.5288, "step": 15336 }, { "epoch": 1.83, "learning_rate": 1.7519532953075407e-06, "loss": 0.5901, "step": 15337 }, { "epoch": 1.83, "learning_rate": 1.751646604211144e-06, "loss": 0.6606, "step": 15338 }, { "epoch": 1.83, "learning_rate": 1.7513399254850928e-06, "loss": 0.7443, "step": 15339 }, { "epoch": 1.83, "learning_rate": 1.7510332591344543e-06, "loss": 0.8377, "step": 15340 }, { "epoch": 1.83, "learning_rate": 1.7507266051642997e-06, "loss": 0.3976, "step": 15341 }, { "epoch": 1.83, "learning_rate": 1.7504199635796965e-06, "loss": 0.5251, "step": 15342 }, { "epoch": 1.83, "learning_rate": 1.7501133343857145e-06, "loss": 0.5206, "step": 15343 }, { "epoch": 1.83, "learning_rate": 1.7498067175874217e-06, "loss": 0.6683, "step": 15344 }, { "epoch": 1.83, "learning_rate": 1.7495001131898879e-06, "loss": 0.6476, "step": 15345 }, { "epoch": 1.83, "learning_rate": 1.7491935211981795e-06, "loss": 0.8922, "step": 15346 }, { "epoch": 1.83, "learning_rate": 1.7488869416173649e-06, "loss": 0.9168, "step": 15347 }, { "epoch": 1.83, "learning_rate": 1.7485803744525117e-06, "loss": 0.3879, "step": 15348 }, { "epoch": 1.83, "learning_rate": 1.7482738197086897e-06, "loss": 0.5971, "step": 15349 }, { "epoch": 1.83, "learning_rate": 1.7479672773909633e-06, "loss": 0.6052, "step": 15350 }, { "epoch": 1.83, "learning_rate": 1.7476607475044007e-06, "loss": 0.5467, "step": 15351 }, { "epoch": 1.83, "learning_rate": 1.7473542300540703e-06, "loss": 0.7546, "step": 15352 }, { "epoch": 1.83, "learning_rate": 1.7470477250450363e-06, "loss": 0.9012, "step": 15353 }, { "epoch": 1.83, "learning_rate": 1.7467412324823675e-06, "loss": 0.4055, "step": 15354 }, { "epoch": 1.83, "learning_rate": 1.7464347523711283e-06, "loss": 0.5247, "step": 15355 }, { "epoch": 1.83, "learning_rate": 1.7461282847163872e-06, "loss": 0.6271, "step": 15356 }, { "epoch": 1.83, "learning_rate": 1.7458218295232082e-06, "loss": 0.6646, "step": 15357 }, { "epoch": 1.83, "learning_rate": 1.7455153867966571e-06, "loss": 0.707, "step": 15358 }, { "epoch": 1.83, "learning_rate": 1.7452089565418e-06, "loss": 0.9234, "step": 15359 }, { "epoch": 1.83, "learning_rate": 1.744902538763703e-06, "loss": 0.3688, "step": 15360 }, { "epoch": 1.83, "learning_rate": 1.7445961334674303e-06, "loss": 0.5398, "step": 15361 }, { "epoch": 1.83, "learning_rate": 1.7442897406580462e-06, "loss": 0.4814, "step": 15362 }, { "epoch": 1.83, "learning_rate": 1.7439833603406158e-06, "loss": 0.5516, "step": 15363 }, { "epoch": 1.83, "learning_rate": 1.7436769925202054e-06, "loss": 0.7476, "step": 15364 }, { "epoch": 1.83, "learning_rate": 1.7433706372018772e-06, "loss": 0.9108, "step": 15365 }, { "epoch": 1.83, "learning_rate": 1.743064294390695e-06, "loss": 0.4752, "step": 15366 }, { "epoch": 1.83, "learning_rate": 1.7427579640917246e-06, "loss": 0.5696, "step": 15367 }, { "epoch": 1.83, "learning_rate": 1.7424516463100282e-06, "loss": 0.5112, "step": 15368 }, { "epoch": 1.83, "learning_rate": 1.7421453410506694e-06, "loss": 0.5991, "step": 15369 }, { "epoch": 1.83, "learning_rate": 1.7418390483187114e-06, "loss": 0.7248, "step": 15370 }, { "epoch": 1.83, "learning_rate": 1.7415327681192187e-06, "loss": 0.7106, "step": 15371 }, { "epoch": 1.83, "learning_rate": 1.7412265004572527e-06, "loss": 0.7154, "step": 15372 }, { "epoch": 1.83, "learning_rate": 1.7409202453378755e-06, "loss": 0.3388, "step": 15373 }, { "epoch": 1.83, "learning_rate": 1.7406140027661505e-06, "loss": 0.4498, "step": 15374 }, { "epoch": 1.83, "learning_rate": 1.7403077727471401e-06, "loss": 0.5282, "step": 15375 }, { "epoch": 1.83, "learning_rate": 1.7400015552859061e-06, "loss": 0.6813, "step": 15376 }, { "epoch": 1.83, "learning_rate": 1.7396953503875094e-06, "loss": 0.7992, "step": 15377 }, { "epoch": 1.83, "learning_rate": 1.739389158057013e-06, "loss": 0.869, "step": 15378 }, { "epoch": 1.83, "learning_rate": 1.7390829782994783e-06, "loss": 0.521, "step": 15379 }, { "epoch": 1.83, "learning_rate": 1.7387768111199643e-06, "loss": 0.5857, "step": 15380 }, { "epoch": 1.83, "learning_rate": 1.738470656523534e-06, "loss": 0.5646, "step": 15381 }, { "epoch": 1.83, "learning_rate": 1.7381645145152482e-06, "loss": 0.5538, "step": 15382 }, { "epoch": 1.83, "learning_rate": 1.7378583851001668e-06, "loss": 0.7851, "step": 15383 }, { "epoch": 1.83, "learning_rate": 1.7375522682833496e-06, "loss": 0.9103, "step": 15384 }, { "epoch": 1.83, "learning_rate": 1.737246164069858e-06, "loss": 0.462, "step": 15385 }, { "epoch": 1.83, "learning_rate": 1.7369400724647517e-06, "loss": 0.4943, "step": 15386 }, { "epoch": 1.83, "learning_rate": 1.736633993473089e-06, "loss": 0.4669, "step": 15387 }, { "epoch": 1.83, "learning_rate": 1.7363279270999306e-06, "loss": 0.6718, "step": 15388 }, { "epoch": 1.83, "learning_rate": 1.7360218733503358e-06, "loss": 0.7047, "step": 15389 }, { "epoch": 1.83, "learning_rate": 1.7357158322293645e-06, "loss": 0.8788, "step": 15390 }, { "epoch": 1.83, "learning_rate": 1.735409803742073e-06, "loss": 0.5853, "step": 15391 }, { "epoch": 1.83, "learning_rate": 1.7351037878935223e-06, "loss": 0.5252, "step": 15392 }, { "epoch": 1.83, "learning_rate": 1.7347977846887705e-06, "loss": 0.6511, "step": 15393 }, { "epoch": 1.83, "learning_rate": 1.7344917941328753e-06, "loss": 0.5373, "step": 15394 }, { "epoch": 1.83, "learning_rate": 1.7341858162308947e-06, "loss": 0.5943, "step": 15395 }, { "epoch": 1.83, "learning_rate": 1.7338798509878873e-06, "loss": 0.8377, "step": 15396 }, { "epoch": 1.83, "learning_rate": 1.7335738984089096e-06, "loss": 0.7914, "step": 15397 }, { "epoch": 1.83, "learning_rate": 1.7332679584990206e-06, "loss": 0.4362, "step": 15398 }, { "epoch": 1.83, "learning_rate": 1.7329620312632756e-06, "loss": 0.5027, "step": 15399 }, { "epoch": 1.83, "learning_rate": 1.732656116706733e-06, "loss": 0.7019, "step": 15400 }, { "epoch": 1.83, "learning_rate": 1.7323502148344498e-06, "loss": 0.7808, "step": 15401 }, { "epoch": 1.83, "learning_rate": 1.732044325651481e-06, "loss": 0.8553, "step": 15402 }, { "epoch": 1.83, "learning_rate": 1.7317384491628843e-06, "loss": 0.7801, "step": 15403 }, { "epoch": 1.83, "learning_rate": 1.7314325853737152e-06, "loss": 0.4133, "step": 15404 }, { "epoch": 1.83, "learning_rate": 1.731126734289031e-06, "loss": 0.4561, "step": 15405 }, { "epoch": 1.83, "learning_rate": 1.7308208959138851e-06, "loss": 0.4747, "step": 15406 }, { "epoch": 1.83, "learning_rate": 1.7305150702533349e-06, "loss": 0.6897, "step": 15407 }, { "epoch": 1.83, "learning_rate": 1.7302092573124347e-06, "loss": 0.6969, "step": 15408 }, { "epoch": 1.84, "learning_rate": 1.729903457096241e-06, "loss": 0.8221, "step": 15409 }, { "epoch": 1.84, "learning_rate": 1.7295976696098072e-06, "loss": 0.5635, "step": 15410 }, { "epoch": 1.84, "learning_rate": 1.7292918948581883e-06, "loss": 0.5074, "step": 15411 }, { "epoch": 1.84, "learning_rate": 1.7289861328464391e-06, "loss": 0.5586, "step": 15412 }, { "epoch": 1.84, "learning_rate": 1.7286803835796146e-06, "loss": 0.571, "step": 15413 }, { "epoch": 1.84, "learning_rate": 1.7283746470627678e-06, "loss": 0.7489, "step": 15414 }, { "epoch": 1.84, "learning_rate": 1.7280689233009523e-06, "loss": 0.8458, "step": 15415 }, { "epoch": 1.84, "learning_rate": 1.727763212299223e-06, "loss": 0.6365, "step": 15416 }, { "epoch": 1.84, "learning_rate": 1.7274575140626318e-06, "loss": 0.4782, "step": 15417 }, { "epoch": 1.84, "learning_rate": 1.727151828596233e-06, "loss": 0.5708, "step": 15418 }, { "epoch": 1.84, "learning_rate": 1.7268461559050792e-06, "loss": 0.6606, "step": 15419 }, { "epoch": 1.84, "learning_rate": 1.726540495994224e-06, "loss": 0.6491, "step": 15420 }, { "epoch": 1.84, "learning_rate": 1.7262348488687183e-06, "loss": 0.7924, "step": 15421 }, { "epoch": 1.84, "learning_rate": 1.7259292145336163e-06, "loss": 0.8114, "step": 15422 }, { "epoch": 1.84, "learning_rate": 1.7256235929939685e-06, "loss": 0.4435, "step": 15423 }, { "epoch": 1.84, "learning_rate": 1.725317984254829e-06, "loss": 0.5837, "step": 15424 }, { "epoch": 1.84, "learning_rate": 1.7250123883212472e-06, "loss": 0.5203, "step": 15425 }, { "epoch": 1.84, "learning_rate": 1.7247068051982758e-06, "loss": 0.7029, "step": 15426 }, { "epoch": 1.84, "learning_rate": 1.7244012348909656e-06, "loss": 0.779, "step": 15427 }, { "epoch": 1.84, "learning_rate": 1.7240956774043687e-06, "loss": 0.8999, "step": 15428 }, { "epoch": 1.84, "learning_rate": 1.7237901327435352e-06, "loss": 0.4188, "step": 15429 }, { "epoch": 1.84, "learning_rate": 1.7234846009135159e-06, "loss": 0.5196, "step": 15430 }, { "epoch": 1.84, "learning_rate": 1.7231790819193616e-06, "loss": 0.5208, "step": 15431 }, { "epoch": 1.84, "learning_rate": 1.7228735757661223e-06, "loss": 0.6309, "step": 15432 }, { "epoch": 1.84, "learning_rate": 1.722568082458847e-06, "loss": 0.7223, "step": 15433 }, { "epoch": 1.84, "learning_rate": 1.7222626020025867e-06, "loss": 0.881, "step": 15434 }, { "epoch": 1.84, "learning_rate": 1.7219571344023922e-06, "loss": 0.4555, "step": 15435 }, { "epoch": 1.84, "learning_rate": 1.7216516796633106e-06, "loss": 0.4566, "step": 15436 }, { "epoch": 1.84, "learning_rate": 1.7213462377903917e-06, "loss": 0.6339, "step": 15437 }, { "epoch": 1.84, "learning_rate": 1.7210408087886848e-06, "loss": 0.6222, "step": 15438 }, { "epoch": 1.84, "learning_rate": 1.7207353926632397e-06, "loss": 0.776, "step": 15439 }, { "epoch": 1.84, "learning_rate": 1.7204299894191033e-06, "loss": 0.8056, "step": 15440 }, { "epoch": 1.84, "learning_rate": 1.7201245990613246e-06, "loss": 0.6198, "step": 15441 }, { "epoch": 1.84, "learning_rate": 1.7198192215949517e-06, "loss": 0.4221, "step": 15442 }, { "epoch": 1.84, "learning_rate": 1.7195138570250334e-06, "loss": 0.6028, "step": 15443 }, { "epoch": 1.84, "learning_rate": 1.719208505356615e-06, "loss": 0.6931, "step": 15444 }, { "epoch": 1.84, "learning_rate": 1.718903166594746e-06, "loss": 0.6407, "step": 15445 }, { "epoch": 1.84, "learning_rate": 1.7185978407444741e-06, "loss": 0.8703, "step": 15446 }, { "epoch": 1.84, "learning_rate": 1.7182925278108448e-06, "loss": 0.7735, "step": 15447 }, { "epoch": 1.84, "learning_rate": 1.7179872277989052e-06, "loss": 0.343, "step": 15448 }, { "epoch": 1.84, "learning_rate": 1.7176819407137023e-06, "loss": 0.6445, "step": 15449 }, { "epoch": 1.84, "learning_rate": 1.7173766665602836e-06, "loss": 0.4803, "step": 15450 }, { "epoch": 1.84, "learning_rate": 1.7170714053436938e-06, "loss": 0.6663, "step": 15451 }, { "epoch": 1.84, "learning_rate": 1.716766157068979e-06, "loss": 0.7529, "step": 15452 }, { "epoch": 1.84, "learning_rate": 1.7164609217411856e-06, "loss": 0.9446, "step": 15453 }, { "epoch": 1.84, "learning_rate": 1.7161556993653595e-06, "loss": 0.5147, "step": 15454 }, { "epoch": 1.84, "learning_rate": 1.7158504899465445e-06, "loss": 0.5255, "step": 15455 }, { "epoch": 1.84, "learning_rate": 1.7155452934897864e-06, "loss": 0.5639, "step": 15456 }, { "epoch": 1.84, "learning_rate": 1.7152401100001314e-06, "loss": 0.6404, "step": 15457 }, { "epoch": 1.84, "learning_rate": 1.7149349394826236e-06, "loss": 0.7581, "step": 15458 }, { "epoch": 1.84, "learning_rate": 1.7146297819423057e-06, "loss": 0.8181, "step": 15459 }, { "epoch": 1.84, "learning_rate": 1.7143246373842243e-06, "loss": 0.3878, "step": 15460 }, { "epoch": 1.84, "learning_rate": 1.714019505813422e-06, "loss": 0.5217, "step": 15461 }, { "epoch": 1.84, "learning_rate": 1.7137143872349443e-06, "loss": 0.4913, "step": 15462 }, { "epoch": 1.84, "learning_rate": 1.7134092816538329e-06, "loss": 0.5255, "step": 15463 }, { "epoch": 1.84, "learning_rate": 1.713104189075132e-06, "loss": 0.7717, "step": 15464 }, { "epoch": 1.84, "learning_rate": 1.712799109503886e-06, "loss": 0.776, "step": 15465 }, { "epoch": 1.84, "learning_rate": 1.7124940429451354e-06, "loss": 0.5838, "step": 15466 }, { "epoch": 1.84, "learning_rate": 1.7121889894039245e-06, "loss": 0.4658, "step": 15467 }, { "epoch": 1.84, "learning_rate": 1.7118839488852963e-06, "loss": 0.5572, "step": 15468 }, { "epoch": 1.84, "learning_rate": 1.7115789213942927e-06, "loss": 0.5726, "step": 15469 }, { "epoch": 1.84, "learning_rate": 1.711273906935955e-06, "loss": 0.668, "step": 15470 }, { "epoch": 1.84, "learning_rate": 1.7109689055153261e-06, "loss": 0.7637, "step": 15471 }, { "epoch": 1.84, "learning_rate": 1.7106639171374468e-06, "loss": 0.9162, "step": 15472 }, { "epoch": 1.84, "learning_rate": 1.7103589418073605e-06, "loss": 0.47, "step": 15473 }, { "epoch": 1.84, "learning_rate": 1.7100539795301058e-06, "loss": 0.5757, "step": 15474 }, { "epoch": 1.84, "learning_rate": 1.709749030310726e-06, "loss": 0.5459, "step": 15475 }, { "epoch": 1.84, "learning_rate": 1.7094440941542603e-06, "loss": 0.7258, "step": 15476 }, { "epoch": 1.84, "learning_rate": 1.7091391710657511e-06, "loss": 0.7463, "step": 15477 }, { "epoch": 1.84, "learning_rate": 1.7088342610502368e-06, "loss": 0.8603, "step": 15478 }, { "epoch": 1.84, "learning_rate": 1.708529364112759e-06, "loss": 0.3598, "step": 15479 }, { "epoch": 1.84, "learning_rate": 1.7082244802583578e-06, "loss": 0.4765, "step": 15480 }, { "epoch": 1.84, "learning_rate": 1.7079196094920714e-06, "loss": 0.5905, "step": 15481 }, { "epoch": 1.84, "learning_rate": 1.7076147518189407e-06, "loss": 0.4999, "step": 15482 }, { "epoch": 1.84, "learning_rate": 1.7073099072440041e-06, "loss": 0.8237, "step": 15483 }, { "epoch": 1.84, "learning_rate": 1.7070050757723022e-06, "loss": 0.8007, "step": 15484 }, { "epoch": 1.84, "learning_rate": 1.706700257408872e-06, "loss": 0.4102, "step": 15485 }, { "epoch": 1.84, "learning_rate": 1.706395452158754e-06, "loss": 0.3564, "step": 15486 }, { "epoch": 1.84, "learning_rate": 1.7060906600269844e-06, "loss": 0.5342, "step": 15487 }, { "epoch": 1.84, "learning_rate": 1.7057858810186046e-06, "loss": 0.6942, "step": 15488 }, { "epoch": 1.84, "learning_rate": 1.7054811151386502e-06, "loss": 0.7917, "step": 15489 }, { "epoch": 1.84, "learning_rate": 1.705176362392159e-06, "loss": 0.8205, "step": 15490 }, { "epoch": 1.84, "learning_rate": 1.7048716227841689e-06, "loss": 0.5598, "step": 15491 }, { "epoch": 1.84, "learning_rate": 1.704566896319719e-06, "loss": 0.6493, "step": 15492 }, { "epoch": 1.85, "learning_rate": 1.7042621830038445e-06, "loss": 0.6286, "step": 15493 }, { "epoch": 1.85, "learning_rate": 1.7039574828415823e-06, "loss": 0.5601, "step": 15494 }, { "epoch": 1.85, "learning_rate": 1.7036527958379706e-06, "loss": 0.7482, "step": 15495 }, { "epoch": 1.85, "learning_rate": 1.7033481219980441e-06, "loss": 0.879, "step": 15496 }, { "epoch": 1.85, "learning_rate": 1.7030434613268406e-06, "loss": 0.7725, "step": 15497 }, { "epoch": 1.85, "learning_rate": 1.7027388138293949e-06, "loss": 0.4013, "step": 15498 }, { "epoch": 1.85, "learning_rate": 1.7024341795107446e-06, "loss": 0.566, "step": 15499 }, { "epoch": 1.85, "learning_rate": 1.702129558375924e-06, "loss": 0.6582, "step": 15500 }, { "epoch": 1.85, "learning_rate": 1.7018249504299677e-06, "loss": 0.6768, "step": 15501 }, { "epoch": 1.85, "learning_rate": 1.7015203556779122e-06, "loss": 0.7177, "step": 15502 }, { "epoch": 1.85, "learning_rate": 1.701215774124793e-06, "loss": 0.8707, "step": 15503 }, { "epoch": 1.85, "learning_rate": 1.700911205775644e-06, "loss": 0.507, "step": 15504 }, { "epoch": 1.85, "learning_rate": 1.700606650635499e-06, "loss": 0.4901, "step": 15505 }, { "epoch": 1.85, "learning_rate": 1.7003021087093936e-06, "loss": 0.546, "step": 15506 }, { "epoch": 1.85, "learning_rate": 1.6999975800023619e-06, "loss": 0.652, "step": 15507 }, { "epoch": 1.85, "learning_rate": 1.699693064519437e-06, "loss": 0.7676, "step": 15508 }, { "epoch": 1.85, "learning_rate": 1.6993885622656526e-06, "loss": 0.8345, "step": 15509 }, { "epoch": 1.85, "learning_rate": 1.699084073246043e-06, "loss": 0.4819, "step": 15510 }, { "epoch": 1.85, "learning_rate": 1.6987795974656406e-06, "loss": 0.443, "step": 15511 }, { "epoch": 1.85, "learning_rate": 1.6984751349294782e-06, "loss": 0.5574, "step": 15512 }, { "epoch": 1.85, "learning_rate": 1.6981706856425892e-06, "loss": 0.6453, "step": 15513 }, { "epoch": 1.85, "learning_rate": 1.6978662496100068e-06, "loss": 0.709, "step": 15514 }, { "epoch": 1.85, "learning_rate": 1.6975618268367621e-06, "loss": 0.7789, "step": 15515 }, { "epoch": 1.85, "learning_rate": 1.6972574173278878e-06, "loss": 0.6035, "step": 15516 }, { "epoch": 1.85, "learning_rate": 1.6969530210884156e-06, "loss": 0.4654, "step": 15517 }, { "epoch": 1.85, "learning_rate": 1.6966486381233776e-06, "loss": 0.555, "step": 15518 }, { "epoch": 1.85, "learning_rate": 1.696344268437805e-06, "loss": 0.5099, "step": 15519 }, { "epoch": 1.85, "learning_rate": 1.6960399120367285e-06, "loss": 0.6174, "step": 15520 }, { "epoch": 1.85, "learning_rate": 1.6957355689251804e-06, "loss": 0.8509, "step": 15521 }, { "epoch": 1.85, "learning_rate": 1.6954312391081911e-06, "loss": 0.7969, "step": 15522 }, { "epoch": 1.85, "learning_rate": 1.69512692259079e-06, "loss": 0.4927, "step": 15523 }, { "epoch": 1.85, "learning_rate": 1.6948226193780082e-06, "loss": 0.5595, "step": 15524 }, { "epoch": 1.85, "learning_rate": 1.6945183294748769e-06, "loss": 0.6013, "step": 15525 }, { "epoch": 1.85, "learning_rate": 1.6942140528864257e-06, "loss": 0.6411, "step": 15526 }, { "epoch": 1.85, "learning_rate": 1.6939097896176825e-06, "loss": 0.7845, "step": 15527 }, { "epoch": 1.85, "learning_rate": 1.6936055396736788e-06, "loss": 0.7077, "step": 15528 }, { "epoch": 1.85, "learning_rate": 1.6933013030594436e-06, "loss": 0.4188, "step": 15529 }, { "epoch": 1.85, "learning_rate": 1.692997079780005e-06, "loss": 0.5171, "step": 15530 }, { "epoch": 1.85, "learning_rate": 1.692692869840392e-06, "loss": 0.5592, "step": 15531 }, { "epoch": 1.85, "learning_rate": 1.692388673245634e-06, "loss": 0.6279, "step": 15532 }, { "epoch": 1.85, "learning_rate": 1.6920844900007597e-06, "loss": 0.7897, "step": 15533 }, { "epoch": 1.85, "learning_rate": 1.6917803201107955e-06, "loss": 0.8324, "step": 15534 }, { "epoch": 1.85, "learning_rate": 1.6914761635807703e-06, "loss": 0.3958, "step": 15535 }, { "epoch": 1.85, "learning_rate": 1.691172020415713e-06, "loss": 0.546, "step": 15536 }, { "epoch": 1.85, "learning_rate": 1.69086789062065e-06, "loss": 0.6085, "step": 15537 }, { "epoch": 1.85, "learning_rate": 1.6905637742006075e-06, "loss": 0.5325, "step": 15538 }, { "epoch": 1.85, "learning_rate": 1.6902596711606145e-06, "loss": 0.6801, "step": 15539 }, { "epoch": 1.85, "learning_rate": 1.6899555815056967e-06, "loss": 0.7268, "step": 15540 }, { "epoch": 1.85, "learning_rate": 1.689651505240882e-06, "loss": 0.5088, "step": 15541 }, { "epoch": 1.85, "learning_rate": 1.689347442371195e-06, "loss": 0.4519, "step": 15542 }, { "epoch": 1.85, "learning_rate": 1.689043392901663e-06, "loss": 0.5695, "step": 15543 }, { "epoch": 1.85, "learning_rate": 1.6887393568373129e-06, "loss": 0.5617, "step": 15544 }, { "epoch": 1.85, "learning_rate": 1.6884353341831678e-06, "loss": 0.6408, "step": 15545 }, { "epoch": 1.85, "learning_rate": 1.6881313249442553e-06, "loss": 0.8199, "step": 15546 }, { "epoch": 1.85, "learning_rate": 1.6878273291255994e-06, "loss": 0.7225, "step": 15547 }, { "epoch": 1.85, "learning_rate": 1.687523346732227e-06, "loss": 0.4895, "step": 15548 }, { "epoch": 1.85, "learning_rate": 1.6872193777691614e-06, "loss": 0.5927, "step": 15549 }, { "epoch": 1.85, "learning_rate": 1.6869154222414274e-06, "loss": 0.5597, "step": 15550 }, { "epoch": 1.85, "learning_rate": 1.6866114801540499e-06, "loss": 0.671, "step": 15551 }, { "epoch": 1.85, "learning_rate": 1.6863075515120537e-06, "loss": 0.6647, "step": 15552 }, { "epoch": 1.85, "learning_rate": 1.6860036363204606e-06, "loss": 0.7463, "step": 15553 }, { "epoch": 1.85, "learning_rate": 1.6856997345842964e-06, "loss": 0.4624, "step": 15554 }, { "epoch": 1.85, "learning_rate": 1.6853958463085834e-06, "loss": 0.5284, "step": 15555 }, { "epoch": 1.85, "learning_rate": 1.6850919714983463e-06, "loss": 0.6719, "step": 15556 }, { "epoch": 1.85, "learning_rate": 1.684788110158607e-06, "loss": 0.5978, "step": 15557 }, { "epoch": 1.85, "learning_rate": 1.684484262294388e-06, "loss": 0.7469, "step": 15558 }, { "epoch": 1.85, "learning_rate": 1.6841804279107135e-06, "loss": 0.8632, "step": 15559 }, { "epoch": 1.85, "learning_rate": 1.6838766070126039e-06, "loss": 0.5601, "step": 15560 }, { "epoch": 1.85, "learning_rate": 1.6835727996050832e-06, "loss": 0.5519, "step": 15561 }, { "epoch": 1.85, "learning_rate": 1.6832690056931721e-06, "loss": 0.553, "step": 15562 }, { "epoch": 1.85, "learning_rate": 1.682965225281894e-06, "loss": 0.5634, "step": 15563 }, { "epoch": 1.85, "learning_rate": 1.6826614583762679e-06, "loss": 0.735, "step": 15564 }, { "epoch": 1.85, "learning_rate": 1.6823577049813172e-06, "loss": 0.8034, "step": 15565 }, { "epoch": 1.85, "learning_rate": 1.6820539651020617e-06, "loss": 0.5684, "step": 15566 }, { "epoch": 1.85, "learning_rate": 1.681750238743524e-06, "loss": 0.4704, "step": 15567 }, { "epoch": 1.85, "learning_rate": 1.6814465259107228e-06, "loss": 0.5826, "step": 15568 }, { "epoch": 1.85, "learning_rate": 1.6811428266086787e-06, "loss": 0.5629, "step": 15569 }, { "epoch": 1.85, "learning_rate": 1.6808391408424123e-06, "loss": 0.6297, "step": 15570 }, { "epoch": 1.85, "learning_rate": 1.680535468616945e-06, "loss": 0.7667, "step": 15571 }, { "epoch": 1.85, "learning_rate": 1.6802318099372944e-06, "loss": 0.8271, "step": 15572 }, { "epoch": 1.85, "learning_rate": 1.6799281648084803e-06, "loss": 0.4545, "step": 15573 }, { "epoch": 1.85, "learning_rate": 1.6796245332355236e-06, "loss": 0.5969, "step": 15574 }, { "epoch": 1.85, "learning_rate": 1.6793209152234408e-06, "loss": 0.5955, "step": 15575 }, { "epoch": 1.85, "learning_rate": 1.6790173107772527e-06, "loss": 0.6244, "step": 15576 }, { "epoch": 1.86, "learning_rate": 1.678713719901977e-06, "loss": 0.7855, "step": 15577 }, { "epoch": 1.86, "learning_rate": 1.678410142602634e-06, "loss": 0.7177, "step": 15578 }, { "epoch": 1.86, "learning_rate": 1.6781065788842389e-06, "loss": 0.4118, "step": 15579 }, { "epoch": 1.86, "learning_rate": 1.677803028751811e-06, "loss": 0.506, "step": 15580 }, { "epoch": 1.86, "learning_rate": 1.6774994922103677e-06, "loss": 0.5885, "step": 15581 }, { "epoch": 1.86, "learning_rate": 1.677195969264928e-06, "loss": 0.72, "step": 15582 }, { "epoch": 1.86, "learning_rate": 1.6768924599205072e-06, "loss": 0.7695, "step": 15583 }, { "epoch": 1.86, "learning_rate": 1.676588964182123e-06, "loss": 0.8866, "step": 15584 }, { "epoch": 1.86, "learning_rate": 1.6762854820547925e-06, "loss": 0.3833, "step": 15585 }, { "epoch": 1.86, "learning_rate": 1.6759820135435328e-06, "loss": 0.4967, "step": 15586 }, { "epoch": 1.86, "learning_rate": 1.6756785586533587e-06, "loss": 0.6382, "step": 15587 }, { "epoch": 1.86, "learning_rate": 1.6753751173892867e-06, "loss": 0.6446, "step": 15588 }, { "epoch": 1.86, "learning_rate": 1.6750716897563345e-06, "loss": 0.7736, "step": 15589 }, { "epoch": 1.86, "learning_rate": 1.6747682757595163e-06, "loss": 0.8543, "step": 15590 }, { "epoch": 1.86, "learning_rate": 1.6744648754038467e-06, "loss": 0.4891, "step": 15591 }, { "epoch": 1.86, "learning_rate": 1.6741614886943422e-06, "loss": 0.4445, "step": 15592 }, { "epoch": 1.86, "learning_rate": 1.6738581156360184e-06, "loss": 0.527, "step": 15593 }, { "epoch": 1.86, "learning_rate": 1.673554756233889e-06, "loss": 0.6114, "step": 15594 }, { "epoch": 1.86, "learning_rate": 1.6732514104929684e-06, "loss": 0.7041, "step": 15595 }, { "epoch": 1.86, "learning_rate": 1.6729480784182717e-06, "loss": 0.7888, "step": 15596 }, { "epoch": 1.86, "learning_rate": 1.6726447600148132e-06, "loss": 0.7999, "step": 15597 }, { "epoch": 1.86, "learning_rate": 1.6723414552876052e-06, "loss": 0.4654, "step": 15598 }, { "epoch": 1.86, "learning_rate": 1.6720381642416623e-06, "loss": 0.4287, "step": 15599 }, { "epoch": 1.86, "learning_rate": 1.6717348868819988e-06, "loss": 0.516, "step": 15600 }, { "epoch": 1.86, "learning_rate": 1.6714316232136275e-06, "loss": 0.6979, "step": 15601 }, { "epoch": 1.86, "learning_rate": 1.6711283732415597e-06, "loss": 0.7197, "step": 15602 }, { "epoch": 1.86, "learning_rate": 1.6708251369708104e-06, "loss": 0.7812, "step": 15603 }, { "epoch": 1.86, "learning_rate": 1.6705219144063903e-06, "loss": 0.3691, "step": 15604 }, { "epoch": 1.86, "learning_rate": 1.6702187055533136e-06, "loss": 0.5087, "step": 15605 }, { "epoch": 1.86, "learning_rate": 1.6699155104165903e-06, "loss": 0.497, "step": 15606 }, { "epoch": 1.86, "learning_rate": 1.669612329001234e-06, "loss": 0.5436, "step": 15607 }, { "epoch": 1.86, "learning_rate": 1.6693091613122559e-06, "loss": 0.8363, "step": 15608 }, { "epoch": 1.86, "learning_rate": 1.6690060073546664e-06, "loss": 0.8272, "step": 15609 }, { "epoch": 1.86, "learning_rate": 1.6687028671334769e-06, "loss": 0.4469, "step": 15610 }, { "epoch": 1.86, "learning_rate": 1.6683997406536995e-06, "loss": 0.3931, "step": 15611 }, { "epoch": 1.86, "learning_rate": 1.6680966279203444e-06, "loss": 0.5825, "step": 15612 }, { "epoch": 1.86, "learning_rate": 1.6677935289384212e-06, "loss": 0.5784, "step": 15613 }, { "epoch": 1.86, "learning_rate": 1.6674904437129407e-06, "loss": 0.5056, "step": 15614 }, { "epoch": 1.86, "learning_rate": 1.667187372248913e-06, "loss": 0.8394, "step": 15615 }, { "epoch": 1.86, "learning_rate": 1.6668843145513491e-06, "loss": 0.7635, "step": 15616 }, { "epoch": 1.86, "learning_rate": 1.6665812706252559e-06, "loss": 0.4526, "step": 15617 }, { "epoch": 1.86, "learning_rate": 1.666278240475645e-06, "loss": 0.5999, "step": 15618 }, { "epoch": 1.86, "learning_rate": 1.6659752241075242e-06, "loss": 0.5512, "step": 15619 }, { "epoch": 1.86, "learning_rate": 1.6656722215259038e-06, "loss": 0.618, "step": 15620 }, { "epoch": 1.86, "learning_rate": 1.6653692327357906e-06, "loss": 0.8502, "step": 15621 }, { "epoch": 1.86, "learning_rate": 1.6650662577421947e-06, "loss": 0.7924, "step": 15622 }, { "epoch": 1.86, "learning_rate": 1.6647632965501242e-06, "loss": 0.4047, "step": 15623 }, { "epoch": 1.86, "learning_rate": 1.6644603491645852e-06, "loss": 0.5265, "step": 15624 }, { "epoch": 1.86, "learning_rate": 1.6641574155905873e-06, "loss": 0.6314, "step": 15625 }, { "epoch": 1.86, "learning_rate": 1.663854495833137e-06, "loss": 0.7009, "step": 15626 }, { "epoch": 1.86, "learning_rate": 1.6635515898972432e-06, "loss": 0.7675, "step": 15627 }, { "epoch": 1.86, "learning_rate": 1.6632486977879106e-06, "loss": 0.8325, "step": 15628 }, { "epoch": 1.86, "learning_rate": 1.6629458195101477e-06, "loss": 0.3875, "step": 15629 }, { "epoch": 1.86, "learning_rate": 1.6626429550689602e-06, "loss": 0.5414, "step": 15630 }, { "epoch": 1.86, "learning_rate": 1.662340104469356e-06, "loss": 0.7023, "step": 15631 }, { "epoch": 1.86, "learning_rate": 1.6620372677163397e-06, "loss": 0.6052, "step": 15632 }, { "epoch": 1.86, "learning_rate": 1.661734444814917e-06, "loss": 0.6983, "step": 15633 }, { "epoch": 1.86, "learning_rate": 1.6614316357700944e-06, "loss": 0.9891, "step": 15634 }, { "epoch": 1.86, "learning_rate": 1.661128840586878e-06, "loss": 0.4038, "step": 15635 }, { "epoch": 1.86, "learning_rate": 1.660826059270272e-06, "loss": 0.5414, "step": 15636 }, { "epoch": 1.86, "learning_rate": 1.6605232918252808e-06, "loss": 0.5673, "step": 15637 }, { "epoch": 1.86, "learning_rate": 1.6602205382569114e-06, "loss": 0.6039, "step": 15638 }, { "epoch": 1.86, "learning_rate": 1.6599177985701658e-06, "loss": 0.7164, "step": 15639 }, { "epoch": 1.86, "learning_rate": 1.65961507277005e-06, "loss": 0.8489, "step": 15640 }, { "epoch": 1.86, "learning_rate": 1.659312360861567e-06, "loss": 0.6149, "step": 15641 }, { "epoch": 1.86, "learning_rate": 1.6590096628497227e-06, "loss": 0.4789, "step": 15642 }, { "epoch": 1.86, "learning_rate": 1.6587069787395182e-06, "loss": 0.5016, "step": 15643 }, { "epoch": 1.86, "learning_rate": 1.6584043085359574e-06, "loss": 0.6255, "step": 15644 }, { "epoch": 1.86, "learning_rate": 1.658101652244044e-06, "loss": 0.815, "step": 15645 }, { "epoch": 1.86, "learning_rate": 1.6577990098687824e-06, "loss": 0.8477, "step": 15646 }, { "epoch": 1.86, "learning_rate": 1.657496381415173e-06, "loss": 0.7873, "step": 15647 }, { "epoch": 1.86, "learning_rate": 1.6571937668882187e-06, "loss": 0.438, "step": 15648 }, { "epoch": 1.86, "learning_rate": 1.656891166292922e-06, "loss": 0.4838, "step": 15649 }, { "epoch": 1.86, "learning_rate": 1.6565885796342861e-06, "loss": 0.5625, "step": 15650 }, { "epoch": 1.86, "learning_rate": 1.6562860069173115e-06, "loss": 0.6333, "step": 15651 }, { "epoch": 1.86, "learning_rate": 1.6559834481469992e-06, "loss": 0.7079, "step": 15652 }, { "epoch": 1.86, "learning_rate": 1.6556809033283523e-06, "loss": 0.8474, "step": 15653 }, { "epoch": 1.86, "learning_rate": 1.6553783724663708e-06, "loss": 0.4207, "step": 15654 }, { "epoch": 1.86, "learning_rate": 1.6550758555660546e-06, "loss": 0.4958, "step": 15655 }, { "epoch": 1.86, "learning_rate": 1.6547733526324055e-06, "loss": 0.5365, "step": 15656 }, { "epoch": 1.86, "learning_rate": 1.6544708636704254e-06, "loss": 0.5705, "step": 15657 }, { "epoch": 1.86, "learning_rate": 1.6541683886851117e-06, "loss": 0.791, "step": 15658 }, { "epoch": 1.86, "learning_rate": 1.6538659276814657e-06, "loss": 0.8428, "step": 15659 }, { "epoch": 1.86, "learning_rate": 1.6535634806644868e-06, "loss": 0.5899, "step": 15660 }, { "epoch": 1.87, "learning_rate": 1.6532610476391746e-06, "loss": 0.455, "step": 15661 }, { "epoch": 1.87, "learning_rate": 1.6529586286105286e-06, "loss": 0.555, "step": 15662 }, { "epoch": 1.87, "learning_rate": 1.652656223583547e-06, "loss": 0.6311, "step": 15663 }, { "epoch": 1.87, "learning_rate": 1.6523538325632293e-06, "loss": 0.7357, "step": 15664 }, { "epoch": 1.87, "learning_rate": 1.6520514555545747e-06, "loss": 0.7826, "step": 15665 }, { "epoch": 1.87, "learning_rate": 1.6517490925625798e-06, "loss": 0.3966, "step": 15666 }, { "epoch": 1.87, "learning_rate": 1.6514467435922431e-06, "loss": 0.4612, "step": 15667 }, { "epoch": 1.87, "learning_rate": 1.651144408648564e-06, "loss": 0.6334, "step": 15668 }, { "epoch": 1.87, "learning_rate": 1.6508420877365394e-06, "loss": 0.7003, "step": 15669 }, { "epoch": 1.87, "learning_rate": 1.650539780861165e-06, "loss": 0.6672, "step": 15670 }, { "epoch": 1.87, "learning_rate": 1.6502374880274402e-06, "loss": 0.7991, "step": 15671 }, { "epoch": 1.87, "learning_rate": 1.6499352092403615e-06, "loss": 0.7407, "step": 15672 }, { "epoch": 1.87, "learning_rate": 1.6496329445049237e-06, "loss": 0.4438, "step": 15673 }, { "epoch": 1.87, "learning_rate": 1.649330693826125e-06, "loss": 0.5305, "step": 15674 }, { "epoch": 1.87, "learning_rate": 1.6490284572089618e-06, "loss": 0.5604, "step": 15675 }, { "epoch": 1.87, "learning_rate": 1.64872623465843e-06, "loss": 0.5713, "step": 15676 }, { "epoch": 1.87, "learning_rate": 1.6484240261795243e-06, "loss": 0.7372, "step": 15677 }, { "epoch": 1.87, "learning_rate": 1.6481218317772402e-06, "loss": 0.8621, "step": 15678 }, { "epoch": 1.87, "learning_rate": 1.6478196514565748e-06, "loss": 0.4962, "step": 15679 }, { "epoch": 1.87, "learning_rate": 1.6475174852225222e-06, "loss": 0.5395, "step": 15680 }, { "epoch": 1.87, "learning_rate": 1.6472153330800767e-06, "loss": 0.6215, "step": 15681 }, { "epoch": 1.87, "learning_rate": 1.6469131950342332e-06, "loss": 0.688, "step": 15682 }, { "epoch": 1.87, "learning_rate": 1.6466110710899857e-06, "loss": 0.7543, "step": 15683 }, { "epoch": 1.87, "learning_rate": 1.64630896125233e-06, "loss": 0.9088, "step": 15684 }, { "epoch": 1.87, "learning_rate": 1.6460068655262578e-06, "loss": 0.5099, "step": 15685 }, { "epoch": 1.87, "learning_rate": 1.6457047839167642e-06, "loss": 0.465, "step": 15686 }, { "epoch": 1.87, "learning_rate": 1.6454027164288427e-06, "loss": 0.627, "step": 15687 }, { "epoch": 1.87, "learning_rate": 1.6451006630674851e-06, "loss": 0.5802, "step": 15688 }, { "epoch": 1.87, "learning_rate": 1.6447986238376856e-06, "loss": 0.6733, "step": 15689 }, { "epoch": 1.87, "learning_rate": 1.6444965987444361e-06, "loss": 0.6943, "step": 15690 }, { "epoch": 1.87, "learning_rate": 1.644194587792731e-06, "loss": 0.4575, "step": 15691 }, { "epoch": 1.87, "learning_rate": 1.6438925909875598e-06, "loss": 0.4728, "step": 15692 }, { "epoch": 1.87, "learning_rate": 1.6435906083339165e-06, "loss": 0.5261, "step": 15693 }, { "epoch": 1.87, "learning_rate": 1.6432886398367917e-06, "loss": 0.5627, "step": 15694 }, { "epoch": 1.87, "learning_rate": 1.6429866855011789e-06, "loss": 0.7241, "step": 15695 }, { "epoch": 1.87, "learning_rate": 1.6426847453320666e-06, "loss": 0.8026, "step": 15696 }, { "epoch": 1.87, "learning_rate": 1.6423828193344482e-06, "loss": 0.871, "step": 15697 }, { "epoch": 1.87, "learning_rate": 1.6420809075133132e-06, "loss": 0.4082, "step": 15698 }, { "epoch": 1.87, "learning_rate": 1.6417790098736538e-06, "loss": 0.6189, "step": 15699 }, { "epoch": 1.87, "learning_rate": 1.6414771264204588e-06, "loss": 0.5804, "step": 15700 }, { "epoch": 1.87, "learning_rate": 1.6411752571587187e-06, "loss": 0.6607, "step": 15701 }, { "epoch": 1.87, "learning_rate": 1.6408734020934245e-06, "loss": 0.847, "step": 15702 }, { "epoch": 1.87, "learning_rate": 1.6405715612295638e-06, "loss": 0.7464, "step": 15703 }, { "epoch": 1.87, "learning_rate": 1.640269734572128e-06, "loss": 0.3735, "step": 15704 }, { "epoch": 1.87, "learning_rate": 1.6399679221261053e-06, "loss": 0.6123, "step": 15705 }, { "epoch": 1.87, "learning_rate": 1.6396661238964857e-06, "loss": 0.5786, "step": 15706 }, { "epoch": 1.87, "learning_rate": 1.6393643398882564e-06, "loss": 0.5122, "step": 15707 }, { "epoch": 1.87, "learning_rate": 1.6390625701064073e-06, "loss": 0.7202, "step": 15708 }, { "epoch": 1.87, "learning_rate": 1.6387608145559258e-06, "loss": 0.777, "step": 15709 }, { "epoch": 1.87, "learning_rate": 1.638459073241801e-06, "loss": 0.326, "step": 15710 }, { "epoch": 1.87, "learning_rate": 1.6381573461690198e-06, "loss": 0.4393, "step": 15711 }, { "epoch": 1.87, "learning_rate": 1.6378556333425693e-06, "loss": 0.4945, "step": 15712 }, { "epoch": 1.87, "learning_rate": 1.6375539347674374e-06, "loss": 0.5854, "step": 15713 }, { "epoch": 1.87, "learning_rate": 1.6372522504486128e-06, "loss": 0.6577, "step": 15714 }, { "epoch": 1.87, "learning_rate": 1.6369505803910798e-06, "loss": 0.9292, "step": 15715 }, { "epoch": 1.87, "learning_rate": 1.636648924599826e-06, "loss": 0.4729, "step": 15716 }, { "epoch": 1.87, "learning_rate": 1.6363472830798388e-06, "loss": 0.4652, "step": 15717 }, { "epoch": 1.87, "learning_rate": 1.6360456558361026e-06, "loss": 0.5901, "step": 15718 }, { "epoch": 1.87, "learning_rate": 1.6357440428736043e-06, "loss": 0.5811, "step": 15719 }, { "epoch": 1.87, "learning_rate": 1.6354424441973293e-06, "loss": 0.6341, "step": 15720 }, { "epoch": 1.87, "learning_rate": 1.6351408598122643e-06, "loss": 0.7762, "step": 15721 }, { "epoch": 1.87, "learning_rate": 1.6348392897233927e-06, "loss": 0.9998, "step": 15722 }, { "epoch": 1.87, "learning_rate": 1.6345377339356999e-06, "loss": 0.4244, "step": 15723 }, { "epoch": 1.87, "learning_rate": 1.6342361924541712e-06, "loss": 0.5559, "step": 15724 }, { "epoch": 1.87, "learning_rate": 1.6339346652837918e-06, "loss": 0.6481, "step": 15725 }, { "epoch": 1.87, "learning_rate": 1.6336331524295443e-06, "loss": 0.5734, "step": 15726 }, { "epoch": 1.87, "learning_rate": 1.633331653896413e-06, "loss": 0.6886, "step": 15727 }, { "epoch": 1.87, "learning_rate": 1.6330301696893829e-06, "loss": 0.8569, "step": 15728 }, { "epoch": 1.87, "learning_rate": 1.6327286998134374e-06, "loss": 0.4982, "step": 15729 }, { "epoch": 1.87, "learning_rate": 1.6324272442735577e-06, "loss": 0.5467, "step": 15730 }, { "epoch": 1.87, "learning_rate": 1.632125803074729e-06, "loss": 0.6049, "step": 15731 }, { "epoch": 1.87, "learning_rate": 1.6318243762219344e-06, "loss": 0.5653, "step": 15732 }, { "epoch": 1.87, "learning_rate": 1.6315229637201552e-06, "loss": 0.7074, "step": 15733 }, { "epoch": 1.87, "learning_rate": 1.6312215655743738e-06, "loss": 0.8347, "step": 15734 }, { "epoch": 1.87, "learning_rate": 1.6309201817895727e-06, "loss": 0.454, "step": 15735 }, { "epoch": 1.87, "learning_rate": 1.630618812370735e-06, "loss": 0.4981, "step": 15736 }, { "epoch": 1.87, "learning_rate": 1.6303174573228409e-06, "loss": 0.563, "step": 15737 }, { "epoch": 1.87, "learning_rate": 1.6300161166508716e-06, "loss": 0.6256, "step": 15738 }, { "epoch": 1.87, "learning_rate": 1.6297147903598095e-06, "loss": 0.6802, "step": 15739 }, { "epoch": 1.87, "learning_rate": 1.6294134784546356e-06, "loss": 0.7965, "step": 15740 }, { "epoch": 1.87, "learning_rate": 1.6291121809403287e-06, "loss": 0.3942, "step": 15741 }, { "epoch": 1.87, "learning_rate": 1.6288108978218704e-06, "loss": 0.3731, "step": 15742 }, { "epoch": 1.87, "learning_rate": 1.6285096291042418e-06, "loss": 0.5775, "step": 15743 }, { "epoch": 1.87, "learning_rate": 1.6282083747924224e-06, "loss": 0.5805, "step": 15744 }, { "epoch": 1.88, "learning_rate": 1.6279071348913907e-06, "loss": 0.7219, "step": 15745 }, { "epoch": 1.88, "learning_rate": 1.627605909406128e-06, "loss": 0.7762, "step": 15746 }, { "epoch": 1.88, "learning_rate": 1.6273046983416124e-06, "loss": 0.8706, "step": 15747 }, { "epoch": 1.88, "learning_rate": 1.627003501702824e-06, "loss": 0.5248, "step": 15748 }, { "epoch": 1.88, "learning_rate": 1.6267023194947402e-06, "loss": 0.6258, "step": 15749 }, { "epoch": 1.88, "learning_rate": 1.6264011517223409e-06, "loss": 0.6161, "step": 15750 }, { "epoch": 1.88, "learning_rate": 1.626099998390604e-06, "loss": 0.6538, "step": 15751 }, { "epoch": 1.88, "learning_rate": 1.6257988595045065e-06, "loss": 0.64, "step": 15752 }, { "epoch": 1.88, "learning_rate": 1.6254977350690276e-06, "loss": 0.8811, "step": 15753 }, { "epoch": 1.88, "learning_rate": 1.625196625089145e-06, "loss": 0.4045, "step": 15754 }, { "epoch": 1.88, "learning_rate": 1.6248955295698358e-06, "loss": 0.553, "step": 15755 }, { "epoch": 1.88, "learning_rate": 1.6245944485160758e-06, "loss": 0.5909, "step": 15756 }, { "epoch": 1.88, "learning_rate": 1.6242933819328438e-06, "loss": 0.6569, "step": 15757 }, { "epoch": 1.88, "learning_rate": 1.6239923298251154e-06, "loss": 0.7081, "step": 15758 }, { "epoch": 1.88, "learning_rate": 1.623691292197868e-06, "loss": 0.7972, "step": 15759 }, { "epoch": 1.88, "learning_rate": 1.623390269056076e-06, "loss": 0.2932, "step": 15760 }, { "epoch": 1.88, "learning_rate": 1.623089260404717e-06, "loss": 0.4747, "step": 15761 }, { "epoch": 1.88, "learning_rate": 1.6227882662487655e-06, "loss": 0.485, "step": 15762 }, { "epoch": 1.88, "learning_rate": 1.6224872865931985e-06, "loss": 0.701, "step": 15763 }, { "epoch": 1.88, "learning_rate": 1.6221863214429895e-06, "loss": 0.6841, "step": 15764 }, { "epoch": 1.88, "learning_rate": 1.6218853708031147e-06, "loss": 0.8303, "step": 15765 }, { "epoch": 1.88, "learning_rate": 1.621584434678549e-06, "loss": 0.6734, "step": 15766 }, { "epoch": 1.88, "learning_rate": 1.621283513074265e-06, "loss": 0.5824, "step": 15767 }, { "epoch": 1.88, "learning_rate": 1.6209826059952389e-06, "loss": 0.5325, "step": 15768 }, { "epoch": 1.88, "learning_rate": 1.6206817134464436e-06, "loss": 0.6134, "step": 15769 }, { "epoch": 1.88, "learning_rate": 1.6203808354328546e-06, "loss": 0.7153, "step": 15770 }, { "epoch": 1.88, "learning_rate": 1.620079971959443e-06, "loss": 0.7083, "step": 15771 }, { "epoch": 1.88, "learning_rate": 1.6197791230311838e-06, "loss": 0.9675, "step": 15772 }, { "epoch": 1.88, "learning_rate": 1.6194782886530488e-06, "loss": 0.4533, "step": 15773 }, { "epoch": 1.88, "learning_rate": 1.619177468830013e-06, "loss": 0.4811, "step": 15774 }, { "epoch": 1.88, "learning_rate": 1.618876663567046e-06, "loss": 0.6145, "step": 15775 }, { "epoch": 1.88, "learning_rate": 1.618575872869123e-06, "loss": 0.6592, "step": 15776 }, { "epoch": 1.88, "learning_rate": 1.6182750967412137e-06, "loss": 0.7648, "step": 15777 }, { "epoch": 1.88, "learning_rate": 1.617974335188292e-06, "loss": 0.8289, "step": 15778 }, { "epoch": 1.88, "learning_rate": 1.6176735882153284e-06, "loss": 0.5094, "step": 15779 }, { "epoch": 1.88, "learning_rate": 1.6173728558272935e-06, "loss": 0.5984, "step": 15780 }, { "epoch": 1.88, "learning_rate": 1.6170721380291605e-06, "loss": 0.5306, "step": 15781 }, { "epoch": 1.88, "learning_rate": 1.6167714348258985e-06, "loss": 0.6566, "step": 15782 }, { "epoch": 1.88, "learning_rate": 1.6164707462224788e-06, "loss": 0.7084, "step": 15783 }, { "epoch": 1.88, "learning_rate": 1.6161700722238716e-06, "loss": 0.9261, "step": 15784 }, { "epoch": 1.88, "learning_rate": 1.615869412835048e-06, "loss": 0.5989, "step": 15785 }, { "epoch": 1.88, "learning_rate": 1.6155687680609769e-06, "loss": 0.4704, "step": 15786 }, { "epoch": 1.88, "learning_rate": 1.6152681379066276e-06, "loss": 0.5435, "step": 15787 }, { "epoch": 1.88, "learning_rate": 1.6149675223769707e-06, "loss": 0.6664, "step": 15788 }, { "epoch": 1.88, "learning_rate": 1.6146669214769752e-06, "loss": 0.5877, "step": 15789 }, { "epoch": 1.88, "learning_rate": 1.6143663352116095e-06, "loss": 0.8811, "step": 15790 }, { "epoch": 1.88, "learning_rate": 1.6140657635858426e-06, "loss": 0.6262, "step": 15791 }, { "epoch": 1.88, "learning_rate": 1.6137652066046425e-06, "loss": 0.4135, "step": 15792 }, { "epoch": 1.88, "learning_rate": 1.613464664272979e-06, "loss": 0.5276, "step": 15793 }, { "epoch": 1.88, "learning_rate": 1.6131641365958184e-06, "loss": 0.5659, "step": 15794 }, { "epoch": 1.88, "learning_rate": 1.6128636235781287e-06, "loss": 0.7273, "step": 15795 }, { "epoch": 1.88, "learning_rate": 1.6125631252248786e-06, "loss": 0.7319, "step": 15796 }, { "epoch": 1.88, "learning_rate": 1.6122626415410342e-06, "loss": 0.7518, "step": 15797 }, { "epoch": 1.88, "learning_rate": 1.611962172531562e-06, "loss": 0.4728, "step": 15798 }, { "epoch": 1.88, "learning_rate": 1.6116617182014297e-06, "loss": 0.607, "step": 15799 }, { "epoch": 1.88, "learning_rate": 1.6113612785556048e-06, "loss": 0.6759, "step": 15800 }, { "epoch": 1.88, "learning_rate": 1.6110608535990519e-06, "loss": 0.7919, "step": 15801 }, { "epoch": 1.88, "learning_rate": 1.6107604433367374e-06, "loss": 0.8002, "step": 15802 }, { "epoch": 1.88, "learning_rate": 1.6104600477736277e-06, "loss": 0.8552, "step": 15803 }, { "epoch": 1.88, "learning_rate": 1.6101596669146883e-06, "loss": 0.47, "step": 15804 }, { "epoch": 1.88, "learning_rate": 1.609859300764884e-06, "loss": 0.4945, "step": 15805 }, { "epoch": 1.88, "learning_rate": 1.6095589493291797e-06, "loss": 0.4848, "step": 15806 }, { "epoch": 1.88, "learning_rate": 1.6092586126125414e-06, "loss": 0.6323, "step": 15807 }, { "epoch": 1.88, "learning_rate": 1.6089582906199333e-06, "loss": 0.7945, "step": 15808 }, { "epoch": 1.88, "learning_rate": 1.6086579833563185e-06, "loss": 0.7382, "step": 15809 }, { "epoch": 1.88, "learning_rate": 1.6083576908266618e-06, "loss": 0.3697, "step": 15810 }, { "epoch": 1.88, "learning_rate": 1.608057413035928e-06, "loss": 0.515, "step": 15811 }, { "epoch": 1.88, "learning_rate": 1.6077571499890803e-06, "loss": 0.5714, "step": 15812 }, { "epoch": 1.88, "learning_rate": 1.6074569016910812e-06, "loss": 0.626, "step": 15813 }, { "epoch": 1.88, "learning_rate": 1.6071566681468946e-06, "loss": 0.7982, "step": 15814 }, { "epoch": 1.88, "learning_rate": 1.6068564493614835e-06, "loss": 0.8296, "step": 15815 }, { "epoch": 1.88, "learning_rate": 1.6065562453398095e-06, "loss": 0.545, "step": 15816 }, { "epoch": 1.88, "learning_rate": 1.6062560560868357e-06, "loss": 0.4432, "step": 15817 }, { "epoch": 1.88, "learning_rate": 1.605955881607525e-06, "loss": 0.5272, "step": 15818 }, { "epoch": 1.88, "learning_rate": 1.6056557219068387e-06, "loss": 0.649, "step": 15819 }, { "epoch": 1.88, "learning_rate": 1.6053555769897376e-06, "loss": 0.7352, "step": 15820 }, { "epoch": 1.88, "learning_rate": 1.6050554468611837e-06, "loss": 0.7858, "step": 15821 }, { "epoch": 1.88, "learning_rate": 1.6047553315261389e-06, "loss": 0.7491, "step": 15822 }, { "epoch": 1.88, "learning_rate": 1.604455230989564e-06, "loss": 0.3567, "step": 15823 }, { "epoch": 1.88, "learning_rate": 1.6041551452564185e-06, "loss": 0.6519, "step": 15824 }, { "epoch": 1.88, "learning_rate": 1.603855074331664e-06, "loss": 0.6734, "step": 15825 }, { "epoch": 1.88, "learning_rate": 1.60355501822026e-06, "loss": 0.7112, "step": 15826 }, { "epoch": 1.88, "learning_rate": 1.6032549769271677e-06, "loss": 0.8714, "step": 15827 }, { "epoch": 1.88, "learning_rate": 1.6029549504573446e-06, "loss": 0.8294, "step": 15828 }, { "epoch": 1.89, "learning_rate": 1.6026549388157522e-06, "loss": 0.4381, "step": 15829 }, { "epoch": 1.89, "learning_rate": 1.6023549420073492e-06, "loss": 0.5359, "step": 15830 }, { "epoch": 1.89, "learning_rate": 1.6020549600370939e-06, "loss": 0.5612, "step": 15831 }, { "epoch": 1.89, "learning_rate": 1.6017549929099451e-06, "loss": 0.5023, "step": 15832 }, { "epoch": 1.89, "learning_rate": 1.6014550406308619e-06, "loss": 0.7686, "step": 15833 }, { "epoch": 1.89, "learning_rate": 1.601155103204803e-06, "loss": 0.8597, "step": 15834 }, { "epoch": 1.89, "learning_rate": 1.6008551806367244e-06, "loss": 0.5666, "step": 15835 }, { "epoch": 1.89, "learning_rate": 1.600555272931586e-06, "loss": 0.5915, "step": 15836 }, { "epoch": 1.89, "learning_rate": 1.6002553800943437e-06, "loss": 0.604, "step": 15837 }, { "epoch": 1.89, "learning_rate": 1.5999555021299565e-06, "loss": 0.6028, "step": 15838 }, { "epoch": 1.89, "learning_rate": 1.5996556390433793e-06, "loss": 0.7825, "step": 15839 }, { "epoch": 1.89, "learning_rate": 1.5993557908395706e-06, "loss": 0.8362, "step": 15840 }, { "epoch": 1.89, "learning_rate": 1.5990559575234857e-06, "loss": 0.6981, "step": 15841 }, { "epoch": 1.89, "learning_rate": 1.598756139100082e-06, "loss": 0.5023, "step": 15842 }, { "epoch": 1.89, "learning_rate": 1.5984563355743152e-06, "loss": 0.4769, "step": 15843 }, { "epoch": 1.89, "learning_rate": 1.5981565469511398e-06, "loss": 0.6467, "step": 15844 }, { "epoch": 1.89, "learning_rate": 1.5978567732355135e-06, "loss": 0.5152, "step": 15845 }, { "epoch": 1.89, "learning_rate": 1.5975570144323892e-06, "loss": 0.827, "step": 15846 }, { "epoch": 1.89, "learning_rate": 1.5972572705467243e-06, "loss": 0.932, "step": 15847 }, { "epoch": 1.89, "learning_rate": 1.5969575415834714e-06, "loss": 0.4759, "step": 15848 }, { "epoch": 1.89, "learning_rate": 1.5966578275475874e-06, "loss": 0.5197, "step": 15849 }, { "epoch": 1.89, "learning_rate": 1.596358128444024e-06, "loss": 0.525, "step": 15850 }, { "epoch": 1.89, "learning_rate": 1.5960584442777371e-06, "loss": 0.5982, "step": 15851 }, { "epoch": 1.89, "learning_rate": 1.5957587750536795e-06, "loss": 0.7448, "step": 15852 }, { "epoch": 1.89, "learning_rate": 1.5954591207768067e-06, "loss": 0.793, "step": 15853 }, { "epoch": 1.89, "learning_rate": 1.5951594814520696e-06, "loss": 0.5581, "step": 15854 }, { "epoch": 1.89, "learning_rate": 1.5948598570844218e-06, "loss": 0.4581, "step": 15855 }, { "epoch": 1.89, "learning_rate": 1.5945602476788163e-06, "loss": 0.5994, "step": 15856 }, { "epoch": 1.89, "learning_rate": 1.5942606532402071e-06, "loss": 0.6797, "step": 15857 }, { "epoch": 1.89, "learning_rate": 1.5939610737735448e-06, "loss": 0.7362, "step": 15858 }, { "epoch": 1.89, "learning_rate": 1.5936615092837816e-06, "loss": 0.7788, "step": 15859 }, { "epoch": 1.89, "learning_rate": 1.5933619597758704e-06, "loss": 0.5517, "step": 15860 }, { "epoch": 1.89, "learning_rate": 1.593062425254761e-06, "loss": 0.494, "step": 15861 }, { "epoch": 1.89, "learning_rate": 1.5927629057254065e-06, "loss": 0.5906, "step": 15862 }, { "epoch": 1.89, "learning_rate": 1.5924634011927567e-06, "loss": 0.6235, "step": 15863 }, { "epoch": 1.89, "learning_rate": 1.5921639116617642e-06, "loss": 0.649, "step": 15864 }, { "epoch": 1.89, "learning_rate": 1.591864437137378e-06, "loss": 0.9338, "step": 15865 }, { "epoch": 1.89, "learning_rate": 1.591564977624548e-06, "loss": 0.6374, "step": 15866 }, { "epoch": 1.89, "learning_rate": 1.5912655331282253e-06, "loss": 0.4055, "step": 15867 }, { "epoch": 1.89, "learning_rate": 1.5909661036533608e-06, "loss": 0.5912, "step": 15868 }, { "epoch": 1.89, "learning_rate": 1.5906666892049022e-06, "loss": 0.6055, "step": 15869 }, { "epoch": 1.89, "learning_rate": 1.5903672897877985e-06, "loss": 0.6879, "step": 15870 }, { "epoch": 1.89, "learning_rate": 1.5900679054070012e-06, "loss": 0.7866, "step": 15871 }, { "epoch": 1.89, "learning_rate": 1.5897685360674581e-06, "loss": 0.8085, "step": 15872 }, { "epoch": 1.89, "learning_rate": 1.589469181774116e-06, "loss": 0.5049, "step": 15873 }, { "epoch": 1.89, "learning_rate": 1.589169842531925e-06, "loss": 0.5393, "step": 15874 }, { "epoch": 1.89, "learning_rate": 1.5888705183458342e-06, "loss": 0.5547, "step": 15875 }, { "epoch": 1.89, "learning_rate": 1.588571209220789e-06, "loss": 0.6536, "step": 15876 }, { "epoch": 1.89, "learning_rate": 1.5882719151617382e-06, "loss": 0.8044, "step": 15877 }, { "epoch": 1.89, "learning_rate": 1.5879726361736292e-06, "loss": 0.9174, "step": 15878 }, { "epoch": 1.89, "learning_rate": 1.5876733722614099e-06, "loss": 0.3338, "step": 15879 }, { "epoch": 1.89, "learning_rate": 1.5873741234300258e-06, "loss": 0.5569, "step": 15880 }, { "epoch": 1.89, "learning_rate": 1.5870748896844235e-06, "loss": 0.4926, "step": 15881 }, { "epoch": 1.89, "learning_rate": 1.5867756710295504e-06, "loss": 0.5536, "step": 15882 }, { "epoch": 1.89, "learning_rate": 1.5864764674703526e-06, "loss": 0.8359, "step": 15883 }, { "epoch": 1.89, "learning_rate": 1.5861772790117744e-06, "loss": 0.8179, "step": 15884 }, { "epoch": 1.89, "learning_rate": 1.5858781056587625e-06, "loss": 0.5215, "step": 15885 }, { "epoch": 1.89, "learning_rate": 1.5855789474162626e-06, "loss": 0.3962, "step": 15886 }, { "epoch": 1.89, "learning_rate": 1.5852798042892201e-06, "loss": 0.4902, "step": 15887 }, { "epoch": 1.89, "learning_rate": 1.584980676282578e-06, "loss": 0.566, "step": 15888 }, { "epoch": 1.89, "learning_rate": 1.5846815634012827e-06, "loss": 0.738, "step": 15889 }, { "epoch": 1.89, "learning_rate": 1.5843824656502776e-06, "loss": 0.7512, "step": 15890 }, { "epoch": 1.89, "learning_rate": 1.584083383034508e-06, "loss": 0.4507, "step": 15891 }, { "epoch": 1.89, "learning_rate": 1.5837843155589161e-06, "loss": 0.4335, "step": 15892 }, { "epoch": 1.89, "learning_rate": 1.5834852632284464e-06, "loss": 0.4829, "step": 15893 }, { "epoch": 1.89, "learning_rate": 1.583186226048043e-06, "loss": 0.6544, "step": 15894 }, { "epoch": 1.89, "learning_rate": 1.5828872040226472e-06, "loss": 0.6801, "step": 15895 }, { "epoch": 1.89, "learning_rate": 1.582588197157203e-06, "loss": 0.7573, "step": 15896 }, { "epoch": 1.89, "learning_rate": 1.5822892054566532e-06, "loss": 0.8438, "step": 15897 }, { "epoch": 1.89, "learning_rate": 1.5819902289259404e-06, "loss": 0.4023, "step": 15898 }, { "epoch": 1.89, "learning_rate": 1.5816912675700053e-06, "loss": 0.4944, "step": 15899 }, { "epoch": 1.89, "learning_rate": 1.581392321393791e-06, "loss": 0.4976, "step": 15900 }, { "epoch": 1.89, "learning_rate": 1.5810933904022379e-06, "loss": 0.6464, "step": 15901 }, { "epoch": 1.89, "learning_rate": 1.5807944746002895e-06, "loss": 0.7297, "step": 15902 }, { "epoch": 1.89, "learning_rate": 1.5804955739928845e-06, "loss": 0.8923, "step": 15903 }, { "epoch": 1.89, "learning_rate": 1.5801966885849652e-06, "loss": 0.363, "step": 15904 }, { "epoch": 1.89, "learning_rate": 1.5798978183814715e-06, "loss": 0.4452, "step": 15905 }, { "epoch": 1.89, "learning_rate": 1.5795989633873448e-06, "loss": 0.6632, "step": 15906 }, { "epoch": 1.89, "learning_rate": 1.5793001236075236e-06, "loss": 0.6575, "step": 15907 }, { "epoch": 1.89, "learning_rate": 1.5790012990469491e-06, "loss": 0.8189, "step": 15908 }, { "epoch": 1.89, "learning_rate": 1.578702489710561e-06, "loss": 0.9212, "step": 15909 }, { "epoch": 1.89, "learning_rate": 1.5784036956032973e-06, "loss": 0.4664, "step": 15910 }, { "epoch": 1.89, "learning_rate": 1.5781049167300977e-06, "loss": 0.479, "step": 15911 }, { "epoch": 1.89, "learning_rate": 1.577806153095901e-06, "loss": 0.554, "step": 15912 }, { "epoch": 1.9, "learning_rate": 1.5775074047056472e-06, "loss": 0.5757, "step": 15913 }, { "epoch": 1.9, "learning_rate": 1.577208671564272e-06, "loss": 0.7621, "step": 15914 }, { "epoch": 1.9, "learning_rate": 1.5769099536767157e-06, "loss": 0.8223, "step": 15915 }, { "epoch": 1.9, "learning_rate": 1.5766112510479144e-06, "loss": 0.5972, "step": 15916 }, { "epoch": 1.9, "learning_rate": 1.576312563682808e-06, "loss": 0.4766, "step": 15917 }, { "epoch": 1.9, "learning_rate": 1.5760138915863314e-06, "loss": 0.5928, "step": 15918 }, { "epoch": 1.9, "learning_rate": 1.575715234763423e-06, "loss": 0.6194, "step": 15919 }, { "epoch": 1.9, "learning_rate": 1.5754165932190191e-06, "loss": 0.6836, "step": 15920 }, { "epoch": 1.9, "learning_rate": 1.5751179669580575e-06, "loss": 0.8629, "step": 15921 }, { "epoch": 1.9, "learning_rate": 1.5748193559854724e-06, "loss": 0.9052, "step": 15922 }, { "epoch": 1.9, "learning_rate": 1.574520760306201e-06, "loss": 0.3195, "step": 15923 }, { "epoch": 1.9, "learning_rate": 1.57422217992518e-06, "loss": 0.3873, "step": 15924 }, { "epoch": 1.9, "learning_rate": 1.573923614847343e-06, "loss": 0.6043, "step": 15925 }, { "epoch": 1.9, "learning_rate": 1.573625065077627e-06, "loss": 0.5959, "step": 15926 }, { "epoch": 1.9, "learning_rate": 1.5733265306209657e-06, "loss": 0.8618, "step": 15927 }, { "epoch": 1.9, "learning_rate": 1.5730280114822958e-06, "loss": 0.7398, "step": 15928 }, { "epoch": 1.9, "learning_rate": 1.57272950766655e-06, "loss": 0.4636, "step": 15929 }, { "epoch": 1.9, "learning_rate": 1.572431019178663e-06, "loss": 0.5468, "step": 15930 }, { "epoch": 1.9, "learning_rate": 1.5721325460235688e-06, "loss": 0.6149, "step": 15931 }, { "epoch": 1.9, "learning_rate": 1.5718340882062025e-06, "loss": 0.5873, "step": 15932 }, { "epoch": 1.9, "learning_rate": 1.571535645731496e-06, "loss": 0.6679, "step": 15933 }, { "epoch": 1.9, "learning_rate": 1.5712372186043828e-06, "loss": 0.8016, "step": 15934 }, { "epoch": 1.9, "learning_rate": 1.5709388068297965e-06, "loss": 0.3382, "step": 15935 }, { "epoch": 1.9, "learning_rate": 1.5706404104126705e-06, "loss": 0.5072, "step": 15936 }, { "epoch": 1.9, "learning_rate": 1.5703420293579357e-06, "loss": 0.6205, "step": 15937 }, { "epoch": 1.9, "learning_rate": 1.5700436636705251e-06, "loss": 0.7473, "step": 15938 }, { "epoch": 1.9, "learning_rate": 1.569745313355372e-06, "loss": 0.6891, "step": 15939 }, { "epoch": 1.9, "learning_rate": 1.569446978417406e-06, "loss": 0.8246, "step": 15940 }, { "epoch": 1.9, "learning_rate": 1.569148658861559e-06, "loss": 0.5405, "step": 15941 }, { "epoch": 1.9, "learning_rate": 1.5688503546927627e-06, "loss": 0.317, "step": 15942 }, { "epoch": 1.9, "learning_rate": 1.5685520659159492e-06, "loss": 0.5842, "step": 15943 }, { "epoch": 1.9, "learning_rate": 1.568253792536048e-06, "loss": 0.5431, "step": 15944 }, { "epoch": 1.9, "learning_rate": 1.5679555345579892e-06, "loss": 0.6867, "step": 15945 }, { "epoch": 1.9, "learning_rate": 1.5676572919867038e-06, "loss": 0.8739, "step": 15946 }, { "epoch": 1.9, "learning_rate": 1.567359064827122e-06, "loss": 0.8608, "step": 15947 }, { "epoch": 1.9, "learning_rate": 1.567060853084173e-06, "loss": 0.3461, "step": 15948 }, { "epoch": 1.9, "learning_rate": 1.5667626567627858e-06, "loss": 0.428, "step": 15949 }, { "epoch": 1.9, "learning_rate": 1.5664644758678907e-06, "loss": 0.5265, "step": 15950 }, { "epoch": 1.9, "learning_rate": 1.5661663104044166e-06, "loss": 0.6894, "step": 15951 }, { "epoch": 1.9, "learning_rate": 1.5658681603772907e-06, "loss": 0.7594, "step": 15952 }, { "epoch": 1.9, "learning_rate": 1.5655700257914427e-06, "loss": 0.8967, "step": 15953 }, { "epoch": 1.9, "learning_rate": 1.565271906651801e-06, "loss": 0.4344, "step": 15954 }, { "epoch": 1.9, "learning_rate": 1.5649738029632936e-06, "loss": 0.5067, "step": 15955 }, { "epoch": 1.9, "learning_rate": 1.5646757147308467e-06, "loss": 0.5094, "step": 15956 }, { "epoch": 1.9, "learning_rate": 1.5643776419593892e-06, "loss": 0.6017, "step": 15957 }, { "epoch": 1.9, "learning_rate": 1.564079584653848e-06, "loss": 0.7514, "step": 15958 }, { "epoch": 1.9, "learning_rate": 1.5637815428191494e-06, "loss": 0.8316, "step": 15959 }, { "epoch": 1.9, "learning_rate": 1.56348351646022e-06, "loss": 0.4746, "step": 15960 }, { "epoch": 1.9, "learning_rate": 1.5631855055819872e-06, "loss": 0.4721, "step": 15961 }, { "epoch": 1.9, "learning_rate": 1.5628875101893772e-06, "loss": 0.5626, "step": 15962 }, { "epoch": 1.9, "learning_rate": 1.5625895302873142e-06, "loss": 0.5748, "step": 15963 }, { "epoch": 1.9, "learning_rate": 1.562291565880725e-06, "loss": 0.7422, "step": 15964 }, { "epoch": 1.9, "learning_rate": 1.5619936169745353e-06, "loss": 0.78, "step": 15965 }, { "epoch": 1.9, "learning_rate": 1.5616956835736708e-06, "loss": 0.5333, "step": 15966 }, { "epoch": 1.9, "learning_rate": 1.5613977656830538e-06, "loss": 0.4096, "step": 15967 }, { "epoch": 1.9, "learning_rate": 1.561099863307611e-06, "loss": 0.5557, "step": 15968 }, { "epoch": 1.9, "learning_rate": 1.5608019764522657e-06, "loss": 0.5534, "step": 15969 }, { "epoch": 1.9, "learning_rate": 1.5605041051219438e-06, "loss": 0.6727, "step": 15970 }, { "epoch": 1.9, "learning_rate": 1.560206249321567e-06, "loss": 0.7592, "step": 15971 }, { "epoch": 1.9, "learning_rate": 1.55990840905606e-06, "loss": 0.7251, "step": 15972 }, { "epoch": 1.9, "learning_rate": 1.5596105843303466e-06, "loss": 0.4913, "step": 15973 }, { "epoch": 1.9, "learning_rate": 1.5593127751493481e-06, "loss": 0.4958, "step": 15974 }, { "epoch": 1.9, "learning_rate": 1.559014981517988e-06, "loss": 0.6106, "step": 15975 }, { "epoch": 1.9, "learning_rate": 1.55871720344119e-06, "loss": 0.7888, "step": 15976 }, { "epoch": 1.9, "learning_rate": 1.5584194409238762e-06, "loss": 0.9568, "step": 15977 }, { "epoch": 1.9, "learning_rate": 1.5581216939709668e-06, "loss": 0.9233, "step": 15978 }, { "epoch": 1.9, "learning_rate": 1.5578239625873854e-06, "loss": 0.5213, "step": 15979 }, { "epoch": 1.9, "learning_rate": 1.5575262467780525e-06, "loss": 0.4296, "step": 15980 }, { "epoch": 1.9, "learning_rate": 1.5572285465478909e-06, "loss": 0.6277, "step": 15981 }, { "epoch": 1.9, "learning_rate": 1.5569308619018191e-06, "loss": 0.5594, "step": 15982 }, { "epoch": 1.9, "learning_rate": 1.5566331928447603e-06, "loss": 0.7171, "step": 15983 }, { "epoch": 1.9, "learning_rate": 1.5563355393816331e-06, "loss": 0.8362, "step": 15984 }, { "epoch": 1.9, "learning_rate": 1.5560379015173603e-06, "loss": 0.5375, "step": 15985 }, { "epoch": 1.9, "learning_rate": 1.555740279256859e-06, "loss": 0.4383, "step": 15986 }, { "epoch": 1.9, "learning_rate": 1.5554426726050498e-06, "loss": 0.4601, "step": 15987 }, { "epoch": 1.9, "learning_rate": 1.5551450815668534e-06, "loss": 0.7012, "step": 15988 }, { "epoch": 1.9, "learning_rate": 1.5548475061471873e-06, "loss": 0.6995, "step": 15989 }, { "epoch": 1.9, "learning_rate": 1.5545499463509717e-06, "loss": 0.8828, "step": 15990 }, { "epoch": 1.9, "learning_rate": 1.554252402183124e-06, "loss": 0.6141, "step": 15991 }, { "epoch": 1.9, "learning_rate": 1.5539548736485648e-06, "loss": 0.4294, "step": 15992 }, { "epoch": 1.9, "learning_rate": 1.55365736075221e-06, "loss": 0.4925, "step": 15993 }, { "epoch": 1.9, "learning_rate": 1.5533598634989787e-06, "loss": 0.639, "step": 15994 }, { "epoch": 1.9, "learning_rate": 1.5530623818937877e-06, "loss": 0.7264, "step": 15995 }, { "epoch": 1.9, "learning_rate": 1.552764915941556e-06, "loss": 0.9089, "step": 15996 }, { "epoch": 1.91, "learning_rate": 1.5524674656471994e-06, "loss": 0.7708, "step": 15997 }, { "epoch": 1.91, "learning_rate": 1.5521700310156345e-06, "loss": 0.4204, "step": 15998 }, { "epoch": 1.91, "learning_rate": 1.5518726120517786e-06, "loss": 0.5417, "step": 15999 }, { "epoch": 1.91, "learning_rate": 1.5515752087605487e-06, "loss": 0.6487, "step": 16000 }, { "epoch": 1.91, "learning_rate": 1.5512778211468598e-06, "loss": 0.6391, "step": 16001 }, { "epoch": 1.91, "learning_rate": 1.5509804492156277e-06, "loss": 0.6759, "step": 16002 }, { "epoch": 1.91, "learning_rate": 1.5506830929717697e-06, "loss": 0.8192, "step": 16003 }, { "epoch": 1.91, "learning_rate": 1.5503857524201984e-06, "loss": 0.3288, "step": 16004 }, { "epoch": 1.91, "learning_rate": 1.550088427565831e-06, "loss": 0.5314, "step": 16005 }, { "epoch": 1.91, "learning_rate": 1.5497911184135811e-06, "loss": 0.5348, "step": 16006 }, { "epoch": 1.91, "learning_rate": 1.5494938249683645e-06, "loss": 0.5244, "step": 16007 }, { "epoch": 1.91, "learning_rate": 1.5491965472350944e-06, "loss": 0.7863, "step": 16008 }, { "epoch": 1.91, "learning_rate": 1.5488992852186846e-06, "loss": 1.0214, "step": 16009 }, { "epoch": 1.91, "learning_rate": 1.5486020389240495e-06, "loss": 0.4038, "step": 16010 }, { "epoch": 1.91, "learning_rate": 1.5483048083561036e-06, "loss": 0.4935, "step": 16011 }, { "epoch": 1.91, "learning_rate": 1.5480075935197584e-06, "loss": 0.6477, "step": 16012 }, { "epoch": 1.91, "learning_rate": 1.547710394419927e-06, "loss": 0.5901, "step": 16013 }, { "epoch": 1.91, "learning_rate": 1.5474132110615233e-06, "loss": 0.734, "step": 16014 }, { "epoch": 1.91, "learning_rate": 1.5471160434494597e-06, "loss": 0.7296, "step": 16015 }, { "epoch": 1.91, "learning_rate": 1.5468188915886468e-06, "loss": 0.5695, "step": 16016 }, { "epoch": 1.91, "learning_rate": 1.5465217554839976e-06, "loss": 0.4259, "step": 16017 }, { "epoch": 1.91, "learning_rate": 1.5462246351404247e-06, "loss": 0.5397, "step": 16018 }, { "epoch": 1.91, "learning_rate": 1.5459275305628386e-06, "loss": 0.6344, "step": 16019 }, { "epoch": 1.91, "learning_rate": 1.5456304417561493e-06, "loss": 0.72, "step": 16020 }, { "epoch": 1.91, "learning_rate": 1.5453333687252692e-06, "loss": 0.7117, "step": 16021 }, { "epoch": 1.91, "learning_rate": 1.5450363114751094e-06, "loss": 0.8298, "step": 16022 }, { "epoch": 1.91, "learning_rate": 1.5447392700105792e-06, "loss": 0.4861, "step": 16023 }, { "epoch": 1.91, "learning_rate": 1.5444422443365886e-06, "loss": 0.5778, "step": 16024 }, { "epoch": 1.91, "learning_rate": 1.5441452344580482e-06, "loss": 0.5875, "step": 16025 }, { "epoch": 1.91, "learning_rate": 1.5438482403798676e-06, "loss": 0.6284, "step": 16026 }, { "epoch": 1.91, "learning_rate": 1.5435512621069554e-06, "loss": 0.6958, "step": 16027 }, { "epoch": 1.91, "learning_rate": 1.5432542996442207e-06, "loss": 0.8634, "step": 16028 }, { "epoch": 1.91, "learning_rate": 1.5429573529965732e-06, "loss": 0.3695, "step": 16029 }, { "epoch": 1.91, "learning_rate": 1.5426604221689217e-06, "loss": 0.6398, "step": 16030 }, { "epoch": 1.91, "learning_rate": 1.5423635071661725e-06, "loss": 0.5997, "step": 16031 }, { "epoch": 1.91, "learning_rate": 1.5420666079932348e-06, "loss": 0.5771, "step": 16032 }, { "epoch": 1.91, "learning_rate": 1.5417697246550172e-06, "loss": 0.7905, "step": 16033 }, { "epoch": 1.91, "learning_rate": 1.541472857156427e-06, "loss": 0.9485, "step": 16034 }, { "epoch": 1.91, "learning_rate": 1.5411760055023694e-06, "loss": 0.4142, "step": 16035 }, { "epoch": 1.91, "learning_rate": 1.5408791696977538e-06, "loss": 0.5031, "step": 16036 }, { "epoch": 1.91, "learning_rate": 1.5405823497474864e-06, "loss": 0.4659, "step": 16037 }, { "epoch": 1.91, "learning_rate": 1.5402855456564724e-06, "loss": 0.6105, "step": 16038 }, { "epoch": 1.91, "learning_rate": 1.5399887574296186e-06, "loss": 0.7056, "step": 16039 }, { "epoch": 1.91, "learning_rate": 1.5396919850718317e-06, "loss": 0.735, "step": 16040 }, { "epoch": 1.91, "learning_rate": 1.5393952285880176e-06, "loss": 0.5033, "step": 16041 }, { "epoch": 1.91, "learning_rate": 1.5390984879830801e-06, "loss": 0.353, "step": 16042 }, { "epoch": 1.91, "learning_rate": 1.5388017632619254e-06, "loss": 0.4953, "step": 16043 }, { "epoch": 1.91, "learning_rate": 1.5385050544294577e-06, "loss": 0.6899, "step": 16044 }, { "epoch": 1.91, "learning_rate": 1.5382083614905832e-06, "loss": 0.6268, "step": 16045 }, { "epoch": 1.91, "learning_rate": 1.5379116844502043e-06, "loss": 0.8267, "step": 16046 }, { "epoch": 1.91, "learning_rate": 1.5376150233132265e-06, "loss": 0.7834, "step": 16047 }, { "epoch": 1.91, "learning_rate": 1.5373183780845524e-06, "loss": 0.3891, "step": 16048 }, { "epoch": 1.91, "learning_rate": 1.5370217487690875e-06, "loss": 0.4151, "step": 16049 }, { "epoch": 1.91, "learning_rate": 1.5367251353717327e-06, "loss": 0.5185, "step": 16050 }, { "epoch": 1.91, "learning_rate": 1.5364285378973925e-06, "loss": 0.6343, "step": 16051 }, { "epoch": 1.91, "learning_rate": 1.5361319563509703e-06, "loss": 0.7474, "step": 16052 }, { "epoch": 1.91, "learning_rate": 1.5358353907373668e-06, "loss": 0.8609, "step": 16053 }, { "epoch": 1.91, "learning_rate": 1.5355388410614854e-06, "loss": 0.3964, "step": 16054 }, { "epoch": 1.91, "learning_rate": 1.5352423073282278e-06, "loss": 0.4733, "step": 16055 }, { "epoch": 1.91, "learning_rate": 1.5349457895424963e-06, "loss": 0.5756, "step": 16056 }, { "epoch": 1.91, "learning_rate": 1.534649287709191e-06, "loss": 0.6894, "step": 16057 }, { "epoch": 1.91, "learning_rate": 1.534352801833215e-06, "loss": 0.8146, "step": 16058 }, { "epoch": 1.91, "learning_rate": 1.5340563319194673e-06, "loss": 0.857, "step": 16059 }, { "epoch": 1.91, "learning_rate": 1.5337598779728502e-06, "loss": 0.3763, "step": 16060 }, { "epoch": 1.91, "learning_rate": 1.533463439998263e-06, "loss": 0.4449, "step": 16061 }, { "epoch": 1.91, "learning_rate": 1.5331670180006065e-06, "loss": 0.5448, "step": 16062 }, { "epoch": 1.91, "learning_rate": 1.5328706119847798e-06, "loss": 0.5589, "step": 16063 }, { "epoch": 1.91, "learning_rate": 1.532574221955684e-06, "loss": 0.7275, "step": 16064 }, { "epoch": 1.91, "learning_rate": 1.5322778479182165e-06, "loss": 0.7623, "step": 16065 }, { "epoch": 1.91, "learning_rate": 1.5319814898772775e-06, "loss": 0.6913, "step": 16066 }, { "epoch": 1.91, "learning_rate": 1.5316851478377665e-06, "loss": 0.4629, "step": 16067 }, { "epoch": 1.91, "learning_rate": 1.5313888218045804e-06, "loss": 0.5479, "step": 16068 }, { "epoch": 1.91, "learning_rate": 1.5310925117826187e-06, "loss": 0.5994, "step": 16069 }, { "epoch": 1.91, "learning_rate": 1.5307962177767785e-06, "loss": 0.6797, "step": 16070 }, { "epoch": 1.91, "learning_rate": 1.5304999397919592e-06, "loss": 0.773, "step": 16071 }, { "epoch": 1.91, "learning_rate": 1.530203677833057e-06, "loss": 0.8206, "step": 16072 }, { "epoch": 1.91, "learning_rate": 1.5299074319049684e-06, "loss": 0.5035, "step": 16073 }, { "epoch": 1.91, "learning_rate": 1.5296112020125917e-06, "loss": 0.4784, "step": 16074 }, { "epoch": 1.91, "learning_rate": 1.5293149881608243e-06, "loss": 0.588, "step": 16075 }, { "epoch": 1.91, "learning_rate": 1.5290187903545609e-06, "loss": 0.5904, "step": 16076 }, { "epoch": 1.91, "learning_rate": 1.5287226085986978e-06, "loss": 0.8516, "step": 16077 }, { "epoch": 1.91, "learning_rate": 1.5284264428981315e-06, "loss": 0.9614, "step": 16078 }, { "epoch": 1.91, "learning_rate": 1.528130293257759e-06, "loss": 0.3439, "step": 16079 }, { "epoch": 1.91, "learning_rate": 1.5278341596824737e-06, "loss": 0.6365, "step": 16080 }, { "epoch": 1.92, "learning_rate": 1.5275380421771704e-06, "loss": 0.727, "step": 16081 }, { "epoch": 1.92, "learning_rate": 1.527241940746746e-06, "loss": 0.6445, "step": 16082 }, { "epoch": 1.92, "learning_rate": 1.5269458553960937e-06, "loss": 0.7964, "step": 16083 }, { "epoch": 1.92, "learning_rate": 1.5266497861301071e-06, "loss": 0.9225, "step": 16084 }, { "epoch": 1.92, "learning_rate": 1.5263537329536818e-06, "loss": 0.399, "step": 16085 }, { "epoch": 1.92, "learning_rate": 1.5260576958717115e-06, "loss": 0.5847, "step": 16086 }, { "epoch": 1.92, "learning_rate": 1.5257616748890891e-06, "loss": 0.6035, "step": 16087 }, { "epoch": 1.92, "learning_rate": 1.5254656700107074e-06, "loss": 0.7352, "step": 16088 }, { "epoch": 1.92, "learning_rate": 1.5251696812414601e-06, "loss": 0.7488, "step": 16089 }, { "epoch": 1.92, "learning_rate": 1.52487370858624e-06, "loss": 0.9062, "step": 16090 }, { "epoch": 1.92, "learning_rate": 1.5245777520499393e-06, "loss": 0.5293, "step": 16091 }, { "epoch": 1.92, "learning_rate": 1.5242818116374496e-06, "loss": 0.4559, "step": 16092 }, { "epoch": 1.92, "learning_rate": 1.5239858873536639e-06, "loss": 0.5639, "step": 16093 }, { "epoch": 1.92, "learning_rate": 1.5236899792034743e-06, "loss": 0.5821, "step": 16094 }, { "epoch": 1.92, "learning_rate": 1.5233940871917697e-06, "loss": 0.6373, "step": 16095 }, { "epoch": 1.92, "learning_rate": 1.5230982113234432e-06, "loss": 0.7855, "step": 16096 }, { "epoch": 1.92, "learning_rate": 1.5228023516033854e-06, "loss": 0.7609, "step": 16097 }, { "epoch": 1.92, "learning_rate": 1.5225065080364876e-06, "loss": 0.4345, "step": 16098 }, { "epoch": 1.92, "learning_rate": 1.5222106806276379e-06, "loss": 0.5648, "step": 16099 }, { "epoch": 1.92, "learning_rate": 1.5219148693817281e-06, "loss": 0.6141, "step": 16100 }, { "epoch": 1.92, "learning_rate": 1.5216190743036485e-06, "loss": 0.5272, "step": 16101 }, { "epoch": 1.92, "learning_rate": 1.5213232953982864e-06, "loss": 0.7508, "step": 16102 }, { "epoch": 1.92, "learning_rate": 1.5210275326705325e-06, "loss": 0.8902, "step": 16103 }, { "epoch": 1.92, "learning_rate": 1.5207317861252762e-06, "loss": 0.3615, "step": 16104 }, { "epoch": 1.92, "learning_rate": 1.5204360557674058e-06, "loss": 0.4722, "step": 16105 }, { "epoch": 1.92, "learning_rate": 1.520140341601809e-06, "loss": 0.6674, "step": 16106 }, { "epoch": 1.92, "learning_rate": 1.519844643633374e-06, "loss": 0.6889, "step": 16107 }, { "epoch": 1.92, "learning_rate": 1.5195489618669906e-06, "loss": 0.7683, "step": 16108 }, { "epoch": 1.92, "learning_rate": 1.5192532963075452e-06, "loss": 0.8587, "step": 16109 }, { "epoch": 1.92, "learning_rate": 1.5189576469599237e-06, "loss": 0.4178, "step": 16110 }, { "epoch": 1.92, "learning_rate": 1.5186620138290158e-06, "loss": 0.4718, "step": 16111 }, { "epoch": 1.92, "learning_rate": 1.5183663969197063e-06, "loss": 0.6403, "step": 16112 }, { "epoch": 1.92, "learning_rate": 1.5180707962368834e-06, "loss": 0.6249, "step": 16113 }, { "epoch": 1.92, "learning_rate": 1.5177752117854321e-06, "loss": 0.5959, "step": 16114 }, { "epoch": 1.92, "learning_rate": 1.5174796435702393e-06, "loss": 0.8845, "step": 16115 }, { "epoch": 1.92, "learning_rate": 1.517184091596191e-06, "loss": 0.6171, "step": 16116 }, { "epoch": 1.92, "learning_rate": 1.516888555868171e-06, "loss": 0.4073, "step": 16117 }, { "epoch": 1.92, "learning_rate": 1.5165930363910657e-06, "loss": 0.5026, "step": 16118 }, { "epoch": 1.92, "learning_rate": 1.5162975331697605e-06, "loss": 0.6417, "step": 16119 }, { "epoch": 1.92, "learning_rate": 1.5160020462091402e-06, "loss": 0.668, "step": 16120 }, { "epoch": 1.92, "learning_rate": 1.5157065755140876e-06, "loss": 0.741, "step": 16121 }, { "epoch": 1.92, "learning_rate": 1.5154111210894883e-06, "loss": 0.6652, "step": 16122 }, { "epoch": 1.92, "learning_rate": 1.5151156829402252e-06, "loss": 0.4445, "step": 16123 }, { "epoch": 1.92, "learning_rate": 1.5148202610711838e-06, "loss": 0.6776, "step": 16124 }, { "epoch": 1.92, "learning_rate": 1.514524855487245e-06, "loss": 0.4627, "step": 16125 }, { "epoch": 1.92, "learning_rate": 1.5142294661932933e-06, "loss": 0.6125, "step": 16126 }, { "epoch": 1.92, "learning_rate": 1.5139340931942108e-06, "loss": 0.7993, "step": 16127 }, { "epoch": 1.92, "learning_rate": 1.5136387364948819e-06, "loss": 0.8799, "step": 16128 }, { "epoch": 1.92, "learning_rate": 1.5133433961001867e-06, "loss": 0.5666, "step": 16129 }, { "epoch": 1.92, "learning_rate": 1.5130480720150076e-06, "loss": 0.5107, "step": 16130 }, { "epoch": 1.92, "learning_rate": 1.5127527642442275e-06, "loss": 0.5107, "step": 16131 }, { "epoch": 1.92, "learning_rate": 1.512457472792726e-06, "loss": 0.6688, "step": 16132 }, { "epoch": 1.92, "learning_rate": 1.5121621976653858e-06, "loss": 0.7162, "step": 16133 }, { "epoch": 1.92, "learning_rate": 1.511866938867087e-06, "loss": 0.8488, "step": 16134 }, { "epoch": 1.92, "learning_rate": 1.5115716964027115e-06, "loss": 0.3972, "step": 16135 }, { "epoch": 1.92, "learning_rate": 1.511276470277138e-06, "loss": 0.4479, "step": 16136 }, { "epoch": 1.92, "learning_rate": 1.510981260495248e-06, "loss": 0.5497, "step": 16137 }, { "epoch": 1.92, "learning_rate": 1.5106860670619198e-06, "loss": 0.673, "step": 16138 }, { "epoch": 1.92, "learning_rate": 1.5103908899820352e-06, "loss": 0.757, "step": 16139 }, { "epoch": 1.92, "learning_rate": 1.5100957292604715e-06, "loss": 0.837, "step": 16140 }, { "epoch": 1.92, "learning_rate": 1.509800584902108e-06, "loss": 0.5468, "step": 16141 }, { "epoch": 1.92, "learning_rate": 1.509505456911824e-06, "loss": 0.4289, "step": 16142 }, { "epoch": 1.92, "learning_rate": 1.509210345294499e-06, "loss": 0.5573, "step": 16143 }, { "epoch": 1.92, "learning_rate": 1.5089152500550096e-06, "loss": 0.7502, "step": 16144 }, { "epoch": 1.92, "learning_rate": 1.5086201711982337e-06, "loss": 0.7051, "step": 16145 }, { "epoch": 1.92, "learning_rate": 1.5083251087290506e-06, "loss": 0.8511, "step": 16146 }, { "epoch": 1.92, "learning_rate": 1.5080300626523359e-06, "loss": 0.8835, "step": 16147 }, { "epoch": 1.92, "learning_rate": 1.5077350329729675e-06, "loss": 0.4135, "step": 16148 }, { "epoch": 1.92, "learning_rate": 1.5074400196958223e-06, "loss": 0.5618, "step": 16149 }, { "epoch": 1.92, "learning_rate": 1.5071450228257773e-06, "loss": 0.547, "step": 16150 }, { "epoch": 1.92, "learning_rate": 1.5068500423677082e-06, "loss": 0.7728, "step": 16151 }, { "epoch": 1.92, "learning_rate": 1.506555078326491e-06, "loss": 0.7767, "step": 16152 }, { "epoch": 1.92, "learning_rate": 1.5062601307070013e-06, "loss": 0.7248, "step": 16153 }, { "epoch": 1.92, "learning_rate": 1.505965199514116e-06, "loss": 0.4434, "step": 16154 }, { "epoch": 1.92, "learning_rate": 1.5056702847527089e-06, "loss": 0.4918, "step": 16155 }, { "epoch": 1.92, "learning_rate": 1.505375386427655e-06, "loss": 0.6744, "step": 16156 }, { "epoch": 1.92, "learning_rate": 1.5050805045438299e-06, "loss": 0.5405, "step": 16157 }, { "epoch": 1.92, "learning_rate": 1.5047856391061083e-06, "loss": 0.7509, "step": 16158 }, { "epoch": 1.92, "learning_rate": 1.5044907901193622e-06, "loss": 0.7677, "step": 16159 }, { "epoch": 1.92, "learning_rate": 1.504195957588467e-06, "loss": 0.3763, "step": 16160 }, { "epoch": 1.92, "learning_rate": 1.5039011415182968e-06, "loss": 0.5463, "step": 16161 }, { "epoch": 1.92, "learning_rate": 1.5036063419137248e-06, "loss": 0.5372, "step": 16162 }, { "epoch": 1.92, "learning_rate": 1.5033115587796223e-06, "loss": 0.5075, "step": 16163 }, { "epoch": 1.92, "learning_rate": 1.5030167921208635e-06, "loss": 0.6505, "step": 16164 }, { "epoch": 1.93, "learning_rate": 1.5027220419423219e-06, "loss": 0.7729, "step": 16165 }, { "epoch": 1.93, "learning_rate": 1.502427308248868e-06, "loss": 0.561, "step": 16166 }, { "epoch": 1.93, "learning_rate": 1.5021325910453738e-06, "loss": 0.5281, "step": 16167 }, { "epoch": 1.93, "learning_rate": 1.5018378903367125e-06, "loss": 0.6355, "step": 16168 }, { "epoch": 1.93, "learning_rate": 1.5015432061277546e-06, "loss": 0.6113, "step": 16169 }, { "epoch": 1.93, "learning_rate": 1.5012485384233708e-06, "loss": 0.724, "step": 16170 }, { "epoch": 1.93, "learning_rate": 1.500953887228432e-06, "loss": 0.6816, "step": 16171 }, { "epoch": 1.93, "learning_rate": 1.5006592525478104e-06, "loss": 0.9319, "step": 16172 }, { "epoch": 1.93, "learning_rate": 1.500364634386375e-06, "loss": 0.4549, "step": 16173 }, { "epoch": 1.93, "learning_rate": 1.5000700327489954e-06, "loss": 0.576, "step": 16174 }, { "epoch": 1.93, "learning_rate": 1.4997754476405418e-06, "loss": 0.6559, "step": 16175 }, { "epoch": 1.93, "learning_rate": 1.499480879065885e-06, "loss": 0.6132, "step": 16176 }, { "epoch": 1.93, "learning_rate": 1.4991863270298935e-06, "loss": 0.7862, "step": 16177 }, { "epoch": 1.93, "learning_rate": 1.4988917915374345e-06, "loss": 0.7714, "step": 16178 }, { "epoch": 1.93, "learning_rate": 1.4985972725933796e-06, "loss": 0.4429, "step": 16179 }, { "epoch": 1.93, "learning_rate": 1.4983027702025955e-06, "loss": 0.452, "step": 16180 }, { "epoch": 1.93, "learning_rate": 1.49800828436995e-06, "loss": 0.4737, "step": 16181 }, { "epoch": 1.93, "learning_rate": 1.4977138151003117e-06, "loss": 0.5593, "step": 16182 }, { "epoch": 1.93, "learning_rate": 1.4974193623985483e-06, "loss": 0.8138, "step": 16183 }, { "epoch": 1.93, "learning_rate": 1.4971249262695278e-06, "loss": 0.8128, "step": 16184 }, { "epoch": 1.93, "learning_rate": 1.4968305067181155e-06, "loss": 0.5579, "step": 16185 }, { "epoch": 1.93, "learning_rate": 1.4965361037491794e-06, "loss": 0.5842, "step": 16186 }, { "epoch": 1.93, "learning_rate": 1.496241717367585e-06, "loss": 0.5703, "step": 16187 }, { "epoch": 1.93, "learning_rate": 1.4959473475782004e-06, "loss": 0.5691, "step": 16188 }, { "epoch": 1.93, "learning_rate": 1.4956529943858895e-06, "loss": 0.762, "step": 16189 }, { "epoch": 1.93, "learning_rate": 1.4953586577955193e-06, "loss": 0.7648, "step": 16190 }, { "epoch": 1.93, "learning_rate": 1.4950643378119546e-06, "loss": 0.6008, "step": 16191 }, { "epoch": 1.93, "learning_rate": 1.4947700344400611e-06, "loss": 0.4487, "step": 16192 }, { "epoch": 1.93, "learning_rate": 1.4944757476847027e-06, "loss": 0.557, "step": 16193 }, { "epoch": 1.93, "learning_rate": 1.494181477550745e-06, "loss": 0.5305, "step": 16194 }, { "epoch": 1.93, "learning_rate": 1.4938872240430522e-06, "loss": 0.7607, "step": 16195 }, { "epoch": 1.93, "learning_rate": 1.4935929871664873e-06, "loss": 0.7238, "step": 16196 }, { "epoch": 1.93, "learning_rate": 1.4932987669259147e-06, "loss": 0.8236, "step": 16197 }, { "epoch": 1.93, "learning_rate": 1.493004563326198e-06, "loss": 0.4244, "step": 16198 }, { "epoch": 1.93, "learning_rate": 1.4927103763722012e-06, "loss": 0.5164, "step": 16199 }, { "epoch": 1.93, "learning_rate": 1.4924162060687852e-06, "loss": 0.5592, "step": 16200 }, { "epoch": 1.93, "learning_rate": 1.4921220524208147e-06, "loss": 0.6024, "step": 16201 }, { "epoch": 1.93, "learning_rate": 1.4918279154331506e-06, "loss": 0.7133, "step": 16202 }, { "epoch": 1.93, "learning_rate": 1.4915337951106567e-06, "loss": 0.8072, "step": 16203 }, { "epoch": 1.93, "learning_rate": 1.4912396914581928e-06, "loss": 0.4402, "step": 16204 }, { "epoch": 1.93, "learning_rate": 1.490945604480622e-06, "loss": 0.5187, "step": 16205 }, { "epoch": 1.93, "learning_rate": 1.4906515341828042e-06, "loss": 0.5839, "step": 16206 }, { "epoch": 1.93, "learning_rate": 1.4903574805696024e-06, "loss": 0.6455, "step": 16207 }, { "epoch": 1.93, "learning_rate": 1.4900634436458757e-06, "loss": 0.7991, "step": 16208 }, { "epoch": 1.93, "learning_rate": 1.4897694234164845e-06, "loss": 0.9688, "step": 16209 }, { "epoch": 1.93, "learning_rate": 1.4894754198862904e-06, "loss": 0.3973, "step": 16210 }, { "epoch": 1.93, "learning_rate": 1.4891814330601518e-06, "loss": 0.4012, "step": 16211 }, { "epoch": 1.93, "learning_rate": 1.488887462942929e-06, "loss": 0.5021, "step": 16212 }, { "epoch": 1.93, "learning_rate": 1.488593509539481e-06, "loss": 0.6128, "step": 16213 }, { "epoch": 1.93, "learning_rate": 1.4882995728546678e-06, "loss": 0.6386, "step": 16214 }, { "epoch": 1.93, "learning_rate": 1.4880056528933474e-06, "loss": 0.8543, "step": 16215 }, { "epoch": 1.93, "learning_rate": 1.487711749660378e-06, "loss": 0.6668, "step": 16216 }, { "epoch": 1.93, "learning_rate": 1.4874178631606182e-06, "loss": 0.5445, "step": 16217 }, { "epoch": 1.93, "learning_rate": 1.4871239933989271e-06, "loss": 0.5175, "step": 16218 }, { "epoch": 1.93, "learning_rate": 1.4868301403801608e-06, "loss": 0.5772, "step": 16219 }, { "epoch": 1.93, "learning_rate": 1.4865363041091769e-06, "loss": 0.691, "step": 16220 }, { "epoch": 1.93, "learning_rate": 1.486242484590833e-06, "loss": 0.7508, "step": 16221 }, { "epoch": 1.93, "learning_rate": 1.4859486818299867e-06, "loss": 0.779, "step": 16222 }, { "epoch": 1.93, "learning_rate": 1.485654895831493e-06, "loss": 0.4804, "step": 16223 }, { "epoch": 1.93, "learning_rate": 1.485361126600209e-06, "loss": 0.6515, "step": 16224 }, { "epoch": 1.93, "learning_rate": 1.4850673741409916e-06, "loss": 0.4382, "step": 16225 }, { "epoch": 1.93, "learning_rate": 1.4847736384586953e-06, "loss": 0.7105, "step": 16226 }, { "epoch": 1.93, "learning_rate": 1.4844799195581748e-06, "loss": 0.7562, "step": 16227 }, { "epoch": 1.93, "learning_rate": 1.4841862174442872e-06, "loss": 0.9752, "step": 16228 }, { "epoch": 1.93, "learning_rate": 1.4838925321218873e-06, "loss": 0.3867, "step": 16229 }, { "epoch": 1.93, "learning_rate": 1.4835988635958287e-06, "loss": 0.5149, "step": 16230 }, { "epoch": 1.93, "learning_rate": 1.4833052118709657e-06, "loss": 0.5435, "step": 16231 }, { "epoch": 1.93, "learning_rate": 1.4830115769521524e-06, "loss": 0.5673, "step": 16232 }, { "epoch": 1.93, "learning_rate": 1.4827179588442447e-06, "loss": 0.6338, "step": 16233 }, { "epoch": 1.93, "learning_rate": 1.4824243575520937e-06, "loss": 0.8291, "step": 16234 }, { "epoch": 1.93, "learning_rate": 1.482130773080553e-06, "loss": 0.4713, "step": 16235 }, { "epoch": 1.93, "learning_rate": 1.4818372054344766e-06, "loss": 0.5475, "step": 16236 }, { "epoch": 1.93, "learning_rate": 1.4815436546187168e-06, "loss": 0.5426, "step": 16237 }, { "epoch": 1.93, "learning_rate": 1.4812501206381252e-06, "loss": 0.6128, "step": 16238 }, { "epoch": 1.93, "learning_rate": 1.4809566034975545e-06, "loss": 0.8914, "step": 16239 }, { "epoch": 1.93, "learning_rate": 1.480663103201857e-06, "loss": 0.9849, "step": 16240 }, { "epoch": 1.93, "learning_rate": 1.4803696197558849e-06, "loss": 0.4695, "step": 16241 }, { "epoch": 1.93, "learning_rate": 1.4800761531644868e-06, "loss": 0.5293, "step": 16242 }, { "epoch": 1.93, "learning_rate": 1.4797827034325162e-06, "loss": 0.5864, "step": 16243 }, { "epoch": 1.93, "learning_rate": 1.479489270564824e-06, "loss": 0.6159, "step": 16244 }, { "epoch": 1.93, "learning_rate": 1.4791958545662588e-06, "loss": 0.7097, "step": 16245 }, { "epoch": 1.93, "learning_rate": 1.4789024554416714e-06, "loss": 0.7867, "step": 16246 }, { "epoch": 1.93, "learning_rate": 1.4786090731959127e-06, "loss": 0.7202, "step": 16247 }, { "epoch": 1.93, "learning_rate": 1.4783157078338324e-06, "loss": 0.5046, "step": 16248 }, { "epoch": 1.94, "learning_rate": 1.4780223593602778e-06, "loss": 0.6058, "step": 16249 }, { "epoch": 1.94, "learning_rate": 1.4777290277800996e-06, "loss": 0.6179, "step": 16250 }, { "epoch": 1.94, "learning_rate": 1.4774357130981469e-06, "loss": 0.7205, "step": 16251 }, { "epoch": 1.94, "learning_rate": 1.4771424153192681e-06, "loss": 0.7562, "step": 16252 }, { "epoch": 1.94, "learning_rate": 1.47684913444831e-06, "loss": 0.8181, "step": 16253 }, { "epoch": 1.94, "learning_rate": 1.4765558704901216e-06, "loss": 0.3835, "step": 16254 }, { "epoch": 1.94, "learning_rate": 1.4762626234495509e-06, "loss": 0.4456, "step": 16255 }, { "epoch": 1.94, "learning_rate": 1.4759693933314454e-06, "loss": 0.5405, "step": 16256 }, { "epoch": 1.94, "learning_rate": 1.475676180140651e-06, "loss": 0.6456, "step": 16257 }, { "epoch": 1.94, "learning_rate": 1.4753829838820155e-06, "loss": 0.8663, "step": 16258 }, { "epoch": 1.94, "learning_rate": 1.475089804560386e-06, "loss": 0.8259, "step": 16259 }, { "epoch": 1.94, "learning_rate": 1.474796642180607e-06, "loss": 0.524, "step": 16260 }, { "epoch": 1.94, "learning_rate": 1.4745034967475252e-06, "loss": 0.5202, "step": 16261 }, { "epoch": 1.94, "learning_rate": 1.4742103682659875e-06, "loss": 0.6151, "step": 16262 }, { "epoch": 1.94, "learning_rate": 1.4739172567408388e-06, "loss": 0.5631, "step": 16263 }, { "epoch": 1.94, "learning_rate": 1.473624162176923e-06, "loss": 0.8881, "step": 16264 }, { "epoch": 1.94, "learning_rate": 1.4733310845790866e-06, "loss": 0.8935, "step": 16265 }, { "epoch": 1.94, "learning_rate": 1.4730380239521727e-06, "loss": 0.6004, "step": 16266 }, { "epoch": 1.94, "learning_rate": 1.4727449803010277e-06, "loss": 0.4962, "step": 16267 }, { "epoch": 1.94, "learning_rate": 1.4724519536304932e-06, "loss": 0.5113, "step": 16268 }, { "epoch": 1.94, "learning_rate": 1.4721589439454149e-06, "loss": 0.6638, "step": 16269 }, { "epoch": 1.94, "learning_rate": 1.4718659512506344e-06, "loss": 0.7645, "step": 16270 }, { "epoch": 1.94, "learning_rate": 1.4715729755509974e-06, "loss": 0.7467, "step": 16271 }, { "epoch": 1.94, "learning_rate": 1.471280016851345e-06, "loss": 0.8808, "step": 16272 }, { "epoch": 1.94, "learning_rate": 1.4709870751565198e-06, "loss": 0.4835, "step": 16273 }, { "epoch": 1.94, "learning_rate": 1.470694150471365e-06, "loss": 0.4882, "step": 16274 }, { "epoch": 1.94, "learning_rate": 1.4704012428007218e-06, "loss": 0.5333, "step": 16275 }, { "epoch": 1.94, "learning_rate": 1.470108352149433e-06, "loss": 0.6463, "step": 16276 }, { "epoch": 1.94, "learning_rate": 1.4698154785223389e-06, "loss": 0.7795, "step": 16277 }, { "epoch": 1.94, "learning_rate": 1.4695226219242822e-06, "loss": 0.93, "step": 16278 }, { "epoch": 1.94, "learning_rate": 1.469229782360102e-06, "loss": 0.391, "step": 16279 }, { "epoch": 1.94, "learning_rate": 1.4689369598346408e-06, "loss": 0.5148, "step": 16280 }, { "epoch": 1.94, "learning_rate": 1.4686441543527374e-06, "loss": 0.5233, "step": 16281 }, { "epoch": 1.94, "learning_rate": 1.468351365919234e-06, "loss": 0.7508, "step": 16282 }, { "epoch": 1.94, "learning_rate": 1.4680585945389687e-06, "loss": 0.8496, "step": 16283 }, { "epoch": 1.94, "learning_rate": 1.4677658402167804e-06, "loss": 0.7595, "step": 16284 }, { "epoch": 1.94, "learning_rate": 1.4674731029575095e-06, "loss": 0.3767, "step": 16285 }, { "epoch": 1.94, "learning_rate": 1.467180382765996e-06, "loss": 0.4713, "step": 16286 }, { "epoch": 1.94, "learning_rate": 1.4668876796470766e-06, "loss": 0.521, "step": 16287 }, { "epoch": 1.94, "learning_rate": 1.4665949936055905e-06, "loss": 0.4879, "step": 16288 }, { "epoch": 1.94, "learning_rate": 1.4663023246463764e-06, "loss": 0.7025, "step": 16289 }, { "epoch": 1.94, "learning_rate": 1.4660096727742713e-06, "loss": 0.8516, "step": 16290 }, { "epoch": 1.94, "learning_rate": 1.4657170379941127e-06, "loss": 0.487, "step": 16291 }, { "epoch": 1.94, "learning_rate": 1.4654244203107381e-06, "loss": 0.485, "step": 16292 }, { "epoch": 1.94, "learning_rate": 1.465131819728986e-06, "loss": 0.6638, "step": 16293 }, { "epoch": 1.94, "learning_rate": 1.464839236253691e-06, "loss": 0.6009, "step": 16294 }, { "epoch": 1.94, "learning_rate": 1.4645466698896896e-06, "loss": 0.7763, "step": 16295 }, { "epoch": 1.94, "learning_rate": 1.4642541206418188e-06, "loss": 0.721, "step": 16296 }, { "epoch": 1.94, "learning_rate": 1.4639615885149152e-06, "loss": 0.7427, "step": 16297 }, { "epoch": 1.94, "learning_rate": 1.4636690735138132e-06, "loss": 0.4257, "step": 16298 }, { "epoch": 1.94, "learning_rate": 1.4633765756433477e-06, "loss": 0.554, "step": 16299 }, { "epoch": 1.94, "learning_rate": 1.4630840949083547e-06, "loss": 0.5566, "step": 16300 }, { "epoch": 1.94, "learning_rate": 1.4627916313136695e-06, "loss": 0.6887, "step": 16301 }, { "epoch": 1.94, "learning_rate": 1.4624991848641246e-06, "loss": 0.7703, "step": 16302 }, { "epoch": 1.94, "learning_rate": 1.462206755564555e-06, "loss": 0.6609, "step": 16303 }, { "epoch": 1.94, "learning_rate": 1.4619143434197952e-06, "loss": 0.4359, "step": 16304 }, { "epoch": 1.94, "learning_rate": 1.4616219484346794e-06, "loss": 0.6362, "step": 16305 }, { "epoch": 1.94, "learning_rate": 1.4613295706140385e-06, "loss": 0.5313, "step": 16306 }, { "epoch": 1.94, "learning_rate": 1.461037209962707e-06, "loss": 0.7103, "step": 16307 }, { "epoch": 1.94, "learning_rate": 1.460744866485519e-06, "loss": 0.8207, "step": 16308 }, { "epoch": 1.94, "learning_rate": 1.4604525401873042e-06, "loss": 0.8863, "step": 16309 }, { "epoch": 1.94, "learning_rate": 1.4601602310728964e-06, "loss": 0.439, "step": 16310 }, { "epoch": 1.94, "learning_rate": 1.4598679391471276e-06, "loss": 0.4975, "step": 16311 }, { "epoch": 1.94, "learning_rate": 1.459575664414829e-06, "loss": 0.5735, "step": 16312 }, { "epoch": 1.94, "learning_rate": 1.4592834068808316e-06, "loss": 0.6165, "step": 16313 }, { "epoch": 1.94, "learning_rate": 1.4589911665499673e-06, "loss": 0.7645, "step": 16314 }, { "epoch": 1.94, "learning_rate": 1.4586989434270657e-06, "loss": 0.7704, "step": 16315 }, { "epoch": 1.94, "learning_rate": 1.458406737516959e-06, "loss": 0.6085, "step": 16316 }, { "epoch": 1.94, "learning_rate": 1.4581145488244753e-06, "loss": 0.4133, "step": 16317 }, { "epoch": 1.94, "learning_rate": 1.4578223773544454e-06, "loss": 0.4582, "step": 16318 }, { "epoch": 1.94, "learning_rate": 1.4575302231116992e-06, "loss": 0.5063, "step": 16319 }, { "epoch": 1.94, "learning_rate": 1.4572380861010666e-06, "loss": 0.6463, "step": 16320 }, { "epoch": 1.94, "learning_rate": 1.4569459663273764e-06, "loss": 0.7606, "step": 16321 }, { "epoch": 1.94, "learning_rate": 1.4566538637954556e-06, "loss": 0.8725, "step": 16322 }, { "epoch": 1.94, "learning_rate": 1.4563617785101347e-06, "loss": 0.4689, "step": 16323 }, { "epoch": 1.94, "learning_rate": 1.4560697104762406e-06, "loss": 0.5185, "step": 16324 }, { "epoch": 1.94, "learning_rate": 1.4557776596986017e-06, "loss": 0.5053, "step": 16325 }, { "epoch": 1.94, "learning_rate": 1.4554856261820458e-06, "loss": 0.6522, "step": 16326 }, { "epoch": 1.94, "learning_rate": 1.4551936099314009e-06, "loss": 0.785, "step": 16327 }, { "epoch": 1.94, "learning_rate": 1.4549016109514934e-06, "loss": 0.9919, "step": 16328 }, { "epoch": 1.94, "learning_rate": 1.454609629247149e-06, "loss": 0.4135, "step": 16329 }, { "epoch": 1.94, "learning_rate": 1.454317664823195e-06, "loss": 0.5965, "step": 16330 }, { "epoch": 1.94, "learning_rate": 1.4540257176844591e-06, "loss": 0.5364, "step": 16331 }, { "epoch": 1.94, "learning_rate": 1.4537337878357645e-06, "loss": 0.5385, "step": 16332 }, { "epoch": 1.95, "learning_rate": 1.4534418752819382e-06, "loss": 0.7913, "step": 16333 }, { "epoch": 1.95, "learning_rate": 1.4531499800278065e-06, "loss": 0.856, "step": 16334 }, { "epoch": 1.95, "learning_rate": 1.4528581020781934e-06, "loss": 0.4972, "step": 16335 }, { "epoch": 1.95, "learning_rate": 1.4525662414379231e-06, "loss": 0.4148, "step": 16336 }, { "epoch": 1.95, "learning_rate": 1.4522743981118204e-06, "loss": 0.6779, "step": 16337 }, { "epoch": 1.95, "learning_rate": 1.451982572104711e-06, "loss": 0.7033, "step": 16338 }, { "epoch": 1.95, "learning_rate": 1.4516907634214167e-06, "loss": 0.6732, "step": 16339 }, { "epoch": 1.95, "learning_rate": 1.4513989720667617e-06, "loss": 0.8896, "step": 16340 }, { "epoch": 1.95, "learning_rate": 1.45110719804557e-06, "loss": 0.5296, "step": 16341 }, { "epoch": 1.95, "learning_rate": 1.4508154413626651e-06, "loss": 0.5112, "step": 16342 }, { "epoch": 1.95, "learning_rate": 1.4505237020228692e-06, "loss": 0.6073, "step": 16343 }, { "epoch": 1.95, "learning_rate": 1.4502319800310034e-06, "loss": 0.571, "step": 16344 }, { "epoch": 1.95, "learning_rate": 1.449940275391891e-06, "loss": 0.6567, "step": 16345 }, { "epoch": 1.95, "learning_rate": 1.449648588110355e-06, "loss": 0.7505, "step": 16346 }, { "epoch": 1.95, "learning_rate": 1.4493569181912148e-06, "loss": 0.7145, "step": 16347 }, { "epoch": 1.95, "learning_rate": 1.4490652656392934e-06, "loss": 0.4458, "step": 16348 }, { "epoch": 1.95, "learning_rate": 1.4487736304594119e-06, "loss": 0.4607, "step": 16349 }, { "epoch": 1.95, "learning_rate": 1.448482012656391e-06, "loss": 0.4582, "step": 16350 }, { "epoch": 1.95, "learning_rate": 1.448190412235049e-06, "loss": 0.6425, "step": 16351 }, { "epoch": 1.95, "learning_rate": 1.4478988292002081e-06, "loss": 0.786, "step": 16352 }, { "epoch": 1.95, "learning_rate": 1.4476072635566888e-06, "loss": 0.8546, "step": 16353 }, { "epoch": 1.95, "learning_rate": 1.4473157153093087e-06, "loss": 0.4488, "step": 16354 }, { "epoch": 1.95, "learning_rate": 1.447024184462888e-06, "loss": 0.4819, "step": 16355 }, { "epoch": 1.95, "learning_rate": 1.446732671022247e-06, "loss": 0.4908, "step": 16356 }, { "epoch": 1.95, "learning_rate": 1.4464411749922037e-06, "loss": 0.7304, "step": 16357 }, { "epoch": 1.95, "learning_rate": 1.4461496963775745e-06, "loss": 0.8278, "step": 16358 }, { "epoch": 1.95, "learning_rate": 1.4458582351831796e-06, "loss": 0.9036, "step": 16359 }, { "epoch": 1.95, "learning_rate": 1.4455667914138367e-06, "loss": 0.3741, "step": 16360 }, { "epoch": 1.95, "learning_rate": 1.445275365074364e-06, "loss": 0.4866, "step": 16361 }, { "epoch": 1.95, "learning_rate": 1.4449839561695772e-06, "loss": 0.5482, "step": 16362 }, { "epoch": 1.95, "learning_rate": 1.4446925647042948e-06, "loss": 0.5113, "step": 16363 }, { "epoch": 1.95, "learning_rate": 1.4444011906833321e-06, "loss": 0.6516, "step": 16364 }, { "epoch": 1.95, "learning_rate": 1.4441098341115072e-06, "loss": 0.8557, "step": 16365 }, { "epoch": 1.95, "learning_rate": 1.4438184949936342e-06, "loss": 0.5544, "step": 16366 }, { "epoch": 1.95, "learning_rate": 1.4435271733345302e-06, "loss": 0.4851, "step": 16367 }, { "epoch": 1.95, "learning_rate": 1.4432358691390119e-06, "loss": 0.5455, "step": 16368 }, { "epoch": 1.95, "learning_rate": 1.4429445824118921e-06, "loss": 0.6156, "step": 16369 }, { "epoch": 1.95, "learning_rate": 1.4426533131579872e-06, "loss": 0.725, "step": 16370 }, { "epoch": 1.95, "learning_rate": 1.4423620613821127e-06, "loss": 0.7366, "step": 16371 }, { "epoch": 1.95, "learning_rate": 1.4420708270890821e-06, "loss": 0.8535, "step": 16372 }, { "epoch": 1.95, "learning_rate": 1.4417796102837084e-06, "loss": 0.4034, "step": 16373 }, { "epoch": 1.95, "learning_rate": 1.4414884109708067e-06, "loss": 0.5464, "step": 16374 }, { "epoch": 1.95, "learning_rate": 1.44119722915519e-06, "loss": 0.6501, "step": 16375 }, { "epoch": 1.95, "learning_rate": 1.4409060648416734e-06, "loss": 0.6809, "step": 16376 }, { "epoch": 1.95, "learning_rate": 1.440614918035067e-06, "loss": 0.8101, "step": 16377 }, { "epoch": 1.95, "learning_rate": 1.440323788740186e-06, "loss": 0.9008, "step": 16378 }, { "epoch": 1.95, "learning_rate": 1.4400326769618408e-06, "loss": 0.3903, "step": 16379 }, { "epoch": 1.95, "learning_rate": 1.4397415827048456e-06, "loss": 0.5955, "step": 16380 }, { "epoch": 1.95, "learning_rate": 1.4394505059740096e-06, "loss": 0.5372, "step": 16381 }, { "epoch": 1.95, "learning_rate": 1.439159446774146e-06, "loss": 0.6186, "step": 16382 }, { "epoch": 1.95, "learning_rate": 1.4388684051100655e-06, "loss": 0.7393, "step": 16383 }, { "epoch": 1.95, "learning_rate": 1.4385773809865805e-06, "loss": 0.9191, "step": 16384 }, { "epoch": 1.95, "learning_rate": 1.4382863744085005e-06, "loss": 0.4441, "step": 16385 }, { "epoch": 1.95, "learning_rate": 1.4379953853806347e-06, "loss": 0.6124, "step": 16386 }, { "epoch": 1.95, "learning_rate": 1.4377044139077953e-06, "loss": 0.5255, "step": 16387 }, { "epoch": 1.95, "learning_rate": 1.4374134599947902e-06, "loss": 0.5453, "step": 16388 }, { "epoch": 1.95, "learning_rate": 1.4371225236464298e-06, "loss": 0.813, "step": 16389 }, { "epoch": 1.95, "learning_rate": 1.436831604867523e-06, "loss": 0.7794, "step": 16390 }, { "epoch": 1.95, "learning_rate": 1.43654070366288e-06, "loss": 0.5178, "step": 16391 }, { "epoch": 1.95, "learning_rate": 1.4362498200373084e-06, "loss": 0.4583, "step": 16392 }, { "epoch": 1.95, "learning_rate": 1.4359589539956158e-06, "loss": 0.5847, "step": 16393 }, { "epoch": 1.95, "learning_rate": 1.4356681055426109e-06, "loss": 0.5496, "step": 16394 }, { "epoch": 1.95, "learning_rate": 1.4353772746831024e-06, "loss": 0.6991, "step": 16395 }, { "epoch": 1.95, "learning_rate": 1.4350864614218956e-06, "loss": 0.775, "step": 16396 }, { "epoch": 1.95, "learning_rate": 1.4347956657637995e-06, "loss": 0.7918, "step": 16397 }, { "epoch": 1.95, "learning_rate": 1.4345048877136202e-06, "loss": 0.4427, "step": 16398 }, { "epoch": 1.95, "learning_rate": 1.4342141272761654e-06, "loss": 0.5263, "step": 16399 }, { "epoch": 1.95, "learning_rate": 1.4339233844562406e-06, "loss": 0.5999, "step": 16400 }, { "epoch": 1.95, "learning_rate": 1.4336326592586507e-06, "loss": 0.6592, "step": 16401 }, { "epoch": 1.95, "learning_rate": 1.4333419516882032e-06, "loss": 0.804, "step": 16402 }, { "epoch": 1.95, "learning_rate": 1.4330512617497023e-06, "loss": 0.8277, "step": 16403 }, { "epoch": 1.95, "learning_rate": 1.4327605894479529e-06, "loss": 0.5634, "step": 16404 }, { "epoch": 1.95, "learning_rate": 1.4324699347877608e-06, "loss": 0.5715, "step": 16405 }, { "epoch": 1.95, "learning_rate": 1.4321792977739312e-06, "loss": 0.5911, "step": 16406 }, { "epoch": 1.95, "learning_rate": 1.4318886784112675e-06, "loss": 0.6558, "step": 16407 }, { "epoch": 1.95, "learning_rate": 1.4315980767045728e-06, "loss": 0.738, "step": 16408 }, { "epoch": 1.95, "learning_rate": 1.4313074926586512e-06, "loss": 0.8228, "step": 16409 }, { "epoch": 1.95, "learning_rate": 1.4310169262783074e-06, "loss": 0.4967, "step": 16410 }, { "epoch": 1.95, "learning_rate": 1.4307263775683428e-06, "loss": 0.4486, "step": 16411 }, { "epoch": 1.95, "learning_rate": 1.4304358465335606e-06, "loss": 0.4515, "step": 16412 }, { "epoch": 1.95, "learning_rate": 1.4301453331787645e-06, "loss": 0.6436, "step": 16413 }, { "epoch": 1.95, "learning_rate": 1.4298548375087562e-06, "loss": 0.6462, "step": 16414 }, { "epoch": 1.95, "learning_rate": 1.429564359528336e-06, "loss": 0.6894, "step": 16415 }, { "epoch": 1.95, "learning_rate": 1.4292738992423066e-06, "loss": 0.6433, "step": 16416 }, { "epoch": 1.96, "learning_rate": 1.4289834566554709e-06, "loss": 0.4532, "step": 16417 }, { "epoch": 1.96, "learning_rate": 1.4286930317726272e-06, "loss": 0.5566, "step": 16418 }, { "epoch": 1.96, "learning_rate": 1.4284026245985778e-06, "loss": 0.6024, "step": 16419 }, { "epoch": 1.96, "learning_rate": 1.4281122351381237e-06, "loss": 0.6723, "step": 16420 }, { "epoch": 1.96, "learning_rate": 1.4278218633960645e-06, "loss": 0.8033, "step": 16421 }, { "epoch": 1.96, "learning_rate": 1.4275315093771987e-06, "loss": 0.9247, "step": 16422 }, { "epoch": 1.96, "learning_rate": 1.4272411730863273e-06, "loss": 0.5093, "step": 16423 }, { "epoch": 1.96, "learning_rate": 1.426950854528249e-06, "loss": 0.5119, "step": 16424 }, { "epoch": 1.96, "learning_rate": 1.4266605537077644e-06, "loss": 0.5977, "step": 16425 }, { "epoch": 1.96, "learning_rate": 1.4263702706296694e-06, "loss": 0.5901, "step": 16426 }, { "epoch": 1.96, "learning_rate": 1.4260800052987644e-06, "loss": 0.8436, "step": 16427 }, { "epoch": 1.96, "learning_rate": 1.425789757719848e-06, "loss": 0.7835, "step": 16428 }, { "epoch": 1.96, "learning_rate": 1.4254995278977169e-06, "loss": 0.4499, "step": 16429 }, { "epoch": 1.96, "learning_rate": 1.4252093158371677e-06, "loss": 0.4428, "step": 16430 }, { "epoch": 1.96, "learning_rate": 1.4249191215429986e-06, "loss": 0.5024, "step": 16431 }, { "epoch": 1.96, "learning_rate": 1.424628945020008e-06, "loss": 0.6559, "step": 16432 }, { "epoch": 1.96, "learning_rate": 1.42433878627299e-06, "loss": 0.8568, "step": 16433 }, { "epoch": 1.96, "learning_rate": 1.424048645306742e-06, "loss": 0.8588, "step": 16434 }, { "epoch": 1.96, "learning_rate": 1.4237585221260613e-06, "loss": 0.4967, "step": 16435 }, { "epoch": 1.96, "learning_rate": 1.4234684167357426e-06, "loss": 0.4305, "step": 16436 }, { "epoch": 1.96, "learning_rate": 1.4231783291405798e-06, "loss": 0.5578, "step": 16437 }, { "epoch": 1.96, "learning_rate": 1.4228882593453703e-06, "loss": 0.5726, "step": 16438 }, { "epoch": 1.96, "learning_rate": 1.4225982073549078e-06, "loss": 0.6465, "step": 16439 }, { "epoch": 1.96, "learning_rate": 1.4223081731739884e-06, "loss": 0.7949, "step": 16440 }, { "epoch": 1.96, "learning_rate": 1.4220181568074042e-06, "loss": 0.6904, "step": 16441 }, { "epoch": 1.96, "learning_rate": 1.4217281582599518e-06, "loss": 0.5536, "step": 16442 }, { "epoch": 1.96, "learning_rate": 1.4214381775364223e-06, "loss": 0.5546, "step": 16443 }, { "epoch": 1.96, "learning_rate": 1.4211482146416105e-06, "loss": 0.6644, "step": 16444 }, { "epoch": 1.96, "learning_rate": 1.4208582695803092e-06, "loss": 0.6259, "step": 16445 }, { "epoch": 1.96, "learning_rate": 1.4205683423573108e-06, "loss": 0.9168, "step": 16446 }, { "epoch": 1.96, "learning_rate": 1.4202784329774084e-06, "loss": 0.8723, "step": 16447 }, { "epoch": 1.96, "learning_rate": 1.4199885414453945e-06, "loss": 0.4854, "step": 16448 }, { "epoch": 1.96, "learning_rate": 1.4196986677660613e-06, "loss": 0.4966, "step": 16449 }, { "epoch": 1.96, "learning_rate": 1.4194088119441984e-06, "loss": 0.6048, "step": 16450 }, { "epoch": 1.96, "learning_rate": 1.4191189739846e-06, "loss": 0.6803, "step": 16451 }, { "epoch": 1.96, "learning_rate": 1.4188291538920543e-06, "loss": 0.7657, "step": 16452 }, { "epoch": 1.96, "learning_rate": 1.4185393516713536e-06, "loss": 0.9061, "step": 16453 }, { "epoch": 1.96, "learning_rate": 1.4182495673272878e-06, "loss": 0.5436, "step": 16454 }, { "epoch": 1.96, "learning_rate": 1.4179598008646486e-06, "loss": 0.5622, "step": 16455 }, { "epoch": 1.96, "learning_rate": 1.4176700522882237e-06, "loss": 0.5155, "step": 16456 }, { "epoch": 1.96, "learning_rate": 1.4173803216028048e-06, "loss": 0.7263, "step": 16457 }, { "epoch": 1.96, "learning_rate": 1.4170906088131787e-06, "loss": 0.7338, "step": 16458 }, { "epoch": 1.96, "learning_rate": 1.4168009139241368e-06, "loss": 0.8752, "step": 16459 }, { "epoch": 1.96, "learning_rate": 1.416511236940466e-06, "loss": 0.4593, "step": 16460 }, { "epoch": 1.96, "learning_rate": 1.4162215778669553e-06, "loss": 0.65, "step": 16461 }, { "epoch": 1.96, "learning_rate": 1.4159319367083924e-06, "loss": 0.5976, "step": 16462 }, { "epoch": 1.96, "learning_rate": 1.4156423134695668e-06, "loss": 0.5483, "step": 16463 }, { "epoch": 1.96, "learning_rate": 1.4153527081552647e-06, "loss": 0.8311, "step": 16464 }, { "epoch": 1.96, "learning_rate": 1.4150631207702726e-06, "loss": 0.7822, "step": 16465 }, { "epoch": 1.96, "learning_rate": 1.414773551319379e-06, "loss": 0.6227, "step": 16466 }, { "epoch": 1.96, "learning_rate": 1.4144839998073684e-06, "loss": 0.47, "step": 16467 }, { "epoch": 1.96, "learning_rate": 1.4141944662390289e-06, "loss": 0.51, "step": 16468 }, { "epoch": 1.96, "learning_rate": 1.4139049506191454e-06, "loss": 0.5746, "step": 16469 }, { "epoch": 1.96, "learning_rate": 1.4136154529525053e-06, "loss": 0.6655, "step": 16470 }, { "epoch": 1.96, "learning_rate": 1.4133259732438934e-06, "loss": 0.8475, "step": 16471 }, { "epoch": 1.96, "learning_rate": 1.413036511498093e-06, "loss": 0.6119, "step": 16472 }, { "epoch": 1.96, "learning_rate": 1.4127470677198901e-06, "loss": 0.5813, "step": 16473 }, { "epoch": 1.96, "learning_rate": 1.4124576419140707e-06, "loss": 0.5808, "step": 16474 }, { "epoch": 1.96, "learning_rate": 1.4121682340854165e-06, "loss": 0.5161, "step": 16475 }, { "epoch": 1.96, "learning_rate": 1.411878844238713e-06, "loss": 0.7007, "step": 16476 }, { "epoch": 1.96, "learning_rate": 1.4115894723787443e-06, "loss": 0.6636, "step": 16477 }, { "epoch": 1.96, "learning_rate": 1.4113001185102928e-06, "loss": 0.8113, "step": 16478 }, { "epoch": 1.96, "learning_rate": 1.411010782638141e-06, "loss": 0.4778, "step": 16479 }, { "epoch": 1.96, "learning_rate": 1.410721464767072e-06, "loss": 0.4876, "step": 16480 }, { "epoch": 1.96, "learning_rate": 1.41043216490187e-06, "loss": 0.4683, "step": 16481 }, { "epoch": 1.96, "learning_rate": 1.4101428830473143e-06, "loss": 0.6204, "step": 16482 }, { "epoch": 1.96, "learning_rate": 1.4098536192081885e-06, "loss": 0.7626, "step": 16483 }, { "epoch": 1.96, "learning_rate": 1.4095643733892735e-06, "loss": 0.8057, "step": 16484 }, { "epoch": 1.96, "learning_rate": 1.4092751455953523e-06, "loss": 0.4279, "step": 16485 }, { "epoch": 1.96, "learning_rate": 1.4089859358312041e-06, "loss": 0.5343, "step": 16486 }, { "epoch": 1.96, "learning_rate": 1.4086967441016092e-06, "loss": 0.5179, "step": 16487 }, { "epoch": 1.96, "learning_rate": 1.4084075704113487e-06, "loss": 0.6883, "step": 16488 }, { "epoch": 1.96, "learning_rate": 1.4081184147652036e-06, "loss": 0.7973, "step": 16489 }, { "epoch": 1.96, "learning_rate": 1.4078292771679517e-06, "loss": 0.8207, "step": 16490 }, { "epoch": 1.96, "learning_rate": 1.4075401576243733e-06, "loss": 0.5478, "step": 16491 }, { "epoch": 1.96, "learning_rate": 1.4072510561392488e-06, "loss": 0.4022, "step": 16492 }, { "epoch": 1.96, "learning_rate": 1.4069619727173562e-06, "loss": 0.6139, "step": 16493 }, { "epoch": 1.96, "learning_rate": 1.406672907363473e-06, "loss": 0.5408, "step": 16494 }, { "epoch": 1.96, "learning_rate": 1.406383860082378e-06, "loss": 0.6499, "step": 16495 }, { "epoch": 1.96, "learning_rate": 1.4060948308788512e-06, "loss": 0.9133, "step": 16496 }, { "epoch": 1.96, "learning_rate": 1.4058058197576671e-06, "loss": 0.8998, "step": 16497 }, { "epoch": 1.96, "learning_rate": 1.4055168267236048e-06, "loss": 0.3405, "step": 16498 }, { "epoch": 1.96, "learning_rate": 1.4052278517814421e-06, "loss": 0.4585, "step": 16499 }, { "epoch": 1.96, "learning_rate": 1.4049388949359548e-06, "loss": 0.5694, "step": 16500 }, { "epoch": 1.97, "learning_rate": 1.4046499561919186e-06, "loss": 0.6614, "step": 16501 }, { "epoch": 1.97, "learning_rate": 1.4043610355541104e-06, "loss": 0.8152, "step": 16502 }, { "epoch": 1.97, "learning_rate": 1.4040721330273063e-06, "loss": 0.9195, "step": 16503 }, { "epoch": 1.97, "learning_rate": 1.4037832486162828e-06, "loss": 0.4682, "step": 16504 }, { "epoch": 1.97, "learning_rate": 1.4034943823258133e-06, "loss": 0.5729, "step": 16505 }, { "epoch": 1.97, "learning_rate": 1.4032055341606742e-06, "loss": 0.574, "step": 16506 }, { "epoch": 1.97, "learning_rate": 1.402916704125639e-06, "loss": 0.5672, "step": 16507 }, { "epoch": 1.97, "learning_rate": 1.4026278922254839e-06, "loss": 0.6726, "step": 16508 }, { "epoch": 1.97, "learning_rate": 1.4023390984649807e-06, "loss": 0.7549, "step": 16509 }, { "epoch": 1.97, "learning_rate": 1.4020503228489041e-06, "loss": 0.6207, "step": 16510 }, { "epoch": 1.97, "learning_rate": 1.4017615653820289e-06, "loss": 0.5613, "step": 16511 }, { "epoch": 1.97, "learning_rate": 1.4014728260691263e-06, "loss": 0.5201, "step": 16512 }, { "epoch": 1.97, "learning_rate": 1.4011841049149698e-06, "loss": 0.6391, "step": 16513 }, { "epoch": 1.97, "learning_rate": 1.4008954019243338e-06, "loss": 0.7094, "step": 16514 }, { "epoch": 1.97, "learning_rate": 1.4006067171019882e-06, "loss": 0.7516, "step": 16515 }, { "epoch": 1.97, "learning_rate": 1.4003180504527052e-06, "loss": 0.4027, "step": 16516 }, { "epoch": 1.97, "learning_rate": 1.4000294019812574e-06, "loss": 0.41, "step": 16517 }, { "epoch": 1.97, "learning_rate": 1.3997407716924155e-06, "loss": 0.5817, "step": 16518 }, { "epoch": 1.97, "learning_rate": 1.3994521595909521e-06, "loss": 0.7594, "step": 16519 }, { "epoch": 1.97, "learning_rate": 1.399163565681636e-06, "loss": 0.8147, "step": 16520 }, { "epoch": 1.97, "learning_rate": 1.3988749899692393e-06, "loss": 0.8091, "step": 16521 }, { "epoch": 1.97, "learning_rate": 1.3985864324585302e-06, "loss": 0.7396, "step": 16522 }, { "epoch": 1.97, "learning_rate": 1.3982978931542812e-06, "loss": 0.392, "step": 16523 }, { "epoch": 1.97, "learning_rate": 1.3980093720612592e-06, "loss": 0.5814, "step": 16524 }, { "epoch": 1.97, "learning_rate": 1.3977208691842351e-06, "loss": 0.646, "step": 16525 }, { "epoch": 1.97, "learning_rate": 1.3974323845279771e-06, "loss": 0.6516, "step": 16526 }, { "epoch": 1.97, "learning_rate": 1.3971439180972557e-06, "loss": 0.7224, "step": 16527 }, { "epoch": 1.97, "learning_rate": 1.3968554698968375e-06, "loss": 0.7668, "step": 16528 }, { "epoch": 1.97, "learning_rate": 1.39656703993149e-06, "loss": 0.4495, "step": 16529 }, { "epoch": 1.97, "learning_rate": 1.3962786282059832e-06, "loss": 0.4756, "step": 16530 }, { "epoch": 1.97, "learning_rate": 1.395990234725082e-06, "loss": 0.5684, "step": 16531 }, { "epoch": 1.97, "learning_rate": 1.395701859493555e-06, "loss": 0.6285, "step": 16532 }, { "epoch": 1.97, "learning_rate": 1.3954135025161688e-06, "loss": 0.6235, "step": 16533 }, { "epoch": 1.97, "learning_rate": 1.3951251637976913e-06, "loss": 0.775, "step": 16534 }, { "epoch": 1.97, "learning_rate": 1.394836843342887e-06, "loss": 0.4358, "step": 16535 }, { "epoch": 1.97, "learning_rate": 1.3945485411565223e-06, "loss": 0.4766, "step": 16536 }, { "epoch": 1.97, "learning_rate": 1.3942602572433622e-06, "loss": 0.5024, "step": 16537 }, { "epoch": 1.97, "learning_rate": 1.3939719916081742e-06, "loss": 0.6126, "step": 16538 }, { "epoch": 1.97, "learning_rate": 1.3936837442557212e-06, "loss": 0.7953, "step": 16539 }, { "epoch": 1.97, "learning_rate": 1.3933955151907683e-06, "loss": 0.9109, "step": 16540 }, { "epoch": 1.97, "learning_rate": 1.3931073044180803e-06, "loss": 0.6368, "step": 16541 }, { "epoch": 1.97, "learning_rate": 1.3928191119424228e-06, "loss": 0.4259, "step": 16542 }, { "epoch": 1.97, "learning_rate": 1.392530937768558e-06, "loss": 0.5777, "step": 16543 }, { "epoch": 1.97, "learning_rate": 1.392242781901249e-06, "loss": 0.4756, "step": 16544 }, { "epoch": 1.97, "learning_rate": 1.3919546443452606e-06, "loss": 0.7061, "step": 16545 }, { "epoch": 1.97, "learning_rate": 1.3916665251053541e-06, "loss": 0.829, "step": 16546 }, { "epoch": 1.97, "learning_rate": 1.3913784241862926e-06, "loss": 0.788, "step": 16547 }, { "epoch": 1.97, "learning_rate": 1.391090341592839e-06, "loss": 0.4738, "step": 16548 }, { "epoch": 1.97, "learning_rate": 1.3908022773297558e-06, "loss": 0.5117, "step": 16549 }, { "epoch": 1.97, "learning_rate": 1.3905142314018044e-06, "loss": 0.6034, "step": 16550 }, { "epoch": 1.97, "learning_rate": 1.3902262038137449e-06, "loss": 0.5763, "step": 16551 }, { "epoch": 1.97, "learning_rate": 1.389938194570339e-06, "loss": 0.7317, "step": 16552 }, { "epoch": 1.97, "learning_rate": 1.389650203676349e-06, "loss": 0.9497, "step": 16553 }, { "epoch": 1.97, "learning_rate": 1.3893622311365335e-06, "loss": 0.4614, "step": 16554 }, { "epoch": 1.97, "learning_rate": 1.389074276955653e-06, "loss": 0.4692, "step": 16555 }, { "epoch": 1.97, "learning_rate": 1.3887863411384694e-06, "loss": 0.5025, "step": 16556 }, { "epoch": 1.97, "learning_rate": 1.3884984236897409e-06, "loss": 0.6105, "step": 16557 }, { "epoch": 1.97, "learning_rate": 1.3882105246142253e-06, "loss": 0.7917, "step": 16558 }, { "epoch": 1.97, "learning_rate": 1.3879226439166832e-06, "loss": 0.9042, "step": 16559 }, { "epoch": 1.97, "learning_rate": 1.3876347816018742e-06, "loss": 0.4349, "step": 16560 }, { "epoch": 1.97, "learning_rate": 1.3873469376745545e-06, "loss": 0.5049, "step": 16561 }, { "epoch": 1.97, "learning_rate": 1.3870591121394834e-06, "loss": 0.4293, "step": 16562 }, { "epoch": 1.97, "learning_rate": 