{ "best_metric": null, "best_model_checkpoint": null, "epoch": 0.6996805111821086, "eval_steps": 500, "global_step": 15024, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 4.4345898004434594e-08, "loss": 4.8175, "step": 1 }, { "epoch": 0.0, "learning_rate": 8.869179600886919e-08, "loss": 4.9065, "step": 2 }, { "epoch": 0.0, "learning_rate": 1.3303769401330377e-07, "loss": 4.8986, "step": 3 }, { "epoch": 0.0, "learning_rate": 1.7738359201773838e-07, "loss": 4.874, "step": 4 }, { "epoch": 0.0, "learning_rate": 2.2172949002217298e-07, "loss": 4.9563, "step": 5 }, { "epoch": 0.0, "learning_rate": 2.6607538802660754e-07, "loss": 4.9716, "step": 6 }, { "epoch": 0.0, "learning_rate": 3.1042128603104217e-07, "loss": 4.9427, "step": 7 }, { "epoch": 0.0, "learning_rate": 3.5476718403547675e-07, "loss": 4.9774, "step": 8 }, { "epoch": 0.0, "learning_rate": 3.9911308203991133e-07, "loss": 4.8781, "step": 9 }, { "epoch": 0.0, "learning_rate": 4.4345898004434597e-07, "loss": 4.8404, "step": 10 }, { "epoch": 0.0, "learning_rate": 4.878048780487805e-07, "loss": 4.843, "step": 11 }, { "epoch": 0.0, "learning_rate": 5.321507760532151e-07, "loss": 4.8867, "step": 12 }, { "epoch": 0.0, "learning_rate": 5.764966740576497e-07, "loss": 4.8705, "step": 13 }, { "epoch": 0.0, "learning_rate": 6.208425720620843e-07, "loss": 4.9949, "step": 14 }, { "epoch": 0.0, "learning_rate": 6.651884700665189e-07, "loss": 4.7571, "step": 15 }, { "epoch": 0.0, "learning_rate": 7.095343680709535e-07, "loss": 4.7447, "step": 16 }, { "epoch": 0.0, "learning_rate": 7.538802660753881e-07, "loss": 4.7472, "step": 17 }, { "epoch": 0.0, "learning_rate": 7.982261640798227e-07, "loss": 4.7241, "step": 18 }, { "epoch": 0.0, "learning_rate": 8.425720620842573e-07, "loss": 4.8774, "step": 19 }, { "epoch": 0.0, "learning_rate": 8.869179600886919e-07, "loss": 4.7116, "step": 20 }, { "epoch": 0.0, "learning_rate": 9.312638580931265e-07, "loss": 4.4686, "step": 21 }, { "epoch": 0.0, "learning_rate": 9.75609756097561e-07, "loss": 4.3648, "step": 22 }, { "epoch": 0.0, "learning_rate": 1.0199556541019955e-06, "loss": 4.404, "step": 23 }, { "epoch": 0.0, "learning_rate": 1.0643015521064302e-06, "loss": 4.5061, "step": 24 }, { "epoch": 0.0, "learning_rate": 1.1086474501108648e-06, "loss": 4.3759, "step": 25 }, { "epoch": 0.0, "learning_rate": 1.1529933481152994e-06, "loss": 4.3672, "step": 26 }, { "epoch": 0.0, "learning_rate": 1.197339246119734e-06, "loss": 4.2962, "step": 27 }, { "epoch": 0.0, "learning_rate": 1.2416851441241687e-06, "loss": 4.0021, "step": 28 }, { "epoch": 0.0, "learning_rate": 1.2860310421286031e-06, "loss": 3.9543, "step": 29 }, { "epoch": 0.0, "learning_rate": 1.3303769401330377e-06, "loss": 3.8272, "step": 30 }, { "epoch": 0.0, "learning_rate": 1.3747228381374724e-06, "loss": 3.8693, "step": 31 }, { "epoch": 0.0, "learning_rate": 1.419068736141907e-06, "loss": 3.8192, "step": 32 }, { "epoch": 0.0, "learning_rate": 1.4634146341463414e-06, "loss": 3.8076, "step": 33 }, { "epoch": 0.0, "learning_rate": 1.5077605321507763e-06, "loss": 3.7096, "step": 34 }, { "epoch": 0.0, "learning_rate": 1.5521064301552107e-06, "loss": 3.7557, "step": 35 }, { "epoch": 0.0, "learning_rate": 1.5964523281596453e-06, "loss": 3.6775, "step": 36 }, { "epoch": 0.0, "learning_rate": 1.6407982261640802e-06, "loss": 3.7345, "step": 37 }, { "epoch": 0.0, "learning_rate": 1.6851441241685146e-06, "loss": 3.5835, "step": 38 }, { "epoch": 0.0, "learning_rate": 1.729490022172949e-06, "loss": 3.6294, "step": 39 }, { "epoch": 0.0, "learning_rate": 1.7738359201773839e-06, "loss": 3.5217, "step": 40 }, { "epoch": 0.0, "learning_rate": 1.8181818181818183e-06, "loss": 3.5784, "step": 41 }, { "epoch": 0.0, "learning_rate": 1.862527716186253e-06, "loss": 3.4816, "step": 42 }, { "epoch": 0.0, "learning_rate": 1.9068736141906873e-06, "loss": 3.5155, "step": 43 }, { "epoch": 0.0, "learning_rate": 1.951219512195122e-06, "loss": 3.45, "step": 44 }, { "epoch": 0.0, "learning_rate": 1.995565410199557e-06, "loss": 3.4511, "step": 45 }, { "epoch": 0.0, "learning_rate": 2.039911308203991e-06, "loss": 3.3953, "step": 46 }, { "epoch": 0.0, "learning_rate": 2.084257206208426e-06, "loss": 3.48, "step": 47 }, { "epoch": 0.0, "learning_rate": 2.1286031042128603e-06, "loss": 3.4064, "step": 48 }, { "epoch": 0.0, "learning_rate": 2.172949002217295e-06, "loss": 3.3884, "step": 49 }, { "epoch": 0.0, "learning_rate": 2.2172949002217296e-06, "loss": 3.3157, "step": 50 }, { "epoch": 0.0, "learning_rate": 2.261640798226164e-06, "loss": 3.4318, "step": 51 }, { "epoch": 0.0, "learning_rate": 2.305986696230599e-06, "loss": 3.2922, "step": 52 }, { "epoch": 0.0, "learning_rate": 2.3503325942350335e-06, "loss": 3.3768, "step": 53 }, { "epoch": 0.0, "learning_rate": 2.394678492239468e-06, "loss": 3.2558, "step": 54 }, { "epoch": 0.0, "learning_rate": 2.4390243902439027e-06, "loss": 3.3012, "step": 55 }, { "epoch": 0.0, "learning_rate": 2.4833702882483374e-06, "loss": 3.2121, "step": 56 }, { "epoch": 0.0, "learning_rate": 2.5277161862527716e-06, "loss": 3.3225, "step": 57 }, { "epoch": 0.0, "learning_rate": 2.5720620842572062e-06, "loss": 3.2427, "step": 58 }, { "epoch": 0.0, "learning_rate": 2.6164079822616413e-06, "loss": 3.2805, "step": 59 }, { "epoch": 0.0, "learning_rate": 2.6607538802660755e-06, "loss": 3.2157, "step": 60 }, { "epoch": 0.0, "learning_rate": 2.70509977827051e-06, "loss": 3.2323, "step": 61 }, { "epoch": 0.0, "learning_rate": 2.7494456762749448e-06, "loss": 3.1427, "step": 62 }, { "epoch": 0.0, "learning_rate": 2.7937915742793794e-06, "loss": 3.1982, "step": 63 }, { "epoch": 0.0, "learning_rate": 2.838137472283814e-06, "loss": 3.1814, "step": 64 }, { "epoch": 0.0, "learning_rate": 2.8824833702882487e-06, "loss": 3.2795, "step": 65 }, { "epoch": 0.0, "learning_rate": 2.926829268292683e-06, "loss": 3.1183, "step": 66 }, { "epoch": 0.0, "learning_rate": 2.971175166297118e-06, "loss": 3.1536, "step": 67 }, { "epoch": 0.0, "learning_rate": 3.0155210643015526e-06, "loss": 3.1286, "step": 68 }, { "epoch": 0.0, "learning_rate": 3.0598669623059868e-06, "loss": 3.1448, "step": 69 }, { "epoch": 0.0, "learning_rate": 3.1042128603104214e-06, "loss": 3.1145, "step": 70 }, { "epoch": 0.0, "learning_rate": 3.1485587583148565e-06, "loss": 3.1417, "step": 71 }, { "epoch": 0.0, "learning_rate": 3.1929046563192907e-06, "loss": 3.0799, "step": 72 }, { "epoch": 0.0, "learning_rate": 3.2372505543237253e-06, "loss": 3.079, "step": 73 }, { "epoch": 0.0, "learning_rate": 3.2815964523281604e-06, "loss": 3.0526, "step": 74 }, { "epoch": 0.0, "learning_rate": 3.3259423503325946e-06, "loss": 3.1142, "step": 75 }, { "epoch": 0.01, "learning_rate": 3.370288248337029e-06, "loss": 3.068, "step": 76 }, { "epoch": 0.01, "learning_rate": 3.414634146341464e-06, "loss": 3.0434, "step": 77 }, { "epoch": 0.01, "learning_rate": 3.458980044345898e-06, "loss": 2.9785, "step": 78 }, { "epoch": 0.01, "learning_rate": 3.503325942350333e-06, "loss": 3.0936, "step": 79 }, { "epoch": 0.01, "learning_rate": 3.5476718403547677e-06, "loss": 2.9895, "step": 80 }, { "epoch": 0.01, "learning_rate": 3.592017738359202e-06, "loss": 3.0492, "step": 81 }, { "epoch": 0.01, "learning_rate": 3.6363636363636366e-06, "loss": 2.9405, "step": 82 }, { "epoch": 0.01, "learning_rate": 3.680709534368071e-06, "loss": 3.0529, "step": 83 }, { "epoch": 0.01, "learning_rate": 3.725055432372506e-06, "loss": 2.9318, "step": 84 }, { "epoch": 0.01, "learning_rate": 3.7694013303769405e-06, "loss": 2.954, "step": 85 }, { "epoch": 0.01, "learning_rate": 3.8137472283813747e-06, "loss": 2.9138, "step": 86 }, { "epoch": 0.01, "learning_rate": 3.858093126385809e-06, "loss": 2.9393, "step": 87 }, { "epoch": 0.01, "learning_rate": 3.902439024390244e-06, "loss": 2.891, "step": 88 }, { "epoch": 0.01, "learning_rate": 3.946784922394679e-06, "loss": 2.9745, "step": 89 }, { "epoch": 0.01, "learning_rate": 3.991130820399114e-06, "loss": 2.921, "step": 90 }, { "epoch": 0.01, "learning_rate": 4.035476718403548e-06, "loss": 2.9222, "step": 91 }, { "epoch": 0.01, "learning_rate": 4.079822616407982e-06, "loss": 2.8907, "step": 92 }, { "epoch": 0.01, "learning_rate": 4.124168514412417e-06, "loss": 2.9163, "step": 93 }, { "epoch": 0.01, "learning_rate": 4.168514412416852e-06, "loss": 2.7732, "step": 94 }, { "epoch": 0.01, "learning_rate": 4.212860310421286e-06, "loss": 2.8666, "step": 95 }, { "epoch": 0.01, "learning_rate": 4.257206208425721e-06, "loss": 2.7829, "step": 96 }, { "epoch": 0.01, "learning_rate": 4.301552106430156e-06, "loss": 2.8673, "step": 97 }, { "epoch": 0.01, "learning_rate": 4.34589800443459e-06, "loss": 2.7658, "step": 98 }, { "epoch": 0.01, "learning_rate": 4.390243902439025e-06, "loss": 2.8648, "step": 99 }, { "epoch": 0.01, "learning_rate": 4.434589800443459e-06, "loss": 2.8146, "step": 100 }, { "epoch": 0.01, "learning_rate": 4.478935698447893e-06, "loss": 2.8125, "step": 101 }, { "epoch": 0.01, "learning_rate": 4.523281596452328e-06, "loss": 2.7512, "step": 102 }, { "epoch": 0.01, "learning_rate": 4.5676274944567635e-06, "loss": 2.7727, "step": 103 }, { "epoch": 0.01, "learning_rate": 4.611973392461198e-06, "loss": 2.7411, "step": 104 }, { "epoch": 0.01, "learning_rate": 4.656319290465633e-06, "loss": 2.8245, "step": 105 }, { "epoch": 0.01, "learning_rate": 4.700665188470067e-06, "loss": 2.7223, "step": 106 }, { "epoch": 0.01, "learning_rate": 4.745011086474501e-06, "loss": 2.7916, "step": 107 }, { "epoch": 0.01, "learning_rate": 4.789356984478936e-06, "loss": 2.7149, "step": 108 }, { "epoch": 0.01, "learning_rate": 4.833702882483371e-06, "loss": 2.7723, "step": 109 }, { "epoch": 0.01, "learning_rate": 4.8780487804878055e-06, "loss": 2.688, "step": 110 }, { "epoch": 0.01, "learning_rate": 4.92239467849224e-06, "loss": 2.7204, "step": 111 }, { "epoch": 0.01, "learning_rate": 4.966740576496675e-06, "loss": 2.7042, "step": 112 }, { "epoch": 0.01, "learning_rate": 5.011086474501109e-06, "loss": 2.7136, "step": 113 }, { "epoch": 0.01, "learning_rate": 5.055432372505543e-06, "loss": 2.6694, "step": 114 }, { "epoch": 0.01, "learning_rate": 5.099778270509978e-06, "loss": 2.7061, "step": 115 }, { "epoch": 0.01, "learning_rate": 5.1441241685144124e-06, "loss": 2.6252, "step": 116 }, { "epoch": 0.01, "learning_rate": 5.188470066518847e-06, "loss": 2.6758, "step": 117 }, { "epoch": 0.01, "learning_rate": 5.2328159645232826e-06, "loss": 2.5827, "step": 118 }, { "epoch": 0.01, "learning_rate": 5.277161862527717e-06, "loss": 2.5884, "step": 119 }, { "epoch": 0.01, "learning_rate": 5.321507760532151e-06, "loss": 2.5919, "step": 120 }, { "epoch": 0.01, "learning_rate": 5.365853658536586e-06, "loss": 2.6568, "step": 121 }, { "epoch": 0.01, "learning_rate": 5.41019955654102e-06, "loss": 2.5799, "step": 122 }, { "epoch": 0.01, "learning_rate": 5.4545454545454545e-06, "loss": 2.6094, "step": 123 }, { "epoch": 0.01, "learning_rate": 5.4988913525498895e-06, "loss": 2.523, "step": 124 }, { "epoch": 0.01, "learning_rate": 5.5432372505543246e-06, "loss": 2.5519, "step": 125 }, { "epoch": 0.01, "learning_rate": 5.587583148558759e-06, "loss": 2.5407, "step": 126 }, { "epoch": 0.01, "learning_rate": 5.631929046563194e-06, "loss": 2.5417, "step": 127 }, { "epoch": 0.01, "learning_rate": 5.676274944567628e-06, "loss": 2.5156, "step": 128 }, { "epoch": 0.01, "learning_rate": 5.720620842572062e-06, "loss": 2.5434, "step": 129 }, { "epoch": 0.01, "learning_rate": 5.764966740576497e-06, "loss": 2.4697, "step": 130 }, { "epoch": 0.01, "learning_rate": 5.8093126385809315e-06, "loss": 2.5331, "step": 131 }, { "epoch": 0.01, "learning_rate": 5.853658536585366e-06, "loss": 2.4299, "step": 132 }, { "epoch": 0.01, "learning_rate": 5.898004434589802e-06, "loss": 2.4642, "step": 133 }, { "epoch": 0.01, "learning_rate": 5.942350332594236e-06, "loss": 2.4163, "step": 134 }, { "epoch": 0.01, "learning_rate": 5.98669623059867e-06, "loss": 2.4794, "step": 135 }, { "epoch": 0.01, "learning_rate": 6.031042128603105e-06, "loss": 2.4429, "step": 136 }, { "epoch": 0.01, "learning_rate": 6.075388026607539e-06, "loss": 2.4157, "step": 137 }, { "epoch": 0.01, "learning_rate": 6.1197339246119735e-06, "loss": 2.4434, "step": 138 }, { "epoch": 0.01, "learning_rate": 6.164079822616409e-06, "loss": 2.4451, "step": 139 }, { "epoch": 0.01, "learning_rate": 6.208425720620843e-06, "loss": 2.381, "step": 140 }, { "epoch": 0.01, "learning_rate": 6.252771618625277e-06, "loss": 2.4419, "step": 141 }, { "epoch": 0.01, "learning_rate": 6.297117516629713e-06, "loss": 2.3609, "step": 142 }, { "epoch": 0.01, "learning_rate": 6.341463414634147e-06, "loss": 2.3762, "step": 143 }, { "epoch": 0.01, "learning_rate": 6.385809312638581e-06, "loss": 2.3376, "step": 144 }, { "epoch": 0.01, "learning_rate": 6.430155210643016e-06, "loss": 2.4116, "step": 145 }, { "epoch": 0.01, "learning_rate": 6.474501108647451e-06, "loss": 2.3498, "step": 146 }, { "epoch": 0.01, "learning_rate": 6.518847006651885e-06, "loss": 2.4143, "step": 147 }, { "epoch": 0.01, "learning_rate": 6.563192904656321e-06, "loss": 2.3161, "step": 148 }, { "epoch": 0.01, "learning_rate": 6.607538802660755e-06, "loss": 2.3299, "step": 149 }, { "epoch": 0.01, "learning_rate": 6.651884700665189e-06, "loss": 2.297, "step": 150 }, { "epoch": 0.01, "learning_rate": 6.696230598669624e-06, "loss": 2.3711, "step": 151 }, { "epoch": 0.01, "learning_rate": 6.740576496674058e-06, "loss": 2.2249, "step": 152 }, { "epoch": 0.01, "learning_rate": 6.784922394678493e-06, "loss": 2.3264, "step": 153 }, { "epoch": 0.01, "learning_rate": 6.829268292682928e-06, "loss": 2.2601, "step": 154 }, { "epoch": 0.01, "learning_rate": 6.873614190687362e-06, "loss": 2.3183, "step": 155 }, { "epoch": 0.01, "learning_rate": 6.917960088691796e-06, "loss": 2.2387, "step": 156 }, { "epoch": 0.01, "learning_rate": 6.962305986696232e-06, "loss": 2.2853, "step": 157 }, { "epoch": 0.01, "learning_rate": 7.006651884700666e-06, "loss": 2.2673, "step": 158 }, { "epoch": 0.01, "learning_rate": 7.0509977827051e-06, "loss": 2.2615, "step": 159 }, { "epoch": 0.01, "learning_rate": 7.0953436807095355e-06, "loss": 2.1728, "step": 160 }, { "epoch": 0.01, "learning_rate": 7.13968957871397e-06, "loss": 2.2367, "step": 161 }, { "epoch": 0.01, "learning_rate": 7.184035476718404e-06, "loss": 2.2481, "step": 162 }, { "epoch": 0.01, "learning_rate": 7.228381374722838e-06, "loss": 2.2719, "step": 163 }, { "epoch": 0.01, "learning_rate": 7.272727272727273e-06, "loss": 2.2049, "step": 164 }, { "epoch": 0.01, "learning_rate": 7.317073170731707e-06, "loss": 2.2364, "step": 165 }, { "epoch": 0.01, "learning_rate": 7.361419068736142e-06, "loss": 2.2188, "step": 166 }, { "epoch": 0.01, "learning_rate": 7.4057649667405775e-06, "loss": 2.2403, "step": 167 }, { "epoch": 0.01, "learning_rate": 7.450110864745012e-06, "loss": 2.1651, "step": 168 }, { "epoch": 0.01, "learning_rate": 7.494456762749446e-06, "loss": 2.2554, "step": 169 }, { "epoch": 0.01, "learning_rate": 7.538802660753881e-06, "loss": 2.1556, "step": 170 }, { "epoch": 0.01, "learning_rate": 7.583148558758315e-06, "loss": 2.1865, "step": 171 }, { "epoch": 0.01, "learning_rate": 7.627494456762749e-06, "loss": 2.1692, "step": 172 }, { "epoch": 0.01, "learning_rate": 7.671840354767184e-06, "loss": 2.198, "step": 173 }, { "epoch": 0.01, "learning_rate": 7.716186252771619e-06, "loss": 2.1498, "step": 174 }, { "epoch": 0.01, "learning_rate": 7.760532150776053e-06, "loss": 2.187, "step": 175 }, { "epoch": 0.01, "learning_rate": 7.804878048780489e-06, "loss": 2.0839, "step": 176 }, { "epoch": 0.01, "learning_rate": 7.849223946784923e-06, "loss": 2.1897, "step": 177 }, { "epoch": 0.01, "learning_rate": 7.893569844789357e-06, "loss": 2.0773, "step": 178 }, { "epoch": 0.01, "learning_rate": 7.937915742793793e-06, "loss": 2.1616, "step": 179 }, { "epoch": 0.01, "learning_rate": 7.982261640798227e-06, "loss": 2.1027, "step": 180 }, { "epoch": 0.01, "learning_rate": 8.026607538802662e-06, "loss": 2.1392, "step": 181 }, { "epoch": 0.01, "learning_rate": 8.070953436807096e-06, "loss": 2.1056, "step": 182 }, { "epoch": 0.01, "learning_rate": 8.11529933481153e-06, "loss": 2.1376, "step": 183 }, { "epoch": 0.01, "learning_rate": 8.159645232815964e-06, "loss": 2.0933, "step": 184 }, { "epoch": 0.01, "learning_rate": 8.2039911308204e-06, "loss": 2.1411, "step": 185 }, { "epoch": 0.01, "learning_rate": 8.248337028824834e-06, "loss": 2.1024, "step": 186 }, { "epoch": 0.01, "learning_rate": 8.292682926829268e-06, "loss": 2.124, "step": 187 }, { "epoch": 0.01, "learning_rate": 8.337028824833704e-06, "loss": 2.0544, "step": 188 }, { "epoch": 0.01, "learning_rate": 8.381374722838139e-06, "loss": 2.1246, "step": 189 }, { "epoch": 0.01, "learning_rate": 8.425720620842573e-06, "loss": 2.0404, "step": 190 }, { "epoch": 0.01, "learning_rate": 8.470066518847007e-06, "loss": 2.0693, "step": 191 }, { "epoch": 0.01, "learning_rate": 8.514412416851441e-06, "loss": 2.1029, "step": 192 }, { "epoch": 0.01, "learning_rate": 8.558758314855875e-06, "loss": 2.1369, "step": 193 }, { "epoch": 0.01, "learning_rate": 8.603104212860311e-06, "loss": 2.0703, "step": 194 }, { "epoch": 0.01, "learning_rate": 8.647450110864746e-06, "loss": 2.1051, "step": 195 }, { "epoch": 0.01, "learning_rate": 8.69179600886918e-06, "loss": 2.015, "step": 196 }, { "epoch": 0.01, "learning_rate": 8.736141906873616e-06, "loss": 2.1873, "step": 197 }, { "epoch": 0.01, "learning_rate": 8.78048780487805e-06, "loss": 2.0934, "step": 198 }, { "epoch": 0.01, "learning_rate": 8.824833702882484e-06, "loss": 2.1003, "step": 199 }, { "epoch": 0.01, "learning_rate": 8.869179600886918e-06, "loss": 2.0735, "step": 200 }, { "epoch": 0.01, "learning_rate": 8.913525498891353e-06, "loss": 2.1056, "step": 201 }, { "epoch": 0.01, "learning_rate": 8.957871396895787e-06, "loss": 2.0909, "step": 202 }, { "epoch": 0.01, "learning_rate": 9.002217294900223e-06, "loss": 2.1188, "step": 203 }, { "epoch": 0.01, "learning_rate": 9.046563192904657e-06, "loss": 2.0783, "step": 204 }, { "epoch": 0.01, "learning_rate": 9.090909090909091e-06, "loss": 2.1094, "step": 205 }, { "epoch": 0.01, "learning_rate": 9.135254988913527e-06, "loss": 2.0776, "step": 206 }, { "epoch": 0.01, "learning_rate": 9.179600886917961e-06, "loss": 2.0563, "step": 207 }, { "epoch": 0.01, "learning_rate": 9.223946784922395e-06, "loss": 2.047, "step": 208 }, { "epoch": 0.01, "learning_rate": 9.268292682926831e-06, "loss": 2.1128, "step": 209 }, { "epoch": 0.01, "learning_rate": 9.312638580931265e-06, "loss": 2.0241, "step": 210 }, { "epoch": 0.01, "learning_rate": 9.3569844789357e-06, "loss": 2.0791, "step": 211 }, { "epoch": 0.01, "learning_rate": 9.401330376940134e-06, "loss": 2.0709, "step": 212 }, { "epoch": 0.01, "learning_rate": 9.445676274944568e-06, "loss": 2.0701, "step": 213 }, { "epoch": 0.01, "learning_rate": 9.490022172949002e-06, "loss": 2.015, "step": 214 }, { "epoch": 0.01, "learning_rate": 9.534368070953438e-06, "loss": 2.0857, "step": 215 }, { "epoch": 0.01, "learning_rate": 9.578713968957872e-06, "loss": 2.0444, "step": 216 }, { "epoch": 0.01, "learning_rate": 9.623059866962307e-06, "loss": 2.1089, "step": 217 }, { "epoch": 0.01, "learning_rate": 9.667405764966743e-06, "loss": 1.9947, "step": 218 }, { "epoch": 0.01, "learning_rate": 9.711751662971177e-06, "loss": 2.1325, "step": 219 }, { "epoch": 0.01, "learning_rate": 9.756097560975611e-06, "loss": 2.0237, "step": 220 }, { "epoch": 0.01, "learning_rate": 9.800443458980045e-06, "loss": 2.0393, "step": 221 }, { "epoch": 0.01, "learning_rate": 9.84478935698448e-06, "loss": 2.0584, "step": 222 }, { "epoch": 0.01, "learning_rate": 9.889135254988914e-06, "loss": 2.0864, "step": 223 }, { "epoch": 0.01, "learning_rate": 9.93348115299335e-06, "loss": 2.0721, "step": 224 }, { "epoch": 0.01, "learning_rate": 9.977827050997784e-06, "loss": 2.0824, "step": 225 }, { "epoch": 0.02, "learning_rate": 1.0022172949002218e-05, "loss": 2.0354, "step": 226 }, { "epoch": 0.02, "learning_rate": 1.0066518847006652e-05, "loss": 2.0446, "step": 227 }, { "epoch": 0.02, "learning_rate": 1.0110864745011086e-05, "loss": 2.0138, "step": 228 }, { "epoch": 0.02, "learning_rate": 1.0155210643015522e-05, "loss": 2.0736, "step": 229 }, { "epoch": 0.02, "learning_rate": 1.0199556541019956e-05, "loss": 2.0348, "step": 230 }, { "epoch": 0.02, "learning_rate": 1.024390243902439e-05, "loss": 2.0796, "step": 231 }, { "epoch": 0.02, "learning_rate": 1.0288248337028825e-05, "loss": 2.0042, "step": 232 }, { "epoch": 0.02, "learning_rate": 1.0332594235033259e-05, "loss": 2.0931, "step": 233 }, { "epoch": 0.02, "learning_rate": 1.0376940133037693e-05, "loss": 2.0102, "step": 234 }, { "epoch": 0.02, "learning_rate": 1.0421286031042131e-05, "loss": 2.0179, "step": 235 }, { "epoch": 0.02, "learning_rate": 1.0465631929046565e-05, "loss": 2.0212, "step": 236 }, { "epoch": 0.02, "learning_rate": 1.0509977827051e-05, "loss": 2.0754, "step": 237 }, { "epoch": 0.02, "learning_rate": 1.0554323725055434e-05, "loss": 2.001, "step": 238 }, { "epoch": 0.02, "learning_rate": 1.0598669623059868e-05, "loss": 2.0521, "step": 239 }, { "epoch": 0.02, "learning_rate": 1.0643015521064302e-05, "loss": 1.9838, "step": 240 }, { "epoch": 0.02, "learning_rate": 1.0687361419068738e-05, "loss": 2.034, "step": 241 }, { "epoch": 0.02, "learning_rate": 1.0731707317073172e-05, "loss": 2.028, "step": 242 }, { "epoch": 0.02, "learning_rate": 1.0776053215077606e-05, "loss": 2.0387, "step": 243 }, { "epoch": 0.02, "learning_rate": 1.082039911308204e-05, "loss": 2.0247, "step": 244 }, { "epoch": 0.02, "learning_rate": 1.0864745011086475e-05, "loss": 2.0403, "step": 245 }, { "epoch": 0.02, "learning_rate": 1.0909090909090909e-05, "loss": 1.9954, "step": 246 }, { "epoch": 0.02, "learning_rate": 1.0953436807095345e-05, "loss": 2.0102, "step": 247 }, { "epoch": 0.02, "learning_rate": 1.0997782705099779e-05, "loss": 2.0263, "step": 248 }, { "epoch": 0.02, "learning_rate": 1.1042128603104215e-05, "loss": 2.0078, "step": 249 }, { "epoch": 0.02, "learning_rate": 1.1086474501108649e-05, "loss": 2.0212, "step": 250 }, { "epoch": 0.02, "learning_rate": 1.1130820399113083e-05, "loss": 2.0445, "step": 251 }, { "epoch": 0.02, "learning_rate": 1.1175166297117518e-05, "loss": 2.0171, "step": 252 }, { "epoch": 0.02, "learning_rate": 1.1219512195121953e-05, "loss": 2.0624, "step": 253 }, { "epoch": 0.02, "learning_rate": 1.1263858093126388e-05, "loss": 1.9875, "step": 254 }, { "epoch": 0.02, "learning_rate": 1.1308203991130822e-05, "loss": 2.0181, "step": 255 }, { "epoch": 0.02, "learning_rate": 1.1352549889135256e-05, "loss": 1.9753, "step": 256 }, { "epoch": 0.02, "learning_rate": 1.139689578713969e-05, "loss": 2.018, "step": 257 }, { "epoch": 0.02, "learning_rate": 1.1441241685144125e-05, "loss": 1.9988, "step": 258 }, { "epoch": 0.02, "learning_rate": 1.1485587583148559e-05, "loss": 1.9902, "step": 259 }, { "epoch": 0.02, "learning_rate": 1.1529933481152995e-05, "loss": 1.9724, "step": 260 }, { "epoch": 0.02, "learning_rate": 1.1574279379157429e-05, "loss": 2.0733, "step": 261 }, { "epoch": 0.02, "learning_rate": 1.1618625277161863e-05, "loss": 1.9515, "step": 262 }, { "epoch": 0.02, "learning_rate": 1.1662971175166297e-05, "loss": 1.9684, "step": 263 }, { "epoch": 0.02, "learning_rate": 1.1707317073170731e-05, "loss": 1.9764, "step": 264 }, { "epoch": 0.02, "learning_rate": 1.1751662971175166e-05, "loss": 1.9758, "step": 265 }, { "epoch": 0.02, "learning_rate": 1.1796008869179603e-05, "loss": 1.9444, "step": 266 }, { "epoch": 0.02, "learning_rate": 1.1840354767184037e-05, "loss": 2.0045, "step": 267 }, { "epoch": 0.02, "learning_rate": 1.1884700665188472e-05, "loss": 1.9953, "step": 268 }, { "epoch": 0.02, "learning_rate": 1.1929046563192906e-05, "loss": 2.0374, "step": 269 }, { "epoch": 0.02, "learning_rate": 1.197339246119734e-05, "loss": 1.9419, "step": 270 }, { "epoch": 0.02, "learning_rate": 1.2017738359201774e-05, "loss": 1.9608, "step": 271 }, { "epoch": 0.02, "learning_rate": 1.206208425720621e-05, "loss": 1.9441, "step": 272 }, { "epoch": 0.02, "learning_rate": 1.2106430155210644e-05, "loss": 2.0198, "step": 273 }, { "epoch": 0.02, "learning_rate": 1.2150776053215079e-05, "loss": 1.9165, "step": 274 }, { "epoch": 0.02, "learning_rate": 1.2195121951219513e-05, "loss": 2.0083, "step": 275 }, { "epoch": 0.02, "learning_rate": 1.2239467849223947e-05, "loss": 1.9016, "step": 276 }, { "epoch": 0.02, "learning_rate": 1.2283813747228381e-05, "loss": 1.9836, "step": 277 }, { "epoch": 0.02, "learning_rate": 1.2328159645232817e-05, "loss": 1.9407, "step": 278 }, { "epoch": 0.02, "learning_rate": 1.2372505543237251e-05, "loss": 2.0017, "step": 279 }, { "epoch": 0.02, "learning_rate": 1.2416851441241686e-05, "loss": 1.9178, "step": 280 }, { "epoch": 0.02, "learning_rate": 1.246119733924612e-05, "loss": 2.0242, "step": 281 }, { "epoch": 0.02, "learning_rate": 1.2505543237250554e-05, "loss": 1.9386, "step": 282 }, { "epoch": 0.02, "learning_rate": 1.2549889135254988e-05, "loss": 1.9905, "step": 283 }, { "epoch": 0.02, "learning_rate": 1.2594235033259426e-05, "loss": 1.949, "step": 284 }, { "epoch": 0.02, "learning_rate": 1.263858093126386e-05, "loss": 2.0047, "step": 285 }, { "epoch": 0.02, "learning_rate": 1.2682926829268294e-05, "loss": 1.9247, "step": 286 }, { "epoch": 0.02, "learning_rate": 1.2727272727272728e-05, "loss": 1.9448, "step": 287 }, { "epoch": 0.02, "learning_rate": 1.2771618625277163e-05, "loss": 1.9319, "step": 288 }, { "epoch": 0.02, "learning_rate": 1.2815964523281597e-05, "loss": 1.9637, "step": 289 }, { "epoch": 0.02, "learning_rate": 1.2860310421286033e-05, "loss": 1.9428, "step": 290 }, { "epoch": 0.02, "learning_rate": 1.2904656319290467e-05, "loss": 1.9681, "step": 291 }, { "epoch": 0.02, "learning_rate": 1.2949002217294901e-05, "loss": 1.9072, "step": 292 }, { "epoch": 0.02, "learning_rate": 1.2993348115299335e-05, "loss": 1.959, "step": 293 }, { "epoch": 0.02, "learning_rate": 1.303769401330377e-05, "loss": 1.9356, "step": 294 }, { "epoch": 0.02, "learning_rate": 1.3082039911308204e-05, "loss": 1.9936, "step": 295 }, { "epoch": 0.02, "learning_rate": 1.3126385809312641e-05, "loss": 1.9423, "step": 296 }, { "epoch": 0.02, "learning_rate": 1.3170731707317076e-05, "loss": 1.9288, "step": 297 }, { "epoch": 0.02, "learning_rate": 1.321507760532151e-05, "loss": 1.9237, "step": 298 }, { "epoch": 0.02, "learning_rate": 1.3259423503325944e-05, "loss": 1.9945, "step": 299 }, { "epoch": 0.02, "learning_rate": 1.3303769401330378e-05, "loss": 1.9161, "step": 300 }, { "epoch": 0.02, "learning_rate": 1.3348115299334812e-05, "loss": 1.9811, "step": 301 }, { "epoch": 0.02, "learning_rate": 1.3392461197339248e-05, "loss": 1.9346, "step": 302 }, { "epoch": 0.02, "learning_rate": 1.3436807095343683e-05, "loss": 1.967, "step": 303 }, { "epoch": 0.02, "learning_rate": 1.3481152993348117e-05, "loss": 1.8775, "step": 304 }, { "epoch": 0.02, "learning_rate": 1.3525498891352551e-05, "loss": 1.955, "step": 305 }, { "epoch": 0.02, "learning_rate": 1.3569844789356985e-05, "loss": 1.8481, "step": 306 }, { "epoch": 0.02, "learning_rate": 1.361419068736142e-05, "loss": 1.9594, "step": 307 }, { "epoch": 0.02, "learning_rate": 1.3658536585365855e-05, "loss": 1.948, "step": 308 }, { "epoch": 0.02, "learning_rate": 1.370288248337029e-05, "loss": 1.9592, "step": 309 }, { "epoch": 0.02, "learning_rate": 1.3747228381374724e-05, "loss": 1.902, "step": 310 }, { "epoch": 0.02, "learning_rate": 1.3791574279379158e-05, "loss": 1.9412, "step": 311 }, { "epoch": 0.02, "learning_rate": 1.3835920177383592e-05, "loss": 1.9026, "step": 312 }, { "epoch": 0.02, "learning_rate": 1.3880266075388026e-05, "loss": 1.957, "step": 313 }, { "epoch": 0.02, "learning_rate": 1.3924611973392464e-05, "loss": 1.8643, "step": 314 }, { "epoch": 0.02, "learning_rate": 1.3968957871396898e-05, "loss": 1.9262, "step": 315 }, { "epoch": 0.02, "learning_rate": 1.4013303769401332e-05, "loss": 1.9232, "step": 316 }, { "epoch": 0.02, "learning_rate": 1.4057649667405767e-05, "loss": 1.9475, "step": 317 }, { "epoch": 0.02, "learning_rate": 1.41019955654102e-05, "loss": 1.9476, "step": 318 }, { "epoch": 0.02, "learning_rate": 1.4146341463414635e-05, "loss": 1.947, "step": 319 }, { "epoch": 0.02, "learning_rate": 1.4190687361419071e-05, "loss": 1.9107, "step": 320 }, { "epoch": 0.02, "learning_rate": 1.4235033259423505e-05, "loss": 1.9425, "step": 321 }, { "epoch": 0.02, "learning_rate": 1.427937915742794e-05, "loss": 1.8985, "step": 322 }, { "epoch": 0.02, "learning_rate": 1.4323725055432374e-05, "loss": 1.8995, "step": 323 }, { "epoch": 0.02, "learning_rate": 1.4368070953436808e-05, "loss": 1.8952, "step": 324 }, { "epoch": 0.02, "learning_rate": 1.4412416851441242e-05, "loss": 1.9628, "step": 325 }, { "epoch": 0.02, "learning_rate": 1.4456762749445676e-05, "loss": 1.8931, "step": 326 }, { "epoch": 0.02, "learning_rate": 1.4501108647450112e-05, "loss": 1.9194, "step": 327 }, { "epoch": 0.02, "learning_rate": 1.4545454545454546e-05, "loss": 1.8896, "step": 328 }, { "epoch": 0.02, "learning_rate": 1.458980044345898e-05, "loss": 1.9574, "step": 329 }, { "epoch": 0.02, "learning_rate": 1.4634146341463415e-05, "loss": 1.9116, "step": 330 }, { "epoch": 0.02, "learning_rate": 1.4678492239467849e-05, "loss": 1.9856, "step": 331 }, { "epoch": 0.02, "learning_rate": 1.4722838137472283e-05, "loss": 1.8853, "step": 332 }, { "epoch": 0.02, "learning_rate": 1.476718403547672e-05, "loss": 1.9599, "step": 333 }, { "epoch": 0.02, "learning_rate": 1.4811529933481155e-05, "loss": 1.8873, "step": 334 }, { "epoch": 0.02, "learning_rate": 1.485587583148559e-05, "loss": 1.9621, "step": 335 }, { "epoch": 0.02, "learning_rate": 1.4900221729490023e-05, "loss": 1.9276, "step": 336 }, { "epoch": 0.02, "learning_rate": 1.4944567627494458e-05, "loss": 1.9419, "step": 337 }, { "epoch": 0.02, "learning_rate": 1.4988913525498892e-05, "loss": 1.8825, "step": 338 }, { "epoch": 0.02, "learning_rate": 1.5033259423503328e-05, "loss": 1.9418, "step": 339 }, { "epoch": 0.02, "learning_rate": 1.5077605321507762e-05, "loss": 1.8848, "step": 340 }, { "epoch": 0.02, "learning_rate": 1.5121951219512196e-05, "loss": 1.9213, "step": 341 }, { "epoch": 0.02, "learning_rate": 1.516629711751663e-05, "loss": 1.8734, "step": 342 }, { "epoch": 0.02, "learning_rate": 1.5210643015521065e-05, "loss": 1.9771, "step": 343 }, { "epoch": 0.02, "learning_rate": 1.5254988913525499e-05, "loss": 1.874, "step": 344 }, { "epoch": 0.02, "learning_rate": 1.5299334811529935e-05, "loss": 1.9789, "step": 345 }, { "epoch": 0.02, "learning_rate": 1.534368070953437e-05, "loss": 1.8871, "step": 346 }, { "epoch": 0.02, "learning_rate": 1.5388026607538803e-05, "loss": 1.9169, "step": 347 }, { "epoch": 0.02, "learning_rate": 1.5432372505543237e-05, "loss": 1.8508, "step": 348 }, { "epoch": 0.02, "learning_rate": 1.547671840354767e-05, "loss": 1.9157, "step": 349 }, { "epoch": 0.02, "learning_rate": 1.5521064301552106e-05, "loss": 1.8988, "step": 350 }, { "epoch": 0.02, "learning_rate": 1.5565410199556543e-05, "loss": 1.8533, "step": 351 }, { "epoch": 0.02, "learning_rate": 1.5609756097560978e-05, "loss": 1.8341, "step": 352 }, { "epoch": 0.02, "learning_rate": 1.5654101995565412e-05, "loss": 1.9159, "step": 353 }, { "epoch": 0.02, "learning_rate": 1.5698447893569846e-05, "loss": 1.8895, "step": 354 }, { "epoch": 0.02, "learning_rate": 1.574279379157428e-05, "loss": 1.9382, "step": 355 }, { "epoch": 0.02, "learning_rate": 1.5787139689578714e-05, "loss": 1.8689, "step": 356 }, { "epoch": 0.02, "learning_rate": 1.5831485587583152e-05, "loss": 1.9198, "step": 357 }, { "epoch": 0.02, "learning_rate": 1.5875831485587586e-05, "loss": 1.8873, "step": 358 }, { "epoch": 0.02, "learning_rate": 1.592017738359202e-05, "loss": 1.9452, "step": 359 }, { "epoch": 0.02, "learning_rate": 1.5964523281596455e-05, "loss": 1.8492, "step": 360 }, { "epoch": 0.02, "learning_rate": 1.600886917960089e-05, "loss": 1.9234, "step": 361 }, { "epoch": 0.02, "learning_rate": 1.6053215077605323e-05, "loss": 1.8911, "step": 362 }, { "epoch": 0.02, "learning_rate": 1.6097560975609757e-05, "loss": 1.8467, "step": 363 }, { "epoch": 0.02, "learning_rate": 1.614190687361419e-05, "loss": 1.8853, "step": 364 }, { "epoch": 0.02, "learning_rate": 1.6186252771618626e-05, "loss": 1.9439, "step": 365 }, { "epoch": 0.02, "learning_rate": 1.623059866962306e-05, "loss": 1.8822, "step": 366 }, { "epoch": 0.02, "learning_rate": 1.6274944567627494e-05, "loss": 1.9469, "step": 367 }, { "epoch": 0.02, "learning_rate": 1.6319290465631928e-05, "loss": 1.8464, "step": 368 }, { "epoch": 0.02, "learning_rate": 1.6363636363636366e-05, "loss": 1.9473, "step": 369 }, { "epoch": 0.02, "learning_rate": 1.64079822616408e-05, "loss": 1.8746, "step": 370 }, { "epoch": 0.02, "learning_rate": 1.6452328159645234e-05, "loss": 1.938, "step": 371 }, { "epoch": 0.02, "learning_rate": 1.649667405764967e-05, "loss": 1.8893, "step": 372 }, { "epoch": 0.02, "learning_rate": 1.6541019955654103e-05, "loss": 1.916, "step": 373 }, { "epoch": 0.02, "learning_rate": 1.6585365853658537e-05, "loss": 1.849, "step": 374 }, { "epoch": 0.02, "learning_rate": 1.6629711751662975e-05, "loss": 1.8748, "step": 375 }, { "epoch": 0.03, "learning_rate": 1.667405764966741e-05, "loss": 1.8555, "step": 376 }, { "epoch": 0.03, "learning_rate": 1.6718403547671843e-05, "loss": 1.9372, "step": 377 }, { "epoch": 0.03, "learning_rate": 1.6762749445676277e-05, "loss": 1.8843, "step": 378 }, { "epoch": 0.03, "learning_rate": 1.680709534368071e-05, "loss": 1.9044, "step": 379 }, { "epoch": 0.03, "learning_rate": 1.6851441241685146e-05, "loss": 1.8315, "step": 380 }, { "epoch": 0.03, "learning_rate": 1.689578713968958e-05, "loss": 1.9283, "step": 381 }, { "epoch": 0.03, "learning_rate": 1.6940133037694014e-05, "loss": 1.8669, "step": 382 }, { "epoch": 0.03, "learning_rate": 1.6984478935698448e-05, "loss": 1.9228, "step": 383 }, { "epoch": 0.03, "learning_rate": 1.7028824833702882e-05, "loss": 1.8837, "step": 384 }, { "epoch": 0.03, "learning_rate": 1.7073170731707317e-05, "loss": 1.9231, "step": 385 }, { "epoch": 0.03, "learning_rate": 1.711751662971175e-05, "loss": 1.867, "step": 386 }, { "epoch": 0.03, "learning_rate": 1.7161862527716185e-05, "loss": 1.9044, "step": 387 }, { "epoch": 0.03, "learning_rate": 1.7206208425720623e-05, "loss": 1.839, "step": 388 }, { "epoch": 0.03, "learning_rate": 1.7250554323725057e-05, "loss": 1.8978, "step": 389 }, { "epoch": 0.03, "learning_rate": 1.729490022172949e-05, "loss": 1.8767, "step": 390 }, { "epoch": 0.03, "learning_rate": 1.7339246119733925e-05, "loss": 1.9412, "step": 391 }, { "epoch": 0.03, "learning_rate": 1.738359201773836e-05, "loss": 1.9017, "step": 392 }, { "epoch": 0.03, "learning_rate": 1.7427937915742794e-05, "loss": 1.8877, "step": 393 }, { "epoch": 0.03, "learning_rate": 1.747228381374723e-05, "loss": 1.8702, "step": 394 }, { "epoch": 0.03, "learning_rate": 1.7516629711751666e-05, "loss": 1.8955, "step": 395 }, { "epoch": 0.03, "learning_rate": 1.75609756097561e-05, "loss": 1.8159, "step": 396 }, { "epoch": 0.03, "learning_rate": 1.7605321507760534e-05, "loss": 1.8665, "step": 397 }, { "epoch": 0.03, "learning_rate": 1.7649667405764968e-05, "loss": 1.8351, "step": 398 }, { "epoch": 0.03, "learning_rate": 1.7694013303769402e-05, "loss": 1.8672, "step": 399 }, { "epoch": 0.03, "learning_rate": 1.7738359201773837e-05, "loss": 1.8384, "step": 400 }, { "epoch": 0.03, "learning_rate": 1.778270509977827e-05, "loss": 1.8928, "step": 401 }, { "epoch": 0.03, "learning_rate": 1.7827050997782705e-05, "loss": 1.8612, "step": 402 }, { "epoch": 0.03, "learning_rate": 1.787139689578714e-05, "loss": 1.8676, "step": 403 }, { "epoch": 0.03, "learning_rate": 1.7915742793791573e-05, "loss": 1.833, "step": 404 }, { "epoch": 0.03, "learning_rate": 1.796008869179601e-05, "loss": 1.9171, "step": 405 }, { "epoch": 0.03, "learning_rate": 1.8004434589800445e-05, "loss": 1.8578, "step": 406 }, { "epoch": 0.03, "learning_rate": 1.804878048780488e-05, "loss": 1.9045, "step": 407 }, { "epoch": 0.03, "learning_rate": 1.8093126385809314e-05, "loss": 1.8346, "step": 408 }, { "epoch": 0.03, "learning_rate": 1.8137472283813748e-05, "loss": 1.8863, "step": 409 }, { "epoch": 0.03, "learning_rate": 1.8181818181818182e-05, "loss": 1.8744, "step": 410 }, { "epoch": 0.03, "learning_rate": 1.8226164079822616e-05, "loss": 1.8836, "step": 411 }, { "epoch": 0.03, "learning_rate": 1.8270509977827054e-05, "loss": 1.8303, "step": 412 }, { "epoch": 0.03, "learning_rate": 1.8314855875831488e-05, "loss": 1.9023, "step": 413 }, { "epoch": 0.03, "learning_rate": 1.8359201773835922e-05, "loss": 1.8109, "step": 414 }, { "epoch": 0.03, "learning_rate": 1.8403547671840357e-05, "loss": 1.8837, "step": 415 }, { "epoch": 0.03, "learning_rate": 1.844789356984479e-05, "loss": 1.8659, "step": 416 }, { "epoch": 0.03, "learning_rate": 1.8492239467849225e-05, "loss": 1.8563, "step": 417 }, { "epoch": 0.03, "learning_rate": 1.8536585365853663e-05, "loss": 1.828, "step": 418 }, { "epoch": 0.03, "learning_rate": 1.8580931263858097e-05, "loss": 1.8523, "step": 419 }, { "epoch": 0.03, "learning_rate": 1.862527716186253e-05, "loss": 1.7823, "step": 420 }, { "epoch": 0.03, "learning_rate": 1.8669623059866965e-05, "loss": 1.8795, "step": 421 }, { "epoch": 0.03, "learning_rate": 1.87139689578714e-05, "loss": 1.8258, "step": 422 }, { "epoch": 0.03, "learning_rate": 1.8758314855875834e-05, "loss": 1.8531, "step": 423 }, { "epoch": 0.03, "learning_rate": 1.8802660753880268e-05, "loss": 1.797, "step": 424 }, { "epoch": 0.03, "learning_rate": 1.8847006651884702e-05, "loss": 1.8891, "step": 425 }, { "epoch": 0.03, "learning_rate": 1.8891352549889136e-05, "loss": 1.8487, "step": 426 }, { "epoch": 0.03, "learning_rate": 1.893569844789357e-05, "loss": 1.8843, "step": 427 }, { "epoch": 0.03, "learning_rate": 1.8980044345898005e-05, "loss": 1.8397, "step": 428 }, { "epoch": 0.03, "learning_rate": 1.902439024390244e-05, "loss": 1.9043, "step": 429 }, { "epoch": 0.03, "learning_rate": 1.9068736141906876e-05, "loss": 1.8669, "step": 430 }, { "epoch": 0.03, "learning_rate": 1.911308203991131e-05, "loss": 1.8846, "step": 431 }, { "epoch": 0.03, "learning_rate": 1.9157427937915745e-05, "loss": 1.8009, "step": 432 }, { "epoch": 0.03, "learning_rate": 1.920177383592018e-05, "loss": 1.8913, "step": 433 }, { "epoch": 0.03, "learning_rate": 1.9246119733924613e-05, "loss": 1.8093, "step": 434 }, { "epoch": 0.03, "learning_rate": 1.9290465631929047e-05, "loss": 1.8493, "step": 435 }, { "epoch": 0.03, "learning_rate": 1.9334811529933485e-05, "loss": 1.8034, "step": 436 }, { "epoch": 0.03, "learning_rate": 1.937915742793792e-05, "loss": 1.8604, "step": 437 }, { "epoch": 0.03, "learning_rate": 1.9423503325942354e-05, "loss": 1.7832, "step": 438 }, { "epoch": 0.03, "learning_rate": 1.9467849223946788e-05, "loss": 1.8211, "step": 439 }, { "epoch": 0.03, "learning_rate": 1.9512195121951222e-05, "loss": 1.8288, "step": 440 }, { "epoch": 0.03, "learning_rate": 1.9556541019955656e-05, "loss": 1.8277, "step": 441 }, { "epoch": 0.03, "learning_rate": 1.960088691796009e-05, "loss": 1.7732, "step": 442 }, { "epoch": 0.03, "learning_rate": 1.9645232815964525e-05, "loss": 1.8692, "step": 443 }, { "epoch": 0.03, "learning_rate": 1.968957871396896e-05, "loss": 1.8429, "step": 444 }, { "epoch": 0.03, "learning_rate": 1.9733924611973393e-05, "loss": 1.8514, "step": 445 }, { "epoch": 0.03, "learning_rate": 1.9778270509977827e-05, "loss": 1.7937, "step": 446 }, { "epoch": 0.03, "learning_rate": 1.982261640798226e-05, "loss": 1.8535, "step": 447 }, { "epoch": 0.03, "learning_rate": 1.98669623059867e-05, "loss": 1.7841, "step": 448 }, { "epoch": 0.03, "learning_rate": 1.9911308203991133e-05, "loss": 1.8499, "step": 449 }, { "epoch": 0.03, "learning_rate": 1.9955654101995567e-05, "loss": 1.8036, "step": 450 }, { "epoch": 0.03, "learning_rate": 2e-05, "loss": 1.8131, "step": 451 }, { "epoch": 0.03, "learning_rate": 1.9999999767634412e-05, "loss": 1.7668, "step": 452 }, { "epoch": 0.03, "learning_rate": 1.9999999070537647e-05, "loss": 1.8442, "step": 453 }, { "epoch": 0.03, "learning_rate": 1.9999997908709746e-05, "loss": 1.7977, "step": 454 }, { "epoch": 0.03, "learning_rate": 1.9999996282150755e-05, "loss": 1.8268, "step": 455 }, { "epoch": 0.03, "learning_rate": 1.9999994190860757e-05, "loss": 1.7879, "step": 456 }, { "epoch": 0.03, "learning_rate": 1.9999991634839847e-05, "loss": 1.7982, "step": 457 }, { "epoch": 0.03, "learning_rate": 1.9999988614088142e-05, "loss": 1.7768, "step": 458 }, { "epoch": 0.03, "learning_rate": 1.9999985128605782e-05, "loss": 1.822, "step": 459 }, { "epoch": 0.03, "learning_rate": 1.9999981178392934e-05, "loss": 1.7957, "step": 460 }, { "epoch": 0.03, "learning_rate": 1.9999976763449775e-05, "loss": 1.8294, "step": 461 }, { "epoch": 0.03, "learning_rate": 1.9999971883776514e-05, "loss": 1.7339, "step": 462 }, { "epoch": 0.03, "learning_rate": 1.9999966539373378e-05, "loss": 1.7973, "step": 463 }, { "epoch": 0.03, "learning_rate": 1.9999960730240615e-05, "loss": 1.7758, "step": 464 }, { "epoch": 0.03, "learning_rate": 1.999995445637849e-05, "loss": 1.7924, "step": 465 }, { "epoch": 0.03, "learning_rate": 1.9999947717787303e-05, "loss": 1.7469, "step": 466 }, { "epoch": 0.03, "learning_rate": 1.999994051446736e-05, "loss": 1.7609, "step": 467 }, { "epoch": 0.03, "learning_rate": 1.9999932846419e-05, "loss": 1.7546, "step": 468 }, { "epoch": 0.03, "learning_rate": 1.999992471364258e-05, "loss": 1.7202, "step": 469 }, { "epoch": 0.03, "learning_rate": 1.9999916116138475e-05, "loss": 1.7145, "step": 470 }, { "epoch": 0.03, "learning_rate": 1.9999907053907085e-05, "loss": 1.7661, "step": 471 }, { "epoch": 0.03, "learning_rate": 1.9999897526948832e-05, "loss": 1.6582, "step": 472 }, { "epoch": 0.03, "learning_rate": 1.9999887535264157e-05, "loss": 1.757, "step": 473 }, { "epoch": 0.03, "learning_rate": 1.9999877078853528e-05, "loss": 1.6761, "step": 474 }, { "epoch": 0.03, "learning_rate": 1.9999866157717428e-05, "loss": 1.7207, "step": 475 }, { "epoch": 0.03, "learning_rate": 1.9999854771856366e-05, "loss": 1.6876, "step": 476 }, { "epoch": 0.03, "learning_rate": 1.999984292127087e-05, "loss": 1.7413, "step": 477 }, { "epoch": 0.03, "learning_rate": 1.9999830605961495e-05, "loss": 1.6619, "step": 478 }, { "epoch": 0.03, "learning_rate": 1.9999817825928804e-05, "loss": 1.6829, "step": 479 }, { "epoch": 0.03, "learning_rate": 1.99998045811734e-05, "loss": 1.6196, "step": 480 }, { "epoch": 0.03, "learning_rate": 1.999979087169589e-05, "loss": 1.6585, "step": 481 }, { "epoch": 0.03, "learning_rate": 1.9999776697496923e-05, "loss": 1.5916, "step": 482 }, { "epoch": 0.03, "learning_rate": 1.9999762058577147e-05, "loss": 1.6409, "step": 483 }, { "epoch": 0.03, "learning_rate": 1.9999746954937244e-05, "loss": 1.5948, "step": 484 }, { "epoch": 0.03, "learning_rate": 1.999973138657792e-05, "loss": 1.6209, "step": 485 }, { "epoch": 0.03, "learning_rate": 1.99997153534999e-05, "loss": 1.5931, "step": 486 }, { "epoch": 0.03, "learning_rate": 1.9999698855703924e-05, "loss": 1.5758, "step": 487 }, { "epoch": 0.03, "learning_rate": 1.999968189319076e-05, "loss": 1.529, "step": 488 }, { "epoch": 0.03, "learning_rate": 1.9999664465961195e-05, "loss": 1.5428, "step": 489 }, { "epoch": 0.03, "learning_rate": 1.9999646574016044e-05, "loss": 1.5418, "step": 490 }, { "epoch": 0.03, "learning_rate": 1.999962821735613e-05, "loss": 1.507, "step": 491 }, { "epoch": 0.03, "learning_rate": 1.9999609395982314e-05, "loss": 1.4518, "step": 492 }, { "epoch": 0.03, "learning_rate": 1.9999590109895467e-05, "loss": 1.5332, "step": 493 }, { "epoch": 0.03, "learning_rate": 1.999957035909649e-05, "loss": 1.5043, "step": 494 }, { "epoch": 0.03, "learning_rate": 1.9999550143586295e-05, "loss": 1.4566, "step": 495 }, { "epoch": 0.03, "learning_rate": 1.999952946336582e-05, "loss": 1.4546, "step": 496 }, { "epoch": 0.03, "learning_rate": 1.9999508318436035e-05, "loss": 1.4871, "step": 497 }, { "epoch": 0.03, "learning_rate": 1.999948670879791e-05, "loss": 1.4202, "step": 498 }, { "epoch": 0.03, "learning_rate": 1.9999464634452467e-05, "loss": 1.4653, "step": 499 }, { "epoch": 0.03, "learning_rate": 1.9999442095400713e-05, "loss": 1.4642, "step": 500 }, { "epoch": 0.03, "learning_rate": 1.999941909164371e-05, "loss": 1.4341, "step": 501 }, { "epoch": 0.03, "learning_rate": 1.9999395623182515e-05, "loss": 1.4278, "step": 502 }, { "epoch": 0.03, "learning_rate": 1.9999371690018227e-05, "loss": 1.4384, "step": 503 }, { "epoch": 0.03, "learning_rate": 1.9999347292151955e-05, "loss": 1.4143, "step": 504 }, { "epoch": 0.03, "learning_rate": 1.9999322429584835e-05, "loss": 1.386, "step": 505 }, { "epoch": 0.03, "learning_rate": 1.999929710231802e-05, "loss": 1.3646, "step": 506 }, { "epoch": 0.03, "learning_rate": 1.999927131035269e-05, "loss": 1.4536, "step": 507 }, { "epoch": 0.03, "learning_rate": 1.999924505369004e-05, "loss": 1.4128, "step": 508 }, { "epoch": 0.03, "learning_rate": 1.9999218332331292e-05, "loss": 1.4431, "step": 509 }, { "epoch": 0.03, "learning_rate": 1.999919114627769e-05, "loss": 1.3771, "step": 510 }, { "epoch": 0.03, "learning_rate": 1.9999163495530495e-05, "loss": 1.4022, "step": 511 }, { "epoch": 0.03, "learning_rate": 1.999913538009099e-05, "loss": 1.3783, "step": 512 }, { "epoch": 0.03, "learning_rate": 1.9999106799960483e-05, "loss": 1.4256, "step": 513 }, { "epoch": 0.03, "learning_rate": 1.9999077755140305e-05, "loss": 1.3518, "step": 514 }, { "epoch": 0.03, "learning_rate": 1.9999048245631803e-05, "loss": 1.3835, "step": 515 }, { "epoch": 0.03, "learning_rate": 1.9999018271436352e-05, "loss": 1.3437, "step": 516 }, { "epoch": 0.03, "learning_rate": 1.999898783255534e-05, "loss": 1.3743, "step": 517 }, { "epoch": 0.03, "learning_rate": 1.9998956928990184e-05, "loss": 1.3762, "step": 518 }, { "epoch": 0.03, "learning_rate": 1.999892556074232e-05, "loss": 1.3648, "step": 519 }, { "epoch": 0.03, "learning_rate": 1.9998893727813206e-05, "loss": 1.3263, "step": 520 }, { "epoch": 0.03, "learning_rate": 1.9998861430204323e-05, "loss": 1.3828, "step": 521 }, { "epoch": 0.03, "learning_rate": 1.9998828667917167e-05, "loss": 1.3406, "step": 522 }, { "epoch": 0.03, "learning_rate": 1.9998795440953266e-05, "loss": 1.3589, "step": 523 }, { "epoch": 0.03, "learning_rate": 1.999876174931416e-05, "loss": 1.368, "step": 524 }, { "epoch": 0.03, "learning_rate": 1.999872759300142e-05, "loss": 1.335, "step": 525 }, { "epoch": 0.04, "learning_rate": 1.999869297201663e-05, "loss": 1.3011, "step": 526 }, { "epoch": 0.04, "learning_rate": 1.9998657886361393e-05, "loss": 1.3353, "step": 527 }, { "epoch": 0.04, "learning_rate": 1.999862233603735e-05, "loss": 1.3615, "step": 528 }, { "epoch": 0.04, "learning_rate": 1.9998586321046148e-05, "loss": 1.3384, "step": 529 }, { "epoch": 0.04, "learning_rate": 1.9998549841389464e-05, "loss": 1.3057, "step": 530 }, { "epoch": 0.04, "learning_rate": 1.9998512897068987e-05, "loss": 1.3186, "step": 531 }, { "epoch": 0.04, "learning_rate": 1.999847548808644e-05, "loss": 1.3215, "step": 532 }, { "epoch": 0.04, "learning_rate": 1.9998437614443554e-05, "loss": 1.3254, "step": 533 }, { "epoch": 0.04, "learning_rate": 1.99983992761421e-05, "loss": 1.3363, "step": 534 }, { "epoch": 0.04, "learning_rate": 1.9998360473183852e-05, "loss": 1.3806, "step": 535 }, { "epoch": 0.04, "learning_rate": 1.9998321205570616e-05, "loss": 1.3803, "step": 536 }, { "epoch": 0.04, "learning_rate": 1.9998281473304216e-05, "loss": 1.3395, "step": 537 }, { "epoch": 0.04, "learning_rate": 1.9998241276386498e-05, "loss": 1.3089, "step": 538 }, { "epoch": 0.04, "learning_rate": 1.999820061481933e-05, "loss": 1.3132, "step": 539 }, { "epoch": 0.04, "learning_rate": 1.9998159488604602e-05, "loss": 1.3122, "step": 540 }, { "epoch": 0.04, "learning_rate": 1.9998117897744228e-05, "loss": 1.3572, "step": 541 }, { "epoch": 0.04, "learning_rate": 1.9998075842240136e-05, "loss": 1.2923, "step": 542 }, { "epoch": 0.04, "learning_rate": 1.9998033322094284e-05, "loss": 1.3538, "step": 543 }, { "epoch": 0.04, "learning_rate": 1.999799033730865e-05, "loss": 1.298, "step": 544 }, { "epoch": 0.04, "learning_rate": 1.9997946887885224e-05, "loss": 1.3426, "step": 545 }, { "epoch": 0.04, "learning_rate": 1.999790297382603e-05, "loss": 1.2845, "step": 546 }, { "epoch": 0.04, "learning_rate": 1.999785859513311e-05, "loss": 1.3251, "step": 547 }, { "epoch": 0.04, "learning_rate": 1.9997813751808523e-05, "loss": 1.2685, "step": 548 }, { "epoch": 0.04, "learning_rate": 1.999776844385436e-05, "loss": 1.3265, "step": 549 }, { "epoch": 0.04, "learning_rate": 1.9997722671272716e-05, "loss": 1.2853, "step": 550 }, { "epoch": 0.04, "learning_rate": 1.9997676434065728e-05, "loss": 1.3201, "step": 551 }, { "epoch": 0.04, "learning_rate": 1.999762973223554e-05, "loss": 1.2778, "step": 552 }, { "epoch": 0.04, "learning_rate": 1.9997582565784324e-05, "loss": 1.2479, "step": 553 }, { "epoch": 0.04, "learning_rate": 1.9997534934714267e-05, "loss": 1.3094, "step": 554 }, { "epoch": 0.04, "learning_rate": 1.999748683902759e-05, "loss": 1.3054, "step": 555 }, { "epoch": 0.04, "learning_rate": 1.9997438278726525e-05, "loss": 1.307, "step": 556 }, { "epoch": 0.04, "learning_rate": 1.9997389253813324e-05, "loss": 1.3287, "step": 557 }, { "epoch": 0.04, "learning_rate": 1.9997339764290275e-05, "loss": 1.3195, "step": 558 }, { "epoch": 0.04, "learning_rate": 1.999728981015967e-05, "loss": 1.324, "step": 559 }, { "epoch": 0.04, "learning_rate": 1.9997239391423834e-05, "loss": 1.2763, "step": 560 }, { "epoch": 0.04, "learning_rate": 1.999718850808511e-05, "loss": 1.3157, "step": 561 }, { "epoch": 0.04, "learning_rate": 1.999713716014586e-05, "loss": 1.2499, "step": 562 }, { "epoch": 0.04, "learning_rate": 1.9997085347608475e-05, "loss": 1.3394, "step": 563 }, { "epoch": 0.04, "learning_rate": 1.999703307047536e-05, "loss": 1.274, "step": 564 }, { "epoch": 0.04, "learning_rate": 1.9996980328748944e-05, "loss": 1.312, "step": 565 }, { "epoch": 0.04, "learning_rate": 1.9996927122431676e-05, "loss": 1.2763, "step": 566 }, { "epoch": 0.04, "learning_rate": 1.9996873451526038e-05, "loss": 1.3456, "step": 567 }, { "epoch": 0.04, "learning_rate": 1.999681931603451e-05, "loss": 1.2527, "step": 568 }, { "epoch": 0.04, "learning_rate": 1.999676471595962e-05, "loss": 1.3215, "step": 569 }, { "epoch": 0.04, "learning_rate": 1.9996709651303903e-05, "loss": 1.3015, "step": 570 }, { "epoch": 0.04, "learning_rate": 1.999665412206991e-05, "loss": 1.2426, "step": 571 }, { "epoch": 0.04, "learning_rate": 1.9996598128260235e-05, "loss": 1.2825, "step": 572 }, { "epoch": 0.04, "learning_rate": 1.999654166987747e-05, "loss": 1.2915, "step": 573 }, { "epoch": 0.04, "learning_rate": 1.9996484746924238e-05, "loss": 1.2553, "step": 574 }, { "epoch": 0.04, "learning_rate": 1.9996427359403192e-05, "loss": 1.2729, "step": 575 }, { "epoch": 0.04, "learning_rate": 1.9996369507316996e-05, "loss": 1.2794, "step": 576 }, { "epoch": 0.04, "learning_rate": 1.9996311190668333e-05, "loss": 1.2751, "step": 577 }, { "epoch": 0.04, "learning_rate": 1.9996252409459924e-05, "loss": 1.2751, "step": 578 }, { "epoch": 0.04, "learning_rate": 1.999619316369449e-05, "loss": 1.2517, "step": 579 }, { "epoch": 0.04, "learning_rate": 1.999613345337479e-05, "loss": 1.2446, "step": 580 }, { "epoch": 0.04, "learning_rate": 1.9996073278503596e-05, "loss": 1.2944, "step": 581 }, { "epoch": 0.04, "learning_rate": 1.999601263908371e-05, "loss": 1.2454, "step": 582 }, { "epoch": 0.04, "learning_rate": 1.9995951535117942e-05, "loss": 1.2839, "step": 583 }, { "epoch": 0.04, "learning_rate": 1.999588996660914e-05, "loss": 1.2596, "step": 584 }, { "epoch": 0.04, "learning_rate": 1.9995827933560158e-05, "loss": 1.283, "step": 585 }, { "epoch": 0.04, "learning_rate": 1.9995765435973886e-05, "loss": 1.2933, "step": 586 }, { "epoch": 0.04, "learning_rate": 1.9995702473853224e-05, "loss": 1.2497, "step": 587 }, { "epoch": 0.04, "learning_rate": 1.9995639047201093e-05, "loss": 1.2596, "step": 588 }, { "epoch": 0.04, "learning_rate": 1.9995575156020454e-05, "loss": 1.2553, "step": 589 }, { "epoch": 0.04, "learning_rate": 1.9995510800314265e-05, "loss": 1.2136, "step": 590 }, { "epoch": 0.04, "learning_rate": 1.999544598008552e-05, "loss": 1.271, "step": 591 }, { "epoch": 0.04, "learning_rate": 1.999538069533723e-05, "loss": 1.2878, "step": 592 }, { "epoch": 0.04, "learning_rate": 1.9995314946072437e-05, "loss": 1.309, "step": 593 }, { "epoch": 0.04, "learning_rate": 1.999524873229419e-05, "loss": 1.2798, "step": 594 }, { "epoch": 0.04, "learning_rate": 1.999518205400556e-05, "loss": 1.2539, "step": 595 }, { "epoch": 0.04, "learning_rate": 1.9995114911209655e-05, "loss": 1.2603, "step": 596 }, { "epoch": 0.04, "learning_rate": 1.9995047303909593e-05, "loss": 1.2668, "step": 597 }, { "epoch": 0.04, "learning_rate": 1.999497923210852e-05, "loss": 1.2969, "step": 598 }, { "epoch": 0.04, "learning_rate": 1.9994910695809588e-05, "loss": 1.2794, "step": 599 }, { "epoch": 0.04, "learning_rate": 1.999484169501599e-05, "loss": 1.2502, "step": 600 }, { "epoch": 0.04, "learning_rate": 1.9994772229730938e-05, "loss": 1.2821, "step": 601 }, { "epoch": 0.04, "learning_rate": 1.9994702299957645e-05, "loss": 1.215, "step": 602 }, { "epoch": 0.04, "learning_rate": 1.999463190569938e-05, "loss": 1.2443, "step": 603 }, { "epoch": 0.04, "learning_rate": 1.9994561046959395e-05, "loss": 1.2819, "step": 604 }, { "epoch": 0.04, "learning_rate": 1.9994489723740996e-05, "loss": 1.2882, "step": 605 }, { "epoch": 0.04, "learning_rate": 1.9994417936047497e-05, "loss": 1.2582, "step": 606 }, { "epoch": 0.04, "learning_rate": 1.9994345683882227e-05, "loss": 1.2587, "step": 607 }, { "epoch": 0.04, "learning_rate": 1.999427296724855e-05, "loss": 1.2282, "step": 608 }, { "epoch": 0.04, "learning_rate": 1.9994199786149845e-05, "loss": 1.264, "step": 609 }, { "epoch": 0.04, "learning_rate": 1.999412614058951e-05, "loss": 1.2961, "step": 610 }, { "epoch": 0.04, "learning_rate": 1.999405203057097e-05, "loss": 1.2826, "step": 611 }, { "epoch": 0.04, "learning_rate": 1.9993977456097663e-05, "loss": 1.2867, "step": 612 }, { "epoch": 0.04, "learning_rate": 1.9993902417173065e-05, "loss": 1.2919, "step": 613 }, { "epoch": 0.04, "learning_rate": 1.9993826913800654e-05, "loss": 1.2563, "step": 614 }, { "epoch": 0.04, "learning_rate": 1.9993750945983944e-05, "loss": 1.2253, "step": 615 }, { "epoch": 0.04, "learning_rate": 1.9993674513726465e-05, "loss": 1.2404, "step": 616 }, { "epoch": 0.04, "learning_rate": 1.9993597617031772e-05, "loss": 1.2322, "step": 617 }, { "epoch": 0.04, "learning_rate": 1.9993520255903428e-05, "loss": 1.2434, "step": 618 }, { "epoch": 0.04, "learning_rate": 1.999344243034504e-05, "loss": 1.2538, "step": 619 }, { "epoch": 0.04, "learning_rate": 1.999336414036022e-05, "loss": 1.2542, "step": 620 }, { "epoch": 0.04, "learning_rate": 1.9993285385952605e-05, "loss": 1.2495, "step": 621 }, { "epoch": 0.04, "learning_rate": 1.9993206167125855e-05, "loss": 1.2357, "step": 622 }, { "epoch": 0.04, "learning_rate": 1.9993126483883656e-05, "loss": 1.2314, "step": 623 }, { "epoch": 0.04, "learning_rate": 1.9993046336229707e-05, "loss": 1.2199, "step": 624 }, { "epoch": 0.04, "learning_rate": 1.9992965724167736e-05, "loss": 1.2609, "step": 625 }, { "epoch": 0.04, "learning_rate": 1.9992884647701483e-05, "loss": 1.24, "step": 626 }, { "epoch": 0.04, "learning_rate": 1.9992803106834718e-05, "loss": 1.2548, "step": 627 }, { "epoch": 0.04, "learning_rate": 1.9992721101571238e-05, "loss": 1.2138, "step": 628 }, { "epoch": 0.04, "learning_rate": 1.9992638631914845e-05, "loss": 1.2632, "step": 629 }, { "epoch": 0.04, "learning_rate": 1.9992555697869377e-05, "loss": 1.2449, "step": 630 }, { "epoch": 0.04, "learning_rate": 1.9992472299438685e-05, "loss": 1.2491, "step": 631 }, { "epoch": 0.04, "learning_rate": 1.9992388436626645e-05, "loss": 1.2184, "step": 632 }, { "epoch": 0.04, "learning_rate": 1.9992304109437157e-05, "loss": 1.2232, "step": 633 }, { "epoch": 0.04, "learning_rate": 1.9992219317874138e-05, "loss": 1.2272, "step": 634 }, { "epoch": 0.04, "learning_rate": 1.9992134061941528e-05, "loss": 1.2475, "step": 635 }, { "epoch": 0.04, "learning_rate": 1.999204834164329e-05, "loss": 1.2374, "step": 636 }, { "epoch": 0.04, "learning_rate": 1.9991962156983412e-05, "loss": 1.2575, "step": 637 }, { "epoch": 0.04, "learning_rate": 1.999187550796589e-05, "loss": 1.1849, "step": 638 }, { "epoch": 0.04, "learning_rate": 1.999178839459476e-05, "loss": 1.275, "step": 639 }, { "epoch": 0.04, "learning_rate": 1.9991700816874062e-05, "loss": 1.2179, "step": 640 }, { "epoch": 0.04, "learning_rate": 1.999161277480787e-05, "loss": 1.2545, "step": 641 }, { "epoch": 0.04, "learning_rate": 1.9991524268400278e-05, "loss": 1.1753, "step": 642 }, { "epoch": 0.04, "learning_rate": 1.9991435297655396e-05, "loss": 1.2162, "step": 643 }, { "epoch": 0.04, "learning_rate": 1.999134586257736e-05, "loss": 1.2514, "step": 644 }, { "epoch": 0.04, "learning_rate": 1.9991255963170326e-05, "loss": 1.2174, "step": 645 }, { "epoch": 0.04, "learning_rate": 1.9991165599438475e-05, "loss": 1.2577, "step": 646 }, { "epoch": 0.04, "learning_rate": 1.9991074771385998e-05, "loss": 1.2567, "step": 647 }, { "epoch": 0.04, "learning_rate": 1.9990983479017124e-05, "loss": 1.2404, "step": 648 }, { "epoch": 0.04, "learning_rate": 1.9990891722336095e-05, "loss": 1.244, "step": 649 }, { "epoch": 0.04, "learning_rate": 1.999079950134717e-05, "loss": 1.2175, "step": 650 }, { "epoch": 0.04, "learning_rate": 1.999070681605464e-05, "loss": 1.283, "step": 651 }, { "epoch": 0.04, "learning_rate": 1.999061366646281e-05, "loss": 1.1889, "step": 652 }, { "epoch": 0.04, "learning_rate": 1.9990520052576012e-05, "loss": 1.2426, "step": 653 }, { "epoch": 0.04, "learning_rate": 1.999042597439859e-05, "loss": 1.2314, "step": 654 }, { "epoch": 0.04, "learning_rate": 1.9990331431934923e-05, "loss": 1.3008, "step": 655 }, { "epoch": 0.04, "learning_rate": 1.99902364251894e-05, "loss": 1.2394, "step": 656 }, { "epoch": 0.04, "learning_rate": 1.999014095416644e-05, "loss": 1.2559, "step": 657 }, { "epoch": 0.04, "learning_rate": 1.9990045018870474e-05, "loss": 1.2295, "step": 658 }, { "epoch": 0.04, "learning_rate": 1.998994861930597e-05, "loss": 1.2081, "step": 659 }, { "epoch": 0.04, "learning_rate": 1.99898517554774e-05, "loss": 1.2117, "step": 660 }, { "epoch": 0.04, "learning_rate": 1.9989754427389265e-05, "loss": 1.2505, "step": 661 }, { "epoch": 0.04, "learning_rate": 1.9989656635046092e-05, "loss": 1.1978, "step": 662 }, { "epoch": 0.04, "learning_rate": 1.9989558378452428e-05, "loss": 1.2589, "step": 663 }, { "epoch": 0.04, "learning_rate": 1.9989459657612836e-05, "loss": 1.1817, "step": 664 }, { "epoch": 0.04, "learning_rate": 1.99893604725319e-05, "loss": 1.2566, "step": 665 }, { "epoch": 0.04, "learning_rate": 1.9989260823214238e-05, "loss": 1.2212, "step": 666 }, { "epoch": 0.04, "learning_rate": 1.998916070966447e-05, "loss": 1.2159, "step": 667 }, { "epoch": 0.04, "learning_rate": 1.998906013188726e-05, "loss": 1.2207, "step": 668 }, { "epoch": 0.04, "learning_rate": 1.9988959089887274e-05, "loss": 1.1901, "step": 669 }, { "epoch": 0.04, "learning_rate": 1.998885758366921e-05, "loss": 1.191, "step": 670 }, { "epoch": 0.04, "learning_rate": 1.998875561323779e-05, "loss": 1.2165, "step": 671 }, { "epoch": 0.04, "learning_rate": 1.998865317859775e-05, "loss": 1.209, "step": 672 }, { "epoch": 0.04, "learning_rate": 1.9988550279753848e-05, "loss": 1.201, "step": 673 }, { "epoch": 0.04, "learning_rate": 1.9988446916710863e-05, "loss": 1.2176, "step": 674 }, { "epoch": 0.04, "learning_rate": 1.998834308947361e-05, "loss": 1.2478, "step": 675 }, { "epoch": 0.04, "learning_rate": 1.99882387980469e-05, "loss": 1.2379, "step": 676 }, { "epoch": 0.05, "learning_rate": 1.998813404243559e-05, "loss": 1.2299, "step": 677 }, { "epoch": 0.05, "learning_rate": 1.998802882264455e-05, "loss": 1.2269, "step": 678 }, { "epoch": 0.05, "learning_rate": 1.998792313867866e-05, "loss": 1.2155, "step": 679 }, { "epoch": 0.05, "learning_rate": 1.998781699054284e-05, "loss": 1.1827, "step": 680 }, { "epoch": 0.05, "learning_rate": 1.9987710378242016e-05, "loss": 1.2264, "step": 681 }, { "epoch": 0.05, "learning_rate": 1.998760330178115e-05, "loss": 1.1613, "step": 682 }, { "epoch": 0.05, "learning_rate": 1.998749576116521e-05, "loss": 1.2187, "step": 683 }, { "epoch": 0.05, "learning_rate": 1.9987387756399204e-05, "loss": 1.1885, "step": 684 }, { "epoch": 0.05, "learning_rate": 1.9987279287488143e-05, "loss": 1.2213, "step": 685 }, { "epoch": 0.05, "learning_rate": 1.998717035443707e-05, "loss": 1.2027, "step": 686 }, { "epoch": 0.05, "learning_rate": 1.9987060957251047e-05, "loss": 1.2245, "step": 687 }, { "epoch": 0.05, "learning_rate": 1.9986951095935158e-05, "loss": 1.2182, "step": 688 }, { "epoch": 0.05, "learning_rate": 1.9986840770494514e-05, "loss": 1.2318, "step": 689 }, { "epoch": 0.05, "learning_rate": 1.9986729980934235e-05, "loss": 1.2418, "step": 690 }, { "epoch": 0.05, "learning_rate": 1.998661872725947e-05, "loss": 1.2079, "step": 691 }, { "epoch": 0.05, "learning_rate": 1.9986507009475397e-05, "loss": 1.2453, "step": 692 }, { "epoch": 0.05, "learning_rate": 1.99863948275872e-05, "loss": 1.1807, "step": 693 }, { "epoch": 0.05, "learning_rate": 1.9986282181600092e-05, "loss": 1.1903, "step": 694 }, { "epoch": 0.05, "learning_rate": 1.9986169071519315e-05, "loss": 1.2085, "step": 695 }, { "epoch": 0.05, "learning_rate": 1.998605549735012e-05, "loss": 1.2256, "step": 696 }, { "epoch": 0.05, "learning_rate": 1.9985941459097793e-05, "loss": 1.2019, "step": 697 }, { "epoch": 0.05, "learning_rate": 1.998582695676762e-05, "loss": 1.1578, "step": 698 }, { "epoch": 0.05, "learning_rate": 1.9985711990364933e-05, "loss": 1.1926, "step": 699 }, { "epoch": 0.05, "learning_rate": 1.998559655989507e-05, "loss": 1.2627, "step": 700 }, { "epoch": 0.05, "learning_rate": 1.99854806653634e-05, "loss": 1.205, "step": 701 }, { "epoch": 0.05, "learning_rate": 1.9985364306775305e-05, "loss": 1.1875, "step": 702 }, { "epoch": 0.05, "learning_rate": 1.998524748413619e-05, "loss": 1.2002, "step": 703 }, { "epoch": 0.05, "learning_rate": 1.9985130197451493e-05, "loss": 1.2108, "step": 704 }, { "epoch": 0.05, "learning_rate": 1.9985012446726656e-05, "loss": 1.2414, "step": 705 }, { "epoch": 0.05, "learning_rate": 1.9984894231967155e-05, "loss": 1.2431, "step": 706 }, { "epoch": 0.05, "learning_rate": 1.9984775553178485e-05, "loss": 1.221, "step": 707 }, { "epoch": 0.05, "learning_rate": 1.9984656410366157e-05, "loss": 1.2158, "step": 708 }, { "epoch": 0.05, "learning_rate": 1.998453680353571e-05, "loss": 1.2264, "step": 709 }, { "epoch": 0.05, "learning_rate": 1.998441673269271e-05, "loss": 1.1928, "step": 710 }, { "epoch": 0.05, "learning_rate": 1.9984296197842722e-05, "loss": 1.2324, "step": 711 }, { "epoch": 0.05, "learning_rate": 1.9984175198991362e-05, "loss": 1.207, "step": 712 }, { "epoch": 0.05, "learning_rate": 1.9984053736144244e-05, "loss": 1.252, "step": 713 }, { "epoch": 0.05, "learning_rate": 1.9983931809307016e-05, "loss": 1.1964, "step": 714 }, { "epoch": 0.05, "learning_rate": 1.998380941848535e-05, "loss": 1.2364, "step": 715 }, { "epoch": 0.05, "learning_rate": 1.998368656368492e-05, "loss": 1.2045, "step": 716 }, { "epoch": 0.05, "learning_rate": 1.9983563244911447e-05, "loss": 1.2645, "step": 717 }, { "epoch": 0.05, "learning_rate": 1.998343946217066e-05, "loss": 1.1889, "step": 718 }, { "epoch": 0.05, "learning_rate": 1.998331521546831e-05, "loss": 1.1966, "step": 719 }, { "epoch": 0.05, "learning_rate": 1.9983190504810167e-05, "loss": 1.2167, "step": 720 }, { "epoch": 0.05, "learning_rate": 1.9983065330202035e-05, "loss": 1.203, "step": 721 }, { "epoch": 0.05, "learning_rate": 1.9982939691649725e-05, "loss": 1.1993, "step": 722 }, { "epoch": 0.05, "learning_rate": 1.998281358915908e-05, "loss": 1.2019, "step": 723 }, { "epoch": 0.05, "learning_rate": 1.998268702273596e-05, "loss": 1.2446, "step": 724 }, { "epoch": 0.05, "learning_rate": 1.998255999238624e-05, "loss": 1.1873, "step": 725 }, { "epoch": 0.05, "learning_rate": 1.9982432498115832e-05, "loss": 1.2085, "step": 726 }, { "epoch": 0.05, "learning_rate": 1.9982304539930657e-05, "loss": 1.2026, "step": 727 }, { "epoch": 0.05, "learning_rate": 1.9982176117836663e-05, "loss": 1.1902, "step": 728 }, { "epoch": 0.05, "learning_rate": 1.9982047231839817e-05, "loss": 1.2483, "step": 729 }, { "epoch": 0.05, "learning_rate": 1.9981917881946113e-05, "loss": 1.1787, "step": 730 }, { "epoch": 0.05, "learning_rate": 1.9981788068161557e-05, "loss": 1.2422, "step": 731 }, { "epoch": 0.05, "learning_rate": 1.9981657790492178e-05, "loss": 1.2233, "step": 732 }, { "epoch": 0.05, "learning_rate": 1.9981527048944043e-05, "loss": 1.1935, "step": 733 }, { "epoch": 0.05, "learning_rate": 1.998139584352322e-05, "loss": 1.2151, "step": 734 }, { "epoch": 0.05, "learning_rate": 1.9981264174235805e-05, "loss": 1.1956, "step": 735 }, { "epoch": 0.05, "learning_rate": 1.998113204108792e-05, "loss": 1.1977, "step": 736 }, { "epoch": 0.05, "learning_rate": 1.9980999444085704e-05, "loss": 1.1673, "step": 737 }, { "epoch": 0.05, "learning_rate": 1.998086638323532e-05, "loss": 1.1799, "step": 738 }, { "epoch": 0.05, "learning_rate": 1.9980732858542957e-05, "loss": 1.2269, "step": 739 }, { "epoch": 0.05, "learning_rate": 1.9980598870014812e-05, "loss": 1.1463, "step": 740 }, { "epoch": 0.05, "learning_rate": 1.9980464417657118e-05, "loss": 1.1757, "step": 741 }, { "epoch": 0.05, "learning_rate": 1.9980329501476117e-05, "loss": 1.1769, "step": 742 }, { "epoch": 0.05, "learning_rate": 1.9980194121478083e-05, "loss": 1.2121, "step": 743 }, { "epoch": 0.05, "learning_rate": 1.998005827766931e-05, "loss": 1.1857, "step": 744 }, { "epoch": 0.05, "learning_rate": 1.9979921970056106e-05, "loss": 1.1886, "step": 745 }, { "epoch": 0.05, "learning_rate": 1.997978519864481e-05, "loss": 1.1763, "step": 746 }, { "epoch": 0.05, "learning_rate": 1.9979647963441776e-05, "loss": 1.2433, "step": 747 }, { "epoch": 0.05, "learning_rate": 1.9979510264453378e-05, "loss": 1.1961, "step": 748 }, { "epoch": 0.05, "learning_rate": 1.9979372101686025e-05, "loss": 1.1967, "step": 749 }, { "epoch": 0.05, "learning_rate": 1.9979233475146127e-05, "loss": 1.253, "step": 750 }, { "epoch": 0.05, "learning_rate": 1.9979094384840133e-05, "loss": 1.2124, "step": 751 }, { "epoch": 0.05, "learning_rate": 1.9978954830774506e-05, "loss": 1.1494, "step": 752 }, { "epoch": 0.05, "learning_rate": 1.9978814812955732e-05, "loss": 1.1759, "step": 753 }, { "epoch": 0.05, "learning_rate": 1.9978674331390315e-05, "loss": 1.1917, "step": 754 }, { "epoch": 0.05, "learning_rate": 1.997853338608479e-05, "loss": 1.1743, "step": 755 }, { "epoch": 0.05, "learning_rate": 1.9978391977045698e-05, "loss": 1.2004, "step": 756 }, { "epoch": 0.05, "learning_rate": 1.9978250104279617e-05, "loss": 1.2044, "step": 757 }, { "epoch": 0.05, "learning_rate": 1.9978107767793137e-05, "loss": 1.1782, "step": 758 }, { "epoch": 0.05, "learning_rate": 1.9977964967592877e-05, "loss": 1.1757, "step": 759 }, { "epoch": 0.05, "learning_rate": 1.997782170368547e-05, "loss": 1.1623, "step": 760 }, { "epoch": 0.05, "learning_rate": 1.9977677976077574e-05, "loss": 1.2248, "step": 761 }, { "epoch": 0.05, "learning_rate": 1.997753378477587e-05, "loss": 1.1897, "step": 762 }, { "epoch": 0.05, "learning_rate": 1.997738912978706e-05, "loss": 1.1852, "step": 763 }, { "epoch": 0.05, "learning_rate": 1.9977244011117866e-05, "loss": 1.1778, "step": 764 }, { "epoch": 0.05, "learning_rate": 1.9977098428775027e-05, "loss": 1.2263, "step": 765 }, { "epoch": 0.05, "learning_rate": 1.9976952382765313e-05, "loss": 1.1709, "step": 766 }, { "epoch": 0.05, "learning_rate": 1.9976805873095513e-05, "loss": 1.1806, "step": 767 }, { "epoch": 0.05, "learning_rate": 1.9976658899772434e-05, "loss": 1.1783, "step": 768 }, { "epoch": 0.05, "learning_rate": 1.9976511462802906e-05, "loss": 1.1721, "step": 769 }, { "epoch": 0.05, "learning_rate": 1.9976363562193778e-05, "loss": 1.1924, "step": 770 }, { "epoch": 0.05, "learning_rate": 1.997621519795193e-05, "loss": 1.2362, "step": 771 }, { "epoch": 0.05, "learning_rate": 1.997606637008425e-05, "loss": 1.1664, "step": 772 }, { "epoch": 0.05, "learning_rate": 1.997591707859766e-05, "loss": 1.1846, "step": 773 }, { "epoch": 0.05, "learning_rate": 1.9975767323499097e-05, "loss": 1.1669, "step": 774 }, { "epoch": 0.05, "learning_rate": 1.9975617104795515e-05, "loss": 1.2369, "step": 775 }, { "epoch": 0.05, "learning_rate": 1.9975466422493904e-05, "loss": 1.2132, "step": 776 }, { "epoch": 0.05, "learning_rate": 1.9975315276601258e-05, "loss": 1.1435, "step": 777 }, { "epoch": 0.05, "learning_rate": 1.997516366712461e-05, "loss": 1.1957, "step": 778 }, { "epoch": 0.05, "learning_rate": 1.9975011594070997e-05, "loss": 1.2092, "step": 779 }, { "epoch": 0.05, "learning_rate": 1.9974859057447494e-05, "loss": 1.1701, "step": 780 }, { "epoch": 0.05, "learning_rate": 1.9974706057261182e-05, "loss": 1.2078, "step": 781 }, { "epoch": 0.05, "learning_rate": 1.997455259351918e-05, "loss": 1.1515, "step": 782 }, { "epoch": 0.05, "learning_rate": 1.9974398666228612e-05, "loss": 1.1712, "step": 783 }, { "epoch": 0.05, "learning_rate": 1.997424427539664e-05, "loss": 1.1645, "step": 784 }, { "epoch": 0.05, "learning_rate": 1.9974089421030428e-05, "loss": 1.1782, "step": 785 }, { "epoch": 0.05, "learning_rate": 1.997393410313718e-05, "loss": 1.1728, "step": 786 }, { "epoch": 0.05, "learning_rate": 1.9973778321724115e-05, "loss": 1.1717, "step": 787 }, { "epoch": 0.05, "learning_rate": 1.997362207679847e-05, "loss": 1.1878, "step": 788 }, { "epoch": 0.05, "learning_rate": 1.9973465368367506e-05, "loss": 1.2067, "step": 789 }, { "epoch": 0.05, "learning_rate": 1.9973308196438507e-05, "loss": 1.2272, "step": 790 }, { "epoch": 0.05, "learning_rate": 1.9973150561018778e-05, "loss": 1.1798, "step": 791 }, { "epoch": 0.05, "learning_rate": 1.997299246211564e-05, "loss": 1.148, "step": 792 }, { "epoch": 0.05, "learning_rate": 1.9972833899736445e-05, "loss": 1.158, "step": 793 }, { "epoch": 0.05, "learning_rate": 1.9972674873888566e-05, "loss": 1.2107, "step": 794 }, { "epoch": 0.05, "learning_rate": 1.9972515384579382e-05, "loss": 1.1847, "step": 795 }, { "epoch": 0.05, "learning_rate": 1.997235543181631e-05, "loss": 1.1736, "step": 796 }, { "epoch": 0.05, "learning_rate": 1.9972195015606787e-05, "loss": 1.139, "step": 797 }, { "epoch": 0.05, "learning_rate": 1.997203413595827e-05, "loss": 1.1832, "step": 798 }, { "epoch": 0.05, "learning_rate": 1.9971872792878226e-05, "loss": 1.1566, "step": 799 }, { "epoch": 0.05, "learning_rate": 1.997171098637416e-05, "loss": 1.1964, "step": 800 }, { "epoch": 0.05, "learning_rate": 1.9971548716453587e-05, "loss": 1.2343, "step": 801 }, { "epoch": 0.05, "learning_rate": 1.9971385983124053e-05, "loss": 1.1995, "step": 802 }, { "epoch": 0.05, "learning_rate": 1.997122278639312e-05, "loss": 1.1891, "step": 803 }, { "epoch": 0.05, "learning_rate": 1.997105912626837e-05, "loss": 1.1965, "step": 804 }, { "epoch": 0.05, "learning_rate": 1.9970895002757413e-05, "loss": 1.2396, "step": 805 }, { "epoch": 0.05, "learning_rate": 1.9970730415867868e-05, "loss": 1.1433, "step": 806 }, { "epoch": 0.05, "learning_rate": 1.997056536560739e-05, "loss": 1.2088, "step": 807 }, { "epoch": 0.05, "learning_rate": 1.9970399851983653e-05, "loss": 1.1163, "step": 808 }, { "epoch": 0.05, "learning_rate": 1.9970233875004344e-05, "loss": 1.2141, "step": 809 }, { "epoch": 0.05, "learning_rate": 1.9970067434677173e-05, "loss": 1.1926, "step": 810 }, { "epoch": 0.05, "learning_rate": 1.996990053100988e-05, "loss": 1.2089, "step": 811 }, { "epoch": 0.05, "learning_rate": 1.9969733164010224e-05, "loss": 1.1829, "step": 812 }, { "epoch": 0.05, "learning_rate": 1.9969565333685977e-05, "loss": 1.1789, "step": 813 }, { "epoch": 0.05, "learning_rate": 1.996939704004494e-05, "loss": 1.1444, "step": 814 }, { "epoch": 0.05, "learning_rate": 1.9969228283094937e-05, "loss": 1.1753, "step": 815 }, { "epoch": 0.05, "learning_rate": 1.9969059062843806e-05, "loss": 1.1513, "step": 816 }, { "epoch": 0.05, "learning_rate": 1.996888937929942e-05, "loss": 1.172, "step": 817 }, { "epoch": 0.05, "learning_rate": 1.9968719232469652e-05, "loss": 1.1351, "step": 818 }, { "epoch": 0.05, "learning_rate": 1.996854862236242e-05, "loss": 1.1631, "step": 819 }, { "epoch": 0.05, "learning_rate": 1.996837754898565e-05, "loss": 1.2062, "step": 820 }, { "epoch": 0.05, "learning_rate": 1.996820601234729e-05, "loss": 1.1827, "step": 821 }, { "epoch": 0.05, "learning_rate": 1.996803401245531e-05, "loss": 1.1652, "step": 822 }, { "epoch": 0.05, "learning_rate": 1.996786154931771e-05, "loss": 1.1947, "step": 823 }, { "epoch": 0.05, "learning_rate": 1.99676886229425e-05, "loss": 1.1324, "step": 824 }, { "epoch": 0.05, "learning_rate": 1.996751523333772e-05, "loss": 1.2151, "step": 825 }, { "epoch": 0.05, "learning_rate": 1.996734138051142e-05, "loss": 1.1431, "step": 826 }, { "epoch": 0.06, "learning_rate": 1.996716706447169e-05, "loss": 1.2058, "step": 827 }, { "epoch": 0.06, "learning_rate": 1.9966992285226627e-05, "loss": 1.1856, "step": 828 }, { "epoch": 0.06, "learning_rate": 1.996681704278435e-05, "loss": 1.1644, "step": 829 }, { "epoch": 0.06, "learning_rate": 1.9966641337153006e-05, "loss": 1.1574, "step": 830 }, { "epoch": 0.06, "learning_rate": 1.996646516834076e-05, "loss": 1.1916, "step": 831 }, { "epoch": 0.06, "learning_rate": 1.9966288536355803e-05, "loss": 1.2038, "step": 832 }, { "epoch": 0.06, "learning_rate": 1.996611144120634e-05, "loss": 1.1928, "step": 833 }, { "epoch": 0.06, "learning_rate": 1.9965933882900596e-05, "loss": 1.1402, "step": 834 }, { "epoch": 0.06, "learning_rate": 1.996575586144683e-05, "loss": 1.1889, "step": 835 }, { "epoch": 0.06, "learning_rate": 1.9965577376853315e-05, "loss": 1.1843, "step": 836 }, { "epoch": 0.06, "learning_rate": 1.9965398429128343e-05, "loss": 1.1653, "step": 837 }, { "epoch": 0.06, "learning_rate": 1.996521901828023e-05, "loss": 1.1629, "step": 838 }, { "epoch": 0.06, "learning_rate": 1.996503914431732e-05, "loss": 1.1217, "step": 839 }, { "epoch": 0.06, "learning_rate": 1.9964858807247963e-05, "loss": 1.1446, "step": 840 }, { "epoch": 0.06, "learning_rate": 1.9964678007080544e-05, "loss": 1.2107, "step": 841 }, { "epoch": 0.06, "learning_rate": 1.9964496743823464e-05, "loss": 1.1746, "step": 842 }, { "epoch": 0.06, "learning_rate": 1.996431501748515e-05, "loss": 1.1839, "step": 843 }, { "epoch": 0.06, "learning_rate": 1.9964132828074048e-05, "loss": 1.1913, "step": 844 }, { "epoch": 0.06, "learning_rate": 1.996395017559862e-05, "loss": 1.1594, "step": 845 }, { "epoch": 0.06, "learning_rate": 1.996376706006736e-05, "loss": 1.1898, "step": 846 }, { "epoch": 0.06, "learning_rate": 1.9963583481488774e-05, "loss": 1.1857, "step": 847 }, { "epoch": 0.06, "learning_rate": 1.9963399439871392e-05, "loss": 1.1088, "step": 848 }, { "epoch": 0.06, "learning_rate": 1.996321493522377e-05, "loss": 1.1923, "step": 849 }, { "epoch": 0.06, "learning_rate": 1.996302996755448e-05, "loss": 1.1513, "step": 850 }, { "epoch": 0.06, "learning_rate": 1.9962844536872127e-05, "loss": 1.1644, "step": 851 }, { "epoch": 0.06, "learning_rate": 1.9962658643185317e-05, "loss": 1.1734, "step": 852 }, { "epoch": 0.06, "learning_rate": 1.996247228650269e-05, "loss": 1.2378, "step": 853 }, { "epoch": 0.06, "learning_rate": 1.9962285466832914e-05, "loss": 1.1618, "step": 854 }, { "epoch": 0.06, "learning_rate": 1.996209818418467e-05, "loss": 1.2073, "step": 855 }, { "epoch": 0.06, "learning_rate": 1.9961910438566657e-05, "loss": 1.1473, "step": 856 }, { "epoch": 0.06, "learning_rate": 1.99617222299876e-05, "loss": 1.1579, "step": 857 }, { "epoch": 0.06, "learning_rate": 1.9961533558456248e-05, "loss": 1.1569, "step": 858 }, { "epoch": 0.06, "learning_rate": 1.996134442398137e-05, "loss": 1.1592, "step": 859 }, { "epoch": 0.06, "learning_rate": 1.9961154826571754e-05, "loss": 1.1552, "step": 860 }, { "epoch": 0.06, "learning_rate": 1.996096476623621e-05, "loss": 1.2078, "step": 861 }, { "epoch": 0.06, "learning_rate": 1.9960774242983577e-05, "loss": 1.1784, "step": 862 }, { "epoch": 0.06, "learning_rate": 1.99605832568227e-05, "loss": 1.1944, "step": 863 }, { "epoch": 0.06, "learning_rate": 1.9960391807762462e-05, "loss": 1.1864, "step": 864 }, { "epoch": 0.06, "learning_rate": 1.9960199895811755e-05, "loss": 1.1856, "step": 865 }, { "epoch": 0.06, "learning_rate": 1.9960007520979504e-05, "loss": 1.1739, "step": 866 }, { "epoch": 0.06, "learning_rate": 1.995981468327464e-05, "loss": 1.2267, "step": 867 }, { "epoch": 0.06, "learning_rate": 1.9959621382706133e-05, "loss": 1.1486, "step": 868 }, { "epoch": 0.06, "learning_rate": 1.9959427619282963e-05, "loss": 1.2184, "step": 869 }, { "epoch": 0.06, "learning_rate": 1.995923339301414e-05, "loss": 1.1801, "step": 870 }, { "epoch": 0.06, "learning_rate": 1.9959038703908683e-05, "loss": 1.1921, "step": 871 }, { "epoch": 0.06, "learning_rate": 1.995884355197564e-05, "loss": 1.1877, "step": 872 }, { "epoch": 0.06, "learning_rate": 1.9958647937224085e-05, "loss": 1.1733, "step": 873 }, { "epoch": 0.06, "learning_rate": 1.9958451859663104e-05, "loss": 1.1751, "step": 874 }, { "epoch": 0.06, "learning_rate": 1.9958255319301814e-05, "loss": 1.1834, "step": 875 }, { "epoch": 0.06, "learning_rate": 1.995805831614935e-05, "loss": 1.1752, "step": 876 }, { "epoch": 0.06, "learning_rate": 1.995786085021486e-05, "loss": 1.1457, "step": 877 }, { "epoch": 0.06, "learning_rate": 1.9957662921507523e-05, "loss": 1.1242, "step": 878 }, { "epoch": 0.06, "learning_rate": 1.9957464530036543e-05, "loss": 1.1695, "step": 879 }, { "epoch": 0.06, "learning_rate": 1.9957265675811133e-05, "loss": 1.1577, "step": 880 }, { "epoch": 0.06, "learning_rate": 1.9957066358840544e-05, "loss": 1.1514, "step": 881 }, { "epoch": 0.06, "learning_rate": 1.9956866579134026e-05, "loss": 1.156, "step": 882 }, { "epoch": 0.06, "learning_rate": 1.995666633670087e-05, "loss": 1.2097, "step": 883 }, { "epoch": 0.06, "learning_rate": 1.9956465631550387e-05, "loss": 1.209, "step": 884 }, { "epoch": 0.06, "learning_rate": 1.9956264463691896e-05, "loss": 1.2097, "step": 885 }, { "epoch": 0.06, "learning_rate": 1.9956062833134746e-05, "loss": 1.1742, "step": 886 }, { "epoch": 0.06, "learning_rate": 1.9955860739888315e-05, "loss": 1.2153, "step": 887 }, { "epoch": 0.06, "learning_rate": 1.995565818396199e-05, "loss": 1.1358, "step": 888 }, { "epoch": 0.06, "learning_rate": 1.9955455165365184e-05, "loss": 1.1712, "step": 889 }, { "epoch": 0.06, "learning_rate": 1.9955251684107333e-05, "loss": 1.1383, "step": 890 }, { "epoch": 0.06, "learning_rate": 1.995504774019789e-05, "loss": 1.1351, "step": 891 }, { "epoch": 0.06, "learning_rate": 1.9954843333646337e-05, "loss": 1.1426, "step": 892 }, { "epoch": 0.06, "learning_rate": 1.9954638464462176e-05, "loss": 1.1821, "step": 893 }, { "epoch": 0.06, "learning_rate": 1.995443313265492e-05, "loss": 1.1257, "step": 894 }, { "epoch": 0.06, "learning_rate": 1.995422733823412e-05, "loss": 1.1703, "step": 895 }, { "epoch": 0.06, "learning_rate": 1.9954021081209334e-05, "loss": 1.1455, "step": 896 }, { "epoch": 0.06, "learning_rate": 1.995381436159015e-05, "loss": 1.1784, "step": 897 }, { "epoch": 0.06, "learning_rate": 1.995360717938617e-05, "loss": 1.122, "step": 898 }, { "epoch": 0.06, "learning_rate": 1.995339953460703e-05, "loss": 1.1929, "step": 899 }, { "epoch": 0.06, "learning_rate": 1.9953191427262375e-05, "loss": 1.1459, "step": 900 }, { "epoch": 0.06, "learning_rate": 1.9952982857361878e-05, "loss": 1.1584, "step": 901 }, { "epoch": 0.06, "learning_rate": 1.995277382491523e-05, "loss": 1.1421, "step": 902 }, { "epoch": 0.06, "learning_rate": 1.9952564329932148e-05, "loss": 1.2041, "step": 903 }, { "epoch": 0.06, "learning_rate": 1.995235437242237e-05, "loss": 1.1356, "step": 904 }, { "epoch": 0.06, "learning_rate": 1.9952143952395646e-05, "loss": 1.1683, "step": 905 }, { "epoch": 0.06, "learning_rate": 1.9951933069861758e-05, "loss": 1.1642, "step": 906 }, { "epoch": 0.06, "learning_rate": 1.995172172483051e-05, "loss": 1.1854, "step": 907 }, { "epoch": 0.06, "learning_rate": 1.9951509917311723e-05, "loss": 1.1801, "step": 908 }, { "epoch": 0.06, "learning_rate": 1.9951297647315235e-05, "loss": 1.1735, "step": 909 }, { "epoch": 0.06, "learning_rate": 1.9951084914850917e-05, "loss": 1.1837, "step": 910 }, { "epoch": 0.06, "learning_rate": 1.995087171992865e-05, "loss": 1.1823, "step": 911 }, { "epoch": 0.06, "learning_rate": 1.9950658062558347e-05, "loss": 1.1485, "step": 912 }, { "epoch": 0.06, "learning_rate": 1.9950443942749936e-05, "loss": 1.1974, "step": 913 }, { "epoch": 0.06, "learning_rate": 1.9950229360513364e-05, "loss": 1.1852, "step": 914 }, { "epoch": 0.06, "learning_rate": 1.995001431585861e-05, "loss": 1.1524, "step": 915 }, { "epoch": 0.06, "learning_rate": 1.994979880879566e-05, "loss": 1.1711, "step": 916 }, { "epoch": 0.06, "learning_rate": 1.9949582839334533e-05, "loss": 1.182, "step": 917 }, { "epoch": 0.06, "learning_rate": 1.994936640748527e-05, "loss": 1.1203, "step": 918 }, { "epoch": 0.06, "learning_rate": 1.9949149513257923e-05, "loss": 1.1264, "step": 919 }, { "epoch": 0.06, "learning_rate": 1.994893215666257e-05, "loss": 1.1574, "step": 920 }, { "epoch": 0.06, "learning_rate": 1.994871433770932e-05, "loss": 1.178, "step": 921 }, { "epoch": 0.06, "learning_rate": 1.9948496056408294e-05, "loss": 1.1297, "step": 922 }, { "epoch": 0.06, "learning_rate": 1.994827731276963e-05, "loss": 1.1804, "step": 923 }, { "epoch": 0.06, "learning_rate": 1.99480581068035e-05, "loss": 1.1788, "step": 924 }, { "epoch": 0.06, "learning_rate": 1.994783843852009e-05, "loss": 1.1761, "step": 925 }, { "epoch": 0.06, "learning_rate": 1.9947618307929607e-05, "loss": 1.1532, "step": 926 }, { "epoch": 0.06, "learning_rate": 1.9947397715042283e-05, "loss": 1.1825, "step": 927 }, { "epoch": 0.06, "learning_rate": 1.9947176659868365e-05, "loss": 1.1445, "step": 928 }, { "epoch": 0.06, "learning_rate": 1.9946955142418135e-05, "loss": 1.1374, "step": 929 }, { "epoch": 0.06, "learning_rate": 1.9946733162701877e-05, "loss": 1.1325, "step": 930 }, { "epoch": 0.06, "learning_rate": 1.9946510720729913e-05, "loss": 1.1775, "step": 931 }, { "epoch": 0.06, "learning_rate": 1.9946287816512582e-05, "loss": 1.1471, "step": 932 }, { "epoch": 0.06, "learning_rate": 1.9946064450060243e-05, "loss": 1.1669, "step": 933 }, { "epoch": 0.06, "learning_rate": 1.994584062138327e-05, "loss": 1.1541, "step": 934 }, { "epoch": 0.06, "learning_rate": 1.994561633049207e-05, "loss": 1.1736, "step": 935 }, { "epoch": 0.06, "learning_rate": 1.9945391577397067e-05, "loss": 1.1269, "step": 936 }, { "epoch": 0.06, "learning_rate": 1.9945166362108706e-05, "loss": 1.1717, "step": 937 }, { "epoch": 0.06, "learning_rate": 1.994494068463745e-05, "loss": 1.1133, "step": 938 }, { "epoch": 0.06, "learning_rate": 1.9944714544993792e-05, "loss": 1.1632, "step": 939 }, { "epoch": 0.06, "learning_rate": 1.994448794318824e-05, "loss": 1.1529, "step": 940 }, { "epoch": 0.06, "learning_rate": 1.994426087923132e-05, "loss": 1.1596, "step": 941 }, { "epoch": 0.06, "learning_rate": 1.9944033353133593e-05, "loss": 1.1286, "step": 942 }, { "epoch": 0.06, "learning_rate": 1.9943805364905623e-05, "loss": 1.1681, "step": 943 }, { "epoch": 0.06, "learning_rate": 1.994357691455801e-05, "loss": 1.1865, "step": 944 }, { "epoch": 0.06, "learning_rate": 1.9943348002101374e-05, "loss": 1.1871, "step": 945 }, { "epoch": 0.06, "learning_rate": 1.9943118627546346e-05, "loss": 1.1478, "step": 946 }, { "epoch": 0.06, "learning_rate": 1.9942888790903593e-05, "loss": 1.1901, "step": 947 }, { "epoch": 0.06, "learning_rate": 1.9942658492183795e-05, "loss": 1.1444, "step": 948 }, { "epoch": 0.06, "learning_rate": 1.994242773139765e-05, "loss": 1.2168, "step": 949 }, { "epoch": 0.06, "learning_rate": 1.9942196508555887e-05, "loss": 1.1526, "step": 950 }, { "epoch": 0.06, "learning_rate": 1.994196482366925e-05, "loss": 1.1799, "step": 951 }, { "epoch": 0.06, "learning_rate": 1.9941732676748503e-05, "loss": 1.1059, "step": 952 }, { "epoch": 0.06, "learning_rate": 1.9941500067804444e-05, "loss": 1.154, "step": 953 }, { "epoch": 0.06, "learning_rate": 1.994126699684787e-05, "loss": 1.1339, "step": 954 }, { "epoch": 0.06, "learning_rate": 1.994103346388962e-05, "loss": 1.1718, "step": 955 }, { "epoch": 0.06, "learning_rate": 1.994079946894055e-05, "loss": 1.1159, "step": 956 }, { "epoch": 0.06, "learning_rate": 1.9940565012011528e-05, "loss": 1.1922, "step": 957 }, { "epoch": 0.06, "learning_rate": 1.9940330093113453e-05, "loss": 1.1533, "step": 958 }, { "epoch": 0.06, "learning_rate": 1.994009471225724e-05, "loss": 1.1685, "step": 959 }, { "epoch": 0.06, "learning_rate": 1.9939858869453832e-05, "loss": 1.1655, "step": 960 }, { "epoch": 0.06, "learning_rate": 1.9939622564714187e-05, "loss": 1.1674, "step": 961 }, { "epoch": 0.06, "learning_rate": 1.9939385798049286e-05, "loss": 1.1611, "step": 962 }, { "epoch": 0.06, "learning_rate": 1.9939148569470132e-05, "loss": 1.1706, "step": 963 }, { "epoch": 0.06, "learning_rate": 1.9938910878987754e-05, "loss": 1.1688, "step": 964 }, { "epoch": 0.06, "learning_rate": 1.9938672726613193e-05, "loss": 1.1409, "step": 965 }, { "epoch": 0.06, "learning_rate": 1.993843411235752e-05, "loss": 1.1625, "step": 966 }, { "epoch": 0.06, "learning_rate": 1.9938195036231825e-05, "loss": 1.1354, "step": 967 }, { "epoch": 0.06, "learning_rate": 1.9937955498247214e-05, "loss": 1.1335, "step": 968 }, { "epoch": 0.06, "learning_rate": 1.993771549841482e-05, "loss": 1.185, "step": 969 }, { "epoch": 0.06, "learning_rate": 1.9937475036745802e-05, "loss": 1.1569, "step": 970 }, { "epoch": 0.06, "learning_rate": 1.993723411325133e-05, "loss": 1.1033, "step": 971 }, { "epoch": 0.06, "learning_rate": 1.9936992727942602e-05, "loss": 1.1692, "step": 972 }, { "epoch": 0.06, "learning_rate": 1.9936750880830838e-05, "loss": 1.18, "step": 973 }, { "epoch": 0.06, "learning_rate": 1.993650857192727e-05, "loss": 1.1174, "step": 974 }, { "epoch": 0.06, "learning_rate": 1.9936265801243168e-05, "loss": 1.1796, "step": 975 }, { "epoch": 0.06, "learning_rate": 1.993602256878981e-05, "loss": 1.1855, "step": 976 }, { "epoch": 0.07, "learning_rate": 1.99357788745785e-05, "loss": 1.1446, "step": 977 }, { "epoch": 0.07, "learning_rate": 1.9935534718620563e-05, "loss": 1.1285, "step": 978 }, { "epoch": 0.07, "learning_rate": 1.993529010092734e-05, "loss": 1.1709, "step": 979 }, { "epoch": 0.07, "learning_rate": 1.993504502151021e-05, "loss": 1.1277, "step": 980 }, { "epoch": 0.07, "learning_rate": 1.993479948038056e-05, "loss": 1.1549, "step": 981 }, { "epoch": 0.07, "learning_rate": 1.9934553477549795e-05, "loss": 1.1705, "step": 982 }, { "epoch": 0.07, "learning_rate": 1.9934307013029355e-05, "loss": 1.203, "step": 983 }, { "epoch": 0.07, "learning_rate": 1.993406008683069e-05, "loss": 1.1347, "step": 984 }, { "epoch": 0.07, "learning_rate": 1.9933812698965274e-05, "loss": 1.1595, "step": 985 }, { "epoch": 0.07, "learning_rate": 1.9933564849444602e-05, "loss": 1.1432, "step": 986 }, { "epoch": 0.07, "learning_rate": 1.99333165382802e-05, "loss": 1.1296, "step": 987 }, { "epoch": 0.07, "learning_rate": 1.9933067765483604e-05, "loss": 1.1141, "step": 988 }, { "epoch": 0.07, "learning_rate": 1.9932818531066377e-05, "loss": 1.1582, "step": 989 }, { "epoch": 0.07, "learning_rate": 1.9932568835040094e-05, "loss": 1.1143, "step": 990 }, { "epoch": 0.07, "learning_rate": 1.9932318677416366e-05, "loss": 1.1613, "step": 991 }, { "epoch": 0.07, "learning_rate": 1.993206805820682e-05, "loss": 1.1724, "step": 992 }, { "epoch": 0.07, "learning_rate": 1.99318169774231e-05, "loss": 1.16, "step": 993 }, { "epoch": 0.07, "learning_rate": 1.9931565435076872e-05, "loss": 1.1434, "step": 994 }, { "epoch": 0.07, "learning_rate": 1.9931313431179834e-05, "loss": 1.1867, "step": 995 }, { "epoch": 0.07, "learning_rate": 1.993106096574369e-05, "loss": 1.1516, "step": 996 }, { "epoch": 0.07, "learning_rate": 1.9930808038780174e-05, "loss": 1.1885, "step": 997 }, { "epoch": 0.07, "learning_rate": 1.9930554650301044e-05, "loss": 1.1181, "step": 998 }, { "epoch": 0.07, "learning_rate": 1.993030080031807e-05, "loss": 1.1715, "step": 999 }, { "epoch": 0.07, "learning_rate": 1.9930046488843056e-05, "loss": 1.1553, "step": 1000 }, { "epoch": 0.07, "learning_rate": 1.992979171588781e-05, "loss": 1.139, "step": 1001 }, { "epoch": 0.07, "learning_rate": 1.9929536481464188e-05, "loss": 1.1363, "step": 1002 }, { "epoch": 0.07, "learning_rate": 1.992928078558404e-05, "loss": 1.1057, "step": 1003 }, { "epoch": 0.07, "learning_rate": 1.992902462825925e-05, "loss": 1.1833, "step": 1004 }, { "epoch": 0.07, "learning_rate": 1.9928768009501725e-05, "loss": 1.1581, "step": 1005 }, { "epoch": 0.07, "learning_rate": 1.992851092932339e-05, "loss": 1.0857, "step": 1006 }, { "epoch": 0.07, "learning_rate": 1.992825338773619e-05, "loss": 1.1825, "step": 1007 }, { "epoch": 0.07, "learning_rate": 1.99279953847521e-05, "loss": 1.1725, "step": 1008 }, { "epoch": 0.07, "learning_rate": 1.9927736920383104e-05, "loss": 1.1384, "step": 1009 }, { "epoch": 0.07, "learning_rate": 1.9927477994641217e-05, "loss": 1.1284, "step": 1010 }, { "epoch": 0.07, "learning_rate": 1.992721860753847e-05, "loss": 1.1622, "step": 1011 }, { "epoch": 0.07, "learning_rate": 1.9926958759086918e-05, "loss": 1.1265, "step": 1012 }, { "epoch": 0.07, "learning_rate": 1.992669844929864e-05, "loss": 1.1207, "step": 1013 }, { "epoch": 0.07, "learning_rate": 1.992643767818573e-05, "loss": 1.1999, "step": 1014 }, { "epoch": 0.07, "learning_rate": 1.9926176445760308e-05, "loss": 1.1436, "step": 1015 }, { "epoch": 0.07, "learning_rate": 1.9925914752034513e-05, "loss": 1.1382, "step": 1016 }, { "epoch": 0.07, "learning_rate": 1.9925652597020507e-05, "loss": 1.1814, "step": 1017 }, { "epoch": 0.07, "learning_rate": 1.9925389980730475e-05, "loss": 1.1394, "step": 1018 }, { "epoch": 0.07, "learning_rate": 1.9925126903176623e-05, "loss": 1.2009, "step": 1019 }, { "epoch": 0.07, "learning_rate": 1.9924863364371173e-05, "loss": 1.1191, "step": 1020 }, { "epoch": 0.07, "learning_rate": 1.992459936432637e-05, "loss": 1.1942, "step": 1021 }, { "epoch": 0.07, "learning_rate": 1.9924334903054495e-05, "loss": 1.0976, "step": 1022 }, { "epoch": 0.07, "learning_rate": 1.9924069980567823e-05, "loss": 1.1549, "step": 1023 }, { "epoch": 0.07, "learning_rate": 1.9923804596878678e-05, "loss": 1.152, "step": 1024 }, { "epoch": 0.07, "learning_rate": 1.9923538751999385e-05, "loss": 1.1536, "step": 1025 }, { "epoch": 0.07, "learning_rate": 1.9923272445942304e-05, "loss": 1.1736, "step": 1026 }, { "epoch": 0.07, "learning_rate": 1.9923005678719805e-05, "loss": 1.1889, "step": 1027 }, { "epoch": 0.07, "learning_rate": 1.9922738450344293e-05, "loss": 1.1093, "step": 1028 }, { "epoch": 0.07, "learning_rate": 1.9922470760828184e-05, "loss": 1.1577, "step": 1029 }, { "epoch": 0.07, "learning_rate": 1.9922202610183917e-05, "loss": 1.1396, "step": 1030 }, { "epoch": 0.07, "learning_rate": 1.9921933998423956e-05, "loss": 1.181, "step": 1031 }, { "epoch": 0.07, "learning_rate": 1.992166492556078e-05, "loss": 1.1493, "step": 1032 }, { "epoch": 0.07, "learning_rate": 1.9921395391606898e-05, "loss": 1.1571, "step": 1033 }, { "epoch": 0.07, "learning_rate": 1.9921125396574835e-05, "loss": 1.1207, "step": 1034 }, { "epoch": 0.07, "learning_rate": 1.9920854940477136e-05, "loss": 1.1612, "step": 1035 }, { "epoch": 0.07, "learning_rate": 1.9920584023326372e-05, "loss": 1.138, "step": 1036 }, { "epoch": 0.07, "learning_rate": 1.9920312645135135e-05, "loss": 1.1595, "step": 1037 }, { "epoch": 0.07, "learning_rate": 1.9920040805916035e-05, "loss": 1.1388, "step": 1038 }, { "epoch": 0.07, "learning_rate": 1.9919768505681706e-05, "loss": 1.1509, "step": 1039 }, { "epoch": 0.07, "learning_rate": 1.99194957444448e-05, "loss": 1.1359, "step": 1040 }, { "epoch": 0.07, "learning_rate": 1.9919222522217998e-05, "loss": 1.1722, "step": 1041 }, { "epoch": 0.07, "learning_rate": 1.9918948839013993e-05, "loss": 1.1452, "step": 1042 }, { "epoch": 0.07, "learning_rate": 1.9918674694845506e-05, "loss": 1.1794, "step": 1043 }, { "epoch": 0.07, "learning_rate": 1.9918400089725277e-05, "loss": 1.1431, "step": 1044 }, { "epoch": 0.07, "learning_rate": 1.9918125023666063e-05, "loss": 1.1813, "step": 1045 }, { "epoch": 0.07, "learning_rate": 1.9917849496680657e-05, "loss": 1.1394, "step": 1046 }, { "epoch": 0.07, "learning_rate": 1.9917573508781856e-05, "loss": 1.1277, "step": 1047 }, { "epoch": 0.07, "learning_rate": 1.991729705998249e-05, "loss": 1.1389, "step": 1048 }, { "epoch": 0.07, "learning_rate": 1.9917020150295406e-05, "loss": 1.1525, "step": 1049 }, { "epoch": 0.07, "learning_rate": 1.991674277973347e-05, "loss": 1.1303, "step": 1050 }, { "epoch": 0.07, "learning_rate": 1.991646494830957e-05, "loss": 1.1465, "step": 1051 }, { "epoch": 0.07, "learning_rate": 1.9916186656036626e-05, "loss": 1.1391, "step": 1052 }, { "epoch": 0.07, "learning_rate": 1.9915907902927567e-05, "loss": 1.1704, "step": 1053 }, { "epoch": 0.07, "learning_rate": 1.991562868899535e-05, "loss": 1.1214, "step": 1054 }, { "epoch": 0.07, "learning_rate": 1.991534901425294e-05, "loss": 1.1143, "step": 1055 }, { "epoch": 0.07, "learning_rate": 1.991506887871335e-05, "loss": 1.1169, "step": 1056 }, { "epoch": 0.07, "learning_rate": 1.991478828238959e-05, "loss": 1.1457, "step": 1057 }, { "epoch": 0.07, "learning_rate": 1.9914507225294697e-05, "loss": 1.1355, "step": 1058 }, { "epoch": 0.07, "learning_rate": 1.991422570744174e-05, "loss": 1.171, "step": 1059 }, { "epoch": 0.07, "learning_rate": 1.99139437288438e-05, "loss": 1.1576, "step": 1060 }, { "epoch": 0.07, "learning_rate": 1.991366128951398e-05, "loss": 1.1778, "step": 1061 }, { "epoch": 0.07, "learning_rate": 1.9913378389465405e-05, "loss": 1.1444, "step": 1062 }, { "epoch": 0.07, "learning_rate": 1.9913095028711226e-05, "loss": 1.1583, "step": 1063 }, { "epoch": 0.07, "learning_rate": 1.991281120726461e-05, "loss": 1.1157, "step": 1064 }, { "epoch": 0.07, "learning_rate": 1.9912526925138744e-05, "loss": 1.1292, "step": 1065 }, { "epoch": 0.07, "learning_rate": 1.991224218234684e-05, "loss": 1.1798, "step": 1066 }, { "epoch": 0.07, "learning_rate": 1.9911956978902135e-05, "loss": 1.1143, "step": 1067 }, { "epoch": 0.07, "learning_rate": 1.9911671314817883e-05, "loss": 1.1616, "step": 1068 }, { "epoch": 0.07, "learning_rate": 1.9911385190107356e-05, "loss": 1.1787, "step": 1069 }, { "epoch": 0.07, "learning_rate": 1.9911098604783853e-05, "loss": 1.1572, "step": 1070 }, { "epoch": 0.07, "learning_rate": 1.9910811558860693e-05, "loss": 1.1421, "step": 1071 }, { "epoch": 0.07, "learning_rate": 1.9910524052351214e-05, "loss": 1.1429, "step": 1072 }, { "epoch": 0.07, "learning_rate": 1.991023608526878e-05, "loss": 1.1589, "step": 1073 }, { "epoch": 0.07, "learning_rate": 1.990994765762677e-05, "loss": 1.1462, "step": 1074 }, { "epoch": 0.07, "learning_rate": 1.9909658769438592e-05, "loss": 1.1559, "step": 1075 }, { "epoch": 0.07, "learning_rate": 1.990936942071767e-05, "loss": 1.2173, "step": 1076 }, { "epoch": 0.07, "learning_rate": 1.990907961147745e-05, "loss": 1.1713, "step": 1077 }, { "epoch": 0.07, "learning_rate": 1.9908789341731402e-05, "loss": 1.1054, "step": 1078 }, { "epoch": 0.07, "learning_rate": 1.9908498611493016e-05, "loss": 1.1211, "step": 1079 }, { "epoch": 0.07, "learning_rate": 1.9908207420775803e-05, "loss": 1.105, "step": 1080 }, { "epoch": 0.07, "learning_rate": 1.9907915769593293e-05, "loss": 1.1765, "step": 1081 }, { "epoch": 0.07, "learning_rate": 1.990762365795904e-05, "loss": 1.1295, "step": 1082 }, { "epoch": 0.07, "learning_rate": 1.9907331085886623e-05, "loss": 1.1591, "step": 1083 }, { "epoch": 0.07, "learning_rate": 1.9907038053389637e-05, "loss": 1.1372, "step": 1084 }, { "epoch": 0.07, "learning_rate": 1.99067445604817e-05, "loss": 1.1652, "step": 1085 }, { "epoch": 0.07, "learning_rate": 1.990645060717645e-05, "loss": 1.1002, "step": 1086 }, { "epoch": 0.07, "learning_rate": 1.9906156193487545e-05, "loss": 1.1955, "step": 1087 }, { "epoch": 0.07, "learning_rate": 1.990586131942868e-05, "loss": 1.0777, "step": 1088 }, { "epoch": 0.07, "learning_rate": 1.9905565985013546e-05, "loss": 1.164, "step": 1089 }, { "epoch": 0.07, "learning_rate": 1.9905270190255874e-05, "loss": 1.1473, "step": 1090 }, { "epoch": 0.07, "learning_rate": 1.9904973935169408e-05, "loss": 1.1436, "step": 1091 }, { "epoch": 0.07, "learning_rate": 1.9904677219767916e-05, "loss": 1.1364, "step": 1092 }, { "epoch": 0.07, "learning_rate": 1.990438004406519e-05, "loss": 1.1263, "step": 1093 }, { "epoch": 0.07, "learning_rate": 1.9904082408075038e-05, "loss": 1.1362, "step": 1094 }, { "epoch": 0.07, "learning_rate": 1.990378431181129e-05, "loss": 1.211, "step": 1095 }, { "epoch": 0.07, "learning_rate": 1.990348575528781e-05, "loss": 1.1478, "step": 1096 }, { "epoch": 0.07, "learning_rate": 1.990318673851846e-05, "loss": 1.1528, "step": 1097 }, { "epoch": 0.07, "learning_rate": 1.990288726151714e-05, "loss": 1.1044, "step": 1098 }, { "epoch": 0.07, "learning_rate": 1.990258732429777e-05, "loss": 1.171, "step": 1099 }, { "epoch": 0.07, "learning_rate": 1.990228692687429e-05, "loss": 1.1302, "step": 1100 }, { "epoch": 0.07, "learning_rate": 1.9901986069260656e-05, "loss": 1.1866, "step": 1101 }, { "epoch": 0.07, "learning_rate": 1.9901684751470855e-05, "loss": 1.13, "step": 1102 }, { "epoch": 0.07, "learning_rate": 1.9901382973518886e-05, "loss": 1.1237, "step": 1103 }, { "epoch": 0.07, "learning_rate": 1.9901080735418776e-05, "loss": 1.1328, "step": 1104 }, { "epoch": 0.07, "learning_rate": 1.9900778037184568e-05, "loss": 1.1926, "step": 1105 }, { "epoch": 0.07, "learning_rate": 1.9900474878830333e-05, "loss": 1.1251, "step": 1106 }, { "epoch": 0.07, "learning_rate": 1.990017126037016e-05, "loss": 1.1507, "step": 1107 }, { "epoch": 0.07, "learning_rate": 1.9899867181818152e-05, "loss": 1.1202, "step": 1108 }, { "epoch": 0.07, "learning_rate": 1.989956264318845e-05, "loss": 1.1723, "step": 1109 }, { "epoch": 0.07, "learning_rate": 1.98992576444952e-05, "loss": 1.137, "step": 1110 }, { "epoch": 0.07, "learning_rate": 1.989895218575258e-05, "loss": 1.1457, "step": 1111 }, { "epoch": 0.07, "learning_rate": 1.9898646266974785e-05, "loss": 1.1298, "step": 1112 }, { "epoch": 0.07, "learning_rate": 1.9898339888176027e-05, "loss": 1.1071, "step": 1113 }, { "epoch": 0.07, "learning_rate": 1.9898033049370554e-05, "loss": 1.1171, "step": 1114 }, { "epoch": 0.07, "learning_rate": 1.9897725750572618e-05, "loss": 1.1718, "step": 1115 }, { "epoch": 0.07, "learning_rate": 1.9897417991796502e-05, "loss": 1.1701, "step": 1116 }, { "epoch": 0.07, "learning_rate": 1.9897109773056513e-05, "loss": 1.1596, "step": 1117 }, { "epoch": 0.07, "learning_rate": 1.9896801094366967e-05, "loss": 1.1181, "step": 1118 }, { "epoch": 0.07, "learning_rate": 1.9896491955742213e-05, "loss": 1.1409, "step": 1119 }, { "epoch": 0.07, "learning_rate": 1.9896182357196618e-05, "loss": 1.1048, "step": 1120 }, { "epoch": 0.07, "learning_rate": 1.989587229874457e-05, "loss": 1.1171, "step": 1121 }, { "epoch": 0.07, "learning_rate": 1.989556178040048e-05, "loss": 1.1549, "step": 1122 }, { "epoch": 0.07, "learning_rate": 1.9895250802178777e-05, "loss": 1.1507, "step": 1123 }, { "epoch": 0.07, "learning_rate": 1.9894939364093916e-05, "loss": 1.1486, "step": 1124 }, { "epoch": 0.07, "learning_rate": 1.9894627466160362e-05, "loss": 1.159, "step": 1125 }, { "epoch": 0.07, "learning_rate": 1.989431510839262e-05, "loss": 1.1292, "step": 1126 }, { "epoch": 0.08, "learning_rate": 1.9894002290805196e-05, "loss": 1.1292, "step": 1127 }, { "epoch": 0.08, "learning_rate": 1.9893689013412637e-05, "loss": 1.1285, "step": 1128 }, { "epoch": 0.08, "learning_rate": 1.98933752762295e-05, "loss": 1.1687, "step": 1129 }, { "epoch": 0.08, "learning_rate": 1.989306107927036e-05, "loss": 1.11, "step": 1130 }, { "epoch": 0.08, "learning_rate": 1.9892746422549825e-05, "loss": 1.152, "step": 1131 }, { "epoch": 0.08, "learning_rate": 1.9892431306082516e-05, "loss": 1.1126, "step": 1132 }, { "epoch": 0.08, "learning_rate": 1.9892115729883076e-05, "loss": 1.1502, "step": 1133 }, { "epoch": 0.08, "learning_rate": 1.989179969396617e-05, "loss": 1.1389, "step": 1134 }, { "epoch": 0.08, "learning_rate": 1.9891483198346488e-05, "loss": 1.1462, "step": 1135 }, { "epoch": 0.08, "learning_rate": 1.9891166243038737e-05, "loss": 1.1108, "step": 1136 }, { "epoch": 0.08, "learning_rate": 1.9890848828057648e-05, "loss": 1.1164, "step": 1137 }, { "epoch": 0.08, "learning_rate": 1.9890530953417976e-05, "loss": 1.1224, "step": 1138 }, { "epoch": 0.08, "learning_rate": 1.989021261913448e-05, "loss": 1.1454, "step": 1139 }, { "epoch": 0.08, "learning_rate": 1.9889893825221972e-05, "loss": 1.1481, "step": 1140 }, { "epoch": 0.08, "learning_rate": 1.9889574571695254e-05, "loss": 1.15, "step": 1141 }, { "epoch": 0.08, "learning_rate": 1.9889254858569167e-05, "loss": 1.1543, "step": 1142 }, { "epoch": 0.08, "learning_rate": 1.988893468585857e-05, "loss": 1.1533, "step": 1143 }, { "epoch": 0.08, "learning_rate": 1.9888614053578347e-05, "loss": 1.1217, "step": 1144 }, { "epoch": 0.08, "learning_rate": 1.9888292961743386e-05, "loss": 1.1412, "step": 1145 }, { "epoch": 0.08, "learning_rate": 1.988797141036862e-05, "loss": 1.141, "step": 1146 }, { "epoch": 0.08, "learning_rate": 1.9887649399468988e-05, "loss": 1.1268, "step": 1147 }, { "epoch": 0.08, "learning_rate": 1.9887326929059458e-05, "loss": 1.1196, "step": 1148 }, { "epoch": 0.08, "learning_rate": 1.9887003999155012e-05, "loss": 1.1447, "step": 1149 }, { "epoch": 0.08, "learning_rate": 1.9886680609770663e-05, "loss": 1.1323, "step": 1150 }, { "epoch": 0.08, "learning_rate": 1.9886356760921433e-05, "loss": 1.1503, "step": 1151 }, { "epoch": 0.08, "learning_rate": 1.9886032452622375e-05, "loss": 1.1233, "step": 1152 }, { "epoch": 0.08, "learning_rate": 1.9885707684888565e-05, "loss": 1.1955, "step": 1153 }, { "epoch": 0.08, "learning_rate": 1.9885382457735095e-05, "loss": 1.1079, "step": 1154 }, { "epoch": 0.08, "learning_rate": 1.9885056771177074e-05, "loss": 1.1453, "step": 1155 }, { "epoch": 0.08, "learning_rate": 1.9884730625229637e-05, "loss": 1.0909, "step": 1156 }, { "epoch": 0.08, "learning_rate": 1.988440401990795e-05, "loss": 1.1792, "step": 1157 }, { "epoch": 0.08, "learning_rate": 1.9884076955227182e-05, "loss": 1.1292, "step": 1158 }, { "epoch": 0.08, "learning_rate": 1.988374943120254e-05, "loss": 1.1423, "step": 1159 }, { "epoch": 0.08, "learning_rate": 1.988342144784924e-05, "loss": 1.1219, "step": 1160 }, { "epoch": 0.08, "learning_rate": 1.9883093005182527e-05, "loss": 1.1511, "step": 1161 }, { "epoch": 0.08, "learning_rate": 1.9882764103217663e-05, "loss": 1.1286, "step": 1162 }, { "epoch": 0.08, "learning_rate": 1.9882434741969936e-05, "loss": 1.1373, "step": 1163 }, { "epoch": 0.08, "learning_rate": 1.9882104921454646e-05, "loss": 1.1484, "step": 1164 }, { "epoch": 0.08, "learning_rate": 1.9881774641687128e-05, "loss": 1.1381, "step": 1165 }, { "epoch": 0.08, "learning_rate": 1.988144390268273e-05, "loss": 1.0824, "step": 1166 }, { "epoch": 0.08, "learning_rate": 1.988111270445682e-05, "loss": 1.1219, "step": 1167 }, { "epoch": 0.08, "learning_rate": 1.9880781047024788e-05, "loss": 1.1501, "step": 1168 }, { "epoch": 0.08, "learning_rate": 1.988044893040205e-05, "loss": 1.1206, "step": 1169 }, { "epoch": 0.08, "learning_rate": 1.9880116354604045e-05, "loss": 1.1377, "step": 1170 }, { "epoch": 0.08, "learning_rate": 1.987978331964622e-05, "loss": 1.1254, "step": 1171 }, { "epoch": 0.08, "learning_rate": 1.9879449825544056e-05, "loss": 1.1453, "step": 1172 }, { "epoch": 0.08, "learning_rate": 1.9879115872313054e-05, "loss": 1.1296, "step": 1173 }, { "epoch": 0.08, "learning_rate": 1.987878145996873e-05, "loss": 1.1201, "step": 1174 }, { "epoch": 0.08, "learning_rate": 1.987844658852663e-05, "loss": 1.1612, "step": 1175 }, { "epoch": 0.08, "learning_rate": 1.987811125800231e-05, "loss": 1.12, "step": 1176 }, { "epoch": 0.08, "learning_rate": 1.987777546841136e-05, "loss": 1.1498, "step": 1177 }, { "epoch": 0.08, "learning_rate": 1.987743921976938e-05, "loss": 1.1575, "step": 1178 }, { "epoch": 0.08, "learning_rate": 1.9877102512092e-05, "loss": 1.1106, "step": 1179 }, { "epoch": 0.08, "learning_rate": 1.9876765345394866e-05, "loss": 1.1033, "step": 1180 }, { "epoch": 0.08, "learning_rate": 1.9876427719693648e-05, "loss": 1.1324, "step": 1181 }, { "epoch": 0.08, "learning_rate": 1.9876089635004037e-05, "loss": 1.1194, "step": 1182 }, { "epoch": 0.08, "learning_rate": 1.9875751091341742e-05, "loss": 1.148, "step": 1183 }, { "epoch": 0.08, "learning_rate": 1.9875412088722502e-05, "loss": 1.1221, "step": 1184 }, { "epoch": 0.08, "learning_rate": 1.9875072627162065e-05, "loss": 1.1719, "step": 1185 }, { "epoch": 0.08, "learning_rate": 1.9874732706676216e-05, "loss": 1.1337, "step": 1186 }, { "epoch": 0.08, "learning_rate": 1.987439232728074e-05, "loss": 1.1209, "step": 1187 }, { "epoch": 0.08, "learning_rate": 1.9874051488991462e-05, "loss": 1.0971, "step": 1188 }, { "epoch": 0.08, "learning_rate": 1.9873710191824224e-05, "loss": 1.1941, "step": 1189 }, { "epoch": 0.08, "learning_rate": 1.987336843579488e-05, "loss": 1.1251, "step": 1190 }, { "epoch": 0.08, "learning_rate": 1.9873026220919324e-05, "loss": 1.1434, "step": 1191 }, { "epoch": 0.08, "learning_rate": 1.9872683547213446e-05, "loss": 1.1159, "step": 1192 }, { "epoch": 0.08, "learning_rate": 1.9872340414693184e-05, "loss": 1.1908, "step": 1193 }, { "epoch": 0.08, "learning_rate": 1.9871996823374475e-05, "loss": 1.1409, "step": 1194 }, { "epoch": 0.08, "learning_rate": 1.9871652773273288e-05, "loss": 1.1238, "step": 1195 }, { "epoch": 0.08, "learning_rate": 1.9871308264405616e-05, "loss": 1.1117, "step": 1196 }, { "epoch": 0.08, "learning_rate": 1.987096329678747e-05, "loss": 1.1289, "step": 1197 }, { "epoch": 0.08, "learning_rate": 1.9870617870434878e-05, "loss": 1.0821, "step": 1198 }, { "epoch": 0.08, "learning_rate": 1.9870271985363893e-05, "loss": 1.1239, "step": 1199 }, { "epoch": 0.08, "learning_rate": 1.9869925641590587e-05, "loss": 1.1164, "step": 1200 }, { "epoch": 0.08, "learning_rate": 1.9869578839131065e-05, "loss": 1.1889, "step": 1201 }, { "epoch": 0.08, "learning_rate": 1.9869231578001437e-05, "loss": 1.1395, "step": 1202 }, { "epoch": 0.08, "learning_rate": 1.986888385821784e-05, "loss": 1.1448, "step": 1203 }, { "epoch": 0.08, "learning_rate": 1.986853567979644e-05, "loss": 1.1244, "step": 1204 }, { "epoch": 0.08, "learning_rate": 1.9868187042753414e-05, "loss": 1.1212, "step": 1205 }, { "epoch": 0.08, "learning_rate": 1.986783794710496e-05, "loss": 1.1067, "step": 1206 }, { "epoch": 0.08, "learning_rate": 1.986748839286731e-05, "loss": 1.1466, "step": 1207 }, { "epoch": 0.08, "learning_rate": 1.9867138380056706e-05, "loss": 1.1456, "step": 1208 }, { "epoch": 0.08, "learning_rate": 1.9866787908689413e-05, "loss": 1.1358, "step": 1209 }, { "epoch": 0.08, "learning_rate": 1.9866436978781714e-05, "loss": 1.106, "step": 1210 }, { "epoch": 0.08, "learning_rate": 1.9866085590349928e-05, "loss": 1.1568, "step": 1211 }, { "epoch": 0.08, "learning_rate": 1.9865733743410374e-05, "loss": 1.1433, "step": 1212 }, { "epoch": 0.08, "learning_rate": 1.986538143797941e-05, "loss": 1.1229, "step": 1213 }, { "epoch": 0.08, "learning_rate": 1.986502867407341e-05, "loss": 1.0779, "step": 1214 }, { "epoch": 0.08, "learning_rate": 1.9864675451708767e-05, "loss": 1.0883, "step": 1215 }, { "epoch": 0.08, "learning_rate": 1.986432177090189e-05, "loss": 1.1092, "step": 1216 }, { "epoch": 0.08, "learning_rate": 1.9863967631669228e-05, "loss": 1.1603, "step": 1217 }, { "epoch": 0.08, "learning_rate": 1.9863613034027224e-05, "loss": 1.1243, "step": 1218 }, { "epoch": 0.08, "learning_rate": 1.9863257977992368e-05, "loss": 1.1087, "step": 1219 }, { "epoch": 0.08, "learning_rate": 1.986290246358116e-05, "loss": 1.0891, "step": 1220 }, { "epoch": 0.08, "learning_rate": 1.9862546490810115e-05, "loss": 1.1236, "step": 1221 }, { "epoch": 0.08, "learning_rate": 1.9862190059695786e-05, "loss": 1.081, "step": 1222 }, { "epoch": 0.08, "learning_rate": 1.9861833170254727e-05, "loss": 1.1331, "step": 1223 }, { "epoch": 0.08, "learning_rate": 1.9861475822503528e-05, "loss": 1.1588, "step": 1224 }, { "epoch": 0.08, "learning_rate": 1.98611180164588e-05, "loss": 1.1095, "step": 1225 }, { "epoch": 0.08, "learning_rate": 1.9860759752137165e-05, "loss": 1.1043, "step": 1226 }, { "epoch": 0.08, "learning_rate": 1.986040102955528e-05, "loss": 1.1166, "step": 1227 }, { "epoch": 0.08, "learning_rate": 1.986004184872981e-05, "loss": 1.12, "step": 1228 }, { "epoch": 0.08, "learning_rate": 1.985968220967745e-05, "loss": 1.1055, "step": 1229 }, { "epoch": 0.08, "learning_rate": 1.985932211241491e-05, "loss": 1.0894, "step": 1230 }, { "epoch": 0.08, "learning_rate": 1.985896155695893e-05, "loss": 1.1294, "step": 1231 }, { "epoch": 0.08, "learning_rate": 1.9858600543326262e-05, "loss": 1.1092, "step": 1232 }, { "epoch": 0.08, "learning_rate": 1.9858239071533683e-05, "loss": 1.1388, "step": 1233 }, { "epoch": 0.08, "learning_rate": 1.9857877141597995e-05, "loss": 1.1241, "step": 1234 }, { "epoch": 0.08, "learning_rate": 1.985751475353602e-05, "loss": 1.0978, "step": 1235 }, { "epoch": 0.08, "learning_rate": 1.9857151907364593e-05, "loss": 1.1476, "step": 1236 }, { "epoch": 0.08, "learning_rate": 1.985678860310058e-05, "loss": 1.1068, "step": 1237 }, { "epoch": 0.08, "learning_rate": 1.9856424840760866e-05, "loss": 1.0991, "step": 1238 }, { "epoch": 0.08, "learning_rate": 1.985606062036235e-05, "loss": 1.1264, "step": 1239 }, { "epoch": 0.08, "learning_rate": 1.985569594192197e-05, "loss": 1.0962, "step": 1240 }, { "epoch": 0.08, "learning_rate": 1.9855330805456662e-05, "loss": 1.1439, "step": 1241 }, { "epoch": 0.08, "learning_rate": 1.98549652109834e-05, "loss": 1.0937, "step": 1242 }, { "epoch": 0.08, "learning_rate": 1.9854599158519175e-05, "loss": 1.1721, "step": 1243 }, { "epoch": 0.08, "learning_rate": 1.9854232648081e-05, "loss": 1.2015, "step": 1244 }, { "epoch": 0.08, "learning_rate": 1.9853865679685906e-05, "loss": 1.1743, "step": 1245 }, { "epoch": 0.08, "learning_rate": 1.9853498253350946e-05, "loss": 1.1432, "step": 1246 }, { "epoch": 0.08, "learning_rate": 1.985313036909319e-05, "loss": 1.131, "step": 1247 }, { "epoch": 0.08, "learning_rate": 1.985276202692975e-05, "loss": 1.0932, "step": 1248 }, { "epoch": 0.08, "learning_rate": 1.985239322687773e-05, "loss": 1.116, "step": 1249 }, { "epoch": 0.08, "learning_rate": 1.9852023968954275e-05, "loss": 1.1306, "step": 1250 }, { "epoch": 0.08, "learning_rate": 1.9851654253176545e-05, "loss": 1.1779, "step": 1251 }, { "epoch": 0.08, "learning_rate": 1.985128407956172e-05, "loss": 1.0984, "step": 1252 }, { "epoch": 0.08, "learning_rate": 1.9850913448127008e-05, "loss": 1.1186, "step": 1253 }, { "epoch": 0.08, "learning_rate": 1.985054235888963e-05, "loss": 1.1089, "step": 1254 }, { "epoch": 0.08, "learning_rate": 1.9850170811866833e-05, "loss": 1.1063, "step": 1255 }, { "epoch": 0.08, "learning_rate": 1.984979880707588e-05, "loss": 1.146, "step": 1256 }, { "epoch": 0.08, "learning_rate": 1.984942634453406e-05, "loss": 1.1469, "step": 1257 }, { "epoch": 0.08, "learning_rate": 1.9849053424258692e-05, "loss": 1.1161, "step": 1258 }, { "epoch": 0.08, "learning_rate": 1.9848680046267095e-05, "loss": 1.1419, "step": 1259 }, { "epoch": 0.08, "learning_rate": 1.9848306210576623e-05, "loss": 1.1331, "step": 1260 }, { "epoch": 0.08, "learning_rate": 1.9847931917204658e-05, "loss": 1.152, "step": 1261 }, { "epoch": 0.08, "learning_rate": 1.9847557166168584e-05, "loss": 1.1287, "step": 1262 }, { "epoch": 0.08, "learning_rate": 1.984718195748582e-05, "loss": 1.1579, "step": 1263 }, { "epoch": 0.08, "learning_rate": 1.9846806291173807e-05, "loss": 1.1095, "step": 1264 }, { "epoch": 0.08, "learning_rate": 1.984643016725e-05, "loss": 1.1457, "step": 1265 }, { "epoch": 0.08, "learning_rate": 1.9846053585731876e-05, "loss": 1.0721, "step": 1266 }, { "epoch": 0.08, "learning_rate": 1.9845676546636944e-05, "loss": 1.1084, "step": 1267 }, { "epoch": 0.08, "learning_rate": 1.9845299049982718e-05, "loss": 1.1317, "step": 1268 }, { "epoch": 0.08, "learning_rate": 1.9844921095786746e-05, "loss": 1.1373, "step": 1269 }, { "epoch": 0.08, "learning_rate": 1.984454268406659e-05, "loss": 1.1209, "step": 1270 }, { "epoch": 0.08, "learning_rate": 1.984416381483984e-05, "loss": 1.13, "step": 1271 }, { "epoch": 0.08, "learning_rate": 1.9843784488124102e-05, "loss": 1.1021, "step": 1272 }, { "epoch": 0.08, "learning_rate": 1.9843404703937e-05, "loss": 1.1232, "step": 1273 }, { "epoch": 0.08, "learning_rate": 1.9843024462296192e-05, "loss": 1.1145, "step": 1274 }, { "epoch": 0.08, "learning_rate": 1.984264376321934e-05, "loss": 1.1202, "step": 1275 }, { "epoch": 0.08, "learning_rate": 1.984226260672414e-05, "loss": 1.142, "step": 1276 }, { "epoch": 0.08, "learning_rate": 1.9841880992828306e-05, "loss": 1.1577, "step": 1277 }, { "epoch": 0.09, "learning_rate": 1.9841498921549574e-05, "loss": 1.0977, "step": 1278 }, { "epoch": 0.09, "learning_rate": 1.9841116392905698e-05, "loss": 1.1504, "step": 1279 }, { "epoch": 0.09, "learning_rate": 1.9840733406914453e-05, "loss": 1.1122, "step": 1280 }, { "epoch": 0.09, "learning_rate": 1.9840349963593647e-05, "loss": 1.1354, "step": 1281 }, { "epoch": 0.09, "learning_rate": 1.983996606296109e-05, "loss": 1.0687, "step": 1282 }, { "epoch": 0.09, "learning_rate": 1.9839581705034626e-05, "loss": 1.125, "step": 1283 }, { "epoch": 0.09, "learning_rate": 1.9839196889832116e-05, "loss": 1.0999, "step": 1284 }, { "epoch": 0.09, "learning_rate": 1.983881161737145e-05, "loss": 1.1241, "step": 1285 }, { "epoch": 0.09, "learning_rate": 1.9838425887670523e-05, "loss": 1.0722, "step": 1286 }, { "epoch": 0.09, "learning_rate": 1.983803970074727e-05, "loss": 1.1143, "step": 1287 }, { "epoch": 0.09, "learning_rate": 1.983765305661963e-05, "loss": 1.1213, "step": 1288 }, { "epoch": 0.09, "learning_rate": 1.983726595530558e-05, "loss": 1.1651, "step": 1289 }, { "epoch": 0.09, "learning_rate": 1.9836878396823104e-05, "loss": 1.1483, "step": 1290 }, { "epoch": 0.09, "learning_rate": 1.9836490381190215e-05, "loss": 1.1426, "step": 1291 }, { "epoch": 0.09, "learning_rate": 1.9836101908424944e-05, "loss": 1.1165, "step": 1292 }, { "epoch": 0.09, "learning_rate": 1.983571297854535e-05, "loss": 1.1189, "step": 1293 }, { "epoch": 0.09, "learning_rate": 1.98353235915695e-05, "loss": 1.1149, "step": 1294 }, { "epoch": 0.09, "learning_rate": 1.9834933747515496e-05, "loss": 1.0911, "step": 1295 }, { "epoch": 0.09, "learning_rate": 1.983454344640145e-05, "loss": 1.1103, "step": 1296 }, { "epoch": 0.09, "learning_rate": 1.9834152688245504e-05, "loss": 1.1234, "step": 1297 }, { "epoch": 0.09, "learning_rate": 1.9833761473065817e-05, "loss": 1.0824, "step": 1298 }, { "epoch": 0.09, "learning_rate": 1.9833369800880572e-05, "loss": 1.0908, "step": 1299 }, { "epoch": 0.09, "learning_rate": 1.9832977671707968e-05, "loss": 1.1249, "step": 1300 }, { "epoch": 0.09, "learning_rate": 1.9832585085566228e-05, "loss": 1.1307, "step": 1301 }, { "epoch": 0.09, "learning_rate": 1.98321920424736e-05, "loss": 1.1214, "step": 1302 }, { "epoch": 0.09, "learning_rate": 1.983179854244835e-05, "loss": 1.1463, "step": 1303 }, { "epoch": 0.09, "learning_rate": 1.983140458550876e-05, "loss": 1.09, "step": 1304 }, { "epoch": 0.09, "learning_rate": 1.9831010171673146e-05, "loss": 1.1168, "step": 1305 }, { "epoch": 0.09, "learning_rate": 1.983061530095983e-05, "loss": 1.1079, "step": 1306 }, { "epoch": 0.09, "learning_rate": 1.9830219973387167e-05, "loss": 1.1197, "step": 1307 }, { "epoch": 0.09, "learning_rate": 1.982982418897353e-05, "loss": 1.1109, "step": 1308 }, { "epoch": 0.09, "learning_rate": 1.982942794773731e-05, "loss": 1.1312, "step": 1309 }, { "epoch": 0.09, "learning_rate": 1.9829031249696918e-05, "loss": 1.0996, "step": 1310 }, { "epoch": 0.09, "learning_rate": 1.98286340948708e-05, "loss": 1.1513, "step": 1311 }, { "epoch": 0.09, "learning_rate": 1.982823648327741e-05, "loss": 1.1256, "step": 1312 }, { "epoch": 0.09, "learning_rate": 1.9827838414935217e-05, "loss": 1.0955, "step": 1313 }, { "epoch": 0.09, "learning_rate": 1.982743988986273e-05, "loss": 1.1011, "step": 1314 }, { "epoch": 0.09, "learning_rate": 1.9827040908078464e-05, "loss": 1.1355, "step": 1315 }, { "epoch": 0.09, "learning_rate": 1.9826641469600968e-05, "loss": 1.1067, "step": 1316 }, { "epoch": 0.09, "learning_rate": 1.98262415744488e-05, "loss": 1.122, "step": 1317 }, { "epoch": 0.09, "learning_rate": 1.982584122264054e-05, "loss": 1.1083, "step": 1318 }, { "epoch": 0.09, "learning_rate": 1.9825440414194803e-05, "loss": 1.1118, "step": 1319 }, { "epoch": 0.09, "learning_rate": 1.9825039149130214e-05, "loss": 1.1307, "step": 1320 }, { "epoch": 0.09, "learning_rate": 1.9824637427465414e-05, "loss": 1.1304, "step": 1321 }, { "epoch": 0.09, "learning_rate": 1.982423524921908e-05, "loss": 1.1304, "step": 1322 }, { "epoch": 0.09, "learning_rate": 1.98238326144099e-05, "loss": 1.153, "step": 1323 }, { "epoch": 0.09, "learning_rate": 1.982342952305658e-05, "loss": 1.1223, "step": 1324 }, { "epoch": 0.09, "learning_rate": 1.9823025975177863e-05, "loss": 1.1326, "step": 1325 }, { "epoch": 0.09, "learning_rate": 1.9822621970792494e-05, "loss": 1.1254, "step": 1326 }, { "epoch": 0.09, "learning_rate": 1.9822217509919256e-05, "loss": 1.1339, "step": 1327 }, { "epoch": 0.09, "learning_rate": 1.982181259257694e-05, "loss": 1.07, "step": 1328 }, { "epoch": 0.09, "learning_rate": 1.9821407218784364e-05, "loss": 1.1352, "step": 1329 }, { "epoch": 0.09, "learning_rate": 1.9821001388560374e-05, "loss": 1.126, "step": 1330 }, { "epoch": 0.09, "learning_rate": 1.9820595101923824e-05, "loss": 1.1296, "step": 1331 }, { "epoch": 0.09, "learning_rate": 1.9820188358893593e-05, "loss": 1.0861, "step": 1332 }, { "epoch": 0.09, "learning_rate": 1.9819781159488587e-05, "loss": 1.1137, "step": 1333 }, { "epoch": 0.09, "learning_rate": 1.981937350372773e-05, "loss": 1.1618, "step": 1334 }, { "epoch": 0.09, "learning_rate": 1.981896539162997e-05, "loss": 1.1142, "step": 1335 }, { "epoch": 0.09, "learning_rate": 1.981855682321427e-05, "loss": 1.1617, "step": 1336 }, { "epoch": 0.09, "learning_rate": 1.981814779849961e-05, "loss": 1.1239, "step": 1337 }, { "epoch": 0.09, "learning_rate": 1.9817738317505014e-05, "loss": 1.1298, "step": 1338 }, { "epoch": 0.09, "learning_rate": 1.98173283802495e-05, "loss": 1.1425, "step": 1339 }, { "epoch": 0.09, "learning_rate": 1.9816917986752126e-05, "loss": 1.0763, "step": 1340 }, { "epoch": 0.09, "learning_rate": 1.9816507137031962e-05, "loss": 1.1472, "step": 1341 }, { "epoch": 0.09, "learning_rate": 1.98160958311081e-05, "loss": 1.1154, "step": 1342 }, { "epoch": 0.09, "learning_rate": 1.9815684068999653e-05, "loss": 1.1344, "step": 1343 }, { "epoch": 0.09, "learning_rate": 1.981527185072576e-05, "loss": 1.0794, "step": 1344 }, { "epoch": 0.09, "learning_rate": 1.9814859176305578e-05, "loss": 1.1232, "step": 1345 }, { "epoch": 0.09, "learning_rate": 1.9814446045758286e-05, "loss": 1.1221, "step": 1346 }, { "epoch": 0.09, "learning_rate": 1.981403245910308e-05, "loss": 1.1405, "step": 1347 }, { "epoch": 0.09, "learning_rate": 1.9813618416359186e-05, "loss": 1.1136, "step": 1348 }, { "epoch": 0.09, "learning_rate": 1.9813203917545838e-05, "loss": 1.1593, "step": 1349 }, { "epoch": 0.09, "learning_rate": 1.9812788962682308e-05, "loss": 1.088, "step": 1350 }, { "epoch": 0.09, "learning_rate": 1.9812373551787877e-05, "loss": 1.1379, "step": 1351 }, { "epoch": 0.09, "learning_rate": 1.9811957684881844e-05, "loss": 1.0798, "step": 1352 }, { "epoch": 0.09, "learning_rate": 1.9811541361983545e-05, "loss": 1.166, "step": 1353 }, { "epoch": 0.09, "learning_rate": 1.981112458311232e-05, "loss": 1.1133, "step": 1354 }, { "epoch": 0.09, "learning_rate": 1.9810707348287546e-05, "loss": 1.1105, "step": 1355 }, { "epoch": 0.09, "learning_rate": 1.9810289657528606e-05, "loss": 1.1278, "step": 1356 }, { "epoch": 0.09, "learning_rate": 1.9809871510854917e-05, "loss": 1.1402, "step": 1357 }, { "epoch": 0.09, "learning_rate": 1.9809452908285907e-05, "loss": 1.115, "step": 1358 }, { "epoch": 0.09, "learning_rate": 1.9809033849841034e-05, "loss": 1.1361, "step": 1359 }, { "epoch": 0.09, "learning_rate": 1.980861433553977e-05, "loss": 1.0966, "step": 1360 }, { "epoch": 0.09, "learning_rate": 1.9808194365401612e-05, "loss": 1.1222, "step": 1361 }, { "epoch": 0.09, "learning_rate": 1.9807773939446078e-05, "loss": 1.1405, "step": 1362 }, { "epoch": 0.09, "learning_rate": 1.9807353057692706e-05, "loss": 1.1201, "step": 1363 }, { "epoch": 0.09, "learning_rate": 1.9806931720161052e-05, "loss": 1.1161, "step": 1364 }, { "epoch": 0.09, "learning_rate": 1.9806509926870704e-05, "loss": 1.096, "step": 1365 }, { "epoch": 0.09, "learning_rate": 1.9806087677841258e-05, "loss": 1.115, "step": 1366 }, { "epoch": 0.09, "learning_rate": 1.980566497309234e-05, "loss": 1.1051, "step": 1367 }, { "epoch": 0.09, "learning_rate": 1.9805241812643596e-05, "loss": 1.1445, "step": 1368 }, { "epoch": 0.09, "learning_rate": 1.980481819651469e-05, "loss": 1.1149, "step": 1369 }, { "epoch": 0.09, "learning_rate": 1.9804394124725303e-05, "loss": 1.1054, "step": 1370 }, { "epoch": 0.09, "learning_rate": 1.9803969597295153e-05, "loss": 1.0787, "step": 1371 }, { "epoch": 0.09, "learning_rate": 1.9803544614243965e-05, "loss": 1.0988, "step": 1372 }, { "epoch": 0.09, "learning_rate": 1.9803119175591484e-05, "loss": 1.1417, "step": 1373 }, { "epoch": 0.09, "learning_rate": 1.9802693281357488e-05, "loss": 1.071, "step": 1374 }, { "epoch": 0.09, "learning_rate": 1.980226693156177e-05, "loss": 1.1361, "step": 1375 }, { "epoch": 0.09, "learning_rate": 1.980184012622414e-05, "loss": 1.0893, "step": 1376 }, { "epoch": 0.09, "learning_rate": 1.9801412865364434e-05, "loss": 1.1131, "step": 1377 }, { "epoch": 0.09, "learning_rate": 1.980098514900251e-05, "loss": 1.1066, "step": 1378 }, { "epoch": 0.09, "learning_rate": 1.9800556977158243e-05, "loss": 1.1679, "step": 1379 }, { "epoch": 0.09, "learning_rate": 1.9800128349851533e-05, "loss": 1.1237, "step": 1380 }, { "epoch": 0.09, "learning_rate": 1.97996992671023e-05, "loss": 1.155, "step": 1381 }, { "epoch": 0.09, "learning_rate": 1.9799269728930483e-05, "loss": 1.0601, "step": 1382 }, { "epoch": 0.09, "learning_rate": 1.9798839735356045e-05, "loss": 1.0945, "step": 1383 }, { "epoch": 0.09, "learning_rate": 1.9798409286398968e-05, "loss": 1.1097, "step": 1384 }, { "epoch": 0.09, "learning_rate": 1.9797978382079258e-05, "loss": 1.1442, "step": 1385 }, { "epoch": 0.09, "learning_rate": 1.979754702241694e-05, "loss": 1.1184, "step": 1386 }, { "epoch": 0.09, "learning_rate": 1.9797115207432062e-05, "loss": 1.1155, "step": 1387 }, { "epoch": 0.09, "learning_rate": 1.9796682937144687e-05, "loss": 1.1152, "step": 1388 }, { "epoch": 0.09, "learning_rate": 1.979625021157491e-05, "loss": 1.1009, "step": 1389 }, { "epoch": 0.09, "learning_rate": 1.9795817030742838e-05, "loss": 1.0547, "step": 1390 }, { "epoch": 0.09, "learning_rate": 1.9795383394668602e-05, "loss": 1.1163, "step": 1391 }, { "epoch": 0.09, "learning_rate": 1.9794949303372355e-05, "loss": 1.093, "step": 1392 }, { "epoch": 0.09, "learning_rate": 1.979451475687427e-05, "loss": 1.1463, "step": 1393 }, { "epoch": 0.09, "learning_rate": 1.9794079755194545e-05, "loss": 1.0774, "step": 1394 }, { "epoch": 0.09, "learning_rate": 1.979364429835339e-05, "loss": 1.1289, "step": 1395 }, { "epoch": 0.09, "learning_rate": 1.9793208386371047e-05, "loss": 1.1452, "step": 1396 }, { "epoch": 0.09, "learning_rate": 1.9792772019267768e-05, "loss": 1.1261, "step": 1397 }, { "epoch": 0.09, "learning_rate": 1.9792335197063844e-05, "loss": 1.1013, "step": 1398 }, { "epoch": 0.09, "learning_rate": 1.9791897919779567e-05, "loss": 1.1216, "step": 1399 }, { "epoch": 0.09, "learning_rate": 1.9791460187435258e-05, "loss": 1.1031, "step": 1400 }, { "epoch": 0.09, "learning_rate": 1.979102200005126e-05, "loss": 1.1335, "step": 1401 }, { "epoch": 0.09, "learning_rate": 1.9790583357647936e-05, "loss": 1.0762, "step": 1402 }, { "epoch": 0.09, "learning_rate": 1.979014426024568e-05, "loss": 1.1282, "step": 1403 }, { "epoch": 0.09, "learning_rate": 1.978970470786489e-05, "loss": 1.0847, "step": 1404 }, { "epoch": 0.09, "learning_rate": 1.9789264700525994e-05, "loss": 1.1509, "step": 1405 }, { "epoch": 0.09, "learning_rate": 1.978882423824944e-05, "loss": 1.1451, "step": 1406 }, { "epoch": 0.09, "learning_rate": 1.9788383321055704e-05, "loss": 1.1617, "step": 1407 }, { "epoch": 0.09, "learning_rate": 1.978794194896527e-05, "loss": 1.1149, "step": 1408 }, { "epoch": 0.09, "learning_rate": 1.9787500121998657e-05, "loss": 1.117, "step": 1409 }, { "epoch": 0.09, "learning_rate": 1.9787057840176386e-05, "loss": 1.0849, "step": 1410 }, { "epoch": 0.09, "learning_rate": 1.978661510351902e-05, "loss": 1.1411, "step": 1411 }, { "epoch": 0.09, "learning_rate": 1.9786171912047135e-05, "loss": 1.0723, "step": 1412 }, { "epoch": 0.09, "learning_rate": 1.9785728265781327e-05, "loss": 1.0921, "step": 1413 }, { "epoch": 0.09, "learning_rate": 1.9785284164742207e-05, "loss": 1.1471, "step": 1414 }, { "epoch": 0.09, "learning_rate": 1.978483960895042e-05, "loss": 1.1107, "step": 1415 }, { "epoch": 0.09, "learning_rate": 1.9784394598426626e-05, "loss": 1.1005, "step": 1416 }, { "epoch": 0.09, "learning_rate": 1.9783949133191506e-05, "loss": 1.0942, "step": 1417 }, { "epoch": 0.09, "learning_rate": 1.978350321326576e-05, "loss": 1.1341, "step": 1418 }, { "epoch": 0.09, "learning_rate": 1.978305683867011e-05, "loss": 1.1086, "step": 1419 }, { "epoch": 0.09, "learning_rate": 1.9782610009425305e-05, "loss": 1.0983, "step": 1420 }, { "epoch": 0.09, "learning_rate": 1.9782162725552107e-05, "loss": 1.1057, "step": 1421 }, { "epoch": 0.09, "learning_rate": 1.97817149870713e-05, "loss": 1.0869, "step": 1422 }, { "epoch": 0.09, "learning_rate": 1.97812667940037e-05, "loss": 1.089, "step": 1423 }, { "epoch": 0.09, "learning_rate": 1.978081814637013e-05, "loss": 1.1319, "step": 1424 }, { "epoch": 0.09, "learning_rate": 1.9780369044191443e-05, "loss": 1.1409, "step": 1425 }, { "epoch": 0.09, "learning_rate": 1.9779919487488507e-05, "loss": 1.0965, "step": 1426 }, { "epoch": 0.09, "learning_rate": 1.9779469476282216e-05, "loss": 1.0968, "step": 1427 }, { "epoch": 0.1, "learning_rate": 1.9779019010593485e-05, "loss": 1.0813, "step": 1428 }, { "epoch": 0.1, "learning_rate": 1.9778568090443247e-05, "loss": 1.1079, "step": 1429 }, { "epoch": 0.1, "learning_rate": 1.9778116715852457e-05, "loss": 1.0818, "step": 1430 }, { "epoch": 0.1, "learning_rate": 1.9777664886842093e-05, "loss": 1.1241, "step": 1431 }, { "epoch": 0.1, "learning_rate": 1.977721260343315e-05, "loss": 1.1466, "step": 1432 }, { "epoch": 0.1, "learning_rate": 1.977675986564665e-05, "loss": 1.1247, "step": 1433 }, { "epoch": 0.1, "learning_rate": 1.977630667350363e-05, "loss": 1.1137, "step": 1434 }, { "epoch": 0.1, "learning_rate": 1.9775853027025162e-05, "loss": 1.158, "step": 1435 }, { "epoch": 0.1, "learning_rate": 1.977539892623231e-05, "loss": 1.1306, "step": 1436 }, { "epoch": 0.1, "learning_rate": 1.9774944371146195e-05, "loss": 1.1424, "step": 1437 }, { "epoch": 0.1, "learning_rate": 1.9774489361787933e-05, "loss": 1.1457, "step": 1438 }, { "epoch": 0.1, "learning_rate": 1.9774033898178668e-05, "loss": 1.1187, "step": 1439 }, { "epoch": 0.1, "learning_rate": 1.9773577980339572e-05, "loss": 1.1085, "step": 1440 }, { "epoch": 0.1, "learning_rate": 1.977312160829183e-05, "loss": 1.1371, "step": 1441 }, { "epoch": 0.1, "learning_rate": 1.9772664782056653e-05, "loss": 1.0889, "step": 1442 }, { "epoch": 0.1, "learning_rate": 1.9772207501655266e-05, "loss": 1.1301, "step": 1443 }, { "epoch": 0.1, "learning_rate": 1.9771749767108924e-05, "loss": 1.0948, "step": 1444 }, { "epoch": 0.1, "learning_rate": 1.97712915784389e-05, "loss": 1.11, "step": 1445 }, { "epoch": 0.1, "learning_rate": 1.9770832935666493e-05, "loss": 1.1215, "step": 1446 }, { "epoch": 0.1, "learning_rate": 1.9770373838813007e-05, "loss": 1.1324, "step": 1447 }, { "epoch": 0.1, "learning_rate": 1.976991428789978e-05, "loss": 1.0901, "step": 1448 }, { "epoch": 0.1, "learning_rate": 1.9769454282948173e-05, "loss": 1.123, "step": 1449 }, { "epoch": 0.1, "learning_rate": 1.976899382397956e-05, "loss": 1.1255, "step": 1450 }, { "epoch": 0.1, "learning_rate": 1.9768532911015342e-05, "loss": 1.0949, "step": 1451 }, { "epoch": 0.1, "learning_rate": 1.9768071544076942e-05, "loss": 1.1257, "step": 1452 }, { "epoch": 0.1, "learning_rate": 1.9767609723185795e-05, "loss": 1.0816, "step": 1453 }, { "epoch": 0.1, "learning_rate": 1.9767147448363366e-05, "loss": 1.118, "step": 1454 }, { "epoch": 0.1, "learning_rate": 1.9766684719631143e-05, "loss": 1.1126, "step": 1455 }, { "epoch": 0.1, "learning_rate": 1.976622153701062e-05, "loss": 1.0972, "step": 1456 }, { "epoch": 0.1, "learning_rate": 1.976575790052333e-05, "loss": 1.1249, "step": 1457 }, { "epoch": 0.1, "learning_rate": 1.976529381019082e-05, "loss": 1.104, "step": 1458 }, { "epoch": 0.1, "learning_rate": 1.976482926603465e-05, "loss": 1.1781, "step": 1459 }, { "epoch": 0.1, "learning_rate": 1.9764364268076422e-05, "loss": 1.1162, "step": 1460 }, { "epoch": 0.1, "learning_rate": 1.9763898816337734e-05, "loss": 1.1454, "step": 1461 }, { "epoch": 0.1, "learning_rate": 1.9763432910840222e-05, "loss": 1.1026, "step": 1462 }, { "epoch": 0.1, "learning_rate": 1.976296655160554e-05, "loss": 1.145, "step": 1463 }, { "epoch": 0.1, "learning_rate": 1.976249973865536e-05, "loss": 1.1248, "step": 1464 }, { "epoch": 0.1, "learning_rate": 1.9762032472011372e-05, "loss": 1.1121, "step": 1465 }, { "epoch": 0.1, "learning_rate": 1.9761564751695295e-05, "loss": 1.0444, "step": 1466 }, { "epoch": 0.1, "learning_rate": 1.9761096577728863e-05, "loss": 1.1114, "step": 1467 }, { "epoch": 0.1, "learning_rate": 1.976062795013384e-05, "loss": 1.0745, "step": 1468 }, { "epoch": 0.1, "learning_rate": 1.9760158868931994e-05, "loss": 1.1119, "step": 1469 }, { "epoch": 0.1, "learning_rate": 1.9759689334145133e-05, "loss": 1.0937, "step": 1470 }, { "epoch": 0.1, "learning_rate": 1.975921934579508e-05, "loss": 1.1029, "step": 1471 }, { "epoch": 0.1, "learning_rate": 1.9758748903903667e-05, "loss": 1.1101, "step": 1472 }, { "epoch": 0.1, "learning_rate": 1.9758278008492762e-05, "loss": 1.1576, "step": 1473 }, { "epoch": 0.1, "learning_rate": 1.975780665958425e-05, "loss": 1.0575, "step": 1474 }, { "epoch": 0.1, "learning_rate": 1.9757334857200034e-05, "loss": 1.1184, "step": 1475 }, { "epoch": 0.1, "learning_rate": 1.9756862601362042e-05, "loss": 1.0844, "step": 1476 }, { "epoch": 0.1, "learning_rate": 1.975638989209222e-05, "loss": 1.1041, "step": 1477 }, { "epoch": 0.1, "learning_rate": 1.9755916729412538e-05, "loss": 1.0995, "step": 1478 }, { "epoch": 0.1, "learning_rate": 1.9755443113344984e-05, "loss": 1.1337, "step": 1479 }, { "epoch": 0.1, "learning_rate": 1.975496904391157e-05, "loss": 1.1246, "step": 1480 }, { "epoch": 0.1, "learning_rate": 1.975449452113432e-05, "loss": 1.1112, "step": 1481 }, { "epoch": 0.1, "learning_rate": 1.9754019545035296e-05, "loss": 1.1179, "step": 1482 }, { "epoch": 0.1, "learning_rate": 1.975354411563657e-05, "loss": 1.1236, "step": 1483 }, { "epoch": 0.1, "learning_rate": 1.9753068232960232e-05, "loss": 1.0564, "step": 1484 }, { "epoch": 0.1, "learning_rate": 1.9752591897028403e-05, "loss": 1.1129, "step": 1485 }, { "epoch": 0.1, "learning_rate": 1.9752115107863213e-05, "loss": 1.0538, "step": 1486 }, { "epoch": 0.1, "learning_rate": 1.9751637865486824e-05, "loss": 1.1361, "step": 1487 }, { "epoch": 0.1, "learning_rate": 1.9751160169921418e-05, "loss": 1.0732, "step": 1488 }, { "epoch": 0.1, "learning_rate": 1.9750682021189195e-05, "loss": 1.0966, "step": 1489 }, { "epoch": 0.1, "learning_rate": 1.975020341931237e-05, "loss": 1.0849, "step": 1490 }, { "epoch": 0.1, "learning_rate": 1.974972436431319e-05, "loss": 1.1038, "step": 1491 }, { "epoch": 0.1, "learning_rate": 1.9749244856213915e-05, "loss": 1.0971, "step": 1492 }, { "epoch": 0.1, "learning_rate": 1.9748764895036833e-05, "loss": 1.0746, "step": 1493 }, { "epoch": 0.1, "learning_rate": 1.9748284480804244e-05, "loss": 1.1028, "step": 1494 }, { "epoch": 0.1, "learning_rate": 1.9747803613538475e-05, "loss": 1.1, "step": 1495 }, { "epoch": 0.1, "learning_rate": 1.974732229326188e-05, "loss": 1.1274, "step": 1496 }, { "epoch": 0.1, "learning_rate": 1.9746840519996824e-05, "loss": 1.1173, "step": 1497 }, { "epoch": 0.1, "learning_rate": 1.9746358293765696e-05, "loss": 1.0552, "step": 1498 }, { "epoch": 0.1, "learning_rate": 1.9745875614590903e-05, "loss": 1.1161, "step": 1499 }, { "epoch": 0.1, "learning_rate": 1.974539248249488e-05, "loss": 1.0731, "step": 1500 }, { "epoch": 0.1, "learning_rate": 1.9744908897500085e-05, "loss": 1.0853, "step": 1501 }, { "epoch": 0.1, "learning_rate": 1.974442485962898e-05, "loss": 1.1139, "step": 1502 }, { "epoch": 0.1, "learning_rate": 1.9743940368904067e-05, "loss": 1.1432, "step": 1503 }, { "epoch": 0.1, "learning_rate": 1.9743455425347862e-05, "loss": 1.1045, "step": 1504 }, { "epoch": 0.1, "learning_rate": 1.97429700289829e-05, "loss": 1.1489, "step": 1505 }, { "epoch": 0.1, "learning_rate": 1.974248417983174e-05, "loss": 1.1124, "step": 1506 }, { "epoch": 0.1, "learning_rate": 1.974199787791696e-05, "loss": 1.0897, "step": 1507 }, { "epoch": 0.1, "learning_rate": 1.9741511123261167e-05, "loss": 1.1217, "step": 1508 }, { "epoch": 0.1, "learning_rate": 1.9741023915886964e-05, "loss": 1.0968, "step": 1509 }, { "epoch": 0.1, "learning_rate": 1.974053625581701e-05, "loss": 1.0999, "step": 1510 }, { "epoch": 0.1, "learning_rate": 1.9740048143073967e-05, "loss": 1.1158, "step": 1511 }, { "epoch": 0.1, "learning_rate": 1.973955957768051e-05, "loss": 1.0862, "step": 1512 }, { "epoch": 0.1, "learning_rate": 1.9739070559659347e-05, "loss": 1.108, "step": 1513 }, { "epoch": 0.1, "learning_rate": 1.973858108903321e-05, "loss": 1.1234, "step": 1514 }, { "epoch": 0.1, "learning_rate": 1.9738091165824838e-05, "loss": 1.071, "step": 1515 }, { "epoch": 0.1, "learning_rate": 1.9737600790057007e-05, "loss": 1.0802, "step": 1516 }, { "epoch": 0.1, "learning_rate": 1.9737109961752507e-05, "loss": 1.1039, "step": 1517 }, { "epoch": 0.1, "learning_rate": 1.9736618680934136e-05, "loss": 1.1092, "step": 1518 }, { "epoch": 0.1, "learning_rate": 1.9736126947624737e-05, "loss": 1.1823, "step": 1519 }, { "epoch": 0.1, "learning_rate": 1.9735634761847156e-05, "loss": 1.1187, "step": 1520 }, { "epoch": 0.1, "learning_rate": 1.9735142123624274e-05, "loss": 1.0966, "step": 1521 }, { "epoch": 0.1, "learning_rate": 1.973464903297898e-05, "loss": 1.0813, "step": 1522 }, { "epoch": 0.1, "learning_rate": 1.9734155489934186e-05, "loss": 1.1294, "step": 1523 }, { "epoch": 0.1, "learning_rate": 1.9733661494512838e-05, "loss": 1.1208, "step": 1524 }, { "epoch": 0.1, "learning_rate": 1.9733167046737885e-05, "loss": 1.0986, "step": 1525 }, { "epoch": 0.1, "learning_rate": 1.9732672146632308e-05, "loss": 1.1116, "step": 1526 }, { "epoch": 0.1, "learning_rate": 1.973217679421911e-05, "loss": 1.1065, "step": 1527 }, { "epoch": 0.1, "learning_rate": 1.9731680989521305e-05, "loss": 1.0927, "step": 1528 }, { "epoch": 0.1, "learning_rate": 1.973118473256194e-05, "loss": 1.1177, "step": 1529 }, { "epoch": 0.1, "learning_rate": 1.9730688023364077e-05, "loss": 1.0998, "step": 1530 }, { "epoch": 0.1, "learning_rate": 1.97301908619508e-05, "loss": 1.1137, "step": 1531 }, { "epoch": 0.1, "learning_rate": 1.972969324834521e-05, "loss": 1.0924, "step": 1532 }, { "epoch": 0.1, "learning_rate": 1.9729195182570434e-05, "loss": 1.1466, "step": 1533 }, { "epoch": 0.1, "learning_rate": 1.9728696664649622e-05, "loss": 1.1455, "step": 1534 }, { "epoch": 0.1, "learning_rate": 1.972819769460594e-05, "loss": 1.15, "step": 1535 }, { "epoch": 0.1, "learning_rate": 1.9727698272462573e-05, "loss": 1.1243, "step": 1536 }, { "epoch": 0.1, "learning_rate": 1.9727198398242736e-05, "loss": 1.1162, "step": 1537 }, { "epoch": 0.1, "learning_rate": 1.9726698071969655e-05, "loss": 1.0851, "step": 1538 }, { "epoch": 0.1, "learning_rate": 1.9726197293666586e-05, "loss": 1.0877, "step": 1539 }, { "epoch": 0.1, "learning_rate": 1.97256960633568e-05, "loss": 1.0868, "step": 1540 }, { "epoch": 0.1, "learning_rate": 1.972519438106359e-05, "loss": 1.1171, "step": 1541 }, { "epoch": 0.1, "learning_rate": 1.9724692246810272e-05, "loss": 1.0671, "step": 1542 }, { "epoch": 0.1, "learning_rate": 1.972418966062018e-05, "loss": 1.1644, "step": 1543 }, { "epoch": 0.1, "learning_rate": 1.9723686622516674e-05, "loss": 1.1207, "step": 1544 }, { "epoch": 0.1, "learning_rate": 1.9723183132523127e-05, "loss": 1.0911, "step": 1545 }, { "epoch": 0.1, "learning_rate": 1.972267919066294e-05, "loss": 1.1242, "step": 1546 }, { "epoch": 0.1, "learning_rate": 1.9722174796959535e-05, "loss": 1.1142, "step": 1547 }, { "epoch": 0.1, "learning_rate": 1.972166995143635e-05, "loss": 1.1292, "step": 1548 }, { "epoch": 0.1, "learning_rate": 1.9721164654116848e-05, "loss": 1.1077, "step": 1549 }, { "epoch": 0.1, "learning_rate": 1.9720658905024513e-05, "loss": 1.1032, "step": 1550 }, { "epoch": 0.1, "learning_rate": 1.9720152704182843e-05, "loss": 1.0878, "step": 1551 }, { "epoch": 0.1, "learning_rate": 1.9719646051615368e-05, "loss": 1.1107, "step": 1552 }, { "epoch": 0.1, "learning_rate": 1.9719138947345632e-05, "loss": 1.109, "step": 1553 }, { "epoch": 0.1, "learning_rate": 1.9718631391397207e-05, "loss": 1.0987, "step": 1554 }, { "epoch": 0.1, "learning_rate": 1.9718123383793674e-05, "loss": 1.089, "step": 1555 }, { "epoch": 0.1, "learning_rate": 1.9717614924558638e-05, "loss": 1.1205, "step": 1556 }, { "epoch": 0.1, "learning_rate": 1.9717106013715738e-05, "loss": 1.0925, "step": 1557 }, { "epoch": 0.1, "learning_rate": 1.9716596651288623e-05, "loss": 1.0778, "step": 1558 }, { "epoch": 0.1, "learning_rate": 1.971608683730096e-05, "loss": 1.1099, "step": 1559 }, { "epoch": 0.1, "learning_rate": 1.9715576571776448e-05, "loss": 1.0794, "step": 1560 }, { "epoch": 0.1, "learning_rate": 1.9715065854738794e-05, "loss": 1.0804, "step": 1561 }, { "epoch": 0.1, "learning_rate": 1.9714554686211735e-05, "loss": 1.1309, "step": 1562 }, { "epoch": 0.1, "learning_rate": 1.971404306621903e-05, "loss": 1.115, "step": 1563 }, { "epoch": 0.1, "learning_rate": 1.971353099478445e-05, "loss": 1.0819, "step": 1564 }, { "epoch": 0.1, "learning_rate": 1.9713018471931794e-05, "loss": 1.124, "step": 1565 }, { "epoch": 0.1, "learning_rate": 1.9712505497684883e-05, "loss": 1.1144, "step": 1566 }, { "epoch": 0.1, "learning_rate": 1.9711992072067557e-05, "loss": 1.1079, "step": 1567 }, { "epoch": 0.1, "learning_rate": 1.9711478195103674e-05, "loss": 1.1008, "step": 1568 }, { "epoch": 0.1, "learning_rate": 1.9710963866817113e-05, "loss": 1.0866, "step": 1569 }, { "epoch": 0.1, "learning_rate": 1.9710449087231784e-05, "loss": 1.0418, "step": 1570 }, { "epoch": 0.1, "learning_rate": 1.9709933856371607e-05, "loss": 1.1279, "step": 1571 }, { "epoch": 0.1, "learning_rate": 1.9709418174260523e-05, "loss": 1.0629, "step": 1572 }, { "epoch": 0.1, "learning_rate": 1.97089020409225e-05, "loss": 1.1187, "step": 1573 }, { "epoch": 0.1, "learning_rate": 1.9708385456381523e-05, "loss": 1.1214, "step": 1574 }, { "epoch": 0.1, "learning_rate": 1.9707868420661604e-05, "loss": 1.0836, "step": 1575 }, { "epoch": 0.1, "learning_rate": 1.9707350933786763e-05, "loss": 1.129, "step": 1576 }, { "epoch": 0.1, "learning_rate": 1.9706832995781057e-05, "loss": 1.0825, "step": 1577 }, { "epoch": 0.11, "learning_rate": 1.9706314606668554e-05, "loss": 1.0869, "step": 1578 }, { "epoch": 0.11, "learning_rate": 1.9705795766473344e-05, "loss": 1.1001, "step": 1579 }, { "epoch": 0.11, "learning_rate": 1.9705276475219538e-05, "loss": 1.0973, "step": 1580 }, { "epoch": 0.11, "learning_rate": 1.9704756732931268e-05, "loss": 1.1276, "step": 1581 }, { "epoch": 0.11, "learning_rate": 1.9704236539632695e-05, "loss": 1.1453, "step": 1582 }, { "epoch": 0.11, "learning_rate": 1.970371589534799e-05, "loss": 1.181, "step": 1583 }, { "epoch": 0.11, "learning_rate": 1.9703194800101344e-05, "loss": 1.0784, "step": 1584 }, { "epoch": 0.11, "learning_rate": 1.970267325391698e-05, "loss": 1.0922, "step": 1585 }, { "epoch": 0.11, "learning_rate": 1.9702151256819138e-05, "loss": 1.0617, "step": 1586 }, { "epoch": 0.11, "learning_rate": 1.970162880883207e-05, "loss": 1.1291, "step": 1587 }, { "epoch": 0.11, "learning_rate": 1.970110590998006e-05, "loss": 1.0834, "step": 1588 }, { "epoch": 0.11, "learning_rate": 1.970058256028741e-05, "loss": 1.1183, "step": 1589 }, { "epoch": 0.11, "learning_rate": 1.9700058759778433e-05, "loss": 1.0548, "step": 1590 }, { "epoch": 0.11, "learning_rate": 1.9699534508477483e-05, "loss": 1.0844, "step": 1591 }, { "epoch": 0.11, "learning_rate": 1.969900980640892e-05, "loss": 1.1093, "step": 1592 }, { "epoch": 0.11, "learning_rate": 1.969848465359712e-05, "loss": 1.0851, "step": 1593 }, { "epoch": 0.11, "learning_rate": 1.9697959050066503e-05, "loss": 1.0678, "step": 1594 }, { "epoch": 0.11, "learning_rate": 1.9697432995841484e-05, "loss": 1.1046, "step": 1595 }, { "epoch": 0.11, "learning_rate": 1.9696906490946514e-05, "loss": 1.0982, "step": 1596 }, { "epoch": 0.11, "learning_rate": 1.9696379535406066e-05, "loss": 1.09, "step": 1597 }, { "epoch": 0.11, "learning_rate": 1.969585212924462e-05, "loss": 1.1043, "step": 1598 }, { "epoch": 0.11, "learning_rate": 1.9695324272486694e-05, "loss": 1.0741, "step": 1599 }, { "epoch": 0.11, "learning_rate": 1.9694795965156818e-05, "loss": 1.0695, "step": 1600 }, { "epoch": 0.11, "learning_rate": 1.9694267207279542e-05, "loss": 1.1089, "step": 1601 }, { "epoch": 0.11, "learning_rate": 1.9693737998879438e-05, "loss": 1.0504, "step": 1602 }, { "epoch": 0.11, "learning_rate": 1.96932083399811e-05, "loss": 1.117, "step": 1603 }, { "epoch": 0.11, "learning_rate": 1.9692678230609147e-05, "loss": 1.0428, "step": 1604 }, { "epoch": 0.11, "learning_rate": 1.969214767078821e-05, "loss": 1.1296, "step": 1605 }, { "epoch": 0.11, "learning_rate": 1.969161666054295e-05, "loss": 1.0784, "step": 1606 }, { "epoch": 0.11, "learning_rate": 1.9691085199898045e-05, "loss": 1.123, "step": 1607 }, { "epoch": 0.11, "learning_rate": 1.9690553288878187e-05, "loss": 1.0958, "step": 1608 }, { "epoch": 0.11, "learning_rate": 1.96900209275081e-05, "loss": 1.1115, "step": 1609 }, { "epoch": 0.11, "learning_rate": 1.9689488115812527e-05, "loss": 1.0717, "step": 1610 }, { "epoch": 0.11, "learning_rate": 1.9688954853816226e-05, "loss": 1.1014, "step": 1611 }, { "epoch": 0.11, "learning_rate": 1.968842114154398e-05, "loss": 1.0564, "step": 1612 }, { "epoch": 0.11, "learning_rate": 1.9687886979020593e-05, "loss": 1.1016, "step": 1613 }, { "epoch": 0.11, "learning_rate": 1.9687352366270886e-05, "loss": 1.0854, "step": 1614 }, { "epoch": 0.11, "learning_rate": 1.968681730331971e-05, "loss": 1.1108, "step": 1615 }, { "epoch": 0.11, "learning_rate": 1.9686281790191927e-05, "loss": 1.0903, "step": 1616 }, { "epoch": 0.11, "learning_rate": 1.9685745826912425e-05, "loss": 1.1453, "step": 1617 }, { "epoch": 0.11, "learning_rate": 1.9685209413506107e-05, "loss": 1.0892, "step": 1618 }, { "epoch": 0.11, "learning_rate": 1.968467254999791e-05, "loss": 1.069, "step": 1619 }, { "epoch": 0.11, "learning_rate": 1.9684135236412783e-05, "loss": 1.0964, "step": 1620 }, { "epoch": 0.11, "learning_rate": 1.9683597472775692e-05, "loss": 1.1262, "step": 1621 }, { "epoch": 0.11, "learning_rate": 1.968305925911163e-05, "loss": 1.0792, "step": 1622 }, { "epoch": 0.11, "learning_rate": 1.968252059544561e-05, "loss": 1.1201, "step": 1623 }, { "epoch": 0.11, "learning_rate": 1.9681981481802664e-05, "loss": 1.1153, "step": 1624 }, { "epoch": 0.11, "learning_rate": 1.9681441918207848e-05, "loss": 1.1071, "step": 1625 }, { "epoch": 0.11, "learning_rate": 1.9680901904686237e-05, "loss": 1.0883, "step": 1626 }, { "epoch": 0.11, "learning_rate": 1.9680361441262924e-05, "loss": 1.1506, "step": 1627 }, { "epoch": 0.11, "learning_rate": 1.9679820527963035e-05, "loss": 1.0979, "step": 1628 }, { "epoch": 0.11, "learning_rate": 1.96792791648117e-05, "loss": 1.1198, "step": 1629 }, { "epoch": 0.11, "learning_rate": 1.9678737351834077e-05, "loss": 1.0803, "step": 1630 }, { "epoch": 0.11, "learning_rate": 1.9678195089055347e-05, "loss": 1.1119, "step": 1631 }, { "epoch": 0.11, "learning_rate": 1.9677652376500716e-05, "loss": 1.0839, "step": 1632 }, { "epoch": 0.11, "learning_rate": 1.9677109214195398e-05, "loss": 1.1063, "step": 1633 }, { "epoch": 0.11, "learning_rate": 1.9676565602164645e-05, "loss": 1.0533, "step": 1634 }, { "epoch": 0.11, "learning_rate": 1.967602154043371e-05, "loss": 1.0957, "step": 1635 }, { "epoch": 0.11, "learning_rate": 1.967547702902788e-05, "loss": 1.0939, "step": 1636 }, { "epoch": 0.11, "learning_rate": 1.9674932067972464e-05, "loss": 1.1077, "step": 1637 }, { "epoch": 0.11, "learning_rate": 1.9674386657292784e-05, "loss": 1.072, "step": 1638 }, { "epoch": 0.11, "learning_rate": 1.967384079701419e-05, "loss": 1.1129, "step": 1639 }, { "epoch": 0.11, "learning_rate": 1.9673294487162045e-05, "loss": 1.0756, "step": 1640 }, { "epoch": 0.11, "learning_rate": 1.9672747727761747e-05, "loss": 1.0873, "step": 1641 }, { "epoch": 0.11, "learning_rate": 1.9672200518838696e-05, "loss": 1.0732, "step": 1642 }, { "epoch": 0.11, "learning_rate": 1.967165286041833e-05, "loss": 1.0778, "step": 1643 }, { "epoch": 0.11, "learning_rate": 1.9671104752526093e-05, "loss": 1.026, "step": 1644 }, { "epoch": 0.11, "learning_rate": 1.9670556195187464e-05, "loss": 1.1073, "step": 1645 }, { "epoch": 0.11, "learning_rate": 1.967000718842793e-05, "loss": 1.1082, "step": 1646 }, { "epoch": 0.11, "learning_rate": 1.966945773227301e-05, "loss": 1.1126, "step": 1647 }, { "epoch": 0.11, "learning_rate": 1.966890782674824e-05, "loss": 1.0761, "step": 1648 }, { "epoch": 0.11, "learning_rate": 1.966835747187917e-05, "loss": 1.0989, "step": 1649 }, { "epoch": 0.11, "learning_rate": 1.966780666769138e-05, "loss": 1.0899, "step": 1650 }, { "epoch": 0.11, "learning_rate": 1.9667255414210466e-05, "loss": 1.0978, "step": 1651 }, { "epoch": 0.11, "learning_rate": 1.9666703711462052e-05, "loss": 1.1244, "step": 1652 }, { "epoch": 0.11, "learning_rate": 1.9666151559471768e-05, "loss": 1.0613, "step": 1653 }, { "epoch": 0.11, "learning_rate": 1.9665598958265285e-05, "loss": 1.052, "step": 1654 }, { "epoch": 0.11, "learning_rate": 1.9665045907868272e-05, "loss": 1.1131, "step": 1655 }, { "epoch": 0.11, "learning_rate": 1.966449240830644e-05, "loss": 1.0575, "step": 1656 }, { "epoch": 0.11, "learning_rate": 1.9663938459605512e-05, "loss": 1.0933, "step": 1657 }, { "epoch": 0.11, "learning_rate": 1.9663384061791228e-05, "loss": 1.0823, "step": 1658 }, { "epoch": 0.11, "learning_rate": 1.9662829214889352e-05, "loss": 1.0711, "step": 1659 }, { "epoch": 0.11, "learning_rate": 1.966227391892567e-05, "loss": 1.0735, "step": 1660 }, { "epoch": 0.11, "learning_rate": 1.9661718173925992e-05, "loss": 1.112, "step": 1661 }, { "epoch": 0.11, "learning_rate": 1.966116197991614e-05, "loss": 1.0584, "step": 1662 }, { "epoch": 0.11, "learning_rate": 1.9660605336921968e-05, "loss": 1.1502, "step": 1663 }, { "epoch": 0.11, "learning_rate": 1.9660048244969338e-05, "loss": 1.0776, "step": 1664 }, { "epoch": 0.11, "learning_rate": 1.9659490704084147e-05, "loss": 1.108, "step": 1665 }, { "epoch": 0.11, "learning_rate": 1.96589327142923e-05, "loss": 1.1036, "step": 1666 }, { "epoch": 0.11, "learning_rate": 1.965837427561973e-05, "loss": 1.1163, "step": 1667 }, { "epoch": 0.11, "learning_rate": 1.9657815388092395e-05, "loss": 1.0732, "step": 1668 }, { "epoch": 0.11, "learning_rate": 1.9657256051736256e-05, "loss": 1.1527, "step": 1669 }, { "epoch": 0.11, "learning_rate": 1.9656696266577318e-05, "loss": 1.0562, "step": 1670 }, { "epoch": 0.11, "learning_rate": 1.9656136032641592e-05, "loss": 1.1007, "step": 1671 }, { "epoch": 0.11, "learning_rate": 1.9655575349955115e-05, "loss": 1.0829, "step": 1672 }, { "epoch": 0.11, "learning_rate": 1.965501421854394e-05, "loss": 1.1046, "step": 1673 }, { "epoch": 0.11, "learning_rate": 1.9654452638434152e-05, "loss": 1.0746, "step": 1674 }, { "epoch": 0.11, "learning_rate": 1.965389060965184e-05, "loss": 1.1004, "step": 1675 }, { "epoch": 0.11, "learning_rate": 1.9653328132223127e-05, "loss": 1.0797, "step": 1676 }, { "epoch": 0.11, "learning_rate": 1.965276520617416e-05, "loss": 1.1052, "step": 1677 }, { "epoch": 0.11, "learning_rate": 1.965220183153109e-05, "loss": 1.0683, "step": 1678 }, { "epoch": 0.11, "learning_rate": 1.9651638008320104e-05, "loss": 1.1233, "step": 1679 }, { "epoch": 0.11, "learning_rate": 1.9651073736567405e-05, "loss": 1.0642, "step": 1680 }, { "epoch": 0.11, "learning_rate": 1.965050901629921e-05, "loss": 1.1023, "step": 1681 }, { "epoch": 0.11, "learning_rate": 1.964994384754177e-05, "loss": 1.0749, "step": 1682 }, { "epoch": 0.11, "learning_rate": 1.964937823032135e-05, "loss": 1.1103, "step": 1683 }, { "epoch": 0.11, "learning_rate": 1.9648812164664232e-05, "loss": 1.0971, "step": 1684 }, { "epoch": 0.11, "learning_rate": 1.9648245650596726e-05, "loss": 1.0278, "step": 1685 }, { "epoch": 0.11, "learning_rate": 1.9647678688145163e-05, "loss": 1.0861, "step": 1686 }, { "epoch": 0.11, "learning_rate": 1.964711127733588e-05, "loss": 1.0973, "step": 1687 }, { "epoch": 0.11, "learning_rate": 1.964654341819526e-05, "loss": 1.0548, "step": 1688 }, { "epoch": 0.11, "learning_rate": 1.9645975110749683e-05, "loss": 1.1096, "step": 1689 }, { "epoch": 0.11, "learning_rate": 1.9645406355025565e-05, "loss": 1.1114, "step": 1690 }, { "epoch": 0.11, "learning_rate": 1.9644837151049336e-05, "loss": 1.126, "step": 1691 }, { "epoch": 0.11, "learning_rate": 1.964426749884745e-05, "loss": 1.0439, "step": 1692 }, { "epoch": 0.11, "learning_rate": 1.964369739844638e-05, "loss": 1.0832, "step": 1693 }, { "epoch": 0.11, "learning_rate": 1.964312684987262e-05, "loss": 1.0904, "step": 1694 }, { "epoch": 0.11, "learning_rate": 1.9642555853152687e-05, "loss": 1.1486, "step": 1695 }, { "epoch": 0.11, "learning_rate": 1.9641984408313112e-05, "loss": 1.0866, "step": 1696 }, { "epoch": 0.11, "learning_rate": 1.9641412515380457e-05, "loss": 1.0751, "step": 1697 }, { "epoch": 0.11, "learning_rate": 1.9640840174381298e-05, "loss": 1.0251, "step": 1698 }, { "epoch": 0.11, "learning_rate": 1.9640267385342232e-05, "loss": 1.1017, "step": 1699 }, { "epoch": 0.11, "learning_rate": 1.9639694148289883e-05, "loss": 1.0998, "step": 1700 }, { "epoch": 0.11, "learning_rate": 1.9639120463250885e-05, "loss": 1.1285, "step": 1701 }, { "epoch": 0.11, "learning_rate": 1.9638546330251903e-05, "loss": 1.0931, "step": 1702 }, { "epoch": 0.11, "learning_rate": 1.9637971749319614e-05, "loss": 1.0735, "step": 1703 }, { "epoch": 0.11, "learning_rate": 1.9637396720480728e-05, "loss": 1.0505, "step": 1704 }, { "epoch": 0.11, "learning_rate": 1.9636821243761966e-05, "loss": 1.0651, "step": 1705 }, { "epoch": 0.11, "learning_rate": 1.9636245319190065e-05, "loss": 1.0745, "step": 1706 }, { "epoch": 0.11, "learning_rate": 1.96356689467918e-05, "loss": 1.0749, "step": 1707 }, { "epoch": 0.11, "learning_rate": 1.9635092126593952e-05, "loss": 1.1033, "step": 1708 }, { "epoch": 0.11, "learning_rate": 1.9634514858623327e-05, "loss": 1.0996, "step": 1709 }, { "epoch": 0.11, "learning_rate": 1.963393714290675e-05, "loss": 1.1084, "step": 1710 }, { "epoch": 0.11, "learning_rate": 1.9633358979471078e-05, "loss": 1.1206, "step": 1711 }, { "epoch": 0.11, "learning_rate": 1.9632780368343175e-05, "loss": 1.0862, "step": 1712 }, { "epoch": 0.11, "learning_rate": 1.9632201309549926e-05, "loss": 1.0992, "step": 1713 }, { "epoch": 0.11, "learning_rate": 1.963162180311825e-05, "loss": 1.1297, "step": 1714 }, { "epoch": 0.11, "learning_rate": 1.9631041849075074e-05, "loss": 1.0722, "step": 1715 }, { "epoch": 0.11, "learning_rate": 1.9630461447447354e-05, "loss": 1.1313, "step": 1716 }, { "epoch": 0.11, "learning_rate": 1.9629880598262056e-05, "loss": 1.0855, "step": 1717 }, { "epoch": 0.11, "learning_rate": 1.9629299301546178e-05, "loss": 1.0805, "step": 1718 }, { "epoch": 0.11, "learning_rate": 1.9628717557326734e-05, "loss": 1.1202, "step": 1719 }, { "epoch": 0.11, "learning_rate": 1.9628135365630766e-05, "loss": 1.07, "step": 1720 }, { "epoch": 0.11, "learning_rate": 1.962755272648532e-05, "loss": 1.0552, "step": 1721 }, { "epoch": 0.11, "learning_rate": 1.9626969639917476e-05, "loss": 1.0509, "step": 1722 }, { "epoch": 0.11, "learning_rate": 1.9626386105954336e-05, "loss": 1.0869, "step": 1723 }, { "epoch": 0.11, "learning_rate": 1.9625802124623015e-05, "loss": 1.0586, "step": 1724 }, { "epoch": 0.11, "learning_rate": 1.9625217695950655e-05, "loss": 1.0621, "step": 1725 }, { "epoch": 0.11, "learning_rate": 1.962463281996441e-05, "loss": 1.0951, "step": 1726 }, { "epoch": 0.11, "learning_rate": 1.9624047496691473e-05, "loss": 1.0865, "step": 1727 }, { "epoch": 0.12, "learning_rate": 1.962346172615903e-05, "loss": 1.0624, "step": 1728 }, { "epoch": 0.12, "learning_rate": 1.9622875508394318e-05, "loss": 1.1076, "step": 1729 }, { "epoch": 0.12, "learning_rate": 1.9622288843424574e-05, "loss": 1.0679, "step": 1730 }, { "epoch": 0.12, "learning_rate": 1.962170173127706e-05, "loss": 1.0919, "step": 1731 }, { "epoch": 0.12, "learning_rate": 1.962111417197907e-05, "loss": 1.0791, "step": 1732 }, { "epoch": 0.12, "learning_rate": 1.9620526165557893e-05, "loss": 1.0777, "step": 1733 }, { "epoch": 0.12, "learning_rate": 1.9619937712040872e-05, "loss": 1.0595, "step": 1734 }, { "epoch": 0.12, "learning_rate": 1.961934881145535e-05, "loss": 1.1282, "step": 1735 }, { "epoch": 0.12, "learning_rate": 1.9618759463828686e-05, "loss": 1.0599, "step": 1736 }, { "epoch": 0.12, "learning_rate": 1.961816966918828e-05, "loss": 1.1268, "step": 1737 }, { "epoch": 0.12, "learning_rate": 1.9617579427561536e-05, "loss": 1.0793, "step": 1738 }, { "epoch": 0.12, "learning_rate": 1.961698873897589e-05, "loss": 1.1521, "step": 1739 }, { "epoch": 0.12, "learning_rate": 1.9616397603458783e-05, "loss": 1.0909, "step": 1740 }, { "epoch": 0.12, "learning_rate": 1.96158060210377e-05, "loss": 1.0994, "step": 1741 }, { "epoch": 0.12, "learning_rate": 1.9615213991740117e-05, "loss": 1.074, "step": 1742 }, { "epoch": 0.12, "learning_rate": 1.9614621515593564e-05, "loss": 1.1076, "step": 1743 }, { "epoch": 0.12, "learning_rate": 1.9614028592625564e-05, "loss": 1.0704, "step": 1744 }, { "epoch": 0.12, "learning_rate": 1.961343522286368e-05, "loss": 1.0688, "step": 1745 }, { "epoch": 0.12, "learning_rate": 1.961284140633548e-05, "loss": 1.0431, "step": 1746 }, { "epoch": 0.12, "learning_rate": 1.961224714306857e-05, "loss": 1.0822, "step": 1747 }, { "epoch": 0.12, "learning_rate": 1.9611652433090557e-05, "loss": 1.0454, "step": 1748 }, { "epoch": 0.12, "learning_rate": 1.9611057276429085e-05, "loss": 1.0868, "step": 1749 }, { "epoch": 0.12, "learning_rate": 1.9610461673111812e-05, "loss": 1.1093, "step": 1750 }, { "epoch": 0.12, "learning_rate": 1.9609865623166416e-05, "loss": 1.1079, "step": 1751 }, { "epoch": 0.12, "learning_rate": 1.96092691266206e-05, "loss": 1.0552, "step": 1752 }, { "epoch": 0.12, "learning_rate": 1.9608672183502083e-05, "loss": 1.151, "step": 1753 }, { "epoch": 0.12, "learning_rate": 1.9608074793838606e-05, "loss": 1.1083, "step": 1754 }, { "epoch": 0.12, "learning_rate": 1.9607476957657933e-05, "loss": 1.0646, "step": 1755 }, { "epoch": 0.12, "learning_rate": 1.9606878674987847e-05, "loss": 1.0563, "step": 1756 }, { "epoch": 0.12, "learning_rate": 1.9606279945856154e-05, "loss": 1.1141, "step": 1757 }, { "epoch": 0.12, "learning_rate": 1.9605680770290674e-05, "loss": 1.093, "step": 1758 }, { "epoch": 0.12, "learning_rate": 1.9605081148319258e-05, "loss": 1.115, "step": 1759 }, { "epoch": 0.12, "learning_rate": 1.960448107996977e-05, "loss": 1.0583, "step": 1760 }, { "epoch": 0.12, "learning_rate": 1.9603880565270096e-05, "loss": 1.0954, "step": 1761 }, { "epoch": 0.12, "learning_rate": 1.9603279604248146e-05, "loss": 1.0909, "step": 1762 }, { "epoch": 0.12, "learning_rate": 1.960267819693184e-05, "loss": 1.1206, "step": 1763 }, { "epoch": 0.12, "learning_rate": 1.9602076343349143e-05, "loss": 1.0672, "step": 1764 }, { "epoch": 0.12, "learning_rate": 1.9601474043528014e-05, "loss": 1.071, "step": 1765 }, { "epoch": 0.12, "learning_rate": 1.9600871297496445e-05, "loss": 1.0764, "step": 1766 }, { "epoch": 0.12, "learning_rate": 1.960026810528245e-05, "loss": 1.1014, "step": 1767 }, { "epoch": 0.12, "learning_rate": 1.959966446691406e-05, "loss": 1.0773, "step": 1768 }, { "epoch": 0.12, "learning_rate": 1.9599060382419327e-05, "loss": 1.0917, "step": 1769 }, { "epoch": 0.12, "learning_rate": 1.9598455851826323e-05, "loss": 1.0823, "step": 1770 }, { "epoch": 0.12, "learning_rate": 1.959785087516315e-05, "loss": 1.1081, "step": 1771 }, { "epoch": 0.12, "learning_rate": 1.9597245452457915e-05, "loss": 1.0403, "step": 1772 }, { "epoch": 0.12, "learning_rate": 1.9596639583738757e-05, "loss": 1.041, "step": 1773 }, { "epoch": 0.12, "learning_rate": 1.9596033269033833e-05, "loss": 1.0519, "step": 1774 }, { "epoch": 0.12, "learning_rate": 1.9595426508371323e-05, "loss": 1.0881, "step": 1775 }, { "epoch": 0.12, "learning_rate": 1.9594819301779417e-05, "loss": 1.0672, "step": 1776 }, { "epoch": 0.12, "learning_rate": 1.9594211649286343e-05, "loss": 1.0548, "step": 1777 }, { "epoch": 0.12, "learning_rate": 1.9593603550920335e-05, "loss": 1.0665, "step": 1778 }, { "epoch": 0.12, "learning_rate": 1.9592995006709653e-05, "loss": 1.0837, "step": 1779 }, { "epoch": 0.12, "learning_rate": 1.9592386016682582e-05, "loss": 1.0558, "step": 1780 }, { "epoch": 0.12, "learning_rate": 1.9591776580867418e-05, "loss": 1.0875, "step": 1781 }, { "epoch": 0.12, "learning_rate": 1.9591166699292488e-05, "loss": 1.082, "step": 1782 }, { "epoch": 0.12, "learning_rate": 1.9590556371986135e-05, "loss": 1.1075, "step": 1783 }, { "epoch": 0.12, "learning_rate": 1.958994559897672e-05, "loss": 1.0444, "step": 1784 }, { "epoch": 0.12, "learning_rate": 1.9589334380292628e-05, "loss": 1.0843, "step": 1785 }, { "epoch": 0.12, "learning_rate": 1.9588722715962267e-05, "loss": 1.0719, "step": 1786 }, { "epoch": 0.12, "learning_rate": 1.9588110606014063e-05, "loss": 1.1562, "step": 1787 }, { "epoch": 0.12, "learning_rate": 1.9587498050476458e-05, "loss": 1.1075, "step": 1788 }, { "epoch": 0.12, "learning_rate": 1.958688504937792e-05, "loss": 1.0672, "step": 1789 }, { "epoch": 0.12, "learning_rate": 1.9586271602746944e-05, "loss": 1.0591, "step": 1790 }, { "epoch": 0.12, "learning_rate": 1.958565771061203e-05, "loss": 1.0889, "step": 1791 }, { "epoch": 0.12, "learning_rate": 1.9585043373001716e-05, "loss": 1.0795, "step": 1792 }, { "epoch": 0.12, "learning_rate": 1.9584428589944548e-05, "loss": 1.1197, "step": 1793 }, { "epoch": 0.12, "learning_rate": 1.958381336146909e-05, "loss": 1.0768, "step": 1794 }, { "epoch": 0.12, "learning_rate": 1.9583197687603946e-05, "loss": 1.082, "step": 1795 }, { "epoch": 0.12, "learning_rate": 1.9582581568377718e-05, "loss": 1.1149, "step": 1796 }, { "epoch": 0.12, "learning_rate": 1.9581965003819045e-05, "loss": 1.055, "step": 1797 }, { "epoch": 0.12, "learning_rate": 1.9581347993956582e-05, "loss": 1.0866, "step": 1798 }, { "epoch": 0.12, "learning_rate": 1.9580730538819e-05, "loss": 1.09, "step": 1799 }, { "epoch": 0.12, "learning_rate": 1.9580112638434992e-05, "loss": 1.0798, "step": 1800 }, { "epoch": 0.12, "learning_rate": 1.9579494292833278e-05, "loss": 1.0776, "step": 1801 }, { "epoch": 0.12, "learning_rate": 1.9578875502042592e-05, "loss": 1.1621, "step": 1802 }, { "epoch": 0.12, "learning_rate": 1.9578256266091692e-05, "loss": 1.1195, "step": 1803 }, { "epoch": 0.12, "learning_rate": 1.9577636585009355e-05, "loss": 1.1008, "step": 1804 }, { "epoch": 0.12, "learning_rate": 1.957701645882438e-05, "loss": 1.0941, "step": 1805 }, { "epoch": 0.12, "learning_rate": 1.957639588756559e-05, "loss": 1.0713, "step": 1806 }, { "epoch": 0.12, "learning_rate": 1.957577487126182e-05, "loss": 1.0907, "step": 1807 }, { "epoch": 0.12, "learning_rate": 1.957515340994193e-05, "loss": 1.0615, "step": 1808 }, { "epoch": 0.12, "learning_rate": 1.95745315036348e-05, "loss": 1.0822, "step": 1809 }, { "epoch": 0.12, "learning_rate": 1.9573909152369342e-05, "loss": 1.0572, "step": 1810 }, { "epoch": 0.12, "learning_rate": 1.9573286356174467e-05, "loss": 1.0898, "step": 1811 }, { "epoch": 0.12, "learning_rate": 1.9572663115079125e-05, "loss": 1.0465, "step": 1812 }, { "epoch": 0.12, "learning_rate": 1.9572039429112277e-05, "loss": 1.084, "step": 1813 }, { "epoch": 0.12, "learning_rate": 1.9571415298302908e-05, "loss": 1.1028, "step": 1814 }, { "epoch": 0.12, "learning_rate": 1.9570790722680025e-05, "loss": 1.0973, "step": 1815 }, { "epoch": 0.12, "learning_rate": 1.9570165702272654e-05, "loss": 1.0454, "step": 1816 }, { "epoch": 0.12, "learning_rate": 1.9569540237109837e-05, "loss": 1.0767, "step": 1817 }, { "epoch": 0.12, "learning_rate": 1.956891432722065e-05, "loss": 1.046, "step": 1818 }, { "epoch": 0.12, "learning_rate": 1.956828797263417e-05, "loss": 1.0891, "step": 1819 }, { "epoch": 0.12, "learning_rate": 1.9567661173379515e-05, "loss": 1.0769, "step": 1820 }, { "epoch": 0.12, "learning_rate": 1.956703392948581e-05, "loss": 1.0655, "step": 1821 }, { "epoch": 0.12, "learning_rate": 1.9566406240982204e-05, "loss": 1.0508, "step": 1822 }, { "epoch": 0.12, "learning_rate": 1.9565778107897874e-05, "loss": 1.0737, "step": 1823 }, { "epoch": 0.12, "learning_rate": 1.9565149530262003e-05, "loss": 1.0448, "step": 1824 }, { "epoch": 0.12, "learning_rate": 1.9564520508103806e-05, "loss": 1.052, "step": 1825 }, { "epoch": 0.12, "learning_rate": 1.9563891041452515e-05, "loss": 1.0786, "step": 1826 }, { "epoch": 0.12, "learning_rate": 1.9563261130337387e-05, "loss": 1.109, "step": 1827 }, { "epoch": 0.12, "learning_rate": 1.9562630774787695e-05, "loss": 1.067, "step": 1828 }, { "epoch": 0.12, "learning_rate": 1.9561999974832727e-05, "loss": 1.0787, "step": 1829 }, { "epoch": 0.12, "learning_rate": 1.9561368730501808e-05, "loss": 1.0739, "step": 1830 }, { "epoch": 0.12, "learning_rate": 1.9560737041824266e-05, "loss": 1.1222, "step": 1831 }, { "epoch": 0.12, "learning_rate": 1.9560104908829466e-05, "loss": 1.0806, "step": 1832 }, { "epoch": 0.12, "learning_rate": 1.9559472331546776e-05, "loss": 1.1562, "step": 1833 }, { "epoch": 0.12, "learning_rate": 1.95588393100056e-05, "loss": 1.0634, "step": 1834 }, { "epoch": 0.12, "learning_rate": 1.955820584423535e-05, "loss": 1.0952, "step": 1835 }, { "epoch": 0.12, "learning_rate": 1.955757193426547e-05, "loss": 1.0793, "step": 1836 }, { "epoch": 0.12, "learning_rate": 1.9556937580125422e-05, "loss": 1.0962, "step": 1837 }, { "epoch": 0.12, "learning_rate": 1.955630278184468e-05, "loss": 1.0786, "step": 1838 }, { "epoch": 0.12, "learning_rate": 1.9555667539452753e-05, "loss": 1.0893, "step": 1839 }, { "epoch": 0.12, "learning_rate": 1.9555031852979156e-05, "loss": 1.0547, "step": 1840 }, { "epoch": 0.12, "learning_rate": 1.9554395722453435e-05, "loss": 1.1086, "step": 1841 }, { "epoch": 0.12, "learning_rate": 1.9553759147905146e-05, "loss": 1.043, "step": 1842 }, { "epoch": 0.12, "learning_rate": 1.9553122129363884e-05, "loss": 1.0891, "step": 1843 }, { "epoch": 0.12, "learning_rate": 1.9552484666859247e-05, "loss": 1.0994, "step": 1844 }, { "epoch": 0.12, "learning_rate": 1.955184676042086e-05, "loss": 1.0898, "step": 1845 }, { "epoch": 0.12, "learning_rate": 1.955120841007837e-05, "loss": 1.0518, "step": 1846 }, { "epoch": 0.12, "learning_rate": 1.9550569615861438e-05, "loss": 1.0701, "step": 1847 }, { "epoch": 0.12, "learning_rate": 1.9549930377799757e-05, "loss": 1.0805, "step": 1848 }, { "epoch": 0.12, "learning_rate": 1.954929069592303e-05, "loss": 1.0901, "step": 1849 }, { "epoch": 0.12, "learning_rate": 1.954865057026099e-05, "loss": 1.035, "step": 1850 }, { "epoch": 0.12, "learning_rate": 1.954801000084338e-05, "loss": 1.0982, "step": 1851 }, { "epoch": 0.12, "learning_rate": 1.9547368987699974e-05, "loss": 1.063, "step": 1852 }, { "epoch": 0.12, "learning_rate": 1.954672753086056e-05, "loss": 1.1147, "step": 1853 }, { "epoch": 0.12, "learning_rate": 1.9546085630354948e-05, "loss": 1.0803, "step": 1854 }, { "epoch": 0.12, "learning_rate": 1.9545443286212967e-05, "loss": 1.0927, "step": 1855 }, { "epoch": 0.12, "learning_rate": 1.9544800498464476e-05, "loss": 1.0561, "step": 1856 }, { "epoch": 0.12, "learning_rate": 1.9544157267139342e-05, "loss": 1.0973, "step": 1857 }, { "epoch": 0.12, "learning_rate": 1.9543513592267452e-05, "loss": 1.0764, "step": 1858 }, { "epoch": 0.12, "learning_rate": 1.9542869473878733e-05, "loss": 1.1206, "step": 1859 }, { "epoch": 0.12, "learning_rate": 1.9542224912003108e-05, "loss": 1.0539, "step": 1860 }, { "epoch": 0.12, "learning_rate": 1.954157990667054e-05, "loss": 1.0495, "step": 1861 }, { "epoch": 0.12, "learning_rate": 1.9540934457911e-05, "loss": 1.0665, "step": 1862 }, { "epoch": 0.12, "learning_rate": 1.9540288565754482e-05, "loss": 1.0349, "step": 1863 }, { "epoch": 0.12, "learning_rate": 1.9539642230231006e-05, "loss": 1.0538, "step": 1864 }, { "epoch": 0.12, "learning_rate": 1.953899545137061e-05, "loss": 1.0753, "step": 1865 }, { "epoch": 0.12, "learning_rate": 1.9538348229203348e-05, "loss": 1.0682, "step": 1866 }, { "epoch": 0.12, "learning_rate": 1.9537700563759303e-05, "loss": 1.0788, "step": 1867 }, { "epoch": 0.12, "learning_rate": 1.953705245506857e-05, "loss": 1.0962, "step": 1868 }, { "epoch": 0.12, "learning_rate": 1.953640390316127e-05, "loss": 1.1118, "step": 1869 }, { "epoch": 0.12, "learning_rate": 1.9535754908067544e-05, "loss": 1.097, "step": 1870 }, { "epoch": 0.12, "learning_rate": 1.9535105469817553e-05, "loss": 1.0924, "step": 1871 }, { "epoch": 0.12, "learning_rate": 1.953445558844148e-05, "loss": 1.0598, "step": 1872 }, { "epoch": 0.12, "learning_rate": 1.953380526396952e-05, "loss": 1.0496, "step": 1873 }, { "epoch": 0.12, "learning_rate": 1.9533154496431904e-05, "loss": 1.0312, "step": 1874 }, { "epoch": 0.12, "learning_rate": 1.953250328585887e-05, "loss": 1.063, "step": 1875 }, { "epoch": 0.12, "learning_rate": 1.953185163228068e-05, "loss": 1.0672, "step": 1876 }, { "epoch": 0.12, "learning_rate": 1.953119953572763e-05, "loss": 1.1105, "step": 1877 }, { "epoch": 0.12, "learning_rate": 1.953054699623001e-05, "loss": 1.0267, "step": 1878 }, { "epoch": 0.13, "learning_rate": 1.952989401381815e-05, "loss": 1.0991, "step": 1879 }, { "epoch": 0.13, "learning_rate": 1.9529240588522404e-05, "loss": 1.1094, "step": 1880 }, { "epoch": 0.13, "learning_rate": 1.9528586720373136e-05, "loss": 1.1505, "step": 1881 }, { "epoch": 0.13, "learning_rate": 1.952793240940072e-05, "loss": 1.082, "step": 1882 }, { "epoch": 0.13, "learning_rate": 1.9527277655635582e-05, "loss": 1.0668, "step": 1883 }, { "epoch": 0.13, "learning_rate": 1.952662245910814e-05, "loss": 1.0845, "step": 1884 }, { "epoch": 0.13, "learning_rate": 1.952596681984885e-05, "loss": 1.073, "step": 1885 }, { "epoch": 0.13, "learning_rate": 1.9525310737888173e-05, "loss": 1.052, "step": 1886 }, { "epoch": 0.13, "learning_rate": 1.9524654213256608e-05, "loss": 1.0875, "step": 1887 }, { "epoch": 0.13, "learning_rate": 1.9523997245984656e-05, "loss": 1.0936, "step": 1888 }, { "epoch": 0.13, "learning_rate": 1.9523339836102858e-05, "loss": 1.0728, "step": 1889 }, { "epoch": 0.13, "learning_rate": 1.9522681983641757e-05, "loss": 1.0909, "step": 1890 }, { "epoch": 0.13, "learning_rate": 1.952202368863193e-05, "loss": 1.1176, "step": 1891 }, { "epoch": 0.13, "learning_rate": 1.9521364951103973e-05, "loss": 1.0639, "step": 1892 }, { "epoch": 0.13, "learning_rate": 1.9520705771088496e-05, "loss": 1.0925, "step": 1893 }, { "epoch": 0.13, "learning_rate": 