diff --git "a/trainer_state.json" "b/trainer_state.json" new file mode 100644--- /dev/null +++ "b/trainer_state.json" @@ -0,0 +1,126016 @@ +{ + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 4.853247053385718, + "global_step": 21000, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.0, + "learning_rate": 4.999807410831215e-05, + "loss": 1.6632, + "step": 1 + }, + { + "epoch": 0.0, + "learning_rate": 4.99961482166243e-05, + "loss": 1.0347, + "step": 2 + }, + { + "epoch": 0.0, + "learning_rate": 4.9994222324936445e-05, + "loss": 1.4717, + "step": 3 + }, + { + "epoch": 0.0, + "learning_rate": 4.99922964332486e-05, + "loss": 1.2105, + "step": 4 + }, + { + "epoch": 0.0, + "learning_rate": 4.9990370541560747e-05, + "loss": 2.3836, + "step": 5 + }, + { + "epoch": 0.0, + "learning_rate": 4.9988444649872894e-05, + "loss": 0.8687, + "step": 6 + }, + { + "epoch": 0.0, + "learning_rate": 4.998651875818504e-05, + "loss": 0.7918, + "step": 7 + }, + { + "epoch": 0.0, + "learning_rate": 4.9984592866497196e-05, + "loss": 0.4978, + "step": 8 + }, + { + "epoch": 0.0, + "learning_rate": 4.9982666974809336e-05, + "loss": 0.3702, + "step": 9 + }, + { + "epoch": 0.0, + "learning_rate": 4.9980741083121484e-05, + "loss": 0.4377, + "step": 10 + }, + { + "epoch": 0.0, + "learning_rate": 4.997881519143364e-05, + "loss": 0.3483, + "step": 11 + }, + { + "epoch": 0.0, + "learning_rate": 4.9976889299745786e-05, + "loss": 0.62, + "step": 12 + }, + { + "epoch": 0.0, + "learning_rate": 4.997496340805793e-05, + "loss": 0.0, + "step": 13 + }, + { + "epoch": 0.0, + "learning_rate": 4.997303751637008e-05, + "loss": 0.8813, + "step": 14 + }, + { + "epoch": 0.0, + "learning_rate": 4.9971111624682235e-05, + "loss": 0.0, + "step": 15 + }, + { + "epoch": 0.0, + "learning_rate": 4.9969185732994376e-05, + "loss": 1.2558, + "step": 16 + }, + { + "epoch": 0.0, + "learning_rate": 4.996725984130653e-05, + "loss": 0.1384, + "step": 17 + }, + { + "epoch": 0.0, + "learning_rate": 4.996533394961868e-05, + "loss": 0.7114, + "step": 18 + }, + { + "epoch": 0.0, + "learning_rate": 4.9963408057930825e-05, + "loss": 0.6478, + "step": 19 + }, + { + "epoch": 0.0, + "learning_rate": 4.996148216624297e-05, + "loss": 0.2608, + "step": 20 + }, + { + "epoch": 0.0, + "learning_rate": 4.995955627455512e-05, + "loss": 1.3041, + "step": 21 + }, + { + "epoch": 0.01, + "learning_rate": 4.9957630382867274e-05, + "loss": 0.7206, + "step": 22 + }, + { + "epoch": 0.01, + "learning_rate": 4.9955704491179415e-05, + "loss": 0.2482, + "step": 23 + }, + { + "epoch": 0.01, + "learning_rate": 4.995377859949157e-05, + "loss": 0.9275, + "step": 24 + }, + { + "epoch": 0.01, + "learning_rate": 4.9951852707803716e-05, + "loss": 0.8095, + "step": 25 + }, + { + "epoch": 0.01, + "learning_rate": 4.9949926816115864e-05, + "loss": 1.1942, + "step": 26 + }, + { + "epoch": 0.01, + "learning_rate": 4.994800092442801e-05, + "loss": 0.0, + "step": 27 + }, + { + "epoch": 0.01, + "learning_rate": 4.994607503274016e-05, + "loss": 0.0, + "step": 28 + }, + { + "epoch": 0.01, + "learning_rate": 4.994414914105231e-05, + "loss": 0.0, + "step": 29 + }, + { + "epoch": 0.01, + "learning_rate": 4.9942223249364454e-05, + "loss": 0.4984, + "step": 30 + }, + { + "epoch": 0.01, + "learning_rate": 4.994029735767661e-05, + "loss": 0.6492, + "step": 31 + }, + { + "epoch": 0.01, + "learning_rate": 4.9938371465988756e-05, + "loss": 0.1573, + "step": 32 + }, + { + "epoch": 0.01, + "learning_rate": 4.99364455743009e-05, + "loss": 0.8866, + "step": 33 + }, + { + "epoch": 0.01, + "learning_rate": 4.993451968261305e-05, + "loss": 0.6603, + "step": 34 + }, + { + "epoch": 0.01, + "learning_rate": 4.9932593790925205e-05, + "loss": 0.1777, + "step": 35 + }, + { + "epoch": 0.01, + "learning_rate": 4.993066789923735e-05, + "loss": 0.401, + "step": 36 + }, + { + "epoch": 0.01, + "learning_rate": 4.99287420075495e-05, + "loss": 0.9486, + "step": 37 + }, + { + "epoch": 0.01, + "learning_rate": 4.992681611586165e-05, + "loss": 1.3843, + "step": 38 + }, + { + "epoch": 0.01, + "learning_rate": 4.9924890224173795e-05, + "loss": 0.6277, + "step": 39 + }, + { + "epoch": 0.01, + "learning_rate": 4.992296433248594e-05, + "loss": 0.4689, + "step": 40 + }, + { + "epoch": 0.01, + "learning_rate": 4.992103844079809e-05, + "loss": 0.4762, + "step": 41 + }, + { + "epoch": 0.01, + "learning_rate": 4.9919112549110244e-05, + "loss": 0.3091, + "step": 42 + }, + { + "epoch": 0.01, + "learning_rate": 4.9917186657422385e-05, + "loss": 0.7119, + "step": 43 + }, + { + "epoch": 0.01, + "learning_rate": 4.991526076573454e-05, + "loss": 0.0, + "step": 44 + }, + { + "epoch": 0.01, + "learning_rate": 4.9913334874046686e-05, + "loss": 0.5097, + "step": 45 + }, + { + "epoch": 0.01, + "learning_rate": 4.9911408982358834e-05, + "loss": 0.0, + "step": 46 + }, + { + "epoch": 0.01, + "learning_rate": 4.990948309067098e-05, + "loss": 0.1624, + "step": 47 + }, + { + "epoch": 0.01, + "learning_rate": 4.990755719898313e-05, + "loss": 1.444, + "step": 48 + }, + { + "epoch": 0.01, + "learning_rate": 4.990563130729528e-05, + "loss": 0.6973, + "step": 49 + }, + { + "epoch": 0.01, + "learning_rate": 4.9903705415607424e-05, + "loss": 0.2316, + "step": 50 + }, + { + "epoch": 0.01, + "learning_rate": 4.990177952391958e-05, + "loss": 0.0, + "step": 51 + }, + { + "epoch": 0.01, + "learning_rate": 4.9899853632231725e-05, + "loss": 0.0, + "step": 52 + }, + { + "epoch": 0.01, + "learning_rate": 4.989792774054387e-05, + "loss": 0.6387, + "step": 53 + }, + { + "epoch": 0.01, + "learning_rate": 4.989600184885602e-05, + "loss": 0.0, + "step": 54 + }, + { + "epoch": 0.01, + "learning_rate": 4.9894075957168175e-05, + "loss": 0.2382, + "step": 55 + }, + { + "epoch": 0.01, + "learning_rate": 4.989215006548032e-05, + "loss": 0.1615, + "step": 56 + }, + { + "epoch": 0.01, + "learning_rate": 4.989022417379246e-05, + "loss": 0.2565, + "step": 57 + }, + { + "epoch": 0.01, + "learning_rate": 4.988829828210462e-05, + "loss": 0.4156, + "step": 58 + }, + { + "epoch": 0.01, + "learning_rate": 4.9886372390416765e-05, + "loss": 0.4102, + "step": 59 + }, + { + "epoch": 0.01, + "learning_rate": 4.988444649872891e-05, + "loss": 0.0, + "step": 60 + }, + { + "epoch": 0.01, + "learning_rate": 4.988252060704106e-05, + "loss": 0.2863, + "step": 61 + }, + { + "epoch": 0.01, + "learning_rate": 4.9880594715353214e-05, + "loss": 0.5744, + "step": 62 + }, + { + "epoch": 0.01, + "learning_rate": 4.987866882366536e-05, + "loss": 0.0, + "step": 63 + }, + { + "epoch": 0.01, + "learning_rate": 4.987674293197751e-05, + "loss": 0.0, + "step": 64 + }, + { + "epoch": 0.02, + "learning_rate": 4.9874817040289656e-05, + "loss": 0.329, + "step": 65 + }, + { + "epoch": 0.02, + "learning_rate": 4.9872891148601804e-05, + "loss": 0.0678, + "step": 66 + }, + { + "epoch": 0.02, + "learning_rate": 4.987096525691395e-05, + "loss": 0.5001, + "step": 67 + }, + { + "epoch": 0.02, + "learning_rate": 4.98690393652261e-05, + "loss": 0.1551, + "step": 68 + }, + { + "epoch": 0.02, + "learning_rate": 4.986711347353825e-05, + "loss": 0.0515, + "step": 69 + }, + { + "epoch": 0.02, + "learning_rate": 4.98651875818504e-05, + "loss": 0.679, + "step": 70 + }, + { + "epoch": 0.02, + "learning_rate": 4.986326169016255e-05, + "loss": 0.0902, + "step": 71 + }, + { + "epoch": 0.02, + "learning_rate": 4.9861335798474695e-05, + "loss": 0.4396, + "step": 72 + }, + { + "epoch": 0.02, + "learning_rate": 4.985940990678685e-05, + "loss": 0.3086, + "step": 73 + }, + { + "epoch": 0.02, + "learning_rate": 4.985748401509899e-05, + "loss": 0.0693, + "step": 74 + }, + { + "epoch": 0.02, + "learning_rate": 4.985555812341114e-05, + "loss": 0.0, + "step": 75 + }, + { + "epoch": 0.02, + "learning_rate": 4.985363223172329e-05, + "loss": 0.7421, + "step": 76 + }, + { + "epoch": 0.02, + "learning_rate": 4.985170634003544e-05, + "loss": 0.1896, + "step": 77 + }, + { + "epoch": 0.02, + "learning_rate": 4.984978044834759e-05, + "loss": 0.4098, + "step": 78 + }, + { + "epoch": 0.02, + "learning_rate": 4.9847854556659734e-05, + "loss": 0.0, + "step": 79 + }, + { + "epoch": 0.02, + "learning_rate": 4.984592866497189e-05, + "loss": 0.3049, + "step": 80 + }, + { + "epoch": 0.02, + "learning_rate": 4.984400277328403e-05, + "loss": 0.4096, + "step": 81 + }, + { + "epoch": 0.02, + "learning_rate": 4.9842076881596184e-05, + "loss": 0.3638, + "step": 82 + }, + { + "epoch": 0.02, + "learning_rate": 4.984015098990833e-05, + "loss": 0.4074, + "step": 83 + }, + { + "epoch": 0.02, + "learning_rate": 4.983822509822048e-05, + "loss": 0.1634, + "step": 84 + }, + { + "epoch": 0.02, + "learning_rate": 4.9836299206532626e-05, + "loss": 0.7539, + "step": 85 + }, + { + "epoch": 0.02, + "learning_rate": 4.9834373314844774e-05, + "loss": 0.2891, + "step": 86 + }, + { + "epoch": 0.02, + "learning_rate": 4.983244742315693e-05, + "loss": 0.1979, + "step": 87 + }, + { + "epoch": 0.02, + "learning_rate": 4.983052153146907e-05, + "loss": 0.1315, + "step": 88 + }, + { + "epoch": 0.02, + "learning_rate": 4.982859563978122e-05, + "loss": 0.0675, + "step": 89 + }, + { + "epoch": 0.02, + "learning_rate": 4.982666974809337e-05, + "loss": 0.1878, + "step": 90 + }, + { + "epoch": 0.02, + "learning_rate": 4.982474385640552e-05, + "loss": 0.4305, + "step": 91 + }, + { + "epoch": 0.02, + "learning_rate": 4.9822817964717665e-05, + "loss": 0.2558, + "step": 92 + }, + { + "epoch": 0.02, + "learning_rate": 4.982089207302982e-05, + "loss": 0.0001, + "step": 93 + }, + { + "epoch": 0.02, + "learning_rate": 4.981896618134197e-05, + "loss": 0.2466, + "step": 94 + }, + { + "epoch": 0.02, + "learning_rate": 4.981704028965411e-05, + "loss": 0.446, + "step": 95 + }, + { + "epoch": 0.02, + "learning_rate": 4.981511439796626e-05, + "loss": 0.5128, + "step": 96 + }, + { + "epoch": 0.02, + "learning_rate": 4.981318850627841e-05, + "loss": 0.1276, + "step": 97 + }, + { + "epoch": 0.02, + "learning_rate": 4.981126261459056e-05, + "loss": 0.1664, + "step": 98 + }, + { + "epoch": 0.02, + "learning_rate": 4.9809336722902704e-05, + "loss": 0.2994, + "step": 99 + }, + { + "epoch": 0.02, + "learning_rate": 4.980741083121486e-05, + "loss": 0.1304, + "step": 100 + }, + { + "epoch": 0.02, + "learning_rate": 4.9805484939527006e-05, + "loss": 0.0507, + "step": 101 + }, + { + "epoch": 0.02, + "learning_rate": 4.9803559047839154e-05, + "loss": 0.5184, + "step": 102 + }, + { + "epoch": 0.02, + "learning_rate": 4.98016331561513e-05, + "loss": 0.2861, + "step": 103 + }, + { + "epoch": 0.02, + "learning_rate": 4.979970726446345e-05, + "loss": 0.4419, + "step": 104 + }, + { + "epoch": 0.02, + "learning_rate": 4.9797781372775596e-05, + "loss": 0.3939, + "step": 105 + }, + { + "epoch": 0.02, + "learning_rate": 4.9795855481087743e-05, + "loss": 0.2823, + "step": 106 + }, + { + "epoch": 0.02, + "learning_rate": 4.97939295893999e-05, + "loss": 0.0003, + "step": 107 + }, + { + "epoch": 0.02, + "learning_rate": 4.9792003697712045e-05, + "loss": 0.104, + "step": 108 + }, + { + "epoch": 0.03, + "learning_rate": 4.979007780602419e-05, + "loss": 0.0867, + "step": 109 + }, + { + "epoch": 0.03, + "learning_rate": 4.978815191433634e-05, + "loss": 0.143, + "step": 110 + }, + { + "epoch": 0.03, + "learning_rate": 4.9786226022648494e-05, + "loss": 0.2694, + "step": 111 + }, + { + "epoch": 0.03, + "learning_rate": 4.9784300130960635e-05, + "loss": 0.1833, + "step": 112 + }, + { + "epoch": 0.03, + "learning_rate": 4.978237423927278e-05, + "loss": 0.1604, + "step": 113 + }, + { + "epoch": 0.03, + "learning_rate": 4.978044834758494e-05, + "loss": 0.0001, + "step": 114 + }, + { + "epoch": 0.03, + "learning_rate": 4.977852245589708e-05, + "loss": 0.3558, + "step": 115 + }, + { + "epoch": 0.03, + "learning_rate": 4.977659656420923e-05, + "loss": 0.2729, + "step": 116 + }, + { + "epoch": 0.03, + "learning_rate": 4.977467067252138e-05, + "loss": 0.2357, + "step": 117 + }, + { + "epoch": 0.03, + "learning_rate": 4.977274478083353e-05, + "loss": 0.1828, + "step": 118 + }, + { + "epoch": 0.03, + "learning_rate": 4.9770818889145674e-05, + "loss": 0.0862, + "step": 119 + }, + { + "epoch": 0.03, + "learning_rate": 4.976889299745783e-05, + "loss": 0.0163, + "step": 120 + }, + { + "epoch": 0.03, + "learning_rate": 4.9766967105769976e-05, + "loss": 0.0002, + "step": 121 + }, + { + "epoch": 0.03, + "learning_rate": 4.976504121408212e-05, + "loss": 0.4404, + "step": 122 + }, + { + "epoch": 0.03, + "learning_rate": 4.976311532239427e-05, + "loss": 0.2727, + "step": 123 + }, + { + "epoch": 0.03, + "learning_rate": 4.976118943070642e-05, + "loss": 0.1448, + "step": 124 + }, + { + "epoch": 0.03, + "learning_rate": 4.9759263539018566e-05, + "loss": 0.3424, + "step": 125 + }, + { + "epoch": 0.03, + "learning_rate": 4.975733764733071e-05, + "loss": 0.2117, + "step": 126 + }, + { + "epoch": 0.03, + "learning_rate": 4.975541175564287e-05, + "loss": 0.1865, + "step": 127 + }, + { + "epoch": 0.03, + "learning_rate": 4.9753485863955015e-05, + "loss": 0.1131, + "step": 128 + }, + { + "epoch": 0.03, + "learning_rate": 4.975155997226716e-05, + "loss": 0.1974, + "step": 129 + }, + { + "epoch": 0.03, + "learning_rate": 4.974963408057931e-05, + "loss": 0.1909, + "step": 130 + }, + { + "epoch": 0.03, + "learning_rate": 4.974770818889146e-05, + "loss": 0.076, + "step": 131 + }, + { + "epoch": 0.03, + "learning_rate": 4.9745782297203605e-05, + "loss": 0.4036, + "step": 132 + }, + { + "epoch": 0.03, + "learning_rate": 4.974385640551575e-05, + "loss": 0.1159, + "step": 133 + }, + { + "epoch": 0.03, + "learning_rate": 4.974193051382791e-05, + "loss": 0.141, + "step": 134 + }, + { + "epoch": 0.03, + "learning_rate": 4.9740004622140054e-05, + "loss": 0.4502, + "step": 135 + }, + { + "epoch": 0.03, + "learning_rate": 4.97380787304522e-05, + "loss": 0.3745, + "step": 136 + }, + { + "epoch": 0.03, + "learning_rate": 4.973615283876435e-05, + "loss": 0.5372, + "step": 137 + }, + { + "epoch": 0.03, + "learning_rate": 4.9734226947076503e-05, + "loss": 0.2066, + "step": 138 + }, + { + "epoch": 0.03, + "learning_rate": 4.9732301055388644e-05, + "loss": 0.1305, + "step": 139 + }, + { + "epoch": 0.03, + "learning_rate": 4.973037516370079e-05, + "loss": 0.0002, + "step": 140 + }, + { + "epoch": 0.03, + "learning_rate": 4.9728449272012946e-05, + "loss": 0.1664, + "step": 141 + }, + { + "epoch": 0.03, + "learning_rate": 4.972652338032509e-05, + "loss": 0.199, + "step": 142 + }, + { + "epoch": 0.03, + "learning_rate": 4.972459748863724e-05, + "loss": 0.4112, + "step": 143 + }, + { + "epoch": 0.03, + "learning_rate": 4.972267159694939e-05, + "loss": 0.3263, + "step": 144 + }, + { + "epoch": 0.03, + "learning_rate": 4.972074570526154e-05, + "loss": 0.2519, + "step": 145 + }, + { + "epoch": 0.03, + "learning_rate": 4.971881981357368e-05, + "loss": 0.0, + "step": 146 + }, + { + "epoch": 0.03, + "learning_rate": 4.971689392188584e-05, + "loss": 0.0, + "step": 147 + }, + { + "epoch": 0.03, + "learning_rate": 4.9714968030197985e-05, + "loss": 0.0001, + "step": 148 + }, + { + "epoch": 0.03, + "learning_rate": 4.971304213851013e-05, + "loss": 0.3544, + "step": 149 + }, + { + "epoch": 0.03, + "learning_rate": 4.971111624682228e-05, + "loss": 0.2228, + "step": 150 + }, + { + "epoch": 0.03, + "learning_rate": 4.970919035513443e-05, + "loss": 0.0734, + "step": 151 + }, + { + "epoch": 0.04, + "learning_rate": 4.970726446344658e-05, + "loss": 0.0001, + "step": 152 + }, + { + "epoch": 0.04, + "learning_rate": 4.970533857175872e-05, + "loss": 0.3112, + "step": 153 + }, + { + "epoch": 0.04, + "learning_rate": 4.9703412680070877e-05, + "loss": 0.1258, + "step": 154 + }, + { + "epoch": 0.04, + "learning_rate": 4.9701486788383024e-05, + "loss": 0.0, + "step": 155 + }, + { + "epoch": 0.04, + "learning_rate": 4.969956089669517e-05, + "loss": 0.1877, + "step": 156 + }, + { + "epoch": 0.04, + "learning_rate": 4.969763500500732e-05, + "loss": 0.1564, + "step": 157 + }, + { + "epoch": 0.04, + "learning_rate": 4.969570911331947e-05, + "loss": 0.1619, + "step": 158 + }, + { + "epoch": 0.04, + "learning_rate": 4.969378322163162e-05, + "loss": 0.3178, + "step": 159 + }, + { + "epoch": 0.04, + "learning_rate": 4.969185732994376e-05, + "loss": 0.243, + "step": 160 + }, + { + "epoch": 0.04, + "learning_rate": 4.9689931438255916e-05, + "loss": 0.1624, + "step": 161 + }, + { + "epoch": 0.04, + "learning_rate": 4.968800554656806e-05, + "loss": 0.1565, + "step": 162 + }, + { + "epoch": 0.04, + "learning_rate": 4.968607965488021e-05, + "loss": 0.3542, + "step": 163 + }, + { + "epoch": 0.04, + "learning_rate": 4.968415376319236e-05, + "loss": 0.1486, + "step": 164 + }, + { + "epoch": 0.04, + "learning_rate": 4.968222787150451e-05, + "loss": 0.0572, + "step": 165 + }, + { + "epoch": 0.04, + "learning_rate": 4.968030197981666e-05, + "loss": 0.0, + "step": 166 + }, + { + "epoch": 0.04, + "learning_rate": 4.967837608812881e-05, + "loss": 0.1755, + "step": 167 + }, + { + "epoch": 0.04, + "learning_rate": 4.9676450196440955e-05, + "loss": 0.2842, + "step": 168 + }, + { + "epoch": 0.04, + "learning_rate": 4.96745243047531e-05, + "loss": 0.2089, + "step": 169 + }, + { + "epoch": 0.04, + "learning_rate": 4.967259841306525e-05, + "loss": 0.0296, + "step": 170 + }, + { + "epoch": 0.04, + "learning_rate": 4.96706725213774e-05, + "loss": 0.0238, + "step": 171 + }, + { + "epoch": 0.04, + "learning_rate": 4.966874662968955e-05, + "loss": 0.0587, + "step": 172 + }, + { + "epoch": 0.04, + "learning_rate": 4.96668207380017e-05, + "loss": 0.3053, + "step": 173 + }, + { + "epoch": 0.04, + "learning_rate": 4.9664894846313846e-05, + "loss": 0.1309, + "step": 174 + }, + { + "epoch": 0.04, + "learning_rate": 4.9662968954625994e-05, + "loss": 0.0407, + "step": 175 + }, + { + "epoch": 0.04, + "learning_rate": 4.966104306293815e-05, + "loss": 0.1515, + "step": 176 + }, + { + "epoch": 0.04, + "learning_rate": 4.965911717125029e-05, + "loss": 0.1587, + "step": 177 + }, + { + "epoch": 0.04, + "learning_rate": 4.9657191279562436e-05, + "loss": 0.003, + "step": 178 + }, + { + "epoch": 0.04, + "learning_rate": 4.965526538787459e-05, + "loss": 0.0239, + "step": 179 + }, + { + "epoch": 0.04, + "learning_rate": 4.965333949618674e-05, + "loss": 0.475, + "step": 180 + }, + { + "epoch": 