1.3867713050014194e-06, "loss": 0.5403, "step": 16563 }, { "epoch": 1.97, "learning_rate": 1.386483516265119e-06, "loss": 0.6698, "step": 16564 }, { "epoch": 1.97, "learning_rate": 1.386195745935339e-06, "loss": 0.8974, "step": 16565 }, { "epoch": 1.97, "learning_rate": 1.3859079940168362e-06, "loss": 0.5092, "step": 16566 }, { "epoch": 1.97, "learning_rate": 1.385620260514368e-06, "loss": 0.549, "step": 16567 }, { "epoch": 1.97, "learning_rate": 1.3853325454326916e-06, "loss": 0.5177, "step": 16568 }, { "epoch": 1.97, "learning_rate": 1.3850448487765606e-06, "loss": 0.5812, "step": 16569 }, { "epoch": 1.97, "learning_rate": 1.3847571705507318e-06, "loss": 0.5458, "step": 16570 }, { "epoch": 1.97, "learning_rate": 1.3844695107599615e-06, "loss": 0.7727, "step": 16571 }, { "epoch": 1.97, "learning_rate": 1.384181869409004e-06, "loss": 0.8563, "step": 16572 }, { "epoch": 1.97, "learning_rate": 1.3838942465026125e-06, "loss": 0.4292, "step": 16573 }, { "epoch": 1.97, "learning_rate": 1.383606642045543e-06, "loss": 0.6269, "step": 16574 }, { "epoch": 1.97, "learning_rate": 1.3833190560425503e-06, "loss": 0.5919, "step": 16575 }, { "epoch": 1.97, "learning_rate": 1.3830314884983864e-06, "loss": 0.6692, "step": 16576 }, { "epoch": 1.97, "learning_rate": 1.3827439394178055e-06, "loss": 0.791, "step": 16577 }, { "epoch": 1.97, "learning_rate": 1.382456408805562e-06, "loss": 0.847, "step": 16578 }, { "epoch": 1.97, "learning_rate": 1.3821688966664082e-06, "loss": 0.4653, "step": 16579 }, { "epoch": 1.97, "learning_rate": 1.3818814030050947e-06, "loss": 0.4917, "step": 16580 }, { "epoch": 1.97, "learning_rate": 1.381593927826376e-06, "loss": 0.5541, "step": 16581 }, { "epoch": 1.97, "learning_rate": 1.3813064711350032e-06, "loss": 0.7205, "step": 16582 }, { "epoch": 1.97, "learning_rate": 1.3810190329357296e-06, "loss": 0.8018, "step": 16583 }, { "epoch": 1.97, "learning_rate": 1.3807316132333042e-06, "loss": 0.8858, "step": 16584 }, { "epoch": 1.98, "learning_rate": 1.3804442120324801e-06, "loss": 0.5223, "step": 16585 }, { "epoch": 1.98, "learning_rate": 1.3801568293380057e-06, "loss": 0.5892, "step": 16586 }, { "epoch": 1.98, "learning_rate": 1.3798694651546345e-06, "loss": 0.5753, "step": 16587 }, { "epoch": 1.98, "learning_rate": 1.3795821194871138e-06, "loss": 0.6404, "step": 16588 }, { "epoch": 1.98, "learning_rate": 1.3792947923401945e-06, "loss": 0.7317, "step": 16589 }, { "epoch": 1.98, "learning_rate": 1.3790074837186266e-06, "loss": 0.813, "step": 16590 }, { "epoch": 1.98, "learning_rate": 1.3787201936271598e-06, "loss": 0.7494, "step": 16591 }, { "epoch": 1.98, "learning_rate": 1.3784329220705422e-06, "loss": 0.4245, "step": 16592 }, { "epoch": 1.98, "learning_rate": 1.3781456690535218e-06, "loss": 0.5332, "step": 16593 }, { "epoch": 1.98, "learning_rate": 1.3778584345808488e-06, "loss": 0.6133, "step": 16594 }, { "epoch": 1.98, "learning_rate": 1.3775712186572687e-06, "loss": 0.7917, "step": 16595 }, { "epoch": 1.98, "learning_rate": 1.3772840212875305e-06, "loss": 0.778, "step": 16596 }, { "epoch": 1.98, "learning_rate": 1.3769968424763816e-06, "loss": 0.8527, "step": 16597 }, { "epoch": 1.98, "learning_rate": 1.3767096822285702e-06, "loss": 0.4088, "step": 16598 }, { "epoch": 1.98, "learning_rate": 1.3764225405488412e-06, "loss": 0.6591, "step": 16599 }, { "epoch": 1.98, "learning_rate": 1.3761354174419427e-06, "loss": 0.5797, "step": 16600 }, { "epoch": 1.98, "learning_rate": 1.375848312912619e-06, "loss": 0.7125, "step": 16601 }, { "epoch": 1.98, "learning_rate": 1.3755612269656182e-06, "loss": 0.7957, "step": 16602 }, { "epoch": 1.98, "learning_rate": 1.3752741596056834e-06, "loss": 0.7303, "step": 16603 }, { "epoch": 1.98, "learning_rate": 1.3749871108375612e-06, "loss": 0.4961, "step": 16604 }, { "epoch": 1.98, "learning_rate": 1.3747000806659966e-06, "loss": 0.4816, "step": 16605 }, { "epoch": 1.98, "learning_rate": 1.3744130690957353e-06, "loss": 0.4787, "step": 16606 }, { "epoch": 1.98, "learning_rate": 1.3741260761315201e-06, "loss": 0.6152, "step": 16607 }, { "epoch": 1.98, "learning_rate": 1.3738391017780941e-06, "loss": 0.7639, "step": 16608 }, { "epoch": 1.98, "learning_rate": 1.3735521460402035e-06, "loss": 0.7527, "step": 16609 }, { "epoch": 1.98, "learning_rate": 1.3732652089225896e-06, "loss": 0.4296, "step": 16610 }, { "epoch": 1.98, "learning_rate": 1.372978290429996e-06, "loss": 0.5507, "step": 16611 }, { "epoch": 1.98, "learning_rate": 1.372691390567166e-06, "loss": 0.4861, "step": 16612 }, { "epoch": 1.98, "learning_rate": 1.372404509338843e-06, "loss": 0.6195, "step": 16613 }, { "epoch": 1.98, "learning_rate": 1.372117646749768e-06, "loss": 0.7801, "step": 16614 }, { "epoch": 1.98, "learning_rate": 1.3718308028046818e-06, "loss": 0.7639, "step": 16615 }, { "epoch": 1.98, "learning_rate": 1.3715439775083272e-06, "loss": 0.5746, "step": 16616 }, { "epoch": 1.98, "learning_rate": 1.3712571708654466e-06, "loss": 0.4454, "step": 16617 }, { "epoch": 1.98, "learning_rate": 1.370970382880778e-06, "loss": 0.6848, "step": 16618 }, { "epoch": 1.98, "learning_rate": 1.3706836135590642e-06, "loss": 0.5909, "step": 16619 }, { "epoch": 1.98, "learning_rate": 1.370396862905046e-06, "loss": 0.781, "step": 16620 }, { "epoch": 1.98, "learning_rate": 1.3701101309234623e-06, "loss": 0.7854, "step": 16621 }, { "epoch": 1.98, "learning_rate": 1.3698234176190523e-06, "loss": 0.7841, "step": 16622 }, { "epoch": 1.98, "learning_rate": 1.3695367229965556e-06, "loss": 0.471, "step": 16623 }, { "epoch": 1.98, "learning_rate": 1.369250047060713e-06, "loss": 0.4913, "step": 16624 }, { "epoch": 1.98, "learning_rate": 1.3689633898162607e-06, "loss": 0.4709, "step": 16625 }, { "epoch": 1.98, "learning_rate": 1.368676751267939e-06, "loss": 0.6921, "step": 16626 }, { "epoch": 1.98, "learning_rate": 1.3683901314204851e-06, "loss": 0.767, "step": 16627 }, { "epoch": 1.98, "learning_rate": 1.3681035302786382e-06, "loss": 0.8491, "step": 16628 }, { "epoch": 1.98, "learning_rate": 1.3678169478471348e-06, "loss": 0.4405, "step": 16629 }, { "epoch": 1.98, "learning_rate": 1.3675303841307117e-06, "loss": 0.476, "step": 16630 }, { "epoch": 1.98, "learning_rate": 1.3672438391341062e-06, "loss": 0.5253, "step": 16631 }, { "epoch": 1.98, "learning_rate": 1.366957312862056e-06, "loss": 0.6945, "step": 16632 }, { "epoch": 1.98, "learning_rate": 1.3666708053192957e-06, "loss": 0.6834, "step": 16633 }, { "epoch": 1.98, "learning_rate": 1.3663843165105616e-06, "loss": 0.8161, "step": 16634 }, { "epoch": 1.98, "learning_rate": 1.3660978464405916e-06, "loss": 0.4588, "step": 16635 }, { "epoch": 1.98, "learning_rate": 1.365811395114119e-06, "loss": 0.398, "step": 16636 }, { "epoch": 1.98, "learning_rate": 1.365524962535878e-06, "loss": 0.5197, "step": 16637 }, { "epoch": 1.98, "learning_rate": 1.3652385487106047e-06, "loss": 0.7256, "step": 16638 }, { "epoch": 1.98, "learning_rate": 1.3649521536430344e-06, "loss": 0.6953, "step": 16639 }, { "epoch": 1.98, "learning_rate": 1.3646657773378993e-06, "loss": 0.8563, "step": 16640 }, { "epoch": 1.98, "learning_rate": 1.3643794197999341e-06, "loss": 0.4035, "step": 16641 }, { "epoch": 1.98, "learning_rate": 1.3640930810338735e-06, "loss": 0.3062, "step": 16642 }, { "epoch": 1.98, "learning_rate": 1.3638067610444494e-06, "loss": 0.4951, "step": 16643 }, { "epoch": 1.98, "learning_rate": 1.3635204598363937e-06, "loss": 0.5868, "step": 16644 }, { "epoch": 1.98, "learning_rate": 1.3632341774144409e-06, "loss": 0.6715, "step": 16645 }, { "epoch": 1.98, "learning_rate": 1.3629479137833218e-06, "loss": 0.7058, "step": 16646 }, { "epoch": 1.98, "learning_rate": 1.3626616689477703e-06, "loss": 0.8606, "step": 16647 }, { "epoch": 1.98, "learning_rate": 1.3623754429125157e-06, "loss": 0.5153, "step": 16648 }, { "epoch": 1.98, "learning_rate": 1.3620892356822919e-06, "loss": 0.5391, "step": 16649 }, { "epoch": 1.98, "learning_rate": 1.3618030472618272e-06, "loss": 0.5393, "step": 16650 }, { "epoch": 1.98, "learning_rate": 1.3615168776558546e-06, "loss": 0.5895, "step": 16651 }, { "epoch": 1.98, "learning_rate": 1.3612307268691027e-06, "loss": 0.7987, "step": 16652 }, { "epoch": 1.98, "learning_rate": 1.3609445949063028e-06, "loss": 0.7878, "step": 16653 }, { "epoch": 1.98, "learning_rate": 1.3606584817721847e-06, "loss": 0.4161, "step": 16654 }, { "epoch": 1.98, "learning_rate": 1.3603723874714766e-06, "loss": 0.5386, "step": 16655 }, { "epoch": 1.98, "learning_rate": 1.3600863120089087e-06, "loss": 0.4993, "step": 16656 }, { "epoch": 1.98, "learning_rate": 1.3598002553892106e-06, "loss": 0.6525, "step": 16657 }, { "epoch": 1.98, "learning_rate": 1.35951421761711e-06, "loss": 0.8571, "step": 16658 }, { "epoch": 1.98, "learning_rate": 1.3592281986973343e-06, "loss": 0.8111, "step": 16659 }, { "epoch": 1.98, "learning_rate": 1.358942198634612e-06, "loss": 0.5455, "step": 16660 }, { "epoch": 1.98, "learning_rate": 1.358656217433671e-06, "loss": 0.4831, "step": 16661 }, { "epoch": 1.98, "learning_rate": 1.358370255099239e-06, "loss": 0.6033, "step": 16662 }, { "epoch": 1.98, "learning_rate": 1.358084311636042e-06, "loss": 0.5983, "step": 16663 }, { "epoch": 1.98, "learning_rate": 1.3577983870488079e-06, "loss": 0.6834, "step": 16664 }, { "epoch": 1.98, "learning_rate": 1.3575124813422612e-06, "loss": 0.7986, "step": 16665 }, { "epoch": 1.98, "learning_rate": 1.3572265945211305e-06, "loss": 0.3646, "step": 16666 }, { "epoch": 1.98, "learning_rate": 1.3569407265901389e-06, "loss": 0.3697, "step": 16667 }, { "epoch": 1.98, "learning_rate": 1.3566548775540129e-06, "loss": 0.4924, "step": 16668 }, { "epoch": 1.99, "learning_rate": 1.3563690474174779e-06, "loss": 0.6193, "step": 16669 }, { "epoch": 1.99, "learning_rate": 1.356083236185259e-06, "loss": 0.6744, "step": 16670 }, { "epoch": 1.99, "learning_rate": 1.3557974438620809e-06, "loss": 0.8875, "step": 16671 }, { "epoch": 1.99, "learning_rate": 1.3555116704526656e-06, "loss": 0.9045, "step": 16672 }, { "epoch": 1.99, "learning_rate": 1.3552259159617393e-06, "loss": 0.472, "step": 16673 }, { "epoch": 1.99, "learning_rate": 1.3549401803940238e-06, "loss": 0.5473, "step": 16674 }, { "epoch": 1.99, "learning_rate": 1.3546544637542433e-06, "loss": 0.5554, "step": 16675 }, { "epoch": 1.99, "learning_rate": 1.3543687660471205e-06, "loss": 0.7236, "step": 16676 }, { "epoch": 1.99, "learning_rate": 1.3540830872773792e-06, "loss": 0.7239, "step": 16677 }, { "epoch": 1.99, "learning_rate": 1.3537974274497406e-06, "loss": 0.8488, "step": 16678 }, { "epoch": 1.99, "learning_rate": 1.3535117865689256e-06, "loss": 0.4217, "step": 16679 }, { "epoch": 1.99, "learning_rate": 1.3532261646396571e-06, "loss": 0.4756, "step": 16680 }, { "epoch": 1.99, "learning_rate": 1.3529405616666573e-06, "loss": 0.5408, "step": 16681 }, { "epoch": 1.99, "learning_rate": 1.3526549776546455e-06, "loss": 0.528, "step": 16682 }, { "epoch": 1.99, "learning_rate": 1.352369412608343e-06, "loss": 0.7859, "step": 16683 }, { "epoch": 1.99, "learning_rate": 1.3520838665324704e-06, "loss": 0.8331, "step": 16684 }, { "epoch": 1.99, "learning_rate": 1.3517983394317485e-06, "loss": 0.4676, "step": 16685 }, { "epoch": 1.99, "learning_rate": 1.3515128313108966e-06, "loss": 0.6516, "step": 16686 }, { "epoch": 1.99, "learning_rate": 1.3512273421746334e-06, "loss": 0.5144, "step": 16687 }, { "epoch": 1.99, "learning_rate": 1.350941872027679e-06, "loss": 0.5968, "step": 16688 }, { "epoch": 1.99, "learning_rate": 1.3506564208747514e-06, "loss": 0.5529, "step": 16689 }, { "epoch": 1.99, "learning_rate": 1.350370988720569e-06, "loss": 0.7982, "step": 16690 }, { "epoch": 1.99, "learning_rate": 1.350085575569851e-06, "loss": 0.6433, "step": 16691 }, { "epoch": 1.99, "learning_rate": 1.349800181427316e-06, "loss": 0.5254, "step": 16692 }, { "epoch": 1.99, "learning_rate": 1.3495148062976803e-06, "loss": 0.5893, "step": 16693 }, { "epoch": 1.99, "learning_rate": 1.3492294501856603e-06, "loss": 0.5298, "step": 16694 }, { "epoch": 1.99, "learning_rate": 1.3489441130959743e-06, "loss": 0.6558, "step": 16695 }, { "epoch": 1.99, "learning_rate": 1.3486587950333397e-06, "loss": 0.8006, "step": 16696 }, { "epoch": 1.99, "learning_rate": 1.3483734960024705e-06, "loss": 0.7897, "step": 16697 }, { "epoch": 1.99, "learning_rate": 1.3480882160080844e-06, "loss": 0.4581, "step": 16698 }, { "epoch": 1.99, "learning_rate": 1.3478029550548978e-06, "loss": 0.5323, "step": 16699 }, { "epoch": 1.99, "learning_rate": 1.3475177131476247e-06, "loss": 0.6059, "step": 16700 }, { "epoch": 1.99, "learning_rate": 1.3472324902909794e-06, "loss": 0.6906, "step": 16701 }, { "epoch": 1.99, "learning_rate": 1.3469472864896782e-06, "loss": 0.7159, "step": 16702 }, { "epoch": 1.99, "learning_rate": 1.3466621017484358e-06, "loss": 0.7182, "step": 16703 }, { "epoch": 1.99, "learning_rate": 1.3463769360719647e-06, "loss": 0.4606, "step": 16704 }, { "epoch": 1.99, "learning_rate": 1.3460917894649797e-06, "loss": 0.5074, "step": 16705 }, { "epoch": 1.99, "learning_rate": 1.3458066619321951e-06, "loss": 0.5437, "step": 16706 }, { "epoch": 1.99, "learning_rate": 1.3455215534783233e-06, "loss": 0.5607, "step": 16707 }, { "epoch": 1.99, "learning_rate": 1.345236464108076e-06, "loss": 0.7562, "step": 16708 }, { "epoch": 1.99, "learning_rate": 1.3449513938261671e-06, "loss": 0.6948, "step": 16709 }, { "epoch": 1.99, "learning_rate": 1.3446663426373085e-06, "loss": 0.4825, "step": 16710 }, { "epoch": 1.99, "learning_rate": 1.3443813105462134e-06, "loss": 0.4393, "step": 16711 }, { "epoch": 1.99, "learning_rate": 1.3440962975575906e-06, "loss": 0.6327, "step": 16712 }, { "epoch": 1.99, "learning_rate": 1.3438113036761536e-06, "loss": 0.627, "step": 16713 }, { "epoch": 1.99, "learning_rate": 1.3435263289066134e-06, "loss": 0.6713, "step": 16714 }, { "epoch": 1.99, "learning_rate": 1.34324137325368e-06, "loss": 0.7963, "step": 16715 }, { "epoch": 1.99, "learning_rate": 1.3429564367220626e-06, "loss": 0.5321, "step": 16716 }, { "epoch": 1.99, "learning_rate": 1.3426715193164729e-06, "loss": 0.369, "step": 16717 }, { "epoch": 1.99, "learning_rate": 1.3423866210416208e-06, "loss": 0.4816, "step": 16718 }, { "epoch": 1.99, "learning_rate": 1.3421017419022143e-06, "loss": 0.5165, "step": 16719 }, { "epoch": 1.99, "learning_rate": 1.3418168819029625e-06, "loss": 0.7944, "step": 16720 }, { "epoch": 1.99, "learning_rate": 1.3415320410485766e-06, "loss": 0.9133, "step": 16721 }, { "epoch": 1.99, "learning_rate": 1.3412472193437627e-06, "loss": 0.8516, "step": 16722 }, { "epoch": 1.99, "learning_rate": 1.3409624167932286e-06, "loss": 0.4314, "step": 16723 }, { "epoch": 1.99, "learning_rate": 1.340677633401683e-06, "loss": 0.4824, "step": 16724 }, { "epoch": 1.99, "learning_rate": 1.3403928691738336e-06, "loss": 0.4638, "step": 16725 }, { "epoch": 1.99, "learning_rate": 1.3401081241143881e-06, "loss": 0.6887, "step": 16726 }, { "epoch": 1.99, "learning_rate": 1.3398233982280518e-06, "loss": 0.8161, "step": 16727 }, { "epoch": 1.99, "learning_rate": 1.339538691519533e-06, "loss": 0.9057, "step": 16728 }, { "epoch": 1.99, "learning_rate": 1.3392540039935361e-06, "loss": 0.4903, "step": 16729 }, { "epoch": 1.99, "learning_rate": 1.338969335654769e-06, "loss": 0.409, "step": 16730 }, { "epoch": 1.99, "learning_rate": 1.338684686507935e-06, "loss": 0.5408, "step": 16731 }, { "epoch": 1.99, "learning_rate": 1.3384000565577407e-06, "loss": 0.6658, "step": 16732 }, { "epoch": 1.99, "learning_rate": 1.3381154458088908e-06, "loss": 0.7021, "step": 16733 }, { "epoch": 1.99, "learning_rate": 1.3378308542660912e-06, "loss": 0.8129, "step": 16734 }, { "epoch": 1.99, "learning_rate": 1.3375462819340452e-06, "loss": 0.3148, "step": 16735 }, { "epoch": 1.99, "learning_rate": 1.3372617288174552e-06, "loss": 0.5033, "step": 16736 }, { "epoch": 1.99, "learning_rate": 1.3369771949210278e-06, "loss": 0.5486, "step": 16737 }, { "epoch": 1.99, "learning_rate": 1.3366926802494639e-06, "loss": 0.5895, "step": 16738 }, { "epoch": 1.99, "learning_rate": 1.3364081848074675e-06, "loss": 0.7432, "step": 16739 }, { "epoch": 1.99, "learning_rate": 1.3361237085997413e-06, "loss": 0.7387, "step": 16740 }, { "epoch": 1.99, "learning_rate": 1.3358392516309893e-06, "loss": 0.4697, "step": 16741 }, { "epoch": 1.99, "learning_rate": 1.3355548139059111e-06, "loss": 0.4872, "step": 16742 }, { "epoch": 1.99, "learning_rate": 1.33527039542921e-06, "loss": 0.5668, "step": 16743 }, { "epoch": 1.99, "learning_rate": 1.3349859962055865e-06, "loss": 0.5656, "step": 16744 }, { "epoch": 1.99, "learning_rate": 1.334701616239743e-06, "loss": 0.726, "step": 16745 }, { "epoch": 1.99, "learning_rate": 1.3344172555363789e-06, "loss": 0.827, "step": 16746 }, { "epoch": 1.99, "learning_rate": 1.3341329141001947e-06, "loss": 0.8879, "step": 16747 }, { "epoch": 1.99, "learning_rate": 1.3338485919358918e-06, "loss": 0.4578, "step": 16748 }, { "epoch": 1.99, "learning_rate": 1.3335642890481703e-06, "loss": 0.5071, "step": 16749 }, { "epoch": 1.99, "learning_rate": 1.3332800054417292e-06, "loss": 0.7033, "step": 16750 }, { "epoch": 1.99, "learning_rate": 1.3329957411212662e-06, "loss": 0.6826, "step": 16751 }, { "epoch": 1.99, "learning_rate": 1.3327114960914827e-06, "loss": 0.7391, "step": 16752 }, { "epoch": 2.0, "learning_rate": 1.3324272703570752e-06, "loss": 0.8057, "step": 16753 }, { "epoch": 2.0, "learning_rate": 1.3321430639227428e-06, "loss": 0.3031, "step": 16754 }, { "epoch": 2.0, "learning_rate": 1.3318588767931835e-06, "loss": 0.5732, "step": 16755 }, { "epoch": 2.0, "learning_rate": 1.3315747089730956e-06, "loss": 0.5449, "step": 16756 }, { "epoch": 2.0, "learning_rate": 1.3312905604671767e-06, "loss": 0.6931, "step": 16757 }, { "epoch": 2.0, "learning_rate": 1.3310064312801214e-06, "loss": 0.756, "step": 16758 }, { "epoch": 2.0, "learning_rate": 1.3307223214166276e-06, "loss": 0.8135, "step": 16759 }, { "epoch": 2.0, "learning_rate": 1.3304382308813935e-06, "loss": 0.4272, "step": 16760 }, { "epoch": 2.0, "learning_rate": 1.330154159679112e-06, "loss": 0.4906, "step": 16761 }, { "epoch": 2.0, "learning_rate": 1.3298701078144807e-06, "loss": 0.5129, "step": 16762 }, { "epoch": 2.0, "learning_rate": 1.3295860752921958e-06, "loss": 0.631, "step": 16763 }, { "epoch": 2.0, "learning_rate": 1.329302062116951e-06, "loss": 0.7853, "step": 16764 }, { "epoch": 2.0, "learning_rate": 1.3290180682934406e-06, "loss": 0.8149, "step": 16765 }, { "epoch": 2.0, "learning_rate": 1.3287340938263598e-06, "loss": 0.3807, "step": 16766 }, { "epoch": 2.0, "learning_rate": 1.3284501387204035e-06, "loss": 0.4693, "step": 16767 }, { "epoch": 2.0, "learning_rate": 1.3281662029802639e-06, "loss": 0.4242, "step": 16768 }, { "epoch": 2.0, "learning_rate": 1.327882286610635e-06, "loss": 0.6543, "step": 16769 }, { "epoch": 2.0, "learning_rate": 1.3275983896162105e-06, "loss": 0.7755, "step": 16770 }, { "epoch": 2.0, "learning_rate": 1.327314512001684e-06, "loss": 0.7562, "step": 16771 }, { "epoch": 2.0, "learning_rate": 1.327030653771747e-06, "loss": 0.8381, "step": 16772 }, { "epoch": 2.0, "learning_rate": 1.3267468149310907e-06, "loss": 0.4067, "step": 16773 }, { "epoch": 2.0, "learning_rate": 1.326462995484408e-06, "loss": 0.4696, "step": 16774 }, { "epoch": 2.0, "learning_rate": 1.326179195436391e-06, "loss": 0.6738, "step": 16775 }, { "epoch": 2.0, "learning_rate": 1.3258954147917301e-06, "loss": 0.5854, "step": 16776 }, { "epoch": 2.0, "learning_rate": 1.325611653555116e-06, "loss": 0.7598, "step": 16777 }, { "epoch": 2.0, "learning_rate": 1.3253279117312412e-06, "loss": 0.6397, "step": 16778 }, { "epoch": 2.0, "learning_rate": 1.3250441893247945e-06, "loss": 0.4237, "step": 16779 }, { "epoch": 2.0, "learning_rate": 1.3247604863404647e-06, "loss": 0.422, "step": 16780 }, { "epoch": 2.0, "learning_rate": 1.324476802782943e-06, "loss": 0.5593, "step": 16781 }, { "epoch": 2.0, "learning_rate": 1.324193138656919e-06, "loss": 0.6339, "step": 16782 }, { "epoch": 2.0, "learning_rate": 1.3239094939670806e-06, "loss": 0.7537, "step": 16783 }, { "epoch": 2.0, "learning_rate": 1.3236258687181164e-06, "loss": 0.8016, "step": 16784 }, { "epoch": 2.0, "learning_rate": 1.3233422629147167e-06, "loss": 0.4366, "step": 16785 }, { "epoch": 2.0, "learning_rate": 1.3230586765615677e-06, "loss": 0.5889, "step": 16786 }, { "epoch": 2.0, "learning_rate": 1.3227751096633567e-06, "loss": 0.5206, "step": 16787 }, { "epoch": 2.0, "learning_rate": 1.3224915622247722e-06, "loss": 0.5921, "step": 16788 }, { "epoch": 2.0, "learning_rate": 1.3222080342505006e-06, "loss": 0.7257, "step": 16789 }, { "epoch": 2.0, "learning_rate": 1.3219245257452303e-06, "loss": 0.7814, "step": 16790 }, { "epoch": 2.0, "learning_rate": 1.3216410367136453e-06, "loss": 0.3978, "step": 16791 }, { "epoch": 2.0, "learning_rate": 1.3213575671604338e-06, "loss": 0.525, "step": 16792 }, { "epoch": 2.0, "learning_rate": 1.3210741170902797e-06, "loss": 0.7417, "step": 16793 }, { "epoch": 2.0, "learning_rate": 1.3207906865078704e-06, "loss": 0.874, "step": 16794 }, { "epoch": 2.0, "learning_rate": 1.320507275417889e-06, "loss": 0.3316, "step": 16795 }, { "epoch": 2.0, "learning_rate": 1.320223883825021e-06, "loss": 0.4536, "step": 16796 }, { "epoch": 2.0, "learning_rate": 1.3199405117339525e-06, "loss": 0.3618, "step": 16797 }, { "epoch": 2.0, "learning_rate": 1.3196571591493653e-06, "loss": 0.4501, "step": 16798 }, { "epoch": 2.0, "learning_rate": 1.3193738260759443e-06, "loss": 0.5938, "step": 16799 }, { "epoch": 2.0, "learning_rate": 1.3190905125183739e-06, "loss": 0.6283, "step": 16800 }, { "epoch": 2.0, "learning_rate": 1.3188072184813365e-06, "loss": 0.447, "step": 16801 }, { "epoch": 2.0, "learning_rate": 1.318523943969514e-06, "loss": 0.3628, "step": 16802 }, { "epoch": 2.0, "learning_rate": 1.3182406889875897e-06, "loss": 0.4319, "step": 16803 }, { "epoch": 2.0, "learning_rate": 1.3179574535402461e-06, "loss": 0.486, "step": 16804 }, { "epoch": 2.0, "learning_rate": 1.317674237632166e-06, "loss": 0.6123, "step": 16805 }, { "epoch": 2.0, "learning_rate": 1.3173910412680288e-06, "loss": 0.6231, "step": 16806 }, { "epoch": 2.0, "learning_rate": 1.3171078644525178e-06, "loss": 0.3264, "step": 16807 }, { "epoch": 2.0, "learning_rate": 1.3168247071903124e-06, "loss": 0.436, "step": 16808 }, { "epoch": 2.0, "learning_rate": 1.3165415694860946e-06, "loss": 0.4858, "step": 16809 }, { "epoch": 2.0, "learning_rate": 1.316258451344543e-06, "loss": 0.4649, "step": 16810 }, { "epoch": 2.0, "learning_rate": 1.3159753527703385e-06, "loss": 0.4815, "step": 16811 }, { "epoch": 2.0, "learning_rate": 1.315692273768161e-06, "loss": 0.545, "step": 16812 }, { "epoch": 2.0, "learning_rate": 1.3154092143426902e-06, "loss": 0.4409, "step": 16813 }, { "epoch": 2.0, "learning_rate": 1.3151261744986049e-06, "loss": 0.302, "step": 16814 }, { "epoch": 2.0, "learning_rate": 1.314843154240582e-06, "loss": 0.4085, "step": 16815 }, { "epoch": 2.0, "learning_rate": 1.3145601535733025e-06, "loss": 0.4732, "step": 16816 }, { "epoch": 2.0, "learning_rate": 1.3142771725014419e-06, "loss": 0.481, "step": 16817 }, { "epoch": 2.0, "learning_rate": 1.3139942110296793e-06, "loss": 0.6103, "step": 16818 }, { "epoch": 2.0, "learning_rate": 1.3137112691626919e-06, "loss": 0.5295, "step": 16819 }, { "epoch": 2.0, "learning_rate": 1.3134283469051573e-06, "loss": 0.3225, "step": 16820 }, { "epoch": 2.0, "learning_rate": 1.3131454442617523e-06, "loss": 0.4056, "step": 16821 }, { "epoch": 2.0, "learning_rate": 1.3128625612371514e-06, "loss": 0.3594, "step": 16822 }, { "epoch": 2.0, "learning_rate": 1.312579697836032e-06, "loss": 0.4615, "step": 16823 }, { "epoch": 2.0, "learning_rate": 1.3122968540630709e-06, "loss": 0.5455, "step": 16824 }, { "epoch": 2.0, "learning_rate": 1.3120140299229414e-06, "loss": 0.5678, "step": 16825 }, { "epoch": 2.0, "learning_rate": 1.3117312254203195e-06, "loss": 0.3219, "step": 16826 }, { "epoch": 2.0, "learning_rate": 1.3114484405598803e-06, "loss": 0.4516, "step": 16827 }, { "epoch": 2.0, "learning_rate": 1.3111656753462992e-06, "loss": 0.3796, "step": 16828 }, { "epoch": 2.0, "learning_rate": 1.3108829297842492e-06, "loss": 0.545, "step": 16829 }, { "epoch": 2.0, "learning_rate": 1.3106002038784032e-06, "loss": 0.5571, "step": 16830 }, { "epoch": 2.0, "learning_rate": 1.3103174976334365e-06, "loss": 0.5893, "step": 16831 }, { "epoch": 2.0, "learning_rate": 1.3100348110540207e-06, "loss": 0.3062, "step": 16832 }, { "epoch": 2.0, "learning_rate": 1.3097521441448289e-06, "loss": 0.4154, "step": 16833 }, { "epoch": 2.0, "learning_rate": 1.3094694969105342e-06, "loss": 0.4211, "step": 16834 }, { "epoch": 2.0, "learning_rate": 1.3091868693558095e-06, "loss": 0.4902, "step": 16835 }, { "epoch": 2.01, "learning_rate": 1.308904261485326e-06, "loss": 0.5301, "step": 16836 }, { "epoch": 2.01, "learning_rate": 1.3086216733037539e-06, "loss": 0.6268, "step": 16837 }, { "epoch": 2.01, "learning_rate": 1.3083391048157652e-06, "loss": 0.4995, "step": 16838 }, { "epoch": 2.01, "learning_rate": 1.3080565560260323e-06, "loss": 0.3205, "step": 16839 }, { "epoch": 2.01, "learning_rate": 1.3077740269392236e-06, "loss": 0.4314, "step": 16840 }, { "epoch": 2.01, "learning_rate": 1.3074915175600103e-06, "loss": 0.5275, "step": 16841 }, { "epoch": 2.01, "learning_rate": 1.3072090278930632e-06, "loss": 0.6039, "step": 16842 }, { "epoch": 2.01, "learning_rate": 1.306926557943051e-06, "loss": 0.564, "step": 16843 }, { "epoch": 2.01, "learning_rate": 1.3066441077146418e-06, "loss": 0.5586, "step": 16844 }, { "epoch": 2.01, "learning_rate": 1.3063616772125059e-06, "loss": 0.3426, "step": 16845 }, { "epoch": 2.01, "learning_rate": 1.3060792664413125e-06, "loss": 0.3905, "step": 16846 }, { "epoch": 2.01, "learning_rate": 1.3057968754057276e-06, "loss": 0.4026, "step": 16847 }, { "epoch": 2.01, "learning_rate": 1.305514504110421e-06, "loss": 0.5663, "step": 16848 }, { "epoch": 2.01, "learning_rate": 1.305232152560061e-06, "loss": 0.5349, "step": 16849 }, { "epoch": 2.01, "learning_rate": 1.3049498207593135e-06, "loss": 0.6568, "step": 16850 }, { "epoch": 2.01, "learning_rate": 1.3046675087128446e-06, "loss": 0.2845, "step": 16851 }, { "epoch": 2.01, "learning_rate": 1.3043852164253222e-06, "loss": 0.4154, "step": 16852 }, { "epoch": 2.01, "learning_rate": 1.3041029439014128e-06, "loss": 0.4679, "step": 16853 }, { "epoch": 2.01, "learning_rate": 1.303820691145783e-06, "loss": 0.5173, "step": 16854 }, { "epoch": 2.01, "learning_rate": 1.303538458163096e-06, "loss": 0.6041, "step": 16855 }, { "epoch": 2.01, "learning_rate": 1.3032562449580192e-06, "loss": 0.5876, "step": 16856 }, { "epoch": 2.01, "learning_rate": 1.302974051535218e-06, "loss": 0.4377, "step": 16857 }, { "epoch": 2.01, "learning_rate": 1.3026918778993558e-06, "loss": 0.3038, "step": 16858 }, { "epoch": 2.01, "learning_rate": 1.3024097240550967e-06, "loss": 0.4261, "step": 16859 }, { "epoch": 2.01, "learning_rate": 1.3021275900071055e-06, "loss": 0.4385, "step": 16860 }, { "epoch": 2.01, "learning_rate": 1.3018454757600463e-06, "loss": 0.5103, "step": 16861 }, { "epoch": 2.01, "learning_rate": 1.3015633813185814e-06, "loss": 0.6239, "step": 16862 }, { "epoch": 2.01, "learning_rate": 1.301281306687374e-06, "loss": 0.3777, "step": 16863 }, { "epoch": 2.01, "learning_rate": 1.300999251871088e-06, "loss": 0.472, "step": 16864 }, { "epoch": 2.01, "learning_rate": 1.3007172168743854e-06, "loss": 0.3919, "step": 16865 }, { "epoch": 2.01, "learning_rate": 1.3004352017019266e-06, "loss": 0.4187, "step": 16866 }, { "epoch": 2.01, "learning_rate": 1.300153206358375e-06, "loss": 0.534, "step": 16867 }, { "epoch": 2.01, "learning_rate": 1.299871230848391e-06, "loss": 0.6118, "step": 16868 }, { "epoch": 2.01, "learning_rate": 1.2995892751766377e-06, "loss": 0.581, "step": 16869 }, { "epoch": 2.01, "learning_rate": 1.299307339347773e-06, "loss": 0.3254, "step": 16870 }, { "epoch": 2.01, "learning_rate": 1.29902542336646e-06, "loss": 0.4215, "step": 16871 }, { "epoch": 2.01, "learning_rate": 1.2987435272373564e-06, "loss": 0.4998, "step": 16872 }, { "epoch": 2.01, "learning_rate": 1.2984616509651241e-06, "loss": 0.5029, "step": 16873 }, { "epoch": 2.01, "learning_rate": 1.2981797945544206e-06, "loss": 0.6616, "step": 16874 }, { "epoch": 2.01, "learning_rate": 1.2978979580099057e-06, "loss": 0.6622, "step": 16875 }, { "epoch": 2.01, "learning_rate": 1.2976161413362387e-06, "loss": 0.2506, "step": 16876 }, { "epoch": 2.01, "learning_rate": 1.2973343445380787e-06, "loss": 0.3042, "step": 16877 }, { "epoch": 2.01, "learning_rate": 1.297052567620083e-06, "loss": 0.4458, "step": 16878 }, { "epoch": 2.01, "learning_rate": 1.296770810586908e-06, "loss": 0.4883, "step": 16879 }, { "epoch": 2.01, "learning_rate": 1.2964890734432137e-06, "loss": 0.5313, "step": 16880 }, { "epoch": 2.01, "learning_rate": 1.296207356193655e-06, "loss": 0.5893, "step": 16881 }, { "epoch": 2.01, "learning_rate": 1.2959256588428895e-06, "loss": 0.3653, "step": 16882 }, { "epoch": 2.01, "learning_rate": 1.295643981395574e-06, "loss": 0.3728, "step": 16883 }, { "epoch": 2.01, "learning_rate": 1.2953623238563657e-06, "loss": 0.5511, "step": 16884 }, { "epoch": 2.01, "learning_rate": 1.2950806862299181e-06, "loss": 0.4659, "step": 16885 }, { "epoch": 2.01, "learning_rate": 1.294799068520889e-06, "loss": 0.5333, "step": 16886 }, { "epoch": 2.01, "learning_rate": 1.294517470733931e-06, "loss": 0.5503, "step": 16887 }, { "epoch": 2.01, "learning_rate": 1.2942358928737018e-06, "loss": 0.4398, "step": 16888 }, { "epoch": 2.01, "learning_rate": 1.2939543349448534e-06, "loss": 0.2883, "step": 16889 }, { "epoch": 2.01, "learning_rate": 1.2936727969520407e-06, "loss": 0.3666, "step": 16890 }, { "epoch": 2.01, "learning_rate": 1.2933912788999184e-06, "loss": 0.4814, "step": 16891 }, { "epoch": 2.01, "learning_rate": 1.2931097807931403e-06, "loss": 0.5082, "step": 16892 }, { "epoch": 2.01, "learning_rate": 1.2928283026363586e-06, "loss": 0.5508, "step": 16893 }, { "epoch": 2.01, "learning_rate": 1.2925468444342253e-06, "loss": 0.5869, "step": 16894 }, { "epoch": 2.01, "learning_rate": 1.2922654061913953e-06, "loss": 0.2753, "step": 16895 }, { "epoch": 2.01, "learning_rate": 1.2919839879125185e-06, "loss": 0.3349, "step": 16896 }, { "epoch": 2.01, "learning_rate": 1.2917025896022474e-06, "loss": 0.4025, "step": 16897 }, { "epoch": 2.01, "learning_rate": 1.2914212112652342e-06, "loss": 0.507, "step": 16898 }, { "epoch": 2.01, "learning_rate": 1.2911398529061306e-06, "loss": 0.4991, "step": 16899 }, { "epoch": 2.01, "learning_rate": 1.2908585145295865e-06, "loss": 0.5416, "step": 16900 }, { "epoch": 2.01, "learning_rate": 1.290577196140252e-06, "loss": 0.2861, "step": 16901 }, { "epoch": 2.01, "learning_rate": 1.2902958977427776e-06, "loss": 0.4145, "step": 16902 }, { "epoch": 2.01, "learning_rate": 1.290014619341815e-06, "loss": 0.4286, "step": 16903 }, { "epoch": 2.01, "learning_rate": 1.289733360942011e-06, "loss": 0.4008, "step": 16904 }, { "epoch": 2.01, "learning_rate": 1.2894521225480158e-06, "loss": 0.5939, "step": 16905 }, { "epoch": 2.01, "learning_rate": 1.2891709041644798e-06, "loss": 0.6482, "step": 16906 }, { "epoch": 2.01, "learning_rate": 1.2888897057960503e-06, "loss": 0.272, "step": 16907 }, { "epoch": 2.01, "learning_rate": 1.2886085274473748e-06, "loss": 0.3347, "step": 16908 }, { "epoch": 2.01, "learning_rate": 1.288327369123102e-06, "loss": 0.4396, "step": 16909 }, { "epoch": 2.01, "learning_rate": 1.2880462308278802e-06, "loss": 0.4651, "step": 16910 }, { "epoch": 2.01, "learning_rate": 1.2877651125663549e-06, "loss": 0.6003, "step": 16911 }, { "epoch": 2.01, "learning_rate": 1.2874840143431741e-06, "loss": 0.6217, "step": 16912 }, { "epoch": 2.01, "learning_rate": 1.2872029361629842e-06, "loss": 0.3934, "step": 16913 }, { "epoch": 2.01, "learning_rate": 1.2869218780304327e-06, "loss": 0.3059, "step": 16914 }, { "epoch": 2.01, "learning_rate": 1.286640839950164e-06, "loss": 0.3876, "step": 16915 }, { "epoch": 2.01, "learning_rate": 1.2863598219268234e-06, "loss": 0.4029, "step": 16916 }, { "epoch": 2.01, "learning_rate": 1.2860788239650568e-06, "loss": 0.4324, "step": 16917 }, { "epoch": 2.01, "learning_rate": 1.2857978460695097e-06, "loss": 0.585, "step": 16918 }, { "epoch": 2.01, "learning_rate": 1.2855168882448254e-06, "loss": 0.4876, "step": 16919 }, { "epoch": 2.02, "learning_rate": 1.285235950495649e-06, "loss": 0.3236, "step": 16920 }, { "epoch": 2.02, "learning_rate": 1.2849550328266247e-06, "loss": 0.4538, "step": 16921 }, { "epoch": 2.02, "learning_rate": 1.284674135242396e-06, "loss": 0.4099, "step": 16922 }, { "epoch": 2.02, "learning_rate": 1.2843932577476048e-06, "loss": 0.449, "step": 16923 }, { "epoch": 2.02, "learning_rate": 1.2841124003468951e-06, "loss": 0.5957, "step": 16924 }, { "epoch": 2.02, "learning_rate": 1.2838315630449105e-06, "loss": 0.5201, "step": 16925 }, { "epoch": 2.02, "learning_rate": 1.283550745846291e-06, "loss": 0.2901, "step": 16926 }, { "epoch": 2.02, "learning_rate": 1.28326994875568e-06, "loss": 0.3844, "step": 16927 }, { "epoch": 2.02, "learning_rate": 1.2829891717777192e-06, "loss": 0.4763, "step": 16928 }, { "epoch": 2.02, "learning_rate": 1.2827084149170504e-06, "loss": 0.4517, "step": 16929 }, { "epoch": 2.02, "learning_rate": 1.282427678178312e-06, "loss": 0.4679, "step": 16930 }, { "epoch": 2.02, "learning_rate": 1.2821469615661466e-06, "loss": 0.5913, "step": 16931 }, { "epoch": 2.02, "learning_rate": 1.281866265085194e-06, "loss": 0.3641, "step": 16932 }, { "epoch": 2.02, "learning_rate": 1.2815855887400952e-06, "loss": 0.305, "step": 16933 }, { "epoch": 2.02, "learning_rate": 1.2813049325354879e-06, "loss": 0.4323, "step": 16934 }, { "epoch": 2.02, "learning_rate": 1.2810242964760134e-06, "loss": 0.439, "step": 16935 }, { "epoch": 2.02, "learning_rate": 1.2807436805663082e-06, "loss": 0.6149, "step": 16936 }, { "epoch": 2.02, "learning_rate": 1.2804630848110137e-06, "loss": 0.5357, "step": 16937 }, { "epoch": 2.02, "learning_rate": 1.2801825092147656e-06, "loss": 0.3522, "step": 16938 }, { "epoch": 2.02, "learning_rate": 1.2799019537822027e-06, "loss": 0.3096, "step": 16939 }, { "epoch": 2.02, "learning_rate": 1.2796214185179634e-06, "loss": 0.4584, "step": 16940 }, { "epoch": 2.02, "learning_rate": 1.2793409034266852e-06, "loss": 0.4298, "step": 16941 }, { "epoch": 2.02, "learning_rate": 1.279060408513003e-06, "loss": 0.5053, "step": 16942 }, { "epoch": 2.02, "learning_rate": 1.278779933781556e-06, "loss": 0.5484, "step": 16943 }, { "epoch": 2.02, "learning_rate": 1.2784994792369793e-06, "loss": 0.5845, "step": 16944 }, { "epoch": 2.02, "learning_rate": 1.2782190448839074e-06, "loss": 0.375, "step": 16945 }, { "epoch": 2.02, "learning_rate": 1.2779386307269773e-06, "loss": 0.2822, "step": 16946 }, { "epoch": 2.02, "learning_rate": 1.277658236770824e-06, "loss": 0.4847, "step": 16947 }, { "epoch": 2.02, "learning_rate": 1.2773778630200839e-06, "loss": 0.5734, "step": 16948 }, { "epoch": 2.02, "learning_rate": 1.277097509479389e-06, "loss": 0.5667, "step": 16949 }, { "epoch": 2.02, "learning_rate": 1.2768171761533765e-06, "loss": 0.5397, "step": 16950 }, { "epoch": 2.02, "learning_rate": 1.2765368630466774e-06, "loss": 0.3484, "step": 16951 }, { "epoch": 2.02, "learning_rate": 1.2762565701639273e-06, "loss": 0.3818, "step": 16952 }, { "epoch": 2.02, "learning_rate": 1.2759762975097578e-06, "loss": 0.5426, "step": 16953 }, { "epoch": 2.02, "learning_rate": 1.2756960450888032e-06, "loss": 0.5625, "step": 16954 }, { "epoch": 2.02, "learning_rate": 1.2754158129056953e-06, "loss": 0.6621, "step": 16955 }, { "epoch": 2.02, "learning_rate": 1.275135600965068e-06, "loss": 0.5456, "step": 16956 }, { "epoch": 2.02, "learning_rate": 1.274855409271552e-06, "loss": 0.3137, "step": 16957 }, { "epoch": 2.02, "learning_rate": 1.2745752378297777e-06, "loss": 0.4145, "step": 16958 }, { "epoch": 2.02, "learning_rate": 1.2742950866443788e-06, "loss": 0.4629, "step": 16959 }, { "epoch": 2.02, "learning_rate": 1.2740149557199839e-06, "loss": 0.4358, "step": 16960 }, { "epoch": 2.02, "learning_rate": 1.2737348450612247e-06, "loss": 0.4684, "step": 16961 }, { "epoch": 2.02, "learning_rate": 1.273454754672731e-06, "loss": 0.6685, "step": 16962 }, { "epoch": 2.02, "learning_rate": 1.2731746845591345e-06, "loss": 0.4192, "step": 16963 }, { "epoch": 2.02, "learning_rate": 1.272894634725063e-06, "loss": 0.3371, "step": 16964 }, { "epoch": 2.02, "learning_rate": 1.272614605175146e-06, "loss": 0.3462, "step": 16965 }, { "epoch": 2.02, "learning_rate": 1.2723345959140118e-06, "loss": 0.463, "step": 16966 }, { "epoch": 2.02, "learning_rate": 1.2720546069462907e-06, "loss": 0.514, "step": 16967 }, { "epoch": 2.02, "learning_rate": 1.2717746382766094e-06, "loss": 0.6011, "step": 16968 }, { "epoch": 2.02, "learning_rate": 1.2714946899095964e-06, "loss": 0.5563, "step": 16969 }, { "epoch": 2.02, "learning_rate": 1.271214761849879e-06, "loss": 0.2786, "step": 16970 }, { "epoch": 2.02, "learning_rate": 1.2709348541020859e-06, "loss": 0.3666, "step": 16971 }, { "epoch": 2.02, "learning_rate": 1.2706549666708426e-06, "loss": 0.4193, "step": 16972 }, { "epoch": 2.02, "learning_rate": 1.270375099560775e-06, "loss": 0.5163, "step": 16973 }, { "epoch": 2.02, "learning_rate": 1.270095252776511e-06, "loss": 0.5846, "step": 16974 }, { "epoch": 2.02, "learning_rate": 1.2698154263226748e-06, "loss": 0.5301, "step": 16975 }, { "epoch": 2.02, "learning_rate": 1.2695356202038928e-06, "loss": 0.2857, "step": 16976 }, { "epoch": 2.02, "learning_rate": 1.2692558344247901e-06, "loss": 0.388, "step": 16977 }, { "epoch": 2.02, "learning_rate": 1.2689760689899927e-06, "loss": 0.4359, "step": 16978 }, { "epoch": 2.02, "learning_rate": 1.2686963239041244e-06, "loss": 0.3806, "step": 16979 }, { "epoch": 2.02, "learning_rate": 1.2684165991718083e-06, "loss": 0.538, "step": 16980 }, { "epoch": 2.02, "learning_rate": 1.2681368947976691e-06, "loss": 0.6505, "step": 16981 }, { "epoch": 2.02, "learning_rate": 1.267857210786331e-06, "loss": 0.3034, "step": 16982 }, { "epoch": 2.02, "learning_rate": 1.267577547142416e-06, "loss": 0.3926, "step": 16983 }, { "epoch": 2.02, "learning_rate": 1.2672979038705475e-06, "loss": 0.3716, "step": 16984 }, { "epoch": 2.02, "learning_rate": 1.2670182809753486e-06, "loss": 0.4453, "step": 16985 }, { "epoch": 2.02, "learning_rate": 1.2667386784614415e-06, "loss": 0.4606, "step": 16986 }, { "epoch": 2.02, "learning_rate": 1.2664590963334463e-06, "loss": 0.6547, "step": 16987 }, { "epoch": 2.02, "learning_rate": 1.266179534595986e-06, "loss": 0.431, "step": 16988 }, { "epoch": 2.02, "learning_rate": 1.2658999932536822e-06, "loss": 0.3704, "step": 16989 }, { "epoch": 2.02, "learning_rate": 1.2656204723111543e-06, "loss": 0.3834, "step": 16990 }, { "epoch": 2.02, "learning_rate": 1.2653409717730233e-06, "loss": 0.5145, "step": 16991 }, { "epoch": 2.02, "learning_rate": 1.2650614916439108e-06, "loss": 0.6733, "step": 16992 }, { "epoch": 2.02, "learning_rate": 1.2647820319284346e-06, "loss": 0.6569, "step": 16993 }, { "epoch": 2.02, "learning_rate": 1.264502592631216e-06, "loss": 0.475, "step": 16994 }, { "epoch": 2.02, "learning_rate": 1.264223173756872e-06, "loss": 0.3673, "step": 16995 }, { "epoch": 2.02, "learning_rate": 1.2639437753100225e-06, "loss": 0.4784, "step": 16996 }, { "epoch": 2.02, "learning_rate": 1.2636643972952875e-06, "loss": 0.4043, "step": 16997 }, { "epoch": 2.02, "learning_rate": 1.2633850397172826e-06, "loss": 0.5662, "step": 16998 }, { "epoch": 2.02, "learning_rate": 1.263105702580627e-06, "loss": 0.6467, "step": 16999 }, { "epoch": 2.02, "learning_rate": 1.2628263858899387e-06, "loss": 0.5315, "step": 17000 }, { "epoch": 2.02, "learning_rate": 1.2625470896498337e-06, "loss": 0.2224, "step": 17001 }, { "epoch": 2.02, "learning_rate": 1.2622678138649286e-06, "loss": 0.365, "step": 17002 }, { "epoch": 2.02, "learning_rate": 1.2619885585398403e-06, "loss": 0.4072, "step": 17003 }, { "epoch": 2.03, "learning_rate": 1.2617093236791861e-06, "loss": 0.5208, "step": 17004 }, { "epoch": 2.03, "learning_rate": 1.2614301092875795e-06, "loss": 0.4543, "step": 17005 }, { "epoch": 2.03, "learning_rate": 1.2611509153696372e-06, "loss": 0.6085, "step": 17006 }, { "epoch": 2.03, "learning_rate": 1.2608717419299748e-06, "loss": 0.335, "step": 17007 }, { "epoch": 2.03, "learning_rate": 1.260592588973207e-06, "loss": 0.3189, "step": 17008 }, { "epoch": 2.03, "learning_rate": 1.2603134565039466e-06, "loss": 0.4457, "step": 17009 }, { "epoch": 2.03, "learning_rate": 1.2600343445268088e-06, "loss": 0.4185, "step": 17010 }, { "epoch": 2.03, "learning_rate": 1.2597552530464074e-06, "loss": 0.5887, "step": 17011 }, { "epoch": 2.03, "learning_rate": 1.2594761820673562e-06, "loss": 0.5824, "step": 17012 }, { "epoch": 2.03, "learning_rate": 1.2591971315942675e-06, "loss": 0.5447, "step": 17013 }, { "epoch": 2.03, "learning_rate": 1.2589181016317547e-06, "loss": 0.304, "step": 17014 }, { "epoch": 2.03, "learning_rate": 1.2586390921844294e-06, "loss": 0.4807, "step": 17015 }, { "epoch": 2.03, "learning_rate": 1.2583601032569048e-06, "loss": 0.4795, "step": 17016 }, { "epoch": 2.03, "learning_rate": 1.2580811348537908e-06, "loss": 0.5158, "step": 17017 }, { "epoch": 2.03, "learning_rate": 1.2578021869797e-06, "loss": 0.5652, "step": 17018 }, { "epoch": 2.03, "learning_rate": 1.2575232596392433e-06, "loss": 0.5395, "step": 17019 }, { "epoch": 2.03, "learning_rate": 1.257244352837032e-06, "loss": 0.2909, "step": 17020 }, { "epoch": 2.03, "learning_rate": 1.2569654665776761e-06, "loss": 0.3901, "step": 17021 }, { "epoch": 2.03, "learning_rate": 1.2566866008657846e-06, "loss": 0.4553, "step": 17022 }, { "epoch": 2.03, "learning_rate": 1.2564077557059683e-06, "loss": 0.547, "step": 17023 }, { "epoch": 2.03, "learning_rate": 1.2561289311028357e-06, "loss": 0.5761, "step": 17024 }, { "epoch": 2.03, "learning_rate": 1.255850127060996e-06, "loss": 0.5565, "step": 17025 }, { "epoch": 2.03, "learning_rate": 1.255571343585058e-06, "loss": 0.4609, "step": 17026 }, { "epoch": 2.03, "learning_rate": 1.2552925806796311e-06, "loss": 0.4121, "step": 17027 }, { "epoch": 2.03, "learning_rate": 1.2550138383493215e-06, "loss": 0.453, "step": 17028 }, { "epoch": 2.03, "learning_rate": 1.2547351165987387e-06, "loss": 0.4405, "step": 17029 }, { "epoch": 2.03, "learning_rate": 1.2544564154324878e-06, "loss": 0.5371, "step": 17030 }, { "epoch": 2.03, "learning_rate": 1.2541777348551775e-06, "loss": 0.59, "step": 17031 }, { "epoch": 2.03, "learning_rate": 1.2538990748714133e-06, "loss": 0.2611, "step": 17032 }, { "epoch": 2.03, "learning_rate": 1.2536204354858016e-06, "loss": 0.3835, "step": 17033 }, { "epoch": 2.03, "learning_rate": 1.253341816702949e-06, "loss": 0.3735, "step": 17034 }, { "epoch": 2.03, "learning_rate": 1.2530632185274615e-06, "loss": 0.4549, "step": 17035 }, { "epoch": 2.03, "learning_rate": 1.252784640963944e-06, "loss": 0.4389, "step": 17036 }, { "epoch": 2.03, "learning_rate": 1.2525060840169994e-06, "loss": 0.5819, "step": 17037 }, { "epoch": 2.03, "learning_rate": 1.2522275476912354e-06, "loss": 0.3957, "step": 17038 }, { "epoch": 2.03, "learning_rate": 1.2519490319912536e-06, "loss": 0.2911, "step": 17039 }, { "epoch": 2.03, "learning_rate": 1.251670536921659e-06, "loss": 0.3605, "step": 17040 }, { "epoch": 2.03, "learning_rate": 1.2513920624870552e-06, "loss": 0.4471, "step": 17041 }, { "epoch": 2.03, "learning_rate": 1.251113608692046e-06, "loss": 0.5958, "step": 17042 }, { "epoch": 2.03, "learning_rate": 1.2508351755412341e-06, "loss": 0.5089, "step": 17043 }, { "epoch": 2.03, "learning_rate": 1.2505567630392205e-06, "loss": 0.6563, "step": 17044 }, { "epoch": 2.03, "learning_rate": 1.2502783711906084e-06, "loss": 0.2822, "step": 17045 }, { "epoch": 2.03, "learning_rate": 1.2500000000000007e-06, "loss": 0.4158, "step": 17046 }, { "epoch": 2.03, "learning_rate": 1.2497216494719968e-06, "loss": 0.4346, "step": 17047 }, { "epoch": 2.03, "learning_rate": 1.249443319611199e-06, "loss": 0.4702, "step": 17048 }, { "epoch": 2.03, "learning_rate": 1.2491650104222088e-06, "loss": 0.5704, "step": 17049 }, { "epoch": 2.03, "learning_rate": 1.2488867219096262e-06, "loss": 0.7126, "step": 17050 }, { "epoch": 2.03, "learning_rate": 1.2486084540780502e-06, "loss": 0.3363, "step": 17051 }, { "epoch": 2.03, "learning_rate": 1.2483302069320813e-06, "loss": 0.4489, "step": 17052 }, { "epoch": 2.03, "learning_rate": 1.24805198047632e-06, "loss": 0.4524, "step": 17053 }, { "epoch": 2.03, "learning_rate": 1.2477737747153637e-06, "loss": 0.5003, "step": 17054 }, { "epoch": 2.03, "learning_rate": 1.247495589653812e-06, "loss": 0.5777, "step": 17055 }, { "epoch": 2.03, "learning_rate": 1.2472174252962632e-06, "loss": 0.6433, "step": 17056 }, { "epoch": 2.03, "learning_rate": 1.2469392816473163e-06, "loss": 0.3244, "step": 17057 }, { "epoch": 2.03, "learning_rate": 1.2466611587115684e-06, "loss": 0.4374, "step": 17058 }, { "epoch": 2.03, "learning_rate": 1.2463830564936159e-06, "loss": 0.4374, "step": 17059 }, { "epoch": 2.03, "learning_rate": 1.246104974998056e-06, "loss": 0.4577, "step": 17060 }, { "epoch": 2.03, "learning_rate": 1.2458269142294877e-06, "loss": 0.5077, "step": 17061 }, { "epoch": 2.03, "learning_rate": 1.2455488741925046e-06, "loss": 0.6479, "step": 17062 }, { "epoch": 2.03, "learning_rate": 1.2452708548917039e-06, "loss": 0.4229, "step": 17063 }, { "epoch": 2.03, "learning_rate": 1.244992856331682e-06, "loss": 0.3793, "step": 17064 }, { "epoch": 2.03, "learning_rate": 1.2447148785170337e-06, "loss": 0.5159, "step": 17065 }, { "epoch": 2.03, "learning_rate": 1.2444369214523528e-06, "loss": 0.5298, "step": 17066 }, { "epoch": 2.03, "learning_rate": 1.2441589851422351e-06, "loss": 0.4291, "step": 17067 }, { "epoch": 2.03, "learning_rate": 1.2438810695912753e-06, "loss": 0.6775, "step": 17068 }, { "epoch": 2.03, "learning_rate": 1.243603174804066e-06, "loss": 0.6494, "step": 17069 }, { "epoch": 2.03, "learning_rate": 1.2433253007852018e-06, "loss": 0.3141, "step": 17070 }, { "epoch": 2.03, "learning_rate": 1.2430474475392767e-06, "loss": 0.457, "step": 17071 }, { "epoch": 2.03, "learning_rate": 1.2427696150708829e-06, "loss": 0.4237, "step": 17072 }, { "epoch": 2.03, "learning_rate": 1.2424918033846118e-06, "loss": 0.5014, "step": 17073 }, { "epoch": 2.03, "learning_rate": 1.2422140124850565e-06, "loss": 0.5704, "step": 17074 }, { "epoch": 2.03, "learning_rate": 1.241936242376809e-06, "loss": 0.5941, "step": 17075 }, { "epoch": 2.03, "learning_rate": 1.2416584930644624e-06, "loss": 0.3438, "step": 17076 }, { "epoch": 2.03, "learning_rate": 1.2413807645526055e-06, "loss": 0.338, "step": 17077 }, { "epoch": 2.03, "learning_rate": 1.2411030568458307e-06, "loss": 0.3887, "step": 17078 }, { "epoch": 2.03, "learning_rate": 1.2408253699487274e-06, "loss": 0.4694, "step": 17079 }, { "epoch": 2.03, "learning_rate": 1.2405477038658872e-06, "loss": 0.5717, "step": 17080 }, { "epoch": 2.03, "learning_rate": 1.240270058601898e-06, "loss": 0.7163, "step": 17081 }, { "epoch": 2.03, "learning_rate": 1.239992434161351e-06, "loss": 0.3029, "step": 17082 }, { "epoch": 2.03, "learning_rate": 1.2397148305488343e-06, "loss": 0.3287, "step": 17083 }, { "epoch": 2.03, "learning_rate": 1.2394372477689385e-06, "loss": 0.3983, "step": 17084 }, { "epoch": 2.03, "learning_rate": 1.2391596858262494e-06, "loss": 0.4085, "step": 17085 }, { "epoch": 2.03, "learning_rate": 1.2388821447253577e-06, "loss": 0.6067, "step": 17086 }, { "epoch": 2.03, "learning_rate": 1.2386046244708499e-06, "loss": 0.5491, "step": 17087 }, { "epoch": 2.04, "learning_rate": 1.2383271250673125e-06, "loss": 0.3427, "step": 17088 }, { "epoch": 2.04, "learning_rate": 1.2380496465193336e-06, "loss": 0.3445, "step": 17089 }, { "epoch": 2.04, "learning_rate": 1.2377721888315e-06, "loss": 0.4446, "step": 17090 }, { "epoch": 2.04, "learning_rate": 1.237494752008399e-06, "loss": 0.4097, "step": 17091 }, { "epoch": 2.04, "learning_rate": 1.2372173360546148e-06, "loss": 0.3955, "step": 17092 }, { "epoch": 2.04, "learning_rate": 1.2369399409747346e-06, "loss": 0.5798, "step": 17093 }, { "epoch": 2.04, "learning_rate": 1.2366625667733427e-06, "loss": 0.5506, "step": 17094 }, { "epoch": 2.04, "learning_rate": 1.236385213455025e-06, "loss": 0.3815, "step": 17095 }, { "epoch": 2.04, "learning_rate": 1.2361078810243651e-06, "loss": 0.3519, "step": 17096 }, { "epoch": 2.04, "learning_rate": 1.235830569485948e-06, "loss": 0.4301, "step": 17097 }, { "epoch": 2.04, "learning_rate": 1.2355532788443576e-06, "loss": 0.48, "step": 17098 }, { "epoch": 2.04, "learning_rate": 1.2352760091041784e-06, "loss": 0.6179, "step": 17099 }, { "epoch": 2.04, "learning_rate": 1.234998760269993e-06, "loss": 0.6061, "step": 17100 }, { "epoch": 2.04, "learning_rate": 1.2347215323463833e-06, "loss": 0.2898, "step": 17101 }, { "epoch": 2.04, "learning_rate": 1.2344443253379337e-06, "loss": 0.3426, "step": 17102 }, { "epoch": 2.04, "learning_rate": 1.234167139249225e-06, "loss": 0.5162, "step": 17103 }, { "epoch": 2.04, "learning_rate": 1.2338899740848392e-06, "loss": 0.4709, "step": 17104 }, { "epoch": 2.04, "learning_rate": 1.2336128298493585e-06, "loss": 0.5107, "step": 17105 }, { "epoch": 2.04, "learning_rate": 1.233335706547365e-06, "loss": 0.7095, "step": 17106 }, { "epoch": 2.04, "learning_rate": 1.2330586041834386e-06, "loss": 0.3349, "step": 17107 }, { "epoch": 2.04, "learning_rate": 1.232781522762159e-06, "loss": 0.3963, "step": 17108 }, { "epoch": 2.04, "learning_rate": 1.232504462288107e-06, "loss": 0.3946, "step": 17109 }, { "epoch": 2.04, "learning_rate": 1.2322274227658635e-06, "loss": 0.3457, "step": 17110 }, { "epoch": 2.04, "learning_rate": 1.231950404200006e-06, "loss": 0.4412, "step": 17111 }, { "epoch": 2.04, "learning_rate": 1.2316734065951146e-06, "loss": 0.6225, "step": 17112 }, { "epoch": 2.04, "learning_rate": 1.2313964299557687e-06, "loss": 0.3108, "step": 17113 }, { "epoch": 2.04, "learning_rate": 1.231119474286547e-06, "loss": 0.3544, "step": 17114 }, { "epoch": 2.04, "learning_rate": 1.2308425395920267e-06, "loss": 0.4204, "step": 17115 }, { "epoch": 2.04, "learning_rate": 1.230565625876785e-06, "loss": 0.42, "step": 17116 }, { "epoch": 2.04, "learning_rate": 1.2302887331454008e-06, "loss": 0.5715, "step": 17117 }, { "epoch": 2.04, "learning_rate": 1.2300118614024495e-06, "loss": 0.5991, "step": 17118 }, { "epoch": 2.04, "learning_rate": 1.2297350106525087e-06, "loss": 0.6766, "step": 17119 }, { "epoch": 2.04, "learning_rate": 1.2294581809001546e-06, "loss": 0.3098, "step": 17120 }, { "epoch": 2.04, "learning_rate": 1.2291813721499646e-06, "loss": 0.4454, "step": 17121 }, { "epoch": 2.04, "learning_rate": 1.2289045844065128e-06, "loss": 0.4728, "step": 17122 }, { "epoch": 2.04, "learning_rate": 1.2286278176743744e-06, "loss": 0.4973, "step": 17123 }, { "epoch": 2.04, "learning_rate": 1.2283510719581243e-06, "loss": 0.59, "step": 17124 }, { "epoch": 2.04, "learning_rate": 1.2280743472623392e-06, "loss": 0.4996, "step": 17125 }, { "epoch": 2.04, "learning_rate": 1.2277976435915904e-06, "loss": 0.3261, "step": 17126 }, { "epoch": 2.04, "learning_rate": 1.2275209609504535e-06, "loss": 0.3823, "step": 17127 }, { "epoch": 2.04, "learning_rate": 1.227244299343503e-06, "loss": 0.4904, "step": 17128 }, { "epoch": 2.04, "learning_rate": 1.2269676587753107e-06, "loss": 0.5196, "step": 17129 }, { "epoch": 2.04, "learning_rate": 1.226691039250449e-06, "loss": 0.5645, "step": 17130 }, { "epoch": 2.04, "learning_rate": 1.2264144407734915e-06, "loss": 0.6442, "step": 17131 }, { "epoch": 2.04, "learning_rate": 1.2261378633490108e-06, "loss": 0.2839, "step": 17132 }, { "epoch": 2.04, "learning_rate": 1.2258613069815772e-06, "loss": 0.421, "step": 17133 }, { "epoch": 2.04, "learning_rate": 1.2255847716757632e-06, "loss": 0.4913, "step": 17134 }, { "epoch": 2.04, "learning_rate": 1.2253082574361408e-06, "loss": 0.4707, "step": 17135 }, { "epoch": 2.04, "learning_rate": 1.225031764267279e-06, "loss": 0.6375, "step": 17136 }, { "epoch": 2.04, "learning_rate": 1.22475529217375e-06, "loss": 0.6978, "step": 17137 }, { "epoch": 2.04, "learning_rate": 1.224478841160122e-06, "loss": 0.3176, "step": 17138 }, { "epoch": 2.04, "learning_rate": 1.2242024112309658e-06, "loss": 0.3322, "step": 17139 }, { "epoch": 2.04, "learning_rate": 1.2239260023908516e-06, "loss": 0.3521, "step": 17140 }, { "epoch": 2.04, "learning_rate": 1.2236496146443469e-06, "loss": 0.4932, "step": 17141 }, { "epoch": 2.04, "learning_rate": 1.2233732479960212e-06, "loss": 0.5282, "step": 17142 }, { "epoch": 2.04, "learning_rate": 1.2230969024504436e-06, "loss": 0.6092, "step": 17143 }, { "epoch": 2.04, "learning_rate": 1.2228205780121815e-06, "loss": 0.6379, "step": 17144 }, { "epoch": 2.04, "learning_rate": 1.2225442746858013e-06, "loss": 0.3608, "step": 17145 }, { "epoch": 2.04, "learning_rate": 1.2222679924758714e-06, "loss": 0.4553, "step": 17146 }, { "epoch": 2.04, "learning_rate": 1.22199173138696e-06, "loss": 0.4048, "step": 17147 }, { "epoch": 2.04, "learning_rate": 1.2217154914236315e-06, "loss": 0.4968, "step": 17148 }, { "epoch": 2.04, "learning_rate": 1.2214392725904531e-06, "loss": 0.6036, "step": 17149 }, { "epoch": 2.04, "learning_rate": 1.2211630748919917e-06, "loss": 0.592, "step": 17150 }, { "epoch": 2.04, "learning_rate": 1.220886898332812e-06, "loss": 0.3077, "step": 17151 }, { "epoch": 2.04, "learning_rate": 1.2206107429174786e-06, "loss": 0.295, "step": 17152 }, { "epoch": 2.04, "learning_rate": 1.2203346086505566e-06, "loss": 0.4172, "step": 17153 }, { "epoch": 2.04, "learning_rate": 1.2200584955366107e-06, "loss": 0.557, "step": 17154 }, { "epoch": 2.04, "learning_rate": 1.2197824035802067e-06, "loss": 0.5785, "step": 17155 }, { "epoch": 2.04, "learning_rate": 1.2195063327859058e-06, "loss": 0.6694, "step": 17156 }, { "epoch": 2.04, "learning_rate": 1.2192302831582736e-06, "loss": 0.3004, "step": 17157 }, { "epoch": 2.04, "learning_rate": 1.2189542547018712e-06, "loss": 0.3205, "step": 17158 }, { "epoch": 2.04, "learning_rate": 1.2186782474212636e-06, "loss": 0.4203, "step": 17159 }, { "epoch": 2.04, "learning_rate": 1.2184022613210112e-06, "loss": 0.3855, "step": 17160 }, { "epoch": 2.04, "learning_rate": 1.2181262964056766e-06, "loss": 0.5434, "step": 17161 }, { "epoch": 2.04, "learning_rate": 1.2178503526798222e-06, "loss": 0.6138, "step": 17162 }, { "epoch": 2.04, "learning_rate": 1.21757443014801e-06, "loss": 0.3349, "step": 17163 }, { "epoch": 2.04, "learning_rate": 1.2172985288148e-06, "loss": 0.402, "step": 17164 }, { "epoch": 2.04, "learning_rate": 1.217022648684752e-06, "loss": 0.4591, "step": 17165 }, { "epoch": 2.04, "learning_rate": 1.2167467897624283e-06, "loss": 0.4337, "step": 17166 }, { "epoch": 2.04, "learning_rate": 1.2164709520523869e-06, "loss": 0.4814, "step": 17167 }, { "epoch": 2.04, "learning_rate": 1.216195135559188e-06, "loss": 0.5298, "step": 17168 }, { "epoch": 2.04, "learning_rate": 1.2159193402873914e-06, "loss": 0.6448, "step": 17169 }, { "epoch": 2.04, "learning_rate": 1.215643566241557e-06, "loss": 0.2895, "step": 17170 }, { "epoch": 2.04, "learning_rate": 1.2153678134262412e-06, "loss": 0.3524, "step": 17171 }, { "epoch": 2.05, "learning_rate": 1.215092081846004e-06, "loss": 0.4674, "step": 17172 }, { "epoch": 2.05, "learning_rate": 1.2148163715054017e-06, "loss": 0.455, "step": 17173 }, { "epoch": 2.05, "learning_rate": 1.2145406824089936e-06, "loss": 0.5343, "step": 17174 }, { "epoch": 2.05, "learning_rate": 1.214265014561335e-06, "loss": 0.5372, "step": 17175 }, { "epoch": 2.05, "learning_rate": 1.2139893679669834e-06, "loss": 0.3332, "step": 17176 }, { "epoch": 2.05, "learning_rate": 1.2137137426304958e-06, "loss": 0.404, "step": 17177 }, { "epoch": 2.05, "learning_rate": 1.2134381385564284e-06, "loss": 0.3896, "step": 17178 }, { "epoch": 2.05, "learning_rate": 1.2131625557493367e-06, "loss": 0.4033, "step": 17179 }, { "epoch": 2.05, "learning_rate": 1.212886994213775e-06, "loss": 0.5518, "step": 17180 }, { "epoch": 2.05, "learning_rate": 1.2126114539543005e-06, "loss": 0.6039, "step": 17181 }, { "epoch": 2.05, "learning_rate": 1.2123359349754655e-06, "loss": 0.3354, "step": 17182 }, { "epoch": 2.05, "learning_rate": 1.2120604372818254e-06, "loss": 0.463, "step": 17183 }, { "epoch": 2.05, "learning_rate": 1.2117849608779345e-06, "loss": 0.4073, "step": 17184 }, { "epoch": 2.05, "learning_rate": 1.2115095057683474e-06, "loss": 0.502, "step": 17185 }, { "epoch": 2.05, "learning_rate": 1.211234071957616e-06, "loss": 0.6143, "step": 17186 }, { "epoch": 2.05, "learning_rate": 1.2109586594502927e-06, "loss": 0.5433, "step": 17187 }, { "epoch": 2.05, "learning_rate": 1.210683268250931e-06, "loss": 0.4207, "step": 17188 }, { "epoch": 2.05, "learning_rate": 1.2104078983640837e-06, "loss": 0.2575, "step": 17189 }, { "epoch": 2.05, "learning_rate": 1.2101325497943014e-06, "loss": 0.4594, "step": 17190 }, { "epoch": 2.05, "learning_rate": 1.209857222546136e-06, "loss": 0.4871, "step": 17191 }, { "epoch": 2.05, "learning_rate": 1.2095819166241401e-06, "loss": 0.5019, "step": 17192 }, { "epoch": 2.05, "learning_rate": 1.2093066320328625e-06, "loss": 0.5172, "step": 17193 }, { "epoch": 2.05, "learning_rate": 1.2090313687768557e-06, "loss": 0.4093, "step": 17194 }, { "epoch": 2.05, "learning_rate": 1.2087561268606673e-06, "loss": 0.2624, "step": 17195 }, { "epoch": 2.05, "learning_rate": 1.2084809062888498e-06, "loss": 0.3265, "step": 17196 }, { "epoch": 2.05, "learning_rate": 1.20820570706595e-06, "loss": 0.4054, "step": 17197 }, { "epoch": 2.05, "learning_rate": 1.2079305291965184e-06, "loss": 0.5937, "step": 17198 }, { "epoch": 2.05, "learning_rate": 1.2076553726851037e-06, "loss": 0.6243, "step": 17199 }, { "epoch": 2.05, "learning_rate": 1.2073802375362551e-06, "loss": 0.5183, "step": 17200 }, { "epoch": 2.05, "learning_rate": 1.2071051237545193e-06, "loss": 0.4536, "step": 17201 }, { "epoch": 2.05, "learning_rate": 1.2068300313444436e-06, "loss": 0.442, "step": 17202 }, { "epoch": 2.05, "learning_rate": 1.2065549603105758e-06, "loss": 0.4919, "step": 17203 }, { "epoch": 2.05, "learning_rate": 1.2062799106574642e-06, "loss": 0.481, "step": 17204 }, { "epoch": 2.05, "learning_rate": 1.206004882389653e-06, "loss": 0.5763, "step": 17205 }, { "epoch": 2.05, "learning_rate": 1.20572987551169e-06, "loss": 0.6672, "step": 17206 }, { "epoch": 2.05, "learning_rate": 1.2054548900281215e-06, "loss": 0.2956, "step": 17207 }, { "epoch": 2.05, "learning_rate": 1.2051799259434925e-06, "loss": 0.4497, "step": 17208 }, { "epoch": 2.05, "learning_rate": 1.204904983262347e-06, "loss": 0.4625, "step": 17209 }, { "epoch": 2.05, "learning_rate": 1.2046300619892305e-06, "loss": 0.423, "step": 17210 }, { "epoch": 2.05, "learning_rate": 1.204355162128689e-06, "loss": 0.6065, "step": 17211 }, { "epoch": 2.05, "learning_rate": 1.204080283685264e-06, "loss": 0.6512, "step": 17212 }, { "epoch": 2.05, "learning_rate": 1.203805426663501e-06, "loss": 0.4736, "step": 17213 }, { "epoch": 2.05, "learning_rate": 1.203530591067944e-06, "loss": 0.3357, "step": 17214 }, { "epoch": 2.05, "learning_rate": 1.203255776903135e-06, "loss": 0.3776, "step": 17215 }, { "epoch": 2.05, "learning_rate": 1.202980984173616e-06, "loss": 0.5063, "step": 17216 }, { "epoch": 2.05, "learning_rate": 1.2027062128839302e-06, "loss": 0.5074, "step": 17217 }, { "epoch": 2.05, "learning_rate": 1.2024314630386194e-06, "loss": 0.6776, "step": 17218 }, { "epoch": 2.05, "learning_rate": 1.2021567346422264e-06, "loss": 0.4775, "step": 17219 }, { "epoch": 2.05, "learning_rate": 1.2018820276992905e-06, "loss": 0.3904, "step": 17220 }, { "epoch": 2.05, "learning_rate": 1.2016073422143546e-06, "loss": 0.4142, "step": 17221 }, { "epoch": 2.05, "learning_rate": 1.2013326781919574e-06, "loss": 0.5449, "step": 17222 }, { "epoch": 2.05, "learning_rate": 1.2010580356366407e-06, "loss": 0.4598, "step": 17223 }, { "epoch": 2.05, "learning_rate": 1.2007834145529429e-06, "loss": 0.5752, "step": 17224 }, { "epoch": 2.05, "learning_rate": 1.2005088149454042e-06, "loss": 0.55, "step": 17225 }, { "epoch": 2.05, "learning_rate": 1.2002342368185638e-06, "loss": 0.2978, "step": 17226 }, { "epoch": 2.05, "learning_rate": 1.1999596801769617e-06, "loss": 0.4155, "step": 17227 }, { "epoch": 2.05, "learning_rate": 1.199685145025134e-06, "loss": 0.4116, "step": 17228 }, { "epoch": 2.05, "learning_rate": 1.199410631367621e-06, "loss": 0.4877, "step": 17229 }, { "epoch": 2.05, "learning_rate": 1.1991361392089592e-06, "loss": 0.5923, "step": 17230 }, { "epoch": 2.05, "learning_rate": 1.1988616685536853e-06, "loss": 0.6585, "step": 17231 }, { "epoch": 2.05, "learning_rate": 1.1985872194063371e-06, "loss": 0.276, "step": 17232 }, { "epoch": 2.05, "learning_rate": 1.1983127917714517e-06, "loss": 0.3347, "step": 17233 }, { "epoch": 2.05, "learning_rate": 1.198038385653566e-06, "loss": 0.4906, "step": 17234 }, { "epoch": 2.05, "learning_rate": 1.1977640010572139e-06, "loss": 0.4176, "step": 17235 }, { "epoch": 2.05, "learning_rate": 1.1974896379869331e-06, "loss": 0.53, "step": 17236 }, { "epoch": 2.05, "learning_rate": 1.1972152964472569e-06, "loss": 0.5927, "step": 17237 }, { "epoch": 2.05, "learning_rate": 1.196940976442722e-06, "loss": 0.3057, "step": 17238 }, { "epoch": 2.05, "learning_rate": 1.1966666779778613e-06, "loss": 0.4036, "step": 17239 }, { "epoch": 2.05, "learning_rate": 1.1963924010572098e-06, "loss": 0.4433, "step": 17240 }, { "epoch": 2.05, "learning_rate": 1.1961181456853013e-06, "loss": 0.464, "step": 17241 }, { "epoch": 2.05, "learning_rate": 1.1958439118666701e-06, "loss": 0.5133, "step": 17242 }, { "epoch": 2.05, "learning_rate": 1.1955696996058485e-06, "loss": 0.5573, "step": 17243 }, { "epoch": 2.05, "learning_rate": 1.1952955089073684e-06, "loss": 0.462, "step": 17244 }, { "epoch": 2.05, "learning_rate": 1.1950213397757638e-06, "loss": 0.4431, "step": 17245 }, { "epoch": 2.05, "learning_rate": 1.194747192215565e-06, "loss": 0.4182, "step": 17246 }, { "epoch": 2.05, "learning_rate": 1.1944730662313047e-06, "loss": 0.45, "step": 17247 }, { "epoch": 2.05, "learning_rate": 1.194198961827514e-06, "loss": 0.4985, "step": 17248 }, { "epoch": 2.05, "learning_rate": 1.1939248790087251e-06, "loss": 0.6306, "step": 17249 }, { "epoch": 2.05, "learning_rate": 1.1936508177794676e-06, "loss": 0.7224, "step": 17250 }, { "epoch": 2.05, "learning_rate": 1.1933767781442706e-06, "loss": 0.2543, "step": 17251 }, { "epoch": 2.05, "learning_rate": 1.1931027601076652e-06, "loss": 0.3522, "step": 17252 }, { "epoch": 2.05, "learning_rate": 1.192828763674182e-06, "loss": 0.3875, "step": 17253 }, { "epoch": 2.05, "learning_rate": 1.1925547888483477e-06, "loss": 0.4922, "step": 17254 }, { "epoch": 2.05, "learning_rate": 1.1922808356346926e-06, "loss": 0.6862, "step": 17255 }, { "epoch": 2.06, "learning_rate": 1.1920069040377449e-06, "loss": 0.5601, "step": 17256 }, { "epoch": 2.06, "learning_rate": 1.191732994062034e-06, "loss": 0.3595, "step": 17257 }, { "epoch": 2.06, "learning_rate": 1.1914591057120863e-06, "loss": 0.3492, "step": 17258 }, { "epoch": 2.06, "learning_rate": 1.1911852389924286e-06, "loss": 0.3925, "step": 17259 }, { "epoch": 2.06, "learning_rate": 1.1909113939075895e-06, "loss": 0.4051, "step": 17260 }, { "epoch": 2.06, "learning_rate": 1.190637570462094e-06, "loss": 0.5602, "step": 17261 }, { "epoch": 2.06, "learning_rate": 1.1903637686604694e-06, "loss": 0.5758, "step": 17262 }, { "epoch": 2.06, "learning_rate": 1.1900899885072414e-06, "loss": 0.4957, "step": 17263 }, { "epoch": 2.06, "learning_rate": 1.1898162300069368e-06, "loss": 0.379, "step": 17264 }, { "epoch": 2.06, "learning_rate": 1.1895424931640801e-06, "loss": 0.411, "step": 17265 }, { "epoch": 2.06, "learning_rate": 1.1892687779831947e-06, "loss": 0.4874, "step": 17266 }, { "epoch": 2.06, "learning_rate": 1.1889950844688064e-06, "loss": 0.556, "step": 17267 }, { "epoch": 2.06, "learning_rate": 1.1887214126254403e-06, "loss": 0.6396, "step": 17268 }, { "epoch": 2.06, "learning_rate": 1.1884477624576184e-06, "loss": 0.606, "step": 17269 }, { "epoch": 2.06, "learning_rate": 1.1881741339698652e-06, "loss": 0.4002, "step": 17270 }, { "epoch": 2.06, "learning_rate": 1.1879005271667045e-06, "loss": 0.4774, "step": 17271 }, { "epoch": 2.06, "learning_rate": 1.187626942052658e-06, "loss": 0.4334, "step": 17272 }, { "epoch": 2.06, "learning_rate": 1.1873533786322475e-06, "loss": 0.5033, "step": 17273 }, { "epoch": 2.06, "learning_rate": 1.1870798369099958e-06, "loss": 0.6126, "step": 17274 }, { "epoch": 2.06, "learning_rate": 1.1868063168904254e-06, "loss": 0.5933, "step": 17275 }, { "epoch": 2.06, "learning_rate": 1.1865328185780558e-06, "loss": 0.331, "step": 17276 }, { "epoch": 2.06, "learning_rate": 1.186259341977409e-06, "loss": 0.372, "step": 17277 }, { "epoch": 2.06, "learning_rate": 1.1859858870930067e-06, "loss": 0.4472, "step": 17278 }, { "epoch": 2.06, "learning_rate": 1.1857124539293666e-06, "loss": 0.5307, "step": 17279 }, { "epoch": 2.06, "learning_rate": 1.1854390424910109e-06, "loss": 0.5683, "step": 17280 }, { "epoch": 2.06, "learning_rate": 1.1851656527824573e-06, "loss": 0.5637, "step": 17281 }, { "epoch": 2.06, "learning_rate": 1.1848922848082255e-06, "loss": 0.3326, "step": 17282 }, { "epoch": 2.06, "learning_rate": 1.184618938572836e-06, "loss": 0.4481, "step": 17283 }, { "epoch": 2.06, "learning_rate": 1.1843456140808045e-06, "loss": 0.4452, "step": 17284 }, { "epoch": 2.06, "learning_rate": 1.1840723113366506e-06, "loss": 0.3988, "step": 17285 }, { "epoch": 2.06, "learning_rate": 1.1837990303448926e-06, "loss": 0.563, "step": 17286 }, { "epoch": 2.06, "learning_rate": 1.1835257711100473e-06, "loss": 0.6432, "step": 17287 }, { "epoch": 2.06, "learning_rate": 1.1832525336366304e-06, "loss": 0.5305, "step": 17288 }, { "epoch": 2.06, "learning_rate": 1.1829793179291595e-06, "loss": 0.3393, "step": 17289 }, { "epoch": 2.06, "learning_rate": 1.182706123992152e-06, "loss": 0.429, "step": 17290 }, { "epoch": 2.06, "learning_rate": 1.182432951830122e-06, "loss": 0.4082, "step": 17291 }, { "epoch": 2.06, "learning_rate": 1.1821598014475858e-06, "loss": 0.5711, "step": 17292 }, { "epoch": 2.06, "learning_rate": 1.1818866728490595e-06, "loss": 0.5929, "step": 17293 }, { "epoch": 2.06, "learning_rate": 1.181613566039057e-06, "loss": 0.4751, "step": 17294 }, { "epoch": 2.06, "learning_rate": 1.1813404810220921e-06, "loss": 0.2985, "step": 17295 }, { "epoch": 2.06, "learning_rate": 1.1810674178026796e-06, "loss": 0.444, "step": 17296 }, { "epoch": 2.06, "learning_rate": 1.1807943763853332e-06, "loss": 0.437, "step": 17297 }, { "epoch": 2.06, "learning_rate": 1.1805213567745677e-06, "loss": 0.4744, "step": 17298 }, { "epoch": 2.06, "learning_rate": 1.1802483589748939e-06, "loss": 0.5721, "step": 17299 }, { "epoch": 2.06, "learning_rate": 1.1799753829908265e-06, "loss": 0.6135, "step": 17300 }, { "epoch": 2.06, "learning_rate": 1.1797024288268755e-06, "loss": 0.3482, "step": 17301 }, { "epoch": 2.06, "learning_rate": 1.1794294964875553e-06, "loss": 0.4059, "step": 17302 }, { "epoch": 2.06, "learning_rate": 1.1791565859773755e-06, "loss": 0.4108, "step": 17303 }, { "epoch": 2.06, "learning_rate": 1.1788836973008478e-06, "loss": 0.5275, "step": 17304 }, { "epoch": 2.06, "learning_rate": 1.1786108304624835e-06, "loss": 0.6108, "step": 17305 }, { "epoch": 2.06, "learning_rate": 1.1783379854667942e-06, "loss": 0.5093, "step": 17306 }, { "epoch": 2.06, "learning_rate": 1.178065162318289e-06, "loss": 0.3113, "step": 17307 }, { "epoch": 2.06, "learning_rate": 1.1777923610214764e-06, "loss": 0.3734, "step": 17308 }, { "epoch": 2.06, "learning_rate": 1.177519581580868e-06, "loss": 0.442, "step": 17309 }, { "epoch": 2.06, "learning_rate": 1.177246824000971e-06, "loss": 0.4278, "step": 17310 }, { "epoch": 2.06, "learning_rate": 1.1769740882862947e-06, "loss": 0.4794, "step": 17311 }, { "epoch": 2.06, "learning_rate": 1.176701374441348e-06, "loss": 0.5209, "step": 17312 }, { "epoch": 2.06, "learning_rate": 1.1764286824706398e-06, "loss": 0.4878, "step": 17313 }, { "epoch": 2.06, "learning_rate": 1.1761560123786753e-06, "loss": 0.3719, "step": 17314 }, { "epoch": 2.06, "learning_rate": 1.1758833641699638e-06, "loss": 0.4395, "step": 17315 }, { "epoch": 2.06, "learning_rate": 1.1756107378490107e-06, "loss": 0.5206, "step": 17316 }, { "epoch": 2.06, "learning_rate": 1.175338133420324e-06, "loss": 0.4474, "step": 17317 }, { "epoch": 2.06, "learning_rate": 1.1750655508884081e-06, "loss": 0.4851, "step": 17318 }, { "epoch": 2.06, "learning_rate": 1.17479299025777e-06, "loss": 0.6085, "step": 17319 }, { "epoch": 2.06, "learning_rate": 1.1745204515329151e-06, "loss": 0.3085, "step": 17320 }, { "epoch": 2.06, "learning_rate": 1.174247934718349e-06, "loss": 0.3284, "step": 17321 }, { "epoch": 2.06, "learning_rate": 1.1739754398185759e-06, "loss": 0.5215, "step": 17322 }, { "epoch": 2.06, "learning_rate": 1.1737029668380991e-06, "loss": 0.5507, "step": 17323 }, { "epoch": 2.06, "learning_rate": 1.1734305157814246e-06, "loss": 0.5971, "step": 17324 }, { "epoch": 2.06, "learning_rate": 1.173158086653054e-06, "loss": 0.5759, "step": 17325 }, { "epoch": 2.06, "learning_rate": 1.1728856794574915e-06, "loss": 0.2534, "step": 17326 }, { "epoch": 2.06, "learning_rate": 1.1726132941992399e-06, "loss": 0.3451, "step": 17327 }, { "epoch": 2.06, "learning_rate": 1.1723409308828032e-06, "loss": 0.4711, "step": 17328 }, { "epoch": 2.06, "learning_rate": 1.172068589512682e-06, "loss": 0.446, "step": 17329 }, { "epoch": 2.06, "learning_rate": 1.1717962700933777e-06, "loss": 0.5313, "step": 17330 }, { "epoch": 2.06, "learning_rate": 1.1715239726293923e-06, "loss": 0.5419, "step": 17331 }, { "epoch": 2.06, "learning_rate": 1.1712516971252282e-06, "loss": 0.3596, "step": 17332 }, { "epoch": 2.06, "learning_rate": 1.1709794435853842e-06, "loss": 0.4301, "step": 17333 }, { "epoch": 2.06, "learning_rate": 1.1707072120143614e-06, "loss": 0.4969, "step": 17334 }, { "epoch": 2.06, "learning_rate": 1.1704350024166608e-06, "loss": 0.5062, "step": 17335 }, { "epoch": 2.06, "learning_rate": 1.1701628147967802e-06, "loss": 0.5238, "step": 17336 }, { "epoch": 2.06, "learning_rate": 1.169890649159221e-06, "loss": 0.5355, "step": 17337 }, { "epoch": 2.06, "learning_rate": 1.16961850550848e-06, "loss": 0.3923, "step": 17338 }, { "epoch": 2.06, "learning_rate": 1.1693463838490573e-06, "loss": 0.3761, "step": 17339 }, { "epoch": 2.07, "learning_rate": 1.1690742841854499e-06, "loss": 0.4506, "step": 17340 }, { "epoch": 2.07, "learning_rate": 1.1688022065221563e-06, "loss": 0.3933, "step": 17341 }, { "epoch": 2.07, "learning_rate": 1.1685301508636738e-06, "loss": 0.5191, "step": 17342 }, { "epoch": 2.07, "learning_rate": 1.1682581172145008e-06, "loss": 0.6451, "step": 17343 }, { "epoch": 2.07, "learning_rate": 1.1679861055791325e-06, "loss": 0.5844, "step": 17344 }, { "epoch": 2.07, "learning_rate": 1.167714115962065e-06, "loss": 0.2741, "step": 17345 }, { "epoch": 2.07, "learning_rate": 1.1674421483677947e-06, "loss": 0.4045, "step": 17346 }, { "epoch": 2.07, "learning_rate": 1.1671702028008185e-06, "loss": 0.4596, "step": 17347 }, { "epoch": 2.07, "learning_rate": 1.16689827926563e-06, "loss": 0.4602, "step": 17348 }, { "epoch": 2.07, "learning_rate": 1.1666263777667245e-06, "loss": 0.6096, "step": 17349 }, { "epoch": 2.07, "learning_rate": 1.166354498308598e-06, "loss": 0.5936, "step": 17350 }, { "epoch": 2.07, "learning_rate": 1.1660826408957431e-06, "loss": 0.3341, "step": 17351 }, { "epoch": 2.07, "learning_rate": 1.1658108055326535e-06, "loss": 0.3364, "step": 17352 }, { "epoch": 2.07, "learning_rate": 1.1655389922238233e-06, "loss": 0.3793, "step": 17353 }, { "epoch": 2.07, "learning_rate": 1.1652672009737464e-06, "loss": 0.4771, "step": 17354 }, { "epoch": 2.07, "learning_rate": 1.1649954317869138e-06, "loss": 0.5938, "step": 17355 }, { "epoch": 2.07, "learning_rate": 1.1647236846678185e-06, "loss": 0.5057, "step": 17356 }, { "epoch": 2.07, "learning_rate": 1.1644519596209539e-06, "loss": 0.3352, "step": 17357 }, { "epoch": 2.07, "learning_rate": 1.1641802566508103e-06, "loss": 0.3417, "step": 17358 }, { "epoch": 2.07, "learning_rate": 1.1639085757618785e-06, "loss": 0.3972, "step": 17359 }, { "epoch": 2.07, "learning_rate": 1.1636369169586497e-06, "loss": 0.4228, "step": 17360 }, { "epoch": 2.07, "learning_rate": 1.163365280245615e-06, "loss": 0.4623, "step": 17361 }, { "epoch": 2.07, "learning_rate": 1.163093665627265e-06, "loss": 0.68, "step": 17362 }, { "epoch": 2.07, "learning_rate": 1.1628220731080882e-06, "loss": 0.3202, "step": 17363 }, { "epoch": 2.07, "learning_rate": 1.1625505026925758e-06, "loss": 0.3803, "step": 17364 }, { "epoch": 2.07, "learning_rate": 1.1622789543852145e-06, "loss": 0.394, "step": 17365 }, { "epoch": 2.07, "learning_rate": 1.1620074281904955e-06, "loss": 0.4842, "step": 17366 }, { "epoch": 2.07, "learning_rate": 1.161735924112905e-06, "loss": 0.4922, "step": 17367 }, { "epoch": 2.07, "learning_rate": 1.1614644421569321e-06, "loss": 0.6114, "step": 17368 }, { "epoch": 2.07, "learning_rate": 1.161192982327064e-06, "loss": 0.6467, "step": 17369 }, { "epoch": 2.07, "learning_rate": 1.1609215446277895e-06, "loss": 0.2968, "step": 17370 }, { "epoch": 2.07, "learning_rate": 1.160650129063593e-06, "loss": 0.4247, "step": 17371 }, { "epoch": 2.07, "learning_rate": 1.1603787356389635e-06, "loss": 0.3947, "step": 17372 }, { "epoch": 2.07, "learning_rate": 1.1601073643583857e-06, "loss": 0.4091, "step": 17373 }, { "epoch": 2.07, "learning_rate": 1.1598360152263446e-06, "loss": 0.6359, "step": 17374 }, { "epoch": 2.07, "learning_rate": 1.1595646882473265e-06, "loss": 0.6347, "step": 17375 }, { "epoch": 2.07, "learning_rate": 1.159293383425817e-06, "loss": 0.2651, "step": 17376 }, { "epoch": 2.07, "learning_rate": 1.1590221007663011e-06, "loss": 0.4525, "step": 17377 }, { "epoch": 2.07, "learning_rate": 1.1587508402732613e-06, "loss": 0.3994, "step": 17378 }, { "epoch": 2.07, "learning_rate": 1.1584796019511837e-06, "loss": 0.4245, "step": 17379 }, { "epoch": 2.07, "learning_rate": 1.1582083858045498e-06, "loss": 0.5211, "step": 17380 }, { "epoch": 2.07, "learning_rate": 1.157937191837845e-06, "loss": 0.5446, "step": 17381 }, { "epoch": 2.07, "learning_rate": 1.15766602005555e-06, "loss": 0.3147, "step": 17382 }, { "epoch": 2.07, "learning_rate": 1.1573948704621485e-06, "loss": 0.4134, "step": 17383 }, { "epoch": 2.07, "learning_rate": 1.1571237430621222e-06, "loss": 0.3922, "step": 17384 }, { "epoch": 2.07, "learning_rate": 1.156852637859954e-06, "loss": 0.4016, "step": 17385 }, { "epoch": 2.07, "learning_rate": 1.1565815548601246e-06, "loss": 0.4642, "step": 17386 }, { "epoch": 2.07, "learning_rate": 1.1563104940671138e-06, "loss": 0.5903, "step": 17387 }, { "epoch": 2.07, "learning_rate": 1.156039455485404e-06, "loss": 0.3661, "step": 17388 }, { "epoch": 2.07, "learning_rate": 1.1557684391194742e-06, "loss": 0.3124, "step": 17389 }, { "epoch": 2.07, "learning_rate": 1.1554974449738046e-06, "loss": 0.3807, "step": 17390 }, { "epoch": 2.07, "learning_rate": 1.1552264730528754e-06, "loss": 0.4755, "step": 17391 }, { "epoch": 2.07, "learning_rate": 1.1549555233611662e-06, "loss": 0.5273, "step": 17392 }, { "epoch": 2.07, "learning_rate": 1.1546845959031538e-06, "loss": 0.5055, "step": 17393 }, { "epoch": 2.07, "learning_rate": 1.1544136906833193e-06, "loss": 0.5615, "step": 17394 }, { "epoch": 2.07, "learning_rate": 1.1541428077061382e-06, "loss": 0.3276, "step": 17395 }, { "epoch": 2.07, "learning_rate": 1.1538719469760902e-06, "loss": 0.4149, "step": 17396 }, { "epoch": 2.07, "learning_rate": 1.1536011084976514e-06, "loss": 0.4572, "step": 17397 }, { "epoch": 2.07, "learning_rate": 1.153330292275299e-06, "loss": 0.4648, "step": 17398 }, { "epoch": 2.07, "learning_rate": 1.1530594983135098e-06, "loss": 0.5455, "step": 17399 }, { "epoch": 2.07, "learning_rate": 1.152788726616761e-06, "loss": 0.599, "step": 17400 }, { "epoch": 2.07, "learning_rate": 1.152517977189528e-06, "loss": 0.2642, "step": 17401 }, { "epoch": 2.07, "learning_rate": 1.1522472500362846e-06, "loss": 0.3672, "step": 17402 }, { "epoch": 2.07, "learning_rate": 1.1519765451615083e-06, "loss": 0.4176, "step": 17403 }, { "epoch": 2.07, "learning_rate": 1.1517058625696718e-06, "loss": 0.4201, "step": 17404 }, { "epoch": 2.07, "learning_rate": 1.1514352022652506e-06, "loss": 0.5285, "step": 17405 }, { "epoch": 2.07, "learning_rate": 1.1511645642527185e-06, "loss": 0.5786, "step": 17406 }, { "epoch": 2.07, "learning_rate": 1.1508939485365506e-06, "loss": 0.3651, "step": 17407 }, { "epoch": 2.07, "learning_rate": 1.1506233551212186e-06, "loss": 0.3764, "step": 17408 }, { "epoch": 2.07, "learning_rate": 1.150352784011195e-06, "loss": 0.4413, "step": 17409 }, { "epoch": 2.07, "learning_rate": 1.1500822352109532e-06, "loss": 0.4294, "step": 17410 }, { "epoch": 2.07, "learning_rate": 1.1498117087249659e-06, "loss": 0.5352, "step": 17411 }, { "epoch": 2.07, "learning_rate": 1.1495412045577038e-06, "loss": 0.6637, "step": 17412 }, { "epoch": 2.07, "learning_rate": 1.1492707227136384e-06, "loss": 0.3728, "step": 17413 }, { "epoch": 2.07, "learning_rate": 1.1490002631972428e-06, "loss": 0.3334, "step": 17414 }, { "epoch": 2.07, "learning_rate": 1.148729826012986e-06, "loss": 0.4242, "step": 17415 }, { "epoch": 2.07, "learning_rate": 1.1484594111653374e-06, "loss": 0.4058, "step": 17416 }, { "epoch": 2.07, "learning_rate": 1.1481890186587683e-06, "loss": 0.5761, "step": 17417 }, { "epoch": 2.07, "learning_rate": 1.1479186484977491e-06, "loss": 0.5785, "step": 17418 }, { "epoch": 2.07, "learning_rate": 1.1476483006867473e-06, "loss": 0.5746, "step": 17419 }, { "epoch": 2.07, "learning_rate": 1.1473779752302325e-06, "loss": 0.3633, "step": 17420 }, { "epoch": 2.07, "learning_rate": 1.147107672132674e-06, "loss": 0.3966, "step": 17421 }, { "epoch": 2.07, "learning_rate": 1.1468373913985383e-06, "loss": 0.3968, "step": 17422 }, { "epoch": 2.07, "learning_rate": 1.146567133032295e-06, "loss": 0.4089, "step": 17423 }, { "epoch": 2.08, "learning_rate": 1.1462968970384098e-06, "loss": 0.5303, "step": 17424 }, { "epoch": 2.08, "learning_rate": 1.1460266834213504e-06, "loss": 0.5618, "step": 17425 }, { "epoch": 2.08, "learning_rate": 1.1457564921855843e-06, "loss": 0.2971, "step": 17426 }, { "epoch": 2.08, "learning_rate": 1.1454863233355762e-06, "loss": 0.3395, "step": 17427 }, { "epoch": 2.08, "learning_rate": 1.145216176875793e-06, "loss": 0.3829, "step": 17428 }, { "epoch": 2.08, "learning_rate": 1.144946052810701e-06, "loss": 0.5164, "step": 17429 }, { "epoch": 2.08, "learning_rate": 1.1446759511447642e-06, "loss": 0.6024, "step": 17430 }, { "epoch": 2.08, "learning_rate": 1.1444058718824466e-06, "loss": 0.6031, "step": 17431 }, { "epoch": 2.08, "learning_rate": 1.144135815028214e-06, "loss": 0.4316, "step": 17432 }, { "epoch": 2.08, "learning_rate": 1.143865780586531e-06, "loss": 0.3732, "step": 17433 }, { "epoch": 2.08, "learning_rate": 1.1435957685618595e-06, "loss": 0.3931, "step": 17434 }, { "epoch": 2.08, "learning_rate": 1.1433257789586639e-06, "loss": 0.4076, "step": 17435 }, { "epoch": 2.08, "learning_rate": 1.1430558117814076e-06, "loss": 0.5608, "step": 17436 }, { "epoch": 2.08, "learning_rate": 1.1427858670345526e-06, "loss": 0.5224, "step": 17437 }, { "epoch": 2.08, "learning_rate": 1.1425159447225603e-06, "loss": 0.396, "step": 17438 }, { "epoch": 2.08, "learning_rate": 1.1422460448498934e-06, "loss": 0.3156, "step": 17439 }, { "epoch": 2.08, "learning_rate": 1.1419761674210134e-06, "loss": 0.3662, "step": 17440 }, { "epoch": 2.08, "learning_rate": 1.141706312440382e-06, "loss": 0.3933, "step": 17441 }, { "epoch": 2.08, "learning_rate": 1.1414364799124583e-06, "loss": 0.6465, "step": 17442 }, { "epoch": 2.08, "learning_rate": 1.1411666698417048e-06, "loss": 0.6109, "step": 17443 }, { "epoch": 2.08, "learning_rate": 1.1408968822325792e-06, "loss": 0.4987, "step": 17444 }, { "epoch": 2.08, "learning_rate": 1.140627117089543e-06, "loss": 0.2814, "step": 17445 }, { "epoch": 2.08, "learning_rate": 1.1403573744170537e-06, "loss": 0.2929, "step": 17446 }, { "epoch": 2.08, "learning_rate": 1.140087654219571e-06, "loss": 0.3777, "step": 17447 }, { "epoch": 2.08, "learning_rate": 1.1398179565015536e-06, "loss": 0.4388, "step": 17448 }, { "epoch": 2.08, "learning_rate": 1.1395482812674606e-06, "loss": 0.5176, "step": 17449 }, { "epoch": 2.08, "learning_rate": 1.1392786285217475e-06, "loss": 0.6423, "step": 17450 }, { "epoch": 2.08, "learning_rate": 1.1390089982688737e-06, "loss": 0.3155, "step": 17451 }, { "epoch": 2.08, "learning_rate": 1.1387393905132954e-06, "loss": 0.4602, "step": 17452 }, { "epoch": 2.08, "learning_rate": 1.1384698052594685e-06, "loss": 0.3214, "step": 17453 }, { "epoch": 2.08, "learning_rate": 1.13820024251185e-06, "loss": 0.4477, "step": 17454 }, { "epoch": 2.08, "learning_rate": 1.1379307022748954e-06, "loss": 0.6246, "step": 17455 }, { "epoch": 2.08, "learning_rate": 1.1376611845530616e-06, "loss": 0.5932, "step": 17456 }, { "epoch": 2.08, "learning_rate": 1.137391689350802e-06, "loss": 0.4112, "step": 17457 }, { "epoch": 2.08, "learning_rate": 1.137122216672573e-06, "loss": 0.382, "step": 17458 }, { "epoch": 2.08, "learning_rate": 1.136852766522827e-06, "loss": 0.3465, "step": 17459 }, { "epoch": 2.08, "learning_rate": 1.1365833389060203e-06, "loss": 0.3816, "step": 17460 }, { "epoch": 2.08, "learning_rate": 1.136313933826604e-06, "loss": 0.5045, "step": 17461 }, { "epoch": 2.08, "learning_rate": 1.1360445512890334e-06, "loss": 0.6225, "step": 17462 }, { "epoch": 2.08, "learning_rate": 1.1357751912977606e-06, "loss": 0.3523, "step": 17463 }, { "epoch": 2.08, "learning_rate": 1.1355058538572392e-06, "loss": 0.4077, "step": 17464 }, { "epoch": 2.08, "learning_rate": 1.1352365389719205e-06, "loss": 0.4089, "step": 17465 }, { "epoch": 2.08, "learning_rate": 1.1349672466462556e-06, "loss": 0.431, "step": 17466 }, { "epoch": 2.08, "learning_rate": 1.134697976884697e-06, "loss": 0.5234, "step": 17467 }, { "epoch": 2.08, "learning_rate": 1.1344287296916954e-06, "loss": 0.5724, "step": 17468 }, { "epoch": 2.08, "learning_rate": 1.1341595050717007e-06, "loss": 0.4847, "step": 17469 }, { "epoch": 2.08, "learning_rate": 1.1338903030291643e-06, "loss": 0.3701, "step": 17470 }, { "epoch": 2.08, "learning_rate": 1.1336211235685369e-06, "loss": 0.4126, "step": 17471 }, { "epoch": 2.08, "learning_rate": 1.1333519666942666e-06, "loss": 0.4151, "step": 17472 }, { "epoch": 2.08, "learning_rate": 1.1330828324108023e-06, "loss": 0.5763, "step": 17473 }, { "epoch": 2.08, "learning_rate": 1.1328137207225934e-06, "loss": 0.6659, "step": 17474 }, { "epoch": 2.08, "learning_rate": 1.1325446316340893e-06, "loss": 0.452, "step": 17475 }, { "epoch": 2.08, "learning_rate": 1.1322755651497363e-06, "loss": 0.2616, "step": 17476 }, { "epoch": 2.08, "learning_rate": 1.1320065212739831e-06, "loss": 0.4275, "step": 17477 }, { "epoch": 2.08, "learning_rate": 1.1317375000112776e-06, "loss": 0.3767, "step": 17478 }, { "epoch": 2.08, "learning_rate": 1.1314685013660653e-06, "loss": 0.504, "step": 17479 }, { "epoch": 2.08, "learning_rate": 1.1311995253427944e-06, "loss": 0.6893, "step": 17480 }, { "epoch": 2.08, "learning_rate": 1.130930571945909e-06, "loss": 0.6293, "step": 17481 }, { "epoch": 2.08, "learning_rate": 1.1306616411798574e-06, "loss": 0.3689, "step": 17482 }, { "epoch": 2.08, "learning_rate": 1.1303927330490826e-06, "loss": 0.4336, "step": 17483 }, { "epoch": 2.08, "learning_rate": 1.1301238475580308e-06, "loss": 0.4767, "step": 17484 }, { "epoch": 2.08, "learning_rate": 1.129854984711147e-06, "loss": 0.4928, "step": 17485 }, { "epoch": 2.08, "learning_rate": 1.1295861445128762e-06, "loss": 0.5409, "step": 17486 }, { "epoch": 2.08, "learning_rate": 1.1293173269676611e-06, "loss": 0.6313, "step": 17487 }, { "epoch": 2.08, "learning_rate": 1.1290485320799449e-06, "loss": 0.4363, "step": 17488 }, { "epoch": 2.08, "learning_rate": 1.1287797598541714e-06, "loss": 0.3494, "step": 17489 }, { "epoch": 2.08, "learning_rate": 1.1285110102947844e-06, "loss": 0.3738, "step": 17490 }, { "epoch": 2.08, "learning_rate": 1.1282422834062247e-06, "loss": 0.4858, "step": 17491 }, { "epoch": 2.08, "learning_rate": 1.127973579192935e-06, "loss": 0.502, "step": 17492 }, { "epoch": 2.08, "learning_rate": 1.1277048976593583e-06, "loss": 0.6047, "step": 17493 }, { "epoch": 2.08, "learning_rate": 1.1274362388099348e-06, "loss": 0.4496, "step": 17494 }, { "epoch": 2.08, "learning_rate": 1.1271676026491044e-06, "loss": 0.3249, "step": 17495 }, { "epoch": 2.08, "learning_rate": 1.1268989891813085e-06, "loss": 0.3831, "step": 17496 }, { "epoch": 2.08, "learning_rate": 1.1266303984109885e-06, "loss": 0.4811, "step": 17497 }, { "epoch": 2.08, "learning_rate": 1.1263618303425826e-06, "loss": 0.5054, "step": 17498 }, { "epoch": 2.08, "learning_rate": 1.126093284980531e-06, "loss": 0.7029, "step": 17499 }, { "epoch": 2.08, "learning_rate": 1.1258247623292732e-06, "loss": 0.5717, "step": 17500 }, { "epoch": 2.08, "learning_rate": 1.1255562623932476e-06, "loss": 0.284, "step": 17501 }, { "epoch": 2.08, "learning_rate": 1.1252877851768913e-06, "loss": 0.5071, "step": 17502 }, { "epoch": 2.08, "learning_rate": 1.1250193306846432e-06, "loss": 0.4495, "step": 17503 }, { "epoch": 2.08, "learning_rate": 1.124750898920941e-06, "loss": 0.4856, "step": 17504 }, { "epoch": 2.08, "learning_rate": 1.1244824898902228e-06, "loss": 0.5799, "step": 17505 }, { "epoch": 2.08, "learning_rate": 1.1242141035969236e-06, "loss": 0.6853, "step": 17506 }, { "epoch": 2.08, "learning_rate": 1.1239457400454817e-06, "loss": 0.3134, "step": 17507 }, { "epoch": 2.09, "learning_rate": 1.123677399240331e-06, "loss": 0.4243, "step": 17508 }, { "epoch": 2.09, "learning_rate": 1.1234090811859095e-06, "loss": 0.4491, "step": 17509 }, { "epoch": 2.09, "learning_rate": 1.1231407858866503e-06, "loss": 0.3843, "step": 17510 }, { "epoch": 2.09, "learning_rate": 1.1228725133469896e-06, "loss": 0.4853, "step": 17511 }, { "epoch": 2.09, "learning_rate": 1.1226042635713616e-06, "loss": 0.5775, "step": 17512 }, { "epoch": 2.09, "learning_rate": 1.122336036564202e-06, "loss": 0.6088, "step": 17513 }, { "epoch": 2.09, "learning_rate": 1.122067832329942e-06, "loss": 0.3142, "step": 17514 }, { "epoch": 2.09, "learning_rate": 1.1217996508730172e-06, "loss": 0.3443, "step": 17515 }, { "epoch": 2.09, "learning_rate": 1.1215314921978602e-06, "loss": 0.3791, "step": 17516 }, { "epoch": 2.09, "learning_rate": 1.1212633563089021e-06, "loss": 0.4414, "step": 17517 }, { "epoch": 2.09, "learning_rate": 1.1209952432105764e-06, "loss": 0.6208, "step": 17518 }, { "epoch": 2.09, "learning_rate": 1.1207271529073152e-06, "loss": 0.5102, "step": 17519 }, { "epoch": 2.09, "learning_rate": 1.1204590854035507e-06, "loss": 0.3926, "step": 17520 }, { "epoch": 2.09, "learning_rate": 1.1201910407037126e-06, "loss": 0.4841, "step": 17521 }, { "epoch": 2.09, "learning_rate": 1.119923018812233e-06, "loss": 0.4734, "step": 17522 }, { "epoch": 2.09, "learning_rate": 1.119655019733541e-06, "loss": 0.515, "step": 17523 }, { "epoch": 2.09, "learning_rate": 1.119387043472068e-06, "loss": 0.6386, "step": 17524 }, { "epoch": 2.09, "learning_rate": 1.1191190900322421e-06, "loss": 0.6594, "step": 17525 }, { "epoch": 2.09, "learning_rate": 1.1188511594184937e-06, "loss": 0.2887, "step": 17526 }, { "epoch": 2.09, "learning_rate": 1.1185832516352515e-06, "loss": 0.4051, "step": 17527 }, { "epoch": 2.09, "learning_rate": 1.1183153666869448e-06, "loss": 0.4449, "step": 17528 }, { "epoch": 2.09, "learning_rate": 1.1180475045780012e-06, "loss": 0.4536, "step": 17529 }, { "epoch": 2.09, "learning_rate": 1.1177796653128475e-06, "loss": 0.5922, "step": 17530 }, { "epoch": 2.09, "learning_rate": 1.1175118488959126e-06, "loss": 0.5195, "step": 17531 }, { "epoch": 2.09, "learning_rate": 1.117244055331622e-06, "loss": 0.3782, "step": 17532 }, { "epoch": 2.09, "learning_rate": 1.116976284624403e-06, "loss": 0.3513, "step": 17533 }, { "epoch": 2.09, "learning_rate": 1.1167085367786823e-06, "loss": 0.411, "step": 17534 }, { "epoch": 2.09, "learning_rate": 1.1164408117988865e-06, "loss": 0.4624, "step": 17535 }, { "epoch": 2.09, "learning_rate": 1.116173109689439e-06, "loss": 0.5725, "step": 17536 }, { "epoch": 2.09, "learning_rate": 1.1159054304547673e-06, "loss": 0.5383, "step": 17537 }, { "epoch": 2.09, "learning_rate": 1.115637774099294e-06, "loss": 0.4166, "step": 17538 }, { "epoch": 2.09, "learning_rate": 1.1153701406274452e-06, "loss": 0.4046, "step": 17539 }, { "epoch": 2.09, "learning_rate": 1.1151025300436436e-06, "loss": 0.3591, "step": 17540 }, { "epoch": 2.09, "learning_rate": 1.114834942352313e-06, "loss": 0.4694, "step": 17541 }, { "epoch": 2.09, "learning_rate": 1.114567377557877e-06, "loss": 0.4729, "step": 17542 }, { "epoch": 2.09, "learning_rate": 1.1142998356647597e-06, "loss": 0.579, "step": 17543 }, { "epoch": 2.09, "learning_rate": 1.1140323166773822e-06, "loss": 0.5471, "step": 17544 }, { "epoch": 2.09, "learning_rate": 1.1137648206001662e-06, "loss": 0.3488, "step": 17545 }, { "epoch": 2.09, "learning_rate": 1.1134973474375347e-06, "loss": 0.424, "step": 17546 }, { "epoch": 2.09, "learning_rate": 1.1132298971939074e-06, "loss": 0.504, "step": 17547 }, { "epoch": 2.09, "learning_rate": 1.1129624698737063e-06, "loss": 0.3935, "step": 17548 }, { "epoch": 2.09, "learning_rate": 1.1126950654813518e-06, "loss": 0.6025, "step": 17549 }, { "epoch": 2.09, "learning_rate": 1.1124276840212654e-06, "loss": 0.5033, "step": 17550 }, { "epoch": 2.09, "learning_rate": 1.1121603254978657e-06, "loss": 0.3129, "step": 17551 }, { "epoch": 2.09, "learning_rate": 1.1118929899155714e-06, "loss": 0.3855, "step": 17552 }, { "epoch": 2.09, "learning_rate": 1.1116256772788025e-06, "loss": 0.3836, "step": 17553 }, { "epoch": 2.09, "learning_rate": 1.1113583875919786e-06, "loss": 0.5859, "step": 17554 }, { "epoch": 2.09, "learning_rate": 1.111091120859516e-06, "loss": 0.597, "step": 17555 }, { "epoch": 2.09, "learning_rate": 1.110823877085834e-06, "loss": 0.593, "step": 17556 }, { "epoch": 2.09, "learning_rate": 1.1105566562753506e-06, "loss": 0.2318, "step": 17557 }, { "epoch": 2.09, "learning_rate": 1.1102894584324822e-06, "loss": 0.2973, "step": 17558 }, { "epoch": 2.09, "learning_rate": 1.1100222835616447e-06, "loss": 0.3558, "step": 17559 }, { "epoch": 2.09, "learning_rate": 1.1097551316672556e-06, "loss": 0.4688, "step": 17560 }, { "epoch": 2.09, "learning_rate": 1.1094880027537317e-06, "loss": 0.4837, "step": 17561 }, { "epoch": 2.09, "learning_rate": 1.1092208968254869e-06, "loss": 0.6122, "step": 17562 }, { "epoch": 2.09, "learning_rate": 1.1089538138869375e-06, "loss": 0.3536, "step": 17563 }, { "epoch": 2.09, "learning_rate": 1.1086867539424988e-06, "loss": 0.3024, "step": 17564 }, { "epoch": 2.09, "learning_rate": 1.108419716996584e-06, "loss": 0.5014, "step": 17565 }, { "epoch": 2.09, "learning_rate": 1.1081527030536092e-06, "loss": 0.5532, "step": 17566 }, { "epoch": 2.09, "learning_rate": 1.1078857121179858e-06, "loss": 0.4465, "step": 17567 }, { "epoch": 2.09, "learning_rate": 1.107618744194129e-06, "loss": 0.58, "step": 17568 }, { "epoch": 2.09, "learning_rate": 1.1073517992864512e-06, "loss": 0.5724, "step": 17569 }, { "epoch": 2.09, "learning_rate": 1.1070848773993649e-06, "loss": 0.3045, "step": 17570 }, { "epoch": 2.09, "learning_rate": 1.106817978537282e-06, "loss": 0.4421, "step": 17571 }, { "epoch": 2.09, "learning_rate": 1.1065511027046158e-06, "loss": 0.4863, "step": 17572 }, { "epoch": 2.09, "learning_rate": 1.1062842499057768e-06, "loss": 0.4572, "step": 17573 }, { "epoch": 2.09, "learning_rate": 1.1060174201451755e-06, "loss": 0.6673, "step": 17574 }, { "epoch": 2.09, "learning_rate": 1.1057506134272228e-06, "loss": 0.6178, "step": 17575 }, { "epoch": 2.09, "learning_rate": 1.1054838297563303e-06, "loss": 0.3555, "step": 17576 }, { "epoch": 2.09, "learning_rate": 1.1052170691369064e-06, "loss": 0.3816, "step": 17577 }, { "epoch": 2.09, "learning_rate": 1.1049503315733614e-06, "loss": 0.4198, "step": 17578 }, { "epoch": 2.09, "learning_rate": 1.1046836170701053e-06, "loss": 0.5286, "step": 17579 }, { "epoch": 2.09, "learning_rate": 1.1044169256315463e-06, "loss": 0.5119, "step": 17580 }, { "epoch": 2.09, "learning_rate": 1.1041502572620913e-06, "loss": 0.5581, "step": 17581 }, { "epoch": 2.09, "learning_rate": 1.10388361196615e-06, "loss": 0.436, "step": 17582 }, { "epoch": 2.09, "learning_rate": 1.1036169897481298e-06, "loss": 0.3755, "step": 17583 }, { "epoch": 2.09, "learning_rate": 1.1033503906124384e-06, "loss": 0.4056, "step": 17584 }, { "epoch": 2.09, "learning_rate": 1.1030838145634817e-06, "loss": 0.5098, "step": 17585 }, { "epoch": 2.09, "learning_rate": 1.1028172616056676e-06, "loss": 0.529, "step": 17586 }, { "epoch": 2.09, "learning_rate": 1.1025507317434003e-06, "loss": 0.6477, "step": 17587 }, { "epoch": 2.09, "learning_rate": 1.1022842249810877e-06, "loss": 0.3941, "step": 17588 }, { "epoch": 2.09, "learning_rate": 1.1020177413231334e-06, "loss": 0.4043, "step": 17589 }, { "epoch": 2.09, "learning_rate": 1.1017512807739428e-06, "loss": 0.4451, "step": 17590 }, { "epoch": 2.09, "learning_rate": 1.101484843337921e-06, "loss": 0.41, "step": 17591 }, { "epoch": 2.1, "learning_rate": 1.101218429019473e-06, "loss": 0.5472, "step": 17592 }, { "epoch": 2.1, "learning_rate": 1.100952037823001e-06, "loss": 0.6074, "step": 17593 }, { "epoch": 2.1, "learning_rate": 1.1006856697529101e-06, "loss": 0.5266, "step": 17594 }, { "epoch": 2.1, "learning_rate": 1.1004193248136021e-06, "loss": 0.3466, "step": 17595 }, { "epoch": 2.1, "learning_rate": 1.1001530030094795e-06, "loss": 0.4008, "step": 17596 }, { "epoch": 2.1, "learning_rate": 1.0998867043449452e-06, "loss": 0.3903, "step": 17597 }, { "epoch": 2.1, "learning_rate": 1.099620428824401e-06, "loss": 0.5314, "step": 17598 }, { "epoch": 2.1, "learning_rate": 1.0993541764522498e-06, "loss": 0.5983, "step": 17599 }, { "epoch": 2.1, "learning_rate": 1.0990879472328904e-06, "loss": 0.4813, "step": 17600 }, { "epoch": 2.1, "learning_rate": 1.0988217411707258e-06, "loss": 0.2726, "step": 17601 }, { "epoch": 2.1, "learning_rate": 1.0985555582701547e-06, "loss": 0.3594, "step": 17602 }, { "epoch": 2.1, "learning_rate": 1.0982893985355783e-06, "loss": 0.434, "step": 17603 }, { "epoch": 2.1, "learning_rate": 1.098023261971395e-06, "loss": 0.5415, "step": 17604 }, { "epoch": 2.1, "learning_rate": 1.097757148582005e-06, "loss": 0.62, "step": 17605 }, { "epoch": 2.1, "learning_rate": 1.0974910583718068e-06, "loss": 0.5932, "step": 17606 }, { "epoch": 2.1, "learning_rate": 1.0972249913452002e-06, "loss": 0.1948, "step": 17607 }, { "epoch": 2.1, "learning_rate": 1.096958947506582e-06, "loss": 0.4214, "step": 17608 }, { "epoch": 2.1, "learning_rate": 1.0966929268603493e-06, "loss": 0.4401, "step": 17609 }, { "epoch": 2.1, "learning_rate": 1.0964269294109011e-06, "loss": 0.4223, "step": 17610 }, { "epoch": 2.1, "learning_rate": 1.0961609551626326e-06, "loss": 0.5607, "step": 17611 }, { "epoch": 2.1, "learning_rate": 1.0958950041199414e-06, "loss": 0.6219, "step": 17612 }, { "epoch": 2.1, "learning_rate": 1.0956290762872235e-06, "loss": 0.4233, "step": 17613 }, { "epoch": 2.1, "learning_rate": 1.095363171668876e-06, "loss": 0.3573, "step": 17614 }, { "epoch": 2.1, "learning_rate": 1.0950972902692928e-06, "loss": 0.3469, "step": 17615 }, { "epoch": 2.1, "learning_rate": 1.0948314320928682e-06, "loss": 0.5218, "step": 17616 }, { "epoch": 2.1, "learning_rate": 1.0945655971439983e-06, "loss": 0.552, "step": 17617 }, { "epoch": 2.1, "learning_rate": 1.0942997854270778e-06, "loss": 0.5808, "step": 17618 }, { "epoch": 2.1, "learning_rate": 1.0940339969464989e-06, "loss": 0.4583, "step": 17619 }, { "epoch": 2.1, "learning_rate": 1.0937682317066562e-06, "loss": 0.322, "step": 17620 }, { "epoch": 2.1, "learning_rate": 1.093502489711943e-06, "loss": 0.3804, "step": 17621 }, { "epoch": 2.1, "learning_rate": 1.0932367709667513e-06, "loss": 0.4731, "step": 17622 }, { "epoch": 2.1, "learning_rate": 1.0929710754754743e-06, "loss": 0.5354, "step": 17623 }, { "epoch": 2.1, "learning_rate": 1.092705403242503e-06, "loss": 0.6489, "step": 17624 }, { "epoch": 2.1, "learning_rate": 1.0924397542722302e-06, "loss": 0.5789, "step": 17625 }, { "epoch": 2.1, "learning_rate": 1.092174128569045e-06, "loss": 0.3197, "step": 17626 }, { "epoch": 2.1, "learning_rate": 1.0919085261373401e-06, "loss": 0.4371, "step": 17627 }, { "epoch": 2.1, "learning_rate": 1.0916429469815053e-06, "loss": 0.4295, "step": 17628 }, { "epoch": 2.1, "learning_rate": 1.0913773911059317e-06, "loss": 0.5028, "step": 17629 }, { "epoch": 2.1, "learning_rate": 1.091111858515008e-06, "loss": 0.6119, "step": 17630 }, { "epoch": 2.1, "learning_rate": 1.0908463492131227e-06, "loss": 0.5004, "step": 17631 }, { "epoch": 2.1, "learning_rate": 1.0905808632046653e-06, "loss": 0.2841, "step": 17632 }, { "epoch": 2.1, "learning_rate": 1.0903154004940255e-06, "loss": 0.3257, "step": 17633 }, { "epoch": 2.1, "learning_rate": 1.0900499610855895e-06, "loss": 0.51, "step": 17634 }, { "epoch": 2.1, "learning_rate": 1.0897845449837457e-06, "loss": 0.4129, "step": 17635 }, { "epoch": 2.1, "learning_rate": 1.089519152192883e-06, "loss": 0.6199, "step": 17636 }, { "epoch": 2.1, "learning_rate": 1.0892537827173869e-06, "loss": 0.5096, "step": 17637 }, { "epoch": 2.1, "learning_rate": 1.0889884365616433e-06, "loss": 0.4229, "step": 17638 }, { "epoch": 2.1, "learning_rate": 1.0887231137300388e-06, "loss": 0.2796, "step": 17639 }, { "epoch": 2.1, "learning_rate": 1.0884578142269611e-06, "loss": 0.464, "step": 17640 }, { "epoch": 2.1, "learning_rate": 1.0881925380567928e-06, "loss": 0.4249, "step": 17641 }, { "epoch": 2.1, "learning_rate": 1.0879272852239205e-06, "loss": 0.5401, "step": 17642 }, { "epoch": 2.1, "learning_rate": 1.08766205573273e-06, "loss": 0.6693, "step": 17643 }, { "epoch": 2.1, "learning_rate": 1.0873968495876036e-06, "loss": 0.445, "step": 17644 }, { "epoch": 2.1, "learning_rate": 1.0871316667929251e-06, "loss": 0.3589, "step": 17645 }, { "epoch": 2.1, "learning_rate": 1.086866507353079e-06, "loss": 0.3484, "step": 17646 }, { "epoch": 2.1, "learning_rate": 1.0866013712724476e-06, "loss": 0.3854, "step": 17647 }, { "epoch": 2.1, "learning_rate": 1.0863362585554155e-06, "loss": 0.4335, "step": 17648 }, { "epoch": 2.1, "learning_rate": 1.0860711692063624e-06, "loss": 0.6733, "step": 17649 }, { "epoch": 2.1, "learning_rate": 1.0858061032296718e-06, "loss": 0.594, "step": 17650 }, { "epoch": 2.1, "learning_rate": 1.0855410606297257e-06, "loss": 0.2523, "step": 17651 }, { "epoch": 2.1, "learning_rate": 1.0852760414109046e-06, "loss": 0.4296, "step": 17652 }, { "epoch": 2.1, "learning_rate": 1.0850110455775883e-06, "loss": 0.3474, "step": 17653 }, { "epoch": 2.1, "learning_rate": 1.084746073134158e-06, "loss": 0.4185, "step": 17654 }, { "epoch": 2.1, "learning_rate": 1.0844811240849941e-06, "loss": 0.5575, "step": 17655 }, { "epoch": 2.1, "learning_rate": 1.0842161984344767e-06, "loss": 0.6336, "step": 17656 }, { "epoch": 2.1, "learning_rate": 1.0839512961869839e-06, "loss": 0.3333, "step": 17657 }, { "epoch": 2.1, "learning_rate": 1.0836864173468952e-06, "loss": 0.3125, "step": 17658 }, { "epoch": 2.1, "learning_rate": 1.0834215619185895e-06, "loss": 0.3964, "step": 17659 }, { "epoch": 2.1, "learning_rate": 1.083156729906443e-06, "loss": 0.3996, "step": 17660 }, { "epoch": 2.1, "learning_rate": 1.0828919213148346e-06, "loss": 0.5676, "step": 17661 }, { "epoch": 2.1, "learning_rate": 1.0826271361481416e-06, "loss": 0.6118, "step": 17662 }, { "epoch": 2.1, "learning_rate": 1.0823623744107423e-06, "loss": 0.355, "step": 17663 }, { "epoch": 2.1, "learning_rate": 1.0820976361070106e-06, "loss": 0.3638, "step": 17664 }, { "epoch": 2.1, "learning_rate": 1.0818329212413248e-06, "loss": 0.4474, "step": 17665 }, { "epoch": 2.1, "learning_rate": 1.0815682298180591e-06, "loss": 0.385, "step": 17666 }, { "epoch": 2.1, "learning_rate": 1.0813035618415904e-06, "loss": 0.4978, "step": 17667 }, { "epoch": 2.1, "learning_rate": 1.0810389173162922e-06, "loss": 0.5702, "step": 17668 }, { "epoch": 2.1, "learning_rate": 1.0807742962465396e-06, "loss": 0.5167, "step": 17669 }, { "epoch": 2.1, "learning_rate": 1.0805096986367067e-06, "loss": 0.4113, "step": 17670 }, { "epoch": 2.1, "learning_rate": 1.0802451244911688e-06, "loss": 0.3661, "step": 17671 }, { "epoch": 2.1, "learning_rate": 1.0799805738142979e-06, "loss": 0.3695, "step": 17672 }, { "epoch": 2.1, "learning_rate": 1.0797160466104664e-06, "loss": 0.4145, "step": 17673 }, { "epoch": 2.1, "learning_rate": 1.079451542884049e-06, "loss": 0.6432, "step": 17674 }, { "epoch": 2.1, "learning_rate": 1.0791870626394155e-06, "loss": 0.5569, "step": 17675 }, { "epoch": 2.11, "learning_rate": 1.0789226058809394e-06, "loss": 0.3049, "step": 17676 }, { "epoch": 2.11, "learning_rate": 1.0786581726129916e-06, "loss": 0.3123, "step": 17677 }, { "epoch": 2.11, "learning_rate": 1.0783937628399447e-06, "loss": 0.3923, "step": 17678 }, { "epoch": 2.11, "learning_rate": 1.078129376566167e-06, "loss": 0.5641, "step": 17679 }, { "epoch": 2.11, "learning_rate": 1.0778650137960309e-06, "loss": 0.6427, "step": 17680 }, { "epoch": 2.11, "learning_rate": 1.0776006745339045e-06, "loss": 0.5143, "step": 17681 }, { "epoch": 2.11, "learning_rate": 1.0773363587841596e-06, "loss": 0.3478, "step": 17682 }, { "epoch": 2.11, "learning_rate": 1.0770720665511628e-06, "loss": 0.3825, "step": 17683 }, { "epoch": 2.11, "learning_rate": 1.0768077978392843e-06, "loss": 0.4937, "step": 17684 }, { "epoch": 2.11, "learning_rate": 1.0765435526528921e-06, "loss": 0.5313, "step": 17685 }, { "epoch": 2.11, "learning_rate": 1.0762793309963555e-06, "loss": 0.4505, "step": 17686 }, { "epoch": 2.11, "learning_rate": 1.076015132874041e-06, "loss": 0.6127, "step": 17687 }, { "epoch": 2.11, "learning_rate": 1.075750958290315e-06, "loss": 0.5078, "step": 17688 }, { "epoch": 2.11, "learning_rate": 1.0754868072495459e-06, "loss": 0.3957, "step": 17689 }, { "epoch": 2.11, "learning_rate": 1.0752226797560986e-06, "loss": 0.4658, "step": 17690 }, { "epoch": 2.11, "learning_rate": 1.0749585758143399e-06, "loss": 0.4242, "step": 17691 }, { "epoch": 2.11, "learning_rate": 1.0746944954286356e-06, "loss": 0.515, "step": 17692 }, { "epoch": 2.11, "learning_rate": 1.0744304386033516e-06, "loss": 0.685, "step": 17693 }, { "epoch": 2.11, "learning_rate": 1.074166405342852e-06, "loss": 0.6426, "step": 17694 }, { "epoch": 2.11, "learning_rate": 1.073902395651501e-06, "loss": 0.3303, "step": 17695 }, { "epoch": 2.11, "learning_rate": 1.0736384095336628e-06, "loss": 0.4281, "step": 17696 }, { "epoch": 2.11, "learning_rate": 1.0733744469937022e-06, "loss": 0.4184, "step": 17697 }, { "epoch": 2.11, "learning_rate": 1.0731105080359804e-06, "loss": 0.5159, "step": 17698 }, { "epoch": 2.11, "learning_rate": 1.0728465926648622e-06, "loss": 0.5671, "step": 17699 }, { "epoch": 2.11, "learning_rate": 1.0725827008847103e-06, "loss": 0.5517, "step": 17700 }, { "epoch": 2.11, "learning_rate": 1.072318832699886e-06, "loss": 0.4268, "step": 17701 }, { "epoch": 2.11, "learning_rate": 1.0720549881147504e-06, "loss": 0.3661, "step": 17702 }, { "epoch": 2.11, "learning_rate": 1.0717911671336657e-06, "loss": 0.5622, "step": 17703 }, { "epoch": 2.11, "learning_rate": 1.0715273697609938e-06, "loss": 0.5644, "step": 17704 }, { "epoch": 2.11, "learning_rate": 1.071263596001093e-06, "loss": 0.5381, "step": 17705 }, { "epoch": 2.11, "learning_rate": 1.0709998458583253e-06, "loss": 0.5709, "step": 17706 }, { "epoch": 2.11, "learning_rate": 1.0707361193370507e-06, "loss": 0.2862, "step": 17707 }, { "epoch": 2.11, "learning_rate": 1.0704724164416274e-06, "loss": 0.3154, "step": 17708 }, { "epoch": 2.11, "learning_rate": 1.0702087371764155e-06, "loss": 0.527, "step": 17709 }, { "epoch": 2.11, "learning_rate": 1.0699450815457726e-06, "loss": 0.5246, "step": 17710 }, { "epoch": 2.11, "learning_rate": 1.0696814495540572e-06, "loss": 0.5332, "step": 17711 }, { "epoch": 2.11, "learning_rate": 1.0694178412056286e-06, "loss": 0.6348, "step": 17712 }, { "epoch": 2.11, "learning_rate": 1.0691542565048422e-06, "loss": 0.3795, "step": 17713 }, { "epoch": 2.11, "learning_rate": 1.0688906954560557e-06, "loss": 0.35, "step": 17714 }, { "epoch": 2.11, "learning_rate": 1.0686271580636273e-06, "loss": 0.4699, "step": 17715 }, { "epoch": 2.11, "learning_rate": 1.068363644331912e-06, "loss": 0.5286, "step": 17716 }, { "epoch": 2.11, "learning_rate": 1.068100154265265e-06, "loss": 0.5271, "step": 17717 }, { "epoch": 2.11, "learning_rate": 1.0678366878680425e-06, "loss": 0.5885, "step": 17718 }, { "epoch": 2.11, "learning_rate": 1.0675732451445995e-06, "loss": 0.5994, "step": 17719 }, { "epoch": 2.11, "learning_rate": 1.067309826099292e-06, "loss": 0.3556, "step": 17720 }, { "epoch": 2.11, "learning_rate": 1.0670464307364723e-06, "loss": 0.3006, "step": 17721 }, { "epoch": 2.11, "learning_rate": 1.0667830590604963e-06, "loss": 0.5269, "step": 17722 }, { "epoch": 2.11, "learning_rate": 1.0665197110757166e-06, "loss": 0.6094, "step": 17723 }, { "epoch": 2.11, "learning_rate": 1.0662563867864856e-06, "loss": 0.6325, "step": 17724 }, { "epoch": 2.11, "learning_rate": 1.0659930861971568e-06, "loss": 0.5529, "step": 17725 }, { "epoch": 2.11, "learning_rate": 1.0657298093120824e-06, "loss": 0.2948, "step": 17726 }, { "epoch": 2.11, "learning_rate": 1.0654665561356157e-06, "loss": 0.5208, "step": 17727 }, { "epoch": 2.11, "learning_rate": 1.0652033266721062e-06, "loss": 0.4591, "step": 17728 }, { "epoch": 2.11, "learning_rate": 1.064940120925907e-06, "loss": 0.4388, "step": 17729 }, { "epoch": 2.11, "learning_rate": 1.064676938901367e-06, "loss": 0.5642, "step": 17730 }, { "epoch": 2.11, "learning_rate": 1.0644137806028388e-06, "loss": 0.7002, "step": 17731 }, { "epoch": 2.11, "learning_rate": 1.0641506460346701e-06, "loss": 0.274, "step": 17732 }, { "epoch": 2.11, "learning_rate": 1.0638875352012117e-06, "loss": 0.4336, "step": 17733 }, { "epoch": 2.11, "learning_rate": 1.063624448106813e-06, "loss": 0.349, "step": 17734 }, { "epoch": 2.11, "learning_rate": 1.0633613847558235e-06, "loss": 0.5155, "step": 17735 }, { "epoch": 2.11, "learning_rate": 1.06309834515259e-06, "loss": 0.5059, "step": 17736 }, { "epoch": 2.11, "learning_rate": 1.0628353293014626e-06, "loss": 0.6728, "step": 17737 }, { "epoch": 2.11, "learning_rate": 1.0625723372067876e-06, "loss": 0.2994, "step": 17738 }, { "epoch": 2.11, "learning_rate": 1.0623093688729114e-06, "loss": 0.2713, "step": 17739 }, { "epoch": 2.11, "learning_rate": 1.062046424304182e-06, "loss": 0.4459, "step": 17740 }, { "epoch": 2.11, "learning_rate": 1.061783503504946e-06, "loss": 0.4475, "step": 17741 }, { "epoch": 2.11, "learning_rate": 1.06152060647955e-06, "loss": 0.5775, "step": 17742 }, { "epoch": 2.11, "learning_rate": 1.0612577332323384e-06, "loss": 0.5878, "step": 17743 }, { "epoch": 2.11, "learning_rate": 1.0609948837676582e-06, "loss": 0.5482, "step": 17744 }, { "epoch": 2.11, "learning_rate": 1.0607320580898523e-06, "loss": 0.3277, "step": 17745 }, { "epoch": 2.11, "learning_rate": 1.060469256203267e-06, "loss": 0.4174, "step": 17746 }, { "epoch": 2.11, "learning_rate": 1.0602064781122448e-06, "loss": 0.3483, "step": 17747 }, { "epoch": 2.11, "learning_rate": 1.0599437238211306e-06, "loss": 0.548, "step": 17748 }, { "epoch": 2.11, "learning_rate": 1.0596809933342671e-06, "loss": 0.5797, "step": 17749 }, { "epoch": 2.11, "learning_rate": 1.0594182866559989e-06, "loss": 0.7099, "step": 17750 }, { "epoch": 2.11, "learning_rate": 1.0591556037906669e-06, "loss": 0.3521, "step": 17751 }, { "epoch": 2.11, "learning_rate": 1.058892944742613e-06, "loss": 0.3083, "step": 17752 }, { "epoch": 2.11, "learning_rate": 1.0586303095161807e-06, "loss": 0.4051, "step": 17753 }, { "epoch": 2.11, "learning_rate": 1.0583676981157092e-06, "loss": 0.4463, "step": 17754 }, { "epoch": 2.11, "learning_rate": 1.0581051105455407e-06, "loss": 0.5641, "step": 17755 }, { "epoch": 2.11, "learning_rate": 1.0578425468100157e-06, "loss": 0.5138, "step": 17756 }, { "epoch": 2.11, "learning_rate": 1.057580006913475e-06, "loss": 0.3764, "step": 17757 }, { "epoch": 2.11, "learning_rate": 1.0573174908602582e-06, "loss": 0.3068, "step": 17758 }, { "epoch": 2.11, "learning_rate": 1.0570549986547036e-06, "loss": 0.423, "step": 17759 }, { "epoch": 2.12, "learning_rate": 1.0567925303011503e-06, "loss": 0.5653, "step": 17760 }, { "epoch": 2.12, "learning_rate": 1.056530085803939e-06, "loss": 0.5413, "step": 17761 }, { "epoch": 2.12, "learning_rate": 1.0562676651674053e-06, "loss": 0.5078, "step": 17762 }, { "epoch": 2.12, "learning_rate": 1.0560052683958886e-06, "loss": 0.4079, "step": 17763 }, { "epoch": 2.12, "learning_rate": 1.0557428954937266e-06, "loss": 0.3306, "step": 17764 }, { "epoch": 2.12, "learning_rate": 1.055480546465255e-06, "loss": 0.3778, "step": 17765 }, { "epoch": 2.12, "learning_rate": 1.0552182213148119e-06, "loss": 0.4629, "step": 17766 }, { "epoch": 2.12, "learning_rate": 1.054955920046732e-06, "loss": 0.52, "step": 17767 }, { "epoch": 2.12, "learning_rate": 1.0546936426653525e-06, "loss": 0.642, "step": 17768 }, { "epoch": 2.12, "learning_rate": 1.054431389175008e-06, "loss": 0.6145, "step": 17769 }, { "epoch": 2.12, "learning_rate": 1.0541691595800338e-06, "loss": 0.4354, "step": 17770 }, { "epoch": 2.12, "learning_rate": 1.0539069538847646e-06, "loss": 0.4625, "step": 17771 }, { "epoch": 2.12, "learning_rate": 1.0536447720935359e-06, "loss": 0.4052, "step": 17772 }, { "epoch": 2.12, "learning_rate": 1.0533826142106802e-06, "loss": 0.5693, "step": 17773 }, { "epoch": 2.12, "learning_rate": 1.0531204802405303e-06, "loss": 0.5938, "step": 17774 }, { "epoch": 2.12, "learning_rate": 1.0528583701874207e-06, "loss": 0.6023, "step": 17775 }, { "epoch": 2.12, "learning_rate": 1.0525962840556841e-06, "loss": 0.2669, "step": 17776 }, { "epoch": 2.12, "learning_rate": 1.0523342218496518e-06, "loss": 0.4538, "step": 17777 }, { "epoch": 2.12, "learning_rate": 1.052072183573656e-06, "loss": 0.3955, "step": 17778 }, { "epoch": 2.12, "learning_rate": 1.0518101692320298e-06, "loss": 0.4547, "step": 17779 }, { "epoch": 2.12, "learning_rate": 1.0515481788291026e-06, "loss": 0.6002, "step": 17780 }, { "epoch": 2.12, "learning_rate": 1.0512862123692049e-06, "loss": 0.5788, "step": 17781 }, { "epoch": 2.12, "learning_rate": 1.0510242698566675e-06, "loss": 0.4863, "step": 17782 }, { "epoch": 2.12, "learning_rate": 1.0507623512958215e-06, "loss": 0.4674, "step": 17783 }, { "epoch": 2.12, "learning_rate": 1.0505004566909946e-06, "loss": 0.459, "step": 17784 }, { "epoch": 2.12, "learning_rate": 1.0502385860465166e-06, "loss": 0.4651, "step": 17785 }, { "epoch": 2.12, "learning_rate": 1.0499767393667175e-06, "loss": 0.4746, "step": 17786 }, { "epoch": 2.12, "learning_rate": 1.0497149166559245e-06, "loss": 0.568, "step": 17787 }, { "epoch": 2.12, "learning_rate": 1.0494531179184647e-06, "loss": 0.3379, "step": 17788 }, { "epoch": 2.12, "learning_rate": 1.0491913431586665e-06, "loss": 0.3701, "step": 17789 }, { "epoch": 2.12, "learning_rate": 1.0489295923808574e-06, "loss": 0.4595, "step": 17790 }, { "epoch": 2.12, "learning_rate": 1.0486678655893645e-06, "loss": 