1.9520046148616134e-05, "loss": 1.0992, "step": 1894 }, { "epoch": 0.13, "learning_rate": 1.951938608371754e-05, "loss": 1.089, "step": 1895 }, { "epoch": 0.13, "learning_rate": 1.951872557642339e-05, "loss": 1.0638, "step": 1896 }, { "epoch": 0.13, "learning_rate": 1.9518064626764384e-05, "loss": 1.0875, "step": 1897 }, { "epoch": 0.13, "learning_rate": 1.951740323477123e-05, "loss": 1.0467, "step": 1898 }, { "epoch": 0.13, "learning_rate": 1.9516741400474675e-05, "loss": 1.0714, "step": 1899 }, { "epoch": 0.13, "learning_rate": 1.951607912390547e-05, "loss": 1.066, "step": 1900 }, { "epoch": 0.13, "learning_rate": 1.951541640509439e-05, "loss": 1.0709, "step": 1901 }, { "epoch": 0.13, "learning_rate": 1.9514753244072242e-05, "loss": 1.0792, "step": 1902 }, { "epoch": 0.13, "learning_rate": 1.951408964086984e-05, "loss": 1.0973, "step": 1903 }, { "epoch": 0.13, "learning_rate": 1.9513425595518027e-05, "loss": 1.0746, "step": 1904 }, { "epoch": 0.13, "learning_rate": 1.951276110804766e-05, "loss": 1.0929, "step": 1905 }, { "epoch": 0.13, "learning_rate": 1.951209617848962e-05, "loss": 1.0266, "step": 1906 }, { "epoch": 0.13, "learning_rate": 1.951143080687481e-05, "loss": 1.0879, "step": 1907 }, { "epoch": 0.13, "learning_rate": 1.951076499323415e-05, "loss": 1.0881, "step": 1908 }, { "epoch": 0.13, "learning_rate": 1.9510098737598584e-05, "loss": 1.1051, "step": 1909 }, { "epoch": 0.13, "learning_rate": 1.9509432039999075e-05, "loss": 1.1052, "step": 1910 }, { "epoch": 0.13, "learning_rate": 1.9508764900466608e-05, "loss": 1.0923, "step": 1911 }, { "epoch": 0.13, "learning_rate": 1.9508097319032184e-05, "loss": 1.0876, "step": 1912 }, { "epoch": 0.13, "learning_rate": 1.9507429295726827e-05, "loss": 1.0866, "step": 1913 }, { "epoch": 0.13, "learning_rate": 1.9506760830581587e-05, "loss": 1.0487, "step": 1914 }, { "epoch": 0.13, "learning_rate": 1.9506091923627523e-05, "loss": 1.076, "step": 1915 }, { "epoch": 0.13, "learning_rate": 1.9505422574895723e-05, "loss": 1.068, "step": 1916 }, { "epoch": 0.13, "learning_rate": 1.9504752784417302e-05, "loss": 1.0397, "step": 1917 }, { "epoch": 0.13, "learning_rate": 1.9504082552223377e-05, "loss": 1.0901, "step": 1918 }, { "epoch": 0.13, "learning_rate": 1.9503411878345103e-05, "loss": 1.1088, "step": 1919 }, { "epoch": 0.13, "learning_rate": 1.950274076281364e-05, "loss": 1.0541, "step": 1920 }, { "epoch": 0.13, "learning_rate": 1.9502069205660185e-05, "loss": 1.0584, "step": 1921 }, { "epoch": 0.13, "learning_rate": 1.950139720691594e-05, "loss": 1.0953, "step": 1922 }, { "epoch": 0.13, "learning_rate": 1.9500724766612143e-05, "loss": 1.0862, "step": 1923 }, { "epoch": 0.13, "learning_rate": 1.950005188478004e-05, "loss": 1.0454, "step": 1924 }, { "epoch": 0.13, "learning_rate": 1.9499378561450898e-05, "loss": 1.1051, "step": 1925 }, { "epoch": 0.13, "learning_rate": 1.949870479665602e-05, "loss": 1.0603, "step": 1926 }, { "epoch": 0.13, "learning_rate": 1.9498030590426702e-05, "loss": 1.0787, "step": 1927 }, { "epoch": 0.13, "learning_rate": 1.949735594279429e-05, "loss": 1.032, "step": 1928 }, { "epoch": 0.13, "learning_rate": 1.9496680853790133e-05, "loss": 1.093, "step": 1929 }, { "epoch": 0.13, "learning_rate": 1.94960053234456e-05, "loss": 1.0772, "step": 1930 }, { "epoch": 0.13, "learning_rate": 1.949532935179209e-05, "loss": 1.1044, "step": 1931 }, { "epoch": 0.13, "learning_rate": 1.949465293886102e-05, "loss": 1.0831, "step": 1932 }, { "epoch": 0.13, "learning_rate": 1.9493976084683814e-05, "loss": 1.0726, "step": 1933 }, { "epoch": 0.13, "learning_rate": 1.9493298789291937e-05, "loss": 1.0745, "step": 1934 }, { "epoch": 0.13, "learning_rate": 1.9492621052716865e-05, "loss": 1.0886, "step": 1935 }, { "epoch": 0.13, "learning_rate": 1.949194287499009e-05, "loss": 1.1009, "step": 1936 }, { "epoch": 0.13, "learning_rate": 1.949126425614313e-05, "loss": 1.0366, "step": 1937 }, { "epoch": 0.13, "learning_rate": 1.9490585196207527e-05, "loss": 1.0596, "step": 1938 }, { "epoch": 0.13, "learning_rate": 1.9489905695214836e-05, "loss": 1.0602, "step": 1939 }, { "epoch": 0.13, "learning_rate": 1.9489225753196633e-05, "loss": 1.0803, "step": 1940 }, { "epoch": 0.13, "learning_rate": 1.948854537018452e-05, "loss": 1.0733, "step": 1941 }, { "epoch": 0.13, "learning_rate": 1.9487864546210113e-05, "loss": 1.0631, "step": 1942 }, { "epoch": 0.13, "learning_rate": 1.9487183281305058e-05, "loss": 1.0714, "step": 1943 }, { "epoch": 0.13, "learning_rate": 1.9486501575501015e-05, "loss": 1.0454, "step": 1944 }, { "epoch": 0.13, "learning_rate": 1.9485819428829655e-05, "loss": 1.0843, "step": 1945 }, { "epoch": 0.13, "learning_rate": 1.9485136841322694e-05, "loss": 1.08, "step": 1946 }, { "epoch": 0.13, "learning_rate": 1.948445381301184e-05, "loss": 1.0978, "step": 1947 }, { "epoch": 0.13, "learning_rate": 1.9483770343928847e-05, "loss": 1.0738, "step": 1948 }, { "epoch": 0.13, "learning_rate": 1.948308643410547e-05, "loss": 1.084, "step": 1949 }, { "epoch": 0.13, "learning_rate": 1.94824020835735e-05, "loss": 1.0622, "step": 1950 }, { "epoch": 0.13, "learning_rate": 1.9481717292364735e-05, "loss": 1.0757, "step": 1951 }, { "epoch": 0.13, "learning_rate": 1.9481032060511e-05, "loss": 1.0407, "step": 1952 }, { "epoch": 0.13, "learning_rate": 1.948034638804414e-05, "loss": 1.055, "step": 1953 }, { "epoch": 0.13, "learning_rate": 1.947966027499602e-05, "loss": 1.0296, "step": 1954 }, { "epoch": 0.13, "learning_rate": 1.947897372139853e-05, "loss": 1.1104, "step": 1955 }, { "epoch": 0.13, "learning_rate": 1.947828672728357e-05, "loss": 1.1252, "step": 1956 }, { "epoch": 0.13, "learning_rate": 1.9477599292683075e-05, "loss": 1.1008, "step": 1957 }, { "epoch": 0.13, "learning_rate": 1.9476911417628985e-05, "loss": 1.0572, "step": 1958 }, { "epoch": 0.13, "learning_rate": 1.9476223102153266e-05, "loss": 1.0726, "step": 1959 }, { "epoch": 0.13, "learning_rate": 1.9475534346287912e-05, "loss": 1.0493, "step": 1960 }, { "epoch": 0.13, "learning_rate": 1.947484515006493e-05, "loss": 1.1092, "step": 1961 }, { "epoch": 0.13, "learning_rate": 1.947415551351635e-05, "loss": 1.063, "step": 1962 }, { "epoch": 0.13, "learning_rate": 1.947346543667422e-05, "loss": 1.1094, "step": 1963 }, { "epoch": 0.13, "learning_rate": 1.9472774919570606e-05, "loss": 1.071, "step": 1964 }, { "epoch": 0.13, "learning_rate": 1.947208396223761e-05, "loss": 1.0553, "step": 1965 }, { "epoch": 0.13, "learning_rate": 1.947139256470733e-05, "loss": 1.0715, "step": 1966 }, { "epoch": 0.13, "learning_rate": 1.9470700727011905e-05, "loss": 1.0929, "step": 1967 }, { "epoch": 0.13, "learning_rate": 1.947000844918349e-05, "loss": 1.0652, "step": 1968 }, { "epoch": 0.13, "learning_rate": 1.9469315731254244e-05, "loss": 1.0688, "step": 1969 }, { "epoch": 0.13, "learning_rate": 1.9468622573256373e-05, "loss": 1.0788, "step": 1970 }, { "epoch": 0.13, "learning_rate": 1.9467928975222087e-05, "loss": 1.0348, "step": 1971 }, { "epoch": 0.13, "learning_rate": 1.946723493718361e-05, "loss": 1.0396, "step": 1972 }, { "epoch": 0.13, "learning_rate": 1.9466540459173213e-05, "loss": 1.1185, "step": 1973 }, { "epoch": 0.13, "learning_rate": 1.9465845541223156e-05, "loss": 1.029, "step": 1974 }, { "epoch": 0.13, "learning_rate": 1.946515018336574e-05, "loss": 1.0855, "step": 1975 }, { "epoch": 0.13, "learning_rate": 1.946445438563328e-05, "loss": 1.0067, "step": 1976 }, { "epoch": 0.13, "learning_rate": 1.9463758148058117e-05, "loss": 1.0871, "step": 1977 }, { "epoch": 0.13, "learning_rate": 1.94630614706726e-05, "loss": 1.0221, "step": 1978 }, { "epoch": 0.13, "learning_rate": 1.9462364353509106e-05, "loss": 1.038, "step": 1979 }, { "epoch": 0.13, "learning_rate": 1.9461666796600034e-05, "loss": 1.0525, "step": 1980 }, { "epoch": 0.13, "learning_rate": 1.9460968799977805e-05, "loss": 1.0932, "step": 1981 }, { "epoch": 0.13, "learning_rate": 1.946027036367485e-05, "loss": 1.0531, "step": 1982 }, { "epoch": 0.13, "learning_rate": 1.9459571487723634e-05, "loss": 1.0868, "step": 1983 }, { "epoch": 0.13, "learning_rate": 1.9458872172156632e-05, "loss": 1.0808, "step": 1984 }, { "epoch": 0.13, "learning_rate": 1.9458172417006347e-05, "loss": 1.0539, "step": 1985 }, { "epoch": 0.13, "learning_rate": 1.9457472222305296e-05, "loss": 1.0393, "step": 1986 }, { "epoch": 0.13, "learning_rate": 1.945677158808602e-05, "loss": 1.0561, "step": 1987 }, { "epoch": 0.13, "learning_rate": 1.945607051438108e-05, "loss": 1.0498, "step": 1988 }, { "epoch": 0.13, "learning_rate": 1.9455369001223053e-05, "loss": 1.0736, "step": 1989 }, { "epoch": 0.13, "learning_rate": 1.9454667048644548e-05, "loss": 1.0397, "step": 1990 }, { "epoch": 0.13, "learning_rate": 1.9453964656678185e-05, "loss": 1.0722, "step": 1991 }, { "epoch": 0.13, "learning_rate": 1.94532618253566e-05, "loss": 1.0666, "step": 1992 }, { "epoch": 0.13, "learning_rate": 1.945255855471246e-05, "loss": 1.0729, "step": 1993 }, { "epoch": 0.13, "learning_rate": 1.945185484477845e-05, "loss": 1.0477, "step": 1994 }, { "epoch": 0.13, "learning_rate": 1.9451150695587272e-05, "loss": 1.1056, "step": 1995 }, { "epoch": 0.13, "learning_rate": 1.945044610717165e-05, "loss": 1.0305, "step": 1996 }, { "epoch": 0.13, "learning_rate": 1.944974107956433e-05, "loss": 1.0902, "step": 1997 }, { "epoch": 0.13, "learning_rate": 1.944903561279807e-05, "loss": 1.0748, "step": 1998 }, { "epoch": 0.13, "learning_rate": 1.9448329706905665e-05, "loss": 1.0508, "step": 1999 }, { "epoch": 0.13, "learning_rate": 1.9447623361919915e-05, "loss": 1.0325, "step": 2000 }, { "epoch": 0.13, "learning_rate": 1.9446916577873645e-05, "loss": 1.0481, "step": 2001 }, { "epoch": 0.13, "learning_rate": 1.9446209354799704e-05, "loss": 1.0594, "step": 2002 }, { "epoch": 0.13, "learning_rate": 1.9445501692730963e-05, "loss": 1.0251, "step": 2003 }, { "epoch": 0.13, "learning_rate": 1.94447935917003e-05, "loss": 1.1179, "step": 2004 }, { "epoch": 0.13, "learning_rate": 1.944408505174063e-05, "loss": 1.0627, "step": 2005 }, { "epoch": 0.13, "learning_rate": 1.9443376072884876e-05, "loss": 1.0751, "step": 2006 }, { "epoch": 0.13, "learning_rate": 1.944266665516599e-05, "loss": 1.0778, "step": 2007 }, { "epoch": 0.13, "learning_rate": 1.944195679861694e-05, "loss": 1.0705, "step": 2008 }, { "epoch": 0.13, "learning_rate": 1.9441246503270716e-05, "loss": 1.0653, "step": 2009 }, { "epoch": 0.13, "learning_rate": 1.9440535769160326e-05, "loss": 1.0679, "step": 2010 }, { "epoch": 0.13, "learning_rate": 1.94398245963188e-05, "loss": 1.0763, "step": 2011 }, { "epoch": 0.13, "learning_rate": 1.9439112984779188e-05, "loss": 1.0267, "step": 2012 }, { "epoch": 0.13, "learning_rate": 1.9438400934574565e-05, "loss": 1.0783, "step": 2013 }, { "epoch": 0.13, "learning_rate": 1.9437688445738016e-05, "loss": 1.0626, "step": 2014 }, { "epoch": 0.13, "learning_rate": 1.9436975518302657e-05, "loss": 1.088, "step": 2015 }, { "epoch": 0.13, "learning_rate": 1.9436262152301618e-05, "loss": 1.0218, "step": 2016 }, { "epoch": 0.13, "learning_rate": 1.9435548347768054e-05, "loss": 1.0854, "step": 2017 }, { "epoch": 0.13, "learning_rate": 1.9434834104735135e-05, "loss": 1.096, "step": 2018 }, { "epoch": 0.13, "learning_rate": 1.9434119423236054e-05, "loss": 1.056, "step": 2019 }, { "epoch": 0.13, "learning_rate": 1.9433404303304027e-05, "loss": 1.0316, "step": 2020 }, { "epoch": 0.13, "learning_rate": 1.9432688744972282e-05, "loss": 1.1452, "step": 2021 }, { "epoch": 0.13, "learning_rate": 1.9431972748274082e-05, "loss": 1.0393, "step": 2022 }, { "epoch": 0.13, "learning_rate": 1.9431256313242692e-05, "loss": 1.1183, "step": 2023 }, { "epoch": 0.13, "learning_rate": 1.9430539439911416e-05, "loss": 1.0974, "step": 2024 }, { "epoch": 0.13, "learning_rate": 1.9429822128313564e-05, "loss": 1.0805, "step": 2025 }, { "epoch": 0.13, "learning_rate": 1.9429104378482474e-05, "loss": 1.0856, "step": 2026 }, { "epoch": 0.13, "learning_rate": 1.94283861904515e-05, "loss": 1.0465, "step": 2027 }, { "epoch": 0.13, "learning_rate": 1.9427667564254017e-05, "loss": 1.0764, "step": 2028 }, { "epoch": 0.14, "learning_rate": 1.942694849992343e-05, "loss": 1.1003, "step": 2029 }, { "epoch": 0.14, "learning_rate": 1.9426228997493145e-05, "loss": 1.0556, "step": 2030 }, { "epoch": 0.14, "learning_rate": 1.942550905699661e-05, "loss": 1.0532, "step": 2031 }, { "epoch": 0.14, "learning_rate": 1.9424788678467273e-05, "loss": 1.0989, "step": 2032 }, { "epoch": 0.14, "learning_rate": 1.942406786193862e-05, "loss": 1.1205, "step": 2033 }, { "epoch": 0.14, "learning_rate": 1.9423346607444145e-05, "loss": 1.1016, "step": 2034 }, { "epoch": 0.14, "learning_rate": 1.942262491501737e-05, "loss": 1.0996, "step": 2035 }, { "epoch": 0.14, "learning_rate": 1.942190278469183e-05, "loss": 1.0691, "step": 2036 }, { "epoch": 0.14, "learning_rate": 1.942118021650109e-05, "loss": 1.0864, "step": 2037 }, { "epoch": 0.14, "learning_rate": 1.942045721047873e-05, "loss": 1.0649, "step": 2038 }, { "epoch": 0.14, "learning_rate": 1.9419733766658347e-05, "loss": 1.0556, "step": 2039 }, { "epoch": 0.14, "learning_rate": 1.941900988507356e-05, "loss": 1.064, "step": 2040 }, { "epoch": 0.14, "learning_rate": 1.9418285565758014e-05, "loss": 1.0663, "step": 2041 }, { "epoch": 0.14, "learning_rate": 1.941756080874537e-05, "loss": 1.0625, "step": 2042 }, { "epoch": 0.14, "learning_rate": 1.9416835614069304e-05, "loss": 1.0708, "step": 2043 }, { "epoch": 0.14, "learning_rate": 1.9416109981763526e-05, "loss": 1.0594, "step": 2044 }, { "epoch": 0.14, "learning_rate": 1.9415383911861756e-05, "loss": 1.0915, "step": 2045 }, { "epoch": 0.14, "learning_rate": 1.9414657404397735e-05, "loss": 1.0706, "step": 2046 }, { "epoch": 0.14, "learning_rate": 1.941393045940523e-05, "loss": 1.1062, "step": 2047 }, { "epoch": 0.14, "learning_rate": 1.9413203076918016e-05, "loss": 1.0353, "step": 2048 }, { "epoch": 0.14, "learning_rate": 1.9412475256969904e-05, "loss": 1.041, "step": 2049 }, { "epoch": 0.14, "learning_rate": 1.9411746999594716e-05, "loss": 1.0661, "step": 2050 }, { "epoch": 0.14, "learning_rate": 1.9411018304826297e-05, "loss": 1.0629, "step": 2051 }, { "epoch": 0.14, "learning_rate": 1.941028917269851e-05, "loss": 1.0867, "step": 2052 }, { "epoch": 0.14, "learning_rate": 1.9409559603245246e-05, "loss": 1.1265, "step": 2053 }, { "epoch": 0.14, "learning_rate": 1.9408829596500404e-05, "loss": 1.0476, "step": 2054 }, { "epoch": 0.14, "learning_rate": 1.9408099152497906e-05, "loss": 1.0841, "step": 2055 }, { "epoch": 0.14, "learning_rate": 1.940736827127171e-05, "loss": 1.0555, "step": 2056 }, { "epoch": 0.14, "learning_rate": 1.9406636952855772e-05, "loss": 1.0888, "step": 2057 }, { "epoch": 0.14, "learning_rate": 1.9405905197284084e-05, "loss": 1.0568, "step": 2058 }, { "epoch": 0.14, "learning_rate": 1.9405173004590652e-05, "loss": 1.0279, "step": 2059 }, { "epoch": 0.14, "learning_rate": 1.94044403748095e-05, "loss": 1.0401, "step": 2060 }, { "epoch": 0.14, "learning_rate": 1.940370730797468e-05, "loss": 1.0578, "step": 2061 }, { "epoch": 0.14, "learning_rate": 1.940297380412026e-05, "loss": 1.0901, "step": 2062 }, { "epoch": 0.14, "learning_rate": 1.9402239863280322e-05, "loss": 1.0792, "step": 2063 }, { "epoch": 0.14, "learning_rate": 1.940150548548898e-05, "loss": 1.0693, "step": 2064 }, { "epoch": 0.14, "learning_rate": 1.9400770670780363e-05, "loss": 1.082, "step": 2065 }, { "epoch": 0.14, "learning_rate": 1.940003541918862e-05, "loss": 1.1306, "step": 2066 }, { "epoch": 0.14, "learning_rate": 1.9399299730747917e-05, "loss": 1.0626, "step": 2067 }, { "epoch": 0.14, "learning_rate": 1.9398563605492446e-05, "loss": 1.0978, "step": 2068 }, { "epoch": 0.14, "learning_rate": 1.9397827043456415e-05, "loss": 1.0637, "step": 2069 }, { "epoch": 0.14, "learning_rate": 1.939709004467406e-05, "loss": 1.0445, "step": 2070 }, { "epoch": 0.14, "learning_rate": 1.9396352609179626e-05, "loss": 1.0967, "step": 2071 }, { "epoch": 0.14, "learning_rate": 1.9395614737007385e-05, "loss": 1.033, "step": 2072 }, { "epoch": 0.14, "learning_rate": 1.9394876428191632e-05, "loss": 1.0552, "step": 2073 }, { "epoch": 0.14, "learning_rate": 1.9394137682766675e-05, "loss": 1.0656, "step": 2074 }, { "epoch": 0.14, "learning_rate": 1.939339850076685e-05, "loss": 1.0811, "step": 2075 }, { "epoch": 0.14, "learning_rate": 1.93926588822265e-05, "loss": 1.0568, "step": 2076 }, { "epoch": 0.14, "learning_rate": 1.9391918827180002e-05, "loss": 1.0469, "step": 2077 }, { "epoch": 0.14, "learning_rate": 1.9391178335661754e-05, "loss": 1.081, "step": 2078 }, { "epoch": 0.14, "learning_rate": 1.939043740770616e-05, "loss": 1.0656, "step": 2079 }, { "epoch": 0.14, "learning_rate": 1.938969604334766e-05, "loss": 1.1031, "step": 2080 }, { "epoch": 0.14, "learning_rate": 1.9388954242620707e-05, "loss": 1.0367, "step": 2081 }, { "epoch": 0.14, "learning_rate": 1.938821200555977e-05, "loss": 1.0842, "step": 2082 }, { "epoch": 0.14, "learning_rate": 1.9387469332199348e-05, "loss": 1.0867, "step": 2083 }, { "epoch": 0.14, "learning_rate": 1.9386726222573953e-05, "loss": 1.0837, "step": 2084 }, { "epoch": 0.14, "learning_rate": 1.938598267671812e-05, "loss": 1.0848, "step": 2085 }, { "epoch": 0.14, "learning_rate": 1.9385238694666403e-05, "loss": 1.0715, "step": 2086 }, { "epoch": 0.14, "learning_rate": 1.938449427645338e-05, "loss": 1.0513, "step": 2087 }, { "epoch": 0.14, "learning_rate": 1.938374942211364e-05, "loss": 1.086, "step": 2088 }, { "epoch": 0.14, "learning_rate": 1.938300413168181e-05, "loss": 1.0718, "step": 2089 }, { "epoch": 0.14, "learning_rate": 1.938225840519252e-05, "loss": 1.0396, "step": 2090 }, { "epoch": 0.14, "learning_rate": 1.938151224268042e-05, "loss": 1.0856, "step": 2091 }, { "epoch": 0.14, "learning_rate": 1.938076564418019e-05, "loss": 1.042, "step": 2092 }, { "epoch": 0.14, "learning_rate": 1.9380018609726537e-05, "loss": 1.0908, "step": 2093 }, { "epoch": 0.14, "learning_rate": 1.9379271139354162e-05, "loss": 1.0987, "step": 2094 }, { "epoch": 0.14, "learning_rate": 1.9378523233097816e-05, "loss": 1.05, "step": 2095 }, { "epoch": 0.14, "learning_rate": 1.9377774890992248e-05, "loss": 1.073, "step": 2096 }, { "epoch": 0.14, "learning_rate": 1.9377026113072236e-05, "loss": 1.0865, "step": 2097 }, { "epoch": 0.14, "learning_rate": 1.9376276899372584e-05, "loss": 1.0448, "step": 2098 }, { "epoch": 0.14, "learning_rate": 1.9375527249928104e-05, "loss": 1.0859, "step": 2099 }, { "epoch": 0.14, "learning_rate": 1.937477716477364e-05, "loss": 1.0039, "step": 2100 }, { "epoch": 0.14, "learning_rate": 1.9374026643944048e-05, "loss": 1.1069, "step": 2101 }, { "epoch": 0.14, "learning_rate": 1.9373275687474208e-05, "loss": 1.0102, "step": 2102 }, { "epoch": 0.14, "learning_rate": 1.9372524295399014e-05, "loss": 1.1016, "step": 2103 }, { "epoch": 0.14, "learning_rate": 1.9371772467753393e-05, "loss": 1.0621, "step": 2104 }, { "epoch": 0.14, "learning_rate": 1.937102020457228e-05, "loss": 1.0429, "step": 2105 }, { "epoch": 0.14, "learning_rate": 1.9370267505890637e-05, "loss": 1.0783, "step": 2106 }, { "epoch": 0.14, "learning_rate": 1.9369514371743445e-05, "loss": 1.0745, "step": 2107 }, { "epoch": 0.14, "learning_rate": 1.9368760802165704e-05, "loss": 1.0503, "step": 2108 }, { "epoch": 0.14, "learning_rate": 1.9368006797192432e-05, "loss": 1.0828, "step": 2109 }, { "epoch": 0.14, "learning_rate": 1.9367252356858676e-05, "loss": 1.0422, "step": 2110 }, { "epoch": 0.14, "learning_rate": 1.936649748119949e-05, "loss": 1.0607, "step": 2111 }, { "epoch": 0.14, "learning_rate": 1.9365742170249958e-05, "loss": 1.1098, "step": 2112 }, { "epoch": 0.14, "learning_rate": 1.9364986424045185e-05, "loss": 1.1198, "step": 2113 }, { "epoch": 0.14, "learning_rate": 1.936423024262029e-05, "loss": 1.0573, "step": 2114 }, { "epoch": 0.14, "learning_rate": 1.9363473626010414e-05, "loss": 1.0906, "step": 2115 }, { "epoch": 0.14, "learning_rate": 1.9362716574250724e-05, "loss": 1.0466, "step": 2116 }, { "epoch": 0.14, "learning_rate": 1.9361959087376395e-05, "loss": 1.0812, "step": 2117 }, { "epoch": 0.14, "learning_rate": 1.9361201165422637e-05, "loss": 1.0274, "step": 2118 }, { "epoch": 0.14, "learning_rate": 1.9360442808424667e-05, "loss": 1.0483, "step": 2119 }, { "epoch": 0.14, "learning_rate": 1.9359684016417735e-05, "loss": 0.9936, "step": 2120 }, { "epoch": 0.14, "learning_rate": 1.93589247894371e-05, "loss": 1.0757, "step": 2121 }, { "epoch": 0.14, "learning_rate": 1.9358165127518043e-05, "loss": 1.0404, "step": 2122 }, { "epoch": 0.14, "learning_rate": 1.9357405030695873e-05, "loss": 1.1166, "step": 2123 }, { "epoch": 0.14, "learning_rate": 1.9356644499005917e-05, "loss": 1.0805, "step": 2124 }, { "epoch": 0.14, "learning_rate": 1.9355883532483507e-05, "loss": 1.037, "step": 2125 }, { "epoch": 0.14, "learning_rate": 1.935512213116402e-05, "loss": 1.0484, "step": 2126 }, { "epoch": 0.14, "learning_rate": 1.9354360295082836e-05, "loss": 1.07, "step": 2127 }, { "epoch": 0.14, "learning_rate": 1.9353598024275356e-05, "loss": 1.0677, "step": 2128 }, { "epoch": 0.14, "learning_rate": 1.9352835318777012e-05, "loss": 1.0572, "step": 2129 }, { "epoch": 0.14, "learning_rate": 1.9352072178623244e-05, "loss": 1.0645, "step": 2130 }, { "epoch": 0.14, "learning_rate": 1.9351308603849525e-05, "loss": 1.0941, "step": 2131 }, { "epoch": 0.14, "learning_rate": 1.935054459449133e-05, "loss": 1.0681, "step": 2132 }, { "epoch": 0.14, "learning_rate": 1.934978015058417e-05, "loss": 1.103, "step": 2133 }, { "epoch": 0.14, "learning_rate": 1.9349015272163573e-05, "loss": 1.0291, "step": 2134 }, { "epoch": 0.14, "learning_rate": 1.9348249959265083e-05, "loss": 1.0514, "step": 2135 }, { "epoch": 0.14, "learning_rate": 1.9347484211924266e-05, "loss": 1.0382, "step": 2136 }, { "epoch": 0.14, "learning_rate": 1.934671803017671e-05, "loss": 1.0753, "step": 2137 }, { "epoch": 0.14, "learning_rate": 1.9345951414058023e-05, "loss": 1.022, "step": 2138 }, { "epoch": 0.14, "learning_rate": 1.934518436360383e-05, "loss": 1.0878, "step": 2139 }, { "epoch": 0.14, "learning_rate": 1.9344416878849776e-05, "loss": 1.0939, "step": 2140 }, { "epoch": 0.14, "learning_rate": 1.9343648959831536e-05, "loss": 1.0407, "step": 2141 }, { "epoch": 0.14, "learning_rate": 1.934288060658479e-05, "loss": 1.0133, "step": 2142 }, { "epoch": 0.14, "learning_rate": 1.9342111819145247e-05, "loss": 1.0652, "step": 2143 }, { "epoch": 0.14, "learning_rate": 1.934134259754864e-05, "loss": 1.0813, "step": 2144 }, { "epoch": 0.14, "learning_rate": 1.9340572941830716e-05, "loss": 1.0506, "step": 2145 }, { "epoch": 0.14, "learning_rate": 1.9339802852027233e-05, "loss": 1.0899, "step": 2146 }, { "epoch": 0.14, "learning_rate": 1.933903232817399e-05, "loss": 1.0401, "step": 2147 }, { "epoch": 0.14, "learning_rate": 1.93382613703068e-05, "loss": 1.0586, "step": 2148 }, { "epoch": 0.14, "learning_rate": 1.933748997846148e-05, "loss": 1.0985, "step": 2149 }, { "epoch": 0.14, "learning_rate": 1.9336718152673884e-05, "loss": 1.0668, "step": 2150 }, { "epoch": 0.14, "learning_rate": 1.933594589297988e-05, "loss": 1.0897, "step": 2151 }, { "epoch": 0.14, "learning_rate": 1.9335173199415358e-05, "loss": 1.0283, "step": 2152 }, { "epoch": 0.14, "learning_rate": 1.9334400072016228e-05, "loss": 1.0717, "step": 2153 }, { "epoch": 0.14, "learning_rate": 1.9333626510818423e-05, "loss": 1.0413, "step": 2154 }, { "epoch": 0.14, "learning_rate": 1.9332852515857886e-05, "loss": 1.0664, "step": 2155 }, { "epoch": 0.14, "learning_rate": 1.933207808717059e-05, "loss": 1.0502, "step": 2156 }, { "epoch": 0.14, "learning_rate": 1.9331303224792526e-05, "loss": 1.0808, "step": 2157 }, { "epoch": 0.14, "learning_rate": 1.9330527928759705e-05, "loss": 1.0483, "step": 2158 }, { "epoch": 0.14, "learning_rate": 1.9329752199108153e-05, "loss": 1.0727, "step": 2159 }, { "epoch": 0.14, "learning_rate": 1.9328976035873927e-05, "loss": 1.0741, "step": 2160 }, { "epoch": 0.14, "learning_rate": 1.9328199439093094e-05, "loss": 1.093, "step": 2161 }, { "epoch": 0.14, "learning_rate": 1.9327422408801744e-05, "loss": 1.0364, "step": 2162 }, { "epoch": 0.14, "learning_rate": 1.9326644945035987e-05, "loss": 1.049, "step": 2163 }, { "epoch": 0.14, "learning_rate": 1.932586704783196e-05, "loss": 1.0841, "step": 2164 }, { "epoch": 0.14, "learning_rate": 1.9325088717225808e-05, "loss": 1.0774, "step": 2165 }, { "epoch": 0.14, "learning_rate": 1.9324309953253706e-05, "loss": 0.9861, "step": 2166 }, { "epoch": 0.14, "learning_rate": 1.9323530755951848e-05, "loss": 1.0642, "step": 2167 }, { "epoch": 0.14, "learning_rate": 1.9322751125356438e-05, "loss": 1.0452, "step": 2168 }, { "epoch": 0.14, "learning_rate": 1.9321971061503716e-05, "loss": 1.019, "step": 2169 }, { "epoch": 0.14, "learning_rate": 1.932119056442993e-05, "loss": 1.0481, "step": 2170 }, { "epoch": 0.14, "learning_rate": 1.932040963417135e-05, "loss": 1.0779, "step": 2171 }, { "epoch": 0.14, "learning_rate": 1.931962827076427e-05, "loss": 1.0417, "step": 2172 }, { "epoch": 0.14, "learning_rate": 1.9318846474245007e-05, "loss": 1.0899, "step": 2173 }, { "epoch": 0.14, "learning_rate": 1.9318064244649887e-05, "loss": 1.0795, "step": 2174 }, { "epoch": 0.14, "learning_rate": 1.9317281582015264e-05, "loss": 1.0775, "step": 2175 }, { "epoch": 0.14, "learning_rate": 1.9316498486377514e-05, "loss": 1.0469, "step": 2176 }, { "epoch": 0.14, "learning_rate": 1.931571495777303e-05, "loss": 1.0662, "step": 2177 }, { "epoch": 0.14, "learning_rate": 1.931493099623822e-05, "loss": 1.0198, "step": 2178 }, { "epoch": 0.15, "learning_rate": 1.931414660180952e-05, "loss": 1.0656, "step": 2179 }, { "epoch": 0.15, "learning_rate": 1.9313361774523387e-05, "loss": 1.0465, "step": 2180 }, { "epoch": 0.15, "learning_rate": 1.931257651441629e-05, "loss": 1.0836, "step": 2181 }, { "epoch": 0.15, "learning_rate": 1.931179082152472e-05, "loss": 1.0496, "step": 2182 }, { "epoch": 0.15, "learning_rate": 1.9311004695885193e-05, "loss": 1.0865, "step": 2183 }, { "epoch": 0.15, "learning_rate": 1.9310218137534247e-05, "loss": 1.0424, "step": 2184 }, { "epoch": 0.15, "learning_rate": 1.9309431146508433e-05, "loss": 1.0693, "step": 2185 }, { "epoch": 0.15, "learning_rate": 1.930864372284432e-05, "loss": 1.0627, "step": 2186 }, { "epoch": 0.15, "learning_rate": 1.9307855866578513e-05, "loss": 1.093, "step": 2187 }, { "epoch": 0.15, "learning_rate": 1.930706757774761e-05, "loss": 1.0467, "step": 2188 }, { "epoch": 0.15, "learning_rate": 1.9306278856388262e-05, "loss": 1.0422, "step": 2189 }, { "epoch": 0.15, "learning_rate": 1.9305489702537116e-05, "loss": 1.079, "step": 2190 }, { "epoch": 0.15, "learning_rate": 1.9304700116230842e-05, "loss": 1.0379, "step": 2191 }, { "epoch": 0.15, "learning_rate": 1.9303910097506144e-05, "loss": 1.0102, "step": 2192 }, { "epoch": 0.15, "learning_rate": 1.930311964639973e-05, "loss": 1.1055, "step": 2193 }, { "epoch": 0.15, "learning_rate": 1.9302328762948335e-05, "loss": 1.08, "step": 2194 }, { "epoch": 0.15, "learning_rate": 1.9301537447188713e-05, "loss": 1.0638, "step": 2195 }, { "epoch": 0.15, "learning_rate": 1.9300745699157647e-05, "loss": 1.0624, "step": 2196 }, { "epoch": 0.15, "learning_rate": 1.9299953518891922e-05, "loss": 1.0434, "step": 2197 }, { "epoch": 0.15, "learning_rate": 1.929916090642836e-05, "loss": 1.0453, "step": 2198 }, { "epoch": 0.15, "learning_rate": 1.929836786180379e-05, "loss": 1.0907, "step": 2199 }, { "epoch": 0.15, "learning_rate": 1.9297574385055074e-05, "loss": 1.0558, "step": 2200 }, { "epoch": 0.15, "learning_rate": 1.9296780476219084e-05, "loss": 1.0434, "step": 2201 }, { "epoch": 0.15, "learning_rate": 1.9295986135332712e-05, "loss": 1.0381, "step": 2202 }, { "epoch": 0.15, "learning_rate": 1.929519136243288e-05, "loss": 1.0633, "step": 2203 }, { "epoch": 0.15, "learning_rate": 1.929439615755652e-05, "loss": 1.0317, "step": 2204 }, { "epoch": 0.15, "learning_rate": 1.9293600520740587e-05, "loss": 1.0207, "step": 2205 }, { "epoch": 0.15, "learning_rate": 1.9292804452022057e-05, "loss": 1.0472, "step": 2206 }, { "epoch": 0.15, "learning_rate": 1.9292007951437927e-05, "loss": 1.0844, "step": 2207 }, { "epoch": 0.15, "learning_rate": 1.9291211019025216e-05, "loss": 1.0439, "step": 2208 }, { "epoch": 0.15, "learning_rate": 1.929041365482095e-05, "loss": 1.0829, "step": 2209 }, { "epoch": 0.15, "learning_rate": 1.92896158588622e-05, "loss": 1.0188, "step": 2210 }, { "epoch": 0.15, "learning_rate": 1.9288817631186025e-05, "loss": 1.0722, "step": 2211 }, { "epoch": 0.15, "learning_rate": 1.9288018971829535e-05, "loss": 0.9826, "step": 2212 }, { "epoch": 0.15, "learning_rate": 1.928721988082984e-05, "loss": 1.0987, "step": 2213 }, { "epoch": 0.15, "learning_rate": 1.9286420358224074e-05, "loss": 1.0655, "step": 2214 }, { "epoch": 0.15, "learning_rate": 1.92856204040494e-05, "loss": 1.0526, "step": 2215 }, { "epoch": 0.15, "learning_rate": 1.928482001834299e-05, "loss": 1.0795, "step": 2216 }, { "epoch": 0.15, "learning_rate": 1.928401920114204e-05, "loss": 1.0598, "step": 2217 }, { "epoch": 0.15, "learning_rate": 1.9283217952483767e-05, "loss": 1.0804, "step": 2218 }, { "epoch": 0.15, "learning_rate": 1.928241627240541e-05, "loss": 1.0455, "step": 2219 }, { "epoch": 0.15, "learning_rate": 1.9281614160944226e-05, "loss": 1.0488, "step": 2220 }, { "epoch": 0.15, "learning_rate": 1.9280811618137486e-05, "loss": 1.1119, "step": 2221 }, { "epoch": 0.15, "learning_rate": 1.928000864402249e-05, "loss": 1.077, "step": 2222 }, { "epoch": 0.15, "learning_rate": 1.9279205238636554e-05, "loss": 1.0455, "step": 2223 }, { "epoch": 0.15, "learning_rate": 1.9278401402017017e-05, "loss": 0.9986, "step": 2224 }, { "epoch": 0.15, "learning_rate": 1.9277597134201232e-05, "loss": 1.0751, "step": 2225 }, { "epoch": 0.15, "learning_rate": 1.9276792435226582e-05, "loss": 1.0563, "step": 2226 }, { "epoch": 0.15, "learning_rate": 1.927598730513046e-05, "loss": 1.0539, "step": 2227 }, { "epoch": 0.15, "learning_rate": 1.927518174395028e-05, "loss": 1.0725, "step": 2228 }, { "epoch": 0.15, "learning_rate": 1.9274375751723483e-05, "loss": 1.0924, "step": 2229 }, { "epoch": 0.15, "learning_rate": 1.9273569328487522e-05, "loss": 1.0488, "step": 2230 }, { "epoch": 0.15, "learning_rate": 1.927276247427988e-05, "loss": 1.0922, "step": 2231 }, { "epoch": 0.15, "learning_rate": 1.927195518913805e-05, "loss": 1.0274, "step": 2232 }, { "epoch": 0.15, "learning_rate": 1.927114747309955e-05, "loss": 1.0387, "step": 2233 }, { "epoch": 0.15, "learning_rate": 1.927033932620192e-05, "loss": 1.0319, "step": 2234 }, { "epoch": 0.15, "learning_rate": 1.926953074848271e-05, "loss": 1.0722, "step": 2235 }, { "epoch": 0.15, "learning_rate": 1.9268721739979504e-05, "loss": 0.9895, "step": 2236 }, { "epoch": 0.15, "learning_rate": 1.9267912300729895e-05, "loss": 1.1001, "step": 2237 }, { "epoch": 0.15, "learning_rate": 1.92671024307715e-05, "loss": 1.049, "step": 2238 }, { "epoch": 0.15, "learning_rate": 1.926629213014196e-05, "loss": 1.0472, "step": 2239 }, { "epoch": 0.15, "learning_rate": 1.926548139887893e-05, "loss": 1.1024, "step": 2240 }, { "epoch": 0.15, "learning_rate": 1.9264670237020086e-05, "loss": 1.0407, "step": 2241 }, { "epoch": 0.15, "learning_rate": 1.9263858644603127e-05, "loss": 1.0538, "step": 2242 }, { "epoch": 0.15, "learning_rate": 1.926304662166577e-05, "loss": 1.073, "step": 2243 }, { "epoch": 0.15, "learning_rate": 1.9262234168245747e-05, "loss": 1.0231, "step": 2244 }, { "epoch": 0.15, "learning_rate": 1.9261421284380825e-05, "loss": 1.0231, "step": 2245 }, { "epoch": 0.15, "learning_rate": 1.9260607970108775e-05, "loss": 1.0799, "step": 2246 }, { "epoch": 0.15, "learning_rate": 1.92597942254674e-05, "loss": 1.0562, "step": 2247 }, { "epoch": 0.15, "learning_rate": 1.9258980050494505e-05, "loss": 1.0237, "step": 2248 }, { "epoch": 0.15, "learning_rate": 1.925816544522794e-05, "loss": 1.076, "step": 2249 }, { "epoch": 0.15, "learning_rate": 1.9257350409705557e-05, "loss": 1.0579, "step": 2250 }, { "epoch": 0.15, "learning_rate": 1.925653494396523e-05, "loss": 1.0708, "step": 2251 }, { "epoch": 0.15, "learning_rate": 1.9255719048044862e-05, "loss": 1.0305, "step": 2252 }, { "epoch": 0.15, "learning_rate": 1.925490272198237e-05, "loss": 1.0808, "step": 2253 }, { "epoch": 0.15, "learning_rate": 1.925408596581569e-05, "loss": 1.0697, "step": 2254 }, { "epoch": 0.15, "learning_rate": 1.9253268779582775e-05, "loss": 1.0974, "step": 2255 }, { "epoch": 0.15, "learning_rate": 1.9252451163321605e-05, "loss": 1.0403, "step": 2256 }, { "epoch": 0.15, "learning_rate": 1.925163311707018e-05, "loss": 1.0085, "step": 2257 }, { "epoch": 0.15, "learning_rate": 1.9250814640866514e-05, "loss": 1.0464, "step": 2258 }, { "epoch": 0.15, "learning_rate": 1.9249995734748647e-05, "loss": 1.0724, "step": 2259 }, { "epoch": 0.15, "learning_rate": 1.924917639875463e-05, "loss": 1.0639, "step": 2260 }, { "epoch": 0.15, "learning_rate": 1.924835663292255e-05, "loss": 1.0587, "step": 2261 }, { "epoch": 0.15, "learning_rate": 1.9247536437290497e-05, "loss": 1.0639, "step": 2262 }, { "epoch": 0.15, "learning_rate": 1.9246715811896585e-05, "loss": 1.0723, "step": 2263 }, { "epoch": 0.15, "learning_rate": 1.924589475677896e-05, "loss": 1.0548, "step": 2264 }, { "epoch": 0.15, "learning_rate": 1.924507327197577e-05, "loss": 1.0677, "step": 2265 }, { "epoch": 0.15, "learning_rate": 1.92442513575252e-05, "loss": 1.0539, "step": 2266 }, { "epoch": 0.15, "learning_rate": 1.9243429013465444e-05, "loss": 1.0081, "step": 2267 }, { "epoch": 0.15, "learning_rate": 1.9242606239834717e-05, "loss": 1.0698, "step": 2268 }, { "epoch": 0.15, "learning_rate": 1.9241783036671258e-05, "loss": 1.0768, "step": 2269 }, { "epoch": 0.15, "learning_rate": 1.924095940401332e-05, "loss": 1.0121, "step": 2270 }, { "epoch": 0.15, "learning_rate": 1.9240135341899186e-05, "loss": 1.091, "step": 2271 }, { "epoch": 0.15, "learning_rate": 1.9239310850367148e-05, "loss": 1.0513, "step": 2272 }, { "epoch": 0.15, "learning_rate": 1.9238485929455522e-05, "loss": 1.0372, "step": 2273 }, { "epoch": 0.15, "learning_rate": 1.9237660579202648e-05, "loss": 1.0815, "step": 2274 }, { "epoch": 0.15, "learning_rate": 1.9236834799646883e-05, "loss": 1.0735, "step": 2275 }, { "epoch": 0.15, "learning_rate": 1.92360085908266e-05, "loss": 1.0523, "step": 2276 }, { "epoch": 0.15, "learning_rate": 1.9235181952780203e-05, "loss": 1.0675, "step": 2277 }, { "epoch": 0.15, "learning_rate": 1.9234354885546096e-05, "loss": 1.0348, "step": 2278 }, { "epoch": 0.15, "learning_rate": 1.9233527389162723e-05, "loss": 1.0524, "step": 2279 }, { "epoch": 0.15, "learning_rate": 1.9232699463668543e-05, "loss": 1.0355, "step": 2280 }, { "epoch": 0.15, "learning_rate": 1.9231871109102027e-05, "loss": 1.0533, "step": 2281 }, { "epoch": 0.15, "learning_rate": 1.923104232550167e-05, "loss": 1.0598, "step": 2282 }, { "epoch": 0.15, "learning_rate": 1.9230213112906e-05, "loss": 1.1024, "step": 2283 }, { "epoch": 0.15, "learning_rate": 1.9229383471353536e-05, "loss": 1.0482, "step": 2284 }, { "epoch": 0.15, "learning_rate": 1.9228553400882844e-05, "loss": 1.0537, "step": 2285 }, { "epoch": 0.15, "learning_rate": 1.9227722901532503e-05, "loss": 1.0345, "step": 2286 }, { "epoch": 0.15, "learning_rate": 1.92268919733411e-05, "loss": 1.0382, "step": 2287 }, { "epoch": 0.15, "learning_rate": 1.9226060616347254e-05, "loss": 1.0515, "step": 2288 }, { "epoch": 0.15, "learning_rate": 1.9225228830589605e-05, "loss": 1.0664, "step": 2289 }, { "epoch": 0.15, "learning_rate": 1.9224396616106803e-05, "loss": 1.0743, "step": 2290 }, { "epoch": 0.15, "learning_rate": 1.922356397293753e-05, "loss": 1.0834, "step": 2291 }, { "epoch": 0.15, "learning_rate": 1.9222730901120473e-05, "loss": 1.0354, "step": 2292 }, { "epoch": 0.15, "learning_rate": 1.9221897400694355e-05, "loss": 1.033, "step": 2293 }, { "epoch": 0.15, "learning_rate": 1.9221063471697907e-05, "loss": 1.0207, "step": 2294 }, { "epoch": 0.15, "learning_rate": 1.9220229114169886e-05, "loss": 1.0852, "step": 2295 }, { "epoch": 0.15, "learning_rate": 1.921939432814907e-05, "loss": 1.0488, "step": 2296 }, { "epoch": 0.15, "learning_rate": 1.9218559113674244e-05, "loss": 1.0841, "step": 2297 }, { "epoch": 0.15, "learning_rate": 1.9217723470784233e-05, "loss": 1.0722, "step": 2298 }, { "epoch": 0.15, "learning_rate": 1.9216887399517872e-05, "loss": 1.0233, "step": 2299 }, { "epoch": 0.15, "learning_rate": 1.921605089991401e-05, "loss": 1.0397, "step": 2300 }, { "epoch": 0.15, "learning_rate": 1.9215213972011524e-05, "loss": 1.0183, "step": 2301 }, { "epoch": 0.15, "learning_rate": 1.9214376615849312e-05, "loss": 1.0825, "step": 2302 }, { "epoch": 0.15, "learning_rate": 1.9213538831466286e-05, "loss": 1.0806, "step": 2303 }, { "epoch": 0.15, "learning_rate": 1.9212700618901377e-05, "loss": 1.0318, "step": 2304 }, { "epoch": 0.15, "learning_rate": 1.9211861978193544e-05, "loss": 1.0577, "step": 2305 }, { "epoch": 0.15, "learning_rate": 1.921102290938176e-05, "loss": 1.0519, "step": 2306 }, { "epoch": 0.15, "learning_rate": 1.921018341250502e-05, "loss": 1.0629, "step": 2307 }, { "epoch": 0.15, "learning_rate": 1.9209343487602333e-05, "loss": 1.0352, "step": 2308 }, { "epoch": 0.15, "learning_rate": 1.9208503134712742e-05, "loss": 1.0706, "step": 2309 }, { "epoch": 0.15, "learning_rate": 1.920766235387529e-05, "loss": 1.0713, "step": 2310 }, { "epoch": 0.15, "learning_rate": 1.920682114512906e-05, "loss": 1.0581, "step": 2311 }, { "epoch": 0.15, "learning_rate": 1.9205979508513145e-05, "loss": 1.0076, "step": 2312 }, { "epoch": 0.15, "learning_rate": 1.920513744406665e-05, "loss": 1.0629, "step": 2313 }, { "epoch": 0.15, "learning_rate": 1.9204294951828716e-05, "loss": 1.0188, "step": 2314 }, { "epoch": 0.15, "learning_rate": 1.9203452031838497e-05, "loss": 1.0655, "step": 2315 }, { "epoch": 0.15, "learning_rate": 1.920260868413516e-05, "loss": 1.0237, "step": 2316 }, { "epoch": 0.15, "learning_rate": 1.92017649087579e-05, "loss": 1.0684, "step": 2317 }, { "epoch": 0.15, "learning_rate": 1.9200920705745933e-05, "loss": 1.0796, "step": 2318 }, { "epoch": 0.15, "learning_rate": 1.920007607513849e-05, "loss": 1.0443, "step": 2319 }, { "epoch": 0.15, "learning_rate": 1.9199231016974824e-05, "loss": 1.0612, "step": 2320 }, { "epoch": 0.15, "learning_rate": 1.9198385531294204e-05, "loss": 1.1014, "step": 2321 }, { "epoch": 0.15, "learning_rate": 1.919753961813593e-05, "loss": 1.0391, "step": 2322 }, { "epoch": 0.15, "learning_rate": 1.9196693277539306e-05, "loss": 1.091, "step": 2323 }, { "epoch": 0.15, "learning_rate": 1.919584650954367e-05, "loss": 1.0397, "step": 2324 }, { "epoch": 0.15, "learning_rate": 1.9194999314188368e-05, "loss": 1.0777, "step": 2325 }, { "epoch": 0.15, "learning_rate": 1.9194151691512778e-05, "loss": 1.0734, "step": 2326 }, { "epoch": 0.15, "learning_rate": 1.919330364155629e-05, "loss": 1.0566, "step": 2327 }, { "epoch": 0.15, "learning_rate": 1.9192455164358314e-05, "loss": 1.0455, "step": 2328 }, { "epoch": 0.16, "learning_rate": 1.9191606259958278e-05, "loss": 1.0552, "step": 2329 }, { "epoch": 0.16, "learning_rate": 1.9190756928395645e-05, "loss": 1.0628, "step": 2330 }, { "epoch": 0.16, "learning_rate": 1.918990716970987e-05, "loss": 1.0945, "step": 2331 }, { "epoch": 0.16, "learning_rate": 1.9189056983940457e-05, "loss": 1.0615, "step": 2332 }, { "epoch": 0.16, "learning_rate": 1.9188206371126913e-05, "loss": 1.0523, "step": 2333 }, { "epoch": 0.16, "learning_rate": 1.9187355331308764e-05, "loss": 1.0594, "step": 2334 }, { "epoch": 0.16, "learning_rate": 1.9186503864525565e-05, "loss": 1.0322, "step": 2335 }, { "epoch": 0.16, "learning_rate": 1.9185651970816886e-05, "loss": 1.0684, "step": 2336 }, { "epoch": 0.16, "learning_rate": 1.9184799650222318e-05, "loss": 1.0615, "step": 2337 }, { "epoch": 0.16, "learning_rate": 1.9183946902781468e-05, "loss": 1.0515, "step": 2338 }, { "epoch": 0.16, "learning_rate": 1.9183093728533966e-05, "loss": 1.0767, "step": 2339 }, { "epoch": 0.16, "learning_rate": 1.9182240127519465e-05, "loss": 1.0213, "step": 2340 }, { "epoch": 0.16, "learning_rate": 1.9181386099777636e-05, "loss": 1.0821, "step": 2341 }, { "epoch": 0.16, "learning_rate": 1.9180531645348157e-05, "loss": 1.0581, "step": 2342 }, { "epoch": 0.16, "learning_rate": 1.917967676427075e-05, "loss": 1.0738, "step": 2343 }, { "epoch": 0.16, "learning_rate": 1.917882145658514e-05, "loss": 1.0569, "step": 2344 }, { "epoch": 0.16, "learning_rate": 1.9177965722331072e-05, "loss": 1.059, "step": 2345 }, { "epoch": 0.16, "learning_rate": 1.917710956154832e-05, "loss": 1.0233, "step": 2346 }, { "epoch": 0.16, "learning_rate": 1.917625297427667e-05, "loss": 1.0572, "step": 2347 }, { "epoch": 0.16, "learning_rate": 1.9175395960555927e-05, "loss": 1.0385, "step": 2348 }, { "epoch": 0.16, "learning_rate": 1.917453852042592e-05, "loss": 1.0492, "step": 2349 }, { "epoch": 0.16, "learning_rate": 1.917368065392651e-05, "loss": 1.033, "step": 2350 }, { "epoch": 0.16, "learning_rate": 1.9172822361097542e-05, "loss": 1.062, "step": 2351 }, { "epoch": 0.16, "learning_rate": 1.9171963641978923e-05, "loss": 1.0477, "step": 2352 }, { "epoch": 0.16, "learning_rate": 1.917110449661055e-05, "loss": 1.0243, "step": 2353 }, { "epoch": 0.16, "learning_rate": 1.9170244925032354e-05, "loss": 1.0486, "step": 2354 }, { "epoch": 0.16, "learning_rate": 1.916938492728428e-05, "loss": 1.0586, "step": 2355 }, { "epoch": 0.16, "learning_rate": 1.9168524503406297e-05, "loss": 1.0103, "step": 2356 }, { "epoch": 0.16, "learning_rate": 1.9167663653438392e-05, "loss": 1.0564, "step": 2357 }, { "epoch": 0.16, "learning_rate": 1.9166802377420568e-05, "loss": 1.0652, "step": 2358 }, { "epoch": 0.16, "learning_rate": 1.9165940675392852e-05, "loss": 1.078, "step": 2359 }, { "epoch": 0.16, "learning_rate": 1.9165078547395293e-05, "loss": 1.0895, "step": 2360 }, { "epoch": 0.16, "learning_rate": 1.9164215993467952e-05, "loss": 1.0887, "step": 2361 }, { "epoch": 0.16, "learning_rate": 1.9163353013650917e-05, "loss": 1.0557, "step": 2362 }, { "epoch": 0.16, "learning_rate": 1.91624896079843e-05, "loss": 1.0399, "step": 2363 }, { "epoch": 0.16, "learning_rate": 1.9161625776508216e-05, "loss": 1.0842, "step": 2364 }, { "epoch": 0.16, "learning_rate": 1.916076151926281e-05, "loss": 1.0862, "step": 2365 }, { "epoch": 0.16, "learning_rate": 1.9159896836288255e-05, "loss": 1.0319, "step": 2366 }, { "epoch": 0.16, "learning_rate": 1.915903172762473e-05, "loss": 1.0144, "step": 2367 }, { "epoch": 0.16, "learning_rate": 1.9158166193312437e-05, "loss": 1.0925, "step": 2368 }, { "epoch": 0.16, "learning_rate": 1.9157300233391608e-05, "loss": 1.0335, "step": 2369 }, { "epoch": 0.16, "learning_rate": 1.915643384790248e-05, "loss": 1.083, "step": 2370 }, { "epoch": 0.16, "learning_rate": 1.915556703688532e-05, "loss": 1.0888, "step": 2371 }, { "epoch": 0.16, "learning_rate": 1.915469980038041e-05, "loss": 1.1247, "step": 2372 }, { "epoch": 0.16, "learning_rate": 1.9153832138428054e-05, "loss": 1.0504, "step": 2373 }, { "epoch": 0.16, "learning_rate": 1.915296405106857e-05, "loss": 1.0181, "step": 2374 }, { "epoch": 0.16, "learning_rate": 1.915209553834231e-05, "loss": 1.0468, "step": 2375 }, { "epoch": 0.16, "learning_rate": 1.915122660028963e-05, "loss": 1.052, "step": 2376 }, { "epoch": 0.16, "learning_rate": 1.9150357236950916e-05, "loss": 1.0889, "step": 2377 }, { "epoch": 0.16, "learning_rate": 1.9149487448366565e-05, "loss": 1.0389, "step": 2378 }, { "epoch": 0.16, "learning_rate": 1.9148617234577004e-05, "loss": 1.0341, "step": 2379 }, { "epoch": 0.16, "learning_rate": 1.9147746595622668e-05, "loss": 1.0273, "step": 2380 }, { "epoch": 0.16, "learning_rate": 1.9146875531544026e-05, "loss": 1.0865, "step": 2381 }, { "epoch": 0.16, "learning_rate": 1.9146004042381555e-05, "loss": 1.0426, "step": 2382 }, { "epoch": 0.16, "learning_rate": 1.9145132128175755e-05, "loss": 1.0736, "step": 2383 }, { "epoch": 0.16, "learning_rate": 1.914425978896715e-05, "loss": 1.0454, "step": 2384 }, { "epoch": 0.16, "learning_rate": 1.9143387024796278e-05, "loss": 1.047, "step": 2385 }, { "epoch": 0.16, "learning_rate": 1.91425138357037e-05, "loss": 1.0539, "step": 2386 }, { "epoch": 0.16, "learning_rate": 1.9141640221729996e-05, "loss": 1.0347, "step": 2387 }, { "epoch": 0.16, "learning_rate": 1.914076618291576e-05, "loss": 1.0215, "step": 2388 }, { "epoch": 0.16, "learning_rate": 1.913989171930162e-05, "loss": 1.1012, "step": 2389 }, { "epoch": 0.16, "learning_rate": 1.9139016830928213e-05, "loss": 1.0683, "step": 2390 }, { "epoch": 0.16, "learning_rate": 1.9138141517836192e-05, "loss": 1.0346, "step": 2391 }, { "epoch": 0.16, "learning_rate": 1.913726578006624e-05, "loss": 1.0261, "step": 2392 }, { "epoch": 0.16, "learning_rate": 1.9136389617659056e-05, "loss": 1.0472, "step": 2393 }, { "epoch": 0.16, "learning_rate": 1.9135513030655358e-05, "loss": 1.0639, "step": 2394 }, { "epoch": 0.16, "learning_rate": 1.9134636019095883e-05, "loss": 1.0706, "step": 2395 }, { "epoch": 0.16, "learning_rate": 1.913375858302138e-05, "loss": 1.0651, "step": 2396 }, { "epoch": 0.16, "learning_rate": 1.9132880722472646e-05, "loss": 1.0683, "step": 2397 }, { "epoch": 0.16, "learning_rate": 1.913200243749046e-05, "loss": 1.0834, "step": 2398 }, { "epoch": 0.16, "learning_rate": 1.9131123728115644e-05, "loss": 1.0272, "step": 2399 }, { "epoch": 0.16, "learning_rate": 1.9130244594389032e-05, "loss": 1.0687, "step": 2400 }, { "epoch": 0.16, "learning_rate": 1.912936503635149e-05, "loss": 1.0556, "step": 2401 }, { "epoch": 0.16, "learning_rate": 1.912848505404388e-05, "loss": 1.0099, "step": 2402 }, { "epoch": 0.16, "learning_rate": 1.912760464750711e-05, "loss": 1.0543, "step": 2403 }, { "epoch": 0.16, "learning_rate": 1.9126723816782088e-05, "loss": 1.059, "step": 2404 }, { "epoch": 0.16, "learning_rate": 1.9125842561909752e-05, "loss": 1.0398, "step": 2405 }, { "epoch": 0.16, "learning_rate": 1.9124960882931055e-05, "loss": 1.0355, "step": 2406 }, { "epoch": 0.16, "learning_rate": 1.912407877988697e-05, "loss": 1.0646, "step": 2407 }, { "epoch": 0.16, "learning_rate": 1.9123196252818494e-05, "loss": 1.0432, "step": 2408 }, { "epoch": 0.16, "learning_rate": 1.9122313301766638e-05, "loss": 1.0376, "step": 2409 }, { "epoch": 0.16, "learning_rate": 1.9121429926772438e-05, "loss": 1.0356, "step": 2410 }, { "epoch": 0.16, "learning_rate": 1.9120546127876944e-05, "loss": 1.0272, "step": 2411 }, { "epoch": 0.16, "learning_rate": 1.9119661905121236e-05, "loss": 1.0423, "step": 2412 }, { "epoch": 0.16, "learning_rate": 1.9118777258546396e-05, "loss": 1.0273, "step": 2413 }, { "epoch": 0.16, "learning_rate": 1.9117892188193547e-05, "loss": 1.0146, "step": 2414 }, { "epoch": 0.16, "learning_rate": 1.911700669410381e-05, "loss": 1.0939, "step": 2415 }, { "epoch": 0.16, "learning_rate": 1.9116120776318354e-05, "loss": 1.061, "step": 2416 }, { "epoch": 0.16, "learning_rate": 1.9115234434878325e-05, "loss": 1.024, "step": 2417 }, { "epoch": 0.16, "learning_rate": 1.9114347669824937e-05, "loss": 1.0136, "step": 2418 }, { "epoch": 0.16, "learning_rate": 1.911346048119939e-05, "loss": 1.0507, "step": 2419 }, { "epoch": 0.16, "learning_rate": 1.9112572869042916e-05, "loss": 1.0695, "step": 2420 }, { "epoch": 0.16, "learning_rate": 1.9111684833396765e-05, "loss": 1.0464, "step": 2421 }, { "epoch": 0.16, "learning_rate": 1.911079637430221e-05, "loss": 1.0901, "step": 2422 }, { "epoch": 0.16, "learning_rate": 1.9109907491800533e-05, "loss": 1.0515, "step": 2423 }, { "epoch": 0.16, "learning_rate": 1.9109018185933048e-05, "loss": 1.0654, "step": 2424 }, { "epoch": 0.16, "learning_rate": 1.9108128456741084e-05, "loss": 1.0507, "step": 2425 }, { "epoch": 0.16, "learning_rate": 1.910723830426599e-05, "loss": 1.0561, "step": 2426 }, { "epoch": 0.16, "learning_rate": 1.9106347728549134e-05, "loss": 1.0371, "step": 2427 }, { "epoch": 0.16, "learning_rate": 1.9105456729631902e-05, "loss": 1.0354, "step": 2428 }, { "epoch": 0.16, "learning_rate": 1.9104565307555702e-05, "loss": 1.0638, "step": 2429 }, { "epoch": 0.16, "learning_rate": 1.910367346236196e-05, "loss": 1.0521, "step": 2430 }, { "epoch": 0.16, "learning_rate": 1.910278119409213e-05, "loss": 1.0507, "step": 2431 }, { "epoch": 0.16, "learning_rate": 1.910188850278767e-05, "loss": 1.0147, "step": 2432 }, { "epoch": 0.16, "learning_rate": 1.9100995388490067e-05, "loss": 1.0397, "step": 2433 }, { "epoch": 0.16, "learning_rate": 1.910010185124083e-05, "loss": 1.0613, "step": 2434 }, { "epoch": 0.16, "learning_rate": 1.9099207891081484e-05, "loss": 1.0831, "step": 2435 }, { "epoch": 0.16, "learning_rate": 1.9098313508053576e-05, "loss": 0.9846, "step": 2436 }, { "epoch": 0.16, "learning_rate": 1.9097418702198665e-05, "loss": 1.0517, "step": 2437 }, { "epoch": 0.16, "learning_rate": 1.909652347355834e-05, "loss": 1.0559, "step": 2438 }, { "epoch": 0.16, "learning_rate": 1.9095627822174202e-05, "loss": 1.0556, "step": 2439 }, { "epoch": 0.16, "learning_rate": 1.9094731748087878e-05, "loss": 1.054, "step": 2440 }, { "epoch": 0.16, "learning_rate": 1.909383525134101e-05, "loss": 1.047, "step": 2441 }, { "epoch": 0.16, "learning_rate": 1.909293833197526e-05, "loss": 1.0224, "step": 2442 }, { "epoch": 0.16, "learning_rate": 1.9092040990032314e-05, "loss": 1.0455, "step": 2443 }, { "epoch": 0.16, "learning_rate": 1.9091143225553868e-05, "loss": 1.0656, "step": 2444 }, { "epoch": 0.16, "learning_rate": 1.909024503858165e-05, "loss": 1.0557, "step": 2445 }, { "epoch": 0.16, "learning_rate": 1.9089346429157397e-05, "loss": 1.0577, "step": 2446 }, { "epoch": 0.16, "learning_rate": 1.9088447397322875e-05, "loss": 1.0263, "step": 2447 }, { "epoch": 0.16, "learning_rate": 1.908754794311986e-05, "loss": 1.0463, "step": 2448 }, { "epoch": 0.16, "learning_rate": 1.908664806659016e-05, "loss": 1.0289, "step": 2449 }, { "epoch": 0.16, "learning_rate": 1.9085747767775584e-05, "loss": 1.0759, "step": 2450 }, { "epoch": 0.16, "learning_rate": 1.908484704671798e-05, "loss": 1.0568, "step": 2451 }, { "epoch": 0.16, "learning_rate": 1.90839459034592e-05, "loss": 1.0373, "step": 2452 }, { "epoch": 0.16, "learning_rate": 1.9083044338041136e-05, "loss": 1.058, "step": 2453 }, { "epoch": 0.16, "learning_rate": 1.9082142350505672e-05, "loss": 1.0753, "step": 2454 }, { "epoch": 0.16, "learning_rate": 1.9081239940894734e-05, "loss": 1.041, "step": 2455 }, { "epoch": 0.16, "learning_rate": 1.9080337109250257e-05, "loss": 1.028, "step": 2456 }, { "epoch": 0.16, "learning_rate": 1.9079433855614203e-05, "loss": 1.0954, "step": 2457 }, { "epoch": 0.16, "learning_rate": 1.9078530180028544e-05, "loss": 1.0409, "step": 2458 }, { "epoch": 0.16, "learning_rate": 1.9077626082535276e-05, "loss": 1.0567, "step": 2459 }, { "epoch": 0.16, "learning_rate": 1.907672156317642e-05, "loss": 1.015, "step": 2460 }, { "epoch": 0.16, "learning_rate": 1.9075816621994008e-05, "loss": 1.0696, "step": 2461 }, { "epoch": 0.16, "learning_rate": 1.9074911259030098e-05, "loss": 1.0749, "step": 2462 }, { "epoch": 0.16, "learning_rate": 1.9074005474326762e-05, "loss": 1.0643, "step": 2463 }, { "epoch": 0.16, "learning_rate": 1.90730992679261e-05, "loss": 1.0227, "step": 2464 }, { "epoch": 0.16, "learning_rate": 1.9072192639870218e-05, "loss": 1.0763, "step": 2465 }, { "epoch": 0.16, "learning_rate": 1.9071285590201255e-05, "loss": 0.9873, "step": 2466 }, { "epoch": 0.16, "learning_rate": 1.9070378118961365e-05, "loss": 1.0472, "step": 2467 }, { "epoch": 0.16, "learning_rate": 1.9069470226192724e-05, "loss": 1.0728, "step": 2468 }, { "epoch": 0.16, "learning_rate": 1.9068561911937517e-05, "loss": 1.0829, "step": 2469 }, { "epoch": 0.16, "learning_rate": 1.906765317623796e-05, "loss": 1.0603, "step": 2470 }, { "epoch": 0.16, "learning_rate": 1.9066744019136282e-05, "loss": 1.0744, "step": 2471 }, { "epoch": 0.16, "learning_rate": 1.906583444067474e-05, "loss": 1.0629, "step": 2472 }, { "epoch": 0.16, "learning_rate": 1.90649244408956e-05, "loss": 1.0484, "step": 2473 }, { "epoch": 0.16, "learning_rate": 1.9064014019841155e-05, "loss": 1.0287, "step": 2474 }, { "epoch": 0.16, "learning_rate": 1.9063103177553714e-05, "loss": 1.0461, "step": 2475 }, { "epoch": 0.16, "learning_rate": 1.9062191914075606e-05, "loss": 1.0093, "step": 2476 }, { "epoch": 0.16, "learning_rate": 1.9061280229449186e-05, "loss": 1.0392, "step": 2477 }, { "epoch": 0.16, "learning_rate": 1.9060368123716816e-05, "loss": 1.0386, "step": 2478 }, { "epoch": 0.17, "learning_rate": 1.9059455596920885e-05, "loss": 1.0821, "step": 2479 }, { "epoch": 0.17, "learning_rate": 1.9058542649103803e-05, "loss": 1.0318, "step": 2480 }, { "epoch": 0.17, "learning_rate": 1.9057629280307998e-05, "loss": 1.0474, "step": 2481 }, { "epoch": 0.17, "learning_rate": 1.9056715490575917e-05, "loss": 1.0204, "step": 2482 }, { "epoch": 0.17, "learning_rate": 1.9055801279950024e-05, "loss": 1.0355, "step": 2483 }, { "epoch": 0.17, "learning_rate": 1.9054886648472806e-05, "loss": 1.0247, "step": 2484 }, { "epoch": 0.17, "learning_rate": 1.9053971596186774e-05, "loss": 1.0652, "step": 2485 }, { "epoch": 0.17, "learning_rate": 1.905305612313445e-05, "loss": 1.0605, "step": 2486 }, { "epoch": 0.17, "learning_rate": 1.9052140229358375e-05, "loss": 1.0656, "step": 2487 }, { "epoch": 0.17, "learning_rate": 1.9051223914901114e-05, "loss": 1.0168, "step": 2488 }, { "epoch": 0.17, "learning_rate": 1.9050307179805252e-05, "loss": 1.0579, "step": 2489 }, { "epoch": 0.17, "learning_rate": 1.9049390024113402e-05, "loss": 1.0427, "step": 2490 }, { "epoch": 0.17, "learning_rate": 1.9048472447868175e-05, "loss": 1.0585, "step": 2491 }, { "epoch": 0.17, "learning_rate": 1.904755445111222e-05, "loss": 1.0384, "step": 2492 }, { "epoch": 0.17, "learning_rate": 1.9046636033888192e-05, "loss": 1.0542, "step": 2493 }, { "epoch": 0.17, "learning_rate": 1.9045717196238783e-05, "loss": 1.0601, "step": 2494 }, { "epoch": 0.17, "learning_rate": 1.9044797938206685e-05, "loss": 1.0831, "step": 2495 }, { "epoch": 0.17, "learning_rate": 1.9043878259834626e-05, "loss": 1.046, "step": 2496 }, { "epoch": 0.17, "learning_rate": 1.904295816116534e-05, "loss": 1.0585, "step": 2497 }, { "epoch": 0.17, "learning_rate": 1.904203764224159e-05, "loss": 1.0598, "step": 2498 }, { "epoch": 0.17, "learning_rate": 1.9041116703106156e-05, "loss": 1.0409, "step": 2499 }, { "epoch": 0.17, "learning_rate": 1.9040195343801838e-05, "loss": 1.0285, "step": 2500 }, { "epoch": 0.17, "learning_rate": 1.903927356437145e-05, "loss": 1.0646, "step": 2501 }, { "epoch": 0.17, "learning_rate": 1.9038351364857836e-05, "loss": 1.0635, "step": 2502 }, { "epoch": 0.17, "learning_rate": 1.9037428745303845e-05, "loss": 1.0183, "step": 2503 }, { "epoch": 0.17, "learning_rate": 1.9036505705752365e-05, "loss": 1.0428, "step": 2504 }, { "epoch": 0.17, "learning_rate": 1.9035582246246283e-05, "loss": 1.0223, "step": 2505 }, { "epoch": 0.17, "learning_rate": 1.9034658366828517e-05, "loss": 1.0321, "step": 2506 }, { "epoch": 0.17, "learning_rate": 1.9033734067542007e-05, "loss": 1.0554, "step": 2507 }, { "epoch": 0.17, "learning_rate": 1.9032809348429703e-05, "loss": 1.0141, "step": 2508 }, { "epoch": 0.17, "learning_rate": 1.9031884209534585e-05, "loss": 1.023, "step": 2509 }, { "epoch": 0.17, "learning_rate": 1.9030958650899642e-05, "loss": 1.0325, "step": 2510 }, { "epoch": 0.17, "learning_rate": 1.9030032672567887e-05, "loss": 1.0797, "step": 2511 }, { "epoch": 0.17, "learning_rate": 1.902910627458236e-05, "loss": 1.0279, "step": 2512 }, { "epoch": 0.17, "learning_rate": 1.9028179456986108e-05, "loss": 1.0623, "step": 2513 }, { "epoch": 0.17, "learning_rate": 1.9027252219822206e-05, "loss": 1.0188, "step": 2514 }, { "epoch": 0.17, "learning_rate": 1.902632456313374e-05, "loss": 1.0455, "step": 2515 }, { "epoch": 0.17, "learning_rate": 1.9025396486963827e-05, "loss": 1.0266, "step": 2516 }, { "epoch": 0.17, "learning_rate": 1.9024467991355595e-05, "loss": 1.0143, "step": 2517 }, { "epoch": 0.17, "learning_rate": 1.90235390763522e-05, "loss": 1.0153, "step": 2518 }, { "epoch": 0.17, "learning_rate": 1.9022609741996797e-05, "loss": 1.0248, "step": 2519 }, { "epoch": 0.17, "learning_rate": 1.902167998833259e-05, "loss": 1.0469, "step": 2520 }, { "epoch": 0.17, "learning_rate": 1.902074981540278e-05, "loss": 1.0654, "step": 2521 }, { "epoch": 0.17, "learning_rate": 1.9019819223250596e-05, "loss": 1.0399, "step": 2522 }, { "epoch": 0.17, "learning_rate": 1.901888821191929e-05, "loss": 1.0459, "step": 2523 }, { "epoch": 0.17, "learning_rate": 1.9017956781452126e-05, "loss": 1.028, "step": 2524 }, { "epoch": 0.17, "learning_rate": 1.9017024931892387e-05, "loss": 1.0411, "step": 2525 }, { "epoch": 0.17, "learning_rate": 1.9016092663283383e-05, "loss": 1.0237, "step": 2526 }, { "epoch": 0.17, "learning_rate": 1.9015159975668438e-05, "loss": 1.0537, "step": 2527 }, { "epoch": 0.17, "learning_rate": 1.9014226869090894e-05, "loss": 1.0378, "step": 2528 }, { "epoch": 0.17, "learning_rate": 1.9013293343594125e-05, "loss": 1.0262, "step": 2529 }, { "epoch": 0.17, "learning_rate": 1.9012359399221504e-05, "loss": 1.0656, "step": 2530 }, { "epoch": 0.17, "learning_rate": 1.901142503601644e-05, "loss": 1.0484, "step": 2531 }, { "epoch": 0.17, "learning_rate": 1.9010490254022355e-05, "loss": 1.0297, "step": 2532 }, { "epoch": 0.17, "learning_rate": 1.9009555053282687e-05, "loss": 1.0799, "step": 2533 }, { "epoch": 0.17, "learning_rate": 1.9008619433840906e-05, "loss": 1.062, "step": 2534 }, { "epoch": 0.17, "learning_rate": 1.900768339574049e-05, "loss": 1.0879, "step": 2535 }, { "epoch": 0.17, "learning_rate": 1.9006746939024935e-05, "loss": 1.0252, "step": 2536 }, { "epoch": 0.17, "learning_rate": 1.9005810063737763e-05, "loss": 1.0372, "step": 2537 }, { "epoch": 0.17, "learning_rate": 1.9004872769922515e-05, "loss": 1.0496, "step": 2538 }, { "epoch": 0.17, "learning_rate": 1.9003935057622752e-05, "loss": 1.0624, "step": 2539 }, { "epoch": 0.17, "learning_rate": 1.9002996926882052e-05, "loss": 1.0581, "step": 2540 }, { "epoch": 0.17, "learning_rate": 1.900205837774401e-05, "loss": 1.0616, "step": 2541 }, { "epoch": 0.17, "learning_rate": 1.900111941025224e-05, "loss": 1.0486, "step": 2542 }, { "epoch": 0.17, "learning_rate": 1.9000180024450385e-05, "loss": 1.0391, "step": 2543 }, { "epoch": 0.17, "learning_rate": 1.89992402203821e-05, "loss": 1.0267, "step": 2544 }, { "epoch": 0.17, "learning_rate": 1.899829999809106e-05, "loss": 1.0156, "step": 2545 }, { "epoch": 0.17, "learning_rate": 1.8997359357620963e-05, "loss": 1.0199, "step": 2546 }, { "epoch": 0.17, "learning_rate": 1.8996418299015515e-05, "loss": 1.089, "step": 2547 }, { "epoch": 0.17, "learning_rate": 1.899547682231846e-05, "loss": 1.0426, "step": 2548 }, { "epoch": 0.17, "learning_rate": 1.899453492757354e-05, "loss": 1.0498, "step": 2549 }, { "epoch": 0.17, "learning_rate": 1.8993592614824542e-05, "loss": 1.0635, "step": 2550 }, { "epoch": 0.17, "learning_rate": 1.899264988411525e-05, "loss": 1.0369, "step": 2551 }, { "epoch": 0.17, "learning_rate": 1.8991706735489472e-05, "loss": 1.0566, "step": 2552 }, { "epoch": 0.17, "learning_rate": 1.8990763168991045e-05, "loss": 1.0908, "step": 2553 }, { "epoch": 0.17, "learning_rate": 1.898981918466382e-05, "loss": 1.0758, "step": 2554 }, { "epoch": 0.17, "learning_rate": 1.8988874782551662e-05, "loss": 1.051, "step": 2555 }, { "epoch": 0.17, "learning_rate": 1.8987929962698465e-05, "loss": 1.0355, "step": 2556 }, { "epoch": 0.17, "learning_rate": 1.8986984725148138e-05, "loss": 1.0359, "step": 2557 }, { "epoch": 0.17, "learning_rate": 1.89860390699446e-05, "loss": 1.0423, "step": 2558 }, { "epoch": 0.17, "learning_rate": 1.8985092997131814e-05, "loss": 1.041, "step": 2559 }, { "epoch": 0.17, "learning_rate": 1.898414650675373e-05, "loss": 1.035, "step": 2560 }, { "epoch": 0.17, "learning_rate": 1.898319959885435e-05, "loss": 1.0779, "step": 2561 }, { "epoch": 0.17, "learning_rate": 1.898225227347767e-05, "loss": 1.0379, "step": 2562 }, { "epoch": 0.17, "learning_rate": 1.8981304530667715e-05, "loss": 1.0571, "step": 2563 }, { "epoch": 0.17, "learning_rate": 1.8980356370468537e-05, "loss": 1.0325, "step": 2564 }, { "epoch": 0.17, "learning_rate": 1.8979407792924194e-05, "loss": 1.0436, "step": 2565 }, { "epoch": 0.17, "learning_rate": 1.897845879807877e-05, "loss": 1.012, "step": 2566 }, { "epoch": 0.17, "learning_rate": 1.8977509385976367e-05, "loss": 1.0579, "step": 2567 }, { "epoch": 0.17, "learning_rate": 1.897655955666111e-05, "loss": 1.0822, "step": 2568 }, { "epoch": 0.17, "learning_rate": 1.897560931017714e-05, "loss": 1.0111, "step": 2569 }, { "epoch": 0.17, "learning_rate": 1.8974658646568612e-05, "loss": 1.0092, "step": 2570 }, { "epoch": 0.17, "learning_rate": 1.8973707565879717e-05, "loss": 1.0659, "step": 2571 }, { "epoch": 0.17, "learning_rate": 1.8972756068154648e-05, "loss": 1.0412, "step": 2572 }, { "epoch": 0.17, "learning_rate": 1.8971804153437624e-05, "loss": 1.1029, "step": 2573 }, { "epoch": 0.17, "learning_rate": 1.8970851821772884e-05, "loss": 1.0627, "step": 2574 }, { "epoch": 0.17, "learning_rate": 1.8969899073204687e-05, "loss": 1.0397, "step": 2575 }, { "epoch": 0.17, "learning_rate": 1.896894590777731e-05, "loss": 1.0389, "step": 2576 }, { "epoch": 0.17, "learning_rate": 1.8967992325535046e-05, "loss": 1.0369, "step": 2577 }, { "epoch": 0.17, "learning_rate": 1.896703832652222e-05, "loss": 1.0487, "step": 2578 }, { "epoch": 0.17, "learning_rate": 1.896608391078316e-05, "loss": 1.0351, "step": 2579 }, { "epoch": 0.17, "learning_rate": 1.8965129078362218e-05, "loss": 1.0469, "step": 2580 }, { "epoch": 0.17, "learning_rate": 1.8964173829303775e-05, "loss": 1.0128, "step": 2581 }, { "epoch": 0.17, "learning_rate": 1.8963218163652218e-05, "loss": 1.0455, "step": 2582 }, { "epoch": 0.17, "learning_rate": 1.8962262081451965e-05, "loss": 1.026, "step": 2583 }, { "epoch": 0.17, "learning_rate": 1.8961305582747447e-05, "loss": 0.9952, "step": 2584 }, { "epoch": 0.17, "learning_rate": 1.8960348667583112e-05, "loss": 1.0631, "step": 2585 }, { "epoch": 0.17, "learning_rate": 1.8959391336003437e-05, "loss": 1.028, "step": 2586 }, { "epoch": 0.17, "learning_rate": 1.895843358805291e-05, "loss": 1.0696, "step": 2587 }, { "epoch": 0.17, "learning_rate": 1.8957475423776036e-05, "loss": 0.9864, "step": 2588 }, { "epoch": 0.17, "learning_rate": 1.895651684321735e-05, "loss": 1.0241, "step": 2589 }, { "epoch": 0.17, "learning_rate": 1.8955557846421394e-05, "loss": 1.0703, "step": 2590 }, { "epoch": 0.17, "learning_rate": 1.8954598433432744e-05, "loss": 1.1136, "step": 2591 }, { "epoch": 0.17, "learning_rate": 1.8953638604295976e-05, "loss": 1.028, "step": 2592 }, { "epoch": 0.17, "learning_rate": 1.8952678359055707e-05, "loss": 1.0241, "step": 2593 }, { "epoch": 0.17, "learning_rate": 1.8951717697756557e-05, "loss": 1.0096, "step": 2594 }, { "epoch": 0.17, "learning_rate": 1.895075662044317e-05, "loss": 1.0264, "step": 2595 }, { "epoch": 0.17, "learning_rate": 1.894979512716021e-05, "loss": 1.0359, "step": 2596 }, { "epoch": 0.17, "learning_rate": 1.894883321795237e-05, "loss": 1.0943, "step": 2597 }, { "epoch": 0.17, "learning_rate": 1.8947870892864338e-05, "loss": 1.0224, "step": 2598 }, { "epoch": 0.17, "learning_rate": 1.8946908151940846e-05, "loss": 1.0563, "step": 2599 }, { "epoch": 0.17, "learning_rate": 1.8945944995226635e-05, "loss": 1.0811, "step": 2600 }, { "epoch": 0.17, "learning_rate": 1.894498142276646e-05, "loss": 1.0396, "step": 2601 }, { "epoch": 0.17, "learning_rate": 1.8944017434605114e-05, "loss": 1.0025, "step": 2602 }, { "epoch": 0.17, "learning_rate": 1.894305303078738e-05, "loss": 1.0863, "step": 2603 }, { "epoch": 0.17, "learning_rate": 1.8942088211358088e-05, "loss": 1.0418, "step": 2604 }, { "epoch": 0.17, "learning_rate": 1.8941122976362072e-05, "loss": 1.0683, "step": 2605 }, { "epoch": 0.17, "learning_rate": 1.894015732584419e-05, "loss": 1.0475, "step": 2606 }, { "epoch": 0.17, "learning_rate": 1.893919125984932e-05, "loss": 1.0636, "step": 2607 }, { "epoch": 0.17, "learning_rate": 1.8938224778422357e-05, "loss": 1.008, "step": 2608 }, { "epoch": 0.17, "learning_rate": 1.8937257881608217e-05, "loss": 1.0589, "step": 2609 }, { "epoch": 0.17, "learning_rate": 1.8936290569451837e-05, "loss": 1.0594, "step": 2610 }, { "epoch": 0.17, "learning_rate": 1.893532284199816e-05, "loss": 1.0596, "step": 2611 }, { "epoch": 0.17, "learning_rate": 1.8934354699292177e-05, "loss": 1.0472, "step": 2612 }, { "epoch": 0.17, "learning_rate": 1.8933386141378868e-05, "loss": 1.0801, "step": 2613 }, { "epoch": 0.17, "learning_rate": 1.8932417168303246e-05, "loss": 1.068, "step": 2614 }, { "epoch": 0.17, "learning_rate": 1.893144778011035e-05, "loss": 1.0692, "step": 2615 }, { "epoch": 0.17, "learning_rate": 1.893047797684522e-05, "loss": 1.0833, "step": 2616 }, { "epoch": 0.17, "learning_rate": 1.8929507758552934e-05, "loss": 1.05, "step": 2617 }, { "epoch": 0.17, "learning_rate": 1.8928537125278576e-05, "loss": 1.0572, "step": 2618 }, { "epoch": 0.17, "learning_rate": 1.8927566077067257e-05, "loss": 1.0038, "step": 2619 }, { "epoch": 0.17, "learning_rate": 1.8926594613964103e-05, "loss": 1.0571, "step": 2620 }, { "epoch": 0.17, "learning_rate": 1.8925622736014264e-05, "loss": 1.0492, "step": 2621 }, { "epoch": 0.17, "learning_rate": 1.89246504432629e-05, "loss": 1.0397, "step": 2622 }, { "epoch": 0.17, "learning_rate": 1.8923677735755204e-05, "loss": 1.0389, "step": 2623 }, { "epoch": 0.17, "learning_rate": 1.8922704613536374e-05, "loss": 0.9945, "step": 2624 }, { "epoch": 0.17, "learning_rate": 1.892173107665164e-05, "loss": 1.0851, "step": 2625 }, { "epoch": 0.17, "learning_rate": 1.8920757125146244e-05, "loss": 1.0111, "step": 2626 }, { "epoch": 0.17, "learning_rate": 1.8919782759065444e-05, "loss": 1.0383, "step": 2627 }, { "epoch": 0.17, "learning_rate": 1.8918807978454525e-05, "loss": 1.0746, "step": 2628 }, { "epoch": 0.17, "learning_rate": 1.8917832783358788e-05, "loss": 0.9904, "step": 2629 }, { "epoch": 0.18, "learning_rate": 1.8916857173823552e-05, "loss": 1.0453, "step": 2630 }, { "epoch": 0.18, "learning_rate": 1.8915881149894163e-05, "loss": 1.06, "step": 2631 }, { "epoch": 0.18, "learning_rate": 1.8914904711615972e-05, "loss": 1.0362, "step": 2632 }, { "epoch": 0.18, "learning_rate": 1.8913927859034356e-05, "loss": 1.0374, "step": 2633 }, { "epoch": 0.18, "learning_rate": 1.891295059219472e-05, "loss": 0.9718, "step": 2634 }, { "epoch": 0.18, "learning_rate": 1.891197291114248e-05, "loss": 1.0734, "step": 2635 }, { "epoch": 0.18, "learning_rate": 1.8910994815923066e-05, "loss": 1.049, "step": 2636 }, { "epoch": 0.18, "learning_rate": 1.8910016306581936e-05, "loss": 1.0157, "step": 2637 }, { "epoch": 0.18, "learning_rate": 1.8909037383164562e-05, "loss": 1.0304, "step": 2638 }, { "epoch": 0.18, "learning_rate": 1.8908058045716443e-05, "loss": 1.0535, "step": 2639 }, { "epoch": 0.18, "learning_rate": 1.890707829428309e-05, "loss": 1.0619, "step": 2640 }, { "epoch": 0.18, "learning_rate": 1.8906098128910032e-05, "loss": 1.0859, "step": 2641 }, { "epoch": 0.18, "learning_rate": 1.8905117549642823e-05, "loss": 0.9794, "step": 2642 }, { "epoch": 0.18, "learning_rate": 1.8904136556527032e-05, "loss": 1.0794, "step": 2643 }, { "epoch": 0.18, "learning_rate": 1.890315514960825e-05, "loss": 1.0609, "step": 2644 }, { "epoch": 0.18, "learning_rate": 1.8902173328932085e-05, "loss": 1.0743, "step": 2645 }, { "epoch": 0.18, "learning_rate": 1.8901191094544167e-05, "loss": 1.0343, "step": 2646 }, { "epoch": 0.18, "learning_rate": 1.8900208446490143e-05, "loss": 1.0593, "step": 2647 }, { "epoch": 0.18, "learning_rate": 1.889922538481568e-05, "loss": 1.0096, "step": 2648 }, { "epoch": 0.18, "learning_rate": 1.8898241909566463e-05, "loss": 1.0408, "step": 2649 }, { "epoch": 0.18, "learning_rate": 1.8897258020788196e-05, "loss": 1.0369, "step": 2650 }, { "epoch": 0.18, "learning_rate": 1.8896273718526604e-05, "loss": 1.0412, "step": 2651 }, { "epoch": 0.18, "learning_rate": 1.889528900282743e-05, "loss": 1.0452, "step": 2652 }, { "epoch": 0.18, "learning_rate": 1.8894303873736438e-05, "loss": 1.0518, "step": 2653 }, { "epoch": 0.18, "learning_rate": 1.8893318331299414e-05, "loss": 1.0891, "step": 2654 }, { "epoch": 0.18, "learning_rate": 1.8892332375562153e-05, "loss": 1.0164, "step": 2655 }, { "epoch": 0.18, "learning_rate": 1.889134600657048e-05, "loss": 1.0521, "step": 2656 }, { "epoch": 0.18, "learning_rate": 1.889035922437023e-05, "loss": 1.0564, "step": 2657 }, { "epoch": 0.18, "learning_rate": 1.888937202900727e-05, "loss": 1.087, "step": 2658 }, { "epoch": 0.18, "learning_rate": 1.8888384420527467e-05, "loss": 1.0563, "step": 2659 }, { "epoch": 0.18, "learning_rate": 1.8887396398976723e-05, "loss": 1.041, "step": 2660 }, { "epoch": 0.18, "learning_rate": 1.8886407964400958e-05, "loss": 1.085, "step": 2661 }, { "epoch": 0.18, "learning_rate": 1.8885419116846104e-05, "loss": 1.0219, "step": 2662 }, { "epoch": 0.18, "learning_rate": 1.8884429856358114e-05, "loss": 1.0343, "step": 2663 }, { "epoch": 0.18, "learning_rate": 1.8883440182982965e-05, "loss": 1.015, "step": 2664 }, { "epoch": 0.18, "learning_rate": 1.888245009676665e-05, "loss": 1.0304, "step": 2665 }, { "epoch": 0.18, "learning_rate": 1.8881459597755186e-05, "loss": 1.045, "step": 2666 }, { "epoch": 0.18, "learning_rate": 1.8880468685994592e-05, "loss": 1.1144, "step": 2667 }, { "epoch": 0.18, "learning_rate": 1.8879477361530933e-05, "loss": 1.0239, "step": 2668 }, { "epoch": 0.18, "learning_rate": 1.8878485624410267e-05, "loss": 1.0053, "step": 2669 }, { "epoch": 0.18, "learning_rate": 1.887749347467869e-05, "loss": 1.0311, "step": 2670 }, { "epoch": 0.18, "learning_rate": 1.8876500912382313e-05, "loss": 1.0789, "step": 2671 }, { "epoch": 0.18, "learning_rate": 1.8875507937567256e-05, "loss": 0.9853, "step": 2672 }, { "epoch": 0.18, "learning_rate": 1.887451455027967e-05, "loss": 1.0379, "step": 2673 }, { "epoch": 0.18, "learning_rate": 1.8873520750565716e-05, "loss": 1.0278, "step": 2674 }, { "epoch": 0.18, "learning_rate": 1.8872526538471588e-05, "loss": 1.076, "step": 2675 }, { "epoch": 0.18, "learning_rate": 1.887153191404348e-05, "loss": 1.0449, "step": 2676 }, { "epoch": 0.18, "learning_rate": 1.8870536877327623e-05, "loss": 1.0079, "step": 2677 }, { "epoch": 0.18, "learning_rate": 1.886954142837026e-05, "loss": 1.0079, "step": 2678 }, { "epoch": 0.18, "learning_rate": 1.8868545567217643e-05, "loss": 1.0462, "step": 2679 }, { "epoch": 0.18, "learning_rate": 1.886754929391606e-05, "loss": 1.0697, "step": 2680 }, { "epoch": 0.18, "learning_rate": 1.886655260851181e-05, "loss": 1.0392, "step": 2681 }, { "epoch": 0.18, "learning_rate": 1.8865555511051215e-05, "loss": 1.0286, "step": 2682 }, { "epoch": 0.18, "learning_rate": 1.8864558001580608e-05, "loss": 1.08, "step": 2683 }, { "epoch": 0.18, "learning_rate": 1.886356008014635e-05, "loss": 0.995, "step": 2684 }, { "epoch": 0.18, "learning_rate": 1.8862561746794815e-05, "loss": 1.0355, "step": 2685 }, { "epoch": 0.18, "learning_rate": 1.88615630015724e-05, "loss": 1.0582, "step": 2686 }, { "epoch": 0.18, "learning_rate": 1.8860563844525515e-05, "loss": 1.0342, "step": 2687 }, { "epoch": 0.18, "learning_rate": 1.8859564275700605e-05, "loss": 1.0026, "step": 2688 }, { "epoch": 0.18, "learning_rate": 1.8858564295144115e-05, "loss": 1.0046, "step": 2689 }, { "epoch": 0.18, "learning_rate": 1.885756390290252e-05, "loss": 1.0461, "step": 2690 }, { "epoch": 0.18, "learning_rate": 1.8856563099022306e-05, "loss": 1.0236, "step": 2691 }, { "epoch": 0.18, "learning_rate": 1.885556188354999e-05, "loss": 0.9837, "step": 2692 }, { "epoch": 0.18, "learning_rate": 1.8854560256532098e-05, "loss": 1.0529, "step": 2693 }, { "epoch": 0.18, "learning_rate": 1.8853558218015185e-05, "loss": 0.9982, "step": 2694 }, { "epoch": 0.18, "learning_rate": 1.885255576804581e-05, "loss": 1.0562, "step": 2695 }, { "epoch": 0.18, "learning_rate": 1.885155290667056e-05, "loss": 1.0238, "step": 2696 }, { "epoch": 0.18, "learning_rate": 1.8850549633936052e-05, "loss": 1.0434, "step": 2697 }, { "epoch": 0.18, "learning_rate": 1.88495459498889e-05, "loss": 1.019, "step": 2698 }, { "epoch": 0.18, "learning_rate": 1.8848541854575755e-05, "loss": 1.0699, "step": 2699 }, { "epoch": 0.18, "learning_rate": 1.8847537348043277e-05, "loss": 1.0054, "step": 2700 }, { "epoch": 0.18, "learning_rate": 1.8846532430338146e-05, "loss": 1.0854, "step": 2701 }, { "epoch": 0.18, "learning_rate": 1.884552710150707e-05, "loss": 1.0512, "step": 2702 }, { "epoch": 0.18, "learning_rate": 1.8844521361596766e-05, "loss": 1.0389, "step": 2703 }, { "epoch": 0.18, "learning_rate": 1.8843515210653978e-05, "loss": 1.02, "step": 2704 }, { "epoch": 0.18, "learning_rate": 1.8842508648725458e-05, "loss": 1.083, "step": 2705 }, { "epoch": 0.18, "learning_rate": 1.884150167585799e-05, "loss": 1.0285, "step": 2706 }, { "epoch": 0.18, "learning_rate": 1.8840494292098367e-05, "loss": 1.0628, "step": 2707 }, { "epoch": 0.18, "learning_rate": 1.8839486497493407e-05, "loss": 1.0216, "step": 2708 }, { "epoch": 0.18, "learning_rate": 1.8838478292089946e-05, "loss": 1.0339, "step": 2709 }, { "epoch": 0.18, "learning_rate": 1.883746967593484e-05, "loss": 1.0354, "step": 2710 }, { "epoch": 0.18, "learning_rate": 1.8836460649074956e-05, "loss": 1.0353, "step": 2711 }, { "epoch": 0.18, "learning_rate": 1.8835451211557197e-05, "loss": 1.0615, "step": 2712 }, { "epoch": 0.18, "learning_rate": 1.8834441363428464e-05, "loss": 1.0429, "step": 2713 }, { "epoch": 0.18, "learning_rate": 1.8833431104735695e-05, "loss": 1.0275, "step": 2714 }, { "epoch": 0.18, "learning_rate": 1.8832420435525836e-05, "loss": 1.0799, "step": 2715 }, { "epoch": 0.18, "learning_rate": 1.883140935584586e-05, "loss": 1.0447, "step": 2716 }, { "epoch": 0.18, "learning_rate": 1.8830397865742753e-05, "loss": 1.0462, "step": 2717 }, { "epoch": 0.18, "learning_rate": 1.882938596526352e-05, "loss": 1.0502, "step": 2718 }, { "epoch": 0.18, "learning_rate": 1.882837365445519e-05, "loss": 1.0462, "step": 2719 }, { "epoch": 0.18, "learning_rate": 1.8827360933364804e-05, "loss": 1.0078, "step": 2720 }, { "epoch": 0.18, "learning_rate": 1.882634780203943e-05, "loss": 1.0678, "step": 2721 }, { "epoch": 0.18, "learning_rate": 1.8825334260526152e-05, "loss": 1.0382, "step": 2722 }, { "epoch": 0.18, "learning_rate": 1.8824320308872072e-05, "loss": 1.0437, "step": 2723 }, { "epoch": 0.18, "learning_rate": 1.882330594712431e-05, "loss": 1.0094, "step": 2724 }, { "epoch": 0.18, "learning_rate": 1.8822291175330006e-05, "loss": 1.0335, "step": 2725 }, { "epoch": 0.18, "learning_rate": 1.8821275993536322e-05, "loss": 1.0121, "step": 2726 }, { "epoch": 0.18, "learning_rate": 1.8820260401790435e-05, "loss": 1.0102, "step": 2727 }, { "epoch": 0.18, "learning_rate": 1.881924440013954e-05, "loss": 1.0712, "step": 2728 }, { "epoch": 0.18, "learning_rate": 1.881822798863086e-05, "loss": 1.077, "step": 2729 }, { "epoch": 0.18, "learning_rate": 1.881721116731163e-05, "loss": 1.0177, "step": 2730 }, { "epoch": 0.18, "learning_rate": 1.88161939362291e-05, "loss": 1.0334, "step": 2731 }, { "epoch": 0.18, "learning_rate": 1.8815176295430547e-05, "loss": 1.0142, "step": 2732 }, { "epoch": 0.18, "learning_rate": 1.8814158244963262e-05, "loss": 0.9903, "step": 2733 }, { "epoch": 0.18, "learning_rate": 1.881313978487456e-05, "loss": 1.0332, "step": 2734 }, { "epoch": 0.18, "learning_rate": 1.881212091521177e-05, "loss": 1.0547, "step": 2735 }, { "epoch": 0.18, "learning_rate": 1.8811101636022244e-05, "loss": 0.9842, "step": 2736 }, { "epoch": 0.18, "learning_rate": 1.881008194735335e-05, "loss": 1.0399, "step": 2737 }, { "epoch": 0.18, "learning_rate": 1.880906184925247e-05, "loss": 1.0765, "step": 2738 }, { "epoch": 0.18, "learning_rate": 1.8808041341767022e-05, "loss": 1.0224, "step": 2739 }, { "epoch": 0.18, "learning_rate": 1.8807020424944428e-05, "loss": 1.0116, "step": 2740 }, { "epoch": 0.18, "learning_rate": 1.880599909883213e-05, "loss": 1.0583, "step": 2741 }, { "epoch": 0.18, "learning_rate": 1.8804977363477593e-05, "loss": 1.0358, "step": 2742 }, { "epoch": 0.18, "learning_rate": 1.8803955218928305e-05, "loss": 1.0567, "step": 2743 }, { "epoch": 0.18, "learning_rate": 1.8802932665231762e-05, "loss": 1.005, "step": 2744 }, { "epoch": 0.18, "learning_rate": 1.880190970243549e-05, "loss": 1.0523, "step": 2745 }, { "epoch": 0.18, "learning_rate": 1.880088633058703e-05, "loss": 1.0433, "step": 2746 }, { "epoch": 0.18, "learning_rate": 1.879986254973393e-05, "loss": 1.0504, "step": 2747 }, { "epoch": 0.18, "learning_rate": 1.8798838359923782e-05, "loss": 1.0641, "step": 2748 }, { "epoch": 0.18, "learning_rate": 1.879781376120418e-05, "loss": 1.0524, "step": 2749 }, { "epoch": 0.18, "learning_rate": 1.8796788753622736e-05, "loss": 1.0371, "step": 2750 }, { "epoch": 0.18, "learning_rate": 1.879576333722709e-05, "loss": 1.0953, "step": 2751 }, { "epoch": 0.18, "learning_rate": 1.879473751206489e-05, "loss": 1.0317, "step": 2752 }, { "epoch": 0.18, "learning_rate": 1.8793711278183817e-05, "loss": 1.0363, "step": 2753 }, { "epoch": 0.18, "learning_rate": 1.8792684635631562e-05, "loss": 1.0361, "step": 2754 }, { "epoch": 0.18, "learning_rate": 1.8791657584455828e-05, "loss": 1.0706, "step": 2755 }, { "epoch": 0.18, "learning_rate": 1.8790630124704356e-05, "loss": 1.0329, "step": 2756 }, { "epoch": 0.18, "learning_rate": 1.878960225642489e-05, "loss": 1.0523, "step": 2757 }, { "epoch": 0.18, "learning_rate": 1.8788573979665196e-05, "loss": 1.0392, "step": 2758 }, { "epoch": 0.18, "learning_rate": 1.878754529447307e-05, "loss": 1.0584, "step": 2759 }, { "epoch": 0.18, "learning_rate": 1.8786516200896304e-05, "loss": 1.0398, "step": 2760 }, { "epoch": 0.18, "learning_rate": 1.878548669898274e-05, "loss": 1.0565, "step": 2761 }, { "epoch": 0.18, "learning_rate": 1.8784456788780207e-05, "loss": 1.0195, "step": 2762 }, { "epoch": 0.18, "learning_rate": 1.8783426470336574e-05, "loss": 1.0548, "step": 2763 }, { "epoch": 0.18, "learning_rate": 1.878239574369973e-05, "loss": 1.0501, "step": 2764 }, { "epoch": 0.18, "learning_rate": 1.8781364608917563e-05, "loss": 1.0731, "step": 2765 }, { "epoch": 0.18, "learning_rate": 1.8780333066038006e-05, "loss": 1.0358, "step": 2766 }, { "epoch": 0.18, "learning_rate": 1.877930111510899e-05, "loss": 1.0513, "step": 2767 }, { "epoch": 0.18, "learning_rate": 1.8778268756178474e-05, "loss": 1.031, "step": 2768 }, { "epoch": 0.18, "learning_rate": 1.8777235989294434e-05, "loss": 1.0366, "step": 2769 }, { "epoch": 0.18, "learning_rate": 1.877620281450487e-05, "loss": 1.0577, "step": 2770 }, { "epoch": 0.18, "learning_rate": 1.8775169231857795e-05, "loss": 0.9734, "step": 2771 }, { "epoch": 0.18, "learning_rate": 1.8774135241401242e-05, "loss": 1.0362, "step": 2772 }, { "epoch": 0.18, "learning_rate": 1.877310084318327e-05, "loss": 1.0267, "step": 2773 }, { "epoch": 0.18, "learning_rate": 1.8772066037251936e-05, "loss": 1.0226, "step": 2774 }, { "epoch": 0.18, "learning_rate": 1.8771030823655342e-05, "loss": 1.0051, "step": 2775 }, { "epoch": 0.18, "learning_rate": 1.8769995202441594e-05, "loss": 1.0617, "step": 2776 }, { "epoch": 0.18, "learning_rate": 1.8768959173658824e-05, "loss": 1.0701, "step": 2777 }, { "epoch": 0.18, "learning_rate": 1.876792273735518e-05, "loss": 1.0356, "step": 2778 }, { "epoch": 0.18, "learning_rate": 1.876688589357882e-05, "loss": 0.9932, "step": 2779 }, { "epoch": 0.19, "learning_rate": 1.8765848642377936e-05, "loss": 1.0341, "step": 2780 }, { "epoch": 0.19, "learning_rate": 1.8764810983800736e-05, "loss": 1.0341, "step": 2781 }, { "epoch": 0.19, "learning_rate": 1.8763772917895434e-05, "loss": 1.0701, "step": 2782 }, { "epoch": 0.19, "learning_rate": 1.876273444471028e-05, "loss": 1.0571, "step": 2783 }, { "epoch": 0.19, "learning_rate": 1.876169556429353e-05, "loss": 1.0669, "step": 2784 }, { "epoch": 0.19, "learning_rate": 1.8760656276693464e-05, "loss": 1.0508, "step": 2785 }, { "epoch": 0.19, "learning_rate": 1.8759616581958387e-05, "loss": 1.0364, "step": 2786 }, { "epoch": 0.19, "learning_rate": 1.875857648013661e-05, "loss": 1.0511, "step": 2787 }, { "epoch": 0.19, "learning_rate": 1.8757535971276472e-05, "loss": 1.0551, "step": 2788 }, { "epoch": 0.19, "learning_rate": 1.8756495055426328e-05, "loss": 1.0982, "step": 2789 }, { "epoch": 0.19, "learning_rate": 1.8755453732634556e-05, "loss": 1.0036, "step": 2790 }, { "epoch": 0.19, "learning_rate": 1.8754412002949546e-05, "loss": 1.0799, "step": 2791 }, { "epoch": 0.19, "learning_rate": 1.8753369866419713e-05, "loss": 1.0211, "step": 2792 }, { "epoch": 0.19, "learning_rate": 1.8752327323093485e-05, "loss": 1.0203, "step": 2793 }, { "epoch": 0.19, "learning_rate": 1.8751284373019314e-05, "loss": 1.0259, "step": 2794 }, { "epoch": 0.19, "learning_rate": 1.8750241016245673e-05, "loss": 1.0791, "step": 2795 }, { "epoch": 0.19, "learning_rate": 1.8749197252821044e-05, "loss": 1.0291, "step": 2796 }, { "epoch": 0.19, "learning_rate": 1.8748153082793933e-05, "loss": 1.0554, "step": 2797 }, { "epoch": 0.19, "learning_rate": 1.8747108506212874e-05, "loss": 1.0352, "step": 2798 }, { "epoch": 0.19, "learning_rate": 1.8746063523126405e-05, "loss": 1.0341, "step": 2799 }, { "epoch": 0.19, "learning_rate": 1.874501813358309e-05, "loss": 1.0444, "step": 2800 }, { "epoch": 0.19, "learning_rate": 1.8743972337631516e-05, "loss": 1.0382, "step": 2801 }, { "epoch": 0.19, "learning_rate": 1.874292613532028e-05, "loss": 1.0222, "step": 2802 }, { "epoch": 0.19, "learning_rate": 1.8741879526698002e-05, "loss": 1.0366, "step": 2803 }, { "epoch": 0.19, "learning_rate": 1.8740832511813325e-05, "loss": 1.0453, "step": 2804 }, { "epoch": 0.19, "learning_rate": 1.8739785090714906e-05, "loss": 1.0453, "step": 2805 }, { "epoch": 0.19, "learning_rate": 1.873873726345142e-05, "loss": 1.0294, "step": 2806 }, { "epoch": 0.19, "learning_rate": 1.873768903007156e-05, "loss": 1.0597, "step": 2807 }, { "epoch": 0.19, "learning_rate": 1.873664039062405e-05, "loss": 1.0534, "step": 2808 }, { "epoch": 0.19, "learning_rate": 1.8735591345157614e-05, "loss": 1.0725, "step": 2809 }, { "epoch": 0.19, "learning_rate": 1.873454189372101e-05, "loss": 0.9791, "step": 2810 }, { "epoch": 0.19, "learning_rate": 1.8733492036363007e-05, "loss": 1.0384, "step": 2811 }, { "epoch": 0.19, "learning_rate": 1.8732441773132395e-05, "loss": 1.0226, "step": 2812 }, { "epoch": 0.19, "learning_rate": 1.8731391104077984e-05, "loss": 1.0445, "step": 2813 }, { "epoch": 0.19, "learning_rate": 1.8730340029248602e-05, "loss": 1.038, "step": 2814 }, { "epoch": 0.19, "learning_rate": 1.8729288548693094e-05, "loss": 1.0419, "step": 2815 }, { "epoch": 0.19, "learning_rate": 1.8728236662460328e-05, "loss": 1.0251, "step": 2816 }, { "epoch": 0.19, "learning_rate": 1.8727184370599187e-05, "loss": 1.0372, "step": 2817 }, { "epoch": 0.19, "learning_rate": 1.8726131673158576e-05, "loss": 1.0202, "step": 2818 }, { "epoch": 0.19, "learning_rate": 1.8725078570187412e-05, "loss": 1.0797, "step": 2819 }, { "epoch": 0.19, "learning_rate": 1.8724025061734642e-05, "loss": 0.985, "step": 2820 }, { "epoch": 0.19, "learning_rate": 1.8722971147849226e-05, "loss": 0.9984, "step": 2821 }, { "epoch": 0.19, "learning_rate": 1.8721916828580136e-05, "loss": 0.956, "step": 2822 }, { "epoch": 0.19, "learning_rate": 1.8720862103976376e-05, "loss": 1.0344, "step": 2823 }, { "epoch": 0.19, "learning_rate": 1.871980697408696e-05, "loss": 0.9955, "step": 2824 }, { "epoch": 0.19, "learning_rate": 1.8718751438960924e-05, "loss": 1.0859, "step": 2825 }, { "epoch": 0.19, "learning_rate": 1.871769549864732e-05, "loss": 1.049, "step": 2826 }, { "epoch": 0.19, "learning_rate": 1.871663915319522e-05, "loss": 1.0408, "step": 2827 }, { "epoch": 0.19, "learning_rate": 1.8715582402653722e-05, "loss": 1.0419, "step": 2828 }, { "epoch": 0.19, "learning_rate": 1.8714525247071928e-05, "loss": 1.0556, "step": 2829 }, { "epoch": 0.19, "learning_rate": 1.8713467686498973e-05, "loss": 1.0315, "step": 2830 }, { "epoch": 0.19, "learning_rate": 1.8712409720984008e-05, "loss": 1.024, "step": 2831 }, { "epoch": 0.19, "learning_rate": 1.8711351350576187e-05, "loss": 1.049, "step": 2832 }, { "epoch": 0.19, "learning_rate": 1.871029257532471e-05, "loss": 1.0752, "step": 2833 }, { "epoch": 0.19, "learning_rate": 1.8709233395278778e-05, "loss": 1.0195, "step": 2834 }, { "epoch": 0.19, "learning_rate": 1.870817381048761e-05, "loss": 1.0142, "step": 2835 }, { "epoch": 0.19, "learning_rate": 1.8707113821000448e-05, "loss": 1.0432, "step": 2836 }, { "epoch": 0.19, "learning_rate": 1.870605342686656e-05, "loss": 1.0217, "step": 2837 }, { "epoch": 0.19, "learning_rate": 1.8704992628135218e-05, "loss": 1.023, "step": 2838 }, { "epoch": 0.19, "learning_rate": 1.8703931424855725e-05, "loss": 1.0412, "step": 2839 }, { "epoch": 0.19, "learning_rate": 1.87028698170774e-05, "loss": 1.0647, "step": 2840 }, { "epoch": 0.19, "learning_rate": 1.870180780484957e-05, "loss": 1.0178, "step": 2841 }, { "epoch": 0.19, "learning_rate": 1.8700745388221602e-05, "loss": 1.0349, "step": 2842 }, { "epoch": 0.19, "learning_rate": 1.8699682567242865e-05, "loss": 1.0433, "step": 2843 }, { "epoch": 0.19, "learning_rate": 1.869861934196275e-05, "loss": 1.0641, "step": 2844 }, { "epoch": 0.19, "learning_rate": 1.869755571243067e-05, "loss": 1.0699, "step": 2845 }, { "epoch": 0.19, "learning_rate": 1.8696491678696054e-05, "loss": 1.0481, "step": 2846 }, { "epoch": 0.19, "learning_rate": 1.869542724080835e-05, "loss": 1.0685, "step": 2847 }, { "epoch": 0.19, "learning_rate": 1.869436239881703e-05, "loss": 1.0565, "step": 2848 }, { "epoch": 0.19, "learning_rate": 1.8693297152771575e-05, "loss": 1.039, "step": 2849 }, { "epoch": 0.19, "learning_rate": 1.86922315027215e-05, "loss": 1.0503, "step": 2850 }, { "epoch": 0.19, "learning_rate": 1.8691165448716316e-05, "loss": 1.0896, "step": 2851 }, { "epoch": 0.19, "learning_rate": 1.8690098990805576e-05, "loss": 1.0569, "step": 2852 }, { "epoch": 0.19, "learning_rate": 1.8689032129038833e-05, "loss": 1.014, "step": 2853 }, { "epoch": 0.19, "learning_rate": 1.8687964863465674e-05, "loss": 1.0339, "step": 2854 }, { "epoch": 0.19, "learning_rate": 1.8686897194135697e-05, "loss": 1.0254, "step": 2855 }, { "epoch": 0.19, "learning_rate": 1.868582912109852e-05, "loss": 1.0112, "step": 2856 }, { "epoch": 0.19, "learning_rate": 1.868476064440378e-05, "loss": 1.0575, "step": 2857 }, { "epoch": 0.19, "learning_rate": 1.8683691764101128e-05, "loss": 1.0585, "step": 2858 }, { "epoch": 0.19, "learning_rate": 1.8682622480240247e-05, "loss": 1.0993, "step": 2859 }, { "epoch": 0.19, "learning_rate": 1.8681552792870817e-05, "loss": 1.0176, "step": 2860 }, { "epoch": 0.19, "learning_rate": 1.8680482702042563e-05, "loss": 1.0247, "step": 2861 }, { "epoch": 0.19, "learning_rate": 1.8679412207805207e-05, "loss": 1.023, "step": 2862 }, { "epoch": 0.19, "learning_rate": 1.8678341310208506e-05, "loss": 1.061, "step": 2863 }, { "epoch": 0.19, "learning_rate": 1.8677270009302217e-05, "loss": 0.9976, "step": 2864 }, { "epoch": 0.19, "learning_rate": 1.8676198305136135e-05, "loss": 1.0136, "step": 2865 }, { "epoch": 0.19, "learning_rate": 1.8675126197760063e-05, "loss": 1.0235, "step": 2866 }, { "epoch": 0.19, "learning_rate": 1.8674053687223827e-05, "loss": 1.07, "step": 2867 }, { "epoch": 0.19, "learning_rate": 1.8672980773577265e-05, "loss": 1.0045, "step": 2868 }, { "epoch": 0.19, "learning_rate": 1.867190745687024e-05, "loss": 1.0133, "step": 2869 }, { "epoch": 0.19, "learning_rate": 1.867083373715264e-05, "loss": 1.0181, "step": 2870 }, { "epoch": 0.19, "learning_rate": 1.8669759614474356e-05, "loss": 1.037, "step": 2871 }, { "epoch": 0.19, "learning_rate": 1.866868508888531e-05, "loss": 1.0234, "step": 2872 }, { "epoch": 0.19, "learning_rate": 1.866761016043543e-05, "loss": 1.0352, "step": 2873 }, { "epoch": 0.19, "learning_rate": 1.8666534829174682e-05, "loss": 1.0594, "step": 2874 }, { "epoch": 0.19, "learning_rate": 1.8665459095153036e-05, "loss": 1.0208, "step": 2875 }, { "epoch": 0.19, "learning_rate": 1.8664382958420485e-05, "loss": 1.0344, "step": 2876 }, { "epoch": 0.19, "learning_rate": 1.8663306419027037e-05, "loss": 1.0131, "step": 2877 }, { "epoch": 0.19, "learning_rate": 1.866222947702273e-05, "loss": 0.9979, "step": 2878 }, { "epoch": 0.19, "learning_rate": 1.8661152132457605e-05, "loss": 1.0507, "step": 2879 }, { "epoch": 0.19, "learning_rate": 1.8660074385381732e-05, "loss": 1.0447, "step": 2880 }, { "epoch": 0.19, "learning_rate": 1.86589962358452e-05, "loss": 1.0517, "step": 2881 }, { "epoch": 0.19, "learning_rate": 1.865791768389811e-05, "loss": 1.0418, "step": 2882 }, { "epoch": 0.19, "learning_rate": 1.865683872959059e-05, "loss": 1.0042, "step": 2883 }, { "epoch": 0.19, "learning_rate": 1.865575937297278e-05, "loss": 1.0462, "step": 2884 }, { "epoch": 0.19, "learning_rate": 1.8654679614094837e-05, "loss": 1.0592, "step": 2885 }, { "epoch": 0.19, "learning_rate": 1.8653599453006947e-05, "loss": 1.0281, "step": 2886 }, { "epoch": 0.19, "learning_rate": 1.8652518889759306e-05, "loss": 1.0422, "step": 2887 }, { "epoch": 0.19, "learning_rate": 1.865143792440213e-05, "loss": 1.0294, "step": 2888 }, { "epoch": 0.19, "learning_rate": 1.8650356556985657e-05, "loss": 1.0224, "step": 2889 }, { "epoch": 0.19, "learning_rate": 1.864927478756014e-05, "loss": 1.0394, "step": 2890 }, { "epoch": 0.19, "learning_rate": 1.864819261617586e-05, "loss": 1.0514, "step": 2891 }, { "epoch": 0.19, "learning_rate": 1.8647110042883093e-05, "loss": 1.0203, "step": 2892 }, { "epoch": 0.19, "learning_rate": 1.864602706773216e-05, "loss": 1.0237, "step": 2893 }, { "epoch": 0.19, "learning_rate": 1.8644943690773393e-05, "loss": 1.0584, "step": 2894 }, { "epoch": 0.19, "learning_rate": 1.864385991205713e-05, "loss": 1.0098, "step": 2895 }, { "epoch": 0.19, "learning_rate": 1.8642775731633743e-05, "loss": 0.9994, "step": 2896 }, { "epoch": 0.19, "learning_rate": 1.864169114955362e-05, "loss": 1.0517, "step": 2897 }, { "epoch": 0.19, "learning_rate": 1.864060616586716e-05, "loss": 1.0729, "step": 2898 }, { "epoch": 0.19, "learning_rate": 1.863952078062479e-05, "loss": 1.0562, "step": 2899 }, { "epoch": 0.19, "learning_rate": 1.863843499387695e-05, "loss": 1.0315, "step": 2900 }, { "epoch": 0.19, "learning_rate": 1.8637348805674097e-05, "loss": 1.0364, "step": 2901 }, { "epoch": 0.19, "learning_rate": 1.863626221606671e-05, "loss": 1.0477, "step": 2902 }, { "epoch": 0.19, "learning_rate": 1.863517522510529e-05, "loss": 1.0426, "step": 2903 }, { "epoch": 0.19, "learning_rate": 1.8634087832840347e-05, "loss": 1.0019, "step": 2904 }, { "epoch": 0.19, "learning_rate": 1.8633000039322423e-05, "loss": 1.0298, "step": 2905 }, { "epoch": 0.19, "learning_rate": 1.8631911844602065e-05, "loss": 1.0257, "step": 2906 }, { "epoch": 0.19, "learning_rate": 1.8630823248729847e-05, "loss": 1.0153, "step": 2907 }, { "epoch": 0.19, "learning_rate": 1.862973425175636e-05, "loss": 0.983, "step": 2908 }, { "epoch": 0.19, "learning_rate": 1.8628644853732212e-05, "loss": 1.0563, "step": 2909 }, { "epoch": 0.19, "learning_rate": 1.8627555054708033e-05, "loss": 1.0088, "step": 2910 }, { "epoch": 0.19, "learning_rate": 1.8626464854734466e-05, "loss": 1.0438, "step": 2911 }, { "epoch": 0.19, "learning_rate": 1.8625374253862176e-05, "loss": 1.0284, "step": 2912 }, { "epoch": 0.19, "learning_rate": 1.862428325214185e-05, "loss": 1.0738, "step": 2913 }, { "epoch": 0.19, "learning_rate": 1.862319184962419e-05, "loss": 1.0235, "step": 2914 }, { "epoch": 0.19, "learning_rate": 1.8622100046359915e-05, "loss": 1.0559, "step": 2915 }, { "epoch": 0.19, "learning_rate": 1.862100784239976e-05, "loss": 1.0201, "step": 2916 }, { "epoch": 0.19, "learning_rate": 1.8619915237794494e-05, "loss": 1.0894, "step": 2917 }, { "epoch": 0.19, "learning_rate": 1.8618822232594884e-05, "loss": 0.9918, "step": 2918 }, { "epoch": 0.19, "learning_rate": 1.861772882685173e-05, "loss": 1.0219, "step": 2919 }, { "epoch": 0.19, "learning_rate": 1.8616635020615846e-05, "loss": 1.0044, "step": 2920 }, { "epoch": 0.19, "learning_rate": 1.8615540813938063e-05, "loss": 1.0436, "step": 2921 }, { "epoch": 0.19, "learning_rate": 1.8614446206869232e-05, "loss": 1.0788, "step": 2922 }, { "epoch": 0.19, "learning_rate": 1.8613351199460226e-05, "loss": 1.0411, "step": 2923 }, { "epoch": 0.19, "learning_rate": 1.8612255791761925e-05, "loss": 1.039, "step": 2924 }, { "epoch": 0.19, "learning_rate": 1.8611159983825248e-05, "loss": 0.9934, "step": 2925 }, { "epoch": 0.19, "learning_rate": 1.8610063775701113e-05, "loss": 1.0432, "step": 2926 }, { "epoch": 0.19, "learning_rate": 1.8608967167440467e-05, "loss": 1.0257, "step": 2927 }, { "epoch": 0.19, "learning_rate": 1.860787015909427e-05, "loss": 1.0311, "step": 2928 }, { "epoch": 0.19, "learning_rate": 1.8606772750713503e-05, "loss": 1.0453, "step": 2929 }, { "epoch": 0.2, "learning_rate": 1.860567494234917e-05, "loss": 0.9807, "step": 2930 }, { "epoch": 0.2, "learning_rate": 1.860457673405229e-05, "loss": 1.075, "step": 2931 }, { "epoch": 0.2, "learning_rate": 1.8603478125873896e-05, "loss": 1.013, "step": 2932 }, { "epoch": 0.2, "learning_rate": 1.8602379117865044e-05, "loss": 1.066, "step": 2933 }, { "epoch": 0.2, "learning_rate": 1.860127971007681e-05, "loss": 1.0499, "step": 2934 }, { "epoch": 0.2, "learning_rate": 1.8600179902560288e-05, "loss": 1.0543, "step": 2935 }, { "epoch": 0.2, "learning_rate": 1.859907969536659e-05, "loss": 1.056, "step": 2936 }, { "epoch": 0.2, "learning_rate": 1.8597979088546842e-05, "loss": 1.0286, "step": 2937 }, { "epoch": 0.2, "learning_rate": 1.8596878082152195e-05, "loss": 1.0148, "step": 2938 }, { "epoch": 0.2, "learning_rate": 1.859577667623382e-05, "loss": 1.0021, "step": 2939 }, { "epoch": 0.2, "learning_rate": 1.8594674870842893e-05, "loss": 1.0196, "step": 2940 }, { "epoch": 0.2, "learning_rate": 1.8593572666030627e-05, "loss": 1.0934, "step": 2941 }, { "epoch": 0.2, "learning_rate": 1.859247006184824e-05, "loss": 1.0209, "step": 2942 }, { "epoch": 0.2, "learning_rate": 1.859136705834698e-05, "loss": 1.0162, "step": 2943 }, { "epoch": 0.2, "learning_rate": 1.85902636555781e-05, "loss": 1.0282, "step": 2944 }, { "epoch": 0.2, "learning_rate": 1.858915985359288e-05, "loss": 1.0406, "step": 2945 }, { "epoch": 0.2, "learning_rate": 1.858805565244262e-05, "loss": 1.0072, "step": 2946 }, { "epoch": 0.2, "learning_rate": 1.858695105217863e-05, "loss": 1.0746, "step": 2947 }, { "epoch": 0.2, "learning_rate": 1.8585846052852254e-05, "loss": 1.0185, "step": 2948 }, { "epoch": 0.2, "learning_rate": 1.8584740654514838e-05, "loss": 1.0253, "step": 2949 }, { "epoch": 0.2, "learning_rate": 1.8583634857217752e-05, "loss": 0.9846, "step": 2950 }, { "epoch": 0.2, "learning_rate": 1.8582528661012386e-05, "loss": 1.0609, "step": 2951 }, { "epoch": 0.2, "learning_rate": 1.8581422065950153e-05, "loss": 1.0627, "step": 2952 }, { "epoch": 0.2, "learning_rate": 1.858031507208248e-05, "loss": 1.0258, "step": 2953 }, { "epoch": 0.2, "learning_rate": 1.8579207679460807e-05, "loss": 1.0257, "step": 2954 }, { "epoch": 0.2, "learning_rate": 1.85780998881366e-05, "loss": 1.0488, "step": 2955 }, { "epoch": 0.2, "learning_rate": 1.8576991698161344e-05, "loss": 1.011, "step": 2956 }, { "epoch": 0.2, "learning_rate": 1.8575883109586537e-05, "loss": 0.9963, "step": 2957 }, { "epoch": 0.2, "learning_rate": 1.8574774122463704e-05, "loss": 1.0132, "step": 2958 }, { "epoch": 0.2, "learning_rate": 1.8573664736844376e-05, "loss": 1.0618, "step": 2959 }, { "epoch": 0.2, "learning_rate": 1.857255495278011e-05, "loss": 1.0559, "step": 2960 }, { "epoch": 0.2, "learning_rate": 1.857144477032249e-05, "loss": 1.0458, "step": 2961 }, { "epoch": 0.2, "learning_rate": 1.8570334189523097e-05, "loss": 1.0255, "step": 2962 }, { "epoch": 0.2, "learning_rate": 1.8569223210433557e-05, "loss": 1.0691, "step": 2963 }, { "epoch": 0.2, "learning_rate": 1.8568111833105485e-05, "loss": 1.0161, "step": 2964 }, { "epoch": 0.2, "learning_rate": 1.8567000057590546e-05, "loss": 1.0459, "step": 2965 }, { "epoch": 0.2, "learning_rate": 1.8565887883940394e-05, "loss": 1.0047, "step": 2966 }, { "epoch": 0.2, "learning_rate": 1.8564775312206726e-05, "loss": 1.027, "step": 2967 }, { "epoch": 0.2, "learning_rate": 1.8563662342441237e-05, "loss": 1.0864, "step": 2968 }, { "epoch": 0.2, "learning_rate": 1.856254897469566e-05, "loss": 1.0257, "step": 2969 }, { "epoch": 0.2, "learning_rate": 1.856143520902173e-05, "loss": 1.0158, "step": 2970 }, { "epoch": 0.2, "learning_rate": 1.856032104547121e-05, "loss": 1.013, "step": 2971 }, { "epoch": 0.2, "learning_rate": 1.8559206484095877e-05, "loss": 1.0261, "step": 2972 }, { "epoch": 0.2, "learning_rate": 1.8558091524947527e-05, "loss": 1.0296, "step": 2973 }, { "epoch": 0.2, "learning_rate": 1.8556976168077977e-05, "loss": 1.0455, "step": 2974 }, { "epoch": 0.2, "learning_rate": 1.8555860413539063e-05, "loss": 1.0627, "step": 2975 }, { "epoch": 0.2, "learning_rate": 1.8554744261382635e-05, "loss": 1.0433, "step": 2976 }, { "epoch": 0.2, "learning_rate": 1.855362771166057e-05, "loss": 1.0409, "step": 2977 }, { "epoch": 0.2, "learning_rate": 1.8552510764424748e-05, "loss": 1.0199, "step": 2978 }, { "epoch": 0.2, "learning_rate": 1.8551393419727083e-05, "loss": 1.0322, "step": 2979 }, { "epoch": 0.2, "learning_rate": 1.85502756776195e-05, "loss": 1.0415, "step": 2980 }, { "epoch": 0.2, "learning_rate": 1.8549157538153944e-05, "loss": 1.0037, "step": 2981 }, { "epoch": 0.2, "learning_rate": 1.8548039001382377e-05, "loss": 1.0178, "step": 2982 }, { "epoch": 0.2, "learning_rate": 1.8546920067356783e-05, "loss": 1.0157, "step": 2983 }, { "epoch": 0.2, "learning_rate": 1.8545800736129166e-05, "loss": 1.048, "step": 2984 }, { "epoch": 0.2, "learning_rate": 1.8544681007751538e-05, "loss": 1.0445, "step": 2985 }, { "epoch": 0.2, "learning_rate": 1.854356088227594e-05, "loss": 1.0066, "step": 2986 }, { "epoch": 0.2, "learning_rate": 1.8542440359754422e-05, "loss": 1.0053, "step": 2987 }, { "epoch": 0.2, "learning_rate": 1.8541319440239066e-05, "loss": 1.0626, "step": 2988 }, { "epoch": 0.2, "learning_rate": 1.8540198123781966e-05, "loss": 1.0162, "step": 2989 }, { "epoch": 0.2, "learning_rate": 1.8539076410435223e-05, "loss": 1.0566, "step": 2990 }, { "epoch": 0.2, "learning_rate": 1.8537954300250973e-05, "loss": 1.0622, "step": 2991 }, { "epoch": 0.2, "learning_rate": 1.853683179328136e-05, "loss": 1.0259, "step": 2992 }, { "epoch": 0.2, "learning_rate": 1.853570888957856e-05, "loss": 1.0298, "step": 2993 }, { "epoch": 0.2, "learning_rate": 1.8534585589194745e-05, "loss": 0.9848, "step": 2994 }, { "epoch": 0.2, "learning_rate": 1.853346189218213e-05, "loss": 1.0319, "step": 2995 }, { "epoch": 0.2, "learning_rate": 1.853233779859293e-05, "loss": 1.0648, "step": 2996 }, { "epoch": 0.2, "learning_rate": 1.8531213308479383e-05, "loss": 1.0361, "step": 2997 }, { "epoch": 0.2, "learning_rate": 1.8530088421893756e-05, "loss": 1.0799, "step": 2998 }, { "epoch": 0.2, "learning_rate": 1.8528963138888315e-05, "loss": 1.0434, "step": 2999 }, { "epoch": 0.2, "learning_rate": 1.8527837459515364e-05, "loss": 1.0253, "step": 3000 }, { "epoch": 0.2, "learning_rate": 1.8526711383827217e-05, "loss": 1.0178, "step": 3001 }, { "epoch": 0.2, "learning_rate": 1.85255849118762e-05, "loss": 0.9928, "step": 3002 }, { "epoch": 0.2, "learning_rate": 1.852445804371467e-05, "loss": 1.0679, "step": 3003 }, { "epoch": 0.2, "learning_rate": 1.852333077939499e-05, "loss": 1.0522, "step": 3004 }, { "epoch": 0.2, "learning_rate": 1.852220311896955e-05, "loss": 1.0372, "step": 3005 }, { "epoch": 0.2, "learning_rate": 1.852107506249076e-05, "loss": 0.9995, "step": 3006 }, { "epoch": 0.2, "learning_rate": 1.851994661001104e-05, "loss": 1.0461, "step": 3007 }, { "epoch": 0.2, "learning_rate": 1.8518817761582833e-05, "loss": 1.0293, "step": 3008 }, { "epoch": 0.2, "learning_rate": 1.8517688517258597e-05, "loss": 0.9978, "step": 3009 }, { "epoch": 0.2, "learning_rate": 1.851655887709082e-05, "loss": 1.0121, "step": 3010 }, { "epoch": 0.2, "learning_rate": 1.8515428841131996e-05, "loss": 1.0626, "step": 3011 }, { "epoch": 0.2, "learning_rate": 1.8514298409434633e-05, "loss": 0.9951, "step": 3012 }, { "epoch": 0.2, "learning_rate": 1.851316758205128e-05, "loss": 1.0985, "step": 3013 }, { "epoch": 0.2, "learning_rate": 1.851203635903448e-05, "loss": 1.0199, "step": 3014 }, { "epoch": 0.2, "learning_rate": 1.851090474043681e-05, "loss": 1.0626, "step": 3015 }, { "epoch": 0.2, "learning_rate": 1.8509772726310857e-05, "loss": 1.035, "step": 3016 }, { "epoch": 0.2, "learning_rate": 1.850864031670923e-05, "loss": 1.0538, "step": 3017 }, { "epoch": 0.2, "learning_rate": 1.850750751168455e-05, "loss": 1.0298, "step": 3018 }, { "epoch": 0.2, "learning_rate": 1.8506374311289472e-05, "loss": 1.0492, "step": 3019 }, { "epoch": 0.2, "learning_rate": 1.8505240715576656e-05, "loss": 1.0077, "step": 3020 }, { "epoch": 0.2, "learning_rate": 1.8504106724598778e-05, "loss": 1.0118, "step": 3021 }, { "epoch": 0.2, "learning_rate": 1.8502972338408547e-05, "loss": 1.0181, "step": 3022 }, { "epoch": 0.2, "learning_rate": 1.8501837557058673e-05, "loss": 1.1111, "step": 3023 }, { "epoch": 0.2, "learning_rate": 1.8500702380601898e-05, "loss": 1.017, "step": 3024 }, { "epoch": 0.2, "learning_rate": 1.8499566809090976e-05, "loss": 1.0426, "step": 3025 }, { "epoch": 0.2, "learning_rate": 1.849843084257868e-05, "loss": 1.0093, "step": 3026 }, { "epoch": 0.2, "learning_rate": 1.84972944811178e-05, "loss": 1.0431, "step": 3027 }, { "epoch": 0.2, "learning_rate": 1.8496157724761156e-05, "loss": 1.0192, "step": 3028 }, { "epoch": 0.2, "learning_rate": 1.8495020573561563e-05, "loss": 0.9915, "step": 3029 }, { "epoch": 0.2, "learning_rate": 1.8493883027571877e-05, "loss": 1.0575, "step": 3030 }, { "epoch": 0.2, "learning_rate": 1.849274508684496e-05, "loss": 1.0507, "step": 3031 }, { "epoch": 0.2, "learning_rate": 1.8491606751433692e-05, "loss": 1.0593, "step": 3032 }, { "epoch": 0.2, "learning_rate": 1.8490468021390985e-05, "loss": 1.0341, "step": 3033 }, { "epoch": 0.2, "learning_rate": 1.848932889676975e-05, "loss": 1.0094, "step": 3034 }, { "epoch": 0.2, "learning_rate": 1.8488189377622928e-05, "loss": 1.0719, "step": 3035 }, { "epoch": 0.2, "learning_rate": 1.848704946400348e-05, "loss": 1.0254, "step": 3036 }, { "epoch": 0.2, "learning_rate": 1.848590915596438e-05, "loss": 1.0716, "step": 3037 }, { "epoch": 0.2, "learning_rate": 1.8484768453558617e-05, "loss": 1.0593, "step": 3038 }, { "epoch": 0.2, "learning_rate": 1.8483627356839206e-05, "loss": 1.0869, "step": 3039 }, { "epoch": 0.2, "learning_rate": 1.848248586585918e-05, "loss": 0.9897, "step": 3040 }, { "epoch": 0.2, "learning_rate": 1.8481343980671585e-05, "loss": 1.0425, "step": 3041 }, { "epoch": 0.2, "learning_rate": 1.8480201701329483e-05, "loss": 1.022, "step": 3042 }, { "epoch": 0.2, "learning_rate": 1.847905902788597e-05, "loss": 1.0673, "step": 3043 }, { "epoch": 0.2, "learning_rate": 1.8477915960394144e-05, "loss": 1.0218, "step": 3044 }, { "epoch": 0.2, "learning_rate": 1.847677249890712e-05, "loss": 1.0219, "step": 3045 }, { "epoch": 0.2, "learning_rate": 1.8475628643478055e-05, "loss": 1.0143, "step": 3046 }, { "epoch": 0.2, "learning_rate": 1.847448439416009e-05, "loss": 1.0333, "step": 3047 }, { "epoch": 0.2, "learning_rate": 1.8473339751006413e-05, "loss": 1.04, "step": 3048 }, { "epoch": 0.2, "learning_rate": 1.8472194714070214e-05, "loss": 1.0313, "step": 3049 }, { "epoch": 0.2, "learning_rate": 1.8471049283404712e-05, "loss": 1.0659, "step": 3050 }, { "epoch": 0.2, "learning_rate": 1.846990345906313e-05, "loss": 1.0516, "step": 3051 }, { "epoch": 0.2, "learning_rate": 1.8468757241098724e-05, "loss": 1.0172, "step": 3052 }, { "epoch": 0.2, "learning_rate": 1.8467610629564763e-05, "loss": 1.0312, "step": 3053 }, { "epoch": 0.2, "learning_rate": 1.846646362451453e-05, "loss": 1.0647, "step": 3054 }, { "epoch": 0.2, "learning_rate": 1.846531622600133e-05, "loss": 1.0468, "step": 3055 }, { "epoch": 0.2, "learning_rate": 1.846416843407849e-05, "loss": 1.0292, "step": 3056 }, { "epoch": 0.2, "learning_rate": 1.8463020248799354e-05, "loss": 1.0396, "step": 3057 }, { "epoch": 0.2, "learning_rate": 1.846187167021727e-05, "loss": 1.0279, "step": 3058 }, { "epoch": 0.2, "learning_rate": 1.8460722698385625e-05, "loss": 1.0158, "step": 3059 }, { "epoch": 0.2, "learning_rate": 1.8459573333357814e-05, "loss": 1.0347, "step": 3060 }, { "epoch": 0.2, "learning_rate": 1.845842357518725e-05, "loss": 1.0868, "step": 3061 }, { "epoch": 0.2, "learning_rate": 1.8457273423927366e-05, "loss": 1.0198, "step": 3062 }, { "epoch": 0.2, "learning_rate": 1.8456122879631618e-05, "loss": 1.0299, "step": 3063 }, { "epoch": 0.2, "learning_rate": 1.8454971942353467e-05, "loss": 1.0496, "step": 3064 }, { "epoch": 0.2, "learning_rate": 1.845382061214641e-05, "loss": 1.0326, "step": 3065 }, { "epoch": 0.2, "learning_rate": 1.845266888906394e-05, "loss": 1.0327, "step": 3066 }, { "epoch": 0.2, "learning_rate": 1.8451516773159597e-05, "loss": 0.9991, "step": 3067 }, { "epoch": 0.2, "learning_rate": 1.8450364264486912e-05, "loss": 1.0293, "step": 3068 }, { "epoch": 0.2, "learning_rate": 1.844921136309945e-05, "loss": 1.065, "step": 3069 }, { "epoch": 0.2, "learning_rate": 1.844805806905079e-05, "loss": 1.0164, "step": 3070 }, { "epoch": 0.2, "learning_rate": 1.8446904382394524e-05, "loss": 1.0409, "step": 3071 }, { "epoch": 0.2, "learning_rate": 1.8445750303184275e-05, "loss": 1.0199, "step": 3072 }, { "epoch": 0.2, "learning_rate": 1.844459583147367e-05, "loss": 1.0687, "step": 3073 }, { "epoch": 0.2, "learning_rate": 1.8443440967316365e-05, "loss": 0.9792, "step": 3074 }, { "epoch": 0.2, "learning_rate": 1.844228571076603e-05, "loss": 1.0455, "step": 3075 }, { "epoch": 0.2, "learning_rate": 1.8441130061876357e-05, "loss": 1.0184, "step": 3076 }, { "epoch": 0.2, "learning_rate": 1.8439974020701043e-05, "loss": 1.0289, "step": 3077 }, { "epoch": 0.2, "learning_rate": 1.843881758729382e-05, "loss": 1.0462, "step": 3078 }, { "epoch": 0.2, "learning_rate": 1.843766076170843e-05, "loss": 1.0206, "step": 3079 }, { "epoch": 0.21, "learning_rate": 1.843650354399863e-05, "loss": 0.9929, "step": 3080 }, { "epoch": 0.21, "learning_rate": 1.8435345934218207e-05, "loss": 1.0543, "step": 3081 }, { "epoch": 0.21, "learning_rate": 1.8434187932420955e-05, "loss": 1.0052, "step": 3082 }, { "epoch": 0.21, "learning_rate": 1.843302953866069e-05, "loss": 1.0019, "step": 3083 }, { "epoch": 0.21, "learning_rate": 1.843187075299124e-05, "loss": 1.0516, "step": 3084 }, { "epoch": 0.21, "learning_rate": 1.843071157546647e-05, "loss": 1.0352, "step": 3085 }, { "epoch": 0.21, "learning_rate": 1.8429552006140243e-05, "loss": 1.021, "step": 3086 }, { "epoch": 0.21, "learning_rate": 1.8428392045066445e-05, "loss": 0.9928, "step": 3087 }, { "epoch": 0.21, "learning_rate": 1.8427231692298985e-05, "loss": 1.0042, "step": 3088 }, { "epoch": 0.21, "learning_rate": 1.8426070947891797e-05, "loss": 1.004, "step": 3089 }, { "epoch": 0.21, "learning_rate": 1.8424909811898812e-05, "loss": 1.0254, "step": 3090 }, { "epoch": 0.21, "learning_rate": 1.8423748284373998e-05, "loss": 0.9893, "step": 3091 }, { "epoch": 0.21, "learning_rate": 1.8422586365371335e-05, "loss": 1.0471, "step": 3092 }, { "epoch": 0.21, "learning_rate": 1.842142405494482e-05, "loss": 1.0257, "step": 3093 }, { "epoch": 0.21, "learning_rate": 1.8420261353148467e-05, "loss": 0.9686, "step": 3094 }, { "epoch": 0.21, "learning_rate": 1.8419098260036312e-05, "loss": 1.0452, "step": 3095 }, { "epoch": 0.21, "learning_rate": 1.841793477566241e-05, "loss": 0.9912, "step": 3096 }, { "epoch": 0.21, "learning_rate": 1.8416770900080826e-05, "loss": 1.0307, "step": 3097 }, { "epoch": 0.21, "learning_rate": 1.8415606633345657e-05, "loss": 1.0292, "step": 3098 }, { "epoch": 0.21, "learning_rate": 1.8414441975511003e-05, "loss": 0.9919, "step": 3099 }, { "epoch": 0.21, "learning_rate": 1.8413276926630995e-05, "loss": 0.9941, "step": 3100 }, { "epoch": 0.21, "learning_rate": 1.8412111486759768e-05, "loss": 1.0485, "step": 3101 }, { "epoch": 0.21, "learning_rate": 1.8410945655951493e-05, "loss": 1.0252, "step": 3102 }, { "epoch": 0.21, "learning_rate": 1.8409779434260345e-05, "loss": 1.0276, "step": 3103 }, { "epoch": 0.21, "learning_rate": 1.840861282174052e-05, "loss": 1.0766, "step": 3104 }, { "epoch": 0.21, "learning_rate": 1.840744581844624e-05, "loss": 1.0667, "step": 3105 }, { "epoch": 0.21, "learning_rate": 1.8406278424431737e-05, "loss": 1.0438, "step": 3106 }, { "epoch": 0.21, "learning_rate": 1.840511063975126e-05, "loss": 1.0598, "step": 3107 }, { "epoch": 0.21, "learning_rate": 1.8403942464459085e-05, "loss": 1.0168, "step": 3108 }, { "epoch": 0.21, "learning_rate": 1.8402773898609494e-05, "loss": 1.0712, "step": 3109 }, { "epoch": 0.21, "learning_rate": 1.8401604942256802e-05, "loss": 1.0228, "step": 3110 }, { "epoch": 0.21, "learning_rate": 1.8400435595455326e-05, "loss": 1.0461, "step": 3111 }, { "epoch": 0.21, "learning_rate": 1.8399265858259413e-05, "loss": 1.0278, "step": 3112 }, { "epoch": 0.21, "learning_rate": 1.839809573072343e-05, "loss": 1.0212, "step": 3113 }, { "epoch": 0.21, "learning_rate": 1.8396925212901747e-05, "loss": 1.0275, "step": 3114 }, { "epoch": 0.21, "learning_rate": 1.8395754304848762e-05, "loss": 1.0456, "step": 3115 }, { "epoch": 0.21, "learning_rate": 1.8394583006618897e-05, "loss": 1.0271, "step": 3116 }, { "epoch": 0.21, "learning_rate": 1.8393411318266588e-05, "loss": 1.0638, "step": 3117 }, { "epoch": 0.21, "learning_rate": 1.8392239239846276e-05, "loss": 1.0442, "step": 3118 }, { "epoch": 0.21, "learning_rate": 1.839106677141244e-05, "loss": 1.0718, "step": 3119 }, { "epoch": 0.21, "learning_rate": 1.8389893913019564e-05, "loss": 1.0648, "step": 3120 }, { "epoch": 0.21, "learning_rate": 1.8388720664722157e-05, "loss": 1.0144, "step": 3121 }, { "epoch": 0.21, "learning_rate": 1.8387547026574742e-05, "loss": 1.0254, "step": 3122 }, { "epoch": 0.21, "learning_rate": 1.8386372998631862e-05, "loss": 1.0536, "step": 3123 }, { "epoch": 0.21, "learning_rate": 1.8385198580948076e-05, "loss": 1.0193, "step": 3124 }, { "epoch": 0.21, "learning_rate": 1.8384023773577967e-05, "loss": 1.0582, "step": 3125 }, { "epoch": 0.21, "learning_rate": 1.8382848576576128e-05, "loss": 1.085, "step": 3126 }, { "epoch": 0.21, "learning_rate": 1.8381672989997178e-05, "loss": 1.0496, "step": 3127 }, { "epoch": 0.21, "learning_rate": 1.8380497013895748e-05, "loss": 1.0441, "step": 3128 }, { "epoch": 0.21, "learning_rate": 1.8379320648326485e-05, "loss": 1.0406, "step": 3129 }, { "epoch": 0.21, "learning_rate": 1.8378143893344067e-05, "loss": 1.0414, "step": 3130 }, { "epoch": 0.21, "learning_rate": 1.8376966749003172e-05, "loss": 1.0367, "step": 3131 }, { "epoch": 0.21, "learning_rate": 1.8375789215358518e-05, "loss": 1.005, "step": 3132 }, { "epoch": 0.21, "learning_rate": 1.8374611292464815e-05, "loss": 1.0282, "step": 3133 }, { "epoch": 0.21, "learning_rate": 1.8373432980376814e-05, "loss": 1.0244, "step": 3134 }, { "epoch": 0.21, "learning_rate": 1.837225427914927e-05, "loss": 1.0456, "step": 3135 }, { "epoch": 0.21, "learning_rate": 1.8371075188836962e-05, "loss": 0.9994, "step": 3136 }, { "epoch": 0.21, "learning_rate": 1.8369895709494686e-05, "loss": 1.0072, "step": 3137 }, { "epoch": 0.21, "learning_rate": 1.8368715841177263e-05, "loss": 1.0204, "step": 3138 }, { "epoch": 0.21, "learning_rate": 1.8367535583939512e-05, "loss": 1.0266, "step": 3139 }, { "epoch": 0.21, "learning_rate": 1.8366354937836294e-05, "loss": 1.0229, "step": 3140 }, { "epoch": 0.21, "learning_rate": 1.8365173902922473e-05, "loss": 1.0587, "step": 3141 }, { "epoch": 0.21, "learning_rate": 1.8363992479252936e-05, "loss": 1.0545, "step": 3142 }, { "epoch": 0.21, "learning_rate": 1.8362810666882586e-05, "loss": 1.0354, "step": 3143 }, { "epoch": 0.21, "learning_rate": 1.8361628465866346e-05, "loss": 0.9868, "step": 3144 }, { "epoch": 0.21, "learning_rate": 1.836044587625916e-05, "loss": 1.0423, "step": 3145 }, { "epoch": 0.21, "learning_rate": 1.835926289811598e-05, "loss": 1.0396, "step": 3146 }, { "epoch": 0.21, "learning_rate": 1.8358079531491794e-05, "loss": 0.9901, "step": 3147 }, { "epoch": 0.21, "learning_rate": 1.8356895776441584e-05, "loss": 1.0327, "step": 3148 }, { "epoch": 0.21, "learning_rate": 1.8355711633020367e-05, "loss": 1.0172, "step": 3149 }, { "epoch": 0.21, "learning_rate": 1.8354527101283178e-05, "loss": 1.0171, "step": 3150 }, { "epoch": 0.21, "learning_rate": 1.8353342181285062e-05, "loss": 1.0026, "step": 3151 }, { "epoch": 0.21, "learning_rate": 1.8352156873081087e-05, "loss": 1.031, "step": 3152 }, { "epoch": 0.21, "learning_rate": 1.8350971176726338e-05, "loss": 1.0195, "step": 3153 }, { "epoch": 0.21, "learning_rate": 1.8349785092275916e-05, "loss": 1.0396, "step": 3154 }, { "epoch": 0.21, "learning_rate": 1.8348598619784948e-05, "loss": 1.0717, "step": 3155 }, { "epoch": 0.21, "learning_rate": 1.8347411759308564e-05, "loss": 1.0327, "step": 3156 }, { "epoch": 0.21, "learning_rate": 1.8346224510901928e-05, "loss": 1.0642, "step": 3157 }, { "epoch": 0.21, "learning_rate": 1.8345036874620214e-05, "loss": 0.9931, "step": 3158 }, { "epoch": 0.21, "learning_rate": 1.8343848850518613e-05, "loss": 1.046, "step": 3159 }, { "epoch": 0.21, "learning_rate": 1.834266043865234e-05, "loss": 1.0251, "step": 3160 }, { "epoch": 0.21, "learning_rate": 1.834147163907662e-05, "loss": 1.0741, "step": 3161 }, { "epoch": 0.21, "learning_rate": 1.8340282451846702e-05, "loss": 0.9857, "step": 3162 }, { "epoch": 0.21, "learning_rate": 1.8339092877017853e-05, "loss": 1.0282, "step": 3163 }, { "epoch": 0.21, "learning_rate": 1.833790291464535e-05, "loss": 1.0105, "step": 3164 }, { "epoch": 0.21, "learning_rate": 1.8336712564784506e-05, "loss": 1.035, "step": 3165 }, { "epoch": 0.21, "learning_rate": 1.8335521827490628e-05, "loss": 1.0438, "step": 3166 }, { "epoch": 0.21, "learning_rate": 1.8334330702819056e-05, "loss": 1.0346, "step": 3167 }, { "epoch": 0.21, "learning_rate": 1.833313919082515e-05, "loss": 0.9697, "step": 3168 }, { "epoch": 0.21, "learning_rate": 1.8331947291564284e-05, "loss": 1.0188, "step": 3169 }, { "epoch": 0.21, "learning_rate": 1.833075500509184e-05, "loss": 1.012, "step": 3170 }, { "epoch": 0.21, "learning_rate": 1.832956233146324e-05, "loss": 1.0529, "step": 3171 }, { "epoch": 0.21, "learning_rate": 1.8328369270733903e-05, "loss": 1.0431, "step": 3172 }, { "epoch": 0.21, "learning_rate": 1.8327175822959273e-05, "loss": 1.0185, "step": 3173 }, { "epoch": 0.21, "learning_rate": 1.8325981988194815e-05, "loss": 1.0144, "step": 3174 }, { "epoch": 0.21, "learning_rate": 1.8324787766496013e-05, "loss": 1.0505, "step": 3175 }, { "epoch": 0.21, "learning_rate": 1.8323593157918366e-05, "loss": 1.0011, "step": 3176 }, { "epoch": 0.21, "learning_rate": 1.832239816251739e-05, "loss": 1.0417, "step": 3177 }, { "epoch": 0.21, "learning_rate": 1.832120278034862e-05, "loss": 0.9865, "step": 3178 }, { "epoch": 0.21, "learning_rate": 1.8320007011467606e-05, "loss": 1.059, "step": 3179 }, { "epoch": 0.21, "learning_rate": 1.8318810855929926e-05, "loss": 1.0263, "step": 3180 }, { "epoch": 0.21, "learning_rate": 1.8317614313791162e-05, "loss": 1.0774, "step": 3181 }, { "epoch": 0.21, "learning_rate": 1.8316417385106927e-05, "loss": 1.0468, "step": 3182 }, { "epoch": 0.21, "learning_rate": 1.831522006993284e-05, "loss": 1.0443, "step": 3183 }, { "epoch": 0.21, "learning_rate": 1.831402236832455e-05, "loss": 1.0117, "step": 3184 }, { "epoch": 0.21, "learning_rate": 1.8312824280337715e-05, "loss": 1.0195, "step": 3185 }, { "epoch": 0.21, "learning_rate": 1.8311625806028012e-05, "loss": 1.0635, "step": 3186 }, { "epoch": 0.21, "learning_rate": 1.8310426945451143e-05, "loss": 1.0222, "step": 3187 }, { "epoch": 0.21, "learning_rate": 1.8309227698662815e-05, "loss": 1.007, "step": 3188 }, { "epoch": 0.21, "learning_rate": 1.8308028065718774e-05, "loss": 1.0245, "step": 3189 }, { "epoch": 0.21, "learning_rate": 1.8306828046674753e-05, "loss": 1.0644, "step": 3190 }, { "epoch": 0.21, "learning_rate": 1.8305627641586533e-05, "loss": 1.0511, "step": 3191 }, { "epoch": 0.21, "learning_rate": 1.8304426850509897e-05, "loss": 1.0011, "step": 3192 }, { "epoch": 0.21, "learning_rate": 1.830322567350065e-05, "loss": 1.0678, "step": 3193 }, { "epoch": 0.21, "learning_rate": 1.8302024110614616e-05, "loss": 1.0355, "step": 3194 }, { "epoch": 0.21, "learning_rate": 1.8300822161907633e-05, "loss": 1.034, "step": 3195 }, { "epoch": 0.21, "learning_rate": 1.8299619827435558e-05, "loss": 0.9955, "step": 3196 }, { "epoch": 0.21, "learning_rate": 1.829841710725427e-05, "loss": 1.017, "step": 3197 }, { "epoch": 0.21, "learning_rate": 1.8297214001419664e-05, "loss": 1.0508, "step": 3198 }, { "epoch": 0.21, "learning_rate": 1.829601050998765e-05, "loss": 1.0647, "step": 3199 }, { "epoch": 0.21, "learning_rate": 1.8294806633014155e-05, "loss": 1.0205, "step": 3200 }, { "epoch": 0.21, "learning_rate": 1.8293602370555135e-05, "loss": 1.0408, "step": 3201 }, { "epoch": 0.21, "learning_rate": 1.829239772266655e-05, "loss": 1.0216, "step": 3202 }, { "epoch": 0.21, "learning_rate": 1.829119268940438e-05, "loss": 1.0466, "step": 3203 }, { "epoch": 0.21, "learning_rate": 1.8289987270824637e-05, "loss": 1.0141, "step": 3204 }, { "epoch": 0.21, "learning_rate": 1.8288781466983333e-05, "loss": 1.0458, "step": 3205 }, { "epoch": 0.21, "learning_rate": 1.828757527793651e-05, "loss": 1.0794, "step": 3206 }, { "epoch": 0.21, "learning_rate": 1.828636870374022e-05, "loss": 1.0306, "step": 3207 }, { "epoch": 0.21, "learning_rate": 1.8285161744450534e-05, "loss": 0.9793, "step": 3208 }, { "epoch": 0.21, "learning_rate": 1.828395440012355e-05, "loss": 1.0477, "step": 3209 }, { "epoch": 0.21, "learning_rate": 1.8282746670815372e-05, "loss": 1.053, "step": 3210 }, { "epoch": 0.21, "learning_rate": 1.828153855658213e-05, "loss": 1.0196, "step": 3211 }, { "epoch": 0.21, "learning_rate": 1.8280330057479965e-05, "loss": 0.9913, "step": 3212 }, { "epoch": 0.21, "learning_rate": 1.8279121173565043e-05, "loss": 1.0156, "step": 3213 }, { "epoch": 0.21, "learning_rate": 1.827791190489354e-05, "loss": 1.0226, "step": 3214 }, { "epoch": 0.21, "learning_rate": 1.8276702251521663e-05, "loss": 1.065, "step": 3215 }, { "epoch": 0.21, "learning_rate": 1.827549221350562e-05, "loss": 0.9926, "step": 3216 }, { "epoch": 0.21, "learning_rate": 1.827428179090165e-05, "loss": 1.0512, "step": 3217 }, { "epoch": 0.21, "learning_rate": 1.8273070983766004e-05, "loss": 1.0263, "step": 3218 }, { "epoch": 0.21, "learning_rate": 1.827185979215495e-05, "loss": 1.0266, "step": 3219 }, { "epoch": 0.21, "learning_rate": 1.8270648216124782e-05, "loss": 1.0114, "step": 3220 }, { "epoch": 0.21, "learning_rate": 1.8269436255731797e-05, "loss": 1.0185, "step": 3221 }, { "epoch": 0.21, "learning_rate": 1.8268223911032326e-05, "loss": 1.0196, "step": 3222 }, { "epoch": 0.21, "learning_rate": 1.8267011182082705e-05, "loss": 1.0435, "step": 3223 }, { "epoch": 0.21, "learning_rate": 1.8265798068939295e-05, "loss": 1.0214, "step": 3224 }, { "epoch": 0.21, "learning_rate": 1.8264584571658475e-05, "loss": 1.0435, "step": 3225 }, { "epoch": 0.21, "learning_rate": 1.826337069029664e-05, "loss": 1.0499, "step": 3226 }, { "epoch": 0.21, "learning_rate": 1.82621564249102e-05, "loss": 1.0103, "step": 3227 }, { "epoch": 0.21, "learning_rate": 1.8260941775555586e-05, "loss": 0.9943, "step": 3228 }, { "epoch": 0.21, "learning_rate": 1.825972674228925e-05, "loss": 1.0075, "step": 3229 }, { "epoch": 0.21, "learning_rate": 1.8258511325167656e-05, "loss": 1.0103, "step": 3230 }, { "epoch": 0.22, "learning_rate": 1.8257295524247287e-05, "loss": 1.0082, "step": 3231 }, { "epoch": 0.22, "learning_rate": 1.825607933958465e-05, "loss": 1.0168, "step": 3232 }, { "epoch": 0.22, "learning_rate": 1.8254862771236258e-05, "loss": 1.0691, "step": 3233 }, { "epoch": 0.22, "learning_rate": 1.8253645819258653e-05, "loss": 1.0271, "step": 3234 }, { "epoch": 0.22, "learning_rate": 1.8252428483708388e-05, "loss": 1.0683, "step": 3235 }, { "epoch": 0.22, "learning_rate": 1.8251210764642042e-05, "loss": 1.0319, "step": 3236 }, { "epoch": 0.22, "learning_rate": 1.82499926621162e-05, "loss": 1.0167, "step": 3237 }, { "epoch": 0.22, "learning_rate": 1.824877417618747e-05, "loss": 0.97, "step": 3238 }, { "epoch": 0.22, "learning_rate": 1.824755530691249e-05, "loss": 1.0341, "step": 3239 }, { "epoch": 0.22, "learning_rate": 1.824633605434789e-05, "loss": 1.0308, "step": 3240 }, { "epoch": 0.22, "learning_rate": 1.8245116418550343e-05, "loss": 1.021, "step": 3241 }, { "epoch": 0.22, "learning_rate": 1.8243896399576523e-05, "loss": 1.0397, "step": 3242 }, { "epoch": 0.22, "learning_rate": 1.824267599748313e-05, "loss": 1.054, "step": 3243 }, { "epoch": 0.22, "learning_rate": 1.8241455212326886e-05, "loss": 1.0133, "step": 3244 }, { "epoch": 0.22, "learning_rate": 1.8240234044164513e-05, "loss": 1.0461, "step": 3245 }, { "epoch": 0.22, "learning_rate": 1.823901249305277e-05, "loss": 0.9952, "step": 3246 }, { "epoch": 0.22, "learning_rate": 1.8237790559048426e-05, "loss": 1.0427, "step": 3247 }, { "epoch": 0.22, "learning_rate": 1.823656824220827e-05, "loss": 1.032, "step": 3248 }, { "epoch": 0.22, "learning_rate": 1.82353455425891e-05, "loss": 1.0502, "step": 3249 }, { "epoch": 0.22, "learning_rate": 1.823412246024774e-05, "loss": 1.0261, "step": 3250 }, { "epoch": 0.22, "learning_rate": 1.823289899524104e-05, "loss": 1.0239, "step": 3251 }, { "epoch": 0.22, "learning_rate": 1.823167514762585e-05, "loss": 1.0371, "step": 3252 }, { "epoch": 0.22, "learning_rate": 1.8230450917459045e-05, "loss": 0.9943, "step": 3253 }, { "epoch": 0.22, "learning_rate": 1.822922630479752e-05, "loss": 0.9996, "step": 3254 }, { "epoch": 0.22, "learning_rate": 1.822800130969819e-05, "loss": 1.0203, "step": 3255 }, { "epoch": 0.22, "learning_rate": 1.822677593221798e-05, "loss": 0.9898, "step": 3256 }, { "epoch": 0.22, "learning_rate": 1.8225550172413838e-05, "loss": 1.0267, "step": 3257 }, { "epoch": 0.22, "learning_rate": 1.8224324030342736e-05, "loss": 1.0088, "step": 3258 }, { "epoch": 0.22, "learning_rate": 1.8223097506061647e-05, "loss": 1.0126, "step": 3259 }, { "epoch": 0.22, "learning_rate": 1.8221870599627577e-05, "loss": 0.9857, "step": 3260 }, { "epoch": 0.22, "learning_rate": 1.8220643311097536e-05, "loss": 1.0317, "step": 3261 }, { "epoch": 0.22, "learning_rate": 1.8219415640528573e-05, "loss": 1.0232, "step": 3262 }, { "epoch": 0.22, "learning_rate": 1.821818758797773e-05, "loss": 0.9942, "step": 3263 }, { "epoch": 0.22, "learning_rate": 1.821695915350209e-05, "loss": 1.0001, "step": 3264 }, { "epoch": 0.22, "learning_rate": 1.8215730337158728e-05, "loss": 1.0086, "step": 3265 }, { "epoch": 0.22, "learning_rate": 1.821450113900476e-05, "loss": 1.0577, "step": 3266 }, { "epoch": 0.22, "learning_rate": 1.8213271559097312e-05, "loss": 1.0135, "step": 3267 }, { "epoch": 0.22, "learning_rate": 1.821204159749352e-05, "loss": 0.9845, "step": 3268 }, { "epoch": 0.22, "learning_rate": 1.8210811254250548e-05, "loss": 0.9745, "step": 3269 }, { "epoch": 0.22, "learning_rate": 1.8209580529425578e-05, "loss": 0.9823, "step": 3270 }, { "epoch": 0.22, "learning_rate": 1.8208349423075797e-05, "loss": 1.0392, "step": 3271 }, { "epoch": 0.22, "learning_rate": 1.8207117935258422e-05, "loss": 1.0489, "step": 3272 }, { "epoch": 0.22, "learning_rate": 1.8205886066030687e-05, "loss": 1.0317, "step": 3273 }, { "epoch": 0.22, "learning_rate": 1.8204653815449837e-05, "loss": 1.0735, "step": 3274 }, { "epoch": 0.22, "learning_rate": 1.820342118357314e-05, "loss": 1.0333, "step": 3275 }, { "epoch": 0.22, "learning_rate": 1.8202188170457883e-05, "loss": 0.9638, "step": 3276 }, { "epoch": 0.22, "learning_rate": 1.820095477616136e-05, "loss": 1.0669, "step": 3277 }, { "epoch": 0.22, "learning_rate": 1.81997210007409e-05, "loss": 1.0275, "step": 3278 }, { "epoch": 0.22, "learning_rate": 1.8198486844253833e-05, "loss": 0.9814, "step": 3279 }, { "epoch": 0.22, "learning_rate": 1.8197252306757517e-05, "loss": 1.0166, "step": 3280 }, { "epoch": 0.22, "learning_rate": 1.819601738830933e-05, "loss": 1.0399, "step": 3281 }, { "epoch": 0.22, "learning_rate": 1.819478208896665e-05, "loss": 1.0042, "step": 3282 }, { "epoch": 0.22, "learning_rate": 1.81935464087869e-05, "loss": 1.0426, "step": 3283 }, { "epoch": 0.22, "learning_rate": 1.8192310347827496e-05, "loss": 1.0193, "step": 3284 }, { "epoch": 0.22, "learning_rate": 1.819107390614588e-05, "loss": 1.025, "step": 3285 }, { "epoch": 0.22, "learning_rate": 1.8189837083799525e-05, "loss": 1.0492, "step": 3286 }, { "epoch": 0.22, "learning_rate": 1.8188599880845897e-05, "loss": 1.0269, "step": 3287 }, { "epoch": 0.22, "learning_rate": 1.81873622973425e-05, "loss": 1.0188, "step": 3288 }, { "epoch": 0.22, "learning_rate": 1.818612433334685e-05, "loss": 1.0372, "step": 3289 }, { "epoch": 0.22, "learning_rate": 1.818488598891647e-05, "loss": 1.0577, "step": 3290 }, { "epoch": 0.22, "learning_rate": 1.818364726410892e-05, "loss": 1.0163, "step": 3291 }, { "epoch": 0.22, "learning_rate": 1.818240815898176e-05, "loss": 1.0386, "step": 3292 }, { "epoch": 0.22, "learning_rate": 1.818116867359258e-05, "loss": 0.9989, "step": 3293 }, { "epoch": 0.22, "learning_rate": 1.8179928807998977e-05, "loss": 0.9988, "step": 3294 }, { "epoch": 0.22, "learning_rate": 1.817868856225858e-05, "loss": 1.046, "step": 3295 }, { "epoch": 0.22, "learning_rate": 1.817744793642902e-05, "loss": 1.0194, "step": 3296 }, { "epoch": 0.22, "learning_rate": 1.8176206930567953e-05, "loss": 1.0523, "step": 3297 }, { "epoch": 0.22, "learning_rate": 1.8174965544733057e-05, "loss": 0.9772, "step": 3298 }, { "epoch": 0.22, "learning_rate": 1.817372377898202e-05, "loss": 1.0546, "step": 3299 }, { "epoch": 0.22, "learning_rate": 1.817248163337255e-05, "loss": 1.0085, "step": 3300 }, { "epoch": 0.22, "learning_rate": 1.8171239107962372e-05, "loss": 1.0334, "step": 3301 }, { "epoch": 0.22, "learning_rate": 1.816999620280924e-05, "loss": 1.061, "step": 3302 }, { "epoch": 0.22, "learning_rate": 1.8168752917970903e-05, "loss": 1.0436, "step": 3303 }, { "epoch": 0.22, "learning_rate": 1.8167509253505143e-05, "loss": 0.9791, "step": 3304 }, { "epoch": 0.22, "learning_rate": 1.8166265209469764e-05, "loss": 1.0322, "step": 3305 }, { "epoch": 0.22, "learning_rate": 1.8165020785922577e-05, "loss": 1.0463, "step": 3306 }, { "epoch": 0.22, "learning_rate": 1.816377598292141e-05, "loss": 1.0293, "step": 3307 }, { "epoch": 0.22, "learning_rate": 1.8162530800524116e-05, "loss": 1.049, "step": 3308 }, { "epoch": 0.22, "learning_rate": 1.8161285238788563e-05, "loss": 1.0366, "step": 3309 }, { "epoch": 0.22, "learning_rate": 1.816003929777264e-05, "loss": 1.0573, "step": 3310 }, { "epoch": 0.22, "learning_rate": 1.815879297753424e-05, "loss": 1.023, "step": 3311 }, { "epoch": 0.22, "learning_rate": 1.8157546278131294e-05, "loss": 1.015, "step": 3312 }, { "epoch": 0.22, "learning_rate": 1.8156299199621733e-05, "loss": 1.0305, "step": 3313 }, { "epoch": 0.22, "learning_rate": 1.8155051742063513e-05, "loss": 1.0232, "step": 3314 }, { "epoch": 0.22, "learning_rate": 1.815380390551461e-05, "loss": 1.026, "step": 3315 }, { "epoch": 0.22, "learning_rate": 1.8152555690033012e-05, "loss": 0.9985, "step": 3316 }, { "epoch": 0.22, "learning_rate": 1.815130709567673e-05, "loss": 1.0101, "step": 3317 }, { "epoch": 0.22, "learning_rate": 1.8150058122503793e-05, "loss": 1.033, "step": 3318 }, { "epoch": 0.22, "learning_rate": 1.814880877057224e-05, "loss": 1.0244, "step": 3319 }, { "epoch": 0.22, "learning_rate": 1.814755903994013e-05, "loss": 1.0342, "step": 3320 }, { "epoch": 0.22, "learning_rate": 1.814630893066555e-05, "loss": 1.0189, "step": 3321 }, { "epoch": 0.22, "learning_rate": 1.8145058442806584e-05, "loss": 1.0293, "step": 3322 }, { "epoch": 0.22, "learning_rate": 1.8143807576421357e-05, "loss": 1.059, "step": 3323 }, { "epoch": 0.22, "learning_rate": 1.8142556331568002e-05, "loss": 1.0432, "step": 3324 }, { "epoch": 0.22, "learning_rate": 1.8141304708304657e-05, "loss": 1.0295, "step": 3325 }, { "epoch": 0.22, "learning_rate": 1.81400527066895e-05, "loss": 0.9929, "step": 3326 }, { "epoch": 0.22, "learning_rate": 1.8138800326780706e-05, "loss": 1.0325, "step": 3327 }, { "epoch": 0.22, "learning_rate": 1.8137547568636486e-05, "loss": 1.0168, "step": 3328 }, { "epoch": 0.22, "learning_rate": 1.8136294432315057e-05, "loss": 1.0318, "step": 3329 }, { "epoch": 0.22, "learning_rate": 1.813504091787465e-05, "loss": 0.9901, "step": 3330 }, { "epoch": 0.22, "learning_rate": 1.813378702537353e-05, "loss": 1.026, "step": 3331 }, { "epoch": 0.22, "learning_rate": 1.8132532754869956e-05, "loss": 1.0294, "step": 3332 }, { "epoch": 0.22, "learning_rate": 1.813127810642223e-05, "loss": 1.0468, "step": 3333 }, { "epoch": 0.22, "learning_rate": 1.8130023080088653e-05, "loss": 1.0078, "step": 3334 }, { "epoch": 0.22, "learning_rate": 1.8128767675927553e-05, "loss": 1.0382, "step": 3335 }, { "epoch": 0.22, "learning_rate": 1.8127511893997274e-05, "loss": 1.072, "step": 3336 }, { "epoch": 0.22, "learning_rate": 1.812625573435617e-05, "loss": 1.0574, "step": 3337 }, { "epoch": 0.22, "learning_rate": 1.8124999197062622e-05, "loss": 1.0435, "step": 3338 }, { "epoch": 0.22, "learning_rate": 1.812374228217503e-05, "loss": 1.0179, "step": 3339 }, { "epoch": 0.22, "learning_rate": 1.81224849897518e-05, "loss": 1.0246, "step": 3340 }, { "epoch": 0.22, "learning_rate": 1.812122731985136e-05, "loss": 0.9976, "step": 3341 }, { "epoch": 0.22, "learning_rate": 1.8119969272532164e-05, "loss": 0.9888, "step": 3342 }, { "epoch": 0.22, "learning_rate": 1.811871084785268e-05, "loss": 1.0419, "step": 3343 }, { "epoch": 0.22, "learning_rate": 1.8117452045871384e-05, "loss": 0.9883, "step": 3344 }, { "epoch": 0.22, "learning_rate": 1.8116192866646777e-05, "loss": 1.0402, "step": 3345 }, { "epoch": 0.22, "learning_rate": 1.811493331023738e-05, "loss": 0.9982, "step": 3346 }, { "epoch": 0.22, "learning_rate": 1.8113673376701727e-05, "loss": 1.0431, "step": 3347 }, { "epoch": 0.22, "learning_rate": 1.8112413066098377e-05, "loss": 1.0403, "step": 3348 }, { "epoch": 0.22, "learning_rate": 1.8111152378485887e-05, "loss": 1.0615, "step": 3349 }, { "epoch": 0.22, "learning_rate": 1.8109891313922858e-05, "loss": 1.0381, "step": 3350 }, { "epoch": 0.22, "learning_rate": 1.810862987246789e-05, "loss": 0.9988, "step": 3351 }, { "epoch": 0.22, "learning_rate": 1.8107368054179607e-05, "loss": 1.0518, "step": 3352 }, { "epoch": 0.22, "learning_rate": 1.8106105859116648e-05, "loss": 1.0588, "step": 3353 }, { "epoch": 0.22, "learning_rate": 1.8104843287337676e-05, "loss": 1.0242, "step": 3354 }, { "epoch": 0.22, "learning_rate": 1.810358033890136e-05, "loss": 1.0451, "step": 3355 }, { "epoch": 0.22, "learning_rate": 1.8102317013866398e-05, "loss": 1.032, "step": 3356 }, { "epoch": 0.22, "learning_rate": 1.81010533122915e-05, "loss": 1.0185, "step": 3357 }, { "epoch": 0.22, "learning_rate": 1.8099789234235393e-05, "loss": 1.0392, "step": 3358 }, { "epoch": 0.22, "learning_rate": 1.8098524779756822e-05, "loss": 1.0276, "step": 3359 }, { "epoch": 0.22, "learning_rate": 1.809725994891455e-05, "loss": 0.9851, "step": 3360 }, { "epoch": 0.22, "learning_rate": 1.809599474176736e-05, "loss": 1.0322, "step": 3361 }, { "epoch": 0.22, "learning_rate": 1.809472915837405e-05, "loss": 1.032, "step": 3362 }, { "epoch": 0.22, "learning_rate": 1.8093463198793433e-05, "loss": 1.0512, "step": 3363 }, { "epoch": 0.22, "learning_rate": 1.8092196863084342e-05, "loss": 1.0014, "step": 3364 }, { "epoch": 0.22, "learning_rate": 1.8090930151305633e-05, "loss": 1.0573, "step": 3365 }, { "epoch": 0.22, "learning_rate": 1.8089663063516168e-05, "loss": 1.0373, "step": 3366 }, { "epoch": 0.22, "learning_rate": 1.8088395599774835e-05, "loss": 1.0378, "step": 3367 }, { "epoch": 0.22, "learning_rate": 1.8087127760140535e-05, "loss": 1.0433, "step": 3368 }, { "epoch": 0.22, "learning_rate": 1.8085859544672194e-05, "loss": 1.0104, "step": 3369 }, { "epoch": 0.22, "learning_rate": 1.808459095342874e-05, "loss": 1.0445, "step": 3370 }, { "epoch": 0.22, "learning_rate": 1.8083321986469144e-05, "loss": 0.9986, "step": 3371 }, { "epoch": 0.22, "learning_rate": 1.8082052643852362e-05, "loss": 1.0056, "step": 3372 }, { "epoch": 0.22, "learning_rate": 1.8080782925637394e-05, "loss": 1.0358, "step": 3373 }, { "epoch": 0.22, "learning_rate": 1.8079512831883243e-05, "loss": 1.0177, "step": 3374 }, { "epoch": 0.22, "learning_rate": 1.8078242362648937e-05, "loss": 1.0624, "step": 3375 }, { "epoch": 0.22, "learning_rate": 1.807697151799352e-05, "loss": 1.0057, "step": 3376 }, { "epoch": 0.22, "learning_rate": 1.807570029797605e-05, "loss": 1.0179, "step": 3377 }, { "epoch": 0.22, "learning_rate": 1.8074428702655606e-05, "loss": 0.9708, "step": 3378 }, { "epoch": 0.22, "learning_rate": 1.807315673209128e-05, "loss": 1.0287, "step": 3379 }, { "epoch": 0.22, "learning_rate": 1.8071884386342188e-05, "loss": 1.0169, "step": 3380 }, { "epoch": 0.23, "learning_rate": 1.8070611665467453e-05, "loss": 1.0106, "step": 3381 }, { "epoch": 0.23, "learning_rate": 1.8069338569526235e-05, "loss": 1.0546, "step": 3382 }, { "epoch": 0.23, "learning_rate": 1.806806509857769e-05, "loss": 1.0198, "step": 3383 }, { "epoch": 0.23, "learning_rate": 1.8066791252681e-05, "loss": 1.0168, "step": 3384 }, { "epoch": 0.23, "learning_rate": 1.8065517031895363e-05, "loss": 1.0277, "step": 3385 }, { "epoch": 0.23, "learning_rate": 1.8064242436280004e-05, "loss": 1.023, "step": 3386 }, { "epoch": 0.23, "learning_rate": 1.806296746589415e-05, "loss": 1.0432, "step": 3387 }, { "epoch": 0.23, "learning_rate": 1.8061692120797054e-05, "loss": 1.0074, "step": 3388 }, { "epoch": 0.23, "learning_rate": 1.8060416401047987e-05, "loss": 1.0208, "step": 3389 }, { "epoch": 0.23, "learning_rate": 1.805914030670624e-05, "loss": 1.0058, "step": 3390 }, { "epoch": 0.23, "learning_rate": 1.8057863837831107e-05, "loss": 1.088, "step": 3391 }, { "epoch": 0.23, "learning_rate": 1.8056586994481916e-05, "loss": 1.0755, "step": 3392 }, { "epoch": 0.23, "learning_rate": 1.8055309776718006e-05, "loss": 1.0248, "step": 3393 }, { "epoch": 0.23, "learning_rate": 1.8054032184598733e-05, "loss": 1.0127, "step": 3394 }, { "epoch": 0.23, "learning_rate": 1.8052754218183465e-05, "loss": 1.0131, "step": 3395 }, { "epoch": 0.23, "learning_rate": 1.8051475877531605e-05, "loss": 1.0123, "step": 3396 }, { "epoch": 0.23, "learning_rate": 1.805019716270255e-05, "loss": 1.057, "step": 3397 }, { "epoch": 0.23, "learning_rate": 1.804891807375573e-05, "loss": 1.0404, "step": 3398 }, { "epoch": 0.23, "learning_rate": 1.804763861075059e-05, "loss": 1.0144, "step": 3399 }, { "epoch": 0.23, "learning_rate": 1.8046358773746585e-05, "loss": 0.9897, "step": 3400 }, { "epoch": 0.23, "learning_rate": 1.8045078562803203e-05, "loss": 1.0827, "step": 3401 }, { "epoch": 0.23, "learning_rate": 1.804379797797993e-05, "loss": 1.0167, "step": 3402 }, { "epoch": 0.23, "learning_rate": 1.8042517019336284e-05, "loss": 1.0413, "step": 3403 }, { "epoch": 0.23, "learning_rate": 1.804123568693179e-05, "loss": 0.9886, "step": 3404 }, { "epoch": 0.23, "learning_rate": 1.8039953980826e-05, "loss": 1.0089, "step": 3405 }, { "epoch": 0.23, "learning_rate": 1.8038671901078482e-05, "loss": 0.9971, "step": 3406 }, { "epoch": 0.23, "learning_rate": 1.803738944774881e-05, "loss": 1.0368, "step": 3407 }, { "epoch": 0.23, "learning_rate": 1.803610662089659e-05, "loss": 1.0179, "step": 3408 }, { "epoch": 0.23, "learning_rate": 1.8034823420581432e-05, "loss": 1.0267, "step": 3409 }, { "epoch": 0.23, "learning_rate": 1.803353984686298e-05, "loss": 1.0103, "step": 3410 }, { "epoch": 0.23, "learning_rate": 1.8032255899800878e-05, "loss": 1.0563, "step": 3411 }, { "epoch": 0.23, "learning_rate": 1.8030971579454797e-05, "loss": 1.0476, "step": 3412 }, { "epoch": 0.23, "learning_rate": 1.8029686885884427e-05, "loss": 1.0069, "step": 3413 }, { "epoch": 0.23, "learning_rate": 1.8028401819149467e-05, "loss": 1.0299, "step": 3414 }, { "epoch": 0.23, "learning_rate": 1.8027116379309637e-05, "loss": 1.0639, "step": 3415 }, { "epoch": 0.23, "learning_rate": 1.8025830566424682e-05, "loss": 1.008, "step": 3416 }, { "epoch": 0.23, "learning_rate": 1.8024544380554353e-05, "loss": 1.0279, "step": 3417 }, { "epoch": 0.23, "learning_rate": 1.8023257821758422e-05, "loss": 0.9975, "step": 3418 }, { "epoch": 0.23, "learning_rate": 1.8021970890096684e-05, "loss": 1.0091, "step": 3419 }, { "epoch": 0.23, "learning_rate": 1.8020683585628942e-05, "loss": 1.0168, "step": 3420 }, { "epoch": 0.23, "learning_rate": 1.8019395908415025e-05, "loss": 0.9997, "step": 3421 }, { "epoch": 0.23, "learning_rate": 1.8018107858514772e-05, "loss": 0.9756, "step": 3422 }, { "epoch": 0.23, "learning_rate": 1.8016819435988043e-05, "loss": 1.0662, "step": 3423 }, { "epoch": 0.23, "learning_rate": 1.801553064089472e-05, "loss": 0.995, "step": 3424 }, { "epoch": 0.23, "learning_rate": 1.8014241473294692e-05, "loss": 1.008, "step": 3425 }, { "epoch": 0.23, "learning_rate": 1.801295193324787e-05, "loss": 0.9595, "step": 3426 }, { "epoch": 0.23, "learning_rate": 1.8011662020814186e-05, "loss": 1.0772, "step": 3427 }, { "epoch": 0.23, "learning_rate": 1.8010371736053586e-05, "loss": 0.9859, "step": 3428 }, { "epoch": 0.23, "learning_rate": 1.8009081079026034e-05, "loss": 1.0578, "step": 3429 }, { "epoch": 0.23, "learning_rate": 1.800779004979151e-05, "loss": 1.021, "step": 3430 }, { "epoch": 0.23, "learning_rate": 1.8006498648410012e-05, "loss": 1.0003, "step": 3431 }, { "epoch": 0.23, "learning_rate": 1.8005206874941553e-05, "loss": 1.0083, "step": 3432 }, { "epoch": 0.23, "learning_rate": 1.8003914729446167e-05, "loss": 1.0067, "step": 3433 }, { "epoch": 0.23, "learning_rate": 1.800262221198391e-05, "loss": 1.0246, "step": 3434 }, { "epoch": 0.23, "learning_rate": 1.8001329322614842e-05, "loss": 0.9898, "step": 3435 }, { "epoch": 0.23, "learning_rate": 1.800003606139905e-05, "loss": 1.0538, "step": 3436 }, { "epoch": 0.23, "learning_rate": 1.7998742428396633e-05, "loss": 1.0376, "step": 3437 }, { "epoch": 0.23, "learning_rate": 1.7997448423667716e-05, "loss": 1.0437, "step": 3438 }, { "epoch": 0.23, "learning_rate": 1.7996154047272434e-05, "loss": 0.9997, "step": 3439 }, { "epoch": 0.23, "learning_rate": 1.7994859299270937e-05, "loss": 0.9697, "step": 3440 }, { "epoch": 0.23, "learning_rate": 1.7993564179723398e-05, "loss": 1.0022, "step": 3441 }, { "epoch": 0.23, "learning_rate": 1.799226868869001e-05, "loss": 1.0315, "step": 3442 }, { "epoch": 0.23, "learning_rate": 1.7990972826230968e-05, "loss": 1.0057, "step": 3443 }, { "epoch": 0.23, "learning_rate": 1.7989676592406504e-05, "loss": 1.0437, "step": 3444 }, { "epoch": 0.23, "learning_rate": 1.7988379987276854e-05, "loss": 1.0056, "step": 3445 }, { "epoch": 0.23, "learning_rate": 1.7987083010902276e-05, "loss": 1.0663, "step": 3446 }, { "epoch": 0.23, "learning_rate": 1.7985785663343047e-05, "loss": 1.0572, "step": 3447 }, { "epoch": 0.23, "learning_rate": 1.7984487944659452e-05, "loss": 0.9635, "step": 3448 }, { "epoch": 0.23, "learning_rate": 1.7983189854911808e-05, "loss": 1.0415, "step": 3449 }, { "epoch": 0.23, "learning_rate": 1.7981891394160436e-05, "loss": 1.0272, "step": 3450 }, { "epoch": 0.23, "learning_rate": 1.798059256246568e-05, "loss": 1.0096, "step": 3451 }, { "epoch": 0.23, "learning_rate": 1.7979293359887904e-05, "loss": 1.0517, "step": 3452 }, { "epoch": 0.23, "learning_rate": 1.7977993786487483e-05, "loss": 1.0298, "step": 3453 }, { "epoch": 0.23, "learning_rate": 1.7976693842324814e-05, "loss": 1.0038, "step": 3454 }, { "epoch": 0.23, "learning_rate": 1.7975393527460306e-05, "loss": 0.9988, "step": 3455 }, { "epoch": 0.23, "learning_rate": 1.7974092841954395e-05, "loss": 1.0162, "step": 3456 }, { "epoch": 0.23, "learning_rate": 1.7972791785867523e-05, "loss": 1.0525, "step": 3457 }, { "epoch": 0.23, "learning_rate": 1.7971490359260156e-05, "loss": 1.0304, "step": 3458 }, { "epoch": 0.23, "learning_rate": 1.7970188562192776e-05, "loss": 1.0461, "step": 3459 }, { "epoch": 0.23, "learning_rate": 1.7968886394725876e-05, "loss": 1.0402, "step": 3460 }, { "epoch": 0.23, "learning_rate": 1.7967583856919977e-05, "loss": 1.0104, "step": 3461 }, { "epoch": 0.23, "learning_rate": 1.7966280948835614e-05, "loss": 0.9926, "step": 3462 }, { "epoch": 0.23, "learning_rate": 1.7964977670533333e-05, "loss": 0.9865, "step": 3463 }, { "epoch": 0.23, "learning_rate": 1.7963674022073703e-05, "loss": 0.9664, "step": 3464 }, { "epoch": 0.23, "learning_rate": 1.796237000351731e-05, "loss": 1.042, "step": 3465 }, { "epoch": 0.23, "learning_rate": 1.7961065614924748e-05, "loss": 1.0158, "step": 3466 }, { "epoch": 0.23, "learning_rate": 1.7959760856356646e-05, "loss": 1.079, "step": 3467 }, { "epoch": 0.23, "learning_rate": 1.7958455727873638e-05, "loss": 0.9778, "step": 3468 }, { "epoch": 0.23, "learning_rate": 1.7957150229536372e-05, "loss": 1.0057, "step": 3469 }, { "epoch": 0.23, "learning_rate": 1.7955844361405525e-05, "loss": 1.0213, "step": 3470 }, { "epoch": 0.23, "learning_rate": 1.795453812354178e-05, "loss": 0.991, "step": 3471 }, { "epoch": 0.23, "learning_rate": 1.795323151600585e-05, "loss": 0.9999, "step": 3472 }, { "epoch": 0.23, "learning_rate": 1.7951924538858443e-05, "loss": 1.0286, "step": 3473 }, { "epoch": 0.23, "learning_rate": 1.795061719216031e-05, "loss": 1.0289, "step": 3474 }, { "epoch": 0.23, "learning_rate": 1.7949309475972207e-05, "loss": 0.9931, "step": 3475 }, { "epoch": 0.23, "learning_rate": 1.7948001390354898e-05, "loss": 1.0052, "step": 3476 }, { "epoch": 0.23, "learning_rate": 1.7946692935369187e-05, "loss": 1.0228, "step": 3477 }, { "epoch": 0.23, "learning_rate": 1.794538411107587e-05, "loss": 0.9988, "step": 3478 }, { "epoch": 0.23, "learning_rate": 1.794407491753578e-05, "loss": 1.0168, "step": 3479 }, { "epoch": 0.23, "learning_rate": 1.7942765354809757e-05, "loss": 1.0149, "step": 3480 }, { "epoch": 0.23, "learning_rate": 1.794145542295866e-05, "loss": 1.0047, "step": 3481 }, { "epoch": 0.23, "learning_rate": 1.7940145122043368e-05, "loss": 1.0153, "step": 3482 }, { "epoch": 0.23, "learning_rate": 1.7938834452124766e-05, "loss": 1.0437, "step": 3483 }, { "epoch": 0.23, "learning_rate": 1.793752341326378e-05, "loss": 1.0327, "step": 3484 }, { "epoch": 0.23, "learning_rate": 1.7936212005521322e-05, "loss": 1.0813, "step": 3485 }, { "epoch": 0.23, "learning_rate": 1.793490022895835e-05, "loss": 1.018, "step": 3486 }, { "epoch": 0.23, "learning_rate": 1.793358808363582e-05, "loss": 1.0424, "step": 3487 }, { "epoch": 0.23, "learning_rate": 1.793227556961471e-05, "loss": 0.9828, "step": 3488 }, { "epoch": 0.23, "learning_rate": 1.793096268695602e-05, "loss": 1.0088, "step": 3489 }, { "epoch": 0.23, "learning_rate": 1.7929649435720768e-05, "loss": 1.0241, "step": 3490 }, { "epoch": 0.23, "learning_rate": 1.7928335815969977e-05, "loss": 0.9838, "step": 3491 }, { "epoch": 0.23, "learning_rate": 1.7927021827764696e-05, "loss": 1.0055, "step": 3492 }, { "epoch": 0.23, "learning_rate": 1.7925707471165993e-05, "loss": 1.0318, "step": 3493 }, { "epoch": 0.23, "learning_rate": 1.792439274623495e-05, "loss": 1.0023, "step": 3494 }, { "epoch": 0.23, "learning_rate": 1.7923077653032666e-05, "loss": 1.0048, "step": 3495 }, { "epoch": 0.23, "learning_rate": 1.792176219162026e-05, "loss": 1.0062, "step": 3496 }, { "epoch": 0.23, "learning_rate": 1.7920446362058858e-05, "loss": 1.0118, "step": 3497 }, { "epoch": 0.23, "learning_rate": 1.7919130164409618e-05, "loss": 1.0304, "step": 3498 }, { "epoch": 0.23, "learning_rate": 1.79178135987337e-05, "loss": 1.0896, "step": 3499 }, { "epoch": 0.23, "learning_rate": 1.7916496665092302e-05, "loss": 0.9986, "step": 3500 }, { "epoch": 0.23, "learning_rate": 1.7915179363546612e-05, "loss": 1.0385, "step": 3501 }, { "epoch": 0.23, "learning_rate": 1.791386169415786e-05, "loss": 0.964, "step": 3502 }, { "epoch": 0.23, "learning_rate": 1.7912543656987274e-05, "loss": 1.03, "step": 3503 }, { "epoch": 0.23, "learning_rate": 1.7911225252096115e-05, "loss": 1.0326, "step": 3504 }, { "epoch": 0.23, "learning_rate": 1.7909906479545646e-05, "loss": 1.0384, "step": 3505 }, { "epoch": 0.23, "learning_rate": 1.790858733939716e-05, "loss": 1.0587, "step": 3506 }, { "epoch": 0.23, "learning_rate": 1.7907267831711954e-05, "loss": 0.9829, "step": 3507 }, { "epoch": 0.23, "learning_rate": 1.7905947956551362e-05, "loss": 1.0096, "step": 3508 }, { "epoch": 0.23, "learning_rate": 1.790462771397671e-05, "loss": 1.0517, "step": 3509 }, { "epoch": 0.23, "learning_rate": 1.7903307104049362e-05, "loss": 1.0204, "step": 3510 }, { "epoch": 0.23, "learning_rate": 1.7901986126830686e-05, "loss": 1.0027, "step": 3511 }, { "epoch": 0.23, "learning_rate": 1.7900664782382074e-05, "loss": 0.9874, "step": 3512 }, { "epoch": 0.23, "learning_rate": 1.7899343070764938e-05, "loss": 1.0579, "step": 3513 }, { "epoch": 0.23, "learning_rate": 1.7898020992040693e-05, "loss": 0.997, "step": 3514 }, { "epoch": 0.23, "learning_rate": 1.7896698546270784e-05, "loss": 1.0371, "step": 3515 }, { "epoch": 0.23, "learning_rate": 1.7895375733516675e-05, "loss": 1.0026, "step": 3516 }, { "epoch": 0.23, "learning_rate": 1.789405255383983e-05, "loss": 1.0191, "step": 3517 }, { "epoch": 0.23, "learning_rate": 1.7892729007301748e-05, "loss": 1.0213, "step": 3518 }, { "epoch": 0.23, "learning_rate": 1.789140509396394e-05, "loss": 1.0423, "step": 3519 }, { "epoch": 0.23, "learning_rate": 1.7890080813887927e-05, "loss": 1.0011, "step": 3520 }, { "epoch": 0.23, "learning_rate": 1.7888756167135254e-05, "loss": 1.0212, "step": 3521 }, { "epoch": 0.23, "learning_rate": 1.7887431153767484e-05, "loss": 0.9907, "step": 3522 }, { "epoch": 0.23, "learning_rate": 1.7886105773846195e-05, "loss": 0.9978, "step": 3523 }, { "epoch": 0.23, "learning_rate": 1.788478002743298e-05, "loss": 0.9644, "step": 3524 }, { "epoch": 0.23, "learning_rate": 1.7883453914589446e-05, "loss": 1.0626, "step": 3525 }, { "epoch": 0.23, "learning_rate": 1.788212743537723e-05, "loss": 1.0104, "step": 3526 }, { "epoch": 0.23, "learning_rate": 1.7880800589857973e-05, "loss": 1.0174, "step": 3527 }, { "epoch": 0.23, "learning_rate": 1.7879473378093336e-05, "loss": 1.0246, "step": 3528 }, { "epoch": 0.23, "learning_rate": 1.7878145800145002e-05, "loss": 0.9867, "step": 3529 }, { "epoch": 0.23, "learning_rate": 1.7876817856074663e-05, "loss": 1.009, "step": 3530 }, { "epoch": 0.24, "learning_rate": 1.7875489545944043e-05, "loss": 1.0404, "step": 3531 }, { "epoch": 0.24, "learning_rate": 1.787416086981486e-05, "loss": 1.0203, "step": 3532 }, { "epoch": 0.24, "learning_rate": 1.7872831827748874e-05, "loss": 1.0118, "step": 3533 }, { "epoch": 0.24, "learning_rate": 1.7871502419807837e-05, "loss": 1.0087, "step": 3534 }, { "epoch": 0.24, "learning_rate": 1.787017264605354e-05, "loss": 1.0436, "step": 3535 }, { "epoch": 0.24, "learning_rate": 1.786884250654778e-05, "loss": 1.0405, "step": 3536 }, { "epoch": 0.24, "learning_rate": 1.7867512001352367e-05, "loss": 1.0418, "step": 3537 }, { "epoch": 0.24, "learning_rate": 1.7866181130529142e-05, "loss": 1.0281, "step": 3538 }, { "epoch": 0.24, "learning_rate": 1.7864849894139952e-05, "loss": 1.0312, "step": 3539 }, { "epoch": 0.24, "learning_rate": 1.786351829224666e-05, "loss": 1.0179, "step": 3540 }, { "epoch": 0.24, "learning_rate": 1.7862186324911154e-05, "loss": 1.0044, "step": 3541 }, { "epoch": 0.24, "learning_rate": 1.786085399219533e-05, "loss": 1.025, "step": 3542 }, { "epoch": 0.24, "learning_rate": 1.7859521294161114e-05, "loss": 1.0275, "step": 3543 }, { "epoch": 0.24, "learning_rate": 1.785818823087043e-05, "loss": 1.0491, "step": 3544 }, { "epoch": 0.24, "learning_rate": 1.7856854802385237e-05, "loss": 1.0131, "step": 3545 }, { "epoch": 0.24, "learning_rate": 1.7855521008767502e-05, "loss": 0.9837, "step": 3546 }, { "epoch": 0.24, "learning_rate": 1.785418685007921e-05, "loss": 1.0237, "step": 3547 }, { "epoch": 0.24, "learning_rate": 1.7852852326382362e-05, "loss": 0.9623, "step": 3548 }, { "epoch": 0.24, "learning_rate": 1.7851517437738983e-05, "loss": 1.041, "step": 3549 }, { "epoch": 0.24, "learning_rate": 1.7850182184211104e-05, "loss": 0.9979, "step": 3550 }, { "epoch": 0.24, "learning_rate": 1.784884656586078e-05, "loss": 1.017, "step": 3551 }, { "epoch": 0.24, "learning_rate": 1.784751058275008e-05, "loss": 1.0024, "step": 3552 }, { "epoch": 0.24, "learning_rate": 1.7846174234941092e-05, "loss": 1.0045, "step": 3553 }, { "epoch": 0.24, "learning_rate": 1.7844837522495923e-05, "loss": 0.9992, "step": 3554 }, { "epoch": 0.24, "learning_rate": 1.784350044547669e-05, "loss": 1.0136, "step": 3555 }, { "epoch": 0.24, "learning_rate": 1.7842163003945532e-05, "loss": 0.9819, "step": 3556 }, { "epoch": 0.24, "learning_rate": 1.784082519796461e-05, "loss": 1.0251, "step": 3557 }, { "epoch": 0.24, "learning_rate": 1.7839487027596088e-05, "loss": 1.0238, "step": 3558 }, { "epoch": 0.24, "learning_rate": 1.7838148492902156e-05, "loss": 1.041, "step": 3559 }, { "epoch": 0.24, "learning_rate": 1.7836809593945025e-05, "loss": 1.0286, "step": 3560 }, { "epoch": 0.24, "learning_rate": 1.7835470330786915e-05, "loss": 1.0825, "step": 3561 }, { "epoch": 0.24, "learning_rate": 1.7834130703490066e-05, "loss": 1.0277, "step": 3562 }, { "epoch": 0.24, "learning_rate": 1.7832790712116735e-05, "loss": 0.9911, "step": 3563 }, { "epoch": 0.24, "learning_rate": 1.7831450356729193e-05, "loss": 1.0569, "step": 3564 }, { "epoch": 0.24, "learning_rate": 1.7830109637389733e-05, "loss": 1.0058, "step": 3565 }, { "epoch": 0.24, "learning_rate": 1.782876855416066e-05, "loss": 1.0379, "step": 3566 }, { "epoch": 0.24, "learning_rate": 1.78274271071043e-05, "loss": 1.0127, "step": 3567 }, { "epoch": 0.24, "learning_rate": 1.7826085296282998e-05, "loss": 1.0256, "step": 3568 }, { "epoch": 0.24, "learning_rate": 1.7824743121759106e-05, "loss": 1.0621, "step": 3569 }, { "epoch": 0.24, "learning_rate": 1.7823400583595e-05, "loss": 1.0192, "step": 3570 }, { "epoch": 0.24, "learning_rate": 1.7822057681853075e-05, "loss": 1.023, "step": 3571 }, { "epoch": 0.24, "learning_rate": 1.782071441659574e-05, "loss": 0.9575, "step": 3572 }, { "epoch": 0.24, "learning_rate": 1.7819370787885418e-05, "loss": 1.0287, "step": 3573 }, { "epoch": 0.24, "learning_rate": 1.781802679578455e-05, "loss": 1.0371, "step": 3574 }, { "epoch": 0.24, "learning_rate": 1.78166824403556e-05, "loss": 1.0009, "step": 3575 }, { "epoch": 0.24, "learning_rate": 1.7815337721661043e-05, "loss": 1.0108, "step": 3576 }, { "epoch": 0.24, "learning_rate": 1.781399263976337e-05, "loss": 1.0226, "step": 3577 }, { "epoch": 0.24, "learning_rate": 1.7812647194725093e-05, "loss": 1.039, "step": 3578 }, { "epoch": 0.24, "learning_rate": 1.781130138660874e-05, "loss": 0.9907, "step": 3579 }, { "epoch": 0.24, "learning_rate": 1.7809955215476855e-05, "loss": 1.0465, "step": 3580 }, { "epoch": 0.24, "learning_rate": 1.7808608681391998e-05, "loss": 1.0441, "step": 3581 }, { "epoch": 0.24, "learning_rate": 1.7807261784416743e-05, "loss": 1.0062, "step": 3582 }, { "epoch": 0.24, "learning_rate": 1.7805914524613688e-05, "loss": 1.0218, "step": 3583 }, { "epoch": 0.24, "learning_rate": 1.780456690204545e-05, "loss": 1.0087, "step": 3584 }, { "epoch": 0.24, "learning_rate": 1.7803218916774643e-05, "loss": 1.0024, "step": 3585 }, { "epoch": 0.24, "learning_rate": 1.7801870568863923e-05, "loss": 0.9824, "step": 3586 }, { "epoch": 0.24, "learning_rate": 1.7800521858375953e-05, "loss": 1.0339, "step": 3587 }, { "epoch": 0.24, "learning_rate": 1.7799172785373405e-05, "loss": 1.004, "step": 3588 }, { "epoch": 0.24, "learning_rate": 1.7797823349918977e-05, "loss": 1.0338, "step": 3589 }, { "epoch": 0.24, "learning_rate": 1.7796473552075385e-05, "loss": 1.0495, "step": 3590 }, { "epoch": 0.24, "learning_rate": 1.7795123391905354e-05, "loss": 1.0418, "step": 3591 }, { "epoch": 0.24, "learning_rate": 1.7793772869471632e-05, "loss": 1.0166, "step": 3592 }, { "epoch": 0.24, "learning_rate": 1.779242198483698e-05, "loss": 0.9768, "step": 3593 }, { "epoch": 0.24, "learning_rate": 1.7791070738064182e-05, "loss": 1.0159, "step": 3594 }, { "epoch": 0.24, "learning_rate": 1.778971912921603e-05, "loss": 1.0272, "step": 3595 }, { "epoch": 0.24, "learning_rate": 1.778836715835534e-05, "loss": 1.0318, "step": 3596 }, { "epoch": 0.24, "learning_rate": 1.778701482554494e-05, "loss": 1.0274, "step": 3597 }, { "epoch": 0.24, "learning_rate": 1.7785662130847685e-05, "loss": 0.9901, "step": 3598 }, { "epoch": 0.24, "learning_rate": 1.778430907432643e-05, "loss": 1.0225, "step": 3599 }, { "epoch": 0.24, "learning_rate": 1.7782955656044056e-05, "loss": 1.032, "step": 3600 }, { "epoch": 0.24, "learning_rate": 1.778160187606347e-05, "loss": 1.0493, "step": 3601 }, { "epoch": 0.24, "learning_rate": 1.7780247734447572e-05, "loss": 1.0104, "step": 3602 }, { "epoch": 0.24, "learning_rate": 1.7778893231259305e-05, "loss": 1.0223, "step": 3603 }, { "epoch": 0.24, "learning_rate": 1.777753836656161e-05, "loss": 1.0431, "step": 3604 }, { "epoch": 0.24, "learning_rate": 1.777618314041746e-05, "loss": 1.0274, "step": 3605 }, { "epoch": 0.24, "learning_rate": 1.7774827552889825e-05, "loss": 0.9814, "step": 3606 }, { "epoch": 0.24, "learning_rate": 1.7773471604041715e-05, "loss": 0.9896, "step": 3607 }, { "epoch": 0.24, "learning_rate": 1.7772115293936137e-05, "loss": 0.9976, "step": 3608 }, { "epoch": 0.24, "learning_rate": 1.7770758622636126e-05, "loss": 1.0012, "step": 3609 }, { "epoch": 0.24, "learning_rate": 1.776940159020473e-05, "loss": 1.0475, "step": 3610 }, { "epoch": 0.24, "learning_rate": 1.7768044196705015e-05, "loss": 1.0141, "step": 3611 }, { "epoch": 0.24, "learning_rate": 1.7766686442200064e-05, "loss": 1.0144, "step": 3612 }, { "epoch": 0.24, "learning_rate": 1.7765328326752975e-05, "loss": 0.9982, "step": 3613 }, { "epoch": 0.24, "learning_rate": 1.7763969850426862e-05, "loss": 1.0229, "step": 3614 }, { "epoch": 0.24, "learning_rate": 1.7762611013284867e-05, "loss": 1.0503, "step": 3615 }, { "epoch": 0.24, "learning_rate": 1.7761251815390125e-05, "loss": 1.0001, "step": 3616 }, { "epoch": 0.24, "learning_rate": 1.7759892256805814e-05, "loss": 0.9794, "step": 3617 }, { "epoch": 0.24, "learning_rate": 1.775853233759511e-05, "loss": 1.0119, "step": 3618 }, { "epoch": 0.24, "learning_rate": 1.7757172057821213e-05, "loss": 0.9955, "step": 3619 }, { "epoch": 0.24, "learning_rate": 1.7755811417547345e-05, "loss": 1.0695, "step": 3620 }, { "epoch": 0.24, "learning_rate": 1.7754450416836735e-05, "loss": 1.0058, "step": 3621 }, { "epoch": 0.24, "learning_rate": 1.7753089055752635e-05, "loss": 0.9797, "step": 3622 }, { "epoch": 0.24, "learning_rate": 1.775172733435831e-05, "loss": 1.0389, "step": 3623 }, { "epoch": 0.24, "learning_rate": 1.775036525271704e-05, "loss": 0.9827, "step": 3624 }, { "epoch": 0.24, "learning_rate": 1.7749002810892133e-05, "loss": 0.9973, "step": 3625 }, { "epoch": 0.24, "learning_rate": 1.77476400089469e-05, "loss": 1.0163, "step": 3626 }, { "epoch": 0.24, "learning_rate": 1.7746276846944677e-05, "loss": 1.0257, "step": 3627 }, { "epoch": 0.24, "learning_rate": 1.7744913324948814e-05, "loss": 0.9941, "step": 3628 }, { "epoch": 0.24, "learning_rate": 1.7743549443022678e-05, "loss": 1.0443, "step": 3629 }, { "epoch": 0.24, "learning_rate": 1.7742185201229654e-05, "loss": 0.998, "step": 3630 }, { "epoch": 0.24, "learning_rate": 1.774082059963314e-05, "loss": 0.9877, "step": 3631 }, { "epoch": 0.24, "learning_rate": 1.7739455638296553e-05, "loss": 0.9667, "step": 3632 }, { "epoch": 0.24, "learning_rate": 1.7738090317283334e-05, "loss": 1.016, "step": 3633 }, { "epoch": 0.24, "learning_rate": 1.7736724636656927e-05, "loss": 1.0111, "step": 3634 }, { "epoch": 0.24, "learning_rate": 1.77353585964808e-05, "loss": 1.0138, "step": 3635 }, { "epoch": 0.24, "learning_rate": 1.7733992196818434e-05, "loss": 1.047, "step": 3636 }, { "epoch": 0.24, "learning_rate": 1.7732625437733338e-05, "loss": 0.9906, "step": 3637 }, { "epoch": 0.24, "learning_rate": 1.7731258319289028e-05, "loss": 1.0202, "step": 3638 }, { "epoch": 0.24, "learning_rate": 1.7729890841549033e-05, "loss": 0.9872, "step": 3639 }, { "epoch": 0.24, "learning_rate": 1.7728523004576903e-05, "loss": 1.0111, "step": 3640 }, { "epoch": 0.24, "learning_rate": 1.7727154808436213e-05, "loss": 1.0434, "step": 3641 }, { "epoch": 0.24, "learning_rate": 1.7725786253190546e-05, "loss": 1.0095, "step": 3642 }, { "epoch": 0.24, "learning_rate": 1.7724417338903495e-05, "loss": 1.0295, "step": 3643 }, { "epoch": 0.24, "learning_rate": 1.7723048065638688e-05, "loss": 1.085, "step": 3644 }, { "epoch": 0.24, "learning_rate": 1.7721678433459753e-05, "loss": 0.992, "step": 3645 }, { "epoch": 0.24, "learning_rate": 1.7720308442430342e-05, "loss": 0.9985, "step": 3646 }, { "epoch": 0.24, "learning_rate": 1.7718938092614126e-05, "loss": 1.0427, "step": 3647 }, { "epoch": 0.24, "learning_rate": 1.771756738407479e-05, "loss": 1.0311, "step": 3648 }, { "epoch": 0.24, "learning_rate": 1.7716196316876026e-05, "loss": 1.0218, "step": 3649 }, { "epoch": 0.24, "learning_rate": 1.7714824891081558e-05, "loss": 1.0535, "step": 3650 }, { "epoch": 0.24, "learning_rate": 1.7713453106755124e-05, "loss": 1.0484, "step": 3651 }, { "epoch": 0.24, "learning_rate": 1.771208096396047e-05, "loss": 1.0204, "step": 3652 }, { "epoch": 0.24, "learning_rate": 1.7710708462761365e-05, "loss": 0.9929, "step": 3653 }, { "epoch": 0.24, "learning_rate": 1.7709335603221597e-05, "loss": 1.0122, "step": 3654 }, { "epoch": 0.24, "learning_rate": 1.7707962385404958e-05, "loss": 1.0444, "step": 3655 }, { "epoch": 0.24, "learning_rate": 1.7706588809375277e-05, "loss": 0.9883, "step": 3656 }, { "epoch": 0.24, "learning_rate": 1.770521487519638e-05, "loss": 1.007, "step": 3657 }, { "epoch": 0.24, "learning_rate": 1.770384058293212e-05, "loss": 1.0268, "step": 3658 }, { "epoch": 0.24, "learning_rate": 1.7702465932646365e-05, "loss": 1.0104, "step": 3659 }, { "epoch": 0.24, "learning_rate": 1.7701090924403e-05, "loss": 1.0088, "step": 3660 }, { "epoch": 0.24, "learning_rate": 1.7699715558265928e-05, "loss": 1.029, "step": 3661 }, { "epoch": 0.24, "learning_rate": 1.7698339834299064e-05, "loss": 0.9792, "step": 3662 }, { "epoch": 0.24, "learning_rate": 1.769696375256634e-05, "loss": 0.9837, "step": 3663 }, { "epoch": 0.24, "learning_rate": 1.769558731313171e-05, "loss": 1.0223, "step": 3664 }, { "epoch": 0.24, "learning_rate": 1.769421051605914e-05, "loss": 1.0367, "step": 3665 }, { "epoch": 0.24, "learning_rate": 1.769283336141262e-05, "loss": 1.0063, "step": 3666 }, { "epoch": 0.24, "learning_rate": 1.769145584925614e-05, "loss": 1.0553, "step": 3667 }, { "epoch": 0.24, "learning_rate": 1.7690077979653725e-05, "loss": 0.9915, "step": 3668 }, { "epoch": 0.24, "learning_rate": 1.7688699752669407e-05, "loss": 1.0433, "step": 3669 }, { "epoch": 0.24, "learning_rate": 1.7687321168367233e-05, "loss": 0.9885, "step": 3670 }, { "epoch": 0.24, "learning_rate": 1.7685942226811274e-05, "loss": 1.0544, "step": 3671 }, { "epoch": 0.24, "learning_rate": 1.7684562928065614e-05, "loss": 1.0237, "step": 3672 }, { "epoch": 0.24, "learning_rate": 1.768318327219435e-05, "loss": 1.0292, "step": 3673 }, { "epoch": 0.24, "learning_rate": 1.7681803259261605e-05, "loss": 1.0201, "step": 3674 }, { "epoch": 0.24, "learning_rate": 1.7680422889331507e-05, "loss": 1.0072, "step": 3675 }, { "epoch": 0.24, "learning_rate": 1.7679042162468207e-05, "loss": 1.0486, "step": 3676 }, { "epoch": 0.24, "learning_rate": 1.767766107873587e-05, "loss": 1.0506, "step": 3677 }, { "epoch": 0.24, "learning_rate": 1.7676279638198687e-05, "loss": 1.023, "step": 3678 }, { "epoch": 0.24, "learning_rate": 1.767489784092085e-05, "loss": 0.9762, "step": 3679 }, { "epoch": 0.24, "learning_rate": 1.7673515686966577e-05, "loss": 1.003, "step": 3680 }, { "epoch": 0.25, "learning_rate": 1.7672133176400103e-05, "loss": 1.0077, "step": 3681 }, { "epoch": 0.25, "learning_rate": 1.7670750309285675e-05, "loss": 1.0097, "step": 3682 }, { "epoch": 0.25, "learning_rate": 1.7669367085687564e-05, "loss": 1.0355, "step": 3683 }, { "epoch": 0.25, "learning_rate": 1.7667983505670044e-05, "loss": 0.9838, "step": 3684 }, { "epoch": 0.25, "learning_rate": 1.7666599569297425e-05, "loss": 1.004, "step": 3685 }, { "epoch": 0.25, "learning_rate": 1.7665215276634014e-05, "loss": 1.0213, "step": 3686 }, { "epoch": 0.25, "learning_rate": 1.766383062774415e-05, "loss": 1.0502, "step": 3687 }, { "epoch": 0.25, "learning_rate": 1.7662445622692177e-05, "loss": 0.9954, "step": 3688 }, { "epoch": 0.25, "learning_rate": 1.7661060261542463e-05, "loss": 1.0638, "step": 3689 }, { "epoch": 0.25, "learning_rate": 1.7659674544359386e-05, "loss": 1.0491, "step": 3690 }, { "epoch": 0.25, "learning_rate": 1.7658288471207353e-05, "loss": 1.017, "step": 3691 }, { "epoch": 0.25, "learning_rate": 1.765690204215077e-05, "loss": 0.9977, "step": 3692 }, { "epoch": 0.25, "learning_rate": 1.7655515257254076e-05, "loss": 1.0078, "step": 3693 }, { "epoch": 0.25, "learning_rate": 1.7654128116581716e-05, "loss": 1.0222, "step": 3694 }, { "epoch": 0.25, "learning_rate": 1.7652740620198155e-05, "loss": 1.0224, "step": 3695 }, { "epoch": 0.25, "learning_rate": 1.765135276816787e-05, "loss": 0.979, "step": 3696 }, { "epoch": 0.25, "learning_rate": 1.7649964560555366e-05, "loss": 1.0292, "step": 3697 }, { "epoch": 0.25, "learning_rate": 1.7648575997425155e-05, "loss": 1.0441, "step": 3698 }, { "epoch": 0.25, "learning_rate": 1.7647187078841766e-05, "loss": 1.0755, "step": 3699 }, { "epoch": 0.25, "learning_rate": 1.7645797804869746e-05, "loss": 1.0022, "step": 3700 }, { "epoch": 0.25, "learning_rate": 1.7644408175573662e-05, "loss": 1.0284, "step": 3701 }, { "epoch": 0.25, "learning_rate": 1.7643018191018093e-05, "loss": 1.0078, "step": 3702 }, { "epoch": 0.25, "learning_rate": 1.7641627851267635e-05, "loss": 1.0151, "step": 3703 }, { "epoch": 0.25, "learning_rate": 1.7640237156386902e-05, "loss": 1.038, "step": 3704 }, { "epoch": 0.25, "learning_rate": 1.7638846106440522e-05, "loss": 0.9895, "step": 3705 }, { "epoch": 0.25, "learning_rate": 1.7637454701493147e-05, "loss": 1.0241, "step": 3706 }, { "epoch": 0.25, "learning_rate": 1.7636062941609436e-05, "loss": 1.0363, "step": 3707 }, { "epoch": 0.25, "learning_rate": 1.7634670826854065e-05, "loss": 1.0056, "step": 3708 }, { "epoch": 0.25, "learning_rate": 1.7633278357291737e-05, "loss": 1.0346, "step": 3709 }, { "epoch": 0.25, "learning_rate": 1.763188553298716e-05, "loss": 0.9707, "step": 3710 }, { "epoch": 0.25, "learning_rate": 1.7630492354005065e-05, "loss": 1.0346, "step": 3711 }, { "epoch": 0.25, "learning_rate": 1.7629098820410195e-05, "loss": 0.9875, "step": 3712 }, { "epoch": 0.25, "learning_rate": 1.7627704932267316e-05, "loss": 1.0568, "step": 3713 }, { "epoch": 0.25, "learning_rate": 1.7626310689641203e-05, "loss": 0.9896, "step": 3714 }, { "epoch": 0.25, "learning_rate": 1.7624916092596647e-05, "loss": 1.0219, "step": 3715 }, { "epoch": 0.25, "learning_rate": 1.762352114119847e-05, "loss": 0.9991, "step": 3716 }, { "epoch": 0.25, "learning_rate": 1.7622125835511487e-05, "loss": 0.9961, "step": 3717 }, { "epoch": 0.25, "learning_rate": 1.7620730175600553e-05, "loss": 1.0125, "step": 3718 }, { "epoch": 0.25, "learning_rate": 1.7619334161530524e-05, "loss": 1.0496, "step": 3719 }, { "epoch": 0.25, "learning_rate": 1.7617937793366275e-05, "loss": 1.0194, "step": 3720 }, { "epoch": 0.25, "learning_rate": 1.7616541071172704e-05, "loss": 1.029, "step": 3721 }, { "epoch": 0.25, "learning_rate": 1.7615143995014717e-05, "loss": 0.9822, "step": 3722 }, { "epoch": 0.25, "learning_rate": 1.7613746564957247e-05, "loss": 1.0536, "step": 3723 }, { "epoch": 0.25, "learning_rate": 1.7612348781065227e-05, "loss": 0.9865, "step": 3724 }, { "epoch": 0.25, "learning_rate": 1.7610950643403622e-05, "loss": 1.0067, "step": 3725 }, { "epoch": 0.25, "learning_rate": 1.7609552152037413e-05, "loss": 1.0421, "step": 3726 }, { "epoch": 0.25, "learning_rate": 1.7608153307031585e-05, "loss": 1.0097, "step": 3727 }, { "epoch": 0.25, "learning_rate": 1.7606754108451144e-05, "loss": 1.0201, "step": 3728 }, { "epoch": 0.25, "learning_rate": 1.760535455636112e-05, "loss": 0.9977, "step": 3729 }, { "epoch": 0.25, "learning_rate": 1.760395465082656e-05, "loss": 0.9843, "step": 3730 }, { "epoch": 0.25, "learning_rate": 1.7602554391912513e-05, "loss": 1.0536, "step": 3731 }, { "epoch": 0.25, "learning_rate": 1.7601153779684055e-05, "loss": 1.0107, "step": 3732 }, { "epoch": 0.25, "learning_rate": 1.7599752814206283e-05, "loss": 1.0306, "step": 3733 }, { "epoch": 0.25, "learning_rate": 1.7598351495544293e-05, "loss": 1.0084, "step": 3734 }, { "epoch": 0.25, "learning_rate": 1.7596949823763222e-05, "loss": 1.0157, "step": 3735 }, { "epoch": 0.25, "learning_rate": 1.7595547798928198e-05, "loss": 0.9864, "step": 3736 }, { "epoch": 0.25, "learning_rate": 1.7594145421104386e-05, "loss": 1.006, "step": 3737 }, { "epoch": 0.25, "learning_rate": 1.7592742690356955e-05, "loss": 1.0315, "step": 3738 }, { "epoch": 0.25, "learning_rate": 1.7591339606751095e-05, "loss": 1.0311, "step": 3739 }, { "epoch": 0.25, "learning_rate": 1.758993617035201e-05, "loss": 0.9998, "step": 3740 }, { "epoch": 0.25, "learning_rate": 1.7588532381224925e-05, "loss": 1.0098, "step": 3741 }, { "epoch": 0.25, "learning_rate": 1.758712823943508e-05, "loss": 1.0, "step": 3742 }, { "epoch": 0.25, "learning_rate": 1.7585723745047724e-05, "loss": 1.0128, "step": 3743 }, { "epoch": 0.25, "learning_rate": 1.758431889812813e-05, "loss": 0.991, "step": 3744 }, { "epoch": 0.25, "learning_rate": 1.7582913698741588e-05, "loss": 1.0476, "step": 3745 }, { "epoch": 0.25, "learning_rate": 1.7581508146953402e-05, "loss": 1.0352, "step": 3746 }, { "epoch": 0.25, "learning_rate": 1.758010224282889e-05, "loss": 1.0273, "step": 3747 }, { "epoch": 0.25, "learning_rate": 1.757869598643339e-05, "loss": 1.0287, "step": 3748 }, { "epoch": 0.25, "learning_rate": 1.7577289377832255e-05, "loss": 1.0207, "step": 3749 }, { "epoch": 0.25, "learning_rate": 1.7575882417090854e-05, "loss": 0.9947, "step": 3750 }, { "epoch": 0.25, "learning_rate": 1.7574475104274575e-05, "loss": 1.0202, "step": 3751 }, { "epoch": 0.25, "learning_rate": 1.757306743944882e-05, "loss": 0.9842, "step": 3752 }, { "epoch": 0.25, "learning_rate": 1.7571659422679003e-05, "loss": 1.0545, "step": 3753 }, { "epoch": 0.25, "learning_rate": 1.7570251054030564e-05, "loss": 0.9995, "step": 3754 }, { "epoch": 0.25, "learning_rate": 1.7568842333568952e-05, "loss": 1.0445, "step": 3755 }, { "epoch": 0.25, "learning_rate": 1.7567433261359633e-05, "loss": 1.0246, "step": 3756 }, { "epoch": 0.25, "learning_rate": 1.7566023837468096e-05, "loss": 0.9927, "step": 3757 }, { "epoch": 0.25, "learning_rate": 1.756461406195984e-05, "loss": 1.0297, "step": 3758 }, { "epoch": 0.25, "learning_rate": 1.7563203934900377e-05, "loss": 0.9755, "step": 3759 }, { "epoch": 0.25, "learning_rate": 1.7561793456355242e-05, "loss": 1.0005, "step": 3760 }, { "epoch": 0.25, "learning_rate": 1.7560382626389988e-05, "loss": 1.0209, "step": 3761 }, { "epoch": 0.25, "learning_rate": 1.7558971445070175e-05, "loss": 1.0069, "step": 3762 }, { "epoch": 0.25, "learning_rate": 1.755755991246139e-05, "loss": 1.0164, "step": 3763 }, { "epoch": 0.25, "learning_rate": 1.755614802862923e-05, "loss": 1.0041, "step": 3764 }, { "epoch": 0.25, "learning_rate": 1.755473579363931e-05, "loss": 1.0221, "step": 3765 }, { "epoch": 0.25, "learning_rate": 1.7553323207557258e-05, "loss": 0.9718, "step": 3766 }, { "epoch": 0.25, "learning_rate": 1.755191027044872e-05, "loss": 1.0011, "step": 3767 }, { "epoch": 0.25, "learning_rate": 1.755049698237937e-05, "loss": 1.0038, "step": 3768 }, { "epoch": 0.25, "learning_rate": 1.7549083343414873e-05, "loss": 1.0428, "step": 3769 }, { "epoch": 0.25, "learning_rate": 1.7547669353620938e-05, "loss": 1.0049, "step": 3770 }, { "epoch": 0.25, "learning_rate": 1.7546255013063273e-05, "loss": 1.0215, "step": 3771 }, { "epoch": 0.25, "learning_rate": 1.7544840321807605e-05, "loss": 1.0322, "step": 3772 }, { "epoch": 0.25, "learning_rate": 1.754342527991968e-05, "loss": 0.9991, "step": 3773 }, { "epoch": 0.25, "learning_rate": 1.754200988746526e-05, "loss": 