0.04, + "learning_rate": 4.9651413604498886e-05, + "loss": 0.284, + "step": 181 + }, + { + "epoch": 0.04, + "learning_rate": 4.964948771281103e-05, + "loss": 0.0775, + "step": 182 + }, + { + "epoch": 0.04, + "learning_rate": 4.964756182112319e-05, + "loss": 0.2412, + "step": 183 + }, + { + "epoch": 0.04, + "learning_rate": 4.964563592943533e-05, + "loss": 0.3058, + "step": 184 + }, + { + "epoch": 0.04, + "learning_rate": 4.964371003774748e-05, + "loss": 0.1608, + "step": 185 + }, + { + "epoch": 0.04, + "learning_rate": 4.964178414605963e-05, + "loss": 0.0546, + "step": 186 + }, + { + "epoch": 0.04, + "learning_rate": 4.963985825437177e-05, + "loss": 0.0007, + "step": 187 + }, + { + "epoch": 0.04, + "learning_rate": 4.9637932362683925e-05, + "loss": 0.2202, + "step": 188 + }, + { + "epoch": 0.04, + "learning_rate": 4.963600647099607e-05, + "loss": 0.1382, + "step": 189 + }, + { + "epoch": 0.04, + "learning_rate": 4.9634080579308226e-05, + "loss": 0.1087, + "step": 190 + }, + { + "epoch": 0.04, + "learning_rate": 4.963215468762037e-05, + "loss": 0.1429, + "step": 191 + }, + { + "epoch": 0.04, + "learning_rate": 4.963022879593252e-05, + "loss": 0.2475, + "step": 192 + }, + { + "epoch": 0.04, + "learning_rate": 4.962830290424467e-05, + "loss": 0.3046, + "step": 193 + }, + { + "epoch": 0.04, + "learning_rate": 4.9626377012556816e-05, + "loss": 0.1171, + "step": 194 + }, + { + "epoch": 0.05, + "learning_rate": 4.9624451120868964e-05, + "loss": 0.0001, + "step": 195 + }, + { + "epoch": 0.05, + "learning_rate": 4.962252522918111e-05, + "loss": 0.0621, + "step": 196 + }, + { + "epoch": 0.05, + "learning_rate": 4.962059933749326e-05, + "loss": 0.1073, + "step": 197 + }, + { + "epoch": 0.05, + "learning_rate": 4.9618673445805406e-05, + "loss": 0.0861, + "step": 198 + }, + { + "epoch": 0.05, + "learning_rate": 4.961674755411756e-05, + "loss": 0.0, + "step": 199 + }, + { + "epoch": 0.05, + "learning_rate": 4.961482166242971e-05, + "loss": 0.0612, + "step": 200 + }, + { + "epoch": 0.05, + "learning_rate": 4.9612895770741855e-05, + "loss": 0.1562, + "step": 201 + }, + { + "epoch": 0.05, + "learning_rate": 4.9610969879054e-05, + "loss": 0.1623, + "step": 202 + }, + { + "epoch": 0.05, + "learning_rate": 4.960904398736616e-05, + "loss": 0.1193, + "step": 203 + }, + { + "epoch": 0.05, + "learning_rate": 4.96071180956783e-05, + "loss": 0.0081, + "step": 204 + }, + { + "epoch": 0.05, + "learning_rate": 4.960519220399045e-05, + "loss": 0.0013, + "step": 205 + }, + { + "epoch": 0.05, + "learning_rate": 4.96032663123026e-05, + "loss": 0.0024, + "step": 206 + }, + { + "epoch": 0.05, + "learning_rate": 4.960134042061475e-05, + "loss": 0.4634, + "step": 207 + }, + { + "epoch": 0.05, + "learning_rate": 4.9599414528926895e-05, + "loss": 0.042, + "step": 208 + }, + { + "epoch": 0.05, + "learning_rate": 4.959748863723904e-05, + "loss": 0.0, + "step": 209 + }, + { + "epoch": 0.05, + "learning_rate": 4.9595562745551196e-05, + "loss": 0.1199, + "step": 210 + }, + { + "epoch": 0.05, + "learning_rate": 4.959363685386334e-05, + "loss": 0.1193, + "step": 211 + }, + { + "epoch": 0.05, + "learning_rate": 4.959171096217549e-05, + "loss": 0.048, + "step": 212 + }, + { + "epoch": 0.05, + "learning_rate": 4.958978507048764e-05, + "loss": 0.0088, + "step": 213 + }, + { + "epoch": 0.05, + "learning_rate": 4.9587859178799786e-05, + "loss": 0.6555, + "step": 214 + }, + { + "epoch": 0.05, + "learning_rate": 4.9585933287111934e-05, + "loss": 0.1461, + "step": 215 + }, + { + "epoch": 0.05, + "learning_rate": 4.958400739542408e-05, + "loss": 0.2031, + "step": 216 + }, + { + "epoch": 0.05, + "learning_rate": 4.9582081503736235e-05, + "loss": 0.0896, + "step": 217 + }, + { + "epoch": 0.05, + "learning_rate": 4.9580155612048376e-05, + "loss": 0.0017, + "step": 218 + }, + { + "epoch": 0.05, + "learning_rate": 4.957822972036053e-05, + "loss": 0.0279, + "step": 219 + }, + { + "epoch": 0.05, + "learning_rate": 4.957630382867268e-05, + "loss": 0.0672, + "step": 220 + }, + { + "epoch": 0.05, + "learning_rate": 4.9574377936984825e-05, + "loss": 0.0001, + "step": 221 + }, + { + "epoch": 0.05, + "learning_rate": 4.957245204529697e-05, + "loss": 0.0167, + "step": 222 + }, + { + "epoch": 0.05, + "learning_rate": 4.957052615360913e-05, + "loss": 0.1114, + "step": 223 + }, + { + "epoch": 0.05, + "learning_rate": 4.9568600261921275e-05, + "loss": 0.0, + "step": 224 + }, + { + "epoch": 0.05, + "learning_rate": 4.9566674370233415e-05, + "loss": 0.043, + "step": 225 + }, + { + "epoch": 0.05, + "learning_rate": 4.956474847854557e-05, + "loss": 0.0357, + "step": 226 + }, + { + "epoch": 0.05, + "learning_rate": 4.956282258685772e-05, + "loss": 0.0, + "step": 227 + }, + { + "epoch": 0.05, + "learning_rate": 4.9560896695169864e-05, + "loss": 0.0014, + "step": 228 + }, + { + "epoch": 0.05, + "learning_rate": 4.955897080348201e-05, + "loss": 0.0391, + "step": 229 + }, + { + "epoch": 0.05, + "learning_rate": 4.9557044911794166e-05, + "loss": 0.1415, + "step": 230 + }, + { + "epoch": 0.05, + "learning_rate": 4.9555119020106314e-05, + "loss": 0.1755, + "step": 231 + }, + { + "epoch": 0.05, + "learning_rate": 4.955319312841846e-05, + "loss": 0.0033, + "step": 232 + }, + { + "epoch": 0.05, + "learning_rate": 4.955126723673061e-05, + "loss": 0.0897, + "step": 233 + }, + { + "epoch": 0.05, + "learning_rate": 4.9549341345042756e-05, + "loss": 0.0, + "step": 234 + }, + { + "epoch": 0.05, + "learning_rate": 4.9547415453354904e-05, + "loss": 0.0998, + "step": 235 + }, + { + "epoch": 0.05, + "learning_rate": 4.954548956166705e-05, + "loss": 0.0439, + "step": 236 + }, + { + "epoch": 0.05, + "learning_rate": 4.9543563669979205e-05, + "loss": 0.0676, + "step": 237 + }, + { + "epoch": 0.06, + "learning_rate": 4.954163777829135e-05, + "loss": 0.2848, + "step": 238 + }, + { + "epoch": 0.06, + "learning_rate": 4.95397118866035e-05, + "loss": 0.209, + "step": 239 + }, + { + "epoch": 0.06, + "learning_rate": 4.953778599491565e-05, + "loss": 0.0, + "step": 240 + }, + { + "epoch": 0.06, + "learning_rate": 4.95358601032278e-05, + "loss": 0.1318, + "step": 241 + }, + { + "epoch": 0.06, + "learning_rate": 4.953393421153994e-05, + "loss": 0.0755, + "step": 242 + }, + { + "epoch": 0.06, + "learning_rate": 4.953200831985209e-05, + "loss": 0.0459, + "step": 243 + }, + { + "epoch": 0.06, + "learning_rate": 4.9530082428164244e-05, + "loss": 0.1674, + "step": 244 + }, + { + "epoch": 0.06, + "learning_rate": 4.952815653647639e-05, + "loss": 0.0009, + "step": 245 + }, + { + "epoch": 0.06, + "learning_rate": 4.952623064478854e-05, + "loss": 0.0, + "step": 246 + }, + { + "epoch": 0.06, + "learning_rate": 4.952430475310069e-05, + "loss": 0.16, + "step": 247 + }, + { + "epoch": 0.06, + "learning_rate": 4.952237886141284e-05, + "loss": 0.0, + "step": 248 + }, + { + "epoch": 0.06, + "learning_rate": 4.952045296972498e-05, + "loss": 0.1262, + "step": 249 + }, + { + "epoch": 0.06, + "learning_rate": 4.9518527078037136e-05, + "loss": 0.0659, + "step": 250 + }, + { + "epoch": 0.06, + "learning_rate": 4.9516601186349284e-05, + "loss": 0.0482, + "step": 251 + }, + { + "epoch": 0.06, + "learning_rate": 4.951467529466143e-05, + "loss": 0.0001, + "step": 252 + }, + { + "epoch": 0.06, + "learning_rate": 4.951274940297358e-05, + "loss": 0.0251, + "step": 253 + }, + { + "epoch": 0.06, + "learning_rate": 4.9510823511285726e-05, + "loss": 0.1693, + "step": 254 + }, + { + "epoch": 0.06, + "learning_rate": 4.950889761959788e-05, + "loss": 0.0, + "step": 255 + }, + { + "epoch": 0.06, + "learning_rate": 4.950697172791002e-05, + "loss": 0.0372, + "step": 256 + }, + { + "epoch": 0.06, + "learning_rate": 4.9505045836222175e-05, + "loss": 0.0185, + "step": 257 + }, + { + "epoch": 0.06, + "learning_rate": 4.950311994453432e-05, + "loss": 0.0748, + "step": 258 + }, + { + "epoch": 0.06, + "learning_rate": 4.950119405284647e-05, + "loss": 0.0393, + "step": 259 + }, + { + "epoch": 0.06, + "learning_rate": 4.949926816115862e-05, + "loss": 0.2993, + "step": 260 + }, + { + "epoch": 0.06, + "learning_rate": 4.949734226947077e-05, + "loss": 0.0878, + "step": 261 + }, + { + "epoch": 0.06, + "learning_rate": 4.949541637778292e-05, + "loss": 0.2308, + "step": 262 + }, + { + "epoch": 0.06, + "learning_rate": 4.949349048609506e-05, + "loss": 0.4115, + "step": 263 + }, + { + "epoch": 0.06, + "learning_rate": 4.9491564594407214e-05, + "loss": 0.0306, + "step": 264 + }, + { + "epoch": 0.06, + "learning_rate": 4.948963870271936e-05, + "loss": 0.0148, + "step": 265 + }, + { + "epoch": 0.06, + "learning_rate": 4.948771281103151e-05, + "loss": 0.071, + "step": 266 + }, + { + "epoch": 0.06, + "learning_rate": 4.948578691934366e-05, + "loss": 0.0646, + "step": 267 + }, + { + "epoch": 0.06, + "learning_rate": 4.948386102765581e-05, + "loss": 0.0637, + "step": 268 + }, + { + "epoch": 0.06, + "learning_rate": 4.948193513596795e-05, + "loss": 0.0289, + "step": 269 + }, + { + "epoch": 0.06, + "learning_rate": 4.9480009244280106e-05, + "loss": 0.0359, + "step": 270 + }, + { + "epoch": 0.06, + "learning_rate": 4.9478083352592253e-05, + "loss": 0.0106, + "step": 271 + }, + { + "epoch": 0.06, + "learning_rate": 4.94761574609044e-05, + "loss": 0.1125, + "step": 272 + }, + { + "epoch": 0.06, + "learning_rate": 4.947423156921655e-05, + "loss": 0.0907, + "step": 273 + }, + { + "epoch": 0.06, + "learning_rate": 4.9472305677528696e-05, + "loss": 0.0715, + "step": 274 + }, + { + "epoch": 0.06, + "learning_rate": 4.947037978584085e-05, + "loss": 0.0004, + "step": 275 + }, + { + "epoch": 0.06, + "learning_rate": 4.946845389415299e-05, + "loss": 0.028, + "step": 276 + }, + { + "epoch": 0.06, + "learning_rate": 4.9466528002465145e-05, + "loss": 0.1113, + "step": 277 + }, + { + "epoch": 0.06, + "learning_rate": 4.946460211077729e-05, + "loss": 0.1859, + "step": 278 + }, + { + "epoch": 0.06, + "learning_rate": 4.946267621908944e-05, + "loss": 0.1255, + "step": 279 + }, + { + "epoch": 0.06, + "learning_rate": 4.946075032740159e-05, + "loss": 0.1759, + "step": 280 + }, + { + "epoch": 0.06, + "learning_rate": 4.9458824435713735e-05, + "loss": 0.0794, + "step": 281 + }, + { + "epoch": 0.07, + "learning_rate": 4.945689854402589e-05, + "loss": 0.0, + "step": 282 + }, + { + "epoch": 0.07, + "learning_rate": 4.945497265233803e-05, + "loss": 0.0452, + "step": 283 + }, + { + "epoch": 0.07, + "learning_rate": 4.9453046760650184e-05, + "loss": 0.0, + "step": 284 + }, + { + "epoch": 0.07, + "learning_rate": 4.945112086896233e-05, + "loss": 0.1033, + "step": 285 + }, + { + "epoch": 0.07, + "learning_rate": 4.944919497727448e-05, + "loss": 0.0886, + "step": 286 + }, + { + "epoch": 0.07, + "learning_rate": 4.944726908558663e-05, + "loss": 0.0039, + "step": 287 + }, + { + "epoch": 0.07, + "learning_rate": 4.944534319389878e-05, + "loss": 0.0111, + "step": 288 + }, + { + "epoch": 0.07, + "learning_rate": 4.944341730221093e-05, + "loss": 0.0929, + "step": 289 + }, + { + "epoch": 0.07, + "learning_rate": 4.944149141052307e-05, + "loss": 0.1342, + "step": 290 + }, + { + "epoch": 0.07, + "learning_rate": 4.943956551883522e-05, + "loss": 0.026, + "step": 291 + }, + { + "epoch": 0.07, + "learning_rate": 4.943763962714737e-05, + "loss": 0.019, + "step": 292 + }, + { + "epoch": 0.07, + "learning_rate": 4.943571373545952e-05, + "loss": 0.122, + "step": 293 + }, + { + "epoch": 0.07, + "learning_rate": 4.9433787843771666e-05, + "loss": 0.104, + "step": 294 + }, + { + "epoch": 0.07, + "learning_rate": 4.943186195208382e-05, + "loss": 0.1905, + "step": 295 + }, + { + "epoch": 0.07, + "learning_rate": 4.942993606039597e-05, + "loss": 0.0872, + "step": 296 + }, + { + "epoch": 0.07, + "learning_rate": 4.9428010168708115e-05, + "loss": 0.0, + "step": 297 + }, + { + "epoch": 0.07, + "learning_rate": 4.942608427702026e-05, + "loss": 0.03, + "step": 298 + }, + { + "epoch": 0.07, + "learning_rate": 4.942415838533241e-05, + "loss": 0.005, + "step": 299 + }, + { + "epoch": 0.07, + "learning_rate": 4.942223249364456e-05, + "loss": 0.0161, + "step": 300 + }, + { + "epoch": 0.07, + "learning_rate": 4.9420306601956705e-05, + "loss": 0.0569, + "step": 301 + }, + { + "epoch": 0.07, + "learning_rate": 4.941838071026886e-05, + "loss": 0.0195, + "step": 302 + }, + { + "epoch": 0.07, + "learning_rate": 4.941645481858101e-05, + "loss": 0.1901, + "step": 303 + }, + { + "epoch": 0.07, + "learning_rate": 4.9414528926893154e-05, + "loss": 0.0363, + "step": 304 + }, + { + "epoch": 0.07, + "learning_rate": 4.94126030352053e-05, + "loss": 0.067, + "step": 305 + }, + { + "epoch": 0.07, + "learning_rate": 4.9410677143517456e-05, + "loss": 0.0331, + "step": 306 + }, + { + "epoch": 0.07, + "learning_rate": 4.9408751251829597e-05, + "loss": 0.0156, + "step": 307 + }, + { + "epoch": 0.07, + "learning_rate": 4.9406825360141744e-05, + "loss": 0.1192, + "step": 308 + }, + { + "epoch": 0.07, + "learning_rate": 4.94048994684539e-05, + "loss": 0.001, + "step": 309 + }, + { + "epoch": 0.07, + "learning_rate": 4.9402973576766046e-05, + "loss": 0.0004, + "step": 310 + }, + { + "epoch": 0.07, + "learning_rate": 4.940104768507819e-05, + "loss": 0.1696, + "step": 311 + }, + { + "epoch": 0.07, + "learning_rate": 4.939912179339034e-05, + "loss": 0.0, + "step": 312 + }, + { + "epoch": 0.07, + "learning_rate": 4.9397195901702495e-05, + "loss": 0.0099, + "step": 313 + }, + { + "epoch": 0.07, + "learning_rate": 4.9395270010014636e-05, + "loss": 0.0, + "step": 314 + }, + { + "epoch": 0.07, + "learning_rate": 4.939334411832679e-05, + "loss": 0.0118, + "step": 315 + }, + { + "epoch": 0.07, + "learning_rate": 4.939141822663894e-05, + "loss": 0.0076, + "step": 316 + }, + { + "epoch": 0.07, + "learning_rate": 4.9389492334951085e-05, + "loss": 0.1233, + "step": 317 + }, + { + "epoch": 0.07, + "learning_rate": 4.938756644326323e-05, + "loss": 0.2803, + "step": 318 + }, + { + "epoch": 0.07, + "learning_rate": 4.938564055157538e-05, + "loss": 0.0467, + "step": 319 + }, + { + "epoch": 0.07, + "learning_rate": 4.9383714659887534e-05, + "loss": 0.0432, + "step": 320 + }, + { + "epoch": 0.07, + "learning_rate": 4.9381788768199675e-05, + "loss": 0.0824, + "step": 321 + }, + { + "epoch": 0.07, + "learning_rate": 4.937986287651183e-05, + "loss": 0.0371, + "step": 322 + }, + { + "epoch": 0.07, + "learning_rate": 4.9377936984823977e-05, + "loss": 0.0, + "step": 323 + }, + { + "epoch": 0.07, + "learning_rate": 4.9376011093136124e-05, + "loss": 0.1086, + "step": 324 + }, + { + "epoch": 0.08, + "learning_rate": 4.937408520144827e-05, + "loss": 0.0001, + "step": 325 + }, + { + "epoch": 0.08, + "learning_rate": 4.9372159309760426e-05, + "loss": 0.0001, + "step": 326 + }, + { + "epoch": 0.08, + "learning_rate": 4.937023341807257e-05, + "loss": 0.1021, + "step": 327 + }, + { + "epoch": 0.08, + "learning_rate": 4.9368307526384714e-05, + "loss": 0.103, + "step": 328 + }, + { + "epoch": 0.08, + "learning_rate": 4.936638163469687e-05, + "loss": 0.0001, + "step": 329 + }, + { + "epoch": 0.08, + "learning_rate": 4.9364455743009016e-05, + "loss": 0.0492, + "step": 330 + }, + { + "epoch": 0.08, + "learning_rate": 4.936252985132116e-05, + "loss": 0.0633, + "step": 331 + }, + { + "epoch": 0.08, + "learning_rate": 4.936060395963331e-05, + "loss": 0.0012, + "step": 332 + }, + { + "epoch": 0.08, + "learning_rate": 4.9358678067945465e-05, + "loss": 0.0063, + "step": 333 + }, + { + "epoch": 0.08, + "learning_rate": 4.935675217625761e-05, + "loss": 0.2497, + "step": 334 + }, + { + "epoch": 0.08, + "learning_rate": 4.935482628456976e-05, + "loss": 0.039, + "step": 335 + }, + { + "epoch": 0.08, + "learning_rate": 4.935290039288191e-05, + "loss": 0.0418, + "step": 336 + }, + { + "epoch": 0.08, + "learning_rate": 4.9350974501194055e-05, + "loss": 0.0374, + "step": 337 + }, + { + "epoch": 0.08, + "learning_rate": 4.93490486095062e-05, + "loss": 0.0004, + "step": 338 + }, + { + "epoch": 0.08, + "learning_rate": 4.934712271781835e-05, + "loss": 0.0466, + "step": 339 + }, + { + "epoch": 0.08, + "learning_rate": 4.9345196826130504e-05, + "loss": 0.0107, + "step": 340 + }, + { + "epoch": 0.08, + "learning_rate": 4.9343270934442645e-05, + "loss": 0.0018, + "step": 341 + }, + { + "epoch": 0.08, + "learning_rate": 4.93413450427548e-05, + "loss": 0.1153, + "step": 342 + }, + { + "epoch": 0.08, + "learning_rate": 4.9339419151066946e-05, + "loss": 0.0, + "step": 343 + }, + { + "epoch": 0.08, + "learning_rate": 4.9337493259379094e-05, + "loss": 0.195, + "step": 344 + }, + { + "epoch": 0.08, + "learning_rate": 4.933556736769124e-05, + "loss": 0.0366, + "step": 345 + }, + { + "epoch": 0.08, + "learning_rate": 4.933364147600339e-05, + "loss": 0.003, + "step": 346 + }, + { + "epoch": 0.08, + "learning_rate": 4.933171558431554e-05, + "loss": 0.0044, + "step": 347 + }, + { + "epoch": 0.08, + "learning_rate": 4.9329789692627684e-05, + "loss": 0.1508, + "step": 348 + }, + { + "epoch": 0.08, + "learning_rate": 4.932786380093984e-05, + "loss": 0.0233, + "step": 349 + }, + { + "epoch": 0.08, + "learning_rate": 4.9325937909251986e-05, + "loss": 0.1082, + "step": 350 + }, + { + "epoch": 0.08, + "learning_rate": 4.932401201756413e-05, + "loss": 0.0, + "step": 351 + }, + { + "epoch": 0.08, + "learning_rate": 4.932208612587628e-05, + "loss": 0.0647, + "step": 352 + }, + { + "epoch": 0.08, + "learning_rate": 4.9320160234188435e-05, + "loss": 0.0243, + "step": 353 + }, + { + "epoch": 0.08, + "learning_rate": 4.931823434250058e-05, + "loss": 0.2509, + "step": 354 + }, + { + "epoch": 0.08, + "learning_rate": 4.931630845081272e-05, + "loss": 0.0887, + "step": 355 + }, + { + "epoch": 0.08, + "learning_rate": 4.931438255912488e-05, + "loss": 0.0052, + "step": 356 + }, + { + "epoch": 0.08, + "learning_rate": 4.9312456667437025e-05, + "loss": 0.2179, + "step": 357 + }, + { + "epoch": 0.08, + "learning_rate": 4.931053077574917e-05, + "loss": 0.0283, + "step": 358 + }, + { + "epoch": 0.08, + "learning_rate": 4.930860488406132e-05, + "loss": 0.2803, + "step": 359 + }, + { + "epoch": 0.08, + "learning_rate": 4.9306678992373474e-05, + "loss": 0.0, + "step": 360 + }, + { + "epoch": 0.08, + "learning_rate": 4.930475310068562e-05, + "loss": 0.0421, + "step": 361 + }, + { + "epoch": 0.08, + "learning_rate": 4.930282720899777e-05, + "loss": 0.1081, + "step": 362 + }, + { + "epoch": 0.08, + "learning_rate": 4.9300901317309916e-05, + "loss": 0.2016, + "step": 363 + }, + { + "epoch": 0.08, + "learning_rate": 4.9298975425622064e-05, + "loss": 0.0326, + "step": 364 + }, + { + "epoch": 0.08, + "learning_rate": 4.929704953393421e-05, + "loss": 0.0, + "step": 365 + }, + { + "epoch": 0.08, + "learning_rate": 4.929512364224636e-05, + "loss": 0.0039, + "step": 366 + }, + { + "epoch": 0.08, + "learning_rate": 4.929319775055851e-05, + "loss": 0.0, + "step": 367 + }, + { + "epoch": 0.09, + "learning_rate": 4.929127185887066e-05, + "loss": 0.0499, + "step": 368 + }, + { + "epoch": 0.09, + "learning_rate": 4.928934596718281e-05, + "loss": 0.0608, + "step": 369 + }, + { + "epoch": 0.09, + "learning_rate": 4.9287420075494955e-05, + "loss": 0.1448, + "step": 370 + }, + { + "epoch": 0.09, + "learning_rate": 4.928549418380711e-05, + "loss": 0.0558, + "step": 371 + }, + { + "epoch": 0.09, + "learning_rate": 4.928356829211925e-05, + "loss": 0.0, + "step": 372 + }, + { + "epoch": 0.09, + "learning_rate": 4.9281642400431405e-05, + "loss": 0.0302, + "step": 373 + }, + { + "epoch": 0.09, + "learning_rate": 4.927971650874355e-05, + "loss": 0.0359, + "step": 374 + }, + { + "epoch": 0.09, + "learning_rate": 4.92777906170557e-05, + "loss": 0.2575, + "step": 375 + }, + { + "epoch": 0.09, + "learning_rate": 4.927586472536785e-05, + "loss": 0.0273, + "step": 376 + }, + { + "epoch": 0.09, + "learning_rate": 4.9273938833679995e-05, + "loss": 0.0247, + "step": 377 + }, + { + "epoch": 0.09, + "learning_rate": 4.927201294199215e-05, + "loss": 0.0121, + "step": 378 + }, + { + "epoch": 0.09, + "learning_rate": 4.927008705030429e-05, + "loss": 0.056, + "step": 379 + }, + { + "epoch": 0.09, + "learning_rate": 4.9268161158616444e-05, + "loss": 0.036, + "step": 380 + }, + { + "epoch": 0.09, + "learning_rate": 4.926623526692859e-05, + "loss": 0.084, + "step": 381 + }, + { + "epoch": 0.09, + "learning_rate": 4.926430937524074e-05, + "loss": 0.0928, + "step": 382 + }, + { + "epoch": 0.09, + "learning_rate": 4.9262383483552886e-05, + "loss": 0.0799, + "step": 383 + }, + { + "epoch": 0.09, + "learning_rate": 4.9260457591865034e-05, + "loss": 0.0394, + "step": 384 + }, + { + "epoch": 0.09, + "learning_rate": 4.925853170017719e-05, + "loss": 0.1354, + "step": 385 + }, + { + "epoch": 0.09, + "learning_rate": 4.925660580848933e-05, + "loss": 0.0177, + "step": 386 + }, + { + "epoch": 0.09, + "learning_rate": 4.925467991680148e-05, + "loss": 0.3162, + "step": 387 + }, + { + "epoch": 0.09, + "learning_rate": 4.925275402511363e-05, + "loss": 0.1136, + "step": 388 + }, + { + "epoch": 0.09, + "learning_rate": 4.925082813342578e-05, + "loss": 0.0009, + "step": 389 + }, + { + "epoch": 0.09, + "learning_rate": 4.9248902241737925e-05, + "loss": 0.225, + "step": 390 + }, + { + "epoch": 0.09, + "learning_rate": 4.924697635005008e-05, + "loss": 0.0, + "step": 391 + }, + { + "epoch": 0.09, + "learning_rate": 4.924505045836223e-05, + "loss": 0.1476, + "step": 392 + }, + { + "epoch": 0.09, + "learning_rate": 4.924312456667437e-05, + "loss": 0.0207, + "step": 393 + }, + { + "epoch": 0.09, + "learning_rate": 4.924119867498652e-05, + "loss": 0.0315, + "step": 394 + }, + { + "epoch": 0.09, + "learning_rate": 4.923927278329867e-05, + "loss": 0.0887, + "step": 395 + }, + { + "epoch": 0.09, + "learning_rate": 4.923734689161082e-05, + "loss": 0.0003, + "step": 396 + }, + { + "epoch": 0.09, + "learning_rate": 4.9235420999922964e-05, + "loss": 0.0005, + "step": 397 + }, + { + "epoch": 0.09, + "learning_rate": 4.923349510823512e-05, + "loss": 0.0054, + "step": 398 + }, + { + "epoch": 0.09, + "learning_rate": 4.9231569216547266e-05, + "loss": 0.0525, + "step": 399 + }, + { + "epoch": 0.09, + "learning_rate": 4.9229643324859414e-05, + "loss": 0.0, + "step": 400 + }, + { + "epoch": 0.09, + "learning_rate": 4.922771743317156e-05, + "loss": 0.0835, + "step": 401 + }, + { + "epoch": 0.09, + "learning_rate": 4.922579154148371e-05, + "loss": 0.2823, + "step": 402 + }, + { + "epoch": 0.09, + "learning_rate": 4.9223865649795856e-05, + "loss": 0.0619, + "step": 403 + }, + { + "epoch": 0.09, + "learning_rate": 4.9221939758108004e-05, + "loss": 0.0001, + "step": 404 + }, + { + "epoch": 0.09, + "learning_rate": 4.922001386642016e-05, + "loss": 0.0821, + "step": 405 + }, + { + "epoch": 0.09, + "learning_rate": 4.9218087974732305e-05, + "loss": 0.0252, + "step": 406 + }, + { + "epoch": 0.09, + "learning_rate": 4.921616208304445e-05, + "loss": 0.0181, + "step": 407 + }, + { + "epoch": 0.09, + "learning_rate": 4.92142361913566e-05, + "loss": 0.072, + "step": 408 + }, + { + "epoch": 0.09, + "learning_rate": 4.9212310299668754e-05, + "loss": 0.1295, + "step": 409 + }, + { + "epoch": 0.09, + "learning_rate": 4.9210384407980895e-05, + "loss": 0.0331, + "step": 410 + }, + { + "epoch": 0.09, + "learning_rate": 4.920845851629304e-05, + "loss": 0.0149, + "step": 411 + }, + { + "epoch": 0.1, + "learning_rate": 4.92065326246052e-05, + "loss": 0.0054, + "step": 412 + }, + { + "epoch": 0.1, + "learning_rate": 4.9204606732917344e-05, + "loss": 0.0967, + "step": 413 + }, + { + "epoch": 0.1, + "learning_rate": 4.920268084122949e-05, + "loss": 0.0986, + "step": 414 + }, + { + "epoch": 0.1, + "learning_rate": 4.920075494954164e-05, + "loss": 0.0001, + "step": 415 + }, + { + "epoch": 0.1, + "learning_rate": 4.9198829057853794e-05, + "loss": 0.0497, + "step": 416 + }, + { + "epoch": 0.1, + "learning_rate": 4.9196903166165934e-05, + "loss": 0.1261, + "step": 417 + }, + { + "epoch": 0.1, + "learning_rate": 4.919497727447809e-05, + "loss": 0.0062, + "step": 418 + }, + { + "epoch": 0.1, + "learning_rate": 4.9193051382790236e-05, + "loss": 0.0498, + "step": 419 + }, + { + "epoch": 0.1, + "learning_rate": 4.919112549110238e-05, + "loss": 0.0309, + "step": 420 + }, + { + "epoch": 0.1, + "learning_rate": 4.918919959941453e-05, + "loss": 0.0343, + "step": 421 + }, + { + "epoch": 0.1, + "learning_rate": 4.918727370772668e-05, + "loss": 0.0, + "step": 422 + }, + { + "epoch": 0.1, + "learning_rate": 4.9185347816038826e-05, + "loss": 0.0, + "step": 423 + }, + { + "epoch": 0.1, + "learning_rate": 4.9183421924350973e-05, + "loss": 0.0, + "step": 424 + }, + { + "epoch": 0.1, + "learning_rate": 4.918149603266313e-05, + "loss": 0.0, + "step": 425 + }, + { + "epoch": 0.1, + "learning_rate": 4.9179570140975275e-05, + "loss": 0.0294, + "step": 426 + }, + { + "epoch": 0.1, + "learning_rate": 4.917764424928742e-05, + "loss": 0.001, + "step": 427 + }, + { + "epoch": 0.1, + "learning_rate": 4.917571835759957e-05, + "loss": 0.0833, + "step": 428 + }, + { + "epoch": 0.1, + "learning_rate": 4.9173792465911724e-05, + "loss": 0.0257, + "step": 429 + }, + { + "epoch": 0.1, + "learning_rate": 4.9171866574223865e-05, + "loss": 0.0881, + "step": 430 + }, + { + "epoch": 0.1, + "learning_rate": 4.916994068253601e-05, + "loss": 0.0438, + "step": 431 + }, + { + "epoch": 0.1, + "learning_rate": 4.916801479084817e-05, + "loss": 0.0, + "step": 432 + }, + { + "epoch": 0.1, + "learning_rate": 4.9166088899160314e-05, + "loss": 0.0153, + "step": 433 + }, + { + "epoch": 0.1, + "learning_rate": 4.916416300747246e-05, + "loss": 0.0017, + "step": 434 + }, + { + "epoch": 0.1, + "learning_rate": 4.916223711578461e-05, + "loss": 0.2491, + "step": 435 + }, + { + "epoch": 0.1, + "learning_rate": 4.9160311224096763e-05, + "loss": 0.0096, + "step": 436 + }, + { + "epoch": 0.1, + "learning_rate": 4.9158385332408904e-05, + "loss": 0.0569, + "step": 437 + }, + { + "epoch": 0.1, + "learning_rate": 4.915645944072106e-05, + "loss": 0.0003, + "step": 438 + }, + { + "epoch": 0.1, + "learning_rate": 4.9154533549033206e-05, + "loss": 0.0272, + "step": 439 + }, + { + "epoch": 0.1, + "learning_rate": 4.9152607657345353e-05, + "loss": 0.0215, + "step": 440 + }, + { + "epoch": 0.1, + "learning_rate": 4.91506817656575e-05, + "loss": 0.0535, + "step": 441 + }, + { + "epoch": 0.1, + "learning_rate": 4.914875587396965e-05, + "loss": 0.0488, + "step": 442 + }, + { + "epoch": 0.1, + "learning_rate": 4.91468299822818e-05, + "loss": 0.0, + "step": 443 + }, + { + "epoch": 0.1, + "learning_rate": 4.914490409059394e-05, + "loss": 0.05, + "step": 444 + }, + { + "epoch": 0.1, + "learning_rate": 4.91429781989061e-05, + "loss": 0.1697, + "step": 445 + }, + { + "epoch": 0.1, + "learning_rate": 4.9141052307218245e-05, + "loss": 0.0596, + "step": 446 + }, + { + "epoch": 0.1, + "learning_rate": 4.913912641553039e-05, + "loss": 0.0418, + "step": 447 + }, + { + "epoch": 0.1, + "learning_rate": 4.913720052384254e-05, + "loss": 0.017, + "step": 448 + }, + { + "epoch": 0.1, + "learning_rate": 4.913527463215469e-05, + "loss": 0.0022, + "step": 449 + }, + { + "epoch": 0.1, + "learning_rate": 4.913334874046684e-05, + "loss": 0.0029, + "step": 450 + }, + { + "epoch": 0.1, + "learning_rate": 4.913142284877898e-05, + "loss": 0.0, + "step": 451 + }, + { + "epoch": 0.1, + "learning_rate": 4.912949695709114e-05, + "loss": 0.0586, + "step": 452 + }, + { + "epoch": 0.1, + "learning_rate": 4.9127571065403284e-05, + "loss": 0.0003, + "step": 453 + }, + { + "epoch": 0.1, + "learning_rate": 4.912564517371543e-05, + "loss": 0.0001, + "step": 454 + }, + { + "epoch": 0.11, + "learning_rate": 4.912371928202758e-05, + "loss": 0.1261, + "step": 455 + }, + { + "epoch": 0.11, + "learning_rate": 4.912179339033973e-05, + "loss": 0.0, + "step": 456 + }, + { + "epoch": 0.11, + "learning_rate": 4.911986749865188e-05, + "loss": 0.0006, + "step": 457 + }, + { + "epoch": 0.11, + "learning_rate": 4.911794160696402e-05, + "loss": 0.0683, + "step": 458 + }, + { + "epoch": 0.11, + "learning_rate": 4.9116015715276176e-05, + "loss": 0.0, + "step": 459 + }, + { + "epoch": 0.11, + "learning_rate": 4.911408982358832e-05, + "loss": 0.0, + "step": 460 + }, + { + "epoch": 0.11, + "learning_rate": 4.911216393190047e-05, + "loss": 0.0504, + "step": 461 + }, + { + "epoch": 0.11, + "learning_rate": 4.911023804021262e-05, + "loss": 0.0, + "step": 462 + }, + { + "epoch": 0.11, + "learning_rate": 4.910831214852477e-05, + "loss": 0.0712, + "step": 463 + }, + { + "epoch": 0.11, + "learning_rate": 4.910638625683692e-05, + "loss": 0.1442, + "step": 464 + }, + { + "epoch": 0.11, + "learning_rate": 4.910446036514907e-05, + "loss": 0.1328, + "step": 465 + }, + { + "epoch": 0.11, + "learning_rate": 4.9102534473461215e-05, + "loss": 0.0123, + "step": 466 + }, + { + "epoch": 0.11, + "learning_rate": 4.910060858177336e-05, + "loss": 0.0567, + "step": 467 + }, + { + "epoch": 0.11, + "learning_rate": 4.909868269008551e-05, + "loss": 0.0016, + "step": 468 + }, + { + "epoch": 0.11, + "learning_rate": 4.909675679839766e-05, + "loss": 0.0796, + "step": 469 + }, + { + "epoch": 0.11, + "learning_rate": 4.909483090670981e-05, + "loss": 0.0094, + "step": 470 + }, + { + "epoch": 0.11, + "learning_rate": 4.909290501502196e-05, + "loss": 0.0544, + "step": 471 + }, + { + "epoch": 0.11, + "learning_rate": 4.9090979123334107e-05, + "loss": 0.0888, + "step": 472 + }, + { + "epoch": 0.11, + "learning_rate": 4.9089053231646254e-05, + "loss": 0.0084, + "step": 473 + }, + { + "epoch": 0.11, + "learning_rate": 4.908712733995841e-05, + "loss": 0.125, + "step": 474 + }, + { + "epoch": 0.11, + "learning_rate": 4.908520144827055e-05, + "loss": 0.0015, + "step": 475 + }, + { + "epoch": 0.11, + "learning_rate": 4.9083275556582696e-05, + "loss": 0.0223, + "step": 476 + }, + { + "epoch": 0.11, + "learning_rate": 4.908134966489485e-05, + "loss": 0.0021, + "step": 477 + }, + { + "epoch": 0.11, + "learning_rate": 4.9079423773207e-05, + "loss": 0.0486, + "step": 478 + }, + { + "epoch": 0.11, + "learning_rate": 4.9077497881519146e-05, + "loss": 0.0188, + "step": 479 + }, + { + "epoch": 0.11, + "learning_rate": 4.907557198983129e-05, + "loss": 0.0794, + "step": 480 + }, + { + "epoch": 0.11, + "learning_rate": 4.907364609814345e-05, + "loss": 0.0, + "step": 481 + }, + { + "epoch": 0.11, + "learning_rate": 4.907172020645559e-05, + "loss": 0.065, + "step": 482 + }, + { + "epoch": 0.11, + "learning_rate": 4.906979431476774e-05, + "loss": 0.0147, + "step": 483 + }, + { + "epoch": 0.11, + "learning_rate": 4.906786842307989e-05, + "loss": 0.0058, + "step": 484 + }, + { + "epoch": 0.11, + "learning_rate": 4.906594253139204e-05, + "loss": 0.1018, + "step": 485 + }, + { + "epoch": 0.11, + "learning_rate": 4.9064016639704185e-05, + "loss": 0.0605, + "step": 486 + }, + { + "epoch": 0.11, + "learning_rate": 4.906209074801633e-05, + "loss": 0.0014, + "step": 487 + }, + { + "epoch": 0.11, + "learning_rate": 4.9060164856328487e-05, + "loss": 0.0993, + "step": 488 + }, + { + "epoch": 0.11, + "learning_rate": 4.905823896464063e-05, + "loss": 0.1544, + "step": 489 + }, + { + "epoch": 0.11, + "learning_rate": 4.905631307295278e-05, + "loss": 0.0017, + "step": 490 + }, + { + "epoch": 0.11, + "learning_rate": 4.905438718126493e-05, + "loss": 0.0632, + "step": 491 + }, + { + "epoch": 0.11, + "learning_rate": 4.9052461289577076e-05, + "loss": 0.0759, + "step": 492 + }, + { + "epoch": 0.11, + "learning_rate": 4.9050535397889224e-05, + "loss": 0.0494, + "step": 493 + }, + { + "epoch": 0.11, + "learning_rate": 4.904860950620138e-05, + "loss": 0.0204, + "step": 494 + }, + { + "epoch": 0.11, + "learning_rate": 4.904668361451352e-05, + "loss": 0.0, + "step": 495 + }, + { + "epoch": 0.11, + "learning_rate": 4.9044757722825666e-05, + "loss": 0.0562, + "step": 496 + }, + { + "epoch": 0.11, + "learning_rate": 4.904283183113782e-05, + "loss": 0.0088, + "step": 497 + }, + { + "epoch": 0.12, + "learning_rate": 4.904090593944997e-05, + "loss": 0.0115, + "step": 498 + }, + { + "epoch": 0.12, + "learning_rate": 4.9038980047762116e-05, + "loss": 0.0, + "step": 499 + }, + { + "epoch": 0.12, + "learning_rate": 4.903705415607426e-05, + "loss": 0.0425, + "step": 500 + }, + { + "epoch": 0.12, + "learning_rate": 4.903512826438642e-05, + "loss": 0.117, + "step": 501 + }, + { + "epoch": 0.12, + "learning_rate": 4.903320237269856e-05, + "loss": 0.0, + "step": 502 + }, + { + "epoch": 0.12, + "learning_rate": 4.903127648101071e-05, + "loss": 0.0238, + "step": 503 + }, + { + "epoch": 0.12, + "learning_rate": 4.902935058932286e-05, + "loss": 0.0479, + "step": 504 + }, + { + "epoch": 0.12, + "learning_rate": 4.902742469763501e-05, + "loss": 0.0, + "step": 505 + }, + { + "epoch": 0.12, + "learning_rate": 4.9025498805947155e-05, + "loss": 0.0574, + "step": 506 + }, + { + "epoch": 0.12, + "learning_rate": 4.90235729142593e-05, + "loss": 0.1001, + "step": 507 + }, + { + "epoch": 0.12, + "learning_rate": 4.9021647022571456e-05, + "loss": 0.1383, + "step": 508 + }, + { + "epoch": 0.12, + "learning_rate": 4.90197211308836e-05, + "loss": 0.0001, + "step": 509 + }, + { + "epoch": 0.12, + "learning_rate": 4.901779523919575e-05, + "loss": 0.0496, + "step": 510 + }, + { + "epoch": 0.12, + "learning_rate": 4.90158693475079e-05, + "loss": 0.047, + "step": 511 + }, + { + "epoch": 0.12, + "learning_rate": 4.9013943455820046e-05, + "loss": 0.0256, + "step": 512 + }, + { + "epoch": 0.12, + "learning_rate": 4.9012017564132194e-05, + "loss": 0.016, + "step": 513 + }, + { + "epoch": 0.12, + "learning_rate": 4.901009167244434e-05, + "loss": 0.2213, + "step": 514 + }, + { + "epoch": 0.12, + "learning_rate": 4.9008165780756496e-05, + "loss": 0.0367, + "step": 515 + }, + { + "epoch": 0.12, + "learning_rate": 4.9006239889068636e-05, + "loss": 0.2303, + "step": 516 + }, + { + "epoch": 0.12, + "learning_rate": 4.900431399738079e-05, + "loss": 0.204, + "step": 517 + }, + { + "epoch": 0.12, + "learning_rate": 4.900238810569294e-05, + "loss": 0.0603, + "step": 518 + }, + { + "epoch": 0.12, + "learning_rate": 4.9000462214005085e-05, + "loss": 0.0274, + "step": 519 + }, + { + "epoch": 0.12, + "learning_rate": 4.899853632231723e-05, + "loss": 0.0176, + "step": 520 + }, + { + "epoch": 0.12, + "learning_rate": 4.899661043062939e-05, + "loss": 0.0411, + "step": 521 + }, + { + "epoch": 0.12, + "learning_rate": 4.8994684538941535e-05, + "loss": 0.2075, + "step": 522 + }, + { + "epoch": 0.12, + "learning_rate": 4.8992758647253675e-05, + "loss": 0.0038, + "step": 523 + }, + { + "epoch": 0.12, + "learning_rate": 4.899083275556583e-05, + "loss": 0.0005, + "step": 524 + }, + { + "epoch": 0.12, + "learning_rate": 4.898890686387798e-05, + "loss": 0.1341, + "step": 525 + }, + { + "epoch": 0.12, + "learning_rate": 4.8986980972190125e-05, + "loss": 0.0, + "step": 526 + }, + { + "epoch": 0.12, + "learning_rate": 4.898505508050227e-05, + "loss": 0.0108, + "step": 527 + }, + { + "epoch": 0.12, + "learning_rate": 4.8983129188814426e-05, + "loss": 0.0391, + "step": 528 + }, + { + "epoch": 0.12, + "learning_rate": 4.8981203297126574e-05, + "loss": 0.0739, + "step": 529 + }, + { + "epoch": 0.12, + "learning_rate": 4.897927740543872e-05, + "loss": 0.0, + "step": 530 + }, + { + "epoch": 0.12, + "learning_rate": 4.897735151375087e-05, + "loss": 0.0251, + "step": 531 + }, + { + "epoch": 0.12, + "learning_rate": 4.8975425622063016e-05, + "loss": 0.0038, + "step": 532 + }, + { + "epoch": 0.12, + "learning_rate": 4.8973499730375164e-05, + "loss": 0.1692, + "step": 533 + }, + { + "epoch": 0.12, + "learning_rate": 4.897157383868731e-05, + "loss": 0.0311, + "step": 534 + }, + { + "epoch": 0.12, + "learning_rate": 4.8969647946999465e-05, + "loss": 0.0416, + "step": 535 + }, + { + "epoch": 0.12, + "learning_rate": 4.896772205531161e-05, + "loss": 0.014, + "step": 536 + }, + { + "epoch": 0.12, + "learning_rate": 4.896579616362376e-05, + "loss": 0.0, + "step": 537 + }, + { + "epoch": 0.12, + "learning_rate": 4.896387027193591e-05, + "loss": 0.0074, + "step": 538 + }, + { + "epoch": 0.12, + "learning_rate": 4.896194438024806e-05, + "loss": 0.0079, + "step": 539 + }, + { + "epoch": 0.12, + "learning_rate": 4.89600184885602e-05, + "loss": 0.001, + "step": 540 + }, + { + "epoch": 0.13, + "learning_rate": 4.895809259687236e-05, + "loss": 0.0, + "step": 541 + }, + { + "epoch": 0.13, + "learning_rate": 4.8956166705184505e-05, + "loss": 0.0337, + "step": 542 + }, + { + "epoch": 0.13, + "learning_rate": 4.895424081349665e-05, + "loss": 0.0191, + "step": 543 + }, + { + "epoch": 0.13, + "learning_rate": 4.89523149218088e-05, + "loss": 0.0065, + "step": 544 + }, + { + "epoch": 0.13, + "learning_rate": 