0.3833, "step": 17791 }, { "epoch": 2.12, "learning_rate": 1.0484061627885128e-06, "loss": 0.5396, "step": 17792 }, { "epoch": 2.12, "learning_rate": 1.0481444839826288e-06, "loss": 0.582, "step": 17793 }, { "epoch": 2.12, "learning_rate": 1.0478828291760395e-06, "loss": 0.4063, "step": 17794 }, { "epoch": 2.12, "learning_rate": 1.0476211983730686e-06, "loss": 0.377, "step": 17795 }, { "epoch": 2.12, "learning_rate": 1.0473595915780405e-06, "loss": 0.4944, "step": 17796 }, { "epoch": 2.12, "learning_rate": 1.0470980087952802e-06, "loss": 0.4549, "step": 17797 }, { "epoch": 2.12, "learning_rate": 1.0468364500291116e-06, "loss": 0.4732, "step": 17798 }, { "epoch": 2.12, "learning_rate": 1.0465749152838594e-06, "loss": 0.6093, "step": 17799 }, { "epoch": 2.12, "learning_rate": 1.0463134045638448e-06, "loss": 0.4441, "step": 17800 }, { "epoch": 2.12, "learning_rate": 1.0460519178733928e-06, "loss": 0.4208, "step": 17801 }, { "epoch": 2.12, "learning_rate": 1.0457904552168243e-06, "loss": 0.3905, "step": 17802 }, { "epoch": 2.12, "learning_rate": 1.0455290165984607e-06, "loss": 0.4362, "step": 17803 }, { "epoch": 2.12, "learning_rate": 1.0452676020226246e-06, "loss": 0.3583, "step": 17804 }, { "epoch": 2.12, "learning_rate": 1.0450062114936372e-06, "loss": 0.5559, "step": 17805 }, { "epoch": 2.12, "learning_rate": 1.04474484501582e-06, "loss": 0.6708, "step": 17806 }, { "epoch": 2.12, "learning_rate": 1.0444835025934916e-06, "loss": 0.264, "step": 17807 }, { "epoch": 2.12, "learning_rate": 1.044222184230974e-06, "loss": 0.3633, "step": 17808 }, { "epoch": 2.12, "learning_rate": 1.0439608899325847e-06, "loss": 0.378, "step": 17809 }, { "epoch": 2.12, "learning_rate": 1.0436996197026456e-06, "loss": 0.5257, "step": 17810 }, { "epoch": 2.12, "learning_rate": 1.0434383735454725e-06, "loss": 0.5868, "step": 17811 }, { "epoch": 2.12, "learning_rate": 1.0431771514653854e-06, "loss": 0.6376, "step": 17812 }, { "epoch": 2.12, "learning_rate": 1.0429159534667024e-06, "loss": 0.3752, "step": 17813 }, { "epoch": 2.12, "learning_rate": 1.042654779553742e-06, "loss": 0.2624, "step": 17814 }, { "epoch": 2.12, "learning_rate": 1.04239362973082e-06, "loss": 0.3723, "step": 17815 }, { "epoch": 2.12, "learning_rate": 1.0421325040022528e-06, "loss": 0.6046, "step": 17816 }, { "epoch": 2.12, "learning_rate": 1.0418714023723586e-06, "loss": 0.4768, "step": 17817 }, { "epoch": 2.12, "learning_rate": 1.0416103248454518e-06, "loss": 0.5983, "step": 17818 }, { "epoch": 2.12, "learning_rate": 1.0413492714258483e-06, "loss": 0.5153, "step": 17819 }, { "epoch": 2.12, "learning_rate": 1.041088242117864e-06, "loss": 0.2801, "step": 17820 }, { "epoch": 2.12, "learning_rate": 1.040827236925814e-06, "loss": 0.3689, "step": 17821 }, { "epoch": 2.12, "learning_rate": 1.0405662558540116e-06, "loss": 0.4111, "step": 17822 }, { "epoch": 2.12, "learning_rate": 1.0403052989067725e-06, "loss": 0.4293, "step": 17823 }, { "epoch": 2.12, "learning_rate": 1.0400443660884082e-06, "loss": 0.6641, "step": 17824 }, { "epoch": 2.12, "learning_rate": 1.0397834574032339e-06, "loss": 0.6005, "step": 17825 }, { "epoch": 2.12, "learning_rate": 1.0395225728555608e-06, "loss": 0.3679, "step": 17826 }, { "epoch": 2.12, "learning_rate": 1.039261712449702e-06, "loss": 0.3594, "step": 17827 }, { "epoch": 2.12, "learning_rate": 1.0390008761899695e-06, "loss": 0.3923, "step": 17828 }, { "epoch": 2.12, "learning_rate": 1.0387400640806763e-06, "loss": 0.5395, "step": 17829 }, { "epoch": 2.12, "learning_rate": 1.0384792761261322e-06, "loss": 0.4855, "step": 17830 }, { "epoch": 2.12, "learning_rate": 1.0382185123306477e-06, "loss": 0.5545, "step": 17831 }, { "epoch": 2.12, "learning_rate": 1.0379577726985345e-06, "loss": 0.2898, "step": 17832 }, { "epoch": 2.12, "learning_rate": 1.0376970572341011e-06, "loss": 0.3278, "step": 17833 }, { "epoch": 2.12, "learning_rate": 1.0374363659416583e-06, "loss": 0.3974, "step": 17834 }, { "epoch": 2.12, "learning_rate": 1.0371756988255149e-06, "loss": 0.429, "step": 17835 }, { "epoch": 2.12, "learning_rate": 1.0369150558899808e-06, "loss": 0.5044, "step": 17836 }, { "epoch": 2.12, "learning_rate": 1.0366544371393636e-06, "loss": 0.7355, "step": 17837 }, { "epoch": 2.12, "learning_rate": 1.0363938425779704e-06, "loss": 0.3543, "step": 17838 }, { "epoch": 2.12, "learning_rate": 1.03613327221011e-06, "loss": 0.3815, "step": 17839 }, { "epoch": 2.12, "learning_rate": 1.0358727260400903e-06, "loss": 0.3504, "step": 17840 }, { "epoch": 2.12, "learning_rate": 1.035612204072216e-06, "loss": 0.4519, "step": 17841 }, { "epoch": 2.12, "learning_rate": 1.035351706310795e-06, "loss": 0.5022, "step": 17842 }, { "epoch": 2.12, "learning_rate": 1.0350912327601346e-06, "loss": 0.6101, "step": 17843 }, { "epoch": 2.13, "learning_rate": 1.0348307834245386e-06, "loss": 0.6304, "step": 17844 }, { "epoch": 2.13, "learning_rate": 1.0345703583083116e-06, "loss": 0.3822, "step": 17845 }, { "epoch": 2.13, "learning_rate": 1.03430995741576e-06, "loss": 0.434, "step": 17846 }, { "epoch": 2.13, "learning_rate": 1.0340495807511882e-06, "loss": 0.5004, "step": 17847 }, { "epoch": 2.13, "learning_rate": 1.0337892283188995e-06, "loss": 0.5406, "step": 17848 }, { "epoch": 2.13, "learning_rate": 1.0335289001231975e-06, "loss": 0.4541, "step": 17849 }, { "epoch": 2.13, "learning_rate": 1.0332685961683858e-06, "loss": 0.7025, "step": 17850 }, { "epoch": 2.13, "learning_rate": 1.0330083164587685e-06, "loss": 0.3198, "step": 17851 }, { "epoch": 2.13, "learning_rate": 1.0327480609986465e-06, "loss": 0.3741, "step": 17852 }, { "epoch": 2.13, "learning_rate": 1.0324878297923213e-06, "loss": 0.464, "step": 17853 }, { "epoch": 2.13, "learning_rate": 1.0322276228440955e-06, "loss": 0.4753, "step": 17854 }, { "epoch": 2.13, "learning_rate": 1.0319674401582712e-06, "loss": 0.5476, "step": 17855 }, { "epoch": 2.13, "learning_rate": 1.031707281739147e-06, "loss": 0.5811, "step": 17856 }, { "epoch": 2.13, "learning_rate": 1.031447147591025e-06, "loss": 0.3516, "step": 17857 }, { "epoch": 2.13, "learning_rate": 1.0311870377182059e-06, "loss": 0.3278, "step": 17858 }, { "epoch": 2.13, "learning_rate": 1.030926952124988e-06, "loss": 0.3213, "step": 17859 }, { "epoch": 2.13, "learning_rate": 1.0306668908156704e-06, "loss": 0.4601, "step": 17860 }, { "epoch": 2.13, "learning_rate": 1.030406853794552e-06, "loss": 0.5015, "step": 17861 }, { "epoch": 2.13, "learning_rate": 1.0301468410659316e-06, "loss": 0.666, "step": 17862 }, { "epoch": 2.13, "learning_rate": 1.0298868526341082e-06, "loss": 0.4288, "step": 17863 }, { "epoch": 2.13, "learning_rate": 1.0296268885033775e-06, "loss": 0.3424, "step": 17864 }, { "epoch": 2.13, "learning_rate": 1.0293669486780389e-06, "loss": 0.4911, "step": 17865 }, { "epoch": 2.13, "learning_rate": 1.0291070331623879e-06, "loss": 0.5022, "step": 17866 }, { "epoch": 2.13, "learning_rate": 1.0288471419607198e-06, "loss": 0.5442, "step": 17867 }, { "epoch": 2.13, "learning_rate": 1.0285872750773323e-06, "loss": 0.6516, "step": 17868 }, { "epoch": 2.13, "learning_rate": 1.0283274325165204e-06, "loss": 0.5377, "step": 17869 }, { "epoch": 2.13, "learning_rate": 1.0280676142825808e-06, "loss": 0.3532, "step": 17870 }, { "epoch": 2.13, "learning_rate": 1.0278078203798057e-06, "loss": 0.3739, "step": 17871 }, { "epoch": 2.13, "learning_rate": 1.027548050812492e-06, "loss": 0.454, "step": 17872 }, { "epoch": 2.13, "learning_rate": 1.0272883055849314e-06, "loss": 0.5126, "step": 17873 }, { "epoch": 2.13, "learning_rate": 1.02702858470142e-06, "loss": 0.5782, "step": 17874 }, { "epoch": 2.13, "learning_rate": 1.0267688881662483e-06, "loss": 0.4485, "step": 17875 }, { "epoch": 2.13, "learning_rate": 1.026509215983711e-06, "loss": 0.3617, "step": 17876 }, { "epoch": 2.13, "learning_rate": 1.0262495681580994e-06, "loss": 0.4418, "step": 17877 }, { "epoch": 2.13, "learning_rate": 1.0259899446937074e-06, "loss": 0.3604, "step": 17878 }, { "epoch": 2.13, "learning_rate": 1.0257303455948245e-06, "loss": 0.5551, "step": 17879 }, { "epoch": 2.13, "learning_rate": 1.0254707708657433e-06, "loss": 0.5591, "step": 17880 }, { "epoch": 2.13, "learning_rate": 1.0252112205107543e-06, "loss": 0.6207, "step": 17881 }, { "epoch": 2.13, "learning_rate": 1.0249516945341464e-06, "loss": 0.3654, "step": 17882 }, { "epoch": 2.13, "learning_rate": 1.0246921929402108e-06, "loss": 0.4042, "step": 17883 }, { "epoch": 2.13, "learning_rate": 1.0244327157332371e-06, "loss": 0.422, "step": 17884 }, { "epoch": 2.13, "learning_rate": 1.0241732629175153e-06, "loss": 0.409, "step": 17885 }, { "epoch": 2.13, "learning_rate": 1.0239138344973323e-06, "loss": 0.5434, "step": 17886 }, { "epoch": 2.13, "learning_rate": 1.0236544304769786e-06, "loss": 0.6618, "step": 17887 }, { "epoch": 2.13, "learning_rate": 1.02339505086074e-06, "loss": 0.3493, "step": 17888 }, { "epoch": 2.13, "learning_rate": 1.0231356956529057e-06, "loss": 0.3401, "step": 17889 }, { "epoch": 2.13, "learning_rate": 1.0228763648577617e-06, "loss": 0.4117, "step": 17890 }, { "epoch": 2.13, "learning_rate": 1.0226170584795954e-06, "loss": 0.4299, "step": 17891 }, { "epoch": 2.13, "learning_rate": 1.0223577765226931e-06, "loss": 0.4532, "step": 17892 }, { "epoch": 2.13, "learning_rate": 1.0220985189913414e-06, "loss": 0.6406, "step": 17893 }, { "epoch": 2.13, "learning_rate": 1.0218392858898253e-06, "loss": 0.5586, "step": 17894 }, { "epoch": 2.13, "learning_rate": 1.0215800772224289e-06, "loss": 0.314, "step": 17895 }, { "epoch": 2.13, "learning_rate": 1.0213208929934388e-06, "loss": 0.4017, "step": 17896 }, { "epoch": 2.13, "learning_rate": 1.0210617332071377e-06, "loss": 0.3453, "step": 17897 }, { "epoch": 2.13, "learning_rate": 1.0208025978678099e-06, "loss": 0.4131, "step": 17898 }, { "epoch": 2.13, "learning_rate": 1.0205434869797395e-06, "loss": 0.5213, "step": 17899 }, { "epoch": 2.13, "learning_rate": 1.0202844005472101e-06, "loss": 0.5658, "step": 17900 }, { "epoch": 2.13, "learning_rate": 1.020025338574504e-06, "loss": 0.3653, "step": 17901 }, { "epoch": 2.13, "learning_rate": 1.019766301065902e-06, "loss": 0.3091, "step": 17902 }, { "epoch": 2.13, "learning_rate": 1.0195072880256877e-06, "loss": 0.4792, "step": 17903 }, { "epoch": 2.13, "learning_rate": 1.0192482994581426e-06, "loss": 0.4113, "step": 17904 }, { "epoch": 2.13, "learning_rate": 1.0189893353675467e-06, "loss": 0.55, "step": 17905 }, { "epoch": 2.13, "learning_rate": 1.0187303957581812e-06, "loss": 0.4948, "step": 17906 }, { "epoch": 2.13, "learning_rate": 1.0184714806343277e-06, "loss": 0.3375, "step": 17907 }, { "epoch": 2.13, "learning_rate": 1.0182125900002638e-06, "loss": 0.4286, "step": 17908 }, { "epoch": 2.13, "learning_rate": 1.017953723860271e-06, "loss": 0.4214, "step": 17909 }, { "epoch": 2.13, "learning_rate": 1.0176948822186267e-06, "loss": 0.3716, "step": 17910 }, { "epoch": 2.13, "learning_rate": 1.0174360650796114e-06, "loss": 0.5496, "step": 17911 }, { "epoch": 2.13, "learning_rate": 1.0171772724475012e-06, "loss": 0.6693, "step": 17912 }, { "epoch": 2.13, "learning_rate": 1.0169185043265752e-06, "loss": 0.42, "step": 17913 }, { "epoch": 2.13, "learning_rate": 1.0166597607211106e-06, "loss": 0.4192, "step": 17914 }, { "epoch": 2.13, "learning_rate": 1.0164010416353859e-06, "loss": 0.4135, "step": 17915 }, { "epoch": 2.13, "learning_rate": 1.016142347073676e-06, "loss": 0.4132, "step": 17916 }, { "epoch": 2.13, "learning_rate": 1.0158836770402572e-06, "loss": 0.4973, "step": 17917 }, { "epoch": 2.13, "learning_rate": 1.0156250315394055e-06, "loss": 0.583, "step": 17918 }, { "epoch": 2.13, "learning_rate": 1.0153664105753975e-06, "loss": 0.5377, "step": 17919 }, { "epoch": 2.13, "learning_rate": 1.0151078141525064e-06, "loss": 0.3521, "step": 17920 }, { "epoch": 2.13, "learning_rate": 1.0148492422750078e-06, "loss": 0.3758, "step": 17921 }, { "epoch": 2.13, "learning_rate": 1.0145906949471765e-06, "loss": 0.4387, "step": 17922 }, { "epoch": 2.13, "learning_rate": 1.014332172173286e-06, "loss": 0.527, "step": 17923 }, { "epoch": 2.13, "learning_rate": 1.0140736739576083e-06, "loss": 0.5809, "step": 17924 }, { "epoch": 2.13, "learning_rate": 1.013815200304417e-06, "loss": 0.5914, "step": 17925 }, { "epoch": 2.13, "learning_rate": 1.0135567512179867e-06, "loss": 0.241, "step": 17926 }, { "epoch": 2.13, "learning_rate": 1.013298326702587e-06, "loss": 0.399, "step": 17927 }, { "epoch": 2.14, "learning_rate": 1.0130399267624905e-06, "loss": 0.5114, "step": 17928 }, { "epoch": 2.14, "learning_rate": 1.0127815514019698e-06, "loss": 0.4464, "step": 17929 }, { "epoch": 2.14, "learning_rate": 1.0125232006252947e-06, "loss": 0.5912, "step": 17930 }, { "epoch": 2.14, "learning_rate": 1.0122648744367354e-06, "loss": 0.6003, "step": 17931 }, { "epoch": 2.14, "learning_rate": 1.012006572840562e-06, "loss": 0.3092, "step": 17932 }, { "epoch": 2.14, "learning_rate": 1.0117482958410452e-06, "loss": 0.4279, "step": 17933 }, { "epoch": 2.14, "learning_rate": 1.0114900434424546e-06, "loss": 0.3374, "step": 17934 }, { "epoch": 2.14, "learning_rate": 1.0112318156490577e-06, "loss": 0.522, "step": 17935 }, { "epoch": 2.14, "learning_rate": 1.0109736124651237e-06, "loss": 0.421, "step": 17936 }, { "epoch": 2.14, "learning_rate": 1.0107154338949217e-06, "loss": 0.6367, "step": 17937 }, { "epoch": 2.14, "learning_rate": 1.010457279942719e-06, "loss": 0.3536, "step": 17938 }, { "epoch": 2.14, "learning_rate": 1.010199150612781e-06, "loss": 0.3421, "step": 17939 }, { "epoch": 2.14, "learning_rate": 1.0099410459093765e-06, "loss": 0.4309, "step": 17940 }, { "epoch": 2.14, "learning_rate": 1.009682965836771e-06, "loss": 0.4383, "step": 17941 }, { "epoch": 2.14, "learning_rate": 1.0094249103992324e-06, "loss": 0.4646, "step": 17942 }, { "epoch": 2.14, "learning_rate": 1.009166879601024e-06, "loss": 0.5603, "step": 17943 }, { "epoch": 2.14, "learning_rate": 1.0089088734464133e-06, "loss": 0.5193, "step": 17944 }, { "epoch": 2.14, "learning_rate": 1.008650891939664e-06, "loss": 0.2908, "step": 17945 }, { "epoch": 2.14, "learning_rate": 1.0083929350850394e-06, "loss": 0.404, "step": 17946 }, { "epoch": 2.14, "learning_rate": 1.008135002886805e-06, "loss": 0.4674, "step": 17947 }, { "epoch": 2.14, "learning_rate": 1.0078770953492242e-06, "loss": 0.5656, "step": 17948 }, { "epoch": 2.14, "learning_rate": 1.007619212476561e-06, "loss": 0.5869, "step": 17949 }, { "epoch": 2.14, "learning_rate": 1.0073613542730765e-06, "loss": 0.6477, "step": 17950 }, { "epoch": 2.14, "learning_rate": 1.0071035207430352e-06, "loss": 0.3492, "step": 17951 }, { "epoch": 2.14, "learning_rate": 1.0068457118906968e-06, "loss": 0.4702, "step": 17952 }, { "epoch": 2.14, "learning_rate": 1.006587927720325e-06, "loss": 0.4134, "step": 17953 }, { "epoch": 2.14, "learning_rate": 1.0063301682361792e-06, "loss": 0.3883, "step": 17954 }, { "epoch": 2.14, "learning_rate": 1.0060724334425212e-06, "loss": 0.5304, "step": 17955 }, { "epoch": 2.14, "learning_rate": 1.005814723343611e-06, "loss": 0.5465, "step": 17956 }, { "epoch": 2.14, "learning_rate": 1.00555703794371e-06, "loss": 0.2842, "step": 17957 }, { "epoch": 2.14, "learning_rate": 1.0052993772470762e-06, "loss": 0.3002, "step": 17958 }, { "epoch": 2.14, "learning_rate": 1.0050417412579686e-06, "loss": 0.4129, "step": 17959 }, { "epoch": 2.14, "learning_rate": 1.004784129980647e-06, "loss": 0.4751, "step": 17960 }, { "epoch": 2.14, "learning_rate": 1.0045265434193683e-06, "loss": 0.535, "step": 17961 }, { "epoch": 2.14, "learning_rate": 1.0042689815783915e-06, "loss": 0.655, "step": 17962 }, { "epoch": 2.14, "learning_rate": 1.004011444461974e-06, "loss": 0.3457, "step": 17963 }, { "epoch": 2.14, "learning_rate": 1.0037539320743733e-06, "loss": 0.3578, "step": 17964 }, { "epoch": 2.14, "learning_rate": 1.0034964444198452e-06, "loss": 0.5194, "step": 17965 }, { "epoch": 2.14, "learning_rate": 1.0032389815026472e-06, "loss": 0.4557, "step": 17966 }, { "epoch": 2.14, "learning_rate": 1.0029815433270335e-06, "loss": 0.5406, "step": 17967 }, { "epoch": 2.14, "learning_rate": 1.0027241298972615e-06, "loss": 0.5923, "step": 17968 }, { "epoch": 2.14, "learning_rate": 1.0024667412175842e-06, "loss": 0.6701, "step": 17969 }, { "epoch": 2.14, "learning_rate": 1.0022093772922574e-06, "loss": 0.3825, "step": 17970 }, { "epoch": 2.14, "learning_rate": 1.0019520381255353e-06, "loss": 0.4053, "step": 17971 }, { "epoch": 2.14, "learning_rate": 1.0016947237216726e-06, "loss": 0.4936, "step": 17972 }, { "epoch": 2.14, "learning_rate": 1.0014374340849217e-06, "loss": 0.4776, "step": 17973 }, { "epoch": 2.14, "learning_rate": 1.0011801692195353e-06, "loss": 0.6075, "step": 17974 }, { "epoch": 2.14, "learning_rate": 1.0009229291297668e-06, "loss": 0.5272, "step": 17975 }, { "epoch": 2.14, "learning_rate": 1.0006657138198676e-06, "loss": 0.2988, "step": 17976 }, { "epoch": 2.14, "learning_rate": 1.0004085232940897e-06, "loss": 0.3578, "step": 17977 }, { "epoch": 2.14, "learning_rate": 1.0001513575566849e-06, "loss": 0.3348, "step": 17978 }, { "epoch": 2.14, "learning_rate": 9.998942166119046e-07, "loss": 0.4226, "step": 17979 }, { "epoch": 2.14, "learning_rate": 9.99637100463999e-07, "loss": 0.6232, "step": 17980 }, { "epoch": 2.14, "learning_rate": 9.99380009117217e-07, "loss": 0.6039, "step": 17981 }, { "epoch": 2.14, "learning_rate": 9.991229425758095e-07, "loss": 0.3031, "step": 17982 }, { "epoch": 2.14, "learning_rate": 9.988659008440267e-07, "loss": 0.3457, "step": 17983 }, { "epoch": 2.14, "learning_rate": 9.986088839261155e-07, "loss": 0.4597, "step": 17984 }, { "epoch": 2.14, "learning_rate": 9.983518918263252e-07, "loss": 0.4485, "step": 17985 }, { "epoch": 2.14, "learning_rate": 9.980949245489051e-07, "loss": 0.523, "step": 17986 }, { "epoch": 2.14, "learning_rate": 9.978379820981022e-07, "loss": 0.6536, "step": 17987 }, { "epoch": 2.14, "learning_rate": 9.975810644781625e-07, "loss": 0.4023, "step": 17988 }, { "epoch": 2.14, "learning_rate": 9.973241716933338e-07, "loss": 0.3827, "step": 17989 }, { "epoch": 2.14, "learning_rate": 9.970673037478634e-07, "loss": 0.3628, "step": 17990 }, { "epoch": 2.14, "learning_rate": 9.968104606459958e-07, "loss": 0.4229, "step": 17991 }, { "epoch": 2.14, "learning_rate": 9.965536423919775e-07, "loss": 0.4967, "step": 17992 }, { "epoch": 2.14, "learning_rate": 9.962968489900535e-07, "loss": 0.5901, "step": 17993 }, { "epoch": 2.14, "learning_rate": 9.960400804444697e-07, "loss": 0.5127, "step": 17994 }, { "epoch": 2.14, "learning_rate": 9.957833367594696e-07, "loss": 0.2931, "step": 17995 }, { "epoch": 2.14, "learning_rate": 9.955266179392961e-07, "loss": 0.5269, "step": 17996 }, { "epoch": 2.14, "learning_rate": 9.95269923988194e-07, "loss": 0.4327, "step": 17997 }, { "epoch": 2.14, "learning_rate": 9.950132549104071e-07, "loss": 0.4724, "step": 17998 }, { "epoch": 2.14, "learning_rate": 9.947566107101767e-07, "loss": 0.5095, "step": 17999 }, { "epoch": 2.14, "learning_rate": 9.944999913917458e-07, "loss": 0.5628, "step": 18000 }, { "epoch": 2.14, "learning_rate": 9.942433969593572e-07, "loss": 0.3381, "step": 18001 }, { "epoch": 2.14, "learning_rate": 9.939868274172515e-07, "loss": 0.4357, "step": 18002 }, { "epoch": 2.14, "learning_rate": 9.93730282769669e-07, "loss": 0.3754, "step": 18003 }, { "epoch": 2.14, "learning_rate": 9.934737630208513e-07, "loss": 0.5265, "step": 18004 }, { "epoch": 2.14, "learning_rate": 9.932172681750388e-07, "loss": 0.6544, "step": 18005 }, { "epoch": 2.14, "learning_rate": 9.929607982364723e-07, "loss": 0.5917, "step": 18006 }, { "epoch": 2.14, "learning_rate": 9.927043532093891e-07, "loss": 0.3395, "step": 18007 }, { "epoch": 2.14, "learning_rate": 9.924479330980304e-07, "loss": 0.3869, "step": 18008 }, { "epoch": 2.14, "learning_rate": 9.92191537906634e-07, "loss": 0.4045, "step": 18009 }, { "epoch": 2.14, "learning_rate": 9.919351676394368e-07, "loss": 0.4171, "step": 18010 }, { "epoch": 2.14, "learning_rate": 9.916788223006779e-07, "loss": 0.5025, "step": 18011 }, { "epoch": 2.15, "learning_rate": 9.91422501894595e-07, "loss": 0.626, "step": 18012 }, { "epoch": 2.15, "learning_rate": 9.911662064254252e-07, "loss": 0.2641, "step": 18013 }, { "epoch": 2.15, "learning_rate": 9.90909935897404e-07, "loss": 0.4072, "step": 18014 }, { "epoch": 2.15, "learning_rate": 9.906536903147687e-07, "loss": 0.4439, "step": 18015 }, { "epoch": 2.15, "learning_rate": 9.903974696817538e-07, "loss": 0.4696, "step": 18016 }, { "epoch": 2.15, "learning_rate": 9.901412740025964e-07, "loss": 0.6046, "step": 18017 }, { "epoch": 2.15, "learning_rate": 9.898851032815295e-07, "loss": 0.6098, "step": 18018 }, { "epoch": 2.15, "learning_rate": 9.896289575227889e-07, "loss": 0.4952, "step": 18019 }, { "epoch": 2.15, "learning_rate": 9.893728367306078e-07, "loss": 0.3412, "step": 18020 }, { "epoch": 2.15, "learning_rate": 9.891167409092216e-07, "loss": 0.4607, "step": 18021 }, { "epoch": 2.15, "learning_rate": 9.888606700628613e-07, "loss": 0.4373, "step": 18022 }, { "epoch": 2.15, "learning_rate": 9.886046241957622e-07, "loss": 0.4971, "step": 18023 }, { "epoch": 2.15, "learning_rate": 9.883486033121553e-07, "loss": 0.6507, "step": 18024 }, { "epoch": 2.15, "learning_rate": 9.880926074162719e-07, "loss": 0.5174, "step": 18025 }, { "epoch": 2.15, "learning_rate": 9.878366365123448e-07, "loss": 0.2028, "step": 18026 }, { "epoch": 2.15, "learning_rate": 9.875806906046048e-07, "loss": 0.3367, "step": 18027 }, { "epoch": 2.15, "learning_rate": 9.873247696972837e-07, "loss": 0.3815, "step": 18028 }, { "epoch": 2.15, "learning_rate": 9.870688737946102e-07, "loss": 0.4591, "step": 18029 }, { "epoch": 2.15, "learning_rate": 9.868130029008161e-07, "loss": 0.5911, "step": 18030 }, { "epoch": 2.15, "learning_rate": 9.86557157020129e-07, "loss": 0.5072, "step": 18031 }, { "epoch": 2.15, "learning_rate": 9.863013361567803e-07, "loss": 0.3528, "step": 18032 }, { "epoch": 2.15, "learning_rate": 9.860455403149965e-07, "loss": 0.4221, "step": 18033 }, { "epoch": 2.15, "learning_rate": 9.857897694990068e-07, "loss": 0.4596, "step": 18034 }, { "epoch": 2.15, "learning_rate": 9.855340237130395e-07, "loss": 0.462, "step": 18035 }, { "epoch": 2.15, "learning_rate": 9.852783029613224e-07, "loss": 0.6261, "step": 18036 }, { "epoch": 2.15, "learning_rate": 9.850226072480822e-07, "loss": 0.5804, "step": 18037 }, { "epoch": 2.15, "learning_rate": 9.847669365775448e-07, "loss": 0.3203, "step": 18038 }, { "epoch": 2.15, "learning_rate": 9.845112909539378e-07, "loss": 0.3302, "step": 18039 }, { "epoch": 2.15, "learning_rate": 9.842556703814854e-07, "loss": 0.3523, "step": 18040 }, { "epoch": 2.15, "learning_rate": 9.840000748644137e-07, "loss": 0.4721, "step": 18041 }, { "epoch": 2.15, "learning_rate": 9.837445044069486e-07, "loss": 0.5047, "step": 18042 }, { "epoch": 2.15, "learning_rate": 9.834889590133144e-07, "loss": 0.6205, "step": 18043 }, { "epoch": 2.15, "learning_rate": 9.832334386877352e-07, "loss": 0.4543, "step": 18044 }, { "epoch": 2.15, "learning_rate": 9.829779434344333e-07, "loss": 0.2765, "step": 18045 }, { "epoch": 2.15, "learning_rate": 9.827224732576338e-07, "loss": 0.3864, "step": 18046 }, { "epoch": 2.15, "learning_rate": 9.824670281615594e-07, "loss": 0.3848, "step": 18047 }, { "epoch": 2.15, "learning_rate": 9.82211608150432e-07, "loss": 0.5322, "step": 18048 }, { "epoch": 2.15, "learning_rate": 9.819562132284738e-07, "loss": 0.6455, "step": 18049 }, { "epoch": 2.15, "learning_rate": 9.81700843399907e-07, "loss": 0.4522, "step": 18050 }, { "epoch": 2.15, "learning_rate": 9.814454986689533e-07, "loss": 0.3207, "step": 18051 }, { "epoch": 2.15, "learning_rate": 9.81190179039833e-07, "loss": 0.5269, "step": 18052 }, { "epoch": 2.15, "learning_rate": 9.809348845167658e-07, "loss": 0.3833, "step": 18053 }, { "epoch": 2.15, "learning_rate": 9.80679615103973e-07, "loss": 0.4675, "step": 18054 }, { "epoch": 2.15, "learning_rate": 9.80424370805673e-07, "loss": 0.5581, "step": 18055 }, { "epoch": 2.15, "learning_rate": 9.801691516260856e-07, "loss": 0.6191, "step": 18056 }, { "epoch": 2.15, "learning_rate": 9.799139575694294e-07, "loss": 0.2928, "step": 18057 }, { "epoch": 2.15, "learning_rate": 9.796587886399242e-07, "loss": 0.3787, "step": 18058 }, { "epoch": 2.15, "learning_rate": 9.794036448417866e-07, "loss": 0.3889, "step": 18059 }, { "epoch": 2.15, "learning_rate": 9.791485261792336e-07, "loss": 0.5068, "step": 18060 }, { "epoch": 2.15, "learning_rate": 9.78893432656483e-07, "loss": 0.5621, "step": 18061 }, { "epoch": 2.15, "learning_rate": 9.786383642777526e-07, "loss": 0.5865, "step": 18062 }, { "epoch": 2.15, "learning_rate": 9.783833210472566e-07, "loss": 0.3973, "step": 18063 }, { "epoch": 2.15, "learning_rate": 9.781283029692121e-07, "loss": 0.3737, "step": 18064 }, { "epoch": 2.15, "learning_rate": 9.778733100478353e-07, "loss": 0.4608, "step": 18065 }, { "epoch": 2.15, "learning_rate": 9.776183422873403e-07, "loss": 0.4066, "step": 18066 }, { "epoch": 2.15, "learning_rate": 9.77363399691941e-07, "loss": 0.5352, "step": 18067 }, { "epoch": 2.15, "learning_rate": 9.771084822658525e-07, "loss": 0.5839, "step": 18068 }, { "epoch": 2.15, "learning_rate": 9.768535900132894e-07, "loss": 0.4783, "step": 18069 }, { "epoch": 2.15, "learning_rate": 9.765987229384635e-07, "loss": 0.3075, "step": 18070 }, { "epoch": 2.15, "learning_rate": 9.763438810455885e-07, "loss": 0.3745, "step": 18071 }, { "epoch": 2.15, "learning_rate": 9.76089064338878e-07, "loss": 0.4813, "step": 18072 }, { "epoch": 2.15, "learning_rate": 9.758342728225427e-07, "loss": 0.501, "step": 18073 }, { "epoch": 2.15, "learning_rate": 9.755795065007945e-07, "loss": 0.7472, "step": 18074 }, { "epoch": 2.15, "learning_rate": 9.753247653778446e-07, "loss": 0.5557, "step": 18075 }, { "epoch": 2.15, "learning_rate": 9.750700494579044e-07, "loss": 0.3026, "step": 18076 }, { "epoch": 2.15, "learning_rate": 9.74815358745185e-07, "loss": 0.3791, "step": 18077 }, { "epoch": 2.15, "learning_rate": 9.745606932438947e-07, "loss": 0.4032, "step": 18078 }, { "epoch": 2.15, "learning_rate": 9.743060529582444e-07, "loss": 0.5205, "step": 18079 }, { "epoch": 2.15, "learning_rate": 9.740514378924437e-07, "loss": 0.6519, "step": 18080 }, { "epoch": 2.15, "learning_rate": 9.73796848050701e-07, "loss": 0.5221, "step": 18081 }, { "epoch": 2.15, "learning_rate": 9.735422834372232e-07, "loss": 0.2774, "step": 18082 }, { "epoch": 2.15, "learning_rate": 9.7328774405622e-07, "loss": 0.4213, "step": 18083 }, { "epoch": 2.15, "learning_rate": 9.73033229911898e-07, "loss": 0.4247, "step": 18084 }, { "epoch": 2.15, "learning_rate": 9.72778741008466e-07, "loss": 0.4229, "step": 18085 }, { "epoch": 2.15, "learning_rate": 9.72524277350129e-07, "loss": 0.5939, "step": 18086 }, { "epoch": 2.15, "learning_rate": 9.722698389410943e-07, "loss": 0.6391, "step": 18087 }, { "epoch": 2.15, "learning_rate": 9.720154257855675e-07, "loss": 0.3408, "step": 18088 }, { "epoch": 2.15, "learning_rate": 9.71761037887753e-07, "loss": 0.3503, "step": 18089 }, { "epoch": 2.15, "learning_rate": 9.715066752518567e-07, "loss": 0.3953, "step": 18090 }, { "epoch": 2.15, "learning_rate": 9.712523378820835e-07, "loss": 0.4936, "step": 18091 }, { "epoch": 2.15, "learning_rate": 9.70998025782638e-07, "loss": 0.5451, "step": 18092 }, { "epoch": 2.15, "learning_rate": 9.70743738957723e-07, "loss": 0.6047, "step": 18093 }, { "epoch": 2.15, "learning_rate": 9.704894774115427e-07, "loss": 0.5353, "step": 18094 }, { "epoch": 2.15, "learning_rate": 9.70235241148299e-07, "loss": 0.3812, "step": 18095 }, { "epoch": 2.16, "learning_rate": 9.69981030172196e-07, "loss": 0.4246, "step": 18096 }, { "epoch": 2.16, "learning_rate": 9.697268444874342e-07, "loss": 0.4709, "step": 18097 }, { "epoch": 2.16, "learning_rate": 9.694726840982158e-07, "loss": 0.5028, "step": 18098 }, { "epoch": 2.16, "learning_rate": 9.692185490087425e-07, "loss": 0.6297, "step": 18099 }, { "epoch": 2.16, "learning_rate": 9.689644392232159e-07, "loss": 0.5745, "step": 18100 }, { "epoch": 2.16, "learning_rate": 9.687103547458356e-07, "loss": 0.3114, "step": 18101 }, { "epoch": 2.16, "learning_rate": 9.684562955808008e-07, "loss": 0.3543, "step": 18102 }, { "epoch": 2.16, "learning_rate": 9.682022617323129e-07, "loss": 0.4413, "step": 18103 }, { "epoch": 2.16, "learning_rate": 9.67948253204569e-07, "loss": 0.4563, "step": 18104 }, { "epoch": 2.16, "learning_rate": 9.67694270001769e-07, "loss": 0.5394, "step": 18105 }, { "epoch": 2.16, "learning_rate": 9.674403121281118e-07, "loss": 0.6125, "step": 18106 }, { "epoch": 2.16, "learning_rate": 9.67186379587795e-07, "loss": 0.4373, "step": 18107 }, { "epoch": 2.16, "learning_rate": 9.669324723850157e-07, "loss": 0.4469, "step": 18108 }, { "epoch": 2.16, "learning_rate": 9.666785905239717e-07, "loss": 0.4303, "step": 18109 }, { "epoch": 2.16, "learning_rate": 9.664247340088586e-07, "loss": 0.4764, "step": 18110 }, { "epoch": 2.16, "learning_rate": 9.661709028438742e-07, "loss": 0.4888, "step": 18111 }, { "epoch": 2.16, "learning_rate": 9.659170970332124e-07, "loss": 0.6385, "step": 18112 }, { "epoch": 2.16, "learning_rate": 9.6566331658107e-07, "loss": 0.4629, "step": 18113 }, { "epoch": 2.16, "learning_rate": 9.654095614916414e-07, "loss": 0.3419, "step": 18114 }, { "epoch": 2.16, "learning_rate": 9.65155831769123e-07, "loss": 0.3921, "step": 18115 }, { "epoch": 2.16, "learning_rate": 9.649021274177067e-07, "loss": 0.4583, "step": 18116 }, { "epoch": 2.16, "learning_rate": 9.646484484415867e-07, "loss": 0.5306, "step": 18117 }, { "epoch": 2.16, "learning_rate": 9.643947948449573e-07, "loss": 0.6324, "step": 18118 }, { "epoch": 2.16, "learning_rate": 9.641411666320099e-07, "loss": 0.5786, "step": 18119 }, { "epoch": 2.16, "learning_rate": 9.63887563806938e-07, "loss": 0.2982, "step": 18120 }, { "epoch": 2.16, "learning_rate": 9.636339863739335e-07, "loss": 0.4697, "step": 18121 }, { "epoch": 2.16, "learning_rate": 9.633804343371891e-07, "loss": 0.4248, "step": 18122 }, { "epoch": 2.16, "learning_rate": 9.631269077008948e-07, "loss": 0.4577, "step": 18123 }, { "epoch": 2.16, "learning_rate": 9.62873406469241e-07, "loss": 0.5816, "step": 18124 }, { "epoch": 2.16, "learning_rate": 9.626199306464187e-07, "loss": 0.6101, "step": 18125 }, { "epoch": 2.16, "learning_rate": 9.62366480236619e-07, "loss": 0.3451, "step": 18126 }, { "epoch": 2.16, "learning_rate": 9.621130552440294e-07, "loss": 0.384, "step": 18127 }, { "epoch": 2.16, "learning_rate": 9.6185965567284e-07, "loss": 0.3713, "step": 18128 }, { "epoch": 2.16, "learning_rate": 9.616062815272404e-07, "loss": 0.4832, "step": 18129 }, { "epoch": 2.16, "learning_rate": 9.61352932811418e-07, "loss": 0.5683, "step": 18130 }, { "epoch": 2.16, "learning_rate": 9.610996095295599e-07, "loss": 0.6377, "step": 18131 }, { "epoch": 2.16, "learning_rate": 9.608463116858544e-07, "loss": 0.3499, "step": 18132 }, { "epoch": 2.16, "learning_rate": 9.605930392844896e-07, "loss": 0.3923, "step": 18133 }, { "epoch": 2.16, "learning_rate": 9.603397923296498e-07, "loss": 0.4175, "step": 18134 }, { "epoch": 2.16, "learning_rate": 9.600865708255225e-07, "loss": 0.4083, "step": 18135 }, { "epoch": 2.16, "learning_rate": 9.598333747762937e-07, "loss": 0.5907, "step": 18136 }, { "epoch": 2.16, "learning_rate": 9.595802041861491e-07, "loss": 0.6535, "step": 18137 }, { "epoch": 2.16, "learning_rate": 9.59327059059273e-07, "loss": 0.4228, "step": 18138 }, { "epoch": 2.16, "learning_rate": 9.590739393998488e-07, "loss": 0.3723, "step": 18139 }, { "epoch": 2.16, "learning_rate": 9.588208452120618e-07, "loss": 0.4162, "step": 18140 }, { "epoch": 2.16, "learning_rate": 9.585677765000963e-07, "loss": 0.4363, "step": 18141 }, { "epoch": 2.16, "learning_rate": 9.583147332681342e-07, "loss": 0.4948, "step": 18142 }, { "epoch": 2.16, "learning_rate": 9.58061715520359e-07, "loss": 0.6014, "step": 18143 }, { "epoch": 2.16, "learning_rate": 9.578087232609536e-07, "loss": 0.4684, "step": 18144 }, { "epoch": 2.16, "learning_rate": 9.575557564940997e-07, "loss": 0.4335, "step": 18145 }, { "epoch": 2.16, "learning_rate": 9.573028152239772e-07, "loss": 0.4131, "step": 18146 }, { "epoch": 2.16, "learning_rate": 9.57049899454769e-07, "loss": 0.4862, "step": 18147 }, { "epoch": 2.16, "learning_rate": 9.567970091906554e-07, "loss": 0.5377, "step": 18148 }, { "epoch": 2.16, "learning_rate": 9.565441444358176e-07, "loss": 0.5208, "step": 18149 }, { "epoch": 2.16, "learning_rate": 9.562913051944336e-07, "loss": 0.51, "step": 18150 }, { "epoch": 2.16, "learning_rate": 9.560384914706846e-07, "loss": 0.2815, "step": 18151 }, { "epoch": 2.16, "learning_rate": 9.557857032687493e-07, "loss": 0.3698, "step": 18152 }, { "epoch": 2.16, "learning_rate": 9.555329405928045e-07, "loss": 0.3505, "step": 18153 }, { "epoch": 2.16, "learning_rate": 9.552802034470302e-07, "loss": 0.4446, "step": 18154 }, { "epoch": 2.16, "learning_rate": 9.550274918356038e-07, "loss": 0.6313, "step": 18155 }, { "epoch": 2.16, "learning_rate": 9.547748057627031e-07, "loss": 0.6385, "step": 18156 }, { "epoch": 2.16, "learning_rate": 9.545221452325037e-07, "loss": 0.3318, "step": 18157 }, { "epoch": 2.16, "learning_rate": 9.542695102491842e-07, "loss": 0.4048, "step": 18158 }, { "epoch": 2.16, "learning_rate": 9.540169008169182e-07, "loss": 0.4017, "step": 18159 }, { "epoch": 2.16, "learning_rate": 9.537643169398832e-07, "loss": 0.4314, "step": 18160 }, { "epoch": 2.16, "learning_rate": 9.535117586222533e-07, "loss": 0.4924, "step": 18161 }, { "epoch": 2.16, "learning_rate": 9.532592258682035e-07, "loss": 0.602, "step": 18162 }, { "epoch": 2.16, "learning_rate": 9.530067186819089e-07, "loss": 0.4231, "step": 18163 }, { "epoch": 2.16, "learning_rate": 9.527542370675436e-07, "loss": 0.3572, "step": 18164 }, { "epoch": 2.16, "learning_rate": 9.525017810292797e-07, "loss": 0.3281, "step": 18165 }, { "epoch": 2.16, "learning_rate": 9.522493505712921e-07, "loss": 0.3716, "step": 18166 }, { "epoch": 2.16, "learning_rate": 9.519969456977527e-07, "loss": 0.5344, "step": 18167 }, { "epoch": 2.16, "learning_rate": 9.517445664128328e-07, "loss": 0.6316, "step": 18168 }, { "epoch": 2.16, "learning_rate": 9.514922127207052e-07, "loss": 0.4271, "step": 18169 }, { "epoch": 2.16, "learning_rate": 9.512398846255411e-07, "loss": 0.2775, "step": 18170 }, { "epoch": 2.16, "learning_rate": 9.509875821315126e-07, "loss": 0.4593, "step": 18171 }, { "epoch": 2.16, "learning_rate": 9.507353052427884e-07, "loss": 0.5212, "step": 18172 }, { "epoch": 2.16, "learning_rate": 9.504830539635404e-07, "loss": 0.551, "step": 18173 }, { "epoch": 2.16, "learning_rate": 9.502308282979367e-07, "loss": 0.5747, "step": 18174 }, { "epoch": 2.16, "learning_rate": 9.499786282501483e-07, "loss": 0.7283, "step": 18175 }, { "epoch": 2.16, "learning_rate": 9.497264538243425e-07, "loss": 0.4131, "step": 18176 }, { "epoch": 2.16, "learning_rate": 9.494743050246885e-07, "loss": 0.3955, "step": 18177 }, { "epoch": 2.16, "learning_rate": 9.492221818553544e-07, "loss": 0.425, "step": 18178 }, { "epoch": 2.16, "learning_rate": 9.489700843205085e-07, "loss": 0.5081, "step": 18179 }, { "epoch": 2.17, "learning_rate": 9.487180124243173e-07, "loss": 0.539, "step": 18180 }, { "epoch": 2.17, "learning_rate": 9.484659661709464e-07, "loss": 0.6054, "step": 18181 }, { "epoch": 2.17, "learning_rate": 9.482139455645645e-07, "loss": 0.4204, "step": 18182 }, { "epoch": 2.17, "learning_rate": 9.479619506093352e-07, "loss": 0.3194, "step": 18183 }, { "epoch": 2.17, "learning_rate": 9.477099813094251e-07, "loss": 0.4774, "step": 18184 }, { "epoch": 2.17, "learning_rate": 9.474580376689993e-07, "loss": 0.4306, "step": 18185 }, { "epoch": 2.17, "learning_rate": 9.472061196922234e-07, "loss": 0.6145, "step": 18186 }, { "epoch": 2.17, "learning_rate": 9.469542273832603e-07, "loss": 0.599, "step": 18187 }, { "epoch": 2.17, "learning_rate": 9.467023607462734e-07, "loss": 0.4151, "step": 18188 }, { "epoch": 2.17, "learning_rate": 9.464505197854268e-07, "loss": 0.3724, "step": 18189 }, { "epoch": 2.17, "learning_rate": 9.461987045048846e-07, "loss": 0.4203, "step": 18190 }, { "epoch": 2.17, "learning_rate": 9.459469149088071e-07, "loss": 0.459, "step": 18191 }, { "epoch": 2.17, "learning_rate": 9.456951510013574e-07, "loss": 0.479, "step": 18192 }, { "epoch": 2.17, "learning_rate": 9.454434127866974e-07, "loss": 0.4972, "step": 18193 }, { "epoch": 2.17, "learning_rate": 9.451917002689892e-07, "loss": 0.6115, "step": 18194 }, { "epoch": 2.17, "learning_rate": 9.449400134523923e-07, "loss": 0.3577, "step": 18195 }, { "epoch": 2.17, "learning_rate": 9.446883523410668e-07, "loss": 0.3402, "step": 18196 }, { "epoch": 2.17, "learning_rate": 9.444367169391744e-07, "loss": 0.487, "step": 18197 }, { "epoch": 2.17, "learning_rate": 9.441851072508726e-07, "loss": 0.5554, "step": 18198 }, { "epoch": 2.17, "learning_rate": 9.439335232803212e-07, "loss": 0.5367, "step": 18199 }, { "epoch": 2.17, "learning_rate": 9.436819650316797e-07, "loss": 0.5809, "step": 18200 }, { "epoch": 2.17, "learning_rate": 9.434304325091065e-07, "loss": 0.3399, "step": 18201 }, { "epoch": 2.17, "learning_rate": 9.43178925716759e-07, "loss": 0.4643, "step": 18202 }, { "epoch": 2.17, "learning_rate": 9.429274446587936e-07, "loss": 0.3509, "step": 18203 }, { "epoch": 2.17, "learning_rate": 9.42675989339368e-07, "loss": 0.4548, "step": 18204 }, { "epoch": 2.17, "learning_rate": 9.4242455976264e-07, "loss": 0.6061, "step": 18205 }, { "epoch": 2.17, "learning_rate": 9.421731559327638e-07, "loss": 0.576, "step": 18206 }, { "epoch": 2.17, "learning_rate": 9.419217778538961e-07, "loss": 0.3468, "step": 18207 }, { "epoch": 2.17, "learning_rate": 9.416704255301929e-07, "loss": 0.3245, "step": 18208 }, { "epoch": 2.17, "learning_rate": 9.414190989658084e-07, "loss": 0.3794, "step": 18209 }, { "epoch": 2.17, "learning_rate": 9.411677981648962e-07, "loss": 0.5253, "step": 18210 }, { "epoch": 2.17, "learning_rate": 9.409165231316109e-07, "loss": 0.5821, "step": 18211 }, { "epoch": 2.17, "learning_rate": 9.406652738701072e-07, "loss": 0.6131, "step": 18212 }, { "epoch": 2.17, "learning_rate": 9.404140503845366e-07, "loss": 0.4257, "step": 18213 }, { "epoch": 2.17, "learning_rate": 9.401628526790524e-07, "loss": 0.3338, "step": 18214 }, { "epoch": 2.17, "learning_rate": 9.399116807578082e-07, "loss": 0.4175, "step": 18215 }, { "epoch": 2.17, "learning_rate": 9.396605346249548e-07, "loss": 0.4788, "step": 18216 }, { "epoch": 2.17, "learning_rate": 9.394094142846427e-07, "loss": 0.4982, "step": 18217 }, { "epoch": 2.17, "learning_rate": 9.391583197410243e-07, "loss": 0.6628, "step": 18218 }, { "epoch": 2.17, "learning_rate": 9.389072509982495e-07, "loss": 0.5232, "step": 18219 }, { "epoch": 2.17, "learning_rate": 9.386562080604697e-07, "loss": 0.3289, "step": 18220 }, { "epoch": 2.17, "learning_rate": 9.384051909318329e-07, "loss": 0.454, "step": 18221 }, { "epoch": 2.17, "learning_rate": 9.381541996164895e-07, "loss": 0.4403, "step": 18222 }, { "epoch": 2.17, "learning_rate": 9.379032341185887e-07, "loss": 0.4915, "step": 18223 }, { "epoch": 2.17, "learning_rate": 9.376522944422789e-07, "loss": 0.6525, "step": 18224 }, { "epoch": 2.17, "learning_rate": 9.374013805917068e-07, "loss": 0.5756, "step": 18225 }, { "epoch": 2.17, "learning_rate": 9.371504925710209e-07, "loss": 0.3666, "step": 18226 }, { "epoch": 2.17, "learning_rate": 9.368996303843686e-07, "loss": 0.4249, "step": 18227 }, { "epoch": 2.17, "learning_rate": 9.366487940358973e-07, "loss": 0.4778, "step": 18228 }, { "epoch": 2.17, "learning_rate": 9.363979835297521e-07, "loss": 0.4944, "step": 18229 }, { "epoch": 2.17, "learning_rate": 9.361471988700799e-07, "loss": 0.5721, "step": 18230 }, { "epoch": 2.17, "learning_rate": 9.35896440061026e-07, "loss": 0.528, "step": 18231 }, { "epoch": 2.17, "learning_rate": 9.356457071067341e-07, "loss": 0.3463, "step": 18232 }, { "epoch": 2.17, "learning_rate": 9.353950000113499e-07, "loss": 0.3876, "step": 18233 }, { "epoch": 2.17, "learning_rate": 9.35144318779018e-07, "loss": 0.3679, "step": 18234 }, { "epoch": 2.17, "learning_rate": 9.348936634138822e-07, "loss": 0.4447, "step": 18235 }, { "epoch": 2.17, "learning_rate": 9.346430339200848e-07, "loss": 0.4785, "step": 18236 }, { "epoch": 2.17, "learning_rate": 9.343924303017704e-07, "loss": 0.5326, "step": 18237 }, { "epoch": 2.17, "learning_rate": 9.341418525630794e-07, "loss": 0.3085, "step": 18238 }, { "epoch": 2.17, "learning_rate": 9.33891300708156e-07, "loss": 0.3749, "step": 18239 }, { "epoch": 2.17, "learning_rate": 9.3364077474114e-07, "loss": 0.3319, "step": 18240 }, { "epoch": 2.17, "learning_rate": 9.33390274666173e-07, "loss": 0.5118, "step": 18241 }, { "epoch": 2.17, "learning_rate": 9.331398004873965e-07, "loss": 0.5321, "step": 18242 }, { "epoch": 2.17, "learning_rate": 9.328893522089516e-07, "loss": 0.6513, "step": 18243 }, { "epoch": 2.17, "learning_rate": 9.32638929834977e-07, "loss": 0.6347, "step": 18244 }, { "epoch": 2.17, "learning_rate": 9.323885333696115e-07, "loss": 0.423, "step": 18245 }, { "epoch": 2.17, "learning_rate": 9.321381628169962e-07, "loss": 0.4109, "step": 18246 }, { "epoch": 2.17, "learning_rate": 9.318878181812674e-07, "loss": 0.3943, "step": 18247 }, { "epoch": 2.17, "learning_rate": 9.316374994665648e-07, "loss": 0.5112, "step": 18248 }, { "epoch": 2.17, "learning_rate": 9.313872066770258e-07, "loss": 0.5272, "step": 18249 }, { "epoch": 2.17, "learning_rate": 9.311369398167888e-07, "loss": 0.5437, "step": 18250 }, { "epoch": 2.17, "learning_rate": 9.308866988899889e-07, "loss": 0.2912, "step": 18251 }, { "epoch": 2.17, "learning_rate": 9.306364839007645e-07, "loss": 0.4225, "step": 18252 }, { "epoch": 2.17, "learning_rate": 9.303862948532499e-07, "loss": 0.4295, "step": 18253 }, { "epoch": 2.17, "learning_rate": 9.301361317515823e-07, "loss": 0.4514, "step": 18254 }, { "epoch": 2.17, "learning_rate": 9.298859945998953e-07, "loss": 0.5906, "step": 18255 }, { "epoch": 2.17, "learning_rate": 9.296358834023244e-07, "loss": 0.4598, "step": 18256 }, { "epoch": 2.17, "learning_rate": 9.293857981630045e-07, "loss": 0.3971, "step": 18257 }, { "epoch": 2.17, "learning_rate": 9.291357388860697e-07, "loss": 0.3462, "step": 18258 }, { "epoch": 2.17, "learning_rate": 9.288857055756528e-07, "loss": 0.3311, "step": 18259 }, { "epoch": 2.17, "learning_rate": 9.286356982358866e-07, "loss": 0.4478, "step": 18260 }, { "epoch": 2.17, "learning_rate": 9.283857168709046e-07, "loss": 0.5114, "step": 18261 }, { "epoch": 2.17, "learning_rate": 9.281357614848379e-07, "loss": 0.7018, "step": 18262 }, { "epoch": 2.17, "learning_rate": 9.27885832081819e-07, "loss": 0.3714, "step": 18263 }, { "epoch": 2.18, "learning_rate": 9.276359286659791e-07, "loss": 0.3335, "step": 18264 }, { "epoch": 2.18, "learning_rate": 9.2738605124145e-07, "loss": 0.357, "step": 18265 }, { "epoch": 2.18, "learning_rate": 9.271361998123618e-07, "loss": 0.4746, "step": 18266 }, { "epoch": 2.18, "learning_rate": 9.26886374382843e-07, "loss": 0.4375, "step": 18267 }, { "epoch": 2.18, "learning_rate": 9.266365749570247e-07, "loss": 0.6438, "step": 18268 }, { "epoch": 2.18, "learning_rate": 9.263868015390365e-07, "loss": 0.4511, "step": 18269 }, { "epoch": 2.18, "learning_rate": 9.261370541330056e-07, "loss": 0.2891, "step": 18270 }, { "epoch": 2.18, "learning_rate": 9.258873327430612e-07, "loss": 0.3067, "step": 18271 }, { "epoch": 2.18, "learning_rate": 9.256376373733322e-07, "loss": 0.4032, "step": 18272 }, { "epoch": 2.18, "learning_rate": 9.253879680279454e-07, "loss": 0.5505, "step": 18273 }, { "epoch": 2.18, "learning_rate": 9.251383247110265e-07, "loss": 0.6745, "step": 18274 }, { "epoch": 2.18, "learning_rate": 9.248887074267033e-07, "loss": 0.559, "step": 18275 }, { "epoch": 2.18, "learning_rate": 9.246391161791027e-07, "loss": 0.3348, "step": 18276 }, { "epoch": 2.18, "learning_rate": 9.243895509723488e-07, "loss": 0.368, "step": 18277 }, { "epoch": 2.18, "learning_rate": 9.241400118105681e-07, "loss": 0.5065, "step": 18278 }, { "epoch": 2.18, "learning_rate": 9.23890498697885e-07, "loss": 0.4851, "step": 18279 }, { "epoch": 2.18, "learning_rate": 9.236410116384251e-07, "loss": 0.6761, "step": 18280 }, { "epoch": 2.18, "learning_rate": 9.233915506363117e-07, "loss": 0.612, "step": 18281 }, { "epoch": 2.18, "learning_rate": 9.231421156956671e-07, "loss": 0.3922, "step": 18282 }, { "epoch": 2.18, "learning_rate": 9.228927068206159e-07, "loss": 0.4209, "step": 18283 }, { "epoch": 2.18, "learning_rate": 9.226433240152813e-07, "loss": 0.3821, "step": 18284 }, { "epoch": 2.18, "learning_rate": 9.223939672837842e-07, "loss": 0.4561, "step": 18285 }, { "epoch": 2.18, "learning_rate": 9.22144636630247e-07, "loss": 0.5645, "step": 18286 }, { "epoch": 2.18, "learning_rate": 9.218953320587923e-07, "loss": 0.6224, "step": 18287 }, { "epoch": 2.18, "learning_rate": 9.216460535735403e-07, "loss": 0.4271, "step": 18288 }, { "epoch": 2.18, "learning_rate": 9.213968011786104e-07, "loss": 0.3912, "step": 18289 }, { "epoch": 2.18, "learning_rate": 9.21147574878124e-07, "loss": 0.4133, "step": 18290 }, { "epoch": 2.18, "learning_rate": 9.208983746762005e-07, "loss": 0.5069, "step": 18291 }, { "epoch": 2.18, "learning_rate": 9.206492005769604e-07, "loss": 0.4584, "step": 18292 }, { "epoch": 2.18, "learning_rate": 9.204000525845205e-07, "loss": 0.672, "step": 18293 }, { "epoch": 2.18, "learning_rate": 9.201509307030012e-07, "loss": 0.5885, "step": 18294 }, { "epoch": 2.18, "learning_rate": 9.199018349365196e-07, "loss": 0.3475, "step": 18295 }, { "epoch": 2.18, "learning_rate": 9.196527652891923e-07, "loss": 0.4355, "step": 18296 }, { "epoch": 2.18, "learning_rate": 9.194037217651374e-07, "loss": 0.3798, "step": 18297 }, { "epoch": 2.18, "learning_rate": 9.191547043684718e-07, "loss": 0.5323, "step": 18298 }, { "epoch": 2.18, "learning_rate": 9.189057131033122e-07, "loss": 0.5242, "step": 18299 }, { "epoch": 2.18, "learning_rate": 9.186567479737732e-07, "loss": 0.5443, "step": 18300 }, { "epoch": 2.18, "learning_rate": 9.184078089839715e-07, "loss": 0.3051, "step": 18301 }, { "epoch": 2.18, "learning_rate": 9.181588961380206e-07, "loss": 0.3644, "step": 18302 }, { "epoch": 2.18, "learning_rate": 9.17910009440037e-07, "loss": 0.431, "step": 18303 }, { "epoch": 2.18, "learning_rate": 9.176611488941325e-07, "loss": 0.5069, "step": 18304 }, { "epoch": 2.18, "learning_rate": 9.17412314504422e-07, "loss": 0.5511, "step": 18305 }, { "epoch": 2.18, "learning_rate": 9.171635062750189e-07, "loss": 0.5529, "step": 18306 }, { "epoch": 2.18, "learning_rate": 9.169147242100365e-07, "loss": 0.2703, "step": 18307 }, { "epoch": 2.18, "learning_rate": 9.166659683135859e-07, "loss": 0.3595, "step": 18308 }, { "epoch": 2.18, "learning_rate": 9.164172385897804e-07, "loss": 0.4835, "step": 18309 }, { "epoch": 2.18, "learning_rate": 9.16168535042731e-07, "loss": 0.4666, "step": 18310 }, { "epoch": 2.18, "learning_rate": 9.159198576765477e-07, "loss": 0.5612, "step": 18311 }, { "epoch": 2.18, "learning_rate": 9.156712064953418e-07, "loss": 0.5906, "step": 18312 }, { "epoch": 2.18, "learning_rate": 9.154225815032242e-07, "loss": 0.2845, "step": 18313 }, { "epoch": 2.18, "learning_rate": 9.151739827043049e-07, "loss": 0.313, "step": 18314 }, { "epoch": 2.18, "learning_rate": 9.149254101026916e-07, "loss": 0.3608, "step": 18315 }, { "epoch": 2.18, "learning_rate": 9.146768637024955e-07, "loss": 0.3897, "step": 18316 }, { "epoch": 2.18, "learning_rate": 9.14428343507823e-07, "loss": 0.4009, "step": 18317 }, { "epoch": 2.18, "learning_rate": 9.141798495227836e-07, "loss": 0.6577, "step": 18318 }, { "epoch": 2.18, "learning_rate": 9.139313817514836e-07, "loss": 0.5315, "step": 18319 }, { "epoch": 2.18, "learning_rate": 9.13682940198031e-07, "loss": 0.3389, "step": 18320 }, { "epoch": 2.18, "learning_rate": 9.134345248665325e-07, "loss": 0.3723, "step": 18321 }, { "epoch": 2.18, "learning_rate": 9.131861357610952e-07, "loss": 0.4161, "step": 18322 }, { "epoch": 2.18, "learning_rate": 9.129377728858244e-07, "loss": 0.5242, "step": 18323 }, { "epoch": 2.18, "learning_rate": 9.126894362448244e-07, "loss": 0.5691, "step": 18324 }, { "epoch": 2.18, "learning_rate": 9.124411258422019e-07, "loss": 0.6201, "step": 18325 }, { "epoch": 2.18, "learning_rate": 9.121928416820603e-07, "loss": 0.3844, "step": 18326 }, { "epoch": 2.18, "learning_rate": 9.119445837685039e-07, "loss": 0.4112, "step": 18327 }, { "epoch": 2.18, "learning_rate": 9.116963521056369e-07, "loss": 0.4069, "step": 18328 }, { "epoch": 2.18, "learning_rate": 9.114481466975633e-07, "loss": 0.5682, "step": 18329 }, { "epoch": 2.18, "learning_rate": 9.111999675483854e-07, "loss": 0.574, "step": 18330 }, { "epoch": 2.18, "learning_rate": 9.109518146622043e-07, "loss": 0.5176, "step": 18331 }, { "epoch": 2.18, "learning_rate": 9.10703688043123e-07, "loss": 0.3114, "step": 18332 }, { "epoch": 2.18, "learning_rate": 9.10455587695244e-07, "loss": 0.3536, "step": 18333 }, { "epoch": 2.18, "learning_rate": 9.102075136226665e-07, "loss": 0.4395, "step": 18334 }, { "epoch": 2.18, "learning_rate": 9.099594658294922e-07, "loss": 0.44, "step": 18335 }, { "epoch": 2.18, "learning_rate": 9.097114443198216e-07, "loss": 0.5367, "step": 18336 }, { "epoch": 2.18, "learning_rate": 9.094634490977549e-07, "loss": 0.6034, "step": 18337 }, { "epoch": 2.18, "learning_rate": 9.092154801673911e-07, "loss": 0.3268, "step": 18338 }, { "epoch": 2.18, "learning_rate": 9.089675375328277e-07, "loss": 0.3079, "step": 18339 }, { "epoch": 2.18, "learning_rate": 9.087196211981656e-07, "loss": 0.3522, "step": 18340 }, { "epoch": 2.18, "learning_rate": 9.084717311675004e-07, "loss": 0.4216, "step": 18341 }, { "epoch": 2.18, "learning_rate": 9.082238674449312e-07, "loss": 0.4518, "step": 18342 }, { "epoch": 2.18, "learning_rate": 9.079760300345552e-07, "loss": 0.5903, "step": 18343 }, { "epoch": 2.18, "learning_rate": 9.077282189404699e-07, "loss": 0.469, "step": 18344 }, { "epoch": 2.18, "learning_rate": 9.074804341667704e-07, "loss": 0.2503, "step": 18345 }, { "epoch": 2.18, "learning_rate": 9.072326757175523e-07, "loss": 0.3809, "step": 18346 }, { "epoch": 2.18, "learning_rate": 9.069849435969116e-07, "loss": 0.5222, "step": 18347 }, { "epoch": 2.19, "learning_rate": 9.067372378089445e-07, "loss": 0.5486, "step": 18348 }, { "epoch": 2.19, "learning_rate": 9.064895583577435e-07, "loss": 0.6136, "step": 18349 }, { "epoch": 2.19, "learning_rate": 9.062419052474039e-07, "loss": 0.6426, "step": 18350 }, { "epoch": 2.19, "learning_rate": 9.059942784820203e-07, "loss": 0.3585, "step": 18351 }, { "epoch": 2.19, "learning_rate": 9.057466780656849e-07, "loss": 0.3857, "step": 18352 }, { "epoch": 2.19, "learning_rate": 9.054991040024899e-07, "loss": 0.4633, "step": 18353 }, { "epoch": 2.19, "learning_rate": 9.052515562965283e-07, "loss": 0.4702, "step": 18354 }, { "epoch": 2.19, "learning_rate": 9.050040349518934e-07, "loss": 0.542, "step": 18355 }, { "epoch": 2.19, "learning_rate": 9.047565399726746e-07, "loss": 0.5439, "step": 18356 }, { "epoch": 2.19, "learning_rate": 9.045090713629643e-07, "loss": 0.3229, "step": 18357 }, { "epoch": 2.19, "learning_rate": 9.042616291268536e-07, "loss": 0.4052, "step": 18358 }, { "epoch": 2.19, "learning_rate": 9.040142132684318e-07, "loss": 0.4889, "step": 18359 }, { "epoch": 2.19, "learning_rate": 9.037668237917885e-07, "loss": 0.5238, "step": 18360 }, { "epoch": 2.19, "learning_rate": 9.035194607010133e-07, "loss": 0.5673, "step": 18361 }, { "epoch": 2.19, "learning_rate": 9.032721240001954e-07, "loss": 0.6699, "step": 18362 }, { "epoch": 2.19, "learning_rate": 9.030248136934245e-07, "loss": 0.3717, "step": 18363 }, { "epoch": 2.19, "learning_rate": 9.02777529784786e-07, "loss": 0.3394, "step": 18364 }, { "epoch": 2.19, "learning_rate": 9.025302722783694e-07, "loss": 0.3953, "step": 18365 }, { "epoch": 2.19, "learning_rate": 9.022830411782621e-07, "loss": 0.4445, "step": 18366 }, { "epoch": 2.19, "learning_rate": 9.020358364885501e-07, "loss": 0.5326, "step": 18367 }, { "epoch": 2.19, "learning_rate": 9.017886582133189e-07, "loss": 0.6525, "step": 18368 }, { "epoch": 2.19, "learning_rate": 9.015415063566555e-07, "loss": 0.5264, "step": 18369 }, { "epoch": 2.19, "learning_rate": 9.012943809226451e-07, "loss": 0.2997, "step": 18370 }, { "epoch": 2.19, "learning_rate": 9.010472819153734e-07, "loss": 0.4866, "step": 18371 }, { "epoch": 2.19, "learning_rate": 9.008002093389238e-07, "loss": 0.4348, "step": 18372 }, { "epoch": 2.19, "learning_rate": 9.005531631973813e-07, "loss": 0.4398, "step": 18373 }, { "epoch": 2.19, "learning_rate": 9.003061434948296e-07, "loss": 0.752, "step": 18374 }, { "epoch": 2.19, "learning_rate": 9.000591502353507e-07, "loss": 0.6639, "step": 18375 }, { "epoch": 2.19, "learning_rate": 8.99812183423028e-07, "loss": 0.2701, "step": 18376 }, { "epoch": 2.19, "learning_rate": 8.995652430619445e-07, "loss": 0.4718, "step": 18377 }, { "epoch": 2.19, "learning_rate": 8.993183291561824e-07, "loss": 0.4753, "step": 18378 }, { "epoch": 2.19, "learning_rate": 8.99071441709822e-07, "loss": 0.498, "step": 18379 }, { "epoch": 2.19, "learning_rate": 8.988245807269455e-07, "loss": 0.6023, "step": 18380 }, { "epoch": 2.19, "learning_rate": 8.985777462116327e-07, "loss": 0.5592, "step": 18381 }, { "epoch": 2.19, "learning_rate": 8.983309381679644e-07, "loss": 0.4002, "step": 18382 }, { "epoch": 2.19, "learning_rate": 8.980841566000198e-07, "loss": 0.4272, "step": 18383 }, { "epoch": 2.19, "learning_rate": 8.978374015118783e-07, "loss": 0.4342, "step": 18384 }, { "epoch": 2.19, "learning_rate": 8.975906729076186e-07, "loss": 0.4959, "step": 18385 }, { "epoch": 2.19, "learning_rate": 8.97343970791321e-07, "loss": 0.5097, "step": 18386 }, { "epoch": 2.19, "learning_rate": 8.970972951670615e-07, "loss": 0.5353, "step": 18387 }, { "epoch": 2.19, "learning_rate": 8.968506460389176e-07, "loss": 0.3733, "step": 18388 }, { "epoch": 2.19, "learning_rate": 8.96604023410968e-07, "loss": 0.2961, "step": 18389 }, { "epoch": 2.19, "learning_rate": 8.963574272872874e-07, "loss": 0.4448, "step": 18390 }, { "epoch": 2.19, "learning_rate": 8.961108576719532e-07, "loss": 0.4638, "step": 18391 }, { "epoch": 2.19, "learning_rate": 8.95864314569041e-07, "loss": 0.5721, "step": 18392 }, { "epoch": 2.19, "learning_rate": 8.956177979826272e-07, "loss": 0.6224, "step": 18393 }, { "epoch": 2.19, "learning_rate": 8.95371307916785e-07, "loss": 0.5464, "step": 18394 }, { "epoch": 2.19, "learning_rate": 8.951248443755903e-07, "loss": 0.3627, "step": 18395 }, { "epoch": 2.19, "learning_rate": 8.948784073631162e-07, "loss": 0.351, "step": 18396 }, { "epoch": 2.19, "learning_rate": 8.946319968834372e-07, "loss": 0.4109, "step": 18397 }, { "epoch": 2.19, "learning_rate": 8.943856129406253e-07, "loss": 0.5759, "step": 18398 }, { "epoch": 2.19, "learning_rate": 8.941392555387538e-07, "loss": 0.5886, "step": 18399 }, { "epoch": 2.19, "learning_rate": 8.938929246818953e-07, "loss": 0.5684, "step": 18400 }, { "epoch": 2.19, "learning_rate": 8.936466203741223e-07, "loss": 0.3158, "step": 18401 }, { "epoch": 2.19, "learning_rate": 8.934003426195054e-07, "loss": 0.3758, "step": 18402 }, { "epoch": 2.19, "learning_rate": 8.931540914221148e-07, "loss": 0.5382, "step": 18403 }, { "epoch": 2.19, "learning_rate": 8.929078667860228e-07, "loss": 0.5074, "step": 18404 }, { "epoch": 2.19, "learning_rate": 8.926616687152978e-07, "loss": 0.5899, "step": 18405 }, { "epoch": 2.19, "learning_rate": 8.924154972140101e-07, "loss": 0.6448, "step": 18406 }, { "epoch": 2.19, "learning_rate": 8.921693522862293e-07, "loss": 0.3054, "step": 18407 }, { "epoch": 2.19, "learning_rate": 8.919232339360249e-07, "loss": 0.4604, "step": 18408 }, { "epoch": 2.19, "learning_rate": 8.916771421674642e-07, "loss": 0.3848, "step": 18409 }, { "epoch": 2.19, "learning_rate": 8.914310769846143e-07, "loss": 0.5151, "step": 18410 }, { "epoch": 2.19, "learning_rate": 8.91185038391544e-07, "loss": 0.4889, "step": 18411 }, { "epoch": 2.19, "learning_rate": 8.909390263923209e-07, "loss": 0.5468, "step": 18412 }, { "epoch": 2.19, "learning_rate": 8.906930409910098e-07, "loss": 0.3643, "step": 18413 }, { "epoch": 2.19, "learning_rate": 8.904470821916775e-07, "loss": 0.3887, "step": 18414 }, { "epoch": 2.19, "learning_rate": 8.90201149998391e-07, "loss": 0.3747, "step": 18415 }, { "epoch": 2.19, "learning_rate": 8.899552444152146e-07, "loss": 0.4335, "step": 18416 }, { "epoch": 2.19, "learning_rate": 8.897093654462122e-07, "loss": 0.5062, "step": 18417 }, { "epoch": 2.19, "learning_rate": 8.894635130954491e-07, "loss": 0.5498, "step": 18418 }, { "epoch": 2.19, "learning_rate": 8.8921768736699e-07, "loss": 0.65, "step": 18419 }, { "epoch": 2.19, "learning_rate": 8.88971888264897e-07, "loss": 0.3433, "step": 18420 }, { "epoch": 2.19, "learning_rate": 8.887261157932339e-07, "loss": 0.3873, "step": 18421 }, { "epoch": 2.19, "learning_rate": 8.884803699560632e-07, "loss": 0.501, "step": 18422 }, { "epoch": 2.19, "learning_rate": 8.882346507574479e-07, "loss": 0.4877, "step": 18423 }, { "epoch": 2.19, "learning_rate": 8.879889582014495e-07, "loss": 0.6743, "step": 18424 }, { "epoch": 2.19, "learning_rate": 8.877432922921278e-07, "loss": 0.5332, "step": 18425 }, { "epoch": 2.19, "learning_rate": 8.874976530335447e-07, "loss": 0.4674, "step": 18426 }, { "epoch": 2.19, "learning_rate": 8.872520404297616e-07, "loss": 0.4423, "step": 18427 }, { "epoch": 2.19, "learning_rate": 8.870064544848367e-07, "loss": 0.4465, "step": 18428 }, { "epoch": 2.19, "learning_rate": 8.867608952028303e-07, "loss": 0.3828, "step": 18429 }, { "epoch": 2.19, "learning_rate": 8.865153625878029e-07, "loss": 0.5747, "step": 18430 }, { "epoch": 2.19, "learning_rate": 8.862698566438115e-07, "loss": 0.5341, "step": 18431 }, { "epoch": 2.2, "learning_rate": 8.860243773749141e-07, "loss": 0.2967, "step": 18432 }, { "epoch": 2.2, "learning_rate": 8.857789247851695e-07, "loss": 0.3739, "step": 18433 }, { "epoch": 2.2, "learning_rate": 8.855334988786344e-07, "loss": 0.3699, "step": 18434 }, { "epoch": 2.2, "learning_rate": 8.852880996593668e-07, "loss": 0.5208, "step": 18435 }, { "epoch": 2.2, "learning_rate": 8.850427271314216e-07, "loss": 0.4973, "step": 18436 }, { "epoch": 2.2, "learning_rate": 8.847973812988566e-07, "loss": 0.6089, "step": 18437 }, { "epoch": 2.2, "learning_rate": 8.845520621657264e-07, "loss": 0.3608, "step": 18438 }, { "epoch": 2.2, "learning_rate": 8.843067697360852e-07, "loss": 0.2832, "step": 18439 }, { "epoch": 2.2, "learning_rate": 8.840615040139888e-07, "loss": 0.3625, "step": 18440 }, { "epoch": 2.2, "learning_rate": 8.838162650034912e-07, "loss": 0.3787, "step": 18441 }, { "epoch": 2.2, "learning_rate": 8.835710527086472e-07, "loss": 0.5738, "step": 18442 }, { "epoch": 2.2, "learning_rate": 8.833258671335085e-07, "loss": 0.5575, "step": 18443 }, { "epoch": 2.2, "learning_rate": 8.830807082821297e-07, "loss": 0.5886, "step": 18444 }, { "epoch": 2.2, "learning_rate": 8.828355761585616e-07, "loss": 0.2893, "step": 18445 }, { "epoch": 2.2, "learning_rate": 8.825904707668581e-07, "loss": 0.3424, "step": 18446 }, { "epoch": 2.2, "learning_rate": 8.823453921110689e-07, "loss": 0.4153, "step": 18447 }, { "epoch": 2.2, "learning_rate": 8.82100340195246e-07, "loss": 0.4875, "step": 18448 }, { "epoch": 2.2, "learning_rate": 8.818553150234405e-07, "loss": 0.6622, "step": 18449 }, { "epoch": 2.2, "learning_rate": 8.816103165997031e-07, "loss": 0.5201, "step": 18450 }, { "epoch": 2.2, "learning_rate": 8.81365344928082e-07, "loss": 0.2884, "step": 18451 }, { "epoch": 2.2, "learning_rate": 8.811204000126286e-07, "loss": 0.419, "step": 18452 }, { "epoch": 2.2, "learning_rate": 8.808754818573909e-07, "loss": 0.4184, "step": 18453 }, { "epoch": 2.2, "learning_rate": 8.806305904664161e-07, "loss": 0.4789, "step": 18454 }, { "epoch": 2.2, "learning_rate": 8.803857258437537e-07, "loss": 0.6238, "step": 18455 }, { "epoch": 2.2, "learning_rate": 8.801408879934514e-07, "loss": 0.5214, "step": 18456 }, { "epoch": 2.2, "learning_rate": 8.798960769195566e-07, "loss": 0.3187, "step": 18457 }, { "epoch": 2.2, "learning_rate": 8.796512926261149e-07, "loss": 0.4122, "step": 18458 }, { "epoch": 2.2, "learning_rate": 8.794065351171743e-07, "loss": 0.4254, "step": 18459 }, { "epoch": 2.2, "learning_rate": 8.791618043967786e-07, "loss": 0.5133, "step": 18460 }, { "epoch": 2.2, "learning_rate": 8.789171004689753e-07, "loss": 0.7674, "step": 18461 }, { "epoch": 2.2, "learning_rate": 8.786724233378074e-07, "loss": 0.6016, "step": 18462 }, { "epoch": 2.2, "learning_rate": 8.784277730073204e-07, "loss": 0.3741, "step": 18463 }, { "epoch": 2.2, "learning_rate": 8.781831494815582e-07, "loss": 0.3332, "step": 18464 }, { "epoch": 2.2, "learning_rate": 8.779385527645656e-07, "loss": 0.4478, "step": 18465 }, { "epoch": 2.2, "learning_rate": 8.77693982860385e-07, "loss": 0.3809, "step": 18466 }, { "epoch": 2.2, "learning_rate": 8.774494397730579e-07, "loss": 0.4471, "step": 18467 }, { "epoch": 2.2, "learning_rate": 8.772049235066285e-07, "loss": 0.5883, "step": 18468 }, { "epoch": 2.2, "learning_rate": 8.769604340651372e-07, "loss": 0.3823, "step": 18469 }, { "epoch": 2.2, "learning_rate": 8.767159714526261e-07, "loss": 0.3888, "step": 18470 }, { "epoch": 2.2, "learning_rate": 8.764715356731362e-07, "loss": 0.3846, "step": 18471 }, { "epoch": 2.2, "learning_rate": 8.762271267307088e-07, "loss": 0.422, "step": 18472 }, { "epoch": 2.2, "learning_rate": 8.759827446293834e-07, "loss": 0.5281, "step": 18473 }, { "epoch": 2.2, "learning_rate": 8.757383893731985e-07, "loss": 0.5753, "step": 18474 }, { "epoch": 2.2, "learning_rate": 8.754940609661944e-07, "loss": 0.614, "step": 18475 }, { "epoch": 2.2, "learning_rate": 8.752497594124107e-07, "loss": 0.3454, "step": 18476 }, { "epoch": 2.2, "learning_rate": 8.750054847158837e-07, "loss": 0.3331, "step": 18477 }, { "epoch": 2.2, "learning_rate": 8.747612368806524e-07, "loss": 0.4247, "step": 18478 }, { "epoch": 2.2, "learning_rate": 8.745170159107544e-07, "loss": 0.4468, "step": 18479 }, { "epoch": 2.2, "learning_rate": 8.742728218102273e-07, "loss": 0.5948, "step": 18480 }, { "epoch": 2.2, "learning_rate": 8.74028654583107e-07, "loss": 0.6262, "step": 18481 }, { "epoch": 2.2, "learning_rate": 8.737845142334284e-07, "loss": 0.3359, "step": 18482 }, { "epoch": 2.2, "learning_rate": 8.735404007652293e-07, "loss": 0.3727, "step": 18483 }, { "epoch": 2.2, "learning_rate": 8.732963141825429e-07, "loss": 0.4084, "step": 18484 }, { "epoch": 2.2, "learning_rate": 8.730522544894049e-07, "loss": 0.3769, "step": 18485 }, { "epoch": 2.2, "learning_rate": 8.728082216898496e-07, "loss": 0.4755, "step": 18486 }, { "epoch": 2.2, "learning_rate": 8.725642157879119e-07, "loss": 0.5481, "step": 18487 }, { "epoch": 2.2, "learning_rate": 8.723202367876241e-07, "loss": 0.4696, "step": 18488 }, { "epoch": 2.2, "learning_rate": 8.720762846930187e-07, "loss": 0.3017, "step": 18489 }, { "epoch": 2.2, "learning_rate": 8.718323595081287e-07, "loss": 0.3623, "step": 18490 }, { "epoch": 2.2, "learning_rate": 8.715884612369873e-07, "loss": 0.4759, "step": 18491 }, { "epoch": 2.2, "learning_rate": 8.713445898836245e-07, "loss": 0.5277, "step": 18492 }, { "epoch": 2.2, "learning_rate": 8.711007454520722e-07, "loss": 0.7135, "step": 18493 }, { "epoch": 2.2, "learning_rate": 8.708569279463622e-07, "loss": 0.5534, "step": 18494 }, { "epoch": 2.2, "learning_rate": 8.70613137370524e-07, "loss": 0.3062, "step": 18495 }, { "epoch": 2.2, "learning_rate": 8.703693737285862e-07, "loss": 0.4556, "step": 18496 }, { "epoch": 2.2, "learning_rate": 8.701256370245797e-07, "loss": 0.4533, "step": 18497 }, { "epoch": 2.2, "learning_rate": 8.698819272625331e-07, "loss": 0.4395, "step": 18498 }, { "epoch": 2.2, "learning_rate": 8.696382444464758e-07, "loss": 0.5464, "step": 18499 }, { "epoch": 2.2, "learning_rate": 8.693945885804341e-07, "loss": 0.7139, "step": 18500 }, { "epoch": 2.2, "learning_rate": 8.691509596684378e-07, "loss": 0.3249, "step": 18501 }, { "epoch": 2.2, "learning_rate": 8.689073577145127e-07, "loss": 0.3811, "step": 18502 }, { "epoch": 2.2, "learning_rate": 8.686637827226852e-07, "loss": 0.4196, "step": 18503 }, { "epoch": 2.2, "learning_rate": 8.684202346969823e-07, "loss": 0.5153, "step": 18504 }, { "epoch": 2.2, "learning_rate": 8.681767136414296e-07, "loss": 0.5718, "step": 18505 }, { "epoch": 2.2, "learning_rate": 8.679332195600537e-07, "loss": 0.6078, "step": 18506 }, { "epoch": 2.2, "learning_rate": 8.676897524568778e-07, "loss": 0.2837, "step": 18507 }, { "epoch": 2.2, "learning_rate": 8.674463123359272e-07, "loss": 0.343, "step": 18508 }, { "epoch": 2.2, "learning_rate": 8.672028992012268e-07, "loss": 0.412, "step": 18509 }, { "epoch": 2.2, "learning_rate": 8.669595130567998e-07, "loss": 0.3974, "step": 18510 }, { "epoch": 2.2, "learning_rate": 8.667161539066679e-07, "loss": 0.5403, "step": 18511 }, { "epoch": 2.2, "learning_rate": 8.664728217548554e-07, "loss": 0.534, "step": 18512 }, { "epoch": 2.2, "learning_rate": 8.662295166053839e-07, "loss": 0.418, "step": 18513 }, { "epoch": 2.2, "learning_rate": 8.659862384622764e-07, "loss": 0.3124, "step": 18514 }, { "epoch": 2.2, "learning_rate": 8.657429873295528e-07, "loss": 0.3502, "step": 18515 }, { "epoch": 2.21, "learning_rate": 8.654997632112355e-07, "loss": 0.4169, "step": 18516 }, { "epoch": 2.21, "learning_rate": 8.652565661113444e-07, "loss": 0.5027, "step": 18517 }, { "epoch": 2.21, "learning_rate": 8.650133960338982e-07, "loss": 0.5529, "step": 18518 }, { "epoch": 2.21, "learning_rate": 8.647702529829183e-07, "loss": 0.6107, "step": 18519 }, { "epoch": 2.21, "learning_rate": 8.645271369624228e-07, "loss": 0.3287, "step": 18520 }, { "epoch": 2.21, "learning_rate": 8.642840479764322e-07, "loss": 0.4157, "step": 18521 }, { "epoch": 2.21, "learning_rate": 8.640409860289625e-07, "loss": 0.3844, "step": 18522 }, { "epoch": 2.21, "learning_rate": 8.637979511240335e-07, "loss": 0.5056, "step": 18523 }, { "epoch": 2.21, "learning_rate": 8.635549432656607e-07, "loss": 0.4933, "step": 18524 }, { "epoch": 2.21, "learning_rate": 8.633119624578629e-07, "loss": 0.4969, "step": 18525 }, { "epoch": 2.21, "learning_rate": 8.630690087046547e-07, "loss": 0.28, "step": 18526 }, { "epoch": 2.21, "learning_rate": 8.628260820100534e-07, "loss": 0.3995, "step": 18527 }, { "epoch": 2.21, "learning_rate": 8.625831823780742e-07, "loss": 0.3958, "step": 18528 }, { "epoch": 2.21, "learning_rate": 8.623403098127334e-07, "loss": 0.4809, "step": 18529 }, { "epoch": 2.21, "learning_rate": 8.620974643180443e-07, "loss": 0.5887, "step": 18530 }, { "epoch": 2.21, "learning_rate": 8.618546458980209e-07, "loss": 0.5522, "step": 18531 }, { "epoch": 2.21, "learning_rate": 8.616118545566787e-07, "loss": 0.3941, "step": 18532 }, { "epoch": 2.21, "learning_rate": 8.61369090298029e-07, "loss": 0.3678, "step": 18533 }, { "epoch": 2.21, "learning_rate": 8.611263531260858e-07, "loss": 0.4381, "step": 18534 }, { "epoch": 2.21, "learning_rate": 8.608836430448613e-07, "loss": 0.5523, "step": 18535 }, { "epoch": 2.21, "learning_rate": 8.606409600583687e-07, "loss": 0.6165, "step": 18536 }, { "epoch": 2.21, "learning_rate": 8.603983041706179e-07, "loss": 0.6881, "step": 18537 }, { "epoch": 2.21, "learning_rate": 8.601556753856214e-07, "loss": 0.3743, "step": 18538 }, { "epoch": 2.21, "learning_rate": 8.599130737073882e-07, "loss": 0.3658, "step": 18539 }, { "epoch": 2.21, "learning_rate": 8.596704991399304e-07, "loss": 0.4157, "step": 18540 }, { "epoch": 2.21, "learning_rate": 8.594279516872561e-07, "loss": 0.4219, "step": 18541 }, { "epoch": 2.21, "learning_rate": 8.591854313533754e-07, "loss": 0.541, "step": 18542 }, { "epoch": 2.21, "learning_rate": 8.589429381422973e-07, "loss": 0.7365, "step": 18543 }, { "epoch": 2.21, "learning_rate": 8.587004720580308e-07, "loss": 0.5951, "step": 18544 }, { "epoch": 2.21, "learning_rate": 8.584580331045833e-07, "loss": 0.2565, "step": 18545 }, { "epoch": 2.21, "learning_rate": 8.582156212859616e-07, "loss": 0.3825, "step": 18546 }, { "epoch": 2.21, "learning_rate": 8.579732366061741e-07, "loss": 0.4398, "step": 18547 }, { "epoch": 2.21, "learning_rate": 8.577308790692257e-07, "loss": 0.5286, "step": 18548 }, { "epoch": 2.21, "learning_rate": 8.57488548679124e-07, "loss": 0.5022, "step": 18549 }, { "epoch": 2.21, "learning_rate": 8.572462454398741e-07, "loss": 0.5368, "step": 18550 }, { "epoch": 2.21, "learning_rate": 8.57003969355483e-07, "loss": 0.4681, "step": 18551 }, { "epoch": 2.21, "learning_rate": 8.567617204299535e-07, "loss": 0.3289, "step": 18552 }, { "epoch": 2.21, "learning_rate": 8.565194986672903e-07, "loss": 0.3527, "step": 18553 }, { "epoch": 2.21, "learning_rate": 8.562773040714975e-07, "loss": 0.3667, "step": 18554 }, { "epoch": 2.21, "learning_rate": 8.560351366465794e-07, "loss": 0.513, "step": 18555 }, { "epoch": 2.21, "learning_rate": 8.557929963965378e-07, "loss": 0.6055, "step": 18556 }, { "epoch": 2.21, "learning_rate": 8.555508833253759e-07, "loss": 0.3737, "step": 18557 }, { "epoch": 2.21, "learning_rate": 8.553087974370969e-07, "loss": 0.3105, "step": 18558 }, { "epoch": 2.21, "learning_rate": 8.550667387357012e-07, "loss": 0.4173, "step": 18559 }, { "epoch": 2.21, "learning_rate": 8.548247072251897e-07, "loss": 0.4554, "step": 18560 }, { "epoch": 2.21, "learning_rate": 8.545827029095638e-07, "loss": 0.6001, "step": 18561 }, { "epoch": 2.21, "learning_rate": 8.543407257928246e-07, "loss": 0.5554, "step": 18562 }, { "epoch": 2.21, "learning_rate": 8.540987758789704e-07, "loss": 0.3537, "step": 18563 }, { "epoch": 2.21, "learning_rate": 8.538568531720018e-07, "loss": 0.2989, "step": 18564 }, { "epoch": 2.21, "learning_rate": 8.536149576759173e-07, "loss": 0.4048, "step": 18565 }, { "epoch": 2.21, "learning_rate": 8.533730893947165e-07, "loss": 0.4522, "step": 18566 }, { "epoch": 2.21, "learning_rate": 8.53131248332397e-07, "loss": 0.498, "step": 18567 }, { "epoch": 2.21, "learning_rate": 8.52889434492955e-07, "loss": 0.6977, "step": 18568 }, { "epoch": 2.21, "learning_rate": 8.526476478803888e-07, "loss": 0.4413, "step": 18569 }, { "epoch": 2.21, "learning_rate": 8.524058884986963e-07, "loss": 0.3846, "step": 18570 }, { "epoch": 2.21, "learning_rate": 8.521641563518717e-07, "loss": 0.3071, "step": 18571 }, { "epoch": 2.21, "learning_rate": 8.51922451443912e-07, "loss": 0.5441, "step": 18572 }, { "epoch": 2.21, "learning_rate": 8.516807737788133e-07, "loss": 0.4202, "step": 18573 }, { "epoch": 2.21, "learning_rate": 8.514391233605698e-07, "loss": 0.4816, "step": 18574 }, { "epoch": 2.21, "learning_rate": 8.511975001931749e-07, "loss": 0.6447, "step": 18575 }, { "epoch": 2.21, "learning_rate": 8.509559042806237e-07, "loss": 0.416, "step": 18576 }, { "epoch": 2.21, "learning_rate": 8.507143356269098e-07, "loss": 0.2946, "step": 18577 }, { "epoch": 2.21, "learning_rate": 8.504727942360271e-07, "loss": 0.4242, "step": 18578 }, { "epoch": 2.21, "learning_rate": 8.502312801119666e-07, "loss": 0.4296, "step": 18579 }, { "epoch": 2.21, "learning_rate": 8.499897932587226e-07, "loss": 0.5361, "step": 18580 }, { "epoch": 2.21, "learning_rate": 8.497483336802856e-07, "loss": 0.5447, "step": 18581 }, { "epoch": 2.21, "learning_rate": 8.495069013806461e-07, "loss": 0.3121, "step": 18582 }, { "epoch": 2.21, "learning_rate": 8.492654963637962e-07, "loss": 0.3721, "step": 18583 }, { "epoch": 2.21, "learning_rate": 8.490241186337261e-07, "loss": 0.5402, "step": 18584 }, { "epoch": 2.21, "learning_rate": 8.487827681944267e-07, "loss": 0.5116, "step": 18585 }, { "epoch": 2.21, "learning_rate": 8.485414450498858e-07, "loss": 0.5403, "step": 18586 }, { "epoch": 2.21, "learning_rate": 8.483001492040943e-07, "loss": 0.5526, "step": 18587 }, { "epoch": 2.21, "learning_rate": 8.48058880661039e-07, "loss": 0.4688, "step": 18588 }, { "epoch": 2.21, "learning_rate": 8.4781763942471e-07, "loss": 0.3021, "step": 18589 }, { "epoch": 2.21, "learning_rate": 8.475764254990932e-07, "loss": 0.3918, "step": 18590 }, { "epoch": 2.21, "learning_rate": 8.473352388881767e-07, "loss": 0.4601, "step": 18591 }, { "epoch": 2.21, "learning_rate": 8.470940795959473e-07, "loss": 0.4208, "step": 18592 }, { "epoch": 2.21, "learning_rate": 8.468529476263923e-07, "loss": 0.6323, "step": 18593 }, { "epoch": 2.21, "learning_rate": 8.46611842983496e-07, "loss": 0.5314, "step": 18594 }, { "epoch": 2.21, "learning_rate": 8.463707656712452e-07, "loss": 0.3796, "step": 18595 }, { "epoch": 2.21, "learning_rate": 8.46129715693625e-07, "loss": 0.4709, "step": 18596 }, { "epoch": 2.21, "learning_rate": 8.458886930546181e-07, "loss": 0.4252, "step": 18597 }, { "epoch": 2.21, "learning_rate": 8.456476977582098e-07, "loss": 0.4715, "step": 18598 }, { "epoch": 2.21, "learning_rate": 8.454067298083843e-07, "loss": 0.5949, "step": 18599 }, { "epoch": 2.22, "learning_rate": 8.451657892091248e-07, "loss": 0.6356, "step": 18600 }, { "epoch": 2.22, "learning_rate": 8.449248759644129e-07, "loss": 0.3421, "step": 18601 }, { "epoch": 2.22, "learning_rate": 8.446839900782328e-07, "loss": 0.4292, "step": 18602 }, { "epoch": 2.22, "learning_rate": 8.444431315545642e-07, "loss": 0.4184, "step": 18603 }, { "epoch": 2.22, "learning_rate": 8.442023003973904e-07, "loss": 0.5298, "step": 18604 }, { "epoch": 2.22, "learning_rate": 8.439614966106907e-07, "loss": 0.6329, "step": 18605 }, { "epoch": 2.22, "learning_rate": 8.437207201984462e-07, "loss": 0.6291, "step": 18606 }, { "epoch": 2.22, "learning_rate": 8.434799711646374e-07, "loss": 0.3607, "step": 18607 }, { "epoch": 2.22, "learning_rate": 8.432392495132447e-07, "loss": 0.4024, "step": 18608 }, { "epoch": 2.22, "learning_rate": 8.429985552482458e-07, "loss": 0.4885, "step": 18609 }, { "epoch": 2.22, "learning_rate": 8.427578883736193e-07, "loss": 0.449, "step": 18610 }, { "epoch": 2.22, "learning_rate": 8.425172488933444e-07, "loss": 0.6144, "step": 18611 }, { "epoch": 2.22, "learning_rate": 8.42276636811398e-07, "loss": 0.5631, "step": 18612 }, { "epoch": 2.22, "learning_rate": 8.420360521317577e-07, "loss": 0.4814, "step": 18613 }, { "epoch": 2.22, "learning_rate": 8.417954948584007e-07, "loss": 0.307, "step": 18614 }, { "epoch": 2.22, "learning_rate": 8.415549649953042e-07, "loss": 0.4191, "step": 18615 }, { "epoch": 2.22, "learning_rate": 8.413144625464431e-07, "loss": 0.4447, "step": 18616 }, { "epoch": 2.22, "learning_rate": 8.410739875157922e-07, "loss": 0.4637, "step": 18617 }, { "epoch": 2.22, "learning_rate": 8.408335399073278e-07, "loss": 0.6403, "step": 18618 }, { "epoch": 2.22, "learning_rate": 8.405931197250247e-07, "loss": 0.5042, "step": 18619 }, { "epoch": 2.22, "learning_rate": 8.40352726972856e-07, "loss": 0.2749, "step": 18620 }, { "epoch": 2.22, "learning_rate": 8.401123616547957e-07, "loss": 0.4262, "step": 18621 }, { "epoch": 2.22, "learning_rate": 8.398720237748176e-07, "loss": 0.3976, "step": 18622 }, { "epoch": 2.22, "learning_rate": 8.39631713336895e-07, "loss": 0.4655, "step": 18623 }, { "epoch": 2.22, "learning_rate": 8.393914303449993e-07, "loss": 0.5713, "step": 18624 }, { "epoch": 2.22, "learning_rate": 8.39151174803102e-07, "loss": 0.5227, "step": 18625 }, { "epoch": 2.22, "learning_rate": 8.38910946715176e-07, "loss": 0.3504, "step": 18626 }, { "epoch": 2.22, "learning_rate": 8.386707460851903e-07, "loss": 0.4685, "step": 18627 }, { "epoch": 2.22, "learning_rate": 8.384305729171169e-07, "loss": 0.3727, "step": 18628 }, { "epoch": 2.22, "learning_rate": 8.381904272149255e-07, "loss": 0.4164, "step": 18629 }, { "epoch": 2.22, "learning_rate": 8.379503089825863e-07, "loss": 0.4659, "step": 18630 }, { "epoch": 2.22, "learning_rate": 8.377102182240682e-07, "loss": 0.5526, "step": 18631 }, { "epoch": 2.22, "learning_rate": 8.37470154943339e-07, "loss": 0.3611, "step": 18632 }, { "epoch": 2.22, "learning_rate": 8.372301191443674e-07, "loss": 0.4355, "step": 18633 }, { "epoch": 2.22, "learning_rate": 8.369901108311221e-07, "loss": 0.384, "step": 18634 }, { "epoch": 2.22, "learning_rate": 8.367501300075692e-07, "loss": 0.4865, "step": 18635 }, { "epoch": 2.22, "learning_rate": 8.365101766776762e-07, "loss": 0.5814, "step": 18636 }, { "epoch": 2.22, "learning_rate": 8.362702508454104e-07, "loss": 0.6807, "step": 18637 }, { "epoch": 2.22, "learning_rate": 8.360303525147371e-07, "loss": 0.3995, "step": 18638 }, { "epoch": 2.22, "learning_rate": 8.357904816896207e-07, "loss": 0.3531, "step": 18639 }, { "epoch": 2.22, "learning_rate": 8.355506383740273e-07, "loss": 0.4639, "step": 18640 }, { "epoch": 2.22, "learning_rate": 8.353108225719215e-07, "loss": 0.3735, "step": 18641 }, { "epoch": 2.22, "learning_rate": 8.350710342872681e-07, "loss": 0.554, "step": 18642 }, { "epoch": 2.22, "learning_rate": 8.348312735240296e-07, "loss": 0.6224, "step": 18643 }, { "epoch": 2.22, "learning_rate": 8.345915402861706e-07, "loss": 0.4753, "step": 18644 }, { "epoch": 2.22, "learning_rate": 8.343518345776533e-07, "loss": 0.4018, "step": 18645 }, { "epoch": 2.22, "learning_rate": 8.34112156402439e-07, "loss": 0.315, "step": 18646 }, { "epoch": 2.22, "learning_rate": 8.338725057644906e-07, "loss": 0.4388, "step": 18647 }, { "epoch": 2.22, "learning_rate": 8.336328826677694e-07, "loss": 0.5074, "step": 18648 }, { "epoch": 2.22, "learning_rate": 8.333932871162375e-07, "loss": 0.64, "step": 18649 }, { "epoch": 2.22, "learning_rate": 8.331537191138534e-07, "loss": 0.4998, "step": 18650 }, { "epoch": 2.22, "learning_rate": 8.329141786645783e-07, "loss": 0.273, "step": 18651 }, { "epoch": 2.22, "learning_rate": 8.326746657723725e-07, "loss": 0.448, "step": 18652 }, { "epoch": 2.22, "learning_rate": 8.324351804411943e-07, "loss": 0.5135, "step": 18653 }, { "epoch": 2.22, "learning_rate": 8.321957226750016e-07, "loss": 0.4806, "step": 18654 }, { "epoch": 2.22, "learning_rate": 8.319562924777536e-07, "loss": 0.4955, "step": 18655 }, { "epoch": 2.22, "learning_rate": 8.317168898534081e-07, "loss": 0.6162, "step": 18656 }, { "epoch": 2.22, "learning_rate": 8.314775148059231e-07, "loss": 0.366, "step": 18657 }, { "epoch": 2.22, "learning_rate": 8.312381673392542e-07, "loss": 0.447, "step": 18658 }, { "epoch": 2.22, "learning_rate": 8.309988474573591e-07, "loss": 0.4125, "step": 18659 }, { "epoch": 2.22, "learning_rate": 8.307595551641931e-07, "loss": 0.4579, "step": 18660 }, { "epoch": 2.22, "learning_rate": 8.305202904637108e-07, "loss": 0.5711, "step": 18661 }, { "epoch": 2.22, "learning_rate": 8.302810533598682e-07, "loss": 0.5024, "step": 18662 }, { "epoch": 2.22, "learning_rate": 8.300418438566199e-07, "loss": 0.3788, "step": 18663 }, { "epoch": 2.22, "learning_rate": 8.298026619579211e-07, "loss": 0.3421, "step": 18664 }, { "epoch": 2.22, "learning_rate": 8.295635076677233e-07, "loss": 0.4184, "step": 18665 }, { "epoch": 2.22, "learning_rate": 8.293243809899818e-07, "loss": 0.4696, "step": 18666 }, { "epoch": 2.22, "learning_rate": 8.290852819286476e-07, "loss": 0.4612, "step": 18667 }, { "epoch": 2.22, "learning_rate": 8.288462104876749e-07, "loss": 0.5422, "step": 18668 }, { "epoch": 2.22, "learning_rate": 8.286071666710135e-07, "loss": 0.5075, "step": 18669 }, { "epoch": 2.22, "learning_rate": 8.28368150482616e-07, "loss": 0.3198, "step": 18670 }, { "epoch": 2.22, "learning_rate": 8.281291619264334e-07, "loss": 0.3939, "step": 18671 }, { "epoch": 2.22, "learning_rate": 8.278902010064166e-07, "loss": 0.4988, "step": 18672 }, { "epoch": 2.22, "learning_rate": 8.276512677265153e-07, "loss": 0.4715, "step": 18673 }, { "epoch": 2.22, "learning_rate": 8.274123620906779e-07, "loss": 0.6272, "step": 18674 }, { "epoch": 2.22, "learning_rate": 8.271734841028553e-07, "loss": 0.5992, "step": 18675 }, { "epoch": 2.22, "learning_rate": 8.269346337669945e-07, "loss": 0.3145, "step": 18676 }, { "epoch": 2.22, "learning_rate": 8.266958110870446e-07, "loss": 0.4359, "step": 18677 }, { "epoch": 2.22, "learning_rate": 8.264570160669536e-07, "loss": 0.3871, "step": 18678 }, { "epoch": 2.22, "learning_rate": 8.262182487106691e-07, "loss": 0.4027, "step": 18679 }, { "epoch": 2.22, "learning_rate": 8.259795090221365e-07, "loss": 0.6479, "step": 18680 }, { "epoch": 2.22, "learning_rate": 8.257407970053041e-07, "loss": 0.5073, "step": 18681 }, { "epoch": 2.22, "learning_rate": 8.255021126641161e-07, "loss": 0.2933, "step": 18682 }, { "epoch": 2.22, "learning_rate": 8.252634560025191e-07, "loss": 0.3951, "step": 18683 }, { "epoch": 2.23, "learning_rate": 8.250248270244571e-07, "loss": 0.4016, "step": 18684 }, { "epoch": 2.23, "learning_rate": 8.247862257338754e-07, "loss": 0.4926, "step": 18685 }, { "epoch": 2.23, "learning_rate": 8.245476521347179e-07, "loss": 0.5012, "step": 18686 }, { "epoch": 2.23, "learning_rate": 8.243091062309289e-07, "loss": 0.6367, "step": 18687 }, { "epoch": 2.23, "learning_rate": 8.24070588026451e-07, "loss": 0.4279, "step": 18688 }, { "epoch": 2.23, "learning_rate": 8.238320975252262e-07, "loss": 0.335, "step": 18689 }, { "epoch": 2.23, "learning_rate": 8.235936347311982e-07, "loss": 0.4037, "step": 18690 }, { "epoch": 2.23, "learning_rate": 8.233551996483074e-07, "loss": 0.3875, "step": 18691 }, { "epoch": 2.23, "learning_rate": 8.231167922804958e-07, "loss": 0.5306, "step": 18692 }, { "epoch": 2.23, "learning_rate": 8.228784126317044e-07, "loss": 0.5795, "step": 18693 }, { "epoch": 2.23, "learning_rate": 8.226400607058741e-07, "loss": 0.5612, "step": 18694 }, { "epoch": 2.23, "learning_rate": 8.224017365069445e-07, "loss": 0.2721, "step": 18695 }, { "epoch": 2.23, "learning_rate": 8.221634400388542e-07, "loss": 0.3233, "step": 18696 }, { "epoch": 2.23, "learning_rate": 8.219251713055428e-07, "loss": 0.323, "step": 18697 }, { "epoch": 2.23, "learning_rate": 8.216869303109501e-07, "loss": 0.4958, "step": 18698 }, { "epoch": 2.23, "learning_rate": 8.214487170590127e-07, "loss": 0.6182, "step": 18699 }, { "epoch": 2.23, "learning_rate": 8.212105315536683e-07, "loss": 0.5852, "step": 18700 }, { "epoch": 2.23, "learning_rate": 8.209723737988559e-07, "loss": 0.2406, "step": 18701 }, { "epoch": 2.23, "learning_rate": 8.20734243798511e-07, "loss": 0.4237, "step": 18702 }, { "epoch": 2.23, "learning_rate": 8.20496141556569e-07, "loss": 0.3815, "step": 18703 }, { "epoch": 2.23, "learning_rate": 8.20258067076967e-07, "loss": 0.4299, "step": 18704 }, { "epoch": 2.23, "learning_rate": 8.200200203636404e-07, "loss": 0.6293, "step": 18705 }, { "epoch": 2.23, "learning_rate": 8.197820014205235e-07, "loss": 0.6526, "step": 18706 }, { "epoch": 2.23, "learning_rate": 8.195440102515506e-07, "loss": 0.3116, "step": 18707 }, { "epoch": 2.23, "learning_rate": 8.193060468606568e-07, "loss": 0.2698, "step": 18708 }, { "epoch": 2.23, "learning_rate": 8.190681112517754e-07, "loss": 0.4136, "step": 18709 }, { "epoch": 2.23, "learning_rate": 8.188302034288392e-07, "loss": 0.47, "step": 18710 }, { "epoch": 2.23, "learning_rate": 8.185923233957802e-07, "loss": 0.4894, "step": 18711 }, { "epoch": 2.23, "learning_rate": 8.183544711565311e-07, "loss": 0.541, "step": 18712 }, { "epoch": 2.23, "learning_rate": 8.181166467150245e-07, "loss": 0.489, "step": 18713 }, { "epoch": 2.23, "learning_rate": 8.178788500751902e-07, "loss": 0.3838, "step": 18714 }, { "epoch": 2.23, "learning_rate": 8.176410812409593e-07, "loss": 0.4682, "step": 18715 }, { "epoch": 2.23, "learning_rate": 8.174033402162638e-07, "loss": 0.4007, "step": 18716 }, { "epoch": 2.23, "learning_rate": 8.171656270050318e-07, "loss": 0.5955, "step": 18717 }, { "epoch": 2.23, "learning_rate": 8.169279416111925e-07, "loss": 0.6201, "step": 18718 }, { "epoch": 2.23, "learning_rate": 8.166902840386756e-07, "loss": 0.4723, "step": 18719 }, { "epoch": 2.23, "learning_rate": 8.164526542914094e-07, "loss": 0.3609, "step": 18720 }, { "epoch": 2.23, "learning_rate": 8.162150523733231e-07, "loss": 0.4032, "step": 18721 }, { "epoch": 2.23, "learning_rate": 8.159774782883426e-07, "loss": 0.4636, "step": 18722 }, { "epoch": 2.23, "learning_rate": 8.15739932040396e-07, "loss": 0.4513, "step": 18723 }, { "epoch": 2.23, "learning_rate": 8.1550241363341e-07, "loss": 0.4968, "step": 18724 }, { "epoch": 2.23, "learning_rate": 8.152649230713098e-07, "loss": 0.5821, "step": 18725 }, { "epoch": 2.23, "learning_rate": 8.150274603580214e-07, "loss": 0.2831, "step": 18726 }, { "epoch": 2.23, "learning_rate": 8.147900254974705e-07, "loss": 0.3962, "step": 18727 }, { "epoch": 2.23, "learning_rate": 8.145526184935829e-07, "loss": 0.4441, "step": 18728 }, { "epoch": 2.23, "learning_rate": 8.143152393502809e-07, "loss": 0.4898, "step": 18729 }, { "epoch": 2.23, "learning_rate": 8.140778880714905e-07, "loss": 0.6518, "step": 18730 }, { "epoch": 2.23, "learning_rate": 8.138405646611331e-07, "loss": 0.4648, "step": 18731 }, { "epoch": 2.23, "learning_rate": 8.136032691231335e-07, "loss": 0.4175, "step": 18732 }, { "epoch": 2.23, "learning_rate": 8.133660014614126e-07, "loss": 0.3526, "step": 18733 }, { "epoch": 2.23, "learning_rate": 8.13128761679893e-07, "loss": 0.4373, "step": 18734 }, { "epoch": 2.23, "learning_rate": 8.128915497824966e-07, "loss": 0.424, "step": 18735 }, { "epoch": 2.23, "learning_rate": 8.126543657731453e-07, "loss": 0.5583, "step": 18736 }, { "epoch": 2.23, "learning_rate": 8.124172096557581e-07, "loss": 0.6725, "step": 18737 }, { "epoch": 2.23, "learning_rate": 8.12180081434257e-07, "loss": 0.4775, "step": 18738 }, { "epoch": 2.23, "learning_rate": 8.119429811125607e-07, "loss": 0.417, "step": 18739 }, { "epoch": 2.23, "learning_rate": 8.117059086945875e-07, "loss": 0.3828, "step": 18740 }, { "epoch": 2.23, "learning_rate": 8.114688641842574e-07, "loss": 0.4136, "step": 18741 }, { "epoch": 2.23, "learning_rate": 8.112318475854888e-07, "loss": 0.5359, "step": 18742 }, { "epoch": 2.23, "learning_rate": 8.109948589022001e-07, "loss": 0.6033, "step": 18743 }, { "epoch": 2.23, "learning_rate": 8.107578981383074e-07, "loss": 0.4949, "step": 18744 }, { "epoch": 2.23, "learning_rate": 8.105209652977292e-07, "loss": 0.4223, "step": 18745 }, { "epoch": 2.23, "learning_rate": 8.102840603843804e-07, "loss": 0.3941, "step": 18746 }, { "epoch": 2.23, "learning_rate": 8.100471834021787e-07, "loss": 0.4211, "step": 18747 }, { "epoch": 2.23, "learning_rate": 8.098103343550378e-07, "loss": 0.4301, "step": 18748 }, { "epoch": 2.23, "learning_rate": 8.095735132468743e-07, "loss": 0.6143, "step": 18749 }, { "epoch": 2.23, "learning_rate": 8.093367200816021e-07, "loss": 0.5748, "step": 18750 }, { "epoch": 2.23, "learning_rate": 8.090999548631368e-07, "loss": 0.3054, "step": 18751 }, { "epoch": 2.23, "learning_rate": 8.088632175953912e-07, "loss": 0.393, "step": 18752 }, { "epoch": 2.23, "learning_rate": 8.086265082822775e-07, "loss": 0.4832, "step": 18753 }, { "epoch": 2.23, "learning_rate": 8.083898269277102e-07, "loss": 0.4967, "step": 18754 }, { "epoch": 2.23, "learning_rate": 8.081531735356004e-07, "loss": 0.6045, "step": 18755 }, { "epoch": 2.23, "learning_rate": 8.079165481098605e-07, "loss": 0.608, "step": 18756 }, { "epoch": 2.23, "learning_rate": 8.076799506544022e-07, "loss": 0.3329, "step": 18757 }, { "epoch": 2.23, "learning_rate": 8.074433811731372e-07, "loss": 0.3867, "step": 18758 }, { "epoch": 2.23, "learning_rate": 8.072068396699748e-07, "loss": 0.5157, "step": 18759 }, { "epoch": 2.23, "learning_rate": 8.069703261488249e-07, "loss": 0.4361, "step": 18760 }, { "epoch": 2.23, "learning_rate": 8.067338406135972e-07, "loss": 0.5538, "step": 18761 }, { "epoch": 2.23, "learning_rate": 8.064973830682024e-07, "loss": 0.6569, "step": 18762 }, { "epoch": 2.23, "learning_rate": 8.062609535165469e-07, "loss": 0.3186, "step": 18763 }, { "epoch": 2.23, "learning_rate": 8.060245519625398e-07, "loss": 0.4768, "step": 18764 }, { "epoch": 2.23, "learning_rate": 8.057881784100893e-07, "loss": 0.4361, "step": 18765 }, { "epoch": 2.23, "learning_rate": 8.055518328631032e-07, "loss": 0.5051, "step": 18766 }, { "epoch": 2.23, "learning_rate": 8.053155153254871e-07, "loss": 0.4668, "step": 18767 }, { "epoch": 2.24, "learning_rate": 8.050792258011472e-07, "loss": 0.6603, "step": 18768 }, { "epoch": 2.24, "learning_rate": 8.048429642939908e-07, "loss": 0.6233, "step": 18769 }, { "epoch": 2.24, "learning_rate": 8.046067308079217e-07, "loss": 0.354, "step": 18770 }, { "epoch": 2.24, "learning_rate": 8.043705253468454e-07, "loss": 0.3675, "step": 18771 }, { "epoch": 2.24, "learning_rate": 8.041343479146666e-07, "loss": 0.3944, "step": 18772 }, { "epoch": 2.24, "learning_rate": 8.038981985152899e-07, "loss": 0.43, "step": 18773 }, { "epoch": 2.24, "learning_rate": 8.036620771526185e-07, "loss": 0.6155, "step": 18774 }, { "epoch": 2.24, "learning_rate": 8.034259838305547e-07, "loss": 0.5984, "step": 18775 }, { "epoch": 2.24, "learning_rate": 8.031899185530015e-07, "loss": 0.3938, "step": 18776 }, { "epoch": 2.24, "learning_rate": 8.029538813238622e-07, "loss": 0.4359, "step": 18777 }, { "epoch": 2.24, "learning_rate": 8.027178721470369e-07, "loss": 0.4621, "step": 18778 }, { "epoch": 2.24, "learning_rate": 8.024818910264273e-07, "loss": 0.5385, "step": 18779 }, { "epoch": 2.24, "learning_rate": 8.022459379659356e-07, "loss": 0.6711, "step": 18780 }, { "epoch": 2.24, "learning_rate": 8.020100129694608e-07, "loss": 0.5489, "step": 18781 }, { "epoch": 2.24, "learning_rate": 8.017741160409023e-07, "loss": 0.3609, "step": 18782 }, { "epoch": 2.24, "learning_rate": 8.015382471841601e-07, "loss": 0.3965, "step": 18783 }, { "epoch": 2.24, "learning_rate": 8.013024064031333e-07, "loss": 0.3997, "step": 18784 }, { "epoch": 2.24, "learning_rate": 8.010665937017209e-07, "loss": 0.4492, "step": 18785 }, { "epoch": 2.24, "learning_rate": 8.008308090838193e-07, "loss": 0.4575, "step": 18786 }, { "epoch": 2.24, "learning_rate": 8.005950525533282e-07, "loss": 0.6088, "step": 18787 }, { "epoch": 2.24, "learning_rate": 8.003593241141433e-07, "loss": 0.3635, "step": 18788 }, { "epoch": 2.24, "learning_rate": 8.001236237701604e-07, "loss": 0.4146, "step": 18789 }, { "epoch": 2.24, "learning_rate": 7.99887951525277e-07, "loss": 0.4058, "step": 18790 }, { "epoch": 2.24, "learning_rate": 7.996523073833881e-07, "loss": 0.4106, "step": 18791 }, { "epoch": 2.24, "learning_rate": 7.9941669134839e-07, "loss": 0.5445, "step": 18792 }, { "epoch": 2.24, "learning_rate": 7.991811034241762e-07, "loss": 0.5298, "step": 18793 }, { "epoch": 2.24, "learning_rate": 7.989455436146412e-07, "loss": 0.4981, "step": 18794 }, { "epoch": 2.24, "learning_rate": 7.987100119236799e-07, "loss": 0.2756, "step": 18795 }, { "epoch": 2.24, "learning_rate": 7.984745083551848e-07, "loss": 0.4261, "step": 18796 }, { "epoch": 2.24, "learning_rate": 7.982390329130482e-07, "loss": 0.4718, "step": 18797 }, { "epoch": 2.24, "learning_rate": 7.980035856011631e-07, "loss": 0.4445, "step": 18798 }, { "epoch": 2.24, "learning_rate": 7.977681664234215e-07, "loss": 0.491, "step": 18799 }, { "epoch": 2.24, "learning_rate": 7.975327753837159e-07, "loss": 0.5902, "step": 18800 }, { "epoch": 2.24, "learning_rate": 7.972974124859356e-07, "loss": 0.3098, "step": 18801 }, { "epoch": 2.24, "learning_rate": 7.970620777339727e-07, "loss": 0.3484, "step": 18802 }, { "epoch": 2.24, "learning_rate": 7.968267711317165e-07, "loss": 0.5001, "step": 18803 }, { "epoch": 2.24, "learning_rate": 7.965914926830559e-07, "loss": 0.461, "step": 18804 }, { "epoch": 2.24, "learning_rate": 7.963562423918808e-07, "loss": 0.5444, "step": 18805 }, { "epoch": 2.24, "learning_rate": 7.961210202620797e-07, "loss": 0.6751, "step": 18806 }, { "epoch": 2.24, "learning_rate": 7.958858262975425e-07, "loss": 0.2987, "step": 18807 }, { "epoch": 2.24, "learning_rate": 7.956506605021544e-07, "loss": 0.3667, "step": 18808 }, { "epoch": 2.24, "learning_rate": 7.954155228798047e-07, "loss": 0.4748, "step": 18809 }, { "epoch": 2.24, "learning_rate": 7.951804134343785e-07, "loss": 0.4157, "step": 18810 }, { "epoch": 2.24, "learning_rate": 7.949453321697643e-07, "loss": 0.5512, "step": 18811 }, { "epoch": 2.24, "learning_rate": 7.947102790898456e-07, "loss": 0.6324, "step": 18812 }, { "epoch": 2.24, "learning_rate": 7.944752541985093e-07, "loss": 0.4909, "step": 18813 }, { "epoch": 2.24, "learning_rate": 7.942402574996403e-07, "loss": 0.3782, "step": 18814 }, { "epoch": 2.24, "learning_rate": 7.940052889971236e-07, "loss": 0.4607, "step": 18815 }, { "epoch": 2.24, "learning_rate": 7.937703486948426e-07, "loss": 0.4355, "step": 18816 }, { "epoch": 2.24, "learning_rate": 7.935354365966802e-07, "loss": 0.5058, "step": 18817 }, { "epoch": 2.24, "learning_rate": 7.933005527065207e-07, "loss": 0.556, "step": 18818 }, { "epoch": 2.24, "learning_rate": 7.930656970282457e-07, "loss": 0.61, "step": 18819 }, { "epoch": 2.24, "learning_rate": 7.92830869565738e-07, "loss": 0.4696, "step": 18820 }, { "epoch": 2.24, "learning_rate": 7.925960703228794e-07, "loss": 0.3634, "step": 18821 }, { "epoch": 2.24, "learning_rate": 7.923612993035514e-07, "loss": 0.4057, "step": 18822 }, { "epoch": 2.24, "learning_rate": 7.92126556511634e-07, "loss": 0.4684, "step": 18823 }, { "epoch": 2.24, "learning_rate": 7.918918419510085e-07, "loss": 0.6234, "step": 18824 }, { "epoch": 2.24, "learning_rate": 7.916571556255537e-07, "loss": 0.5398, "step": 18825 }, { "epoch": 2.24, "learning_rate": 7.9142249753915e-07, "loss": 0.3237, "step": 18826 }, { "epoch": 2.24, "learning_rate": 7.91187867695675e-07, "loss": 0.4115, "step": 18827 }, { "epoch": 2.24, "learning_rate": 7.909532660990077e-07, "loss": 0.3577, "step": 18828 }, { "epoch": 2.24, "learning_rate": 7.907186927530266e-07, "loss": 0.4484, "step": 18829 }, { "epoch": 2.24, "learning_rate": 7.904841476616096e-07, "loss": 0.6078, "step": 18830 }, { "epoch": 2.24, "learning_rate": 7.902496308286334e-07, "loss": 0.5785, "step": 18831 }, { "epoch": 2.24, "learning_rate": 7.90015142257973e-07, "loss": 0.4394, "step": 18832 }, { "epoch": 2.24, "learning_rate": 7.897806819535067e-07, "loss": 0.3904, "step": 18833 }, { "epoch": 2.24, "learning_rate": 7.895462499191086e-07, "loss": 0.4309, "step": 18834 }, { "epoch": 2.24, "learning_rate": 7.893118461586543e-07, "loss": 0.4113, "step": 18835 }, { "epoch": 2.24, "learning_rate": 7.89077470676019e-07, "loss": 0.6001, "step": 18836 }, { "epoch": 2.24, "learning_rate": 7.888431234750774e-07, "loss": 0.5353, "step": 18837 }, { "epoch": 2.24, "learning_rate": 7.886088045597023e-07, "loss": 0.3273, "step": 18838 }, { "epoch": 2.24, "learning_rate": 7.883745139337667e-07, "loss": 0.3172, "step": 18839 }, { "epoch": 2.24, "learning_rate": 7.881402516011439e-07, "loss": 0.5198, "step": 18840 }, { "epoch": 2.24, "learning_rate": 7.879060175657072e-07, "loss": 0.4253, "step": 18841 }, { "epoch": 2.24, "learning_rate": 7.876718118313267e-07, "loss": 0.5258, "step": 18842 }, { "epoch": 2.24, "learning_rate": 7.874376344018752e-07, "loss": 0.6156, "step": 18843 }, { "epoch": 2.24, "learning_rate": 7.872034852812238e-07, "loss": 0.6383, "step": 18844 }, { "epoch": 2.24, "learning_rate": 7.869693644732426e-07, "loss": 0.306, "step": 18845 }, { "epoch": 2.24, "learning_rate": 7.867352719818008e-07, "loss": 0.4173, "step": 18846 }, { "epoch": 2.24, "learning_rate": 7.865012078107687e-07, "loss": 0.456, "step": 18847 }, { "epoch": 2.24, "learning_rate": 7.86267171964016e-07, "loss": 0.5486, "step": 18848 }, { "epoch": 2.24, "learning_rate": 7.860331644454103e-07, "loss": 0.5437, "step": 18849 }, { "epoch": 2.24, "learning_rate": 7.857991852588198e-07, "loss": 0.5573, "step": 18850 }, { "epoch": 2.24, "learning_rate": 7.85565234408113e-07, "loss": 0.2799, "step": 18851 }, { "epoch": 2.25, "learning_rate": 7.853313118971573e-07, "loss": 0.4515, "step": 18852 }, { "epoch": 2.25, "learning_rate": 7.85097417729819e-07, "loss": 0.362, "step": 18853 }, { "epoch": 2.25, "learning_rate": 7.848635519099635e-07, "loss": 0.4244, "step": 18854 }, { "epoch": 2.25, "learning_rate": 7.846297144414572e-07, "loss": 0.5327, "step": 18855 }, { "epoch": 2.25, "learning_rate": 7.843959053281663e-07, "loss": 0.6437, "step": 18856 }, { "epoch": 2.25, "learning_rate": 7.841621245739547e-07, "loss": 0.3286, "step": 18857 }, { "epoch": 2.25, "learning_rate": 7.839283721826868e-07, "loss": 0.3094, "step": 18858 }, { "epoch": 2.25, "learning_rate": 7.836946481582281e-07, "loss": 0.3837, "step": 18859 }, { "epoch": 2.25, "learning_rate": 7.834609525044404e-07, "loss": 0.4608, "step": 18860 }, { "epoch": 2.25, "learning_rate": 7.832272852251865e-07, "loss": 0.5812, "step": 18861 }, { "epoch": 2.25, "learning_rate": 7.829936463243299e-07, "loss": 0.6367, "step": 18862 }, { "epoch": 2.25, "learning_rate": 7.827600358057322e-07, "loss": 0.3256, "step": 18863 }, { "epoch": 2.25, "learning_rate": 7.82526453673256e-07, "loss": 0.3183, "step": 18864 }, { "epoch": 2.25, "learning_rate": 7.822928999307608e-07, "loss": 0.4191, "step": 18865 }, { "epoch": 2.25, "learning_rate": 7.82059374582109e-07, "loss": 0.5229, "step": 18866 }, { "epoch": 2.25, "learning_rate": 7.8182587763116e-07, "loss": 0.5578, "step": 18867 }, { "epoch": 2.25, "learning_rate": 7.815924090817723e-07, "loss": 0.5318, "step": 18868 }, { "epoch": 2.25, "learning_rate": 7.813589689378065e-07, "loss": 0.515, "step": 18869 }, { "epoch": 2.25, "learning_rate": 7.811255572031209e-07, "loss": 0.4436, "step": 18870 }, { "epoch": 2.25, "learning_rate": 7.80892173881575e-07, "loss": 0.4471, "step": 18871 }, { "epoch": 2.25, "learning_rate": 7.806588189770251e-07, "loss": 0.3756, "step": 18872 }, { "epoch": 2.25, "learning_rate": 7.804254924933297e-07, "loss": 0.5176, "step": 18873 }, { "epoch": 2.25, "learning_rate": 7.801921944343443e-07, "loss": 0.563, "step": 18874 }, { "epoch": 2.25, "learning_rate": 7.799589248039275e-07, "loss": 0.5666, "step": 18875 }, { "epoch": 2.25, "learning_rate": 7.797256836059328e-07, "loss": 0.3312, "step": 18876 }, { "epoch": 2.25, "learning_rate": 7.79492470844217e-07, "loss": 0.3024, "step": 18877 }, { "epoch": 2.25, "learning_rate": 7.79259286522635e-07, "loss": 0.3966, "step": 18878 }, { "epoch": 2.25, "learning_rate": 7.790261306450423e-07, "loss": 0.5089, "step": 18879 }, { "epoch": 2.25, "learning_rate": 7.787930032152913e-07, "loss": 0.5298, "step": 18880 }, { "epoch": 2.25, "learning_rate": 7.785599042372369e-07, "loss": 0.6504, "step": 18881 }, { "epoch": 2.25, "learning_rate": 7.783268337147318e-07, "loss": 0.2905, "step": 18882 }, { "epoch": 2.25, "learning_rate": 7.780937916516279e-07, "loss": 0.3975, "step": 18883 }, { "epoch": 2.25, "learning_rate": 7.778607780517783e-07, "loss": 0.3974, "step": 18884 }, { "epoch": 2.25, "learning_rate": 7.776277929190342e-07, "loss": 0.379, "step": 18885 }, { "epoch": 2.25, "learning_rate": 7.773948362572481e-07, "loss": 0.5121, "step": 18886 }, { "epoch": 2.25, "learning_rate": 7.771619080702692e-07, "loss": 0.5861, "step": 18887 }, { "epoch": 2.25, "learning_rate": 7.76929008361949e-07, "loss": 0.3555, "step": 18888 }, { "epoch": 2.25, "learning_rate": 7.766961371361365e-07, "loss": 0.358, "step": 18889 }, { "epoch": 2.25, "learning_rate": 7.764632943966821e-07, "loss": 0.5005, "step": 18890 }, { "epoch": 2.25, "learning_rate": 7.762304801474333e-07, "loss": 0.5109, "step": 18891 }, { "epoch": 2.25, "learning_rate": 7.75997694392239e-07, "loss": 0.5249, "step": 18892 }, { "epoch": 2.25, "learning_rate": 7.757649371349476e-07, "loss": 0.5942, "step": 18893 }, { "epoch": 2.25, "learning_rate": 7.755322083794075e-07, "loss": 0.6208, "step": 18894 }, { "epoch": 2.25, "learning_rate": 7.752995081294646e-07, "loss": 0.2837, "step": 18895 }, { "epoch": 2.25, "learning_rate": 7.750668363889646e-07, "loss": 0.4157, "step": 18896 }, { "epoch": 2.25, "learning_rate": 7.748341931617554e-07, "loss": 0.4076, "step": 18897 }, { "epoch": 2.25, "learning_rate": 7.746015784516808e-07, "loss": 0.5241, "step": 18898 }, { "epoch": 2.25, "learning_rate": 7.743689922625871e-07, "loss": 0.6335, "step": 18899 }, { "epoch": 2.25, "learning_rate": 7.741364345983188e-07, "loss": 0.6536, "step": 18900 }, { "epoch": 2.25, "learning_rate": 7.739039054627206e-07, "loss": 0.3466, "step": 18901 }, { "epoch": 2.25, "learning_rate": 7.736714048596359e-07, "loss": 0.3709, "step": 18902 }, { "epoch": 2.25, "learning_rate": 7.734389327929068e-07, "loss": 0.5251, "step": 18903 }, { "epoch": 2.25, "learning_rate": 7.732064892663771e-07, "loss": 0.4271, "step": 18904 }, { "epoch": 2.25, "learning_rate": 7.729740742838898e-07, "loss": 0.5162, "step": 18905 }, { "epoch": 2.25, "learning_rate": 7.727416878492849e-07, "loss": 0.6424, "step": 18906 }, { "epoch": 2.25, "learning_rate": 7.725093299664052e-07, "loss": 0.3747, "step": 18907 }, { "epoch": 2.25, "learning_rate": 7.722770006390911e-07, "loss": 0.3378, "step": 18908 }, { "epoch": 2.25, "learning_rate": 7.720446998711839e-07, "loss": 0.376, "step": 18909 }, { "epoch": 2.25, "learning_rate": 7.718124276665229e-07, "loss": 0.4431, "step": 18910 }, { "epoch": 2.25, "learning_rate": 7.715801840289467e-07, "loss": 0.5883, "step": 18911 }, { "epoch": 2.25, "learning_rate": 7.713479689622958e-07, "loss": 0.6049, "step": 18912 }, { "epoch": 2.25, "learning_rate": 7.711157824704071e-07, "loss": 0.3242, "step": 18913 }, { "epoch": 2.25, "learning_rate": 7.708836245571197e-07, "loss": 0.3753, "step": 18914 }, { "epoch": 2.25, "learning_rate": 7.706514952262711e-07, "loss": 0.3833, "step": 18915 }, { "epoch": 2.25, "learning_rate": 7.70419394481699e-07, "loss": 0.4936, "step": 18916 }, { "epoch": 2.25, "learning_rate": 7.701873223272397e-07, "loss": 0.4209, "step": 18917 }, { "epoch": 2.25, "learning_rate": 7.69955278766728e-07, "loss": 0.711, "step": 18918 }, { "epoch": 2.25, "learning_rate": 7.697232638040011e-07, "loss": 0.5139, "step": 18919 }, { "epoch": 2.25, "learning_rate": 7.694912774428945e-07, "loss": 0.3647, "step": 18920 }, { "epoch": 2.25, "learning_rate": 7.692593196872414e-07, "loss": 0.4239, "step": 18921 }, { "epoch": 2.25, "learning_rate": 7.690273905408768e-07, "loss": 0.4997, "step": 18922 }, { "epoch": 2.25, "learning_rate": 7.687954900076356e-07, "loss": 0.6988, "step": 18923 }, { "epoch": 2.25, "learning_rate": 7.685636180913503e-07, "loss": 0.5803, "step": 18924 }, { "epoch": 2.25, "learning_rate": 7.68331774795853e-07, "loss": 0.5718, "step": 18925 }, { "epoch": 2.25, "learning_rate": 7.680999601249764e-07, "loss": 0.2623, "step": 18926 }, { "epoch": 2.25, "learning_rate": 7.67868174082553e-07, "loss": 0.4474, "step": 18927 }, { "epoch": 2.25, "learning_rate": 7.676364166724148e-07, "loss": 0.4559, "step": 18928 }, { "epoch": 2.25, "learning_rate": 7.674046878983912e-07, "loss": 0.4071, "step": 18929 }, { "epoch": 2.25, "learning_rate": 7.671729877643141e-07, "loss": 0.5816, "step": 18930 }, { "epoch": 2.25, "learning_rate": 7.669413162740133e-07, "loss": 0.6527, "step": 18931 }, { "epoch": 2.25, "learning_rate": 7.667096734313167e-07, "loss": 0.3757, "step": 18932 }, { "epoch": 2.25, "learning_rate": 7.66478059240055e-07, "loss": 0.3324, "step": 18933 }, { "epoch": 2.25, "learning_rate": 7.662464737040564e-07, "loss": 0.3647, "step": 18934 }, { "epoch": 2.25, "learning_rate": 7.660149168271496e-07, "loss": 0.4612, "step": 18935 }, { "epoch": 2.26, "learning_rate": 7.657833886131614e-07, "loss": 0.5098, "step": 18936 }, { "epoch": 2.26, "learning_rate": 7.655518890659188e-07, "loss": 0.6699, "step": 18937 }, { "epoch": 2.26, "learning_rate": 7.653204181892499e-07, "loss": 0.3463, "step": 18938 }, { "epoch": 2.26, "learning_rate": 7.650889759869801e-07, "loss": 0.3194, "step": 18939 }, { "epoch": 2.26, "learning_rate": 7.648575624629345e-07, "loss": 0.4442, "step": 18940 }, { "epoch": 2.26, "learning_rate": 7.646261776209387e-07, "loss": 0.4785, "step": 18941 }, { "epoch": 2.26, "learning_rate": 7.643948214648178e-07, "loss": 0.5791, "step": 18942 }, { "epoch": 2.26, "learning_rate": 7.641634939983972e-07, "loss": 0.5594, "step": 18943 }, { "epoch": 2.26, "learning_rate": 7.639321952254988e-07, "loss": 0.5289, "step": 18944 }, { "epoch": 2.26, "learning_rate": 7.637009251499474e-07, "loss": 0.3077, "step": 18945 }, { "epoch": 2.26, "learning_rate": 7.634696837755656e-07, "loss": 0.3468, "step": 18946 }, { "epoch": 2.26, "learning_rate": 7.632384711061749e-07, "loss": 0.4766, "step": 18947 }, { "epoch": 2.26, "learning_rate": 7.630072871455979e-07, "loss": 0.5511, "step": 18948 }, { "epoch": 2.26, "learning_rate": 7.627761318976562e-07, "loss": 0.5233, "step": 18949 }, { "epoch": 2.26, "learning_rate": 7.625450053661718e-07, "loss": 0.5283, "step": 18950 }, { "epoch": 2.26, "learning_rate": 7.623139075549632e-07, "loss": 0.3301, "step": 18951 }, { "epoch": 2.26, "learning_rate": 7.620828384678525e-07, "loss": 0.3494, "step": 18952 }, { "epoch": 2.26, "learning_rate": 7.618517981086573e-07, "loss": 0.433, "step": 18953 }, { "epoch": 2.26, "learning_rate": 7.616207864811986e-07, "loss": 0.463, "step": 18954 }, { "epoch": 2.26, "learning_rate": 7.613898035892933e-07, "loss": 0.6576, "step": 18955 }, { "epoch": 2.26, "learning_rate": 7.611588494367605e-07, "loss": 0.548, "step": 18956 }, { "epoch": 2.26, "learning_rate": 7.609279240274178e-07, "loss": 0.4194, "step": 18957 }, { "epoch": 2.26, "learning_rate": 7.606970273650833e-07, "loss": 0.4029, "step": 18958 }, { "epoch": 2.26, "learning_rate": 7.604661594535728e-07, "loss": 0.3857, "step": 18959 }, { "epoch": 2.26, "learning_rate": 7.602353202967019e-07, "loss": 0.4372, "step": 18960 }, { "epoch": 2.26, "learning_rate": 7.60004509898288e-07, "loss": 0.5448, "step": 18961 }, { "epoch": 2.26, "learning_rate": 7.597737282621445e-07, "loss": 0.6273, "step": 18962 }, { "epoch": 2.26, "learning_rate": 7.595429753920875e-07, "loss": 0.4473, "step": 18963 }, { "epoch": 2.26, "learning_rate": 7.593122512919312e-07, "loss": 0.44, "step": 18964 }, { "epoch": 2.26, "learning_rate": 7.590815559654902e-07, "loss": 0.422, "step": 18965 }, { "epoch": 2.26, "learning_rate": 7.588508894165761e-07, "loss": 0.4178, "step": 18966 }, { "epoch": 2.26, "learning_rate": 7.586202516490041e-07, "loss": 0.4691, "step": 18967 }, { "epoch": 2.26, "learning_rate": 7.583896426665844e-07, "loss": 0.5608, "step": 18968 }, { "epoch": 2.26, "learning_rate": 7.581590624731308e-07, "loss": 0.6132, "step": 18969 }, { "epoch": 2.26, "learning_rate": 7.579285110724532e-07, "loss": 0.2673, "step": 18970 }, { "epoch": 2.26, "learning_rate": 7.576979884683639e-07, "loss": 0.3668, "step": 18971 }, { "epoch": 2.26, "learning_rate": 7.574674946646727e-07, "loss": 0.3508, "step": 18972 }, { "epoch": 2.26, "learning_rate": 7.572370296651907e-07, "loss": 0.502, "step": 18973 }, { "epoch": 2.26, "learning_rate": 7.570065934737272e-07, "loss": 0.6147, "step": 18974 }, { "epoch": 2.26, "learning_rate": 7.567761860940898e-07, "loss": 0.7242, "step": 18975 }, { "epoch": 2.26, "learning_rate": 7.565458075300893e-07, "loss": 0.3483, "step": 18976 }, { "epoch": 2.26, "learning_rate": 7.563154577855322e-07, "loss": 0.4, "step": 18977 }, { "epoch": 2.26, "learning_rate": 7.560851368642267e-07, "loss": 0.4206, "step": 18978 }, { "epoch": 2.26, "learning_rate": 7.558548447699804e-07, "loss": 0.5389, "step": 18979 }, { "epoch": 2.26, "learning_rate": 7.556245815066005e-07, "loss": 0.681, "step": 18980 }, { "epoch": 2.26, "learning_rate": 7.553943470778927e-07, "loss": 0.5911, "step": 18981 }, { "epoch": 2.26, "learning_rate": 7.551641414876618e-07, "loss": 0.4083, "step": 18982 }, { "epoch": 2.26, "learning_rate": 7.549339647397141e-07, "loss": 0.3949, "step": 18983 }, { "epoch": 2.26, "learning_rate": 7.547038168378554e-07, "loss": 0.3864, "step": 18984 }, { "epoch": 2.26, "learning_rate": 7.54473697785888e-07, "loss": 0.4423, "step": 18985 }, { "epoch": 2.26, "learning_rate": 7.54243607587617e-07, "loss": 0.5095, "step": 18986 }, { "epoch": 2.26, "learning_rate": 7.540135462468462e-07, "loss": 0.5817, "step": 18987 }, { "epoch": 2.26, "learning_rate": 7.537835137673783e-07, "loss": 0.4166, "step": 18988 }, { "epoch": 2.26, "learning_rate": 7.535535101530145e-07, "loss": 0.3686, "step": 18989 }, { "epoch": 2.26, "learning_rate": 7.533235354075574e-07, "loss": 0.4622, "step": 18990 }, { "epoch": 2.26, "learning_rate": 7.530935895348101e-07, "loss": 0.4315, "step": 18991 }, { "epoch": 2.26, "learning_rate": 7.528636725385712e-07, "loss": 0.4975, "step": 18992 }, { "epoch": 2.26, "learning_rate": 7.526337844226428e-07, "loss": 0.5522, "step": 18993 }, { "epoch": 2.26, "learning_rate": 7.524039251908241e-07, "loss": 0.4654, "step": 18994 }, { "epoch": 2.26, "learning_rate": 7.521740948469164e-07, "loss": 0.3393, "step": 18995 }, { "epoch": 2.26, "learning_rate": 7.519442933947171e-07, "loss": 0.4162, "step": 18996 }, { "epoch": 2.26, "learning_rate": 7.517145208380247e-07, "loss": 0.5393, "step": 18997 }, { "epoch": 2.26, "learning_rate": 7.51484777180638e-07, "loss": 0.5179, "step": 18998 }, { "epoch": 2.26, "learning_rate": 7.512550624263557e-07, "loss": 0.5759, "step": 18999 }, { "epoch": 2.26, "learning_rate": 7.510253765789727e-07, "loss": 0.5696, "step": 19000 }, { "epoch": 2.26, "learning_rate": 7.507957196422872e-07, "loss": 0.3059, "step": 19001 }, { "epoch": 2.26, "learning_rate": 7.505660916200961e-07, "loss": 0.2857, "step": 19002 }, { "epoch": 2.26, "learning_rate": 7.50336492516194e-07, "loss": 0.4588, "step": 19003 }, { "epoch": 2.26, "learning_rate": 7.501069223343759e-07, "loss": 0.4039, "step": 19004 }, { "epoch": 2.26, "learning_rate": 7.498773810784374e-07, "loss": 0.5745, "step": 19005 }, { "epoch": 2.26, "learning_rate": 7.496478687521721e-07, "loss": 0.6836, "step": 19006 }, { "epoch": 2.26, "learning_rate": 7.494183853593756e-07, "loss": 0.2797, "step": 19007 }, { "epoch": 2.26, "learning_rate": 7.491889309038391e-07, "loss": 0.4577, "step": 19008 }, { "epoch": 2.26, "learning_rate": 7.489595053893575e-07, "loss": 0.4899, "step": 19009 }, { "epoch": 2.26, "learning_rate": 7.487301088197221e-07, "loss": 0.4972, "step": 19010 }, { "epoch": 2.26, "learning_rate": 7.485007411987241e-07, "loss": 0.5234, "step": 19011 }, { "epoch": 2.26, "learning_rate": 7.482714025301557e-07, "loss": 0.6786, "step": 19012 }, { "epoch": 2.26, "learning_rate": 7.480420928178084e-07, "loss": 0.4208, "step": 19013 }, { "epoch": 2.26, "learning_rate": 7.478128120654726e-07, "loss": 0.3722, "step": 19014 }, { "epoch": 2.26, "learning_rate": 7.475835602769377e-07, "loss": 0.3635, "step": 19015 }, { "epoch": 2.26, "learning_rate": 7.473543374559941e-07, "loss": 0.4306, "step": 19016 }, { "epoch": 2.26, "learning_rate": 7.471251436064295e-07, "loss": 0.6742, "step": 19017 }, { "epoch": 2.26, "learning_rate": 7.468959787320343e-07, "loss": 0.6396, "step": 19018 }, { "epoch": 2.26, "learning_rate": 7.466668428365947e-07, "loss": 0.5966, "step": 19019 }, { "epoch": 2.27, "learning_rate": 7.464377359238994e-07, "loss": 0.3535, "step": 19020 }, { "epoch": 2.27, "learning_rate": 7.462086579977357e-07, "loss": 0.3784, "step": 19021 }, { "epoch": 2.27, "learning_rate": 7.459796090618907e-07, "loss": 0.3661, "step": 19022 }, { "epoch": 2.27, "learning_rate": 7.457505891201489e-07, "loss": 0.5285, "step": 19023 }, { "epoch": 2.27, "learning_rate": 7.455215981762982e-07, "loss": 0.6386, "step": 19024 }, { "epoch": 2.27, "learning_rate": 7.452926362341226e-07, "loss": 0.5718, "step": 19025 }, { "epoch": 2.27, "learning_rate": 7.450637032974059e-07, "loss": 0.4115, "step": 19026 }, { "epoch": 2.27, "learning_rate": 7.448347993699337e-07, "loss": 0.4036, "step": 19027 }, { "epoch": 2.27, "learning_rate": 7.446059244554892e-07, "loss": 