1.0069, "step": 3774 }, { "epoch": 0.25, "learning_rate": 1.7540594144510122e-05, "loss": 0.993, "step": 3775 }, { "epoch": 0.25, "learning_rate": 1.753917805112006e-05, "loss": 1.0432, "step": 3776 }, { "epoch": 0.25, "learning_rate": 1.753776160736089e-05, "loss": 1.0599, "step": 3777 }, { "epoch": 0.25, "learning_rate": 1.7536344813298427e-05, "loss": 0.9979, "step": 3778 }, { "epoch": 0.25, "learning_rate": 1.7534927668998522e-05, "loss": 1.0284, "step": 3779 }, { "epoch": 0.25, "learning_rate": 1.753351017452703e-05, "loss": 1.0291, "step": 3780 }, { "epoch": 0.25, "learning_rate": 1.7532092329949832e-05, "loss": 1.0159, "step": 3781 }, { "epoch": 0.25, "learning_rate": 1.7530674135332816e-05, "loss": 0.9776, "step": 3782 }, { "epoch": 0.25, "learning_rate": 1.752925559074189e-05, "loss": 1.0648, "step": 3783 }, { "epoch": 0.25, "learning_rate": 1.752783669624298e-05, "loss": 1.0498, "step": 3784 }, { "epoch": 0.25, "learning_rate": 1.7526417451902024e-05, "loss": 1.0306, "step": 3785 }, { "epoch": 0.25, "learning_rate": 1.7524997857784978e-05, "loss": 1.0148, "step": 3786 }, { "epoch": 0.25, "learning_rate": 1.7523577913957813e-05, "loss": 1.0069, "step": 3787 }, { "epoch": 0.25, "learning_rate": 1.7522157620486527e-05, "loss": 0.9661, "step": 3788 }, { "epoch": 0.25, "learning_rate": 1.752073697743712e-05, "loss": 1.033, "step": 3789 }, { "epoch": 0.25, "learning_rate": 1.7519315984875614e-05, "loss": 0.9835, "step": 3790 }, { "epoch": 0.25, "learning_rate": 1.7517894642868046e-05, "loss": 1.0244, "step": 3791 }, { "epoch": 0.25, "learning_rate": 1.7516472951480472e-05, "loss": 1.0124, "step": 3792 }, { "epoch": 0.25, "learning_rate": 1.751505091077896e-05, "loss": 1.0503, "step": 3793 }, { "epoch": 0.25, "learning_rate": 1.75136285208296e-05, "loss": 1.0263, "step": 3794 }, { "epoch": 0.25, "learning_rate": 1.7512205781698493e-05, "loss": 1.0717, "step": 3795 }, { "epoch": 0.25, "learning_rate": 1.7510782693451758e-05, "loss": 1.0161, "step": 3796 }, { "epoch": 0.25, "learning_rate": 1.750935925615553e-05, "loss": 1.0416, "step": 3797 }, { "epoch": 0.25, "learning_rate": 1.750793546987596e-05, "loss": 0.9635, "step": 3798 }, { "epoch": 0.25, "learning_rate": 1.7506511334679217e-05, "loss": 1.0494, "step": 3799 }, { "epoch": 0.25, "learning_rate": 1.7505086850631487e-05, "loss": 0.9852, "step": 3800 }, { "epoch": 0.25, "learning_rate": 1.7503662017798967e-05, "loss": 1.0058, "step": 3801 }, { "epoch": 0.25, "learning_rate": 1.750223683624788e-05, "loss": 1.0084, "step": 3802 }, { "epoch": 0.25, "learning_rate": 1.7500811306044442e-05, "loss": 1.0346, "step": 3803 }, { "epoch": 0.25, "learning_rate": 1.749938542725492e-05, "loss": 1.0384, "step": 3804 }, { "epoch": 0.25, "learning_rate": 1.7497959199945575e-05, "loss": 1.0607, "step": 3805 }, { "epoch": 0.25, "learning_rate": 1.7496532624182677e-05, "loss": 1.0193, "step": 3806 }, { "epoch": 0.25, "learning_rate": 1.749510570003254e-05, "loss": 0.9904, "step": 3807 }, { "epoch": 0.25, "learning_rate": 1.7493678427561464e-05, "loss": 1.0085, "step": 3808 }, { "epoch": 0.25, "learning_rate": 1.7492250806835788e-05, "loss": 1.0326, "step": 3809 }, { "epoch": 0.25, "learning_rate": 1.749082283792185e-05, "loss": 0.9709, "step": 3810 }, { "epoch": 0.25, "learning_rate": 1.7489394520886017e-05, "loss": 0.999, "step": 3811 }, { "epoch": 0.25, "learning_rate": 1.748796585579467e-05, "loss": 0.9893, "step": 3812 }, { "epoch": 0.25, "learning_rate": 1.74865368427142e-05, "loss": 0.9917, "step": 3813 }, { "epoch": 0.25, "learning_rate": 1.7485107481711014e-05, "loss": 0.9814, "step": 3814 }, { "epoch": 0.25, "learning_rate": 1.7483677772851544e-05, "loss": 1.0421, "step": 3815 }, { "epoch": 0.25, "learning_rate": 1.7482247716202232e-05, "loss": 1.0451, "step": 3816 }, { "epoch": 0.25, "learning_rate": 1.748081731182954e-05, "loss": 1.0309, "step": 3817 }, { "epoch": 0.25, "learning_rate": 1.7479386559799937e-05, "loss": 1.0249, "step": 3818 }, { "epoch": 0.25, "learning_rate": 1.7477955460179917e-05, "loss": 1.0221, "step": 3819 }, { "epoch": 0.25, "learning_rate": 1.747652401303599e-05, "loss": 1.0372, "step": 3820 }, { "epoch": 0.25, "learning_rate": 1.747509221843468e-05, "loss": 1.0338, "step": 3821 }, { "epoch": 0.25, "learning_rate": 1.7473660076442524e-05, "loss": 0.9692, "step": 3822 }, { "epoch": 0.25, "learning_rate": 1.7472227587126077e-05, "loss": 1.0348, "step": 3823 }, { "epoch": 0.25, "learning_rate": 1.7470794750551916e-05, "loss": 1.0011, "step": 3824 }, { "epoch": 0.25, "learning_rate": 1.7469361566786624e-05, "loss": 0.9935, "step": 3825 }, { "epoch": 0.25, "learning_rate": 1.746792803589681e-05, "loss": 1.0538, "step": 3826 }, { "epoch": 0.25, "learning_rate": 1.74664941579491e-05, "loss": 1.0339, "step": 3827 }, { "epoch": 0.25, "learning_rate": 1.7465059933010116e-05, "loss": 1.0146, "step": 3828 }, { "epoch": 0.25, "learning_rate": 1.746362536114652e-05, "loss": 1.0435, "step": 3829 }, { "epoch": 0.25, "learning_rate": 1.7462190442424983e-05, "loss": 0.9539, "step": 3830 }, { "epoch": 0.25, "learning_rate": 1.7460755176912183e-05, "loss": 1.0128, "step": 3831 }, { "epoch": 0.26, "learning_rate": 1.745931956467483e-05, "loss": 1.0148, "step": 3832 }, { "epoch": 0.26, "learning_rate": 1.7457883605779632e-05, "loss": 1.0441, "step": 3833 }, { "epoch": 0.26, "learning_rate": 1.745644730029333e-05, "loss": 1.036, "step": 3834 }, { "epoch": 0.26, "learning_rate": 1.745501064828267e-05, "loss": 1.0069, "step": 3835 }, { "epoch": 0.26, "learning_rate": 1.7453573649814415e-05, "loss": 0.9909, "step": 3836 }, { "epoch": 0.26, "learning_rate": 1.7452136304955355e-05, "loss": 1.0124, "step": 3837 }, { "epoch": 0.26, "learning_rate": 1.7450698613772284e-05, "loss": 0.9748, "step": 3838 }, { "epoch": 0.26, "learning_rate": 1.744926057633201e-05, "loss": 1.053, "step": 3839 }, { "epoch": 0.26, "learning_rate": 1.744782219270137e-05, "loss": 0.9987, "step": 3840 }, { "epoch": 0.26, "learning_rate": 1.7446383462947212e-05, "loss": 1.0169, "step": 3841 }, { "epoch": 0.26, "learning_rate": 1.7444944387136388e-05, "loss": 0.9976, "step": 3842 }, { "epoch": 0.26, "learning_rate": 1.744350496533579e-05, "loss": 1.0553, "step": 3843 }, { "epoch": 0.26, "learning_rate": 1.74420651976123e-05, "loss": 0.9849, "step": 3844 }, { "epoch": 0.26, "learning_rate": 1.7440625084032836e-05, "loss": 0.9881, "step": 3845 }, { "epoch": 0.26, "learning_rate": 1.7439184624664325e-05, "loss": 1.0059, "step": 3846 }, { "epoch": 0.26, "learning_rate": 1.7437743819573706e-05, "loss": 1.0037, "step": 3847 }, { "epoch": 0.26, "learning_rate": 1.7436302668827935e-05, "loss": 1.0086, "step": 3848 }, { "epoch": 0.26, "learning_rate": 1.7434861172493995e-05, "loss": 1.0097, "step": 3849 }, { "epoch": 0.26, "learning_rate": 1.743341933063887e-05, "loss": 0.9992, "step": 3850 }, { "epoch": 0.26, "learning_rate": 1.7431977143329568e-05, "loss": 1.0496, "step": 3851 }, { "epoch": 0.26, "learning_rate": 1.7430534610633118e-05, "loss": 1.0283, "step": 3852 }, { "epoch": 0.26, "learning_rate": 1.742909173261655e-05, "loss": 1.03, "step": 3853 }, { "epoch": 0.26, "learning_rate": 1.7427648509346926e-05, "loss": 0.9928, "step": 3854 }, { "epoch": 0.26, "learning_rate": 1.7426204940891313e-05, "loss": 1.0556, "step": 3855 }, { "epoch": 0.26, "learning_rate": 1.7424761027316795e-05, "loss": 0.9881, "step": 3856 }, { "epoch": 0.26, "learning_rate": 1.7423316768690487e-05, "loss": 0.9986, "step": 3857 }, { "epoch": 0.26, "learning_rate": 1.7421872165079498e-05, "loss": 0.9746, "step": 3858 }, { "epoch": 0.26, "learning_rate": 1.7420427216550967e-05, "loss": 1.021, "step": 3859 }, { "epoch": 0.26, "learning_rate": 1.7418981923172043e-05, "loss": 1.0164, "step": 3860 }, { "epoch": 0.26, "learning_rate": 1.7417536285009896e-05, "loss": 1.0068, "step": 3861 }, { "epoch": 0.26, "learning_rate": 1.7416090302131706e-05, "loss": 0.991, "step": 3862 }, { "epoch": 0.26, "learning_rate": 1.7414643974604676e-05, "loss": 1.0493, "step": 3863 }, { "epoch": 0.26, "learning_rate": 1.7413197302496018e-05, "loss": 0.9872, "step": 3864 }, { "epoch": 0.26, "learning_rate": 1.7411750285872967e-05, "loss": 0.9966, "step": 3865 }, { "epoch": 0.26, "learning_rate": 1.7410302924802766e-05, "loss": 0.98, "step": 3866 }, { "epoch": 0.26, "learning_rate": 1.7408855219352682e-05, "loss": 1.0282, "step": 3867 }, { "epoch": 0.26, "learning_rate": 1.7407407169589994e-05, "loss": 1.0112, "step": 3868 }, { "epoch": 0.26, "learning_rate": 1.7405958775581995e-05, "loss": 1.0196, "step": 3869 }, { "epoch": 0.26, "learning_rate": 1.7404510037396e-05, "loss": 0.9976, "step": 3870 }, { "epoch": 0.26, "learning_rate": 1.7403060955099332e-05, "loss": 0.9769, "step": 3871 }, { "epoch": 0.26, "learning_rate": 1.7401611528759338e-05, "loss": 0.9842, "step": 3872 }, { "epoch": 0.26, "learning_rate": 1.7400161758443377e-05, "loss": 1.0166, "step": 3873 }, { "epoch": 0.26, "learning_rate": 1.739871164421882e-05, "loss": 0.9713, "step": 3874 }, { "epoch": 0.26, "learning_rate": 1.739726118615306e-05, "loss": 0.9823, "step": 3875 }, { "epoch": 0.26, "learning_rate": 1.7395810384313513e-05, "loss": 1.0046, "step": 3876 }, { "epoch": 0.26, "learning_rate": 1.7394359238767595e-05, "loss": 1.0112, "step": 3877 }, { "epoch": 0.26, "learning_rate": 1.7392907749582742e-05, "loss": 0.9919, "step": 3878 }, { "epoch": 0.26, "learning_rate": 1.7391455916826414e-05, "loss": 1.0394, "step": 3879 }, { "epoch": 0.26, "learning_rate": 1.739000374056608e-05, "loss": 0.9646, "step": 3880 }, { "epoch": 0.26, "learning_rate": 1.738855122086923e-05, "loss": 1.0366, "step": 3881 }, { "epoch": 0.26, "learning_rate": 1.7387098357803362e-05, "loss": 1.0189, "step": 3882 }, { "epoch": 0.26, "learning_rate": 1.7385645151436e-05, "loss": 1.0396, "step": 3883 }, { "epoch": 0.26, "learning_rate": 1.738419160183468e-05, "loss": 0.9863, "step": 3884 }, { "epoch": 0.26, "learning_rate": 1.7382737709066947e-05, "loss": 1.0019, "step": 3885 }, { "epoch": 0.26, "learning_rate": 1.738128347320037e-05, "loss": 0.9704, "step": 3886 }, { "epoch": 0.26, "learning_rate": 1.7379828894302538e-05, "loss": 1.0237, "step": 3887 }, { "epoch": 0.26, "learning_rate": 1.737837397244104e-05, "loss": 1.0009, "step": 3888 }, { "epoch": 0.26, "learning_rate": 1.73769187076835e-05, "loss": 1.0054, "step": 3889 }, { "epoch": 0.26, "learning_rate": 1.737546310009754e-05, "loss": 1.0066, "step": 3890 }, { "epoch": 0.26, "learning_rate": 1.7374007149750816e-05, "loss": 1.0447, "step": 3891 }, { "epoch": 0.26, "learning_rate": 1.737255085671098e-05, "loss": 1.016, "step": 3892 }, { "epoch": 0.26, "learning_rate": 1.737109422104572e-05, "loss": 0.9893, "step": 3893 }, { "epoch": 0.26, "learning_rate": 1.7369637242822728e-05, "loss": 0.9733, "step": 3894 }, { "epoch": 0.26, "learning_rate": 1.7368179922109713e-05, "loss": 1.0254, "step": 3895 }, { "epoch": 0.26, "learning_rate": 1.7366722258974396e-05, "loss": 1.0413, "step": 3896 }, { "epoch": 0.26, "learning_rate": 1.7365264253484526e-05, "loss": 1.0182, "step": 3897 }, { "epoch": 0.26, "learning_rate": 1.7363805905707863e-05, "loss": 1.0488, "step": 3898 }, { "epoch": 0.26, "learning_rate": 1.7362347215712173e-05, "loss": 1.068, "step": 3899 }, { "epoch": 0.26, "learning_rate": 1.736088818356525e-05, "loss": 1.0165, "step": 3900 }, { "epoch": 0.26, "learning_rate": 1.73594288093349e-05, "loss": 1.0401, "step": 3901 }, { "epoch": 0.26, "learning_rate": 1.7357969093088945e-05, "loss": 0.9935, "step": 3902 }, { "epoch": 0.26, "learning_rate": 1.7356509034895227e-05, "loss": 0.9897, "step": 3903 }, { "epoch": 0.26, "learning_rate": 1.735504863482159e-05, "loss": 0.9754, "step": 3904 }, { "epoch": 0.26, "learning_rate": 1.735358789293591e-05, "loss": 1.0068, "step": 3905 }, { "epoch": 0.26, "learning_rate": 1.7352126809306066e-05, "loss": 1.0254, "step": 3906 }, { "epoch": 0.26, "learning_rate": 1.7350665383999968e-05, "loss": 1.0226, "step": 3907 }, { "epoch": 0.26, "learning_rate": 1.7349203617085523e-05, "loss": 1.0223, "step": 3908 }, { "epoch": 0.26, "learning_rate": 1.7347741508630673e-05, "loss": 1.0733, "step": 3909 }, { "epoch": 0.26, "learning_rate": 1.7346279058703364e-05, "loss": 1.0088, "step": 3910 }, { "epoch": 0.26, "learning_rate": 1.7344816267371558e-05, "loss": 1.0124, "step": 3911 }, { "epoch": 0.26, "learning_rate": 1.7343353134703235e-05, "loss": 0.9761, "step": 3912 }, { "epoch": 0.26, "learning_rate": 1.7341889660766395e-05, "loss": 1.0275, "step": 3913 }, { "epoch": 0.26, "learning_rate": 1.734042584562905e-05, "loss": 1.0103, "step": 3914 }, { "epoch": 0.26, "learning_rate": 1.7338961689359222e-05, "loss": 1.0805, "step": 3915 }, { "epoch": 0.26, "learning_rate": 1.7337497192024964e-05, "loss": 1.0182, "step": 3916 }, { "epoch": 0.26, "learning_rate": 1.733603235369433e-05, "loss": 1.0492, "step": 3917 }, { "epoch": 0.26, "learning_rate": 1.7334567174435395e-05, "loss": 0.9722, "step": 3918 }, { "epoch": 0.26, "learning_rate": 1.7333101654316253e-05, "loss": 0.9998, "step": 3919 }, { "epoch": 0.26, "learning_rate": 1.7331635793405012e-05, "loss": 1.0414, "step": 3920 }, { "epoch": 0.26, "learning_rate": 1.733016959176979e-05, "loss": 1.0236, "step": 3921 }, { "epoch": 0.26, "learning_rate": 1.7328703049478736e-05, "loss": 0.9819, "step": 3922 }, { "epoch": 0.26, "learning_rate": 1.732723616659999e-05, "loss": 0.9875, "step": 3923 }, { "epoch": 0.26, "learning_rate": 1.732576894320174e-05, "loss": 1.0197, "step": 3924 }, { "epoch": 0.26, "learning_rate": 1.7324301379352158e-05, "loss": 1.025, "step": 3925 }, { "epoch": 0.26, "learning_rate": 1.7322833475119455e-05, "loss": 1.0128, "step": 3926 }, { "epoch": 0.26, "learning_rate": 1.732136523057184e-05, "loss": 1.0095, "step": 3927 }, { "epoch": 0.26, "learning_rate": 1.7319896645777555e-05, "loss": 1.0312, "step": 3928 }, { "epoch": 0.26, "learning_rate": 1.7318427720804852e-05, "loss": 1.0542, "step": 3929 }, { "epoch": 0.26, "learning_rate": 1.731695845572199e-05, "loss": 0.9915, "step": 3930 }, { "epoch": 0.26, "learning_rate": 1.731548885059725e-05, "loss": 0.9976, "step": 3931 }, { "epoch": 0.26, "learning_rate": 1.7314018905498932e-05, "loss": 0.9755, "step": 3932 }, { "epoch": 0.26, "learning_rate": 1.731254862049535e-05, "loss": 1.0176, "step": 3933 }, { "epoch": 0.26, "learning_rate": 1.731107799565483e-05, "loss": 0.998, "step": 3934 }, { "epoch": 0.26, "learning_rate": 1.730960703104572e-05, "loss": 0.9822, "step": 3935 }, { "epoch": 0.26, "learning_rate": 1.7308135726736376e-05, "loss": 0.9639, "step": 3936 }, { "epoch": 0.26, "learning_rate": 1.7306664082795173e-05, "loss": 1.038, "step": 3937 }, { "epoch": 0.26, "learning_rate": 1.730519209929051e-05, "loss": 1.0082, "step": 3938 }, { "epoch": 0.26, "learning_rate": 1.7303719776290792e-05, "loss": 1.0394, "step": 3939 }, { "epoch": 0.26, "learning_rate": 1.7302247113864434e-05, "loss": 1.0621, "step": 3940 }, { "epoch": 0.26, "learning_rate": 1.730077411207989e-05, "loss": 1.0107, "step": 3941 }, { "epoch": 0.26, "learning_rate": 1.7299300771005603e-05, "loss": 0.9914, "step": 3942 }, { "epoch": 0.26, "learning_rate": 1.729782709071005e-05, "loss": 1.0077, "step": 3943 }, { "epoch": 0.26, "learning_rate": 1.7296353071261713e-05, "loss": 0.9787, "step": 3944 }, { "epoch": 0.26, "learning_rate": 1.7294878712729096e-05, "loss": 1.03, "step": 3945 }, { "epoch": 0.26, "learning_rate": 1.729340401518072e-05, "loss": 1.0118, "step": 3946 }, { "epoch": 0.26, "learning_rate": 1.7291928978685116e-05, "loss": 1.0659, "step": 3947 }, { "epoch": 0.26, "learning_rate": 1.7290453603310836e-05, "loss": 1.0198, "step": 3948 }, { "epoch": 0.26, "learning_rate": 1.7288977889126443e-05, "loss": 1.0073, "step": 3949 }, { "epoch": 0.26, "learning_rate": 1.7287501836200517e-05, "loss": 1.0322, "step": 3950 }, { "epoch": 0.26, "learning_rate": 1.7286025444601657e-05, "loss": 1.0136, "step": 3951 }, { "epoch": 0.26, "learning_rate": 1.7284548714398472e-05, "loss": 0.9969, "step": 3952 }, { "epoch": 0.26, "learning_rate": 1.7283071645659597e-05, "loss": 0.9955, "step": 3953 }, { "epoch": 0.26, "learning_rate": 1.728159423845367e-05, "loss": 1.0012, "step": 3954 }, { "epoch": 0.26, "learning_rate": 1.7280116492849354e-05, "loss": 1.0297, "step": 3955 }, { "epoch": 0.26, "learning_rate": 1.727863840891532e-05, "loss": 1.0708, "step": 3956 }, { "epoch": 0.26, "learning_rate": 1.727715998672027e-05, "loss": 0.9902, "step": 3957 }, { "epoch": 0.26, "learning_rate": 1.7275681226332898e-05, "loss": 0.995, "step": 3958 }, { "epoch": 0.26, "learning_rate": 1.7274202127821934e-05, "loss": 1.0283, "step": 3959 }, { "epoch": 0.26, "learning_rate": 1.727272269125611e-05, "loss": 0.9803, "step": 3960 }, { "epoch": 0.26, "learning_rate": 1.727124291670419e-05, "loss": 1.0132, "step": 3961 }, { "epoch": 0.26, "learning_rate": 1.7269762804234932e-05, "loss": 1.0429, "step": 3962 }, { "epoch": 0.26, "learning_rate": 1.7268282353917128e-05, "loss": 1.0093, "step": 3963 }, { "epoch": 0.26, "learning_rate": 1.7266801565819582e-05, "loss": 1.0322, "step": 3964 }, { "epoch": 0.26, "learning_rate": 1.7265320440011106e-05, "loss": 0.9522, "step": 3965 }, { "epoch": 0.26, "learning_rate": 1.7263838976560534e-05, "loss": 0.9953, "step": 3966 }, { "epoch": 0.26, "learning_rate": 1.726235717553671e-05, "loss": 0.9893, "step": 3967 }, { "epoch": 0.26, "learning_rate": 1.726087503700851e-05, "loss": 1.0321, "step": 3968 }, { "epoch": 0.26, "learning_rate": 1.7259392561044803e-05, "loss": 1.0131, "step": 3969 }, { "epoch": 0.26, "learning_rate": 1.7257909747714486e-05, "loss": 1.0305, "step": 3970 }, { "epoch": 0.26, "learning_rate": 1.7256426597086468e-05, "loss": 0.9874, "step": 3971 }, { "epoch": 0.26, "learning_rate": 1.7254943109229683e-05, "loss": 0.991, "step": 3972 }, { "epoch": 0.26, "learning_rate": 1.7253459284213067e-05, "loss": 1.0659, "step": 3973 }, { "epoch": 0.26, "learning_rate": 1.7251975122105578e-05, "loss": 1.0275, "step": 3974 }, { "epoch": 0.26, "learning_rate": 1.7250490622976195e-05, "loss": 0.991, "step": 3975 }, { "epoch": 0.26, "learning_rate": 1.7249005786893903e-05, "loss": 0.9912, "step": 3976 }, { "epoch": 0.26, "learning_rate": 1.7247520613927707e-05, "loss": 1.0346, "step": 3977 }, { "epoch": 0.26, "learning_rate": 1.7246035104146627e-05, "loss": 0.9646, "step": 3978 }, { "epoch": 0.26, "learning_rate": 1.7244549257619702e-05, "loss": 1.0229, "step": 3979 }, { "epoch": 0.26, "learning_rate": 1.724306307441598e-05, "loss": 0.9934, "step": 3980 }, { "epoch": 0.26, "learning_rate": 1.7241576554604535e-05, "loss": 1.0175, "step": 3981 }, { "epoch": 0.27, "learning_rate": 1.7240089698254444e-05, "loss": 1.0291, "step": 3982 }, { "epoch": 0.27, "learning_rate": 1.723860250543481e-05, "loss": 0.9666, "step": 3983 }, { "epoch": 0.27, "learning_rate": 1.7237114976214745e-05, "loss": 1.0167, "step": 3984 }, { "epoch": 0.27, "learning_rate": 1.723562711066338e-05, "loss": 1.043, "step": 3985 }, { "epoch": 0.27, "learning_rate": 1.7234138908849858e-05, "loss": 0.9854, "step": 3986 }, { "epoch": 0.27, "learning_rate": 1.7232650370843345e-05, "loss": 1.0156, "step": 3987 }, { "epoch": 0.27, "learning_rate": 1.7231161496713017e-05, "loss": 1.0205, "step": 3988 }, { "epoch": 0.27, "learning_rate": 1.7229672286528063e-05, "loss": 1.0099, "step": 3989 }, { "epoch": 0.27, "learning_rate": 1.7228182740357695e-05, "loss": 1.0216, "step": 3990 }, { "epoch": 0.27, "learning_rate": 1.7226692858271133e-05, "loss": 0.9808, "step": 3991 }, { "epoch": 0.27, "learning_rate": 1.7225202640337624e-05, "loss": 1.03, "step": 3992 }, { "epoch": 0.27, "learning_rate": 1.7223712086626417e-05, "loss": 1.0069, "step": 3993 }, { "epoch": 0.27, "learning_rate": 1.722222119720678e-05, "loss": 0.9952, "step": 3994 }, { "epoch": 0.27, "learning_rate": 1.722072997214801e-05, "loss": 0.9976, "step": 3995 }, { "epoch": 0.27, "learning_rate": 1.7219238411519397e-05, "loss": 1.0164, "step": 3996 }, { "epoch": 0.27, "learning_rate": 1.7217746515390263e-05, "loss": 1.0454, "step": 3997 }, { "epoch": 0.27, "learning_rate": 1.7216254283829943e-05, "loss": 0.977, "step": 3998 }, { "epoch": 0.27, "learning_rate": 1.7214761716907785e-05, "loss": 1.0058, "step": 3999 }, { "epoch": 0.27, "learning_rate": 1.721326881469315e-05, "loss": 1.0048, "step": 4000 }, { "epoch": 0.27, "learning_rate": 1.7211775577255422e-05, "loss": 0.9687, "step": 4001 }, { "epoch": 0.27, "learning_rate": 1.7210282004663995e-05, "loss": 1.0681, "step": 4002 }, { "epoch": 0.27, "learning_rate": 1.7208788096988282e-05, "loss": 1.0191, "step": 4003 }, { "epoch": 0.27, "learning_rate": 1.72072938542977e-05, "loss": 1.0127, "step": 4004 }, { "epoch": 0.27, "learning_rate": 1.7205799276661704e-05, "loss": 1.063, "step": 4005 }, { "epoch": 0.27, "learning_rate": 1.7204304364149745e-05, "loss": 1.0392, "step": 4006 }, { "epoch": 0.27, "learning_rate": 1.7202809116831297e-05, "loss": 1.009, "step": 4007 }, { "epoch": 0.27, "learning_rate": 1.7201313534775847e-05, "loss": 0.9646, "step": 4008 }, { "epoch": 0.27, "learning_rate": 1.71998176180529e-05, "loss": 0.9945, "step": 4009 }, { "epoch": 0.27, "learning_rate": 1.7198321366731986e-05, "loss": 1.0102, "step": 4010 }, { "epoch": 0.27, "learning_rate": 1.7196824780882625e-05, "loss": 1.0129, "step": 4011 }, { "epoch": 0.27, "learning_rate": 1.7195327860574377e-05, "loss": 0.9815, "step": 4012 }, { "epoch": 0.27, "learning_rate": 1.7193830605876806e-05, "loss": 1.0212, "step": 4013 }, { "epoch": 0.27, "learning_rate": 1.719233301685949e-05, "loss": 0.9592, "step": 4014 }, { "epoch": 0.27, "learning_rate": 1.719083509359204e-05, "loss": 1.0211, "step": 4015 }, { "epoch": 0.27, "learning_rate": 1.7189336836144054e-05, "loss": 0.9952, "step": 4016 }, { "epoch": 0.27, "learning_rate": 1.7187838244585167e-05, "loss": 1.0108, "step": 4017 }, { "epoch": 0.27, "learning_rate": 1.7186339318985027e-05, "loss": 0.9905, "step": 4018 }, { "epoch": 0.27, "learning_rate": 1.7184840059413284e-05, "loss": 1.0052, "step": 4019 }, { "epoch": 0.27, "learning_rate": 1.7183340465939622e-05, "loss": 1.0284, "step": 4020 }, { "epoch": 0.27, "learning_rate": 1.7181840538633728e-05, "loss": 1.0506, "step": 4021 }, { "epoch": 0.27, "learning_rate": 1.7180340277565312e-05, "loss": 0.9859, "step": 4022 }, { "epoch": 0.27, "learning_rate": 1.7178839682804093e-05, "loss": 1.0544, "step": 4023 }, { "epoch": 0.27, "learning_rate": 1.7177338754419808e-05, "loss": 0.9984, "step": 4024 }, { "epoch": 0.27, "learning_rate": 1.7175837492482208e-05, "loss": 1.0021, "step": 4025 }, { "epoch": 0.27, "learning_rate": 1.7174335897061066e-05, "loss": 0.968, "step": 4026 }, { "epoch": 0.27, "learning_rate": 1.717283396822616e-05, "loss": 1.0286, "step": 4027 }, { "epoch": 0.27, "learning_rate": 1.7171331706047296e-05, "loss": 1.0074, "step": 4028 }, { "epoch": 0.27, "learning_rate": 1.7169829110594282e-05, "loss": 1.0055, "step": 4029 }, { "epoch": 0.27, "learning_rate": 1.7168326181936955e-05, "loss": 1.0482, "step": 4030 }, { "epoch": 0.27, "learning_rate": 1.716682292014516e-05, "loss": 1.0169, "step": 4031 }, { "epoch": 0.27, "learning_rate": 1.7165319325288748e-05, "loss": 1.0478, "step": 4032 }, { "epoch": 0.27, "learning_rate": 1.716381539743761e-05, "loss": 1.0334, "step": 4033 }, { "epoch": 0.27, "learning_rate": 1.716231113666163e-05, "loss": 1.0277, "step": 4034 }, { "epoch": 0.27, "learning_rate": 1.7160806543030718e-05, "loss": 1.0218, "step": 4035 }, { "epoch": 0.27, "learning_rate": 1.7159301616614794e-05, "loss": 1.0021, "step": 4036 }, { "epoch": 0.27, "learning_rate": 1.7157796357483804e-05, "loss": 1.002, "step": 4037 }, { "epoch": 0.27, "learning_rate": 1.7156290765707694e-05, "loss": 1.0188, "step": 4038 }, { "epoch": 0.27, "learning_rate": 1.7154784841356434e-05, "loss": 1.0095, "step": 4039 }, { "epoch": 0.27, "learning_rate": 1.7153278584500015e-05, "loss": 1.0185, "step": 4040 }, { "epoch": 0.27, "learning_rate": 1.7151771995208434e-05, "loss": 1.0564, "step": 4041 }, { "epoch": 0.27, "learning_rate": 1.7150265073551708e-05, "loss": 1.0073, "step": 4042 }, { "epoch": 0.27, "learning_rate": 1.7148757819599867e-05, "loss": 1.0423, "step": 4043 }, { "epoch": 0.27, "learning_rate": 1.714725023342296e-05, "loss": 0.9885, "step": 4044 }, { "epoch": 0.27, "learning_rate": 1.7145742315091044e-05, "loss": 1.0165, "step": 4045 }, { "epoch": 0.27, "learning_rate": 1.7144234064674202e-05, "loss": 0.9958, "step": 4046 }, { "epoch": 0.27, "learning_rate": 1.7142725482242528e-05, "loss": 1.0598, "step": 4047 }, { "epoch": 0.27, "learning_rate": 1.7141216567866122e-05, "loss": 0.9668, "step": 4048 }, { "epoch": 0.27, "learning_rate": 1.713970732161512e-05, "loss": 0.9816, "step": 4049 }, { "epoch": 0.27, "learning_rate": 1.7138197743559656e-05, "loss": 1.024, "step": 4050 }, { "epoch": 0.27, "learning_rate": 1.713668783376988e-05, "loss": 1.0178, "step": 4051 }, { "epoch": 0.27, "learning_rate": 1.7135177592315966e-05, "loss": 1.0155, "step": 4052 }, { "epoch": 0.27, "learning_rate": 1.71336670192681e-05, "loss": 0.9666, "step": 4053 }, { "epoch": 0.27, "learning_rate": 1.7132156114696485e-05, "loss": 0.9997, "step": 4054 }, { "epoch": 0.27, "learning_rate": 1.713064487867133e-05, "loss": 1.0507, "step": 4055 }, { "epoch": 0.27, "learning_rate": 1.7129133311262882e-05, "loss": 0.971, "step": 4056 }, { "epoch": 0.27, "learning_rate": 1.7127621412541372e-05, "loss": 1.0211, "step": 4057 }, { "epoch": 0.27, "learning_rate": 1.7126109182577067e-05, "loss": 1.0152, "step": 4058 }, { "epoch": 0.27, "learning_rate": 1.712459662144025e-05, "loss": 0.996, "step": 4059 }, { "epoch": 0.27, "learning_rate": 1.7123083729201214e-05, "loss": 0.9929, "step": 4060 }, { "epoch": 0.27, "learning_rate": 1.7121570505930262e-05, "loss": 1.0421, "step": 4061 }, { "epoch": 0.27, "learning_rate": 1.7120056951697724e-05, "loss": 0.9984, "step": 4062 }, { "epoch": 0.27, "learning_rate": 1.7118543066573937e-05, "loss": 1.041, "step": 4063 }, { "epoch": 0.27, "learning_rate": 1.7117028850629255e-05, "loss": 1.0052, "step": 4064 }, { "epoch": 0.27, "learning_rate": 1.711551430393405e-05, "loss": 0.9714, "step": 4065 }, { "epoch": 0.27, "learning_rate": 1.7113999426558704e-05, "loss": 1.0022, "step": 4066 }, { "epoch": 0.27, "learning_rate": 1.7112484218573625e-05, "loss": 1.0305, "step": 4067 }, { "epoch": 0.27, "learning_rate": 1.7110968680049226e-05, "loss": 1.0043, "step": 4068 }, { "epoch": 0.27, "learning_rate": 1.710945281105594e-05, "loss": 1.0181, "step": 4069 }, { "epoch": 0.27, "learning_rate": 1.7107936611664207e-05, "loss": 0.9783, "step": 4070 }, { "epoch": 0.27, "learning_rate": 1.7106420081944496e-05, "loss": 1.017, "step": 4071 }, { "epoch": 0.27, "learning_rate": 1.710490322196729e-05, "loss": 0.9963, "step": 4072 }, { "epoch": 0.27, "learning_rate": 1.710338603180307e-05, "loss": 1.0017, "step": 4073 }, { "epoch": 0.27, "learning_rate": 1.7101868511522352e-05, "loss": 0.9947, "step": 4074 }, { "epoch": 0.27, "learning_rate": 1.7100350661195663e-05, "loss": 1.0634, "step": 4075 }, { "epoch": 0.27, "learning_rate": 1.7098832480893534e-05, "loss": 1.014, "step": 4076 }, { "epoch": 0.27, "learning_rate": 1.7097313970686526e-05, "loss": 1.0137, "step": 4077 }, { "epoch": 0.27, "learning_rate": 1.7095795130645203e-05, "loss": 1.0131, "step": 4078 }, { "epoch": 0.27, "learning_rate": 1.7094275960840157e-05, "loss": 1.0289, "step": 4079 }, { "epoch": 0.27, "learning_rate": 1.7092756461341983e-05, "loss": 0.9922, "step": 4080 }, { "epoch": 0.27, "learning_rate": 1.70912366322213e-05, "loss": 1.0182, "step": 4081 }, { "epoch": 0.27, "learning_rate": 1.708971647354874e-05, "loss": 1.025, "step": 4082 }, { "epoch": 0.27, "learning_rate": 1.7088195985394945e-05, "loss": 0.9971, "step": 4083 }, { "epoch": 0.27, "learning_rate": 1.7086675167830578e-05, "loss": 0.9596, "step": 4084 }, { "epoch": 0.27, "learning_rate": 1.7085154020926323e-05, "loss": 1.0186, "step": 4085 }, { "epoch": 0.27, "learning_rate": 1.7083632544752866e-05, "loss": 1.0135, "step": 4086 }, { "epoch": 0.27, "learning_rate": 1.7082110739380912e-05, "loss": 0.9952, "step": 4087 }, { "epoch": 0.27, "learning_rate": 1.7080588604881194e-05, "loss": 1.0068, "step": 4088 }, { "epoch": 0.27, "learning_rate": 1.707906614132444e-05, "loss": 1.033, "step": 4089 }, { "epoch": 0.27, "learning_rate": 1.707754334878141e-05, "loss": 1.0627, "step": 4090 }, { "epoch": 0.27, "learning_rate": 1.707602022732287e-05, "loss": 1.0246, "step": 4091 }, { "epoch": 0.27, "learning_rate": 1.7074496777019606e-05, "loss": 1.0239, "step": 4092 }, { "epoch": 0.27, "learning_rate": 1.7072972997942415e-05, "loss": 1.0687, "step": 4093 }, { "epoch": 0.27, "learning_rate": 1.7071448890162112e-05, "loss": 0.9953, "step": 4094 }, { "epoch": 0.27, "learning_rate": 1.7069924453749532e-05, "loss": 1.0389, "step": 4095 }, { "epoch": 0.27, "learning_rate": 1.7068399688775515e-05, "loss": 0.9777, "step": 4096 }, { "epoch": 0.27, "learning_rate": 1.7066874595310923e-05, "loss": 1.0267, "step": 4097 }, { "epoch": 0.27, "learning_rate": 1.7065349173426633e-05, "loss": 0.9732, "step": 4098 }, { "epoch": 0.27, "learning_rate": 1.7063823423193534e-05, "loss": 1.0151, "step": 4099 }, { "epoch": 0.27, "learning_rate": 1.7062297344682534e-05, "loss": 0.9612, "step": 4100 }, { "epoch": 0.27, "learning_rate": 1.7060770937964556e-05, "loss": 1.0191, "step": 4101 }, { "epoch": 0.27, "learning_rate": 1.7059244203110533e-05, "loss": 0.9653, "step": 4102 }, { "epoch": 0.27, "learning_rate": 1.705771714019142e-05, "loss": 0.9512, "step": 4103 }, { "epoch": 0.27, "learning_rate": 1.7056189749278184e-05, "loss": 1.0142, "step": 4104 }, { "epoch": 0.27, "learning_rate": 1.7054662030441806e-05, "loss": 1.0459, "step": 4105 }, { "epoch": 0.27, "learning_rate": 1.705313398375329e-05, "loss": 1.0122, "step": 4106 }, { "epoch": 0.27, "learning_rate": 1.7051605609283638e-05, "loss": 1.0276, "step": 4107 }, { "epoch": 0.27, "learning_rate": 1.7050076907103886e-05, "loss": 0.9852, "step": 4108 }, { "epoch": 0.27, "learning_rate": 1.7048547877285078e-05, "loss": 1.0082, "step": 4109 }, { "epoch": 0.27, "learning_rate": 1.704701851989827e-05, "loss": 1.0149, "step": 4110 }, { "epoch": 0.27, "learning_rate": 1.7045488835014534e-05, "loss": 0.9584, "step": 4111 }, { "epoch": 0.27, "learning_rate": 1.7043958822704968e-05, "loss": 0.9835, "step": 4112 }, { "epoch": 0.27, "learning_rate": 1.7042428483040665e-05, "loss": 0.9933, "step": 4113 }, { "epoch": 0.27, "learning_rate": 1.704089781609275e-05, "loss": 0.9869, "step": 4114 }, { "epoch": 0.27, "learning_rate": 1.7039366821932364e-05, "loss": 0.9926, "step": 4115 }, { "epoch": 0.27, "learning_rate": 1.7037835500630645e-05, "loss": 1.0276, "step": 4116 }, { "epoch": 0.27, "learning_rate": 1.7036303852258768e-05, "loss": 1.0118, "step": 4117 }, { "epoch": 0.27, "learning_rate": 1.703477187688791e-05, "loss": 1.0189, "step": 4118 }, { "epoch": 0.27, "learning_rate": 1.7033239574589267e-05, "loss": 1.0172, "step": 4119 }, { "epoch": 0.27, "learning_rate": 1.7031706945434047e-05, "loss": 0.9648, "step": 4120 }, { "epoch": 0.27, "learning_rate": 1.703017398949348e-05, "loss": 1.0036, "step": 4121 }, { "epoch": 0.27, "learning_rate": 1.70286407068388e-05, "loss": 1.0024, "step": 4122 }, { "epoch": 0.27, "learning_rate": 1.702710709754128e-05, "loss": 1.0271, "step": 4123 }, { "epoch": 0.27, "learning_rate": 1.7025573161672174e-05, "loss": 1.0246, "step": 4124 }, { "epoch": 0.27, "learning_rate": 1.7024038899302773e-05, "loss": 1.0382, "step": 4125 }, { "epoch": 0.27, "learning_rate": 1.702250431050439e-05, "loss": 0.9634, "step": 4126 }, { "epoch": 0.27, "learning_rate": 1.7020969395348326e-05, "loss": 0.9658, "step": 4127 }, { "epoch": 0.27, "learning_rate": 1.7019434153905922e-05, "loss": 1.033, "step": 4128 }, { "epoch": 0.27, "learning_rate": 1.7017898586248528e-05, "loss": 1.0316, "step": 4129 }, { "epoch": 0.27, "learning_rate": 1.7016362692447498e-05, "loss": 0.9761, "step": 4130 }, { "epoch": 0.27, "learning_rate": 1.7014826472574223e-05, "loss": 1.0418, "step": 4131 }, { "epoch": 0.28, "learning_rate": 1.701328992670008e-05, "loss": 0.9987, "step": 4132 }, { "epoch": 0.28, "learning_rate": 1.7011753054896487e-05, "loss": 1.0475, "step": 4133 }, { "epoch": 0.28, "learning_rate": 1.7010215857234866e-05, "loss": 0.9882, "step": 4134 }, { "epoch": 0.28, "learning_rate": 1.7008678333786653e-05, "loss": 0.9757, "step": 4135 }, { "epoch": 0.28, "learning_rate": 1.7007140484623302e-05, "loss": 0.9619, "step": 4136 }, { "epoch": 0.28, "learning_rate": 1.7005602309816284e-05, "loss": 1.0199, "step": 4137 }, { "epoch": 0.28, "learning_rate": 1.7004063809437082e-05, "loss": 0.9629, "step": 4138 }, { "epoch": 0.28, "learning_rate": 1.7002524983557193e-05, "loss": 0.994, "step": 4139 }, { "epoch": 0.28, "learning_rate": 1.7000985832248137e-05, "loss": 1.0078, "step": 4140 }, { "epoch": 0.28, "learning_rate": 1.699944635558143e-05, "loss": 1.0312, "step": 4141 }, { "epoch": 0.28, "learning_rate": 1.699790655362863e-05, "loss": 0.9808, "step": 4142 }, { "epoch": 0.28, "learning_rate": 1.699636642646129e-05, "loss": 1.0015, "step": 4143 }, { "epoch": 0.28, "learning_rate": 1.6994825974150987e-05, "loss": 0.9822, "step": 4144 }, { "epoch": 0.28, "learning_rate": 1.6993285196769305e-05, "loss": 0.9944, "step": 4145 }, { "epoch": 0.28, "learning_rate": 1.699174409438786e-05, "loss": 0.9828, "step": 4146 }, { "epoch": 0.28, "learning_rate": 1.699020266707826e-05, "loss": 1.0179, "step": 4147 }, { "epoch": 0.28, "learning_rate": 1.6988660914912144e-05, "loss": 1.0384, "step": 4148 }, { "epoch": 0.28, "learning_rate": 1.698711883796116e-05, "loss": 0.9796, "step": 4149 }, { "epoch": 0.28, "learning_rate": 1.698557643629698e-05, "loss": 0.9786, "step": 4150 }, { "epoch": 0.28, "learning_rate": 1.6984033709991282e-05, "loss": 1.0376, "step": 4151 }, { "epoch": 0.28, "learning_rate": 1.698249065911576e-05, "loss": 0.9892, "step": 4152 }, { "epoch": 0.28, "learning_rate": 1.698094728374212e-05, "loss": 1.0228, "step": 4153 }, { "epoch": 0.28, "learning_rate": 1.6979403583942095e-05, "loss": 1.0129, "step": 4154 }, { "epoch": 0.28, "learning_rate": 1.697785955978742e-05, "loss": 0.9989, "step": 4155 }, { "epoch": 0.28, "learning_rate": 1.6976315211349848e-05, "loss": 1.0111, "step": 4156 }, { "epoch": 0.28, "learning_rate": 1.6974770538701157e-05, "loss": 1.0032, "step": 4157 }, { "epoch": 0.28, "learning_rate": 1.6973225541913133e-05, "loss": 0.9677, "step": 4158 }, { "epoch": 0.28, "learning_rate": 1.697168022105757e-05, "loss": 1.0122, "step": 4159 }, { "epoch": 0.28, "learning_rate": 1.6970134576206287e-05, "loss": 0.997, "step": 4160 }, { "epoch": 0.28, "learning_rate": 1.6968588607431117e-05, "loss": 0.9745, "step": 4161 }, { "epoch": 0.28, "learning_rate": 1.6967042314803908e-05, "loss": 0.9828, "step": 4162 }, { "epoch": 0.28, "learning_rate": 1.6965495698396512e-05, "loss": 1.0016, "step": 4163 }, { "epoch": 0.28, "learning_rate": 1.6963948758280815e-05, "loss": 1.0301, "step": 4164 }, { "epoch": 0.28, "learning_rate": 1.6962401494528702e-05, "loss": 1.034, "step": 4165 }, { "epoch": 0.28, "learning_rate": 1.696085390721208e-05, "loss": 1.0053, "step": 4166 }, { "epoch": 0.28, "learning_rate": 1.695930599640287e-05, "loss": 1.0445, "step": 4167 }, { "epoch": 0.28, "learning_rate": 1.695775776217301e-05, "loss": 1.0272, "step": 4168 }, { "epoch": 0.28, "learning_rate": 1.6956209204594452e-05, "loss": 1.0283, "step": 4169 }, { "epoch": 0.28, "learning_rate": 1.6954660323739163e-05, "loss": 1.0015, "step": 4170 }, { "epoch": 0.28, "learning_rate": 1.695311111967912e-05, "loss": 0.9939, "step": 4171 }, { "epoch": 0.28, "learning_rate": 1.695156159248632e-05, "loss": 1.0242, "step": 4172 }, { "epoch": 0.28, "learning_rate": 1.6950011742232778e-05, "loss": 1.024, "step": 4173 }, { "epoch": 0.28, "learning_rate": 1.694846156899052e-05, "loss": 1.019, "step": 4174 }, { "epoch": 0.28, "learning_rate": 1.694691107283158e-05, "loss": 0.9984, "step": 4175 }, { "epoch": 0.28, "learning_rate": 1.694536025382802e-05, "loss": 0.9975, "step": 4176 }, { "epoch": 0.28, "learning_rate": 1.6943809112051922e-05, "loss": 1.0054, "step": 4177 }, { "epoch": 0.28, "learning_rate": 1.694225764757535e-05, "loss": 0.9926, "step": 4178 }, { "epoch": 0.28, "learning_rate": 1.6940705860470426e-05, "loss": 1.0272, "step": 4179 }, { "epoch": 0.28, "learning_rate": 1.6939153750809252e-05, "loss": 0.9943, "step": 4180 }, { "epoch": 0.28, "learning_rate": 1.6937601318663967e-05, "loss": 1.0076, "step": 4181 }, { "epoch": 0.28, "learning_rate": 1.6936048564106714e-05, "loss": 0.9641, "step": 4182 }, { "epoch": 0.28, "learning_rate": 1.6934495487209656e-05, "loss": 1.0243, "step": 4183 }, { "epoch": 0.28, "learning_rate": 1.6932942088044972e-05, "loss": 1.03, "step": 4184 }, { "epoch": 0.28, "learning_rate": 1.6931388366684848e-05, "loss": 1.0164, "step": 4185 }, { "epoch": 0.28, "learning_rate": 1.6929834323201496e-05, "loss": 0.9808, "step": 4186 }, { "epoch": 0.28, "learning_rate": 1.692827995766713e-05, "loss": 1.0149, "step": 4187 }, { "epoch": 0.28, "learning_rate": 1.692672527015399e-05, "loss": 1.0269, "step": 4188 }, { "epoch": 0.28, "learning_rate": 1.692517026073433e-05, "loss": 1.0118, "step": 4189 }, { "epoch": 0.28, "learning_rate": 1.692361492948041e-05, "loss": 1.0227, "step": 4190 }, { "epoch": 0.28, "learning_rate": 1.6922059276464516e-05, "loss": 1.0608, "step": 4191 }, { "epoch": 0.28, "learning_rate": 1.6920503301758943e-05, "loss": 0.976, "step": 4192 }, { "epoch": 0.28, "learning_rate": 1.6918947005436e-05, "loss": 1.0671, "step": 4193 }, { "epoch": 0.28, "learning_rate": 1.691739038756802e-05, "loss": 0.9879, "step": 4194 }, { "epoch": 0.28, "learning_rate": 1.6915833448227334e-05, "loss": 1.0198, "step": 4195 }, { "epoch": 0.28, "learning_rate": 1.6914276187486304e-05, "loss": 0.9603, "step": 4196 }, { "epoch": 0.28, "learning_rate": 1.6912718605417295e-05, "loss": 0.9761, "step": 4197 }, { "epoch": 0.28, "learning_rate": 1.69111607020927e-05, "loss": 1.0359, "step": 4198 }, { "epoch": 0.28, "learning_rate": 1.6909602477584915e-05, "loss": 0.9947, "step": 4199 }, { "epoch": 0.28, "learning_rate": 1.690804393196636e-05, "loss": 1.0253, "step": 4200 }, { "epoch": 0.28, "learning_rate": 1.6906485065309456e-05, "loss": 1.0168, "step": 4201 }, { "epoch": 0.28, "learning_rate": 1.690492587768666e-05, "loss": 1.0031, "step": 4202 }, { "epoch": 0.28, "learning_rate": 1.6903366369170425e-05, "loss": 1.0522, "step": 4203 }, { "epoch": 0.28, "learning_rate": 1.690180653983323e-05, "loss": 0.9817, "step": 4204 }, { "epoch": 0.28, "learning_rate": 1.690024638974756e-05, "loss": 0.9831, "step": 4205 }, { "epoch": 0.28, "learning_rate": 1.6898685918985926e-05, "loss": 1.0386, "step": 4206 }, { "epoch": 0.28, "learning_rate": 1.6897125127620844e-05, "loss": 1.0002, "step": 4207 }, { "epoch": 0.28, "learning_rate": 1.6895564015724854e-05, "loss": 0.9833, "step": 4208 }, { "epoch": 0.28, "learning_rate": 1.68940025833705e-05, "loss": 1.0211, "step": 4209 }, { "epoch": 0.28, "learning_rate": 1.6892440830630348e-05, "loss": 0.9936, "step": 4210 }, { "epoch": 0.28, "learning_rate": 1.689087875757698e-05, "loss": 1.0182, "step": 4211 }, { "epoch": 0.28, "learning_rate": 1.6889316364282983e-05, "loss": 0.9985, "step": 4212 }, { "epoch": 0.28, "learning_rate": 1.6887753650820976e-05, "loss": 1.0217, "step": 4213 }, { "epoch": 0.28, "learning_rate": 1.688619061726358e-05, "loss": 1.0001, "step": 4214 }, { "epoch": 0.28, "learning_rate": 1.6884627263683428e-05, "loss": 0.9631, "step": 4215 }, { "epoch": 0.28, "learning_rate": 1.6883063590153184e-05, "loss": 1.0184, "step": 4216 }, { "epoch": 0.28, "learning_rate": 1.688149959674551e-05, "loss": 1.0068, "step": 4217 }, { "epoch": 0.28, "learning_rate": 1.687993528353309e-05, "loss": 1.0423, "step": 4218 }, { "epoch": 0.28, "learning_rate": 1.6878370650588624e-05, "loss": 1.018, "step": 4219 }, { "epoch": 0.28, "learning_rate": 1.687680569798483e-05, "loss": 0.9947, "step": 4220 }, { "epoch": 0.28, "learning_rate": 1.6875240425794427e-05, "loss": 1.0115, "step": 4221 }, { "epoch": 0.28, "learning_rate": 1.687367483409016e-05, "loss": 1.0122, "step": 4222 }, { "epoch": 0.28, "learning_rate": 1.6872108922944798e-05, "loss": 1.0105, "step": 4223 }, { "epoch": 0.28, "learning_rate": 1.68705426924311e-05, "loss": 0.9934, "step": 4224 }, { "epoch": 0.28, "learning_rate": 1.6868976142621858e-05, "loss": 0.9965, "step": 4225 }, { "epoch": 0.28, "learning_rate": 1.6867409273589875e-05, "loss": 1.0297, "step": 4226 }, { "epoch": 0.28, "learning_rate": 1.686584208540797e-05, "loss": 0.999, "step": 4227 }, { "epoch": 0.28, "learning_rate": 1.6864274578148975e-05, "loss": 0.9852, "step": 4228 }, { "epoch": 0.28, "learning_rate": 1.6862706751885733e-05, "loss": 1.039, "step": 4229 }, { "epoch": 0.28, "learning_rate": 1.686113860669111e-05, "loss": 0.9703, "step": 4230 }, { "epoch": 0.28, "learning_rate": 1.6859570142637985e-05, "loss": 1.0029, "step": 4231 }, { "epoch": 0.28, "learning_rate": 1.685800135979924e-05, "loss": 1.0076, "step": 4232 }, { "epoch": 0.28, "learning_rate": 1.6856432258247787e-05, "loss": 0.9675, "step": 4233 }, { "epoch": 0.28, "learning_rate": 1.6854862838056546e-05, "loss": 1.0219, "step": 4234 }, { "epoch": 0.28, "learning_rate": 1.685329309929846e-05, "loss": 1.0375, "step": 4235 }, { "epoch": 0.28, "learning_rate": 1.6851723042046468e-05, "loss": 1.0243, "step": 4236 }, { "epoch": 0.28, "learning_rate": 1.685015266637354e-05, "loss": 0.9777, "step": 4237 }, { "epoch": 0.28, "learning_rate": 1.684858197235266e-05, "loss": 1.0194, "step": 4238 }, { "epoch": 0.28, "learning_rate": 1.6847010960056816e-05, "loss": 0.9802, "step": 4239 }, { "epoch": 0.28, "learning_rate": 1.6845439629559027e-05, "loss": 1.0324, "step": 4240 }, { "epoch": 0.28, "learning_rate": 1.6843867980932308e-05, "loss": 1.0349, "step": 4241 }, { "epoch": 0.28, "learning_rate": 1.6842296014249705e-05, "loss": 1.0041, "step": 4242 }, { "epoch": 0.28, "learning_rate": 1.684072372958427e-05, "loss": 1.0273, "step": 4243 }, { "epoch": 0.28, "learning_rate": 1.6839151127009072e-05, "loss": 0.9995, "step": 4244 }, { "epoch": 0.28, "learning_rate": 1.6837578206597193e-05, "loss": 0.9907, "step": 4245 }, { "epoch": 0.28, "learning_rate": 1.6836004968421737e-05, "loss": 1.0088, "step": 4246 }, { "epoch": 0.28, "learning_rate": 1.683443141255581e-05, "loss": 1.0209, "step": 4247 }, { "epoch": 0.28, "learning_rate": 1.6832857539072545e-05, "loss": 0.9741, "step": 4248 }, { "epoch": 0.28, "learning_rate": 1.6831283348045085e-05, "loss": 1.0444, "step": 4249 }, { "epoch": 0.28, "learning_rate": 1.6829708839546587e-05, "loss": 0.997, "step": 4250 }, { "epoch": 0.28, "learning_rate": 1.6828134013650216e-05, "loss": 1.0045, "step": 4251 }, { "epoch": 0.28, "learning_rate": 1.682655887042917e-05, "loss": 0.9746, "step": 4252 }, { "epoch": 0.28, "learning_rate": 1.6824983409956648e-05, "loss": 1.0052, "step": 4253 }, { "epoch": 0.28, "learning_rate": 1.6823407632305865e-05, "loss": 1.0286, "step": 4254 }, { "epoch": 0.28, "learning_rate": 1.682183153755005e-05, "loss": 1.0495, "step": 4255 }, { "epoch": 0.28, "learning_rate": 1.682025512576245e-05, "loss": 0.9953, "step": 4256 }, { "epoch": 0.28, "learning_rate": 1.6818678397016325e-05, "loss": 0.9738, "step": 4257 }, { "epoch": 0.28, "learning_rate": 1.681710135138496e-05, "loss": 0.9882, "step": 4258 }, { "epoch": 0.28, "learning_rate": 1.681552398894163e-05, "loss": 0.9789, "step": 4259 }, { "epoch": 0.28, "learning_rate": 1.681394630975965e-05, "loss": 1.0249, "step": 4260 }, { "epoch": 0.28, "learning_rate": 1.6812368313912336e-05, "loss": 0.9874, "step": 4261 }, { "epoch": 0.28, "learning_rate": 1.681079000147303e-05, "loss": 0.9628, "step": 4262 }, { "epoch": 0.28, "learning_rate": 1.680921137251507e-05, "loss": 0.9999, "step": 4263 }, { "epoch": 0.28, "learning_rate": 1.6807632427111823e-05, "loss": 1.0121, "step": 4264 }, { "epoch": 0.28, "learning_rate": 1.680605316533667e-05, "loss": 0.9867, "step": 4265 }, { "epoch": 0.28, "learning_rate": 1.6804473587263004e-05, "loss": 0.9893, "step": 4266 }, { "epoch": 0.28, "learning_rate": 1.680289369296423e-05, "loss": 1.0061, "step": 4267 }, { "epoch": 0.28, "learning_rate": 1.6801313482513775e-05, "loss": 1.0297, "step": 4268 }, { "epoch": 0.28, "learning_rate": 1.6799732955985076e-05, "loss": 1.0152, "step": 4269 }, { "epoch": 0.28, "learning_rate": 1.679815211345158e-05, "loss": 0.9474, "step": 4270 }, { "epoch": 0.28, "learning_rate": 1.679657095498676e-05, "loss": 1.0433, "step": 4271 }, { "epoch": 0.28, "learning_rate": 1.6794989480664093e-05, "loss": 1.0089, "step": 4272 }, { "epoch": 0.28, "learning_rate": 1.6793407690557075e-05, "loss": 1.0016, "step": 4273 }, { "epoch": 0.28, "learning_rate": 1.679182558473922e-05, "loss": 1.0146, "step": 4274 }, { "epoch": 0.28, "learning_rate": 1.6790243163284047e-05, "loss": 1.0071, "step": 4275 }, { "epoch": 0.28, "learning_rate": 1.6788660426265104e-05, "loss": 1.0442, "step": 4276 }, { "epoch": 0.28, "learning_rate": 1.6787077373755942e-05, "loss": 1.0262, "step": 4277 }, { "epoch": 0.28, "learning_rate": 1.678549400583013e-05, "loss": 0.9572, "step": 4278 }, { "epoch": 0.28, "learning_rate": 1.6783910322561254e-05, "loss": 1.041, "step": 4279 }, { "epoch": 0.28, "learning_rate": 1.6782326324022908e-05, "loss": 1.0089, "step": 4280 }, { "epoch": 0.28, "learning_rate": 1.678074201028871e-05, "loss": 0.991, "step": 4281 }, { "epoch": 0.29, "learning_rate": 1.6779157381432285e-05, "loss": 0.9869, "step": 4282 }, { "epoch": 0.29, "learning_rate": 1.677757243752728e-05, "loss": 1.0199, "step": 4283 }, { "epoch": 0.29, "learning_rate": 1.6775987178647347e-05, "loss": 0.9973, "step": 4284 }, { "epoch": 0.29, "learning_rate": 1.6774401604866163e-05, "loss": 0.9674, "step": 4285 }, { "epoch": 0.29, "learning_rate": 1.6772815716257414e-05, "loss": 0.987, "step": 4286 }, { "epoch": 0.29, "learning_rate": 1.6771229512894796e-05, "loss": 1.001, "step": 4287 }, { "epoch": 0.29, "learning_rate": 1.6769642994852026e-05, "loss": 0.9676, "step": 4288 }, { "epoch": 0.29, "learning_rate": 1.6768056162202838e-05, "loss": 1.0259, "step": 4289 }, { "epoch": 0.29, "learning_rate": 1.676646901502098e-05, "loss": 1.0352, "step": 4290 }, { "epoch": 0.29, "learning_rate": 1.67648815533802e-05, "loss": 1.0317, "step": 4291 }, { "epoch": 0.29, "learning_rate": 1.6763293777354286e-05, "loss": 0.9772, "step": 4292 }, { "epoch": 0.29, "learning_rate": 1.6761705687017015e-05, "loss": 1.049, "step": 4293 }, { "epoch": 0.29, "learning_rate": 1.67601172824422e-05, "loss": 1.0418, "step": 4294 }, { "epoch": 0.29, "learning_rate": 1.6758528563703654e-05, "loss": 0.9814, "step": 4295 }, { "epoch": 0.29, "learning_rate": 1.6756939530875208e-05, "loss": 0.9841, "step": 4296 }, { "epoch": 0.29, "learning_rate": 1.6755350184030714e-05, "loss": 1.0201, "step": 4297 }, { "epoch": 0.29, "learning_rate": 1.675376052324403e-05, "loss": 0.9868, "step": 4298 }, { "epoch": 0.29, "learning_rate": 1.6752170548589037e-05, "loss": 0.9915, "step": 4299 }, { "epoch": 0.29, "learning_rate": 1.6750580260139628e-05, "loss": 1.0308, "step": 4300 }, { "epoch": 0.29, "learning_rate": 1.6748989657969696e-05, "loss": 0.9909, "step": 4301 }, { "epoch": 0.29, "learning_rate": 1.6747398742153175e-05, "loss": 0.9714, "step": 4302 }, { "epoch": 0.29, "learning_rate": 1.674580751276399e-05, "loss": 1.02, "step": 4303 }, { "epoch": 0.29, "learning_rate": 1.6744215969876095e-05, "loss": 1.0349, "step": 4304 }, { "epoch": 0.29, "learning_rate": 1.6742624113563456e-05, "loss": 0.9742, "step": 4305 }, { "epoch": 0.29, "learning_rate": 1.6741031943900048e-05, "loss": 1.0092, "step": 4306 }, { "epoch": 0.29, "learning_rate": 1.6739439460959864e-05, "loss": 1.0252, "step": 4307 }, { "epoch": 0.29, "learning_rate": 1.6737846664816916e-05, "loss": 0.9779, "step": 4308 }, { "epoch": 0.29, "learning_rate": 1.673625355554522e-05, "loss": 1.0215, "step": 4309 }, { "epoch": 0.29, "learning_rate": 1.673466013321882e-05, "loss": 0.9865, "step": 4310 }, { "epoch": 0.29, "learning_rate": 1.6733066397911756e-05, "loss": 1.0093, "step": 4311 }, { "epoch": 0.29, "learning_rate": 1.6731472349698105e-05, "loss": 0.9217, "step": 4312 }, { "epoch": 0.29, "learning_rate": 1.6729877988651944e-05, "loss": 1.0188, "step": 4313 }, { "epoch": 0.29, "learning_rate": 1.6728283314847368e-05, "loss": 1.0057, "step": 4314 }, { "epoch": 0.29, "learning_rate": 1.6726688328358486e-05, "loss": 1.0263, "step": 4315 }, { "epoch": 0.29, "learning_rate": 1.672509302925942e-05, "loss": 1.0475, "step": 4316 }, { "epoch": 0.29, "learning_rate": 1.6723497417624308e-05, "loss": 0.9924, "step": 4317 }, { "epoch": 0.29, "learning_rate": 1.6721901493527307e-05, "loss": 0.9745, "step": 4318 }, { "epoch": 0.29, "learning_rate": 1.6720305257042583e-05, "loss": 1.0212, "step": 4319 }, { "epoch": 0.29, "learning_rate": 1.671870870824432e-05, "loss": 0.9904, "step": 4320 }, { "epoch": 0.29, "learning_rate": 1.6717111847206714e-05, "loss": 1.0137, "step": 4321 }, { "epoch": 0.29, "learning_rate": 1.6715514674003972e-05, "loss": 0.9983, "step": 4322 }, { "epoch": 0.29, "learning_rate": 1.6713917188710324e-05, "loss": 1.051, "step": 4323 }, { "epoch": 0.29, "learning_rate": 1.6712319391400004e-05, "loss": 1.0014, "step": 4324 }, { "epoch": 0.29, "learning_rate": 1.6710721282147274e-05, "loss": 1.0235, "step": 4325 }, { "epoch": 0.29, "learning_rate": 1.6709122861026398e-05, "loss": 1.022, "step": 4326 }, { "epoch": 0.29, "learning_rate": 1.6707524128111664e-05, "loss": 1.0495, "step": 4327 }, { "epoch": 0.29, "learning_rate": 1.670592508347737e-05, "loss": 0.9683, "step": 4328 }, { "epoch": 0.29, "learning_rate": 1.6704325727197823e-05, "loss": 0.9945, "step": 4329 }, { "epoch": 0.29, "learning_rate": 1.6702726059347356e-05, "loss": 0.9943, "step": 4330 }, { "epoch": 0.29, "learning_rate": 1.6701126080000306e-05, "loss": 1.011, "step": 4331 }, { "epoch": 0.29, "learning_rate": 1.6699525789231033e-05, "loss": 1.0061, "step": 4332 }, { "epoch": 0.29, "learning_rate": 1.6697925187113905e-05, "loss": 0.9959, "step": 4333 }, { "epoch": 0.29, "learning_rate": 1.6696324273723305e-05, "loss": 1.0138, "step": 4334 }, { "epoch": 0.29, "learning_rate": 1.6694723049133638e-05, "loss": 1.0344, "step": 4335 }, { "epoch": 0.29, "learning_rate": 1.6693121513419316e-05, "loss": 1.0095, "step": 4336 }, { "epoch": 0.29, "learning_rate": 1.6691519666654763e-05, "loss": 0.999, "step": 4337 }, { "epoch": 0.29, "learning_rate": 1.6689917508914428e-05, "loss": 1.0412, "step": 4338 }, { "epoch": 0.29, "learning_rate": 1.6688315040272766e-05, "loss": 1.0265, "step": 4339 }, { "epoch": 0.29, "learning_rate": 1.6686712260804247e-05, "loss": 1.0111, "step": 4340 }, { "epoch": 0.29, "learning_rate": 1.6685109170583357e-05, "loss": 1.0532, "step": 4341 }, { "epoch": 0.29, "learning_rate": 1.6683505769684596e-05, "loss": 1.0277, "step": 4342 }, { "epoch": 0.29, "learning_rate": 1.6681902058182486e-05, "loss": 1.0159, "step": 4343 }, { "epoch": 0.29, "learning_rate": 1.6680298036151544e-05, "loss": 1.0508, "step": 4344 }, { "epoch": 0.29, "learning_rate": 1.6678693703666327e-05, "loss": 0.9937, "step": 4345 }, { "epoch": 0.29, "learning_rate": 1.6677089060801387e-05, "loss": 1.051, "step": 4346 }, { "epoch": 0.29, "learning_rate": 1.66754841076313e-05, "loss": 0.9779, "step": 4347 }, { "epoch": 0.29, "learning_rate": 1.6673878844230643e-05, "loss": 1.0415, "step": 4348 }, { "epoch": 0.29, "learning_rate": 1.667227327067403e-05, "loss": 0.9886, "step": 4349 }, { "epoch": 0.29, "learning_rate": 1.6670667387036075e-05, "loss": 0.9938, "step": 4350 }, { "epoch": 0.29, "learning_rate": 1.66690611933914e-05, "loss": 0.9855, "step": 4351 }, { "epoch": 0.29, "learning_rate": 1.666745468981466e-05, "loss": 0.9784, "step": 4352 }, { "epoch": 0.29, "learning_rate": 1.6665847876380505e-05, "loss": 1.0184, "step": 4353 }, { "epoch": 0.29, "learning_rate": 1.666424075316362e-05, "loss": 1.0021, "step": 4354 }, { "epoch": 0.29, "learning_rate": 1.666263332023868e-05, "loss": 1.0062, "step": 4355 }, { "epoch": 0.29, "learning_rate": 1.6661025577680394e-05, "loss": 0.9809, "step": 4356 }, { "epoch": 0.29, "learning_rate": 1.665941752556348e-05, "loss": 1.0379, "step": 4357 }, { "epoch": 0.29, "learning_rate": 1.6657809163962668e-05, "loss": 0.9716, "step": 4358 }, { "epoch": 0.29, "learning_rate": 1.6656200492952702e-05, "loss": 1.0062, "step": 4359 }, { "epoch": 0.29, "learning_rate": 1.6654591512608344e-05, "loss": 1.0022, "step": 4360 }, { "epoch": 0.29, "learning_rate": 1.6652982223004373e-05, "loss": 0.9847, "step": 4361 }, { "epoch": 0.29, "learning_rate": 1.665137262421557e-05, "loss": 0.9744, "step": 4362 }, { "epoch": 0.29, "learning_rate": 1.6649762716316733e-05, "loss": 0.9759, "step": 4363 }, { "epoch": 0.29, "learning_rate": 1.6648152499382694e-05, "loss": 0.9946, "step": 4364 }, { "epoch": 0.29, "learning_rate": 1.6646541973488274e-05, "loss": 0.9782, "step": 4365 }, { "epoch": 0.29, "learning_rate": 1.6644931138708322e-05, "loss": 1.0166, "step": 4366 }, { "epoch": 0.29, "learning_rate": 1.66433199951177e-05, "loss": 1.029, "step": 4367 }, { "epoch": 0.29, "learning_rate": 1.6641708542791285e-05, "loss": 0.9746, "step": 4368 }, { "epoch": 0.29, "learning_rate": 1.6640096781803966e-05, "loss": 1.0154, "step": 4369 }, { "epoch": 0.29, "learning_rate": 1.6638484712230635e-05, "loss": 0.961, "step": 4370 }, { "epoch": 0.29, "learning_rate": 1.6636872334146227e-05, "loss": 1.0242, "step": 4371 }, { "epoch": 0.29, "learning_rate": 1.663525964762566e-05, "loss": 0.9747, "step": 4372 }, { "epoch": 0.29, "learning_rate": 1.663364665274389e-05, "loss": 1.0435, "step": 4373 }, { "epoch": 0.29, "learning_rate": 1.6632033349575874e-05, "loss": 0.9808, "step": 4374 }, { "epoch": 0.29, "learning_rate": 1.6630419738196587e-05, "loss": 1.0047, "step": 4375 }, { "epoch": 0.29, "learning_rate": 1.6628805818681023e-05, "loss": 0.9889, "step": 4376 }, { "epoch": 0.29, "learning_rate": 1.662719159110418e-05, "loss": 1.0481, "step": 4377 }, { "epoch": 0.29, "learning_rate": 1.6625577055541074e-05, "loss": 1.0112, "step": 4378 }, { "epoch": 0.29, "learning_rate": 1.6623962212066748e-05, "loss": 1.0229, "step": 4379 }, { "epoch": 0.29, "learning_rate": 1.6622347060756237e-05, "loss": 0.9762, "step": 4380 }, { "epoch": 0.29, "learning_rate": 1.6620731601684612e-05, "loss": 1.0213, "step": 4381 }, { "epoch": 0.29, "learning_rate": 1.661911583492694e-05, "loss": 0.9957, "step": 4382 }, { "epoch": 0.29, "learning_rate": 1.6617499760558322e-05, "loss": 0.9948, "step": 4383 }, { "epoch": 0.29, "learning_rate": 1.6615883378653853e-05, "loss": 0.9824, "step": 4384 }, { "epoch": 0.29, "learning_rate": 1.661426668928865e-05, "loss": 1.0101, "step": 4385 }, { "epoch": 0.29, "learning_rate": 1.661264969253785e-05, "loss": 0.9743, "step": 4386 }, { "epoch": 0.29, "learning_rate": 1.66110323884766e-05, "loss": 1.0026, "step": 4387 }, { "epoch": 0.29, "learning_rate": 1.660941477718006e-05, "loss": 0.991, "step": 4388 }, { "epoch": 0.29, "learning_rate": 1.66077968587234e-05, "loss": 0.9814, "step": 4389 }, { "epoch": 0.29, "learning_rate": 1.660617863318182e-05, "loss": 0.9834, "step": 4390 }, { "epoch": 0.29, "learning_rate": 1.6604560100630524e-05, "loss": 1.0153, "step": 4391 }, { "epoch": 0.29, "learning_rate": 1.660294126114472e-05, "loss": 0.9677, "step": 4392 }, { "epoch": 0.29, "learning_rate": 1.6601322114799642e-05, "loss": 1.0068, "step": 4393 }, { "epoch": 0.29, "learning_rate": 1.6599702661670546e-05, "loss": 0.9881, "step": 4394 }, { "epoch": 0.29, "learning_rate": 1.6598082901832686e-05, "loss": 1.0242, "step": 4395 }, { "epoch": 0.29, "learning_rate": 1.6596462835361337e-05, "loss": 0.9739, "step": 4396 }, { "epoch": 0.29, "learning_rate": 1.6594842462331793e-05, "loss": 1.0301, "step": 4397 }, { "epoch": 0.29, "learning_rate": 1.6593221782819356e-05, "loss": 0.9356, "step": 4398 }, { "epoch": 0.29, "learning_rate": 1.659160079689934e-05, "loss": 1.0145, "step": 4399 }, { "epoch": 0.29, "learning_rate": 1.6589979504647086e-05, "loss": 1.0416, "step": 4400 }, { "epoch": 0.29, "learning_rate": 1.658835790613793e-05, "loss": 1.0081, "step": 4401 }, { "epoch": 0.29, "learning_rate": 1.658673600144724e-05, "loss": 1.0291, "step": 4402 }, { "epoch": 0.29, "learning_rate": 1.658511379065039e-05, "loss": 0.9973, "step": 4403 }, { "epoch": 0.29, "learning_rate": 1.6583491273822763e-05, "loss": 0.9905, "step": 4404 }, { "epoch": 0.29, "learning_rate": 1.6581868451039773e-05, "loss": 1.0142, "step": 4405 }, { "epoch": 0.29, "learning_rate": 1.6580245322376827e-05, "loss": 0.9786, "step": 4406 }, { "epoch": 0.29, "learning_rate": 1.6578621887909366e-05, "loss": 1.0079, "step": 4407 }, { "epoch": 0.29, "learning_rate": 1.657699814771283e-05, "loss": 1.0076, "step": 4408 }, { "epoch": 0.29, "learning_rate": 1.657537410186268e-05, "loss": 1.0041, "step": 4409 }, { "epoch": 0.29, "learning_rate": 1.6573749750434397e-05, "loss": 1.0069, "step": 4410 }, { "epoch": 0.29, "learning_rate": 1.657212509350346e-05, "loss": 1.031, "step": 4411 }, { "epoch": 0.29, "learning_rate": 1.6570500131145376e-05, "loss": 1.0134, "step": 4412 }, { "epoch": 0.29, "learning_rate": 1.6568874863435664e-05, "loss": 0.9997, "step": 4413 }, { "epoch": 0.29, "learning_rate": 1.6567249290449854e-05, "loss": 0.9811, "step": 4414 }, { "epoch": 0.29, "learning_rate": 1.656562341226349e-05, "loss": 0.9898, "step": 4415 }, { "epoch": 0.29, "learning_rate": 1.6563997228952133e-05, "loss": 0.9722, "step": 4416 }, { "epoch": 0.29, "learning_rate": 1.6562370740591355e-05, "loss": 1.0403, "step": 4417 }, { "epoch": 0.29, "learning_rate": 1.656074394725675e-05, "loss": 1.0323, "step": 4418 }, { "epoch": 0.29, "learning_rate": 1.655911684902391e-05, "loss": 1.0114, "step": 4419 }, { "epoch": 0.29, "learning_rate": 1.6557489445968456e-05, "loss": 1.0022, "step": 4420 }, { "epoch": 0.29, "learning_rate": 1.6555861738166027e-05, "loss": 0.9891, "step": 4421 }, { "epoch": 0.29, "learning_rate": 1.6554233725692254e-05, "loss": 1.014, "step": 4422 }, { "epoch": 0.29, "learning_rate": 1.6552605408622804e-05, "loss": 1.0135, "step": 4423 }, { "epoch": 0.29, "learning_rate": 1.6550976787033347e-05, "loss": 0.977, "step": 4424 }, { "epoch": 0.29, "learning_rate": 1.654934786099957e-05, "loss": 1.0024, "step": 4425 }, { "epoch": 0.29, "learning_rate": 1.6547718630597178e-05, "loss": 1.0088, "step": 4426 }, { "epoch": 0.29, "learning_rate": 1.6546089095901884e-05, "loss": 1.036, "step": 4427 }, { "epoch": 0.29, "learning_rate": 1.6544459256989415e-05, "loss": 0.9928, "step": 4428 }, { "epoch": 0.29, "learning_rate": 1.6542829113935516e-05, "loss": 1.0374, "step": 4429 }, { "epoch": 0.29, "learning_rate": 1.654119866681595e-05, "loss": 1.009, "step": 4430 }, { "epoch": 0.29, "learning_rate": 1.6539567915706482e-05, "loss": 1.0253, "step": 4431 }, { "epoch": 0.29, "learning_rate": 1.6537936860682898e-05, "loss": 0.9853, "step": 4432 }, { "epoch": 0.3, "learning_rate": 1.6536305501821007e-05, "loss": 0.9761, "step": 4433 }, { "epoch": 0.3, "learning_rate": 1.653467383919661e-05, "loss": 0.9624, "step": 4434 }, { "epoch": 0.3, "learning_rate": 1.653304187288555e-05, "loss": 1.0176, "step": 4435 }, { "epoch": 0.3, "learning_rate": 1.6531409602963656e-05, "loss": 1.0076, "step": 4436 }, { "epoch": 0.3, "learning_rate": 1.6529777029506795e-05, "loss": 0.9816, "step": 4437 }, { "epoch": 0.3, "learning_rate": 1.6528144152590836e-05, "loss": 1.0054, "step": 4438 }, { "epoch": 0.3, "learning_rate": 1.6526510972291657e-05, "loss": 1.028, "step": 4439 }, { "epoch": 0.3, "learning_rate": 1.6524877488685165e-05, "loss": 0.9847, "step": 4440 }, { "epoch": 0.3, "learning_rate": 1.6523243701847267e-05, "loss": 1.0182, "step": 4441 }, { "epoch": 0.3, "learning_rate": 1.6521609611853896e-05, "loss": 1.0146, "step": 4442 }, { "epoch": 0.3, "learning_rate": 1.651997521878099e-05, "loss": 1.0086, "step": 4443 }, { "epoch": 0.3, "learning_rate": 1.65183405227045e-05, "loss": 1.0097, "step": 4444 }, { "epoch": 0.3, "learning_rate": 1.65167055237004e-05, "loss": 1.0322, "step": 4445 }, { "epoch": 0.3, "learning_rate": 1.6515070221844683e-05, "loss": 0.944, "step": 4446 }, { "epoch": 0.3, "learning_rate": 1.651343461721333e-05, "loss": 1.0306, "step": 4447 }, { "epoch": 0.3, "learning_rate": 1.651179870988236e-05, "loss": 0.9621, "step": 4448 }, { "epoch": 0.3, "learning_rate": 1.6510162499927796e-05, "loss": 1.031, "step": 4449 }, { "epoch": 0.3, "learning_rate": 1.6508525987425685e-05, "loss": 1.0264, "step": 4450 }, { "epoch": 0.3, "learning_rate": 1.6506889172452072e-05, "loss": 0.9958, "step": 4451 }, { "epoch": 0.3, "learning_rate": 1.650525205508303e-05, "loss": 1.057, "step": 4452 }, { "epoch": 0.3, "learning_rate": 1.650361463539464e-05, "loss": 1.014, "step": 4453 }, { "epoch": 0.3, "learning_rate": 1.6501976913463004e-05, "loss": 1.0072, "step": 4454 }, { "epoch": 0.3, "learning_rate": 1.6500338889364218e-05, "loss": 1.006, "step": 4455 }, { "epoch": 0.3, "learning_rate": 1.6498700563174417e-05, "loss": 1.0014, "step": 4456 }, { "epoch": 0.3, "learning_rate": 1.6497061934969734e-05, "loss": 1.0423, "step": 4457 }, { "epoch": 0.3, "learning_rate": 1.6495423004826327e-05, "loss": 0.9863, "step": 4458 }, { "epoch": 0.3, "learning_rate": 1.6493783772820357e-05, "loss": 1.0016, "step": 4459 }, { "epoch": 0.3, "learning_rate": 1.6492144239028003e-05, "loss": 0.9929, "step": 4460 }, { "epoch": 0.3, "learning_rate": 1.6490504403525464e-05, "loss": 1.0137, "step": 4461 }, { "epoch": 0.3, "learning_rate": 1.6488864266388947e-05, "loss": 1.0174, "step": 4462 }, { "epoch": 0.3, "learning_rate": 1.6487223827694673e-05, "loss": 0.9839, "step": 4463 }, { "epoch": 0.3, "learning_rate": 1.6485583087518877e-05, "loss": 1.0049, "step": 4464 }, { "epoch": 0.3, "learning_rate": 1.6483942045937813e-05, "loss": 0.9883, "step": 4465 }, { "epoch": 0.3, "learning_rate": 1.6482300703027744e-05, "loss": 1.011, "step": 4466 }, { "epoch": 0.3, "learning_rate": 1.648065905886495e-05, "loss": 0.972, "step": 4467 }, { "epoch": 0.3, "learning_rate": 1.6479017113525714e-05, "loss": 0.9897, "step": 4468 }, { "epoch": 0.3, "learning_rate": 1.6477374867086355e-05, "loss": 1.0088, "step": 4469 }, { "epoch": 0.3, "learning_rate": 1.6475732319623186e-05, "loss": 0.9937, "step": 4470 }, { "epoch": 0.3, "learning_rate": 1.6474089471212545e-05, "loss": 0.9803, "step": 4471 }, { "epoch": 0.3, "learning_rate": 1.6472446321930776e-05, "loss": 0.9959, "step": 4472 }, { "epoch": 0.3, "learning_rate": 1.6470802871854242e-05, "loss": 0.9826, "step": 4473 }, { "epoch": 0.3, "learning_rate": 1.6469159121059324e-05, "loss": 1.0102, "step": 4474 }, { "epoch": 0.3, "learning_rate": 1.6467515069622407e-05, "loss": 0.9766, "step": 4475 }, { "epoch": 0.3, "learning_rate": 1.6465870717619898e-05, "loss": 1.0216, "step": 4476 }, { "epoch": 0.3, "learning_rate": 1.6464226065128216e-05, "loss": 1.051, "step": 4477 }, { "epoch": 0.3, "learning_rate": 1.646258111222379e-05, "loss": 0.9671, "step": 4478 }, { "epoch": 0.3, "learning_rate": 1.6460935858983067e-05, "loss": 1.0056, "step": 4479 }, { "epoch": 0.3, "learning_rate": 1.645929030548251e-05, "loss": 0.9873, "step": 4480 }, { "epoch": 0.3, "learning_rate": 1.645764445179859e-05, "loss": 1.0417, "step": 4481 }, { "epoch": 0.3, "learning_rate": 1.6455998298007794e-05, "loss": 0.9917, "step": 4482 }, { "epoch": 0.3, "learning_rate": 1.6454351844186624e-05, "loss": 1.0417, "step": 4483 }, { "epoch": 0.3, "learning_rate": 1.6452705090411602e-05, "loss": 1.0231, "step": 4484 }, { "epoch": 0.3, "learning_rate": 1.6451058036759252e-05, "loss": 1.0108, "step": 4485 }, { "epoch": 0.3, "learning_rate": 1.6449410683306116e-05, "loss": 1.0008, "step": 4486 }, { "epoch": 0.3, "learning_rate": 1.6447763030128756e-05, "loss": 0.9901, "step": 4487 }, { "epoch": 0.3, "learning_rate": 1.6446115077303743e-05, "loss": 0.9546, "step": 4488 }, { "epoch": 0.3, "learning_rate": 1.6444466824907662e-05, "loss": 1.0124, "step": 4489 }, { "epoch": 0.3, "learning_rate": 1.644281827301711e-05, "loss": 1.0157, "step": 4490 }, { "epoch": 0.3, "learning_rate": 1.6441169421708705e-05, "loss": 1.0365, "step": 4491 }, { "epoch": 0.3, "learning_rate": 1.643952027105907e-05, "loss": 0.9488, "step": 4492 }, { "epoch": 0.3, "learning_rate": 1.6437870821144847e-05, "loss": 1.0171, "step": 4493 }, { "epoch": 0.3, "learning_rate": 1.6436221072042694e-05, "loss": 1.0133, "step": 4494 }, { "epoch": 0.3, "learning_rate": 1.6434571023829277e-05, "loss": 1.0547, "step": 4495 }, { "epoch": 0.3, "learning_rate": 1.643292067658128e-05, "loss": 0.9702, "step": 4496 }, { "epoch": 0.3, "learning_rate": 1.64312700303754e-05, "loss": 0.9936, "step": 4497 }, { "epoch": 0.3, "learning_rate": 1.6429619085288345e-05, "loss": 0.9964, "step": 4498 }, { "epoch": 0.3, "learning_rate": 1.6427967841396847e-05, "loss": 1.042, "step": 4499 }, { "epoch": 0.3, "learning_rate": 1.6426316298777634e-05, "loss": 0.9956, "step": 4500 }, { "epoch": 0.3, "learning_rate": 1.6424664457507466e-05, "loss": 1.018, "step": 4501 }, { "epoch": 0.3, "learning_rate": 1.6423012317663106e-05, "loss": 0.9766, "step": 4502 }, { "epoch": 0.3, "learning_rate": 1.642135987932133e-05, "loss": 0.9955, "step": 4503 }, { "epoch": 0.3, "learning_rate": 1.6419707142558946e-05, "loss": 1.017, "step": 4504 }, { "epoch": 0.3, "learning_rate": 1.641805410745275e-05, "loss": 1.0099, "step": 4505 }, { "epoch": 0.3, "learning_rate": 1.6416400774079562e-05, "loss": 1.0235, "step": 4506 }, { "epoch": 0.3, "learning_rate": 1.6414747142516227e-05, "loss": 1.0053, "step": 4507 }, { "epoch": 0.3, "learning_rate": 1.6413093212839587e-05, "loss": 1.0174, "step": 4508 }, { "epoch": 0.3, "learning_rate": 1.6411438985126508e-05, "loss": 1.0245, "step": 4509 }, { "epoch": 0.3, "learning_rate": 1.6409784459453865e-05, "loss": 0.9978, "step": 4510 }, { "epoch": 0.3, "learning_rate": 1.6408129635898556e-05, "loss": 1.013, "step": 4511 }, { "epoch": 0.3, "learning_rate": 1.6406474514537477e-05, "loss": 1.0139, "step": 4512 }, { "epoch": 0.0, "learning_rate": 1.640481909544755e-05, "loss": 1.06, "step": 4513 }, { "epoch": 0.0, "learning_rate": 1.6403163378705708e-05, "loss": 1.0141, "step": 4514 }, { "epoch": 0.0, "learning_rate": 1.6401507364388896e-05, "loss": 1.0126, "step": 4515 }, { "epoch": 0.0, "learning_rate": 1.6399851052574076e-05, "loss": 1.0049, "step": 4516 }, { "epoch": 0.0, "learning_rate": 1.639819444333822e-05, "loss": 0.9987, "step": 4517 }, { "epoch": 0.0, "learning_rate": 1.639653753675832e-05, "loss": 1.0151, "step": 4518 }, { "epoch": 0.0, "learning_rate": 1.639488033291137e-05, "loss": 1.004, "step": 4519 }, { "epoch": 0.0, "learning_rate": 1.6393222831874392e-05, "loss": 0.9905, "step": 4520 }, { "epoch": 0.0, "learning_rate": 1.6391565033724415e-05, "loss": 1.0433, "step": 4521 }, { "epoch": 0.0, "learning_rate": 1.638990693853848e-05, "loss": 1.0162, "step": 4522 }, { "epoch": 0.0, "learning_rate": 1.6388248546393644e-05, "loss": 1.0252, "step": 4523 }, { "epoch": 0.0, "learning_rate": 1.6386589857366976e-05, "loss": 1.0303, "step": 4524 }, { "epoch": 0.0, "learning_rate": 1.6384930871535566e-05, "loss": 1.0109, "step": 4525 }, { "epoch": 0.0, "learning_rate": 1.6383271588976503e-05, "loss": 1.0476, "step": 4526 }, { "epoch": 0.0, "learning_rate": 1.638161200976691e-05, "loss": 1.038, "step": 4527 }, { "epoch": 0.0, "learning_rate": 1.6379952133983903e-05, "loss": 1.0324, "step": 4528 }, { "epoch": 0.0, "learning_rate": 1.6378291961704628e-05, "loss": 1.0263, "step": 4529 }, { "epoch": 0.0, "learning_rate": 1.6376631493006233e-05, "loss": 1.009, "step": 4530 }, { "epoch": 0.0, "learning_rate": 1.637497072796589e-05, "loss": 1.0379, "step": 4531 }, { "epoch": 0.0, "learning_rate": 1.637330966666078e-05, "loss": 1.0164, "step": 4532 }, { "epoch": 0.0, "learning_rate": 1.63716483091681e-05, "loss": 1.0003, "step": 4533 }, { "epoch": 0.0, "learning_rate": 1.636998665556505e-05, "loss": 0.9941, "step": 4534 }, { "epoch": 0.0, "learning_rate": 1.6368324705928852e-05, "loss": 1.0197, "step": 4535 }, { "epoch": 0.0, "learning_rate": 1.636666246033675e-05, "loss": 0.9512, "step": 4536 }, { "epoch": 0.0, "learning_rate": 1.6364999918865997e-05, "loss": 0.9794, "step": 4537 }, { "epoch": 0.0, "learning_rate": 1.6363337081593848e-05, "loss": 0.9846, "step": 4538 }, { "epoch": 0.0, "learning_rate": 1.6361673948597577e-05, "loss": 1.0149, "step": 4539 }, { "epoch": 0.0, "learning_rate": 1.6360010519954485e-05, "loss": 0.9978, "step": 4540 }, { "epoch": 0.0, "learning_rate": 1.6358346795741868e-05, "loss": 1.0344, "step": 4541 }, { "epoch": 0.0, "learning_rate": 1.635668277603705e-05, "loss": 1.0047, "step": 4542 }, { "epoch": 0.0, "learning_rate": 1.635501846091736e-05, "loss": 1.0654, "step": 4543 }, { "epoch": 0.0, "learning_rate": 1.6353353850460147e-05, "loss": 1.0174, "step": 4544 }, { "epoch": 0.0, "learning_rate": 1.6351688944742767e-05, "loss": 1.0415, "step": 4545 }, { "epoch": 0.0, "learning_rate": 1.63500237438426e-05, "loss": 1.021, "step": 4546 }, { "epoch": 0.0, "learning_rate": 1.6348358247837025e-05, "loss": 1.0167, "step": 4547 }, { "epoch": 0.0, "learning_rate": 1.6346692456803447e-05, "loss": 0.99, "step": 4548 }, { "epoch": 0.0, "learning_rate": 1.6345026370819282e-05, "loss": 1.0036, "step": 4549 }, { "epoch": 0.0, "learning_rate": 1.6343359989961952e-05, "loss": 1.0201, "step": 4550 }, { "epoch": 0.0, "learning_rate": 1.6341693314308906e-05, "loss": 1.0063, "step": 4551 }, { "epoch": 0.0, "learning_rate": 1.6340026343937597e-05, "loss": 1.0199, "step": 4552 }, { "epoch": 0.0, "learning_rate": 1.633835907892549e-05, "loss": 0.9989, "step": 4553 }, { "epoch": 0.0, "learning_rate": 1.6336691519350077e-05, "loss": 1.0091, "step": 4554 }, { "epoch": 0.0, "learning_rate": 1.6335023665288847e-05, "loss": 1.023, "step": 4555 }, { "epoch": 0.0, "learning_rate": 1.633335551681931e-05, "loss": 1.0189, "step": 4556 }, { "epoch": 0.0, "learning_rate": 1.6331687074018997e-05, "loss": 1.0299, "step": 4557 }, { "epoch": 0.0, "learning_rate": 1.633001833696544e-05, "loss": 1.0125, "step": 4558 }, { "epoch": 0.0, "learning_rate": 1.6328349305736195e-05, "loss": 1.0185, "step": 4559 }, { "epoch": 0.0, "learning_rate": 1.6326679980408822e-05, "loss": 0.9969, "step": 4560 }, { "epoch": 0.0, "learning_rate": 1.6325010361060904e-05, "loss": 1.0225, "step": 4561 }, { "epoch": 0.0, "learning_rate": 1.6323340447770028e-05, "loss": 0.9857, "step": 4562 }, { "epoch": 0.0, "learning_rate": 1.6321670240613805e-05, "loss": 1.0001, "step": 4563 }, { "epoch": 0.0, "learning_rate": 1.6319999739669854e-05, "loss": 0.9888, "step": 4564 }, { "epoch": 0.0, "learning_rate": 1.6318328945015807e-05, "loss": 0.9852, "step": 4565 }, { "epoch": 0.0, "learning_rate": 1.631665785672931e-05, "loss": 0.9834, "step": 4566 }, { "epoch": 0.0, "learning_rate": 1.631498647488803e-05, "loss": 1.0097, "step": 4567 }, { "epoch": 0.0, "learning_rate": 1.6313314799569632e-05, "loss": 0.9576, "step": 4568 }, { "epoch": 0.0, "learning_rate": 1.6311642830851812e-05, "loss": 1.0433, "step": 4569 }, { "epoch": 0.0, "learning_rate": 1.6309970568812266e-05, "loss": 1.0107, "step": 4570 }, { "epoch": 0.0, "learning_rate": 1.630829801352871e-05, "loss": 1.0092, "step": 4571 }, { "epoch": 0.0, "learning_rate": 1.6306625165078877e-05, "loss": 1.0368, "step": 4572 }, { "epoch": 0.0, "learning_rate": 1.630495202354051e-05, "loss": 0.9969, "step": 4573 }, { "epoch": 0.0, "learning_rate": 1.6303278588991356e-05, "loss": 0.9896, "step": 4574 }, { "epoch": 0.0, "learning_rate": 1.6301604861509192e-05, "loss": 1.0049, "step": 4575 }, { "epoch": 0.0, "learning_rate": 1.6299930841171803e-05, "loss": 1.0274, "step": 4576 }, { "epoch": 0.0, "learning_rate": 1.629825652805698e-05, "loss": 1.0161, "step": 4577 }, { "epoch": 0.0, "learning_rate": 1.6296581922242535e-05, "loss": 0.983, "step": 4578 }, { "epoch": 0.0, "learning_rate": 1.62949070238063e-05, "loss": 1.0307, "step": 4579 }, { "epoch": 0.0, "learning_rate": 1.62932318328261e-05, "loss": 0.984, "step": 4580 }, { "epoch": 0.0, "learning_rate": 1.6291556349379794e-05, "loss": 1.0493, "step": 4581 }, { "epoch": 0.0, "learning_rate": 1.6289880573545246e-05, "loss": 0.9552, "step": 4582 }, { "epoch": 0.0, "learning_rate": 1.6288204505400336e-05, "loss": 1.0175, "step": 4583 }, { "epoch": 0.0, "learning_rate": 1.6286528145022954e-05, "loss": 0.9555, "step": 4584 }, { "epoch": 0.0, "learning_rate": 1.6284851492491007e-05, "loss": 1.0245, "step": 4585 }, { "epoch": 0.0, "learning_rate": 1.628317454788241e-05, "loss": 1.0128, "step": 4586 }, { "epoch": 0.0, "learning_rate": 1.62814973112751e-05, "loss": 1.0362, "step": 4587 }, { "epoch": 0.01, "learning_rate": 1.6279819782747024e-05, "loss": 0.9967, "step": 4588 }, { "epoch": 0.01, "learning_rate": 1.627814196237614e-05, "loss": 1.0107, "step": 4589 }, { "epoch": 0.01, "learning_rate": 1.6276463850240422e-05, "loss": 1.0073, "step": 4590 }, { "epoch": 0.01, "learning_rate": 1.6274785446417857e-05, "loss": 1.0275, "step": 4591 }, { "epoch": 0.01, "learning_rate": 1.6273106750986446e-05, "loss": 1.0026, "step": 4592 }, { "epoch": 0.01, "learning_rate": 1.6271427764024205e-05, "loss": 1.0147, "step": 4593 }, { "epoch": 0.01, "learning_rate": 1.6269748485609155e-05, "loss": 1.0142, "step": 4594 }, { "epoch": 0.01, "learning_rate": 1.6268068915819344e-05, "loss": 0.9922, "step": 4595 }, { "epoch": 0.01, "learning_rate": 1.6266389054732827e-05, "loss": 0.9945, "step": 4596 }, { "epoch": 0.01, "learning_rate": 1.6264708902427668e-05, "loss": 1.0191, "step": 4597 }, { "epoch": 0.01, "learning_rate": 1.626302845898195e-05, "loss": 0.9564, "step": 4598 }, { "epoch": 0.01, "learning_rate": 1.6261347724473776e-05, "loss": 1.0281, "step": 4599 }, { "epoch": 0.01, "learning_rate": 1.625966669898124e-05, "loss": 0.9735, "step": 4600 }, { "epoch": 0.01, "learning_rate": 1.625798538258248e-05, "loss": 1.0289, "step": 4601 }, { "epoch": 0.01, "learning_rate": 1.625630377535562e-05, "loss": 1.0083, "step": 4602 }, { "epoch": 0.01, "learning_rate": 1.625462187737882e-05, "loss": 1.0158, "step": 4603 }, { "epoch": 0.01, "learning_rate": 1.6252939688730236e-05, "loss": 0.9639, "step": 4604 }, { "epoch": 0.01, "learning_rate": 1.6251257209488047e-05, "loss": 0.9817, "step": 4605 }, { "epoch": 0.01, "learning_rate": 1.624957443973044e-05, "loss": 0.9718, "step": 4606 }, { "epoch": 0.01, "learning_rate": 1.6247891379535622e-05, "loss": 1.016, "step": 4607 }, { "epoch": 0.01, "learning_rate": 1.624620802898181e-05, "loss": 0.9773, "step": 4608 }, { "epoch": 0.01, "learning_rate": 1.6244524388147233e-05, "loss": 1.0186, "step": 4609 }, { "epoch": 0.01, "learning_rate": 1.6242840457110135e-05, "loss": 0.9893, "step": 4610 }, { "epoch": 0.01, "learning_rate": 1.6241156235948775e-05, "loss": 1.0103, "step": 4611 }, { "epoch": 0.01, "learning_rate": 1.623947172474142e-05, "loss": 0.9775, "step": 4612 }, { "epoch": 0.01, "learning_rate": 1.623778692356636e-05, "loss": 0.9729, "step": 4613 }, { "epoch": 0.01, "learning_rate": 1.6236101832501892e-05, "loss": 0.9612, "step": 4614 }, { "epoch": 0.01, "learning_rate": 1.6234416451626325e-05, "loss": 1.0123, "step": 4615 }, { "epoch": 0.01, "learning_rate": 1.623273078101798e-05, "loss": 0.9976, "step": 4616 }, { "epoch": 0.01, "learning_rate": 1.6231044820755205e-05, "loss": 1.0063, "step": 4617 }, { "epoch": 0.01, "learning_rate": 1.622935857091635e-05, "loss": 0.9516, "step": 4618 }, { "epoch": 0.01, "learning_rate": 1.6227672031579774e-05, "loss": 1.0269, "step": 4619 }, { "epoch": 0.01, "learning_rate": 1.6225985202823858e-05, "loss": 1.0307, "step": 4620 }, { "epoch": 0.01, "learning_rate": 1.6224298084726996e-05, "loss": 1.0001, "step": 4621 }, { "epoch": 0.01, "learning_rate": 1.6222610677367596e-05, "loss": 1.024, "step": 4622 }, { "epoch": 0.01, "learning_rate": 1.622092298082407e-05, "loss": 1.0318, "step": 4623 }, { "epoch": 0.01, "learning_rate": 1.6219234995174858e-05, "loss": 0.9711, "step": 4624 }, { "epoch": 0.01, "learning_rate": 1.62175467204984e-05, "loss": 0.9743, "step": 4625 }, { "epoch": 0.01, "learning_rate": 1.6215858156873155e-05, "loss": 0.9684, "step": 4626 }, { "epoch": 0.01, "learning_rate": 1.6214169304377607e-05, "loss": 1.004, "step": 4627 }, { "epoch": 0.01, "learning_rate": 1.6212480163090227e-05, "loss": 0.9856, "step": 4628 }, { "epoch": 0.01, "learning_rate": 1.6210790733089523e-05, "loss": 1.0032, "step": 4629 }, { "epoch": 0.01, "learning_rate": 1.620910101445401e-05, "loss": 0.988, "step": 4630 }, { "epoch": 0.01, "learning_rate": 1.6207411007262207e-05, "loss": 1.0011, "step": 4631 }, { "epoch": 0.01, "learning_rate": 1.620572071159266e-05, "loss": 0.9954, "step": 4632 }, { "epoch": 0.01, "learning_rate": 1.6204030127523922e-05, "loss": 0.9997, "step": 4633 }, { "epoch": 0.01, "learning_rate": 1.6202339255134556e-05, "loss": 0.9565, "step": 4634 }, { "epoch": 0.01, "learning_rate": 1.6200648094503144e-05, "loss": 1.0009, "step": 4635 }, { "epoch": 0.01, "learning_rate": 1.6198956645708278e-05, "loss": 1.0302, "step": 4636 }, { "epoch": 0.01, "learning_rate": 1.619726490882857e-05, "loss": 0.9803, "step": 4637 }, { "epoch": 0.01, "learning_rate": 1.619557288394264e-05, "loss": 0.9596, "step": 4638 }, { "epoch": 0.01, "learning_rate": 1.619388057112911e-05, "loss": 1.0232, "step": 4639 }, { "epoch": 0.01, "learning_rate": 1.6192187970466646e-05, "loss": 0.9655, "step": 4640 }, { "epoch": 0.01, "learning_rate": 1.619049508203389e-05, "loss": 0.9943, "step": 4641 }, { "epoch": 0.01, "learning_rate": 1.6188801905909526e-05, "loss": 1.0209, "step": 4642 }, { "epoch": 0.01, "learning_rate": 1.618710844217224e-05, "loss": 1.0106, "step": 4643 }, { "epoch": 0.01, "learning_rate": 1.618541469090073e-05, "loss": 1.0219, "step": 4644 }, { "epoch": 0.01, "learning_rate": 1.6183720652173716e-05, "loss": 0.9874, "step": 4645 }, { "epoch": 0.01, "learning_rate": 1.6182026326069913e-05, "loss": 0.9802, "step": 4646 }, { "epoch": 0.01, "learning_rate": 1.6180331712668077e-05, "loss": 0.9616, "step": 4647 }, { "epoch": 0.01, "learning_rate": 1.617863681204695e-05, "loss": 0.973, "step": 4648 }, { "epoch": 0.01, "learning_rate": 1.6176941624285305e-05, "loss": 1.0108, "step": 4649 }, { "epoch": 0.01, "learning_rate": 1.617524614946192e-05, "loss": 1.0259, "step": 4650 }, { "epoch": 0.01, "learning_rate": 1.6173550387655593e-05, "loss": 1.0049, "step": 4651 }, { "epoch": 0.01, "learning_rate": 1.6171854338945124e-05, "loss": 0.9812, "step": 4652 }, { "epoch": 0.01, "learning_rate": 1.6170158003409343e-05, "loss": 1.0527, "step": 4653 }, { "epoch": 0.01, "learning_rate": 1.616846138112708e-05, "loss": 0.9769, "step": 4654 }, { "epoch": 0.01, "learning_rate": 1.616676447217718e-05, "loss": 0.9805, "step": 4655 }, { "epoch": 0.01, "learning_rate": 1.6165067276638503e-05, "loss": 0.9413, "step": 4656 }, { "epoch": 0.01, "learning_rate": 1.6163369794589926e-05, "loss": 1.0211, "step": 4657 }, { "epoch": 0.01, "learning_rate": 1.6161672026110338e-05, "loss": 0.9914, "step": 4658 }, { "epoch": 0.01, "learning_rate": 1.6159973971278637e-05, "loss": 1.0031, "step": 4659 }, { "epoch": 0.01, "learning_rate": 1.6158275630173735e-05, "loss": 1.0004, "step": 4660 }, { "epoch": 0.01, "learning_rate": 1.6156577002874566e-05, "loss": 1.0352, "step": 4661 }, { "epoch": 0.01, "learning_rate": 1.615487808946006e-05, "loss": 0.9629, "step": 4662 }, { "epoch": 0.01, "learning_rate": 1.615317889000918e-05, "loss": 1.013, "step": 4663 }, { "epoch": 0.01, "learning_rate": 1.6151479404600888e-05, "loss": 0.9907, "step": 4664 }, { "epoch": 0.01, "learning_rate": 1.6149779633314166e-05, "loss": 1.0214, "step": 4665 }, { "epoch": 0.01, "learning_rate": 1.6148079576228008e-05, "loss": 0.9892, "step": 4666 }, { "epoch": 0.01, "learning_rate": 1.614637923342142e-05, "loss": 0.9803, "step": 4667 }, { "epoch": 0.01, "learning_rate": 1.6144678604973426e-05, "loss": 1.0241, "step": 4668 }, { "epoch": 0.01, "learning_rate": 1.6142977690963056e-05, "loss": 0.9947, "step": 4669 }, { "epoch": 0.01, "learning_rate": 1.6141276491469355e-05, "loss": 0.9951, "step": 4670 }, { "epoch": 0.01, "learning_rate": 1.6139575006571385e-05, "loss": 1.0179, "step": 4671 }, { "epoch": 0.01, "learning_rate": 1.613787323634822e-05, "loss": 0.9952, "step": 4672 }, { "epoch": 0.01, "learning_rate": 1.6136171180878946e-05, "loss": 0.9893, "step": 4673 }, { "epoch": 0.01, "learning_rate": 1.6134468840242666e-05, "loss": 1.0229, "step": 4674 }, { "epoch": 0.01, "learning_rate": 1.6132766214518486e-05, "loss": 1.0454, "step": 4675 }, { "epoch": 0.01, "learning_rate": 1.6131063303785542e-05, "loss": 0.9792, "step": 4676 }, { "epoch": 0.01, "learning_rate": 1.6129360108122965e-05, "loss": 1.0731, "step": 4677 }, { "epoch": 0.01, "learning_rate": 1.612765662760991e-05, "loss": 1.0634, "step": 4678 }, { "epoch": 0.01, "learning_rate": 1.612595286232554e-05, "loss": 0.9589, "step": 4679 }, { "epoch": 0.01, "learning_rate": 1.6124248812349044e-05, "loss": 0.9781, "step": 4680 }, { "epoch": 0.01, "learning_rate": 1.6122544477759604e-05, "loss": 1.0212, "step": 4681 }, { "epoch": 0.01, "learning_rate": 1.6120839858636435e-05, "loss": 1.0099, "step": 4682 }, { "epoch": 0.01, "learning_rate": 1.6119134955058748e-05, "loss": 0.9982, "step": 4683 }, { "epoch": 0.01, "learning_rate": 1.6117429767105782e-05, "loss": 0.9566, "step": 4684 }, { "epoch": 0.01, "learning_rate": 1.6115724294856775e-05, "loss": 1.0296, "step": 4685 }, { "epoch": 0.01, "learning_rate": 1.6114018538390995e-05, "loss": 0.9798, "step": 4686 }, { "epoch": 0.01, "learning_rate": 1.6112312497787705e-05, "loss": 1.0094, "step": 4687 }, { "epoch": 0.01, "learning_rate": 1.611060617312619e-05, "loss": 0.9759, "step": 4688 }, { "epoch": 0.01, "learning_rate": 1.6108899564485753e-05, "loss": 1.0179, "step": 4689 }, { "epoch": 0.01, "learning_rate": 1.6107192671945706e-05, "loss": 0.9624, "step": 4690 }, { "epoch": 0.01, "learning_rate": 1.6105485495585373e-05, "loss": 0.9966, "step": 4691 }, { "epoch": 0.01, "learning_rate": 1.610377803548409e-05, "loss": 1.0016, "step": 4692 }, { "epoch": 0.01, "learning_rate": 1.6102070291721207e-05, "loss": 1.0244, "step": 4693 }, { "epoch": 0.01, "learning_rate": 1.610036226437609e-05, "loss": 0.981, "step": 4694 }, { "epoch": 0.01, "learning_rate": 1.6098653953528116e-05, "loss": 0.9864, "step": 4695 }, { "epoch": 0.01, "learning_rate": 1.6096945359256676e-05, "loss": 1.0346, "step": 4696 }, { "epoch": 0.01, "learning_rate": 1.609523648164117e-05, "loss": 1.0423, "step": 4697 }, { "epoch": 0.01, "learning_rate": 1.609352732076102e-05, "loss": 0.9619, "step": 4698 }, { "epoch": 0.01, "learning_rate": 1.6091817876695655e-05, "loss": 1.0194, "step": 4699 }, { "epoch": 0.01, "learning_rate": 1.609010814952452e-05, "loss": 0.9776, "step": 4700 }, { "epoch": 0.01, "learning_rate": 1.608839813932707e-05, "loss": 1.0012, "step": 4701 }, { "epoch": 0.01, "learning_rate": 1.6086687846182764e-05, "loss": 0.9743, "step": 4702 }, { "epoch": 0.01, "learning_rate": 1.60849772701711e-05, "loss": 1.0368, "step": 4703 }, { "epoch": 0.01, "learning_rate": 1.608326641137157e-05, "loss": 0.9782, "step": 4704 }, { "epoch": 0.01, "learning_rate": 1.6081555269863676e-05, "loss": 1.0315, "step": 4705 }, { "epoch": 0.01, "learning_rate": 1.6079843845726943e-05, "loss": 0.9882, "step": 4706 }, { "epoch": 0.01, "learning_rate": 1.607813213904091e-05, "loss": 1.0421, "step": 4707 }, { "epoch": 0.01, "learning_rate": 1.6076420149885125e-05, "loss": 1.0105, "step": 4708 }, { "epoch": 0.01, "learning_rate": 1.607470787833915e-05, "loss": 0.9924, "step": 4709 }, { "epoch": 0.01, "learning_rate": 1.6072995324482553e-05, "loss": 0.9995, "step": 4710 }, { "epoch": 0.01, "learning_rate": 1.6071282488394925e-05, "loss": 0.9842, "step": 4711 }, { "epoch": 0.01, "learning_rate": 1.606956937015587e-05, "loss": 0.9785, "step": 4712 }, { "epoch": 0.01, "learning_rate": 1.6067855969845002e-05, "loss": 1.0024, "step": 4713 }, { "epoch": 0.01, "learning_rate": 1.6066142287541943e-05, "loss": 0.9382, "step": 4714 }, { "epoch": 0.01, "learning_rate": 1.606442832332634e-05, "loss": 0.9602, "step": 4715 }, { "epoch": 0.01, "learning_rate": 1.6062714077277837e-05, "loss": 0.9564, "step": 4716 }, { "epoch": 0.01, "learning_rate": 1.606099954947611e-05, "loss": 0.9821, "step": 4717 }, { "epoch": 0.01, "learning_rate": 1.6059284740000827e-05, "loss": 1.0041, "step": 4718 }, { "epoch": 0.01, "learning_rate": 1.6057569648931696e-05, "loss": 1.0277, "step": 4719 }, { "epoch": 0.01, "learning_rate": 1.605585427634841e-05, "loss": 0.9308, "step": 4720 }, { "epoch": 0.01, "learning_rate": 1.6054138622330692e-05, "loss": 1.0218, "step": 4721 }, { "epoch": 0.01, "learning_rate": 1.6052422686958275e-05, "loss": 0.9947, "step": 4722 }, { "epoch": 0.01, "learning_rate": 1.6050706470310902e-05, "loss": 1.0163, "step": 4723 }, { "epoch": 0.01, "learning_rate": 1.6048989972468333e-05, "loss": 0.948, "step": 4724 }, { "epoch": 0.01, "learning_rate": 1.6047273193510332e-05, "loss": 0.9975, "step": 4725 }, { "epoch": 0.01, "learning_rate": 1.6045556133516696e-05, "loss": 0.9814, "step": 4726 }, { "epoch": 0.01, "learning_rate": 1.6043838792567212e-05, "loss": 0.9801, "step": 4727 }, { "epoch": 0.01, "learning_rate": 1.604212117074169e-05, "loss": 1.0004, "step": 4728 }, { "epoch": 0.01, "learning_rate": 1.6040403268119963e-05, "loss": 1.0213, "step": 4729 }, { "epoch": 0.01, "learning_rate": 1.6038685084781854e-05, "loss": 1.0289, "step": 4730 }, { "epoch": 0.01, "learning_rate": 1.603696662080722e-05, "loss": 0.9588, "step": 4731 }, { "epoch": 0.01, "learning_rate": 1.6035247876275927e-05, "loss": 0.9893, "step": 4732 }, { "epoch": 0.01, "learning_rate": 1.6033528851267842e-05, "loss": 0.9867, "step": 4733 }, { "epoch": 0.01, "learning_rate": 1.6031809545862857e-05, "loss": 0.9701, "step": 4734 }, { "epoch": 0.01, "learning_rate": 1.6030089960140873e-05, "loss": 0.978, "step": 4735 }, { "epoch": 0.01, "learning_rate": 1.6028370094181805e-05, "loss": 1.0235, "step": 4736 }, { "epoch": 0.01, "learning_rate": 1.6026649948065584e-05, "loss": 0.9904, "step": 4737 }, { "epoch": 0.02, "learning_rate": 1.6024929521872143e-05, "loss": 0.9285, "step": 4738 }, { "epoch": 0.02, "learning_rate": 1.6023208815681444e-05, "loss": 0.9959, "step": 4739 }, { "epoch": 0.02, "learning_rate": 1.6021487829573447e-05, "loss": 0.9726, "step": 4740 }, { "epoch": 0.02, "learning_rate": 1.6019766563628133e-05, "loss": 0.997, "step": 4741 }, { "epoch": 0.02, "learning_rate": 1.6018045017925495e-05, "loss": 1.0072, "step": 4742 }, { "epoch": 0.02, "learning_rate": 1.601632319254554e-05, "loss": 1.0223, "step": 4743 }, { "epoch": 0.02, "learning_rate": 1.601460108756829e-05, "loss": 0.9686, "step": 4744 }, { "epoch": 0.02, "learning_rate": 1.601287870307377e-05, "loss": 0.9999, "step": 4745 }, { "epoch": 0.02, "learning_rate": 1.6011156039142022e-05, "loss": 1.0165, "step": 4746 }, { "epoch": 0.02, "learning_rate": 1.6009433095853115e-05, "loss": 1.0161, "step": 4747 }, { "epoch": 0.02, "learning_rate": 1.6007709873287107e-05, "loss": 0.938, "step": 4748 }, { "epoch": 0.02, "learning_rate": 1.600598637152409e-05, "loss": 0.9707, "step": 4749 }, { "epoch": 0.02, "learning_rate": 1.600426259064416e-05, "loss": 0.9793, "step": 4750 }, { "epoch": 0.02, "learning_rate": 1.600253853072742e-05, "loss": 1.0014, "step": 4751 }, { "epoch": 0.02, "learning_rate": 1.6000814191854002e-05, "loss": 0.9866, "step": 4752 }, { "epoch": 0.02, "learning_rate": 1.5999089574104038e-05, "loss": 0.9997, "step": 4753 }, { "epoch": 0.02, "learning_rate": 1.599736467755767e-05, "loss": 1.0136, "step": 4754 }, { "epoch": 0.02, "learning_rate": 1.5995639502295067e-05, "loss": 0.9953, "step": 4755 }, { "epoch": 0.02, "learning_rate": 1.5993914048396394e-05, "loss": 0.9851, "step": 4756 }, { "epoch": 0.02, "learning_rate": 1.5992188315941852e-05, "loss": 0.9981, "step": 4757 }, { "epoch": 0.02, "learning_rate": 1.599046230501163e-05, "loss": 0.9404, "step": 4758 }, { "epoch": 0.02, "learning_rate": 1.5988736015685942e-05, "loss": 1.0225, "step": 4759 }, { "epoch": 0.02, "learning_rate": 1.5987009448045022e-05, "loss": 0.9393, "step": 4760 }, { "epoch": 0.02, "learning_rate": 1.5985282602169104e-05, "loss": 1.0077, "step": 4761 }, { "epoch": 0.02, "learning_rate": 1.598355547813844e-05, "loss": 0.9878, "step": 4762 }, { "epoch": 0.02, "learning_rate": 1.598182807603329e-05, "loss": 1.016, "step": 4763 }, { "epoch": 0.02, "learning_rate": 1.598010039593394e-05, "loss": 0.9717, "step": 4764 }, { "epoch": 0.02, "learning_rate": 1.597837243792067e-05, "loss": 1.0164, "step": 4765 }, { "epoch": 0.02, "learning_rate": 1.59766442020738e-05, "loss": 0.9543, "step": 4766 }, { "epoch": 0.02, "learning_rate": 1.5974915688473634e-05, "loss": 0.9861, "step": 4767 }, { "epoch": 0.02, "learning_rate": 1.5973186897200506e-05, "loss": 0.9568, "step": 4768 }, { "epoch": 0.02, "learning_rate": 1.5971457828334756e-05, "loss": 1.0306, "step": 4769 }, { "epoch": 0.02, "learning_rate": 1.596972848195674e-05, "loss": 0.9669, "step": 4770 }, { "epoch": 0.02, "learning_rate": 1.5967998858146824e-05, "loss": 1.0187, "step": 4771 }, { "epoch": 0.02, "learning_rate": 1.5966268956985392e-05, "loss": 0.992, "step": 4772 }, { "epoch": 0.02, "learning_rate": 1.5964538778552837e-05, "loss": 0.9711, "step": 4773 }, { "epoch": 0.02, "learning_rate": 1.5962808322929567e-05, "loss": 1.0468, "step": 4774 }, { "epoch": 0.02, "learning_rate": 1.5961077590196e-05, "loss": 1.0125, "step": 4775 }, { "epoch": 0.02, "learning_rate": 1.595934658043257e-05, "loss": 0.9939, "step": 4776 }, { "epoch": 0.02, "learning_rate": 1.5957615293719722e-05, "loss": 0.9594, "step": 4777 }, { "epoch": 0.02, "learning_rate": 1.5955883730137912e-05, "loss": 1.01, "step": 4778 }, { "epoch": 0.02, "learning_rate": 1.5954151889767616e-05, "loss": 0.9977, "step": 4779 }, { "epoch": 0.02, "learning_rate": 1.5952419772689308e-05, "loss": 0.9879, "step": 4780 }, { "epoch": 0.02, "learning_rate": 1.5950687378983497e-05, "loss": 0.9607, "step": 4781 }, { "epoch": 0.02, "learning_rate": 1.594895470873069e-05, "loss": 0.9791, "step": 4782 }, { "epoch": 0.02, "learning_rate": 1.5947221762011405e-05, "loss": 0.9619, "step": 4783 }, { "epoch": 0.02, "learning_rate": 1.594548853890618e-05, "loss": 0.989, "step": 4784 }, { "epoch": 0.02, "learning_rate": 1.5943755039495562e-05, "loss": 0.9796, "step": 4785 }, { "epoch": 0.02, "learning_rate": 1.5942021263860113e-05, "loss": 0.9862, "step": 4786 }, { "epoch": 0.02, "learning_rate": 1.594028721208041e-05, "loss": 0.9961, "step": 4787 }, { "epoch": 0.02, "learning_rate": 1.5938552884237032e-05, "loss": 0.9802, "step": 4788 }, { "epoch": 0.02, "learning_rate": 1.5936818280410585e-05, "loss": 0.9839, "step": 4789 }, { "epoch": 0.02, "learning_rate": 1.5935083400681678e-05, "loss": 0.9575, "step": 4790 }, { "epoch": 0.02, "learning_rate": 1.5933348245130945e-05, "loss": 0.9936, "step": 4791 }, { "epoch": 0.02, "learning_rate": 1.593161281383901e-05, "loss": 0.9651, "step": 4792 }, { "epoch": 0.02, "learning_rate": 1.5929877106886534e-05, "loss": 1.0003, "step": 4793 }, { "epoch": 0.02, "learning_rate": 1.592814112435418e-05, "loss": 0.9849, "step": 4794 }, { "epoch": 0.02, "learning_rate": 1.592640486632262e-05, "loss": 0.963, "step": 4795 }, { "epoch": 0.02, "learning_rate": 1.5924668332872547e-05, "loss": 0.9988, "step": 4796 }, { "epoch": 0.02, "learning_rate": 1.592293152408466e-05, "loss": 0.9379, "step": 4797 }, { "epoch": 0.02, "learning_rate": 1.5921194440039677e-05, "loss": 0.9774, "step": 4798 }, { "epoch": 0.02, "learning_rate": 1.5919457080818322e-05, "loss": 1.0167, "step": 4799 }, { "epoch": 0.02, "learning_rate": 1.5917719446501343e-05, "loss": 0.9957, "step": 4800 }, { "epoch": 0.02, "learning_rate": 1.5915981537169484e-05, "loss": 0.9955, "step": 4801 }, { "epoch": 0.02, "learning_rate": 1.591424335290352e-05, "loss": 0.9829, "step": 4802 }, { "epoch": 0.02, "learning_rate": 1.591250489378422e-05, "loss": 1.005, "step": 4803 }, { "epoch": 0.02, "learning_rate": 1.5910766159892384e-05, "loss": 0.9866, "step": 4804 }, { "epoch": 0.02, "learning_rate": 1.5909027151308813e-05, "loss": 0.9586, "step": 4805 }, { "epoch": 0.02, "learning_rate": 1.5907287868114324e-05, "loss": 0.9783, "step": 4806 }, { "epoch": 0.02, "learning_rate": 1.590554831038975e-05, "loss": 1.0127, "step": 4807 }, { "epoch": 0.02, "learning_rate": 1.5903808478215924e-05, "loss": 1.024, "step": 4808 }, { "epoch": 0.02, "learning_rate": 1.5902068371673714e-05, "loss": 0.9625, "step": 4809 }, { "epoch": 0.02, "learning_rate": 1.5900327990843978e-05, "loss": 1.0037, "step": 4810 }, { "epoch": 0.02, "learning_rate": 1.5898587335807603e-05, "loss": 0.9917, "step": 4811 }, { "epoch": 0.02, "learning_rate": 1.5896846406645484e-05, "loss": 0.969, "step": 4812 }, { "epoch": 0.02, "learning_rate": 1.589510520343852e-05, "loss": 0.9981, "step": 4813 }, { "epoch": 0.02, "learning_rate": 1.589336372626764e-05, "loss": 0.9463, "step": 4814 }, { "epoch": 0.02, "learning_rate": 1.589162197521377e-05, "loss": 0.9849, "step": 4815 }, { "epoch": 0.02, "learning_rate": 1.588987995035785e-05, "loss": 0.9836, "step": 4816 }, { "epoch": 0.02, "learning_rate": 1.5888137651780847e-05, "loss": 0.972, "step": 4817 }, { "epoch": 0.02, "learning_rate": 1.5886395079563723e-05, "loss": 1.0175, "step": 4818 }, { "epoch": 0.02, "learning_rate": 1.5884652233787466e-05, "loss": 1.0181, "step": 4819 }, { "epoch": 0.02, "learning_rate": 1.588290911453307e-05, "loss": 0.9467, "step": 4820 }, { "epoch": 0.02, "learning_rate": 1.5881165721881543e-05, "loss": 1.0157, "step": 4821 }, { "epoch": 0.02, "learning_rate": 1.5879422055913908e-05, "loss": 0.9289, "step": 4822 }, { "epoch": 0.02, "learning_rate": 1.587767811671119e-05, "loss": 0.9905, "step": 4823 }, { "epoch": 0.02, "learning_rate": 1.587593390435445e-05, "loss": 0.9516, "step": 4824 }, { "epoch": 0.02, "learning_rate": 1.5874189418924736e-05, "loss": 1.0164, "step": 4825 }, { "epoch": 0.02, "learning_rate": 1.587244466050312e-05, "loss": 0.9763, "step": 4826 }, { "epoch": 0.02, "learning_rate": 1.587069962917069e-05, "loss": 0.9698, "step": 4827 }, { "epoch": 0.02, "learning_rate": 1.5868954325008542e-05, "loss": 0.9923, "step": 4828 }, { "epoch": 0.02, "learning_rate": 1.5867208748097783e-05, "loss": 0.9567, "step": 4829 }, { "epoch": 0.02, "learning_rate": 1.5865462898519542e-05, "loss": 0.9406, "step": 4830 }, { "epoch": 0.02, "learning_rate": 1.586371677635495e-05, "loss": 0.9909, "step": 4831 }, { "epoch": 0.02, "learning_rate": 1.586197038168515e-05, "loss": 0.9621, "step": 4832 }, { "epoch": 0.02, "learning_rate": 1.586022371459131e-05, "loss": 0.9703, "step": 4833 }, { "epoch": 0.02, "learning_rate": 1.58584767751546e-05, "loss": 1.0202, "step": 4834 }, { "epoch": 0.02, "learning_rate": 1.5856729563456206e-05, "loss": 1.0006, "step": 4835 }, { "epoch": 0.02, "learning_rate": 1.5854982079577326e-05, "loss": 0.9788, "step": 4836 }, { "epoch": 0.02, "learning_rate": 1.5853234323599175e-05, "loss": 0.9863, "step": 4837 }, { "epoch": 0.02, "learning_rate": 1.5851486295602966e-05, "loss": 0.9522, "step": 4838 }, { "epoch": 0.02, "learning_rate": 1.5849737995669946e-05, "loss": 0.9775, "step": 4839 }, { "epoch": 0.02, "learning_rate": 1.5847989423881363e-05, "loss": 1.0053, "step": 4840 }, { "epoch": 0.02, "learning_rate": 1.584624058031847e-05, "loss": 1.0025, "step": 4841 }, { "epoch": 0.02, "learning_rate": 1.5844491465062555e-05, "loss": 0.9684, "step": 4842 }, { "epoch": 0.02, "learning_rate": 1.584274207819489e-05, "loss": 0.9818, "step": 4843 }, { "epoch": 0.02, "learning_rate": 1.5840992419796784e-05, "loss": 0.9671, "step": 4844 }, { "epoch": 0.02, "learning_rate": 1.5839242489949546e-05, "loss": 0.9879, "step": 4845 }, { "epoch": 0.02, "learning_rate": 1.58374922887345e-05, "loss": 1.0017, "step": 4846 }, { "epoch": 0.02, "learning_rate": 1.5835741816232986e-05, "loss": 0.9947, "step": 4847 }, { "epoch": 0.02, "learning_rate": 1.5833991072526352e-05, "loss": 0.9335, "step": 4848 }, { "epoch": 0.02, "learning_rate": 1.5832240057695964e-05, "loss": 0.9817, "step": 4849 }, { "epoch": 0.02, "learning_rate": 1.5830488771823192e-05, "loss": 1.0087, "step": 4850 }, { "epoch": 0.02, "learning_rate": 1.582873721498942e-05, "loss": 1.0184, "step": 4851 }, { "epoch": 0.02, "learning_rate": 1.5826985387276062e-05, "loss": 0.9646, "step": 4852 }, { "epoch": 0.02, "learning_rate": 1.5825233288764522e-05, "loss": 0.9692, "step": 4853 }, { "epoch": 0.02, "learning_rate": 1.5823480919536224e-05, "loss": 0.9553, "step": 4854 }, { "epoch": 0.02, "learning_rate": 1.5821728279672605e-05, "loss": 0.9993, "step": 4855 }, { "epoch": 0.02, "learning_rate": 1.5819975369255126e-05, "loss": 0.9908, "step": 4856 }, { "epoch": 0.02, "learning_rate": 1.581822218836524e-05, "loss": 0.9731, "step": 4857 }, { "epoch": 0.02, "learning_rate": 1.5816468737084428e-05, "loss": 0.972, "step": 4858 }, { "epoch": 0.02, "learning_rate": 1.5814715015494175e-05, "loss": 0.971, "step": 4859 }, { "epoch": 0.02, "learning_rate": 1.581296102367598e-05, "loss": 0.9964, "step": 4860 }, { "epoch": 0.02, "learning_rate": 1.5811206761711364e-05, "loss": 1.0062, "step": 4861 }, { "epoch": 0.02, "learning_rate": 1.5809452229681847e-05, "loss": 1.0095, "step": 4862 }, { "epoch": 0.02, "learning_rate": 1.580769742766897e-05, "loss": 0.9989, "step": 4863 }, { "epoch": 0.02, "learning_rate": 1.5805942355754287e-05, "loss": 0.9932, "step": 4864 }, { "epoch": 0.02, "learning_rate": 1.5804187014019352e-05, "loss": 0.9756, "step": 4865 }, { "epoch": 0.02, "learning_rate": 1.580243140254575e-05, "loss": 1.0114, "step": 4866 }, { "epoch": 0.02, "learning_rate": 1.580067552141507e-05, "loss": 1.0143, "step": 4867 }, { "epoch": 0.02, "learning_rate": 1.5798919370708907e-05, "loss": 0.9986, "step": 4868 }, { "epoch": 0.02, "learning_rate": 1.5797162950508882e-05, "loss": 0.983, "step": 4869 }, { "epoch": 0.02, "learning_rate": 1.579540626089661e-05, "loss": 0.9738, "step": 4870 }, { "epoch": 0.02, "learning_rate": 1.5793649301953745e-05, "loss": 0.9979, "step": 4871 }, { "epoch": 0.02, "learning_rate": 1.5791892073761928e-05, "loss": 0.9406, "step": 4872 }, { "epoch": 0.02, "learning_rate": 1.5790134576402827e-05, "loss": 0.9813, "step": 4873 }, { "epoch": 0.02, "learning_rate": 1.5788376809958116e-05, "loss": 0.9547, "step": 4874 }, { "epoch": 0.02, "learning_rate": 1.5786618774509483e-05, "loss": 0.9919, "step": 4875 }, { "epoch": 0.02, "learning_rate": 1.5784860470138633e-05, "loss": 0.9976, "step": 4876 }, { "epoch": 0.02, "learning_rate": 1.5783101896927278e-05, "loss": 0.9732, "step": 4877 }, { "epoch": 0.02, "learning_rate": 1.578134305495715e-05, "loss": 0.9901, "step": 4878 }, { "epoch": 0.02, "learning_rate": 1.5779583944309976e-05, "loss": 0.9935, "step": 4879 }, { "epoch": 0.02, "learning_rate": 1.5777824565067513e-05, "loss": 0.9908, "step": 4880 }, { "epoch": 0.02, "learning_rate": 1.577606491731153e-05, "loss": 0.9821, "step": 4881 }, { "epoch": 0.02, "learning_rate": 1.5774305001123795e-05, "loss": 0.9548, "step": 4882 }, { "epoch": 0.02, "learning_rate": 1.5772544816586103e-05, "loss": 0.9808, "step": 4883 }, { "epoch": 0.02, "learning_rate": 1.5770784363780252e-05, "loss": 0.9813, "step": 4884 }, { "epoch": 0.02, "learning_rate": 1.576902364278806e-05, "loss": 1.0031, "step": 4885 }, { "epoch": 0.02, "learning_rate": 1.5767262653691348e-05, "loss": 0.9968, "step": 4886 }, { "epoch": 0.02, "learning_rate": 1.5765501396571955e-05, "loss": 1.0268, "step": 4887 }, { "epoch": 0.03, "learning_rate": 1.576373987151173e-05, "loss": 0.9593, "step": 4888 }, { "epoch": 0.03, "learning_rate": 1.5761978078592547e-05, "loss": 0.9936, "step": 4889 }, { "epoch": 0.03, "learning_rate": 1.5760216017896274e-05, "loss": 0.9331, "step": 4890 }, { "epoch": 0.03, "learning_rate": 1.57584536895048e-05, "loss": 0.9898, "step": 4891 }, { "epoch": 0.03, "learning_rate": 1.5756691093500026e-05, "loss": 0.9707, "step": 4892 }, { "epoch": 0.03, "learning_rate": 1.575492822996387e-05, "loss": 0.9392, "step": 4893 }, { "epoch": 0.03, "learning_rate": 1.575316509897825e-05, "loss": 0.9597, "step": 4894 }, { "epoch": 0.03, "learning_rate": 1.5751401700625107e-05, "loss": 0.9961, "step": 4895 }, { "epoch": 0.03, "learning_rate": 1.574963803498639e-05, "loss": 1.0082, "step": 4896 }, { "epoch": 0.03, "learning_rate": 1.5747874102144073e-05, "loss": 0.9858, "step": 4897 }, { "epoch": 0.03, "learning_rate": 1.5746109902180114e-05, "loss": 0.9608, "step": 4898 }, { "epoch": 0.03, "learning_rate": 1.5744345435176517e-05, "loss": 0.9892, "step": 4899 }, { "epoch": 0.03, "learning_rate": 1.5742580701215273e-05, "loss": 0.9876, "step": 4900 }, { "epoch": 0.03, "learning_rate": 1.5740815700378392e-05, "loss": 1.0269, "step": 4901 }, { "epoch": 0.03, "learning_rate": 1.573905043274791e-05, "loss": 0.9873, "step": 4902 }, { "epoch": 0.03, "learning_rate": 1.5737284898405857e-05, "loss": 0.9779, "step": 4903 }, { "epoch": 0.03, "learning_rate": 1.5735519097434285e-05, "loss": 0.931, "step": 4904 }, { "epoch": 0.03, "learning_rate": 1.5733753029915252e-05, "loss": 0.9782, "step": 4905 }, { "epoch": 0.03, "learning_rate": 1.5731986695930844e-05, "loss": 0.9609, "step": 4906 }, { "epoch": 0.03, "learning_rate": 1.5730220095563137e-05, "loss": 1.0034, "step": 4907 }, { "epoch": 0.03, "learning_rate": 1.572845322889423e-05, "loss": 0.993, "step": 4908 }, { "epoch": 0.03, "learning_rate": 1.572668609600624e-05, "loss": 0.9792, "step": 4909 }, { "epoch": 0.03, "learning_rate": 1.572491869698129e-05, "loss": 1.0019, "step": 4910 }, { "epoch": 0.03, "learning_rate": 1.572315103190152e-05, "loss": 0.9802, "step": 4911 }, { "epoch": 0.03, "learning_rate": 1.572138310084907e-05, "loss": 0.9551, "step": 4912 }, { "epoch": 0.03, "learning_rate": 1.571961490390611e-05, "loss": 0.9942, "step": 4913 }, { "epoch": 0.03, "learning_rate": 1.5717846441154806e-05, "loss": 0.989, "step": 4914 }, { "epoch": 0.03, "learning_rate": 1.5716077712677354e-05, "loss": 1.0504, "step": 4915 }, { "epoch": 0.03, "learning_rate": 1.5714308718555943e-05, "loss": 0.9946, "step": 4916 }, { "epoch": 0.03, "learning_rate": 1.5712539458872783e-05, "loss": 1.0047, "step": 4917 }, { "epoch": 0.03, "learning_rate": 1.5710769933710107e-05, "loss": 0.9729, "step": 4918 }, { "epoch": 0.03, "learning_rate": 1.5709000143150145e-05, "loss": 0.9769, "step": 4919 }, { "epoch": 0.03, "learning_rate": 1.570723008727514e-05, "loss": 0.9816, "step": 4920 }, { "epoch": 0.03, "learning_rate": 1.570545976616736e-05, "loss": 1.015, "step": 4921 }, { "epoch": 0.03, "learning_rate": 1.5703689179909073e-05, "loss": 0.9499, "step": 4922 }, { "epoch": 0.03, "learning_rate": 1.570191832858256e-05, "loss": 0.9874, "step": 4923 }, { "epoch": 0.03, "learning_rate": 1.5700147212270127e-05, "loss": 0.9941, "step": 4924 }, { "epoch": 0.03, "learning_rate": 1.5698375831054078e-05, "loss": 0.9972, "step": 4925 }, { "epoch": 0.03, "learning_rate": 1.569660418501673e-05, "loss": 0.9701, "step": 4926 }, { "epoch": 0.03, "learning_rate": 1.5694832274240427e-05, "loss": 0.9743, "step": 4927 }, { "epoch": 0.03, "learning_rate": 1.569306009880751e-05, "loss": 0.9979, "step": 4928 }, { "epoch": 0.03, "learning_rate": 1.5691287658800337e-05, "loss": 0.9939, "step": 4929 }, { "epoch": 0.03, "learning_rate": 1.568951495430128e-05, "loss": 0.9704, "step": 4930 }, { "epoch": 0.03, "learning_rate": 1.568774198539272e-05, "loss": 0.9567, "step": 4931 }, { "epoch": 0.03, "learning_rate": 1.5685968752157052e-05, "loss": 0.9796, "step": 4932 }, { "epoch": 0.03, "learning_rate": 1.568419525467669e-05, "loss": 0.989, "step": 4933 }, { "epoch": 0.03, "learning_rate": 1.5682421493034045e-05, "loss": 0.9448, "step": 4934 }, { "epoch": 0.03, "learning_rate": 1.568064746731156e-05, "loss": 1.0118, "step": 4935 }, { "epoch": 0.03, "learning_rate": 1.5678873177591668e-05, "loss": 0.9903, "step": 4936 }, { "epoch": 0.03, "learning_rate": 1.5677098623956836e-05, "loss": 0.9926, "step": 4937 }, { "epoch": 0.03, "learning_rate": 1.5675323806489524e-05, "loss": 0.9711, "step": 4938 }, { "epoch": 0.03, "learning_rate": 1.567354872527222e-05, "loss": 0.9686, "step": 4939 }, { "epoch": 0.03, "learning_rate": 1.5671773380387413e-05, "loss": 0.936, "step": 4940 }, { "epoch": 0.03, "learning_rate": 1.5669997771917614e-05, "loss": 0.9557, "step": 4941 }, { "epoch": 0.03, "learning_rate": 1.5668221899945338e-05, "loss": 0.978, "step": 4942 }, { "epoch": 0.03, "learning_rate": 1.566644576455311e-05, "loss": 0.9848, "step": 4943 }, { "epoch": 0.03, "learning_rate": 1.5664669365823482e-05, "loss": 0.9865, "step": 4944 }, { "epoch": 0.03, "learning_rate": 1.5662892703839006e-05, "loss": 0.9847, "step": 4945 }, { "epoch": 0.03, "learning_rate": 1.5661115778682246e-05, "loss": 0.9394, "step": 4946 }, { "epoch": 0.03, "learning_rate": 1.565933859043578e-05, "loss": 1.0159, "step": 4947 }, { "epoch": 0.03, "learning_rate": 1.5657561139182205e-05, "loss": 0.9688, "step": 4948 }, { "epoch": 0.03, "learning_rate": 1.565578342500412e-05, "loss": 1.0097, "step": 4949 }, { "epoch": 0.03, "learning_rate": 1.5654005447984145e-05, "loss": 1.0045, "step": 4950 }, { "epoch": 0.03, "learning_rate": 1.5652227208204906e-05, "loss": 0.9683, "step": 4951 }, { "epoch": 0.03, "learning_rate": 1.5650448705749047e-05, "loss": 0.9775, "step": 4952 }, { "epoch": 0.03, "learning_rate": 1.564866994069921e-05, "loss": 0.9611, "step": 4953 }, { "epoch": 0.03, "learning_rate": 1.5646890913138072e-05, "loss": 1.0004, "step": 4954 }, { "epoch": 0.03, "learning_rate": 1.5645111623148303e-05, "loss": 1.0012, "step": 4955 }, { "epoch": 0.03, "learning_rate": 1.5643332070812592e-05, "loss": 1.0269, "step": 4956 }, { "epoch": 0.03, "learning_rate": 1.5641552256213644e-05, "loss": 0.9824, "step": 4957 }, { "epoch": 0.03, "learning_rate": 1.563977217943417e-05, "loss": 0.9677, "step": 4958 }, { "epoch": 0.03, "learning_rate": 1.56379918405569e-05, "loss": 0.986, "step": 4959 }, { "epoch": 0.03, "learning_rate": 1.5636211239664565e-05, "loss": 0.9184, "step": 4960 }, { "epoch": 0.03, "learning_rate": 1.563443037683992e-05, "loss": 0.9578, "step": 4961 }, { "epoch": 0.03, "learning_rate": 1.5632649252165727e-05, "loss": 1.024, "step": 4962 }, { "epoch": 0.03, "learning_rate": 1.5630867865724755e-05, "loss": 0.9866, "step": 4963 }, { "epoch": 0.03, "learning_rate": 1.5629086217599794e-05, "loss": 0.9659, "step": 4964 }, { "epoch": 0.03, "learning_rate": 1.5627304307873647e-05, "loss": 0.9813, "step": 4965 }, { "epoch": 0.03, "learning_rate": 1.562552213662912e-05, "loss": 0.9634, "step": 4966 }, { "epoch": 0.03, "learning_rate": 1.5623739703949036e-05, "loss": 0.9696, "step": 4967 }, { "epoch": 0.03, "learning_rate": 1.5621957009916238e-05, "loss": 0.9628, "step": 4968 }, { "epoch": 0.03, "learning_rate": 1.5620174054613563e-05, "loss": 1.0106, "step": 4969 }, { "epoch": 0.03, "learning_rate": 1.561839083812387e-05, "loss": 1.018, "step": 4970 }, { "epoch": 0.03, "learning_rate": 1.561660736053004e-05, "loss": 1.0359, "step": 4971 }, { "epoch": 0.03, "learning_rate": 1.5614823621914952e-05, "loss": 0.9488, "step": 4972 }, { "epoch": 0.03, "learning_rate": 1.56130396223615e-05, "loss": 0.9763, "step": 4973 }, { "epoch": 0.03, "learning_rate": 1.5611255361952595e-05, "loss": 0.9633, "step": 4974 }, { "epoch": 0.03, "learning_rate": 1.5609470840771156e-05, "loss": 0.9556, "step": 4975 }, { "epoch": 0.03, "learning_rate": 1.5607686058900115e-05, "loss": 0.9721, "step": 4976 }, { "epoch": 0.03, "learning_rate": 1.5605901016422416e-05, "loss": 0.9291, "step": 4977 }, { "epoch": 0.03, "learning_rate": 1.5604115713421014e-05, "loss": 1.032, "step": 4978 }, { "epoch": 0.03, "learning_rate": 1.5602330149978883e-05, "loss": 0.9989, "step": 4979 }, { "epoch": 0.03, "learning_rate": 1.5600544326179e-05, "loss": 0.9791, "step": 4980 }, { "epoch": 0.03, "learning_rate": 1.5598758242104353e-05, "loss": 0.9933, "step": 4981 }, { "epoch": 0.03, "learning_rate": 1.5596971897837957e-05, "loss": 0.9702, "step": 4982 }, { "epoch": 0.03, "learning_rate": 1.5595185293462827e-05, "loss": 0.9532, "step": 4983 }, { "epoch": 0.03, "learning_rate": 1.5593398429061985e-05, "loss": 1.0014, "step": 4984 }, { "epoch": 0.03, "learning_rate": 1.5591611304718478e-05, "loss": 0.9921, "step": 4985 }, { "epoch": 0.03, "learning_rate": 1.5589823920515353e-05, "loss": 1.0133, "step": 4986 }, { "epoch": 0.03, "learning_rate": 1.5588036276535684e-05, "loss": 0.9834, "step": 4987 }, { "epoch": 0.03, "learning_rate": 1.5586248372862543e-05, "loss": 0.9831, "step": 4988 }, { "epoch": 0.03, "learning_rate": 1.558446020957902e-05, "loss": 0.9854, "step": 4989 }, { "epoch": 0.03, "learning_rate": 1.558267178676822e-05, "loss": 0.9621, "step": 4990 }, { "epoch": 0.03, "learning_rate": 1.5580883104513253e-05, "loss": 0.9758, "step": 4991 }, { "epoch": 0.03, "learning_rate": 1.5579094162897244e-05, "loss": 0.9555, "step": 4992 }, { "epoch": 0.03, "learning_rate": 1.5577304962003333e-05, "loss": 0.9419, "step": 4993 }, { "epoch": 0.03, "learning_rate": 1.557551550191467e-05, "loss": 0.9529, "step": 4994 }, { "epoch": 0.03, "learning_rate": 1.5573725782714413e-05, "loss": 0.9807, "step": 4995 }, { "epoch": 0.03, "learning_rate": 1.557193580448574e-05, "loss": 1.0085, "step": 4996 }, { "epoch": 0.03, "learning_rate": 1.557014556731183e-05, "loss": 0.9262, "step": 4997 }, { "epoch": 0.03, "learning_rate": 1.5568355071275896e-05, "loss": 0.9679, "step": 4998 }, { "epoch": 0.03, "learning_rate": 1.5566564316461134e-05, "loss": 0.9599, "step": 4999 }, { "epoch": 0.03, "learning_rate": 1.556477330295077e-05, "loss": 0.9557, "step": 5000 }, { "epoch": 0.03, "learning_rate": 1.5562982030828036e-05, "loss": 0.9918, "step": 5001 }, { "epoch": 0.03, "learning_rate": 1.5561190500176183e-05, "loss": 0.95, "step": 5002 }, { "epoch": 0.03, "learning_rate": 1.5559398711078464e-05, "loss": 0.974, "step": 5003 }, { "epoch": 0.03, "learning_rate": 1.5557606663618152e-05, "loss": 0.9643, "step": 5004 }, { "epoch": 0.03, "learning_rate": 1.5555814357878528e-05, "loss": 0.9683, "step": 5005 }, { "epoch": 0.03, "learning_rate": 1.555402179394289e-05, "loss": 0.9772, "step": 5006 }, { "epoch": 0.03, "learning_rate": 1.5552228971894535e-05, "loss": 0.9751, "step": 5007 }, { "epoch": 0.03, "learning_rate": 1.555043589181679e-05, "loss": 0.9495, "step": 5008 }, { "epoch": 0.03, "learning_rate": 1.5548642553792977e-05, "loss": 0.9834, "step": 5009 }, { "epoch": 0.03, "learning_rate": 1.5546848957906446e-05, "loss": 0.9703, "step": 5010 }, { "epoch": 0.03, "learning_rate": 1.5545055104240548e-05, "loss": 0.9815, "step": 5011 }, { "epoch": 0.03, "learning_rate": 1.5543260992878646e-05, "loss": 0.9776, "step": 5012 }, { "epoch": 0.03, "learning_rate": 1.554146662390412e-05, "loss": 0.9443, "step": 5013 }, { "epoch": 0.03, "learning_rate": 1.5539671997400364e-05, "loss": 0.9397, "step": 5014 }, { "epoch": 0.03, "learning_rate": 1.553787711345077e-05, "loss": 0.9263, "step": 5015 }, { "epoch": 0.03, "learning_rate": 1.553608197213876e-05, "loss": 0.9716, "step": 5016 }, { "epoch": 0.03, "learning_rate": 1.5534286573547763e-05, "loss": 0.9662, "step": 5017 }, { "epoch": 0.03, "learning_rate": 1.5532490917761208e-05, "loss": 0.9573, "step": 5018 }, { "epoch": 0.03, "learning_rate": 1.5530695004862545e-05, "loss": 0.9518, "step": 5019 }, { "epoch": 0.03, "learning_rate": 1.552889883493524e-05, "loss": 0.9482, "step": 5020 }, { "epoch": 0.03, "learning_rate": 1.5527102408062772e-05, "loss": 0.9754, "step": 5021 }, { "epoch": 0.03, "learning_rate": 1.5525305724328613e-05, "loss": 0.9725, "step": 5022 }, { "epoch": 0.03, "learning_rate": 1.5523508783816273e-05, "loss": 1.0141, "step": 5023 }, { "epoch": 0.03, "learning_rate": 1.552171158660925e-05, "loss": 0.9806, "step": 5024 }, { "epoch": 0.03, "learning_rate": 1.5519914132791075e-05, "loss": 1.001, "step": 5025 }, { "epoch": 0.03, "learning_rate": 1.551811642244528e-05, "loss": 0.9718, "step": 5026 }, { "epoch": 0.03, "learning_rate": 1.5516318455655402e-05, "loss": 0.9804, "step": 5027 }, { "epoch": 0.03, "learning_rate": 1.5514520232505004e-05, "loss": 0.9484, "step": 5028 }, { "epoch": 0.03, "learning_rate": 1.5512721753077663e-05, "loss": 0.9694, "step": 5029 }, { "epoch": 0.03, "learning_rate": 1.5510923017456946e-05, "loss": 0.9827, "step": 5030 }, { "epoch": 0.03, "learning_rate": 1.5509124025726453e-05, "loss": 0.9853, "step": 5031 }, { "epoch": 0.03, "learning_rate": 1.5507324777969786e-05, "loss": 0.9825, "step": 5032 }, { "epoch": 0.03, "learning_rate": 1.5505525274270567e-05, "loss": 0.9603, "step": 5033 }, { "epoch": 0.03, "learning_rate": 1.5503725514712415e-05, "loss": 0.9514, "step": 5034 }, { "epoch": 0.03, "learning_rate": 1.550192549937898e-05, "loss": 0.9466, "step": 5035 }, { "epoch": 0.03, "learning_rate": 1.5500125228353912e-05, "loss": 0.9561, "step": 5036 }, { "epoch": 0.03, "learning_rate": 1.5498324701720873e-05, "loss": 0.9735, "step": 5037 }, { "epoch": 0.04, "learning_rate": 1.5496523919563538e-05, "loss": 0.9823, "step": 5038 }, { "epoch": 0.04, "learning_rate": 1.54947228819656e-05, "loss": 1.0032, "step": 5039 }, { "epoch": 0.04, "learning_rate": 1.549292158901075e-05, "loss": 0.9587, "step": 5040 }, { "epoch": 0.04, "learning_rate": 1.5491120040782708e-05, "loss": 0.982, "step": 5041 }, { "epoch": 0.04, "learning_rate": 1.5489318237365195e-05, "loss": 0.9792, "step": 5042 }, { "epoch": 0.04, "learning_rate": 1.5487516178841948e-05, "loss": 0.9973, "step": 5043 }, { "epoch": 0.04, "learning_rate": 1.548571386529671e-05, "loss": 0.9429, "step": 5044 }, { "epoch": 0.04, "learning_rate": 1.5483911296813246e-05, "loss": 1.003, "step": 5045 }, { "epoch": 0.04, "learning_rate": 1.5482108473475324e-05, "loss": 0.9767, "step": 5046 }, { "epoch": 0.04, "learning_rate": 1.548030539536672e-05, "loss": 0.9871, "step": 5047 }, { "epoch": 0.04, "learning_rate": 1.5478502062571242e-05, "loss": 0.9341, "step": 5048 }, { "epoch": 0.04, "learning_rate": 1.5476698475172688e-05, "loss": 1.0341, "step": 5049 }, { "epoch": 0.04, "learning_rate": 1.5474894633254876e-05, "loss": 0.9793, "step": 5050 }, { "epoch": 0.04, "learning_rate": 1.547309053690164e-05, "loss": 0.9792, "step": 5051 }, { "epoch": 0.04, "learning_rate": 1.5471286186196823e-05, "loss": 0.9642, "step": 5052 }, { "epoch": 0.04, "learning_rate": 1.5469481581224274e-05, "loss": 0.9842, "step": 5053 }, { "epoch": 0.04, "learning_rate": 1.5467676722067858e-05, "loss": 0.9169, "step": 5054 }, { "epoch": 0.04, "learning_rate": 1.5465871608811454e-05, "loss": 1.0032, "step": 5055 }, { "epoch": 0.04, "learning_rate": 1.5464066241538957e-05, "loss": 0.9556, "step": 5056 }, { "epoch": 0.04, "learning_rate": 1.546226062033426e-05, "loss": 1.004, "step": 5057 }, { "epoch": 0.04, "learning_rate": 1.5460454745281278e-05, "loss": 0.9557, "step": 5058 }, { "epoch": 0.04, "learning_rate": 1.5458648616463938e-05, "loss": 1.0194, "step": 5059 }, { "epoch": 0.04, "learning_rate": 1.545684223396618e-05, "loss": 0.9588, "step": 5060 }, { "epoch": 0.04, "learning_rate": 1.545503559787194e-05, "loss": 1.0202, "step": 5061 }, { "epoch": 0.04, "learning_rate": 1.5453228708265193e-05, "loss": 0.9438, "step": 5062 }, { "epoch": 0.04, "learning_rate": 1.5451421565229897e-05, "loss": 1.0105, "step": 5063 }, { "epoch": 0.04, "learning_rate": 1.544961416885005e-05, "loss": 0.9636, "step": 5064 }, { "epoch": 0.04, "learning_rate": 1.544780651920963e-05, "loss": 0.9726, "step": 5065 }, { "epoch": 0.04, "learning_rate": 1.5445998616392657e-05, "loss": 0.9658, "step": 5066 }, { "epoch": 0.04, "learning_rate": 1.5444190460483146e-05, "loss": 0.9952, "step": 5067 }, { "epoch": 0.04, "learning_rate": 1.5442382051565127e-05, "loss": 0.9864, "step": 5068 }, { "epoch": 0.04, "learning_rate": 1.5440573389722646e-05, "loss": 1.0176, "step": 5069 }, { "epoch": 0.04, "learning_rate": 1.543876447503975e-05, "loss": 0.9999, "step": 5070 }, { "epoch": 0.04, "learning_rate": 1.543695530760051e-05, "loss": 0.944, "step": 5071 }, { "epoch": 0.04, "learning_rate": 1.5435145887489006e-05, "loss": 0.9426, "step": 5072 }, { "epoch": 0.04, "learning_rate": 1.5433336214789326e-05, "loss": 0.9691, "step": 5073 }, { "epoch": 0.04, "learning_rate": 1.5431526289585566e-05, "loss": 0.9192, "step": 5074 }, { "epoch": 0.04, "learning_rate": 1.5429716111961844e-05, "loss": 0.9394, "step": 5075 }, { "epoch": 0.04, "learning_rate": 1.542790568200228e-05, "loss": 0.9643, "step": 5076 }, { "epoch": 0.04, "learning_rate": 1.5426094999791014e-05, "loss": 0.9809, "step": 5077 }, { "epoch": 0.04, "learning_rate": 1.5424284065412198e-05, "loss": 1.0149, "step": 5078 }, { "epoch": 0.04, "learning_rate": 1.5422472878949982e-05, "loss": 1.015, "step": 5079 }, { "epoch": 0.04, "learning_rate": 1.5420661440488544e-05, "loss": 0.9539, "step": 5080 }, { "epoch": 0.04, "learning_rate": 1.5418849750112066e-05, "loss": 0.9836, "step": 5081 }, { "epoch": 0.04, "learning_rate": 1.541703780790474e-05, "loss": 0.9409, "step": 5082 }, { "epoch": 0.04, "learning_rate": 1.541522561395078e-05, "loss": 0.9667, "step": 5083 }, { "epoch": 0.04, "learning_rate": 1.54134131683344e-05, "loss": 0.9232, "step": 5084 }, { "epoch": 0.04, "learning_rate": 1.5411600471139827e-05, "loss": 0.9822, "step": 5085 }, { "epoch": 0.04, "learning_rate": 1.5409787522451308e-05, "loss": 1.0135, "step": 5086 }, { "epoch": 0.04, "learning_rate": 1.540797432235309e-05, "loss": 1.0137, "step": 5087 }, { "epoch": 0.04, "learning_rate": 1.540616087092944e-05, "loss": 0.9481, "step": 5088 }, { "epoch": 0.04, "learning_rate": 1.540434716826464e-05, "loss": 0.9779, "step": 5089 }, { "epoch": 0.04, "learning_rate": 1.5402533214442976e-05, "loss": 0.9231, "step": 5090 }, { "epoch": 0.04, "learning_rate": 1.5400719009548747e-05, "loss": 0.9626, "step": 5091 }, { "epoch": 0.04, "learning_rate": 1.539890455366626e-05, "loss": 0.9579, "step": 5092 }, { "epoch": 0.04, "learning_rate": 1.5397089846879846e-05, "loss": 0.9798, "step": 5093 }, { "epoch": 0.04, "learning_rate": 1.5395274889273837e-05, "loss": 0.9811, "step": 5094 }, { "epoch": 0.04, "learning_rate": 1.539345968093258e-05, "loss": 0.9821, "step": 5095 }, { "epoch": 0.04, "learning_rate": 1.5391644221940434e-05, "loss": 0.9623, "step": 5096 }, { "epoch": 0.04, "learning_rate": 1.538982851238177e-05, "loss": 0.9709, "step": 5097 }, { "epoch": 0.04, "learning_rate": 1.5388012552340964e-05, "loss": 0.9601, "step": 5098 }, { "epoch": 0.04, "learning_rate": 1.5386196341902417e-05, "loss": 0.9573, "step": 5099 }, { "epoch": 0.04, "learning_rate": 1.5384379881150527e-05, "loss": 0.9354, "step": 5100 }, { "epoch": 0.04, "learning_rate": 1.5382563170169718e-05, "loss": 0.9967, "step": 5101 }, { "epoch": 0.04, "learning_rate": 1.5380746209044414e-05, "loss": 0.9434, "step": 5102 }, { "epoch": 0.04, "learning_rate": 1.537892899785905e-05, "loss": 1.024, "step": 5103 }, { "epoch": 0.04, "learning_rate": 1.5377111536698086e-05, "loss": 0.949, "step": 5104 }, { "epoch": 0.04, "learning_rate": 1.537529382564598e-05, "loss": 0.9756, "step": 5105 }, { "epoch": 0.04, "learning_rate": 1.537347586478721e-05, "loss": 0.9646, "step": 5106 }, { "epoch": 0.04, "learning_rate": 1.5371657654206262e-05, "loss": 0.9822, "step": 5107 }, { "epoch": 0.04, "learning_rate": 1.536983919398763e-05, "loss": 0.9389, "step": 5108 }, { "epoch": 0.04, "learning_rate": 1.5368020484215828e-05, "loss": 0.9986, "step": 5109 }, { "epoch": 0.04, "learning_rate": 1.5366201524975372e-05, "loss": 0.9689, "step": 5110 }, { "epoch": 0.04, "learning_rate": 1.5364382316350803e-05, "loss": 0.9714, "step": 5111 }, { "epoch": 0.04, "learning_rate": 1.5362562858426655e-05, "loss": 0.9642, "step": 5112 }, { "epoch": 0.04, "learning_rate": 1.5360743151287492e-05, "loss": 0.9678, "step": 5113 }, { "epoch": 0.04, "learning_rate": 1.5358923195017877e-05, "loss": 0.9639, "step": 5114 }, { "epoch": 0.04, "learning_rate": 1.5357102989702387e-05, "loss": 0.9499, "step": 5115 }, { "epoch": 0.04, "learning_rate": 1.5355282535425626e-05, "loss": 0.9865, "step": 5116 }, { "epoch": 0.04, "learning_rate": 1.5353461832272176e-05, "loss": 0.9771, "step": 5117 }, { "epoch": 0.04, "learning_rate": 1.5351640880326668e-05, "loss": 0.9646, "step": 5118 }, { "epoch": 0.04, "learning_rate": 1.534981967967372e-05, "loss": 0.9699, "step": 5119 }, { "epoch": 0.04, "learning_rate": 1.5347998230397963e-05, "loss": 0.9252, "step": 5120 }, { "epoch": 0.04, "learning_rate": 1.5346176532584058e-05, "loss": 0.9577, "step": 5121 }, { "epoch": 0.04, "learning_rate": 1.534435458631666e-05, "loss": 0.9517, "step": 5122 }, { "epoch": 0.04, "learning_rate": 1.5342532391680434e-05, "loss": 0.9846, "step": 5123 }, { "epoch": 0.04, "learning_rate": 1.534070994876007e-05, "loss": 0.9465, "step": 5124 }, { "epoch": 0.04, "learning_rate": 1.533888725764026e-05, "loss": 1.0135, "step": 5125 }, { "epoch": 0.04, "learning_rate": 1.5337064318405713e-05, "loss": 0.9619, "step": 5126 }, { "epoch": 0.04, "learning_rate": 1.5335241131141145e-05, "loss": 0.9554, "step": 5127 }, { "epoch": 0.04, "learning_rate": 1.5333417695931283e-05, "loss": 0.9351, "step": 5128 }, { "epoch": 0.04, "learning_rate": 1.533159401286087e-05, "loss": 0.9706, "step": 5129 }, { "epoch": 0.04, "learning_rate": 1.532977008201466e-05, "loss": 0.9203, "step": 5130 }, { "epoch": 0.04, "learning_rate": 1.5327945903477416e-05, "loss": 0.9499, "step": 5131 }, { "epoch": 0.04, "learning_rate": 1.5326121477333908e-05, "loss": 0.9632, "step": 5132 }, { "epoch": 0.04, "learning_rate": 1.5324296803668927e-05, "loss": 0.962, "step": 5133 }, { "epoch": 0.04, "learning_rate": 1.5322471882567272e-05, "loss": 0.9459, "step": 5134 }, { "epoch": 0.04, "learning_rate": 1.5320646714113748e-05, "loss": 1.016, "step": 5135 }, { "epoch": 0.04, "learning_rate": 1.5318821298393184e-05, "loss": 0.9597, "step": 5136 }, { "epoch": 0.04, "learning_rate": 1.531699563549041e-05, "loss": 0.9738, "step": 5137 }, { "epoch": 0.04, "learning_rate": 1.5315169725490266e-05, "loss": 0.9358, "step": 5138 }, { "epoch": 0.04, "learning_rate": 1.5313343568477613e-05, "loss": 1.0117, "step": 5139 }, { "epoch": 0.04, "learning_rate": 1.5311517164537316e-05, "loss": 0.937, "step": 5140 }, { "epoch": 0.04, "learning_rate": 1.530969051375425e-05, "loss": 1.0135, "step": 5141 }, { "epoch": 0.04, "learning_rate": 1.5307863616213313e-05, "loss": 0.9284, "step": 5142 }, { "epoch": 0.04, "learning_rate": 1.5306036471999397e-05, "loss": 0.9704, "step": 5143 }, { "epoch": 0.04, "learning_rate": 1.5304209081197425e-05, "loss": 0.9886, "step": 5144 }, { "epoch": 0.04, "learning_rate": 1.5302381443892318e-05, "loss": 0.963, "step": 5145 }, { "epoch": 0.04, "learning_rate": 1.530055356016901e-05, "loss": 0.9387, "step": 5146 }, { "epoch": 0.04, "learning_rate": 1.529872543011245e-05, "loss": 0.9688, "step": 5147 }, { "epoch": 0.04, "learning_rate": 1.5296897053807593e-05, "loss": 0.9569, "step": 5148 }, { "epoch": 0.04, "learning_rate": 1.5295068431339414e-05, "loss": 1.0194, "step": 5149 }, { "epoch": 0.04, "learning_rate": 1.5293239562792893e-05, "loss": 0.9452, "step": 5150 }, { "epoch": 0.04, "learning_rate": 1.5291410448253025e-05, "loss": 0.9955, "step": 5151 }, { "epoch": 0.04, "learning_rate": 1.5289581087804816e-05, "loss": 0.9739, "step": 5152 }, { "epoch": 0.04, "learning_rate": 1.5287751481533276e-05, "loss": 0.983, "step": 5153 }, { "epoch": 0.04, "learning_rate": 1.5285921629523438e-05, "loss": 0.9694, "step": 5154 }, { "epoch": 0.04, "learning_rate": 1.5284091531860336e-05, "loss": 0.9838, "step": 5155 }, { "epoch": 0.04, "learning_rate": 1.5282261188629027e-05, "loss": 0.9565, "step": 5156 }, { "epoch": 0.04, "learning_rate": 1.5280430599914567e-05, "loss": 0.9885, "step": 5157 }, { "epoch": 0.04, "learning_rate": 1.527859976580203e-05, "loss": 0.9456, "step": 5158 }, { "epoch": 0.04, "learning_rate": 1.5276768686376505e-05, "loss": 0.9726, "step": 5159 }, { "epoch": 0.04, "learning_rate": 1.5274937361723082e-05, "loss": 0.9612, "step": 5160 }, { "epoch": 0.04, "learning_rate": 1.5273105791926872e-05, "loss": 0.9656, "step": 5161 }, { "epoch": 0.04, "learning_rate": 1.5271273977072993e-05, "loss": 0.9503, "step": 5162 }, { "epoch": 0.04, "learning_rate": 1.5269441917246573e-05, "loss": 1.002, "step": 5163 }, { "epoch": 0.04, "learning_rate": 1.526760961253276e-05, "loss": 0.9515, "step": 5164 }, { "epoch": 0.04, "learning_rate": 1.5265777063016703e-05, "loss": 0.933, "step": 5165 }, { "epoch": 0.04, "learning_rate": 1.5263944268783563e-05, "loss": 0.921, "step": 5166 }, { "epoch": 0.04, "learning_rate": 1.5262111229918516e-05, "loss": 0.9799, "step": 5167 }, { "epoch": 0.04, "learning_rate": 1.5260277946506756e-05, "loss": 0.9279, "step": 5168 }, { "epoch": 0.04, "learning_rate": 1.5258444418633474e-05, "loss": 0.9613, "step": 5169 }, { "epoch": 0.04, "learning_rate": 1.5256610646383884e-05, "loss": 0.963, "step": 5170 }, { "epoch": 0.04, "learning_rate": 1.5254776629843204e-05, "loss": 0.9951, "step": 5171 }, { "epoch": 0.04, "learning_rate": 1.5252942369096668e-05, "loss": 0.9625, "step": 5172 }, { "epoch": 0.04, "learning_rate": 1.5251107864229521e-05, "loss": 0.9304, "step": 5173 }, { "epoch": 0.04, "learning_rate": 1.524927311532702e-05, "loss": 0.9963, "step": 5174 }, { "epoch": 0.04, "learning_rate": 1.5247438122474425e-05, "loss": 0.9635, "step": 5175 }, { "epoch": 0.04, "learning_rate": 1.5245602885757017e-05, "loss": 0.9472, "step": 5176 }, { "epoch": 0.04, "learning_rate": 1.5243767405260089e-05, "loss": 0.9826, "step": 5177 }, { "epoch": 0.04, "learning_rate": 1.5241931681068936e-05, "loss": 0.9302, "step": 5178 }, { "epoch": 0.04, "learning_rate": 1.5240095713268872e-05, "loss": 0.9742, "step": 5179 }, { "epoch": 0.04, "learning_rate": 1.523825950194522e-05, "loss": 0.9659, "step": 5180 }, { "epoch": 0.04, "learning_rate": 1.5236423047183315e-05, "loss": 0.9754, "step": 5181 }, { "epoch": 0.04, "learning_rate": 1.5234586349068505e-05, "loss": 0.957, "step": 5182 }, { "epoch": 0.04, "learning_rate": 1.5232749407686142e-05, "loss": 1.0021, "step": 5183 }, { "epoch": 0.04, "learning_rate": 1.5230912223121597e-05, "loss": 0.9734, "step": 5184 }, { "epoch": 0.04, "learning_rate": 1.522907479546025e-05, "loss": 0.9752, "step": 5185 }, { "epoch": 0.04, "learning_rate": 1.522723712478749e-05, "loss": 0.9477, "step": 5186 }, { "epoch": 0.04, "learning_rate": 1.5225399211188723e-05, "loss": 0.9361, "step": 5187 }, { "epoch": 0.04, "learning_rate": 1.5223561054749361e-05, "loss": 0.9707, "step": 5188 }, { "epoch": 0.05, "learning_rate": 1.5221722655554824e-05, "loss": 0.9362, "step": 5189 }, { "epoch": 0.05, "learning_rate": 1.5219884013690557e-05, "loss": 0.9994, "step": 5190 }, { "epoch": 0.05, "learning_rate": 1.5218045129242004e-05, "loss": 1.0088, "step": 5191 }, { "epoch": 0.05, "learning_rate": 1.5216206002294617e-05, "loss": 0.9455, "step": 5192 }, { "epoch": 0.05, "learning_rate": 1.5214366632933877e-05, "loss": 0.9883, "step": 5193 }, { "epoch": 0.05, "learning_rate": 1.5212527021245256e-05, "loss": 0.9693, "step": 5194 }, { "epoch": 0.05, "learning_rate": 1.5210687167314251e-05, "loss": 0.9893, "step": 5195 }, { "epoch": 0.05, "learning_rate": 1.5208847071226366e-05, "loss": 0.961, "step": 5196 }, { "epoch": 0.05, "learning_rate": 1.5207006733067114e-05, "loss": 0.9809, "step": 5197 }, { "epoch": 0.05, "learning_rate": 1.5205166152922024e-05, "loss": 0.9519, "step": 5198 }, { "epoch": 0.05, "learning_rate": 1.520332533087663e-05, "loss": 1.006, "step": 5199 }, { "epoch": 0.05, "learning_rate": 1.5201484267016482e-05, "loss": 0.9464, "step": 5200 }, { "epoch": 0.05, "learning_rate": 1.5199642961427141e-05, "loss": 1.0111, "step": 5201 }, { "epoch": 0.05, "learning_rate": 1.519780141419418e-05, "loss": 0.9539, "step": 5202 }, { "epoch": 0.05, "learning_rate": 1.5195959625403176e-05, "loss": 0.9477, "step": 5203 }, { "epoch": 0.05, "learning_rate": 1.5194117595139725e-05, "loss": 0.9987, "step": 5204 }, { "epoch": 0.05, "learning_rate": 1.5192275323489437e-05, "loss": 0.9594, "step": 5205 }, { "epoch": 0.05, "learning_rate": 1.5190432810537923e-05, "loss": 1.0232, "step": 5206 }, { "epoch": 0.05, "learning_rate": 1.5188590056370808e-05, "loss": 0.9554, "step": 5207 }, { "epoch": 0.05, "learning_rate": 1.5186747061073737e-05, "loss": 0.981, "step": 5208 }, { "epoch": 0.05, "learning_rate": 1.5184903824732353e-05, "loss": 0.9769, "step": 5209 }, { "epoch": 0.05, "learning_rate": 1.5183060347432323e-05, "loss": 0.9831, "step": 5210 }, { "epoch": 0.05, "learning_rate": 1.5181216629259315e-05, "loss": 0.9635, "step": 5211 }, { "epoch": 0.05, "learning_rate": 1.5179372670299014e-05, "loss": 0.9883, "step": 5212 }, { "epoch": 0.05, "learning_rate": 1.5177528470637115e-05, "loss": 0.9505, "step": 5213 }, { "epoch": 0.05, "learning_rate": 1.5175684030359325e-05, "loss": 0.99, "step": 5214 }, { "epoch": 0.05, "learning_rate": 1.5173839349551356e-05, "loss": 0.9465, "step": 5215 }, { "epoch": 0.05, "learning_rate": 1.517199442829894e-05, "loss": 0.9599, "step": 5216 }, { "epoch": 0.05, "learning_rate": 1.5170149266687815e-05, "loss": 0.9998, "step": 5217 }, { "epoch": 0.05, "learning_rate": 1.5168303864803729e-05, "loss": 0.9681, "step": 5218 }, { "epoch": 0.05, "learning_rate": 1.5166458222732451e-05, "loss": 0.9528, "step": 5219 }, { "epoch": 0.05, "learning_rate": 1.5164612340559746e-05, "loss": 0.9433, "step": 5220 }, { "epoch": 0.05, "learning_rate": 1.5162766218371401e-05, "loss": 0.9626, "step": 5221 }, { "epoch": 0.05, "learning_rate": 1.5160919856253214e-05, "loss": 0.9412, "step": 5222 }, { "epoch": 0.05, "learning_rate": 1.5159073254290983e-05, "loss": 0.966, "step": 5223 }, { "epoch": 0.05, "learning_rate": 1.5157226412570533e-05, "loss": 0.9419, "step": 5224 }, { "epoch": 0.05, "learning_rate": 1.5155379331177692e-05, "loss": 0.9563, "step": 5225 }, { "epoch": 0.05, "learning_rate": 1.5153532010198296e-05, "loss": 0.9586, "step": 5226 }, { "epoch": 0.05, "learning_rate": 1.5151684449718197e-05, "loss": 0.9845, "step": 5227 }, { "epoch": 0.05, "learning_rate": 1.5149836649823257e-05, "loss": 0.9156, "step": 5228 }, { "epoch": 0.05, "learning_rate": 1.5147988610599352e-05, "loss": 0.952, "step": 5229 }, { "epoch": 0.05, "learning_rate": 1.5146140332132359e-05, "loss": 0.985, "step": 5230 }, { "epoch": 0.05, "learning_rate": 1.5144291814508184e-05, "loss": 0.9761, "step": 5231 }, { "epoch": 0.05, "learning_rate": 1.514244305781272e-05, "loss": 0.9465, "step": 5232 }, { "epoch": 0.05, "learning_rate": 1.5140594062131895e-05, "loss": 1.0034, "step": 5233 }, { "epoch": 0.05, "learning_rate": 1.5138744827551635e-05, "loss": 0.9648, "step": 5234 }, { "epoch": 0.05, "learning_rate": 1.5136895354157874e-05, "loss": 0.946, "step": 5235 }, { "epoch": 0.05, "learning_rate": 1.5135045642036574e-05, "loss": 0.9237, "step": 5236 }, { "epoch": 0.05, "learning_rate": 1.5133195691273687e-05, "loss": 0.9646, "step": 5237 }, { "epoch": 0.05, "learning_rate": 1.513134550195519e-05, "loss": 0.9329, "step": 5238 }, { "epoch": 0.05, "learning_rate": 1.5129495074167067e-05, "loss": 0.9304, "step": 5239 }, { "epoch": 0.05, "learning_rate": 1.5127644407995314e-05, "loss": 0.98, "step": 5240 }, { "epoch": 0.05, "learning_rate": 1.5125793503525935e-05, "loss": 0.9653, "step": 5241 }, { "epoch": 0.05, "learning_rate": 1.5123942360844947e-05, "loss": 0.9699, "step": 5242 }, { "epoch": 0.05, "learning_rate": 1.512209098003838e-05, "loss": 0.9418, "step": 5243 }, { "epoch": 0.05, "learning_rate": 1.5120239361192275e-05, "loss": 0.9703, "step": 5244 }, { "epoch": 0.05, "learning_rate": 1.5118387504392682e-05, "loss": 0.9791, "step": 5245 }, { "epoch": 0.05, "learning_rate": 1.5116535409725657e-05, "loss": 0.9687, "step": 5246 }, { "epoch": 0.05, "learning_rate": 1.5114683077277277e-05, "loss": 0.9798, "step": 5247 }, { "epoch": 0.05, "learning_rate": 1.5112830507133627e-05, "loss": 0.9768, "step": 5248 }, { "epoch": 0.05, "learning_rate": 1.51109776993808e-05, "loss": 0.9988, "step": 5249 }, { "epoch": 0.05, "learning_rate": 1.5109124654104901e-05, "loss": 0.9681, "step": 5250 }, { "epoch": 0.05, "learning_rate": 1.5107271371392049e-05, "loss": 0.9553, "step": 5251 }, { "epoch": 0.05, "learning_rate": 1.5105417851328371e-05, "loss": 0.9903, "step": 5252 }, { "epoch": 0.05, "learning_rate": 1.5103564094000005e-05, "loss": 0.9635, "step": 5253 }, { "epoch": 0.05, "learning_rate": 1.51017100994931e-05, "loss": 0.9498, "step": 5254 }, { "epoch": 0.05, "learning_rate": 1.509985586789382e-05, "loss": 0.9861, "step": 5255 }, { "epoch": 0.05, "learning_rate": 1.509800139928833e-05, "loss": 0.9552, "step": 5256 }, { "epoch": 0.05, "learning_rate": 1.5096146693762825e-05, "loss": 0.9755, "step": 5257 }, { "epoch": 0.05, "learning_rate": 1.5094291751403487e-05, "loss": 0.9729, "step": 5258 }, { "epoch": 0.05, "learning_rate": 1.509243657229653e-05, "loss": 0.9848, "step": 5259 }, { "epoch": 0.05, "learning_rate": 1.5090581156528164e-05, "loss": 0.949, "step": 5260 }, { "epoch": 0.05, "learning_rate": 1.5088725504184619e-05, "loss": 0.9909, "step": 5261 }, { "epoch": 0.05, "learning_rate": 1.508686961535213e-05, "loss": 0.9405, "step": 5262 }, { "epoch": 0.05, "learning_rate": 1.5085013490116947e-05, "loss": 0.9678, "step": 5263 }, { "epoch": 0.05, "learning_rate": 1.5083157128565331e-05, "loss": 0.9831, "step": 5264 }, { "epoch": 0.05, "learning_rate": 1.5081300530783553e-05, "loss": 0.9731, "step": 5265 }, { "epoch": 0.05, "learning_rate": 1.5079443696857895e-05, "loss": 0.9695, "step": 5266 }, { "epoch": 0.05, "learning_rate": 1.507758662687465e-05, "loss": 1.0149, "step": 5267 }, { "epoch": 0.05, "learning_rate": 1.5075729320920117e-05, "loss": 0.9552, "step": 5268 }, { "epoch": 0.05, "learning_rate": 1.5073871779080622e-05, "loss": 0.9886, "step": 5269 }, { "epoch": 0.05, "learning_rate": 1.5072014001442479e-05, "loss": 0.925, "step": 5270 }, { "epoch": 0.05, "learning_rate": 1.5070155988092033e-05, "loss": 0.9426, "step": 5271 }, { "epoch": 0.05, "learning_rate": 1.5068297739115623e-05, "loss": 0.9646, "step": 5272 }, { "epoch": 0.05, "learning_rate": 1.5066439254599617e-05, "loss": 0.9695, "step": 5273 }, { "epoch": 0.05, "learning_rate": 1.5064580534630378e-05, "loss": 0.9908, "step": 5274 }, { "epoch": 0.05, "learning_rate": 1.506272157929429e-05, "loss": 0.9219, "step": 5275 }, { "epoch": 0.05, "learning_rate": 1.5060862388677744e-05, "loss": 0.9526, "step": 5276 }, { "epoch": 0.05, "learning_rate": 1.505900296286714e-05, "loss": 0.9476, "step": 5277 }, { "epoch": 0.05, "learning_rate": 1.5057143301948892e-05, "loss": 0.9512, "step": 5278 }, { "epoch": 0.05, "learning_rate": 1.5055283406009427e-05, "loss": 0.9576, "step": 5279 }, { "epoch": 0.05, "learning_rate": 1.505342327513518e-05, "loss": 0.9743, "step": 5280 }, { "epoch": 0.05, "learning_rate": 1.5051562909412591e-05, "loss": 0.95, "step": 5281 }, { "epoch": 0.05, "learning_rate": 1.5049702308928125e-05, "loss": 0.9353, "step": 5282 }, { "epoch": 0.05, "learning_rate": 1.5047841473768245e-05, "loss": 0.989, "step": 5283 }, { "epoch": 0.05, "learning_rate": 1.5045980404019432e-05, "loss": 0.9588, "step": 5284 }, { "epoch": 0.05, "learning_rate": 1.5044119099768173e-05, "loss": 0.9563, "step": 5285 }, { "epoch": 0.05, "learning_rate": 1.5042257561100973e-05, "loss": 0.9711, "step": 5286 }, { "epoch": 0.05, "learning_rate": 1.504039578810434e-05, "loss": 0.9234, "step": 5287 }, { "epoch": 0.05, "learning_rate": 1.5038533780864799e-05, "loss": 0.959, "step": 5288 }, { "epoch": 0.05, "learning_rate": 1.5036671539468879e-05, "loss": 0.9728, "step": 5289 }, { "epoch": 0.05, "learning_rate": 1.503480906400313e-05, "loss": 0.9347, "step": 5290 }, { "epoch": 0.05, "learning_rate": 1.50329463545541e-05, "loss": 0.9611, "step": 5291 }, { "epoch": 0.05, "learning_rate": 1.503108341120836e-05, "loss": 0.9579, "step": 5292 }, { "epoch": 0.05, "learning_rate": 1.5029220234052486e-05, "loss": 0.9471, "step": 5293 }, { "epoch": 0.05, "learning_rate": 1.5027356823173067e-05, "loss": 0.9224, "step": 5294 }, { "epoch": 0.05, "learning_rate": 1.5025493178656695e-05, "loss": 1.0309, "step": 5295 }, { "epoch": 0.05, "learning_rate": 1.5023629300589988e-05, "loss": 0.9497, "step": 5296 }, { "epoch": 0.05, "learning_rate": 1.5021765189059563e-05, "loss": 0.9873, "step": 5297 }, { "epoch": 0.05, "learning_rate": 1.501990084415205e-05, "loss": 0.936, "step": 5298 }, { "epoch": 0.05, "learning_rate": 1.5018036265954091e-05, "loss": 0.9611, "step": 5299 }, { "epoch": 0.05, "learning_rate": 1.5016171454552338e-05, "loss": 0.9303, "step": 5300 }, { "epoch": 0.05, "learning_rate": 1.5014306410033457e-05, "loss": 0.9392, "step": 5301 }, { "epoch": 0.05, "learning_rate": 1.501244113248412e-05, "loss": 0.8968, "step": 5302 }, { "epoch": 0.05, "learning_rate": 1.5010575621991014e-05, "loss": 0.9593, "step": 5303 }, { "epoch": 0.05, "learning_rate": 1.5008709878640834e-05, "loss": 0.908, "step": 5304 }, { "epoch": 0.05, "learning_rate": 1.5006843902520288e-05, "loss": 0.9577, "step": 5305 }, { "epoch": 0.05, "learning_rate": 1.5004977693716094e-05, "loss": 0.9534, "step": 5306 }, { "epoch": 0.05, "learning_rate": 1.500311125231498e-05, "loss": 0.9504, "step": 5307 }, { "epoch": 0.05, "learning_rate": 1.5001244578403682e-05, "loss": 0.9378, "step": 5308 }, { "epoch": 0.05, "learning_rate": 1.4999377672068955e-05, "loss": 0.9804, "step": 5309 }, { "epoch": 0.05, "learning_rate": 1.4997510533397559e-05, "loss": 0.9311, "step": 5310 }, { "epoch": 0.05, "learning_rate": 1.4995643162476266e-05, "loss": 1.004, "step": 5311 }, { "epoch": 0.05, "learning_rate": 1.4993775559391855e-05, "loss": 0.9738, "step": 5312 }, { "epoch": 0.05, "learning_rate": 1.4991907724231122e-05, "loss": 0.9782, "step": 5313 }, { "epoch": 0.05, "learning_rate": 1.4990039657080871e-05, "loss": 0.955, "step": 5314 }, { "epoch": 0.05, "learning_rate": 1.4988171358027919e-05, "loss": 0.9433, "step": 5315 }, { "epoch": 0.05, "learning_rate": 1.4986302827159086e-05, "loss": 0.9136, "step": 5316 }, { "epoch": 0.05, "learning_rate": 1.4984434064561215e-05, "loss": 0.9913, "step": 5317 }, { "epoch": 0.05, "learning_rate": 1.4982565070321147e-05, "loss": 0.9516, "step": 5318 }, { "epoch": 0.05, "learning_rate": 1.4980695844525745e-05, "loss": 0.9733, "step": 5319 }, { "epoch": 0.05, "learning_rate": 1.4978826387261877e-05, "loss": 0.9196, "step": 5320 }, { "epoch": 0.05, "learning_rate": 1.4976956698616422e-05, "loss": 0.9877, "step": 5321 }, { "epoch": 0.05, "learning_rate": 1.4975086778676269e-05, "loss": 0.9509, "step": 5322 }, { "epoch": 0.05, "learning_rate": 1.4973216627528318e-05, "loss": 0.9366, "step": 5323 }, { "epoch": 0.05, "learning_rate": 1.4971346245259484e-05, "loss": 0.9701, "step": 5324 }, { "epoch": 0.05, "learning_rate": 1.4969475631956685e-05, "loss": 0.9538, "step": 5325 }, { "epoch": 0.05, "learning_rate": 1.4967604787706863e-05, "loss": 0.9186, "step": 5326 }, { "epoch": 0.05, "learning_rate": 1.496573371259695e-05, "loss": 0.971, "step": 5327 }, { "epoch": 0.05, "learning_rate": 1.4963862406713912e-05, "loss": 0.8934, "step": 5328 }, { "epoch": 0.05, "learning_rate": 1.4961990870144706e-05, "loss": 0.9463, "step": 5329 }, { "epoch": 0.05, "learning_rate": 1.4960119102976314e-05, "loss": 0.9387, "step": 5330 }, { "epoch": 0.05, "learning_rate": 1.4958247105295718e-05, "loss": 0.9689, "step": 5331 }, { "epoch": 0.05, "learning_rate": 1.4956374877189919e-05, "loss": 0.9481, "step": 5332 }, { "epoch": 0.05, "learning_rate": 1.4954502418745924e-05, "loss": 1.0139, "step": 5333 }, { "epoch": 0.05, "learning_rate": 1.4952629730050752e-05, "loss": 0.9933, "step": 5334 }, { "epoch": 0.05, "learning_rate": 1.4950756811191431e-05, "loss": 0.9987, "step": 5335 }, { "epoch": 0.05, "learning_rate": 1.4948883662255007e-05, "loss": 0.9708, "step": 5336 }, { "epoch": 0.05, "learning_rate": 1.4947010283328525e-05, "loss": 0.9827, "step": 5337 }, { "epoch": 0.05, "learning_rate": 1.494513667449905e-05, "loss": 0.9987, "step": 5338 }, { "epoch": 0.06, "learning_rate": 1.4943262835853651e-05, "loss": 0.9194, "step": 5339 }, { "epoch": 0.06, "learning_rate": 1.4941388767479417e-05, "loss": 0.9531, "step": 5340 }, { "epoch": 0.06, "learning_rate": 1.4939514469463437e-05, "loss": 0.9813, "step": 5341 }, { "epoch": 0.06, "learning_rate": 1.4937639941892815e-05, "loss": 0.9557, "step": 5342 }, { "epoch": 0.06, "learning_rate": 1.4935765184854669e-05, "loss": 1.0076, "step": 5343 }, { "epoch": 0.06, "learning_rate": 1.4933890198436127e-05, "loss": 0.9495, "step": 5344 }, { "epoch": 0.06, "learning_rate": 1.4932014982724319e-05, "loss": 0.9879, "step": 5345 }, { "epoch": 0.06, "learning_rate": 1.4930139537806397e-05, "loss": 0.923, "step": 5346 }, { "epoch": 0.06, "learning_rate": 1.4928263863769515e-05, "loss": 0.9752, "step": 5347 }, { "epoch": 0.06, "learning_rate": 1.4926387960700843e-05, "loss": 0.9537, "step": 5348 }, { "epoch": 0.06, "learning_rate": 1.4924511828687565e-05, "loss": 0.9681, "step": 5349 }, { "epoch": 0.06, "learning_rate": 1.4922635467816862e-05, "loss": 0.9697, "step": 5350 }, { "epoch": 0.06, "learning_rate": 1.4920758878175941e-05, "loss": 0.9503, "step": 5351 }, { "epoch": 0.06, "learning_rate": 1.4918882059852012e-05, "loss": 0.9194, "step": 5352 }, { "epoch": 0.06, "learning_rate": 1.4917005012932292e-05, "loss": 0.977, "step": 5353 }, { "epoch": 0.06, "learning_rate": 1.4915127737504017e-05, "loss": 0.9599, "step": 5354 }, { "epoch": 0.06, "learning_rate": 1.4913250233654433e-05, "loss": 1.0009, "step": 5355 }, { "epoch": 0.06, "learning_rate": 1.4911372501470784e-05, "loss": 0.9785, "step": 5356 }, { "epoch": 0.06, "learning_rate": 1.4909494541040346e-05, "loss": 0.9365, "step": 5357 }, { "epoch": 0.06, "learning_rate": 1.4907616352450382e-05, "loss": 0.956, "step": 5358 }, { "epoch": 0.06, "learning_rate": 1.4905737935788187e-05, "loss": 0.9885, "step": 5359 }, { "epoch": 0.06, "learning_rate": 1.4903859291141053e-05, "loss": 0.9753, "step": 5360 }, { "epoch": 0.06, "learning_rate": 1.4901980418596282e-05, "loss": 0.9652, "step": 5361 }, { "epoch": 0.06, "learning_rate": 1.4900101318241197e-05, "loss": 0.9751, "step": 5362 }, { "epoch": 0.06, "learning_rate": 1.4898221990163126e-05, "loss": 0.9329, "step": 5363 }, { "epoch": 0.06, "learning_rate": 1.4896342434449402e-05, "loss": 0.9186, "step": 5364 }, { "epoch": 0.06, "learning_rate": 1.4894462651187379e-05, "loss": 0.9646, "step": 5365 }, { "epoch": 0.06, "learning_rate": 1.4892582640464415e-05, "loss": 0.9441, "step": 5366 }, { "epoch": 0.06, "learning_rate": 1.4890702402367879e-05, "loss": 0.9585, "step": 5367 }, { "epoch": 0.06, "learning_rate": 1.4888821936985149e-05, "loss": 0.9608, "step": 5368 }, { "epoch": 0.06, "learning_rate": 1.4886941244403624e-05, "loss": 0.9666, "step": 5369 }, { "epoch": 0.06, "learning_rate": 1.4885060324710697e-05, "loss": 1.0226, "step": 5370 }, { "epoch": 0.06, "learning_rate": 1.4883179177993783e-05, "loss": 0.9623, "step": 5371 }, { "epoch": 0.06, "learning_rate": 1.4881297804340306e-05, "loss": 0.9805, "step": 5372 }, { "epoch": 0.06, "learning_rate": 1.48794162038377e-05, "loss": 0.9607, "step": 5373 }, { "epoch": 0.06, "learning_rate": 1.4877534376573409e-05, "loss": 0.9781, "step": 5374 }, { "epoch": 0.06, "learning_rate": 1.4875652322634884e-05, "loss": 0.9248, "step": 5375 }, { "epoch": 0.06, "learning_rate": 1.4873770042109592e-05, "loss": 0.9572, "step": 5376 }, { "epoch": 0.06, "learning_rate": 1.4871887535085007e-05, "loss": 0.9788, "step": 5377 }, { "epoch": 0.06, "learning_rate": 1.4870004801648618e-05, "loss": 0.9502, "step": 5378 }, { "epoch": 0.06, "learning_rate": 1.4868121841887922e-05, "loss": 1.0079, "step": 5379 }, { "epoch": 0.06, "learning_rate": 1.486623865589042e-05, "loss": 0.9656, "step": 5380 }, { "epoch": 0.06, "learning_rate": 1.4864355243743637e-05, "loss": 0.9948, "step": 5381 }, { "epoch": 0.06, "learning_rate": 1.4862471605535097e-05, "loss": 0.9597, "step": 5382 }, { "epoch": 0.06, "learning_rate": 1.4860587741352337e-05, "loss": 0.9418, "step": 5383 }, { "epoch": 0.06, "learning_rate": 1.4858703651282909e-05, "loss": 0.9735, "step": 5384 }, { "epoch": 0.06, "learning_rate": 1.485681933541437e-05, "loss": 0.945, "step": 5385 }, { "epoch": 0.06, "learning_rate": 1.4854934793834295e-05, "loss": 0.9147, "step": 5386 }, { "epoch": 0.06, "learning_rate": 1.4853050026630258e-05, "loss": 0.9775, "step": 5387 }, { "epoch": 0.06, "learning_rate": 1.4851165033889855e-05, "loss": 0.9511, "step": 5388 }, { "epoch": 0.06, "learning_rate": 1.4849279815700683e-05, "loss": 0.9917, "step": 5389 }, { "epoch": 0.06, "learning_rate": 1.4847394372150361e-05, "loss": 0.9732, "step": 5390 }, { "epoch": 0.06, "learning_rate": 1.4845508703326504e-05, "loss": 0.9976, "step": 5391 }, { "epoch": 0.06, "learning_rate": 1.4843622809316749e-05, "loss": 0.9453, "step": 5392 }, { "epoch": 0.06, "learning_rate": 1.4841736690208736e-05, "loss": 0.9442, "step": 5393 }, { "epoch": 0.06, "learning_rate": 1.4839850346090124e-05, "loss": 0.9274, "step": 5394 }, { "epoch": 0.06, "learning_rate": 1.4837963777048574e-05, "loss": 1.0421, "step": 5395 }, { "epoch": 0.06, "learning_rate": 1.4836076983171758e-05, "loss": 0.9705, "step": 5396 }, { "epoch": 0.06, "learning_rate": 1.483418996454737e-05, "loss": 0.9774, "step": 5397 }, { "epoch": 0.06, "learning_rate": 1.4832302721263097e-05, "loss": 0.9614, "step": 5398 }, { "epoch": 0.06, "learning_rate": 1.4830415253406647e-05, "loss": 0.9655, "step": 5399 }, { "epoch": 0.06, "learning_rate": 1.4828527561065738e-05, "loss": 0.9529, "step": 5400 }, { "epoch": 0.06, "learning_rate": 1.4826639644328098e-05, "loss": 0.9641, "step": 5401 }, { "epoch": 0.06, "learning_rate": 1.4824751503281463e-05, "loss": 0.9436, "step": 5402 }, { "epoch": 0.06, "learning_rate": 1.4822863138013579e-05, "loss": 0.9681, "step": 5403 }, { "epoch": 0.06, "learning_rate": 1.4820974548612208e-05, "loss": 0.9443, "step": 5404 }, { "epoch": 0.06, "learning_rate": 1.4819085735165117e-05, "loss": 0.9577, "step": 5405 }, { "epoch": 0.06, "learning_rate": 1.4817196697760082e-05, "loss": 0.9218, "step": 5406 }, { "epoch": 0.06, "learning_rate": 1.4815307436484898e-05, "loss": 0.963, "step": 5407 }, { "epoch": 0.06, "learning_rate": 1.481341795142736e-05, "loss": 0.9006, "step": 5408 }, { "epoch": 0.06, "learning_rate": 1.4811528242675279e-05, "loss": 0.9778, "step": 5409 }, { "epoch": 0.06, "learning_rate": 1.480963831031648e-05, "loss": 0.96, "step": 5410 }, { "epoch": 0.06, "learning_rate": 1.480774815443879e-05, "loss": 0.9821, "step": 5411 }, { "epoch": 0.06, "learning_rate": 1.4805857775130052e-05, "loss": 0.9201, "step": 5412 }, { "epoch": 0.06, "learning_rate": 1.4803967172478118e-05, "loss": 0.9753, "step": 5413 }, { "epoch": 0.06, "learning_rate": 1.4802076346570847e-05, "loss": 0.9574, "step": 5414 }, { "epoch": 0.06, "learning_rate": 1.4800185297496115e-05, "loss": 1.0009, "step": 5415 }, { "epoch": 0.06, "learning_rate": 1.4798294025341802e-05, "loss": 0.9254, "step": 5416 }, { "epoch": 0.06, "learning_rate": 1.4796402530195804e-05, "loss": 0.9338, "step": 5417 }, { "epoch": 0.06, "learning_rate": 1.4794510812146028e-05, "loss": 0.9686, "step": 5418 }, { "epoch": 0.06, "learning_rate": 1.479261887128038e-05, "loss": 0.984, "step": 5419 }, { "epoch": 0.06, "learning_rate": 1.479072670768679e-05, "loss": 0.9564, "step": 5420 }, { "epoch": 0.06, "learning_rate": 1.4788834321453189e-05, "loss": 0.9246, "step": 5421 }, { "epoch": 0.06, "learning_rate": 1.4786941712667526e-05, "loss": 0.9191, "step": 5422 }, { "epoch": 0.06, "learning_rate": 1.4785048881417755e-05, "loss": 0.9863, "step": 5423 }, { "epoch": 0.06, "learning_rate": 1.478315582779184e-05, "loss": 0.9458, "step": 5424 }, { "epoch": 0.06, "learning_rate": 1.478126255187776e-05, "loss": 0.9862, "step": 5425 }, { "epoch": 0.06, "learning_rate": 1.4779369053763497e-05, "loss": 0.9337, "step": 5426 }, { "epoch": 0.06, "learning_rate": 1.4777475333537052e-05, "loss": 0.9868, "step": 5427 }, { "epoch": 0.06, "learning_rate": 1.4775581391286436e-05, "loss": 0.9326, "step": 5428 }, { "epoch": 0.06, "learning_rate": 1.4773687227099655e-05, "loss": 0.95, "step": 5429 }, { "epoch": 0.06, "learning_rate": 1.4771792841064744e-05, "loss": 0.9094, "step": 5430 }, { "epoch": 0.06, "learning_rate": 1.4769898233269742e-05, "loss": 0.947, "step": 5431 }, { "epoch": 0.06, "learning_rate": 1.4768003403802693e-05, "loss": 0.9308, "step": 5432 }, { "epoch": 0.06, "learning_rate": 1.4766108352751661e-05, "loss": 0.9422, "step": 5433 }, { "epoch": 0.06, "learning_rate": 1.4764213080204708e-05, "loss": 0.9352, "step": 5434 }, { "epoch": 0.06, "learning_rate": 1.476231758624992e-05, "loss": 0.9098, "step": 5435 }, { "epoch": 0.06, "learning_rate": 1.4760421870975381e-05, "loss": 0.9324, "step": 5436 }, { "epoch": 0.06, "learning_rate": 1.4758525934469195e-05, "loss": 0.954, "step": 5437 }, { "epoch": 0.06, "learning_rate": 1.475662977681947e-05, "loss": 0.967, "step": 5438 }, { "epoch": 0.06, "learning_rate": 1.4754733398114326e-05, "loss": 0.9359, "step": 5439 }, { "epoch": 0.06, "learning_rate": 1.4752836798441897e-05, "loss": 0.9513, "step": 5440 }, { "epoch": 0.06, "learning_rate": 1.475093997789032e-05, "loss": 0.9864, "step": 5441 }, { "epoch": 0.06, "learning_rate": 1.4749042936547745e-05, "loss": 0.9273, "step": 5442 }, { "epoch": 0.06, "learning_rate": 1.474714567450234e-05, "loss": 0.9674, "step": 5443 }, { "epoch": 0.06, "learning_rate": 1.4745248191842272e-05, "loss": 0.9172, "step": 5444 }, { "epoch": 0.06, "learning_rate": 1.4743350488655721e-05, "loss": 0.9696, "step": 5445 }, { "epoch": 0.06, "learning_rate": 1.4741452565030883e-05, "loss": 0.9466, "step": 5446 }, { "epoch": 0.06, "learning_rate": 1.473955442105596e-05, "loss": 0.9805, "step": 5447 }, { "epoch": 0.06, "learning_rate": 1.4737656056819162e-05, "loss": 0.9675, "step": 5448 }, { "epoch": 0.06, "learning_rate": 1.4735757472408714e-05, "loss": 0.9957, "step": 5449 }, { "epoch": 0.06, "learning_rate": 1.473385866791285e-05, "loss": 0.9144, "step": 5450 }, { "epoch": 0.06, "learning_rate": 1.4731959643419815e-05, "loss": 0.978, "step": 5451 }, { "epoch": 0.06, "learning_rate": 1.4730060399017855e-05, "loss": 0.9533, "step": 5452 }, { "epoch": 0.06, "learning_rate": 1.4728160934795242e-05, "loss": 0.9549, "step": 5453 }, { "epoch": 0.06, "learning_rate": 1.4726261250840244e-05, "loss": 0.9424, "step": 5454 }, { "epoch": 0.06, "learning_rate": 1.4724361347241147e-05, "loss": 1.0223, "step": 5455 }, { "epoch": 0.06, "learning_rate": 1.472246122408625e-05, "loss": 0.9753, "step": 5456 }, { "epoch": 0.06, "learning_rate": 1.4720560881463852e-05, "loss": 0.989, "step": 5457 }, { "epoch": 0.06, "learning_rate": 1.4718660319462272e-05, "loss": 0.9386, "step": 5458 }, { "epoch": 0.06, "learning_rate": 1.471675953816983e-05, "loss": 0.9615, "step": 5459 }, { "epoch": 0.06, "learning_rate": 1.4714858537674867e-05, "loss": 0.9751, "step": 5460 }, { "epoch": 0.06, "learning_rate": 1.4712957318065724e-05, "loss": 0.9499, "step": 5461 }, { "epoch": 0.06, "learning_rate": 1.4711055879430757e-05, "loss": 0.9572, "step": 5462 }, { "epoch": 0.06, "learning_rate": 1.4709154221858334e-05, "loss": 0.9673, "step": 5463 }, { "epoch": 0.06, "learning_rate": 1.470725234543683e-05, "loss": 0.9807, "step": 5464 }, { "epoch": 0.06, "learning_rate": 1.470535025025463e-05, "loss": 0.9961, "step": 5465 }, { "epoch": 0.06, "learning_rate": 1.4703447936400135e-05, "loss": 0.9469, "step": 5466 }, { "epoch": 0.06, "learning_rate": 1.4701545403961745e-05, "loss": 0.9416, "step": 5467 }, { "epoch": 0.06, "learning_rate": 1.469964265302788e-05, "loss": 0.9499, "step": 5468 }, { "epoch": 0.06, "learning_rate": 1.4697739683686965e-05, "loss": 0.9819, "step": 5469 }, { "epoch": 0.06, "learning_rate": 1.4695836496027438e-05, "loss": 0.9744, "step": 5470 }, { "epoch": 0.06, "learning_rate": 1.4693933090137747e-05, "loss": 0.9866, "step": 5471 }, { "epoch": 0.06, "learning_rate": 1.4692029466106348e-05, "loss": 0.9123, "step": 5472 }, { "epoch": 0.06, "learning_rate": 1.4690125624021709e-05, "loss": 0.9451, "step": 5473 }, { "epoch": 0.06, "learning_rate": 1.4688221563972307e-05, "loss": 0.9729, "step": 5474 }, { "epoch": 0.06, "learning_rate": 1.468631728604663e-05, "loss": 0.9563, "step": 5475 }, { "epoch": 0.06, "learning_rate": 1.4684412790333177e-05, "loss": 0.938, "step": 5476 }, { "epoch": 0.06, "learning_rate": 1.4682508076920452e-05, "loss": 0.9868, "step": 5477 }, { "epoch": 0.06, "learning_rate": 1.4680603145896974e-05, "loss": 0.8986, "step": 5478 }, { "epoch": 0.06, "learning_rate": 1.4678697997351277e-05, "loss": 0.9667, "step": 5479 }, { "epoch": 0.06, "learning_rate": 1.467679263137189e-05, "loss": 0.9642, "step": 5480 }, { "epoch": 0.06, "learning_rate": 1.4674887048047368e-05, "loss": 1.0024, "step": 5481 }, { "epoch": 0.06, "learning_rate": 1.4672981247466269e-05, "loss": 0.9742, "step": 5482 }, { "epoch": 0.06, "learning_rate": 1.4671075229717157e-05, "loss": 0.9335, "step": 5483 }, { "epoch": 0.06, "learning_rate": 1.4669168994888614e-05, "loss": 0.9393, "step": 5484 }, { "epoch": 0.06, "learning_rate": 1.4667262543069232e-05, "loss": 0.9508, "step": 5485 }, { "epoch": 0.06, "learning_rate": 1.46653558743476e-05, "loss": 0.9884, "step": 5486 }, { "epoch": 0.06, "learning_rate": 1.4663448988812336e-05, "loss": 0.9863, "step": 5487 }, { "epoch": 0.06, "learning_rate": 1.4661541886552053e-05, "loss": 0.9338, "step": 5488 }, { "epoch": 0.07, "learning_rate": 1.4659634567655385e-05, "loss": 0.9557, "step": 5489 }, { "epoch": 0.07, "learning_rate": 1.4657727032210971e-05, "loss": 0.9575, "step": 5490 }, { "epoch": 0.07, "learning_rate": 1.4655819280307453e-05, "loss": 0.9601, "step": 5491 }, { "epoch": 0.07, "learning_rate": 1.4653911312033496e-05, "loss": 0.953, "step": 5492 }, { "epoch": 0.07, "learning_rate": 1.465200312747777e-05, "loss": 0.9551, "step": 5493 }, { "epoch": 0.07, "learning_rate": 1.465009472672895e-05, "loss": 0.894, "step": 5494 }, { "epoch": 0.07, "learning_rate": 1.4648186109875728e-05, "loss": 0.9864, "step": 5495 }, { "epoch": 0.07, "learning_rate": 1.4646277277006805e-05, "loss": 1.0162, "step": 5496 }, { "epoch": 0.07, "learning_rate": 1.464436822821089e-05, "loss": 0.9808, "step": 5497 }, { "epoch": 0.07, "learning_rate": 1.4642458963576698e-05, "loss": 0.9099, "step": 5498 }, { "epoch": 0.07, "learning_rate": 1.4640549483192966e-05, "loss": 0.9722, "step": 5499 }, { "epoch": 0.07, "learning_rate": 1.4638639787148423e-05, "loss": 0.9817, "step": 5500 }, { "epoch": 0.07, "learning_rate": 1.4636729875531826e-05, "loss": 0.9771, "step": 5501 }, { "epoch": 0.07, "learning_rate": 1.4634819748431934e-05, "loss": 0.9998, "step": 5502 }, { "epoch": 0.07, "learning_rate": 1.4632909405937516e-05, "loss": 0.9901, "step": 5503 }, { "epoch": 0.07, "learning_rate": 1.4630998848137352e-05, "loss": 0.9969, "step": 5504 }, { "epoch": 0.07, "learning_rate": 1.462908807512023e-05, "loss": 0.9915, "step": 5505 }, { "epoch": 0.07, "learning_rate": 1.462717708697495e-05, "loss": 0.9653, "step": 5506 }, { "epoch": 0.07, "learning_rate": 1.4625265883790323e-05, "loss": 0.9627, "step": 5507 }, { "epoch": 0.07, "learning_rate": 1.4623354465655166e-05, "loss": 0.9579, "step": 5508 }, { "epoch": 0.07, "learning_rate": 1.462144283265831e-05, "loss": 0.9449, "step": 5509 }, { "epoch": 0.07, "learning_rate": 1.4619530984888596e-05, "loss": 0.9619, "step": 5510 }, { "epoch": 0.07, "learning_rate": 1.4617618922434873e-05, "loss": 1.0179, "step": 5511 }, { "epoch": 0.07, "learning_rate": 1.4615706645385996e-05, "loss": 0.9876, "step": 5512 }, { "epoch": 0.07, "learning_rate": 1.4613794153830842e-05, "loss": 0.9535, "step": 5513 }, { "epoch": 0.07, "learning_rate": 1.4611881447858287e-05, "loss": 0.9806, "step": 5514 }, { "epoch": 0.07, "learning_rate": 1.4609968527557217e-05, "loss": 0.9536, "step": 5515 }, { "epoch": 0.07, "learning_rate": 1.4608055393016539e-05, "loss": 0.956, "step": 5516 }, { "epoch": 0.07, "learning_rate": 1.4606142044325152e-05, "loss": 1.024, "step": 5517 }, { "epoch": 0.07, "learning_rate": 1.4604228481571983e-05, "loss": 0.9115, "step": 5518 }, { "epoch": 0.07, "learning_rate": 1.460231470484596e-05, "loss": 0.9814, "step": 5519 }, { "epoch": 0.07, "learning_rate": 1.4600400714236023e-05, "loss": 1.0258, "step": 5520 }, { "epoch": 0.07, "learning_rate": 1.4598486509831116e-05, "loss": 0.9117, "step": 5521 }, { "epoch": 0.07, "learning_rate": 1.4596572091720204e-05, "loss": 0.9256, "step": 5522 }, { "epoch": 0.07, "learning_rate": 1.4594657459992252e-05, "loss": 0.9463, "step": 5523 }, { "epoch": 0.07, "learning_rate": 1.459274261473624e-05, "loss": 0.9571, "step": 5524 }, { "epoch": 0.07, "learning_rate": 1.4590827556041158e-05, "loss": 0.9469, "step": 5525 }, { "epoch": 0.07, "learning_rate": 1.4588912283996007e-05, "loss": 0.9446, "step": 5526 }, { "epoch": 0.07, "learning_rate": 1.458699679868979e-05, "loss": 1.0172, "step": 5527 }, { "epoch": 0.07, "learning_rate": 1.458508110021153e-05, "loss": 0.9626, "step": 5528 }, { "epoch": 0.07, "learning_rate": 1.458316518865025e-05, "loss": 0.9954, "step": 5529 }, { "epoch": 0.07, "learning_rate": 1.4581249064094993e-05, "loss": 0.9507, "step": 5530 }, { "epoch": 0.07, "learning_rate": 1.4579332726634809e-05, "loss": 0.9894, "step": 5531 }, { "epoch": 0.07, "learning_rate": 1.4577416176358752e-05, "loss": 0.9997, "step": 5532 }, { "epoch": 0.07, "learning_rate": 1.4575499413355894e-05, "loss": 0.9716, "step": 5533 }, { "epoch": 0.07, "learning_rate": 1.4573582437715309e-05, "loss": 0.9949, "step": 5534 }, { "epoch": 0.07, "learning_rate": 1.4571665249526083e-05, "loss": 0.9566, "step": 5535 }, { "epoch": 0.07, "learning_rate": 1.456974784887732e-05, "loss": 0.9297, "step": 5536 }, { "epoch": 0.07, "learning_rate": 1.4567830235858127e-05, "loss": 0.979, "step": 5537 }, { "epoch": 0.07, "learning_rate": 1.4565912410557617e-05, "loss": 0.9673, "step": 5538 }, { "epoch": 0.07, "learning_rate": 1.456399437306492e-05, "loss": 0.9336, "step": 5539 }, { "epoch": 0.07, "learning_rate": 1.4562076123469174e-05, "loss": 0.9552, "step": 5540 }, { "epoch": 0.07, "learning_rate": 1.456015766185952e-05, "loss": 0.9715, "step": 5541 }, { "epoch": 0.07, "learning_rate": 1.4558238988325124e-05, "loss": 0.9888, "step": 5542 }, { "epoch": 0.07, "learning_rate": 1.4556320102955148e-05, "loss": 0.9674, "step": 5543 }, { "epoch": 0.07, "learning_rate": 1.4554401005838768e-05, "loss": 0.9996, "step": 5544 }, { "epoch": 0.07, "learning_rate": 1.4552481697065172e-05, "loss": 1.0276, "step": 5545 }, { "epoch": 0.07, "learning_rate": 1.4550562176723555e-05, "loss": 0.9646, "step": 5546 }, { "epoch": 0.07, "learning_rate": 1.4548642444903122e-05, "loss": 0.975, "step": 5547 }, { "epoch": 0.07, "learning_rate": 1.4546722501693096e-05, "loss": 0.9456, "step": 5548 }, { "epoch": 0.07, "learning_rate": 1.4544802347182692e-05, "loss": 0.9881, "step": 5549 }, { "epoch": 0.07, "learning_rate": 1.4542881981461154e-05, "loss": 0.9441, "step": 5550 }, { "epoch": 0.07, "learning_rate": 1.4540961404617725e-05, "loss": 0.9676, "step": 5551 }, { "epoch": 0.07, "learning_rate": 1.4539040616741657e-05, "loss": 0.9502, "step": 5552 }, { "epoch": 0.07, "learning_rate": 1.453711961792222e-05, "loss": 0.9787, "step": 5553 }, { "epoch": 0.07, "learning_rate": 1.4535198408248685e-05, "loss": 0.955, "step": 5554 }, { "epoch": 0.07, "learning_rate": 1.453327698781034e-05, "loss": 0.97, "step": 5555 }, { "epoch": 0.07, "learning_rate": 1.4531355356696475e-05, "loss": 0.9456, "step": 5556 }, { "epoch": 0.07, "learning_rate": 1.4529433514996397e-05, "loss": 0.9256, "step": 5557 }, { "epoch": 0.07, "learning_rate": 1.4527511462799417e-05, "loss": 0.9326, "step": 5558 }, { "epoch": 0.07, "learning_rate": 1.4525589200194865e-05, "loss": 0.9517, "step": 5559 }, { "epoch": 0.07, "learning_rate": 1.4523666727272072e-05, "loss": 0.9449, "step": 5560 }, { "epoch": 0.07, "learning_rate": 1.4521744044120376e-05, "loss": 0.9696, "step": 5561 }, { "epoch": 0.07, "learning_rate": 1.4519821150829138e-05, "loss": 0.932, "step": 5562 }, { "epoch": 0.07, "learning_rate": 1.4517898047487715e-05, "loss": 0.9646, "step": 5563 }, { "epoch": 0.07, "learning_rate": 1.451597473418548e-05, "loss": 0.9286, "step": 5564 }, { "epoch": 0.07, "learning_rate": 1.4514051211011821e-05, "loss": 0.9591, "step": 5565 }, { "epoch": 0.07, "learning_rate": 1.4512127478056126e-05, "loss": 0.9759, "step": 5566 }, { "epoch": 0.07, "learning_rate": 1.4510203535407795e-05, "loss": 0.976, "step": 5567 }, { "epoch": 0.07, "learning_rate": 1.4508279383156244e-05, "loss": 0.9451, "step": 5568 }, { "epoch": 0.07, "learning_rate": 1.4506355021390888e-05, "loss": 0.9518, "step": 5569 }, { "epoch": 0.07, "learning_rate": 1.4504430450201164e-05, "loss": 0.9425, "step": 5570 }, { "epoch": 0.07, "learning_rate": 1.4502505669676512e-05, "loss": 0.981, "step": 5571 }, { "epoch": 0.07, "learning_rate": 1.4500580679906379e-05, "loss": 0.9633, "step": 5572 }, { "epoch": 0.07, "learning_rate": 1.4498655480980228e-05, "loss": 0.9927, "step": 5573 }, { "epoch": 0.07, "learning_rate": 1.4496730072987535e-05, "loss": 0.9522, "step": 5574 }, { "epoch": 0.07, "learning_rate": 1.4494804456017768e-05, "loss": 1.0092, "step": 5575 }, { "epoch": 0.07, "learning_rate": 1.449287863016042e-05, "loss": 0.9633, "step": 5576 }, { "epoch": 0.07, "learning_rate": 1.4490952595504997e-05, "loss": 1.0007, "step": 5577 }, { "epoch": 0.07, "learning_rate": 1.4489026352141001e-05, "loss": 0.9892, "step": 5578 }, { "epoch": 0.07, "learning_rate": 1.4487099900157952e-05, "loss": 0.9622, "step": 5579 }, { "epoch": 0.07, "learning_rate": 1.4485173239645378e-05, "loss": 0.9807, "step": 5580 }, { "epoch": 0.07, "learning_rate": 1.4483246370692819e-05, "loss": 0.9746, "step": 5581 }, { "epoch": 0.07, "learning_rate": 1.4481319293389819e-05, "loss": 0.9816, "step": 5582 }, { "epoch": 0.07, "learning_rate": 1.4479392007825941e-05, "loss": 0.956, "step": 5583 }, { "epoch": 0.07, "learning_rate": 1.4477464514090745e-05, "loss": 0.9514, "step": 5584 }, { "epoch": 0.07, "learning_rate": 1.4475536812273814e-05, "loss": 0.9634, "step": 5585 }, { "epoch": 0.07, "learning_rate": 1.4473608902464728e-05, "loss": 0.9613, "step": 5586 }, { "epoch": 0.07, "learning_rate": 1.4471680784753089e-05, "loss": 0.989, "step": 5587 }, { "epoch": 0.07, "learning_rate": 1.44697524592285e-05, "loss": 0.9437, "step": 5588 }, { "epoch": 0.07, "learning_rate": 1.4467823925980575e-05, "loss": 0.9978, "step": 5589 }, { "epoch": 0.07, "learning_rate": 1.4465895185098942e-05, "loss": 0.9626, "step": 5590 }, { "epoch": 0.07, "learning_rate": 1.4463966236673233e-05, "loss": 0.9577, "step": 5591 }, { "epoch": 0.07, "learning_rate": 1.4462037080793092e-05, "loss": 0.9368, "step": 5592 }, { "epoch": 0.07, "learning_rate": 1.4460107717548172e-05, "loss": 0.9911, "step": 5593 }, { "epoch": 0.07, "learning_rate": 1.4458178147028146e-05, "loss": 0.9372, "step": 5594 }, { "epoch": 0.07, "learning_rate": 1.4456248369322672e-05, "loss": 0.9725, "step": 5595 }, { "epoch": 0.07, "learning_rate": 1.4454318384521444e-05, "loss": 0.9391, "step": 5596 }, { "epoch": 0.07, "learning_rate": 1.4452388192714152e-05, "loss": 0.9523, "step": 5597 }, { "epoch": 0.07, "learning_rate": 1.4450457793990496e-05, "loss": 0.9929, "step": 5598 }, { "epoch": 0.07, "learning_rate": 1.4448527188440187e-05, "loss": 1.0087, "step": 5599 }, { "epoch": 0.07, "learning_rate": 1.444659637615295e-05, "loss": 1.0074, "step": 5600 }, { "epoch": 0.07, "learning_rate": 1.4444665357218512e-05, "loss": 1.0272, "step": 5601 }, { "epoch": 0.07, "learning_rate": 1.4442734131726616e-05, "loss": 1.0097, "step": 5602 }, { "epoch": 0.07, "learning_rate": 1.444080269976701e-05, "loss": 0.9892, "step": 5603 }, { "epoch": 0.07, "learning_rate": 1.4438871061429456e-05, "loss": 0.9664, "step": 5604 }, { "epoch": 0.07, "learning_rate": 1.4436939216803723e-05, "loss": 0.9676, "step": 5605 }, { "epoch": 0.07, "learning_rate": 1.4435007165979591e-05, "loss": 0.968, "step": 5606 }, { "epoch": 0.07, "learning_rate": 1.443307490904684e-05, "loss": 1.0223, "step": 5607 }, { "epoch": 0.07, "learning_rate": 1.4431142446095282e-05, "loss": 0.9631, "step": 5608 }, { "epoch": 0.07, "learning_rate": 1.442920977721471e-05, "loss": 0.9419, "step": 5609 }, { "epoch": 0.07, "learning_rate": 1.4427276902494951e-05, "loss": 0.9024, "step": 5610 }, { "epoch": 0.07, "learning_rate": 1.442534382202583e-05, "loss": 0.9366, "step": 5611 }, { "epoch": 0.07, "learning_rate": 1.4423410535897184e-05, "loss": 1.0125, "step": 5612 }, { "epoch": 0.07, "learning_rate": 1.4421477044198853e-05, "loss": 0.9898, "step": 5613 }, { "epoch": 0.07, "learning_rate": 1.4419543347020698e-05, "loss": 0.9778, "step": 5614 }, { "epoch": 0.07, "learning_rate": 1.441760944445258e-05, "loss": 0.9802, "step": 5615 }, { "epoch": 0.07, "learning_rate": 1.4415675336584378e-05, "loss": 0.9482, "step": 5616 }, { "epoch": 0.07, "learning_rate": 1.4413741023505977e-05, "loss": 0.9552, "step": 5617 }, { "epoch": 0.07, "learning_rate": 1.4411806505307262e-05, "loss": 0.9199, "step": 5618 }, { "epoch": 0.07, "learning_rate": 1.4409871782078142e-05, "loss": 0.9668, "step": 5619 }, { "epoch": 0.07, "learning_rate": 1.4407936853908535e-05, "loss": 0.9571, "step": 5620 }, { "epoch": 0.07, "learning_rate": 1.4406001720888352e-05, "loss": 0.9737, "step": 5621 }, { "epoch": 0.07, "learning_rate": 1.440406638310753e-05, "loss": 0.9223, "step": 5622 }, { "epoch": 0.07, "learning_rate": 1.4402130840656012e-05, "loss": 0.9556, "step": 5623 }, { "epoch": 0.07, "learning_rate": 1.4400195093623745e-05, "loss": 0.9518, "step": 5624 }, { "epoch": 0.07, "learning_rate": 1.439825914210069e-05, "loss": 0.9815, "step": 5625 }, { "epoch": 0.07, "learning_rate": 1.4396322986176823e-05, "loss": 0.9623, "step": 5626 }, { "epoch": 0.07, "learning_rate": 1.4394386625942112e-05, "loss": 0.9605, "step": 5627 }, { "epoch": 0.07, "learning_rate": 1.4392450061486555e-05, "loss": 0.9655, "step": 5628 }, { "epoch": 0.07, "learning_rate": 1.4390513292900147e-05, "loss": 0.9963, "step": 5629 }, { "epoch": 0.07, "learning_rate": 1.4388576320272894e-05, "loss": 0.9398, "step": 5630 }, { "epoch": 0.07, "learning_rate": 1.4386639143694817e-05, "loss": 0.9374, "step": 5631 }, { "epoch": 0.07, "learning_rate": 1.4384701763255936e-05, "loss": 0.9397, "step": 5632 }, { "epoch": 0.07, "learning_rate": 1.4382764179046294e-05, "loss": 0.9328, "step": 5633 }, { "epoch": 0.07, "learning_rate": 1.4380826391155935e-05, "loss": 0.9596, "step": 5634 }, { "epoch": 0.07, "learning_rate": 1.4378888399674913e-05, "loss": 0.9985, "step": 5635 }, { "epoch": 0.07, "learning_rate": 1.4376950204693289e-05, "loss": 0.9332, "step": 5636 }, { "epoch": 0.07, "learning_rate": 1.4375011806301144e-05, "loss": 0.997, "step": 5637 }, { "epoch": 0.07, "learning_rate": 1.4373073204588556e-05, "loss": 0.9956, "step": 5638 }, { "epoch": 0.08, "learning_rate": 1.4371134399645618e-05, "loss": 0.9906, "step": 5639 }, { "epoch": 0.08, "learning_rate": 1.4369195391562438e-05, "loss": 0.9551, "step": 5640 }, { "epoch": 0.08, "learning_rate": 1.436725618042912e-05, "loss": 0.9746, "step": 5641 }, { "epoch": 0.08, "learning_rate": 1.4365316766335793e-05, "loss": 0.9095, "step": 5642 }, { "epoch": 0.08, "learning_rate": 1.4363377149372584e-05, "loss": 1.0089, "step": 5643 }, { "epoch": 0.08, "learning_rate": 1.4361437329629628e-05, "loss": 0.9667, "step": 5644 }, { "epoch": 0.08, "learning_rate": 1.4359497307197083e-05, "loss": 1.0175, "step": 5645 }, { "epoch": 0.08, "learning_rate": 1.4357557082165103e-05, "loss": 0.9462, "step": 5646 }, { "epoch": 0.08, "learning_rate": 1.4355616654623856e-05, "loss": 0.9828, "step": 5647 }, { "epoch": 0.08, "learning_rate": 1.4353676024663521e-05, "loss": 0.9538, "step": 5648 }, { "epoch": 0.08, "learning_rate": 1.4351735192374287e-05, "loss": 0.9315, "step": 5649 }, { "epoch": 0.08, "learning_rate": 1.4349794157846348e-05, "loss": 0.9696, "step": 5650 }, { "epoch": 0.08, "learning_rate": 1.4347852921169912e-05, "loss": 1.0108, "step": 5651 }, { "epoch": 0.08, "learning_rate": 1.4345911482435194e-05, "loss": 0.9553, "step": 5652 }, { "epoch": 0.08, "learning_rate": 1.4343969841732412e-05, "loss": 