4.895038903012095e-05, + "loss": 0.0053, + "step": 545 + }, + { + "epoch": 0.13, + "learning_rate": 4.89484631384331e-05, + "loss": 0.0152, + "step": 546 + }, + { + "epoch": 0.13, + "learning_rate": 4.894653724674524e-05, + "loss": 0.0459, + "step": 547 + }, + { + "epoch": 0.13, + "learning_rate": 4.8944611355057396e-05, + "loss": 0.0169, + "step": 548 + }, + { + "epoch": 0.13, + "learning_rate": 4.8942685463369544e-05, + "loss": 0.1101, + "step": 549 + }, + { + "epoch": 0.13, + "learning_rate": 4.894075957168169e-05, + "loss": 0.0828, + "step": 550 + }, + { + "epoch": 0.13, + "learning_rate": 4.893883367999384e-05, + "loss": 0.0015, + "step": 551 + }, + { + "epoch": 0.13, + "learning_rate": 4.8936907788305986e-05, + "loss": 0.0088, + "step": 552 + }, + { + "epoch": 0.13, + "learning_rate": 4.893498189661814e-05, + "loss": 0.0308, + "step": 553 + }, + { + "epoch": 0.13, + "learning_rate": 4.893305600493028e-05, + "loss": 0.0872, + "step": 554 + }, + { + "epoch": 0.13, + "learning_rate": 4.8931130113242435e-05, + "loss": 0.011, + "step": 555 + }, + { + "epoch": 0.13, + "learning_rate": 4.892920422155458e-05, + "loss": 0.0369, + "step": 556 + }, + { + "epoch": 0.13, + "learning_rate": 4.892727832986673e-05, + "loss": 0.0495, + "step": 557 + }, + { + "epoch": 0.13, + "learning_rate": 4.892535243817888e-05, + "loss": 0.0, + "step": 558 + }, + { + "epoch": 0.13, + "learning_rate": 4.892342654649103e-05, + "loss": 0.0486, + "step": 559 + }, + { + "epoch": 0.13, + "learning_rate": 4.892150065480318e-05, + "loss": 0.0798, + "step": 560 + }, + { + "epoch": 0.13, + "learning_rate": 4.891957476311532e-05, + "loss": 0.0288, + "step": 561 + }, + { + "epoch": 0.13, + "learning_rate": 4.8917648871427474e-05, + "loss": 0.0473, + "step": 562 + }, + { + "epoch": 0.13, + "learning_rate": 4.891572297973962e-05, + "loss": 0.0007, + "step": 563 + }, + { + "epoch": 0.13, + "learning_rate": 4.891379708805177e-05, + "loss": 0.0751, + "step": 564 + }, + { + "epoch": 0.13, + "learning_rate": 4.891187119636392e-05, + "loss": 0.0225, + "step": 565 + }, + { + "epoch": 0.13, + "learning_rate": 4.890994530467607e-05, + "loss": 0.1462, + "step": 566 + }, + { + "epoch": 0.13, + "learning_rate": 4.890801941298821e-05, + "loss": 0.0955, + "step": 567 + }, + { + "epoch": 0.13, + "learning_rate": 4.8906093521300366e-05, + "loss": 0.0, + "step": 568 + }, + { + "epoch": 0.13, + "learning_rate": 4.8904167629612514e-05, + "loss": 0.0647, + "step": 569 + }, + { + "epoch": 0.13, + "learning_rate": 4.890224173792466e-05, + "loss": 0.0411, + "step": 570 + }, + { + "epoch": 0.13, + "learning_rate": 4.890031584623681e-05, + "loss": 0.0154, + "step": 571 + }, + { + "epoch": 0.13, + "learning_rate": 4.8898389954548956e-05, + "loss": 0.0121, + "step": 572 + }, + { + "epoch": 0.13, + "learning_rate": 4.889646406286111e-05, + "loss": 0.0993, + "step": 573 + }, + { + "epoch": 0.13, + "learning_rate": 4.889453817117325e-05, + "loss": 0.0597, + "step": 574 + }, + { + "epoch": 0.13, + "learning_rate": 4.8892612279485405e-05, + "loss": 0.0654, + "step": 575 + }, + { + "epoch": 0.13, + "learning_rate": 4.889068638779755e-05, + "loss": 0.0, + "step": 576 + }, + { + "epoch": 0.13, + "learning_rate": 4.88887604961097e-05, + "loss": 0.0, + "step": 577 + }, + { + "epoch": 0.13, + "learning_rate": 4.888683460442185e-05, + "loss": 0.0337, + "step": 578 + }, + { + "epoch": 0.13, + "learning_rate": 4.8884908712733995e-05, + "loss": 0.0027, + "step": 579 + }, + { + "epoch": 0.13, + "learning_rate": 4.888298282104615e-05, + "loss": 0.0011, + "step": 580 + }, + { + "epoch": 0.13, + "learning_rate": 4.888105692935829e-05, + "loss": 0.0383, + "step": 581 + }, + { + "epoch": 0.13, + "learning_rate": 4.8879131037670444e-05, + "loss": 0.2108, + "step": 582 + }, + { + "epoch": 0.13, + "learning_rate": 4.887720514598259e-05, + "loss": 0.0061, + "step": 583 + }, + { + "epoch": 0.13, + "learning_rate": 4.887527925429474e-05, + "loss": 0.016, + "step": 584 + }, + { + "epoch": 0.14, + "learning_rate": 4.887335336260689e-05, + "loss": 0.0, + "step": 585 + }, + { + "epoch": 0.14, + "learning_rate": 4.887142747091904e-05, + "loss": 0.0332, + "step": 586 + }, + { + "epoch": 0.14, + "learning_rate": 4.886950157923119e-05, + "loss": 0.0194, + "step": 587 + }, + { + "epoch": 0.14, + "learning_rate": 4.886757568754333e-05, + "loss": 0.0004, + "step": 588 + }, + { + "epoch": 0.14, + "learning_rate": 4.8865649795855483e-05, + "loss": 0.0205, + "step": 589 + }, + { + "epoch": 0.14, + "learning_rate": 4.886372390416763e-05, + "loss": 0.0, + "step": 590 + }, + { + "epoch": 0.14, + "learning_rate": 4.886179801247978e-05, + "loss": 0.0274, + "step": 591 + }, + { + "epoch": 0.14, + "learning_rate": 4.8859872120791926e-05, + "loss": 0.017, + "step": 592 + }, + { + "epoch": 0.14, + "learning_rate": 4.885794622910408e-05, + "loss": 0.0151, + "step": 593 + }, + { + "epoch": 0.14, + "learning_rate": 4.885602033741623e-05, + "loss": 0.0, + "step": 594 + }, + { + "epoch": 0.14, + "learning_rate": 4.8854094445728375e-05, + "loss": 0.0523, + "step": 595 + }, + { + "epoch": 0.14, + "learning_rate": 4.885216855404052e-05, + "loss": 0.0887, + "step": 596 + }, + { + "epoch": 0.14, + "learning_rate": 4.885024266235268e-05, + "loss": 0.0001, + "step": 597 + }, + { + "epoch": 0.14, + "learning_rate": 4.884831677066482e-05, + "loss": 0.0403, + "step": 598 + }, + { + "epoch": 0.14, + "learning_rate": 4.8846390878976965e-05, + "loss": 0.0273, + "step": 599 + }, + { + "epoch": 0.14, + "learning_rate": 4.884446498728912e-05, + "loss": 0.0158, + "step": 600 + }, + { + "epoch": 0.14, + "learning_rate": 4.884253909560127e-05, + "loss": 0.0208, + "step": 601 + }, + { + "epoch": 0.14, + "learning_rate": 4.8840613203913414e-05, + "loss": 0.0051, + "step": 602 + }, + { + "epoch": 0.14, + "learning_rate": 4.883868731222556e-05, + "loss": 0.0659, + "step": 603 + }, + { + "epoch": 0.14, + "learning_rate": 4.8836761420537716e-05, + "loss": 0.031, + "step": 604 + }, + { + "epoch": 0.14, + "learning_rate": 4.883483552884986e-05, + "loss": 0.0006, + "step": 605 + }, + { + "epoch": 0.14, + "learning_rate": 4.883290963716201e-05, + "loss": 0.0123, + "step": 606 + }, + { + "epoch": 0.14, + "learning_rate": 4.883098374547416e-05, + "loss": 0.0001, + "step": 607 + }, + { + "epoch": 0.14, + "learning_rate": 4.8829057853786306e-05, + "loss": 0.0486, + "step": 608 + }, + { + "epoch": 0.14, + "learning_rate": 4.882713196209845e-05, + "loss": 0.2122, + "step": 609 + }, + { + "epoch": 0.14, + "learning_rate": 4.88252060704106e-05, + "loss": 0.0, + "step": 610 + }, + { + "epoch": 0.14, + "learning_rate": 4.8823280178722755e-05, + "loss": 0.0, + "step": 611 + }, + { + "epoch": 0.14, + "learning_rate": 4.8821354287034896e-05, + "loss": 0.0, + "step": 612 + }, + { + "epoch": 0.14, + "learning_rate": 4.881942839534705e-05, + "loss": 0.0278, + "step": 613 + }, + { + "epoch": 0.14, + "learning_rate": 4.88175025036592e-05, + "loss": 0.0, + "step": 614 + }, + { + "epoch": 0.14, + "learning_rate": 4.8815576611971345e-05, + "loss": 0.0263, + "step": 615 + }, + { + "epoch": 0.14, + "learning_rate": 4.881365072028349e-05, + "loss": 0.0951, + "step": 616 + }, + { + "epoch": 0.14, + "learning_rate": 4.881172482859564e-05, + "loss": 0.0005, + "step": 617 + }, + { + "epoch": 0.14, + "learning_rate": 4.8809798936907794e-05, + "loss": 0.0038, + "step": 618 + }, + { + "epoch": 0.14, + "learning_rate": 4.8807873045219935e-05, + "loss": 0.0291, + "step": 619 + }, + { + "epoch": 0.14, + "learning_rate": 4.880594715353209e-05, + "loss": 0.0101, + "step": 620 + }, + { + "epoch": 0.14, + "learning_rate": 4.880402126184424e-05, + "loss": 0.0, + "step": 621 + }, + { + "epoch": 0.14, + "learning_rate": 4.8802095370156384e-05, + "loss": 0.0026, + "step": 622 + }, + { + "epoch": 0.14, + "learning_rate": 4.880016947846853e-05, + "loss": 0.0408, + "step": 623 + }, + { + "epoch": 0.14, + "learning_rate": 4.8798243586780686e-05, + "loss": 0.0073, + "step": 624 + }, + { + "epoch": 0.14, + "learning_rate": 4.879631769509283e-05, + "loss": 0.0, + "step": 625 + }, + { + "epoch": 0.14, + "learning_rate": 4.8794391803404974e-05, + "loss": 0.0424, + "step": 626 + }, + { + "epoch": 0.14, + "learning_rate": 4.879246591171713e-05, + "loss": 0.0288, + "step": 627 + }, + { + "epoch": 0.15, + "learning_rate": 4.8790540020029276e-05, + "loss": 0.0362, + "step": 628 + }, + { + "epoch": 0.15, + "learning_rate": 4.878861412834142e-05, + "loss": 0.0317, + "step": 629 + }, + { + "epoch": 0.15, + "learning_rate": 4.878668823665357e-05, + "loss": 0.01, + "step": 630 + }, + { + "epoch": 0.15, + "learning_rate": 4.8784762344965725e-05, + "loss": 0.0706, + "step": 631 + }, + { + "epoch": 0.15, + "learning_rate": 4.878283645327787e-05, + "loss": 0.057, + "step": 632 + }, + { + "epoch": 0.15, + "learning_rate": 4.878091056159002e-05, + "loss": 0.0189, + "step": 633 + }, + { + "epoch": 0.15, + "learning_rate": 4.877898466990217e-05, + "loss": 0.0, + "step": 634 + }, + { + "epoch": 0.15, + "learning_rate": 4.8777058778214315e-05, + "loss": 0.0, + "step": 635 + }, + { + "epoch": 0.15, + "learning_rate": 4.877513288652646e-05, + "loss": 0.0283, + "step": 636 + }, + { + "epoch": 0.15, + "learning_rate": 4.877320699483861e-05, + "loss": 0.0203, + "step": 637 + }, + { + "epoch": 0.15, + "learning_rate": 4.8771281103150764e-05, + "loss": 0.0, + "step": 638 + }, + { + "epoch": 0.15, + "learning_rate": 4.876935521146291e-05, + "loss": 0.0002, + "step": 639 + }, + { + "epoch": 0.15, + "learning_rate": 4.876742931977506e-05, + "loss": 0.0389, + "step": 640 + }, + { + "epoch": 0.15, + "learning_rate": 4.8765503428087207e-05, + "loss": 0.0136, + "step": 641 + }, + { + "epoch": 0.15, + "learning_rate": 4.876357753639936e-05, + "loss": 0.0003, + "step": 642 + }, + { + "epoch": 0.15, + "learning_rate": 4.87616516447115e-05, + "loss": 0.0577, + "step": 643 + }, + { + "epoch": 0.15, + "learning_rate": 4.875972575302365e-05, + "loss": 0.0, + "step": 644 + }, + { + "epoch": 0.15, + "learning_rate": 4.87577998613358e-05, + "loss": 0.0041, + "step": 645 + }, + { + "epoch": 0.15, + "learning_rate": 4.8755873969647944e-05, + "loss": 0.111, + "step": 646 + }, + { + "epoch": 0.15, + "learning_rate": 4.87539480779601e-05, + "loss": 0.0, + "step": 647 + }, + { + "epoch": 0.15, + "learning_rate": 4.8752022186272246e-05, + "loss": 0.0696, + "step": 648 + }, + { + "epoch": 0.15, + "learning_rate": 4.875009629458439e-05, + "loss": 0.0353, + "step": 649 + }, + { + "epoch": 0.15, + "learning_rate": 4.874817040289654e-05, + "loss": 0.0002, + "step": 650 + }, + { + "epoch": 0.15, + "learning_rate": 4.8746244511208695e-05, + "loss": 0.0, + "step": 651 + }, + { + "epoch": 0.15, + "learning_rate": 4.874431861952084e-05, + "loss": 0.2208, + "step": 652 + }, + { + "epoch": 0.15, + "learning_rate": 4.874239272783299e-05, + "loss": 0.0258, + "step": 653 + }, + { + "epoch": 0.15, + "learning_rate": 4.874046683614514e-05, + "loss": 0.0116, + "step": 654 + }, + { + "epoch": 0.15, + "learning_rate": 4.8738540944457285e-05, + "loss": 0.0, + "step": 655 + }, + { + "epoch": 0.15, + "learning_rate": 4.873661505276943e-05, + "loss": 0.0, + "step": 656 + }, + { + "epoch": 0.15, + "learning_rate": 4.873468916108158e-05, + "loss": 0.009, + "step": 657 + }, + { + "epoch": 0.15, + "learning_rate": 4.8732763269393734e-05, + "loss": 0.0196, + "step": 658 + }, + { + "epoch": 0.15, + "learning_rate": 4.873083737770588e-05, + "loss": 0.0079, + "step": 659 + }, + { + "epoch": 0.15, + "learning_rate": 4.872891148601803e-05, + "loss": 0.1185, + "step": 660 + }, + { + "epoch": 0.15, + "learning_rate": 4.8726985594330176e-05, + "loss": 0.0437, + "step": 661 + }, + { + "epoch": 0.15, + "learning_rate": 4.872505970264233e-05, + "loss": 0.0297, + "step": 662 + }, + { + "epoch": 0.15, + "learning_rate": 4.872313381095447e-05, + "loss": 0.0176, + "step": 663 + }, + { + "epoch": 0.15, + "learning_rate": 4.872120791926662e-05, + "loss": 0.0022, + "step": 664 + }, + { + "epoch": 0.15, + "learning_rate": 4.871928202757877e-05, + "loss": 0.002, + "step": 665 + }, + { + "epoch": 0.15, + "learning_rate": 4.871735613589092e-05, + "loss": 0.0822, + "step": 666 + }, + { + "epoch": 0.15, + "learning_rate": 4.871543024420307e-05, + "loss": 0.0443, + "step": 667 + }, + { + "epoch": 0.15, + "learning_rate": 4.8713504352515216e-05, + "loss": 0.0, + "step": 668 + }, + { + "epoch": 0.15, + "learning_rate": 4.871157846082737e-05, + "loss": 0.0204, + "step": 669 + }, + { + "epoch": 0.15, + "learning_rate": 4.870965256913951e-05, + "loss": 0.0447, + "step": 670 + }, + { + "epoch": 0.16, + "learning_rate": 4.8707726677451665e-05, + "loss": 0.0, + "step": 671 + }, + { + "epoch": 0.16, + "learning_rate": 4.870580078576381e-05, + "loss": 0.0058, + "step": 672 + }, + { + "epoch": 0.16, + "learning_rate": 4.870387489407596e-05, + "loss": 0.0047, + "step": 673 + }, + { + "epoch": 0.16, + "learning_rate": 4.870194900238811e-05, + "loss": 0.0, + "step": 674 + }, + { + "epoch": 0.16, + "learning_rate": 4.8700023110700255e-05, + "loss": 0.0111, + "step": 675 + }, + { + "epoch": 0.16, + "learning_rate": 4.869809721901241e-05, + "loss": 0.0424, + "step": 676 + }, + { + "epoch": 0.16, + "learning_rate": 4.869617132732455e-05, + "loss": 0.0097, + "step": 677 + }, + { + "epoch": 0.16, + "learning_rate": 4.8694245435636704e-05, + "loss": 0.009, + "step": 678 + }, + { + "epoch": 0.16, + "learning_rate": 4.869231954394885e-05, + "loss": 0.0375, + "step": 679 + }, + { + "epoch": 0.16, + "learning_rate": 4.8690393652261e-05, + "loss": 0.0458, + "step": 680 + }, + { + "epoch": 0.16, + "learning_rate": 4.8688467760573146e-05, + "loss": 0.2009, + "step": 681 + }, + { + "epoch": 0.16, + "learning_rate": 4.8686541868885294e-05, + "loss": 0.0, + "step": 682 + }, + { + "epoch": 0.16, + "learning_rate": 4.868461597719745e-05, + "loss": 0.04, + "step": 683 + }, + { + "epoch": 0.16, + "learning_rate": 4.868269008550959e-05, + "loss": 0.043, + "step": 684 + }, + { + "epoch": 0.16, + "learning_rate": 4.868076419382174e-05, + "loss": 0.0063, + "step": 685 + }, + { + "epoch": 0.16, + "learning_rate": 4.867883830213389e-05, + "loss": 0.0, + "step": 686 + }, + { + "epoch": 0.16, + "learning_rate": 4.867691241044604e-05, + "loss": 0.0, + "step": 687 + }, + { + "epoch": 0.16, + "learning_rate": 4.8674986518758185e-05, + "loss": 0.0, + "step": 688 + }, + { + "epoch": 0.16, + "learning_rate": 4.867306062707034e-05, + "loss": 0.0082, + "step": 689 + }, + { + "epoch": 0.16, + "learning_rate": 4.867113473538249e-05, + "loss": 0.0232, + "step": 690 + }, + { + "epoch": 0.16, + "learning_rate": 4.866920884369463e-05, + "loss": 0.0552, + "step": 691 + }, + { + "epoch": 0.16, + "learning_rate": 4.866728295200678e-05, + "loss": 0.0036, + "step": 692 + }, + { + "epoch": 0.16, + "learning_rate": 4.866535706031893e-05, + "loss": 0.0122, + "step": 693 + }, + { + "epoch": 0.16, + "learning_rate": 4.866343116863108e-05, + "loss": 0.1008, + "step": 694 + }, + { + "epoch": 0.16, + "learning_rate": 4.8661505276943225e-05, + "loss": 0.0195, + "step": 695 + }, + { + "epoch": 0.16, + "learning_rate": 4.865957938525538e-05, + "loss": 0.0126, + "step": 696 + }, + { + "epoch": 0.16, + "learning_rate": 4.8657653493567526e-05, + "loss": 0.0014, + "step": 697 + }, + { + "epoch": 0.16, + "learning_rate": 4.8655727601879674e-05, + "loss": 0.0457, + "step": 698 + }, + { + "epoch": 0.16, + "learning_rate": 4.865380171019182e-05, + "loss": 0.0112, + "step": 699 + }, + { + "epoch": 0.16, + "learning_rate": 4.865187581850397e-05, + "loss": 0.0561, + "step": 700 + }, + { + "epoch": 0.16, + "learning_rate": 4.8649949926816116e-05, + "loss": 0.0, + "step": 701 + }, + { + "epoch": 0.16, + "learning_rate": 4.8648024035128264e-05, + "loss": 0.0007, + "step": 702 + }, + { + "epoch": 0.16, + "learning_rate": 4.864609814344042e-05, + "loss": 0.0169, + "step": 703 + }, + { + "epoch": 0.16, + "learning_rate": 4.8644172251752565e-05, + "loss": 0.0, + "step": 704 + }, + { + "epoch": 0.16, + "learning_rate": 4.864224636006471e-05, + "loss": 0.0313, + "step": 705 + }, + { + "epoch": 0.16, + "learning_rate": 4.864032046837686e-05, + "loss": 0.0791, + "step": 706 + }, + { + "epoch": 0.16, + "learning_rate": 4.8638394576689015e-05, + "loss": 0.002, + "step": 707 + }, + { + "epoch": 0.16, + "learning_rate": 4.8636468685001155e-05, + "loss": 0.0076, + "step": 708 + }, + { + "epoch": 0.16, + "learning_rate": 4.863454279331331e-05, + "loss": 0.0736, + "step": 709 + }, + { + "epoch": 0.16, + "learning_rate": 4.863261690162546e-05, + "loss": 0.0001, + "step": 710 + }, + { + "epoch": 0.16, + "learning_rate": 4.8630691009937604e-05, + "loss": 0.0181, + "step": 711 + }, + { + "epoch": 0.16, + "learning_rate": 4.862876511824975e-05, + "loss": 0.0727, + "step": 712 + }, + { + "epoch": 0.16, + "learning_rate": 4.86268392265619e-05, + "loss": 0.0364, + "step": 713 + }, + { + "epoch": 0.17, + "learning_rate": 4.8624913334874054e-05, + "loss": 0.0143, + "step": 714 + }, + { + "epoch": 0.17, + "learning_rate": 4.8622987443186194e-05, + "loss": 0.0, + "step": 715 + }, + { + "epoch": 0.17, + "learning_rate": 4.862106155149835e-05, + "loss": 0.0199, + "step": 716 + }, + { + "epoch": 0.17, + "learning_rate": 4.8619135659810496e-05, + "loss": 0.0235, + "step": 717 + }, + { + "epoch": 0.17, + "learning_rate": 4.8617209768122644e-05, + "loss": 0.02, + "step": 718 + }, + { + "epoch": 0.17, + "learning_rate": 4.861528387643479e-05, + "loss": 0.0, + "step": 719 + }, + { + "epoch": 0.17, + "learning_rate": 4.861335798474694e-05, + "loss": 0.0134, + "step": 720 + }, + { + "epoch": 0.17, + "learning_rate": 4.8611432093059086e-05, + "loss": 0.0404, + "step": 721 + }, + { + "epoch": 0.17, + "learning_rate": 4.8609506201371234e-05, + "loss": 0.0315, + "step": 722 + }, + { + "epoch": 0.17, + "learning_rate": 4.860758030968339e-05, + "loss": 0.0, + "step": 723 + }, + { + "epoch": 0.17, + "learning_rate": 4.8605654417995535e-05, + "loss": 0.0, + "step": 724 + }, + { + "epoch": 0.17, + "learning_rate": 4.860372852630768e-05, + "loss": 0.0001, + "step": 725 + }, + { + "epoch": 0.17, + "learning_rate": 4.860180263461983e-05, + "loss": 0.013, + "step": 726 + }, + { + "epoch": 0.17, + "learning_rate": 