0.549, "step": 19028 }, { "epoch": 2.27, "learning_rate": 7.443770785578575e-07, "loss": 0.417, "step": 19029 }, { "epoch": 2.27, "learning_rate": 7.441482616808188e-07, "loss": 0.4958, "step": 19030 }, { "epoch": 2.27, "learning_rate": 7.439194738281575e-07, "loss": 0.5978, "step": 19031 }, { "epoch": 2.27, "learning_rate": 7.436907150036537e-07, "loss": 0.4157, "step": 19032 }, { "epoch": 2.27, "learning_rate": 7.434619852110906e-07, "loss": 0.3841, "step": 19033 }, { "epoch": 2.27, "learning_rate": 7.432332844542475e-07, "loss": 0.3746, "step": 19034 }, { "epoch": 2.27, "learning_rate": 7.430046127369056e-07, "loss": 0.4728, "step": 19035 }, { "epoch": 2.27, "learning_rate": 7.427759700628451e-07, "loss": 0.61, "step": 19036 }, { "epoch": 2.27, "learning_rate": 7.425473564358457e-07, "loss": 0.6343, "step": 19037 }, { "epoch": 2.27, "learning_rate": 7.423187718596863e-07, "loss": 0.367, "step": 19038 }, { "epoch": 2.27, "learning_rate": 7.420902163381441e-07, "loss": 0.3524, "step": 19039 }, { "epoch": 2.27, "learning_rate": 7.418616898749989e-07, "loss": 0.3841, "step": 19040 }, { "epoch": 2.27, "learning_rate": 7.416331924740269e-07, "loss": 0.3731, "step": 19041 }, { "epoch": 2.27, "learning_rate": 7.414047241390054e-07, "loss": 0.5497, "step": 19042 }, { "epoch": 2.27, "learning_rate": 7.411762848737117e-07, "loss": 0.6803, "step": 19043 }, { "epoch": 2.27, "learning_rate": 7.409478746819224e-07, "loss": 0.4685, "step": 19044 }, { "epoch": 2.27, "learning_rate": 7.40719493567412e-07, "loss": 0.2807, "step": 19045 }, { "epoch": 2.27, "learning_rate": 7.404911415339555e-07, "loss": 0.4549, "step": 19046 }, { "epoch": 2.27, "learning_rate": 7.402628185853278e-07, "loss": 0.4224, "step": 19047 }, { "epoch": 2.27, "learning_rate": 7.400345247253043e-07, "loss": 0.4665, "step": 19048 }, { "epoch": 2.27, "learning_rate": 7.398062599576572e-07, "loss": 0.5773, "step": 19049 }, { "epoch": 2.27, "learning_rate": 7.3957802428616e-07, "loss": 0.6017, "step": 19050 }, { "epoch": 2.27, "learning_rate": 7.393498177145858e-07, "loss": 0.3315, "step": 19051 }, { "epoch": 2.27, "learning_rate": 7.391216402467078e-07, "loss": 0.415, "step": 19052 }, { "epoch": 2.27, "learning_rate": 7.388934918862967e-07, "loss": 0.4951, "step": 19053 }, { "epoch": 2.27, "learning_rate": 7.386653726371232e-07, "loss": 0.45, "step": 19054 }, { "epoch": 2.27, "learning_rate": 7.384372825029598e-07, "loss": 0.5438, "step": 19055 }, { "epoch": 2.27, "learning_rate": 7.38209221487575e-07, "loss": 0.6327, "step": 19056 }, { "epoch": 2.27, "learning_rate": 7.379811895947394e-07, "loss": 0.3712, "step": 19057 }, { "epoch": 2.27, "learning_rate": 7.377531868282228e-07, "loss": 0.4894, "step": 19058 }, { "epoch": 2.27, "learning_rate": 7.375252131917948e-07, "loss": 0.4172, "step": 19059 }, { "epoch": 2.27, "learning_rate": 7.372972686892227e-07, "loss": 0.4298, "step": 19060 }, { "epoch": 2.27, "learning_rate": 7.37069353324274e-07, "loss": 0.6099, "step": 19061 }, { "epoch": 2.27, "learning_rate": 7.368414671007168e-07, "loss": 0.5849, "step": 19062 }, { "epoch": 2.27, "learning_rate": 7.366136100223187e-07, "loss": 0.3199, "step": 19063 }, { "epoch": 2.27, "learning_rate": 7.363857820928449e-07, "loss": 0.2845, "step": 19064 }, { "epoch": 2.27, "learning_rate": 7.361579833160623e-07, "loss": 0.4624, "step": 19065 }, { "epoch": 2.27, "learning_rate": 7.359302136957369e-07, "loss": 0.4612, "step": 19066 }, { "epoch": 2.27, "learning_rate": 7.35702473235633e-07, "loss": 0.4517, "step": 19067 }, { "epoch": 2.27, "learning_rate": 7.354747619395145e-07, "loss": 0.6078, "step": 19068 }, { "epoch": 2.27, "learning_rate": 7.35247079811146e-07, "loss": 0.5803, "step": 19069 }, { "epoch": 2.27, "learning_rate": 7.350194268542918e-07, "loss": 0.3428, "step": 19070 }, { "epoch": 2.27, "learning_rate": 7.34791803072715e-07, "loss": 0.4082, "step": 19071 }, { "epoch": 2.27, "learning_rate": 7.345642084701773e-07, "loss": 0.5558, "step": 19072 }, { "epoch": 2.27, "learning_rate": 7.34336643050442e-07, "loss": 0.5053, "step": 19073 }, { "epoch": 2.27, "learning_rate": 7.341091068172701e-07, "loss": 0.6156, "step": 19074 }, { "epoch": 2.27, "learning_rate": 7.338815997744222e-07, "loss": 0.497, "step": 19075 }, { "epoch": 2.27, "learning_rate": 7.336541219256596e-07, "loss": 0.3403, "step": 19076 }, { "epoch": 2.27, "learning_rate": 7.334266732747425e-07, "loss": 0.3976, "step": 19077 }, { "epoch": 2.27, "learning_rate": 7.331992538254315e-07, "loss": 0.3944, "step": 19078 }, { "epoch": 2.27, "learning_rate": 7.329718635814842e-07, "loss": 0.4827, "step": 19079 }, { "epoch": 2.27, "learning_rate": 7.327445025466606e-07, "loss": 0.5689, "step": 19080 }, { "epoch": 2.27, "learning_rate": 7.325171707247192e-07, "loss": 0.6719, "step": 19081 }, { "epoch": 2.27, "learning_rate": 7.322898681194171e-07, "loss": 0.3126, "step": 19082 }, { "epoch": 2.27, "learning_rate": 7.320625947345114e-07, "loss": 0.4168, "step": 19083 }, { "epoch": 2.27, "learning_rate": 7.318353505737591e-07, "loss": 0.5046, "step": 19084 }, { "epoch": 2.27, "learning_rate": 7.316081356409169e-07, "loss": 0.4745, "step": 19085 }, { "epoch": 2.27, "learning_rate": 7.313809499397415e-07, "loss": 0.5043, "step": 19086 }, { "epoch": 2.27, "learning_rate": 7.311537934739863e-07, "loss": 0.6044, "step": 19087 }, { "epoch": 2.27, "learning_rate": 7.309266662474085e-07, "loss": 0.39, "step": 19088 }, { "epoch": 2.27, "learning_rate": 7.30699568263761e-07, "loss": 0.3858, "step": 19089 }, { "epoch": 2.27, "learning_rate": 7.304724995267975e-07, "loss": 0.3409, "step": 19090 }, { "epoch": 2.27, "learning_rate": 7.302454600402719e-07, "loss": 0.4938, "step": 19091 }, { "epoch": 2.27, "learning_rate": 7.300184498079374e-07, "loss": 0.5788, "step": 19092 }, { "epoch": 2.27, "learning_rate": 7.297914688335473e-07, "loss": 0.5847, "step": 19093 }, { "epoch": 2.27, "learning_rate": 7.295645171208518e-07, "loss": 0.6094, "step": 19094 }, { "epoch": 2.27, "learning_rate": 7.293375946736045e-07, "loss": 0.3303, "step": 19095 }, { "epoch": 2.27, "learning_rate": 7.291107014955542e-07, "loss": 0.3422, "step": 19096 }, { "epoch": 2.27, "learning_rate": 7.288838375904536e-07, "loss": 0.466, "step": 19097 }, { "epoch": 2.27, "learning_rate": 7.28657002962051e-07, "loss": 0.4814, "step": 19098 }, { "epoch": 2.27, "learning_rate": 7.284301976140965e-07, "loss": 0.6574, "step": 19099 }, { "epoch": 2.27, "learning_rate": 7.282034215503397e-07, "loss": 0.5167, "step": 19100 }, { "epoch": 2.27, "learning_rate": 7.279766747745298e-07, "loss": 0.3231, "step": 19101 }, { "epoch": 2.27, "learning_rate": 7.27749957290414e-07, "loss": 0.4299, "step": 19102 }, { "epoch": 2.27, "learning_rate": 7.275232691017395e-07, "loss": 0.4168, "step": 19103 }, { "epoch": 2.28, "learning_rate": 7.272966102122547e-07, "loss": 0.4979, "step": 19104 }, { "epoch": 2.28, "learning_rate": 7.270699806257051e-07, "loss": 0.5609, "step": 19105 }, { "epoch": 2.28, "learning_rate": 7.268433803458374e-07, "loss": 0.6178, "step": 19106 }, { "epoch": 2.28, "learning_rate": 7.266168093763972e-07, "loss": 0.3833, "step": 19107 }, { "epoch": 2.28, "learning_rate": 7.263902677211309e-07, "loss": 0.3258, "step": 19108 }, { "epoch": 2.28, "learning_rate": 7.261637553837817e-07, "loss": 0.4156, "step": 19109 }, { "epoch": 2.28, "learning_rate": 7.259372723680949e-07, "loss": 0.4692, "step": 19110 }, { "epoch": 2.28, "learning_rate": 7.257108186778133e-07, "loss": 0.5652, "step": 19111 }, { "epoch": 2.28, "learning_rate": 7.254843943166815e-07, "loss": 0.5397, "step": 19112 }, { "epoch": 2.28, "learning_rate": 7.252579992884406e-07, "loss": 0.3331, "step": 19113 }, { "epoch": 2.28, "learning_rate": 7.25031633596834e-07, "loss": 0.3536, "step": 19114 }, { "epoch": 2.28, "learning_rate": 7.248052972456035e-07, "loss": 0.4502, "step": 19115 }, { "epoch": 2.28, "learning_rate": 7.245789902384908e-07, "loss": 0.494, "step": 19116 }, { "epoch": 2.28, "learning_rate": 7.243527125792368e-07, "loss": 0.6368, "step": 19117 }, { "epoch": 2.28, "learning_rate": 7.241264642715806e-07, "loss": 0.5903, "step": 19118 }, { "epoch": 2.28, "learning_rate": 7.239002453192637e-07, "loss": 0.4837, "step": 19119 }, { "epoch": 2.28, "learning_rate": 7.236740557260239e-07, "loss": 0.2264, "step": 19120 }, { "epoch": 2.28, "learning_rate": 7.234478954956009e-07, "loss": 0.3898, "step": 19121 }, { "epoch": 2.28, "learning_rate": 7.232217646317336e-07, "loss": 0.4388, "step": 19122 }, { "epoch": 2.28, "learning_rate": 7.229956631381601e-07, "loss": 0.4297, "step": 19123 }, { "epoch": 2.28, "learning_rate": 7.227695910186175e-07, "loss": 0.5391, "step": 19124 }, { "epoch": 2.28, "learning_rate": 7.22543548276842e-07, "loss": 0.5659, "step": 19125 }, { "epoch": 2.28, "learning_rate": 7.223175349165707e-07, "loss": 0.3357, "step": 19126 }, { "epoch": 2.28, "learning_rate": 7.220915509415405e-07, "loss": 0.3282, "step": 19127 }, { "epoch": 2.28, "learning_rate": 7.218655963554852e-07, "loss": 0.4496, "step": 19128 }, { "epoch": 2.28, "learning_rate": 7.21639671162141e-07, "loss": 0.509, "step": 19129 }, { "epoch": 2.28, "learning_rate": 7.214137753652428e-07, "loss": 0.5783, "step": 19130 }, { "epoch": 2.28, "learning_rate": 7.211879089685245e-07, "loss": 0.6422, "step": 19131 }, { "epoch": 2.28, "learning_rate": 7.209620719757182e-07, "loss": 0.3321, "step": 19132 }, { "epoch": 2.28, "learning_rate": 7.207362643905583e-07, "loss": 0.3609, "step": 19133 }, { "epoch": 2.28, "learning_rate": 7.205104862167778e-07, "loss": 0.3647, "step": 19134 }, { "epoch": 2.28, "learning_rate": 7.202847374581076e-07, "loss": 0.414, "step": 19135 }, { "epoch": 2.28, "learning_rate": 7.200590181182801e-07, "loss": 0.541, "step": 19136 }, { "epoch": 2.28, "learning_rate": 7.198333282010264e-07, "loss": 0.5558, "step": 19137 }, { "epoch": 2.28, "learning_rate": 7.19607667710078e-07, "loss": 0.4661, "step": 19138 }, { "epoch": 2.28, "learning_rate": 7.193820366491641e-07, "loss": 0.4812, "step": 19139 }, { "epoch": 2.28, "learning_rate": 7.191564350220137e-07, "loss": 0.5315, "step": 19140 }, { "epoch": 2.28, "learning_rate": 7.189308628323571e-07, "loss": 0.4077, "step": 19141 }, { "epoch": 2.28, "learning_rate": 7.187053200839233e-07, "loss": 0.5923, "step": 19142 }, { "epoch": 2.28, "learning_rate": 7.184798067804394e-07, "loss": 0.5858, "step": 19143 }, { "epoch": 2.28, "learning_rate": 7.182543229256337e-07, "loss": 0.5061, "step": 19144 }, { "epoch": 2.28, "learning_rate": 7.180288685232345e-07, "loss": 0.3236, "step": 19145 }, { "epoch": 2.28, "learning_rate": 7.178034435769673e-07, "loss": 0.3578, "step": 19146 }, { "epoch": 2.28, "learning_rate": 7.175780480905581e-07, "loss": 0.3732, "step": 19147 }, { "epoch": 2.28, "learning_rate": 7.173526820677334e-07, "loss": 0.4434, "step": 19148 }, { "epoch": 2.28, "learning_rate": 7.171273455122183e-07, "loss": 0.5164, "step": 19149 }, { "epoch": 2.28, "learning_rate": 7.169020384277387e-07, "loss": 0.6564, "step": 19150 }, { "epoch": 2.28, "learning_rate": 7.166767608180172e-07, "loss": 0.3429, "step": 19151 }, { "epoch": 2.28, "learning_rate": 7.164515126867791e-07, "loss": 0.2891, "step": 19152 }, { "epoch": 2.28, "learning_rate": 7.162262940377474e-07, "loss": 0.468, "step": 19153 }, { "epoch": 2.28, "learning_rate": 7.160011048746438e-07, "loss": 0.4287, "step": 19154 }, { "epoch": 2.28, "learning_rate": 7.157759452011917e-07, "loss": 0.4619, "step": 19155 }, { "epoch": 2.28, "learning_rate": 7.15550815021113e-07, "loss": 0.6652, "step": 19156 }, { "epoch": 2.28, "learning_rate": 7.153257143381296e-07, "loss": 0.3473, "step": 19157 }, { "epoch": 2.28, "learning_rate": 7.151006431559613e-07, "loss": 0.2966, "step": 19158 }, { "epoch": 2.28, "learning_rate": 7.148756014783298e-07, "loss": 0.3743, "step": 19159 }, { "epoch": 2.28, "learning_rate": 7.146505893089536e-07, "loss": 0.4901, "step": 19160 }, { "epoch": 2.28, "learning_rate": 7.144256066515536e-07, "loss": 0.5811, "step": 19161 }, { "epoch": 2.28, "learning_rate": 7.142006535098478e-07, "loss": 0.6311, "step": 19162 }, { "epoch": 2.28, "learning_rate": 7.139757298875547e-07, "loss": 0.455, "step": 19163 }, { "epoch": 2.28, "learning_rate": 7.137508357883927e-07, "loss": 0.3699, "step": 19164 }, { "epoch": 2.28, "learning_rate": 7.135259712160802e-07, "loss": 0.5039, "step": 19165 }, { "epoch": 2.28, "learning_rate": 7.133011361743322e-07, "loss": 0.4653, "step": 19166 }, { "epoch": 2.28, "learning_rate": 7.130763306668676e-07, "loss": 0.4401, "step": 19167 }, { "epoch": 2.28, "learning_rate": 7.128515546974008e-07, "loss": 0.5524, "step": 19168 }, { "epoch": 2.28, "learning_rate": 7.126268082696469e-07, "loss": 0.5225, "step": 19169 }, { "epoch": 2.28, "learning_rate": 7.124020913873222e-07, "loss": 0.2573, "step": 19170 }, { "epoch": 2.28, "learning_rate": 7.121774040541407e-07, "loss": 0.387, "step": 19171 }, { "epoch": 2.28, "learning_rate": 7.119527462738174e-07, "loss": 0.4496, "step": 19172 }, { "epoch": 2.28, "learning_rate": 7.117281180500646e-07, "loss": 0.5079, "step": 19173 }, { "epoch": 2.28, "learning_rate": 7.115035193865969e-07, "loss": 0.6176, "step": 19174 }, { "epoch": 2.28, "learning_rate": 7.112789502871254e-07, "loss": 0.5795, "step": 19175 }, { "epoch": 2.28, "learning_rate": 7.110544107553635e-07, "loss": 0.3646, "step": 19176 }, { "epoch": 2.28, "learning_rate": 7.108299007950215e-07, "loss": 0.404, "step": 19177 }, { "epoch": 2.28, "learning_rate": 7.106054204098115e-07, "loss": 0.4142, "step": 19178 }, { "epoch": 2.28, "learning_rate": 7.103809696034439e-07, "loss": 0.5294, "step": 19179 }, { "epoch": 2.28, "learning_rate": 7.101565483796299e-07, "loss": 0.559, "step": 19180 }, { "epoch": 2.28, "learning_rate": 7.099321567420783e-07, "loss": 0.5974, "step": 19181 }, { "epoch": 2.28, "learning_rate": 7.097077946944974e-07, "loss": 0.349, "step": 19182 }, { "epoch": 2.28, "learning_rate": 7.09483462240598e-07, "loss": 0.3961, "step": 19183 }, { "epoch": 2.28, "learning_rate": 7.092591593840864e-07, "loss": 0.47, "step": 19184 }, { "epoch": 2.28, "learning_rate": 7.090348861286708e-07, "loss": 0.4979, "step": 19185 }, { "epoch": 2.28, "learning_rate": 7.088106424780591e-07, "loss": 0.4775, "step": 19186 }, { "epoch": 2.28, "learning_rate": 7.085864284359586e-07, "loss": 0.67, "step": 19187 }, { "epoch": 2.29, "learning_rate": 7.083622440060744e-07, "loss": 0.4043, "step": 19188 }, { "epoch": 2.29, "learning_rate": 7.081380891921122e-07, "loss": 0.3891, "step": 19189 }, { "epoch": 2.29, "learning_rate": 7.079139639977778e-07, "loss": 0.4713, "step": 19190 }, { "epoch": 2.29, "learning_rate": 7.076898684267764e-07, "loss": 0.4594, "step": 19191 }, { "epoch": 2.29, "learning_rate": 7.074658024828113e-07, "loss": 0.4941, "step": 19192 }, { "epoch": 2.29, "learning_rate": 7.072417661695869e-07, "loss": 0.5589, "step": 19193 }, { "epoch": 2.29, "learning_rate": 7.070177594908065e-07, "loss": 0.4211, "step": 19194 }, { "epoch": 2.29, "learning_rate": 7.067937824501736e-07, "loss": 0.4182, "step": 19195 }, { "epoch": 2.29, "learning_rate": 7.0656983505139e-07, "loss": 0.4078, "step": 19196 }, { "epoch": 2.29, "learning_rate": 7.063459172981568e-07, "loss": 0.3961, "step": 19197 }, { "epoch": 2.29, "learning_rate": 7.061220291941767e-07, "loss": 0.5547, "step": 19198 }, { "epoch": 2.29, "learning_rate": 7.058981707431495e-07, "loss": 0.5783, "step": 19199 }, { "epoch": 2.29, "learning_rate": 7.056743419487758e-07, "loss": 0.6133, "step": 19200 }, { "epoch": 2.29, "learning_rate": 7.054505428147559e-07, "loss": 0.2843, "step": 19201 }, { "epoch": 2.29, "learning_rate": 7.052267733447898e-07, "loss": 0.416, "step": 19202 }, { "epoch": 2.29, "learning_rate": 7.050030335425759e-07, "loss": 0.4188, "step": 19203 }, { "epoch": 2.29, "learning_rate": 7.047793234118116e-07, "loss": 0.5161, "step": 19204 }, { "epoch": 2.29, "learning_rate": 7.045556429561959e-07, "loss": 0.6023, "step": 19205 }, { "epoch": 2.29, "learning_rate": 7.043319921794264e-07, "loss": 0.5902, "step": 19206 }, { "epoch": 2.29, "learning_rate": 7.041083710851993e-07, "loss": 0.3721, "step": 19207 }, { "epoch": 2.29, "learning_rate": 7.038847796772116e-07, "loss": 0.3874, "step": 19208 }, { "epoch": 2.29, "learning_rate": 7.036612179591598e-07, "loss": 0.376, "step": 19209 }, { "epoch": 2.29, "learning_rate": 7.034376859347386e-07, "loss": 0.4458, "step": 19210 }, { "epoch": 2.29, "learning_rate": 7.032141836076428e-07, "loss": 0.5579, "step": 19211 }, { "epoch": 2.29, "learning_rate": 7.02990710981567e-07, "loss": 0.5856, "step": 19212 }, { "epoch": 2.29, "learning_rate": 7.027672680602058e-07, "loss": 0.3122, "step": 19213 }, { "epoch": 2.29, "learning_rate": 7.02543854847253e-07, "loss": 0.3508, "step": 19214 }, { "epoch": 2.29, "learning_rate": 7.023204713464004e-07, "loss": 0.4879, "step": 19215 }, { "epoch": 2.29, "learning_rate": 7.020971175613422e-07, "loss": 0.4635, "step": 19216 }, { "epoch": 2.29, "learning_rate": 7.018737934957684e-07, "loss": 0.5977, "step": 19217 }, { "epoch": 2.29, "learning_rate": 7.016504991533727e-07, "loss": 0.5832, "step": 19218 }, { "epoch": 2.29, "learning_rate": 7.014272345378442e-07, "loss": 0.5446, "step": 19219 }, { "epoch": 2.29, "learning_rate": 7.012039996528746e-07, "loss": 0.3079, "step": 19220 }, { "epoch": 2.29, "learning_rate": 7.009807945021546e-07, "loss": 0.3748, "step": 19221 }, { "epoch": 2.29, "learning_rate": 7.00757619089372e-07, "loss": 0.4349, "step": 19222 }, { "epoch": 2.29, "learning_rate": 7.00534473418217e-07, "loss": 0.5757, "step": 19223 }, { "epoch": 2.29, "learning_rate": 7.003113574923792e-07, "loss": 0.5833, "step": 19224 }, { "epoch": 2.29, "learning_rate": 7.000882713155455e-07, "loss": 0.4716, "step": 19225 }, { "epoch": 2.29, "learning_rate": 6.998652148914028e-07, "loss": 0.3283, "step": 19226 }, { "epoch": 2.29, "learning_rate": 6.996421882236392e-07, "loss": 0.4256, "step": 19227 }, { "epoch": 2.29, "learning_rate": 6.994191913159413e-07, "loss": 0.3498, "step": 19228 }, { "epoch": 2.29, "learning_rate": 6.99196224171996e-07, "loss": 0.4283, "step": 19229 }, { "epoch": 2.29, "learning_rate": 6.989732867954874e-07, "loss": 0.6003, "step": 19230 }, { "epoch": 2.29, "learning_rate": 6.987503791901023e-07, "loss": 0.6332, "step": 19231 }, { "epoch": 2.29, "learning_rate": 6.985275013595247e-07, "loss": 0.3304, "step": 19232 }, { "epoch": 2.29, "learning_rate": 6.983046533074378e-07, "loss": 0.3884, "step": 19233 }, { "epoch": 2.29, "learning_rate": 6.980818350375262e-07, "loss": 0.2943, "step": 19234 }, { "epoch": 2.29, "learning_rate": 6.978590465534729e-07, "loss": 0.4227, "step": 19235 }, { "epoch": 2.29, "learning_rate": 6.976362878589618e-07, "loss": 0.6785, "step": 19236 }, { "epoch": 2.29, "learning_rate": 6.974135589576731e-07, "loss": 0.5932, "step": 19237 }, { "epoch": 2.29, "learning_rate": 6.971908598532904e-07, "loss": 0.4486, "step": 19238 }, { "epoch": 2.29, "learning_rate": 6.969681905494932e-07, "loss": 0.4833, "step": 19239 }, { "epoch": 2.29, "learning_rate": 6.967455510499638e-07, "loss": 0.4545, "step": 19240 }, { "epoch": 2.29, "learning_rate": 6.965229413583812e-07, "loss": 0.5297, "step": 19241 }, { "epoch": 2.29, "learning_rate": 6.963003614784255e-07, "loss": 0.4614, "step": 19242 }, { "epoch": 2.29, "learning_rate": 6.960778114137764e-07, "loss": 0.6451, "step": 19243 }, { "epoch": 2.29, "learning_rate": 6.95855291168113e-07, "loss": 0.5516, "step": 19244 }, { "epoch": 2.29, "learning_rate": 6.956328007451132e-07, "loss": 0.2928, "step": 19245 }, { "epoch": 2.29, "learning_rate": 6.954103401484535e-07, "loss": 0.3579, "step": 19246 }, { "epoch": 2.29, "learning_rate": 6.951879093818134e-07, "loss": 0.42, "step": 19247 }, { "epoch": 2.29, "learning_rate": 6.949655084488677e-07, "loss": 0.4933, "step": 19248 }, { "epoch": 2.29, "learning_rate": 6.947431373532937e-07, "loss": 0.561, "step": 19249 }, { "epoch": 2.29, "learning_rate": 6.945207960987671e-07, "loss": 0.5733, "step": 19250 }, { "epoch": 2.29, "learning_rate": 6.942984846889639e-07, "loss": 0.3031, "step": 19251 }, { "epoch": 2.29, "learning_rate": 6.940762031275575e-07, "loss": 0.4509, "step": 19252 }, { "epoch": 2.29, "learning_rate": 6.938539514182239e-07, "loss": 0.5191, "step": 19253 }, { "epoch": 2.29, "learning_rate": 6.936317295646349e-07, "loss": 0.5579, "step": 19254 }, { "epoch": 2.29, "learning_rate": 6.934095375704661e-07, "loss": 0.5767, "step": 19255 }, { "epoch": 2.29, "learning_rate": 6.931873754393883e-07, "loss": 0.5948, "step": 19256 }, { "epoch": 2.29, "learning_rate": 6.929652431750747e-07, "loss": 0.4238, "step": 19257 }, { "epoch": 2.29, "learning_rate": 6.927431407811975e-07, "loss": 0.3824, "step": 19258 }, { "epoch": 2.29, "learning_rate": 6.925210682614284e-07, "loss": 0.3938, "step": 19259 }, { "epoch": 2.29, "learning_rate": 6.922990256194378e-07, "loss": 0.4706, "step": 19260 }, { "epoch": 2.29, "learning_rate": 6.920770128588953e-07, "loss": 0.525, "step": 19261 }, { "epoch": 2.29, "learning_rate": 6.91855029983472e-07, "loss": 0.663, "step": 19262 }, { "epoch": 2.29, "learning_rate": 6.916330769968363e-07, "loss": 0.4062, "step": 19263 }, { "epoch": 2.29, "learning_rate": 6.914111539026575e-07, "loss": 0.3335, "step": 19264 }, { "epoch": 2.29, "learning_rate": 6.911892607046039e-07, "loss": 0.4247, "step": 19265 }, { "epoch": 2.29, "learning_rate": 6.909673974063444e-07, "loss": 0.4276, "step": 19266 }, { "epoch": 2.29, "learning_rate": 6.907455640115457e-07, "loss": 0.6026, "step": 19267 }, { "epoch": 2.29, "learning_rate": 6.905237605238738e-07, "loss": 0.5939, "step": 19268 }, { "epoch": 2.29, "learning_rate": 6.90301986946996e-07, "loss": 0.6155, "step": 19269 }, { "epoch": 2.29, "learning_rate": 6.90080243284579e-07, "loss": 0.2423, "step": 19270 }, { "epoch": 2.29, "learning_rate": 6.898585295402865e-07, "loss": 0.3732, "step": 19271 }, { "epoch": 2.3, "learning_rate": 6.896368457177846e-07, "loss": 0.4376, "step": 19272 }, { "epoch": 2.3, "learning_rate": 6.894151918207384e-07, "loss": 0.3937, "step": 19273 }, { "epoch": 2.3, "learning_rate": 6.891935678528108e-07, "loss": 0.6278, "step": 19274 }, { "epoch": 2.3, "learning_rate": 6.889719738176648e-07, "loss": 0.6964, "step": 19275 }, { "epoch": 2.3, "learning_rate": 6.887504097189638e-07, "loss": 0.2671, "step": 19276 }, { "epoch": 2.3, "learning_rate": 6.885288755603714e-07, "loss": 0.3826, "step": 19277 }, { "epoch": 2.3, "learning_rate": 6.883073713455479e-07, "loss": 0.4451, "step": 19278 }, { "epoch": 2.3, "learning_rate": 6.880858970781556e-07, "loss": 0.5185, "step": 19279 }, { "epoch": 2.3, "learning_rate": 6.878644527618555e-07, "loss": 0.5548, "step": 19280 }, { "epoch": 2.3, "learning_rate": 6.876430384003089e-07, "loss": 0.5766, "step": 19281 }, { "epoch": 2.3, "learning_rate": 6.874216539971748e-07, "loss": 0.2891, "step": 19282 }, { "epoch": 2.3, "learning_rate": 6.872002995561123e-07, "loss": 0.3311, "step": 19283 }, { "epoch": 2.3, "learning_rate": 6.869789750807807e-07, "loss": 0.4736, "step": 19284 }, { "epoch": 2.3, "learning_rate": 6.867576805748397e-07, "loss": 0.4125, "step": 19285 }, { "epoch": 2.3, "learning_rate": 6.865364160419455e-07, "loss": 0.4793, "step": 19286 }, { "epoch": 2.3, "learning_rate": 6.863151814857564e-07, "loss": 0.584, "step": 19287 }, { "epoch": 2.3, "learning_rate": 6.860939769099304e-07, "loss": 0.4769, "step": 19288 }, { "epoch": 2.3, "learning_rate": 6.858728023181232e-07, "loss": 0.3377, "step": 19289 }, { "epoch": 2.3, "learning_rate": 6.856516577139899e-07, "loss": 0.3414, "step": 19290 }, { "epoch": 2.3, "learning_rate": 6.854305431011871e-07, "loss": 0.439, "step": 19291 }, { "epoch": 2.3, "learning_rate": 6.852094584833696e-07, "loss": 0.5758, "step": 19292 }, { "epoch": 2.3, "learning_rate": 6.84988403864193e-07, "loss": 0.5894, "step": 19293 }, { "epoch": 2.3, "learning_rate": 6.847673792473092e-07, "loss": 0.5775, "step": 19294 }, { "epoch": 2.3, "learning_rate": 6.845463846363739e-07, "loss": 0.2654, "step": 19295 }, { "epoch": 2.3, "learning_rate": 6.843254200350394e-07, "loss": 0.407, "step": 19296 }, { "epoch": 2.3, "learning_rate": 6.84104485446957e-07, "loss": 0.4223, "step": 19297 }, { "epoch": 2.3, "learning_rate": 6.838835808757798e-07, "loss": 0.5295, "step": 19298 }, { "epoch": 2.3, "learning_rate": 6.836627063251597e-07, "loss": 0.612, "step": 19299 }, { "epoch": 2.3, "learning_rate": 6.834418617987478e-07, "loss": 0.6001, "step": 19300 }, { "epoch": 2.3, "learning_rate": 6.832210473001941e-07, "loss": 0.3181, "step": 19301 }, { "epoch": 2.3, "learning_rate": 6.830002628331495e-07, "loss": 0.2907, "step": 19302 }, { "epoch": 2.3, "learning_rate": 6.82779508401262e-07, "loss": 0.4413, "step": 19303 }, { "epoch": 2.3, "learning_rate": 6.825587840081827e-07, "loss": 0.4573, "step": 19304 }, { "epoch": 2.3, "learning_rate": 6.823380896575585e-07, "loss": 0.5417, "step": 19305 }, { "epoch": 2.3, "learning_rate": 6.82117425353038e-07, "loss": 0.6833, "step": 19306 }, { "epoch": 2.3, "learning_rate": 6.81896791098269e-07, "loss": 0.3282, "step": 19307 }, { "epoch": 2.3, "learning_rate": 6.816761868968993e-07, "loss": 0.3394, "step": 19308 }, { "epoch": 2.3, "learning_rate": 6.814556127525742e-07, "loss": 0.3889, "step": 19309 }, { "epoch": 2.3, "learning_rate": 6.812350686689412e-07, "loss": 0.4873, "step": 19310 }, { "epoch": 2.3, "learning_rate": 6.810145546496449e-07, "loss": 0.5424, "step": 19311 }, { "epoch": 2.3, "learning_rate": 6.807940706983302e-07, "loss": 0.6185, "step": 19312 }, { "epoch": 2.3, "learning_rate": 6.805736168186419e-07, "loss": 0.3017, "step": 19313 }, { "epoch": 2.3, "learning_rate": 6.803531930142246e-07, "loss": 0.3012, "step": 19314 }, { "epoch": 2.3, "learning_rate": 6.801327992887221e-07, "loss": 0.3955, "step": 19315 }, { "epoch": 2.3, "learning_rate": 6.799124356457767e-07, "loss": 0.4602, "step": 19316 }, { "epoch": 2.3, "learning_rate": 6.796921020890321e-07, "loss": 0.4307, "step": 19317 }, { "epoch": 2.3, "learning_rate": 6.794717986221289e-07, "loss": 0.5268, "step": 19318 }, { "epoch": 2.3, "learning_rate": 6.792515252487106e-07, "loss": 0.5939, "step": 19319 }, { "epoch": 2.3, "learning_rate": 6.790312819724165e-07, "loss": 0.3743, "step": 19320 }, { "epoch": 2.3, "learning_rate": 6.788110687968883e-07, "loss": 0.4251, "step": 19321 }, { "epoch": 2.3, "learning_rate": 6.785908857257656e-07, "loss": 0.5363, "step": 19322 }, { "epoch": 2.3, "learning_rate": 6.783707327626893e-07, "loss": 0.5219, "step": 19323 }, { "epoch": 2.3, "learning_rate": 6.781506099112978e-07, "loss": 0.6551, "step": 19324 }, { "epoch": 2.3, "learning_rate": 6.779305171752284e-07, "loss": 0.5106, "step": 19325 }, { "epoch": 2.3, "learning_rate": 6.777104545581214e-07, "loss": 0.402, "step": 19326 }, { "epoch": 2.3, "learning_rate": 6.774904220636131e-07, "loss": 0.4143, "step": 19327 }, { "epoch": 2.3, "learning_rate": 6.772704196953406e-07, "loss": 0.4294, "step": 19328 }, { "epoch": 2.3, "learning_rate": 6.770504474569411e-07, "loss": 0.4101, "step": 19329 }, { "epoch": 2.3, "learning_rate": 6.768305053520516e-07, "loss": 0.6149, "step": 19330 }, { "epoch": 2.3, "learning_rate": 6.766105933843065e-07, "loss": 0.5512, "step": 19331 }, { "epoch": 2.3, "learning_rate": 6.763907115573407e-07, "loss": 0.2751, "step": 19332 }, { "epoch": 2.3, "learning_rate": 6.761708598747896e-07, "loss": 0.4373, "step": 19333 }, { "epoch": 2.3, "learning_rate": 6.759510383402879e-07, "loss": 0.3739, "step": 19334 }, { "epoch": 2.3, "learning_rate": 6.757312469574675e-07, "loss": 0.4163, "step": 19335 }, { "epoch": 2.3, "learning_rate": 6.75511485729963e-07, "loss": 0.5256, "step": 19336 }, { "epoch": 2.3, "learning_rate": 6.752917546614066e-07, "loss": 0.5278, "step": 19337 }, { "epoch": 2.3, "learning_rate": 6.750720537554315e-07, "loss": 0.3846, "step": 19338 }, { "epoch": 2.3, "learning_rate": 6.748523830156684e-07, "loss": 0.2612, "step": 19339 }, { "epoch": 2.3, "learning_rate": 6.746327424457477e-07, "loss": 0.4371, "step": 19340 }, { "epoch": 2.3, "learning_rate": 6.744131320493019e-07, "loss": 0.4268, "step": 19341 }, { "epoch": 2.3, "learning_rate": 6.741935518299591e-07, "loss": 0.5562, "step": 19342 }, { "epoch": 2.3, "learning_rate": 6.739740017913504e-07, "loss": 0.5858, "step": 19343 }, { "epoch": 2.3, "learning_rate": 6.737544819371045e-07, "loss": 0.644, "step": 19344 }, { "epoch": 2.3, "learning_rate": 6.735349922708511e-07, "loss": 0.3241, "step": 19345 }, { "epoch": 2.3, "learning_rate": 6.733155327962176e-07, "loss": 0.3656, "step": 19346 }, { "epoch": 2.3, "learning_rate": 6.730961035168307e-07, "loss": 0.3832, "step": 19347 }, { "epoch": 2.3, "learning_rate": 6.728767044363186e-07, "loss": 0.435, "step": 19348 }, { "epoch": 2.3, "learning_rate": 6.726573355583085e-07, "loss": 0.5444, "step": 19349 }, { "epoch": 2.3, "learning_rate": 6.724379968864253e-07, "loss": 0.5718, "step": 19350 }, { "epoch": 2.3, "learning_rate": 6.722186884242953e-07, "loss": 0.3235, "step": 19351 }, { "epoch": 2.3, "learning_rate": 6.719994101755448e-07, "loss": 0.3799, "step": 19352 }, { "epoch": 2.3, "learning_rate": 6.71780162143797e-07, "loss": 0.3848, "step": 19353 }, { "epoch": 2.3, "learning_rate": 6.715609443326759e-07, "loss": 0.4905, "step": 19354 }, { "epoch": 2.3, "learning_rate": 6.71341756745806e-07, "loss": 0.5862, "step": 19355 }, { "epoch": 2.31, "learning_rate": 6.7112259938681e-07, "loss": 0.5687, "step": 19356 }, { "epoch": 2.31, "learning_rate": 6.709034722593119e-07, "loss": 0.2522, "step": 19357 }, { "epoch": 2.31, "learning_rate": 6.706843753669321e-07, "loss": 0.4406, "step": 19358 }, { "epoch": 2.31, "learning_rate": 6.704653087132937e-07, "loss": 0.407, "step": 19359 }, { "epoch": 2.31, "learning_rate": 6.702462723020167e-07, "loss": 0.4474, "step": 19360 }, { "epoch": 2.31, "learning_rate": 6.700272661367229e-07, "loss": 0.5952, "step": 19361 }, { "epoch": 2.31, "learning_rate": 6.698082902210313e-07, "loss": 0.5648, "step": 19362 }, { "epoch": 2.31, "learning_rate": 6.695893445585622e-07, "loss": 0.3089, "step": 19363 }, { "epoch": 2.31, "learning_rate": 6.693704291529354e-07, "loss": 0.4075, "step": 19364 }, { "epoch": 2.31, "learning_rate": 6.691515440077684e-07, "loss": 0.45, "step": 19365 }, { "epoch": 2.31, "learning_rate": 6.689326891266801e-07, "loss": 0.5011, "step": 19366 }, { "epoch": 2.31, "learning_rate": 6.687138645132888e-07, "loss": 0.5644, "step": 19367 }, { "epoch": 2.31, "learning_rate": 6.684950701712109e-07, "loss": 0.6471, "step": 19368 }, { "epoch": 2.31, "learning_rate": 6.682763061040623e-07, "loss": 0.576, "step": 19369 }, { "epoch": 2.31, "learning_rate": 6.680575723154603e-07, "loss": 0.2599, "step": 19370 }, { "epoch": 2.31, "learning_rate": 6.678388688090201e-07, "loss": 0.3279, "step": 19371 }, { "epoch": 2.31, "learning_rate": 6.676201955883579e-07, "loss": 0.3628, "step": 19372 }, { "epoch": 2.31, "learning_rate": 6.67401552657087e-07, "loss": 0.5509, "step": 19373 }, { "epoch": 2.31, "learning_rate": 6.671829400188232e-07, "loss": 0.5254, "step": 19374 }, { "epoch": 2.31, "learning_rate": 6.669643576771786e-07, "loss": 0.5731, "step": 19375 }, { "epoch": 2.31, "learning_rate": 6.667458056357668e-07, "loss": 0.3173, "step": 19376 }, { "epoch": 2.31, "learning_rate": 6.665272838982004e-07, "loss": 0.4111, "step": 19377 }, { "epoch": 2.31, "learning_rate": 6.663087924680917e-07, "loss": 0.3879, "step": 19378 }, { "epoch": 2.31, "learning_rate": 6.660903313490535e-07, "loss": 0.4861, "step": 19379 }, { "epoch": 2.31, "learning_rate": 6.658719005446951e-07, "loss": 0.5424, "step": 19380 }, { "epoch": 2.31, "learning_rate": 6.656535000586289e-07, "loss": 0.6155, "step": 19381 }, { "epoch": 2.31, "learning_rate": 6.654351298944634e-07, "loss": 0.3432, "step": 19382 }, { "epoch": 2.31, "learning_rate": 6.652167900558102e-07, "loss": 0.3655, "step": 19383 }, { "epoch": 2.31, "learning_rate": 6.649984805462761e-07, "loss": 0.4406, "step": 19384 }, { "epoch": 2.31, "learning_rate": 6.647802013694715e-07, "loss": 0.4694, "step": 19385 }, { "epoch": 2.31, "learning_rate": 6.645619525290043e-07, "loss": 0.5687, "step": 19386 }, { "epoch": 2.31, "learning_rate": 6.643437340284825e-07, "loss": 0.6233, "step": 19387 }, { "epoch": 2.31, "learning_rate": 6.641255458715129e-07, "loss": 0.4271, "step": 19388 }, { "epoch": 2.31, "learning_rate": 6.639073880617011e-07, "loss": 0.4002, "step": 19389 }, { "epoch": 2.31, "learning_rate": 6.636892606026552e-07, "loss": 0.4203, "step": 19390 }, { "epoch": 2.31, "learning_rate": 6.634711634979793e-07, "loss": 0.466, "step": 19391 }, { "epoch": 2.31, "learning_rate": 6.632530967512787e-07, "loss": 0.4924, "step": 19392 }, { "epoch": 2.31, "learning_rate": 6.63035060366159e-07, "loss": 0.6511, "step": 19393 }, { "epoch": 2.31, "learning_rate": 6.628170543462245e-07, "loss": 0.6375, "step": 19394 }, { "epoch": 2.31, "learning_rate": 6.625990786950773e-07, "loss": 0.3721, "step": 19395 }, { "epoch": 2.31, "learning_rate": 6.623811334163225e-07, "loss": 0.4167, "step": 19396 }, { "epoch": 2.31, "learning_rate": 6.62163218513561e-07, "loss": 0.4838, "step": 19397 }, { "epoch": 2.31, "learning_rate": 6.619453339903961e-07, "loss": 0.5085, "step": 19398 }, { "epoch": 2.31, "learning_rate": 6.617274798504286e-07, "loss": 0.5808, "step": 19399 }, { "epoch": 2.31, "learning_rate": 6.6150965609726e-07, "loss": 0.6109, "step": 19400 }, { "epoch": 2.31, "learning_rate": 6.612918627344911e-07, "loss": 0.3875, "step": 19401 }, { "epoch": 2.31, "learning_rate": 6.610740997657228e-07, "loss": 0.3839, "step": 19402 }, { "epoch": 2.31, "learning_rate": 6.608563671945539e-07, "loss": 0.3828, "step": 19403 }, { "epoch": 2.31, "learning_rate": 6.606386650245827e-07, "loss": 0.4696, "step": 19404 }, { "epoch": 2.31, "learning_rate": 6.604209932594096e-07, "loss": 0.6252, "step": 19405 }, { "epoch": 2.31, "learning_rate": 6.602033519026311e-07, "loss": 0.5663, "step": 19406 }, { "epoch": 2.31, "learning_rate": 6.599857409578456e-07, "loss": 0.3811, "step": 19407 }, { "epoch": 2.31, "learning_rate": 6.597681604286498e-07, "loss": 0.2814, "step": 19408 }, { "epoch": 2.31, "learning_rate": 6.595506103186419e-07, "loss": 0.4303, "step": 19409 }, { "epoch": 2.31, "learning_rate": 6.593330906314166e-07, "loss": 0.5018, "step": 19410 }, { "epoch": 2.31, "learning_rate": 6.591156013705691e-07, "loss": 0.5505, "step": 19411 }, { "epoch": 2.31, "learning_rate": 6.588981425396954e-07, "loss": 0.6069, "step": 19412 }, { "epoch": 2.31, "learning_rate": 6.586807141423903e-07, "loss": 0.3463, "step": 19413 }, { "epoch": 2.31, "learning_rate": 6.584633161822468e-07, "loss": 0.2948, "step": 19414 }, { "epoch": 2.31, "learning_rate": 6.582459486628592e-07, "loss": 0.4839, "step": 19415 }, { "epoch": 2.31, "learning_rate": 6.580286115878215e-07, "loss": 0.4848, "step": 19416 }, { "epoch": 2.31, "learning_rate": 6.578113049607246e-07, "loss": 0.5167, "step": 19417 }, { "epoch": 2.31, "learning_rate": 6.575940287851621e-07, "loss": 0.6597, "step": 19418 }, { "epoch": 2.31, "learning_rate": 6.57376783064724e-07, "loss": 0.4934, "step": 19419 }, { "epoch": 2.31, "learning_rate": 6.571595678030024e-07, "loss": 0.322, "step": 19420 }, { "epoch": 2.31, "learning_rate": 6.569423830035887e-07, "loss": 0.4363, "step": 19421 }, { "epoch": 2.31, "learning_rate": 6.567252286700712e-07, "loss": 0.4336, "step": 19422 }, { "epoch": 2.31, "learning_rate": 6.565081048060401e-07, "loss": 0.5809, "step": 19423 }, { "epoch": 2.31, "learning_rate": 6.562910114150858e-07, "loss": 0.6742, "step": 19424 }, { "epoch": 2.31, "learning_rate": 6.560739485007955e-07, "loss": 0.5302, "step": 19425 }, { "epoch": 2.31, "learning_rate": 6.558569160667571e-07, "loss": 0.3387, "step": 19426 }, { "epoch": 2.31, "learning_rate": 6.556399141165584e-07, "loss": 0.3992, "step": 19427 }, { "epoch": 2.31, "learning_rate": 6.554229426537876e-07, "loss": 0.4883, "step": 19428 }, { "epoch": 2.31, "learning_rate": 6.552060016820294e-07, "loss": 0.4551, "step": 19429 }, { "epoch": 2.31, "learning_rate": 6.549890912048706e-07, "loss": 0.616, "step": 19430 }, { "epoch": 2.31, "learning_rate": 6.54772211225898e-07, "loss": 0.6826, "step": 19431 }, { "epoch": 2.31, "learning_rate": 6.545553617486955e-07, "loss": 0.2958, "step": 19432 }, { "epoch": 2.31, "learning_rate": 6.54338542776847e-07, "loss": 0.3603, "step": 19433 }, { "epoch": 2.31, "learning_rate": 6.541217543139372e-07, "loss": 0.4433, "step": 19434 }, { "epoch": 2.31, "learning_rate": 6.539049963635496e-07, "loss": 0.5474, "step": 19435 }, { "epoch": 2.31, "learning_rate": 6.536882689292684e-07, "loss": 0.6969, "step": 19436 }, { "epoch": 2.31, "learning_rate": 6.53471572014674e-07, "loss": 0.6184, "step": 19437 }, { "epoch": 2.31, "learning_rate": 6.5325490562335e-07, "loss": 0.3414, "step": 19438 }, { "epoch": 2.31, "learning_rate": 6.530382697588777e-07, "loss": 0.3279, "step": 19439 }, { "epoch": 2.32, "learning_rate": 6.528216644248369e-07, "loss": 0.4659, "step": 19440 }, { "epoch": 2.32, "learning_rate": 6.52605089624809e-07, "loss": 0.3071, "step": 19441 }, { "epoch": 2.32, "learning_rate": 6.523885453623741e-07, "loss": 0.5647, "step": 19442 }, { "epoch": 2.32, "learning_rate": 6.521720316411123e-07, "loss": 0.5662, "step": 19443 }, { "epoch": 2.32, "learning_rate": 6.519555484646009e-07, "loss": 0.488, "step": 19444 }, { "epoch": 2.32, "learning_rate": 6.517390958364203e-07, "loss": 0.3109, "step": 19445 }, { "epoch": 2.32, "learning_rate": 6.515226737601465e-07, "loss": 0.4287, "step": 19446 }, { "epoch": 2.32, "learning_rate": 6.51306282239359e-07, "loss": 0.388, "step": 19447 }, { "epoch": 2.32, "learning_rate": 6.510899212776331e-07, "loss": 0.4467, "step": 19448 }, { "epoch": 2.32, "learning_rate": 6.508735908785458e-07, "loss": 0.654, "step": 19449 }, { "epoch": 2.32, "learning_rate": 6.506572910456732e-07, "loss": 0.6113, "step": 19450 }, { "epoch": 2.32, "learning_rate": 6.504410217825915e-07, "loss": 0.3686, "step": 19451 }, { "epoch": 2.32, "learning_rate": 6.50224783092874e-07, "loss": 0.415, "step": 19452 }, { "epoch": 2.32, "learning_rate": 6.500085749800972e-07, "loss": 0.3877, "step": 19453 }, { "epoch": 2.32, "learning_rate": 6.497923974478335e-07, "loss": 0.4994, "step": 19454 }, { "epoch": 2.32, "learning_rate": 6.495762504996561e-07, "loss": 0.564, "step": 19455 }, { "epoch": 2.32, "learning_rate": 6.493601341391387e-07, "loss": 0.6471, "step": 19456 }, { "epoch": 2.32, "learning_rate": 6.491440483698533e-07, "loss": 0.2778, "step": 19457 }, { "epoch": 2.32, "learning_rate": 6.48927993195373e-07, "loss": 0.3326, "step": 19458 }, { "epoch": 2.32, "learning_rate": 6.487119686192675e-07, "loss": 0.435, "step": 19459 }, { "epoch": 2.32, "learning_rate": 6.484959746451092e-07, "loss": 0.5012, "step": 19460 }, { "epoch": 2.32, "learning_rate": 6.48280011276467e-07, "loss": 0.5176, "step": 19461 }, { "epoch": 2.32, "learning_rate": 6.480640785169124e-07, "loss": 0.578, "step": 19462 }, { "epoch": 2.32, "learning_rate": 6.478481763700131e-07, "loss": 0.3344, "step": 19463 }, { "epoch": 2.32, "learning_rate": 6.476323048393391e-07, "loss": 0.2976, "step": 19464 }, { "epoch": 2.32, "learning_rate": 6.474164639284583e-07, "loss": 0.4042, "step": 19465 }, { "epoch": 2.32, "learning_rate": 6.472006536409397e-07, "loss": 0.3282, "step": 19466 }, { "epoch": 2.32, "learning_rate": 6.469848739803497e-07, "loss": 0.5627, "step": 19467 }, { "epoch": 2.32, "learning_rate": 6.467691249502544e-07, "loss": 0.5738, "step": 19468 }, { "epoch": 2.32, "learning_rate": 6.46553406554222e-07, "loss": 0.6883, "step": 19469 }, { "epoch": 2.32, "learning_rate": 6.463377187958161e-07, "loss": 0.4291, "step": 19470 }, { "epoch": 2.32, "learning_rate": 6.461220616786038e-07, "loss": 0.3406, "step": 19471 }, { "epoch": 2.32, "learning_rate": 6.45906435206149e-07, "loss": 0.3506, "step": 19472 }, { "epoch": 2.32, "learning_rate": 6.456908393820172e-07, "loss": 0.4548, "step": 19473 }, { "epoch": 2.32, "learning_rate": 6.454752742097706e-07, "loss": 0.5276, "step": 19474 }, { "epoch": 2.32, "learning_rate": 6.452597396929741e-07, "loss": 0.5272, "step": 19475 }, { "epoch": 2.32, "learning_rate": 6.450442358351891e-07, "loss": 0.3032, "step": 19476 }, { "epoch": 2.32, "learning_rate": 6.44828762639979e-07, "loss": 0.4013, "step": 19477 }, { "epoch": 2.32, "learning_rate": 6.446133201109047e-07, "loss": 0.4486, "step": 19478 }, { "epoch": 2.32, "learning_rate": 6.443979082515278e-07, "loss": 0.4498, "step": 19479 }, { "epoch": 2.32, "learning_rate": 6.44182527065409e-07, "loss": 0.6538, "step": 19480 }, { "epoch": 2.32, "learning_rate": 6.439671765561098e-07, "loss": 0.6012, "step": 19481 }, { "epoch": 2.32, "learning_rate": 6.437518567271886e-07, "loss": 0.2981, "step": 19482 }, { "epoch": 2.32, "learning_rate": 6.435365675822044e-07, "loss": 0.3646, "step": 19483 }, { "epoch": 2.32, "learning_rate": 6.433213091247173e-07, "loss": 0.5065, "step": 19484 }, { "epoch": 2.32, "learning_rate": 6.431060813582837e-07, "loss": 0.47, "step": 19485 }, { "epoch": 2.32, "learning_rate": 6.428908842864626e-07, "loss": 0.5361, "step": 19486 }, { "epoch": 2.32, "learning_rate": 6.42675717912811e-07, "loss": 0.6376, "step": 19487 }, { "epoch": 2.32, "learning_rate": 6.424605822408864e-07, "loss": 0.3711, "step": 19488 }, { "epoch": 2.32, "learning_rate": 6.422454772742443e-07, "loss": 0.3255, "step": 19489 }, { "epoch": 2.32, "learning_rate": 6.420304030164396e-07, "loss": 0.315, "step": 19490 }, { "epoch": 2.32, "learning_rate": 6.41815359471028e-07, "loss": 0.5152, "step": 19491 }, { "epoch": 2.32, "learning_rate": 6.416003466415652e-07, "loss": 0.4482, "step": 19492 }, { "epoch": 2.32, "learning_rate": 6.41385364531604e-07, "loss": 0.5561, "step": 19493 }, { "epoch": 2.32, "learning_rate": 6.411704131446986e-07, "loss": 0.5258, "step": 19494 }, { "epoch": 2.32, "learning_rate": 6.40955492484403e-07, "loss": 0.2747, "step": 19495 }, { "epoch": 2.32, "learning_rate": 6.407406025542692e-07, "loss": 0.3846, "step": 19496 }, { "epoch": 2.32, "learning_rate": 6.405257433578485e-07, "loss": 0.5038, "step": 19497 }, { "epoch": 2.32, "learning_rate": 6.40310914898693e-07, "loss": 0.4574, "step": 19498 }, { "epoch": 2.32, "learning_rate": 6.400961171803544e-07, "loss": 0.5932, "step": 19499 }, { "epoch": 2.32, "learning_rate": 6.398813502063836e-07, "loss": 0.5505, "step": 19500 }, { "epoch": 2.32, "learning_rate": 6.396666139803296e-07, "loss": 0.2356, "step": 19501 }, { "epoch": 2.32, "learning_rate": 6.394519085057433e-07, "loss": 0.3461, "step": 19502 }, { "epoch": 2.32, "learning_rate": 6.392372337861722e-07, "loss": 0.3289, "step": 19503 }, { "epoch": 2.32, "learning_rate": 6.390225898251665e-07, "loss": 0.4122, "step": 19504 }, { "epoch": 2.32, "learning_rate": 6.388079766262728e-07, "loss": 0.5231, "step": 19505 }, { "epoch": 2.32, "learning_rate": 6.385933941930394e-07, "loss": 0.6424, "step": 19506 }, { "epoch": 2.32, "learning_rate": 6.383788425290144e-07, "loss": 0.4798, "step": 19507 }, { "epoch": 2.32, "learning_rate": 6.381643216377423e-07, "loss": 0.3445, "step": 19508 }, { "epoch": 2.32, "learning_rate": 6.379498315227703e-07, "loss": 0.4394, "step": 19509 }, { "epoch": 2.32, "learning_rate": 6.377353721876445e-07, "loss": 0.427, "step": 19510 }, { "epoch": 2.32, "learning_rate": 6.375209436359092e-07, "loss": 0.5361, "step": 19511 }, { "epoch": 2.32, "learning_rate": 6.373065458711083e-07, "loss": 0.6698, "step": 19512 }, { "epoch": 2.32, "learning_rate": 6.370921788967865e-07, "loss": 0.3032, "step": 19513 }, { "epoch": 2.32, "learning_rate": 6.368778427164874e-07, "loss": 0.267, "step": 19514 }, { "epoch": 2.32, "learning_rate": 6.366635373337543e-07, "loss": 0.4132, "step": 19515 }, { "epoch": 2.32, "learning_rate": 6.364492627521288e-07, "loss": 0.4367, "step": 19516 }, { "epoch": 2.32, "learning_rate": 6.36235018975154e-07, "loss": 0.5014, "step": 19517 }, { "epoch": 2.32, "learning_rate": 6.360208060063708e-07, "loss": 0.6146, "step": 19518 }, { "epoch": 2.32, "learning_rate": 6.358066238493193e-07, "loss": 0.6839, "step": 19519 }, { "epoch": 2.32, "learning_rate": 6.355924725075408e-07, "loss": 0.346, "step": 19520 }, { "epoch": 2.32, "learning_rate": 6.353783519845752e-07, "loss": 0.4153, "step": 19521 }, { "epoch": 2.32, "learning_rate": 6.351642622839624e-07, "loss": 0.4085, "step": 19522 }, { "epoch": 2.32, "learning_rate": 6.349502034092403e-07, "loss": 0.5033, "step": 19523 }, { "epoch": 2.33, "learning_rate": 6.347361753639483e-07, "loss": 0.6587, "step": 19524 }, { "epoch": 2.33, "learning_rate": 6.345221781516234e-07, "loss": 0.6035, "step": 19525 }, { "epoch": 2.33, "learning_rate": 6.34308211775804e-07, "loss": 0.3616, "step": 19526 }, { "epoch": 2.33, "learning_rate": 6.340942762400257e-07, "loss": 0.4175, "step": 19527 }, { "epoch": 2.33, "learning_rate": 6.338803715478256e-07, "loss": 0.5138, "step": 19528 }, { "epoch": 2.33, "learning_rate": 6.336664977027396e-07, "loss": 0.4829, "step": 19529 }, { "epoch": 2.33, "learning_rate": 6.334526547083034e-07, "loss": 0.6321, "step": 19530 }, { "epoch": 2.33, "learning_rate": 6.332388425680516e-07, "loss": 0.6804, "step": 19531 }, { "epoch": 2.33, "learning_rate": 6.33025061285518e-07, "loss": 0.3685, "step": 19532 }, { "epoch": 2.33, "learning_rate": 6.32811310864237e-07, "loss": 0.3976, "step": 19533 }, { "epoch": 2.33, "learning_rate": 6.325975913077412e-07, "loss": 0.447, "step": 19534 }, { "epoch": 2.33, "learning_rate": 6.323839026195639e-07, "loss": 0.4492, "step": 19535 }, { "epoch": 2.33, "learning_rate": 6.321702448032374e-07, "loss": 0.5477, "step": 19536 }, { "epoch": 2.33, "learning_rate": 6.319566178622944e-07, "loss": 0.5484, "step": 19537 }, { "epoch": 2.33, "learning_rate": 6.317430218002644e-07, "loss": 0.5253, "step": 19538 }, { "epoch": 2.33, "learning_rate": 6.315294566206795e-07, "loss": 0.3879, "step": 19539 }, { "epoch": 2.33, "learning_rate": 6.313159223270687e-07, "loss": 0.469, "step": 19540 }, { "epoch": 2.33, "learning_rate": 6.311024189229637e-07, "loss": 0.4386, "step": 19541 }, { "epoch": 2.33, "learning_rate": 6.308889464118914e-07, "loss": 0.5656, "step": 19542 }, { "epoch": 2.33, "learning_rate": 6.306755047973817e-07, "loss": 0.5782, "step": 19543 }, { "epoch": 2.33, "learning_rate": 6.304620940829628e-07, "loss": 0.4432, "step": 19544 }, { "epoch": 2.33, "learning_rate": 6.30248714272163e-07, "loss": 0.2919, "step": 19545 }, { "epoch": 2.33, "learning_rate": 6.300353653685091e-07, "loss": 0.3805, "step": 19546 }, { "epoch": 2.33, "learning_rate": 6.298220473755265e-07, "loss": 0.3777, "step": 19547 }, { "epoch": 2.33, "learning_rate": 6.296087602967433e-07, "loss": 0.5019, "step": 19548 }, { "epoch": 2.33, "learning_rate": 6.293955041356833e-07, "loss": 0.6167, "step": 19549 }, { "epoch": 2.33, "learning_rate": 6.291822788958729e-07, "loss": 0.5131, "step": 19550 }, { "epoch": 2.33, "learning_rate": 6.28969084580836e-07, "loss": 0.2759, "step": 19551 }, { "epoch": 2.33, "learning_rate": 6.287559211940982e-07, "loss": 0.5216, "step": 19552 }, { "epoch": 2.33, "learning_rate": 6.285427887391818e-07, "loss": 0.4586, "step": 19553 }, { "epoch": 2.33, "learning_rate": 6.283296872196096e-07, "loss": 0.5268, "step": 19554 }, { "epoch": 2.33, "learning_rate": 6.281166166389047e-07, "loss": 0.5545, "step": 19555 }, { "epoch": 2.33, "learning_rate": 6.279035770005898e-07, "loss": 0.6186, "step": 19556 }, { "epoch": 2.33, "learning_rate": 6.276905683081852e-07, "loss": 0.3537, "step": 19557 }, { "epoch": 2.33, "learning_rate": 6.274775905652125e-07, "loss": 0.3514, "step": 19558 }, { "epoch": 2.33, "learning_rate": 6.272646437751931e-07, "loss": 0.4243, "step": 19559 }, { "epoch": 2.33, "learning_rate": 6.270517279416455e-07, "loss": 0.4004, "step": 19560 }, { "epoch": 2.33, "learning_rate": 6.268388430680908e-07, "loss": 0.555, "step": 19561 }, { "epoch": 2.33, "learning_rate": 6.266259891580465e-07, "loss": 0.5301, "step": 19562 }, { "epoch": 2.33, "learning_rate": 6.264131662150319e-07, "loss": 0.2671, "step": 19563 }, { "epoch": 2.33, "learning_rate": 6.262003742425654e-07, "loss": 0.3105, "step": 19564 }, { "epoch": 2.33, "learning_rate": 6.259876132441634e-07, "loss": 0.447, "step": 19565 }, { "epoch": 2.33, "learning_rate": 6.257748832233432e-07, "loss": 0.4628, "step": 19566 }, { "epoch": 2.33, "learning_rate": 6.255621841836221e-07, "loss": 0.5528, "step": 19567 }, { "epoch": 2.33, "learning_rate": 6.253495161285156e-07, "loss": 0.5551, "step": 19568 }, { "epoch": 2.33, "learning_rate": 6.251368790615381e-07, "loss": 0.6348, "step": 19569 }, { "epoch": 2.33, "learning_rate": 6.249242729862051e-07, "loss": 0.4064, "step": 19570 }, { "epoch": 2.33, "learning_rate": 6.247116979060322e-07, "loss": 0.379, "step": 19571 }, { "epoch": 2.33, "learning_rate": 6.244991538245315e-07, "loss": 0.3795, "step": 19572 }, { "epoch": 2.33, "learning_rate": 6.24286640745217e-07, "loss": 0.452, "step": 19573 }, { "epoch": 2.33, "learning_rate": 6.240741586716026e-07, "loss": 0.5232, "step": 19574 }, { "epoch": 2.33, "learning_rate": 6.238617076071998e-07, "loss": 0.6323, "step": 19575 }, { "epoch": 2.33, "learning_rate": 6.236492875555195e-07, "loss": 0.347, "step": 19576 }, { "epoch": 2.33, "learning_rate": 6.234368985200739e-07, "loss": 0.4201, "step": 19577 }, { "epoch": 2.33, "learning_rate": 6.232245405043741e-07, "loss": 0.3986, "step": 19578 }, { "epoch": 2.33, "learning_rate": 6.230122135119305e-07, "loss": 0.5811, "step": 19579 }, { "epoch": 2.33, "learning_rate": 6.227999175462521e-07, "loss": 0.5577, "step": 19580 }, { "epoch": 2.33, "learning_rate": 6.22587652610849e-07, "loss": 0.5837, "step": 19581 }, { "epoch": 2.33, "learning_rate": 6.223754187092295e-07, "loss": 0.4103, "step": 19582 }, { "epoch": 2.33, "learning_rate": 6.221632158449015e-07, "loss": 0.3258, "step": 19583 }, { "epoch": 2.33, "learning_rate": 6.21951044021373e-07, "loss": 0.3654, "step": 19584 }, { "epoch": 2.33, "learning_rate": 6.217389032421512e-07, "loss": 0.4748, "step": 19585 }, { "epoch": 2.33, "learning_rate": 6.215267935107436e-07, "loss": 0.5497, "step": 19586 }, { "epoch": 2.33, "learning_rate": 6.213147148306554e-07, "loss": 0.4788, "step": 19587 }, { "epoch": 2.33, "learning_rate": 6.211026672053927e-07, "loss": 0.4286, "step": 19588 }, { "epoch": 2.33, "learning_rate": 6.208906506384601e-07, "loss": 0.382, "step": 19589 }, { "epoch": 2.33, "learning_rate": 6.206786651333638e-07, "loss": 0.4082, "step": 19590 }, { "epoch": 2.33, "learning_rate": 6.204667106936057e-07, "loss": 0.4853, "step": 19591 }, { "epoch": 2.33, "learning_rate": 6.202547873226908e-07, "loss": 0.4753, "step": 19592 }, { "epoch": 2.33, "learning_rate": 6.200428950241219e-07, "loss": 0.5908, "step": 19593 }, { "epoch": 2.33, "learning_rate": 6.198310338014024e-07, "loss": 0.4505, "step": 19594 }, { "epoch": 2.33, "learning_rate": 6.196192036580329e-07, "loss": 0.2656, "step": 19595 }, { "epoch": 2.33, "learning_rate": 6.194074045975165e-07, "loss": 0.404, "step": 19596 }, { "epoch": 2.33, "learning_rate": 6.191956366233534e-07, "loss": 0.3555, "step": 19597 }, { "epoch": 2.33, "learning_rate": 6.189838997390437e-07, "loss": 0.5177, "step": 19598 }, { "epoch": 2.33, "learning_rate": 6.187721939480878e-07, "loss": 0.5797, "step": 19599 }, { "epoch": 2.33, "learning_rate": 6.185605192539856e-07, "loss": 0.4635, "step": 19600 }, { "epoch": 2.33, "learning_rate": 6.183488756602366e-07, "loss": 0.3147, "step": 19601 }, { "epoch": 2.33, "learning_rate": 6.181372631703375e-07, "loss": 0.3147, "step": 19602 }, { "epoch": 2.33, "learning_rate": 6.179256817877885e-07, "loss": 0.4624, "step": 19603 }, { "epoch": 2.33, "learning_rate": 6.177141315160847e-07, "loss": 0.4832, "step": 19604 }, { "epoch": 2.33, "learning_rate": 6.175026123587255e-07, "loss": 0.5508, "step": 19605 }, { "epoch": 2.33, "learning_rate": 6.172911243192051e-07, "loss": 0.682, "step": 19606 }, { "epoch": 2.34, "learning_rate": 6.170796674010202e-07, "loss": 0.331, "step": 19607 }, { "epoch": 2.34, "learning_rate": 6.168682416076666e-07, "loss": 0.5123, "step": 19608 }, { "epoch": 2.34, "learning_rate": 6.166568469426396e-07, "loss": 0.4269, "step": 19609 }, { "epoch": 2.34, "learning_rate": 6.16445483409433e-07, "loss": 0.5047, "step": 19610 }, { "epoch": 2.34, "learning_rate": 6.162341510115399e-07, "loss": 0.5432, "step": 19611 }, { "epoch": 2.34, "learning_rate": 6.16022849752455e-07, "loss": 0.6033, "step": 19612 }, { "epoch": 2.34, "learning_rate": 6.1581157963567e-07, "loss": 0.2681, "step": 19613 }, { "epoch": 2.34, "learning_rate": 6.156003406646774e-07, "loss": 0.3692, "step": 19614 }, { "epoch": 2.34, "learning_rate": 6.153891328429692e-07, "loss": 0.403, "step": 19615 }, { "epoch": 2.34, "learning_rate": 6.151779561740379e-07, "loss": 0.5051, "step": 19616 }, { "epoch": 2.34, "learning_rate": 6.149668106613721e-07, "loss": 0.5032, "step": 19617 }, { "epoch": 2.34, "learning_rate": 6.147556963084639e-07, "loss": 0.4947, "step": 19618 }, { "epoch": 2.34, "learning_rate": 6.145446131188015e-07, "loss": 0.4527, "step": 19619 }, { "epoch": 2.34, "learning_rate": 6.143335610958756e-07, "loss": 0.2795, "step": 19620 }, { "epoch": 2.34, "learning_rate": 6.141225402431733e-07, "loss": 0.3514, "step": 19621 }, { "epoch": 2.34, "learning_rate": 6.139115505641838e-07, "loss": 0.3797, "step": 19622 }, { "epoch": 2.34, "learning_rate": 6.137005920623945e-07, "loss": 0.414, "step": 19623 }, { "epoch": 2.34, "learning_rate": 6.134896647412936e-07, "loss": 0.5541, "step": 19624 }, { "epoch": 2.34, "learning_rate": 6.132787686043665e-07, "loss": 0.6039, "step": 19625 }, { "epoch": 2.34, "learning_rate": 6.130679036550993e-07, "loss": 0.316, "step": 19626 }, { "epoch": 2.34, "learning_rate": 6.128570698969786e-07, "loss": 0.4561, "step": 19627 }, { "epoch": 2.34, "learning_rate": 6.12646267333488e-07, "loss": 0.4346, "step": 19628 }, { "epoch": 2.34, "learning_rate": 6.124354959681131e-07, "loss": 0.4197, "step": 19629 }, { "epoch": 2.34, "learning_rate": 6.122247558043376e-07, "loss": 0.534, "step": 19630 }, { "epoch": 2.34, "learning_rate": 6.120140468456462e-07, "loss": 0.6609, "step": 19631 }, { "epoch": 2.34, "learning_rate": 6.11803369095521e-07, "loss": 0.3109, "step": 19632 }, { "epoch": 2.34, "learning_rate": 6.115927225574439e-07, "loss": 0.4583, "step": 19633 }, { "epoch": 2.34, "learning_rate": 6.113821072348975e-07, "loss": 0.4416, "step": 19634 }, { "epoch": 2.34, "learning_rate": 6.111715231313642e-07, "loss": 0.5786, "step": 19635 }, { "epoch": 2.34, "learning_rate": 6.10960970250323e-07, "loss": 0.594, "step": 19636 }, { "epoch": 2.34, "learning_rate": 6.107504485952559e-07, "loss": 0.5555, "step": 19637 }, { "epoch": 2.34, "learning_rate": 6.105399581696428e-07, "loss": 0.2944, "step": 19638 }, { "epoch": 2.34, "learning_rate": 6.103294989769629e-07, "loss": 0.3687, "step": 19639 }, { "epoch": 2.34, "learning_rate": 6.101190710206942e-07, "loss": 0.4256, "step": 19640 }, { "epoch": 2.34, "learning_rate": 6.099086743043159e-07, "loss": 0.5217, "step": 19641 }, { "epoch": 2.34, "learning_rate": 6.096983088313055e-07, "loss": 0.5601, "step": 19642 }, { "epoch": 2.34, "learning_rate": 6.094879746051416e-07, "loss": 0.6329, "step": 19643 }, { "epoch": 2.34, "learning_rate": 6.092776716292995e-07, "loss": 0.487, "step": 19644 }, { "epoch": 2.34, "learning_rate": 6.090673999072566e-07, "loss": 0.3118, "step": 19645 }, { "epoch": 2.34, "learning_rate": 6.088571594424877e-07, "loss": 0.3915, "step": 19646 }, { "epoch": 2.34, "learning_rate": 6.086469502384692e-07, "loss": 0.397, "step": 19647 }, { "epoch": 2.34, "learning_rate": 6.084367722986748e-07, "loss": 0.4753, "step": 19648 }, { "epoch": 2.34, "learning_rate": 6.082266256265793e-07, "loss": 0.5753, "step": 19649 }, { "epoch": 2.34, "learning_rate": 6.080165102256569e-07, "loss": 0.5717, "step": 19650 }, { "epoch": 2.34, "learning_rate": 6.078064260993797e-07, "loss": 0.2952, "step": 19651 }, { "epoch": 2.34, "learning_rate": 6.075963732512211e-07, "loss": 0.3786, "step": 19652 }, { "epoch": 2.34, "learning_rate": 6.07386351684654e-07, "loss": 0.5044, "step": 19653 }, { "epoch": 2.34, "learning_rate": 6.071763614031492e-07, "loss": 0.5572, "step": 19654 }, { "epoch": 2.34, "learning_rate": 6.06966402410177e-07, "loss": 0.5552, "step": 19655 }, { "epoch": 2.34, "learning_rate": 6.067564747092095e-07, "loss": 0.6487, "step": 19656 }, { "epoch": 2.34, "learning_rate": 6.065465783037161e-07, "loss": 0.29, "step": 19657 }, { "epoch": 2.34, "learning_rate": 6.063367131971674e-07, "loss": 0.3688, "step": 19658 }, { "epoch": 2.34, "learning_rate": 6.061268793930308e-07, "loss": 0.4161, "step": 19659 }, { "epoch": 2.34, "learning_rate": 6.059170768947767e-07, "loss": 0.4785, "step": 19660 }, { "epoch": 2.34, "learning_rate": 6.057073057058724e-07, "loss": 0.5902, "step": 19661 }, { "epoch": 2.34, "learning_rate": 6.054975658297843e-07, "loss": 0.6137, "step": 19662 }, { "epoch": 2.34, "learning_rate": 6.052878572699805e-07, "loss": 0.3714, "step": 19663 }, { "epoch": 2.34, "learning_rate": 6.050781800299275e-07, "loss": 0.3735, "step": 19664 }, { "epoch": 2.34, "learning_rate": 6.048685341130916e-07, "loss": 0.3698, "step": 19665 }, { "epoch": 2.34, "learning_rate": 6.046589195229375e-07, "loss": 0.3654, "step": 19666 }, { "epoch": 2.34, "learning_rate": 6.044493362629308e-07, "loss": 0.5445, "step": 19667 }, { "epoch": 2.34, "learning_rate": 6.042397843365353e-07, "loss": 0.57, "step": 19668 }, { "epoch": 2.34, "learning_rate": 6.040302637472157e-07, "loss": 0.6214, "step": 19669 }, { "epoch": 2.34, "learning_rate": 6.038207744984342e-07, "loss": 0.332, "step": 19670 }, { "epoch": 2.34, "learning_rate": 6.036113165936544e-07, "loss": 0.4089, "step": 19671 }, { "epoch": 2.34, "learning_rate": 6.03401890036339e-07, "loss": 0.4508, "step": 19672 }, { "epoch": 2.34, "learning_rate": 6.031924948299498e-07, "loss": 0.5042, "step": 19673 }, { "epoch": 2.34, "learning_rate": 6.029831309779474e-07, "loss": 0.5712, "step": 19674 }, { "epoch": 2.34, "learning_rate": 6.027737984837934e-07, "loss": 0.5954, "step": 19675 }, { "epoch": 2.34, "learning_rate": 6.025644973509479e-07, "loss": 0.3015, "step": 19676 }, { "epoch": 2.34, "learning_rate": 6.023552275828698e-07, "loss": 0.3616, "step": 19677 }, { "epoch": 2.34, "learning_rate": 6.021459891830189e-07, "loss": 0.4075, "step": 19678 }, { "epoch": 2.34, "learning_rate": 6.019367821548542e-07, "loss": 0.4844, "step": 19679 }, { "epoch": 2.34, "learning_rate": 6.017276065018343e-07, "loss": 0.6341, "step": 19680 }, { "epoch": 2.34, "learning_rate": 6.015184622274156e-07, "loss": 0.5843, "step": 19681 }, { "epoch": 2.34, "learning_rate": 6.013093493350569e-07, "loss": 0.3135, "step": 19682 }, { "epoch": 2.34, "learning_rate": 6.011002678282132e-07, "loss": 0.3968, "step": 19683 }, { "epoch": 2.34, "learning_rate": 6.008912177103421e-07, "loss": 0.354, "step": 19684 }, { "epoch": 2.34, "learning_rate": 6.006821989848977e-07, "loss": 0.4878, "step": 19685 }, { "epoch": 2.34, "learning_rate": 6.004732116553361e-07, "loss": 0.5858, "step": 19686 }, { "epoch": 2.34, "learning_rate": 6.002642557251115e-07, "loss": 0.6779, "step": 19687 }, { "epoch": 2.34, "learning_rate": 6.000553311976789e-07, "loss": 0.4034, "step": 19688 }, { "epoch": 2.34, "learning_rate": 5.998464380764913e-07, "loss": 0.2919, "step": 19689 }, { "epoch": 2.34, "learning_rate": 5.996375763650006e-07, "loss": 0.401, "step": 19690 }, { "epoch": 2.35, "learning_rate": 5.994287460666612e-07, "loss": 0.445, "step": 19691 }, { "epoch": 2.35, "learning_rate": 5.99219947184923e-07, "loss": 0.5651, "step": 19692 }, { "epoch": 2.35, "learning_rate": 5.990111797232387e-07, "loss": 0.6803, "step": 19693 }, { "epoch": 2.35, "learning_rate": 5.988024436850592e-07, "loss": 0.5171, "step": 19694 }, { "epoch": 2.35, "learning_rate": 5.985937390738356e-07, "loss": 0.3649, "step": 19695 }, { "epoch": 2.35, "learning_rate": 5.983850658930168e-07, "loss": 0.3642, "step": 19696 }, { "epoch": 2.35, "learning_rate": 5.98176424146052e-07, "loss": 0.4753, "step": 19697 }, { "epoch": 2.35, "learning_rate": 5.979678138363904e-07, "loss": 0.5209, "step": 19698 }, { "epoch": 2.35, "learning_rate": 5.97759234967481e-07, "loss": 0.5959, "step": 19699 }, { "epoch": 2.35, "learning_rate": 5.975506875427703e-07, "loss": 0.6444, "step": 19700 }, { "epoch": 2.35, "learning_rate": 5.973421715657066e-07, "loss": 0.4706, "step": 19701 }, { "epoch": 2.35, "learning_rate": 5.971336870397368e-07, "loss": 0.4847, "step": 19702 }, { "epoch": 2.35, "learning_rate": 5.969252339683066e-07, "loss": 0.3506, "step": 19703 }, { "epoch": 2.35, "learning_rate": 5.967168123548623e-07, "loss": 0.5035, "step": 19704 }, { "epoch": 2.35, "learning_rate": 5.965084222028481e-07, "loss": 0.5402, "step": 19705 }, { "epoch": 2.35, "learning_rate": 5.963000635157096e-07, "loss": 0.6066, "step": 19706 }, { "epoch": 2.35, "learning_rate": 5.960917362968913e-07, "loss": 0.356, "step": 19707 }, { "epoch": 2.35, "learning_rate": 5.958834405498356e-07, "loss": 0.3038, "step": 19708 }, { "epoch": 2.35, "learning_rate": 5.956751762779867e-07, "loss": 0.4907, "step": 19709 }, { "epoch": 2.35, "learning_rate": 5.954669434847873e-07, "loss": 0.4571, "step": 19710 }, { "epoch": 2.35, "learning_rate": 5.952587421736794e-07, "loss": 0.5406, "step": 19711 }, { "epoch": 2.35, "learning_rate": 5.950505723481034e-07, "loss": 0.6457, "step": 19712 }, { "epoch": 2.35, "learning_rate": 5.948424340115014e-07, "loss": 0.534, "step": 19713 }, { "epoch": 2.35, "learning_rate": 5.946343271673144e-07, "loss": 0.3313, "step": 19714 }, { "epoch": 2.35, "learning_rate": 5.944262518189813e-07, "loss": 0.3504, "step": 19715 }, { "epoch": 2.35, "learning_rate": 5.942182079699421e-07, "loss": 0.4971, "step": 19716 }, { "epoch": 2.35, "learning_rate": 5.940101956236366e-07, "loss": 0.5005, "step": 19717 }, { "epoch": 2.35, "learning_rate": 5.938022147835023e-07, "loss": 0.6806, "step": 19718 }, { "epoch": 2.35, "learning_rate": 5.935942654529767e-07, "loss": 0.5899, "step": 19719 }, { "epoch": 2.35, "learning_rate": 5.933863476354978e-07, "loss": 0.2599, "step": 19720 }, { "epoch": 2.35, "learning_rate": 5.931784613345029e-07, "loss": 0.4437, "step": 19721 }, { "epoch": 2.35, "learning_rate": 5.929706065534282e-07, "loss": 0.4487, "step": 19722 }, { "epoch": 2.35, "learning_rate": 5.927627832957092e-07, "loss": 0.4695, "step": 19723 }, { "epoch": 2.35, "learning_rate": 5.925549915647818e-07, "loss": 0.5794, "step": 19724 }, { "epoch": 2.35, "learning_rate": 5.923472313640804e-07, "loss": 0.4123, "step": 19725 }, { "epoch": 2.35, "learning_rate": 5.921395026970389e-07, "loss": 0.35, "step": 19726 }, { "epoch": 2.35, "learning_rate": 5.919318055670914e-07, "loss": 0.3401, "step": 19727 }, { "epoch": 2.35, "learning_rate": 5.91724139977671e-07, "loss": 0.4212, "step": 19728 }, { "epoch": 2.35, "learning_rate": 5.915165059322118e-07, "loss": 0.4912, "step": 19729 }, { "epoch": 2.35, "learning_rate": 5.913089034341437e-07, "loss": 0.5856, "step": 19730 }, { "epoch": 2.35, "learning_rate": 5.911013324869008e-07, "loss": 0.6158, "step": 19731 }, { "epoch": 2.35, "learning_rate": 5.908937930939121e-07, "loss": 0.3689, "step": 19732 }, { "epoch": 2.35, "learning_rate": 5.906862852586098e-07, "loss": 0.3988, "step": 19733 }, { "epoch": 2.35, "learning_rate": 5.90478808984423e-07, "loss": 0.3381, "step": 19734 }, { "epoch": 2.35, "learning_rate": 5.902713642747812e-07, "loss": 0.4796, "step": 19735 }, { "epoch": 2.35, "learning_rate": 5.900639511331144e-07, "loss": 0.5712, "step": 19736 }, { "epoch": 2.35, "learning_rate": 5.898565695628513e-07, "loss": 0.6154, "step": 19737 }, { "epoch": 2.35, "learning_rate": 5.896492195674189e-07, "loss": 0.3674, "step": 19738 }, { "epoch": 2.35, "learning_rate": 5.894419011502456e-07, "loss": 0.359, "step": 19739 }, { "epoch": 2.35, "learning_rate": 5.892346143147582e-07, "loss": 0.3633, "step": 19740 }, { "epoch": 2.35, "learning_rate": 5.890273590643822e-07, "loss": 0.4591, "step": 19741 }, { "epoch": 2.35, "learning_rate": 5.888201354025441e-07, "loss": 0.5668, "step": 19742 }, { "epoch": 2.35, "learning_rate": 5.886129433326696e-07, "loss": 0.623, "step": 19743 }, { "epoch": 2.35, "learning_rate": 5.884057828581843e-07, "loss": 0.5184, "step": 19744 }, { "epoch": 2.35, "learning_rate": 5.88198653982511e-07, "loss": 0.3785, "step": 19745 }, { "epoch": 2.35, "learning_rate": 5.879915567090752e-07, "loss": 0.4166, "step": 19746 }, { "epoch": 2.35, "learning_rate": 5.877844910412989e-07, "loss": 0.4257, "step": 19747 }, { "epoch": 2.35, "learning_rate": 5.875774569826056e-07, "loss": 0.5587, "step": 19748 }, { "epoch": 2.35, "learning_rate": 5.87370454536417e-07, "loss": 0.5409, "step": 19749 }, { "epoch": 2.35, "learning_rate": 5.871634837061554e-07, "loss": 0.5852, "step": 19750 }, { "epoch": 2.35, "learning_rate": 5.869565444952416e-07, "loss": 0.3761, "step": 19751 }, { "epoch": 2.35, "learning_rate": 5.867496369070977e-07, "loss": 0.4765, "step": 19752 }, { "epoch": 2.35, "learning_rate": 5.865427609451427e-07, "loss": 0.4159, "step": 19753 }, { "epoch": 2.35, "learning_rate": 5.863359166127955e-07, "loss": 0.514, "step": 19754 }, { "epoch": 2.35, "learning_rate": 5.86129103913477e-07, "loss": 0.5887, "step": 19755 }, { "epoch": 2.35, "learning_rate": 5.859223228506045e-07, "loss": 0.583, "step": 19756 }, { "epoch": 2.35, "learning_rate": 5.857155734275962e-07, "loss": 0.3273, "step": 19757 }, { "epoch": 2.35, "learning_rate": 5.855088556478703e-07, "loss": 0.4048, "step": 19758 }, { "epoch": 2.35, "learning_rate": 5.853021695148445e-07, "loss": 0.4131, "step": 19759 }, { "epoch": 2.35, "learning_rate": 5.850955150319334e-07, "loss": 0.4462, "step": 19760 }, { "epoch": 2.35, "learning_rate": 5.848888922025553e-07, "loss": 0.5455, "step": 19761 }, { "epoch": 2.35, "learning_rate": 5.846823010301236e-07, "loss": 0.631, "step": 19762 }, { "epoch": 2.35, "learning_rate": 5.844757415180549e-07, "loss": 0.4565, "step": 19763 }, { "epoch": 2.35, "learning_rate": 5.84269213669762e-07, "loss": 0.3364, "step": 19764 }, { "epoch": 2.35, "learning_rate": 5.840627174886599e-07, "loss": 0.3599, "step": 19765 }, { "epoch": 2.35, "learning_rate": 5.838562529781619e-07, "loss": 0.3831, "step": 19766 }, { "epoch": 2.35, "learning_rate": 5.836498201416815e-07, "loss": 0.6028, "step": 19767 }, { "epoch": 2.35, "learning_rate": 5.834434189826302e-07, "loss": 0.7039, "step": 19768 }, { "epoch": 2.35, "learning_rate": 5.832370495044196e-07, "loss": 0.5328, "step": 19769 }, { "epoch": 2.35, "learning_rate": 5.830307117104619e-07, "loss": 0.3031, "step": 19770 }, { "epoch": 2.35, "learning_rate": 5.828244056041669e-07, "loss": 0.3776, "step": 19771 }, { "epoch": 2.35, "learning_rate": 5.826181311889451e-07, "loss": 0.4647, "step": 19772 }, { "epoch": 2.35, "learning_rate": 5.824118884682064e-07, "loss": 0.5237, "step": 19773 }, { "epoch": 2.35, "learning_rate": 5.822056774453608e-07, "loss": 0.6655, "step": 19774 }, { "epoch": 2.36, "learning_rate": 5.819994981238164e-07, "loss": 0.5849, "step": 19775 }, { "epoch": 2.36, "learning_rate": 5.817933505069806e-07, "loss": 0.331, "step": 19776 }, { "epoch": 2.36, "learning_rate": 5.815872345982615e-07, "loss": 0.4013, "step": 19777 }, { "epoch": 2.36, "learning_rate": 5.813811504010672e-07, "loss": 0.4282, "step": 19778 }, { "epoch": 2.36, "learning_rate": 5.811750979188027e-07, "loss": 0.5482, "step": 19779 }, { "epoch": 2.36, "learning_rate": 5.809690771548749e-07, "loss": 0.5616, "step": 19780 }, { "epoch": 2.36, "learning_rate": 5.8076308811269e-07, "loss": 0.7395, "step": 19781 }, { "epoch": 2.36, "learning_rate": 5.805571307956521e-07, "loss": 0.4117, "step": 19782 }, { "epoch": 2.36, "learning_rate": 5.803512052071652e-07, "loss": 0.4192, "step": 19783 }, { "epoch": 2.36, "learning_rate": 5.801453113506342e-07, "loss": 0.4053, "step": 19784 }, { "epoch": 2.36, "learning_rate": 5.799394492294622e-07, "loss": 0.4577, "step": 19785 }, { "epoch": 2.36, "learning_rate": 5.797336188470529e-07, "loss": 0.504, "step": 19786 }, { "epoch": 2.36, "learning_rate": 5.795278202068072e-07, "loss": 0.6865, "step": 19787 }, { "epoch": 2.36, "learning_rate": 5.793220533121285e-07, "loss": 0.4614, "step": 19788 }, { "epoch": 2.36, "learning_rate": 5.791163181664167e-07, "loss": 0.3697, "step": 19789 }, { "epoch": 2.36, "learning_rate": 5.789106147730741e-07, "loss": 0.4617, "step": 19790 }, { "epoch": 2.36, "learning_rate": 5.787049431354996e-07, "loss": 0.4472, "step": 19791 }, { "epoch": 2.36, "learning_rate": 5.784993032570935e-07, "loss": 0.4576, "step": 19792 }, { "epoch": 2.36, "learning_rate": 5.782936951412557e-07, "loss": 0.5714, "step": 19793 }, { "epoch": 2.36, "learning_rate": 5.78088118791384e-07, "loss": 0.5752, "step": 19794 }, { "epoch": 2.36, "learning_rate": 5.778825742108768e-07, "loss": 0.3427, "step": 19795 }, { "epoch": 2.36, "learning_rate": 5.776770614031324e-07, "loss": 0.3782, "step": 19796 }, { "epoch": 2.36, "learning_rate": 5.774715803715477e-07, "loss": 0.5309, "step": 19797 }, { "epoch": 2.36, "learning_rate": 5.772661311195185e-07, "loss": 0.5621, "step": 19798 }, { "epoch": 2.36, "learning_rate": 5.770607136504413e-07, "loss": 0.5434, "step": 19799 }, { "epoch": 2.36, "learning_rate": 5.768553279677119e-07, "loss": 0.5363, "step": 19800 }, { "epoch": 2.36, "learning_rate": 5.766499740747261e-07, "loss": 0.2916, "step": 19801 }, { "epoch": 2.36, "learning_rate": 5.764446519748768e-07, "loss": 0.4186, "step": 19802 }, { "epoch": 2.36, "learning_rate": 5.762393616715594e-07, "loss": 0.4802, "step": 19803 }, { "epoch": 2.36, "learning_rate": 5.760341031681669e-07, "loss": 0.4677, "step": 19804 }, { "epoch": 2.36, "learning_rate": 5.758288764680914e-07, "loss": 0.5197, "step": 19805 }, { "epoch": 2.36, "learning_rate": 5.75623681574726e-07, "loss": 0.5226, "step": 19806 }, { "epoch": 2.36, "learning_rate": 5.754185184914627e-07, "loss": 0.2919, "step": 19807 }, { "epoch": 2.36, "learning_rate": 5.752133872216934e-07, "loss": 0.4028, "step": 19808 }, { "epoch": 2.36, "learning_rate": 5.750082877688076e-07, "loss": 0.4299, "step": 19809 }, { "epoch": 2.36, "learning_rate": 5.74803220136197e-07, "loss": 0.501, "step": 19810 }, { "epoch": 2.36, "learning_rate": 5.7459818432725e-07, "loss": 0.4838, "step": 19811 }, { "epoch": 2.36, "learning_rate": 5.743931803453572e-07, "loss": 0.6381, "step": 19812 }, { "epoch": 2.36, "learning_rate": 5.741882081939062e-07, "loss": 0.398, "step": 19813 }, { "epoch": 2.36, "learning_rate": 5.739832678762858e-07, "loss": 0.314, "step": 19814 }, { "epoch": 2.36, "learning_rate": 5.737783593958835e-07, "loss": 0.3932, "step": 19815 }, { "epoch": 2.36, "learning_rate": 5.735734827560871e-07, "loss": 0.4455, "step": 19816 }, { "epoch": 2.36, "learning_rate": 5.733686379602818e-07, "loss": 0.5125, "step": 19817 }, { "epoch": 2.36, "learning_rate": 5.731638250118556e-07, "loss": 0.619, "step": 19818 }, { "epoch": 2.36, "learning_rate": 5.72959043914193e-07, "loss": 0.5453, "step": 19819 }, { "epoch": 2.36, "learning_rate": 5.727542946706782e-07, "loss": 0.3299, "step": 19820 }, { "epoch": 2.36, "learning_rate": 5.725495772846967e-07, "loss": 0.3729, "step": 19821 }, { "epoch": 2.36, "learning_rate": 5.723448917596325e-07, "loss": 0.4119, "step": 19822 }, { "epoch": 2.36, "learning_rate": 5.721402380988697e-07, "loss": 0.4606, "step": 19823 }, { "epoch": 2.36, "learning_rate": 5.719356163057899e-07, "loss": 0.5681, "step": 19824 }, { "epoch": 2.36, "learning_rate": 5.717310263837766e-07, "loss": 0.5905, "step": 19825 }, { "epoch": 2.36, "learning_rate": 5.715264683362107e-07, "loss": 0.3683, "step": 19826 }, { "epoch": 2.36, "learning_rate": 5.713219421664748e-07, "loss": 0.3776, "step": 19827 }, { "epoch": 2.36, "learning_rate": 5.711174478779483e-07, "loss": 0.3539, "step": 19828 }, { "epoch": 2.36, "learning_rate": 5.709129854740122e-07, "loss": 0.4559, "step": 19829 }, { "epoch": 2.36, "learning_rate": 5.707085549580463e-07, "loss": 0.5881, "step": 19830 }, { "epoch": 2.36, "learning_rate": 5.705041563334304e-07, "loss": 0.6003, "step": 19831 }, { "epoch": 2.36, "learning_rate": 5.702997896035431e-07, "loss": 0.4053, "step": 19832 }, { "epoch": 2.36, "learning_rate": 5.70095454771761e-07, "loss": 0.438, "step": 19833 }, { "epoch": 2.36, "learning_rate": 5.69891151841464e-07, "loss": 0.4733, "step": 19834 }, { "epoch": 2.36, "learning_rate": 5.696868808160275e-07, "loss": 0.4199, "step": 19835 }, { "epoch": 2.36, "learning_rate": 5.694826416988289e-07, "loss": 0.4833, "step": 19836 }, { "epoch": 2.36, "learning_rate": 5.692784344932439e-07, "loss": 0.6088, "step": 19837 }, { "epoch": 2.36, "learning_rate": 5.690742592026493e-07, "loss": 0.5889, "step": 19838 }, { "epoch": 2.36, "learning_rate": 5.688701158304191e-07, "loss": 0.3876, "step": 19839 }, { "epoch": 2.36, "learning_rate": 5.686660043799275e-07, "loss": 0.3773, "step": 19840 }, { "epoch": 2.36, "learning_rate": 5.684619248545487e-07, "loss": 0.4366, "step": 19841 }, { "epoch": 2.36, "learning_rate": 5.68257877257657e-07, "loss": 0.5199, "step": 19842 }, { "epoch": 2.36, "learning_rate": 5.68053861592624e-07, "loss": 0.4649, "step": 19843 }, { "epoch": 2.36, "learning_rate": 5.678498778628228e-07, "loss": 0.5653, "step": 19844 }, { "epoch": 2.36, "learning_rate": 5.676459260716258e-07, "loss": 0.3457, "step": 19845 }, { "epoch": 2.36, "learning_rate": 5.674420062224034e-07, "loss": 0.4367, "step": 19846 }, { "epoch": 2.36, "learning_rate": 5.672381183185271e-07, "loss": 0.419, "step": 19847 }, { "epoch": 2.36, "learning_rate": 5.670342623633665e-07, "loss": 0.4487, "step": 19848 }, { "epoch": 2.36, "learning_rate": 5.668304383602916e-07, "loss": 0.6426, "step": 19849 }, { "epoch": 2.36, "learning_rate": 5.666266463126724e-07, "loss": 0.5749, "step": 19850 }, { "epoch": 2.36, "learning_rate": 5.664228862238763e-07, "loss": 0.2756, "step": 19851 }, { "epoch": 2.36, "learning_rate": 5.662191580972723e-07, "loss": 0.4308, "step": 19852 }, { "epoch": 2.36, "learning_rate": 5.660154619362285e-07, "loss": 0.455, "step": 19853 }, { "epoch": 2.36, "learning_rate": 5.658117977441113e-07, "loss": 0.5179, "step": 19854 }, { "epoch": 2.36, "learning_rate": 5.656081655242867e-07, "loss": 0.6156, "step": 19855 }, { "epoch": 2.36, "learning_rate": 5.654045652801213e-07, "loss": 0.5563, "step": 19856 }, { "epoch": 2.36, "learning_rate": 5.652009970149816e-07, "loss": 0.3761, "step": 19857 }, { "epoch": 2.36, "learning_rate": 5.649974607322312e-07, "loss": 0.3739, "step": 19858 }, { "epoch": 2.37, "learning_rate": 5.64793956435235e-07, "loss": 0.4431, "step": 19859 }, { "epoch": 2.37, "learning_rate": 5.645904841273578e-07, "loss": 0.5005, "step": 19860 }, { "epoch": 2.37, "learning_rate": 5.643870438119622e-07, "loss": 0.5331, "step": 19861 }, { "epoch": 2.37, "learning_rate": 5.641836354924105e-07, "loss": 0.5729, "step": 19862 }, { "epoch": 2.37, "learning_rate": 5.639802591720658e-07, "loss": 0.4947, "step": 19863 }, { "epoch": 2.37, "learning_rate": 5.637769148542899e-07, "loss": 0.4564, "step": 19864 }, { "epoch": 2.37, "learning_rate": 5.635736025424446e-07, "loss": 0.3928, "step": 19865 }, { "epoch": 2.37, "learning_rate": 5.633703222398895e-07, "loss": 0.4277, "step": 19866 }, { "epoch": 2.37, "learning_rate": 5.631670739499862e-07, "loss": 0.507, "step": 19867 }, { "epoch": 2.37, "learning_rate": 5.629638576760938e-07, "loss": 0.6283, "step": 19868 }, { "epoch": 2.37, "learning_rate": 5.627606734215704e-07, "loss": 0.4961, "step": 19869 }, { "epoch": 2.37, "learning_rate": 5.625575211897757e-07, "loss": 0.329, "step": 19870 }, { "epoch": 2.37, "learning_rate": 5.623544009840678e-07, "loss": 0.404, "step": 19871 }, { "epoch": 2.37, "learning_rate": 5.621513128078051e-07, "loss": 0.4116, "step": 19872 }, { "epoch": 2.37, "learning_rate": 5.619482566643428e-07, "loss": 0.4299, "step": 19873 }, { "epoch": 2.37, "learning_rate": 5.617452325570385e-07, "loss": 0.6048, "step": 19874 }, { "epoch": 2.37, "learning_rate": 5.615422404892489e-07, "loss": 0.6809, "step": 19875 }, { "epoch": 2.37, "learning_rate": 5.613392804643289e-07, "loss": 0.2977, "step": 19876 }, { "epoch": 2.37, "learning_rate": 5.611363524856322e-07, "loss": 0.4218, "step": 19877 }, { "epoch": 2.37, "learning_rate": 5.609334565565148e-07, "loss": 0.3653, "step": 19878 }, { "epoch": 2.37, "learning_rate": 5.607305926803294e-07, "loss": 0.4287, "step": 19879 }, { "epoch": 2.37, "learning_rate": 5.605277608604312e-07, "loss": 0.5059, "step": 19880 }, { "epoch": 2.37, "learning_rate": 5.603249611001709e-07, "loss": 0.6186, "step": 19881 }, { "epoch": 2.37, "learning_rate": 5.601221934029024e-07, "loss": 0.3216, "step": 19882 }, { "epoch": 2.37, "learning_rate": 5.599194577719771e-07, "loss": 0.3038, "step": 19883 }, { "epoch": 2.37, "learning_rate": 5.597167542107451e-07, "loss": 0.5611, "step": 19884 }, { "epoch": 2.37, "learning_rate": 5.595140827225579e-07, "loss": 0.4375, "step": 19885 }, { "epoch": 2.37, "learning_rate": 5.593114433107658e-07, "loss": 0.5766, "step": 19886 }, { "epoch": 2.37, "learning_rate": 5.591088359787192e-07, "loss": 0.5912, "step": 19887 }, { "epoch": 2.37, "learning_rate": 5.589062607297655e-07, "loss": 0.3884, "step": 19888 }, { "epoch": 2.37, "learning_rate": 5.58703717567255e-07, "loss": 0.3151, "step": 19889 }, { "epoch": 2.37, "learning_rate": 5.585012064945341e-07, "loss": 0.3928, "step": 19890 }, { "epoch": 2.37, "learning_rate": 5.58298727514952e-07, "loss": 0.4611, "step": 19891 }, { "epoch": 2.37, "learning_rate": 5.58096280631854e-07, "loss": 0.5413, "step": 19892 }, { "epoch": 2.37, "learning_rate": 5.578938658485874e-07, "loss": 0.679, "step": 19893 }, { "epoch": 2.37, "learning_rate": 5.576914831684985e-07, "loss": 0.5883, "step": 19894 }, { "epoch": 2.37, "learning_rate": 5.574891325949328e-07, "loss": 0.3165, "step": 19895 }, { "epoch": 2.37, "learning_rate": 5.572868141312346e-07, "loss": 0.3591, "step": 19896 }, { "epoch": 2.37, "learning_rate": 5.57084527780748e-07, "loss": 0.4289, "step": 19897 }, { "epoch": 2.37, "learning_rate": 5.568822735468177e-07, "loss": 0.4077, "step": 19898 }, { "epoch": 2.37, "learning_rate": 5.566800514327858e-07, "loss": 0.5929, "step": 19899 }, { "epoch": 2.37, "learning_rate": 5.564778614419955e-07, "loss": 0.6053, "step": 19900 }, { "epoch": 2.37, "learning_rate": 5.562757035777894e-07, "loss": 0.3881, "step": 19901 }, { "epoch": 2.37, "learning_rate": 5.560735778435097e-07, "loss": 0.4183, "step": 19902 }, { "epoch": 2.37, "learning_rate": 5.558714842424962e-07, "loss": 0.4338, "step": 19903 }, { "epoch": 2.37, "learning_rate": 5.556694227780909e-07, "loss": 0.4798, "step": 19904 }, { "epoch": 2.37, "learning_rate": 5.554673934536325e-07, "loss": 0.4945, "step": 19905 }, { "epoch": 2.37, "learning_rate": 5.552653962724619e-07, "loss": 0.55, "step": 19906 }, { "epoch": 2.37, "learning_rate": 5.550634312379166e-07, "loss": 0.3514, "step": 19907 }, { "epoch": 2.37, "learning_rate": 5.548614983533363e-07, "loss": 0.3379, "step": 19908 }, { "epoch": 2.37, "learning_rate": 5.546595976220584e-07, "loss": 0.4343, "step": 19909 }, { "epoch": 2.37, "learning_rate": 5.544577290474215e-07, "loss": 0.4514, "step": 19910 }, { "epoch": 2.37, "learning_rate": 5.542558926327613e-07, "loss": 0.4598, "step": 19911 }, { "epoch": 2.37, "learning_rate": 5.540540883814139e-07, "loss": 0.6758, "step": 19912 }, { "epoch": 2.37, "learning_rate": 5.538523162967166e-07, "loss": 0.4324, "step": 19913 }, { "epoch": 2.37, "learning_rate": 5.536505763820027e-07, "loss": 0.3148, "step": 19914 }, { "epoch": 2.37, "learning_rate": 5.534488686406084e-07, "loss": 0.4458, "step": 19915 }, { "epoch": 2.37, "learning_rate": 5.532471930758676e-07, "loss": 0.4387, "step": 19916 }, { "epoch": 2.37, "learning_rate": 5.530455496911147e-07, "loss": 0.5385, "step": 19917 }, { "epoch": 2.37, "learning_rate": 5.528439384896822e-07, "loss": 0.5557, "step": 19918 }, { "epoch": 2.37, "learning_rate": 5.52642359474902e-07, "loss": 0.5714, "step": 19919 }, { "epoch": 2.37, "learning_rate": 5.524408126501071e-07, "loss": 0.242, "step": 19920 }, { "epoch": 2.37, "learning_rate": 5.522392980186295e-07, "loss": 0.4016, "step": 19921 }, { "epoch": 2.37, "learning_rate": 5.520378155837994e-07, "loss": 0.3549, "step": 19922 }, { "epoch": 2.37, "learning_rate": 5.518363653489472e-07, "loss": 0.4998, "step": 19923 }, { "epoch": 2.37, "learning_rate": 5.516349473174043e-07, "loss": 0.5352, "step": 19924 }, { "epoch": 2.37, "learning_rate": 5.514335614924993e-07, "loss": 0.5291, "step": 19925 }, { "epoch": 2.37, "learning_rate": 5.512322078775603e-07, "loss": 0.3574, "step": 19926 }, { "epoch": 2.37, "learning_rate": 5.510308864759162e-07, "loss": 0.4258, "step": 19927 }, { "epoch": 2.37, "learning_rate": 5.508295972908953e-07, "loss": 0.3373, "step": 19928 }, { "epoch": 2.37, "learning_rate": 5.506283403258253e-07, "loss": 0.4503, "step": 19929 }, { "epoch": 2.37, "learning_rate": 5.504271155840316e-07, "loss": 0.4865, "step": 19930 }, { "epoch": 2.37, "learning_rate": 5.502259230688422e-07, "loss": 0.5018, "step": 19931 }, { "epoch": 2.37, "learning_rate": 5.500247627835811e-07, "loss": 0.3158, "step": 19932 }, { "epoch": 2.37, "learning_rate": 5.498236347315749e-07, "loss": 0.5409, "step": 19933 }, { "epoch": 2.37, "learning_rate": 5.496225389161469e-07, "loss": 0.3954, "step": 19934 }, { "epoch": 2.37, "learning_rate": 5.494214753406221e-07, "loss": 0.4476, "step": 19935 }, { "epoch": 2.37, "learning_rate": 5.492204440083249e-07, "loss": 0.582, "step": 19936 }, { "epoch": 2.37, "learning_rate": 5.490194449225764e-07, "loss": 0.6009, "step": 19937 }, { "epoch": 2.37, "learning_rate": 5.488184780867003e-07, "loss": 0.4472, "step": 19938 }, { "epoch": 2.37, "learning_rate": 5.486175435040192e-07, "loss": 0.3831, "step": 19939 }, { "epoch": 2.37, "learning_rate": 5.484166411778539e-07, "loss": 0.4709, "step": 19940 }, { "epoch": 2.37, "learning_rate": 5.482157711115243e-07, "loss": 0.4019, "step": 19941 }, { "epoch": 2.37, "learning_rate": 5.48014933308352e-07, "loss": 0.4267, "step": 19942 }, { "epoch": 2.38, "learning_rate": 5.478141277716564e-07, "loss": 0.6066, "step": 19943 }, { "epoch": 2.38, "learning_rate": 5.476133545047583e-07, "loss": 0.5007, "step": 19944 }, { "epoch": 2.38, "learning_rate": 5.474126135109742e-07, "loss": 0.2444, "step": 19945 }, { "epoch": 2.38, "learning_rate": 5.472119047936239e-07, "loss": 0.4175, "step": 19946 }, { "epoch": 2.38, "learning_rate": 5.47011228356025e-07, "loss": 0.3833, "step": 19947 }, { "epoch": 2.38, "learning_rate": 5.468105842014936e-07, "loss": 0.4817, "step": 19948 }, { "epoch": 2.38, "learning_rate": 5.46609972333347e-07, "loss": 0.6432, "step": 19949 }, { "epoch": 2.38, "learning_rate": 5.464093927549016e-07, "loss": 0.6434, "step": 19950 }, { "epoch": 2.38, "learning_rate": 5.462088454694736e-07, "loss": 0.3121, "step": 19951 }, { "epoch": 2.38, "learning_rate": 5.460083304803767e-07, "loss": 0.4149, "step": 19952 }, { "epoch": 2.38, "learning_rate": 5.458078477909265e-07, "loss": 0.4457, "step": 19953 }, { "epoch": 2.38, "learning_rate": 5.45607397404436e-07, "loss": 0.4934, "step": 19954 }, { "epoch": 2.38, "learning_rate": 5.454069793242198e-07, "loss": 0.5805, "step": 19955 }, { "epoch": 2.38, "learning_rate": 5.452065935535897e-07, "loss": 0.5868, "step": 19956 }, { "epoch": 2.38, "learning_rate": 5.450062400958587e-07, "loss": 0.3493, "step": 19957 }, { "epoch": 2.38, "learning_rate": 5.448059189543384e-07, "loss": 0.3673, "step": 19958 }, { "epoch": 2.38, "learning_rate": 5.446056301323408e-07, "loss": 0.4455, "step": 19959 }, { "epoch": 2.38, "learning_rate": 5.444053736331759e-07, "loss": 0.4698, "step": 19960 }, { "epoch": 2.38, "learning_rate": 5.442051494601547e-07, "loss": 0.687, "step": 19961 }, { "epoch": 2.38, "learning_rate": 5.440049576165865e-07, "loss": 0.5444, "step": 19962 }, { "epoch": 2.38, "learning_rate": 5.4380479810578e-07, "loss": 0.3997, "step": 19963 }, { "epoch": 2.38, "learning_rate": 5.436046709310441e-07, "loss": 0.3815, "step": 19964 }, { "epoch": 2.38, "learning_rate": 5.434045760956871e-07, "loss": 0.3768, "step": 19965 }, { "epoch": 2.38, "learning_rate": 5.432045136030176e-07, "loss": 0.5238, "step": 19966 }, { "epoch": 2.38, "learning_rate": 5.430044834563408e-07, "loss": 0.4872, "step": 19967 }, { "epoch": 2.38, "learning_rate": 5.428044856589648e-07, "loss": 0.6156, "step": 19968 }, { "epoch": 2.38, "learning_rate": 5.426045202141944e-07, "loss": 0.5065, "step": 19969 }, { "epoch": 2.38, "learning_rate": 5.424045871253358e-07, "loss": 0.346, "step": 19970 }, { "epoch": 2.38, "learning_rate": 5.422046863956934e-07, "loss": 0.466, "step": 19971 }, { "epoch": 2.38, "learning_rate": 5.420048180285714e-07, "loss": 0.4486, "step": 19972 }, { "epoch": 2.38, "learning_rate": 5.418049820272742e-07, "loss": 0.4758, "step": 19973 }, { "epoch": 2.38, "learning_rate": 5.416051783951057e-07, "loss": 0.6157, "step": 19974 }, { "epoch": 2.38, "learning_rate": 5.414054071353678e-07, "loss": 0.6309, "step": 19975 }, { "epoch": 2.38, "learning_rate": 5.412056682513622e-07, "loss": 0.3721, "step": 19976 }, { "epoch": 2.38, "learning_rate": 5.410059617463922e-07, "loss": 0.34, "step": 19977 }, { "epoch": 2.38, "learning_rate": 5.408062876237572e-07, "loss": 0.4691, "step": 19978 }, { "epoch": 2.38, "learning_rate": 5.406066458867584e-07, "loss": 0.492, "step": 19979 }, { "epoch": 2.38, "learning_rate": 5.404070365386965e-07, "loss": 0.4698, "step": 19980 }, { "epoch": 2.38, "learning_rate": 5.40207459582871e-07, "loss": 0.5904, "step": 19981 }, { "epoch": 2.38, "learning_rate": 5.40007915022581e-07, "loss": 0.1852, "step": 19982 }, { "epoch": 2.38, "learning_rate": 5.398084028611239e-07, "loss": 0.2897, "step": 19983 }, { "epoch": 2.38, "learning_rate": 5.396089231017982e-07, "loss": 0.4632, "step": 19984 }, { "epoch": 2.38, "learning_rate": 5.394094757479024e-07, "loss": 0.4486, "step": 19985 }, { "epoch": 2.38, "learning_rate": 5.392100608027315e-07, "loss": 0.5518, "step": 19986 }, { "epoch": 2.38, "learning_rate": 5.390106782695831e-07, "loss": 0.7047, "step": 19987 }, { "epoch": 2.38, "learning_rate": 5.388113281517532e-07, "loss": 0.6269, "step": 19988 }, { "epoch": 2.38, "learning_rate": 5.386120104525358e-07, "loss": 0.3313, "step": 19989 }, { "epoch": 2.38, "learning_rate": 5.384127251752275e-07, "loss": 0.3634, "step": 19990 }, { "epoch": 2.38, "learning_rate": 5.382134723231206e-07, "loss": 0.5042, "step": 19991 }, { "epoch": 2.38, "learning_rate": 5.380142518995096e-07, "loss": 0.499, "step": 19992 }, { "epoch": 2.38, "learning_rate": 5.378150639076882e-07, "loss": 0.6809, "step": 19993 }, { "epoch": 2.38, "learning_rate": 5.37615908350948e-07, "loss": 0.5016, "step": 19994 }, { "epoch": 2.38, "learning_rate": 5.374167852325812e-07, "loss": 0.3546, "step": 19995 }, { "epoch": 2.38, "learning_rate": 5.372176945558804e-07, "loss": 0.4562, "step": 19996 }, { "epoch": 2.38, "learning_rate": 5.370186363241358e-07, "loss": 0.4631, "step": 19997 }, { "epoch": 2.38, "learning_rate": 5.368196105406373e-07, "loss": 0.461, "step": 19998 }, { "epoch": 2.38, "learning_rate": 5.366206172086755e-07, "loss": 0.5745, "step": 19999 }, { "epoch": 2.38, "learning_rate": 5.3642165633154e-07, "loss": 0.5107, "step": 20000 }, { "epoch": 2.38, "learning_rate": 5.362227279125188e-07, "loss": 0.4043, "step": 20001 }, { "epoch": 2.38, "learning_rate": 5.360238319549005e-07, "loss": 0.415, "step": 20002 }, { "epoch": 2.38, "learning_rate": 5.358249684619737e-07, "loss": 0.5227, "step": 20003 }, { "epoch": 2.38, "learning_rate": 5.356261374370253e-07, "loss": 0.447, "step": 20004 }, { "epoch": 2.38, "learning_rate": 5.354273388833405e-07, "loss": 0.5082, "step": 20005 }, { "epoch": 2.38, "learning_rate": 5.35228572804207e-07, "loss": 0.5815, "step": 20006 }, { "epoch": 2.38, "learning_rate": 5.350298392029102e-07, "loss": 0.234, "step": 20007 }, { "epoch": 2.38, "learning_rate": 5.348311380827354e-07, "loss": 0.3041, "step": 20008 }, { "epoch": 2.38, "learning_rate": 5.346324694469662e-07, "loss": 0.4215, "step": 20009 }, { "epoch": 2.38, "learning_rate": 5.344338332988877e-07, "loss": 0.4494, "step": 20010 }, { "epoch": 2.38, "learning_rate": 5.34235229641783e-07, "loss": 0.5397, "step": 20011 }, { "epoch": 2.38, "learning_rate": 5.340366584789341e-07, "loss": 0.5115, "step": 20012 }, { "epoch": 2.38, "learning_rate": 5.338381198136245e-07, "loss": 0.374, "step": 20013 }, { "epoch": 2.38, "learning_rate": 5.336396136491356e-07, "loss": 0.4326, "step": 20014 }, { "epoch": 2.38, "learning_rate": 5.334411399887496e-07, "loss": 0.5805, "step": 20015 }, { "epoch": 2.38, "learning_rate": 5.332426988357456e-07, "loss": 0.5079, "step": 20016 }, { "epoch": 2.38, "learning_rate": 5.330442901934052e-07, "loss": 0.5836, "step": 20017 }, { "epoch": 2.38, "learning_rate": 5.32845914065008e-07, "loss": 0.5799, "step": 20018 }, { "epoch": 2.38, "learning_rate": 5.326475704538333e-07, "loss": 0.5951, "step": 20019 }, { "epoch": 2.38, "learning_rate": 5.324492593631583e-07, "loss": 0.3665, "step": 20020 }, { "epoch": 2.38, "learning_rate": 5.322509807962626e-07, "loss": 0.4053, "step": 20021 }, { "epoch": 2.38, "learning_rate": 5.320527347564228e-07, "loss": 0.4618, "step": 20022 }, { "epoch": 2.38, "learning_rate": 5.318545212469172e-07, "loss": 0.4678, "step": 20023 }, { "epoch": 2.38, "learning_rate": 5.316563402710212e-07, "loss": 0.6066, "step": 20024 }, { "epoch": 2.38, "learning_rate": 5.314581918320113e-07, "loss": 0.7004, "step": 20025 }, { "epoch": 2.38, "learning_rate": 5.312600759331629e-07, "loss": 0.2942, "step": 20026 }, { "epoch": 2.39, "learning_rate": 5.310619925777499e-07, "loss": 0.3306, "step": 20027 }, { "epoch": 2.39, "learning_rate": 5.308639417690473e-07, "loss": 0.4711, "step": 20028 }, { "epoch": 2.39, "learning_rate": 5.30665923510329e-07, "loss": 0.546, "step": 20029 }, { "epoch": 2.39, "learning_rate": 5.304679378048691e-07, "loss": 0.5863, "step": 20030 }, { "epoch": 2.39, "learning_rate": 5.302699846559384e-07, "loss": 0.5884, "step": 20031 }, { "epoch": 2.39, "learning_rate": 5.300720640668111e-07, "loss": 0.3437, "step": 20032 }, { "epoch": 2.39, "learning_rate": 5.298741760407572e-07, "loss": 0.3104, "step": 20033 }, { "epoch": 2.39, "learning_rate": 5.29676320581049e-07, "loss": 0.42, "step": 20034 }, { "epoch": 2.39, "learning_rate": 5.294784976909559e-07, "loss": 0.3995, "step": 20035 }, { "epoch": 2.39, "learning_rate": 5.292807073737486e-07, "loss": 0.5275, "step": 20036 }, { "epoch": 2.39, "learning_rate": 5.290829496326966e-07, "loss": 0.6105, "step": 20037 }, { "epoch": 2.39, "learning_rate": 5.288852244710693e-07, "loss": 0.3899, "step": 20038 }, { "epoch": 2.39, "learning_rate": 5.286875318921347e-07, "loss": 0.3707, "step": 20039 }, { "epoch": 2.39, "learning_rate": 5.284898718991602e-07, "loss": 0.2904, "step": 20040 }, { "epoch": 2.39, "learning_rate": 5.282922444954139e-07, "loss": 0.4017, "step": 20041 }, { "epoch": 2.39, "learning_rate": 5.280946496841619e-07, "loss": 0.6107, "step": 20042 }, { "epoch": 2.39, "learning_rate": 5.278970874686704e-07, "loss": 0.5961, "step": 20043 }, { "epoch": 2.39, "learning_rate": 5.276995578522057e-07, "loss": 0.4733, "step": 20044 }, { "epoch": 2.39, "learning_rate": 5.275020608380335e-07, "loss": 0.2734, "step": 20045 }, { "epoch": 2.39, "learning_rate": 5.273045964294171e-07, "loss": 0.3419, "step": 20046 }, { "epoch": 2.39, "learning_rate": 5.271071646296217e-07, "loss": 0.4505, "step": 20047 }, { "epoch": 2.39, "learning_rate": 5.2690976544191e-07, "loss": 0.4019, "step": 20048 }, { "epoch": 2.39, "learning_rate": 5.267123988695461e-07, "loss": 0.6619, "step": 20049 }, { "epoch": 2.39, "learning_rate": 5.265150649157907e-07, "loss": 0.6007, "step": 20050 }, { "epoch": 2.39, "learning_rate": 5.263177635839072e-07, "loss": 0.4379, "step": 20051 }, { "epoch": 2.39, "learning_rate": 5.26120494877157e-07, "loss": 0.386, "step": 20052 }, { "epoch": 2.39, "learning_rate": 5.259232587988009e-07, "loss": 0.4571, "step": 20053 }, { "epoch": 2.39, "learning_rate": 5.257260553520991e-07, "loss": 0.5222, "step": 20054 }, { "epoch": 2.39, "learning_rate": 5.255288845403106e-07, "loss": 0.4873, "step": 20055 }, { "epoch": 2.39, "learning_rate": 5.253317463666957e-07, "loss": 0.624, "step": 20056 }, { "epoch": 2.39, "learning_rate": 5.251346408345124e-07, "loss": 0.213, "step": 20057 }, { "epoch": 2.39, "learning_rate": 5.24937567947019e-07, "loss": 0.3128, "step": 20058 }, { "epoch": 2.39, "learning_rate": 5.247405277074736e-07, "loss": 0.4335, "step": 20059 }, { "epoch": 2.39, "learning_rate": 5.245435201191334e-07, "loss": 0.4105, "step": 20060 }, { "epoch": 2.39, "learning_rate": 5.243465451852548e-07, "loss": 0.4656, "step": 20061 }, { "epoch": 2.39, "learning_rate": 5.241496029090925e-07, "loss": 0.5908, "step": 20062 }, { "epoch": 2.39, "learning_rate": 5.239526932939035e-07, "loss": 0.4271, "step": 20063 }, { "epoch": 2.39, "learning_rate": 5.237558163429427e-07, "loss": 0.3515, "step": 20064 }, { "epoch": 2.39, "learning_rate": 5.235589720594633e-07, "loss": 0.4334, "step": 20065 }, { "epoch": 2.39, "learning_rate": 5.233621604467201e-07, "loss": 0.4299, "step": 20066 }, { "epoch": 2.39, "learning_rate": 5.231653815079671e-07, "loss": 0.446, "step": 20067 }, { "epoch": 2.39, "learning_rate": 5.22968635246456e-07, "loss": 0.5987, "step": 20068 }, { "epoch": 2.39, "learning_rate": 5.227719216654387e-07, "loss": 0.4981, "step": 20069 }, { "epoch": 2.39, "learning_rate": 5.225752407681672e-07, "loss": 0.2966, "step": 20070 }, { "epoch": 2.39, "learning_rate": 5.223785925578931e-07, "loss": 0.4554, "step": 20071 }, { "epoch": 2.39, "learning_rate": 5.221819770378675e-07, "loss": 0.4202, "step": 20072 }, { "epoch": 2.39, "learning_rate": 5.21985394211339e-07, "loss": 0.4543, "step": 20073 }, { "epoch": 2.39, "learning_rate": 5.217888440815577e-07, "loss": 0.5816, "step": 20074 }, { "epoch": 2.39, "learning_rate": 5.215923266517737e-07, "loss": 0.6654, "step": 20075 }, { "epoch": 2.39, "learning_rate": 5.213958419252346e-07, "loss": 0.2694, "step": 20076 }, { "epoch": 2.39, "learning_rate": 5.211993899051876e-07, "loss": 0.3926, "step": 20077 }, { "epoch": 2.39, "learning_rate": 5.210029705948805e-07, "loss": 0.4143, "step": 20078 }, { "epoch": 2.39, "learning_rate": 5.208065839975612e-07, "loss": 0.4597, "step": 20079 }, { "epoch": 2.39, "learning_rate": 5.206102301164742e-07, "loss": 0.5463, "step": 20080 }, { "epoch": 2.39, "learning_rate": 5.204139089548666e-07, "loss": 0.5298, "step": 20081 }, { "epoch": 2.39, "learning_rate": 5.202176205159834e-07, "loss": 0.28, "step": 20082 }, { "epoch": 2.39, "learning_rate": 5.200213648030694e-07, "loss": 0.3866, "step": 20083 }, { "epoch": 2.39, "learning_rate": 5.198251418193676e-07, "loss": 0.4142, "step": 20084 }, { "epoch": 2.39, "learning_rate": 5.196289515681224e-07, "loss": 0.4246, "step": 20085 }, { "epoch": 2.39, "learning_rate": 5.194327940525767e-07, "loss": 0.6054, "step": 20086 }, { "epoch": 2.39, "learning_rate": 5.192366692759737e-07, "loss": 0.4549, "step": 20087 }, { "epoch": 2.39, "learning_rate": 5.190405772415544e-07, "loss": 0.4195, "step": 20088 }, { "epoch": 2.39, "learning_rate": 5.188445179525609e-07, "loss": 0.4306, "step": 20089 }, { "epoch": 2.39, "learning_rate": 5.186484914122339e-07, "loss": 0.4223, "step": 20090 }, { "epoch": 2.39, "learning_rate": 5.184524976238128e-07, "loss": 0.5105, "step": 20091 }, { "epoch": 2.39, "learning_rate": 5.182565365905384e-07, "loss": 0.4283, "step": 20092 }, { "epoch": 2.39, "learning_rate": 5.180606083156498e-07, "loss": 0.5193, "step": 20093 }, { "epoch": 2.39, "learning_rate": 5.178647128023859e-07, "loss": 0.4565, "step": 20094 }, { "epoch": 2.39, "learning_rate": 5.176688500539842e-07, "loss": 0.3538, "step": 20095 }, { "epoch": 2.39, "learning_rate": 5.174730200736836e-07, "loss": 0.3579, "step": 20096 }, { "epoch": 2.39, "learning_rate": 5.172772228647194e-07, "loss": 0.4214, "step": 20097 }, { "epoch": 2.39, "learning_rate": 5.170814584303297e-07, "loss": 0.4776, "step": 20098 }, { "epoch": 2.39, "learning_rate": 5.168857267737492e-07, "loss": 0.6206, "step": 20099 }, { "epoch": 2.39, "learning_rate": 5.166900278982143e-07, "loss": 0.5501, "step": 20100 }, { "epoch": 2.39, "learning_rate": 5.164943618069593e-07, "loss": 0.3696, "step": 20101 }, { "epoch": 2.39, "learning_rate": 5.162987285032198e-07, "loss": 0.2811, "step": 20102 }, { "epoch": 2.39, "learning_rate": 5.161031279902281e-07, "loss": 0.3933, "step": 20103 }, { "epoch": 2.39, "learning_rate": 5.159075602712188e-07, "loss": 0.4434, "step": 20104 }, { "epoch": 2.39, "learning_rate": 5.157120253494238e-07, "loss": 0.5742, "step": 20105 }, { "epoch": 2.39, "learning_rate": 5.155165232280751e-07, "loss": 0.5995, "step": 20106 }, { "epoch": 2.39, "learning_rate": 5.153210539104048e-07, "loss": 0.3626, "step": 20107 }, { "epoch": 2.39, "learning_rate": 5.151256173996438e-07, "loss": 0.3974, "step": 20108 }, { "epoch": 2.39, "learning_rate": 5.149302136990239e-07, "loss": 0.4501, "step": 20109 }, { "epoch": 2.39, "learning_rate": 5.147348428117732e-07, "loss": 0.5027, "step": 20110 }, { "epoch": 2.4, "learning_rate": 5.145395047411228e-07, "loss": 0.5461, "step": 20111 }, { "epoch": 2.4, "learning_rate": 5.143441994903004e-07, "loss": 0.5583, "step": 20112 }, { "epoch": 2.4, "learning_rate": 5.141489270625355e-07, "loss": 0.4351, "step": 20113 }, { "epoch": 2.4, "learning_rate": 5.139536874610551e-07, "loss": 0.3087, "step": 20114 }, { "epoch": 2.4, "learning_rate": 5.137584806890866e-07, "loss": 0.4502, "step": 20115 }, { "epoch": 2.4, "learning_rate": 5.135633067498571e-07, "loss": 0.4182, "step": 20116 }, { "epoch": 2.4, "learning_rate": 5.133681656465936e-07, "loss": 0.5141, "step": 20117 }, { "epoch": 2.4, "learning_rate": 5.131730573825211e-07, "loss": 0.5863, "step": 20118 }, { "epoch": 2.4, "learning_rate": 5.129779819608638e-07, "loss": 0.5106, "step": 20119 }, { "epoch": 2.4, "learning_rate": 5.12782939384848e-07, "loss": 0.2809, "step": 20120 }, { "epoch": 2.4, "learning_rate": 5.12587929657696e-07, "loss": 0.3971, "step": 20121 }, { "epoch": 2.4, "learning_rate": 5.123929527826327e-07, "loss": 0.396, "step": 20122 }, { "epoch": 2.4, "learning_rate": 5.121980087628802e-07, "loss": 0.4104, "step": 20123 }, { "epoch": 2.4, "learning_rate": 5.120030976016624e-07, "loss": 0.5358, "step": 20124 }, { "epoch": 2.4, "learning_rate": 5.118082193022004e-07, "loss": 0.5627, "step": 20125 }, { "epoch": 2.4, "learning_rate": 5.116133738677142e-07, "loss": 0.2692, "step": 20126 }, { "epoch": 2.4, "learning_rate": 5.114185613014261e-07, "loss": 0.3038, "step": 20127 }, { "epoch": 2.4, "learning_rate": 5.112237816065565e-07, "loss": 0.4976, "step": 20128 }, { "epoch": 2.4, "learning_rate": 5.110290347863242e-07, "loss": 0.3778, "step": 20129 }, { "epoch": 2.4, "learning_rate": 5.108343208439487e-07, "loss": 0.5636, "step": 20130 }, { "epoch": 2.4, "learning_rate": 5.106396397826488e-07, "loss": 0.6136, "step": 20131 }, { "epoch": 2.4, "learning_rate": 5.104449916056433e-07, "loss": 0.3582, "step": 20132 }, { "epoch": 2.4, "learning_rate": 5.10250376316149e-07, "loss": 0.3606, "step": 20133 }, { "epoch": 2.4, "learning_rate": 5.100557939173823e-07, "loss": 0.3446, "step": 20134 }, { "epoch": 2.4, "learning_rate": 5.098612444125603e-07, "loss": 0.3704, "step": 20135 }, { "epoch": 2.4, "learning_rate": 5.096667278048995e-07, "loss": 0.4968, "step": 20136 }, { "epoch": 2.4, "learning_rate": 5.09472244097614e-07, "loss": 0.65, "step": 20137 }, { "epoch": 2.4, "learning_rate": 5.092777932939194e-07, "loss": 0.4638, "step": 20138 }, { "epoch": 2.4, "learning_rate": 5.090833753970307e-07, "loss": 0.3314, "step": 20139 }, { "epoch": 2.4, "learning_rate": 5.088889904101604e-07, "loss": 0.3955, "step": 20140 }, { "epoch": 2.4, "learning_rate": 5.086946383365219e-07, "loss": 0.3893, "step": 20141 }, { "epoch": 2.4, "learning_rate": 5.085003191793281e-07, "loss": 0.449, "step": 20142 }, { "epoch": 2.4, "learning_rate": 5.083060329417916e-07, "loss": 0.5804, "step": 20143 }, { "epoch": 2.4, "learning_rate": 5.081117796271229e-07, "loss": 0.5112, "step": 20144 }, { "epoch": 2.4, "learning_rate": 5.079175592385338e-07, "loss": 0.3684, "step": 20145 }, { "epoch": 2.4, "learning_rate": 5.077233717792352e-07, "loss": 0.3228, "step": 20146 }, { "epoch": 2.4, "learning_rate": 5.075292172524363e-07, "loss": 0.5841, "step": 20147 }, { "epoch": 2.4, "learning_rate": 5.07335095661346e-07, "loss": 0.5558, "step": 20148 }, { "epoch": 2.4, "learning_rate": 5.071410070091737e-07, "loss": 0.5945, "step": 20149 }, { "epoch": 2.4, "learning_rate": 5.069469512991279e-07, "loss": 0.6228, "step": 20150 }, { "epoch": 2.4, "learning_rate": 5.067529285344167e-07, "loss": 0.3463, "step": 20151 }, { "epoch": 2.4, "learning_rate": 5.065589387182462e-07, "loss": 0.3944, "step": 20152 }, { "epoch": 2.4, "learning_rate": 5.063649818538246e-07, "loss": 0.421, "step": 20153 }, { "epoch": 2.4, "learning_rate": 5.061710579443571e-07, "loss": 0.4322, "step": 20154 }, { "epoch": 2.4, "learning_rate": 5.059771669930486e-07, "loss": 0.558, "step": 20155 }, { "epoch": 2.4, "learning_rate": 5.057833090031047e-07, "loss": 0.5896, "step": 20156 }, { "epoch": 2.4, "learning_rate": 5.055894839777301e-07, "loss": 0.2962, "step": 20157 }, { "epoch": 2.4, "learning_rate": 5.053956919201295e-07, "loss": 0.3405, "step": 20158 }, { "epoch": 2.4, "learning_rate": 5.052019328335045e-07, "loss": 0.4314, "step": 20159 }, { "epoch": 2.4, "learning_rate": 5.050082067210591e-07, "loss": 0.4978, "step": 20160 }, { "epoch": 2.4, "learning_rate": 5.048145135859961e-07, "loss": 0.5302, "step": 20161 }, { "epoch": 2.4, "learning_rate": 5.046208534315166e-07, "loss": 0.6568, "step": 20162 }, { "epoch": 2.4, "learning_rate": 5.04427226260821e-07, "loss": 0.479, "step": 20163 }, { "epoch": 2.4, "learning_rate": 5.04233632077111e-07, "loss": 0.3917, "step": 20164 }, { "epoch": 2.4, "learning_rate": 5.040400708835866e-07, "loss": 0.5123, "step": 20165 }, { "epoch": 2.4, "learning_rate": 5.038465426834477e-07, "loss": 0.4679, "step": 20166 }, { "epoch": 2.4, "learning_rate": 5.036530474798925e-07, "loss": 0.3893, "step": 20167 }, { "epoch": 2.4, "learning_rate": 5.034595852761204e-07, "loss": 0.6504, "step": 20168 }, { "epoch": 2.4, "learning_rate": 5.03266156075329e-07, "loss": 0.411, "step": 20169 }, { "epoch": 2.4, "learning_rate": 5.030727598807151e-07, "loss": 0.3094, "step": 20170 }, { "epoch": 2.4, "learning_rate": 5.028793966954757e-07, "loss": 0.4246, "step": 20171 }, { "epoch": 2.4, "learning_rate": 5.026860665228075e-07, "loss": 0.4464, "step": 20172 }, { "epoch": 2.4, "learning_rate": 5.024927693659071e-07, "loss": 0.5551, "step": 20173 }, { "epoch": 2.4, "learning_rate": 5.022995052279681e-07, "loss": 0.5631, "step": 20174 }, { "epoch": 2.4, "learning_rate": 5.021062741121863e-07, "loss": 0.577, "step": 20175 }, { "epoch": 2.4, "learning_rate": 5.01913076021755e-07, "loss": 0.3316, "step": 20176 }, { "epoch": 2.4, "learning_rate": 5.017199109598689e-07, "loss": 0.4586, "step": 20177 }, { "epoch": 2.4, "learning_rate": 5.015267789297196e-07, "loss": 0.3672, "step": 20178 }, { "epoch": 2.4, "learning_rate": 5.013336799345001e-07, "loss": 0.5561, "step": 20179 }, { "epoch": 2.4, "learning_rate": 5.011406139774028e-07, "loss": 0.5478, "step": 20180 }, { "epoch": 2.4, "learning_rate": 5.009475810616196e-07, "loss": 0.5639, "step": 20181 }, { "epoch": 2.4, "learning_rate": 5.007545811903405e-07, "loss": 0.3047, "step": 20182 }, { "epoch": 2.4, "learning_rate": 5.005616143667552e-07, "loss": 0.4203, "step": 20183 }, { "epoch": 2.4, "learning_rate": 5.00368680594055e-07, "loss": 0.4647, "step": 20184 }, { "epoch": 2.4, "learning_rate": 5.001757798754275e-07, "loss": 0.4256, "step": 20185 }, { "epoch": 2.4, "learning_rate": 4.999829122140623e-07, "loss": 0.5712, "step": 20186 }, { "epoch": 2.4, "learning_rate": 4.997900776131473e-07, "loss": 0.5038, "step": 20187 }, { "epoch": 2.4, "learning_rate": 4.995972760758708e-07, "loss": 0.5599, "step": 20188 }, { "epoch": 2.4, "learning_rate": 4.994045076054186e-07, "loss": 0.3903, "step": 20189 }, { "epoch": 2.4, "learning_rate": 4.992117722049783e-07, "loss": 0.4415, "step": 20190 }, { "epoch": 2.4, "learning_rate": 4.990190698777348e-07, "loss": 0.4488, "step": 20191 }, { "epoch": 2.4, "learning_rate": 4.988264006268745e-07, "loss": 0.6688, "step": 20192 }, { "epoch": 2.4, "learning_rate": 4.986337644555811e-07, "loss": 0.5856, "step": 20193 }, { "epoch": 2.4, "learning_rate": 4.984411613670393e-07, "loss": 0.5517, "step": 20194 }, { "epoch": 2.41, "learning_rate": 4.982485913644333e-07, "loss": 0.2579, "step": 20195 }, { "epoch": 2.41, "learning_rate": 4.980560544509467e-07, "loss": 0.4374, "step": 20196 }, { "epoch": 2.41, "learning_rate": 4.978635506297613e-07, "loss": 0.4126, "step": 20197 }, { "epoch": 2.41, "learning_rate": 4.976710799040587e-07, "loss": 0.4687, "step": 20198 }, { "epoch": 2.41, "learning_rate": 4.974786422770214e-07, "loss": 0.6326, "step": 20199 }, { "epoch": 2.41, "learning_rate": 4.972862377518309e-07, "loss": 0.4928, "step": 20200 }, { "epoch": 2.41, "learning_rate": 4.97093866331666e-07, "loss": 0.4318, "step": 20201 }, { "epoch": 2.41, "learning_rate": 4.969015280197076e-07, "loss": 0.4426, "step": 20202 }, { "epoch": 2.41, "learning_rate": 4.967092228191359e-07, "loss": 0.4215, "step": 20203 }, { "epoch": 2.41, "learning_rate": 4.965169507331285e-07, "loss": 0.4455, "step": 20204 }, { "epoch": 2.41, "learning_rate": 4.963247117648634e-07, "loss": 0.5685, "step": 20205 }, { "epoch": 2.41, "learning_rate": 4.961325059175193e-07, "loss": 0.6076, "step": 20206 }, { "epoch": 2.41, "learning_rate": 4.959403331942733e-07, "loss": 0.3351, "step": 20207 }, { "epoch": 2.41, "learning_rate": 4.957481935983014e-07, "loss": 0.3523, "step": 20208 }, { "epoch": 2.41, "learning_rate": 4.955560871327797e-07, "loss": 0.4095, "step": 20209 }, { "epoch": 2.41, "learning_rate": 4.953640138008847e-07, "loss": 0.4429, "step": 20210 }, { "epoch": 2.41, "learning_rate": 4.951719736057911e-07, "loss": 0.5758, "step": 20211 }, { "epoch": 2.41, "learning_rate": 4.949799665506722e-07, "loss": 0.642, "step": 20212 }, { "epoch": 2.41, "learning_rate": 4.947879926387025e-07, "loss": 0.3122, "step": 20213 }, { "epoch": 2.41, "learning_rate": 4.945960518730558e-07, "loss": 0.3237, "step": 20214 }, { "epoch": 2.41, "learning_rate": 4.944041442569051e-07, "loss": 0.4401, "step": 20215 }, { "epoch": 2.41, "learning_rate": 4.942122697934215e-07, "loss": 0.4192, "step": 20216 }, { "epoch": 2.41, "learning_rate": 4.940204284857775e-07, "loss": 0.5846, "step": 20217 }, { "epoch": 2.41, "learning_rate": 4.938286203371448e-07, "loss": 0.5162, "step": 20218 }, { "epoch": 2.41, "learning_rate": 4.936368453506934e-07, "loss": 0.6747, "step": 20219 }, { "epoch": 2.41, "learning_rate": 4.934451035295926e-07, "loss": 0.2461, "step": 20220 }, { "epoch": 2.41, "learning_rate": 4.932533948770124e-07, "loss": 0.4073, "step": 20221 }, { "epoch": 2.41, "learning_rate": 4.930617193961229e-07, "loss": 0.436, "step": 20222 }, { "epoch": 2.41, "learning_rate": 4.928700770900907e-07, "loss": 0.5182, "step": 20223 }, { "epoch": 2.41, "learning_rate": 4.926784679620847e-07, "loss": 0.6567, "step": 20224 }, { "epoch": 2.41, "learning_rate": 4.924868920152726e-07, "loss": 0.5206, "step": 20225 }, { "epoch": 2.41, "learning_rate": 4.922953492528207e-07, "loss": 0.3048, "step": 20226 }, { "epoch": 2.41, "learning_rate": 4.921038396778943e-07, "loss": 0.3563, "step": 20227 }, { "epoch": 2.41, "learning_rate": 4.919123632936601e-07, "loss": 0.4373, "step": 20228 }, { "epoch": 2.41, "learning_rate": 4.91720920103283e-07, "loss": 0.3919, "step": 20229 }, { "epoch": 2.41, "learning_rate": 4.915295101099282e-07, "loss": 0.6267, "step": 20230 }, { "epoch": 2.41, "learning_rate": 4.913381333167582e-07, "loss": 0.5393, "step": 20231 }, { "epoch": 2.41, "learning_rate": 4.911467897269384e-07, "loss": 0.2897, "step": 20232 }, { "epoch": 2.41, "learning_rate": 4.909554793436308e-07, "loss": 0.3507, "step": 20233 }, { "epoch": 2.41, "learning_rate": 4.907642021699968e-07, "loss": 0.4178, "step": 20234 }, { "epoch": 2.41, "learning_rate": 4.905729582091992e-07, "loss": 0.4185, "step": 20235 }, { "epoch": 2.41, "learning_rate": 4.903817474643993e-07, "loss": 0.5655, "step": 20236 }, { "epoch": 2.41, "learning_rate": 4.901905699387582e-07, "loss": 0.5527, "step": 20237 }, { "epoch": 2.41, "learning_rate": 4.899994256354351e-07, "loss": 0.3957, "step": 20238 }, { "epoch": 2.41, "learning_rate": 4.898083145575907e-07, "loss": 0.3021, "step": 20239 }, { "epoch": 2.41, "learning_rate": 4.896172367083832e-07, "loss": 0.3675, "step": 20240 }, { "epoch": 2.41, "learning_rate": 4.894261920909718e-07, "loss": 0.3714, "step": 20241 }, { "epoch": 2.41, "learning_rate": 4.892351807085136e-07, "loss": 0.6455, "step": 20242 }, { "epoch": 2.41, "learning_rate": 4.890442025641664e-07, "loss": 0.6235, "step": 20243 }, { "epoch": 2.41, "learning_rate": 4.888532576610875e-07, "loss": 0.5562, "step": 20244 }, { "epoch": 2.41, "learning_rate": 4.886623460024337e-07, "loss": 0.396, "step": 20245 }, { "epoch": 2.41, "learning_rate": 4.884714675913593e-07, "loss": 0.4126, "step": 20246 }, { "epoch": 2.41, "learning_rate": 4.882806224310207e-07, "loss": 0.3986, "step": 20247 }, { "epoch": 2.41, "learning_rate": 4.880898105245726e-07, "loss": 0.5685, "step": 20248 }, { "epoch": 2.41, "learning_rate": 4.878990318751681e-07, "loss": 0.6796, "step": 20249 }, { "epoch": 2.41, "learning_rate": 4.877082864859611e-07, "loss": 0.5676, "step": 20250 }, { "epoch": 2.41, "learning_rate": 4.875175743601055e-07, "loss": 0.3387, "step": 20251 }, { "epoch": 2.41, "learning_rate": 4.873268955007535e-07, "loss": 0.3984, "step": 20252 }, { "epoch": 2.41, "learning_rate": 4.87136249911056e-07, "loss": 0.4755, "step": 20253 }, { "epoch": 2.41, "learning_rate": 4.869456375941661e-07, "loss": 0.5002, "step": 20254 }, { "epoch": 2.41, "learning_rate": 4.867550585532333e-07, "loss": 0.5589, "step": 20255 }, { "epoch": 2.41, "learning_rate": 4.865645127914084e-07, "loss": 0.6271, "step": 20256 }, { "epoch": 2.41, "learning_rate": 4.863740003118408e-07, "loss": 0.3081, "step": 20257 }, { "epoch": 2.41, "learning_rate": 4.861835211176799e-07, "loss": 0.4064, "step": 20258 }, { "epoch": 2.41, "learning_rate": 4.859930752120742e-07, "loss": 0.3373, "step": 20259 }, { "epoch": 2.41, "learning_rate": 4.858026625981726e-07, "loss": 0.4327, "step": 20260 }, { "epoch": 2.41, "learning_rate": 4.856122832791222e-07, "loss": 0.4616, "step": 20261 }, { "epoch": 2.41, "learning_rate": 4.85421937258069e-07, "loss": 0.658, "step": 20262 }, { "epoch": 2.41, "learning_rate": 4.852316245381608e-07, "loss": 0.31, "step": 20263 }, { "epoch": 2.41, "learning_rate": 4.850413451225424e-07, "loss": 0.3921, "step": 20264 }, { "epoch": 2.41, "learning_rate": 4.848510990143596e-07, "loss": 0.4368, "step": 20265 }, { "epoch": 2.41, "learning_rate": 4.846608862167571e-07, "loss": 0.4012, "step": 20266 }, { "epoch": 2.41, "learning_rate": 4.844707067328799e-07, "loss": 0.4618, "step": 20267 }, { "epoch": 2.41, "learning_rate": 4.842805605658712e-07, "loss": 