0.948, "step": 5653 }, { "epoch": 0.08, "learning_rate": 1.4342027999151811e-05, "loss": 0.9489, "step": 5654 }, { "epoch": 0.08, "learning_rate": 1.4340085954783626e-05, "loss": 0.9633, "step": 5655 }, { "epoch": 0.08, "learning_rate": 1.4338143708718115e-05, "loss": 0.9543, "step": 5656 }, { "epoch": 0.08, "learning_rate": 1.4336201261045538e-05, "loss": 0.9533, "step": 5657 }, { "epoch": 0.08, "learning_rate": 1.4334258611856168e-05, "loss": 0.9698, "step": 5658 }, { "epoch": 0.08, "learning_rate": 1.4332315761240283e-05, "loss": 0.9331, "step": 5659 }, { "epoch": 0.08, "learning_rate": 1.4330372709288178e-05, "loss": 0.9562, "step": 5660 }, { "epoch": 0.08, "learning_rate": 1.4328429456090146e-05, "loss": 1.0009, "step": 5661 }, { "epoch": 0.08, "learning_rate": 1.4326486001736501e-05, "loss": 0.9513, "step": 5662 }, { "epoch": 0.08, "learning_rate": 1.4324542346317562e-05, "loss": 0.9727, "step": 5663 }, { "epoch": 0.08, "learning_rate": 1.4322598489923655e-05, "loss": 0.96, "step": 5664 }, { "epoch": 0.08, "learning_rate": 1.4320654432645116e-05, "loss": 1.0178, "step": 5665 }, { "epoch": 0.08, "learning_rate": 1.4318710174572294e-05, "loss": 0.9266, "step": 5666 }, { "epoch": 0.08, "learning_rate": 1.4316765715795542e-05, "loss": 0.9473, "step": 5667 }, { "epoch": 0.08, "learning_rate": 1.4314821056405226e-05, "loss": 0.9232, "step": 5668 }, { "epoch": 0.08, "learning_rate": 1.4312876196491722e-05, "loss": 0.9538, "step": 5669 }, { "epoch": 0.08, "learning_rate": 1.4310931136145412e-05, "loss": 0.9652, "step": 5670 }, { "epoch": 0.08, "learning_rate": 1.4308985875456687e-05, "loss": 0.9703, "step": 5671 }, { "epoch": 0.08, "learning_rate": 1.4307040414515956e-05, "loss": 0.9298, "step": 5672 }, { "epoch": 0.08, "learning_rate": 1.4305094753413624e-05, "loss": 0.9794, "step": 5673 }, { "epoch": 0.08, "learning_rate": 1.4303148892240115e-05, "loss": 0.9312, "step": 5674 }, { "epoch": 0.08, "learning_rate": 1.4301202831085861e-05, "loss": 0.9286, "step": 5675 }, { "epoch": 0.08, "learning_rate": 1.4299256570041295e-05, "loss": 0.9563, "step": 5676 }, { "epoch": 0.08, "learning_rate": 1.4297310109196872e-05, "loss": 0.9722, "step": 5677 }, { "epoch": 0.08, "learning_rate": 1.4295363448643048e-05, "loss": 0.9686, "step": 5678 }, { "epoch": 0.08, "learning_rate": 1.4293416588470289e-05, "loss": 0.9987, "step": 5679 }, { "epoch": 0.08, "learning_rate": 1.4291469528769074e-05, "loss": 0.9441, "step": 5680 }, { "epoch": 0.08, "learning_rate": 1.428952226962989e-05, "loss": 0.9879, "step": 5681 }, { "epoch": 0.08, "learning_rate": 1.4287574811143227e-05, "loss": 0.9564, "step": 5682 }, { "epoch": 0.08, "learning_rate": 1.4285627153399594e-05, "loss": 1.006, "step": 5683 }, { "epoch": 0.08, "learning_rate": 1.4283679296489505e-05, "loss": 0.9838, "step": 5684 }, { "epoch": 0.08, "learning_rate": 1.4281731240503478e-05, "loss": 0.9793, "step": 5685 }, { "epoch": 0.08, "learning_rate": 1.427978298553205e-05, "loss": 0.9703, "step": 5686 }, { "epoch": 0.08, "learning_rate": 1.4277834531665761e-05, "loss": 0.9931, "step": 5687 }, { "epoch": 0.08, "learning_rate": 1.4275885878995164e-05, "loss": 0.9462, "step": 5688 }, { "epoch": 0.08, "learning_rate": 1.4273937027610818e-05, "loss": 0.9894, "step": 5689 }, { "epoch": 0.08, "learning_rate": 1.4271987977603287e-05, "loss": 1.0138, "step": 5690 }, { "epoch": 0.08, "learning_rate": 1.4270038729063153e-05, "loss": 0.9426, "step": 5691 }, { "epoch": 0.08, "learning_rate": 1.426808928208101e-05, "loss": 0.9616, "step": 5692 }, { "epoch": 0.08, "learning_rate": 1.4266139636747442e-05, "loss": 1.0003, "step": 5693 }, { "epoch": 0.08, "learning_rate": 1.4264189793153068e-05, "loss": 0.9473, "step": 5694 }, { "epoch": 0.08, "learning_rate": 1.4262239751388493e-05, "loss": 0.9597, "step": 5695 }, { "epoch": 0.08, "learning_rate": 1.4260289511544347e-05, "loss": 0.9498, "step": 5696 }, { "epoch": 0.08, "learning_rate": 1.4258339073711266e-05, "loss": 0.9254, "step": 5697 }, { "epoch": 0.08, "learning_rate": 1.4256388437979887e-05, "loss": 0.9725, "step": 5698 }, { "epoch": 0.08, "learning_rate": 1.4254437604440863e-05, "loss": 0.9601, "step": 5699 }, { "epoch": 0.08, "learning_rate": 1.4252486573184859e-05, "loss": 0.9669, "step": 5700 }, { "epoch": 0.08, "learning_rate": 1.4250535344302543e-05, "loss": 0.9455, "step": 5701 }, { "epoch": 0.08, "learning_rate": 1.4248583917884595e-05, "loss": 0.9715, "step": 5702 }, { "epoch": 0.08, "learning_rate": 1.4246632294021705e-05, "loss": 0.9599, "step": 5703 }, { "epoch": 0.08, "learning_rate": 1.424468047280457e-05, "loss": 0.9617, "step": 5704 }, { "epoch": 0.08, "learning_rate": 1.4242728454323897e-05, "loss": 0.9489, "step": 5705 }, { "epoch": 0.08, "learning_rate": 1.4240776238670402e-05, "loss": 0.9637, "step": 5706 }, { "epoch": 0.08, "learning_rate": 1.4238823825934813e-05, "loss": 0.9673, "step": 5707 }, { "epoch": 0.08, "learning_rate": 1.4236871216207862e-05, "loss": 0.9709, "step": 5708 }, { "epoch": 0.08, "learning_rate": 1.4234918409580294e-05, "loss": 0.9496, "step": 5709 }, { "epoch": 0.08, "learning_rate": 1.423296540614286e-05, "loss": 0.9689, "step": 5710 }, { "epoch": 0.08, "learning_rate": 1.4231012205986328e-05, "loss": 0.9982, "step": 5711 }, { "epoch": 0.08, "learning_rate": 1.4229058809201463e-05, "loss": 0.9453, "step": 5712 }, { "epoch": 0.08, "learning_rate": 1.4227105215879049e-05, "loss": 0.9568, "step": 5713 }, { "epoch": 0.08, "learning_rate": 1.4225151426109873e-05, "loss": 0.9473, "step": 5714 }, { "epoch": 0.08, "learning_rate": 1.4223197439984735e-05, "loss": 0.9506, "step": 5715 }, { "epoch": 0.08, "learning_rate": 1.4221243257594443e-05, "loss": 0.9337, "step": 5716 }, { "epoch": 0.08, "learning_rate": 1.4219288879029813e-05, "loss": 0.9805, "step": 5717 }, { "epoch": 0.08, "learning_rate": 1.4217334304381676e-05, "loss": 0.9271, "step": 5718 }, { "epoch": 0.08, "learning_rate": 1.4215379533740858e-05, "loss": 0.9725, "step": 5719 }, { "epoch": 0.08, "learning_rate": 1.4213424567198212e-05, "loss": 0.978, "step": 5720 }, { "epoch": 0.08, "learning_rate": 1.4211469404844587e-05, "loss": 0.9779, "step": 5721 }, { "epoch": 0.08, "learning_rate": 1.4209514046770845e-05, "loss": 0.973, "step": 5722 }, { "epoch": 0.08, "learning_rate": 1.420755849306786e-05, "loss": 0.9274, "step": 5723 }, { "epoch": 0.08, "learning_rate": 1.420560274382651e-05, "loss": 1.0081, "step": 5724 }, { "epoch": 0.08, "learning_rate": 1.4203646799137688e-05, "loss": 0.9863, "step": 5725 }, { "epoch": 0.08, "learning_rate": 1.420169065909229e-05, "loss": 0.9882, "step": 5726 }, { "epoch": 0.08, "learning_rate": 1.4199734323781228e-05, "loss": 0.9463, "step": 5727 }, { "epoch": 0.08, "learning_rate": 1.419777779329541e-05, "loss": 0.9478, "step": 5728 }, { "epoch": 0.08, "learning_rate": 1.4195821067725773e-05, "loss": 0.9777, "step": 5729 }, { "epoch": 0.08, "learning_rate": 1.4193864147163248e-05, "loss": 0.9817, "step": 5730 }, { "epoch": 0.08, "learning_rate": 1.4191907031698774e-05, "loss": 1.0163, "step": 5731 }, { "epoch": 0.08, "learning_rate": 1.4189949721423313e-05, "loss": 0.9584, "step": 5732 }, { "epoch": 0.08, "learning_rate": 1.418799221642782e-05, "loss": 0.992, "step": 5733 }, { "epoch": 0.08, "learning_rate": 1.4186034516803269e-05, "loss": 0.9649, "step": 5734 }, { "epoch": 0.08, "learning_rate": 1.4184076622640646e-05, "loss": 0.9735, "step": 5735 }, { "epoch": 0.08, "learning_rate": 1.4182118534030931e-05, "loss": 0.9479, "step": 5736 }, { "epoch": 0.08, "learning_rate": 1.4180160251065127e-05, "loss": 0.9757, "step": 5737 }, { "epoch": 0.08, "learning_rate": 1.417820177383424e-05, "loss": 0.9192, "step": 5738 }, { "epoch": 0.08, "learning_rate": 1.417624310242929e-05, "loss": 0.9646, "step": 5739 }, { "epoch": 0.08, "learning_rate": 1.4174284236941301e-05, "loss": 0.9483, "step": 5740 }, { "epoch": 0.08, "learning_rate": 1.4172325177461305e-05, "loss": 0.9653, "step": 5741 }, { "epoch": 0.08, "learning_rate": 1.4170365924080347e-05, "loss": 0.997, "step": 5742 }, { "epoch": 0.08, "learning_rate": 1.4168406476889484e-05, "loss": 0.9398, "step": 5743 }, { "epoch": 0.08, "learning_rate": 1.4166446835979772e-05, "loss": 0.9308, "step": 5744 }, { "epoch": 0.08, "learning_rate": 1.4164487001442283e-05, "loss": 0.9899, "step": 5745 }, { "epoch": 0.08, "learning_rate": 1.4162526973368096e-05, "loss": 0.9435, "step": 5746 }, { "epoch": 0.08, "learning_rate": 1.41605667518483e-05, "loss": 0.9376, "step": 5747 }, { "epoch": 0.08, "learning_rate": 1.4158606336973996e-05, "loss": 0.9347, "step": 5748 }, { "epoch": 0.08, "learning_rate": 1.4156645728836287e-05, "loss": 1.0191, "step": 5749 }, { "epoch": 0.08, "learning_rate": 1.4154684927526291e-05, "loss": 0.9551, "step": 5750 }, { "epoch": 0.08, "learning_rate": 1.415272393313513e-05, "loss": 1.0018, "step": 5751 }, { "epoch": 0.08, "learning_rate": 1.4150762745753937e-05, "loss": 0.8792, "step": 5752 }, { "epoch": 0.08, "learning_rate": 1.4148801365473856e-05, "loss": 0.9982, "step": 5753 }, { "epoch": 0.08, "learning_rate": 1.4146839792386037e-05, "loss": 0.9526, "step": 5754 }, { "epoch": 0.08, "learning_rate": 1.4144878026581648e-05, "loss": 0.9331, "step": 5755 }, { "epoch": 0.08, "learning_rate": 1.4142916068151846e-05, "loss": 0.9527, "step": 5756 }, { "epoch": 0.08, "learning_rate": 1.4140953917187816e-05, "loss": 0.9706, "step": 5757 }, { "epoch": 0.08, "learning_rate": 1.4138991573780749e-05, "loss": 0.9281, "step": 5758 }, { "epoch": 0.08, "learning_rate": 1.4137029038021835e-05, "loss": 0.9824, "step": 5759 }, { "epoch": 0.08, "learning_rate": 1.4135066310002281e-05, "loss": 0.9925, "step": 5760 }, { "epoch": 0.08, "learning_rate": 1.4133103389813302e-05, "loss": 0.9518, "step": 5761 }, { "epoch": 0.08, "learning_rate": 1.4131140277546119e-05, "loss": 0.9832, "step": 5762 }, { "epoch": 0.08, "learning_rate": 1.4129176973291964e-05, "loss": 0.9674, "step": 5763 }, { "epoch": 0.08, "learning_rate": 1.4127213477142084e-05, "loss": 0.8999, "step": 5764 }, { "epoch": 0.08, "learning_rate": 1.4125249789187718e-05, "loss": 0.9726, "step": 5765 }, { "epoch": 0.08, "learning_rate": 1.4123285909520136e-05, "loss": 0.9503, "step": 5766 }, { "epoch": 0.08, "learning_rate": 1.41213218382306e-05, "loss": 0.9998, "step": 5767 }, { "epoch": 0.08, "learning_rate": 1.4119357575410382e-05, "loss": 0.9403, "step": 5768 }, { "epoch": 0.08, "learning_rate": 1.4117393121150776e-05, "loss": 0.9789, "step": 5769 }, { "epoch": 0.08, "learning_rate": 1.4115428475543072e-05, "loss": 0.9576, "step": 5770 }, { "epoch": 0.08, "learning_rate": 1.4113463638678573e-05, "loss": 0.9579, "step": 5771 }, { "epoch": 0.08, "learning_rate": 1.4111498610648591e-05, "loss": 0.9688, "step": 5772 }, { "epoch": 0.08, "learning_rate": 1.4109533391544452e-05, "loss": 0.9656, "step": 5773 }, { "epoch": 0.08, "learning_rate": 1.4107567981457477e-05, "loss": 0.8822, "step": 5774 }, { "epoch": 0.08, "learning_rate": 1.4105602380479009e-05, "loss": 0.9533, "step": 5775 }, { "epoch": 0.08, "learning_rate": 1.41036365887004e-05, "loss": 0.9498, "step": 5776 }, { "epoch": 0.08, "learning_rate": 1.4101670606213e-05, "loss": 0.9492, "step": 5777 }, { "epoch": 0.08, "learning_rate": 1.4099704433108177e-05, "loss": 0.9724, "step": 5778 }, { "epoch": 0.08, "learning_rate": 1.4097738069477304e-05, "loss": 0.9738, "step": 5779 }, { "epoch": 0.08, "learning_rate": 1.4095771515411766e-05, "loss": 0.9375, "step": 5780 }, { "epoch": 0.08, "learning_rate": 1.4093804771002955e-05, "loss": 0.9508, "step": 5781 }, { "epoch": 0.08, "learning_rate": 1.409183783634227e-05, "loss": 0.9948, "step": 5782 }, { "epoch": 0.08, "learning_rate": 1.4089870711521121e-05, "loss": 0.96, "step": 5783 }, { "epoch": 0.08, "learning_rate": 1.4087903396630929e-05, "loss": 0.9742, "step": 5784 }, { "epoch": 0.08, "learning_rate": 1.4085935891763117e-05, "loss": 1.0077, "step": 5785 }, { "epoch": 0.08, "learning_rate": 1.4083968197009122e-05, "loss": 0.9232, "step": 5786 }, { "epoch": 0.08, "learning_rate": 1.4082000312460392e-05, "loss": 1.0035, "step": 5787 }, { "epoch": 0.08, "learning_rate": 1.4080032238208374e-05, "loss": 0.9585, "step": 5788 }, { "epoch": 0.08, "learning_rate": 1.4078063974344542e-05, "loss": 0.9903, "step": 5789 }, { "epoch": 0.09, "learning_rate": 1.4076095520960357e-05, "loss": 0.9488, "step": 5790 }, { "epoch": 0.09, "learning_rate": 1.4074126878147302e-05, "loss": 0.9621, "step": 5791 }, { "epoch": 0.09, "learning_rate": 1.407215804599687e-05, "loss": 0.9354, "step": 5792 }, { "epoch": 0.09, "learning_rate": 1.4070189024600554e-05, "loss": 0.9789, "step": 5793 }, { "epoch": 0.09, "learning_rate": 1.406821981404986e-05, "loss": 0.957, "step": 5794 }, { "epoch": 0.09, "learning_rate": 1.4066250414436309e-05, "loss": 0.9817, "step": 5795 }, { "epoch": 0.09, "learning_rate": 1.406428082585142e-05, "loss": 0.9472, "step": 5796 }, { "epoch": 0.09, "learning_rate": 1.4062311048386726e-05, "loss": 0.9628, "step": 5797 }, { "epoch": 0.09, "learning_rate": 1.406034108213377e-05, "loss": 0.9691, "step": 5798 }, { "epoch": 0.09, "learning_rate": 1.4058370927184106e-05, "loss": 0.9922, "step": 5799 }, { "epoch": 0.09, "learning_rate": 1.4056400583629285e-05, "loss": 0.9739, "step": 5800 }, { "epoch": 0.09, "learning_rate": 1.4054430051560885e-05, "loss": 0.9746, "step": 5801 }, { "epoch": 0.09, "learning_rate": 1.4052459331070475e-05, "loss": 0.9884, "step": 5802 }, { "epoch": 0.09, "learning_rate": 1.4050488422249644e-05, "loss": 0.9947, "step": 5803 }, { "epoch": 0.09, "learning_rate": 1.4048517325189985e-05, "loss": 0.9384, "step": 5804 }, { "epoch": 0.09, "learning_rate": 1.4046546039983104e-05, "loss": 0.9994, "step": 5805 }, { "epoch": 0.09, "learning_rate": 1.4044574566720605e-05, "loss": 0.9436, "step": 5806 }, { "epoch": 0.09, "learning_rate": 1.404260290549412e-05, "loss": 0.9668, "step": 5807 }, { "epoch": 0.09, "learning_rate": 1.4040631056395268e-05, "loss": 0.9531, "step": 5808 }, { "epoch": 0.09, "learning_rate": 1.403865901951569e-05, "loss": 0.9811, "step": 5809 }, { "epoch": 0.09, "learning_rate": 1.4036686794947039e-05, "loss": 0.9058, "step": 5810 }, { "epoch": 0.09, "learning_rate": 1.4034714382780958e-05, "loss": 0.9488, "step": 5811 }, { "epoch": 0.09, "learning_rate": 1.4032741783109125e-05, "loss": 0.9487, "step": 5812 }, { "epoch": 0.09, "learning_rate": 1.4030768996023204e-05, "loss": 0.9514, "step": 5813 }, { "epoch": 0.09, "learning_rate": 1.4028796021614877e-05, "loss": 0.9376, "step": 5814 }, { "epoch": 0.09, "learning_rate": 1.4026822859975837e-05, "loss": 0.9542, "step": 5815 }, { "epoch": 0.09, "learning_rate": 1.4024849511197783e-05, "loss": 0.9892, "step": 5816 }, { "epoch": 0.09, "learning_rate": 1.4022875975372419e-05, "loss": 0.9471, "step": 5817 }, { "epoch": 0.09, "learning_rate": 1.4020902252591464e-05, "loss": 0.944, "step": 5818 }, { "epoch": 0.09, "learning_rate": 1.401892834294665e-05, "loss": 0.9697, "step": 5819 }, { "epoch": 0.09, "learning_rate": 1.4016954246529697e-05, "loss": 0.9291, "step": 5820 }, { "epoch": 0.09, "learning_rate": 1.4014979963432355e-05, "loss": 0.9974, "step": 5821 }, { "epoch": 0.09, "learning_rate": 1.4013005493746374e-05, "loss": 0.9859, "step": 5822 }, { "epoch": 0.09, "learning_rate": 1.4011030837563518e-05, "loss": 0.9326, "step": 5823 }, { "epoch": 0.09, "learning_rate": 1.4009055994975547e-05, "loss": 0.9786, "step": 5824 }, { "epoch": 0.09, "learning_rate": 1.4007080966074245e-05, "loss": 0.9921, "step": 5825 }, { "epoch": 0.09, "learning_rate": 1.4005105750951394e-05, "loss": 0.9784, "step": 5826 }, { "epoch": 0.09, "learning_rate": 1.4003130349698791e-05, "loss": 0.9522, "step": 5827 }, { "epoch": 0.09, "learning_rate": 1.4001154762408238e-05, "loss": 0.9489, "step": 5828 }, { "epoch": 0.09, "learning_rate": 1.3999178989171545e-05, "loss": 1.0015, "step": 5829 }, { "epoch": 0.09, "learning_rate": 1.3997203030080536e-05, "loss": 0.9487, "step": 5830 }, { "epoch": 0.09, "learning_rate": 1.3995226885227036e-05, "loss": 0.9946, "step": 5831 }, { "epoch": 0.09, "learning_rate": 1.3993250554702882e-05, "loss": 0.9772, "step": 5832 }, { "epoch": 0.09, "learning_rate": 1.3991274038599927e-05, "loss": 0.9528, "step": 5833 }, { "epoch": 0.09, "learning_rate": 1.3989297337010019e-05, "loss": 0.9832, "step": 5834 }, { "epoch": 0.09, "learning_rate": 1.3987320450025026e-05, "loss": 0.9266, "step": 5835 }, { "epoch": 0.09, "learning_rate": 1.3985343377736817e-05, "loss": 0.9504, "step": 5836 }, { "epoch": 0.09, "learning_rate": 1.3983366120237273e-05, "loss": 1.0043, "step": 5837 }, { "epoch": 0.09, "learning_rate": 1.3981388677618286e-05, "loss": 0.9533, "step": 5838 }, { "epoch": 0.09, "learning_rate": 1.3979411049971754e-05, "loss": 0.9151, "step": 5839 }, { "epoch": 0.09, "learning_rate": 1.3977433237389577e-05, "loss": 0.9238, "step": 5840 }, { "epoch": 0.09, "learning_rate": 1.3975455239963676e-05, "loss": 0.992, "step": 5841 }, { "epoch": 0.09, "learning_rate": 1.3973477057785977e-05, "loss": 0.9326, "step": 5842 }, { "epoch": 0.09, "learning_rate": 1.3971498690948403e-05, "loss": 1.0034, "step": 5843 }, { "epoch": 0.09, "learning_rate": 1.3969520139542905e-05, "loss": 0.9865, "step": 5844 }, { "epoch": 0.09, "learning_rate": 1.3967541403661429e-05, "loss": 1.0135, "step": 5845 }, { "epoch": 0.09, "learning_rate": 1.3965562483395926e-05, "loss": 0.9414, "step": 5846 }, { "epoch": 0.09, "learning_rate": 1.3963583378838374e-05, "loss": 1.0143, "step": 5847 }, { "epoch": 0.09, "learning_rate": 1.3961604090080742e-05, "loss": 0.9674, "step": 5848 }, { "epoch": 0.09, "learning_rate": 1.3959624617215013e-05, "loss": 0.9533, "step": 5849 }, { "epoch": 0.09, "learning_rate": 1.3957644960333183e-05, "loss": 0.9997, "step": 5850 }, { "epoch": 0.09, "learning_rate": 1.395566511952725e-05, "loss": 1.0184, "step": 5851 }, { "epoch": 0.09, "learning_rate": 1.3953685094889222e-05, "loss": 0.9652, "step": 5852 }, { "epoch": 0.09, "learning_rate": 1.3951704886511123e-05, "loss": 0.9444, "step": 5853 }, { "epoch": 0.09, "learning_rate": 1.3949724494484974e-05, "loss": 0.9715, "step": 5854 }, { "epoch": 0.09, "learning_rate": 1.3947743918902808e-05, "loss": 0.9679, "step": 5855 }, { "epoch": 0.09, "learning_rate": 1.3945763159856674e-05, "loss": 0.9676, "step": 5856 }, { "epoch": 0.09, "learning_rate": 1.3943782217438622e-05, "loss": 0.9428, "step": 5857 }, { "epoch": 0.09, "learning_rate": 1.3941801091740714e-05, "loss": 1.0001, "step": 5858 }, { "epoch": 0.09, "learning_rate": 1.3939819782855018e-05, "loss": 1.0061, "step": 5859 }, { "epoch": 0.09, "learning_rate": 1.3937838290873609e-05, "loss": 0.9807, "step": 5860 }, { "epoch": 0.09, "learning_rate": 1.3935856615888574e-05, "loss": 0.9924, "step": 5861 }, { "epoch": 0.09, "learning_rate": 1.3933874757992015e-05, "loss": 0.8974, "step": 5862 }, { "epoch": 0.09, "learning_rate": 1.3931892717276024e-05, "loss": 0.9521, "step": 5863 }, { "epoch": 0.09, "learning_rate": 1.3929910493832716e-05, "loss": 0.9303, "step": 5864 }, { "epoch": 0.09, "learning_rate": 1.3927928087754218e-05, "loss": 0.9601, "step": 5865 }, { "epoch": 0.09, "learning_rate": 1.3925945499132653e-05, "loss": 0.9442, "step": 5866 }, { "epoch": 0.09, "learning_rate": 1.3923962728060157e-05, "loss": 0.9951, "step": 5867 }, { "epoch": 0.09, "learning_rate": 1.3921979774628878e-05, "loss": 0.9427, "step": 5868 }, { "epoch": 0.09, "learning_rate": 1.3919996638930967e-05, "loss": 1.0003, "step": 5869 }, { "epoch": 0.09, "learning_rate": 1.3918013321058587e-05, "loss": 0.9401, "step": 5870 }, { "epoch": 0.09, "learning_rate": 1.3916029821103916e-05, "loss": 0.9614, "step": 5871 }, { "epoch": 0.09, "learning_rate": 1.3914046139159123e-05, "loss": 0.961, "step": 5872 }, { "epoch": 0.09, "learning_rate": 1.3912062275316406e-05, "loss": 0.9877, "step": 5873 }, { "epoch": 0.09, "learning_rate": 1.3910078229667953e-05, "loss": 0.9916, "step": 5874 }, { "epoch": 0.09, "learning_rate": 1.3908094002305971e-05, "loss": 1.0059, "step": 5875 }, { "epoch": 0.09, "learning_rate": 1.3906109593322676e-05, "loss": 0.9221, "step": 5876 }, { "epoch": 0.09, "learning_rate": 1.3904125002810287e-05, "loss": 0.989, "step": 5877 }, { "epoch": 0.09, "learning_rate": 1.3902140230861033e-05, "loss": 0.9908, "step": 5878 }, { "epoch": 0.09, "learning_rate": 1.3900155277567157e-05, "loss": 0.994, "step": 5879 }, { "epoch": 0.09, "learning_rate": 1.3898170143020903e-05, "loss": 0.968, "step": 5880 }, { "epoch": 0.09, "learning_rate": 1.3896184827314529e-05, "loss": 0.9378, "step": 5881 }, { "epoch": 0.09, "learning_rate": 1.3894199330540297e-05, "loss": 0.9492, "step": 5882 }, { "epoch": 0.09, "learning_rate": 1.3892213652790475e-05, "loss": 0.9612, "step": 5883 }, { "epoch": 0.09, "learning_rate": 1.389022779415735e-05, "loss": 0.975, "step": 5884 }, { "epoch": 0.09, "learning_rate": 1.388824175473321e-05, "loss": 0.9193, "step": 5885 }, { "epoch": 0.09, "learning_rate": 1.3886255534610347e-05, "loss": 0.9905, "step": 5886 }, { "epoch": 0.09, "learning_rate": 1.3884269133881073e-05, "loss": 0.9694, "step": 5887 }, { "epoch": 0.09, "learning_rate": 1.38822825526377e-05, "loss": 0.9491, "step": 5888 }, { "epoch": 0.09, "learning_rate": 1.388029579097255e-05, "loss": 1.0108, "step": 5889 }, { "epoch": 0.09, "learning_rate": 1.3878308848977958e-05, "loss": 0.9821, "step": 5890 }, { "epoch": 0.09, "learning_rate": 1.3876321726746258e-05, "loss": 0.9834, "step": 5891 }, { "epoch": 0.09, "learning_rate": 1.38743344243698e-05, "loss": 0.9543, "step": 5892 }, { "epoch": 0.09, "learning_rate": 1.3872346941940938e-05, "loss": 1.0055, "step": 5893 }, { "epoch": 0.09, "learning_rate": 1.3870359279552041e-05, "loss": 0.97, "step": 5894 }, { "epoch": 0.09, "learning_rate": 1.3868371437295479e-05, "loss": 0.9898, "step": 5895 }, { "epoch": 0.09, "learning_rate": 1.3866383415263635e-05, "loss": 0.9451, "step": 5896 }, { "epoch": 0.09, "learning_rate": 1.3864395213548899e-05, "loss": 1.0032, "step": 5897 }, { "epoch": 0.09, "learning_rate": 1.3862406832243662e-05, "loss": 0.9769, "step": 5898 }, { "epoch": 0.09, "learning_rate": 1.3860418271440339e-05, "loss": 0.9789, "step": 5899 }, { "epoch": 0.09, "learning_rate": 1.3858429531231338e-05, "loss": 0.9782, "step": 5900 }, { "epoch": 0.09, "learning_rate": 1.3856440611709086e-05, "loss": 0.9763, "step": 5901 }, { "epoch": 0.09, "learning_rate": 1.3854451512966016e-05, "loss": 0.9647, "step": 5902 }, { "epoch": 0.09, "learning_rate": 1.3852462235094561e-05, "loss": 0.987, "step": 5903 }, { "epoch": 0.09, "learning_rate": 1.3850472778187176e-05, "loss": 0.9871, "step": 5904 }, { "epoch": 0.09, "learning_rate": 1.3848483142336316e-05, "loss": 0.9683, "step": 5905 }, { "epoch": 0.09, "learning_rate": 1.384649332763444e-05, "loss": 0.9657, "step": 5906 }, { "epoch": 0.09, "learning_rate": 1.3844503334174024e-05, "loss": 0.9606, "step": 5907 }, { "epoch": 0.09, "learning_rate": 1.3842513162047553e-05, "loss": 0.9749, "step": 5908 }, { "epoch": 0.09, "learning_rate": 1.384052281134751e-05, "loss": 0.9867, "step": 5909 }, { "epoch": 0.09, "learning_rate": 1.3838532282166398e-05, "loss": 0.942, "step": 5910 }, { "epoch": 0.09, "learning_rate": 1.3836541574596722e-05, "loss": 0.9774, "step": 5911 }, { "epoch": 0.09, "learning_rate": 1.3834550688730997e-05, "loss": 0.9942, "step": 5912 }, { "epoch": 0.09, "learning_rate": 1.3832559624661742e-05, "loss": 1.0273, "step": 5913 }, { "epoch": 0.09, "learning_rate": 1.383056838248149e-05, "loss": 0.9153, "step": 5914 }, { "epoch": 0.09, "learning_rate": 1.382857696228278e-05, "loss": 1.0081, "step": 5915 }, { "epoch": 0.09, "learning_rate": 1.3826585364158161e-05, "loss": 0.9732, "step": 5916 }, { "epoch": 0.09, "learning_rate": 1.3824593588200191e-05, "loss": 0.9741, "step": 5917 }, { "epoch": 0.09, "learning_rate": 1.3822601634501425e-05, "loss": 0.9776, "step": 5918 }, { "epoch": 0.09, "learning_rate": 1.3820609503154446e-05, "loss": 0.9863, "step": 5919 }, { "epoch": 0.09, "learning_rate": 1.381861719425183e-05, "loss": 0.9797, "step": 5920 }, { "epoch": 0.09, "learning_rate": 1.3816624707886159e-05, "loss": 0.9354, "step": 5921 }, { "epoch": 0.09, "learning_rate": 1.381463204415004e-05, "loss": 0.9572, "step": 5922 }, { "epoch": 0.09, "learning_rate": 1.3812639203136077e-05, "loss": 0.9965, "step": 5923 }, { "epoch": 0.09, "learning_rate": 1.381064618493688e-05, "loss": 0.9797, "step": 5924 }, { "epoch": 0.09, "learning_rate": 1.3808652989645073e-05, "loss": 0.9712, "step": 5925 }, { "epoch": 0.09, "learning_rate": 1.3806659617353285e-05, "loss": 0.9339, "step": 5926 }, { "epoch": 0.09, "learning_rate": 1.3804666068154154e-05, "loss": 0.9833, "step": 5927 }, { "epoch": 0.09, "learning_rate": 1.3802672342140327e-05, "loss": 1.0079, "step": 5928 }, { "epoch": 0.09, "learning_rate": 1.3800678439404455e-05, "loss": 0.9975, "step": 5929 }, { "epoch": 0.09, "learning_rate": 1.3798684360039208e-05, "loss": 0.927, "step": 5930 }, { "epoch": 0.09, "learning_rate": 1.3796690104137253e-05, "loss": 0.9773, "step": 5931 }, { "epoch": 0.09, "learning_rate": 1.379469567179127e-05, "loss": 0.9361, "step": 5932 }, { "epoch": 0.09, "learning_rate": 1.3792701063093945e-05, "loss": 0.9991, "step": 5933 }, { "epoch": 0.09, "learning_rate": 1.3790706278137977e-05, "loss": 0.9501, "step": 5934 }, { "epoch": 0.09, "learning_rate": 1.3788711317016067e-05, "loss": 0.9884, "step": 5935 }, { "epoch": 0.09, "learning_rate": 1.3786716179820927e-05, "loss": 0.9343, "step": 5936 }, { "epoch": 0.09, "learning_rate": 1.378472086664528e-05, "loss": 0.9553, "step": 5937 }, { "epoch": 0.09, "learning_rate": 1.378272537758185e-05, "loss": 0.9685, "step": 5938 }, { "epoch": 0.09, "learning_rate": 1.3780729712723377e-05, "loss": 0.9681, "step": 5939 }, { "epoch": 0.1, "learning_rate": 1.3778733872162607e-05, "loss": 1.0135, "step": 5940 }, { "epoch": 0.1, "learning_rate": 1.3776737855992287e-05, "loss": 0.9925, "step": 5941 }, { "epoch": 0.1, "learning_rate": 1.3774741664305185e-05, "loss": 0.9475, "step": 5942 }, { "epoch": 0.1, "learning_rate": 1.3772745297194069e-05, "loss": 0.9922, "step": 5943 }, { "epoch": 0.1, "learning_rate": 1.3770748754751708e-05, "loss": 1.007, "step": 5944 }, { "epoch": 0.1, "learning_rate": 1.3768752037070898e-05, "loss": 0.9765, "step": 5945 }, { "epoch": 0.1, "learning_rate": 1.376675514424443e-05, "loss": 0.9452, "step": 5946 }, { "epoch": 0.1, "learning_rate": 1.3764758076365103e-05, "loss": 0.9411, "step": 5947 }, { "epoch": 0.1, "learning_rate": 1.376276083352573e-05, "loss": 0.9985, "step": 5948 }, { "epoch": 0.1, "learning_rate": 1.3760763415819125e-05, "loss": 0.9306, "step": 5949 }, { "epoch": 0.1, "learning_rate": 1.375876582333812e-05, "loss": 0.9783, "step": 5950 }, { "epoch": 0.1, "learning_rate": 1.3756768056175545e-05, "loss": 0.9872, "step": 5951 }, { "epoch": 0.1, "learning_rate": 1.3754770114424245e-05, "loss": 1.0033, "step": 5952 }, { "epoch": 0.1, "learning_rate": 1.3752771998177068e-05, "loss": 0.9522, "step": 5953 }, { "epoch": 0.1, "learning_rate": 1.3750773707526875e-05, "loss": 0.9895, "step": 5954 }, { "epoch": 0.1, "learning_rate": 1.374877524256653e-05, "loss": 0.9376, "step": 5955 }, { "epoch": 0.1, "learning_rate": 1.3746776603388912e-05, "loss": 0.9764, "step": 5956 }, { "epoch": 0.1, "learning_rate": 1.3744777790086903e-05, "loss": 0.9833, "step": 5957 }, { "epoch": 0.1, "learning_rate": 1.3742778802753392e-05, "loss": 0.9871, "step": 5958 }, { "epoch": 0.1, "learning_rate": 1.3740779641481275e-05, "loss": 0.9757, "step": 5959 }, { "epoch": 0.1, "learning_rate": 1.3738780306363469e-05, "loss": 1.0062, "step": 5960 }, { "epoch": 0.1, "learning_rate": 1.373678079749288e-05, "loss": 0.979, "step": 5961 }, { "epoch": 0.1, "learning_rate": 1.3734781114962434e-05, "loss": 0.9517, "step": 5962 }, { "epoch": 0.1, "learning_rate": 1.3732781258865066e-05, "loss": 0.9881, "step": 5963 }, { "epoch": 0.1, "learning_rate": 1.3730781229293713e-05, "loss": 0.9545, "step": 5964 }, { "epoch": 0.1, "learning_rate": 1.3728781026341323e-05, "loss": 0.9925, "step": 5965 }, { "epoch": 0.1, "learning_rate": 1.372678065010085e-05, "loss": 0.9557, "step": 5966 }, { "epoch": 0.1, "learning_rate": 1.3724780100665261e-05, "loss": 0.9487, "step": 5967 }, { "epoch": 0.1, "learning_rate": 1.3722779378127521e-05, "loss": 0.9984, "step": 5968 }, { "epoch": 0.1, "learning_rate": 1.372077848258062e-05, "loss": 0.9511, "step": 5969 }, { "epoch": 0.1, "learning_rate": 1.3718777414117539e-05, "loss": 0.9954, "step": 5970 }, { "epoch": 0.1, "learning_rate": 1.3716776172831274e-05, "loss": 1.023, "step": 5971 }, { "epoch": 0.1, "learning_rate": 1.3714774758814834e-05, "loss": 0.9663, "step": 5972 }, { "epoch": 0.1, "learning_rate": 1.3712773172161225e-05, "loss": 0.9943, "step": 5973 }, { "epoch": 0.1, "learning_rate": 1.3710771412963466e-05, "loss": 0.9771, "step": 5974 }, { "epoch": 0.1, "learning_rate": 1.3708769481314592e-05, "loss": 0.9405, "step": 5975 }, { "epoch": 0.1, "learning_rate": 1.3706767377307635e-05, "loss": 0.9797, "step": 5976 }, { "epoch": 0.1, "learning_rate": 1.3704765101035639e-05, "loss": 0.9988, "step": 5977 }, { "epoch": 0.1, "learning_rate": 1.3702762652591655e-05, "loss": 0.9403, "step": 5978 }, { "epoch": 0.1, "learning_rate": 1.3700760032068745e-05, "loss": 0.9838, "step": 5979 }, { "epoch": 0.1, "learning_rate": 1.3698757239559978e-05, "loss": 1.0006, "step": 5980 }, { "epoch": 0.1, "learning_rate": 1.3696754275158427e-05, "loss": 0.9405, "step": 5981 }, { "epoch": 0.1, "learning_rate": 1.3694751138957179e-05, "loss": 0.9279, "step": 5982 }, { "epoch": 0.1, "learning_rate": 1.3692747831049322e-05, "loss": 0.9682, "step": 5983 }, { "epoch": 0.1, "learning_rate": 1.3690744351527958e-05, "loss": 0.962, "step": 5984 }, { "epoch": 0.1, "learning_rate": 1.3688740700486196e-05, "loss": 0.9835, "step": 5985 }, { "epoch": 0.1, "learning_rate": 1.3686736878017152e-05, "loss": 1.0121, "step": 5986 }, { "epoch": 0.1, "learning_rate": 1.3684732884213947e-05, "loss": 0.9841, "step": 5987 }, { "epoch": 0.1, "learning_rate": 1.3682728719169717e-05, "loss": 0.9707, "step": 5988 }, { "epoch": 0.1, "learning_rate": 1.3680724382977602e-05, "loss": 0.9795, "step": 5989 }, { "epoch": 0.1, "learning_rate": 1.3678719875730744e-05, "loss": 0.9908, "step": 5990 }, { "epoch": 0.1, "learning_rate": 1.3676715197522302e-05, "loss": 0.987, "step": 5991 }, { "epoch": 0.1, "learning_rate": 1.3674710348445441e-05, "loss": 0.9913, "step": 5992 }, { "epoch": 0.1, "learning_rate": 1.3672705328593329e-05, "loss": 0.9748, "step": 5993 }, { "epoch": 0.1, "learning_rate": 1.3670700138059151e-05, "loss": 0.9629, "step": 5994 }, { "epoch": 0.1, "learning_rate": 1.366869477693609e-05, "loss": 1.0178, "step": 5995 }, { "epoch": 0.1, "learning_rate": 1.3666689245317344e-05, "loss": 0.9564, "step": 5996 }, { "epoch": 0.1, "learning_rate": 1.3664683543296114e-05, "loss": 0.9961, "step": 5997 }, { "epoch": 0.1, "learning_rate": 1.3662677670965612e-05, "loss": 0.9699, "step": 5998 }, { "epoch": 0.1, "learning_rate": 1.3660671628419055e-05, "loss": 0.9929, "step": 5999 }, { "epoch": 0.1, "learning_rate": 1.3658665415749677e-05, "loss": 0.9665, "step": 6000 }, { "epoch": 0.1, "learning_rate": 1.3656659033050706e-05, "loss": 0.9617, "step": 6001 }, { "epoch": 0.1, "learning_rate": 1.3654652480415387e-05, "loss": 0.9841, "step": 6002 }, { "epoch": 0.1, "learning_rate": 1.3652645757936974e-05, "loss": 0.9694, "step": 6003 }, { "epoch": 0.1, "learning_rate": 1.3650638865708721e-05, "loss": 0.9885, "step": 6004 }, { "epoch": 0.1, "learning_rate": 1.3648631803823892e-05, "loss": 0.9508, "step": 6005 }, { "epoch": 0.1, "learning_rate": 1.3646624572375773e-05, "loss": 0.9476, "step": 6006 }, { "epoch": 0.1, "learning_rate": 1.3644617171457632e-05, "loss": 1.0202, "step": 6007 }, { "epoch": 0.1, "learning_rate": 1.3642609601162771e-05, "loss": 0.948, "step": 6008 }, { "epoch": 0.1, "learning_rate": 1.3640601861584481e-05, "loss": 0.9719, "step": 6009 }, { "epoch": 0.1, "learning_rate": 1.363859395281607e-05, "loss": 0.9635, "step": 6010 }, { "epoch": 0.1, "learning_rate": 1.3636585874950853e-05, "loss": 0.9493, "step": 6011 }, { "epoch": 0.1, "learning_rate": 1.3634577628082152e-05, "loss": 0.9359, "step": 6012 }, { "epoch": 0.1, "learning_rate": 1.3632569212303293e-05, "loss": 0.9568, "step": 6013 }, { "epoch": 0.1, "learning_rate": 1.3630560627707615e-05, "loss": 0.9325, "step": 6014 }, { "epoch": 0.1, "learning_rate": 1.3628551874388465e-05, "loss": 0.965, "step": 6015 }, { "epoch": 0.1, "learning_rate": 1.3626542952439192e-05, "loss": 0.9802, "step": 6016 }, { "epoch": 0.1, "learning_rate": 1.3624533861953162e-05, "loss": 0.9845, "step": 6017 }, { "epoch": 0.1, "learning_rate": 1.3622524603023739e-05, "loss": 0.9768, "step": 6018 }, { "epoch": 0.1, "learning_rate": 1.3620515175744307e-05, "loss": 0.9878, "step": 6019 }, { "epoch": 0.1, "learning_rate": 1.3618505580208238e-05, "loss": 0.995, "step": 6020 }, { "epoch": 0.1, "learning_rate": 1.3616495816508936e-05, "loss": 0.9793, "step": 6021 }, { "epoch": 0.1, "learning_rate": 1.3614485884739795e-05, "loss": 0.9682, "step": 6022 }, { "epoch": 0.1, "learning_rate": 1.3612475784994223e-05, "loss": 1.0225, "step": 6023 }, { "epoch": 0.1, "learning_rate": 1.3610465517365636e-05, "loss": 0.9514, "step": 6024 }, { "epoch": 0.1, "learning_rate": 1.3608455081947458e-05, "loss": 0.9535, "step": 6025 }, { "epoch": 0.1, "learning_rate": 1.3606444478833123e-05, "loss": 0.92, "step": 6026 }, { "epoch": 0.1, "learning_rate": 1.3604433708116068e-05, "loss": 0.962, "step": 6027 }, { "epoch": 0.1, "learning_rate": 1.3602422769889734e-05, "loss": 0.9337, "step": 6028 }, { "epoch": 0.1, "learning_rate": 1.3600411664247584e-05, "loss": 0.9445, "step": 6029 }, { "epoch": 0.1, "learning_rate": 1.3598400391283075e-05, "loss": 0.9824, "step": 6030 }, { "epoch": 0.1, "learning_rate": 1.3596388951089679e-05, "loss": 0.9692, "step": 6031 }, { "epoch": 0.1, "learning_rate": 1.3594377343760876e-05, "loss": 0.9745, "step": 6032 }, { "epoch": 0.1, "learning_rate": 1.3592365569390146e-05, "loss": 0.996, "step": 6033 }, { "epoch": 0.1, "learning_rate": 1.3590353628070991e-05, "loss": 0.9302, "step": 6034 }, { "epoch": 0.1, "learning_rate": 1.3588341519896906e-05, "loss": 0.9822, "step": 6035 }, { "epoch": 0.1, "learning_rate": 1.3586329244961398e-05, "loss": 0.9574, "step": 6036 }, { "epoch": 0.1, "learning_rate": 1.3584316803357989e-05, "loss": 0.9827, "step": 6037 }, { "epoch": 0.1, "learning_rate": 1.3582304195180202e-05, "loss": 0.9757, "step": 6038 }, { "epoch": 0.1, "learning_rate": 1.3580291420521565e-05, "loss": 0.9591, "step": 6039 }, { "epoch": 0.1, "learning_rate": 1.3578278479475626e-05, "loss": 0.9614, "step": 6040 }, { "epoch": 0.1, "learning_rate": 1.3576265372135926e-05, "loss": 0.9771, "step": 6041 }, { "epoch": 0.1, "learning_rate": 1.3574252098596024e-05, "loss": 0.9501, "step": 6042 }, { "epoch": 0.1, "learning_rate": 1.3572238658949477e-05, "loss": 0.9925, "step": 6043 }, { "epoch": 0.1, "learning_rate": 1.3570225053289864e-05, "loss": 0.9578, "step": 6044 }, { "epoch": 0.1, "learning_rate": 1.356821128171076e-05, "loss": 0.9652, "step": 6045 }, { "epoch": 0.1, "learning_rate": 1.356619734430575e-05, "loss": 0.964, "step": 6046 }, { "epoch": 0.1, "learning_rate": 1.3564183241168427e-05, "loss": 0.9865, "step": 6047 }, { "epoch": 0.1, "learning_rate": 1.3562168972392398e-05, "loss": 0.962, "step": 6048 }, { "epoch": 0.1, "learning_rate": 1.3560154538071268e-05, "loss": 0.9841, "step": 6049 }, { "epoch": 0.1, "learning_rate": 1.3558139938298655e-05, "loss": 0.952, "step": 6050 }, { "epoch": 0.1, "learning_rate": 1.3556125173168182e-05, "loss": 1.0386, "step": 6051 }, { "epoch": 0.1, "learning_rate": 1.3554110242773489e-05, "loss": 0.9671, "step": 6052 }, { "epoch": 0.1, "learning_rate": 1.3552095147208204e-05, "loss": 0.9496, "step": 6053 }, { "epoch": 0.1, "learning_rate": 1.355007988656598e-05, "loss": 0.942, "step": 6054 }, { "epoch": 0.1, "learning_rate": 1.354806446094048e-05, "loss": 0.9789, "step": 6055 }, { "epoch": 0.1, "learning_rate": 1.3546048870425356e-05, "loss": 0.9539, "step": 6056 }, { "epoch": 0.1, "learning_rate": 1.3544033115114288e-05, "loss": 0.95, "step": 6057 }, { "epoch": 0.1, "learning_rate": 1.3542017195100948e-05, "loss": 0.955, "step": 6058 }, { "epoch": 0.1, "learning_rate": 1.3540001110479023e-05, "loss": 0.9927, "step": 6059 }, { "epoch": 0.1, "learning_rate": 1.3537984861342205e-05, "loss": 0.9554, "step": 6060 }, { "epoch": 0.1, "learning_rate": 1.3535968447784203e-05, "loss": 0.9561, "step": 6061 }, { "epoch": 0.1, "learning_rate": 1.3533951869898716e-05, "loss": 0.9745, "step": 6062 }, { "epoch": 0.1, "learning_rate": 1.353193512777947e-05, "loss": 0.9491, "step": 6063 }, { "epoch": 0.1, "learning_rate": 1.3529918221520185e-05, "loss": 0.9971, "step": 6064 }, { "epoch": 0.1, "learning_rate": 1.3527901151214593e-05, "loss": 1.0172, "step": 6065 }, { "epoch": 0.1, "learning_rate": 1.352588391695643e-05, "loss": 0.9392, "step": 6066 }, { "epoch": 0.1, "learning_rate": 1.352386651883945e-05, "loss": 0.9896, "step": 6067 }, { "epoch": 0.1, "learning_rate": 1.3521848956957401e-05, "loss": 0.9608, "step": 6068 }, { "epoch": 0.1, "learning_rate": 1.3519831231404055e-05, "loss": 0.9695, "step": 6069 }, { "epoch": 0.1, "learning_rate": 1.3517813342273171e-05, "loss": 0.9685, "step": 6070 }, { "epoch": 0.1, "learning_rate": 1.351579528965853e-05, "loss": 1.0508, "step": 6071 }, { "epoch": 0.1, "learning_rate": 1.3513777073653924e-05, "loss": 0.9623, "step": 6072 }, { "epoch": 0.1, "learning_rate": 1.3511758694353141e-05, "loss": 1.0298, "step": 6073 }, { "epoch": 0.1, "learning_rate": 1.3509740151849976e-05, "loss": 0.9399, "step": 6074 }, { "epoch": 0.1, "learning_rate": 1.3507721446238244e-05, "loss": 0.9337, "step": 6075 }, { "epoch": 0.1, "learning_rate": 1.3505702577611759e-05, "loss": 0.9693, "step": 6076 }, { "epoch": 0.1, "learning_rate": 1.3503683546064341e-05, "loss": 0.939, "step": 6077 }, { "epoch": 0.1, "learning_rate": 1.3501664351689824e-05, "loss": 0.9512, "step": 6078 }, { "epoch": 0.1, "learning_rate": 1.3499644994582047e-05, "loss": 0.9783, "step": 6079 }, { "epoch": 0.1, "learning_rate": 1.3497625474834855e-05, "loss": 0.9792, "step": 6080 }, { "epoch": 0.1, "learning_rate": 1.3495605792542098e-05, "loss": 0.9737, "step": 6081 }, { "epoch": 0.1, "learning_rate": 1.349358594779764e-05, "loss": 0.945, "step": 6082 }, { "epoch": 0.1, "learning_rate": 1.349156594069535e-05, "loss": 0.9905, "step": 6083 }, { "epoch": 0.1, "learning_rate": 1.3489545771329105e-05, "loss": 0.961, "step": 6084 }, { "epoch": 0.1, "learning_rate": 1.3487525439792784e-05, "loss": 0.9729, "step": 6085 }, { "epoch": 0.1, "learning_rate": 1.3485504946180279e-05, "loss": 0.9591, "step": 6086 }, { "epoch": 0.1, "learning_rate": 1.3483484290585493e-05, "loss": 0.9863, "step": 6087 }, { "epoch": 0.1, "learning_rate": 1.3481463473102331e-05, "loss": 0.9714, "step": 6088 }, { "epoch": 0.1, "learning_rate": 1.3479442493824704e-05, "loss": 0.9612, "step": 6089 }, { "epoch": 0.11, "learning_rate": 1.3477421352846534e-05, "loss": 0.9847, "step": 6090 }, { "epoch": 0.11, "learning_rate": 1.3475400050261749e-05, "loss": 0.9795, "step": 6091 }, { "epoch": 0.11, "learning_rate": 1.347337858616429e-05, "loss": 0.9668, "step": 6092 }, { "epoch": 0.11, "learning_rate": 1.3471356960648093e-05, "loss": 0.9749, "step": 6093 }, { "epoch": 0.11, "learning_rate": 1.3469335173807114e-05, "loss": 0.9524, "step": 6094 }, { "epoch": 0.11, "learning_rate": 1.3467313225735312e-05, "loss": 1.0492, "step": 6095 }, { "epoch": 0.11, "learning_rate": 1.3465291116526655e-05, "loss": 0.9765, "step": 6096 }, { "epoch": 0.11, "learning_rate": 1.3463268846275113e-05, "loss": 0.9876, "step": 6097 }, { "epoch": 0.11, "learning_rate": 1.3461246415074667e-05, "loss": 1.0098, "step": 6098 }, { "epoch": 0.11, "learning_rate": 1.3459223823019305e-05, "loss": 0.9521, "step": 6099 }, { "epoch": 0.11, "learning_rate": 1.3457201070203027e-05, "loss": 1.0047, "step": 6100 }, { "epoch": 0.11, "learning_rate": 1.3455178156719838e-05, "loss": 0.9526, "step": 6101 }, { "epoch": 0.11, "learning_rate": 1.345315508266374e-05, "loss": 0.9667, "step": 6102 }, { "epoch": 0.11, "learning_rate": 1.3451131848128763e-05, "loss": 0.9975, "step": 6103 }, { "epoch": 0.11, "learning_rate": 1.3449108453208926e-05, "loss": 0.9753, "step": 6104 }, { "epoch": 0.11, "learning_rate": 1.3447084897998259e-05, "loss": 0.9927, "step": 6105 }, { "epoch": 0.11, "learning_rate": 1.3445061182590813e-05, "loss": 0.9582, "step": 6106 }, { "epoch": 0.11, "learning_rate": 1.344303730708063e-05, "loss": 0.9751, "step": 6107 }, { "epoch": 0.11, "learning_rate": 1.3441013271561768e-05, "loss": 1.0208, "step": 6108 }, { "epoch": 0.11, "learning_rate": 1.3438989076128287e-05, "loss": 0.9676, "step": 6109 }, { "epoch": 0.11, "learning_rate": 1.3436964720874261e-05, "loss": 0.9499, "step": 6110 }, { "epoch": 0.11, "learning_rate": 1.3434940205893767e-05, "loss": 0.9793, "step": 6111 }, { "epoch": 0.11, "learning_rate": 1.343291553128089e-05, "loss": 0.9342, "step": 6112 }, { "epoch": 0.11, "learning_rate": 1.3430890697129727e-05, "loss": 0.9672, "step": 6113 }, { "epoch": 0.11, "learning_rate": 1.342886570353437e-05, "loss": 0.9649, "step": 6114 }, { "epoch": 0.11, "learning_rate": 1.3426840550588933e-05, "loss": 0.9466, "step": 6115 }, { "epoch": 0.11, "learning_rate": 1.3424815238387531e-05, "loss": 0.9504, "step": 6116 }, { "epoch": 0.11, "learning_rate": 1.3422789767024285e-05, "loss": 0.9735, "step": 6117 }, { "epoch": 0.11, "learning_rate": 1.3420764136593324e-05, "loss": 0.9695, "step": 6118 }, { "epoch": 0.11, "learning_rate": 1.3418738347188791e-05, "loss": 0.9979, "step": 6119 }, { "epoch": 0.11, "learning_rate": 1.3416712398904821e-05, "loss": 0.9786, "step": 6120 }, { "epoch": 0.11, "learning_rate": 1.3414686291835577e-05, "loss": 0.9583, "step": 6121 }, { "epoch": 0.11, "learning_rate": 1.3412660026075208e-05, "loss": 0.9709, "step": 6122 }, { "epoch": 0.11, "learning_rate": 1.3410633601717886e-05, "loss": 0.9725, "step": 6123 }, { "epoch": 0.11, "learning_rate": 1.3408607018857789e-05, "loss": 0.9322, "step": 6124 }, { "epoch": 0.11, "learning_rate": 1.3406580277589092e-05, "loss": 0.9682, "step": 6125 }, { "epoch": 0.11, "learning_rate": 1.3404553378005988e-05, "loss": 0.9661, "step": 6126 }, { "epoch": 0.11, "learning_rate": 1.3402526320202676e-05, "loss": 0.993, "step": 6127 }, { "epoch": 0.11, "learning_rate": 1.340049910427335e-05, "loss": 0.9759, "step": 6128 }, { "epoch": 0.11, "learning_rate": 1.3398471730312228e-05, "loss": 0.9992, "step": 6129 }, { "epoch": 0.11, "learning_rate": 1.3396444198413529e-05, "loss": 0.9799, "step": 6130 }, { "epoch": 0.11, "learning_rate": 1.3394416508671478e-05, "loss": 0.9633, "step": 6131 }, { "epoch": 0.11, "learning_rate": 1.3392388661180303e-05, "loss": 0.9537, "step": 6132 }, { "epoch": 0.11, "learning_rate": 1.3390360656034252e-05, "loss": 0.9619, "step": 6133 }, { "epoch": 0.11, "learning_rate": 1.338833249332757e-05, "loss": 0.9835, "step": 6134 }, { "epoch": 0.11, "learning_rate": 1.3386304173154509e-05, "loss": 0.9907, "step": 6135 }, { "epoch": 0.11, "learning_rate": 1.3384275695609335e-05, "loss": 0.9696, "step": 6136 }, { "epoch": 0.11, "learning_rate": 1.3382247060786316e-05, "loss": 0.9925, "step": 6137 }, { "epoch": 0.11, "learning_rate": 1.3380218268779728e-05, "loss": 0.9923, "step": 6138 }, { "epoch": 0.11, "learning_rate": 1.3378189319683858e-05, "loss": 1.0135, "step": 6139 }, { "epoch": 0.11, "learning_rate": 1.3376160213592995e-05, "loss": 1.0073, "step": 6140 }, { "epoch": 0.11, "learning_rate": 1.3374130950601441e-05, "loss": 0.9301, "step": 6141 }, { "epoch": 0.11, "learning_rate": 1.3372101530803501e-05, "loss": 0.9915, "step": 6142 }, { "epoch": 0.11, "learning_rate": 1.3370071954293485e-05, "loss": 1.0018, "step": 6143 }, { "epoch": 0.11, "learning_rate": 1.3368042221165718e-05, "loss": 0.9811, "step": 6144 }, { "epoch": 0.11, "learning_rate": 1.3366012331514525e-05, "loss": 0.9292, "step": 6145 }, { "epoch": 0.11, "learning_rate": 1.3363982285434242e-05, "loss": 0.9725, "step": 6146 }, { "epoch": 0.11, "learning_rate": 1.3361952083019216e-05, "loss": 1.0236, "step": 6147 }, { "epoch": 0.11, "learning_rate": 1.335992172436379e-05, "loss": 0.9722, "step": 6148 }, { "epoch": 0.11, "learning_rate": 1.3357891209562328e-05, "loss": 0.963, "step": 6149 }, { "epoch": 0.11, "learning_rate": 1.335586053870919e-05, "loss": 0.9581, "step": 6150 }, { "epoch": 0.11, "learning_rate": 1.3353829711898744e-05, "loss": 0.9788, "step": 6151 }, { "epoch": 0.11, "learning_rate": 1.3351798729225377e-05, "loss": 1.0244, "step": 6152 }, { "epoch": 0.11, "learning_rate": 1.3349767590783472e-05, "loss": 0.9973, "step": 6153 }, { "epoch": 0.11, "learning_rate": 1.3347736296667418e-05, "loss": 0.9865, "step": 6154 }, { "epoch": 0.11, "learning_rate": 1.334570484697162e-05, "loss": 1.0216, "step": 6155 }, { "epoch": 0.11, "learning_rate": 1.3343673241790485e-05, "loss": 0.9249, "step": 6156 }, { "epoch": 0.11, "learning_rate": 1.334164148121843e-05, "loss": 0.9886, "step": 6157 }, { "epoch": 0.11, "learning_rate": 1.3339609565349874e-05, "loss": 0.9464, "step": 6158 }, { "epoch": 0.11, "learning_rate": 1.3337577494279247e-05, "loss": 0.9424, "step": 6159 }, { "epoch": 0.11, "learning_rate": 1.3335545268100987e-05, "loss": 0.9971, "step": 6160 }, { "epoch": 0.11, "learning_rate": 1.3333512886909537e-05, "loss": 0.9434, "step": 6161 }, { "epoch": 0.11, "learning_rate": 1.3331480350799349e-05, "loss": 0.9664, "step": 6162 }, { "epoch": 0.11, "learning_rate": 1.332944765986488e-05, "loss": 0.9733, "step": 6163 }, { "epoch": 0.11, "learning_rate": 1.3327414814200597e-05, "loss": 0.9881, "step": 6164 }, { "epoch": 0.11, "learning_rate": 1.3325381813900973e-05, "loss": 0.9876, "step": 6165 }, { "epoch": 0.11, "learning_rate": 1.3323348659060483e-05, "loss": 0.9867, "step": 6166 }, { "epoch": 0.11, "learning_rate": 1.3321315349773621e-05, "loss": 0.994, "step": 6167 }, { "epoch": 0.11, "learning_rate": 1.3319281886134876e-05, "loss": 0.9277, "step": 6168 }, { "epoch": 0.11, "learning_rate": 1.3317248268238751e-05, "loss": 1.0073, "step": 6169 }, { "epoch": 0.11, "learning_rate": 1.3315214496179758e-05, "loss": 0.955, "step": 6170 }, { "epoch": 0.11, "learning_rate": 1.3313180570052406e-05, "loss": 0.9406, "step": 6171 }, { "epoch": 0.11, "learning_rate": 1.3311146489951224e-05, "loss": 0.9594, "step": 6172 }, { "epoch": 0.11, "learning_rate": 1.330911225597074e-05, "loss": 0.996, "step": 6173 }, { "epoch": 0.11, "learning_rate": 1.3307077868205487e-05, "loss": 0.9468, "step": 6174 }, { "epoch": 0.11, "learning_rate": 1.3305043326750013e-05, "loss": 0.9606, "step": 6175 }, { "epoch": 0.11, "learning_rate": 1.3303008631698874e-05, "loss": 0.9844, "step": 6176 }, { "epoch": 0.11, "learning_rate": 1.3300973783146622e-05, "loss": 1.0101, "step": 6177 }, { "epoch": 0.11, "learning_rate": 1.3298938781187824e-05, "loss": 0.9642, "step": 6178 }, { "epoch": 0.11, "learning_rate": 1.3296903625917058e-05, "loss": 0.9711, "step": 6179 }, { "epoch": 0.11, "learning_rate": 1.3294868317428895e-05, "loss": 0.9494, "step": 6180 }, { "epoch": 0.11, "learning_rate": 1.329283285581793e-05, "loss": 1.0077, "step": 6181 }, { "epoch": 0.11, "learning_rate": 1.3290797241178752e-05, "loss": 0.9986, "step": 6182 }, { "epoch": 0.11, "learning_rate": 1.3288761473605967e-05, "loss": 0.971, "step": 6183 }, { "epoch": 0.11, "learning_rate": 1.3286725553194178e-05, "loss": 0.9964, "step": 6184 }, { "epoch": 0.11, "learning_rate": 1.3284689480038006e-05, "loss": 0.9947, "step": 6185 }, { "epoch": 0.11, "learning_rate": 1.3282653254232073e-05, "loss": 0.9651, "step": 6186 }, { "epoch": 0.11, "learning_rate": 1.3280616875871004e-05, "loss": 0.981, "step": 6187 }, { "epoch": 0.11, "learning_rate": 1.3278580345049441e-05, "loss": 1.0218, "step": 6188 }, { "epoch": 0.11, "learning_rate": 1.3276543661862023e-05, "loss": 0.9528, "step": 6189 }, { "epoch": 0.11, "learning_rate": 1.3274506826403407e-05, "loss": 0.9301, "step": 6190 }, { "epoch": 0.11, "learning_rate": 1.3272469838768244e-05, "loss": 0.9888, "step": 6191 }, { "epoch": 0.11, "learning_rate": 1.3270432699051206e-05, "loss": 1.0044, "step": 6192 }, { "epoch": 0.11, "learning_rate": 1.3268395407346963e-05, "loss": 0.9887, "step": 6193 }, { "epoch": 0.11, "learning_rate": 1.3266357963750192e-05, "loss": 0.9737, "step": 6194 }, { "epoch": 0.11, "learning_rate": 1.3264320368355581e-05, "loss": 0.9584, "step": 6195 }, { "epoch": 0.11, "learning_rate": 1.3262282621257827e-05, "loss": 0.937, "step": 6196 }, { "epoch": 0.11, "learning_rate": 1.3260244722551623e-05, "loss": 0.9673, "step": 6197 }, { "epoch": 0.11, "learning_rate": 1.3258206672331681e-05, "loss": 0.9955, "step": 6198 }, { "epoch": 0.11, "learning_rate": 1.3256168470692717e-05, "loss": 0.9858, "step": 6199 }, { "epoch": 0.11, "learning_rate": 1.3254130117729448e-05, "loss": 0.9353, "step": 6200 }, { "epoch": 0.11, "learning_rate": 1.3252091613536604e-05, "loss": 0.9994, "step": 6201 }, { "epoch": 0.11, "learning_rate": 1.3250052958208926e-05, "loss": 0.9417, "step": 6202 }, { "epoch": 0.11, "learning_rate": 1.3248014151841151e-05, "loss": 1.005, "step": 6203 }, { "epoch": 0.11, "learning_rate": 1.324597519452803e-05, "loss": 1.0079, "step": 6204 }, { "epoch": 0.11, "learning_rate": 1.324393608636432e-05, "loss": 0.983, "step": 6205 }, { "epoch": 0.11, "learning_rate": 1.3241896827444785e-05, "loss": 0.966, "step": 6206 }, { "epoch": 0.11, "learning_rate": 1.3239857417864193e-05, "loss": 0.9821, "step": 6207 }, { "epoch": 0.11, "learning_rate": 1.3237817857717326e-05, "loss": 0.9791, "step": 6208 }, { "epoch": 0.11, "learning_rate": 1.3235778147098965e-05, "loss": 0.9684, "step": 6209 }, { "epoch": 0.11, "learning_rate": 1.3233738286103905e-05, "loss": 0.9738, "step": 6210 }, { "epoch": 0.11, "learning_rate": 1.3231698274826945e-05, "loss": 0.9834, "step": 6211 }, { "epoch": 0.11, "learning_rate": 1.3229658113362886e-05, "loss": 0.9609, "step": 6212 }, { "epoch": 0.11, "learning_rate": 1.3227617801806543e-05, "loss": 0.9621, "step": 6213 }, { "epoch": 0.11, "learning_rate": 1.322557734025274e-05, "loss": 0.9623, "step": 6214 }, { "epoch": 0.11, "learning_rate": 1.3223536728796296e-05, "loss": 0.9685, "step": 6215 }, { "epoch": 0.11, "learning_rate": 1.322149596753205e-05, "loss": 0.9575, "step": 6216 }, { "epoch": 0.11, "learning_rate": 1.321945505655484e-05, "loss": 1.0194, "step": 6217 }, { "epoch": 0.11, "learning_rate": 1.3217413995959514e-05, "loss": 0.9772, "step": 6218 }, { "epoch": 0.11, "learning_rate": 1.321537278584093e-05, "loss": 0.9747, "step": 6219 }, { "epoch": 0.11, "learning_rate": 1.3213331426293945e-05, "loss": 0.9482, "step": 6220 }, { "epoch": 0.11, "learning_rate": 1.3211289917413426e-05, "loss": 1.0114, "step": 6221 }, { "epoch": 0.11, "learning_rate": 1.3209248259294255e-05, "loss": 0.9699, "step": 6222 }, { "epoch": 0.11, "learning_rate": 1.3207206452031305e-05, "loss": 1.024, "step": 6223 }, { "epoch": 0.11, "learning_rate": 1.3205164495719471e-05, "loss": 0.9978, "step": 6224 }, { "epoch": 0.11, "learning_rate": 1.3203122390453653e-05, "loss": 0.9637, "step": 6225 }, { "epoch": 0.11, "learning_rate": 1.3201080136328748e-05, "loss": 0.946, "step": 6226 }, { "epoch": 0.11, "learning_rate": 1.3199037733439664e-05, "loss": 0.9997, "step": 6227 }, { "epoch": 0.11, "learning_rate": 1.3196995181881325e-05, "loss": 0.9416, "step": 6228 }, { "epoch": 0.11, "learning_rate": 1.3194952481748646e-05, "loss": 1.0036, "step": 6229 }, { "epoch": 0.11, "learning_rate": 1.3192909633136564e-05, "loss": 0.9548, "step": 6230 }, { "epoch": 0.11, "learning_rate": 1.3190866636140017e-05, "loss": 0.9733, "step": 6231 }, { "epoch": 0.11, "learning_rate": 1.3188823490853946e-05, "loss": 0.9591, "step": 6232 }, { "epoch": 0.11, "learning_rate": 1.3186780197373306e-05, "loss": 0.9901, "step": 6233 }, { "epoch": 0.11, "learning_rate": 1.3184736755793051e-05, "loss": 0.8972, "step": 6234 }, { "epoch": 0.11, "learning_rate": 1.3182693166208146e-05, "loss": 0.9667, "step": 6235 }, { "epoch": 0.11, "learning_rate": 1.3180649428713567e-05, "loss": 0.9662, "step": 6236 }, { "epoch": 0.11, "learning_rate": 1.3178605543404292e-05, "loss": 1.0054, "step": 6237 }, { "epoch": 0.11, "learning_rate": 1.3176561510375305e-05, "loss": 1.0192, "step": 6238 }, { "epoch": 0.11, "learning_rate": 1.31745173297216e-05, "loss": 0.9787, "step": 6239 }, { "epoch": 0.12, "learning_rate": 1.3172473001538175e-05, "loss": 0.9911, "step": 6240 }, { "epoch": 0.12, "learning_rate": 1.3170428525920036e-05, "loss": 1.0178, "step": 6241 }, { "epoch": 0.12, "learning_rate": 1.3168383902962198e-05, "loss": 1.011, "step": 6242 }, { "epoch": 0.12, "learning_rate": 1.3166339132759682e-05, "loss": 0.9972, "step": 6243 }, { "epoch": 0.12, "learning_rate": 1.3164294215407511e-05, "loss": 0.9627, "step": 6244 }, { "epoch": 0.12, "learning_rate": 1.3162249151000721e-05, "loss": 1.0292, "step": 6245 }, { "epoch": 0.12, "learning_rate": 1.3160203939634351e-05, "loss": 0.9456, "step": 6246 }, { "epoch": 0.12, "learning_rate": 1.3158158581403448e-05, "loss": 0.9769, "step": 6247 }, { "epoch": 0.12, "learning_rate": 1.3156113076403071e-05, "loss": 0.9489, "step": 6248 }, { "epoch": 0.12, "learning_rate": 1.3154067424728281e-05, "loss": 0.9596, "step": 6249 }, { "epoch": 0.12, "learning_rate": 1.3152021626474136e-05, "loss": 0.984, "step": 6250 }, { "epoch": 0.12, "learning_rate": 1.3149975681735722e-05, "loss": 0.977, "step": 6251 }, { "epoch": 0.12, "learning_rate": 1.3147929590608113e-05, "loss": 0.9853, "step": 6252 }, { "epoch": 0.12, "learning_rate": 1.3145883353186399e-05, "loss": 0.9741, "step": 6253 }, { "epoch": 0.12, "learning_rate": 1.3143836969565678e-05, "loss": 0.9644, "step": 6254 }, { "epoch": 0.12, "learning_rate": 1.3141790439841046e-05, "loss": 0.9862, "step": 6255 }, { "epoch": 0.12, "learning_rate": 1.3139743764107617e-05, "loss": 1.0158, "step": 6256 }, { "epoch": 0.12, "learning_rate": 1.3137696942460511e-05, "loss": 1.0251, "step": 6257 }, { "epoch": 0.12, "learning_rate": 1.3135649974994837e-05, "loss": 0.9644, "step": 6258 }, { "epoch": 0.12, "learning_rate": 1.3133602861805732e-05, "loss": 0.9985, "step": 6259 }, { "epoch": 0.12, "learning_rate": 1.3131555602988334e-05, "loss": 0.9461, "step": 6260 }, { "epoch": 0.12, "learning_rate": 1.3129508198637777e-05, "loss": 0.9877, "step": 6261 }, { "epoch": 0.12, "learning_rate": 1.3127460648849218e-05, "loss": 0.9691, "step": 6262 }, { "epoch": 0.12, "learning_rate": 1.3125412953717808e-05, "loss": 1.0212, "step": 6263 }, { "epoch": 0.12, "learning_rate": 1.3123365113338715e-05, "loss": 0.966, "step": 6264 }, { "epoch": 0.12, "learning_rate": 1.3121317127807105e-05, "loss": 0.9959, "step": 6265 }, { "epoch": 0.12, "learning_rate": 1.3119268997218156e-05, "loss": 0.9676, "step": 6266 }, { "epoch": 0.12, "learning_rate": 1.3117220721667048e-05, "loss": 0.9894, "step": 6267 }, { "epoch": 0.12, "learning_rate": 1.3115172301248975e-05, "loss": 0.9614, "step": 6268 }, { "epoch": 0.12, "learning_rate": 1.3113123736059128e-05, "loss": 0.9881, "step": 6269 }, { "epoch": 0.12, "learning_rate": 1.3111075026192714e-05, "loss": 1.0031, "step": 6270 }, { "epoch": 0.12, "learning_rate": 1.3109026171744945e-05, "loss": 0.9769, "step": 6271 }, { "epoch": 0.12, "learning_rate": 1.3106977172811036e-05, "loss": 0.9456, "step": 6272 }, { "epoch": 0.12, "learning_rate": 1.3104928029486203e-05, "loss": 0.976, "step": 6273 }, { "epoch": 0.12, "learning_rate": 1.3102878741865688e-05, "loss": 0.9491, "step": 6274 }, { "epoch": 0.12, "learning_rate": 1.310082931004472e-05, "loss": 0.9897, "step": 6275 }, { "epoch": 0.12, "learning_rate": 1.3098779734118546e-05, "loss": 0.9462, "step": 6276 }, { "epoch": 0.12, "learning_rate": 1.3096730014182415e-05, "loss": 0.9545, "step": 6277 }, { "epoch": 0.12, "learning_rate": 1.3094680150331582e-05, "loss": 1.0212, "step": 6278 }, { "epoch": 0.12, "learning_rate": 1.3092630142661312e-05, "loss": 0.9642, "step": 6279 }, { "epoch": 0.12, "learning_rate": 1.309057999126688e-05, "loss": 0.9682, "step": 6280 }, { "epoch": 0.12, "learning_rate": 1.3088529696243555e-05, "loss": 1.0251, "step": 6281 }, { "epoch": 0.12, "learning_rate": 1.3086479257686623e-05, "loss": 0.9867, "step": 6282 }, { "epoch": 0.12, "learning_rate": 1.3084428675691378e-05, "loss": 0.9995, "step": 6283 }, { "epoch": 0.12, "learning_rate": 1.3082377950353114e-05, "loss": 0.9947, "step": 6284 }, { "epoch": 0.12, "learning_rate": 1.3080327081767134e-05, "loss": 1.0248, "step": 6285 }, { "epoch": 0.12, "learning_rate": 1.3078276070028748e-05, "loss": 0.9357, "step": 6286 }, { "epoch": 0.12, "learning_rate": 1.3076224915233275e-05, "loss": 0.9341, "step": 6287 }, { "epoch": 0.12, "learning_rate": 1.307417361747604e-05, "loss": 0.9896, "step": 6288 }, { "epoch": 0.12, "learning_rate": 1.3072122176852372e-05, "loss": 1.0014, "step": 6289 }, { "epoch": 0.12, "learning_rate": 1.3070070593457602e-05, "loss": 0.9918, "step": 6290 }, { "epoch": 0.12, "learning_rate": 1.3068018867387083e-05, "loss": 1.0201, "step": 6291 }, { "epoch": 0.12, "learning_rate": 1.3065966998736155e-05, "loss": 0.9664, "step": 6292 }, { "epoch": 0.12, "learning_rate": 1.3063914987600183e-05, "loss": 1.0059, "step": 6293 }, { "epoch": 0.12, "learning_rate": 1.3061862834074528e-05, "loss": 1.003, "step": 6294 }, { "epoch": 0.12, "learning_rate": 1.3059810538254558e-05, "loss": 1.0028, "step": 6295 }, { "epoch": 0.12, "learning_rate": 1.3057758100235652e-05, "loss": 0.953, "step": 6296 }, { "epoch": 0.12, "learning_rate": 1.3055705520113193e-05, "loss": 0.9746, "step": 6297 }, { "epoch": 0.12, "learning_rate": 1.3053652797982567e-05, "loss": 0.9974, "step": 6298 }, { "epoch": 0.12, "learning_rate": 1.3051599933939174e-05, "loss": 0.9797, "step": 6299 }, { "epoch": 0.12, "learning_rate": 1.3049546928078418e-05, "loss": 1.0081, "step": 6300 }, { "epoch": 0.12, "learning_rate": 1.3047493780495706e-05, "loss": 0.9729, "step": 6301 }, { "epoch": 0.12, "learning_rate": 1.3045440491286455e-05, "loss": 1.0249, "step": 6302 }, { "epoch": 0.12, "learning_rate": 1.3043387060546092e-05, "loss": 0.9586, "step": 6303 }, { "epoch": 0.12, "learning_rate": 1.3041333488370036e-05, "loss": 0.9336, "step": 6304 }, { "epoch": 0.12, "learning_rate": 1.303927977485373e-05, "loss": 0.962, "step": 6305 }, { "epoch": 0.12, "learning_rate": 1.303722592009262e-05, "loss": 0.9402, "step": 6306 }, { "epoch": 0.12, "learning_rate": 1.3035171924182145e-05, "loss": 0.9621, "step": 6307 }, { "epoch": 0.12, "learning_rate": 1.3033117787217768e-05, "loss": 0.9758, "step": 6308 }, { "epoch": 0.12, "learning_rate": 1.3031063509294951e-05, "loss": 1.0193, "step": 6309 }, { "epoch": 0.12, "learning_rate": 1.3029009090509162e-05, "loss": 0.9581, "step": 6310 }, { "epoch": 0.12, "learning_rate": 1.3026954530955872e-05, "loss": 0.9901, "step": 6311 }, { "epoch": 0.12, "learning_rate": 1.302489983073057e-05, "loss": 0.9459, "step": 6312 }, { "epoch": 0.12, "learning_rate": 1.3022844989928737e-05, "loss": 0.9934, "step": 6313 }, { "epoch": 0.12, "learning_rate": 1.3020790008645874e-05, "loss": 0.9662, "step": 6314 }, { "epoch": 0.12, "learning_rate": 1.3018734886977477e-05, "loss": 0.9724, "step": 6315 }, { "epoch": 0.12, "learning_rate": 1.3016679625019057e-05, "loss": 0.9212, "step": 6316 }, { "epoch": 0.12, "learning_rate": 1.301462422286613e-05, "loss": 1.0079, "step": 6317 }, { "epoch": 0.12, "learning_rate": 1.3012568680614215e-05, "loss": 0.9379, "step": 6318 }, { "epoch": 0.12, "learning_rate": 1.3010512998358835e-05, "loss": 0.9699, "step": 6319 }, { "epoch": 0.12, "learning_rate": 1.3008457176195535e-05, "loss": 0.9814, "step": 6320 }, { "epoch": 0.12, "learning_rate": 1.3006401214219844e-05, "loss": 0.97, "step": 6321 }, { "epoch": 0.12, "learning_rate": 1.3004345112527312e-05, "loss": 0.976, "step": 6322 }, { "epoch": 0.12, "learning_rate": 1.30022888712135e-05, "loss": 1.0073, "step": 6323 }, { "epoch": 0.12, "learning_rate": 1.3000232490373959e-05, "loss": 1.0048, "step": 6324 }, { "epoch": 0.12, "learning_rate": 1.2998175970104256e-05, "loss": 0.9912, "step": 6325 }, { "epoch": 0.12, "learning_rate": 1.2996119310499972e-05, "loss": 0.956, "step": 6326 }, { "epoch": 0.12, "learning_rate": 1.299406251165668e-05, "loss": 0.9724, "step": 6327 }, { "epoch": 0.12, "learning_rate": 1.2992005573669964e-05, "loss": 1.0111, "step": 6328 }, { "epoch": 0.12, "learning_rate": 1.2989948496635423e-05, "loss": 0.9878, "step": 6329 }, { "epoch": 0.12, "learning_rate": 1.2987891280648648e-05, "loss": 0.9781, "step": 6330 }, { "epoch": 0.12, "learning_rate": 1.2985833925805247e-05, "loss": 1.0004, "step": 6331 }, { "epoch": 0.12, "learning_rate": 1.2983776432200837e-05, "loss": 0.9813, "step": 6332 }, { "epoch": 0.12, "learning_rate": 1.298171879993103e-05, "loss": 1.0132, "step": 6333 }, { "epoch": 0.12, "learning_rate": 1.2979661029091452e-05, "loss": 0.9436, "step": 6334 }, { "epoch": 0.12, "learning_rate": 1.2977603119777737e-05, "loss": 0.9938, "step": 6335 }, { "epoch": 0.12, "learning_rate": 1.2975545072085519e-05, "loss": 0.9306, "step": 6336 }, { "epoch": 0.12, "learning_rate": 1.2973486886110442e-05, "loss": 0.9678, "step": 6337 }, { "epoch": 0.12, "learning_rate": 1.2971428561948158e-05, "loss": 0.9692, "step": 6338 }, { "epoch": 0.12, "learning_rate": 1.2969370099694322e-05, "loss": 0.9315, "step": 6339 }, { "epoch": 0.12, "learning_rate": 1.2967311499444599e-05, "loss": 0.9604, "step": 6340 }, { "epoch": 0.12, "learning_rate": 1.296525276129466e-05, "loss": 0.9514, "step": 6341 }, { "epoch": 0.12, "learning_rate": 1.2963193885340176e-05, "loss": 0.9778, "step": 6342 }, { "epoch": 0.12, "learning_rate": 1.2961134871676832e-05, "loss": 0.9845, "step": 6343 }, { "epoch": 0.12, "learning_rate": 1.2959075720400316e-05, "loss": 0.9399, "step": 6344 }, { "epoch": 0.12, "learning_rate": 1.2957016431606326e-05, "loss": 0.9873, "step": 6345 }, { "epoch": 0.12, "learning_rate": 1.2954957005390563e-05, "loss": 0.9967, "step": 6346 }, { "epoch": 0.12, "learning_rate": 1.295289744184873e-05, "loss": 0.9904, "step": 6347 }, { "epoch": 0.12, "learning_rate": 1.2950837741076545e-05, "loss": 0.9434, "step": 6348 }, { "epoch": 0.12, "learning_rate": 1.2948777903169733e-05, "loss": 0.97, "step": 6349 }, { "epoch": 0.12, "learning_rate": 1.2946717928224012e-05, "loss": 0.9702, "step": 6350 }, { "epoch": 0.12, "learning_rate": 1.2944657816335124e-05, "loss": 0.9494, "step": 6351 }, { "epoch": 0.12, "learning_rate": 1.2942597567598805e-05, "loss": 0.975, "step": 6352 }, { "epoch": 0.12, "learning_rate": 1.2940537182110797e-05, "loss": 1.0028, "step": 6353 }, { "epoch": 0.12, "learning_rate": 1.2938476659966858e-05, "loss": 0.9523, "step": 6354 }, { "epoch": 0.12, "learning_rate": 1.2936416001262749e-05, "loss": 1.0129, "step": 6355 }, { "epoch": 0.12, "learning_rate": 1.2934355206094229e-05, "loss": 1.0029, "step": 6356 }, { "epoch": 0.12, "learning_rate": 1.2932294274557075e-05, "loss": 0.9809, "step": 6357 }, { "epoch": 0.12, "learning_rate": 1.293023320674706e-05, "loss": 0.9431, "step": 6358 }, { "epoch": 0.12, "learning_rate": 1.2928172002759972e-05, "loss": 0.9706, "step": 6359 }, { "epoch": 0.12, "learning_rate": 1.2926110662691597e-05, "loss": 0.9745, "step": 6360 }, { "epoch": 0.12, "learning_rate": 1.292404918663774e-05, "loss": 0.9656, "step": 6361 }, { "epoch": 0.12, "learning_rate": 1.2921987574694196e-05, "loss": 1.0104, "step": 6362 }, { "epoch": 0.12, "learning_rate": 1.291992582695678e-05, "loss": 0.9892, "step": 6363 }, { "epoch": 0.12, "learning_rate": 1.2917863943521305e-05, "loss": 0.9506, "step": 6364 }, { "epoch": 0.12, "learning_rate": 1.2915801924483594e-05, "loss": 0.9492, "step": 6365 }, { "epoch": 0.12, "learning_rate": 1.2913739769939476e-05, "loss": 0.9411, "step": 6366 }, { "epoch": 0.12, "learning_rate": 1.2911677479984781e-05, "loss": 1.0055, "step": 6367 }, { "epoch": 0.12, "learning_rate": 1.2909615054715357e-05, "loss": 0.976, "step": 6368 }, { "epoch": 0.12, "learning_rate": 1.2907552494227049e-05, "loss": 1.0178, "step": 6369 }, { "epoch": 0.12, "learning_rate": 1.2905489798615709e-05, "loss": 0.9756, "step": 6370 }, { "epoch": 0.12, "learning_rate": 1.2903426967977196e-05, "loss": 1.0208, "step": 6371 }, { "epoch": 0.12, "learning_rate": 1.2901364002407384e-05, "loss": 0.9726, "step": 6372 }, { "epoch": 0.12, "learning_rate": 1.2899300902002137e-05, "loss": 0.9622, "step": 6373 }, { "epoch": 0.12, "learning_rate": 1.2897237666857332e-05, "loss": 0.979, "step": 6374 }, { "epoch": 0.12, "learning_rate": 1.2895174297068861e-05, "loss": 1.009, "step": 6375 }, { "epoch": 0.12, "learning_rate": 1.2893110792732615e-05, "loss": 0.979, "step": 6376 }, { "epoch": 0.12, "learning_rate": 1.2891047153944485e-05, "loss": 0.9592, "step": 6377 }, { "epoch": 0.12, "learning_rate": 1.2888983380800382e-05, "loss": 0.9627, "step": 6378 }, { "epoch": 0.12, "learning_rate": 1.2886919473396212e-05, "loss": 0.9479, "step": 6379 }, { "epoch": 0.12, "learning_rate": 1.2884855431827892e-05, "loss": 0.9868, "step": 6380 }, { "epoch": 0.12, "learning_rate": 1.2882791256191344e-05, "loss": 0.9837, "step": 6381 }, { "epoch": 0.12, "learning_rate": 1.2880726946582496e-05, "loss": 0.98, "step": 6382 }, { "epoch": 0.12, "learning_rate": 1.2878662503097283e-05, "loss": 0.9792, "step": 6383 }, { "epoch": 0.12, "learning_rate": 1.2876597925831648e-05, "loss": 0.9928, "step": 6384 }, { "epoch": 0.12, "learning_rate": 1.2874533214881537e-05, "loss": 0.9571, "step": 6385 }, { "epoch": 0.12, "learning_rate": 1.2872468370342908e-05, "loss": 1.0139, "step": 6386 }, { "epoch": 0.12, "learning_rate": 1.2870403392311715e-05, "loss": 0.9886, "step": 6387 }, { "epoch": 0.12, "learning_rate": 1.2868338280883924e-05, "loss": 0.9957, "step": 6388 }, { "epoch": 0.12, "learning_rate": 1.2866273036155508e-05, "loss": 0.9849, "step": 6389 }, { "epoch": 0.12, "learning_rate": 1.2864207658222444e-05, "loss": 0.9824, "step": 6390 }, { "epoch": 0.13, "learning_rate": 1.2862142147180723e-05, "loss": 1.0218, "step": 6391 }, { "epoch": 0.13, "learning_rate": 1.2860076503126332e-05, "loss": 0.9141, "step": 6392 }, { "epoch": 0.13, "learning_rate": 1.2858010726155263e-05, "loss": 1.008, "step": 6393 }, { "epoch": 0.13, "learning_rate": 1.2855944816363528e-05, "loss": 0.9564, "step": 6394 }, { "epoch": 0.13, "learning_rate": 1.2853878773847133e-05, "loss": 0.9591, "step": 6395 }, { "epoch": 0.13, "learning_rate": 1.2851812598702087e-05, "loss": 0.9717, "step": 6396 }, { "epoch": 0.13, "learning_rate": 1.284974629102442e-05, "loss": 0.9582, "step": 6397 }, { "epoch": 0.13, "learning_rate": 1.2847679850910156e-05, "loss": 0.9842, "step": 6398 }, { "epoch": 0.13, "learning_rate": 1.2845613278455329e-05, "loss": 0.9792, "step": 6399 }, { "epoch": 0.13, "learning_rate": 1.2843546573755979e-05, "loss": 1.0134, "step": 6400 }, { "epoch": 0.13, "learning_rate": 1.2841479736908154e-05, "loss": 0.9899, "step": 6401 }, { "epoch": 0.13, "learning_rate": 1.2839412768007905e-05, "loss": 0.972, "step": 6402 }, { "epoch": 0.13, "learning_rate": 1.283734566715129e-05, "loss": 0.9871, "step": 6403 }, { "epoch": 0.13, "learning_rate": 1.2835278434434375e-05, "loss": 0.9907, "step": 6404 }, { "epoch": 0.13, "learning_rate": 1.2833211069953228e-05, "loss": 0.9624, "step": 6405 }, { "epoch": 0.13, "learning_rate": 1.2831143573803931e-05, "loss": 0.9504, "step": 6406 }, { "epoch": 0.13, "learning_rate": 1.2829075946082563e-05, "loss": 0.9673, "step": 6407 }, { "epoch": 0.13, "learning_rate": 1.282700818688521e-05, "loss": 1.0, "step": 6408 }, { "epoch": 0.13, "learning_rate": 1.2824940296307977e-05, "loss": 0.9653, "step": 6409 }, { "epoch": 0.13, "learning_rate": 1.2822872274446958e-05, "loss": 0.9176, "step": 6410 }, { "epoch": 0.13, "learning_rate": 1.282080412139826e-05, "loss": 0.996, "step": 6411 }, { "epoch": 0.13, "learning_rate": 1.2818735837257998e-05, "loss": 1.0069, "step": 6412 }, { "epoch": 0.13, "learning_rate": 1.2816667422122295e-05, "loss": 0.9681, "step": 6413 }, { "epoch": 0.13, "learning_rate": 1.281459887608727e-05, "loss": 0.988, "step": 6414 }, { "epoch": 0.13, "learning_rate": 1.2812530199249063e-05, "loss": 0.9808, "step": 6415 }, { "epoch": 0.13, "learning_rate": 1.2810461391703804e-05, "loss": 0.9618, "step": 6416 }, { "epoch": 0.13, "learning_rate": 1.280839245354764e-05, "loss": 0.9631, "step": 6417 }, { "epoch": 0.13, "learning_rate": 1.2806323384876727e-05, "loss": 0.9905, "step": 6418 }, { "epoch": 0.13, "learning_rate": 1.2804254185787208e-05, "loss": 0.9699, "step": 6419 }, { "epoch": 0.13, "learning_rate": 1.2802184856375255e-05, "loss": 0.9691, "step": 6420 }, { "epoch": 0.13, "learning_rate": 1.2800115396737036e-05, "loss": 0.9965, "step": 6421 }, { "epoch": 0.13, "learning_rate": 1.2798045806968722e-05, "loss": 0.9711, "step": 6422 }, { "epoch": 0.13, "learning_rate": 1.279597608716649e-05, "loss": 1.0086, "step": 6423 }, { "epoch": 0.13, "learning_rate": 1.279390623742654e-05, "loss": 0.9777, "step": 6424 }, { "epoch": 0.13, "learning_rate": 1.2791836257845046e-05, "loss": 0.9776, "step": 6425 }, { "epoch": 0.13, "learning_rate": 1.2789766148518221e-05, "loss": 0.9617, "step": 6426 }, { "epoch": 0.13, "learning_rate": 1.2787695909542262e-05, "loss": 1.0059, "step": 6427 }, { "epoch": 0.13, "learning_rate": 1.278562554101338e-05, "loss": 0.9685, "step": 6428 }, { "epoch": 0.13, "learning_rate": 1.2783555043027794e-05, "loss": 0.971, "step": 6429 }, { "epoch": 0.13, "learning_rate": 1.2781484415681725e-05, "loss": 0.9817, "step": 6430 }, { "epoch": 0.13, "learning_rate": 1.27794136590714e-05, "loss": 0.9825, "step": 6431 }, { "epoch": 0.13, "learning_rate": 1.277734277329306e-05, "loss": 0.9957, "step": 6432 }, { "epoch": 0.13, "learning_rate": 1.2775271758442937e-05, "loss": 0.9732, "step": 6433 }, { "epoch": 0.13, "learning_rate": 1.2773200614617283e-05, "loss": 0.9273, "step": 6434 }, { "epoch": 0.13, "learning_rate": 1.2771129341912345e-05, "loss": 0.9597, "step": 6435 }, { "epoch": 0.13, "learning_rate": 1.2769057940424391e-05, "loss": 0.9928, "step": 6436 }, { "epoch": 0.13, "learning_rate": 1.2766986410249676e-05, "loss": 0.9669, "step": 6437 }, { "epoch": 0.13, "learning_rate": 1.2764914751484477e-05, "loss": 0.9547, "step": 6438 }, { "epoch": 0.13, "learning_rate": 1.2762842964225066e-05, "loss": 1.0038, "step": 6439 }, { "epoch": 0.13, "learning_rate": 1.2760771048567726e-05, "loss": 0.9616, "step": 6440 }, { "epoch": 0.13, "learning_rate": 1.275869900460875e-05, "loss": 0.9919, "step": 6441 }, { "epoch": 0.13, "learning_rate": 1.2756626832444428e-05, "loss": 0.9159, "step": 6442 }, { "epoch": 0.13, "learning_rate": 1.2754554532171057e-05, "loss": 0.9563, "step": 6443 }, { "epoch": 0.13, "learning_rate": 1.2752482103884953e-05, "loss": 0.9941, "step": 6444 }, { "epoch": 0.13, "learning_rate": 1.2750409547682419e-05, "loss": 1.0091, "step": 6445 }, { "epoch": 0.13, "learning_rate": 1.2748336863659775e-05, "loss": 1.0033, "step": 6446 }, { "epoch": 0.13, "learning_rate": 1.2746264051913353e-05, "loss": 0.9692, "step": 6447 }, { "epoch": 0.13, "learning_rate": 1.2744191112539475e-05, "loss": 0.9139, "step": 6448 }, { "epoch": 0.13, "learning_rate": 1.2742118045634476e-05, "loss": 0.9789, "step": 6449 }, { "epoch": 0.13, "learning_rate": 1.2740044851294704e-05, "loss": 0.9703, "step": 6450 }, { "epoch": 0.13, "learning_rate": 1.2737971529616501e-05, "loss": 0.9423, "step": 6451 }, { "epoch": 0.13, "learning_rate": 1.2735898080696223e-05, "loss": 0.9909, "step": 6452 }, { "epoch": 0.13, "learning_rate": 1.2733824504630233e-05, "loss": 0.9897, "step": 6453 }, { "epoch": 0.13, "learning_rate": 1.273175080151489e-05, "loss": 0.9718, "step": 6454 }, { "epoch": 0.13, "learning_rate": 1.2729676971446572e-05, "loss": 0.941, "step": 6455 }, { "epoch": 0.13, "learning_rate": 1.2727603014521655e-05, "loss": 0.992, "step": 6456 }, { "epoch": 0.13, "learning_rate": 1.2725528930836515e-05, "loss": 0.985, "step": 6457 }, { "epoch": 0.13, "learning_rate": 1.272345472048755e-05, "loss": 0.9824, "step": 6458 }, { "epoch": 0.13, "learning_rate": 1.2721380383571149e-05, "loss": 0.9724, "step": 6459 }, { "epoch": 0.13, "learning_rate": 1.2719305920183718e-05, "loss": 0.9649, "step": 6460 }, { "epoch": 0.13, "learning_rate": 1.2717231330421661e-05, "loss": 0.9366, "step": 6461 }, { "epoch": 0.13, "learning_rate": 1.271515661438139e-05, "loss": 0.9683, "step": 6462 }, { "epoch": 0.13, "learning_rate": 1.2713081772159324e-05, "loss": 0.9806, "step": 6463 }, { "epoch": 0.13, "learning_rate": 1.271100680385189e-05, "loss": 1.0041, "step": 6464 }, { "epoch": 0.13, "learning_rate": 1.2708931709555517e-05, "loss": 0.979, "step": 6465 }, { "epoch": 0.13, "learning_rate": 1.2706856489366638e-05, "loss": 1.0032, "step": 6466 }, { "epoch": 0.13, "learning_rate": 1.2704781143381699e-05, "loss": 1.0067, "step": 6467 }, { "epoch": 0.13, "learning_rate": 1.2702705671697144e-05, "loss": 0.9815, "step": 6468 }, { "epoch": 0.13, "learning_rate": 1.2700630074409427e-05, "loss": 0.9826, "step": 6469 }, { "epoch": 0.13, "learning_rate": 1.2698554351615013e-05, "loss": 1.0006, "step": 6470 }, { "epoch": 0.13, "learning_rate": 1.2696478503410364e-05, "loss": 1.0279, "step": 6471 }, { "epoch": 0.13, "learning_rate": 1.2694402529891947e-05, "loss": 0.9212, "step": 6472 }, { "epoch": 0.13, "learning_rate": 1.2692326431156245e-05, "loss": 0.9777, "step": 6473 }, { "epoch": 0.13, "learning_rate": 1.269025020729974e-05, "loss": 0.9579, "step": 6474 }, { "epoch": 0.13, "learning_rate": 1.2688173858418913e-05, "loss": 1.0014, "step": 6475 }, { "epoch": 0.13, "learning_rate": 1.268609738461027e-05, "loss": 0.9895, "step": 6476 }, { "epoch": 0.13, "learning_rate": 1.2684020785970302e-05, "loss": 0.989, "step": 6477 }, { "epoch": 0.13, "learning_rate": 1.2681944062595521e-05, "loss": 0.9519, "step": 6478 }, { "epoch": 0.13, "learning_rate": 1.2679867214582438e-05, "loss": 1.0018, "step": 6479 }, { "epoch": 0.13, "learning_rate": 1.2677790242027564e-05, "loss": 0.9821, "step": 6480 }, { "epoch": 0.13, "learning_rate": 1.2675713145027428e-05, "loss": 0.966, "step": 6481 }, { "epoch": 0.13, "learning_rate": 1.2673635923678562e-05, "loss": 0.9371, "step": 6482 }, { "epoch": 0.13, "learning_rate": 1.2671558578077497e-05, "loss": 0.9426, "step": 6483 }, { "epoch": 0.13, "learning_rate": 1.2669481108320773e-05, "loss": 0.9779, "step": 6484 }, { "epoch": 0.13, "learning_rate": 1.2667403514504937e-05, "loss": 1.0039, "step": 6485 }, { "epoch": 0.13, "learning_rate": 1.2665325796726541e-05, "loss": 0.9314, "step": 6486 }, { "epoch": 0.13, "learning_rate": 1.2663247955082147e-05, "loss": 1.0251, "step": 6487 }, { "epoch": 0.13, "learning_rate": 1.2661169989668315e-05, "loss": 0.9787, "step": 6488 }, { "epoch": 0.13, "learning_rate": 1.2659091900581615e-05, "loss": 0.959, "step": 6489 }, { "epoch": 0.13, "learning_rate": 1.2657013687918624e-05, "loss": 0.9961, "step": 6490 }, { "epoch": 0.13, "learning_rate": 1.265493535177592e-05, "loss": 0.9713, "step": 6491 }, { "epoch": 0.13, "learning_rate": 1.265285689225009e-05, "loss": 0.9208, "step": 6492 }, { "epoch": 0.13, "learning_rate": 1.265077830943773e-05, "loss": 0.984, "step": 6493 }, { "epoch": 0.13, "learning_rate": 1.2648699603435435e-05, "loss": 0.9661, "step": 6494 }, { "epoch": 0.13, "learning_rate": 1.2646620774339812e-05, "loss": 0.9706, "step": 6495 }, { "epoch": 0.13, "learning_rate": 1.2644541822247468e-05, "loss": 0.9675, "step": 6496 }, { "epoch": 0.13, "learning_rate": 1.2642462747255018e-05, "loss": 0.96, "step": 6497 }, { "epoch": 0.13, "learning_rate": 1.2640383549459086e-05, "loss": 1.0058, "step": 6498 }, { "epoch": 0.13, "learning_rate": 1.2638304228956295e-05, "loss": 0.9893, "step": 6499 }, { "epoch": 0.13, "learning_rate": 1.263622478584328e-05, "loss": 0.9566, "step": 6500 }, { "epoch": 0.13, "learning_rate": 1.263414522021668e-05, "loss": 1.0218, "step": 6501 }, { "epoch": 0.13, "learning_rate": 1.2632065532173137e-05, "loss": 0.9735, "step": 6502 }, { "epoch": 0.13, "learning_rate": 1.26299857218093e-05, "loss": 0.9911, "step": 6503 }, { "epoch": 0.13, "learning_rate": 1.2627905789221826e-05, "loss": 0.9716, "step": 6504 }, { "epoch": 0.13, "learning_rate": 1.2625825734507377e-05, "loss": 0.9781, "step": 6505 }, { "epoch": 0.13, "learning_rate": 1.2623745557762616e-05, "loss": 1.0114, "step": 6506 }, { "epoch": 0.13, "learning_rate": 1.262166525908422e-05, "loss": 1.0435, "step": 6507 }, { "epoch": 0.13, "learning_rate": 1.2619584838568862e-05, "loss": 0.9626, "step": 6508 }, { "epoch": 0.13, "learning_rate": 1.2617504296313228e-05, "loss": 0.9823, "step": 6509 }, { "epoch": 0.13, "learning_rate": 1.261542363241401e-05, "loss": 0.9357, "step": 6510 }, { "epoch": 0.13, "learning_rate": 1.2613342846967898e-05, "loss": 0.9226, "step": 6511 }, { "epoch": 0.13, "learning_rate": 1.2611261940071592e-05, "loss": 0.9257, "step": 6512 }, { "epoch": 0.13, "learning_rate": 1.2609180911821808e-05, "loss": 0.9536, "step": 6513 }, { "epoch": 0.13, "learning_rate": 1.2607099762315245e-05, "loss": 0.9788, "step": 6514 }, { "epoch": 0.13, "learning_rate": 1.2605018491648626e-05, "loss": 0.9596, "step": 6515 }, { "epoch": 0.13, "learning_rate": 1.260293709991868e-05, "loss": 0.9736, "step": 6516 }, { "epoch": 0.13, "learning_rate": 1.2600855587222125e-05, "loss": 0.995, "step": 6517 }, { "epoch": 0.13, "learning_rate": 1.2598773953655705e-05, "loss": 0.9711, "step": 6518 }, { "epoch": 0.13, "learning_rate": 1.2596692199316154e-05, "loss": 0.9835, "step": 6519 }, { "epoch": 0.13, "learning_rate": 1.2594610324300218e-05, "loss": 0.9811, "step": 6520 }, { "epoch": 0.13, "learning_rate": 1.2592528328704649e-05, "loss": 0.9585, "step": 6521 }, { "epoch": 0.13, "learning_rate": 1.2590446212626208e-05, "loss": 0.997, "step": 6522 }, { "epoch": 0.13, "learning_rate": 1.258836397616165e-05, "loss": 0.9612, "step": 6523 }, { "epoch": 0.13, "learning_rate": 1.2586281619407747e-05, "loss": 0.9311, "step": 6524 }, { "epoch": 0.13, "learning_rate": 1.2584199142461275e-05, "loss": 0.951, "step": 6525 }, { "epoch": 0.13, "learning_rate": 1.2582116545419012e-05, "loss": 0.974, "step": 6526 }, { "epoch": 0.13, "learning_rate": 1.2580033828377736e-05, "loss": 0.9751, "step": 6527 }, { "epoch": 0.13, "learning_rate": 1.2577950991434249e-05, "loss": 0.9487, "step": 6528 }, { "epoch": 0.13, "learning_rate": 1.2575868034685335e-05, "loss": 0.9688, "step": 6529 }, { "epoch": 0.13, "learning_rate": 1.2573784958227803e-05, "loss": 0.995, "step": 6530 }, { "epoch": 0.13, "learning_rate": 1.2571701762158458e-05, "loss": 0.9664, "step": 6531 }, { "epoch": 0.13, "learning_rate": 1.2569618446574115e-05, "loss": 0.9803, "step": 6532 }, { "epoch": 0.13, "learning_rate": 1.2567535011571588e-05, "loss": 0.9283, "step": 6533 }, { "epoch": 0.13, "learning_rate": 1.2565451457247705e-05, "loss": 0.9475, "step": 6534 }, { "epoch": 0.13, "learning_rate": 1.256336778369929e-05, "loss": 0.9624, "step": 6535 }, { "epoch": 0.13, "learning_rate": 1.2561283991023183e-05, "loss": 0.9775, "step": 6536 }, { "epoch": 0.13, "learning_rate": 1.2559200079316221e-05, "loss": 0.9641, "step": 6537 }, { "epoch": 0.13, "learning_rate": 1.255711604867525e-05, "loss": 0.9634, "step": 6538 }, { "epoch": 0.13, "learning_rate": 1.2555031899197125e-05, "loss": 1.0022, "step": 6539 }, { "epoch": 0.13, "learning_rate": 1.2552947630978698e-05, "loss": 0.9688, "step": 6540 }, { "epoch": 0.14, "learning_rate": 1.2550863244116836e-05, "loss": 0.9504, "step": 6541 }, { "epoch": 0.14, "learning_rate": 1.2548778738708405e-05, "loss": 0.9894, "step": 6542 }, { "epoch": 0.14, "learning_rate": 1.2546694114850275e-05, "loss": 0.9971, "step": 6543 }, { "epoch": 0.14, "learning_rate": 1.2544609372639331e-05, "loss": 0.9567, "step": 6544 }, { "epoch": 0.14, "learning_rate": 1.2542524512172454e-05, "loss": 0.9787, "step": 6545 }, { "epoch": 0.14, "learning_rate": 1.2540439533546534e-05, "loss": 0.9962, "step": 6546 }, { "epoch": 0.14, "learning_rate": 1.2538354436858467e-05, "loss": 0.9739, "step": 6547 }, { "epoch": 0.14, "learning_rate": 1.2536269222205156e-05, "loss": 0.967, "step": 6548 }, { "epoch": 0.14, "learning_rate": 1.2534183889683504e-05, "loss": 0.9613, "step": 6549 }, { "epoch": 0.14, "learning_rate": 1.2532098439390423e-05, "loss": 0.969, "step": 6550 }, { "epoch": 0.14, "learning_rate": 1.2530012871422834e-05, "loss": 0.9428, "step": 6551 }, { "epoch": 0.14, "learning_rate": 1.2527927185877658e-05, "loss": 0.9513, "step": 6552 }, { "epoch": 0.14, "learning_rate": 1.252584138285182e-05, "loss": 0.9622, "step": 6553 }, { "epoch": 0.14, "learning_rate": 1.252375546244226e-05, "loss": 0.9633, "step": 6554 }, { "epoch": 0.14, "learning_rate": 1.252166942474591e-05, "loss": 0.9655, "step": 6555 }, { "epoch": 0.14, "learning_rate": 1.2519583269859723e-05, "loss": 0.9478, "step": 6556 }, { "epoch": 0.14, "learning_rate": 1.2517496997880642e-05, "loss": 0.9728, "step": 6557 }, { "epoch": 0.14, "learning_rate": 1.2515410608905626e-05, "loss": 0.9391, "step": 6558 }, { "epoch": 0.14, "learning_rate": 1.2513324103031636e-05, "loss": 0.9792, "step": 6559 }, { "epoch": 0.14, "learning_rate": 1.2511237480355636e-05, "loss": 0.993, "step": 6560 }, { "epoch": 0.14, "learning_rate": 1.2509150740974599e-05, "loss": 0.9527, "step": 6561 }, { "epoch": 0.14, "learning_rate": 1.2507063884985507e-05, "loss": 0.9762, "step": 6562 }, { "epoch": 0.14, "learning_rate": 1.2504976912485334e-05, "loss": 1.0109, "step": 6563 }, { "epoch": 0.14, "learning_rate": 1.2502889823571077e-05, "loss": 0.9657, "step": 6564 }, { "epoch": 0.14, "learning_rate": 1.2500802618339723e-05, "loss": 0.944, "step": 6565 }, { "epoch": 0.14, "learning_rate": 1.249871529688827e-05, "loss": 0.9756, "step": 6566 }, { "epoch": 0.14, "learning_rate": 1.2496627859313728e-05, "loss": 0.9509, "step": 6567 }, { "epoch": 0.14, "learning_rate": 1.2494540305713105e-05, "loss": 0.9793, "step": 6568 }, { "epoch": 0.14, "learning_rate": 1.2492452636183416e-05, "loss": 0.9863, "step": 6569 }, { "epoch": 0.14, "learning_rate": 1.249036485082168e-05, "loss": 0.9796, "step": 6570 }, { "epoch": 0.14, "learning_rate": 1.2488276949724923e-05, "loss": 0.9614, "step": 6571 }, { "epoch": 0.14, "learning_rate": 1.2486188932990179e-05, "loss": 0.9827, "step": 6572 }, { "epoch": 0.14, "learning_rate": 1.248410080071448e-05, "loss": 1.0104, "step": 6573 }, { "epoch": 0.14, "learning_rate": 1.2482012552994873e-05, "loss": 0.9475, "step": 6574 }, { "epoch": 0.14, "learning_rate": 1.24799241899284e-05, "loss": 1.0082, "step": 6575 }, { "epoch": 0.14, "learning_rate": 1.2477835711612118e-05, "loss": 0.9548, "step": 6576 }, { "epoch": 0.14, "learning_rate": 1.2475747118143087e-05, "loss": 0.9657, "step": 6577 }, { "epoch": 0.14, "learning_rate": 1.2473658409618363e-05, "loss": 0.9285, "step": 6578 }, { "epoch": 0.14, "learning_rate": 1.2471569586135024e-05, "loss": 1.0049, "step": 6579 }, { "epoch": 0.14, "learning_rate": 1.2469480647790137e-05, "loss": 0.9838, "step": 6580 }, { "epoch": 0.14, "learning_rate": 1.2467391594680782e-05, "loss": 1.0081, "step": 6581 }, { "epoch": 0.14, "learning_rate": 1.246530242690405e-05, "loss": 0.9626, "step": 6582 }, { "epoch": 0.14, "learning_rate": 1.246321314455702e-05, "loss": 0.9635, "step": 6583 }, { "epoch": 0.14, "learning_rate": 1.2461123747736799e-05, "loss": 0.9611, "step": 6584 }, { "epoch": 0.14, "learning_rate": 1.2459034236540481e-05, "loss": 0.9802, "step": 6585 }, { "epoch": 0.14, "learning_rate": 1.2456944611065175e-05, "loss": 0.984, "step": 6586 }, { "epoch": 0.14, "learning_rate": 1.2454854871407993e-05, "loss": 1.0019, "step": 6587 }, { "epoch": 0.14, "learning_rate": 1.2452765017666051e-05, "loss": 1.0166, "step": 6588 }, { "epoch": 0.14, "learning_rate": 1.2450675049936467e-05, "loss": 0.9893, "step": 6589 }, { "epoch": 0.14, "learning_rate": 1.244858496831637e-05, "loss": 1.0128, "step": 6590 }, { "epoch": 0.14, "learning_rate": 1.24464947729029e-05, "loss": 0.9936, "step": 6591 }, { "epoch": 0.14, "learning_rate": 1.2444404463793185e-05, "loss": 0.9854, "step": 6592 }, { "epoch": 0.14, "learning_rate": 1.2442314041084372e-05, "loss": 0.9788, "step": 6593 }, { "epoch": 0.14, "learning_rate": 1.2440223504873612e-05, "loss": 0.9095, "step": 6594 }, { "epoch": 0.14, "learning_rate": 1.2438132855258055e-05, "loss": 0.962, "step": 6595 }, { "epoch": 0.14, "learning_rate": 1.243604209233486e-05, "loss": 0.9755, "step": 6596 }, { "epoch": 0.14, "learning_rate": 1.2433951216201197e-05, "loss": 1.0367, "step": 6597 }, { "epoch": 0.14, "learning_rate": 1.2431860226954227e-05, "loss": 0.9587, "step": 6598 }, { "epoch": 0.14, "learning_rate": 1.2429769124691128e-05, "loss": 1.0277, "step": 6599 }, { "epoch": 0.14, "learning_rate": 1.2427677909509086e-05, "loss": 0.9884, "step": 6600 }, { "epoch": 0.14, "learning_rate": 1.242558658150528e-05, "loss": 0.9842, "step": 6601 }, { "epoch": 0.14, "learning_rate": 1.2423495140776901e-05, "loss": 1.0009, "step": 6602 }, { "epoch": 0.14, "learning_rate": 1.2421403587421148e-05, "loss": 0.9917, "step": 6603 }, { "epoch": 0.14, "learning_rate": 1.2419311921535217e-05, "loss": 0.9903, "step": 6604 }, { "epoch": 0.14, "learning_rate": 1.2417220143216315e-05, "loss": 0.9767, "step": 6605 }, { "epoch": 0.14, "learning_rate": 1.241512825256166e-05, "loss": 1.0034, "step": 6606 }, { "epoch": 0.14, "learning_rate": 1.2413036249668462e-05, "loss": 0.969, "step": 6607 }, { "epoch": 0.14, "learning_rate": 1.2410944134633944e-05, "loss": 0.9845, "step": 6608 }, { "epoch": 0.14, "learning_rate": 1.2408851907555335e-05, "loss": 0.9593, "step": 6609 }, { "epoch": 0.14, "learning_rate": 1.2406759568529867e-05, "loss": 1.0356, "step": 6610 }, { "epoch": 0.14, "learning_rate": 1.2404667117654779e-05, "loss": 1.0099, "step": 6611 }, { "epoch": 0.14, "learning_rate": 1.2402574555027306e-05, "loss": 0.9592, "step": 6612 }, { "epoch": 0.14, "learning_rate": 1.2400481880744705e-05, "loss": 0.9827, "step": 6613 }, { "epoch": 0.14, "learning_rate": 1.2398389094904226e-05, "loss": 0.98, "step": 6614 }, { "epoch": 0.14, "learning_rate": 1.2396296197603126e-05, "loss": 0.9669, "step": 6615 }, { "epoch": 0.14, "learning_rate": 1.239420318893867e-05, "loss": 0.9803, "step": 6616 }, { "epoch": 0.14, "learning_rate": 1.2392110069008125e-05, "loss": 0.9416, "step": 6617 }, { "epoch": 0.14, "learning_rate": 1.2390016837908766e-05, "loss": 0.9841, "step": 6618 }, { "epoch": 0.14, "learning_rate": 1.2387923495737872e-05, "loss": 0.9704, "step": 6619 }, { "epoch": 0.14, "learning_rate": 1.238583004259273e-05, "loss": 0.9932, "step": 6620 }, { "epoch": 0.14, "learning_rate": 1.2383736478570623e-05, "loss": 1.0085, "step": 6621 }, { "epoch": 0.14, "learning_rate": 1.2381642803768847e-05, "loss": 0.9802, "step": 6622 }, { "epoch": 0.14, "learning_rate": 1.2379549018284707e-05, "loss": 0.9183, "step": 6623 }, { "epoch": 0.14, "learning_rate": 1.23774551222155e-05, "loss": 0.9652, "step": 6624 }, { "epoch": 0.14, "learning_rate": 1.2375361115658544e-05, "loss": 0.9498, "step": 6625 }, { "epoch": 0.14, "learning_rate": 1.2373266998711152e-05, "loss": 0.9911, "step": 6626 }, { "epoch": 0.14, "learning_rate": 1.2371172771470637e-05, "loss": 0.9788, "step": 6627 }, { "epoch": 0.14, "learning_rate": 1.2369078434034329e-05, "loss": 0.9961, "step": 6628 }, { "epoch": 0.14, "learning_rate": 1.2366983986499562e-05, "loss": 1.0088, "step": 6629 }, { "epoch": 0.14, "learning_rate": 1.2364889428963667e-05, "loss": 0.9759, "step": 6630 }, { "epoch": 0.14, "learning_rate": 1.2362794761523987e-05, "loss": 0.9438, "step": 6631 }, { "epoch": 0.14, "learning_rate": 1.2360699984277864e-05, "loss": 0.9567, "step": 6632 }, { "epoch": 0.14, "learning_rate": 1.2358605097322653e-05, "loss": 1.0052, "step": 6633 }, { "epoch": 0.14, "learning_rate": 1.235651010075571e-05, "loss": 0.9664, "step": 6634 }, { "epoch": 0.14, "learning_rate": 1.2354414994674391e-05, "loss": 0.9871, "step": 6635 }, { "epoch": 0.14, "learning_rate": 1.2352319779176064e-05, "loss": 0.938, "step": 6636 }, { "epoch": 0.14, "learning_rate": 1.2350224454358109e-05, "loss": 1.002, "step": 6637 }, { "epoch": 0.14, "learning_rate": 1.234812902031789e-05, "loss": 0.9899, "step": 6638 }, { "epoch": 0.14, "learning_rate": 1.2346033477152792e-05, "loss": 0.9918, "step": 6639 }, { "epoch": 0.14, "learning_rate": 1.2343937824960206e-05, "loss": 1.0357, "step": 6640 }, { "epoch": 0.14, "learning_rate": 1.234184206383752e-05, "loss": 1.0071, "step": 6641 }, { "epoch": 0.14, "learning_rate": 1.2339746193882127e-05, "loss": 0.9764, "step": 6642 }, { "epoch": 0.14, "learning_rate": 1.233765021519144e-05, "loss": 0.9884, "step": 6643 }, { "epoch": 0.14, "learning_rate": 1.233555412786285e-05, "loss": 0.9408, "step": 6644 }, { "epoch": 0.14, "learning_rate": 1.233345793199378e-05, "loss": 1.0117, "step": 6645 }, { "epoch": 0.14, "learning_rate": 1.2331361627681645e-05, "loss": 0.9471, "step": 6646 }, { "epoch": 0.14, "learning_rate": 1.2329265215023862e-05, "loss": 0.9263, "step": 6647 }, { "epoch": 0.14, "learning_rate": 1.2327168694117864e-05, "loss": 0.9404, "step": 6648 }, { "epoch": 0.14, "learning_rate": 1.2325072065061082e-05, "loss": 0.9604, "step": 6649 }, { "epoch": 0.14, "learning_rate": 1.2322975327950946e-05, "loss": 0.9746, "step": 6650 }, { "epoch": 0.14, "learning_rate": 1.2320878482884904e-05, "loss": 0.9675, "step": 6651 }, { "epoch": 0.14, "learning_rate": 1.2318781529960405e-05, "loss": 0.9444, "step": 6652 }, { "epoch": 0.14, "learning_rate": 1.2316684469274894e-05, "loss": 0.9785, "step": 6653 }, { "epoch": 0.14, "learning_rate": 1.2314587300925834e-05, "loss": 0.9511, "step": 6654 }, { "epoch": 0.14, "learning_rate": 1.2312490025010681e-05, "loss": 0.9994, "step": 6655 }, { "epoch": 0.14, "learning_rate": 1.2310392641626912e-05, "loss": 0.9949, "step": 6656 }, { "epoch": 0.14, "learning_rate": 1.2308295150871987e-05, "loss": 1.0295, "step": 6657 }, { "epoch": 0.14, "learning_rate": 1.2306197552843391e-05, "loss": 0.9739, "step": 6658 }, { "epoch": 0.14, "learning_rate": 1.2304099847638601e-05, "loss": 1.0061, "step": 6659 }, { "epoch": 0.14, "learning_rate": 1.230200203535511e-05, "loss": 0.9099, "step": 6660 }, { "epoch": 0.14, "learning_rate": 1.2299904116090402e-05, "loss": 0.9828, "step": 6661 }, { "epoch": 0.14, "learning_rate": 1.229780608994198e-05, "loss": 0.9736, "step": 6662 }, { "epoch": 0.14, "learning_rate": 1.2295707957007345e-05, "loss": 0.9965, "step": 6663 }, { "epoch": 0.14, "learning_rate": 1.2293609717384002e-05, "loss": 0.9624, "step": 6664 }, { "epoch": 0.14, "learning_rate": 1.2291511371169462e-05, "loss": 1.0245, "step": 6665 }, { "epoch": 0.14, "learning_rate": 1.2289412918461246e-05, "loss": 0.9639, "step": 6666 }, { "epoch": 0.14, "learning_rate": 1.228731435935687e-05, "loss": 0.942, "step": 6667 }, { "epoch": 0.14, "learning_rate": 1.2285215693953863e-05, "loss": 0.9664, "step": 6668 }, { "epoch": 0.14, "learning_rate": 1.2283116922349758e-05, "loss": 1.0151, "step": 6669 }, { "epoch": 0.14, "learning_rate": 1.2281018044642088e-05, "loss": 0.9571, "step": 6670 }, { "epoch": 0.14, "learning_rate": 1.22789190609284e-05, "loss": 0.9624, "step": 6671 }, { "epoch": 0.14, "learning_rate": 1.2276819971306238e-05, "loss": 0.9756, "step": 6672 }, { "epoch": 0.14, "learning_rate": 1.2274720775873148e-05, "loss": 0.9201, "step": 6673 }, { "epoch": 0.14, "learning_rate": 1.2272621474726692e-05, "loss": 0.9446, "step": 6674 }, { "epoch": 0.14, "learning_rate": 1.227052206796443e-05, "loss": 0.9636, "step": 6675 }, { "epoch": 0.14, "learning_rate": 1.2268422555683925e-05, "loss": 0.9788, "step": 6676 }, { "epoch": 0.14, "learning_rate": 1.2266322937982752e-05, "loss": 0.9572, "step": 6677 }, { "epoch": 0.14, "learning_rate": 1.2264223214958486e-05, "loss": 0.9466, "step": 6678 }, { "epoch": 0.14, "learning_rate": 1.2262123386708706e-05, "loss": 1.0137, "step": 6679 }, { "epoch": 0.14, "learning_rate": 1.2260023453330997e-05, "loss": 0.9729, "step": 6680 }, { "epoch": 0.14, "learning_rate": 1.2257923414922955e-05, "loss": 0.9539, "step": 6681 }, { "epoch": 0.14, "learning_rate": 1.2255823271582166e-05, "loss": 0.9613, "step": 6682 }, { "epoch": 0.14, "learning_rate": 1.2253723023406238e-05, "loss": 0.989, "step": 6683 }, { "epoch": 0.14, "learning_rate": 1.2251622670492772e-05, "loss": 0.963, "step": 6684 }, { "epoch": 0.14, "learning_rate": 1.224952221293938e-05, "loss": 0.9886, "step": 6685 }, { "epoch": 0.14, "learning_rate": 1.2247421650843676e-05, "loss": 0.9341, "step": 6686 }, { "epoch": 0.14, "learning_rate": 1.2245320984303282e-05, "loss": 0.9402, "step": 6687 }, { "epoch": 0.14, "learning_rate": 1.2243220213415816e-05, "loss": 0.998, "step": 6688 }, { "epoch": 0.14, "learning_rate": 1.2241119338278917e-05, "loss": 0.9387, "step": 6689 }, { "epoch": 0.14, "learning_rate": 1.2239018358990209e-05, "loss": 0.964, "step": 6690 }, { "epoch": 0.15, "learning_rate": 1.2236917275647338e-05, "loss": 0.9623, "step": 6691 }, { "epoch": 0.15, "learning_rate": 1.2234816088347948e-05, "loss": 0.969, "step": 6692 }, { "epoch": 0.15, "learning_rate": 1.2232714797189682e-05, "loss": 0.9788, "step": 6693 }, { "epoch": 0.15, "learning_rate": 1.2230613402270199e-05, "loss": 0.9841, "step": 6694 }, { "epoch": 0.15, "learning_rate": 1.2228511903687158e-05, "loss": 0.9985, "step": 6695 }, { "epoch": 0.15, "learning_rate": 1.2226410301538214e-05, "loss": 0.9449, "step": 6696 }, { "epoch": 0.15, "learning_rate": 1.2224308595921044e-05, "loss": 1.0236, "step": 6697 }, { "epoch": 0.15, "learning_rate": 1.2222206786933318e-05, "loss": 0.9568, "step": 6698 }, { "epoch": 0.15, "learning_rate": 1.222010487467271e-05, "loss": 0.9671, "step": 6699 }, { "epoch": 0.15, "learning_rate": 1.2218002859236911e-05, "loss": 0.9312, "step": 6700 }, { "epoch": 0.15, "learning_rate": 1.2215900740723598e-05, "loss": 0.9578, "step": 6701 }, { "epoch": 0.15, "learning_rate": 1.2213798519230471e-05, "loss": 0.9941, "step": 6702 }, { "epoch": 0.15, "learning_rate": 1.2211696194855221e-05, "loss": 1.0373, "step": 6703 }, { "epoch": 0.15, "learning_rate": 1.2209593767695554e-05, "loss": 0.931, "step": 6704 }, { "epoch": 0.15, "learning_rate": 1.2207491237849174e-05, "loss": 0.9746, "step": 6705 }, { "epoch": 0.15, "learning_rate": 1.2205388605413793e-05, "loss": 0.9434, "step": 6706 }, { "epoch": 0.15, "learning_rate": 1.2203285870487125e-05, "loss": 0.9872, "step": 6707 }, { "epoch": 0.15, "learning_rate": 1.2201183033166893e-05, "loss": 0.9863, "step": 6708 }, { "epoch": 0.15, "learning_rate": 1.2199080093550822e-05, "loss": 1.0028, "step": 6709 }, { "epoch": 0.15, "learning_rate": 1.2196977051736643e-05, "loss": 1.0007, "step": 6710 }, { "epoch": 0.15, "learning_rate": 1.2194873907822086e-05, "loss": 0.996, "step": 6711 }, { "epoch": 0.15, "learning_rate": 1.2192770661904899e-05, "loss": 0.9678, "step": 6712 }, { "epoch": 0.15, "learning_rate": 1.2190667314082818e-05, "loss": 0.9967, "step": 6713 }, { "epoch": 0.15, "learning_rate": 1.2188563864453595e-05, "loss": 0.97, "step": 6714 }, { "epoch": 0.15, "learning_rate": 1.218646031311499e-05, "loss": 0.9781, "step": 6715 }, { "epoch": 0.15, "learning_rate": 1.2184356660164751e-05, "loss": 0.9558, "step": 6716 }, { "epoch": 0.15, "learning_rate": 1.2182252905700649e-05, "loss": 0.9973, "step": 6717 }, { "epoch": 0.15, "learning_rate": 1.2180149049820451e-05, "loss": 0.9748, "step": 6718 }, { "epoch": 0.15, "learning_rate": 1.2178045092621924e-05, "loss": 0.9357, "step": 6719 }, { "epoch": 0.15, "learning_rate": 1.2175941034202852e-05, "loss": 1.0039, "step": 6720 }, { "epoch": 0.15, "learning_rate": 1.2173836874661016e-05, "loss": 0.9484, "step": 6721 }, { "epoch": 0.15, "learning_rate": 1.2171732614094201e-05, "loss": 1.0031, "step": 6722 }, { "epoch": 0.15, "learning_rate": 1.2169628252600202e-05, "loss": 0.9621, "step": 6723 }, { "epoch": 0.15, "learning_rate": 1.2167523790276808e-05, "loss": 0.9722, "step": 6724 }, { "epoch": 0.15, "learning_rate": 1.2165419227221827e-05, "loss": 0.9525, "step": 6725 }, { "epoch": 0.15, "learning_rate": 1.216331456353306e-05, "loss": 1.0094, "step": 6726 }, { "epoch": 0.15, "learning_rate": 1.2161209799308322e-05, "loss": 0.9692, "step": 6727 }, { "epoch": 0.15, "learning_rate": 1.2159104934645425e-05, "loss": 0.9507, "step": 6728 }, { "epoch": 0.15, "learning_rate": 1.2156999969642189e-05, "loss": 0.9955, "step": 6729 }, { "epoch": 0.15, "learning_rate": 1.2154894904396437e-05, "loss": 0.9688, "step": 6730 }, { "epoch": 0.15, "learning_rate": 1.2152789739005999e-05, "loss": 0.9474, "step": 6731 }, { "epoch": 0.15, "learning_rate": 1.2150684473568713e-05, "loss": 0.9298, "step": 6732 }, { "epoch": 0.15, "learning_rate": 1.214857910818241e-05, "loss": 0.9857, "step": 6733 }, { "epoch": 0.15, "learning_rate": 1.2146473642944933e-05, "loss": 0.9463, "step": 6734 }, { "epoch": 0.15, "learning_rate": 1.2144368077954139e-05, "loss": 0.9635, "step": 6735 }, { "epoch": 0.15, "learning_rate": 1.2142262413307868e-05, "loss": 0.9773, "step": 6736 }, { "epoch": 0.15, "learning_rate": 1.2140156649103984e-05, "loss": 1.0086, "step": 6737 }, { "epoch": 0.15, "learning_rate": 1.213805078544035e-05, "loss": 0.9626, "step": 6738 }, { "epoch": 0.15, "learning_rate": 1.2135944822414825e-05, "loss": 0.9903, "step": 6739 }, { "epoch": 0.15, "learning_rate": 1.2133838760125288e-05, "loss": 0.9546, "step": 6740 }, { "epoch": 0.15, "learning_rate": 1.2131732598669608e-05, "loss": 0.9594, "step": 6741 }, { "epoch": 0.15, "learning_rate": 1.2129626338145664e-05, "loss": 0.9903, "step": 6742 }, { "epoch": 0.15, "learning_rate": 1.2127519978651346e-05, "loss": 0.9903, "step": 6743 }, { "epoch": 0.15, "learning_rate": 1.212541352028454e-05, "loss": 0.9838, "step": 6744 }, { "epoch": 0.15, "learning_rate": 1.2123306963143137e-05, "loss": 0.9595, "step": 6745 }, { "epoch": 0.15, "learning_rate": 1.2121200307325044e-05, "loss": 0.9471, "step": 6746 }, { "epoch": 0.15, "learning_rate": 1.2119093552928154e-05, "loss": 0.9658, "step": 6747 }, { "epoch": 0.15, "learning_rate": 1.2116986700050382e-05, "loss": 0.9713, "step": 6748 }, { "epoch": 0.15, "learning_rate": 1.2114879748789633e-05, "loss": 0.9722, "step": 6749 }, { "epoch": 0.15, "learning_rate": 1.2112772699243832e-05, "loss": 0.9844, "step": 6750 }, { "epoch": 0.15, "learning_rate": 1.2110665551510892e-05, "loss": 0.9453, "step": 6751 }, { "epoch": 0.15, "learning_rate": 1.2108558305688742e-05, "loss": 1.0035, "step": 6752 }, { "epoch": 0.15, "learning_rate": 1.2106450961875314e-05, "loss": 1.0306, "step": 6753 }, { "epoch": 0.15, "learning_rate": 1.210434352016854e-05, "loss": 0.9703, "step": 6754 }, { "epoch": 0.15, "learning_rate": 1.2102235980666362e-05, "loss": 0.97, "step": 6755 }, { "epoch": 0.15, "learning_rate": 1.2100128343466723e-05, "loss": 1.0012, "step": 6756 }, { "epoch": 0.15, "learning_rate": 1.2098020608667569e-05, "loss": 1.0219, "step": 6757 }, { "epoch": 0.15, "learning_rate": 1.2095912776366859e-05, "loss": 0.9932, "step": 6758 }, { "epoch": 0.15, "learning_rate": 1.2093804846662542e-05, "loss": 0.9889, "step": 6759 }, { "epoch": 0.15, "learning_rate": 1.2091696819652586e-05, "loss": 0.9707, "step": 6760 }, { "epoch": 0.15, "learning_rate": 1.2089588695434957e-05, "loss": 1.0052, "step": 6761 }, { "epoch": 0.15, "learning_rate": 1.2087480474107625e-05, "loss": 0.9664, "step": 6762 }, { "epoch": 0.15, "learning_rate": 1.208537215576857e-05, "loss": 0.9683, "step": 6763 }, { "epoch": 0.15, "learning_rate": 1.2083263740515764e-05, "loss": 0.915, "step": 6764 }, { "epoch": 0.15, "learning_rate": 1.2081155228447194e-05, "loss": 0.9955, "step": 6765 }, { "epoch": 0.15, "learning_rate": 1.2079046619660854e-05, "loss": 0.99, "step": 6766 }, { "epoch": 0.15, "learning_rate": 1.2076937914254736e-05, "loss": 0.945, "step": 6767 }, { "epoch": 0.15, "learning_rate": 1.2074829112326834e-05, "loss": 0.9546, "step": 6768 }, { "epoch": 0.15, "learning_rate": 1.2072720213975151e-05, "loss": 0.9568, "step": 6769 }, { "epoch": 0.15, "learning_rate": 1.20706112192977e-05, "loss": 0.9003, "step": 6770 }, { "epoch": 0.15, "learning_rate": 1.206850212839249e-05, "loss": 0.9863, "step": 6771 }, { "epoch": 0.15, "learning_rate": 1.2066392941357532e-05, "loss": 0.9563, "step": 6772 }, { "epoch": 0.15, "learning_rate": 1.2064283658290855e-05, "loss": 0.9581, "step": 6773 }, { "epoch": 0.15, "learning_rate": 1.2062174279290475e-05, "loss": 0.979, "step": 6774 }, { "epoch": 0.15, "learning_rate": 1.206006480445443e-05, "loss": 0.9756, "step": 6775 }, { "epoch": 0.15, "learning_rate": 1.2057955233880748e-05, "loss": 0.9678, "step": 6776 }, { "epoch": 0.15, "learning_rate": 1.2055845567667465e-05, "loss": 0.9554, "step": 6777 }, { "epoch": 0.15, "learning_rate": 1.2053735805912634e-05, "loss": 0.9653, "step": 6778 }, { "epoch": 0.15, "learning_rate": 1.2051625948714294e-05, "loss": 0.9911, "step": 6779 }, { "epoch": 0.15, "learning_rate": 1.2049515996170497e-05, "loss": 1.0052, "step": 6780 }, { "epoch": 0.15, "learning_rate": 1.2047405948379302e-05, "loss": 0.9522, "step": 6781 }, { "epoch": 0.15, "learning_rate": 1.2045295805438769e-05, "loss": 0.9949, "step": 6782 }, { "epoch": 0.15, "learning_rate": 1.2043185567446957e-05, "loss": 0.9929, "step": 6783 }, { "epoch": 0.15, "learning_rate": 1.2041075234501946e-05, "loss": 0.9938, "step": 6784 }, { "epoch": 0.15, "learning_rate": 1.2038964806701803e-05, "loss": 0.959, "step": 6785 }, { "epoch": 0.15, "learning_rate": 1.2036854284144608e-05, "loss": 0.9125, "step": 6786 }, { "epoch": 0.15, "learning_rate": 1.2034743666928443e-05, "loss": 0.9692, "step": 6787 }, { "epoch": 0.15, "learning_rate": 1.2032632955151394e-05, "loss": 0.9273, "step": 6788 }, { "epoch": 0.15, "learning_rate": 1.2030522148911553e-05, "loss": 0.9843, "step": 6789 }, { "epoch": 0.15, "learning_rate": 1.2028411248307016e-05, "loss": 0.9566, "step": 6790 }, { "epoch": 0.15, "learning_rate": 1.2026300253435883e-05, "loss": 0.9739, "step": 6791 }, { "epoch": 0.15, "learning_rate": 1.2024189164396258e-05, "loss": 1.0136, "step": 6792 }, { "epoch": 0.15, "learning_rate": 1.2022077981286252e-05, "loss": 1.0066, "step": 6793 }, { "epoch": 0.15, "learning_rate": 1.2019966704203979e-05, "loss": 0.9783, "step": 6794 }, { "epoch": 0.15, "learning_rate": 1.201785533324755e-05, "loss": 0.984, "step": 6795 }, { "epoch": 0.15, "learning_rate": 1.2015743868515095e-05, "loss": 1.0186, "step": 6796 }, { "epoch": 0.15, "learning_rate": 1.2013632310104732e-05, "loss": 0.9952, "step": 6797 }, { "epoch": 0.15, "learning_rate": 1.20115206581146e-05, "loss": 0.923, "step": 6798 }, { "epoch": 0.15, "learning_rate": 1.2009408912642831e-05, "loss": 1.0193, "step": 6799 }, { "epoch": 0.15, "learning_rate": 1.2007297073787562e-05, "loss": 0.9611, "step": 6800 }, { "epoch": 0.15, "learning_rate": 1.2005185141646942e-05, "loss": 1.0154, "step": 6801 }, { "epoch": 0.15, "learning_rate": 1.2003073116319114e-05, "loss": 0.9729, "step": 6802 }, { "epoch": 0.15, "learning_rate": 1.200096099790223e-05, "loss": 0.9709, "step": 6803 }, { "epoch": 0.15, "learning_rate": 1.1998848786494453e-05, "loss": 0.9716, "step": 6804 }, { "epoch": 0.15, "learning_rate": 1.1996736482193936e-05, "loss": 0.9552, "step": 6805 }, { "epoch": 0.15, "learning_rate": 1.199462408509885e-05, "loss": 0.974, "step": 6806 }, { "epoch": 0.15, "learning_rate": 1.1992511595307363e-05, "loss": 0.9828, "step": 6807 }, { "epoch": 0.15, "learning_rate": 1.1990399012917649e-05, "loss": 0.9507, "step": 6808 }, { "epoch": 0.15, "learning_rate": 1.1988286338027888e-05, "loss": 0.9525, "step": 6809 }, { "epoch": 0.15, "learning_rate": 1.198617357073626e-05, "loss": 0.96, "step": 6810 }, { "epoch": 0.15, "learning_rate": 1.1984060711140951e-05, "loss": 0.9487, "step": 6811 }, { "epoch": 0.15, "learning_rate": 1.1981947759340153e-05, "loss": 0.9761, "step": 6812 }, { "epoch": 0.15, "learning_rate": 1.1979834715432064e-05, "loss": 1.0144, "step": 6813 }, { "epoch": 0.15, "learning_rate": 1.1977721579514884e-05, "loss": 1.001, "step": 6814 }, { "epoch": 0.15, "learning_rate": 1.1975608351686809e-05, "loss": 0.9723, "step": 6815 }, { "epoch": 0.15, "learning_rate": 1.1973495032046061e-05, "loss": 0.9754, "step": 6816 }, { "epoch": 0.15, "learning_rate": 1.1971381620690843e-05, "loss": 1.0265, "step": 6817 }, { "epoch": 0.15, "learning_rate": 1.196926811771937e-05, "loss": 0.9475, "step": 6818 }, { "epoch": 0.15, "learning_rate": 1.1967154523229872e-05, "loss": 0.9725, "step": 6819 }, { "epoch": 0.15, "learning_rate": 1.1965040837320565e-05, "loss": 0.9833, "step": 6820 }, { "epoch": 0.15, "learning_rate": 1.1962927060089682e-05, "loss": 0.9642, "step": 6821 }, { "epoch": 0.15, "learning_rate": 1.1960813191635461e-05, "loss": 1.0204, "step": 6822 }, { "epoch": 0.15, "learning_rate": 1.1958699232056135e-05, "loss": 0.9745, "step": 6823 }, { "epoch": 0.15, "learning_rate": 1.1956585181449948e-05, "loss": 0.9963, "step": 6824 }, { "epoch": 0.15, "learning_rate": 1.195447103991515e-05, "loss": 1.0059, "step": 6825 }, { "epoch": 0.15, "learning_rate": 1.1952356807549986e-05, "loss": 0.9452, "step": 6826 }, { "epoch": 0.15, "learning_rate": 1.1950242484452713e-05, "loss": 0.9926, "step": 6827 }, { "epoch": 0.15, "learning_rate": 1.1948128070721588e-05, "loss": 0.9661, "step": 6828 }, { "epoch": 0.15, "learning_rate": 1.1946013566454877e-05, "loss": 0.9668, "step": 6829 }, { "epoch": 0.15, "learning_rate": 1.194389897175085e-05, "loss": 1.0087, "step": 6830 }, { "epoch": 0.15, "learning_rate": 1.1941784286707774e-05, "loss": 0.9799, "step": 6831 }, { "epoch": 0.15, "learning_rate": 1.1939669511423928e-05, "loss": 0.9706, "step": 6832 }, { "epoch": 0.15, "learning_rate": 1.1937554645997593e-05, "loss": 0.9984, "step": 6833 }, { "epoch": 0.15, "learning_rate": 1.1935439690527048e-05, "loss": 1.0167, "step": 6834 }, { "epoch": 0.15, "learning_rate": 1.193332464511059e-05, "loss": 0.9896, "step": 6835 }, { "epoch": 0.15, "learning_rate": 1.1931209509846503e-05, "loss": 0.9591, "step": 6836 }, { "epoch": 0.15, "learning_rate": 1.192909428483309e-05, "loss": 0.9842, "step": 6837 }, { "epoch": 0.15, "learning_rate": 1.192697897016865e-05, "loss": 0.9566, "step": 6838 }, { "epoch": 0.15, "learning_rate": 1.1924863565951488e-05, "loss": 1.0163, "step": 6839 }, { "epoch": 0.15, "learning_rate": 1.1922748072279917e-05, "loss": 0.9695, "step": 6840 }, { "epoch": 0.16, "learning_rate": 1.1920632489252242e-05, "loss": 0.9548, "step": 6841 }, { "epoch": 0.16, "learning_rate": 1.191851681696679e-05, "loss": 0.9519, "step": 6842 }, { "epoch": 0.16, "learning_rate": 1.1916401055521877e-05, "loss": 0.99, "step": 6843 }, { "epoch": 0.16, "learning_rate": 1.1914285205015832e-05, "loss": 0.9728, "step": 6844 }, { "epoch": 0.16, "learning_rate": 1.1912169265546985e-05, "loss": 0.9947, "step": 6845 }, { "epoch": 0.16, "learning_rate": 1.1910053237213669e-05, "loss": 1.004, "step": 6846 }, { "epoch": 0.16, "learning_rate": 1.1907937120114223e-05, "loss": 0.9924, "step": 6847 }, { "epoch": 0.16, "learning_rate": 1.1905820914346991e-05, "loss": 0.9699, "step": 6848 }, { "epoch": 0.16, "learning_rate": 1.1903704620010315e-05, "loss": 0.9258, "step": 6849 }, { "epoch": 0.16, "learning_rate": 1.1901588237202551e-05, "loss": 0.9594, "step": 6850 }, { "epoch": 0.16, "learning_rate": 1.1899471766022053e-05, "loss": 0.9827, "step": 6851 }, { "epoch": 0.16, "learning_rate": 1.1897355206567176e-05, "loss": 1.0126, "step": 6852 }, { "epoch": 0.16, "learning_rate": 1.1895238558936288e-05, "loss": 0.973, "step": 6853 }, { "epoch": 0.16, "learning_rate": 1.1893121823227754e-05, "loss": 0.9396, "step": 6854 }, { "epoch": 0.16, "learning_rate": 1.1891004999539946e-05, "loss": 0.9703, "step": 6855 }, { "epoch": 0.16, "learning_rate": 1.188888808797124e-05, "loss": 1.0014, "step": 6856 }, { "epoch": 0.16, "learning_rate": 1.1886771088620012e-05, "loss": 0.9699, "step": 6857 }, { "epoch": 0.16, "learning_rate": 1.1884654001584649e-05, "loss": 0.9553, "step": 6858 }, { "epoch": 0.16, "learning_rate": 1.1882536826963537e-05, "loss": 0.9967, "step": 6859 }, { "epoch": 0.16, "learning_rate": 1.188041956485507e-05, "loss": 0.9323, "step": 6860 }, { "epoch": 0.16, "learning_rate": 1.187830221535764e-05, "loss": 0.9902, "step": 6861 }, { "epoch": 0.16, "learning_rate": 1.1876184778569652e-05, "loss": 0.9656, "step": 6862 }, { "epoch": 0.16, "learning_rate": 1.1874067254589505e-05, "loss": 0.964, "step": 6863 }, { "epoch": 0.16, "learning_rate": 1.187194964351561e-05, "loss": 0.9871, "step": 6864 }, { "epoch": 0.16, "learning_rate": 1.1869831945446377e-05, "loss": 0.9787, "step": 6865 }, { "epoch": 0.16, "learning_rate": 1.186771416048022e-05, "loss": 0.9811, "step": 6866 }, { "epoch": 0.16, "learning_rate": 1.1865596288715565e-05, "loss": 0.9744, "step": 6867 }, { "epoch": 0.16, "learning_rate": 1.1863478330250835e-05, "loss": 0.9466, "step": 6868 }, { "epoch": 0.16, "learning_rate": 1.1861360285184454e-05, "loss": 0.9694, "step": 6869 }, { "epoch": 0.16, "learning_rate": 1.1859242153614856e-05, "loss": 0.9697, "step": 6870 }, { "epoch": 0.16, "learning_rate": 1.185712393564048e-05, "loss": 1.0078, "step": 6871 }, { "epoch": 0.16, "learning_rate": 1.185500563135976e-05, "loss": 0.9724, "step": 6872 }, { "epoch": 0.16, "learning_rate": 1.1852887240871145e-05, "loss": 0.9436, "step": 6873 }, { "epoch": 0.16, "learning_rate": 1.1850768764273082e-05, "loss": 0.9532, "step": 6874 }, { "epoch": 0.16, "learning_rate": 1.1848650201664022e-05, "loss": 0.9765, "step": 6875 }, { "epoch": 0.16, "learning_rate": 1.1846531553142426e-05, "loss": 0.9746, "step": 6876 }, { "epoch": 0.16, "learning_rate": 1.1844412818806749e-05, "loss": 0.9796, "step": 6877 }, { "epoch": 0.16, "learning_rate": 1.1842293998755457e-05, "loss": 0.9979, "step": 6878 }, { "epoch": 0.16, "learning_rate": 1.184017509308702e-05, "loss": 0.9884, "step": 6879 }, { "epoch": 0.16, "learning_rate": 1.1838056101899906e-05, "loss": 0.9573, "step": 6880 }, { "epoch": 0.16, "learning_rate": 1.1835937025292593e-05, "loss": 0.9768, "step": 6881 }, { "epoch": 0.16, "learning_rate": 1.1833817863363563e-05, "loss": 0.9772, "step": 6882 }, { "epoch": 0.16, "learning_rate": 1.1831698616211296e-05, "loss": 0.9391, "step": 6883 }, { "epoch": 0.16, "learning_rate": 1.1829579283934286e-05, "loss": 0.9583, "step": 6884 }, { "epoch": 0.16, "learning_rate": 1.1827459866631019e-05, "loss": 0.9807, "step": 6885 }, { "epoch": 0.16, "learning_rate": 1.1825340364399996e-05, "loss": 0.9752, "step": 6886 }, { "epoch": 0.16, "learning_rate": 1.1823220777339712e-05, "loss": 0.9914, "step": 6887 }, { "epoch": 0.16, "learning_rate": 1.1821101105548676e-05, "loss": 0.9641, "step": 6888 }, { "epoch": 0.16, "learning_rate": 1.181898134912539e-05, "loss": 0.972, "step": 6889 }, { "epoch": 0.16, "learning_rate": 1.1816861508168367e-05, "loss": 0.986, "step": 6890 }, { "epoch": 0.16, "learning_rate": 1.1814741582776128e-05, "loss": 0.9712, "step": 6891 }, { "epoch": 0.16, "learning_rate": 1.1812621573047188e-05, "loss": 0.9895, "step": 6892 }, { "epoch": 0.16, "learning_rate": 1.1810501479080072e-05, "loss": 0.9893, "step": 6893 }, { "epoch": 0.16, "learning_rate": 1.1808381300973307e-05, "loss": 0.9765, "step": 6894 }, { "epoch": 0.16, "learning_rate": 1.180626103882542e-05, "loss": 0.9999, "step": 6895 }, { "epoch": 0.16, "learning_rate": 1.1804140692734951e-05, "loss": 0.9575, "step": 6896 }, { "epoch": 0.16, "learning_rate": 1.1802020262800442e-05, "loss": 0.9991, "step": 6897 }, { "epoch": 0.16, "learning_rate": 1.1799899749120429e-05, "loss": 1.0024, "step": 6898 }, { "epoch": 0.16, "learning_rate": 1.1797779151793463e-05, "loss": 1.0189, "step": 6899 }, { "epoch": 0.16, "learning_rate": 1.1795658470918094e-05, "loss": 0.9564, "step": 6900 }, { "epoch": 0.16, "learning_rate": 1.1793537706592875e-05, "loss": 0.9964, "step": 6901 }, { "epoch": 0.16, "learning_rate": 1.179141685891637e-05, "loss": 0.9885, "step": 6902 }, { "epoch": 0.16, "learning_rate": 1.1789295927987134e-05, "loss": 0.9768, "step": 6903 }, { "epoch": 0.16, "learning_rate": 1.1787174913903735e-05, "loss": 0.9772, "step": 6904 }, { "epoch": 0.16, "learning_rate": 1.1785053816764752e-05, "loss": 1.0191, "step": 6905 }, { "epoch": 0.16, "learning_rate": 1.1782932636668745e-05, "loss": 1.0101, "step": 6906 }, { "epoch": 0.16, "learning_rate": 1.17808113737143e-05, "loss": 0.9588, "step": 6907 }, { "epoch": 0.16, "learning_rate": 1.1778690028e-05, "loss": 0.9954, "step": 6908 }, { "epoch": 0.16, "learning_rate": 1.1776568599624428e-05, "loss": 1.0287, "step": 6909 }, { "epoch": 0.16, "learning_rate": 1.177444708868617e-05, "loss": 0.9839, "step": 6910 }, { "epoch": 0.16, "learning_rate": 1.1772325495283824e-05, "loss": 0.9713, "step": 6911 }, { "epoch": 0.16, "learning_rate": 1.1770203819515987e-05, "loss": 0.9587, "step": 6912 }, { "epoch": 0.16, "learning_rate": 1.1768082061481257e-05, "loss": 0.9928, "step": 6913 }, { "epoch": 0.16, "learning_rate": 1.1765960221278241e-05, "loss": 0.9598, "step": 6914 }, { "epoch": 0.16, "learning_rate": 1.1763838299005546e-05, "loss": 0.9862, "step": 6915 }, { "epoch": 0.16, "learning_rate": 1.1761716294761787e-05, "loss": 0.9673, "step": 6916 }, { "epoch": 0.16, "learning_rate": 1.1759594208645577e-05, "loss": 0.9492, "step": 6917 }, { "epoch": 0.16, "learning_rate": 1.1757472040755535e-05, "loss": 0.9649, "step": 6918 }, { "epoch": 0.16, "learning_rate": 1.1755349791190289e-05, "loss": 0.967, "step": 6919 }, { "epoch": 0.16, "learning_rate": 1.1753227460048466e-05, "loss": 0.9682, "step": 6920 }, { "epoch": 0.16, "learning_rate": 1.1751105047428695e-05, "loss": 0.977, "step": 6921 }, { "epoch": 0.16, "learning_rate": 1.1748982553429611e-05, "loss": 1.0124, "step": 6922 }, { "epoch": 0.16, "learning_rate": 1.1746859978149853e-05, "loss": 0.9653, "step": 6923 }, { "epoch": 0.16, "learning_rate": 1.1744737321688067e-05, "loss": 0.9636, "step": 6924 }, { "epoch": 0.16, "learning_rate": 1.1742614584142897e-05, "loss": 1.0123, "step": 6925 }, { "epoch": 0.16, "learning_rate": 1.1740491765612994e-05, "loss": 0.9718, "step": 6926 }, { "epoch": 0.16, "learning_rate": 1.173836886619701e-05, "loss": 1.0105, "step": 6927 }, { "epoch": 0.16, "learning_rate": 1.1736245885993604e-05, "loss": 0.9227, "step": 6928 }, { "epoch": 0.16, "learning_rate": 1.1734122825101437e-05, "loss": 1.0036, "step": 6929 }, { "epoch": 0.16, "learning_rate": 1.1731999683619173e-05, "loss": 0.9619, "step": 6930 }, { "epoch": 0.16, "learning_rate": 1.1729876461645486e-05, "loss": 0.9789, "step": 6931 }, { "epoch": 0.16, "learning_rate": 1.1727753159279048e-05, "loss": 0.9529, "step": 6932 }, { "epoch": 0.16, "learning_rate": 1.1725629776618528e-05, "loss": 0.9843, "step": 6933 }, { "epoch": 0.16, "learning_rate": 1.1723506313762615e-05, "loss": 0.9473, "step": 6934 }, { "epoch": 0.16, "learning_rate": 1.1721382770809986e-05, "loss": 0.9496, "step": 6935 }, { "epoch": 0.16, "learning_rate": 1.1719259147859333e-05, "loss": 0.9453, "step": 6936 }, { "epoch": 0.16, "learning_rate": 1.171713544500935e-05, "loss": 1.0183, "step": 6937 }, { "epoch": 0.16, "learning_rate": 1.1715011662358722e-05, "loss": 0.974, "step": 6938 }, { "epoch": 0.16, "learning_rate": 1.1712887800006158e-05, "loss": 0.9575, "step": 6939 }, { "epoch": 0.16, "learning_rate": 1.1710763858050358e-05, "loss": 0.9737, "step": 6940 }, { "epoch": 0.16, "learning_rate": 1.1708639836590024e-05, "loss": 0.9872, "step": 6941 }, { "epoch": 0.16, "learning_rate": 1.170651573572387e-05, "loss": 0.9912, "step": 6942 }, { "epoch": 0.16, "learning_rate": 1.1704391555550607e-05, "loss": 1.0118, "step": 6943 }, { "epoch": 0.16, "learning_rate": 1.1702267296168954e-05, "loss": 0.9801, "step": 6944 }, { "epoch": 0.16, "learning_rate": 1.1700142957677634e-05, "loss": 0.9526, "step": 6945 }, { "epoch": 0.16, "learning_rate": 1.1698018540175367e-05, "loss": 0.9453, "step": 6946 }, { "epoch": 0.16, "learning_rate": 1.1695894043760883e-05, "loss": 0.9855, "step": 6947 }, { "epoch": 0.16, "learning_rate": 1.1693769468532915e-05, "loss": 0.9771, "step": 6948 }, { "epoch": 0.16, "learning_rate": 1.1691644814590202e-05, "loss": 0.9931, "step": 6949 }, { "epoch": 0.16, "learning_rate": 1.1689520082031475e-05, "loss": 0.9887, "step": 6950 }, { "epoch": 0.16, "learning_rate": 1.1687395270955485e-05, "loss": 1.0109, "step": 6951 }, { "epoch": 0.16, "learning_rate": 1.1685270381460972e-05, "loss": 0.9916, "step": 6952 }, { "epoch": 0.16, "learning_rate": 1.1683145413646688e-05, "loss": 0.9983, "step": 6953 }, { "epoch": 0.16, "learning_rate": 1.1681020367611391e-05, "loss": 0.9579, "step": 6954 }, { "epoch": 0.16, "learning_rate": 1.1678895243453836e-05, "loss": 0.9792, "step": 6955 }, { "epoch": 0.16, "learning_rate": 1.1676770041272783e-05, "loss": 0.9653, "step": 6956 }, { "epoch": 0.16, "learning_rate": 1.1674644761166999e-05, "loss": 0.9423, "step": 6957 }, { "epoch": 0.16, "learning_rate": 1.1672519403235248e-05, "loss": 1.0113, "step": 6958 }, { "epoch": 0.16, "learning_rate": 1.1670393967576303e-05, "loss": 0.9769, "step": 6959 }, { "epoch": 0.16, "learning_rate": 1.1668268454288947e-05, "loss": 0.9733, "step": 6960 }, { "epoch": 0.16, "learning_rate": 1.1666142863471949e-05, "loss": 0.936, "step": 6961 }, { "epoch": 0.16, "learning_rate": 1.1664017195224098e-05, "loss": 1.0145, "step": 6962 }, { "epoch": 0.16, "learning_rate": 1.1661891449644185e-05, "loss": 0.971, "step": 6963 }, { "epoch": 0.16, "learning_rate": 1.1659765626830987e-05, "loss": 0.9883, "step": 6964 }, { "epoch": 0.16, "learning_rate": 1.1657639726883306e-05, "loss": 0.9533, "step": 6965 }, { "epoch": 0.16, "learning_rate": 1.165551374989994e-05, "loss": 0.9958, "step": 6966 }, { "epoch": 0.16, "learning_rate": 1.1653387695979685e-05, "loss": 0.9366, "step": 6967 }, { "epoch": 0.16, "learning_rate": 1.1651261565221352e-05, "loss": 1.0102, "step": 6968 }, { "epoch": 0.16, "learning_rate": 1.1649135357723742e-05, "loss": 0.9786, "step": 6969 }, { "epoch": 0.16, "learning_rate": 1.1647009073585669e-05, "loss": 0.9667, "step": 6970 }, { "epoch": 0.16, "learning_rate": 1.164488271290595e-05, "loss": 0.9812, "step": 6971 }, { "epoch": 0.16, "learning_rate": 1.1642756275783405e-05, "loss": 0.9879, "step": 6972 }, { "epoch": 0.16, "learning_rate": 1.1640629762316848e-05, "loss": 0.9512, "step": 6973 }, { "epoch": 0.16, "learning_rate": 1.1638503172605113e-05, "loss": 0.9596, "step": 6974 }, { "epoch": 0.16, "learning_rate": 1.1636376506747024e-05, "loss": 0.9726, "step": 6975 }, { "epoch": 0.16, "learning_rate": 1.1634249764841417e-05, "loss": 0.965, "step": 6976 }, { "epoch": 0.16, "learning_rate": 1.163212294698713e-05, "loss": 0.9455, "step": 6977 }, { "epoch": 0.16, "learning_rate": 1.1629996053282998e-05, "loss": 1.0241, "step": 6978 }, { "epoch": 0.16, "learning_rate": 1.1627869083827865e-05, "loss": 0.9827, "step": 6979 }, { "epoch": 0.16, "learning_rate": 1.1625742038720584e-05, "loss": 0.9363, "step": 6980 }, { "epoch": 0.16, "learning_rate": 1.1623614918059997e-05, "loss": 0.9154, "step": 6981 }, { "epoch": 0.16, "learning_rate": 1.1621487721944961e-05, "loss": 0.977, "step": 6982 }, { "epoch": 0.16, "learning_rate": 1.1619360450474339e-05, "loss": 0.969, "step": 6983 }, { "epoch": 0.16, "learning_rate": 1.1617233103746984e-05, "loss": 0.9878, "step": 6984 }, { "epoch": 0.16, "learning_rate": 1.1615105681861762e-05, "loss": 1.0314, "step": 6985 }, { "epoch": 0.16, "learning_rate": 1.1612978184917549e-05, "loss": 0.9811, "step": 6986 }, { "epoch": 0.16, "learning_rate": 1.1610850613013205e-05, "loss": 0.9632, "step": 6987 }, { "epoch": 0.16, "learning_rate": 1.1608722966247611e-05, "loss": 0.9258, "step": 6988 }, { "epoch": 0.16, "learning_rate": 1.1606595244719644e-05, "loss": 1.0034, "step": 6989 }, { "epoch": 0.16, "learning_rate": 1.1604467448528188e-05, "loss": 0.9571, "step": 6990 }, { "epoch": 0.17, "learning_rate": 1.1602339577772126e-05, "loss": 0.9472, "step": 6991 }, { "epoch": 0.17, "learning_rate": 1.1600211632550344e-05, "loss": 0.9606, "step": 6992 }, { "epoch": 0.17, "learning_rate": 1.1598083612961738e-05, "loss": 0.9814, "step": 6993 }, { "epoch": 0.17, "learning_rate": 1.1595955519105208e-05, "loss": 1.0016, "step": 6994 }, { "epoch": 0.17, "learning_rate": 1.1593827351079647e-05, "loss": 0.9874, "step": 6995 }, { "epoch": 0.17, "learning_rate": 1.1591699108983957e-05, "loss": 0.9344, "step": 6996 }, { "epoch": 0.17, "learning_rate": 1.1589570792917047e-05, "loss": 0.9604, "step": 6997 }, { "epoch": 0.17, "learning_rate": 1.1587442402977824e-05, "loss": 1.0025, "step": 6998 }, { "epoch": 0.17, "learning_rate": 1.1585313939265204e-05, "loss": 0.9594, "step": 6999 }, { "epoch": 0.17, "learning_rate": 1.15831854018781e-05, "loss": 1.0209, "step": 7000 }, { "epoch": 0.17, "learning_rate": 1.1581056790915437e-05, "loss": 0.9814, "step": 7001 }, { "epoch": 0.17, "learning_rate": 1.1578928106476133e-05, "loss": 1.0031, "step": 7002 }, { "epoch": 0.17, "learning_rate": 1.1576799348659118e-05, "loss": 0.9726, "step": 7003 }, { "epoch": 0.17, "learning_rate": 1.1574670517563315e-05, "loss": 0.9491, "step": 7004 }, { "epoch": 0.17, "learning_rate": 1.1572541613287667e-05, "loss": 0.9756, "step": 7005 }, { "epoch": 0.17, "learning_rate": 1.1570412635931108e-05, "loss": 1.026, "step": 7006 }, { "epoch": 0.17, "learning_rate": 1.1568283585592573e-05, "loss": 0.9802, "step": 7007 }, { "epoch": 0.17, "learning_rate": 1.156615446237101e-05, "loss": 0.9469, "step": 7008 }, { "epoch": 0.17, "learning_rate": 1.156402526636537e-05, "loss": 0.9401, "step": 7009 }, { "epoch": 0.17, "learning_rate": 1.1561895997674594e-05, "loss": 0.9587, "step": 7010 }, { "epoch": 0.17, "learning_rate": 1.1559766656397641e-05, "loss": 0.9976, "step": 7011 }, { "epoch": 0.17, "learning_rate": 1.155763724263347e-05, "loss": 0.9806, "step": 7012 }, { "epoch": 0.17, "learning_rate": 1.1555507756481036e-05, "loss": 0.9625, "step": 7013 }, { "epoch": 0.17, "learning_rate": 1.1553378198039308e-05, "loss": 1.0025, "step": 7014 }, { "epoch": 0.17, "learning_rate": 1.1551248567407254e-05, "loss": 1.0016, "step": 7015 }, { "epoch": 0.17, "learning_rate": 1.1549118864683837e-05, "loss": 0.9941, "step": 7016 }, { "epoch": 0.17, "learning_rate": 1.154698908996804e-05, "loss": 0.9325, "step": 7017 }, { "epoch": 0.17, "learning_rate": 1.1544859243358835e-05, "loss": 0.9889, "step": 7018 }, { "epoch": 0.17, "learning_rate": 1.1542729324955201e-05, "loss": 0.9412, "step": 7019 }, { "epoch": 0.17, "learning_rate": 1.1540599334856129e-05, "loss": 0.9942, "step": 7020 }, { "epoch": 0.17, "learning_rate": 1.1538469273160597e-05, "loss": 1.0449, "step": 7021 }, { "epoch": 0.17, "learning_rate": 1.1536339139967604e-05, "loss": 0.9677, "step": 7022 }, { "epoch": 0.17, "learning_rate": 1.1534208935376138e-05, "loss": 0.9999, "step": 7023 }, { "epoch": 0.17, "learning_rate": 1.1532078659485202e-05, "loss": 0.9852, "step": 7024 }, { "epoch": 0.17, "learning_rate": 1.152994831239379e-05, "loss": 0.9997, "step": 7025 }, { "epoch": 0.17, "learning_rate": 1.1527817894200912e-05, "loss": 0.9887, "step": 7026 }, { "epoch": 0.17, "learning_rate": 1.152568740500557e-05, "loss": 0.9578, "step": 7027 }, { "epoch": 0.17, "learning_rate": 1.1523556844906777e-05, "loss": 0.9526, "step": 7028 }, { "epoch": 0.17, "learning_rate": 1.1521426214003546e-05, "loss": 0.9761, "step": 7029 }, { "epoch": 0.17, "learning_rate": 1.1519295512394895e-05, "loss": 0.9426, "step": 7030 }, { "epoch": 0.17, "learning_rate": 1.1517164740179843e-05, "loss": 0.9981, "step": 7031 }, { "epoch": 0.17, "learning_rate": 1.151503389745742e-05, "loss": 1.0351, "step": 7032 }, { "epoch": 0.17, "learning_rate": 1.1512902984326644e-05, "loss": 0.9669, "step": 7033 }, { "epoch": 0.17, "learning_rate": 1.1510772000886545e-05, "loss": 0.9774, "step": 7034 }, { "epoch": 0.17, "learning_rate": 1.1508640947236165e-05, "loss": 0.9614, "step": 7035 }, { "epoch": 0.17, "learning_rate": 1.1506509823474534e-05, "loss": 0.9883, "step": 7036 }, { "epoch": 0.17, "learning_rate": 1.1504378629700693e-05, "loss": 0.9832, "step": 7037 }, { "epoch": 0.17, "learning_rate": 1.1502247366013689e-05, "loss": 0.9732, "step": 7038 }, { "epoch": 0.17, "learning_rate": 1.1500116032512562e-05, "loss": 1.0135, "step": 7039 }, { "epoch": 0.17, "learning_rate": 1.1497984629296367e-05, "loss": 0.9891, "step": 7040 }, { "epoch": 0.17, "learning_rate": 1.1495853156464156e-05, "loss": 0.9632, "step": 7041 }, { "epoch": 0.17, "learning_rate": 1.1493721614114983e-05, "loss": 0.9627, "step": 7042 }, { "epoch": 0.17, "learning_rate": 1.149159000234791e-05, "loss": 0.9579, "step": 7043 }, { "epoch": 0.17, "learning_rate": 1.1489458321261996e-05, "loss": 0.9605, "step": 7044 }, { "epoch": 0.17, "learning_rate": 1.148732657095631e-05, "loss": 0.9395, "step": 7045 }, { "epoch": 0.17, "learning_rate": 1.1485194751529921e-05, "loss": 1.0102, "step": 7046 }, { "epoch": 0.17, "learning_rate": 1.1483062863081904e-05, "loss": 0.9784, "step": 7047 }, { "epoch": 0.17, "learning_rate": 1.1480930905711327e-05, "loss": 1.0055, "step": 7048 }, { "epoch": 0.17, "learning_rate": 1.1478798879517275e-05, "loss": 0.9288, "step": 7049 }, { "epoch": 0.17, "learning_rate": 1.1476666784598825e-05, "loss": 0.938, "step": 7050 }, { "epoch": 0.17, "learning_rate": 1.1474534621055065e-05, "loss": 0.9616, "step": 7051 }, { "epoch": 0.17, "learning_rate": 1.1472402388985086e-05, "loss": 0.9792, "step": 7052 }, { "epoch": 0.17, "learning_rate": 1.1470270088487974e-05, "loss": 0.9644, "step": 7053 }, { "epoch": 0.17, "learning_rate": 1.1468137719662827e-05, "loss": 0.9764, "step": 7054 }, { "epoch": 0.17, "learning_rate": 1.1466005282608745e-05, "loss": 0.9903, "step": 7055 }, { "epoch": 0.17, "learning_rate": 1.1463872777424822e-05, "loss": 0.9537, "step": 7056 }, { "epoch": 0.17, "learning_rate": 1.1461740204210167e-05, "loss": 0.9747, "step": 7057 }, { "epoch": 0.17, "learning_rate": 1.145960756306389e-05, "loss": 0.9728, "step": 7058 }, { "epoch": 0.17, "learning_rate": 1.1457474854085095e-05, "loss": 0.9578, "step": 7059 }, { "epoch": 0.17, "learning_rate": 1.1455342077372897e-05, "loss": 0.9719, "step": 7060 }, { "epoch": 0.17, "learning_rate": 1.145320923302642e-05, "loss": 1.0053, "step": 7061 }, { "epoch": 0.17, "learning_rate": 1.1451076321144774e-05, "loss": 0.9519, "step": 7062 }, { "epoch": 0.17, "learning_rate": 1.144894334182709e-05, "loss": 0.9581, "step": 7063 }, { "epoch": 0.17, "learning_rate": 1.144681029517249e-05, "loss": 0.9811, "step": 7064 }, { "epoch": 0.17, "learning_rate": 1.1444677181280102e-05, "loss": 0.9693, "step": 7065 }, { "epoch": 0.17, "learning_rate": 1.1442544000249058e-05, "loss": 1.0072, "step": 7066 }, { "epoch": 0.17, "learning_rate": 1.1440410752178502e-05, "loss": 1.0008, "step": 7067 }, { "epoch": 0.17, "learning_rate": 1.1438277437167562e-05, "loss": 1.0126, "step": 7068 }, { "epoch": 0.17, "learning_rate": 1.1436144055315388e-05, "loss": 1.0239, "step": 7069 }, { "epoch": 0.17, "learning_rate": 1.1434010606721121e-05, "loss": 0.978, "step": 7070 }, { "epoch": 0.17, "learning_rate": 1.1431877091483907e-05, "loss": 1.0046, "step": 7071 }, { "epoch": 0.17, "learning_rate": 1.1429743509702904e-05, "loss": 1.0141, "step": 7072 }, { "epoch": 0.17, "learning_rate": 1.1427609861477258e-05, "loss": 0.9705, "step": 7073 }, { "epoch": 0.17, "learning_rate": 1.1425476146906129e-05, "loss": 0.956, "step": 7074 }, { "epoch": 0.17, "learning_rate": 1.1423342366088681e-05, "loss": 0.9881, "step": 7075 }, { "epoch": 0.17, "learning_rate": 1.1421208519124074e-05, "loss": 0.9596, "step": 7076 }, { "epoch": 0.17, "learning_rate": 1.1419074606111474e-05, "loss": 0.9895, "step": 7077 }, { "epoch": 0.17, "learning_rate": 1.1416940627150057e-05, "loss": 0.9768, "step": 7078 }, { "epoch": 0.17, "learning_rate": 1.1414806582338986e-05, "loss": 0.972, "step": 7079 }, { "epoch": 0.17, "learning_rate": 1.1412672471777441e-05, "loss": 0.9726, "step": 7080 }, { "epoch": 0.17, "learning_rate": 1.1410538295564606e-05, "loss": 0.9731, "step": 7081 }, { "epoch": 0.17, "learning_rate": 1.1408404053799653e-05, "loss": 0.9434, "step": 7082 }, { "epoch": 0.17, "learning_rate": 1.1406269746581773e-05, "loss": 0.9772, "step": 7083 }, { "epoch": 0.17, "learning_rate": 1.1404135374010155e-05, "loss": 0.9816, "step": 7084 }, { "epoch": 0.17, "learning_rate": 1.1402000936183983e-05, "loss": 0.9996, "step": 7085 }, { "epoch": 0.17, "learning_rate": 1.139986643320246e-05, "loss": 0.9437, "step": 7086 }, { "epoch": 0.17, "learning_rate": 1.1397731865164778e-05, "loss": 0.9377, "step": 7087 }, { "epoch": 0.17, "learning_rate": 1.1395597232170137e-05, "loss": 0.9349, "step": 7088 }, { "epoch": 0.17, "learning_rate": 1.139346253431774e-05, "loss": 0.9562, "step": 7089 }, { "epoch": 0.17, "learning_rate": 1.1391327771706798e-05, "loss": 0.9506, "step": 7090 }, { "epoch": 0.17, "learning_rate": 1.1389192944436509e-05, "loss": 1.0007, "step": 7091 }, { "epoch": 0.17, "learning_rate": 1.1387058052606097e-05, "loss": 0.9464, "step": 7092 }, { "epoch": 0.17, "learning_rate": 1.1384923096314773e-05, "loss": 0.9501, "step": 7093 }, { "epoch": 0.17, "learning_rate": 1.1382788075661751e-05, "loss": 0.9509, "step": 7094 }, { "epoch": 0.17, "learning_rate": 1.1380652990746256e-05, "loss": 1.0204, "step": 7095 }, { "epoch": 0.17, "learning_rate": 1.1378517841667511e-05, "loss": 0.9762, "step": 7096 }, { "epoch": 0.17, "learning_rate": 1.1376382628524743e-05, "loss": 0.9601, "step": 7097 }, { "epoch": 0.17, "learning_rate": 1.1374247351417186e-05, "loss": 0.9868, "step": 7098 }, { "epoch": 0.17, "learning_rate": 1.1372112010444062e-05, "loss": 0.9889, "step": 7099 }, { "epoch": 0.17, "learning_rate": 1.136997660570462e-05, "loss": 0.934, "step": 7100 }, { "epoch": 0.17, "learning_rate": 1.1367841137298095e-05, "loss": 0.9808, "step": 7101 }, { "epoch": 0.17, "learning_rate": 1.1365705605323723e-05, "loss": 0.955, "step": 7102 }, { "epoch": 0.17, "learning_rate": 1.1363570009880752e-05, "loss": 0.9856, "step": 7103 }, { "epoch": 0.17, "learning_rate": 1.1361434351068432e-05, "loss": 0.9478, "step": 7104 }, { "epoch": 0.17, "learning_rate": 1.1359298628986014e-05, "loss": 0.9585, "step": 7105 }, { "epoch": 0.17, "learning_rate": 1.1357162843732747e-05, "loss": 0.9803, "step": 7106 }, { "epoch": 0.17, "learning_rate": 1.1355026995407893e-05, "loss": 0.9504, "step": 7107 }, { "epoch": 0.17, "learning_rate": 1.1352891084110707e-05, "loss": 0.9222, "step": 7108 }, { "epoch": 0.17, "learning_rate": 1.1350755109940457e-05, "loss": 0.9859, "step": 7109 }, { "epoch": 0.17, "learning_rate": 1.1348619072996405e-05, "loss": 0.9596, "step": 7110 }, { "epoch": 0.17, "learning_rate": 1.1346482973377816e-05, "loss": 0.9873, "step": 7111 }, { "epoch": 0.17, "learning_rate": 1.1344346811183967e-05, "loss": 0.9987, "step": 7112 }, { "epoch": 0.17, "learning_rate": 1.134221058651413e-05, "loss": 0.9714, "step": 7113 }, { "epoch": 0.17, "learning_rate": 1.1340074299467578e-05, "loss": 0.962, "step": 7114 }, { "epoch": 0.17, "learning_rate": 1.1337937950143598e-05, "loss": 0.991, "step": 7115 }, { "epoch": 0.17, "learning_rate": 1.1335801538641468e-05, "loss": 0.9671, "step": 7116 }, { "epoch": 0.17, "learning_rate": 1.1333665065060476e-05, "loss": 0.9819, "step": 7117 }, { "epoch": 0.17, "learning_rate": 1.1331528529499909e-05, "loss": 0.9722, "step": 7118 }, { "epoch": 0.17, "learning_rate": 1.1329391932059059e-05, "loss": 0.9725, "step": 7119 }, { "epoch": 0.17, "learning_rate": 1.1327255272837221e-05, "loss": 0.9854, "step": 7120 }, { "epoch": 0.17, "learning_rate": 1.1325118551933692e-05, "loss": 0.9694, "step": 7121 }, { "epoch": 0.17, "learning_rate": 1.1322981769447771e-05, "loss": 0.9537, "step": 7122 }, { "epoch": 0.17, "learning_rate": 1.1320844925478763e-05, "loss": 0.9975, "step": 7123 }, { "epoch": 0.17, "learning_rate": 1.1318708020125973e-05, "loss": 1.0242, "step": 7124 }, { "epoch": 0.17, "learning_rate": 1.1316571053488708e-05, "loss": 1.0598, "step": 7125 }, { "epoch": 0.17, "learning_rate": 1.1314434025666278e-05, "loss": 1.0042, "step": 7126 }, { "epoch": 0.17, "learning_rate": 1.1312296936758005e-05, "loss": 0.9713, "step": 7127 }, { "epoch": 0.17, "learning_rate": 1.1310159786863198e-05, "loss": 0.9483, "step": 7128 }, { "epoch": 0.17, "learning_rate": 1.1308022576081181e-05, "loss": 0.9615, "step": 7129 }, { "epoch": 0.17, "learning_rate": 1.1305885304511277e-05, "loss": 1.004, "step": 7130 }, { "epoch": 0.17, "learning_rate": 1.130374797225281e-05, "loss": 1.0057, "step": 7131 }, { "epoch": 0.17, "learning_rate": 1.130161057940511e-05, "loss": 0.9563, "step": 7132 }, { "epoch": 0.17, "learning_rate": 1.1299473126067509e-05, "loss": 0.9964, "step": 7133 }, { "epoch": 0.17, "learning_rate": 1.1297335612339335e-05, "loss": 0.9597, "step": 7134 }, { "epoch": 0.17, "learning_rate": 1.1295198038319933e-05, "loss": 0.9702, "step": 7135 }, { "epoch": 0.17, "learning_rate": 1.129306040410864e-05, "loss": 0.9829, "step": 7136 }, { "epoch": 0.17, "learning_rate": 1.1290922709804795e-05, "loss": 0.956, "step": 7137 }, { "epoch": 0.17, "learning_rate": 1.128878495550775e-05, "loss": 0.9941, "step": 7138 }, { "epoch": 0.17, "learning_rate": 1.1286647141316849e-05, "loss": 0.9539, "step": 7139 }, { "epoch": 0.17, "learning_rate": 1.128450926733144e-05, "loss": 1.0029, "step": 7140 }, { "epoch": 0.17, "learning_rate": 1.128237133365088e-05, "loss": 0.9639, "step": 7141 }, { "epoch": 0.18, "learning_rate": 1.128023334037453e-05, "loss": 0.957, "step": 7142 }, { "epoch": 0.18, "learning_rate": 1.1278095287601737e-05, "loss": 0.9986, "step": 7143 }, { "epoch": 0.18, "learning_rate": 1.1275957175431878e-05, "loss": 0.9632, "step": 7144 }, { "epoch": 0.18, "learning_rate": 1.1273819003964306e-05, "loss": 0.9597, "step": 7145 }, { "epoch": 0.18, "learning_rate": 1.1271680773298393e-05, "loss": 0.9559, "step": 7146 }, { "epoch": 0.18, "learning_rate": 1.126954248353351e-05, "loss": 0.9751, "step": 7147 }, { "epoch": 0.18, "learning_rate": 1.1267404134769031e-05, "loss": 0.9778, "step": 7148 }, { "epoch": 0.18, "learning_rate": 1.1265265727104327e-05, "loss": 0.9977, "step": 7149 }, { "epoch": 0.18, "learning_rate": 1.1263127260638781e-05, "loss": 0.9442, "step": 7150 }, { "epoch": 0.18, "learning_rate": 1.1260988735471773e-05, "loss": 0.9421, "step": 7151 }, { "epoch": 0.18, "learning_rate": 1.125885015170268e-05, "loss": 0.9742, "step": 7152 }, { "epoch": 0.18, "learning_rate": 1.1256711509430902e-05, "loss": 0.9186, "step": 7153 }, { "epoch": 0.18, "learning_rate": 1.1254572808755819e-05, "loss": 1.0188, "step": 7154 }, { "epoch": 0.18, "learning_rate": 1.1252434049776825e-05, "loss": 0.9679, "step": 7155 }, { "epoch": 0.18, "learning_rate": 1.1250295232593317e-05, "loss": 1.0125, "step": 7156 }, { "epoch": 0.18, "learning_rate": 1.1248156357304692e-05, "loss": 0.9924, "step": 7157 }, { "epoch": 0.18, "learning_rate": 1.1246017424010345e-05, "loss": 0.9774, "step": 7158 }, { "epoch": 0.18, "learning_rate": 1.1243878432809684e-05, "loss": 1.0285, "step": 7159 }, { "epoch": 0.18, "learning_rate": 1.1241739383802112e-05, "loss": 0.9848, "step": 7160 }, { "epoch": 0.18, "learning_rate": 1.1239600277087044e-05, "loss": 0.9865, "step": 7161 }, { "epoch": 0.18, "learning_rate": 1.1237461112763884e-05, "loss": 0.9621, "step": 7162 }, { "epoch": 0.18, "learning_rate": 1.1235321890932046e-05, "loss": 0.9812, "step": 7163 }, { "epoch": 0.18, "learning_rate": 1.1233182611690945e-05, "loss": 0.9723, "step": 7164 }, { "epoch": 0.18, "learning_rate": 1.1231043275140005e-05, "loss": 0.9496, "step": 7165 }, { "epoch": 0.18, "learning_rate": 1.1228903881378646e-05, "loss": 0.9361, "step": 7166 }, { "epoch": 0.18, "learning_rate": 1.1226764430506292e-05, "loss": 0.9889, "step": 7167 }, { "epoch": 0.18, "learning_rate": 1.1224624922622367e-05, "loss": 0.9701, "step": 7168 }, { "epoch": 0.18, "learning_rate": 1.1222485357826304e-05, "loss": 0.9619, "step": 7169 }, { "epoch": 0.18, "learning_rate": 1.1220345736217537e-05, "loss": 0.9515, "step": 7170 }, { "epoch": 0.18, "learning_rate": 1.1218206057895498e-05, "loss": 0.9687, "step": 7171 }, { "epoch": 0.18, "learning_rate": 1.1216066322959622e-05, "loss": 0.9742, "step": 7172 }, { "epoch": 0.18, "learning_rate": 1.1213926531509353e-05, "loss": 0.9903, "step": 7173 }, { "epoch": 0.18, "learning_rate": 1.1211786683644132e-05, "loss": 0.9841, "step": 7174 }, { "epoch": 0.18, "learning_rate": 1.1209646779463406e-05, "loss": 1.0365, "step": 7175 }, { "epoch": 0.18, "learning_rate": 1.1207506819066622e-05, "loss": 0.9485, "step": 7176 }, { "epoch": 0.18, "learning_rate": 1.1205366802553231e-05, "loss": 0.9892, "step": 7177 }, { "epoch": 0.18, "learning_rate": 1.1203226730022687e-05, "loss": 0.9756, "step": 7178 }, { "epoch": 0.18, "learning_rate": 1.1201086601574444e-05, "loss": 0.9582, "step": 7179 }, { "epoch": 0.18, "learning_rate": 1.119894641730796e-05, "loss": 0.9641, "step": 7180 }, { "epoch": 0.18, "learning_rate": 1.11968061773227e-05, "loss": 0.9785, "step": 7181 }, { "epoch": 0.18, "learning_rate": 1.1194665881718124e-05, "loss": 0.9785, "step": 7182 }, { "epoch": 0.18, "learning_rate": 1.1192525530593696e-05, "loss": 0.9558, "step": 7183 }, { "epoch": 0.18, "learning_rate": 1.1190385124048893e-05, "loss": 0.9988, "step": 7184 }, { "epoch": 0.18, "learning_rate": 1.1188244662183181e-05, "loss": 1.0112, "step": 7185 }, { "epoch": 0.18, "learning_rate": 1.1186104145096031e-05, "loss": 0.9402, "step": 7186 }, { "epoch": 0.18, "learning_rate": 1.1183963572886923e-05, "loss": 1.0103, "step": 7187 }, { "epoch": 0.18, "learning_rate": 1.1181822945655338e-05, "loss": 0.9676, "step": 7188 }, { "epoch": 0.18, "learning_rate": 1.1179682263500754e-05, "loss": 0.9782, "step": 7189 }, { "epoch": 0.18, "learning_rate": 1.1177541526522658e-05, "loss": 0.9549, "step": 7190 }, { "epoch": 0.18, "learning_rate": 1.1175400734820531e-05, "loss": 0.9959, "step": 7191 }, { "epoch": 0.18, "learning_rate": 1.117325988849387e-05, "loss": 0.9571, "step": 7192 }, { "epoch": 0.18, "learning_rate": 1.1171118987642164e-05, "loss": 0.9749, "step": 7193 }, { "epoch": 0.18, "learning_rate": 1.1168978032364908e-05, "loss": 1.0029, "step": 7194 }, { "epoch": 0.18, "learning_rate": 1.1166837022761595e-05, "loss": 1.0028, "step": 7195 }, { "epoch": 0.18, "learning_rate": 1.1164695958931729e-05, "loss": 0.9758, "step": 7196 }, { "epoch": 0.18, "learning_rate": 1.1162554840974808e-05, "loss": 0.9722, "step": 7197 }, { "epoch": 0.18, "learning_rate": 1.1160413668990338e-05, "loss": 0.9885, "step": 7198 }, { "epoch": 0.18, "learning_rate": 1.1158272443077828e-05, "loss": 0.989, "step": 7199 }, { "epoch": 0.18, "learning_rate": 1.1156131163336785e-05, "loss": 0.9432, "step": 7200 }, { "epoch": 0.18, "learning_rate": 1.1153989829866725e-05, "loss": 0.9811, "step": 7201 }, { "epoch": 0.18, "learning_rate": 1.1151848442767158e-05, "loss": 0.9355, "step": 7202 }, { "epoch": 0.18, "learning_rate": 1.1149707002137598e-05, "loss": 0.9496, "step": 7203 }, { "epoch": 0.18, "learning_rate": 1.114756550807757e-05, "loss": 0.9915, "step": 7204 }, { "epoch": 0.18, "learning_rate": 1.1145423960686596e-05, "loss": 0.9755, "step": 7205 }, { "epoch": 0.18, "learning_rate": 1.1143282360064197e-05, "loss": 0.9802, "step": 7206 }, { "epoch": 0.18, "learning_rate": 1.11411407063099e-05, "loss": 0.9296, "step": 7207 }, { "epoch": 0.18, "learning_rate": 1.1138998999523243e-05, "loss": 0.9652, "step": 7208 }, { "epoch": 0.18, "learning_rate": 1.1136857239803747e-05, "loss": 0.9699, "step": 7209 }, { "epoch": 0.18, "learning_rate": 1.1134715427250948e-05, "loss": 0.9887, "step": 7210 }, { "epoch": 0.18, "learning_rate": 1.1132573561964385e-05, "loss": 1.0092, "step": 7211 }, { "epoch": 0.18, "learning_rate": 1.11304316440436e-05, "loss": 0.9634, "step": 7212 }, { "epoch": 0.18, "learning_rate": 1.1128289673588131e-05, "loss": 0.9939, "step": 7213 }, { "epoch": 0.18, "learning_rate": 1.1126147650697519e-05, "loss": 0.9752, "step": 7214 }, { "epoch": 0.18, "learning_rate": 1.1124005575471316e-05, "loss": 0.979, "step": 7215 }, { "epoch": 0.18, "learning_rate": 1.112186344800907e-05, "loss": 0.9562, "step": 7216 }, { "epoch": 0.18, "learning_rate": 1.111972126841033e-05, "loss": 0.9872, "step": 7217 }, { "epoch": 0.18, "learning_rate": 1.1117579036774653e-05, "loss": 0.9615, "step": 7218 }, { "epoch": 0.18, "learning_rate": 1.1115436753201592e-05, "loss": 0.97, "step": 7219 }, { "epoch": 0.18, "learning_rate": 1.1113294417790705e-05, "loss": 0.9976, "step": 7220 }, { "epoch": 0.18, "learning_rate": 1.1111152030641556e-05, "loss": 1.0063, "step": 7221 }, { "epoch": 0.18, "learning_rate": 1.110900959185371e-05, "loss": 1.0071, "step": 7222 }, { "epoch": 0.18, "learning_rate": 1.1106867101526726e-05, "loss": 0.9464, "step": 7223 }, { "epoch": 0.18, "learning_rate": 1.1104724559760181e-05, "loss": 0.9952, "step": 7224 }, { "epoch": 0.18, "learning_rate": 1.1102581966653638e-05, "loss": 0.9772, "step": 7225 }, { "epoch": 0.18, "learning_rate": 1.1100439322306671e-05, "loss": 0.9827, "step": 7226 }, { "epoch": 0.18, "learning_rate": 1.109829662681886e-05, "loss": 0.9769, "step": 7227 }, { "epoch": 0.18, "learning_rate": 1.109615388028978e-05, "loss": 0.9662, "step": 7228 }, { "epoch": 0.18, "learning_rate": 1.1094011082819006e-05, "loss": 0.9787, "step": 7229 }, { "epoch": 0.18, "learning_rate": 1.1091868234506132e-05, "loss": 0.9665, "step": 7230 }, { "epoch": 0.18, "learning_rate": 1.1089725335450734e-05, "loss": 0.9351, "step": 7231 }, { "epoch": 0.18, "learning_rate": 1.1087582385752402e-05, "loss": 0.9728, "step": 7232 }, { "epoch": 0.18, "learning_rate": 1.1085439385510726e-05, "loss": 0.9473, "step": 7233 }, { "epoch": 0.18, "learning_rate": 1.1083296334825295e-05, "loss": 0.9846, "step": 7234 }, { "epoch": 0.18, "learning_rate": 1.1081153233795708e-05, "loss": 1.0092, "step": 7235 }, { "epoch": 0.18, "learning_rate": 1.1079010082521557e-05, "loss": 0.9546, "step": 7236 }, { "epoch": 0.18, "learning_rate": 1.1076866881102441e-05, "loss": 0.9484, "step": 7237 }, { "epoch": 0.18, "learning_rate": 1.1074723629637964e-05, "loss": 0.938, "step": 7238 }, { "epoch": 0.18, "learning_rate": 1.1072580328227732e-05, "loss": 0.9695, "step": 7239 }, { "epoch": 0.18, "learning_rate": 1.1070436976971345e-05, "loss": 0.9104, "step": 7240 }, { "epoch": 0.18, "learning_rate": 1.1068293575968414e-05, "loss": 0.9499, "step": 7241 }, { "epoch": 0.18, "learning_rate": 1.106615012531855e-05, "loss": 0.9725, "step": 7242 }, { "epoch": 0.18, "learning_rate": 1.1064006625121365e-05, "loss": 0.9783, "step": 7243 }, { "epoch": 0.18, "learning_rate": 1.106186307547647e-05, "loss": 0.9484, "step": 7244 }, { "epoch": 0.18, "learning_rate": 1.1059719476483492e-05, "loss": 0.9799, "step": 7245 }, { "epoch": 0.18, "learning_rate": 1.1057575828242044e-05, "loss": 1.0, "step": 7246 }, { "epoch": 0.18, "learning_rate": 1.105543213085175e-05, "loss": 0.9772, "step": 7247 }, { "epoch": 0.18, "learning_rate": 1.1053288384412236e-05, "loss": 0.9475, "step": 7248 }, { "epoch": 0.18, "learning_rate": 1.1051144589023122e-05, "loss": 0.9745, "step": 7249 }, { "epoch": 0.18, "learning_rate": 1.1049000744784041e-05, "loss": 1.0099, "step": 7250 }, { "epoch": 0.18, "learning_rate": 1.1046856851794627e-05, "loss": 0.9825, "step": 7251 }, { "epoch": 0.18, "learning_rate": 1.1044712910154509e-05, "loss": 0.9718, "step": 7252 }, { "epoch": 0.18, "learning_rate": 1.1042568919963323e-05, "loss": 1.0067, "step": 7253 }, { "epoch": 0.18, "learning_rate": 1.1040424881320713e-05, "loss": 0.9607, "step": 7254 }, { "epoch": 0.18, "learning_rate": 1.1038280794326313e-05, "loss": 0.9913, "step": 7255 }, { "epoch": 0.18, "learning_rate": 1.1036136659079764e-05, "loss": 1.0354, "step": 7256 }, { "epoch": 0.18, "learning_rate": 1.1033992475680718e-05, "loss": 0.9682, "step": 7257 }, { "epoch": 0.18, "learning_rate": 1.103184824422881e-05, "loss": 0.9541, "step": 7258 }, { "epoch": 0.18, "learning_rate": 1.10297039648237e-05, "loss": 0.9667, "step": 7259 }, { "epoch": 0.18, "learning_rate": 1.1027559637565039e-05, "loss": 0.9585, "step": 7260 }, { "epoch": 0.18, "learning_rate": 1.1025415262552474e-05, "loss": 0.9751, "step": 7261 }, { "epoch": 0.18, "learning_rate": 1.1023270839885668e-05, "loss": 1.0345, "step": 7262 }, { "epoch": 0.18, "learning_rate": 1.1021126369664275e-05, "loss": 1.032, "step": 7263 }, { "epoch": 0.18, "learning_rate": 1.1018981851987951e-05, "loss": 1.0241, "step": 7264 }, { "epoch": 0.18, "learning_rate": 1.1016837286956365e-05, "loss": 1.0142, "step": 7265 }, { "epoch": 0.18, "learning_rate": 1.1014692674669179e-05, "loss": 0.9519, "step": 7266 }, { "epoch": 0.18, "learning_rate": 1.101254801522606e-05, "loss": 1.0191, "step": 7267 }, { "epoch": 0.18, "learning_rate": 1.101040330872668e-05, "loss": 0.9401, "step": 7268 }, { "epoch": 0.18, "learning_rate": 1.1008258555270705e-05, "loss": 0.9622, "step": 7269 }, { "epoch": 0.18, "learning_rate": 1.1006113754957813e-05, "loss": 1.0149, "step": 7270 }, { "epoch": 0.18, "learning_rate": 1.1003968907887676e-05, "loss": 0.9705, "step": 7271 }, { "epoch": 0.18, "learning_rate": 1.1001824014159974e-05, "loss": 0.9539, "step": 7272 }, { "epoch": 0.18, "learning_rate": 1.0999679073874383e-05, "loss": 1.0188, "step": 7273 }, { "epoch": 0.18, "learning_rate": 1.0997534087130592e-05, "loss": 0.9817, "step": 7274 }, { "epoch": 0.18, "learning_rate": 1.099538905402828e-05, "loss": 0.954, "step": 7275 }, { "epoch": 0.18, "learning_rate": 1.0993243974667134e-05, "loss": 0.9729, "step": 7276 }, { "epoch": 0.18, "learning_rate": 1.0991098849146843e-05, "loss": 0.9688, "step": 7277 }, { "epoch": 0.18, "learning_rate": 1.0988953677567102e-05, "loss": 0.9642, "step": 7278 }, { "epoch": 0.18, "learning_rate": 1.0986808460027597e-05, "loss": 0.9652, "step": 7279 }, { "epoch": 0.18, "learning_rate": 1.0984663196628025e-05, "loss": 0.9797, "step": 7280 }, { "epoch": 0.18, "learning_rate": 1.0982517887468083e-05, "loss": 0.9754, "step": 7281 }, { "epoch": 0.18, "learning_rate": 1.0980372532647473e-05, "loss": 0.9831, "step": 7282 }, { "epoch": 0.18, "learning_rate": 1.0978227132265892e-05, "loss": 0.9784, "step": 7283 }, { "epoch": 0.18, "learning_rate": 1.0976081686423049e-05, "loss": 0.9645, "step": 7284 }, { "epoch": 0.18, "learning_rate": 1.0973936195218647e-05, "loss": 0.9595, "step": 7285 }, { "epoch": 0.18, "learning_rate": 1.097179065875239e-05, "loss": 0.9888, "step": 7286 }, { "epoch": 0.18, "learning_rate": 1.0969645077123992e-05, "loss": 0.9788, "step": 7287 }, { "epoch": 0.18, "learning_rate": 1.0967499450433163e-05, "loss": 0.9247, "step": 7288 }, { "epoch": 0.18, "learning_rate": 1.0965353778779617e-05, "loss": 1.0146, "step": 7289 }, { "epoch": 0.18, "learning_rate": 1.0963208062263072e-05, "loss": 0.9692, "step": 7290 }, { "epoch": 0.18, "learning_rate": 1.0961062300983244e-05, "loss": 0.9681, "step": 7291 }, { "epoch": 0.19, "learning_rate": 1.0958916495039856e-05, "loss": 0.9944, "step": 7292 }, { "epoch": 0.19, "learning_rate": 1.0956770644532626e-05, "loss": 0.9389, "step": 7293 }, { "epoch": 0.19, "learning_rate": 1.0954624749561282e-05, "loss": 1.0025, "step": 7294 }, { "epoch": 0.19, "learning_rate": 1.095247881022555e-05, "loss": 0.946, "step": 7295 }, { "epoch": 0.19, "learning_rate": 1.0950332826625154e-05, "loss": 0.9875, "step": 7296 }, { "epoch": 0.19, "learning_rate": 1.094818679885983e-05, "loss": 0.9545, "step": 7297 }, { "epoch": 0.19, "learning_rate": 1.0946040727029308e-05, "loss": 0.9795, "step": 7298 }, { "epoch": 0.19, "learning_rate": 1.0943894611233324e-05, "loss": 0.9721, "step": 7299 }, { "epoch": 0.19, "learning_rate": 1.0941748451571616e-05, "loss": 0.9687, "step": 7300 }, { "epoch": 0.19, "learning_rate": 1.0939602248143921e-05, "loss": 1.0054, "step": 7301 }, { "epoch": 0.19, "learning_rate": 1.0937456001049975e-05, "loss": 0.9668, "step": 7302 }, { "epoch": 0.19, "learning_rate": 1.093530971038953e-05, "loss": 1.0179, "step": 7303 }, { "epoch": 0.19, "learning_rate": 1.0933163376262324e-05, "loss": 0.9349, "step": 7304 }, { "epoch": 0.19, "learning_rate": 1.0931016998768106e-05, "loss": 0.9544, "step": 7305 }, { "epoch": 0.19, "learning_rate": 1.0928870578006627e-05, "loss": 0.9981, "step": 7306 }, { "epoch": 0.19, "learning_rate": 1.0926724114077634e-05, "loss": 0.9602, "step": 7307 }, { "epoch": 0.19, "learning_rate": 1.0924577607080882e-05, "loss": 0.999, "step": 7308 }, { "epoch": 0.19, "learning_rate": 1.0922431057116127e-05, "loss": 0.9934, "step": 7309 }, { "epoch": 0.19, "learning_rate": 1.0920284464283121e-05, "loss": 0.9651, "step": 7310 }, { "epoch": 0.19, "learning_rate": 1.0918137828681627e-05, "loss": 1.0027, "step": 7311 }, { "epoch": 0.19, "learning_rate": 1.0915991150411407e-05, "loss": 0.9915, "step": 7312 }, { "epoch": 0.19, "learning_rate": 1.091384442957222e-05, "loss": 1.0041, "step": 7313 }, { "epoch": 0.19, "learning_rate": 1.0911697666263837e-05, "loss": 0.9688, "step": 7314 }, { "epoch": 0.19, "learning_rate": 1.0909550860586016e-05, "loss": 0.979, "step": 7315 }, { "epoch": 0.19, "learning_rate": 1.0907404012638532e-05, "loss": 0.9369, "step": 7316 }, { "epoch": 0.19, "learning_rate": 1.0905257122521155e-05, "loss": 0.9979, "step": 7317 }, { "epoch": 0.19, "learning_rate": 1.0903110190333655e-05, "loss": 0.9574, "step": 7318 }, { "epoch": 0.19, "learning_rate": 1.0900963216175807e-05, "loss": 1.0028, "step": 7319 }, { "epoch": 0.19, "learning_rate": 1.0898816200147392e-05, "loss": 0.9544, "step": 7320 }, { "epoch": 0.19, "learning_rate": 1.0896669142348182e-05, "loss": 0.9828, "step": 7321 }, { "epoch": 0.19, "learning_rate": 1.0894522042877962e-05, "loss": 0.9678, "step": 7322 }, { "epoch": 0.19, "learning_rate": 1.0892374901836516e-05, "loss": 0.9932, "step": 7323 }, { "epoch": 0.19, "learning_rate": 1.0890227719323625e-05, "loss": 0.9045, "step": 7324 }, { "epoch": 0.19, "learning_rate": 1.0888080495439073e-05, "loss": 0.9745, "step": 7325 }, { "epoch": 0.19, "learning_rate": 1.0885933230282655e-05, "loss": 0.9861, "step": 7326 }, { "epoch": 0.19, "learning_rate": 1.0883785923954154e-05, "loss": 0.9524, "step": 7327 }, { "epoch": 0.19, "learning_rate": 1.0881638576553365e-05, "loss": 0.987, "step": 7328 }, { "epoch": 0.19, "learning_rate": 1.0879491188180084e-05, "loss": 0.9983, "step": 7329 }, { "epoch": 0.19, "learning_rate": 1.0877343758934105e-05, "loss": 0.9587, "step": 7330 }, { "epoch": 0.19, "learning_rate": 1.0875196288915226e-05, "loss": 0.998, "step": 7331 }, { "epoch": 0.19, "learning_rate": 1.0873048778223247e-05, "loss": 0.9991, "step": 7332 }, { "epoch": 0.19, "learning_rate": 1.0870901226957966e-05, "loss": 0.9952, "step": 7333 }, { "epoch": 0.19, "learning_rate": 1.0868753635219189e-05, "loss": 0.9651, "step": 7334 }, { "epoch": 0.19, "learning_rate": 1.0866606003106725e-05, "loss": 0.9669, "step": 7335 }, { "epoch": 0.19, "learning_rate": 1.0864458330720376e-05, "loss": 0.9849, "step": 7336 }, { "epoch": 0.19, "learning_rate": 1.0862310618159952e-05, "loss": 1.001, "step": 7337 }, { "epoch": 0.19, "learning_rate": 1.0860162865525266e-05, "loss": 0.9601, "step": 7338 }, { "epoch": 0.19, "learning_rate": 1.0858015072916128e-05, "loss": 0.9402, "step": 7339 }, { "epoch": 0.19, "learning_rate": 1.0855867240432354e-05, "loss": 0.9573, "step": 7340 }, { "epoch": 0.19, "learning_rate": 1.0853719368173763e-05, "loss": 0.9856, "step": 7341 }, { "epoch": 0.19, "learning_rate": 1.0851571456240169e-05, "loss": 0.9701, "step": 7342 }, { "epoch": 0.19, "learning_rate": 1.0849423504731396e-05, "loss": 0.9763, "step": 7343 }, { "epoch": 0.19, "learning_rate": 1.084727551374726e-05, "loss": 0.9798, "step": 7344 }, { "epoch": 0.19, "learning_rate": 1.0845127483387589e-05, "loss": 0.9586, "step": 7345 }, { "epoch": 0.19, "learning_rate": 1.0842979413752212e-05, "loss": 0.9692, "step": 7346 }, { "epoch": 0.19, "learning_rate": 1.0840831304940952e-05, "loss": 1.0214, "step": 7347 }, { "epoch": 0.19, "learning_rate": 1.083868315705364e-05, "loss": 0.99, "step": 7348 }, { "epoch": 0.19, "learning_rate": 1.0836534970190106e-05, "loss": 0.9735, "step": 7349 }, { "epoch": 0.19, "learning_rate": 1.0834386744450182e-05, "loss": 0.9272, "step": 7350 }, { "epoch": 0.19, "learning_rate": 1.0832238479933703e-05, "loss": 0.9916, "step": 7351 }, { "epoch": 0.19, "learning_rate": 1.0830090176740512e-05, "loss": 0.9722, "step": 7352 }, { "epoch": 0.19, "learning_rate": 1.0827941834970436e-05, "loss": 1.0144, "step": 7353 }, { "epoch": 0.19, "learning_rate": 1.0825793454723325e-05, "loss": 0.9395, "step": 7354 }, { "epoch": 0.19, "learning_rate": 1.0823645036099017e-05, "loss": 0.9692, "step": 7355 }, { "epoch": 0.19, "learning_rate": 1.0821496579197355e-05, "loss": 0.9587, "step": 7356 }, { "epoch": 0.19, "learning_rate": 1.0819348084118183e-05, "loss": 0.9576, "step": 7357 }, { "epoch": 0.19, "learning_rate": 1.0817199550961352e-05, "loss": 0.9748, "step": 7358 }, { "epoch": 0.19, "learning_rate": 1.0815050979826711e-05, "loss": 0.9708, "step": 7359 }, { "epoch": 0.19, "learning_rate": 1.0812902370814109e-05, "loss": 0.9523, "step": 7360 }, { "epoch": 0.19, "learning_rate": 1.0810753724023397e-05, "loss": 0.9774, "step": 7361 }, { "epoch": 0.19, "learning_rate": 1.0808605039554437e-05, "loss": 0.9662, "step": 7362 }, { "epoch": 0.19, "learning_rate": 1.0806456317507072e-05, "loss": 0.9928, "step": 7363 }, { "epoch": 0.19, "learning_rate": 1.0804307557981174e-05, "loss": 0.9584, "step": 7364 }, { "epoch": 0.19, "learning_rate": 1.0802158761076589e-05, "loss": 0.9216, "step": 7365 }, { "epoch": 0.19, "learning_rate": 1.0800009926893188e-05, "loss": 0.9994, "step": 7366 }, { "epoch": 0.19, "learning_rate": 1.0797861055530832e-05, "loss": 0.9682, "step": 7367 }, { "epoch": 0.19, "learning_rate": 1.0795712147089384e-05, "loss": 0.9616, "step": 7368 }, { "epoch": 0.19, "learning_rate": 1.079356320166871e-05, "loss": 0.9952, "step": 7369 }, { "epoch": 0.19, "learning_rate": 1.0791414219368684e-05, "loss": 0.9486, "step": 7370 }, { "epoch": 0.19, "learning_rate": 1.0789265200289168e-05, "loss": 0.9494, "step": 7371 }, { "epoch": 0.19, "learning_rate": 1.0787116144530037e-05, "loss": 0.9819, "step": 7372 }, { "epoch": 0.19, "learning_rate": 1.0784967052191164e-05, "loss": 0.9839, "step": 7373 }, { "epoch": 0.19, "learning_rate": 1.0782817923372425e-05, "loss": 1.0012, "step": 7374 }, { "epoch": 0.19, "learning_rate": 1.0780668758173696e-05, "loss": 0.9432, "step": 7375 }, { "epoch": 0.19, "learning_rate": 1.0778519556694853e-05, "loss": 0.983, "step": 7376 }, { "epoch": 0.19, "learning_rate": 1.0776370319035783e-05, "loss": 0.963, "step": 7377 }, { "epoch": 0.19, "learning_rate": 1.0774221045296363e-05, "loss": 0.9602, "step": 7378 }, { "epoch": 0.19, "learning_rate": 1.0772071735576475e-05, "loss": 0.9943, "step": 7379 }, { "epoch": 0.19, "learning_rate": 1.0769922389976004e-05, "loss": 0.9912, "step": 7380 }, { "epoch": 0.19, "learning_rate": 1.0767773008594842e-05, "loss": 0.968, "step": 7381 }, { "epoch": 0.19, "learning_rate": 1.0765623591532872e-05, "loss": 1.0289, "step": 7382 }, { "epoch": 0.19, "learning_rate": 1.0763474138889988e-05, "loss": 0.963, "step": 7383 }, { "epoch": 0.19, "learning_rate": 1.0761324650766078e-05, "loss": 0.9734, "step": 7384 }, { "epoch": 0.19, "learning_rate": 1.075917512726104e-05, "loss": 0.984, "step": 7385 }, { "epoch": 0.19, "learning_rate": 1.0757025568474764e-05, "loss": 0.9602, "step": 7386 }, { "epoch": 0.19, "learning_rate": 1.0754875974507153e-05, "loss": 0.9765, "step": 7387 }, { "epoch": 0.19, "learning_rate": 1.0752726345458097e-05, "loss": 0.9835, "step": 7388 }, { "epoch": 0.19, "learning_rate": 1.0750576681427504e-05, "loss": 0.9985, "step": 7389 }, { "epoch": 0.19, "learning_rate": 1.074842698251527e-05, "loss": 0.9589, "step": 7390 }, { "epoch": 0.19, "learning_rate": 1.07462772488213e-05, "loss": 0.9679, "step": 7391 }, { "epoch": 0.19, "learning_rate": 1.07441274804455e-05, "loss": 0.9858, "step": 7392 }, { "epoch": 0.19, "learning_rate": 1.0741977677487778e-05, "loss": 0.9464, "step": 7393 }, { "epoch": 0.19, "learning_rate": 1.0739827840048036e-05, "loss": 0.983, "step": 7394 }, { "epoch": 0.19, "learning_rate": 1.073767796822619e-05, "loss": 0.9825, "step": 7395 }, { "epoch": 0.19, "learning_rate": 1.0735528062122148e-05, "loss": 1.0078, "step": 7396 }, { "epoch": 0.19, "learning_rate": 1.0733378121835821e-05, "loss": 0.951, "step": 7397 }, { "epoch": 0.19, "learning_rate": 1.0731228147467128e-05, "loss": 0.9716, "step": 7398 }, { "epoch": 0.19, "learning_rate": 1.072907813911598e-05, "loss": 1.0227, "step": 7399 }, { "epoch": 0.19, "learning_rate": 1.0726928096882301e-05, "loss": 0.9476, "step": 7400 }, { "epoch": 0.19, "learning_rate": 1.0724778020866007e-05, "loss": 1.0194, "step": 7401 }, { "epoch": 0.19, "learning_rate": 1.0722627911167014e-05, "loss": 0.9664, "step": 7402 }, { "epoch": 0.19, "learning_rate": 1.072047776788525e-05, "loss": 0.951, "step": 7403 }, { "epoch": 0.19, "learning_rate": 1.071832759112064e-05, "loss": 1.0112, "step": 7404 }, { "epoch": 0.19, "learning_rate": 1.0716177380973105e-05, "loss": 0.9872, "step": 7405 }, { "epoch": 0.19, "learning_rate": 1.0714027137542575e-05, "loss": 0.954, "step": 7406 }, { "epoch": 0.19, "learning_rate": 1.0711876860928976e-05, "loss": 1.0184, "step": 7407 }, { "epoch": 0.19, "learning_rate": 1.0709726551232242e-05, "loss": 0.95, "step": 7408 }, { "epoch": 0.19, "learning_rate": 1.07075762085523e-05, "loss": 0.969, "step": 7409 }, { "epoch": 0.19, "learning_rate": 1.0705425832989084e-05, "loss": 0.9766, "step": 7410 }, { "epoch": 0.19, "learning_rate": 1.0703275424642532e-05, "loss": 0.9789, "step": 7411 }, { "epoch": 0.19, "learning_rate": 1.0701124983612579e-05, "loss": 0.9781, "step": 7412 }, { "epoch": 0.19, "learning_rate": 1.0698974509999159e-05, "loss": 0.9656, "step": 7413 }, { "epoch": 0.19, "learning_rate": 1.0696824003902213e-05, "loss": 0.9857, "step": 7414 }, { "epoch": 0.19, "learning_rate": 1.0694673465421686e-05, "loss": 0.9748, "step": 7415 }, { "epoch": 0.19, "learning_rate": 1.0692522894657517e-05, "loss": 0.9271, "step": 7416 }, { "epoch": 0.19, "learning_rate": 1.0690372291709648e-05, "loss": 0.988, "step": 7417 }, { "epoch": 0.19, "learning_rate": 1.0688221656678026e-05, "loss": 0.978, "step": 7418 }, { "epoch": 0.19, "learning_rate": 1.0686070989662596e-05, "loss": 0.9859, "step": 7419 }, { "epoch": 0.19, "learning_rate": 1.068392029076331e-05, "loss": 0.9921, "step": 7420 }, { "epoch": 0.19, "learning_rate": 1.0681769560080115e-05, "loss": 0.9379, "step": 7421 }, { "epoch": 0.19, "learning_rate": 1.0679618797712961e-05, "loss": 0.9778, "step": 7422 }, { "epoch": 0.19, "learning_rate": 1.0677468003761804e-05, "loss": 1.0026, "step": 7423 }, { "epoch": 0.19, "learning_rate": 1.06753171783266e-05, "loss": 0.9898, "step": 7424 }, { "epoch": 0.19, "learning_rate": 1.0673166321507293e-05, "loss": 1.0072, "step": 7425 }, { "epoch": 0.19, "learning_rate": 1.0671015433403851e-05, "loss": 0.9578, "step": 7426 }, { "epoch": 0.19, "learning_rate": 1.0668864514116231e-05, "loss": 0.9589, "step": 7427 }, { "epoch": 0.19, "learning_rate": 1.0666713563744389e-05, "loss": 0.9411, "step": 7428 }, { "epoch": 0.19, "learning_rate": 1.0664562582388291e-05, "loss": 0.933, "step": 7429 }, { "epoch": 0.19, "learning_rate": 1.0662411570147897e-05, "loss": 0.955, "step": 7430 }, { "epoch": 0.19, "learning_rate": 1.0660260527123171e-05, "loss": 0.964, "step": 7431 }, { "epoch": 0.19, "learning_rate": 1.065810945341408e-05, "loss": 0.9065, "step": 7432 }, { "epoch": 0.19, "learning_rate": 1.0655958349120592e-05, "loss": 0.9712, "step": 7433 }, { "epoch": 0.19, "learning_rate": 1.065380721434267e-05, "loss": 0.961, "step": 7434 }, { "epoch": 0.19, "learning_rate": 1.0651656049180293e-05, "loss": 0.9878, "step": 7435 }, { "epoch": 0.19, "learning_rate": 1.0649504853733423e-05, "loss": 0.9571, "step": 7436 }, { "epoch": 0.19, "learning_rate": 1.0647353628102038e-05, "loss": 0.9907, "step": 7437 }, { "epoch": 0.19, "learning_rate": 1.0645202372386114e-05, "loss": 0.9843, "step": 7438 }, { "epoch": 0.19, "learning_rate": 1.0643051086685624e-05, "loss": 0.9734, "step": 7439 }, { "epoch": 0.19, "learning_rate": 1.0640899771100539e-05, "loss": 1.0105, "step": 7440 }, { "epoch": 0.19, "learning_rate": 1.063874842573085e-05, "loss": 0.9541, "step": 7441 }, { "epoch": 0.2, "learning_rate": 1.0636597050676525e-05, "loss": 1.0062, "step": 7442 }, { "epoch": 0.2, "learning_rate": 1.063444564603755e-05, "loss": 0.9586, "step": 7443 }, { "epoch": 0.2, "learning_rate": 1.0632294211913911e-05, "loss": 0.936, "step": 7444 }, { "epoch": 0.2, "learning_rate": 1.0630142748405585e-05, "loss": 1.0234, "step": 7445 }, { "epoch": 0.2, "learning_rate": 1.0627991255612564e-05, "loss": 0.9617, "step": 7446 }, { "epoch": 0.2, "learning_rate": 1.062583973363483e-05, "loss": 0.9994, "step": 7447 }, { "epoch": 0.2, "learning_rate": 1.0623688182572372e-05, "loss": 0.9883, "step": 7448 }, { "epoch": 0.2, "learning_rate": 1.0621536602525176e-05, "loss": 0.9877, "step": 7449 }, { "epoch": 0.2, "learning_rate": 1.061938499359324e-05, "loss": 0.9405, "step": 7450 }, { "epoch": 0.2, "learning_rate": 1.0617233355876548e-05, "loss": 0.9674, "step": 7451 }, { "epoch": 0.2, "learning_rate": 1.06150816894751e-05, "loss": 0.9476, "step": 7452 }, { "epoch": 0.2, "learning_rate": 1.0612929994488887e-05, "loss": 0.9823, "step": 7453 }, { "epoch": 0.2, "learning_rate": 1.061077827101791e-05, "loss": 0.9739, "step": 7454 }, { "epoch": 0.2, "learning_rate": 1.0608626519162157e-05, "loss": 1.0312, "step": 7455 }, { "epoch": 0.2, "learning_rate": 1.0606474739021635e-05, "loss": 0.9641, "step": 7456 }, { "epoch": 0.2, "learning_rate": 1.0604322930696338e-05, "loss": 0.9058, "step": 7457 }, { "epoch": 0.2, "learning_rate": 1.060217109428627e-05, "loss": 0.9737, "step": 7458 }, { "epoch": 0.2, "learning_rate": 1.0600019229891436e-05, "loss": 0.9796, "step": 7459 }, { "epoch": 0.2, "learning_rate": 1.0597867337611834e-05, "loss": 0.9787, "step": 7460 }, { "epoch": 0.2, "learning_rate": 1.0595715417547477e-05, "loss": 0.9589, "step": 7461 }, { "epoch": 0.2, "learning_rate": 1.0593563469798364e-05, "loss": 0.9385, "step": 7462 }, { "epoch": 0.2, "learning_rate": 1.0591411494464506e-05, "loss": 1.0288, "step": 7463 }, { "epoch": 0.2, "learning_rate": 1.0589259491645914e-05, "loss": 0.9981, "step": 7464 }, { "epoch": 0.2, "learning_rate": 1.0587107461442592e-05, "loss": 0.9425, "step": 7465 }, { "epoch": 0.2, "learning_rate": 1.0584955403954556e-05, "loss": 0.9856, "step": 7466 }, { "epoch": 0.2, "learning_rate": 1.058280331928182e-05, "loss": 0.9928, "step": 7467 }, { "epoch": 0.2, "learning_rate": 1.0580651207524395e-05, "loss": 0.9214, "step": 7468 }, { "epoch": 0.2, "learning_rate": 1.05784990687823e-05, "loss": 1.0067, "step": 7469 }, { "epoch": 0.2, "learning_rate": 1.057634690315555e-05, "loss": 0.9477, "step": 7470 }, { "epoch": 0.2, "learning_rate": 1.0574194710744158e-05, "loss": 0.9666, "step": 7471 }, { "epoch": 0.2, "learning_rate": 1.057204249164815e-05, "loss": 1.0187, "step": 7472 }, { "epoch": 0.2, "learning_rate": 1.0569890245967544e-05, "loss": 0.9704, "step": 7473 }, { "epoch": 0.2, "learning_rate": 1.056773797380236e-05, "loss": 0.9553, "step": 7474 }, { "epoch": 0.2, "learning_rate": 1.056558567525262e-05, "loss": 0.9774, "step": 7475 }, { "epoch": 0.2, "learning_rate": 1.0563433350418355e-05, "loss": 0.9848, "step": 7476 }, { "epoch": 0.2, "learning_rate": 1.0561280999399584e-05, "loss": 1.0017, "step": 7477 }, { "epoch": 0.2, "learning_rate": 1.0559128622296333e-05, "loss": 0.98, "step": 7478 }, { "epoch": 0.2, "learning_rate": 1.0556976219208632e-05, "loss": 0.9803, "step": 7479 }, { "epoch": 0.2, "learning_rate": 1.0554823790236509e-05, "loss": 0.9779, "step": 7480 }, { "epoch": 0.2, "learning_rate": 1.0552671335479996e-05, "loss": 1.0027, "step": 7481 }, { "epoch": 0.2, "learning_rate": 1.0550518855039117e-05, "loss": 0.9966, "step": 7482 }, { "epoch": 0.2, "learning_rate": 1.0548366349013915e-05, "loss": 0.9493, "step": 7483 }, { "epoch": 0.2, "learning_rate": 1.0546213817504421e-05, "loss": 0.9817, "step": 7484 }, { "epoch": 0.2, "learning_rate": 1.0544061260610665e-05, "loss": 1.017, "step": 7485 }, { "epoch": 0.2, "learning_rate": 1.0541908678432684e-05, "loss": 1.0218, "step": 7486 }, { "epoch": 0.2, "learning_rate": 1.053975607107052e-05, "loss": 0.9569, "step": 7487 }, { "epoch": 0.2, "learning_rate": 1.0537603438624205e-05, "loss": 0.983, "step": 7488 }, { "epoch": 0.2, "learning_rate": 1.0535450781193784e-05, "loss": 0.99, "step": 7489 }, { "epoch": 0.2, "learning_rate": 1.0533298098879295e-05, "loss": 0.9446, "step": 7490 }, { "epoch": 0.2, "learning_rate": 1.053114539178078e-05, "loss": 0.983, "step": 7491 }, { "epoch": 0.2, "learning_rate": 1.0528992659998283e-05, "loss": 1.0026, "step": 7492 }, { "epoch": 0.2, "learning_rate": 1.0526839903631849e-05, "loss": 0.97, "step": 7493 }, { "epoch": 0.2, "learning_rate": 1.0524687122781519e-05, "loss": 0.9742, "step": 7494 }, { "epoch": 0.2, "learning_rate": 1.0522534317547342e-05, "loss": 0.9534, "step": 7495 }, { "epoch": 0.2, "learning_rate": 1.0520381488029369e-05, "loss": 1.014, "step": 7496 }, { "epoch": 0.2, "learning_rate": 1.0518228634327643e-05, "loss": 1.0035, "step": 7497 }, { "epoch": 0.2, "learning_rate": 1.0516075756542216e-05, "loss": 0.9697, "step": 7498 }, { "epoch": 0.2, "learning_rate": 1.0513922854773142e-05, "loss": 1.0047, "step": 7499 }, { "epoch": 0.2, "learning_rate": 1.051176992912047e-05, "loss": 0.9662, "step": 7500 }, { "epoch": 0.2, "learning_rate": 1.0509616979684252e-05, "loss": 0.9464, "step": 7501 }, { "epoch": 0.2, "learning_rate": 1.0507464006564545e-05, "loss": 0.9779, "step": 7502 }, { "epoch": 0.2, "learning_rate": 1.05053110098614e-05, "loss": 1.0032, "step": 7503 }, { "epoch": 0.2, "learning_rate": 1.050315798967488e-05, "loss": 0.9451, "step": 7504 }, { "epoch": 0.2, "learning_rate": 1.050100494610504e-05, "loss": 0.9562, "step": 7505 }, { "epoch": 0.2, "learning_rate": 1.0498851879251934e-05, "loss": 0.9753, "step": 7506 }, { "epoch": 0.2, "learning_rate": 1.0496698789215629e-05, "loss": 0.9707, "step": 7507 }, { "epoch": 0.2, "learning_rate": 1.0494545676096183e-05, "loss": 0.9769, "step": 7508 }, { "epoch": 0.2, "learning_rate": 1.0492392539993654e-05, "loss": 0.9929, "step": 7509 }, { "epoch": 0.2, "learning_rate": 1.0490239381008112e-05, "loss": 0.9404, "step": 7510 }, { "epoch": 0.2, "learning_rate": 1.0488086199239615e-05, "loss": 0.9407, "step": 7511 }, { "epoch": 0.2, "learning_rate": 1.0485932994788228e-05, "loss": 0.9442, "step": 7512 }, { "epoch": 0.2, "learning_rate": 1.0483779767754025e-05, "loss": 0.9581, "step": 7513 }, { "epoch": 0.2, "learning_rate": 1.0481626518237062e-05, "loss": 0.9565, "step": 7514 }, { "epoch": 0.2, "learning_rate": 1.0479473246337416e-05, "loss": 0.9926, "step": 7515 }, { "epoch": 0.2, "learning_rate": 1.0477319952155156e-05, "loss": 0.9507, "step": 7516 }, { "epoch": 0.2, "learning_rate": 1.0475166635790343e-05, "loss": 0.9648, "step": 7517 }, { "epoch": 0.2, "learning_rate": 1.0473013297343057e-05, "loss": 0.9812, "step": 7518 }, { "epoch": 0.2, "learning_rate": 1.0470859936913371e-05, "loss": 1.0261, "step": 7519 }, { "epoch": 0.2, "learning_rate": 1.0468706554601353e-05, "loss": 0.9801, "step": 7520 }, { "epoch": 0.2, "learning_rate": 1.046655315050708e-05, "loss": 0.9616, "step": 7521 }, { "epoch": 0.2, "learning_rate": 1.0464399724730627e-05, "loss": 0.9382, "step": 7522 }, { "epoch": 0.2, "learning_rate": 1.0462246277372072e-05, "loss": 0.96, "step": 7523 }, { "epoch": 0.2, "learning_rate": 1.046009280853149e-05, "loss": 0.9894, "step": 7524 }, { "epoch": 0.2, "learning_rate": 1.0457939318308963e-05, "loss": 1.0003, "step": 7525 }, { "epoch": 0.2, "learning_rate": 1.0455785806804564e-05, "loss": 0.9818, "step": 7526 }, { "epoch": 0.2, "learning_rate": 1.0453632274118376e-05, "loss": 0.9866, "step": 7527 }, { "epoch": 0.2, "learning_rate": 1.0451478720350488e-05, "loss": 0.9602, "step": 7528 }, { "epoch": 0.2, "learning_rate": 1.044932514560097e-05, "loss": 1.0091, "step": 7529 }, { "epoch": 0.2, "learning_rate": 1.0447171549969917e-05, "loss": 0.9895, "step": 7530 }, { "epoch": 0.2, "learning_rate": 1.0445017933557404e-05, "loss": 0.9962, "step": 7531 }, { "epoch": 0.2, "learning_rate": 1.044286429646352e-05, "loss": 0.9591, "step": 7532 }, { "epoch": 0.2, "learning_rate": 1.0440710638788354e-05, "loss": 0.9994, "step": 7533 }, { "epoch": 0.2, "learning_rate": 1.0438556960631987e-05, "loss": 0.9953, "step": 7534 }, { "epoch": 0.2, "learning_rate": 1.043640326209451e-05, "loss": 0.9948, "step": 7535 }, { "epoch": 0.2, "learning_rate": 1.0434249543276016e-05, "loss": 0.9682, "step": 7536 }, { "epoch": 0.2, "learning_rate": 1.0432095804276587e-05, "loss": 0.9826, "step": 7537 }, { "epoch": 0.2, "learning_rate": 1.0429942045196324e-05, "loss": 0.9963, "step": 7538 }, { "epoch": 0.2, "learning_rate": 1.0427788266135312e-05, "loss": 0.9726, "step": 7539 }, { "epoch": 0.2, "learning_rate": 1.0425634467193643e-05, "loss": 0.9222, "step": 7540 }, { "epoch": 0.2, "learning_rate": 1.0423480648471414e-05, "loss": 0.9735, "step": 7541 }, { "epoch": 0.2, "learning_rate": 1.042132681006872e-05, "loss": 0.9297, "step": 7542 }, { "epoch": 0.2, "learning_rate": 1.0419172952085651e-05, "loss": 0.9159, "step": 7543 }, { "epoch": 0.2, "learning_rate": 1.0417019074622312e-05, "loss": 0.951, "step": 7544 }, { "epoch": 0.2, "learning_rate": 1.0414865177778797e-05, "loss": 0.9831, "step": 7545 }, { "epoch": 0.2, "learning_rate": 1.0412711261655202e-05, "loss": 0.9817, "step": 7546 }, { "epoch": 0.2, "learning_rate": 1.0410557326351627e-05, "loss": 0.9887, "step": 7547 }, { "epoch": 0.2, "learning_rate": 1.0408403371968175e-05, "loss": 0.9773, "step": 7548 }, { "epoch": 0.2, "learning_rate": 1.0406249398604941e-05, "loss": 0.9341, "step": 7549 }, { "epoch": 0.2, "learning_rate": 1.0404095406362034e-05, "loss": 0.975, "step": 7550 }, { "epoch": 0.2, "learning_rate": 1.0401941395339552e-05, "loss": 0.9586, "step": 7551 }, { "epoch": 0.2, "learning_rate": 1.0399787365637599e-05, "loss": 0.9815, "step": 7552 }, { "epoch": 0.2, "learning_rate": 1.0397633317356283e-05, "loss": 1.0211, "step": 7553 }, { "epoch": 0.2, "learning_rate": 1.0395479250595709e-05, "loss": 0.9294, "step": 7554 }, { "epoch": 0.2, "learning_rate": 1.0393325165455975e-05, "loss": 0.9497, "step": 7555 }, { "epoch": 0.2, "learning_rate": 1.0391171062037195e-05, "loss": 0.9834, "step": 7556 }, { "epoch": 0.2, "learning_rate": 1.0389016940439479e-05, "loss": 0.9694, "step": 7557 }, { "epoch": 0.2, "learning_rate": 1.0386862800762932e-05, "loss": 0.9867, "step": 7558 }, { "epoch": 0.2, "learning_rate": 1.0384708643107664e-05, "loss": 0.9403, "step": 7559 }, { "epoch": 0.2, "learning_rate": 1.0382554467573785e-05, "loss": 0.9667, "step": 7560 }, { "epoch": 0.2, "learning_rate": 1.0380400274261406e-05, "loss": 1.0204, "step": 7561 }, { "epoch": 0.2, "learning_rate": 1.0378246063270643e-05, "loss": 0.984, "step": 7562 }, { "epoch": 0.2, "learning_rate": 1.0376091834701605e-05, "loss": 0.959, "step": 7563 }, { "epoch": 0.2, "learning_rate": 1.0373937588654404e-05, "loss": 0.9673, "step": 7564 }, { "epoch": 0.2, "learning_rate": 1.037178332522916e-05, "loss": 0.9991, "step": 7565 }, { "epoch": 0.2, "learning_rate": 1.0369629044525983e-05, "loss": 0.9714, "step": 7566 }, { "epoch": 0.2, "learning_rate": 1.0367474746644993e-05, "loss": 0.9643, "step": 7567 }, { "epoch": 0.2, "learning_rate": 1.0365320431686308e-05, "loss": 0.9662, "step": 7568 }, { "epoch": 0.2, "learning_rate": 1.0363166099750044e-05, "loss": 0.9728, "step": 7569 }, { "epoch": 0.2, "learning_rate": 1.0361011750936314e-05, "loss": 0.9541, "step": 7570 }, { "epoch": 0.2, "learning_rate": 1.0358857385345246e-05, "loss": 0.9577, "step": 7571 }, { "epoch": 0.2, "learning_rate": 1.0356703003076954e-05, "loss": 0.9571, "step": 7572 }, { "epoch": 0.2, "learning_rate": 1.0354548604231563e-05, "loss": 1.0036, "step": 7573 }, { "epoch": 0.2, "learning_rate": 1.0352394188909192e-05, "loss": 0.9681, "step": 7574 }, { "epoch": 0.2, "learning_rate": 1.0350239757209963e-05, "loss": 0.9304, "step": 7575 }, { "epoch": 0.2, "learning_rate": 1.0348085309234001e-05, "loss": 1.0342, "step": 7576 }, { "epoch": 0.2, "learning_rate": 1.0345930845081432e-05, "loss": 0.9792, "step": 7577 }, { "epoch": 0.2, "learning_rate": 1.0343776364852373e-05, "loss": 0.9671, "step": 7578 }, { "epoch": 0.2, "learning_rate": 1.0341621868646958e-05, "loss": 0.9684, "step": 7579 }, { "epoch": 0.2, "learning_rate": 1.0339467356565309e-05, "loss": 0.9575, "step": 7580 }, { "epoch": 0.2, "learning_rate": 1.033731282870755e-05, "loss": 0.9583, "step": 7581 }, { "epoch": 0.2, "learning_rate": 1.0335158285173819e-05, "loss": 0.9746, "step": 7582 }, { "epoch": 0.2, "learning_rate": 1.033300372606423e-05, "loss": 0.9467, "step": 7583 }, { "epoch": 0.2, "learning_rate": 1.0330849151478924e-05, "loss": 0.976, "step": 7584 }, { "epoch": 0.2, "learning_rate": 1.0328694561518028e-05, "loss": 0.9629, "step": 7585 }, { "epoch": 0.2, "learning_rate": 1.0326539956281665e-05, "loss": 0.962, "step": 7586 }, { "epoch": 0.2, "learning_rate": 1.0324385335869973e-05, "loss": 0.9455, "step": 7587 }, { "epoch": 0.2, "learning_rate": 1.0322230700383087e-05, "loss": 0.9769, "step": 7588 }, { "epoch": 0.2, "learning_rate": 1.0320076049921133e-05, "loss": 0.9718, "step": 7589 }, { "epoch": 0.2, "learning_rate": 1.0317921384584245e-05, "loss": 0.9474, "step": 7590 }, { "epoch": 0.2, "learning_rate": 1.0315766704472563e-05, "loss": 0.9495, "step": 7591 }, { "epoch": 0.21, "learning_rate": 1.0313612009686217e-05, "loss": 0.9564, "step": 7592 }, { "epoch": 0.21, "learning_rate": 1.0311457300325339e-05, "loss": 0.9637, "step": 7593 }, { "epoch": 0.21, "learning_rate": 1.0309302576490072e-05, "loss": 0.9856, "step": 7594 }, { "epoch": 0.21, "learning_rate": 1.030714783828055e-05, "loss": 0.9556, "step": 7595 }, { "epoch": 0.21, "learning_rate": 1.0304993085796909e-05, "loss": 0.9767, "step": 7596 }, { "epoch": 0.21, "learning_rate": 1.0302838319139288e-05, "loss": 0.9844, "step": 7597 }, { "epoch": 0.21, "learning_rate": 1.0300683538407826e-05, "loss": 0.9735, "step": 7598 }, { "epoch": 0.21, "learning_rate": 1.0298528743702664e-05, "loss": 0.9411, "step": 7599 }, { "epoch": 0.21, "learning_rate": 1.029637393512394e-05, "loss": 0.9646, "step": 7600 }, { "epoch": 0.21, "learning_rate": 1.0294219112771796e-05, "loss": 0.967, "step": 7601 }, { "epoch": 0.21, "learning_rate": 1.0292064276746368e-05, "loss": 0.9588, "step": 7602 }, { "epoch": 0.21, "learning_rate": 1.0289909427147806e-05, "loss": 0.9845, "step": 7603 }, { "epoch": 0.21, "learning_rate": 1.0287754564076249e-05, "loss": 1.0011, "step": 7604 }, { "epoch": 0.21, "learning_rate": 1.028559968763184e-05, "loss": 0.9741, "step": 7605 }, { "epoch": 0.21, "learning_rate": 1.028344479791472e-05, "loss": 0.9823, "step": 7606 }, { "epoch": 0.21, "learning_rate": 1.0281289895025037e-05, "loss": 0.9976, "step": 7607 }, { "epoch": 0.21, "learning_rate": 1.0279134979062938e-05, "loss": 0.9395, "step": 7608 }, { "epoch": 0.21, "learning_rate": 1.0276980050128567e-05, "loss": 0.9851, "step": 7609 }, { "epoch": 0.21, "learning_rate": 1.0274825108322067e-05, "loss": 0.9763, "step": 7610 }, { "epoch": 0.21, "learning_rate": 1.0272670153743589e-05, "loss": 0.968, "step": 7611 }, { "epoch": 0.21, "learning_rate": 1.0270515186493276e-05, "loss": 0.9823, "step": 7612 }, { "epoch": 0.21, "learning_rate": 1.026836020667128e-05, "loss": 0.9387, "step": 7613 }, { "epoch": 0.21, "learning_rate": 1.026620521437775e-05, "loss": 1.0318, "step": 7614 }, { "epoch": 0.21, "learning_rate": 1.0264050209712832e-05, "loss": 0.9804, "step": 7615 }, { "epoch": 0.21, "learning_rate": 1.0261895192776676e-05, "loss": 0.9832, "step": 7616 }, { "epoch": 0.21, "learning_rate": 1.0259740163669438e-05, "loss": 1.0019, "step": 7617 }, { "epoch": 0.21, "learning_rate": 1.0257585122491261e-05, "loss": 0.9858, "step": 7618 }, { "epoch": 0.21, "learning_rate": 1.0255430069342299e-05, "loss": 0.9789, "step": 7619 }, { "epoch": 0.21, "learning_rate": 1.025327500432271e-05, "loss": 1.0045, "step": 7620 }, { "epoch": 0.21, "learning_rate": 1.0251119927532635e-05, "loss": 0.9173, "step": 7621 }, { "epoch": 0.21, "learning_rate": 1.024896483907224e-05, "loss": 0.9785, "step": 7622 }, { "epoch": 0.21, "learning_rate": 1.0246809739041672e-05, "loss": 0.9657, "step": 7623 }, { "epoch": 0.21, "learning_rate": 1.0244654627541084e-05, "loss": 0.9771, "step": 7624 }, { "epoch": 0.21, "learning_rate": 1.024249950467063e-05, "loss": 1.0286, "step": 7625 }, { "epoch": 0.21, "learning_rate": 1.0240344370530473e-05, "loss": 0.9673, "step": 7626 }, { "epoch": 0.21, "learning_rate": 1.0238189225220761e-05, "loss": 1.0124, "step": 7627 }, { "epoch": 0.21, "learning_rate": 1.0236034068841654e-05, "loss": 0.9955, "step": 7628 }, { "epoch": 0.21, "learning_rate": 1.0233878901493305e-05, "loss": 0.9868, "step": 7629 }, { "epoch": 0.21, "learning_rate": 1.0231723723275877e-05, "loss": 0.971, "step": 7630 }, { "epoch": 0.21, "learning_rate": 1.0229568534289525e-05, "loss": 0.9683, "step": 7631 }, { "epoch": 0.21, "learning_rate": 1.0227413334634406e-05, "loss": 0.9669, "step": 7632 }, { "epoch": 0.21, "learning_rate": 1.0225258124410682e-05, "loss": 0.9818, "step": 7633 }, { "epoch": 0.21, "learning_rate": 1.0223102903718509e-05, "loss": 0.9565, "step": 7634 }, { "epoch": 0.21, "learning_rate": 1.0220947672658048e-05, "loss": 0.9945, "step": 7635 }, { "epoch": 0.21, "learning_rate": 1.0218792431329459e-05, "loss": 0.9307, "step": 7636 }, { "epoch": 0.21, "learning_rate": 1.0216637179832907e-05, "loss": 0.989, "step": 7637 }, { "epoch": 0.21, "learning_rate": 1.0214481918268548e-05, "loss": 0.9485, "step": 7638 }, { "epoch": 0.21, "learning_rate": 1.0212326646736543e-05, "loss": 1.0077, "step": 7639 }, { "epoch": 0.21, "learning_rate": 1.021017136533706e-05, "loss": 0.9506, "step": 7640 }, { "epoch": 0.21, "learning_rate": 1.0208016074170257e-05, "loss": 0.9689, "step": 7641 }, { "epoch": 0.21, "learning_rate": 1.0205860773336297e-05, "loss": 0.9346, "step": 7642 }, { "epoch": 0.21, "learning_rate": 1.0203705462935348e-05, "loss": 0.9553, "step": 7643 }, { "epoch": 0.21, "learning_rate": 1.0201550143067567e-05, "loss": 0.9951, "step": 7644 }, { "epoch": 0.21, "learning_rate": 1.0199394813833123e-05, "loss": 0.9719, "step": 7645 }, { "epoch": 0.21, "learning_rate": 1.0197239475332186e-05, "loss": 0.9748, "step": 7646 }, { "epoch": 0.21, "learning_rate": 1.0195084127664912e-05, "loss": 1.0206, "step": 7647 }, { "epoch": 0.21, "learning_rate": 1.0192928770931467e-05, "loss": 0.9245, "step": 7648 }, { "epoch": 0.21, "learning_rate": 1.0190773405232024e-05, "loss": 0.9691, "step": 7649 }, { "epoch": 0.21, "learning_rate": 1.0188618030666743e-05, "loss": 0.9692, "step": 7650 }, { "epoch": 0.21, "learning_rate": 1.0186462647335798e-05, "loss": 0.9883, "step": 7651 }, { "epoch": 0.21, "learning_rate": 1.018430725533935e-05, "loss": 0.9649, "step": 7652 }, { "epoch": 0.21, "learning_rate": 1.018215185477757e-05, "loss": 0.9362, "step": 7653 }, { "epoch": 0.21, "learning_rate": 1.0179996445750624e-05, "loss": 0.968, "step": 7654 }, { "epoch": 0.21, "learning_rate": 1.0177841028358683e-05, "loss": 0.9765, "step": 7655 }, { "epoch": 0.21, "learning_rate": 1.0175685602701915e-05, "loss": 0.9829, "step": 7656 }, { "epoch": 0.21, "learning_rate": 1.017353016888049e-05, "loss": 0.9694, "step": 7657 }, { "epoch": 0.21, "learning_rate": 1.0171374726994573e-05, "loss": 0.9867, "step": 7658 }, { "epoch": 0.21, "learning_rate": 1.016921927714434e-05, "loss": 0.9419, "step": 7659 }, { "epoch": 0.21, "learning_rate": 1.0167063819429961e-05, "loss": 0.9939, "step": 7660 }, { "epoch": 0.21, "learning_rate": 1.0164908353951606e-05, "loss": 0.9693, "step": 7661 }, { "epoch": 0.21, "learning_rate": 1.0162752880809444e-05, "loss": 0.9631, "step": 7662 }, { "epoch": 0.21, "learning_rate": 1.0160597400103649e-05, "loss": 0.9858, "step": 7663 }, { "epoch": 0.21, "learning_rate": 1.0158441911934388e-05, "loss": 0.9331, "step": 7664 }, { "epoch": 0.21, "learning_rate": 1.0156286416401842e-05, "loss": 0.9933, "step": 7665 }, { "epoch": 0.21, "learning_rate": 1.0154130913606178e-05, "loss": 0.9803, "step": 7666 }, { "epoch": 0.21, "learning_rate": 1.0151975403647568e-05, "loss": 0.9226, "step": 7667 }, { "epoch": 0.21, "learning_rate": 1.0149819886626187e-05, "loss": 0.9863, "step": 7668 }, { "epoch": 0.21, "learning_rate": 1.0147664362642214e-05, "loss": 0.9723, "step": 7669 }, { "epoch": 0.21, "learning_rate": 1.0145508831795812e-05, "loss": 0.9715, "step": 7670 }, { "epoch": 0.21, "learning_rate": 1.0143353294187162e-05, "loss": 0.9634, "step": 7671 }, { "epoch": 0.21, "learning_rate": 1.0141197749916438e-05, "loss": 0.9594, "step": 7672 }, { "epoch": 0.21, "learning_rate": 1.0139042199083812e-05, "loss": 0.9951, "step": 7673 }, { "epoch": 0.21, "learning_rate": 1.0136886641789467e-05, "loss": 0.9792, "step": 7674 }, { "epoch": 0.21, "learning_rate": 1.0134731078133568e-05, "loss": 1.005, "step": 7675 }, { "epoch": 0.21, "learning_rate": 1.0132575508216296e-05, "loss": 0.9453, "step": 7676 }, { "epoch": 0.21, "learning_rate": 1.013041993213783e-05, "loss": 0.9662, "step": 7677 }, { "epoch": 0.21, "learning_rate": 1.012826434999834e-05, "loss": 0.9611, "step": 7678 }, { "epoch": 0.21, "learning_rate": 1.0126108761898006e-05, "loss": 0.9711, "step": 7679 }, { "epoch": 0.21, "learning_rate": 1.0123953167937006e-05, "loss": 0.9889, "step": 7680 }, { "epoch": 0.21, "learning_rate": 1.0121797568215514e-05, "loss": 0.9744, "step": 7681 }, { "epoch": 0.21, "learning_rate": 1.011964196283371e-05, "loss": 1.0292, "step": 7682 }, { "epoch": 0.21, "learning_rate": 1.0117486351891772e-05, "loss": 0.9694, "step": 7683 }, { "epoch": 0.21, "learning_rate": 1.0115330735489877e-05, "loss": 0.9559, "step": 7684 }, { "epoch": 0.21, "learning_rate": 1.0113175113728202e-05, "loss": 0.9448, "step": 7685 }, { "epoch": 0.21, "learning_rate": 1.0111019486706928e-05, "loss": 0.9663, "step": 7686 }, { "epoch": 0.21, "learning_rate": 1.0108863854526229e-05, "loss": 0.9704, "step": 7687 }, { "epoch": 0.21, "learning_rate": 1.0106708217286287e-05, "loss": 0.9565, "step": 7688 }, { "epoch": 0.21, "learning_rate": 1.0104552575087285e-05, "loss": 0.9766, "step": 7689 }, { "epoch": 0.21, "learning_rate": 1.0102396928029396e-05, "loss": 1.0017, "step": 7690 }, { "epoch": 0.21, "learning_rate": 1.0100241276212803e-05, "loss": 0.9904, "step": 7691 }, { "epoch": 0.21, "learning_rate": 1.009808561973769e-05, "loss": 0.9167, "step": 7692 }, { "epoch": 0.21, "learning_rate": 1.0095929958704226e-05, "loss": 0.9232, "step": 7693 }, { "epoch": 0.21, "learning_rate": 1.0093774293212602e-05, "loss": 0.9483, "step": 7694 }, { "epoch": 0.21, "learning_rate": 1.0091618623362994e-05, "loss": 0.9928, "step": 7695 }, { "epoch": 0.21, "learning_rate": 1.008946294925558e-05, "loss": 0.9344, "step": 7696 }, { "epoch": 0.21, "learning_rate": 1.0087307270990543e-05, "loss": 0.9805, "step": 7697 }, { "epoch": 0.21, "learning_rate": 1.0085151588668066e-05, "loss": 0.9655, "step": 7698 }, { "epoch": 0.21, "learning_rate": 1.008299590238833e-05, "loss": 0.9624, "step": 7699 }, { "epoch": 0.21, "learning_rate": 1.0080840212251518e-05, "loss": 0.9373, "step": 7700 }, { "epoch": 0.21, "learning_rate": 1.007868451835781e-05, "loss": 0.9768, "step": 7701 }, { "epoch": 0.21, "learning_rate": 1.0076528820807382e-05, "loss": 0.9838, "step": 7702 }, { "epoch": 0.21, "learning_rate": 1.0074373119700422e-05, "loss": 0.9856, "step": 7703 }, { "epoch": 0.21, "learning_rate": 1.0072217415137112e-05, "loss": 0.9351, "step": 7704 }, { "epoch": 0.21, "learning_rate": 1.0070061707217634e-05, "loss": 1.0404, "step": 7705 }, { "epoch": 0.21, "learning_rate": 1.006790599604217e-05, "loss": 0.9687, "step": 7706 }, { "epoch": 0.21, "learning_rate": 1.0065750281710904e-05, "loss": 0.9609, "step": 7707 }, { "epoch": 0.21, "learning_rate": 1.0063594564324014e-05, "loss": 0.9598, "step": 7708 }, { "epoch": 0.21, "learning_rate": 1.006143884398169e-05, "loss": 0.9214, "step": 7709 }, { "epoch": 0.21, "learning_rate": 1.0059283120784107e-05, "loss": 0.9974, "step": 7710 }, { "epoch": 0.21, "learning_rate": 1.0057127394831453e-05, "loss": 0.9689, "step": 7711 }, { "epoch": 0.21, "learning_rate": 1.0054971666223915e-05, "loss": 0.9461, "step": 7712 }, { "epoch": 0.21, "learning_rate": 1.0052815935061669e-05, "loss": 0.9218, "step": 7713 }, { "epoch": 0.21, "learning_rate": 1.0050660201444902e-05, "loss": 0.9662, "step": 7714 }, { "epoch": 0.21, "learning_rate": 1.0048504465473801e-05, "loss": 0.9823, "step": 7715 }, { "epoch": 0.21, "learning_rate": 1.0046348727248543e-05, "loss": 0.9552, "step": 7716 }, { "epoch": 0.21, "learning_rate": 1.0044192986869314e-05, "loss": 1.007, "step": 7717 }, { "epoch": 0.21, "learning_rate": 1.0042037244436303e-05, "loss": 0.9411, "step": 7718 }, { "epoch": 0.21, "learning_rate": 1.0039881500049687e-05, "loss": 0.9648, "step": 7719 }, { "epoch": 0.21, "learning_rate": 1.0037725753809654e-05, "loss": 0.9774, "step": 7720 }, { "epoch": 0.21, "learning_rate": 1.0035570005816386e-05, "loss": 1.0109, "step": 7721 }, { "epoch": 0.21, "learning_rate": 1.0033414256170071e-05, "loss": 0.9738, "step": 7722 }, { "epoch": 0.21, "learning_rate": 1.0031258504970894e-05, "loss": 0.9464, "step": 7723 }, { "epoch": 0.21, "learning_rate": 1.0029102752319033e-05, "loss": 0.9746, "step": 7724 }, { "epoch": 0.21, "learning_rate": 1.0026946998314678e-05, "loss": 0.9723, "step": 7725 }, { "epoch": 0.21, "learning_rate": 1.002479124305801e-05, "loss": 0.9683, "step": 7726 }, { "epoch": 0.21, "learning_rate": 1.0022635486649217e-05, "loss": 0.9537, "step": 7727 }, { "epoch": 0.21, "learning_rate": 1.0020479729188481e-05, "loss": 0.9915, "step": 7728 }, { "epoch": 0.21, "learning_rate": 1.0018323970775989e-05, "loss": 0.961, "step": 7729 }, { "epoch": 0.21, "learning_rate": 1.0016168211511928e-05, "loss": 0.986, "step": 7730 }, { "epoch": 0.21, "learning_rate": 1.0014012451496476e-05, "loss": 1.0138, "step": 7731 }, { "epoch": 0.21, "learning_rate": 1.0011856690829822e-05, "loss": 0.9875, "step": 7732 }, { "epoch": 0.21, "learning_rate": 1.0009700929612151e-05, "loss": 0.9529, "step": 7733 }, { "epoch": 0.21, "learning_rate": 1.0007545167943647e-05, "loss": 0.9641, "step": 7734 }, { "epoch": 0.21, "learning_rate": 1.0005389405924498e-05, "loss": 0.9153, "step": 7735 }, { "epoch": 0.21, "learning_rate": 1.0003233643654883e-05, "loss": 0.9882, "step": 7736 }, { "epoch": 0.21, "learning_rate": 1.0001077881234991e-05, "loss": 0.9427, "step": 7737 }, { "epoch": 0.21, "learning_rate": 9.99892211876501e-06, "loss": 0.971, "step": 7738 }, { "epoch": 0.21, "learning_rate": 9.99676635634512e-06, "loss": 0.9631, "step": 7739 }, { "epoch": 0.21, "learning_rate": 9.994610594075505e-06, "loss": 0.963, "step": 7740 }, { "epoch": 0.21, "learning_rate": 9.992454832056357e-06, "loss": 0.9753, "step": 7741 }, { "epoch": 0.21, "learning_rate": 9.990299070387854e-06, "loss": 0.9518, "step": 7742 }, { "epoch": 0.22, "learning_rate": 9.98814330917018e-06, "loss": 0.9787, "step": 7743 }, { "epoch": 0.22, "learning_rate": 9.985987548503529e-06, "loss": 0.9838, "step": 7744 }, { "epoch": 0.22, "learning_rate": 9.983831788488075e-06, "loss": 0.9681, "step": 7745 }, { "epoch": 0.22, "learning_rate": 9.981676029224013e-06, "loss": 0.9241, "step": 7746 }, { "epoch": 0.22, "learning_rate": 9.97952027081152e-06, "loss": 0.983, "step": 7747 }, { "epoch": 0.22, "learning_rate": 9.977364513350788e-06, "loss": 0.9283, "step": 7748 }, { "epoch": 0.22, "learning_rate": 9.975208756941992e-06, "loss": 1.0116, "step": 7749 }, { "epoch": 0.22, "learning_rate": 9.973053001685327e-06, "loss": 0.9495, "step": 7750 }, { "epoch": 0.22, "learning_rate": 9.970897247680972e-06, "loss": 0.9771, "step": 7751 }, { "epoch": 0.22, "learning_rate": 9.968741495029111e-06, "loss": 0.9832, "step": 7752 }, { "epoch": 0.22, "learning_rate": 9.966585743829932e-06, "loss": 0.9742, "step": 7753 }, { "epoch": 0.22, "learning_rate": 9.964429994183614e-06, "loss": 0.9398, "step": 7754 }, { "epoch": 0.22, "learning_rate": 9.962274246190351e-06, "loss": 1.037, "step": 7755 }, { "epoch": 0.22, "learning_rate": 9.960118499950313e-06, "loss": 0.9729, "step": 7756 }, { "epoch": 0.22, "learning_rate": 9.9579627555637e-06, "loss": 0.9156, "step": 7757 }, { "epoch": 0.22, "learning_rate": 9.955807013130689e-06, "loss": 0.9712, "step": 7758 }, { "epoch": 0.22, "learning_rate": 9.95365127275146e-06, "loss": 0.9848, "step": 7759 }, { "epoch": 0.22, "learning_rate": 9.951495534526204e-06, "loss": 0.9377, "step": 7760 }, { "epoch": 0.22, "learning_rate": 9.949339798555098e-06, "loss": 1.0124, "step": 7761 }, { "epoch": 0.22, "learning_rate": 9.947184064938333e-06, "loss": 0.9967, "step": 7762 }, { "epoch": 0.22, "learning_rate": 9.945028333776088e-06, "loss": 0.9719, "step": 7763 }, { "epoch": 0.22, "learning_rate": 9.942872605168548e-06, "loss": 0.9839, "step": 7764 }, { "epoch": 0.22, "learning_rate": 9.940716879215898e-06, "loss": 0.9948, "step": 7765 }, { "epoch": 0.22, "learning_rate": 9.938561156018314e-06, "loss": 0.9846, "step": 7766 }, { "epoch": 0.22, "learning_rate": 9.936405435675991e-06, "loss": 0.9313, "step": 7767 }, { "epoch": 0.22, "learning_rate": 9.934249718289099e-06, "loss": 0.9868, "step": 7768 }, { "epoch": 0.22, "learning_rate": 9.932094003957832e-06, "loss": 0.992, "step": 7769 }, { "epoch": 0.22, "learning_rate": 9.929938292782368e-06, "loss": 0.9474, "step": 7770 }, { "epoch": 0.22, "learning_rate": 9.92778258486289e-06, "loss": 0.9788, "step": 7771 }, { "epoch": 0.22, "learning_rate": 9.92562688029958e-06, "loss": 1.0056, "step": 7772 }, { "epoch": 0.22, "learning_rate": 9.923471179192622e-06, "loss": 0.9813, "step": 7773 }, { "epoch": 0.22, "learning_rate": 9.921315481642197e-06, "loss": 0.9702, "step": 7774 }, { "epoch": 0.22, "learning_rate": 9.919159787748485e-06, "loss": 0.9522, "step": 7775 }, { "epoch": 0.22, "learning_rate": 9.917004097611671e-06, "loss": 0.9538, "step": 7776 }, { "epoch": 0.22, "learning_rate": 9.914848411331933e-06, "loss": 0.9929, "step": 7777 }, { "epoch": 0.22, "learning_rate": 9.91269272900946e-06, "loss": 0.9834, "step": 7778 }, { "epoch": 0.22, "learning_rate": 9.910537050744421e-06, "loss": 0.9583, "step": 7779 }, { "epoch": 0.22, "learning_rate": 9.908381376637011e-06, "loss": 0.9566, "step": 7780 }, { "epoch": 0.22, "learning_rate": 9.906225706787402e-06, "loss": 0.9753, "step": 7781 }, { "epoch": 0.22, "learning_rate": 9.904070041295775e-06, "loss": 0.9483, "step": 7782 }, { "epoch": 0.22, "learning_rate": 9.901914380262313e-06, "loss": 0.9989, "step": 7783 }, { "epoch": 0.22, "learning_rate": 9.899758723787197e-06, "loss": 0.9857, "step": 7784 }, { "epoch": 0.22, "learning_rate": 9.897603071970605e-06, "loss": 0.9714, "step": 7785 }, { "epoch": 0.22, "learning_rate": 9.895447424912716e-06, "loss": 0.9465, "step": 7786 }, { "epoch": 0.22, "learning_rate": 9.893291782713715e-06, "loss": 0.9992, "step": 7787 }, { "epoch": 0.22, "learning_rate": 9.891136145473775e-06, "loss": 0.9709, "step": 7788 }, { "epoch": 0.22, "learning_rate": 9.888980513293076e-06, "loss": 0.9442, "step": 7789 }, { "epoch": 0.22, "learning_rate": 9.886824886271803e-06, "loss": 0.9467, "step": 7790 }, { "epoch": 0.22, "learning_rate": 9.884669264510125e-06, "loss": 0.9717, "step": 7791 }, { "epoch": 0.22, "learning_rate": 9.88251364810823e-06, "loss": 0.9564, "step": 7792 }, { "epoch": 0.22, "learning_rate": 9.88035803716629e-06, "loss": 0.9637, "step": 7793 }, { "epoch": 0.22, "learning_rate": 9.878202431784488e-06, "loss": 0.9703, "step": 7794 }, { "epoch": 0.22, "learning_rate": 9.876046832062996e-06, "loss": 0.9493, "step": 7795 }, { "epoch": 0.22, "learning_rate": 9.873891238101996e-06, "loss": 0.9407, "step": 7796 }, { "epoch": 0.22, "learning_rate": 9.871735650001664e-06, "loss": 0.9505, "step": 7797 }, { "epoch": 0.22, "learning_rate": 9.869580067862174e-06, "loss": 0.9617, "step": 7798 }, { "epoch": 0.22, "learning_rate": 9.867424491783708e-06, "loss": 0.952, "step": 7799 }, { "epoch": 0.22, "learning_rate": 9.865268921866434e-06, "loss": 0.9374, "step": 7800 }, { "epoch": 0.22, "learning_rate": 9.863113358210536e-06, "loss": 0.9594, "step": 7801 }, { "epoch": 0.22, "learning_rate": 9.860957800916186e-06, "loss": 1.0221, "step": 7802 }, { "epoch": 0.22, "learning_rate": 9.858802250083564e-06, "loss": 0.9683, "step": 7803 }, { "epoch": 0.22, "learning_rate": 9.856646705812842e-06, "loss": 0.977, "step": 7804 }, { "epoch": 0.22, "learning_rate": 9.854491168204191e-06, "loss": 0.9758, "step": 7805 }, { "epoch": 0.22, "learning_rate": 9.852335637357791e-06, "loss": 0.9461, "step": 7806 }, { "epoch": 0.22, "learning_rate": 9.850180113373815e-06, "loss": 0.9769, "step": 7807 }, { "epoch": 0.22, "learning_rate": 9.848024596352434e-06, "loss": 0.9418, "step": 7808 }, { "epoch": 0.22, "learning_rate": 9.845869086393824e-06, "loss": 0.9632, "step": 7809 }, { "epoch": 0.22, "learning_rate": 9.843713583598163e-06, "loss": 0.9714, "step": 7810 }, { "epoch": 0.22, "learning_rate": 9.841558088065617e-06, "loss": 0.9989, "step": 7811 }, { "epoch": 0.22, "learning_rate": 9.839402599896355e-06, "loss": 0.9744, "step": 7812 }, { "epoch": 0.22, "learning_rate": 9.837247119190561e-06, "loss": 0.9465, "step": 7813 }, { "epoch": 0.22, "learning_rate": 9.835091646048397e-06, "loss": 0.9594, "step": 7814 }, { "epoch": 0.22, "learning_rate": 9.832936180570042e-06, "loss": 0.9661, "step": 7815 }, { "epoch": 0.22, "learning_rate": 9.83078072285566e-06, "loss": 0.9012, "step": 7816 }, { "epoch": 0.22, "learning_rate": 9.82862527300543e-06, "loss": 0.9883, "step": 7817 }, { "epoch": 0.22, "learning_rate": 9.826469831119514e-06, "loss": 0.9889, "step": 7818 }, { "epoch": 0.22, "learning_rate": 9.824314397298088e-06, "loss": 0.9586, "step": 7819 }, { "epoch": 0.22, "learning_rate": 9.822158971641322e-06, "loss": 0.9478, "step": 7820 }, { "epoch": 0.22, "learning_rate": 9.82000355424938e-06, "loss": 0.9812, "step": 7821 }, { "epoch": 0.22, "learning_rate": 9.817848145222436e-06, "loss": 0.9254, "step": 7822 }, { "epoch": 0.22, "learning_rate": 9.815692744660652e-06, "loss": 0.9716, "step": 7823 }, { "epoch": 0.22, "learning_rate": 9.813537352664204e-06, "loss": 0.9723, "step": 7824 }, { "epoch": 0.22, "learning_rate": 9.811381969333257e-06, "loss": 1.0011, "step": 7825 }, { "epoch": 0.22, "learning_rate": 9.809226594767979e-06, "loss": 0.958, "step": 7826 }, { "epoch": 0.22, "learning_rate": 9.807071229068536e-06, "loss": 1.0089, "step": 7827 }, { "epoch": 0.22, "learning_rate": 9.804915872335093e-06, "loss": 0.9971, "step": 7828 }, { "epoch": 0.22, "learning_rate": 9.802760524667818e-06, "loss": 0.9804, "step": 7829 }, { "epoch": 0.22, "learning_rate": 9.800605186166875e-06, "loss": 0.9471, "step": 7830 }, { "epoch": 0.22, "learning_rate": 9.798449856932435e-06, "loss": 0.9635, "step": 7831 }, { "epoch": 0.22, "learning_rate": 9.796294537064654e-06, "loss": 0.9145, "step": 7832 }, { "epoch": 0.22, "learning_rate": 9.794139226663706e-06, "loss": 0.9884, "step": 7833 }, { "epoch": 0.22, "learning_rate": 9.791983925829748e-06, "loss": 0.9633, "step": 7834 }, { "epoch": 0.22, "learning_rate": 9.789828634662942e-06, "loss": 0.9935, "step": 7835 }, { "epoch": 0.22, "learning_rate": 9.78767335326346e-06, "loss": 0.9969, "step": 7836 }, { "epoch": 0.22, "learning_rate": 9.785518081731455e-06, "loss": 0.9949, "step": 7837 }, { "epoch": 0.22, "learning_rate": 9.783362820167096e-06, "loss": 0.9495, "step": 7838 }, { "epoch": 0.22, "learning_rate": 9.781207568670541e-06, "loss": 0.9568, "step": 7839 }, { "epoch": 0.22, "learning_rate": 9.779052327341955e-06, "loss": 1.0171, "step": 7840 }, { "epoch": 0.22, "learning_rate": 9.776897096281493e-06, "loss": 0.9544, "step": 7841 }, { "epoch": 0.22, "learning_rate": 9.774741875589322e-06, "loss": 0.9833, "step": 7842 }, { "epoch": 0.22, "learning_rate": 9.772586665365597e-06, "loss": 0.9753, "step": 7843 }, { "epoch": 0.22, "learning_rate": 9.770431465710478e-06, "loss": 0.987, "step": 7844 }, { "epoch": 0.22, "learning_rate": 9.768276276724126e-06, "loss": 0.9512, "step": 7845 }, { "epoch": 0.22, "learning_rate": 9.766121098506695e-06, "loss": 0.973, "step": 7846 }, { "epoch": 0.22, "learning_rate": 9.76396593115835e-06, "loss": 0.9739, "step": 7847 }, { "epoch": 0.22, "learning_rate": 9.76181077477924e-06, "loss": 0.9805, "step": 7848 }, { "epoch": 0.22, "learning_rate": 9.75965562946953e-06, "loss": 1.0144, "step": 7849 }, { "epoch": 0.22, "learning_rate": 9.757500495329371e-06, "loss": 0.9961, "step": 7850 }, { "epoch": 0.22, "learning_rate": 9.75534537245892e-06, "loss": 0.9844, "step": 7851 }, { "epoch": 0.22, "learning_rate": 9.753190260958333e-06, "loss": 0.9811, "step": 7852 }, { "epoch": 0.22, "learning_rate": 9.751035160927762e-06, "loss": 0.9488, "step": 7853 }, { "epoch": 0.22, "learning_rate": 9.748880072467366e-06, "loss": 0.9485, "step": 7854 }, { "epoch": 0.22, "learning_rate": 9.746724995677292e-06, "loss": 0.965, "step": 7855 }, { "epoch": 0.22, "learning_rate": 9.744569930657703e-06, "loss": 0.9848, "step": 7856 }, { "epoch": 0.22, "learning_rate": 9.742414877508744e-06, "loss": 0.9407, "step": 7857 }, { "epoch": 0.22, "learning_rate": 9.740259836330566e-06, "loss": 0.9563, "step": 7858 }, { "epoch": 0.22, "learning_rate": 9.738104807223329e-06, "loss": 0.9618, "step": 7859 }, { "epoch": 0.22, "learning_rate": 9.735949790287171e-06, "loss": 0.9393, "step": 7860 }, { "epoch": 0.22, "learning_rate": 9.733794785622254e-06, "loss": 0.9575, "step": 7861 }, { "epoch": 0.22, "learning_rate": 9.73163979332872e-06, "loss": 0.9886, "step": 7862 }, { "epoch": 0.22, "learning_rate": 9.729484813506726e-06, "loss": 0.971, "step": 7863 }, { "epoch": 0.22, "learning_rate": 9.727329846256413e-06, "loss": 0.9677, "step": 7864 }, { "epoch": 0.22, "learning_rate": 9.725174891677937e-06, "loss": 0.9675, "step": 7865 }, { "epoch": 0.22, "learning_rate": 9.723019949871438e-06, "loss": 0.9467, "step": 7866 }, { "epoch": 0.22, "learning_rate": 9.720865020937063e-06, "loss": 0.9792, "step": 7867 }, { "epoch": 0.22, "learning_rate": 9.718710104974965e-06, "loss": 0.9075, "step": 7868 }, { "epoch": 0.22, "learning_rate": 9.716555202085281e-06, "loss": 0.987, "step": 7869 }, { "epoch": 0.22, "learning_rate": 9.714400312368165e-06, "loss": 0.9368, "step": 7870 }, { "epoch": 0.22, "learning_rate": 9.712245435923753e-06, "loss": 1.0438, "step": 7871 }, { "epoch": 0.22, "learning_rate": 9.710090572852197e-06, "loss": 0.9522, "step": 7872 }, { "epoch": 0.22, "learning_rate": 9.707935723253634e-06, "loss": 0.9477, "step": 7873 }, { "epoch": 0.22, "learning_rate": 9.70578088722821e-06, "loss": 0.9344, "step": 7874 }, { "epoch": 0.22, "learning_rate": 9.703626064876064e-06, "loss": 0.9904, "step": 7875 }, { "epoch": 0.22, "learning_rate": 9.70147125629734e-06, "loss": 0.9752, "step": 7876 }, { "epoch": 0.22, "learning_rate": 9.699316461592176e-06, "loss": 1.005, "step": 7877 }, { "epoch": 0.22, "learning_rate": 9.697161680860713e-06, "loss": 0.9687, "step": 7878 }, { "epoch": 0.22, "learning_rate": 9.695006914203096e-06, "loss": 0.9168, "step": 7879 }, { "epoch": 0.22, "learning_rate": 9.692852161719456e-06, "loss": 0.9836, "step": 7880 }, { "epoch": 0.22, "learning_rate": 9.69069742350993e-06, "loss": 0.9866, "step": 7881 }, { "epoch": 0.22, "learning_rate": 9.688542699674663e-06, "loss": 0.9499, "step": 7882 }, { "epoch": 0.22, "learning_rate": 9.686387990313787e-06, "loss": 1.0065, "step": 7883 }, { "epoch": 0.22, "learning_rate": 9.68423329552744e-06, "loss": 0.9172, "step": 7884 }, { "epoch": 0.22, "learning_rate": 9.682078615415755e-06, "loss": 0.9898, "step": 7885 }, { "epoch": 0.22, "learning_rate": 9.67992395007887e-06, "loss": 0.9514, "step": 7886 }, { "epoch": 0.22, "learning_rate": 9.677769299616917e-06, "loss": 0.9773, "step": 7887 }, { "epoch": 0.22, "learning_rate": 9.675614664130029e-06, "loss": 0.9578, "step": 7888 }, { "epoch": 0.22, "learning_rate": 9.67346004371834e-06, "loss": 0.9399, "step": 7889 }, { "epoch": 0.22, "learning_rate": 9.671305438481976e-06, "loss": 0.9338, "step": 7890 }, { "epoch": 0.22, "learning_rate": 9.669150848521081e-06, "loss": 0.9639, "step": 7891 }, { "epoch": 0.22, "learning_rate": 9.666996273935771e-06, "loss": 0.9383, "step": 7892 }, { "epoch": 0.23, "learning_rate": 9.664841714826185e-06, "loss": 1.0019, "step": 7893 }, { "epoch": 0.23, "learning_rate": 9.662687171292448e-06, "loss": 0.9553, "step": 7894 }, { "epoch": 0.23, "learning_rate": 9.660532643434693e-06, "loss": 0.9894, "step": 7895 }, { "epoch": 0.23, "learning_rate": 9.658378131353045e-06, "loss": 1.0083, "step": 7896 }, { "epoch": 0.23, "learning_rate": 9.656223635147629e-06, "loss": 0.9713, "step": 7897 }, { "epoch": 0.23, "learning_rate": 9.654069154918574e-06, "loss": 0.9969, "step": 7898 }, { "epoch": 0.23, "learning_rate": 9.651914690766e-06, "loss": 1.0003, "step": 7899 }, { "epoch": 0.23, "learning_rate": 9.64976024279004e-06, "loss": 0.9584, "step": 7900 }, { "epoch": 0.23, "learning_rate": 9.64760581109081e-06, "loss": 0.9787, "step": 7901 }, { "epoch": 0.23, "learning_rate": 9.645451395768442e-06, "loss": 0.9281, "step": 7902 }, { "epoch": 0.23, "learning_rate": 9.643296996923051e-06, "loss": 0.9591, "step": 7903 }, { "epoch": 0.23, "learning_rate": 9.641142614654755e-06, "loss": 0.9832, "step": 7904 }, { "epoch": 0.23, "learning_rate": 9.638988249063688e-06, "loss": 0.9598, "step": 7905 }, { "epoch": 0.23, "learning_rate": 9.63683390024996e-06, "loss": 0.9532, "step": 7906 }, { "epoch": 0.23, "learning_rate": 9.634679568313693e-06, "loss": 0.9608, "step": 7907 }, { "epoch": 0.23, "learning_rate": 9.632525253355006e-06, "loss": 0.9951, "step": 7908 }, { "epoch": 0.23, "learning_rate": 9.630370955474018e-06, "loss": 0.9834, "step": 7909 }, { "epoch": 0.23, "learning_rate": 9.628216674770841e-06, "loss": 0.9488, "step": 7910 }, { "epoch": 0.23, "learning_rate": 9.6260624113456e-06, "loss": 0.9455, "step": 7911 }, { "epoch": 0.23, "learning_rate": 9.623908165298401e-06, "loss": 0.9459, "step": 7912 }, { "epoch": 0.23, "learning_rate": 9.621753936729359e-06, "loss": 0.9848, "step": 7913 }, { "epoch": 0.23, "learning_rate": 9.619599725738597e-06, "loss": 0.9615, "step": 7914 }, { "epoch": 0.23, "learning_rate": 9.617445532426217e-06, "loss": 1.014, "step": 7915 }, { "epoch": 0.23, "learning_rate": 9.61529135689234e-06, "loss": 1.0341, "step": 7916 }, { "epoch": 0.23, "learning_rate": 9.61313719923707e-06, "loss": 0.9698, "step": 7917 }, { "epoch": 0.23, "learning_rate": 9.610983059560523e-06, "loss": 0.9237, "step": 7918 }, { "epoch": 0.23, "learning_rate": 9.608828937962807e-06, "loss": 0.9873, "step": 7919 }, { "epoch": 0.23, "learning_rate": 9.606674834544027e-06, "loss": 0.981, "step": 7920 }, { "epoch": 0.23, "learning_rate": 9.604520749404298e-06, "loss": 1.0105, "step": 7921 }, { "epoch": 0.23, "learning_rate": 9.602366682643719e-06, "loss": 0.9332, "step": 7922 }, { "epoch": 0.23, "learning_rate": 9.600212634362403e-06, "loss": 0.9121, "step": 7923 }, { "epoch": 0.23, "learning_rate": 9.598058604660448e-06, "loss": 0.9398, "step": 7924 }, { "epoch": 0.23, "learning_rate": 9.59590459363797e-06, "loss": 0.9923, "step": 7925 }, { "epoch": 0.23, "learning_rate": 9.593750601395064e-06, "loss": 0.9481, "step": 7926 }, { "epoch": 0.23, "learning_rate": 9.591596628031829e-06, "loss": 0.966, "step": 7927 }, { "epoch": 0.23, "learning_rate": 9.589442673648376e-06, "loss": 0.978, "step": 7928 }, { "epoch": 0.23, "learning_rate": 9.5872887383448e-06, "loss": 0.9617, "step": 7929 }, { "epoch": 0.23, "learning_rate": 9.585134822221206e-06, "loss": 0.9519, "step": 7930 }, { "epoch": 0.23, "learning_rate": 9.58298092537769e-06, "loss": 1.0143, "step": 7931 }, { "epoch": 0.23, "learning_rate": 9.58082704791435e-06, "loss": 0.943, "step": 7932 }, { "epoch": 0.23, "learning_rate": 9.578673189931283e-06, "loss": 1.0403, "step": 7933 }, { "epoch": 0.23, "learning_rate": 9.57651935152859e-06, "loss": 0.9638, "step": 7934 }, { "epoch": 0.23, "learning_rate": 9.574365532806362e-06, "loss": 0.9716, "step": 7935 }, { "epoch": 0.23, "learning_rate": 9.572211733864692e-06, "loss": 0.9767, "step": 7936 }, { "epoch": 0.23, "learning_rate": 9.57005795480368e-06, "loss": 0.9876, "step": 7937 }, { "epoch": 0.23, "learning_rate": 9.567904195723411e-06, "loss": 0.9747, "step": 7938 }, { "epoch": 0.23, "learning_rate": 9.565750456723988e-06, "loss": 0.9544, "step": 7939 }, { "epoch": 0.23, "learning_rate": 9.56359673790549e-06, "loss": 0.9686, "step": 7940 }, { "epoch": 0.23, "learning_rate": 9.561443039368017e-06, "loss": 0.9829, "step": 7941 }, { "epoch": 0.23, "learning_rate": 9.559289361211652e-06, "loss": 0.9717, "step": 7942 }, { "epoch": 0.23, "learning_rate": 9.557135703536482e-06, "loss": 1.0047, "step": 7943 }, { "epoch": 0.23, "learning_rate": 9.554982066442601e-06, "loss": 1.0356, "step": 7944 }, { "epoch": 0.23, "learning_rate": 9.552828450030087e-06, "loss": 0.9871, "step": 7945 }, { "epoch": 0.23, "learning_rate": 9.550674854399032e-06, "loss": 0.9288, "step": 7946 }, { "epoch": 0.23, "learning_rate": 9.548521279649512e-06, "loss": 0.9676, "step": 7947 }, { "epoch": 0.23, "learning_rate": 9.546367725881625e-06, "loss": 0.9322, "step": 7948 }, { "epoch": 0.23, "learning_rate": 9.544214193195441e-06, "loss": 0.9783, "step": 7949 }, { "epoch": 0.23, "learning_rate": 9.54206068169104e-06, "loss": 0.9816, "step": 7950 }, { "epoch": 0.23, "learning_rate": 9.539907191468512e-06, "loss": 1.0046, "step": 7951 }, { "epoch": 0.23, "learning_rate": 9.537753722627931e-06, "loss": 0.9476, "step": 7952 }, { "epoch": 0.23, "learning_rate": 9.535600275269376e-06, "loss": 0.9855, "step": 7953 }, { "epoch": 0.23, "learning_rate": 9.533446849492922e-06, "loss": 0.9771, "step": 7954 }, { "epoch": 0.23, "learning_rate": 9.53129344539865e-06, "loss": 0.9708, "step": 7955 }, { "epoch": 0.23, "learning_rate": 9.529140063086632e-06, "loss": 0.9899, "step": 7956 }, { "epoch": 0.23, "learning_rate": 9.526986702656944e-06, "loss": 0.9613, "step": 7957 }, { "epoch": 0.23, "learning_rate": 9.52483336420966e-06, "loss": 0.9764, "step": 7958 }, { "epoch": 0.23, "learning_rate": 9.522680047844847e-06, "loss": 0.9684, "step": 7959 }, { "epoch": 0.23, "learning_rate": 9.520526753662587e-06, "loss": 0.9288, "step": 7960 }, { "epoch": 0.23, "learning_rate": 9.518373481762938e-06, "loss": 0.9731, "step": 7961 }, { "epoch": 0.23, "learning_rate": 9.516220232245979e-06, "loss": 0.9182, "step": 7962 }, { "epoch": 0.23, "learning_rate": 9.51406700521177e-06, "loss": 0.9719, "step": 7963 }, { "epoch": 0.23, "learning_rate": 9.511913800760388e-06, "loss": 0.9475, "step": 7964 }, { "epoch": 0.23, "learning_rate": 9.509760618991892e-06, "loss": 0.9526, "step": 7965 }, { "epoch": 0.23, "learning_rate": 9.507607460006347e-06, "loss": 0.9999, "step": 7966 }, { "epoch": 0.23, "learning_rate": 9.505454323903822e-06, "loss": 0.9737, "step": 7967 }, { "epoch": 0.23, "learning_rate": 9.503301210784373e-06, "loss": 0.9522, "step": 7968 }, { "epoch": 0.23, "learning_rate": 9.501148120748068e-06, "loss": 0.9537, "step": 7969 }, { "epoch": 0.23, "learning_rate": 9.498995053894961e-06, "loss": 0.9813, "step": 7970 }, { "epoch": 0.23, "learning_rate": 9.496842010325123e-06, "loss": 1.0055, "step": 7971 }, { "epoch": 0.23, "learning_rate": 9.494688990138603e-06, "loss": 0.9226, "step": 7972 }, { "epoch": 0.23, "learning_rate": 9.492535993435459e-06, "loss": 0.949, "step": 7973 }, { "epoch": 0.23, "learning_rate": 9.490383020315751e-06, "loss": 0.9762, "step": 7974 }, { "epoch": 0.23, "learning_rate": 9.488230070879533e-06, "loss": 0.9648, "step": 7975 }, { "epoch": 0.23, "learning_rate": 9.486077145226861e-06, "loss": 0.9716, "step": 7976 }, { "epoch": 0.23, "learning_rate": 9.483924243457786e-06, "loss": 0.9317, "step": 7977 }, { "epoch": 0.23, "learning_rate": 9.48177136567236e-06, "loss": 0.9498, "step": 7978 }, { "epoch": 0.23, "learning_rate": 9.479618511970634e-06, "loss": 0.9421, "step": 7979 }, { "epoch": 0.23, "learning_rate": 9.47746568245266e-06, "loss": 0.9515, "step": 7980 }, { "epoch": 0.23, "learning_rate": 9.475312877218486e-06, "loss": 0.9689, "step": 7981 }, { "epoch": 0.23, "learning_rate": 9.473160096368153e-06, "loss": 0.9452, "step": 7982 }, { "epoch": 0.23, "learning_rate": 9.471007340001722e-06, "loss": 0.9533, "step": 7983 }, { "epoch": 0.23, "learning_rate": 9.46885460821922e-06, "loss": 0.9504, "step": 7984 }, { "epoch": 0.23, "learning_rate": 9.466701901120707e-06, "loss": 1.017, "step": 7985 }, { "epoch": 0.23, "learning_rate": 9.464549218806218e-06, "loss": 0.9512, "step": 7986 }, { "epoch": 0.23, "learning_rate": 9.462396561375797e-06, "loss": 0.9675, "step": 7987 }, { "epoch": 0.23, "learning_rate": 9.460243928929485e-06, "loss": 0.9561, "step": 7988 }, { "epoch": 0.23, "learning_rate": 9.45809132156732e-06, "loss": 0.9774, "step": 7989 }, { "epoch": 0.23, "learning_rate": 9.455938739389342e-06, "loss": 0.9611, "step": 7990 }, { "epoch": 0.23, "learning_rate": 9.453786182495584e-06, "loss": 0.9683, "step": 7991 }, { "epoch": 0.23, "learning_rate": 9.451633650986087e-06, "loss": 0.9954, "step": 7992 }, { "epoch": 0.23, "learning_rate": 9.449481144960881e-06, "loss": 0.9559, "step": 7993 }, { "epoch": 0.23, "learning_rate": 9.447328664520009e-06, "loss": 0.9363, "step": 7994 }, { "epoch": 0.23, "learning_rate": 9.445176209763496e-06, "loss": 0.9734, "step": 7995 }, { "epoch": 0.23, "learning_rate": 9.44302378079137e-06, "loss": 0.9876, "step": 7996 }, { "epoch": 0.23, "learning_rate": 9.44087137770367e-06, "loss": 0.9839, "step": 7997 }, { "epoch": 0.23, "learning_rate": 9.43871900060042e-06, "loss": 0.9657, "step": 7998 }, { "epoch": 0.23, "learning_rate": 9.436566649581648e-06, "loss": 0.968, "step": 7999 }, { "epoch": 0.23, "learning_rate": 9.43441432474738e-06, "loss": 0.9693, "step": 8000 }, { "epoch": 0.23, "learning_rate": 9.432262026197644e-06, "loss": 0.9491, "step": 8001 }, { "epoch": 0.23, "learning_rate": 9.430109754032458e-06, "loss": 0.9854, "step": 8002 }, { "epoch": 0.23, "learning_rate": 9.427957508351852e-06, "loss": 1.0095, "step": 8003 }, { "epoch": 0.23, "learning_rate": 9.425805289255847e-06, "loss": 0.9274, "step": 8004 }, { "epoch": 0.23, "learning_rate": 9.423653096844452e-06, "loss": 0.993, "step": 8005 }, { "epoch": 0.23, "learning_rate": 9.421500931217705e-06, "loss": 0.9711, "step": 8006 }, { "epoch": 0.23, "learning_rate": 9.419348792475605e-06, "loss": 0.9741, "step": 8007 }, { "epoch": 0.23, "learning_rate": 9.417196680718183e-06, "loss": 0.9338, "step": 8008 }, { "epoch": 0.23, "learning_rate": 9.415044596045446e-06, "loss": 0.9626, "step": 8009 }, { "epoch": 0.23, "learning_rate": 9.412892538557411e-06, "loss": 0.9865, "step": 8010 }, { "epoch": 0.23, "learning_rate": 9.410740508354093e-06, "loss": 1.0176, "step": 8011 }, { "epoch": 0.23, "learning_rate": 9.408588505535498e-06, "loss": 0.9818, "step": 8012 }, { "epoch": 0.23, "learning_rate": 9.406436530201641e-06, "loss": 0.9893, "step": 8013 }, { "epoch": 0.23, "learning_rate": 9.404284582452527e-06, "loss": 0.9539, "step": 8014 }, { "epoch": 0.23, "learning_rate": 9.402132662388168e-06, "loss": 0.9417, "step": 8015 }, { "epoch": 0.23, "learning_rate": 9.399980770108566e-06, "loss": 0.8957, "step": 8016 }, { "epoch": 0.23, "learning_rate": 9.397828905713731e-06, "loss": 0.9845, "step": 8017 }, { "epoch": 0.23, "learning_rate": 9.395677069303667e-06, "loss": 1.0139, "step": 8018 }, { "epoch": 0.23, "learning_rate": 9.393525260978368e-06, "loss": 1.0138, "step": 8019 }, { "epoch": 0.23, "learning_rate": 9.391373480837847e-06, "loss": 0.9547, "step": 8020 }, { "epoch": 0.23, "learning_rate": 9.389221728982095e-06, "loss": 0.9542, "step": 8021 }, { "epoch": 0.23, "learning_rate": 9.387070005511115e-06, "loss": 1.0039, "step": 8022 }, { "epoch": 0.23, "learning_rate": 9.3849183105249e-06, "loss": 0.9298, "step": 8023 }, { "epoch": 0.23, "learning_rate": 9.382766644123453e-06, "loss": 0.9377, "step": 8024 }, { "epoch": 0.23, "learning_rate": 9.380615006406761e-06, "loss": 1.03, "step": 8025 }, { "epoch": 0.23, "learning_rate": 9.378463397474827e-06, "loss": 0.9663, "step": 8026 }, { "epoch": 0.23, "learning_rate": 9.376311817427635e-06, "loss": 0.9488, "step": 8027 }, { "epoch": 0.23, "learning_rate": 9.374160266365172e-06, "loss": 0.936, "step": 8028 }, { "epoch": 0.23, "learning_rate": 9.372008744387441e-06, "loss": 0.9762, "step": 8029 }, { "epoch": 0.23, "learning_rate": 9.369857251594415e-06, "loss": 0.9289, "step": 8030 }, { "epoch": 0.23, "learning_rate": 9.367705788086092e-06, "loss": 0.9705, "step": 8031 }, { "epoch": 0.23, "learning_rate": 9.36555435396245e-06, "loss": 0.9339, "step": 8032 }, { "epoch": 0.23, "learning_rate": 9.363402949323478e-06, "loss": 0.9585, "step": 8033 }, { "epoch": 0.23, "learning_rate": 9.361251574269155e-06, "loss": 1.0136, "step": 8034 }, { "epoch": 0.23, "learning_rate": 9.359100228899463e-06, "loss": 0.9774, "step": 8035 }, { "epoch": 0.23, "learning_rate": 9.356948913314383e-06, "loss": 0.9692, "step": 8036 }, { "epoch": 0.23, "learning_rate": 9.35479762761389e-06, "loss": 0.9322, "step": 8037 }, { "epoch": 0.23, "learning_rate": 9.352646371897966e-06, "loss": 0.947, "step": 8038 }, { "epoch": 0.23, "learning_rate": 9.350495146266577e-06, "loss": 0.975, "step": 8039 }, { "epoch": 0.23, "learning_rate": 9.34834395081971e-06, "loss": 0.9819, "step": 8040 }, { "epoch": 0.23, "learning_rate": 9.346192785657335e-06, "loss": 0.966, "step": 8041 }, { "epoch": 0.23, "learning_rate": 9.344041650879412e-06, "loss": 0.9659, "step": 8042 }, { "epoch": 0.24, "learning_rate": 9.341890546585923e-06, "loss": 0.9821, "step": 8043 }, { "epoch": 0.24, "learning_rate": 9.33973947287683e-06, "loss": 0.9581, "step": 8044 }, { "epoch": 0.24, "learning_rate": 9.337588429852106e-06, "loss": 0.9564, "step": 8045 }, { "epoch": 0.24, "learning_rate": 9.335437417611712e-06, "loss": 0.9896, "step": 8046 }, { "epoch": 0.24, "learning_rate": 9.333286436255613e-06, "loss": 1.0006, "step": 8047 }, { "epoch": 0.24, "learning_rate": 9.331135485883769e-06, "loss": 0.9399, "step": 8048 }, { "epoch": 0.24, "learning_rate": 9.328984566596152e-06, "loss": 0.9497, "step": 8049 }, { "epoch": 0.24, "learning_rate": 9.326833678492712e-06, "loss": 0.9858, "step": 8050 }, { "epoch": 0.24, "learning_rate": 9.324682821673406e-06, "loss": 0.9891, "step": 8051 }, { "epoch": 0.24, "learning_rate": 9.3225319962382e-06, "loss": 0.9396, "step": 8052 }, { "epoch": 0.24, "learning_rate": 9.320381202287039e-06, "loss": 0.9643, "step": 8053 }, { "epoch": 0.24, "learning_rate": 9.318230439919889e-06, "loss": 0.9605, "step": 8054 }, { "epoch": 0.24, "learning_rate": 9.316079709236692e-06, "loss": 1.0309, "step": 8055 }, { "epoch": 0.24, "learning_rate": 9.313929010337405e-06, "loss": 0.9338, "step": 8056 }, { "epoch": 0.24, "learning_rate": 9.311778343321979e-06, "loss": 0.9785, "step": 8057 }, { "epoch": 0.24, "learning_rate": 9.309627708290357e-06, "loss": 0.9799, "step": 8058 }, { "epoch": 0.24, "learning_rate": 9.307477105342488e-06, "loss": 0.931, "step": 8059 }, { "epoch": 0.24, "learning_rate": 9.305326534578316e-06, "loss": 0.9776, "step": 8060 }, { "epoch": 0.24, "learning_rate": 9.303175996097789e-06, "loss": 0.9764, "step": 8061 }, { "epoch": 0.24, "learning_rate": 9.301025490000843e-06, "loss": 0.9583, "step": 8062 }, { "epoch": 0.24, "learning_rate": 9.298875016387424e-06, "loss": 0.9565, "step": 8063 }, { "epoch": 0.24, "learning_rate": 9.296724575357473e-06, "loss": 0.9434, "step": 8064 }, { "epoch": 0.24, "learning_rate": 9.294574167010917e-06, "loss": 0.9513, "step": 8065 }, { "epoch": 0.24, "learning_rate": 9.292423791447703e-06, "loss": 0.9766, "step": 8066 }, { "epoch": 0.24, "learning_rate": 9.29027344876776e-06, "loss": 0.9802, "step": 8067 }, { "epoch": 0.24, "learning_rate": 9.288123139071026e-06, "loss": 0.9443, "step": 8068 }, { "epoch": 0.24, "learning_rate": 9.285972862457426e-06, "loss": 0.9536, "step": 8069 }, { "epoch": 0.24, "learning_rate": 9.283822619026898e-06, "loss": 0.9599, "step": 8070 }, { "epoch": 0.24, "learning_rate": 9.28167240887936e-06, "loss": 0.955, "step": 8071 }, { "epoch": 0.24, "learning_rate": 9.279522232114751e-06, "loss": 0.9987, "step": 8072 }, { "epoch": 0.24, "learning_rate": 9.27737208883299e-06, "loss": 0.9771, "step": 8073 }, { "epoch": 0.24, "learning_rate": 9.275221979133996e-06, "loss": 0.9714, "step": 8074 }, { "epoch": 0.24, "learning_rate": 9.273071903117704e-06, "loss": 0.9593, "step": 8075 }, { "epoch": 0.24, "learning_rate": 9.27092186088402e-06, "loss": 0.9433, "step": 8076 }, { "epoch": 0.24, "learning_rate": 9.268771852532877e-06, "loss": 0.9683, "step": 8077 }, { "epoch": 0.24, "learning_rate": 9.26662187816418e-06, "loss": 0.9718, "step": 8078 }, { "epoch": 0.24, "learning_rate": 9.264471937877857e-06, "loss": 0.9866, "step": 8079 }, { "epoch": 0.24, "learning_rate": 9.262322031773815e-06, "loss": 0.9675, "step": 8080 }, { "epoch": 0.24, "learning_rate": 9.260172159951967e-06, "loss": 0.9436, "step": 8081 }, { "epoch": 0.24, "learning_rate": 9.258022322512228e-06, "loss": 0.9808, "step": 8082 }, { "epoch": 0.24, "learning_rate": 9.255872519554501e-06, "loss": 1.0093, "step": 8083 }, { "epoch": 0.24, "learning_rate": 9.253722751178704e-06, "loss": 0.9613, "step": 8084 }, { "epoch": 0.24, "learning_rate": 9.251573017484731e-06, "loss": 0.9313, "step": 8085 }, { "epoch": 0.24, "learning_rate": 9.2494233185725e-06, "loss": 0.9436, "step": 8086 }, { "epoch": 0.24, "learning_rate": 9.247273654541908e-06, "loss": 0.9892, "step": 8087 }, { "epoch": 0.24, "learning_rate": 9.24512402549285e-06, "loss": 0.9484, "step": 8088 }, { "epoch": 0.24, "learning_rate": 9.242974431525238e-06, "loss": 0.9768, "step": 8089 }, { "epoch": 0.24, "learning_rate": 9.240824872738963e-06, "loss": 0.9613, "step": 8090 }, { "epoch": 0.24, "learning_rate": 9.238675349233924e-06, "loss": 0.9845, "step": 8091 }, { "epoch": 0.24, "learning_rate": 9.236525861110015e-06, "loss": 0.9749, "step": 8092 }, { "epoch": 0.24, "learning_rate": 9.234376408467131e-06, "loss": 0.948, "step": 8093 }, { "epoch": 0.24, "learning_rate": 9.232226991405158e-06, "loss": 0.9943, "step": 8094 }, { "epoch": 0.24, "learning_rate": 9.230077610023998e-06, "loss": 0.9916, "step": 8095 }, { "epoch": 0.24, "learning_rate": 9.227928264423532e-06, "loss": 0.9703, "step": 8096 }, { "epoch": 0.24, "learning_rate": 9.22577895470364e-06, "loss": 0.9729, "step": 8097 }, { "epoch": 0.24, "learning_rate": 9.22362968096422e-06, "loss": 0.9903, "step": 8098 }, { "epoch": 0.24, "learning_rate": 9.221480443305147e-06, "loss": 0.9626, "step": 8099 }, { "epoch": 0.24, "learning_rate": 9.219331241826307e-06, "loss": 0.9917, "step": 8100 }, { "epoch": 0.24, "learning_rate": 9.217182076627577e-06, "loss": 0.9937, "step": 8101 }, { "epoch": 0.24, "learning_rate": 9.21503294780884e-06, "loss": 0.9876, "step": 8102 }, { "epoch": 0.24, "learning_rate": 9.212883855469968e-06, "loss": 1.0063, "step": 8103 }, { "epoch": 0.24, "learning_rate": 9.210734799710837e-06, "loss": 0.9234, "step": 8104 }, { "epoch": 0.24, "learning_rate": 9.208585780631321e-06, "loss": 0.9908, "step": 8105 }, { "epoch": 0.24, "learning_rate": 9.20643679833129e-06, "loss": 0.9448, "step": 8106 }, { "epoch": 0.24, "learning_rate": 9.20428785291062e-06, "loss": 0.9891, "step": 8107 }, { "epoch": 0.24, "learning_rate": 9.202138944469168e-06, "loss": 0.9552, "step": 8108 }, { "epoch": 0.24, "learning_rate": 9.199990073106813e-06, "loss": 0.9892, "step": 8109 }, { "epoch": 0.24, "learning_rate": 9.197841238923416e-06, "loss": 0.9761, "step": 8110 }, { "epoch": 0.24, "learning_rate": 9.195692442018831e-06, "loss": 0.9555, "step": 8111 }, { "epoch": 0.24, "learning_rate": 9.19354368249293e-06, "loss": 0.9353, "step": 8112 }, { "epoch": 0.24, "learning_rate": 9.191394960445568e-06, "loss": 0.951, "step": 8113 }, { "epoch": 0.24, "learning_rate": 9.189246275976605e-06, "loss": 0.9852, "step": 8114 }, { "epoch": 0.24, "learning_rate": 9.187097629185894e-06, "loss": 0.9155, "step": 8115 }, { "epoch": 0.24, "learning_rate": 9.184949020173292e-06, "loss": 0.9406, "step": 8116 }, { "epoch": 0.24, "learning_rate": 9.182800449038647e-06, "loss": 0.9718, "step": 8117 }, { "epoch": 0.24, "learning_rate": 9.180651915881819e-06, "loss": 0.9543, "step": 8118 }, { "epoch": 0.24, "learning_rate": 9.178503420802651e-06, "loss": 0.9769, "step": 8119 }, { "epoch": 0.24, "learning_rate": 9.176354963900987e-06, "loss": 0.9779, "step": 8120 }, { "epoch": 0.24, "learning_rate": 9.174206545276678e-06, "loss": 0.9677, "step": 8121 }, { "epoch": 0.24, "learning_rate": 9.172058165029565e-06, "loss": 0.9801, "step": 8122 }, { "epoch": 0.24, "learning_rate": 9.169909823259493e-06, "loss": 0.9643, "step": 8123 }, { "epoch": 0.24, "learning_rate": 9.167761520066297e-06, "loss": 0.994, "step": 8124 }, { "epoch": 0.24, "learning_rate": 9.165613255549821e-06, "loss": 0.9495, "step": 8125 }, { "epoch": 0.24, "learning_rate": 9.163465029809899e-06, "loss": 0.9661, "step": 8126 }, { "epoch": 0.24, "learning_rate": 9.161316842946364e-06, "loss": 0.9725, "step": 8127 }, { "epoch": 0.24, "learning_rate": 9.159168695059051e-06, "loss": 0.9472, "step": 8128 }, { "epoch": 0.24, "learning_rate": 9.157020586247788e-06, "loss": 0.9691, "step": 8129 }, { "epoch": 0.24, "learning_rate": 9.154872516612412e-06, "loss": 0.9619, "step": 8130 }, { "epoch": 0.24, "learning_rate": 9.152724486252742e-06, "loss": 1.0187, "step": 8131 }, { "epoch": 0.24, "learning_rate": 9.150576495268609e-06, "loss": 0.9348, "step": 8132 }, { "epoch": 0.24, "learning_rate": 9.148428543759836e-06, "loss": 0.9626, "step": 8133 }, { "epoch": 0.24, "learning_rate": 9.14628063182624e-06, "loss": 0.9774, "step": 8134 }, { "epoch": 0.24, "learning_rate": 9.144132759567648e-06, "loss": 0.9264, "step": 8135 }, { "epoch": 0.24, "learning_rate": 9.141984927083874e-06, "loss": 0.9809, "step": 8136 }, { "epoch": 0.24, "learning_rate": 9.139837134474739e-06, "loss": 0.9724, "step": 8137 }, { "epoch": 0.24, "learning_rate": 9.137689381840051e-06, "loss": 0.9899, "step": 8138 }, { "epoch": 0.24, "learning_rate": 9.135541669279629e-06, "loss": 0.964, "step": 8139 }, { "epoch": 0.24, "learning_rate": 9.133393996893276e-06, "loss": 0.9491, "step": 8140 }, { "epoch": 0.24, "learning_rate": 9.131246364780814e-06, "loss": 0.9607, "step": 8141 }, { "epoch": 0.24, "learning_rate": 9.12909877304204e-06, "loss": 0.9984, "step": 8142 }, { "epoch": 0.24, "learning_rate": 9.126951221776756e-06, "loss": 0.9932, "step": 8143 }, { "epoch": 0.24, "learning_rate": 9.124803711084777e-06, "loss": 0.942, "step": 8144 }, { "epoch": 0.24, "learning_rate": 9.122656241065897e-06, "loss": 0.985, "step": 8145 }, { "epoch": 0.24, "learning_rate": 9.120508811819918e-06, "loss": 0.981, "step": 8146 }, { "epoch": 0.24, "learning_rate": 9.118361423446636e-06, "loss": 0.9851, "step": 8147 }, { "epoch": 0.24, "learning_rate": 9.116214076045848e-06, "loss": 0.9562, "step": 8148 }, { "epoch": 0.24, "learning_rate": 9.11406676971735e-06, "loss": 0.9621, "step": 8149 }, { "epoch": 0.24, "learning_rate": 9.111919504560928e-06, "loss": 0.9709, "step": 8150 }, { "epoch": 0.24, "learning_rate": 9.10977228067638e-06, "loss": 0.9966, "step": 8151 }, { "epoch": 0.24, "learning_rate": 9.107625098163484e-06, "loss": 0.959, "step": 8152 }, { "epoch": 0.24, "learning_rate": 9.10547795712204e-06, "loss": 1.0195, "step": 8153 }, { "epoch": 0.24, "learning_rate": 9.103330857651818e-06, "loss": 0.9493, "step": 8154 }, { "epoch": 0.24, "learning_rate": 9.101183799852611e-06, "loss": 0.9724, "step": 8155 }, { "epoch": 0.24, "learning_rate": 9.099036783824198e-06, "loss": 1.0049, "step": 8156 }, { "epoch": 0.24, "learning_rate": 9.096889809666349e-06, "loss": 0.9595, "step": 8157 }, { "epoch": 0.24, "learning_rate": 9.09474287747885e-06, "loss": 0.9593, "step": 8158 }, { "epoch": 0.24, "learning_rate": 9.09259598736147e-06, "loss": 0.9731, "step": 8159 }, { "epoch": 0.24, "learning_rate": 9.090449139413987e-06, "loss": 0.9957, "step": 8160 }, { "epoch": 0.24, "learning_rate": 9.088302333736166e-06, "loss": 0.9548, "step": 8161 }, { "epoch": 0.24, "learning_rate": 9.086155570427782e-06, "loss": 0.9204, "step": 8162 }, { "epoch": 0.24, "learning_rate": 9.084008849588593e-06, "loss": 0.9643, "step": 8163 }, { "epoch": 0.24, "learning_rate": 9.081862171318375e-06, "loss": 0.9214, "step": 8164 }, { "epoch": 0.24, "learning_rate": 9.079715535716884e-06, "loss": 0.9764, "step": 8165 }, { "epoch": 0.24, "learning_rate": 9.077568942883877e-06, "loss": 0.9674, "step": 8166 }, { "epoch": 0.24, "learning_rate": 9.07542239291912e-06, "loss": 0.9506, "step": 8167 }, { "epoch": 0.24, "learning_rate": 9.073275885922368e-06, "loss": 0.9552, "step": 8168 }, { "epoch": 0.24, "learning_rate": 9.071129421993376e-06, "loss": 0.985, "step": 8169 }, { "epoch": 0.24, "learning_rate": 9.068983001231895e-06, "loss": 1.001, "step": 8170 }, { "epoch": 0.24, "learning_rate": 9.066836623737677e-06, "loss": 0.9707, "step": 8171 }, { "epoch": 0.24, "learning_rate": 9.064690289610474e-06, "loss": 0.9772, "step": 8172 }, { "epoch": 0.24, "learning_rate": 9.062543998950027e-06, "loss": 0.9546, "step": 8173 }, { "epoch": 0.24, "learning_rate": 9.060397751856084e-06, "loss": 0.9607, "step": 8174 }, { "epoch": 0.24, "learning_rate": 9.058251548428385e-06, "loss": 0.983, "step": 8175 }, { "epoch": 0.24, "learning_rate": 9.056105388766678e-06, "loss": 0.9575, "step": 8176 }, { "epoch": 0.24, "learning_rate": 9.053959272970692e-06, "loss": 0.9567, "step": 8177 }, { "epoch": 0.24, "learning_rate": 9.051813201140171e-06, "loss": 0.9794, "step": 8178 }, { "epoch": 0.24, "learning_rate": 9.04966717337485e-06, "loss": 0.9794, "step": 8179 }, { "epoch": 0.24, "learning_rate": 9.047521189774456e-06, "loss": 0.9696, "step": 8180 }, { "epoch": 0.24, "learning_rate": 9.045375250438723e-06, "loss": 0.9737, "step": 8181 }, { "epoch": 0.24, "learning_rate": 9.043229355467377e-06, "loss": 0.967, "step": 8182 }, { "epoch": 0.24, "learning_rate": 9.041083504960149e-06, "loss": 0.9467, "step": 8183 }, { "epoch": 0.24, "learning_rate": 9.038937699016757e-06, "loss": 0.9603, "step": 8184 }, { "epoch": 0.24, "learning_rate": 9.036791937736932e-06, "loss": 0.947, "step": 8185 }, { "epoch": 0.24, "learning_rate": 9.034646221220383e-06, "loss": 0.9297, "step": 8186 }, { "epoch": 0.24, "learning_rate": 9.03250054956684e-06, "loss": 0.9863, "step": 8187 }, { "epoch": 0.24, "learning_rate": 9.030354922876013e-06, "loss": 0.9581, "step": 8188 }, { "epoch": 0.24, "learning_rate": 9.028209341247612e-06, "loss": 0.9793, "step": 8189 }, { "epoch": 0.24, "learning_rate": 9.026063804781357e-06, "loss": 0.9643, "step": 8190 }, { "epoch": 0.24, "learning_rate": 9.023918313576953e-06, "loss": 0.959, "step": 8191 }, { "epoch": 0.24, "learning_rate": 9.02177286773411e-06, "loss": 0.9712, "step": 8192 }, { "epoch": 0.25, "learning_rate": 9.019627467352529e-06, "loss": 0.9422, "step": 8193 }, { "epoch": 0.25, "learning_rate": 9.017482112531919e-06, "loss": 0.9661, "step": 8194 }, { "epoch": 0.25, "learning_rate": 9.01533680337198e-06, "loss": 0.9559, "step": 8195 }, { "epoch": 0.25, "learning_rate": 9.013191539972407e-06, "loss": 0.9821, "step": 8196 }, { "epoch": 0.25, "learning_rate": 9.011046322432903e-06, "loss": 0.9711, "step": 8197 }, { "epoch": 0.25, "learning_rate": 9.008901150853157e-06, "loss": 0.9474, "step": 8198 }, { "epoch": 0.25, "learning_rate": 9.00675602533287e-06, "loss": 0.9344, "step": 8199 }, { "epoch": 0.25, "learning_rate": 9.00461094597172e-06, "loss": 0.9583, "step": 8200 }, { "epoch": 0.25, "learning_rate": 9.00246591286941e-06, "loss": 0.9814, "step": 8201 }, { "epoch": 0.25, "learning_rate": 9.000320926125618e-06, "loss": 0.9319, "step": 8202 }, { "epoch": 0.25, "learning_rate": 8.998175985840029e-06, "loss": 0.9326, "step": 8203 }, { "epoch": 0.25, "learning_rate": 8.996031092112329e-06, "loss": 0.9909, "step": 8204 }, { "epoch": 0.25, "learning_rate": 8.99388624504219e-06, "loss": 0.9616, "step": 8205 }, { "epoch": 0.25, "learning_rate": 8.991741444729299e-06, "loss": 0.9584, "step": 8206 }, { "epoch": 0.25, "learning_rate": 8.989596691273324e-06, "loss": 0.9654, "step": 8207 }, { "epoch": 0.25, "learning_rate": 8.987451984773943e-06, "loss": 0.9678, "step": 8208 }, { "epoch": 0.25, "learning_rate": 8.985307325330821e-06, "loss": 0.9545, "step": 8209 }, { "epoch": 0.25, "learning_rate": 8.983162713043638e-06, "loss": 0.9603, "step": 8210 }, { "epoch": 0.25, "learning_rate": 8.981018148012054e-06, "loss": 0.962, "step": 8211 }, { "epoch": 0.25, "learning_rate": 8.97887363033573e-06, "loss": 0.995, "step": 8212 }, { "epoch": 0.25, "learning_rate": 8.976729160114334e-06, "loss": 0.98, "step": 8213 }, { "epoch": 0.25, "learning_rate": 8.974584737447526e-06, "loss": 0.984, "step": 8214 }, { "epoch": 0.25, "learning_rate": 8.972440362434963e-06, "loss": 0.9538, "step": 8215 }, { "epoch": 0.25, "learning_rate": 8.9702960351763e-06, "loss": 0.9838, "step": 8216 }, { "epoch": 0.25, "learning_rate": 8.968151755771191e-06, "loss": 0.9764, "step": 8217 }, { "epoch": 0.25, "learning_rate": 8.966007524319289e-06, "loss": 0.9169, "step": 8218 }, { "epoch": 0.25, "learning_rate": 8.963863340920239e-06, "loss": 0.982, "step": 8219 }, { "epoch": 0.25, "learning_rate": 8.961719205673692e-06, "loss": 0.9466, "step": 8220 }, { "epoch": 0.25, "learning_rate": 8.959575118679289e-06, "loss": 1.0028, "step": 8221 }, { "epoch": 0.25, "learning_rate": 8.95743108003668e-06, "loss": 0.9332, "step": 8222 }, { "epoch": 0.25, "learning_rate": 8.955287089845491e-06, "loss": 0.9635, "step": 8223 }, { "epoch": 0.25, "learning_rate": 8.953143148205375e-06, "loss": 0.9647, "step": 8224 }, { "epoch": 0.25, "learning_rate": 8.950999255215962e-06, "loss": 0.9756, "step": 8225 }, { "epoch": 0.25, "learning_rate": 8.948855410976881e-06, "loss": 1.0245, "step": 8226 }, { "epoch": 0.25, "learning_rate": 8.94671161558777e-06, "loss": 0.9738, "step": 8227 }, { "epoch": 0.25, "learning_rate": 8.944567869148251e-06, "loss": 0.9332, "step": 8228 }, { "epoch": 0.25, "learning_rate": 8.942424171757958e-06, "loss": 0.9457, "step": 8229 }, { "epoch": 0.25, "learning_rate": 8.94028052351651e-06, "loss": 0.9288, "step": 8230 }, { "epoch": 0.25, "learning_rate": 8.938136924523531e-06, "loss": 0.9859, "step": 8231 }, { "epoch": 0.25, "learning_rate": 8.935993374878637e-06, "loss": 0.971, "step": 8232 }, { "epoch": 0.25, "learning_rate": 8.933849874681453e-06, "loss": 0.9412, "step": 8233 }, { "epoch": 0.25, "learning_rate": 8.931706424031591e-06, "loss": 0.9993, "step": 8234 }, { "epoch": 0.25, "learning_rate": 8.929563023028656e-06, "loss": 0.907, "step": 8235 }, { "epoch": 0.25, "learning_rate": 8.927419671772271e-06, "loss": 0.9583, "step": 8236 }, { "epoch": 0.25, "learning_rate": 8.925276370362036e-06, "loss": 0.9868, "step": 8237 }, { "epoch": 0.25, "learning_rate": 8.92313311889756e-06, "loss": 0.9339, "step": 8238 }, { "epoch": 0.25, "learning_rate": 8.920989917478446e-06, "loss": 0.9602, "step": 8239 }, { "epoch": 0.25, "learning_rate": 8.918846766204297e-06, "loss": 0.9604, "step": 8240 }, { "epoch": 0.25, "learning_rate": 8.916703665174708e-06, "loss": 0.9587, "step": 8241 }, { "epoch": 0.25, "learning_rate": 8.914560614489279e-06, "loss": 0.9819, "step": 8242 }, { "epoch": 0.25, "learning_rate": 8.912417614247601e-06, "loss": 0.9927, "step": 8243 }, { "epoch": 0.25, "learning_rate": 8.910274664549266e-06, "loss": 0.9761, "step": 8244 }, { "epoch": 0.25, "learning_rate": 8.908131765493873e-06, "loss": 0.9663, "step": 8245 }, { "epoch": 0.25, "learning_rate": 8.905988917180992e-06, "loss": 0.9474, "step": 8246 }, { "epoch": 0.25, "learning_rate": 8.903846119710225e-06, "loss": 0.9869, "step": 8247 }, { "epoch": 0.25, "learning_rate": 8.901703373181144e-06, "loss": 0.9938, "step": 8248 }, { "epoch": 0.25, "learning_rate": 8.89956067769333e-06, "loss": 0.9749, "step": 8249 }, { "epoch": 0.25, "learning_rate": 8.897418033346367e-06, "loss": 1.011, "step": 8250 }, { "epoch": 0.25, "learning_rate": 8.895275440239824e-06, "loss": 0.9787, "step": 8251 }, { "epoch": 0.25, "learning_rate": 8.893132898473277e-06, "loss": 0.9645, "step": 8252 }, { "epoch": 0.25, "learning_rate": 8.890990408146294e-06, "loss": 0.9399, "step": 8253 }, { "epoch": 0.25, "learning_rate": 8.888847969358446e-06, "loss": 0.9926, "step": 8254 }, { "epoch": 0.25, "learning_rate": 8.886705582209295e-06, "loss": 0.9559, "step": 8255 }, { "epoch": 0.25, "learning_rate": 8.884563246798411e-06, "loss": 0.9668, "step": 8256 }, { "epoch": 0.25, "learning_rate": 8.882420963225354e-06, "loss": 0.9688, "step": 8257 }, { "epoch": 0.25, "learning_rate": 8.880278731589672e-06, "loss": 0.9407, "step": 8258 }, { "epoch": 0.25, "learning_rate": 8.878136551990934e-06, "loss": 0.9697, "step": 8259 }, { "epoch": 0.25, "learning_rate": 8.875994424528686e-06, "loss": 0.9741, "step": 8260 }, { "epoch": 0.25, "learning_rate": 8.873852349302484e-06, "loss": 0.9584, "step": 8261 }, { "epoch": 0.25, "learning_rate": 8.871710326411874e-06, "loss": 0.9283, "step": 8262 }, { "epoch": 0.25, "learning_rate": 8.869568355956404e-06, "loss": 0.9572, "step": 8263 }, { "epoch": 0.25, "learning_rate": 8.867426438035618e-06, "loss": 0.9483, "step": 8264 }, { "epoch": 0.25, "learning_rate": 8.865284572749055e-06, "loss": 0.9853, "step": 8265 }, { "epoch": 0.25, "learning_rate": 8.86314276019626e-06, "loss": 0.9466, "step": 8266 }, { "epoch": 0.25, "learning_rate": 8.86100100047676e-06, "loss": 0.9501, "step": 8267 }, { "epoch": 0.25, "learning_rate": 8.858859293690101e-06, "loss": 1.0062, "step": 8268 }, { "epoch": 0.25, "learning_rate": 8.856717639935803e-06, "loss": 0.9782, "step": 8269 }, { "epoch": 0.25, "learning_rate": 8.854576039313407e-06, "loss": 0.9406, "step": 8270 }, { "epoch": 0.25, "learning_rate": 8.852434491922434e-06, "loss": 0.9412, "step": 8271 }, { "epoch": 0.25, "learning_rate": 8.850292997862405e-06, "loss": 0.9292, "step": 8272 }, { "epoch": 0.25, "learning_rate": 8.848151557232849e-06, "loss": 0.9503, "step": 8273 }, { "epoch": 0.25, "learning_rate": 8.84601017013328e-06, "loss": 0.9415, "step": 8274 }, { "epoch": 0.25, "learning_rate": 8.843868836663217e-06, "loss": 1.0327, "step": 8275 }, { "epoch": 0.25, "learning_rate": 8.841727556922173e-06, "loss": 0.9704, "step": 8276 }, { "epoch": 0.25, "learning_rate": 8.839586331009663e-06, "loss": 0.9797, "step": 8277 }, { "epoch": 0.25, "learning_rate": 8.837445159025194e-06, "loss": 0.9404, "step": 8278 }, { "epoch": 0.25, "learning_rate": 8.835304041068273e-06, "loss": 1.0064, "step": 8279 }, { "epoch": 0.25, "learning_rate": 8.83316297723841e-06, "loss": 0.988, "step": 8280 }, { "epoch": 0.25, "learning_rate": 8.831021967635094e-06, "loss": 0.9556, "step": 8281 }, { "epoch": 0.25, "learning_rate": 8.828881012357839e-06, "loss": 0.9663, "step": 8282 }, { "epoch": 0.25, "learning_rate": 8.82674011150613e-06, "loss": 0.935, "step": 8283 }, { "epoch": 0.25, "learning_rate": 8.82459926517947e-06, "loss": 0.9523, "step": 8284 }, { "epoch": 0.25, "learning_rate": 8.822458473477346e-06, "loss": 0.9238, "step": 8285 }, { "epoch": 0.25, "learning_rate": 8.82031773649925e-06, "loss": 0.9446, "step": 8286 }, { "epoch": 0.25, "learning_rate": 8.818177054344668e-06, "loss": 0.9669, "step": 8287 }, { "epoch": 0.25, "learning_rate": 8.81603642711308e-06, "loss": 0.9859, "step": 8288 }, { "epoch": 0.25, "learning_rate": 8.813895854903974e-06, "loss": 0.9804, "step": 8289 }, { "epoch": 0.25, "learning_rate": 8.811755337816824e-06, "loss": 0.9614, "step": 8290 }, { "epoch": 0.25, "learning_rate": 8.80961487595111e-06, "loss": 0.9232, "step": 8291 }, { "epoch": 0.25, "learning_rate": 8.807474469406303e-06, "loss": 0.9926, "step": 8292 }, { "epoch": 0.25, "learning_rate": 8.80533411828188e-06, "loss": 0.9716, "step": 8293 }, { "epoch": 0.25, "learning_rate": 8.803193822677304e-06, "loss": 0.9726, "step": 8294 }, { "epoch": 0.25, "learning_rate": 8.801053582692041e-06, "loss": 0.9915, "step": 8295 }, { "epoch": 0.25, "learning_rate": 8.79891339842556e-06, "loss": 0.9822, "step": 8296 }, { "epoch": 0.25, "learning_rate": 8.796773269977315e-06, "loss": 1.0219, "step": 8297 }, { "epoch": 0.25, "learning_rate": 8.79463319744677e-06, "loss": 0.9556, "step": 8298 }, { "epoch": 0.25, "learning_rate": 8.792493180933378e-06, "loss": 0.954, "step": 8299 }, { "epoch": 0.25, "learning_rate": 8.790353220536598e-06, "loss": 1.0092, "step": 8300 }, { "epoch": 0.25, "learning_rate": 8.788213316355868e-06, "loss": 0.9579, "step": 8301 }, { "epoch": 0.25, "learning_rate": 8.786073468490648e-06, "loss": 0.9747, "step": 8302 }, { "epoch": 0.25, "learning_rate": 8.783933677040383e-06, "loss": 0.9501, "step": 8303 }, { "epoch": 0.25, "learning_rate": 8.781793942104505e-06, "loss": 0.9642, "step": 8304 }, { "epoch": 0.25, "learning_rate": 8.779654263782466e-06, "loss": 0.9258, "step": 8305 }, { "epoch": 0.25, "learning_rate": 8.777514642173696e-06, "loss": 0.9486, "step": 8306 }, { "epoch": 0.25, "learning_rate": 8.775375077377634e-06, "loss": 0.9583, "step": 8307 }, { "epoch": 0.25, "learning_rate": 8.773235569493712e-06, "loss": 0.9445, "step": 8308 }, { "epoch": 0.25, "learning_rate": 8.771096118621357e-06, "loss": 0.9866, "step": 8309 }, { "epoch": 0.25, "learning_rate": 8.768956724859998e-06, "loss": 0.9886, "step": 8310 }, { "epoch": 0.25, "learning_rate": 8.766817388309058e-06, "loss": 0.966, "step": 8311 }, { "epoch": 0.25, "learning_rate": 8.76467810906796e-06, "loss": 0.9659, "step": 8312 }, { "epoch": 0.25, "learning_rate": 8.76253888723612e-06, "loss": 0.9539, "step": 8313 }, { "epoch": 0.25, "learning_rate": 8.76039972291296e-06, "loss": 0.9389, "step": 8314 }, { "epoch": 0.25, "learning_rate": 8.758260616197886e-06, "loss": 0.9299, "step": 8315 }, { "epoch": 0.25, "learning_rate": 8.756121567190317e-06, "loss": 0.9251, "step": 8316 }, { "epoch": 0.25, "learning_rate": 8.753982575989658e-06, "loss": 0.9633, "step": 8317 }, { "epoch": 0.25, "learning_rate": 8.751843642695313e-06, "loss": 0.9913, "step": 8318 }, { "epoch": 0.25, "learning_rate": 8.749704767406686e-06, "loss": 0.963, "step": 8319 }, { "epoch": 0.25, "learning_rate": 8.747565950223177e-06, "loss": 0.9965, "step": 8320 }, { "epoch": 0.25, "learning_rate": 8.745427191244185e-06, "loss": 0.9743, "step": 8321 }, { "epoch": 0.25, "learning_rate": 8.743288490569098e-06, "loss": 0.9491, "step": 8322 }, { "epoch": 0.25, "learning_rate": 8.741149848297321e-06, "loss": 0.9503, "step": 8323 }, { "epoch": 0.25, "learning_rate": 8.739011264528229e-06, "loss": 0.9962, "step": 8324 }, { "epoch": 0.25, "learning_rate": 8.736872739361222e-06, "loss": 0.9667, "step": 8325 }, { "epoch": 0.25, "learning_rate": 8.734734272895678e-06, "loss": 0.9685, "step": 8326 }, { "epoch": 0.25, "learning_rate": 8.732595865230972e-06, "loss": 0.9843, "step": 8327 }, { "epoch": 0.25, "learning_rate": 8.730457516466493e-06, "loss": 0.9449, "step": 8328 }, { "epoch": 0.25, "learning_rate": 8.728319226701608e-06, "loss": 1.0024, "step": 8329 }, { "epoch": 0.25, "learning_rate": 8.726180996035697e-06, "loss": 0.9942, "step": 8330 }, { "epoch": 0.25, "learning_rate": 8.724042824568125e-06, "loss": 0.9641, "step": 8331 }, { "epoch": 0.25, "learning_rate": 8.721904712398264e-06, "loss": 0.9484, "step": 8332 }, { "epoch": 0.25, "learning_rate": 8.719766659625477e-06, "loss": 0.9481, "step": 8333 }, { "epoch": 0.25, "learning_rate": 8.717628666349123e-06, "loss": 0.9707, "step": 8334 }, { "epoch": 0.25, "learning_rate": 8.715490732668565e-06, "loss": 0.9605, "step": 8335 }, { "epoch": 0.25, "learning_rate": 8.713352858683154e-06, "loss": 0.9961, "step": 8336 }, { "epoch": 0.25, "learning_rate": 8.711215044492254e-06, "loss": 0.9671, "step": 8337 }, { "epoch": 0.25, "learning_rate": 8.709077290195205e-06, "loss": 0.981, "step": 8338 }, { "epoch": 0.25, "learning_rate": 8.706939595891364e-06, "loss": 0.9952, "step": 8339 }, { "epoch": 0.25, "learning_rate": 8.70480196168007e-06, "loss": 0.984, "step": 8340 }, { "epoch": 0.25, "learning_rate": 8.702664387660666e-06, "loss": 0.9867, "step": 8341 }, { "epoch": 0.25, "learning_rate": 8.700526873932496e-06, "loss": 0.9826, "step": 8342 }, { "epoch": 0.25, "learning_rate": 8.698389420594892e-06, "loss": 0.9696, "step": 8343 }, { "epoch": 0.26, "learning_rate": 8.696252027747194e-06, "loss": 0.9635, "step": 8344 }, { "epoch": 0.26, "learning_rate": 8.694114695488723e-06, "loss": 0.9496, "step": 8345 }, { "epoch": 0.26, "learning_rate": 8.691977423918822e-06, "loss": 0.9641, "step": 8346 }, { "epoch": 0.26, "learning_rate": 8.689840213136802e-06, "loss": 0.9729, "step": 8347 }, { "epoch": 0.26, "learning_rate": 8.687703063241999e-06, "loss": 0.9689, "step": 8348 }, { "epoch": 0.26, "learning_rate": 8.685565974333727e-06, "loss": 0.9424, "step": 8349 }, { "epoch": 0.26, "learning_rate": 8.683428946511296e-06, "loss": 0.9629, "step": 8350 }, { "epoch": 0.26, "learning_rate": 8.681291979874032e-06, "loss": 0.9841, "step": 8351 }, { "epoch": 0.26, "learning_rate": 8.679155074521239e-06, "loss": 0.9329, "step": 8352 }, { "epoch": 0.26, "learning_rate": 8.677018230552232e-06, "loss": 0.9864, "step": 8353 }, { "epoch": 0.26, "learning_rate": 8.67488144806631e-06, "loss": 0.9869, "step": 8354 }, { "epoch": 0.26, "learning_rate": 8.672744727162782e-06, "loss": 0.9332, "step": 8355 }, { "epoch": 0.26, "learning_rate": 8.670608067940946e-06, "loss": 0.936, "step": 8356 }, { "epoch": 0.26, "learning_rate": 8.668471470500094e-06, "loss": 0.9786, "step": 8357 }, { "epoch": 0.26, "learning_rate": 8.66633493493953e-06, "loss": 0.9588, "step": 8358 }, { "epoch": 0.26, "learning_rate": 8.664198461358533e-06, "loss": 0.9756, "step": 8359 }, { "epoch": 0.26, "learning_rate": 8.662062049856406e-06, "loss": 0.9388, "step": 8360 }, { "epoch": 0.26, "learning_rate": 8.659925700532424e-06, "loss": 1.0013, "step": 8361 }, { "epoch": 0.26, "learning_rate": 8.657789413485876e-06, "loss": 0.9651, "step": 8362 }, { "epoch": 0.26, "learning_rate": 8.655653188816038e-06, "loss": 0.9857, "step": 8363 }, { "epoch": 0.26, "learning_rate": 8.653517026622187e-06, "loss": 0.9552, "step": 8364 }, { "epoch": 0.26, "learning_rate": 8.651380927003601e-06, "loss": 0.9652, "step": 8365 }, { "epoch": 0.26, "learning_rate": 8.649244890059547e-06, "loss": 0.9644, "step": 8366 }, { "epoch": 0.26, "learning_rate": 8.647108915889295e-06, "loss": 0.9979, "step": 8367 }, { "epoch": 0.26, "learning_rate": 8.644973004592108e-06, "loss": 0.9553, "step": 8368 }, { "epoch": 0.26, "learning_rate": 8.642837156267256e-06, "loss": 0.9515, "step": 8369 }, { "epoch": 0.26, "learning_rate": 8.640701371013988e-06, "loss": 0.968, "step": 8370 }, { "epoch": 0.26, "learning_rate": 8.63856564893157e-06, "loss": 0.9481, "step": 8371 }, { "epoch": 0.26, "learning_rate": 8.636429990119254e-06, "loss": 0.9373, "step": 8372 }, { "epoch": 0.26, "learning_rate": 8.63429439467628e-06, "loss": 0.9849, "step": 8373 }, { "epoch": 0.26, "learning_rate": 8.632158862701912e-06, "loss": 0.917, "step": 8374 }, { "epoch": 0.26, "learning_rate": 8.630023394295382e-06, "loss": 0.9425, "step": 8375 }, { "epoch": 0.26, "learning_rate": 8.62788798955594e-06, "loss": 0.944, "step": 8376 }, { "epoch": 0.26, "learning_rate": 8.62575264858282e-06, "loss": 0.9464, "step": 8377 }, { "epoch": 0.26, "learning_rate": 8.62361737147526e-06, "loss": 0.9652, "step": 8378 }, { "epoch": 0.26, "learning_rate": 8.621482158332494e-06, "loss": 0.9323, "step": 8379 }, { "epoch": 0.26, "learning_rate": 8.619347009253748e-06, "loss": 0.9705, "step": 8380 }, { "epoch": 0.26, "learning_rate": 8.617211924338254e-06, "loss": 0.9875, "step": 8381 }, { "epoch": 0.26, "learning_rate": 8.61507690368523e-06, "loss": 0.9119, "step": 8382 }, { "epoch": 0.26, "learning_rate": 8.612941947393904e-06, "loss": 0.9516, "step": 8383 }, { "epoch": 0.26, "learning_rate": 8.610807055563491e-06, "loss": 0.9738, "step": 8384 }, { "epoch": 0.26, "learning_rate": 8.608672228293207e-06, "loss": 0.938, "step": 8385 }, { "epoch": 0.26, "learning_rate": 8.606537465682263e-06, "loss": 0.9495, "step": 8386 }, { "epoch": 0.26, "learning_rate": 8.604402767829865e-06, "loss": 0.9959, "step": 8387 }, { "epoch": 0.26, "learning_rate": 8.602268134835226e-06, "loss": 0.9592, "step": 8388 }, { "epoch": 0.26, "learning_rate": 8.600133566797541e-06, "loss": 0.9957, "step": 8389 }, { "epoch": 0.26, "learning_rate": 8.597999063816018e-06, "loss": 0.9552, "step": 8390 }, { "epoch": 0.26, "learning_rate": 8.595864625989847e-06, "loss": 0.9264, "step": 8391 }, { "epoch": 0.26, "learning_rate": 8.593730253418232e-06, "loss": 0.9983, "step": 8392 }, { "epoch": 0.26, "learning_rate": 8.591595946200348e-06, "loss": 0.9233, "step": 8393 }, { "epoch": 0.26, "learning_rate": 8.589461704435397e-06, "loss": 0.9598, "step": 8394 }, { "epoch": 0.26, "learning_rate": 8.58732752822256e-06, "loss": 0.9698, "step": 8395 }, { "epoch": 0.26, "learning_rate": 8.585193417661016e-06, "loss": 0.9144, "step": 8396 }, { "epoch": 0.26, "learning_rate": 8.583059372849948e-06, "loss": 0.9033, "step": 8397 }, { "epoch": 0.26, "learning_rate": 8.580925393888526e-06, "loss": 0.9644, "step": 8398 }, { "epoch": 0.26, "learning_rate": 8.578791480875929e-06, "loss": 0.9663, "step": 8399 }, { "epoch": 0.26, "learning_rate": 8.576657633911322e-06, "loss": 0.9304, "step": 8400 }, { "epoch": 0.26, "learning_rate": 8.574523853093875e-06, "loss": 0.972, "step": 8401 }, { "epoch": 0.26, "learning_rate": 8.572390138522749e-06, "loss": 0.952, "step": 8402 }, { "epoch": 0.26, "learning_rate": 8.5702564902971e-06, "loss": 0.976, "step": 8403 }, { "epoch": 0.26, "learning_rate": 8.568122908516096e-06, "loss": 0.981, "step": 8404 }, { "epoch": 0.26, "learning_rate": 8.565989393278882e-06, "loss": 0.9448, "step": 8405 }, { "epoch": 0.26, "learning_rate": 8.563855944684614e-06, "loss": 0.9677, "step": 8406 }, { "epoch": 0.26, "learning_rate": 8.561722562832438e-06, "loss": 1.0086, "step": 8407 }, { "epoch": 0.26, "learning_rate": 8.559589247821501e-06, "loss": 0.989, "step": 8408 }, { "epoch": 0.26, "learning_rate": 8.557455999750944e-06, "loss": 0.9651, "step": 8409 }, { "epoch": 0.26, "learning_rate": 8.555322818719901e-06, "loss": 0.9469, "step": 8410 }, { "epoch": 0.26, "learning_rate": 8.553189704827516e-06, "loss": 0.9488, "step": 8411 }, { "epoch": 0.26, "learning_rate": 8.551056658172914e-06, "loss": 0.9763, "step": 8412 }, { "epoch": 0.26, "learning_rate": 8.548923678855228e-06, "loss": 0.9304, "step": 8413 }, { "epoch": 0.26, "learning_rate": 8.54679076697358e-06, "loss": 0.9323, "step": 8414 }, { "epoch": 0.26, "learning_rate": 8.544657922627104e-06, "loss": 0.9543, "step": 8415 }, { "epoch": 0.26, "learning_rate": 8.542525145914907e-06, "loss": 0.9658, "step": 8416 }, { "epoch": 0.26, "learning_rate": 8.540392436936112e-06, "loss": 0.9347, "step": 8417 }, { "epoch": 0.26, "learning_rate": 8.538259795789835e-06, "loss": 0.9806, "step": 8418 }, { "epoch": 0.26, "learning_rate": 8.53612722257518e-06, "loss": 0.9312, "step": 8419 }, { "epoch": 0.26, "learning_rate": 8.53399471739126e-06, "loss": 1.0024, "step": 8420 }, { "epoch": 0.26, "learning_rate": 8.531862280337175e-06, "loss": 0.9572, "step": 8421 }, { "epoch": 0.26, "learning_rate": 8.529729911512028e-06, "loss": 0.9819, "step": 8422 }, { "epoch": 0.26, "learning_rate": 8.527597611014917e-06, "loss": 1.005, "step": 8423 }, { "epoch": 0.26, "learning_rate": 8.525465378944937e-06, "loss": 0.959, "step": 8424 }, { "epoch": 0.26, "learning_rate": 8.52333321540118e-06, "loss": 0.9793, "step": 8425 }, { "epoch": 0.26, "learning_rate": 8.521201120482729e-06, "loss": 0.9508, "step": 8426 }, { "epoch": 0.26, "learning_rate": 8.519069094288678e-06, "loss": 0.9816, "step": 8427 }, { "epoch": 0.26, "learning_rate": 8.5169371369181e-06, "loss": 0.9621, "step": 8428 }, { "epoch": 0.26, "learning_rate": 8.51480524847008e-06, "loss": 0.9828, "step": 8429 }, { "epoch": 0.26, "learning_rate": 8.51267342904369e-06, "loss": 0.9883, "step": 8430 }, { "epoch": 0.26, "learning_rate": 8.510541678738007e-06, "loss": 0.9356, "step": 8431 }, { "epoch": 0.26, "learning_rate": 8.508409997652093e-06, "loss": 0.9872, "step": 8432 }, { "epoch": 0.26, "learning_rate": 8.50627838588502e-06, "loss": 0.9615, "step": 8433 }, { "epoch": 0.26, "learning_rate": 8.50414684353585e-06, "loss": 0.9354, "step": 8434 }, { "epoch": 0.26, "learning_rate": 8.502015370703636e-06, "loss": 0.9477, "step": 8435 }, { "epoch": 0.26, "learning_rate": 8.499883967487441e-06, "loss": 0.937, "step": 8436 }, { "epoch": 0.26, "learning_rate": 8.497752633986313e-06, "loss": 0.9933, "step": 8437 }, { "epoch": 0.26, "learning_rate": 8.495621370299309e-06, "loss": 0.9399, "step": 8438 }, { "epoch": 0.26, "learning_rate": 8.493490176525468e-06, "loss": 0.9133, "step": 8439 }, { "epoch": 0.26, "learning_rate": 8.491359052763836e-06, "loss": 0.9593, "step": 8440 }, { "epoch": 0.26, "learning_rate": 8.489227999113458e-06, "loss": 0.9595, "step": 8441 }, { "epoch": 0.26, "learning_rate": 8.48709701567336e-06, "loss": 0.9479, "step": 8442 }, { "epoch": 0.26, "learning_rate": 8.484966102542584e-06, "loss": 0.9621, "step": 8443 }, { "epoch": 0.26, "learning_rate": 8.482835259820157e-06, "loss": 0.9267, "step": 8444 }, { "epoch": 0.26, "learning_rate": 8.480704487605108e-06, "loss": 1.0003, "step": 8445 }, { "epoch": 0.26, "learning_rate": 8.478573785996456e-06, "loss": 0.9425, "step": 8446 }, { "epoch": 0.26, "learning_rate": 8.476443155093228e-06, "loss": 0.9533, "step": 8447 }, { "epoch": 0.26, "learning_rate": 8.474312594994435e-06, "loss": 0.9264, "step": 8448 }, { "epoch": 0.26, "learning_rate": 8.472182105799091e-06, "loss": 0.9635, "step": 8449 }, { "epoch": 0.26, "learning_rate": 8.470051687606215e-06, "loss": 0.9441, "step": 8450 }, { "epoch": 0.26, "learning_rate": 8.467921340514802e-06, "loss": 0.9861, "step": 8451 }, { "epoch": 0.26, "learning_rate": 8.465791064623865e-06, "loss": 0.9477, "step": 8452 }, { "epoch": 0.26, "learning_rate": 8.463660860032398e-06, "loss": 0.9967, "step": 8453 }, { "epoch": 0.26, "learning_rate": 8.461530726839405e-06, "loss": 0.9157, "step": 8454 }, { "epoch": 0.26, "learning_rate": 8.459400665143876e-06, "loss": 0.9517, "step": 8455 }, { "epoch": 0.26, "learning_rate": 8.457270675044802e-06, "loss": 0.9629, "step": 8456 }, { "epoch": 0.26, "learning_rate": 8.45514075664117e-06, "loss": 0.9477, "step": 8457 }, { "epoch": 0.26, "learning_rate": 8.453010910031964e-06, "loss": 1.0073, "step": 8458 }, { "epoch": 0.26, "learning_rate": 8.450881135316165e-06, "loss": 0.9651, "step": 8459 }, { "epoch": 0.26, "learning_rate": 8.448751432592747e-06, "loss": 0.945, "step": 8460 }, { "epoch": 0.26, "learning_rate": 8.446621801960693e-06, "loss": 0.9799, "step": 8461 }, { "epoch": 0.26, "learning_rate": 8.444492243518962e-06, "loss": 1.0193, "step": 8462 }, { "epoch": 0.26, "learning_rate": 8.442362757366532e-06, "loss": 0.9759, "step": 8463 }, { "epoch": 0.26, "learning_rate": 8.44023334360236e-06, "loss": 0.9443, "step": 8464 }, { "epoch": 0.26, "learning_rate": 8.438104002325408e-06, "loss": 0.9567, "step": 8465 }, { "epoch": 0.26, "learning_rate": 8.435974733634635e-06, "loss": 0.9481, "step": 8466 }, { "epoch": 0.26, "learning_rate": 8.433845537628991e-06, "loss": 0.9324, "step": 8467 }, { "epoch": 0.26, "learning_rate": 8.43171641440743e-06, "loss": 0.9354, "step": 8468 }, { "epoch": 0.26, "learning_rate": 8.429587364068897e-06, "loss": 0.9348, "step": 8469 }, { "epoch": 0.26, "learning_rate": 8.427458386712337e-06, "loss": 0.932, "step": 8470 }, { "epoch": 0.26, "learning_rate": 8.425329482436688e-06, "loss": 1.0124, "step": 8471 }, { "epoch": 0.26, "learning_rate": 8.423200651340887e-06, "loss": 0.9835, "step": 8472 }, { "epoch": 0.26, "learning_rate": 8.421071893523872e-06, "loss": 0.9839, "step": 8473 }, { "epoch": 0.26, "learning_rate": 8.418943209084565e-06, "loss": 0.9676, "step": 8474 }, { "epoch": 0.26, "learning_rate": 8.416814598121901e-06, "loss": 0.9308, "step": 8475 }, { "epoch": 0.26, "learning_rate": 8.414686060734798e-06, "loss": 0.9678, "step": 8476 }, { "epoch": 0.26, "learning_rate": 8.412557597022178e-06, "loss": 0.9435, "step": 8477 }, { "epoch": 0.26, "learning_rate": 8.410429207082955e-06, "loss": 0.9674, "step": 8478 }, { "epoch": 0.26, "learning_rate": 8.408300891016047e-06, "loss": 0.977, "step": 8479 }, { "epoch": 0.26, "learning_rate": 8.406172648920358e-06, "loss": 0.9142, "step": 8480 }, { "epoch": 0.26, "learning_rate": 8.404044480894795e-06, "loss": 0.9517, "step": 8481 }, { "epoch": 0.26, "learning_rate": 8.401916387038263e-06, "loss": 0.952, "step": 8482 }, { "epoch": 0.26, "learning_rate": 8.399788367449656e-06, "loss": 0.9422, "step": 8483 }, { "epoch": 0.26, "learning_rate": 8.397660422227877e-06, "loss": 0.9499, "step": 8484 }, { "epoch": 0.26, "learning_rate": 8.395532551471814e-06, "loss": 0.938, "step": 8485 }, { "epoch": 0.26, "learning_rate": 8.393404755280357e-06, "loss": 0.988, "step": 8486 }, { "epoch": 0.26, "learning_rate": 8.391277033752392e-06, "loss": 0.9889, "step": 8487 }, { "epoch": 0.26, "learning_rate": 8.389149386986797e-06, "loss": 0.9552, "step": 8488 }, { "epoch": 0.26, "learning_rate": 8.387021815082455e-06, "loss": 0.9092, "step": 8489 }, { "epoch": 0.26, "learning_rate": 8.384894318138237e-06, "loss": 1.0003, "step": 8490 }, { "epoch": 0.26, "learning_rate": 8.38276689625302e-06, "loss": 0.9175, "step": 8491 }, { "epoch": 0.26, "learning_rate": 8.380639549525663e-06, "loss": 0.9346, "step": 8492 }, { "epoch": 0.26, "learning_rate": 8.37851227805504e-06, "loss": 1.0032, "step": 8493 }, { "epoch": 0.27, "learning_rate": 8.376385081940008e-06, "loss": 0.9584, "step": 8494 }, { "epoch": 0.27, "learning_rate": 8.374257961279419e-06, "loss": 0.9223, "step": 8495 }, { "epoch": 0.27, "learning_rate": 8.372130916172138e-06, "loss": 0.9236, "step": 8496 }, { "epoch": 0.27, "learning_rate": 8.370003946717005e-06, "loss": 0.9899, "step": 8497 }, { "epoch": 0.27, "learning_rate": 8.367877053012874e-06, "loss": 0.9273, "step": 8498 }, { "epoch": 0.27, "learning_rate": 8.365750235158584e-06, "loss": 0.9714, "step": 8499 }, { "epoch": 0.27, "learning_rate": 8.363623493252978e-06, "loss": 0.9641, "step": 8500 }, { "epoch": 0.27, "learning_rate": 8.36149682739489e-06, "loss": 0.9959, "step": 8501 }, { "epoch": 0.27, "learning_rate": 8.359370237683155e-06, "loss": 0.972, "step": 8502 }, { "epoch": 0.27, "learning_rate": 8.357243724216602e-06, "loss": 0.9234, "step": 8503 }, { "epoch": 0.27, "learning_rate": 8.355117287094052e-06, "loss": 0.9252, "step": 8504 }, { "epoch": 0.27, "learning_rate": 8.352990926414334e-06, "loss": 0.9655, "step": 8505 }, { "epoch": 0.27, "learning_rate": 8.35086464227626e-06, "loss": 0.966, "step": 8506 }, { "epoch": 0.27, "learning_rate": 8.348738434778652e-06, "loss": 0.9397, "step": 8507 }, { "epoch": 0.27, "learning_rate": 8.346612304020313e-06, "loss": 0.9347, "step": 8508 }, { "epoch": 0.27, "learning_rate": 8.344486250100062e-06, "loss": 0.9372, "step": 8509 }, { "epoch": 0.27, "learning_rate": 8.342360273116696e-06, "loss": 0.9282, "step": 8510 }, { "epoch": 0.27, "learning_rate": 8.340234373169016e-06, "loss": 0.9854, "step": 8511 }, { "epoch": 0.27, "learning_rate": 8.338108550355821e-06, "loss": 0.9375, "step": 8512 }, { "epoch": 0.27, "learning_rate": 8.335982804775902e-06, "loss": 0.9474, "step": 8513 }, { "epoch": 0.27, "learning_rate": 8.333857136528054e-06, "loss": 0.9551, "step": 8514 }, { "epoch": 0.27, "learning_rate": 8.331731545711054e-06, "loss": 0.9699, "step": 8515 }, { "epoch": 0.27, "learning_rate": 8.3296060324237e-06, "loss": 0.9258, "step": 8516 }, { "epoch": 0.27, "learning_rate": 8.327480596764759e-06, "loss": 0.9859, "step": 8517 }, { "epoch": 0.27, "learning_rate": 8.325355238833006e-06, "loss": 0.9328, "step": 8518 }, { "epoch": 0.27, "learning_rate": 8.323229958727224e-06, "loss": 1.0168, "step": 8519 }, { "epoch": 0.27, "learning_rate": 8.321104756546167e-06, "loss": 0.9681, "step": 8520 }, { "epoch": 0.27, "learning_rate": 8.318979632388612e-06, "loss": 0.9453, "step": 8521 }, { "epoch": 0.27, "learning_rate": 8.316854586353312e-06, "loss": 0.9437, "step": 8522 }, { "epoch": 0.27, "learning_rate": 8.314729618539033e-06, "loss": 0.978, "step": 8523 }, { "epoch": 0.27, "learning_rate": 8.312604729044519e-06, "loss": 0.9956, "step": 8524 }, { "epoch": 0.27, "learning_rate": 8.310479917968528e-06, "loss": 0.9772, "step": 8525 }, { "epoch": 0.27, "learning_rate": 8.308355185409803e-06, "loss": 0.9541, "step": 8526 }, { "epoch": 0.27, "learning_rate": 8.306230531467086e-06, "loss": 0.9162, "step": 8527 }, { "epoch": 0.27, "learning_rate": 8.30410595623912e-06, "loss": 0.9552, "step": 8528 }, { "epoch": 0.27, "learning_rate": 8.301981459824633e-06, "loss": 0.9496, "step": 8529 }, { "epoch": 0.27, "learning_rate": 8.29985704232237e-06, "loss": 0.9562, "step": 8530 }, { "epoch": 0.27, "learning_rate": 8.297732703831046e-06, "loss": 0.9754, "step": 8531 }, { "epoch": 0.27, "learning_rate": 8.295608444449395e-06, "loss": 0.9463, "step": 8532 }, { "epoch": 0.27, "learning_rate": 8.293484264276135e-06, "loss": 0.9564, "step": 8533 }, { "epoch": 0.27, "learning_rate": 8.291360163409978e-06, "loss": 0.9638, "step": 8534 }, { "epoch": 0.27, "learning_rate": 8.289236141949647e-06, "loss": 0.9475, "step": 8535 }, { "epoch": 0.27, "learning_rate": 8.287112199993844e-06, "loss": 0.9502, "step": 8536 }, { "epoch": 0.27, "learning_rate": 8.28498833764128e-06, "loss": 0.9209, "step": 8537 }, { "epoch": 0.27, "learning_rate": 8.282864554990652e-06, "loss": 0.9262, "step": 8538 }, { "epoch": 0.27, "learning_rate": 8.280740852140669e-06, "loss": 0.9509, "step": 8539 }, { "epoch": 0.27, "learning_rate": 8.278617229190018e-06, "loss": 0.9733, "step": 8540 }, { "epoch": 0.27, "learning_rate": 8.276493686237388e-06, "loss": 0.9795, "step": 8541 }, { "epoch": 0.27, "learning_rate": 8.274370223381477e-06, "loss": 0.9566, "step": 8542 }, { "epoch": 0.27, "learning_rate": 8.272246840720955e-06, "loss": 0.9901, "step": 8543 }, { "epoch": 0.27, "learning_rate": 8.270123538354516e-06, "loss": 0.9775, "step": 8544 }, { "epoch": 0.27, "learning_rate": 8.268000316380827e-06, "loss": 0.9797, "step": 8545 }, { "epoch": 0.27, "learning_rate": 8.265877174898567e-06, "loss": 0.9781, "step": 8546 }, { "epoch": 0.27, "learning_rate": 8.263754114006399e-06, "loss": 0.9879, "step": 8547 }, { "epoch": 0.27, "learning_rate": 8.261631133802995e-06, "loss": 0.9737, "step": 8548 }, { "epoch": 0.27, "learning_rate": 8.259508234387011e-06, "loss": 0.9845, "step": 8549 }, { "epoch": 0.27, "learning_rate": 8.257385415857106e-06, "loss": 0.9388, "step": 8550 }, { "epoch": 0.27, "learning_rate": 8.255262678311936e-06, "loss": 0.9849, "step": 8551 }, { "epoch": 0.27, "learning_rate": 8.253140021850147e-06, "loss": 0.957, "step": 8552 }, { "epoch": 0.27, "learning_rate": 8.25101744657039e-06, "loss": 0.9681, "step": 8553 }, { "epoch": 0.27, "learning_rate": 8.248894952571306e-06, "loss": 0.984, "step": 8554 }, { "epoch": 0.27, "learning_rate": 8.246772539951536e-06, "loss": 0.9787, "step": 8555 }, { "epoch": 0.27, "learning_rate": 8.244650208809714e-06, "loss": 0.9096, "step": 8556 }, { "epoch": 0.27, "learning_rate": 8.242527959244467e-06, "loss": 0.9383, "step": 8557 }, { "epoch": 0.27, "learning_rate": 8.240405791354428e-06, "loss": 0.9233, "step": 8558 }, { "epoch": 0.27, "learning_rate": 8.238283705238216e-06, "loss": 0.894, "step": 8559 }, { "epoch": 0.27, "learning_rate": 8.236161700994458e-06, "loss": 0.9332, "step": 8560 }, { "epoch": 0.27, "learning_rate": 8.234039778721759e-06, "loss": 0.9955, "step": 8561 }, { "epoch": 0.27, "learning_rate": 8.231917938518748e-06, "loss": 0.9537, "step": 8562 }, { "epoch": 0.27, "learning_rate": 8.229796180484018e-06, "loss": 0.956, "step": 8563 }, { "epoch": 0.27, "learning_rate": 8.227674504716177e-06, "loss": 0.9287, "step": 8564 }, { "epoch": 0.27, "learning_rate": 8.225552911313835e-06, "loss": 0.976, "step": 8565 }, { "epoch": 0.27, "learning_rate": 8.223431400375576e-06, "loss": 0.9708, "step": 8566 }, { "epoch": 0.27, "learning_rate": 8.221309972000004e-06, "loss": 0.9887, "step": 8567 }, { "epoch": 0.27, "learning_rate": 8.219188626285702e-06, "loss": 0.9217, "step": 8568 }, { "epoch": 0.27, "learning_rate": 8.217067363331258e-06, "loss": 0.9471, "step": 8569 }, { "epoch": 0.27, "learning_rate": 8.214946183235253e-06, "loss": 0.9338, "step": 8570 }, { "epoch": 0.27, "learning_rate": 8.212825086096266e-06, "loss": 0.9612, "step": 8571 }, { "epoch": 0.27, "learning_rate": 8.210704072012871e-06, "loss": 0.9625, "step": 8572 }, { "epoch": 0.27, "learning_rate": 8.208583141083635e-06, "loss": 0.9736, "step": 8573 }, { "epoch": 0.27, "learning_rate": 8.206462293407129e-06, "loss": 0.9714, "step": 8574 }, { "epoch": 0.27, "learning_rate": 8.204341529081908e-06, "loss": 0.9609, "step": 8575 }, { "epoch": 0.27, "learning_rate": 8.202220848206539e-06, "loss": 0.9811, "step": 8576 }, { "epoch": 0.27, "learning_rate": 8.200100250879573e-06, "loss": 0.9459, "step": 8577 }, { "epoch": 0.27, "learning_rate": 8.197979737199562e-06, "loss": 0.9589, "step": 8578 }, { "epoch": 0.27, "learning_rate": 8.195859307265052e-06, "loss": 0.9709, "step": 8579 }, { "epoch": 0.27, "learning_rate": 8.193738961174584e-06, "loss": 0.9572, "step": 8580 }, { "epoch": 0.27, "learning_rate": 8.1916186990267e-06, "loss": 0.9808, "step": 8581 }, { "epoch": 0.27, "learning_rate": 8.189498520919931e-06, "loss": 0.9557, "step": 8582 }, { "epoch": 0.27, "learning_rate": 8.187378426952816e-06, "loss": 0.9735, "step": 8583 }, { "epoch": 0.27, "learning_rate": 8.185258417223872e-06, "loss": 0.9296, "step": 8584 }, { "epoch": 0.27, "learning_rate": 8.183138491831636e-06, "loss": 0.9682, "step": 8585 }, { "epoch": 0.27, "learning_rate": 8.181018650874616e-06, "loss": 1.0049, "step": 8586 }, { "epoch": 0.27, "learning_rate": 8.178898894451327e-06, "loss": 0.9855, "step": 8587 }, { "epoch": 0.27, "learning_rate": 8.176779222660291e-06, "loss": 0.9715, "step": 8588 }, { "epoch": 0.27, "learning_rate": 8.174659635600006e-06, "loss": 0.9671, "step": 8589 }, { "epoch": 0.27, "learning_rate": 8.172540133368984e-06, "loss": 0.9626, "step": 8590 }, { "epoch": 0.27, "learning_rate": 8.170420716065716e-06, "loss": 0.9996, "step": 8591 }, { "epoch": 0.27, "learning_rate": 8.168301383788706e-06, "loss": 0.9811, "step": 8592 }, { "epoch": 0.27, "learning_rate": 8.16618213663644e-06, "loss": 0.9627, "step": 8593 }, { "epoch": 0.27, "learning_rate": 8.164062974707409e-06, "loss": 0.9533, "step": 8594 }, { "epoch": 0.27, "learning_rate": 8.1619438981001e-06, "loss": 0.9948, "step": 8595 }, { "epoch": 0.27, "learning_rate": 8.159824906912983e-06, "loss": 0.9811, "step": 8596 }, { "epoch": 0.27, "learning_rate": 8.157706001244547e-06, "loss": 0.9502, "step": 8597 }, { "epoch": 0.27, "learning_rate": 8.155587181193253e-06, "loss": 0.9871, "step": 8598 }, { "epoch": 0.27, "learning_rate": 8.153468446857576e-06, "loss": 0.9733, "step": 8599 }, { "epoch": 0.27, "learning_rate": 8.151349798335976e-06, "loss": 0.9131, "step": 8600 }, { "epoch": 0.27, "learning_rate": 8.14923123572692e-06, "loss": 0.9909, "step": 8601 }, { "epoch": 0.27, "learning_rate": 8.147112759128859e-06, "loss": 0.9571, "step": 8602 }, { "epoch": 0.27, "learning_rate": 8.144994368640244e-06, "loss": 0.9759, "step": 8603 }, { "epoch": 0.27, "learning_rate": 8.142876064359526e-06, "loss": 0.9415, "step": 8604 }, { "epoch": 0.27, "learning_rate": 8.140757846385147e-06, "loss": 0.928, "step": 8605 }, { "epoch": 0.27, "learning_rate": 8.13863971481555e-06, "loss": 0.9713, "step": 8606 }, { "epoch": 0.27, "learning_rate": 8.136521669749165e-06, "loss": 0.9859, "step": 8607 }, { "epoch": 0.27, "learning_rate": 8.134403711284436e-06, "loss": 0.9623, "step": 8608 }, { "epoch": 0.27, "learning_rate": 8.132285839519783e-06, "loss": 0.9678, "step": 8609 }, { "epoch": 0.27, "learning_rate": 8.130168054553626e-06, "loss": 0.9571, "step": 8610 }, { "epoch": 0.27, "learning_rate": 8.128050356484394e-06, "loss": 0.9589, "step": 8611 }, { "epoch": 0.27, "learning_rate": 8.125932745410497e-06, "loss": 0.9592, "step": 8612 }, { "epoch": 0.27, "learning_rate": 8.123815221430351e-06, "loss": 0.9203, "step": 8613 }, { "epoch": 0.27, "learning_rate": 8.121697784642361e-06, "loss": 0.959, "step": 8614 }, { "epoch": 0.27, "learning_rate": 8.119580435144933e-06, "loss": 0.978, "step": 8615 }, { "epoch": 0.27, "learning_rate": 8.117463173036465e-06, "loss": 0.9515, "step": 8616 }, { "epoch": 0.27, "learning_rate": 8.115345998415355e-06, "loss": 0.9839, "step": 8617 }, { "epoch": 0.27, "learning_rate": 8.113228911379993e-06, "loss": 0.9486, "step": 8618 }, { "epoch": 0.27, "learning_rate": 8.111111912028763e-06, "loss": 0.984, "step": 8619 }, { "epoch": 0.27, "learning_rate": 8.108995000460059e-06, "loss": 0.9426, "step": 8620 }, { "epoch": 0.27, "learning_rate": 8.106878176772247e-06, "loss": 0.9617, "step": 8621 }, { "epoch": 0.27, "learning_rate": 8.104761441063714e-06, "loss": 0.9713, "step": 8622 }, { "epoch": 0.27, "learning_rate": 8.102644793432826e-06, "loss": 0.9582, "step": 8623 }, { "epoch": 0.27, "learning_rate": 8.10052823397795e-06, "loss": 0.9621, "step": 8624 }, { "epoch": 0.27, "learning_rate": 8.098411762797452e-06, "loss": 0.989, "step": 8625 }, { "epoch": 0.27, "learning_rate": 8.096295379989687e-06, "loss": 0.9435, "step": 8626 }, { "epoch": 0.27, "learning_rate": 8.094179085653014e-06, "loss": 0.9683, "step": 8627 }, { "epoch": 0.27, "learning_rate": 8.09206287988578e-06, "loss": 0.9691, "step": 8628 }, { "epoch": 0.27, "learning_rate": 8.089946762786335e-06, "loss": 0.9366, "step": 8629 }, { "epoch": 0.27, "learning_rate": 8.087830734453015e-06, "loss": 0.9442, "step": 8630 }, { "epoch": 0.27, "learning_rate": 8.085714794984171e-06, "loss": 0.9273, "step": 8631 }, { "epoch": 0.27, "learning_rate": 8.083598944478128e-06, "loss": 0.9407, "step": 8632 }, { "epoch": 0.27, "learning_rate": 8.081483183033212e-06, "loss": 0.9647, "step": 8633 }, { "epoch": 0.27, "learning_rate": 8.07936751074776e-06, "loss": 0.9751, "step": 8634 }, { "epoch": 0.27, "learning_rate": 8.077251927720087e-06, "loss": 0.9371, "step": 8635 }, { "epoch": 0.27, "learning_rate": 8.075136434048514e-06, "loss": 1.0042, "step": 8636 }, { "epoch": 0.27, "learning_rate": 8.073021029831351e-06, "loss": 0.9691, "step": 8637 }, { "epoch": 0.27, "learning_rate": 8.070905715166912e-06, "loss": 0.9673, "step": 8638 }, { "epoch": 0.27, "learning_rate": 8.068790490153498e-06, "loss": 0.9823, "step": 8639 }, { "epoch": 0.27, "learning_rate": 8.066675354889415e-06, "loss": 0.9865, "step": 8640 }, { "epoch": 0.27, "learning_rate": 8.064560309472955e-06, "loss": 0.9377, "step": 8641 }, { "epoch": 0.27, "learning_rate": 8.06244535400241e-06, "loss": 0.9432, "step": 8642 }, { "epoch": 0.27, "learning_rate": 8.060330488576077e-06, "loss": 0.9781, "step": 8643 }, { "epoch": 0.28, "learning_rate": 8.058215713292228e-06, "loss": 0.9602, "step": 8644 }, { "epoch": 0.28, "learning_rate": 8.056101028249154e-06, "loss": 0.9812, "step": 8645 }, { "epoch": 0.28, "learning_rate": 8.053986433545125e-06, "loss": 0.9412, "step": 8646 }, { "epoch": 0.28, "learning_rate": 8.051871929278417e-06, "loss": 0.9864, "step": 8647 }, { "epoch": 0.28, "learning_rate": 8.049757515547294e-06, "loss": 0.9335, "step": 8648 }, { "epoch": 0.28, "learning_rate": 8.047643192450019e-06, "loss": 0.9243, "step": 8649 }, { "epoch": 0.28, "learning_rate": 8.045528960084855e-06, "loss": 0.9295, "step": 8650 }, { "epoch": 0.28, "learning_rate": 8.043414818550053e-06, "loss": 0.948, "step": 8651 }, { "epoch": 0.28, "learning_rate": 8.041300767943867e-06, "loss": 0.8974, "step": 8652 }, { "epoch": 0.28, "learning_rate": 8.039186808364539e-06, "loss": 0.9533, "step": 8653 }, { "epoch": 0.28, "learning_rate": 8.03707293991032e-06, "loss": 0.9276, "step": 8654 }, { "epoch": 0.28, "learning_rate": 8.034959162679442e-06, "loss": 0.993, "step": 8655 }, { "epoch": 0.28, "learning_rate": 8.032845476770133e-06, "loss": 0.9664, "step": 8656 }, { "epoch": 0.28, "learning_rate": 8.030731882280632e-06, "loss": 0.9789, "step": 8657 }, { "epoch": 0.28, "learning_rate": 8.02861837930916e-06, "loss": 0.9475, "step": 8658 }, { "epoch": 0.28, "learning_rate": 8.026504967953944e-06, "loss": 0.9268, "step": 8659 }, { "epoch": 0.28, "learning_rate": 8.02439164831319e-06, "loss": 0.9256, "step": 8660 }, { "epoch": 0.28, "learning_rate": 8.022278420485122e-06, "loss": 0.9709, "step": 8661 }, { "epoch": 0.28, "learning_rate": 8.020165284567938e-06, "loss": 0.9912, "step": 8662 }, { "epoch": 0.28, "learning_rate": 8.01805224065985e-06, "loss": 0.9038, "step": 8663 }, { "epoch": 0.28, "learning_rate": 8.015939288859056e-06, "loss": 0.9383, "step": 8664 }, { "epoch": 0.28, "learning_rate": 8.013826429263743e-06, "loss": 0.9692, "step": 8665 }, { "epoch": 0.28, "learning_rate": 8.011713661972119e-06, "loss": 0.9344, "step": 8666 }, { "epoch": 0.28, "learning_rate": 8.009600987082353e-06, "loss": 0.9238, "step": 8667 }, { "epoch": 0.28, "learning_rate": 8.00748840469264e-06, "loss": 0.9693, "step": 8668 }, { "epoch": 0.28, "learning_rate": 8.005375914901152e-06, "loss": 0.9452, "step": 8669 }, { "epoch": 0.28, "learning_rate": 8.003263517806066e-06, "loss": 0.9453, "step": 8670 }, { "epoch": 0.28, "learning_rate": 8.001151213505552e-06, "loss": 1.0028, "step": 8671 }, { "epoch": 0.28, "learning_rate": 7.999039002097771e-06, "loss": 0.9438, "step": 8672 }, { "epoch": 0.28, "learning_rate": 7.99692688368089e-06, "loss": 0.9447, "step": 8673 }, { "epoch": 0.28, "learning_rate": 7.994814858353062e-06, "loss": 0.9614, "step": 8674 }, { "epoch": 0.28, "learning_rate": 7.99270292621244e-06, "loss": 1.0173, "step": 8675 }, { "epoch": 0.28, "learning_rate": 7.990591087357169e-06, "loss": 0.9838, "step": 8676 }, { "epoch": 0.28, "learning_rate": 7.988479341885402e-06, "loss": 0.979, "step": 8677 }, { "epoch": 0.28, "learning_rate": 7.986367689895271e-06, "loss": 0.9546, "step": 8678 }, { "epoch": 0.28, "learning_rate": 7.98425613148491e-06, "loss": 0.9866, "step": 8679 }, { "epoch": 0.28, "learning_rate": 7.982144666752453e-06, "loss": 0.9444, "step": 8680 }, { "epoch": 0.28, "learning_rate": 7.980033295796026e-06, "loss": 0.935, "step": 8681 }, { "epoch": 0.28, "learning_rate": 7.97792201871375e-06, "loss": 0.9222, "step": 8682 }, { "epoch": 0.28, "learning_rate": 7.975810835603743e-06, "loss": 0.9259, "step": 8683 }, { "epoch": 0.28, "learning_rate": 7.973699746564122e-06, "loss": 0.9302, "step": 8684 }, { "epoch": 0.28, "learning_rate": 7.971588751692986e-06, "loss": 0.9707, "step": 8685 }, { "epoch": 0.28, "learning_rate": 7.969477851088452e-06, "loss": 0.9484, "step": 8686 }, { "epoch": 0.28, "learning_rate": 7.967367044848611e-06, "loss": 0.982, "step": 8687 }, { "epoch": 0.28, "learning_rate": 7.96525633307156e-06, "loss": 0.9475, "step": 8688 }, { "epoch": 0.28, "learning_rate": 7.963145715855397e-06, "loss": 0.9631, "step": 8689 }, { "epoch": 0.28, "learning_rate": 7.961035193298199e-06, "loss": 1.0003, "step": 8690 }, { "epoch": 0.28, "learning_rate": 7.958924765498055e-06, "loss": 0.9593, "step": 8691 }, { "epoch": 0.28, "learning_rate": 7.956814432553043e-06, "loss": 0.9595, "step": 8692 }, { "epoch": 0.28, "learning_rate": 7.954704194561236e-06, "loss": 0.9666, "step": 8693 }, { "epoch": 0.28, "learning_rate": 7.952594051620703e-06, "loss": 0.9825, "step": 8694 }, { "epoch": 0.28, "learning_rate": 7.950484003829506e-06, "loss": 0.9163, "step": 8695 }, { "epoch": 0.28, "learning_rate": 7.948374051285711e-06, "loss": 0.9276, "step": 8696 }, { "epoch": 0.28, "learning_rate": 7.94626419408737e-06, "loss": 0.9497, "step": 8697 }, { "epoch": 0.28, "learning_rate": 7.944154432332537e-06, "loss": 0.9465, "step": 8698 }, { "epoch": 0.28, "learning_rate": 7.942044766119254e-06, "loss": 0.9509, "step": 8699 }, { "epoch": 0.28, "learning_rate": 7.939935195545573e-06, "loss": 0.9097, "step": 8700 }, { "epoch": 0.28, "learning_rate": 7.937825720709529e-06, "loss": 0.9413, "step": 8701 }, { "epoch": 0.28, "learning_rate": 7.935716341709147e-06, "loss": 0.9546, "step": 8702 }, { "epoch": 0.28, "learning_rate": 7.93360705864247e-06, "loss": 0.9817, "step": 8703 }, { "epoch": 0.28, "learning_rate": 7.931497871607513e-06, "loss": 0.9604, "step": 8704 }, { "epoch": 0.28, "learning_rate": 7.929388780702301e-06, "loss": 0.9475, "step": 8705 }, { "epoch": 0.28, "learning_rate": 7.927279786024849e-06, "loss": 0.9408, "step": 8706 }, { "epoch": 0.28, "learning_rate": 7.925170887673171e-06, "loss": 0.9668, "step": 8707 }, { "epoch": 0.28, "learning_rate": 7.923062085745265e-06, "loss": 0.9923, "step": 8708 }, { "epoch": 0.28, "learning_rate": 7.92095338033915e-06, "loss": 0.989, "step": 8709 }, { "epoch": 0.28, "learning_rate": 7.91884477155281e-06, "loss": 0.9677, "step": 8710 }, { "epoch": 0.28, "learning_rate": 7.916736259484239e-06, "loss": 0.9598, "step": 8711 }, { "epoch": 0.28, "learning_rate": 7.914627844231438e-06, "loss": 0.9124, "step": 8712 }, { "epoch": 0.28, "learning_rate": 7.912519525892375e-06, "loss": 0.9484, "step": 8713 }, { "epoch": 0.28, "learning_rate": 7.910411304565045e-06, "loss": 0.9813, "step": 8714 }, { "epoch": 0.28, "learning_rate": 7.908303180347414e-06, "loss": 0.9608, "step": 8715 }, { "epoch": 0.28, "learning_rate": 7.90619515333746e-06, "loss": 0.9436, "step": 8716 }, { "epoch": 0.28, "learning_rate": 7.904087223633146e-06, "loss": 0.9762, "step": 8717 }, { "epoch": 0.28, "learning_rate": 7.901979391332433e-06, "loss": 1.0052, "step": 8718 }, { "epoch": 0.28, "learning_rate": 7.899871656533282e-06, "loss": 0.9557, "step": 8719 }, { "epoch": 0.28, "learning_rate": 7.89776401933364e-06, "loss": 0.9221, "step": 8720 }, { "epoch": 0.28, "learning_rate": 7.895656479831463e-06, "loss": 0.9665, "step": 8721 }, { "epoch": 0.28, "learning_rate": 7.893549038124688e-06, "loss": 0.9646, "step": 8722 }, { "epoch": 0.28, "learning_rate": 7.89144169431126e-06, "loss": 0.9856, "step": 8723 }, { "epoch": 0.28, "learning_rate": 7.889334448489115e-06, "loss": 0.9511, "step": 8724 }, { "epoch": 0.28, "learning_rate": 7.887227300756172e-06, "loss": 0.9536, "step": 8725 }, { "epoch": 0.28, "learning_rate": 7.885120251210368e-06, "loss": 0.9438, "step": 8726 }, { "epoch": 0.28, "learning_rate": 7.88301329994962e-06, "loss": 0.9774, "step": 8727 }, { "epoch": 0.28, "learning_rate": 7.880906447071847e-06, "loss": 0.9671, "step": 8728 }, { "epoch": 0.28, "learning_rate": 7.878799692674959e-06, "loss": 0.9831, "step": 8729 }, { "epoch": 0.28, "learning_rate": 7.876693036856865e-06, "loss": 0.9435, "step": 8730 }, { "epoch": 0.28, "learning_rate": 7.87458647971546e-06, "loss": 0.9546, "step": 8731 }, { "epoch": 0.28, "learning_rate": 7.872480021348659e-06, "loss": 0.9709, "step": 8732 }, { "epoch": 0.28, "learning_rate": 7.87037366185434e-06, "loss": 0.9873, "step": 8733 }, { "epoch": 0.28, "learning_rate": 7.868267401330396e-06, "loss": 0.933, "step": 8734 }, { "epoch": 0.28, "learning_rate": 7.866161239874719e-06, "loss": 0.9934, "step": 8735 }, { "epoch": 0.28, "learning_rate": 7.864055177585175e-06, "loss": 0.9521, "step": 8736 }, { "epoch": 0.28, "learning_rate": 7.861949214559653e-06, "loss": 0.951, "step": 8737 }, { "epoch": 0.28, "learning_rate": 7.859843350896016e-06, "loss": 0.9301, "step": 8738 }, { "epoch": 0.28, "learning_rate": 7.857737586692133e-06, "loss": 0.9458, "step": 8739 }, { "epoch": 0.28, "learning_rate": 7.855631922045866e-06, "loss": 0.992, "step": 8740 }, { "epoch": 0.28, "learning_rate": 7.853526357055068e-06, "loss": 0.9897, "step": 8741 }, { "epoch": 0.28, "learning_rate": 7.851420891817596e-06, "loss": 0.9545, "step": 8742 }, { "epoch": 0.28, "learning_rate": 7.84931552643129e-06, "loss": 0.9669, "step": 8743 }, { "epoch": 0.28, "learning_rate": 7.847210260994005e-06, "loss": 0.9656, "step": 8744 }, { "epoch": 0.28, "learning_rate": 7.845105095603563e-06, "loss": 0.96, "step": 8745 }, { "epoch": 0.28, "learning_rate": 7.843000030357814e-06, "loss": 0.9599, "step": 8746 }, { "epoch": 0.28, "learning_rate": 7.840895065354582e-06, "loss": 1.0014, "step": 8747 }, { "epoch": 0.28, "learning_rate": 7.83879020069168e-06, "loss": 0.9725, "step": 8748 }, { "epoch": 0.28, "learning_rate": 7.836685436466941e-06, "loss": 0.962, "step": 8749 }, { "epoch": 0.28, "learning_rate": 7.834580772778176e-06, "loss": 0.968, "step": 8750 }, { "epoch": 0.28, "learning_rate": 7.832476209723195e-06, "loss": 0.9549, "step": 8751 }, { "epoch": 0.28, "learning_rate": 7.830371747399803e-06, "loss": 0.9922, "step": 8752 }, { "epoch": 0.28, "learning_rate": 7.828267385905802e-06, "loss": 0.9726, "step": 8753 }, { "epoch": 0.28, "learning_rate": 7.826163125338984e-06, "loss": 0.9701, "step": 8754 }, { "epoch": 0.28, "learning_rate": 7.824058965797151e-06, "loss": 0.9322, "step": 8755 }, { "epoch": 0.28, "learning_rate": 7.82195490737808e-06, "loss": 0.9341, "step": 8756 }, { "epoch": 0.28, "learning_rate": 7.819850950179552e-06, "loss": 0.9919, "step": 8757 }, { "epoch": 0.28, "learning_rate": 7.817747094299356e-06, "loss": 0.9565, "step": 8758 }, { "epoch": 0.28, "learning_rate": 7.81564333983525e-06, "loss": 0.9823, "step": 8759 }, { "epoch": 0.28, "learning_rate": 7.813539686885015e-06, "loss": 0.9875, "step": 8760 }, { "epoch": 0.28, "learning_rate": 7.811436135546405e-06, "loss": 0.9495, "step": 8761 }, { "epoch": 0.28, "learning_rate": 7.809332685917185e-06, "loss": 0.9557, "step": 8762 }, { "epoch": 0.28, "learning_rate": 7.807229338095106e-06, "loss": 0.9849, "step": 8763 }, { "epoch": 0.28, "learning_rate": 7.805126092177915e-06, "loss": 0.9746, "step": 8764 }, { "epoch": 0.28, "learning_rate": 7.803022948263362e-06, "loss": 0.9653, "step": 8765 }, { "epoch": 0.28, "learning_rate": 7.80091990644918e-06, "loss": 0.9427, "step": 8766 }, { "epoch": 0.28, "learning_rate": 7.79881696683311e-06, "loss": 1.0123, "step": 8767 }, { "epoch": 0.28, "learning_rate": 7.796714129512875e-06, "loss": 0.975, "step": 8768 }, { "epoch": 0.28, "learning_rate": 7.79461139458621e-06, "loss": 0.992, "step": 8769 }, { "epoch": 0.28, "learning_rate": 7.792508762150833e-06, "loss": 0.9565, "step": 8770 }, { "epoch": 0.28, "learning_rate": 7.790406232304448e-06, "loss": 0.964, "step": 8771 }, { "epoch": 0.28, "learning_rate": 7.78830380514478e-06, "loss": 0.9208, "step": 8772 }, { "epoch": 0.28, "learning_rate": 7.786201480769532e-06, "loss": 0.9567, "step": 8773 }, { "epoch": 0.28, "learning_rate": 7.784099259276405e-06, "loss": 0.9586, "step": 8774 }, { "epoch": 0.28, "learning_rate": 7.781997140763092e-06, "loss": 0.9749, "step": 8775 }, { "epoch": 0.28, "learning_rate": 7.779895125327293e-06, "loss": 0.9112, "step": 8776 }, { "epoch": 0.28, "learning_rate": 7.777793213066684e-06, "loss": 0.9409, "step": 8777 }, { "epoch": 0.28, "learning_rate": 7.77569140407896e-06, "loss": 0.9266, "step": 8778 }, { "epoch": 0.28, "learning_rate": 7.77358969846179e-06, "loss": 0.9123, "step": 8779 }, { "epoch": 0.28, "learning_rate": 7.771488096312847e-06, "loss": 0.9847, "step": 8780 }, { "epoch": 0.28, "learning_rate": 7.769386597729805e-06, "loss": 0.95, "step": 8781 }, { "epoch": 0.28, "learning_rate": 7.76728520281032e-06, "loss": 0.9528, "step": 8782 }, { "epoch": 0.28, "learning_rate": 7.765183911652057e-06, "loss": 0.9593, "step": 8783 }, { "epoch": 0.28, "learning_rate": 7.763082724352664e-06, "loss": 0.9395, "step": 8784 }, { "epoch": 0.28, "learning_rate": 7.760981641009793e-06, "loss": 0.9461, "step": 8785 }, { "epoch": 0.28, "learning_rate": 7.758880661721088e-06, "loss": 0.9445, "step": 8786 }, { "epoch": 0.28, "learning_rate": 7.756779786584186e-06, "loss": 0.9553, "step": 8787 }, { "epoch": 0.28, "learning_rate": 7.754679015696723e-06, "loss": 0.9458, "step": 8788 }, { "epoch": 0.28, "learning_rate": 7.752578349156324e-06, "loss": 0.9934, "step": 8789 }, { "epoch": 0.28, "learning_rate": 7.750477787060624e-06, "loss": 0.9881, "step": 8790 }, { "epoch": 0.28, "learning_rate": 7.748377329507228e-06, "loss": 0.9352, "step": 8791 }, { "epoch": 0.28, "learning_rate": 7.746276976593765e-06, "loss": 0.9377, "step": 8792 }, { "epoch": 0.28, "learning_rate": 7.744176728417839e-06, "loss": 0.944, "step": 8793 }, { "epoch": 0.29, "learning_rate": 7.74207658507705e-06, "loss": 0.9803, "step": 8794 }, { "epoch": 0.29, "learning_rate": 7.739976546669006e-06, "loss": 0.938, "step": 8795 }, { "epoch": 0.29, "learning_rate": 7.737876613291297e-06, "loss": 0.9684, "step": 8796 }, { "epoch": 0.29, "learning_rate": 7.735776785041517e-06, "loss": 0.975, "step": 8797 }, { "epoch": 0.29, "learning_rate": 7.73367706201725e-06, "loss": 0.9539, "step": 8798 }, { "epoch": 0.29, "learning_rate": 7.731577444316079e-06, "loss": 0.959, "step": 8799 }, { "epoch": 0.29, "learning_rate": 7.729477932035571e-06, "loss": 0.9336, "step": 8800 }, { "epoch": 0.29, "learning_rate": 7.727378525273313e-06, "loss": 0.9758, "step": 8801 }, { "epoch": 0.29, "learning_rate": 7.725279224126857e-06, "loss": 0.9296, "step": 8802 }, { "epoch": 0.29, "learning_rate": 7.723180028693766e-06, "loss": 0.9719, "step": 8803 }, { "epoch": 0.29, "learning_rate": 7.721080939071603e-06, "loss": 0.946, "step": 8804 }, { "epoch": 0.29, "learning_rate": 7.718981955357912e-06, "loss": 0.9675, "step": 8805 }, { "epoch": 0.29, "learning_rate": 7.716883077650245e-06, "loss": 0.9395, "step": 8806 }, { "epoch": 0.29, "learning_rate": 7.714784306046139e-06, "loss": 0.9366, "step": 8807 }, { "epoch": 0.29, "learning_rate": 7.712685640643135e-06, "loss": 0.9673, "step": 8808 }, { "epoch": 0.29, "learning_rate": 7.710587081538759e-06, "loss": 1.0024, "step": 8809 }, { "epoch": 0.29, "learning_rate": 7.708488628830541e-06, "loss": 0.9348, "step": 8810 }, { "epoch": 0.29, "learning_rate": 7.706390282616003e-06, "loss": 0.9378, "step": 8811 }, { "epoch": 0.29, "learning_rate": 7.704292042992656e-06, "loss": 0.959, "step": 8812 }, { "epoch": 0.29, "learning_rate": 7.702193910058022e-06, "loss": 0.9609, "step": 8813 }, { "epoch": 0.29, "learning_rate": 7.700095883909598e-06, "loss": 0.9462, "step": 8814 }, { "epoch": 0.29, "learning_rate": 7.697997964644893e-06, "loss": 1.002, "step": 8815 }, { "epoch": 0.29, "learning_rate": 7.695900152361404e-06, "loss": 0.9382, "step": 8816 }, { "epoch": 0.29, "learning_rate": 7.693802447156612e-06, "loss": 0.9694, "step": 8817 }, { "epoch": 0.29, "learning_rate": 7.691704849128016e-06, "loss": 0.9777, "step": 8818 }, { "epoch": 0.29, "learning_rate": 7.689607358373092e-06, "loss": 0.9273, "step": 8819 }, { "epoch": 0.29, "learning_rate": 7.68750997498932e-06, "loss": 0.9557, "step": 8820 }, { "epoch": 0.29, "learning_rate": 7.68541269907417e-06, "loss": 0.986, "step": 8821 }, { "epoch": 0.29, "learning_rate": 7.68331553072511e-06, "loss": 0.9892, "step": 8822 }, { "epoch": 0.29, "learning_rate": 7.681218470039597e-06, "loss": 0.9703, "step": 8823 }, { "epoch": 0.29, "learning_rate": 7.6791215171151e-06, "loss": 0.9349, "step": 8824 }, { "epoch": 0.29, "learning_rate": 7.67702467204906e-06, "loss": 0.94, "step": 8825 }, { "epoch": 0.29, "learning_rate": 7.674927934938923e-06, "loss": 0.9847, "step": 8826 }, { "epoch": 0.29, "learning_rate": 7.672831305882137e-06, "loss": 0.9771, "step": 8827 }, { "epoch": 0.29, "learning_rate": 7.67073478497614e-06, "loss": 0.9671, "step": 8828 }, { "epoch": 0.29, "learning_rate": 7.668638372318359e-06, "loss": 0.9864, "step": 8829 }, { "epoch": 0.29, "learning_rate": 7.666542068006222e-06, "loss": 0.9329, "step": 8830 }, { "epoch": 0.29, "learning_rate": 7.664445872137153e-06, "loss": 0.95, "step": 8831 }, { "epoch": 0.29, "learning_rate": 7.662349784808567e-06, "loss": 0.9365, "step": 8832 }, { "epoch": 0.29, "learning_rate": 7.660253806117874e-06, "loss": 0.9842, "step": 8833 }, { "epoch": 0.29, "learning_rate": 7.658157936162485e-06, "loss": 0.8913, "step": 8834 }, { "epoch": 0.29, "learning_rate": 7.656062175039796e-06, "loss": 0.9632, "step": 8835 }, { "epoch": 0.29, "learning_rate": 7.653966522847211e-06, "loss": 0.9516, "step": 8836 }, { "epoch": 0.29, "learning_rate": 7.651870979682112e-06, "loss": 0.9378, "step": 8837 }, { "epoch": 0.29, "learning_rate": 7.649775545641895e-06, "loss": 0.9271, "step": 8838 }, { "epoch": 0.29, "learning_rate": 7.647680220823939e-06, "loss": 0.9641, "step": 8839 }, { "epoch": 0.29, "learning_rate": 7.645585005325612e-06, "loss": 0.9975, "step": 8840 }, { "epoch": 0.29, "learning_rate": 7.643489899244296e-06, "loss": 0.9285, "step": 8841 }, { "epoch": 0.29, "learning_rate": 7.64139490267735e-06, "loss": 0.9327, "step": 8842 }, { "epoch": 0.29, "learning_rate": 7.63930001572214e-06, "loss": 0.9695, "step": 8843 }, { "epoch": 0.29, "learning_rate": 7.637205238476018e-06, "loss": 0.9296, "step": 8844 }, { "epoch": 0.29, "learning_rate": 7.635110571036336e-06, "loss": 0.9423, "step": 8845 }, { "epoch": 0.29, "learning_rate": 7.633016013500438e-06, "loss": 0.9469, "step": 8846 }, { "epoch": 0.29, "learning_rate": 7.630921565965673e-06, "loss": 0.9862, "step": 8847 }, { "epoch": 0.29, "learning_rate": 7.628827228529368e-06, "loss": 0.9422, "step": 8848 }, { "epoch": 0.29, "learning_rate": 7.626733001288852e-06, "loss": 1.0096, "step": 8849 }, { "epoch": 0.29, "learning_rate": 7.6246388843414574e-06, "loss": 0.9372, "step": 8850 }, { "epoch": 0.29, "learning_rate": 7.622544877784499e-06, "loss": 0.9256, "step": 8851 }, { "epoch": 0.29, "learning_rate": 7.620450981715296e-06, "loss": 0.9519, "step": 8852 }, { "epoch": 0.29, "learning_rate": 7.6183571962311535e-06, "loss": 0.9751, "step": 8853 }, { "epoch": 0.29, "learning_rate": 7.616263521429381e-06, "loss": 0.9322, "step": 8854 }, { "epoch": 0.29, "learning_rate": 7.6141699574072755e-06, "loss": 0.9236, "step": 8855 }, { "epoch": 0.29, "learning_rate": 7.612076504262129e-06, "loss": 0.9461, "step": 8856 }, { "epoch": 0.29, "learning_rate": 7.609983162091237e-06, "loss": 0.9632, "step": 8857 }, { "epoch": 0.29, "learning_rate": 7.607889930991875e-06, "loss": 0.9848, "step": 8858 }, { "epoch": 0.29, "learning_rate": 7.605796811061334e-06, "loss": 0.9747, "step": 8859 }, { "epoch": 0.29, "learning_rate": 7.6037038023968745e-06, "loss": 0.9557, "step": 8860 }, { "epoch": 0.29, "learning_rate": 7.601610905095777e-06, "loss": 0.9259, "step": 8861 }, { "epoch": 0.29, "learning_rate": 7.5995181192553e-06, "loss": 0.898, "step": 8862 }, { "epoch": 0.29, "learning_rate": 7.597425444972694e-06, "loss": 0.9463, "step": 8863 }, { "epoch": 0.29, "learning_rate": 7.5953328823452265e-06, "loss": 0.9652, "step": 8864 }, { "epoch": 0.29, "learning_rate": 7.593240431470135e-06, "loss": 0.9674, "step": 8865 }, { "epoch": 0.29, "learning_rate": 7.591148092444668e-06, "loss": 0.9324, "step": 8866 }, { "epoch": 0.29, "learning_rate": 7.589055865366057e-06, "loss": 0.9897, "step": 8867 }, { "epoch": 0.29, "learning_rate": 7.586963750331542e-06, "loss": 0.9862, "step": 8868 }, { "epoch": 0.29, "learning_rate": 7.5848717474383405e-06, "loss": 0.9417, "step": 8869 }, { "epoch": 0.29, "learning_rate": 7.582779856783686e-06, "loss": 0.97, "step": 8870 }, { "epoch": 0.29, "learning_rate": 7.5806880784647884e-06, "loss": 0.9583, "step": 8871 }, { "epoch": 0.29, "learning_rate": 7.5785964125788545e-06, "loss": 0.919, "step": 8872 }, { "epoch": 0.29, "learning_rate": 7.576504859223102e-06, "loss": 1.005, "step": 8873 }, { "epoch": 0.29, "learning_rate": 7.5744134184947215e-06, "loss": 0.9797, "step": 8874 }, { "epoch": 0.29, "learning_rate": 7.5723220904909155e-06, "loss": 0.9355, "step": 8875 }, { "epoch": 0.29, "learning_rate": 7.570230875308871e-06, "loss": 0.9487, "step": 8876 }, { "epoch": 0.29, "learning_rate": 7.568139773045775e-06, "loss": 0.9636, "step": 8877 }, { "epoch": 0.29, "learning_rate": 7.5660487837988085e-06, "loss": 0.9384, "step": 8878 }, { "epoch": 0.29, "learning_rate": 7.563957907665142e-06, "loss": 0.9362, "step": 8879 }, { "epoch": 0.29, "learning_rate": 7.56186714474195e-06, "loss": 0.939, "step": 8880 }, { "epoch": 0.29, "learning_rate": 7.55977649512639e-06, "loss": 0.9569, "step": 8881 }, { "epoch": 0.29, "learning_rate": 7.557685958915632e-06, "loss": 0.9442, "step": 8882 }, { "epoch": 0.29, "learning_rate": 7.555595536206817e-06, "loss": 0.9835, "step": 8883 }, { "epoch": 0.29, "learning_rate": 7.553505227097104e-06, "loss": 0.9285, "step": 8884 }, { "epoch": 0.29, "learning_rate": 7.551415031683632e-06, "loss": 0.965, "step": 8885 }, { "epoch": 0.29, "learning_rate": 7.549324950063537e-06, "loss": 0.9334, "step": 8886 }, { "epoch": 0.29, "learning_rate": 7.547234982333955e-06, "loss": 0.9625, "step": 8887 }, { "epoch": 0.29, "learning_rate": 7.545145128592009e-06, "loss": 0.9559, "step": 8888 }, { "epoch": 0.29, "learning_rate": 7.543055388934826e-06, "loss": 0.9799, "step": 8889 }, { "epoch": 0.29, "learning_rate": 7.54096576345952e-06, "loss": 0.9924, "step": 8890 }, { "epoch": 0.29, "learning_rate": 7.538876252263204e-06, "loss": 0.9472, "step": 8891 }, { "epoch": 0.29, "learning_rate": 7.5367868554429794e-06, "loss": 0.9552, "step": 8892 }, { "epoch": 0.29, "learning_rate": 7.534697573095954e-06, "loss": 0.9792, "step": 8893 }, { "epoch": 0.29, "learning_rate": 7.5326084053192226e-06, "loss": 0.9404, "step": 8894 }, { "epoch": 0.29, "learning_rate": 7.530519352209866e-06, "loss": 0.9493, "step": 8895 }, { "epoch": 0.29, "learning_rate": 7.52843041386498e-06, "loss": 0.9807, "step": 8896 }, { "epoch": 0.29, "learning_rate": 7.526341590381636e-06, "loss": 1.0106, "step": 8897 }, { "epoch": 0.29, "learning_rate": 7.524252881856916e-06, "loss": 0.9247, "step": 8898 }, { "epoch": 0.29, "learning_rate": 7.5221642883878824e-06, "loss": 0.9359, "step": 8899 }, { "epoch": 0.29, "learning_rate": 7.520075810071602e-06, "loss": 0.959, "step": 8900 }, { "epoch": 0.29, "learning_rate": 7.517987447005132e-06, "loss": 0.9433, "step": 8901 }, { "epoch": 0.29, "learning_rate": 7.5158991992855234e-06, "loss": 0.9523, "step": 8902 }, { "epoch": 0.29, "learning_rate": 7.513811067009826e-06, "loss": 0.9317, "step": 8903 }, { "epoch": 0.29, "learning_rate": 7.511723050275078e-06, "loss": 0.9868, "step": 8904 }, { "epoch": 0.29, "learning_rate": 7.5096351491783246e-06, "loss": 0.9295, "step": 8905 }, { "epoch": 0.29, "learning_rate": 7.507547363816585e-06, "loss": 0.9572, "step": 8906 }, { "epoch": 0.29, "learning_rate": 7.505459694286896e-06, "loss": 0.9577, "step": 8907 }, { "epoch": 0.29, "learning_rate": 7.503372140686274e-06, "loss": 0.9472, "step": 8908 }, { "epoch": 0.29, "learning_rate": 7.501284703111731e-06, "loss": 0.9119, "step": 8909 }, { "epoch": 0.29, "learning_rate": 7.499197381660282e-06, "loss": 0.9339, "step": 8910 }, { "epoch": 0.29, "learning_rate": 7.4971101764289276e-06, "loss": 0.9279, "step": 8911 }, { "epoch": 0.29, "learning_rate": 7.495023087514669e-06, "loss": 0.9423, "step": 8912 }, { "epoch": 0.29, "learning_rate": 7.4929361150144975e-06, "loss": 0.9503, "step": 8913 }, { "epoch": 0.29, "learning_rate": 7.490849259025404e-06, "loss": 0.9916, "step": 8914 }, { "epoch": 0.29, "learning_rate": 7.4887625196443645e-06, "loss": 0.9416, "step": 8915 }, { "epoch": 0.29, "learning_rate": 7.4866758969683675e-06, "loss": 0.9456, "step": 8916 }, { "epoch": 0.29, "learning_rate": 7.4845893910943786e-06, "loss": 0.9697, "step": 8917 }, { "epoch": 0.29, "learning_rate": 7.482503002119359e-06, "loss": 0.919, "step": 8918 }, { "epoch": 0.29, "learning_rate": 7.48041673014028e-06, "loss": 0.9666, "step": 8919 }, { "epoch": 0.29, "learning_rate": 7.47833057525409e-06, "loss": 0.9693, "step": 8920 }, { "epoch": 0.29, "learning_rate": 7.476244537557744e-06, "loss": 0.9723, "step": 8921 }, { "epoch": 0.29, "learning_rate": 7.474158617148181e-06, "loss": 0.993, "step": 8922 }, { "epoch": 0.29, "learning_rate": 7.472072814122346e-06, "loss": 0.9577, "step": 8923 }, { "epoch": 0.29, "learning_rate": 7.4699871285771695e-06, "loss": 0.9892, "step": 8924 }, { "epoch": 0.29, "learning_rate": 7.46790156060958e-06, "loss": 0.9388, "step": 8925 }, { "epoch": 0.29, "learning_rate": 7.4658161103165014e-06, "loss": 0.965, "step": 8926 }, { "epoch": 0.29, "learning_rate": 7.463730777794847e-06, "loss": 0.9575, "step": 8927 }, { "epoch": 0.29, "learning_rate": 7.4616455631415376e-06, "loss": 0.931, "step": 8928 }, { "epoch": 0.29, "learning_rate": 7.459560466453467e-06, "loss": 0.9439, "step": 8929 }, { "epoch": 0.29, "learning_rate": 7.457475487827548e-06, "loss": 0.9393, "step": 8930 }, { "epoch": 0.29, "learning_rate": 7.455390627360672e-06, "loss": 0.9639, "step": 8931 }, { "epoch": 0.29, "learning_rate": 7.453305885149726e-06, "loss": 0.9608, "step": 8932 }, { "epoch": 0.29, "learning_rate": 7.4512212612916e-06, "loss": 0.9595, "step": 8933 }, { "epoch": 0.29, "learning_rate": 7.4491367558831665e-06, "loss": 0.9645, "step": 8934 }, { "epoch": 0.29, "learning_rate": 7.447052369021305e-06, "loss": 0.9418, "step": 8935 }, { "epoch": 0.29, "learning_rate": 7.444968100802879e-06, "loss": 0.9573, "step": 8936 }, { "epoch": 0.29, "learning_rate": 7.442883951324753e-06, "loss": 0.9214, "step": 8937 }, { "epoch": 0.29, "learning_rate": 7.44079992068378e-06, "loss": 0.9755, "step": 8938 }, { "epoch": 0.29, "learning_rate": 7.43871600897682e-06, "loss": 0.9586, "step": 8939 }, { "epoch": 0.29, "learning_rate": 7.4366322163007145e-06, "loss": 0.9644, "step": 8940 }, { "epoch": 0.29, "learning_rate": 7.434548542752298e-06, "loss": 0.9921, "step": 8941 }, { "epoch": 0.29, "learning_rate": 7.432464988428414e-06, "loss": 0.9335, "step": 8942 }, { "epoch": 0.29, "learning_rate": 7.430381553425887e-06, "loss": 0.9581, "step": 8943 }, { "epoch": 0.29, "learning_rate": 7.428298237841543e-06, "loss": 0.9229, "step": 8944 }, { "epoch": 0.3, "learning_rate": 7.4262150417721976e-06, "loss": 0.9301, "step": 8945 }, { "epoch": 0.3, "learning_rate": 7.424131965314667e-06, "loss": 0.9392, "step": 8946 }, { "epoch": 0.3, "learning_rate": 7.422049008565757e-06, "loss": 0.9615, "step": 8947 }, { "epoch": 0.3, "learning_rate": 7.419966171622266e-06, "loss": 0.9752, "step": 8948 }, { "epoch": 0.3, "learning_rate": 7.417883454580994e-06, "loss": 0.9529, "step": 8949 }, { "epoch": 0.3, "learning_rate": 7.415800857538725e-06, "loss": 0.9251, "step": 8950 }, { "epoch": 0.3, "learning_rate": 7.413718380592256e-06, "loss": 0.9372, "step": 8951 }, { "epoch": 0.3, "learning_rate": 7.411636023838351e-06, "loss": 0.95, "step": 8952 }, { "epoch": 0.3, "learning_rate": 7.409553787373795e-06, "loss": 0.9527, "step": 8953 }, { "epoch": 0.3, "learning_rate": 7.407471671295353e-06, "loss": 0.9765, "step": 8954 }, { "epoch": 0.3, "learning_rate": 7.405389675699784e-06, "loss": 0.9725, "step": 8955 }, { "epoch": 0.3, "learning_rate": 7.40330780068385e-06, "loss": 0.9863, "step": 8956 }, { "epoch": 0.3, "learning_rate": 7.401226046344299e-06, "loss": 0.9529, "step": 8957 }, { "epoch": 0.3, "learning_rate": 7.399144412777877e-06, "loss": 0.9919, "step": 8958 }, { "epoch": 0.3, "learning_rate": 7.397062900081323e-06, "loss": 0.9464, "step": 8959 }, { "epoch": 0.3, "learning_rate": 7.394981508351375e-06, "loss": 0.9626, "step": 8960 }, { "epoch": 0.3, "learning_rate": 7.392900237684755e-06, "loss": 0.9382, "step": 8961 }, { "epoch": 0.3, "learning_rate": 7.390819088178196e-06, "loss": 0.9603, "step": 8962 }, { "epoch": 0.3, "learning_rate": 7.388738059928411e-06, "loss": 0.9802, "step": 8963 }, { "epoch": 0.3, "learning_rate": 7.386657153032104e-06, "loss": 0.9487, "step": 8964 }, { "epoch": 0.3, "learning_rate": 7.384576367585994e-06, "loss": 0.9912, "step": 8965 }, { "epoch": 0.3, "learning_rate": 7.382495703686774e-06, "loss": 0.9219, "step": 8966 }, { "epoch": 0.3, "learning_rate": 7.3804151614311405e-06, "loss": 0.9813, "step": 8967 }, { "epoch": 0.3, "learning_rate": 7.378334740915783e-06, "loss": 0.9223, "step": 8968 }, { "epoch": 0.3, "learning_rate": 7.3762544422373875e-06, "loss": 0.9694, "step": 8969 }, { "epoch": 0.3, "learning_rate": 7.374174265492627e-06, "loss": 0.964, "step": 8970 }, { "epoch": 0.3, "learning_rate": 7.372094210778177e-06, "loss": 0.9319, "step": 8971 }, { "epoch": 0.3, "learning_rate": 7.3700142781907045e-06, "loss": 0.9639, "step": 8972 }, { "epoch": 0.3, "learning_rate": 7.367934467826865e-06, "loss": 0.9449, "step": 8973 }, { "epoch": 0.3, "learning_rate": 7.365854779783324e-06, "loss": 0.9642, "step": 8974 }, { "epoch": 0.3, "learning_rate": 7.363775214156721e-06, "loss": 0.9886, "step": 8975 }, { "epoch": 0.3, "learning_rate": 7.361695771043708e-06, "loss": 0.92, "step": 8976 }, { "epoch": 0.3, "learning_rate": 7.359616450540919e-06, "loss": 0.9116, "step": 8977 }, { "epoch": 0.3, "learning_rate": 7.357537252744984e-06, "loss": 0.9236, "step": 8978 }, { "epoch": 0.3, "learning_rate": 7.355458177752537e-06, "loss": 0.9793, "step": 8979 }, { "epoch": 0.3, "learning_rate": 7.353379225660191e-06, "loss": 0.9716, "step": 8980 }, { "epoch": 0.3, "learning_rate": 7.3513003965645665e-06, "loss": 0.9499, "step": 8981 }, { "epoch": 0.3, "learning_rate": 7.349221690562269e-06, "loss": 0.9243, "step": 8982 }, { "epoch": 0.3, "learning_rate": 7.347143107749912e-06, "loss": 0.9385, "step": 8983 }, { "epoch": 0.3, "learning_rate": 7.345064648224081e-06, "loss": 0.9741, "step": 8984 }, { "epoch": 0.3, "learning_rate": 7.34298631208138e-06, "loss": 0.9727, "step": 8985 }, { "epoch": 0.3, "learning_rate": 7.340908099418389e-06, "loss": 0.9196, "step": 8986 }, { "epoch": 0.3, "learning_rate": 7.338830010331686e-06, "loss": 0.956, "step": 8987 }, { "epoch": 0.3, "learning_rate": 7.336752044917855e-06, "loss": 0.9119, "step": 8988 }, { "epoch": 0.3, "learning_rate": 7.3346742032734595e-06, "loss": 0.9936, "step": 8989 }, { "epoch": 0.3, "learning_rate": 7.332596485495066e-06, "loss": 0.9586, "step": 8990 }, { "epoch": 0.3, "learning_rate": 7.33051889167923e-06, "loss": 0.9812, "step": 8991 }, { "epoch": 0.3, "learning_rate": 7.328441421922507e-06, "loss": 0.9439, "step": 8992 }, { "epoch": 0.3, "learning_rate": 7.326364076321442e-06, "loss": 0.9561, "step": 8993 }, { "epoch": 0.3, "learning_rate": 7.324286854972573e-06, "loss": 0.9488, "step": 8994 }, { "epoch": 0.3, "learning_rate": 7.32220975797244e-06, "loss": 0.9814, "step": 8995 }, { "epoch": 0.3, "learning_rate": 7.320132785417566e-06, "loss": 0.9538, "step": 8996 }, { "epoch": 0.3, "learning_rate": 7.318055937404481e-06, "loss": 0.9439, "step": 8997 }, { "epoch": 0.3, "learning_rate": 7.315979214029698e-06, "loss": 0.967, "step": 8998 }, { "epoch": 0.3, "learning_rate": 7.313902615389733e-06, "loss": 0.9561, "step": 8999 }, { "epoch": 0.3, "learning_rate": 7.311826141581089e-06, "loss": 0.9294, "step": 9000 }, { "epoch": 0.3, "learning_rate": 7.309749792700265e-06, "loss": 0.9572, "step": 9001 }, { "epoch": 0.3, "learning_rate": 7.3076735688437585e-06, "loss": 0.9302, "step": 9002 }, { "epoch": 0.3, "learning_rate": 7.305597470108055e-06, "loss": 0.9914, "step": 9003 }, { "epoch": 0.3, "learning_rate": 7.3035214965896405e-06, "loss": 0.8938, "step": 9004 }, { "epoch": 0.3, "learning_rate": 7.301445648384986e-06, "loss": 0.9379, "step": 9005 }, { "epoch": 0.3, "learning_rate": 7.299369925590575e-06, "loss": 0.9961, "step": 9006 }, { "epoch": 0.3, "learning_rate": 7.297294328302858e-06, "loss": 0.9762, "step": 9007 }, { "epoch": 0.3, "learning_rate": 7.295218856618304e-06, "loss": 0.9633, "step": 9008 }, { "epoch": 0.3, "learning_rate": 7.293143510633367e-06, "loss": 0.958, "step": 9009 }, { "epoch": 0.3, "learning_rate": 7.291068290444485e-06, "loss": 0.9011, "step": 9010 }, { "epoch": 0.3, "learning_rate": 7.288993196148112e-06, "loss": 0.9702, "step": 9011 }, { "epoch": 0.3, "learning_rate": 7.286918227840676e-06, "loss": 0.9332, "step": 9012 }, { "epoch": 0.3, "learning_rate": 7.284843385618613e-06, "loss": 0.9333, "step": 9013 }, { "epoch": 0.3, "learning_rate": 7.282768669578342e-06, "loss": 1.0037, "step": 9014 }, { "epoch": 0.3, "learning_rate": 7.2806940798162855e-06, "loss": 0.9623, "step": 9015 }, { "epoch": 0.3, "learning_rate": 7.2786196164288546e-06, "loss": 0.9556, "step": 9016 }, { "epoch": 0.3, "learning_rate": 7.276545279512455e-06, "loss": 0.938, "step": 9017 }, { "epoch": 0.3, "learning_rate": 7.274471069163489e-06, "loss": 0.9363, "step": 9018 }, { "epoch": 0.3, "learning_rate": 7.2723969854783496e-06, "loss": 0.9363, "step": 9019 }, { "epoch": 0.3, "learning_rate": 7.27032302855343e-06, "loss": 0.9525, "step": 9020 }, { "epoch": 0.3, "learning_rate": 7.2682491984851086e-06, "loss": 0.9394, "step": 9021 }, { "epoch": 0.3, "learning_rate": 7.266175495369769e-06, "loss": 0.9489, "step": 9022 }, { "epoch": 0.3, "learning_rate": 7.2641019193037785e-06, "loss": 0.9569, "step": 9023 }, { "epoch": 0.3, "learning_rate": 7.262028470383501e-06, "loss": 0.9261, "step": 9024 }, { "epoch": 0.3, "learning_rate": 7.2599551487053e-06, "loss": 0.9591, "step": 9025 }, { "epoch": 0.3, "learning_rate": 7.257881954365525e-06, "loss": 0.9424, "step": 9026 }, { "epoch": 0.3, "learning_rate": 7.255808887460529e-06, "loss": 1.0067, "step": 9027 }, { "epoch": 0.3, "learning_rate": 7.2537359480866465e-06, "loss": 0.9655, "step": 9028 }, { "epoch": 0.3, "learning_rate": 7.251663136340226e-06, "loss": 0.9545, "step": 9029 }, { "epoch": 0.3, "learning_rate": 7.249590452317581e-06, "loss": 0.9559, "step": 9030 }, { "epoch": 0.3, "learning_rate": 7.24751789611505e-06, "loss": 0.9938, "step": 9031 }, { "epoch": 0.3, "learning_rate": 7.245445467828946e-06, "loss": 0.9657, "step": 9032 }, { "epoch": 0.3, "learning_rate": 7.243373167555576e-06, "loss": 0.9281, "step": 9033 }, { "epoch": 0.3, "learning_rate": 7.241300995391254e-06, "loss": 0.9441, "step": 9034 }, { "epoch": 0.3, "learning_rate": 7.239228951432275e-06, "loss": 0.9656, "step": 9035 }, { "epoch": 0.3, "learning_rate": 7.237157035774938e-06, "loss": 0.9721, "step": 9036 }, { "epoch": 0.3, "learning_rate": 7.235085248515526e-06, "loss": 0.9442, "step": 9037 }, { "epoch": 0.3, "learning_rate": 7.2330135897503264e-06, "loss": 0.9021, "step": 9038 }, { "epoch": 0.3, "learning_rate": 7.230942059575614e-06, "loss": 0.9812, "step": 9039 }, { "epoch": 0.3, "learning_rate": 7.2288706580876565e-06, "loss": 0.9392, "step": 9040 }, { "epoch": 0.3, "learning_rate": 7.226799385382723e-06, "loss": 0.9528, "step": 9041 }, { "epoch": 0.3, "learning_rate": 7.2247282415570655e-06, "loss": 0.9157, "step": 9042 }, { "epoch": 0.3, "learning_rate": 7.222657226706944e-06, "loss": 0.9413, "step": 9043 }, { "epoch": 0.3, "learning_rate": 7.2205863409286005e-06, "loss": 0.9837, "step": 9044 }, { "epoch": 0.3, "learning_rate": 7.218515584318277e-06, "loss": 0.977, "step": 9045 }, { "epoch": 0.3, "learning_rate": 7.216444956972211e-06, "loss": 0.9834, "step": 9046 }, { "epoch": 0.3, "learning_rate": 7.214374458986624e-06, "loss": 0.9607, "step": 9047 }, { "epoch": 0.3, "learning_rate": 7.212304090457742e-06, "loss": 0.9106, "step": 9048 }, { "epoch": 0.3, "learning_rate": 7.210233851481782e-06, "loss": 0.9592, "step": 9049 }, { "epoch": 0.3, "learning_rate": 7.208163742154957e-06, "loss": 0.9275, "step": 9050 }, { "epoch": 0.3, "learning_rate": 7.206093762573463e-06, "loss": 1.0284, "step": 9051 }, { "epoch": 0.3, "learning_rate": 7.204023912833511e-06, "loss": 0.9362, "step": 9052 }, { "epoch": 0.3, "learning_rate": 7.20195419303128e-06, "loss": 0.9494, "step": 9053 }, { "epoch": 0.3, "learning_rate": 7.199884603262965e-06, "loss": 0.9667, "step": 9054 }, { "epoch": 0.3, "learning_rate": 7.197815143624749e-06, "loss": 0.9637, "step": 9055 }, { "epoch": 0.3, "learning_rate": 7.195745814212793e-06, "loss": 1.0024, "step": 9056 }, { "epoch": 0.3, "learning_rate": 7.1936766151232795e-06, "loss": 0.9587, "step": 9057 }, { "epoch": 0.3, "learning_rate": 7.191607546452361e-06, "loss": 0.9749, "step": 9058 }, { "epoch": 0.3, "learning_rate": 7.1895386082962e-06, "loss": 0.9693, "step": 9059 }, { "epoch": 0.3, "learning_rate": 7.18746980075094e-06, "loss": 0.909, "step": 9060 }, { "epoch": 0.3, "learning_rate": 7.185401123912732e-06, "loss": 0.9397, "step": 9061 }, { "epoch": 0.3, "learning_rate": 7.183332577877711e-06, "loss": 0.9703, "step": 9062 }, { "epoch": 0.3, "learning_rate": 7.1812641627420055e-06, "loss": 0.9159, "step": 9063 }, { "epoch": 0.3, "learning_rate": 7.179195878601746e-06, "loss": 0.9467, "step": 9064 }, { "epoch": 0.3, "learning_rate": 7.1771277255530456e-06, "loss": 0.9505, "step": 9065 }, { "epoch": 0.3, "learning_rate": 7.175059703692027e-06, "loss": 0.9324, "step": 9066 }, { "epoch": 0.3, "learning_rate": 7.1729918131147895e-06, "loss": 0.9233, "step": 9067 }, { "epoch": 0.3, "learning_rate": 7.170924053917441e-06, "loss": 0.9341, "step": 9068 }, { "epoch": 0.3, "learning_rate": 7.168856426196074e-06, "loss": 0.9131, "step": 9069 }, { "epoch": 0.3, "learning_rate": 7.166788930046773e-06, "loss": 0.8924, "step": 9070 }, { "epoch": 0.3, "learning_rate": 7.164721565565629e-06, "loss": 0.9214, "step": 9071 }, { "epoch": 0.3, "learning_rate": 7.1626543328487134e-06, "loss": 0.9613, "step": 9072 }, { "epoch": 0.3, "learning_rate": 7.1605872319921e-06, "loss": 0.9784, "step": 9073 }, { "epoch": 0.3, "learning_rate": 7.158520263091847e-06, "loss": 0.9471, "step": 9074 }, { "epoch": 0.3, "learning_rate": 7.156453426244025e-06, "loss": 0.9662, "step": 9075 }, { "epoch": 0.3, "learning_rate": 7.154386721544673e-06, "loss": 0.9512, "step": 9076 }, { "epoch": 0.3, "learning_rate": 7.152320149089847e-06, "loss": 0.9241, "step": 9077 }, { "epoch": 0.3, "learning_rate": 7.150253708975583e-06, "loss": 0.9474, "step": 9078 }, { "epoch": 0.3, "learning_rate": 7.148187401297915e-06, "loss": 0.9652, "step": 9079 }, { "epoch": 0.3, "learning_rate": 7.146121226152872e-06, "loss": 0.9525, "step": 9080 }, { "epoch": 0.3, "learning_rate": 7.144055183636474e-06, "loss": 0.9361, "step": 9081 }, { "epoch": 0.3, "learning_rate": 7.141989273844738e-06, "loss": 0.9702, "step": 9082 }, { "epoch": 0.3, "learning_rate": 7.139923496873671e-06, "loss": 0.9867, "step": 9083 }, { "epoch": 0.3, "learning_rate": 7.137857852819279e-06, "loss": 0.9425, "step": 9084 }, { "epoch": 0.3, "learning_rate": 7.135792341777559e-06, "loss": 0.9729, "step": 9085 }, { "epoch": 0.3, "learning_rate": 7.133726963844495e-06, "loss": 0.9521, "step": 9086 }, { "epoch": 0.3, "learning_rate": 7.131661719116083e-06, "loss": 0.9281, "step": 9087 }, { "epoch": 0.3, "learning_rate": 7.129596607688289e-06, "loss": 0.92, "step": 9088 }, { "epoch": 0.3, "learning_rate": 7.127531629657096e-06, "loss": 0.9536, "step": 9089 }, { "epoch": 0.3, "learning_rate": 7.1254667851184625e-06, "loss": 0.9573, "step": 9090 }, { "epoch": 0.3, "learning_rate": 7.123402074168353e-06, "loss": 0.9504, "step": 9091 }, { "epoch": 0.3, "learning_rate": 7.121337496902721e-06, "loss": 0.9161, "step": 9092 }, { "epoch": 0.3, "learning_rate": 7.119273053417508e-06, "loss": 0.9801, "step": 9093 }, { "epoch": 0.3, "learning_rate": 7.117208743808661e-06, "loss": 0.9635, "step": 9094 }, { "epoch": 0.31, "learning_rate": 7.115144568172112e-06, "loss": 0.953, "step": 9095 }, { "epoch": 0.31, "learning_rate": 7.113080526603793e-06, "loss": 0.9518, "step": 9096 }, { "epoch": 0.31, "learning_rate": 7.111016619199619e-06, "loss": 0.9717, "step": 9097 }, { "epoch": 0.31, "learning_rate": 7.1089528460555165e-06, "loss": 0.9583, "step": 9098 }, { "epoch": 0.31, "learning_rate": 7.106889207267387e-06, "loss": 0.9792, "step": 9099 }, { "epoch": 0.31, "learning_rate": 7.1048257029311395e-06, "loss": 0.9425, "step": 9100 }, { "epoch": 0.31, "learning_rate": 7.10276233314267e-06, "loss": 0.957, "step": 9101 }, { "epoch": 0.31, "learning_rate": 7.1006990979978675e-06, "loss": 0.9134, "step": 9102 }, { "epoch": 0.31, "learning_rate": 7.09863599759262e-06, "loss": 0.9292, "step": 9103 }, { "epoch": 0.31, "learning_rate": 7.096573032022804e-06, "loss": 0.9558, "step": 9104 }, { "epoch": 0.31, "learning_rate": 7.094510201384295e-06, "loss": 0.9584, "step": 9105 }, { "epoch": 0.31, "learning_rate": 7.092447505772955e-06, "loss": 0.912, "step": 9106 }, { "epoch": 0.31, "learning_rate": 7.090384945284646e-06, "loss": 0.9137, "step": 9107 }, { "epoch": 0.31, "learning_rate": 7.088322520015223e-06, "loss": 0.9664, "step": 9108 }, { "epoch": 0.31, "learning_rate": 7.086260230060529e-06, "loss": 0.9227, "step": 9109 }, { "epoch": 0.31, "learning_rate": 7.084198075516412e-06, "loss": 0.967, "step": 9110 }, { "epoch": 0.31, "learning_rate": 7.082136056478697e-06, "loss": 0.9845, "step": 9111 }, { "epoch": 0.31, "learning_rate": 7.080074173043223e-06, "loss": 0.9314, "step": 9112 }, { "epoch": 0.31, "learning_rate": 7.078012425305805e-06, "loss": 0.9434, "step": 9113 }, { "epoch": 0.31, "learning_rate": 7.075950813362262e-06, "loss": 0.96, "step": 9114 }, { "epoch": 0.31, "learning_rate": 7.073889337308404e-06, "loss": 0.9149, "step": 9115 }, { "epoch": 0.31, "learning_rate": 7.071827997240031e-06, "loss": 0.9553, "step": 9116 }, { "epoch": 0.31, "learning_rate": 7.069766793252943e-06, "loss": 0.9383, "step": 9117 }, { "epoch": 0.31, "learning_rate": 7.067705725442928e-06, "loss": 0.9501, "step": 9118 }, { "epoch": 0.31, "learning_rate": 7.065644793905774e-06, "loss": 0.9459, "step": 9119 }, { "epoch": 0.31, "learning_rate": 7.063583998737251e-06, "loss": 0.9163, "step": 9120 }, { "epoch": 0.31, "learning_rate": 7.061523340033144e-06, "loss": 0.9815, "step": 9121 }, { "epoch": 0.31, "learning_rate": 7.059462817889203e-06, "loss": 0.9431, "step": 9122 }, { "epoch": 0.31, "learning_rate": 7.057402432401199e-06, "loss": 0.9379, "step": 9123 }, { "epoch": 0.31, "learning_rate": 7.05534218366488e-06, "loss": 0.9762, "step": 9124 }, { "epoch": 0.31, "learning_rate": 7.053282071775989e-06, "loss": 0.9731, "step": 9125 }, { "epoch": 0.31, "learning_rate": 7.05122209683027e-06, "loss": 0.9715, "step": 9126 }, { "epoch": 0.31, "learning_rate": 7.049162258923455e-06, "loss": 0.9544, "step": 9127 }, { "epoch": 0.31, "learning_rate": 7.047102558151273e-06, "loss": 0.9378, "step": 9128 }, { "epoch": 0.31, "learning_rate": 7.045042994609441e-06, "loss": 0.9625, "step": 9129 }, { "epoch": 0.31, "learning_rate": 7.042983568393678e-06, "loss": 0.9407, "step": 9130 }, { "epoch": 0.31, "learning_rate": 7.040924279599688e-06, "loss": 0.9234, "step": 9131 }, { "epoch": 0.31, "learning_rate": 7.03886512832317e-06, "loss": 0.9386, "step": 9132 }, { "epoch": 0.31, "learning_rate": 7.03680611465983e-06, "loss": 0.9868, "step": 9133 }, { "epoch": 0.31, "learning_rate": 7.034747238705344e-06, "loss": 0.9592, "step": 9134 }, { "epoch": 0.31, "learning_rate": 7.032688500555403e-06, "loss": 0.9246, "step": 9135 }, { "epoch": 0.31, "learning_rate": 7.030629900305679e-06, "loss": 0.9552, "step": 9136 }, { "epoch": 0.31, "learning_rate": 7.028571438051846e-06, "loss": 0.9393, "step": 9137 }, { "epoch": 0.31, "learning_rate": 7.026513113889559e-06, "loss": 0.9542, "step": 9138 }, { "epoch": 0.31, "learning_rate": 7.024454927914484e-06, "loss": 0.9652, "step": 9139 }, { "epoch": 0.31, "learning_rate": 7.022396880222266e-06, "loss": 1.0003, "step": 9140 }, { "epoch": 0.31, "learning_rate": 7.020338970908549e-06, "loss": 0.9297, "step": 9141 }, { "epoch": 0.31, "learning_rate": 7.018281200068973e-06, "loss": 0.9313, "step": 9142 }, { "epoch": 0.31, "learning_rate": 7.016223567799163e-06, "loss": 0.9697, "step": 9143 }, { "epoch": 0.31, "learning_rate": 7.014166074194755e-06, "loss": 0.9646, "step": 9144 }, { "epoch": 0.31, "learning_rate": 7.012108719351354e-06, "loss": 0.9053, "step": 9145 }, { "epoch": 0.31, "learning_rate": 7.010051503364581e-06, "loss": 0.9689, "step": 9146 }, { "epoch": 0.31, "learning_rate": 7.007994426330038e-06, "loss": 0.9249, "step": 9147 }, { "epoch": 0.31, "learning_rate": 7.005937488343324e-06, "loss": 0.9494, "step": 9148 }, { "epoch": 0.31, "learning_rate": 7.003880689500031e-06, "loss": 0.9438, "step": 9149 }, { "epoch": 0.31, "learning_rate": 7.001824029895743e-06, "loss": 0.9482, "step": 9150 }, { "epoch": 0.31, "learning_rate": 6.9997675096260455e-06, "loss": 0.954, "step": 9151 }, { "epoch": 0.31, "learning_rate": 6.9977111287865045e-06, "loss": 0.945, "step": 9152 }, { "epoch": 0.31, "learning_rate": 6.995654887472689e-06, "loss": 0.9731, "step": 9153 }, { "epoch": 0.31, "learning_rate": 6.993598785780163e-06, "loss": 0.9629, "step": 9154 }, { "epoch": 0.31, "learning_rate": 6.9915428238044694e-06, "loss": 0.9881, "step": 9155 }, { "epoch": 0.31, "learning_rate": 6.989487001641169e-06, "loss": 0.9494, "step": 9156 }, { "epoch": 0.31, "learning_rate": 6.987431319385789e-06, "loss": 0.9536, "step": 9157 }, { "epoch": 0.31, "learning_rate": 6.985375777133873e-06, "loss": 0.9796, "step": 9158 }, { "epoch": 0.31, "learning_rate": 6.983320374980944e-06, "loss": 0.9349, "step": 9159 }, { "epoch": 0.31, "learning_rate": 6.981265113022526e-06, "loss": 0.9586, "step": 9160 }, { "epoch": 0.31, "learning_rate": 6.979209991354129e-06, "loss": 0.9332, "step": 9161 }, { "epoch": 0.31, "learning_rate": 6.977155010071266e-06, "loss": 0.9615, "step": 9162 }, { "epoch": 0.31, "learning_rate": 6.975100169269434e-06, "loss": 0.9522, "step": 9163 }, { "epoch": 0.31, "learning_rate": 6.973045469044128e-06, "loss": 0.9246, "step": 9164 }, { "epoch": 0.31, "learning_rate": 6.970990909490843e-06, "loss": 0.9301, "step": 9165 }, { "epoch": 0.31, "learning_rate": 6.9689364907050485e-06, "loss": 0.9249, "step": 9166 }, { "epoch": 0.31, "learning_rate": 6.966882212782234e-06, "loss": 0.9107, "step": 9167 }, { "epoch": 0.31, "learning_rate": 6.9648280758178545e-06, "loss": 0.9291, "step": 9168 }, { "epoch": 0.31, "learning_rate": 6.962774079907384e-06, "loss": 0.9581, "step": 9169 }, { "epoch": 0.31, "learning_rate": 6.960720225146272e-06, "loss": 0.9269, "step": 9170 }, { "epoch": 0.31, "learning_rate": 6.958666511629966e-06, "loss": 0.9464, "step": 9171 }, { "epoch": 0.31, "learning_rate": 6.956612939453914e-06, "loss": 0.9449, "step": 9172 }, { "epoch": 0.31, "learning_rate": 6.954559508713545e-06, "loss": 0.9323, "step": 9173 }, { "epoch": 0.31, "learning_rate": 6.952506219504297e-06, "loss": 0.918, "step": 9174 }, { "epoch": 0.31, "learning_rate": 6.950453071921583e-06, "loss": 0.9397, "step": 9175 }, { "epoch": 0.31, "learning_rate": 6.948400066060828e-06, "loss": 0.9317, "step": 9176 }, { "epoch": 0.31, "learning_rate": 6.9463472020174385e-06, "loss": 0.9687, "step": 9177 }, { "epoch": 0.31, "learning_rate": 6.94429447988681e-06, "loss": 0.9262, "step": 9178 }, { "epoch": 0.31, "learning_rate": 6.942241899764352e-06, "loss": 0.9603, "step": 9179 }, { "epoch": 0.31, "learning_rate": 6.940189461745444e-06, "loss": 0.9302, "step": 9180 }, { "epoch": 0.31, "learning_rate": 6.938137165925475e-06, "loss": 0.9479, "step": 9181 }, { "epoch": 0.31, "learning_rate": 6.936085012399818e-06, "loss": 0.9036, "step": 9182 }, { "epoch": 0.31, "learning_rate": 6.934033001263847e-06, "loss": 0.9182, "step": 9183 }, { "epoch": 0.31, "learning_rate": 6.931981132612922e-06, "loss": 0.9822, "step": 9184 }, { "epoch": 0.31, "learning_rate": 6.9299294065424e-06, "loss": 0.9511, "step": 9185 }, { "epoch": 0.31, "learning_rate": 6.927877823147633e-06, "loss": 0.9728, "step": 9186 }, { "epoch": 0.31, "learning_rate": 6.925826382523961e-06, "loss": 0.9776, "step": 9187 }, { "epoch": 0.31, "learning_rate": 6.923775084766726e-06, "loss": 0.9349, "step": 9188 }, { "epoch": 0.31, "learning_rate": 6.921723929971251e-06, "loss": 0.967, "step": 9189 }, { "epoch": 0.31, "learning_rate": 6.919672918232868e-06, "loss": 0.9532, "step": 9190 }, { "epoch": 0.31, "learning_rate": 6.917622049646888e-06, "loss": 0.9604, "step": 9191 }, { "epoch": 0.31, "learning_rate": 6.915571324308624e-06, "loss": 0.9251, "step": 9192 }, { "epoch": 0.31, "learning_rate": 6.91352074231338e-06, "loss": 0.9577, "step": 9193 }, { "epoch": 0.31, "learning_rate": 6.911470303756449e-06, "loss": 0.9333, "step": 9194 }, { "epoch": 0.31, "learning_rate": 6.909420008733126e-06, "loss": 0.9639, "step": 9195 }, { "epoch": 0.31, "learning_rate": 6.907369857338688e-06, "loss": 0.9067, "step": 9196 }, { "epoch": 0.31, "learning_rate": 6.905319849668422e-06, "loss": 0.9552, "step": 9197 }, { "epoch": 0.31, "learning_rate": 6.903269985817587e-06, "loss": 0.9392, "step": 9198 }, { "epoch": 0.31, "learning_rate": 6.901220265881458e-06, "loss": 0.9323, "step": 9199 }, { "epoch": 0.31, "learning_rate": 6.899170689955284e-06, "loss": 0.9216, "step": 9200 }, { "epoch": 0.31, "learning_rate": 6.897121258134314e-06, "loss": 0.9242, "step": 9201 }, { "epoch": 0.31, "learning_rate": 6.8950719705138e-06, "loss": 0.9034, "step": 9202 }, { "epoch": 0.31, "learning_rate": 6.893022827188969e-06, "loss": 0.9704, "step": 9203 }, { "epoch": 0.31, "learning_rate": 6.890973828255059e-06, "loss": 1.0046, "step": 9204 }, { "epoch": 0.31, "learning_rate": 6.888924973807287e-06, "loss": 0.959, "step": 9205 }, { "epoch": 0.31, "learning_rate": 6.886876263940875e-06, "loss": 0.9925, "step": 9206 }, { "epoch": 0.31, "learning_rate": 6.8848276987510286e-06, "loss": 0.9537, "step": 9207 }, { "epoch": 0.31, "learning_rate": 6.882779278332955e-06, "loss": 0.9224, "step": 9208 }, { "epoch": 0.31, "learning_rate": 6.880731002781848e-06, "loss": 0.9497, "step": 9209 }, { "epoch": 0.31, "learning_rate": 6.878682872192898e-06, "loss": 0.9524, "step": 9210 }, { "epoch": 0.31, "learning_rate": 6.876634886661289e-06, "loss": 0.9664, "step": 9211 }, { "epoch": 0.31, "learning_rate": 6.874587046282192e-06, "loss": 0.9433, "step": 9212 }, { "epoch": 0.31, "learning_rate": 6.8725393511507845e-06, "loss": 0.9472, "step": 9213 }, { "epoch": 0.31, "learning_rate": 6.870491801362224e-06, "loss": 0.9066, "step": 9214 }, { "epoch": 0.31, "learning_rate": 6.868444397011671e-06, "loss": 0.899, "step": 9215 }, { "epoch": 0.31, "learning_rate": 6.866397138194271e-06, "loss": 0.9284, "step": 9216 }, { "epoch": 0.31, "learning_rate": 6.864350025005165e-06, "loss": 0.9607, "step": 9217 }, { "epoch": 0.31, "learning_rate": 6.862303057539495e-06, "loss": 0.9246, "step": 9218 }, { "epoch": 0.31, "learning_rate": 6.860256235892382e-06, "loss": 0.9341, "step": 9219 }, { "epoch": 0.31, "learning_rate": 6.858209560158955e-06, "loss": 0.9312, "step": 9220 }, { "epoch": 0.31, "learning_rate": 6.856163030434323e-06, "loss": 0.9256, "step": 9221 }, { "epoch": 0.31, "learning_rate": 6.8541166468136046e-06, "loss": 0.9858, "step": 9222 }, { "epoch": 0.31, "learning_rate": 6.852070409391893e-06, "loss": 0.9462, "step": 9223 }, { "epoch": 0.31, "learning_rate": 6.850024318264281e-06, "loss": 0.9818, "step": 9224 }, { "epoch": 0.31, "learning_rate": 6.847978373525867e-06, "loss": 0.9748, "step": 9225 }, { "epoch": 0.31, "learning_rate": 6.845932575271722e-06, "loss": 0.9612, "step": 9226 }, { "epoch": 0.31, "learning_rate": 6.8438869235969294e-06, "loss": 0.9738, "step": 9227 }, { "epoch": 0.31, "learning_rate": 6.841841418596551e-06, "loss": 0.9272, "step": 9228 }, { "epoch": 0.31, "learning_rate": 6.839796060365651e-06, "loss": 0.9906, "step": 9229 }, { "epoch": 0.31, "learning_rate": 6.837750848999282e-06, "loss": 0.9416, "step": 9230 }, { "epoch": 0.31, "learning_rate": 6.835705784592492e-06, "loss": 0.9458, "step": 9231 }, { "epoch": 0.31, "learning_rate": 6.833660867240323e-06, "loss": 0.9625, "step": 9232 }, { "epoch": 0.31, "learning_rate": 6.831616097037804e-06, "loss": 0.9691, "step": 9233 }, { "epoch": 0.31, "learning_rate": 6.829571474079968e-06, "loss": 0.9767, "step": 9234 }, { "epoch": 0.31, "learning_rate": 6.827526998461826e-06, "loss": 0.9729, "step": 9235 }, { "epoch": 0.31, "learning_rate": 6.825482670278403e-06, "loss": 0.9513, "step": 9236 }, { "epoch": 0.31, "learning_rate": 6.823438489624695e-06, "loss": 0.9699, "step": 9237 }, { "epoch": 0.31, "learning_rate": 6.821394456595709e-06, "loss": 0.9549, "step": 9238 }, { "epoch": 0.31, "learning_rate": 6.819350571286434e-06, "loss": 0.9382, "step": 9239 }, { "epoch": 0.31, "learning_rate": 6.817306833791854e-06, "loss": 0.9135, "step": 9240 }, { "epoch": 0.31, "learning_rate": 6.815263244206953e-06, "loss": 0.9263, "step": 9241 }, { "epoch": 0.31, "learning_rate": 6.813219802626698e-06, "loss": 0.9373, "step": 9242 }, { "epoch": 0.31, "learning_rate": 6.811176509146057e-06, "loss": 0.9665, "step": 9243 }, { "epoch": 0.31, "learning_rate": 6.809133363859983e-06, "loss": 0.934, "step": 9244 }, { "epoch": 0.32, "learning_rate": 6.807090366863438e-06, "loss": 0.999, "step": 9245 }, { "epoch": 0.32, "learning_rate": 6.805047518251359e-06, "loss": 0.9586, "step": 9246 }, { "epoch": 0.32, "learning_rate": 6.803004818118679e-06, "loss": 0.9863, "step": 9247 }, { "epoch": 0.32, "learning_rate": 6.800962266560342e-06, "loss": 0.894, "step": 9248 }, { "epoch": 0.32, "learning_rate": 6.798919863671256e-06, "loss": 0.9357, "step": 9249 }, { "epoch": 0.32, "learning_rate": 6.79687760954635e-06, "loss": 0.9719, "step": 9250 }, { "epoch": 0.32, "learning_rate": 6.794835504280528e-06, "loss": 0.9467, "step": 9251 }, { "epoch": 0.32, "learning_rate": 6.7927935479686965e-06, "loss": 0.948, "step": 9252 }, { "epoch": 0.32, "learning_rate": 6.79075174070575e-06, "loss": 0.9687, "step": 9253 }, { "epoch": 0.32, "learning_rate": 6.788710082586577e-06, "loss": 0.9408, "step": 9254 }, { "epoch": 0.32, "learning_rate": 6.786668573706061e-06, "loss": 0.9476, "step": 9255 }, { "epoch": 0.32, "learning_rate": 6.784627214159074e-06, "loss": 0.9408, "step": 9256 }, { "epoch": 0.32, "learning_rate": 6.782586004040489e-06, "loss": 0.9388, "step": 9257 }, { "epoch": 0.32, "learning_rate": 6.78054494344516e-06, "loss": 0.9836, "step": 9258 }, { "epoch": 0.32, "learning_rate": 6.778504032467952e-06, "loss": 0.9613, "step": 9259 }, { "epoch": 0.32, "learning_rate": 6.776463271203705e-06, "loss": 0.9219, "step": 9260 }, { "epoch": 0.32, "learning_rate": 6.774422659747264e-06, "loss": 0.9721, "step": 9261 }, { "epoch": 0.32, "learning_rate": 6.772382198193459e-06, "loss": 0.9744, "step": 9262 }, { "epoch": 0.32, "learning_rate": 6.770341886637118e-06, "loss": 0.9428, "step": 9263 }, { "epoch": 0.32, "learning_rate": 6.7683017251730584e-06, "loss": 0.9265, "step": 9264 }, { "epoch": 0.32, "learning_rate": 6.766261713896096e-06, "loss": 0.9216, "step": 9265 }, { "epoch": 0.32, "learning_rate": 6.764221852901038e-06, "loss": 0.9118, "step": 9266 }, { "epoch": 0.32, "learning_rate": 6.762182142282675e-06, "loss": 0.9165, "step": 9267 }, { "epoch": 0.32, "learning_rate": 6.760142582135811e-06, "loss": 0.9339, "step": 9268 }, { "epoch": 0.32, "learning_rate": 6.758103172555221e-06, "loss": 0.9769, "step": 9269 }, { "epoch": 0.32, "learning_rate": 6.756063913635682e-06, "loss": 0.9289, "step": 9270 }, { "epoch": 0.32, "learning_rate": 6.754024805471973e-06, "loss": 0.9679, "step": 9271 }, { "epoch": 0.32, "learning_rate": 6.751985848158851e-06, "loss": 1.0206, "step": 9272 }, { "epoch": 0.32, "learning_rate": 6.7499470417910765e-06, "loss": 0.9898, "step": 9273 }, { "epoch": 0.32, "learning_rate": 6.7479083864633956e-06, "loss": 0.925, "step": 9274 }, { "epoch": 0.32, "learning_rate": 6.745869882270555e-06, "loss": 0.9399, "step": 9275 }, { "epoch": 0.32, "learning_rate": 6.743831529307287e-06, "loss": 0.9412, "step": 9276 }, { "epoch": 0.32, "learning_rate": 6.741793327668322e-06, "loss": 0.9689, "step": 9277 }, { "epoch": 0.32, "learning_rate": 6.739755277448381e-06, "loss": 0.9573, "step": 9278 }, { "epoch": 0.32, "learning_rate": 6.737717378742175e-06, "loss": 0.92, "step": 9279 }, { "epoch": 0.32, "learning_rate": 6.735679631644423e-06, "loss": 0.9564, "step": 9280 }, { "epoch": 0.32, "learning_rate": 6.733642036249809e-06, "loss": 0.9764, "step": 9281 }, { "epoch": 0.32, "learning_rate": 6.731604592653039e-06, "loss": 0.9646, "step": 9282 }, { "epoch": 0.32, "learning_rate": 6.729567300948794e-06, "loss": 0.9249, "step": 9283 }, { "epoch": 0.32, "learning_rate": 6.727530161231757e-06, "loss": 0.9596, "step": 9284 }, { "epoch": 0.32, "learning_rate": 6.725493173596597e-06, "loss": 0.9604, "step": 9285 }, { "epoch": 0.32, "learning_rate": 6.723456338137979e-06, "loss": 0.9583, "step": 9286 }, { "epoch": 0.32, "learning_rate": 6.721419654950565e-06, "loss": 0.9691, "step": 9287 }, { "epoch": 0.32, "learning_rate": 6.7193831241289995e-06, "loss": 0.9471, "step": 9288 }, { "epoch": 0.32, "learning_rate": 6.717346745767932e-06, "loss": 0.9489, "step": 9289 }, { "epoch": 0.32, "learning_rate": 6.7153105199619936e-06, "loss": 0.9301, "step": 9290 }, { "epoch": 0.32, "learning_rate": 6.713274446805824e-06, "loss": 0.939, "step": 9291 }, { "epoch": 0.32, "learning_rate": 6.711238526394039e-06, "loss": 0.9377, "step": 9292 }, { "epoch": 0.32, "learning_rate": 6.709202758821249e-06, "loss": 0.9191, "step": 9293 }, { "epoch": 0.32, "learning_rate": 6.707167144182074e-06, "loss": 0.9612, "step": 9294 }, { "epoch": 0.32, "learning_rate": 6.705131682571107e-06, "loss": 0.9606, "step": 9295 }, { "epoch": 0.32, "learning_rate": 6.7030963740829466e-06, "loss": 0.9355, "step": 9296 }, { "epoch": 0.32, "learning_rate": 6.701061218812177e-06, "loss": 0.9567, "step": 9297 }, { "epoch": 0.32, "learning_rate": 6.699026216853381e-06, "loss": 0.9424, "step": 9298 }, { "epoch": 0.32, "learning_rate": 6.6969913683011274e-06, "loss": 0.9254, "step": 9299 }, { "epoch": 0.32, "learning_rate": 6.694956673249988e-06, "loss": 0.9498, "step": 9300 }, { "epoch": 0.32, "learning_rate": 6.692922131794517e-06, "loss": 0.9909, "step": 9301 }, { "epoch": 0.32, "learning_rate": 6.690887744029264e-06, "loss": 0.9557, "step": 9302 }, { "epoch": 0.32, "learning_rate": 6.688853510048782e-06, "loss": 0.9437, "step": 9303 }, { "epoch": 0.32, "learning_rate": 6.686819429947595e-06, "loss": 0.9836, "step": 9304 }, { "epoch": 0.32, "learning_rate": 6.684785503820246e-06, "loss": 0.9994, "step": 9305 }, { "epoch": 0.32, "learning_rate": 6.68275173176125e-06, "loss": 0.9326, "step": 9306 }, { "epoch": 0.32, "learning_rate": 6.680718113865127e-06, "loss": 0.9532, "step": 9307 }, { "epoch": 0.32, "learning_rate": 6.678684650226383e-06, "loss": 0.976, "step": 9308 }, { "epoch": 0.32, "learning_rate": 6.67665134093952e-06, "loss": 0.9764, "step": 9309 }, { "epoch": 0.32, "learning_rate": 6.674618186099032e-06, "loss": 0.9453, "step": 9310 }, { "epoch": 0.32, "learning_rate": 6.672585185799405e-06, "loss": 0.961, "step": 9311 }, { "epoch": 0.32, "learning_rate": 6.670552340135123e-06, "loss": 0.9276, "step": 9312 }, { "epoch": 0.32, "learning_rate": 6.668519649200651e-06, "loss": 0.9788, "step": 9313 }, { "epoch": 0.32, "learning_rate": 6.666487113090466e-06, "loss": 0.9896, "step": 9314 }, { "epoch": 0.32, "learning_rate": 6.664454731899017e-06, "loss": 0.899, "step": 9315 }, { "epoch": 0.32, "learning_rate": 6.6624225057207545e-06, "loss": 0.9034, "step": 9316 }, { "epoch": 0.32, "learning_rate": 6.6603904346501304e-06, "loss": 0.9696, "step": 9317 }, { "epoch": 0.32, "learning_rate": 6.658358518781573e-06, "loss": 0.8926, "step": 9318 }, { "epoch": 0.32, "learning_rate": 6.656326758209517e-06, "loss": 0.9582, "step": 9319 }, { "epoch": 0.32, "learning_rate": 6.654295153028383e-06, "loss": 0.95, "step": 9320 }, { "epoch": 0.32, "learning_rate": 6.652263703332587e-06, "loss": 0.9689, "step": 9321 }, { "epoch": 0.32, "learning_rate": 6.6502324092165326e-06, "loss": 0.9615, "step": 9322 }, { "epoch": 0.32, "learning_rate": 6.648201270774628e-06, "loss": 0.9222, "step": 9323 }, { "epoch": 0.32, "learning_rate": 6.64617028810126e-06, "loss": 0.9602, "step": 9324 }, { "epoch": 0.32, "learning_rate": 6.644139461290814e-06, "loss": 0.982, "step": 9325 }, { "epoch": 0.32, "learning_rate": 6.642108790437678e-06, "loss": 0.9187, "step": 9326 }, { "epoch": 0.32, "learning_rate": 6.64007827563621e-06, "loss": 0.9654, "step": 9327 }, { "epoch": 0.32, "learning_rate": 6.638047916980787e-06, "loss": 0.9294, "step": 9328 }, { "epoch": 0.32, "learning_rate": 6.636017714565758e-06, "loss": 0.92, "step": 9329 }, { "epoch": 0.32, "learning_rate": 6.633987668485478e-06, "loss": 0.9701, "step": 9330 }, { "epoch": 0.32, "learning_rate": 6.631957778834287e-06, "loss": 0.9229, "step": 9331 }, { "epoch": 0.32, "learning_rate": 6.6299280457065185e-06, "loss": 0.9238, "step": 9332 }, { "epoch": 0.32, "learning_rate": 6.627898469196506e-06, "loss": 0.9127, "step": 9333 }, { "epoch": 0.32, "learning_rate": 6.625869049398562e-06, "loss": 0.9409, "step": 9334 }, { "epoch": 0.32, "learning_rate": 6.623839786407008e-06, "loss": 0.9414, "step": 9335 }, { "epoch": 0.32, "learning_rate": 6.621810680316143e-06, "loss": 0.8825, "step": 9336 }, { "epoch": 0.32, "learning_rate": 6.6197817312202765e-06, "loss": 0.9098, "step": 9337 }, { "epoch": 0.32, "learning_rate": 6.61775293921369e-06, "loss": 0.9187, "step": 9338 }, { "epoch": 0.32, "learning_rate": 6.615724304390667e-06, "loss": 0.9927, "step": 9339 }, { "epoch": 0.32, "learning_rate": 6.6136958268454945e-06, "loss": 0.9587, "step": 9340 }, { "epoch": 0.32, "learning_rate": 6.611667506672433e-06, "loss": 0.9771, "step": 9341 }, { "epoch": 0.32, "learning_rate": 6.60963934396575e-06, "loss": 0.9716, "step": 9342 }, { "epoch": 0.32, "learning_rate": 6.607611338819697e-06, "loss": 0.935, "step": 9343 }, { "epoch": 0.32, "learning_rate": 6.605583491328526e-06, "loss": 0.9533, "step": 9344 }, { "epoch": 0.32, "learning_rate": 6.603555801586473e-06, "loss": 0.9435, "step": 9345 }, { "epoch": 0.32, "learning_rate": 6.601528269687775e-06, "loss": 0.9377, "step": 9346 }, { "epoch": 0.32, "learning_rate": 6.599500895726655e-06, "loss": 0.9303, "step": 9347 }, { "epoch": 0.32, "learning_rate": 6.597473679797328e-06, "loss": 0.919, "step": 9348 }, { "epoch": 0.32, "learning_rate": 6.595446621994015e-06, "loss": 0.9284, "step": 9349 }, { "epoch": 0.32, "learning_rate": 6.593419722410907e-06, "loss": 0.9704, "step": 9350 }, { "epoch": 0.32, "learning_rate": 6.5913929811422126e-06, "loss": 0.9544, "step": 9351 }, { "epoch": 0.32, "learning_rate": 6.589366398282113e-06, "loss": 0.9296, "step": 9352 }, { "epoch": 0.32, "learning_rate": 6.587339973924795e-06, "loss": 0.9619, "step": 9353 }, { "epoch": 0.32, "learning_rate": 6.585313708164429e-06, "loss": 0.9381, "step": 9354 }, { "epoch": 0.32, "learning_rate": 6.583287601095181e-06, "loss": 0.9686, "step": 9355 }, { "epoch": 0.32, "learning_rate": 6.581261652811215e-06, "loss": 0.9768, "step": 9356 }, { "epoch": 0.32, "learning_rate": 6.579235863406678e-06, "loss": 0.9632, "step": 9357 }, { "epoch": 0.32, "learning_rate": 6.5772102329757194e-06, "loss": 0.9558, "step": 9358 }, { "epoch": 0.32, "learning_rate": 6.5751847616124695e-06, "loss": 0.9103, "step": 9359 }, { "epoch": 0.32, "learning_rate": 6.573159449411071e-06, "loss": 0.9255, "step": 9360 }, { "epoch": 0.32, "learning_rate": 6.5711342964656356e-06, "loss": 0.9131, "step": 9361 }, { "epoch": 0.32, "learning_rate": 6.569109302870277e-06, "loss": 0.9206, "step": 9362 }, { "epoch": 0.32, "learning_rate": 6.567084468719112e-06, "loss": 0.9383, "step": 9363 }, { "epoch": 0.32, "learning_rate": 6.5650597941062345e-06, "loss": 0.9438, "step": 9364 }, { "epoch": 0.32, "learning_rate": 6.563035279125742e-06, "loss": 0.9381, "step": 9365 }, { "epoch": 0.32, "learning_rate": 6.561010923871716e-06, "loss": 0.9271, "step": 9366 }, { "epoch": 0.32, "learning_rate": 6.5589867284382366e-06, "loss": 0.9756, "step": 9367 }, { "epoch": 0.32, "learning_rate": 6.556962692919373e-06, "loss": 0.9543, "step": 9368 }, { "epoch": 0.32, "learning_rate": 6.554938817409191e-06, "loss": 0.9623, "step": 9369 }, { "epoch": 0.32, "learning_rate": 6.552915102001744e-06, "loss": 0.9848, "step": 9370 }, { "epoch": 0.32, "learning_rate": 6.550891546791079e-06, "loss": 0.9862, "step": 9371 }, { "epoch": 0.32, "learning_rate": 6.548868151871244e-06, "loss": 0.9006, "step": 9372 }, { "epoch": 0.32, "learning_rate": 6.54684491733626e-06, "loss": 0.927, "step": 9373 }, { "epoch": 0.32, "learning_rate": 6.544821843280167e-06, "loss": 0.9869, "step": 9374 }, { "epoch": 0.32, "learning_rate": 6.542798929796974e-06, "loss": 0.97, "step": 9375 }, { "epoch": 0.32, "learning_rate": 6.5407761769806965e-06, "loss": 0.9444, "step": 9376 }, { "epoch": 0.32, "learning_rate": 6.538753584925338e-06, "loss": 0.953, "step": 9377 }, { "epoch": 0.32, "learning_rate": 6.536731153724891e-06, "loss": 0.9103, "step": 9378 }, { "epoch": 0.32, "learning_rate": 6.5347088834733496e-06, "loss": 0.9782, "step": 9379 }, { "epoch": 0.32, "learning_rate": 6.532686774264689e-06, "loss": 0.9595, "step": 9380 }, { "epoch": 0.32, "learning_rate": 6.530664826192888e-06, "loss": 0.9633, "step": 9381 }, { "epoch": 0.32, "learning_rate": 6.528643039351907e-06, "loss": 0.9344, "step": 9382 }, { "epoch": 0.32, "learning_rate": 6.526621413835714e-06, "loss": 0.968, "step": 9383 }, { "epoch": 0.32, "learning_rate": 6.524599949738255e-06, "loss": 0.9412, "step": 9384 }, { "epoch": 0.32, "learning_rate": 6.522578647153468e-06, "loss": 0.9001, "step": 9385 }, { "epoch": 0.32, "learning_rate": 6.5205575061753e-06, "loss": 0.986, "step": 9386 }, { "epoch": 0.32, "learning_rate": 6.518536526897672e-06, "loss": 0.9802, "step": 9387 }, { "epoch": 0.32, "learning_rate": 6.516515709414508e-06, "loss": 0.9745, "step": 9388 }, { "epoch": 0.32, "learning_rate": 6.514495053819722e-06, "loss": 0.9465, "step": 9389 }, { "epoch": 0.32, "learning_rate": 6.512474560207221e-06, "loss": 0.9422, "step": 9390 }, { "epoch": 0.32, "learning_rate": 6.510454228670897e-06, "loss": 0.9274, "step": 9391 }, { "epoch": 0.32, "learning_rate": 6.5084340593046526e-06, "loss": 0.928, "step": 9392 }, { "epoch": 0.32, "learning_rate": 6.506414052202364e-06, "loss": 0.9083, "step": 9393 }, { "epoch": 0.32, "learning_rate": 6.504394207457902e-06, "loss": 0.9414, "step": 9394 }, { "epoch": 0.33, "learning_rate": 6.502374525165151e-06, "loss": 0.9207, "step": 9395 }, { "epoch": 0.33, "learning_rate": 6.500355005417955e-06, "loss": 0.9277, "step": 9396 }, { "epoch": 0.33, "learning_rate": 6.4983356483101765e-06, "loss": 0.968, "step": 9397 }, { "epoch": 0.33, "learning_rate": 6.496316453935661e-06, "loss": 0.9663, "step": 9398 }, { "epoch": 0.33, "learning_rate": 6.494297422388245e-06, "loss": 0.9034, "step": 9399 }, { "epoch": 0.33, "learning_rate": 6.49227855376176e-06, "loss": 0.9573, "step": 9400 }, { "epoch": 0.33, "learning_rate": 6.490259848150027e-06, "loss": 0.9797, "step": 9401 }, { "epoch": 0.33, "learning_rate": 6.488241305646865e-06, "loss": 0.9455, "step": 9402 }, { "epoch": 0.33, "learning_rate": 6.4862229263460775e-06, "loss": 0.9901, "step": 9403 }, { "epoch": 0.33, "learning_rate": 6.48420471034147e-06, "loss": 0.9712, "step": 9404 }, { "epoch": 0.33, "learning_rate": 6.48218665772683e-06, "loss": 0.9572, "step": 9405 }, { "epoch": 0.33, "learning_rate": 6.480168768595948e-06, "loss": 0.9548, "step": 9406 }, { "epoch": 0.33, "learning_rate": 6.478151043042602e-06, "loss": 0.9639, "step": 9407 }, { "epoch": 0.33, "learning_rate": 6.476133481160553e-06, "loss": 0.9335, "step": 9408 }, { "epoch": 0.33, "learning_rate": 6.4741160830435735e-06, "loss": 0.9677, "step": 9409 }, { "epoch": 0.33, "learning_rate": 6.472098848785411e-06, "loss": 0.9512, "step": 9410 }, { "epoch": 0.33, "learning_rate": 6.4700817784798195e-06, "loss": 0.9961, "step": 9411 }, { "epoch": 0.33, "learning_rate": 6.468064872220532e-06, "loss": 0.9163, "step": 9412 }, { "epoch": 0.33, "learning_rate": 6.466048130101286e-06, "loss": 0.9589, "step": 9413 }, { "epoch": 0.33, "learning_rate": 6.464031552215799e-06, "loss": 0.9621, "step": 9414 }, { "epoch": 0.33, "learning_rate": 6.462015138657798e-06, "loss": 0.9448, "step": 9415 }, { "epoch": 0.33, "learning_rate": 6.459998889520984e-06, "loss": 0.9579, "step": 9416 }, { "epoch": 0.33, "learning_rate": 6.457982804899056e-06, "loss": 0.9905, "step": 9417 }, { "epoch": 0.33, "learning_rate": 6.455966884885718e-06, "loss": 0.9695, "step": 9418 }, { "epoch": 0.33, "learning_rate": 6.453951129574644e-06, "loss": 0.9512, "step": 9419 }, { "epoch": 0.33, "learning_rate": 6.451935539059522e-06, "loss": 0.9447, "step": 9420 }, { "epoch": 0.33, "learning_rate": 6.449920113434018e-06, "loss": 0.9424, "step": 9421 }, { "epoch": 0.33, "learning_rate": 6.4479048527917996e-06, "loss": 0.974, "step": 9422 }, { "epoch": 0.33, "learning_rate": 6.445889757226517e-06, "loss": 0.9337, "step": 9423 }, { "epoch": 0.33, "learning_rate": 6.44387482683182e-06, "loss": 0.937, "step": 9424 }, { "epoch": 