4.8599876742931984e-05, + "loss": 0.0233, + "step": 727 + }, + { + "epoch": 0.17, + "learning_rate": 4.8597950851244125e-05, + "loss": 0.0633, + "step": 728 + }, + { + "epoch": 0.17, + "learning_rate": 4.859602495955627e-05, + "loss": 0.0, + "step": 729 + }, + { + "epoch": 0.17, + "learning_rate": 4.859409906786843e-05, + "loss": 0.0093, + "step": 730 + }, + { + "epoch": 0.17, + "learning_rate": 4.8592173176180574e-05, + "loss": 0.0, + "step": 731 + }, + { + "epoch": 0.17, + "learning_rate": 4.859024728449272e-05, + "loss": 0.0408, + "step": 732 + }, + { + "epoch": 0.17, + "learning_rate": 4.858832139280487e-05, + "loss": 0.0683, + "step": 733 + }, + { + "epoch": 0.17, + "learning_rate": 4.8586395501117024e-05, + "loss": 0.1194, + "step": 734 + }, + { + "epoch": 0.17, + "learning_rate": 4.8584469609429164e-05, + "loss": 0.0052, + "step": 735 + }, + { + "epoch": 0.17, + "learning_rate": 4.858254371774132e-05, + "loss": 0.0, + "step": 736 + }, + { + "epoch": 0.17, + "learning_rate": 4.8580617826053466e-05, + "loss": 0.0032, + "step": 737 + }, + { + "epoch": 0.17, + "learning_rate": 4.8578691934365613e-05, + "loss": 0.0787, + "step": 738 + }, + { + "epoch": 0.17, + "learning_rate": 4.857676604267776e-05, + "loss": 0.0145, + "step": 739 + }, + { + "epoch": 0.17, + "learning_rate": 4.857484015098991e-05, + "loss": 0.0026, + "step": 740 + }, + { + "epoch": 0.17, + "learning_rate": 4.857291425930206e-05, + "loss": 0.0, + "step": 741 + }, + { + "epoch": 0.17, + "learning_rate": 4.8570988367614203e-05, + "loss": 0.067, + "step": 742 + }, + { + "epoch": 0.17, + "learning_rate": 4.856906247592636e-05, + "loss": 0.0288, + "step": 743 + }, + { + "epoch": 0.17, + "learning_rate": 4.8567136584238505e-05, + "loss": 0.0034, + "step": 744 + }, + { + "epoch": 0.17, + "learning_rate": 4.856521069255065e-05, + "loss": 0.1162, + "step": 745 + }, + { + "epoch": 0.17, + "learning_rate": 4.85632848008628e-05, + "loss": 0.0433, + "step": 746 + }, + { + "epoch": 0.17, + "learning_rate": 4.856135890917495e-05, + "loss": 0.1102, + "step": 747 + }, + { + "epoch": 0.17, + "learning_rate": 4.85594330174871e-05, + "loss": 0.0011, + "step": 748 + }, + { + "epoch": 0.17, + "learning_rate": 4.855750712579924e-05, + "loss": 0.0834, + "step": 749 + }, + { + "epoch": 0.17, + "learning_rate": 4.85555812341114e-05, + "loss": 0.0715, + "step": 750 + }, + { + "epoch": 0.17, + "learning_rate": 4.8553655342423544e-05, + "loss": 0.0, + "step": 751 + }, + { + "epoch": 0.17, + "learning_rate": 4.855172945073569e-05, + "loss": 0.0239, + "step": 752 + }, + { + "epoch": 0.17, + "learning_rate": 4.854980355904784e-05, + "loss": 0.0239, + "step": 753 + }, + { + "epoch": 0.17, + "learning_rate": 4.8547877667359993e-05, + "loss": 0.0874, + "step": 754 + }, + { + "epoch": 0.17, + "learning_rate": 4.854595177567214e-05, + "loss": 0.0267, + "step": 755 + }, + { + "epoch": 0.17, + "learning_rate": 4.854402588398428e-05, + "loss": 0.0, + "step": 756 + }, + { + "epoch": 0.17, + "learning_rate": 4.8542099992296436e-05, + "loss": 0.0068, + "step": 757 + }, + { + "epoch": 0.18, + "learning_rate": 4.854017410060858e-05, + "loss": 0.0035, + "step": 758 + }, + { + "epoch": 0.18, + "learning_rate": 4.853824820892073e-05, + "loss": 0.0511, + "step": 759 + }, + { + "epoch": 0.18, + "learning_rate": 4.853632231723288e-05, + "loss": 0.0002, + "step": 760 + }, + { + "epoch": 0.18, + "learning_rate": 4.853439642554503e-05, + "loss": 0.0133, + "step": 761 + }, + { + "epoch": 0.18, + "learning_rate": 4.853247053385718e-05, + "loss": 0.0043, + "step": 762 + }, + { + "epoch": 0.18, + "learning_rate": 4.853054464216933e-05, + "loss": 0.0016, + "step": 763 + }, + { + "epoch": 0.18, + "learning_rate": 4.8528618750481475e-05, + "loss": 0.0044, + "step": 764 + }, + { + "epoch": 0.18, + "learning_rate": 4.852669285879363e-05, + "loss": 0.0, + "step": 765 + }, + { + "epoch": 0.18, + "learning_rate": 4.852476696710577e-05, + "loss": 0.0514, + "step": 766 + }, + { + "epoch": 0.18, + "learning_rate": 4.852284107541792e-05, + "loss": 0.1247, + "step": 767 + }, + { + "epoch": 0.18, + "learning_rate": 4.852091518373007e-05, + "loss": 0.0005, + "step": 768 + }, + { + "epoch": 0.18, + "learning_rate": 4.851898929204222e-05, + "loss": 0.0147, + "step": 769 + }, + { + "epoch": 0.18, + "learning_rate": 4.851706340035437e-05, + "loss": 0.0011, + "step": 770 + }, + { + "epoch": 0.18, + "learning_rate": 4.8515137508666514e-05, + "loss": 0.0075, + "step": 771 + }, + { + "epoch": 0.18, + "learning_rate": 4.851321161697867e-05, + "loss": 0.0277, + "step": 772 + }, + { + "epoch": 0.18, + "learning_rate": 4.851128572529081e-05, + "loss": 0.0038, + "step": 773 + }, + { + "epoch": 0.18, + "learning_rate": 4.850935983360296e-05, + "loss": 0.1305, + "step": 774 + }, + { + "epoch": 0.18, + "learning_rate": 4.850743394191511e-05, + "loss": 0.0424, + "step": 775 + }, + { + "epoch": 0.18, + "learning_rate": 4.850550805022726e-05, + "loss": 0.0, + "step": 776 + }, + { + "epoch": 0.18, + "learning_rate": 4.8503582158539406e-05, + "loss": 0.0717, + "step": 777 + }, + { + "epoch": 0.18, + "learning_rate": 4.850165626685155e-05, + "loss": 0.0203, + "step": 778 + }, + { + "epoch": 0.18, + "learning_rate": 4.849973037516371e-05, + "loss": 0.0142, + "step": 779 + }, + { + "epoch": 0.18, + "learning_rate": 4.849780448347585e-05, + "loss": 0.0202, + "step": 780 + }, + { + "epoch": 0.18, + "learning_rate": 4.8495878591788e-05, + "loss": 0.0018, + "step": 781 + }, + { + "epoch": 0.18, + "learning_rate": 4.849395270010015e-05, + "loss": 0.0438, + "step": 782 + }, + { + "epoch": 0.18, + "learning_rate": 4.84920268084123e-05, + "loss": 0.0119, + "step": 783 + }, + { + "epoch": 0.18, + "learning_rate": 4.8490100916724445e-05, + "loss": 0.0523, + "step": 784 + }, + { + "epoch": 0.18, + "learning_rate": 4.848817502503659e-05, + "loss": 0.0183, + "step": 785 + }, + { + "epoch": 0.18, + "learning_rate": 4.848624913334875e-05, + "loss": 0.0, + "step": 786 + }, + { + "epoch": 0.18, + "learning_rate": 4.848432324166089e-05, + "loss": 0.0411, + "step": 787 + }, + { + "epoch": 0.18, + "learning_rate": 4.848239734997304e-05, + "loss": 0.0207, + "step": 788 + }, + { + "epoch": 0.18, + "learning_rate": 4.848047145828519e-05, + "loss": 0.0, + "step": 789 + }, + { + "epoch": 0.18, + "learning_rate": 4.8478545566597337e-05, + "loss": 0.0224, + "step": 790 + }, + { + "epoch": 0.18, + "learning_rate": 4.8476619674909484e-05, + "loss": 0.0323, + "step": 791 + }, + { + "epoch": 0.18, + "learning_rate": 4.847469378322164e-05, + "loss": 0.1266, + "step": 792 + }, + { + "epoch": 0.18, + "learning_rate": 4.8472767891533786e-05, + "loss": 0.0, + "step": 793 + }, + { + "epoch": 0.18, + "learning_rate": 4.8470841999845926e-05, + "loss": 0.0084, + "step": 794 + }, + { + "epoch": 0.18, + "learning_rate": 4.846891610815808e-05, + "loss": 0.0299, + "step": 795 + }, + { + "epoch": 0.18, + "learning_rate": 4.846699021647023e-05, + "loss": 0.0033, + "step": 796 + }, + { + "epoch": 0.18, + "learning_rate": 4.8465064324782376e-05, + "loss": 0.0, + "step": 797 + }, + { + "epoch": 0.18, + "learning_rate": 4.846313843309452e-05, + "loss": 0.011, + "step": 798 + }, + { + "epoch": 0.18, + "learning_rate": 4.846121254140668e-05, + "loss": 0.0138, + "step": 799 + }, + { + "epoch": 0.18, + "learning_rate": 4.845928664971882e-05, + "loss": 0.0, + "step": 800 + }, + { + "epoch": 0.19, + "learning_rate": 4.845736075803097e-05, + "loss": 0.0004, + "step": 801 + }, + { + "epoch": 0.19, + "learning_rate": 4.845543486634312e-05, + "loss": 0.0363, + "step": 802 + }, + { + "epoch": 0.19, + "learning_rate": 4.845350897465527e-05, + "loss": 0.0791, + "step": 803 + }, + { + "epoch": 0.19, + "learning_rate": 4.8451583082967415e-05, + "loss": 0.0258, + "step": 804 + }, + { + "epoch": 0.19, + "learning_rate": 4.844965719127956e-05, + "loss": 0.0835, + "step": 805 + }, + { + "epoch": 0.19, + "learning_rate": 4.8447731299591717e-05, + "loss": 0.0011, + "step": 806 + }, + { + "epoch": 0.19, + "learning_rate": 4.844580540790386e-05, + "loss": 0.0102, + "step": 807 + }, + { + "epoch": 0.19, + "learning_rate": 4.844387951621601e-05, + "loss": 0.0252, + "step": 808 + }, + { + "epoch": 0.19, + "learning_rate": 4.844195362452816e-05, + "loss": 0.0002, + "step": 809 + }, + { + "epoch": 0.19, + "learning_rate": 4.8440027732840306e-05, + "loss": 0.0002, + "step": 810 + }, + { + "epoch": 0.19, + "learning_rate": 4.8438101841152454e-05, + "loss": 0.0001, + "step": 811 + }, + { + "epoch": 0.19, + "learning_rate": 4.84361759494646e-05, + "loss": 0.0611, + "step": 812 + }, + { + "epoch": 0.19, + "learning_rate": 4.8434250057776756e-05, + "loss": 0.0, + "step": 813 + }, + { + "epoch": 0.19, + "learning_rate": 4.8432324166088896e-05, + "loss": 0.0248, + "step": 814 + }, + { + "epoch": 0.19, + "learning_rate": 4.843039827440105e-05, + "loss": 0.0586, + "step": 815 + }, + { + "epoch": 0.19, + "learning_rate": 4.84284723827132e-05, + "loss": 0.0408, + "step": 816 + }, + { + "epoch": 0.19, + "learning_rate": 4.8426546491025346e-05, + "loss": 0.0001, + "step": 817 + }, + { + "epoch": 0.19, + "learning_rate": 4.842462059933749e-05, + "loss": 0.0088, + "step": 818 + }, + { + "epoch": 0.19, + "learning_rate": 4.842269470764965e-05, + "loss": 0.0001, + "step": 819 + }, + { + "epoch": 0.19, + "learning_rate": 4.8420768815961795e-05, + "loss": 0.0246, + "step": 820 + }, + { + "epoch": 0.19, + "learning_rate": 4.841884292427394e-05, + "loss": 0.0608, + "step": 821 + }, + { + "epoch": 0.19, + "learning_rate": 4.841691703258609e-05, + "loss": 0.043, + "step": 822 + }, + { + "epoch": 0.19, + "learning_rate": 4.841499114089824e-05, + "loss": 0.0226, + "step": 823 + }, + { + "epoch": 0.19, + "learning_rate": 4.8413065249210385e-05, + "loss": 0.0001, + "step": 824 + }, + { + "epoch": 0.19, + "learning_rate": 4.841113935752253e-05, + "loss": 0.005, + "step": 825 + }, + { + "epoch": 0.19, + "learning_rate": 4.8409213465834686e-05, + "loss": 0.0062, + "step": 826 + }, + { + "epoch": 0.19, + "learning_rate": 4.8407287574146834e-05, + "loss": 0.0046, + "step": 827 + }, + { + "epoch": 0.19, + "learning_rate": 4.840536168245898e-05, + "loss": 0.0471, + "step": 828 + }, + { + "epoch": 0.19, + "learning_rate": 4.840343579077113e-05, + "loss": 0.0, + "step": 829 + }, + { + "epoch": 0.19, + "learning_rate": 4.840150989908328e-05, + "loss": 0.0549, + "step": 830 + }, + { + "epoch": 0.19, + "learning_rate": 4.8399584007395424e-05, + "loss": 0.0247, + "step": 831 + }, + { + "epoch": 0.19, + "learning_rate": 4.839765811570757e-05, + "loss": 0.0111, + "step": 832 + }, + { + "epoch": 0.19, + "learning_rate": 4.8395732224019726e-05, + "loss": 0.0039, + "step": 833 + }, + { + "epoch": 0.19, + "learning_rate": 4.839380633233187e-05, + "loss": 0.0005, + "step": 834 + }, + { + "epoch": 0.19, + "learning_rate": 4.839188044064402e-05, + "loss": 0.0021, + "step": 835 + }, + { + "epoch": 0.19, + "learning_rate": 4.838995454895617e-05, + "loss": 0.0001, + "step": 836 + }, + { + "epoch": 0.19, + "learning_rate": 4.838802865726832e-05, + "loss": 0.0002, + "step": 837 + }, + { + "epoch": 0.19, + "learning_rate": 4.838610276558046e-05, + "loss": 0.1488, + "step": 838 + }, + { + "epoch": 0.19, + "learning_rate": 4.838417687389262e-05, + "loss": 0.001, + "step": 839 + }, + { + "epoch": 0.19, + "learning_rate": 4.8382250982204765e-05, + "loss": 0.0062, + "step": 840 + }, + { + "epoch": 0.19, + "learning_rate": 4.838032509051691e-05, + "loss": 0.0001, + "step": 841 + }, + { + "epoch": 0.19, + "learning_rate": 4.837839919882906e-05, + "loss": 0.0001, + "step": 842 + }, + { + "epoch": 0.19, + "learning_rate": 4.837647330714121e-05, + "loss": 0.0046, + "step": 843 + }, + { + "epoch": 0.2, + "learning_rate": 4.837454741545336e-05, + "loss": 0.0003, + "step": 844 + }, + { + "epoch": 0.2, + "learning_rate": 4.83726215237655e-05, + "loss": 0.0, + "step": 845 + }, + { + "epoch": 0.2, + "learning_rate": 4.8370695632077656e-05, + "loss": 0.0278, + "step": 846 + }, + { + "epoch": 0.2, + "learning_rate": 4.8368769740389804e-05, + "loss": 0.045, + "step": 847 + }, + { + "epoch": 0.2, + "learning_rate": 4.836684384870195e-05, + "loss": 0.0794, + "step": 848 + }, + { + "epoch": 0.2, + "learning_rate": 4.83649179570141e-05, + "loss": 0.0, + "step": 849 + }, + { + "epoch": 0.2, + "learning_rate": 4.8362992065326246e-05, + "loss": 0.0016, + "step": 850 + }, + { + "epoch": 0.2, + "learning_rate": 4.83610661736384e-05, + "loss": 0.0198, + "step": 851 + }, + { + "epoch": 0.2, + "learning_rate": 4.835914028195054e-05, + "loss": 0.005, + "step": 852 + }, + { + "epoch": 0.2, + "learning_rate": 4.8357214390262695e-05, + "loss": 0.0083, + "step": 853 + }, + { + "epoch": 0.2, + "learning_rate": 4.835528849857484e-05, + "loss": 0.0055, + "step": 854 + }, + { + "epoch": 0.2, + "learning_rate": 4.835336260688699e-05, + "loss": 0.019, + "step": 855 + }, + { + "epoch": 0.2, + "learning_rate": 4.835143671519914e-05, + "loss": 0.4881, + "step": 856 + }, + { + "epoch": 0.2, + "learning_rate": 4.834951082351129e-05, + "loss": 0.046, + "step": 857 + }, + { + "epoch": 0.2, + "learning_rate": 4.834758493182344e-05, + "loss": 0.0, + "step": 858 + }, + { + "epoch": 0.2, + "learning_rate": 4.834565904013558e-05, + "loss": 0.0182, + "step": 859 + }, + { + "epoch": 0.2, + "learning_rate": 4.8343733148447735e-05, + "loss": 0.0071, + "step": 860 + }, + { + "epoch": 0.2, + "learning_rate": 4.834180725675988e-05, + "loss": 0.0137, + "step": 861 + }, + { + "epoch": 0.2, + "learning_rate": 4.833988136507203e-05, + "loss": 0.0405, + "step": 862 + }, + { + "epoch": 0.2, + "learning_rate": 4.833795547338418e-05, + "loss": 0.0045, + "step": 863 + }, + { + "epoch": 0.2, + "learning_rate": 4.833602958169633e-05, + "loss": 0.3273, + "step": 864 + }, + { + "epoch": 0.2, + "learning_rate": 4.833410369000848e-05, + "loss": 0.0119, + "step": 865 + }, + { + "epoch": 0.2, + "learning_rate": 4.8332177798320626e-05, + "loss": 0.0048, + "step": 866 + }, + { + "epoch": 0.2, + "learning_rate": 4.8330251906632774e-05, + "loss": 0.0227, + "step": 867 + }, + { + "epoch": 0.2, + "learning_rate": 4.832832601494492e-05, + "loss": 0.0002, + "step": 868 + }, + { + "epoch": 0.2, + "learning_rate": 4.832640012325707e-05, + "loss": 0.0164, + "step": 869 + }, + { + "epoch": 0.2, + "learning_rate": 4.8324474231569216e-05, + "loss": 0.025, + "step": 870 + }, + { + "epoch": 0.2, + "learning_rate": 4.832254833988137e-05, + "loss": 0.0166, + "step": 871 + }, + { + "epoch": 0.2, + "learning_rate": 4.832062244819351e-05, + "loss": 0.0196, + "step": 872 + }, + { + "epoch": 0.2, + "learning_rate": 4.8318696556505665e-05, + "loss": 0.0001, + "step": 873 + }, + { + "epoch": 0.2, + "learning_rate": 4.831677066481781e-05, + "loss": 0.1163, + "step": 874 + }, + { + "epoch": 0.2, + "learning_rate": 4.831484477312996e-05, + "loss": 0.0043, + "step": 875 + }, + { + "epoch": 0.2, + "learning_rate": 4.831291888144211e-05, + "loss": 0.0458, + "step": 876 + }, + { + "epoch": 0.2, + "learning_rate": 4.831099298975426e-05, + "loss": 0.0, + "step": 877 + }, + { + "epoch": 0.2, + "learning_rate": 4.830906709806641e-05, + "loss": 0.0001, + "step": 878 + }, + { + "epoch": 0.2, + "learning_rate": 4.830714120637855e-05, + "loss": 0.0086, + "step": 879 + }, + { + "epoch": 0.2, + "learning_rate": 4.8305215314690704e-05, + "loss": 0.0133, + "step": 880 + }, + { + "epoch": 0.2, + "learning_rate": 4.830328942300285e-05, + "loss": 0.3412, + "step": 881 + }, + { + "epoch": 0.2, + "learning_rate": 4.8301363531315e-05, + "loss": 0.0261, + "step": 882 + }, + { + "epoch": 0.2, + "learning_rate": 4.829943763962715e-05, + "loss": 0.0032, + "step": 883 + }, + { + "epoch": 0.2, + "learning_rate": 4.82975117479393e-05, + "loss": 0.1151, + "step": 884 + }, + { + "epoch": 0.2, + "learning_rate": 4.829558585625145e-05, + "loss": 0.0003, + "step": 885 + }, + { + "epoch": 0.2, + "learning_rate": 4.8293659964563596e-05, + "loss": 0.0004, + "step": 886 + }, + { + "epoch": 0.2, + "learning_rate": 4.8291734072875744e-05, + "loss": 0.0445, + "step": 887 + }, + { + "epoch": 0.21, + "learning_rate": 4.828980818118789e-05, + "loss": 0.0, + "step": 888 + }, + { + "epoch": 0.21, + "learning_rate": 4.828788228950004e-05, + "loss": 0.0011, + "step": 889 + }, + { + "epoch": 0.21, + "learning_rate": 4.8285956397812186e-05, + "loss": 0.0839, + "step": 890 + }, + { + "epoch": 0.21, + "learning_rate": 4.828403050612434e-05, + "loss": 0.3702, + "step": 891 + }, + { + "epoch": 0.21, + "learning_rate": 4.828210461443649e-05, + "loss": 0.0041, + "step": 892 + }, + { + "epoch": 0.21, + "learning_rate": 4.8280178722748635e-05, + "loss": 0.0, + "step": 893 + }, + { + "epoch": 0.21, + "learning_rate": 4.827825283106078e-05, + "loss": 0.0, + "step": 894 + }, + { + "epoch": 0.21, + "learning_rate": 4.827632693937294e-05, + "loss": 0.0383, + "step": 895 + }, + { + "epoch": 0.21, + "learning_rate": 4.827440104768508e-05, + "loss": 0.0175, + "step": 896 + }, + { + "epoch": 0.21, + "learning_rate": 4.8272475155997225e-05, + "loss": 0.0892, + "step": 897 + }, + { + "epoch": 0.21, + "learning_rate": 4.827054926430938e-05, + "loss": 0.001, + "step": 898 + }, + { + "epoch": 0.21, + "learning_rate": 4.826862337262153e-05, + "loss": 0.0151, + "step": 899 + }, + { + "epoch": 0.21, + "learning_rate": 4.8266697480933674e-05, + "loss": 0.0, + "step": 900 + }, + { + "epoch": 0.21, + "learning_rate": 4.826477158924582e-05, + "loss": 0.0004, + "step": 901 + }, + { + "epoch": 0.21, + "learning_rate": 4.8262845697557976e-05, + "loss": 0.0522, + "step": 902 + }, + { + "epoch": 0.21, + "learning_rate": 4.826091980587012e-05, + "loss": 0.0033, + "step": 903 + }, + { + "epoch": 0.21, + "learning_rate": 4.825899391418227e-05, + "loss": 0.0127, + "step": 904 + }, + { + "epoch": 0.21, + "learning_rate": 4.825706802249442e-05, + "loss": 0.0156, + "step": 905 + }, + { + "epoch": 0.21, + "learning_rate": 4.8255142130806566e-05, + "loss": 0.0, + "step": 906 + }, + { + "epoch": 0.21, + "learning_rate": 4.8253216239118713e-05, + "loss": 0.014, + "step": 907 + }, + { + "epoch": 0.21, + "learning_rate": 4.825129034743086e-05, + "loss": 0.0089, + "step": 908 + }, + { + "epoch": 0.21, + "learning_rate": 4.8249364455743015e-05, + "loss": 0.003, + "step": 