0.5736, "step": 20268 }, { "epoch": 2.41, "learning_rate": 4.840904477188732e-07, "loss": 0.5589, "step": 20269 }, { "epoch": 2.41, "learning_rate": 4.839003681950294e-07, "loss": 0.3183, "step": 20270 }, { "epoch": 2.41, "learning_rate": 4.837103219974821e-07, "loss": 0.4349, "step": 20271 }, { "epoch": 2.41, "learning_rate": 4.835203091293719e-07, "loss": 0.4621, "step": 20272 }, { "epoch": 2.41, "learning_rate": 4.833303295938399e-07, "loss": 0.479, "step": 20273 }, { "epoch": 2.41, "learning_rate": 4.831403833940271e-07, "loss": 0.627, "step": 20274 }, { "epoch": 2.41, "learning_rate": 4.829504705330734e-07, "loss": 0.5367, "step": 20275 }, { "epoch": 2.41, "learning_rate": 4.827605910141179e-07, "loss": 0.2883, "step": 20276 }, { "epoch": 2.41, "learning_rate": 4.825707448402983e-07, "loss": 0.4265, "step": 20277 }, { "epoch": 2.41, "learning_rate": 4.823809320147535e-07, "loss": 0.4337, "step": 20278 }, { "epoch": 2.42, "learning_rate": 4.82191152540622e-07, "loss": 0.5177, "step": 20279 }, { "epoch": 2.42, "learning_rate": 4.820014064210393e-07, "loss": 0.5392, "step": 20280 }, { "epoch": 2.42, "learning_rate": 4.818116936591427e-07, "loss": 0.6031, "step": 20281 }, { "epoch": 2.42, "learning_rate": 4.816220142580688e-07, "loss": 0.3493, "step": 20282 }, { "epoch": 2.42, "learning_rate": 4.814323682209523e-07, "loss": 0.3977, "step": 20283 }, { "epoch": 2.42, "learning_rate": 4.812427555509275e-07, "loss": 0.4498, "step": 20284 }, { "epoch": 2.42, "learning_rate": 4.810531762511295e-07, "loss": 0.4613, "step": 20285 }, { "epoch": 2.42, "learning_rate": 4.808636303246922e-07, "loss": 0.5654, "step": 20286 }, { "epoch": 2.42, "learning_rate": 4.806741177747481e-07, "loss": 0.6314, "step": 20287 }, { "epoch": 2.42, "learning_rate": 4.804846386044304e-07, "loss": 0.31, "step": 20288 }, { "epoch": 2.42, "learning_rate": 4.802951928168714e-07, "loss": 0.394, "step": 20289 }, { "epoch": 2.42, "learning_rate": 4.801057804152026e-07, "loss": 0.3734, "step": 20290 }, { "epoch": 2.42, "learning_rate": 4.799164014025539e-07, "loss": 0.3856, "step": 20291 }, { "epoch": 2.42, "learning_rate": 4.797270557820566e-07, "loss": 0.5681, "step": 20292 }, { "epoch": 2.42, "learning_rate": 4.795377435568408e-07, "loss": 0.6189, "step": 20293 }, { "epoch": 2.42, "learning_rate": 4.793484647300362e-07, "loss": 0.4478, "step": 20294 }, { "epoch": 2.42, "learning_rate": 4.791592193047704e-07, "loss": 0.2589, "step": 20295 }, { "epoch": 2.42, "learning_rate": 4.789700072841727e-07, "loss": 0.4178, "step": 20296 }, { "epoch": 2.42, "learning_rate": 4.787808286713705e-07, "loss": 0.5929, "step": 20297 }, { "epoch": 2.42, "learning_rate": 4.785916834694901e-07, "loss": 0.4761, "step": 20298 }, { "epoch": 2.42, "learning_rate": 4.784025716816587e-07, "loss": 0.5853, "step": 20299 }, { "epoch": 2.42, "learning_rate": 4.782134933110025e-07, "loss": 0.4563, "step": 20300 }, { "epoch": 2.42, "learning_rate": 4.780244483606474e-07, "loss": 0.2574, "step": 20301 }, { "epoch": 2.42, "learning_rate": 4.778354368337176e-07, "loss": 0.3988, "step": 20302 }, { "epoch": 2.42, "learning_rate": 4.776464587333374e-07, "loss": 0.4866, "step": 20303 }, { "epoch": 2.42, "learning_rate": 4.774575140626317e-07, "loss": 0.6026, "step": 20304 }, { "epoch": 2.42, "learning_rate": 4.772686028247228e-07, "loss": 0.5712, "step": 20305 }, { "epoch": 2.42, "learning_rate": 4.770797250227333e-07, "loss": 0.4934, "step": 20306 }, { "epoch": 2.42, "learning_rate": 4.768908806597852e-07, "loss": 0.4327, "step": 20307 }, { "epoch": 2.42, "learning_rate": 4.7670206973900095e-07, "loss": 0.3725, "step": 20308 }, { "epoch": 2.42, "learning_rate": 4.7651329226350184e-07, "loss": 0.3669, "step": 20309 }, { "epoch": 2.42, "learning_rate": 4.7632454823640705e-07, "loss": 0.4247, "step": 20310 }, { "epoch": 2.42, "learning_rate": 4.7613583766083797e-07, "loss": 0.4619, "step": 20311 }, { "epoch": 2.42, "learning_rate": 4.759471605399135e-07, "loss": 0.6627, "step": 20312 }, { "epoch": 2.42, "learning_rate": 4.7575851687675145e-07, "loss": 0.401, "step": 20313 }, { "epoch": 2.42, "learning_rate": 4.7556990667447094e-07, "loss": 0.2719, "step": 20314 }, { "epoch": 2.42, "learning_rate": 4.753813299361898e-07, "loss": 0.4158, "step": 20315 }, { "epoch": 2.42, "learning_rate": 4.751927866650258e-07, "loss": 0.4601, "step": 20316 }, { "epoch": 2.42, "learning_rate": 4.7500427686409447e-07, "loss": 0.5687, "step": 20317 }, { "epoch": 2.42, "learning_rate": 4.7481580053651276e-07, "loss": 0.6587, "step": 20318 }, { "epoch": 2.42, "learning_rate": 4.746273576853952e-07, "loss": 0.5934, "step": 20319 }, { "epoch": 2.42, "learning_rate": 4.744389483138581e-07, "loss": 0.302, "step": 20320 }, { "epoch": 2.42, "learning_rate": 4.742505724250143e-07, "loss": 0.4399, "step": 20321 }, { "epoch": 2.42, "learning_rate": 4.7406223002197853e-07, "loss": 0.5083, "step": 20322 }, { "epoch": 2.42, "learning_rate": 4.738739211078644e-07, "loss": 0.5823, "step": 20323 }, { "epoch": 2.42, "learning_rate": 4.736856456857847e-07, "loss": 0.6423, "step": 20324 }, { "epoch": 2.42, "learning_rate": 4.734974037588513e-07, "loss": 0.5417, "step": 20325 }, { "epoch": 2.42, "learning_rate": 4.7330919533017523e-07, "loss": 0.4507, "step": 20326 }, { "epoch": 2.42, "learning_rate": 4.7312102040286913e-07, "loss": 0.4529, "step": 20327 }, { "epoch": 2.42, "learning_rate": 4.7293287898004163e-07, "loss": 0.485, "step": 20328 }, { "epoch": 2.42, "learning_rate": 4.7274477106480393e-07, "loss": 0.5209, "step": 20329 }, { "epoch": 2.42, "learning_rate": 4.7255669666026517e-07, "loss": 0.4827, "step": 20330 }, { "epoch": 2.42, "learning_rate": 4.723686557695351e-07, "loss": 0.5399, "step": 20331 }, { "epoch": 2.42, "learning_rate": 4.7218064839572056e-07, "loss": 0.3685, "step": 20332 }, { "epoch": 2.42, "learning_rate": 4.7199267454193067e-07, "loss": 0.442, "step": 20333 }, { "epoch": 2.42, "learning_rate": 4.718047342112714e-07, "loss": 0.4007, "step": 20334 }, { "epoch": 2.42, "learning_rate": 4.7161682740685085e-07, "loss": 0.4898, "step": 20335 }, { "epoch": 2.42, "learning_rate": 4.714289541317735e-07, "loss": 0.5542, "step": 20336 }, { "epoch": 2.42, "learning_rate": 4.712411143891457e-07, "loss": 0.5851, "step": 20337 }, { "epoch": 2.42, "learning_rate": 4.7105330818207255e-07, "loss": 0.3033, "step": 20338 }, { "epoch": 2.42, "learning_rate": 4.7086553551365934e-07, "loss": 0.3251, "step": 20339 }, { "epoch": 2.42, "learning_rate": 4.706777963870088e-07, "loss": 0.3679, "step": 20340 }, { "epoch": 2.42, "learning_rate": 4.70490090805224e-07, "loss": 0.5381, "step": 20341 }, { "epoch": 2.42, "learning_rate": 4.7030241877140837e-07, "loss": 0.5403, "step": 20342 }, { "epoch": 2.42, "learning_rate": 4.7011478028866464e-07, "loss": 0.5536, "step": 20343 }, { "epoch": 2.42, "learning_rate": 4.6992717536009313e-07, "loss": 0.5055, "step": 20344 }, { "epoch": 2.42, "learning_rate": 4.697396039887961e-07, "loss": 0.3198, "step": 20345 }, { "epoch": 2.42, "learning_rate": 4.6955206617787406e-07, "loss": 0.3936, "step": 20346 }, { "epoch": 2.42, "learning_rate": 4.693645619304268e-07, "loss": 0.4674, "step": 20347 }, { "epoch": 2.42, "learning_rate": 4.691770912495533e-07, "loss": 0.3918, "step": 20348 }, { "epoch": 2.42, "learning_rate": 4.6898965413835265e-07, "loss": 0.616, "step": 20349 }, { "epoch": 2.42, "learning_rate": 4.688022505999243e-07, "loss": 0.5441, "step": 20350 }, { "epoch": 2.42, "learning_rate": 4.686148806373644e-07, "loss": 0.3081, "step": 20351 }, { "epoch": 2.42, "learning_rate": 4.6842754425377086e-07, "loss": 0.4694, "step": 20352 }, { "epoch": 2.42, "learning_rate": 4.6824024145224136e-07, "loss": 0.4148, "step": 20353 }, { "epoch": 2.42, "learning_rate": 4.6805297223587093e-07, "loss": 0.4874, "step": 20354 }, { "epoch": 2.42, "learning_rate": 4.6786573660775494e-07, "loss": 0.5395, "step": 20355 }, { "epoch": 2.42, "learning_rate": 4.676785345709886e-07, "loss": 0.7063, "step": 20356 }, { "epoch": 2.42, "learning_rate": 4.6749136612866673e-07, "loss": 0.2784, "step": 20357 }, { "epoch": 2.42, "learning_rate": 4.673042312838838e-07, "loss": 0.3382, "step": 20358 }, { "epoch": 2.42, "learning_rate": 4.6711713003973176e-07, "loss": 0.3994, "step": 20359 }, { "epoch": 2.42, "learning_rate": 4.669300623993042e-07, "loss": 0.529, "step": 20360 }, { "epoch": 2.42, "learning_rate": 4.6674302836569375e-07, "loss": 0.5091, "step": 20361 }, { "epoch": 2.42, "learning_rate": 4.6655602794199175e-07, "loss": 0.6252, "step": 20362 }, { "epoch": 2.43, "learning_rate": 4.663690611312888e-07, "loss": 0.3966, "step": 20363 }, { "epoch": 2.43, "learning_rate": 4.661821279366757e-07, "loss": 0.4095, "step": 20364 }, { "epoch": 2.43, "learning_rate": 4.659952283612434e-07, "loss": 0.4979, "step": 20365 }, { "epoch": 2.43, "learning_rate": 4.658083624080798e-07, "loss": 0.418, "step": 20366 }, { "epoch": 2.43, "learning_rate": 4.6562153008027495e-07, "loss": 0.5522, "step": 20367 }, { "epoch": 2.43, "learning_rate": 4.654347313809174e-07, "loss": 0.6505, "step": 20368 }, { "epoch": 2.43, "learning_rate": 4.6524796631309434e-07, "loss": 0.4295, "step": 20369 }, { "epoch": 2.43, "learning_rate": 4.6506123487989273e-07, "loss": 0.3988, "step": 20370 }, { "epoch": 2.43, "learning_rate": 4.648745370843993e-07, "loss": 0.3924, "step": 20371 }, { "epoch": 2.43, "learning_rate": 4.646878729297008e-07, "loss": 0.4557, "step": 20372 }, { "epoch": 2.43, "learning_rate": 4.6450124241888327e-07, "loss": 0.4679, "step": 20373 }, { "epoch": 2.43, "learning_rate": 4.6431464555503017e-07, "loss": 0.5854, "step": 20374 }, { "epoch": 2.43, "learning_rate": 4.6412808234122753e-07, "loss": 0.5925, "step": 20375 }, { "epoch": 2.43, "learning_rate": 4.639415527805585e-07, "loss": 0.343, "step": 20376 }, { "epoch": 2.43, "learning_rate": 4.637550568761057e-07, "loss": 0.3911, "step": 20377 }, { "epoch": 2.43, "learning_rate": 4.635685946309526e-07, "loss": 0.4946, "step": 20378 }, { "epoch": 2.43, "learning_rate": 4.6338216604818186e-07, "loss": 0.4619, "step": 20379 }, { "epoch": 2.43, "learning_rate": 4.6319577113087514e-07, "loss": 0.6705, "step": 20380 }, { "epoch": 2.43, "learning_rate": 4.6300940988211254e-07, "loss": 0.6102, "step": 20381 }, { "epoch": 2.43, "learning_rate": 4.628230823049762e-07, "loss": 0.3419, "step": 20382 }, { "epoch": 2.43, "learning_rate": 4.6263678840254453e-07, "loss": 0.4, "step": 20383 }, { "epoch": 2.43, "learning_rate": 4.624505281778982e-07, "loss": 0.4075, "step": 20384 }, { "epoch": 2.43, "learning_rate": 4.6226430163411496e-07, "loss": 0.3806, "step": 20385 }, { "epoch": 2.43, "learning_rate": 4.62078108774274e-07, "loss": 0.5997, "step": 20386 }, { "epoch": 2.43, "learning_rate": 4.618919496014526e-07, "loss": 0.6055, "step": 20387 }, { "epoch": 2.43, "learning_rate": 4.6170582411872915e-07, "loss": 0.3817, "step": 20388 }, { "epoch": 2.43, "learning_rate": 4.6151973232917897e-07, "loss": 0.2968, "step": 20389 }, { "epoch": 2.43, "learning_rate": 4.61333674235879e-07, "loss": 0.3148, "step": 20390 }, { "epoch": 2.43, "learning_rate": 4.6114764984190455e-07, "loss": 0.4202, "step": 20391 }, { "epoch": 2.43, "learning_rate": 4.6096165915032993e-07, "loss": 0.6131, "step": 20392 }, { "epoch": 2.43, "learning_rate": 4.6077570216423005e-07, "loss": 0.4739, "step": 20393 }, { "epoch": 2.43, "learning_rate": 4.605897788866789e-07, "loss": 0.6129, "step": 20394 }, { "epoch": 2.43, "learning_rate": 4.6040388932075046e-07, "loss": 0.3715, "step": 20395 }, { "epoch": 2.43, "learning_rate": 4.602180334695164e-07, "loss": 0.4241, "step": 20396 }, { "epoch": 2.43, "learning_rate": 4.6003221133605005e-07, "loss": 0.4761, "step": 20397 }, { "epoch": 2.43, "learning_rate": 4.5984642292342174e-07, "loss": 0.5959, "step": 20398 }, { "epoch": 2.43, "learning_rate": 4.596606682347038e-07, "loss": 0.5247, "step": 20399 }, { "epoch": 2.43, "learning_rate": 4.5947494727296565e-07, "loss": 0.5494, "step": 20400 }, { "epoch": 2.43, "learning_rate": 4.5928926004127766e-07, "loss": 0.2694, "step": 20401 }, { "epoch": 2.43, "learning_rate": 4.5910360654270955e-07, "loss": 0.3244, "step": 20402 }, { "epoch": 2.43, "learning_rate": 4.589179867803306e-07, "loss": 0.4324, "step": 20403 }, { "epoch": 2.43, "learning_rate": 4.5873240075720886e-07, "loss": 0.4397, "step": 20404 }, { "epoch": 2.43, "learning_rate": 4.585468484764111e-07, "loss": 0.6091, "step": 20405 }, { "epoch": 2.43, "learning_rate": 4.583613299410056e-07, "loss": 0.6072, "step": 20406 }, { "epoch": 2.43, "learning_rate": 4.5817584515405825e-07, "loss": 0.4668, "step": 20407 }, { "epoch": 2.43, "learning_rate": 4.579903941186353e-07, "loss": 0.3584, "step": 20408 }, { "epoch": 2.43, "learning_rate": 4.578049768378026e-07, "loss": 0.3852, "step": 20409 }, { "epoch": 2.43, "learning_rate": 4.576195933146257e-07, "loss": 0.5106, "step": 20410 }, { "epoch": 2.43, "learning_rate": 4.574342435521681e-07, "loss": 0.5506, "step": 20411 }, { "epoch": 2.43, "learning_rate": 4.5724892755349333e-07, "loss": 0.6063, "step": 20412 }, { "epoch": 2.43, "learning_rate": 4.5706364532166535e-07, "loss": 0.378, "step": 20413 }, { "epoch": 2.43, "learning_rate": 4.568783968597471e-07, "loss": 0.3402, "step": 20414 }, { "epoch": 2.43, "learning_rate": 4.5669318217079985e-07, "loss": 0.475, "step": 20415 }, { "epoch": 2.43, "learning_rate": 4.565080012578857e-07, "loss": 0.4316, "step": 20416 }, { "epoch": 2.43, "learning_rate": 4.563228541240658e-07, "loss": 0.5199, "step": 20417 }, { "epoch": 2.43, "learning_rate": 4.561377407724013e-07, "loss": 0.6455, "step": 20418 }, { "epoch": 2.43, "learning_rate": 4.5595266120595116e-07, "loss": 0.5166, "step": 20419 }, { "epoch": 2.43, "learning_rate": 4.557676154277746e-07, "loss": 0.3523, "step": 20420 }, { "epoch": 2.43, "learning_rate": 4.555826034409311e-07, "loss": 0.4084, "step": 20421 }, { "epoch": 2.43, "learning_rate": 4.5539762524847905e-07, "loss": 0.4567, "step": 20422 }, { "epoch": 2.43, "learning_rate": 4.5521268085347515e-07, "loss": 0.4252, "step": 20423 }, { "epoch": 2.43, "learning_rate": 4.550277702589773e-07, "loss": 0.4985, "step": 20424 }, { "epoch": 2.43, "learning_rate": 4.548428934680427e-07, "loss": 0.4986, "step": 20425 }, { "epoch": 2.43, "learning_rate": 4.546580504837267e-07, "loss": 0.2406, "step": 20426 }, { "epoch": 2.43, "learning_rate": 4.5447324130908383e-07, "loss": 0.3891, "step": 20427 }, { "epoch": 2.43, "learning_rate": 4.5428846594717026e-07, "loss": 0.4162, "step": 20428 }, { "epoch": 2.43, "learning_rate": 4.5410372440104043e-07, "loss": 0.395, "step": 20429 }, { "epoch": 2.43, "learning_rate": 4.53919016673747e-07, "loss": 0.4938, "step": 20430 }, { "epoch": 2.43, "learning_rate": 4.537343427683441e-07, "loss": 0.5863, "step": 20431 }, { "epoch": 2.43, "learning_rate": 4.5354970268788463e-07, "loss": 0.2903, "step": 20432 }, { "epoch": 2.43, "learning_rate": 4.5336509643542056e-07, "loss": 0.3864, "step": 20433 }, { "epoch": 2.43, "learning_rate": 4.531805240140025e-07, "loss": 0.4205, "step": 20434 }, { "epoch": 2.43, "learning_rate": 4.529959854266819e-07, "loss": 0.5825, "step": 20435 }, { "epoch": 2.43, "learning_rate": 4.5281148067650966e-07, "loss": 0.5643, "step": 20436 }, { "epoch": 2.43, "learning_rate": 4.5262700976653586e-07, "loss": 0.5525, "step": 20437 }, { "epoch": 2.43, "learning_rate": 4.5244257269980883e-07, "loss": 0.411, "step": 20438 }, { "epoch": 2.43, "learning_rate": 4.5225816947937843e-07, "loss": 0.2922, "step": 20439 }, { "epoch": 2.43, "learning_rate": 4.5207380010829223e-07, "loss": 0.3661, "step": 20440 }, { "epoch": 2.43, "learning_rate": 4.518894645895974e-07, "loss": 0.4554, "step": 20441 }, { "epoch": 2.43, "learning_rate": 4.5170516292634163e-07, "loss": 0.4778, "step": 20442 }, { "epoch": 2.43, "learning_rate": 4.515208951215716e-07, "loss": 0.5906, "step": 20443 }, { "epoch": 2.43, "learning_rate": 4.5133666117833344e-07, "loss": 0.6092, "step": 20444 }, { "epoch": 2.43, "learning_rate": 4.5115246109967147e-07, "loss": 0.2717, "step": 20445 }, { "epoch": 2.43, "learning_rate": 4.5096829488863124e-07, "loss": 0.411, "step": 20446 }, { "epoch": 2.44, "learning_rate": 4.507841625482579e-07, "loss": 0.464, "step": 20447 }, { "epoch": 2.44, "learning_rate": 4.5060006408159415e-07, "loss": 0.4844, "step": 20448 }, { "epoch": 2.44, "learning_rate": 4.5041599949168264e-07, "loss": 0.5607, "step": 20449 }, { "epoch": 2.44, "learning_rate": 4.502319687815665e-07, "loss": 0.6394, "step": 20450 }, { "epoch": 2.44, "learning_rate": 4.5004797195428826e-07, "loss": 0.3258, "step": 20451 }, { "epoch": 2.44, "learning_rate": 4.498640090128895e-07, "loss": 0.5209, "step": 20452 }, { "epoch": 2.44, "learning_rate": 4.496800799604098e-07, "loss": 0.405, "step": 20453 }, { "epoch": 2.44, "learning_rate": 4.494961847998913e-07, "loss": 0.4637, "step": 20454 }, { "epoch": 2.44, "learning_rate": 4.4931232353437297e-07, "loss": 0.557, "step": 20455 }, { "epoch": 2.44, "learning_rate": 4.491284961668932e-07, "loss": 0.5636, "step": 20456 }, { "epoch": 2.44, "learning_rate": 4.4894470270049175e-07, "loss": 0.4235, "step": 20457 }, { "epoch": 2.44, "learning_rate": 4.4876094313820617e-07, "loss": 0.3728, "step": 20458 }, { "epoch": 2.44, "learning_rate": 4.485772174830752e-07, "loss": 0.4597, "step": 20459 }, { "epoch": 2.44, "learning_rate": 4.4839352573813414e-07, "loss": 0.4098, "step": 20460 }, { "epoch": 2.44, "learning_rate": 4.482098679064209e-07, "loss": 0.5964, "step": 20461 }, { "epoch": 2.44, "learning_rate": 4.4802624399096994e-07, "loss": 0.6349, "step": 20462 }, { "epoch": 2.44, "learning_rate": 4.47842653994818e-07, "loss": 0.3241, "step": 20463 }, { "epoch": 2.44, "learning_rate": 4.476590979209988e-07, "loss": 0.3759, "step": 20464 }, { "epoch": 2.44, "learning_rate": 4.4747557577254685e-07, "loss": 0.3827, "step": 20465 }, { "epoch": 2.44, "learning_rate": 4.472920875524958e-07, "loss": 0.5102, "step": 20466 }, { "epoch": 2.44, "learning_rate": 4.4710863326387935e-07, "loss": 0.4904, "step": 20467 }, { "epoch": 2.44, "learning_rate": 4.4692521290972957e-07, "loss": 0.5744, "step": 20468 }, { "epoch": 2.44, "learning_rate": 4.4674182649307757e-07, "loss": 0.5744, "step": 20469 }, { "epoch": 2.44, "learning_rate": 4.46558474016956e-07, "loss": 0.3131, "step": 20470 }, { "epoch": 2.44, "learning_rate": 4.4637515548439454e-07, "loss": 0.4482, "step": 20471 }, { "epoch": 2.44, "learning_rate": 4.461918708984242e-07, "loss": 0.44, "step": 20472 }, { "epoch": 2.44, "learning_rate": 4.4600862026207446e-07, "loss": 0.5888, "step": 20473 }, { "epoch": 2.44, "learning_rate": 4.458254035783752e-07, "loss": 0.4977, "step": 20474 }, { "epoch": 2.44, "learning_rate": 4.4564222085035386e-07, "loss": 0.5207, "step": 20475 }, { "epoch": 2.44, "learning_rate": 4.454590720810395e-07, "loss": 0.3106, "step": 20476 }, { "epoch": 2.44, "learning_rate": 4.452759572734586e-07, "loss": 0.4303, "step": 20477 }, { "epoch": 2.44, "learning_rate": 4.4509287643063894e-07, "loss": 0.4151, "step": 20478 }, { "epoch": 2.44, "learning_rate": 4.449098295556059e-07, "loss": 0.4767, "step": 20479 }, { "epoch": 2.44, "learning_rate": 4.44726816651386e-07, "loss": 0.5818, "step": 20480 }, { "epoch": 2.44, "learning_rate": 4.44543837721004e-07, "loss": 0.6666, "step": 20481 }, { "epoch": 2.44, "learning_rate": 4.4436089276748554e-07, "loss": 0.2602, "step": 20482 }, { "epoch": 2.44, "learning_rate": 4.441779817938541e-07, "loss": 0.3529, "step": 20483 }, { "epoch": 2.44, "learning_rate": 4.4399510480313243e-07, "loss": 0.4503, "step": 20484 }, { "epoch": 2.44, "learning_rate": 4.438122617983442e-07, "loss": 0.5046, "step": 20485 }, { "epoch": 2.44, "learning_rate": 4.4362945278251247e-07, "loss": 0.6619, "step": 20486 }, { "epoch": 2.44, "learning_rate": 4.434466777586577e-07, "loss": 0.6837, "step": 20487 }, { "epoch": 2.44, "learning_rate": 4.432639367298022e-07, "loss": 0.3387, "step": 20488 }, { "epoch": 2.44, "learning_rate": 4.430812296989667e-07, "loss": 0.3636, "step": 20489 }, { "epoch": 2.44, "learning_rate": 4.4289855666917125e-07, "loss": 0.4334, "step": 20490 }, { "epoch": 2.44, "learning_rate": 4.4271591764343454e-07, "loss": 0.4543, "step": 20491 }, { "epoch": 2.44, "learning_rate": 4.4253331262477635e-07, "loss": 0.5891, "step": 20492 }, { "epoch": 2.44, "learning_rate": 4.4235074161621565e-07, "loss": 0.6165, "step": 20493 }, { "epoch": 2.44, "learning_rate": 4.4216820462076926e-07, "loss": 0.493, "step": 20494 }, { "epoch": 2.44, "learning_rate": 4.4198570164145527e-07, "loss": 0.3321, "step": 20495 }, { "epoch": 2.44, "learning_rate": 4.418032326812907e-07, "loss": 0.3791, "step": 20496 }, { "epoch": 2.44, "learning_rate": 4.4162079774329183e-07, "loss": 0.4017, "step": 20497 }, { "epoch": 2.44, "learning_rate": 4.414383968304728e-07, "loss": 0.5376, "step": 20498 }, { "epoch": 2.44, "learning_rate": 4.412560299458499e-07, "loss": 0.5725, "step": 20499 }, { "epoch": 2.44, "learning_rate": 4.410736970924376e-07, "loss": 0.588, "step": 20500 }, { "epoch": 2.44, "learning_rate": 4.408913982732507e-07, "loss": 0.2693, "step": 20501 }, { "epoch": 2.44, "learning_rate": 4.4070913349130077e-07, "loss": 0.4541, "step": 20502 }, { "epoch": 2.44, "learning_rate": 4.40526902749602e-07, "loss": 0.4124, "step": 20503 }, { "epoch": 2.44, "learning_rate": 4.4034470605116673e-07, "loss": 0.5199, "step": 20504 }, { "epoch": 2.44, "learning_rate": 4.4016254339900615e-07, "loss": 0.5166, "step": 20505 }, { "epoch": 2.44, "learning_rate": 4.399804147961312e-07, "loss": 0.6326, "step": 20506 }, { "epoch": 2.44, "learning_rate": 4.3979832024555274e-07, "loss": 0.2942, "step": 20507 }, { "epoch": 2.44, "learning_rate": 4.3961625975028177e-07, "loss": 0.3287, "step": 20508 }, { "epoch": 2.44, "learning_rate": 4.3943423331332587e-07, "loss": 0.4161, "step": 20509 }, { "epoch": 2.44, "learning_rate": 4.392522409376954e-07, "loss": 0.4998, "step": 20510 }, { "epoch": 2.44, "learning_rate": 4.3907028262639875e-07, "loss": 0.6303, "step": 20511 }, { "epoch": 2.44, "learning_rate": 4.38888358382443e-07, "loss": 0.5398, "step": 20512 }, { "epoch": 2.44, "learning_rate": 4.3870646820883517e-07, "loss": 0.3995, "step": 20513 }, { "epoch": 2.44, "learning_rate": 4.385246121085826e-07, "loss": 0.3639, "step": 20514 }, { "epoch": 2.44, "learning_rate": 4.3834279008469096e-07, "loss": 0.4554, "step": 20515 }, { "epoch": 2.44, "learning_rate": 4.381610021401667e-07, "loss": 0.4593, "step": 20516 }, { "epoch": 2.44, "learning_rate": 4.379792482780132e-07, "loss": 0.4853, "step": 20517 }, { "epoch": 2.44, "learning_rate": 4.3779752850123677e-07, "loss": 0.6461, "step": 20518 }, { "epoch": 2.44, "learning_rate": 4.376158428128399e-07, "loss": 0.5828, "step": 20519 }, { "epoch": 2.44, "learning_rate": 4.374341912158259e-07, "loss": 0.2528, "step": 20520 }, { "epoch": 2.44, "learning_rate": 4.3725257371319744e-07, "loss": 0.4542, "step": 20521 }, { "epoch": 2.44, "learning_rate": 4.37070990307957e-07, "loss": 0.424, "step": 20522 }, { "epoch": 2.44, "learning_rate": 4.3688944100310716e-07, "loss": 0.5139, "step": 20523 }, { "epoch": 2.44, "learning_rate": 4.367079258016471e-07, "loss": 0.6536, "step": 20524 }, { "epoch": 2.44, "learning_rate": 4.3652644470657897e-07, "loss": 0.5528, "step": 20525 }, { "epoch": 2.44, "learning_rate": 4.363449977209008e-07, "loss": 0.2974, "step": 20526 }, { "epoch": 2.44, "learning_rate": 4.3616358484761395e-07, "loss": 0.3403, "step": 20527 }, { "epoch": 2.44, "learning_rate": 4.3598220608971564e-07, "loss": 0.4714, "step": 20528 }, { "epoch": 2.44, "learning_rate": 4.3580086145020434e-07, "loss": 0.4787, "step": 20529 }, { "epoch": 2.44, "learning_rate": 4.356195509320782e-07, "loss": 0.5502, "step": 20530 }, { "epoch": 2.45, "learning_rate": 4.3543827453833483e-07, "loss": 0.5846, "step": 20531 }, { "epoch": 2.45, "learning_rate": 4.352570322719693e-07, "loss": 0.3785, "step": 20532 }, { "epoch": 2.45, "learning_rate": 4.350758241359787e-07, "loss": 0.3699, "step": 20533 }, { "epoch": 2.45, "learning_rate": 4.348946501333581e-07, "loss": 0.3522, "step": 20534 }, { "epoch": 2.45, "learning_rate": 4.3471351026710187e-07, "loss": 0.5067, "step": 20535 }, { "epoch": 2.45, "learning_rate": 4.345324045402044e-07, "loss": 0.5681, "step": 20536 }, { "epoch": 2.45, "learning_rate": 4.343513329556595e-07, "loss": 0.5711, "step": 20537 }, { "epoch": 2.45, "learning_rate": 4.341702955164609e-07, "loss": 0.4091, "step": 20538 }, { "epoch": 2.45, "learning_rate": 4.339892922256003e-07, "loss": 0.3615, "step": 20539 }, { "epoch": 2.45, "learning_rate": 4.3380832308607067e-07, "loss": 0.466, "step": 20540 }, { "epoch": 2.45, "learning_rate": 4.336273881008621e-07, "loss": 0.4396, "step": 20541 }, { "epoch": 2.45, "learning_rate": 4.334464872729668e-07, "loss": 0.5045, "step": 20542 }, { "epoch": 2.45, "learning_rate": 4.33265620605374e-07, "loss": 0.57, "step": 20543 }, { "epoch": 2.45, "learning_rate": 4.330847881010736e-07, "loss": 0.4814, "step": 20544 }, { "epoch": 2.45, "learning_rate": 4.329039897630552e-07, "loss": 0.4332, "step": 20545 }, { "epoch": 2.45, "learning_rate": 4.3272322559430815e-07, "loss": 0.467, "step": 20546 }, { "epoch": 2.45, "learning_rate": 4.325424955978194e-07, "loss": 0.3921, "step": 20547 }, { "epoch": 2.45, "learning_rate": 4.323617997765761e-07, "loss": 0.4818, "step": 20548 }, { "epoch": 2.45, "learning_rate": 4.321811381335664e-07, "loss": 0.5938, "step": 20549 }, { "epoch": 2.45, "learning_rate": 4.320005106717753e-07, "loss": 0.6088, "step": 20550 }, { "epoch": 2.45, "learning_rate": 4.318199173941892e-07, "loss": 0.2725, "step": 20551 }, { "epoch": 2.45, "learning_rate": 4.316393583037937e-07, "loss": 0.3689, "step": 20552 }, { "epoch": 2.45, "learning_rate": 4.3145883340357353e-07, "loss": 0.4708, "step": 20553 }, { "epoch": 2.45, "learning_rate": 4.3127834269651255e-07, "loss": 0.4992, "step": 20554 }, { "epoch": 2.45, "learning_rate": 4.310978861855933e-07, "loss": 0.5696, "step": 20555 }, { "epoch": 2.45, "learning_rate": 4.309174638737998e-07, "loss": 0.5134, "step": 20556 }, { "epoch": 2.45, "learning_rate": 4.3073707576411465e-07, "loss": 0.4117, "step": 20557 }, { "epoch": 2.45, "learning_rate": 4.305567218595189e-07, "loss": 0.4323, "step": 20558 }, { "epoch": 2.45, "learning_rate": 4.3037640216299417e-07, "loss": 0.4236, "step": 20559 }, { "epoch": 2.45, "learning_rate": 4.301961166775209e-07, "loss": 0.4775, "step": 20560 }, { "epoch": 2.45, "learning_rate": 4.3001586540608016e-07, "loss": 0.6238, "step": 20561 }, { "epoch": 2.45, "learning_rate": 4.298356483516511e-07, "loss": 0.5976, "step": 20562 }, { "epoch": 2.45, "learning_rate": 4.2965546551721167e-07, "loss": 0.3268, "step": 20563 }, { "epoch": 2.45, "learning_rate": 4.294753169057411e-07, "loss": 0.3914, "step": 20564 }, { "epoch": 2.45, "learning_rate": 4.2929520252021796e-07, "loss": 0.3633, "step": 20565 }, { "epoch": 2.45, "learning_rate": 4.2911512236361833e-07, "loss": 0.4436, "step": 20566 }, { "epoch": 2.45, "learning_rate": 4.289350764389194e-07, "loss": 0.5455, "step": 20567 }, { "epoch": 2.45, "learning_rate": 4.287550647490979e-07, "loss": 0.601, "step": 20568 }, { "epoch": 2.45, "learning_rate": 4.285750872971292e-07, "loss": 0.577, "step": 20569 }, { "epoch": 2.45, "learning_rate": 4.2839514408598736e-07, "loss": 0.3403, "step": 20570 }, { "epoch": 2.45, "learning_rate": 4.2821523511864775e-07, "loss": 0.295, "step": 20571 }, { "epoch": 2.45, "learning_rate": 4.280353603980847e-07, "loss": 0.42, "step": 20572 }, { "epoch": 2.45, "learning_rate": 4.278555199272702e-07, "loss": 0.4586, "step": 20573 }, { "epoch": 2.45, "learning_rate": 4.2767571370917806e-07, "loss": 0.5154, "step": 20574 }, { "epoch": 2.45, "learning_rate": 4.274959417467808e-07, "loss": 0.5954, "step": 20575 }, { "epoch": 2.45, "learning_rate": 4.2731620404304953e-07, "loss": 0.3444, "step": 20576 }, { "epoch": 2.45, "learning_rate": 4.2713650060095453e-07, "loss": 0.4026, "step": 20577 }, { "epoch": 2.45, "learning_rate": 4.2695683142346734e-07, "loss": 0.49, "step": 20578 }, { "epoch": 2.45, "learning_rate": 4.267771965135575e-07, "loss": 0.5203, "step": 20579 }, { "epoch": 2.45, "learning_rate": 4.2659759587419526e-07, "loss": 0.6414, "step": 20580 }, { "epoch": 2.45, "learning_rate": 4.264180295083481e-07, "loss": 0.5922, "step": 20581 }, { "epoch": 2.45, "learning_rate": 4.262384974189857e-07, "loss": 0.3651, "step": 20582 }, { "epoch": 2.45, "learning_rate": 4.260589996090747e-07, "loss": 0.394, "step": 20583 }, { "epoch": 2.45, "learning_rate": 4.258795360815818e-07, "loss": 0.3684, "step": 20584 }, { "epoch": 2.45, "learning_rate": 4.2570010683947447e-07, "loss": 0.4717, "step": 20585 }, { "epoch": 2.45, "learning_rate": 4.2552071188571827e-07, "loss": 0.4651, "step": 20586 }, { "epoch": 2.45, "learning_rate": 4.253413512232793e-07, "loss": 0.6642, "step": 20587 }, { "epoch": 2.45, "learning_rate": 4.251620248551214e-07, "loss": 0.3168, "step": 20588 }, { "epoch": 2.45, "learning_rate": 4.249827327842093e-07, "loss": 0.2609, "step": 20589 }, { "epoch": 2.45, "learning_rate": 4.2480347501350716e-07, "loss": 0.3791, "step": 20590 }, { "epoch": 2.45, "learning_rate": 4.246242515459778e-07, "loss": 0.3956, "step": 20591 }, { "epoch": 2.45, "learning_rate": 4.2444506238458335e-07, "loss": 0.5179, "step": 20592 }, { "epoch": 2.45, "learning_rate": 4.2426590753228583e-07, "loss": 0.5357, "step": 20593 }, { "epoch": 2.45, "learning_rate": 4.24086786992047e-07, "loss": 0.4801, "step": 20594 }, { "epoch": 2.45, "learning_rate": 4.239077007668285e-07, "loss": 0.3603, "step": 20595 }, { "epoch": 2.45, "learning_rate": 4.237286488595896e-07, "loss": 0.4431, "step": 20596 }, { "epoch": 2.45, "learning_rate": 4.235496312732906e-07, "loss": 0.4469, "step": 20597 }, { "epoch": 2.45, "learning_rate": 4.233706480108904e-07, "loss": 0.5444, "step": 20598 }, { "epoch": 2.45, "learning_rate": 4.2319169907534706e-07, "loss": 0.5406, "step": 20599 }, { "epoch": 2.45, "learning_rate": 4.2301278446961943e-07, "loss": 0.5872, "step": 20600 }, { "epoch": 2.45, "learning_rate": 4.228339041966645e-07, "loss": 0.3219, "step": 20601 }, { "epoch": 2.45, "learning_rate": 4.226550582594402e-07, "loss": 0.379, "step": 20602 }, { "epoch": 2.45, "learning_rate": 4.2247624666090116e-07, "loss": 0.3772, "step": 20603 }, { "epoch": 2.45, "learning_rate": 4.2229746940400496e-07, "loss": 0.4972, "step": 20604 }, { "epoch": 2.45, "learning_rate": 4.221187264917054e-07, "loss": 0.5572, "step": 20605 }, { "epoch": 2.45, "learning_rate": 4.219400179269581e-07, "loss": 0.5822, "step": 20606 }, { "epoch": 2.45, "learning_rate": 4.2176134371271606e-07, "loss": 0.4833, "step": 20607 }, { "epoch": 2.45, "learning_rate": 4.215827038519335e-07, "loss": 0.3355, "step": 20608 }, { "epoch": 2.45, "learning_rate": 4.214040983475631e-07, "loss": 0.441, "step": 20609 }, { "epoch": 2.45, "learning_rate": 4.212255272025578e-07, "loss": 0.5106, "step": 20610 }, { "epoch": 2.45, "learning_rate": 4.2104699041986943e-07, "loss": 0.5129, "step": 20611 }, { "epoch": 2.45, "learning_rate": 4.208684880024477e-07, "loss": 0.5964, "step": 20612 }, { "epoch": 2.45, "learning_rate": 4.2069001995324517e-07, "loss": 0.3504, "step": 20613 }, { "epoch": 2.45, "learning_rate": 4.2051158627521046e-07, "loss": 0.3167, "step": 20614 }, { "epoch": 2.46, "learning_rate": 4.203331869712937e-07, "loss": 0.5099, "step": 20615 }, { "epoch": 2.46, "learning_rate": 4.2015482204444375e-07, "loss": 0.4221, "step": 20616 }, { "epoch": 2.46, "learning_rate": 4.1997649149760984e-07, "loss": 0.4954, "step": 20617 }, { "epoch": 2.46, "learning_rate": 4.1979819533373824e-07, "loss": 0.6299, "step": 20618 }, { "epoch": 2.46, "learning_rate": 4.19619933555778e-07, "loss": 0.4902, "step": 20619 }, { "epoch": 2.46, "learning_rate": 4.194417061666739e-07, "loss": 0.3085, "step": 20620 }, { "epoch": 2.46, "learning_rate": 4.192635131693737e-07, "loss": 0.303, "step": 20621 }, { "epoch": 2.46, "learning_rate": 4.190853545668214e-07, "loss": 0.4524, "step": 20622 }, { "epoch": 2.46, "learning_rate": 4.189072303619629e-07, "loss": 0.526, "step": 20623 }, { "epoch": 2.46, "learning_rate": 4.187291405577426e-07, "loss": 0.5701, "step": 20624 }, { "epoch": 2.46, "learning_rate": 4.185510851571048e-07, "loss": 0.5863, "step": 20625 }, { "epoch": 2.46, "learning_rate": 4.183730641629924e-07, "loss": 0.2496, "step": 20626 }, { "epoch": 2.46, "learning_rate": 4.181950775783469e-07, "loss": 0.3982, "step": 20627 }, { "epoch": 2.46, "learning_rate": 4.180171254061119e-07, "loss": 0.4831, "step": 20628 }, { "epoch": 2.46, "learning_rate": 4.178392076492288e-07, "loss": 0.4705, "step": 20629 }, { "epoch": 2.46, "learning_rate": 4.1766132431063783e-07, "loss": 0.5572, "step": 20630 }, { "epoch": 2.46, "learning_rate": 4.174834753932797e-07, "loss": 0.6215, "step": 20631 }, { "epoch": 2.46, "learning_rate": 4.173056609000953e-07, "loss": 0.3954, "step": 20632 }, { "epoch": 2.46, "learning_rate": 4.171278808340229e-07, "loss": 0.3788, "step": 20633 }, { "epoch": 2.46, "learning_rate": 4.16950135198001e-07, "loss": 0.4662, "step": 20634 }, { "epoch": 2.46, "learning_rate": 4.1677242399496804e-07, "loss": 0.4154, "step": 20635 }, { "epoch": 2.46, "learning_rate": 4.1659474722786225e-07, "loss": 0.4894, "step": 20636 }, { "epoch": 2.46, "learning_rate": 4.164171048996196e-07, "loss": 0.5564, "step": 20637 }, { "epoch": 2.46, "learning_rate": 4.16239497013177e-07, "loss": 0.3404, "step": 20638 }, { "epoch": 2.46, "learning_rate": 4.160619235714708e-07, "loss": 0.3425, "step": 20639 }, { "epoch": 2.46, "learning_rate": 4.1588438457743603e-07, "loss": 0.4119, "step": 20640 }, { "epoch": 2.46, "learning_rate": 4.1570688003400665e-07, "loss": 0.3925, "step": 20641 }, { "epoch": 2.46, "learning_rate": 4.155294099441173e-07, "loss": 0.5545, "step": 20642 }, { "epoch": 2.46, "learning_rate": 4.153519743107018e-07, "loss": 0.5136, "step": 20643 }, { "epoch": 2.46, "learning_rate": 4.151745731366938e-07, "loss": 0.4734, "step": 20644 }, { "epoch": 2.46, "learning_rate": 4.149972064250243e-07, "loss": 0.284, "step": 20645 }, { "epoch": 2.46, "learning_rate": 4.1481987417862577e-07, "loss": 0.4109, "step": 20646 }, { "epoch": 2.46, "learning_rate": 4.146425764004305e-07, "loss": 0.4817, "step": 20647 }, { "epoch": 2.46, "learning_rate": 4.1446531309336846e-07, "loss": 0.5683, "step": 20648 }, { "epoch": 2.46, "learning_rate": 4.1428808426036916e-07, "loss": 0.5575, "step": 20649 }, { "epoch": 2.46, "learning_rate": 4.1411088990436277e-07, "loss": 0.6964, "step": 20650 }, { "epoch": 2.46, "learning_rate": 4.139337300282789e-07, "loss": 0.2959, "step": 20651 }, { "epoch": 2.46, "learning_rate": 4.13756604635045e-07, "loss": 0.3902, "step": 20652 }, { "epoch": 2.46, "learning_rate": 4.135795137275894e-07, "loss": 0.4393, "step": 20653 }, { "epoch": 2.46, "learning_rate": 4.1340245730884023e-07, "loss": 0.5468, "step": 20654 }, { "epoch": 2.46, "learning_rate": 4.132254353817236e-07, "loss": 0.5757, "step": 20655 }, { "epoch": 2.46, "learning_rate": 4.130484479491648e-07, "loss": 0.5271, "step": 20656 }, { "epoch": 2.46, "learning_rate": 4.1287149501409054e-07, "loss": 0.4165, "step": 20657 }, { "epoch": 2.46, "learning_rate": 4.1269457657942517e-07, "loss": 0.3703, "step": 20658 }, { "epoch": 2.46, "learning_rate": 4.125176926480945e-07, "loss": 0.4032, "step": 20659 }, { "epoch": 2.46, "learning_rate": 4.123408432230208e-07, "loss": 0.6016, "step": 20660 }, { "epoch": 2.46, "learning_rate": 4.1216402830712875e-07, "loss": 0.5719, "step": 20661 }, { "epoch": 2.46, "learning_rate": 4.119872479033407e-07, "loss": 0.6072, "step": 20662 }, { "epoch": 2.46, "learning_rate": 4.1181050201457765e-07, "loss": 0.4143, "step": 20663 }, { "epoch": 2.46, "learning_rate": 4.116337906437626e-07, "loss": 0.3503, "step": 20664 }, { "epoch": 2.46, "learning_rate": 4.1145711379381616e-07, "loss": 0.4045, "step": 20665 }, { "epoch": 2.46, "learning_rate": 4.1128047146765936e-07, "loss": 0.4608, "step": 20666 }, { "epoch": 2.46, "learning_rate": 4.11103863668211e-07, "loss": 0.5155, "step": 20667 }, { "epoch": 2.46, "learning_rate": 4.109272903983916e-07, "loss": 0.5963, "step": 20668 }, { "epoch": 2.46, "learning_rate": 4.107507516611189e-07, "loss": 0.559, "step": 20669 }, { "epoch": 2.46, "learning_rate": 4.105742474593122e-07, "loss": 0.3202, "step": 20670 }, { "epoch": 2.46, "learning_rate": 4.1039777779588757e-07, "loss": 0.415, "step": 20671 }, { "epoch": 2.46, "learning_rate": 4.1022134267376333e-07, "loss": 0.4221, "step": 20672 }, { "epoch": 2.46, "learning_rate": 4.1004494209585577e-07, "loss": 0.4904, "step": 20673 }, { "epoch": 2.46, "learning_rate": 4.0986857606508085e-07, "loss": 0.5887, "step": 20674 }, { "epoch": 2.46, "learning_rate": 4.0969224458435347e-07, "loss": 0.6069, "step": 20675 }, { "epoch": 2.46, "learning_rate": 4.095159476565894e-07, "loss": 0.3967, "step": 20676 }, { "epoch": 2.46, "learning_rate": 4.0933968528470176e-07, "loss": 0.2925, "step": 20677 }, { "epoch": 2.46, "learning_rate": 4.091634574716041e-07, "loss": 0.3629, "step": 20678 }, { "epoch": 2.46, "learning_rate": 4.0898726422021e-07, "loss": 0.4068, "step": 20679 }, { "epoch": 2.46, "learning_rate": 4.0881110553343176e-07, "loss": 0.5701, "step": 20680 }, { "epoch": 2.46, "learning_rate": 4.086349814141821e-07, "loss": 0.5366, "step": 20681 }, { "epoch": 2.46, "learning_rate": 4.084588918653709e-07, "loss": 0.2397, "step": 20682 }, { "epoch": 2.46, "learning_rate": 4.0828283688991057e-07, "loss": 0.4136, "step": 20683 }, { "epoch": 2.46, "learning_rate": 4.0810681649070954e-07, "loss": 0.414, "step": 20684 }, { "epoch": 2.46, "learning_rate": 4.0793083067067917e-07, "loss": 0.4475, "step": 20685 }, { "epoch": 2.46, "learning_rate": 4.077548794327271e-07, "loss": 0.475, "step": 20686 }, { "epoch": 2.46, "learning_rate": 4.0757896277976243e-07, "loss": 0.5777, "step": 20687 }, { "epoch": 2.46, "learning_rate": 4.074030807146931e-07, "loss": 0.3313, "step": 20688 }, { "epoch": 2.46, "learning_rate": 4.07227233240427e-07, "loss": 0.3318, "step": 20689 }, { "epoch": 2.46, "learning_rate": 4.070514203598702e-07, "loss": 0.316, "step": 20690 }, { "epoch": 2.46, "learning_rate": 4.0687564207592874e-07, "loss": 0.4853, "step": 20691 }, { "epoch": 2.46, "learning_rate": 4.066998983915088e-07, "loss": 0.575, "step": 20692 }, { "epoch": 2.46, "learning_rate": 4.065241893095148e-07, "loss": 0.544, "step": 20693 }, { "epoch": 2.46, "learning_rate": 4.063485148328514e-07, "loss": 0.5916, "step": 20694 }, { "epoch": 2.46, "learning_rate": 4.061728749644228e-07, "loss": 0.2806, "step": 20695 }, { "epoch": 2.46, "learning_rate": 4.059972697071329e-07, "loss": 0.4, "step": 20696 }, { "epoch": 2.46, "learning_rate": 4.058216990638841e-07, "loss": 0.3755, "step": 20697 }, { "epoch": 2.46, "learning_rate": 4.0564616303757744e-07, "loss": 0.549, "step": 20698 }, { "epoch": 2.47, "learning_rate": 4.0547066163111547e-07, "loss": 0.5824, "step": 20699 }, { "epoch": 2.47, "learning_rate": 4.0529519484739986e-07, "loss": 0.5258, "step": 20700 }, { "epoch": 2.47, "learning_rate": 4.0511976268932965e-07, "loss": 0.2594, "step": 20701 }, { "epoch": 2.47, "learning_rate": 4.0494436515980587e-07, "loss": 0.3424, "step": 20702 }, { "epoch": 2.47, "learning_rate": 4.0476900226172733e-07, "loss": 0.3828, "step": 20703 }, { "epoch": 2.47, "learning_rate": 4.045936739979936e-07, "loss": 0.4902, "step": 20704 }, { "epoch": 2.47, "learning_rate": 4.044183803715021e-07, "loss": 0.5395, "step": 20705 }, { "epoch": 2.47, "learning_rate": 4.0424312138515024e-07, "loss": 0.5753, "step": 20706 }, { "epoch": 2.47, "learning_rate": 4.0406789704183547e-07, "loss": 0.3132, "step": 20707 }, { "epoch": 2.47, "learning_rate": 4.0389270734445455e-07, "loss": 0.3967, "step": 20708 }, { "epoch": 2.47, "learning_rate": 4.0371755229590273e-07, "loss": 0.3732, "step": 20709 }, { "epoch": 2.47, "learning_rate": 4.035424318990755e-07, "loss": 0.458, "step": 20710 }, { "epoch": 2.47, "learning_rate": 4.033673461568685e-07, "loss": 0.6184, "step": 20711 }, { "epoch": 2.47, "learning_rate": 4.0319229507217505e-07, "loss": 0.5876, "step": 20712 }, { "epoch": 2.47, "learning_rate": 4.0301727864788835e-07, "loss": 0.2942, "step": 20713 }, { "epoch": 2.47, "learning_rate": 4.02842296886902e-07, "loss": 0.4209, "step": 20714 }, { "epoch": 2.47, "learning_rate": 4.0266734979210906e-07, "loss": 0.3928, "step": 20715 }, { "epoch": 2.47, "learning_rate": 4.024924373664002e-07, "loss": 0.538, "step": 20716 }, { "epoch": 2.47, "learning_rate": 4.0231755961266746e-07, "loss": 0.6418, "step": 20717 }, { "epoch": 2.47, "learning_rate": 4.0214271653380184e-07, "loss": 0.5824, "step": 20718 }, { "epoch": 2.47, "learning_rate": 4.0196790813269324e-07, "loss": 0.6777, "step": 20719 }, { "epoch": 2.47, "learning_rate": 4.0179313441223054e-07, "loss": 0.2938, "step": 20720 }, { "epoch": 2.47, "learning_rate": 4.016183953753036e-07, "loss": 0.3935, "step": 20721 }, { "epoch": 2.47, "learning_rate": 4.014436910248004e-07, "loss": 0.4697, "step": 20722 }, { "epoch": 2.47, "learning_rate": 4.0126902136360974e-07, "loss": 0.5112, "step": 20723 }, { "epoch": 2.47, "learning_rate": 4.0109438639461764e-07, "loss": 0.6244, "step": 20724 }, { "epoch": 2.47, "learning_rate": 4.00919786120712e-07, "loss": 0.5313, "step": 20725 }, { "epoch": 2.47, "learning_rate": 4.007452205447784e-07, "loss": 0.3569, "step": 20726 }, { "epoch": 2.47, "learning_rate": 4.0057068966970205e-07, "loss": 0.3493, "step": 20727 }, { "epoch": 2.47, "learning_rate": 4.0039619349836827e-07, "loss": 0.4372, "step": 20728 }, { "epoch": 2.47, "learning_rate": 4.0022173203366157e-07, "loss": 0.3938, "step": 20729 }, { "epoch": 2.47, "learning_rate": 4.000473052784665e-07, "loss": 0.5512, "step": 20730 }, { "epoch": 2.47, "learning_rate": 3.998729132356649e-07, "loss": 0.5889, "step": 20731 }, { "epoch": 2.47, "learning_rate": 3.996985559081404e-07, "loss": 0.2062, "step": 20732 }, { "epoch": 2.47, "learning_rate": 3.9952423329877533e-07, "loss": 0.3676, "step": 20733 }, { "epoch": 2.47, "learning_rate": 3.993499454104513e-07, "loss": 0.4044, "step": 20734 }, { "epoch": 2.47, "learning_rate": 3.99175692246048e-07, "loss": 0.4192, "step": 20735 }, { "epoch": 2.47, "learning_rate": 3.9900147380844716e-07, "loss": 0.5511, "step": 20736 }, { "epoch": 2.47, "learning_rate": 3.98827290100528e-07, "loss": 0.5343, "step": 20737 }, { "epoch": 2.47, "learning_rate": 3.9865314112517066e-07, "loss": 0.4656, "step": 20738 }, { "epoch": 2.47, "learning_rate": 3.984790268852526e-07, "loss": 0.3133, "step": 20739 }, { "epoch": 2.47, "learning_rate": 3.983049473836531e-07, "loss": 0.4193, "step": 20740 }, { "epoch": 2.47, "learning_rate": 3.981309026232494e-07, "loss": 0.5077, "step": 20741 }, { "epoch": 2.47, "learning_rate": 3.979568926069174e-07, "loss": 0.5879, "step": 20742 }, { "epoch": 2.47, "learning_rate": 3.977829173375344e-07, "loss": 0.5722, "step": 20743 }, { "epoch": 2.47, "learning_rate": 3.9760897681797613e-07, "loss": 0.4398, "step": 20744 }, { "epoch": 2.47, "learning_rate": 3.9743507105111876e-07, "loss": 0.2946, "step": 20745 }, { "epoch": 2.47, "learning_rate": 3.972612000398351e-07, "loss": 0.2968, "step": 20746 }, { "epoch": 2.47, "learning_rate": 3.9708736378700095e-07, "loss": 0.4936, "step": 20747 }, { "epoch": 2.47, "learning_rate": 3.9691356229548864e-07, "loss": 0.5708, "step": 20748 }, { "epoch": 2.47, "learning_rate": 3.967397955681723e-07, "loss": 0.4682, "step": 20749 }, { "epoch": 2.47, "learning_rate": 3.965660636079227e-07, "loss": 0.6921, "step": 20750 }, { "epoch": 2.47, "learning_rate": 3.963923664176128e-07, "loss": 0.3053, "step": 20751 }, { "epoch": 2.47, "learning_rate": 3.9621870400011366e-07, "loss": 0.4151, "step": 20752 }, { "epoch": 2.47, "learning_rate": 3.9604507635829624e-07, "loss": 0.4639, "step": 20753 }, { "epoch": 2.47, "learning_rate": 3.958714834950303e-07, "loss": 0.4436, "step": 20754 }, { "epoch": 2.47, "learning_rate": 3.9569792541318486e-07, "loss": 0.5537, "step": 20755 }, { "epoch": 2.47, "learning_rate": 3.955244021156296e-07, "loss": 0.6306, "step": 20756 }, { "epoch": 2.47, "learning_rate": 3.953509136052322e-07, "loss": 0.2972, "step": 20757 }, { "epoch": 2.47, "learning_rate": 3.9517745988486066e-07, "loss": 0.4179, "step": 20758 }, { "epoch": 2.47, "learning_rate": 3.950040409573824e-07, "loss": 0.4171, "step": 20759 }, { "epoch": 2.47, "learning_rate": 3.9483065682566436e-07, "loss": 0.5146, "step": 20760 }, { "epoch": 2.47, "learning_rate": 3.9465730749257193e-07, "loss": 0.5507, "step": 20761 }, { "epoch": 2.47, "learning_rate": 3.9448399296097116e-07, "loss": 0.5836, "step": 20762 }, { "epoch": 2.47, "learning_rate": 3.9431071323372623e-07, "loss": 0.3868, "step": 20763 }, { "epoch": 2.47, "learning_rate": 3.9413746831370255e-07, "loss": 0.3392, "step": 20764 }, { "epoch": 2.47, "learning_rate": 3.9396425820376225e-07, "loss": 0.4409, "step": 20765 }, { "epoch": 2.47, "learning_rate": 3.9379108290676983e-07, "loss": 0.4009, "step": 20766 }, { "epoch": 2.47, "learning_rate": 3.936179424255873e-07, "loss": 0.4937, "step": 20767 }, { "epoch": 2.47, "learning_rate": 3.9344483676307777e-07, "loss": 0.5234, "step": 20768 }, { "epoch": 2.47, "learning_rate": 3.9327176592210164e-07, "loss": 0.5265, "step": 20769 }, { "epoch": 2.47, "learning_rate": 3.9309872990551945e-07, "loss": 0.3648, "step": 20770 }, { "epoch": 2.47, "learning_rate": 3.929257287161922e-07, "loss": 0.4028, "step": 20771 }, { "epoch": 2.47, "learning_rate": 3.927527623569799e-07, "loss": 0.3901, "step": 20772 }, { "epoch": 2.47, "learning_rate": 3.9257983083074075e-07, "loss": 0.4958, "step": 20773 }, { "epoch": 2.47, "learning_rate": 3.924069341403336e-07, "loss": 0.6666, "step": 20774 }, { "epoch": 2.47, "learning_rate": 3.9223407228861754e-07, "loss": 0.5204, "step": 20775 }, { "epoch": 2.47, "learning_rate": 3.920612452784489e-07, "loss": 0.307, "step": 20776 }, { "epoch": 2.47, "learning_rate": 3.918884531126843e-07, "loss": 0.3602, "step": 20777 }, { "epoch": 2.47, "learning_rate": 3.9171569579418033e-07, "loss": 0.4338, "step": 20778 }, { "epoch": 2.47, "learning_rate": 3.915429733257936e-07, "loss": 0.4674, "step": 20779 }, { "epoch": 2.47, "learning_rate": 3.9137028571037765e-07, "loss": 0.6125, "step": 20780 }, { "epoch": 2.47, "learning_rate": 3.911976329507877e-07, "loss": 0.6401, "step": 20781 }, { "epoch": 2.47, "learning_rate": 3.910250150498787e-07, "loss": 0.3012, "step": 20782 }, { "epoch": 2.48, "learning_rate": 3.908524320105031e-07, "loss": 0.4552, "step": 20783 }, { "epoch": 2.48, "learning_rate": 3.9067988383551304e-07, "loss": 0.4088, "step": 20784 }, { "epoch": 2.48, "learning_rate": 3.905073705277615e-07, "loss": 0.5057, "step": 20785 }, { "epoch": 2.48, "learning_rate": 3.903348920901004e-07, "loss": 0.5675, "step": 20786 }, { "epoch": 2.48, "learning_rate": 3.9016244852538074e-07, "loss": 0.6779, "step": 20787 }, { "epoch": 2.48, "learning_rate": 3.8999003983645255e-07, "loss": 0.4451, "step": 20788 }, { "epoch": 2.48, "learning_rate": 3.89817666026166e-07, "loss": 0.3137, "step": 20789 }, { "epoch": 2.48, "learning_rate": 3.8964532709737103e-07, "loss": 0.4007, "step": 20790 }, { "epoch": 2.48, "learning_rate": 3.8947302305291564e-07, "loss": 0.4631, "step": 20791 }, { "epoch": 2.48, "learning_rate": 3.893007538956481e-07, "loss": 0.5378, "step": 20792 }, { "epoch": 2.48, "learning_rate": 3.891285196284159e-07, "loss": 0.6534, "step": 20793 }, { "epoch": 2.48, "learning_rate": 3.8895632025406695e-07, "loss": 0.5409, "step": 20794 }, { "epoch": 2.48, "learning_rate": 3.887841557754468e-07, "loss": 0.281, "step": 20795 }, { "epoch": 2.48, "learning_rate": 3.886120261954016e-07, "loss": 0.3849, "step": 20796 }, { "epoch": 2.48, "learning_rate": 3.884399315167772e-07, "loss": 0.4687, "step": 20797 }, { "epoch": 2.48, "learning_rate": 3.882678717424182e-07, "loss": 0.4602, "step": 20798 }, { "epoch": 2.48, "learning_rate": 3.880958468751678e-07, "loss": 0.4905, "step": 20799 }, { "epoch": 2.48, "learning_rate": 3.8792385691787005e-07, "loss": 0.5283, "step": 20800 }, { "epoch": 2.48, "learning_rate": 3.877519018733683e-07, "loss": 0.2963, "step": 20801 }, { "epoch": 2.48, "learning_rate": 3.8757998174450556e-07, "loss": 0.3788, "step": 20802 }, { "epoch": 2.48, "learning_rate": 3.87408096534122e-07, "loss": 0.418, "step": 20803 }, { "epoch": 2.48, "learning_rate": 3.87236246245061e-07, "loss": 0.4868, "step": 20804 }, { "epoch": 2.48, "learning_rate": 3.870644308801616e-07, "loss": 0.5338, "step": 20805 }, { "epoch": 2.48, "learning_rate": 3.868926504422643e-07, "loss": 0.5316, "step": 20806 }, { "epoch": 2.48, "learning_rate": 3.8672090493420853e-07, "loss": 0.3671, "step": 20807 }, { "epoch": 2.48, "learning_rate": 3.865491943588334e-07, "loss": 0.3631, "step": 20808 }, { "epoch": 2.48, "learning_rate": 3.8637751871897834e-07, "loss": 0.3528, "step": 20809 }, { "epoch": 2.48, "learning_rate": 3.862058780174793e-07, "loss": 0.4437, "step": 20810 }, { "epoch": 2.48, "learning_rate": 3.8603427225717557e-07, "loss": 0.4768, "step": 20811 }, { "epoch": 2.48, "learning_rate": 3.8586270144090166e-07, "loss": 0.6358, "step": 20812 }, { "epoch": 2.48, "learning_rate": 3.856911655714954e-07, "loss": 0.4136, "step": 20813 }, { "epoch": 2.48, "learning_rate": 3.8551966465179087e-07, "loss": 0.3436, "step": 20814 }, { "epoch": 2.48, "learning_rate": 3.853481986846239e-07, "loss": 0.4028, "step": 20815 }, { "epoch": 2.48, "learning_rate": 3.8517676767282854e-07, "loss": 0.4583, "step": 20816 }, { "epoch": 2.48, "learning_rate": 3.8500537161923923e-07, "loss": 0.5761, "step": 20817 }, { "epoch": 2.48, "learning_rate": 3.848340105266882e-07, "loss": 0.5622, "step": 20818 }, { "epoch": 2.48, "learning_rate": 3.8466268439800866e-07, "loss": 0.5406, "step": 20819 }, { "epoch": 2.48, "learning_rate": 3.844913932360328e-07, "loss": 0.3456, "step": 20820 }, { "epoch": 2.48, "learning_rate": 3.8432013704359087e-07, "loss": 0.4765, "step": 20821 }, { "epoch": 2.48, "learning_rate": 3.841489158235148e-07, "loss": 0.4983, "step": 20822 }, { "epoch": 2.48, "learning_rate": 3.8397772957863457e-07, "loss": 0.4779, "step": 20823 }, { "epoch": 2.48, "learning_rate": 3.8380657831178057e-07, "loss": 0.5607, "step": 20824 }, { "epoch": 2.48, "learning_rate": 3.8363546202578095e-07, "loss": 0.5222, "step": 20825 }, { "epoch": 2.48, "learning_rate": 3.834643807234653e-07, "loss": 0.2809, "step": 20826 }, { "epoch": 2.48, "learning_rate": 3.832933344076603e-07, "loss": 0.3716, "step": 20827 }, { "epoch": 2.48, "learning_rate": 3.831223230811948e-07, "loss": 0.5143, "step": 20828 }, { "epoch": 2.48, "learning_rate": 3.8295134674689426e-07, "loss": 0.5032, "step": 20829 }, { "epoch": 2.48, "learning_rate": 3.8278040540758567e-07, "loss": 0.5467, "step": 20830 }, { "epoch": 2.48, "learning_rate": 3.8260949906609445e-07, "loss": 0.5511, "step": 20831 }, { "epoch": 2.48, "learning_rate": 3.8243862772524647e-07, "loss": 0.4499, "step": 20832 }, { "epoch": 2.48, "learning_rate": 3.8226779138786584e-07, "loss": 0.3729, "step": 20833 }, { "epoch": 2.48, "learning_rate": 3.8209699005677556e-07, "loss": 0.5637, "step": 20834 }, { "epoch": 2.48, "learning_rate": 3.8192622373480065e-07, "loss": 0.4529, "step": 20835 }, { "epoch": 2.48, "learning_rate": 3.8175549242476215e-07, "loss": 0.612, "step": 20836 }, { "epoch": 2.48, "learning_rate": 3.81584796129483e-07, "loss": 0.6409, "step": 20837 }, { "epoch": 2.48, "learning_rate": 3.8141413485178495e-07, "loss": 0.4051, "step": 20838 }, { "epoch": 2.48, "learning_rate": 3.812435085944896e-07, "loss": 0.2724, "step": 20839 }, { "epoch": 2.48, "learning_rate": 3.810729173604169e-07, "loss": 0.349, "step": 20840 }, { "epoch": 2.48, "learning_rate": 3.809023611523863e-07, "loss": 0.4202, "step": 20841 }, { "epoch": 2.48, "learning_rate": 3.8073183997321714e-07, "loss": 0.5479, "step": 20842 }, { "epoch": 2.48, "learning_rate": 3.805613538257291e-07, "loss": 0.5751, "step": 20843 }, { "epoch": 2.48, "learning_rate": 3.803909027127395e-07, "loss": 0.4764, "step": 20844 }, { "epoch": 2.48, "learning_rate": 3.8022048663706567e-07, "loss": 0.314, "step": 20845 }, { "epoch": 2.48, "learning_rate": 3.8005010560152545e-07, "loss": 0.4622, "step": 20846 }, { "epoch": 2.48, "learning_rate": 3.798797596089351e-07, "loss": 0.4348, "step": 20847 }, { "epoch": 2.48, "learning_rate": 3.797094486621103e-07, "loss": 0.4313, "step": 20848 }, { "epoch": 2.48, "learning_rate": 3.795391727638659e-07, "loss": 0.6307, "step": 20849 }, { "epoch": 2.48, "learning_rate": 3.793689319170166e-07, "loss": 0.6227, "step": 20850 }, { "epoch": 2.48, "learning_rate": 3.7919872612437736e-07, "loss": 0.2768, "step": 20851 }, { "epoch": 2.48, "learning_rate": 3.790285553887607e-07, "loss": 0.4198, "step": 20852 }, { "epoch": 2.48, "learning_rate": 3.7885841971297995e-07, "loss": 0.382, "step": 20853 }, { "epoch": 2.48, "learning_rate": 3.7868831909984803e-07, "loss": 0.4862, "step": 20854 }, { "epoch": 2.48, "learning_rate": 3.7851825355217617e-07, "loss": 0.5906, "step": 20855 }, { "epoch": 2.48, "learning_rate": 3.783482230727753e-07, "loss": 0.5335, "step": 20856 }, { "epoch": 2.48, "learning_rate": 3.7817822766445604e-07, "loss": 0.3224, "step": 20857 }, { "epoch": 2.48, "learning_rate": 3.7800826733002923e-07, "loss": 0.3777, "step": 20858 }, { "epoch": 2.48, "learning_rate": 3.7783834207230334e-07, "loss": 0.4527, "step": 20859 }, { "epoch": 2.48, "learning_rate": 3.7766845189408765e-07, "loss": 0.5297, "step": 20860 }, { "epoch": 2.48, "learning_rate": 3.7749859679819115e-07, "loss": 0.5706, "step": 20861 }, { "epoch": 2.48, "learning_rate": 3.773287767874212e-07, "loss": 0.5734, "step": 20862 }, { "epoch": 2.48, "learning_rate": 3.7715899186458375e-07, "loss": 0.3314, "step": 20863 }, { "epoch": 2.48, "learning_rate": 3.769892420324864e-07, "loss": 0.3354, "step": 20864 }, { "epoch": 2.48, "learning_rate": 3.7681952729393514e-07, "loss": 0.5204, "step": 20865 }, { "epoch": 2.48, "learning_rate": 3.7664984765173553e-07, "loss": 0.3778, "step": 20866 }, { "epoch": 2.49, "learning_rate": 3.7648020310869173e-07, "loss": 0.4765, "step": 20867 }, { "epoch": 2.49, "learning_rate": 3.7631059366760897e-07, "loss": 0.6656, "step": 20868 }, { "epoch": 2.49, "learning_rate": 3.7614101933129007e-07, "loss": 0.4698, "step": 20869 }, { "epoch": 2.49, "learning_rate": 3.7597148010253796e-07, "loss": 0.3623, "step": 20870 }, { "epoch": 2.49, "learning_rate": 3.7580197598415523e-07, "loss": 0.4106, "step": 20871 }, { "epoch": 2.49, "learning_rate": 3.756325069789443e-07, "loss": 0.4847, "step": 20872 }, { "epoch": 2.49, "learning_rate": 3.7546307308970686e-07, "loss": 0.4878, "step": 20873 }, { "epoch": 2.49, "learning_rate": 3.752936743192426e-07, "loss": 0.6425, "step": 20874 }, { "epoch": 2.49, "learning_rate": 3.75124310670352e-07, "loss": 0.5526, "step": 20875 }, { "epoch": 2.49, "learning_rate": 3.749549821458354e-07, "loss": 0.4337, "step": 20876 }, { "epoch": 2.49, "learning_rate": 3.747856887484913e-07, "loss": 0.4469, "step": 20877 }, { "epoch": 2.49, "learning_rate": 3.7461643048111754e-07, "loss": 0.4352, "step": 20878 }, { "epoch": 2.49, "learning_rate": 3.7444720734651244e-07, "loss": 0.5211, "step": 20879 }, { "epoch": 2.49, "learning_rate": 3.742780193474735e-07, "loss": 0.589, "step": 20880 }, { "epoch": 2.49, "learning_rate": 3.7410886648679785e-07, "loss": 0.6741, "step": 20881 }, { "epoch": 2.49, "learning_rate": 3.7393974876728025e-07, "loss": 0.3271, "step": 20882 }, { "epoch": 2.49, "learning_rate": 3.737706661917179e-07, "loss": 0.3875, "step": 20883 }, { "epoch": 2.49, "learning_rate": 3.7360161876290437e-07, "loss": 0.3474, "step": 20884 }, { "epoch": 2.49, "learning_rate": 3.734326064836341e-07, "loss": 0.4464, "step": 20885 }, { "epoch": 2.49, "learning_rate": 3.732636293567016e-07, "loss": 0.5173, "step": 20886 }, { "epoch": 2.49, "learning_rate": 3.7309468738489926e-07, "loss": 0.6288, "step": 20887 }, { "epoch": 2.49, "learning_rate": 3.7292578057102126e-07, "loss": 0.3248, "step": 20888 }, { "epoch": 2.49, "learning_rate": 3.7275690891785757e-07, "loss": 0.3606, "step": 20889 }, { "epoch": 2.49, "learning_rate": 3.725880724282016e-07, "loss": 0.374, "step": 20890 }, { "epoch": 2.49, "learning_rate": 3.7241927110484245e-07, "loss": 0.4726, "step": 20891 }, { "epoch": 2.49, "learning_rate": 3.72250504950572e-07, "loss": 0.4793, "step": 20892 }, { "epoch": 2.49, "learning_rate": 3.720817739681787e-07, "loss": 0.5998, "step": 20893 }, { "epoch": 2.49, "learning_rate": 3.7191307816045186e-07, "loss": 0.6222, "step": 20894 }, { "epoch": 2.49, "learning_rate": 3.717444175301807e-07, "loss": 0.2871, "step": 20895 }, { "epoch": 2.49, "learning_rate": 3.7157579208015354e-07, "loss": 0.5589, "step": 20896 }, { "epoch": 2.49, "learning_rate": 3.7140720181315706e-07, "loss": 0.4756, "step": 20897 }, { "epoch": 2.49, "learning_rate": 3.712386467319776e-07, "loss": 0.4591, "step": 20898 }, { "epoch": 2.49, "learning_rate": 3.710701268394024e-07, "loss": 0.6343, "step": 20899 }, { "epoch": 2.49, "learning_rate": 3.7090164213821613e-07, "loss": 0.5384, "step": 20900 }, { "epoch": 2.49, "learning_rate": 3.707331926312041e-07, "loss": 0.3684, "step": 20901 }, { "epoch": 2.49, "learning_rate": 3.7056477832115135e-07, "loss": 0.3785, "step": 20902 }, { "epoch": 2.49, "learning_rate": 3.7039639921084195e-07, "loss": 0.4894, "step": 20903 }, { "epoch": 2.49, "learning_rate": 3.702280553030582e-07, "loss": 0.4702, "step": 20904 }, { "epoch": 2.49, "learning_rate": 3.7005974660058394e-07, "loss": 0.5518, "step": 20905 }, { "epoch": 2.49, "learning_rate": 3.6989147310620034e-07, "loss": 0.6118, "step": 20906 }, { "epoch": 2.49, "learning_rate": 3.6972323482269005e-07, "loss": 0.4178, "step": 20907 }, { "epoch": 2.49, "learning_rate": 3.695550317528326e-07, "loss": 0.3997, "step": 20908 }, { "epoch": 2.49, "learning_rate": 3.6938686389940944e-07, "loss": 0.3668, "step": 20909 }, { "epoch": 2.49, "learning_rate": 3.6921873126520016e-07, "loss": 0.4833, "step": 20910 }, { "epoch": 2.49, "learning_rate": 3.690506338529845e-07, "loss": 0.5907, "step": 20911 }, { "epoch": 2.49, "learning_rate": 3.6888257166554085e-07, "loss": 0.5909, "step": 20912 }, { "epoch": 2.49, "learning_rate": 3.6871454470564646e-07, "loss": 0.3498, "step": 20913 }, { "epoch": 2.49, "learning_rate": 3.6854655297607925e-07, "loss": 0.3805, "step": 20914 }, { "epoch": 2.49, "learning_rate": 3.6837859647961717e-07, "loss": 0.3827, "step": 20915 }, { "epoch": 2.49, "learning_rate": 3.682106752190351e-07, "loss": 0.3706, "step": 20916 }, { "epoch": 2.49, "learning_rate": 3.6804278919710945e-07, "loss": 0.584, "step": 20917 }, { "epoch": 2.49, "learning_rate": 3.678749384166161e-07, "loss": 0.6306, "step": 20918 }, { "epoch": 2.49, "learning_rate": 3.6770712288032864e-07, "loss": 0.5847, "step": 20919 }, { "epoch": 2.49, "learning_rate": 3.675393425910209e-07, "loss": 0.3415, "step": 20920 }, { "epoch": 2.49, "learning_rate": 3.673715975514669e-07, "loss": 0.3682, "step": 20921 }, { "epoch": 2.49, "learning_rate": 3.672038877644399e-07, "loss": 0.3858, "step": 20922 }, { "epoch": 2.49, "learning_rate": 3.670362132327107e-07, "loss": 0.4704, "step": 20923 }, { "epoch": 2.49, "learning_rate": 3.6686857395905217e-07, "loss": 0.6117, "step": 20924 }, { "epoch": 2.49, "learning_rate": 3.667009699462357e-07, "loss": 0.6773, "step": 20925 }, { "epoch": 2.49, "learning_rate": 3.665334011970309e-07, "loss": 0.2767, "step": 20926 }, { "epoch": 2.49, "learning_rate": 3.663658677142076e-07, "loss": 0.4007, "step": 20927 }, { "epoch": 2.49, "learning_rate": 3.6619836950053553e-07, "loss": 0.3827, "step": 20928 }, { "epoch": 2.49, "learning_rate": 3.660309065587836e-07, "loss": 0.5457, "step": 20929 }, { "epoch": 2.49, "learning_rate": 3.6586347889172006e-07, "loss": 0.6235, "step": 20930 }, { "epoch": 2.49, "learning_rate": 3.6569608650211196e-07, "loss": 0.7363, "step": 20931 }, { "epoch": 2.49, "learning_rate": 3.655287293927268e-07, "loss": 0.4016, "step": 20932 }, { "epoch": 2.49, "learning_rate": 3.6536140756633116e-07, "loss": 0.357, "step": 20933 }, { "epoch": 2.49, "learning_rate": 3.651941210256904e-07, "loss": 0.4454, "step": 20934 }, { "epoch": 2.49, "learning_rate": 3.6502686977356957e-07, "loss": 0.4923, "step": 20935 }, { "epoch": 2.49, "learning_rate": 3.6485965381273384e-07, "loss": 0.4818, "step": 20936 }, { "epoch": 2.49, "learning_rate": 3.646924731459475e-07, "loss": 0.6166, "step": 20937 }, { "epoch": 2.49, "learning_rate": 3.6452532777597333e-07, "loss": 0.3801, "step": 20938 }, { "epoch": 2.49, "learning_rate": 3.6435821770557464e-07, "loss": 0.3687, "step": 20939 }, { "epoch": 2.49, "learning_rate": 3.6419114293751443e-07, "loss": 0.4722, "step": 20940 }, { "epoch": 2.49, "learning_rate": 3.64024103474554e-07, "loss": 0.4973, "step": 20941 }, { "epoch": 2.49, "learning_rate": 3.6385709931945344e-07, "loss": 0.4141, "step": 20942 }, { "epoch": 2.49, "learning_rate": 3.636901304749743e-07, "loss": 0.5244, "step": 20943 }, { "epoch": 2.49, "learning_rate": 3.6352319694387683e-07, "loss": 0.592, "step": 20944 }, { "epoch": 2.49, "learning_rate": 3.6335629872892046e-07, "loss": 0.3391, "step": 20945 }, { "epoch": 2.49, "learning_rate": 3.6318943583286333e-07, "loss": 0.4498, "step": 20946 }, { "epoch": 2.49, "learning_rate": 3.630226082584648e-07, "loss": 0.5286, "step": 20947 }, { "epoch": 2.49, "learning_rate": 3.6285581600848163e-07, "loss": 0.5306, "step": 20948 }, { "epoch": 2.49, "learning_rate": 3.6268905908567077e-07, "loss": 0.5832, "step": 20949 }, { "epoch": 2.49, "learning_rate": 3.625223374927889e-07, "loss": 0.7282, "step": 20950 }, { "epoch": 2.5, "learning_rate": 3.6235565123259236e-07, "loss": 0.339, "step": 20951 }, { "epoch": 2.5, "learning_rate": 3.6218900030783683e-07, "loss": 0.3877, "step": 20952 }, { "epoch": 2.5, "learning_rate": 3.620223847212756e-07, "loss": 0.4677, "step": 20953 }, { "epoch": 2.5, "learning_rate": 3.6185580447566454e-07, "loss": 0.5095, "step": 20954 }, { "epoch": 2.5, "learning_rate": 3.616892595737559e-07, "loss": 0.6492, "step": 20955 }, { "epoch": 2.5, "learning_rate": 3.6152275001830385e-07, "loss": 0.6152, "step": 20956 }, { "epoch": 2.5, "learning_rate": 3.6135627581205953e-07, "loss": 0.264, "step": 20957 }, { "epoch": 2.5, "learning_rate": 3.611898369577754e-07, "loss": 0.4024, "step": 20958 }, { "epoch": 2.5, "learning_rate": 3.6102343345820274e-07, "loss": 0.4294, "step": 20959 }, { "epoch": 2.5, "learning_rate": 3.6085706531609254e-07, "loss": 0.3875, "step": 20960 }, { "epoch": 2.5, "learning_rate": 3.6069073253419434e-07, "loss": 0.5866, "step": 20961 }, { "epoch": 2.5, "learning_rate": 3.6052443511525815e-07, "loss": 0.6485, "step": 20962 }, { "epoch": 2.5, "learning_rate": 3.6035817306203285e-07, "loss": 0.4452, "step": 20963 }, { "epoch": 2.5, "learning_rate": 3.6019194637726574e-07, "loss": 0.3775, "step": 20964 }, { "epoch": 2.5, "learning_rate": 3.600257550637051e-07, "loss": 0.3927, "step": 20965 }, { "epoch": 2.5, "learning_rate": 3.5985959912409855e-07, "loss": 0.4775, "step": 20966 }, { "epoch": 2.5, "learning_rate": 3.5969347856119276e-07, "loss": 0.6197, "step": 20967 }, { "epoch": 2.5, "learning_rate": 3.59527393377733e-07, "loss": 0.6142, "step": 20968 }, { "epoch": 2.5, "learning_rate": 3.5936134357646565e-07, "loss": 0.7327, "step": 20969 }, { "epoch": 2.5, "learning_rate": 3.591953291601341e-07, "loss": 0.3058, "step": 20970 }, { "epoch": 2.5, "learning_rate": 3.59029350131484e-07, "loss": 0.4476, "step": 20971 }, { "epoch": 2.5, "learning_rate": 3.5886340649325777e-07, "loss": 0.3835, "step": 20972 }, { "epoch": 2.5, "learning_rate": 3.586974982481994e-07, "loss": 0.5485, "step": 20973 }, { "epoch": 2.5, "learning_rate": 3.585316253990506e-07, "loss": 0.5916, "step": 20974 }, { "epoch": 2.5, "learning_rate": 3.583657879485547e-07, "loss": 0.4849, "step": 20975 }, { "epoch": 2.5, "learning_rate": 3.581999858994517e-07, "loss": 0.3878, "step": 20976 }, { "epoch": 2.5, "learning_rate": 3.5803421925448225e-07, "loss": 0.4071, "step": 20977 }, { "epoch": 2.5, "learning_rate": 3.5786848801638684e-07, "loss": 0.4298, "step": 20978 }, { "epoch": 2.5, "learning_rate": 3.577027921879059e-07, "loss": 0.5422, "step": 20979 }, { "epoch": 2.5, "learning_rate": 3.575371317717766e-07, "loss": 0.6896, "step": 20980 }, { "epoch": 2.5, "learning_rate": 3.573715067707384e-07, "loss": 0.5989, "step": 20981 }, { "epoch": 2.5, "learning_rate": 3.5720591718752946e-07, "loss": 0.336, "step": 20982 }, { "epoch": 2.5, "learning_rate": 3.570403630248867e-07, "loss": 0.4187, "step": 20983 }, { "epoch": 2.5, "learning_rate": 3.56874844285546e-07, "loss": 0.4798, "step": 20984 }, { "epoch": 2.5, "learning_rate": 3.567093609722438e-07, "loss": 0.4609, "step": 20985 }, { "epoch": 2.5, "learning_rate": 3.565439130877163e-07, "loss": 0.5376, "step": 20986 }, { "epoch": 2.5, "learning_rate": 3.56378500634697e-07, "loss": 0.6293, "step": 20987 }, { "epoch": 2.5, "learning_rate": 3.5621312361592137e-07, "loss": 0.4498, "step": 20988 }, { "epoch": 2.5, "learning_rate": 3.5604778203412237e-07, "loss": 0.4072, "step": 20989 }, { "epoch": 2.5, "learning_rate": 3.55882475892034e-07, "loss": 0.3974, "step": 20990 }, { "epoch": 2.5, "learning_rate": 3.557172051923885e-07, "loss": 0.4163, "step": 20991 }, { "epoch": 2.5, "learning_rate": 3.5555196993791675e-07, "loss": 0.4645, "step": 20992 }, { "epoch": 2.5, "learning_rate": 3.553867701313507e-07, "loss": 0.6125, "step": 20993 }, { "epoch": 2.5, "learning_rate": 3.552216057754221e-07, "loss": 0.5154, "step": 20994 }, { "epoch": 2.5, "learning_rate": 3.5505647687285954e-07, "loss": 0.3388, "step": 20995 }, { "epoch": 2.5, "learning_rate": 3.548913834263937e-07, "loss": 0.4758, "step": 20996 }, { "epoch": 2.5, "learning_rate": 3.547263254387537e-07, "loss": 0.42, "step": 20997 }, { "epoch": 2.5, "learning_rate": 3.545613029126674e-07, "loss": 0.45, "step": 20998 }, { "epoch": 2.5, "learning_rate": 3.543963158508623e-07, "loss": 0.6604, "step": 20999 }, { "epoch": 2.5, "learning_rate": 3.542313642560663e-07, "loss": 0.5009, "step": 21000 }, { "epoch": 2.5, "learning_rate": 3.540664481310063e-07, "loss": 0.3217, "step": 21001 }, { "epoch": 2.5, "learning_rate": 3.539015674784074e-07, "loss": 0.3631, "step": 21002 }, { "epoch": 2.5, "learning_rate": 3.5373672230099576e-07, "loss": 0.3384, "step": 21003 }, { "epoch": 2.5, "learning_rate": 3.5357191260149663e-07, "loss": 0.5319, "step": 21004 }, { "epoch": 2.5, "learning_rate": 3.5340713838263404e-07, "loss": 0.5752, "step": 21005 }, { "epoch": 2.5, "learning_rate": 3.532423996471307e-07, "loss": 0.4831, "step": 21006 }, { "epoch": 2.5, "learning_rate": 3.530776963977109e-07, "loss": 0.3221, "step": 21007 }, { "epoch": 2.5, "learning_rate": 3.5291302863709686e-07, "loss": 0.4953, "step": 21008 }, { "epoch": 2.5, "learning_rate": 3.527483963680112e-07, "loss": 0.4911, "step": 21009 }, { "epoch": 2.5, "learning_rate": 3.525837995931741e-07, "loss": 0.5149, "step": 21010 }, { "epoch": 2.5, "learning_rate": 3.5241923831530745e-07, "loss": 0.5718, "step": 21011 }, { "epoch": 2.5, "learning_rate": 3.5225471253713116e-07, "loss": 0.5828, "step": 21012 }, { "epoch": 2.5, "learning_rate": 3.5209022226136396e-07, "loss": 0.5266, "step": 21013 }, { "epoch": 2.5, "learning_rate": 3.519257674907259e-07, "loss": 0.3582, "step": 21014 }, { "epoch": 2.5, "learning_rate": 3.5176134822793475e-07, "loss": 0.3958, "step": 21015 }, { "epoch": 2.5, "learning_rate": 3.515969644757095e-07, "loss": 0.456, "step": 21016 }, { "epoch": 2.5, "learning_rate": 3.5143261623676623e-07, "loss": 0.5006, "step": 21017 }, { "epoch": 2.5, "learning_rate": 3.51268303513822e-07, "loss": 0.5508, "step": 21018 }, { "epoch": 2.5, "learning_rate": 3.5110402630959357e-07, "loss": 0.629, "step": 21019 }, { "epoch": 2.5, "learning_rate": 3.509397846267959e-07, "loss": 0.3146, "step": 21020 }, { "epoch": 2.5, "learning_rate": 3.5077557846814327e-07, "loss": 0.4489, "step": 21021 }, { "epoch": 2.5, "learning_rate": 3.5061140783635097e-07, "loss": 0.4181, "step": 21022 }, { "epoch": 2.5, "learning_rate": 3.504472727341321e-07, "loss": 0.5578, "step": 21023 }, { "epoch": 2.5, "learning_rate": 3.502831731642009e-07, "loss": 0.5958, "step": 21024 }, { "epoch": 2.5, "learning_rate": 3.5011910912926857e-07, "loss": 0.4501, "step": 21025 }, { "epoch": 2.5, "learning_rate": 3.499550806320484e-07, "loss": 0.37, "step": 21026 }, { "epoch": 2.5, "learning_rate": 3.497910876752511e-07, "loss": 0.4511, "step": 21027 }, { "epoch": 2.5, "learning_rate": 3.4962713026158697e-07, "loss": 0.4574, "step": 21028 }, { "epoch": 2.5, "learning_rate": 3.494632083937669e-07, "loss": 0.5259, "step": 21029 }, { "epoch": 2.5, "learning_rate": 3.4929932207450027e-07, "loss": 0.5826, "step": 21030 }, { "epoch": 2.5, "learning_rate": 3.4913547130649707e-07, "loss": 0.6217, "step": 21031 }, { "epoch": 2.5, "learning_rate": 3.4897165609246413e-07, "loss": 0.4011, "step": 21032 }, { "epoch": 2.5, "learning_rate": 3.4880787643511107e-07, "loss": 0.3108, "step": 21033 }, { "epoch": 2.5, "learning_rate": 3.4864413233714366e-07, "loss": 0.4137, "step": 21034 }, { "epoch": 2.51, "learning_rate": 3.4848042380126983e-07, "loss": 0.5491, "step": 21035 }, { "epoch": 2.51, "learning_rate": 3.483167508301946e-07, "loss": 0.6673, "step": 21036 }, { "epoch": 2.51, "learning_rate": 3.481531134266239e-07, "loss": 0.6711, "step": 21037 }, { "epoch": 2.51, "learning_rate": 3.4798951159326273e-07, "loss": 0.4014, "step": 21038 }, { "epoch": 2.51, "learning_rate": 3.478259453328164e-07, "loss": 0.3577, "step": 21039 }, { "epoch": 2.51, "learning_rate": 3.476624146479876e-07, "loss": 0.4708, "step": 21040 }, { "epoch": 2.51, "learning_rate": 3.4749891954147905e-07, "loss": 0.4517, "step": 21041 }, { "epoch": 2.51, "learning_rate": 3.473354600159945e-07, "loss": 0.511, "step": 21042 }, { "epoch": 2.51, "learning_rate": 3.471720360742348e-07, "loss": 0.6472, "step": 21043 }, { "epoch": 2.51, "learning_rate": 3.470086477189022e-07, "loss": 0.6229, "step": 21044 }, { "epoch": 2.51, "learning_rate": 3.4684529495269707e-07, "loss": 0.292, "step": 21045 }, { "epoch": 2.51, "learning_rate": 3.4668197777832063e-07, "loss": 0.356, "step": 21046 }, { "epoch": 2.51, "learning_rate": 3.4651869619847124e-07, "loss": 0.3729, "step": 21047 }, { "epoch": 2.51, "learning_rate": 3.4635545021584893e-07, "loss": 0.5179, "step": 21048 }, { "epoch": 2.51, "learning_rate": 3.461922398331513e-07, "loss": 0.529, "step": 21049 }, { "epoch": 2.51, "learning_rate": 3.46029065053077e-07, "loss": 0.5472, "step": 21050 }, { "epoch": 2.51, "learning_rate": 3.4586592587832256e-07, "loss": 0.4193, "step": 21051 }, { "epoch": 2.51, "learning_rate": 3.457028223115852e-07, "loss": 0.431, "step": 21052 }, { "epoch": 2.51, "learning_rate": 3.455397543555608e-07, "loss": 0.4337, "step": 21053 }, { "epoch": 2.51, "learning_rate": 3.4537672201294585e-07, "loss": 0.4955, "step": 21054 }, { "epoch": 2.51, "learning_rate": 3.452137252864343e-07, "loss": 0.5932, "step": 21055 }, { "epoch": 2.51, "learning_rate": 3.450507641787199e-07, "loss": 0.5799, "step": 21056 }, { "epoch": 2.51, "learning_rate": 3.448878386924975e-07, "loss": 0.4231, "step": 21057 }, { "epoch": 2.51, "learning_rate": 3.4472494883046046e-07, "loss": 0.4158, "step": 21058 }, { "epoch": 2.51, "learning_rate": 3.4456209459530014e-07, "loss": 0.5559, "step": 21059 }, { "epoch": 2.51, "learning_rate": 3.443992759897094e-07, "loss": 0.6014, "step": 21060 }, { "epoch": 2.51, "learning_rate": 3.4423649301638e-07, "loss": 0.4951, "step": 21061 }, { "epoch": 2.51, "learning_rate": 3.440737456780022e-07, "loss": 0.6778, "step": 21062 }, { "epoch": 2.51, "learning_rate": 3.4391103397726585e-07, "loss": 0.4013, "step": 21063 }, { "epoch": 2.51, "learning_rate": 3.4374835791686103e-07, "loss": 0.4049, "step": 21064 }, { "epoch": 2.51, "learning_rate": 3.435857174994772e-07, "loss": 0.3521, "step": 21065 }, { "epoch": 2.51, "learning_rate": 3.434231127278018e-07, "loss": 0.4984, "step": 21066 }, { "epoch": 2.51, "learning_rate": 3.432605436045233e-07, "loss": 0.4768, "step": 21067 }, { "epoch": 2.51, "learning_rate": 3.4309801013232976e-07, "loss": 0.6113, "step": 21068 }, { "epoch": 2.51, "learning_rate": 3.429355123139069e-07, "loss": 0.5539, "step": 21069 }, { "epoch": 2.51, "learning_rate": 3.427730501519405e-07, "loss": 0.2986, "step": 21070 }, { "epoch": 2.51, "learning_rate": 3.4261062364911666e-07, "loss": 0.3831, "step": 21071 }, { "epoch": 2.51, "learning_rate": 3.4244823280812025e-07, "loss": 0.4502, "step": 21072 }, { "epoch": 2.51, "learning_rate": 3.4228587763163597e-07, "loss": 0.4903, "step": 21073 }, { "epoch": 2.51, "learning_rate": 3.421235581223467e-07, "loss": 0.5699, "step": 21074 }, { "epoch": 2.51, "learning_rate": 3.419612742829362e-07, "loss": 0.6203, "step": 21075 }, { "epoch": 2.51, "learning_rate": 3.417990261160872e-07, "loss": 0.351, "step": 21076 }, { "epoch": 2.51, "learning_rate": 3.4163681362448157e-07, "loss": 0.3602, "step": 21077 }, { "epoch": 2.51, "learning_rate": 3.414746368108002e-07, "loss": 0.3674, "step": 21078 }, { "epoch": 2.51, "learning_rate": 3.413124956777239e-07, "loss": 0.4297, "step": 21079 }, { "epoch": 2.51, "learning_rate": 3.4115039022793365e-07, "loss": 0.5747, "step": 21080 }, { "epoch": 2.51, "learning_rate": 3.409883204641082e-07, "loss": 0.6487, "step": 21081 }, { "epoch": 2.51, "learning_rate": 3.4082628638892697e-07, "loss": 0.2824, "step": 21082 }, { "epoch": 2.51, "learning_rate": 3.4066428800506893e-07, "loss": 0.2803, "step": 21083 }, { "epoch": 2.51, "learning_rate": 3.405023253152112e-07, "loss": 0.4068, "step": 21084 }, { "epoch": 2.51, "learning_rate": 3.40340398322031e-07, "loss": 0.4445, "step": 21085 }, { "epoch": 2.51, "learning_rate": 3.401785070282049e-07, "loss": 0.4741, "step": 21086 }, { "epoch": 2.51, "learning_rate": 3.400166514364095e-07, "loss": 0.5126, "step": 21087 }, { "epoch": 2.51, "learning_rate": 3.398548315493208e-07, "loss": 0.5492, "step": 21088 }, { "epoch": 2.51, "learning_rate": 3.3969304736961223e-07, "loss": 0.2374, "step": 21089 }, { "epoch": 2.51, "learning_rate": 3.395312988999594e-07, "loss": 0.4381, "step": 21090 }, { "epoch": 2.51, "learning_rate": 3.3936958614303547e-07, "loss": 0.4511, "step": 21091 }, { "epoch": 2.51, "learning_rate": 3.3920790910151294e-07, "loss": 0.4488, "step": 21092 }, { "epoch": 2.51, "learning_rate": 3.390462677780651e-07, "loss": 0.6415, "step": 21093 }, { "epoch": 2.51, "learning_rate": 3.388846621753639e-07, "loss": 0.5383, "step": 21094 }, { "epoch": 2.51, "learning_rate": 3.3872309229608074e-07, "loss": 0.3246, "step": 21095 }, { "epoch": 2.51, "learning_rate": 3.385615581428861e-07, "loss": 0.2916, "step": 21096 }, { "epoch": 2.51, "learning_rate": 3.3840005971845055e-07, "loss": 0.5079, "step": 21097 }, { "epoch": 2.51, "learning_rate": 3.382385970254429e-07, "loss": 0.4558, "step": 21098 }, { "epoch": 2.51, "learning_rate": 3.380771700665331e-07, "loss": 0.5667, "step": 21099 }, { "epoch": 2.51, "learning_rate": 3.3791577884438876e-07, "loss": 0.6502, "step": 21100 }, { "epoch": 2.51, "learning_rate": 3.377544233616778e-07, "loss": 0.4763, "step": 21101 }, { "epoch": 2.51, "learning_rate": 3.3759310362106777e-07, "loss": 0.3548, "step": 21102 }, { "epoch": 2.51, "learning_rate": 3.3743181962522546e-07, "loss": 0.3348, "step": 21103 }, { "epoch": 2.51, "learning_rate": 3.3727057137681654e-07, "loss": 0.5343, "step": 21104 }, { "epoch": 2.51, "learning_rate": 3.3710935887850655e-07, "loss": 0.6392, "step": 21105 }, { "epoch": 2.51, "learning_rate": 3.3694818213296065e-07, "loss": 0.6002, "step": 21106 }, { "epoch": 2.51, "learning_rate": 3.3678704114284226e-07, "loss": 0.3805, "step": 21107 }, { "epoch": 2.51, "learning_rate": 3.366259359108157e-07, "loss": 0.3736, "step": 21108 }, { "epoch": 2.51, "learning_rate": 3.3646486643954367e-07, "loss": 0.4222, "step": 21109 }, { "epoch": 2.51, "learning_rate": 3.363038327316895e-07, "loss": 0.4846, "step": 21110 }, { "epoch": 2.51, "learning_rate": 3.361428347899143e-07, "loss": 0.503, "step": 21111 }, { "epoch": 2.51, "learning_rate": 3.3598187261687984e-07, "loss": 0.6252, "step": 21112 }, { "epoch": 2.51, "learning_rate": 3.358209462152462e-07, "loss": 0.4117, "step": 21113 }, { "epoch": 2.51, "learning_rate": 3.3566005558767433e-07, "loss": 0.3636, "step": 21114 }, { "epoch": 2.51, "learning_rate": 3.354992007368227e-07, "loss": 0.3311, "step": 21115 }, { "epoch": 2.51, "learning_rate": 3.3533838166535104e-07, "loss": 0.5099, "step": 21116 }, { "epoch": 2.51, "learning_rate": 3.3517759837591786e-07, "loss": 0.4517, "step": 21117 }, { "epoch": 2.51, "learning_rate": 3.350168508711807e-07, "loss": 0.5166, "step": 21118 }, { "epoch": 2.52, "learning_rate": 3.348561391537969e-07, "loss": 0.5294, "step": 21119 }, { "epoch": 2.52, "learning_rate": 3.346954632264221e-07, "loss": 0.3703, "step": 21120 }, { "epoch": 2.52, "learning_rate": 3.3453482309171307e-07, "loss": 0.3112, "step": 21121 }, { "epoch": 2.52, "learning_rate": 3.3437421875232577e-07, "loss": 0.4602, "step": 21122 }, { "epoch": 2.52, "learning_rate": 3.342136502109136e-07, "loss": 0.4326, "step": 21123 }, { "epoch": 2.52, "learning_rate": 3.340531174701317e-07, "loss": 0.5742, "step": 21124 }, { "epoch": 2.52, "learning_rate": 3.33892620532634e-07, "loss": 0.5461, "step": 21125 }, { "epoch": 2.52, "learning_rate": 3.3373215940107286e-07, "loss": 0.325, "step": 21126 }, { "epoch": 2.52, "learning_rate": 3.3357173407810055e-07, "loss": 0.3846, "step": 21127 }, { "epoch": 2.52, "learning_rate": 3.334113445663692e-07, "loss": 0.4652, "step": 21128 }, { "epoch": 2.52, "learning_rate": 3.332509908685305e-07, "loss": 0.4196, "step": 21129 }, { "epoch": 2.52, "learning_rate": 3.3309067298723435e-07, "loss": 0.6006, "step": 21130 }, { "epoch": 2.52, "learning_rate": 3.3293039092513107e-07, "loss": 0.6294, "step": 21131 }, { "epoch": 2.52, "learning_rate": 3.3277014468487053e-07, "loss": 0.1906, "step": 21132 }, { "epoch": 2.52, "learning_rate": 3.3260993426910166e-07, "loss": 0.3504, "step": 21133 }, { "epoch": 2.52, "learning_rate": 3.324497596804724e-07, "loss": 0.4312, "step": 21134 }, { "epoch": 2.52, "learning_rate": 3.3228962092163004e-07, "loss": 0.3994, "step": 21135 }, { "epoch": 2.52, "learning_rate": 3.321295179952222e-07, "loss": 0.5334, "step": 21136 }, { "epoch": 2.52, "learning_rate": 3.3196945090389564e-07, "loss": 0.5218, "step": 21137 }, { "epoch": 2.52, "learning_rate": 3.3180941965029553e-07, "loss": 0.2953, "step": 21138 }, { "epoch": 2.52, "learning_rate": 3.316494242370677e-07, "loss": 0.3511, "step": 21139 }, { "epoch": 2.52, "learning_rate": 3.314894646668573e-07, "loss": 0.4908, "step": 21140 }, { "epoch": 2.52, "learning_rate": 3.3132954094230813e-07, "loss": 0.5211, "step": 21141 }, { "epoch": 2.52, "learning_rate": 3.31169653066063e-07, "loss": 0.4471, "step": 21142 }, { "epoch": 2.52, "learning_rate": 3.310098010407656e-07, "loss": 0.6772, "step": 21143 }, { "epoch": 2.52, "learning_rate": 3.308499848690586e-07, "loss": 0.5833, "step": 21144 }, { "epoch": 2.52, "learning_rate": 3.306902045535829e-07, "loss": 0.4795, "step": 21145 }, { "epoch": 2.52, "learning_rate": 3.3053046009698e-07, "loss": 0.3088, "step": 21146 }, { "epoch": 2.52, "learning_rate": 3.3037075150189115e-07, "loss": 0.5171, "step": 21147 }, { "epoch": 2.52, "learning_rate": 3.3021107877095587e-07, "loss": 0.4362, "step": 21148 }, { "epoch": 2.52, "learning_rate": 3.3005144190681294e-07, "loss": 0.602, "step": 21149 }, { "epoch": 2.52, "learning_rate": 3.2989184091210186e-07, "loss": 0.5287, "step": 21150 }, { "epoch": 2.52, "learning_rate": 3.2973227578946083e-07, "loss": 0.295, "step": 21151 }, { "epoch": 2.52, "learning_rate": 3.295727465415277e-07, "loss": 0.4142, "step": 21152 }, { "epoch": 2.52, "learning_rate": 3.2941325317093874e-07, "loss": 0.4823, "step": 21153 }, { "epoch": 2.52, "learning_rate": 3.2925379568033127e-07, "loss": 0.4439, "step": 21154 }, { "epoch": 2.52, "learning_rate": 3.29094374072341e-07, "loss": 0.4806, "step": 21155 }, { "epoch": 2.52, "learning_rate": 3.2893498834960206e-07, "loss": 0.5802, "step": 21156 }, { "epoch": 2.52, "learning_rate": 3.2877563851475004e-07, "loss": 0.3764, "step": 21157 }, { "epoch": 2.52, "learning_rate": 3.286163245704188e-07, "loss": 0.3526, "step": 21158 }, { "epoch": 2.52, "learning_rate": 3.284570465192424e-07, "loss": 0.5011, "step": 21159 }, { "epoch": 2.52, "learning_rate": 3.282978043638529e-07, "loss": 0.4276, "step": 21160 }, { "epoch": 2.52, "learning_rate": 3.281385981068827e-07, "loss": 0.4917, "step": 21161 }, { "epoch": 2.52, "learning_rate": 3.279794277509643e-07, "loss": 0.5264, "step": 21162 }, { "epoch": 2.52, "learning_rate": 3.2782029329872844e-07, "loss": 0.3398, "step": 21163 }, { "epoch": 2.52, "learning_rate": 3.2766119475280466e-07, "loss": 0.2883, "step": 21164 }, { "epoch": 2.52, "learning_rate": 3.2750213211582365e-07, "loss": 0.4041, "step": 21165 }, { "epoch": 2.52, "learning_rate": 3.273431053904147e-07, "loss": 0.5134, "step": 21166 }, { "epoch": 2.52, "learning_rate": 3.271841145792068e-07, "loss": 0.5785, "step": 21167 }, { "epoch": 2.52, "learning_rate": 3.2702515968482757e-07, "loss": 0.6334, "step": 21168 }, { "epoch": 2.52, "learning_rate": 3.268662407099052e-07, "loss": 0.5295, "step": 21169 }, { "epoch": 2.52, "learning_rate": 3.267073576570662e-07, "loss": 0.297, "step": 21170 }, { "epoch": 2.52, "learning_rate": 3.265485105289365e-07, "loss": 0.4594, "step": 21171 }, { "epoch": 2.52, "learning_rate": 3.2638969932814207e-07, "loss": 0.432, "step": 21172 }, { "epoch": 2.52, "learning_rate": 3.262309240573086e-07, "loss": 0.5312, "step": 21173 }, { "epoch": 2.52, "learning_rate": 3.2607218471906066e-07, "loss": 0.5566, "step": 21174 }, { "epoch": 2.52, "learning_rate": 3.259134813160214e-07, "loss": 0.554, "step": 21175 }, { "epoch": 2.52, "learning_rate": 3.2575481385081543e-07, "loss": 0.2897, "step": 21176 }, { "epoch": 2.52, "learning_rate": 3.2559618232606393e-07, "loss": 0.4201, "step": 21177 }, { "epoch": 2.52, "learning_rate": 3.25437586744391e-07, "loss": 0.6564, "step": 21178 }, { "epoch": 2.52, "learning_rate": 3.252790271084164e-07, "loss": 0.4498, "step": 21179 }, { "epoch": 2.52, "learning_rate": 3.251205034207619e-07, "loss": 0.6187, "step": 21180 }, { "epoch": 2.52, "learning_rate": 3.24962015684048e-07, "loss": 0.583, "step": 21181 }, { "epoch": 2.52, "learning_rate": 3.248035639008951e-07, "loss": 0.3041, "step": 21182 }, { "epoch": 2.52, "learning_rate": 3.2464514807392185e-07, "loss": 0.3888, "step": 21183 }, { "epoch": 2.52, "learning_rate": 3.244867682057462e-07, "loss": 0.4097, "step": 21184 }, { "epoch": 2.52, "learning_rate": 3.243284242989872e-07, "loss": 0.4271, "step": 21185 }, { "epoch": 2.52, "learning_rate": 3.241701163562616e-07, "loss": 0.5242, "step": 21186 }, { "epoch": 2.52, "learning_rate": 3.240118443801865e-07, "loss": 0.637, "step": 21187 }, { "epoch": 2.52, "learning_rate": 3.2385360837337815e-07, "loss": 0.3796, "step": 21188 }, { "epoch": 2.52, "learning_rate": 3.2369540833845275e-07, "loss": 0.3461, "step": 21189 }, { "epoch": 2.52, "learning_rate": 3.2353724427802433e-07, "loss": 0.4452, "step": 21190 }, { "epoch": 2.52, "learning_rate": 3.233791161947086e-07, "loss": 0.506, "step": 21191 }, { "epoch": 2.52, "learning_rate": 3.2322102409111794e-07, "loss": 0.5987, "step": 21192 }, { "epoch": 2.52, "learning_rate": 3.230629679698671e-07, "loss": 0.5806, "step": 21193 }, { "epoch": 2.52, "learning_rate": 3.2290494783356746e-07, "loss": 0.6484, "step": 21194 }, { "epoch": 2.52, "learning_rate": 3.2274696368483153e-07, "loss": 0.2918, "step": 21195 }, { "epoch": 2.52, "learning_rate": 3.225890155262712e-07, "loss": 0.3671, "step": 21196 }, { "epoch": 2.52, "learning_rate": 3.2243110336049763e-07, "loss": 0.3735, "step": 21197 }, { "epoch": 2.52, "learning_rate": 3.222732271901205e-07, "loss": 0.5214, "step": 21198 }, { "epoch": 2.52, "learning_rate": 3.221153870177493e-07, "loss": 0.5881, "step": 21199 }, { "epoch": 2.52, "learning_rate": 3.219575828459934e-07, "loss": 0.6449, "step": 21200 }, { "epoch": 2.52, "learning_rate": 3.2179981467746203e-07, "loss": 0.366, "step": 21201 }, { "epoch": 2.52, "learning_rate": 3.2164208251476203e-07, "loss": 0.3529, "step": 21202 }, { "epoch": 2.53, "learning_rate": 3.2148438636050105e-07, "loss": 0.472, "step": 21203 }, { "epoch": 2.53, "learning_rate": 3.2132672621728646e-07, "loss": 0.4509, "step": 21204 }, { "epoch": 2.53, "learning_rate": 3.211691020877239e-07, "loss": 0.5602, "step": 21205 }, { "epoch": 2.53, "learning_rate": 3.2101151397441826e-07, "loss": 0.6921, "step": 21206 }, { "epoch": 2.53, "learning_rate": 3.2085396187997526e-07, "loss": 0.3868, "step": 21207 }, { "epoch": 2.53, "learning_rate": 3.206964458069997e-07, "loss": 0.2955, "step": 21208 }, { "epoch": 2.53, "learning_rate": 3.2053896575809426e-07, "loss": 0.3818, "step": 21209 }, { "epoch": 2.53, "learning_rate": 3.2038152173586244e-07, "loss": 0.4858, "step": 21210 }, { "epoch": 2.53, "learning_rate": 3.202241137429074e-07, "loss": 0.5444, "step": 21211 }, { "epoch": 2.53, "learning_rate": 3.2006674178183096e-07, "loss": 0.5881, "step": 21212 }, { "epoch": 2.53, "learning_rate": 3.199094058552335e-07, "loss": 0.3006, "step": 21213 }, { "epoch": 2.53, "learning_rate": 3.1975210596571635e-07, "loss": 0.3391, "step": 21214 }, { "epoch": 2.53, "learning_rate": 3.195948421158801e-07, "loss": 0.4679, "step": 21215 }, { "epoch": 2.53, "learning_rate": 3.194376143083244e-07, "loss": 0.4731, "step": 21216 }, { "epoch": 2.53, "learning_rate": 3.192804225456475e-07, "loss": 0.6375, "step": 21217 }, { "epoch": 2.53, "learning_rate": 3.191232668304481e-07, "loss": 0.6137, "step": 21218 }, { "epoch": 2.53, "learning_rate": 3.1896614716532463e-07, "loss": 0.4007, "step": 21219 }, { "epoch": 2.53, "learning_rate": 3.1880906355287373e-07, "loss": 0.3305, "step": 21220 }, { "epoch": 2.53, "learning_rate": 3.186520159956916e-07, "loss": 0.4199, "step": 21221 }, { "epoch": 2.53, "learning_rate": 3.184950044963744e-07, "loss": 0.5039, "step": 21222 }, { "epoch": 2.53, "learning_rate": 3.183380290575186e-07, "loss": 0.4678, "step": 21223 }, { "epoch": 2.53, "learning_rate": 3.181810896817175e-07, "loss": 0.6294, "step": 21224 }, { "epoch": 2.53, "learning_rate": 3.1802418637156633e-07, "loss": 0.5887, "step": 21225 }, { "epoch": 2.53, "learning_rate": 3.1786731912965887e-07, "loss": 0.334, "step": 21226 }, { "epoch": 2.53, "learning_rate": 3.177104879585877e-07, "loss": 0.2939, "step": 21227 }, { "epoch": 2.53, "learning_rate": 3.175536928609446e-07, "loss": 0.4601, "step": 21228 }, { "epoch": 2.53, "learning_rate": 3.1739693383932237e-07, "loss": 0.4401, "step": 21229 }, { "epoch": 2.53, "learning_rate": 3.1724021089631186e-07, "loss": 0.4818, "step": 21230 }, { "epoch": 2.53, "learning_rate": 3.170835240345041e-07, "loss": 0.6069, "step": 21231 }, { "epoch": 2.53, "learning_rate": 3.169268732564887e-07, "loss": 0.308, "step": 21232 }, { "epoch": 2.53, "learning_rate": 3.167702585648555e-07, "loss": 0.4172, "step": 21233 }, { "epoch": 2.53, "learning_rate": 3.166136799621933e-07, "loss": 0.4556, "step": 21234 }, { "epoch": 2.53, "learning_rate": 3.164571374510894e-07, "loss": 0.3973, "step": 21235 }, { "epoch": 2.53, "learning_rate": 3.1630063103413236e-07, "loss": 0.547, "step": 21236 }, { "epoch": 2.53, "learning_rate": 3.161441607139093e-07, "loss": 0.5547, "step": 21237 }, { "epoch": 2.53, "learning_rate": 3.159877264930067e-07, "loss": 0.4133, "step": 21238 }, { "epoch": 2.53, "learning_rate": 3.1583132837401e-07, "loss": 0.3831, "step": 21239 }, { "epoch": 2.53, "learning_rate": 3.1567496635950517e-07, "loss": 0.4601, "step": 21240 }, { "epoch": 2.53, "learning_rate": 3.155186404520758e-07, "loss": 0.441, "step": 21241 }, { "epoch": 2.53, "learning_rate": 3.1536235065430716e-07, "loss": 0.5468, "step": 21242 }, { "epoch": 2.53, "learning_rate": 3.1520609696878176e-07, "loss": 0.6128, "step": 21243 }, { "epoch": 2.53, "learning_rate": 3.150498793980827e-07, "loss": 0.4936, "step": 21244 }, { "epoch": 2.53, "learning_rate": 3.148936979447925e-07, "loss": 0.4422, "step": 21245 }, { "epoch": 2.53, "learning_rate": 3.1473755261149345e-07, "loss": 0.3331, "step": 21246 }, { "epoch": 2.53, "learning_rate": 3.145814434007655e-07, "loss": 0.4249, "step": 21247 }, { "epoch": 2.53, "learning_rate": 3.1442537031519004e-07, "loss": 0.4787, "step": 21248 }, { "epoch": 2.53, "learning_rate": 3.142693333573468e-07, "loss": 0.5892, "step": 21249 }, { "epoch": 2.53, "learning_rate": 3.141133325298143e-07, "loss": 0.6473, "step": 21250 }, { "epoch": 2.53, "learning_rate": 3.139573678351715e-07, "loss": 0.323, "step": 21251 }, { "epoch": 2.53, "learning_rate": 3.138014392759972e-07, "loss": 0.4626, "step": 21252 }, { "epoch": 2.53, "learning_rate": 3.136455468548691e-07, "loss": 0.4221, "step": 21253 }, { "epoch": 2.53, "learning_rate": 3.1348969057436285e-07, "loss": 0.4905, "step": 21254 }, { "epoch": 2.53, "learning_rate": 3.133338704370559e-07, "loss": 0.5527, "step": 21255 }, { "epoch": 2.53, "learning_rate": 3.1317808644552334e-07, "loss": 0.5568, "step": 21256 }, { "epoch": 2.53, "learning_rate": 3.1302233860234095e-07, "loss": 0.3561, "step": 21257 }, { "epoch": 2.53, "learning_rate": 3.128666269100819e-07, "loss": 0.3875, "step": 21258 }, { "epoch": 2.53, "learning_rate": 3.127109513713214e-07, "loss": 0.4584, "step": 21259 }, { "epoch": 2.53, "learning_rate": 3.1255531198863237e-07, "loss": 0.4843, "step": 21260 }, { "epoch": 2.53, "learning_rate": 3.12399708764588e-07, "loss": 0.4792, "step": 21261 }, { "epoch": 2.53, "learning_rate": 3.122441417017602e-07, "loss": 0.5937, "step": 21262 }, { "epoch": 2.53, "learning_rate": 3.1208861080271965e-07, "loss": 0.4915, "step": 21263 }, { "epoch": 2.53, "learning_rate": 3.1193311607003786e-07, "loss": 0.3762, "step": 21264 }, { "epoch": 2.53, "learning_rate": 3.1177765750628586e-07, "loss": 0.3819, "step": 21265 }, { "epoch": 2.53, "learning_rate": 3.1162223511403217e-07, "loss": 0.4927, "step": 21266 }, { "epoch": 2.53, "learning_rate": 3.114668488958467e-07, "loss": 0.4653, "step": 21267 }, { "epoch": 2.53, "learning_rate": 3.1131149885429814e-07, "loss": 0.5981, "step": 21268 }, { "epoch": 2.53, "learning_rate": 3.111561849919545e-07, "loss": 0.5508, "step": 21269 }, { "epoch": 2.53, "learning_rate": 3.1100090731138204e-07, "loss": 0.3904, "step": 21270 }, { "epoch": 2.53, "learning_rate": 3.108456658151482e-07, "loss": 0.3432, "step": 21271 }, { "epoch": 2.53, "learning_rate": 3.1069046050581977e-07, "loss": 0.4723, "step": 21272 }, { "epoch": 2.53, "learning_rate": 3.1053529138596135e-07, "loss": 0.4862, "step": 21273 }, { "epoch": 2.53, "learning_rate": 3.1038015845813815e-07, "loss": 0.6444, "step": 21274 }, { "epoch": 2.53, "learning_rate": 3.1022506172491443e-07, "loss": 0.5852, "step": 21275 }, { "epoch": 2.53, "learning_rate": 3.1007000118885523e-07, "loss": 0.3995, "step": 21276 }, { "epoch": 2.53, "learning_rate": 3.0991497685252225e-07, "loss": 0.4693, "step": 21277 }, { "epoch": 2.53, "learning_rate": 3.0975998871847796e-07, "loss": 0.4612, "step": 21278 }, { "epoch": 2.53, "learning_rate": 3.09605036789285e-07, "loss": 0.5072, "step": 21279 }, { "epoch": 2.53, "learning_rate": 3.0945012106750495e-07, "loss": 0.5591, "step": 21280 }, { "epoch": 2.53, "learning_rate": 3.0929524155569797e-07, "loss": 0.5269, "step": 21281 }, { "epoch": 2.53, "learning_rate": 3.0914039825642423e-07, "loss": 0.4204, "step": 21282 }, { "epoch": 2.53, "learning_rate": 3.089855911722439e-07, "loss": 0.4662, "step": 21283 }, { "epoch": 2.53, "learning_rate": 3.088308203057158e-07, "loss": 0.5011, "step": 21284 }, { "epoch": 2.53, "learning_rate": 3.086760856593976e-07, "loss": 0.5502, "step": 21285 }, { "epoch": 2.53, "learning_rate": 3.085213872358475e-07, "loss": 0.625, "step": 21286 }, { "epoch": 2.54, "learning_rate": 3.083667250376232e-07, "loss": 0.5429, "step": 21287 }, { "epoch": 2.54, "learning_rate": 3.0821209906728043e-07, "loss": 0.3542, "step": 21288 }, { "epoch": 2.54, "learning_rate": 3.080575093273755e-07, "loss": 0.3597, "step": 21289 }, { "epoch": 2.54, "learning_rate": 3.0790295582046435e-07, "loss": 0.3983, "step": 21290 }, { "epoch": 2.54, "learning_rate": 3.0774843854910143e-07, "loss": 0.409, "step": 21291 }, { "epoch": 2.54, "learning_rate": 3.075939575158401e-07, "loss": 0.5643, "step": 21292 }, { "epoch": 2.54, "learning_rate": 3.074395127232346e-07, "loss": 0.5692, "step": 21293 }, { "epoch": 2.54, "learning_rate": 3.0728510417383813e-07, "loss": 0.6001, "step": 21294 }, { "epoch": 2.54, "learning_rate": 3.07130731870203e-07, "loss": 0.3421, "step": 21295 }, { "epoch": 2.54, "learning_rate": 3.0697639581488054e-07, "loss": 0.4431, "step": 21296 }, { "epoch": 2.54, "learning_rate": 3.068220960104229e-07, "loss": 0.3756, "step": 21297 }, { "epoch": 2.54, "learning_rate": 3.0666783245937996e-07, "loss": 0.4767, "step": 21298 }, { "epoch": 2.54, "learning_rate": 3.065136051643014e-07, "loss": 0.5903, "step": 21299 }, { "epoch": 2.54, "learning_rate": 3.063594141277368e-07, "loss": 0.481, "step": 21300 }, { "epoch": 2.54, "learning_rate": 3.062052593522352e-07, "loss": 0.3414, "step": 21301 }, { "epoch": 2.54, "learning_rate": 3.060511408403452e-07, "loss": 0.4393, "step": 21302 }, { "epoch": 2.54, "learning_rate": 3.058970585946136e-07, "loss": 0.411, "step": 21303 }, { "epoch": 2.54, "learning_rate": 3.0574301261758755e-07, "loss": 0.4959, "step": 21304 }, { "epoch": 2.54, "learning_rate": 3.0558900291181414e-07, "loss": 0.6171, "step": 21305 }, { "epoch": 2.54, "learning_rate": 3.054350294798389e-07, "loss": 0.6045, "step": 21306 }, { "epoch": 2.54, "learning_rate": 3.052810923242058e-07, "loss": 0.3683, "step": 21307 }, { "epoch": 2.54, "learning_rate": 3.051271914474607e-07, "loss": 0.3555, "step": 21308 }, { "epoch": 2.54, "learning_rate": 3.049733268521474e-07, "loss": 0.453, "step": 21309 }, { "epoch": 2.54, "learning_rate": 3.048194985408093e-07, "loss": 0.4706, "step": 21310 }, { "epoch": 2.54, "learning_rate": 3.046657065159889e-07, "loss": 0.5408, "step": 21311 }, { "epoch": 2.54, "learning_rate": 3.0451195078022883e-07, "loss": 0.6711, "step": 21312 }, { "epoch": 2.54, "learning_rate": 3.0435823133607075e-07, "loss": 0.3639, "step": 21313 }, { "epoch": 2.54, "learning_rate": 3.042045481860545e-07, "loss": 0.3536, "step": 21314 }, { "epoch": 2.54, "learning_rate": 3.0405090133272144e-07, "loss": 0.4351, "step": 21315 }, { "epoch": 2.54, "learning_rate": 3.038972907786114e-07, "loss": 0.4407, "step": 21316 }, { "epoch": 2.54, "learning_rate": 3.037437165262638e-07, "loss": 0.5714, "step": 21317 }, { "epoch": 2.54, "learning_rate": 3.035901785782161e-07, "loss": 0.5635, "step": 21318 }, { "epoch": 2.54, "learning_rate": 3.0343667693700784e-07, "loss": 0.6382, "step": 21319 }, { "epoch": 2.54, "learning_rate": 3.032832116051751e-07, "loss": 0.2921, "step": 21320 }, { "epoch": 2.54, "learning_rate": 3.0312978258525555e-07, "loss": 0.3752, "step": 21321 }, { "epoch": 2.54, "learning_rate": 3.0297638987978465e-07, "loss": 0.3957, "step": 21322 }, { "epoch": 2.54, "learning_rate": 3.028230334912982e-07, "loss": 0.407, "step": 21323 }, { "epoch": 2.54, "learning_rate": 3.0266971342233165e-07, "loss": 0.6079, "step": 21324 }, { "epoch": 2.54, "learning_rate": 3.025164296754193e-07, "loss": 0.549, "step": 21325 }, { "epoch": 2.54, "learning_rate": 3.023631822530951e-07, "loss": 0.3198, "step": 21326 }, { "epoch": 2.54, "learning_rate": 3.02209971157891e-07, "loss": 0.46, "step": 21327 }, { "epoch": 2.54, "learning_rate": 3.0205679639234123e-07, "loss": 0.3761, "step": 21328 }, { "epoch": 2.54, "learning_rate": 3.019036579589765e-07, "loss": 0.378, "step": 21329 }, { "epoch": 2.54, "learning_rate": 3.017505558603287e-07, "loss": 0.592, "step": 21330 }, { "epoch": 2.54, "learning_rate": 3.015974900989288e-07, "loss": 0.6394, "step": 21331 }, { "epoch": 2.54, "learning_rate": 3.0144446067730704e-07, "loss": 0.3603, "step": 21332 }, { "epoch": 2.54, "learning_rate": 3.0129146759799276e-07, "loss": 0.3557, "step": 21333 }, { "epoch": 2.54, "learning_rate": 3.011385108635151e-07, "loss": 0.4019, "step": 21334 }, { "epoch": 2.54, "learning_rate": 3.0098559047640197e-07, "loss": 0.4582, "step": 21335 }, { "epoch": 2.54, "learning_rate": 3.0083270643918224e-07, "loss": 0.5065, "step": 21336 }, { "epoch": 2.54, "learning_rate": 3.006798587543816e-07, "loss": 0.6258, "step": 21337 }, { "epoch": 2.54, "learning_rate": 3.0052704742452755e-07, "loss": 0.3846, "step": 21338 }, { "epoch": 2.54, "learning_rate": 3.0037427245214605e-07, "loss": 0.2448, "step": 21339 }, { "epoch": 2.54, "learning_rate": 3.0022153383976294e-07, "loss": 0.3625, "step": 21340 }, { "epoch": 2.54, "learning_rate": 3.0006883158990223e-07, "loss": 0.4266, "step": 21341 }, { "epoch": 2.54, "learning_rate": 2.999161657050881e-07, "loss": 0.6122, "step": 21342 }, { "epoch": 2.54, "learning_rate": 2.99763536187844e-07, "loss": 0.5846, "step": 21343 }, { "epoch": 2.54, "learning_rate": 2.9961094304069406e-07, "loss": 0.5388, "step": 21344 }, { "epoch": 2.54, "learning_rate": 2.9945838626615933e-07, "loss": 0.2597, "step": 21345 }, { "epoch": 2.54, "learning_rate": 2.993058658667619e-07, "loss": 0.3601, "step": 21346 }, { "epoch": 2.54, "learning_rate": 2.9915338184502374e-07, "loss": 0.4992, "step": 21347 }, { "epoch": 2.54, "learning_rate": 2.99000934203465e-07, "loss": 0.4765, "step": 21348 }, { "epoch": 2.54, "learning_rate": 2.988485229446048e-07, "loss": 0.6631, "step": 21349 }, { "epoch": 2.54, "learning_rate": 2.9869614807096325e-07, "loss": 0.5729, "step": 21350 }, { "epoch": 2.54, "learning_rate": 2.9854380958505987e-07, "loss": 0.2492, "step": 21351 }, { "epoch": 2.54, "learning_rate": 2.9839150748941116e-07, "loss": 0.3817, "step": 21352 }, { "epoch": 2.54, "learning_rate": 2.982392417865357e-07, "loss": 0.4472, "step": 21353 }, { "epoch": 2.54, "learning_rate": 2.9808701247895087e-07, "loss": 0.426, "step": 21354 }, { "epoch": 2.54, "learning_rate": 2.979348195691725e-07, "loss": 0.5788, "step": 21355 }, { "epoch": 2.54, "learning_rate": 2.9778266305971554e-07, "loss": 0.6726, "step": 21356 }, { "epoch": 2.54, "learning_rate": 2.9763054295309625e-07, "loss": 0.2755, "step": 21357 }, { "epoch": 2.54, "learning_rate": 2.9747845925182873e-07, "loss": 0.3426, "step": 21358 }, { "epoch": 2.54, "learning_rate": 2.9732641195842766e-07, "loss": 0.4175, "step": 21359 }, { "epoch": 2.54, "learning_rate": 2.971744010754052e-07, "loss": 0.455, "step": 21360 }, { "epoch": 2.54, "learning_rate": 2.970224266052749e-07, "loss": 0.4875, "step": 21361 }, { "epoch": 2.54, "learning_rate": 2.9687048855054914e-07, "loss": 0.4743, "step": 21362 }, { "epoch": 2.54, "learning_rate": 2.9671858691373926e-07, "loss": 0.538, "step": 21363 }, { "epoch": 2.54, "learning_rate": 2.9656672169735527e-07, "loss": 0.3686, "step": 21364 }, { "epoch": 2.54, "learning_rate": 2.964148929039082e-07, "loss": 0.4031, "step": 21365 }, { "epoch": 2.54, "learning_rate": 2.9626310053590843e-07, "loss": 0.4336, "step": 21366 }, { "epoch": 2.54, "learning_rate": 2.961113445958641e-07, "loss": 0.4976, "step": 21367 }, { "epoch": 2.54, "learning_rate": 2.9595962508628426e-07, "loss": 0.5399, "step": 21368 }, { "epoch": 2.54, "learning_rate": 2.958079420096768e-07, "loss": 0.6462, "step": 21369 }, { "epoch": 2.54, "learning_rate": 2.9565629536854964e-07, "loss": 0.2854, "step": 21370 }, { "epoch": 2.55, "learning_rate": 2.955046851654078e-07, "loss": 0.4684, "step": 21371 }, { "epoch": 2.55, "learning_rate": 2.9535311140275884e-07, "loss": 0.4522, "step": 21372 }, { "epoch": 2.55, "learning_rate": 2.952015740831077e-07, "loss": 0.5773, "step": 21373 }, { "epoch": 2.55, "learning_rate": 2.9505007320896037e-07, "loss": 0.6415, "step": 21374 }, { "epoch": 2.55, "learning_rate": 2.9489860878281954e-07, "loss": 0.6433, "step": 21375 }, { "epoch": 2.55, "learning_rate": 2.9474718080719027e-07, "loss": 0.2572, "step": 21376 }, { "epoch": 2.55, "learning_rate": 2.945957892845752e-07, "loss": 0.4377, "step": 21377 }, { "epoch": 2.55, "learning_rate": 2.9444443421747626e-07, "loss": 0.3489, "step": 21378 }, { "epoch": 2.55, "learning_rate": 2.9429311560839586e-07, "loss": 0.469, "step": 21379 }, { "epoch": 2.55, "learning_rate": 2.9414183345983533e-07, "loss": 0.5201, "step": 21380 }, { "epoch": 2.55, "learning_rate": 2.93990587774296e-07, "loss": 0.5735, "step": 21381 }, { "epoch": 2.55, "learning_rate": 2.938393785542767e-07, "loss": 0.4173, "step": 21382 }, { "epoch": 2.55, "learning_rate": 2.936882058022783e-07, "loss": 0.3023, "step": 21383 }, { "epoch": 2.55, "learning_rate": 2.9353706952079813e-07, "loss": 0.3448, "step": 21384 }, { "epoch": 2.55, "learning_rate": 2.933859697123362e-07, "loss": 0.3483, "step": 21385 }, { "epoch": 2.55, "learning_rate": 2.932349063793888e-07, "loss": 0.4668, "step": 21386 }, { "epoch": 2.55, "learning_rate": 2.930838795244534e-07, "loss": 0.5967, "step": 21387 }, { "epoch": 2.55, "learning_rate": 2.929328891500266e-07, "loss": 0.3345, "step": 21388 }, { "epoch": 2.55, "learning_rate": 2.92781935258605e-07, "loss": 0.2457, "step": 21389 }, { "epoch": 2.55, "learning_rate": 2.9263101785268253e-07, "loss": 0.4551, "step": 21390 }, { "epoch": 2.55, "learning_rate": 2.9248013693475517e-07, "loss": 0.464, "step": 21391 }, { "epoch": 2.55, "learning_rate": 2.923292925073165e-07, "loss": 0.4017, "step": 21392 }, { "epoch": 2.55, "learning_rate": 2.9217848457285925e-07, "loss": 0.6136, "step": 21393 }, { "epoch": 2.55, "learning_rate": 2.920277131338767e-07, "loss": 0.5437, "step": 21394 }, { "epoch": 2.55, "learning_rate": 2.9187697819286155e-07, "loss": 0.3225, "step": 21395 }, { "epoch": 2.55, "learning_rate": 2.9172627975230574e-07, "loss": 0.4754, "step": 21396 }, { "epoch": 2.55, "learning_rate": 2.9157561781469916e-07, "loss": 0.5615, "step": 21397 }, { "epoch": 2.55, "learning_rate": 2.9142499238253353e-07, "loss": 0.5052, "step": 21398 }, { "epoch": 2.55, "learning_rate": 2.9127440345829763e-07, "loss": 0.7097, "step": 21399 }, { "epoch": 2.55, "learning_rate": 2.911238510444814e-07, "loss": 0.5548, "step": 21400 }, { "epoch": 2.55, "learning_rate": 2.9097333514357314e-07, "loss": 0.3678, "step": 21401 }, { "epoch": 2.55, "learning_rate": 2.908228557580606e-07, "loss": 0.3719, "step": 21402 }, { "epoch": 2.55, "learning_rate": 2.90672412890432e-07, "loss": 0.4009, "step": 21403 }, { "epoch": 2.55, "learning_rate": 2.9052200654317407e-07, "loss": 0.4808, "step": 21404 }, { "epoch": 2.55, "learning_rate": 2.903716367187728e-07, "loss": 0.5374, "step": 21405 }, { "epoch": 2.55, "learning_rate": 2.902213034197132e-07, "loss": 0.5542, "step": 21406 }, { "epoch": 2.55, "learning_rate": 2.90071006648481e-07, "loss": 0.3271, "step": 21407 }, { "epoch": 2.55, "learning_rate": 2.8992074640756096e-07, "loss": 0.3699, "step": 21408 }, { "epoch": 2.55, "learning_rate": 2.89770522699436e-07, "loss": 0.4681, "step": 21409 }, { "epoch": 2.55, "learning_rate": 2.8962033552658976e-07, "loss": 0.5012, "step": 21410 }, { "epoch": 2.55, "learning_rate": 2.894701848915052e-07, "loss": 0.6108, "step": 21411 }, { "epoch": 2.55, "learning_rate": 2.8932007079666396e-07, "loss": 0.7517, "step": 21412 }, { "epoch": 2.55, "learning_rate": 2.8916999324454716e-07, "loss": 0.4318, "step": 21413 }, { "epoch": 2.55, "learning_rate": 2.8901995223763553e-07, "loss": 0.3305, "step": 21414 }, { "epoch": 2.55, "learning_rate": 2.8886994777841045e-07, "loss": 0.3999, "step": 21415 }, { "epoch": 2.55, "learning_rate": 2.887199798693499e-07, "loss": 0.4246, "step": 21416 }, { "epoch": 2.55, "learning_rate": 2.885700485129336e-07, "loss": 0.4954, "step": 21417 }, { "epoch": 2.55, "learning_rate": 2.884201537116402e-07, "loss": 0.5725, "step": 21418 }, { "epoch": 2.55, "learning_rate": 2.8827029546794757e-07, "loss": 0.5761, "step": 21419 }, { "epoch": 2.55, "learning_rate": 2.8812047378433264e-07, "loss": 0.2926, "step": 21420 }, { "epoch": 2.55, "learning_rate": 2.8797068866327133e-07, "loss": 0.4376, "step": 21421 }, { "epoch": 2.55, "learning_rate": 2.878209401072399e-07, "loss": 0.4033, "step": 21422 }, { "epoch": 2.55, "learning_rate": 2.876712281187147e-07, "loss": 0.42, "step": 21423 }, { "epoch": 2.55, "learning_rate": 2.875215527001693e-07, "loss": 0.7087, "step": 21424 }, { "epoch": 2.55, "learning_rate": 2.8737191385407816e-07, "loss": 0.526, "step": 21425 }, { "epoch": 2.55, "learning_rate": 2.8722231158291533e-07, "loss": 0.2446, "step": 21426 }, { "epoch": 2.55, "learning_rate": 2.870727458891534e-07, "loss": 0.3224, "step": 21427 }, { "epoch": 2.55, "learning_rate": 2.869232167752645e-07, "loss": 0.4023, "step": 21428 }, { "epoch": 2.55, "learning_rate": 2.867737242437202e-07, "loss": 0.4325, "step": 21429 }, { "epoch": 2.55, "learning_rate": 2.866242682969925e-07, "loss": 0.5228, "step": 21430 }, { "epoch": 2.55, "learning_rate": 2.864748489375507e-07, "loss": 0.6202, "step": 21431 }, { "epoch": 2.55, "learning_rate": 2.863254661678655e-07, "loss": 0.368, "step": 21432 }, { "epoch": 2.55, "learning_rate": 2.861761199904067e-07, "loss": 0.487, "step": 21433 }, { "epoch": 2.55, "learning_rate": 2.8602681040764217e-07, "loss": 0.465, "step": 21434 }, { "epoch": 2.55, "learning_rate": 2.858775374220399e-07, "loss": 0.4091, "step": 21435 }, { "epoch": 2.55, "learning_rate": 2.8572830103606786e-07, "loss": 0.5652, "step": 21436 }, { "epoch": 2.55, "learning_rate": 2.855791012521925e-07, "loss": 0.6614, "step": 21437 }, { "epoch": 2.55, "learning_rate": 2.8542993807288117e-07, "loss": 0.4382, "step": 21438 }, { "epoch": 2.55, "learning_rate": 2.852808115005981e-07, "loss": 0.363, "step": 21439 }, { "epoch": 2.55, "learning_rate": 2.8513172153780966e-07, "loss": 0.4409, "step": 21440 }, { "epoch": 2.55, "learning_rate": 2.8498266818697913e-07, "loss": 0.3649, "step": 21441 }, { "epoch": 2.55, "learning_rate": 2.848336514505712e-07, "loss": 0.4825, "step": 21442 }, { "epoch": 2.55, "learning_rate": 2.846846713310486e-07, "loss": 0.724, "step": 21443 }, { "epoch": 2.55, "learning_rate": 2.845357278308744e-07, "loss": 0.4835, "step": 21444 }, { "epoch": 2.55, "learning_rate": 2.843868209525108e-07, "loss": 0.2749, "step": 21445 }, { "epoch": 2.55, "learning_rate": 2.8423795069841857e-07, "loss": 0.3193, "step": 21446 }, { "epoch": 2.55, "learning_rate": 2.8408911707105915e-07, "loss": 0.4936, "step": 21447 }, { "epoch": 2.55, "learning_rate": 2.839403200728927e-07, "loss": 0.4647, "step": 21448 }, { "epoch": 2.55, "learning_rate": 2.8379155970637896e-07, "loss": 0.5754, "step": 21449 }, { "epoch": 2.55, "learning_rate": 2.8364283597397595e-07, "loss": 0.5517, "step": 21450 }, { "epoch": 2.55, "learning_rate": 2.8349414887814284e-07, "loss": 0.2822, "step": 21451 }, { "epoch": 2.55, "learning_rate": 2.833454984213377e-07, "loss": 0.4403, "step": 21452 }, { "epoch": 2.55, "learning_rate": 2.8319688460601763e-07, "loss": 0.4617, "step": 21453 }, { "epoch": 2.55, "learning_rate": 2.830483074346388e-07, "loss": 0.5471, "step": 21454 }, { "epoch": 2.56, "learning_rate": 2.828997669096578e-07, "loss": 0.5807, "step": 21455 }, { "epoch": 2.56, "learning_rate": 2.8275126303352995e-07, "loss": 0.617, "step": 21456 }, { "epoch": 2.56, "learning_rate": 2.826027958087091e-07, "loss": 0.3408, "step": 21457 }, { "epoch": 2.56, "learning_rate": 2.824543652376499e-07, "loss": 0.3577, "step": 21458 }, { "epoch": 2.56, "learning_rate": 2.823059713228063e-07, "loss": 0.415, "step": 21459 }, { "epoch": 2.56, "learning_rate": 2.8215761406663154e-07, "loss": 0.4299, "step": 21460 }, { "epoch": 2.56, "learning_rate": 2.8200929347157707e-07, "loss": 0.4929, "step": 21461 }, { "epoch": 2.56, "learning_rate": 2.818610095400956e-07, "loss": 0.5678, "step": 21462 }, { "epoch": 2.56, "learning_rate": 2.817127622746371e-07, "loss": 0.3684, "step": 21463 }, { "epoch": 2.56, "learning_rate": 2.8156455167765355e-07, "loss": 0.3944, "step": 21464 }, { "epoch": 2.56, "learning_rate": 2.8141637775159327e-07, "loss": 0.489, "step": 21465 }, { "epoch": 2.56, "learning_rate": 2.8126824049890675e-07, "loss": 0.3822, "step": 21466 }, { "epoch": 2.56, "learning_rate": 2.811201399220423e-07, "loss": 0.5267, "step": 21467 }, { "epoch": 2.56, "learning_rate": 2.809720760234483e-07, "loss": 0.619, "step": 21468 }, { "epoch": 2.56, "learning_rate": 2.8082404880557245e-07, "loss": 0.4804, "step": 21469 }, { "epoch": 2.56, "learning_rate": 2.8067605827086064e-07, "loss": 0.3224, "step": 21470 }, { "epoch": 2.56, "learning_rate": 2.805281044217603e-07, "loss": 0.4188, "step": 21471 }, { "epoch": 2.56, "learning_rate": 2.803801872607159e-07, "loss": 0.4144, "step": 21472 }, { "epoch": 2.56, "learning_rate": 2.8023230679017355e-07, "loss": 0.5175, "step": 21473 }, { "epoch": 2.56, "learning_rate": 2.800844630125774e-07, "loss": 0.5026, "step": 21474 }, { "epoch": 2.56, "learning_rate": 2.7993665593037177e-07, "loss": 0.5416, "step": 21475 }, { "epoch": 2.56, "learning_rate": 2.7978888554599905e-07, "loss": 0.3725, "step": 21476 }, { "epoch": 2.56, "learning_rate": 2.796411518619027e-07, "loss": 0.3947, "step": 21477 }, { "epoch": 2.56, "learning_rate": 2.794934548805242e-07, "loss": 0.3405, "step": 21478 }, { "epoch": 2.56, "learning_rate": 2.7934579460430545e-07, "loss": 0.42, "step": 21479 }, { "epoch": 2.56, "learning_rate": 2.791981710356864e-07, "loss": 0.5354, "step": 21480 }, { "epoch": 2.56, "learning_rate": 2.7905058417710824e-07, "loss": 0.592, "step": 21481 }, { "epoch": 2.56, "learning_rate": 2.7890303403101034e-07, "loss": 0.2457, "step": 21482 }, { "epoch": 2.56, "learning_rate": 2.787555205998319e-07, "loss": 0.3557, "step": 21483 }, { "epoch": 2.56, "learning_rate": 2.7860804388601115e-07, "loss": 0.4485, "step": 21484 }, { "epoch": 2.56, "learning_rate": 2.7846060389198545e-07, "loss": 0.4493, "step": 21485 }, { "epoch": 2.56, "learning_rate": 2.783132006201922e-07, "loss": 0.6022, "step": 21486 }, { "epoch": 2.56, "learning_rate": 2.7816583407306865e-07, "loss": 0.5888, "step": 21487 }, { "epoch": 2.56, "learning_rate": 2.780185042530498e-07, "loss": 0.3221, "step": 21488 }, { "epoch": 2.56, "learning_rate": 2.778712111625717e-07, "loss": 0.3106, "step": 21489 }, { "epoch": 2.56, "learning_rate": 2.777239548040692e-07, "loss": 0.3555, "step": 21490 }, { "epoch": 2.56, "learning_rate": 2.775767351799763e-07, "loss": 0.4759, "step": 21491 }, { "epoch": 2.56, "learning_rate": 2.774295522927259e-07, "loss": 0.5394, "step": 21492 }, { "epoch": 2.56, "learning_rate": 2.772824061447515e-07, "loss": 0.5622, "step": 21493 }, { "epoch": 2.56, "learning_rate": 2.7713529673848603e-07, "loss": 0.5125, "step": 21494 }, { "epoch": 2.56, "learning_rate": 2.7698822407636024e-07, "loss": 0.3966, "step": 21495 }, { "epoch": 2.56, "learning_rate": 2.768411881608052e-07, "loss": 0.4132, "step": 21496 }, { "epoch": 2.56, "learning_rate": 2.766941889942526e-07, "loss": 0.412, "step": 21497 }, { "epoch": 2.56, "learning_rate": 2.7654722657913106e-07, "loss": 0.3942, "step": 21498 }, { "epoch": 2.56, "learning_rate": 2.764003009178712e-07, "loss": 0.6114, "step": 21499 }, { "epoch": 