0.33, "learning_rate": 6.441860061701349e-06, "loss": 0.9615, "step": 9425 }, { "epoch": 0.33, "learning_rate": 6.439845461928735e-06, "loss": 0.9698, "step": 9426 }, { "epoch": 0.33, "learning_rate": 6.437831027607606e-06, "loss": 0.952, "step": 9427 }, { "epoch": 0.33, "learning_rate": 6.435816758831572e-06, "loss": 0.9334, "step": 9428 }, { "epoch": 0.33, "learning_rate": 6.433802655694252e-06, "loss": 0.9269, "step": 9429 }, { "epoch": 0.33, "learning_rate": 6.431788718289247e-06, "loss": 0.9047, "step": 9430 }, { "epoch": 0.33, "learning_rate": 6.429774946710139e-06, "loss": 0.9296, "step": 9431 }, { "epoch": 0.33, "learning_rate": 6.427761341050525e-06, "loss": 0.9658, "step": 9432 }, { "epoch": 0.33, "learning_rate": 6.425747901403981e-06, "loss": 0.9743, "step": 9433 }, { "epoch": 0.33, "learning_rate": 6.423734627864078e-06, "loss": 0.9598, "step": 9434 }, { "epoch": 0.33, "learning_rate": 6.421721520524377e-06, "loss": 0.9346, "step": 9435 }, { "epoch": 0.33, "learning_rate": 6.419708579478436e-06, "loss": 0.9681, "step": 9436 }, { "epoch": 0.33, "learning_rate": 6.417695804819799e-06, "loss": 0.9357, "step": 9437 }, { "epoch": 0.33, "learning_rate": 6.415683196642014e-06, "loss": 0.9388, "step": 9438 }, { "epoch": 0.33, "learning_rate": 6.413670755038606e-06, "loss": 0.9505, "step": 9439 }, { "epoch": 0.33, "learning_rate": 6.411658480103097e-06, "loss": 0.9289, "step": 9440 }, { "epoch": 0.33, "learning_rate": 6.409646371929013e-06, "loss": 0.9169, "step": 9441 }, { "epoch": 0.33, "learning_rate": 6.407634430609853e-06, "loss": 0.9967, "step": 9442 }, { "epoch": 0.33, "learning_rate": 6.405622656239127e-06, "loss": 0.9023, "step": 9443 }, { "epoch": 0.33, "learning_rate": 6.403611048910321e-06, "loss": 0.9417, "step": 9444 }, { "epoch": 0.33, "learning_rate": 6.401599608716929e-06, "loss": 0.9723, "step": 9445 }, { "epoch": 0.33, "learning_rate": 6.3995883357524205e-06, "loss": 0.9702, "step": 9446 }, { "epoch": 0.33, "learning_rate": 6.397577230110269e-06, "loss": 0.9449, "step": 9447 }, { "epoch": 0.33, "learning_rate": 6.395566291883939e-06, "loss": 0.9362, "step": 9448 }, { "epoch": 0.33, "learning_rate": 6.39355552116688e-06, "loss": 0.9446, "step": 9449 }, { "epoch": 0.33, "learning_rate": 6.391544918052544e-06, "loss": 0.9269, "step": 9450 }, { "epoch": 0.33, "learning_rate": 6.389534482634365e-06, "loss": 0.9521, "step": 9451 }, { "epoch": 0.33, "learning_rate": 6.387524215005781e-06, "loss": 0.9627, "step": 9452 }, { "epoch": 0.33, "learning_rate": 6.385514115260212e-06, "loss": 0.9695, "step": 9453 }, { "epoch": 0.33, "learning_rate": 6.383504183491067e-06, "loss": 0.9726, "step": 9454 }, { "epoch": 0.33, "learning_rate": 6.381494419791765e-06, "loss": 0.9795, "step": 9455 }, { "epoch": 0.33, "learning_rate": 6.379484824255697e-06, "loss": 0.9199, "step": 9456 }, { "epoch": 0.33, "learning_rate": 6.377475396976263e-06, "loss": 0.9591, "step": 9457 }, { "epoch": 0.33, "learning_rate": 6.37546613804684e-06, "loss": 0.9711, "step": 9458 }, { "epoch": 0.33, "learning_rate": 6.37345704756081e-06, "loss": 0.9544, "step": 9459 }, { "epoch": 0.33, "learning_rate": 6.371448125611535e-06, "loss": 0.935, "step": 9460 }, { "epoch": 0.33, "learning_rate": 6.369439372292388e-06, "loss": 0.9247, "step": 9461 }, { "epoch": 0.33, "learning_rate": 6.367430787696711e-06, "loss": 0.952, "step": 9462 }, { "epoch": 0.33, "learning_rate": 6.36542237191785e-06, "loss": 0.9643, "step": 9463 }, { "epoch": 0.33, "learning_rate": 6.36341412504915e-06, "loss": 0.969, "step": 9464 }, { "epoch": 0.33, "learning_rate": 6.36140604718393e-06, "loss": 0.9664, "step": 9465 }, { "epoch": 0.33, "learning_rate": 6.35939813841552e-06, "loss": 0.9787, "step": 9466 }, { "epoch": 0.33, "learning_rate": 6.357390398837231e-06, "loss": 0.9412, "step": 9467 }, { "epoch": 0.33, "learning_rate": 6.355382828542369e-06, "loss": 0.9595, "step": 9468 }, { "epoch": 0.33, "learning_rate": 6.3533754276242336e-06, "loss": 0.947, "step": 9469 }, { "epoch": 0.33, "learning_rate": 6.3513681961761084e-06, "loss": 0.9376, "step": 9470 }, { "epoch": 0.33, "learning_rate": 6.349361134291286e-06, "loss": 0.9823, "step": 9471 }, { "epoch": 0.33, "learning_rate": 6.3473542420630305e-06, "loss": 0.945, "step": 9472 }, { "epoch": 0.33, "learning_rate": 6.3453475195846155e-06, "loss": 0.955, "step": 9473 }, { "epoch": 0.33, "learning_rate": 6.343340966949294e-06, "loss": 0.9811, "step": 9474 }, { "epoch": 0.33, "learning_rate": 6.341334584250325e-06, "loss": 0.967, "step": 9475 }, { "epoch": 0.33, "learning_rate": 6.339328371580949e-06, "loss": 0.94, "step": 9476 }, { "epoch": 0.33, "learning_rate": 6.337322329034391e-06, "loss": 0.952, "step": 9477 }, { "epoch": 0.33, "learning_rate": 6.335316456703891e-06, "loss": 0.9479, "step": 9478 }, { "epoch": 0.33, "learning_rate": 6.333310754682659e-06, "loss": 0.9463, "step": 9479 }, { "epoch": 0.33, "learning_rate": 6.331305223063913e-06, "loss": 0.9701, "step": 9480 }, { "epoch": 0.33, "learning_rate": 6.329299861940852e-06, "loss": 0.9535, "step": 9481 }, { "epoch": 0.33, "learning_rate": 6.327294671406673e-06, "loss": 0.973, "step": 9482 }, { "epoch": 0.33, "learning_rate": 6.32528965155456e-06, "loss": 0.9377, "step": 9483 }, { "epoch": 0.33, "learning_rate": 6.323284802477702e-06, "loss": 0.9671, "step": 9484 }, { "epoch": 0.33, "learning_rate": 6.321280124269261e-06, "loss": 0.9564, "step": 9485 }, { "epoch": 0.33, "learning_rate": 6.3192756170224e-06, "loss": 0.9235, "step": 9486 }, { "epoch": 0.33, "learning_rate": 6.317271280830284e-06, "loss": 0.9501, "step": 9487 }, { "epoch": 0.33, "learning_rate": 6.315267115786052e-06, "loss": 0.915, "step": 9488 }, { "epoch": 0.33, "learning_rate": 6.31326312198285e-06, "loss": 0.9807, "step": 9489 }, { "epoch": 0.33, "learning_rate": 6.311259299513804e-06, "loss": 0.9206, "step": 9490 }, { "epoch": 0.33, "learning_rate": 6.309255648472044e-06, "loss": 0.9554, "step": 9491 }, { "epoch": 0.33, "learning_rate": 6.307252168950682e-06, "loss": 0.9124, "step": 9492 }, { "epoch": 0.33, "learning_rate": 6.3052488610428255e-06, "loss": 0.9539, "step": 9493 }, { "epoch": 0.33, "learning_rate": 6.303245724841577e-06, "loss": 0.9519, "step": 9494 }, { "epoch": 0.33, "learning_rate": 6.301242760440023e-06, "loss": 0.9625, "step": 9495 }, { "epoch": 0.33, "learning_rate": 6.299239967931258e-06, "loss": 0.9688, "step": 9496 }, { "epoch": 0.33, "learning_rate": 6.297237347408345e-06, "loss": 0.934, "step": 9497 }, { "epoch": 0.33, "learning_rate": 6.295234898964363e-06, "loss": 0.9523, "step": 9498 }, { "epoch": 0.33, "learning_rate": 6.29323262269237e-06, "loss": 0.9451, "step": 9499 }, { "epoch": 0.33, "learning_rate": 6.29123051868541e-06, "loss": 0.9656, "step": 9500 }, { "epoch": 0.33, "learning_rate": 6.289228587036535e-06, "loss": 0.9091, "step": 9501 }, { "epoch": 0.33, "learning_rate": 6.287226827838779e-06, "loss": 0.9663, "step": 9502 }, { "epoch": 0.33, "learning_rate": 6.285225241185171e-06, "loss": 0.9885, "step": 9503 }, { "epoch": 0.33, "learning_rate": 6.283223827168728e-06, "loss": 0.9436, "step": 9504 }, { "epoch": 0.33, "learning_rate": 6.2812225858824645e-06, "loss": 0.9314, "step": 9505 }, { "epoch": 0.33, "learning_rate": 6.27922151741938e-06, "loss": 0.9228, "step": 9506 }, { "epoch": 0.33, "learning_rate": 6.27722062187248e-06, "loss": 0.9112, "step": 9507 }, { "epoch": 0.33, "learning_rate": 6.275219899334746e-06, "loss": 0.9771, "step": 9508 }, { "epoch": 0.33, "learning_rate": 6.273219349899152e-06, "loss": 0.9788, "step": 9509 }, { "epoch": 0.33, "learning_rate": 6.271218973658681e-06, "loss": 0.977, "step": 9510 }, { "epoch": 0.33, "learning_rate": 6.269218770706289e-06, "loss": 0.921, "step": 9511 }, { "epoch": 0.33, "learning_rate": 6.2672187411349374e-06, "loss": 0.897, "step": 9512 }, { "epoch": 0.33, "learning_rate": 6.265218885037567e-06, "loss": 0.9658, "step": 9513 }, { "epoch": 0.33, "learning_rate": 6.263219202507125e-06, "loss": 0.974, "step": 9514 }, { "epoch": 0.33, "learning_rate": 6.261219693636538e-06, "loss": 0.9577, "step": 9515 }, { "epoch": 0.33, "learning_rate": 6.259220358518727e-06, "loss": 0.8861, "step": 9516 }, { "epoch": 0.33, "learning_rate": 6.257221197246614e-06, "loss": 0.984, "step": 9517 }, { "epoch": 0.33, "learning_rate": 6.255222209913099e-06, "loss": 0.9262, "step": 9518 }, { "epoch": 0.33, "learning_rate": 6.253223396611092e-06, "loss": 0.9524, "step": 9519 }, { "epoch": 0.33, "learning_rate": 6.251224757433469e-06, "loss": 0.9504, "step": 9520 }, { "epoch": 0.33, "learning_rate": 6.2492262924731275e-06, "loss": 0.9482, "step": 9521 }, { "epoch": 0.33, "learning_rate": 6.247228001822937e-06, "loss": 0.932, "step": 9522 }, { "epoch": 0.33, "learning_rate": 6.245229885575758e-06, "loss": 0.967, "step": 9523 }, { "epoch": 0.33, "learning_rate": 6.243231943824457e-06, "loss": 0.9556, "step": 9524 }, { "epoch": 0.33, "learning_rate": 6.241234176661881e-06, "loss": 0.9029, "step": 9525 }, { "epoch": 0.33, "learning_rate": 6.239236584180876e-06, "loss": 0.9384, "step": 9526 }, { "epoch": 0.33, "learning_rate": 6.237239166474273e-06, "loss": 0.9152, "step": 9527 }, { "epoch": 0.33, "learning_rate": 6.2352419236349e-06, "loss": 0.9295, "step": 9528 }, { "epoch": 0.33, "learning_rate": 6.233244855755572e-06, "loss": 0.9574, "step": 9529 }, { "epoch": 0.33, "learning_rate": 6.231247962929104e-06, "loss": 0.8966, "step": 9530 }, { "epoch": 0.33, "learning_rate": 6.229251245248296e-06, "loss": 0.9612, "step": 9531 }, { "epoch": 0.33, "learning_rate": 6.227254702805936e-06, "loss": 0.9697, "step": 9532 }, { "epoch": 0.33, "learning_rate": 6.225258335694818e-06, "loss": 0.9703, "step": 9533 }, { "epoch": 0.33, "learning_rate": 6.223262144007713e-06, "loss": 0.9197, "step": 9534 }, { "epoch": 0.33, "learning_rate": 6.221266127837398e-06, "loss": 0.9617, "step": 9535 }, { "epoch": 0.33, "learning_rate": 6.219270287276624e-06, "loss": 0.9309, "step": 9536 }, { "epoch": 0.33, "learning_rate": 6.2172746224181524e-06, "loss": 0.9475, "step": 9537 }, { "epoch": 0.33, "learning_rate": 6.2152791333547256e-06, "loss": 0.9376, "step": 9538 }, { "epoch": 0.33, "learning_rate": 6.213283820179076e-06, "loss": 0.9783, "step": 9539 }, { "epoch": 0.33, "learning_rate": 6.211288682983938e-06, "loss": 0.9698, "step": 9540 }, { "epoch": 0.33, "learning_rate": 6.209293721862024e-06, "loss": 0.9633, "step": 9541 }, { "epoch": 0.33, "learning_rate": 6.207298936906059e-06, "loss": 0.9495, "step": 9542 }, { "epoch": 0.33, "learning_rate": 6.205304328208731e-06, "loss": 0.9669, "step": 9543 }, { "epoch": 0.33, "learning_rate": 6.203309895862748e-06, "loss": 0.9609, "step": 9544 }, { "epoch": 0.33, "learning_rate": 6.201315639960796e-06, "loss": 0.9512, "step": 9545 }, { "epoch": 0.34, "learning_rate": 6.199321560595546e-06, "loss": 0.9402, "step": 9546 }, { "epoch": 0.34, "learning_rate": 6.197327657859677e-06, "loss": 0.9248, "step": 9547 }, { "epoch": 0.34, "learning_rate": 6.195333931845851e-06, "loss": 0.9622, "step": 9548 }, { "epoch": 0.34, "learning_rate": 6.19334038264672e-06, "loss": 0.9493, "step": 9549 }, { "epoch": 0.34, "learning_rate": 6.19134701035493e-06, "loss": 0.9304, "step": 9550 }, { "epoch": 0.34, "learning_rate": 6.189353815063124e-06, "loss": 0.9617, "step": 9551 }, { "epoch": 0.34, "learning_rate": 6.1873607968639236e-06, "loss": 0.9197, "step": 9552 }, { "epoch": 0.34, "learning_rate": 6.185367955849961e-06, "loss": 0.9589, "step": 9553 }, { "epoch": 0.34, "learning_rate": 6.183375292113844e-06, "loss": 0.9197, "step": 9554 }, { "epoch": 0.34, "learning_rate": 6.181382805748175e-06, "loss": 0.9405, "step": 9555 }, { "epoch": 0.34, "learning_rate": 6.179390496845557e-06, "loss": 0.9045, "step": 9556 }, { "epoch": 0.34, "learning_rate": 6.177398365498576e-06, "loss": 0.9685, "step": 9557 }, { "epoch": 0.34, "learning_rate": 6.175406411799813e-06, "loss": 0.9527, "step": 9558 }, { "epoch": 0.34, "learning_rate": 6.173414635841839e-06, "loss": 0.9428, "step": 9559 }, { "epoch": 0.34, "learning_rate": 6.171423037717221e-06, "loss": 0.9373, "step": 9560 }, { "epoch": 0.34, "learning_rate": 6.169431617518514e-06, "loss": 0.9007, "step": 9561 }, { "epoch": 0.34, "learning_rate": 6.167440375338262e-06, "loss": 0.9162, "step": 9562 }, { "epoch": 0.34, "learning_rate": 6.1654493112690085e-06, "loss": 0.9457, "step": 9563 }, { "epoch": 0.34, "learning_rate": 6.163458425403279e-06, "loss": 0.8876, "step": 9564 }, { "epoch": 0.34, "learning_rate": 6.161467717833605e-06, "loss": 0.9206, "step": 9565 }, { "epoch": 0.34, "learning_rate": 6.15947718865249e-06, "loss": 0.9405, "step": 9566 }, { "epoch": 0.34, "learning_rate": 6.15748683795245e-06, "loss": 0.9317, "step": 9567 }, { "epoch": 0.34, "learning_rate": 6.155496665825978e-06, "loss": 0.9428, "step": 9568 }, { "epoch": 0.34, "learning_rate": 6.1535066723655634e-06, "loss": 0.9503, "step": 9569 }, { "epoch": 0.34, "learning_rate": 6.15151685766369e-06, "loss": 0.9657, "step": 9570 }, { "epoch": 0.34, "learning_rate": 6.149527221812825e-06, "loss": 0.9186, "step": 9571 }, { "epoch": 0.34, "learning_rate": 6.147537764905441e-06, "loss": 0.9426, "step": 9572 }, { "epoch": 0.34, "learning_rate": 6.145548487033987e-06, "loss": 0.9762, "step": 9573 }, { "epoch": 0.34, "learning_rate": 6.143559388290916e-06, "loss": 0.9252, "step": 9574 }, { "epoch": 0.34, "learning_rate": 6.141570468768663e-06, "loss": 0.983, "step": 9575 }, { "epoch": 0.34, "learning_rate": 6.139581728559665e-06, "loss": 0.9, "step": 9576 }, { "epoch": 0.34, "learning_rate": 6.1375931677563435e-06, "loss": 0.988, "step": 9577 }, { "epoch": 0.34, "learning_rate": 6.1356047864511056e-06, "loss": 0.9375, "step": 9578 }, { "epoch": 0.34, "learning_rate": 6.133616584736368e-06, "loss": 0.9326, "step": 9579 }, { "epoch": 0.34, "learning_rate": 6.131628562704521e-06, "loss": 0.9433, "step": 9580 }, { "epoch": 0.34, "learning_rate": 6.12964072044796e-06, "loss": 0.9659, "step": 9581 }, { "epoch": 0.34, "learning_rate": 6.127653058059062e-06, "loss": 0.958, "step": 9582 }, { "epoch": 0.34, "learning_rate": 6.125665575630204e-06, "loss": 0.9627, "step": 9583 }, { "epoch": 0.34, "learning_rate": 6.123678273253746e-06, "loss": 0.9427, "step": 9584 }, { "epoch": 0.34, "learning_rate": 6.121691151022046e-06, "loss": 0.9817, "step": 9585 }, { "epoch": 0.34, "learning_rate": 6.119704209027453e-06, "loss": 0.9352, "step": 9586 }, { "epoch": 0.34, "learning_rate": 6.1177174473623e-06, "loss": 0.9298, "step": 9587 }, { "epoch": 0.34, "learning_rate": 6.115730866118931e-06, "loss": 0.9438, "step": 9588 }, { "epoch": 0.34, "learning_rate": 6.113744465389654e-06, "loss": 0.909, "step": 9589 }, { "epoch": 0.34, "learning_rate": 6.111758245266795e-06, "loss": 0.9211, "step": 9590 }, { "epoch": 0.34, "learning_rate": 6.109772205842653e-06, "loss": 0.9236, "step": 9591 }, { "epoch": 0.34, "learning_rate": 6.107786347209528e-06, "loss": 0.9799, "step": 9592 }, { "epoch": 0.34, "learning_rate": 6.105800669459708e-06, "loss": 0.9597, "step": 9593 }, { "epoch": 0.34, "learning_rate": 6.103815172685474e-06, "loss": 0.9583, "step": 9594 }, { "epoch": 0.34, "learning_rate": 6.101829856979098e-06, "loss": 0.9709, "step": 9595 }, { "epoch": 0.34, "learning_rate": 6.099844722432844e-06, "loss": 0.9344, "step": 9596 }, { "epoch": 0.34, "learning_rate": 6.09785976913897e-06, "loss": 0.9501, "step": 9597 }, { "epoch": 0.34, "learning_rate": 6.095874997189714e-06, "loss": 0.9218, "step": 9598 }, { "epoch": 0.34, "learning_rate": 6.093890406677326e-06, "loss": 0.9183, "step": 9599 }, { "epoch": 0.34, "learning_rate": 6.091905997694034e-06, "loss": 0.9431, "step": 9600 }, { "epoch": 0.34, "learning_rate": 6.08992177033205e-06, "loss": 0.9366, "step": 9601 }, { "epoch": 0.34, "learning_rate": 6.0879377246835994e-06, "loss": 0.9216, "step": 9602 }, { "epoch": 0.34, "learning_rate": 6.085953860840877e-06, "loss": 0.9584, "step": 9603 }, { "epoch": 0.34, "learning_rate": 6.083970178896088e-06, "loss": 0.9435, "step": 9604 }, { "epoch": 0.34, "learning_rate": 6.0819866789414124e-06, "loss": 0.9286, "step": 9605 }, { "epoch": 0.34, "learning_rate": 6.080003361069037e-06, "loss": 0.9463, "step": 9606 }, { "epoch": 0.34, "learning_rate": 6.078020225371128e-06, "loss": 0.9555, "step": 9607 }, { "epoch": 0.34, "learning_rate": 6.0760372719398476e-06, "loss": 0.9534, "step": 9608 }, { "epoch": 0.34, "learning_rate": 6.074054500867352e-06, "loss": 0.946, "step": 9609 }, { "epoch": 0.34, "learning_rate": 6.072071912245783e-06, "loss": 0.9613, "step": 9610 }, { "epoch": 0.34, "learning_rate": 6.070089506167285e-06, "loss": 0.94, "step": 9611 }, { "epoch": 0.34, "learning_rate": 6.068107282723977e-06, "loss": 0.937, "step": 9612 }, { "epoch": 0.34, "learning_rate": 6.066125242007989e-06, "loss": 0.9672, "step": 9613 }, { "epoch": 0.34, "learning_rate": 6.064143384111427e-06, "loss": 0.9507, "step": 9614 }, { "epoch": 0.34, "learning_rate": 6.062161709126394e-06, "loss": 0.9406, "step": 9615 }, { "epoch": 0.34, "learning_rate": 6.060180217144986e-06, "loss": 0.9025, "step": 9616 }, { "epoch": 0.34, "learning_rate": 6.058198908259288e-06, "loss": 0.9432, "step": 9617 }, { "epoch": 0.34, "learning_rate": 6.05621778256138e-06, "loss": 0.9604, "step": 9618 }, { "epoch": 0.34, "learning_rate": 6.054236840143328e-06, "loss": 0.9543, "step": 9619 }, { "epoch": 0.34, "learning_rate": 6.052256081097195e-06, "loss": 0.9679, "step": 9620 }, { "epoch": 0.34, "learning_rate": 6.050275505515029e-06, "loss": 0.9292, "step": 9621 }, { "epoch": 0.34, "learning_rate": 6.04829511348888e-06, "loss": 0.9318, "step": 9622 }, { "epoch": 0.34, "learning_rate": 6.046314905110782e-06, "loss": 0.9526, "step": 9623 }, { "epoch": 0.34, "learning_rate": 6.044334880472753e-06, "loss": 0.9479, "step": 9624 }, { "epoch": 0.34, "learning_rate": 6.042355039666821e-06, "loss": 0.9387, "step": 9625 }, { "epoch": 0.34, "learning_rate": 6.040375382784988e-06, "loss": 0.9379, "step": 9626 }, { "epoch": 0.34, "learning_rate": 6.038395909919261e-06, "loss": 0.962, "step": 9627 }, { "epoch": 0.34, "learning_rate": 6.036416621161627e-06, "loss": 0.9374, "step": 9628 }, { "epoch": 0.34, "learning_rate": 6.034437516604075e-06, "loss": 0.9465, "step": 9629 }, { "epoch": 0.34, "learning_rate": 6.032458596338577e-06, "loss": 0.9575, "step": 9630 }, { "epoch": 0.34, "learning_rate": 6.030479860457098e-06, "loss": 0.9776, "step": 9631 }, { "epoch": 0.34, "learning_rate": 6.0285013090516e-06, "loss": 0.9539, "step": 9632 }, { "epoch": 0.34, "learning_rate": 6.026522942214026e-06, "loss": 1.0068, "step": 9633 }, { "epoch": 0.34, "learning_rate": 6.024544760036327e-06, "loss": 0.931, "step": 9634 }, { "epoch": 0.34, "learning_rate": 6.0225667626104245e-06, "loss": 0.9953, "step": 9635 }, { "epoch": 0.34, "learning_rate": 6.02058895002825e-06, "loss": 0.9444, "step": 9636 }, { "epoch": 0.34, "learning_rate": 6.018611322381717e-06, "loss": 0.9667, "step": 9637 }, { "epoch": 0.34, "learning_rate": 6.016633879762728e-06, "loss": 0.9426, "step": 9638 }, { "epoch": 0.34, "learning_rate": 6.0146566222631864e-06, "loss": 0.917, "step": 9639 }, { "epoch": 0.34, "learning_rate": 6.012679549974976e-06, "loss": 0.965, "step": 9640 }, { "epoch": 0.34, "learning_rate": 6.010702662989984e-06, "loss": 0.9784, "step": 9641 }, { "epoch": 0.34, "learning_rate": 6.008725961400076e-06, "loss": 0.9396, "step": 9642 }, { "epoch": 0.34, "learning_rate": 6.00674944529712e-06, "loss": 0.979, "step": 9643 }, { "epoch": 0.34, "learning_rate": 6.004773114772966e-06, "loss": 0.9589, "step": 9644 }, { "epoch": 0.34, "learning_rate": 6.002796969919468e-06, "loss": 0.974, "step": 9645 }, { "epoch": 0.34, "learning_rate": 6.00082101082846e-06, "loss": 0.9929, "step": 9646 }, { "epoch": 0.34, "learning_rate": 5.998845237591765e-06, "loss": 0.9871, "step": 9647 }, { "epoch": 0.34, "learning_rate": 5.9968696503012125e-06, "loss": 0.9759, "step": 9648 }, { "epoch": 0.34, "learning_rate": 5.9948942490486084e-06, "loss": 0.9588, "step": 9649 }, { "epoch": 0.34, "learning_rate": 5.992919033925759e-06, "loss": 0.9501, "step": 9650 }, { "epoch": 0.34, "learning_rate": 5.990944005024455e-06, "loss": 0.9816, "step": 9651 }, { "epoch": 0.34, "learning_rate": 5.9889691624364866e-06, "loss": 0.9487, "step": 9652 }, { "epoch": 0.34, "learning_rate": 5.986994506253629e-06, "loss": 0.9667, "step": 9653 }, { "epoch": 0.34, "learning_rate": 5.985020036567648e-06, "loss": 0.962, "step": 9654 }, { "epoch": 0.34, "learning_rate": 5.983045753470308e-06, "loss": 0.9498, "step": 9655 }, { "epoch": 0.34, "learning_rate": 5.981071657053355e-06, "loss": 0.9247, "step": 9656 }, { "epoch": 0.34, "learning_rate": 5.979097747408538e-06, "loss": 0.9243, "step": 9657 }, { "epoch": 0.34, "learning_rate": 5.977124024627581e-06, "loss": 0.9295, "step": 9658 }, { "epoch": 0.34, "learning_rate": 5.975150488802221e-06, "loss": 0.9406, "step": 9659 }, { "epoch": 0.34, "learning_rate": 5.973177140024167e-06, "loss": 0.9476, "step": 9660 }, { "epoch": 0.34, "learning_rate": 5.971203978385125e-06, "loss": 0.9508, "step": 9661 }, { "epoch": 0.34, "learning_rate": 5.969231003976801e-06, "loss": 0.9414, "step": 9662 }, { "epoch": 0.34, "learning_rate": 5.967258216890878e-06, "loss": 0.9847, "step": 9663 }, { "epoch": 0.34, "learning_rate": 5.965285617219043e-06, "loss": 0.9739, "step": 9664 }, { "epoch": 0.34, "learning_rate": 5.9633132050529655e-06, "loss": 0.9447, "step": 9665 }, { "epoch": 0.34, "learning_rate": 5.961340980484311e-06, "loss": 0.9578, "step": 9666 }, { "epoch": 0.34, "learning_rate": 5.959368943604733e-06, "loss": 0.9523, "step": 9667 }, { "epoch": 0.34, "learning_rate": 5.957397094505883e-06, "loss": 0.9282, "step": 9668 }, { "epoch": 0.34, "learning_rate": 5.955425433279398e-06, "loss": 0.954, "step": 9669 }, { "epoch": 0.34, "learning_rate": 5.9534539600169e-06, "loss": 0.9801, "step": 9670 }, { "epoch": 0.34, "learning_rate": 5.951482674810018e-06, "loss": 0.9507, "step": 9671 }, { "epoch": 0.34, "learning_rate": 5.949511577750357e-06, "loss": 0.9225, "step": 9672 }, { "epoch": 0.34, "learning_rate": 5.947540668929528e-06, "loss": 0.9621, "step": 9673 }, { "epoch": 0.34, "learning_rate": 5.945569948439117e-06, "loss": 0.961, "step": 9674 }, { "epoch": 0.34, "learning_rate": 5.9435994163707156e-06, "loss": 0.9625, "step": 9675 }, { "epoch": 0.34, "learning_rate": 5.9416290728159e-06, "loss": 0.9396, "step": 9676 }, { "epoch": 0.34, "learning_rate": 5.939658917866233e-06, "loss": 0.9398, "step": 9677 }, { "epoch": 0.34, "learning_rate": 5.937688951613278e-06, "loss": 0.9458, "step": 9678 }, { "epoch": 0.34, "learning_rate": 5.935719174148583e-06, "loss": 0.9701, "step": 9679 }, { "epoch": 0.34, "learning_rate": 5.933749585563695e-06, "loss": 0.9261, "step": 9680 }, { "epoch": 0.34, "learning_rate": 5.93178018595014e-06, "loss": 0.9199, "step": 9681 }, { "epoch": 0.34, "learning_rate": 5.929810975399449e-06, "loss": 0.9424, "step": 9682 }, { "epoch": 0.34, "learning_rate": 5.927841954003134e-06, "loss": 0.9893, "step": 9683 }, { "epoch": 0.34, "learning_rate": 5.9258731218526984e-06, "loss": 0.9536, "step": 9684 }, { "epoch": 0.34, "learning_rate": 5.923904479039646e-06, "loss": 0.9628, "step": 9685 }, { "epoch": 0.34, "learning_rate": 5.921936025655462e-06, "loss": 0.9368, "step": 9686 }, { "epoch": 0.34, "learning_rate": 5.919967761791627e-06, "loss": 0.9187, "step": 9687 }, { "epoch": 0.34, "learning_rate": 5.917999687539611e-06, "loss": 0.923, "step": 9688 }, { "epoch": 0.34, "learning_rate": 5.916031802990883e-06, "loss": 0.9374, "step": 9689 }, { "epoch": 0.34, "learning_rate": 5.914064108236885e-06, "loss": 0.939, "step": 9690 }, { "epoch": 0.34, "learning_rate": 5.9120966033690745e-06, "loss": 0.9289, "step": 9691 }, { "epoch": 0.34, "learning_rate": 5.910129288478883e-06, "loss": 0.9772, "step": 9692 }, { "epoch": 0.34, "learning_rate": 5.908162163657731e-06, "loss": 0.9635, "step": 9693 }, { "epoch": 0.34, "learning_rate": 5.906195228997047e-06, "loss": 0.9537, "step": 9694 }, { "epoch": 0.34, "learning_rate": 5.9042284845882345e-06, "loss": 0.9405, "step": 9695 }, { "epoch": 0.35, "learning_rate": 5.902261930522698e-06, "loss": 0.948, "step": 9696 }, { "epoch": 0.35, "learning_rate": 5.900295566891825e-06, "loss": 0.968, "step": 9697 }, { "epoch": 0.35, "learning_rate": 5.898329393787003e-06, "loss": 0.934, "step": 9698 }, { "epoch": 0.35, "learning_rate": 5.896363411299606e-06, "loss": 0.9738, "step": 9699 }, { "epoch": 0.35, "learning_rate": 5.894397619520993e-06, "loss": 0.9543, "step": 9700 }, { "epoch": 0.35, "learning_rate": 5.892432018542528e-06, "loss": 0.9653, "step": 9701 }, { "epoch": 0.35, "learning_rate": 5.890466608455552e-06, "loss": 0.9547, "step": 9702 }, { "epoch": 0.35, "learning_rate": 5.8885013893514106e-06, "loss": 0.9355, "step": 9703 }, { "epoch": 0.35, "learning_rate": 5.886536361321429e-06, "loss": 0.974, "step": 9704 }, { "epoch": 0.35, "learning_rate": 5.884571524456931e-06, "loss": 0.9415, "step": 9705 }, { "epoch": 0.35, "learning_rate": 5.8826068788492284e-06, "loss": 0.9018, "step": 9706 }, { "epoch": 0.35, "learning_rate": 5.88064242458962e-06, "loss": 0.9566, "step": 9707 }, { "epoch": 0.35, "learning_rate": 5.878678161769406e-06, "loss": 0.9273, "step": 9708 }, { "epoch": 0.35, "learning_rate": 5.876714090479867e-06, "loss": 0.9401, "step": 9709 }, { "epoch": 0.35, "learning_rate": 5.874750210812283e-06, "loss": 0.9578, "step": 9710 }, { "epoch": 0.35, "learning_rate": 5.872786522857918e-06, "loss": 1.003, "step": 9711 }, { "epoch": 0.35, "learning_rate": 5.870823026708038e-06, "loss": 0.9344, "step": 9712 }, { "epoch": 0.35, "learning_rate": 5.868859722453883e-06, "loss": 0.9764, "step": 9713 }, { "epoch": 0.35, "learning_rate": 5.866896610186701e-06, "loss": 0.9565, "step": 9714 }, { "epoch": 0.35, "learning_rate": 5.864933689997725e-06, "loss": 0.9828, "step": 9715 }, { "epoch": 0.35, "learning_rate": 5.862970961978167e-06, "loss": 0.9579, "step": 9716 }, { "epoch": 0.35, "learning_rate": 5.861008426219255e-06, "loss": 0.9446, "step": 9717 }, { "epoch": 0.35, "learning_rate": 5.8590460828121845e-06, "loss": 0.9464, "step": 9718 }, { "epoch": 0.35, "learning_rate": 5.8570839318481574e-06, "loss": 0.9305, "step": 9719 }, { "epoch": 0.35, "learning_rate": 5.855121973418356e-06, "loss": 0.9443, "step": 9720 }, { "epoch": 0.35, "learning_rate": 5.8531602076139645e-06, "loss": 1.0109, "step": 9721 }, { "epoch": 0.35, "learning_rate": 5.85119863452615e-06, "loss": 0.9588, "step": 9722 }, { "epoch": 0.35, "learning_rate": 5.849237254246067e-06, "loss": 0.9344, "step": 9723 }, { "epoch": 0.35, "learning_rate": 5.847276066864877e-06, "loss": 0.9484, "step": 9724 }, { "epoch": 0.35, "learning_rate": 5.845315072473712e-06, "loss": 0.9419, "step": 9725 }, { "epoch": 0.35, "learning_rate": 5.843354271163715e-06, "loss": 0.9103, "step": 9726 }, { "epoch": 0.35, "learning_rate": 5.841393663026005e-06, "loss": 0.9849, "step": 9727 }, { "epoch": 0.35, "learning_rate": 5.8394332481517e-06, "loss": 0.959, "step": 9728 }, { "epoch": 0.35, "learning_rate": 5.837473026631909e-06, "loss": 0.9176, "step": 9729 }, { "epoch": 0.35, "learning_rate": 5.835512998557717e-06, "loss": 0.9525, "step": 9730 }, { "epoch": 0.35, "learning_rate": 5.8335531640202325e-06, "loss": 1.0088, "step": 9731 }, { "epoch": 0.35, "learning_rate": 5.831593523110519e-06, "loss": 0.9266, "step": 9732 }, { "epoch": 0.35, "learning_rate": 5.8296340759196526e-06, "loss": 0.9296, "step": 9733 }, { "epoch": 0.35, "learning_rate": 5.827674822538696e-06, "loss": 0.8964, "step": 9734 }, { "epoch": 0.35, "learning_rate": 5.825715763058703e-06, "loss": 0.9728, "step": 9735 }, { "epoch": 0.35, "learning_rate": 5.823756897570711e-06, "loss": 0.9578, "step": 9736 }, { "epoch": 0.35, "learning_rate": 5.82179822616576e-06, "loss": 0.9653, "step": 9737 }, { "epoch": 0.35, "learning_rate": 5.819839748934879e-06, "loss": 0.9363, "step": 9738 }, { "epoch": 0.35, "learning_rate": 5.817881465969074e-06, "loss": 0.9676, "step": 9739 }, { "epoch": 0.35, "learning_rate": 5.815923377359358e-06, "loss": 0.9425, "step": 9740 }, { "epoch": 0.35, "learning_rate": 5.813965483196732e-06, "loss": 0.9531, "step": 9741 }, { "epoch": 0.35, "learning_rate": 5.8120077835721855e-06, "loss": 0.9736, "step": 9742 }, { "epoch": 0.35, "learning_rate": 5.810050278576689e-06, "loss": 0.9788, "step": 9743 }, { "epoch": 0.35, "learning_rate": 5.808092968301229e-06, "loss": 0.9274, "step": 9744 }, { "epoch": 0.35, "learning_rate": 5.806135852836756e-06, "loss": 0.9601, "step": 9745 }, { "epoch": 0.35, "learning_rate": 5.804178932274229e-06, "loss": 0.9701, "step": 9746 }, { "epoch": 0.35, "learning_rate": 5.802222206704593e-06, "loss": 0.9911, "step": 9747 }, { "epoch": 0.35, "learning_rate": 5.800265676218778e-06, "loss": 0.9279, "step": 9748 }, { "epoch": 0.35, "learning_rate": 5.798309340907713e-06, "loss": 0.9624, "step": 9749 }, { "epoch": 0.35, "learning_rate": 5.796353200862315e-06, "loss": 0.9627, "step": 9750 }, { "epoch": 0.35, "learning_rate": 5.794397256173494e-06, "loss": 0.9782, "step": 9751 }, { "epoch": 0.35, "learning_rate": 5.792441506932145e-06, "loss": 0.9107, "step": 9752 }, { "epoch": 0.35, "learning_rate": 5.790485953229158e-06, "loss": 0.9563, "step": 9753 }, { "epoch": 0.35, "learning_rate": 5.788530595155419e-06, "loss": 0.977, "step": 9754 }, { "epoch": 0.35, "learning_rate": 5.786575432801788e-06, "loss": 0.9506, "step": 9755 }, { "epoch": 0.35, "learning_rate": 5.7846204662591455e-06, "loss": 0.9593, "step": 9756 }, { "epoch": 0.35, "learning_rate": 5.7826656956183284e-06, "loss": 0.9882, "step": 9757 }, { "epoch": 0.35, "learning_rate": 5.780711120970187e-06, "loss": 0.957, "step": 9758 }, { "epoch": 0.35, "learning_rate": 5.778756742405558e-06, "loss": 0.9518, "step": 9759 }, { "epoch": 0.35, "learning_rate": 5.776802560015269e-06, "loss": 0.9723, "step": 9760 }, { "epoch": 0.35, "learning_rate": 5.774848573890131e-06, "loss": 0.944, "step": 9761 }, { "epoch": 0.35, "learning_rate": 5.7728947841209546e-06, "loss": 0.9565, "step": 9762 }, { "epoch": 0.35, "learning_rate": 5.770941190798543e-06, "loss": 0.9284, "step": 9763 }, { "epoch": 0.35, "learning_rate": 5.768987794013672e-06, "loss": 0.9144, "step": 9764 }, { "epoch": 0.35, "learning_rate": 5.767034593857142e-06, "loss": 0.9951, "step": 9765 }, { "epoch": 0.35, "learning_rate": 5.765081590419708e-06, "loss": 0.9723, "step": 9766 }, { "epoch": 0.35, "learning_rate": 5.76312878379214e-06, "loss": 0.9383, "step": 9767 }, { "epoch": 0.35, "learning_rate": 5.761176174065192e-06, "loss": 0.9674, "step": 9768 }, { "epoch": 0.35, "learning_rate": 5.759223761329601e-06, "loss": 0.9453, "step": 9769 }, { "epoch": 0.35, "learning_rate": 5.757271545676106e-06, "loss": 0.9225, "step": 9770 }, { "epoch": 0.35, "learning_rate": 5.755319527195431e-06, "loss": 0.9556, "step": 9771 }, { "epoch": 0.35, "learning_rate": 5.753367705978299e-06, "loss": 0.9512, "step": 9772 }, { "epoch": 0.35, "learning_rate": 5.7514160821154085e-06, "loss": 0.9565, "step": 9773 }, { "epoch": 0.35, "learning_rate": 5.74946465569746e-06, "loss": 1.0083, "step": 9774 }, { "epoch": 0.35, "learning_rate": 5.747513426815147e-06, "loss": 0.9778, "step": 9775 }, { "epoch": 0.35, "learning_rate": 5.745562395559138e-06, "loss": 0.9331, "step": 9776 }, { "epoch": 0.35, "learning_rate": 5.7436115620201185e-06, "loss": 0.9995, "step": 9777 }, { "epoch": 0.35, "learning_rate": 5.741660926288738e-06, "loss": 0.9109, "step": 9778 }, { "epoch": 0.35, "learning_rate": 5.739710488455654e-06, "loss": 0.965, "step": 9779 }, { "epoch": 0.35, "learning_rate": 5.737760248611507e-06, "loss": 0.9369, "step": 9780 }, { "epoch": 0.35, "learning_rate": 5.7358102068469366e-06, "loss": 0.9573, "step": 9781 }, { "epoch": 0.35, "learning_rate": 5.733860363252559e-06, "loss": 0.9278, "step": 9782 }, { "epoch": 0.35, "learning_rate": 5.731910717918993e-06, "loss": 0.9948, "step": 9783 }, { "epoch": 0.35, "learning_rate": 5.72996127093685e-06, "loss": 0.9377, "step": 9784 }, { "epoch": 0.35, "learning_rate": 5.728012022396716e-06, "loss": 0.9458, "step": 9785 }, { "epoch": 0.35, "learning_rate": 5.7260629723891854e-06, "loss": 0.9474, "step": 9786 }, { "epoch": 0.35, "learning_rate": 5.724114121004836e-06, "loss": 0.9362, "step": 9787 }, { "epoch": 0.35, "learning_rate": 5.722165468334237e-06, "loss": 0.984, "step": 9788 }, { "epoch": 0.35, "learning_rate": 5.720217014467948e-06, "loss": 0.9445, "step": 9789 }, { "epoch": 0.35, "learning_rate": 5.718268759496524e-06, "loss": 0.9359, "step": 9790 }, { "epoch": 0.35, "learning_rate": 5.716320703510498e-06, "loss": 0.982, "step": 9791 }, { "epoch": 0.35, "learning_rate": 5.7143728466004064e-06, "loss": 0.9383, "step": 9792 }, { "epoch": 0.35, "learning_rate": 5.712425188856777e-06, "loss": 0.9477, "step": 9793 }, { "epoch": 0.35, "learning_rate": 5.7104777303701145e-06, "loss": 0.9265, "step": 9794 }, { "epoch": 0.35, "learning_rate": 5.708530471230929e-06, "loss": 0.965, "step": 9795 }, { "epoch": 0.35, "learning_rate": 5.706583411529712e-06, "loss": 0.9492, "step": 9796 }, { "epoch": 0.35, "learning_rate": 5.704636551356957e-06, "loss": 0.9487, "step": 9797 }, { "epoch": 0.35, "learning_rate": 5.702689890803132e-06, "loss": 0.9323, "step": 9798 }, { "epoch": 0.35, "learning_rate": 5.7007434299587085e-06, "loss": 0.9169, "step": 9799 }, { "epoch": 0.35, "learning_rate": 5.698797168914146e-06, "loss": 0.9654, "step": 9800 }, { "epoch": 0.35, "learning_rate": 5.6968511077598845e-06, "loss": 0.9471, "step": 9801 }, { "epoch": 0.35, "learning_rate": 5.69490524658638e-06, "loss": 0.9825, "step": 9802 }, { "epoch": 0.35, "learning_rate": 5.692959585484046e-06, "loss": 0.9022, "step": 9803 }, { "epoch": 0.35, "learning_rate": 5.6910141245433126e-06, "loss": 0.9559, "step": 9804 }, { "epoch": 0.35, "learning_rate": 5.689068863854589e-06, "loss": 0.9677, "step": 9805 }, { "epoch": 0.35, "learning_rate": 5.687123803508282e-06, "loss": 0.9309, "step": 9806 }, { "epoch": 0.35, "learning_rate": 5.685178943594776e-06, "loss": 0.955, "step": 9807 }, { "epoch": 0.35, "learning_rate": 5.68323428420446e-06, "loss": 0.9601, "step": 9808 }, { "epoch": 0.35, "learning_rate": 5.681289825427712e-06, "loss": 0.9987, "step": 9809 }, { "epoch": 0.35, "learning_rate": 5.679345567354884e-06, "loss": 0.9241, "step": 9810 }, { "epoch": 0.35, "learning_rate": 5.677401510076348e-06, "loss": 0.974, "step": 9811 }, { "epoch": 0.35, "learning_rate": 5.675457653682439e-06, "loss": 0.9479, "step": 9812 }, { "epoch": 0.35, "learning_rate": 5.673513998263499e-06, "loss": 0.9408, "step": 9813 }, { "epoch": 0.35, "learning_rate": 5.671570543909858e-06, "loss": 0.9221, "step": 9814 }, { "epoch": 0.35, "learning_rate": 5.6696272907118274e-06, "loss": 0.9269, "step": 9815 }, { "epoch": 0.35, "learning_rate": 5.667684238759719e-06, "loss": 0.9235, "step": 9816 }, { "epoch": 0.35, "learning_rate": 5.665741388143834e-06, "loss": 0.9465, "step": 9817 }, { "epoch": 0.35, "learning_rate": 5.663798738954466e-06, "loss": 0.9485, "step": 9818 }, { "epoch": 0.35, "learning_rate": 5.661856291281884e-06, "loss": 0.9697, "step": 9819 }, { "epoch": 0.35, "learning_rate": 5.659914045216376e-06, "loss": 0.9085, "step": 9820 }, { "epoch": 0.35, "learning_rate": 5.657972000848191e-06, "loss": 0.9479, "step": 9821 }, { "epoch": 0.35, "learning_rate": 5.656030158267587e-06, "loss": 0.9705, "step": 9822 }, { "epoch": 0.35, "learning_rate": 5.654088517564812e-06, "loss": 0.9551, "step": 9823 }, { "epoch": 0.35, "learning_rate": 5.65214707883009e-06, "loss": 0.9537, "step": 9824 }, { "epoch": 0.35, "learning_rate": 5.650205842153653e-06, "loss": 0.9766, "step": 9825 }, { "epoch": 0.35, "learning_rate": 5.648264807625712e-06, "loss": 0.9372, "step": 9826 }, { "epoch": 0.35, "learning_rate": 5.64632397533648e-06, "loss": 0.9607, "step": 9827 }, { "epoch": 0.35, "learning_rate": 5.644383345376147e-06, "loss": 0.9606, "step": 9828 }, { "epoch": 0.35, "learning_rate": 5.642442917834899e-06, "loss": 0.914, "step": 9829 }, { "epoch": 0.35, "learning_rate": 5.6405026928029225e-06, "loss": 0.9544, "step": 9830 }, { "epoch": 0.35, "learning_rate": 5.638562670370375e-06, "loss": 0.9339, "step": 9831 }, { "epoch": 0.35, "learning_rate": 5.63662285062742e-06, "loss": 0.9569, "step": 9832 }, { "epoch": 0.35, "learning_rate": 5.634683233664209e-06, "loss": 0.9872, "step": 9833 }, { "epoch": 0.35, "learning_rate": 5.632743819570878e-06, "loss": 0.9296, "step": 9834 }, { "epoch": 0.35, "learning_rate": 5.630804608437562e-06, "loss": 0.9856, "step": 9835 }, { "epoch": 0.35, "learning_rate": 5.628865600354383e-06, "loss": 0.9327, "step": 9836 }, { "epoch": 0.35, "learning_rate": 5.626926795411447e-06, "loss": 0.9645, "step": 9837 }, { "epoch": 0.35, "learning_rate": 5.624988193698859e-06, "loss": 0.9651, "step": 9838 }, { "epoch": 0.35, "learning_rate": 5.623049795306715e-06, "loss": 0.9129, "step": 9839 }, { "epoch": 0.35, "learning_rate": 5.621111600325093e-06, "loss": 0.9283, "step": 9840 }, { "epoch": 0.35, "learning_rate": 5.619173608844069e-06, "loss": 0.9763, "step": 9841 }, { "epoch": 0.35, "learning_rate": 5.617235820953707e-06, "loss": 0.9578, "step": 9842 }, { "epoch": 0.35, "learning_rate": 5.615298236744068e-06, "loss": 0.9255, "step": 9843 }, { "epoch": 0.35, "learning_rate": 5.613360856305184e-06, "loss": 0.9178, "step": 9844 }, { "epoch": 0.35, "learning_rate": 5.6114236797271085e-06, "loss": 0.9822, "step": 9845 }, { "epoch": 0.36, "learning_rate": 5.609486707099858e-06, "loss": 0.9151, "step": 9846 }, { "epoch": 0.36, "learning_rate": 5.607549938513446e-06, "loss": 0.9191, "step": 9847 }, { "epoch": 0.36, "learning_rate": 5.60561337405789e-06, "loss": 0.9246, "step": 9848 }, { "epoch": 0.36, "learning_rate": 5.603677013823181e-06, "loss": 0.9324, "step": 9849 }, { "epoch": 0.36, "learning_rate": 5.60174085789931e-06, "loss": 0.9251, "step": 9850 }, { "epoch": 0.36, "learning_rate": 5.599804906376256e-06, "loss": 0.9582, "step": 9851 }, { "epoch": 0.36, "learning_rate": 5.597869159343993e-06, "loss": 0.9328, "step": 9852 }, { "epoch": 0.36, "learning_rate": 5.595933616892473e-06, "loss": 0.9362, "step": 9853 }, { "epoch": 0.36, "learning_rate": 5.593998279111651e-06, "loss": 0.9735, "step": 9854 }, { "epoch": 0.36, "learning_rate": 5.592063146091472e-06, "loss": 0.9625, "step": 9855 }, { "epoch": 0.36, "learning_rate": 5.590128217921855e-06, "loss": 0.9162, "step": 9856 }, { "epoch": 0.36, "learning_rate": 5.5881934946927404e-06, "loss": 0.9515, "step": 9857 }, { "epoch": 0.36, "learning_rate": 5.586258976494028e-06, "loss": 0.9625, "step": 9858 }, { "epoch": 0.36, "learning_rate": 5.5843246634156215e-06, "loss": 0.9622, "step": 9859 }, { "epoch": 0.36, "learning_rate": 5.582390555547424e-06, "loss": 0.9416, "step": 9860 }, { "epoch": 0.36, "learning_rate": 5.5804566529793065e-06, "loss": 0.952, "step": 9861 }, { "epoch": 0.36, "learning_rate": 5.5785229558011496e-06, "loss": 0.9837, "step": 9862 }, { "epoch": 0.36, "learning_rate": 5.57658946410282e-06, "loss": 0.94, "step": 9863 }, { "epoch": 0.36, "learning_rate": 5.574656177974173e-06, "loss": 0.9436, "step": 9864 }, { "epoch": 0.36, "learning_rate": 5.572723097505047e-06, "loss": 0.9981, "step": 9865 }, { "epoch": 0.36, "learning_rate": 5.5707902227852914e-06, "loss": 0.9689, "step": 9866 }, { "epoch": 0.36, "learning_rate": 5.568857553904722e-06, "loss": 0.9702, "step": 9867 }, { "epoch": 0.36, "learning_rate": 5.566925090953159e-06, "loss": 0.9152, "step": 9868 }, { "epoch": 0.36, "learning_rate": 5.564992834020415e-06, "loss": 0.9935, "step": 9869 }, { "epoch": 0.36, "learning_rate": 5.5630607831962794e-06, "loss": 0.923, "step": 9870 }, { "epoch": 0.36, "learning_rate": 5.561128938570545e-06, "loss": 0.9105, "step": 9871 }, { "epoch": 0.36, "learning_rate": 5.55919730023299e-06, "loss": 0.9448, "step": 9872 }, { "epoch": 0.36, "learning_rate": 5.557265868273387e-06, "loss": 0.9573, "step": 9873 }, { "epoch": 0.36, "learning_rate": 5.555334642781491e-06, "loss": 0.9195, "step": 9874 }, { "epoch": 0.36, "learning_rate": 5.553403623847053e-06, "loss": 0.8979, "step": 9875 }, { "epoch": 0.36, "learning_rate": 5.5514728115598175e-06, "loss": 0.9337, "step": 9876 }, { "epoch": 0.36, "learning_rate": 5.549542206009505e-06, "loss": 0.9404, "step": 9877 }, { "epoch": 0.36, "learning_rate": 5.547611807285851e-06, "loss": 0.9296, "step": 9878 }, { "epoch": 0.36, "learning_rate": 5.5456816154785575e-06, "loss": 0.9677, "step": 9879 }, { "epoch": 0.36, "learning_rate": 5.543751630677328e-06, "loss": 0.9023, "step": 9880 }, { "epoch": 0.36, "learning_rate": 5.541821852971856e-06, "loss": 0.9731, "step": 9881 }, { "epoch": 0.36, "learning_rate": 5.539892282451828e-06, "loss": 0.9676, "step": 9882 }, { "epoch": 0.36, "learning_rate": 5.537962919206911e-06, "loss": 0.944, "step": 9883 }, { "epoch": 0.36, "learning_rate": 5.536033763326769e-06, "loss": 0.9159, "step": 9884 }, { "epoch": 0.36, "learning_rate": 5.5341048149010635e-06, "loss": 0.934, "step": 9885 }, { "epoch": 0.36, "learning_rate": 5.532176074019429e-06, "loss": 0.9155, "step": 9886 }, { "epoch": 0.36, "learning_rate": 5.530247540771503e-06, "loss": 0.9341, "step": 9887 }, { "epoch": 0.36, "learning_rate": 5.5283192152469135e-06, "loss": 0.937, "step": 9888 }, { "epoch": 0.36, "learning_rate": 5.526391097535275e-06, "loss": 0.9451, "step": 9889 }, { "epoch": 0.36, "learning_rate": 5.5244631877261865e-06, "loss": 0.9532, "step": 9890 }, { "epoch": 0.36, "learning_rate": 5.522535485909258e-06, "loss": 0.9346, "step": 9891 }, { "epoch": 0.36, "learning_rate": 5.520607992174062e-06, "loss": 0.9342, "step": 9892 }, { "epoch": 0.36, "learning_rate": 5.518680706610181e-06, "loss": 0.9326, "step": 9893 }, { "epoch": 0.36, "learning_rate": 5.5167536293071855e-06, "loss": 0.9138, "step": 9894 }, { "epoch": 0.36, "learning_rate": 5.514826760354625e-06, "loss": 0.9261, "step": 9895 }, { "epoch": 0.36, "learning_rate": 5.512900099842051e-06, "loss": 0.9314, "step": 9896 }, { "epoch": 0.36, "learning_rate": 5.510973647859002e-06, "loss": 0.9878, "step": 9897 }, { "epoch": 0.36, "learning_rate": 5.5090474044950074e-06, "loss": 1.0165, "step": 9898 }, { "epoch": 0.36, "learning_rate": 5.507121369839581e-06, "loss": 0.9304, "step": 9899 }, { "epoch": 0.36, "learning_rate": 5.505195543982235e-06, "loss": 0.9499, "step": 9900 }, { "epoch": 0.36, "learning_rate": 5.503269927012472e-06, "loss": 0.9659, "step": 9901 }, { "epoch": 0.36, "learning_rate": 5.501344519019769e-06, "loss": 1.0, "step": 9902 }, { "epoch": 0.36, "learning_rate": 5.499419320093624e-06, "loss": 0.9731, "step": 9903 }, { "epoch": 0.36, "learning_rate": 5.497494330323492e-06, "loss": 0.9653, "step": 9904 }, { "epoch": 0.36, "learning_rate": 5.495569549798838e-06, "loss": 0.9181, "step": 9905 }, { "epoch": 0.36, "learning_rate": 5.493644978609116e-06, "loss": 0.9363, "step": 9906 }, { "epoch": 0.36, "learning_rate": 5.491720616843762e-06, "loss": 0.9349, "step": 9907 }, { "epoch": 0.36, "learning_rate": 5.489796464592208e-06, "loss": 0.918, "step": 9908 }, { "epoch": 0.36, "learning_rate": 5.487872521943878e-06, "loss": 0.9394, "step": 9909 }, { "epoch": 0.36, "learning_rate": 5.485948788988183e-06, "loss": 0.8849, "step": 9910 }, { "epoch": 0.36, "learning_rate": 5.484025265814518e-06, "loss": 0.9335, "step": 9911 }, { "epoch": 0.36, "learning_rate": 5.482101952512289e-06, "loss": 0.9457, "step": 9912 }, { "epoch": 0.36, "learning_rate": 5.4801788491708655e-06, "loss": 0.9401, "step": 9913 }, { "epoch": 0.36, "learning_rate": 5.4782559558796235e-06, "loss": 0.932, "step": 9914 }, { "epoch": 0.36, "learning_rate": 5.476333272727934e-06, "loss": 0.9159, "step": 9915 }, { "epoch": 0.36, "learning_rate": 5.4744107998051365e-06, "loss": 0.9391, "step": 9916 }, { "epoch": 0.36, "learning_rate": 5.472488537200583e-06, "loss": 0.9288, "step": 9917 }, { "epoch": 0.36, "learning_rate": 5.470566485003606e-06, "loss": 0.9311, "step": 9918 }, { "epoch": 0.36, "learning_rate": 5.468644643303532e-06, "loss": 0.9131, "step": 9919 }, { "epoch": 0.36, "learning_rate": 5.466723012189666e-06, "loss": 0.986, "step": 9920 }, { "epoch": 0.36, "learning_rate": 5.4648015917513175e-06, "loss": 0.9472, "step": 9921 }, { "epoch": 0.36, "learning_rate": 5.462880382077786e-06, "loss": 0.9358, "step": 9922 }, { "epoch": 0.36, "learning_rate": 5.460959383258342e-06, "loss": 0.9504, "step": 9923 }, { "epoch": 0.36, "learning_rate": 5.459038595382279e-06, "loss": 0.9452, "step": 9924 }, { "epoch": 0.36, "learning_rate": 5.457118018538847e-06, "loss": 0.9461, "step": 9925 }, { "epoch": 0.36, "learning_rate": 5.4551976528173074e-06, "loss": 0.9334, "step": 9926 }, { "epoch": 0.36, "learning_rate": 5.453277498306906e-06, "loss": 0.9727, "step": 9927 }, { "epoch": 0.36, "learning_rate": 5.451357555096879e-06, "loss": 0.9404, "step": 9928 }, { "epoch": 0.36, "learning_rate": 5.4494378232764465e-06, "loss": 0.9495, "step": 9929 }, { "epoch": 0.36, "learning_rate": 5.44751830293483e-06, "loss": 0.9154, "step": 9930 }, { "epoch": 0.36, "learning_rate": 5.4455989941612365e-06, "loss": 1.0068, "step": 9931 }, { "epoch": 0.36, "learning_rate": 5.443679897044856e-06, "loss": 0.9245, "step": 9932 }, { "epoch": 0.36, "learning_rate": 5.441761011674877e-06, "loss": 0.9284, "step": 9933 }, { "epoch": 0.36, "learning_rate": 5.43984233814048e-06, "loss": 0.9035, "step": 9934 }, { "epoch": 0.36, "learning_rate": 5.437923876530833e-06, "loss": 0.9228, "step": 9935 }, { "epoch": 0.36, "learning_rate": 5.43600562693508e-06, "loss": 0.9631, "step": 9936 }, { "epoch": 0.36, "learning_rate": 5.434087589442386e-06, "loss": 0.9275, "step": 9937 }, { "epoch": 0.36, "learning_rate": 5.432169764141875e-06, "loss": 0.9134, "step": 9938 }, { "epoch": 0.36, "learning_rate": 5.430252151122679e-06, "loss": 0.9585, "step": 9939 }, { "epoch": 0.36, "learning_rate": 5.42833475047392e-06, "loss": 0.9097, "step": 9940 }, { "epoch": 0.36, "learning_rate": 5.426417562284696e-06, "loss": 0.9359, "step": 9941 }, { "epoch": 0.36, "learning_rate": 5.424500586644111e-06, "loss": 0.9352, "step": 9942 }, { "epoch": 0.36, "learning_rate": 5.422583823641249e-06, "loss": 0.9422, "step": 9943 }, { "epoch": 0.36, "learning_rate": 5.420667273365195e-06, "loss": 0.9549, "step": 9944 }, { "epoch": 0.36, "learning_rate": 5.418750935905008e-06, "loss": 0.9521, "step": 9945 }, { "epoch": 0.36, "learning_rate": 5.416834811349752e-06, "loss": 0.8928, "step": 9946 }, { "epoch": 0.36, "learning_rate": 5.414918899788477e-06, "loss": 0.9596, "step": 9947 }, { "epoch": 0.36, "learning_rate": 5.413003201310212e-06, "loss": 0.8897, "step": 9948 }, { "epoch": 0.36, "learning_rate": 5.411087716003998e-06, "loss": 0.9133, "step": 9949 }, { "epoch": 0.36, "learning_rate": 5.409172443958844e-06, "loss": 0.941, "step": 9950 }, { "epoch": 0.36, "learning_rate": 5.40725738526376e-06, "loss": 0.981, "step": 9951 }, { "epoch": 0.36, "learning_rate": 5.405342540007753e-06, "loss": 0.9145, "step": 9952 }, { "epoch": 0.36, "learning_rate": 5.403427908279801e-06, "loss": 0.9627, "step": 9953 }, { "epoch": 0.36, "learning_rate": 5.401513490168887e-06, "loss": 0.9278, "step": 9954 }, { "epoch": 0.36, "learning_rate": 5.399599285763981e-06, "loss": 0.9941, "step": 9955 }, { "epoch": 0.36, "learning_rate": 5.3976852951540445e-06, "loss": 0.9631, "step": 9956 }, { "epoch": 0.36, "learning_rate": 5.395771518428017e-06, "loss": 0.9553, "step": 9957 }, { "epoch": 0.36, "learning_rate": 5.393857955674851e-06, "loss": 0.9436, "step": 9958 }, { "epoch": 0.36, "learning_rate": 5.391944606983466e-06, "loss": 0.9205, "step": 9959 }, { "epoch": 0.36, "learning_rate": 5.390031472442783e-06, "loss": 0.9392, "step": 9960 }, { "epoch": 0.36, "learning_rate": 5.388118552141718e-06, "loss": 0.9699, "step": 9961 }, { "epoch": 0.36, "learning_rate": 5.3862058461691605e-06, "loss": 0.9505, "step": 9962 }, { "epoch": 0.36, "learning_rate": 5.384293354614003e-06, "loss": 0.9419, "step": 9963 }, { "epoch": 0.36, "learning_rate": 5.382381077565129e-06, "loss": 0.9328, "step": 9964 }, { "epoch": 0.36, "learning_rate": 5.380469015111407e-06, "loss": 0.9278, "step": 9965 }, { "epoch": 0.36, "learning_rate": 5.378557167341691e-06, "loss": 0.9259, "step": 9966 }, { "epoch": 0.36, "learning_rate": 5.376645534344835e-06, "loss": 0.9852, "step": 9967 }, { "epoch": 0.36, "learning_rate": 5.374734116209683e-06, "loss": 0.9903, "step": 9968 }, { "epoch": 0.36, "learning_rate": 5.372822913025051e-06, "loss": 0.9783, "step": 9969 }, { "epoch": 0.36, "learning_rate": 5.3709119248797735e-06, "loss": 0.9608, "step": 9970 }, { "epoch": 0.36, "learning_rate": 5.369001151862651e-06, "loss": 0.9391, "step": 9971 }, { "epoch": 0.36, "learning_rate": 5.367090594062485e-06, "loss": 0.9241, "step": 9972 }, { "epoch": 0.36, "learning_rate": 5.365180251568065e-06, "loss": 0.9553, "step": 9973 }, { "epoch": 0.36, "learning_rate": 5.3632701244681776e-06, "loss": 0.971, "step": 9974 }, { "epoch": 0.36, "learning_rate": 5.361360212851581e-06, "loss": 0.9243, "step": 9975 }, { "epoch": 0.36, "learning_rate": 5.359450516807038e-06, "loss": 0.9235, "step": 9976 }, { "epoch": 0.36, "learning_rate": 5.357541036423306e-06, "loss": 0.9281, "step": 9977 }, { "epoch": 0.36, "learning_rate": 5.355631771789114e-06, "loss": 0.9201, "step": 9978 }, { "epoch": 0.36, "learning_rate": 5.353722722993196e-06, "loss": 0.9525, "step": 9979 }, { "epoch": 0.36, "learning_rate": 5.351813890124271e-06, "loss": 0.9692, "step": 9980 }, { "epoch": 0.36, "learning_rate": 5.34990527327105e-06, "loss": 0.9537, "step": 9981 }, { "epoch": 0.36, "learning_rate": 5.34799687252223e-06, "loss": 0.9476, "step": 9982 }, { "epoch": 0.36, "learning_rate": 5.346088687966507e-06, "loss": 0.9517, "step": 9983 }, { "epoch": 0.36, "learning_rate": 5.344180719692549e-06, "loss": 0.9425, "step": 9984 }, { "epoch": 0.36, "learning_rate": 5.342272967789033e-06, "loss": 1.0045, "step": 9985 }, { "epoch": 0.36, "learning_rate": 5.3403654323446184e-06, "loss": 0.8979, "step": 9986 }, { "epoch": 0.36, "learning_rate": 5.338458113447949e-06, "loss": 0.9498, "step": 9987 }, { "epoch": 0.36, "learning_rate": 5.336551011187666e-06, "loss": 0.9162, "step": 9988 }, { "epoch": 0.36, "learning_rate": 5.3346441256524005e-06, "loss": 0.9436, "step": 9989 }, { "epoch": 0.36, "learning_rate": 5.332737456930774e-06, "loss": 0.9084, "step": 9990 }, { "epoch": 0.36, "learning_rate": 5.330831005111388e-06, "loss": 0.9781, "step": 9991 }, { "epoch": 0.36, "learning_rate": 5.3289247702828436e-06, "loss": 0.9307, "step": 9992 }, { "epoch": 0.36, "learning_rate": 5.327018752533737e-06, "loss": 0.9251, "step": 9993 }, { "epoch": 0.36, "learning_rate": 5.325112951952631e-06, "loss": 1.0109, "step": 9994 }, { "epoch": 0.36, "learning_rate": 5.323207368628112e-06, "loss": 0.9212, "step": 9995 }, { "epoch": 0.37, "learning_rate": 5.321302002648727e-06, "loss": 0.9418, "step": 9996 }, { "epoch": 0.37, "learning_rate": 5.319396854103026e-06, "loss": 0.9683, "step": 9997 }, { "epoch": 0.37, "learning_rate": 5.317491923079554e-06, "loss": 0.9299, "step": 9998 }, { "epoch": 0.37, "learning_rate": 5.3155872096668284e-06, "loss": 0.918, "step": 9999 }, { "epoch": 0.37, "learning_rate": 5.313682713953372e-06, "loss": 0.9076, "step": 10000 }, { "epoch": 0.37, "learning_rate": 5.311778436027694e-06, "loss": 0.9294, "step": 10001 }, { "epoch": 0.37, "learning_rate": 5.309874375978296e-06, "loss": 0.9164, "step": 10002 }, { "epoch": 0.37, "learning_rate": 5.3079705338936515e-06, "loss": 0.9634, "step": 10003 }, { "epoch": 0.37, "learning_rate": 5.306066909862255e-06, "loss": 0.9006, "step": 10004 }, { "epoch": 0.37, "learning_rate": 5.3041635039725635e-06, "loss": 0.9256, "step": 10005 }, { "epoch": 0.37, "learning_rate": 5.302260316313037e-06, "loss": 0.9555, "step": 10006 }, { "epoch": 0.37, "learning_rate": 5.300357346972125e-06, "loss": 0.9112, "step": 10007 }, { "epoch": 0.37, "learning_rate": 5.298454596038258e-06, "loss": 0.936, "step": 10008 }, { "epoch": 0.37, "learning_rate": 5.2965520635998676e-06, "loss": 0.9223, "step": 10009 }, { "epoch": 0.37, "learning_rate": 5.294649749745369e-06, "loss": 0.9783, "step": 10010 }, { "epoch": 0.37, "learning_rate": 5.292747654563173e-06, "loss": 0.9306, "step": 10011 }, { "epoch": 0.37, "learning_rate": 5.290845778141665e-06, "loss": 0.9257, "step": 10012 }, { "epoch": 0.37, "learning_rate": 5.288944120569244e-06, "loss": 0.8966, "step": 10013 }, { "epoch": 0.37, "learning_rate": 5.287042681934282e-06, "loss": 0.9376, "step": 10014 }, { "epoch": 0.37, "learning_rate": 5.285141462325134e-06, "loss": 0.8895, "step": 10015 }, { "epoch": 0.37, "learning_rate": 5.2832404618301725e-06, "loss": 0.9167, "step": 10016 }, { "epoch": 0.37, "learning_rate": 5.28133968053773e-06, "loss": 0.9871, "step": 10017 }, { "epoch": 0.37, "learning_rate": 5.279439118536149e-06, "loss": 0.9423, "step": 10018 }, { "epoch": 0.37, "learning_rate": 5.27753877591375e-06, "loss": 0.9765, "step": 10019 }, { "epoch": 0.37, "learning_rate": 5.275638652758854e-06, "loss": 0.9621, "step": 10020 }, { "epoch": 0.37, "learning_rate": 5.273738749159758e-06, "loss": 0.915, "step": 10021 }, { "epoch": 0.37, "learning_rate": 5.27183906520476e-06, "loss": 0.968, "step": 10022 }, { "epoch": 0.37, "learning_rate": 5.269939600982151e-06, "loss": 0.9096, "step": 10023 }, { "epoch": 0.37, "learning_rate": 5.268040356580191e-06, "loss": 0.9655, "step": 10024 }, { "epoch": 0.37, "learning_rate": 5.266141332087151e-06, "loss": 0.972, "step": 10025 }, { "epoch": 0.37, "learning_rate": 5.264242527591287e-06, "loss": 0.9559, "step": 10026 }, { "epoch": 0.37, "learning_rate": 5.262343943180838e-06, "loss": 0.9799, "step": 10027 }, { "epoch": 0.37, "learning_rate": 5.26044557894404e-06, "loss": 0.9584, "step": 10028 }, { "epoch": 0.37, "learning_rate": 5.2585474349691205e-06, "loss": 0.9383, "step": 10029 }, { "epoch": 0.37, "learning_rate": 5.25664951134428e-06, "loss": 0.9458, "step": 10030 }, { "epoch": 0.37, "learning_rate": 5.254751808157731e-06, "loss": 0.9537, "step": 10031 }, { "epoch": 0.37, "learning_rate": 5.252854325497665e-06, "loss": 0.9543, "step": 10032 }, { "epoch": 0.37, "learning_rate": 5.250957063452258e-06, "loss": 0.965, "step": 10033 }, { "epoch": 0.37, "learning_rate": 5.2490600221096835e-06, "loss": 0.92, "step": 10034 }, { "epoch": 0.37, "learning_rate": 5.247163201558105e-06, "loss": 0.9613, "step": 10035 }, { "epoch": 0.37, "learning_rate": 5.245266601885678e-06, "loss": 0.9387, "step": 10036 }, { "epoch": 0.37, "learning_rate": 5.243370223180534e-06, "loss": 0.9517, "step": 10037 }, { "epoch": 0.37, "learning_rate": 5.241474065530808e-06, "loss": 0.9116, "step": 10038 }, { "epoch": 0.37, "learning_rate": 5.239578129024625e-06, "loss": 0.9578, "step": 10039 }, { "epoch": 0.37, "learning_rate": 5.237682413750083e-06, "loss": 0.964, "step": 10040 }, { "epoch": 0.37, "learning_rate": 5.235786919795296e-06, "loss": 0.9558, "step": 10041 }, { "epoch": 0.37, "learning_rate": 5.233891647248344e-06, "loss": 0.9296, "step": 10042 }, { "epoch": 0.37, "learning_rate": 5.231996596197309e-06, "loss": 0.9443, "step": 10043 }, { "epoch": 0.37, "learning_rate": 5.230101766730264e-06, "loss": 0.9595, "step": 10044 }, { "epoch": 0.37, "learning_rate": 5.228207158935259e-06, "loss": 0.9506, "step": 10045 }, { "epoch": 0.37, "learning_rate": 5.226312772900348e-06, "loss": 0.9276, "step": 10046 }, { "epoch": 0.37, "learning_rate": 5.2244186087135686e-06, "loss": 0.9371, "step": 10047 }, { "epoch": 0.37, "learning_rate": 5.222524666462952e-06, "loss": 0.8978, "step": 10048 }, { "epoch": 0.37, "learning_rate": 5.2206309462365025e-06, "loss": 0.9429, "step": 10049 }, { "epoch": 0.37, "learning_rate": 5.218737448122244e-06, "loss": 0.9086, "step": 10050 }, { "epoch": 0.37, "learning_rate": 5.216844172208162e-06, "loss": 0.962, "step": 10051 }, { "epoch": 0.37, "learning_rate": 5.214951118582248e-06, "loss": 0.9447, "step": 10052 }, { "epoch": 0.37, "learning_rate": 5.213058287332478e-06, "loss": 0.9409, "step": 10053 }, { "epoch": 0.37, "learning_rate": 5.211165678546814e-06, "loss": 0.9563, "step": 10054 }, { "epoch": 0.37, "learning_rate": 5.209273292313214e-06, "loss": 0.9236, "step": 10055 }, { "epoch": 0.37, "learning_rate": 5.207381128719622e-06, "loss": 0.9376, "step": 10056 }, { "epoch": 0.37, "learning_rate": 5.205489187853978e-06, "loss": 0.9923, "step": 10057 }, { "epoch": 0.37, "learning_rate": 5.203597469804194e-06, "loss": 0.9679, "step": 10058 }, { "epoch": 0.37, "learning_rate": 5.2017059746582e-06, "loss": 0.9883, "step": 10059 }, { "epoch": 0.37, "learning_rate": 5.199814702503891e-06, "loss": 0.9489, "step": 10060 }, { "epoch": 0.37, "learning_rate": 5.197923653429155e-06, "loss": 0.9102, "step": 10061 }, { "epoch": 0.37, "learning_rate": 5.196032827521889e-06, "loss": 0.9367, "step": 10062 }, { "epoch": 0.37, "learning_rate": 5.194142224869951e-06, "loss": 0.9854, "step": 10063 }, { "epoch": 0.37, "learning_rate": 5.192251845561212e-06, "loss": 0.9709, "step": 10064 }, { "epoch": 0.37, "learning_rate": 5.190361689683521e-06, "loss": 0.9541, "step": 10065 }, { "epoch": 0.37, "learning_rate": 5.188471757324722e-06, "loss": 0.951, "step": 10066 }, { "epoch": 0.37, "learning_rate": 5.186582048572644e-06, "loss": 0.916, "step": 10067 }, { "epoch": 0.37, "learning_rate": 5.184692563515104e-06, "loss": 0.9169, "step": 10068 }, { "epoch": 0.37, "learning_rate": 5.182803302239923e-06, "loss": 0.9866, "step": 10069 }, { "epoch": 0.37, "learning_rate": 5.180914264834887e-06, "loss": 0.9641, "step": 10070 }, { "epoch": 0.37, "learning_rate": 5.1790254513877945e-06, "loss": 0.9541, "step": 10071 }, { "epoch": 0.37, "learning_rate": 5.177136861986423e-06, "loss": 0.9672, "step": 10072 }, { "epoch": 0.37, "learning_rate": 5.175248496718539e-06, "loss": 0.9506, "step": 10073 }, { "epoch": 0.37, "learning_rate": 5.173360355671901e-06, "loss": 0.9637, "step": 10074 }, { "epoch": 0.37, "learning_rate": 5.171472438934264e-06, "loss": 0.9946, "step": 10075 }, { "epoch": 0.37, "learning_rate": 5.169584746593355e-06, "loss": 0.9441, "step": 10076 }, { "epoch": 0.37, "learning_rate": 5.167697278736905e-06, "loss": 0.9439, "step": 10077 }, { "epoch": 0.37, "learning_rate": 5.165810035452636e-06, "loss": 0.9537, "step": 10078 }, { "epoch": 0.37, "learning_rate": 5.163923016828243e-06, "loss": 0.8904, "step": 10079 }, { "epoch": 0.37, "learning_rate": 5.16203622295143e-06, "loss": 0.9163, "step": 10080 }, { "epoch": 0.37, "learning_rate": 5.160149653909878e-06, "loss": 0.9504, "step": 10081 }, { "epoch": 0.37, "learning_rate": 5.158263309791267e-06, "loss": 0.9012, "step": 10082 }, { "epoch": 0.37, "learning_rate": 5.156377190683256e-06, "loss": 0.9716, "step": 10083 }, { "epoch": 0.37, "learning_rate": 5.1544912966735e-06, "loss": 0.9303, "step": 10084 }, { "epoch": 0.37, "learning_rate": 5.1526056278496424e-06, "loss": 0.9534, "step": 10085 }, { "epoch": 0.37, "learning_rate": 5.150720184299316e-06, "loss": 0.9088, "step": 10086 }, { "epoch": 0.37, "learning_rate": 5.14883496611015e-06, "loss": 0.9808, "step": 10087 }, { "epoch": 0.37, "learning_rate": 5.146949973369745e-06, "loss": 0.9717, "step": 10088 }, { "epoch": 0.37, "learning_rate": 5.145065206165708e-06, "loss": 0.9631, "step": 10089 }, { "epoch": 0.37, "learning_rate": 5.143180664585634e-06, "loss": 0.9358, "step": 10090 }, { "epoch": 0.37, "learning_rate": 5.141296348717096e-06, "loss": 0.9081, "step": 10091 }, { "epoch": 0.37, "learning_rate": 5.139412258647667e-06, "loss": 0.8985, "step": 10092 }, { "epoch": 0.37, "learning_rate": 5.137528394464906e-06, "loss": 0.9276, "step": 10093 }, { "epoch": 0.37, "learning_rate": 5.135644756256369e-06, "loss": 0.9275, "step": 10094 }, { "epoch": 0.37, "learning_rate": 5.1337613441095775e-06, "loss": 0.9468, "step": 10095 }, { "epoch": 0.37, "learning_rate": 5.131878158112082e-06, "loss": 0.9279, "step": 10096 }, { "epoch": 0.37, "learning_rate": 5.129995198351383e-06, "loss": 0.9508, "step": 10097 }, { "epoch": 0.37, "learning_rate": 5.1281124649149925e-06, "loss": 0.9332, "step": 10098 }, { "epoch": 0.37, "learning_rate": 5.126229957890413e-06, "loss": 0.9863, "step": 10099 }, { "epoch": 0.37, "learning_rate": 5.124347677365121e-06, "loss": 0.951, "step": 10100 }, { "epoch": 0.37, "learning_rate": 5.122465623426595e-06, "loss": 0.9444, "step": 10101 }, { "epoch": 0.37, "learning_rate": 5.1205837961623015e-06, "loss": 0.9274, "step": 10102 }, { "epoch": 0.37, "learning_rate": 5.1187021956596975e-06, "loss": 0.9084, "step": 10103 }, { "epoch": 0.37, "learning_rate": 5.116820822006217e-06, "loss": 0.9261, "step": 10104 }, { "epoch": 0.37, "learning_rate": 5.1149396752893075e-06, "loss": 0.9446, "step": 10105 }, { "epoch": 0.37, "learning_rate": 5.113058755596383e-06, "loss": 0.9414, "step": 10106 }, { "epoch": 0.37, "learning_rate": 5.11117806301485e-06, "loss": 0.9636, "step": 10107 }, { "epoch": 0.37, "learning_rate": 5.1092975976321255e-06, "loss": 0.9034, "step": 10108 }, { "epoch": 0.37, "learning_rate": 5.107417359535588e-06, "loss": 0.9654, "step": 10109 }, { "epoch": 0.37, "learning_rate": 5.105537348812621e-06, "loss": 0.9673, "step": 10110 }, { "epoch": 0.37, "learning_rate": 5.103657565550597e-06, "loss": 0.9177, "step": 10111 }, { "epoch": 0.37, "learning_rate": 5.101778009836877e-06, "loss": 0.9325, "step": 10112 }, { "epoch": 0.37, "learning_rate": 5.099898681758805e-06, "loss": 0.95, "step": 10113 }, { "epoch": 0.37, "learning_rate": 5.0980195814037195e-06, "loss": 0.8892, "step": 10114 }, { "epoch": 0.37, "learning_rate": 5.096140708858954e-06, "loss": 0.9392, "step": 10115 }, { "epoch": 0.37, "learning_rate": 5.0942620642118126e-06, "loss": 0.9697, "step": 10116 }, { "epoch": 0.37, "learning_rate": 5.092383647549619e-06, "loss": 0.9711, "step": 10117 }, { "epoch": 0.37, "learning_rate": 5.090505458959657e-06, "loss": 0.9569, "step": 10118 }, { "epoch": 0.37, "learning_rate": 5.088627498529215e-06, "loss": 1.0156, "step": 10119 }, { "epoch": 0.37, "learning_rate": 5.0867497663455685e-06, "loss": 0.9193, "step": 10120 }, { "epoch": 0.37, "learning_rate": 5.084872262495986e-06, "loss": 0.9028, "step": 10121 }, { "epoch": 0.37, "learning_rate": 5.082994987067711e-06, "loss": 0.9151, "step": 10122 }, { "epoch": 0.37, "learning_rate": 5.08111794014799e-06, "loss": 0.94, "step": 10123 }, { "epoch": 0.37, "learning_rate": 5.079241121824063e-06, "loss": 0.8713, "step": 10124 }, { "epoch": 0.37, "learning_rate": 5.07736453218314e-06, "loss": 0.9173, "step": 10125 }, { "epoch": 0.37, "learning_rate": 5.0754881713124394e-06, "loss": 0.919, "step": 10126 }, { "epoch": 0.37, "learning_rate": 5.073612039299157e-06, "loss": 0.9623, "step": 10127 }, { "epoch": 0.37, "learning_rate": 5.071736136230489e-06, "loss": 0.9142, "step": 10128 }, { "epoch": 0.37, "learning_rate": 5.0698604621936075e-06, "loss": 0.9672, "step": 10129 }, { "epoch": 0.37, "learning_rate": 5.067985017275685e-06, "loss": 0.9204, "step": 10130 }, { "epoch": 0.37, "learning_rate": 5.0661098015638766e-06, "loss": 0.9627, "step": 10131 }, { "epoch": 0.37, "learning_rate": 5.06423481514533e-06, "loss": 0.9363, "step": 10132 }, { "epoch": 0.37, "learning_rate": 5.062360058107188e-06, "loss": 0.9403, "step": 10133 }, { "epoch": 0.37, "learning_rate": 5.060485530536567e-06, "loss": 0.9769, "step": 10134 }, { "epoch": 0.37, "learning_rate": 5.058611232520586e-06, "loss": 0.9588, "step": 10135 }, { "epoch": 0.37, "learning_rate": 5.056737164146353e-06, "loss": 0.9565, "step": 10136 }, { "epoch": 0.37, "learning_rate": 5.054863325500955e-06, "loss": 0.9495, "step": 10137 }, { "epoch": 0.37, "learning_rate": 5.052989716671478e-06, "loss": 0.9668, "step": 10138 }, { "epoch": 0.37, "learning_rate": 5.051116337744996e-06, "loss": 0.9621, "step": 10139 }, { "epoch": 0.37, "learning_rate": 5.049243188808573e-06, "loss": 0.9046, "step": 10140 }, { "epoch": 0.37, "learning_rate": 5.0473702699492484e-06, "loss": 0.9662, "step": 10141 }, { "epoch": 0.37, "learning_rate": 5.045497581254079e-06, "loss": 0.9331, "step": 10142 }, { "epoch": 0.37, "learning_rate": 5.0436251228100826e-06, "loss": 0.9361, "step": 10143 }, { "epoch": 0.37, "learning_rate": 5.0417528947042835e-06, "loss": 0.9259, "step": 10144 }, { "epoch": 0.37, "learning_rate": 5.039880897023691e-06, "loss": 0.9109, "step": 10145 }, { "epoch": 0.38, "learning_rate": 5.038009129855297e-06, "loss": 0.9353, "step": 10146 }, { "epoch": 0.38, "learning_rate": 5.0361375932860916e-06, "loss": 0.9101, "step": 10147 }, { "epoch": 0.38, "learning_rate": 5.03426628740305e-06, "loss": 0.9556, "step": 10148 }, { "epoch": 0.38, "learning_rate": 5.032395212293143e-06, "loss": 0.9564, "step": 10149 }, { "epoch": 0.38, "learning_rate": 5.030524368043313e-06, "loss": 0.9076, "step": 10150 }, { "epoch": 0.38, "learning_rate": 5.02865375474052e-06, "loss": 0.9255, "step": 10151 }, { "epoch": 0.38, "learning_rate": 5.026783372471688e-06, "loss": 0.9707, "step": 10152 }, { "epoch": 0.38, "learning_rate": 5.024913221323733e-06, "loss": 0.9007, "step": 10153 }, { "epoch": 0.38, "learning_rate": 5.023043301383583e-06, "loss": 0.9623, "step": 10154 }, { "epoch": 0.38, "learning_rate": 5.021173612738126e-06, "loss": 0.9319, "step": 10155 }, { "epoch": 0.38, "learning_rate": 5.019304155474256e-06, "loss": 0.943, "step": 10156 }, { "epoch": 0.38, "learning_rate": 5.0174349296788524e-06, "loss": 0.9138, "step": 10157 }, { "epoch": 0.38, "learning_rate": 5.01556593543879e-06, "loss": 0.9537, "step": 10158 }, { "epoch": 0.38, "learning_rate": 5.013697172840917e-06, "loss": 0.9273, "step": 10159 }, { "epoch": 0.38, "learning_rate": 5.0118286419720856e-06, "loss": 0.9228, "step": 10160 }, { "epoch": 0.38, "learning_rate": 5.009960342919133e-06, "loss": 0.9116, "step": 10161 }, { "epoch": 0.38, "learning_rate": 5.0080922757688785e-06, "loss": 0.9132, "step": 10162 }, { "epoch": 0.38, "learning_rate": 5.0062244406081495e-06, "loss": 0.9425, "step": 10163 }, { "epoch": 0.38, "learning_rate": 5.004356837523738e-06, "loss": 0.924, "step": 10164 }, { "epoch": 0.38, "learning_rate": 5.0024894666024425e-06, "loss": 0.9096, "step": 10165 }, { "epoch": 0.38, "learning_rate": 5.000622327931044e-06, "loss": 0.9599, "step": 10166 }, { "epoch": 0.38, "learning_rate": 4.998755421596321e-06, "loss": 0.9302, "step": 10167 }, { "epoch": 0.38, "learning_rate": 4.996888747685024e-06, "loss": 0.959, "step": 10168 }, { "epoch": 0.38, "learning_rate": 4.9950223062839075e-06, "loss": 0.9848, "step": 10169 }, { "epoch": 0.38, "learning_rate": 4.993156097479716e-06, "loss": 0.9275, "step": 10170 }, { "epoch": 0.38, "learning_rate": 4.991290121359169e-06, "loss": 0.9114, "step": 10171 }, { "epoch": 0.38, "learning_rate": 4.989424378008989e-06, "loss": 0.9637, "step": 10172 }, { "epoch": 0.38, "learning_rate": 4.9875588675158815e-06, "loss": 0.9808, "step": 10173 }, { "epoch": 0.38, "learning_rate": 4.985693589966545e-06, "loss": 0.9748, "step": 10174 }, { "epoch": 0.38, "learning_rate": 4.983828545447666e-06, "loss": 0.9117, "step": 10175 }, { "epoch": 0.38, "learning_rate": 4.9819637340459124e-06, "loss": 0.958, "step": 10176 }, { "epoch": 0.38, "learning_rate": 4.980099155847952e-06, "loss": 0.9415, "step": 10177 }, { "epoch": 0.38, "learning_rate": 4.978234810940438e-06, "loss": 0.9368, "step": 10178 }, { "epoch": 0.38, "learning_rate": 4.976370699410015e-06, "loss": 0.9573, "step": 10179 }, { "epoch": 0.38, "learning_rate": 4.9745068213433055e-06, "loss": 0.9232, "step": 10180 }, { "epoch": 0.38, "learning_rate": 4.972643176826937e-06, "loss": 0.9368, "step": 10181 }, { "epoch": 0.38, "learning_rate": 4.970779765947519e-06, "loss": 0.9148, "step": 10182 }, { "epoch": 0.38, "learning_rate": 4.968916588791644e-06, "loss": 0.9752, "step": 10183 }, { "epoch": 0.38, "learning_rate": 4.967053645445904e-06, "loss": 0.9929, "step": 10184 }, { "epoch": 0.38, "learning_rate": 4.965190935996875e-06, "loss": 0.8847, "step": 10185 }, { "epoch": 0.38, "learning_rate": 4.963328460531127e-06, "loss": 0.9628, "step": 10186 }, { "epoch": 0.38, "learning_rate": 4.961466219135203e-06, "loss": 0.9357, "step": 10187 }, { "epoch": 0.38, "learning_rate": 4.959604211895664e-06, "loss": 0.919, "step": 10188 }, { "epoch": 0.38, "learning_rate": 4.957742438899029e-06, "loss": 0.9664, "step": 10189 }, { "epoch": 0.38, "learning_rate": 4.955880900231828e-06, "loss": 0.9356, "step": 10190 }, { "epoch": 0.38, "learning_rate": 4.954019595980572e-06, "loss": 0.8753, "step": 10191 }, { "epoch": 0.38, "learning_rate": 4.952158526231759e-06, "loss": 0.9706, "step": 10192 }, { "epoch": 0.38, "learning_rate": 4.950297691071878e-06, "loss": 0.9727, "step": 10193 }, { "epoch": 0.38, "learning_rate": 4.94843709058741e-06, "loss": 0.9183, "step": 10194 }, { "epoch": 0.38, "learning_rate": 4.946576724864827e-06, "loss": 0.9549, "step": 10195 }, { "epoch": 0.38, "learning_rate": 4.944716593990573e-06, "loss": 0.9622, "step": 10196 }, { "epoch": 0.38, "learning_rate": 4.942856698051111e-06, "loss": 0.9458, "step": 10197 }, { "epoch": 0.38, "learning_rate": 4.940997037132866e-06, "loss": 0.9695, "step": 10198 }, { "epoch": 0.38, "learning_rate": 4.939137611322259e-06, "loss": 0.9557, "step": 10199 }, { "epoch": 0.38, "learning_rate": 4.937278420705714e-06, "loss": 0.8965, "step": 10200 }, { "epoch": 0.38, "learning_rate": 4.935419465369625e-06, "loss": 0.9716, "step": 10201 }, { "epoch": 0.38, "learning_rate": 4.9335607454003856e-06, "loss": 0.94, "step": 10202 }, { "epoch": 0.38, "learning_rate": 4.931702260884377e-06, "loss": 0.9799, "step": 10203 }, { "epoch": 0.38, "learning_rate": 4.929844011907973e-06, "loss": 0.914, "step": 10204 }, { "epoch": 0.38, "learning_rate": 4.9279859985575225e-06, "loss": 0.9378, "step": 10205 }, { "epoch": 0.38, "learning_rate": 4.92612822091938e-06, "loss": 0.9356, "step": 10206 }, { "epoch": 0.38, "learning_rate": 4.924270679079884e-06, "loss": 0.9434, "step": 10207 }, { "epoch": 0.38, "learning_rate": 4.922413373125351e-06, "loss": 0.9336, "step": 10208 }, { "epoch": 0.38, "learning_rate": 4.920556303142108e-06, "loss": 0.9622, "step": 10209 }, { "epoch": 0.38, "learning_rate": 4.918699469216449e-06, "loss": 0.959, "step": 10210 }, { "epoch": 0.38, "learning_rate": 4.9168428714346705e-06, "loss": 0.9603, "step": 10211 }, { "epoch": 0.38, "learning_rate": 4.914986509883054e-06, "loss": 0.9603, "step": 10212 }, { "epoch": 0.38, "learning_rate": 4.913130384647875e-06, "loss": 0.967, "step": 10213 }, { "epoch": 0.38, "learning_rate": 4.9112744958153855e-06, "loss": 0.9221, "step": 10214 }, { "epoch": 0.38, "learning_rate": 4.909418843471838e-06, "loss": 0.9978, "step": 10215 }, { "epoch": 0.38, "learning_rate": 4.907563427703476e-06, "loss": 0.9476, "step": 10216 }, { "epoch": 0.38, "learning_rate": 4.905708248596514e-06, "loss": 0.9567, "step": 10217 }, { "epoch": 0.38, "learning_rate": 4.903853306237178e-06, "loss": 0.94, "step": 10218 }, { "epoch": 0.38, "learning_rate": 4.901998600711669e-06, "loss": 0.9152, "step": 10219 }, { "epoch": 0.38, "learning_rate": 4.900144132106183e-06, "loss": 0.9479, "step": 10220 }, { "epoch": 0.38, "learning_rate": 4.898289900506904e-06, "loss": 0.959, "step": 10221 }, { "epoch": 0.38, "learning_rate": 4.896435905999999e-06, "loss": 0.9173, "step": 10222 }, { "epoch": 0.38, "learning_rate": 4.894582148671632e-06, "loss": 0.9454, "step": 10223 }, { "epoch": 0.38, "learning_rate": 4.892728628607951e-06, "loss": 0.9387, "step": 10224 }, { "epoch": 0.38, "learning_rate": 4.890875345895102e-06, "loss": 0.945, "step": 10225 }, { "epoch": 0.38, "learning_rate": 4.889022300619202e-06, "loss": 0.9635, "step": 10226 }, { "epoch": 0.38, "learning_rate": 4.887169492866375e-06, "loss": 0.9259, "step": 10227 }, { "epoch": 0.38, "learning_rate": 4.885316922722728e-06, "loss": 0.9924, "step": 10228 }, { "epoch": 0.38, "learning_rate": 4.8834645902743485e-06, "loss": 0.936, "step": 10229 }, { "epoch": 0.38, "learning_rate": 4.881612495607324e-06, "loss": 0.9146, "step": 10230 }, { "epoch": 0.38, "learning_rate": 4.879760638807728e-06, "loss": 0.9882, "step": 10231 }, { "epoch": 0.38, "learning_rate": 4.877909019961624e-06, "loss": 0.9177, "step": 10232 }, { "epoch": 0.38, "learning_rate": 4.876057639155053e-06, "loss": 0.9368, "step": 10233 }, { "epoch": 0.38, "learning_rate": 4.874206496474069e-06, "loss": 0.9199, "step": 10234 }, { "epoch": 0.38, "learning_rate": 4.872355592004689e-06, "loss": 0.9695, "step": 10235 }, { "epoch": 0.38, "learning_rate": 4.870504925832934e-06, "loss": 0.9482, "step": 10236 }, { "epoch": 0.38, "learning_rate": 4.8686544980448135e-06, "loss": 0.9526, "step": 10237 }, { "epoch": 0.38, "learning_rate": 4.866804308726316e-06, "loss": 0.9116, "step": 10238 }, { "epoch": 0.38, "learning_rate": 4.864954357963429e-06, "loss": 0.9111, "step": 10239 }, { "epoch": 0.38, "learning_rate": 4.863104645842126e-06, "loss": 0.9103, "step": 10240 }, { "epoch": 0.38, "learning_rate": 4.861255172448372e-06, "loss": 0.9354, "step": 10241 }, { "epoch": 0.38, "learning_rate": 4.859405937868106e-06, "loss": 0.975, "step": 10242 }, { "epoch": 0.38, "learning_rate": 4.857556942187283e-06, "loss": 0.9468, "step": 10243 }, { "epoch": 0.38, "learning_rate": 4.855708185491825e-06, "loss": 0.9935, "step": 10244 }, { "epoch": 0.38, "learning_rate": 4.853859667867641e-06, "loss": 0.9659, "step": 10245 }, { "epoch": 0.38, "learning_rate": 4.852011389400653e-06, "loss": 0.9395, "step": 10246 }, { "epoch": 0.38, "learning_rate": 4.850163350176745e-06, "loss": 0.9273, "step": 10247 }, { "epoch": 0.38, "learning_rate": 4.848315550281805e-06, "loss": 0.931, "step": 10248 }, { "epoch": 0.38, "learning_rate": 4.846467989801706e-06, "loss": 0.9391, "step": 10249 }, { "epoch": 0.38, "learning_rate": 4.844620668822312e-06, "loss": 0.909, "step": 10250 }, { "epoch": 0.38, "learning_rate": 4.842773587429469e-06, "loss": 0.927, "step": 10251 }, { "epoch": 0.38, "learning_rate": 4.840926745709018e-06, "loss": 0.9182, "step": 10252 }, { "epoch": 0.38, "learning_rate": 4.839080143746793e-06, "loss": 0.9637, "step": 10253 }, { "epoch": 0.38, "learning_rate": 4.837233781628599e-06, "loss": 0.9587, "step": 10254 }, { "epoch": 0.38, "learning_rate": 4.835387659440257e-06, "loss": 0.9591, "step": 10255 }, { "epoch": 0.38, "learning_rate": 4.833541777267552e-06, "loss": 0.899, "step": 10256 }, { "epoch": 0.38, "learning_rate": 4.831696135196271e-06, "loss": 0.9586, "step": 10257 }, { "epoch": 0.38, "learning_rate": 4.829850733312187e-06, "loss": 0.9549, "step": 10258 }, { "epoch": 0.38, "learning_rate": 4.828005571701064e-06, "loss": 0.9697, "step": 10259 }, { "epoch": 0.38, "learning_rate": 4.826160650448647e-06, "loss": 0.9542, "step": 10260 }, { "epoch": 0.38, "learning_rate": 4.824315969640678e-06, "loss": 0.9404, "step": 10261 }, { "epoch": 0.38, "learning_rate": 4.822471529362889e-06, "loss": 0.9146, "step": 10262 }, { "epoch": 0.38, "learning_rate": 4.820627329700989e-06, "loss": 0.9598, "step": 10263 }, { "epoch": 0.38, "learning_rate": 4.818783370740687e-06, "loss": 0.9372, "step": 10264 }, { "epoch": 0.38, "learning_rate": 4.81693965256768e-06, "loss": 0.9662, "step": 10265 }, { "epoch": 0.38, "learning_rate": 4.815096175267647e-06, "loss": 0.9834, "step": 10266 }, { "epoch": 0.38, "learning_rate": 4.813252938926267e-06, "loss": 1.0068, "step": 10267 }, { "epoch": 0.38, "learning_rate": 4.811409943629194e-06, "loss": 0.9254, "step": 10268 }, { "epoch": 0.38, "learning_rate": 4.80956718946208e-06, "loss": 0.9674, "step": 10269 }, { "epoch": 0.38, "learning_rate": 4.807724676510563e-06, "loss": 0.9178, "step": 10270 }, { "epoch": 0.38, "learning_rate": 4.805882404860276e-06, "loss": 0.9743, "step": 10271 }, { "epoch": 0.38, "learning_rate": 4.804040374596826e-06, "loss": 0.9224, "step": 10272 }, { "epoch": 0.38, "learning_rate": 4.802198585805823e-06, "loss": 0.9615, "step": 10273 }, { "epoch": 0.38, "learning_rate": 4.800357038572863e-06, "loss": 0.9169, "step": 10274 }, { "epoch": 0.38, "learning_rate": 4.798515732983522e-06, "loss": 0.926, "step": 10275 }, { "epoch": 0.38, "learning_rate": 4.796674669123373e-06, "loss": 0.9756, "step": 10276 }, { "epoch": 0.38, "learning_rate": 4.7948338470779794e-06, "loss": 0.9563, "step": 10277 }, { "epoch": 0.38, "learning_rate": 4.792993266932887e-06, "loss": 0.9147, "step": 10278 }, { "epoch": 0.38, "learning_rate": 4.7911529287736345e-06, "loss": 0.9952, "step": 10279 }, { "epoch": 0.38, "learning_rate": 4.789312832685752e-06, "loss": 0.9409, "step": 10280 }, { "epoch": 0.38, "learning_rate": 4.7874729787547465e-06, "loss": 0.9558, "step": 10281 }, { "epoch": 0.38, "learning_rate": 4.785633367066126e-06, "loss": 0.9507, "step": 10282 }, { "epoch": 0.38, "learning_rate": 4.783793997705386e-06, "loss": 0.9485, "step": 10283 }, { "epoch": 0.38, "learning_rate": 4.781954870758001e-06, "loss": 0.9031, "step": 10284 }, { "epoch": 0.38, "learning_rate": 4.780115986309445e-06, "loss": 0.9183, "step": 10285 }, { "epoch": 0.38, "learning_rate": 4.7782773444451756e-06, "loss": 0.9545, "step": 10286 }, { "epoch": 0.38, "learning_rate": 4.7764389452506445e-06, "loss": 0.9322, "step": 10287 }, { "epoch": 0.38, "learning_rate": 4.774600788811277e-06, "loss": 0.9609, "step": 10288 }, { "epoch": 0.38, "learning_rate": 4.772762875212511e-06, "loss": 0.9699, "step": 10289 }, { "epoch": 0.38, "learning_rate": 4.7709252045397555e-06, "loss": 0.9714, "step": 10290 }, { "epoch": 0.38, "learning_rate": 4.769087776878404e-06, "loss": 0.9383, "step": 10291 }, { "epoch": 0.38, "learning_rate": 4.7672505923138626e-06, "loss": 0.8988, "step": 10292 }, { "epoch": 0.38, "learning_rate": 4.7654136509314985e-06, "loss": 0.928, "step": 10293 }, { "epoch": 0.38, "learning_rate": 4.763576952816685e-06, "loss": 0.9386, "step": 10294 }, { "epoch": 0.38, "learning_rate": 4.76174049805478e-06, "loss": 0.9759, "step": 10295 }, { "epoch": 0.38, "learning_rate": 4.759904286731132e-06, "loss": 0.9344, "step": 10296 }, { "epoch": 0.39, "learning_rate": 4.758068318931067e-06, "loss": 0.9101, "step": 10297 }, { "epoch": 0.39, "learning_rate": 4.7562325947399135e-06, "loss": 0.9308, "step": 10298 }, { "epoch": 0.39, "learning_rate": 4.754397114242987e-06, "loss": 0.922, "step": 10299 }, { "epoch": 0.39, "learning_rate": 4.752561877525575e-06, "loss": 0.8913, "step": 10300 }, { "epoch": 0.39, "learning_rate": 4.750726884672984e-06, "loss": 0.9235, "step": 10301 }, { "epoch": 0.39, "learning_rate": 4.74889213577048e-06, "loss": 0.9719, "step": 10302 }, { "epoch": 0.39, "learning_rate": 4.747057630903332e-06, "loss": 0.9295, "step": 10303 }, { "epoch": 0.39, "learning_rate": 4.745223370156797e-06, "loss": 0.9413, "step": 10304 }, { "epoch": 0.39, "learning_rate": 4.743389353616122e-06, "loss": 0.9117, "step": 10305 }, { "epoch": 0.39, "learning_rate": 4.741555581366529e-06, "loss": 0.9339, "step": 10306 }, { "epoch": 0.39, "learning_rate": 4.739722053493248e-06, "loss": 0.9487, "step": 10307 }, { "epoch": 0.39, "learning_rate": 4.737888770081488e-06, "loss": 0.9226, "step": 10308 }, { "epoch": 0.39, "learning_rate": 4.736055731216439e-06, "loss": 0.9767, "step": 10309 }, { "epoch": 0.39, "learning_rate": 4.734222936983302e-06, "loss": 0.9758, "step": 10310 }, { "epoch": 0.39, "learning_rate": 4.732390387467241e-06, "loss": 0.9254, "step": 10311 }, { "epoch": 0.39, "learning_rate": 4.730558082753425e-06, "loss": 0.9135, "step": 10312 }, { "epoch": 0.39, "learning_rate": 4.72872602292701e-06, "loss": 0.9187, "step": 10313 }, { "epoch": 0.39, "learning_rate": 4.726894208073129e-06, "loss": 0.9369, "step": 10314 }, { "epoch": 0.39, "learning_rate": 4.72506263827692e-06, "loss": 0.9282, "step": 10315 }, { "epoch": 0.39, "learning_rate": 4.723231313623496e-06, "loss": 0.9459, "step": 10316 }, { "epoch": 0.39, "learning_rate": 4.721400234197973e-06, "loss": 0.9142, "step": 10317 }, { "epoch": 0.39, "learning_rate": 4.7195694000854365e-06, "loss": 0.9418, "step": 10318 }, { "epoch": 0.39, "learning_rate": 4.717738811370975e-06, "loss": 0.9715, "step": 10319 }, { "epoch": 0.39, "learning_rate": 4.715908468139668e-06, "loss": 0.9504, "step": 10320 }, { "epoch": 0.39, "learning_rate": 4.7140783704765666e-06, "loss": 0.9576, "step": 10321 }, { "epoch": 0.39, "learning_rate": 4.712248518466727e-06, "loss": 0.9256, "step": 10322 }, { "epoch": 0.39, "learning_rate": 4.710418912195187e-06, "loss": 0.9357, "step": 10323 }, { "epoch": 0.39, "learning_rate": 4.708589551746975e-06, "loss": 0.9344, "step": 10324 }, { "epoch": 0.39, "learning_rate": 4.706760437207106e-06, "loss": 0.9616, "step": 10325 }, { "epoch": 0.39, "learning_rate": 4.704931568660589e-06, "loss": 0.9097, "step": 10326 }, { "epoch": 0.39, "learning_rate": 4.7031029461924105e-06, "loss": 0.9409, "step": 10327 }, { "epoch": 0.39, "learning_rate": 4.701274569887554e-06, "loss": 0.9283, "step": 10328 }, { "epoch": 0.39, "learning_rate": 4.699446439830996e-06, "loss": 0.9572, "step": 10329 }, { "epoch": 0.39, "learning_rate": 4.6976185561076855e-06, "loss": 0.9688, "step": 10330 }, { "epoch": 0.39, "learning_rate": 4.695790918802577e-06, "loss": 0.9462, "step": 10331 }, { "epoch": 0.39, "learning_rate": 4.693963528000602e-06, "loss": 0.9945, "step": 10332 }, { "epoch": 0.39, "learning_rate": 4.692136383786693e-06, "loss": 0.9289, "step": 10333 }, { "epoch": 0.39, "learning_rate": 4.69030948624575e-06, "loss": 0.9495, "step": 10334 }, { "epoch": 0.39, "learning_rate": 4.688482835462689e-06, "loss": 0.9522, "step": 10335 }, { "epoch": 0.39, "learning_rate": 4.686656431522393e-06, "loss": 0.9823, "step": 10336 }, { "epoch": 0.39, "learning_rate": 4.684830274509734e-06, "loss": 0.9786, "step": 10337 }, { "epoch": 0.39, "learning_rate": 4.6830043645095945e-06, "loss": 0.9383, "step": 10338 }, { "epoch": 0.39, "learning_rate": 4.681178701606818e-06, "loss": 0.9403, "step": 10339 }, { "epoch": 0.39, "learning_rate": 4.6793532858862526e-06, "loss": 0.9681, "step": 10340 }, { "epoch": 0.39, "learning_rate": 4.6775281174327304e-06, "loss": 0.9868, "step": 10341 }, { "epoch": 0.39, "learning_rate": 4.675703196331078e-06, "loss": 0.9474, "step": 10342 }, { "epoch": 0.39, "learning_rate": 4.673878522666097e-06, "loss": 0.9718, "step": 10343 }, { "epoch": 0.39, "learning_rate": 4.672054096522589e-06, "loss": 0.9653, "step": 10344 }, { "epoch": 0.39, "learning_rate": 4.670229917985345e-06, "loss": 0.9458, "step": 10345 }, { "epoch": 0.39, "learning_rate": 4.668405987139129e-06, "loss": 0.9252, "step": 10346 }, { "epoch": 0.39, "learning_rate": 4.6665823040687195e-06, "loss": 0.9603, "step": 10347 }, { "epoch": 0.39, "learning_rate": 4.664758868858858e-06, "loss": 0.959, "step": 10348 }, { "epoch": 0.39, "learning_rate": 4.6629356815942885e-06, "loss": 0.9293, "step": 10349 }, { "epoch": 0.39, "learning_rate": 4.661112742359739e-06, "loss": 0.9728, "step": 10350 }, { "epoch": 0.39, "learning_rate": 4.659290051239934e-06, "loss": 0.9413, "step": 10351 }, { "epoch": 0.39, "learning_rate": 4.657467608319569e-06, "loss": 0.9068, "step": 10352 }, { "epoch": 0.39, "learning_rate": 4.655645413683345e-06, "loss": 0.9599, "step": 10353 }, { "epoch": 0.39, "learning_rate": 4.6538234674159465e-06, "loss": 0.9064, "step": 10354 }, { "epoch": 0.39, "learning_rate": 4.652001769602035e-06, "loss": 0.9348, "step": 10355 }, { "epoch": 0.39, "learning_rate": 4.650180320326285e-06, "loss": 0.9389, "step": 10356 }, { "epoch": 0.39, "learning_rate": 4.648359119673334e-06, "loss": 0.9698, "step": 10357 }, { "epoch": 0.39, "learning_rate": 4.646538167727823e-06, "loss": 0.9155, "step": 10358 }, { "epoch": 0.39, "learning_rate": 4.644717464574381e-06, "loss": 0.9594, "step": 10359 }, { "epoch": 0.39, "learning_rate": 4.6428970102976135e-06, "loss": 0.9586, "step": 10360 }, { "epoch": 0.39, "learning_rate": 4.641076804982127e-06, "loss": 0.9117, "step": 10361 }, { "epoch": 0.39, "learning_rate": 4.639256848712511e-06, "loss": 0.8776, "step": 10362 }, { "epoch": 0.39, "learning_rate": 4.63743714157335e-06, "loss": 0.9466, "step": 10363 }, { "epoch": 0.39, "learning_rate": 4.6356176836492025e-06, "loss": 0.9266, "step": 10364 }, { "epoch": 0.39, "learning_rate": 4.63379847502463e-06, "loss": 0.9635, "step": 10365 }, { "epoch": 0.39, "learning_rate": 4.631979515784178e-06, "loss": 0.9395, "step": 10366 }, { "epoch": 0.39, "learning_rate": 4.6301608060123706e-06, "loss": 0.942, "step": 10367 }, { "epoch": 0.39, "learning_rate": 4.628342345793743e-06, "loss": 0.9574, "step": 10368 }, { "epoch": 0.39, "learning_rate": 4.626524135212792e-06, "loss": 0.9293, "step": 10369 }, { "epoch": 0.39, "learning_rate": 4.62470617435402e-06, "loss": 0.9053, "step": 10370 }, { "epoch": 0.39, "learning_rate": 4.622888463301914e-06, "loss": 0.9559, "step": 10371 }, { "epoch": 0.39, "learning_rate": 4.621071002140953e-06, "loss": 0.946, "step": 10372 }, { "epoch": 0.39, "learning_rate": 4.6192537909555915e-06, "loss": 0.9285, "step": 10373 }, { "epoch": 0.39, "learning_rate": 4.617436829830285e-06, "loss": 0.953, "step": 10374 }, { "epoch": 0.39, "learning_rate": 4.6156201188494765e-06, "loss": 0.9268, "step": 10375 }, { "epoch": 0.39, "learning_rate": 4.613803658097588e-06, "loss": 1.0048, "step": 10376 }, { "epoch": 0.39, "learning_rate": 4.611987447659039e-06, "loss": 0.9634, "step": 10377 }, { "epoch": 0.39, "learning_rate": 4.610171487618233e-06, "loss": 0.8927, "step": 10378 }, { "epoch": 0.39, "learning_rate": 4.60835577805957e-06, "loss": 0.9417, "step": 10379 }, { "epoch": 0.39, "learning_rate": 4.60654031906742e-06, "loss": 0.9137, "step": 10380 }, { "epoch": 0.39, "learning_rate": 4.6047251107261655e-06, "loss": 0.9568, "step": 10381 }, { "epoch": 0.39, "learning_rate": 4.602910153120156e-06, "loss": 0.9049, "step": 10382 }, { "epoch": 0.39, "learning_rate": 4.601095446333741e-06, "loss": 0.9606, "step": 10383 }, { "epoch": 0.39, "learning_rate": 4.59928099045126e-06, "loss": 0.9383, "step": 10384 }, { "epoch": 0.39, "learning_rate": 4.597466785557029e-06, "loss": 0.9268, "step": 10385 }, { "epoch": 0.39, "learning_rate": 4.595652831735362e-06, "loss": 0.9343, "step": 10386 }, { "epoch": 0.39, "learning_rate": 4.5938391290705595e-06, "loss": 0.9266, "step": 10387 }, { "epoch": 0.39, "learning_rate": 4.5920256776469164e-06, "loss": 0.9213, "step": 10388 }, { "epoch": 0.39, "learning_rate": 4.590212477548698e-06, "loss": 0.9536, "step": 10389 }, { "epoch": 0.39, "learning_rate": 4.5883995288601755e-06, "loss": 0.939, "step": 10390 }, { "epoch": 0.39, "learning_rate": 4.586586831665606e-06, "loss": 0.9542, "step": 10391 }, { "epoch": 0.39, "learning_rate": 4.584774386049219e-06, "loss": 0.9467, "step": 10392 }, { "epoch": 0.39, "learning_rate": 4.582962192095259e-06, "loss": 0.938, "step": 10393 }, { "epoch": 0.39, "learning_rate": 4.581150249887936e-06, "loss": 0.9418, "step": 10394 }, { "epoch": 0.39, "learning_rate": 4.579338559511457e-06, "loss": 0.9492, "step": 10395 }, { "epoch": 0.39, "learning_rate": 4.577527121050018e-06, "loss": 0.9986, "step": 10396 }, { "epoch": 0.39, "learning_rate": 4.575715934587807e-06, "loss": 0.9683, "step": 10397 }, { "epoch": 0.39, "learning_rate": 4.573905000208987e-06, "loss": 0.9555, "step": 10398 }, { "epoch": 0.39, "learning_rate": 4.572094317997722e-06, "loss": 0.983, "step": 10399 }, { "epoch": 0.39, "learning_rate": 4.5702838880381615e-06, "loss": 0.9574, "step": 10400 }, { "epoch": 0.39, "learning_rate": 4.568473710414435e-06, "loss": 0.9651, "step": 10401 }, { "epoch": 0.39, "learning_rate": 4.5666637852106775e-06, "loss": 0.9628, "step": 10402 }, { "epoch": 0.39, "learning_rate": 4.564854112510994e-06, "loss": 0.9413, "step": 10403 }, { "epoch": 0.39, "learning_rate": 4.563044692399489e-06, "loss": 0.9196, "step": 10404 }, { "epoch": 0.39, "learning_rate": 4.5612355249602535e-06, "loss": 0.8988, "step": 10405 }, { "epoch": 0.39, "learning_rate": 4.559426610277359e-06, "loss": 0.9394, "step": 10406 }, { "epoch": 0.39, "learning_rate": 4.557617948434875e-06, "loss": 0.9507, "step": 10407 }, { "epoch": 0.39, "learning_rate": 4.555809539516857e-06, "loss": 0.902, "step": 10408 }, { "epoch": 0.39, "learning_rate": 4.5540013836073484e-06, "loss": 0.9572, "step": 10409 }, { "epoch": 0.39, "learning_rate": 4.552193480790374e-06, "loss": 0.9666, "step": 10410 }, { "epoch": 0.39, "learning_rate": 4.5503858311499575e-06, "loss": 0.9164, "step": 10411 }, { "epoch": 0.39, "learning_rate": 4.548578434770108e-06, "loss": 0.9267, "step": 10412 }, { "epoch": 0.39, "learning_rate": 4.546771291734809e-06, "loss": 0.9152, "step": 10413 }, { "epoch": 0.39, "learning_rate": 4.544964402128061e-06, "loss": 0.9585, "step": 10414 }, { "epoch": 0.39, "learning_rate": 4.543157766033824e-06, "loss": 0.95, "step": 10415 }, { "epoch": 0.39, "learning_rate": 4.5413513835360615e-06, "loss": 0.9384, "step": 10416 }, { "epoch": 0.39, "learning_rate": 4.539545254718721e-06, "loss": 0.9475, "step": 10417 }, { "epoch": 0.39, "learning_rate": 4.537739379665745e-06, "loss": 0.9452, "step": 10418 }, { "epoch": 0.39, "learning_rate": 4.5359337584610476e-06, "loss": 0.9407, "step": 10419 }, { "epoch": 0.39, "learning_rate": 4.534128391188548e-06, "loss": 0.9217, "step": 10420 }, { "epoch": 0.39, "learning_rate": 4.532323277932149e-06, "loss": 0.9592, "step": 10421 }, { "epoch": 0.39, "learning_rate": 4.530518418775734e-06, "loss": 0.954, "step": 10422 }, { "epoch": 0.39, "learning_rate": 4.5287138138031824e-06, "loss": 0.9105, "step": 10423 }, { "epoch": 0.39, "learning_rate": 4.526909463098362e-06, "loss": 0.9082, "step": 10424 }, { "epoch": 0.39, "learning_rate": 4.525105366745128e-06, "loss": 0.9145, "step": 10425 }, { "epoch": 0.39, "learning_rate": 4.523301524827313e-06, "loss": 0.9329, "step": 10426 }, { "epoch": 0.39, "learning_rate": 4.521497937428761e-06, "loss": 0.9126, "step": 10427 }, { "epoch": 0.39, "learning_rate": 4.51969460463328e-06, "loss": 0.9088, "step": 10428 }, { "epoch": 0.39, "learning_rate": 4.517891526524679e-06, "loss": 0.9542, "step": 10429 }, { "epoch": 0.39, "learning_rate": 4.516088703186758e-06, "loss": 0.9269, "step": 10430 }, { "epoch": 0.39, "learning_rate": 4.5142861347032915e-06, "loss": 0.9526, "step": 10431 }, { "epoch": 0.39, "learning_rate": 4.512483821158055e-06, "loss": 0.9208, "step": 10432 }, { "epoch": 0.39, "learning_rate": 4.510681762634805e-06, "loss": 0.9591, "step": 10433 }, { "epoch": 0.39, "learning_rate": 4.508879959217295e-06, "loss": 0.9141, "step": 10434 }, { "epoch": 0.39, "learning_rate": 4.507078410989253e-06, "loss": 0.9123, "step": 10435 }, { "epoch": 0.39, "learning_rate": 4.505277118034406e-06, "loss": 0.9633, "step": 10436 }, { "epoch": 0.39, "learning_rate": 4.503476080436468e-06, "loss": 0.9787, "step": 10437 }, { "epoch": 0.39, "learning_rate": 4.501675298279129e-06, "loss": 0.9308, "step": 10438 }, { "epoch": 0.39, "learning_rate": 4.499874771646092e-06, "loss": 0.9837, "step": 10439 }, { "epoch": 0.39, "learning_rate": 4.498074500621021e-06, "loss": 0.9505, "step": 10440 }, { "epoch": 0.39, "learning_rate": 4.496274485287585e-06, "loss": 0.9333, "step": 10441 }, { "epoch": 0.39, "learning_rate": 4.4944747257294356e-06, "loss": 0.954, "step": 10442 }, { "epoch": 0.39, "learning_rate": 4.492675222030217e-06, "loss": 0.9301, "step": 10443 }, { "epoch": 0.39, "learning_rate": 4.490875974273551e-06, "loss": 0.9005, "step": 10444 }, { "epoch": 0.39, "learning_rate": 4.489076982543057e-06, "loss": 0.9349, "step": 10445 }, { "epoch": 0.39, "learning_rate": 4.487278246922343e-06, "loss": 0.9471, "step": 10446 }, { "epoch": 0.4, "learning_rate": 4.485479767494992e-06, "loss": 0.952, "step": 10447 }, { "epoch": 0.4, "learning_rate": 4.4836815443446e-06, "loss": 0.9453, "step": 10448 }, { "epoch": 0.4, "learning_rate": 4.481883577554724e-06, "loss": 0.9337, "step": 10449 }, { "epoch": 0.4, "learning_rate": 4.480085867208927e-06, "loss": 0.9312, "step": 10450 }, { "epoch": 0.4, "learning_rate": 4.478288413390753e-06, "loss": 0.9358, "step": 10451 }, { "epoch": 0.4, "learning_rate": 4.476491216183733e-06, "loss": 0.9372, "step": 10452 }, { "epoch": 0.4, "learning_rate": 4.474694275671388e-06, "loss": 0.903, "step": 10453 }, { "epoch": 0.4, "learning_rate": 4.472897591937231e-06, "loss": 0.9374, "step": 10454 }, { "epoch": 0.4, "learning_rate": 4.471101165064761e-06, "loss": 0.9159, "step": 10455 }, { "epoch": 0.4, "learning_rate": 4.469304995137457e-06, "loss": 0.9666, "step": 10456 }, { "epoch": 0.4, "learning_rate": 4.467509082238795e-06, "loss": 0.9412, "step": 10457 }, { "epoch": 0.4, "learning_rate": 4.465713426452244e-06, "loss": 0.9713, "step": 10458 }, { "epoch": 0.4, "learning_rate": 4.4639180278612385e-06, "loss": 0.9192, "step": 10459 }, { "epoch": 0.4, "learning_rate": 4.462122886549232e-06, "loss": 0.9296, "step": 10460 }, { "epoch": 0.4, "learning_rate": 4.460328002599641e-06, "loss": 0.9033, "step": 10461 }, { "epoch": 0.4, "learning_rate": 4.4585333760958804e-06, "loss": 0.9461, "step": 10462 }, { "epoch": 0.4, "learning_rate": 4.456739007121356e-06, "loss": 0.947, "step": 10463 }, { "epoch": 0.4, "learning_rate": 4.454944895759457e-06, "loss": 0.9722, "step": 10464 }, { "epoch": 0.4, "learning_rate": 4.4531510420935565e-06, "loss": 0.9735, "step": 10465 }, { "epoch": 0.4, "learning_rate": 4.4513574462070245e-06, "loss": 0.9625, "step": 10466 }, { "epoch": 0.4, "learning_rate": 4.449564108183216e-06, "loss": 0.9434, "step": 10467 }, { "epoch": 0.4, "learning_rate": 4.447771028105469e-06, "loss": 0.9382, "step": 10468 }, { "epoch": 0.4, "learning_rate": 4.445978206057116e-06, "loss": 0.978, "step": 10469 }, { "epoch": 0.4, "learning_rate": 4.444185642121472e-06, "loss": 0.9462, "step": 10470 }, { "epoch": 0.4, "learning_rate": 4.442393336381848e-06, "loss": 0.9644, "step": 10471 }, { "epoch": 0.4, "learning_rate": 4.440601288921535e-06, "loss": 0.9432, "step": 10472 }, { "epoch": 0.4, "learning_rate": 4.4388094998238204e-06, "loss": 0.9034, "step": 10473 }, { "epoch": 0.4, "learning_rate": 4.437017969171966e-06, "loss": 0.9137, "step": 10474 }, { "epoch": 0.4, "learning_rate": 4.435226697049233e-06, "loss": 0.9433, "step": 10475 }, { "epoch": 0.4, "learning_rate": 4.433435683538871e-06, "loss": 0.9207, "step": 10476 }, { "epoch": 0.4, "learning_rate": 4.4316449287241085e-06, "loss": 0.9488, "step": 10477 }, { "epoch": 0.4, "learning_rate": 4.429854432688169e-06, "loss": 0.9261, "step": 10478 }, { "epoch": 0.4, "learning_rate": 4.428064195514262e-06, "loss": 0.8944, "step": 10479 }, { "epoch": 0.4, "learning_rate": 4.426274217285592e-06, "loss": 0.9213, "step": 10480 }, { "epoch": 0.4, "learning_rate": 4.424484498085335e-06, "loss": 0.9252, "step": 10481 }, { "epoch": 0.4, "learning_rate": 4.4226950379966695e-06, "loss": 0.9305, "step": 10482 }, { "epoch": 0.4, "learning_rate": 4.420905837102761e-06, "loss": 0.9247, "step": 10483 }, { "epoch": 0.4, "learning_rate": 4.419116895486748e-06, "loss": 0.8939, "step": 10484 }, { "epoch": 0.4, "learning_rate": 4.417328213231783e-06, "loss": 0.9719, "step": 10485 }, { "epoch": 0.4, "learning_rate": 4.4155397904209805e-06, "loss": 0.9684, "step": 10486 }, { "epoch": 0.4, "learning_rate": 4.413751627137458e-06, "loss": 0.9829, "step": 10487 }, { "epoch": 0.4, "learning_rate": 4.4119637234643164e-06, "loss": 0.9528, "step": 10488 }, { "epoch": 0.4, "learning_rate": 4.410176079484649e-06, "loss": 0.9551, "step": 10489 }, { "epoch": 0.4, "learning_rate": 4.408388695281527e-06, "loss": 0.957, "step": 10490 }, { "epoch": 0.4, "learning_rate": 4.406601570938017e-06, "loss": 0.8915, "step": 10491 }, { "epoch": 0.4, "learning_rate": 4.4048147065371785e-06, "loss": 0.9379, "step": 10492 }, { "epoch": 0.4, "learning_rate": 4.4030281021620405e-06, "loss": 0.9459, "step": 10493 }, { "epoch": 0.4, "learning_rate": 4.401241757895647e-06, "loss": 0.9607, "step": 10494 }, { "epoch": 0.4, "learning_rate": 4.399455673821004e-06, "loss": 0.8999, "step": 10495 }, { "epoch": 0.4, "learning_rate": 4.397669850021119e-06, "loss": 0.9094, "step": 10496 }, { "epoch": 0.4, "learning_rate": 4.39588428657899e-06, "loss": 0.9613, "step": 10497 }, { "epoch": 0.4, "learning_rate": 4.394098983577589e-06, "loss": 0.9439, "step": 10498 }, { "epoch": 0.4, "learning_rate": 4.392313941099888e-06, "loss": 0.9491, "step": 10499 }, { "epoch": 0.4, "learning_rate": 4.3905291592288464e-06, "loss": 0.9648, "step": 10500 }, { "epoch": 0.4, "learning_rate": 4.38874463804741e-06, "loss": 0.9219, "step": 10501 }, { "epoch": 0.4, "learning_rate": 4.3869603776385e-06, "loss": 0.9282, "step": 10502 }, { "epoch": 0.4, "learning_rate": 4.385176378085051e-06, "loss": 0.9743, "step": 10503 }, { "epoch": 0.4, "learning_rate": 4.383392639469965e-06, "loss": 0.9559, "step": 10504 }, { "epoch": 0.4, "learning_rate": 4.38160916187613e-06, "loss": 0.9678, "step": 10505 }, { "epoch": 0.4, "learning_rate": 4.379825945386443e-06, "loss": 0.9567, "step": 10506 }, { "epoch": 0.4, "learning_rate": 4.378042990083766e-06, "loss": 0.9056, "step": 10507 }, { "epoch": 0.4, "learning_rate": 4.3762602960509625e-06, "loss": 0.9384, "step": 10508 }, { "epoch": 0.4, "learning_rate": 4.374477863370881e-06, "loss": 0.9605, "step": 10509 }, { "epoch": 0.4, "learning_rate": 4.372695692126357e-06, "loss": 0.9531, "step": 10510 }, { "epoch": 0.4, "learning_rate": 4.370913782400208e-06, "loss": 0.919, "step": 10511 }, { "epoch": 0.4, "learning_rate": 4.369132134275248e-06, "loss": 0.923, "step": 10512 }, { "epoch": 0.4, "learning_rate": 4.367350747834282e-06, "loss": 0.9428, "step": 10513 }, { "epoch": 0.4, "learning_rate": 4.365569623160086e-06, "loss": 0.9228, "step": 10514 }, { "epoch": 0.4, "learning_rate": 4.363788760335438e-06, "loss": 0.9107, "step": 10515 }, { "epoch": 0.4, "learning_rate": 4.362008159443103e-06, "loss": 0.9293, "step": 10516 }, { "epoch": 0.4, "learning_rate": 4.36022782056583e-06, "loss": 0.9627, "step": 10517 }, { "epoch": 0.4, "learning_rate": 4.358447743786356e-06, "loss": 0.9545, "step": 10518 }, { "epoch": 0.4, "learning_rate": 4.3566679291874104e-06, "loss": 0.9059, "step": 10519 }, { "epoch": 0.4, "learning_rate": 4.3548883768517e-06, "loss": 0.9251, "step": 10520 }, { "epoch": 0.4, "learning_rate": 4.35310908686193e-06, "loss": 0.957, "step": 10521 }, { "epoch": 0.4, "learning_rate": 4.351330059300793e-06, "loss": 0.9288, "step": 10522 }, { "epoch": 0.4, "learning_rate": 4.3495512942509585e-06, "loss": 0.9464, "step": 10523 }, { "epoch": 0.4, "learning_rate": 4.347772791795095e-06, "loss": 0.8946, "step": 10524 }, { "epoch": 0.4, "learning_rate": 4.345994552015855e-06, "loss": 0.9692, "step": 10525 }, { "epoch": 0.4, "learning_rate": 4.344216574995882e-06, "loss": 0.9105, "step": 10526 }, { "epoch": 0.4, "learning_rate": 4.342438860817795e-06, "loss": 0.9258, "step": 10527 }, { "epoch": 0.4, "learning_rate": 4.340661409564222e-06, "loss": 0.8997, "step": 10528 }, { "epoch": 0.4, "learning_rate": 4.338884221317761e-06, "loss": 0.9578, "step": 10529 }, { "epoch": 0.4, "learning_rate": 4.337107296160996e-06, "loss": 0.9504, "step": 10530 }, { "epoch": 0.4, "learning_rate": 4.3353306341765214e-06, "loss": 0.9483, "step": 10531 }, { "epoch": 0.4, "learning_rate": 4.333554235446892e-06, "loss": 0.9218, "step": 10532 }, { "epoch": 0.4, "learning_rate": 4.331778100054665e-06, "loss": 0.9552, "step": 10533 }, { "epoch": 0.4, "learning_rate": 4.330002228082388e-06, "loss": 0.9395, "step": 10534 }, { "epoch": 0.4, "learning_rate": 4.328226619612589e-06, "loss": 0.9409, "step": 10535 }, { "epoch": 0.4, "learning_rate": 4.3264512747277835e-06, "loss": 0.9139, "step": 10536 }, { "epoch": 0.4, "learning_rate": 4.324676193510478e-06, "loss": 0.9119, "step": 10537 }, { "epoch": 0.4, "learning_rate": 4.3229013760431695e-06, "loss": 0.9718, "step": 10538 }, { "epoch": 0.4, "learning_rate": 4.32112682240833e-06, "loss": 0.9958, "step": 10539 }, { "epoch": 0.4, "learning_rate": 4.319352532688444e-06, "loss": 0.9555, "step": 10540 }, { "epoch": 0.4, "learning_rate": 4.317578506965955e-06, "loss": 0.9406, "step": 10541 }, { "epoch": 0.4, "learning_rate": 4.3158047453233125e-06, "loss": 0.9349, "step": 10542 }, { "epoch": 0.4, "learning_rate": 4.314031247842951e-06, "loss": 0.9247, "step": 10543 }, { "epoch": 0.4, "learning_rate": 4.312258014607284e-06, "loss": 0.8971, "step": 10544 }, { "epoch": 0.4, "learning_rate": 4.310485045698725e-06, "loss": 0.9256, "step": 10545 }, { "epoch": 0.4, "learning_rate": 4.308712341199666e-06, "loss": 0.9452, "step": 10546 }, { "epoch": 0.4, "learning_rate": 4.306939901192495e-06, "loss": 0.9821, "step": 10547 }, { "epoch": 0.4, "learning_rate": 4.305167725759572e-06, "loss": 0.9309, "step": 10548 }, { "epoch": 0.4, "learning_rate": 4.30339581498327e-06, "loss": 0.9488, "step": 10549 }, { "epoch": 0.4, "learning_rate": 4.301624168945925e-06, "loss": 0.9345, "step": 10550 }, { "epoch": 0.4, "learning_rate": 4.299852787729873e-06, "loss": 0.9326, "step": 10551 }, { "epoch": 0.4, "learning_rate": 4.298081671417442e-06, "loss": 0.9273, "step": 10552 }, { "epoch": 0.4, "learning_rate": 4.296310820090931e-06, "loss": 0.9302, "step": 10553 }, { "epoch": 0.4, "learning_rate": 4.294540233832642e-06, "loss": 0.9224, "step": 10554 }, { "epoch": 0.4, "learning_rate": 4.292769912724859e-06, "loss": 0.9592, "step": 10555 }, { "epoch": 0.4, "learning_rate": 4.290999856849859e-06, "loss": 0.8924, "step": 10556 }, { "epoch": 0.4, "learning_rate": 4.2892300662898944e-06, "loss": 0.9316, "step": 10557 }, { "epoch": 0.4, "learning_rate": 4.287460541127217e-06, "loss": 0.898, "step": 10558 }, { "epoch": 0.4, "learning_rate": 4.285691281444063e-06, "loss": 0.8901, "step": 10559 }, { "epoch": 0.4, "learning_rate": 4.283922287322652e-06, "loss": 0.9306, "step": 10560 }, { "epoch": 0.4, "learning_rate": 4.2821535588451955e-06, "loss": 0.9184, "step": 10561 }, { "epoch": 0.4, "learning_rate": 4.2803850960938935e-06, "loss": 0.9327, "step": 10562 }, { "epoch": 0.4, "learning_rate": 4.2786168991509315e-06, "loss": 0.9742, "step": 10563 }, { "epoch": 0.4, "learning_rate": 4.276848968098483e-06, "loss": 0.9642, "step": 10564 }, { "epoch": 0.4, "learning_rate": 4.275081303018712e-06, "loss": 0.937, "step": 10565 }, { "epoch": 0.4, "learning_rate": 4.273313903993762e-06, "loss": 0.9826, "step": 10566 }, { "epoch": 0.4, "learning_rate": 4.2715467711057724e-06, "loss": 0.9444, "step": 10567 }, { "epoch": 0.4, "learning_rate": 4.26977990443687e-06, "loss": 0.9296, "step": 10568 }, { "epoch": 0.4, "learning_rate": 4.268013304069161e-06, "loss": 0.9278, "step": 10569 }, { "epoch": 0.4, "learning_rate": 4.266246970084747e-06, "loss": 0.9018, "step": 10570 }, { "epoch": 0.4, "learning_rate": 4.264480902565716e-06, "loss": 0.8937, "step": 10571 }, { "epoch": 0.4, "learning_rate": 4.262715101594146e-06, "loss": 0.9153, "step": 10572 }, { "epoch": 0.4, "learning_rate": 4.260949567252088e-06, "loss": 0.9347, "step": 10573 }, { "epoch": 0.4, "learning_rate": 4.259184299621608e-06, "loss": 0.9513, "step": 10574 }, { "epoch": 0.4, "learning_rate": 4.257419298784731e-06, "loss": 0.9327, "step": 10575 }, { "epoch": 0.4, "learning_rate": 4.255654564823485e-06, "loss": 0.9469, "step": 10576 }, { "epoch": 0.4, "learning_rate": 4.253890097819888e-06, "loss": 0.955, "step": 10577 }, { "epoch": 0.4, "learning_rate": 4.2521258978559324e-06, "loss": 0.9264, "step": 10578 }, { "epoch": 0.4, "learning_rate": 4.250361965013609e-06, "loss": 0.9342, "step": 10579 }, { "epoch": 0.4, "learning_rate": 4.248598299374895e-06, "loss": 0.9341, "step": 10580 }, { "epoch": 0.4, "learning_rate": 4.246834901021757e-06, "loss": 0.9619, "step": 10581 }, { "epoch": 0.4, "learning_rate": 4.245071770036136e-06, "loss": 0.976, "step": 10582 }, { "epoch": 0.4, "learning_rate": 4.243308906499975e-06, "loss": 0.9196, "step": 10583 }, { "epoch": 0.4, "learning_rate": 4.241546310495205e-06, "loss": 0.94, "step": 10584 }, { "epoch": 0.4, "learning_rate": 4.239783982103727e-06, "loss": 0.9539, "step": 10585 }, { "epoch": 0.4, "learning_rate": 4.238021921407456e-06, "loss": 0.911, "step": 10586 }, { "epoch": 0.4, "learning_rate": 4.23626012848827e-06, "loss": 0.9364, "step": 10587 }, { "epoch": 0.4, "learning_rate": 4.234498603428048e-06, "loss": 0.9157, "step": 10588 }, { "epoch": 0.4, "learning_rate": 4.232737346308659e-06, "loss": 0.9459, "step": 10589 }, { "epoch": 0.4, "learning_rate": 4.230976357211946e-06, "loss": 0.9627, "step": 10590 }, { "epoch": 0.4, "learning_rate": 4.2292156362197505e-06, "loss": 0.9911, "step": 10591 }, { "epoch": 0.4, "learning_rate": 4.227455183413899e-06, "loss": 0.9367, "step": 10592 }, { "epoch": 0.4, "learning_rate": 4.225694998876209e-06, "loss": 0.9651, "step": 10593 }, { "epoch": 0.4, "learning_rate": 4.22393508268847e-06, "loss": 0.927, "step": 10594 }, { "epoch": 0.4, "learning_rate": 4.222175434932488e-06, "loss": 0.9473, "step": 10595 }, { "epoch": 0.4, "learning_rate": 4.220416055690027e-06, "loss": 0.9436, "step": 10596 }, { "epoch": 0.41, "learning_rate": 4.218656945042854e-06, "loss": 0.9464, "step": 10597 }, { "epoch": 0.41, "learning_rate": 4.216898103072723e-06, "loss": 0.9644, "step": 10598 }, { "epoch": 0.41, "learning_rate": 4.2151395298613675e-06, "loss": 0.9318, "step": 10599 }, { "epoch": 0.41, "learning_rate": 4.213381225490517e-06, "loss": 0.9288, "step": 10600 }, { "epoch": 0.41, "learning_rate": 4.211623190041885e-06, "loss": 0.9651, "step": 10601 }, { "epoch": 0.41, "learning_rate": 4.209865423597177e-06, "loss": 0.9699, "step": 10602 }, { "epoch": 0.41, "learning_rate": 4.2081079262380744e-06, "loss": 0.9248, "step": 10603 }, { "epoch": 0.41, "learning_rate": 4.206350698046256e-06, "loss": 0.9502, "step": 10604 }, { "epoch": 0.41, "learning_rate": 4.204593739103391e-06, "loss": 0.9489, "step": 10605 }, { "epoch": 0.41, "learning_rate": 4.20283704949112e-06, "loss": 0.9351, "step": 10606 }, { "epoch": 0.41, "learning_rate": 4.201080629291096e-06, "loss": 0.9143, "step": 10607 }, { "epoch": 0.41, "learning_rate": 4.199324478584933e-06, "loss": 0.9118, "step": 10608 }, { "epoch": 0.41, "learning_rate": 4.197568597454249e-06, "loss": 0.9206, "step": 10609 }, { "epoch": 0.41, "learning_rate": 4.195812985980647e-06, "loss": 0.908, "step": 10610 }, { "epoch": 0.41, "learning_rate": 4.194057644245718e-06, "loss": 0.9364, "step": 10611 }, { "epoch": 0.41, "learning_rate": 4.192302572331031e-06, "loss": 0.9076, "step": 10612 }, { "epoch": 0.41, "learning_rate": 4.190547770318154e-06, "loss": 0.993, "step": 10613 }, { "epoch": 0.41, "learning_rate": 4.188793238288641e-06, "loss": 0.8969, "step": 10614 }, { "epoch": 0.41, "learning_rate": 4.187038976324023e-06, "loss": 0.8987, "step": 10615 }, { "epoch": 0.41, "learning_rate": 4.18528498450583e-06, "loss": 0.9356, "step": 10616 }, { "epoch": 0.41, "learning_rate": 4.183531262915576e-06, "loss": 0.9154, "step": 10617 }, { "epoch": 0.41, "learning_rate": 4.181777811634765e-06, "loss": 0.9495, "step": 10618 }, { "epoch": 0.41, "learning_rate": 4.180024630744874e-06, "loss": 0.9108, "step": 10619 }, { "epoch": 0.41, "learning_rate": 4.178271720327396e-06, "loss": 0.9245, "step": 10620 }, { "epoch": 0.41, "learning_rate": 4.176519080463779e-06, "loss": 1.0018, "step": 10621 }, { "epoch": 0.41, "learning_rate": 4.1747667112354805e-06, "loss": 0.9568, "step": 10622 }, { "epoch": 0.41, "learning_rate": 4.173014612723941e-06, "loss": 0.9336, "step": 10623 }, { "epoch": 0.41, "learning_rate": 4.171262785010579e-06, "loss": 0.955, "step": 10624 }, { "epoch": 0.41, "learning_rate": 4.1695112281768115e-06, "loss": 0.9506, "step": 10625 }, { "epoch": 0.41, "learning_rate": 4.167759942304038e-06, "loss": 0.9093, "step": 10626 }, { "epoch": 0.41, "learning_rate": 4.166008927473649e-06, "loss": 0.947, "step": 10627 }, { "epoch": 0.41, "learning_rate": 4.164258183767015e-06, "loss": 0.8939, "step": 10628 }, { "epoch": 0.41, "learning_rate": 4.1625077112655e-06, "loss": 0.956, "step": 10629 }, { "epoch": 0.41, "learning_rate": 4.160757510050458e-06, "loss": 0.9218, "step": 10630 }, { "epoch": 0.41, "learning_rate": 4.159007580203216e-06, "loss": 0.9623, "step": 10631 }, { "epoch": 0.41, "learning_rate": 4.1572579218051126e-06, "loss": 0.9237, "step": 10632 }, { "epoch": 0.41, "learning_rate": 4.155508534937449e-06, "loss": 0.9478, "step": 10633 }, { "epoch": 0.41, "learning_rate": 4.153759419681529e-06, "loss": 0.9276, "step": 10634 }, { "epoch": 0.41, "learning_rate": 4.152010576118643e-06, "loss": 0.8991, "step": 10635 }, { "epoch": 0.41, "learning_rate": 4.150262004330055e-06, "loss": 0.9276, "step": 10636 }, { "epoch": 0.41, "learning_rate": 4.148513704397035e-06, "loss": 0.9121, "step": 10637 }, { "epoch": 0.41, "learning_rate": 4.146765676400829e-06, "loss": 0.961, "step": 10638 }, { "epoch": 0.41, "learning_rate": 4.145017920422678e-06, "loss": 0.9741, "step": 10639 }, { "epoch": 0.41, "learning_rate": 4.143270436543794e-06, "loss": 0.8915, "step": 10640 }, { "epoch": 0.41, "learning_rate": 4.141523224845403e-06, "loss": 0.9503, "step": 10641 }, { "epoch": 0.41, "learning_rate": 4.139776285408692e-06, "loss": 0.9374, "step": 10642 }, { "epoch": 0.41, "learning_rate": 4.138029618314852e-06, "loss": 0.9291, "step": 10643 }, { "epoch": 0.41, "learning_rate": 4.136283223645057e-06, "loss": 0.9219, "step": 10644 }, { "epoch": 0.41, "learning_rate": 4.134537101480462e-06, "loss": 0.9129, "step": 10645 }, { "epoch": 0.41, "learning_rate": 4.132791251902218e-06, "loss": 0.9227, "step": 10646 }, { "epoch": 0.41, "learning_rate": 4.131045674991461e-06, "loss": 0.9486, "step": 10647 }, { "epoch": 0.41, "learning_rate": 4.129300370829315e-06, "loss": 0.8807, "step": 10648 }, { "epoch": 0.41, "learning_rate": 4.127555339496885e-06, "loss": 0.9163, "step": 10649 }, { "epoch": 0.41, "learning_rate": 4.125810581075269e-06, "loss": 0.9388, "step": 10650 }, { "epoch": 0.41, "learning_rate": 4.1240660956455555e-06, "loss": 0.9427, "step": 10651 }, { "epoch": 0.41, "learning_rate": 4.122321883288808e-06, "loss": 0.9634, "step": 10652 }, { "epoch": 0.41, "learning_rate": 4.120577944086097e-06, "loss": 0.9433, "step": 10653 }, { "epoch": 0.41, "learning_rate": 4.118834278118459e-06, "loss": 0.8825, "step": 10654 }, { "epoch": 0.41, "learning_rate": 4.117090885466931e-06, "loss": 0.9439, "step": 10655 }, { "epoch": 0.41, "learning_rate": 4.115347766212534e-06, "loss": 0.9582, "step": 10656 }, { "epoch": 0.41, "learning_rate": 4.11360492043628e-06, "loss": 1.0106, "step": 10657 }, { "epoch": 0.41, "learning_rate": 4.111862348219158e-06, "loss": 0.959, "step": 10658 }, { "epoch": 0.41, "learning_rate": 4.110120049642153e-06, "loss": 0.9881, "step": 10659 }, { "epoch": 0.41, "learning_rate": 4.108378024786238e-06, "loss": 0.9418, "step": 10660 }, { "epoch": 0.41, "learning_rate": 4.106636273732365e-06, "loss": 0.926, "step": 10661 }, { "epoch": 0.41, "learning_rate": 4.104894796561482e-06, "loss": 0.9906, "step": 10662 }, { "epoch": 0.41, "learning_rate": 4.1031535933545185e-06, "loss": 0.9607, "step": 10663 }, { "epoch": 0.41, "learning_rate": 4.101412664192397e-06, "loss": 0.8969, "step": 10664 }, { "epoch": 0.41, "learning_rate": 4.099672009156023e-06, "loss": 0.9313, "step": 10665 }, { "epoch": 0.41, "learning_rate": 4.097931628326292e-06, "loss": 0.9384, "step": 10666 }, { "epoch": 0.41, "learning_rate": 4.096191521784079e-06, "loss": 0.9677, "step": 10667 }, { "epoch": 0.41, "learning_rate": 4.094451689610256e-06, "loss": 0.9337, "step": 10668 }, { "epoch": 0.41, "learning_rate": 4.092712131885681e-06, "loss": 0.9332, "step": 10669 }, { "epoch": 0.41, "learning_rate": 4.09097284869119e-06, "loss": 0.9315, "step": 10670 }, { "epoch": 0.41, "learning_rate": 4.089233840107618e-06, "loss": 0.973, "step": 10671 }, { "epoch": 0.41, "learning_rate": 4.08749510621578e-06, "loss": 0.9344, "step": 10672 }, { "epoch": 0.41, "learning_rate": 4.0857566470964846e-06, "loss": 0.9387, "step": 10673 }, { "epoch": 0.41, "learning_rate": 4.084018462830517e-06, "loss": 0.9367, "step": 10674 }, { "epoch": 0.41, "learning_rate": 4.082280553498658e-06, "loss": 0.9344, "step": 10675 }, { "epoch": 0.41, "learning_rate": 4.080542919181681e-06, "loss": 0.9554, "step": 10676 }, { "epoch": 0.41, "learning_rate": 4.078805559960323e-06, "loss": 0.902, "step": 10677 }, { "epoch": 0.41, "learning_rate": 4.077068475915343e-06, "loss": 0.9365, "step": 10678 }, { "epoch": 0.41, "learning_rate": 4.075331667127456e-06, "loss": 0.8915, "step": 10679 }, { "epoch": 0.41, "learning_rate": 4.073595133677382e-06, "loss": 0.9457, "step": 10680 }, { "epoch": 0.41, "learning_rate": 4.0718588756458244e-06, "loss": 0.9276, "step": 10681 }, { "epoch": 0.41, "learning_rate": 4.070122893113469e-06, "loss": 0.9491, "step": 10682 }, { "epoch": 0.41, "learning_rate": 4.068387186160991e-06, "loss": 0.9382, "step": 10683 }, { "epoch": 0.41, "learning_rate": 4.066651754869059e-06, "loss": 1.0048, "step": 10684 }, { "epoch": 0.41, "learning_rate": 4.064916599318325e-06, "loss": 1.0118, "step": 10685 }, { "epoch": 0.41, "learning_rate": 4.063181719589415e-06, "loss": 0.9626, "step": 10686 }, { "epoch": 0.41, "learning_rate": 4.061447115762972e-06, "loss": 0.9515, "step": 10687 }, { "epoch": 0.41, "learning_rate": 4.059712787919595e-06, "loss": 0.9314, "step": 10688 }, { "epoch": 0.41, "learning_rate": 4.057978736139889e-06, "loss": 0.9421, "step": 10689 }, { "epoch": 0.41, "learning_rate": 4.056244960504443e-06, "loss": 0.9461, "step": 10690 }, { "epoch": 0.41, "learning_rate": 4.054511461093824e-06, "loss": 0.9366, "step": 10691 }, { "epoch": 0.41, "learning_rate": 4.0527782379885985e-06, "loss": 0.9724, "step": 10692 }, { "epoch": 0.41, "learning_rate": 4.051045291269312e-06, "loss": 0.9345, "step": 10693 }, { "epoch": 0.41, "learning_rate": 4.049312621016505e-06, "loss": 0.9335, "step": 10694 }, { "epoch": 0.41, "learning_rate": 4.047580227310689e-06, "loss": 0.9446, "step": 10695 }, { "epoch": 0.41, "learning_rate": 4.045848110232389e-06, "loss": 0.9142, "step": 10696 }, { "epoch": 0.41, "learning_rate": 4.044116269862094e-06, "loss": 0.9395, "step": 10697 }, { "epoch": 0.41, "learning_rate": 4.0423847062802805e-06, "loss": 0.9648, "step": 10698 }, { "epoch": 0.41, "learning_rate": 4.040653419567434e-06, "loss": 0.9633, "step": 10699 }, { "epoch": 0.41, "learning_rate": 4.038922409804003e-06, "loss": 0.9472, "step": 10700 }, { "epoch": 0.41, "learning_rate": 4.0371916770704344e-06, "loss": 0.9563, "step": 10701 }, { "epoch": 0.41, "learning_rate": 4.0354612214471635e-06, "loss": 0.9319, "step": 10702 }, { "epoch": 0.41, "learning_rate": 4.033731043014611e-06, "loss": 0.9817, "step": 10703 }, { "epoch": 0.41, "learning_rate": 4.03200114185318e-06, "loss": 0.9452, "step": 10704 }, { "epoch": 0.41, "learning_rate": 4.0302715180432635e-06, "loss": 0.9307, "step": 10705 }, { "epoch": 0.41, "learning_rate": 4.028542171665251e-06, "loss": 0.9429, "step": 10706 }, { "epoch": 0.41, "learning_rate": 4.026813102799499e-06, "loss": 0.945, "step": 10707 }, { "epoch": 0.41, "learning_rate": 4.0250843115263685e-06, "loss": 0.9661, "step": 10708 }, { "epoch": 0.41, "learning_rate": 4.023355797926201e-06, "loss": 0.941, "step": 10709 }, { "epoch": 0.41, "learning_rate": 4.021627562079327e-06, "loss": 0.9103, "step": 10710 }, { "epoch": 0.41, "learning_rate": 4.019899604066061e-06, "loss": 0.9333, "step": 10711 }, { "epoch": 0.41, "learning_rate": 4.018171923966713e-06, "loss": 0.922, "step": 10712 }, { "epoch": 0.41, "learning_rate": 4.016444521861564e-06, "loss": 0.873, "step": 10713 }, { "epoch": 0.41, "learning_rate": 4.014717397830897e-06, "loss": 0.9389, "step": 10714 }, { "epoch": 0.41, "learning_rate": 4.01299055195498e-06, "loss": 0.9409, "step": 10715 }, { "epoch": 0.41, "learning_rate": 4.011263984314058e-06, "loss": 0.9275, "step": 10716 }, { "epoch": 0.41, "learning_rate": 4.009537694988372e-06, "loss": 0.9531, "step": 10717 }, { "epoch": 0.41, "learning_rate": 4.00781168405815e-06, "loss": 0.9258, "step": 10718 }, { "epoch": 0.41, "learning_rate": 4.006085951603608e-06, "loss": 0.9615, "step": 10719 }, { "epoch": 0.41, "learning_rate": 4.004360497704939e-06, "loss": 0.915, "step": 10720 }, { "epoch": 0.41, "learning_rate": 4.002635322442333e-06, "loss": 0.9557, "step": 10721 }, { "epoch": 0.41, "learning_rate": 4.000910425895969e-06, "loss": 0.9579, "step": 10722 }, { "epoch": 0.41, "learning_rate": 3.999185808145996e-06, "loss": 0.948, "step": 10723 }, { "epoch": 0.41, "learning_rate": 3.99746146927258e-06, "loss": 0.9658, "step": 10724 }, { "epoch": 0.41, "learning_rate": 3.995737409355842e-06, "loss": 0.9584, "step": 10725 }, { "epoch": 0.41, "learning_rate": 3.99401362847591e-06, "loss": 0.9302, "step": 10726 }, { "epoch": 0.41, "learning_rate": 3.992290126712896e-06, "loss": 0.9091, "step": 10727 }, { "epoch": 0.41, "learning_rate": 3.99056690414689e-06, "loss": 0.9046, "step": 10728 }, { "epoch": 0.41, "learning_rate": 3.98884396085798e-06, "loss": 0.9505, "step": 10729 }, { "epoch": 0.41, "learning_rate": 3.987121296926234e-06, "loss": 0.9664, "step": 10730 }, { "epoch": 0.41, "learning_rate": 3.985398912431716e-06, "loss": 0.9721, "step": 10731 }, { "epoch": 0.41, "learning_rate": 3.983676807454458e-06, "loss": 0.9238, "step": 10732 }, { "epoch": 0.41, "learning_rate": 3.981954982074506e-06, "loss": 0.9512, "step": 10733 }, { "epoch": 0.41, "learning_rate": 3.980233436371869e-06, "loss": 0.9615, "step": 10734 }, { "epoch": 0.41, "learning_rate": 3.978512170426555e-06, "loss": 0.9597, "step": 10735 }, { "epoch": 0.41, "learning_rate": 3.976791184318561e-06, "loss": 0.9321, "step": 10736 }, { "epoch": 0.41, "learning_rate": 3.975070478127858e-06, "loss": 0.9298, "step": 10737 }, { "epoch": 0.41, "learning_rate": 3.973350051934418e-06, "loss": 0.9224, "step": 10738 }, { "epoch": 0.41, "learning_rate": 3.9716299058181955e-06, "loss": 0.9394, "step": 10739 }, { "epoch": 0.41, "learning_rate": 3.969910039859131e-06, "loss": 0.9237, "step": 10740 }, { "epoch": 0.41, "learning_rate": 3.968190454137143e-06, "loss": 0.9362, "step": 10741 }, { "epoch": 0.41, "learning_rate": 3.966471148732162e-06, "loss": 0.9478, "step": 10742 }, { "epoch": 0.41, "learning_rate": 3.964752123724079e-06, "loss": 0.9147, "step": 10743 }, { "epoch": 0.41, "learning_rate": 3.963033379192779e-06, "loss": 0.9198, "step": 10744 }, { "epoch": 0.41, "learning_rate": 3.961314915218149e-06, "loss": 0.9203, "step": 10745 }, { "epoch": 0.41, "learning_rate": 3.959596731880041e-06, "loss": 0.9092, "step": 10746 }, { "epoch": 0.42, "learning_rate": 3.9578788292583095e-06, "loss": 0.9865, "step": 10747 }, { "epoch": 0.42, "learning_rate": 3.95616120743279e-06, "loss": 0.9152, "step": 10748 }, { "epoch": 0.42, "learning_rate": 3.954443866483309e-06, "loss": 0.919, "step": 10749 }, { "epoch": 0.42, "learning_rate": 3.952726806489668e-06, "loss": 0.94, "step": 10750 }, { "epoch": 0.42, "learning_rate": 3.951010027531671e-06, "loss": 0.9288, "step": 10751 }, { "epoch": 0.42, "learning_rate": 3.949293529689103e-06, "loss": 0.9077, "step": 10752 }, { "epoch": 0.42, "learning_rate": 3.947577313041729e-06, "loss": 0.9277, "step": 10753 }, { "epoch": 0.42, "learning_rate": 3.945861377669311e-06, "loss": 0.9605, "step": 10754 }, { "epoch": 0.42, "learning_rate": 3.944145723651594e-06, "loss": 0.8874, "step": 10755 }, { "epoch": 0.42, "learning_rate": 3.942430351068307e-06, "loss": 0.9285, "step": 10756 }, { "epoch": 0.42, "learning_rate": 3.940715259999172e-06, "loss": 0.9624, "step": 10757 }, { "epoch": 0.42, "learning_rate": 3.939000450523896e-06, "loss": 0.9023, "step": 10758 }, { "epoch": 0.42, "learning_rate": 3.937285922722166e-06, "loss": 0.9522, "step": 10759 }, { "epoch": 0.42, "learning_rate": 3.935571676673664e-06, "loss": 0.9224, "step": 10760 }, { "epoch": 0.42, "learning_rate": 3.93385771245806e-06, "loss": 0.9448, "step": 10761 }, { "epoch": 0.42, "learning_rate": 3.932144030155002e-06, "loss": 0.9277, "step": 10762 }, { "epoch": 0.42, "learning_rate": 3.930430629844131e-06, "loss": 0.973, "step": 10763 }, { "epoch": 0.42, "learning_rate": 3.9287175116050745e-06, "loss": 0.9248, "step": 10764 }, { "epoch": 0.42, "learning_rate": 3.9270046755174515e-06, "loss": 0.953, "step": 10765 }, { "epoch": 0.42, "learning_rate": 3.925292121660855e-06, "loss": 0.9085, "step": 10766 }, { "epoch": 0.42, "learning_rate": 3.923579850114876e-06, "loss": 0.8993, "step": 10767 }, { "epoch": 0.42, "learning_rate": 3.92186786095909e-06, "loss": 0.9768, "step": 10768 }, { "epoch": 0.42, "learning_rate": 3.920156154273056e-06, "loss": 0.9447, "step": 10769 }, { "epoch": 0.42, "learning_rate": 3.918444730136329e-06, "loss": 0.9121, "step": 10770 }, { "epoch": 0.42, "learning_rate": 3.916733588628435e-06, "loss": 0.8921, "step": 10771 }, { "epoch": 0.42, "learning_rate": 3.915022729828901e-06, "loss": 0.9874, "step": 10772 }, { "epoch": 0.42, "learning_rate": 3.913312153817239e-06, "loss": 0.9621, "step": 10773 }, { "epoch": 0.42, "learning_rate": 3.9116018606729375e-06, "loss": 0.9197, "step": 10774 }, { "epoch": 0.42, "learning_rate": 3.909891850475482e-06, "loss": 0.9551, "step": 10775 }, { "epoch": 0.42, "learning_rate": 3.908182123304344e-06, "loss": 0.8872, "step": 10776 }, { "epoch": 0.42, "learning_rate": 3.906472679238982e-06, "loss": 0.9059, "step": 10777 }, { "epoch": 0.42, "learning_rate": 3.904763518358828e-06, "loss": 0.9434, "step": 10778 }, { "epoch": 0.42, "learning_rate": 3.903054640743328e-06, "loss": 0.9011, "step": 10779 }, { "epoch": 0.42, "learning_rate": 3.901346046471887e-06, "loss": 0.9397, "step": 10780 }, { "epoch": 0.42, "learning_rate": 3.899637735623911e-06, "loss": 0.9001, "step": 10781 }, { "epoch": 0.42, "learning_rate": 3.897929708278797e-06, "loss": 0.9273, "step": 10782 }, { "epoch": 0.42, "learning_rate": 3.896221964515914e-06, "loss": 0.9592, "step": 10783 }, { "epoch": 0.42, "learning_rate": 3.894514504414629e-06, "loss": 0.9327, "step": 10784 }, { "epoch": 0.42, "learning_rate": 3.892807328054294e-06, "loss": 0.9215, "step": 10785 }, { "epoch": 0.42, "learning_rate": 3.891100435514249e-06, "loss": 0.9226, "step": 10786 }, { "epoch": 0.42, "learning_rate": 3.88939382687381e-06, "loss": 0.9304, "step": 10787 }, { "epoch": 0.42, "learning_rate": 3.887687502212299e-06, "loss": 0.9329, "step": 10788 }, { "epoch": 0.42, "learning_rate": 3.885981461609012e-06, "loss": 0.9692, "step": 10789 }, { "epoch": 0.42, "learning_rate": 3.8842757051432245e-06, "loss": 0.9206, "step": 10790 }, { "epoch": 0.42, "learning_rate": 3.8825702328942215e-06, "loss": 0.9175, "step": 10791 }, { "epoch": 0.42, "learning_rate": 3.880865044941253e-06, "loss": 0.9333, "step": 10792 }, { "epoch": 0.42, "learning_rate": 3.879160141363566e-06, "loss": 0.9318, "step": 10793 }, { "epoch": 0.42, "learning_rate": 3.877455522240394e-06, "loss": 0.9184, "step": 10794 }, { "epoch": 0.42, "learning_rate": 3.87575118765096e-06, "loss": 0.9358, "step": 10795 }, { "epoch": 0.42, "learning_rate": 3.874047137674462e-06, "loss": 0.9186, "step": 10796 }, { "epoch": 0.42, "learning_rate": 3.872343372390094e-06, "loss": 0.9348, "step": 10797 }, { "epoch": 0.42, "learning_rate": 3.8706398918770414e-06, "loss": 0.9161, "step": 10798 }, { "epoch": 0.42, "learning_rate": 3.868936696214461e-06, "loss": 0.9459, "step": 10799 }, { "epoch": 0.42, "learning_rate": 3.8672337854815146e-06, "loss": 0.9257, "step": 10800 }, { "epoch": 0.42, "learning_rate": 3.865531159757336e-06, "loss": 0.9019, "step": 10801 }, { "epoch": 0.42, "learning_rate": 3.863828819121053e-06, "loss": 0.9171, "step": 10802 }, { "epoch": 0.42, "learning_rate": 3.862126763651779e-06, "loss": 0.9737, "step": 10803 }, { "epoch": 0.42, "learning_rate": 3.860424993428617e-06, "loss": 0.9381, "step": 10804 }, { "epoch": 0.42, "learning_rate": 3.858723508530648e-06, "loss": 0.923, "step": 10805 }, { "epoch": 0.42, "learning_rate": 3.857022309036946e-06, "loss": 0.9291, "step": 10806 }, { "epoch": 0.42, "learning_rate": 3.855321395026578e-06, "loss": 0.9062, "step": 10807 }, { "epoch": 0.42, "learning_rate": 3.853620766578582e-06, "loss": 0.9472, "step": 10808 }, { "epoch": 0.42, "learning_rate": 3.851920423771993e-06, "loss": 0.9403, "step": 10809 }, { "epoch": 0.42, "learning_rate": 3.850220366685835e-06, "loss": 0.9535, "step": 10810 }, { "epoch": 0.42, "learning_rate": 3.848520595399117e-06, "loss": 0.9734, "step": 10811 }, { "epoch": 0.42, "learning_rate": 3.846821109990825e-06, "loss": 0.9322, "step": 10812 }, { "epoch": 0.42, "learning_rate": 3.845121910539943e-06, "loss": 0.9863, "step": 10813 }, { "epoch": 0.42, "learning_rate": 3.843422997125438e-06, "loss": 0.944, "step": 10814 }, { "epoch": 0.42, "learning_rate": 3.841724369826265e-06, "loss": 0.9343, "step": 10815 }, { "epoch": 0.42, "learning_rate": 3.840026028721367e-06, "loss": 0.9514, "step": 10816 }, { "epoch": 0.42, "learning_rate": 3.838327973889664e-06, "loss": 0.9117, "step": 10817 }, { "epoch": 0.42, "learning_rate": 3.836630205410073e-06, "loss": 0.9072, "step": 10818 }, { "epoch": 0.42, "learning_rate": 3.834932723361501e-06, "loss": 0.9244, "step": 10819 }, { "epoch": 0.42, "learning_rate": 3.833235527822825e-06, "loss": 0.9176, "step": 10820 }, { "epoch": 0.42, "learning_rate": 3.831538618872924e-06, "loss": 0.9228, "step": 10821 }, { "epoch": 0.42, "learning_rate": 3.829841996590658e-06, "loss": 0.9582, "step": 10822 }, { "epoch": 0.42, "learning_rate": 3.828145661054879e-06, "loss": 0.9416, "step": 10823 }, { "epoch": 0.42, "learning_rate": 3.826449612344408e-06, "loss": 0.9315, "step": 10824 }, { "epoch": 0.42, "learning_rate": 3.824753850538082e-06, "loss": 0.9473, "step": 10825 }, { "epoch": 0.42, "learning_rate": 3.823058375714698e-06, "loss": 0.9421, "step": 10826 }, { "epoch": 0.42, "learning_rate": 3.821363187953052e-06, "loss": 0.9351, "step": 10827 }, { "epoch": 0.42, "learning_rate": 3.819668287331929e-06, "loss": 0.9742, "step": 10828 }, { "epoch": 0.42, "learning_rate": 3.817973673930089e-06, "loss": 0.9221, "step": 10829 }, { "epoch": 0.42, "learning_rate": 3.816279347826289e-06, "loss": 0.9138, "step": 10830 }, { "epoch": 0.42, "learning_rate": 3.8145853090992712e-06, "loss": 0.944, "step": 10831 }, { "epoch": 0.42, "learning_rate": 3.8128915578277648e-06, "loss": 0.9285, "step": 10832 }, { "epoch": 0.42, "learning_rate": 3.8111980940904734e-06, "loss": 0.9641, "step": 10833 }, { "epoch": 0.42, "learning_rate": 3.8095049179661127e-06, "loss": 0.9106, "step": 10834 }, { "epoch": 0.42, "learning_rate": 3.807812029533362e-06, "loss": 0.9272, "step": 10835 }, { "epoch": 0.42, "learning_rate": 3.806119428870888e-06, "loss": 0.9216, "step": 10836 }, { "epoch": 0.42, "learning_rate": 3.8044271160573654e-06, "loss": 0.9521, "step": 10837 }, { "epoch": 0.42, "learning_rate": 3.802735091171431e-06, "loss": 0.9322, "step": 10838 }, { "epoch": 0.42, "learning_rate": 3.801043354291721e-06, "loss": 0.8861, "step": 10839 }, { "epoch": 0.42, "learning_rate": 3.7993519054968563e-06, "loss": 0.9528, "step": 10840 }, { "epoch": 0.42, "learning_rate": 3.797660744865448e-06, "loss": 0.9418, "step": 10841 }, { "epoch": 0.42, "learning_rate": 3.7959698724760818e-06, "loss": 0.92, "step": 10842 }, { "epoch": 0.42, "learning_rate": 3.7942792884073408e-06, "loss": 0.9337, "step": 10843 }, { "epoch": 0.42, "learning_rate": 3.792588992737797e-06, "loss": 0.9507, "step": 10844 }, { "epoch": 0.42, "learning_rate": 3.790898985545991e-06, "loss": 0.9393, "step": 10845 }, { "epoch": 0.42, "learning_rate": 3.7892092669104787e-06, "loss": 0.9353, "step": 10846 }, { "epoch": 0.42, "learning_rate": 3.7875198369097745e-06, "loss": 0.9462, "step": 10847 }, { "epoch": 0.42, "learning_rate": 3.7858306956223967e-06, "loss": 0.9144, "step": 10848 }, { "epoch": 0.42, "learning_rate": 3.784141843126843e-06, "loss": 0.9721, "step": 10849 }, { "epoch": 0.42, "learning_rate": 3.7824532795016043e-06, "loss": 0.9075, "step": 10850 }, { "epoch": 0.42, "learning_rate": 3.7807650048251456e-06, "loss": 0.9579, "step": 10851 }, { "epoch": 0.42, "learning_rate": 3.7790770191759306e-06, "loss": 0.9591, "step": 10852 }, { "epoch": 0.42, "learning_rate": 3.7773893226324097e-06, "loss": 0.9569, "step": 10853 }, { "epoch": 0.42, "learning_rate": 3.7757019152730055e-06, "loss": 0.9577, "step": 10854 }, { "epoch": 0.42, "learning_rate": 3.7740147971761443e-06, "loss": 0.9787, "step": 10855 }, { "epoch": 0.42, "learning_rate": 3.7723279684202284e-06, "loss": 0.9198, "step": 10856 }, { "epoch": 0.42, "learning_rate": 3.7706414290836544e-06, "loss": 0.9397, "step": 10857 }, { "epoch": 0.42, "learning_rate": 3.7689551792447963e-06, "loss": 0.913, "step": 10858 }, { "epoch": 0.42, "learning_rate": 3.76726921898202e-06, "loss": 0.9696, "step": 10859 }, { "epoch": 0.42, "learning_rate": 3.765583548373678e-06, "loss": 0.9395, "step": 10860 }, { "epoch": 0.42, "learning_rate": 3.76389816749811e-06, "loss": 0.9314, "step": 10861 }, { "epoch": 0.42, "learning_rate": 3.762213076433643e-06, "loss": 0.9656, "step": 10862 }, { "epoch": 0.42, "learning_rate": 3.7605282752585826e-06, "loss": 0.9177, "step": 10863 }, { "epoch": 0.42, "learning_rate": 3.758843764051229e-06, "loss": 0.9517, "step": 10864 }, { "epoch": 0.42, "learning_rate": 3.757159542889871e-06, "loss": 0.9587, "step": 10865 }, { "epoch": 0.42, "learning_rate": 3.7554756118527725e-06, "loss": 0.9409, "step": 10866 }, { "epoch": 0.42, "learning_rate": 3.7537919710181936e-06, "loss": 0.9535, "step": 10867 }, { "epoch": 0.42, "learning_rate": 3.75210862046438e-06, "loss": 0.938, "step": 10868 }, { "epoch": 0.42, "learning_rate": 3.750425560269565e-06, "loss": 0.9202, "step": 10869 }, { "epoch": 0.42, "learning_rate": 3.7487427905119545e-06, "loss": 0.9451, "step": 10870 }, { "epoch": 0.42, "learning_rate": 3.7470603112697667e-06, "loss": 0.9349, "step": 10871 }, { "epoch": 0.42, "learning_rate": 3.7453781226211816e-06, "loss": 0.8974, "step": 10872 }, { "epoch": 0.42, "learning_rate": 3.7436962246443787e-06, "loss": 0.957, "step": 10873 }, { "epoch": 0.42, "learning_rate": 3.7420146174175245e-06, "loss": 0.8788, "step": 10874 }, { "epoch": 0.42, "learning_rate": 3.7403333010187613e-06, "loss": 0.9381, "step": 10875 }, { "epoch": 0.42, "learning_rate": 3.738652275526229e-06, "loss": 0.9463, "step": 10876 }, { "epoch": 0.42, "learning_rate": 3.7369715410180494e-06, "loss": 0.9738, "step": 10877 }, { "epoch": 0.42, "learning_rate": 3.7352910975723366e-06, "loss": 0.9772, "step": 10878 }, { "epoch": 0.42, "learning_rate": 3.7336109452671744e-06, "loss": 0.9211, "step": 10879 }, { "epoch": 0.42, "learning_rate": 3.731931084180659e-06, "loss": 0.9456, "step": 10880 }, { "epoch": 0.42, "learning_rate": 3.7302515143908513e-06, "loss": 0.9409, "step": 10881 }, { "epoch": 0.42, "learning_rate": 3.7285722359757993e-06, "loss": 0.9632, "step": 10882 }, { "epoch": 0.42, "learning_rate": 3.726893249013559e-06, "loss": 0.9345, "step": 10883 }, { "epoch": 0.42, "learning_rate": 3.725214553582147e-06, "loss": 0.8918, "step": 10884 }, { "epoch": 0.42, "learning_rate": 3.723536149759581e-06, "loss": 0.9558, "step": 10885 }, { "epoch": 0.42, "learning_rate": 3.7218580376238622e-06, "loss": 0.9132, "step": 10886 }, { "epoch": 0.42, "learning_rate": 3.72018021725298e-06, "loss": 0.9536, "step": 10887 }, { "epoch": 0.42, "learning_rate": 3.718502688724902e-06, "loss": 0.9243, "step": 10888 }, { "epoch": 0.42, "learning_rate": 3.716825452117593e-06, "loss": 0.9623, "step": 10889 }, { "epoch": 0.42, "learning_rate": 3.715148507509e-06, "loss": 0.9442, "step": 10890 }, { "epoch": 0.42, "learning_rate": 3.7134718549770464e-06, "loss": 0.9269, "step": 10891 }, { "epoch": 0.42, "learning_rate": 3.711795494599666e-06, "loss": 0.9433, "step": 10892 }, { "epoch": 0.42, "learning_rate": 3.710119426454755e-06, "loss": 0.9335, "step": 10893 }, { "epoch": 0.42, "learning_rate": 3.708443650620206e-06, "loss": 0.9741, "step": 10894 }, { "epoch": 0.42, "learning_rate": 3.706768167173901e-06, "loss": 0.9867, "step": 10895 }, { "epoch": 0.42, "learning_rate": 3.705092976193705e-06, "loss": 0.9162, "step": 10896 }, { "epoch": 0.42, "learning_rate": 3.703418077757466e-06, "loss": 0.9198, "step": 10897 }, { "epoch": 0.43, "learning_rate": 3.7017434719430224e-06, "loss": 0.9789, "step": 10898 }, { "epoch": 0.43, "learning_rate": 3.7000691588282022e-06, "loss": 0.9046, "step": 10899 }, { "epoch": 0.43, "learning_rate": 3.6983951384908102e-06, "loss": 0.9831, "step": 10900 }, { "epoch": 0.43, "learning_rate": 3.696721411008647e-06, "loss": 0.9366, "step": 10901 }, { "epoch": 0.43, "learning_rate": 3.695047976459494e-06, "loss": 0.9376, "step": 10902 }, { "epoch": 0.43, "learning_rate": 3.693374834921123e-06, "loss": 0.9208, "step": 10903 }, { "epoch": 0.43, "learning_rate": 3.691701986471291e-06, "loss": 0.9153, "step": 10904 }, { "epoch": 0.43, "learning_rate": 3.690029431187737e-06, "loss": 0.9638, "step": 10905 }, { "epoch": 0.43, "learning_rate": 3.688357169148191e-06, "loss": 0.9847, "step": 10906 }, { "epoch": 0.43, "learning_rate": 3.686685200430369e-06, "loss": 0.9348, "step": 10907 }, { "epoch": 0.43, "learning_rate": 3.6850135251119745e-06, "loss": 0.9749, "step": 10908 }, { "epoch": 0.43, "learning_rate": 3.6833421432706906e-06, "loss": 0.9704, "step": 10909 }, { "epoch": 0.43, "learning_rate": 3.6816710549841948e-06, "loss": 0.9227, "step": 10910 }, { "epoch": 0.43, "learning_rate": 3.680000260330151e-06, "loss": 0.9631, "step": 10911 }, { "epoch": 0.43, "learning_rate": 3.678329759386199e-06, "loss": 0.9601, "step": 10912 }, { "epoch": 0.43, "learning_rate": 3.6766595522299755e-06, "loss": 0.9199, "step": 10913 }, { "epoch": 0.43, "learning_rate": 3.6749896389390994e-06, "loss": 0.9371, "step": 10914 }, { "epoch": 0.43, "learning_rate": 3.6733200195911823e-06, "loss": 0.9285, "step": 10915 }, { "epoch": 0.43, "learning_rate": 3.6716506942638055e-06, "loss": 0.9425, "step": 10916 }, { "epoch": 0.43, "learning_rate": 3.6699816630345607e-06, "loss": 0.9455, "step": 10917 }, { "epoch": 0.43, "learning_rate": 3.6683129259810046e-06, "loss": 0.9449, "step": 10918 }, { "epoch": 0.43, "learning_rate": 3.666644483180689e-06, "loss": 0.9609, "step": 10919 }, { "epoch": 0.43, "learning_rate": 3.664976334711158e-06, "loss": 0.9189, "step": 10920 }, { "epoch": 0.43, "learning_rate": 3.6633084806499275e-06, "loss": 0.9562, "step": 10921 }, { "epoch": 0.43, "learning_rate": 3.661640921074511e-06, "loss": 0.9268, "step": 10922 }, { "epoch": 0.43, "learning_rate": 3.6599736560624054e-06, "loss": 0.9143, "step": 10923 }, { "epoch": 0.43, "learning_rate": 3.6583066856910978e-06, "loss": 0.9276, "step": 10924 }, { "epoch": 0.43, "learning_rate": 3.6566400100380474e-06, "loss": 0.9612, "step": 10925 }, { "epoch": 0.43, "learning_rate": 3.6549736291807214e-06, "loss": 0.9257, "step": 10926 }, { "epoch": 0.43, "learning_rate": 3.653307543196558e-06, "loss": 0.94, "step": 10927 }, { "epoch": 0.43, "learning_rate": 3.6516417521629757e-06, "loss": 0.8801, "step": 10928 }, { "epoch": 0.43, "learning_rate": 3.6499762561574046e-06, "loss": 0.9781, "step": 10929 }, { "epoch": 0.43, "learning_rate": 3.6483110552572344e-06, "loss": 0.9508, "step": 10930 }, { "epoch": 0.43, "learning_rate": 3.646646149539855e-06, "loss": 0.9204, "step": 10931 }, { "epoch": 0.43, "learning_rate": 3.6449815390826415e-06, "loss": 0.9417, "step": 10932 }, { "epoch": 0.43, "learning_rate": 3.6433172239629556e-06, "loss": 0.9692, "step": 10933 }, { "epoch": 0.43, "learning_rate": 3.6416532042581366e-06, "loss": 0.9666, "step": 10934 }, { "epoch": 0.43, "learning_rate": 3.6399894800455202e-06, "loss": 0.8994, "step": 10935 }, { "epoch": 0.43, "learning_rate": 3.638326051402429e-06, "loss": 0.9826, "step": 10936 }, { "epoch": 0.43, "learning_rate": 3.636662918406155e-06, "loss": 0.9048, "step": 10937 }, { "epoch": 0.43, "learning_rate": 3.635000081134006e-06, "loss": 0.9195, "step": 10938 }, { "epoch": 0.43, "learning_rate": 3.6333375396632475e-06, "loss": 0.971, "step": 10939 }, { "epoch": 0.43, "learning_rate": 3.6316752940711464e-06, "loss": 0.9405, "step": 10940 }, { "epoch": 0.43, "learning_rate": 3.6300133444349528e-06, "loss": 0.9555, "step": 10941 }, { "epoch": 0.43, "learning_rate": 3.628351690831906e-06, "loss": 0.9562, "step": 10942 }, { "epoch": 0.43, "learning_rate": 3.62669033333922e-06, "loss": 0.9779, "step": 10943 }, { "epoch": 0.43, "learning_rate": 3.6250292720341095e-06, "loss": 0.9085, "step": 10944 }, { "epoch": 0.43, "learning_rate": 3.6233685069937698e-06, "loss": 0.9827, "step": 10945 }, { "epoch": 0.43, "learning_rate": 3.6217080382953775e-06, "loss": 0.9645, "step": 10946 }, { "epoch": 0.43, "learning_rate": 3.620047866016101e-06, "loss": 0.9394, "step": 10947 }, { "epoch": 0.43, "learning_rate": 3.6183879902330944e-06, "loss": 0.9058, "step": 10948 }, { "epoch": 0.43, "learning_rate": 3.616728411023498e-06, "loss": 0.9391, "step": 10949 }, { "epoch": 0.43, "learning_rate": 3.61506912846444e-06, "loss": 0.9599, "step": 10950 }, { "epoch": 0.43, "learning_rate": 3.613410142633026e-06, "loss": 0.9852, "step": 10951 }, { "epoch": 0.43, "learning_rate": 3.611751453606358e-06, "loss": 0.9085, "step": 10952 }, { "epoch": 0.43, "learning_rate": 3.6100930614615204e-06, "loss": 0.9364, "step": 10953 }, { "epoch": 0.43, "learning_rate": 3.608434966275587e-06, "loss": 0.8824, "step": 10954 }, { "epoch": 0.43, "learning_rate": 3.606777168125608e-06, "loss": 0.9335, "step": 10955 }, { "epoch": 0.43, "learning_rate": 3.60511966708863e-06, "loss": 0.9421, "step": 10956 }, { "epoch": 0.43, "learning_rate": 3.603462463241686e-06, "loss": 0.8852, "step": 10957 }, { "epoch": 0.43, "learning_rate": 3.601805556661783e-06, "loss": 0.9479, "step": 10958 }, { "epoch": 0.43, "learning_rate": 3.600148947425928e-06, "loss": 0.9662, "step": 10959 }, { "epoch": 0.43, "learning_rate": 3.598492635611106e-06, "loss": 0.9367, "step": 10960 }, { "epoch": 0.43, "learning_rate": 3.596836621294295e-06, "loss": 0.9473, "step": 10961 }, { "epoch": 0.43, "learning_rate": 3.5951809045524523e-06, "loss": 0.9672, "step": 10962 }, { "epoch": 0.43, "learning_rate": 3.593525485462528e-06, "loss": 0.942, "step": 10963 }, { "epoch": 0.43, "learning_rate": 3.5918703641014474e-06, "loss": 0.9408, "step": 10964 }, { "epoch": 0.43, "learning_rate": 3.5902155405461346e-06, "loss": 0.9745, "step": 10965 }, { "epoch": 0.43, "learning_rate": 3.588561014873496e-06, "loss": 0.916, "step": 10966 }, { "epoch": 0.43, "learning_rate": 3.586906787160416e-06, "loss": 0.9274, "step": 10967 }, { "epoch": 0.43, "learning_rate": 3.5852528574837754e-06, "loss": 0.9512, "step": 10968 }, { "epoch": 0.43, "learning_rate": 3.5835992259204376e-06, "loss": 0.8975, "step": 10969 }, { "epoch": 0.43, "learning_rate": 3.581945892547256e-06, "loss": 0.9668, "step": 10970 }, { "epoch": 0.43, "learning_rate": 3.5802928574410544e-06, "loss": 0.9346, "step": 10971 }, { "epoch": 0.43, "learning_rate": 3.578640120678668e-06, "loss": 1.0111, "step": 10972 }, { "epoch": 0.43, "learning_rate": 3.5769876823369e-06, "loss": 0.9402, "step": 10973 }, { "epoch": 0.43, "learning_rate": 3.5753355424925364e-06, "loss": 0.9366, "step": 10974 }, { "epoch": 0.43, "learning_rate": 3.57368370122237e-06, "loss": 0.9621, "step": 10975 }, { "epoch": 0.43, "learning_rate": 3.572032158603158e-06, "loss": 0.9588, "step": 10976 }, { "epoch": 0.43, "learning_rate": 3.5703809147116552e-06, "loss": 0.9623, "step": 10977 }, { "epoch": 0.43, "learning_rate": 3.5687299696246012e-06, "loss": 0.9252, "step": 10978 }, { "epoch": 0.43, "learning_rate": 3.567079323418724e-06, "loss": 0.9568, "step": 10979 }, { "epoch": 0.43, "learning_rate": 3.565428976170725e-06, "loss": 0.9607, "step": 10980 }, { "epoch": 0.43, "learning_rate": 3.5637789279573077e-06, "loss": 0.9333, "step": 10981 }, { "epoch": 0.43, "learning_rate": 3.562129178855156e-06, "loss": 0.9014, "step": 10982 }, { "epoch": 0.43, "learning_rate": 3.5604797289409297e-06, "loss": 0.9636, "step": 10983 }, { "epoch": 0.43, "learning_rate": 3.558830578291298e-06, "loss": 0.984, "step": 10984 }, { "epoch": 0.43, "learning_rate": 3.557181726982891e-06, "loss": 0.9347, "step": 10985 }, { "epoch": 0.43, "learning_rate": 3.5555331750923396e-06, "loss": 0.9138, "step": 10986 }, { "epoch": 0.43, "learning_rate": 3.553884922696257e-06, "loss": 0.9315, "step": 10987 }, { "epoch": 0.43, "learning_rate": 3.5522369698712465e-06, "loss": 0.9414, "step": 10988 }, { "epoch": 0.43, "learning_rate": 3.5505893166938853e-06, "loss": 0.9439, "step": 10989 }, { "epoch": 0.43, "learning_rate": 3.5489419632407508e-06, "loss": 0.97, "step": 10990 }, { "epoch": 0.43, "learning_rate": 3.5472949095884024e-06, "loss": 0.9302, "step": 10991 }, { "epoch": 0.43, "learning_rate": 3.545648155813374e-06, "loss": 0.9329, "step": 10992 }, { "epoch": 0.43, "learning_rate": 3.544001701992209e-06, "loss": 0.9656, "step": 10993 }, { "epoch": 0.43, "learning_rate": 3.5423555482014137e-06, "loss": 0.9258, "step": 10994 }, { "epoch": 0.43, "learning_rate": 3.5407096945174914e-06, "loss": 0.9221, "step": 10995 }, { "epoch": 0.43, "learning_rate": 3.5390641410169355e-06, "loss": 0.9382, "step": 10996 }, { "epoch": 0.43, "learning_rate": 3.5374188877762127e-06, "loss": 0.9129, "step": 10997 }, { "epoch": 0.43, "learning_rate": 3.535773934871787e-06, "loss": 0.9055, "step": 10998 }, { "epoch": 0.43, "learning_rate": 3.534129282380102e-06, "loss": 0.9597, "step": 10999 }, { "epoch": 0.43, "learning_rate": 3.5324849303775965e-06, "loss": 0.9442, "step": 11000 }, { "epoch": 0.43, "learning_rate": 3.5308408789406788e-06, "loss": 0.979, "step": 11001 }, { "epoch": 0.43, "learning_rate": 3.5291971281457593e-06, "loss": 0.9172, "step": 11002 }, { "epoch": 0.43, "learning_rate": 3.52755367806923e-06, "loss": 0.959, "step": 11003 }, { "epoch": 0.43, "learning_rate": 3.5259105287874608e-06, "loss": 0.9291, "step": 11004 }, { "epoch": 0.43, "learning_rate": 3.5242676803768174e-06, "loss": 0.9081, "step": 11005 }, { "epoch": 0.43, "learning_rate": 3.5226251329136475e-06, "loss": 0.91, "step": 11006 }, { "epoch": 0.43, "learning_rate": 3.5209828864742866e-06, "loss": 0.931, "step": 11007 }, { "epoch": 0.43, "learning_rate": 3.519340941135053e-06, "loss": 0.9293, "step": 11008 }, { "epoch": 0.43, "learning_rate": 3.517699296972259e-06, "loss": 0.9572, "step": 11009 }, { "epoch": 0.43, "learning_rate": 3.5160579540621886e-06, "loss": 0.9458, "step": 11010 }, { "epoch": 0.43, "learning_rate": 3.5144169124811233e-06, "loss": 0.9503, "step": 11011 }, { "epoch": 0.43, "learning_rate": 3.5127761723053313e-06, "loss": 0.9201, "step": 11012 }, { "epoch": 0.43, "learning_rate": 3.5111357336110564e-06, "loss": 0.9276, "step": 11013 }, { "epoch": 0.43, "learning_rate": 3.509495596474538e-06, "loss": 0.9305, "step": 11014 }, { "epoch": 0.43, "learning_rate": 3.5078557609719986e-06, "loss": 0.936, "step": 11015 }, { "epoch": 0.43, "learning_rate": 3.5062162271796486e-06, "loss": 0.9567, "step": 11016 }, { "epoch": 0.43, "learning_rate": 3.5045769951736743e-06, "loss": 0.9324, "step": 11017 }, { "epoch": 0.43, "learning_rate": 3.502938065030268e-06, "loss": 0.9217, "step": 11018 }, { "epoch": 0.43, "learning_rate": 3.5012994368255894e-06, "loss": 0.9324, "step": 11019 }, { "epoch": 0.43, "learning_rate": 3.4996611106357837e-06, "loss": 0.955, "step": 11020 }, { "epoch": 0.43, "learning_rate": 3.4980230865370023e-06, "loss": 0.9814, "step": 11021 }, { "epoch": 0.43, "learning_rate": 3.4963853646053604e-06, "loss": 0.935, "step": 11022 }, { "epoch": 0.43, "learning_rate": 3.4947479449169695e-06, "loss": 0.9341, "step": 11023 }, { "epoch": 0.43, "learning_rate": 3.4931108275479287e-06, "loss": 0.9244, "step": 11024 }, { "epoch": 0.43, "learning_rate": 3.49147401257432e-06, "loss": 0.9213, "step": 11025 }, { "epoch": 0.43, "learning_rate": 3.4898375000722063e-06, "loss": 0.9362, "step": 11026 }, { "epoch": 0.43, "learning_rate": 3.4882012901176433e-06, "loss": 0.95, "step": 11027 }, { "epoch": 0.43, "learning_rate": 3.486565382786676e-06, "loss": 0.9708, "step": 11028 }, { "epoch": 0.43, "learning_rate": 3.4849297781553205e-06, "loss": 0.9343, "step": 11029 }, { "epoch": 0.43, "learning_rate": 3.483294476299599e-06, "loss": 0.9454, "step": 11030 }, { "epoch": 0.43, "learning_rate": 3.4816594772955013e-06, "loss": 0.9289, "step": 11031 }, { "epoch": 0.43, "learning_rate": 3.4800247812190136e-06, "loss": 0.9716, "step": 11032 }, { "epoch": 0.43, "learning_rate": 3.4783903881461054e-06, "loss": 0.9225, "step": 11033 }, { "epoch": 0.43, "learning_rate": 3.4767562981527357e-06, "loss": 0.8953, "step": 11034 }, { "epoch": 0.43, "learning_rate": 3.475122511314838e-06, "loss": 0.9143, "step": 11035 }, { "epoch": 0.43, "learning_rate": 3.4734890277083444e-06, "loss": 0.9685, "step": 11036 }, { "epoch": 0.43, "learning_rate": 3.47185584740917e-06, "loss": 0.9735, "step": 11037 }, { "epoch": 0.43, "learning_rate": 3.4702229704932043e-06, "loss": 0.9362, "step": 11038 }, { "epoch": 0.43, "learning_rate": 3.4685903970363457e-06, "loss": 0.9374, "step": 11039 }, { "epoch": 0.43, "learning_rate": 3.4669581271144535e-06, "loss": 0.9691, "step": 11040 }, { "epoch": 0.43, "learning_rate": 3.4653261608033893e-06, "loss": 0.9607, "step": 11041 }, { "epoch": 0.43, "learning_rate": 3.463694498179e-06, "loss": 0.9813, "step": 11042 }, { "epoch": 0.43, "learning_rate": 3.4620631393171046e-06, "loss": 0.9348, "step": 11043 }, { "epoch": 0.43, "learning_rate": 3.4604320842935226e-06, "loss": 0.9257, "step": 11044 }, { "epoch": 0.43, "learning_rate": 3.4588013331840542e-06, "loss": 0.9793, "step": 11045 }, { "epoch": 0.43, "learning_rate": 3.457170886064488e-06, "loss": 0.9526, "step": 11046 }, { "epoch": 0.43, "learning_rate": 3.455540743010589e-06, "loss": 0.9734, "step": 11047 }, { "epoch": 0.44, "learning_rate": 3.45391090409812e-06, "loss": 0.9304, "step": 11048 }, { "epoch": 0.44, "learning_rate": 3.452281369402827e-06, "loss": 0.9843, "step": 11049 }, { "epoch": 0.44, "learning_rate": 3.4506521390004332e-06, "loss": 0.9, "step": 11050 }, { "epoch": 0.44, "learning_rate": 3.4490232129666567e-06, "loss": 0.9442, "step": 11051 }, { "epoch": 0.44, "learning_rate": 3.447394591377199e-06, "loss": 0.8601, "step": 11052 }, { "epoch": 0.44, "learning_rate": 3.445766274307748e-06, "loss": 0.932, "step": 11053 }, { "epoch": 0.44, "learning_rate": 3.444138261833976e-06, "loss": 0.921, "step": 11054 }, { "epoch": 0.44, "learning_rate": 3.4425105540315452e-06, "loss": 0.9366, "step": 11055 }, { "epoch": 0.44, "learning_rate": 3.4408831509760933e-06, "loss": 0.9392, "step": 11056 }, { "epoch": 0.44, "learning_rate": 3.439256052743254e-06, "loss": 0.8867, "step": 11057 }, { "epoch": 0.44, "learning_rate": 3.437629259408648e-06, "loss": 0.9483, "step": 11058 }, { "epoch": 0.44, "learning_rate": 3.4360027710478707e-06, "loss": 0.9091, "step": 11059 }, { "epoch": 0.44, "learning_rate": 3.4343765877365133e-06, "loss": 0.8917, "step": 11060 }, { "epoch": 0.44, "learning_rate": 3.4327507095501477e-06, "loss": 0.9203, "step": 11061 }, { "epoch": 0.44, "learning_rate": 3.4311251365643393e-06, "loss": 0.94, "step": 11062 }, { "epoch": 0.44, "learning_rate": 3.429499868854623e-06, "loss": 0.9173, "step": 11063 }, { "epoch": 0.44, "learning_rate": 3.4278749064965422e-06, "loss": 0.945, "step": 11064 }, { "epoch": 0.44, "learning_rate": 3.426250249565606e-06, "loss": 0.9366, "step": 11065 }, { "epoch": 0.44, "learning_rate": 3.4246258981373184e-06, "loss": 0.9306, "step": 11066 }, { "epoch": 0.44, "learning_rate": 3.423001852287172e-06, "loss": 0.9348, "step": 11067 }, { "epoch": 0.44, "learning_rate": 3.4213781120906365e-06, "loss": 0.9574, "step": 11068 }, { "epoch": 0.44, "learning_rate": 3.4197546776231737e-06, "loss": 0.9427, "step": 11069 }, { "epoch": 0.44, "learning_rate": 3.4181315489602296e-06, "loss": 0.9591, "step": 11070 }, { "epoch": 0.44, "learning_rate": 3.41650872617724e-06, "loss": 0.9105, "step": 11071 }, { "epoch": 0.44, "learning_rate": 3.414886209349615e-06, "loss": 0.9216, "step": 11072 }, { "epoch": 0.44, "learning_rate": 3.4132639985527627e-06, "loss": 0.9203, "step": 11073 }, { "epoch": 0.44, "learning_rate": 3.4116420938620743e-06, "loss": 0.9574, "step": 11074 }, { "epoch": 0.44, "learning_rate": 3.4100204953529157e-06, "loss": 0.9558, "step": 11075 }, { "epoch": 0.44, "learning_rate": 3.408399203100662e-06, "loss": 0.9393, "step": 11076 }, { "epoch": 0.44, "learning_rate": 3.406778217180646e-06, "loss": 0.9204, "step": 11077 }, { "epoch": 0.44, "learning_rate": 3.405157537668208e-06, "loss": 0.913, "step": 11078 }, { "epoch": 0.44, "learning_rate": 3.4035371646386627e-06, "loss": 0.9119, "step": 11079 }, { "epoch": 0.44, "learning_rate": 3.4019170981673178e-06, "loss": 0.9823, "step": 11080 }, { "epoch": 0.44, "learning_rate": 3.400297338329457e-06, "loss": 0.9216, "step": 11081 }, { "epoch": 0.44, "learning_rate": 3.3986778852003588e-06, "loss": 0.9173, "step": 11082 }, { "epoch": 0.44, "learning_rate": 3.3970587388552865e-06, "loss": 0.9173, "step": 11083 }, { "epoch": 0.44, "learning_rate": 3.3954398993694793e-06, "loss": 0.9287, "step": 11084 }, { "epoch": 0.44, "learning_rate": 3.39382136681818e-06, "loss": 0.941, "step": 11085 }, { "epoch": 0.44, "learning_rate": 3.3922031412765988e-06, "loss": 0.8981, "step": 11086 }, { "epoch": 0.44, "learning_rate": 3.390585222819942e-06, "loss": 0.924, "step": 11087 }, { "epoch": 0.44, "learning_rate": 3.3889676115234037e-06, "loss": 0.9229, "step": 11088 }, { "epoch": 0.44, "learning_rate": 3.387350307462153e-06, "loss": 0.9476, "step": 11089 }, { "epoch": 0.44, "learning_rate": 3.385733310711352e-06, "loss": 0.9308, "step": 11090 }, { "epoch": 0.44, "learning_rate": 3.3841166213461497e-06, "loss": 0.914, "step": 11091 }, { "epoch": 0.44, "learning_rate": 3.382500239441682e-06, "loss": 0.9365, "step": 11092 }, { "epoch": 0.44, "learning_rate": 3.3808841650730584e-06, "loss": 0.9452, "step": 11093 }, { "epoch": 0.44, "learning_rate": 3.3792683983153897e-06, "loss": 0.9359, "step": 11094 }, { "epoch": 0.44, "learning_rate": 3.3776529392437663e-06, "loss": 0.9348, "step": 11095 }, { "epoch": 0.44, "learning_rate": 3.376037787933254e-06, "loss": 0.9245, "step": 11096 }, { "epoch": 0.44, "learning_rate": 3.3744229444589284e-06, "loss": 0.947, "step": 11097 }, { "epoch": 0.44, "learning_rate": 3.3728084088958247e-06, "loss": 0.9343, "step": 11098 }, { "epoch": 0.44, "learning_rate": 3.3711941813189807e-06, "loss": 0.9483, "step": 11099 }, { "epoch": 0.44, "learning_rate": 3.369580261803412e-06, "loss": 0.9222, "step": 11100 }, { "epoch": 0.44, "learning_rate": 3.3679666504241294e-06, "loss": 0.9559, "step": 11101 }, { "epoch": 0.44, "learning_rate": 3.3663533472561118e-06, "loss": 0.978, "step": 11102 }, { "epoch": 0.44, "learning_rate": 3.3647403523743403e-06, "loss": 0.9289, "step": 11103 }, { "epoch": 0.44, "learning_rate": 3.363127665853779e-06, "loss": 0.9056, "step": 11104 }, { "epoch": 0.44, "learning_rate": 3.361515287769367e-06, "loss": 0.9212, "step": 11105 }, { "epoch": 0.44, "learning_rate": 3.35990321819604e-06, "loss": 0.9153, "step": 11106 }, { "epoch": 0.44, "learning_rate": 3.3582914572087156e-06, "loss": 0.9318, "step": 11107 }, { "epoch": 0.44, "learning_rate": 3.3566800048823024e-06, "loss": 0.9558, "step": 11108 }, { "epoch": 0.44, "learning_rate": 3.355068861291677e-06, "loss": 0.9494, "step": 11109 }, { "epoch": 0.44, "learning_rate": 3.35345802651173e-06, "loss": 0.9044, "step": 11110 }, { "epoch": 0.44, "learning_rate": 3.35184750061731e-06, "loss": 0.9432, "step": 11111 }, { "epoch": 0.44, "learning_rate": 3.350237283683269e-06, "loss": 0.9454, "step": 11112 }, { "epoch": 0.44, "learning_rate": 3.348627375784439e-06, "loss": 0.9073, "step": 11113 }, { "epoch": 0.44, "learning_rate": 3.3470177769956326e-06, "loss": 0.905, "step": 11114 }, { "epoch": 0.44, "learning_rate": 3.345408487391656e-06, "loss": 0.9421, "step": 11115 }, { "epoch": 0.44, "learning_rate": 3.343799507047297e-06, "loss": 0.9591, "step": 11116 }, { "epoch": 0.44, "learning_rate": 3.3421908360373357e-06, "loss": 0.9375, "step": 11117 }, { "epoch": 0.44, "learning_rate": 3.3405824744365222e-06, "loss": 0.9139, "step": 11118 }, { "epoch": 0.44, "learning_rate": 3.3389744223196087e-06, "loss": 0.8937, "step": 11119 }, { "epoch": 0.44, "learning_rate": 3.337366679761326e-06, "loss": 0.9406, "step": 11120 }, { "epoch": 0.44, "learning_rate": 3.335759246836384e-06, "loss": 0.9685, "step": 11121 }, { "epoch": 0.44, "learning_rate": 3.3341521236194973e-06, "loss": 0.9114, "step": 11122 }, { "epoch": 0.44, "learning_rate": 3.3325453101853445e-06, "loss": 0.9226, "step": 11123 }, { "epoch": 0.44, "learning_rate": 3.3309388066086013e-06, "loss": 0.9286, "step": 11124 }, { "epoch": 0.44, "learning_rate": 3.329332612963928e-06, "loss": 0.9382, "step": 11125 }, { "epoch": 0.44, "learning_rate": 3.327726729325972e-06, "loss": 0.9117, "step": 11126 }, { "epoch": 0.44, "learning_rate": 3.3261211557693583e-06, "loss": 0.9111, "step": 11127 }, { "epoch": 0.44, "learning_rate": 3.3245158923687047e-06, "loss": 0.9325, "step": 11128 }, { "epoch": 0.44, "learning_rate": 3.322910939198617e-06, "loss": 0.9619, "step": 11129 }, { "epoch": 0.44, "learning_rate": 3.321306296333673e-06, "loss": 0.9594, "step": 11130 }, { "epoch": 0.44, "learning_rate": 3.3197019638484574e-06, "loss": 0.9438, "step": 11131 }, { "epoch": 0.44, "learning_rate": 3.3180979418175186e-06, "loss": 0.9407, "step": 11132 }, { "epoch": 0.44, "learning_rate": 3.316494230315406e-06, "loss": 0.9108, "step": 11133 }, { "epoch": 0.44, "learning_rate": 3.3148908294166495e-06, "loss": 0.9258, "step": 11134 }, { "epoch": 0.44, "learning_rate": 3.3132877391957597e-06, "loss": 0.9271, "step": 11135 }, { "epoch": 0.44, "learning_rate": 3.311684959727238e-06, "loss": 0.9299, "step": 11136 }, { "epoch": 0.44, "learning_rate": 3.3100824910855745e-06, "loss": 0.9648, "step": 11137 }, { "epoch": 0.44, "learning_rate": 3.30848033334524e-06, "loss": 0.9477, "step": 11138 }, { "epoch": 0.44, "learning_rate": 3.306878486580688e-06, "loss": 0.9215, "step": 11139 }, { "epoch": 0.44, "learning_rate": 3.305276950866364e-06, "loss": 0.9335, "step": 11140 }, { "epoch": 0.44, "learning_rate": 3.303675726276698e-06, "loss": 0.9831, "step": 11141 }, { "epoch": 0.44, "learning_rate": 3.302074812886097e-06, "loss": 0.9132, "step": 11142 }, { "epoch": 0.44, "learning_rate": 3.3004742107689702e-06, "loss": 0.9497, "step": 11143 }, { "epoch": 0.44, "learning_rate": 3.2988739199996956e-06, "loss": 0.9231, "step": 11144 }, { "epoch": 0.44, "learning_rate": 3.2972739406526466e-06, "loss": 0.953, "step": 11145 }, { "epoch": 0.44, "learning_rate": 3.2956742728021774e-06, "loss": 0.9408, "step": 11146 }, { "epoch": 0.44, "learning_rate": 3.2940749165226337e-06, "loss": 0.9383, "step": 11147 }, { "epoch": 0.44, "learning_rate": 3.2924758718883365e-06, "loss": 0.9261, "step": 11148 }, { "epoch": 0.44, "learning_rate": 3.2908771389736017e-06, "loss": 0.9568, "step": 11149 }, { "epoch": 0.44, "learning_rate": 3.2892787178527307e-06, "loss": 0.9338, "step": 11150 }, { "epoch": 0.44, "learning_rate": 3.287680608599999e-06, "loss": 0.9306, "step": 11151 }, { "epoch": 0.44, "learning_rate": 3.2860828112896812e-06, "loss": 0.9639, "step": 11152 }, { "epoch": 0.44, "learning_rate": 3.2844853259960308e-06, "loss": 0.9126, "step": 11153 }, { "epoch": 0.44, "learning_rate": 3.2828881527932923e-06, "loss": 0.9195, "step": 11154 }, { "epoch": 0.44, "learning_rate": 3.2812912917556784e-06, "loss": 0.9429, "step": 11155 }, { "epoch": 0.44, "learning_rate": 3.2796947429574165e-06, "loss": 0.9495, "step": 11156 }, { "epoch": 0.44, "learning_rate": 3.2780985064726934e-06, "loss": 0.9597, "step": 11157 }, { "epoch": 0.44, "learning_rate": 3.2765025823756923e-06, "loss": 0.9419, "step": 11158 }, { "epoch": 0.44, "learning_rate": 3.2749069707405857e-06, "loss": 0.9471, "step": 11159 }, { "epoch": 0.44, "learning_rate": 3.273311671641519e-06, "loss": 0.9312, "step": 11160 }, { "epoch": 0.44, "learning_rate": 3.271716685152635e-06, "loss": 0.9402, "step": 11161 }, { "epoch": 0.44, "learning_rate": 3.270122011348057e-06, "loss": 0.9376, "step": 11162 }, { "epoch": 0.44, "learning_rate": 3.268527650301898e-06, "loss": 0.9228, "step": 11163 }, { "epoch": 0.44, "learning_rate": 3.2669336020882457e-06, "loss": 0.9511, "step": 11164 }, { "epoch": 0.44, "learning_rate": 3.2653398667811853e-06, "loss": 0.9398, "step": 11165 }, { "epoch": 0.44, "learning_rate": 3.2637464444547852e-06, "loss": 0.8902, "step": 11166 }, { "epoch": 0.44, "learning_rate": 3.262153335183086e-06, "loss": 0.9317, "step": 11167 }, { "epoch": 0.44, "learning_rate": 3.2605605390401386e-06, "loss": 0.9483, "step": 11168 }, { "epoch": 0.44, "learning_rate": 3.258968056099956e-06, "loss": 0.8823, "step": 11169 }, { "epoch": 0.44, "learning_rate": 3.2573758864365467e-06, "loss": 0.9311, "step": 11170 }, { "epoch": 0.44, "learning_rate": 3.2557840301239062e-06, "loss": 0.9709, "step": 11171 }, { "epoch": 0.44, "learning_rate": 3.2541924872360152e-06, "loss": 0.952, "step": 11172 }, { "epoch": 0.44, "learning_rate": 3.2526012578468303e-06, "loss": 0.948, "step": 11173 }, { "epoch": 0.44, "learning_rate": 3.2510103420303063e-06, "loss": 0.9181, "step": 11174 }, { "epoch": 0.44, "learning_rate": 3.24941973986038e-06, "loss": 0.9065, "step": 11175 }, { "epoch": 0.44, "learning_rate": 3.247829451410962e-06, "loss": 0.931, "step": 11176 }, { "epoch": 0.44, "learning_rate": 3.2462394767559714e-06, "loss": 0.9505, "step": 11177 }, { "epoch": 0.44, "learning_rate": 3.2446498159692884e-06, "loss": 0.9438, "step": 11178 }, { "epoch": 0.44, "learning_rate": 3.2430604691247933e-06, "loss": 0.9693, "step": 11179 }, { "epoch": 0.44, "learning_rate": 3.2414714362963518e-06, "loss": 0.9139, "step": 11180 }, { "epoch": 0.44, "learning_rate": 3.239882717557804e-06, "loss": 0.912, "step": 11181 }, { "epoch": 0.44, "learning_rate": 3.238294312982987e-06, "loss": 0.9073, "step": 11182 }, { "epoch": 0.44, "learning_rate": 3.2367062226457167e-06, "loss": 0.9282, "step": 11183 }, { "epoch": 0.44, "learning_rate": 3.2351184466198028e-06, "loss": 0.9047, "step": 11184 }, { "epoch": 0.44, "learning_rate": 3.233530984979022e-06, "loss": 0.9807, "step": 11185 }, { "epoch": 0.44, "learning_rate": 3.231943837797162e-06, "loss": 0.9185, "step": 11186 }, { "epoch": 0.44, "learning_rate": 3.230357005147978e-06, "loss": 0.9631, "step": 11187 }, { "epoch": 0.44, "learning_rate": 3.2287704871052062e-06, "loss": 0.9436, "step": 11188 }, { "epoch": 0.44, "learning_rate": 3.2271842837425917e-06, "loss": 0.9439, "step": 11189 }, { "epoch": 0.44, "learning_rate": 3.2255983951338387e-06, "loss": 0.9295, "step": 11190 }, { "epoch": 0.44, "learning_rate": 3.224012821352652e-06, "loss": 0.9651, "step": 11191 }, { "epoch": 0.44, "learning_rate": 3.2224275624727207e-06, "loss": 0.877, "step": 11192 }, { "epoch": 0.44, "learning_rate": 3.2208426185677166e-06, "loss": 0.9473, "step": 11193 }, { "epoch": 0.44, "learning_rate": 3.219257989711293e-06, "loss": 0.934, "step": 11194 }, { "epoch": 0.44, "learning_rate": 3.2176736759770945e-06, "loss": 0.9189, "step": 11195 }, { "epoch": 0.44, "learning_rate": 3.216089677438753e-06, "loss": 0.8967, "step": 11196 }, { "epoch": 0.44, "learning_rate": 3.2145059941698744e-06, "loss": 0.946, "step": 11197 }, { "epoch": 0.45, "learning_rate": 3.2129226262440614e-06, "loss": 0.947, "step": 11198 }, { "epoch": 0.45, "learning_rate": 3.2113395737348984e-06, "loss": 0.9285, "step": 11199 }, { "epoch": 0.45, "learning_rate": 3.2097568367159536e-06, "loss": 0.9292, "step": 11200 }, { "epoch": 0.45, "learning_rate": 3.208174415260783e-06, "loss": 0.9423, "step": 11201 }, { "epoch": 0.45, "learning_rate": 3.2065923094429286e-06, "loss": 0.9547, "step": 11202 }, { "epoch": 0.45, "learning_rate": 3.2050105193359104e-06, "loss": 0.9348, "step": 11203 }, { "epoch": 0.45, "learning_rate": 3.203429045013242e-06, "loss": 0.9587, "step": 11204 }, { "epoch": 0.45, "learning_rate": 3.2018478865484228e-06, "loss": 0.9659, "step": 11205 }, { "epoch": 0.45, "learning_rate": 3.2002670440149277e-06, "loss": 0.9418, "step": 11206 }, { "epoch": 0.45, "learning_rate": 3.198686517486226e-06, "loss": 0.9186, "step": 11207 }, { "epoch": 0.45, "learning_rate": 3.19710630703577e-06, "loss": 0.9547, "step": 11208 }, { "epoch": 0.45, "learning_rate": 3.1955264127370002e-06, "loss": 0.9516, "step": 11209 }, { "epoch": 0.45, "learning_rate": 3.193946834663331e-06, "loss": 0.9194, "step": 11210 }, { "epoch": 0.45, "learning_rate": 3.1923675728881808e-06, "loss": 0.9022, "step": 11211 }, { "epoch": 0.45, "learning_rate": 3.1907886274849377e-06, "loss": 0.9289, "step": 11212 }, { "epoch": 0.45, "learning_rate": 3.1892099985269743e-06, "loss": 0.9356, "step": 11213 }, { "epoch": 0.45, "learning_rate": 3.1876316860876655e-06, "loss": 0.939, "step": 11214 }, { "epoch": 0.45, "learning_rate": 3.1860536902403526e-06, "loss": 0.9345, "step": 11215 }, { "epoch": 0.45, "learning_rate": 3.184476011058373e-06, "loss": 0.9219, "step": 11216 }, { "epoch": 0.45, "learning_rate": 3.182898648615045e-06, "loss": 0.9366, "step": 11217 }, { "epoch": 0.45, "learning_rate": 3.181321602983677e-06, "loss": 0.9509, "step": 11218 }, { "epoch": 0.45, "learning_rate": 3.1797448742375535e-06, "loss": 0.9077, "step": 11219 }, { "epoch": 0.45, "learning_rate": 3.178168462449954e-06, "loss": 0.9551, "step": 11220 }, { "epoch": 0.45, "learning_rate": 3.176592367694141e-06, "loss": 0.9379, "step": 11221 }, { "epoch": 0.45, "learning_rate": 3.1750165900433515e-06, "loss": 0.9159, "step": 11222 }, { "epoch": 0.45, "learning_rate": 3.17344112957083e-06, "loss": 0.9221, "step": 11223 }, { "epoch": 0.45, "learning_rate": 3.1718659863497837e-06, "loss": 0.9571, "step": 11224 }, { "epoch": 0.45, "learning_rate": 3.170291160453416e-06, "loss": 0.8772, "step": 11225 }, { "epoch": 0.45, "learning_rate": 3.1687166519549185e-06, "loss": 0.9475, "step": 11226 }, { "epoch": 0.45, "learning_rate": 3.1671424609274558e-06, "loss": 0.9551, "step": 11227 }, { "epoch": 0.45, "learning_rate": 3.165568587444191e-06, "loss": 0.9472, "step": 11228 }, { "epoch": 0.45, "learning_rate": 3.1639950315782643e-06, "loss": 0.8954, "step": 11229 }, { "epoch": 0.45, "learning_rate": 3.162421793402809e-06, "loss": 0.92, "step": 11230 }, { "epoch": 0.45, "learning_rate": 3.1608488729909283e-06, "loss": 0.9408, "step": 11231 }, { "epoch": 0.45, "learning_rate": 3.159276270415732e-06, "loss": 0.922, "step": 11232 }, { "epoch": 0.45, "learning_rate": 3.1577039857502965e-06, "loss": 0.8958, "step": 11233 }, { "epoch": 0.45, "learning_rate": 3.1561320190676927e-06, "loss": 0.9522, "step": 11234 }, { "epoch": 0.45, "learning_rate": 3.154560370440978e-06, "loss": 0.971, "step": 11235 }, { "epoch": 0.45, "learning_rate": 3.152989039943185e-06, "loss": 0.911, "step": 11236 }, { "epoch": 0.45, "learning_rate": 3.1514180276473427e-06, "loss": 0.9063, "step": 11237 }, { "epoch": 0.45, "learning_rate": 3.1498473336264613e-06, "loss": 0.9948, "step": 11238 }, { "epoch": 0.45, "learning_rate": 3.1482769579535365e-06, "loss": 0.9686, "step": 11239 }, { "epoch": 0.45, "learning_rate": 3.146706900701545e-06, "loss": 0.9027, "step": 11240 }, { "epoch": 0.45, "learning_rate": 3.1451371619434534e-06, "loss": 0.9482, "step": 11241 }, { "epoch": 0.45, "learning_rate": 3.143567741752217e-06, "loss": 0.8949, "step": 11242 }, { "epoch": 0.45, "learning_rate": 3.1419986402007653e-06, "loss": 0.9631, "step": 11243 }, { "epoch": 0.45, "learning_rate": 3.140429857362021e-06, "loss": 0.9507, "step": 11244 }, { "epoch": 0.45, "learning_rate": 3.1388613933088906e-06, "loss": 0.9501, "step": 11245 }, { "epoch": 0.45, "learning_rate": 3.137293248114267e-06, "loss": 0.918, "step": 11246 }, { "epoch": 0.45, "learning_rate": 3.135725421851026e-06, "loss": 0.9541, "step": 11247 }, { "epoch": 0.45, "learning_rate": 3.134157914592032e-06, "loss": 0.9601, "step": 11248 }, { "epoch": 0.45, "learning_rate": 3.1325907264101265e-06, "loss": 0.9317, "step": 11249 }, { "epoch": 0.45, "learning_rate": 3.1310238573781437e-06, "loss": 0.9164, "step": 11250 }, { "epoch": 0.45, "learning_rate": 3.129457307568905e-06, "loss": 0.9479, "step": 11251 }, { "epoch": 0.45, "learning_rate": 3.1278910770552074e-06, "loss": 1.0102, "step": 11252 }, { "epoch": 0.45, "learning_rate": 3.1263251659098393e-06, "loss": 0.9213, "step": 11253 }, { "epoch": 0.45, "learning_rate": 3.1247595742055747e-06, "loss": 0.932, "step": 11254 }, { "epoch": 0.45, "learning_rate": 3.1231943020151758e-06, "loss": 0.9349, "step": 11255 }, { "epoch": 0.45, "learning_rate": 3.1216293494113738e-06, "loss": 0.954, "step": 11256 }, { "epoch": 0.45, "learning_rate": 3.120064716466912e-06, "loss": 0.9457, "step": 11257 }, { "epoch": 0.45, "learning_rate": 3.118500403254492e-06, "loss": 0.9786, "step": 11258 }, { "epoch": 0.45, "learning_rate": 3.116936409846817e-06, "loss": 0.9255, "step": 11259 }, { "epoch": 0.45, "learning_rate": 3.1153727363165744e-06, "loss": 0.8845, "step": 11260 }, { "epoch": 0.45, "learning_rate": 3.1138093827364245e-06, "loss": 0.9347, "step": 11261 }, { "epoch": 0.45, "learning_rate": 3.1122463491790254e-06, "loss": 0.9478, "step": 11262 }, { "epoch": 0.45, "learning_rate": 3.110683635717018e-06, "loss": 0.874, "step": 11263 }, { "epoch": 0.45, "learning_rate": 3.1091212424230267e-06, "loss": 0.9003, "step": 11264 }, { "epoch": 0.45, "learning_rate": 3.107559169369656e-06, "loss": 0.9197, "step": 11265 }, { "epoch": 0.45, "learning_rate": 3.1059974166295025e-06, "loss": 0.9514, "step": 11266 }, { "epoch": 0.45, "learning_rate": 3.1044359842751506e-06, "loss": 0.9156, "step": 11267 }, { "epoch": 0.45, "learning_rate": 3.1028748723791534e-06, "loss": 0.9497, "step": 11268 }, { "epoch": 0.45, "learning_rate": 3.101314081014075e-06, "loss": 0.902, "step": 11269 }, { "epoch": 0.45, "learning_rate": 3.0997536102524404e-06, "loss": 0.9158, "step": 11270 }, { "epoch": 0.45, "learning_rate": 3.0981934601667717e-06, "loss": 0.9322, "step": 11271 }, { "epoch": 0.45, "learning_rate": 3.0966336308295785e-06, "loss": 0.9568, "step": 11272 }, { "epoch": 0.45, "learning_rate": 3.095074122313343e-06, "loss": 0.8692, "step": 11273 }, { "epoch": 0.45, "learning_rate": 3.0935149346905456e-06, "loss": 0.9704, "step": 11274 }, { "epoch": 0.45, "learning_rate": 3.0919560680336447e-06, "loss": 0.9631, "step": 11275 }, { "epoch": 0.45, "learning_rate": 3.0903975224150896e-06, "loss": 0.9298, "step": 11276 }, { "epoch": 0.45, "learning_rate": 3.088839297907301e-06, "loss": 0.9683, "step": 11277 }, { "epoch": 0.45, "learning_rate": 3.0872813945827076e-06, "loss": 0.8988, "step": 11278 }, { "epoch": 0.45, "learning_rate": 3.0857238125137013e-06, "loss": 0.9315, "step": 11279 }, { "epoch": 0.45, "learning_rate": 3.084166551772668e-06, "loss": 0.9183, "step": 11280 }, { "epoch": 0.45, "learning_rate": 3.082609612431986e-06, "loss": 0.9451, "step": 11281 }, { "epoch": 0.45, "learning_rate": 3.081052994564001e-06, "loss": 0.9503, "step": 11282 }, { "epoch": 0.45, "learning_rate": 3.0794966982410592e-06, "loss": 0.9618, "step": 11283 }, { "epoch": 0.45, "learning_rate": 3.077940723535485e-06, "loss": 0.9406, "step": 11284 }, { "epoch": 0.45, "learning_rate": 3.0763850705195943e-06, "loss": 0.9352, "step": 11285 }, { "epoch": 0.45, "learning_rate": 3.0748297392656746e-06, "loss": 0.9174, "step": 11286 }, { "epoch": 0.45, "learning_rate": 3.073274729846013e-06, "loss": 0.9453, "step": 11287 }, { "epoch": 0.45, "learning_rate": 3.071720042332876e-06, "loss": 0.9289, "step": 11288 }, { "epoch": 0.45, "learning_rate": 3.0701656767985075e-06, "loss": 0.9211, "step": 11289 }, { "epoch": 0.45, "learning_rate": 3.068611633315155e-06, "loss": 0.9479, "step": 11290 }, { "epoch": 0.45, "learning_rate": 3.0670579119550302e-06, "loss": 0.9566, "step": 11291 }, { "epoch": 0.45, "learning_rate": 3.065504512790344e-06, "loss": 0.9596, "step": 11292 }, { "epoch": 0.45, "learning_rate": 3.0639514358932864e-06, "loss": 0.9392, "step": 11293 }, { "epoch": 0.45, "learning_rate": 3.062398681336036e-06, "loss": 0.9409, "step": 11294 }, { "epoch": 0.45, "learning_rate": 3.060846249190751e-06, "loss": 0.9153, "step": 11295 }, { "epoch": 0.45, "learning_rate": 3.0592941395295782e-06, "loss": 0.9415, "step": 11296 }, { "epoch": 0.45, "learning_rate": 3.0577423524246518e-06, "loss": 0.9542, "step": 11297 }, { "epoch": 0.45, "learning_rate": 3.056190887948084e-06, "loss": 0.8928, "step": 11298 }, { "epoch": 0.45, "learning_rate": 3.054639746171979e-06, "loss": 0.9134, "step": 11299 }, { "epoch": 0.45, "learning_rate": 3.0530889271684217e-06, "loss": 0.9183, "step": 