909 + }, + { + "epoch": 0.21, + "learning_rate": 4.8247438564055156e-05, + "loss": 0.0022, + "step": 910 + }, + { + "epoch": 0.21, + "learning_rate": 4.824551267236731e-05, + "loss": 0.0003, + "step": 911 + }, + { + "epoch": 0.21, + "learning_rate": 4.824358678067946e-05, + "loss": 0.0118, + "step": 912 + }, + { + "epoch": 0.21, + "learning_rate": 4.8241660888991605e-05, + "loss": 0.0031, + "step": 913 + }, + { + "epoch": 0.21, + "learning_rate": 4.823973499730375e-05, + "loss": 0.0541, + "step": 914 + }, + { + "epoch": 0.21, + "learning_rate": 4.82378091056159e-05, + "loss": 0.0001, + "step": 915 + }, + { + "epoch": 0.21, + "learning_rate": 4.8235883213928054e-05, + "loss": 0.0, + "step": 916 + }, + { + "epoch": 0.21, + "learning_rate": 4.8233957322240195e-05, + "loss": 0.0, + "step": 917 + }, + { + "epoch": 0.21, + "learning_rate": 4.823203143055235e-05, + "loss": 0.0282, + "step": 918 + }, + { + "epoch": 0.21, + "learning_rate": 4.82301055388645e-05, + "loss": 0.1314, + "step": 919 + }, + { + "epoch": 0.21, + "learning_rate": 4.8228179647176644e-05, + "loss": 0.216, + "step": 920 + }, + { + "epoch": 0.21, + "learning_rate": 4.822625375548879e-05, + "loss": 0.0001, + "step": 921 + }, + { + "epoch": 0.21, + "learning_rate": 4.8224327863800946e-05, + "loss": 0.0002, + "step": 922 + }, + { + "epoch": 0.21, + "learning_rate": 4.8222401972113093e-05, + "loss": 0.056, + "step": 923 + }, + { + "epoch": 0.21, + "learning_rate": 4.8220476080425234e-05, + "loss": 0.0003, + "step": 924 + }, + { + "epoch": 0.21, + "learning_rate": 4.821855018873739e-05, + "loss": 0.1017, + "step": 925 + }, + { + "epoch": 0.21, + "learning_rate": 4.8216624297049536e-05, + "loss": 0.0052, + "step": 926 + }, + { + "epoch": 0.21, + "learning_rate": 4.821469840536168e-05, + "loss": 0.0161, + "step": 927 + }, + { + "epoch": 0.21, + "learning_rate": 4.821277251367383e-05, + "loss": 0.0061, + "step": 928 + }, + { + "epoch": 0.21, + "learning_rate": 4.8210846621985985e-05, + "loss": 0.0, + "step": 929 + }, + { + "epoch": 0.21, + "learning_rate": 4.820892073029813e-05, + "loss": 0.0, + "step": 930 + }, + { + "epoch": 0.22, + "learning_rate": 4.820699483861028e-05, + "loss": 0.0048, + "step": 931 + }, + { + "epoch": 0.22, + "learning_rate": 4.820506894692243e-05, + "loss": 0.0423, + "step": 932 + }, + { + "epoch": 0.22, + "learning_rate": 4.820314305523458e-05, + "loss": 0.0012, + "step": 933 + }, + { + "epoch": 0.22, + "learning_rate": 4.820121716354672e-05, + "loss": 0.0, + "step": 934 + }, + { + "epoch": 0.22, + "learning_rate": 4.819929127185887e-05, + "loss": 0.0018, + "step": 935 + }, + { + "epoch": 0.22, + "learning_rate": 4.8197365380171024e-05, + "loss": 0.0119, + "step": 936 + }, + { + "epoch": 0.22, + "learning_rate": 4.819543948848317e-05, + "loss": 0.0062, + "step": 937 + }, + { + "epoch": 0.22, + "learning_rate": 4.819351359679532e-05, + "loss": 0.0005, + "step": 938 + }, + { + "epoch": 0.22, + "learning_rate": 4.8191587705107467e-05, + "loss": 0.0674, + "step": 939 + }, + { + "epoch": 0.22, + "learning_rate": 4.818966181341962e-05, + "loss": 0.0007, + "step": 940 + }, + { + "epoch": 0.22, + "learning_rate": 4.818773592173176e-05, + "loss": 0.0123, + "step": 941 + }, + { + "epoch": 0.22, + "learning_rate": 4.8185810030043916e-05, + "loss": 0.0082, + "step": 942 + }, + { + "epoch": 0.22, + "learning_rate": 4.818388413835606e-05, + "loss": 0.0, + "step": 943 + }, + { + "epoch": 0.22, + "learning_rate": 4.8181958246668204e-05, + "loss": 0.0003, + "step": 944 + }, + { + "epoch": 0.22, + "learning_rate": 4.818003235498036e-05, + "loss": 0.0547, + "step": 945 + }, + { + "epoch": 0.22, + "learning_rate": 4.8178106463292506e-05, + "loss": 0.0551, + "step": 946 + }, + { + "epoch": 0.22, + "learning_rate": 4.817618057160466e-05, + "loss": 0.0029, + "step": 947 + }, + { + "epoch": 0.22, + "learning_rate": 4.81742546799168e-05, + "loss": 0.0002, + "step": 948 + }, + { + "epoch": 0.22, + "learning_rate": 4.8172328788228955e-05, + "loss": 0.0433, + "step": 949 + }, + { + "epoch": 0.22, + "learning_rate": 4.81704028965411e-05, + "loss": 0.0238, + "step": 950 + }, + { + "epoch": 0.22, + "learning_rate": 4.816847700485325e-05, + "loss": 0.0305, + "step": 951 + }, + { + "epoch": 0.22, + "learning_rate": 4.81665511131654e-05, + "loss": 0.0, + "step": 952 + }, + { + "epoch": 0.22, + "learning_rate": 4.8164625221477545e-05, + "loss": 0.1204, + "step": 953 + }, + { + "epoch": 0.22, + "learning_rate": 4.816269932978969e-05, + "loss": 0.0, + "step": 954 + }, + { + "epoch": 0.22, + "learning_rate": 4.816077343810184e-05, + "loss": 0.0019, + "step": 955 + }, + { + "epoch": 0.22, + "learning_rate": 4.8158847546413994e-05, + "loss": 0.0273, + "step": 956 + }, + { + "epoch": 0.22, + "learning_rate": 4.815692165472614e-05, + "loss": 0.0086, + "step": 957 + }, + { + "epoch": 0.22, + "learning_rate": 4.815499576303829e-05, + "loss": 0.0238, + "step": 958 + }, + { + "epoch": 0.22, + "learning_rate": 4.8153069871350436e-05, + "loss": 0.0171, + "step": 959 + }, + { + "epoch": 0.22, + "learning_rate": 4.815114397966259e-05, + "loss": 0.001, + "step": 960 + }, + { + "epoch": 0.22, + "learning_rate": 4.814921808797473e-05, + "loss": 0.0, + "step": 961 + }, + { + "epoch": 0.22, + "learning_rate": 4.814729219628688e-05, + "loss": 0.0, + "step": 962 + }, + { + "epoch": 0.22, + "learning_rate": 4.814536630459903e-05, + "loss": 0.0506, + "step": 963 + }, + { + "epoch": 0.22, + "learning_rate": 4.814344041291118e-05, + "loss": 0.0126, + "step": 964 + }, + { + "epoch": 0.22, + "learning_rate": 4.814151452122333e-05, + "loss": 0.0344, + "step": 965 + }, + { + "epoch": 0.22, + "learning_rate": 4.8139588629535476e-05, + "loss": 0.0703, + "step": 966 + }, + { + "epoch": 0.22, + "learning_rate": 4.813766273784763e-05, + "loss": 0.0, + "step": 967 + }, + { + "epoch": 0.22, + "learning_rate": 4.813573684615977e-05, + "loss": 0.002, + "step": 968 + }, + { + "epoch": 0.22, + "learning_rate": 4.8133810954471925e-05, + "loss": 0.001, + "step": 969 + }, + { + "epoch": 0.22, + "learning_rate": 4.813188506278407e-05, + "loss": 0.0019, + "step": 970 + }, + { + "epoch": 0.22, + "learning_rate": 4.812995917109622e-05, + "loss": 0.0007, + "step": 971 + }, + { + "epoch": 0.22, + "learning_rate": 4.812803327940837e-05, + "loss": 0.0073, + "step": 972 + }, + { + "epoch": 0.22, + "learning_rate": 4.8126107387720515e-05, + "loss": 0.0001, + "step": 973 + }, + { + "epoch": 0.23, + "learning_rate": 4.812418149603267e-05, + "loss": 0.0, + "step": 974 + }, + { + "epoch": 0.23, + "learning_rate": 4.812225560434481e-05, + "loss": 0.038, + "step": 975 + }, + { + "epoch": 0.23, + "learning_rate": 4.8120329712656964e-05, + "loss": 0.0456, + "step": 976 + }, + { + "epoch": 0.23, + "learning_rate": 4.811840382096911e-05, + "loss": 0.0169, + "step": 977 + }, + { + "epoch": 0.23, + "learning_rate": 4.811647792928126e-05, + "loss": 0.0328, + "step": 978 + }, + { + "epoch": 0.23, + "learning_rate": 4.8114552037593406e-05, + "loss": 0.0044, + "step": 979 + }, + { + "epoch": 0.23, + "learning_rate": 4.8112626145905554e-05, + "loss": 0.0022, + "step": 980 + }, + { + "epoch": 0.23, + "learning_rate": 4.811070025421771e-05, + "loss": 0.0063, + "step": 981 + }, + { + "epoch": 0.23, + "learning_rate": 4.810877436252985e-05, + "loss": 0.0019, + "step": 982 + }, + { + "epoch": 0.23, + "learning_rate": 4.8106848470842e-05, + "loss": 0.0, + "step": 983 + }, + { + "epoch": 0.23, + "learning_rate": 4.810492257915415e-05, + "loss": 0.0027, + "step": 984 + }, + { + "epoch": 0.23, + "learning_rate": 4.81029966874663e-05, + "loss": 0.0695, + "step": 985 + }, + { + "epoch": 0.23, + "learning_rate": 4.8101070795778445e-05, + "loss": 0.0047, + "step": 986 + }, + { + "epoch": 0.23, + "learning_rate": 4.80991449040906e-05, + "loss": 0.0083, + "step": 987 + }, + { + "epoch": 0.23, + "learning_rate": 4.809721901240275e-05, + "loss": 0.0373, + "step": 988 + }, + { + "epoch": 0.23, + "learning_rate": 4.8095293120714895e-05, + "loss": 0.0136, + "step": 989 + }, + { + "epoch": 0.23, + "learning_rate": 4.809336722902704e-05, + "loss": 0.0395, + "step": 990 + }, + { + "epoch": 0.23, + "learning_rate": 4.809144133733919e-05, + "loss": 0.134, + "step": 991 + }, + { + "epoch": 0.23, + "learning_rate": 4.808951544565134e-05, + "loss": 0.0123, + "step": 992 + }, + { + "epoch": 0.23, + "learning_rate": 4.8087589553963485e-05, + "loss": 0.0, + "step": 993 + }, + { + "epoch": 0.23, + "learning_rate": 4.808566366227564e-05, + "loss": 0.0585, + "step": 994 + }, + { + "epoch": 0.23, + "learning_rate": 4.8083737770587786e-05, + "loss": 0.0025, + "step": 995 + }, + { + "epoch": 0.23, + "learning_rate": 4.8081811878899934e-05, + "loss": 0.0, + "step": 996 + }, + { + "epoch": 0.23, + "learning_rate": 4.807988598721208e-05, + "loss": 0.0, + "step": 997 + }, + { + "epoch": 0.23, + "learning_rate": 4.8077960095524236e-05, + "loss": 0.0006, + "step": 998 + }, + { + "epoch": 0.23, + "learning_rate": 4.8076034203836376e-05, + "loss": 0.0446, + "step": 999 + }, + { + "epoch": 0.23, + "learning_rate": 4.8074108312148524e-05, + "loss": 0.0081, + "step": 1000 + }, + { + "epoch": 0.23, + "learning_rate": 4.807218242046068e-05, + "loss": 0.0028, + "step": 1001 + }, + { + "epoch": 0.23, + "learning_rate": 4.8070256528772825e-05, + "loss": 0.0001, + "step": 1002 + }, + { + "epoch": 0.23, + "learning_rate": 4.806833063708497e-05, + "loss": 0.0984, + "step": 1003 + }, + { + "epoch": 0.23, + "learning_rate": 4.806640474539712e-05, + "loss": 0.0255, + "step": 1004 + }, + { + "epoch": 0.23, + "learning_rate": 4.8064478853709275e-05, + "loss": 0.0386, + "step": 1005 + }, + { + "epoch": 0.23, + "learning_rate": 4.8062552962021415e-05, + "loss": 0.0109, + "step": 1006 + }, + { + "epoch": 0.23, + "learning_rate": 4.806062707033357e-05, + "loss": 0.0, + "step": 1007 + }, + { + "epoch": 0.23, + "learning_rate": 4.805870117864572e-05, + "loss": 0.0108, + "step": 1008 + }, + { + "epoch": 0.23, + "learning_rate": 4.8056775286957865e-05, + "loss": 0.0112, + "step": 1009 + }, + { + "epoch": 0.23, + "learning_rate": 4.805484939527001e-05, + "loss": 0.0192, + "step": 1010 + }, + { + "epoch": 0.23, + "learning_rate": 4.805292350358216e-05, + "loss": 0.0009, + "step": 1011 + }, + { + "epoch": 0.23, + "learning_rate": 4.8050997611894314e-05, + "loss": 0.0303, + "step": 1012 + }, + { + "epoch": 0.23, + "learning_rate": 4.8049071720206454e-05, + "loss": 0.0015, + "step": 1013 + }, + { + "epoch": 0.23, + "learning_rate": 4.804714582851861e-05, + "loss": 0.016, + "step": 1014 + }, + { + "epoch": 0.23, + "learning_rate": 4.8045219936830756e-05, + "loss": 0.0811, + "step": 1015 + }, + { + "epoch": 0.23, + "learning_rate": 4.8043294045142904e-05, + "loss": 0.0071, + "step": 1016 + }, + { + "epoch": 0.24, + "learning_rate": 4.804136815345505e-05, + "loss": 0.0, + "step": 1017 + }, + { + "epoch": 0.24, + "learning_rate": 4.80394422617672e-05, + "loss": 0.011, + "step": 1018 + }, + { + "epoch": 0.24, + "learning_rate": 4.803751637007935e-05, + "loss": 0.0513, + "step": 1019 + }, + { + "epoch": 0.24, + "learning_rate": 4.8035590478391494e-05, + "loss": 0.0256, + "step": 1020 + }, + { + "epoch": 0.24, + "learning_rate": 4.803366458670365e-05, + "loss": 0.0, + "step": 1021 + }, + { + "epoch": 0.24, + "learning_rate": 4.8031738695015795e-05, + "loss": 0.0118, + "step": 1022 + }, + { + "epoch": 0.24, + "learning_rate": 4.802981280332794e-05, + "loss": 0.0189, + "step": 1023 + }, + { + "epoch": 0.24, + "learning_rate": 4.802788691164009e-05, + "loss": 0.0033, + "step": 1024 + }, + { + "epoch": 0.24, + "learning_rate": 4.8025961019952245e-05, + "loss": 0.0041, + "step": 1025 + }, + { + "epoch": 0.24, + "learning_rate": 4.8024035128264385e-05, + "loss": 0.0008, + "step": 1026 + }, + { + "epoch": 0.24, + "learning_rate": 4.802210923657653e-05, + "loss": 0.0, + "step": 1027 + }, + { + "epoch": 0.24, + "learning_rate": 4.802018334488869e-05, + "loss": 0.0, + "step": 1028 + }, + { + "epoch": 0.24, + "learning_rate": 4.8018257453200834e-05, + "loss": 0.0236, + "step": 1029 + }, + { + "epoch": 0.24, + "learning_rate": 4.801633156151298e-05, + "loss": 0.0, + "step": 1030 + }, + { + "epoch": 0.24, + "learning_rate": 4.801440566982513e-05, + "loss": 0.0007, + "step": 1031 + }, + { + "epoch": 0.24, + "learning_rate": 4.8012479778137284e-05, + "loss": 0.0151, + "step": 1032 + }, + { + "epoch": 0.24, + "learning_rate": 4.8010553886449424e-05, + "loss": 0.0001, + "step": 1033 + }, + { + "epoch": 0.24, + "learning_rate": 4.800862799476158e-05, + "loss": 0.0113, + "step": 1034 + }, + { + "epoch": 0.24, + "learning_rate": 4.8006702103073726e-05, + "loss": 0.0024, + "step": 1035 + }, + { + "epoch": 0.24, + "learning_rate": 4.8004776211385874e-05, + "loss": 0.0004, + "step": 1036 + }, + { + "epoch": 0.24, + "learning_rate": 4.800285031969802e-05, + "loss": 0.0, + "step": 1037 + }, + { + "epoch": 0.24, + "learning_rate": 4.800092442801017e-05, + "loss": 0.0004, + "step": 1038 + }, + { + "epoch": 0.24, + "learning_rate": 4.799899853632232e-05, + "loss": 0.0011, + "step": 1039 + }, + { + "epoch": 0.24, + "learning_rate": 4.7997072644634463e-05, + "loss": 0.0, + "step": 1040 + }, + { + "epoch": 0.24, + "learning_rate": 4.799514675294662e-05, + "loss": 0.0001, + "step": 1041 + }, + { + "epoch": 0.24, + "learning_rate": 4.7993220861258765e-05, + "loss": 0.0408, + "step": 1042 + }, + { + "epoch": 0.24, + "learning_rate": 4.799129496957091e-05, + "loss": 0.019, + "step": 1043 + }, + { + "epoch": 0.24, + "learning_rate": 4.798936907788306e-05, + "loss": 0.002, + "step": 1044 + }, + { + "epoch": 0.24, + "learning_rate": 4.7987443186195214e-05, + "loss": 0.0012, + "step": 1045 + }, + { + "epoch": 0.24, + "learning_rate": 4.798551729450736e-05, + "loss": 0.0092, + "step": 1046 + }, + { + "epoch": 0.24, + "learning_rate": 4.79835914028195e-05, + "loss": 0.0405, + "step": 1047 + }, + { + "epoch": 0.24, + "learning_rate": 4.798166551113166e-05, + "loss": 0.0258, + "step": 1048 + }, + { + "epoch": 0.24, + "learning_rate": 4.7979739619443804e-05, + "loss": 0.0134, + "step": 1049 + }, + { + "epoch": 0.24, + "learning_rate": 4.797781372775595e-05, + "loss": 0.0034, + "step": 1050 + }, + { + "epoch": 0.24, + "learning_rate": 4.79758878360681e-05, + "loss": 0.0081, + "step": 1051 + }, + { + "epoch": 0.24, + "learning_rate": 4.7973961944380254e-05, + "loss": 0.0, + "step": 1052 + }, + { + "epoch": 0.24, + "learning_rate": 4.79720360526924e-05, + "loss": 0.0327, + "step": 1053 + }, + { + "epoch": 0.24, + "learning_rate": 4.797011016100455e-05, + "loss": 0.0089, + "step": 1054 + }, + { + "epoch": 0.24, + "learning_rate": 4.7968184269316696e-05, + "loss": 0.0043, + "step": 1055 + }, + { + "epoch": 0.24, + "learning_rate": 4.7966258377628843e-05, + "loss": 0.0001, + "step": 1056 + }, + { + "epoch": 0.24, + "learning_rate": 4.796433248594099e-05, + "loss": 0.0496, + "step": 1057 + }, + { + "epoch": 0.24, + "learning_rate": 4.796240659425314e-05, + "loss": 0.2024, + "step": 1058 + }, + { + "epoch": 0.24, + "learning_rate": 4.796048070256529e-05, + "loss": 0.0022, + "step": 1059 + }, + { + "epoch": 0.24, + "learning_rate": 4.795855481087744e-05, + "loss": 0.0, + "step": 1060 + }, + { + "epoch": 0.25, + "learning_rate": 4.795662891918959e-05, + "loss": 0.0004, + "step": 1061 + }, + { + "epoch": 0.25, + "learning_rate": 4.7954703027501735e-05, + "loss": 0.0159, + "step": 1062 + }, + { + "epoch": 0.25, + "learning_rate": 4.795277713581389e-05, + "loss": 0.001, + "step": 1063 + }, + { + "epoch": 0.25, + "learning_rate": 4.795085124412603e-05, + "loss": 0.0012, + "step": 1064 + }, + { + "epoch": 0.25, + "learning_rate": 4.794892535243818e-05, + "loss": 0.4181, + "step": 1065 + }, + { + "epoch": 0.25, + "learning_rate": 4.794699946075033e-05, + "loss": 0.045, + "step": 1066 + }, + { + "epoch": 0.25, + "learning_rate": 4.794507356906248e-05, + "loss": 0.0058, + "step": 1067 + }, + { + "epoch": 0.25, + "learning_rate": 4.794314767737463e-05, + "loss": 0.0009, + "step": 1068 + }, + { + "epoch": 0.25, + "learning_rate": 4.7941221785686774e-05, + "loss": 0.0495, + "step": 1069 + }, + { + "epoch": 0.25, + "learning_rate": 4.793929589399893e-05, + "loss": 0.0034, + "step": 1070 + }, + { + "epoch": 0.25, + "learning_rate": 4.793737000231107e-05, + "loss": 0.0013, + "step": 1071 + }, + { + "epoch": 0.25, + "learning_rate": 4.7935444110623223e-05, + "loss": 0.0217, + "step": 1072 + }, + { + "epoch": 0.25, + "learning_rate": 4.793351821893537e-05, + "loss": 0.0053, + "step": 1073 + }, + { + "epoch": 0.25, + "learning_rate": 4.793159232724752e-05, + "loss": 0.0037, + "step": 1074 + }, + { + "epoch": 0.25, + "learning_rate": 4.7929666435559666e-05, + "loss": 0.0028, + "step": 1075 + }, + { + "epoch": 0.25, + "learning_rate": 4.792774054387181e-05, + "loss": 0.0017, + "step": 1076 + }, + { + "epoch": 0.25, + "learning_rate": 4.792581465218397e-05, + "loss": 0.0148, + "step": 1077 + }, + { + "epoch": 0.25, + "learning_rate": 4.792388876049611e-05, + "loss": 0.0556, + "step": 1078 + }, + { + "epoch": 0.25, + "learning_rate": 4.792196286880826e-05, + "loss": 0.0334, + "step": 1079 + }, + { + "epoch": 0.25, + "learning_rate": 4.792003697712041e-05, + "loss": 0.0, + "step": 1080 + }, + { + "epoch": 0.25, + "learning_rate": 4.791811108543256e-05, + "loss": 0.0008, + "step": 1081 + }, + { + "epoch": 0.25, + "learning_rate": 4.7916185193744705e-05, + "loss": 0.0015, + "step": 1082 + }, + { + "epoch": 0.25, + "learning_rate": 4.791425930205685e-05, + "loss": 0.0104, + "step": 1083 + }, + { + "epoch": 0.25, + "learning_rate": 4.791233341036901e-05, + "loss": 0.0011, + "step": 1084 + }, + { + "epoch": 0.25, + "learning_rate": 4.791040751868115e-05, + "loss": 0.0154, + "step": 1085 + }, + { + "epoch": 0.25, + "learning_rate": 4.79084816269933e-05, + "loss": 0.0, + "step": 1086 + }, + { + "epoch": 0.25, + "learning_rate": 4.790655573530545e-05, + "loss": 0.0052, + "step": 1087 + }, + { + "epoch": 0.25, + "learning_rate": 4.79046298436176e-05, + "loss": 0.0018, + "step": 1088 + }, + { + "epoch": 0.25, + "learning_rate": 4.7902703951929744e-05, + "loss": 0.0534, + "step": 1089 + }, + { + "epoch": 0.25, + "learning_rate": 4.79007780602419e-05, + "loss": 0.0004, + "step": 1090 + }, + { + "epoch": 0.25, + "learning_rate": 4.7898852168554046e-05, + "loss": 0.0879, + "step": 1091 + }, + { + "epoch": 0.25, + "learning_rate": 4.7896926276866187e-05, + "loss": 0.0244, + "step": 1092 + }, + { + "epoch": 0.25, + "learning_rate": 4.789500038517834e-05, + "loss": 0.0057, + "step": 1093 + }, + { + "epoch": 0.25, + "learning_rate": 4.789307449349049e-05, + "loss": 0.0, + "step": 1094 + }, + { + "epoch": 0.25, + "learning_rate": 4.7891148601802636e-05, + "loss": 0.0, + "step": 1095 + }, + { + "epoch": 0.25, + "learning_rate": 4.788922271011478e-05, + "loss": 0.049, + "step": 1096 + }, + { + "epoch": 0.25, + "learning_rate": 4.788729681842694e-05, + "loss": 0.0012, + "step": 1097 + }, + { + "epoch": 0.25, + "learning_rate": 4.788537092673908e-05, + "loss": 0.0001, + "step": 1098 + }, + { + "epoch": 0.25, + "learning_rate": 4.788344503505123e-05, + "loss": 0.0303, + "step": 1099 + }, + { + "epoch": 0.25, + "learning_rate": 4.788151914336338e-05, + "loss": 0.0019, + "step": 1100 + }, + { + "epoch": 0.25, + "learning_rate": 4.7879593251675534e-05, + "loss": 0.0333, + "step": 1101 + }, + { + "epoch": 0.25, + "learning_rate": 4.7877667359987675e-05, + "loss": 0.0046, + "step": 1102 + }, + { + "epoch": 0.25, + "learning_rate": 4.787574146829982e-05, + "loss": 0.0021, + "step": 1103 + }, + { + "epoch": 0.26, + "learning_rate": 4.7873815576611977e-05, + "loss": 0.0078, + "step": 1104 + }, + { + "epoch": 0.26, + "learning_rate": 4.787188968492412e-05, + "loss": 0.0001, + "step": 1105 + }, + { + "epoch": 0.26, + "learning_rate": 4.786996379323627e-05, + "loss": 0.0083, + "step": 1106 + }, + { + "epoch": 0.26, + "learning_rate": 4.786803790154842e-05, + "loss": 0.0369, + "step": 1107 + }, + { + "epoch": 0.26, + "learning_rate": 4.7866112009860567e-05, + "loss": 0.0696, + "step": 1108 + }, + { + "epoch": 0.26, + "learning_rate": 4.7864186118172714e-05, + "loss": 0.0001, + "step": 1109 + }, + { + "epoch": 0.26, + "learning_rate": 4.786226022648487e-05, + "loss": 0.0001, + "step": 1110 + }, + { + "epoch": 0.26, + "learning_rate": 4.7860334334797016e-05, + "loss": 0.0384, + "step": 1111 + }, + { + "epoch": 0.26, + "learning_rate": 4.7858408443109156e-05, + "loss": 0.0, + "step": 1112 + }, + { + "epoch": 0.26, + "learning_rate": 4.785648255142131e-05, + "loss": 0.0, + "step": 1113 + }, + { + "epoch": 0.26, + "learning_rate": 4.785455665973346e-05, + "loss": 0.001, + "step": 1114 + }, + { + "epoch": 0.26, + "learning_rate": 4.7852630768045606e-05, + "loss": 0.0085, + "step": 1115 + }, + { + "epoch": 0.26, + "learning_rate": 4.785070487635775e-05, + "loss": 0.0125, + "step": 1116 + }, + { + "epoch": 0.26, + "learning_rate": 4.784877898466991e-05, + "loss": 0.0001, + "step": 1117 + }, + { + "epoch": 0.26, + "learning_rate": 4.7846853092982055e-05, + "loss": 0.0043, + "step": 1118 + }, + { + "epoch": 0.26, + "learning_rate": 4.78449272012942e-05, + "loss": 0.0123, + "step": 1119 + }, + { + "epoch": 0.26, + "learning_rate": 4.784300130960635e-05, + "loss": 0.0001, + "step": 1120 + }, + { + "epoch": 0.26, + "learning_rate": 4.78410754179185e-05, + "loss": 0.0006, + "step": 1121 + }, + { + "epoch": 0.26, + "learning_rate": 4.7839149526230645e-05, + "loss": 0.0738, + "step": 1122 + }, + { + "epoch": 0.26, + "learning_rate": 4.783722363454279e-05, + "loss": 0.0139, + "step": 1123 + }, + { + "epoch": 0.26, + "learning_rate": 4.7835297742854946e-05, + "loss": 0.0017, + "step": 1124 + }, + { + "epoch": 0.26, + "learning_rate": 4.7833371851167094e-05, + "loss": 0.0, + "step": 1125 + }, + { + "epoch": 0.26, + "learning_rate": 4.783144595947924e-05, + "loss": 0.0374, + "step": 1126 + }, + { + "epoch": 0.26, + "learning_rate": 4.782952006779139e-05, + "loss": 0.0041, + "step": 1127 + }, + { + "epoch": 0.26, + "learning_rate": 4.782759417610354e-05, + "loss": 0.001, + "step": 1128 + }, + { + "epoch": 0.26, + "learning_rate": 4.7825668284415684e-05, + "loss": 0.0001, + "step": 1129 + }, + { + "epoch": 0.26, + "learning_rate": 4.782374239272783e-05, + "loss": 0.0019, + "step": 1130 + }, + { + "epoch": 0.26, + "learning_rate": 4.7821816501039986e-05, + "loss": 0.0011, + "step": 1131 + }, + { + "epoch": 0.26, + "learning_rate": 4.781989060935213e-05, + "loss": 0.0001, + "step": 1132 + }, + { + "epoch": 0.26, + "learning_rate": 4.781796471766428e-05, + "loss": 0.0232, + "step": 1133 + }, + { + "epoch": 0.26, + "learning_rate": 4.781603882597643e-05, + "loss": 0.0637, + "step": 1134 + }, + { + "epoch": 0.26, + "learning_rate": 4.781411293428858e-05, + "loss": 0.0013, + "step": 1135 + }, + { + "epoch": 0.26, + "learning_rate": 4.781218704260072e-05, + "loss": 0.0138, + "step": 1136 + }, + { + "epoch": 0.26, + "learning_rate": 4.781026115091288e-05, + "loss": 0.0001, + "step": 1137 + }, + { + "epoch": 0.26, + "learning_rate": 4.7808335259225025e-05, + "loss": 0.0087, + "step": 1138 + }, + { + "epoch": 0.26, + "learning_rate": 4.780640936753717e-05, + "loss": 0.0172, + "step": 1139 + }, + { + "epoch": 0.26, + "learning_rate": 4.780448347584932e-05, + "loss": 0.0, + "step": 1140 + }, + { + "epoch": 0.26, + "learning_rate": 4.780255758416147e-05, + "loss": 0.0024, + "step": 1141 + }, + { + "epoch": 0.26, + "learning_rate": 4.780063169247362e-05, + "loss": 0.0132, + "step": 1142 + }, + { + "epoch": 0.26, + "learning_rate": 4.779870580078576e-05, + "loss": 0.0174, + "step": 1143 + }, + { + "epoch": 0.26, + "learning_rate": 4.7796779909097916e-05, + "loss": 0.0003, + "step": 1144 + }, + { + "epoch": 0.26, + "learning_rate": 4.7794854017410064e-05, + "loss": 0.0019, + "step": 1145 + }, + { + "epoch": 0.26, + "learning_rate": 4.779292812572221e-05, + "loss": 0.0227, + "step": 1146 + }, + { + "epoch": 0.27, + "learning_rate": 4.779100223403436e-05, + "loss": 0.0, + "step": 1147 + }, + { + "epoch": 0.27, + "learning_rate": 4.7789076342346506e-05, + "loss": 0.0058, + "step": 1148 + }, + { + "epoch": 0.27, + "learning_rate": 4.778715045065866e-05, + "loss": 0.0027, + "step": 1149 + }, + { + "epoch": 0.27, + "learning_rate": 4.77852245589708e-05, + "loss": 0.0413, + "step": 1150 + }, + { + "epoch": 0.27, + "learning_rate": 4.7783298667282955e-05, + "loss": 0.004, + "step": 1151 + }, + { + "epoch": 0.27, + "learning_rate": 4.77813727755951e-05, + "loss": 0.0001, + "step": 1152 + }, + { + "epoch": 0.27, + "learning_rate": 4.777944688390725e-05, + "loss": 0.0048, + "step": 1153 + }, + { + "epoch": 0.27, + "learning_rate": 4.77775209922194e-05, + "loss": 0.0601, + "step": 1154 + }, + { + "epoch": 0.27, + "learning_rate": 4.777559510053155e-05, + "loss": 0.086, + "step": 1155 + }, + { + "epoch": 0.27, + "learning_rate": 4.77736692088437e-05, + "loss": 0.0055, + "step": 1156 + }, + { + "epoch": 0.27, + "learning_rate": 4.777174331715585e-05, + "loss": 0.0, + "step": 1157 + }, + { + "epoch": 0.27, + "learning_rate": 4.7769817425467995e-05, + "loss": 0.0233, + "step": 1158 + }, + { + "epoch": 0.27, + "learning_rate": 4.776789153378014e-05, + "loss": 0.0136, + "step": 1159 + }, + { + "epoch": 0.27, + "learning_rate": 4.776596564209229e-05, + "loss": 0.0383, + "step": 1160 + }, + { + "epoch": 0.27, + "learning_rate": 4.776403975040444e-05, + "loss": 0.0164, + "step": 1161 + }, + { + "epoch": 0.27, + "learning_rate": 4.776211385871659e-05, + "loss": 0.0258, + "step": 1162 + }, + { + "epoch": 0.27, + "learning_rate": 4.776018796702874e-05, + "loss": 0.0297, + "step": 1163 + }, + { + "epoch": 0.27, + "learning_rate": 4.7758262075340886e-05, + "loss": 0.0042, + "step": 1164 + }, + { + "epoch": 0.27, + "learning_rate": 4.7756336183653034e-05, + "loss": 0.0, + "step": 1165 + }, + { + "epoch": 0.27, + "learning_rate": 4.775441029196519e-05, + "loss": 0.0395, + "step": 1166 + }, + { + "epoch": 0.27, + "learning_rate": 4.775248440027733e-05, + "loss": 0.0022, + "step": 1167 + }, + { + "epoch": 0.27, + "learning_rate": 4.7750558508589476e-05, + "loss": 0.0258, + "step": 1168 + }, + { + "epoch": 0.27, + "learning_rate": 4.774863261690163e-05, + "loss": 0.0046, + "step": 1169 + }, + { + "epoch": 0.27, + "learning_rate": 4.774670672521378e-05, + "loss": 0.0003, + "step": 1170 + }, + { + "epoch": 0.27, + "learning_rate": 4.7744780833525925e-05, + "loss": 0.0078, + "step": 1171 + }, + { + "epoch": 0.27, + "learning_rate": 4.774285494183807e-05, + "loss": 0.0001, + "step": 1172 + }, + { + "epoch": 0.27, + "learning_rate": 4.774092905015023e-05, + "loss": 0.015, + "step": 1173 + }, + { + "epoch": 0.27, + "learning_rate": 4.773900315846237e-05, + "loss": 0.0321, + "step": 1174 + }, + { + "epoch": 0.27, + "learning_rate": 4.773707726677452e-05, + "loss": 0.0093, + "step": 1175 + }, + { + "epoch": 0.27, + "learning_rate": 4.773515137508667e-05, + "loss": 0.0452, + "step": 1176 + }, + { + "epoch": 0.27, + "learning_rate": 4.773322548339881e-05, + "loss": 0.0037, + "step": 1177 + }, + { + "epoch": 0.27, + "learning_rate": 4.7731299591710964e-05, + "loss": 0.0, + "step": 1178 + }, + { + "epoch": 0.27, + "learning_rate": 4.772937370002311e-05, + "loss": 0.0569, + "step": 1179 + }, + { + "epoch": 0.27, + "learning_rate": 4.772744780833526e-05, + "loss": 0.0162, + "step": 1180 + }, + { + "epoch": 0.27, + "learning_rate": 4.772552191664741e-05, + "loss": 0.0011, + "step": 1181 + }, + { + "epoch": 0.27, + "learning_rate": 4.772359602495956e-05, + "loss": 0.0, + "step": 1182 + }, + { + "epoch": 0.27, + "learning_rate": 4.772167013327171e-05, + "loss": 0.0, + "step": 1183 + }, + { + "epoch": 0.27, + "learning_rate": 4.7719744241583856e-05, + "loss": 0.0259, + "step": 1184 + }, + { + "epoch": 0.27, + "learning_rate": 4.7717818349896004e-05, + "loss": 0.0344, + "step": 1185 + }, + { + "epoch": 0.27, + "learning_rate": 4.771589245820815e-05, + "loss": 0.0124, + "step": 1186 + }, + { + "epoch": 0.27, + "learning_rate": 4.77139665665203e-05, + "loss": 0.0085, + "step": 1187 + }, + { + "epoch": 0.27, + "learning_rate": 4.7712040674832446e-05, + "loss": 0.0022, + "step": 1188 + }, + { + "epoch": 0.27, + "learning_rate": 4.77101147831446e-05, + "loss": 0.0004, + "step": 1189 + }, + { + "epoch": 0.28, + "learning_rate": 4.770818889145675e-05, + "loss": 0.0001, + "step": 1190 + }, + { + "epoch": 0.28, + "learning_rate": 4.7706262999768895e-05, + "loss": 0.0056, + "step": 1191 + }, + { + "epoch": 0.28, + "learning_rate": 4.770433710808104e-05, + "loss": 0.0, + "step": 1192 + }, + { + "epoch": 0.28, + "learning_rate": 4.77024112163932e-05, + "loss": 0.0537, + "step": 1193 + }, + { + "epoch": 0.28, + "learning_rate": 4.770048532470534e-05, + "loss": 0.0208, + "step": 1194 + }, + { + "epoch": 0.28, + "learning_rate": 4.7698559433017485e-05, + "loss": 0.0001, + "step": 1195 + }, + { + "epoch": 0.28, + "learning_rate": 4.769663354132964e-05, + "loss": 0.005, + "step": 1196 + }, + { + "epoch": 0.28, + "learning_rate": 4.769470764964179e-05, + "loss": 0.0784, + "step": 1197 + }, + { + "epoch": 0.28, + "learning_rate": 4.7692781757953934e-05, + "loss": 0.0097, + "step": 1198 + }, + { + "epoch": 0.28, + "learning_rate": 4.769085586626608e-05, + "loss": 0.002, + "step": 1199 + }, + { + "epoch": 0.28, + "learning_rate": 4.7688929974578236e-05, + "loss": 0.0001, + "step": 1200 + }, + { + "epoch": 0.28, + "learning_rate": 4.768700408289038e-05, + "loss": 0.0, + "step": 1201 + }, + { + "epoch": 0.28, + "learning_rate": 4.768507819120253e-05, + "loss": 0.0, + "step": 1202 + }, + { + "epoch": 0.28, + "learning_rate": 4.768315229951468e-05, + "loss": 0.0008, + "step": 1203 + }, + { + "epoch": 0.28, + "learning_rate": 4.7681226407826826e-05, + "loss": 0.001, + "step": 1204 + }, + { + "epoch": 0.28, + "learning_rate": 4.7679300516138974e-05, + "loss": 0.0092, + "step": 1205 + }, + { + "epoch": 0.28, + "learning_rate": 4.767737462445112e-05, + "loss": 0.0004, + "step": 1206 + }, + { + "epoch": 0.28, + "learning_rate": 4.7675448732763275e-05, + "loss": 0.0301, + "step": 1207 + }, + { + "epoch": 0.28, + "learning_rate": 4.7673522841075416e-05, + "loss": 0.0, + "step": 1208 + }, + { + "epoch": 0.28, + "learning_rate": 4.767159694938757e-05, + "loss": 0.0005, + "step": 1209 + }, + { + "epoch": 0.28, + "learning_rate": 4.766967105769972e-05, + "loss": 0.0, + "step": 1210 + }, + { + "epoch": 0.28, + "learning_rate": 4.7667745166011865e-05, + "loss": 0.0, + "step": 1211 + }, + { + "epoch": 0.28, + "learning_rate": 4.766581927432401e-05, + "loss": 0.0559, + "step": 1212 + }, + { + "epoch": 0.28, + "learning_rate": 4.766389338263617e-05, + "loss": 0.0489, + "step": 1213 + }, + { + "epoch": 0.28, + "learning_rate": 4.7661967490948314e-05, + "loss": 0.0042, + "step": 1214 + }, + { + "epoch": 0.28, + "learning_rate": 4.7660041599260455e-05, + "loss": 0.0792, + "step": 1215 + }, + { + "epoch": 0.28, + "learning_rate": 4.765811570757261e-05, + "loss": 0.0277, + "step": 1216 + }, + { + "epoch": 0.28, + "learning_rate": 4.765618981588476e-05, + "loss": 0.1134, + "step": 1217 + }, + { + "epoch": 0.28, + "learning_rate": 4.7654263924196904e-05, + "loss": 0.0, + "step": 1218 + }, + { + "epoch": 0.28, + "learning_rate": 4.765233803250905e-05, + "loss": 0.0055, + "step": 1219 + }, + { + "epoch": 0.28, + "learning_rate": 4.7650412140821206e-05, + "loss": 0.0167, + "step": 1220 + }, + { + "epoch": 0.28, + "learning_rate": 4.7648486249133353e-05, + "loss": 0.0, + "step": 1221 + }, + { + "epoch": 0.28, + "learning_rate": 4.76465603574455e-05, + "loss": 0.0208, + "step": 1222 + }, + { + "epoch": 0.28, + "learning_rate": 4.764463446575765e-05, + "loss": 0.0012, + "step": 1223 + }, + { + "epoch": 0.28, + "learning_rate": 4.7642708574069796e-05, + "loss": 0.0011, + "step": 1224 + }, + { + "epoch": 0.28, + "learning_rate": 4.764078268238194e-05, + "loss": 0.0056, + "step": 1225 + }, + { + "epoch": 0.28, + "learning_rate": 4.763885679069409e-05, + "loss": 0.0077, + "step": 1226 + }, + { + "epoch": 0.28, + "learning_rate": 4.7636930899006245e-05, + "loss": 0.0032, + "step": 1227 + }, + { + "epoch": 0.28, + "learning_rate": 4.763500500731839e-05, + "loss": 0.0291, + "step": 1228 + }, + { + "epoch": 0.28, + "learning_rate": 4.763307911563054e-05, + "loss": 0.0001, + "step": 1229 + }, + { + "epoch": 0.28, + "learning_rate": 4.763115322394269e-05, + "loss": 0.0096, + "step": 1230 + }, + { + "epoch": 0.28, + "learning_rate": 4.762922733225484e-05, + "loss": 0.067, + "step": 1231 + }, + { + "epoch": 0.28, + "learning_rate": 4.762730144056698e-05, + "loss": 0.0, + "step": 1232 + }, + { + "epoch": 0.28, + "learning_rate": 4.762537554887913e-05, + "loss": 0.0147, + "step": 1233 + }, + { + "epoch": 0.29, + "learning_rate": 4.7623449657191284e-05, + "loss": 0.0102, + "step": 1234 + }, + { + "epoch": 0.29, + "learning_rate": 4.762152376550343e-05, + "loss": 0.0196, + "step": 1235 + }, + { + "epoch": 0.29, + "learning_rate": 4.761959787381558e-05, + "loss": 0.0, + "step": 1236 + }, + { + "epoch": 0.29, + "learning_rate": 4.761767198212773e-05, + "loss": 0.0007, + "step": 1237 + }, + { + "epoch": 0.29, + "learning_rate": 4.761574609043988e-05, + "loss": 0.0515, + "step": 1238 + }, + { + "epoch": 0.29, + "learning_rate": 4.761382019875202e-05, + "loss": 0.0056, + "step": 1239 + }, + { + "epoch": 0.29, + "learning_rate": 4.7611894307064176e-05, + "loss": 0.0008, + "step": 1240 + }, + { + "epoch": 0.29, + "learning_rate": 4.760996841537632e-05, + "loss": 0.0246, + "step": 1241 + }, + { + "epoch": 0.29, + "learning_rate": 4.760804252368847e-05, + "loss": 0.0012, + "step": 1242 + }, + { + "epoch": 0.29, + "learning_rate": 4.760611663200062e-05, + "loss": 0.0166, + "step": 1243 + }, + { + "epoch": 0.29, + "learning_rate": 4.7604190740312766e-05, + "loss": 0.0228, + "step": 1244 + }, + { + "epoch": 0.29, + "learning_rate": 4.760226484862492e-05, + "loss": 0.0, + "step": 1245 + }, + { + "epoch": 0.29, + "learning_rate": 4.760033895693706e-05, + "loss": 0.0283, + "step": 1246 + }, + { + "epoch": 0.29, + "learning_rate": 4.7598413065249215e-05, + "loss": 0.0045, + "step": 1247 + }, + { + "epoch": 0.29, + "learning_rate": 4.759648717356136e-05, + "loss": 0.0313, + "step": 1248 + }, + { + "epoch": 0.29, + "learning_rate": 4.759456128187351e-05, + "loss": 0.0472, + "step": 1249 + }, + { + "epoch": 0.29, + "learning_rate": 4.759263539018566e-05, + "loss": 0.0059, + "step": 1250 + }, + { + "epoch": 0.29, + "learning_rate": 4.7590709498497805e-05, + "loss": 0.0123, + "step": 1251 + }, + { + "epoch": 0.29, + "learning_rate": 4.758878360680995e-05, + "loss": 0.0011, + "step": 1252 + }, + { + "epoch": 0.29, + "learning_rate": 4.75868577151221e-05, + "loss": 0.0001, + "step": 1253 + }, + { + "epoch": 0.29, + "learning_rate": 4.7584931823434254e-05, + "loss": 0.0082, + "step": 1254 + }, + { + "epoch": 0.29, + "learning_rate": 4.75830059317464e-05, + "loss": 0.0098, + "step": 1255 + }, + { + "epoch": 0.29, + "learning_rate": 4.758108004005855e-05, + "loss": 0.0067, + "step": 1256 + }, + { + "epoch": 0.29, + "learning_rate": 4.7579154148370697e-05, + "loss": 0.0134, + "step": 1257 + }, + { + "epoch": 0.29, + "learning_rate": 4.757722825668285e-05, + "loss": 0.0, + "step": 1258 + }, + { + "epoch": 0.29, + "learning_rate": 4.757530236499499e-05, + "loss": 0.0009, + "step": 1259 + }, + { + "epoch": 0.29, + "learning_rate": 4.757337647330714e-05, + "loss": 0.0022, + "step": 1260 + }, + { + "epoch": 0.29, + "learning_rate": 4.757145058161929e-05, + "loss": 0.0, + "step": 1261 + }, + { + "epoch": 0.29, + "learning_rate": 4.756952468993144e-05, + "loss": 0.0129, + "step": 1262 + }, + { + "epoch": 0.29, + "learning_rate": 4.756759879824359e-05, + "loss": 0.0, + "step": 1263 + }, + { + "epoch": 0.29, + "learning_rate": 4.7565672906555736e-05, + "loss": 0.0034, + "step": 1264 + }, + { + "epoch": 0.29, + "learning_rate": 4.756374701486789e-05, + "loss": 0.0164, + "step": 1265 + }, + { + "epoch": 0.29, + "learning_rate": 4.756182112318003e-05, + "loss": 0.0075, + "step": 1266 + }, + { + "epoch": 0.29, + "learning_rate": 4.7559895231492185e-05, + "loss": 0.0, + "step": 1267 + }, + { + "epoch": 0.29, + "learning_rate": 4.755796933980433e-05, + "loss": 0.0001, + "step": 1268 + }, + { + "epoch": 0.29, + "learning_rate": 4.755604344811648e-05, + "loss": 0.0, + "step": 1269 + }, + { + "epoch": 0.29, + "learning_rate": 4.755411755642863e-05, + "loss": 0.0009, + "step": 1270 + }, + { + "epoch": 