2.56, "learning_rate": 2.7625341201290014e-07, "loss": 0.5773, "step": 21500 }, { "epoch": 2.56, "learning_rate": 2.761065598666468e-07, "loss": 0.3278, "step": 21501 }, { "epoch": 2.56, "learning_rate": 2.759597444815393e-07, "loss": 0.3697, "step": 21502 }, { "epoch": 2.56, "learning_rate": 2.7581296586000313e-07, "loss": 0.3967, "step": 21503 }, { "epoch": 2.56, "learning_rate": 2.7566622400446523e-07, "loss": 0.4456, "step": 21504 }, { "epoch": 2.56, "learning_rate": 2.7551951891735205e-07, "loss": 0.4902, "step": 21505 }, { "epoch": 2.56, "learning_rate": 2.753728506010875e-07, "loss": 0.5565, "step": 21506 }, { "epoch": 2.56, "learning_rate": 2.752262190580962e-07, "loss": 0.3254, "step": 21507 }, { "epoch": 2.56, "learning_rate": 2.7507962429080215e-07, "loss": 0.3402, "step": 21508 }, { "epoch": 2.56, "learning_rate": 2.749330663016289e-07, "loss": 0.4335, "step": 21509 }, { "epoch": 2.56, "learning_rate": 2.7478654509299845e-07, "loss": 0.4492, "step": 21510 }, { "epoch": 2.56, "learning_rate": 2.74640060667333e-07, "loss": 0.6024, "step": 21511 }, { "epoch": 2.56, "learning_rate": 2.744936130270545e-07, "loss": 0.6245, "step": 21512 }, { "epoch": 2.56, "learning_rate": 2.7434720217458353e-07, "loss": 0.3522, "step": 21513 }, { "epoch": 2.56, "learning_rate": 2.7420082811233954e-07, "loss": 0.2952, "step": 21514 }, { "epoch": 2.56, "learning_rate": 2.7405449084274254e-07, "loss": 0.4006, "step": 21515 }, { "epoch": 2.56, "learning_rate": 2.739081903682114e-07, "loss": 0.4531, "step": 21516 }, { "epoch": 2.56, "learning_rate": 2.7376192669116537e-07, "loss": 0.5396, "step": 21517 }, { "epoch": 2.56, "learning_rate": 2.7361569981402106e-07, "loss": 0.6133, "step": 21518 }, { "epoch": 2.56, "learning_rate": 2.734695097391965e-07, "loss": 0.5116, "step": 21519 }, { "epoch": 2.56, "learning_rate": 2.733233564691076e-07, "loss": 0.3465, "step": 21520 }, { "epoch": 2.56, "learning_rate": 2.731772400061702e-07, "loss": 0.4446, "step": 21521 }, { "epoch": 2.56, "learning_rate": 2.7303116035279963e-07, "loss": 0.4633, "step": 21522 }, { "epoch": 2.56, "learning_rate": 2.7288511751141083e-07, "loss": 0.5512, "step": 21523 }, { "epoch": 2.56, "learning_rate": 2.7273911148441856e-07, "loss": 0.689, "step": 21524 }, { "epoch": 2.56, "learning_rate": 2.7259314227423507e-07, "loss": 0.4868, "step": 21525 }, { "epoch": 2.56, "learning_rate": 2.724472098832745e-07, "loss": 0.2847, "step": 21526 }, { "epoch": 2.56, "learning_rate": 2.7230131431394776e-07, "loss": 0.4272, "step": 21527 }, { "epoch": 2.56, "learning_rate": 2.7215545556866785e-07, "loss": 0.4844, "step": 21528 }, { "epoch": 2.56, "learning_rate": 2.720096336498446e-07, "loss": 0.4492, "step": 21529 }, { "epoch": 2.56, "learning_rate": 2.71863848559889e-07, "loss": 0.5264, "step": 21530 }, { "epoch": 2.56, "learning_rate": 2.7171810030121094e-07, "loss": 0.5536, "step": 21531 }, { "epoch": 2.56, "learning_rate": 2.7157238887622035e-07, "loss": 0.2885, "step": 21532 }, { "epoch": 2.56, "learning_rate": 2.7142671428732446e-07, "loss": 0.3803, "step": 21533 }, { "epoch": 2.56, "learning_rate": 2.712810765369325e-07, "loss": 0.4144, "step": 21534 }, { "epoch": 2.56, "learning_rate": 2.7113547562745144e-07, "loss": 0.411, "step": 21535 }, { "epoch": 2.56, "learning_rate": 2.709899115612874e-07, "loss": 0.482, "step": 21536 }, { "epoch": 2.56, "learning_rate": 2.708443843408473e-07, "loss": 0.576, "step": 21537 }, { "epoch": 2.56, "learning_rate": 2.706988939685365e-07, "loss": 0.4305, "step": 21538 }, { "epoch": 2.57, "learning_rate": 2.705534404467605e-07, "loss": 0.3078, "step": 21539 }, { "epoch": 2.57, "learning_rate": 2.7040802377792244e-07, "loss": 0.4145, "step": 21540 }, { "epoch": 2.57, "learning_rate": 2.7026264396442763e-07, "loss": 0.4336, "step": 21541 }, { "epoch": 2.57, "learning_rate": 2.70117301008678e-07, "loss": 0.5635, "step": 21542 }, { "epoch": 2.57, "learning_rate": 2.699719949130769e-07, "loss": 0.606, "step": 21543 }, { "epoch": 2.57, "learning_rate": 2.698267256800252e-07, "loss": 0.5702, "step": 21544 }, { "epoch": 2.57, "learning_rate": 2.696814933119249e-07, "loss": 0.4164, "step": 21545 }, { "epoch": 2.57, "learning_rate": 2.6953629781117654e-07, "loss": 0.3757, "step": 21546 }, { "epoch": 2.57, "learning_rate": 2.6939113918018096e-07, "loss": 0.4074, "step": 21547 }, { "epoch": 2.57, "learning_rate": 2.692460174213371e-07, "loss": 0.5729, "step": 21548 }, { "epoch": 2.57, "learning_rate": 2.6910093253704333e-07, "loss": 0.621, "step": 21549 }, { "epoch": 2.57, "learning_rate": 2.6895588452969803e-07, "loss": 0.5997, "step": 21550 }, { "epoch": 2.57, "learning_rate": 2.6881087340169974e-07, "loss": 0.2894, "step": 21551 }, { "epoch": 2.57, "learning_rate": 2.6866589915544445e-07, "loss": 0.4122, "step": 21552 }, { "epoch": 2.57, "learning_rate": 2.6852096179332933e-07, "loss": 0.2678, "step": 21553 }, { "epoch": 2.57, "learning_rate": 2.6837606131775025e-07, "loss": 0.5247, "step": 21554 }, { "epoch": 2.57, "learning_rate": 2.6823119773110194e-07, "loss": 0.6458, "step": 21555 }, { "epoch": 2.57, "learning_rate": 2.6808637103577897e-07, "loss": 0.5497, "step": 21556 }, { "epoch": 2.57, "learning_rate": 2.679415812341754e-07, "loss": 0.419, "step": 21557 }, { "epoch": 2.57, "learning_rate": 2.6779682832868524e-07, "loss": 0.4221, "step": 21558 }, { "epoch": 2.57, "learning_rate": 2.6765211232170044e-07, "loss": 0.3879, "step": 21559 }, { "epoch": 2.57, "learning_rate": 2.6750743321561356e-07, "loss": 0.4659, "step": 21560 }, { "epoch": 2.57, "learning_rate": 2.6736279101281597e-07, "loss": 0.5297, "step": 21561 }, { "epoch": 2.57, "learning_rate": 2.672181857156994e-07, "loss": 0.5782, "step": 21562 }, { "epoch": 2.57, "learning_rate": 2.6707361732665335e-07, "loss": 0.5075, "step": 21563 }, { "epoch": 2.57, "learning_rate": 2.669290858480675e-07, "loss": 0.3742, "step": 21564 }, { "epoch": 2.57, "learning_rate": 2.667845912823311e-07, "loss": 0.3485, "step": 21565 }, { "epoch": 2.57, "learning_rate": 2.6664013363183334e-07, "loss": 0.4123, "step": 21566 }, { "epoch": 2.57, "learning_rate": 2.664957128989612e-07, "loss": 0.4971, "step": 21567 }, { "epoch": 2.57, "learning_rate": 2.663513290861022e-07, "loss": 0.6106, "step": 21568 }, { "epoch": 2.57, "learning_rate": 2.6620698219564366e-07, "loss": 0.6121, "step": 21569 }, { "epoch": 2.57, "learning_rate": 2.6606267222997113e-07, "loss": 0.282, "step": 21570 }, { "epoch": 2.57, "learning_rate": 2.6591839919146963e-07, "loss": 0.4104, "step": 21571 }, { "epoch": 2.57, "learning_rate": 2.6577416308252445e-07, "loss": 0.4286, "step": 21572 }, { "epoch": 2.57, "learning_rate": 2.6562996390552015e-07, "loss": 0.4038, "step": 21573 }, { "epoch": 2.57, "learning_rate": 2.6548580166283975e-07, "loss": 0.6326, "step": 21574 }, { "epoch": 2.57, "learning_rate": 2.653416763568664e-07, "loss": 0.5254, "step": 21575 }, { "epoch": 2.57, "learning_rate": 2.651975879899832e-07, "loss": 0.2965, "step": 21576 }, { "epoch": 2.57, "learning_rate": 2.650535365645712e-07, "loss": 0.3514, "step": 21577 }, { "epoch": 2.57, "learning_rate": 2.649095220830114e-07, "loss": 0.4183, "step": 21578 }, { "epoch": 2.57, "learning_rate": 2.647655445476849e-07, "loss": 0.472, "step": 21579 }, { "epoch": 2.57, "learning_rate": 2.646216039609714e-07, "loss": 0.5837, "step": 21580 }, { "epoch": 2.57, "learning_rate": 2.6447770032525077e-07, "loss": 0.6839, "step": 21581 }, { "epoch": 2.57, "learning_rate": 2.643338336429008e-07, "loss": 0.3392, "step": 21582 }, { "epoch": 2.57, "learning_rate": 2.641900039163009e-07, "loss": 0.4126, "step": 21583 }, { "epoch": 2.57, "learning_rate": 2.64046211147827e-07, "loss": 0.4225, "step": 21584 }, { "epoch": 2.57, "learning_rate": 2.6390245533985756e-07, "loss": 0.4854, "step": 21585 }, { "epoch": 2.57, "learning_rate": 2.6375873649476745e-07, "loss": 0.5855, "step": 21586 }, { "epoch": 2.57, "learning_rate": 2.6361505461493325e-07, "loss": 0.5572, "step": 21587 }, { "epoch": 2.57, "learning_rate": 2.634714097027305e-07, "loss": 0.3926, "step": 21588 }, { "epoch": 2.57, "learning_rate": 2.633278017605323e-07, "loss": 0.3439, "step": 21589 }, { "epoch": 2.57, "learning_rate": 2.6318423079071314e-07, "loss": 0.3788, "step": 21590 }, { "epoch": 2.57, "learning_rate": 2.63040696795647e-07, "loss": 0.4067, "step": 21591 }, { "epoch": 2.57, "learning_rate": 2.628971997777058e-07, "loss": 0.4304, "step": 21592 }, { "epoch": 2.57, "learning_rate": 2.62753739739261e-07, "loss": 0.5898, "step": 21593 }, { "epoch": 2.57, "learning_rate": 2.626103166826846e-07, "loss": 0.5714, "step": 21594 }, { "epoch": 2.57, "learning_rate": 2.624669306103475e-07, "loss": 0.3534, "step": 21595 }, { "epoch": 2.57, "learning_rate": 2.6232358152462e-07, "loss": 0.408, "step": 21596 }, { "epoch": 2.57, "learning_rate": 2.6218026942787134e-07, "loss": 0.4119, "step": 21597 }, { "epoch": 2.57, "learning_rate": 2.62036994322471e-07, "loss": 0.5168, "step": 21598 }, { "epoch": 2.57, "learning_rate": 2.618937562107868e-07, "loss": 0.6174, "step": 21599 }, { "epoch": 2.57, "learning_rate": 2.6175055509518625e-07, "loss": 0.5409, "step": 21600 }, { "epoch": 2.57, "learning_rate": 2.6160739097803674e-07, "loss": 0.3493, "step": 21601 }, { "epoch": 2.57, "learning_rate": 2.6146426386170485e-07, "loss": 0.4243, "step": 21602 }, { "epoch": 2.57, "learning_rate": 2.613211737485569e-07, "loss": 0.4013, "step": 21603 }, { "epoch": 2.57, "learning_rate": 2.611781206409575e-07, "loss": 0.471, "step": 21604 }, { "epoch": 2.57, "learning_rate": 2.6103510454127204e-07, "loss": 0.5442, "step": 21605 }, { "epoch": 2.57, "learning_rate": 2.608921254518637e-07, "loss": 0.6033, "step": 21606 }, { "epoch": 2.57, "learning_rate": 2.607491833750969e-07, "loss": 0.4374, "step": 21607 }, { "epoch": 2.57, "learning_rate": 2.606062783133334e-07, "loss": 0.388, "step": 21608 }, { "epoch": 2.57, "learning_rate": 2.60463410268936e-07, "loss": 0.3483, "step": 21609 }, { "epoch": 2.57, "learning_rate": 2.6032057924426644e-07, "loss": 0.3935, "step": 21610 }, { "epoch": 2.57, "learning_rate": 2.601777852416862e-07, "loss": 0.5724, "step": 21611 }, { "epoch": 2.57, "learning_rate": 2.60035028263555e-07, "loss": 0.6313, "step": 21612 }, { "epoch": 2.57, "learning_rate": 2.598923083122323e-07, "loss": 0.3008, "step": 21613 }, { "epoch": 2.57, "learning_rate": 2.5974962539007804e-07, "loss": 0.38, "step": 21614 }, { "epoch": 2.57, "learning_rate": 2.596069794994502e-07, "loss": 0.4314, "step": 21615 }, { "epoch": 2.57, "learning_rate": 2.594643706427066e-07, "loss": 0.4159, "step": 21616 }, { "epoch": 2.57, "learning_rate": 2.5932179882220523e-07, "loss": 0.5138, "step": 21617 }, { "epoch": 2.57, "learning_rate": 2.5917926404030294e-07, "loss": 0.5337, "step": 21618 }, { "epoch": 2.57, "learning_rate": 2.590367662993551e-07, "loss": 0.5634, "step": 21619 }, { "epoch": 2.57, "learning_rate": 2.588943056017179e-07, "loss": 0.2319, "step": 21620 }, { "epoch": 2.57, "learning_rate": 2.5875188194974533e-07, "loss": 0.395, "step": 21621 }, { "epoch": 2.57, "learning_rate": 2.586094953457929e-07, "loss": 0.391, "step": 21622 }, { "epoch": 2.58, "learning_rate": 2.584671457922128e-07, "loss": 0.4416, "step": 21623 }, { "epoch": 2.58, "learning_rate": 2.5832483329135907e-07, "loss": 0.6169, "step": 21624 }, { "epoch": 2.58, "learning_rate": 2.5818255784558406e-07, "loss": 0.5211, "step": 21625 }, { "epoch": 2.58, "learning_rate": 2.5804031945723976e-07, "loss": 0.2996, "step": 21626 }, { "epoch": 2.58, "learning_rate": 2.5789811812867717e-07, "loss": 0.5298, "step": 21627 }, { "epoch": 2.58, "learning_rate": 2.577559538622462e-07, "loss": 0.511, "step": 21628 }, { "epoch": 2.58, "learning_rate": 2.5761382666029733e-07, "loss": 0.4713, "step": 21629 }, { "epoch": 2.58, "learning_rate": 2.574717365251808e-07, "loss": 0.5402, "step": 21630 }, { "epoch": 2.58, "learning_rate": 2.573296834592443e-07, "loss": 0.594, "step": 21631 }, { "epoch": 2.58, "learning_rate": 2.57187667464836e-07, "loss": 0.4077, "step": 21632 }, { "epoch": 2.58, "learning_rate": 2.570456885443043e-07, "loss": 0.5188, "step": 21633 }, { "epoch": 2.58, "learning_rate": 2.569037466999957e-07, "loss": 0.4359, "step": 21634 }, { "epoch": 2.58, "learning_rate": 2.5676184193425573e-07, "loss": 0.4358, "step": 21635 }, { "epoch": 2.58, "learning_rate": 2.5661997424943076e-07, "loss": 0.5262, "step": 21636 }, { "epoch": 2.58, "learning_rate": 2.564781436478664e-07, "loss": 0.7722, "step": 21637 }, { "epoch": 2.58, "learning_rate": 2.5633635013190603e-07, "loss": 0.402, "step": 21638 }, { "epoch": 2.58, "learning_rate": 2.5619459370389397e-07, "loss": 0.3783, "step": 21639 }, { "epoch": 2.58, "learning_rate": 2.5605287436617416e-07, "loss": 0.4263, "step": 21640 }, { "epoch": 2.58, "learning_rate": 2.559111921210883e-07, "loss": 0.4201, "step": 21641 }, { "epoch": 2.58, "learning_rate": 2.557695469709789e-07, "loss": 0.4167, "step": 21642 }, { "epoch": 2.58, "learning_rate": 2.556279389181868e-07, "loss": 0.5989, "step": 21643 }, { "epoch": 2.58, "learning_rate": 2.554863679650532e-07, "loss": 0.5381, "step": 21644 }, { "epoch": 2.58, "learning_rate": 2.553448341139189e-07, "loss": 0.3283, "step": 21645 }, { "epoch": 2.58, "learning_rate": 2.552033373671223e-07, "loss": 0.4053, "step": 21646 }, { "epoch": 2.58, "learning_rate": 2.550618777270028e-07, "loss": 0.4299, "step": 21647 }, { "epoch": 2.58, "learning_rate": 2.549204551958992e-07, "loss": 0.4874, "step": 21648 }, { "epoch": 2.58, "learning_rate": 2.54779069776149e-07, "loss": 0.6696, "step": 21649 }, { "epoch": 2.58, "learning_rate": 2.5463772147008873e-07, "loss": 0.5121, "step": 21650 }, { "epoch": 2.58, "learning_rate": 2.544964102800551e-07, "loss": 0.2946, "step": 21651 }, { "epoch": 2.58, "learning_rate": 2.5435513620838487e-07, "loss": 0.4118, "step": 21652 }, { "epoch": 2.58, "learning_rate": 2.5421389925741226e-07, "loss": 0.3955, "step": 21653 }, { "epoch": 2.58, "learning_rate": 2.5407269942947233e-07, "loss": 0.5202, "step": 21654 }, { "epoch": 2.58, "learning_rate": 2.5393153672689933e-07, "loss": 0.6217, "step": 21655 }, { "epoch": 2.58, "learning_rate": 2.5379041115202675e-07, "loss": 0.5847, "step": 21656 }, { "epoch": 2.58, "learning_rate": 2.536493227071868e-07, "loss": 0.3455, "step": 21657 }, { "epoch": 2.58, "learning_rate": 2.5350827139471177e-07, "loss": 0.3333, "step": 21658 }, { "epoch": 2.58, "learning_rate": 2.533672572169335e-07, "loss": 0.4273, "step": 21659 }, { "epoch": 2.58, "learning_rate": 2.5322628017618365e-07, "loss": 0.5451, "step": 21660 }, { "epoch": 2.58, "learning_rate": 2.5308534027479123e-07, "loss": 0.52, "step": 21661 }, { "epoch": 2.58, "learning_rate": 2.529444375150872e-07, "loss": 0.5698, "step": 21662 }, { "epoch": 2.58, "learning_rate": 2.5280357189940015e-07, "loss": 0.4368, "step": 21663 }, { "epoch": 2.58, "learning_rate": 2.5266274343005825e-07, "loss": 0.3871, "step": 21664 }, { "epoch": 2.58, "learning_rate": 2.525219521093897e-07, "loss": 0.3384, "step": 21665 }, { "epoch": 2.58, "learning_rate": 2.523811979397217e-07, "loss": 0.4014, "step": 21666 }, { "epoch": 2.58, "learning_rate": 2.5224048092338167e-07, "loss": 0.4435, "step": 21667 }, { "epoch": 2.58, "learning_rate": 2.520998010626949e-07, "loss": 0.5836, "step": 21668 }, { "epoch": 2.58, "learning_rate": 2.5195915835998705e-07, "loss": 0.618, "step": 21669 }, { "epoch": 2.58, "learning_rate": 2.5181855281758286e-07, "loss": 0.3557, "step": 21670 }, { "epoch": 2.58, "learning_rate": 2.5167798443780675e-07, "loss": 0.3429, "step": 21671 }, { "epoch": 2.58, "learning_rate": 2.5153745322298197e-07, "loss": 0.4424, "step": 21672 }, { "epoch": 2.58, "learning_rate": 2.513969591754317e-07, "loss": 0.4823, "step": 21673 }, { "epoch": 2.58, "learning_rate": 2.5125650229747835e-07, "loss": 0.575, "step": 21674 }, { "epoch": 2.58, "learning_rate": 2.511160825914441e-07, "loss": 0.5385, "step": 21675 }, { "epoch": 2.58, "learning_rate": 2.509757000596494e-07, "loss": 0.3605, "step": 21676 }, { "epoch": 2.58, "learning_rate": 2.508353547044154e-07, "loss": 0.4212, "step": 21677 }, { "epoch": 2.58, "learning_rate": 2.506950465280619e-07, "loss": 0.4848, "step": 21678 }, { "epoch": 2.58, "learning_rate": 2.5055477553290743e-07, "loss": 0.5122, "step": 21679 }, { "epoch": 2.58, "learning_rate": 2.504145417212711e-07, "loss": 0.5616, "step": 21680 }, { "epoch": 2.58, "learning_rate": 2.5027434509547143e-07, "loss": 0.5524, "step": 21681 }, { "epoch": 2.58, "learning_rate": 2.50134185657826e-07, "loss": 0.2825, "step": 21682 }, { "epoch": 2.58, "learning_rate": 2.499940634106507e-07, "loss": 0.4019, "step": 21683 }, { "epoch": 2.58, "learning_rate": 2.4985397835626286e-07, "loss": 0.5037, "step": 21684 }, { "epoch": 2.58, "learning_rate": 2.49713930496977e-07, "loss": 0.4611, "step": 21685 }, { "epoch": 2.58, "learning_rate": 2.495739198351091e-07, "loss": 0.5968, "step": 21686 }, { "epoch": 2.58, "learning_rate": 2.494339463729728e-07, "loss": 0.6192, "step": 21687 }, { "epoch": 2.58, "learning_rate": 2.4929401011288214e-07, "loss": 0.4668, "step": 21688 }, { "epoch": 2.58, "learning_rate": 2.491541110571505e-07, "loss": 0.3105, "step": 21689 }, { "epoch": 2.58, "learning_rate": 2.4901424920809046e-07, "loss": 0.3545, "step": 21690 }, { "epoch": 2.58, "learning_rate": 2.4887442456801414e-07, "loss": 0.395, "step": 21691 }, { "epoch": 2.58, "learning_rate": 2.487346371392316e-07, "loss": 0.4611, "step": 21692 }, { "epoch": 2.58, "learning_rate": 2.485948869240545e-07, "loss": 0.5913, "step": 21693 }, { "epoch": 2.58, "learning_rate": 2.484551739247934e-07, "loss": 0.5569, "step": 21694 }, { "epoch": 2.58, "learning_rate": 2.4831549814375686e-07, "loss": 0.2887, "step": 21695 }, { "epoch": 2.58, "learning_rate": 2.4817585958325397e-07, "loss": 0.3997, "step": 21696 }, { "epoch": 2.58, "learning_rate": 2.4803625824559336e-07, "loss": 0.4011, "step": 21697 }, { "epoch": 2.58, "learning_rate": 2.4789669413308207e-07, "loss": 0.422, "step": 21698 }, { "epoch": 2.58, "learning_rate": 2.477571672480278e-07, "loss": 0.6035, "step": 21699 }, { "epoch": 2.58, "learning_rate": 2.4761767759273606e-07, "loss": 0.5784, "step": 21700 }, { "epoch": 2.58, "learning_rate": 2.474782251695135e-07, "loss": 0.4025, "step": 21701 }, { "epoch": 2.58, "learning_rate": 2.4733880998066453e-07, "loss": 0.381, "step": 21702 }, { "epoch": 2.58, "learning_rate": 2.47199432028494e-07, "loss": 0.4231, "step": 21703 }, { "epoch": 2.58, "learning_rate": 2.470600913153057e-07, "loss": 0.4151, "step": 21704 }, { "epoch": 2.58, "learning_rate": 2.4692078784340377e-07, "loss": 0.5781, "step": 21705 }, { "epoch": 2.58, "learning_rate": 2.467815216150904e-07, "loss": 0.5763, "step": 21706 }, { "epoch": 2.59, "learning_rate": 2.4664229263266695e-07, "loss": 0.2677, "step": 21707 }, { "epoch": 2.59, "learning_rate": 2.465031008984353e-07, "loss": 0.3454, "step": 21708 }, { "epoch": 2.59, "learning_rate": 2.463639464146972e-07, "loss": 0.3896, "step": 21709 }, { "epoch": 2.59, "learning_rate": 2.4622482918375155e-07, "loss": 0.4951, "step": 21710 }, { "epoch": 2.59, "learning_rate": 2.4608574920789856e-07, "loss": 0.4874, "step": 21711 }, { "epoch": 2.59, "learning_rate": 2.4594670648943805e-07, "loss": 0.6105, "step": 21712 }, { "epoch": 2.59, "learning_rate": 2.458077010306673e-07, "loss": 0.2984, "step": 21713 }, { "epoch": 2.59, "learning_rate": 2.456687328338839e-07, "loss": 0.3555, "step": 21714 }, { "epoch": 2.59, "learning_rate": 2.455298019013855e-07, "loss": 0.3627, "step": 21715 }, { "epoch": 2.59, "learning_rate": 2.4539090823546914e-07, "loss": 0.4131, "step": 21716 }, { "epoch": 2.59, "learning_rate": 2.4525205183842993e-07, "loss": 0.6013, "step": 21717 }, { "epoch": 2.59, "learning_rate": 2.4511323271256326e-07, "loss": 0.5993, "step": 21718 }, { "epoch": 2.59, "learning_rate": 2.449744508601648e-07, "loss": 0.4538, "step": 21719 }, { "epoch": 2.59, "learning_rate": 2.448357062835277e-07, "loss": 0.2847, "step": 21720 }, { "epoch": 2.59, "learning_rate": 2.4469699898494513e-07, "loss": 0.3886, "step": 21721 }, { "epoch": 2.59, "learning_rate": 2.4455832896671053e-07, "loss": 0.3869, "step": 21722 }, { "epoch": 2.59, "learning_rate": 2.444196962311157e-07, "loss": 0.4744, "step": 21723 }, { "epoch": 2.59, "learning_rate": 2.4428110078045323e-07, "loss": 0.5596, "step": 21724 }, { "epoch": 2.59, "learning_rate": 2.4414254261701297e-07, "loss": 0.6351, "step": 21725 }, { "epoch": 2.59, "learning_rate": 2.4400402174308643e-07, "loss": 0.4269, "step": 21726 }, { "epoch": 2.59, "learning_rate": 2.43865538160962e-07, "loss": 0.4863, "step": 21727 }, { "epoch": 2.59, "learning_rate": 2.437270918729298e-07, "loss": 0.4618, "step": 21728 }, { "epoch": 2.59, "learning_rate": 2.435886828812781e-07, "loss": 0.4916, "step": 21729 }, { "epoch": 2.59, "learning_rate": 2.4345031118829435e-07, "loss": 0.5115, "step": 21730 }, { "epoch": 2.59, "learning_rate": 2.4331197679626687e-07, "loss": 0.6396, "step": 21731 }, { "epoch": 2.59, "learning_rate": 2.4317367970748155e-07, "loss": 0.2838, "step": 21732 }, { "epoch": 2.59, "learning_rate": 2.4303541992422437e-07, "loss": 0.3221, "step": 21733 }, { "epoch": 2.59, "learning_rate": 2.4289719744878175e-07, "loss": 0.4008, "step": 21734 }, { "epoch": 2.59, "learning_rate": 2.4275901228343777e-07, "loss": 0.3983, "step": 21735 }, { "epoch": 2.59, "learning_rate": 2.4262086443047617e-07, "loss": 0.6283, "step": 21736 }, { "epoch": 2.59, "learning_rate": 2.424827538921812e-07, "loss": 0.6473, "step": 21737 }, { "epoch": 2.59, "learning_rate": 2.423446806708357e-07, "loss": 0.4581, "step": 21738 }, { "epoch": 2.59, "learning_rate": 2.422066447687227e-07, "loss": 0.2728, "step": 21739 }, { "epoch": 2.59, "learning_rate": 2.420686461881225e-07, "loss": 0.3714, "step": 21740 }, { "epoch": 2.59, "learning_rate": 2.419306849313177e-07, "loss": 0.3883, "step": 21741 }, { "epoch": 2.59, "learning_rate": 2.417927610005882e-07, "loss": 0.524, "step": 21742 }, { "epoch": 2.59, "learning_rate": 2.4165487439821305e-07, "loss": 0.5151, "step": 21743 }, { "epoch": 2.59, "learning_rate": 2.4151702512647257e-07, "loss": 0.632, "step": 21744 }, { "epoch": 2.59, "learning_rate": 2.4137921318764523e-07, "loss": 0.2917, "step": 21745 }, { "epoch": 2.59, "learning_rate": 2.412414385840095e-07, "loss": 0.3436, "step": 21746 }, { "epoch": 2.59, "learning_rate": 2.4110370131784164e-07, "loss": 0.4561, "step": 21747 }, { "epoch": 2.59, "learning_rate": 2.409660013914197e-07, "loss": 0.4853, "step": 21748 }, { "epoch": 2.59, "learning_rate": 2.408283388070187e-07, "loss": 0.6348, "step": 21749 }, { "epoch": 2.59, "learning_rate": 2.4069071356691567e-07, "loss": 0.5889, "step": 21750 }, { "epoch": 2.59, "learning_rate": 2.4055312567338397e-07, "loss": 0.3364, "step": 21751 }, { "epoch": 2.59, "learning_rate": 2.404155751286988e-07, "loss": 0.397, "step": 21752 }, { "epoch": 2.59, "learning_rate": 2.4027806193513357e-07, "loss": 0.4681, "step": 21753 }, { "epoch": 2.59, "learning_rate": 2.401405860949621e-07, "loss": 0.5046, "step": 21754 }, { "epoch": 2.59, "learning_rate": 2.400031476104564e-07, "loss": 0.4758, "step": 21755 }, { "epoch": 2.59, "learning_rate": 2.3986574648388797e-07, "loss": 0.5726, "step": 21756 }, { "epoch": 2.59, "learning_rate": 2.397283827175281e-07, "loss": 0.3646, "step": 21757 }, { "epoch": 2.59, "learning_rate": 2.395910563136486e-07, "loss": 0.3817, "step": 21758 }, { "epoch": 2.59, "learning_rate": 2.3945376727451787e-07, "loss": 0.4273, "step": 21759 }, { "epoch": 2.59, "learning_rate": 2.3931651560240603e-07, "loss": 0.4882, "step": 21760 }, { "epoch": 2.59, "learning_rate": 2.3917930129958243e-07, "loss": 0.5624, "step": 21761 }, { "epoch": 2.59, "learning_rate": 2.3904212436831415e-07, "loss": 0.714, "step": 21762 }, { "epoch": 2.59, "learning_rate": 2.389049848108696e-07, "loss": 0.3473, "step": 21763 }, { "epoch": 2.59, "learning_rate": 2.3876788262951505e-07, "loss": 0.4253, "step": 21764 }, { "epoch": 2.59, "learning_rate": 2.386308178265179e-07, "loss": 0.3957, "step": 21765 }, { "epoch": 2.59, "learning_rate": 2.384937904041421e-07, "loss": 0.4854, "step": 21766 }, { "epoch": 2.59, "learning_rate": 2.3835680036465386e-07, "loss": 0.5136, "step": 21767 }, { "epoch": 2.59, "learning_rate": 2.382198477103176e-07, "loss": 0.5723, "step": 21768 }, { "epoch": 2.59, "learning_rate": 2.3808293244339752e-07, "loss": 0.5609, "step": 21769 }, { "epoch": 2.59, "learning_rate": 2.3794605456615633e-07, "loss": 0.4491, "step": 21770 }, { "epoch": 2.59, "learning_rate": 2.3780921408085606e-07, "loss": 0.3921, "step": 21771 }, { "epoch": 2.59, "learning_rate": 2.3767241098975934e-07, "loss": 0.4546, "step": 21772 }, { "epoch": 2.59, "learning_rate": 2.37535645295128e-07, "loss": 0.5551, "step": 21773 }, { "epoch": 2.59, "learning_rate": 2.373989169992219e-07, "loss": 0.6426, "step": 21774 }, { "epoch": 2.59, "learning_rate": 2.3726222610430144e-07, "loss": 0.5881, "step": 21775 }, { "epoch": 2.59, "learning_rate": 2.3712557261262646e-07, "loss": 0.2566, "step": 21776 }, { "epoch": 2.59, "learning_rate": 2.3698895652645605e-07, "loss": 0.388, "step": 21777 }, { "epoch": 2.59, "learning_rate": 2.3685237784804748e-07, "loss": 0.4536, "step": 21778 }, { "epoch": 2.59, "learning_rate": 2.3671583657965903e-07, "loss": 0.4462, "step": 21779 }, { "epoch": 2.59, "learning_rate": 2.36579332723548e-07, "loss": 0.593, "step": 21780 }, { "epoch": 2.59, "learning_rate": 2.364428662819701e-07, "loss": 0.5439, "step": 21781 }, { "epoch": 2.59, "learning_rate": 2.3630643725718162e-07, "loss": 0.3998, "step": 21782 }, { "epoch": 2.59, "learning_rate": 2.361700456514382e-07, "loss": 0.4005, "step": 21783 }, { "epoch": 2.59, "learning_rate": 2.3603369146699335e-07, "loss": 0.3957, "step": 21784 }, { "epoch": 2.59, "learning_rate": 2.358973747061019e-07, "loss": 0.4207, "step": 21785 }, { "epoch": 2.59, "learning_rate": 2.3576109537101654e-07, "loss": 0.5668, "step": 21786 }, { "epoch": 2.59, "learning_rate": 2.3562485346399016e-07, "loss": 0.5684, "step": 21787 }, { "epoch": 2.59, "learning_rate": 2.3548864898727542e-07, "loss": 0.3803, "step": 21788 }, { "epoch": 2.59, "learning_rate": 2.3535248194312272e-07, "loss": 0.3126, "step": 21789 }, { "epoch": 2.59, "learning_rate": 2.3521635233378388e-07, "loss": 0.2815, "step": 21790 }, { "epoch": 2.6, "learning_rate": 2.3508026016150875e-07, "loss": 0.4311, "step": 21791 }, { "epoch": 2.6, "learning_rate": 2.3494420542854723e-07, "loss": 0.5735, "step": 21792 }, { "epoch": 2.6, "learning_rate": 2.3480818813714778e-07, "loss": 0.5931, "step": 21793 }, { "epoch": 2.6, "learning_rate": 2.3467220828955862e-07, "loss": 0.5679, "step": 21794 }, { "epoch": 2.6, "learning_rate": 2.3453626588802874e-07, "loss": 0.2743, "step": 21795 }, { "epoch": 2.6, "learning_rate": 2.344003609348039e-07, "loss": 0.4096, "step": 21796 }, { "epoch": 2.6, "learning_rate": 2.3426449343213114e-07, "loss": 0.3827, "step": 21797 }, { "epoch": 2.6, "learning_rate": 2.3412866338225704e-07, "loss": 0.448, "step": 21798 }, { "epoch": 2.6, "learning_rate": 2.339928707874259e-07, "loss": 0.6664, "step": 21799 }, { "epoch": 2.6, "learning_rate": 2.3385711564988256e-07, "loss": 0.6514, "step": 21800 }, { "epoch": 2.6, "learning_rate": 2.337213979718711e-07, "loss": 0.3076, "step": 21801 }, { "epoch": 2.6, "learning_rate": 2.3358571775563528e-07, "loss": 0.3605, "step": 21802 }, { "epoch": 2.6, "learning_rate": 2.3345007500341799e-07, "loss": 0.4581, "step": 21803 }, { "epoch": 2.6, "learning_rate": 2.333144697174608e-07, "loss": 0.4912, "step": 21804 }, { "epoch": 2.6, "learning_rate": 2.3317890190000608e-07, "loss": 0.5138, "step": 21805 }, { "epoch": 2.6, "learning_rate": 2.3304337155329426e-07, "loss": 0.6071, "step": 21806 }, { "epoch": 2.6, "learning_rate": 2.329078786795652e-07, "loss": 0.3855, "step": 21807 }, { "epoch": 2.6, "learning_rate": 2.3277242328105936e-07, "loss": 0.284, "step": 21808 }, { "epoch": 2.6, "learning_rate": 2.3263700536001548e-07, "loss": 0.3983, "step": 21809 }, { "epoch": 2.6, "learning_rate": 2.3250162491867262e-07, "loss": 0.4715, "step": 21810 }, { "epoch": 2.6, "learning_rate": 2.3236628195926758e-07, "loss": 0.4898, "step": 21811 }, { "epoch": 2.6, "learning_rate": 2.322309764840386e-07, "loss": 0.6667, "step": 21812 }, { "epoch": 2.6, "learning_rate": 2.320957084952216e-07, "loss": 0.2843, "step": 21813 }, { "epoch": 2.6, "learning_rate": 2.3196047799505323e-07, "loss": 0.3887, "step": 21814 }, { "epoch": 2.6, "learning_rate": 2.3182528498576773e-07, "loss": 0.4256, "step": 21815 }, { "epoch": 2.6, "learning_rate": 2.3169012946960083e-07, "loss": 0.4458, "step": 21816 }, { "epoch": 2.6, "learning_rate": 2.3155501144878661e-07, "loss": 0.6165, "step": 21817 }, { "epoch": 2.6, "learning_rate": 2.3141993092555854e-07, "loss": 0.5757, "step": 21818 }, { "epoch": 2.6, "learning_rate": 2.3128488790214897e-07, "loss": 0.5079, "step": 21819 }, { "epoch": 2.6, "learning_rate": 2.3114988238079118e-07, "loss": 0.2586, "step": 21820 }, { "epoch": 2.6, "learning_rate": 2.310149143637161e-07, "loss": 0.3607, "step": 21821 }, { "epoch": 2.6, "learning_rate": 2.308799838531542e-07, "loss": 0.4636, "step": 21822 }, { "epoch": 2.6, "learning_rate": 2.3074509085133672e-07, "loss": 0.5142, "step": 21823 }, { "epoch": 2.6, "learning_rate": 2.306102353604933e-07, "loss": 0.5911, "step": 21824 }, { "epoch": 2.6, "learning_rate": 2.3047541738285384e-07, "loss": 0.5785, "step": 21825 }, { "epoch": 2.6, "learning_rate": 2.303406369206454e-07, "loss": 0.4138, "step": 21826 }, { "epoch": 2.6, "learning_rate": 2.3020589397609705e-07, "loss": 0.372, "step": 21827 }, { "epoch": 2.6, "learning_rate": 2.3007118855143562e-07, "loss": 0.4623, "step": 21828 }, { "epoch": 2.6, "learning_rate": 2.2993652064888794e-07, "loss": 0.4583, "step": 21829 }, { "epoch": 2.6, "learning_rate": 2.2980189027068e-07, "loss": 0.6117, "step": 21830 }, { "epoch": 2.6, "learning_rate": 2.2966729741903697e-07, "loss": 0.6077, "step": 21831 }, { "epoch": 2.6, "learning_rate": 2.2953274209618404e-07, "loss": 0.2956, "step": 21832 }, { "epoch": 2.6, "learning_rate": 2.2939822430434605e-07, "loss": 0.3605, "step": 21833 }, { "epoch": 2.6, "learning_rate": 2.2926374404574598e-07, "loss": 0.3814, "step": 21834 }, { "epoch": 2.6, "learning_rate": 2.291293013226059e-07, "loss": 0.4872, "step": 21835 }, { "epoch": 2.6, "learning_rate": 2.2899489613714936e-07, "loss": 0.5535, "step": 21836 }, { "epoch": 2.6, "learning_rate": 2.2886052849159818e-07, "loss": 0.6494, "step": 21837 }, { "epoch": 2.6, "learning_rate": 2.287261983881725e-07, "loss": 0.558, "step": 21838 }, { "epoch": 2.6, "learning_rate": 2.2859190582909307e-07, "loss": 0.4011, "step": 21839 }, { "epoch": 2.6, "learning_rate": 2.284576508165809e-07, "loss": 0.3661, "step": 21840 }, { "epoch": 2.6, "learning_rate": 2.2832343335285363e-07, "loss": 0.4138, "step": 21841 }, { "epoch": 2.6, "learning_rate": 2.2818925344013115e-07, "loss": 0.5052, "step": 21842 }, { "epoch": 2.6, "learning_rate": 2.2805511108063062e-07, "loss": 0.6119, "step": 21843 }, { "epoch": 2.6, "learning_rate": 2.2792100627656993e-07, "loss": 0.4758, "step": 21844 }, { "epoch": 2.6, "learning_rate": 2.277869390301654e-07, "loss": 0.3417, "step": 21845 }, { "epoch": 2.6, "learning_rate": 2.2765290934363332e-07, "loss": 0.4012, "step": 21846 }, { "epoch": 2.6, "learning_rate": 2.275189172191894e-07, "loss": 0.4094, "step": 21847 }, { "epoch": 2.6, "learning_rate": 2.273849626590488e-07, "loss": 0.4983, "step": 21848 }, { "epoch": 2.6, "learning_rate": 2.272510456654256e-07, "loss": 0.6229, "step": 21849 }, { "epoch": 2.6, "learning_rate": 2.2711716624053277e-07, "loss": 0.527, "step": 21850 }, { "epoch": 2.6, "learning_rate": 2.2698332438658377e-07, "loss": 0.2881, "step": 21851 }, { "epoch": 2.6, "learning_rate": 2.2684952010579188e-07, "loss": 0.3211, "step": 21852 }, { "epoch": 2.6, "learning_rate": 2.267157534003675e-07, "loss": 0.4526, "step": 21853 }, { "epoch": 2.6, "learning_rate": 2.2658202427252252e-07, "loss": 0.453, "step": 21854 }, { "epoch": 2.6, "learning_rate": 2.2644833272446792e-07, "loss": 0.5043, "step": 21855 }, { "epoch": 2.6, "learning_rate": 2.2631467875841307e-07, "loss": 0.6714, "step": 21856 }, { "epoch": 2.6, "learning_rate": 2.2618106237656702e-07, "loss": 0.3281, "step": 21857 }, { "epoch": 2.6, "learning_rate": 2.2604748358113855e-07, "loss": 0.3685, "step": 21858 }, { "epoch": 2.6, "learning_rate": 2.2591394237433673e-07, "loss": 0.4225, "step": 21859 }, { "epoch": 2.6, "learning_rate": 2.2578043875836758e-07, "loss": 0.4729, "step": 21860 }, { "epoch": 2.6, "learning_rate": 2.2564697273543878e-07, "loss": 0.481, "step": 21861 }, { "epoch": 2.6, "learning_rate": 2.2551354430775661e-07, "loss": 0.5547, "step": 21862 }, { "epoch": 2.6, "learning_rate": 2.2538015347752657e-07, "loss": 0.2838, "step": 21863 }, { "epoch": 2.6, "learning_rate": 2.2524680024695295e-07, "loss": 0.3441, "step": 21864 }, { "epoch": 2.6, "learning_rate": 2.251134846182404e-07, "loss": 0.3654, "step": 21865 }, { "epoch": 2.6, "learning_rate": 2.2498020659359299e-07, "loss": 0.4015, "step": 21866 }, { "epoch": 2.6, "learning_rate": 2.2484696617521423e-07, "loss": 0.572, "step": 21867 }, { "epoch": 2.6, "learning_rate": 2.2471376336530543e-07, "loss": 0.6458, "step": 21868 }, { "epoch": 2.6, "learning_rate": 2.2458059816606953e-07, "loss": 0.5266, "step": 21869 }, { "epoch": 2.6, "learning_rate": 2.2444747057970674e-07, "loss": 0.3465, "step": 21870 }, { "epoch": 2.6, "learning_rate": 2.2431438060841859e-07, "loss": 0.408, "step": 21871 }, { "epoch": 2.6, "learning_rate": 2.2418132825440448e-07, "loss": 0.4162, "step": 21872 }, { "epoch": 2.6, "learning_rate": 2.2404831351986374e-07, "loss": 0.5815, "step": 21873 }, { "epoch": 2.6, "learning_rate": 2.2391533640699597e-07, "loss": 0.6515, "step": 21874 }, { "epoch": 2.61, "learning_rate": 2.2378239691799807e-07, "loss": 0.6129, "step": 21875 }, { "epoch": 2.61, "learning_rate": 2.2364949505506827e-07, "loss": 0.2873, "step": 21876 }, { "epoch": 2.61, "learning_rate": 2.235166308204037e-07, "loss": 0.4257, "step": 21877 }, { "epoch": 2.61, "learning_rate": 2.233838042162001e-07, "loss": 0.3942, "step": 21878 }, { "epoch": 2.61, "learning_rate": 2.2325101524465291e-07, "loss": 0.3609, "step": 21879 }, { "epoch": 2.61, "learning_rate": 2.2311826390795738e-07, "loss": 0.5722, "step": 21880 }, { "epoch": 2.61, "learning_rate": 2.2298555020830808e-07, "loss": 0.6993, "step": 21881 }, { "epoch": 2.61, "learning_rate": 2.228528741478994e-07, "loss": 0.3444, "step": 21882 }, { "epoch": 2.61, "learning_rate": 2.2272023572892294e-07, "loss": 0.3024, "step": 21883 }, { "epoch": 2.61, "learning_rate": 2.2258763495357273e-07, "loss": 0.4206, "step": 21884 }, { "epoch": 2.61, "learning_rate": 2.224550718240398e-07, "loss": 0.4423, "step": 21885 }, { "epoch": 2.61, "learning_rate": 2.223225463425155e-07, "loss": 0.5707, "step": 21886 }, { "epoch": 2.61, "learning_rate": 2.2219005851119024e-07, "loss": 0.5422, "step": 21887 }, { "epoch": 2.61, "learning_rate": 2.220576083322548e-07, "loss": 0.461, "step": 21888 }, { "epoch": 2.61, "learning_rate": 2.2192519580789856e-07, "loss": 0.39, "step": 21889 }, { "epoch": 2.61, "learning_rate": 2.2179282094030945e-07, "loss": 0.4138, "step": 21890 }, { "epoch": 2.61, "learning_rate": 2.2166048373167658e-07, "loss": 0.4622, "step": 21891 }, { "epoch": 2.61, "learning_rate": 2.2152818418418681e-07, "loss": 0.5052, "step": 21892 }, { "epoch": 2.61, "learning_rate": 2.2139592230002755e-07, "loss": 0.6597, "step": 21893 }, { "epoch": 2.61, "learning_rate": 2.2126369808138455e-07, "loss": 0.6019, "step": 21894 }, { "epoch": 2.61, "learning_rate": 2.2113151153044382e-07, "loss": 0.3562, "step": 21895 }, { "epoch": 2.61, "learning_rate": 2.209993626493906e-07, "loss": 0.4601, "step": 21896 }, { "epoch": 2.61, "learning_rate": 2.2086725144040922e-07, "loss": 0.3999, "step": 21897 }, { "epoch": 2.61, "learning_rate": 2.207351779056832e-07, "loss": 0.5202, "step": 21898 }, { "epoch": 2.61, "learning_rate": 2.206031420473964e-07, "loss": 0.591, "step": 21899 }, { "epoch": 2.61, "learning_rate": 2.204711438677304e-07, "loss": 0.5468, "step": 21900 }, { "epoch": 2.61, "learning_rate": 2.203391833688681e-07, "loss": 0.3846, "step": 21901 }, { "epoch": 2.61, "learning_rate": 2.2020726055299006e-07, "loss": 0.4059, "step": 21902 }, { "epoch": 2.61, "learning_rate": 2.2007537542227703e-07, "loss": 0.4215, "step": 21903 }, { "epoch": 2.61, "learning_rate": 2.1994352797891e-07, "loss": 0.3904, "step": 21904 }, { "epoch": 2.61, "learning_rate": 2.1981171822506726e-07, "loss": 0.5161, "step": 21905 }, { "epoch": 2.61, "learning_rate": 2.196799461629287e-07, "loss": 0.5518, "step": 21906 }, { "epoch": 2.61, "learning_rate": 2.1954821179467122e-07, "loss": 0.3794, "step": 21907 }, { "epoch": 2.61, "learning_rate": 2.1941651512247392e-07, "loss": 0.3629, "step": 21908 }, { "epoch": 2.61, "learning_rate": 2.1928485614851254e-07, "loss": 0.509, "step": 21909 }, { "epoch": 2.61, "learning_rate": 2.1915323487496365e-07, "loss": 0.5265, "step": 21910 }, { "epoch": 2.61, "learning_rate": 2.1902165130400305e-07, "loss": 0.512, "step": 21911 }, { "epoch": 2.61, "learning_rate": 2.1889010543780675e-07, "loss": 0.4993, "step": 21912 }, { "epoch": 2.61, "learning_rate": 2.1875859727854832e-07, "loss": 0.344, "step": 21913 }, { "epoch": 2.61, "learning_rate": 2.1862712682840097e-07, "loss": 0.3315, "step": 21914 }, { "epoch": 2.61, "learning_rate": 2.1849569408953885e-07, "loss": 0.4445, "step": 21915 }, { "epoch": 2.61, "learning_rate": 2.1836429906413493e-07, "loss": 0.4645, "step": 21916 }, { "epoch": 2.61, "learning_rate": 2.1823294175435994e-07, "loss": 0.4575, "step": 21917 }, { "epoch": 2.61, "learning_rate": 2.1810162216238578e-07, "loss": 0.5822, "step": 21918 }, { "epoch": 2.61, "learning_rate": 2.1797034029038378e-07, "loss": 0.6539, "step": 21919 }, { "epoch": 2.61, "learning_rate": 2.1783909614052385e-07, "loss": 0.3177, "step": 21920 }, { "epoch": 2.61, "learning_rate": 2.1770788971497426e-07, "loss": 0.3956, "step": 21921 }, { "epoch": 2.61, "learning_rate": 2.1757672101590493e-07, "loss": 0.4707, "step": 21922 }, { "epoch": 2.61, "learning_rate": 2.1744559004548442e-07, "loss": 0.472, "step": 21923 }, { "epoch": 2.61, "learning_rate": 2.1731449680587936e-07, "loss": 0.6336, "step": 21924 }, { "epoch": 2.61, "learning_rate": 2.1718344129925712e-07, "loss": 0.5482, "step": 21925 }, { "epoch": 2.61, "learning_rate": 2.1705242352778462e-07, "loss": 0.3098, "step": 21926 }, { "epoch": 2.61, "learning_rate": 2.1692144349362675e-07, "loss": 0.4905, "step": 21927 }, { "epoch": 2.61, "learning_rate": 2.1679050119894934e-07, "loss": 0.3967, "step": 21928 }, { "epoch": 2.61, "learning_rate": 2.1665959664591617e-07, "loss": 0.4412, "step": 21929 }, { "epoch": 2.61, "learning_rate": 2.1652872983669138e-07, "loss": 0.5785, "step": 21930 }, { "epoch": 2.61, "learning_rate": 2.1639790077343874e-07, "loss": 0.6127, "step": 21931 }, { "epoch": 2.61, "learning_rate": 2.162671094583199e-07, "loss": 0.3455, "step": 21932 }, { "epoch": 2.61, "learning_rate": 2.1613635589349756e-07, "loss": 0.3484, "step": 21933 }, { "epoch": 2.61, "learning_rate": 2.1600564008113334e-07, "loss": 0.4094, "step": 21934 }, { "epoch": 2.61, "learning_rate": 2.1587496202338741e-07, "loss": 0.3914, "step": 21935 }, { "epoch": 2.61, "learning_rate": 2.157443217224195e-07, "loss": 0.561, "step": 21936 }, { "epoch": 2.61, "learning_rate": 2.1561371918038975e-07, "loss": 0.5562, "step": 21937 }, { "epoch": 2.61, "learning_rate": 2.1548315439945706e-07, "loss": 0.3965, "step": 21938 }, { "epoch": 2.61, "learning_rate": 2.1535262738177935e-07, "loss": 0.3788, "step": 21939 }, { "epoch": 2.61, "learning_rate": 2.1522213812951443e-07, "loss": 0.457, "step": 21940 }, { "epoch": 2.61, "learning_rate": 2.1509168664481966e-07, "loss": 0.4374, "step": 21941 }, { "epoch": 2.61, "learning_rate": 2.1496127292985087e-07, "loss": 0.5619, "step": 21942 }, { "epoch": 2.61, "learning_rate": 2.1483089698676378e-07, "loss": 0.557, "step": 21943 }, { "epoch": 2.61, "learning_rate": 2.147005588177134e-07, "loss": 0.6502, "step": 21944 }, { "epoch": 2.61, "learning_rate": 2.145702584248549e-07, "loss": 0.2788, "step": 21945 }, { "epoch": 2.61, "learning_rate": 2.1443999581034185e-07, "loss": 0.3914, "step": 21946 }, { "epoch": 2.61, "learning_rate": 2.1430977097632726e-07, "loss": 0.4612, "step": 21947 }, { "epoch": 2.61, "learning_rate": 2.141795839249644e-07, "loss": 0.4663, "step": 21948 }, { "epoch": 2.61, "learning_rate": 2.140494346584046e-07, "loss": 0.4897, "step": 21949 }, { "epoch": 2.61, "learning_rate": 2.139193231787992e-07, "loss": 0.4701, "step": 21950 }, { "epoch": 2.61, "learning_rate": 2.1378924948829926e-07, "loss": 0.3035, "step": 21951 }, { "epoch": 2.61, "learning_rate": 2.1365921358905472e-07, "loss": 0.3466, "step": 21952 }, { "epoch": 2.61, "learning_rate": 2.1352921548321553e-07, "loss": 0.3546, "step": 21953 }, { "epoch": 2.61, "learning_rate": 2.1339925517292997e-07, "loss": 0.3814, "step": 21954 }, { "epoch": 2.61, "learning_rate": 2.1326933266034717e-07, "loss": 0.6293, "step": 21955 }, { "epoch": 2.61, "learning_rate": 2.1313944794761342e-07, "loss": 0.6012, "step": 21956 }, { "epoch": 2.61, "learning_rate": 2.1300960103687733e-07, "loss": 0.2811, "step": 21957 }, { "epoch": 2.61, "learning_rate": 2.1287979193028357e-07, "loss": 0.2999, "step": 21958 }, { "epoch": 2.62, "learning_rate": 2.1275002062997902e-07, "loss": 0.4391, "step": 21959 }, { "epoch": 2.62, "learning_rate": 2.1262028713810834e-07, "loss": 0.5314, "step": 21960 }, { "epoch": 2.62, "learning_rate": 2.1249059145681677e-07, "loss": 0.5354, "step": 21961 }, { "epoch": 2.62, "learning_rate": 2.1236093358824734e-07, "loss": 0.6249, "step": 21962 }, { "epoch": 2.62, "learning_rate": 2.1223131353454414e-07, "loss": 0.3684, "step": 21963 }, { "epoch": 2.62, "learning_rate": 2.1210173129784906e-07, "loss": 0.2224, "step": 21964 }, { "epoch": 2.62, "learning_rate": 2.1197218688030403e-07, "loss": 0.4283, "step": 21965 }, { "epoch": 2.62, "learning_rate": 2.1184268028405065e-07, "loss": 0.464, "step": 21966 }, { "epoch": 2.62, "learning_rate": 2.117132115112297e-07, "loss": 0.5848, "step": 21967 }, { "epoch": 2.62, "learning_rate": 2.1158378056398171e-07, "loss": 0.6046, "step": 21968 }, { "epoch": 2.62, "learning_rate": 2.114543874444458e-07, "loss": 0.5002, "step": 21969 }, { "epoch": 2.62, "learning_rate": 2.1132503215476108e-07, "loss": 0.3098, "step": 21970 }, { "epoch": 2.62, "learning_rate": 2.1119571469706502e-07, "loss": 0.4336, "step": 21971 }, { "epoch": 2.62, "learning_rate": 2.1106643507349645e-07, "loss": 0.4905, "step": 21972 }, { "epoch": 2.62, "learning_rate": 2.1093719328619143e-07, "loss": 0.5145, "step": 21973 }, { "epoch": 2.62, "learning_rate": 2.1080798933728664e-07, "loss": 0.5828, "step": 21974 }, { "epoch": 2.62, "learning_rate": 2.106788232289178e-07, "loss": 0.5795, "step": 21975 }, { "epoch": 2.62, "learning_rate": 2.1054969496322047e-07, "loss": 0.2789, "step": 21976 }, { "epoch": 2.62, "learning_rate": 2.1042060454232877e-07, "loss": 0.3368, "step": 21977 }, { "epoch": 2.62, "learning_rate": 2.1029155196837624e-07, "loss": 0.433, "step": 21978 }, { "epoch": 2.62, "learning_rate": 2.101625372434965e-07, "loss": 0.4462, "step": 21979 }, { "epoch": 2.62, "learning_rate": 2.1003356036982253e-07, "loss": 0.4734, "step": 21980 }, { "epoch": 2.62, "learning_rate": 2.0990462134948542e-07, "loss": 0.628, "step": 21981 }, { "epoch": 2.62, "learning_rate": 2.0977572018461734e-07, "loss": 0.4016, "step": 21982 }, { "epoch": 2.62, "learning_rate": 2.0964685687734908e-07, "loss": 0.3503, "step": 21983 }, { "epoch": 2.62, "learning_rate": 2.0951803142981008e-07, "loss": 0.3657, "step": 21984 }, { "epoch": 2.62, "learning_rate": 2.0938924384413057e-07, "loss": 0.4875, "step": 21985 }, { "epoch": 2.62, "learning_rate": 2.0926049412243853e-07, "loss": 0.5592, "step": 21986 }, { "epoch": 2.62, "learning_rate": 2.0913178226686316e-07, "loss": 0.4974, "step": 21987 }, { "epoch": 2.62, "learning_rate": 2.090031082795313e-07, "loss": 0.4424, "step": 21988 }, { "epoch": 2.62, "learning_rate": 2.0887447216257046e-07, "loss": 0.2739, "step": 21989 }, { "epoch": 2.62, "learning_rate": 2.0874587391810669e-07, "loss": 0.4184, "step": 21990 }, { "epoch": 2.62, "learning_rate": 2.0861731354826607e-07, "loss": 0.4364, "step": 21991 }, { "epoch": 2.62, "learning_rate": 2.084887910551739e-07, "loss": 0.5782, "step": 21992 }, { "epoch": 2.62, "learning_rate": 2.0836030644095367e-07, "loss": 0.5608, "step": 21993 }, { "epoch": 2.62, "learning_rate": 2.0823185970772986e-07, "loss": 0.4454, "step": 21994 }, { "epoch": 2.62, "learning_rate": 2.081034508576263e-07, "loss": 0.3007, "step": 21995 }, { "epoch": 2.62, "learning_rate": 2.0797507989276434e-07, "loss": 0.5233, "step": 21996 }, { "epoch": 2.62, "learning_rate": 2.0784674681526672e-07, "loss": 0.4304, "step": 21997 }, { "epoch": 2.62, "learning_rate": 2.0771845162725514e-07, "loss": 0.4898, "step": 21998 }, { "epoch": 2.62, "learning_rate": 2.0759019433084976e-07, "loss": 0.6024, "step": 21999 }, { "epoch": 2.62, "learning_rate": 2.074619749281706e-07, "loss": 0.4772, "step": 22000 }, { "epoch": 2.62, "learning_rate": 2.0733379342133736e-07, "loss": 0.2435, "step": 22001 }, { "epoch": 2.62, "learning_rate": 2.0720564981246917e-07, "loss": 0.3097, "step": 22002 }, { "epoch": 2.62, "learning_rate": 2.0707754410368346e-07, "loss": 0.5104, "step": 22003 }, { "epoch": 2.62, "learning_rate": 2.069494762970986e-07, "loss": 0.4726, "step": 22004 }, { "epoch": 2.62, "learning_rate": 2.068214463948315e-07, "loss": 0.5852, "step": 22005 }, { "epoch": 2.62, "learning_rate": 2.0669345439899846e-07, "loss": 0.688, "step": 22006 }, { "epoch": 2.62, "learning_rate": 2.0656550031171451e-07, "loss": 0.3753, "step": 22007 }, { "epoch": 2.62, "learning_rate": 2.0643758413509547e-07, "loss": 0.395, "step": 22008 }, { "epoch": 2.62, "learning_rate": 2.0630970587125543e-07, "loss": 0.3711, "step": 22009 }, { "epoch": 2.62, "learning_rate": 2.0618186552230884e-07, "loss": 0.461, "step": 22010 }, { "epoch": 2.62, "learning_rate": 2.060540630903679e-07, "loss": 0.5852, "step": 22011 }, { "epoch": 2.62, "learning_rate": 2.0592629857754649e-07, "loss": 0.5595, "step": 22012 }, { "epoch": 2.62, "learning_rate": 2.0579857198595538e-07, "loss": 0.318, "step": 22013 }, { "epoch": 2.62, "learning_rate": 2.056708833177068e-07, "loss": 0.3188, "step": 22014 }, { "epoch": 2.62, "learning_rate": 2.0554323257491043e-07, "loss": 0.4451, "step": 22015 }, { "epoch": 2.62, "learning_rate": 2.054156197596771e-07, "loss": 0.4345, "step": 22016 }, { "epoch": 2.62, "learning_rate": 2.0528804487411651e-07, "loss": 0.5789, "step": 22017 }, { "epoch": 2.62, "learning_rate": 2.051605079203367e-07, "loss": 0.5557, "step": 22018 }, { "epoch": 2.62, "learning_rate": 2.0503300890044598e-07, "loss": 0.4332, "step": 22019 }, { "epoch": 2.62, "learning_rate": 2.049055478165529e-07, "loss": 0.2959, "step": 22020 }, { "epoch": 2.62, "learning_rate": 2.0477812467076363e-07, "loss": 0.3812, "step": 22021 }, { "epoch": 2.62, "learning_rate": 2.0465073946518422e-07, "loss": 0.4746, "step": 22022 }, { "epoch": 2.62, "learning_rate": 2.0452339220192046e-07, "loss": 0.601, "step": 22023 }, { "epoch": 2.62, "learning_rate": 2.0439608288307767e-07, "loss": 0.5723, "step": 22024 }, { "epoch": 2.62, "learning_rate": 2.0426881151076078e-07, "loss": 0.5234, "step": 22025 }, { "epoch": 2.62, "learning_rate": 2.0414157808707259e-07, "loss": 0.4749, "step": 22026 }, { "epoch": 2.62, "learning_rate": 2.040143826141172e-07, "loss": 0.4449, "step": 22027 }, { "epoch": 2.62, "learning_rate": 2.0388722509399662e-07, "loss": 0.3869, "step": 22028 }, { "epoch": 2.62, "learning_rate": 2.037601055288127e-07, "loss": 0.5201, "step": 22029 }, { "epoch": 2.62, "learning_rate": 2.036330239206666e-07, "loss": 0.4926, "step": 22030 }, { "epoch": 2.62, "learning_rate": 2.0350598027165935e-07, "loss": 0.5344, "step": 22031 }, { "epoch": 2.62, "learning_rate": 2.0337897458389154e-07, "loss": 0.4429, "step": 22032 }, { "epoch": 2.62, "learning_rate": 2.032520068594615e-07, "loss": 0.393, "step": 22033 }, { "epoch": 2.62, "learning_rate": 2.031250771004689e-07, "loss": 0.3932, "step": 22034 }, { "epoch": 2.62, "learning_rate": 2.02998185309011e-07, "loss": 0.5835, "step": 22035 }, { "epoch": 2.62, "learning_rate": 2.0287133148718635e-07, "loss": 0.5083, "step": 22036 }, { "epoch": 2.62, "learning_rate": 2.027445156370908e-07, "loss": 0.5798, "step": 22037 }, { "epoch": 2.62, "learning_rate": 2.0261773776082127e-07, "loss": 0.4481, "step": 22038 }, { "epoch": 2.62, "learning_rate": 2.0249099786047306e-07, "loss": 0.3086, "step": 22039 }, { "epoch": 2.62, "learning_rate": 2.0236429593814223e-07, "loss": 0.4759, "step": 22040 }, { "epoch": 2.62, "learning_rate": 2.0223763199592183e-07, "loss": 0.4732, "step": 22041 }, { "epoch": 2.62, "learning_rate": 2.0211100603590633e-07, "loss": 0.5289, "step": 22042 }, { "epoch": 2.63, "learning_rate": 2.019844180601885e-07, "loss": 0.6409, "step": 22043 }, { "epoch": 2.63, "learning_rate": 2.0185786807086162e-07, "loss": 0.5221, "step": 22044 }, { "epoch": 2.63, "learning_rate": 2.0173135607001653e-07, "loss": 0.3343, "step": 22045 }, { "epoch": 2.63, "learning_rate": 2.0160488205974465e-07, "loss": 0.4376, "step": 22046 }, { "epoch": 2.63, "learning_rate": 2.0147844604213762e-07, "loss": 0.4136, "step": 22047 }, { "epoch": 2.63, "learning_rate": 2.0135204801928432e-07, "loss": 0.5239, "step": 22048 }, { "epoch": 2.63, "learning_rate": 2.0122568799327502e-07, "loss": 0.5166, "step": 22049 }, { "epoch": 2.63, "learning_rate": 2.0109936596619722e-07, "loss": 0.7349, "step": 22050 }, { "epoch": 2.63, "learning_rate": 2.0097308194014065e-07, "loss": 0.2823, "step": 22051 }, { "epoch": 2.63, "learning_rate": 2.008468359171911e-07, "loss": 0.4844, "step": 22052 }, { "epoch": 2.63, "learning_rate": 2.007206278994367e-07, "loss": 0.4107, "step": 22053 }, { "epoch": 2.63, "learning_rate": 2.0059445788896292e-07, "loss": 0.4803, "step": 22054 }, { "epoch": 2.63, "learning_rate": 2.004683258878562e-07, "loss": 0.6092, "step": 22055 }, { "epoch": 2.63, "learning_rate": 2.0034223189820096e-07, "loss": 0.5386, "step": 22056 }, { "epoch": 2.63, "learning_rate": 2.0021617592208136e-07, "loss": 0.3262, "step": 22057 }, { "epoch": 2.63, "learning_rate": 2.0009015796158105e-07, "loss": 0.3816, "step": 22058 }, { "epoch": 2.63, "learning_rate": 1.9996417801878416e-07, "loss": 0.4299, "step": 22059 }, { "epoch": 2.63, "learning_rate": 1.998382360957718e-07, "loss": 0.4676, "step": 22060 }, { "epoch": 2.63, "learning_rate": 1.9971233219462622e-07, "loss": 0.6357, "step": 22061 }, { "epoch": 2.63, "learning_rate": 1.9958646631742961e-07, "loss": 0.6267, "step": 22062 }, { "epoch": 2.63, "learning_rate": 1.9946063846626146e-07, "loss": 0.3884, "step": 22063 }, { "epoch": 2.63, "learning_rate": 1.9933484864320147e-07, "loss": 0.375, "step": 22064 }, { "epoch": 2.63, "learning_rate": 1.9920909685032962e-07, "loss": 0.4059, "step": 22065 }, { "epoch": 2.63, "learning_rate": 1.9908338308972485e-07, "loss": 0.3665, "step": 22066 }, { "epoch": 2.63, "learning_rate": 1.9895770736346464e-07, "loss": 0.4968, "step": 22067 }, { "epoch": 2.63, "learning_rate": 1.9883206967362622e-07, "loss": 0.6271, "step": 22068 }, { "epoch": 2.63, "learning_rate": 1.9870647002228739e-07, "loss": 0.5691, "step": 22069 }, { "epoch": 2.63, "learning_rate": 1.985809084115234e-07, "loss": 0.2654, "step": 22070 }, { "epoch": 2.63, "learning_rate": 1.9845538484341038e-07, "loss": 0.2895, "step": 22071 }, { "epoch": 2.63, "learning_rate": 1.983298993200225e-07, "loss": 0.3922, "step": 22072 }, { "epoch": 2.63, "learning_rate": 1.982044518434345e-07, "loss": 0.5174, "step": 22073 }, { "epoch": 2.63, "learning_rate": 1.9807904241572057e-07, "loss": 0.5335, "step": 22074 }, { "epoch": 2.63, "learning_rate": 1.9795367103895263e-07, "loss": 0.4681, "step": 22075 }, { "epoch": 2.63, "learning_rate": 1.9782833771520378e-07, "loss": 0.2678, "step": 22076 }, { "epoch": 2.63, "learning_rate": 1.977030424465462e-07, "loss": 0.3859, "step": 22077 }, { "epoch": 2.63, "learning_rate": 1.9757778523505023e-07, "loss": 0.4964, "step": 22078 }, { "epoch": 2.63, "learning_rate": 1.974525660827864e-07, "loss": 0.4865, "step": 22079 }, { "epoch": 2.63, "learning_rate": 1.9732738499182502e-07, "loss": 0.7009, "step": 22080 }, { "epoch": 2.63, "learning_rate": 1.9720224196423528e-07, "loss": 0.6014, "step": 22081 }, { "epoch": 2.63, "learning_rate": 1.9707713700208548e-07, "loss": 0.3698, "step": 22082 }, { "epoch": 2.63, "learning_rate": 1.969520701074437e-07, "loss": 0.3878, "step": 22083 }, { "epoch": 2.63, "learning_rate": 1.9682704128237806e-07, "loss": 0.4232, "step": 22084 }, { "epoch": 2.63, "learning_rate": 1.9670205052895464e-07, "loss": 0.4062, "step": 22085 }, { "epoch": 2.63, "learning_rate": 1.96577097849239e-07, "loss": 0.5591, "step": 22086 }, { "epoch": 2.63, "learning_rate": 1.964521832452973e-07, "loss": 0.6016, "step": 22087 }, { "epoch": 2.63, "learning_rate": 1.9632730671919427e-07, "loss": 0.5471, "step": 22088 }, { "epoch": 2.63, "learning_rate": 1.9620246827299462e-07, "loss": 0.3972, "step": 22089 }, { "epoch": 2.63, "learning_rate": 1.9607766790876116e-07, "loss": 0.4037, "step": 22090 }, { "epoch": 2.63, "learning_rate": 1.959529056285575e-07, "loss": 0.4997, "step": 22091 }, { "epoch": 2.63, "learning_rate": 1.9582818143444565e-07, "loss": 0.5319, "step": 22092 }, { "epoch": 2.63, "learning_rate": 1.957034953284867e-07, "loss": 0.6043, "step": 22093 }, { "epoch": 2.63, "learning_rate": 1.9557884731274236e-07, "loss": 0.5258, "step": 22094 }, { "epoch": 2.63, "learning_rate": 1.9545423738927317e-07, "loss": 0.2753, "step": 22095 }, { "epoch": 2.63, "learning_rate": 1.9532966556013915e-07, "loss": 0.3666, "step": 22096 }, { "epoch": 2.63, "learning_rate": 1.952051318273987e-07, "loss": 0.384, "step": 22097 }, { "epoch": 2.63, "learning_rate": 1.9508063619311097e-07, "loss": 0.4827, "step": 22098 }, { "epoch": 2.63, "learning_rate": 1.9495617865933402e-07, "loss": 0.5691, "step": 22099 }, { "epoch": 2.63, "learning_rate": 1.9483175922812485e-07, "loss": 0.7159, "step": 22100 }, { "epoch": 2.63, "learning_rate": 1.9470737790153959e-07, "loss": 0.2865, "step": 22101 }, { "epoch": 2.63, "learning_rate": 1.9458303468163492e-07, "loss": 0.3557, "step": 22102 }, { "epoch": 2.63, "learning_rate": 1.9445872957046613e-07, "loss": 0.4669, "step": 22103 }, { "epoch": 2.63, "learning_rate": 1.9433446257008854e-07, "loss": 0.4694, "step": 22104 }, { "epoch": 2.63, "learning_rate": 1.942102336825552e-07, "loss": 0.5668, "step": 22105 }, { "epoch": 2.63, "learning_rate": 1.9408604290992033e-07, "loss": 0.6686, "step": 22106 }, { "epoch": 2.63, "learning_rate": 1.9396189025423702e-07, "loss": 0.44, "step": 22107 }, { "epoch": 2.63, "learning_rate": 1.9383777571755663e-07, "loss": 0.3728, "step": 22108 }, { "epoch": 2.63, "learning_rate": 1.9371369930193117e-07, "loss": 0.3931, "step": 22109 }, { "epoch": 2.63, "learning_rate": 1.9358966100941206e-07, "loss": 0.4919, "step": 22110 }, { "epoch": 2.63, "learning_rate": 1.934656608420496e-07, "loss": 0.6333, "step": 22111 }, { "epoch": 2.63, "learning_rate": 1.9334169880189297e-07, "loss": 0.5838, "step": 22112 }, { "epoch": 2.63, "learning_rate": 1.9321777489099192e-07, "loss": 0.3413, "step": 22113 }, { "epoch": 2.63, "learning_rate": 1.9309388911139427e-07, "loss": 0.3534, "step": 22114 }, { "epoch": 2.63, "learning_rate": 1.9297004146514865e-07, "loss": 0.4376, "step": 22115 }, { "epoch": 2.63, "learning_rate": 1.928462319543012e-07, "loss": 0.3768, "step": 22116 }, { "epoch": 2.63, "learning_rate": 1.9272246058089916e-07, "loss": 0.5312, "step": 22117 }, { "epoch": 2.63, "learning_rate": 1.925987273469887e-07, "loss": 0.5108, "step": 22118 }, { "epoch": 2.63, "learning_rate": 1.924750322546151e-07, "loss": 0.4441, "step": 22119 }, { "epoch": 2.63, "learning_rate": 1.9235137530582287e-07, "loss": 0.3876, "step": 22120 }, { "epoch": 2.63, "learning_rate": 1.9222775650265562e-07, "loss": 0.4239, "step": 22121 }, { "epoch": 2.63, "learning_rate": 1.9210417584715702e-07, "loss": 0.4898, "step": 22122 }, { "epoch": 2.63, "learning_rate": 1.9198063334137068e-07, "loss": 0.4439, "step": 22123 }, { "epoch": 2.63, "learning_rate": 1.9185712898733777e-07, "loss": 0.5875, "step": 22124 }, { "epoch": 2.63, "learning_rate": 1.9173366278709997e-07, "loss": 0.6657, "step": 22125 }, { "epoch": 2.63, "learning_rate": 1.91610234742699e-07, "loss": 0.2754, "step": 22126 }, { "epoch": 2.64, "learning_rate": 1.914868448561741e-07, "loss": 0.4466, "step": 22127 }, { "epoch": 2.64, "learning_rate": 1.9136349312956577e-07, "loss": 0.3744, "step": 22128 }, { "epoch": 2.64, "learning_rate": 1.9124017956491213e-07, "loss": 0.481, "step": 22129 }, { "epoch": 2.64, "learning_rate": 1.9111690416425245e-07, "loss": 0.6104, "step": 22130 }, { "epoch": 2.64, "learning_rate": 1.909936669296236e-07, "loss": 0.5691, "step": 22131 }, { "epoch": 2.64, "learning_rate": 1.9087046786306325e-07, "loss": 0.3069, "step": 22132 }, { "epoch": 2.64, "learning_rate": 1.9074730696660799e-07, "loss": 0.4033, "step": 22133 }, { "epoch": 2.64, "learning_rate": 1.9062418424229373e-07, "loss": 0.395, "step": 22134 }, { "epoch": 2.64, "learning_rate": 1.9050109969215552e-07, "loss": 0.5315, "step": 22135 }, { "epoch": 2.64, "learning_rate": 1.9037805331822728e-07, "loss": 0.5513, "step": 22136 }, { "epoch": 2.64, "learning_rate": 1.9025504512254405e-07, "loss": 0.6441, "step": 22137 }, { "epoch": 2.64, "learning_rate": 1.9013207510713893e-07, "loss": 0.285, "step": 22138 }, { "epoch": 2.64, "learning_rate": 1.900091432740439e-07, "loss": 0.3199, "step": 22139 }, { "epoch": 2.64, "learning_rate": 1.8988624962529151e-07, "loss": 0.3613, "step": 22140 }, { "epoch": 2.64, "learning_rate": 1.89763394162914e-07, "loss": 0.4694, "step": 22141 }, { "epoch": 2.64, "learning_rate": 1.8964057688894117e-07, "loss": 0.5947, "step": 22142 }, { "epoch": 2.64, "learning_rate": 1.8951779780540302e-07, "loss": 0.6393, "step": 22143 }, { "epoch": 2.64, "learning_rate": 1.8939505691432964e-07, "loss": 0.5096, "step": 22144 }, { "epoch": 2.64, "learning_rate": 1.892723542177502e-07, "loss": 0.35, "step": 22145 }, { "epoch": 2.64, "learning_rate": 1.8914968971769227e-07, "loss": 0.3476, "step": 22146 }, { "epoch": 2.64, "learning_rate": 1.8902706341618393e-07, "loss": 0.4796, "step": 22147 }, { "epoch": 2.64, "learning_rate": 1.8890447531525247e-07, "loss": 0.5411, "step": 22148 }, { "epoch": 2.64, "learning_rate": 1.8878192541692404e-07, "loss": 0.5827, "step": 22149 }, { "epoch": 2.64, "learning_rate": 1.8865941372322366e-07, "loss": 0.6424, "step": 22150 }, { "epoch": 2.64, "learning_rate": 1.8853694023617725e-07, "loss": 0.2932, "step": 22151 }, { "epoch": 2.64, "learning_rate": 1.8841450495780928e-07, "loss": 0.4106, "step": 22152 }, { "epoch": 2.64, "learning_rate": 1.882921078901437e-07, "loss": 0.4023, "step": 22153 }, { "epoch": 2.64, "learning_rate": 1.8816974903520328e-07, "loss": 0.4929, "step": 22154 }, { "epoch": 2.64, "learning_rate": 1.8804742839501094e-07, "loss": 0.618, "step": 22155 }, { "epoch": 2.64, "learning_rate": 1.8792514597158888e-07, "loss": 0.5915, "step": 22156 }, { "epoch": 2.64, "learning_rate": 1.878029017669583e-07, "loss": 0.3125, "step": 22157 }, { "epoch": 2.64, "learning_rate": 1.8768069578313924e-07, "loss": 0.3387, "step": 22158 }, { "epoch": 2.64, "learning_rate": 1.875585280221523e-07, "loss": 0.3717, "step": 22159 }, { "epoch": 2.64, "learning_rate": 1.8743639848601723e-07, "loss": 0.3848, "step": 22160 }, { "epoch": 2.64, "learning_rate": 1.873143071767522e-07, "loss": 0.5622, "step": 22161 }, { "epoch": 2.64, "learning_rate": 1.8719225409637583e-07, "loss": 0.5346, "step": 22162 }, { "epoch": 2.64, "learning_rate": 1.8707023924690594e-07, "loss": 0.385, "step": 22163 }, { "epoch": 2.64, "learning_rate": 1.8694826263035897e-07, "loss": 0.2936, "step": 22164 }, { "epoch": 2.64, "learning_rate": 1.8682632424875086e-07, "loss": 0.3323, "step": 22165 }, { "epoch": 2.64, "learning_rate": 1.86704424104098e-07, "loss": 0.4233, "step": 22166 }, { "epoch": 2.64, "learning_rate": 1.8658256219841463e-07, "loss": 0.4543, "step": 22167 }, { "epoch": 2.64, "learning_rate": 1.8646073853371638e-07, "loss": 0.6754, "step": 22168 }, { "epoch": 2.64, "learning_rate": 1.8633895311201577e-07, "loss": 0.5648, "step": 22169 }, { "epoch": 2.64, "learning_rate": 1.8621720593532677e-07, "loss": 0.2697, "step": 22170 }, { "epoch": 2.64, "learning_rate": 1.8609549700566165e-07, "loss": 0.3542, "step": 22171 }, { "epoch": 2.64, "learning_rate": 1.8597382632503158e-07, "loss": 0.4631, "step": 22172 }, { "epoch": 2.64, "learning_rate": 1.858521938954483e-07, "loss": 0.4494, "step": 22173 }, { "epoch": 2.64, "learning_rate": 1.8573059971892238e-07, "loss": 0.4907, "step": 22174 }, { "epoch": 2.64, "learning_rate": 1.8560904379746418e-07, "loss": 0.5556, "step": 22175 }, { "epoch": 2.64, "learning_rate": 1.854875261330824e-07, "loss": 0.2754, "step": 22176 }, { "epoch": 2.64, "learning_rate": 1.853660467277865e-07, "loss": 0.4668, "step": 22177 }, { "epoch": 2.64, "learning_rate": 1.8524460558358353e-07, "loss": 0.4468, "step": 22178 }, { "epoch": 2.64, "learning_rate": 1.8512320270248185e-07, "loss": 0.4428, "step": 22179 }, { "epoch": 2.64, "learning_rate": 1.8500183808648765e-07, "loss": 0.5655, "step": 22180 }, { "epoch": 2.64, "learning_rate": 1.8488051173760706e-07, "loss": 0.5425, "step": 22181 }, { "epoch": 2.64, "learning_rate": 1.8475922365784606e-07, "loss": 0.47, "step": 22182 }, { "epoch": 2.64, "learning_rate": 1.846379738492096e-07, "loss": 0.3254, "step": 22183 }, { "epoch": 2.64, "learning_rate": 1.8451676231370148e-07, "loss": 0.4001, "step": 22184 }, { "epoch": 2.64, "learning_rate": 1.8439558905332583e-07, "loss": 0.408, "step": 22185 }, { "epoch": 2.64, "learning_rate": 1.84274454070085e-07, "loss": 0.4791, "step": 22186 }, { "epoch": 2.64, "learning_rate": 1.8415335736598204e-07, "loss": 0.6212, "step": 22187 }, { "epoch": 2.64, "learning_rate": 1.8403229894301823e-07, "loss": 0.4697, "step": 22188 }, { "epoch": 2.64, "learning_rate": 1.8391127880319464e-07, "loss": 0.4658, "step": 22189 }, { "epoch": 2.64, "learning_rate": 1.8379029694851252e-07, "loss": 0.4162, "step": 22190 }, { "epoch": 2.64, "learning_rate": 1.8366935338097081e-07, "loss": 0.5014, "step": 22191 }, { "epoch": 2.64, "learning_rate": 1.8354844810256927e-07, "loss": 0.5265, "step": 22192 }, { "epoch": 2.64, "learning_rate": 1.8342758111530578e-07, "loss": 0.6265, "step": 22193 }, { "epoch": 2.64, "learning_rate": 1.8330675242117929e-07, "loss": 0.4937, "step": 22194 }, { "epoch": 2.64, "learning_rate": 1.8318596202218626e-07, "loss": 0.3578, "step": 22195 }, { "epoch": 2.64, "learning_rate": 1.830652099203234e-07, "loss": 0.4054, "step": 22196 }, { "epoch": 2.64, "learning_rate": 1.8294449611758718e-07, "loss": 0.3415, "step": 22197 }, { "epoch": 2.64, "learning_rate": 1.8282382061597327e-07, "loss": 0.4983, "step": 22198 }, { "epoch": 2.64, "learning_rate": 1.8270318341747612e-07, "loss": 0.6078, "step": 22199 }, { "epoch": 2.64, "learning_rate": 1.8258258452408916e-07, "loss": 0.6142, "step": 22200 }, { "epoch": 2.64, "learning_rate": 1.8246202393780665e-07, "loss": 0.3786, "step": 22201 }, { "epoch": 2.64, "learning_rate": 1.823415016606217e-07, "loss": 0.3851, "step": 22202 }, { "epoch": 2.64, "learning_rate": 1.8222101769452605e-07, "loss": 0.4696, "step": 22203 }, { "epoch": 2.64, "learning_rate": 1.8210057204151116e-07, "loss": 0.5086, "step": 22204 }, { "epoch": 2.64, "learning_rate": 1.8198016470356878e-07, "loss": 0.5594, "step": 22205 }, { "epoch": 2.64, "learning_rate": 1.818597956826887e-07, "loss": 0.61, "step": 22206 }, { "epoch": 2.64, "learning_rate": 1.8173946498086042e-07, "loss": 0.3642, "step": 22207 }, { "epoch": 2.64, "learning_rate": 1.8161917260007323e-07, "loss": 0.4616, "step": 22208 }, { "epoch": 2.64, "learning_rate": 1.8149891854231604e-07, "loss": 0.4537, "step": 22209 }, { "epoch": 2.64, "learning_rate": 1.8137870280957594e-07, "loss": 0.4198, "step": 22210 }, { "epoch": 2.65, "learning_rate": 1.8125852540384042e-07, "loss": 0.4962, "step": 22211 }, { "epoch": 2.65, "learning_rate": 1.811383863270963e-07, "loss": 0.6224, "step": 22212 }, { "epoch": 2.65, "learning_rate": 1.810182855813289e-07, "loss": 0.4218, "step": 22213 }, { "epoch": 2.65, "learning_rate": 1.808982231685244e-07, "loss": 0.3015, "step": 22214 }, { "epoch": 2.65, "learning_rate": 1.8077819909066623e-07, "loss": 0.3572, "step": 22215 }, { "epoch": 2.65, "learning_rate": 1.806582133497392e-07, "loss": 0.4202, "step": 22216 }, { "epoch": 2.65, "learning_rate": 1.8053826594772672e-07, "loss": 0.4898, "step": 22217 }, { "epoch": 2.65, "learning_rate": 1.8041835688661108e-07, "loss": 0.6235, "step": 22218 }, { "epoch": 2.65, "learning_rate": 1.802984861683746e-07, "loss": 0.5695, "step": 22219 }, { "epoch": 2.65, "learning_rate": 1.80178653794999e-07, "loss": 0.3186, "step": 22220 }, { "epoch": 2.65, "learning_rate": 1.8005885976846522e-07, "loss": 0.3518, "step": 22221 }, { "epoch": 2.65, "learning_rate": 1.799391040907525e-07, "loss": 0.4145, "step": 22222 }, { "epoch": 2.65, "learning_rate": 1.798193867638412e-07, "loss": 0.5159, "step": 22223 }, { "epoch": 2.65, "learning_rate": 1.796997077897103e-07, "loss": 0.6019, "step": 22224 }, { "epoch": 2.65, "learning_rate": 1.7958006717033767e-07, "loss": 0.5843, "step": 22225 }, { "epoch": 2.65, "learning_rate": 1.7946046490770143e-07, "loss": 0.3715, "step": 22226 }, { "epoch": 2.65, "learning_rate": 1.793409010037786e-07, "loss": 0.3678, "step": 22227 }, { "epoch": 2.65, "learning_rate": 1.7922137546054542e-07, "loss": 0.4813, "step": 22228 }, { "epoch": 2.65, "learning_rate": 1.7910188827997721e-07, "loss": 0.3911, "step": 22229 }, { "epoch": 2.65, "learning_rate": 1.7898243946404937e-07, "loss": 0.5304, "step": 22230 }, { "epoch": 2.65, "learning_rate": 1.7886302901473695e-07, "loss": 0.5807, "step": 22231 }, { "epoch": 2.65, "learning_rate": 1.787436569340134e-07, "loss": 0.2731, "step": 22232 }, { "epoch": 2.65, "learning_rate": 1.7862432322385186e-07, "loss": 0.2436, "step": 22233 }, { "epoch": 2.65, "learning_rate": 1.785050278862252e-07, "loss": 0.3817, "step": 22234 }, { "epoch": 2.65, "learning_rate": 1.7838577092310543e-07, "loss": 0.4843, "step": 22235 }, { "epoch": 2.65, "learning_rate": 1.7826655233646323e-07, "loss": 0.51, "step": 22236 }, { "epoch": 2.65, "learning_rate": 1.7814737212826976e-07, "loss": 0.6082, "step": 22237 }, { "epoch": 2.65, "learning_rate": 1.7802823030049486e-07, "loss": 0.2851, "step": 22238 }, { "epoch": 2.65, "learning_rate": 1.7790912685510863e-07, "loss": 0.3246, "step": 22239 }, { "epoch": 2.65, "learning_rate": 1.7779006179407922e-07, "loss": 0.4481, "step": 22240 }, { "epoch": 2.65, "learning_rate": 1.7767103511937473e-07, "loss": 0.4396, "step": 22241 }, { "epoch": 2.65, "learning_rate": 1.7755204683296334e-07, "loss": 0.5063, "step": 22242 }, { "epoch": 2.65, "learning_rate": 1.7743309693681127e-07, "loss": 0.5792, "step": 22243 }, { "epoch": 2.65, "learning_rate": 1.773141854328847e-07, "loss": 0.3978, "step": 22244 }, { "epoch": 2.65, "learning_rate": 1.7719531232314957e-07, "loss": 0.3403, "step": 22245 }, { "epoch": 2.65, "learning_rate": 1.7707647760957097e-07, "loss": 0.419, "step": 22246 }, { "epoch": 2.65, "learning_rate": 1.7695768129411345e-07, "loss": 0.3778, "step": 22247 }, { "epoch": 2.65, "learning_rate": 1.7683892337873987e-07, "loss": 0.4576, "step": 22248 }, { "epoch": 2.65, "learning_rate": 1.7672020386541423e-07, "loss": 0.6546, "step": 22249 }, { "epoch": 2.65, "learning_rate": 1.7660152275609853e-07, "loss": 0.5903, "step": 22250 }, { "epoch": 2.65, "learning_rate": 1.7648288005275432e-07, "loss": 0.3194, "step": 22251 }, { "epoch": 2.65, "learning_rate": 1.7636427575734306e-07, "loss": 0.4193, "step": 22252 }, { "epoch": 2.65, "learning_rate": 1.762457098718251e-07, "loss": 0.4161, "step": 22253 }, { "epoch": 2.65, "learning_rate": 1.7612718239816112e-07, "loss": 0.4809, "step": 22254 }, { "epoch": 2.65, "learning_rate": 1.7600869333830927e-07, "loss": 0.6399, "step": 22255 }, { "epoch": 2.65, "learning_rate": 1.758902426942294e-07, "loss": 0.689, "step": 22256 }, { "epoch": 2.65, "learning_rate": 1.7577183046787793e-07, "loss": 0.3835, "step": 22257 }, { "epoch": 2.65, "learning_rate": 1.756534566612139e-07, "loss": 0.3127, "step": 22258 }, { "epoch": 2.65, "learning_rate": 1.755351212761927e-07, "loss": 0.3929, "step": 22259 }, { "epoch": 2.65, "learning_rate": 1.754168243147711e-07, "loss": 0.3185, "step": 22260 }, { "epoch": 2.65, "learning_rate": 1.7529856577890442e-07, "loss": 0.5154, "step": 22261 }, { "epoch": 2.65, "learning_rate": 1.751803456705481e-07, "loss": 0.5988, "step": 22262 }, { "epoch": 2.65, "learning_rate": 1.7506216399165527e-07, "loss": 0.4617, "step": 22263 }, { "epoch": 2.65, "learning_rate": 1.7494402074417993e-07, "loss": 0.2665, "step": 22264 }, { "epoch": 2.65, "learning_rate": 1.7482591593007496e-07, "loss": 0.3907, "step": 22265 }, { "epoch": 2.65, "learning_rate": 1.74707849551293e-07, "loss": 0.503, "step": 22266 }, { "epoch": 2.65, "learning_rate": 1.7458982160978494e-07, "loss": 0.4953, "step": 22267 }, { "epoch": 2.65, "learning_rate": 1.744718321075023e-07, "loss": 0.6543, "step": 22268 }, { "epoch": 2.65, "learning_rate": 1.74353881046396e-07, "loss": 0.5207, "step": 22269 }, { "epoch": 2.65, "learning_rate": 1.742359684284145e-07, "loss": 0.2914, "step": 22270 }, { "epoch": 2.65, "learning_rate": 1.741180942555082e-07, "loss": 0.3503, "step": 22271 }, { "epoch": 2.65, "learning_rate": 1.7400025852962438e-07, "loss": 0.4141, "step": 22272 }, { "epoch": 2.65, "learning_rate": 1.738824612527118e-07, "loss": 0.5176, "step": 22273 }, { "epoch": 2.65, "learning_rate": 1.7376470242671722e-07, "loss": 0.5253, "step": 22274 }, { "epoch": 2.65, "learning_rate": 1.7364698205358715e-07, "loss": 0.5022, "step": 22275 }, { "epoch": 2.65, "learning_rate": 1.7352930013526753e-07, "loss": 0.4925, "step": 22276 }, { "epoch": 2.65, "learning_rate": 1.734116566737043e-07, "loss": 0.3911, "step": 22277 }, { "epoch": 2.65, "learning_rate": 1.7329405167084175e-07, "loss": 0.4225, "step": 22278 }, { "epoch": 2.65, "learning_rate": 1.73176485128623e-07, "loss": 0.535, "step": 22279 }, { "epoch": 2.65, "learning_rate": 1.730589570489924e-07, "loss": 0.4203, "step": 22280 }, { "epoch": 2.65, "learning_rate": 1.7294146743389306e-07, "loss": 0.4723, "step": 22281 }, { "epoch": 2.65, "learning_rate": 1.7282401628526568e-07, "loss": 0.2999, "step": 22282 }, { "epoch": 2.65, "learning_rate": 1.7270660360505287e-07, "loss": 0.392, "step": 22283 }, { "epoch": 2.65, "learning_rate": 1.7258922939519558e-07, "loss": 0.4223, "step": 22284 }, { "epoch": 2.65, "learning_rate": 1.7247189365763362e-07, "loss": 0.4728, "step": 22285 }, { "epoch": 2.65, "learning_rate": 1.7235459639430601e-07, "loss": 0.478, "step": 22286 }, { "epoch": 2.65, "learning_rate": 1.7223733760715233e-07, "loss": 0.5693, "step": 22287 }, { "epoch": 2.65, "learning_rate": 1.721201172981113e-07, "loss": 0.444, "step": 22288 }, { "epoch": 2.65, "learning_rate": 1.720029354691194e-07, "loss": 0.3661, "step": 22289 }, { "epoch": 2.65, "learning_rate": 1.7188579212211454e-07, "loss": 0.4466, "step": 22290 }, { "epoch": 2.65, "learning_rate": 1.7176868725903322e-07, "loss": 0.5765, "step": 22291 }, { "epoch": 2.65, "learning_rate": 1.7165162088181085e-07, "loss": 0.4799, "step": 22292 }, { "epoch": 2.65, "learning_rate": 1.71534592992382e-07, "loss": 0.5062, "step": 22293 }, { "epoch": 2.65, "learning_rate": 1.714176035926818e-07, "loss": 0.5265, "step": 22294 }, { "epoch": 2.66, "learning_rate": 1.713006526846439e-07, "loss": 0.3494, "step": 22295 }, { "epoch": 2.66, "learning_rate": 1.7118374027020212e-07, "loss": 0.4779, "step": 22296 }, { "epoch": 2.66, "learning_rate": 1.7106686635128822e-07, "loss": 0.4128, "step": 22297 }, { "epoch": 2.66, "learning_rate": 1.7095003092983426e-07, "loss": 0.5408, "step": 22298 }, { "epoch": 2.66, "learning_rate": 1.708332340077723e-07, "loss": 0.6089, "step": 22299 }, { "epoch": 2.66, "learning_rate": 1.707164755870322e-07, "loss": 0.6138, "step": 22300 }, { "epoch": 2.66, "learning_rate": 1.7059975566954407e-07, "loss": 0.4254, "step": 22301 }, { "epoch": 2.66, "learning_rate": 1.704830742572372e-07, "loss": 0.4889, "step": 22302 }, { "epoch": 2.66, "learning_rate": 1.703664313520409e-07, "loss": 0.4274, "step": 22303 }, { "epoch": 2.66, "learning_rate": 1.7024982695588278e-07, "loss": 0.5129, "step": 22304 }, { "epoch": 2.66, "learning_rate": 1.7013326107069046e-07, "loss": 0.5863, "step": 22305 }, { "epoch": 2.66, "learning_rate": 1.7001673369839128e-07, "loss": 0.6122, "step": 22306 }, { "epoch": 2.66, "learning_rate": 1.6990024484091066e-07, "loss": 0.3258, "step": 22307 }, { "epoch": 2.66, "learning_rate": 1.6978379450017457e-07, "loss": 0.3196, "step": 22308 }, { "epoch": 2.66, "learning_rate": 1.6966738267810757e-07, "loss": 0.3782, "step": 22309 }, { "epoch": 2.66, "learning_rate": 1.6955100937663422e-07, "loss": 0.478, "step": 22310 }, { "epoch": 2.66, "learning_rate": 1.6943467459767883e-07, "loss": 0.5481, "step": 22311 }, { "epoch": 2.66, "learning_rate": 1.6931837834316344e-07, "loss": 0.6705, "step": 22312 }, { "epoch": 2.66, "learning_rate": 1.69202120615011e-07, "loss": 0.3941, "step": 22313 }, { "epoch": 2.66, "learning_rate": 1.690859014151433e-07, "loss": 0.3598, "step": 22314 }, { "epoch": 2.66, "learning_rate": 1.6896972074548074e-07, "loss": 0.4224, "step": 22315 }, { "epoch": 2.66, "learning_rate": 1.6885357860794428e-07, "loss": 0.3728, "step": 22316 }, { "epoch": 2.66, "learning_rate": 1.6873747500445377e-07, "loss": 0.5592, "step": 22317 }, { "epoch": 2.66, "learning_rate": 1.6862140993692878e-07, "loss": 0.658, "step": 22318 }, { "epoch": 2.66, "learning_rate": 1.6850538340728695e-07, "loss": 0.5237, "step": 22319 }, { "epoch": 2.66, "learning_rate": 1.6838939541744731e-07, "loss": 0.2612, "step": 22320 }, { "epoch": 2.66, "learning_rate": 1.6827344596932639e-07, "loss": 0.4099, "step": 22321 }, { "epoch": 2.66, "learning_rate": 1.6815753506484122e-07, "loss": 0.4552, "step": 22322 }, { "epoch": 2.66, "learning_rate": 1.6804166270590726e-07, "loss": 0.557, "step": 22323 }, { "epoch": 2.66, "learning_rate": 1.679258288944402e-07, "loss": 0.6158, "step": 22324 }, { "epoch": 2.66, "learning_rate": 1.6781003363235515e-07, "loss": 0.4857, "step": 22325 }, { "epoch": 2.66, "learning_rate": 1.6769427692156614e-07, "loss": 0.3494, "step": 22326 }, { "epoch": 2.66, "learning_rate": 1.675785587639861e-07, "loss": 0.4467, "step": 22327 }, { "epoch": 2.66, "learning_rate": 1.6746287916152847e-07, "loss": 0.3412, "step": 22328 }, { "epoch": 2.66, "learning_rate": 1.673472381161051e-07, "loss": 0.5846, "step": 22329 }, { "epoch": 2.66, "learning_rate": 1.6723163562962775e-07, "loss": 0.5366, "step": 22330 }, { "epoch": 2.66, "learning_rate": 1.6711607170400656e-07, "loss": 0.5554, "step": 22331 }, { "epoch": 2.66, "learning_rate": 1.670005463411528e-07, "loss": 0.2775, "step": 22332 }, { "epoch": 2.66, "learning_rate": 1.6688505954297608e-07, "loss": 0.3148, "step": 22333 }, { "epoch": 2.66, "learning_rate": 1.667696113113848e-07, "loss": 0.354, "step": 22334 }, { "epoch": 2.66, "learning_rate": 1.6665420164828778e-07, "loss": 0.4203, "step": 22335 }, { "epoch": 2.66, "learning_rate": 1.6653883055559233e-07, "loss": 0.5022, "step": 22336 }, { "epoch": 2.66, "learning_rate": 1.6642349803520614e-07, "loss": 0.6012, "step": 22337 }, { "epoch": 2.66, "learning_rate": 1.6630820408903514e-07, "loss": 0.4326, "step": 22338 }, { "epoch": 2.66, "learning_rate": 1.6619294871898505e-07, "loss": 0.3846, "step": 22339 }, { "epoch": 2.66, "learning_rate": 1.660777319269616e-07, "loss": 0.4753, "step": 22340 }, { "epoch": 2.66, "learning_rate": 1.659625537148693e-07, "loss": 0.4599, "step": 22341 }, { "epoch": 2.66, "learning_rate": 1.658474140846117e-07, "loss": 0.6622, "step": 22342 }, { "epoch": 2.66, "learning_rate": 1.6573231303809196e-07, "loss": 0.5598, "step": 22343 }, { "epoch": 2.66, "learning_rate": 1.65617250577213e-07, "loss": 0.463, "step": 22344 }, { "epoch": 2.66, "learning_rate": 1.6550222670387723e-07, "loss": 0.3434, "step": 22345 }, { "epoch": 2.66, "learning_rate": 1.6538724141998503e-07, "loss": 0.517, "step": 22346 }, { "epoch": 2.66, "learning_rate": 1.6527229472743738e-07, "loss": 0.5078, "step": 22347 }, { "epoch": 2.66, "learning_rate": 1.651573866281353e-07, "loss": 0.4568, "step": 22348 }, { "epoch": 2.66, "learning_rate": 1.650425171239775e-07, "loss": 0.6155, "step": 22349 }, { "epoch": 2.66, "learning_rate": 1.6492768621686222e-07, "loss": 0.5934, "step": 22350 }, { "epoch": 2.66, "learning_rate": 1.648128939086885e-07, "loss": 0.3719, "step": 22351 }, { "epoch": 2.66, "learning_rate": 1.6469814020135393e-07, "loss": 0.4265, "step": 22352 }, { "epoch": 2.66, "learning_rate": 1.6458342509675458e-07, "loss": 0.396, "step": 22353 }, { "epoch": 2.66, "learning_rate": 1.644687485967869e-07, "loss": 0.4631, "step": 22354 }, { "epoch": 2.66, "learning_rate": 1.6435411070334722e-07, "loss": 0.6315, "step": 22355 }, { "epoch": 2.66, "learning_rate": 1.642395114183301e-07, "loss": 0.5747, "step": 22356 }, { "epoch": 2.66, "learning_rate": 1.6412495074363016e-07, "loss": 0.2529, "step": 22357 }, { "epoch": 2.66, "learning_rate": 1.6401042868114002e-07, "loss": 0.3086, "step": 22358 }, { "epoch": 2.66, "learning_rate": 1.6389594523275376e-07, "loss": 0.4341, "step": 22359 }, { "epoch": 2.66, "learning_rate": 1.63781500400364e-07, "loss": 0.4341, "step": 22360 }, { "epoch": 2.66, "learning_rate": 1.6366709418586145e-07, "loss": 0.6166, "step": 22361 }, { "epoch": 2.66, "learning_rate": 1.6355272659113796e-07, "loss": 0.7675, "step": 22362 }, { "epoch": 2.66, "learning_rate": 1.634383976180845e-07, "loss": 0.3953, "step": 22363 }, { "epoch": 2.66, "learning_rate": 1.633241072685901e-07, "loss": 0.4142, "step": 22364 }, { "epoch": 2.66, "learning_rate": 1.632098555445441e-07, "loss": 0.4229, "step": 22365 }, { "epoch": 2.66, "learning_rate": 1.6309564244783526e-07, "loss": 0.416, "step": 22366 }, { "epoch": 2.66, "learning_rate": 1.629814679803518e-07, "loss": 0.5496, "step": 22367 }, { "epoch": 2.66, "learning_rate": 1.6286733214398026e-07, "loss": 0.6469, "step": 22368 }, { "epoch": 2.66, "learning_rate": 1.6275323494060802e-07, "loss": 0.5042, "step": 22369 }, { "epoch": 2.66, "learning_rate": 1.6263917637212135e-07, "loss": 0.2837, "step": 22370 }, { "epoch": 2.66, "learning_rate": 1.625251564404051e-07, "loss": 0.4887, "step": 22371 }, { "epoch": 2.66, "learning_rate": 1.6241117514734367e-07, "loss": 0.457, "step": 22372 }, { "epoch": 2.66, "learning_rate": 1.6229723249482189e-07, "loss": 0.5359, "step": 22373 }, { "epoch": 2.66, "learning_rate": 1.6218332848472296e-07, "loss": 0.4911, "step": 22374 }, { "epoch": 2.66, "learning_rate": 1.6206946311892984e-07, "loss": 0.5955, "step": 22375 }, { "epoch": 2.66, "learning_rate": 1.6195563639932466e-07, "loss": 0.348, "step": 22376 }, { "epoch": 2.66, "learning_rate": 1.618418483277892e-07, "loss": 0.4098, "step": 22377 }, { "epoch": 2.66, "learning_rate": 1.6172809890620422e-07, "loss": 0.4181, "step": 22378 }, { "epoch": 2.67, "learning_rate": 1.6161438813644958e-07, "loss": 0.413, "step": 22379 }, { "epoch": 2.67, "learning_rate": 1.6150071602040517e-07, "loss": 0.5409, "step": 22380 }, { "epoch": 2.67, "learning_rate": 1.6138708255995034e-07, "loss": 0.6815, "step": 22381 }, { "epoch": 2.67, "learning_rate": 1.6127348775696355e-07, "loss": 0.3426, "step": 22382 }, { "epoch": 2.67, "learning_rate": 1.6115993161332193e-07, "loss": 0.321, "step": 22383 }, { "epoch": 2.67, "learning_rate": 1.6104641413090287e-07, "loss": 0.4086, "step": 22384 }, { "epoch": 2.67, "learning_rate": 1.6093293531158322e-07, "loss": 0.4138, "step": 22385 }, { "epoch": 2.67, "learning_rate": 1.6081949515723838e-07, "loss": 0.5686, "step": 22386 }, { "epoch": 2.67, "learning_rate": 1.60706093669743e-07, "loss": 0.6643, "step": 22387 }, { "epoch": 2.67, "learning_rate": 1.6059273085097248e-07, "loss": 0.45, "step": 22388 }, { "epoch": 2.67, "learning_rate": 1.604794067028001e-07, "loss": 0.4113, "step": 22389 }, { "epoch": 2.67, "learning_rate": 1.6036612122710016e-07, "loss": 0.4041, "step": 22390 }, { "epoch": 2.67, "learning_rate": 1.6025287442574395e-07, "loss": 0.4854, "step": 22391 }, { "epoch": 2.67, "learning_rate": 1.601396663006044e-07, "loss": 0.6046, "step": 22392 }, { "epoch": 2.67, "learning_rate": 1.6002649685355254e-07, "loss": 0.5903, "step": 22393 }, { "epoch": 2.67, "learning_rate": 1.599133660864588e-07, "loss": 0.488, "step": 22394 }, { "epoch": 2.67, "learning_rate": 1.5980027400119337e-07, "loss": 0.322, "step": 22395 }, { "epoch": 2.67, "learning_rate": 1.5968722059962582e-07, "loss": 0.5104, "step": 22396 }, { "epoch": 2.67, "learning_rate": 1.5957420588362498e-07, "loss": 0.4458, "step": 22397 }, { "epoch": 2.67, "learning_rate": 1.594612298550588e-07, "loss": 0.4791, "step": 22398 }, { "epoch": 2.67, "learning_rate": 1.5934829251579519e-07, "loss": 0.4848, "step": 22399 }, { "epoch": 2.67, "learning_rate": 1.592353938677002e-07, "loss": 0.5814, "step": 22400 }, { "epoch": 2.67, "learning_rate": 1.591225339126412e-07, "loss": 0.2924, "step": 22401 }, { "epoch": 2.67, "learning_rate": 1.5900971265248256e-07, "loss": 0.3215, "step": 22402 }, { "epoch": 2.67, "learning_rate": 1.588969300890897e-07, "loss": 0.4777, "step": 22403 }, { "epoch": 2.67, "learning_rate": 1.5878418622432724e-07, "loss": 0.5133, "step": 22404 }, { "epoch": 2.67, "learning_rate": 1.5867148106005898e-07, "loss": 0.5438, "step": 22405 }, { "epoch": 2.67, "learning_rate": 1.5855881459814759e-07, "loss": 0.5843, "step": 22406 }, { "epoch": 2.67, "learning_rate": 1.5844618684045492e-07, "loss": 0.351, "step": 22407 }, { "epoch": 2.67, "learning_rate": 1.5833359778884339e-07, "loss": 0.3346, "step": 22408 }, { "epoch": 2.67, "learning_rate": 1.5822104744517424e-07, "loss": 0.4426, "step": 22409 }, { "epoch": 2.67, "learning_rate": 1.581085358113074e-07, "loss": 0.4449, "step": 22410 }, { "epoch": 2.67, "learning_rate": 1.579960628891028e-07, "loss": 0.4774, "step": 22411 }, { "epoch": 2.67, "learning_rate": 1.5788362868042028e-07, "loss": 0.5455, "step": 22412 }, { "epoch": 2.67, "learning_rate": 1.5777123318711756e-07, "loss": 0.2949, "step": 22413 }, { "epoch": 2.67, "learning_rate": 1.5765887641105344e-07, "loss": 0.3451, "step": 22414 }, { "epoch": 2.67, "learning_rate": 1.575465583540839e-07, "loss": 0.4003, "step": 22415 }, { "epoch": 2.67, "learning_rate": 1.574342790180669e-07, "loss": 0.4349, "step": 22416 }, { "epoch": 2.67, "learning_rate": 1.5732203840485739e-07, "loss": 0.5407, "step": 22417 }, { "epoch": 2.67, "learning_rate": 1.5720983651631105e-07, "loss": 0.6284, "step": 22418 }, { "epoch": 2.67, "learning_rate": 1.5709767335428283e-07, "loss": 0.5659, "step": 22419 }, { "epoch": 2.67, "learning_rate": 1.5698554892062706e-07, "loss": 0.3362, "step": 22420 }, { "epoch": 2.67, "learning_rate": 1.568734632171967e-07, "loss": 0.4056, "step": 22421 }, { "epoch": 2.67, "learning_rate": 1.5676141624584418e-07, "loss": 0.3764, "step": 22422 }, { "epoch": 2.67, "learning_rate": 1.5664940800842188e-07, "loss": 0.4354, "step": 22423 }, { "epoch": 2.67, "learning_rate": 1.5653743850678222e-07, "loss": 0.5712, "step": 22424 }, { "epoch": 2.67, "learning_rate": 1.5642550774277482e-07, "loss": 0.5552, "step": 22425 }, { "epoch": 2.67, "learning_rate": 1.5631361571825044e-07, "loss": 0.3038, "step": 22426 }, { "epoch": 2.67, "learning_rate": 1.5620176243505896e-07, "loss": 0.4451, "step": 22427 }, { "epoch": 2.67, "learning_rate": 1.5608994789504895e-07, "loss": 0.3708, "step": 22428 }, { "epoch": 2.67, "learning_rate": 1.5597817210006862e-07, "loss": 0.424, "step": 22429 }, { "epoch": 2.67, "learning_rate": 1.5586643505196565e-07, "loss": 0.6415, "step": 22430 }, { "epoch": 2.67, "learning_rate": 1.5575473675258745e-07, "loss": 0.5612, "step": 22431 }, { "epoch": 2.67, "learning_rate": 1.556430772037798e-07, "loss": 0.3322, "step": 22432 }, { "epoch": 2.67, "learning_rate": 1.5553145640738897e-07, "loss": 0.4095, "step": 22433 }, { "epoch": 2.67, "learning_rate": 1.5541987436526019e-07, "loss": 0.4515, "step": 22434 }, { "epoch": 2.67, "learning_rate": 1.5530833107923748e-07, "loss": 0.4341, "step": 22435 }, { "epoch": 2.67, "learning_rate": 1.551968265511647e-07, "loss": 0.4737, "step": 22436 }, { "epoch": 2.67, "learning_rate": 1.5508536078288478e-07, "loss": 0.6365, "step": 22437 }, { "epoch": 2.67, "learning_rate": 1.5497393377624097e-07, "loss": 0.3571, "step": 22438 }, { "epoch": 2.67, "learning_rate": 1.5486254553307488e-07, "loss": 0.3237, "step": 22439 }, { "epoch": 2.67, "learning_rate": 1.5475119605522727e-07, "loss": 0.4411, "step": 22440 }, { "epoch": 2.67, "learning_rate": 1.5463988534453938e-07, "loss": 0.4058, "step": 22441 }, { "epoch": 2.67, "learning_rate": 1.545286134028512e-07, "loss": 0.4709, "step": 22442 }, { "epoch": 2.67, "learning_rate": 1.5441738023200203e-07, "loss": 0.6102, "step": 22443 }, { "epoch": 2.67, "learning_rate": 1.543061858338299e-07, "loss": 0.5148, "step": 22444 }, { "epoch": 2.67, "learning_rate": 1.541950302101733e-07, "loss": 0.3548, "step": 22445 }, { "epoch": 2.67, "learning_rate": 1.540839133628702e-07, "loss": 0.3487, "step": 22446 }, { "epoch": 2.67, "learning_rate": 1.539728352937564e-07, "loss": 0.5493, "step": 22447 }, { "epoch": 2.67, "learning_rate": 1.5386179600466843e-07, "loss": 0.4544, "step": 22448 }, { "epoch": 2.67, "learning_rate": 1.537507954974421e-07, "loss": 0.6641, "step": 22449 }, { "epoch": 2.67, "learning_rate": 1.53639833773912e-07, "loss": 0.6449, "step": 22450 }, { "epoch": 2.67, "learning_rate": 1.5352891083591227e-07, "loss": 0.3107, "step": 22451 }, { "epoch": 2.67, "learning_rate": 1.5341802668527612e-07, "loss": 0.4682, "step": 22452 }, { "epoch": 2.67, "learning_rate": 1.5330718132383709e-07, "loss": 0.4614, "step": 22453 }, { "epoch": 2.67, "learning_rate": 1.5319637475342763e-07, "loss": 0.5513, "step": 22454 }, { "epoch": 2.67, "learning_rate": 1.5308560697587875e-07, "loss": 0.5895, "step": 22455 }, { "epoch": 2.67, "learning_rate": 1.5297487799302178e-07, "loss": 0.6537, "step": 22456 }, { "epoch": 2.67, "learning_rate": 1.528641878066872e-07, "loss": 0.3024, "step": 22457 }, { "epoch": 2.67, "learning_rate": 1.5275353641870438e-07, "loss": 0.3427, "step": 22458 }, { "epoch": 2.67, "learning_rate": 1.5264292383090212e-07, "loss": 0.3897, "step": 22459 }, { "epoch": 2.67, "learning_rate": 1.525323500451095e-07, "loss": 0.4102, "step": 22460 }, { "epoch": 2.67, "learning_rate": 1.5242181506315457e-07, "loss": 0.571, "step": 22461 }, { "epoch": 2.68, "learning_rate": 1.5231131888686357e-07, "loss": 0.6665, "step": 22462 }, { "epoch": 2.68, "learning_rate": 1.522008615180637e-07, "loss": 0.4374, "step": 22463 }, { "epoch": 2.68, "learning_rate": 1.5209044295858038e-07, "loss": 0.328, "step": 22464 }, { "epoch": 2.68, "learning_rate": 1.5198006321023917e-07, "loss": 0.4197, "step": 22465 }, { "epoch": 2.68, "learning_rate": 1.518697222748644e-07, "loss": 0.5208, "step": 22466 }, { "epoch": 2.68, "learning_rate": 1.5175942015428019e-07, "loss": 0.5531, "step": 22467 }, { "epoch": 2.68, "learning_rate": 1.516491568503095e-07, "loss": 0.5559, "step": 22468 }, { "epoch": 2.68, "learning_rate": 1.515389323647759e-07, "loss": 0.5026, "step": 22469 }, { "epoch": 2.68, "learning_rate": 1.514287466995004e-07, "loss": 0.3201, "step": 22470 }, { "epoch": 2.68, "learning_rate": 1.513185998563052e-07, "loss": 0.5121, "step": 22471 }, { "epoch": 2.68, "learning_rate": 1.5120849183701019e-07, "loss": 0.3698, "step": 22472 }, { "epoch": 2.68, "learning_rate": 1.5109842264343615e-07, "loss": 0.4572, "step": 22473 }, { "epoch": 2.68, "learning_rate": 1.509883922774019e-07, "loss": 0.6154, "step": 22474 }, { "epoch": 2.68, "learning_rate": 1.5087840074072685e-07, "loss": 0.5139, "step": 22475 }, { "epoch": 2.68, "learning_rate": 1.507684480352292e-07, "loss": 0.2816, "step": 22476 }, { "epoch": 2.68, "learning_rate": 1.506585341627259e-07, "loss": 0.4097, "step": 22477 }, { "epoch": 2.68, "learning_rate": 1.5054865912503436e-07, "loss": 0.4026, "step": 22478 }, { "epoch": 2.68, "learning_rate": 1.5043882292397033e-07, "loss": 0.4913, "step": 22479 }, { "epoch": 2.68, "learning_rate": 1.5032902556135016e-07, "loss": 0.5122, "step": 22480 }, { "epoch": 2.68, "learning_rate": 1.502192670389879e-07, "loss": 0.5099, "step": 22481 }, { "epoch": 2.68, "learning_rate": 1.5010954735869832e-07, "loss": 0.3483, "step": 22482 }, { "epoch": 2.68, "learning_rate": 1.4999986652229514e-07, "loss": 0.4256, "step": 22483 }, { "epoch": 2.68, "learning_rate": 1.4989022453159168e-07, "loss": 0.4701, "step": 22484 }, { "epoch": 2.68, "learning_rate": 1.4978062138839982e-07, "loss": 0.4715, "step": 22485 }, { "epoch": 2.68, "learning_rate": 1.4967105709453146e-07, "loss": 0.4497, "step": 22486 }, { "epoch": 2.68, "learning_rate": 1.495615316517976e-07, "loss": 0.6355, "step": 22487 }, { "epoch": 2.68, "learning_rate": 1.4945204506200904e-07, "loss": 0.4262, "step": 22488 }, { "epoch": 2.68, "learning_rate": 1.4934259732697522e-07, "loss": 0.3039, "step": 22489 }, { "epoch": 2.68, "learning_rate": 1.4923318844850542e-07, "loss": 0.3463, "step": 22490 }, { "epoch": 2.68, "learning_rate": 1.4912381842840884e-07, "loss": 0.4363, "step": 22491 }, { "epoch": 2.68, "learning_rate": 1.490144872684926e-07, "loss": 0.5298, "step": 22492 }, { "epoch": 2.68, "learning_rate": 1.489051949705636e-07, "loss": 0.5906, "step": 22493 }, { "epoch": 2.68, "learning_rate": 1.4879594153642928e-07, "loss": 0.5238, "step": 22494 }, { "epoch": 2.68, "learning_rate": 1.486867269678957e-07, "loss": 0.3403, "step": 22495 }, { "epoch": 2.68, "learning_rate": 1.4857755126676726e-07, "loss": 0.507, "step": 22496 }, { "epoch": 2.68, "learning_rate": 1.4846841443484943e-07, "loss": 0.4797, "step": 22497 }, { "epoch": 2.68, "learning_rate": 1.483593164739458e-07, "loss": 0.5153, "step": 22498 }, { "epoch": 2.68, "learning_rate": 1.4825025738586046e-07, "loss": 0.5332, "step": 22499 }, { "epoch": 2.68, "learning_rate": 1.481412371723956e-07, "loss": 0.5105, "step": 22500 }, { "epoch": 2.68, "learning_rate": 1.4803225583535335e-07, "loss": 0.293, "step": 22501 }, { "epoch": 2.68, "learning_rate": 1.479233133765351e-07, "loss": 0.4468, "step": 22502 }, { "epoch": 2.68, "learning_rate": 1.4781440979774215e-07, "loss": 0.4527, "step": 22503 }, { "epoch": 2.68, "learning_rate": 1.477055451007739e-07, "loss": 0.462, "step": 22504 }, { "epoch": 2.68, "learning_rate": 1.4759671928743058e-07, "loss": 0.484, "step": 22505 }, { "epoch": 2.68, "learning_rate": 1.4748793235951103e-07, "loss": 0.5895, "step": 22506 }, { "epoch": 2.68, "learning_rate": 1.4737918431881353e-07, "loss": 0.3186, "step": 22507 }, { "epoch": 2.68, "learning_rate": 1.47270475167135e-07, "loss": 0.2707, "step": 22508 }, { "epoch": 2.68, "learning_rate": 1.4716180490627285e-07, "loss": 0.5463, "step": 22509 }, { "epoch": 2.68, "learning_rate": 1.47053173538024e-07, "loss": 0.408, "step": 22510 }, { "epoch": 2.68, "learning_rate": 1.469445810641834e-07, "loss": 0.5804, "step": 22511 }, { "epoch": 2.68, "learning_rate": 1.4683602748654602e-07, "loss": 0.5882, "step": 22512 }, { "epoch": 2.68, "learning_rate": 1.4672751280690706e-07, "loss": 0.3561, "step": 22513 }, { "epoch": 2.68, "learning_rate": 1.4661903702705982e-07, "loss": 0.3897, "step": 22514 }, { "epoch": 2.68, "learning_rate": 1.4651060014879705e-07, "loss": 0.4064, "step": 22515 }, { "epoch": 2.68, "learning_rate": 1.4640220217391144e-07, "loss": 0.4873, "step": 22516 }, { "epoch": 2.68, "learning_rate": 1.462938431041949e-07, "loss": 0.565, "step": 22517 }, { "epoch": 2.68, "learning_rate": 1.4618552294143884e-07, "loss": 0.5782, "step": 22518 }, { "epoch": 2.68, "learning_rate": 1.4607724168743343e-07, "loss": 0.5177, "step": 22519 }, { "epoch": 2.68, "learning_rate": 1.459689993439692e-07, "loss": 0.3717, "step": 22520 }, { "epoch": 2.68, "learning_rate": 1.4586079591283468e-07, "loss": 0.4005, "step": 22521 }, { "epoch": 2.68, "learning_rate": 1.4575263139581853e-07, "loss": 0.4247, "step": 22522 }, { "epoch": 2.68, "learning_rate": 1.456445057947087e-07, "loss": 0.4796, "step": 22523 }, { "epoch": 2.68, "learning_rate": 1.455364191112929e-07, "loss": 0.6564, "step": 22524 }, { "epoch": 2.68, "learning_rate": 1.454283713473581e-07, "loss": 0.5656, "step": 22525 }, { "epoch": 2.68, "learning_rate": 1.4532036250468923e-07, "loss": 0.3358, "step": 22526 }, { "epoch": 2.68, "learning_rate": 1.4521239258507264e-07, "loss": 0.4221, "step": 22527 }, { "epoch": 2.68, "learning_rate": 1.45104461590293e-07, "loss": 0.4175, "step": 22528 }, { "epoch": 2.68, "learning_rate": 1.449965695221342e-07, "loss": 0.4373, "step": 22529 }, { "epoch": 2.68, "learning_rate": 1.448887163823795e-07, "loss": 0.584, "step": 22530 }, { "epoch": 2.68, "learning_rate": 1.4478090217281166e-07, "loss": 0.5305, "step": 22531 }, { "epoch": 2.68, "learning_rate": 1.446731268952134e-07, "loss": 0.4228, "step": 22532 }, { "epoch": 2.68, "learning_rate": 1.445653905513661e-07, "loss": 0.3639, "step": 22533 }, { "epoch": 2.68, "learning_rate": 1.4445769314305024e-07, "loss": 0.4976, "step": 22534 }, { "epoch": 2.68, "learning_rate": 1.4435003467204694e-07, "loss": 0.4322, "step": 22535 }, { "epoch": 2.68, "learning_rate": 1.4424241514013448e-07, "loss": 0.4924, "step": 22536 }, { "epoch": 2.68, "learning_rate": 1.441348345490931e-07, "loss": 0.6239, "step": 22537 }, { "epoch": 2.68, "learning_rate": 1.440272929007003e-07, "loss": 0.3869, "step": 22538 }, { "epoch": 2.68, "learning_rate": 1.4391979019673403e-07, "loss": 0.3154, "step": 22539 }, { "epoch": 2.68, "learning_rate": 1.4381232643897153e-07, "loss": 0.3801, "step": 22540 }, { "epoch": 2.68, "learning_rate": 1.4370490162918887e-07, "loss": 0.4415, "step": 22541 }, { "epoch": 2.68, "learning_rate": 1.4359751576916214e-07, "loss": 0.4951, "step": 22542 }, { "epoch": 2.68, "learning_rate": 1.4349016886066575e-07, "loss": 0.6225, "step": 22543 }, { "epoch": 2.68, "learning_rate": 1.4338286090547494e-07, "loss": 0.4603, "step": 22544 }, { "epoch": 2.68, "learning_rate": 1.43275591905363e-07, "loss": 0.3268, "step": 22545 }, { "epoch": 2.69, "learning_rate": 1.4316836186210297e-07, "loss": 0.3564, "step": 22546 }, { "epoch": 2.69, "learning_rate": 1.430611707774679e-07, "loss": 0.4429, "step": 22547 }, { "epoch": 2.69, "learning_rate": 1.4295401865322967e-07, "loss": 0.4187, "step": 22548 }, { "epoch": 2.69, "learning_rate": 1.4284690549115937e-07, "loss": 0.6517, "step": 22549 }, { "epoch": 2.69, "learning_rate": 1.4273983129302725e-07, "loss": 0.499, "step": 22550 }, { "epoch": 2.69, "learning_rate": 1.426327960606033e-07, "loss": 0.3823, "step": 22551 }, { "epoch": 2.69, "learning_rate": 1.4252579979565751e-07, "loss": 0.3767, "step": 22552 }, { "epoch": 2.69, "learning_rate": 1.4241884249995758e-07, "loss": 0.5073, "step": 22553 }, { "epoch": 2.69, "learning_rate": 1.4231192417527213e-07, "loss": 0.5683, "step": 22554 }, { "epoch": 2.69, "learning_rate": 1.4220504482336888e-07, "loss": 0.5479, "step": 22555 }, { "epoch": 2.69, "learning_rate": 1.420982044460134e-07, "loss": 0.587, "step": 22556 }, { "epoch": 2.69, "learning_rate": 1.4199140304497316e-07, "loss": 0.2893, "step": 22557 }, { "epoch": 2.69, "learning_rate": 1.4188464062201229e-07, "loss": 0.3547, "step": 22558 }, { "epoch": 2.69, "learning_rate": 1.4177791717889662e-07, "loss": 0.4046, "step": 22559 }, { "epoch": 2.69, "learning_rate": 1.416712327173894e-07, "loss": 0.4513, "step": 22560 }, { "epoch": 2.69, "learning_rate": 1.415645872392546e-07, "loss": 0.5828, "step": 22561 }, { "epoch": 2.69, "learning_rate": 1.4145798074625488e-07, "loss": 0.6109, "step": 22562 }, { "epoch": 2.69, "learning_rate": 1.4135141324015334e-07, "loss": 0.3733, "step": 22563 }, { "epoch": 2.69, "learning_rate": 1.4124488472271074e-07, "loss": 0.3255, "step": 22564 }, { "epoch": 2.69, "learning_rate": 1.4113839519568767e-07, "loss": 0.433, "step": 22565 }, { "epoch": 2.69, "learning_rate": 1.410319446608449e-07, "loss": 0.4023, "step": 22566 }, { "epoch": 2.69, "learning_rate": 1.4092553311994246e-07, "loss": 0.5078, "step": 22567 }, { "epoch": 2.69, "learning_rate": 1.4081916057473833e-07, "loss": 0.5141, "step": 22568 }, { "epoch": 2.69, "learning_rate": 1.407128270269917e-07, "loss": 0.4191, "step": 22569 }, { "epoch": 2.69, "learning_rate": 1.4060653247846e-07, "loss": 0.3081, "step": 22570 }, { "epoch": 2.69, "learning_rate": 1.405002769309005e-07, "loss": 0.3911, "step": 22571 }, { "epoch": 2.69, "learning_rate": 1.40394060386069e-07, "loss": 0.3515, "step": 22572 }, { "epoch": 2.69, "learning_rate": 1.4028788284572186e-07, "loss": 0.613, "step": 22573 }, { "epoch": 2.69, "learning_rate": 1.4018174431161408e-07, "loss": 0.5893, "step": 22574 }, { "epoch": 2.69, "learning_rate": 1.4007564478550008e-07, "loss": 0.5071, "step": 22575 }, { "epoch": 2.69, "learning_rate": 1.3996958426913342e-07, "loss": 0.2514, "step": 22576 }, { "epoch": 2.69, "learning_rate": 1.3986356276426776e-07, "loss": 0.4332, "step": 22577 }, { "epoch": 2.69, "learning_rate": 1.3975758027265584e-07, "loss": 0.4875, "step": 22578 }, { "epoch": 2.69, "learning_rate": 1.396516367960485e-07, "loss": 0.3513, "step": 22579 }, { "epoch": 2.69, "learning_rate": 1.395457323361979e-07, "loss": 0.4951, "step": 22580 }, { "epoch": 2.69, "learning_rate": 1.3943986689485407e-07, "loss": 0.5631, "step": 22581 }, { "epoch": 2.69, "learning_rate": 1.393340404737678e-07, "loss": 0.4735, "step": 22582 }, { "epoch": 2.69, "learning_rate": 1.3922825307468774e-07, "loss": 0.427, "step": 22583 }, { "epoch": 2.69, "learning_rate": 1.3912250469936244e-07, "loss": 0.3756, "step": 22584 }, { "epoch": 2.69, "learning_rate": 1.3901679534954083e-07, "loss": 0.4038, "step": 22585 }, { "epoch": 2.69, "learning_rate": 1.389111250269698e-07, "loss": 0.4969, "step": 22586 }, { "epoch": 2.69, "learning_rate": 1.3880549373339519e-07, "loss": 0.5748, "step": 22587 }, { "epoch": 2.69, "learning_rate": 1.3869990147056423e-07, "loss": 0.4472, "step": 22588 }, { "epoch": 2.69, "learning_rate": 1.3859434824022243e-07, "loss": 0.2781, "step": 22589 }, { "epoch": 2.69, "learning_rate": 1.3848883404411374e-07, "loss": 0.3635, "step": 22590 }, { "epoch": 2.69, "learning_rate": 1.3838335888398251e-07, "loss": 0.3911, "step": 22591 }, { "epoch": 2.69, "learning_rate": 1.3827792276157325e-07, "loss": 0.4388, "step": 22592 }, { "epoch": 2.69, "learning_rate": 1.3817252567862787e-07, "loss": 0.5196, "step": 22593 }, { "epoch": 2.69, "learning_rate": 1.380671676368886e-07, "loss": 0.484, "step": 22594 }, { "epoch": 2.69, "learning_rate": 1.3796184863809708e-07, "loss": 0.3488, "step": 22595 }, { "epoch": 2.69, "learning_rate": 1.3785656868399444e-07, "loss": 0.4596, "step": 22596 }, { "epoch": 2.69, "learning_rate": 1.377513277763215e-07, "loss": 0.4544, "step": 22597 }, { "epoch": 2.69, "learning_rate": 1.376461259168166e-07, "loss": 0.5197, "step": 22598 }, { "epoch": 2.69, "learning_rate": 1.3754096310721997e-07, "loss": 0.5418, "step": 22599 }, { "epoch": 2.69, "learning_rate": 1.3743583934926947e-07, "loss": 0.6785, "step": 22600 }, { "epoch": 2.69, "learning_rate": 1.373307546447025e-07, "loss": 0.3388, "step": 22601 }, { "epoch": 2.69, "learning_rate": 1.372257089952564e-07, "loss": 0.3914, "step": 22602 }, { "epoch": 2.69, "learning_rate": 1.3712070240266745e-07, "loss": 0.4937, "step": 22603 }, { "epoch": 2.69, "learning_rate": 1.370157348686718e-07, "loss": 0.5339, "step": 22604 }, { "epoch": 2.69, "learning_rate": 1.369108063950042e-07, "loss": 0.5877, "step": 22605 }, { "epoch": 2.69, "learning_rate": 1.3680591698339961e-07, "loss": 0.5539, "step": 22606 }, { "epoch": 2.69, "learning_rate": 1.3670106663559113e-07, "loss": 0.3751, "step": 22607 }, { "epoch": 2.69, "learning_rate": 1.3659625535331262e-07, "loss": 0.3761, "step": 22608 }, { "epoch": 2.69, "learning_rate": 1.3649148313829574e-07, "loss": 0.4229, "step": 22609 }, { "epoch": 2.69, "learning_rate": 1.363867499922733e-07, "loss": 0.5441, "step": 22610 }, { "epoch": 2.69, "learning_rate": 1.3628205591697613e-07, "loss": 0.5947, "step": 22611 }, { "epoch": 2.69, "learning_rate": 1.3617740091413507e-07, "loss": 0.5841, "step": 22612 }, { "epoch": 2.69, "learning_rate": 1.3607278498547983e-07, "loss": 0.4215, "step": 22613 }, { "epoch": 2.69, "learning_rate": 1.3596820813273986e-07, "loss": 0.2845, "step": 22614 }, { "epoch": 2.69, "learning_rate": 1.358636703576438e-07, "loss": 0.4044, "step": 22615 }, { "epoch": 2.69, "learning_rate": 1.357591716619197e-07, "loss": 0.3858, "step": 22616 }, { "epoch": 2.69, "learning_rate": 1.3565471204729453e-07, "loss": 0.4842, "step": 22617 }, { "epoch": 2.69, "learning_rate": 1.3555029151549548e-07, "loss": 0.4743, "step": 22618 }, { "epoch": 2.69, "learning_rate": 1.354459100682487e-07, "loss": 0.4787, "step": 22619 }, { "epoch": 2.69, "learning_rate": 1.353415677072792e-07, "loss": 0.3633, "step": 22620 }, { "epoch": 2.69, "learning_rate": 1.3523726443431224e-07, "loss": 0.3647, "step": 22621 }, { "epoch": 2.69, "learning_rate": 1.3513300025107147e-07, "loss": 0.4037, "step": 22622 }, { "epoch": 2.69, "learning_rate": 1.3502877515928104e-07, "loss": 0.4913, "step": 22623 }, { "epoch": 2.69, "learning_rate": 1.3492458916066292e-07, "loss": 0.5908, "step": 22624 }, { "epoch": 2.69, "learning_rate": 1.348204422569399e-07, "loss": 0.4831, "step": 22625 }, { "epoch": 2.69, "learning_rate": 1.3471633444983312e-07, "loss": 0.3022, "step": 22626 }, { "epoch": 2.69, "learning_rate": 1.3461226574106452e-07, "loss": 0.333, "step": 22627 }, { "epoch": 2.69, "learning_rate": 1.3450823613235354e-07, "loss": 0.4023, "step": 22628 }, { "epoch": 2.69, "learning_rate": 1.3440424562541938e-07, "loss": 0.5182, "step": 22629 }, { "epoch": 2.7, "learning_rate": 1.3430029422198175e-07, "loss": 0.6046, "step": 22630 }, { "epoch": 2.7, "learning_rate": 1.3419638192375904e-07, "loss": 0.6828, "step": 22631 }, { "epoch": 2.7, "learning_rate": 1.3409250873246822e-07, "loss": 0.4389, "step": 22632 }, { "epoch": 2.7, "learning_rate": 1.3398867464982674e-07, "loss": 0.3431, "step": 22633 }, { "epoch": 2.7, "learning_rate": 1.338848796775513e-07, "loss": 0.3499, "step": 22634 }, { "epoch": 2.7, "learning_rate": 1.3378112381735754e-07, "loss": 0.4113, "step": 22635 }, { "epoch": 2.7, "learning_rate": 1.3367740707095982e-07, "loss": 0.496, "step": 22636 }, { "epoch": 2.7, "learning_rate": 1.3357372944007324e-07, "loss": 0.5816, "step": 22637 }, { "epoch": 2.7, "learning_rate": 1.3347009092641166e-07, "loss": 0.2455, "step": 22638 }, { "epoch": 2.7, "learning_rate": 1.3336649153168791e-07, "loss": 0.3647, "step": 22639 }, { "epoch": 2.7, "learning_rate": 1.3326293125761446e-07, "loss": 0.4386, "step": 22640 }, { "epoch": 2.7, "learning_rate": 1.3315941010590332e-07, "loss": 0.4962, "step": 22641 }, { "epoch": 2.7, "learning_rate": 1.3305592807826616e-07, "loss": 0.5479, "step": 22642 }, { "epoch": 2.7, "learning_rate": 1.32952485176413e-07, "loss": 0.5957, "step": 22643 }, { "epoch": 2.7, "learning_rate": 1.3284908140205388e-07, "loss": 0.5297, "step": 22644 }, { "epoch": 2.7, "learning_rate": 1.3274571675689768e-07, "loss": 0.3273, "step": 22645 }, { "epoch": 2.7, "learning_rate": 1.326423912426539e-07, "loss": 0.4115, "step": 22646 }, { "epoch": 2.7, "learning_rate": 1.3253910486102972e-07, "loss": 0.4132, "step": 22647 }, { "epoch": 2.7, "learning_rate": 1.3243585761373272e-07, "loss": 0.4004, "step": 22648 }, { "epoch": 2.7, "learning_rate": 1.3233264950246987e-07, "loss": 0.5087, "step": 22649 }, { "epoch": 2.7, "learning_rate": 1.322294805289473e-07, "loss": 0.5515, "step": 22650 }, { "epoch": 2.7, "learning_rate": 1.3212635069486946e-07, "loss": 0.3167, "step": 22651 }, { "epoch": 2.7, "learning_rate": 1.3202326000194194e-07, "loss": 0.4334, "step": 22652 }, { "epoch": 2.7, "learning_rate": 1.3192020845186864e-07, "loss": 0.4831, "step": 22653 }, { "epoch": 2.7, "learning_rate": 1.3181719604635297e-07, "loss": 0.5326, "step": 22654 }, { "epoch": 2.7, "learning_rate": 1.3171422278709766e-07, "loss": 0.6146, "step": 22655 }, { "epoch": 2.7, "learning_rate": 1.316112886758053e-07, "loss": 0.6167, "step": 22656 }, { "epoch": 2.7, "learning_rate": 1.31508393714177e-07, "loss": 0.3191, "step": 22657 }, { "epoch": 2.7, "learning_rate": 1.3140553790391335e-07, "loss": 0.3497, "step": 22658 }, { "epoch": 2.7, "learning_rate": 1.3130272124671496e-07, "loss": 0.458, "step": 22659 }, { "epoch": 2.7, "learning_rate": 1.311999437442815e-07, "loss": 0.5069, "step": 22660 }, { "epoch": 2.7, "learning_rate": 1.310972053983117e-07, "loss": 0.5327, "step": 22661 }, { "epoch": 2.7, "learning_rate": 1.309945062105039e-07, "loss": 0.575, "step": 22662 }, { "epoch": 2.7, "learning_rate": 1.308918461825559e-07, "loss": 0.3932, "step": 22663 }, { "epoch": 2.7, "learning_rate": 1.3078922531616438e-07, "loss": 0.3728, "step": 22664 }, { "epoch": 2.7, "learning_rate": 1.3068664361302552e-07, "loss": 0.4778, "step": 22665 }, { "epoch": 2.7, "learning_rate": 1.3058410107483514e-07, "loss": 0.4554, "step": 22666 }, { "epoch": 2.7, "learning_rate": 1.3048159770328834e-07, "loss": 0.5139, "step": 22667 }, { "epoch": 2.7, "learning_rate": 1.3037913350008008e-07, "loss": 0.5449, "step": 22668 }, { "epoch": 2.7, "learning_rate": 1.3027670846690293e-07, "loss": 0.5446, "step": 22669 }, { "epoch": 2.7, "learning_rate": 1.3017432260545083e-07, "loss": 0.2733, "step": 22670 }, { "epoch": 2.7, "learning_rate": 1.3007197591741656e-07, "loss": 0.3756, "step": 22671 }, { "epoch": 2.7, "learning_rate": 1.2996966840449098e-07, "loss": 0.4219, "step": 22672 }, { "epoch": 2.7, "learning_rate": 1.2986740006836557e-07, "loss": 0.5115, "step": 22673 }, { "epoch": 2.7, "learning_rate": 1.2976517091073086e-07, "loss": 0.6459, "step": 22674 }, { "epoch": 2.7, "learning_rate": 1.2966298093327667e-07, "loss": 0.5386, "step": 22675 }, { "epoch": 2.7, "learning_rate": 1.2956083013769272e-07, "loss": 0.3031, "step": 22676 }, { "epoch": 2.7, "learning_rate": 1.2945871852566682e-07, "loss": 0.2886, "step": 22677 }, { "epoch": 2.7, "learning_rate": 1.2935664609888736e-07, "loss": 0.3876, "step": 22678 }, { "epoch": 2.7, "learning_rate": 1.2925461285904133e-07, "loss": 0.472, "step": 22679 }, { "epoch": 2.7, "learning_rate": 1.291526188078157e-07, "loss": 0.4758, "step": 22680 }, { "epoch": 2.7, "learning_rate": 1.290506639468958e-07, "loss": 0.5728, "step": 22681 }, { "epoch": 2.7, "learning_rate": 1.289487482779675e-07, "loss": 0.3781, "step": 22682 }, { "epoch": 2.7, "learning_rate": 1.288468718027158e-07, "loss": 0.3188, "step": 22683 }, { "epoch": 2.7, "learning_rate": 1.2874503452282362e-07, "loss": 0.3277, "step": 22684 }, { "epoch": 2.7, "learning_rate": 1.2864323643997563e-07, "loss": 0.5225, "step": 22685 }, { "epoch": 2.7, "learning_rate": 1.2854147755585327e-07, "loss": 0.4843, "step": 22686 }, { "epoch": 2.7, "learning_rate": 1.284397578721397e-07, "loss": 0.595, "step": 22687 }, { "epoch": 2.7, "learning_rate": 1.2833807739051573e-07, "loss": 0.413, "step": 22688 }, { "epoch": 2.7, "learning_rate": 1.282364361126623e-07, "loss": 0.2155, "step": 22689 }, { "epoch": 2.7, "learning_rate": 1.281348340402594e-07, "loss": 0.3852, "step": 22690 }, { "epoch": 2.7, "learning_rate": 1.2803327117498737e-07, "loss": 0.4151, "step": 22691 }, { "epoch": 2.7, "learning_rate": 1.2793174751852405e-07, "loss": 0.5182, "step": 22692 }, { "epoch": 2.7, "learning_rate": 1.2783026307254782e-07, "loss": 0.4982, "step": 22693 }, { "epoch": 2.7, "learning_rate": 1.2772881783873646e-07, "loss": 0.6596, "step": 22694 }, { "epoch": 2.7, "learning_rate": 1.2762741181876702e-07, "loss": 0.3096, "step": 22695 }, { "epoch": 2.7, "learning_rate": 1.275260450143151e-07, "loss": 0.4165, "step": 22696 }, { "epoch": 2.7, "learning_rate": 1.274247174270568e-07, "loss": 0.47, "step": 22697 }, { "epoch": 2.7, "learning_rate": 1.2732342905866752e-07, "loss": 0.4306, "step": 22698 }, { "epoch": 2.7, "learning_rate": 1.272221799108203e-07, "loss": 0.5747, "step": 22699 }, { "epoch": 2.7, "learning_rate": 1.2712096998519025e-07, "loss": 0.6212, "step": 22700 }, { "epoch": 2.7, "learning_rate": 1.2701979928344933e-07, "loss": 0.3106, "step": 22701 }, { "epoch": 2.7, "learning_rate": 1.2691866780727034e-07, "loss": 0.3465, "step": 22702 }, { "epoch": 2.7, "learning_rate": 1.2681757555832453e-07, "loss": 0.4964, "step": 22703 }, { "epoch": 2.7, "learning_rate": 1.2671652253828353e-07, "loss": 0.4254, "step": 22704 }, { "epoch": 2.7, "learning_rate": 1.2661550874881745e-07, "loss": 0.5408, "step": 22705 }, { "epoch": 2.7, "learning_rate": 1.265145341915966e-07, "loss": 0.6627, "step": 22706 }, { "epoch": 2.7, "learning_rate": 1.2641359886828935e-07, "loss": 0.2782, "step": 22707 }, { "epoch": 2.7, "learning_rate": 1.263127027805644e-07, "loss": 0.4224, "step": 22708 }, { "epoch": 2.7, "learning_rate": 1.2621184593008957e-07, "loss": 0.4625, "step": 22709 }, { "epoch": 2.7, "learning_rate": 1.2611102831853238e-07, "loss": 0.4371, "step": 22710 }, { "epoch": 2.7, "learning_rate": 1.260102499475588e-07, "loss": 0.6219, "step": 22711 }, { "epoch": 2.7, "learning_rate": 1.2590951081883518e-07, "loss": 0.687, "step": 22712 }, { "epoch": 2.7, "learning_rate": 1.2580881093402663e-07, "loss": 0.4082, "step": 22713 }, { "epoch": 2.71, "learning_rate": 1.2570815029479766e-07, "loss": 0.3532, "step": 22714 }, { "epoch": 2.71, "learning_rate": 1.2560752890281192e-07, "loss": 0.4219, "step": 22715 }, { "epoch": 2.71, "learning_rate": 1.2550694675973278e-07, "loss": 0.4236, "step": 22716 }, { "epoch": 2.71, "learning_rate": 1.2540640386722368e-07, "loss": 0.5297, "step": 22717 }, { "epoch": 2.71, "learning_rate": 1.253059002269455e-07, "loss": 0.5212, "step": 22718 }, { "epoch": 2.71, "learning_rate": 1.2520543584055995e-07, "loss": 0.4641, "step": 22719 }, { "epoch": 2.71, "learning_rate": 1.2510501070972818e-07, "loss": 0.3032, "step": 22720 }, { "epoch": 2.71, "learning_rate": 1.2500462483611005e-07, "loss": 0.4346, "step": 22721 }, { "epoch": 2.71, "learning_rate": 1.2490427822136414e-07, "loss": 0.4438, "step": 22722 }, { "epoch": 2.71, "learning_rate": 1.2480397086715001e-07, "loss": 0.49, "step": 22723 }, { "epoch": 2.71, "learning_rate": 1.247037027751255e-07, "loss": 0.5927, "step": 22724 }, { "epoch": 2.71, "learning_rate": 1.2460347394694844e-07, "loss": 0.5832, "step": 22725 }, { "epoch": 2.71, "learning_rate": 1.2450328438427472e-07, "loss": 0.2787, "step": 22726 }, { "epoch": 2.71, "learning_rate": 1.244031340887611e-07, "loss": 0.4302, "step": 22727 }, { "epoch": 2.71, "learning_rate": 1.2430302306206343e-07, "loss": 0.4556, "step": 22728 }, { "epoch": 2.71, "learning_rate": 1.2420295130583625e-07, "loss": 0.4709, "step": 