11300 }, { "epoch": 0.45, "learning_rate": 3.0515384310094877e-06, "loss": 0.9302, "step": 11301 }, { "epoch": 0.45, "learning_rate": 3.049988257767222e-06, "loss": 0.9432, "step": 11302 }, { "epoch": 0.45, "learning_rate": 3.048438407513683e-06, "loss": 0.9262, "step": 11303 }, { "epoch": 0.45, "learning_rate": 3.046888880320883e-06, "loss": 0.9065, "step": 11304 }, { "epoch": 0.45, "learning_rate": 3.045339676260839e-06, "loss": 0.9163, "step": 11305 }, { "epoch": 0.45, "learning_rate": 3.04379079540555e-06, "loss": 0.9396, "step": 11306 }, { "epoch": 0.45, "learning_rate": 3.042242237826991e-06, "loss": 0.9316, "step": 11307 }, { "epoch": 0.45, "learning_rate": 3.040694003597131e-06, "loss": 0.9089, "step": 11308 }, { "epoch": 0.45, "learning_rate": 3.0391460927879225e-06, "loss": 0.9202, "step": 11309 }, { "epoch": 0.45, "learning_rate": 3.037598505471303e-06, "loss": 0.9458, "step": 11310 }, { "epoch": 0.45, "learning_rate": 3.036051241719189e-06, "loss": 0.9565, "step": 11311 }, { "epoch": 0.45, "learning_rate": 3.034504301603489e-06, "loss": 0.8992, "step": 11312 }, { "epoch": 0.45, "learning_rate": 3.032957685196097e-06, "loss": 0.9535, "step": 11313 }, { "epoch": 0.45, "learning_rate": 3.0314113925688814e-06, "loss": 0.892, "step": 11314 }, { "epoch": 0.45, "learning_rate": 3.0298654237937144e-06, "loss": 0.94, "step": 11315 }, { "epoch": 0.45, "learning_rate": 3.0283197789424324e-06, "loss": 0.9287, "step": 11316 }, { "epoch": 0.45, "learning_rate": 3.0267744580868694e-06, "loss": 0.9318, "step": 11317 }, { "epoch": 0.45, "learning_rate": 3.0252294612988454e-06, "loss": 0.9343, "step": 11318 }, { "epoch": 0.45, "learning_rate": 3.023684788650154e-06, "loss": 0.9099, "step": 11319 }, { "epoch": 0.45, "learning_rate": 3.0221404402125854e-06, "loss": 0.928, "step": 11320 }, { "epoch": 0.45, "learning_rate": 3.020596416057909e-06, "loss": 0.8988, "step": 11321 }, { "epoch": 0.45, "learning_rate": 3.0190527162578843e-06, "loss": 0.9182, "step": 11322 }, { "epoch": 0.45, "learning_rate": 3.0175093408842413e-06, "loss": 0.9311, "step": 11323 }, { "epoch": 0.45, "learning_rate": 3.0159662900087193e-06, "loss": 0.9852, "step": 11324 }, { "epoch": 0.45, "learning_rate": 3.014423563703018e-06, "loss": 0.9098, "step": 11325 }, { "epoch": 0.45, "learning_rate": 3.0128811620388378e-06, "loss": 0.8991, "step": 11326 }, { "epoch": 0.45, "learning_rate": 3.01133908508786e-06, "loss": 0.9311, "step": 11327 }, { "epoch": 0.45, "learning_rate": 3.0097973329217447e-06, "loss": 0.9387, "step": 11328 }, { "epoch": 0.45, "learning_rate": 3.008255905612144e-06, "loss": 0.9164, "step": 11329 }, { "epoch": 0.45, "learning_rate": 3.0067148032306936e-06, "loss": 0.94, "step": 11330 }, { "epoch": 0.45, "learning_rate": 3.005174025849017e-06, "loss": 0.9545, "step": 11331 }, { "epoch": 0.45, "learning_rate": 3.003633573538712e-06, "loss": 0.9326, "step": 11332 }, { "epoch": 0.45, "learning_rate": 3.002093446371371e-06, "loss": 0.9376, "step": 11333 }, { "epoch": 0.45, "learning_rate": 3.0005536444185724e-06, "loss": 0.864, "step": 11334 }, { "epoch": 0.45, "learning_rate": 2.9990141677518657e-06, "loss": 0.9146, "step": 11335 }, { "epoch": 0.45, "learning_rate": 2.9974750164428078e-06, "loss": 0.9672, "step": 11336 }, { "epoch": 0.45, "learning_rate": 2.9959361905629192e-06, "loss": 0.9288, "step": 11337 }, { "epoch": 0.45, "learning_rate": 2.994397690183716e-06, "loss": 0.936, "step": 11338 }, { "epoch": 0.45, "learning_rate": 2.992859515376697e-06, "loss": 0.8935, "step": 11339 }, { "epoch": 0.45, "learning_rate": 2.9913216662133503e-06, "loss": 0.9653, "step": 11340 }, { "epoch": 0.45, "learning_rate": 2.9897841427651363e-06, "loss": 0.9411, "step": 11341 }, { "epoch": 0.45, "learning_rate": 2.9882469451035146e-06, "loss": 0.9579, "step": 11342 }, { "epoch": 0.45, "learning_rate": 2.9867100732999243e-06, "loss": 0.958, "step": 11343 }, { "epoch": 0.45, "learning_rate": 2.985173527425783e-06, "loss": 0.9217, "step": 11344 }, { "epoch": 0.45, "learning_rate": 2.9836373075525014e-06, "loss": 0.9871, "step": 11345 }, { "epoch": 0.45, "learning_rate": 2.982101413751474e-06, "loss": 0.9828, "step": 11346 }, { "epoch": 0.45, "learning_rate": 2.9805658460940802e-06, "loss": 0.9143, "step": 11347 }, { "epoch": 0.46, "learning_rate": 2.9790306046516735e-06, "loss": 0.941, "step": 11348 }, { "epoch": 0.46, "learning_rate": 2.9774956894956153e-06, "loss": 0.903, "step": 11349 }, { "epoch": 0.46, "learning_rate": 2.9759611006972266e-06, "loss": 0.9327, "step": 11350 }, { "epoch": 0.46, "learning_rate": 2.9744268383278284e-06, "loss": 0.9183, "step": 11351 }, { "epoch": 0.46, "learning_rate": 2.972892902458726e-06, "loss": 0.9153, "step": 11352 }, { "epoch": 0.46, "learning_rate": 2.971359293161199e-06, "loss": 0.9535, "step": 11353 }, { "epoch": 0.46, "learning_rate": 2.9698260105065236e-06, "loss": 0.9034, "step": 11354 }, { "epoch": 0.46, "learning_rate": 2.968293054565956e-06, "loss": 0.9474, "step": 11355 }, { "epoch": 0.46, "learning_rate": 2.9667604254107386e-06, "loss": 0.9412, "step": 11356 }, { "epoch": 0.46, "learning_rate": 2.965228123112094e-06, "loss": 0.9539, "step": 11357 }, { "epoch": 0.46, "learning_rate": 2.963696147741234e-06, "loss": 0.9613, "step": 11358 }, { "epoch": 0.46, "learning_rate": 2.9621644993693587e-06, "loss": 0.9722, "step": 11359 }, { "epoch": 0.46, "learning_rate": 2.9606331780676376e-06, "loss": 0.9227, "step": 11360 }, { "epoch": 0.46, "learning_rate": 2.959102183907251e-06, "loss": 0.9312, "step": 11361 }, { "epoch": 0.46, "learning_rate": 2.957571516959338e-06, "loss": 0.8596, "step": 11362 }, { "epoch": 0.46, "learning_rate": 2.956041177295036e-06, "loss": 0.9504, "step": 11363 }, { "epoch": 0.46, "learning_rate": 2.9545111649854687e-06, "loss": 0.9539, "step": 11364 }, { "epoch": 0.46, "learning_rate": 2.9529814801017355e-06, "loss": 0.9409, "step": 11365 }, { "epoch": 0.46, "learning_rate": 2.951452122714926e-06, "loss": 0.9618, "step": 11366 }, { "epoch": 0.46, "learning_rate": 2.9499230928961166e-06, "loss": 0.944, "step": 11367 }, { "epoch": 0.46, "learning_rate": 2.948394390716367e-06, "loss": 0.9438, "step": 11368 }, { "epoch": 0.46, "learning_rate": 2.9468660162467132e-06, "loss": 0.9226, "step": 11369 }, { "epoch": 0.46, "learning_rate": 2.9453379695581964e-06, "loss": 0.9126, "step": 11370 }, { "epoch": 0.46, "learning_rate": 2.943810250721818e-06, "loss": 0.9526, "step": 11371 }, { "epoch": 0.46, "learning_rate": 2.9422828598085806e-06, "loss": 0.9647, "step": 11372 }, { "epoch": 0.46, "learning_rate": 2.9407557968894706e-06, "loss": 0.9812, "step": 11373 }, { "epoch": 0.46, "learning_rate": 2.939229062035447e-06, "loss": 0.8828, "step": 11374 }, { "epoch": 0.46, "learning_rate": 2.937702655317467e-06, "loss": 0.916, "step": 11375 }, { "epoch": 0.46, "learning_rate": 2.9361765768064665e-06, "loss": 0.903, "step": 11376 }, { "epoch": 0.46, "learning_rate": 2.9346508265733708e-06, "loss": 0.9464, "step": 11377 }, { "epoch": 0.46, "learning_rate": 2.9331254046890767e-06, "loss": 0.9536, "step": 11378 }, { "epoch": 0.46, "learning_rate": 2.9316003112244874e-06, "loss": 0.9392, "step": 11379 }, { "epoch": 0.46, "learning_rate": 2.9300755462504724e-06, "loss": 0.942, "step": 11380 }, { "epoch": 0.46, "learning_rate": 2.9285511098378873e-06, "loss": 0.9474, "step": 11381 }, { "epoch": 0.46, "learning_rate": 2.9270270020575885e-06, "loss": 0.9046, "step": 11382 }, { "epoch": 0.46, "learning_rate": 2.925503222980398e-06, "loss": 0.9148, "step": 11383 }, { "epoch": 0.46, "learning_rate": 2.9239797726771323e-06, "loss": 0.9691, "step": 11384 }, { "epoch": 0.46, "learning_rate": 2.9224566512185916e-06, "loss": 0.9447, "step": 11385 }, { "epoch": 0.46, "learning_rate": 2.920933858675563e-06, "loss": 0.9564, "step": 11386 }, { "epoch": 0.46, "learning_rate": 2.91941139511881e-06, "loss": 0.966, "step": 11387 }, { "epoch": 0.46, "learning_rate": 2.917889260619088e-06, "loss": 0.9333, "step": 11388 }, { "epoch": 0.46, "learning_rate": 2.9163674552471387e-06, "loss": 0.9283, "step": 11389 }, { "epoch": 0.46, "learning_rate": 2.9148459790736805e-06, "loss": 0.9446, "step": 11390 }, { "epoch": 0.46, "learning_rate": 2.913324832169422e-06, "loss": 0.9424, "step": 11391 }, { "epoch": 0.46, "learning_rate": 2.9118040146050574e-06, "loss": 0.9337, "step": 11392 }, { "epoch": 0.46, "learning_rate": 2.910283526451262e-06, "loss": 0.9794, "step": 11393 }, { "epoch": 0.46, "learning_rate": 2.9087633677786996e-06, "loss": 0.9208, "step": 11394 }, { "epoch": 0.46, "learning_rate": 2.907243538658019e-06, "loss": 0.9204, "step": 11395 }, { "epoch": 0.46, "learning_rate": 2.9057240391598453e-06, "loss": 0.9231, "step": 11396 }, { "epoch": 0.46, "learning_rate": 2.9042048693547974e-06, "loss": 0.9207, "step": 11397 }, { "epoch": 0.46, "learning_rate": 2.9026860293134786e-06, "loss": 0.9864, "step": 11398 }, { "epoch": 0.46, "learning_rate": 2.9011675191064694e-06, "loss": 0.9138, "step": 11399 }, { "epoch": 0.46, "learning_rate": 2.89964933880434e-06, "loss": 0.9247, "step": 11400 }, { "epoch": 0.46, "learning_rate": 2.898131488477648e-06, "loss": 0.9573, "step": 11401 }, { "epoch": 0.46, "learning_rate": 2.896613968196933e-06, "loss": 0.9525, "step": 11402 }, { "epoch": 0.46, "learning_rate": 2.8950967780327145e-06, "loss": 0.9654, "step": 11403 }, { "epoch": 0.46, "learning_rate": 2.8935799180555047e-06, "loss": 0.9202, "step": 11404 }, { "epoch": 0.46, "learning_rate": 2.8920633883357986e-06, "loss": 0.9426, "step": 11405 }, { "epoch": 0.46, "learning_rate": 2.890547188944064e-06, "loss": 0.959, "step": 11406 }, { "epoch": 0.46, "learning_rate": 2.889031319950778e-06, "loss": 0.9583, "step": 11407 }, { "epoch": 0.46, "learning_rate": 2.8875157814263768e-06, "loss": 0.9371, "step": 11408 }, { "epoch": 0.46, "learning_rate": 2.8860005734412964e-06, "loss": 0.9595, "step": 11409 }, { "epoch": 0.46, "learning_rate": 2.884485696065955e-06, "loss": 0.9491, "step": 11410 }, { "epoch": 0.46, "learning_rate": 2.882971149370749e-06, "loss": 0.9575, "step": 11411 }, { "epoch": 0.46, "learning_rate": 2.8814569334260667e-06, "loss": 0.9327, "step": 11412 }, { "epoch": 0.46, "learning_rate": 2.879943048302278e-06, "loss": 0.916, "step": 11413 }, { "epoch": 0.46, "learning_rate": 2.878429494069741e-06, "loss": 0.9191, "step": 11414 }, { "epoch": 0.46, "learning_rate": 2.8769162707987862e-06, "loss": 0.8946, "step": 11415 }, { "epoch": 0.46, "learning_rate": 2.87540337855975e-06, "loss": 0.9153, "step": 11416 }, { "epoch": 0.46, "learning_rate": 2.873890817422933e-06, "loss": 0.9552, "step": 11417 }, { "epoch": 0.46, "learning_rate": 2.8723785874586295e-06, "loss": 0.9523, "step": 11418 }, { "epoch": 0.46, "learning_rate": 2.870866688737123e-06, "loss": 0.9492, "step": 11419 }, { "epoch": 0.46, "learning_rate": 2.8693551213286684e-06, "loss": 0.9343, "step": 11420 }, { "epoch": 0.46, "learning_rate": 2.8678438853035175e-06, "loss": 0.9515, "step": 11421 }, { "epoch": 0.46, "learning_rate": 2.8663329807319006e-06, "loss": 0.9073, "step": 11422 }, { "epoch": 0.46, "learning_rate": 2.864822407684038e-06, "loss": 0.9528, "step": 11423 }, { "epoch": 0.46, "learning_rate": 2.863312166230121e-06, "loss": 0.9481, "step": 11424 }, { "epoch": 0.46, "learning_rate": 2.861802256440348e-06, "loss": 0.9176, "step": 11425 }, { "epoch": 0.46, "learning_rate": 2.860292678384884e-06, "loss": 0.9425, "step": 11426 }, { "epoch": 0.46, "learning_rate": 2.858783432133875e-06, "loss": 0.965, "step": 11427 }, { "epoch": 0.46, "learning_rate": 2.8572745177574757e-06, "loss": 0.9355, "step": 11428 }, { "epoch": 0.46, "learning_rate": 2.855765935325798e-06, "loss": 0.9323, "step": 11429 }, { "epoch": 0.46, "learning_rate": 2.8542576849089563e-06, "loss": 0.9614, "step": 11430 }, { "epoch": 0.46, "learning_rate": 2.852749766577041e-06, "loss": 0.9289, "step": 11431 }, { "epoch": 0.46, "learning_rate": 2.8512421804001354e-06, "loss": 0.9107, "step": 11432 }, { "epoch": 0.46, "learning_rate": 2.849734926448294e-06, "loss": 0.9599, "step": 11433 }, { "epoch": 0.46, "learning_rate": 2.8482280047915667e-06, "loss": 0.9459, "step": 11434 }, { "epoch": 0.46, "learning_rate": 2.846721415499989e-06, "loss": 0.9396, "step": 11435 }, { "epoch": 0.46, "learning_rate": 2.8452151586435693e-06, "loss": 0.9607, "step": 11436 }, { "epoch": 0.46, "learning_rate": 2.8437092342923113e-06, "loss": 0.9348, "step": 11437 }, { "epoch": 0.46, "learning_rate": 2.8422036425162013e-06, "loss": 0.9546, "step": 11438 }, { "epoch": 0.46, "learning_rate": 2.8406983833852076e-06, "loss": 0.9447, "step": 11439 }, { "epoch": 0.46, "learning_rate": 2.8391934569692846e-06, "loss": 0.9342, "step": 11440 }, { "epoch": 0.46, "learning_rate": 2.837688863338374e-06, "loss": 0.9704, "step": 11441 }, { "epoch": 0.46, "learning_rate": 2.836184602562393e-06, "loss": 0.9157, "step": 11442 }, { "epoch": 0.46, "learning_rate": 2.834680674711252e-06, "loss": 0.9056, "step": 11443 }, { "epoch": 0.46, "learning_rate": 2.833177079854846e-06, "loss": 0.9573, "step": 11444 }, { "epoch": 0.46, "learning_rate": 2.8316738180630477e-06, "loss": 0.9372, "step": 11445 }, { "epoch": 0.46, "learning_rate": 2.830170889405719e-06, "loss": 0.9176, "step": 11446 }, { "epoch": 0.46, "learning_rate": 2.8286682939527064e-06, "loss": 0.9503, "step": 11447 }, { "epoch": 0.46, "learning_rate": 2.8271660317738437e-06, "loss": 0.9195, "step": 11448 }, { "epoch": 0.46, "learning_rate": 2.8256641029389385e-06, "loss": 0.9374, "step": 11449 }, { "epoch": 0.46, "learning_rate": 2.8241625075177947e-06, "loss": 0.9294, "step": 11450 }, { "epoch": 0.46, "learning_rate": 2.822661245580195e-06, "loss": 0.9545, "step": 11451 }, { "epoch": 0.46, "learning_rate": 2.821160317195909e-06, "loss": 0.9169, "step": 11452 }, { "epoch": 0.46, "learning_rate": 2.819659722434691e-06, "loss": 0.9716, "step": 11453 }, { "epoch": 0.46, "learning_rate": 2.818159461366272e-06, "loss": 0.9129, "step": 11454 }, { "epoch": 0.46, "learning_rate": 2.8166595340603787e-06, "loss": 0.9225, "step": 11455 }, { "epoch": 0.46, "learning_rate": 2.8151599405867204e-06, "loss": 0.9546, "step": 11456 }, { "epoch": 0.46, "learning_rate": 2.813660681014979e-06, "loss": 0.9333, "step": 11457 }, { "epoch": 0.46, "learning_rate": 2.8121617554148363e-06, "loss": 0.9118, "step": 11458 }, { "epoch": 0.46, "learning_rate": 2.8106631638559487e-06, "loss": 0.9274, "step": 11459 }, { "epoch": 0.46, "learning_rate": 2.809164906407967e-06, "loss": 0.9503, "step": 11460 }, { "epoch": 0.46, "learning_rate": 2.807666983140508e-06, "loss": 0.9702, "step": 11461 }, { "epoch": 0.46, "learning_rate": 2.806169394123198e-06, "loss": 0.9281, "step": 11462 }, { "epoch": 0.46, "learning_rate": 2.804672139425626e-06, "loss": 0.9271, "step": 11463 }, { "epoch": 0.46, "learning_rate": 2.803175219117377e-06, "loss": 0.9104, "step": 11464 }, { "epoch": 0.46, "learning_rate": 2.801678633268019e-06, "loss": 0.9753, "step": 11465 }, { "epoch": 0.46, "learning_rate": 2.8001823819470987e-06, "loss": 0.9198, "step": 11466 }, { "epoch": 0.46, "learning_rate": 2.798686465224154e-06, "loss": 0.9021, "step": 11467 }, { "epoch": 0.46, "learning_rate": 2.7971908831687054e-06, "loss": 0.948, "step": 11468 }, { "epoch": 0.46, "learning_rate": 2.7956956358502594e-06, "loss": 0.989, "step": 11469 }, { "epoch": 0.46, "learning_rate": 2.794200723338296e-06, "loss": 0.895, "step": 11470 }, { "epoch": 0.46, "learning_rate": 2.792706145702302e-06, "loss": 0.8885, "step": 11471 }, { "epoch": 0.46, "learning_rate": 2.7912119030117256e-06, "loss": 0.9055, "step": 11472 }, { "epoch": 0.46, "learning_rate": 2.7897179953360064e-06, "loss": 0.9463, "step": 11473 }, { "epoch": 0.46, "learning_rate": 2.788224422744581e-06, "loss": 0.9773, "step": 11474 }, { "epoch": 0.46, "learning_rate": 2.7867311853068524e-06, "loss": 0.9283, "step": 11475 }, { "epoch": 0.46, "learning_rate": 2.7852382830922174e-06, "loss": 0.939, "step": 11476 }, { "epoch": 0.46, "learning_rate": 2.7837457161700588e-06, "loss": 0.9353, "step": 11477 }, { "epoch": 0.46, "learning_rate": 2.7822534846097403e-06, "loss": 0.9542, "step": 11478 }, { "epoch": 0.46, "learning_rate": 2.780761588480607e-06, "loss": 0.9002, "step": 11479 }, { "epoch": 0.46, "learning_rate": 2.7792700278519945e-06, "loss": 0.9378, "step": 11480 }, { "epoch": 0.46, "learning_rate": 2.777778802793223e-06, "loss": 0.9571, "step": 11481 }, { "epoch": 0.46, "learning_rate": 2.7762879133735844e-06, "loss": 0.9526, "step": 11482 }, { "epoch": 0.46, "learning_rate": 2.774797359662379e-06, "loss": 0.9099, "step": 11483 }, { "epoch": 0.46, "learning_rate": 2.773307141728867e-06, "loss": 0.9189, "step": 11484 }, { "epoch": 0.46, "learning_rate": 2.7718172596423067e-06, "loss": 0.9369, "step": 11485 }, { "epoch": 0.46, "learning_rate": 2.7703277134719376e-06, "loss": 0.9341, "step": 11486 }, { "epoch": 0.46, "learning_rate": 2.7688385032869867e-06, "loss": 0.9233, "step": 11487 }, { "epoch": 0.46, "learning_rate": 2.767349629156657e-06, "loss": 0.931, "step": 11488 }, { "epoch": 0.46, "learning_rate": 2.7658610911501426e-06, "loss": 0.9852, "step": 11489 }, { "epoch": 0.46, "learning_rate": 2.7643728893366253e-06, "loss": 0.9753, "step": 11490 }, { "epoch": 0.46, "learning_rate": 2.762885023785259e-06, "loss": 0.9586, "step": 11491 }, { "epoch": 0.46, "learning_rate": 2.7613974945651924e-06, "loss": 0.8939, "step": 11492 }, { "epoch": 0.46, "learning_rate": 2.7599103017455563e-06, "loss": 0.9243, "step": 11493 }, { "epoch": 0.46, "learning_rate": 2.758423445395468e-06, "loss": 0.9064, "step": 11494 }, { "epoch": 0.46, "learning_rate": 2.756936925584022e-06, "loss": 0.962, "step": 11495 }, { "epoch": 0.46, "learning_rate": 2.7554507423803012e-06, "loss": 0.9246, "step": 11496 }, { "epoch": 0.46, "learning_rate": 2.7539648958533747e-06, "loss": 0.9382, "step": 11497 }, { "epoch": 0.46, "learning_rate": 2.7524793860722954e-06, "loss": 0.9679, "step": 11498 }, { "epoch": 0.47, "learning_rate": 2.750994213106102e-06, "loss": 0.9027, "step": 11499 }, { "epoch": 0.47, "learning_rate": 2.749509377023808e-06, "loss": 0.9695, "step": 11500 }, { "epoch": 0.47, "learning_rate": 2.748024877894423e-06, "loss": 0.9542, "step": 11501 }, { "epoch": 0.47, "learning_rate": 2.746540715786937e-06, "loss": 0.9451, "step": 11502 }, { "epoch": 0.47, "learning_rate": 2.7450568907703213e-06, "loss": 0.9341, "step": 11503 }, { "epoch": 0.47, "learning_rate": 2.743573402913533e-06, "loss": 0.9394, "step": 11504 }, { "epoch": 0.47, "learning_rate": 2.7420902522855176e-06, "loss": 0.9423, "step": 11505 }, { "epoch": 0.47, "learning_rate": 2.740607438955203e-06, "loss": 0.9124, "step": 11506 }, { "epoch": 0.47, "learning_rate": 2.7391249629914905e-06, "loss": 0.952, "step": 11507 }, { "epoch": 0.47, "learning_rate": 2.737642824463289e-06, "loss": 0.9256, "step": 11508 }, { "epoch": 0.47, "learning_rate": 2.7361610234394686e-06, "loss": 0.9361, "step": 11509 }, { "epoch": 0.47, "learning_rate": 2.734679559988895e-06, "loss": 0.9463, "step": 11510 }, { "epoch": 0.47, "learning_rate": 2.733198434180422e-06, "loss": 0.9201, "step": 11511 }, { "epoch": 0.47, "learning_rate": 2.731717646082873e-06, "loss": 0.9349, "step": 11512 }, { "epoch": 0.47, "learning_rate": 2.7302371957650707e-06, "loss": 0.9391, "step": 11513 }, { "epoch": 0.47, "learning_rate": 2.7287570832958143e-06, "loss": 0.9101, "step": 11514 }, { "epoch": 0.47, "learning_rate": 2.7272773087438944e-06, "loss": 0.9659, "step": 11515 }, { "epoch": 0.47, "learning_rate": 2.7257978721780685e-06, "loss": 0.9797, "step": 11516 }, { "epoch": 0.47, "learning_rate": 2.724318773667105e-06, "loss": 0.9328, "step": 11517 }, { "epoch": 0.47, "learning_rate": 2.7228400132797362e-06, "loss": 0.9676, "step": 11518 }, { "epoch": 0.47, "learning_rate": 2.721361591084677e-06, "loss": 0.9523, "step": 11519 }, { "epoch": 0.47, "learning_rate": 2.7198835071506478e-06, "loss": 0.9426, "step": 11520 }, { "epoch": 0.47, "learning_rate": 2.718405761546331e-06, "loss": 0.9231, "step": 11521 }, { "epoch": 0.47, "learning_rate": 2.7169283543404047e-06, "loss": 0.9566, "step": 11522 }, { "epoch": 0.47, "learning_rate": 2.715451285601527e-06, "loss": 0.9279, "step": 11523 }, { "epoch": 0.47, "learning_rate": 2.7139745553983464e-06, "loss": 0.9418, "step": 11524 }, { "epoch": 0.47, "learning_rate": 2.712498163799486e-06, "loss": 0.9323, "step": 11525 }, { "epoch": 0.47, "learning_rate": 2.7110221108735602e-06, "loss": 0.9642, "step": 11526 }, { "epoch": 0.47, "learning_rate": 2.7095463966891687e-06, "loss": 0.9162, "step": 11527 }, { "epoch": 0.47, "learning_rate": 2.708071021314883e-06, "loss": 0.9461, "step": 11528 }, { "epoch": 0.47, "learning_rate": 2.7065959848192823e-06, "loss": 0.9107, "step": 11529 }, { "epoch": 0.47, "learning_rate": 2.7051212872709053e-06, "loss": 0.8848, "step": 11530 }, { "epoch": 0.47, "learning_rate": 2.7036469287382894e-06, "loss": 0.934, "step": 11531 }, { "epoch": 0.47, "learning_rate": 2.702172909289953e-06, "loss": 0.9264, "step": 11532 }, { "epoch": 0.47, "learning_rate": 2.700699228994401e-06, "loss": 0.9317, "step": 11533 }, { "epoch": 0.47, "learning_rate": 2.699225887920115e-06, "loss": 0.9639, "step": 11534 }, { "epoch": 0.47, "learning_rate": 2.697752886135566e-06, "loss": 0.9716, "step": 11535 }, { "epoch": 0.47, "learning_rate": 2.6962802237092146e-06, "loss": 0.9392, "step": 11536 }, { "epoch": 0.47, "learning_rate": 2.694807900709494e-06, "loss": 0.9379, "step": 11537 }, { "epoch": 0.47, "learning_rate": 2.693335917204829e-06, "loss": 0.9045, "step": 11538 }, { "epoch": 0.47, "learning_rate": 2.691864273263628e-06, "loss": 0.9447, "step": 11539 }, { "epoch": 0.47, "learning_rate": 2.6903929689542864e-06, "loss": 0.9708, "step": 11540 }, { "epoch": 0.47, "learning_rate": 2.6889220043451727e-06, "loss": 0.9463, "step": 11541 }, { "epoch": 0.47, "learning_rate": 2.687451379504652e-06, "loss": 0.9252, "step": 11542 }, { "epoch": 0.47, "learning_rate": 2.6859810945010687e-06, "loss": 0.9774, "step": 11543 }, { "epoch": 0.47, "learning_rate": 2.684511149402751e-06, "loss": 0.9451, "step": 11544 }, { "epoch": 0.47, "learning_rate": 2.6830415442780143e-06, "loss": 0.9386, "step": 11545 }, { "epoch": 0.47, "learning_rate": 2.6815722791951503e-06, "loss": 0.9229, "step": 11546 }, { "epoch": 0.47, "learning_rate": 2.680103354222444e-06, "loss": 0.9549, "step": 11547 }, { "epoch": 0.47, "learning_rate": 2.6786347694281623e-06, "loss": 0.92, "step": 11548 }, { "epoch": 0.47, "learning_rate": 2.677166524880551e-06, "loss": 0.924, "step": 11549 }, { "epoch": 0.47, "learning_rate": 2.6756986206478452e-06, "loss": 0.9338, "step": 11550 }, { "epoch": 0.47, "learning_rate": 2.674231056798263e-06, "loss": 0.9495, "step": 11551 }, { "epoch": 0.47, "learning_rate": 2.6727638334000118e-06, "loss": 0.9131, "step": 11552 }, { "epoch": 0.47, "learning_rate": 2.671296950521266e-06, "loss": 0.9342, "step": 11553 }, { "epoch": 0.47, "learning_rate": 2.669830408230211e-06, "loss": 0.9034, "step": 11554 }, { "epoch": 0.47, "learning_rate": 2.668364206594991e-06, "loss": 0.9739, "step": 11555 }, { "epoch": 0.47, "learning_rate": 2.6668983456837484e-06, "loss": 0.9466, "step": 11556 }, { "epoch": 0.47, "learning_rate": 2.6654328255646088e-06, "loss": 0.9449, "step": 11557 }, { "epoch": 0.47, "learning_rate": 2.6639676463056743e-06, "loss": 0.9047, "step": 11558 }, { "epoch": 0.47, "learning_rate": 2.662502807975039e-06, "loss": 0.8995, "step": 11559 }, { "epoch": 0.47, "learning_rate": 2.661038310640779e-06, "loss": 0.9475, "step": 11560 }, { "epoch": 0.47, "learning_rate": 2.659574154370955e-06, "loss": 0.9106, "step": 11561 }, { "epoch": 0.47, "learning_rate": 2.658110339233605e-06, "loss": 0.8956, "step": 11562 }, { "epoch": 0.47, "learning_rate": 2.656646865296767e-06, "loss": 0.9497, "step": 11563 }, { "epoch": 0.47, "learning_rate": 2.655183732628448e-06, "loss": 0.8999, "step": 11564 }, { "epoch": 0.47, "learning_rate": 2.653720941296638e-06, "loss": 0.9335, "step": 11565 }, { "epoch": 0.47, "learning_rate": 2.6522584913693295e-06, "loss": 0.935, "step": 11566 }, { "epoch": 0.47, "learning_rate": 2.650796382914478e-06, "loss": 0.8988, "step": 11567 }, { "epoch": 0.47, "learning_rate": 2.6493346160000356e-06, "loss": 0.9552, "step": 11568 }, { "epoch": 0.47, "learning_rate": 2.6478731906939335e-06, "loss": 0.9472, "step": 11569 }, { "epoch": 0.47, "learning_rate": 2.6464121070640947e-06, "loss": 0.9254, "step": 11570 }, { "epoch": 0.47, "learning_rate": 2.6449513651784133e-06, "loss": 0.9277, "step": 11571 }, { "epoch": 0.47, "learning_rate": 2.6434909651047758e-06, "loss": 0.8858, "step": 11572 }, { "epoch": 0.47, "learning_rate": 2.6420309069110563e-06, "loss": 0.948, "step": 11573 }, { "epoch": 0.47, "learning_rate": 2.6405711906650975e-06, "loss": 0.9325, "step": 11574 }, { "epoch": 0.47, "learning_rate": 2.639111816434752e-06, "loss": 0.958, "step": 11575 }, { "epoch": 0.47, "learning_rate": 2.63765278428783e-06, "loss": 0.9748, "step": 11576 }, { "epoch": 0.47, "learning_rate": 2.6361940942921406e-06, "loss": 0.9184, "step": 11577 }, { "epoch": 0.47, "learning_rate": 2.634735746515473e-06, "loss": 0.9063, "step": 11578 }, { "epoch": 0.47, "learning_rate": 2.6332777410256074e-06, "loss": 0.9752, "step": 11579 }, { "epoch": 0.47, "learning_rate": 2.631820077890291e-06, "loss": 0.9743, "step": 11580 }, { "epoch": 0.47, "learning_rate": 2.630362757177274e-06, "loss": 0.9217, "step": 11581 }, { "epoch": 0.47, "learning_rate": 2.6289057789542816e-06, "loss": 0.9069, "step": 11582 }, { "epoch": 0.47, "learning_rate": 2.6274491432890205e-06, "loss": 0.9106, "step": 11583 }, { "epoch": 0.47, "learning_rate": 2.625992850249187e-06, "loss": 0.957, "step": 11584 }, { "epoch": 0.47, "learning_rate": 2.62453689990246e-06, "loss": 0.9309, "step": 11585 }, { "epoch": 0.47, "learning_rate": 2.6230812923165017e-06, "loss": 0.9559, "step": 11586 }, { "epoch": 0.47, "learning_rate": 2.6216260275589622e-06, "loss": 0.9419, "step": 11587 }, { "epoch": 0.47, "learning_rate": 2.6201711056974664e-06, "loss": 0.8859, "step": 11588 }, { "epoch": 0.47, "learning_rate": 2.6187165267996305e-06, "loss": 0.9644, "step": 11589 }, { "epoch": 0.47, "learning_rate": 2.617262290933056e-06, "loss": 0.9081, "step": 11590 }, { "epoch": 0.47, "learning_rate": 2.6158083981653258e-06, "loss": 0.907, "step": 11591 }, { "epoch": 0.47, "learning_rate": 2.6143548485640024e-06, "loss": 0.9423, "step": 11592 }, { "epoch": 0.47, "learning_rate": 2.6129016421966402e-06, "loss": 0.9615, "step": 11593 }, { "epoch": 0.47, "learning_rate": 2.6114487791307765e-06, "loss": 0.8917, "step": 11594 }, { "epoch": 0.47, "learning_rate": 2.6099962594339245e-06, "loss": 0.9124, "step": 11595 }, { "epoch": 0.47, "learning_rate": 2.6085440831735897e-06, "loss": 0.9273, "step": 11596 }, { "epoch": 0.47, "learning_rate": 2.607092250417261e-06, "loss": 0.9405, "step": 11597 }, { "epoch": 0.47, "learning_rate": 2.6056407612324106e-06, "loss": 0.9093, "step": 11598 }, { "epoch": 0.47, "learning_rate": 2.6041896156864864e-06, "loss": 0.9219, "step": 11599 }, { "epoch": 0.47, "learning_rate": 2.602738813846938e-06, "loss": 0.9094, "step": 11600 }, { "epoch": 0.47, "learning_rate": 2.601288355781182e-06, "loss": 0.9302, "step": 11601 }, { "epoch": 0.47, "learning_rate": 2.5998382415566258e-06, "loss": 0.9151, "step": 11602 }, { "epoch": 0.47, "learning_rate": 2.598388471240666e-06, "loss": 0.9256, "step": 11603 }, { "epoch": 0.47, "learning_rate": 2.5969390449006703e-06, "loss": 0.9685, "step": 11604 }, { "epoch": 0.47, "learning_rate": 2.5954899626040033e-06, "loss": 0.9454, "step": 11605 }, { "epoch": 0.47, "learning_rate": 2.5940412244180058e-06, "loss": 0.9202, "step": 11606 }, { "epoch": 0.47, "learning_rate": 2.59259283041001e-06, "loss": 0.9837, "step": 11607 }, { "epoch": 0.47, "learning_rate": 2.5911447806473177e-06, "loss": 0.9332, "step": 11608 }, { "epoch": 0.47, "learning_rate": 2.5896970751972363e-06, "loss": 0.9049, "step": 11609 }, { "epoch": 0.47, "learning_rate": 2.588249714127039e-06, "loss": 0.9251, "step": 11610 }, { "epoch": 0.47, "learning_rate": 2.586802697503983e-06, "loss": 0.9342, "step": 11611 }, { "epoch": 0.47, "learning_rate": 2.585356025395328e-06, "loss": 0.9197, "step": 11612 }, { "epoch": 0.47, "learning_rate": 2.5839096978682963e-06, "loss": 1.003, "step": 11613 }, { "epoch": 0.47, "learning_rate": 2.5824637149901067e-06, "loss": 0.9077, "step": 11614 }, { "epoch": 0.47, "learning_rate": 2.581018076827958e-06, "loss": 0.9633, "step": 11615 }, { "epoch": 0.47, "learning_rate": 2.579572783449037e-06, "loss": 0.9185, "step": 11616 }, { "epoch": 0.47, "learning_rate": 2.578127834920504e-06, "loss": 0.904, "step": 11617 }, { "epoch": 0.47, "learning_rate": 2.576683231309515e-06, "loss": 0.9499, "step": 11618 }, { "epoch": 0.47, "learning_rate": 2.575238972683206e-06, "loss": 0.9165, "step": 11619 }, { "epoch": 0.47, "learning_rate": 2.573795059108689e-06, "loss": 0.9387, "step": 11620 }, { "epoch": 0.47, "learning_rate": 2.572351490653078e-06, "loss": 0.9349, "step": 11621 }, { "epoch": 0.47, "learning_rate": 2.5709082673834517e-06, "loss": 0.9141, "step": 11622 }, { "epoch": 0.47, "learning_rate": 2.5694653893668854e-06, "loss": 0.9253, "step": 11623 }, { "epoch": 0.47, "learning_rate": 2.5680228566704314e-06, "loss": 0.9473, "step": 11624 }, { "epoch": 0.47, "learning_rate": 2.5665806693611327e-06, "loss": 0.9439, "step": 11625 }, { "epoch": 0.47, "learning_rate": 2.5651388275060083e-06, "loss": 0.9261, "step": 11626 }, { "epoch": 0.47, "learning_rate": 2.5636973311720657e-06, "loss": 0.9419, "step": 11627 }, { "epoch": 0.47, "learning_rate": 2.5622561804262993e-06, "loss": 0.9379, "step": 11628 }, { "epoch": 0.47, "learning_rate": 2.560815375335679e-06, "loss": 0.9207, "step": 11629 }, { "epoch": 0.47, "learning_rate": 2.5593749159671644e-06, "loss": 0.9424, "step": 11630 }, { "epoch": 0.47, "learning_rate": 2.5579348023877005e-06, "loss": 0.9069, "step": 11631 }, { "epoch": 0.47, "learning_rate": 2.5564950346642115e-06, "loss": 0.9364, "step": 11632 }, { "epoch": 0.47, "learning_rate": 2.555055612863613e-06, "loss": 0.9137, "step": 11633 }, { "epoch": 0.47, "learning_rate": 2.553616537052792e-06, "loss": 0.961, "step": 11634 }, { "epoch": 0.47, "learning_rate": 2.5521778072986303e-06, "loss": 0.904, "step": 11635 }, { "epoch": 0.47, "learning_rate": 2.5507394236679905e-06, "loss": 0.9665, "step": 11636 }, { "epoch": 0.47, "learning_rate": 2.5493013862277215e-06, "loss": 0.9292, "step": 11637 }, { "epoch": 0.47, "learning_rate": 2.547863695044648e-06, "loss": 0.9284, "step": 11638 }, { "epoch": 0.47, "learning_rate": 2.5464263501855857e-06, "loss": 0.9546, "step": 11639 }, { "epoch": 0.47, "learning_rate": 2.5449893517173363e-06, "loss": 0.9365, "step": 11640 }, { "epoch": 0.47, "learning_rate": 2.543552699706675e-06, "loss": 0.9244, "step": 11641 }, { "epoch": 0.47, "learning_rate": 2.542116394220371e-06, "loss": 0.9158, "step": 11642 }, { "epoch": 0.47, "learning_rate": 2.5406804353251746e-06, "loss": 0.9186, "step": 11643 }, { "epoch": 0.47, "learning_rate": 2.5392448230878206e-06, "loss": 0.9747, "step": 11644 }, { "epoch": 0.47, "learning_rate": 2.5378095575750197e-06, "loss": 0.9441, "step": 11645 }, { "epoch": 0.47, "learning_rate": 2.5363746388534828e-06, "loss": 0.9281, "step": 11646 }, { "epoch": 0.47, "learning_rate": 2.534940066989887e-06, "loss": 0.9507, "step": 11647 }, { "epoch": 0.47, "learning_rate": 2.533505842050904e-06, "loss": 0.9022, "step": 11648 }, { "epoch": 0.48, "learning_rate": 2.532071964103191e-06, "loss": 0.8962, "step": 11649 }, { "epoch": 0.48, "learning_rate": 2.5306384332133773e-06, "loss": 0.9841, "step": 11650 }, { "epoch": 0.48, "learning_rate": 2.529205249448087e-06, "loss": 0.9329, "step": 11651 }, { "epoch": 0.48, "learning_rate": 2.5277724128739246e-06, "loss": 0.9126, "step": 11652 }, { "epoch": 0.48, "learning_rate": 2.5263399235574827e-06, "loss": 0.9037, "step": 11653 }, { "epoch": 0.48, "learning_rate": 2.5249077815653224e-06, "loss": 0.9082, "step": 11654 }, { "epoch": 0.48, "learning_rate": 2.5234759869640124e-06, "loss": 0.9094, "step": 11655 }, { "epoch": 0.48, "learning_rate": 2.5220445398200876e-06, "loss": 0.927, "step": 11656 }, { "epoch": 0.48, "learning_rate": 2.520613440200066e-06, "loss": 0.9162, "step": 11657 }, { "epoch": 0.48, "learning_rate": 2.519182688170466e-06, "loss": 0.9121, "step": 11658 }, { "epoch": 0.48, "learning_rate": 2.517752283797771e-06, "loss": 0.9401, "step": 11659 }, { "epoch": 0.48, "learning_rate": 2.516322227148459e-06, "loss": 0.9257, "step": 11660 }, { "epoch": 0.48, "learning_rate": 2.514892518288988e-06, "loss": 0.9284, "step": 11661 }, { "epoch": 0.48, "learning_rate": 2.513463157285807e-06, "loss": 0.961, "step": 11662 }, { "epoch": 0.48, "learning_rate": 2.5120341442053344e-06, "loss": 0.9591, "step": 11663 }, { "epoch": 0.48, "learning_rate": 2.5106054791139846e-06, "loss": 0.8905, "step": 11664 }, { "epoch": 0.48, "learning_rate": 2.5091771620781545e-06, "loss": 0.9308, "step": 11665 }, { "epoch": 0.48, "learning_rate": 2.507749193164215e-06, "loss": 0.9565, "step": 11666 }, { "epoch": 0.48, "learning_rate": 2.5063215724385394e-06, "loss": 0.965, "step": 11667 }, { "epoch": 0.48, "learning_rate": 2.504894299967464e-06, "loss": 0.891, "step": 11668 }, { "epoch": 0.48, "learning_rate": 2.5034673758173234e-06, "loss": 0.9623, "step": 11669 }, { "epoch": 0.48, "learning_rate": 2.502040800054429e-06, "loss": 0.9476, "step": 11670 }, { "epoch": 0.48, "learning_rate": 2.5006145727450825e-06, "loss": 0.9446, "step": 11671 }, { "epoch": 0.48, "learning_rate": 2.4991886939555587e-06, "loss": 0.9015, "step": 11672 }, { "epoch": 0.48, "learning_rate": 2.4977631637521262e-06, "loss": 0.9309, "step": 11673 }, { "epoch": 0.48, "learning_rate": 2.496337982201037e-06, "loss": 0.9987, "step": 11674 }, { "epoch": 0.48, "learning_rate": 2.494913149368513e-06, "loss": 0.9459, "step": 11675 }, { "epoch": 0.48, "learning_rate": 2.4934886653207847e-06, "loss": 0.9134, "step": 11676 }, { "epoch": 0.48, "learning_rate": 2.492064530124042e-06, "loss": 0.9286, "step": 11677 }, { "epoch": 0.48, "learning_rate": 2.4906407438444724e-06, "loss": 0.9296, "step": 11678 }, { "epoch": 0.48, "learning_rate": 2.4892173065482473e-06, "loss": 0.8897, "step": 11679 }, { "epoch": 0.48, "learning_rate": 2.4877942183015104e-06, "loss": 0.9274, "step": 11680 }, { "epoch": 0.48, "learning_rate": 2.486371479170402e-06, "loss": 0.9694, "step": 11681 }, { "epoch": 0.48, "learning_rate": 2.4849490892210402e-06, "loss": 0.9485, "step": 11682 }, { "epoch": 0.48, "learning_rate": 2.4835270485195318e-06, "loss": 0.9417, "step": 11683 }, { "epoch": 0.48, "learning_rate": 2.4821053571319565e-06, "loss": 0.9138, "step": 11684 }, { "epoch": 0.48, "learning_rate": 2.4806840151243884e-06, "loss": 0.9307, "step": 11685 }, { "epoch": 0.48, "learning_rate": 2.4792630225628835e-06, "loss": 0.9247, "step": 11686 }, { "epoch": 0.48, "learning_rate": 2.4778423795134755e-06, "loss": 0.9118, "step": 11687 }, { "epoch": 0.48, "learning_rate": 2.4764220860421873e-06, "loss": 0.9531, "step": 11688 }, { "epoch": 0.48, "learning_rate": 2.475002142215026e-06, "loss": 0.9281, "step": 11689 }, { "epoch": 0.48, "learning_rate": 2.4735825480979802e-06, "loss": 0.9227, "step": 11690 }, { "epoch": 0.48, "learning_rate": 2.4721633037570223e-06, "loss": 0.9928, "step": 11691 }, { "epoch": 0.48, "learning_rate": 2.4707444092581133e-06, "loss": 0.9406, "step": 11692 }, { "epoch": 0.48, "learning_rate": 2.469325864667186e-06, "loss": 0.9284, "step": 11693 }, { "epoch": 0.48, "learning_rate": 2.467907670050169e-06, "loss": 0.9343, "step": 11694 }, { "epoch": 0.48, "learning_rate": 2.466489825472973e-06, "loss": 0.8851, "step": 11695 }, { "epoch": 0.48, "learning_rate": 2.4650723310014823e-06, "loss": 0.9385, "step": 11696 }, { "epoch": 0.48, "learning_rate": 2.4636551867015767e-06, "loss": 0.9829, "step": 11697 }, { "epoch": 0.48, "learning_rate": 2.462238392639116e-06, "loss": 0.9641, "step": 11698 }, { "epoch": 0.48, "learning_rate": 2.4608219488799434e-06, "loss": 0.9678, "step": 11699 }, { "epoch": 0.48, "learning_rate": 2.459405855489878e-06, "loss": 0.918, "step": 11700 }, { "epoch": 0.48, "learning_rate": 2.4579901125347426e-06, "loss": 0.9106, "step": 11701 }, { "epoch": 0.48, "learning_rate": 2.456574720080326e-06, "loss": 0.9229, "step": 11702 }, { "epoch": 0.48, "learning_rate": 2.455159678192397e-06, "loss": 0.9127, "step": 11703 }, { "epoch": 0.48, "learning_rate": 2.4537449869367314e-06, "loss": 0.9365, "step": 11704 }, { "epoch": 0.48, "learning_rate": 2.452330646379064e-06, "loss": 0.9361, "step": 11705 }, { "epoch": 0.48, "learning_rate": 2.4509166565851274e-06, "loss": 0.9436, "step": 11706 }, { "epoch": 0.48, "learning_rate": 2.4495030176206336e-06, "loss": 0.9003, "step": 11707 }, { "epoch": 0.48, "learning_rate": 2.448089729551282e-06, "loss": 0.9595, "step": 11708 }, { "epoch": 0.48, "learning_rate": 2.446676792442747e-06, "loss": 0.9264, "step": 11709 }, { "epoch": 0.48, "learning_rate": 2.4452642063606947e-06, "loss": 0.904, "step": 11710 }, { "epoch": 0.48, "learning_rate": 2.4438519713707742e-06, "loss": 0.9414, "step": 11711 }, { "epoch": 0.48, "learning_rate": 2.44244008753861e-06, "loss": 0.9309, "step": 11712 }, { "epoch": 0.48, "learning_rate": 2.4410285549298275e-06, "loss": 0.9044, "step": 11713 }, { "epoch": 0.48, "learning_rate": 2.439617373610016e-06, "loss": 0.928, "step": 11714 }, { "epoch": 0.48, "learning_rate": 2.43820654364476e-06, "loss": 0.8948, "step": 11715 }, { "epoch": 0.48, "learning_rate": 2.436796065099626e-06, "loss": 0.9521, "step": 11716 }, { "epoch": 0.48, "learning_rate": 2.4353859380401646e-06, "loss": 0.9468, "step": 11717 }, { "epoch": 0.48, "learning_rate": 2.433976162531906e-06, "loss": 0.8888, "step": 11718 }, { "epoch": 0.48, "learning_rate": 2.432566738640367e-06, "loss": 0.9448, "step": 11719 }, { "epoch": 0.48, "learning_rate": 2.431157666431052e-06, "loss": 0.9345, "step": 11720 }, { "epoch": 0.48, "learning_rate": 2.4297489459694366e-06, "loss": 0.9388, "step": 11721 }, { "epoch": 0.48, "learning_rate": 2.428340577320999e-06, "loss": 0.9457, "step": 11722 }, { "epoch": 0.48, "learning_rate": 2.426932560551183e-06, "loss": 0.9026, "step": 11723 }, { "epoch": 0.48, "learning_rate": 2.425524895725425e-06, "loss": 0.9419, "step": 11724 }, { "epoch": 0.48, "learning_rate": 2.424117582909148e-06, "loss": 0.942, "step": 11725 }, { "epoch": 0.48, "learning_rate": 2.4227106221677476e-06, "loss": 0.9266, "step": 11726 }, { "epoch": 0.48, "learning_rate": 2.421304013566612e-06, "loss": 0.9221, "step": 11727 }, { "epoch": 0.48, "learning_rate": 2.4198977571711114e-06, "loss": 0.9177, "step": 11728 }, { "epoch": 0.48, "learning_rate": 2.4184918530466027e-06, "loss": 0.9222, "step": 11729 }, { "epoch": 0.48, "learning_rate": 2.4170863012584144e-06, "loss": 0.8894, "step": 11730 }, { "epoch": 0.48, "learning_rate": 2.415681101871873e-06, "loss": 0.9385, "step": 11731 }, { "epoch": 0.48, "learning_rate": 2.4142762549522823e-06, "loss": 0.9241, "step": 11732 }, { "epoch": 0.48, "learning_rate": 2.4128717605649255e-06, "loss": 0.9109, "step": 11733 }, { "epoch": 0.48, "learning_rate": 2.4114676187750773e-06, "loss": 0.9357, "step": 11734 }, { "epoch": 0.48, "learning_rate": 2.4100638296479916e-06, "loss": 0.9347, "step": 11735 }, { "epoch": 0.48, "learning_rate": 2.408660393248907e-06, "loss": 0.9084, "step": 11736 }, { "epoch": 0.48, "learning_rate": 2.407257309643046e-06, "loss": 0.9422, "step": 11737 }, { "epoch": 0.48, "learning_rate": 2.405854578895618e-06, "loss": 0.9133, "step": 11738 }, { "epoch": 0.48, "learning_rate": 2.404452201071804e-06, "loss": 0.9258, "step": 11739 }, { "epoch": 0.48, "learning_rate": 2.403050176236781e-06, "loss": 0.9219, "step": 11740 }, { "epoch": 0.48, "learning_rate": 2.401648504455709e-06, "loss": 0.9468, "step": 11741 }, { "epoch": 0.48, "learning_rate": 2.4002471857937227e-06, "loss": 0.9526, "step": 11742 }, { "epoch": 0.48, "learning_rate": 2.398846220315947e-06, "loss": 0.9065, "step": 11743 }, { "epoch": 0.48, "learning_rate": 2.3974456080874896e-06, "loss": 0.9465, "step": 11744 }, { "epoch": 0.48, "learning_rate": 2.396045349173445e-06, "loss": 0.9258, "step": 11745 }, { "epoch": 0.48, "learning_rate": 2.3946454436388777e-06, "loss": 0.9466, "step": 11746 }, { "epoch": 0.48, "learning_rate": 2.3932458915488587e-06, "loss": 0.9506, "step": 11747 }, { "epoch": 0.48, "learning_rate": 2.3918466929684205e-06, "loss": 0.926, "step": 11748 }, { "epoch": 0.48, "learning_rate": 2.3904478479625903e-06, "loss": 0.9077, "step": 11749 }, { "epoch": 0.48, "learning_rate": 2.389049356596379e-06, "loss": 0.9398, "step": 11750 }, { "epoch": 0.48, "learning_rate": 2.3876512189347757e-06, "loss": 0.9558, "step": 11751 }, { "epoch": 0.48, "learning_rate": 2.386253435042757e-06, "loss": 0.8998, "step": 11752 }, { "epoch": 0.48, "learning_rate": 2.384856004985283e-06, "loss": 0.9478, "step": 11753 }, { "epoch": 0.48, "learning_rate": 2.3834589288273003e-06, "loss": 0.9507, "step": 11754 }, { "epoch": 0.48, "learning_rate": 2.3820622066337275e-06, "loss": 0.953, "step": 11755 }, { "epoch": 0.48, "learning_rate": 2.3806658384694793e-06, "loss": 0.9449, "step": 11756 }, { "epoch": 0.48, "learning_rate": 2.3792698243994506e-06, "loss": 0.9169, "step": 11757 }, { "epoch": 0.48, "learning_rate": 2.377874164488512e-06, "loss": 0.8923, "step": 11758 }, { "epoch": 0.48, "learning_rate": 2.3764788588015344e-06, "loss": 0.9041, "step": 11759 }, { "epoch": 0.48, "learning_rate": 2.375083907403354e-06, "loss": 0.9259, "step": 11760 }, { "epoch": 0.48, "learning_rate": 2.3736893103588e-06, "loss": 0.9113, "step": 11761 }, { "epoch": 0.48, "learning_rate": 2.372295067732685e-06, "loss": 0.9587, "step": 11762 }, { "epoch": 0.48, "learning_rate": 2.370901179589806e-06, "loss": 0.9482, "step": 11763 }, { "epoch": 0.48, "learning_rate": 2.3695076459949364e-06, "loss": 1.0306, "step": 11764 }, { "epoch": 0.48, "learning_rate": 2.368114467012841e-06, "loss": 0.9734, "step": 11765 }, { "epoch": 0.48, "learning_rate": 2.366721642708266e-06, "loss": 0.9219, "step": 11766 }, { "epoch": 0.48, "learning_rate": 2.365329173145934e-06, "loss": 0.9566, "step": 11767 }, { "epoch": 0.48, "learning_rate": 2.3639370583905685e-06, "loss": 0.8977, "step": 11768 }, { "epoch": 0.48, "learning_rate": 2.3625452985068552e-06, "loss": 0.9624, "step": 11769 }, { "epoch": 0.48, "learning_rate": 2.361153893559478e-06, "loss": 0.9415, "step": 11770 }, { "epoch": 0.48, "learning_rate": 2.3597628436131026e-06, "loss": 0.9457, "step": 11771 }, { "epoch": 0.48, "learning_rate": 2.3583721487323683e-06, "loss": 0.9502, "step": 11772 }, { "epoch": 0.48, "learning_rate": 2.356981808981911e-06, "loss": 0.909, "step": 11773 }, { "epoch": 0.48, "learning_rate": 2.35559182442634e-06, "loss": 0.9639, "step": 11774 }, { "epoch": 0.48, "learning_rate": 2.3542021951302574e-06, "loss": 0.9156, "step": 11775 }, { "epoch": 0.48, "learning_rate": 2.3528129211582384e-06, "loss": 0.9333, "step": 11776 }, { "epoch": 0.48, "learning_rate": 2.351424002574848e-06, "loss": 0.9271, "step": 11777 }, { "epoch": 0.48, "learning_rate": 2.350035439444638e-06, "loss": 0.9119, "step": 11778 }, { "epoch": 0.48, "learning_rate": 2.348647231832131e-06, "loss": 0.9775, "step": 11779 }, { "epoch": 0.48, "learning_rate": 2.3472593798018506e-06, "loss": 0.9365, "step": 11780 }, { "epoch": 0.48, "learning_rate": 2.345871883418287e-06, "loss": 0.9203, "step": 11781 }, { "epoch": 0.48, "learning_rate": 2.344484742745925e-06, "loss": 0.9825, "step": 11782 }, { "epoch": 0.48, "learning_rate": 2.3430979578492295e-06, "loss": 0.9128, "step": 11783 }, { "epoch": 0.48, "learning_rate": 2.3417115287926507e-06, "loss": 0.8803, "step": 11784 }, { "epoch": 0.48, "learning_rate": 2.340325455640614e-06, "loss": 0.951, "step": 11785 }, { "epoch": 0.48, "learning_rate": 2.3389397384575398e-06, "loss": 0.9319, "step": 11786 }, { "epoch": 0.48, "learning_rate": 2.337554377307827e-06, "loss": 0.8998, "step": 11787 }, { "epoch": 0.48, "learning_rate": 2.336169372255853e-06, "loss": 0.8998, "step": 11788 }, { "epoch": 0.48, "learning_rate": 2.334784723365987e-06, "loss": 0.9587, "step": 11789 }, { "epoch": 0.48, "learning_rate": 2.333400430702577e-06, "loss": 0.8977, "step": 11790 }, { "epoch": 0.48, "learning_rate": 2.3320164943299572e-06, "loss": 0.918, "step": 11791 }, { "epoch": 0.48, "learning_rate": 2.3306329143124374e-06, "loss": 0.9501, "step": 11792 }, { "epoch": 0.48, "learning_rate": 2.329249690714326e-06, "loss": 0.9097, "step": 11793 }, { "epoch": 0.48, "learning_rate": 2.327866823599899e-06, "loss": 0.9398, "step": 11794 }, { "epoch": 0.48, "learning_rate": 2.3264843130334237e-06, "loss": 0.9047, "step": 11795 }, { "epoch": 0.48, "learning_rate": 2.3251021590791534e-06, "loss": 0.9055, "step": 11796 }, { "epoch": 0.48, "learning_rate": 2.323720361801316e-06, "loss": 0.9024, "step": 11797 }, { "epoch": 0.48, "learning_rate": 2.322338921264129e-06, "loss": 0.9231, "step": 11798 }, { "epoch": 0.49, "learning_rate": 2.320957837531794e-06, "loss": 0.9342, "step": 11799 }, { "epoch": 0.49, "learning_rate": 2.3195771106684974e-06, "loss": 0.9117, "step": 11800 }, { "epoch": 0.49, "learning_rate": 2.3181967407383977e-06, "loss": 0.9797, "step": 11801 }, { "epoch": 0.49, "learning_rate": 2.3168167278056497e-06, "loss": 0.9363, "step": 11802 }, { "epoch": 0.49, "learning_rate": 2.3154370719343898e-06, "loss": 0.8758, "step": 11803 }, { "epoch": 0.49, "learning_rate": 2.3140577731887257e-06, "loss": 0.908, "step": 11804 }, { "epoch": 0.49, "learning_rate": 2.31267883163277e-06, "loss": 0.9343, "step": 11805 }, { "epoch": 0.49, "learning_rate": 2.311300247330597e-06, "loss": 0.9581, "step": 11806 }, { "epoch": 0.49, "learning_rate": 2.3099220203462768e-06, "loss": 0.9384, "step": 11807 }, { "epoch": 0.49, "learning_rate": 2.30854415074386e-06, "loss": 0.9379, "step": 11808 }, { "epoch": 0.49, "learning_rate": 2.3071666385873835e-06, "loss": 0.9193, "step": 11809 }, { "epoch": 0.49, "learning_rate": 2.305789483940859e-06, "loss": 0.942, "step": 11810 }, { "epoch": 0.49, "learning_rate": 2.3044126868682892e-06, "loss": 0.9297, "step": 11811 }, { "epoch": 0.49, "learning_rate": 2.303036247433662e-06, "loss": 0.8698, "step": 11812 }, { "epoch": 0.49, "learning_rate": 2.3016601657009364e-06, "loss": 1.0068, "step": 11813 }, { "epoch": 0.49, "learning_rate": 2.3002844417340743e-06, "loss": 0.9053, "step": 11814 }, { "epoch": 0.49, "learning_rate": 2.298909075597e-06, "loss": 0.8862, "step": 11815 }, { "epoch": 0.49, "learning_rate": 2.297534067353635e-06, "loss": 0.9365, "step": 11816 }, { "epoch": 0.49, "learning_rate": 2.2961594170678835e-06, "loss": 0.9041, "step": 11817 }, { "epoch": 0.49, "learning_rate": 2.2947851248036245e-06, "loss": 0.9202, "step": 11818 }, { "epoch": 0.49, "learning_rate": 2.293411190624727e-06, "loss": 0.9042, "step": 11819 }, { "epoch": 0.49, "learning_rate": 2.292037614595043e-06, "loss": 0.9339, "step": 11820 }, { "epoch": 0.49, "learning_rate": 2.290664396778408e-06, "loss": 0.8783, "step": 11821 }, { "epoch": 0.49, "learning_rate": 2.289291537238637e-06, "loss": 0.9027, "step": 11822 }, { "epoch": 0.49, "learning_rate": 2.287919036039532e-06, "loss": 0.9809, "step": 11823 }, { "epoch": 0.49, "learning_rate": 2.286546893244881e-06, "loss": 0.9439, "step": 11824 }, { "epoch": 0.49, "learning_rate": 2.2851751089184417e-06, "loss": 0.9635, "step": 11825 }, { "epoch": 0.49, "learning_rate": 2.2838036831239785e-06, "loss": 0.9567, "step": 11826 }, { "epoch": 0.49, "learning_rate": 2.2824326159252164e-06, "loss": 0.96, "step": 11827 }, { "epoch": 0.49, "learning_rate": 2.2810619073858752e-06, "loss": 0.9073, "step": 11828 }, { "epoch": 0.49, "learning_rate": 2.2796915575696578e-06, "loss": 0.897, "step": 11829 }, { "epoch": 0.49, "learning_rate": 2.2783215665402503e-06, "loss": 0.9325, "step": 11830 }, { "epoch": 0.49, "learning_rate": 2.2769519343613158e-06, "loss": 0.9307, "step": 11831 }, { "epoch": 0.49, "learning_rate": 2.275582661096506e-06, "loss": 0.9011, "step": 11832 }, { "epoch": 0.49, "learning_rate": 2.2742137468094605e-06, "loss": 0.9536, "step": 11833 }, { "epoch": 0.49, "learning_rate": 2.2728451915637904e-06, "loss": 1.0038, "step": 11834 }, { "epoch": 0.49, "learning_rate": 2.2714769954230996e-06, "loss": 0.9279, "step": 11835 }, { "epoch": 0.49, "learning_rate": 2.2701091584509727e-06, "loss": 0.9019, "step": 11836 }, { "epoch": 0.49, "learning_rate": 2.2687416807109787e-06, "loss": 0.9555, "step": 11837 }, { "epoch": 0.49, "learning_rate": 2.267374562266662e-06, "loss": 0.8886, "step": 11838 }, { "epoch": 0.49, "learning_rate": 2.266007803181568e-06, "loss": 0.9518, "step": 11839 }, { "epoch": 0.49, "learning_rate": 2.2646414035192056e-06, "loss": 0.92, "step": 11840 }, { "epoch": 0.49, "learning_rate": 2.2632753633430772e-06, "loss": 0.9353, "step": 11841 }, { "epoch": 0.49, "learning_rate": 2.261909682716671e-06, "loss": 0.853, "step": 11842 }, { "epoch": 0.49, "learning_rate": 2.2605443617034474e-06, "loss": 0.9282, "step": 11843 }, { "epoch": 0.49, "learning_rate": 2.259179400366862e-06, "loss": 0.9069, "step": 11844 }, { "epoch": 0.49, "learning_rate": 2.257814798770348e-06, "loss": 0.9043, "step": 11845 }, { "epoch": 0.49, "learning_rate": 2.2564505569773254e-06, "loss": 0.9151, "step": 11846 }, { "epoch": 0.49, "learning_rate": 2.255086675051189e-06, "loss": 0.931, "step": 11847 }, { "epoch": 0.49, "learning_rate": 2.253723153055325e-06, "loss": 0.9533, "step": 11848 }, { "epoch": 0.49, "learning_rate": 2.2523599910531048e-06, "loss": 0.8721, "step": 11849 }, { "epoch": 0.49, "learning_rate": 2.2509971891078684e-06, "loss": 0.9442, "step": 11850 }, { "epoch": 0.49, "learning_rate": 2.2496347472829615e-06, "loss": 0.9367, "step": 11851 }, { "epoch": 0.49, "learning_rate": 2.2482726656416944e-06, "loss": 0.9406, "step": 11852 }, { "epoch": 0.49, "learning_rate": 2.246910944247367e-06, "loss": 0.9252, "step": 11853 }, { "epoch": 0.49, "learning_rate": 2.2455495831632647e-06, "loss": 0.985, "step": 11854 }, { "epoch": 0.49, "learning_rate": 2.2441885824526564e-06, "loss": 0.9174, "step": 11855 }, { "epoch": 0.49, "learning_rate": 2.2428279421787878e-06, "loss": 0.9062, "step": 11856 }, { "epoch": 0.49, "learning_rate": 2.2414676624048926e-06, "loss": 0.931, "step": 11857 }, { "epoch": 0.49, "learning_rate": 2.240107743194191e-06, "loss": 0.8941, "step": 11858 }, { "epoch": 0.49, "learning_rate": 2.2387481846098745e-06, "loss": 0.9646, "step": 11859 }, { "epoch": 0.49, "learning_rate": 2.2373889867151377e-06, "loss": 0.9229, "step": 11860 }, { "epoch": 0.49, "learning_rate": 2.2360301495731375e-06, "loss": 0.9194, "step": 11861 }, { "epoch": 0.49, "learning_rate": 2.2346716732470266e-06, "loss": 0.9645, "step": 11862 }, { "epoch": 0.49, "learning_rate": 2.23331355779994e-06, "loss": 0.9436, "step": 11863 }, { "epoch": 0.49, "learning_rate": 2.231955803294987e-06, "loss": 0.9359, "step": 11864 }, { "epoch": 0.49, "learning_rate": 2.230598409795273e-06, "loss": 0.9527, "step": 11865 }, { "epoch": 0.49, "learning_rate": 2.2292413773638767e-06, "loss": 0.9618, "step": 11866 }, { "epoch": 0.49, "learning_rate": 2.2278847060638676e-06, "loss": 0.8844, "step": 11867 }, { "epoch": 0.49, "learning_rate": 2.226528395958286e-06, "loss": 0.9613, "step": 11868 }, { "epoch": 0.49, "learning_rate": 2.2251724471101753e-06, "loss": 0.9, "step": 11869 }, { "epoch": 0.49, "learning_rate": 2.223816859582546e-06, "loss": 0.9148, "step": 11870 }, { "epoch": 0.49, "learning_rate": 2.2224616334383886e-06, "loss": 0.9485, "step": 11871 }, { "epoch": 0.49, "learning_rate": 2.2211067687406983e-06, "loss": 0.9293, "step": 11872 }, { "epoch": 0.49, "learning_rate": 2.2197522655524297e-06, "loss": 0.9263, "step": 11873 }, { "epoch": 0.49, "learning_rate": 2.218398123936535e-06, "loss": 0.9258, "step": 11874 }, { "epoch": 0.49, "learning_rate": 2.217044343955943e-06, "loss": 0.9384, "step": 11875 }, { "epoch": 0.49, "learning_rate": 2.2156909256735735e-06, "loss": 0.9046, "step": 11876 }, { "epoch": 0.49, "learning_rate": 2.2143378691523174e-06, "loss": 0.9627, "step": 11877 }, { "epoch": 0.49, "learning_rate": 2.212985174455059e-06, "loss": 0.9484, "step": 11878 }, { "epoch": 0.49, "learning_rate": 2.211632841644663e-06, "loss": 0.9215, "step": 11879 }, { "epoch": 0.49, "learning_rate": 2.2102808707839727e-06, "loss": 0.8901, "step": 11880 }, { "epoch": 0.49, "learning_rate": 2.2089292619358218e-06, "loss": 0.9034, "step": 11881 }, { "epoch": 0.49, "learning_rate": 2.2075780151630223e-06, "loss": 0.9202, "step": 11882 }, { "epoch": 0.49, "learning_rate": 2.20622713052837e-06, "loss": 0.9275, "step": 11883 }, { "epoch": 0.49, "learning_rate": 2.204876608094647e-06, "loss": 0.9623, "step": 11884 }, { "epoch": 0.49, "learning_rate": 2.2035264479246187e-06, "loss": 0.9182, "step": 11885 }, { "epoch": 0.49, "learning_rate": 2.2021766500810247e-06, "loss": 0.9316, "step": 11886 }, { "epoch": 0.49, "learning_rate": 2.2008272146265973e-06, "loss": 0.9366, "step": 11887 }, { "epoch": 0.49, "learning_rate": 2.199478141624052e-06, "loss": 0.9204, "step": 11888 }, { "epoch": 0.49, "learning_rate": 2.198129431136078e-06, "loss": 0.9466, "step": 11889 }, { "epoch": 0.49, "learning_rate": 2.196781083225359e-06, "loss": 0.9606, "step": 11890 }, { "epoch": 0.49, "learning_rate": 2.1954330979545548e-06, "loss": 0.9136, "step": 11891 }, { "epoch": 0.49, "learning_rate": 2.1940854753863138e-06, "loss": 0.9274, "step": 11892 }, { "epoch": 0.49, "learning_rate": 2.19273821558326e-06, "loss": 0.9199, "step": 11893 }, { "epoch": 0.49, "learning_rate": 2.191391318608006e-06, "loss": 0.9114, "step": 11894 }, { "epoch": 0.49, "learning_rate": 2.19004478452315e-06, "loss": 0.982, "step": 11895 }, { "epoch": 0.49, "learning_rate": 2.18869861339126e-06, "loss": 0.932, "step": 11896 }, { "epoch": 0.49, "learning_rate": 2.1873528052749094e-06, "loss": 0.9281, "step": 11897 }, { "epoch": 0.49, "learning_rate": 2.1860073602366317e-06, "loss": 0.9358, "step": 11898 }, { "epoch": 0.49, "learning_rate": 2.1846622783389595e-06, "loss": 0.9176, "step": 11899 }, { "epoch": 0.49, "learning_rate": 2.1833175596444013e-06, "loss": 0.9152, "step": 11900 }, { "epoch": 0.49, "learning_rate": 2.181973204215453e-06, "loss": 0.9434, "step": 11901 }, { "epoch": 0.49, "learning_rate": 2.1806292121145866e-06, "loss": 0.9384, "step": 11902 }, { "epoch": 0.49, "learning_rate": 2.1792855834042638e-06, "loss": 0.9444, "step": 11903 }, { "epoch": 0.49, "learning_rate": 2.1779423181469284e-06, "loss": 0.9168, "step": 11904 }, { "epoch": 0.49, "learning_rate": 2.1765994164049995e-06, "loss": 0.9539, "step": 11905 }, { "epoch": 0.49, "learning_rate": 2.1752568782408977e-06, "loss": 0.8913, "step": 11906 }, { "epoch": 0.49, "learning_rate": 2.173914703717005e-06, "loss": 0.9487, "step": 11907 }, { "epoch": 0.49, "learning_rate": 2.1725728928957e-06, "loss": 0.9303, "step": 11908 }, { "epoch": 0.49, "learning_rate": 2.1712314458393437e-06, "loss": 0.9158, "step": 11909 }, { "epoch": 0.49, "learning_rate": 2.1698903626102717e-06, "loss": 0.904, "step": 11910 }, { "epoch": 0.49, "learning_rate": 2.1685496432708107e-06, "loss": 0.9363, "step": 11911 }, { "epoch": 0.49, "learning_rate": 2.1672092878832683e-06, "loss": 0.9087, "step": 11912 }, { "epoch": 0.49, "learning_rate": 2.1658692965099383e-06, "loss": 0.8937, "step": 11913 }, { "epoch": 0.49, "learning_rate": 2.1645296692130844e-06, "loss": 0.8835, "step": 11914 }, { "epoch": 0.49, "learning_rate": 2.163190406054976e-06, "loss": 0.8887, "step": 11915 }, { "epoch": 0.49, "learning_rate": 2.161851507097844e-06, "loss": 0.8877, "step": 11916 }, { "epoch": 0.49, "learning_rate": 2.1605129724039144e-06, "loss": 0.9629, "step": 11917 }, { "epoch": 0.49, "learning_rate": 2.159174802035394e-06, "loss": 0.9474, "step": 11918 }, { "epoch": 0.49, "learning_rate": 2.1578369960544687e-06, "loss": 0.9535, "step": 11919 }, { "epoch": 0.49, "learning_rate": 2.1564995545233125e-06, "loss": 0.9647, "step": 11920 }, { "epoch": 0.49, "learning_rate": 2.1551624775040793e-06, "loss": 0.9601, "step": 11921 }, { "epoch": 0.49, "learning_rate": 2.1538257650589114e-06, "loss": 0.9476, "step": 11922 }, { "epoch": 0.49, "learning_rate": 2.152489417249923e-06, "loss": 0.9331, "step": 11923 }, { "epoch": 0.49, "learning_rate": 2.1511534341392227e-06, "loss": 0.9205, "step": 11924 }, { "epoch": 0.49, "learning_rate": 2.1498178157889003e-06, "loss": 0.9143, "step": 11925 }, { "epoch": 0.49, "learning_rate": 2.1484825622610195e-06, "loss": 0.9347, "step": 11926 }, { "epoch": 0.49, "learning_rate": 2.1471476736176376e-06, "loss": 0.9053, "step": 11927 }, { "epoch": 0.49, "learning_rate": 2.1458131499207913e-06, "loss": 0.9291, "step": 11928 }, { "epoch": 0.49, "learning_rate": 2.144478991232498e-06, "loss": 0.9586, "step": 11929 }, { "epoch": 0.49, "learning_rate": 2.1431451976147623e-06, "loss": 0.9496, "step": 11930 }, { "epoch": 0.49, "learning_rate": 2.1418117691295713e-06, "loss": 0.9611, "step": 11931 }, { "epoch": 0.49, "learning_rate": 2.14047870583889e-06, "loss": 0.9267, "step": 11932 }, { "epoch": 0.49, "learning_rate": 2.13914600780467e-06, "loss": 0.9129, "step": 11933 }, { "epoch": 0.49, "learning_rate": 2.13781367508885e-06, "loss": 0.9399, "step": 11934 }, { "epoch": 0.49, "learning_rate": 2.1364817077533427e-06, "loss": 0.9193, "step": 11935 }, { "epoch": 0.49, "learning_rate": 2.135150105860051e-06, "loss": 0.9155, "step": 11936 }, { "epoch": 0.49, "learning_rate": 2.1338188694708584e-06, "loss": 0.9164, "step": 11937 }, { "epoch": 0.49, "learning_rate": 2.1324879986476352e-06, "loss": 0.8614, "step": 11938 }, { "epoch": 0.49, "learning_rate": 2.1311574934522215e-06, "loss": 0.9293, "step": 11939 }, { "epoch": 0.49, "learning_rate": 2.129827353946462e-06, "loss": 0.8803, "step": 11940 }, { "epoch": 0.49, "learning_rate": 2.128497580192167e-06, "loss": 0.9286, "step": 11941 }, { "epoch": 0.49, "learning_rate": 2.1271681722511282e-06, "loss": 0.9251, "step": 11942 }, { "epoch": 0.49, "learning_rate": 2.125839130185141e-06, "loss": 0.9191, "step": 11943 }, { "epoch": 0.49, "learning_rate": 2.1245104540559603e-06, "loss": 0.9236, "step": 11944 }, { "epoch": 0.49, "learning_rate": 2.1231821439253365e-06, "loss": 0.9347, "step": 11945 }, { "epoch": 0.49, "learning_rate": 2.121854199855e-06, "loss": 0.9564, "step": 11946 }, { "epoch": 0.49, "learning_rate": 2.1205266219066676e-06, "loss": 0.9433, "step": 11947 }, { "epoch": 0.49, "learning_rate": 2.119199410142031e-06, "loss": 0.9012, "step": 11948 }, { "epoch": 0.5, "learning_rate": 2.1178725646227726e-06, "loss": 0.9708, "step": 11949 }, { "epoch": 0.5, "learning_rate": 2.1165460854105566e-06, "loss": 0.9289, "step": 11950 }, { "epoch": 0.5, "learning_rate": 2.115219972567023e-06, "loss": 0.9524, "step": 11951 }, { "epoch": 0.5, "learning_rate": 2.1138942261538074e-06, "loss": 0.9411, "step": 11952 }, { "epoch": 0.5, "learning_rate": 2.1125688462325167e-06, "loss": 0.9056, "step": 11953 }, { "epoch": 0.5, "learning_rate": 2.1112438328647465e-06, "loss": 0.9226, "step": 11954 }, { "epoch": 0.5, "learning_rate": 2.1099191861120784e-06, "loss": 0.9545, "step": 11955 }, { "epoch": 0.5, "learning_rate": 2.1085949060360654e-06, "loss": 0.9179, "step": 11956 }, { "epoch": 0.5, "learning_rate": 2.107270992698255e-06, "loss": 0.9164, "step": 11957 }, { "epoch": 0.5, "learning_rate": 2.1059474461601726e-06, "loss": 0.913, "step": 11958 }, { "epoch": 0.5, "learning_rate": 2.1046242664833317e-06, "loss": 0.895, "step": 11959 }, { "epoch": 0.5, "learning_rate": 2.1033014537292152e-06, "loss": 0.9415, "step": 11960 }, { "epoch": 0.5, "learning_rate": 2.1019790079593094e-06, "loss": 0.9217, "step": 11961 }, { "epoch": 0.5, "learning_rate": 2.1006569292350643e-06, "loss": 0.9297, "step": 11962 }, { "epoch": 0.5, "learning_rate": 2.099335217617925e-06, "loss": 0.9012, "step": 11963 }, { "epoch": 0.5, "learning_rate": 2.0980138731693167e-06, "loss": 0.933, "step": 11964 }, { "epoch": 0.5, "learning_rate": 2.0966928959506415e-06, "loss": 0.9526, "step": 11965 }, { "epoch": 0.5, "learning_rate": 2.0953722860232916e-06, "loss": 0.8829, "step": 11966 }, { "epoch": 0.5, "learning_rate": 2.094052043448641e-06, "loss": 0.9094, "step": 11967 }, { "epoch": 0.5, "learning_rate": 2.0927321682880476e-06, "loss": 0.9379, "step": 11968 }, { "epoch": 0.5, "learning_rate": 2.091412660602845e-06, "loss": 0.9089, "step": 11969 }, { "epoch": 0.5, "learning_rate": 2.090093520454356e-06, "loss": 0.9247, "step": 11970 }, { "epoch": 0.5, "learning_rate": 2.08877474790389e-06, "loss": 0.944, "step": 11971 }, { "epoch": 0.5, "learning_rate": 2.0874563430127246e-06, "loss": 0.9141, "step": 11972 }, { "epoch": 0.5, "learning_rate": 2.0861383058421423e-06, "loss": 0.941, "step": 11973 }, { "epoch": 0.5, "learning_rate": 2.0848206364533873e-06, "loss": 0.9325, "step": 11974 }, { "epoch": 0.5, "learning_rate": 2.083503334907699e-06, "loss": 0.8924, "step": 11975 }, { "epoch": 0.5, "learning_rate": 2.0821864012662973e-06, "loss": 0.9362, "step": 11976 }, { "epoch": 0.5, "learning_rate": 2.0808698355903856e-06, "loss": 0.8928, "step": 11977 }, { "epoch": 0.5, "learning_rate": 2.079553637941144e-06, "loss": 0.9141, "step": 11978 }, { "epoch": 0.5, "learning_rate": 2.0782378083797437e-06, "loss": 0.9567, "step": 11979 }, { "epoch": 0.5, "learning_rate": 2.076922346967336e-06, "loss": 0.9252, "step": 11980 }, { "epoch": 0.5, "learning_rate": 2.0756072537650508e-06, "loss": 0.9202, "step": 11981 }, { "epoch": 0.5, "learning_rate": 2.0742925288340075e-06, "loss": 0.9324, "step": 11982 }, { "epoch": 0.5, "learning_rate": 2.072978172235305e-06, "loss": 0.9629, "step": 11983 }, { "epoch": 0.5, "learning_rate": 2.0716641840300266e-06, "loss": 0.963, "step": 11984 }, { "epoch": 0.5, "learning_rate": 2.070350564279232e-06, "loss": 0.9417, "step": 11985 }, { "epoch": 0.5, "learning_rate": 2.069037313043979e-06, "loss": 0.9422, "step": 11986 }, { "epoch": 0.5, "learning_rate": 2.0677244303852907e-06, "loss": 0.9478, "step": 11987 }, { "epoch": 0.5, "learning_rate": 2.0664119163641817e-06, "loss": 0.9474, "step": 11988 }, { "epoch": 0.5, "learning_rate": 2.0650997710416533e-06, "loss": 0.9245, "step": 11989 }, { "epoch": 0.5, "learning_rate": 2.06378799447868e-06, "loss": 0.8837, "step": 11990 }, { "epoch": 0.5, "learning_rate": 2.062476586736224e-06, "loss": 0.9599, "step": 11991 }, { "epoch": 0.5, "learning_rate": 2.0611655478752336e-06, "loss": 0.8843, "step": 11992 }, { "epoch": 0.5, "learning_rate": 2.0598548779566387e-06, "loss": 0.9103, "step": 11993 }, { "epoch": 0.5, "learning_rate": 2.0585445770413425e-06, "loss": 0.8994, "step": 11994 }, { "epoch": 0.5, "learning_rate": 2.0572346451902446e-06, "loss": 0.9142, "step": 11995 }, { "epoch": 0.5, "learning_rate": 2.0559250824642243e-06, "loss": 0.9542, "step": 11996 }, { "epoch": 0.5, "learning_rate": 2.054615888924131e-06, "loss": 0.9337, "step": 11997 }, { "epoch": 0.5, "learning_rate": 2.0533070646308173e-06, "loss": 0.9558, "step": 11998 }, { "epoch": 0.5, "learning_rate": 2.0519986096451028e-06, "loss": 0.9086, "step": 11999 }, { "epoch": 0.5, "learning_rate": 2.0506905240277963e-06, "loss": 0.9226, "step": 12000 }, { "epoch": 0.5, "learning_rate": 2.0493828078396928e-06, "loss": 0.8847, "step": 12001 }, { "epoch": 0.5, "learning_rate": 2.048075461141559e-06, "loss": 0.94, "step": 12002 }, { "epoch": 0.5, "learning_rate": 2.0467684839941547e-06, "loss": 0.944, "step": 12003 }, { "epoch": 0.5, "learning_rate": 2.04546187645822e-06, "loss": 0.906, "step": 12004 }, { "epoch": 0.5, "learning_rate": 2.0441556385944783e-06, "loss": 0.9486, "step": 12005 }, { "epoch": 0.5, "learning_rate": 2.0428497704636274e-06, "loss": 0.8929, "step": 12006 }, { "epoch": 0.5, "learning_rate": 2.041544272126366e-06, "loss": 0.9453, "step": 12007 }, { "epoch": 0.5, "learning_rate": 2.0402391436433556e-06, "loss": 0.925, "step": 12008 }, { "epoch": 0.5, "learning_rate": 2.038934385075253e-06, "loss": 0.9442, "step": 12009 }, { "epoch": 0.5, "learning_rate": 2.0376299964826963e-06, "loss": 0.9451, "step": 12010 }, { "epoch": 0.5, "learning_rate": 2.0363259779263012e-06, "loss": 0.949, "step": 12011 }, { "epoch": 0.5, "learning_rate": 2.0350223294666694e-06, "loss": 0.8808, "step": 12012 }, { "epoch": 0.5, "learning_rate": 2.0337190511643878e-06, "loss": 0.898, "step": 12013 }, { "epoch": 0.5, "learning_rate": 2.0324161430800247e-06, "loss": 0.9195, "step": 12014 }, { "epoch": 0.5, "learning_rate": 2.0311136052741274e-06, "loss": 0.931, "step": 12015 }, { "epoch": 0.5, "learning_rate": 2.0298114378072287e-06, "loss": 0.9599, "step": 12016 }, { "epoch": 0.5, "learning_rate": 2.0285096407398485e-06, "loss": 0.9688, "step": 12017 }, { "epoch": 0.5, "learning_rate": 2.027208214132478e-06, "loss": 0.9529, "step": 12018 }, { "epoch": 0.5, "learning_rate": 2.025907158045608e-06, "loss": 0.9332, "step": 12019 }, { "epoch": 0.5, "learning_rate": 2.0246064725396953e-06, "loss": 0.9264, "step": 12020 }, { "epoch": 0.5, "learning_rate": 2.0233061576751877e-06, "loss": 0.9285, "step": 12021 }, { "epoch": 0.5, "learning_rate": 2.0220062135125183e-06, "loss": 0.9378, "step": 12022 }, { "epoch": 0.5, "learning_rate": 2.0207066401121e-06, "loss": 0.9374, "step": 12023 }, { "epoch": 0.5, "learning_rate": 2.019407437534322e-06, "loss": 0.9102, "step": 12024 }, { "epoch": 0.5, "learning_rate": 2.018108605839567e-06, "loss": 0.905, "step": 12025 }, { "epoch": 0.5, "learning_rate": 2.016810145088197e-06, "loss": 0.9377, "step": 12026 }, { "epoch": 0.5, "learning_rate": 2.0155120553405515e-06, "loss": 0.9201, "step": 12027 }, { "epoch": 0.5, "learning_rate": 2.014214336656958e-06, "loss": 0.9455, "step": 12028 }, { "epoch": 0.5, "learning_rate": 2.0129169890977264e-06, "loss": 0.9124, "step": 12029 }, { "epoch": 0.5, "learning_rate": 2.01162001272315e-06, "loss": 0.939, "step": 12030 }, { "epoch": 0.5, "learning_rate": 2.0103234075934964e-06, "loss": 0.935, "step": 12031 }, { "epoch": 0.5, "learning_rate": 2.009027173769034e-06, "loss": 0.8982, "step": 12032 }, { "epoch": 0.5, "learning_rate": 2.0077313113099938e-06, "loss": 0.9339, "step": 12033 }, { "epoch": 0.5, "learning_rate": 2.0064358202766023e-06, "loss": 0.9122, "step": 12034 }, { "epoch": 0.5, "learning_rate": 2.0051407007290657e-06, "loss": 0.9214, "step": 12035 }, { "epoch": 0.5, "learning_rate": 2.0038459527275687e-06, "loss": 0.9567, "step": 12036 }, { "epoch": 0.5, "learning_rate": 2.0025515763322845e-06, "loss": 0.9135, "step": 12037 }, { "epoch": 0.5, "learning_rate": 2.0012575716033676e-06, "loss": 0.9387, "step": 12038 }, { "epoch": 0.5, "learning_rate": 1.9999639386009552e-06, "loss": 0.8784, "step": 12039 }, { "epoch": 0.5, "learning_rate": 1.9986706773851627e-06, "loss": 0.9434, "step": 12040 }, { "epoch": 0.5, "learning_rate": 1.9973777880160937e-06, "loss": 0.9221, "step": 12041 }, { "epoch": 0.5, "learning_rate": 1.996085270553836e-06, "loss": 0.8675, "step": 12042 }, { "epoch": 0.5, "learning_rate": 1.9947931250584485e-06, "loss": 0.9149, "step": 12043 }, { "epoch": 0.5, "learning_rate": 1.9935013515899927e-06, "loss": 0.9512, "step": 12044 }, { "epoch": 0.5, "learning_rate": 1.9922099502084934e-06, "loss": 0.9494, "step": 12045 }, { "epoch": 0.5, "learning_rate": 1.9909189209739675e-06, "loss": 0.9122, "step": 12046 }, { "epoch": 0.5, "learning_rate": 1.989628263946416e-06, "loss": 0.9424, "step": 12047 }, { "epoch": 0.5, "learning_rate": 1.988337979185816e-06, "loss": 0.9395, "step": 12048 }, { "epoch": 0.5, "learning_rate": 1.987048066752132e-06, "loss": 0.889, "step": 12049 }, { "epoch": 0.5, "learning_rate": 1.98575852670531e-06, "loss": 0.9246, "step": 12050 }, { "epoch": 0.5, "learning_rate": 1.9844693591052842e-06, "loss": 0.9112, "step": 12051 }, { "epoch": 0.5, "learning_rate": 1.983180564011956e-06, "loss": 0.9265, "step": 12052 }, { "epoch": 0.5, "learning_rate": 1.98189214148523e-06, "loss": 0.944, "step": 12053 }, { "epoch": 0.5, "learning_rate": 1.9806040915849777e-06, "loss": 0.9209, "step": 12054 }, { "epoch": 0.5, "learning_rate": 1.9793164143710588e-06, "loss": 0.9085, "step": 12055 }, { "epoch": 0.5, "learning_rate": 1.97802910990332e-06, "loss": 0.9462, "step": 12056 }, { "epoch": 0.5, "learning_rate": 1.9767421782415797e-06, "loss": 0.913, "step": 12057 }, { "epoch": 0.5, "learning_rate": 1.97545561944565e-06, "loss": 0.9371, "step": 12058 }, { "epoch": 0.5, "learning_rate": 1.9741694335753192e-06, "loss": 0.9459, "step": 12059 }, { "epoch": 0.5, "learning_rate": 1.972883620690366e-06, "loss": 0.9266, "step": 12060 }, { "epoch": 0.5, "learning_rate": 1.9715981808505346e-06, "loss": 0.9487, "step": 12061 }, { "epoch": 0.5, "learning_rate": 1.9703131141155763e-06, "loss": 0.9224, "step": 12062 }, { "epoch": 0.5, "learning_rate": 1.9690284205452058e-06, "loss": 0.9442, "step": 12063 }, { "epoch": 0.5, "learning_rate": 1.9677441001991227e-06, "loss": 0.9244, "step": 12064 }, { "epoch": 0.5, "learning_rate": 1.966460153137023e-06, "loss": 0.964, "step": 12065 }, { "epoch": 0.5, "learning_rate": 1.9651765794185686e-06, "loss": 0.957, "step": 12066 }, { "epoch": 0.5, "learning_rate": 1.963893379103413e-06, "loss": 0.9345, "step": 12067 }, { "epoch": 0.5, "learning_rate": 1.9626105522511917e-06, "loss": 0.9383, "step": 12068 }, { "epoch": 0.5, "learning_rate": 1.961328098921522e-06, "loss": 0.9047, "step": 12069 }, { "epoch": 0.5, "learning_rate": 1.960046019174e-06, "loss": 0.9428, "step": 12070 }, { "epoch": 0.5, "learning_rate": 1.958764313068211e-06, "loss": 0.9409, "step": 12071 }, { "epoch": 0.5, "learning_rate": 1.9574829806637206e-06, "loss": 0.9218, "step": 12072 }, { "epoch": 0.5, "learning_rate": 1.956202022020073e-06, "loss": 0.9172, "step": 12073 }, { "epoch": 0.5, "learning_rate": 1.9549214371968008e-06, "loss": 0.9365, "step": 12074 }, { "epoch": 0.5, "learning_rate": 1.9536412262534145e-06, "loss": 0.9206, "step": 12075 }, { "epoch": 0.5, "learning_rate": 1.952361389249412e-06, "loss": 0.8818, "step": 12076 }, { "epoch": 0.5, "learning_rate": 1.9510819262442705e-06, "loss": 0.9306, "step": 12077 }, { "epoch": 0.5, "learning_rate": 1.949802837297453e-06, "loss": 0.9788, "step": 12078 }, { "epoch": 0.5, "learning_rate": 1.9485241224683983e-06, "loss": 0.9277, "step": 12079 }, { "epoch": 0.5, "learning_rate": 1.9472457818165335e-06, "loss": 0.9649, "step": 12080 }, { "epoch": 0.5, "learning_rate": 1.9459678154012708e-06, "loss": 0.9373, "step": 12081 }, { "epoch": 0.5, "learning_rate": 1.9446902232819963e-06, "loss": 0.9512, "step": 12082 }, { "epoch": 0.5, "learning_rate": 1.9434130055180843e-06, "loss": 0.9464, "step": 12083 }, { "epoch": 0.5, "learning_rate": 1.942136162168894e-06, "loss": 0.943, "step": 12084 }, { "epoch": 0.5, "learning_rate": 1.940859693293765e-06, "loss": 0.9091, "step": 12085 }, { "epoch": 0.5, "learning_rate": 1.939583598952014e-06, "loss": 0.9079, "step": 12086 }, { "epoch": 0.5, "learning_rate": 1.9383078792029474e-06, "loss": 0.9539, "step": 12087 }, { "epoch": 0.5, "learning_rate": 1.937032534105855e-06, "loss": 0.9407, "step": 12088 }, { "epoch": 0.5, "learning_rate": 1.935757563719999e-06, "loss": 0.9192, "step": 12089 }, { "epoch": 0.5, "learning_rate": 1.9344829681046394e-06, "loss": 0.9208, "step": 12090 }, { "epoch": 0.5, "learning_rate": 1.933208747319004e-06, "loss": 0.9004, "step": 12091 }, { "epoch": 0.5, "learning_rate": 1.9319349014223132e-06, "loss": 0.9068, "step": 12092 }, { "epoch": 0.5, "learning_rate": 1.930661430473768e-06, "loss": 0.9316, "step": 12093 }, { "epoch": 0.5, "learning_rate": 1.929388334532547e-06, "loss": 0.9255, "step": 12094 }, { "epoch": 0.5, "learning_rate": 1.928115613657816e-06, "loss": 0.9434, "step": 12095 }, { "epoch": 0.5, "learning_rate": 1.926843267908722e-06, "loss": 0.9547, "step": 12096 }, { "epoch": 0.5, "learning_rate": 1.9255712973443987e-06, "loss": 0.8891, "step": 12097 }, { "epoch": 0.5, "learning_rate": 1.9242997020239506e-06, "loss": 0.9343, "step": 12098 }, { "epoch": 0.5, "learning_rate": 1.9230284820064826e-06, "loss": 0.914, "step": 12099 }, { "epoch": 0.51, "learning_rate": 1.9217576373510637e-06, "loss": 0.8839, "step": 12100 }, { "epoch": 0.51, "learning_rate": 1.9204871681167594e-06, "loss": 0.9475, "step": 12101 }, { "epoch": 0.51, "learning_rate": 1.9192170743626115e-06, "loss": 0.9144, "step": 12102 }, { "epoch": 0.51, "learning_rate": 1.9179473561476414e-06, "loss": 0.9691, "step": 12103 }, { "epoch": 0.51, "learning_rate": 1.9166780135308605e-06, "loss": 0.9358, "step": 12104 }, { "epoch": 0.51, "learning_rate": 1.9154090465712584e-06, "loss": 0.9173, "step": 12105 }, { "epoch": 0.51, "learning_rate": 1.9141404553278097e-06, "loss": 0.9185, "step": 12106 }, { "epoch": 0.51, "learning_rate": 1.9128722398594633e-06, "loss": 0.9441, "step": 12107 }, { "epoch": 0.51, "learning_rate": 1.9116044002251675e-06, "loss": 0.9465, "step": 12108 }, { "epoch": 0.51, "learning_rate": 1.9103369364838365e-06, "loss": 0.9001, "step": 12109 }, { "epoch": 0.51, "learning_rate": 1.9090698486943683e-06, "loss": 0.9306, "step": 12110 }, { "epoch": 0.51, "learning_rate": 1.9078031369156593e-06, "loss": 0.9342, "step": 12111 }, { "epoch": 0.51, "learning_rate": 1.90653680120657e-06, "loss": 0.9337, "step": 12112 }, { "epoch": 0.51, "learning_rate": 1.9052708416259525e-06, "loss": 0.9245, "step": 12113 }, { "epoch": 0.51, "learning_rate": 1.9040052582326407e-06, "loss": 0.92, "step": 12114 }, { "epoch": 0.51, "learning_rate": 1.9027400510854533e-06, "loss": 0.8853, "step": 12115 }, { "epoch": 0.51, "learning_rate": 1.901475220243182e-06, "loss": 0.9301, "step": 12116 }, { "epoch": 0.51, "learning_rate": 1.9002107657646107e-06, "loss": 0.959, "step": 12117 }, { "epoch": 0.51, "learning_rate": 1.898946687708505e-06, "loss": 0.9228, "step": 12118 }, { "epoch": 0.51, "learning_rate": 1.8976829861336054e-06, "loss": 0.9374, "step": 12119 }, { "epoch": 0.51, "learning_rate": 1.896419661098643e-06, "loss": 0.8886, "step": 12120 }, { "epoch": 0.51, "learning_rate": 1.8951567126623272e-06, "loss": 0.9412, "step": 12121 }, { "epoch": 0.51, "learning_rate": 1.8938941408833521e-06, "loss": 0.9374, "step": 12122 }, { "epoch": 0.51, "learning_rate": 1.8926319458203946e-06, "loss": 0.938, "step": 12123 }, { "epoch": 0.51, "learning_rate": 1.8913701275321128e-06, "loss": 0.9337, "step": 12124 }, { "epoch": 0.51, "learning_rate": 1.8901086860771445e-06, "loss": 0.8944, "step": 12125 }, { "epoch": 0.51, "learning_rate": 1.888847621514114e-06, "loss": 0.9076, "step": 12126 }, { "epoch": 0.51, "learning_rate": 1.8875869339016296e-06, "loss": 0.9613, "step": 12127 }, { "epoch": 0.51, "learning_rate": 1.8863266232982746e-06, "loss": 0.9551, "step": 12128 }, { "epoch": 0.51, "learning_rate": 1.8850666897626224e-06, "loss": 0.9429, "step": 12129 }, { "epoch": 0.51, "learning_rate": 1.8838071333532248e-06, "loss": 0.9247, "step": 12130 }, { "epoch": 0.51, "learning_rate": 1.8825479541286218e-06, "loss": 0.9411, "step": 12131 }, { "epoch": 0.51, "learning_rate": 1.8812891521473242e-06, "loss": 0.9192, "step": 12132 }, { "epoch": 0.51, "learning_rate": 1.8800307274678364e-06, "loss": 0.9312, "step": 12133 }, { "epoch": 0.51, "learning_rate": 1.8787726801486438e-06, "loss": 0.8928, "step": 12134 }, { "epoch": 0.51, "learning_rate": 1.8775150102482042e-06, "loss": 0.9246, "step": 12135 }, { "epoch": 0.51, "learning_rate": 1.8762577178249753e-06, "loss": 0.8957, "step": 12136 }, { "epoch": 0.51, "learning_rate": 1.875000802937379e-06, "loss": 0.8881, "step": 12137 }, { "epoch": 0.51, "learning_rate": 1.873744265643831e-06, "loss": 0.9228, "step": 12138 }, { "epoch": 0.51, "learning_rate": 1.8724881060027301e-06, "loss": 0.9226, "step": 12139 }, { "epoch": 0.51, "learning_rate": 1.8712323240724483e-06, "loss": 0.9342, "step": 12140 }, { "epoch": 0.51, "learning_rate": 1.869976919911347e-06, "loss": 0.8731, "step": 12141 }, { "epoch": 0.51, "learning_rate": 1.8687218935777706e-06, "loss": 0.9635, "step": 12142 }, { "epoch": 0.51, "learning_rate": 1.8674672451300469e-06, "loss": 0.9631, "step": 12143 }, { "epoch": 0.51, "learning_rate": 1.8662129746264735e-06, "loss": 0.9114, "step": 12144 }, { "epoch": 0.51, "learning_rate": 1.8649590821253515e-06, "loss": 0.9443, "step": 12145 }, { "epoch": 0.51, "learning_rate": 1.8637055676849458e-06, "loss": 0.97, "step": 12146 }, { "epoch": 0.51, "learning_rate": 1.862452431363514e-06, "loss": 0.9299, "step": 12147 }, { "epoch": 0.51, "learning_rate": 1.861199673219295e-06, "loss": 0.95, "step": 12148 }, { "epoch": 0.51, "learning_rate": 1.8599472933105033e-06, "loss": 0.9444, "step": 12149 }, { "epoch": 0.51, "learning_rate": 1.8586952916953438e-06, "loss": 0.9154, "step": 12150 }, { "epoch": 0.51, "learning_rate": 1.8574436684320008e-06, "loss": 0.9833, "step": 12151 }, { "epoch": 0.51, "learning_rate": 1.8561924235786445e-06, "loss": 0.9614, "step": 12152 }, { "epoch": 0.51, "learning_rate": 1.8549415571934148e-06, "loss": 0.9126, "step": 12153 }, { "epoch": 0.51, "learning_rate": 1.8536910693344557e-06, "loss": 0.91, "step": 12154 }, { "epoch": 0.51, "learning_rate": 1.8524409600598748e-06, "loss": 0.9416, "step": 12155 }, { "epoch": 0.51, "learning_rate": 1.8511912294277622e-06, "loss": 0.9094, "step": 12156 }, { "epoch": 0.51, "learning_rate": 1.8499418774962097e-06, "loss": 0.9623, "step": 12157 }, { "epoch": 0.51, "learning_rate": 1.8486929043232693e-06, "loss": 0.9141, "step": 12158 }, { "epoch": 0.51, "learning_rate": 1.847444309966988e-06, "loss": 0.9238, "step": 12159 }, { "epoch": 0.51, "learning_rate": 1.8461960944853918e-06, "loss": 0.9111, "step": 12160 }, { "epoch": 0.51, "learning_rate": 1.8449482579364908e-06, "loss": 0.9242, "step": 12161 }, { "epoch": 0.51, "learning_rate": 1.843700800378271e-06, "loss": 0.8925, "step": 12162 }, { "epoch": 0.51, "learning_rate": 1.8424537218687088e-06, "loss": 0.9238, "step": 12163 }, { "epoch": 0.51, "learning_rate": 1.841207022465763e-06, "loss": 0.9478, "step": 12164 }, { "epoch": 0.51, "learning_rate": 1.8399607022273624e-06, "loss": 0.9558, "step": 12165 }, { "epoch": 0.51, "learning_rate": 1.8387147612114376e-06, "loss": 0.9228, "step": 12166 }, { "epoch": 0.51, "learning_rate": 1.8374691994758853e-06, "loss": 0.8964, "step": 12167 }, { "epoch": 0.51, "learning_rate": 1.8362240170785916e-06, "loss": 0.9393, "step": 12168 }, { "epoch": 0.51, "learning_rate": 1.8349792140774258e-06, "loss": 0.9386, "step": 12169 }, { "epoch": 0.51, "learning_rate": 1.8337347905302383e-06, "loss": 0.9431, "step": 12170 }, { "epoch": 0.51, "learning_rate": 1.8324907464948571e-06, "loss": 0.9078, "step": 12171 }, { "epoch": 0.51, "learning_rate": 1.8312470820290994e-06, "loss": 0.9158, "step": 12172 }, { "epoch": 0.51, "learning_rate": 1.8300037971907658e-06, "loss": 0.9633, "step": 12173 }, { "epoch": 0.51, "learning_rate": 1.8287608920376288e-06, "loss": 0.9112, "step": 12174 }, { "epoch": 0.51, "learning_rate": 1.8275183666274532e-06, "loss": 0.9159, "step": 12175 }, { "epoch": 0.51, "learning_rate": 1.826276221017983e-06, "loss": 0.9642, "step": 12176 }, { "epoch": 0.51, "learning_rate": 1.8250344552669473e-06, "loss": 0.9062, "step": 12177 }, { "epoch": 0.51, "learning_rate": 1.8237930694320494e-06, "loss": 0.9548, "step": 12178 }, { "epoch": 0.51, "learning_rate": 1.8225520635709836e-06, "loss": 0.9565, "step": 12179 }, { "epoch": 0.51, "learning_rate": 1.821311437741422e-06, "loss": 0.9131, "step": 12180 }, { "epoch": 0.51, "learning_rate": 1.8200711920010228e-06, "loss": 0.917, "step": 12181 }, { "epoch": 0.51, "learning_rate": 1.8188313264074243e-06, "loss": 0.9247, "step": 12182 }, { "epoch": 0.51, "learning_rate": 1.8175918410182425e-06, "loss": 0.933, "step": 12183 }, { "epoch": 0.51, "learning_rate": 1.8163527358910825e-06, "loss": 0.9664, "step": 12184 }, { "epoch": 0.51, "learning_rate": 1.815114011083532e-06, "loss": 0.9461, "step": 12185 }, { "epoch": 0.51, "learning_rate": 1.8138756666531542e-06, "loss": 0.9341, "step": 12186 }, { "epoch": 0.51, "learning_rate": 1.8126377026575015e-06, "loss": 0.9635, "step": 12187 }, { "epoch": 0.51, "learning_rate": 1.8114001191541041e-06, "loss": 0.9492, "step": 12188 }, { "epoch": 0.51, "learning_rate": 1.8101629162004798e-06, "loss": 0.9105, "step": 12189 }, { "epoch": 0.51, "learning_rate": 1.8089260938541186e-06, "loss": 0.9469, "step": 12190 }, { "epoch": 0.51, "learning_rate": 1.8076896521725084e-06, "loss": 0.9027, "step": 12191 }, { "epoch": 0.51, "learning_rate": 1.8064535912131032e-06, "loss": 0.9574, "step": 12192 }, { "epoch": 0.51, "learning_rate": 1.805217911033349e-06, "loss": 0.9215, "step": 12193 }, { "epoch": 0.51, "learning_rate": 1.8039826116906755e-06, "loss": 0.9343, "step": 12194 }, { "epoch": 0.51, "learning_rate": 1.8027476932424848e-06, "loss": 0.9265, "step": 12195 }, { "epoch": 0.51, "learning_rate": 1.8015131557461695e-06, "loss": 0.9021, "step": 12196 }, { "epoch": 0.51, "learning_rate": 1.800278999259103e-06, "loss": 0.9332, "step": 12197 }, { "epoch": 0.51, "learning_rate": 1.7990452238386436e-06, "loss": 0.9522, "step": 12198 }, { "epoch": 0.51, "learning_rate": 1.79781182954212e-06, "loss": 0.9262, "step": 12199 }, { "epoch": 0.51, "learning_rate": 1.7965788164268615e-06, "loss": 0.9224, "step": 12200 }, { "epoch": 0.51, "learning_rate": 1.7953461845501663e-06, "loss": 0.8847, "step": 12201 }, { "epoch": 0.51, "learning_rate": 1.7941139339693137e-06, "loss": 0.9303, "step": 12202 }, { "epoch": 0.51, "learning_rate": 1.79288206474158e-06, "loss": 0.9334, "step": 12203 }, { "epoch": 0.51, "learning_rate": 1.7916505769242055e-06, "loss": 0.8914, "step": 12204 }, { "epoch": 0.51, "learning_rate": 1.7904194705744249e-06, "loss": 0.9608, "step": 12205 }, { "epoch": 0.51, "learning_rate": 1.7891887457494517e-06, "loss": 0.9241, "step": 12206 }, { "epoch": 0.51, "learning_rate": 1.7879584025064822e-06, "loss": 0.9039, "step": 12207 }, { "epoch": 0.51, "learning_rate": 1.786728440902692e-06, "loss": 0.9687, "step": 12208 }, { "epoch": 0.51, "learning_rate": 1.7854988609952417e-06, "loss": 0.9695, "step": 12209 }, { "epoch": 0.51, "learning_rate": 1.784269662841277e-06, "loss": 0.8782, "step": 12210 }, { "epoch": 0.51, "learning_rate": 1.7830408464979144e-06, "loss": 0.9598, "step": 12211 }, { "epoch": 0.51, "learning_rate": 1.7818124120222712e-06, "loss": 0.9072, "step": 12212 }, { "epoch": 0.51, "learning_rate": 1.78058435947143e-06, "loss": 0.966, "step": 12213 }, { "epoch": 0.51, "learning_rate": 1.779356688902464e-06, "loss": 0.9451, "step": 12214 }, { "epoch": 0.51, "learning_rate": 1.778129400372427e-06, "loss": 0.9088, "step": 12215 }, { "epoch": 0.51, "learning_rate": 1.7769024939383561e-06, "loss": 0.9244, "step": 12216 }, { "epoch": 0.51, "learning_rate": 1.7756759696572668e-06, "loss": 0.9104, "step": 12217 }, { "epoch": 0.51, "learning_rate": 1.7744498275861598e-06, "loss": 0.9422, "step": 12218 }, { "epoch": 0.51, "learning_rate": 1.7732240677820223e-06, "loss": 0.916, "step": 12219 }, { "epoch": 0.51, "learning_rate": 1.7719986903018127e-06, "loss": 0.9215, "step": 12220 }, { "epoch": 0.51, "learning_rate": 1.7707736952024811e-06, "loss": 0.9066, "step": 12221 }, { "epoch": 0.51, "learning_rate": 1.7695490825409577e-06, "loss": 0.9444, "step": 12222 }, { "epoch": 0.51, "learning_rate": 1.7683248523741548e-06, "loss": 0.9138, "step": 12223 }, { "epoch": 0.51, "learning_rate": 1.767101004758962e-06, "loss": 0.9085, "step": 12224 }, { "epoch": 0.51, "learning_rate": 1.7658775397522587e-06, "loss": 0.9103, "step": 12225 }, { "epoch": 0.51, "learning_rate": 1.7646544574109015e-06, "loss": 0.9789, "step": 12226 }, { "epoch": 0.51, "learning_rate": 1.7634317577917325e-06, "loss": 0.9465, "step": 12227 }, { "epoch": 0.51, "learning_rate": 1.7622094409515754e-06, "loss": 0.9419, "step": 12228 }, { "epoch": 0.51, "learning_rate": 1.7609875069472303e-06, "loss": 0.9274, "step": 12229 }, { "epoch": 0.51, "learning_rate": 1.7597659558354884e-06, "loss": 0.9743, "step": 12230 }, { "epoch": 0.51, "learning_rate": 1.7585447876731198e-06, "loss": 0.921, "step": 12231 }, { "epoch": 0.51, "learning_rate": 1.7573240025168715e-06, "loss": 0.9612, "step": 12232 }, { "epoch": 0.51, "learning_rate": 1.75610360042348e-06, "loss": 0.9228, "step": 12233 }, { "epoch": 0.51, "learning_rate": 1.7548835814496601e-06, "loss": 0.9189, "step": 12234 }, { "epoch": 0.51, "learning_rate": 1.753663945652113e-06, "loss": 0.9632, "step": 12235 }, { "epoch": 0.51, "learning_rate": 1.7524446930875127e-06, "loss": 0.9352, "step": 12236 }, { "epoch": 0.51, "learning_rate": 1.7512258238125302e-06, "loss": 0.94, "step": 12237 }, { "epoch": 0.51, "learning_rate": 1.7500073378838033e-06, "loss": 0.8721, "step": 12238 }, { "epoch": 0.51, "learning_rate": 1.7487892353579605e-06, "loss": 0.9316, "step": 12239 }, { "epoch": 0.51, "learning_rate": 1.747571516291614e-06, "loss": 0.9295, "step": 12240 }, { "epoch": 0.51, "learning_rate": 1.746354180741351e-06, "loss": 0.9165, "step": 12241 }, { "epoch": 0.51, "learning_rate": 1.7451372287637446e-06, "loss": 0.9115, "step": 12242 }, { "epoch": 0.51, "learning_rate": 1.7439206604153535e-06, "loss": 0.9321, "step": 12243 }, { "epoch": 0.51, "learning_rate": 1.7427044757527157e-06, "loss": 0.9445, "step": 12244 }, { "epoch": 0.51, "learning_rate": 1.7414886748323445e-06, "loss": 0.9825, "step": 12245 }, { "epoch": 0.51, "learning_rate": 1.7402732577107517e-06, "loss": 0.9154, "step": 12246 }, { "epoch": 0.51, "learning_rate": 1.7390582244444177e-06, "loss": 0.9789, "step": 12247 }, { "epoch": 0.51, "learning_rate": 1.7378435750898015e-06, "loss": 0.9664, "step": 12248 }, { "epoch": 0.51, "learning_rate": 1.7366293097033638e-06, "loss": 0.9073, "step": 12249 }, { "epoch": 0.52, "learning_rate": 1.735415428341527e-06, "loss": 0.9567, "step": 12250 }, { "epoch": 0.52, "learning_rate": 1.7342019310607062e-06, "loss": 0.9134, "step": 12251 }, { "epoch": 0.52, "learning_rate": 1.732988817917297e-06, "loss": 0.9527, "step": 12252 }, { "epoch": 0.52, "learning_rate": 1.7317760889676781e-06, "loss": 0.9337, "step": 12253 }, { "epoch": 0.52, "learning_rate": 1.7305637442682056e-06, "loss": 0.9186, "step": 12254 }, { "epoch": 0.52, "learning_rate": 1.7293517838752205e-06, "loss": 0.9122, "step": 12255 }, { "epoch": 0.52, "learning_rate": 1.7281402078450516e-06, "loss": 0.9238, "step": 12256 }, { "epoch": 0.52, "learning_rate": 1.7269290162339958e-06, "loss": 0.974, "step": 12257 }, { "epoch": 0.52, "learning_rate": 1.7257182090983515e-06, "loss": 0.9196, "step": 12258 }, { "epoch": 0.52, "learning_rate": 1.7245077864943816e-06, "loss": 0.9347, "step": 12259 }, { "epoch": 0.52, "learning_rate": 1.7232977484783398e-06, "loss": 0.9109, "step": 12260 }, { "epoch": 0.52, "learning_rate": 1.72208809510646e-06, "loss": 0.9342, "step": 12261 }, { "epoch": 0.52, "learning_rate": 1.7208788264349618e-06, "loss": 0.9227, "step": 12262 }, { "epoch": 0.52, "learning_rate": 1.719669942520039e-06, "loss": 0.9133, "step": 12263 }, { "epoch": 0.52, "learning_rate": 1.7184614434178737e-06, "loss": 0.9338, "step": 12264 }, { "epoch": 0.52, "learning_rate": 1.7172533291846316e-06, "loss": 0.903, "step": 12265 }, { "epoch": 0.52, "learning_rate": 1.7160455998764537e-06, "loss": 0.9559, "step": 12266 }, { "epoch": 0.52, "learning_rate": 1.714838255549468e-06, "loss": 0.9095, "step": 12267 }, { "epoch": 0.52, "learning_rate": 1.7136312962597833e-06, "loss": 0.9379, "step": 12268 }, { "epoch": 0.52, "learning_rate": 1.7124247220634915e-06, "loss": 0.9074, "step": 12269 }, { "epoch": 0.52, "learning_rate": 1.7112185330166686e-06, "loss": 0.933, "step": 12270 }, { "epoch": 0.52, "learning_rate": 1.7100127291753642e-06, "loss": 0.9652, "step": 12271 }, { "epoch": 0.52, "learning_rate": 1.7088073105956193e-06, "loss": 0.9273, "step": 12272 }, { "epoch": 0.52, "learning_rate": 1.7076022773334522e-06, "loss": 0.9426, "step": 12273 }, { "epoch": 0.52, "learning_rate": 1.7063976294448681e-06, "loss": 0.927, "step": 12274 }, { "epoch": 0.52, "learning_rate": 1.7051933669858456e-06, "loss": 0.9357, "step": 12275 }, { "epoch": 0.52, "learning_rate": 1.7039894900123532e-06, "loss": 0.9353, "step": 12276 }, { "epoch": 0.52, "learning_rate": 1.7027859985803396e-06, "loss": 0.9499, "step": 12277 }, { "epoch": 0.52, "learning_rate": 1.7015828927457324e-06, "loss": 0.9473, "step": 12278 }, { "epoch": 0.52, "learning_rate": 1.7003801725644432e-06, "loss": 0.8922, "step": 12279 }, { "epoch": 0.52, "learning_rate": 1.6991778380923696e-06, "loss": 0.9413, "step": 12280 }, { "epoch": 0.52, "learning_rate": 1.697975889385387e-06, "loss": 0.8619, "step": 12281 }, { "epoch": 0.52, "learning_rate": 1.6967743264993486e-06, "loss": 0.9367, "step": 12282 }, { "epoch": 0.52, "learning_rate": 1.695573149490104e-06, "loss": 0.9315, "step": 12283 }, { "epoch": 0.52, "learning_rate": 1.6943723584134687e-06, "loss": 0.9334, "step": 12284 }, { "epoch": 0.52, "learning_rate": 1.6931719533252477e-06, "loss": 0.9301, "step": 12285 }, { "epoch": 0.52, "learning_rate": 1.6919719342812324e-06, "loss": 0.9174, "step": 12286 }, { "epoch": 0.52, "learning_rate": 1.6907723013371858e-06, "loss": 0.9371, "step": 12287 }, { "epoch": 0.52, "learning_rate": 1.6895730545488597e-06, "loss": 0.917, "step": 12288 }, { "epoch": 0.52, "learning_rate": 1.6883741939719888e-06, "loss": 0.919, "step": 12289 }, { "epoch": 0.52, "learning_rate": 1.6871757196622885e-06, "loss": 0.9543, "step": 12290 }, { "epoch": 0.52, "learning_rate": 1.6859776316754506e-06, "loss": 0.962, "step": 12291 }, { "epoch": 0.52, "learning_rate": 1.6847799300671608e-06, "loss": 0.9437, "step": 12292 }, { "epoch": 0.52, "learning_rate": 1.6835826148930778e-06, "loss": 0.964, "step": 12293 }, { "epoch": 0.52, "learning_rate": 1.6823856862088384e-06, "loss": 0.9449, "step": 12294 }, { "epoch": 0.52, "learning_rate": 1.681189144070078e-06, "loss": 0.9259, "step": 12295 }, { "epoch": 0.52, "learning_rate": 1.6799929885323951e-06, "loss": 0.9457, "step": 12296 }, { "epoch": 0.52, "learning_rate": 1.6787972196513824e-06, "loss": 0.9076, "step": 12297 }, { "epoch": 0.52, "learning_rate": 1.6776018374826108e-06, "loss": 0.9188, "step": 12298 }, { "epoch": 0.52, "learning_rate": 1.676406842081636e-06, "loss": 0.9051, "step": 12299 }, { "epoch": 0.52, "learning_rate": 1.6752122335039879e-06, "loss": 0.8958, "step": 12300 }, { "epoch": 0.52, "learning_rate": 1.6740180118051863e-06, "loss": 0.9417, "step": 12301 }, { "epoch": 0.52, "learning_rate": 1.6728241770407328e-06, "loss": 0.9349, "step": 12302 }, { "epoch": 0.52, "learning_rate": 1.6716307292661005e-06, "loss": 0.9089, "step": 12303 }, { "epoch": 0.52, "learning_rate": 1.6704376685367641e-06, "loss": 0.9246, "step": 12304 }, { "epoch": 0.52, "learning_rate": 1.6692449949081601e-06, "loss": 0.933, "step": 12305 }, { "epoch": 0.52, "learning_rate": 1.668052708435719e-06, "loss": 0.9386, "step": 12306 }, { "epoch": 0.52, "learning_rate": 1.6668608091748495e-06, "loss": 0.9117, "step": 12307 }, { "epoch": 0.52, "learning_rate": 1.6656692971809463e-06, "loss": 0.8782, "step": 12308 }, { "epoch": 0.52, "learning_rate": 1.664478172509376e-06, "loss": 0.8845, "step": 12309 }, { "epoch": 0.52, "learning_rate": 1.6632874352154982e-06, "loss": 0.951, "step": 12310 }, { "epoch": 0.52, "learning_rate": 1.6620970853546515e-06, "loss": 0.9416, "step": 12311 }, { "epoch": 0.52, "learning_rate": 1.6609071229821517e-06, "loss": 0.9221, "step": 12312 }, { "epoch": 0.52, "learning_rate": 1.6597175481533001e-06, "loss": 0.9267, "step": 12313 }, { "epoch": 0.52, "learning_rate": 1.6585283609233816e-06, "loss": 0.9367, "step": 12314 }, { "epoch": 0.52, "learning_rate": 1.6573395613476617e-06, "loss": 0.9253, "step": 12315 }, { "epoch": 0.52, "learning_rate": 1.6561511494813886e-06, "loss": 0.9585, "step": 12316 }, { "epoch": 0.52, "learning_rate": 1.6549631253797882e-06, "loss": 0.8864, "step": 12317 }, { "epoch": 0.52, "learning_rate": 1.653775489098073e-06, "loss": 0.9059, "step": 12318 }, { "epoch": 0.52, "learning_rate": 1.6525882406914372e-06, "loss": 0.945, "step": 12319 }, { "epoch": 0.52, "learning_rate": 1.6514013802150575e-06, "loss": 0.9493, "step": 12320 }, { "epoch": 0.52, "learning_rate": 1.650214907724086e-06, "loss": 0.959, "step": 12321 }, { "epoch": 0.52, "learning_rate": 1.6490288232736652e-06, "loss": 0.9133, "step": 12322 }, { "epoch": 0.52, "learning_rate": 1.6478431269189177e-06, "loss": 0.9595, "step": 12323 }, { "epoch": 0.52, "learning_rate": 1.6466578187149419e-06, "loss": 0.9491, "step": 12324 }, { "epoch": 0.52, "learning_rate": 1.6454728987168257e-06, "loss": 0.9233, "step": 12325 }, { "epoch": 0.52, "learning_rate": 1.644288366979635e-06, "loss": 0.9032, "step": 12326 }, { "epoch": 0.52, "learning_rate": 1.6431042235584217e-06, "loss": 0.9274, "step": 12327 }, { "epoch": 0.52, "learning_rate": 1.641920468508209e-06, "loss": 0.969, "step": 12328 }, { "epoch": 0.52, "learning_rate": 1.6407371018840212e-06, "loss": 0.9336, "step": 12329 }, { "epoch": 0.52, "learning_rate": 1.639554123740843e-06, "loss": 0.903, "step": 12330 }, { "epoch": 0.52, "learning_rate": 1.6383715341336547e-06, "loss": 0.893, "step": 12331 }, { "epoch": 0.52, "learning_rate": 1.6371893331174182e-06, "loss": 0.9193, "step": 12332 }, { "epoch": 0.52, "learning_rate": 1.6360075207470683e-06, "loss": 0.9391, "step": 12333 }, { "epoch": 0.52, "learning_rate": 1.6348260970775299e-06, "loss": 0.8881, "step": 12334 }, { "epoch": 0.52, "learning_rate": 1.6336450621637079e-06, "loss": 0.9455, "step": 12335 }, { "epoch": 0.52, "learning_rate": 1.6324644160604908e-06, "loss": 0.8964, "step": 12336 }, { "epoch": 0.52, "learning_rate": 1.631284158822739e-06, "loss": 0.8827, "step": 12337 }, { "epoch": 0.52, "learning_rate": 1.6301042905053133e-06, "loss": 0.8998, "step": 12338 }, { "epoch": 0.52, "learning_rate": 1.6289248111630419e-06, "loss": 0.9531, "step": 12339 }, { "epoch": 0.52, "learning_rate": 1.6277457208507319e-06, "loss": 0.9225, "step": 12340 }, { "epoch": 0.52, "learning_rate": 1.6265670196231898e-06, "loss": 0.9479, "step": 12341 }, { "epoch": 0.52, "learning_rate": 1.6253887075351871e-06, "loss": 0.9547, "step": 12342 }, { "epoch": 0.52, "learning_rate": 1.6242107846414856e-06, "loss": 0.9053, "step": 12343 }, { "epoch": 0.52, "learning_rate": 1.6230332509968262e-06, "loss": 0.8648, "step": 12344 }, { "epoch": 0.52, "learning_rate": 1.621856106655937e-06, "loss": 0.917, "step": 12345 }, { "epoch": 0.52, "learning_rate": 1.6206793516735165e-06, "loss": 0.9753, "step": 12346 }, { "epoch": 0.52, "learning_rate": 1.6195029861042555e-06, "loss": 0.934, "step": 12347 }, { "epoch": 0.52, "learning_rate": 1.6183270100028259e-06, "loss": 0.8949, "step": 12348 }, { "epoch": 0.52, "learning_rate": 1.6171514234238716e-06, "loss": 0.9476, "step": 12349 }, { "epoch": 0.52, "learning_rate": 1.6159762264220346e-06, "loss": 0.9161, "step": 12350 }, { "epoch": 0.52, "learning_rate": 1.6148014190519245e-06, "loss": 0.9694, "step": 12351 }, { "epoch": 0.52, "learning_rate": 1.61362700136814e-06, "loss": 0.9165, "step": 12352 }, { "epoch": 0.52, "learning_rate": 1.6124529734252592e-06, "loss": 0.9098, "step": 12353 }, { "epoch": 0.52, "learning_rate": 1.6112793352778465e-06, "loss": 0.9076, "step": 12354 }, { "epoch": 0.52, "learning_rate": 1.6101060869804386e-06, "loss": 0.9262, "step": 12355 }, { "epoch": 0.52, "learning_rate": 1.6089332285875624e-06, "loss": 0.9278, "step": 12356 }, { "epoch": 0.52, "learning_rate": 1.6077607601537283e-06, "loss": 0.921, "step": 12357 }, { "epoch": 0.52, "learning_rate": 1.6065886817334142e-06, "loss": 0.9138, "step": 12358 }, { "epoch": 0.52, "learning_rate": 1.6054169933811037e-06, "loss": 0.9585, "step": 12359 }, { "epoch": 0.52, "learning_rate": 1.6042456951512387e-06, "loss": 0.8907, "step": 12360 }, { "epoch": 0.52, "learning_rate": 1.6030747870982565e-06, "loss": 0.9415, "step": 12361 }, { "epoch": 0.52, "learning_rate": 1.601904269276575e-06, "loss": 0.934, "step": 12362 }, { "epoch": 0.52, "learning_rate": 1.6007341417405874e-06, "loss": 0.9326, "step": 12363 }, { "epoch": 0.52, "learning_rate": 1.5995644045446756e-06, "loss": 0.8869, "step": 12364 }, { "epoch": 0.52, "learning_rate": 1.5983950577432007e-06, "loss": 0.9446, "step": 12365 }, { "epoch": 0.52, "learning_rate": 1.5972261013905087e-06, "loss": 0.9015, "step": 12366 }, { "epoch": 0.52, "learning_rate": 1.5960575355409191e-06, "loss": 0.9318, "step": 12367 }, { "epoch": 0.52, "learning_rate": 1.5948893602487415e-06, "loss": 0.9585, "step": 12368 }, { "epoch": 0.52, "learning_rate": 1.5937215755682667e-06, "loss": 1.0009, "step": 12369 }, { "epoch": 0.52, "learning_rate": 1.5925541815537615e-06, "loss": 0.9485, "step": 12370 }, { "epoch": 0.52, "learning_rate": 1.5913871782594803e-06, "loss": 0.9101, "step": 12371 }, { "epoch": 0.52, "learning_rate": 1.5902205657396575e-06, "loss": 0.8983, "step": 12372 }, { "epoch": 0.52, "learning_rate": 1.5890543440485085e-06, "loss": 0.9098, "step": 12373 }, { "epoch": 0.52, "learning_rate": 1.587888513240231e-06, "loss": 0.9736, "step": 12374 }, { "epoch": 0.52, "learning_rate": 1.5867230733690086e-06, "loss": 0.9018, "step": 12375 }, { "epoch": 0.52, "learning_rate": 1.585558024488998e-06, "loss": 0.9502, "step": 12376 }, { "epoch": 0.52, "learning_rate": 1.5843933666543443e-06, "loss": 0.9208, "step": 12377 }, { "epoch": 0.52, "learning_rate": 1.5832290999191747e-06, "loss": 0.8904, "step": 12378 }, { "epoch": 0.52, "learning_rate": 1.5820652243375934e-06, "loss": 0.9282, "step": 12379 }, { "epoch": 0.52, "learning_rate": 1.5809017399636895e-06, "loss": 0.9343, "step": 12380 }, { "epoch": 0.52, "learning_rate": 1.5797386468515352e-06, "loss": 0.9554, "step": 12381 }, { "epoch": 0.52, "learning_rate": 1.578575945055184e-06, "loss": 0.939, "step": 12382 }, { "epoch": 0.52, "learning_rate": 1.577413634628665e-06, "loss": 0.9572, "step": 12383 }, { "epoch": 0.52, "learning_rate": 1.5762517156260039e-06, "loss": 0.9223, "step": 12384 }, { "epoch": 0.52, "learning_rate": 1.5750901881011916e-06, "loss": 0.9282, "step": 12385 }, { "epoch": 0.52, "learning_rate": 1.573929052108205e-06, "loss": 0.9436, "step": 12386 }, { "epoch": 0.52, "learning_rate": 1.5727683077010158e-06, "loss": 0.8789, "step": 12387 }, { "epoch": 0.52, "learning_rate": 1.5716079549335584e-06, "loss": 0.9482, "step": 12388 }, { "epoch": 0.52, "learning_rate": 1.5704479938597617e-06, "loss": 0.914, "step": 12389 }, { "epoch": 0.52, "learning_rate": 1.5692884245335326e-06, "loss": 0.961, "step": 12390 }, { "epoch": 0.52, "learning_rate": 1.5681292470087617e-06, "loss": 0.915, "step": 12391 }, { "epoch": 0.52, "learning_rate": 1.5669704613393145e-06, "loss": 0.8951, "step": 12392 }, { "epoch": 0.52, "learning_rate": 1.5658120675790478e-06, "loss": 0.9316, "step": 12393 }, { "epoch": 0.52, "learning_rate": 1.5646540657817955e-06, "loss": 0.933, "step": 12394 }, { "epoch": 0.52, "learning_rate": 1.5634964560013688e-06, "loss": 0.8997, "step": 12395 }, { "epoch": 0.52, "learning_rate": 1.5623392382915726e-06, "loss": 0.8962, "step": 12396 }, { "epoch": 0.52, "learning_rate": 1.5611824127061814e-06, "loss": 0.9352, "step": 12397 }, { "epoch": 0.52, "learning_rate": 1.560025979298958e-06, "loss": 0.9437, "step": 12398 }, { "epoch": 0.52, "learning_rate": 1.558869938123645e-06, "loss": 0.902, "step": 12399 }, { "epoch": 0.53, "learning_rate": 1.5577142892339702e-06, "loss": 0.9201, "step": 12400 }, { "epoch": 0.53, "learning_rate": 1.5565590326836355e-06, "loss": 0.9426, "step": 12401 }, { "epoch": 0.53, "learning_rate": 1.555404168526331e-06, "loss": 0.9121, "step": 12402 }, { "epoch": 0.53, "learning_rate": 1.5542496968157306e-06, "loss": 0.9449, "step": 12403 }, { "epoch": 0.53, "learning_rate": 1.5530956176054779e-06, "loss": 0.9226, "step": 12404 }, { "epoch": 0.53, "learning_rate": 1.5519419309492156e-06, "loss": 0.9231, "step": 12405 }, { "epoch": 0.53, "learning_rate": 1.550788636900553e-06, "loss": 0.9052, "step": 12406 }, { "epoch": 0.53, "learning_rate": 1.5496357355130897e-06, "loss": 0.8991, "step": 12407 }, { "epoch": 0.53, "learning_rate": 1.548483226840407e-06, "loss": 0.9494, "step": 12408 }, { "epoch": 0.53, "learning_rate": 1.5473311109360601e-06, "loss": 0.9114, "step": 12409 }, { "epoch": 0.53, "learning_rate": 1.5461793878535936e-06, "loss": 0.9147, "step": 12410 }, { "epoch": 0.53, "learning_rate": 1.5450280576465337e-06, "loss": 0.9111, "step": 12411 }, { "epoch": 0.53, "learning_rate": 1.5438771203683856e-06, "loss": 0.9238, "step": 12412 }, { "epoch": 0.53, "learning_rate": 1.542726576072635e-06, "loss": 0.9602, "step": 12413 }, { "epoch": 0.53, "learning_rate": 1.5415764248127518e-06, "loss": 0.9332, "step": 12414 }, { "epoch": 0.53, "learning_rate": 1.5404266666421908e-06, "loss": 0.9438, "step": 12415 }, { "epoch": 0.53, "learning_rate": 1.5392773016143781e-06, "loss": 0.9247, "step": 12416 }, { "epoch": 0.53, "learning_rate": 1.538128329782733e-06, "loss": 0.9357, "step": 12417 }, { "epoch": 0.53, "learning_rate": 1.5369797512006513e-06, "loss": 0.9432, "step": 12418 }, { "epoch": 0.53, "learning_rate": 1.5358315659215094e-06, "loss": 0.9073, "step": 12419 }, { "epoch": 0.53, "learning_rate": 1.5346837739986686e-06, "loss": 0.9167, "step": 12420 }, { "epoch": 0.53, "learning_rate": 1.5335363754854726e-06, "loss": 0.9476, "step": 12421 }, { "epoch": 0.53, "learning_rate": 1.53238937043524e-06, "loss": 0.947, "step": 12422 }, { "epoch": 0.53, "learning_rate": 1.5312427589012769e-06, "loss": 0.9312, "step": 12423 }, { "epoch": 0.53, "learning_rate": 1.5300965409368728e-06, "loss": 0.8904, "step": 12424 }, { "epoch": 0.53, "learning_rate": 1.528950716595292e-06, "loss": 0.9122, "step": 12425 }, { "epoch": 0.53, "learning_rate": 1.5278052859297866e-06, "loss": 0.9324, "step": 12426 }, { "epoch": 0.53, "learning_rate": 1.5266602489935878e-06, "loss": 0.9222, "step": 12427 }, { "epoch": 0.53, "learning_rate": 1.5255156058399124e-06, "loss": 0.9177, "step": 12428 }, { "epoch": 0.53, "learning_rate": 1.5243713565219477e-06, "loss": 0.9348, "step": 12429 }, { "epoch": 0.53, "learning_rate": 1.5232275010928789e-06, "loss": 0.9392, "step": 12430 }, { "epoch": 0.53, "learning_rate": 1.5220840396058624e-06, "loss": 0.9379, "step": 12431 }, { "epoch": 0.53, "learning_rate": 1.520940972114031e-06, "loss": 0.9086, "step": 12432 }, { "epoch": 0.53, "learning_rate": 1.5197982986705185e-06, "loss": 0.9368, "step": 12433 }, { "epoch": 0.53, "learning_rate": 1.5186560193284194e-06, "loss": 0.917, "step": 12434 }, { "epoch": 0.53, "learning_rate": 1.5175141341408227e-06, "loss": 0.9164, "step": 12435 }, { "epoch": 0.53, "learning_rate": 1.516372643160795e-06, "loss": 0.9107, "step": 12436 }, { "epoch": 0.53, "learning_rate": 1.515231546441387e-06, "loss": 0.9527, "step": 12437 }, { "epoch": 0.53, "learning_rate": 1.514090844035624e-06, "loss": 0.8709, "step": 12438 }, { "epoch": 0.53, "learning_rate": 1.5129505359965213e-06, "loss": 0.9521, "step": 12439 }, { "epoch": 0.53, "learning_rate": 1.5118106223770746e-06, "loss": 0.9409, "step": 12440 }, { "epoch": 0.53, "learning_rate": 1.5106711032302512e-06, "loss": 0.9517, "step": 12441 }, { "epoch": 0.53, "learning_rate": 1.509531978609019e-06, "loss": 0.9659, "step": 12442 }, { "epoch": 0.53, "learning_rate": 1.508393248566309e-06, "loss": 0.9415, "step": 12443 }, { "epoch": 0.53, "learning_rate": 1.507254913155044e-06, "loss": 0.9117, "step": 12444 }, { "epoch": 0.53, "learning_rate": 1.5061169724281254e-06, "loss": 0.9688, "step": 12445 }, { "epoch": 0.53, "learning_rate": 1.5049794264384398e-06, "loss": 0.9111, "step": 12446 }, { "epoch": 0.53, "learning_rate": 1.503842275238848e-06, "loss": 0.9123, "step": 12447 }, { "epoch": 0.53, "learning_rate": 1.502705518882198e-06, "loss": 0.8997, "step": 12448 }, { "epoch": 0.53, "learning_rate": 1.501569157421323e-06, "loss": 0.9181, "step": 12449 }, { "epoch": 0.53, "learning_rate": 1.500433190909024e-06, "loss": 0.9565, "step": 12450 }, { "epoch": 0.53, "learning_rate": 1.4992976193981046e-06, "loss": 0.9246, "step": 12451 }, { "epoch": 0.53, "learning_rate": 1.498162442941329e-06, "loss": 0.9222, "step": 12452 }, { "epoch": 0.53, "learning_rate": 1.4970276615914559e-06, "loss": 0.9395, "step": 12453 }, { "epoch": 0.53, "learning_rate": 1.4958932754012234e-06, "loss": 0.9424, "step": 12454 }, { "epoch": 0.53, "learning_rate": 1.4947592844233472e-06, "loss": 0.9166, "step": 12455 }, { "epoch": 0.53, "learning_rate": 1.4936256887105283e-06, "loss": 0.9656, "step": 12456 }, { "epoch": 0.53, "learning_rate": 1.492492488315449e-06, "loss": 0.9252, "step": 12457 }, { "epoch": 0.53, "learning_rate": 1.491359683290775e-06, "loss": 0.9318, "step": 12458 }, { "epoch": 0.53, "learning_rate": 1.4902272736891455e-06, "loss": 0.9347, "step": 12459 }, { "epoch": 0.53, "learning_rate": 1.4890952595631913e-06, "loss": 0.9073, "step": 12460 }, { "epoch": 0.53, "learning_rate": 1.4879636409655219e-06, "loss": 0.9242, "step": 12461 }, { "epoch": 0.53, "learning_rate": 1.4868324179487204e-06, "loss": 0.9151, "step": 12462 }, { "epoch": 0.53, "learning_rate": 1.485701590565367e-06, "loss": 0.9445, "step": 12463 }, { "epoch": 0.53, "learning_rate": 1.484571158868008e-06, "loss": 0.9223, "step": 12464 }, { "epoch": 0.53, "learning_rate": 1.4834411229091816e-06, "loss": 0.9659, "step": 12465 }, { "epoch": 0.53, "learning_rate": 1.482311482741402e-06, "loss": 0.9036, "step": 12466 }, { "epoch": 0.53, "learning_rate": 1.4811822384171714e-06, "loss": 0.9373, "step": 12467 }, { "epoch": 0.53, "learning_rate": 1.4800533899889635e-06, "loss": 0.8891, "step": 12468 }, { "epoch": 0.53, "learning_rate": 1.4789249375092418e-06, "loss": 0.9105, "step": 12469 }, { "epoch": 0.53, "learning_rate": 1.4777968810304522e-06, "loss": 0.9527, "step": 12470 }, { "epoch": 0.53, "learning_rate": 1.4766692206050137e-06, "loss": 0.9094, "step": 12471 }, { "epoch": 0.53, "learning_rate": 1.4755419562853345e-06, "loss": 0.9404, "step": 12472 }, { "epoch": 0.53, "learning_rate": 1.4744150881238018e-06, "loss": 0.9441, "step": 12473 }, { "epoch": 0.53, "learning_rate": 1.4732886161727877e-06, "loss": 0.9042, "step": 12474 }, { "epoch": 0.53, "learning_rate": 1.4721625404846351e-06, "loss": 0.9488, "step": 12475 }, { "epoch": 0.53, "learning_rate": 1.4710368611116866e-06, "loss": 0.8631, "step": 12476 }, { "epoch": 0.53, "learning_rate": 1.4699115781062478e-06, "loss": 0.904, "step": 12477 }, { "epoch": 0.53, "learning_rate": 1.4687866915206173e-06, "loss": 0.9473, "step": 12478 }, { "epoch": 0.53, "learning_rate": 1.467662201407074e-06, "loss": 0.9138, "step": 12479 }, { "epoch": 0.53, "learning_rate": 1.4665381078178731e-06, "loss": 0.9771, "step": 12480 }, { "epoch": 0.53, "learning_rate": 1.465414410805255e-06, "loss": 0.962, "step": 12481 }, { "epoch": 0.53, "learning_rate": 1.464291110421442e-06, "loss": 0.9, "step": 12482 }, { "epoch": 0.53, "learning_rate": 1.4631682067186404e-06, "loss": 0.9445, "step": 12483 }, { "epoch": 0.53, "learning_rate": 1.4620456997490306e-06, "loss": 0.9223, "step": 12484 }, { "epoch": 0.53, "learning_rate": 1.46092358956478e-06, "loss": 0.9204, "step": 12485 }, { "epoch": 0.53, "learning_rate": 1.4598018762180411e-06, "loss": 0.8905, "step": 12486 }, { "epoch": 0.53, "learning_rate": 1.4586805597609333e-06, "loss": 0.9438, "step": 12487 }, { "epoch": 0.53, "learning_rate": 1.4575596402455794e-06, "loss": 0.9328, "step": 12488 }, { "epoch": 0.53, "learning_rate": 1.4564391177240644e-06, "loss": 0.9536, "step": 12489 }, { "epoch": 0.53, "learning_rate": 1.4553189922484645e-06, "loss": 0.9117, "step": 12490 }, { "epoch": 0.53, "learning_rate": 1.4541992638708358e-06, "loss": 0.9564, "step": 12491 }, { "epoch": 0.53, "learning_rate": 1.4530799326432177e-06, "loss": 0.9398, "step": 12492 }, { "epoch": 0.53, "learning_rate": 1.4519609986176253e-06, "loss": 0.9199, "step": 12493 }, { "epoch": 0.53, "learning_rate": 1.4508424618460592e-06, "loss": 0.9428, "step": 12494 }, { "epoch": 0.53, "learning_rate": 1.4497243223805046e-06, "loss": 0.9387, "step": 12495 }, { "epoch": 0.53, "learning_rate": 1.4486065802729188e-06, "loss": 0.9448, "step": 12496 }, { "epoch": 0.53, "learning_rate": 1.4474892355752556e-06, "loss": 0.9534, "step": 12497 }, { "epoch": 0.53, "learning_rate": 1.4463722883394337e-06, "loss": 0.936, "step": 12498 }, { "epoch": 0.53, "learning_rate": 1.445255738617365e-06, "loss": 0.9484, "step": 12499 }, { "epoch": 0.53, "learning_rate": 1.4441395864609398e-06, "loss": 0.9217, "step": 12500 }, { "epoch": 0.53, "learning_rate": 1.4430238319220247e-06, "loss": 0.9044, "step": 12501 }, { "epoch": 0.53, "learning_rate": 1.4419084750524759e-06, "loss": 0.9085, "step": 12502 }, { "epoch": 0.53, "learning_rate": 1.4407935159041265e-06, "loss": 0.938, "step": 12503 }, { "epoch": 0.53, "learning_rate": 1.4396789545287949e-06, "loss": 0.8896, "step": 12504 }, { "epoch": 0.53, "learning_rate": 1.438564790978273e-06, "loss": 0.9548, "step": 12505 }, { "epoch": 0.53, "learning_rate": 1.4374510253043416e-06, "loss": 0.9101, "step": 12506 }, { "epoch": 0.53, "learning_rate": 1.436337657558765e-06, "loss": 0.8986, "step": 12507 }, { "epoch": 0.53, "learning_rate": 1.4352246877932753e-06, "loss": 0.9292, "step": 12508 }, { "epoch": 0.53, "learning_rate": 1.4341121160596073e-06, "loss": 0.9226, "step": 12509 }, { "epoch": 0.53, "learning_rate": 1.4329999424094575e-06, "loss": 0.9446, "step": 12510 }, { "epoch": 0.53, "learning_rate": 1.4318881668945139e-06, "loss": 0.9123, "step": 12511 }, { "epoch": 0.53, "learning_rate": 1.430776789566446e-06, "loss": 0.9448, "step": 12512 }, { "epoch": 0.53, "learning_rate": 1.4296658104769034e-06, "loss": 0.9395, "step": 12513 }, { "epoch": 0.53, "learning_rate": 1.428555229677513e-06, "loss": 0.8947, "step": 12514 }, { "epoch": 0.53, "learning_rate": 1.4274450472198886e-06, "loss": 0.937, "step": 12515 }, { "epoch": 0.53, "learning_rate": 1.4263352631556272e-06, "loss": 0.9089, "step": 12516 }, { "epoch": 0.53, "learning_rate": 1.4252258775362992e-06, "loss": 0.9357, "step": 12517 }, { "epoch": 0.53, "learning_rate": 1.4241168904134627e-06, "loss": 0.9511, "step": 12518 }, { "epoch": 0.53, "learning_rate": 1.4230083018386565e-06, "loss": 0.9323, "step": 12519 }, { "epoch": 0.53, "learning_rate": 1.4219001118634025e-06, "loss": 0.895, "step": 12520 }, { "epoch": 0.53, "learning_rate": 1.4207923205391938e-06, "loss": 0.9139, "step": 12521 }, { "epoch": 0.53, "learning_rate": 1.4196849279175228e-06, "loss": 0.9352, "step": 12522 }, { "epoch": 0.53, "learning_rate": 1.4185779340498473e-06, "loss": 0.8881, "step": 12523 }, { "epoch": 0.53, "learning_rate": 1.4174713389876137e-06, "loss": 0.9544, "step": 12524 }, { "epoch": 0.53, "learning_rate": 1.4163651427822522e-06, "loss": 0.9326, "step": 12525 }, { "epoch": 0.53, "learning_rate": 1.4152593454851672e-06, "loss": 0.961, "step": 12526 }, { "epoch": 0.53, "learning_rate": 1.4141539471477482e-06, "loss": 0.9456, "step": 12527 }, { "epoch": 0.53, "learning_rate": 1.4130489478213694e-06, "loss": 0.9542, "step": 12528 }, { "epoch": 0.53, "learning_rate": 1.4119443475573845e-06, "loss": 0.8738, "step": 12529 }, { "epoch": 0.53, "learning_rate": 1.410840146407123e-06, "loss": 0.9485, "step": 12530 }, { "epoch": 0.53, "learning_rate": 1.4097363444219037e-06, "loss": 0.9336, "step": 12531 }, { "epoch": 0.53, "learning_rate": 1.4086329416530252e-06, "loss": 0.9246, "step": 12532 }, { "epoch": 0.53, "learning_rate": 1.4075299381517604e-06, "loss": 0.9399, "step": 12533 }, { "epoch": 0.53, "learning_rate": 1.4064273339693768e-06, "loss": 0.9752, "step": 12534 }, { "epoch": 0.53, "learning_rate": 1.4053251291571101e-06, "loss": 0.9616, "step": 12535 }, { "epoch": 0.53, "learning_rate": 1.4042233237661852e-06, "loss": 0.9086, "step": 12536 }, { "epoch": 0.53, "learning_rate": 1.403121917847805e-06, "loss": 0.8984, "step": 12537 }, { "epoch": 0.53, "learning_rate": 1.4020209114531601e-06, "loss": 0.9038, "step": 12538 }, { "epoch": 0.53, "learning_rate": 1.400920304633413e-06, "loss": 0.8908, "step": 12539 }, { "epoch": 0.53, "learning_rate": 1.3998200974397125e-06, "loss": 0.9436, "step": 12540 }, { "epoch": 0.53, "learning_rate": 1.398720289923191e-06, "loss": 0.9284, "step": 12541 }, { "epoch": 0.53, "learning_rate": 1.3976208821349558e-06, "loss": 0.9082, "step": 12542 }, { "epoch": 0.53, "learning_rate": 1.3965218741261067e-06, "loss": 0.9083, "step": 12543 }, { "epoch": 0.53, "learning_rate": 1.3954232659477117e-06, "loss": 0.9197, "step": 12544 }, { "epoch": 0.53, "learning_rate": 1.3943250576508293e-06, "loss": 0.9423, "step": 12545 }, { "epoch": 0.53, "learning_rate": 1.3932272492864984e-06, "loss": 0.9035, "step": 12546 }, { "epoch": 0.53, "learning_rate": 1.3921298409057338e-06, "loss": 0.9392, "step": 12547 }, { "epoch": 0.53, "learning_rate": 1.3910328325595358e-06, "loss": 0.9252, "step": 12548 }, { "epoch": 0.53, "learning_rate": 1.3899362242988878e-06, "loss": 0.9186, "step": 12549 }, { "epoch": 0.54, "learning_rate": 1.388840016174755e-06, "loss": 0.9113, "step": 12550 }, { "epoch": 0.54, "learning_rate": 1.3877442082380755e-06, "loss": 0.8871, "step": 12551 }, { "epoch": 0.54, "learning_rate": 1.3866488005397783e-06, "loss": 0.9176, "step": 12552 }, { "epoch": 0.54, "learning_rate": 1.3855537931307716e-06, "loss": 0.9933, "step": 12553 }, { "epoch": 0.54, "learning_rate": 1.3844591860619382e-06, "loss": 0.9499, "step": 12554 }, { "epoch": 0.54, "learning_rate": 1.3833649793841564e-06, "loss": 0.9591, "step": 12555 }, { "epoch": 0.54, "learning_rate": 1.3822711731482719e-06, "loss": 0.913, "step": 12556 }, { "epoch": 0.54, "learning_rate": 1.3811777674051174e-06, "loss": 0.9128, "step": 12557 }, { "epoch": 0.54, "learning_rate": 1.3800847622055081e-06, "loss": 0.9248, "step": 12558 }, { "epoch": 0.54, "learning_rate": 1.3789921576002407e-06, "loss": 0.9445, "step": 12559 }, { "epoch": 0.54, "learning_rate": 1.3778999536400883e-06, "loss": 0.9147, "step": 12560 }, { "epoch": 0.54, "learning_rate": 1.3768081503758114e-06, "loss": 0.9196, "step": 12561 }, { "epoch": 0.54, "learning_rate": 1.3757167478581512e-06, "loss": 0.8955, "step": 12562 }, { "epoch": 0.54, "learning_rate": 1.3746257461378253e-06, "loss": 0.9314, "step": 12563 }, { "epoch": 0.54, "learning_rate": 1.3735351452655365e-06, "loss": 0.9391, "step": 12564 }, { "epoch": 0.54, "learning_rate": 1.3724449452919685e-06, "loss": 0.9305, "step": 12565 }, { "epoch": 0.54, "learning_rate": 1.3713551462677876e-06, "loss": 0.9359, "step": 12566 }, { "epoch": 0.54, "learning_rate": 1.3702657482436398e-06, "loss": 0.923, "step": 12567 }, { "epoch": 0.54, "learning_rate": 1.3691767512701548e-06, "loss": 0.9318, "step": 12568 }, { "epoch": 0.54, "learning_rate": 1.3680881553979363e-06, "loss": 0.9367, "step": 12569 }, { "epoch": 0.54, "learning_rate": 1.3669999606775786e-06, "loss": 0.957, "step": 12570 }, { "epoch": 0.54, "learning_rate": 1.3659121671596542e-06, "loss": 0.9311, "step": 12571 }, { "epoch": 0.54, "learning_rate": 1.364824774894714e-06, "loss": 0.8952, "step": 12572 }, { "epoch": 0.54, "learning_rate": 1.3637377839332921e-06, "loss": 0.9335, "step": 12573 }, { "epoch": 0.54, "learning_rate": 1.3626511943259058e-06, "loss": 0.8972, "step": 12574 }, { "epoch": 0.54, "learning_rate": 1.3615650061230545e-06, "loss": 0.9057, "step": 12575 }, { "epoch": 0.54, "learning_rate": 1.3604792193752126e-06, "loss": 0.9198, "step": 12576 }, { "epoch": 0.54, "learning_rate": 1.3593938341328405e-06, "loss": 0.914, "step": 12577 }, { "epoch": 0.54, "learning_rate": 1.3583088504463837e-06, "loss": 0.9312, "step": 12578 }, { "epoch": 0.54, "learning_rate": 1.3572242683662573e-06, "loss": 0.9291, "step": 12579 }, { "epoch": 0.54, "learning_rate": 1.3561400879428744e-06, "loss": 0.9356, "step": 12580 }, { "epoch": 0.54, "learning_rate": 1.3550563092266122e-06, "loss": 0.9427, "step": 12581 }, { "epoch": 0.54, "learning_rate": 1.3539729322678409e-06, "loss": 0.8988, "step": 12582 }, { "epoch": 0.54, "learning_rate": 1.3528899571169084e-06, "loss": 0.9362, "step": 12583 }, { "epoch": 0.54, "learning_rate": 1.3518073838241462e-06, "loss": 0.91, "step": 12584 }, { "epoch": 0.54, "learning_rate": 1.3507252124398596e-06, "loss": 0.9144, "step": 12585 }, { "epoch": 0.54, "learning_rate": 1.3496434430143436e-06, "loss": 0.975, "step": 12586 }, { "epoch": 0.54, "learning_rate": 1.3485620755978723e-06, "loss": 0.912, "step": 12587 }, { "epoch": 0.54, "learning_rate": 1.3474811102406949e-06, "loss": 0.9357, "step": 12588 }, { "epoch": 0.54, "learning_rate": 1.3464005469930564e-06, "loss": 0.9153, "step": 12589 }, { "epoch": 0.54, "learning_rate": 1.3453203859051657e-06, "loss": 0.9563, "step": 12590 }, { "epoch": 0.54, "learning_rate": 1.3442406270272235e-06, "loss": 0.9292, "step": 12591 }, { "epoch": 0.54, "learning_rate": 1.3431612704094145e-06, "loss": 0.9193, "step": 12592 }, { "epoch": 0.54, "learning_rate": 1.3420823161018913e-06, "loss": 0.8867, "step": 12593 }, { "epoch": 0.54, "learning_rate": 1.3410037641548023e-06, "loss": 0.9005, "step": 12594 }, { "epoch": 0.54, "learning_rate": 1.3399256146182682e-06, "loss": 0.8951, "step": 12595 }, { "epoch": 0.54, "learning_rate": 1.338847867542399e-06, "loss": 0.9018, "step": 12596 }, { "epoch": 0.54, "learning_rate": 1.337770522977272e-06, "loss": 0.9188, "step": 12597 }, { "epoch": 0.54, "learning_rate": 1.3366935809729641e-06, "loss": 0.9504, "step": 12598 }, { "epoch": 0.54, "learning_rate": 1.3356170415795178e-06, "loss": 0.9293, "step": 12599 }, { "epoch": 0.54, "learning_rate": 1.3345409048469648e-06, "loss": 0.9316, "step": 12600 }, { "epoch": 0.54, "learning_rate": 1.333465170825321e-06, "loss": 0.9133, "step": 12601 }, { "epoch": 0.54, "learning_rate": 1.3323898395645718e-06, "loss": 0.9311, "step": 12602 }, { "epoch": 0.54, "learning_rate": 1.331314911114695e-06, "loss": 0.891, "step": 12603 }, { "epoch": 0.54, "learning_rate": 1.3302403855256463e-06, "loss": 0.9299, "step": 12604 }, { "epoch": 0.54, "learning_rate": 1.3291662628473634e-06, "loss": 0.9201, "step": 12605 }, { "epoch": 0.54, "learning_rate": 1.3280925431297597e-06, "loss": 0.9468, "step": 12606 }, { "epoch": 0.54, "learning_rate": 1.3270192264227366e-06, "loss": 0.9242, "step": 12607 }, { "epoch": 0.54, "learning_rate": 1.3259463127761784e-06, "loss": 0.9252, "step": 12608 }, { "epoch": 0.54, "learning_rate": 1.32487380223994e-06, "loss": 0.8941, "step": 12609 }, { "epoch": 0.54, "learning_rate": 1.3238016948638666e-06, "loss": 0.8966, "step": 12610 }, { "epoch": 0.54, "learning_rate": 1.3227299906977843e-06, "loss": 0.9309, "step": 12611 }, { "epoch": 0.54, "learning_rate": 1.3216586897914963e-06, "loss": 0.9407, "step": 12612 }, { "epoch": 0.54, "learning_rate": 1.3205877921947918e-06, "loss": 0.8883, "step": 12613 }, { "epoch": 0.54, "learning_rate": 1.3195172979574388e-06, "loss": 0.8861, "step": 12614 }, { "epoch": 0.54, "learning_rate": 1.318447207129182e-06, "loss": 0.9736, "step": 12615 }, { "epoch": 0.54, "learning_rate": 1.3173775197597561e-06, "loss": 0.8932, "step": 12616 }, { "epoch": 0.54, "learning_rate": 1.3163082358988732e-06, "loss": 0.9271, "step": 12617 }, { "epoch": 0.54, "learning_rate": 1.3152393555962227e-06, "loss": 0.8453, "step": 12618 }, { "epoch": 0.54, "learning_rate": 1.3141708789014818e-06, "loss": 0.9339, "step": 12619 }, { "epoch": 0.54, "learning_rate": 1.3131028058643036e-06, "loss": 0.9322, "step": 12620 }, { "epoch": 0.54, "learning_rate": 1.3120351365343286e-06, "loss": 0.9199, "step": 12621 }, { "epoch": 0.54, "learning_rate": 1.3109678709611672e-06, "loss": 0.916, "step": 12622 }, { "epoch": 0.54, "learning_rate": 1.3099010091944287e-06, "loss": 0.9322, "step": 12623 }, { "epoch": 0.54, "learning_rate": 1.3088345512836886e-06, "loss": 0.9457, "step": 12624 }, { "epoch": 0.54, "learning_rate": 1.3077684972785044e-06, "loss": 0.9027, "step": 12625 }, { "epoch": 0.54, "learning_rate": 1.306702847228426e-06, "loss": 0.9342, "step": 12626 }, { "epoch": 0.54, "learning_rate": 1.305637601182973e-06, "loss": 0.8904, "step": 12627 }, { "epoch": 0.54, "learning_rate": 1.3045727591916513e-06, "loss": 0.9156, "step": 12628 }, { "epoch": 0.54, "learning_rate": 1.303508321303949e-06, "loss": 0.9543, "step": 12629 }, { "epoch": 0.54, "learning_rate": 1.3024442875693343e-06, "loss": 0.8959, "step": 12630 }, { "epoch": 0.54, "learning_rate": 1.3013806580372534e-06, "loss": 0.9117, "step": 12631 }, { "epoch": 0.54, "learning_rate": 1.3003174327571378e-06, "loss": 0.9229, "step": 12632 }, { "epoch": 0.54, "learning_rate": 1.299254611778401e-06, "loss": 0.9127, "step": 12633 }, { "epoch": 0.54, "learning_rate": 1.2981921951504295e-06, "loss": 0.8911, "step": 12634 }, { "epoch": 0.54, "learning_rate": 1.2971301829226057e-06, "loss": 0.9399, "step": 12635 }, { "epoch": 0.54, "learning_rate": 1.296068575144277e-06, "loss": 0.9144, "step": 12636 }, { "epoch": 0.54, "learning_rate": 1.2950073718647838e-06, "loss": 0.9595, "step": 12637 }, { "epoch": 0.54, "learning_rate": 1.2939465731334444e-06, "loss": 0.9266, "step": 12638 }, { "epoch": 0.54, "learning_rate": 1.292886178999554e-06, "loss": 0.9073, "step": 12639 }, { "epoch": 0.54, "learning_rate": 1.291826189512394e-06, "loss": 0.9183, "step": 12640 }, { "epoch": 0.54, "learning_rate": 1.2907666047212252e-06, "loss": 0.8985, "step": 12641 }, { "epoch": 0.54, "learning_rate": 1.2897074246752917e-06, "loss": 0.8824, "step": 12642 }, { "epoch": 0.54, "learning_rate": 1.2886486494238115e-06, "loss": 0.952, "step": 12643 }, { "epoch": 0.54, "learning_rate": 1.2875902790159966e-06, "loss": 0.9449, "step": 12644 }, { "epoch": 0.54, "learning_rate": 1.2865323135010277e-06, "loss": 0.9203, "step": 12645 }, { "epoch": 0.54, "learning_rate": 1.285474752928072e-06, "loss": 0.8971, "step": 12646 }, { "epoch": 0.54, "learning_rate": 1.2844175973462814e-06, "loss": 0.9262, "step": 12647 }, { "epoch": 0.54, "learning_rate": 1.283360846804781e-06, "loss": 0.9114, "step": 12648 }, { "epoch": 0.54, "learning_rate": 1.2823045013526824e-06, "loss": 0.9202, "step": 12649 }, { "epoch": 0.54, "learning_rate": 1.2812485610390778e-06, "loss": 0.914, "step": 12650 }, { "epoch": 0.54, "learning_rate": 1.2801930259130425e-06, "loss": 0.9073, "step": 12651 }, { "epoch": 0.54, "learning_rate": 1.2791378960236257e-06, "loss": 0.9564, "step": 12652 }, { "epoch": 0.54, "learning_rate": 1.2780831714198648e-06, "loss": 0.895, "step": 12653 }, { "epoch": 0.54, "learning_rate": 1.2770288521507779e-06, "loss": 0.986, "step": 12654 }, { "epoch": 0.54, "learning_rate": 1.2759749382653575e-06, "loss": 0.9223, "step": 12655 }, { "epoch": 0.54, "learning_rate": 1.274921429812589e-06, "loss": 0.9053, "step": 12656 }, { "epoch": 0.54, "learning_rate": 1.2738683268414276e-06, "loss": 0.9205, "step": 12657 }, { "epoch": 0.54, "learning_rate": 1.2728156294008143e-06, "loss": 0.9162, "step": 12658 }, { "epoch": 0.54, "learning_rate": 1.2717633375396732e-06, "loss": 0.926, "step": 12659 }, { "epoch": 0.54, "learning_rate": 1.2707114513069085e-06, "loss": 0.9442, "step": 12660 }, { "epoch": 0.54, "learning_rate": 1.2696599707514012e-06, "loss": 0.9266, "step": 12661 }, { "epoch": 0.54, "learning_rate": 1.2686088959220189e-06, "loss": 0.9176, "step": 12662 }, { "epoch": 0.54, "learning_rate": 1.2675582268676089e-06, "loss": 0.9067, "step": 12663 }, { "epoch": 0.54, "learning_rate": 1.266507963636997e-06, "loss": 0.9686, "step": 12664 }, { "epoch": 0.54, "learning_rate": 1.2654581062789939e-06, "loss": 0.8865, "step": 12665 }, { "epoch": 0.54, "learning_rate": 1.2644086548423884e-06, "loss": 0.8964, "step": 12666 }, { "epoch": 0.54, "learning_rate": 1.263359609375955e-06, "loss": 0.8814, "step": 12667 }, { "epoch": 0.54, "learning_rate": 1.2623109699284396e-06, "loss": 0.9378, "step": 12668 }, { "epoch": 0.54, "learning_rate": 1.261262736548584e-06, "loss": 0.9389, "step": 12669 }, { "epoch": 0.54, "learning_rate": 1.2602149092850969e-06, "loss": 0.9271, "step": 12670 }, { "epoch": 0.54, "learning_rate": 1.2591674881866756e-06, "loss": 0.9251, "step": 12671 }, { "epoch": 0.54, "learning_rate": 1.2581204733019991e-06, "loss": 0.9606, "step": 12672 }, { "epoch": 0.54, "learning_rate": 1.2570738646797232e-06, "loss": 0.9662, "step": 12673 }, { "epoch": 0.54, "learning_rate": 1.2560276623684863e-06, "loss": 0.9299, "step": 12674 }, { "epoch": 0.54, "learning_rate": 1.25498186641691e-06, "loss": 0.9348, "step": 12675 }, { "epoch": 0.54, "learning_rate": 1.2539364768735984e-06, "loss": 0.9087, "step": 12676 }, { "epoch": 0.54, "learning_rate": 1.2528914937871296e-06, "loss": 0.9178, "step": 12677 }, { "epoch": 0.54, "learning_rate": 1.2518469172060677e-06, "loss": 0.9071, "step": 12678 }, { "epoch": 0.54, "learning_rate": 1.2508027471789608e-06, "loss": 0.9411, "step": 12679 }, { "epoch": 0.54, "learning_rate": 1.2497589837543288e-06, "loss": 0.9658, "step": 12680 }, { "epoch": 0.54, "learning_rate": 1.2487156269806877e-06, "loss": 0.941, "step": 12681 }, { "epoch": 0.54, "learning_rate": 1.2476726769065172e-06, "loss": 0.9235, "step": 12682 }, { "epoch": 0.54, "learning_rate": 1.246630133580289e-06, "loss": 0.9499, "step": 12683 }, { "epoch": 0.54, "learning_rate": 1.2455879970504569e-06, "loss": 0.9238, "step": 12684 }, { "epoch": 0.54, "learning_rate": 1.2445462673654464e-06, "loss": 0.9188, "step": 12685 }, { "epoch": 0.54, "learning_rate": 1.2435049445736736e-06, "loss": 0.9368, "step": 12686 }, { "epoch": 0.54, "learning_rate": 1.2424640287235302e-06, "loss": 0.9846, "step": 12687 }, { "epoch": 0.54, "learning_rate": 1.241423519863395e-06, "loss": 0.9274, "step": 12688 }, { "epoch": 0.54, "learning_rate": 1.2403834180416152e-06, "loss": 0.9303, "step": 12689 }, { "epoch": 0.54, "learning_rate": 1.2393437233065374e-06, "loss": 0.9224, "step": 12690 }, { "epoch": 0.54, "learning_rate": 1.2383044357064722e-06, "loss": 0.9216, "step": 12691 }, { "epoch": 0.54, "learning_rate": 1.2372655552897218e-06, "loss": 0.9175, "step": 12692 }, { "epoch": 0.54, "learning_rate": 1.2362270821045674e-06, "loss": 0.9039, "step": 12693 }, { "epoch": 0.54, "learning_rate": 1.2351890161992664e-06, "loss": 0.92, "step": 12694 }, { "epoch": 0.54, "learning_rate": 1.2341513576220631e-06, "loss": 0.9105, "step": 12695 }, { "epoch": 0.54, "learning_rate": 1.2331141064211804e-06, "loss": 0.9234, "step": 12696 }, { "epoch": 0.54, "learning_rate": 1.232077262644824e-06, "loss": 0.8901, "step": 12697 }, { "epoch": 0.54, "learning_rate": 1.2310408263411766e-06, "loss": 0.9101, "step": 12698 }, { "epoch": 0.54, "learning_rate": 1.2300047975584063e-06, "loss": 0.9598, "step": 12699 }, { "epoch": 0.54, "learning_rate": 1.2289691763446622e-06, "loss": 0.8904, "step": 12700 }, { "epoch": 0.55, "learning_rate": 1.2279339627480657e-06, "loss": 0.8972, "step": 12701 }, { "epoch": 0.55, "learning_rate": 1.2268991568167364e-06, "loss": 0.9015, "step": 12702 }, { "epoch": 0.55, "learning_rate": 1.2258647585987594e-06, "loss": 0.9353, "step": 12703 }, { "epoch": 0.55, "learning_rate": 1.2248307681422055e-06, "loss": 0.9473, "step": 12704 }, { "epoch": 0.55, "learning_rate": 1.2237971854951302e-06, "loss": 0.9555, "step": 12705 }, { "epoch": 0.55, "learning_rate": 1.222764010705567e-06, "loss": 0.9276, "step": 12706 }, { "epoch": 0.55, "learning_rate": 1.221731243821529e-06, "loss": 0.9392, "step": 12707 }, { "epoch": 0.55, "learning_rate": 1.2206988848910128e-06, "loss": 0.9165, "step": 12708 }, { "epoch": 0.55, "learning_rate": 1.2196669339619981e-06, "loss": 0.9033, "step": 12709 }, { "epoch": 0.55, "learning_rate": 1.2186353910824377e-06, "loss": 0.9136, "step": 12710 }, { "epoch": 0.55, "learning_rate": 1.2176042563002744e-06, "loss": 0.9279, "step": 12711 }, { "epoch": 0.55, "learning_rate": 1.2165735296634262e-06, "loss": 0.9102, "step": 12712 }, { "epoch": 0.55, "learning_rate": 1.2155432112197974e-06, "loss": 0.9231, "step": 12713 }, { "epoch": 0.55, "learning_rate": 1.2145133010172639e-06, "loss": 0.9283, "step": 12714 }, { "epoch": 0.55, "learning_rate": 1.2134837991036975e-06, "loss": 0.9259, "step": 12715 }, { "epoch": 0.55, "learning_rate": 1.2124547055269343e-06, "loss": 0.9113, "step": 12716 }, { "epoch": 0.55, "learning_rate": 1.2114260203348039e-06, "loss": 0.9212, "step": 12717 }, { "epoch": 0.55, "learning_rate": 1.2103977435751136e-06, "loss": 0.9104, "step": 12718 }, { "epoch": 0.55, "learning_rate": 1.2093698752956462e-06, "loss": 0.9563, "step": 12719 }, { "epoch": 0.55, "learning_rate": 1.2083424155441714e-06, "loss": 0.8931, "step": 12720 }, { "epoch": 0.55, "learning_rate": 1.207315364368441e-06, "loss": 0.9217, "step": 12721 }, { "epoch": 0.55, "learning_rate": 1.2062887218161845e-06, "loss": 0.8921, "step": 12722 }, { "epoch": 0.55, "learning_rate": 1.2052624879351105e-06, "loss": 0.9557, "step": 12723 }, { "epoch": 0.55, "learning_rate": 1.204236662772913e-06, "loss": 0.9327, "step": 12724 }, { "epoch": 0.55, "learning_rate": 1.203211246377267e-06, "loss": 0.9521, "step": 12725 }, { "epoch": 0.55, "learning_rate": 1.2021862387958217e-06, "loss": 0.9073, "step": 12726 }, { "epoch": 0.55, "learning_rate": 1.2011616400762194e-06, "loss": 0.887, "step": 12727 }, { "epoch": 0.55, "learning_rate": 1.2001374502660711e-06, "loss": 0.9143, "step": 12728 }, { "epoch": 0.55, "learning_rate": 1.1991136694129756e-06, "loss": 0.9016, "step": 12729 }, { "epoch": 0.55, "learning_rate": 1.1980902975645137e-06, "loss": 0.908, "step": 12730 }, { "epoch": 0.55, "learning_rate": 1.1970673347682405e-06, "loss": 0.9215, "step": 12731 }, { "epoch": 0.55, "learning_rate": 1.196044781071698e-06, "loss": 0.9114, "step": 12732 }, { "epoch": 0.55, "learning_rate": 1.195022636522407e-06, "loss": 0.8882, "step": 12733 }, { "epoch": 0.55, "learning_rate": 1.1940009011678734e-06, "loss": 0.9353, "step": 12734 }, { "epoch": 0.55, "learning_rate": 1.1929795750555727e-06, "loss": 0.9372, "step": 12735 }, { "epoch": 0.55, "learning_rate": 1.1919586582329789e-06, "loss": 0.946, "step": 12736 }, { "epoch": 0.55, "learning_rate": 1.1909381507475292e-06, "loss": 0.9485, "step": 12737 }, { "epoch": 0.55, "learning_rate": 1.1899180526466526e-06, "loss": 0.9283, "step": 12738 }, { "epoch": 0.55, "learning_rate": 1.1888983639777586e-06, "loss": 0.9488, "step": 12739 }, { "epoch": 0.55, "learning_rate": 1.1878790847882305e-06, "loss": 0.9472, "step": 12740 }, { "epoch": 0.55, "learning_rate": 1.1868602151254405e-06, "loss": 0.9619, "step": 12741 }, { "epoch": 0.55, "learning_rate": 1.1858417550367374e-06, "loss": 0.9336, "step": 12742 }, { "epoch": 0.55, "learning_rate": 1.1848237045694555e-06, "loss": 0.9679, "step": 12743 }, { "epoch": 0.55, "learning_rate": 1.1838060637709025e-06, "loss": 0.9147, "step": 12744 }, { "epoch": 0.55, "learning_rate": 1.1827888326883719e-06, "loss": 0.942, "step": 12745 }, { "epoch": 0.55, "learning_rate": 1.1817720113691421e-06, "loss": 0.9279, "step": 12746 }, { "epoch": 0.55, "learning_rate": 1.1807555998604592e-06, "loss": 0.9117, "step": 12747 }, { "epoch": 0.55, "learning_rate": 1.1797395982095695e-06, "loss": 0.9632, "step": 12748 }, { "epoch": 0.55, "learning_rate": 1.1787240064636818e-06, "loss": 0.9283, "step": 12749 }, { "epoch": 0.55, "learning_rate": 1.1777088246699963e-06, "loss": 0.954, "step": 12750 }, { "epoch": 0.55, "learning_rate": 1.176694052875692e-06, "loss": 0.9264, "step": 12751 }, { "epoch": 0.55, "learning_rate": 1.1756796911279322e-06, "loss": 0.9028, "step": 12752 }, { "epoch": 0.55, "learning_rate": 1.1746657394738504e-06, "loss": 0.9054, "step": 12753 }, { "epoch": 0.55, "learning_rate": 1.173652197960571e-06, "loss": 0.8843, "step": 12754 }, { "epoch": 0.55, "learning_rate": 1.1726390666351994e-06, "loss": 0.9342, "step": 12755 }, { "epoch": 0.55, "learning_rate": 1.171626345544815e-06, "loss": 0.9082, "step": 12756 }, { "epoch": 0.55, "learning_rate": 1.1706140347364837e-06, "loss": 0.9171, "step": 12757 }, { "epoch": 0.55, "learning_rate": 1.1696021342572495e-06, "loss": 0.9289, "step": 12758 }, { "epoch": 0.55, "learning_rate": 1.1685906441541406e-06, "loss": 0.9379, "step": 12759 }, { "epoch": 0.55, "learning_rate": 1.1675795644741627e-06, "loss": 0.9215, "step": 12760 }, { "epoch": 0.55, "learning_rate": 1.1665688952643062e-06, "loss": 0.9083, "step": 12761 }, { "epoch": 0.55, "learning_rate": 1.1655586365715365e-06, "loss": 0.9013, "step": 12762 }, { "epoch": 0.55, "learning_rate": 1.1645487884428053e-06, "loss": 0.9376, "step": 12763 }, { "epoch": 0.55, "learning_rate": 1.1635393509250449e-06, "loss": 0.9265, "step": 12764 }, { "epoch": 0.55, "learning_rate": 1.1625303240651641e-06, "loss": 0.9317, "step": 12765 }, { "epoch": 0.55, "learning_rate": 1.161521707910056e-06, "loss": 0.9287, "step": 12766 }, { "epoch": 0.55, "learning_rate": 1.160513502506595e-06, "loss": 0.9426, "step": 12767 }, { "epoch": 0.55, "learning_rate": 1.159505707901637e-06, "loss": 0.9451, "step": 12768 }, { "epoch": 0.55, "learning_rate": 1.1584983241420145e-06, "loss": 0.9467, "step": 12769 }, { "epoch": 0.55, "learning_rate": 1.1574913512745446e-06, "loss": 0.9279, "step": 12770 }, { "epoch": 0.55, "learning_rate": 1.1564847893460274e-06, "loss": 0.9391, "step": 12771 }, { "epoch": 0.55, "learning_rate": 1.1554786384032334e-06, "loss": 0.945, "step": 12772 }, { "epoch": 0.55, "learning_rate": 1.1544728984929322e-06, "loss": 0.9541, "step": 12773 }, { "epoch": 0.55, "learning_rate": 1.1534675696618557e-06, "loss": 0.9129, "step": 12774 }, { "epoch": 0.55, "learning_rate": 1.1524626519567262e-06, "loss": 0.912, "step": 12775 }, { "epoch": 0.55, "learning_rate": 1.151458145424249e-06, "loss": 0.9127, "step": 12776 }, { "epoch": 0.55, "learning_rate": 1.1504540501111017e-06, "loss": 0.8725, "step": 12777 }, { "epoch": 0.55, "learning_rate": 1.1494503660639511e-06, "loss": 0.9101, "step": 12778 }, { "epoch": 0.55, "learning_rate": 1.148447093329439e-06, "loss": 0.9016, "step": 12779 }, { "epoch": 0.55, "learning_rate": 1.1474442319541945e-06, "loss": 0.9773, "step": 12780 }, { "epoch": 0.55, "learning_rate": 1.1464417819848172e-06, "loss": 0.8929, "step": 12781 }, { "epoch": 0.55, "learning_rate": 1.1454397434679022e-06, "loss": 0.9415, "step": 12782 }, { "epoch": 0.55, "learning_rate": 1.1444381164500106e-06, "loss": 0.9304, "step": 12783 }, { "epoch": 0.55, "learning_rate": 1.1434369009776947e-06, "loss": 0.9175, "step": 12784 }, { "epoch": 0.55, "learning_rate": 1.1424360970974846e-06, "loss": 0.9054, "step": 12785 }, { "epoch": 0.55, "learning_rate": 1.1414357048558867e-06, "loss": 0.9043, "step": 12786 }, { "epoch": 0.55, "learning_rate": 1.1404357242993958e-06, "loss": 0.8978, "step": 12787 }, { "epoch": 0.55, "learning_rate": 1.1394361554744837e-06, "loss": 0.9399, "step": 12788 }, { "epoch": 0.55, "learning_rate": 1.1384369984276044e-06, "loss": 0.9102, "step": 12789 }, { "epoch": 0.55, "learning_rate": 1.1374382532051864e-06, "loss": 0.945, "step": 12790 }, { "epoch": 0.55, "learning_rate": 1.1364399198536535e-06, "loss": 0.9613, "step": 12791 }, { "epoch": 0.55, "learning_rate": 1.1354419984193955e-06, "loss": 0.9081, "step": 12792 }, { "epoch": 0.55, "learning_rate": 1.134444488948786e-06, "loss": 0.9601, "step": 12793 }, { "epoch": 0.55, "learning_rate": 1.1334473914881905e-06, "loss": 0.9747, "step": 12794 }, { "epoch": 0.55, "learning_rate": 1.1324507060839419e-06, "loss": 0.9062, "step": 12795 }, { "epoch": 0.55, "learning_rate": 1.1314544327823596e-06, "loss": 0.917, "step": 12796 }, { "epoch": 0.55, "learning_rate": 1.1304585716297444e-06, "loss": 0.9083, "step": 12797 }, { "epoch": 0.55, "learning_rate": 1.1294631226723784e-06, "loss": 0.8904, "step": 12798 }, { "epoch": 0.55, "learning_rate": 1.128468085956521e-06, "loss": 0.9177, "step": 12799 }, { "epoch": 0.55, "learning_rate": 1.1274734615284145e-06, "loss": 0.9241, "step": 12800 }, { "epoch": 0.55, "learning_rate": 1.1264792494342858e-06, "loss": 0.9016, "step": 12801 }, { "epoch": 0.55, "learning_rate": 1.1254854497203339e-06, "loss": 0.9073, "step": 12802 }, { "epoch": 0.55, "learning_rate": 1.124492062432747e-06, "loss": 0.9132, "step": 12803 }, { "epoch": 0.55, "learning_rate": 1.12349908761769e-06, "loss": 0.9678, "step": 12804 }, { "epoch": 0.55, "learning_rate": 1.1225065253213096e-06, "loss": 0.9153, "step": 12805 }, { "epoch": 0.55, "learning_rate": 1.1215143755897328e-06, "loss": 0.953, "step": 12806 }, { "epoch": 0.55, "learning_rate": 1.1205226384690704e-06, "loss": 0.9096, "step": 12807 }, { "epoch": 0.55, "learning_rate": 1.1195313140054086e-06, "loss": 0.9091, "step": 12808 }, { "epoch": 0.55, "learning_rate": 1.1185404022448177e-06, "loss": 0.9347, "step": 12809 }, { "epoch": 0.55, "learning_rate": 1.1175499032333514e-06, "loss": 0.9185, "step": 12810 }, { "epoch": 0.55, "learning_rate": 1.1165598170170367e-06, "loss": 0.9246, "step": 12811 }, { "epoch": 0.55, "learning_rate": 1.1155701436418887e-06, "loss": 0.9106, "step": 12812 }, { "epoch": 0.55, "learning_rate": 1.1145808831538996e-06, "loss": 0.9192, "step": 12813 }, { "epoch": 0.55, "learning_rate": 1.1135920355990471e-06, "loss": 0.8942, "step": 12814 }, { "epoch": 0.55, "learning_rate": 1.1126036010232798e-06, "loss": 0.9063, "step": 12815 }, { "epoch": 0.55, "learning_rate": 1.111615579472537e-06, "loss": 0.9226, "step": 12816 }, { "epoch": 0.55, "learning_rate": 1.110627970992736e-06, "loss": 0.9375, "step": 12817 }, { "epoch": 0.55, "learning_rate": 1.109640775629769e-06, "loss": 0.9264, "step": 12818 }, { "epoch": 0.55, "learning_rate": 1.1086539934295215e-06, "loss": 0.9501, "step": 12819 }, { "epoch": 0.55, "learning_rate": 1.107667624437847e-06, "loss": 0.971, "step": 12820 }, { "epoch": 0.55, "learning_rate": 1.1066816687005866e-06, "loss": 0.8665, "step": 12821 }, { "epoch": 0.55, "learning_rate": 1.1056961262635623e-06, "loss": 0.9779, "step": 12822 }, { "epoch": 0.55, "learning_rate": 1.1047109971725723e-06, "loss": 0.9351, "step": 12823 }, { "epoch": 0.55, "learning_rate": 1.1037262814733996e-06, "loss": 0.9542, "step": 12824 }, { "epoch": 0.55, "learning_rate": 1.1027419792118088e-06, "loss": 0.9496, "step": 12825 }, { "epoch": 0.55, "learning_rate": 1.1017580904335434e-06, "loss": 0.934, "step": 12826 }, { "epoch": 0.55, "learning_rate": 1.1007746151843225e-06, "loss": 0.9248, "step": 12827 }, { "epoch": 0.55, "learning_rate": 1.0997915535098603e-06, "loss": 0.9618, "step": 12828 }, { "epoch": 0.55, "learning_rate": 1.0988089054558348e-06, "loss": 0.9387, "step": 12829 }, { "epoch": 0.55, "learning_rate": 1.0978266710679163e-06, "loss": 0.8949, "step": 12830 }, { "epoch": 0.55, "learning_rate": 1.0968448503917539e-06, "loss": 0.9269, "step": 12831 }, { "epoch": 0.55, "learning_rate": 1.095863443472972e-06, "loss": 0.9115, "step": 12832 }, { "epoch": 0.55, "learning_rate": 1.0948824503571808e-06, "loss": 0.9627, "step": 12833 }, { "epoch": 0.55, "learning_rate": 1.0939018710899707e-06, "loss": 0.8953, "step": 12834 }, { "epoch": 0.55, "learning_rate": 1.0929217057169139e-06, "loss": 0.9735, "step": 12835 }, { "epoch": 0.55, "learning_rate": 1.0919419542835574e-06, "loss": 0.9482, "step": 12836 }, { "epoch": 0.55, "learning_rate": 1.0909626168354393e-06, "loss": 0.8831, "step": 12837 }, { "epoch": 0.55, "learning_rate": 1.0899836934180685e-06, "loss": 0.9056, "step": 12838 }, { "epoch": 0.55, "learning_rate": 1.0890051840769356e-06, "loss": 0.9449, "step": 12839 }, { "epoch": 0.55, "learning_rate": 1.0880270888575239e-06, "loss": 0.9671, "step": 12840 }, { "epoch": 0.55, "learning_rate": 1.0870494078052796e-06, "loss": 0.9677, "step": 12841 }, { "epoch": 0.55, "learning_rate": 1.0860721409656438e-06, "loss": 0.9091, "step": 12842 }, { "epoch": 0.55, "learning_rate": 1.0850952883840304e-06, "loss": 0.9079, "step": 12843 }, { "epoch": 0.55, "learning_rate": 1.084118850105841e-06, "loss": 0.9329, "step": 12844 }, { "epoch": 0.55, "learning_rate": 1.0831428261764488e-06, "loss": 0.9051, "step": 12845 }, { "epoch": 0.55, "learning_rate": 1.0821672166412133e-06, "loss": 0.9443, "step": 12846 }, { "epoch": 0.55, "learning_rate": 1.0811920215454785e-06, "loss": 0.9138, "step": 12847 }, { "epoch": 0.55, "learning_rate": 1.0802172409345578e-06, "loss": 0.9091, "step": 12848 }, { "epoch": 0.55, "learning_rate": 1.0792428748537597e-06, "loss": 0.9553, "step": 12849 }, { "epoch": 0.55, "learning_rate": 1.0782689233483613e-06, "loss": 0.9043, "step": 12850 }, { "epoch": 0.56, "learning_rate": 1.077295386463626e-06, "loss": 0.886, "step": 12851 }, { "epoch": 0.56, "learning_rate": 1.0763222642447968e-06, "loss": 0.8994, "step": 12852 }, { "epoch": 0.56, "learning_rate": 1.0753495567371009e-06, "loss": 0.9318, "step": 12853 }, { "epoch": 0.56, "learning_rate": 1.0743772639857387e-06, "loss": 0.9305, "step": 12854 }, { "epoch": 0.56, "learning_rate": 1.0734053860358983e-06, "loss": 0.9255, "step": 12855 }, { "epoch": 0.56, "learning_rate": 1.0724339229327462e-06, "loss": 0.9414, "step": 12856 }, { "epoch": 0.56, "learning_rate": 1.0714628747214263e-06, "loss": 0.9074, "step": 12857 }, { "epoch": 0.56, "learning_rate": 1.0704922414470687e-06, "loss": 0.9032, "step": 12858 }, { "epoch": 0.56, "learning_rate": 1.0695220231547799e-06, "loss": 0.9307, "step": 12859 }, { "epoch": 0.56, "learning_rate": 1.0685522198896536e-06, "loss": 0.9077, "step": 12860 }, { "epoch": 0.56, "learning_rate": 1.067582831696754e-06, "loss": 0.9351, "step": 12861 }, { "epoch": 0.56, "learning_rate": 1.0666138586211339e-06, "loss": 0.9022, "step": 12862 }, { "epoch": 0.56, "learning_rate": 1.065645300707825e-06, "loss": 0.8883, "step": 12863 }, { "epoch": 0.56, "learning_rate": 1.0646771580018377e-06, "loss": 0.8899, "step": 12864 }, { "epoch": 0.56, "learning_rate": 1.0637094305481688e-06, "loss": 0.9478, "step": 12865 }, { "epoch": 0.56, "learning_rate": 1.0627421183917852e-06, "loss": 0.9071, "step": 12866 }, { "epoch": 0.56, "learning_rate": 1.0617752215776444e-06, "loss": 0.9045, "step": 12867 }, { "epoch": 0.56, "learning_rate": 1.060808740150684e-06, "loss": 0.9261, "step": 12868 }, { "epoch": 0.56, "learning_rate": 1.0598426741558132e-06, "loss": 0.9228, "step": 12869 }, { "epoch": 0.56, "learning_rate": 1.058877023637931e-06, "loss": 0.8825, "step": 12870 }, { "epoch": 0.56, "learning_rate": 1.0579117886419144e-06, "loss": 0.9281, "step": 12871 }, { "epoch": 0.56, "learning_rate": 1.0569469692126233e-06, "loss": 0.9553, "step": 12872 }, { "epoch": 0.56, "learning_rate": 1.0559825653948896e-06, "loss": 0.9158, "step": 12873 }, { "epoch": 0.56, "learning_rate": 1.0550185772335397e-06, "loss": 0.9609, "step": 12874 }, { "epoch": 0.56, "learning_rate": 1.0540550047733667e-06, "loss": 0.9233, "step": 12875 }, { "epoch": 0.56, "learning_rate": 1.0530918480591544e-06, "loss": 0.9212, "step": 12876 }, { "epoch": 0.56, "learning_rate": 1.0521291071356643e-06, "loss": 0.9409, "step": 12877 }, { "epoch": 0.56, "learning_rate": 1.0511667820476346e-06, "loss": 0.8825, "step": 12878 }, { "epoch": 0.56, "learning_rate": 1.0502048728397896e-06, "loss": 0.9249, "step": 12879 }, { "epoch": 0.56, "learning_rate": 1.0492433795568314e-06, "loss": 0.9257, "step": 12880 }, { "epoch": 0.56, "learning_rate": 1.0482823022434463e-06, "loss": 0.8843, "step": 12881 }, { "epoch": 0.56, "learning_rate": 1.0473216409442931e-06, "loss": 0.9272, "step": 12882 }, { "epoch": 0.56, "learning_rate": 1.0463613957040242e-06, "loss": 0.9373, "step": 12883 }, { "epoch": 0.56, "learning_rate": 1.0454015665672613e-06, "loss": 0.9027, "step": 12884 }, { "epoch": 0.56, "learning_rate": 1.0444421535786053e-06, "loss": 0.9085, "step": 12885 }, { "epoch": 0.56, "learning_rate": 1.043483156782653e-06, "loss": 0.8832, "step": 12886 }, { "epoch": 0.56, "learning_rate": 1.0425245762239654e-06, "loss": 0.9284, "step": 12887 }, { "epoch": 0.56, "learning_rate": 1.0415664119470925e-06, "loss": 0.9389, "step": 12888 }, { "epoch": 0.56, "learning_rate": 1.0406086639965629e-06, "loss": 0.9154, "step": 12889 }, { "epoch": 0.56, "learning_rate": 1.039651332416889e-06, "loss": 0.9115, "step": 12890 }, { "epoch": 0.56, "learning_rate": 1.0386944172525559e-06, "loss": 0.9379, "step": 12891 }, { "epoch": 0.56, "learning_rate": 1.0377379185480375e-06, "loss": 0.9438, "step": 12892 }, { "epoch": 0.56, "learning_rate": 1.0367818363477855e-06, "loss": 0.926, "step": 12893 }, { "epoch": 0.56, "learning_rate": 1.035826170696229e-06, "loss": 0.9081, "step": 12894 }, { "epoch": 0.56, "learning_rate": 1.034870921637786e-06, "loss": 0.967, "step": 12895 }, { "epoch": 0.56, "learning_rate": 1.0339160892168453e-06, "loss": 0.9356, "step": 12896 }, { "epoch": 0.56, "learning_rate": 1.0329616734777826e-06, "loss": 0.9199, "step": 12897 }, { "epoch": 0.56, "learning_rate": 1.0320076744649532e-06, "loss": 0.8992, "step": 12898 }, { "epoch": 0.56, "learning_rate": 1.031054092222693e-06, "loss": 0.9273, "step": 12899 }, { "epoch": 0.56, "learning_rate": 1.0301009267953145e-06, "loss": 0.8979, "step": 12900 }, { "epoch": 0.56, "learning_rate": 1.0291481782271174e-06, "loss": 0.9275, "step": 12901 }, { "epoch": 0.56, "learning_rate": 1.0281958465623788e-06, "loss": 0.9341, "step": 12902 }, { "epoch": 0.56, "learning_rate": 1.0272439318453553e-06, "loss": 0.9141, "step": 12903 }, { "epoch": 0.56, "learning_rate": 1.026292434120284e-06, "loss": 0.8965, "step": 12904 }, { "epoch": 0.56, "learning_rate": 1.0253413534313872e-06, "loss": 0.9538, "step": 12905 }, { "epoch": 0.56, "learning_rate": 1.024390689822864e-06, "loss": 0.8937, "step": 12906 }, { "epoch": 0.56, "learning_rate": 1.0234404433388922e-06, "loss": 0.888, "step": 12907 }, { "epoch": 0.56, "learning_rate": 1.0224906140236345e-06, "loss": 0.9193, "step": 12908 }, { "epoch": 0.56, "learning_rate": 1.0215412019212323e-06, "loss": 0.9344, "step": 12909 }, { "epoch": 0.56, "learning_rate": 1.0205922070758078e-06, "loss": 0.8957, "step": 12910 }, { "epoch": 0.56, "learning_rate": 1.019643629531466e-06, "loss": 0.9194, "step": 12911 }, { "epoch": 0.56, "learning_rate": 1.0186954693322847e-06, "loss": 0.9096, "step": 12912 }, { "epoch": 0.56, "learning_rate": 1.0177477265223323e-06, "loss": 0.9128, "step": 12913 }, { "epoch": 0.56, "learning_rate": 1.0168004011456533e-06, "loss": 0.9235, "step": 12914 }, { "epoch": 0.56, "learning_rate": 1.0158534932462704e-06, "loss": 0.9541, "step": 12915 }, { "epoch": 0.56, "learning_rate": 1.0149070028681906e-06, "loss": 0.9269, "step": 12916 }, { "epoch": 0.56, "learning_rate": 1.0139609300553999e-06, "loss": 0.9216, "step": 12917 }, { "epoch": 0.56, "learning_rate": 1.0130152748518674e-06, "loss": 0.9185, "step": 12918 }, { "epoch": 0.56, "learning_rate": 1.0120700373015357e-06, "loss": 0.9227, "step": 12919 }, { "epoch": 0.56, "learning_rate": 1.0111252174483399e-06, "loss": 0.8892, "step": 12920 }, { "epoch": 0.56, "learning_rate": 1.0101808153361826e-06, "loss": 0.8739, "step": 12921 }, { "epoch": 0.56, "learning_rate": 1.0092368310089551e-06, "loss": 0.911, "step": 12922 }, { "epoch": 0.56, "learning_rate": 1.0082932645105303e-06, "loss": 0.9183, "step": 12923 }, { "epoch": 0.56, "learning_rate": 1.0073501158847544e-06, "loss": 0.9401, "step": 12924 }, { "epoch": 0.56, "learning_rate": 1.0064073851754597e-06, "loss": 0.9338, "step": 12925 }, { "epoch": 0.56, "learning_rate": 1.0054650724264592e-06, "loss": 0.9494, "step": 12926 }, { "epoch": 0.56, "learning_rate": 1.0045231776815445e-06, "loss": 0.9148, "step": 12927 }, { "epoch": 0.56, "learning_rate": 1.0035817009844851e-06, "loss": 0.9378, "step": 12928 }, { "epoch": 0.56, "learning_rate": 1.0026406423790413e-06, "loss": 0.9414, "step": 12929 }, { "epoch": 0.56, "learning_rate": 1.001700001908943e-06, "loss": 0.9223, "step": 12930 }, { "epoch": 0.56, "learning_rate": 1.0007597796179003e-06, "loss": 0.8832, "step": 12931 }, { "epoch": 0.56, "learning_rate": 9.998199755496173e-07, "loss": 0.8937, "step": 12932 }, { "epoch": 0.56, "learning_rate": 9.988805897477627e-07, "loss": 0.9412, "step": 12933 }, { "epoch": 0.56, "learning_rate": 9.979416222559945e-07, "loss": 0.9491, "step": 12934 }, { "epoch": 0.56, "learning_rate": 9.970030731179514e-07, "loss": 0.9328, "step": 12935 }, { "epoch": 0.56, "learning_rate": 9.960649423772494e-07, "loss": 0.9279, "step": 12936 }, { "epoch": 0.56, "learning_rate": 9.95127230077486e-07, "loss": 0.9404, "step": 12937 }, { "epoch": 0.56, "learning_rate": 9.941899362622387e-07, "loss": 0.9097, "step": 12938 }, { "epoch": 0.56, "learning_rate": 9.9325306097507e-07, "loss": 0.9509, "step": 12939 }, { "epoch": 0.56, "learning_rate": 9.923166042595134e-07, "loss": 0.8984, "step": 12940 }, { "epoch": 0.56, "learning_rate": 9.913805661590958e-07, "loss": 0.923, "step": 12941 }, { "epoch": 0.56, "learning_rate": 9.904449467173127e-07, "loss": 0.9384, "step": 12942 }, { "epoch": 0.56, "learning_rate": 9.89509745977648e-07, "loss": 0.9085, "step": 12943 }, { "epoch": 0.56, "learning_rate": 9.885749639835618e-07, "loss": 0.9067, "step": 12944 }, { "epoch": 0.56, "learning_rate": 9.876406007784989e-07, "loss": 0.9273, "step": 12945 }, { "epoch": 0.56, "learning_rate": 9.867066564058791e-07, "loss": 0.8892, "step": 12946 }, { "epoch": 0.56, "learning_rate": 9.857731309091057e-07, "loss": 0.9049, "step": 12947 }, { "epoch": 0.56, "learning_rate": 9.84840024331567e-07, "loss": 0.8746, "step": 12948 }, { "epoch": 0.56, "learning_rate": 9.839073367166207e-07, "loss": 0.9476, "step": 12949 }, { "epoch": 0.56, "learning_rate": 9.829750681076155e-07, "loss": 0.9005, "step": 12950 }, { "epoch": 0.56, "learning_rate": 9.820432185478767e-07, "loss": 0.8895, "step": 12951 }, { "epoch": 0.56, "learning_rate": 9.811117880807108e-07, "loss": 0.8982, "step": 12952 }, { "epoch": 0.56, "learning_rate": 9.801807767494042e-07, "loss": 0.9498, "step": 12953 }, { "epoch": 0.56, "learning_rate": 9.792501845972213e-07, "loss": 0.9315, "step": 12954 }, { "epoch": 0.56, "learning_rate": 9.783200116674119e-07, "loss": 0.9322, "step": 12955 }, { "epoch": 0.56, "learning_rate": 9.773902580032035e-07, "loss": 0.9281, "step": 12956 }, { "epoch": 0.56, "learning_rate": 9.764609236478063e-07, "loss": 0.9552, "step": 12957 }, { "epoch": 0.56, "learning_rate": 9.755320086444064e-07, "loss": 0.9092, "step": 12958 }, { "epoch": 0.56, "learning_rate": 9.746035130361741e-07, "loss": 0.9242, "step": 12959 }, { "epoch": 0.56, "learning_rate": 9.736754368662626e-07, "loss": 0.9531, "step": 12960 }, { "epoch": 0.56, "learning_rate": 9.727477801777985e-07, "loss": 0.9209, "step": 12961 }, { "epoch": 0.56, "learning_rate": 9.71820543013894e-07, "loss": 0.9397, "step": 12962 }, { "epoch": 0.56, "learning_rate": 9.70893725417641e-07, "loss": 0.9224, "step": 12963 }, { "epoch": 0.56, "learning_rate": 9.699673274321143e-07, "loss": 0.926, "step": 12964 }, { "epoch": 0.56, "learning_rate": 9.690413491003591e-07, "loss": 0.9463, "step": 12965 }, { "epoch": 0.56, "learning_rate": 9.681157904654181e-07, "loss": 0.8946, "step": 12966 }, { "epoch": 0.56, "learning_rate": 9.671906515702979e-07, "loss": 0.9214, "step": 12967 }, { "epoch": 0.56, "learning_rate": 9.662659324579948e-07, "loss": 0.8922, "step": 12968 }, { "epoch": 0.56, "learning_rate": 9.65341633171486e-07, "loss": 0.961, "step": 12969 }, { "epoch": 0.56, "learning_rate": 9.644177537537214e-07, "loss": 0.9289, "step": 12970 }, { "epoch": 0.56, "learning_rate": 9.634942942476388e-07, "loss": 0.9173, "step": 12971 }, { "epoch": 0.56, "learning_rate": 9.625712546961552e-07, "loss": 0.8907, "step": 12972 }, { "epoch": 0.56, "learning_rate": 9.616486351421683e-07, "loss": 0.9468, "step": 12973 }, { "epoch": 0.56, "learning_rate": 9.607264356285495e-07, "loss": 0.8812, "step": 12974 }, { "epoch": 0.56, "learning_rate": 9.598046561981645e-07, "loss": 0.9621, "step": 12975 }, { "epoch": 0.56, "learning_rate": 9.588832968938466e-07, "loss": 0.9865, "step": 12976 }, { "epoch": 0.56, "learning_rate": 9.579623577584108e-07, "loss": 0.9305, "step": 12977 }, { "epoch": 0.56, "learning_rate": 9.570418388346625e-07, "loss": 0.9307, "step": 12978 }, { "epoch": 0.56, "learning_rate": 9.561217401653778e-07, "loss": 0.9173, "step": 12979 }, { "epoch": 0.56, "learning_rate": 9.552020617933166e-07, "loss": 0.9168, "step": 12980 }, { "epoch": 0.56, "learning_rate": 9.542828037612195e-07, "loss": 0.9557, "step": 12981 }, { "epoch": 0.56, "learning_rate": 9.533639661118099e-07, "loss": 0.9, "step": 12982 }, { "epoch": 0.56, "learning_rate": 9.524455488877837e-07, "loss": 0.9537, "step": 12983 }, { "epoch": 0.56, "learning_rate": 9.515275521318268e-07, "loss": 0.8975, "step": 12984 }, { "epoch": 0.56, "learning_rate": 9.506099758866017e-07, "loss": 0.8977, "step": 12985 }, { "epoch": 0.56, "learning_rate": 9.496928201947453e-07, "loss": 0.9198, "step": 12986 }, { "epoch": 0.56, "learning_rate": 9.487760850988892e-07, "loss": 0.8991, "step": 12987 }, { "epoch": 0.56, "learning_rate": 9.478597706416303e-07, "loss": 0.9196, "step": 12988 }, { "epoch": 0.56, "learning_rate": 9.469438768655548e-07, "loss": 0.9174, "step": 12989 }, { "epoch": 0.56, "learning_rate": 9.460284038132273e-07, "loss": 0.8981, "step": 12990 }, { "epoch": 0.56, "learning_rate": 9.451133515271937e-07, "loss": 0.9579, "step": 12991 }, { "epoch": 0.56, "learning_rate": 9.44198720049978e-07, "loss": 0.9258, "step": 12992 }, { "epoch": 0.56, "learning_rate": 9.43284509424085e-07, "loss": 0.9546, "step": 12993 }, { "epoch": 0.56, "learning_rate": 9.42370719692004e-07, "loss": 0.9317, "step": 12994 }, { "epoch": 0.56, "learning_rate": 9.414573508961989e-07, "loss": 0.9133, "step": 12995 }, { "epoch": 0.56, "learning_rate": 9.405444030791166e-07, "loss": 0.9329, "step": 12996 }, { "epoch": 0.56, "learning_rate": 9.396318762831868e-07, "loss": 0.8955, "step": 12997 }, { "epoch": 0.56, "learning_rate": 9.387197705508155e-07, "loss": 0.9203, "step": 12998 }, { "epoch": 0.56, "learning_rate": 9.378080859243943e-07, "loss": 0.9033, "step": 12999 }, { "epoch": 0.56, "learning_rate": 9.368968224462871e-07, "loss": 0.9217, "step": 13000 }, { "epoch": 0.57, "learning_rate": 9.359859801588467e-07, "loss": 0.9435, "step": 13001 }, { "epoch": 0.57, "learning_rate": 9.350755591044014e-07, "loss": 0.9037, "step": 13002 }, { "epoch": 0.57, "learning_rate": 9.34165559325263e-07, "loss": 0.9379, "step": 13003 }, { "epoch": 0.57, "learning_rate": 9.332559808637198e-07, "loss": 0.9466, "step": 13004 }, { "epoch": 0.57, "learning_rate": 9.323468237620436e-07, "loss": 0.9035, "step": 13005 }, { "epoch": 0.57, "learning_rate": 9.314380880624874e-07, "loss": 0.8783, "step": 13006 }, { "epoch": 0.57, "learning_rate": 9.305297738072805e-07, "loss": 0.9413, "step": 13007 }, { "epoch": 0.57, "learning_rate": 9.296218810386348e-07, "loss": 0.9564, "step": 13008 }, { "epoch": 0.57, "learning_rate": 9.287144097987455e-07, "loss": 0.9033, "step": 13009 }, { "epoch": 0.57, "learning_rate": 9.278073601297854e-07, "loss": 0.9417, "step": 13010 }, { "epoch": 0.57, "learning_rate": 9.269007320739032e-07, "loss": 0.9718, "step": 13011 }, { "epoch": 0.57, "learning_rate": 9.259945256732394e-07, "loss": 0.9357, "step": 13012 }, { "epoch": 0.57, "learning_rate": 9.250887409699038e-07, "loss": 0.9303, "step": 13013 }, { "epoch": 0.57, "learning_rate": 9.241833780059928e-07, "loss": 0.9159, "step": 13014 }, { "epoch": 0.57, "learning_rate": 9.232784368235826e-07, "loss": 0.9484, "step": 13015 }, { "epoch": 0.57, "learning_rate": 9.223739174647262e-07, "loss": 0.9623, "step": 13016 }, { "epoch": 0.57, "learning_rate": 9.214698199714589e-07, "loss": 0.9197, "step": 13017 }, { "epoch": 0.57, "learning_rate": 9.205661443857994e-07, "loss": 0.9331, "step": 13018 }, { "epoch": 0.57, "learning_rate": 9.19662890749744e-07, "loss": 0.9694, "step": 13019 }, { "epoch": 0.57, "learning_rate": 9.187600591052659e-07, "loss": 0.9104, "step": 13020 }, { "epoch": 0.57, "learning_rate": 9.178576494943303e-07, "loss": 0.945, "step": 13021 }, { "epoch": 0.57, "learning_rate": 9.169556619588693e-07, "loss": 0.9188, "step": 13022 }, { "epoch": 0.57, "learning_rate": 9.160540965407983e-07, "loss": 0.9212, "step": 13023 }, { "epoch": 0.57, "learning_rate": 9.151529532820236e-07, "loss": 0.9192, "step": 13024 }, { "epoch": 0.57, "learning_rate": 9.142522322244185e-07, "loss": 0.938, "step": 13025 }, { "epoch": 0.57, "learning_rate": 9.133519334098439e-07, "loss": 0.922, "step": 13026 }, { "epoch": 0.57, "learning_rate": 9.124520568801387e-07, "loss": 0.9027, "step": 13027 }, { "epoch": 0.57, "learning_rate": 9.115526026771271e-07, "loss": 0.8907, "step": 13028 }, { "epoch": 0.57, "learning_rate": 9.106535708426034e-07, "loss": 0.9265, "step": 13029 }, { "epoch": 0.57, "learning_rate": 9.097549614183521e-07, "loss": 0.9005, "step": 13030 }, { "epoch": 0.57, "learning_rate": 9.08856774446134e-07, "loss": 0.9307, "step": 13031 }, { "epoch": 0.57, "learning_rate": 9.079590099676882e-07, "loss": 0.957, "step": 13032 }, { "epoch": 0.57, "learning_rate": 9.070616680247412e-07, "loss": 0.9578, "step": 13033 }, { "epoch": 0.57, "learning_rate": 9.06164748658992e-07, "loss": 0.8955, "step": 13034 }, { "epoch": 0.57, "learning_rate": 9.052682519121225e-07, "loss": 0.924, "step": 13035 }, { "epoch": 0.57, "learning_rate": 9.043721778257985e-07, "loss": 0.8979, "step": 13036 }, { "epoch": 0.57, "learning_rate": 9.034765264416634e-07, "loss": 0.9043, "step": 13037 }, { "epoch": 0.57, "learning_rate": 9.025812978013371e-07, "loss": 0.9596, "step": 13038 }, { "epoch": 0.57, "learning_rate": 9.016864919464263e-07, "loss": 0.9251, "step": 13039 }, { "epoch": 0.57, "learning_rate": 9.007921089185178e-07, "loss": 0.9166, "step": 13040 }, { "epoch": 0.57, "learning_rate": 8.998981487591718e-07, "loss": 0.8987, "step": 13041 }, { "epoch": 0.57, "learning_rate": 8.990046115099349e-07, "loss": 0.9315, "step": 13042 }, { "epoch": 0.57, "learning_rate": 8.981114972123329e-07, "loss": 0.9406, "step": 13043 }, { "epoch": 0.57, "learning_rate": 8.972188059078723e-07, "loss": 0.9173, "step": 13044 }, { "epoch": 0.57, "learning_rate": 8.963265376380403e-07, "loss": 0.9407, "step": 13045 }, { "epoch": 0.57, "learning_rate": 8.954346924443002e-07, "loss": 0.9261, "step": 13046 }, { "epoch": 0.57, "learning_rate": 8.945432703681001e-07, "loss": 0.8905, "step": 13047 }, { "epoch": 0.57, "learning_rate": 8.936522714508678e-07, "loss": 0.9068, "step": 13048 }, { "epoch": 0.57, "learning_rate": 8.927616957340124e-07, "loss": 0.9205, "step": 13049 }, { "epoch": 0.57, "learning_rate": 8.918715432589175e-07, "loss": 0.93, "step": 13050 }, { "epoch": 0.57, "learning_rate": 8.909818140669546e-07, "loss": 0.8955, "step": 13051 }, { "epoch": 0.57, "learning_rate": 8.900925081994716e-07, "loss": 0.9335, "step": 13052 }, { "epoch": 0.57, "learning_rate": 8.892036256977965e-07, "loss": 0.9133, "step": 13053 }, { "epoch": 0.57, "learning_rate": 8.883151666032375e-07, "loss": 0.911, "step": 13054 }, { "epoch": 0.57, "learning_rate": 8.874271309570858e-07, "loss": 0.8984, "step": 13055 }, { "epoch": 0.57, "learning_rate": 8.86539518800611e-07, "loss": 0.9209, "step": 13056 }, { "epoch": 0.57, "learning_rate": 8.85652330175063e-07, "loss": 0.907, "step": 13057 }, { "epoch": 0.57, "learning_rate": 8.847655651216747e-07, "loss": 0.9492, "step": 13058 }, { "epoch": 0.57, "learning_rate": 8.838792236816518e-07, "loss": 0.9414, "step": 13059 }, { "epoch": 0.57, "learning_rate": 8.829933058961882e-07, "loss": 0.9015, "step": 13060 }, { "epoch": 0.57, "learning_rate": 8.821078118064563e-07, "loss": 0.9424, "step": 13061 }, { "epoch": 0.57, "learning_rate": 8.812227414536045e-07, "loss": 0.9257, "step": 13062 }, { "epoch": 0.57, "learning_rate": 8.803380948787676e-07, "loss": 0.9392, "step": 13063 }, { "epoch": 0.57, "learning_rate": 8.794538721230561e-07, "loss": 0.9153, "step": 13064 }, { "epoch": 0.57, "learning_rate": 8.785700732275659e-07, "loss": 0.9657, "step": 13065 }, { "epoch": 0.57, "learning_rate": 8.776866982333632e-07, "loss": 0.9501, "step": 13066 }, { "epoch": 0.57, "learning_rate": 8.768037471815094e-07, "loss": 0.8892, "step": 13067 }, { "epoch": 0.57, "learning_rate": 8.759212201130341e-07, "loss": 0.9497, "step": 13068 }, { "epoch": 0.57, "learning_rate": 8.750391170689476e-07, "loss": 0.9143, "step": 13069 }, { "epoch": 0.57, "learning_rate": 8.741574380902506e-07, "loss": 0.9047, "step": 13070 }, { "epoch": 0.57, "learning_rate": 8.732761832179126e-07, "loss": 0.913, "step": 13071 }, { "epoch": 0.57, "learning_rate": 8.723953524928908e-07, "loss": 0.9221, "step": 13072 }, { "epoch": 0.57, "learning_rate": 8.71514945956119e-07, "loss": 0.9305, "step": 13073 }, { "epoch": 0.57, "learning_rate": 8.706349636485134e-07, "loss": 0.9884, "step": 13074 }, { "epoch": 0.57, "learning_rate": 8.697554056109681e-07, "loss": 0.9424, "step": 13075 }, { "epoch": 0.57, "learning_rate": 8.688762718843591e-07, "loss": 0.9224, "step": 13076 }, { "epoch": 0.57, "learning_rate": 8.67997562509546e-07, "loss": 0.9564, "step": 13077 }, { "epoch": 0.57, "learning_rate": 8.671192775273574e-07, "loss": 0.9518, "step": 13078 }, { "epoch": 0.57, "learning_rate": 8.662414169786193e-07, "loss": 0.9292, "step": 13079 }, { "epoch": 0.57, "learning_rate": 8.653639809041215e-07, "loss": 0.9463, "step": 13080 }, { "epoch": 0.57, "learning_rate": 8.644869693446434e-07, "loss": 0.9481, "step": 13081 }, { "epoch": 0.57, "learning_rate": 8.636103823409436e-07, "loss": 0.9143, "step": 13082 }, { "epoch": 0.57, "learning_rate": 8.627342199337607e-07, "loss": 0.9414, "step": 13083 }, { "epoch": 0.57, "learning_rate": 8.618584821638098e-07, "loss": 0.9533, "step": 13084 }, { "epoch": 0.57, "learning_rate": 8.609831690717896e-07, "loss": 0.9143, "step": 13085 }, { "epoch": 0.57, "learning_rate": 8.601082806983818e-07, "loss": 0.9045, "step": 13086 }, { "epoch": 0.57, "learning_rate": 8.592338170842385e-07, "loss": 0.9689, "step": 13087 }, { "epoch": 0.57, "learning_rate": 8.58359778270007e-07, "loss": 0.9068, "step": 13088 }, { "epoch": 0.57, "learning_rate": 8.574861642963017e-07, "loss": 0.8954, "step": 13089 }, { "epoch": 0.57, "learning_rate": 8.566129752037223e-07, "loss": 0.9232, "step": 13090 }, { "epoch": 0.57, "learning_rate": 8.557402110328516e-07, "loss": 0.9396, "step": 13091 }, { "epoch": 0.57, "learning_rate": 8.548678718242453e-07, "loss": 0.9253, "step": 13092 }, { "epoch": 0.57, "learning_rate": 8.539959576184464e-07, "loss": 0.9155, "step": 13093 }, { "epoch": 0.57, "learning_rate": 8.531244684559759e-07, "loss": 0.9372, "step": 13094 }, { "epoch": 0.57, "learning_rate": 8.522534043773345e-07, "loss": 0.9144, "step": 13095 }, { "epoch": 0.57, "learning_rate": 8.513827654230011e-07, "loss": 0.941, "step": 13096 }, { "epoch": 0.57, "learning_rate": 8.505125516334379e-07, "loss": 0.9053, "step": 13097 }, { "epoch": 0.57, "learning_rate": 8.496427630490888e-07, "loss": 0.9259, "step": 13098 }, { "epoch": 0.57, "learning_rate": 8.487733997103731e-07, "loss": 0.9377, "step": 13099 }, { "epoch": 0.57, "learning_rate": 8.479044616576925e-07, "loss": 0.9561, "step": 13100 }, { "epoch": 0.57, "learning_rate": 8.470359489314306e-07, "loss": 0.9035, "step": 13101 }, { "epoch": 0.57, "learning_rate": 8.461678615719493e-07, "loss": 0.9226, "step": 13102 }, { "epoch": 0.57, "learning_rate": 8.45300199619592e-07, "loss": 0.9298, "step": 13103 }, { "epoch": 0.57, "learning_rate": 8.44432963114683e-07, "loss": 0.9599, "step": 13104 }, { "epoch": 0.57, "learning_rate": 8.435661520975213e-07, "loss": 0.9152, "step": 13105 }, { "epoch": 0.57, "learning_rate": 8.426997666083936e-07, "loss": 0.9412, "step": 13106 }, { "epoch": 0.57, "learning_rate": 8.418338066875642e-07, "loss": 0.9484, "step": 13107 }, { "epoch": 0.57, "learning_rate": 8.409682723752743e-07, "loss": 0.9401, "step": 13108 }, { "epoch": 0.57, "learning_rate": 8.401031637117473e-07, "loss": 0.9447, "step": 13109 }, { "epoch": 0.57, "learning_rate": 8.392384807371912e-07, "loss": 0.889, "step": 13110 }, { "epoch": 0.57, "learning_rate": 8.383742234917891e-07, "loss": 0.9371, "step": 13111 }, { "epoch": 0.57, "learning_rate": 8.375103920157024e-07, "loss": 0.9009, "step": 13112 }, { "epoch": 0.57, "learning_rate": 8.366469863490822e-07, "loss": 0.9142, "step": 13113 }, { "epoch": 0.57, "learning_rate": 8.35784006532051e-07, "loss": 0.9281, "step": 13114 }, { "epoch": 0.57, "learning_rate": 8.349214526047089e-07, "loss": 0.9878, "step": 13115 }, { "epoch": 0.57, "learning_rate": 8.340593246071505e-07, "loss": 0.94, "step": 13116 }, { "epoch": 0.57, "learning_rate": 8.331976225794347e-07, "loss": 0.9025, "step": 13117 }, { "epoch": 0.57, "learning_rate": 8.323363465616108e-07, "loss": 0.8745, "step": 13118 }, { "epoch": 0.57, "learning_rate": 8.314754965937033e-07, "loss": 0.9471, "step": 13119 }, { "epoch": 0.57, "learning_rate": 8.306150727157214e-07, "loss": 0.9696, "step": 13120 }, { "epoch": 0.57, "learning_rate": 8.297550749676475e-07, "loss": 0.9375, "step": 13121 }, { "epoch": 0.57, "learning_rate": 8.288955033894519e-07, "loss": 0.9118, "step": 13122 }, { "epoch": 0.57, "learning_rate": 8.280363580210804e-07, "loss": 0.9325, "step": 13123 }, { "epoch": 0.57, "learning_rate": 8.271776389024577e-07, "loss": 0.8997, "step": 13124 }, { "epoch": 0.57, "learning_rate": 8.263193460734964e-07, "loss": 0.9266, "step": 13125 }, { "epoch": 0.57, "learning_rate": 8.254614795740789e-07, "loss": 0.911, "step": 13126 }, { "epoch": 0.57, "learning_rate": 8.246040394440757e-07, "loss": 0.9203, "step": 13127 }, { "epoch": 0.57, "learning_rate": 8.237470257233349e-07, "loss": 0.9515, "step": 13128 }, { "epoch": 0.57, "learning_rate": 8.228904384516845e-07, "loss": 0.9202, "step": 13129 }, { "epoch": 0.57, "learning_rate": 8.220342776689305e-07, "loss": 0.9381, "step": 13130 }, { "epoch": 0.57, "learning_rate": 8.211785434148634e-07, "loss": 0.9323, "step": 13131 }, { "epoch": 0.57, "learning_rate": 8.203232357292523e-07, "loss": 0.891, "step": 13132 }, { "epoch": 0.57, "learning_rate": 8.194683546518433e-07, "loss": 0.9148, "step": 13133 }, { "epoch": 0.57, "learning_rate": 8.186139002223692e-07, "loss": 0.8839, "step": 13134 }, { "epoch": 0.57, "learning_rate": 8.177598724805358e-07, "loss": 0.8906, "step": 13135 }, { "epoch": 0.57, "learning_rate": 8.169062714660347e-07, "loss": 0.8907, "step": 13136 }, { "epoch": 0.57, "learning_rate": 8.160530972185355e-07, "loss": 0.9675, "step": 13137 }, { "epoch": 0.57, "learning_rate": 8.15200349777685e-07, "loss": 0.9177, "step": 13138 }, { "epoch": 0.57, "learning_rate": 8.143480291831152e-07, "loss": 0.8822, "step": 13139 }, { "epoch": 0.57, "learning_rate": 8.134961354744353e-07, "loss": 0.9128, "step": 13140 }, { "epoch": 0.57, "learning_rate": 8.126446686912382e-07, "loss": 0.9428, "step": 13141 }, { "epoch": 0.57, "learning_rate": 8.11793628873091e-07, "loss": 0.891, "step": 13142 }, { "epoch": 0.57, "learning_rate": 8.109430160595444e-07, "loss": 0.9204, "step": 13143 }, { "epoch": 0.57, "learning_rate": 8.100928302901311e-07, "loss": 0.9399, "step": 13144 }, { "epoch": 0.57, "learning_rate": 8.092430716043576e-07, "loss": 0.9276, "step": 13145 }, { "epoch": 0.57, "learning_rate": 8.083937400417219e-07, "loss": 0.9314, "step": 13146 }, { "epoch": 0.57, "learning_rate": 8.075448356416881e-07, "loss": 0.975, "step": 13147 }, { "epoch": 0.57, "learning_rate": 8.066963584437115e-07, "loss": 0.9683, "step": 13148 }, { "epoch": 0.57, "learning_rate": 8.058483084872215e-07, "loss": 0.8753, "step": 13149 }, { "epoch": 0.57, "learning_rate": 8.050006858116321e-07, "loss": 0.9451, "step": 13150 }, { "epoch": 0.58, "learning_rate": 8.041534904563319e-07, "loss": 0.9058, "step": 13151 }, { "epoch": 0.58, "learning_rate": 8.03306722460695e-07, "loss": 0.8874, "step": 13152 }, { "epoch": 0.58, "learning_rate": 8.02460381864073e-07, "loss": 0.9075, "step": 13153 }, { "epoch": 0.58, "learning_rate": 8.016144687057959e-07, "loss": 0.949, "step": 13154 }, { "epoch": 0.58, "learning_rate": 8.007689830251775e-07, "loss": 0.9342, "step": 13155 }, { "epoch": 0.58, "learning_rate": 7.99923924861511e-07, "loss": 0.8944, "step": 13156 }, { "epoch": 0.58, "learning_rate": 7.990792942540693e-07, "loss": 0.927, "step": 13157 }, { "epoch": 0.58, "learning_rate": 7.982350912421e-07, "loss": 0.8887, "step": 13158 }, { "epoch": 0.58, "learning_rate": 7.973913158648428e-07, "loss": 0.9286, "step": 13159 }, { "epoch": 0.58, "learning_rate": 7.965479681615063e-07, "loss": 0.9454, "step": 13160 }, { "epoch": 0.58, "learning_rate": 7.957050481712836e-07, "loss": 0.9612, "step": 13161 }, { "epoch": 0.58, "learning_rate": 7.948625559333511e-07, "loss": 0.9394, "step": 13162 }, { "epoch": 0.58, "learning_rate": 7.940204914868588e-07, "loss": 0.9679, "step": 13163 }, { "epoch": 0.58, "learning_rate": 7.931788548709396e-07, "loss": 0.9158, "step": 13164 }, { "epoch": 0.58, "learning_rate": 7.92337646124709e-07, "loss": 0.9031, "step": 13165 }, { "epoch": 0.58, "learning_rate": 7.914968652872624e-07, "loss": 0.9013, "step": 13166 }, { "epoch": 0.58, "learning_rate": 7.906565123976684e-07, "loss": 0.8841, "step": 13167 }, { "epoch": 0.58, "learning_rate": 7.898165874949837e-07, "loss": 0.9442, "step": 13168 }, { "epoch": 0.58, "learning_rate": 7.889770906182437e-07, "loss": 0.8912, "step": 13169 }, { "epoch": 0.58, "learning_rate": 7.881380218064572e-07, "loss": 0.9033, "step": 13170 }, { "epoch": 0.58, "learning_rate": 7.872993810986252e-07, "loss": 0.8926, "step": 13171 }, { "epoch": 0.58, "learning_rate": 7.864611685337175e-07, "loss": 0.9214, "step": 13172 }, { "epoch": 0.58, "learning_rate": 7.8562338415069e-07, "loss": 0.9883, "step": 13173 }, { "epoch": 0.58, "learning_rate": 7.847860279884756e-07, "loss": 0.9089, "step": 13174 }, { "epoch": 0.58, "learning_rate": 7.839491000859922e-07, "loss": 0.9248, "step": 13175 }, { "epoch": 0.58, "learning_rate": 7.83112600482131e-07, "loss": 0.9347, "step": 13176 }, { "epoch": 0.58, "learning_rate": 7.822765292157675e-07, "loss": 0.9565, "step": 13177 }, { "epoch": 0.58, "learning_rate": 7.814408863257583e-07, "loss": 0.9547, "step": 13178 }, { "epoch": 0.58, "learning_rate": 7.806056718509347e-07, "loss": 0.9511, "step": 13179 }, { "epoch": 0.58, "learning_rate": 7.797708858301167e-07, "loss": 0.9136, "step": 13180 }, { "epoch": 0.58, "learning_rate": 7.789365283020956e-07, "loss": 0.9325, "step": 13181 }, { "epoch": 0.58, "learning_rate": 7.781025993056468e-07, "loss": 0.9312, "step": 13182 }, { "epoch": 0.58, "learning_rate": 7.772690988795295e-07, "loss": 0.9349, "step": 13183 }, { "epoch": 0.58, "learning_rate": 7.764360270624738e-07, "loss": 0.9451, "step": 13184 }, { "epoch": 0.58, "learning_rate": 7.756033838931975e-07, "loss": 0.8686, "step": 13185 }, { "epoch": 0.58, "learning_rate": 7.747711694103966e-07, "loss": 0.8783, "step": 13186 }, { "epoch": 0.58, "learning_rate": 7.739393836527476e-07, "loss": 0.9578, "step": 13187 }, { "epoch": 0.58, "learning_rate": 7.731080266589031e-07, "loss": 0.9462, "step": 13188 }, { "epoch": 0.58, "learning_rate": 7.722770984674999e-07, "loss": 0.9492, "step": 13189 }, { "epoch": 0.58, "learning_rate": 7.714465991171571e-07, "loss": 0.9832, "step": 13190 }, { "epoch": 0.58, "learning_rate": 7.706165286464639e-07, "loss": 0.8532, "step": 13191 }, { "epoch": 0.58, "learning_rate": 7.697868870940039e-07, "loss": 0.9191, "step": 13192 }, { "epoch": 0.58, "learning_rate": 7.689576744983285e-07, "loss": 0.9196, "step": 13193 }, { "epoch": 0.58, "learning_rate": 7.681288908979744e-07, "loss": 0.9255, "step": 13194 }, { "epoch": 0.58, "learning_rate": 7.673005363314578e-07, "loss": 0.9475, "step": 13195 }, { "epoch": 0.58, "learning_rate": 7.664726108372777e-07, "loss": 0.9359, "step": 13196 }, { "epoch": 0.58, "learning_rate": 7.656451144539057e-07, "loss": 0.937, "step": 13197 }, { "epoch": 0.58, "learning_rate": 7.648180472198008e-07, "loss": 0.937, "step": 13198 }, { "epoch": 0.58, "learning_rate": 7.639914091734013e-07, "loss": 0.9202, "step": 13199 }, { "epoch": 0.58, "learning_rate": 7.631652003531187e-07, "loss": 0.9115, "step": 13200 }, { "epoch": 0.58, "learning_rate": 7.623394207973522e-07, "loss": 0.8936, "step": 13201 }, { "epoch": 0.58, "learning_rate": 7.61514070544479e-07, "loss": 0.8803, "step": 13202 }, { "epoch": 0.58, "learning_rate": 7.606891496328561e-07, "loss": 0.9245, "step": 13203 }, { "epoch": 0.58, "learning_rate": 7.598646581008162e-07, "loss": 0.9267, "step": 13204 }, { "epoch": 0.58, "learning_rate": 7.59040595986682e-07, "loss": 0.8976, "step": 13205 }, { "epoch": 0.58, "learning_rate": 7.58216963328745e-07, "loss": 0.9252, "step": 13206 }, { "epoch": 0.58, "learning_rate": 7.573937601652848e-07, "loss": 0.8772, "step": 13207 }, { "epoch": 0.58, "learning_rate": 7.565709865345594e-07, "loss": 0.9288, "step": 13208 }, { "epoch": 0.58, "learning_rate": 7.557486424748017e-07, "loss": 0.9452, "step": 13209 }, { "epoch": 0.58, "learning_rate": 7.5492672802423e-07, "loss": 0.9536, "step": 13210 }, { "epoch": 0.58, "learning_rate": 7.541052432210427e-07, "loss": 0.9018, "step": 13211 }, { "epoch": 0.58, "learning_rate": 7.53284188103417e-07, "loss": 0.9507, "step": 13212 }, { "epoch": 0.58, "learning_rate": 7.524635627095078e-07, "loss": 0.9033, "step": 13213 }, { "epoch": 0.58, "learning_rate": 7.516433670774526e-07, "loss": 0.9171, "step": 13214 }, { "epoch": 0.58, "learning_rate": 7.508236012453707e-07, "loss": 0.911, "step": 13215 }, { "epoch": 0.58, "learning_rate": 7.50004265251355e-07, "loss": 0.9245, "step": 13216 }, { "epoch": 0.58, "learning_rate": 7.491853591334874e-07, "loss": 0.958, "step": 13217 }, { "epoch": 0.58, "learning_rate": 7.483668829298219e-07, "loss": 0.9594, "step": 13218 }, { "epoch": 0.58, "learning_rate": 7.475488366783956e-07, "loss": 0.9493, "step": 13219 }, { "epoch": 0.58, "learning_rate": 7.467312204172272e-07, "loss": 0.9257, "step": 13220 }, { "epoch": 0.58, "learning_rate": 7.459140341843141e-07, "loss": 0.9268, "step": 13221 }, { "epoch": 0.58, "learning_rate": 7.450972780176314e-07, "loss": 0.9008, "step": 13222 }, { "epoch": 0.58, "learning_rate": 7.442809519551375e-07, "loss": 0.9636, "step": 13223 }, { "epoch": 0.58, "learning_rate": 7.434650560347712e-07, "loss": 0.9006, "step": 13224 }, { "epoch": 0.58, "learning_rate": 7.426495902944453e-07, "loss": 0.9167, "step": 13225 }, { "epoch": 0.58, "learning_rate": 7.418345547720618e-07, "loss": 0.9553, "step": 13226 }, { "epoch": 0.58, "learning_rate": 7.41019949505496e-07, "loss": 0.9141, "step": 13227 }, { "epoch": 0.58, "learning_rate": 7.402057745326042e-07, "loss": 0.917, "step": 13228 }, { "epoch": 0.58, "learning_rate": 7.39392029891226e-07, "loss": 0.929, "step": 13229 }, { "epoch": 0.58, "learning_rate": 7.38578715619176e-07, "loss": 0.9196, "step": 13230 }, { "epoch": 0.58, "learning_rate": 7.377658317542524e-07, "loss": 0.9344, "step": 13231 }, { "epoch": 0.58, "learning_rate": 7.369533783342331e-07, "loss": 0.8986, "step": 13232 }, { "epoch": 0.58, "learning_rate": 7.361413553968766e-07, "loss": 0.9472, "step": 13233 }, { "epoch": 0.58, "learning_rate": 7.353297629799172e-07, "loss": 0.9314, "step": 13234 }, { "epoch": 0.58, "learning_rate": 7.345186011210725e-07, "loss": 0.9311, "step": 13235 }, { "epoch": 0.58, "learning_rate": 7.337078698580436e-07, "loss": 0.9544, "step": 13236 }, { "epoch": 0.58, "learning_rate": 7.328975692285001e-07, "loss": 0.9445, "step": 13237 }, { "epoch": 0.58, "learning_rate": 7.320876992701086e-07, "loss": 0.9126, "step": 13238 }, { "epoch": 0.58, "learning_rate": 7.312782600204993e-07, "loss": 0.9113, "step": 13239 }, { "epoch": 0.58, "learning_rate": 7.304692515172906e-07, "loss": 0.9621, "step": 13240 }, { "epoch": 0.58, "learning_rate": 7.296606737980816e-07, "loss": 0.8899, "step": 13241 }, { "epoch": 0.58, "learning_rate": 7.288525269004509e-07, "loss": 0.9278, "step": 13242 }, { "epoch": 0.58, "learning_rate": 7.280448108619509e-07, "loss": 0.8912, "step": 13243 }, { "epoch": 0.58, "learning_rate": 7.272375257201203e-07, "loss": 0.9611, "step": 13244 }, { "epoch": 0.58, "learning_rate": 7.264306715124791e-07, "loss": 0.9497, "step": 13245 }, { "epoch": 0.58, "learning_rate": 7.256242482765207e-07, "loss": 0.895, "step": 13246 }, { "epoch": 0.58, "learning_rate": 7.24818256049723e-07, "loss": 0.9685, "step": 13247 }, { "epoch": 0.58, "learning_rate": 7.240126948695437e-07, "loss": 0.9127, "step": 13248 }, { "epoch": 0.58, "learning_rate": 7.232075647734194e-07, "loss": 0.9294, "step": 13249 }, { "epoch": 0.58, "learning_rate": 7.22402865798768e-07, "loss": 0.8858, "step": 13250 }, { "epoch": 0.58, "learning_rate": 7.215985979829853e-07, "loss": 0.9562, "step": 13251 }, { "epoch": 0.58, "learning_rate": 7.207947613634481e-07, "loss": 0.9206, "step": 13252 }, { "epoch": 0.58, "learning_rate": 7.199913559775129e-07, "loss": 0.9465, "step": 13253 }, { "epoch": 0.58, "learning_rate": 7.19188381862519e-07, "loss": 0.9422, "step": 13254 }, { "epoch": 0.58, "learning_rate": 7.183858390557796e-07, "loss": 0.8809, "step": 13255 }, { "epoch": 0.58, "learning_rate": 7.175837275945918e-07, "loss": 0.91, "step": 13256 }, { "epoch": 0.58, "learning_rate": 7.167820475162335e-07, "loss": 0.9162, "step": 13257 }, { "epoch": 0.58, "learning_rate": 7.159807988579637e-07, "loss": 0.9065, "step": 13258 }, { "epoch": 0.58, "learning_rate": 7.151799816570138e-07, "loss": 0.9172, "step": 13259 }, { "epoch": 0.58, "learning_rate": 7.143795959506028e-07, "loss": 0.8951, "step": 13260 }, { "epoch": 0.58, "learning_rate": 7.135796417759289e-07, "loss": 0.8795, "step": 13261 }, { "epoch": 0.58, "learning_rate": 7.127801191701633e-07, "loss": 0.9232, "step": 13262 }, { "epoch": 0.58, "learning_rate": 7.119810281704687e-07, "loss": 0.9356, "step": 13263 }, { "epoch": 0.58, "learning_rate": 7.111823688139763e-07, "loss": 0.9069, "step": 13264 }, { "epoch": 0.58, "learning_rate": 7.103841411378055e-07, "loss": 0.9436, "step": 13265 }, { "epoch": 0.58, "learning_rate": 7.095863451790497e-07, "loss": 0.9158, "step": 13266 }, { "epoch": 0.58, "learning_rate": 7.087889809747883e-07, "loss": 0.8994, "step": 13267 }, { "epoch": 0.58, "learning_rate": 7.079920485620751e-07, "loss": 0.9604, "step": 13268 }, { "epoch": 0.58, "learning_rate": 7.071955479779457e-07, "loss": 0.9085, "step": 13269 }, { "epoch": 0.58, "learning_rate": 7.063994792594175e-07, "loss": 0.8993, "step": 13270 }, { "epoch": 0.58, "learning_rate": 7.05603842443483e-07, "loss": 0.9092, "step": 13271 }, { "epoch": 0.58, "learning_rate": 7.048086375671237e-07, "loss": 0.9024, "step": 13272 }, { "epoch": 0.58, "learning_rate": 7.040138646672901e-07, "loss": 0.9511, "step": 13273 }, { "epoch": 0.58, "learning_rate": 7.032195237809192e-07, "loss": 0.9123, "step": 13274 }, { "epoch": 0.58, "learning_rate": 7.024256149449293e-07, "loss": 0.962, "step": 13275 }, { "epoch": 0.58, "learning_rate": 7.01632138196211e-07, "loss": 0.9186, "step": 13276 }, { "epoch": 0.58, "learning_rate": 7.008390935716425e-07, "loss": 0.9487, "step": 13277 }, { "epoch": 0.58, "learning_rate": 7.000464811080787e-07, "loss": 0.866, "step": 13278 }, { "epoch": 0.58, "learning_rate": 6.992543008423558e-07, "loss": 0.9099, "step": 13279 }, { "epoch": 0.58, "learning_rate": 6.984625528112854e-07, "loss": 0.9603, "step": 13280 }, { "epoch": 0.58, "learning_rate": 6.97671237051668e-07, "loss": 0.929, "step": 13281 }, { "epoch": 0.58, "learning_rate": 6.968803536002744e-07, "loss": 0.9155, "step": 13282 }, { "epoch": 0.58, "learning_rate": 6.960899024938583e-07, "loss": 0.9105, "step": 13283 }, { "epoch": 0.58, "learning_rate": 6.952998837691593e-07, "loss": 0.9169, "step": 13284 }, { "epoch": 0.58, "learning_rate": 6.94510297462887e-07, "loss": 0.889, "step": 13285 }, { "epoch": 0.58, "learning_rate": 6.937211436117386e-07, "loss": 0.9446, "step": 13286 }, { "epoch": 0.58, "learning_rate": 6.929324222523892e-07, "loss": 0.9253, "step": 13287 }, { "epoch": 0.58, "learning_rate": 6.921441334214929e-07, "loss": 0.9313, "step": 13288 }, { "epoch": 0.58, "learning_rate": 6.913562771556814e-07, "loss": 0.9099, "step": 13289 }, { "epoch": 0.58, "learning_rate": 6.9056885349157e-07, "loss": 0.9618, "step": 13290 }, { "epoch": 0.58, "learning_rate": 6.89781862465756e-07, "loss": 0.8954, "step": 13291 }, { "epoch": 0.58, "learning_rate": 6.889953041148079e-07, "loss": 0.9264, "step": 13292 }, { "epoch": 0.58, "learning_rate": 6.882091784752831e-07, "loss": 0.9516, "step": 13293 }, { "epoch": 0.58, "learning_rate": 6.874234855837148e-07, "loss": 0.9001, "step": 13294 }, { "epoch": 0.58, "learning_rate": 6.866382254766158e-07, "loss": 0.9133, "step": 13295 }, { "epoch": 0.58, "learning_rate": 6.858533981904791e-07, "loss": 0.9089, "step": 13296 }, { "epoch": 0.58, "learning_rate": 6.850690037617824e-07, "loss": 0.9093, "step": 13297 }, { "epoch": 0.58, "learning_rate": 6.842850422269731e-07, "loss": 0.9985, "step": 13298 }, { "epoch": 0.58, "learning_rate": 6.835015136224865e-07, "loss": 0.9363, "step": 13299 }, { "epoch": 0.58, "learning_rate": 6.827184179847379e-07, "loss": 0.9462, "step": 13300 }, { "epoch": 0.58, "learning_rate": 6.819357553501161e-07, "loss": 0.9055, "step": 13301 }, { "epoch": 0.59, "learning_rate": 6.811535257549962e-07, "loss": 0.9199, "step": 13302 }, { "epoch": 0.59, "learning_rate": 6.803717292357304e-07, "loss": 0.9136, "step": 13303 }, { "epoch": 0.59, "learning_rate": 6.79590365828654e-07, "loss": 0.9442, "step": 13304 }, { "epoch": 0.59, "learning_rate": 6.788094355700725e-07, "loss": 0.9158, "step": 13305 }, { "epoch": 0.59, "learning_rate": 6.780289384962868e-07, "loss": 0.9211, "step": 13306 }, { "epoch": 0.59, "learning_rate": 6.772488746435646e-07, "loss": 0.9373, "step": 13307 }, { "epoch": 0.59, "learning_rate": 6.764692440481546e-07, "loss": 0.872, "step": 13308 }, { "epoch": 0.59, "learning_rate": 6.756900467462946e-07, "loss": 0.9177, "step": 13309 }, { "epoch": 0.59, "learning_rate": 6.749112827741932e-07, "loss": 0.9398, "step": 13310 }, { "epoch": 0.59, "learning_rate": 6.741329521680417e-07, "loss": 0.9323, "step": 13311 }, { "epoch": 0.59, "learning_rate": 6.733550549640133e-07, "loss": 0.9604, "step": 13312 }, { "epoch": 0.59, "learning_rate": 6.725775911982602e-07, "loss": 0.8811, "step": 13313 }, { "epoch": 0.59, "learning_rate": 6.7180056090691e-07, "loss": 0.9239, "step": 13314 }, { "epoch": 0.59, "learning_rate": 6.710239641260752e-07, "loss": 0.9151, "step": 13315 }, { "epoch": 0.59, "learning_rate": 6.70247800891849e-07, "loss": 0.9304, "step": 13316 }, { "epoch": 0.59, "learning_rate": 6.694720712402969e-07, "loss": 0.928, "step": 13317 }, { "epoch": 0.59, "learning_rate": 6.686967752074757e-07, "loss": 0.9082, "step": 13318 }, { "epoch": 0.59, "learning_rate": 6.679219128294123e-07, "loss": 0.8943, "step": 13319 }, { "epoch": 0.59, "learning_rate": 6.671474841421166e-07, "loss": 0.8953, "step": 13320 }, { "epoch": 0.59, "learning_rate": 6.66373489181581e-07, "loss": 0.9333, "step": 13321 }, { "epoch": 0.59, "learning_rate": 6.655999279837733e-07, "loss": 0.9329, "step": 13322 }, { "epoch": 0.59, "learning_rate": 6.648268005846436e-07, "loss": 0.9183, "step": 13323 }, { "epoch": 0.59, "learning_rate": 6.640541070201223e-07, "loss": 0.949, "step": 13324 }, { "epoch": 0.59, "learning_rate": 6.632818473261205e-07, "loss": 0.9197, "step": 13325 }, { "epoch": 0.59, "learning_rate": 6.625100215385227e-07, "loss": 0.9577, "step": 13326 }, { "epoch": 0.59, "learning_rate": 6.617386296932038e-07, "loss": 0.9353, "step": 13327 }, { "epoch": 0.59, "learning_rate": 6.609676718260083e-07, "loss": 0.947, "step": 13328 }, { "epoch": 0.59, "learning_rate": 6.601971479727665e-07, "loss": 0.8964, "step": 13329 }, { "epoch": 0.59, "learning_rate": 6.594270581692885e-07, "loss": 0.8945, "step": 13330 }, { "epoch": 0.59, "learning_rate": 6.586574024513603e-07, "loss": 0.936, "step": 13331 }, { "epoch": 0.59, "learning_rate": 6.578881808547522e-07, "loss": 0.9411, "step": 13332 }, { "epoch": 0.59, "learning_rate": 6.57119393415211e-07, "loss": 0.9476, "step": 13333 }, { "epoch": 0.59, "learning_rate": 6.56351040168467e-07, "loss": 0.9259, "step": 13334 }, { "epoch": 0.59, "learning_rate": 6.555831211502239e-07, "loss": 0.9183, "step": 13335 }, { "epoch": 0.59, "learning_rate": 6.548156363961722e-07, "loss": 0.9714, "step": 13336 }, { "epoch": 0.59, "learning_rate": 6.54048585941981e-07, "loss": 0.9047, "step": 13337 }, { "epoch": 0.59, "learning_rate": 6.532819698232918e-07, "loss": 0.9325, "step": 13338 }, { "epoch": 0.59, "learning_rate": 6.525157880757361e-07, "loss": 0.9174, "step": 13339 }, { "epoch": 0.59, "learning_rate": 6.517500407349198e-07, "loss": 0.9058, "step": 13340 }, { "epoch": 0.59, "learning_rate": 6.50984727836429e-07, "loss": 0.9327, "step": 13341 }, { "epoch": 0.59, "learning_rate": 6.502198494158318e-07, "loss": 0.9013, "step": 13342 }, { "epoch": 0.59, "learning_rate": 6.494554055086732e-07, "loss": 0.9396, "step": 13343 }, { "epoch": 0.59, "learning_rate": 6.486913961504793e-07, "loss": 0.9267, "step": 13344 }, { "epoch": 0.59, "learning_rate": 6.479278213767559e-07, "loss": 0.9642, "step": 13345 }, { "epoch": 0.59, "learning_rate": 6.471646812229893e-07, "loss": 0.923, "step": 13346 }, { "epoch": 0.59, "learning_rate": 6.464019757246442e-07, "loss": 0.9584, "step": 13347 }, { "epoch": 0.59, "learning_rate": 6.456397049171658e-07, "loss": 0.9301, "step": 13348 }, { "epoch": 0.59, "learning_rate": 6.448778688359803e-07, "loss": 0.94, "step": 13349 }, { "epoch": 0.59, "learning_rate": 6.441164675164935e-07, "loss": 0.9096, "step": 13350 }, { "epoch": 0.59, "learning_rate": 6.433555009940851e-07, "loss": 0.9196, "step": 13351 }, { "epoch": 0.59, "learning_rate": 6.425949693041267e-07, "loss": 0.9024, "step": 13352 }, { "epoch": 0.59, "learning_rate": 6.418348724819568e-07, "loss": 0.914, "step": 13353 }, { "epoch": 0.59, "learning_rate": 6.410752105629026e-07, "loss": 0.8584, "step": 13354 }, { "epoch": 0.59, "learning_rate": 6.40315983582267e-07, "loss": 0.9552, "step": 13355 }, { "epoch": 0.59, "learning_rate": 6.39557191575334e-07, "loss": 0.9667, "step": 13356 }, { "epoch": 0.59, "learning_rate": 6.387988345773654e-07, "loss": 0.9456, "step": 13357 }, { "epoch": 0.59, "learning_rate": 6.380409126236064e-07, "loss": 0.9446, "step": 13358 }, { "epoch": 0.59, "learning_rate": 6.372834257492799e-07, "loss": 0.9198, "step": 13359 }, { "epoch": 0.59, "learning_rate": 6.365263739895878e-07, "loss": 0.9118, "step": 13360 }, { "epoch": 0.59, "learning_rate": 6.357697573797117e-07, "loss": 0.9168, "step": 13361 }, { "epoch": 0.59, "learning_rate": 6.35013575954817e-07, "loss": 0.89, "step": 13362 }, { "epoch": 0.59, "learning_rate": 6.342578297500412e-07, "loss": 0.9147, "step": 13363 }, { "epoch": 0.59, "learning_rate": 6.335025188005128e-07, "loss": 0.8959, "step": 13364 }, { "epoch": 0.59, "learning_rate": 6.32747643141327e-07, "loss": 0.94, "step": 13365 }, { "epoch": 0.59, "learning_rate": 6.319932028075682e-07, "loss": 0.9001, "step": 13366 }, { "epoch": 0.59, "learning_rate": 6.312391978342991e-07, "loss": 0.8925, "step": 13367 }, { "epoch": 0.59, "learning_rate": 6.304856282565563e-07, "loss": 0.9402, "step": 13368 }, { "epoch": 0.59, "learning_rate": 6.297324941093641e-07, "loss": 0.9861, "step": 13369 }, { "epoch": 0.59, "learning_rate": 6.28979795427721e-07, "loss": 0.9087, "step": 13370 }, { "epoch": 0.59, "learning_rate": 6.282275322466103e-07, "loss": 0.9119, "step": 13371 }, { "epoch": 0.59, "learning_rate": 6.274757046009871e-07, "loss": 0.9003, "step": 13372 }, { "epoch": 0.59, "learning_rate": 6.267243125257971e-07, "loss": 0.9334, "step": 13373 }, { "epoch": 0.59, "learning_rate": 6.259733560559544e-07, "loss": 0.9398, "step": 13374 }, { "epoch": 0.59, "learning_rate": 6.252228352263612e-07, "loss": 0.9196, "step": 13375 }, { "epoch": 0.59, "learning_rate": 6.244727500718973e-07, "loss": 0.9426, "step": 13376 }, { "epoch": 0.59, "learning_rate": 6.237231006274181e-07, "loss": 0.9161, "step": 13377 }, { "epoch": 0.59, "learning_rate": 6.229738869277647e-07, "loss": 0.9029, "step": 13378 }, { "epoch": 0.59, "learning_rate": 6.222251090077547e-07, "loss": 0.8728, "step": 13379 }, { "epoch": 0.59, "learning_rate": 6.214767669021871e-07, "loss": 0.9178, "step": 13380 }, { "epoch": 0.59, "learning_rate": 6.207288606458384e-07, "loss": 0.9498, "step": 13381 }, { "epoch": 0.59, "learning_rate": 6.199813902734664e-07, "loss": 0.9355, "step": 13382 }, { "epoch": 0.59, "learning_rate": 6.1923435581981e-07, "loss": 0.9462, "step": 13383 }, { "epoch": 0.59, "learning_rate": 6.184877573195824e-07, "loss": 0.9364, "step": 13384 }, { "epoch": 0.59, "learning_rate": 6.17741594807485e-07, "loss": 0.9183, "step": 13385 }, { "epoch": 0.59, "learning_rate": 6.169958683181909e-07, "loss": 0.9075, "step": 13386 }, { "epoch": 0.59, "learning_rate": 6.162505778863582e-07, "loss": 0.9134, "step": 13387 }, { "epoch": 0.59, "learning_rate": 6.15505723546621e-07, "loss": 0.9339, "step": 13388 }, { "epoch": 0.59, "learning_rate": 6.147613053335988e-07, "loss": 0.9431, "step": 13389 }, { "epoch": 0.59, "learning_rate": 6.140173232818824e-07, "loss": 0.9059, "step": 13390 }, { "epoch": 0.59, "learning_rate": 6.132737774260489e-07, "loss": 0.9366, "step": 13391 }, { "epoch": 0.59, "learning_rate": 6.125306678006549e-07, "loss": 0.9303, "step": 13392 }, { "epoch": 0.59, "learning_rate": 6.117879944402328e-07, "loss": 0.9527, "step": 13393 }, { "epoch": 0.59, "learning_rate": 6.110457573792961e-07, "loss": 0.9566, "step": 13394 }, { "epoch": 0.59, "learning_rate": 6.103039566523416e-07, "loss": 0.9044, "step": 13395 }, { "epoch": 0.59, "learning_rate": 6.09562592293842e-07, "loss": 0.9065, "step": 13396 }, { "epoch": 0.59, "learning_rate": 6.088216643382483e-07, "loss": 0.9375, "step": 13397 }, { "epoch": 0.59, "learning_rate": 6.080811728199998e-07, "loss": 0.8987, "step": 13398 }, { "epoch": 0.59, "learning_rate": 6.073411177735034e-07, "loss": 0.9306, "step": 13399 }, { "epoch": 0.59, "learning_rate": 6.066014992331548e-07, "loss": 0.9356, "step": 13400 }, { "epoch": 0.59, "learning_rate": 6.058623172333266e-07, "loss": 0.934, "step": 13401 }, { "epoch": 0.59, "learning_rate": 6.051235718083681e-07, "loss": 0.9187, "step": 13402 }, { "epoch": 0.59, "learning_rate": 6.043852629926139e-07, "loss": 0.95, "step": 13403 }, { "epoch": 0.59, "learning_rate": 6.036473908203744e-07, "loss": 0.9577, "step": 13404 }, { "epoch": 0.59, "learning_rate": 6.029099553259421e-07, "loss": 0.8878, "step": 13405 }, { "epoch": 0.59, "learning_rate": 6.021729565435852e-07, "loss": 0.969, "step": 13406 }, { "epoch": 0.59, "learning_rate": 6.014363945075563e-07, "loss": 0.9564, "step": 13407 }, { "epoch": 0.59, "learning_rate": 6.007002692520869e-07, "loss": 0.9353, "step": 13408 }, { "epoch": 0.59, "learning_rate": 5.999645808113829e-07, "loss": 0.916, "step": 13409 }, { "epoch": 0.59, "learning_rate": 5.992293292196393e-07, "loss": 0.9177, "step": 13410 }, { "epoch": 0.59, "learning_rate": 5.984945145110221e-07, "loss": 0.9354, "step": 13411 }, { "epoch": 0.59, "learning_rate": 5.977601367196806e-07, "loss": 0.9265, "step": 13412 }, { "epoch": 0.59, "learning_rate": 5.970261958797452e-07, "loss": 0.9218, "step": 13413 }, { "epoch": 0.59, "learning_rate": 5.962926920253232e-07, "loss": 0.913, "step": 13414 }, { "epoch": 0.59, "learning_rate": 5.955596251905016e-07, "loss": 0.9282, "step": 13415 }, { "epoch": 0.59, "learning_rate": 5.948269954093511e-07, "loss": 0.9257, "step": 13416 }, { "epoch": 0.59, "learning_rate": 5.940948027159188e-07, "loss": 0.9225, "step": 13417 }, { "epoch": 0.59, "learning_rate": 5.933630471442286e-07, "loss": 0.9322, "step": 13418 }, { "epoch": 0.59, "learning_rate": 5.926317287282923e-07, "loss": 0.8912, "step": 13419 }, { "epoch": 0.59, "learning_rate": 5.919008475020937e-07, "loss": 0.9244, "step": 13420 }, { "epoch": 0.59, "learning_rate": 5.91170403499599e-07, "loss": 0.9433, "step": 13421 }, { "epoch": 0.59, "learning_rate": 5.904403967547567e-07, "loss": 0.9379, "step": 13422 }, { "epoch": 0.59, "learning_rate": 5.897108273014896e-07, "loss": 0.9568, "step": 13423 }, { "epoch": 0.59, "learning_rate": 5.889816951737038e-07, "loss": 0.911, "step": 13424 }, { "epoch": 0.59, "learning_rate": 5.882530004052856e-07, "loss": 0.911, "step": 13425 }, { "epoch": 0.59, "learning_rate": 5.87524743030099e-07, "loss": 0.9594, "step": 13426 }, { "epoch": 0.59, "learning_rate": 5.86796923081987e-07, "loss": 0.9426, "step": 13427 }, { "epoch": 0.59, "learning_rate": 5.860695405947759e-07, "loss": 0.9265, "step": 13428 }, { "epoch": 0.59, "learning_rate": 5.853425956022684e-07, "loss": 0.9852, "step": 13429 }, { "epoch": 0.59, "learning_rate": 5.846160881382456e-07, "loss": 0.9167, "step": 13430 }, { "epoch": 0.59, "learning_rate": 5.83890018236476e-07, "loss": 0.8834, "step": 13431 }, { "epoch": 0.59, "learning_rate": 5.83164385930698e-07, "loss": 0.935, "step": 13432 }, { "epoch": 0.59, "learning_rate": 5.824391912546345e-07, "loss": 0.9243, "step": 13433 }, { "epoch": 0.59, "learning_rate": 5.817144342419889e-07, "loss": 0.9839, "step": 13434 }, { "epoch": 0.59, "learning_rate": 5.80990114926444e-07, "loss": 0.9286, "step": 13435 }, { "epoch": 0.59, "learning_rate": 5.802662333416575e-07, "loss": 0.9248, "step": 13436 }, { "epoch": 0.59, "learning_rate": 5.795427895212725e-07, "loss": 0.9937, "step": 13437 }, { "epoch": 0.59, "learning_rate": 5.788197834989106e-07, "loss": 0.9212, "step": 13438 }, { "epoch": 0.59, "learning_rate": 5.780972153081698e-07, "loss": 0.8865, "step": 13439 }, { "epoch": 0.59, "learning_rate": 5.77375084982632e-07, "loss": 0.922, "step": 13440 }, { "epoch": 0.59, "learning_rate": 5.76653392555856e-07, "loss": 0.9451, "step": 13441 }, { "epoch": 0.59, "learning_rate": 5.759321380613814e-07, "loss": 0.92, "step": 13442 }, { "epoch": 0.59, "learning_rate": 5.752113215327271e-07, "loss": 0.9129, "step": 13443 }, { "epoch": 0.59, "learning_rate": 5.74490943003394e-07, "loss": 0.931, "step": 13444 }, { "epoch": 0.59, "learning_rate": 5.737710025068555e-07, "loss": 0.9187, "step": 13445 }, { "epoch": 0.59, "learning_rate": 5.730515000765724e-07, "loss": 0.8896, "step": 13446 }, { "epoch": 0.59, "learning_rate": 5.723324357459836e-07, "loss": 0.9721, "step": 13447 }, { "epoch": 0.59, "learning_rate": 5.716138095485024e-07, "loss": 0.8758, "step": 13448 }, { "epoch": 0.59, "learning_rate": 5.708956215175276e-07, "loss": 0.9381, "step": 13449 }, { "epoch": 0.59, "learning_rate": 5.70177871686437e-07, "loss": 0.9025, "step": 13450 }, { "epoch": 0.59, "learning_rate": 5.694605600885861e-07, "loss": 0.9147, "step": 13451 }, { "epoch": 0.6, "learning_rate": 5.687436867573081e-07, "loss": 0.94, "step": 13452 }, { "epoch": 0.6, "learning_rate": 5.680272517259211e-07, "loss": 0.9291, "step": 13453 }, { "epoch": 0.6, "learning_rate": 5.673112550277194e-07, "loss": 0.9243, "step": 13454 }, { "epoch": 0.6, "learning_rate": 5.665956966959752e-07, "loss": 0.8934, "step": 13455 }, { "epoch": 0.6, "learning_rate": 5.658805767639486e-07, "loss": 0.9099, "step": 13456 }, { "epoch": 0.6, "learning_rate": 5.651658952648664e-07, "loss": 0.973, "step": 13457 }, { "epoch": 0.6, "learning_rate": 5.644516522319465e-07, "loss": 0.9245, "step": 13458 }, { "epoch": 0.6, "learning_rate": 5.637378476983823e-07, "loss": 0.9315, "step": 13459 }, { "epoch": 0.6, "learning_rate": 5.63024481697344e-07, "loss": 0.8936, "step": 13460 }, { "epoch": 0.6, "learning_rate": 5.62311554261985e-07, "loss": 0.9157, "step": 13461 }, { "epoch": 0.6, "learning_rate": 5.615990654254366e-07, "loss": 0.8644, "step": 13462 }, { "epoch": 0.6, "learning_rate": 5.608870152208135e-07, "loss": 0.939, "step": 13463 }, { "epoch": 0.6, "learning_rate": 5.601754036812012e-07, "loss": 0.9859, "step": 13464 }, { "epoch": 0.6, "learning_rate": 5.594642308396758e-07, "loss": 0.9248, "step": 13465 }, { "epoch": 0.6, "learning_rate": 5.587534967292851e-07, "loss": 0.9499, "step": 13466 }, { "epoch": 0.6, "learning_rate": 5.580432013830606e-07, "loss": 0.92, "step": 13467 }, { "epoch": 0.6, "learning_rate": 5.573333448340113e-07, "loss": 0.9106, "step": 13468 }, { "epoch": 0.6, "learning_rate": 5.566239271151253e-07, "loss": 0.9034, "step": 13469 }, { "epoch": 0.6, "learning_rate": 5.559149482593717e-07, "loss": 0.9588, "step": 13470 }, { "epoch": 0.6, "learning_rate": 5.552064082997011e-07, "loss": 0.8931, "step": 13471 }, { "epoch": 0.6, "learning_rate": 5.544983072690402e-07, "loss": 0.916, "step": 13472 }, { "epoch": 0.6, "learning_rate": 5.537906452002951e-07, "loss": 0.9216, "step": 13473 }, { "epoch": 0.6, "learning_rate": 5.530834221263559e-07, "loss": 0.9303, "step": 13474 }, { "epoch": 0.6, "learning_rate": 5.523766380800888e-07, "loss": 0.8909, "step": 13475 }, { "epoch": 0.6, "learning_rate": 5.516702930943362e-07, "loss": 0.9131, "step": 13476 }, { "epoch": 0.6, "learning_rate": 5.50964387201931e-07, "loss": 0.8922, "step": 13477 }, { "epoch": 0.6, "learning_rate": 5.502589204356746e-07, "loss": 0.9375, "step": 13478 }, { "epoch": 0.6, "learning_rate": 5.495538928283517e-07, "loss": 0.9244, "step": 13479 }, { "epoch": 0.6, "learning_rate": 5.488493044127297e-07, "loss": 0.8882, "step": 13480 }, { "epoch": 0.6, "learning_rate": 5.481451552215522e-07, "loss": 0.9148, "step": 13481 }, { "epoch": 0.6, "learning_rate": 5.474414452875409e-07, "loss": 0.9455, "step": 13482 }, { "epoch": 0.6, "learning_rate": 5.46738174643402e-07, "loss": 0.9301, "step": 13483 }, { "epoch": 0.6, "learning_rate": 5.460353433218201e-07, "loss": 0.9308, "step": 13484 }, { "epoch": 0.6, "learning_rate": 5.453329513554528e-07, "loss": 0.908, "step": 13485 }, { "epoch": 0.6, "learning_rate": 5.446309987769472e-07, "loss": 0.8867, "step": 13486 }, { "epoch": 0.6, "learning_rate": 5.439294856189225e-07, "loss": 0.9265, "step": 13487 }, { "epoch": 0.6, "learning_rate": 5.432284119139808e-07, "loss": 0.9643, "step": 13488 }, { "epoch": 0.6, "learning_rate": 5.425277776947036e-07, "loss": 0.9303, "step": 13489 }, { "epoch": 0.6, "learning_rate": 5.418275829936537e-07, "loss": 0.9301, "step": 13490 }, { "epoch": 0.6, "learning_rate": 5.411278278433673e-07, "loss": 0.8931, "step": 13491 }, { "epoch": 0.6, "learning_rate": 5.404285122763663e-07, "loss": 0.8968, "step": 13492 }, { "epoch": 0.6, "learning_rate": 5.397296363251514e-07, "loss": 0.9551, "step": 13493 }, { "epoch": 0.6, "learning_rate": 5.390312000221976e-07, "loss": 0.9439, "step": 13494 }, { "epoch": 0.6, "learning_rate": 5.383332033999667e-07, "loss": 0.9595, "step": 13495 }, { "epoch": 0.6, "learning_rate": 5.37635646490896e-07, "loss": 0.8811, "step": 13496 }, { "epoch": 0.6, "learning_rate": 5.369385293274032e-07, "loss": 0.9384, "step": 13497 }, { "epoch": 0.6, "learning_rate": 5.362418519418855e-07, "loss": 0.9196, "step": 13498 }, { "epoch": 0.6, "learning_rate": 5.355456143667181e-07, "loss": 0.9781, "step": 13499 }, { "epoch": 0.6, "learning_rate": 5.348498166342608e-07, "loss": 0.9588, "step": 13500 }, { "epoch": 0.6, "learning_rate": 5.341544587768444e-07, "loss": 0.9242, "step": 13501 }, { "epoch": 0.6, "learning_rate": 5.334595408267906e-07, "loss": 0.9213, "step": 13502 }, { "epoch": 0.6, "learning_rate": 5.327650628163894e-07, "loss": 0.9311, "step": 13503 }, { "epoch": 0.6, "learning_rate": 5.32071024777917e-07, "loss": 0.8857, "step": 13504 }, { "epoch": 0.6, "learning_rate": 5.3137742674363e-07, "loss": 0.9201, "step": 13505 }, { "epoch": 0.6, "learning_rate": 5.30684268745757e-07, "loss": 0.8843, "step": 13506 }, { "epoch": 0.6, "learning_rate": 5.299915508165143e-07, "loss": 0.9274, "step": 13507 }, { "epoch": 0.6, "learning_rate": 5.292992729880952e-07, "loss": 0.9508, "step": 13508 }, { "epoch": 0.6, "learning_rate": 5.286074352926718e-07, "loss": 0.9127, "step": 13509 }, { "epoch": 0.6, "learning_rate": 5.279160377623915e-07, "loss": 0.9424, "step": 13510 }, { "epoch": 0.6, "learning_rate": 5.272250804293932e-07, "loss": 0.9454, "step": 13511 }, { "epoch": 0.6, "learning_rate": 5.265345633257835e-07, "loss": 0.9335, "step": 13512 }, { "epoch": 0.6, "learning_rate": 5.25844486483652e-07, "loss": 0.9269, "step": 13513 }, { "epoch": 0.6, "learning_rate": 5.251548499350723e-07, "loss": 0.9087, "step": 13514 }, { "epoch": 0.6, "learning_rate": 5.244656537120907e-07, "loss": 0.901, "step": 13515 }, { "epoch": 0.6, "learning_rate": 5.237768978467362e-07, "loss": 0.9156, "step": 13516 }, { "epoch": 0.6, "learning_rate": 5.230885823710197e-07, "loss": 0.9664, "step": 13517 }, { "epoch": 0.6, "learning_rate": 5.224007073169302e-07, "loss": 0.9178, "step": 13518 }, { "epoch": 0.6, "learning_rate": 5.217132727164298e-07, "loss": 0.8981, "step": 13519 }, { "epoch": 0.6, "learning_rate": 5.21026278601472e-07, "loss": 0.9049, "step": 13520 }, { "epoch": 0.6, "learning_rate": 5.203397250039821e-07, "loss": 0.9391, "step": 13521 }, { "epoch": 0.6, "learning_rate": 5.196536119558615e-07, "loss": 0.9296, "step": 13522 }, { "epoch": 0.6, "learning_rate": 5.189679394890034e-07, "loss": 0.8988, "step": 13523 }, { "epoch": 0.6, "learning_rate": 5.182827076352681e-07, "loss": 0.9139, "step": 13524 }, { "epoch": 0.6, "learning_rate": 5.175979164265021e-07, "loss": 0.9114, "step": 13525 }, { "epoch": 0.6, "learning_rate": 5.16913565894529e-07, "loss": 0.9305, "step": 13526 }, { "epoch": 0.6, "learning_rate": 5.162296560711544e-07, "loss": 0.926, "step": 13527 }, { "epoch": 0.6, "learning_rate": 5.155461869881606e-07, "loss": 0.9482, "step": 13528 }, { "epoch": 0.6, "learning_rate": 5.148631586773089e-07, "loss": 0.9279, "step": 13529 }, { "epoch": 0.6, "learning_rate": 5.141805711703463e-07, "loss": 0.9105, "step": 13530 }, { "epoch": 0.6, "learning_rate": 5.134984244989904e-07, "loss": 0.912, "step": 13531 }, { "epoch": 0.6, "learning_rate": 5.128167186949428e-07, "loss": 0.9, "step": 13532 }, { "epoch": 0.6, "learning_rate": 5.12135453789887e-07, "loss": 0.9005, "step": 13533 }, { "epoch": 0.6, "learning_rate": 5.114546298154821e-07, "loss": 0.9593, "step": 13534 }, { "epoch": 0.6, "learning_rate": 5.107742468033683e-07, "loss": 0.9245, "step": 13535 }, { "epoch": 0.6, "learning_rate": 5.10094304785167e-07, "loss": 0.9304, "step": 13536 }, { "epoch": 0.6, "learning_rate": 5.09414803792474e-07, "loss": 0.9378, "step": 13537 }, { "epoch": 0.6, "learning_rate": 5.087357438568696e-07, "loss": 0.9187, "step": 13538 }, { "epoch": 0.6, "learning_rate": 5.08057125009912e-07, "loss": 0.9089, "step": 13539 }, { "epoch": 0.6, "learning_rate": 5.073789472831369e-07, "loss": 0.8553, "step": 13540 }, { "epoch": 0.6, "learning_rate": 5.067012107080637e-07, "loss": 0.9521, "step": 13541 }, { "epoch": 0.6, "learning_rate": 5.060239153161872e-07, "loss": 0.9226, "step": 13542 }, { "epoch": 0.6, "learning_rate": 5.053470611389855e-07, "loss": 0.95, "step": 13543 }, { "epoch": 0.6, "learning_rate": 5.046706482079111e-07, "loss": 0.9317, "step": 13544 }, { "epoch": 0.6, "learning_rate": 5.039946765544013e-07, "loss": 0.9075, "step": 13545 }, { "epoch": 0.6, "learning_rate": 5.033191462098697e-07, "loss": 0.9132, "step": 13546 }, { "epoch": 0.6, "learning_rate": 5.026440572057101e-07, "loss": 0.9893, "step": 13547 }, { "epoch": 0.6, "learning_rate": 5.019694095732985e-07, "loss": 0.9073, "step": 13548 }, { "epoch": 0.6, "learning_rate": 5.012952033439844e-07, "loss": 0.9218, "step": 13549 }, { "epoch": 0.6, "learning_rate": 5.006214385491015e-07, "loss": 0.9414, "step": 13550 }, { "epoch": 0.6, "learning_rate": 4.999481152199637e-07, "loss": 0.9582, "step": 13551 }, { "epoch": 0.6, "learning_rate": 4.992752333878592e-07, "loss": 0.9628, "step": 13552 }, { "epoch": 0.6, "learning_rate": 4.986027930840609e-07, "loss": 0.9183, "step": 13553 }, { "epoch": 0.6, "learning_rate": 4.979307943398182e-07, "loss": 0.9163, "step": 13554 }, { "epoch": 0.6, "learning_rate": 4.972592371863627e-07, "loss": 0.921, "step": 13555 }, { "epoch": 0.6, "learning_rate": 4.965881216548996e-07, "loss": 0.9275, "step": 13556 }, { "epoch": 0.6, "learning_rate": 4.959174477766237e-07, "loss": 0.9564, "step": 13557 }, { "epoch": 0.6, "learning_rate": 4.952472155826993e-07, "loss": 0.9129, "step": 13558 }, { "epoch": 0.6, "learning_rate": 4.945774251042746e-07, "loss": 0.9199, "step": 13559 }, { "epoch": 0.6, "learning_rate": 4.93908076372479e-07, "loss": 0.9231, "step": 13560 }, { "epoch": 0.6, "learning_rate": 4.932391694184169e-07, "loss": 0.9282, "step": 13561 }, { "epoch": 0.6, "learning_rate": 4.925707042731742e-07, "loss": 0.9088, "step": 13562 }, { "epoch": 0.6, "learning_rate": 4.919026809678185e-07, "loss": 0.9371, "step": 13563 }, { "epoch": 0.6, "learning_rate": 4.912350995333948e-07, "loss": 0.9709, "step": 13564 }, { "epoch": 0.6, "learning_rate": 4.90567960000925e-07, "loss": 0.8834, "step": 13565 }, { "epoch": 0.6, "learning_rate": 4.899012624014177e-07, "loss": 0.9422, "step": 13566 }, { "epoch": 0.6, "learning_rate": 4.892350067658524e-07, "loss": 0.9233, "step": 13567 }, { "epoch": 0.6, "learning_rate": 4.885691931251923e-07, "loss": 0.8999, "step": 13568 }, { "epoch": 0.6, "learning_rate": 4.879038215103826e-07, "loss": 0.9013, "step": 13569 }, { "epoch": 0.6, "learning_rate": 4.87238891952343e-07, "loss": 0.9649, "step": 13570 }, { "epoch": 0.6, "learning_rate": 4.865744044819753e-07, "loss": 0.9561, "step": 13571 }, { "epoch": 0.6, "learning_rate": 4.859103591301606e-07, "loss": 0.9434, "step": 13572 }, { "epoch": 0.6, "learning_rate": 4.852467559277596e-07, "loss": 0.9167, "step": 13573 }, { "epoch": 0.6, "learning_rate": 4.84583594905611e-07, "loss": 0.8893, "step": 13574 }, { "epoch": 0.6, "learning_rate": 4.839208760945335e-07, "loss": 0.9171, "step": 13575 }, { "epoch": 0.6, "learning_rate": 4.832585995253281e-07, "loss": 0.9049, "step": 13576 }, { "epoch": 0.6, "learning_rate": 4.825967652287688e-07, "loss": 0.9705, "step": 13577 }, { "epoch": 0.6, "learning_rate": 4.819353732356191e-07, "loss": 0.9357, "step": 13578 }, { "epoch": 0.6, "learning_rate": 4.812744235766109e-07, "loss": 0.9194, "step": 13579 }, { "epoch": 0.6, "learning_rate": 4.806139162824619e-07, "loss": 0.901, "step": 13580 }, { "epoch": 0.6, "learning_rate": 4.799538513838675e-07, "loss": 0.9409, "step": 13581 }, { "epoch": 0.6, "learning_rate": 4.792942289115054e-07, "loss": 0.8944, "step": 13582 }, { "epoch": 0.6, "learning_rate": 4.786350488960278e-07, "loss": 0.8942, "step": 13583 }, { "epoch": 0.6, "learning_rate": 4.779763113680702e-07, "loss": 0.8812, "step": 13584 }, { "epoch": 0.6, "learning_rate": 4.773180163582458e-07, "loss": 0.9202, "step": 13585 }, { "epoch": 0.6, "learning_rate": 4.7666016389714685e-07, "loss": 0.89, "step": 13586 }, { "epoch": 0.6, "learning_rate": 4.7600275401534555e-07, "loss": 0.8795, "step": 13587 }, { "epoch": 0.6, "learning_rate": 4.753457867433953e-07, "loss": 0.8913, "step": 13588 }, { "epoch": 0.6, "learning_rate": 4.746892621118282e-07, "loss": 0.9402, "step": 13589 }, { "epoch": 0.6, "learning_rate": 4.740331801511522e-07, "loss": 0.9652, "step": 13590 }, { "epoch": 0.6, "learning_rate": 4.733775408918584e-07, "loss": 0.9166, "step": 13591 }, { "epoch": 0.6, "learning_rate": 4.7272234436441797e-07, "loss": 0.9029, "step": 13592 }, { "epoch": 0.6, "learning_rate": 4.720675905992778e-07, "loss": 0.9316, "step": 13593 }, { "epoch": 0.6, "learning_rate": 4.714132796268689e-07, "loss": 0.9385, "step": 13594 }, { "epoch": 0.6, "learning_rate": 4.70759411477596e-07, "loss": 0.9064, "step": 13595 }, { "epoch": 0.6, "learning_rate": 4.701059861818491e-07, "loss": 0.8844, "step": 13596 }, { "epoch": 0.6, "learning_rate": 4.6945300376999403e-07, "loss": 0.9311, "step": 13597 }, { "epoch": 0.6, "learning_rate": 4.6880046427237536e-07, "loss": 0.9507, "step": 13598 }, { "epoch": 0.6, "learning_rate": 4.6814836771932105e-07, "loss": 0.9287, "step": 13599 }, { "epoch": 0.6, "learning_rate": 4.674967141411335e-07, "loss": 0.955, "step": 13600 }, { "epoch": 0.6, "learning_rate": 4.668455035681008e-07, "loss": 0.895, "step": 13601 }, { "epoch": 0.61, "learning_rate": 4.661947360304819e-07, "loss": 0.9349, "step": 13602 }, { "epoch": 0.61, "learning_rate": 4.6554441155852504e-07, "loss": 0.9567, "step": 13603 }, { "epoch": 0.61, "learning_rate": 4.6489453018244925e-07, "loss": 0.9544, "step": 13604 }, { "epoch": 0.61, "learning_rate": 4.6424509193245704e-07, "loss": 0.9115, "step": 13605 }, { "epoch": 0.61, "learning_rate": 4.635960968387321e-07, "loss": 0.9408, "step": 13606 }, { "epoch": 0.61, "learning_rate": 4.6294754493143246e-07, "loss": 0.9185, "step": 13607 }, { "epoch": 0.61, "learning_rate": 4.6229943624069963e-07, "loss": 0.8819, "step": 13608 }, { "epoch": 0.61, "learning_rate": 4.616517707966528e-07, "loss": 0.9681, "step": 13609 }, { "epoch": 0.61, "learning_rate": 4.610045486293924e-07, "loss": 0.9536, "step": 13610 }, { "epoch": 0.61, "learning_rate": 4.6035776976899425e-07, "loss": 0.9501, "step": 13611 }, { "epoch": 0.61, "learning_rate": 4.597114342455189e-07, "loss": 0.9315, "step": 13612 }, { "epoch": 0.61, "learning_rate": 4.5906554208900333e-07, "loss": 0.9236, "step": 13613 }, { "epoch": 0.61, "learning_rate": 4.5842009332946026e-07, "loss": 0.9312, "step": 13614 }, { "epoch": 0.61, "learning_rate": 4.577750879968912e-07, "loss": 0.9202, "step": 13615 }, { "epoch": 0.61, "learning_rate": 4.5713052612126883e-07, "loss": 0.9382, "step": 13616 }, { "epoch": 0.61, "learning_rate": 4.5648640773254704e-07, "loss": 0.9352, "step": 13617 }, { "epoch": 0.61, "learning_rate": 4.558427328606607e-07, "loss": 0.8804, "step": 13618 }, { "epoch": 0.61, "learning_rate": 4.5519950153552593e-07, "loss": 0.9559, "step": 13619 }, { "epoch": 0.61, "learning_rate": 4.5455671378703215e-07, "loss": 0.9371, "step": 13620 }, { "epoch": 0.61, "learning_rate": 4.539143696450532e-07, "loss": 0.9328, "step": 13621 }, { "epoch": 0.61, "learning_rate": 4.532724691394419e-07, "loss": 0.933, "step": 13622 }, { "epoch": 0.61, "learning_rate": 4.526310123000266e-07, "loss": 0.9518, "step": 13623 }, { "epoch": 0.61, "learning_rate": 4.519899991566212e-07, "loss": 0.9342, "step": 13624 }, { "epoch": 0.61, "learning_rate": 4.5134942973901197e-07, "loss": 0.9094, "step": 13625 }, { "epoch": 0.61, "learning_rate": 4.507093040769706e-07, "loss": 0.9154, "step": 13626 }, { "epoch": 0.61, "learning_rate": 4.5006962220024543e-07, "loss": 0.9437, "step": 13627 }, { "epoch": 0.61, "learning_rate": 4.49430384138565e-07, "loss": 0.8968, "step": 13628 }, { "epoch": 0.61, "learning_rate": 4.487915899216344e-07, "loss": 0.915, "step": 13629 }, { "epoch": 0.61, "learning_rate": 4.4815323957914215e-07, "loss": 0.9445, "step": 13630 }, { "epoch": 0.61, "learning_rate": 4.475153331407556e-07, "loss": 0.8974, "step": 13631 }, { "epoch": 0.61, "learning_rate": 4.4687787063611766e-07, "loss": 0.8984, "step": 13632 }, { "epoch": 0.61, "learning_rate": 4.462408520948536e-07, "loss": 0.9639, "step": 13633 }, { "epoch": 0.61, "learning_rate": 4.456042775465685e-07, "loss": 0.8959, "step": 13634 }, { "epoch": 0.61, "learning_rate": 4.449681470208467e-07, "loss": 0.9387, "step": 13635 }, { "epoch": 0.61, "learning_rate": 4.4433246054724987e-07, "loss": 0.9358, "step": 13636 }, { "epoch": 0.61, "learning_rate": 4.4369721815532006e-07, "loss": 0.9335, "step": 13637 }, { "epoch": 0.61, "learning_rate": 4.430624198745803e-07, "loss": 0.9295, "step": 13638 }, { "epoch": 0.61, "learning_rate": 4.4242806573453034e-07, "loss": 0.9154, "step": 13639 }, { "epoch": 0.61, "learning_rate": 4.4179415576465325e-07, "loss": 0.8956, "step": 13640 }, { "epoch": 0.61, "learning_rate": 4.4116068999440543e-07, "loss": 0.9177, "step": 13641 }, { "epoch": 0.61, "learning_rate": 4.405276684532267e-07, "loss": 0.9268, "step": 13642 }, { "epoch": 0.61, "learning_rate": 4.3989509117053796e-07, "loss": 0.9193, "step": 13643 }, { "epoch": 0.61, "learning_rate": 4.3926295817573463e-07, "loss": 0.9534, "step": 13644 }, { "epoch": 0.61, "learning_rate": 4.386312694981931e-07, "loss": 0.913, "step": 13645 }, { "epoch": 0.61, "learning_rate": 4.3800002516727224e-07, "loss": 0.8764, "step": 13646 }, { "epoch": 0.61, "learning_rate": 4.373692252123085e-07, "loss": 0.9463, "step": 13647 }, { "epoch": 0.61, "learning_rate": 4.367388696626129e-07, "loss": 0.9471, "step": 13648 }, { "epoch": 0.61, "learning_rate": 4.3610895854748645e-07, "loss": 0.8889, "step": 13649 }, { "epoch": 0.61, "learning_rate": 4.354794918961969e-07, "loss": 0.9555, "step": 13650 }, { "epoch": 0.61, "learning_rate": 4.348504697380007e-07, "loss": 0.8893, "step": 13651 }, { "epoch": 0.61, "learning_rate": 4.342218921021313e-07, "loss": 0.9095, "step": 13652 }, { "epoch": 0.61, "learning_rate": 4.335937590177974e-07, "loss": 0.9351, "step": 13653 }, { "epoch": 0.61, "learning_rate": 4.329660705141925e-07, "loss": 0.9036, "step": 13654 }, { "epoch": 0.61, "learning_rate": 4.323388266204864e-07, "loss": 0.92, "step": 13655 }, { "epoch": 0.61, "learning_rate": 4.317120273658315e-07, "loss": 0.9504, "step": 13656 }, { "epoch": 0.61, "learning_rate": 4.310856727793522e-07, "loss": 0.9877, "step": 13657 }, { "epoch": 0.61, "learning_rate": 4.30459762890163e-07, "loss": 0.9796, "step": 13658 }, { "epoch": 0.61, "learning_rate": 4.298342977273495e-07, "loss": 0.948, "step": 13659 }, { "epoch": 0.61, "learning_rate": 4.292092773199752e-07, "loss": 0.9411, "step": 13660 }, { "epoch": 0.61, "learning_rate": 4.2858470169709345e-07, "loss": 0.9215, "step": 13661 }, { "epoch": 0.61, "learning_rate": 4.2796057088772434e-07, "loss": 0.8895, "step": 13662 }, { "epoch": 0.61, "learning_rate": 4.27336884920877e-07, "loss": 0.9232, "step": 13663 }, { "epoch": 0.61, "learning_rate": 4.267136438255337e-07, "loss": 0.9382, "step": 13664 }, { "epoch": 0.61, "learning_rate": 4.260908476306613e-07, "loss": 0.9157, "step": 13665 }, { "epoch": 0.61, "learning_rate": 4.2546849636520006e-07, "loss": 0.9466, "step": 13666 }, { "epoch": 0.61, "learning_rate": 4.248465900580734e-07, "loss": 0.9242, "step": 13667 }, { "epoch": 0.61, "learning_rate": 4.2422512873818486e-07, "loss": 0.9489, "step": 13668 }, { "epoch": 0.61, "learning_rate": 4.236041124344126e-07, "loss": 0.9206, "step": 13669 }, { "epoch": 0.61, "learning_rate": 4.2298354117562114e-07, "loss": 0.9139, "step": 13670 }, { "epoch": 0.61, "learning_rate": 4.2236341499064635e-07, "loss": 0.9267, "step": 13671 }, { "epoch": 0.61, "learning_rate": 4.217437339083097e-07, "loss": 0.9435, "step": 13672 }, { "epoch": 0.61, "learning_rate": 4.211244979574103e-07, "loss": 0.8999, "step": 13673 }, { "epoch": 0.61, "learning_rate": 4.2050570716672514e-07, "loss": 0.9463, "step": 13674 }, { "epoch": 0.61, "learning_rate": 4.198873615650101e-07, "loss": 0.9484, "step": 13675 }, { "epoch": 0.61, "learning_rate": 4.192694611810033e-07, "loss": 0.9406, "step": 13676 }, { "epoch": 0.61, "learning_rate": 4.186520060434207e-07, "loss": 0.9201, "step": 13677 }, { "epoch": 0.61, "learning_rate": 4.180349961809549e-07, "loss": 0.9665, "step": 13678 }, { "epoch": 0.61, "learning_rate": 4.174184316222829e-07, "loss": 0.9367, "step": 13679 }, { "epoch": 0.61, "learning_rate": 4.168023123960574e-07, "loss": 0.9256, "step": 13680 }, { "epoch": 0.61, "learning_rate": 4.16186638530911e-07, "loss": 0.905, "step": 13681 }, { "epoch": 0.61, "learning_rate": 4.155714100554575e-07, "loss": 0.9216, "step": 13682 }, { "epoch": 0.61, "learning_rate": 4.149566269982863e-07, "loss": 0.9053, "step": 13683 }, { "epoch": 0.61, "learning_rate": 4.1434228938797003e-07, "loss": 0.9389, "step": 13684 }, { "epoch": 0.61, "learning_rate": 4.1372839725305813e-07, "loss": 0.9334, "step": 13685 }, { "epoch": 0.61, "learning_rate": 4.1311495062208104e-07, "loss": 0.9349, "step": 13686 }, { "epoch": 0.61, "learning_rate": 4.125019495235449e-07, "loss": 0.8887, "step": 13687 }, { "epoch": 0.61, "learning_rate": 4.118893939859403e-07, "loss": 0.9226, "step": 13688 }, { "epoch": 0.61, "learning_rate": 4.112772840377355e-07, "loss": 0.9211, "step": 13689 }, { "epoch": 0.61, "learning_rate": 4.106656197073733e-07, "loss": 0.9469, "step": 13690 }, { "epoch": 0.61, "learning_rate": 4.100544010232821e-07, "loss": 0.9114, "step": 13691 }, { "epoch": 0.61, "learning_rate": 4.0944362801386807e-07, "loss": 0.9626, "step": 13692 }, { "epoch": 0.61, "learning_rate": 4.08833300707514e-07, "loss": 0.8792, "step": 13693 }, { "epoch": 0.61, "learning_rate": 4.082234191325829e-07, "loss": 0.9325, "step": 13694 }, { "epoch": 0.61, "learning_rate": 4.07613983317422e-07, "loss": 0.9056, "step": 13695 }, { "epoch": 0.61, "learning_rate": 4.070049932903497e-07, "loss": 0.912, "step": 13696 }, { "epoch": 0.61, "learning_rate": 4.063964490796679e-07, "loss": 0.919, "step": 13697 }, { "epoch": 0.61, "learning_rate": 4.0578835071366065e-07, "loss": 0.9308, "step": 13698 }, { "epoch": 0.61, "learning_rate": 4.0518069822058413e-07, "loss": 0.9293, "step": 13699 }, { "epoch": 0.61, "learning_rate": 4.0457349162868033e-07, "loss": 0.9348, "step": 13700 }, { "epoch": 0.61, "learning_rate": 4.039667309661677e-07, "loss": 0.9203, "step": 13701 }, { "epoch": 0.61, "learning_rate": 4.033604162612459e-07, "loss": 1.0022, "step": 13702 }, { "epoch": 0.61, "learning_rate": 4.0275454754208685e-07, "loss": 0.9313, "step": 13703 }, { "epoch": 0.61, "learning_rate": 4.0214912483685366e-07, "loss": 0.9327, "step": 13704 }, { "epoch": 0.61, "learning_rate": 4.015441481736793e-07, "loss": 0.9122, "step": 13705 }, { "epoch": 0.61, "learning_rate": 4.009396175806757e-07, "loss": 0.9242, "step": 13706 }, { "epoch": 0.61, "learning_rate": 4.0033553308594377e-07, "loss": 0.9081, "step": 13707 }, { "epoch": 0.61, "learning_rate": 3.9973189471755325e-07, "loss": 0.9119, "step": 13708 }, { "epoch": 0.61, "learning_rate": 3.991287025035573e-07, "loss": 0.9515, "step": 13709 }, { "epoch": 0.61, "learning_rate": 3.985259564719879e-07, "loss": 0.8875, "step": 13710 }, { "epoch": 0.61, "learning_rate": 3.979236566508593e-07, "loss": 0.8929, "step": 13711 }, { "epoch": 0.61, "learning_rate": 3.9732180306815914e-07, "loss": 0.9041, "step": 13712 }, { "epoch": 0.61, "learning_rate": 3.967203957518584e-07, "loss": 0.9041, "step": 13713 }, { "epoch": 0.61, "learning_rate": 3.9611943472990797e-07, "loss": 0.9249, "step": 13714 }, { "epoch": 0.61, "learning_rate": 3.955189200302323e-07, "loss": 0.9398, "step": 13715 }, { "epoch": 0.61, "learning_rate": 3.949188516807434e-07, "loss": 0.947, "step": 13716 }, { "epoch": 0.61, "learning_rate": 3.943192297093268e-07, "loss": 0.8819, "step": 13717 }, { "epoch": 0.61, "learning_rate": 3.9372005414384794e-07, "loss": 0.9159, "step": 13718 }, { "epoch": 0.61, "learning_rate": 3.931213250121535e-07, "loss": 0.9312, "step": 13719 }, { "epoch": 0.61, "learning_rate": 3.9252304234206896e-07, "loss": 0.9523, "step": 13720 }, { "epoch": 0.61, "learning_rate": 3.919252061613954e-07, "loss": 0.9233, "step": 13721 }, { "epoch": 0.61, "learning_rate": 3.913278164979184e-07, "loss": 0.9111, "step": 13722 }, { "epoch": 0.61, "learning_rate": 3.907308733794024e-07, "loss": 0.8889, "step": 13723 }, { "epoch": 0.61, "learning_rate": 3.9013437683358525e-07, "loss": 0.8762, "step": 13724 }, { "epoch": 0.61, "learning_rate": 3.8953832688818916e-07, "loss": 0.9237, "step": 13725 }, { "epoch": 0.61, "learning_rate": 3.889427235709153e-07, "loss": 0.8789, "step": 13726 }, { "epoch": 0.61, "learning_rate": 3.883475669094439e-07, "loss": 0.9064, "step": 13727 }, { "epoch": 0.61, "learning_rate": 3.8775285693143263e-07, "loss": 0.9352, "step": 13728 }, { "epoch": 0.61, "learning_rate": 3.871585936645195e-07, "loss": 0.9159, "step": 13729 }, { "epoch": 0.61, "learning_rate": 3.865647771363212e-07, "loss": 0.9404, "step": 13730 }, { "epoch": 0.61, "learning_rate": 3.859714073744358e-07, "loss": 0.9284, "step": 13731 }, { "epoch": 0.61, "learning_rate": 3.85378484406439e-07, "loss": 0.9457, "step": 13732 }, { "epoch": 0.61, "learning_rate": 3.8478600825988423e-07, "loss": 0.9512, "step": 13733 }, { "epoch": 0.61, "learning_rate": 3.8419397896230506e-07, "loss": 0.9128, "step": 13734 }, { "epoch": 0.61, "learning_rate": 3.8360239654121946e-07, "loss": 0.927, "step": 13735 }, { "epoch": 0.61, "learning_rate": 3.8301126102411435e-07, "loss": 0.8896, "step": 13736 }, { "epoch": 0.61, "learning_rate": 3.824205724384655e-07, "loss": 0.9429, "step": 13737 }, { "epoch": 0.61, "learning_rate": 3.8183033081172216e-07, "loss": 0.9441, "step": 13738 }, { "epoch": 0.61, "learning_rate": 3.812405361713156e-07, "loss": 0.9096, "step": 13739 }, { "epoch": 0.61, "learning_rate": 3.806511885446551e-07, "loss": 0.9322, "step": 13740 }, { "epoch": 0.61, "learning_rate": 3.8006228795912983e-07, "loss": 0.9209, "step": 13741 }, { "epoch": 0.61, "learning_rate": 3.794738344421078e-07, "loss": 0.9415, "step": 13742 }, { "epoch": 0.61, "learning_rate": 3.788858280209362e-07, "loss": 0.9177, "step": 13743 }, { "epoch": 0.61, "learning_rate": 3.7829826872294194e-07, "loss": 0.9309, "step": 13744 }, { "epoch": 0.61, "learning_rate": 3.777111565754288e-07, "loss": 0.9031, "step": 13745 }, { "epoch": 0.61, "learning_rate": 3.7712449160568374e-07, "loss": 0.8919, "step": 13746 }, { "epoch": 0.61, "learning_rate": 3.765382738409695e-07, "loss": 0.9381, "step": 13747 }, { "epoch": 0.61, "learning_rate": 3.7595250330853207e-07, "loss": 0.944, "step": 13748 }, { "epoch": 0.61, "learning_rate": 3.7536718003558956e-07, "loss": 0.8869, "step": 13749 }, { "epoch": 0.61, "learning_rate": 3.747823040493481e-07, "loss": 0.959, "step": 13750 }, { "epoch": 0.61, "learning_rate": 3.7419787537698813e-07, "loss": 0.9033, "step": 13751 }, { "epoch": 0.62, "learning_rate": 3.7361389404566573e-07, "loss": 0.9075, "step": 13752 }, { "epoch": 0.62, "learning_rate": 3.730303600825258e-07, "loss": 0.9102, "step": 13753 }, { "epoch": 0.62, "learning_rate": 3.7244727351468447e-07, "loss": 0.9203, "step": 13754 }, { "epoch": 0.62, "learning_rate": 3.7186463436923783e-07, "loss": 0.9061, "step": 13755 }, { "epoch": 0.62, "learning_rate": 3.712824426732653e-07, "loss": 0.8664, "step": 13756 }, { "epoch": 0.62, "learning_rate": 3.7070069845382415e-07, "loss": 0.9182, "step": 13757 }, { "epoch": 0.62, "learning_rate": 3.7011940173794726e-07, "loss": 0.9363, "step": 13758 }, { "epoch": 0.62, "learning_rate": 3.695385525526507e-07, "loss": 0.909, "step": 13759 }, { "epoch": 0.62, "learning_rate": 3.689581509249285e-07, "loss": 0.8942, "step": 13760 }, { "epoch": 0.62, "learning_rate": 3.683781968817501e-07, "loss": 0.9179, "step": 13761 }, { "epoch": 0.62, "learning_rate": 3.677986904500741e-07, "loss": 0.9181, "step": 13762 }, { "epoch": 0.62, "learning_rate": 3.6721963165682774e-07, "loss": 0.9372, "step": 13763 }, { "epoch": 0.62, "learning_rate": 3.666410205289228e-07, "loss": 0.9409, "step": 13764 }, { "epoch": 0.62, "learning_rate": 3.66062857093249e-07, "loss": 0.8972, "step": 13765 }, { "epoch": 0.62, "learning_rate": 3.654851413766758e-07, "loss": 0.9453, "step": 13766 }, { "epoch": 0.62, "learning_rate": 3.649078734060507e-07, "loss": 0.9259, "step": 13767 }, { "epoch": 0.62, "learning_rate": 3.6433105320820116e-07, "loss": 0.9353, "step": 13768 }, { "epoch": 0.62, "learning_rate": 3.637546808099357e-07, "loss": 0.928, "step": 13769 }, { "epoch": 0.62, "learning_rate": 3.631787562380362e-07, "loss": 0.9113, "step": 13770 }, { "epoch": 0.62, "learning_rate": 3.6260327951927245e-07, "loss": 0.9138, "step": 13771 }, { "epoch": 0.62, "learning_rate": 3.6202825068038517e-07, "loss": 0.917, "step": 13772 }, { "epoch": 0.62, "learning_rate": 3.6145366974809856e-07, "loss": 0.964, "step": 13773 }, { "epoch": 0.62, "learning_rate": 3.608795367491169e-07, "loss": 0.9305, "step": 13774 }, { "epoch": 0.62, "learning_rate": 3.603058517101199e-07, "loss": 0.8927, "step": 13775 }, { "epoch": 0.62, "learning_rate": 3.597326146577684e-07, "loss": 0.9347, "step": 13776 }, { "epoch": 0.62, "learning_rate": 3.5915982561870346e-07, "loss": 0.9215, "step": 13777 }, { "epoch": 0.62, "learning_rate": 3.5858748461954474e-07, "loss": 0.8506, "step": 13778 }, { "epoch": 0.62, "learning_rate": 3.5801559168688995e-07, "loss": 0.9259, "step": 13779 }, { "epoch": 0.62, "learning_rate": 3.5744414684731666e-07, "loss": 0.9405, "step": 13780 }, { "epoch": 0.62, "learning_rate": 3.5687315012738257e-07, "loss": 0.9192, "step": 13781 }, { "epoch": 0.62, "learning_rate": 3.5630260155362306e-07, "loss": 0.9431, "step": 13782 }, { "epoch": 0.62, "learning_rate": 3.5573250115255253e-07, "loss": 0.9278, "step": 13783 }, { "epoch": 0.62, "learning_rate": 3.5516284895066533e-07, "loss": 0.8944, "step": 13784 }, { "epoch": 0.62, "learning_rate": 3.5459364497443696e-07, "loss": 0.901, "step": 13785 }, { "epoch": 0.62, "learning_rate": 3.540248892503184e-07, "loss": 0.9522, "step": 13786 }, { "epoch": 0.62, "learning_rate": 3.5345658180474306e-07, "loss": 0.9507, "step": 13787 }, { "epoch": 0.62, "learning_rate": 3.528887226641209e-07, "loss": 0.9556, "step": 13788 }, { "epoch": 0.62, "learning_rate": 3.5232131185484075e-07, "loss": 0.9348, "step": 13789 }, { "epoch": 0.62, "learning_rate": 3.517543494032749e-07, "loss": 0.8769, "step": 13790 }, { "epoch": 0.62, "learning_rate": 3.5118783533576893e-07, "loss": 0.9816, "step": 13791 }, { "epoch": 0.62, "learning_rate": 3.506217696786518e-07, "loss": 0.8967, "step": 13792 }, { "epoch": 0.62, "learning_rate": 3.5005615245823023e-07, "loss": 0.9332, "step": 13793 }, { "epoch": 0.62, "learning_rate": 3.4949098370079203e-07, "loss": 0.8899, "step": 13794 }, { "epoch": 0.62, "learning_rate": 3.489262634325985e-07, "loss": 0.8891, "step": 13795 }, { "epoch": 0.62, "learning_rate": 3.483619916798986e-07, "loss": 0.9111, "step": 13796 }, { "epoch": 0.62, "learning_rate": 3.4779816846891247e-07, "loss": 0.9093, "step": 13797 }, { "epoch": 0.62, "learning_rate": 3.4723479382584134e-07, "loss": 0.9264, "step": 13798 }, { "epoch": 0.62, "learning_rate": 3.466718677768721e-07, "loss": 0.9224, "step": 13799 }, { "epoch": 0.62, "learning_rate": 3.461093903481616e-07, "loss": 0.9582, "step": 13800 }, { "epoch": 0.62, "learning_rate": 3.455473615658511e-07, "loss": 0.913, "step": 13801 }, { "epoch": 0.62, "learning_rate": 3.4498578145606087e-07, "loss": 0.8834, "step": 13802 }, { "epoch": 0.62, "learning_rate": 3.444246500448878e-07, "loss": 0.9135, "step": 13803 }, { "epoch": 0.62, "learning_rate": 3.4386396735840987e-07, "loss": 0.9277, "step": 13804 }, { "epoch": 0.62, "learning_rate": 3.433037334226841e-07, "loss": 0.8904, "step": 13805 }, { "epoch": 0.62, "learning_rate": 3.4274394826374625e-07, "loss": 0.9621, "step": 13806 }, { "epoch": 0.62, "learning_rate": 3.421846119076089e-07, "loss": 0.931, "step": 13807 }, { "epoch": 0.62, "learning_rate": 3.416257243802712e-07, "loss": 0.9121, "step": 13808 }, { "epoch": 0.62, "learning_rate": 3.410672857077013e-07, "loss": 1.0353, "step": 13809 }, { "epoch": 0.62, "learning_rate": 3.405092959158551e-07, "loss": 0.9018, "step": 13810 }, { "epoch": 0.62, "learning_rate": 3.399517550306619e-07, "loss": 0.8952, "step": 13811 }, { "epoch": 0.62, "learning_rate": 3.393946630780343e-07, "loss": 0.9302, "step": 13812 }, { "epoch": 0.62, "learning_rate": 3.3883802008386034e-07, "loss": 0.9462, "step": 13813 }, { "epoch": 0.62, "learning_rate": 3.382818260740106e-07, "loss": 0.9493, "step": 13814 }, { "epoch": 0.62, "learning_rate": 3.3772608107433214e-07, "loss": 0.9041, "step": 13815 }, { "epoch": 0.62, "learning_rate": 3.371707851106498e-07, "loss": 0.9649, "step": 13816 }, { "epoch": 0.62, "learning_rate": 3.3661593820877524e-07, "loss": 0.9751, "step": 13817 }, { "epoch": 0.62, "learning_rate": 3.3606154039449e-07, "loss": 0.9309, "step": 13818 }, { "epoch": 0.62, "learning_rate": 3.3550759169356017e-07, "loss": 0.9155, "step": 13819 }, { "epoch": 0.62, "learning_rate": 3.349540921317296e-07, "loss": 0.9286, "step": 13820 }, { "epoch": 0.62, "learning_rate": 3.3440104173471876e-07, "loss": 0.9478, "step": 13821 }, { "epoch": 0.62, "learning_rate": 3.338484405282327e-07, "loss": 0.9271, "step": 13822 }, { "epoch": 0.62, "learning_rate": 3.332962885379509e-07, "loss": 0.9256, "step": 13823 }, { "epoch": 0.62, "learning_rate": 3.3274458578953504e-07, "loss": 0.9467, "step": 13824 }, { "epoch": 0.62, "learning_rate": 3.3219333230862236e-07, "loss": 0.9222, "step": 13825 }, { "epoch": 0.62, "learning_rate": 3.316425281208313e-07, "loss": 0.9413, "step": 13826 }, { "epoch": 0.62, "learning_rate": 3.310921732517636e-07, "loss": 0.9127, "step": 13827 }, { "epoch": 0.62, "learning_rate": 3.305422677269898e-07, "loss": 0.9421, "step": 13828 }, { "epoch": 0.62, "learning_rate": 3.299928115720696e-07, "loss": 0.9115, "step": 13829 }, { "epoch": 0.62, "learning_rate": 3.29443804812537e-07, "loss": 0.9416, "step": 13830 }, { "epoch": 0.62, "learning_rate": 3.288952474739071e-07, "loss": 0.9225, "step": 13831 }, { "epoch": 0.62, "learning_rate": 3.2834713958167176e-07, "loss": 0.9382, "step": 13832 }, { "epoch": 0.62, "learning_rate": 3.277994811613039e-07, "loss": 0.9274, "step": 13833 }, { "epoch": 0.62, "learning_rate": 3.272522722382543e-07, "loss": 0.9159, "step": 13834 }, { "epoch": 0.62, "learning_rate": 3.267055128379537e-07, "loss": 0.9422, "step": 13835 }, { "epoch": 0.62, "learning_rate": 3.2615920298581296e-07, "loss": 0.9041, "step": 13836 }, { "epoch": 0.62, "learning_rate": 3.256133427072183e-07, "loss": 0.9455, "step": 13837 }, { "epoch": 0.62, "learning_rate": 3.2506793202753964e-07, "loss": 0.9538, "step": 13838 }, { "epoch": 0.62, "learning_rate": 3.2452297097212317e-07, "loss": 0.9001, "step": 13839 }, { "epoch": 0.62, "learning_rate": 3.2397845956629535e-07, "loss": 0.9205, "step": 13840 }, { "epoch": 0.62, "learning_rate": 3.2343439783535934e-07, "loss": 0.9431, "step": 13841 }, { "epoch": 0.62, "learning_rate": 3.228907858046026e-07, "loss": 0.9278, "step": 13842 }, { "epoch": 0.62, "learning_rate": 3.2234762349928614e-07, "loss": 0.9363, "step": 13843 }, { "epoch": 0.62, "learning_rate": 3.2180491094465414e-07, "loss": 0.9555, "step": 13844 }, { "epoch": 0.62, "learning_rate": 3.212626481659264e-07, "loss": 0.9184, "step": 13845 }, { "epoch": 0.62, "learning_rate": 3.207208351883051e-07, "loss": 0.9658, "step": 13846 }, { "epoch": 0.62, "learning_rate": 3.2017947203696774e-07, "loss": 0.9057, "step": 13847 }, { "epoch": 0.62, "learning_rate": 3.196385587370754e-07, "loss": 0.9243, "step": 13848 }, { "epoch": 0.62, "learning_rate": 3.190980953137657e-07, "loss": 0.9239, "step": 13849 }, { "epoch": 0.62, "learning_rate": 3.185580817921541e-07, "loss": 0.9077, "step": 13850 }, { "epoch": 0.62, "learning_rate": 3.180185181973383e-07, "loss": 0.9056, "step": 13851 }, { "epoch": 0.62, "learning_rate": 3.174794045543939e-07, "loss": 0.865, "step": 13852 }, { "epoch": 0.62, "learning_rate": 3.169407408883729e-07, "loss": 0.9096, "step": 13853 }, { "epoch": 0.62, "learning_rate": 3.1640252722431096e-07, "loss": 0.8969, "step": 13854 }, { "epoch": 0.62, "learning_rate": 3.158647635872192e-07, "loss": 0.8775, "step": 13855 }, { "epoch": 0.62, "learning_rate": 3.1532745000208976e-07, "loss": 0.8987, "step": 13856 }, { "epoch": 0.62, "learning_rate": 3.147905864938927e-07, "loss": 0.9187, "step": 13857 }, { "epoch": 0.62, "learning_rate": 3.1425417308757923e-07, "loss": 0.9157, "step": 13858 }, { "epoch": 0.62, "learning_rate": 3.13718209808076e-07, "loss": 0.941, "step": 13859 }, { "epoch": 0.62, "learning_rate": 3.13182696680292e-07, "loss": 0.9281, "step": 13860 }, { "epoch": 0.62, "learning_rate": 3.126476337291162e-07, "loss": 0.9721, "step": 13861 }, { "epoch": 0.62, "learning_rate": 3.1211302097940874e-07, "loss": 0.8795, "step": 13862 }, { "epoch": 0.62, "learning_rate": 3.1157885845602197e-07, "loss": 0.9159, "step": 13863 }, { "epoch": 0.62, "learning_rate": 3.11045146183776e-07, "loss": 0.925, "step": 13864 }, { "epoch": 0.62, "learning_rate": 3.1051188418747433e-07, "loss": 0.9276, "step": 13865 }, { "epoch": 0.62, "learning_rate": 3.099790724919005e-07, "loss": 0.8924, "step": 13866 }, { "epoch": 0.62, "learning_rate": 3.0944671112181466e-07, "loss": 0.8925, "step": 13867 }, { "epoch": 0.62, "learning_rate": 3.089148001019582e-07, "loss": 0.9288, "step": 13868 }, { "epoch": 0.62, "learning_rate": 3.0838333945705014e-07, "loss": 0.8888, "step": 13869 }, { "epoch": 0.62, "learning_rate": 3.0785232921179077e-07, "loss": 0.9517, "step": 13870 }, { "epoch": 0.62, "learning_rate": 3.0732176939085476e-07, "loss": 0.9061, "step": 13871 }, { "epoch": 0.62, "learning_rate": 3.0679166001890027e-07, "loss": 0.9536, "step": 13872 }, { "epoch": 0.62, "learning_rate": 3.0626200112056524e-07, "loss": 0.9568, "step": 13873 }, { "epoch": 0.62, "learning_rate": 3.0573279272046006e-07, "loss": 0.9247, "step": 13874 }, { "epoch": 0.62, "learning_rate": 3.0520403484318395e-07, "loss": 0.9332, "step": 13875 }, { "epoch": 0.62, "learning_rate": 3.046757275133061e-07, "loss": 0.9476, "step": 13876 }, { "epoch": 0.62, "learning_rate": 3.041478707553791e-07, "loss": 0.8919, "step": 13877 }, { "epoch": 0.62, "learning_rate": 3.036204645939356e-07, "loss": 0.9419, "step": 13878 }, { "epoch": 0.62, "learning_rate": 3.030935090534859e-07, "loss": 0.9602, "step": 13879 }, { "epoch": 0.62, "learning_rate": 3.0256700415851827e-07, "loss": 0.9153, "step": 13880 }, { "epoch": 0.62, "learning_rate": 3.0204094993349974e-07, "loss": 0.9004, "step": 13881 }, { "epoch": 0.62, "learning_rate": 3.015153464028808e-07, "loss": 0.9424, "step": 13882 }, { "epoch": 0.62, "learning_rate": 3.009901935910853e-07, "loss": 0.9443, "step": 13883 }, { "epoch": 0.62, "learning_rate": 3.0046549152251916e-07, "loss": 0.9083, "step": 13884 }, { "epoch": 0.62, "learning_rate": 2.999412402215673e-07, "loss": 0.9041, "step": 13885 }, { "epoch": 0.62, "learning_rate": 2.9941743971259483e-07, "loss": 0.9258, "step": 13886 }, { "epoch": 0.62, "learning_rate": 2.9889409001994107e-07, "loss": 0.9225, "step": 13887 }, { "epoch": 0.62, "learning_rate": 2.9837119116793214e-07, "loss": 0.8964, "step": 13888 }, { "epoch": 0.62, "learning_rate": 2.978487431808641e-07, "loss": 0.9007, "step": 13889 }, { "epoch": 0.62, "learning_rate": 2.973267460830187e-07, "loss": 0.9068, "step": 13890 }, { "epoch": 0.62, "learning_rate": 2.968051998986576e-07, "loss": 0.9516, "step": 13891 }, { "epoch": 0.62, "learning_rate": 2.9628410465201374e-07, "loss": 0.9242, "step": 13892 }, { "epoch": 0.62, "learning_rate": 2.957634603673065e-07, "loss": 0.8828, "step": 13893 }, { "epoch": 0.62, "learning_rate": 2.952432670687322e-07, "loss": 0.9531, "step": 13894 }, { "epoch": 0.62, "learning_rate": 2.9472352478046585e-07, "loss": 0.9226, "step": 13895 }, { "epoch": 0.62, "learning_rate": 2.942042335266604e-07, "loss": 0.9204, "step": 13896 }, { "epoch": 0.62, "learning_rate": 2.9368539333144875e-07, "loss": 0.9395, "step": 13897 }, { "epoch": 0.62, "learning_rate": 2.9316700421894495e-07, "loss": 0.9755, "step": 13898 }, { "epoch": 0.62, "learning_rate": 2.9264906621323754e-07, "loss": 0.8991, "step": 13899 }, { "epoch": 0.62, "learning_rate": 2.921315793383994e-07, "loss": 0.8912, "step": 13900 }, { "epoch": 0.62, "learning_rate": 2.9161454361847806e-07, "loss": 0.9401, "step": 13901 }, { "epoch": 0.62, "learning_rate": 2.9109795907750205e-07, "loss": 0.9456, "step": 13902 }, { "epoch": 0.63, "learning_rate": 2.905818257394799e-07, "loss": 0.9463, "step": 13903 }, { "epoch": 0.63, "learning_rate": 2.900661436283969e-07, "loss": 0.9218, "step": 13904 }, { "epoch": 0.63, "learning_rate": 2.8955091276821724e-07, "loss": 0.9264, "step": 13905 }, { "epoch": 0.63, "learning_rate": 2.890361331828873e-07, "loss": 0.8942, "step": 13906 }, { "epoch": 0.63, "learning_rate": 2.885218048963301e-07, "loss": 0.937, "step": 13907 }, { "epoch": 0.63, "learning_rate": 2.8800792793244547e-07, "loss": 0.9418, "step": 13908 }, { "epoch": 0.63, "learning_rate": 2.8749450231511877e-07, "loss": 0.9264, "step": 13909 }, { "epoch": 0.63, "learning_rate": 2.869815280682087e-07, "loss": 0.9457, "step": 13910 }, { "epoch": 0.63, "learning_rate": 2.864690052155539e-07, "loss": 0.9216, "step": 13911 }, { "epoch": 0.63, "learning_rate": 2.8595693378097533e-07, "loss": 0.8986, "step": 13912 }, { "epoch": 0.63, "learning_rate": 2.8544531378826733e-07, "loss": 0.9416, "step": 13913 }, { "epoch": 0.63, "learning_rate": 2.8493414526120977e-07, "loss": 0.9378, "step": 13914 }, { "epoch": 0.63, "learning_rate": 2.8442342822355587e-07, "loss": 0.9131, "step": 13915 }, { "epoch": 0.63, "learning_rate": 2.8391316269904103e-07, "loss": 0.9362, "step": 13916 }, { "epoch": 0.63, "learning_rate": 2.8340334871137965e-07, "loss": 0.9043, "step": 13917 }, { "epoch": 0.63, "learning_rate": 2.828939862842628e-07, "loss": 0.9184, "step": 13918 }, { "epoch": 0.63, "learning_rate": 2.8238507544136373e-07, "loss": 0.9408, "step": 13919 }, { "epoch": 0.63, "learning_rate": 2.818766162063302e-07, "loss": 0.9541, "step": 13920 }, { "epoch": 0.63, "learning_rate": 2.813686086027967e-07, "loss": 0.9286, "step": 13921 }, { "epoch": 0.63, "learning_rate": 2.8086105265436757e-07, "loss": 0.94, "step": 13922 }, { "epoch": 0.63, "learning_rate": 2.803539483846329e-07, "loss": 0.9546, "step": 13923 }, { "epoch": 0.63, "learning_rate": 2.798472958171583e-07, "loss": 0.9423, "step": 13924 }, { "epoch": 0.63, "learning_rate": 2.7934109497549154e-07, "loss": 0.9156, "step": 13925 }, { "epoch": 0.63, "learning_rate": 2.7883534588315385e-07, "loss": 0.9477, "step": 13926 }, { "epoch": 0.63, "learning_rate": 2.78330048563652e-07, "loss": 0.9221, "step": 13927 }, { "epoch": 0.63, "learning_rate": 2.778252030404671e-07, "loss": 0.9521, "step": 13928 }, { "epoch": 0.63, "learning_rate": 2.7732080933706164e-07, "loss": 0.9007, "step": 13929 }, { "epoch": 0.63, "learning_rate": 2.7681686747687566e-07, "loss": 0.9227, "step": 13930 }, { "epoch": 0.63, "learning_rate": 2.763133774833293e-07, "loss": 0.902, "step": 13931 }, { "epoch": 0.63, "learning_rate": 2.758103393798206e-07, "loss": 0.9292, "step": 13932 }, { "epoch": 0.63, "learning_rate": 2.753077531897297e-07, "loss": 0.9353, "step": 13933 }, { "epoch": 0.63, "learning_rate": 2.748056189364112e-07, "loss": 0.9153, "step": 13934 }, { "epoch": 0.63, "learning_rate": 2.74303936643201e-07, "loss": 0.9222, "step": 13935 }, { "epoch": 0.63, "learning_rate": 2.7380270633341477e-07, "loss": 0.9062, "step": 13936 }, { "epoch": 0.63, "learning_rate": 2.7330192803034617e-07, "loss": 0.9159, "step": 13937 }, { "epoch": 0.63, "learning_rate": 2.7280160175726656e-07, "loss": 0.951, "step": 13938 }, { "epoch": 0.63, "learning_rate": 2.7230172753742847e-07, "loss": 0.9079, "step": 13939 }, { "epoch": 0.63, "learning_rate": 2.7180230539406216e-07, "loss": 0.9329, "step": 13940 }, { "epoch": 0.63, "learning_rate": 2.7130333535038024e-07, "loss": 0.9542, "step": 13941 }, { "epoch": 0.63, "learning_rate": 2.708048174295663e-07, "loss": 0.9014, "step": 13942 }, { "epoch": 0.63, "learning_rate": 2.7030675165479195e-07, "loss": 0.9484, "step": 13943 }, { "epoch": 0.63, "learning_rate": 2.6980913804920293e-07, "loss": 0.9069, "step": 13944 }, { "epoch": 0.63, "learning_rate": 2.693119766359231e-07, "loss": 0.9334, "step": 13945 }, { "epoch": 0.63, "learning_rate": 2.6881526743806064e-07, "loss": 0.9112, "step": 13946 }, { "epoch": 0.63, "learning_rate": 2.683190104786959e-07, "loss": 0.9453, "step": 13947 }, { "epoch": 0.63, "learning_rate": 2.678232057808927e-07, "loss": 0.9121, "step": 13948 }, { "epoch": 0.63, "learning_rate": 2.673278533676937e-07, "loss": 0.9195, "step": 13949 }, { "epoch": 0.63, "learning_rate": 2.6683295326211835e-07, "loss": 0.9302, "step": 13950 }, { "epoch": 0.63, "learning_rate": 2.663385054871659e-07, "loss": 0.9445, "step": 13951 }, { "epoch": 0.63, "learning_rate": 2.658445100658147e-07, "loss": 0.891, "step": 13952 }, { "epoch": 0.63, "learning_rate": 2.6535096702102414e-07, "loss": 0.9283, "step": 13953 }, { "epoch": 0.63, "learning_rate": 2.64857876375727e-07, "loss": 0.9701, "step": 13954 }, { "epoch": 0.63, "learning_rate": 2.643652381528439e-07, "loss": 0.9591, "step": 13955 }, { "epoch": 0.63, "learning_rate": 2.638730523752653e-07, "loss": 0.9347, "step": 13956 }, { "epoch": 0.63, "learning_rate": 2.6338131906586516e-07, "loss": 0.9391, "step": 13957 }, { "epoch": 0.63, "learning_rate": 2.628900382474986e-07, "loss": 0.9407, "step": 13958 }, { "epoch": 0.63, "learning_rate": 2.6239920994299284e-07, "loss": 0.9358, "step": 13959 }, { "epoch": 0.63, "learning_rate": 2.619088341751608e-07, "loss": 0.9583, "step": 13960 }, { "epoch": 0.63, "learning_rate": 2.6141891096679196e-07, "loss": 0.9071, "step": 13961 }, { "epoch": 0.63, "learning_rate": 2.609294403406537e-07, "loss": 0.9202, "step": 13962 }, { "epoch": 0.63, "learning_rate": 2.6044042231949116e-07, "loss": 0.9327, "step": 13963 }, { "epoch": 0.63, "learning_rate": 2.599518569260362e-07, "loss": 0.9311, "step": 13964 }, { "epoch": 0.63, "learning_rate": 2.5946374418298945e-07, "loss": 0.9132, "step": 13965 }, { "epoch": 0.63, "learning_rate": 2.58976084113034e-07, "loss": 0.9307, "step": 13966 }, { "epoch": 0.63, "learning_rate": 2.584888767388383e-07, "loss": 0.8742, "step": 13967 }, { "epoch": 0.63, "learning_rate": 2.5800212208303864e-07, "loss": 0.9407, "step": 13968 }, { "epoch": 0.63, "learning_rate": 2.5751582016825925e-07, "loss": 0.8897, "step": 13969 }, { "epoch": 0.63, "learning_rate": 2.570299710170998e-07, "loss": 0.9139, "step": 13970 }, { "epoch": 0.63, "learning_rate": 2.565445746521389e-07, "loss": 0.8831, "step": 13971 }, { "epoch": 0.63, "learning_rate": 2.56059631095934e-07, "loss": 0.9442, "step": 13972 }, { "epoch": 0.63, "learning_rate": 2.5557514037102273e-07, "loss": 0.9564, "step": 13973 }, { "epoch": 0.63, "learning_rate": 2.550911024999203e-07, "loss": 0.951, "step": 13974 }, { "epoch": 0.63, "learning_rate": 2.54607517505121e-07, "loss": 0.9321, "step": 13975 }, { "epoch": 0.63, "learning_rate": 2.541243854090991e-07, "loss": 0.9385, "step": 13976 }, { "epoch": 0.63, "learning_rate": 2.536417062343077e-07, "loss": 0.9336, "step": 13977 }, { "epoch": 0.63, "learning_rate": 2.531594800031778e-07, "loss": 0.8963, "step": 13978 }, { "epoch": 0.63, "learning_rate": 2.526777067381203e-07, "loss": 0.9261, "step": 13979 }, { "epoch": 0.63, "learning_rate": 2.5219638646152513e-07, "loss": 0.9377, "step": 13980 }, { "epoch": 0.63, "learning_rate": 2.517155191957599e-07, "loss": 0.8746, "step": 13981 }, { "epoch": 0.63, "learning_rate": 2.5123510496317116e-07, "loss": 0.8975, "step": 13982 }, { "epoch": 0.63, "learning_rate": 2.507551437860878e-07, "loss": 0.962, "step": 13983 }, { "epoch": 0.63, "learning_rate": 2.50275635686813e-07, "loss": 0.9237, "step": 13984 }, { "epoch": 0.63, "learning_rate": 2.4979658068763125e-07, "loss": 0.9651, "step": 13985 }, { "epoch": 0.63, "learning_rate": 2.4931797881080687e-07, "loss": 0.9402, "step": 13986 }, { "epoch": 0.63, "learning_rate": 2.48839830078581e-07, "loss": 0.9052, "step": 13987 }, { "epoch": 0.63, "learning_rate": 2.483621345131748e-07, "loss": 0.9146, "step": 13988 }, { "epoch": 0.63, "learning_rate": 2.4788489213678934e-07, "loss": 0.9122, "step": 13989 }, { "epoch": 0.63, "learning_rate": 2.4740810297160244e-07, "loss": 0.9308, "step": 13990 }, { "epoch": 0.63, "learning_rate": 2.4693176703977085e-07, "loss": 0.8952, "step": 13991 }, { "epoch": 0.63, "learning_rate": 2.4645588436343346e-07, "loss": 0.8778, "step": 13992 }, { "epoch": 0.63, "learning_rate": 2.4598045496470604e-07, "loss": 0.9032, "step": 13993 }, { "epoch": 0.63, "learning_rate": 2.455054788656819e-07, "loss": 0.9294, "step": 13994 }, { "epoch": 0.63, "learning_rate": 2.450309560884345e-07, "loss": 0.9258, "step": 13995 }, { "epoch": 0.63, "learning_rate": 2.445568866550196e-07, "loss": 0.9633, "step": 13996 }, { "epoch": 0.63, "learning_rate": 2.4408327058746404e-07, "loss": 0.9414, "step": 13997 }, { "epoch": 0.63, "learning_rate": 2.4361010790778126e-07, "loss": 0.9138, "step": 13998 }, { "epoch": 0.63, "learning_rate": 2.4313739863796036e-07, "loss": 0.8619, "step": 13999 }, { "epoch": 0.63, "learning_rate": 2.4266514279996713e-07, "loss": 0.9324, "step": 14000 }, { "epoch": 0.63, "learning_rate": 2.421933404157528e-07, "loss": 0.9348, "step": 14001 }, { "epoch": 0.63, "learning_rate": 2.4172199150724107e-07, "loss": 0.9146, "step": 14002 }, { "epoch": 0.63, "learning_rate": 2.412510960963366e-07, "loss": 0.9346, "step": 14003 }, { "epoch": 0.63, "learning_rate": 2.407806542049251e-07, "loss": 0.9329, "step": 14004 }, { "epoch": 0.63, "learning_rate": 2.4031066585486704e-07, "loss": 0.9312, "step": 14005 }, { "epoch": 0.63, "learning_rate": 2.398411310680071e-07, "loss": 0.9124, "step": 14006 }, { "epoch": 0.63, "learning_rate": 2.3937204986616337e-07, "loss": 0.9067, "step": 14007 }, { "epoch": 0.63, "learning_rate": 2.389034222711384e-07, "loss": 0.9291, "step": 14008 }, { "epoch": 0.63, "learning_rate": 2.38435248304707e-07, "loss": 0.9251, "step": 14009 }, { "epoch": 0.63, "learning_rate": 2.3796752798863066e-07, "loss": 0.9239, "step": 14010 }, { "epoch": 0.63, "learning_rate": 2.375002613446431e-07, "loss": 0.8838, "step": 14011 }, { "epoch": 0.63, "learning_rate": 2.370334483944603e-07, "loss": 0.917, "step": 14012 }, { "epoch": 0.63, "learning_rate": 2.365670891597771e-07, "loss": 0.9297, "step": 14013 }, { "epoch": 0.63, "learning_rate": 2.361011836622662e-07, "loss": 0.9586, "step": 14014 }, { "epoch": 0.63, "learning_rate": 2.3563573192357914e-07, "loss": 0.8981, "step": 14015 }, { "epoch": 0.63, "learning_rate": 2.3517073396534862e-07, "loss": 0.8975, "step": 14016 }, { "epoch": 0.63, "learning_rate": 2.3470618980918403e-07, "loss": 0.8868, "step": 14017 }, { "epoch": 0.63, "learning_rate": 2.3424209947667253e-07, "loss": 0.8977, "step": 14018 }, { "epoch": 0.63, "learning_rate": 2.3377846298938244e-07, "loss": 0.9243, "step": 14019 }, { "epoch": 0.63, "learning_rate": 2.3331528036886207e-07, "loss": 0.9305, "step": 14020 }, { "epoch": 0.63, "learning_rate": 2.3285255163663535e-07, "loss": 0.8891, "step": 14021 }, { "epoch": 0.63, "learning_rate": 2.323902768142061e-07, "loss": 0.9421, "step": 14022 }, { "epoch": 0.63, "learning_rate": 2.3192845592305947e-07, "loss": 0.9135, "step": 14023 }, { "epoch": 0.63, "learning_rate": 2.3146708898465708e-07, "loss": 0.93, "step": 14024 }, { "epoch": 0.63, "learning_rate": 2.3100617602043962e-07, "loss": 0.9411, "step": 14025 }, { "epoch": 0.63, "learning_rate": 2.3054571705182882e-07, "loss": 0.912, "step": 14026 }, { "epoch": 0.63, "learning_rate": 2.300857121002209e-07, "loss": 0.9085, "step": 14027 }, { "epoch": 0.63, "learning_rate": 2.2962616118699544e-07, "loss": 0.963, "step": 14028 }, { "epoch": 0.63, "learning_rate": 2.2916706433351088e-07, "loss": 0.9352, "step": 14029 }, { "epoch": 0.63, "learning_rate": 2.2870842156109906e-07, "loss": 0.9108, "step": 14030 }, { "epoch": 0.63, "learning_rate": 2.282502328910763e-07, "loss": 0.938, "step": 14031 }, { "epoch": 0.63, "learning_rate": 2.277924983447355e-07, "loss": 0.9505, "step": 14032 }, { "epoch": 0.63, "learning_rate": 2.273352179433519e-07, "loss": 0.9321, "step": 14033 }, { "epoch": 0.63, "learning_rate": 2.2687839170817184e-07, "loss": 0.9553, "step": 14034 }, { "epoch": 0.63, "learning_rate": 2.2642201966043053e-07, "loss": 0.9149, "step": 14035 }, { "epoch": 0.63, "learning_rate": 2.2596610182133328e-07, "loss": 0.884, "step": 14036 }, { "epoch": 0.63, "learning_rate": 2.2551063821206976e-07, "loss": 0.8885, "step": 14037 }, { "epoch": 0.63, "learning_rate": 2.250556288538075e-07, "loss": 0.9192, "step": 14038 }, { "epoch": 0.63, "learning_rate": 2.2460107376768959e-07, "loss": 0.9359, "step": 14039 }, { "epoch": 0.63, "learning_rate": 2.2414697297484244e-07, "loss": 0.913, "step": 14040 }, { "epoch": 0.63, "learning_rate": 2.236933264963692e-07, "loss": 0.9087, "step": 14041 }, { "epoch": 0.63, "learning_rate": 2.2324013435335302e-07, "loss": 0.8985, "step": 14042 }, { "epoch": 0.63, "learning_rate": 2.2278739656685256e-07, "loss": 0.9121, "step": 14043 }, { "epoch": 0.63, "learning_rate": 2.22335113157911e-07, "loss": 0.9191, "step": 14044 }, { "epoch": 0.63, "learning_rate": 2.218832841475471e-07, "loss": 0.9255, "step": 14045 }, { "epoch": 0.63, "learning_rate": 2.214319095567552e-07, "loss": 0.9365, "step": 14046 }, { "epoch": 0.63, "learning_rate": 2.2098098940651734e-07, "loss": 0.9175, "step": 14047 }, { "epoch": 0.63, "learning_rate": 2.205305237177846e-07, "loss": 0.9133, "step": 14048 }, { "epoch": 0.63, "learning_rate": 2.2008051251149464e-07, "loss": 0.9285, "step": 14049 }, { "epoch": 0.63, "learning_rate": 2.1963095580855964e-07, "loss": 0.9031, "step": 14050 }, { "epoch": 0.63, "learning_rate": 2.1918185362987176e-07, "loss": 0.8967, "step": 14051 }, { "epoch": 0.63, "learning_rate": 2.18733205996301e-07, "loss": 0.9224, "step": 14052 }, { "epoch": 0.64, "learning_rate": 2.1828501292870062e-07, "loss": 0.9219, "step": 14053 }, { "epoch": 0.64, "learning_rate": 2.1783727444789737e-07, "loss": 0.9645, "step": 14054 }, { "epoch": 0.64, "learning_rate": 2.1738999057469789e-07, "loss": 0.8867, "step": 14055 }, { "epoch": 0.64, "learning_rate": 2.1694316132989224e-07, "loss": 0.9305, "step": 14056 }, { "epoch": 0.64, "learning_rate": 2.164967867342438e-07, "loss": 0.9273, "step": 14057 }, { "epoch": 0.64, "learning_rate": 2.1605086680849596e-07, "loss": 0.9325, "step": 14058 }, { "epoch": 0.64, "learning_rate": 2.156054015733755e-07, "loss": 0.8856, "step": 14059 }, { "epoch": 0.64, "learning_rate": 2.151603910495803e-07, "loss": 0.9346, "step": 14060 }, { "epoch": 0.64, "learning_rate": 2.1471583525779494e-07, "loss": 0.9005, "step": 14061 }, { "epoch": 0.64, "learning_rate": 2.1427173421867732e-07, "loss": 0.9119, "step": 14062 }, { "epoch": 0.64, "learning_rate": 2.1382808795286646e-07, "loss": 0.8998, "step": 14063 }, { "epoch": 0.64, "learning_rate": 2.1338489648098037e-07, "loss": 0.9345, "step": 14064 }, { "epoch": 0.64, "learning_rate": 2.1294215982361587e-07, "loss": 0.9333, "step": 14065 }, { "epoch": 0.64, "learning_rate": 2.1249987800134875e-07, "loss": 0.9101, "step": 14066 }, { "epoch": 0.64, "learning_rate": 2.1205805103473031e-07, "loss": 0.8947, "step": 14067 }, { "epoch": 0.64, "learning_rate": 2.116166789442964e-07, "loss": 0.9262, "step": 14068 }, { "epoch": 0.64, "learning_rate": 2.1117576175055832e-07, "loss": 0.9322, "step": 14069 }, { "epoch": 0.64, "learning_rate": 2.107352994740075e-07, "loss": 0.9697, "step": 14070 }, { "epoch": 0.64, "learning_rate": 2.1029529213511202e-07, "loss": 0.8989, "step": 14071 }, { "epoch": 0.64, "learning_rate": 2.0985573975432215e-07, "loss": 0.9237, "step": 14072 }, { "epoch": 0.64, "learning_rate": 2.0941664235206382e-07, "loss": 0.9453, "step": 14073 }, { "epoch": 0.64, "learning_rate": 2.089779999487429e-07, "loss": 0.92, "step": 14074 }, { "epoch": 0.64, "learning_rate": 2.0853981256474754e-07, "loss": 0.9291, "step": 14075 }, { "epoch": 0.64, "learning_rate": 2.081020802204381e-07, "loss": 0.9031, "step": 14076 }, { "epoch": 0.64, "learning_rate": 2.076648029361583e-07, "loss": 0.9019, "step": 14077 }, { "epoch": 0.64, "learning_rate": 2.0722798073223083e-07, "loss": 0.9283, "step": 14078 }, { "epoch": 0.64, "learning_rate": 2.0679161362895605e-07, "loss": 0.9091, "step": 14079 }, { "epoch": 0.64, "learning_rate": 2.063557016466111e-07, "loss": 0.883, "step": 14080 }, { "epoch": 0.64, "learning_rate": 2.0592024480545868e-07, "loss": 0.9236, "step": 14081 }, { "epoch": 0.64, "learning_rate": 2.0548524312573148e-07, "loss": 0.9457, "step": 14082 }, { "epoch": 0.64, "learning_rate": 2.0505069662764666e-07, "loss": 0.8812, "step": 14083 }, { "epoch": 0.64, "learning_rate": 2.0461660533140027e-07, "loss": 0.9198, "step": 14084 }, { "epoch": 0.64, "learning_rate": 2.0418296925716396e-07, "loss": 0.9603, "step": 14085 }, { "epoch": 0.64, "learning_rate": 2.0374978842509164e-07, "loss": 0.9126, "step": 14086 }, { "epoch": 0.64, "learning_rate": 2.0331706285531384e-07, "loss": 0.9074, "step": 14087 }, { "epoch": 0.64, "learning_rate": 2.0288479256794114e-07, "loss": 0.9697, "step": 14088 }, { "epoch": 0.64, "learning_rate": 2.0245297758306192e-07, "loss": 0.9359, "step": 14089 }, { "epoch": 0.64, "learning_rate": 2.020216179207435e-07, "loss": 0.9608, "step": 14090 }, { "epoch": 0.64, "learning_rate": 2.015907136010342e-07, "loss": 0.9332, "step": 14091 }, { "epoch": 0.64, "learning_rate": 2.01160264643957e-07, "loss": 0.931, "step": 14092 }, { "epoch": 0.64, "learning_rate": 2.0073027106952026e-07, "loss": 0.909, "step": 14093 }, { "epoch": 0.64, "learning_rate": 2.0030073289770245e-07, "loss": 0.9437, "step": 14094 }, { "epoch": 0.64, "learning_rate": 1.9987165014846876e-07, "loss": 0.9523, "step": 14095 }, { "epoch": 0.64, "learning_rate": 1.9944302284175987e-07, "loss": 0.9103, "step": 14096 }, { "epoch": 0.64, "learning_rate": 1.9901485099749207e-07, "loss": 0.8982, "step": 14097 }, { "epoch": 0.64, "learning_rate": 1.9858713463556721e-07, "loss": 0.9177, "step": 14098 }, { "epoch": 0.64, "learning_rate": 1.9815987377586276e-07, "loss": 0.8904, "step": 14099 }, { "epoch": 0.64, "learning_rate": 1.9773306843823282e-07, "loss": 0.8927, "step": 14100 }, { "epoch": 0.64, "learning_rate": 1.973067186425126e-07, "loss": 0.9602, "step": 14101 }, { "epoch": 0.64, "learning_rate": 1.9688082440851854e-07, "loss": 0.9277, "step": 14102 }, { "epoch": 0.64, "learning_rate": 1.9645538575603918e-07, "loss": 0.9393, "step": 14103 }, { "epoch": 0.64, "learning_rate": 1.9603040270484986e-07, "loss": 0.9046, "step": 14104 }, { "epoch": 0.64, "learning_rate": 1.956058752746992e-07, "loss": 0.9618, "step": 14105 }, { "epoch": 0.64, "learning_rate": 1.951818034853148e-07, "loss": 0.9096, "step": 14106 }, { "epoch": 0.64, "learning_rate": 1.947581873564064e-07, "loss": 0.926, "step": 14107 }, { "epoch": 0.64, "learning_rate": 1.9433502690766048e-07, "loss": 0.9663, "step": 14108 }, { "epoch": 0.64, "learning_rate": 1.9391232215874356e-07, "loss": 0.9276, "step": 14109 }, { "epoch": 0.64, "learning_rate": 1.934900731292988e-07, "loss": 0.9593, "step": 14110 }, { "epoch": 0.64, "learning_rate": 1.930682798389494e-07, "loss": 0.9111, "step": 14111 }, { "epoch": 0.64, "learning_rate": 1.926469423072974e-07, "loss": 0.941, "step": 14112 }, { "epoch": 0.64, "learning_rate": 1.922260605539239e-07, "loss": 0.8895, "step": 14113 }, { "epoch": 0.64, "learning_rate": 1.9180563459838874e-07, "loss": 0.9362, "step": 14114 }, { "epoch": 0.64, "learning_rate": 1.9138566446023076e-07, "loss": 0.914, "step": 14115 }, { "epoch": 0.64, "learning_rate": 1.9096615015896657e-07, "loss": 0.9029, "step": 14116 }, { "epoch": 0.64, "learning_rate": 1.905470917140917e-07, "loss": 0.8989, "step": 14117 }, { "epoch": 0.64, "learning_rate": 1.9012848914508387e-07, "loss": 0.9679, "step": 14118 }, { "epoch": 0.64, "learning_rate": 1.8971034247139418e-07, "loss": 0.8979, "step": 14119 }, { "epoch": 0.64, "learning_rate": 1.8929265171245604e-07, "loss": 0.9154, "step": 14120 }, { "epoch": 0.64, "learning_rate": 1.8887541688768053e-07, "loss": 0.912, "step": 14121 }, { "epoch": 0.64, "learning_rate": 1.884586380164577e-07, "loss": 0.9174, "step": 14122 }, { "epoch": 0.64, "learning_rate": 1.8804231511815762e-07, "loss": 0.9271, "step": 14123 }, { "epoch": 0.64, "learning_rate": 1.8762644821212707e-07, "loss": 0.9379, "step": 14124 }, { "epoch": 0.64, "learning_rate": 1.8721103731769498e-07, "loss": 0.9054, "step": 14125 }, { "epoch": 0.64, "learning_rate": 1.8679608245416147e-07, "loss": 0.9668, "step": 14126 }, { "epoch": 0.64, "learning_rate": 1.8638158364081672e-07, "loss": 0.9161, "step": 14127 }, { "epoch": 0.64, "learning_rate": 1.8596754089692082e-07, "loss": 0.9462, "step": 14128 }, { "epoch": 0.64, "learning_rate": 1.855539542417162e-07, "loss": 0.9232, "step": 14129 }, { "epoch": 0.64, "learning_rate": 1.85140823694423e-07, "loss": 0.8734, "step": 14130 }, { "epoch": 0.64, "learning_rate": 1.8472814927424142e-07, "loss": 0.9472, "step": 14131 }, { "epoch": 0.64, "learning_rate": 1.843159310003495e-07, "loss": 0.9611, "step": 14132 }, { "epoch": 0.64, "learning_rate": 1.8390416889190299e-07, "loss": 0.9701, "step": 14133 }, { "epoch": 0.64, "learning_rate": 1.8349286296804104e-07, "loss": 0.938, "step": 14134 }, { "epoch": 0.64, "learning_rate": 1.83082013247875e-07, "loss": 0.9383, "step": 14135 }, { "epoch": 0.64, "learning_rate": 1.8267161975049962e-07, "loss": 0.9183, "step": 14136 }, { "epoch": 0.64, "learning_rate": 1.8226168249498854e-07, "loss": 0.9231, "step": 14137 }, { "epoch": 0.64, "learning_rate": 1.8185220150038873e-07, "loss": 0.9527, "step": 14138 }, { "epoch": 0.64, "learning_rate": 1.8144317678573497e-07, "loss": 0.954, "step": 14139 }, { "epoch": 0.64, "learning_rate": 1.8103460837003316e-07, "loss": 0.9418, "step": 14140 }, { "epoch": 0.64, "learning_rate": 1.8062649627227037e-07, "loss": 0.9067, "step": 14141 }, { "epoch": 0.64, "learning_rate": 1.802188405114147e-07, "loss": 0.9635, "step": 14142 }, { "epoch": 0.64, "learning_rate": 1.798116411064099e-07, "loss": 0.9404, "step": 14143 }, { "epoch": 0.64, "learning_rate": 1.7940489807618088e-07, "loss": 0.9001, "step": 14144 }, { "epoch": 0.64, "learning_rate": 1.7899861143962804e-07, "loss": 0.9389, "step": 14145 }, { "epoch": 0.64, "learning_rate": 1.785927812156363e-07, "loss": 0.9148, "step": 14146 }, { "epoch": 0.64, "learning_rate": 1.7818740742306162e-07, "loss": 0.8937, "step": 14147 }, { "epoch": 0.64, "learning_rate": 1.7778249008074678e-07, "loss": 0.9263, "step": 14148 }, { "epoch": 0.64, "learning_rate": 1.7737802920750668e-07, "loss": 0.9577, "step": 14149 }, { "epoch": 0.64, "learning_rate": 1.7697402482213964e-07, "loss": 0.9071, "step": 14150 }, { "epoch": 0.64, "learning_rate": 1.7657047694342178e-07, "loss": 0.9308, "step": 14151 }, { "epoch": 0.64, "learning_rate": 1.761673855901047e-07, "loss": 0.9327, "step": 14152 }, { "epoch": 0.64, "learning_rate": 1.7576475078092238e-07, "loss": 0.9083, "step": 14153 }, { "epoch": 0.64, "learning_rate": 1.7536257253458644e-07, "loss": 0.9276, "step": 14154 }, { "epoch": 0.64, "learning_rate": 1.7496085086978974e-07, "loss": 0.9256, "step": 14155 }, { "epoch": 0.64, "learning_rate": 1.7455958580519627e-07, "loss": 0.8732, "step": 14156 }, { "epoch": 0.64, "learning_rate": 1.7415877735945995e-07, "loss": 0.9856, "step": 14157 }, { "epoch": 0.64, "learning_rate": 1.737584255512048e-07, "loss": 0.8982, "step": 14158 }, { "epoch": 0.64, "learning_rate": 1.7335853039903373e-07, "loss": 0.8878, "step": 14159 }, { "epoch": 0.64, "learning_rate": 1.7295909192153626e-07, "loss": 0.9533, "step": 14160 }, { "epoch": 0.64, "learning_rate": 1.7256011013727313e-07, "loss": 0.9195, "step": 14161 }, { "epoch": 0.64, "learning_rate": 1.7216158506478508e-07, "loss": 0.9598, "step": 14162 }, { "epoch": 0.64, "learning_rate": 1.7176351672259395e-07, "loss": 0.9039, "step": 14163 }, { "epoch": 0.64, "learning_rate": 1.713659051292016e-07, "loss": 0.9454, "step": 14164 }, { "epoch": 0.64, "learning_rate": 1.7096875030308212e-07, "loss": 0.8878, "step": 14165 }, { "epoch": 0.64, "learning_rate": 1.7057205226269414e-07, "loss": 0.908, "step": 14166 }, { "epoch": 0.64, "learning_rate": 1.7017581102647508e-07, "loss": 0.9423, "step": 14167 }, { "epoch": 0.64, "learning_rate": 1.6978002661283688e-07, "loss": 0.9089, "step": 14168 }, { "epoch": 0.64, "learning_rate": 1.6938469904017374e-07, "loss": 0.9246, "step": 14169 }, { "epoch": 0.64, "learning_rate": 1.6898982832685873e-07, "loss": 0.9341, "step": 14170 }, { "epoch": 0.64, "learning_rate": 1.6859541449124271e-07, "loss": 0.9291, "step": 14171 }, { "epoch": 0.64, "learning_rate": 1.682014575516544e-07, "loss": 0.918, "step": 14172 }, { "epoch": 0.64, "learning_rate": 1.6780795752640243e-07, "loss": 0.9406, "step": 14173 }, { "epoch": 0.64, "learning_rate": 1.6741491443377444e-07, "loss": 0.8955, "step": 14174 }, { "epoch": 0.64, "learning_rate": 1.670223282920358e-07, "loss": 0.9353, "step": 14175 }, { "epoch": 0.64, "learning_rate": 1.666301991194319e-07, "loss": 0.9284, "step": 14176 }, { "epoch": 0.64, "learning_rate": 1.6623852693418484e-07, "loss": 0.9219, "step": 14177 }, { "epoch": 0.64, "learning_rate": 1.6584731175449786e-07, "loss": 0.94, "step": 14178 }, { "epoch": 0.64, "learning_rate": 1.6545655359855196e-07, "loss": 0.9147, "step": 14179 }, { "epoch": 0.64, "learning_rate": 1.6506625248450702e-07, "loss": 0.9425, "step": 14180 }, { "epoch": 0.64, "learning_rate": 1.646764084305019e-07, "loss": 0.9466, "step": 14181 }, { "epoch": 0.64, "learning_rate": 1.6428702145465215e-07, "loss": 0.9329, "step": 14182 }, { "epoch": 0.64, "learning_rate": 1.6389809157505653e-07, "loss": 0.9046, "step": 14183 }, { "epoch": 0.64, "learning_rate": 1.635096188097851e-07, "loss": 0.9059, "step": 14184 }, { "epoch": 0.64, "learning_rate": 1.6312160317689784e-07, "loss": 0.9271, "step": 14185 }, { "epoch": 0.64, "learning_rate": 1.6273404469442144e-07, "loss": 0.8766, "step": 14186 }, { "epoch": 0.64, "learning_rate": 1.6234694338037038e-07, "loss": 0.933, "step": 14187 }, { "epoch": 0.64, "learning_rate": 1.619602992527336e-07, "loss": 0.9563, "step": 14188 }, { "epoch": 0.64, "learning_rate": 1.6157411232947895e-07, "loss": 0.9404, "step": 14189 }, { "epoch": 0.64, "learning_rate": 1.611883826285543e-07, "loss": 0.9785, "step": 14190 }, { "epoch": 0.64, "learning_rate": 1.608031101678853e-07, "loss": 0.9428, "step": 14191 }, { "epoch": 0.64, "learning_rate": 1.6041829496537765e-07, "loss": 0.9161, "step": 14192 }, { "epoch": 0.64, "learning_rate": 1.6003393703891256e-07, "loss": 0.9071, "step": 14193 }, { "epoch": 0.64, "learning_rate": 1.596500364063558e-07, "loss": 0.9195, "step": 14194 }, { "epoch": 0.64, "learning_rate": 1.5926659308554638e-07, "loss": 0.9048, "step": 14195 }, { "epoch": 0.64, "learning_rate": 1.588836070943034e-07, "loss": 0.9223, "step": 14196 }, { "epoch": 0.64, "learning_rate": 1.5850107845042818e-07, "loss": 0.9041, "step": 14197 }, { "epoch": 0.64, "learning_rate": 1.5811900717169537e-07, "loss": 0.9593, "step": 14198 }, { "epoch": 0.64, "learning_rate": 1.577373932758619e-07, "loss": 0.9507, "step": 14199 }, { "epoch": 0.64, "learning_rate": 1.5735623678066247e-07, "loss": 0.9214, "step": 14200 }, { "epoch": 0.64, "learning_rate": 1.569755377038118e-07, "loss": 0.9123, "step": 14201 }, { "epoch": 0.64, "learning_rate": 1.5659529606299905e-07, "loss": 0.9233, "step": 14202 }, { "epoch": 0.65, "learning_rate": 1.56215511875899e-07, "loss": 0.9108, "step": 14203 }, { "epoch": 0.65, "learning_rate": 1.5583618516015975e-07, "loss": 0.9117, "step": 14204 }, { "epoch": 0.65, "learning_rate": 1.554573159334083e-07, "loss": 0.9285, "step": 14205 }, { "epoch": 0.65, "learning_rate": 1.55078904213255e-07, "loss": 0.9244, "step": 14206 }, { "epoch": 0.65, "learning_rate": 1.5470095001728358e-07, "loss": 0.945, "step": 14207 }, { "epoch": 0.65, "learning_rate": 1.5432345336305886e-07, "loss": 0.9423, "step": 14208 }, { "epoch": 0.65, "learning_rate": 1.5394641426812461e-07, "loss": 0.9193, "step": 14209 }, { "epoch": 0.65, "learning_rate": 1.535698327500035e-07, "loss": 0.9351, "step": 14210 }, { "epoch": 0.65, "learning_rate": 1.531937088261959e-07, "loss": 0.9369, "step": 14211 }, { "epoch": 0.65, "learning_rate": 1.528180425141823e-07, "loss": 0.9149, "step": 14212 }, { "epoch": 0.65, "learning_rate": 1.524428338314199e-07, "loss": 0.8944, "step": 14213 }, { "epoch": 0.65, "learning_rate": 1.520680827953469e-07, "loss": 0.9276, "step": 14214 }, { "epoch": 0.65, "learning_rate": 1.516937894233772e-07, "loss": 0.9319, "step": 14215 }, { "epoch": 0.65, "learning_rate": 1.5131995373290799e-07, "loss": 0.9171, "step": 14216 }, { "epoch": 0.65, "learning_rate": 1.5094657574131088e-07, "loss": 0.933, "step": 14217 }, { "epoch": 0.65, "learning_rate": 1.5057365546593872e-07, "loss": 0.9232, "step": 14218 }, { "epoch": 0.65, "learning_rate": 1.5020119292412204e-07, "loss": 0.8721, "step": 14219 }, { "epoch": 0.65, "learning_rate": 1.4982918813317037e-07, "loss": 0.9204, "step": 14220 }, { "epoch": 0.65, "learning_rate": 1.4945764111037097e-07, "loss": 0.9123, "step": 14221 }, { "epoch": 0.65, "learning_rate": 1.4908655187299337e-07, "loss": 0.9218, "step": 14222 }, { "epoch": 0.65, "learning_rate": 1.4871592043827933e-07, "loss": 0.9047, "step": 14223 }, { "epoch": 0.65, "learning_rate": 1.4834574682345616e-07, "loss": 0.9531, "step": 14224 }, { "epoch": 0.65, "learning_rate": 1.4797603104572676e-07, "loss": 0.9325, "step": 14225 }, { "epoch": 0.65, "learning_rate": 1.47606773122273e-07, "loss": 0.9096, "step": 14226 }, { "epoch": 0.65, "learning_rate": 1.4723797307025333e-07, "loss": 0.9224, "step": 14227 }, { "epoch": 0.65, "learning_rate": 1.4686963090680962e-07, "loss": 0.9037, "step": 14228 }, { "epoch": 0.65, "learning_rate": 1.4650174664905814e-07, "loss": 0.925, "step": 14229 }, { "epoch": 0.65, "learning_rate": 1.4613432031409747e-07, "loss": 0.8986, "step": 14230 }, { "epoch": 0.65, "learning_rate": 1.457673519190017e-07, "loss": 0.9267, "step": 14231 }, { "epoch": 0.65, "learning_rate": 1.45400841480825e-07, "loss": 0.9193, "step": 14232 }, { "epoch": 0.65, "learning_rate": 1.450347890166004e-07, "loss": 0.915, "step": 14233 }, { "epoch": 0.65, "learning_rate": 1.4466919454333982e-07, "loss": 0.9284, "step": 14234 }, { "epoch": 0.65, "learning_rate": 1.4430405807803306e-07, "loss": 0.9315, "step": 14235 }, { "epoch": 0.65, "learning_rate": 1.439393796376487e-07, "loss": 0.9317, "step": 14236 }, { "epoch": 0.65, "learning_rate": 1.4357515923913657e-07, "loss": 0.8974, "step": 14237 }, { "epoch": 0.65, "learning_rate": 1.43211396899422e-07, "loss": 0.9329, "step": 14238 }, { "epoch": 0.65, "learning_rate": 1.4284809263540812e-07, "loss": 0.9287, "step": 14239 }, { "epoch": 0.65, "learning_rate": 1.4248524646398253e-07, "loss": 0.9091, "step": 14240 }, { "epoch": 0.65, "learning_rate": 1.4212285840200511e-07, "loss": 0.9204, "step": 14241 }, { "epoch": 0.65, "learning_rate": 1.4176092846631683e-07, "loss": 0.9186, "step": 14242 }, { "epoch": 0.65, "learning_rate": 1.4139945667374088e-07, "loss": 0.9214, "step": 14243 }, { "epoch": 0.65, "learning_rate": 1.4103844304107272e-07, "loss": 0.9009, "step": 14244 }, { "epoch": 0.65, "learning_rate": 1.4067788758509116e-07, "loss": 0.9135, "step": 14245 }, { "epoch": 0.65, "learning_rate": 1.4031779032255277e-07, "loss": 0.9448, "step": 14246 }, { "epoch": 0.65, "learning_rate": 1.3995815127019198e-07, "loss": 0.9301, "step": 14247 }, { "epoch": 0.65, "learning_rate": 1.3959897044472094e-07, "loss": 0.934, "step": 14248 }, { "epoch": 0.65, "learning_rate": 1.3924024786283407e-07, "loss": 0.9401, "step": 14249 }, { "epoch": 0.65, "learning_rate": 1.388819835412014e-07, "loss": 0.9182, "step": 14250 }, { "epoch": 0.65, "learning_rate": 1.3852417749647186e-07, "loss": 0.8999, "step": 14251 }, { "epoch": 0.65, "learning_rate": 1.381668297452754e-07, "loss": 0.9202, "step": 14252 }, { "epoch": 0.65, "learning_rate": 1.378099403042177e-07, "loss": 0.8993, "step": 14253 }, { "epoch": 0.65, "learning_rate": 1.3745350918988544e-07, "loss": 0.9251, "step": 14254 }, { "epoch": 0.65, "learning_rate": 1.3709753641884205e-07, "loss": 0.9189, "step": 14255 }, { "epoch": 0.65, "learning_rate": 1.3674202200763208e-07, "loss": 0.9122, "step": 14256 }, { "epoch": 0.65, "learning_rate": 1.3638696597277678e-07, "loss": 0.9397, "step": 14257 }, { "epoch": 0.65, "learning_rate": 1.3603236833077515e-07, "loss": 0.9098, "step": 14258 }, { "epoch": 0.65, "learning_rate": 1.3567822909810957e-07, "loss": 0.8852, "step": 14259 }, { "epoch": 0.65, "learning_rate": 1.3532454829123465e-07, "loss": 0.9099, "step": 14260 }, { "epoch": 0.65, "learning_rate": 1.3497132592659056e-07, "loss": 0.9169, "step": 14261 }, { "epoch": 0.65, "learning_rate": 1.3461856202058975e-07, "loss": 0.9243, "step": 14262 }, { "epoch": 0.65, "learning_rate": 1.3426625658962688e-07, "loss": 0.9055, "step": 14263 }, { "epoch": 0.65, "learning_rate": 1.3391440965007552e-07, "loss": 0.9433, "step": 14264 }, { "epoch": 0.65, "learning_rate": 1.3356302121828702e-07, "loss": 0.9474, "step": 14265 }, { "epoch": 0.65, "learning_rate": 1.3321209131059055e-07, "loss": 0.9226, "step": 14266 }, { "epoch": 0.65, "learning_rate": 1.3286161994329637e-07, "loss": 0.9243, "step": 14267 }, { "epoch": 0.65, "learning_rate": 1.325116071326904e-07, "loss": 0.884, "step": 14268 }, { "epoch": 0.65, "learning_rate": 1.321620528950396e-07, "loss": 0.9536, "step": 14269 }, { "epoch": 0.65, "learning_rate": 1.318129572465887e-07, "loss": 0.9227, "step": 14270 }, { "epoch": 0.65, "learning_rate": 1.314643202035615e-07, "loss": 0.9172, "step": 14271 }, { "epoch": 0.65, "learning_rate": 1.3111614178216047e-07, "loss": 0.9368, "step": 14272 }, { "epoch": 0.65, "learning_rate": 1.307684219985661e-07, "loss": 0.9187, "step": 14273 }, { "epoch": 0.65, "learning_rate": 1.3042116086893764e-07, "loss": 0.926, "step": 14274 }, { "epoch": 0.65, "learning_rate": 1.300743584094133e-07, "loss": 0.931, "step": 14275 }, { "epoch": 0.65, "learning_rate": 1.297280146361113e-07, "loss": 0.9348, "step": 14276 }, { "epoch": 0.65, "learning_rate": 1.2938212956512652e-07, "loss": 0.921, "step": 14277 }, { "epoch": 0.65, "learning_rate": 1.290367032125339e-07, "loss": 0.8828, "step": 14278 }, { "epoch": 0.65, "learning_rate": 1.2869173559438509e-07, "loss": 0.9456, "step": 14279 }, { "epoch": 0.65, "learning_rate": 1.2834722672671385e-07, "loss": 0.9288, "step": 14280 }, { "epoch": 0.65, "learning_rate": 1.2800317662552853e-07, "loss": 0.914, "step": 14281 }, { "epoch": 0.65, "learning_rate": 1.2765958530681966e-07, "loss": 0.9244, "step": 14282 }, { "epoch": 0.65, "learning_rate": 1.2731645278655448e-07, "loss": 0.923, "step": 14283 }, { "epoch": 0.65, "learning_rate": 1.269737790806802e-07, "loss": 0.8958, "step": 14284 }, { "epoch": 0.65, "learning_rate": 1.2663156420511858e-07, "loss": 0.8927, "step": 14285 }, { "epoch": 0.65, "learning_rate": 1.2628980817577908e-07, "loss": 0.9364, "step": 14286 }, { "epoch": 0.65, "learning_rate": 1.25948511008539e-07, "loss": 0.9026, "step": 14287 }, { "epoch": 0.65, "learning_rate": 1.2560767271926234e-07, "loss": 0.8849, "step": 14288 }, { "epoch": 0.65, "learning_rate": 1.2526729332378862e-07, "loss": 0.9278, "step": 14289 }, { "epoch": 0.65, "learning_rate": 1.2492737283793522e-07, "loss": 0.9355, "step": 14290 }, { "epoch": 0.65, "learning_rate": 1.2458791127749948e-07, "loss": 0.9724, "step": 14291 }, { "epoch": 0.65, "learning_rate": 1.2424890865825768e-07, "loss": 0.9142, "step": 14292 }, { "epoch": 0.65, "learning_rate": 1.239103649959661e-07, "loss": 0.9345, "step": 14293 }, { "epoch": 0.65, "learning_rate": 1.2357228030635437e-07, "loss": 0.9198, "step": 14294 }, { "epoch": 0.65, "learning_rate": 1.232346546051366e-07, "loss": 0.9075, "step": 14295 }, { "epoch": 0.65, "learning_rate": 1.228974879080036e-07, "loss": 0.93, "step": 14296 }, { "epoch": 0.65, "learning_rate": 1.2256078023062278e-07, "loss": 0.9231, "step": 14297 }, { "epoch": 0.65, "learning_rate": 1.222245315886439e-07, "loss": 0.8843, "step": 14298 }, { "epoch": 0.65, "learning_rate": 1.2188874199769218e-07, "loss": 0.9294, "step": 14299 }, { "epoch": 0.65, "learning_rate": 1.2155341147337295e-07, "loss": 0.9107, "step": 14300 }, { "epoch": 0.65, "learning_rate": 1.2121854003127044e-07, "loss": 0.8944, "step": 14301 }, { "epoch": 0.65, "learning_rate": 1.208841276869477e-07, "loss": 0.916, "step": 14302 }, { "epoch": 0.65, "learning_rate": 1.2055017445594452e-07, "loss": 0.923, "step": 14303 }, { "epoch": 0.65, "learning_rate": 1.2021668035378187e-07, "loss": 0.8768, "step": 14304 }, { "epoch": 0.65, "learning_rate": 1.1988364539595841e-07, "loss": 0.9256, "step": 14305 }, { "epoch": 0.65, "learning_rate": 1.195510695979496e-07, "loss": 0.9051, "step": 14306 }, { "epoch": 0.65, "learning_rate": 1.1921895297521303e-07, "loss": 0.8926, "step": 14307 }, { "epoch": 0.65, "learning_rate": 1.1888729554318301e-07, "loss": 0.9604, "step": 14308 }, { "epoch": 0.65, "learning_rate": 1.1855609731727169e-07, "loss": 0.886, "step": 14309 }, { "epoch": 0.65, "learning_rate": 1.182253583128723e-07, "loss": 0.9257, "step": 14310 }, { "epoch": 0.65, "learning_rate": 1.1789507854535475e-07, "loss": 0.9105, "step": 14311 }, { "epoch": 0.65, "learning_rate": 1.175652580300668e-07, "loss": 0.9147, "step": 14312 }, { "epoch": 0.65, "learning_rate": 1.1723589678233838e-07, "loss": 0.8994, "step": 14313 }, { "epoch": 0.65, "learning_rate": 1.1690699481747503e-07, "loss": 0.9009, "step": 14314 }, { "epoch": 0.65, "learning_rate": 1.1657855215076119e-07, "loss": 0.934, "step": 14315 }, { "epoch": 0.65, "learning_rate": 1.1625056879746133e-07, "loss": 0.9233, "step": 14316 }, { "epoch": 0.65, "learning_rate": 1.1592304477281768e-07, "loss": 0.9247, "step": 14317 }, { "epoch": 0.65, "learning_rate": 1.1559598009205253e-07, "loss": 0.9261, "step": 14318 }, { "epoch": 0.65, "learning_rate": 1.152693747703626e-07, "loss": 0.9102, "step": 14319 }, { "epoch": 0.65, "learning_rate": 1.1494322882292908e-07, "loss": 0.9226, "step": 14320 }, { "epoch": 0.65, "learning_rate": 1.1461754226490762e-07, "loss": 0.9415, "step": 14321 }, { "epoch": 0.65, "learning_rate": 1.1429231511143501e-07, "loss": 0.9093, "step": 14322 }, { "epoch": 0.65, "learning_rate": 1.1396754737762472e-07, "loss": 0.9493, "step": 14323 }, { "epoch": 0.65, "learning_rate": 1.1364323907856911e-07, "loss": 0.9359, "step": 14324 }, { "epoch": 0.65, "learning_rate": 1.1331939022934057e-07, "loss": 0.9429, "step": 14325 }, { "epoch": 0.65, "learning_rate": 1.129960008449893e-07, "loss": 0.9481, "step": 14326 }, { "epoch": 0.65, "learning_rate": 1.1267307094054436e-07, "loss": 0.8803, "step": 14327 }, { "epoch": 0.65, "learning_rate": 1.1235060053101376e-07, "loss": 0.9268, "step": 14328 }, { "epoch": 0.65, "learning_rate": 1.1202858963138219e-07, "loss": 0.9174, "step": 14329 }, { "epoch": 0.65, "learning_rate": 1.1170703825661655e-07, "loss": 0.9512, "step": 14330 }, { "epoch": 0.65, "learning_rate": 1.1138594642165712e-07, "loss": 0.8903, "step": 14331 }, { "epoch": 0.65, "learning_rate": 1.1106531414142974e-07, "loss": 0.8931, "step": 14332 }, { "epoch": 0.65, "learning_rate": 1.1074514143083359e-07, "loss": 0.924, "step": 14333 }, { "epoch": 0.65, "learning_rate": 1.104254283047479e-07, "loss": 0.9329, "step": 14334 }, { "epoch": 0.65, "learning_rate": 1.1010617477803076e-07, "loss": 0.9044, "step": 14335 }, { "epoch": 0.65, "learning_rate": 1.0978738086551921e-07, "loss": 0.9473, "step": 14336 }, { "epoch": 0.65, "learning_rate": 1.0946904658202806e-07, "loss": 0.9038, "step": 14337 }, { "epoch": 0.65, "learning_rate": 1.0915117194235214e-07, "loss": 0.8927, "step": 14338 }, { "epoch": 0.65, "learning_rate": 1.0883375696126408e-07, "loss": 0.9328, "step": 14339 }, { "epoch": 0.65, "learning_rate": 1.085168016535132e-07, "loss": 0.9628, "step": 14340 }, { "epoch": 0.65, "learning_rate": 1.0820030603383214e-07, "loss": 0.9162, "step": 14341 }, { "epoch": 0.65, "learning_rate": 1.0788427011692803e-07, "loss": 0.9223, "step": 14342 }, { "epoch": 0.65, "learning_rate": 1.0756869391748692e-07, "loss": 0.9146, "step": 14343 }, { "epoch": 0.65, "learning_rate": 1.0725357745017706e-07, "loss": 0.9435, "step": 14344 }, { "epoch": 0.65, "learning_rate": 1.0693892072964119e-07, "loss": 0.8981, "step": 14345 }, { "epoch": 0.65, "learning_rate": 1.0662472377050314e-07, "loss": 0.9145, "step": 14346 }, { "epoch": 0.65, "learning_rate": 1.0631098658736349e-07, "loss": 0.9122, "step": 14347 }, { "epoch": 0.65, "learning_rate": 1.0599770919480502e-07, "loss": 0.8911, "step": 14348 }, { "epoch": 0.65, "learning_rate": 1.0568489160738493e-07, "loss": 0.9499, "step": 14349 }, { "epoch": 0.65, "learning_rate": 1.0537253383964052e-07, "loss": 0.932, "step": 14350 }, { "epoch": 0.65, "learning_rate": 1.0506063590608905e-07, "loss": 0.9181, "step": 14351 }, { "epoch": 0.65, "learning_rate": 1.047491978212234e-07, "loss": 0.9435, "step": 14352 }, { "epoch": 0.66, "learning_rate": 1.0443821959952083e-07, "loss": 0.9324, "step": 14353 }, { "epoch": 0.66, "learning_rate": 1.0412770125542982e-07, "loss": 0.9528, "step": 14354 }, { "epoch": 0.66, "learning_rate": 1.0381764280338213e-07, "loss": 0.9098, "step": 14355 }, { "epoch": 0.66, "learning_rate": 1.0350804425778849e-07, "loss": 0.9254, "step": 14356 }, { "epoch": 0.66, "learning_rate": 1.0319890563303625e-07, "loss": 0.9032, "step": 14357 }, { "epoch": 0.66, "learning_rate": 1.028902269434906e-07, "loss": 0.9225, "step": 14358 }, { "epoch": 0.66, "learning_rate": 1.0258200820349895e-07, "loss": 0.877, "step": 14359 }, { "epoch": 0.66, "learning_rate": 1.0227424942738429e-07, "loss": 0.9463, "step": 14360 }, { "epoch": 0.66, "learning_rate": 1.019669506294485e-07, "loss": 0.9205, "step": 14361 }, { "epoch": 0.66, "learning_rate": 1.0166011182397351e-07, "loss": 0.9579, "step": 14362 }, { "epoch": 0.66, "learning_rate": 1.0135373302521901e-07, "loss": 0.8754, "step": 14363 }, { "epoch": 0.66, "learning_rate": 1.010478142474225e-07, "loss": 0.971, "step": 14364 }, { "epoch": 0.66, "learning_rate": 1.007423555048026e-07, "loss": 0.9287, "step": 14365 }, { "epoch": 0.66, "learning_rate": 1.0043735681155243e-07, "loss": 0.8911, "step": 14366 }, { "epoch": 0.66, "learning_rate": 1.001328181818495e-07, "loss": 0.9255, "step": 14367 }, { "epoch": 0.66, "learning_rate": 9.982873962984363e-08, "loss": 0.9104, "step": 14368 }, { "epoch": 0.66, "learning_rate": 9.952512116966907e-08, "loss": 0.9457, "step": 14369 }, { "epoch": 0.66, "learning_rate": 9.92219628154334e-08, "loss": 0.9194, "step": 14370 }, { "epoch": 0.66, "learning_rate": 9.891926458122647e-08, "loss": 0.9001, "step": 14371 }, { "epoch": 0.66, "learning_rate": 9.861702648111594e-08, "loss": 0.9414, "step": 14372 }, { "epoch": 0.66, "learning_rate": 9.83152485291472e-08, "loss": 0.8774, "step": 14373 }, { "epoch": 0.66, "learning_rate": 9.801393073934462e-08, "loss": 0.9302, "step": 14374 }, { "epoch": 0.66, "learning_rate": 9.771307312571254e-08, "loss": 0.9543, "step": 14375 }, { "epoch": 0.66, "learning_rate": 9.7412675702232e-08, "loss": 0.9891, "step": 14376 }, { "epoch": 0.66, "learning_rate": 9.711273848286185e-08, "loss": 0.9123, "step": 14377 }, { "epoch": 0.66, "learning_rate": 9.681326148154424e-08, "loss": 0.9354, "step": 14378 }, { "epoch": 0.66, "learning_rate": 9.651424471219472e-08, "loss": 0.9024, "step": 14379 }, { "epoch": 0.66, "learning_rate": 9.621568818870997e-08, "loss": 0.8906, "step": 14380 }, { "epoch": 0.66, "learning_rate": 9.591759192496552e-08, "loss": 0.882, "step": 14381 }, { "epoch": 0.66, "learning_rate": 9.561995593481255e-08, "loss": 0.9231, "step": 14382 }, { "epoch": 0.66, "learning_rate": 9.532278023208553e-08, "loss": 0.8815, "step": 14383 }, { "epoch": 0.66, "learning_rate": 9.502606483059451e-08, "loss": 0.9162, "step": 14384 }, { "epoch": 0.66, "learning_rate": 9.47298097441296e-08, "loss": 0.8897, "step": 14385 }, { "epoch": 0.66, "learning_rate": 9.443401498645532e-08, "loss": 0.9146, "step": 14386 }, { "epoch": 0.66, "learning_rate": 9.413868057132291e-08, "loss": 0.9158, "step": 14387 }, { "epoch": 0.66, "learning_rate": 9.384380651245472e-08, "loss": 0.9679, "step": 14388 }, { "epoch": 0.66, "learning_rate": 9.354939282355313e-08, "loss": 0.9313, "step": 14389 }, { "epoch": 0.66, "learning_rate": 9.325543951830495e-08, "loss": 0.9526, "step": 14390 }, { "epoch": 0.66, "learning_rate": 9.296194661036594e-08, "loss": 0.8893, "step": 14391 }, { "epoch": 0.66, "learning_rate": 9.266891411337963e-08, "loss": 0.9201, "step": 14392 }, { "epoch": 0.66, "learning_rate": 9.23763420409618e-08, "loss": 0.9046, "step": 14393 }, { "epoch": 0.66, "learning_rate": 9.208423040671045e-08, "loss": 0.8927, "step": 14394 }, { "epoch": 0.66, "learning_rate": 9.179257922420027e-08, "loss": 0.9065, "step": 14395 }, { "epoch": 0.66, "learning_rate": 9.150138850698598e-08, "loss": 0.9453, "step": 14396 }, { "epoch": 0.66, "learning_rate": 9.121065826859899e-08, "loss": 0.9245, "step": 14397 }, { "epoch": 0.66, "learning_rate": 9.092038852255069e-08, "loss": 0.9418, "step": 14398 }, { "epoch": 0.66, "learning_rate": 9.063057928233143e-08, "loss": 0.9373, "step": 14399 }, { "epoch": 0.66, "learning_rate": 9.034123056140931e-08, "loss": 0.9158, "step": 14400 }, { "epoch": 0.66, "learning_rate": 9.005234237323134e-08, "loss": 0.9352, "step": 14401 }, { "epoch": 0.66, "learning_rate": 8.976391473122237e-08, "loss": 0.9629, "step": 14402 }, { "epoch": 0.66, "learning_rate": 8.947594764878831e-08, "loss": 0.9155, "step": 14403 }, { "epoch": 0.66, "learning_rate": 8.918844113930958e-08, "loss": 0.9476, "step": 14404 }, { "epoch": 0.66, "learning_rate": 8.890139521614882e-08, "loss": 0.9, "step": 14405 }, { "epoch": 0.66, "learning_rate": 8.861480989264648e-08, "loss": 0.9127, "step": 14406 }, { "epoch": 0.66, "learning_rate": 8.832868518211857e-08, "loss": 0.9069, "step": 14407 }, { "epoch": 0.66, "learning_rate": 8.804302109786444e-08, "loss": 0.9054, "step": 14408 }, { "epoch": 0.66, "learning_rate": 8.775781765316016e-08, "loss": 0.9388, "step": 14409 }, { "epoch": 0.66, "learning_rate": 8.747307486125845e-08, "loss": 0.9152, "step": 14410 }, { "epoch": 0.66, "learning_rate": 8.718879273539316e-08, "loss": 0.8887, "step": 14411 }, { "epoch": 0.66, "learning_rate": 8.690497128877595e-08, "loss": 0.9317, "step": 14412 }, { "epoch": 0.66, "learning_rate": 8.662161053459516e-08, "loss": 0.9396, "step": 14413 }, { "epoch": 0.66, "learning_rate": 8.633871048602139e-08, "loss": 0.9168, "step": 14414 }, { "epoch": 0.66, "learning_rate": 8.605627115620185e-08, "loss": 0.941, "step": 14415 }, { "epoch": 0.66, "learning_rate": 8.577429255826164e-08, "loss": 0.9076, "step": 14416 }, { "epoch": 0.66, "learning_rate": 8.54927747053047e-08, "loss": 0.9787, "step": 14417 }, { "epoch": 0.66, "learning_rate": 8.521171761041502e-08, "loss": 0.916, "step": 14418 }, { "epoch": 0.66, "learning_rate": 8.493112128665326e-08, "loss": 0.924, "step": 14419 }, { "epoch": 0.66, "learning_rate": 8.46509857470601e-08, "loss": 0.9301, "step": 14420 }, { "epoch": 0.66, "learning_rate": 8.437131100465512e-08, "loss": 0.9067, "step": 14421 }, { "epoch": 0.66, "learning_rate": 8.40920970724346e-08, "loss": 0.9457, "step": 14422 }, { "epoch": 0.66, "learning_rate": 8.38133439633737e-08, "loss": 0.9411, "step": 14423 }, { "epoch": 0.66, "learning_rate": 8.353505169042986e-08, "loss": 0.9326, "step": 14424 }, { "epoch": 0.66, "learning_rate": 8.325722026653383e-08, "loss": 0.9424, "step": 14425 }, { "epoch": 0.66, "learning_rate": 8.297984970459749e-08, "loss": 0.9585, "step": 14426 }, { "epoch": 0.66, "learning_rate": 8.270294001751167e-08, "loss": 0.9249, "step": 14427 }, { "epoch": 0.66, "learning_rate": 8.242649121814494e-08, "loss": 0.9324, "step": 14428 }, { "epoch": 0.66, "learning_rate": 8.215050331934482e-08, "loss": 0.9356, "step": 14429 }, { "epoch": 0.66, "learning_rate": 8.18749763339377e-08, "loss": 0.9187, "step": 14430 }, { "epoch": 0.66, "learning_rate": 8.15999102747278e-08, "loss": 0.9391, "step": 14431 }, { "epoch": 0.66, "learning_rate": 8.13253051544971e-08, "loss": 0.9304, "step": 14432 }, { "epoch": 0.66, "learning_rate": 8.105116098600985e-08, "loss": 0.8866, "step": 14433 }, { "epoch": 0.66, "learning_rate": 8.077747778200474e-08, "loss": 0.9045, "step": 14434 }, { "epoch": 0.66, "learning_rate": 8.050425555520047e-08, "loss": 0.9303, "step": 14435 }, { "epoch": 0.66, "learning_rate": 8.023149431829691e-08, "loss": 0.9195, "step": 14436 }, { "epoch": 0.66, "learning_rate": 7.995919408396614e-08, "loss": 0.9177, "step": 14437 }, { "epoch": 0.66, "learning_rate": 7.968735486486579e-08, "loss": 0.9006, "step": 14438 }, { "epoch": 0.66, "learning_rate": 7.94159766736291e-08, "loss": 0.914, "step": 14439 }, { "epoch": 0.66, "learning_rate": 7.914505952286711e-08, "loss": 0.9384, "step": 14440 }, { "epoch": 0.66, "learning_rate": 7.887460342516861e-08, "loss": 0.8796, "step": 14441 }, { "epoch": 0.66, "learning_rate": 7.860460839310468e-08, "loss": 0.9402, "step": 14442 }, { "epoch": 0.66, "learning_rate": 7.833507443922306e-08, "loss": 0.8921, "step": 14443 }, { "epoch": 0.66, "learning_rate": 7.806600157604704e-08, "loss": 0.9194, "step": 14444 }, { "epoch": 0.66, "learning_rate": 7.779738981608554e-08, "loss": 0.8957, "step": 14445 }, { "epoch": 0.66, "learning_rate": 7.752923917181743e-08, "loss": 0.9398, "step": 14446 }, { "epoch": 0.66, "learning_rate": 7.72615496557072e-08, "loss": 0.8918, "step": 14447 }, { "epoch": 0.66, "learning_rate": 7.69943212801949e-08, "loss": 0.9261, "step": 14448 }, { "epoch": 0.66, "learning_rate": 7.67275540576995e-08, "loss": 0.8953, "step": 14449 }, { "epoch": 0.66, "learning_rate": 7.646124800061771e-08, "loss": 1.0034, "step": 14450 }, { "epoch": 0.66, "learning_rate": 7.619540312132524e-08, "loss": 0.9074, "step": 14451 }, { "epoch": 0.66, "learning_rate": 7.593001943217881e-08, "loss": 0.931, "step": 14452 }, { "epoch": 0.66, "learning_rate": 7.566509694550861e-08, "loss": 0.9131, "step": 14453 }, { "epoch": 0.66, "learning_rate": 7.540063567362921e-08, "loss": 0.9171, "step": 14454 }, { "epoch": 0.66, "learning_rate": 7.513663562882967e-08, "loss": 0.9206, "step": 14455 }, { "epoch": 0.66, "learning_rate": 7.487309682337906e-08, "loss": 0.8888, "step": 14456 }, { "epoch": 0.66, "learning_rate": 7.461001926952537e-08, "loss": 0.9329, "step": 14457 }, { "epoch": 0.66, "learning_rate": 7.434740297949328e-08, "loss": 0.9282, "step": 14458 }, { "epoch": 0.66, "learning_rate": 7.408524796548855e-08, "loss": 0.9447, "step": 14459 }, { "epoch": 0.66, "learning_rate": 7.382355423969367e-08, "loss": 0.8623, "step": 14460 }, { "epoch": 0.66, "learning_rate": 7.356232181427225e-08, "loss": 0.8946, "step": 14461 }, { "epoch": 0.66, "learning_rate": 7.330155070136125e-08, "loss": 0.9093, "step": 14462 }, { "epoch": 0.66, "learning_rate": 7.304124091308207e-08, "loss": 0.9633, "step": 14463 }, { "epoch": 0.66, "learning_rate": 7.278139246153171e-08, "loss": 0.9248, "step": 14464 }, { "epoch": 0.66, "learning_rate": 7.252200535878496e-08, "loss": 0.9137, "step": 14465 }, { "epoch": 0.66, "learning_rate": 7.226307961689771e-08, "loss": 0.9006, "step": 14466 }, { "epoch": 0.66, "learning_rate": 7.200461524790258e-08, "loss": 0.9407, "step": 14467 }, { "epoch": 0.66, "learning_rate": 7.174661226380997e-08, "loss": 0.9246, "step": 14468 }, { "epoch": 0.66, "learning_rate": 7.148907067661249e-08, "loss": 0.9219, "step": 14469 }, { "epoch": 0.66, "learning_rate": 7.123199049827723e-08, "loss": 0.9107, "step": 14470 }, { "epoch": 0.66, "learning_rate": 7.097537174075242e-08, "loss": 0.9027, "step": 14471 }, { "epoch": 0.66, "learning_rate": 7.071921441596297e-08, "loss": 0.947, "step": 14472 }, { "epoch": 0.66, "learning_rate": 7.04635185358149e-08, "loss": 0.9768, "step": 14473 }, { "epoch": 0.66, "learning_rate": 7.020828411218872e-08, "loss": 0.9306, "step": 14474 }, { "epoch": 0.66, "learning_rate": 6.995351115694826e-08, "loss": 0.8855, "step": 14475 }, { "epoch": 0.66, "learning_rate": 6.969919968193184e-08, "loss": 0.9208, "step": 14476 }, { "epoch": 0.66, "learning_rate": 6.944534969896e-08, "loss": 0.916, "step": 14477 }, { "epoch": 0.66, "learning_rate": 6.919196121982774e-08, "loss": 0.9114, "step": 14478 }, { "epoch": 0.66, "learning_rate": 6.893903425631233e-08, "loss": 0.9582, "step": 14479 }, { "epoch": 0.66, "learning_rate": 6.868656882016877e-08, "loss": 0.9548, "step": 14480 }, { "epoch": 0.66, "learning_rate": 6.843456492312772e-08, "loss": 0.9492, "step": 14481 }, { "epoch": 0.66, "learning_rate": 6.818302257690201e-08, "loss": 0.9012, "step": 14482 }, { "epoch": 0.66, "learning_rate": 6.793194179318119e-08, "loss": 0.9304, "step": 14483 }, { "epoch": 0.66, "learning_rate": 6.768132258363369e-08, "loss": 0.9202, "step": 14484 }, { "epoch": 0.66, "learning_rate": 6.743116495990798e-08, "loss": 0.9271, "step": 14485 }, { "epoch": 0.66, "learning_rate": 6.718146893362809e-08, "loss": 0.9155, "step": 14486 }, { "epoch": 0.66, "learning_rate": 6.693223451639808e-08, "loss": 0.935, "step": 14487 }, { "epoch": 0.66, "learning_rate": 6.668346171980089e-08, "loss": 0.9224, "step": 14488 }, { "epoch": 0.66, "learning_rate": 6.643515055539839e-08, "loss": 0.9473, "step": 14489 }, { "epoch": 0.66, "learning_rate": 6.618730103472915e-08, "loss": 0.9449, "step": 14490 }, { "epoch": 0.66, "learning_rate": 6.593991316931391e-08, "loss": 0.8861, "step": 14491 }, { "epoch": 0.66, "learning_rate": 6.569298697064685e-08, "loss": 0.8978, "step": 14492 }, { "epoch": 0.66, "learning_rate": 6.544652245020433e-08, "loss": 0.9289, "step": 14493 }, { "epoch": 0.66, "learning_rate": 6.520051961944052e-08, "loss": 0.9661, "step": 14494 }, { "epoch": 0.66, "learning_rate": 6.49549784897896e-08, "loss": 0.9073, "step": 14495 }, { "epoch": 0.66, "learning_rate": 6.470989907265913e-08, "loss": 0.898, "step": 14496 }, { "epoch": 0.66, "learning_rate": 6.446528137944108e-08, "loss": 0.897, "step": 14497 }, { "epoch": 0.66, "learning_rate": 6.422112542150415e-08, "loss": 0.9449, "step": 14498 }, { "epoch": 0.66, "learning_rate": 6.397743121019262e-08, "loss": 0.9616, "step": 14499 }, { "epoch": 0.66, "learning_rate": 6.373419875683406e-08, "loss": 0.93, "step": 14500 }, { "epoch": 0.66, "learning_rate": 6.349142807273056e-08, "loss": 0.9343, "step": 14501 }, { "epoch": 0.66, "learning_rate": 6.324911916916531e-08, "loss": 0.9131, "step": 14502 }, { "epoch": 0.67, "learning_rate": 6.300727205739932e-08, "loss": 0.9196, "step": 14503 }, { "epoch": 0.67, "learning_rate": 6.276588674867023e-08, "loss": 0.9495, "step": 14504 }, { "epoch": 0.67, "learning_rate": 6.25249632541991e-08, "loss": 0.9124, "step": 14505 }, { "epoch": 0.67, "learning_rate": 6.228450158517918e-08, "loss": 0.9373, "step": 14506 }, { "epoch": 0.67, "learning_rate": 6.20445017527882e-08, "loss": 0.9568, "step": 14507 }, { "epoch": 0.67, "learning_rate": 6.180496376817835e-08, "loss": 0.9271, "step": 14508 }, { "epoch": 0.67, "learning_rate": 6.156588764248072e-08, "loss": 0.9418, "step": 14509 }, { "epoch": 0.67, "learning_rate": 6.132727338680866e-08, "loss": 0.9409, "step": 14510 }, { "epoch": 0.67, "learning_rate": 6.108912101224773e-08, "loss": 0.8903, "step": 14511 }, { "epoch": 0.67, "learning_rate": 6.085143052986909e-08, "loss": 0.9358, "step": 14512 }, { "epoch": 0.67, "learning_rate": 6.06142019507161e-08, "loss": 0.9161, "step": 14513 }, { "epoch": 0.67, "learning_rate": 6.037743528581552e-08, "loss": 0.94, "step": 14514 }, { "epoch": 0.67, "learning_rate": 6.014113054616965e-08, "loss": 0.9462, "step": 14515 }, { "epoch": 0.67, "learning_rate": 5.990528774276194e-08, "loss": 0.9595, "step": 14516 }, { "epoch": 0.67, "learning_rate": 5.966990688655027e-08, "loss": 0.9239, "step": 14517 }, { "epoch": 0.67, "learning_rate": 5.94349879884748e-08, "loss": 0.9351, "step": 14518 }, { "epoch": 0.67, "learning_rate": 5.920053105945345e-08, "loss": 0.9486, "step": 14519 }, { "epoch": 0.67, "learning_rate": 5.896653611037972e-08, "loss": 0.9642, "step": 14520 }, { "epoch": 0.67, "learning_rate": 5.8733003152131596e-08, "loss": 0.9421, "step": 14521 }, { "epoch": 0.67, "learning_rate": 5.8499932195559274e-08, "loss": 0.9166, "step": 14522 }, { "epoch": 0.67, "learning_rate": 5.826732325149631e-08, "loss": 0.9243, "step": 14523 }, { "epoch": 0.67, "learning_rate": 5.803517633075073e-08, "loss": 0.939, "step": 14524 }, { "epoch": 0.67, "learning_rate": 5.78034914441139e-08, "loss": 0.8985, "step": 14525 }, { "epoch": 0.67, "learning_rate": 5.757226860235055e-08, "loss": 0.9596, "step": 14526 }, { "epoch": 0.67, "learning_rate": 5.734150781620651e-08, "loss": 0.9599, "step": 14527 }, { "epoch": 0.67, "learning_rate": 5.711120909640655e-08, "loss": 0.9049, "step": 14528 }, { "epoch": 0.67, "learning_rate": 5.688137245365433e-08, "loss": 0.9573, "step": 14529 }, { "epoch": 0.67, "learning_rate": 5.665199789862907e-08, "loss": 0.9111, "step": 14530 }, { "epoch": 0.67, "learning_rate": 5.6423085441991155e-08, "loss": 0.9243, "step": 14531 }, { "epoch": 0.67, "learning_rate": 5.619463509437983e-08, "loss": 0.8984, "step": 14532 }, { "epoch": 0.67, "learning_rate": 5.5966646866412175e-08, "loss": 0.8553, "step": 14533 }, { "epoch": 0.67, "learning_rate": 5.5739120768680824e-08, "loss": 0.9492, "step": 14534 }, { "epoch": 0.67, "learning_rate": 5.551205681176286e-08, "loss": 0.9041, "step": 14535 }, { "epoch": 0.67, "learning_rate": 5.528545500620874e-08, "loss": 0.8988, "step": 14536 }, { "epoch": 0.67, "learning_rate": 5.505931536255005e-08, "loss": 0.931, "step": 14537 }, { "epoch": 0.67, "learning_rate": 5.483363789129503e-08, "loss": 0.9084, "step": 14538 }, { "epoch": 0.67, "learning_rate": 5.4608422602933084e-08, "loss": 0.8593, "step": 14539 }, { "epoch": 0.67, "learning_rate": 5.4383669507930284e-08, "loss": 0.9528, "step": 14540 }, { "epoch": 0.67, "learning_rate": 5.4159378616731594e-08, "loss": 0.9245, "step": 14541 }, { "epoch": 0.67, "learning_rate": 5.393554993976091e-08, "loss": 0.935, "step": 14542 }, { "epoch": 0.67, "learning_rate": 5.3712183487418803e-08, "loss": 0.8948, "step": 14543 }, { "epoch": 0.67, "learning_rate": 5.348927927008696e-08, "loss": 0.9209, "step": 14544 }, { "epoch": 0.67, "learning_rate": 5.326683729812376e-08, "loss": 0.9281, "step": 14545 }, { "epoch": 0.67, "learning_rate": 5.304485758186873e-08, "loss": 0.9174, "step": 14546 }, { "epoch": 0.67, "learning_rate": 5.28233401316347e-08, "loss": 0.9271, "step": 14547 }, { "epoch": 0.67, "learning_rate": 5.260228495771902e-08, "loss": 0.9553, "step": 14548 }, { "epoch": 0.67, "learning_rate": 5.2381692070393454e-08, "loss": 0.9295, "step": 14549 }, { "epoch": 0.67, "learning_rate": 5.216156147991092e-08, "loss": 0.926, "step": 14550 }, { "epoch": 0.67, "learning_rate": 5.194189319649989e-08, "loss": 0.882, "step": 14551 }, { "epoch": 0.67, "learning_rate": 5.1722687230369995e-08, "loss": 0.9356, "step": 14552 }, { "epoch": 0.67, "learning_rate": 5.150394359170752e-08, "loss": 0.9282, "step": 14553 }, { "epoch": 0.67, "learning_rate": 5.128566229067988e-08, "loss": 0.9163, "step": 14554 }, { "epoch": 0.67, "learning_rate": 5.106784333743009e-08, "loss": 0.9663, "step": 14555 }, { "epoch": 0.67, "learning_rate": 5.085048674208115e-08, "loss": 0.9237, "step": 14556 }, { "epoch": 0.67, "learning_rate": 5.063359251473277e-08, "loss": 0.923, "step": 14557 }, { "epoch": 0.67, "learning_rate": 5.041716066546798e-08, "loss": 0.9541, "step": 14558 }, { "epoch": 0.67, "learning_rate": 5.020119120434208e-08, "loss": 0.9145, "step": 14559 }, { "epoch": 0.67, "learning_rate": 4.998568414139371e-08, "loss": 0.9723, "step": 14560 }, { "epoch": 0.67, "learning_rate": 4.9770639486637075e-08, "loss": 0.9202, "step": 14561 }, { "epoch": 0.67, "learning_rate": 4.9556057250066406e-08, "loss": 0.9546, "step": 14562 }, { "epoch": 0.67, "learning_rate": 4.934193744165483e-08, "loss": 0.9412, "step": 14563 }, { "epoch": 0.67, "learning_rate": 4.912828007135106e-08, "loss": 0.9271, "step": 14564 }, { "epoch": 0.67, "learning_rate": 4.891508514908605e-08, "loss": 0.922, "step": 14565 }, { "epoch": 0.67, "learning_rate": 4.8702352684767416e-08, "loss": 0.9223, "step": 14566 }, { "epoch": 0.67, "learning_rate": 4.8490082688280595e-08, "loss": 0.9543, "step": 14567 }, { "epoch": 0.67, "learning_rate": 4.827827516949102e-08, "loss": 0.9431, "step": 14568 }, { "epoch": 0.67, "learning_rate": 4.806693013824304e-08, "loss": 0.9173, "step": 14569 }, { "epoch": 0.67, "learning_rate": 4.785604760435658e-08, "loss": 0.9008, "step": 14570 }, { "epoch": 0.67, "learning_rate": 4.764562757763491e-08, "loss": 0.958, "step": 14571 }, { "epoch": 0.67, "learning_rate": 4.743567006785354e-08, "loss": 0.9479, "step": 14572 }, { "epoch": 0.67, "learning_rate": 4.7226175084771344e-08, "loss": 0.9351, "step": 14573 }, { "epoch": 0.67, "learning_rate": 4.701714263812496e-08, "loss": 0.8957, "step": 14574 }, { "epoch": 0.67, "learning_rate": 4.680857273762773e-08, "loss": 0.9073, "step": 14575 }, { "epoch": 0.67, "learning_rate": 4.6600465392971916e-08, "loss": 0.9118, "step": 14576 }, { "epoch": 0.67, "learning_rate": 4.639282061383088e-08, "loss": 0.9034, "step": 14577 }, { "epoch": 0.67, "learning_rate": 4.618563840985357e-08, "loss": 0.9276, "step": 14578 }, { "epoch": 0.67, "learning_rate": 4.5978918790667846e-08, "loss": 0.967, "step": 14579 }, { "epoch": 0.67, "learning_rate": 4.5772661765881576e-08, "loss": 0.9371, "step": 14580 }, { "epoch": 0.67, "learning_rate": 4.556686734507931e-08, "loss": 0.9349, "step": 14581 }, { "epoch": 0.67, "learning_rate": 4.536153553782563e-08, "loss": 0.8812, "step": 14582 }, { "epoch": 0.67, "learning_rate": 4.5156666353662895e-08, "loss": 0.9143, "step": 14583 }, { "epoch": 0.67, "learning_rate": 4.495225980211127e-08, "loss": 0.9438, "step": 14584 }, { "epoch": 0.67, "learning_rate": 4.4748315892669814e-08, "loss": 0.9089, "step": 14585 }, { "epoch": 0.67, "learning_rate": 4.4544834634818735e-08, "loss": 0.9205, "step": 14586 }, { "epoch": 0.67, "learning_rate": 4.434181603801268e-08, "loss": 0.9039, "step": 14587 }, { "epoch": 0.67, "learning_rate": 4.413926011168634e-08, "loss": 0.9359, "step": 14588 }, { "epoch": 0.67, "learning_rate": 4.39371668652544e-08, "loss": 0.956, "step": 14589 }, { "epoch": 0.67, "learning_rate": 4.3735536308107116e-08, "loss": 0.9227, "step": 14590 }, { "epoch": 0.67, "learning_rate": 4.353436844961589e-08, "loss": 0.9122, "step": 14591 }, { "epoch": 0.67, "learning_rate": 4.3333663299129915e-08, "loss": 0.9833, "step": 14592 }, { "epoch": 0.67, "learning_rate": 4.313342086597616e-08, "loss": 0.9319, "step": 14593 }, { "epoch": 0.67, "learning_rate": 4.2933641159460524e-08, "loss": 0.9155, "step": 14594 }, { "epoch": 0.67, "learning_rate": 4.27343241888678e-08, "loss": 0.9401, "step": 14595 }, { "epoch": 0.67, "learning_rate": 4.2535469963459476e-08, "loss": 0.8838, "step": 14596 }, { "epoch": 0.67, "learning_rate": 4.233707849247926e-08, "loss": 0.916, "step": 14597 }, { "epoch": 0.67, "learning_rate": 4.2139149785144216e-08, "loss": 0.9256, "step": 14598 }, { "epoch": 0.67, "learning_rate": 4.194168385065478e-08, "loss": 0.9258, "step": 14599 }, { "epoch": 0.67, "learning_rate": 4.174468069818805e-08, "loss": 0.9409, "step": 14600 }, { "epoch": 0.67, "learning_rate": 4.1548140336896695e-08, "loss": 0.9247, "step": 14601 }, { "epoch": 0.67, "learning_rate": 4.135206277591785e-08, "loss": 0.9116, "step": 14602 }, { "epoch": 0.67, "learning_rate": 4.115644802436203e-08, "loss": 0.9301, "step": 14603 }, { "epoch": 0.67, "learning_rate": 4.0961296091320826e-08, "loss": 0.927, "step": 14604 }, { "epoch": 0.67, "learning_rate": 4.0766606985862544e-08, "loss": 0.9189, "step": 14605 }, { "epoch": 0.67, "learning_rate": 4.057238071703551e-08, "loss": 0.9446, "step": 14606 }, { "epoch": 0.67, "learning_rate": 4.037861729386694e-08, "loss": 0.9559, "step": 14607 }, { "epoch": 0.67, "learning_rate": 4.018531672536075e-08, "loss": 0.9156, "step": 14608 }, { "epoch": 0.67, "learning_rate": 3.999247902049974e-08, "loss": 0.9148, "step": 14609 }, { "epoch": 0.67, "learning_rate": 3.980010418824565e-08, "loss": 0.94, "step": 14610 }, { "epoch": 0.67, "learning_rate": 3.96081922375402e-08, "loss": 0.9334, "step": 14611 }, { "epoch": 0.67, "learning_rate": 3.941674317730071e-08, "loss": 0.9636, "step": 14612 }, { "epoch": 0.67, "learning_rate": 3.922575701642561e-08, "loss": 0.9454, "step": 14613 }, { "epoch": 0.67, "learning_rate": 3.903523376379004e-08, "loss": 0.9536, "step": 14614 }, { "epoch": 0.67, "learning_rate": 3.8845173428248007e-08, "loss": 0.9241, "step": 14615 }, { "epoch": 0.67, "learning_rate": 3.8655576018631346e-08, "loss": 0.9617, "step": 14616 }, { "epoch": 0.67, "learning_rate": 3.8466441543753006e-08, "loss": 0.944, "step": 14617 }, { "epoch": 0.67, "learning_rate": 3.8277770012401515e-08, "loss": 0.9399, "step": 14618 }, { "epoch": 0.67, "learning_rate": 3.808956143334541e-08, "loss": 0.8901, "step": 14619 }, { "epoch": 0.67, "learning_rate": 3.7901815815332145e-08, "loss": 0.893, "step": 14620 }, { "epoch": 0.67, "learning_rate": 3.771453316708584e-08, "loss": 0.9126, "step": 14621 }, { "epoch": 0.67, "learning_rate": 3.752771349730955e-08, "loss": 0.9184, "step": 14622 }, { "epoch": 0.67, "learning_rate": 3.734135681468631e-08, "loss": 0.9311, "step": 14623 }, { "epoch": 0.67, "learning_rate": 3.715546312787699e-08, "loss": 0.9544, "step": 14624 }, { "epoch": 0.67, "learning_rate": 3.697003244551911e-08, "loss": 0.9214, "step": 14625 }, { "epoch": 0.67, "learning_rate": 3.678506477623245e-08, "loss": 0.9211, "step": 14626 }, { "epoch": 0.67, "learning_rate": 3.660056012861124e-08, "loss": 0.8899, "step": 14627 }, { "epoch": 0.67, "learning_rate": 3.641651851122974e-08, "loss": 0.9473, "step": 14628 }, { "epoch": 0.67, "learning_rate": 3.6232939932643316e-08, "loss": 0.928, "step": 14629 }, { "epoch": 0.67, "learning_rate": 3.6049824401380715e-08, "loss": 0.9191, "step": 14630 }, { "epoch": 0.67, "learning_rate": 3.5867171925954016e-08, "loss": 0.9524, "step": 14631 }, { "epoch": 0.67, "learning_rate": 3.568498251484975e-08, "loss": 0.9471, "step": 14632 }, { "epoch": 0.67, "learning_rate": 3.5503256176536714e-08, "loss": 0.9223, "step": 14633 }, { "epoch": 0.67, "learning_rate": 3.532199291945815e-08, "loss": 0.9272, "step": 14634 }, { "epoch": 0.67, "learning_rate": 3.514119275203953e-08, "loss": 0.8941, "step": 14635 }, { "epoch": 0.67, "learning_rate": 3.4960855682684145e-08, "loss": 0.9363, "step": 14636 }, { "epoch": 0.67, "learning_rate": 3.478098171976974e-08, "loss": 0.9289, "step": 14637 }, { "epoch": 0.67, "learning_rate": 3.4601570871658494e-08, "loss": 0.9214, "step": 14638 }, { "epoch": 0.67, "learning_rate": 3.442262314668599e-08, "loss": 0.9398, "step": 14639 }, { "epoch": 0.67, "learning_rate": 3.424413855316999e-08, "loss": 0.9527, "step": 14640 }, { "epoch": 0.67, "learning_rate": 3.40661170994061e-08, "loss": 0.9476, "step": 14641 }, { "epoch": 0.67, "learning_rate": 3.388855879366437e-08, "loss": 0.9084, "step": 14642 }, { "epoch": 0.67, "learning_rate": 3.37114636441993e-08, "loss": 0.9168, "step": 14643 }, { "epoch": 0.67, "learning_rate": 3.353483165923988e-08, "loss": 0.921, "step": 14644 }, { "epoch": 0.67, "learning_rate": 3.335866284699507e-08, "loss": 0.9228, "step": 14645 }, { "epoch": 0.67, "learning_rate": 3.318295721565168e-08, "loss": 0.8681, "step": 14646 }, { "epoch": 0.67, "learning_rate": 3.300771477337539e-08, "loss": 0.8681, "step": 14647 }, { "epoch": 0.67, "learning_rate": 3.28329355283108e-08, "loss": 0.9499, "step": 14648 }, { "epoch": 0.67, "learning_rate": 3.265861948857918e-08, "loss": 0.9158, "step": 14649 }, { "epoch": 0.67, "learning_rate": 3.248476666228295e-08, "loss": 0.9149, "step": 14650 }, { "epoch": 0.67, "learning_rate": 3.2311377057501204e-08, "loss": 0.9171, "step": 14651 }, { "epoch": 0.67, "learning_rate": 3.213845068229082e-08, "loss": 0.8917, "step": 14652 }, { "epoch": 0.67, "learning_rate": 3.1965987544689826e-08, "loss": 0.9158, "step": 14653 }, { "epoch": 0.68, "learning_rate": 3.179398765271291e-08, "loss": 0.9052, "step": 14654 }, { "epoch": 0.68, "learning_rate": 3.1622451014352575e-08, "loss": 0.9176, "step": 14655 }, { "epoch": 0.68, "learning_rate": 3.1451377637580214e-08, "loss": 0.9139, "step": 14656 }, { "epoch": 0.68, "learning_rate": 3.128076753034837e-08, "loss": 0.9065, "step": 14657 }, { "epoch": 0.68, "learning_rate": 3.111062070058401e-08, "loss": 0.9529, "step": 14658 }, { "epoch": 0.68, "learning_rate": 3.0940937156194175e-08, "loss": 0.9228, "step": 14659 }, { "epoch": 0.68, "learning_rate": 3.077171690506475e-08, "loss": 0.9185, "step": 14660 }, { "epoch": 0.68, "learning_rate": 3.0602959955061686e-08, "loss": 0.8767, "step": 14661 }, { "epoch": 0.68, "learning_rate": 3.043466631402536e-08, "loss": 0.9203, "step": 14662 }, { "epoch": 0.68, "learning_rate": 3.026683598977842e-08, "loss": 0.9757, "step": 14663 }, { "epoch": 0.68, "learning_rate": 3.009946899012017e-08, "loss": 0.8998, "step": 14664 }, { "epoch": 0.68, "learning_rate": 2.993256532282773e-08, "loss": 0.8766, "step": 14665 }, { "epoch": 0.68, "learning_rate": 2.976612499565934e-08, "loss": 0.9027, "step": 14666 }, { "epoch": 0.68, "learning_rate": 2.9600148016347697e-08, "loss": 0.8866, "step": 14667 }, { "epoch": 0.68, "learning_rate": 2.9434634392607743e-08, "loss": 0.9138, "step": 14668 }, { "epoch": 0.68, "learning_rate": 2.9269584132132213e-08, "loss": 0.9071, "step": 14669 }, { "epoch": 0.68, "learning_rate": 2.9104997242590528e-08, "loss": 0.9317, "step": 14670 }, { "epoch": 0.68, "learning_rate": 2.8940873731631015e-08, "loss": 0.9728, "step": 14671 }, { "epoch": 0.68, "learning_rate": 2.8777213606880904e-08, "loss": 0.9246, "step": 14672 }, { "epoch": 0.68, "learning_rate": 2.861401687594745e-08, "loss": 0.912, "step": 14673 }, { "epoch": 0.68, "learning_rate": 2.8451283546413465e-08, "loss": 0.9059, "step": 14674 }, { "epoch": 0.68, "learning_rate": 2.828901362584402e-08, "loss": 0.9512, "step": 14675 }, { "epoch": 0.68, "learning_rate": 2.8127207121776412e-08, "loss": 0.951, "step": 14676 }, { "epoch": 0.68, "learning_rate": 2.7965864041733515e-08, "loss": 0.9428, "step": 14677 }, { "epoch": 0.68, "learning_rate": 2.7804984393212664e-08, "loss": 0.9029, "step": 14678 }, { "epoch": 0.68, "learning_rate": 2.76445681836901e-08, "loss": 0.9412, "step": 14679 }, { "epoch": 0.68, "learning_rate": 2.7484615420620975e-08, "loss": 0.9586, "step": 14680 }, { "epoch": 0.68, "learning_rate": 2.7325126111438227e-08, "loss": 0.9099, "step": 14681 }, { "epoch": 0.68, "learning_rate": 2.7166100263554816e-08, "loss": 0.9034, "step": 14682 }, { "epoch": 0.68, "learning_rate": 2.700753788436039e-08, "loss": 0.9202, "step": 14683 }, { "epoch": 0.68, "learning_rate": 2.6849438981224607e-08, "loss": 0.9308, "step": 14684 }, { "epoch": 0.68, "learning_rate": 2.669180356149381e-08, "loss": 0.9146, "step": 14685 }, { "epoch": 0.68, "learning_rate": 2.6534631632494368e-08, "loss": 0.9329, "step": 14686 }, { "epoch": 0.68, "learning_rate": 2.637792320153154e-08, "loss": 0.9126, "step": 14687 }, { "epoch": 0.68, "learning_rate": 2.6221678275886176e-08, "loss": 0.9525, "step": 14688 }, { "epoch": 0.68, "learning_rate": 2.606589686282024e-08, "loss": 0.9393, "step": 14689 }, { "epoch": 0.68, "learning_rate": 2.5910578969574607e-08, "loss": 0.9593, "step": 14690 }, { "epoch": 0.68, "learning_rate": 2.575572460336573e-08, "loss": 0.9341, "step": 14691 }, { "epoch": 0.68, "learning_rate": 2.560133377139007e-08, "loss": 0.9303, "step": 14692 }, { "epoch": 0.68, "learning_rate": 2.5447406480823e-08, "loss": 0.9827, "step": 14693 }, { "epoch": 0.68, "learning_rate": 2.52939427388188e-08, "loss": 0.8951, "step": 14694 }, { "epoch": 0.68, "learning_rate": 2.514094255250954e-08, "loss": 0.9184, "step": 14695 }, { "epoch": 0.68, "learning_rate": 2.498840592900509e-08, "loss": 0.9266, "step": 14696 }, { "epoch": 0.68, "learning_rate": 2.483633287539311e-08, "loss": 0.9359, "step": 14697 }, { "epoch": 0.68, "learning_rate": 2.4684723398742394e-08, "loss": 0.9044, "step": 14698 }, { "epoch": 0.68, "learning_rate": 2.4533577506098415e-08, "loss": 0.8457, "step": 14699 }, { "epoch": 0.68, "learning_rate": 2.4382895204485556e-08, "loss": 0.9442, "step": 14700 }, { "epoch": 0.68, "learning_rate": 2.4232676500905995e-08, "loss": 0.9194, "step": 14701 }, { "epoch": 0.68, "learning_rate": 2.4082921402340808e-08, "loss": 0.9682, "step": 14702 }, { "epoch": 0.68, "learning_rate": 2.39336299157511e-08, "loss": 0.9088, "step": 14703 }, { "epoch": 0.68, "learning_rate": 2.378480204807243e-08, "loss": 0.8712, "step": 14704 }, { "epoch": 0.68, "learning_rate": 2.36364378062226e-08, "loss": 0.923, "step": 14705 }, { "epoch": 0.68, "learning_rate": 2.3488537197096094e-08, "loss": 0.9366, "step": 14706 }, { "epoch": 0.68, "learning_rate": 2.3341100227567414e-08, "loss": 0.9163, "step": 14707 }, { "epoch": 0.68, "learning_rate": 2.3194126904487748e-08, "loss": 0.9054, "step": 14708 }, { "epoch": 0.68, "learning_rate": 2.304761723468718e-08, "loss": 0.9472, "step": 14709 }, { "epoch": 0.68, "learning_rate": 2.2901571224974718e-08, "loss": 0.9197, "step": 14710 }, { "epoch": 0.68, "learning_rate": 2.275598888213826e-08, "loss": 0.9189, "step": 14711 }, { "epoch": 0.68, "learning_rate": 2.2610870212942394e-08, "loss": 0.9143, "step": 14712 }, { "epoch": 0.68, "learning_rate": 2.2466215224130615e-08, "loss": 0.9563, "step": 14713 }, { "epoch": 0.68, "learning_rate": 2.232202392242755e-08, "loss": 0.9157, "step": 14714 }, { "epoch": 0.68, "learning_rate": 2.2178296314532278e-08, "loss": 0.9136, "step": 14715 }, { "epoch": 0.68, "learning_rate": 2.2035032407126123e-08, "loss": 0.8787, "step": 14716 }, { "epoch": 0.68, "learning_rate": 2.189223220686487e-08, "loss": 0.9035, "step": 14717 }, { "epoch": 0.68, "learning_rate": 2.174989572038655e-08, "loss": 0.9256, "step": 14718 }, { "epoch": 0.68, "learning_rate": 2.1608022954305863e-08, "loss": 0.9361, "step": 14719 }, { "epoch": 0.68, "learning_rate": 2.1466613915214208e-08, "loss": 0.9286, "step": 14720 }, { "epoch": 0.68, "learning_rate": 2.1325668609686323e-08, "loss": 0.8939, "step": 14721 }, { "epoch": 0.68, "learning_rate": 2.1185187044270307e-08, "loss": 0.9705, "step": 14722 }, { "epoch": 0.68, "learning_rate": 2.1045169225495376e-08, "loss": 0.9189, "step": 14723 }, { "epoch": 0.68, "learning_rate": 2.0905615159868554e-08, "loss": 0.9206, "step": 14724 }, { "epoch": 0.68, "learning_rate": 2.0766524853874648e-08, "loss": 0.9174, "step": 14725 }, { "epoch": 0.68, "learning_rate": 2.062789831397849e-08, "loss": 0.9215, "step": 14726 }, { "epoch": 0.68, "learning_rate": 2.0489735546622702e-08, "loss": 0.9573, "step": 14727 }, { "epoch": 0.68, "learning_rate": 2.0352036558227704e-08, "loss": 0.9078, "step": 14728 }, { "epoch": 0.68, "learning_rate": 2.0214801355192826e-08, "loss": 0.9081, "step": 14729 }, { "epoch": 0.68, "learning_rate": 2.0078029943895184e-08, "loss": 0.9398, "step": 14730 }, { "epoch": 0.68, "learning_rate": 1.9941722330691915e-08, "loss": 0.9279, "step": 14731 }, { "epoch": 0.68, "learning_rate": 1.9805878521917955e-08, "loss": 0.9222, "step": 14732 }, { "epoch": 0.68, "learning_rate": 1.9670498523884918e-08, "loss": 0.9002, "step": 14733 }, { "epoch": 0.68, "learning_rate": 1.953558234288555e-08, "loss": 0.9311, "step": 14734 }, { "epoch": 0.68, "learning_rate": 1.9401129985189283e-08, "loss": 0.9178, "step": 14735 }, { "epoch": 0.68, "learning_rate": 1.9267141457045557e-08, "loss": 0.9274, "step": 14736 }, { "epoch": 0.68, "learning_rate": 1.9133616764679396e-08, "loss": 0.9256, "step": 14737 }, { "epoch": 0.68, "learning_rate": 1.9000555914296948e-08, "loss": 0.9379, "step": 14738 }, { "epoch": 0.68, "learning_rate": 1.886795891208326e-08, "loss": 0.9333, "step": 14739 }, { "epoch": 0.68, "learning_rate": 1.8735825764197857e-08, "loss": 0.9071, "step": 14740 }, { "epoch": 0.68, "learning_rate": 1.8604156476783595e-08, "loss": 0.8841, "step": 14741 }, { "epoch": 0.68, "learning_rate": 1.847295105595892e-08, "loss": 0.9239, "step": 14742 }, { "epoch": 0.68, "learning_rate": 1.8342209507822285e-08, "loss": 0.9058, "step": 14743 }, { "epoch": 0.68, "learning_rate": 1.8211931838447715e-08, "loss": 0.9265, "step": 14744 }, { "epoch": 0.68, "learning_rate": 1.8082118053890374e-08, "loss": 0.9121, "step": 14745 }, { "epoch": 0.68, "learning_rate": 1.795276816018321e-08, "loss": 0.9637, "step": 14746 }, { "epoch": 0.68, "learning_rate": 1.7823882163338078e-08, "loss": 0.9158, "step": 14747 }, { "epoch": 0.68, "learning_rate": 1.769546006934464e-08, "loss": 0.9334, "step": 14748 }, { "epoch": 0.68, "learning_rate": 1.7567501884170335e-08, "loss": 0.8978, "step": 14749 }, { "epoch": 0.68, "learning_rate": 1.744000761376152e-08, "loss": 0.9399, "step": 14750 }, { "epoch": 0.68, "learning_rate": 1.7312977264044572e-08, "loss": 0.9102, "step": 14751 }, { "epoch": 0.68, "learning_rate": 1.718641084092143e-08, "loss": 0.9281, "step": 14752 }, { "epoch": 0.68, "learning_rate": 1.706030835027628e-08, "loss": 0.9101, "step": 14753 }, { "epoch": 0.68, "learning_rate": 1.693466979796665e-08, "loss": 0.9005, "step": 14754 }, { "epoch": 0.68, "learning_rate": 1.6809495189833435e-08, "loss": 0.9398, "step": 14755 }, { "epoch": 0.68, "learning_rate": 1.6684784531693087e-08, "loss": 0.9197, "step": 14756 }, { "epoch": 0.68, "learning_rate": 1.6560537829342082e-08, "loss": 0.9474, "step": 14757 }, { "epoch": 0.68, "learning_rate": 1.643675508855358e-08, "loss": 0.9581, "step": 14758 }, { "epoch": 0.68, "learning_rate": 1.6313436315080754e-08, "loss": 0.9154, "step": 14759 }, { "epoch": 0.68, "learning_rate": 1.619058151465458e-08, "loss": 0.9303, "step": 14760 }, { "epoch": 0.68, "learning_rate": 1.606819069298271e-08, "loss": 0.9094, "step": 14761 }, { "epoch": 0.68, "learning_rate": 1.594626385575615e-08, "loss": 0.958, "step": 14762 }, { "epoch": 0.68, "learning_rate": 1.5824801008639257e-08, "loss": 0.9233, "step": 14763 }, { "epoch": 0.68, "learning_rate": 1.5703802157277514e-08, "loss": 0.9399, "step": 14764 }, { "epoch": 0.68, "learning_rate": 1.558326730729309e-08, "loss": 0.9275, "step": 14765 }, { "epoch": 0.68, "learning_rate": 1.546319646428929e-08, "loss": 0.9576, "step": 14766 }, { "epoch": 0.68, "learning_rate": 1.5343589633844968e-08, "loss": 0.9606, "step": 14767 }, { "epoch": 0.68, "learning_rate": 1.5224446821517914e-08, "loss": 0.9557, "step": 14768 }, { "epoch": 0.68, "learning_rate": 1.5105768032847023e-08, "loss": 0.9419, "step": 14769 }, { "epoch": 0.68, "learning_rate": 1.4987553273346778e-08, "loss": 0.9516, "step": 14770 }, { "epoch": 0.68, "learning_rate": 1.486980254851056e-08, "loss": 0.9158, "step": 14771 }, { "epoch": 0.68, "learning_rate": 1.4752515863810657e-08, "loss": 0.9459, "step": 14772 }, { "epoch": 0.68, "learning_rate": 1.4635693224698266e-08, "loss": 0.9091, "step": 14773 }, { "epoch": 0.68, "learning_rate": 1.4519334636602378e-08, "loss": 0.9261, "step": 14774 }, { "epoch": 0.68, "learning_rate": 1.4403440104929778e-08, "loss": 0.9303, "step": 14775 }, { "epoch": 0.68, "learning_rate": 1.4288009635068378e-08, "loss": 0.8747, "step": 14776 }, { "epoch": 0.68, "learning_rate": 1.4173043232380557e-08, "loss": 0.9221, "step": 14777 }, { "epoch": 0.68, "learning_rate": 1.4058540902209816e-08, "loss": 0.9479, "step": 14778 }, { "epoch": 0.68, "learning_rate": 1.3944502649878566e-08, "loss": 0.9422, "step": 14779 }, { "epoch": 0.68, "learning_rate": 1.383092848068368e-08, "loss": 0.9692, "step": 14780 }, { "epoch": 0.68, "learning_rate": 1.3717818399906491e-08, "loss": 0.9659, "step": 14781 }, { "epoch": 0.68, "learning_rate": 1.3605172412801681e-08, "loss": 0.9111, "step": 14782 }, { "epoch": 0.68, "learning_rate": 1.3492990524605065e-08, "loss": 0.9248, "step": 14783 }, { "epoch": 0.68, "learning_rate": 1.3381272740529139e-08, "loss": 0.9685, "step": 14784 }, { "epoch": 0.68, "learning_rate": 1.3270019065767526e-08, "loss": 0.9059, "step": 14785 }, { "epoch": 0.68, "learning_rate": 1.3159229505488313e-08, "loss": 0.9978, "step": 14786 }, { "epoch": 0.68, "learning_rate": 1.3048904064841827e-08, "loss": 0.9409, "step": 14787 }, { "epoch": 0.68, "learning_rate": 1.2939042748955078e-08, "loss": 0.9261, "step": 14788 }, { "epoch": 0.68, "learning_rate": 1.2829645562933978e-08, "loss": 0.9244, "step": 14789 }, { "epoch": 0.68, "learning_rate": 1.2720712511861133e-08, "loss": 0.9135, "step": 14790 }, { "epoch": 0.68, "learning_rate": 1.2612243600799156e-08, "loss": 0.9353, "step": 14791 }, { "epoch": 0.68, "learning_rate": 1.2504238834790683e-08, "loss": 0.9728, "step": 14792 }, { "epoch": 0.68, "learning_rate": 1.239669821885281e-08, "loss": 0.9614, "step": 14793 }, { "epoch": 0.68, "learning_rate": 1.2289621757985981e-08, "loss": 0.9187, "step": 14794 }, { "epoch": 0.68, "learning_rate": 1.2183009457163997e-08, "loss": 0.9133, "step": 14795 }, { "epoch": 0.68, "learning_rate": 1.2076861321341783e-08, "loss": 0.9281, "step": 14796 }, { "epoch": 0.68, "learning_rate": 1.1971177355453168e-08, "loss": 0.9466, "step": 14797 }, { "epoch": 0.68, "learning_rate": 1.186595756440978e-08, "loss": 0.9442, "step": 14798 }, { "epoch": 0.68, "learning_rate": 1.1761201953099931e-08, "loss": 0.9435, "step": 14799 }, { "epoch": 0.68, "learning_rate": 1.1656910526394171e-08, "loss": 0.933, "step": 14800 }, { "epoch": 0.68, "learning_rate": 1.1553083289137512e-08, "loss": 0.9416, "step": 14801 }, { "epoch": 0.68, "learning_rate": 1.1449720246156093e-08, "loss": 0.9602, "step": 14802 }, { "epoch": 0.68, "learning_rate": 1.1346821402252739e-08, "loss": 0.8971, "step": 14803 }, { "epoch": 0.69, "learning_rate": 1.124438676221029e-08, "loss": 0.9109, "step": 14804 }, { "epoch": 0.69, "learning_rate": 1.1142416330789385e-08, "loss": 0.9186, "step": 14805 }, { "epoch": 0.69, "learning_rate": 1.1040910112727344e-08, "loss": 0.9832, "step": 14806 }, { "epoch": 0.69, "learning_rate": 1.0939868112742613e-08, "loss": 0.9291, "step": 14807 }, { "epoch": 0.69, "learning_rate": 1.0839290335531438e-08, "loss": 0.9578, "step": 14808 }, { "epoch": 0.69, "learning_rate": 1.0739176785766747e-08, "loss": 0.9236, "step": 14809 }, { "epoch": 0.69, "learning_rate": 1.0639527468102595e-08, "loss": 0.967, "step": 14810 }, { "epoch": 0.69, "learning_rate": 1.0540342387167501e-08, "loss": 0.9545, "step": 14811 }, { "epoch": 0.69, "learning_rate": 1.0441621547574443e-08, "loss": 0.9109, "step": 14812 }, { "epoch": 0.69, "learning_rate": 1.0343364953908642e-08, "loss": 0.9584, "step": 14813 }, { "epoch": 0.69, "learning_rate": 1.0245572610736443e-08, "loss": 0.9287, "step": 14814 }, { "epoch": 0.69, "learning_rate": 1.0148244522603102e-08, "loss": 0.9401, "step": 14815 }, { "epoch": 0.69, "learning_rate": 1.0051380694032774e-08, "loss": 0.9303, "step": 14816 }, { "epoch": 0.69, "learning_rate": 9.954981129526309e-09, "loss": 0.9103, "step": 14817 }, { "epoch": 0.69, "learning_rate": 9.859045833562341e-09, "loss": 0.9116, "step": 14818 }, { "epoch": 0.69, "learning_rate": 9.76357481060175e-09, "loss": 0.9174, "step": 14819 }, { "epoch": 0.69, "learning_rate": 9.668568065079876e-09, "loss": 0.9552, "step": 14820 }, { "epoch": 0.69, "learning_rate": 9.574025601410963e-09, "loss": 0.8979, "step": 14821 }, { "epoch": 0.69, "learning_rate": 9.479947423991498e-09, "loss": 0.9398, "step": 14822 }, { "epoch": 0.69, "learning_rate": 9.386333537191317e-09, "loss": 0.908, "step": 14823 }, { "epoch": 0.69, "learning_rate": 9.293183945361384e-09, "loss": 0.9578, "step": 14824 }, { "epoch": 0.69, "learning_rate": 9.200498652831569e-09, "loss": 0.9182, "step": 14825 }, { "epoch": 0.69, "learning_rate": 9.108277663907316e-09, "loss": 0.97, "step": 14826 }, { "epoch": 0.69, "learning_rate": 9.01652098287742e-09, "loss": 0.9137, "step": 14827 }, { "epoch": 0.69, "learning_rate": 8.925228614002913e-09, "loss": 0.8956, "step": 14828 }, { "epoch": 0.69, "learning_rate": 8.834400561529289e-09, "loss": 0.9399, "step": 14829 }, { "epoch": 0.69, "learning_rate": 8.744036829674284e-09, "loss": 0.8958, "step": 14830 }, { "epoch": 0.69, "learning_rate": 8.654137422641207e-09, "loss": 0.9556, "step": 14831 }, { "epoch": 0.69, "learning_rate": 8.564702344605602e-09, "loss": 0.9133, "step": 14832 }, { "epoch": 0.69, "learning_rate": 8.475731599724147e-09, "loss": 0.9315, "step": 14833 }, { "epoch": 0.69, "learning_rate": 8.38722519213131e-09, "loss": 0.9276, "step": 14834 }, { "epoch": 0.69, "learning_rate": 8.29918312594047e-09, "loss": 0.9178, "step": 14835 }, { "epoch": 0.69, "learning_rate": 8.211605405243906e-09, "loss": 0.9691, "step": 14836 }, { "epoch": 0.69, "learning_rate": 8.124492034111698e-09, "loss": 0.897, "step": 14837 }, { "epoch": 0.69, "learning_rate": 8.037843016590608e-09, "loss": 0.9328, "step": 14838 }, { "epoch": 0.69, "learning_rate": 7.951658356709634e-09, "loss": 0.9689, "step": 14839 }, { "epoch": 0.69, "learning_rate": 7.865938058472245e-09, "loss": 0.9283, "step": 14840 }, { "epoch": 0.69, "learning_rate": 7.780682125863026e-09, "loss": 0.9286, "step": 14841 }, { "epoch": 0.69, "learning_rate": 7.695890562844365e-09, "loss": 0.9153, "step": 14842 }, { "epoch": 0.69, "learning_rate": 7.611563373355335e-09, "loss": 0.9536, "step": 14843 }, { "epoch": 0.69, "learning_rate": 7.527700561317241e-09, "loss": 0.9332, "step": 14844 }, { "epoch": 0.69, "learning_rate": 7.444302130624747e-09, "loss": 0.8962, "step": 14845 }, { "epoch": 0.69, "learning_rate": 7.361368085155862e-09, "loss": 0.8993, "step": 14846 }, { "epoch": 0.69, "learning_rate": 7.278898428764169e-09, "loss": 0.9097, "step": 14847 }, { "epoch": 0.69, "learning_rate": 7.196893165282159e-09, "loss": 0.9565, "step": 14848 }, { "epoch": 0.69, "learning_rate": 7.115352298520117e-09, "loss": 0.9183, "step": 14849 }, { "epoch": 0.69, "learning_rate": 7.034275832268344e-09, "loss": 0.879, "step": 14850 }, { "epoch": 0.69, "learning_rate": 6.9536637702949385e-09, "loss": 0.9079, "step": 14851 }, { "epoch": 0.69, "learning_rate": 6.873516116345791e-09, "loss": 0.9259, "step": 14852 }, { "epoch": 0.69, "learning_rate": 6.793832874144591e-09, "loss": 0.9376, "step": 14853 }, { "epoch": 0.69, "learning_rate": 6.714614047396151e-09, "loss": 0.9345, "step": 14854 }, { "epoch": 0.69, "learning_rate": 6.635859639781972e-09, "loss": 0.8964, "step": 14855 }, { "epoch": 0.69, "learning_rate": 6.557569654960238e-09, "loss": 0.932, "step": 14856 }, { "epoch": 0.69, "learning_rate": 6.479744096571372e-09, "loss": 0.927, "step": 14857 }, { "epoch": 0.69, "learning_rate": 6.402382968231369e-09, "loss": 0.9404, "step": 14858 }, { "epoch": 0.69, "learning_rate": 6.3254862735351304e-09, "loss": 0.9167, "step": 14859 }, { "epoch": 0.69, "learning_rate": 6.249054016056466e-09, "loss": 0.9423, "step": 14860 }, { "epoch": 0.69, "learning_rate": 6.173086199346978e-09, "loss": 0.9342, "step": 14861 }, { "epoch": 0.69, "learning_rate": 6.097582826938286e-09, "loss": 0.905, "step": 14862 }, { "epoch": 0.69, "learning_rate": 6.022543902337586e-09, "loss": 0.9325, "step": 14863 }, { "epoch": 0.69, "learning_rate": 5.947969429034306e-09, "loss": 0.9451, "step": 14864 }, { "epoch": 0.69, "learning_rate": 5.873859410492344e-09, "loss": 0.9258, "step": 14865 }, { "epoch": 0.69, "learning_rate": 5.800213850157832e-09, "loss": 0.9249, "step": 14866 }, { "epoch": 0.69, "learning_rate": 5.727032751450256e-09, "loss": 0.9192, "step": 14867 }, { "epoch": 0.69, "learning_rate": 5.6543161177735615e-09, "loss": 0.9279, "step": 14868 }, { "epoch": 0.69, "learning_rate": 5.582063952505046e-09, "loss": 0.9022, "step": 14869 }, { "epoch": 0.69, "learning_rate": 5.510276259004244e-09, "loss": 0.901, "step": 14870 }, { "epoch": 0.69, "learning_rate": 5.438953040605155e-09, "loss": 0.9093, "step": 14871 }, { "epoch": 0.69, "learning_rate": 5.368094300625126e-09, "loss": 0.918, "step": 14872 }, { "epoch": 0.69, "learning_rate": 5.2977000423548585e-09, "loss": 0.9007, "step": 14873 }, { "epoch": 0.69, "learning_rate": 5.227770269066179e-09, "loss": 0.9362, "step": 14874 }, { "epoch": 0.69, "learning_rate": 5.158304984009821e-09, "loss": 0.9208, "step": 14875 }, { "epoch": 0.69, "learning_rate": 5.089304190414313e-09, "loss": 0.9557, "step": 14876 }, { "epoch": 0.69, "learning_rate": 5.0207678914848675e-09, "loss": 0.9615, "step": 14877 }, { "epoch": 0.69, "learning_rate": 4.952696090407827e-09, "loss": 0.9477, "step": 14878 }, { "epoch": 0.69, "learning_rate": 4.8850887903462155e-09, "loss": 0.9406, "step": 14879 }, { "epoch": 0.69, "learning_rate": 4.817945994441964e-09, "loss": 0.9129, "step": 14880 }, { "epoch": 0.69, "learning_rate": 4.751267705814799e-09, "loss": 0.9595, "step": 14881 }, { "epoch": 0.69, "learning_rate": 4.685053927564465e-09, "loss": 0.9019, "step": 14882 }, { "epoch": 0.69, "learning_rate": 4.619304662768498e-09, "loss": 0.9468, "step": 14883 }, { "epoch": 0.69, "learning_rate": 4.554019914481123e-09, "loss": 0.9203, "step": 14884 }, { "epoch": 0.69, "learning_rate": 4.489199685736579e-09, "loss": 0.9351, "step": 14885 }, { "epoch": 0.69, "learning_rate": 4.4248439795480105e-09, "loss": 0.9074, "step": 14886 }, { "epoch": 0.69, "learning_rate": 4.36095279890636e-09, "loss": 0.9355, "step": 14887 }, { "epoch": 0.69, "learning_rate": 4.297526146779252e-09, "loss": 0.9557, "step": 14888 }, { "epoch": 0.69, "learning_rate": 4.23456402611655e-09, "loss": 0.8996, "step": 14889 }, { "epoch": 0.69, "learning_rate": 4.172066439842582e-09, "loss": 0.9243, "step": 14890 }, { "epoch": 0.69, "learning_rate": 4.11003339086169e-09, "loss": 0.9248, "step": 14891 }, { "epoch": 0.69, "learning_rate": 4.048464882058234e-09, "loss": 0.96, "step": 14892 }, { "epoch": 0.69, "learning_rate": 3.987360916293259e-09, "loss": 0.9334, "step": 14893 }, { "epoch": 0.69, "learning_rate": 3.926721496404496e-09, "loss": 0.9491, "step": 14894 }, { "epoch": 0.69, "learning_rate": 3.866546625213019e-09, "loss": 0.9177, "step": 14895 }, { "epoch": 0.69, "learning_rate": 3.806836305512152e-09, "loss": 0.9384, "step": 14896 }, { "epoch": 0.69, "learning_rate": 3.747590540079671e-09, "loss": 0.9195, "step": 14897 }, { "epoch": 0.69, "learning_rate": 3.6888093316667097e-09, "loss": 0.8894, "step": 14898 }, { "epoch": 0.69, "learning_rate": 3.630492683006637e-09, "loss": 0.9521, "step": 14899 }, { "epoch": 0.69, "learning_rate": 3.5726405968083968e-09, "loss": 0.9245, "step": 14900 }, { "epoch": 0.69, "learning_rate": 3.5152530757620596e-09, "loss": 0.9052, "step": 14901 }, { "epoch": 0.69, "learning_rate": 3.458330122533271e-09, "loss": 0.9229, "step": 14902 }, { "epoch": 0.69, "learning_rate": 3.4018717397665823e-09, "loss": 0.8853, "step": 14903 }, { "epoch": 0.69, "learning_rate": 3.345877930088781e-09, "loss": 0.9395, "step": 14904 }, { "epoch": 0.69, "learning_rate": 3.2903486960988995e-09, "loss": 0.8879, "step": 14905 }, { "epoch": 0.69, "learning_rate": 3.2352840403804264e-09, "loss": 0.9018, "step": 14906 }, { "epoch": 0.69, "learning_rate": 3.1806839654890954e-09, "loss": 0.935, "step": 14907 }, { "epoch": 0.69, "learning_rate": 3.1265484739650966e-09, "loss": 0.9572, "step": 14908 }, { "epoch": 0.69, "learning_rate": 3.0728775683241953e-09, "loss": 0.8979, "step": 14909 }, { "epoch": 0.69, "learning_rate": 3.0196712510588422e-09, "loss": 0.9452, "step": 14910 }, { "epoch": 0.69, "learning_rate": 2.9669295246426146e-09, "loss": 0.9048, "step": 14911 }, { "epoch": 0.69, "learning_rate": 2.914652391526884e-09, "loss": 0.9614, "step": 14912 }, { "epoch": 0.69, "learning_rate": 2.8628398541408198e-09, "loss": 0.8786, "step": 14913 }, { "epoch": 0.69, "learning_rate": 2.8114919148924948e-09, "loss": 0.9365, "step": 14914 }, { "epoch": 0.69, "learning_rate": 2.760608576167778e-09, "loss": 0.9274, "step": 14915 }, { "epoch": 0.69, "learning_rate": 2.7101898403314453e-09, "loss": 0.9026, "step": 14916 }, { "epoch": 0.69, "learning_rate": 2.660235709727177e-09, "loss": 0.9309, "step": 14917 }, { "epoch": 0.69, "learning_rate": 2.6107461866764496e-09, "loss": 0.9042, "step": 14918 }, { "epoch": 0.69, "learning_rate": 2.5617212734785346e-09, "loss": 0.9113, "step": 14919 }, { "epoch": 0.69, "learning_rate": 2.5131609724116103e-09, "loss": 0.9575, "step": 14920 }, { "epoch": 0.69, "learning_rate": 2.4650652857338697e-09, "loss": 0.9035, "step": 14921 }, { "epoch": 0.69, "learning_rate": 2.417434215679082e-09, "loss": 0.9926, "step": 14922 }, { "epoch": 0.69, "learning_rate": 2.3702677644610316e-09, "loss": 0.9843, "step": 14923 }, { "epoch": 0.69, "learning_rate": 2.3235659342724093e-09, "loss": 0.9238, "step": 14924 }, { "epoch": 0.69, "learning_rate": 2.277328727283701e-09, "loss": 0.9203, "step": 14925 }, { "epoch": 0.69, "learning_rate": 2.2315561456420776e-09, "loss": 0.9087, "step": 14926 }, { "epoch": 0.69, "learning_rate": 2.1862481914769475e-09, "loss": 0.9175, "step": 14927 }, { "epoch": 0.69, "learning_rate": 2.1414048668910723e-09, "loss": 0.9533, "step": 14928 }, { "epoch": 0.69, "learning_rate": 2.0970261739716724e-09, "loss": 0.915, "step": 14929 }, { "epoch": 0.69, "learning_rate": 2.053112114779321e-09, "loss": 0.9434, "step": 14930 }, { "epoch": 0.69, "learning_rate": 2.009662691354608e-09, "loss": 0.9122, "step": 14931 }, { "epoch": 0.69, "learning_rate": 1.9666779057170293e-09, "loss": 0.9407, "step": 14932 }, { "epoch": 0.69, "learning_rate": 1.9241577598649864e-09, "loss": 0.9521, "step": 14933 }, { "epoch": 0.69, "learning_rate": 1.882102255774676e-09, "loss": 0.9552, "step": 14934 }, { "epoch": 0.69, "learning_rate": 1.8405113953989806e-09, "loss": 0.9647, "step": 14935 }, { "epoch": 0.69, "learning_rate": 1.7993851806719086e-09, "loss": 0.9131, "step": 14936 }, { "epoch": 0.69, "learning_rate": 1.758723613504154e-09, "loss": 0.9132, "step": 14937 }, { "epoch": 0.69, "learning_rate": 1.7185266957864266e-09, "loss": 0.938, "step": 14938 }, { "epoch": 0.69, "learning_rate": 1.6787944293861214e-09, "loss": 0.9258, "step": 14939 }, { "epoch": 0.69, "learning_rate": 1.6395268161495393e-09, "loss": 0.912, "step": 14940 }, { "epoch": 0.69, "learning_rate": 1.6007238579007768e-09, "loss": 0.9345, "step": 14941 }, { "epoch": 0.69, "learning_rate": 1.5623855564450564e-09, "loss": 0.9379, "step": 14942 }, { "epoch": 0.69, "learning_rate": 1.5245119135631758e-09, "loss": 0.9327, "step": 14943 }, { "epoch": 0.69, "learning_rate": 1.4871029310148387e-09, "loss": 0.9364, "step": 14944 }, { "epoch": 0.69, "learning_rate": 1.4501586105386545e-09, "loss": 0.9212, "step": 14945 }, { "epoch": 0.69, "learning_rate": 1.4136789538521378e-09, "loss": 0.9374, "step": 14946 }, { "epoch": 0.69, "learning_rate": 1.3776639626505994e-09, "loss": 0.9332, "step": 14947 }, { "epoch": 0.69, "learning_rate": 1.3421136386071454e-09, "loss": 0.9535, "step": 14948 }, { "epoch": 0.69, "learning_rate": 1.3070279833737875e-09, "loss": 0.908, "step": 14949 }, { "epoch": 0.69, "learning_rate": 1.2724069985825539e-09, "loss": 0.9641, "step": 14950 }, { "epoch": 0.69, "learning_rate": 1.2382506858399368e-09, "loss": 0.9199, "step": 14951 }, { "epoch": 0.69, "learning_rate": 1.2045590467357759e-09, "loss": 0.9667, "step": 14952 }, { "epoch": 0.69, "learning_rate": 1.1713320828332653e-09, "loss": 0.9137, "step": 14953 }, { "epoch": 0.7, "learning_rate": 1.1385697956789454e-09, "loss": 0.9615, "step": 14954 }, { "epoch": 0.7, "learning_rate": 1.1062721867949322e-09, "loss": 0.9155, "step": 14955 }, { "epoch": 0.7, "learning_rate": 1.074439257681137e-09, "loss": 0.9147, "step": 14956 }, { "epoch": 0.7, "learning_rate": 1.0430710098174867e-09, "loss": 0.9045, "step": 14957 }, { "epoch": 0.7, "learning_rate": 1.0121674446617048e-09, "loss": 0.9371, "step": 14958 }, { "epoch": 0.7, "learning_rate": 9.81728563649309e-10, "loss": 0.8951, "step": 14959 }, { "epoch": 0.7, "learning_rate": 9.517543681969443e-10, "loss": 0.9663, "step": 14960 }, { "epoch": 0.7, "learning_rate": 9.222448596957201e-10, "loss": 0.9391, "step": 14961 }, { "epoch": 0.7, "learning_rate": 8.932000395178719e-10, "loss": 0.9458, "step": 14962 }, { "epoch": 0.7, "learning_rate": 8.646199090123209e-10, "loss": 0.9216, "step": 14963 }, { "epoch": 0.7, "learning_rate": 8.365044695080037e-10, "loss": 0.9396, "step": 14964 }, { "epoch": 0.7, "learning_rate": 8.088537223116533e-10, "loss": 0.9201, "step": 14965 }, { "epoch": 0.7, "learning_rate": 7.816676687089076e-10, "loss": 0.902, "step": 14966 }, { "epoch": 0.7, "learning_rate": 7.5494630996098e-10, "loss": 0.944, "step": 14967 }, { "epoch": 0.7, "learning_rate": 7.286896473113203e-10, "loss": 0.9637, "step": 14968 }, { "epoch": 0.7, "learning_rate": 7.028976819800637e-10, "loss": 0.9296, "step": 14969 }, { "epoch": 0.7, "learning_rate": 6.77570415166251e-10, "loss": 0.9149, "step": 14970 }, { "epoch": 0.7, "learning_rate": 6.527078480456084e-10, "loss": 0.9701, "step": 14971 }, { "epoch": 0.7, "learning_rate": 6.283099817749882e-10, "loss": 0.9417, "step": 14972 }, { "epoch": 0.7, "learning_rate": 6.043768174868181e-10, "loss": 0.948, "step": 14973 }, { "epoch": 0.7, "learning_rate": 5.809083562946516e-10, "loss": 0.9654, "step": 14974 }, { "epoch": 0.7, "learning_rate": 5.579045992876175e-10, "loss": 0.9286, "step": 14975 }, { "epoch": 0.7, "learning_rate": 5.353655475359709e-10, "loss": 0.9623, "step": 14976 }, { "epoch": 0.7, "learning_rate": 5.132912020877623e-10, "loss": 0.9189, "step": 14977 }, { "epoch": 0.7, "learning_rate": 4.916815639677274e-10, "loss": 0.8939, "step": 14978 }, { "epoch": 0.7, "learning_rate": 4.705366341795081e-10, "loss": 0.9558, "step": 14979 }, { "epoch": 0.7, "learning_rate": 4.49856413707872e-10, "loss": 0.9995, "step": 14980 }, { "epoch": 0.7, "learning_rate": 4.296409035120519e-10, "loss": 0.9367, "step": 14981 }, { "epoch": 0.7, "learning_rate": 4.098901045324066e-10, "loss": 0.9546, "step": 14982 }, { "epoch": 0.7, "learning_rate": 3.9060401768598043e-10, "loss": 0.9452, "step": 14983 }, { "epoch": 0.7, "learning_rate": 3.717826438709438e-10, "loss": 0.9472, "step": 14984 }, { "epoch": 0.7, "learning_rate": 3.534259839599319e-10, "loss": 0.8751, "step": 14985 }, { "epoch": 0.7, "learning_rate": 3.3553403880670634e-10, "loss": 0.9712, "step": 14986 }, { "epoch": 0.7, "learning_rate": 3.181068092428241e-10, "loss": 0.9287, "step": 14987 }, { "epoch": 0.7, "learning_rate": 3.0114429607763783e-10, "loss": 0.8883, "step": 14988 }, { "epoch": 0.7, "learning_rate": 2.846465001016263e-10, "loss": 0.934, "step": 14989 }, { "epoch": 0.7, "learning_rate": 2.6861342207862294e-10, "loss": 0.9432, "step": 14990 }, { "epoch": 0.7, "learning_rate": 2.530450627558079e-10, "loss": 0.9379, "step": 14991 }, { "epoch": 0.7, "learning_rate": 2.3794142285482604e-10, "loss": 0.9519, "step": 14992 }, { "epoch": 0.7, "learning_rate": 2.233025030795588e-10, "loss": 0.9079, "step": 14993 }, { "epoch": 0.7, "learning_rate": 2.0912830410946273e-10, "loss": 0.9717, "step": 14994 }, { "epoch": 0.7, "learning_rate": 1.9541882660290001e-10, "loss": 0.9871, "step": 14995 }, { "epoch": 0.7, "learning_rate": 1.821740711971387e-10, "loss": 0.9133, "step": 14996 }, { "epoch": 0.7, "learning_rate": 1.6939403850835257e-10, "loss": 0.9136, "step": 14997 }, { "epoch": 0.7, "learning_rate": 1.5707872913051093e-10, "loss": 0.9028, "step": 14998 }, { "epoch": 0.7, "learning_rate": 1.452281436342684e-10, "loss": 0.9439, "step": 14999 }, { "epoch": 0.7, "learning_rate": 1.3384228257140586e-10, "loss": 0.8927, "step": 15000 }, { "epoch": 0.7, "learning_rate": 1.2292114647260988e-10, "loss": 0.951, "step": 15001 }, { "epoch": 0.7, "learning_rate": 1.1246473584303197e-10, "loss": 0.9537, "step": 15002 }, { "epoch": 0.7, "learning_rate": 1.0247305116894979e-10, "loss": 0.9163, "step": 15003 }, { "epoch": 0.7, "learning_rate": 9.294609291665702e-11, "loss": 0.9293, "step": 15004 }, { "epoch": 0.7, "learning_rate": 8.388386152691219e-11, "loss": 0.9264, "step": 15005 }, { "epoch": 0.7, "learning_rate": 7.528635742160007e-11, "loss": 0.9136, "step": 15006 }, { "epoch": 0.7, "learning_rate": 6.715358100040093e-11, "loss": 0.8755, "step": 15007 }, { "epoch": 0.7, "learning_rate": 5.94855326407906e-11, "loss": 0.9122, "step": 15008 }, { "epoch": 0.7, "learning_rate": 5.228221269915068e-11, "loss": 0.949, "step": 15009 }, { "epoch": 0.7, "learning_rate": 4.554362151076852e-11, "loss": 0.964, "step": 15010 }, { "epoch": 0.7, "learning_rate": 3.9269759388727015e-11, "loss": 0.9011, "step": 15011 }, { "epoch": 0.7, "learning_rate": 3.346062662390459e-11, "loss": 0.9167, "step": 15012 }, { "epoch": 0.7, "learning_rate": 2.811622348719567e-11, "loss": 0.9465, "step": 15013 }, { "epoch": 0.7, "learning_rate": 2.323655022617999e-11, "loss": 0.8985, "step": 15014 }, { "epoch": 0.7, "learning_rate": 1.8821607068453263e-11, "loss": 0.9304, "step": 15015 }, { "epoch": 0.7, "learning_rate": 1.487139421829653e-11, "loss": 0.9351, "step": 15016 }, { "epoch": 0.7, "learning_rate": 1.1385911860006815e-11, "loss": 0.9268, "step": 15017 }, { "epoch": 0.7, "learning_rate": 8.365160154566454e-12, "loss": 0.9138, "step": 15018 }, { "epoch": 0.7, "learning_rate": 5.809139244083994e-12, "loss": 0.8971, "step": 15019 }, { "epoch": 0.7, "learning_rate": 3.717849245132854e-12, "loss": 0.9095, "step": 15020 }, { "epoch": 0.7, "learning_rate": 2.091290257633105e-12, "loss": 0.8802, "step": 15021 }, { "epoch": 0.7, "learning_rate": 9.294623548594673e-13, "loss": 0.9258, "step": 15022 }, { "epoch": 0.7, "learning_rate": 2.323655912128686e-13, "loss": 0.8891, "step": 15023 }, { "epoch": 0.7, "learning_rate": 0.0, "loss": 0.9323, "step": 15024 }, { "epoch": 0.7, "step": 15024, "total_flos": 3.809287425747953e+24, "train_loss": 0.664228332345002, "train_runtime": 58818.6896, "train_samples_per_second": 261.559, "train_steps_per_second": 0.255 } ], "logging_steps": 1.0, "max_steps": 15024, "num_input_tokens_seen": 0, "num_train_epochs": 9223372036854775807, "save_steps": 752, "total_flos": 3.809287425747953e+24, "train_batch_size": 1, "trial_name": null, "trial_params": null }