22729 }, { "epoch": 2.71, "learning_rate": 1.2410291882173298e-07, "loss": 0.5292, "step": 22730 }, { "epoch": 2.71, "learning_rate": 1.240029256114081e-07, "loss": 0.621, "step": 22731 }, { "epoch": 2.71, "learning_rate": 1.2390297167651444e-07, "loss": 0.3633, "step": 22732 }, { "epoch": 2.71, "learning_rate": 1.2380305701870377e-07, "loss": 0.4442, "step": 22733 }, { "epoch": 2.71, "learning_rate": 1.237031816396278e-07, "loss": 0.3596, "step": 22734 }, { "epoch": 2.71, "learning_rate": 1.23603345540938e-07, "loss": 0.4599, "step": 22735 }, { "epoch": 2.71, "learning_rate": 1.2350354872428445e-07, "loss": 0.5843, "step": 22736 }, { "epoch": 2.71, "learning_rate": 1.2340379119131607e-07, "loss": 0.6851, "step": 22737 }, { "epoch": 2.71, "learning_rate": 1.233040729436827e-07, "loss": 0.3973, "step": 22738 }, { "epoch": 2.71, "learning_rate": 1.2320439398303213e-07, "loss": 0.4362, "step": 22739 }, { "epoch": 2.71, "learning_rate": 1.2310475431101281e-07, "loss": 0.3843, "step": 22740 }, { "epoch": 2.71, "learning_rate": 1.230051539292712e-07, "loss": 0.4765, "step": 22741 }, { "epoch": 2.71, "learning_rate": 1.2290559283945402e-07, "loss": 0.5267, "step": 22742 }, { "epoch": 2.71, "learning_rate": 1.2280607104320664e-07, "loss": 0.6666, "step": 22743 }, { "epoch": 2.71, "learning_rate": 1.2270658854217438e-07, "loss": 0.5161, "step": 22744 }, { "epoch": 2.71, "learning_rate": 1.226071453380015e-07, "loss": 0.3047, "step": 22745 }, { "epoch": 2.71, "learning_rate": 1.2250774143233196e-07, "loss": 0.3874, "step": 22746 }, { "epoch": 2.71, "learning_rate": 1.2240837682680946e-07, "loss": 0.3476, "step": 22747 }, { "epoch": 2.71, "learning_rate": 1.2230905152307542e-07, "loss": 0.5649, "step": 22748 }, { "epoch": 2.71, "learning_rate": 1.2220976552277275e-07, "loss": 0.7189, "step": 22749 }, { "epoch": 2.71, "learning_rate": 1.2211051882754177e-07, "loss": 0.5049, "step": 22750 }, { "epoch": 2.71, "learning_rate": 1.2201131143902396e-07, "loss": 0.2819, "step": 22751 }, { "epoch": 2.71, "learning_rate": 1.2191214335885798e-07, "loss": 0.3256, "step": 22752 }, { "epoch": 2.71, "learning_rate": 1.2181301458868422e-07, "loss": 0.3901, "step": 22753 }, { "epoch": 2.71, "learning_rate": 1.2171392513014052e-07, "loss": 0.4785, "step": 22754 }, { "epoch": 2.71, "learning_rate": 1.2161487498486584e-07, "loss": 0.514, "step": 22755 }, { "epoch": 2.71, "learning_rate": 1.2151586415449613e-07, "loss": 0.5205, "step": 22756 }, { "epoch": 2.71, "learning_rate": 1.2141689264066948e-07, "loss": 0.2655, "step": 22757 }, { "epoch": 2.71, "learning_rate": 1.213179604450207e-07, "loss": 0.3403, "step": 22758 }, { "epoch": 2.71, "learning_rate": 1.21219067569186e-07, "loss": 0.4376, "step": 22759 }, { "epoch": 2.71, "learning_rate": 1.2112021401479907e-07, "loss": 0.4323, "step": 22760 }, { "epoch": 2.71, "learning_rate": 1.2102139978349497e-07, "loss": 0.5072, "step": 22761 }, { "epoch": 2.71, "learning_rate": 1.2092262487690686e-07, "loss": 0.6483, "step": 22762 }, { "epoch": 2.71, "learning_rate": 1.2082388929666733e-07, "loss": 0.3257, "step": 22763 }, { "epoch": 2.71, "learning_rate": 1.2072519304440867e-07, "loss": 0.3357, "step": 22764 }, { "epoch": 2.71, "learning_rate": 1.2062653612176178e-07, "loss": 0.4922, "step": 22765 }, { "epoch": 2.71, "learning_rate": 1.2052791853035846e-07, "loss": 0.4546, "step": 22766 }, { "epoch": 2.71, "learning_rate": 1.2042934027182763e-07, "loss": 0.5105, "step": 22767 }, { "epoch": 2.71, "learning_rate": 1.2033080134779968e-07, "loss": 0.619, "step": 22768 }, { "epoch": 2.71, "learning_rate": 1.202323017599033e-07, "loss": 0.438, "step": 22769 }, { "epoch": 2.71, "learning_rate": 1.2013384150976691e-07, "loss": 0.2834, "step": 22770 }, { "epoch": 2.71, "learning_rate": 1.2003542059901785e-07, "loss": 0.4316, "step": 22771 }, { "epoch": 2.71, "learning_rate": 1.1993703902928284e-07, "loss": 0.4382, "step": 22772 }, { "epoch": 2.71, "learning_rate": 1.1983869680218806e-07, "loss": 0.5457, "step": 22773 }, { "epoch": 2.71, "learning_rate": 1.1974039391935948e-07, "loss": 0.6229, "step": 22774 }, { "epoch": 2.71, "learning_rate": 1.196421303824219e-07, "loss": 0.6175, "step": 22775 }, { "epoch": 2.71, "learning_rate": 1.1954390619299956e-07, "loss": 0.388, "step": 22776 }, { "epoch": 2.71, "learning_rate": 1.1944572135271642e-07, "loss": 0.4265, "step": 22777 }, { "epoch": 2.71, "learning_rate": 1.193475758631954e-07, "loss": 0.3595, "step": 22778 }, { "epoch": 2.71, "learning_rate": 1.192494697260582e-07, "loss": 0.4974, "step": 22779 }, { "epoch": 2.71, "learning_rate": 1.1915140294292693e-07, "loss": 0.6238, "step": 22780 }, { "epoch": 2.71, "learning_rate": 1.1905337551542329e-07, "loss": 0.4839, "step": 22781 }, { "epoch": 2.71, "learning_rate": 1.1895538744516683e-07, "loss": 0.329, "step": 22782 }, { "epoch": 2.71, "learning_rate": 1.1885743873377737e-07, "loss": 0.3311, "step": 22783 }, { "epoch": 2.71, "learning_rate": 1.1875952938287444e-07, "loss": 0.3186, "step": 22784 }, { "epoch": 2.71, "learning_rate": 1.1866165939407676e-07, "loss": 0.4259, "step": 22785 }, { "epoch": 2.71, "learning_rate": 1.1856382876900136e-07, "loss": 0.5042, "step": 22786 }, { "epoch": 2.71, "learning_rate": 1.1846603750926555e-07, "loss": 0.6266, "step": 22787 }, { "epoch": 2.71, "learning_rate": 1.1836828561648583e-07, "loss": 0.4541, "step": 22788 }, { "epoch": 2.71, "learning_rate": 1.1827057309227868e-07, "loss": 0.2771, "step": 22789 }, { "epoch": 2.71, "learning_rate": 1.1817289993825837e-07, "loss": 0.4269, "step": 22790 }, { "epoch": 2.71, "learning_rate": 1.180752661560397e-07, "loss": 0.468, "step": 22791 }, { "epoch": 2.71, "learning_rate": 1.1797767174723723e-07, "loss": 0.4653, "step": 22792 }, { "epoch": 2.71, "learning_rate": 1.1788011671346384e-07, "loss": 0.554, "step": 22793 }, { "epoch": 2.71, "learning_rate": 1.1778260105633154e-07, "loss": 0.4078, "step": 22794 }, { "epoch": 2.71, "learning_rate": 1.176851247774527e-07, "loss": 0.3476, "step": 22795 }, { "epoch": 2.71, "learning_rate": 1.1758768787843905e-07, "loss": 0.4039, "step": 22796 }, { "epoch": 2.71, "learning_rate": 1.1749029036090043e-07, "loss": 0.4258, "step": 22797 }, { "epoch": 2.72, "learning_rate": 1.1739293222644721e-07, "loss": 0.4985, "step": 22798 }, { "epoch": 2.72, "learning_rate": 1.1729561347668922e-07, "loss": 0.5798, "step": 22799 }, { "epoch": 2.72, "learning_rate": 1.1719833411323462e-07, "loss": 0.5427, "step": 22800 }, { "epoch": 2.72, "learning_rate": 1.17101094137691e-07, "loss": 0.2999, "step": 22801 }, { "epoch": 2.72, "learning_rate": 1.1700389355166625e-07, "loss": 0.4071, "step": 22802 }, { "epoch": 2.72, "learning_rate": 1.1690673235676714e-07, "loss": 0.3977, "step": 22803 }, { "epoch": 2.72, "learning_rate": 1.1680961055460016e-07, "loss": 0.4423, "step": 22804 }, { "epoch": 2.72, "learning_rate": 1.1671252814676986e-07, "loss": 0.6195, "step": 22805 }, { "epoch": 2.72, "learning_rate": 1.1661548513488163e-07, "loss": 0.6297, "step": 22806 }, { "epoch": 2.72, "learning_rate": 1.1651848152053974e-07, "loss": 0.362, "step": 22807 }, { "epoch": 2.72, "learning_rate": 1.1642151730534679e-07, "loss": 0.45, "step": 22808 }, { "epoch": 2.72, "learning_rate": 1.1632459249090622e-07, "loss": 0.3175, "step": 22809 }, { "epoch": 2.72, "learning_rate": 1.1622770707882008e-07, "loss": 0.4487, "step": 22810 }, { "epoch": 2.72, "learning_rate": 1.1613086107069016e-07, "loss": 0.527, "step": 22811 }, { "epoch": 2.72, "learning_rate": 1.160340544681171e-07, "loss": 0.5907, "step": 22812 }, { "epoch": 2.72, "learning_rate": 1.1593728727270104e-07, "loss": 0.4021, "step": 22813 }, { "epoch": 2.72, "learning_rate": 1.1584055948604179e-07, "loss": 0.3065, "step": 22814 }, { "epoch": 2.72, "learning_rate": 1.1574387110973834e-07, "loss": 0.4526, "step": 22815 }, { "epoch": 2.72, "learning_rate": 1.1564722214538831e-07, "loss": 0.4232, "step": 22816 }, { "epoch": 2.72, "learning_rate": 1.1555061259458989e-07, "loss": 0.5345, "step": 22817 }, { "epoch": 2.72, "learning_rate": 1.1545404245893982e-07, "loss": 0.6551, "step": 22818 }, { "epoch": 2.72, "learning_rate": 1.153575117400349e-07, "loss": 0.5562, "step": 22819 }, { "epoch": 2.72, "learning_rate": 1.1526102043947024e-07, "loss": 0.269, "step": 22820 }, { "epoch": 2.72, "learning_rate": 1.1516456855884095e-07, "loss": 0.5604, "step": 22821 }, { "epoch": 2.72, "learning_rate": 1.1506815609974131e-07, "loss": 0.4618, "step": 22822 }, { "epoch": 2.72, "learning_rate": 1.149717830637656e-07, "loss": 0.5495, "step": 22823 }, { "epoch": 2.72, "learning_rate": 1.148754494525059e-07, "loss": 0.6103, "step": 22824 }, { "epoch": 2.72, "learning_rate": 1.1477915526755535e-07, "loss": 0.6278, "step": 22825 }, { "epoch": 2.72, "learning_rate": 1.1468290051050573e-07, "loss": 0.3463, "step": 22826 }, { "epoch": 2.72, "learning_rate": 1.1458668518294775e-07, "loss": 0.4735, "step": 22827 }, { "epoch": 2.72, "learning_rate": 1.1449050928647204e-07, "loss": 0.4932, "step": 22828 }, { "epoch": 2.72, "learning_rate": 1.1439437282266819e-07, "loss": 0.3424, "step": 22829 }, { "epoch": 2.72, "learning_rate": 1.1429827579312575e-07, "loss": 0.6285, "step": 22830 }, { "epoch": 2.72, "learning_rate": 1.1420221819943289e-07, "loss": 0.6508, "step": 22831 }, { "epoch": 2.72, "learning_rate": 1.1410620004317752e-07, "loss": 0.336, "step": 22832 }, { "epoch": 2.72, "learning_rate": 1.1401022132594669e-07, "loss": 0.3352, "step": 22833 }, { "epoch": 2.72, "learning_rate": 1.1391428204932747e-07, "loss": 0.4703, "step": 22834 }, { "epoch": 2.72, "learning_rate": 1.1381838221490527e-07, "loss": 0.4855, "step": 22835 }, { "epoch": 2.72, "learning_rate": 1.1372252182426519e-07, "loss": 0.5482, "step": 22836 }, { "epoch": 2.72, "learning_rate": 1.1362670087899236e-07, "loss": 0.6497, "step": 22837 }, { "epoch": 2.72, "learning_rate": 1.1353091938067024e-07, "loss": 0.3753, "step": 22838 }, { "epoch": 2.72, "learning_rate": 1.1343517733088227e-07, "loss": 0.371, "step": 22839 }, { "epoch": 2.72, "learning_rate": 1.1333947473121082e-07, "loss": 0.358, "step": 22840 }, { "epoch": 2.72, "learning_rate": 1.1324381158323877e-07, "loss": 0.4798, "step": 22841 }, { "epoch": 2.72, "learning_rate": 1.1314818788854626e-07, "loss": 0.5411, "step": 22842 }, { "epoch": 2.72, "learning_rate": 1.1305260364871478e-07, "loss": 0.5131, "step": 22843 }, { "epoch": 2.72, "learning_rate": 1.1295705886532365e-07, "loss": 0.4632, "step": 22844 }, { "epoch": 2.72, "learning_rate": 1.1286155353995299e-07, "loss": 0.325, "step": 22845 }, { "epoch": 2.72, "learning_rate": 1.1276608767418096e-07, "loss": 0.41, "step": 22846 }, { "epoch": 2.72, "learning_rate": 1.1267066126958576e-07, "loss": 0.4321, "step": 22847 }, { "epoch": 2.72, "learning_rate": 1.1257527432774474e-07, "loss": 0.4143, "step": 22848 }, { "epoch": 2.72, "learning_rate": 1.1247992685023523e-07, "loss": 0.5424, "step": 22849 }, { "epoch": 2.72, "learning_rate": 1.1238461883863266e-07, "loss": 0.5686, "step": 22850 }, { "epoch": 2.72, "learning_rate": 1.1228935029451244e-07, "loss": 0.3206, "step": 22851 }, { "epoch": 2.72, "learning_rate": 1.1219412121944967e-07, "loss": 0.4198, "step": 22852 }, { "epoch": 2.72, "learning_rate": 1.1209893161501866e-07, "loss": 0.4112, "step": 22853 }, { "epoch": 2.72, "learning_rate": 1.1200378148279234e-07, "loss": 0.5485, "step": 22854 }, { "epoch": 2.72, "learning_rate": 1.1190867082434387e-07, "loss": 0.5869, "step": 22855 }, { "epoch": 2.72, "learning_rate": 1.1181359964124588e-07, "loss": 0.652, "step": 22856 }, { "epoch": 2.72, "learning_rate": 1.1171856793506963e-07, "loss": 0.3688, "step": 22857 }, { "epoch": 2.72, "learning_rate": 1.1162357570738525e-07, "loss": 0.4198, "step": 22858 }, { "epoch": 2.72, "learning_rate": 1.1152862295976369e-07, "loss": 0.4388, "step": 22859 }, { "epoch": 2.72, "learning_rate": 1.1143370969377454e-07, "loss": 0.4583, "step": 22860 }, { "epoch": 2.72, "learning_rate": 1.1133883591098654e-07, "loss": 0.5576, "step": 22861 }, { "epoch": 2.72, "learning_rate": 1.1124400161296789e-07, "loss": 0.5873, "step": 22862 }, { "epoch": 2.72, "learning_rate": 1.1114920680128649e-07, "loss": 0.3227, "step": 22863 }, { "epoch": 2.72, "learning_rate": 1.1105445147750943e-07, "loss": 0.3734, "step": 22864 }, { "epoch": 2.72, "learning_rate": 1.1095973564320212e-07, "loss": 0.3844, "step": 22865 }, { "epoch": 2.72, "learning_rate": 1.1086505929993108e-07, "loss": 0.4592, "step": 22866 }, { "epoch": 2.72, "learning_rate": 1.1077042244926089e-07, "loss": 0.4528, "step": 22867 }, { "epoch": 2.72, "learning_rate": 1.1067582509275643e-07, "loss": 0.6006, "step": 22868 }, { "epoch": 2.72, "learning_rate": 1.1058126723198088e-07, "loss": 0.5025, "step": 22869 }, { "epoch": 2.72, "learning_rate": 1.1048674886849741e-07, "loss": 0.2872, "step": 22870 }, { "epoch": 2.72, "learning_rate": 1.103922700038687e-07, "loss": 0.4129, "step": 22871 }, { "epoch": 2.72, "learning_rate": 1.1029783063965655e-07, "loss": 0.4263, "step": 22872 }, { "epoch": 2.72, "learning_rate": 1.1020343077742108e-07, "loss": 0.4775, "step": 22873 }, { "epoch": 2.72, "learning_rate": 1.1010907041872354e-07, "loss": 0.5556, "step": 22874 }, { "epoch": 2.72, "learning_rate": 1.1001474956512409e-07, "loss": 0.6237, "step": 22875 }, { "epoch": 2.72, "learning_rate": 1.0992046821818091e-07, "loss": 0.2513, "step": 22876 }, { "epoch": 2.72, "learning_rate": 1.0982622637945306e-07, "loss": 0.3151, "step": 22877 }, { "epoch": 2.72, "learning_rate": 1.0973202405049844e-07, "loss": 0.4546, "step": 22878 }, { "epoch": 2.72, "learning_rate": 1.0963786123287385e-07, "loss": 0.5297, "step": 22879 }, { "epoch": 2.72, "learning_rate": 1.095437379281361e-07, "loss": 0.593, "step": 22880 }, { "epoch": 2.72, "learning_rate": 1.0944965413784064e-07, "loss": 0.5173, "step": 22881 }, { "epoch": 2.73, "learning_rate": 1.0935560986354288e-07, "loss": 0.3338, "step": 22882 }, { "epoch": 2.73, "learning_rate": 1.0926160510679795e-07, "loss": 0.4142, "step": 22883 }, { "epoch": 2.73, "learning_rate": 1.0916763986915907e-07, "loss": 0.4255, "step": 22884 }, { "epoch": 2.73, "learning_rate": 1.0907371415217999e-07, "loss": 0.3837, "step": 22885 }, { "epoch": 2.73, "learning_rate": 1.089798279574128e-07, "loss": 0.5336, "step": 22886 }, { "epoch": 2.73, "learning_rate": 1.0888598128640959e-07, "loss": 0.5694, "step": 22887 }, { "epoch": 2.73, "learning_rate": 1.0879217414072163e-07, "loss": 0.313, "step": 22888 }, { "epoch": 2.73, "learning_rate": 1.0869840652189962e-07, "loss": 0.388, "step": 22889 }, { "epoch": 2.73, "learning_rate": 1.0860467843149425e-07, "loss": 0.4189, "step": 22890 }, { "epoch": 2.73, "learning_rate": 1.0851098987105346e-07, "loss": 0.491, "step": 22891 }, { "epoch": 2.73, "learning_rate": 1.0841734084212713e-07, "loss": 0.571, "step": 22892 }, { "epoch": 2.73, "learning_rate": 1.0832373134626262e-07, "loss": 0.6665, "step": 22893 }, { "epoch": 2.73, "learning_rate": 1.0823016138500785e-07, "loss": 0.4762, "step": 22894 }, { "epoch": 2.73, "learning_rate": 1.0813663095990884e-07, "loss": 0.3574, "step": 22895 }, { "epoch": 2.73, "learning_rate": 1.0804314007251182e-07, "loss": 0.4594, "step": 22896 }, { "epoch": 2.73, "learning_rate": 1.0794968872436252e-07, "loss": 0.3484, "step": 22897 }, { "epoch": 2.73, "learning_rate": 1.0785627691700607e-07, "loss": 0.5257, "step": 22898 }, { "epoch": 2.73, "learning_rate": 1.077629046519854e-07, "loss": 0.618, "step": 22899 }, { "epoch": 2.73, "learning_rate": 1.0766957193084515e-07, "loss": 0.5408, "step": 22900 }, { "epoch": 2.73, "learning_rate": 1.0757627875512738e-07, "loss": 0.4888, "step": 22901 }, { "epoch": 2.73, "learning_rate": 1.0748302512637476e-07, "loss": 0.3724, "step": 22902 }, { "epoch": 2.73, "learning_rate": 1.07389811046128e-07, "loss": 0.4217, "step": 22903 }, { "epoch": 2.73, "learning_rate": 1.0729663651592864e-07, "loss": 0.4903, "step": 22904 }, { "epoch": 2.73, "learning_rate": 1.0720350153731685e-07, "loss": 0.5742, "step": 22905 }, { "epoch": 2.73, "learning_rate": 1.0711040611183166e-07, "loss": 0.6671, "step": 22906 }, { "epoch": 2.73, "learning_rate": 1.070173502410124e-07, "loss": 0.4249, "step": 22907 }, { "epoch": 2.73, "learning_rate": 1.0692433392639701e-07, "loss": 0.4547, "step": 22908 }, { "epoch": 2.73, "learning_rate": 1.0683135716952342e-07, "loss": 0.4643, "step": 22909 }, { "epoch": 2.73, "learning_rate": 1.0673841997192791e-07, "loss": 0.5006, "step": 22910 }, { "epoch": 2.73, "learning_rate": 1.066455223351473e-07, "loss": 0.6077, "step": 22911 }, { "epoch": 2.73, "learning_rate": 1.0655266426071675e-07, "loss": 0.6208, "step": 22912 }, { "epoch": 2.73, "learning_rate": 1.0645984575017199e-07, "loss": 0.2843, "step": 22913 }, { "epoch": 2.73, "learning_rate": 1.0636706680504705e-07, "loss": 0.3995, "step": 22914 }, { "epoch": 2.73, "learning_rate": 1.062743274268746e-07, "loss": 0.4055, "step": 22915 }, { "epoch": 2.73, "learning_rate": 1.0618162761718869e-07, "loss": 0.4107, "step": 22916 }, { "epoch": 2.73, "learning_rate": 1.0608896737752145e-07, "loss": 0.5533, "step": 22917 }, { "epoch": 2.73, "learning_rate": 1.0599634670940413e-07, "loss": 0.623, "step": 22918 }, { "epoch": 2.73, "learning_rate": 1.0590376561436827e-07, "loss": 0.4668, "step": 22919 }, { "epoch": 2.73, "learning_rate": 1.0581122409394407e-07, "loss": 0.3205, "step": 22920 }, { "epoch": 2.73, "learning_rate": 1.0571872214966138e-07, "loss": 0.4137, "step": 22921 }, { "epoch": 2.73, "learning_rate": 1.0562625978304875e-07, "loss": 0.3209, "step": 22922 }, { "epoch": 2.73, "learning_rate": 1.0553383699563491e-07, "loss": 0.5095, "step": 22923 }, { "epoch": 2.73, "learning_rate": 1.054414537889481e-07, "loss": 0.4955, "step": 22924 }, { "epoch": 2.73, "learning_rate": 1.053491101645146e-07, "loss": 0.6662, "step": 22925 }, { "epoch": 2.73, "learning_rate": 1.0525680612386124e-07, "loss": 0.349, "step": 22926 }, { "epoch": 2.73, "learning_rate": 1.0516454166851404e-07, "loss": 0.4984, "step": 22927 }, { "epoch": 2.73, "learning_rate": 1.0507231679999785e-07, "loss": 0.4194, "step": 22928 }, { "epoch": 2.73, "learning_rate": 1.0498013151983733e-07, "loss": 0.4422, "step": 22929 }, { "epoch": 2.73, "learning_rate": 1.0488798582955595e-07, "loss": 0.55, "step": 22930 }, { "epoch": 2.73, "learning_rate": 1.0479587973067723e-07, "loss": 0.5379, "step": 22931 }, { "epoch": 2.73, "learning_rate": 1.0470381322472356e-07, "loss": 0.3418, "step": 22932 }, { "epoch": 2.73, "learning_rate": 1.0461178631321678e-07, "loss": 0.396, "step": 22933 }, { "epoch": 2.73, "learning_rate": 1.0451979899767817e-07, "loss": 0.423, "step": 22934 }, { "epoch": 2.73, "learning_rate": 1.0442785127962846e-07, "loss": 0.3506, "step": 22935 }, { "epoch": 2.73, "learning_rate": 1.0433594316058754e-07, "loss": 0.4567, "step": 22936 }, { "epoch": 2.73, "learning_rate": 1.042440746420742e-07, "loss": 0.6194, "step": 22937 }, { "epoch": 2.73, "learning_rate": 1.0415224572560722e-07, "loss": 0.329, "step": 22938 }, { "epoch": 2.73, "learning_rate": 1.0406045641270485e-07, "loss": 0.3768, "step": 22939 }, { "epoch": 2.73, "learning_rate": 1.039687067048839e-07, "loss": 0.4448, "step": 22940 }, { "epoch": 2.73, "learning_rate": 1.0387699660366152e-07, "loss": 0.4614, "step": 22941 }, { "epoch": 2.73, "learning_rate": 1.0378532611055342e-07, "loss": 0.6037, "step": 22942 }, { "epoch": 2.73, "learning_rate": 1.0369369522707506e-07, "loss": 0.5764, "step": 22943 }, { "epoch": 2.73, "learning_rate": 1.0360210395474079e-07, "loss": 0.459, "step": 22944 }, { "epoch": 2.73, "learning_rate": 1.0351055229506468e-07, "loss": 0.306, "step": 22945 }, { "epoch": 2.73, "learning_rate": 1.0341904024956023e-07, "loss": 0.358, "step": 22946 }, { "epoch": 2.73, "learning_rate": 1.0332756781974069e-07, "loss": 0.458, "step": 22947 }, { "epoch": 2.73, "learning_rate": 1.0323613500711705e-07, "loss": 0.4742, "step": 22948 }, { "epoch": 2.73, "learning_rate": 1.0314474181320145e-07, "loss": 0.598, "step": 22949 }, { "epoch": 2.73, "learning_rate": 1.0305338823950434e-07, "loss": 0.5527, "step": 22950 }, { "epoch": 2.73, "learning_rate": 1.0296207428753563e-07, "loss": 0.3602, "step": 22951 }, { "epoch": 2.73, "learning_rate": 1.0287079995880495e-07, "loss": 0.4032, "step": 22952 }, { "epoch": 2.73, "learning_rate": 1.0277956525482108e-07, "loss": 0.3676, "step": 22953 }, { "epoch": 2.73, "learning_rate": 1.0268837017709254e-07, "loss": 0.4717, "step": 22954 }, { "epoch": 2.73, "learning_rate": 1.0259721472712592e-07, "loss": 0.658, "step": 22955 }, { "epoch": 2.73, "learning_rate": 1.025060989064286e-07, "loss": 0.5356, "step": 22956 }, { "epoch": 2.73, "learning_rate": 1.0241502271650689e-07, "loss": 0.4594, "step": 22957 }, { "epoch": 2.73, "learning_rate": 1.0232398615886624e-07, "loss": 0.3286, "step": 22958 }, { "epoch": 2.73, "learning_rate": 1.0223298923501073e-07, "loss": 0.4164, "step": 22959 }, { "epoch": 2.73, "learning_rate": 1.02142031946445e-07, "loss": 0.4418, "step": 22960 }, { "epoch": 2.73, "learning_rate": 1.0205111429467284e-07, "loss": 0.5319, "step": 22961 }, { "epoch": 2.73, "learning_rate": 1.019602362811975e-07, "loss": 0.609, "step": 22962 }, { "epoch": 2.73, "learning_rate": 1.0186939790751999e-07, "loss": 0.4178, "step": 22963 }, { "epoch": 2.73, "learning_rate": 1.0177859917514299e-07, "loss": 0.3553, "step": 22964 }, { "epoch": 2.73, "learning_rate": 1.016878400855667e-07, "loss": 0.4038, "step": 22965 }, { "epoch": 2.74, "learning_rate": 1.0159712064029186e-07, "loss": 0.4425, "step": 22966 }, { "epoch": 2.74, "learning_rate": 1.0150644084081785e-07, "loss": 0.5964, "step": 22967 }, { "epoch": 2.74, "learning_rate": 1.0141580068864315e-07, "loss": 0.6275, "step": 22968 }, { "epoch": 2.74, "learning_rate": 1.0132520018526715e-07, "loss": 0.5341, "step": 22969 }, { "epoch": 2.74, "learning_rate": 1.0123463933218641e-07, "loss": 0.4018, "step": 22970 }, { "epoch": 2.74, "learning_rate": 1.0114411813089891e-07, "loss": 0.4283, "step": 22971 }, { "epoch": 2.74, "learning_rate": 1.0105363658289985e-07, "loss": 0.4444, "step": 22972 }, { "epoch": 2.74, "learning_rate": 1.009631946896858e-07, "loss": 0.4098, "step": 22973 }, { "epoch": 2.74, "learning_rate": 1.0087279245275111e-07, "loss": 0.5576, "step": 22974 }, { "epoch": 2.74, "learning_rate": 1.0078242987359072e-07, "loss": 0.5502, "step": 22975 }, { "epoch": 2.74, "learning_rate": 1.0069210695369786e-07, "loss": 0.2751, "step": 22976 }, { "epoch": 2.74, "learning_rate": 1.0060182369456606e-07, "loss": 0.4512, "step": 22977 }, { "epoch": 2.74, "learning_rate": 1.0051158009768747e-07, "loss": 0.395, "step": 22978 }, { "epoch": 2.74, "learning_rate": 1.0042137616455366e-07, "loss": 0.3895, "step": 22979 }, { "epoch": 2.74, "learning_rate": 1.0033121189665568e-07, "loss": 0.6164, "step": 22980 }, { "epoch": 2.74, "learning_rate": 1.0024108729548453e-07, "loss": 0.6315, "step": 22981 }, { "epoch": 2.74, "learning_rate": 1.0015100236252906e-07, "loss": 0.454, "step": 22982 }, { "epoch": 2.74, "learning_rate": 1.0006095709927888e-07, "loss": 0.2801, "step": 22983 }, { "epoch": 2.74, "learning_rate": 9.997095150722309e-08, "loss": 0.4149, "step": 22984 }, { "epoch": 2.74, "learning_rate": 9.988098558784826e-08, "loss": 0.4283, "step": 22985 }, { "epoch": 2.74, "learning_rate": 9.979105934264239e-08, "loss": 0.4645, "step": 22986 }, { "epoch": 2.74, "learning_rate": 9.970117277309149e-08, "loss": 0.703, "step": 22987 }, { "epoch": 2.74, "learning_rate": 9.961132588068217e-08, "loss": 0.5426, "step": 22988 }, { "epoch": 2.74, "learning_rate": 9.952151866689852e-08, "loss": 0.3538, "step": 22989 }, { "epoch": 2.74, "learning_rate": 9.943175113322545e-08, "loss": 0.3796, "step": 22990 }, { "epoch": 2.74, "learning_rate": 9.934202328114706e-08, "loss": 0.4077, "step": 22991 }, { "epoch": 2.74, "learning_rate": 9.925233511214688e-08, "loss": 0.5701, "step": 22992 }, { "epoch": 2.74, "learning_rate": 9.916268662770679e-08, "loss": 0.6406, "step": 22993 }, { "epoch": 2.74, "learning_rate": 9.907307782930892e-08, "loss": 0.4735, "step": 22994 }, { "epoch": 2.74, "learning_rate": 9.898350871843432e-08, "loss": 0.3801, "step": 22995 }, { "epoch": 2.74, "learning_rate": 9.889397929656402e-08, "loss": 0.4461, "step": 22996 }, { "epoch": 2.74, "learning_rate": 9.880448956517768e-08, "loss": 0.412, "step": 22997 }, { "epoch": 2.74, "learning_rate": 9.87150395257544e-08, "loss": 0.5439, "step": 22998 }, { "epoch": 2.74, "learning_rate": 9.862562917977353e-08, "loss": 0.542, "step": 22999 }, { "epoch": 2.74, "learning_rate": 9.853625852871223e-08, "loss": 0.6924, "step": 23000 }, { "epoch": 2.74, "learning_rate": 9.844692757404795e-08, "loss": 0.3676, "step": 23001 }, { "epoch": 2.74, "learning_rate": 9.835763631725725e-08, "loss": 0.3909, "step": 23002 }, { "epoch": 2.74, "learning_rate": 9.826838475981676e-08, "loss": 0.421, "step": 23003 }, { "epoch": 2.74, "learning_rate": 9.817917290320112e-08, "loss": 0.4849, "step": 23004 }, { "epoch": 2.74, "learning_rate": 9.809000074888525e-08, "loss": 0.6174, "step": 23005 }, { "epoch": 2.74, "learning_rate": 9.800086829834355e-08, "loss": 0.6523, "step": 23006 }, { "epoch": 2.74, "learning_rate": 9.791177555304899e-08, "loss": 0.2825, "step": 23007 }, { "epoch": 2.74, "learning_rate": 9.782272251447427e-08, "loss": 0.4535, "step": 23008 }, { "epoch": 2.74, "learning_rate": 9.77337091840913e-08, "loss": 0.4272, "step": 23009 }, { "epoch": 2.74, "learning_rate": 9.764473556337168e-08, "loss": 0.411, "step": 23010 }, { "epoch": 2.74, "learning_rate": 9.755580165378641e-08, "loss": 0.544, "step": 23011 }, { "epoch": 2.74, "learning_rate": 9.746690745680492e-08, "loss": 0.6252, "step": 23012 }, { "epoch": 2.74, "learning_rate": 9.73780529738974e-08, "loss": 0.2804, "step": 23013 }, { "epoch": 2.74, "learning_rate": 9.728923820653213e-08, "loss": 0.3394, "step": 23014 }, { "epoch": 2.74, "learning_rate": 9.720046315617765e-08, "loss": 0.4086, "step": 23015 }, { "epoch": 2.74, "learning_rate": 9.711172782430084e-08, "loss": 0.4898, "step": 23016 }, { "epoch": 2.74, "learning_rate": 9.702303221236858e-08, "loss": 0.6031, "step": 23017 }, { "epoch": 2.74, "learning_rate": 9.693437632184777e-08, "loss": 0.5739, "step": 23018 }, { "epoch": 2.74, "learning_rate": 9.684576015420277e-08, "loss": 0.5504, "step": 23019 }, { "epoch": 2.74, "learning_rate": 9.675718371089937e-08, "loss": 0.3208, "step": 23020 }, { "epoch": 2.74, "learning_rate": 9.66686469934014e-08, "loss": 0.4409, "step": 23021 }, { "epoch": 2.74, "learning_rate": 9.658015000317239e-08, "loss": 0.3829, "step": 23022 }, { "epoch": 2.74, "learning_rate": 9.64916927416748e-08, "loss": 0.5155, "step": 23023 }, { "epoch": 2.74, "learning_rate": 9.640327521037135e-08, "loss": 0.5771, "step": 23024 }, { "epoch": 2.74, "learning_rate": 9.631489741072336e-08, "loss": 0.5863, "step": 23025 }, { "epoch": 2.74, "learning_rate": 9.622655934419218e-08, "loss": 0.2716, "step": 23026 }, { "epoch": 2.74, "learning_rate": 9.613826101223717e-08, "loss": 0.4363, "step": 23027 }, { "epoch": 2.74, "learning_rate": 9.605000241631885e-08, "loss": 0.4029, "step": 23028 }, { "epoch": 2.74, "learning_rate": 9.596178355789577e-08, "loss": 0.4425, "step": 23029 }, { "epoch": 2.74, "learning_rate": 9.587360443842564e-08, "loss": 0.6367, "step": 23030 }, { "epoch": 2.74, "learning_rate": 9.578546505936675e-08, "loss": 0.5391, "step": 23031 }, { "epoch": 2.74, "learning_rate": 9.569736542217573e-08, "loss": 0.3468, "step": 23032 }, { "epoch": 2.74, "learning_rate": 9.560930552830943e-08, "loss": 0.3464, "step": 23033 }, { "epoch": 2.74, "learning_rate": 9.552128537922257e-08, "loss": 0.4155, "step": 23034 }, { "epoch": 2.74, "learning_rate": 9.543330497637088e-08, "loss": 0.5182, "step": 23035 }, { "epoch": 2.74, "learning_rate": 9.534536432120823e-08, "loss": 0.6401, "step": 23036 }, { "epoch": 2.74, "learning_rate": 9.525746341518871e-08, "loss": 0.6914, "step": 23037 }, { "epoch": 2.74, "learning_rate": 9.516960225976507e-08, "loss": 0.3673, "step": 23038 }, { "epoch": 2.74, "learning_rate": 9.508178085638947e-08, "loss": 0.3055, "step": 23039 }, { "epoch": 2.74, "learning_rate": 9.499399920651381e-08, "loss": 0.3849, "step": 23040 }, { "epoch": 2.74, "learning_rate": 9.490625731158942e-08, "loss": 0.5689, "step": 23041 }, { "epoch": 2.74, "learning_rate": 9.481855517306598e-08, "loss": 0.5112, "step": 23042 }, { "epoch": 2.74, "learning_rate": 9.473089279239427e-08, "loss": 0.6026, "step": 23043 }, { "epoch": 2.74, "learning_rate": 9.464327017102204e-08, "loss": 0.6244, "step": 23044 }, { "epoch": 2.74, "learning_rate": 9.455568731039894e-08, "loss": 0.2816, "step": 23045 }, { "epoch": 2.74, "learning_rate": 9.446814421197186e-08, "loss": 0.3992, "step": 23046 }, { "epoch": 2.74, "learning_rate": 9.438064087718802e-08, "loss": 0.435, "step": 23047 }, { "epoch": 2.74, "learning_rate": 9.429317730749455e-08, "loss": 0.4799, "step": 23048 }, { "epoch": 2.74, "learning_rate": 9.420575350433642e-08, "loss": 0.5763, "step": 23049 }, { "epoch": 2.75, "learning_rate": 9.411836946915915e-08, "loss": 0.5217, "step": 23050 }, { "epoch": 2.75, "learning_rate": 9.403102520340685e-08, "loss": 0.3138, "step": 23051 }, { "epoch": 2.75, "learning_rate": 9.394372070852392e-08, "loss": 0.5205, "step": 23052 }, { "epoch": 2.75, "learning_rate": 9.385645598595284e-08, "loss": 0.4492, "step": 23053 }, { "epoch": 2.75, "learning_rate": 9.376923103713659e-08, "loss": 0.4767, "step": 23054 }, { "epoch": 2.75, "learning_rate": 9.368204586351681e-08, "loss": 0.5841, "step": 23055 }, { "epoch": 2.75, "learning_rate": 9.359490046653485e-08, "loss": 0.5306, "step": 23056 }, { "epoch": 2.75, "learning_rate": 9.350779484763123e-08, "loss": 0.3542, "step": 23057 }, { "epoch": 2.75, "learning_rate": 9.342072900824533e-08, "loss": 0.4112, "step": 23058 }, { "epoch": 2.75, "learning_rate": 9.333370294981686e-08, "loss": 0.459, "step": 23059 }, { "epoch": 2.75, "learning_rate": 9.324671667378438e-08, "loss": 0.5153, "step": 23060 }, { "epoch": 2.75, "learning_rate": 9.315977018158506e-08, "loss": 0.5646, "step": 23061 }, { "epoch": 2.75, "learning_rate": 9.307286347465722e-08, "loss": 0.5569, "step": 23062 }, { "epoch": 2.75, "learning_rate": 9.298599655443691e-08, "loss": 0.4278, "step": 23063 }, { "epoch": 2.75, "learning_rate": 9.289916942235994e-08, "loss": 0.3847, "step": 23064 }, { "epoch": 2.75, "learning_rate": 9.281238207986154e-08, "loss": 0.3644, "step": 23065 }, { "epoch": 2.75, "learning_rate": 9.27256345283764e-08, "loss": 0.4963, "step": 23066 }, { "epoch": 2.75, "learning_rate": 9.263892676933866e-08, "loss": 0.6631, "step": 23067 }, { "epoch": 2.75, "learning_rate": 9.255225880418134e-08, "loss": 0.5881, "step": 23068 }, { "epoch": 2.75, "learning_rate": 9.246563063433716e-08, "loss": 0.5784, "step": 23069 }, { "epoch": 2.75, "learning_rate": 9.237904226123807e-08, "loss": 0.356, "step": 23070 }, { "epoch": 2.75, "learning_rate": 9.229249368631593e-08, "loss": 0.4431, "step": 23071 }, { "epoch": 2.75, "learning_rate": 9.220598491100075e-08, "loss": 0.4643, "step": 23072 }, { "epoch": 2.75, "learning_rate": 9.211951593672219e-08, "loss": 0.517, "step": 23073 }, { "epoch": 2.75, "learning_rate": 9.203308676491024e-08, "loss": 0.5199, "step": 23074 }, { "epoch": 2.75, "learning_rate": 9.194669739699375e-08, "loss": 0.672, "step": 23075 }, { "epoch": 2.75, "learning_rate": 9.186034783439989e-08, "loss": 0.2709, "step": 23076 }, { "epoch": 2.75, "learning_rate": 9.177403807855672e-08, "loss": 0.4437, "step": 23077 }, { "epoch": 2.75, "learning_rate": 9.168776813089114e-08, "loss": 0.4848, "step": 23078 }, { "epoch": 2.75, "learning_rate": 9.160153799282867e-08, "loss": 0.4536, "step": 23079 }, { "epoch": 2.75, "learning_rate": 9.151534766579456e-08, "loss": 0.5697, "step": 23080 }, { "epoch": 2.75, "learning_rate": 9.142919715121407e-08, "loss": 0.6335, "step": 23081 }, { "epoch": 2.75, "learning_rate": 9.134308645051104e-08, "loss": 0.3383, "step": 23082 }, { "epoch": 2.75, "learning_rate": 9.12570155651088e-08, "loss": 0.3684, "step": 23083 }, { "epoch": 2.75, "learning_rate": 9.117098449643009e-08, "loss": 0.3761, "step": 23084 }, { "epoch": 2.75, "learning_rate": 9.108499324589765e-08, "loss": 0.5067, "step": 23085 }, { "epoch": 2.75, "learning_rate": 9.099904181493202e-08, "loss": 0.52, "step": 23086 }, { "epoch": 2.75, "learning_rate": 9.09131302049543e-08, "loss": 0.6013, "step": 23087 }, { "epoch": 2.75, "learning_rate": 9.082725841738472e-08, "loss": 0.4205, "step": 23088 }, { "epoch": 2.75, "learning_rate": 9.074142645364242e-08, "loss": 0.3294, "step": 23089 }, { "epoch": 2.75, "learning_rate": 9.065563431514713e-08, "loss": 0.3581, "step": 23090 }, { "epoch": 2.75, "learning_rate": 9.056988200331601e-08, "loss": 0.3624, "step": 23091 }, { "epoch": 2.75, "learning_rate": 9.048416951956712e-08, "loss": 0.4649, "step": 23092 }, { "epoch": 2.75, "learning_rate": 9.039849686531709e-08, "loss": 0.5818, "step": 23093 }, { "epoch": 2.75, "learning_rate": 9.031286404198175e-08, "loss": 0.4504, "step": 23094 }, { "epoch": 2.75, "learning_rate": 9.02272710509769e-08, "loss": 0.2738, "step": 23095 }, { "epoch": 2.75, "learning_rate": 9.01417178937175e-08, "loss": 0.3782, "step": 23096 }, { "epoch": 2.75, "learning_rate": 9.005620457161801e-08, "loss": 0.3784, "step": 23097 }, { "epoch": 2.75, "learning_rate": 8.997073108609145e-08, "loss": 0.517, "step": 23098 }, { "epoch": 2.75, "learning_rate": 8.988529743855085e-08, "loss": 0.647, "step": 23099 }, { "epoch": 2.75, "learning_rate": 8.979990363040897e-08, "loss": 0.362, "step": 23100 }, { "epoch": 2.75, "learning_rate": 8.971454966307691e-08, "loss": 0.3873, "step": 23101 }, { "epoch": 2.75, "learning_rate": 8.962923553796493e-08, "loss": 0.3631, "step": 23102 }, { "epoch": 2.75, "learning_rate": 8.95439612564844e-08, "loss": 0.4108, "step": 23103 }, { "epoch": 2.75, "learning_rate": 8.94587268200442e-08, "loss": 0.52, "step": 23104 }, { "epoch": 2.75, "learning_rate": 8.937353223005374e-08, "loss": 0.5823, "step": 23105 }, { "epoch": 2.75, "learning_rate": 8.92883774879208e-08, "loss": 0.6436, "step": 23106 }, { "epoch": 2.75, "learning_rate": 8.920326259505369e-08, "loss": 0.3026, "step": 23107 }, { "epoch": 2.75, "learning_rate": 8.91181875528585e-08, "loss": 0.339, "step": 23108 }, { "epoch": 2.75, "learning_rate": 8.903315236274219e-08, "loss": 0.3653, "step": 23109 }, { "epoch": 2.75, "learning_rate": 8.894815702610999e-08, "loss": 0.438, "step": 23110 }, { "epoch": 2.75, "learning_rate": 8.886320154436717e-08, "loss": 0.5624, "step": 23111 }, { "epoch": 2.75, "learning_rate": 8.877828591891819e-08, "loss": 0.6417, "step": 23112 }, { "epoch": 2.75, "learning_rate": 8.869341015116606e-08, "loss": 0.413, "step": 23113 }, { "epoch": 2.75, "learning_rate": 8.86085742425144e-08, "loss": 0.297, "step": 23114 }, { "epoch": 2.75, "learning_rate": 8.852377819436514e-08, "loss": 0.3885, "step": 23115 }, { "epoch": 2.75, "learning_rate": 8.84390220081205e-08, "loss": 0.4406, "step": 23116 }, { "epoch": 2.75, "learning_rate": 8.835430568518072e-08, "loss": 0.6008, "step": 23117 }, { "epoch": 2.75, "learning_rate": 8.826962922694693e-08, "loss": 0.5917, "step": 23118 }, { "epoch": 2.75, "learning_rate": 8.818499263481828e-08, "loss": 0.4997, "step": 23119 }, { "epoch": 2.75, "learning_rate": 8.810039591019448e-08, "loss": 0.2934, "step": 23120 }, { "epoch": 2.75, "learning_rate": 8.801583905447331e-08, "loss": 0.4234, "step": 23121 }, { "epoch": 2.75, "learning_rate": 8.793132206905281e-08, "loss": 0.4183, "step": 23122 }, { "epoch": 2.75, "learning_rate": 8.784684495532964e-08, "loss": 0.5005, "step": 23123 }, { "epoch": 2.75, "learning_rate": 8.776240771470073e-08, "loss": 0.5412, "step": 23124 }, { "epoch": 2.75, "learning_rate": 8.767801034856166e-08, "loss": 0.7305, "step": 23125 }, { "epoch": 2.75, "learning_rate": 8.759365285830712e-08, "loss": 0.2556, "step": 23126 }, { "epoch": 2.75, "learning_rate": 8.750933524533239e-08, "loss": 0.3436, "step": 23127 }, { "epoch": 2.75, "learning_rate": 8.742505751103053e-08, "loss": 0.3917, "step": 23128 }, { "epoch": 2.75, "learning_rate": 8.734081965679514e-08, "loss": 0.4758, "step": 23129 }, { "epoch": 2.75, "learning_rate": 8.725662168401816e-08, "loss": 0.5238, "step": 23130 }, { "epoch": 2.75, "learning_rate": 8.71724635940921e-08, "loss": 0.5708, "step": 23131 }, { "epoch": 2.75, "learning_rate": 8.708834538840721e-08, "loss": 0.31, "step": 23132 }, { "epoch": 2.75, "learning_rate": 8.700426706835436e-08, "loss": 0.3349, "step": 23133 }, { "epoch": 2.76, "learning_rate": 8.692022863532351e-08, "loss": 0.4125, "step": 23134 }, { "epoch": 2.76, "learning_rate": 8.683623009070414e-08, "loss": 0.4631, "step": 23135 }, { "epoch": 2.76, "learning_rate": 8.675227143588428e-08, "loss": 0.5571, "step": 23136 }, { "epoch": 2.76, "learning_rate": 8.666835267225171e-08, "loss": 0.4615, "step": 23137 }, { "epoch": 2.76, "learning_rate": 8.658447380119367e-08, "loss": 0.3777, "step": 23138 }, { "epoch": 2.76, "learning_rate": 8.650063482409681e-08, "loss": 0.4215, "step": 23139 }, { "epoch": 2.76, "learning_rate": 8.641683574234699e-08, "loss": 0.3912, "step": 23140 }, { "epoch": 2.76, "learning_rate": 8.633307655732948e-08, "loss": 0.4956, "step": 23141 }, { "epoch": 2.76, "learning_rate": 8.624935727042871e-08, "loss": 0.4931, "step": 23142 }, { "epoch": 2.76, "learning_rate": 8.616567788302888e-08, "loss": 0.6328, "step": 23143 }, { "epoch": 2.76, "learning_rate": 8.608203839651246e-08, "loss": 0.5096, "step": 23144 }, { "epoch": 2.76, "learning_rate": 8.599843881226255e-08, "loss": 0.3047, "step": 23145 }, { "epoch": 2.76, "learning_rate": 8.591487913166108e-08, "loss": 0.4249, "step": 23146 }, { "epoch": 2.76, "learning_rate": 8.583135935608916e-08, "loss": 0.4338, "step": 23147 }, { "epoch": 2.76, "learning_rate": 8.574787948692736e-08, "loss": 0.5304, "step": 23148 }, { "epoch": 2.76, "learning_rate": 8.566443952555625e-08, "loss": 0.5937, "step": 23149 }, { "epoch": 2.76, "learning_rate": 8.558103947335417e-08, "loss": 0.6017, "step": 23150 }, { "epoch": 2.76, "learning_rate": 8.54976793317e-08, "loss": 0.316, "step": 23151 }, { "epoch": 2.76, "learning_rate": 8.541435910197155e-08, "loss": 0.3704, "step": 23152 }, { "epoch": 2.76, "learning_rate": 8.533107878554658e-08, "loss": 0.4602, "step": 23153 }, { "epoch": 2.76, "learning_rate": 8.524783838380179e-08, "loss": 0.4762, "step": 23154 }, { "epoch": 2.76, "learning_rate": 8.516463789811246e-08, "loss": 0.6328, "step": 23155 }, { "epoch": 2.76, "learning_rate": 8.508147732985417e-08, "loss": 0.5701, "step": 23156 }, { "epoch": 2.76, "learning_rate": 8.499835668040219e-08, "loss": 0.274, "step": 23157 }, { "epoch": 2.76, "learning_rate": 8.491527595112986e-08, "loss": 0.4432, "step": 23158 }, { "epoch": 2.76, "learning_rate": 8.483223514341055e-08, "loss": 0.4394, "step": 23159 }, { "epoch": 2.76, "learning_rate": 8.474923425861703e-08, "loss": 0.4843, "step": 23160 }, { "epoch": 2.76, "learning_rate": 8.466627329812128e-08, "loss": 0.5544, "step": 23161 }, { "epoch": 2.76, "learning_rate": 8.458335226329468e-08, "loss": 0.5813, "step": 23162 }, { "epoch": 2.76, "learning_rate": 8.450047115550807e-08, "loss": 0.4641, "step": 23163 }, { "epoch": 2.76, "learning_rate": 8.441762997613124e-08, "loss": 0.2527, "step": 23164 }, { "epoch": 2.76, "learning_rate": 8.433482872653415e-08, "loss": 0.4305, "step": 23165 }, { "epoch": 2.76, "learning_rate": 8.425206740808434e-08, "loss": 0.5115, "step": 23166 }, { "epoch": 2.76, "learning_rate": 8.416934602215071e-08, "loss": 0.5487, "step": 23167 }, { "epoch": 2.76, "learning_rate": 8.40866645701005e-08, "loss": 0.7036, "step": 23168 }, { "epoch": 2.76, "learning_rate": 8.400402305330069e-08, "loss": 0.6772, "step": 23169 }, { "epoch": 2.76, "learning_rate": 8.392142147311682e-08, "loss": 0.3157, "step": 23170 }, { "epoch": 2.76, "learning_rate": 8.383885983091505e-08, "loss": 0.3535, "step": 23171 }, { "epoch": 2.76, "learning_rate": 8.375633812805928e-08, "loss": 0.4188, "step": 23172 }, { "epoch": 2.76, "learning_rate": 8.367385636591396e-08, "loss": 0.4925, "step": 23173 }, { "epoch": 2.76, "learning_rate": 8.359141454584246e-08, "loss": 0.6613, "step": 23174 }, { "epoch": 2.76, "learning_rate": 8.350901266920758e-08, "loss": 0.467, "step": 23175 }, { "epoch": 2.76, "learning_rate": 8.342665073737182e-08, "loss": 0.3705, "step": 23176 }, { "epoch": 2.76, "learning_rate": 8.334432875169607e-08, "loss": 0.3752, "step": 23177 }, { "epoch": 2.76, "learning_rate": 8.326204671354171e-08, "loss": 0.3842, "step": 23178 }, { "epoch": 2.76, "learning_rate": 8.317980462426795e-08, "loss": 0.4552, "step": 23179 }, { "epoch": 2.76, "learning_rate": 8.309760248523535e-08, "loss": 0.6064, "step": 23180 }, { "epoch": 2.76, "learning_rate": 8.301544029780174e-08, "loss": 0.5759, "step": 23181 }, { "epoch": 2.76, "learning_rate": 8.293331806332571e-08, "loss": 0.3954, "step": 23182 }, { "epoch": 2.76, "learning_rate": 8.285123578316485e-08, "loss": 0.3789, "step": 23183 }, { "epoch": 2.76, "learning_rate": 8.276919345867607e-08, "loss": 0.4258, "step": 23184 }, { "epoch": 2.76, "learning_rate": 8.268719109121498e-08, "loss": 0.5332, "step": 23185 }, { "epoch": 2.76, "learning_rate": 8.260522868213772e-08, "loss": 0.5619, "step": 23186 }, { "epoch": 2.76, "learning_rate": 8.252330623279875e-08, "loss": 0.595, "step": 23187 }, { "epoch": 2.76, "learning_rate": 8.244142374455255e-08, "loss": 0.3605, "step": 23188 }, { "epoch": 2.76, "learning_rate": 8.235958121875221e-08, "loss": 0.3666, "step": 23189 }, { "epoch": 2.76, "learning_rate": 8.227777865675108e-08, "loss": 0.4124, "step": 23190 }, { "epoch": 2.76, "learning_rate": 8.219601605990113e-08, "loss": 0.469, "step": 23191 }, { "epoch": 2.76, "learning_rate": 8.211429342955351e-08, "loss": 0.5519, "step": 23192 }, { "epoch": 2.76, "learning_rate": 8.203261076705993e-08, "loss": 0.6052, "step": 23193 }, { "epoch": 2.76, "learning_rate": 8.195096807377012e-08, "loss": 0.4121, "step": 23194 }, { "epoch": 2.76, "learning_rate": 8.186936535103384e-08, "loss": 0.321, "step": 23195 }, { "epoch": 2.76, "learning_rate": 8.178780260019947e-08, "loss": 0.3499, "step": 23196 }, { "epoch": 2.76, "learning_rate": 8.170627982261564e-08, "loss": 0.4346, "step": 23197 }, { "epoch": 2.76, "learning_rate": 8.162479701962989e-08, "loss": 0.4803, "step": 23198 }, { "epoch": 2.76, "learning_rate": 8.154335419258946e-08, "loss": 0.5195, "step": 23199 }, { "epoch": 2.76, "learning_rate": 8.146195134284052e-08, "loss": 0.5485, "step": 23200 }, { "epoch": 2.76, "learning_rate": 8.138058847172781e-08, "loss": 0.3102, "step": 23201 }, { "epoch": 2.76, "learning_rate": 8.12992655805972e-08, "loss": 0.4367, "step": 23202 }, { "epoch": 2.76, "learning_rate": 8.12179826707929e-08, "loss": 0.389, "step": 23203 }, { "epoch": 2.76, "learning_rate": 8.113673974365799e-08, "loss": 0.4933, "step": 23204 }, { "epoch": 2.76, "learning_rate": 8.105553680053557e-08, "loss": 0.5168, "step": 23205 }, { "epoch": 2.76, "learning_rate": 8.097437384276846e-08, "loss": 0.5583, "step": 23206 }, { "epoch": 2.76, "learning_rate": 8.089325087169809e-08, "loss": 0.2313, "step": 23207 }, { "epoch": 2.76, "learning_rate": 8.081216788866503e-08, "loss": 0.3292, "step": 23208 }, { "epoch": 2.76, "learning_rate": 8.073112489500962e-08, "loss": 0.4945, "step": 23209 }, { "epoch": 2.76, "learning_rate": 8.065012189207189e-08, "loss": 0.3844, "step": 23210 }, { "epoch": 2.76, "learning_rate": 8.056915888119049e-08, "loss": 0.6152, "step": 23211 }, { "epoch": 2.76, "learning_rate": 8.048823586370408e-08, "loss": 0.6175, "step": 23212 }, { "epoch": 2.76, "learning_rate": 8.040735284094992e-08, "loss": 0.3414, "step": 23213 }, { "epoch": 2.76, "learning_rate": 8.032650981426554e-08, "loss": 0.3507, "step": 23214 }, { "epoch": 2.76, "learning_rate": 8.02457067849871e-08, "loss": 0.3983, "step": 23215 }, { "epoch": 2.76, "learning_rate": 8.016494375444994e-08, "loss": 0.4765, "step": 23216 }, { "epoch": 2.76, "learning_rate": 8.008422072398908e-08, "loss": 0.5349, "step": 23217 }, { "epoch": 2.77, "learning_rate": 8.00035376949393e-08, "loss": 0.5828, "step": 23218 }, { "epoch": 2.77, "learning_rate": 7.992289466863424e-08, "loss": 0.5594, "step": 23219 }, { "epoch": 2.77, "learning_rate": 7.984229164640645e-08, "loss": 0.2667, "step": 23220 }, { "epoch": 2.77, "learning_rate": 7.976172862958903e-08, "loss": 0.5564, "step": 23221 }, { "epoch": 2.77, "learning_rate": 7.968120561951315e-08, "loss": 0.4441, "step": 23222 }, { "epoch": 2.77, "learning_rate": 7.960072261750995e-08, "loss": 0.4743, "step": 23223 }, { "epoch": 2.77, "learning_rate": 7.952027962490977e-08, "loss": 0.6488, "step": 23224 }, { "epoch": 2.77, "learning_rate": 7.943987664304292e-08, "loss": 0.6612, "step": 23225 }, { "epoch": 2.77, "learning_rate": 7.935951367323752e-08, "loss": 0.3763, "step": 23226 }, { "epoch": 2.77, "learning_rate": 7.927919071682277e-08, "loss": 0.3537, "step": 23227 }, { "epoch": 2.77, "learning_rate": 7.919890777512595e-08, "loss": 0.3222, "step": 23228 }, { "epoch": 2.77, "learning_rate": 7.911866484947462e-08, "loss": 0.514, "step": 23229 }, { "epoch": 2.77, "learning_rate": 7.903846194119464e-08, "loss": 0.5798, "step": 23230 }, { "epoch": 2.77, "learning_rate": 7.89582990516119e-08, "loss": 0.6783, "step": 23231 }, { "epoch": 2.77, "learning_rate": 7.887817618205146e-08, "loss": 0.2889, "step": 23232 }, { "epoch": 2.77, "learning_rate": 7.879809333383836e-08, "loss": 0.369, "step": 23233 }, { "epoch": 2.77, "learning_rate": 7.871805050829572e-08, "loss": 0.3595, "step": 23234 }, { "epoch": 2.77, "learning_rate": 7.863804770674721e-08, "loss": 0.4061, "step": 23235 }, { "epoch": 2.77, "learning_rate": 7.855808493051482e-08, "loss": 0.5634, "step": 23236 }, { "epoch": 2.77, "learning_rate": 7.847816218092024e-08, "loss": 0.6299, "step": 23237 }, { "epoch": 2.77, "learning_rate": 7.839827945928468e-08, "loss": 0.3003, "step": 23238 }, { "epoch": 2.77, "learning_rate": 7.831843676692874e-08, "loss": 0.3624, "step": 23239 }, { "epoch": 2.77, "learning_rate": 7.823863410517275e-08, "loss": 0.4365, "step": 23240 }, { "epoch": 2.77, "learning_rate": 7.81588714753348e-08, "loss": 0.4313, "step": 23241 }, { "epoch": 2.77, "learning_rate": 7.807914887873414e-08, "loss": 0.5059, "step": 23242 }, { "epoch": 2.77, "learning_rate": 7.799946631668859e-08, "loss": 0.6426, "step": 23243 }, { "epoch": 2.77, "learning_rate": 7.791982379051488e-08, "loss": 0.6189, "step": 23244 }, { "epoch": 2.77, "learning_rate": 7.784022130152946e-08, "loss": 0.3851, "step": 23245 }, { "epoch": 2.77, "learning_rate": 7.776065885104878e-08, "loss": 0.3605, "step": 23246 }, { "epoch": 2.77, "learning_rate": 7.768113644038733e-08, "loss": 0.4736, "step": 23247 }, { "epoch": 2.77, "learning_rate": 7.760165407086018e-08, "loss": 0.5225, "step": 23248 }, { "epoch": 2.77, "learning_rate": 7.752221174378099e-08, "loss": 0.481, "step": 23249 }, { "epoch": 2.77, "learning_rate": 7.74428094604629e-08, "loss": 0.5045, "step": 23250 }, { "epoch": 2.77, "learning_rate": 7.736344722221816e-08, "loss": 0.2581, "step": 23251 }, { "epoch": 2.77, "learning_rate": 7.728412503035936e-08, "loss": 0.3215, "step": 23252 }, { "epoch": 2.77, "learning_rate": 7.720484288619684e-08, "loss": 0.4475, "step": 23253 }, { "epoch": 2.77, "learning_rate": 7.712560079104176e-08, "loss": 0.4647, "step": 23254 }, { "epoch": 2.77, "learning_rate": 7.70463987462039e-08, "loss": 0.5683, "step": 23255 }, { "epoch": 2.77, "learning_rate": 7.696723675299195e-08, "loss": 0.5827, "step": 23256 }, { "epoch": 2.77, "learning_rate": 7.68881148127154e-08, "loss": 0.2991, "step": 23257 }, { "epoch": 2.77, "learning_rate": 7.680903292668129e-08, "loss": 0.3888, "step": 23258 }, { "epoch": 2.77, "learning_rate": 7.672999109619745e-08, "loss": 0.3972, "step": 23259 }, { "epoch": 2.77, "learning_rate": 7.665098932257004e-08, "loss": 0.4381, "step": 23260 }, { "epoch": 2.77, "learning_rate": 7.657202760710497e-08, "loss": 0.6588, "step": 23261 }, { "epoch": 2.77, "learning_rate": 7.64931059511076e-08, "loss": 0.6818, "step": 23262 }, { "epoch": 2.77, "learning_rate": 7.64142243558827e-08, "loss": 0.4681, "step": 23263 }, { "epoch": 2.77, "learning_rate": 7.633538282273423e-08, "loss": 0.2903, "step": 23264 }, { "epoch": 2.77, "learning_rate": 7.625658135296504e-08, "loss": 0.4044, "step": 23265 }, { "epoch": 2.77, "learning_rate": 7.617781994787771e-08, "loss": 0.4134, "step": 23266 }, { "epoch": 2.77, "learning_rate": 7.609909860877479e-08, "loss": 0.586, "step": 23267 }, { "epoch": 2.77, "learning_rate": 7.602041733695693e-08, "loss": 0.5962, "step": 23268 }, { "epoch": 2.77, "learning_rate": 7.594177613372472e-08, "loss": 0.6039, "step": 23269 }, { "epoch": 2.77, "learning_rate": 7.586317500037882e-08, "loss": 0.2186, "step": 23270 }, { "epoch": 2.77, "learning_rate": 7.57846139382179e-08, "loss": 0.4087, "step": 23271 }, { "epoch": 2.77, "learning_rate": 7.570609294854064e-08, "loss": 0.4183, "step": 23272 }, { "epoch": 2.77, "learning_rate": 7.562761203264518e-08, "loss": 0.4011, "step": 23273 }, { "epoch": 2.77, "learning_rate": 7.55491711918288e-08, "loss": 0.6586, "step": 23274 }, { "epoch": 2.77, "learning_rate": 7.54707704273877e-08, "loss": 0.6114, "step": 23275 }, { "epoch": 2.77, "learning_rate": 7.539240974061834e-08, "loss": 0.2044, "step": 23276 }, { "epoch": 2.77, "learning_rate": 7.53140891328158e-08, "loss": 0.3599, "step": 23277 }, { "epoch": 2.77, "learning_rate": 7.523580860527512e-08, "loss": 0.4308, "step": 23278 }, { "epoch": 2.77, "learning_rate": 7.515756815928976e-08, "loss": 0.5251, "step": 23279 }, { "epoch": 2.77, "learning_rate": 7.507936779615337e-08, "loss": 0.6111, "step": 23280 }, { "epoch": 2.77, "learning_rate": 7.500120751715828e-08, "loss": 0.5753, "step": 23281 }, { "epoch": 2.77, "learning_rate": 7.492308732359704e-08, "loss": 0.3192, "step": 23282 }, { "epoch": 2.77, "learning_rate": 7.48450072167603e-08, "loss": 0.3826, "step": 23283 }, { "epoch": 2.77, "learning_rate": 7.476696719793925e-08, "loss": 0.4187, "step": 23284 }, { "epoch": 2.77, "learning_rate": 7.468896726842367e-08, "loss": 0.4116, "step": 23285 }, { "epoch": 2.77, "learning_rate": 7.461100742950311e-08, "loss": 0.527, "step": 23286 }, { "epoch": 2.77, "learning_rate": 7.453308768246598e-08, "loss": 0.6129, "step": 23287 }, { "epoch": 2.77, "learning_rate": 7.44552080286004e-08, "loss": 0.3107, "step": 23288 }, { "epoch": 2.77, "learning_rate": 7.437736846919397e-08, "loss": 0.3241, "step": 23289 }, { "epoch": 2.77, "learning_rate": 7.429956900553287e-08, "loss": 0.3933, "step": 23290 }, { "epoch": 2.77, "learning_rate": 7.42218096389033e-08, "loss": 0.4518, "step": 23291 }, { "epoch": 2.77, "learning_rate": 7.414409037059117e-08, "loss": 0.4666, "step": 23292 }, { "epoch": 2.77, "learning_rate": 7.406641120188102e-08, "loss": 0.6324, "step": 23293 }, { "epoch": 2.77, "learning_rate": 7.398877213405597e-08, "loss": 0.445, "step": 23294 }, { "epoch": 2.77, "learning_rate": 7.39111731684003e-08, "loss": 0.361, "step": 23295 }, { "epoch": 2.77, "learning_rate": 7.383361430619628e-08, "loss": 0.4064, "step": 23296 }, { "epoch": 2.77, "learning_rate": 7.375609554872682e-08, "loss": 0.4048, "step": 23297 }, { "epoch": 2.77, "learning_rate": 7.367861689727224e-08, "loss": 0.5178, "step": 23298 }, { "epoch": 2.77, "learning_rate": 7.360117835311347e-08, "loss": 0.5423, "step": 23299 }, { "epoch": 2.77, "learning_rate": 7.352377991753146e-08, "loss": 0.5997, "step": 23300 }, { "epoch": 2.77, "learning_rate": 7.34464215918046e-08, "loss": 0.367, "step": 23301 }, { "epoch": 2.78, "learning_rate": 7.336910337721187e-08, "loss": 0.4742, "step": 23302 }, { "epoch": 2.78, "learning_rate": 7.32918252750317e-08, "loss": 0.3616, "step": 23303 }, { "epoch": 2.78, "learning_rate": 7.321458728654141e-08, "loss": 0.4944, "step": 23304 }, { "epoch": 2.78, "learning_rate": 7.313738941301718e-08, "loss": 0.6581, "step": 23305 }, { "epoch": 2.78, "learning_rate": 7.306023165573578e-08, "loss": 0.5247, "step": 23306 }, { "epoch": 2.78, "learning_rate": 7.29831140159723e-08, "loss": 0.2964, "step": 23307 }, { "epoch": 2.78, "learning_rate": 7.290603649500184e-08, "loss": 0.3645, "step": 23308 }, { "epoch": 2.78, "learning_rate": 7.282899909409807e-08, "loss": 0.3816, "step": 23309 }, { "epoch": 2.78, "learning_rate": 7.275200181453446e-08, "loss": 0.4164, "step": 23310 }, { "epoch": 2.78, "learning_rate": 7.267504465758385e-08, "loss": 0.6102, "step": 23311 }, { "epoch": 2.78, "learning_rate": 7.259812762451884e-08, "loss": 0.6149, "step": 23312 }, { "epoch": 2.78, "learning_rate": 7.25212507166101e-08, "loss": 0.3701, "step": 23313 }, { "epoch": 2.78, "learning_rate": 7.244441393512907e-08, "loss": 0.3413, "step": 23314 }, { "epoch": 2.78, "learning_rate": 7.236761728134533e-08, "loss": 0.3889, "step": 23315 }, { "epoch": 2.78, "learning_rate": 7.229086075652896e-08, "loss": 0.4053, "step": 23316 }, { "epoch": 2.78, "learning_rate": 7.221414436194785e-08, "loss": 0.606, "step": 23317 }, { "epoch": 2.78, "learning_rate": 7.213746809887067e-08, "loss": 0.6418, "step": 23318 }, { "epoch": 2.78, "learning_rate": 7.206083196856534e-08, "loss": 0.5125, "step": 23319 }, { "epoch": 2.78, "learning_rate": 7.198423597229776e-08, "loss": 0.433, "step": 23320 }, { "epoch": 2.78, "learning_rate": 7.19076801113347e-08, "loss": 0.364, "step": 23321 }, { "epoch": 2.78, "learning_rate": 7.183116438694126e-08, "loss": 0.4665, "step": 23322 }, { "epoch": 2.78, "learning_rate": 7.175468880038256e-08, "loss": 0.5219, "step": 23323 }, { "epoch": 2.78, "learning_rate": 7.167825335292256e-08, "loss": 0.5851, "step": 23324 }, { "epoch": 2.78, "learning_rate": 7.160185804582443e-08, "loss": 0.6089, "step": 23325 }, { "epoch": 2.78, "learning_rate": 7.152550288035159e-08, "loss": 0.2928, "step": 23326 }, { "epoch": 2.78, "learning_rate": 7.144918785776611e-08, "loss": 0.4129, "step": 23327 }, { "epoch": 2.78, "learning_rate": 7.13729129793292e-08, "loss": 0.4199, "step": 23328 }, { "epoch": 2.78, "learning_rate": 7.129667824630204e-08, "loss": 0.468, "step": 23329 }, { "epoch": 2.78, "learning_rate": 7.122048365994422e-08, "loss": 0.6512, "step": 23330 }, { "epoch": 2.78, "learning_rate": 7.114432922151609e-08, "loss": 0.6185, "step": 23331 }, { "epoch": 2.78, "learning_rate": 7.106821493227556e-08, "loss": 0.3469, "step": 23332 }, { "epoch": 2.78, "learning_rate": 7.099214079348105e-08, "loss": 0.3168, "step": 23333 }, { "epoch": 2.78, "learning_rate": 7.091610680639072e-08, "loss": 0.3774, "step": 23334 }, { "epoch": 2.78, "learning_rate": 7.084011297226079e-08, "loss": 0.3921, "step": 23335 }, { "epoch": 2.78, "learning_rate": 7.076415929234776e-08, "loss": 0.4974, "step": 23336 }, { "epoch": 2.78, "learning_rate": 7.068824576790672e-08, "loss": 0.5642, "step": 23337 }, { "epoch": 2.78, "learning_rate": 7.061237240019309e-08, "loss": 0.4755, "step": 23338 }, { "epoch": 2.78, "learning_rate": 7.053653919046055e-08, "loss": 0.3634, "step": 23339 }, { "epoch": 2.78, "learning_rate": 7.046074613996312e-08, "loss": 0.4497, "step": 23340 }, { "epoch": 2.78, "learning_rate": 7.038499324995313e-08, "loss": 0.4585, "step": 23341 }, { "epoch": 2.78, "learning_rate": 7.030928052168318e-08, "loss": 0.3664, "step": 23342 }, { "epoch": 2.78, "learning_rate": 7.023360795640505e-08, "loss": 0.5346, "step": 23343 }, { "epoch": 2.78, "learning_rate": 7.015797555536885e-08, "loss": 0.5966, "step": 23344 }, { "epoch": 2.78, "learning_rate": 7.008238331982498e-08, "loss": 0.2671, "step": 23345 }, { "epoch": 2.78, "learning_rate": 7.000683125102381e-08, "loss": 0.3783, "step": 23346 }, { "epoch": 2.78, "learning_rate": 6.993131935021297e-08, "loss": 0.3976, "step": 23347 }, { "epoch": 2.78, "learning_rate": 6.985584761864172e-08, "loss": 0.4864, "step": 23348 }, { "epoch": 2.78, "learning_rate": 6.978041605755714e-08, "loss": 0.6599, "step": 23349 }, { "epoch": 2.78, "learning_rate": 6.970502466820655e-08, "loss": 0.5895, "step": 23350 }, { "epoch": 2.78, "learning_rate": 6.962967345183535e-08, "loss": 0.2471, "step": 23351 }, { "epoch": 2.78, "learning_rate": 6.955436240968949e-08, "loss": 0.4331, "step": 23352 }, { "epoch": 2.78, "learning_rate": 6.947909154301436e-08, "loss": 0.4237, "step": 23353 }, { "epoch": 2.78, "learning_rate": 6.940386085305339e-08, "loss": 0.5176, "step": 23354 }, { "epoch": 2.78, "learning_rate": 6.932867034105062e-08, "loss": 0.5781, "step": 23355 }, { "epoch": 2.78, "learning_rate": 6.925352000824892e-08, "loss": 0.5111, "step": 23356 }, { "epoch": 2.78, "learning_rate": 6.917840985589064e-08, "loss": 0.2463, "step": 23357 }, { "epoch": 2.78, "learning_rate": 6.910333988521728e-08, "loss": 0.4224, "step": 23358 }, { "epoch": 2.78, "learning_rate": 6.902831009746952e-08, "loss": 0.5468, "step": 23359 }, { "epoch": 2.78, "learning_rate": 6.895332049388747e-08, "loss": 0.4948, "step": 23360 }, { "epoch": 2.78, "learning_rate": 6.887837107571155e-08, "loss": 0.5759, "step": 23361 }, { "epoch": 2.78, "learning_rate": 6.88034618441799e-08, "loss": 0.6048, "step": 23362 }, { "epoch": 2.78, "learning_rate": 6.872859280053102e-08, "loss": 0.4567, "step": 23363 }, { "epoch": 2.78, "learning_rate": 6.865376394600249e-08, "loss": 0.3101, "step": 23364 }, { "epoch": 2.78, "learning_rate": 6.857897528183139e-08, "loss": 0.3789, "step": 23365 }, { "epoch": 2.78, "learning_rate": 6.850422680925366e-08, "loss": 0.4461, "step": 23366 }, { "epoch": 2.78, "learning_rate": 6.842951852950503e-08, "loss": 0.5206, "step": 23367 }, { "epoch": 2.78, "learning_rate": 6.835485044382084e-08, "loss": 0.6862, "step": 23368 }, { "epoch": 2.78, "learning_rate": 6.828022255343458e-08, "loss": 0.6101, "step": 23369 }, { "epoch": 2.78, "learning_rate": 6.820563485958026e-08, "loss": 0.3157, "step": 23370 }, { "epoch": 2.78, "learning_rate": 6.813108736349134e-08, "loss": 0.4005, "step": 23371 }, { "epoch": 2.78, "learning_rate": 6.805658006639932e-08, "loss": 0.3131, "step": 23372 }, { "epoch": 2.78, "learning_rate": 6.7982112969536e-08, "loss": 0.525, "step": 23373 }, { "epoch": 2.78, "learning_rate": 6.790768607413234e-08, "loss": 0.5749, "step": 23374 }, { "epoch": 2.78, "learning_rate": 6.783329938141875e-08, "loss": 0.4925, "step": 23375 }, { "epoch": 2.78, "learning_rate": 6.775895289262507e-08, "loss": 0.3586, "step": 23376 }, { "epoch": 2.78, "learning_rate": 6.768464660897978e-08, "loss": 0.4197, "step": 23377 }, { "epoch": 2.78, "learning_rate": 6.761038053171159e-08, "loss": 0.4664, "step": 23378 }, { "epoch": 2.78, "learning_rate": 6.753615466204788e-08, "loss": 0.4864, "step": 23379 }, { "epoch": 2.78, "learning_rate": 6.746196900121515e-08, "loss": 0.447, "step": 23380 }, { "epoch": 2.78, "learning_rate": 6.738782355044048e-08, "loss": 0.6106, "step": 23381 }, { "epoch": 2.78, "learning_rate": 6.731371831094901e-08, "loss": 0.3749, "step": 23382 }, { "epoch": 2.78, "learning_rate": 6.723965328396642e-08, "loss": 0.346, "step": 23383 }, { "epoch": 2.78, "learning_rate": 6.71656284707159e-08, "loss": 0.4343, "step": 23384 }, { "epoch": 2.78, "learning_rate": 6.709164387242201e-08, "loss": 0.4571, "step": 23385 }, { "epoch": 2.79, "learning_rate": 6.70176994903074e-08, "loss": 0.5104, "step": 23386 }, { "epoch": 2.79, "learning_rate": 6.694379532559441e-08, "loss": 0.6218, "step": 23387 }, { "epoch": 2.79, "learning_rate": 6.686993137950431e-08, "loss": 0.3172, "step": 23388 }, { "epoch": 2.79, "learning_rate": 6.679610765325861e-08, "loss": 0.3927, "step": 23389 }, { "epoch": 2.79, "learning_rate": 6.672232414807717e-08, "loss": 0.4308, "step": 23390 }, { "epoch": 2.79, "learning_rate": 6.664858086518039e-08, "loss": 0.4283, "step": 23391 }, { "epoch": 2.79, "learning_rate": 6.65748778057862e-08, "loss": 0.5149, "step": 23392 }, { "epoch": 2.79, "learning_rate": 6.65012149711139e-08, "loss": 0.6336, "step": 23393 }, { "epoch": 2.79, "learning_rate": 6.642759236238028e-08, "loss": 0.5257, "step": 23394 }, { "epoch": 2.79, "learning_rate": 6.635400998080327e-08, "loss": 0.2994, "step": 23395 }, { "epoch": 2.79, "learning_rate": 6.628046782759829e-08, "loss": 0.3548, "step": 23396 }, { "epoch": 2.79, "learning_rate": 6.620696590398129e-08, "loss": 0.3935, "step": 23397 }, { "epoch": 2.79, "learning_rate": 6.613350421116771e-08, "loss": 0.5569, "step": 23398 }, { "epoch": 2.79, "learning_rate": 6.606008275037128e-08, "loss": 0.5902, "step": 23399 }, { "epoch": 2.79, "learning_rate": 6.598670152280607e-08, "loss": 0.6146, "step": 23400 }, { "epoch": 2.79, "learning_rate": 6.591336052968494e-08, "loss": 0.3555, "step": 23401 }, { "epoch": 2.79, "learning_rate": 6.584005977222002e-08, "loss": 0.2953, "step": 23402 }, { "epoch": 2.79, "learning_rate": 6.576679925162311e-08, "loss": 0.4012, "step": 23403 }, { "epoch": 2.79, "learning_rate": 6.569357896910544e-08, "loss": 0.4402, "step": 23404 }, { "epoch": 2.79, "learning_rate": 6.562039892587691e-08, "loss": 0.5149, "step": 23405 }, { "epoch": 2.79, "learning_rate": 6.554725912314791e-08, "loss": 0.6233, "step": 23406 }, { "epoch": 2.79, "learning_rate": 6.547415956212665e-08, "loss": 0.4156, "step": 23407 }, { "epoch": 2.79, "learning_rate": 6.54011002440219e-08, "loss": 0.4213, "step": 23408 }, { "epoch": 2.79, "learning_rate": 6.532808117004102e-08, "loss": 0.4189, "step": 23409 }, { "epoch": 2.79, "learning_rate": 6.525510234139165e-08, "loss": 0.3273, "step": 23410 }, { "epoch": 2.79, "learning_rate": 6.518216375927949e-08, "loss": 0.5878, "step": 23411 }, { "epoch": 2.79, "learning_rate": 6.510926542491025e-08, "loss": 0.5925, "step": 23412 }, { "epoch": 2.79, "learning_rate": 6.503640733948962e-08, "loss": 0.4291, "step": 23413 }, { "epoch": 2.79, "learning_rate": 6.49635895042211e-08, "loss": 0.2968, "step": 23414 }, { "epoch": 2.79, "learning_rate": 6.4890811920309e-08, "loss": 0.4498, "step": 23415 }, { "epoch": 2.79, "learning_rate": 6.481807458895595e-08, "loss": 0.4154, "step": 23416 }, { "epoch": 2.79, "learning_rate": 6.474537751136461e-08, "loss": 0.5807, "step": 23417 }, { "epoch": 2.79, "learning_rate": 6.467272068873653e-08, "loss": 0.6535, "step": 23418 }, { "epoch": 2.79, "learning_rate": 6.460010412227241e-08, "loss": 0.5208, "step": 23419 }, { "epoch": 2.79, "learning_rate": 6.452752781317295e-08, "loss": 0.3398, "step": 23420 }, { "epoch": 2.79, "learning_rate": 6.445499176263831e-08, "loss": 0.4553, "step": 23421 }, { "epoch": 2.79, "learning_rate": 6.438249597186669e-08, "loss": 0.3782, "step": 23422 }, { "epoch": 2.79, "learning_rate": 6.431004044205658e-08, "loss": 0.4739, "step": 23423 }, { "epoch": 2.79, "learning_rate": 6.423762517440619e-08, "loss": 0.5664, "step": 23424 }, { "epoch": 2.79, "learning_rate": 6.416525017011205e-08, "loss": 0.6227, "step": 23425 }, { "epoch": 2.79, "learning_rate": 6.409291543037072e-08, "loss": 0.3937, "step": 23426 }, { "epoch": 2.79, "learning_rate": 6.40206209563779e-08, "loss": 0.3665, "step": 23427 }, { "epoch": 2.79, "learning_rate": 6.394836674932874e-08, "loss": 0.5351, "step": 23428 }, { "epoch": 2.79, "learning_rate": 6.38761528104176e-08, "loss": 0.476, "step": 23429 }, { "epoch": 2.79, "learning_rate": 6.380397914083763e-08, "loss": 0.5774, "step": 23430 }, { "epoch": 2.79, "learning_rate": 6.373184574178238e-08, "loss": 0.6919, "step": 23431 }, { "epoch": 2.79, "learning_rate": 6.365975261444445e-08, "loss": 0.242, "step": 23432 }, { "epoch": 2.79, "learning_rate": 6.358769976001517e-08, "loss": 0.2963, "step": 23433 }, { "epoch": 2.79, "learning_rate": 6.35156871796852e-08, "loss": 0.4769, "step": 23434 }, { "epoch": 2.79, "learning_rate": 6.344371487464584e-08, "loss": 0.4114, "step": 23435 }, { "epoch": 2.79, "learning_rate": 6.337178284608641e-08, "loss": 0.5338, "step": 23436 }, { "epoch": 2.79, "learning_rate": 6.32998910951954e-08, "loss": 0.6958, "step": 23437 }, { "epoch": 2.79, "learning_rate": 6.322803962316187e-08, "loss": 0.3816, "step": 23438 }, { "epoch": 2.79, "learning_rate": 6.31562284311732e-08, "loss": 0.3732, "step": 23439 }, { "epoch": 2.79, "learning_rate": 6.308445752041675e-08, "loss": 0.4496, "step": 23440 }, { "epoch": 2.79, "learning_rate": 6.301272689207855e-08, "loss": 0.4566, "step": 23441 }, { "epoch": 2.79, "learning_rate": 6.294103654734429e-08, "loss": 0.4658, "step": 23442 }, { "epoch": 2.79, "learning_rate": 6.286938648739971e-08, "loss": 0.6396, "step": 23443 }, { "epoch": 2.79, "learning_rate": 6.27977767134283e-08, "loss": 0.4704, "step": 23444 }, { "epoch": 2.79, "learning_rate": 6.27262072266141e-08, "loss": 0.3352, "step": 23445 }, { "epoch": 2.79, "learning_rate": 6.265467802814035e-08, "loss": 0.3807, "step": 23446 }, { "epoch": 2.79, "learning_rate": 6.258318911918915e-08, "loss": 0.4331, "step": 23447 }, { "epoch": 2.79, "learning_rate": 6.251174050094233e-08, "loss": 0.5066, "step": 23448 }, { "epoch": 2.79, "learning_rate": 6.24403321745809e-08, "loss": 0.5404, "step": 23449 }, { "epoch": 2.79, "learning_rate": 6.236896414128585e-08, "loss": 0.5594, "step": 23450 }, { "epoch": 2.79, "learning_rate": 6.229763640223596e-08, "loss": 0.3645, "step": 23451 }, { "epoch": 2.79, "learning_rate": 6.222634895861057e-08, "loss": 0.3792, "step": 23452 }, { "epoch": 2.79, "learning_rate": 6.215510181158818e-08, "loss": 0.4079, "step": 23453 }, { "epoch": 2.79, "learning_rate": 6.208389496234645e-08, "loss": 0.4699, "step": 23454 }, { "epoch": 2.79, "learning_rate": 6.20127284120628e-08, "loss": 0.613, "step": 23455 }, { "epoch": 2.79, "learning_rate": 6.194160216191319e-08, "loss": 0.6682, "step": 23456 }, { "epoch": 2.79, "learning_rate": 6.187051621307366e-08, "loss": 0.2504, "step": 23457 }, { "epoch": 2.79, "learning_rate": 6.179947056671853e-08, "loss": 0.2841, "step": 23458 }, { "epoch": 2.79, "learning_rate": 6.172846522402326e-08, "loss": 0.4181, "step": 23459 }, { "epoch": 2.79, "learning_rate": 6.165750018616079e-08, "loss": 0.3867, "step": 23460 }, { "epoch": 2.79, "learning_rate": 6.158657545430435e-08, "loss": 0.6106, "step": 23461 }, { "epoch": 2.79, "learning_rate": 6.151569102962691e-08, "loss": 0.5074, "step": 23462 }, { "epoch": 2.79, "learning_rate": 6.144484691329916e-08, "loss": 0.3085, "step": 23463 }, { "epoch": 2.79, "learning_rate": 6.137404310649298e-08, "loss": 0.3494, "step": 23464 }, { "epoch": 2.79, "learning_rate": 6.130327961037824e-08, "loss": 0.4458, "step": 23465 }, { "epoch": 2.79, "learning_rate": 6.123255642612541e-08, "loss": 0.5853, "step": 23466 }, { "epoch": 2.79, "learning_rate": 6.116187355490245e-08, "loss": 0.5334, "step": 23467 }, { "epoch": 2.79, "learning_rate": 6.109123099787866e-08, "loss": 0.644, "step": 23468 }, { "epoch": 2.79, "learning_rate": 6.102062875622122e-08, "loss": 0.4609, "step": 23469 }, { "epoch": 2.8, "learning_rate": 6.095006683109777e-08, "loss": 0.3494, "step": 23470 }, { "epoch": 2.8, "learning_rate": 6.087954522367406e-08, "loss": 0.4587, "step": 23471 }, { "epoch": 2.8, "learning_rate": 6.080906393511665e-08, "loss": 0.3804, "step": 23472 }, { "epoch": 2.8, "learning_rate": 6.073862296658961e-08, "loss": 0.5595, "step": 23473 }, { "epoch": 2.8, "learning_rate": 6.066822231925812e-08, "loss": 0.5873, "step": 23474 }, { "epoch": 2.8, "learning_rate": 6.059786199428513e-08, "loss": 0.6619, "step": 23475 }, { "epoch": 2.8, "learning_rate": 6.052754199283445e-08, "loss": 0.3301, "step": 23476 }, { "epoch": 2.8, "learning_rate": 6.04572623160682e-08, "loss": 0.4149, "step": 23477 }, { "epoch": 2.8, "learning_rate": 6.038702296514792e-08, "loss": 0.3779, "step": 23478 }, { "epoch": 2.8, "learning_rate": 6.031682394123523e-08, "loss": 0.5121, "step": 23479 }, { "epoch": 2.8, "learning_rate": 6.024666524548972e-08, "loss": 0.5182, "step": 23480 }, { "epoch": 2.8, "learning_rate": 6.017654687907187e-08, "loss": 0.5643, "step": 23481 }, { "epoch": 2.8, "learning_rate": 6.01064688431402e-08, "loss": 0.351, "step": 23482 }, { "epoch": 2.8, "learning_rate": 6.003643113885349e-08, "loss": 0.387, "step": 23483 }, { "epoch": 2.8, "learning_rate": 5.996643376736916e-08, "loss": 0.3949, "step": 23484 }, { "epoch": 2.8, "learning_rate": 5.98964767298446e-08, "loss": 0.4904, "step": 23485 }, { "epoch": 2.8, "learning_rate": 5.982656002743586e-08, "loss": 0.5562, "step": 23486 }, { "epoch": 2.8, "learning_rate": 5.975668366129894e-08, "loss": 0.5485, "step": 23487 }, { "epoch": 2.8, "learning_rate": 5.968684763258848e-08, "loss": 0.3021, "step": 23488 }, { "epoch": 2.8, "learning_rate": 5.961705194245965e-08, "loss": 0.3562, "step": 23489 }, { "epoch": 2.8, "learning_rate": 5.954729659206543e-08, "loss": 0.3426, "step": 23490 }, { "epoch": 2.8, "learning_rate": 5.9477581582559066e-08, "loss": 0.5188, "step": 23491 }, { "epoch": 2.8, "learning_rate": 5.9407906915093515e-08, "loss": 0.4872, "step": 23492 }, { "epoch": 2.8, "learning_rate": 5.933827259081981e-08, "loss": 0.5173, "step": 23493 }, { "epoch": 2.8, "learning_rate": 5.926867861088897e-08, "loss": 0.4976, "step": 23494 }, { "epoch": 2.8, "learning_rate": 5.9199124976452024e-08, "loss": 0.244, "step": 23495 }, { "epoch": 2.8, "learning_rate": 5.9129611688658326e-08, "loss": 0.3381, "step": 23496 }, { "epoch": 2.8, "learning_rate": 5.906013874865668e-08, "loss": 0.425, "step": 23497 }, { "epoch": 2.8, "learning_rate": 5.899070615759589e-08, "loss": 0.4927, "step": 23498 }, { "epoch": 2.8, "learning_rate": 5.892131391662337e-08, "loss": 0.4789, "step": 23499 }, { "epoch": 2.8, "learning_rate": 5.885196202688681e-08, "loss": 0.6336, "step": 23500 }, { "epoch": 2.8, "learning_rate": 5.878265048953197e-08, "loss": 0.3528, "step": 23501 }, { "epoch": 2.8, "learning_rate": 5.871337930570459e-08, "loss": 0.3461, "step": 23502 }, { "epoch": 2.8, "learning_rate": 5.864414847654987e-08, "loss": 0.3466, "step": 23503 }, { "epoch": 2.8, "learning_rate": 5.8574958003212445e-08, "loss": 0.4251, "step": 23504 }, { "epoch": 2.8, "learning_rate": 5.850580788683558e-08, "loss": 0.4865, "step": 23505 }, { "epoch": 2.8, "learning_rate": 5.843669812856251e-08, "loss": 0.6022, "step": 23506 }, { "epoch": 2.8, "learning_rate": 5.836762872953594e-08, "loss": 0.2839, "step": 23507 }, { "epoch": 2.8, "learning_rate": 5.8298599690897464e-08, "loss": 0.3824, "step": 23508 }, { "epoch": 2.8, "learning_rate": 5.822961101378782e-08, "loss": 0.4169, "step": 23509 }, { "epoch": 2.8, "learning_rate": 5.8160662699347504e-08, "loss": 0.3918, "step": 23510 }, { "epoch": 2.8, "learning_rate": 5.809175474871642e-08, "loss": 0.5123, "step": 23511 }, { "epoch": 2.8, "learning_rate": 5.802288716303339e-08, "loss": 0.6729, "step": 23512 }, { "epoch": 2.8, "learning_rate": 5.795405994343695e-08, "loss": 0.3664, "step": 23513 }, { "epoch": 2.8, "learning_rate": 5.7885273091065064e-08, "loss": 0.2733, "step": 23514 }, { "epoch": 2.8, "learning_rate": 5.7816526607054334e-08, "loss": 0.4773, "step": 23515 }, { "epoch": 2.8, "learning_rate": 5.774782049254135e-08, "loss": 0.416, "step": 23516 }, { "epoch": 2.8, "learning_rate": 5.767915474866159e-08, "loss": 0.4451, "step": 23517 }, { "epoch": 2.8, "learning_rate": 5.761052937655054e-08, "loss": 0.5966, "step": 23518 }, { "epoch": 2.8, "learning_rate": 5.754194437734256e-08, "loss": 0.4964, "step": 23519 }, { "epoch": 2.8, "learning_rate": 5.747339975217092e-08, "loss": 0.3076, "step": 23520 }, { "epoch": 2.8, "learning_rate": 5.7404895502169154e-08, "loss": 0.3867, "step": 23521 }, { "epoch": 2.8, "learning_rate": 5.7336431628469416e-08, "loss": 0.4822, "step": 23522 }, { "epoch": 2.8, "learning_rate": 5.726800813220301e-08, "loss": 0.5148, "step": 23523 }, { "epoch": 2.8, "learning_rate": 5.719962501450155e-08, "loss": 0.5259, "step": 23524 }, { "epoch": 2.8, "learning_rate": 5.7131282276495504e-08, "loss": 0.6198, "step": 23525 }, { "epoch": 2.8, "learning_rate": 5.7062979919314255e-08, "loss": 0.3375, "step": 23526 }, { "epoch": 2.8, "learning_rate": 5.699471794408662e-08, "loss": 0.4778, "step": 23527 }, { "epoch": 2.8, "learning_rate": 5.692649635194142e-08, "loss": 0.4116, "step": 23528 }, { "epoch": 2.8, "learning_rate": 5.6858315144006635e-08, "loss": 0.4275, "step": 23529 }, { "epoch": 2.8, "learning_rate": 5.67901743214086e-08, "loss": 0.5937, "step": 23530 }, { "epoch": 2.8, "learning_rate": 5.67220738852739e-08, "loss": 0.5223, "step": 23531 }, { "epoch": 2.8, "learning_rate": 5.665401383672831e-08, "loss": 0.2758, "step": 23532 }, { "epoch": 2.8, "learning_rate": 5.658599417689703e-08, "loss": 0.4553, "step": 23533 }, { "epoch": 2.8, "learning_rate": 5.6518014906904453e-08, "loss": 0.3678, "step": 23534 }, { "epoch": 2.8, "learning_rate": 5.6450076027873555e-08, "loss": 0.4137, "step": 23535 }, { "epoch": 2.8, "learning_rate": 5.638217754092845e-08, "loss": 0.6403, "step": 23536 }, { "epoch": 2.8, "learning_rate": 5.6314319447190736e-08, "loss": 0.6257, "step": 23537 }, { "epoch": 2.8, "learning_rate": 5.624650174778257e-08, "loss": 0.4425, "step": 23538 }, { "epoch": 2.8, "learning_rate": 5.617872444382472e-08, "loss": 0.3297, "step": 23539 }, { "epoch": 2.8, "learning_rate": 5.61109875364374e-08, "loss": 0.4156, "step": 23540 }, { "epoch": 2.8, "learning_rate": 5.6043291026740835e-08, "loss": 0.4312, "step": 23541 }, { "epoch": 2.8, "learning_rate": 5.597563491585328e-08, "loss": 0.5107, "step": 23542 }, { "epoch": 2.8, "learning_rate": 5.5908019204894136e-08, "loss": 0.6066, "step": 23543 }, { "epoch": 2.8, "learning_rate": 5.584044389498e-08, "loss": 0.5994, "step": 23544 }, { "epoch": 2.8, "learning_rate": 5.577290898722887e-08, "loss": 0.3353, "step": 23545 }, { "epoch": 2.8, "learning_rate": 5.570541448275624e-08, "loss": 0.4854, "step": 23546 }, { "epoch": 2.8, "learning_rate": 5.5637960382678155e-08, "loss": 0.403, "step": 23547 }, { "epoch": 2.8, "learning_rate": 5.557054668810985e-08, "loss": 0.4826, "step": 23548 }, { "epoch": 2.8, "learning_rate": 5.55031734001657e-08, "loss": 0.6636, "step": 23549 }, { "epoch": 2.8, "learning_rate": 5.5435840519958986e-08, "loss": 0.5058, "step": 23550 }, { "epoch": 2.8, "learning_rate": 5.5368548048602975e-08, "loss": 0.2927, "step": 23551 }, { "epoch": 2.8, "learning_rate": 5.5301295987209844e-08, "loss": 0.3727, "step": 23552 }, { "epoch": 2.8, "learning_rate": 5.5234084336891755e-08, "loss": 0.4674, "step": 23553 }, { "epoch": 2.81, "learning_rate": 5.516691309875921e-08, "loss": 0.4829, "step": 23554 }, { "epoch": 2.81, "learning_rate": 5.5099782273922706e-08, "loss": 0.4144, "step": 23555 }, { "epoch": 2.81, "learning_rate": 5.5032691863491914e-08, "loss": 0.4834, "step": 23556 }, { "epoch": 2.81, "learning_rate": 5.4965641868575945e-08, "loss": 0.3837, "step": 23557 }, { "epoch": 2.81, "learning_rate": 5.4898632290283084e-08, "loss": 0.3619, "step": 23558 }, { "epoch": 2.81, "learning_rate": 5.4831663129721044e-08, "loss": 0.432, "step": 23559 }, { "epoch": 2.81, "learning_rate": 5.476473438799673e-08, "loss": 0.4584, "step": 23560 }, { "epoch": 2.81, "learning_rate": 5.469784606621648e-08, "loss": 0.5257, "step": 23561 }, { "epoch": 2.81, "learning_rate": 5.463099816548578e-08, "loss": 0.6083, "step": 23562 }, { "epoch": 2.81, "learning_rate": 5.4564190686910156e-08, "loss": 0.4272, "step": 23563 }, { "epoch": 2.81, "learning_rate": 5.449742363159372e-08, "loss": 0.4178, "step": 23564 }, { "epoch": 2.81, "learning_rate": 5.443069700064002e-08, "loss": 0.4306, "step": 23565 }, { "epoch": 2.81, "learning_rate": 5.436401079515208e-08, "loss": 0.4505, "step": 23566 }, { "epoch": 2.81, "learning_rate": 5.429736501623206e-08, "loss": 0.5204, "step": 23567 }, { "epoch": 2.81, "learning_rate": 5.4230759664982145e-08, "loss": 0.5856, "step": 23568 }, { "epoch": 2.81, "learning_rate": 5.416419474250256e-08, "loss": 0.5833, "step": 23569 }, { "epoch": 2.81, "learning_rate": 5.409767024989437e-08, "loss": 0.3032, "step": 23570 }, { "epoch": 2.81, "learning_rate": 5.40311861882567e-08, "loss": 0.4267, "step": 23571 }, { "epoch": 2.81, "learning_rate": 5.396474255868922e-08, "loss": 0.4642, "step": 23572 }, { "epoch": 2.81, "learning_rate": 5.389833936228939e-08, "loss": 0.4762, "step": 23573 }, { "epoch": 2.81, "learning_rate": 5.383197660015521e-08, "loss": 0.5598, "step": 23574 }, { "epoch": 2.81, "learning_rate": 5.376565427338387e-08, "loss": 0.4841, "step": 23575 }, { "epoch": 2.81, "learning_rate": 5.3699372383071434e-08, "loss": 0.2946, "step": 23576 }, { "epoch": 2.81, "learning_rate": 5.363313093031369e-08, "loss": 0.4737, "step": 23577 }, { "epoch": 2.81, "learning_rate": 5.35669299162056e-08, "loss": 0.4561, "step": 23578 }, { "epoch": 2.81, "learning_rate": 5.3500769341841566e-08, "loss": 0.4573, "step": 23579 }, { "epoch": 2.81, "learning_rate": 5.343464920831459e-08, "loss": 0.5835, "step": 23580 }, { "epoch": 2.81, "learning_rate": 5.336856951671854e-08, "loss": 0.621, "step": 23581 }, { "epoch": 2.81, "learning_rate": 5.330253026814503e-08, "loss": 0.2555, "step": 23582 }, { "epoch": 2.81, "learning_rate": 5.323653146368624e-08, "loss": 0.4008, "step": 23583 }, { "epoch": 2.81, "learning_rate": 5.317057310443297e-08, "loss": 0.4072, "step": 23584 }, { "epoch": 2.81, "learning_rate": 5.3104655191475177e-08, "loss": 0.4203, "step": 23585 }, { "epoch": 2.81, "learning_rate": 5.30387777259031e-08, "loss": 0.4509, "step": 23586 }, { "epoch": 2.81, "learning_rate": 5.297294070880504e-08, "loss": 0.5846, "step": 23587 }, { "epoch": 2.81, "learning_rate": 5.2907144141269576e-08, "loss": 0.3158, "step": 23588 }, { "epoch": 2.81, "learning_rate": 5.284138802438443e-08, "loss": 0.3195, "step": 23589 }, { "epoch": 2.81, "learning_rate": 5.277567235923653e-08, "loss": 0.3527, "step": 23590 }, { "epoch": 2.81, "learning_rate": 5.270999714691194e-08, "loss": 0.4863, "step": 23591 }, { "epoch": 2.81, "learning_rate": 5.264436238849674e-08, "loss": 0.4744, "step": 23592 }, { "epoch": 2.81, "learning_rate": 5.257876808507534e-08, "loss": 0.5479, "step": 23593 }, { "epoch": 2.81, "learning_rate": 5.2513214237732427e-08, "loss": 0.5238, "step": 23594 }, { "epoch": 2.81, "learning_rate": 5.2447700847551296e-08, "loss": 0.3127, "step": 23595 }, { "epoch": 2.81, "learning_rate": 5.238222791561498e-08, "loss": 0.3534, "step": 23596 }, { "epoch": 2.81, "learning_rate": 5.231679544300594e-08, "loss": 0.4049, "step": 23597 }, { "epoch": 2.81, "learning_rate": 5.2251403430805816e-08, "loss": 0.5033, "step": 23598 }, { "epoch": 2.81, "learning_rate": 5.218605188009485e-08, "loss": 0.5613, "step": 23599 }, { "epoch": 2.81, "learning_rate": 5.2120740791954395e-08, "loss": 0.5446, "step": 23600 }, { "epoch": 2.81, "learning_rate": 5.205547016746304e-08, "loss": 0.2957, "step": 23601 }, { "epoch": 2.81, "learning_rate": 5.1990240007700755e-08, "loss": 0.4051, "step": 23602 }, { "epoch": 2.81, "learning_rate": 5.192505031374473e-08, "loss": 0.399, "step": 23603 }, { "epoch": 2.81, "learning_rate": 5.185990108667299e-08, "loss": 0.487, "step": 23604 }, { "epoch": 2.81, "learning_rate": 5.1794792327562737e-08, "loss": 0.5997, "step": 23605 }, { "epoch": 2.81, "learning_rate": 5.172972403748977e-08, "loss": 0.6355, "step": 23606 }, { "epoch": 2.81, "learning_rate": 5.166469621753017e-08, "loss": 0.331, "step": 23607 }, { "epoch": 2.81, "learning_rate": 5.159970886875837e-08, "loss": 0.34, "step": 23608 }, { "epoch": 2.81, "learning_rate": 5.153476199224905e-08, "loss": 0.3839, "step": 23609 }, { "epoch": 2.81, "learning_rate": 5.1469855589075244e-08, "loss": 0.462, "step": 23610 }, { "epoch": 2.81, "learning_rate": 5.140498966031027e-08, "loss": 0.4522, "step": 23611 }, { "epoch": 2.81, "learning_rate": 5.134016420702603e-08, "loss": 0.6122, "step": 23612 }, { "epoch": 2.81, "learning_rate": 5.1275379230294745e-08, "loss": 0.404, "step": 23613 }, { "epoch": 2.81, "learning_rate": 5.121063473118693e-08, "loss": 0.3281, "step": 23614 }, { "epoch": 2.81, "learning_rate": 5.114593071077256e-08, "loss": 0.4156, "step": 23615 }, { "epoch": 2.81, "learning_rate": 5.1081267170121626e-08, "loss": 0.4504, "step": 23616 }, { "epoch": 2.81, "learning_rate": 5.1016644110302713e-08, "loss": 0.5914, "step": 23617 }, { "epoch": 2.81, "learning_rate": 5.095206153238413e-08, "loss": 0.6236, "step": 23618 }, { "epoch": 2.81, "learning_rate": 5.0887519437433365e-08, "loss": 0.5384, "step": 23619 }, { "epoch": 2.81, "learning_rate": 5.0823017826517604e-08, "loss": 0.3475, "step": 23620 }, { "epoch": 2.81, "learning_rate": 5.0758556700702674e-08, "loss": 0.4611, "step": 23621 }, { "epoch": 2.81, "learning_rate": 5.0694136061054665e-08, "loss": 0.4413, "step": 23622 }, { "epoch": 2.81, "learning_rate": 5.062975590863772e-08, "loss": 0.5492, "step": 23623 }, { "epoch": 2.81, "learning_rate": 5.056541624451683e-08, "loss": 0.5754, "step": 23624 }, { "epoch": 2.81, "learning_rate": 5.050111706975475e-08, "loss": 0.5539, "step": 23625 }, { "epoch": 2.81, "learning_rate": 5.0436858385414786e-08, "loss": 0.2755, "step": 23626 }, { "epoch": 2.81, "learning_rate": 5.0372640192558887e-08, "loss": 0.3914, "step": 23627 }, { "epoch": 2.81, "learning_rate": 5.0308462492249245e-08, "loss": 0.456, "step": 23628 }, { "epoch": 2.81, "learning_rate": 5.0244325285546126e-08, "loss": 0.6034, "step": 23629 }, { "epoch": 2.81, "learning_rate": 5.018022857350979e-08, "loss": 0.6282, "step": 23630 }, { "epoch": 2.81, "learning_rate": 5.0116172357199676e-08, "loss": 0.5932, "step": 23631 }, { "epoch": 2.81, "learning_rate": 5.005215663767521e-08, "loss": 0.3585, "step": 23632 }, { "epoch": 2.81, "learning_rate": 4.9988181415993874e-08, "loss": 0.3046, "step": 23633 }, { "epoch": 2.81, "learning_rate": 4.992424669321344e-08, "loss": 0.4251, "step": 23634 }, { "epoch": 2.81, "learning_rate": 4.986035247039084e-08, "loss": 0.4521, "step": 23635 }, { "epoch": 2.81, "learning_rate": 4.9796498748582445e-08, "loss": 0.61, "step": 23636 }, { "epoch": 2.81, "learning_rate": 4.973268552884325e-08, "loss": 0.5118, "step": 23637 }, { "epoch": 2.82, "learning_rate": 4.966891281222824e-08, "loss": 0.3046, "step": 23638 }, { "epoch": 2.82, "learning_rate": 4.9605180599792134e-08, "loss": 0.2712, "step": 23639 }, { "epoch": 2.82, "learning_rate": 4.95414888925877e-08, "loss": 0.3612, "step": 23640 }, { "epoch": 2.82, "learning_rate": 4.947783769166797e-08, "loss": 0.4607, "step": 23641 }, { "epoch": 2.82, "learning_rate": 4.941422699808518e-08, "loss": 0.5827, "step": 23642 }, { "epoch": 2.82, "learning_rate": 4.9350656812890974e-08, "loss": 0.6307, "step": 23643 }, { "epoch": 2.82, "learning_rate": 4.9287127137136183e-08, "loss": 0.617, "step": 23644 }, { "epoch": 2.82, "learning_rate": 4.92236379718708e-08, "loss": 0.3286, "step": 23645 }, { "epoch": 2.82, "learning_rate": 4.916018931814398e-08, "loss": 0.3807, "step": 23646 }, { "epoch": 2.82, "learning_rate": 4.909678117700517e-08, "loss": 0.4568, "step": 23647 }, { "epoch": 2.82, "learning_rate": 4.903341354950214e-08, "loss": 0.4557, "step": 23648 }, { "epoch": 2.82, "learning_rate": 4.8970086436682385e-08, "loss": 0.6048, "step": 23649 }, { "epoch": 2.82, "learning_rate": 4.890679983959284e-08, "loss": 0.6455, "step": 23650 }, { "epoch": 2.82, "learning_rate": 4.884355375927963e-08, "loss": 0.2841, "step": 23651 }, { "epoch": 2.82, "learning_rate": 4.878034819678801e-08, "loss": 0.3162, "step": 23652 }, { "epoch": 2.82, "learning_rate": 4.871718315316271e-08, "loss": 0.4038, "step": 23653 }, { "epoch": 2.82, "learning_rate": 4.865405862944844e-08, "loss": 0.3959, "step": 23654 }, { "epoch": 2.82, "learning_rate": 4.859097462668799e-08, "loss": 0.4811, "step": 23655 }, { "epoch": 2.82, "learning_rate": 4.852793114592441e-08, "loss": 0.5298, "step": 23656 }, { "epoch": 2.82, "learning_rate": 4.84649281882002e-08, "loss": 0.4684, "step": 23657 }, { "epoch": 2.82, "learning_rate": 4.840196575455619e-08, "loss": 0.3894, "step": 23658 }, { "epoch": 2.82, "learning_rate": 4.833904384603349e-08, "loss": 0.3652, "step": 23659 }, { "epoch": 2.82, "learning_rate": 4.827616246367184e-08, "loss": 0.4411, "step": 23660 }, { "epoch": 2.82, "learning_rate": 4.821332160851094e-08, "loss": 0.5864, "step": 23661 }, { "epoch": 2.82, "learning_rate": 4.815052128158998e-08, "loss": 0.6002, "step": 23662 }, { "epoch": 2.82, "learning_rate": 4.808776148394617e-08, "loss": 0.4027, "step": 23663 }, { "epoch": 2.82, "learning_rate": 4.802504221661758e-08, "loss": 0.3351, "step": 23664 }, { "epoch": 2.82, "learning_rate": 4.7962363480640875e-08, "loss": 0.3815, "step": 23665 }, { "epoch": 2.82, "learning_rate": 4.789972527705189e-08, "loss": 0.4505, "step": 23666 }, { "epoch": 2.82, "learning_rate": 4.7837127606886194e-08, "loss": 0.4929, "step": 23667 }, { "epoch": 2.82, "learning_rate": 4.7774570471178506e-08, "loss": 0.5686, "step": 23668 }, { "epoch": 2.82, "learning_rate": 4.771205387096328e-08, "loss": 0.4463, "step": 23669 }, { "epoch": 2.82, "learning_rate": 4.764957780727331e-08, "loss": 0.3816, "step": 23670 }, { "epoch": 2.82, "learning_rate": 4.7587142281141366e-08, "loss": 0.4989, "step": 23671 }, { "epoch": 2.82, "learning_rate": 4.752474729359996e-08, "loss": 0.438, "step": 23672 }, { "epoch": 2.82, "learning_rate": 4.746239284568049e-08, "loss": 0.4264, "step": 23673 }, { "epoch": 2.82, "learning_rate": 4.740007893841325e-08, "loss": 0.5632, "step": 23674 }, { "epoch": 2.82, "learning_rate": 4.733780557282824e-08, "loss": 0.5903, "step": 23675 }, { "epoch": 2.82, "learning_rate": 4.72755727499552e-08, "loss": 0.4142, "step": 23676 }, { "epoch": 2.82, "learning_rate": 4.721338047082302e-08, "loss": 0.3599, "step": 23677 }, { "epoch": 2.82, "learning_rate": 4.715122873645922e-08, "loss": 0.4847, "step": 23678 }, { "epoch": 2.82, "learning_rate": 4.708911754789158e-08, "loss": 0.5502, "step": 23679 }, { "epoch": 2.82, "learning_rate": 4.702704690614651e-08, "loss": 0.5518, "step": 23680 }, { "epoch": 2.82, "learning_rate": 4.696501681225041e-08, "loss": 0.5223, "step": 23681 }, { "epoch": 2.82, "learning_rate": 4.6903027267228286e-08, "loss": 0.3106, "step": 23682 }, { "epoch": 2.82, "learning_rate": 4.6841078272104604e-08, "loss": 0.342, "step": 23683 }, { "epoch": 2.82, "learning_rate": 4.677916982790437e-08, "loss": 0.3784, "step": 23684 }, { "epoch": 2.82, "learning_rate": 4.6717301935649826e-08, "loss": 0.4425, "step": 23685 }, { "epoch": 2.82, "learning_rate": 4.665547459636433e-08, "loss": 0.4916, "step": 23686 }, { "epoch": 2.82, "learning_rate": 4.6593687811069546e-08, "loss": 0.6527, "step": 23687 }, { "epoch": 2.82, "learning_rate": 4.6531941580787166e-08, "loss": 0.4503, "step": 23688 }, { "epoch": 2.82, "learning_rate": 4.6470235906537486e-08, "loss": 0.3766, "step": 23689 }, { "epoch": 2.82, "learning_rate": 4.6408570789340525e-08, "loss": 0.428, "step": 23690 }, { "epoch": 2.82, "learning_rate": 4.634694623021602e-08, "loss": 0.4346, "step": 23691 }, { "epoch": 2.82, "learning_rate": 4.628536223018232e-08, "loss": 0.651, "step": 23692 }, { "epoch": 2.82, "learning_rate": 4.62238187902575e-08, "loss": 0.6465, "step": 23693 }, { "epoch": 2.82, "learning_rate": 4.616231591145881e-08, "loss": 0.4309, "step": 23694 }, { "epoch": 2.82, "learning_rate": 4.6100853594802933e-08, "loss": 0.4301, "step": 23695 }, { "epoch": 2.82, "learning_rate": 4.6039431841305993e-08, "loss": 0.4036, "step": 23696 }, { "epoch": 2.82, "learning_rate": 4.597805065198274e-08, "loss": 0.4244, "step": 23697 }, { "epoch": 2.82, "learning_rate": 4.591671002784848e-08, "loss": 0.4977, "step": 23698 }, { "epoch": 2.82, "learning_rate": 4.585540996991683e-08, "loss": 0.6719, "step": 23699 }, { "epoch": 2.82, "learning_rate": 4.579415047920116e-08, "loss": 0.5121, "step": 23700 }, { "epoch": 2.82, "learning_rate": 4.5732931556714546e-08, "loss": 0.3067, "step": 23701 }, { "epoch": 2.82, "learning_rate": 4.567175320346812e-08, "loss": 0.4945, "step": 23702 }, { "epoch": 2.82, "learning_rate": 4.561061542047357e-08, "loss": 0.4508, "step": 23703 }, { "epoch": 2.82, "learning_rate": 4.554951820874148e-08, "loss": 0.3636, "step": 23704 }, { "epoch": 2.82, "learning_rate": 4.548846156928161e-08, "loss": 0.5848, "step": 23705 }, { "epoch": 2.82, "learning_rate": 4.542744550310368e-08, "loss": 0.548, "step": 23706 }, { "epoch": 2.82, "learning_rate": 4.536647001121608e-08, "loss": 0.3175, "step": 23707 }, { "epoch": 2.82, "learning_rate": 4.530553509462659e-08, "loss": 0.359, "step": 23708 }, { "epoch": 2.82, "learning_rate": 4.5244640754342486e-08, "loss": 0.3957, "step": 23709 }, { "epoch": 2.82, "learning_rate": 4.518378699137044e-08, "loss": 0.3689, "step": 23710 }, { "epoch": 2.82, "learning_rate": 4.5122973806716595e-08, "loss": 0.5668, "step": 23711 }, { "epoch": 2.82, "learning_rate": 4.5062201201385714e-08, "loss": 0.554, "step": 23712 }, { "epoch": 2.82, "learning_rate": 4.500146917638282e-08, "loss": 0.4924, "step": 23713 }, { "epoch": 2.82, "learning_rate": 4.494077773271155e-08, "loss": 0.3893, "step": 23714 }, { "epoch": 2.82, "learning_rate": 4.488012687137555e-08, "loss": 0.4078, "step": 23715 }, { "epoch": 2.82, "learning_rate": 4.481951659337652e-08, "loss": 0.3908, "step": 23716 }, { "epoch": 2.82, "learning_rate": 4.4758946899716994e-08, "loss": 0.5379, "step": 23717 }, { "epoch": 2.82, "learning_rate": 4.4698417791398386e-08, "loss": 0.6902, "step": 23718 }, { "epoch": 2.82, "learning_rate": 4.463792926942073e-08, "loss": 0.4847, "step": 23719 }, { "epoch": 2.82, "learning_rate": 4.457748133478407e-08, "loss": 0.3279, "step": 23720 }, { "epoch": 2.82, "learning_rate": 4.451707398848787e-08, "loss": 0.4699, "step": 23721 }, { "epoch": 2.83, "learning_rate": 4.4456707231530224e-08, "loss": 0.4187, "step": 23722 }, { "epoch": 2.83, "learning_rate": 4.439638106490951e-08, "loss": 0.4779, "step": 23723 }, { "epoch": 2.83, "learning_rate": 4.433609548962242e-08, "loss": 0.452, "step": 23724 }, { "epoch": 2.83, "learning_rate": 4.4275850506665943e-08, "loss": 0.5273, "step": 23725 }, { "epoch": 2.83, "learning_rate": 4.421564611703566e-08, "loss": 0.3144, "step": 23726 }, { "epoch": 2.83, "learning_rate": 4.4155482321726614e-08, "loss": 0.332, "step": 23727 }, { "epoch": 2.83, "learning_rate": 4.4095359121733574e-08, "loss": 0.3847, "step": 23728 }, { "epoch": 2.83, "learning_rate": 4.403527651805045e-08, "loss": 0.4413, "step": 23729 }, { "epoch": 2.83, "learning_rate": 4.397523451167035e-08, "loss": 0.5496, "step": 23730 }, { "epoch": 2.83, "learning_rate": 4.391523310358553e-08, "loss": 0.6006, "step": 23731 }, { "epoch": 2.83, "learning_rate": 4.385527229478798e-08, "loss": 0.3692, "step": 23732 }, { "epoch": 2.83, "learning_rate": 4.379535208626884e-08, "loss": 0.3479, "step": 23733 }, { "epoch": 2.83, "learning_rate": 4.373547247901872e-08, "loss": 0.5356, "step": 23734 }, { "epoch": 2.83, "learning_rate": 4.367563347402709e-08, "loss": 0.5089, "step": 23735 }, { "epoch": 2.83, "learning_rate": 4.361583507228373e-08, "loss": 0.5093, "step": 23736 }, { "epoch": 2.83, "learning_rate": 4.355607727477673e-08, "loss": 0.6541, "step": 23737 }, { "epoch": 2.83, "learning_rate": 4.349636008249364e-08, "loss": 0.4311, "step": 23738 }, { "epoch": 2.83, "learning_rate": 4.343668349642172e-08, "loss": 0.3161, "step": 23739 }, { "epoch": 2.83, "learning_rate": 4.3377047517547676e-08, "loss": 0.3995, "step": 23740 }, { "epoch": 2.83, "learning_rate": 4.3317452146857396e-08, "loss": 0.4818, "step": 23741 }, { "epoch": 2.83, "learning_rate": 4.3257897385335365e-08, "loss": 0.555, "step": 23742 }, { "epoch": 2.83, "learning_rate": 4.319838323396691e-08, "loss": 0.5315, "step": 23743 }, { "epoch": 2.83, "learning_rate": 4.3138909693734854e-08, "loss": 0.5946, "step": 23744 }, { "epoch": 2.83, "learning_rate": 4.307947676562313e-08, "loss": 0.4347, "step": 23745 }, { "epoch": 2.83, "learning_rate": 4.302008445061373e-08, "loss": 0.4106, "step": 23746 }, { "epoch": 2.83, "learning_rate": 4.2960732749688093e-08, "loss": 0.4871, "step": 23747 }, { "epoch": 2.83, "learning_rate": 4.290142166382821e-08, "loss": 0.5255, "step": 23748 }, { "epoch": 2.83, "learning_rate": 4.284215119401386e-08, "loss": 0.6756, "step": 23749 }, { "epoch": 2.83, "learning_rate": 4.2782921341224805e-08, "loss": 0.6259, "step": 23750 }, { "epoch": 2.83, "learning_rate": 4.2723732106440264e-08, "loss": 0.3724, "step": 23751 }, { "epoch": 2.83, "learning_rate": 4.26645834906389e-08, "loss": 0.3477, "step": 23752 }, { "epoch": 2.83, "learning_rate": 4.260547549479771e-08, "loss": 0.317, "step": 23753 }, { "epoch": 2.83, "learning_rate": 4.2546408119894245e-08, "loss": 0.4995, "step": 23754 }, { "epoch": 2.83, "learning_rate": 4.248738136690495e-08, "loss": 0.5058, "step": 23755 }, { "epoch": 2.83, "learning_rate": 4.24283952368057e-08, "loss": 0.6078, "step": 23756 }, { "epoch": 2.83, "learning_rate": 4.2369449730571e-08, "loss": 0.3619, "step": 23757 }, { "epoch": 2.83, "learning_rate": 4.231054484917563e-08, "loss": 0.3955, "step": 23758 }, { "epoch": 2.83, "learning_rate": 4.225168059359325e-08, "loss": 0.3838, "step": 23759 }, { "epoch": 2.83, "learning_rate": 4.219285696479669e-08, "loss": 0.4391, "step": 23760 }, { "epoch": 2.83, "learning_rate": 4.213407396375851e-08, "loss": 0.5924, "step": 23761 }, { "epoch": 2.83, "learning_rate": 4.2075331591450166e-08, "loss": 0.5526, "step": 23762 }, { "epoch": 2.83, "learning_rate": 4.201662984884281e-08, "loss": 0.4732, "step": 23763 }, { "epoch": 2.83, "learning_rate": 4.195796873690705e-08, "loss": 0.2941, "step": 23764 }, { "epoch": 2.83, "learning_rate": 4.189934825661213e-08, "loss": 0.3534, "step": 23765 }, { "epoch": 2.83, "learning_rate": 4.184076840892726e-08, "loss": 0.5143, "step": 23766 }, { "epoch": 2.83, "learning_rate": 4.1782229194820834e-08, "loss": 0.4893, "step": 23767 }, { "epoch": 2.83, "learning_rate": 4.172373061526014e-08, "loss": 0.588, "step": 23768 }, { "epoch": 2.83, "learning_rate": 4.1665272671212734e-08, "loss": 0.5669, "step": 23769 }, { "epoch": 2.83, "learning_rate": 4.1606855363644506e-08, "loss": 0.3578, "step": 23770 }, { "epoch": 2.83, "learning_rate": 4.154847869352136e-08, "loss": 0.4009, "step": 23771 }, { "epoch": 2.83, "learning_rate": 4.1490142661808066e-08, "loss": 0.4374, "step": 23772 }, { "epoch": 2.83, "learning_rate": 4.143184726946914e-08, "loss": 0.5765, "step": 23773 }, { "epoch": 2.83, "learning_rate": 4.13735925174677e-08, "loss": 0.4721, "step": 23774 }, { "epoch": 2.83, "learning_rate": 4.1315378406767414e-08, "loss": 0.6175, "step": 23775 }, { "epoch": 2.83, "learning_rate": 4.125720493833002e-08, "loss": 0.3437, "step": 23776 }, { "epoch": 2.83, "learning_rate": 4.1199072113117525e-08, "loss": 0.3576, "step": 23777 }, { "epoch": 2.83, "learning_rate": 4.114097993209054e-08, "loss": 0.4925, "step": 23778 }, { "epoch": 2.83, "learning_rate": 4.10829283962097e-08, "loss": 0.4496, "step": 23779 }, { "epoch": 2.83, "learning_rate": 4.1024917506434225e-08, "loss": 0.576, "step": 23780 }, { "epoch": 2.83, "learning_rate": 4.09669472637228e-08, "loss": 0.5638, "step": 23781 }, { "epoch": 2.83, "learning_rate": 4.090901766903466e-08, "loss": 0.3047, "step": 23782 }, { "epoch": 2.83, "learning_rate": 4.0851128723326526e-08, "loss": 0.4042, "step": 23783 }, { "epoch": 2.83, "learning_rate": 4.079328042755543e-08, "loss": 0.3835, "step": 23784 }, { "epoch": 2.83, "learning_rate": 4.073547278267781e-08, "loss": 0.4382, "step": 23785 }, { "epoch": 2.83, "learning_rate": 4.067770578964958e-08, "loss": 0.5154, "step": 23786 }, { "epoch": 2.83, "learning_rate": 4.0619979449424974e-08, "loss": 0.6622, "step": 23787 }, { "epoch": 2.83, "learning_rate": 4.0562293762958224e-08, "loss": 0.3638, "step": 23788 }, { "epoch": 2.83, "learning_rate": 4.05046487312033e-08, "loss": 0.2944, "step": 23789 }, { "epoch": 2.83, "learning_rate": 4.044704435511304e-08, "loss": 0.3961, "step": 23790 }, { "epoch": 2.83, "learning_rate": 4.0389480635639463e-08, "loss": 0.5028, "step": 23791 }, { "epoch": 2.83, "learning_rate": 4.033195757373376e-08, "loss": 0.5996, "step": 23792 }, { "epoch": 2.83, "learning_rate": 4.027447517034766e-08, "loss": 0.5723, "step": 23793 }, { "epoch": 2.83, "learning_rate": 4.0217033426430964e-08, "loss": 0.5457, "step": 23794 }, { "epoch": 2.83, "learning_rate": 4.015963234293263e-08, "loss": 0.2863, "step": 23795 }, { "epoch": 2.83, "learning_rate": 4.0102271920802186e-08, "loss": 0.3295, "step": 23796 }, { "epoch": 2.83, "learning_rate": 4.004495216098775e-08, "loss": 0.4198, "step": 23797 }, { "epoch": 2.83, "learning_rate": 3.998767306443635e-08, "loss": 0.4498, "step": 23798 }, { "epoch": 2.83, "learning_rate": 3.993043463209528e-08, "loss": 0.676, "step": 23799 }, { "epoch": 2.83, "learning_rate": 3.9873236864910734e-08, "loss": 0.5401, "step": 23800 }, { "epoch": 2.83, "learning_rate": 3.981607976382779e-08, "loss": 0.3741, "step": 23801 }, { "epoch": 2.83, "learning_rate": 3.9758963329791514e-08, "loss": 0.3459, "step": 23802 }, { "epoch": 2.83, "learning_rate": 3.9701887563745876e-08, "loss": 0.4291, "step": 23803 }, { "epoch": 2.83, "learning_rate": 3.9644852466634574e-08, "loss": 0.4795, "step": 23804 }, { "epoch": 2.83, "learning_rate": 3.9587858039400463e-08, "loss": 0.5153, "step": 23805 }, { "epoch": 2.84, "learning_rate": 3.9530904282985285e-08, "loss": 0.5428, "step": 23806 }, { "epoch": 2.84, "learning_rate": 3.9473991198330786e-08, "loss": 0.281, "step": 23807 }, { "epoch": 2.84, "learning_rate": 3.9417118786377606e-08, "loss": 0.3514, "step": 23808 }, { "epoch": 2.84, "learning_rate": 3.93602870480661e-08, "loss": 0.4467, "step": 23809 }, { "epoch": 2.84, "learning_rate": 3.930349598433525e-08, "loss": 0.4216, "step": 23810 }, { "epoch": 2.84, "learning_rate": 3.9246745596124294e-08, "loss": 0.4566, "step": 23811 }, { "epoch": 2.84, "learning_rate": 3.919003588437109e-08, "loss": 0.6213, "step": 23812 }, { "epoch": 2.84, "learning_rate": 3.9133366850012956e-08, "loss": 0.4165, "step": 23813 }, { "epoch": 2.84, "learning_rate": 3.907673849398663e-08, "loss": 0.3359, "step": 23814 }, { "epoch": 2.84, "learning_rate": 3.90201508172286e-08, "loss": 0.4357, "step": 23815 }, { "epoch": 2.84, "learning_rate": 3.896360382067394e-08, "loss": 0.4103, "step": 23816 }, { "epoch": 2.84, "learning_rate": 3.890709750525717e-08, "loss": 0.5482, "step": 23817 }, { "epoch": 2.84, "learning_rate": 3.8850631871912846e-08, "loss": 0.6356, "step": 23818 }, { "epoch": 2.84, "learning_rate": 3.879420692157382e-08, "loss": 0.52, "step": 23819 }, { "epoch": 2.84, "learning_rate": 3.873782265517323e-08, "loss": 0.331, "step": 23820 }, { "epoch": 2.84, "learning_rate": 3.8681479073642836e-08, "loss": 0.3706, "step": 23821 }, { "epoch": 2.84, "learning_rate": 3.8625176177914393e-08, "loss": 0.4811, "step": 23822 }, { "epoch": 2.84, "learning_rate": 3.8568913968917986e-08, "loss": 0.4807, "step": 23823 }, { "epoch": 2.84, "learning_rate": 3.851269244758426e-08, "loss": 0.6242, "step": 23824 }, { "epoch": 2.84, "learning_rate": 3.845651161484221e-08, "loss": 0.4162, "step": 23825 }, { "epoch": 2.84, "learning_rate": 3.840037147162024e-08, "loss": 0.2325, "step": 23826 }, { "epoch": 2.84, "learning_rate": 3.834427201884705e-08, "loss": 0.347, "step": 23827 }, { "epoch": 2.84, "learning_rate": 3.8288213257449414e-08, "loss": 0.4093, "step": 23828 }, { "epoch": 2.84, "learning_rate": 3.823219518835436e-08, "loss": 0.531, "step": 23829 }, { "epoch": 2.84, "learning_rate": 3.817621781248726e-08, "loss": 0.5523, "step": 23830 }, { "epoch": 2.84, "learning_rate": 3.812028113077432e-08, "loss": 0.5866, "step": 23831 }, { "epoch": 2.84, "learning_rate": 3.8064385144139236e-08, "loss": 0.3397, "step": 23832 }, { "epoch": 2.84, "learning_rate": 3.800852985350684e-08, "loss": 0.2925, "step": 23833 }, { "epoch": 2.84, "learning_rate": 3.795271525979971e-08, "loss": 0.3436, "step": 23834 }, { "epoch": 2.84, "learning_rate": 3.789694136394101e-08, "loss": 0.4294, "step": 23835 }, { "epoch": 2.84, "learning_rate": 3.7841208166852495e-08, "loss": 0.529, "step": 23836 }, { "epoch": 2.84, "learning_rate": 3.7785515669455096e-08, "loss": 0.5854, "step": 23837 }, { "epoch": 2.84, "learning_rate": 3.772986387267003e-08, "loss": 0.3832, "step": 23838 }, { "epoch": 2.84, "learning_rate": 3.767425277741682e-08, "loss": 0.2525, "step": 23839 }, { "epoch": 2.84, "learning_rate": 3.7618682384614744e-08, "loss": 0.4062, "step": 23840 }, { "epoch": 2.84, "learning_rate": 3.756315269518251e-08, "loss": 0.4041, "step": 23841 }, { "epoch": 2.84, "learning_rate": 3.7507663710038e-08, "loss": 0.4865, "step": 23842 }, { "epoch": 2.84, "learning_rate": 3.7452215430098246e-08, "loss": 0.5578, "step": 23843 }, { "epoch": 2.84, "learning_rate": 3.7396807856280306e-08, "loss": 0.5743, "step": 23844 }, { "epoch": 2.84, "learning_rate": 3.7341440989499546e-08, "loss": 0.2566, "step": 23845 }, { "epoch": 2.84, "learning_rate": 3.728611483067163e-08, "loss": 0.4717, "step": 23846 }, { "epoch": 2.84, "learning_rate": 3.7230829380710545e-08, "loss": 0.38, "step": 23847 }, { "epoch": 2.84, "learning_rate": 3.717558464053056e-08, "loss": 0.4725, "step": 23848 }, { "epoch": 2.84, "learning_rate": 3.7120380611044836e-08, "loss": 0.5567, "step": 23849 }, { "epoch": 2.84, "learning_rate": 3.706521729316626e-08, "loss": 0.6187, "step": 23850 }, { "epoch": 2.84, "learning_rate": 3.701009468780603e-08, "loss": 0.2844, "step": 23851 }, { "epoch": 2.84, "learning_rate": 3.695501279587538e-08, "loss": 0.3387, "step": 23852 }, { "epoch": 2.84, "learning_rate": 3.6899971618285235e-08, "loss": 0.5033, "step": 23853 }, { "epoch": 2.84, "learning_rate": 3.6844971155945433e-08, "loss": 0.4294, "step": 23854 }, { "epoch": 2.84, "learning_rate": 3.6790011409764405e-08, "loss": 0.4995, "step": 23855 }, { "epoch": 2.84, "learning_rate": 3.6735092380651426e-08, "loss": 0.5959, "step": 23856 }, { "epoch": 2.84, "learning_rate": 3.668021406951438e-08, "loss": 0.327, "step": 23857 }, { "epoch": 2.84, "learning_rate": 3.662537647725978e-08, "loss": 0.2844, "step": 23858 }, { "epoch": 2.84, "learning_rate": 3.6570579604794374e-08, "loss": 0.4063, "step": 23859 }, { "epoch": 2.84, "learning_rate": 3.6515823453023846e-08, "loss": 0.5498, "step": 23860 }, { "epoch": 2.84, "learning_rate": 3.6461108022853855e-08, "loss": 0.5794, "step": 23861 }, { "epoch": 2.84, "learning_rate": 3.640643331518812e-08, "loss": 0.7101, "step": 23862 }, { "epoch": 2.84, "learning_rate": 3.6351799330930914e-08, "loss": 0.4409, "step": 23863 }, { "epoch": 2.84, "learning_rate": 3.6297206070985136e-08, "loss": 0.2957, "step": 23864 }, { "epoch": 2.84, "learning_rate": 3.624265353625339e-08, "loss": 0.3834, "step": 23865 }, { "epoch": 2.84, "learning_rate": 3.618814172763718e-08, "loss": 0.5411, "step": 23866 }, { "epoch": 2.84, "learning_rate": 3.613367064603773e-08, "loss": 0.5012, "step": 23867 }, { "epoch": 2.84, "learning_rate": 3.607924029235543e-08, "loss": 0.6277, "step": 23868 }, { "epoch": 2.84, "learning_rate": 3.602485066749012e-08, "loss": 0.4408, "step": 23869 }, { "epoch": 2.84, "learning_rate": 3.597050177234052e-08, "loss": 0.3246, "step": 23870 }, { "epoch": 2.84, "learning_rate": 3.591619360780535e-08, "loss": 0.4292, "step": 23871 }, { "epoch": 2.84, "learning_rate": 3.586192617478251e-08, "loss": 0.4903, "step": 23872 }, { "epoch": 2.84, "learning_rate": 3.580769947416879e-08, "loss": 0.5851, "step": 23873 }, { "epoch": 2.84, "learning_rate": 3.5753513506860404e-08, "loss": 0.6318, "step": 23874 }, { "epoch": 2.84, "learning_rate": 3.5699368273753034e-08, "loss": 0.6076, "step": 23875 }, { "epoch": 2.84, "learning_rate": 3.564526377574207e-08, "loss": 0.2328, "step": 23876 }, { "epoch": 2.84, "learning_rate": 3.5591200013721797e-08, "loss": 0.4445, "step": 23877 }, { "epoch": 2.84, "learning_rate": 3.5537176988585387e-08, "loss": 0.4404, "step": 23878 }, { "epoch": 2.84, "learning_rate": 3.548319470122658e-08, "loss": 0.5179, "step": 23879 }, { "epoch": 2.84, "learning_rate": 3.542925315253742e-08, "loss": 0.6103, "step": 23880 }, { "epoch": 2.84, "learning_rate": 3.5375352343409164e-08, "loss": 0.5619, "step": 23881 }, { "epoch": 2.84, "learning_rate": 3.5321492274733304e-08, "loss": 0.4243, "step": 23882 }, { "epoch": 2.84, "learning_rate": 3.526767294739997e-08, "loss": 0.3721, "step": 23883 }, { "epoch": 2.84, "learning_rate": 3.5213894362298715e-08, "loss": 0.4874, "step": 23884 }, { "epoch": 2.84, "learning_rate": 3.516015652031884e-08, "loss": 0.4525, "step": 23885 }, { "epoch": 2.84, "learning_rate": 3.510645942234825e-08, "loss": 0.5453, "step": 23886 }, { "epoch": 2.84, "learning_rate": 3.505280306927455e-08, "loss": 0.5463, "step": 23887 }, { "epoch": 2.84, "learning_rate": 3.499918746198511e-08, "loss": 0.373, "step": 23888 }, { "epoch": 2.84, "learning_rate": 3.494561260136587e-08, "loss": 0.2886, "step": 23889 }, { "epoch": 2.85, "learning_rate": 3.489207848830223e-08, "loss": 0.3959, "step": 23890 }, { "epoch": 2.85, "learning_rate": 3.483858512367988e-08, "loss": 0.4735, "step": 23891 }, { "epoch": 2.85, "learning_rate": 3.4785132508382e-08, "loss": 0.5362, "step": 23892 }, { "epoch": 2.85, "learning_rate": 3.473172064329317e-08, "loss": 0.5874, "step": 23893 }, { "epoch": 2.85, "learning_rate": 3.467834952929544e-08, "loss": 0.5376, "step": 23894 }, { "epoch": 2.85, "learning_rate": 3.462501916727201e-08, "loss": 0.3384, "step": 23895 }, { "epoch": 2.85, "learning_rate": 3.457172955810328e-08, "loss": 0.3734, "step": 23896 }, { "epoch": 2.85, "learning_rate": 3.451848070267105e-08, "loss": 0.4661, "step": 23897 }, { "epoch": 2.85, "learning_rate": 3.446527260185517e-08, "loss": 0.5237, "step": 23898 }, { "epoch": 2.85, "learning_rate": 3.441210525653521e-08, "loss": 0.6704, "step": 23899 }, { "epoch": 2.85, "learning_rate": 3.4358978667589914e-08, "loss": 0.5579, "step": 23900 }, { "epoch": 2.85, "learning_rate": 3.430589283589775e-08, "loss": 0.301, "step": 23901 }, { "epoch": 2.85, "learning_rate": 3.4252847762335784e-08, "loss": 0.3463, "step": 23902 }, { "epoch": 2.85, "learning_rate": 3.419984344778165e-08, "loss": 0.3615, "step": 23903 }, { "epoch": 2.85, "learning_rate": 3.414687989311049e-08, "loss": 0.4282, "step": 23904 }, { "epoch": 2.85, "learning_rate": 3.4093957099198547e-08, "loss": 0.5633, "step": 23905 }, { "epoch": 2.85, "learning_rate": 3.4041075066920395e-08, "loss": 0.646, "step": 23906 }, { "epoch": 2.85, "learning_rate": 3.398823379715005e-08, "loss": 0.2724, "step": 23907 }, { "epoch": 2.85, "learning_rate": 3.3935433290761266e-08, "loss": 0.3322, "step": 23908 }, { "epoch": 2.85, "learning_rate": 3.388267354862668e-08, "loss": 0.3713, "step": 23909 }, { "epoch": 2.85, "learning_rate": 3.382995457161864e-08, "loss": 0.4267, "step": 23910 }, { "epoch": 2.85, "learning_rate": 3.377727636060785e-08, "loss": 0.5049, "step": 23911 }, { "epoch": 2.85, "learning_rate": 3.372463891646582e-08, "loss": 0.7301, "step": 23912 }, { "epoch": 2.85, "learning_rate": 3.367204224006243e-08, "loss": 0.4565, "step": 23913 }, { "epoch": 2.85, "learning_rate": 3.361948633226753e-08, "loss": 0.37, "step": 23914 }, { "epoch": 2.85, "learning_rate": 3.3566971193949024e-08, "loss": 0.4639, "step": 23915 }, { "epoch": 2.85, "learning_rate": 3.3514496825975686e-08, "loss": 0.4214, "step": 23916 }, { "epoch": 2.85, "learning_rate": 3.3462063229214304e-08, "loss": 0.5456, "step": 23917 }, { "epoch": 2.85, "learning_rate": 3.3409670404532526e-08, "loss": 0.6329, "step": 23918 }, { "epoch": 2.85, "learning_rate": 3.335731835279549e-08, "loss": 0.477, "step": 23919 }, { "epoch": 2.85, "learning_rate": 3.330500707486889e-08, "loss": 0.3695, "step": 23920 }, { "epoch": 2.85, "learning_rate": 3.325273657161759e-08, "loss": 0.3961, "step": 23921 }, { "epoch": 2.85, "learning_rate": 3.320050684390563e-08, "loss": 0.4868, "step": 23922 }, { "epoch": 2.85, "learning_rate": 3.3148317892596205e-08, "loss": 0.5266, "step": 23923 }, { "epoch": 2.85, "learning_rate": 3.309616971855195e-08, "loss": 0.5263, "step": 23924 }, { "epoch": 2.85, "learning_rate": 3.304406232263524e-08, "loss": 0.5188, "step": 23925 }, { "epoch": 2.85, "learning_rate": 3.2991995705707045e-08, "loss": 0.4913, "step": 23926 }, { "epoch": 2.85, "learning_rate": 3.2939969868628075e-08, "loss": 0.3807, "step": 23927 }, { "epoch": 2.85, "learning_rate": 3.2887984812258464e-08, "loss": 0.4127, "step": 23928 }, { "epoch": 2.85, "learning_rate": 3.2836040537457536e-08, "loss": 0.4147, "step": 23929 }, { "epoch": 2.85, "learning_rate": 3.278413704508404e-08, "loss": 0.4877, "step": 23930 }, { "epoch": 2.85, "learning_rate": 3.273227433599563e-08, "loss": 0.5994, "step": 23931 }, { "epoch": 2.85, "learning_rate": 3.268045241104939e-08, "loss": 0.2452, "step": 23932 }, { "epoch": 2.85, "learning_rate": 3.2628671271102693e-08, "loss": 0.3587, "step": 23933 }, { "epoch": 2.85, "learning_rate": 3.2576930917010977e-08, "loss": 0.4254, "step": 23934 }, { "epoch": 2.85, "learning_rate": 3.252523134962965e-08, "loss": 0.4247, "step": 23935 }, { "epoch": 2.85, "learning_rate": 3.247357256981332e-08, "loss": 0.5251, "step": 23936 }, { "epoch": 2.85, "learning_rate": 3.242195457841601e-08, "loss": 0.5385, "step": 23937 }, { "epoch": 2.85, "learning_rate": 3.237037737629068e-08, "loss": 0.4375, "step": 23938 }, { "epoch": 2.85, "learning_rate": 3.231884096428994e-08, "loss": 0.333, "step": 23939 }, { "epoch": 2.85, "learning_rate": 3.226734534326592e-08, "loss": 0.5017, "step": 23940 }, { "epoch": 2.85, "learning_rate": 3.221589051406987e-08, "loss": 0.4309, "step": 23941 }, { "epoch": 2.85, "learning_rate": 3.2164476477551943e-08, "loss": 0.5558, "step": 23942 }, { "epoch": 2.85, "learning_rate": 3.211310323456257e-08, "loss": 0.6662, "step": 23943 }, { "epoch": 2.85, "learning_rate": 3.206177078595052e-08, "loss": 0.5566, "step": 23944 }, { "epoch": 2.85, "learning_rate": 3.201047913256455e-08, "loss": 0.3153, "step": 23945 }, { "epoch": 2.85, "learning_rate": 3.195922827525233e-08, "loss": 0.4184, "step": 23946 }, { "epoch": 2.85, "learning_rate": 3.190801821486095e-08, "loss": 0.448, "step": 23947 }, { "epoch": 2.85, "learning_rate": 3.1856848952237506e-08, "loss": 0.492, "step": 23948 }, { "epoch": 2.85, "learning_rate": 3.1805720488227445e-08, "loss": 0.6052, "step": 23949 }, { "epoch": 2.85, "learning_rate": 3.175463282367591e-08, "loss": 0.6254, "step": 23950 }, { "epoch": 2.85, "learning_rate": 3.170358595942724e-08, "loss": 0.3489, "step": 23951 }, { "epoch": 2.85, "learning_rate": 3.1652579896325474e-08, "loss": 0.4279, "step": 23952 }, { "epoch": 2.85, "learning_rate": 3.160161463521355e-08, "loss": 0.5441, "step": 23953 }, { "epoch": 2.85, "learning_rate": 3.155069017693413e-08, "loss": 0.5518, "step": 23954 }, { "epoch": 2.85, "learning_rate": 3.1499806522329036e-08, "loss": 0.6651, "step": 23955 }, { "epoch": 2.85, "learning_rate": 3.144896367223899e-08, "loss": 0.5504, "step": 23956 }, { "epoch": 2.85, "learning_rate": 3.1398161627504695e-08, "loss": 0.3005, "step": 23957 }, { "epoch": 2.85, "learning_rate": 3.134740038896633e-08, "loss": 0.3792, "step": 23958 }, { "epoch": 2.85, "learning_rate": 3.1296679957462385e-08, "loss": 0.3545, "step": 23959 }, { "epoch": 2.85, "learning_rate": 3.124600033383135e-08, "loss": 0.4476, "step": 23960 }, { "epoch": 2.85, "learning_rate": 3.119536151891117e-08, "loss": 0.4429, "step": 23961 }, { "epoch": 2.85, "learning_rate": 3.114476351353868e-08, "loss": 0.7225, "step": 23962 }, { "epoch": 2.85, "learning_rate": 3.10942063185507e-08, "loss": 0.2808, "step": 23963 }, { "epoch": 2.85, "learning_rate": 3.1043689934782404e-08, "loss": 0.3611, "step": 23964 }, { "epoch": 2.85, "learning_rate": 3.099321436306924e-08, "loss": 0.4313, "step": 23965 }, { "epoch": 2.85, "learning_rate": 3.094277960424524e-08, "loss": 0.5228, "step": 23966 }, { "epoch": 2.85, "learning_rate": 3.0892385659144486e-08, "loss": 0.5098, "step": 23967 }, { "epoch": 2.85, "learning_rate": 3.084203252859935e-08, "loss": 0.6133, "step": 23968 }, { "epoch": 2.85, "learning_rate": 3.0791720213442786e-08, "loss": 0.5598, "step": 23969 }, { "epoch": 2.85, "learning_rate": 3.074144871450663e-08, "loss": 0.2843, "step": 23970 }, { "epoch": 2.85, "learning_rate": 3.069121803262104e-08, "loss": 0.4145, "step": 23971 }, { "epoch": 2.85, "learning_rate": 3.064102816861703e-08, "loss": 0.421, "step": 23972 }, { "epoch": 2.85, "learning_rate": 3.0590879123323924e-08, "loss": 0.5402, "step": 23973 }, { "epoch": 2.86, "learning_rate": 3.054077089757107e-08, "loss": 0.5415, "step": 23974 }, { "epoch": 2.86, "learning_rate": 3.0490703492186136e-08, "loss": 0.537, "step": 23975 }, { "epoch": 2.86, "learning_rate": 3.044067690799707e-08, "loss": 0.3947, "step": 23976 }, { "epoch": 2.86, "learning_rate": 3.0390691145830984e-08, "loss": 0.3272, "step": 23977 }, { "epoch": 2.86, "learning_rate": 3.034074620651417e-08, "loss": 0.4219, "step": 23978 }, { "epoch": 2.86, "learning_rate": 3.02908420908718e-08, "loss": 0.4823, "step": 23979 }, { "epoch": 2.86, "learning_rate": 3.0240978799729035e-08, "loss": 0.6245, "step": 23980 }, { "epoch": 2.86, "learning_rate": 3.0191156333910233e-08, "loss": 0.5551, "step": 23981 }, { "epoch": 2.86, "learning_rate": 3.014137469423889e-08, "loss": 0.3904, "step": 23982 }, { "epoch": 2.86, "learning_rate": 3.009163388153768e-08, "loss": 0.3918, "step": 23983 }, { "epoch": 2.86, "learning_rate": 3.0041933896629016e-08, "loss": 0.4338, "step": 23984 }, { "epoch": 2.86, "learning_rate": 2.9992274740334446e-08, "loss": 0.4087, "step": 23985 }, { "epoch": 2.86, "learning_rate": 2.994265641347499e-08, "loss": 0.5776, "step": 23986 }, { "epoch": 2.86, "learning_rate": 2.989307891687054e-08, "loss": 0.582, "step": 23987 }, { "epoch": 2.86, "learning_rate": 2.984354225134073e-08, "loss": 0.2376, "step": 23988 }, { "epoch": 2.86, "learning_rate": 2.9794046417704613e-08, "loss": 0.3458, "step": 23989 }, { "epoch": 2.86, "learning_rate": 2.9744591416779876e-08, "loss": 0.4696, "step": 23990 }, { "epoch": 2.86, "learning_rate": 2.9695177249384478e-08, "loss": 0.4981, "step": 23991 }, { "epoch": 2.86, "learning_rate": 2.9645803916334982e-08, "loss": 0.5201, "step": 23992 }, { "epoch": 2.86, "learning_rate": 2.959647141844768e-08, "loss": 0.5617, "step": 23993 }, { "epoch": 2.86, "learning_rate": 2.9547179756538314e-08, "loss": 0.4906, "step": 23994 }, { "epoch": 2.86, "learning_rate": 2.9497928931420948e-08, "loss": 0.2987, "step": 23995 }, { "epoch": 2.86, "learning_rate": 2.9448718943910214e-08, "loss": 0.4167, "step": 23996 }, { "epoch": 2.86, "learning_rate": 2.9399549794819627e-08, "loss": 0.4336, "step": 23997 }, { "epoch": 2.86, "learning_rate": 2.935042148496159e-08, "loss": 0.4847, "step": 23998 }, { "epoch": 2.86, "learning_rate": 2.930133401514823e-08, "loss": 0.6807, "step": 23999 }, { "epoch": 2.86, "learning_rate": 2.9252287386191403e-08, "loss": 0.6218, "step": 24000 }, { "epoch": 2.86, "learning_rate": 2.9203281598901568e-08, "loss": 0.3205, "step": 24001 }, { "epoch": 2.86, "learning_rate": 2.915431665408863e-08, "loss": 0.3741, "step": 24002 }, { "epoch": 2.86, "learning_rate": 2.910539255256223e-08, "loss": 0.4988, "step": 24003 }, { "epoch": 2.86, "learning_rate": 2.9056509295131153e-08, "loss": 0.4742, "step": 24004 }, { "epoch": 2.86, "learning_rate": 2.900766688260309e-08, "loss": 0.5606, "step": 24005 }, { "epoch": 2.86, "learning_rate": 2.8958865315785733e-08, "loss": 0.567, "step": 24006 }, { "epoch": 2.86, "learning_rate": 2.8910104595485932e-08, "loss": 0.3506, "step": 24007 }, { "epoch": 2.86, "learning_rate": 2.886138472250888e-08, "loss": 0.3913, "step": 24008 }, { "epoch": 2.86, "learning_rate": 2.881270569766115e-08, "loss": 0.4474, "step": 24009 }, { "epoch": 2.86, "learning_rate": 2.8764067521746263e-08, "loss": 0.408, "step": 24010 }, { "epoch": 2.86, "learning_rate": 2.8715470195568585e-08, "loss": 0.5806, "step": 24011 }, { "epoch": 2.86, "learning_rate": 2.866691371993191e-08, "loss": 0.6496, "step": 24012 }, { "epoch": 2.86, "learning_rate": 2.8618398095638377e-08, "loss": 0.2488, "step": 24013 }, { "epoch": 2.86, "learning_rate": 2.8569923323490124e-08, "loss": 0.2731, "step": 24014 }, { "epoch": 2.86, "learning_rate": 2.852148940428845e-08, "loss": 0.4198, "step": 24015 }, { "epoch": 2.86, "learning_rate": 2.847309633883383e-08, "loss": 0.4545, "step": 24016 }, { "epoch": 2.86, "learning_rate": 2.8424744127926174e-08, "loss": 0.5259, "step": 24017 }, { "epoch": 2.86, "learning_rate": 2.8376432772365124e-08, "loss": 0.6097, "step": 24018 }, { "epoch": 2.86, "learning_rate": 2.8328162272948935e-08, "loss": 0.4356, "step": 24019 }, { "epoch": 2.86, "learning_rate": 2.827993263047557e-08, "loss": 0.2931, "step": 24020 }, { "epoch": 2.86, "learning_rate": 2.8231743845742455e-08, "loss": 0.3567, "step": 24021 }, { "epoch": 2.86, "learning_rate": 2.8183595919545892e-08, "loss": 0.4748, "step": 24022 }, { "epoch": 2.86, "learning_rate": 2.813548885268219e-08, "loss": 0.5111, "step": 24023 }, { "epoch": 2.86, "learning_rate": 2.8087422645945715e-08, "loss": 0.6354, "step": 24024 }, { "epoch": 2.86, "learning_rate": 2.8039397300131942e-08, "loss": 0.5486, "step": 24025 }, { "epoch": 2.86, "learning_rate": 2.7991412816034123e-08, "loss": 0.2678, "step": 24026 }, { "epoch": 2.86, "learning_rate": 2.7943469194445795e-08, "loss": 0.4387, "step": 24027 }, { "epoch": 2.86, "learning_rate": 2.789556643615937e-08, "loss": 0.4291, "step": 24028 }, { "epoch": 2.86, "learning_rate": 2.7847704541966726e-08, "loss": 0.5131, "step": 24029 }, { "epoch": 2.86, "learning_rate": 2.779988351265861e-08, "loss": 0.5059, "step": 24030 }, { "epoch": 2.86, "learning_rate": 2.7752103349026337e-08, "loss": 0.6695, "step": 24031 }, { "epoch": 2.86, "learning_rate": 2.7704364051859e-08, "loss": 0.2898, "step": 24032 }, { "epoch": 2.86, "learning_rate": 2.7656665621945687e-08, "loss": 0.297, "step": 24033 }, { "epoch": 2.86, "learning_rate": 2.7609008060075493e-08, "loss": 0.3259, "step": 24034 }, { "epoch": 2.86, "learning_rate": 2.756139136703584e-08, "loss": 0.4933, "step": 24035 }, { "epoch": 2.86, "learning_rate": 2.7513815543613875e-08, "loss": 0.4473, "step": 24036 }, { "epoch": 2.86, "learning_rate": 2.746628059059592e-08, "loss": 0.541, "step": 24037 }, { "epoch": 2.86, "learning_rate": 2.7418786508767726e-08, "loss": 0.3595, "step": 24038 }, { "epoch": 2.86, "learning_rate": 2.7371333298914504e-08, "loss": 0.3951, "step": 24039 }, { "epoch": 2.86, "learning_rate": 2.7323920961820627e-08, "loss": 0.3383, "step": 24040 }, { "epoch": 2.86, "learning_rate": 2.7276549498269634e-08, "loss": 0.5037, "step": 24041 }, { "epoch": 2.86, "learning_rate": 2.722921890904506e-08, "loss": 0.4713, "step": 24042 }, { "epoch": 2.86, "learning_rate": 2.7181929194928793e-08, "loss": 0.5599, "step": 24043 }, { "epoch": 2.86, "learning_rate": 2.713468035670297e-08, "loss": 0.5452, "step": 24044 }, { "epoch": 2.86, "learning_rate": 2.7087472395148084e-08, "loss": 0.3593, "step": 24045 }, { "epoch": 2.86, "learning_rate": 2.70403053110449e-08, "loss": 0.4059, "step": 24046 }, { "epoch": 2.86, "learning_rate": 2.6993179105172795e-08, "loss": 0.4514, "step": 24047 }, { "epoch": 2.86, "learning_rate": 2.6946093778311144e-08, "loss": 0.5818, "step": 24048 }, { "epoch": 2.86, "learning_rate": 2.6899049331237936e-08, "loss": 0.5803, "step": 24049 }, { "epoch": 2.86, "learning_rate": 2.6852045764731162e-08, "loss": 0.5979, "step": 24050 }, { "epoch": 2.86, "learning_rate": 2.6805083079567417e-08, "loss": 0.3001, "step": 24051 }, { "epoch": 2.86, "learning_rate": 2.675816127652303e-08, "loss": 0.4217, "step": 24052 }, { "epoch": 2.86, "learning_rate": 2.6711280356374048e-08, "loss": 0.455, "step": 24053 }, { "epoch": 2.86, "learning_rate": 2.6664440319895123e-08, "loss": 0.5316, "step": 24054 }, { "epoch": 2.86, "learning_rate": 2.6617641167860362e-08, "loss": 0.6277, "step": 24055 }, { "epoch": 2.86, "learning_rate": 2.6570882901043317e-08, "loss": 0.6403, "step": 24056 }, { "epoch": 2.86, "learning_rate": 2.6524165520217526e-08, "loss": 0.2586, "step": 24057 }, { "epoch": 2.87, "learning_rate": 2.647748902615488e-08, "loss": 0.3873, "step": 24058 }, { "epoch": 2.87, "learning_rate": 2.6430853419626425e-08, "loss": 0.4528, "step": 24059 }, { "epoch": 2.87, "learning_rate": 2.6384258701403763e-08, "loss": 0.5167, "step": 24060 }, { "epoch": 2.87, "learning_rate": 2.633770487225712e-08, "loss": 0.4809, "step": 24061 }, { "epoch": 2.87, "learning_rate": 2.629119193295532e-08, "loss": 0.487, "step": 24062 }, { "epoch": 2.87, "learning_rate": 2.6244719884267746e-08, "loss": 0.2069, "step": 24063 }, { "epoch": 2.87, "learning_rate": 2.619828872696295e-08, "loss": 0.3749, "step": 24064 }, { "epoch": 2.87, "learning_rate": 2.6151898461807546e-08, "loss": 0.4371, "step": 24065 }, { "epoch": 2.87, "learning_rate": 2.610554908956897e-08, "loss": 0.4175, "step": 24066 }, { "epoch": 2.87, "learning_rate": 2.605924061101328e-08, "loss": 0.5179, "step": 24067 }, { "epoch": 2.87, "learning_rate": 2.6012973026905976e-08, "loss": 0.5277, "step": 24068 }, { "epoch": 2.87, "learning_rate": 2.5966746338011716e-08, "loss": 0.5835, "step": 24069 }, { "epoch": 2.87, "learning_rate": 2.592056054509462e-08, "loss": 0.2687, "step": 24070 }, { "epoch": 2.87, "learning_rate": 2.5874415648918516e-08, "loss": 0.3302, "step": 24071 }, { "epoch": 2.87, "learning_rate": 2.5828311650245854e-08, "loss": 0.4656, "step": 24072 }, { "epoch": 2.87, "learning_rate": 2.5782248549838797e-08, "loss": 0.4959, "step": 24073 }, { "epoch": 2.87, "learning_rate": 2.5736226348458683e-08, "loss": 0.571, "step": 24074 }, { "epoch": 2.87, "learning_rate": 2.5690245046866292e-08, "loss": 0.5418, "step": 24075 }, { "epoch": 2.87, "learning_rate": 2.5644304645821848e-08, "loss": 0.317, "step": 24076 }, { "epoch": 2.87, "learning_rate": 2.5598405146084747e-08, "loss": 0.4156, "step": 24077 }, { "epoch": 2.87, "learning_rate": 2.555254654841327e-08, "loss": 0.3878, "step": 24078 }, { "epoch": 2.87, "learning_rate": 2.5506728853566255e-08, "loss": 0.4561, "step": 24079 }, { "epoch": 2.87, "learning_rate": 2.5460952062300594e-08, "loss": 0.6306, "step": 24080 }, { "epoch": 2.87, "learning_rate": 2.5415216175372903e-08, "loss": 0.4846, "step": 24081 }, { "epoch": 2.87, "learning_rate": 2.5369521193539247e-08, "loss": 0.3978, "step": 24082 }, { "epoch": 2.87, "learning_rate": 2.532386711755541e-08, "loss": 0.3307, "step": 24083 }, { "epoch": 2.87, "learning_rate": 2.5278253948175237e-08, "loss": 0.4735, "step": 24084 }, { "epoch": 2.87, "learning_rate": 2.523268168615339e-08, "loss": 0.4636, "step": 24085 }, { "epoch": 2.87, "learning_rate": 2.5187150332242894e-08, "loss": 0.4694, "step": 24086 }, { "epoch": 2.87, "learning_rate": 2.5141659887196746e-08, "loss": 0.6771, "step": 24087 }, { "epoch": 2.87, "learning_rate": 2.5096210351766294e-08, "loss": 0.4376, "step": 24088 }, { "epoch": 2.87, "learning_rate": 2.5050801726703155e-08, "loss": 0.3528, "step": 24089 }, { "epoch": 2.87, "learning_rate": 2.5005434012757845e-08, "loss": 0.3838, "step": 24090 }, { "epoch": 2.87, "learning_rate": 2.4960107210680597e-08, "loss": 0.4635, "step": 24091 }, { "epoch": 2.87, "learning_rate": 2.491482132122025e-08, "loss": 0.4468, "step": 24092 }, { "epoch": 2.87, "learning_rate": 2.4869576345125934e-08, "loss": 0.6097, "step": 24093 }, { "epoch": 2.87, "learning_rate": 2.4824372283144828e-08, "loss": 0.5134, "step": 24094 }, { "epoch": 2.87, "learning_rate": 2.477920913602494e-08, "loss": 0.5019, "step": 24095 }, { "epoch": 2.87, "learning_rate": 2.4734086904512066e-08, "loss": 0.4409, "step": 24096 }, { "epoch": 2.87, "learning_rate": 2.468900558935228e-08, "loss": 0.4071, "step": 24097 }, { "epoch": 2.87, "learning_rate": 2.4643965191291364e-08, "loss": 0.6277, "step": 24098 }, { "epoch": 2.87, "learning_rate": 2.4598965711073174e-08, "loss": 0.6307, "step": 24099 }, { "epoch": 2.87, "learning_rate": 2.455400714944184e-08, "loss": 0.5496, "step": 24100 }, { "epoch": 2.87, "learning_rate": 2.450908950714037e-08, "loss": 0.3596, "step": 24101 }, { "epoch": 2.87, "learning_rate": 2.4464212784911512e-08, "loss": 0.3371, "step": 24102 }, { "epoch": 2.87, "learning_rate": 2.441937698349689e-08, "loss": 0.4006, "step": 24103 }, { "epoch": 2.87, "learning_rate": 2.437458210363758e-08, "loss": 0.549, "step": 24104 }, { "epoch": 2.87, "learning_rate": 2.4329828146074096e-08, "loss": 0.5379, "step": 24105 }, { "epoch": 2.87, "learning_rate": 2.4285115111546688e-08, "loss": 0.5228, "step": 24106 }, { "epoch": 2.87, "learning_rate": 2.4240443000793646e-08, "loss": 0.3824, "step": 24107 }, { "epoch": 2.87, "learning_rate": 2.4195811814554104e-08, "loss": 0.4116, "step": 24108 }, { "epoch": 2.87, "learning_rate": 2.415122155356525e-08, "loss": 0.3393, "step": 24109 }, { "epoch": 2.87, "learning_rate": 2.4106672218564554e-08, "loss": 0.4926, "step": 24110 }, { "epoch": 2.87, "learning_rate": 2.4062163810288363e-08, "loss": 0.5283, "step": 24111 }, { "epoch": 2.87, "learning_rate": 2.4017696329472205e-08, "loss": 0.6168, "step": 24112 }, { "epoch": 2.87, "learning_rate": 2.3973269776851604e-08, "loss": 0.4482, "step": 24113 }, { "epoch": 2.87, "learning_rate": 2.3928884153160138e-08, "loss": 0.3513, "step": 24114 }, { "epoch": 2.87, "learning_rate": 2.3884539459132496e-08, "loss": 0.3743, "step": 24115 }, { "epoch": 2.87, "learning_rate": 2.384023569550087e-08, "loss": 0.4956, "step": 24116 }, { "epoch": 2.87, "learning_rate": 2.3795972862998007e-08, "loss": 0.551, "step": 24117 }, { "epoch": 2.87, "learning_rate": 2.3751750962355268e-08, "loss": 0.5394, "step": 24118 }, { "epoch": 2.87, "learning_rate": 2.3707569994304013e-08, "loss": 0.6914, "step": 24119 }, { "epoch": 2.87, "learning_rate": 2.3663429959574492e-08, "loss": 0.3388, "step": 24120 }, { "epoch": 2.87, "learning_rate": 2.3619330858896117e-08, "loss": 0.4016, "step": 24121 }, { "epoch": 2.87, "learning_rate": 2.357527269299803e-08, "loss": 0.4349, "step": 24122 }, { "epoch": 2.87, "learning_rate": 2.353125546260854e-08, "loss": 0.46, "step": 24123 }, { "epoch": 2.87, "learning_rate": 2.3487279168455113e-08, "loss": 0.5269, "step": 24124 }, { "epoch": 2.87, "learning_rate": 2.344334381126495e-08, "loss": 0.6638, "step": 24125 }, { "epoch": 2.87, "learning_rate": 2.3399449391763863e-08, "loss": 0.3714, "step": 24126 }, { "epoch": 2.87, "learning_rate": 2.3355595910677654e-08, "loss": 0.3149, "step": 24127 }, { "epoch": 2.87, "learning_rate": 2.3311783368731578e-08, "loss": 0.3418, "step": 24128 }, { "epoch": 2.87, "learning_rate": 2.3268011766649223e-08, "loss": 0.3978, "step": 24129 }, { "epoch": 2.87, "learning_rate": 2.3224281105154733e-08, "loss": 0.5348, "step": 24130 }, { "epoch": 2.87, "learning_rate": 2.3180591384970307e-08, "loss": 0.6313, "step": 24131 }, { "epoch": 2.87, "learning_rate": 2.3136942606818978e-08, "loss": 0.2644, "step": 24132 }, { "epoch": 2.87, "learning_rate": 2.309333477142156e-08, "loss": 0.3073, "step": 24133 }, { "epoch": 2.87, "learning_rate": 2.3049767879498863e-08, "loss": 0.4404, "step": 24134 }, { "epoch": 2.87, "learning_rate": 2.30062419317717e-08, "loss": 0.5292, "step": 24135 }, { "epoch": 2.87, "learning_rate": 2.2962756928959218e-08, "loss": 0.5347, "step": 24136 }, { "epoch": 2.87, "learning_rate": 2.2919312871780287e-08, "loss": 0.5909, "step": 24137 }, { "epoch": 2.87, "learning_rate": 2.2875909760952665e-08, "loss": 0.3683, "step": 24138 }, { "epoch": 2.87, "learning_rate": 2.283254759719411e-08, "loss": 0.396, "step": 24139 }, { "epoch": 2.87, "learning_rate": 2.2789226381221553e-08, "loss": 0.3638, "step": 24140 }, { "epoch": 2.87, "learning_rate": 2.2745946113750804e-08, "loss": 0.4086, "step": 24141 }, { "epoch": 2.88, "learning_rate": 2.2702706795497686e-08, "loss": 0.4756, "step": 24142 }, { "epoch": 2.88, "learning_rate": 2.2659508427176623e-08, "loss": 0.6422, "step": 24143 }, { "epoch": 2.88, "learning_rate": 2.261635100950149e-08, "loss": 0.5968, "step": 24144 }, { "epoch": 2.88, "learning_rate": 2.2573234543186163e-08, "loss": 0.4125, "step": 24145 }, { "epoch": 2.88, "learning_rate": 2.253015902894312e-08, "loss": 0.3997, "step": 24146 }, { "epoch": 2.88, "learning_rate": 2.24871244674843e-08, "loss": 0.4448, "step": 24147 }, { "epoch": 2.88, "learning_rate": 2.244413085952135e-08, "loss": 0.4232, "step": 24148 }, { "epoch": 2.88, "learning_rate": 2.240117820576454e-08, "loss": 0.6539, "step": 24149 }, { "epoch": 2.88, "learning_rate": 2.235826650692441e-08, "loss": 0.7395, "step": 24150 }, { "epoch": 2.88, "learning_rate": 2.2315395763710113e-08, "loss": 0.3608, "step": 24151 }, { "epoch": 2.88, "learning_rate": 2.2272565976829974e-08, "loss": 0.4027, "step": 24152 }, { "epoch": 2.88, "learning_rate": 2.2229777146992314e-08, "loss": 0.493, "step": 24153 }, { "epoch": 2.88, "learning_rate": 2.218702927490435e-08, "loss": 0.4776, "step": 24154 }, { "epoch": 2.88, "learning_rate": 2.214432236127273e-08, "loss": 0.5843, "step": 24155 }, { "epoch": 2.88, "learning_rate": 2.2101656406803284e-08, "loss": 0.6811, "step": 24156 }, { "epoch": 2.88, "learning_rate": 2.205903141220128e-08, "loss": 0.2866, "step": 24157 }, { "epoch": 2.88, "learning_rate": 2.2016447378171435e-08, "loss": 0.4304, "step": 24158 }, { "epoch": 2.88, "learning_rate": 2.1973904305417904e-08, "loss": 0.4093, "step": 24159 }, { "epoch": 2.88, "learning_rate": 2.193140219464346e-08, "loss": 0.5549, "step": 24160 }, { "epoch": 2.88, "learning_rate": 2.1888941046550872e-08, "loss": 0.5308, "step": 24161 }, { "epoch": 2.88, "learning_rate": 2.184652086184208e-08, "loss": 0.6621, "step": 24162 }, { "epoch": 2.88, "learning_rate": 2.180414164121819e-08, "loss": 0.4059, "step": 24163 }, { "epoch": 2.88, "learning_rate": 2.1761803385379755e-08, "loss": 0.3962, "step": 24164 }, { "epoch": 2.88, "learning_rate": 2.1719506095026767e-08, "loss": 0.4376, "step": 24165 }, { "epoch": 2.88, "learning_rate": 2.1677249770858112e-08, "loss": 0.4561, "step": 24166 }, { "epoch": 2.88, "learning_rate": 2.16350344135724e-08, "loss": 0.5685, "step": 24167 }, { "epoch": 2.88, "learning_rate": 2.159286002386768e-08, "loss": 0.5635, "step": 24168 }, { "epoch": 2.88, "learning_rate": 2.1550726602440895e-08, "loss": 0.5296, "step": 24169 }, { "epoch": 2.88, "learning_rate": 2.1508634149988717e-08, "loss": 0.3137, "step": 24170 }, { "epoch": 2.88, "learning_rate": 2.1466582667206415e-08, "loss": 0.3667, "step": 24171 }, { "epoch": 2.88, "learning_rate": 2.1424572154789826e-08, "loss": 0.3937, "step": 24172 }, { "epoch": 2.88, "learning_rate": 2.1382602613432833e-08, "loss": 0.4183, "step": 24173 }, { "epoch": 2.88, "learning_rate": 2.134067404382961e-08, "loss": 0.6244, "step": 24174 }, { "epoch": 2.88, "learning_rate": 2.1298786446672937e-08, "loss": 0.519, "step": 24175 }, { "epoch": 2.88, "learning_rate": 2.125693982265503e-08, "loss": 0.3491, "step": 24176 }, { "epoch": 2.88, "learning_rate": 2.1215134172468122e-08, "loss": 0.3082, "step": 24177 }, { "epoch": 2.88, "learning_rate": 2.1173369496803044e-08, "loss": 0.3921, "step": 24178 }, { "epoch": 2.88, "learning_rate": 2.1131645796350075e-08, "loss": 0.5708, "step": 24179 }, { "epoch": 2.88, "learning_rate": 2.1089963071799225e-08, "loss": 0.6057, "step": 24180 }, { "epoch": 2.88, "learning_rate": 2.1048321323839106e-08, "loss": 0.6133, "step": 24181 }, { "epoch": 2.88, "learning_rate": 2.100672055315833e-08, "loss": 0.3127, "step": 24182 }, { "epoch": 2.88, "learning_rate": 2.0965160760444135e-08, "loss": 0.3342, "step": 24183 }, { "epoch": 2.88, "learning_rate": 2.092364194638402e-08, "loss": 0.3604, "step": 24184 }, { "epoch": 2.88, "learning_rate": 2.0882164111664384e-08, "loss": 0.4312, "step": 24185 }, { "epoch": 2.88, "learning_rate": 2.084072725697023e-08, "loss": 0.5435, "step": 24186 }, { "epoch": 2.88, "learning_rate": 2.079933138298712e-08, "loss": 0.6833, "step": 24187 }, { "epoch": 2.88, "learning_rate": 2.0757976490398958e-08, "loss": 0.466, "step": 24188 }, { "epoch": 2.88, "learning_rate": 2.0716662579889636e-08, "loss": 0.3419, "step": 24189 }, { "epoch": 2.88, "learning_rate": 2.0675389652141664e-08, "loss": 0.3569, "step": 24190 }, { "epoch": 2.88, "learning_rate": 2.0634157707837553e-08, "loss": 0.5461, "step": 24191 }, { "epoch": 2.88, "learning_rate": 2.0592966747658983e-08, "loss": 0.4183, "step": 24192 }, { "epoch": 2.88, "learning_rate": 2.055181677228679e-08, "loss": 0.5793, "step": 24193 }, { "epoch": 2.88, "learning_rate": 2.0510707782400995e-08, "loss": 0.5851, "step": 24194 }, { "epoch": 2.88, "learning_rate": 2.046963977868105e-08, "loss": 0.312, "step": 24195 }, { "epoch": 2.88, "learning_rate": 2.042861276180641e-08, "loss": 0.4608, "step": 24196 }, { "epoch": 2.88, "learning_rate": 2.0387626732454592e-08, "loss": 0.5052, "step": 24197 }, { "epoch": 2.88, "learning_rate": 2.0346681691303383e-08, "loss": 0.4074, "step": 24198 }, { "epoch": 2.88, "learning_rate": 2.0305777639029467e-08, "loss": 0.6738, "step": 24199 }, { "epoch": 2.88, "learning_rate": 2.0264914576309523e-08, "loss": 0.5958, "step": 24200 }, { "epoch": 2.88, "learning_rate": 2.0224092503818293e-08, "loss": 0.3464, "step": 24201 }, { "epoch": 2.88, "learning_rate": 2.018331142223079e-08, "loss": 0.3685, "step": 24202 }, { "epoch": 2.88, "learning_rate": 2.0142571332221472e-08, "loss": 0.4148, "step": 24203 }, { "epoch": 2.88, "learning_rate": 2.0101872234463414e-08, "loss": 0.5349, "step": 24204 }, { "epoch": 2.88, "learning_rate": 2.0061214129629415e-08, "loss": 0.674, "step": 24205 }, { "epoch": 2.88, "learning_rate": 2.0020597018391707e-08, "loss": 0.4971, "step": 24206 }, { "epoch": 2.88, "learning_rate": 1.9980020901421704e-08, "loss": 0.315, "step": 24207 }, { "epoch": 2.88, "learning_rate": 1.9939485779389978e-08, "loss": 0.3116, "step": 24208 }, { "epoch": 2.88, "learning_rate": 1.9898991652966825e-08, "loss": 0.4758, "step": 24209 }, { "epoch": 2.88, "learning_rate": 1.9858538522821158e-08, "loss": 0.4818, "step": 24210 }, { "epoch": 2.88, "learning_rate": 1.9818126389622162e-08, "loss": 0.5999, "step": 24211 }, { "epoch": 2.88, "learning_rate": 1.9777755254037355e-08, "loss": 0.5613, "step": 24212 }, { "epoch": 2.88, "learning_rate": 1.9737425116734544e-08, "loss": 0.3888, "step": 24213 }, { "epoch": 2.88, "learning_rate": 1.9697135978380134e-08, "loss": 0.3714, "step": 24214 }, { "epoch": 2.88, "learning_rate": 1.965688783964026e-08, "loss": 0.419, "step": 24215 }, { "epoch": 2.88, "learning_rate": 1.9616680701180223e-08, "loss": 0.4989, "step": 24216 }, { "epoch": 2.88, "learning_rate": 1.9576514563664495e-08, "loss": 0.5254, "step": 24217 }, { "epoch": 2.88, "learning_rate": 1.9536389427756984e-08, "loss": 0.6126, "step": 24218 }, { "epoch": 2.88, "learning_rate": 1.949630529412133e-08, "loss": 0.5757, "step": 24219 }, { "epoch": 2.88, "learning_rate": 1.9456262163419502e-08, "loss": 0.325, "step": 24220 }, { "epoch": 2.88, "learning_rate": 1.9416260036314026e-08, "loss": 0.4326, "step": 24221 }, { "epoch": 2.88, "learning_rate": 1.937629891346604e-08, "loss": 0.4395, "step": 24222 }, { "epoch": 2.88, "learning_rate": 1.933637879553585e-08, "loss": 0.5428, "step": 24223 }, { "epoch": 2.88, "learning_rate": 1.929649968318348e-08, "loss": 0.5322, "step": 24224 }, { "epoch": 2.88, "learning_rate": 1.9256661577068124e-08, "loss": 0.5391, "step": 24225 }, { "epoch": 2.89, "learning_rate": 1.9216864477848428e-08, "loss": 0.3326, "step": 24226 }, { "epoch": 2.89, "learning_rate": 1.917710838618192e-08, "loss": 0.426, "step": 24227 }, { "epoch": 2.89, "learning_rate": 1.9137393302725847e-08, "loss": 0.416, "step": 24228 }, { "epoch": 2.89, "learning_rate": 1.909771922813719e-08, "loss": 0.5165, "step": 24229 }, { "epoch": 2.89, "learning_rate": 1.9058086163071255e-08, "loss": 0.5097, "step": 24230 }, { "epoch": 2.89, "learning_rate": 1.9018494108183072e-08, "loss": 0.6237, "step": 24231 }, { "epoch": 2.89, "learning_rate": 1.8978943064127676e-08, "loss": 0.3236, "step": 24232 }, { "epoch": 2.89, "learning_rate": 1.8939433031558152e-08, "loss": 0.3476, "step": 24233 }, { "epoch": 2.89, "learning_rate": 1.8899964011128424e-08, "loss": 0.4334, "step": 24234 }, { "epoch": 2.89, "learning_rate": 1.8860536003490193e-08, "loss": 0.48, "step": 24235 }, { "epoch": 2.89, "learning_rate": 1.8821149009295715e-08, "loss": 0.5806, "step": 24236 }, { "epoch": 2.89, "learning_rate": 1.8781803029195577e-08, "loss": 0.5009, "step": 24237 }, { "epoch": 2.89, "learning_rate": 1.874249806384065e-08, "loss": 0.4017, "step": 24238 }, { "epoch": 2.89, "learning_rate": 1.8703234113880143e-08, "loss": 0.356, "step": 24239 }, { "epoch": 2.89, "learning_rate": 1.8664011179963526e-08, "loss": 0.3738, "step": 24240 }, { "epoch": 2.89, "learning_rate": 1.8624829262739176e-08, "loss": 0.4339, "step": 24241 }, { "epoch": 2.89, "learning_rate": 1.8585688362854348e-08, "loss": 0.6048, "step": 24242 }, { "epoch": 2.89, "learning_rate": 1.8546588480956307e-08, "loss": 0.6768, "step": 24243 }, { "epoch": 2.89, "learning_rate": 1.8507529617691754e-08, "loss": 0.4859, "step": 24244 }, { "epoch": 2.89, "learning_rate": 1.8468511773705733e-08, "loss": 0.3159, "step": 24245 }, { "epoch": 2.89, "learning_rate": 1.842953494964328e-08, "loss": 0.3782, "step": 24246 }, { "epoch": 2.89, "learning_rate": 1.8390599146148878e-08, "loss": 0.3997, "step": 24247 }, { "epoch": 2.89, "learning_rate": 1.8351704363865908e-08, "loss": 0.4655, "step": 24248 }, { "epoch": 2.89, "learning_rate": 1.8312850603438016e-08, "loss": 0.588, "step": 24249 }, { "epoch": 2.89, "learning_rate": 1.8274037865506356e-08, "loss": 0.5978, "step": 24250 }, { "epoch": 2.89, "learning_rate": 1.8235266150713472e-08, "loss": 0.3204, "step": 24251 }, { "epoch": 2.89, "learning_rate": 1.819653545969968e-08, "loss": 0.3052, "step": 24252 }, { "epoch": 2.89, "learning_rate": 1.815784579310559e-08, "loss": 0.4601, "step": 24253 }, { "epoch": 2.89, "learning_rate": 1.81191971515704e-08, "loss": 0.4175, "step": 24254 }, { "epoch": 2.89, "learning_rate": 1.8080589535733327e-08, "loss": 0.5421, "step": 24255 }, { "epoch": 2.89, "learning_rate": 1.8042022946232196e-08, "loss": 0.5923, "step": 24256 }, { "epoch": 2.89, "learning_rate": 1.8003497383704547e-08, "loss": 0.3517, "step": 24257 }, { "epoch": 2.89, "learning_rate": 1.7965012848787655e-08, "loss": 0.3697, "step": 24258 }, { "epoch": 2.89, "learning_rate": 1.7926569342117118e-08, "loss": 0.4084, "step": 24259 }, { "epoch": 2.89, "learning_rate": 1.7888166864328815e-08, "loss": 0.4403, "step": 24260 }, { "epoch": 2.89, "learning_rate": 1.7849805416057242e-08, "loss": 0.537, "step": 24261 }, { "epoch": 2.89, "learning_rate": 1.7811484997936613e-08, "loss": 0.6293, "step": 24262 }, { "epoch": 2.89, "learning_rate": 1.7773205610600587e-08, "loss": 0.3507, "step": 24263 }, { "epoch": 2.89, "learning_rate": 1.7734967254681713e-08, "loss": 0.3788, "step": 24264 }, { "epoch": 2.89, "learning_rate": 1.7696769930811985e-08, "loss": 0.3007, "step": 24265 }, { "epoch": 2.89, "learning_rate": 1.7658613639623125e-08, "loss": 0.4614, "step": 24266 }, { "epoch": 2.89, "learning_rate": 1.7620498381745456e-08, "loss": 0.5092, "step": 24267 }, { "epoch": 2.89, "learning_rate": 1.7582424157809307e-08, "loss": 0.6069, "step": 24268 }, { "epoch": 2.89, "learning_rate": 1.7544390968443904e-08, "loss": 0.5577, "step": 24269 }, { "epoch": 2.89, "learning_rate": 1.750639881427818e-08, "loss": 0.348, "step": 24270 }, { "epoch": 2.89, "learning_rate": 1.7468447695939973e-08, "loss": 0.4178, "step": 24271 }, { "epoch": 2.89, "learning_rate": 1.7430537614056552e-08, "loss": 0.4199, "step": 24272 }, { "epoch": 2.89, "learning_rate": 1.7392668569254645e-08, "loss": 0.4323, "step": 24273 }, { "epoch": 2.89, "learning_rate": 1.7354840562160412e-08, "loss": 0.5605, "step": 24274 }, { "epoch": 2.89, "learning_rate": 1.7317053593398913e-08, "loss": 0.576, "step": 24275 }, { "epoch": 2.89, "learning_rate": 1.7279307663594646e-08, "loss": 0.3087, "step": 24276 }, { "epoch": 2.89, "learning_rate": 1.7241602773371556e-08, "loss": 0.4212, "step": 24277 }, { "epoch": 2.89, "learning_rate": 1.7203938923353312e-08, "loss": 0.4199, "step": 24278 }, { "epoch": 2.89, "learning_rate": 1.7166316114162473e-08, "loss": 0.4245, "step": 24279 }, { "epoch": 2.89, "learning_rate": 1.712873434642076e-08, "loss": 0.5638, "step": 24280 }, { "epoch": 2.89, "learning_rate": 1.7091193620749345e-08, "loss": 0.6344, "step": 24281 }, { "epoch": 2.89, "learning_rate": 1.705369393776857e-08, "loss": 0.3783, "step": 24282 }, { "epoch": 2.89, "learning_rate": 1.701623529809876e-08, "loss": 0.4105, "step": 24283 }, { "epoch": 2.89, "learning_rate": 1.6978817702358873e-08, "loss": 0.3754, "step": 24284 }, { "epoch": 2.89, "learning_rate": 1.6941441151167582e-08, "loss": 0.4495, "step": 24285 }, { "epoch": 2.89, "learning_rate": 1.6904105645142443e-08, "loss": 0.5318, "step": 24286 }, { "epoch": 2.89, "learning_rate": 1.6866811184901023e-08, "loss": 0.6347, "step": 24287 }, { "epoch": 2.89, "learning_rate": 1.6829557771059492e-08, "loss": 0.5369, "step": 24288 }, { "epoch": 2.89, "learning_rate": 1.6792345404233467e-08, "loss": 0.3589, "step": 24289 }, { "epoch": 2.89, "learning_rate": 1.6755174085038574e-08, "loss": 0.3752, "step": 24290 }, { "epoch": 2.89, "learning_rate": 1.6718043814089035e-08, "loss": 0.4339, "step": 24291 }, { "epoch": 2.89, "learning_rate": 1.6680954591998534e-08, "loss": 0.5065, "step": 24292 }, { "epoch": 2.89, "learning_rate": 1.6643906419380183e-08, "loss": 0.6624, "step": 24293 }, { "epoch": 2.89, "learning_rate": 1.6606899296846556e-08, "loss": 0.6224, "step": 24294 }, { "epoch": 2.89, "learning_rate": 1.6569933225009105e-08, "loss": 0.4792, "step": 24295 }, { "epoch": 2.89, "learning_rate": 1.6533008204479007e-08, "loss": 0.3849, "step": 24296 }, { "epoch": 2.89, "learning_rate": 1.6496124235866883e-08, "loss": 0.4086, "step": 24297 }, { "epoch": 2.89, "learning_rate": 1.6459281319781972e-08, "loss": 0.5275, "step": 24298 }, { "epoch": 2.89, "learning_rate": 1.6422479456833784e-08, "loss": 0.5919, "step": 24299 }, { "epoch": 2.89, "learning_rate": 1.6385718647630166e-08, "loss": 0.5757, "step": 24300 }, { "epoch": 2.89, "learning_rate": 1.6348998892778966e-08, "loss": 0.2802, "step": 24301 }, { "epoch": 2.89, "learning_rate": 1.6312320192887478e-08, "loss": 0.3483, "step": 24302 }, { "epoch": 2.89, "learning_rate": 1.62756825485616e-08, "loss": 0.4918, "step": 24303 }, { "epoch": 2.89, "learning_rate": 1.6239085960406965e-08, "loss": 0.4586, "step": 24304 }, { "epoch": 2.89, "learning_rate": 1.6202530429028916e-08, "loss": 0.587, "step": 24305 }, { "epoch": 2.89, "learning_rate": 1.6166015955031146e-08, "loss": 0.6647, "step": 24306 }, { "epoch": 2.89, "learning_rate": 1.6129542539017607e-08, "loss": 0.2934, "step": 24307 }, { "epoch": 2.89, "learning_rate": 1.6093110181591432e-08, "loss": 0.3506, "step": 24308 }, { "epoch": 2.89, "learning_rate": 1.605671888335436e-08, "loss": 0.3085, "step": 24309 }, { "epoch": 2.9, "learning_rate": 1.6020368644908134e-08, "loss": 0.4449, "step": 24310 }, { "epoch": 2.9, "learning_rate": 1.5984059466853662e-08, "loss": 0.436, "step": 24311 }, { "epoch": 2.9, "learning_rate": 1.5947791349791018e-08, "loss": 0.662, "step": 24312 }, { "epoch": 2.9, "learning_rate": 1.5911564294320004e-08, "loss": 0.311, "step": 24313 }, { "epoch": 2.9, "learning_rate": 1.5875378301039023e-08, "loss": 0.3304, "step": 24314 }, { "epoch": 2.9, "learning_rate": 1.583923337054677e-08, "loss": 0.3284, "step": 24315 }, { "epoch": 2.9, "learning_rate": 1.5803129503440263e-08, "loss": 0.4309, "step": 24316 }, { "epoch": 2.9, "learning_rate": 1.5767066700316524e-08, "loss": 0.5791, "step": 24317 }, { "epoch": 2.9, "learning_rate": 1.573104496177147e-08, "loss": 0.5764, "step": 24318 }, { "epoch": 2.9, "learning_rate": 1.569506428840073e-08, "loss": 0.541, "step": 24319 }, { "epoch": 2.9, "learning_rate": 1.565912468079911e-08, "loss": 0.3036, "step": 24320 }, { "epoch": 2.9, "learning_rate": 1.5623226139560576e-08, "loss": 0.4674, "step": 24321 }, { "epoch": 2.9, "learning_rate": 1.5587368665278547e-08, "loss": 0.4096, "step": 24322 }, { "epoch": 2.9, "learning_rate": 1.5551552258545877e-08, "loss": 0.4774, "step": 24323 }, { "epoch": 2.9, "learning_rate": 1.5515776919954318e-08, "loss": 0.5487, "step": 24324 }, { "epoch": 2.9, "learning_rate": 1.5480042650095617e-08, "loss": 0.5979, "step": 24325 }, { "epoch": 2.9, "learning_rate": 1.5444349449560137e-08, "loss": 0.353, "step": 24326 }, { "epoch": 2.9, "learning_rate": 1.540869731893796e-08, "loss": 0.3532, "step": 24327 }, { "epoch": 2.9, "learning_rate": 1.5373086258818893e-08, "loss": 0.4426, "step": 24328 }, { "epoch": 2.9, "learning_rate": 1.5337516269790796e-08, "loss": 0.4851, "step": 24329 }, { "epoch": 2.9, "learning_rate": 1.5301987352442372e-08, "loss": 0.5951, "step": 24330 }, { "epoch": 2.9, "learning_rate": 1.526649950736009e-08, "loss": 0.5757, "step": 24331 }, { "epoch": 2.9, "learning_rate": 1.5231052735131536e-08, "loss": 0.4238, "step": 24332 }, { "epoch": 2.9, "learning_rate": 1.51956470363418e-08, "loss": 0.3463, "step": 24333 }, { "epoch": 2.9, "learning_rate": 1.51602824115768e-08, "loss": 0.3912, "step": 24334 }, { "epoch": 2.9, "learning_rate": 1.5124958861420792e-08, "loss": 0.4058, "step": 24335 }, { "epoch": 2.9, "learning_rate": 1.5089676386457476e-08, "loss": 0.5193, "step": 24336 }, { "epoch": 2.9, "learning_rate": 1.505443498727055e-08, "loss": 0.6072, "step": 24337 }, { "epoch": 2.9, "learning_rate": 1.501923466444233e-08, "loss": 0.4372, "step": 24338 }, { "epoch": 2.9, "learning_rate": 1.4984075418554566e-08, "loss": 0.2814, "step": 24339 }, { "epoch": 2.9, "learning_rate": 1.4948957250188466e-08, "loss": 0.3282, "step": 24340 }, { "epoch": 2.9, "learning_rate": 1.4913880159924676e-08, "loss": 0.389, "step": 24341 }, { "epoch": 2.9, "learning_rate": 1.4878844148343008e-08, "loss": 0.4753, "step": 24342 }, { "epoch": 2.9, "learning_rate": 1.484384921602272e-08, "loss": 0.6298, "step": 24343 }, { "epoch": 2.9, "learning_rate": 1.4808895363541963e-08, "loss": 0.5201, "step": 24344 }, { "epoch": 2.9, "learning_rate": 1.4773982591478609e-08, "loss": 0.3362, "step": 24345 }, { "epoch": 2.9, "learning_rate": 1.473911090040997e-08, "loss": 0.4319, "step": 24346 }, { "epoch": 2.9, "learning_rate": 1.4704280290912254e-08, "loss": 0.5138, "step": 24347 }, { "epoch": 2.9, "learning_rate": 1.4669490763561389e-08, "loss": 0.4591, "step": 24348 }, { "epoch": 2.9, "learning_rate": 1.4634742318932193e-08, "loss": 0.6217, "step": 24349 }, { "epoch": 2.9, "learning_rate": 1.4600034957599484e-08, "loss": 0.4668, "step": 24350 }, { "epoch": 2.9, "learning_rate": 1.4565368680136694e-08, "loss": 0.3394, "step": 24351 }, { "epoch": 2.9, "learning_rate": 1.4530743487116972e-08, "loss": 0.3486, "step": 24352 }, { "epoch": 2.9, "learning_rate": 1.4496159379112362e-08, "loss": 0.4312, "step": 24353 }, { "epoch": 2.9, "learning_rate": 1.4461616356695185e-08, "loss": 0.4641, "step": 24354 }, { "epoch": 2.9, "learning_rate": 1.4427114420435817e-08, "loss": 0.5067, "step": 24355 }, { "epoch": 2.9, "learning_rate": 1.4392653570904636e-08, "loss": 0.6286, "step": 24356 }, { "epoch": 2.9, "learning_rate": 1.4358233808671462e-08, "loss": 0.3655, "step": 24357 }, { "epoch": 2.9, "learning_rate": 1.4323855134305564e-08, "loss": 0.399, "step": 24358 }, { "epoch": 2.9, "learning_rate": 1.428951754837482e-08, "loss": 0.4152, "step": 24359 }, { "epoch": 2.9, "learning_rate": 1.4255221051446833e-08, "loss": 0.4277, "step": 24360 }, { "epoch": 2.9, "learning_rate": 1.4220965644088647e-08, "loss": 0.5444, "step": 24361 }, { "epoch": 2.9, "learning_rate": 1.4186751326866477e-08, "loss": 0.5108, "step": 24362 }, { "epoch": 2.9, "learning_rate": 1.4152578100345704e-08, "loss": 0.3172, "step": 24363 }, { "epoch": 2.9, "learning_rate": 1.4118445965091432e-08, "loss": 0.3108, "step": 24364 }, { "epoch": 2.9, "learning_rate": 1.4084354921668209e-08, "loss": 0.4854, "step": 24365 }, { "epoch": 2.9, "learning_rate": 1.4050304970638918e-08, "loss": 0.4531, "step": 24366 }, { "epoch": 2.9, "learning_rate": 1.4016296112566442e-08, "loss": 0.513, "step": 24367 }, { "epoch": 2.9, "learning_rate": 1.3982328348013385e-08, "loss": 0.6432, "step": 24368 }, { "epoch": 2.9, "learning_rate": 1.3948401677541245e-08, "loss": 0.5479, "step": 24369 }, { "epoch": 2.9, "learning_rate": 1.3914516101710406e-08, "loss": 0.4419, "step": 24370 }, { "epoch": 2.9, "learning_rate": 1.3880671621081254e-08, "loss": 0.4086, "step": 24371 }, { "epoch": 2.9, "learning_rate": 1.3846868236213063e-08, "loss": 0.4733, "step": 24372 }, { "epoch": 2.9, "learning_rate": 1.3813105947665107e-08, "loss": 0.5719, "step": 24373 }, { "epoch": 2.9, "learning_rate": 1.3779384755994718e-08, "loss": 0.6508, "step": 24374 }, { "epoch": 2.9, "learning_rate": 1.3745704661760062e-08, "loss": 0.6975, "step": 24375 }, { "epoch": 2.9, "learning_rate": 1.3712065665517082e-08, "loss": 0.3513, "step": 24376 }, { "epoch": 2.9, "learning_rate": 1.3678467767822557e-08, "loss": 0.4576, "step": 24377 }, { "epoch": 2.9, "learning_rate": 1.3644910969231595e-08, "loss": 0.4389, "step": 24378 }, { "epoch": 2.9, "learning_rate": 1.3611395270298755e-08, "loss": 0.5067, "step": 24379 }, { "epoch": 2.9, "learning_rate": 1.3577920671578314e-08, "loss": 0.621, "step": 24380 }, { "epoch": 2.9, "learning_rate": 1.3544487173623444e-08, "loss": 0.5266, "step": 24381 }, { "epoch": 2.9, "learning_rate": 1.3511094776986755e-08, "loss": 0.2795, "step": 24382 }, { "epoch": 2.9, "learning_rate": 1.3477743482220306e-08, "loss": 0.3636, "step": 24383 }, { "epoch": 2.9, "learning_rate": 1.3444433289875602e-08, "loss": 0.4725, "step": 24384 }, { "epoch": 2.9, "learning_rate": 1.3411164200502758e-08, "loss": 0.4359, "step": 24385 }, { "epoch": 2.9, "learning_rate": 1.3377936214652165e-08, "loss": 0.5667, "step": 24386 }, { "epoch": 2.9, "learning_rate": 1.3344749332873108e-08, "loss": 0.6352, "step": 24387 }, { "epoch": 2.9, "learning_rate": 1.3311603555713759e-08, "loss": 0.4732, "step": 24388 }, { "epoch": 2.9, "learning_rate": 1.327849888372229e-08, "loss": 0.3203, "step": 24389 }, { "epoch": 2.9, "learning_rate": 1.3245435317445765e-08, "loss": 0.4669, "step": 24390 }, { "epoch": 2.9, "learning_rate": 1.3212412857430967e-08, "loss": 0.4141, "step": 24391 }, { "epoch": 2.9, "learning_rate": 1.317943150422385e-08, "loss": 0.4774, "step": 24392 }, { "epoch": 2.9, "learning_rate": 1.3146491258368976e-08, "loss": 0.582, "step": 24393 }, { "epoch": 2.91, "learning_rate": 1.3113592120411745e-08, "loss": 0.5275, "step": 24394 }, { "epoch": 2.91, "learning_rate": 1.3080734090895054e-08, "loss": 0.2923, "step": 24395 }, { "epoch": 2.91, "learning_rate": 1.3047917170362912e-08, "loss": 0.4793, "step": 24396 }, { "epoch": 2.91, "learning_rate": 1.3015141359357109e-08, "loss": 0.4146, "step": 24397 }, { "epoch": 2.91, "learning_rate": 1.2982406658419433e-08, "loss": 0.5401, "step": 24398 }, { "epoch": 2.91, "learning_rate": 1.2949713068091675e-08, "loss": 0.7013, "step": 24399 }, { "epoch": 2.91, "learning_rate": 1.2917060588913677e-08, "loss": 0.4996, "step": 24400 }, { "epoch": 2.91, "learning_rate": 1.2884449221425565e-08, "loss": 0.3122, "step": 24401 }, { "epoch": 2.91, "learning_rate": 1.2851878966165798e-08, "loss": 0.398, "step": 24402 }, { "epoch": 2.91, "learning_rate": 1.2819349823673665e-08, "loss": 0.3219, "step": 24403 }, { "epoch": 2.91, "learning_rate": 1.278686179448596e-08, "loss": 0.4009, "step": 24404 }, { "epoch": 2.91, "learning_rate": 1.275441487914031e-08, "loss": 0.7121, "step": 24405 }, { "epoch": 2.91, "learning_rate": 1.272200907817267e-08, "loss": 0.5976, "step": 24406 }, { "epoch": 2.91, "learning_rate": 1.2689644392119004e-08, "loss": 0.3099, "step": 24407 }, { "epoch": 2.91, "learning_rate": 1.265732082151444e-08, "loss": 0.3811, "step": 24408 }, { "epoch": 2.91, "learning_rate": 1.2625038366892994e-08, "loss": 0.415, "step": 24409 }, { "epoch": 2.91, "learning_rate": 1.2592797028788129e-08, "loss": 0.4596, "step": 24410 }, { "epoch": 2.91, "learning_rate": 1.2560596807733305e-08, "loss": 0.4867, "step": 24411 }, { "epoch": 2.91, "learning_rate": 1.2528437704260321e-08, "loss": 0.594, "step": 24412 }, { "epoch": 2.91, "learning_rate": 1.2496319718900974e-08, "loss": 0.3382, "step": 24413 }, { "epoch": 2.91, "learning_rate": 1.2464242852186504e-08, "loss": 0.4184, "step": 24414 }, { "epoch": 2.91, "learning_rate": 1.2432207104646487e-08, "loss": 0.4208, "step": 24415 }, { "epoch": 2.91, "learning_rate": 1.2400212476811057e-08, "loss": 0.4908, "step": 24416 }, { "epoch": 2.91, "learning_rate": 1.236825896920868e-08, "loss": 0.6172, "step": 24417 }, { "epoch": 2.91, "learning_rate": 1.233634658236782e-08, "loss": 0.6356, "step": 24418 }, { "epoch": 2.91, "learning_rate": 1.2304475316815834e-08, "loss": 0.49, "step": 24419 }, { "epoch": 2.91, "learning_rate": 1.2272645173079522e-08, "loss": 0.3164, "step": 24420 }, { "epoch": 2.91, "learning_rate": 1.2240856151685132e-08, "loss": 0.429, "step": 24421 }, { "epoch": 2.91, "learning_rate": 1.2209108253158076e-08, "loss": 0.3794, "step": 24422 }, { "epoch": 2.91, "learning_rate": 1.2177401478023488e-08, "loss": 0.4425, "step": 24423 }, { "epoch": 2.91, "learning_rate": 1.214573582680484e-08, "loss": 0.5568, "step": 24424 }, { "epoch": 2.91, "learning_rate": 1.2114111300026155e-08, "loss": 0.6321, "step": 24425 }, { "epoch": 2.91, "learning_rate": 1.2082527898210072e-08, "loss": 0.3253, "step": 24426 }, { "epoch": 2.91, "learning_rate": 1.2050985621878396e-08, "loss": 0.4268, "step": 24427 }, { "epoch": 2.91, "learning_rate": 1.2019484471552655e-08, "loss": 0.4101, "step": 24428 }, { "epoch": 2.91, "learning_rate": 1.1988024447753543e-08, "loss": 0.4763, "step": 24429 }, { "epoch": 2.91, "learning_rate": 1.19566055510012e-08, "loss": 0.4624, "step": 24430 }, { "epoch": 2.91, "learning_rate": 1.192522778181493e-08, "loss": 0.5866, "step": 24431 }, { "epoch": 2.91, "learning_rate": 1.1893891140713488e-08, "loss": 0.3119, "step": 24432 }, { "epoch": 2.91, "learning_rate": 1.186259562821479e-08, "loss": 0.31, "step": 24433 }, { "epoch": 2.91, "learning_rate": 1.1831341244836203e-08, "loss": 0.4039, "step": 24434 }, { "epoch": 2.91, "learning_rate": 1.1800127991093979e-08, "loss": 0.3983, "step": 24435 }, { "epoch": 2.91, "learning_rate": 1.1768955867504928e-08, "loss": 0.5194, "step": 24436 }, { "epoch": 2.91, "learning_rate": 1.173782487458336e-08, "loss": 0.4946, "step": 24437 }, { "epoch": 2.91, "learning_rate": 1.1706735012844694e-08, "loss": 0.51, "step": 24438 }, { "epoch": 2.91, "learning_rate": 1.1675686282802411e-08, "loss": 0.3691, "step": 24439 }, { "epoch": 2.91, "learning_rate": 1.1644678684969435e-08, "loss": 0.4818, "step": 24440 }, { "epoch": 2.91, "learning_rate": 1.1613712219859242e-08, "loss": 0.4358, "step": 24441 }, { "epoch": 2.91, "learning_rate": 1.1582786887982811e-08, "loss": 0.5199, "step": 24442 }, { "epoch": 2.91, "learning_rate": 1.1551902689851957e-08, "loss": 0.6198, "step": 24443 }, { "epoch": 2.91, "learning_rate": 1.1521059625976826e-08, "loss": 0.4983, "step": 24444 }, { "epoch": 2.91, "learning_rate": 1.149025769686729e-08, "loss": 0.3105, "step": 24445 }, { "epoch": 2.91, "learning_rate": 1.1459496903032662e-08, "loss": 0.471, "step": 24446 }, { "epoch": 2.91, "learning_rate": 1.1428777244981426e-08, "loss": 0.4275, "step": 24447 }, { "epoch": 2.91, "learning_rate": 1.1398098723221229e-08, "loss": 0.5342, "step": 24448 }, { "epoch": 2.91, "learning_rate": 1.1367461338259444e-08, "loss": 0.5229, "step": 24449 }, { "epoch": 2.91, "learning_rate": 1.1336865090602056e-08, "loss": 0.656, "step": 24450 }, { "epoch": 2.91, "learning_rate": 1.1306309980755325e-08, "loss": 0.3251, "step": 24451 }, { "epoch": 2.91, "learning_rate": 1.1275796009224127e-08, "loss": 0.4214, "step": 24452 }, { "epoch": 2.91, "learning_rate": 1.1245323176512778e-08, "loss": 0.418, "step": 24453 }, { "epoch": 2.91, "learning_rate": 1.1214891483125046e-08, "loss": 0.491, "step": 24454 }, { "epoch": 2.91, "learning_rate": 1.1184500929563857e-08, "loss": 0.5007, "step": 24455 }, { "epoch": 2.91, "learning_rate": 1.1154151516331868e-08, "loss": 0.6005, "step": 24456 }, { "epoch": 2.91, "learning_rate": 1.1123843243930344e-08, "loss": 0.3534, "step": 24457 }, { "epoch": 2.91, "learning_rate": 1.1093576112860826e-08, "loss": 0.3835, "step": 24458 }, { "epoch": 2.91, "learning_rate": 1.1063350123623195e-08, "loss": 0.4031, "step": 24459 }, { "epoch": 2.91, "learning_rate": 1.1033165276717327e-08, "loss": 0.414, "step": 24460 }, { "epoch": 2.91, "learning_rate": 1.1003021572641992e-08, "loss": 0.5279, "step": 24461 }, { "epoch": 2.91, "learning_rate": 1.09729190118954e-08, "loss": 0.629, "step": 24462 }, { "epoch": 2.91, "learning_rate": 1.0942857594975765e-08, "loss": 0.3486, "step": 24463 }, { "epoch": 2.91, "learning_rate": 1.0912837322379078e-08, "loss": 0.3176, "step": 24464 }, { "epoch": 2.91, "learning_rate": 1.0882858194602163e-08, "loss": 0.46, "step": 24465 }, { "epoch": 2.91, "learning_rate": 1.0852920212140461e-08, "loss": 0.4852, "step": 24466 }, { "epoch": 2.91, "learning_rate": 1.0823023375489128e-08, "loss": 0.5123, "step": 24467 }, { "epoch": 2.91, "learning_rate": 1.0793167685141659e-08, "loss": 0.5493, "step": 24468 }, { "epoch": 2.91, "learning_rate": 1.0763353141592104e-08, "loss": 0.5192, "step": 24469 }, { "epoch": 2.91, "learning_rate": 1.07335797453334e-08, "loss": 0.3001, "step": 24470 }, { "epoch": 2.91, "learning_rate": 1.0703847496857378e-08, "loss": 0.3486, "step": 24471 }, { "epoch": 2.91, "learning_rate": 1.0674156396655588e-08, "loss": 0.4172, "step": 24472 }, { "epoch": 2.91, "learning_rate": 1.064450644521875e-08, "loss": 0.4376, "step": 24473 }, { "epoch": 2.91, "learning_rate": 1.0614897643037302e-08, "loss": 0.6009, "step": 24474 }, { "epoch": 2.91, "learning_rate": 1.0585329990600301e-08, "loss": 0.5431, "step": 24475 }, { "epoch": 2.91, "learning_rate": 1.0555803488396799e-08, "loss": 0.3593, "step": 24476 }, { "epoch": 2.91, "learning_rate": 1.0526318136914736e-08, "loss": 0.3732, "step": 24477 }, { "epoch": 2.92, "learning_rate": 1.0496873936641505e-08, "loss": 0.4489, "step": 24478 }, { "epoch": 2.92, "learning_rate": 1.0467470888063658e-08, "loss": 0.4781, "step": 24479 }, { "epoch": 2.92, "learning_rate": 1.043810899166775e-08, "loss": 0.5289, "step": 24480 }, { "epoch": 2.92, "learning_rate": 1.0408788247938395e-08, "loss": 0.5596, "step": 24481 }, { "epoch": 2.92, "learning_rate": 1.0379508657361036e-08, "loss": 0.3416, "step": 24482 }, { "epoch": 2.92, "learning_rate": 1.03502702204189e-08, "loss": 0.4405, "step": 24483 }, { "epoch": 2.92, "learning_rate": 1.0321072937596044e-08, "loss": 0.3793, "step": 24484 }, { "epoch": 2.92, "learning_rate": 1.029191680937458e-08, "loss": 0.3994, "step": 24485 }, { "epoch": 2.92, "learning_rate": 1.0262801836236624e-08, "loss": 0.5972, "step": 24486 }, { "epoch": 2.92, "learning_rate": 1.0233728018663736e-08, "loss": 0.6028, "step": 24487 }, { "epoch": 2.92, "learning_rate": 1.0204695357135808e-08, "loss": 0.3066, "step": 24488 }, { "epoch": 2.92, "learning_rate": 1.0175703852133567e-08, "loss": 0.4491, "step": 24489 }, { "epoch": 2.92, "learning_rate": 1.014675350413552e-08, "loss": 0.3949, "step": 24490 }, { "epoch": 2.92, "learning_rate": 1.0117844313620728e-08, "loss": 0.3836, "step": 24491 }, { "epoch": 2.92, "learning_rate": 1.0088976281066864e-08, "loss": 0.5719, "step": 24492 }, { "epoch": 2.92, "learning_rate": 1.0060149406951325e-08, "loss": 0.5188, "step": 24493 }, { "epoch": 2.92, "learning_rate": 1.0031363691750118e-08, "loss": 0.4226, "step": 24494 }, { "epoch": 2.92, "learning_rate": 1.0002619135939528e-08, "loss": 0.4176, "step": 24495 }, { "epoch": 2.92, "learning_rate": 9.973915739994734e-09, "loss": 0.3353, "step": 24496 }, { "epoch": 2.92, "learning_rate": 9.945253504389795e-09, "loss": 0.383, "step": 24497 }, { "epoch": 2.92, "learning_rate": 9.916632429598783e-09, "loss": 0.4864, "step": 24498 }, { "epoch": 2.92, "learning_rate": 9.888052516094925e-09, "loss": 0.6001, "step": 24499 }, { "epoch": 2.92, "learning_rate": 9.859513764350348e-09, "loss": 0.5239, "step": 24500 }, { "epoch": 2.92, "learning_rate": 9.831016174837171e-09, "loss": 0.3313, "step": 24501 }, { "epoch": 2.92, "learning_rate": 9.80255974802613e-09, "loss": 0.4527, "step": 24502 }, { "epoch": 2.92, "learning_rate": 9.774144484387404e-09, "loss": 0.5154, "step": 24503 }, { "epoch": 2.92, "learning_rate": 9.74577038439145e-09, "loss": 0.5121, "step": 24504 }, { "epoch": 2.92, "learning_rate": 9.717437448506505e-09, "loss": 0.5786, "step": 24505 }, { "epoch": 2.92, "learning_rate": 9.689145677201638e-09, "loss": 0.5903, "step": 24506 }, { "epoch": 2.92, "learning_rate": 9.660895070943977e-09, "loss": 0.3309, "step": 24507 }, { "epoch": 2.92, "learning_rate": 9.632685630200644e-09, "loss": 0.315, "step": 24508 }, { "epoch": 2.92, "learning_rate": 9.604517355437936e-09, "loss": 0.4811, "step": 24509 }, { "epoch": 2.92, "learning_rate": 9.576390247121592e-09, "loss": 0.4694, "step": 24510 }, { "epoch": 2.92, "learning_rate": 9.548304305716516e-09, "loss": 0.5263, "step": 24511 }, { "epoch": 2.92, "learning_rate": 9.520259531686781e-09, "loss": 0.657, "step": 24512 }, { "epoch": 2.92, "learning_rate": 9.492255925496185e-09, "loss": 0.4393, "step": 24513 }, { "epoch": 2.92, "learning_rate": 9.46429348760769e-09, "loss": 0.272, "step": 24514 }, { "epoch": 2.92, "learning_rate": 9.436372218483426e-09, "loss": 0.3661, "step": 24515 }, { "epoch": 2.92, "learning_rate": 9.408492118584689e-09, "loss": 0.4394, "step": 24516 }, { "epoch": 2.92, "learning_rate": 9.380653188372779e-09, "loss": 0.3912, "step": 24517 }, { "epoch": 2.92, "learning_rate": 9.352855428307606e-09, "loss": 0.5457, "step": 24518 }, { "epoch": 2.92, "learning_rate": 9.325098838848523e-09, "loss": 0.5689, "step": 24519 }, { "epoch": 2.92, "learning_rate": 9.297383420454608e-09, "loss": 0.3172, "step": 24520 }, { "epoch": 2.92, "learning_rate": 9.269709173584107e-09, "loss": 0.3792, "step": 24521 }, { "epoch": 2.92, "learning_rate": 9.242076098694152e-09, "loss": 0.4917, "step": 24522 }, { "epoch": 2.92, "learning_rate": 9.214484196241602e-09, "loss": 0.4745, "step": 24523 }, { "epoch": 2.92, "learning_rate": 9.186933466682757e-09, "loss": 0.5535, "step": 24524 }, { "epoch": 2.92, "learning_rate": 9.159423910473087e-09, "loss": 0.6397, "step": 24525 }, { "epoch": 2.92, "learning_rate": 9.131955528066671e-09, "loss": 0.364, "step": 24526 }, { "epoch": 2.92, "learning_rate": 9.104528319918426e-09, "loss": 0.4008, "step": 24527 }, { "epoch": 2.92, "learning_rate": 9.077142286481045e-09, "loss": 0.3712, "step": 24528 }, { "epoch": 2.92, "learning_rate": 9.049797428207774e-09, "loss": 0.4627, "step": 24529 }, { "epoch": 2.92, "learning_rate": 9.022493745550197e-09, "loss": 0.6369, "step": 24530 }, { "epoch": 2.92, "learning_rate": 8.9952312389599e-09, "loss": 0.549, "step": 24531 }, { "epoch": 2.92, "learning_rate": 8.968009908887354e-09, "loss": 0.2843, "step": 24532 }, { "epoch": 2.92, "learning_rate": 8.940829755782753e-09, "loss": 0.4323, "step": 24533 }, { "epoch": 2.92, "learning_rate": 8.913690780094908e-09, "loss": 0.3645, "step": 24534 }, { "epoch": 2.92, "learning_rate": 8.886592982273178e-09, "loss": 0.4645, "step": 24535 }, { "epoch": 2.92, "learning_rate": 8.859536362764987e-09, "loss": 0.4817, "step": 24536 }, { "epoch": 2.92, "learning_rate": 8.832520922017473e-09, "loss": 0.6125, "step": 24537 }, { "epoch": 2.92, "learning_rate": 8.805546660477782e-09, "loss": 0.4081, "step": 24538 }, { "epoch": 2.92, "learning_rate": 8.778613578591111e-09, "loss": 0.287, "step": 24539 }, { "epoch": 2.92, "learning_rate": 8.751721676803215e-09, "loss": 0.4279, "step": 24540 }, { "epoch": 2.92, "learning_rate": 8.724870955558185e-09, "loss": 0.4986, "step": 24541 }, { "epoch": 2.92, "learning_rate": 8.698061415300108e-09, "loss": 0.5703, "step": 24542 }, { "epoch": 2.92, "learning_rate": 8.67129305647224e-09, "loss": 0.6087, "step": 24543 }, { "epoch": 2.92, "learning_rate": 8.644565879517009e-09, "loss": 0.5112, "step": 24544 }, { "epoch": 2.92, "learning_rate": 8.617879884875723e-09, "loss": 0.3653, "step": 24545 }, { "epoch": 2.92, "learning_rate": 8.591235072990256e-09, "loss": 0.4562, "step": 24546 }, { "epoch": 2.92, "learning_rate": 8.564631444300808e-09, "loss": 0.4369, "step": 24547 }, { "epoch": 2.92, "learning_rate": 8.538068999246752e-09, "loss": 0.4536, "step": 24548 }, { "epoch": 2.92, "learning_rate": 8.511547738267734e-09, "loss": 0.6327, "step": 24549 }, { "epoch": 2.92, "learning_rate": 8.485067661801739e-09, "loss": 0.6143, "step": 24550 }, { "epoch": 2.92, "learning_rate": 8.458628770286748e-09, "loss": 0.3191, "step": 24551 }, { "epoch": 2.92, "learning_rate": 8.432231064159635e-09, "loss": 0.4092, "step": 24552 }, { "epoch": 2.92, "learning_rate": 8.405874543856718e-09, "loss": 0.4154, "step": 24553 }, { "epoch": 2.92, "learning_rate": 8.37955920981376e-09, "loss": 0.4481, "step": 24554 }, { "epoch": 2.92, "learning_rate": 8.353285062465687e-09, "loss": 0.5881, "step": 24555 }, { "epoch": 2.92, "learning_rate": 8.327052102247158e-09, "loss": 0.5913, "step": 24556 }, { "epoch": 2.92, "learning_rate": 8.300860329591431e-09, "loss": 0.3248, "step": 24557 }, { "epoch": 2.92, "learning_rate": 8.274709744931498e-09, "loss": 0.4104, "step": 24558 }, { "epoch": 2.92, "learning_rate": 8.248600348699509e-09, "loss": 0.4021, "step": 24559 }, { "epoch": 2.92, "learning_rate": 8.222532141327345e-09, "loss": 0.4308, "step": 24560 }, { "epoch": 2.92, "learning_rate": 8.196505123246045e-09, "loss": 0.5762, "step": 24561 }, { "epoch": 2.93, "learning_rate": 8.17051929488527e-09, "loss": 0.6288, "step": 24562 }, { "epoch": 2.93, "learning_rate": 8.144574656674953e-09, "loss": 0.3026, "step": 24563 }, { "epoch": 2.93, "learning_rate": 8.118671209043916e-09, "loss": 0.289, "step": 24564 }, { "epoch": 2.93, "learning_rate": 8.092808952420428e-09, "loss": 0.3335, "step": 24565 }, { "epoch": 2.93, "learning_rate": 8.066987887231924e-09, "loss": 0.4174, "step": 24566 }, { "epoch": 2.93, "learning_rate": 8.041208013905289e-09, "loss": 0.4833, "step": 24567 }, { "epoch": 2.93, "learning_rate": 8.015469332866288e-09, "loss": 0.6731, "step": 24568 }, { "epoch": 2.93, "learning_rate": 7.989771844540972e-09, "loss": 0.5659, "step": 24569 }, { "epoch": 2.93, "learning_rate": 7.964115549353447e-09, "loss": 0.3357, "step": 24570 }, { "epoch": 2.93, "learning_rate": 7.93850044772837e-09, "loss": 0.4502, "step": 24571 }, { "epoch": 2.93, "learning_rate": 7.912926540089294e-09, "loss": 0.498, "step": 24572 }, { "epoch": 2.93, "learning_rate": 7.88739382685838e-09, "loss": 0.4954, "step": 24573 }, { "epoch": 2.93, "learning_rate": 7.861902308457791e-09, "loss": 0.5686, "step": 24574 }, { "epoch": 2.93, "learning_rate": 7.836451985309135e-09, "loss": 0.6189, "step": 24575 }, { "epoch": 2.93, "learning_rate": 7.811042857833185e-09, "loss": 0.314, "step": 24576 }, { "epoch": 2.93, "learning_rate": 7.785674926449883e-09, "loss": 0.3266, "step": 24577 }, { "epoch": 2.93, "learning_rate": 7.76034819157806e-09, "loss": 0.4104, "step": 24578 }, { "epoch": 2.93, "learning_rate": 7.735062653637104e-09, "loss": 0.5092, "step": 24579 }, { "epoch": 2.93, "learning_rate": 7.709818313044737e-09, "loss": 0.6591, "step": 24580 }, { "epoch": 2.93, "learning_rate": 7.684615170218123e-09, "loss": 0.6141, "step": 24581 }, { "epoch": 2.93, "learning_rate": 7.659453225573877e-09, "loss": 0.3842, "step": 24582 }, { "epoch": 2.93, "learning_rate": 7.63433247952805e-09, "loss": 0.4092, "step": 24583 }, { "epoch": 2.93, "learning_rate": 7.60925293249587e-09, "loss": 0.4323, "step": 24584 }, { "epoch": 2.93, "learning_rate": 7.584214584892002e-09, "loss": 0.4662, "step": 24585 }, { "epoch": 2.93, "learning_rate": 7.559217437129729e-09, "loss": 0.4627, "step": 24586 }, { "epoch": 2.93, "learning_rate": 7.534261489623162e-09, "loss": 0.6609, "step": 24587 }, { "epoch": 2.93, "learning_rate": 7.509346742784474e-09, "loss": 0.3312, "step": 24588 }, { "epoch": 2.93, "learning_rate": 7.484473197025e-09, "loss": 0.3267, "step": 24589 }, { "epoch": 2.93, "learning_rate": 7.459640852756633e-09, "loss": 0.3315, "step": 24590 }, { "epoch": 2.93, "learning_rate": 7.434849710389325e-09, "loss": 0.5063, "step": 24591 }, { "epoch": 2.93, "learning_rate": 7.410099770333301e-09, "loss": 0.6417, "step": 24592 }, { "epoch": 2.93, "learning_rate": 7.385391032997125e-09, "loss": 0.5472, "step": 24593 }, { "epoch": 2.93, "learning_rate": 7.360723498789913e-09, "loss": 0.5606, "step": 24594 }, { "epoch": 2.93, "learning_rate": 7.3360971681191165e-09, "loss": 0.3192, "step": 24595 }, { "epoch": 2.93, "learning_rate": 7.311512041391633e-09, "loss": 0.4593, "step": 24596 }, { "epoch": 2.93, "learning_rate": 7.286968119014081e-09, "loss": 0.3971, "step": 24597 }, { "epoch": 2.93, "learning_rate": 7.2624654013919715e-09, "loss": 0.4878, "step": 24598 }, { "epoch": 2.93, "learning_rate": 7.238003888930811e-09, "loss": 0.5978, "step": 24599 }, { "epoch": 2.93, "learning_rate": 7.213583582034167e-09, "loss": 0.5988, "step": 24600 }, { "epoch": 2.93, "learning_rate": 7.189204481106438e-09, "loss": 0.2747, "step": 24601 }, { "epoch": 2.93, "learning_rate": 7.1648665865500784e-09, "loss": 0.4029, "step": 24602 }, { "epoch": 2.93, "learning_rate": 7.140569898768102e-09, "loss": 0.364, "step": 24603 }, { "epoch": 2.93, "learning_rate": 7.116314418161296e-09, "loss": 0.4103, "step": 24604 }, { "epoch": 2.93, "learning_rate": 7.092100145131009e-09, "loss": 0.5003, "step": 24605 }, { "epoch": 2.93, "learning_rate": 7.067927080077475e-09, "loss": 0.5601, "step": 24606 }, { "epoch": 2.93, "learning_rate": 7.043795223400374e-09, "loss": 0.3791, "step": 24607 }, { "epoch": 2.93, "learning_rate": 7.019704575498276e-09, "loss": 0.4258, "step": 24608 }, { "epoch": 2.93, "learning_rate": 6.99565513677003e-09, "loss": 0.3851, "step": 24609 }, { "epoch": 2.93, "learning_rate": 6.971646907612539e-09, "loss": 0.5161, "step": 24610 }, { "epoch": 2.93, "learning_rate": 6.947679888422987e-09, "loss": 0.5381, "step": 24611 }, { "epoch": 2.93, "learning_rate": 6.9237540795974445e-09, "loss": 0.5513, "step": 24612 }, { "epoch": 2.93, "learning_rate": 6.89986948153143e-09, "loss": 0.3296, "step": 24613 }, { "epoch": 2.93, "learning_rate": 6.876026094619626e-09, "loss": 0.4201, "step": 24614 }, { "epoch": 2.93, "learning_rate": 6.85222391925644e-09, "loss": 0.4938, "step": 24615 }, { "epoch": 2.93, "learning_rate": 6.828462955835169e-09, "loss": 0.6406, "step": 24616 }, { "epoch": 2.93, "learning_rate": 6.804743204748832e-09, "loss": 0.5677, "step": 24617 }, { "epoch": 2.93, "learning_rate": 6.781064666389059e-09, "loss": 0.5584, "step": 24618 }, { "epoch": 2.93, "learning_rate": 6.757427341147482e-09, "loss": 0.4573, "step": 24619 }, { "epoch": 2.93, "learning_rate": 6.733831229414622e-09, "loss": 0.3606, "step": 24620 }, { "epoch": 2.93, "learning_rate": 6.710276331581e-09, "loss": 0.356, "step": 24621 }, { "epoch": 2.93, "learning_rate": 6.68676264803575e-09, "loss": 0.3392, "step": 24622 }, { "epoch": 2.93, "learning_rate": 6.663290179167448e-09, "loss": 0.483, "step": 24623 }, { "epoch": 2.93, "learning_rate": 6.639858925364118e-09, "loss": 0.5557, "step": 24624 }, { "epoch": 2.93, "learning_rate": 6.616468887013228e-09, "loss": 0.5459, "step": 24625 }, { "epoch": 2.93, "learning_rate": 6.593120064501135e-09, "loss": 0.2683, "step": 24626 }, { "epoch": 2.93, "learning_rate": 6.569812458214197e-09, "loss": 0.385, "step": 24627 }, { "epoch": 2.93, "learning_rate": 6.5465460685373846e-09, "loss": 0.4023, "step": 24628 }, { "epoch": 2.93, "learning_rate": 6.52332089585539e-09, "loss": 0.4121, "step": 24629 }, { "epoch": 2.93, "learning_rate": 6.5001369405520705e-09, "loss": 0.6385, "step": 24630 }, { "epoch": 2.93, "learning_rate": 6.476994203010734e-09, "loss": 0.5852, "step": 24631 }, { "epoch": 2.93, "learning_rate": 6.453892683613849e-09, "loss": 0.3491, "step": 24632 }, { "epoch": 2.93, "learning_rate": 6.430832382743612e-09, "loss": 0.3479, "step": 24633 }, { "epoch": 2.93, "learning_rate": 6.40781330078083e-09, "loss": 0.4531, "step": 24634 }, { "epoch": 2.93, "learning_rate": 6.38483543810603e-09, "loss": 0.4333, "step": 24635 }, { "epoch": 2.93, "learning_rate": 6.361898795099186e-09, "loss": 0.4695, "step": 24636 }, { "epoch": 2.93, "learning_rate": 6.339003372139163e-09, "loss": 0.4858, "step": 24637 }, { "epoch": 2.93, "learning_rate": 6.316149169604824e-09, "loss": 0.3537, "step": 24638 }, { "epoch": 2.93, "learning_rate": 6.293336187873922e-09, "loss": 0.2785, "step": 24639 }, { "epoch": 2.93, "learning_rate": 6.2705644273233775e-09, "loss": 0.3319, "step": 24640 }, { "epoch": 2.93, "learning_rate": 6.24783388832928e-09, "loss": 0.463, "step": 24641 }, { "epoch": 2.93, "learning_rate": 6.225144571267994e-09, "loss": 0.5056, "step": 24642 }, { "epoch": 2.93, "learning_rate": 6.202496476514497e-09, "loss": 0.6841, "step": 24643 }, { "epoch": 2.93, "learning_rate": 6.179889604442657e-09, "loss": 0.5116, "step": 24644 }, { "epoch": 2.93, "learning_rate": 6.157323955426619e-09, "loss": 0.3257, "step": 24645 }, { "epoch": 2.94, "learning_rate": 6.1347995298394196e-09, "loss": 0.4166, "step": 24646 }, { "epoch": 2.94, "learning_rate": 6.1123163280529804e-09, "loss": 0.5168, "step": 24647 }, { "epoch": 2.94, "learning_rate": 6.089874350439507e-09, "loss": 0.5173, "step": 24648 }, { "epoch": 2.94, "learning_rate": 6.067473597369256e-09, "loss": 0.5587, "step": 24649 }, { "epoch": 2.94, "learning_rate": 6.04511406921332e-09, "loss": 0.4967, "step": 24650 }, { "epoch": 2.94, "learning_rate": 6.022795766340572e-09, "loss": 0.3587, "step": 24651 }, { "epoch": 2.94, "learning_rate": 6.000518689120715e-09, "loss": 0.3809, "step": 24652 }, { "epoch": 2.94, "learning_rate": 5.978282837921235e-09, "loss": 0.505, "step": 24653 }, { "epoch": 2.94, "learning_rate": 5.956088213109889e-09, "loss": 0.4189, "step": 24654 }, { "epoch": 2.94, "learning_rate": 5.9339348150538876e-09, "loss": 0.554, "step": 24655 }, { "epoch": 2.94, "learning_rate": 5.911822644119048e-09, "loss": 0.7261, "step": 24656 }, { "epoch": 2.94, "learning_rate": 5.889751700671187e-09, "loss": 0.2899, "step": 24657 }, { "epoch": 2.94, "learning_rate": 5.867721985075015e-09, "loss": 0.461, "step": 24658 }, { "epoch": 2.94, "learning_rate": 5.845733497694683e-09, "loss": 0.4039, "step": 24659 }, { "epoch": 2.94, "learning_rate": 5.823786238893514e-09, "loss": 0.4171, "step": 24660 }, { "epoch": 2.94, "learning_rate": 5.80188020903455e-09, "loss": 0.5849, "step": 24661 }, { "epoch": 2.94, "learning_rate": 5.780015408479722e-09, "loss": 0.5955, "step": 24662 }, { "epoch": 2.94, "learning_rate": 5.758191837590688e-09, "loss": 0.4677, "step": 24663 }, { "epoch": 2.94, "learning_rate": 5.73640949672799e-09, "loss": 0.3271, "step": 24664 }, { "epoch": 2.94, "learning_rate": 5.714668386251343e-09, "loss": 0.3405, "step": 24665 }, { "epoch": 2.94, "learning_rate": 5.692968506521013e-09, "loss": 0.3914, "step": 24666 }, { "epoch": 2.94, "learning_rate": 5.671309857895046e-09, "loss": 0.5377, "step": 24667 }, { "epoch": 2.94, "learning_rate": 5.649692440731491e-09, "loss": 0.5773, "step": 24668 }, { "epoch": 2.94, "learning_rate": 5.628116255387839e-09, "loss": 0.4489, "step": 24669 }, { "epoch": 2.94, "learning_rate": 5.606581302221026e-09, "loss": 0.2842, "step": 24670 }, { "epoch": 2.94, "learning_rate": 5.585087581586323e-09, "loss": 0.4518, "step": 24671 }, { "epoch": 2.94, "learning_rate": 5.563635093839559e-09, "loss": 0.4113, "step": 24672 }, { "epoch": 2.94, "learning_rate": 5.54222383933517e-09, "loss": 0.4697, "step": 24673 }, { "epoch": 2.94, "learning_rate": 5.520853818427318e-09, "loss": 0.5334, "step": 24674 }, { "epoch": 2.94, "learning_rate": 5.499525031468777e-09, "loss": 0.5456, "step": 24675 }, { "epoch": 2.94, "learning_rate": 5.478237478812598e-09, "loss": 0.2494, "step": 24676 }, { "epoch": 2.94, "learning_rate": 5.4569911608104434e-09, "loss": 0.3484, "step": 24677 }, { "epoch": 2.94, "learning_rate": 5.435786077813421e-09, "loss": 0.4024, "step": 24678 }, { "epoch": 2.94, "learning_rate": 5.414622230172084e-09, "loss": 0.5009, "step": 24679 }, { "epoch": 2.94, "learning_rate": 5.393499618236708e-09, "loss": 0.5817, "step": 24680 }, { "epoch": 2.94, "learning_rate": 5.372418242355903e-09, "loss": 0.6324, "step": 24681 }, { "epoch": 2.94, "learning_rate": 5.35137810287828e-09, "loss": 0.3951, "step": 24682 }, { "epoch": 2.94, "learning_rate": 5.3303792001516145e-09, "loss": 0.4557, "step": 24683 }, { "epoch": 2.94, "learning_rate": 5.309421534523407e-09, "loss": 0.5128, "step": 24684 }, { "epoch": 2.94, "learning_rate": 5.288505106339492e-09, "loss": 0.4696, "step": 24685 }, { "epoch": 2.94, "learning_rate": 5.2676299159459824e-09, "loss": 0.5477, "step": 24686 }, { "epoch": 2.94, "learning_rate": 5.246795963687879e-09, "loss": 0.4707, "step": 24687 }, { "epoch": 2.94, "learning_rate": 5.226003249909628e-09, "loss": 0.3614, "step": 24688 }, { "epoch": 2.94, "learning_rate": 5.205251774954845e-09, "loss": 0.2913, "step": 24689 }, { "epoch": 2.94, "learning_rate": 5.184541539166588e-09, "loss": 0.5523, "step": 24690 }, { "epoch": 2.94, "learning_rate": 5.163872542887361e-09, "loss": 0.4645, "step": 24691 }, { "epoch": 2.94, "learning_rate": 5.143244786458279e-09, "loss": 0.5256, "step": 24692 }, { "epoch": 2.94, "learning_rate": 5.122658270221015e-09, "loss": 0.6732, "step": 24693 }, { "epoch": 2.94, "learning_rate": 5.102112994515296e-09, "loss": 0.6432, "step": 24694 }, { "epoch": 2.94, "learning_rate": 5.081608959681128e-09, "loss": 0.3153, "step": 24695 }, { "epoch": 2.94, "learning_rate": 5.061146166057407e-09, "loss": 0.404, "step": 24696 }, { "epoch": 2.94, "learning_rate": 5.0407246139821955e-09, "loss": 0.3925, "step": 24697 }, { "epoch": 2.94, "learning_rate": 5.020344303793001e-09, "loss": 0.4456, "step": 24698 }, { "epoch": 2.94, "learning_rate": 5.0000052358270544e-09, "loss": 0.5639, "step": 24699 }, { "epoch": 2.94, "learning_rate": 4.9797074104204756e-09, "loss": 0.5803, "step": 24700 }, { "epoch": 2.94, "learning_rate": 4.9594508279082745e-09, "loss": 0.3042, "step": 24701 }, { "epoch": 2.94, "learning_rate": 4.939235488626015e-09, "loss": 0.3398, "step": 24702 }, { "epoch": 2.94, "learning_rate": 4.919061392907598e-09, "loss": 0.398, "step": 24703 }, { "epoch": 2.94, "learning_rate": 4.898928541086367e-09, "loss": 0.4132, "step": 24704 }, { "epoch": 2.94, "learning_rate": 4.8788369334951126e-09, "loss": 0.5743, "step": 24705 }, { "epoch": 2.94, "learning_rate": 4.858786570466068e-09, "loss": 0.6215, "step": 24706 }, { "epoch": 2.94, "learning_rate": 4.838777452330634e-09, "loss": 0.3442, "step": 24707 }, { "epoch": 2.94, "learning_rate": 4.818809579419658e-09, "loss": 0.397, "step": 24708 }, { "epoch": 2.94, "learning_rate": 4.798882952062878e-09, "loss": 0.3732, "step": 24709 }, { "epoch": 2.94, "learning_rate": 4.778997570590305e-09, "loss": 0.4169, "step": 24710 }, { "epoch": 2.94, "learning_rate": 4.7591534353300105e-09, "loss": 0.4908, "step": 24711 }, { "epoch": 2.94, "learning_rate": 4.739350546610344e-09, "loss": 0.6232, "step": 24712 }, { "epoch": 2.94, "learning_rate": 4.719588904758543e-09, "loss": 0.4345, "step": 24713 }, { "epoch": 2.94, "learning_rate": 4.699868510101291e-09, "loss": 0.3846, "step": 24714 }, { "epoch": 2.94, "learning_rate": 4.680189362964716e-09, "loss": 0.399, "step": 24715 }, { "epoch": 2.94, "learning_rate": 4.6605514636738345e-09, "loss": 0.3331, "step": 24716 }, { "epoch": 2.94, "learning_rate": 4.6409548125536665e-09, "loss": 0.4729, "step": 24717 }, { "epoch": 2.94, "learning_rate": 4.621399409927562e-09, "loss": 0.5492, "step": 24718 }, { "epoch": 2.94, "learning_rate": 4.601885256119432e-09, "loss": 0.5024, "step": 24719 }, { "epoch": 2.94, "learning_rate": 4.582412351451237e-09, "loss": 0.3091, "step": 24720 }, { "epoch": 2.94, "learning_rate": 4.562980696245223e-09, "loss": 0.379, "step": 24721 }, { "epoch": 2.94, "learning_rate": 4.54359029082252e-09, "loss": 0.395, "step": 24722 }, { "epoch": 2.94, "learning_rate": 4.524241135503704e-09, "loss": 0.5803, "step": 24723 }, { "epoch": 2.94, "learning_rate": 4.504933230608521e-09, "loss": 0.5096, "step": 24724 }, { "epoch": 2.94, "learning_rate": 4.485666576456438e-09, "loss": 0.593, "step": 24725 }, { "epoch": 2.94, "learning_rate": 4.466441173365254e-09, "loss": 0.2515, "step": 24726 }, { "epoch": 2.94, "learning_rate": 4.447257021653606e-09, "loss": 0.4047, "step": 24727 }, { "epoch": 2.94, "learning_rate": 4.428114121637905e-09, "loss": 0.4321, "step": 24728 }, { "epoch": 2.94, "learning_rate": 4.409012473634844e-09, "loss": 0.4917, "step": 24729 }, { "epoch": 2.95, "learning_rate": 4.38995207796028e-09, "loss": 0.5871, "step": 24730 }, { "epoch": 2.95, "learning_rate": 4.37093293492924e-09, "loss": 0.6691, "step": 24731 }, { "epoch": 2.95, "learning_rate": 4.3519550448561935e-09, "loss": 0.2634, "step": 24732 }, { "epoch": 2.95, "learning_rate": 4.3330184080545015e-09, "loss": 0.3988, "step": 24733 }, { "epoch": 2.95, "learning_rate": 4.314123024837524e-09, "loss": 0.3371, "step": 24734 }, { "epoch": 2.95, "learning_rate": 4.295268895517513e-09, "loss": 0.4086, "step": 24735 }, { "epoch": 2.95, "learning_rate": 4.276456020406161e-09, "loss": 0.5734, "step": 24736 }, { "epoch": 2.95, "learning_rate": 4.257684399814332e-09, "loss": 0.5992, "step": 24737 }, { "epoch": 2.95, "learning_rate": 4.2389540340526114e-09, "loss": 0.4307, "step": 24738 }, { "epoch": 2.95, "learning_rate": 4.220264923430195e-09, "loss": 0.2676, "step": 24739 }, { "epoch": 2.95, "learning_rate": 4.201617068256003e-09, "loss": 0.4611, "step": 24740 }, { "epoch": 2.95, "learning_rate": 4.183010468838677e-09, "loss": 0.3834, "step": 24741 }, { "epoch": 2.95, "learning_rate": 4.1644451254857495e-09, "loss": 0.5124, "step": 24742 }, { "epoch": 2.95, "learning_rate": 4.145921038503642e-09, "loss": 0.5896, "step": 24743 }, { "epoch": 2.95, "learning_rate": 4.1274382081993304e-09, "loss": 0.5474, "step": 24744 }, { "epoch": 2.95, "learning_rate": 4.108996634877571e-09, "loss": 0.3238, "step": 24745 }, { "epoch": 2.95, "learning_rate": 4.090596318843676e-09, "loss": 0.4367, "step": 24746 }, { "epoch": 2.95, "learning_rate": 4.07223726040129e-09, "loss": 0.4236, "step": 24747 }, { "epoch": 2.95, "learning_rate": 4.0539194598546165e-09, "loss": 0.4316, "step": 24748 }, { "epoch": 2.95, "learning_rate": 4.035642917505911e-09, "loss": 0.6544, "step": 24749 }, { "epoch": 2.95, "learning_rate": 4.017407633657711e-09, "loss": 0.6599, "step": 24750 }, { "epoch": 2.95, "learning_rate": 3.999213608610886e-09, "loss": 0.2601, "step": 24751 }, { "epoch": 2.95, "learning_rate": 3.9810608426665845e-09, "loss": 0.4633, "step": 24752 }, { "epoch": 2.95, "learning_rate": 3.9629493361248435e-09, "loss": 0.4214, "step": 24753 }, { "epoch": 2.95, "learning_rate": 3.94487908928487e-09, "loss": 0.4808, "step": 24754 }, { "epoch": 2.95, "learning_rate": 3.92685010244559e-09, "loss": 0.5356, "step": 24755 }, { "epoch": 2.95, "learning_rate": 3.908862375904821e-09, "loss": 0.6713, "step": 24756 }, { "epoch": 2.95, "learning_rate": 3.890915909960102e-09, "loss": 0.2879, "step": 24757 }, { "epoch": 2.95, "learning_rate": 3.873010704907865e-09, "loss": 0.2352, "step": 24758 }, { "epoch": 2.95, "learning_rate": 3.855146761044259e-09, "loss": 0.4791, "step": 24759 }, { "epoch": 2.95, "learning_rate": 3.837324078664606e-09, "loss": 0.4209, "step": 24760 }, { "epoch": 2.95, "learning_rate": 3.819542658063391e-09, "loss": 0.6512, "step": 24761 }, { "epoch": 2.95, "learning_rate": 3.801802499534546e-09, "loss": 0.5257, "step": 24762 }, { "epoch": 2.95, "learning_rate": 3.784103603371448e-09, "loss": 0.4536, "step": 24763 }, { "epoch": 2.95, "learning_rate": 3.7664459698663615e-09, "loss": 0.3345, "step": 24764 }, { "epoch": 2.95, "learning_rate": 3.7488295993115546e-09, "loss": 0.4764, "step": 24765 }, { "epoch": 2.95, "learning_rate": 3.731254491997904e-09, "loss": 0.3899, "step": 24766 }, { "epoch": 2.95, "learning_rate": 3.71372064821629e-09, "loss": 0.5515, "step": 24767 }, { "epoch": 2.95, "learning_rate": 3.6962280682562024e-09, "loss": 0.6163, "step": 24768 }, { "epoch": 2.95, "learning_rate": 3.6787767524068542e-09, "loss": 0.6018, "step": 24769 }, { "epoch": 2.95, "learning_rate": 3.661366700956903e-09, "loss": 0.3026, "step": 24770 }, { "epoch": 2.95, "learning_rate": 3.6439979141938973e-09, "loss": 0.3848, "step": 24771 }, { "epoch": 2.95, "learning_rate": 3.626670392405107e-09, "loss": 0.4427, "step": 24772 }, { "epoch": 2.95, "learning_rate": 3.60938413587697e-09, "loss": 0.4615, "step": 24773 }, { "epoch": 2.95, "learning_rate": 3.5921391448950905e-09, "loss": 0.5991, "step": 24774 }, { "epoch": 2.95, "learning_rate": 3.574935419744796e-09, "loss": 0.6591, "step": 24775 }, { "epoch": 2.95, "learning_rate": 3.5577729607100264e-09, "loss": 0.4111, "step": 24776 }, { "epoch": 2.95, "learning_rate": 3.5406517680749985e-09, "loss": 0.3134, "step": 24777 }, { "epoch": 2.95, "learning_rate": 3.523571842122264e-09, "loss": 0.4441, "step": 24778 }, { "epoch": 2.95, "learning_rate": 3.5065331831346526e-09, "loss": 0.4315, "step": 24779 }, { "epoch": 2.95, "learning_rate": 3.4895357913930507e-09, "loss": 0.5684, "step": 24780 }, { "epoch": 2.95, "learning_rate": 3.472579667179177e-09, "loss": 0.5569, "step": 24781 }, { "epoch": 2.95, "learning_rate": 3.455664810773085e-09, "loss": 0.3934, "step": 24782 }, { "epoch": 2.95, "learning_rate": 3.438791222454274e-09, "loss": 0.4468, "step": 24783 }, { "epoch": 2.95, "learning_rate": 3.4219589025016875e-09, "loss": 0.4415, "step": 24784 }, { "epoch": 2.95, "learning_rate": 3.4051678511937137e-09, "loss": 0.4151, "step": 24785 }, { "epoch": 2.95, "learning_rate": 3.3884180688076308e-09, "loss": 0.6515, "step": 24786 }, { "epoch": 2.95, "learning_rate": 3.371709555620717e-09, "loss": 0.6015, "step": 24787 }, { "epoch": 2.95, "learning_rate": 3.355042311908585e-09, "loss": 0.2442, "step": 24788 }, { "epoch": 2.95, "learning_rate": 3.3384163379474034e-09, "loss": 0.3443, "step": 24789 }, { "epoch": 2.95, "learning_rate": 3.3218316340113966e-09, "loss": 0.4648, "step": 24790 }, { "epoch": 2.95, "learning_rate": 3.305288200375345e-09, "loss": 0.36, "step": 24791 }, { "epoch": 2.95, "learning_rate": 3.288786037312086e-09, "loss": 0.6014, "step": 24792 }, { "epoch": 2.95, "learning_rate": 3.2723251450947347e-09, "loss": 0.4969, "step": 24793 }, { "epoch": 2.95, "learning_rate": 3.2559055239955728e-09, "loss": 0.6646, "step": 24794 }, { "epoch": 2.95, "learning_rate": 3.239527174285495e-09, "loss": 0.3043, "step": 24795 }, { "epoch": 2.95, "learning_rate": 3.2231900962356734e-09, "loss": 0.4293, "step": 24796 }, { "epoch": 2.95, "learning_rate": 3.206894290115892e-09, "loss": 0.4152, "step": 24797 }, { "epoch": 2.95, "learning_rate": 3.1906397561959347e-09, "loss": 0.4487, "step": 24798 }, { "epoch": 2.95, "learning_rate": 3.1744264947439206e-09, "loss": 0.594, "step": 24799 }, { "epoch": 2.95, "learning_rate": 3.1582545060282468e-09, "loss": 0.5435, "step": 24800 }, { "epoch": 2.95, "learning_rate": 3.142123790316198e-09, "loss": 0.3649, "step": 24801 }, { "epoch": 2.95, "learning_rate": 3.1260343478742293e-09, "loss": 0.3872, "step": 24802 }, { "epoch": 2.95, "learning_rate": 3.1099861789682385e-09, "loss": 0.3955, "step": 24803 }, { "epoch": 2.95, "learning_rate": 3.093979283863846e-09, "loss": 0.3734, "step": 24804 }, { "epoch": 2.95, "learning_rate": 3.078013662825563e-09, "loss": 0.5743, "step": 24805 }, { "epoch": 2.95, "learning_rate": 3.062089316117345e-09, "loss": 0.5501, "step": 24806 }, { "epoch": 2.95, "learning_rate": 3.046206244002037e-09, "loss": 0.4365, "step": 24807 }, { "epoch": 2.95, "learning_rate": 3.030364446742484e-09, "loss": 0.2555, "step": 24808 }, { "epoch": 2.95, "learning_rate": 3.0145639246006996e-09, "loss": 0.4486, "step": 24809 }, { "epoch": 2.95, "learning_rate": 2.9988046778375855e-09, "loss": 0.4527, "step": 24810 }, { "epoch": 2.95, "learning_rate": 2.983086706713767e-09, "loss": 0.5323, "step": 24811 }, { "epoch": 2.95, "learning_rate": 2.967410011489036e-09, "loss": 0.6312, "step": 24812 }, { "epoch": 2.95, "learning_rate": 2.9517745924223517e-09, "loss": 0.4064, "step": 24813 }, { "epoch": 2.96, "learning_rate": 2.936180449772674e-09, "loss": 0.3329, "step": 24814 }, { "epoch": 2.96, "learning_rate": 2.9206275837972976e-09, "loss": 0.4742, "step": 24815 }, { "epoch": 2.96, "learning_rate": 2.9051159947535155e-09, "loss": 0.4185, "step": 24816 }, { "epoch": 2.96, "learning_rate": 2.889645682897513e-09, "loss": 0.475, "step": 24817 }, { "epoch": 2.96, "learning_rate": 2.8742166484854726e-09, "loss": 0.6674, "step": 24818 }, { "epoch": 2.96, "learning_rate": 2.8588288917719143e-09, "loss": 0.4482, "step": 24819 }, { "epoch": 2.96, "learning_rate": 2.843482413011356e-09, "loss": 0.274, "step": 24820 }, { "epoch": 2.96, "learning_rate": 2.828177212457761e-09, "loss": 0.3409, "step": 24821 }, { "epoch": 2.96, "learning_rate": 2.8129132903639834e-09, "loss": 0.3913, "step": 24822 }, { "epoch": 2.96, "learning_rate": 2.7976906469823205e-09, "loss": 0.5052, "step": 24823 }, { "epoch": 2.96, "learning_rate": 2.7825092825642386e-09, "loss": 0.6066, "step": 24824 }, { "epoch": 2.96, "learning_rate": 2.7673691973606477e-09, "loss": 0.5549, "step": 24825 }, { "epoch": 2.96, "learning_rate": 2.7522703916221805e-09, "loss": 0.3513, "step": 24826 }, { "epoch": 2.96, "learning_rate": 2.7372128655980823e-09, "loss": 0.4513, "step": 24827 }, { "epoch": 2.96, "learning_rate": 2.722196619537598e-09, "loss": 0.4134, "step": 24828 }, { "epoch": 2.96, "learning_rate": 2.7072216536885855e-09, "loss": 0.4753, "step": 24829 }, { "epoch": 2.96, "learning_rate": 2.6922879682989013e-09, "loss": 0.6333, "step": 24830 }, { "epoch": 2.96, "learning_rate": 2.677395563615015e-09, "loss": 0.6306, "step": 24831 }, { "epoch": 2.96, "learning_rate": 2.662544439883674e-09, "loss": 0.3846, "step": 24832 }, { "epoch": 2.96, "learning_rate": 2.6477345973496824e-09, "loss": 0.3558, "step": 24833 }, { "epoch": 2.96, "learning_rate": 2.6329660362584e-09, "loss": 0.3849, "step": 24834 }, { "epoch": 2.96, "learning_rate": 2.6182387568535196e-09, "loss": 0.4877, "step": 24835 }, { "epoch": 2.96, "learning_rate": 2.6035527593787357e-09, "loss": 0.5943, "step": 24836 }, { "epoch": 2.96, "learning_rate": 2.58890804407691e-09, "loss": 0.561, "step": 24837 }, { "epoch": 2.96, "learning_rate": 2.5743046111897928e-09, "loss": 0.2889, "step": 24838 }, { "epoch": 2.96, "learning_rate": 2.559742460959136e-09, "loss": 0.3559, "step": 24839 }, { "epoch": 2.96, "learning_rate": 2.5452215936255796e-09, "loss": 0.3692, "step": 24840 }, { "epoch": 2.96, "learning_rate": 2.5307420094289327e-09, "loss": 0.4429, "step": 24841 }, { "epoch": 2.96, "learning_rate": 2.5163037086084476e-09, "loss": 0.4898, "step": 24842 }, { "epoch": 2.96, "learning_rate": 2.5019066914033773e-09, "loss": 0.4752, "step": 24843 }, { "epoch": 2.96, "learning_rate": 2.48755095805131e-09, "loss": 0.6671, "step": 24844 }, { "epoch": 2.96, "learning_rate": 2.473236508789556e-09, "loss": 0.308, "step": 24845 }, { "epoch": 2.96, "learning_rate": 2.4589633438545925e-09, "loss": 0.4324, "step": 24846 }, { "epoch": 2.96, "learning_rate": 2.444731463482619e-09, "loss": 0.4108, "step": 24847 }, { "epoch": 2.96, "learning_rate": 2.4305408679087262e-09, "loss": 0.4977, "step": 24848 }, { "epoch": 2.96, "learning_rate": 2.4163915573677253e-09, "loss": 0.4998, "step": 24849 }, { "epoch": 2.96, "learning_rate": 2.4022835320930415e-09, "loss": 0.6612, "step": 24850 }, { "epoch": 2.96, "learning_rate": 2.388216792318654e-09, "loss": 0.3128, "step": 24851 }, { "epoch": 2.96, "learning_rate": 2.3741913382763217e-09, "loss": 0.353, "step": 24852 }, { "epoch": 2.96, "learning_rate": 2.3602071701980813e-09, "loss": 0.3684, "step": 24853 }, { "epoch": 2.96, "learning_rate": 2.346264288315414e-09, "loss": 0.4496, "step": 24854 }, { "epoch": 2.96, "learning_rate": 2.332362692858414e-09, "loss": 0.5606, "step": 24855 }, { "epoch": 2.96, "learning_rate": 2.3185023840568975e-09, "loss": 0.5825, "step": 24856 }, { "epoch": 2.96, "learning_rate": 2.304683362140403e-09, "loss": 0.2661, "step": 24857 }, { "epoch": 2.96, "learning_rate": 2.2909056273368036e-09, "loss": 0.4021, "step": 24858 }, { "epoch": 2.96, "learning_rate": 2.277169179874528e-09, "loss": 0.3799, "step": 24859 }, { "epoch": 2.96, "learning_rate": 2.2634740199797833e-09, "loss": 0.5132, "step": 24860 }, { "epoch": 2.96, "learning_rate": 2.249820147879611e-09, "loss": 0.6062, "step": 24861 }, { "epoch": 2.96, "learning_rate": 2.236207563799386e-09, "loss": 0.5864, "step": 24862 }, { "epoch": 2.96, "learning_rate": 2.2226362679642066e-09, "loss": 0.3796, "step": 24863 }, { "epoch": 2.96, "learning_rate": 2.209106260598615e-09, "loss": 0.3213, "step": 24864 }, { "epoch": 2.96, "learning_rate": 2.195617541925765e-09, "loss": 0.3945, "step": 24865 }, { "epoch": 2.96, "learning_rate": 2.182170112169091e-09, "loss": 0.3777, "step": 24866 }, { "epoch": 2.96, "learning_rate": 2.1687639715509136e-09, "loss": 0.4323, "step": 24867 }, { "epoch": 2.96, "learning_rate": 2.155399120292445e-09, "loss": 0.5775, "step": 24868 }, { "epoch": 2.96, "learning_rate": 2.142075558614898e-09, "loss": 0.4999, "step": 24869 }, { "epoch": 2.96, "learning_rate": 2.128793286738373e-09, "loss": 0.3475, "step": 24870 }, { "epoch": 2.96, "learning_rate": 2.1155523048826955e-09, "loss": 0.5702, "step": 24871 }, { "epoch": 2.96, "learning_rate": 2.1023526132663006e-09, "loss": 0.4089, "step": 24872 }, { "epoch": 2.96, "learning_rate": 2.089194212107626e-09, "loss": 0.4952, "step": 24873 }, { "epoch": 2.96, "learning_rate": 2.0760771016242742e-09, "loss": 0.6056, "step": 24874 }, { "epoch": 2.96, "learning_rate": 2.0630012820327396e-09, "loss": 0.6208, "step": 24875 }, { "epoch": 2.96, "learning_rate": 2.049966753549515e-09, "loss": 0.418, "step": 24876 }, { "epoch": 2.96, "learning_rate": 2.036973516389984e-09, "loss": 0.3644, "step": 24877 }, { "epoch": 2.96, "learning_rate": 2.0240215707689748e-09, "loss": 0.4277, "step": 24878 }, { "epoch": 2.96, "learning_rate": 2.0111109169002052e-09, "loss": 0.4706, "step": 24879 }, { "epoch": 2.96, "learning_rate": 1.998241554997671e-09, "loss": 0.5247, "step": 24880 }, { "epoch": 2.96, "learning_rate": 1.985413485273702e-09, "loss": 0.8065, "step": 24881 }, { "epoch": 2.96, "learning_rate": 1.9726267079403505e-09, "loss": 0.2444, "step": 24882 }, { "epoch": 2.96, "learning_rate": 1.9598812232091147e-09, "loss": 0.4127, "step": 24883 }, { "epoch": 2.96, "learning_rate": 1.947177031290659e-09, "loss": 0.4579, "step": 24884 }, { "epoch": 2.96, "learning_rate": 1.9345141323950934e-09, "loss": 0.4015, "step": 24885 }, { "epoch": 2.96, "learning_rate": 1.921892526731417e-09, "loss": 0.5461, "step": 24886 }, { "epoch": 2.96, "learning_rate": 1.9093122145086294e-09, "loss": 0.6361, "step": 24887 }, { "epoch": 2.96, "learning_rate": 1.896773195934343e-09, "loss": 0.3143, "step": 24888 }, { "epoch": 2.96, "learning_rate": 1.884275471216168e-09, "loss": 0.3454, "step": 24889 }, { "epoch": 2.96, "learning_rate": 1.8718190405603298e-09, "loss": 0.385, "step": 24890 }, { "epoch": 2.96, "learning_rate": 1.8594039041730516e-09, "loss": 0.435, "step": 24891 }, { "epoch": 2.96, "learning_rate": 1.8470300622594472e-09, "loss": 0.5447, "step": 24892 }, { "epoch": 2.96, "learning_rate": 1.8346975150240754e-09, "loss": 0.6066, "step": 24893 }, { "epoch": 2.96, "learning_rate": 1.8224062626706618e-09, "loss": 0.695, "step": 24894 }, { "epoch": 2.96, "learning_rate": 1.810156305402655e-09, "loss": 0.3694, "step": 24895 }, { "epoch": 2.96, "learning_rate": 1.797947643422393e-09, "loss": 0.4158, "step": 24896 }, { "epoch": 2.96, "learning_rate": 1.7857802769313814e-09, "loss": 0.4219, "step": 24897 }, { "epoch": 2.97, "learning_rate": 1.7736542061314034e-09, "loss": 0.433, "step": 24898 }, { "epoch": 2.97, "learning_rate": 1.7615694312222986e-09, "loss": 0.5779, "step": 24899 }, { "epoch": 2.97, "learning_rate": 1.7495259524041852e-09, "loss": 0.5984, "step": 24900 }, { "epoch": 2.97, "learning_rate": 1.7375237698757929e-09, "loss": 0.3145, "step": 24901 }, { "epoch": 2.97, "learning_rate": 1.7255628838361293e-09, "loss": 0.5645, "step": 24902 }, { "epoch": 2.97, "learning_rate": 1.713643294482259e-09, "loss": 0.4292, "step": 24903 }, { "epoch": 2.97, "learning_rate": 1.701765002011524e-09, "loss": 0.5215, "step": 24904 }, { "epoch": 2.97, "learning_rate": 1.6899280066204337e-09, "loss": 0.551, "step": 24905 }, { "epoch": 2.97, "learning_rate": 1.67813230850411e-09, "loss": 0.625, "step": 24906 }, { "epoch": 2.97, "learning_rate": 1.6663779078582298e-09, "loss": 0.4117, "step": 24907 }, { "epoch": 2.97, "learning_rate": 1.654664804876527e-09, "loss": 0.4473, "step": 24908 }, { "epoch": 2.97, "learning_rate": 1.6429929997530126e-09, "loss": 0.4986, "step": 24909 }, { "epoch": 2.97, "learning_rate": 1.6313624926803106e-09, "loss": 0.4565, "step": 24910 }, { "epoch": 2.97, "learning_rate": 1.619773283851045e-09, "loss": 0.5185, "step": 24911 }, { "epoch": 2.97, "learning_rate": 1.6082253734561736e-09, "loss": 0.5608, "step": 24912 }, { "epoch": 2.97, "learning_rate": 1.5967187616872103e-09, "loss": 0.4285, "step": 24913 }, { "epoch": 2.97, "learning_rate": 1.5852534487340031e-09, "loss": 0.4409, "step": 24914 }, { "epoch": 2.97, "learning_rate": 1.5738294347864003e-09, "loss": 0.4773, "step": 24915 }, { "epoch": 2.97, "learning_rate": 1.562446720032862e-09, "loss": 0.3874, "step": 24916 }, { "epoch": 2.97, "learning_rate": 1.5511053046618485e-09, "loss": 0.517, "step": 24917 }, { "epoch": 2.97, "learning_rate": 1.5398051888607102e-09, "loss": 0.4749, "step": 24918 }, { "epoch": 2.97, "learning_rate": 1.5285463728159644e-09, "loss": 0.5029, "step": 24919 }, { "epoch": 2.97, "learning_rate": 1.5173288567141287e-09, "loss": 0.3781, "step": 24920 }, { "epoch": 2.97, "learning_rate": 1.5061526407406102e-09, "loss": 0.2871, "step": 24921 }, { "epoch": 2.97, "learning_rate": 1.4950177250799836e-09, "loss": 0.4901, "step": 24922 }, { "epoch": 2.97, "learning_rate": 1.4839241099162683e-09, "loss": 0.4189, "step": 24923 }, { "epoch": 2.97, "learning_rate": 1.4728717954329285e-09, "loss": 0.5956, "step": 24924 }, { "epoch": 2.97, "learning_rate": 1.4618607818128738e-09, "loss": 0.6058, "step": 24925 }, { "epoch": 2.97, "learning_rate": 1.4508910692376254e-09, "loss": 0.3023, "step": 24926 }, { "epoch": 2.97, "learning_rate": 1.4399626578889825e-09, "loss": 0.3748, "step": 24927 }, { "epoch": 2.97, "learning_rate": 1.429075547947356e-09, "loss": 0.4418, "step": 24928 }, { "epoch": 2.97, "learning_rate": 1.4182297395926026e-09, "loss": 0.3991, "step": 24929 }, { "epoch": 2.97, "learning_rate": 1.4074252330043004e-09, "loss": 0.5421, "step": 24930 }, { "epoch": 2.97, "learning_rate": 1.3966620283609178e-09, "loss": 0.552, "step": 24931 }, { "epoch": 2.97, "learning_rate": 1.3859401258403683e-09, "loss": 0.4247, "step": 24932 }, { "epoch": 2.97, "learning_rate": 1.3752595256197321e-09, "loss": 0.3142, "step": 24933 }, { "epoch": 2.97, "learning_rate": 1.3646202278758126e-09, "loss": 0.4587, "step": 24934 }, { "epoch": 2.97, "learning_rate": 1.3540222327843022e-09, "loss": 0.4514, "step": 24935 }, { "epoch": 2.97, "learning_rate": 1.3434655405206164e-09, "loss": 0.5211, "step": 24936 }, { "epoch": 2.97, "learning_rate": 1.3329501512590604e-09, "loss": 0.643, "step": 24937 }, { "epoch": 2.97, "learning_rate": 1.3224760651733836e-09, "loss": 0.6209, "step": 24938 }, { "epoch": 2.97, "learning_rate": 1.3120432824370588e-09, "loss": 0.3861, "step": 24939 }, { "epoch": 2.97, "learning_rate": 1.3016518032221703e-09, "loss": 0.4585, "step": 24940 }, { "epoch": 2.97, "learning_rate": 1.2913016277005253e-09, "loss": 0.447, "step": 24941 }, { "epoch": 2.97, "learning_rate": 1.2809927560433755e-09, "loss": 0.5521, "step": 24942 }, { "epoch": 2.97, "learning_rate": 1.2707251884211402e-09, "loss": 0.5447, "step": 24943 }, { "epoch": 2.97, "learning_rate": 1.2604989250034061e-09, "loss": 0.5634, "step": 24944 }, { "epoch": 2.97, "learning_rate": 1.2503139659594821e-09, "loss": 0.3844, "step": 24945 }, { "epoch": 2.97, "learning_rate": 1.2401703114572894e-09, "loss": 0.3318, "step": 24946 }, { "epoch": 2.97, "learning_rate": 1.2300679616650268e-09, "loss": 0.4553, "step": 24947 }, { "epoch": 2.97, "learning_rate": 1.220006916749228e-09, "loss": 0.4995, "step": 24948 }, { "epoch": 2.97, "learning_rate": 1.209987176876426e-09, "loss": 0.5637, "step": 24949 }, { "epoch": 2.97, "learning_rate": 1.2000087422120444e-09, "loss": 0.4775, "step": 24950 }, { "epoch": 2.97, "learning_rate": 1.1900716129212286e-09, "loss": 0.3628, "step": 24951 }, { "epoch": 2.97, "learning_rate": 1.1801757891682918e-09, "loss": 0.4094, "step": 24952 }, { "epoch": 2.97, "learning_rate": 1.1703212711167145e-09, "loss": 0.4366, "step": 24953 }, { "epoch": 2.97, "learning_rate": 1.1605080589294215e-09, "loss": 0.5063, "step": 24954 }, { "epoch": 2.97, "learning_rate": 1.1507361527685058e-09, "loss": 0.6182, "step": 24955 }, { "epoch": 2.97, "learning_rate": 1.1410055527955044e-09, "loss": 0.579, "step": 24956 }, { "epoch": 2.97, "learning_rate": 1.1313162591716775e-09, "loss": 0.2661, "step": 24957 }, { "epoch": 2.97, "learning_rate": 1.1216682720566196e-09, "loss": 0.418, "step": 24958 }, { "epoch": 2.97, "learning_rate": 1.1120615916102028e-09, "loss": 0.4074, "step": 24959 }, { "epoch": 2.97, "learning_rate": 1.1024962179909117e-09, "loss": 0.4855, "step": 24960 }, { "epoch": 2.97, "learning_rate": 1.0929721513569524e-09, "loss": 0.5207, "step": 24961 }, { "epoch": 2.97, "learning_rate": 1.0834893918659773e-09, "loss": 0.6046, "step": 24962 }, { "epoch": 2.97, "learning_rate": 1.0740479396748049e-09, "loss": 0.3815, "step": 24963 }, { "epoch": 2.97, "learning_rate": 1.0646477949391444e-09, "loss": 0.2897, "step": 24964 }, { "epoch": 2.97, "learning_rate": 1.0552889578144266e-09, "loss": 0.3472, "step": 24965 }, { "epoch": 2.97, "learning_rate": 1.0459714284555279e-09, "loss": 0.3615, "step": 24966 }, { "epoch": 2.97, "learning_rate": 1.0366952070164916e-09, "loss": 0.5461, "step": 24967 }, { "epoch": 2.97, "learning_rate": 1.0274602936505285e-09, "loss": 0.5667, "step": 24968 }, { "epoch": 2.97, "learning_rate": 1.0182666885100167e-09, "loss": 0.6171, "step": 24969 }, { "epoch": 2.97, "learning_rate": 1.0091143917476121e-09, "loss": 0.3153, "step": 24970 }, { "epoch": 2.97, "learning_rate": 1.0000034035140271e-09, "loss": 0.4318, "step": 24971 }, { "epoch": 2.97, "learning_rate": 9.909337239602523e-10, "loss": 0.4008, "step": 24972 }, { "epoch": 2.97, "learning_rate": 9.819053532358902e-10, "loss": 0.4915, "step": 24973 }, { "epoch": 2.97, "learning_rate": 9.729182914905433e-10, "loss": 0.5932, "step": 24974 }, { "epoch": 2.97, "learning_rate": 9.639725388721488e-10, "loss": 0.5186, "step": 24975 }, { "epoch": 2.97, "learning_rate": 9.550680955291991e-10, "loss": 0.3639, "step": 24976 }, { "epoch": 2.97, "learning_rate": 9.462049616087987e-10, "loss": 0.3968, "step": 24977 }, { "epoch": 2.97, "learning_rate": 9.373831372572195e-10, "loss": 0.4176, "step": 24978 }, { "epoch": 2.97, "learning_rate": 9.286026226201783e-10, "loss": 0.4726, "step": 24979 }, { "epoch": 2.97, "learning_rate": 9.198634178431142e-10, "loss": 0.4472, "step": 24980 }, { "epoch": 2.97, "learning_rate": 9.11165523070634e-10, "loss": 0.6055, "step": 24981 }, { "epoch": 2.98, "learning_rate": 9.025089384459562e-10, "loss": 0.3531, "step": 24982 }, { "epoch": 2.98, "learning_rate": 8.938936641128548e-10, "loss": 0.4044, "step": 24983 }, { "epoch": 2.98, "learning_rate": 8.853197002131608e-10, "loss": 0.4455, "step": 24984 }, { "epoch": 2.98, "learning_rate": 8.767870468887052e-10, "loss": 0.4512, "step": 24985 }, { "epoch": 2.98, "learning_rate": 8.682957042807639e-10, "loss": 0.5428, "step": 24986 }, { "epoch": 2.98, "learning_rate": 8.598456725297799e-10, "loss": 0.5431, "step": 24987 }, { "epoch": 2.98, "learning_rate": 8.514369517750864e-10, "loss": 0.4862, "step": 24988 }, { "epoch": 2.98, "learning_rate": 8.430695421557388e-10, "loss": 0.3503, "step": 24989 }, { "epoch": 2.98, "learning_rate": 8.347434438102375e-10, "loss": 0.3694, "step": 24990 }, { "epoch": 2.98, "learning_rate": 8.2645865687625e-10, "loss": 0.3851, "step": 24991 }, { "epoch": 2.98, "learning_rate": 8.182151814903338e-10, "loss": 0.5513, "step": 24992 }, { "epoch": 2.98, "learning_rate": 8.100130177890464e-10, "loss": 0.6409, "step": 24993 }, { "epoch": 2.98, "learning_rate": 8.018521659081125e-10, "loss": 0.584, "step": 24994 }, { "epoch": 2.98, "learning_rate": 7.93732625982424e-10, "loss": 0.3842, "step": 24995 }, { "epoch": 2.98, "learning_rate": 7.856543981457631e-10, "loss": 0.4221, "step": 24996 }, { "epoch": 2.98, "learning_rate": 7.776174825319116e-10, "loss": 0.3912, "step": 24997 }, { "epoch": 2.98, "learning_rate": 7.696218792738186e-10, "loss": 0.4609, "step": 24998 }, { "epoch": 2.98, "learning_rate": 7.616675885036006e-10, "loss": 0.5663, "step": 24999 }, { "epoch": 2.98, "learning_rate": 7.537546103528193e-10, "loss": 0.5255, "step": 25000 }, { "epoch": 2.98, "learning_rate": 7.458829449519256e-10, "loss": 0.2845, "step": 25001 }, { "epoch": 2.98, "learning_rate": 7.380525924313709e-10, "loss": 0.4704, "step": 25002 }, { "epoch": 2.98, "learning_rate": 7.302635529207735e-10, "loss": 0.4134, "step": 25003 }, { "epoch": 2.98, "learning_rate": 7.225158265483645e-10, "loss": 0.4444, "step": 25004 }, { "epoch": 2.98, "learning_rate": 7.148094134423745e-10, "loss": 0.6096, "step": 25005 }, { "epoch": 2.98, "learning_rate": 7.07144313730479e-10, "loss": 0.523, "step": 25006 }, { "epoch": 2.98, "learning_rate": 6.995205275392436e-10, "loss": 0.2315, "step": 25007 }, { "epoch": 2.98, "learning_rate": 6.919380549946786e-10, "loss": 0.2825, "step": 25008 }, { "epoch": 2.98, "learning_rate": 6.843968962219616e-10, "loss": 0.4586, "step": 25009 }, { "epoch": 2.98, "learning_rate": 6.768970513457151e-10, "loss": 0.4941, "step": 25010 }, { "epoch": 2.98, "learning_rate": 6.694385204902843e-10, "loss": 0.4707, "step": 25011 }, { "epoch": 2.98, "learning_rate": 6.62021303778626e-10, "loss": 0.6439, "step": 25012 }, { "epoch": 2.98, "learning_rate": 6.546454013336978e-10, "loss": 0.5865, "step": 25013 }, { "epoch": 2.98, "learning_rate": 6.473108132770689e-10, "loss": 0.2759, "step": 25014 }, { "epoch": 2.98, "learning_rate": 6.400175397300312e-10, "loss": 0.4124, "step": 25015 }, { "epoch": 2.98, "learning_rate": 6.327655808133215e-10, "loss": 0.4496, "step": 25016 }, { "epoch": 2.98, "learning_rate": 6.255549366465663e-10, "loss": 0.4788, "step": 25017 }, { "epoch": 2.98, "learning_rate": 6.183856073491146e-10, "loss": 0.561, "step": 25018 }, { "epoch": 2.98, "learning_rate": 6.112575930397602e-10, "loss": 0.4669, "step": 25019 }, { "epoch": 2.98, "learning_rate": 6.041708938356317e-10, "loss": 0.4673, "step": 25020 }, { "epoch": 2.98, "learning_rate": 5.971255098544126e-10, "loss": 0.4894, "step": 25021 }, { "epoch": 2.98, "learning_rate": 5.901214412126766e-10, "loss": 0.4717, "step": 25022 }, { "epoch": 2.98, "learning_rate": 5.83158688025609e-10, "loss": 0.4688, "step": 25023 }, { "epoch": 2.98, "learning_rate": 5.762372504089509e-10, "loss": 0.6594, "step": 25024 }, { "epoch": 2.98, "learning_rate": 5.693571284764998e-10, "loss": 0.6676, "step": 25025 }, { "epoch": 2.98, "learning_rate": 5.62518322342609e-10, "loss": 0.3823, "step": 25026 }, { "epoch": 2.98, "learning_rate": 5.557208321196883e-10, "loss": 0.3602, "step": 25027 }, { "epoch": 2.98, "learning_rate": 5.489646579207031e-10, "loss": 0.4438, "step": 25028 }, { "epoch": 2.98, "learning_rate": 5.422497998569532e-10, "loss": 0.4621, "step": 25029 }, { "epoch": 2.98, "learning_rate": 5.355762580394608e-10, "loss": 0.5443, "step": 25030 }, { "epoch": 2.98, "learning_rate": 5.289440325786932e-10, "loss": 0.59, "step": 25031 }, { "epoch": 2.98, "learning_rate": 5.223531235840073e-10, "loss": 0.3209, "step": 25032 }, { "epoch": 2.98, "learning_rate": 5.158035311647602e-10, "loss": 0.3896, "step": 25033 }, { "epoch": 2.98, "learning_rate": 5.09295255428921e-10, "loss": 0.3636, "step": 25034 }, { "epoch": 2.98, "learning_rate": 5.028282964841813e-10, "loss": 0.4526, "step": 25035 }, { "epoch": 2.98, "learning_rate": 4.964026544374001e-10, "loss": 0.5708, "step": 25036 }, { "epoch": 2.98, "learning_rate": 4.900183293948812e-10, "loss": 0.5432, "step": 25037 }, { "epoch": 2.98, "learning_rate": 4.83675321462096e-10, "loss": 0.428, "step": 25038 }, { "epoch": 2.98, "learning_rate": 4.773736307436827e-10, "loss": 0.3481, "step": 25039 }, { "epoch": 2.98, "learning_rate": 4.711132573442801e-10, "loss": 0.4347, "step": 25040 }, { "epoch": 2.98, "learning_rate": 4.6489420136686114e-10, "loss": 0.4538, "step": 25041 }, { "epoch": 2.98, "learning_rate": 4.5871646291467676e-10, "loss": 0.5554, "step": 25042 }, { "epoch": 2.98, "learning_rate": 4.5258004208958983e-10, "loss": 0.7193, "step": 25043 }, { "epoch": 2.98, "learning_rate": 4.4648493899290825e-10, "loss": 0.5242, "step": 25044 }, { "epoch": 2.98, "learning_rate": 4.404311537259398e-10, "loss": 0.2903, "step": 25045 }, { "epoch": 2.98, "learning_rate": 4.344186863880495e-10, "loss": 0.3699, "step": 25046 }, { "epoch": 2.98, "learning_rate": 4.2844753707887986e-10, "loss": 0.4024, "step": 25047 }, { "epoch": 2.98, "learning_rate": 4.225177058975183e-10, "loss": 0.4411, "step": 25048 }, { "epoch": 2.98, "learning_rate": 4.1662919294138684e-10, "loss": 0.5687, "step": 25049 }, { "epoch": 2.98, "learning_rate": 4.107819983081851e-10, "loss": 0.4546, "step": 25050 }, { "epoch": 2.98, "learning_rate": 4.0497612209450255e-10, "loss": 0.3143, "step": 25051 }, { "epoch": 2.98, "learning_rate": 3.9921156439637343e-10, "loss": 0.4975, "step": 25052 }, { "epoch": 2.98, "learning_rate": 3.934883253089994e-10, "loss": 0.4309, "step": 25053 }, { "epoch": 2.98, "learning_rate": 3.8780640492702694e-10, "loss": 0.4576, "step": 25054 }, { "epoch": 2.98, "learning_rate": 3.821658033442699e-10, "loss": 0.5435, "step": 25055 }, { "epoch": 2.98, "learning_rate": 3.76566520653987e-10, "loss": 0.5715, "step": 25056 }, { "epoch": 2.98, "learning_rate": 3.7100855694888195e-10, "loss": 0.2971, "step": 25057 }, { "epoch": 2.98, "learning_rate": 3.654919123208256e-10, "loss": 0.3579, "step": 25058 }, { "epoch": 2.98, "learning_rate": 3.6001658686085627e-10, "loss": 0.4089, "step": 25059 }, { "epoch": 2.98, "learning_rate": 3.545825806594572e-10, "loss": 0.5032, "step": 25060 }, { "epoch": 2.98, "learning_rate": 3.4918989380655633e-10, "loss": 0.6203, "step": 25061 }, { "epoch": 2.98, "learning_rate": 3.438385263915267e-10, "loss": 0.5916, "step": 25062 }, { "epoch": 2.98, "learning_rate": 3.385284785023535e-10, "loss": 0.3049, "step": 25063 }, { "epoch": 2.98, "learning_rate": 3.3325975022702186e-10, "loss": 0.2929, "step": 25064 }, { "epoch": 2.98, "learning_rate": 3.2803234165268426e-10, "loss": 0.4121, "step": 25065 }, { "epoch": 2.99, "learning_rate": 3.2284625286566065e-10, "loss": 0.4035, "step": 25066 }, { "epoch": 2.99, "learning_rate": 3.177014839519932e-10, "loss": 0.4776, "step": 25067 }, { "epoch": 2.99, "learning_rate": 3.125980349960589e-10, "loss": 0.6753, "step": 25068 }, { "epoch": 2.99, "learning_rate": 3.075359060830674e-10, "loss": 0.4807, "step": 25069 }, { "epoch": 2.99, "learning_rate": 3.025150972960078e-10, "loss": 0.3143, "step": 25070 }, { "epoch": 2.99, "learning_rate": 2.9753560871786935e-10, "loss": 0.4176, "step": 25071 }, { "epoch": 2.99, "learning_rate": 2.9259744043164117e-10, "loss": 0.4229, "step": 25072 }, { "epoch": 2.99, "learning_rate": 2.8770059251836956e-10, "loss": 0.4781, "step": 25073 }, { "epoch": 2.99, "learning_rate": 2.828450650591008e-10, "loss": 0.6608, "step": 25074 }, { "epoch": 2.99, "learning_rate": 2.7803085813404853e-10, "loss": 0.5817, "step": 25075 }, { "epoch": 2.99, "learning_rate": 2.732579718231487e-10, "loss": 0.3157, "step": 25076 }, { "epoch": 2.99, "learning_rate": 2.685264062049497e-10, "loss": 0.3173, "step": 25077 }, { "epoch": 2.99, "learning_rate": 2.6383616135772226e-10, "loss": 0.4457, "step": 25078 }, { "epoch": 2.99, "learning_rate": 2.591872373589044e-10, "loss": 0.4737, "step": 25079 }, { "epoch": 2.99, "learning_rate": 2.545796342856566e-10, "loss": 0.5665, "step": 25080 }, { "epoch": 2.99, "learning_rate": 2.5001335221375156e-10, "loss": 0.65, "step": 25081 }, { "epoch": 2.99, "learning_rate": 2.4548839121896205e-10, "loss": 0.3657, "step": 25082 }, { "epoch": 2.99, "learning_rate": 2.4100475137622816e-10, "loss": 0.3787, "step": 25083 }, { "epoch": 2.99, "learning_rate": 2.3656243275910206e-10, "loss": 0.4113, "step": 25084 }, { "epoch": 2.99, "learning_rate": 2.3216143544141367e-10, "loss": 0.3857, "step": 25085 }, { "epoch": 2.99, "learning_rate": 2.2780175949588257e-10, "loss": 0.5566, "step": 25086 }, { "epoch": 2.99, "learning_rate": 2.2348340499467325e-10, "loss": 0.4961, "step": 25087 }, { "epoch": 2.99, "learning_rate": 2.1920637200883999e-10, "loss": 0.3133, "step": 25088 }, { "epoch": 2.99, "learning_rate": 2.1497066060943705e-10, "loss": 0.3579, "step": 25089 }, { "epoch": 2.99, "learning_rate": 2.1077627086613095e-10, "loss": 0.3398, "step": 25090 }, { "epoch": 2.99, "learning_rate": 2.0662320284858816e-10, "loss": 0.4539, "step": 25091 }, { "epoch": 2.99, "learning_rate": 2.0251145662536498e-10, "loss": 0.5671, "step": 25092 }, { "epoch": 2.99, "learning_rate": 1.9844103226418499e-10, "loss": 0.6097, "step": 25093 }, { "epoch": 2.99, "learning_rate": 1.9441192983277178e-10, "loss": 0.5822, "step": 25094 }, { "epoch": 2.99, "learning_rate": 1.9042414939718368e-10, "loss": 0.3568, "step": 25095 }, { "epoch": 2.99, "learning_rate": 1.8647769102375645e-10, "loss": 0.3586, "step": 25096 }, { "epoch": 2.99, "learning_rate": 1.8257255477771574e-10, "loss": 0.5049, "step": 25097 }, { "epoch": 2.99, "learning_rate": 1.7870874072345446e-10, "loss": 0.5415, "step": 25098 }, { "epoch": 2.99, "learning_rate": 1.7488624892481042e-10, "loss": 0.4901, "step": 25099 }, { "epoch": 2.99, "learning_rate": 1.711050794450664e-10, "loss": 0.5102, "step": 25100 }, { "epoch": 2.99, "learning_rate": 1.6736523234667235e-10, "loss": 0.2879, "step": 25101 }, { "epoch": 2.99, "learning_rate": 1.6366670769180083e-10, "loss": 0.4522, "step": 25102 }, { "epoch": 2.99, "learning_rate": 1.6000950554095895e-10, "loss": 0.3961, "step": 25103 }, { "epoch": 2.99, "learning_rate": 1.5639362595493145e-10, "loss": 0.4633, "step": 25104 }, { "epoch": 2.99, "learning_rate": 1.5281906899339284e-10, "loss": 0.5324, "step": 25105 }, { "epoch": 2.99, "learning_rate": 1.4928583471574e-10, "loss": 0.6156, "step": 25106 }, { "epoch": 2.99, "learning_rate": 1.4579392317998208e-10, "loss": 0.3145, "step": 25107 }, { "epoch": 2.99, "learning_rate": 1.4234333444412828e-10, "loss": 0.3453, "step": 25108 }, { "epoch": 2.99, "learning_rate": 1.3893406856507753e-10, "loss": 0.3718, "step": 25109 }, { "epoch": 2.99, "learning_rate": 1.3556612559917358e-10, "loss": 0.4987, "step": 25110 }, { "epoch": 2.99, "learning_rate": 1.3223950560192767e-10, "loss": 0.6169, "step": 25111 }, { "epoch": 2.99, "learning_rate": 1.2895420862885088e-10, "loss": 0.5851, "step": 25112 }, { "epoch": 2.99, "learning_rate": 1.2571023473378907e-10, "loss": 0.3293, "step": 25113 }, { "epoch": 2.99, "learning_rate": 1.2250758397031047e-10, "loss": 0.3432, "step": 25114 }, { "epoch": 2.99, "learning_rate": 1.1934625639170584e-10, "loss": 0.4572, "step": 25115 }, { "epoch": 2.99, "learning_rate": 1.1622625205015558e-10, "loss": 0.4053, "step": 25116 }, { "epoch": 2.99, "learning_rate": 1.1314757099700758e-10, "loss": 0.5305, "step": 25117 }, { "epoch": 2.99, "learning_rate": 1.1011021328333205e-10, "loss": 0.5923, "step": 25118 }, { "epoch": 2.99, "learning_rate": 1.0711417895908904e-10, "loss": 0.6606, "step": 25119 }, { "epoch": 2.99, "learning_rate": 1.041594680742386e-10, "loss": 0.3376, "step": 25120 }, { "epoch": 2.99, "learning_rate": 1.0124608067735297e-10, "loss": 0.4888, "step": 25121 }, { "epoch": 2.99, "learning_rate": 9.837401681672687e-11, "loss": 0.3754, "step": 25122 }, { "epoch": 2.99, "learning_rate": 9.554327653954477e-11, "loss": 0.5142, "step": 25123 }, { "epoch": 2.99, "learning_rate": 9.275385989299113e-11, "loss": 0.6224, "step": 25124 }, { "epoch": 2.99, "learning_rate": 9.000576692286267e-11, "loss": 0.5375, "step": 25125 }, { "epoch": 2.99, "learning_rate": 8.729899767467853e-11, "loss": 0.2955, "step": 25126 }, { "epoch": 2.99, "learning_rate": 8.463355219312518e-11, "loss": 0.3507, "step": 25127 }, { "epoch": 2.99, "learning_rate": 8.200943052261156e-11, "loss": 0.451, "step": 25128 }, { "epoch": 2.99, "learning_rate": 7.942663270588124e-11, "loss": 0.4551, "step": 25129 }, { "epoch": 2.99, "learning_rate": 7.688515878623293e-11, "loss": 0.5598, "step": 25130 }, { "epoch": 2.99, "learning_rate": 7.438500880529997e-11, "loss": 0.6376, "step": 25131 }, { "epoch": 2.99, "learning_rate": 7.192618280471575e-11, "loss": 0.344, "step": 25132 }, { "epoch": 2.99, "learning_rate": 6.95086808250034e-11, "loss": 0.3873, "step": 25133 }, { "epoch": 2.99, "learning_rate": 6.713250290585337e-11, "loss": 0.333, "step": 25134 }, { "epoch": 2.99, "learning_rate": 6.479764908667863e-11, "loss": 0.445, "step": 25135 }, { "epoch": 2.99, "learning_rate": 6.250411940633694e-11, "loss": 0.4698, "step": 25136 }, { "epoch": 2.99, "learning_rate": 6.02519139025759e-11, "loss": 0.5053, "step": 25137 }, { "epoch": 2.99, "learning_rate": 5.804103261258798e-11, "loss": 0.3817, "step": 25138 }, { "epoch": 2.99, "learning_rate": 5.5871475573010533e-11, "loss": 0.4961, "step": 25139 }, { "epoch": 2.99, "learning_rate": 5.3743242819648265e-11, "loss": 0.4468, "step": 25140 }, { "epoch": 2.99, "learning_rate": 5.165633438747319e-11, "loss": 0.4042, "step": 25141 }, { "epoch": 2.99, "learning_rate": 4.9610750311457344e-11, "loss": 0.5005, "step": 25142 }, { "epoch": 2.99, "learning_rate": 4.760649062490741e-11, "loss": 0.5861, "step": 25143 }, { "epoch": 2.99, "learning_rate": 4.564355536113008e-11, "loss": 0.5012, "step": 25144 }, { "epoch": 2.99, "learning_rate": 4.3721944552876926e-11, "loss": 0.3791, "step": 25145 }, { "epoch": 2.99, "learning_rate": 4.1841658231511764e-11, "loss": 0.372, "step": 25146 }, { "epoch": 2.99, "learning_rate": 4.000269642839838e-11, "loss": 0.4399, "step": 25147 }, { "epoch": 2.99, "learning_rate": 3.820505917379036e-11, "loss": 0.5002, "step": 25148 }, { "epoch": 2.99, "learning_rate": 3.6448746497108614e-11, "loss": 0.519, "step": 25149 }, { "epoch": 3.0, "learning_rate": 3.47337584280516e-11, "loss": 0.4939, "step": 25150 }, { "epoch": 3.0, "learning_rate": 3.3060094994374904e-11, "loss": 0.3097, "step": 25151 }, { "epoch": 3.0, "learning_rate": 3.142775622411165e-11, "loss": 0.4263, "step": 25152 }, { "epoch": 3.0, "learning_rate": 2.983674214418475e-11, "loss": 0.4599, "step": 25153 }, { "epoch": 3.0, "learning_rate": 2.8287052780684442e-11, "loss": 0.3897, "step": 25154 }, { "epoch": 3.0, "learning_rate": 2.6778688159423417e-11, "loss": 0.6198, "step": 25155 }, { "epoch": 3.0, "learning_rate": 2.5311648305381685e-11, "loss": 0.5809, "step": 25156 }, { "epoch": 3.0, "learning_rate": 2.388593324242905e-11, "loss": 0.2787, "step": 25157 }, { "epoch": 3.0, "learning_rate": 2.2501542994712854e-11, "loss": 0.3747, "step": 25158 }, { "epoch": 3.0, "learning_rate": 2.1158477584437564e-11, "loss": 0.4362, "step": 25159 }, { "epoch": 3.0, "learning_rate": 1.985673703436275e-11, "loss": 0.4444, "step": 25160 }, { "epoch": 3.0, "learning_rate": 1.8596321365860203e-11, "loss": 0.545, "step": 25161 }, { "epoch": 3.0, "learning_rate": 1.7377230599746608e-11, "loss": 0.5168, "step": 25162 }, { "epoch": 3.0, "learning_rate": 1.6199464756283535e-11, "loss": 0.4369, "step": 25163 }, { "epoch": 3.0, "learning_rate": 1.5063023854622327e-11, "loss": 0.4201, "step": 25164 }, { "epoch": 3.0, "learning_rate": 1.396790791363678e-11, "loss": 0.3772, "step": 25165 }, { "epoch": 3.0, "learning_rate": 1.2914116951645573e-11, "loss": 0.3652, "step": 25166 }, { "epoch": 3.0, "learning_rate": 1.1901650986134717e-11, "loss": 0.4988, "step": 25167 }, { "epoch": 3.0, "learning_rate": 1.0930510033480001e-11, "loss": 0.6481, "step": 25168 }, { "epoch": 3.0, "learning_rate": 1.0000694109779662e-11, "loss": 0.6157, "step": 25169 }, { "epoch": 3.0, "learning_rate": 9.112203230854377e-12, "loss": 0.3779, "step": 25170 }, { "epoch": 3.0, "learning_rate": 8.26503741085949e-12, "loss": 0.3937, "step": 25171 }, { "epoch": 3.0, "learning_rate": 7.45919666395034e-12, "loss": 0.4286, "step": 25172 }, { "epoch": 3.0, "learning_rate": 6.694681003727166e-12, "loss": 0.6003, "step": 25173 }, { "epoch": 3.0, "learning_rate": 5.971490442402417e-12, "loss": 0.5283, "step": 25174 }, { "epoch": 3.0, "learning_rate": 5.289624992188547e-12, "loss": 0.507, "step": 25175 }, { "epoch": 3.0, "learning_rate": 4.649084664187786e-12, "loss": 0.3313, "step": 25176 }, { "epoch": 3.0, "learning_rate": 4.04986946922481e-12, "loss": 0.3312, "step": 25177 }, { "epoch": 3.0, "learning_rate": 3.4919794172916245e-12, "loss": 0.433, "step": 25178 }, { "epoch": 3.0, "learning_rate": 2.9754145172700145e-12, "loss": 0.4659, "step": 25179 }, { "epoch": 3.0, "learning_rate": 2.5001747777642084e-12, "loss": 0.5487, "step": 25180 }, { "epoch": 3.0, "learning_rate": 2.066260206545767e-12, "loss": 0.596, "step": 25181 }, { "epoch": 3.0, "learning_rate": 1.6736708111086963e-12, "loss": 0.3219, "step": 25182 }, { "epoch": 3.0, "learning_rate": 1.3224065978367784e-12, "loss": 0.3112, "step": 25183 }, { "epoch": 3.0, "learning_rate": 1.0124675722811285e-12, "loss": 0.3853, "step": 25184 }, { "epoch": 3.0, "learning_rate": 7.438537399928614e-13, "loss": 0.491, "step": 25185 }, { "epoch": 3.0, "learning_rate": 5.165651051353137e-13, "loss": 0.5107, "step": 25186 }, { "epoch": 3.0, "learning_rate": 3.3060167131671037e-13, "loss": 0.6429, "step": 25187 }, { "epoch": 3.0, "learning_rate": 1.8596344186772032e-13, "loss": 0.3725, "step": 25188 }, { "epoch": 3.0, "learning_rate": 8.265041928634531e-14, "loss": 0.4766, "step": 25189 }, { "epoch": 3.0, "learning_rate": 2.0662604960364206e-14, "loss": 0.4945, "step": 25190 }, { "epoch": 3.0, "learning_rate": 0.0, "loss": 0.5034, "step": 25191 }, { "epoch": 3.0, "step": 25191, "total_flos": 4967667393127424.0, "train_loss": 0.650876347066808, "train_runtime": 60670.5076, "train_samples_per_second": 13.286, "train_steps_per_second": 0.415 } ], "max_steps": 25191, "num_train_epochs": 3, "total_flos": 4967667393127424.0, "trial_name": null, "trial_params": null }