0.29, + "learning_rate": 4.7552191664740775e-05, + "loss": 0.0004, + "step": 1271 + }, + { + "epoch": 0.29, + "learning_rate": 4.755026577305293e-05, + "loss": 0.0, + "step": 1272 + }, + { + "epoch": 0.29, + "learning_rate": 4.754833988136507e-05, + "loss": 0.0706, + "step": 1273 + }, + { + "epoch": 0.29, + "learning_rate": 4.7546413989677224e-05, + "loss": 0.0081, + "step": 1274 + }, + { + "epoch": 0.29, + "learning_rate": 4.754448809798937e-05, + "loss": 0.0013, + "step": 1275 + }, + { + "epoch": 0.29, + "learning_rate": 4.754256220630152e-05, + "loss": 0.0026, + "step": 1276 + }, + { + "epoch": 0.3, + "learning_rate": 4.7540636314613666e-05, + "loss": 0.0135, + "step": 1277 + }, + { + "epoch": 0.3, + "learning_rate": 4.753871042292582e-05, + "loss": 0.0326, + "step": 1278 + }, + { + "epoch": 0.3, + "learning_rate": 4.753678453123797e-05, + "loss": 0.0142, + "step": 1279 + }, + { + "epoch": 0.3, + "learning_rate": 4.753485863955011e-05, + "loss": 0.0, + "step": 1280 + }, + { + "epoch": 0.3, + "learning_rate": 4.753293274786226e-05, + "loss": 0.0059, + "step": 1281 + }, + { + "epoch": 0.3, + "learning_rate": 4.753100685617441e-05, + "loss": 0.0191, + "step": 1282 + }, + { + "epoch": 0.3, + "learning_rate": 4.752908096448656e-05, + "loss": 0.0, + "step": 1283 + }, + { + "epoch": 0.3, + "learning_rate": 4.7527155072798706e-05, + "loss": 0.0002, + "step": 1284 + }, + { + "epoch": 0.3, + "learning_rate": 4.752522918111086e-05, + "loss": 0.0101, + "step": 1285 + }, + { + "epoch": 0.3, + "learning_rate": 4.752330328942301e-05, + "loss": 0.1116, + "step": 1286 + }, + { + "epoch": 0.3, + "learning_rate": 4.7521377397735155e-05, + "loss": 0.0001, + "step": 1287 + }, + { + "epoch": 0.3, + "learning_rate": 4.75194515060473e-05, + "loss": 0.003, + "step": 1288 + }, + { + "epoch": 0.3, + "learning_rate": 4.751752561435945e-05, + "loss": 0.0, + "step": 1289 + }, + { + "epoch": 0.3, + "learning_rate": 4.75155997226716e-05, + "loss": 0.0002, + "step": 1290 + }, + { + "epoch": 0.3, + "learning_rate": 4.7513673830983745e-05, + "loss": 0.0185, + "step": 1291 + }, + { + "epoch": 0.3, + "learning_rate": 4.75117479392959e-05, + "loss": 0.0126, + "step": 1292 + }, + { + "epoch": 0.3, + "learning_rate": 4.7509822047608046e-05, + "loss": 0.0, + "step": 1293 + }, + { + "epoch": 0.3, + "learning_rate": 4.7507896155920194e-05, + "loss": 0.0, + "step": 1294 + }, + { + "epoch": 0.3, + "learning_rate": 4.750597026423234e-05, + "loss": 0.0031, + "step": 1295 + }, + { + "epoch": 0.3, + "learning_rate": 4.7504044372544496e-05, + "loss": 0.0015, + "step": 1296 + }, + { + "epoch": 0.3, + "learning_rate": 4.7502118480856636e-05, + "loss": 0.0005, + "step": 1297 + }, + { + "epoch": 0.3, + "learning_rate": 4.7500192589168784e-05, + "loss": 0.0009, + "step": 1298 + }, + { + "epoch": 0.3, + "learning_rate": 4.749826669748094e-05, + "loss": 0.0308, + "step": 1299 + }, + { + "epoch": 0.3, + "learning_rate": 4.7496340805793086e-05, + "loss": 0.0002, + "step": 1300 + }, + { + "epoch": 0.3, + "learning_rate": 4.749441491410523e-05, + "loss": 0.0, + "step": 1301 + }, + { + "epoch": 0.3, + "learning_rate": 4.749248902241738e-05, + "loss": 0.0013, + "step": 1302 + }, + { + "epoch": 0.3, + "learning_rate": 4.7490563130729535e-05, + "loss": 0.0035, + "step": 1303 + }, + { + "epoch": 0.3, + "learning_rate": 4.7488637239041675e-05, + "loss": 0.0006, + "step": 1304 + }, + { + "epoch": 0.3, + "learning_rate": 4.748671134735383e-05, + "loss": 0.0003, + "step": 1305 + }, + { + "epoch": 0.3, + "learning_rate": 4.748478545566598e-05, + "loss": 0.0296, + "step": 1306 + }, + { + "epoch": 0.3, + "learning_rate": 4.7482859563978125e-05, + "loss": 0.0091, + "step": 1307 + }, + { + "epoch": 0.3, + "learning_rate": 4.748093367229027e-05, + "loss": 0.0, + "step": 1308 + }, + { + "epoch": 0.3, + "learning_rate": 4.747900778060242e-05, + "loss": 0.0192, + "step": 1309 + }, + { + "epoch": 0.3, + "learning_rate": 4.7477081888914574e-05, + "loss": 0.0085, + "step": 1310 + }, + { + "epoch": 0.3, + "learning_rate": 4.7475155997226715e-05, + "loss": 0.0013, + "step": 1311 + }, + { + "epoch": 0.3, + "learning_rate": 4.747323010553887e-05, + "loss": 0.0084, + "step": 1312 + }, + { + "epoch": 0.3, + "learning_rate": 4.7471304213851016e-05, + "loss": 0.0071, + "step": 1313 + }, + { + "epoch": 0.3, + "learning_rate": 4.7469378322163164e-05, + "loss": 0.0354, + "step": 1314 + }, + { + "epoch": 0.3, + "learning_rate": 4.746745243047531e-05, + "loss": 0.0091, + "step": 1315 + }, + { + "epoch": 0.3, + "learning_rate": 4.746552653878746e-05, + "loss": 0.0, + "step": 1316 + }, + { + "epoch": 0.3, + "learning_rate": 4.746360064709961e-05, + "loss": 0.0381, + "step": 1317 + }, + { + "epoch": 0.3, + "learning_rate": 4.7461674755411754e-05, + "loss": 0.0, + "step": 1318 + }, + { + "epoch": 0.3, + "learning_rate": 4.745974886372391e-05, + "loss": 0.0026, + "step": 1319 + }, + { + "epoch": 0.31, + "learning_rate": 4.7457822972036055e-05, + "loss": 0.0008, + "step": 1320 + }, + { + "epoch": 0.31, + "learning_rate": 4.74558970803482e-05, + "loss": 0.0198, + "step": 1321 + }, + { + "epoch": 0.31, + "learning_rate": 4.745397118866035e-05, + "loss": 0.0, + "step": 1322 + }, + { + "epoch": 0.31, + "learning_rate": 4.7452045296972505e-05, + "loss": 0.0167, + "step": 1323 + }, + { + "epoch": 0.31, + "learning_rate": 4.745011940528465e-05, + "loss": 0.0212, + "step": 1324 + }, + { + "epoch": 0.31, + "learning_rate": 4.74481935135968e-05, + "loss": 0.0283, + "step": 1325 + }, + { + "epoch": 0.31, + "learning_rate": 4.744626762190895e-05, + "loss": 0.0008, + "step": 1326 + }, + { + "epoch": 0.31, + "learning_rate": 4.7444341730221095e-05, + "loss": 0.0, + "step": 1327 + }, + { + "epoch": 0.31, + "learning_rate": 4.744241583853324e-05, + "loss": 0.0005, + "step": 1328 + }, + { + "epoch": 0.31, + "learning_rate": 4.744048994684539e-05, + "loss": 0.0007, + "step": 1329 + }, + { + "epoch": 0.31, + "learning_rate": 4.7438564055157544e-05, + "loss": 0.0, + "step": 1330 + }, + { + "epoch": 0.31, + "learning_rate": 4.7436638163469684e-05, + "loss": 0.0011, + "step": 1331 + }, + { + "epoch": 0.31, + "learning_rate": 4.743471227178184e-05, + "loss": 0.0, + "step": 1332 + }, + { + "epoch": 0.31, + "learning_rate": 4.7432786380093986e-05, + "loss": 0.0232, + "step": 1333 + }, + { + "epoch": 0.31, + "learning_rate": 4.7430860488406134e-05, + "loss": 0.0054, + "step": 1334 + }, + { + "epoch": 0.31, + "learning_rate": 4.742893459671828e-05, + "loss": 0.0001, + "step": 1335 + }, + { + "epoch": 0.31, + "learning_rate": 4.742700870503043e-05, + "loss": 0.0, + "step": 1336 + }, + { + "epoch": 0.31, + "learning_rate": 4.742508281334258e-05, + "loss": 0.0342, + "step": 1337 + }, + { + "epoch": 0.31, + "learning_rate": 4.7423156921654724e-05, + "loss": 0.0008, + "step": 1338 + }, + { + "epoch": 0.31, + "learning_rate": 4.742123102996688e-05, + "loss": 0.0003, + "step": 1339 + }, + { + "epoch": 0.31, + "learning_rate": 4.7419305138279025e-05, + "loss": 0.0125, + "step": 1340 + }, + { + "epoch": 0.31, + "learning_rate": 4.741737924659117e-05, + "loss": 0.0, + "step": 1341 + }, + { + "epoch": 0.31, + "learning_rate": 4.741545335490332e-05, + "loss": 0.0001, + "step": 1342 + }, + { + "epoch": 0.31, + "learning_rate": 4.7413527463215475e-05, + "loss": 0.0095, + "step": 1343 + }, + { + "epoch": 0.31, + "learning_rate": 4.741160157152762e-05, + "loss": 0.0219, + "step": 1344 + }, + { + "epoch": 0.31, + "learning_rate": 4.740967567983976e-05, + "loss": 0.0001, + "step": 1345 + }, + { + "epoch": 0.31, + "learning_rate": 4.740774978815192e-05, + "loss": 0.0091, + "step": 1346 + }, + { + "epoch": 0.31, + "learning_rate": 4.7405823896464064e-05, + "loss": 0.0049, + "step": 1347 + }, + { + "epoch": 0.31, + "learning_rate": 4.740389800477621e-05, + "loss": 0.0092, + "step": 1348 + }, + { + "epoch": 0.31, + "learning_rate": 4.740197211308836e-05, + "loss": 0.0131, + "step": 1349 + }, + { + "epoch": 0.31, + "learning_rate": 4.7400046221400514e-05, + "loss": 0.0277, + "step": 1350 + }, + { + "epoch": 0.31, + "learning_rate": 4.739812032971266e-05, + "loss": 0.0256, + "step": 1351 + }, + { + "epoch": 0.31, + "learning_rate": 4.739619443802481e-05, + "loss": 0.0062, + "step": 1352 + }, + { + "epoch": 0.31, + "learning_rate": 4.7394268546336956e-05, + "loss": 0.0, + "step": 1353 + }, + { + "epoch": 0.31, + "learning_rate": 4.7392342654649104e-05, + "loss": 0.0129, + "step": 1354 + }, + { + "epoch": 0.31, + "learning_rate": 4.739041676296125e-05, + "loss": 0.0029, + "step": 1355 + }, + { + "epoch": 0.31, + "learning_rate": 4.73884908712734e-05, + "loss": 0.0, + "step": 1356 + }, + { + "epoch": 0.31, + "learning_rate": 4.738656497958555e-05, + "loss": 0.0457, + "step": 1357 + }, + { + "epoch": 0.31, + "learning_rate": 4.73846390878977e-05, + "loss": 0.0005, + "step": 1358 + }, + { + "epoch": 0.31, + "learning_rate": 4.738271319620985e-05, + "loss": 0.0005, + "step": 1359 + }, + { + "epoch": 0.31, + "learning_rate": 4.7380787304521995e-05, + "loss": 0.0, + "step": 1360 + }, + { + "epoch": 0.31, + "learning_rate": 4.737886141283415e-05, + "loss": 0.0001, + "step": 1361 + }, + { + "epoch": 0.31, + "learning_rate": 4.737693552114629e-05, + "loss": 0.0009, + "step": 1362 + }, + { + "epoch": 0.31, + "learning_rate": 4.737500962945844e-05, + "loss": 0.0, + "step": 1363 + }, + { + "epoch": 0.32, + "learning_rate": 4.737308373777059e-05, + "loss": 0.0084, + "step": 1364 + }, + { + "epoch": 0.32, + "learning_rate": 4.737115784608274e-05, + "loss": 0.0011, + "step": 1365 + }, + { + "epoch": 0.32, + "learning_rate": 4.736923195439489e-05, + "loss": 0.0488, + "step": 1366 + }, + { + "epoch": 0.32, + "learning_rate": 4.7367306062707034e-05, + "loss": 0.0018, + "step": 1367 + }, + { + "epoch": 0.32, + "learning_rate": 4.736538017101919e-05, + "loss": 0.007, + "step": 1368 + }, + { + "epoch": 0.32, + "learning_rate": 4.736345427933133e-05, + "loss": 0.0325, + "step": 1369 + }, + { + "epoch": 0.32, + "learning_rate": 4.7361528387643484e-05, + "loss": 0.0017, + "step": 1370 + }, + { + "epoch": 0.32, + "learning_rate": 4.735960249595563e-05, + "loss": 0.0018, + "step": 1371 + }, + { + "epoch": 0.32, + "learning_rate": 4.735767660426778e-05, + "loss": 0.0004, + "step": 1372 + }, + { + "epoch": 0.32, + "learning_rate": 4.7355750712579926e-05, + "loss": 0.0, + "step": 1373 + }, + { + "epoch": 0.32, + "learning_rate": 4.7353824820892073e-05, + "loss": 0.0175, + "step": 1374 + }, + { + "epoch": 0.32, + "learning_rate": 4.735189892920423e-05, + "loss": 0.0019, + "step": 1375 + }, + { + "epoch": 0.32, + "learning_rate": 4.734997303751637e-05, + "loss": 0.0086, + "step": 1376 + }, + { + "epoch": 0.32, + "learning_rate": 4.734804714582852e-05, + "loss": 0.0142, + "step": 1377 + }, + { + "epoch": 0.32, + "learning_rate": 4.734612125414067e-05, + "loss": 0.0, + "step": 1378 + }, + { + "epoch": 0.32, + "learning_rate": 4.734419536245282e-05, + "loss": 0.0002, + "step": 1379 + }, + { + "epoch": 0.32, + "learning_rate": 4.7342269470764965e-05, + "loss": 0.0, + "step": 1380 + }, + { + "epoch": 0.32, + "learning_rate": 4.734034357907712e-05, + "loss": 0.0352, + "step": 1381 + }, + { + "epoch": 0.32, + "learning_rate": 4.733841768738927e-05, + "loss": 0.0093, + "step": 1382 + }, + { + "epoch": 0.32, + "learning_rate": 4.733649179570141e-05, + "loss": 0.0125, + "step": 1383 + }, + { + "epoch": 0.32, + "learning_rate": 4.733456590401356e-05, + "loss": 0.0093, + "step": 1384 + }, + { + "epoch": 0.32, + "learning_rate": 4.733264001232571e-05, + "loss": 0.0009, + "step": 1385 + }, + { + "epoch": 0.32, + "learning_rate": 4.733071412063786e-05, + "loss": 0.0095, + "step": 1386 + }, + { + "epoch": 0.32, + "learning_rate": 4.7328788228950004e-05, + "loss": 0.0001, + "step": 1387 + }, + { + "epoch": 0.32, + "learning_rate": 4.732686233726216e-05, + "loss": 0.0234, + "step": 1388 + }, + { + "epoch": 0.32, + "learning_rate": 4.7324936445574306e-05, + "loss": 0.0218, + "step": 1389 + }, + { + "epoch": 0.32, + "learning_rate": 4.7323010553886453e-05, + "loss": 0.0013, + "step": 1390 + }, + { + "epoch": 0.32, + "learning_rate": 4.73210846621986e-05, + "loss": 0.0001, + "step": 1391 + }, + { + "epoch": 0.32, + "learning_rate": 4.731915877051075e-05, + "loss": 0.01, + "step": 1392 + }, + { + "epoch": 0.32, + "learning_rate": 4.7317232878822896e-05, + "loss": 0.0473, + "step": 1393 + }, + { + "epoch": 0.32, + "learning_rate": 4.731530698713504e-05, + "loss": 0.0118, + "step": 1394 + }, + { + "epoch": 0.32, + "learning_rate": 4.73133810954472e-05, + "loss": 0.0008, + "step": 1395 + }, + { + "epoch": 0.32, + "learning_rate": 4.7311455203759345e-05, + "loss": 0.011, + "step": 1396 + }, + { + "epoch": 0.32, + "learning_rate": 4.730952931207149e-05, + "loss": 0.0, + "step": 1397 + }, + { + "epoch": 0.32, + "learning_rate": 4.730760342038364e-05, + "loss": 0.0, + "step": 1398 + }, + { + "epoch": 0.32, + "learning_rate": 4.7305677528695794e-05, + "loss": 0.0001, + "step": 1399 + }, + { + "epoch": 0.32, + "learning_rate": 4.7303751637007935e-05, + "loss": 0.0243, + "step": 1400 + }, + { + "epoch": 0.32, + "learning_rate": 4.730182574532008e-05, + "loss": 0.0143, + "step": 1401 + }, + { + "epoch": 0.32, + "learning_rate": 4.729989985363224e-05, + "loss": 0.0003, + "step": 1402 + }, + { + "epoch": 0.32, + "learning_rate": 4.729797396194438e-05, + "loss": 0.0174, + "step": 1403 + }, + { + "epoch": 0.32, + "learning_rate": 4.729604807025653e-05, + "loss": 0.0006, + "step": 1404 + }, + { + "epoch": 0.32, + "learning_rate": 4.729412217856868e-05, + "loss": 0.0198, + "step": 1405 + }, + { + "epoch": 0.32, + "learning_rate": 4.7292196286880827e-05, + "loss": 0.0021, + "step": 1406 + }, + { + "epoch": 0.33, + "learning_rate": 4.7290270395192974e-05, + "loss": 0.0043, + "step": 1407 + }, + { + "epoch": 0.33, + "learning_rate": 4.728834450350513e-05, + "loss": 0.0001, + "step": 1408 + }, + { + "epoch": 0.33, + "learning_rate": 4.7286418611817276e-05, + "loss": 0.0005, + "step": 1409 + }, + { + "epoch": 0.33, + "learning_rate": 4.7284492720129417e-05, + "loss": 0.0, + "step": 1410 + }, + { + "epoch": 0.33, + "learning_rate": 4.728256682844157e-05, + "loss": 0.0004, + "step": 1411 + }, + { + "epoch": 0.33, + "learning_rate": 4.728064093675372e-05, + "loss": 0.0008, + "step": 1412 + }, + { + "epoch": 0.33, + "learning_rate": 4.7278715045065866e-05, + "loss": 0.0002, + "step": 1413 + }, + { + "epoch": 0.33, + "learning_rate": 4.727678915337801e-05, + "loss": 0.0019, + "step": 1414 + }, + { + "epoch": 0.33, + "learning_rate": 4.727486326169017e-05, + "loss": 0.0053, + "step": 1415 + }, + { + "epoch": 0.33, + "learning_rate": 4.7272937370002315e-05, + "loss": 0.0082, + "step": 1416 + }, + { + "epoch": 0.33, + "learning_rate": 4.727101147831446e-05, + "loss": 0.0001, + "step": 1417 + }, + { + "epoch": 0.33, + "learning_rate": 4.726908558662661e-05, + "loss": 0.0301, + "step": 1418 + }, + { + "epoch": 0.33, + "learning_rate": 4.726715969493876e-05, + "loss": 0.0177, + "step": 1419 + }, + { + "epoch": 0.33, + "learning_rate": 4.7265233803250905e-05, + "loss": 0.0, + "step": 1420 + }, + { + "epoch": 0.33, + "learning_rate": 4.726330791156305e-05, + "loss": 0.0, + "step": 1421 + }, + { + "epoch": 0.33, + "learning_rate": 4.7261382019875207e-05, + "loss": 0.0182, + "step": 1422 + }, + { + "epoch": 0.33, + "learning_rate": 4.7259456128187354e-05, + "loss": 0.0, + "step": 1423 + }, + { + "epoch": 0.33, + "learning_rate": 4.72575302364995e-05, + "loss": 0.0209, + "step": 1424 + }, + { + "epoch": 0.33, + "learning_rate": 4.725560434481165e-05, + "loss": 0.0004, + "step": 1425 + }, + { + "epoch": 0.33, + "learning_rate": 4.72536784531238e-05, + "loss": 0.0484, + "step": 1426 + }, + { + "epoch": 0.33, + "learning_rate": 4.7251752561435944e-05, + "loss": 0.0093, + "step": 1427 + }, + { + "epoch": 0.33, + "learning_rate": 4.724982666974809e-05, + "loss": 0.0028, + "step": 1428 + }, + { + "epoch": 0.33, + "learning_rate": 4.7247900778060246e-05, + "loss": 0.0088, + "step": 1429 + }, + { + "epoch": 0.33, + "learning_rate": 4.724597488637239e-05, + "loss": 0.0, + "step": 1430 + }, + { + "epoch": 0.33, + "learning_rate": 4.724404899468454e-05, + "loss": 0.0215, + "step": 1431 + }, + { + "epoch": 0.33, + "learning_rate": 4.724212310299669e-05, + "loss": 0.0008, + "step": 1432 + }, + { + "epoch": 0.33, + "learning_rate": 4.724019721130884e-05, + "loss": 0.0, + "step": 1433 + }, + { + "epoch": 0.33, + "learning_rate": 4.723827131962098e-05, + "loss": 0.0016, + "step": 1434 + }, + { + "epoch": 0.33, + "learning_rate": 4.723634542793314e-05, + "loss": 0.0001, + "step": 1435 + }, + { + "epoch": 0.33, + "learning_rate": 4.7234419536245285e-05, + "loss": 0.0025, + "step": 1436 + }, + { + "epoch": 0.33, + "learning_rate": 4.723249364455743e-05, + "loss": 0.0, + "step": 1437 + }, + { + "epoch": 0.33, + "learning_rate": 4.723056775286958e-05, + "loss": 0.0063, + "step": 1438 + }, + { + "epoch": 0.33, + "learning_rate": 4.722864186118173e-05, + "loss": 0.0001, + "step": 1439 + }, + { + "epoch": 0.33, + "learning_rate": 4.722671596949388e-05, + "loss": 0.0207, + "step": 1440 + }, + { + "epoch": 0.33, + "learning_rate": 4.722479007780602e-05, + "loss": 0.0069, + "step": 1441 + }, + { + "epoch": 0.33, + "learning_rate": 4.7222864186118176e-05, + "loss": 0.0789, + "step": 1442 + }, + { + "epoch": 0.33, + "learning_rate": 4.7220938294430324e-05, + "loss": 0.0, + "step": 1443 + }, + { + "epoch": 0.33, + "learning_rate": 4.721901240274247e-05, + "loss": 0.012, + "step": 1444 + }, + { + "epoch": 0.33, + "learning_rate": 4.721708651105462e-05, + "loss": 0.0118, + "step": 1445 + }, + { + "epoch": 0.33, + "learning_rate": 4.721516061936677e-05, + "loss": 0.0, + "step": 1446 + }, + { + "epoch": 0.33, + "learning_rate": 4.721323472767892e-05, + "loss": 0.0, + "step": 1447 + }, + { + "epoch": 0.33, + "learning_rate": 4.721130883599106e-05, + "loss": 0.0674, + "step": 1448 + }, + { + "epoch": 0.33, + "learning_rate": 4.7209382944303216e-05, + "loss": 0.0, + "step": 1449 + }, + { + "epoch": 0.34, + "learning_rate": 4.720745705261536e-05, + "loss": 0.0, + "step": 1450 + }, + { + "epoch": 0.34, + "learning_rate": 4.720553116092751e-05, + "loss": 0.0001, + "step": 1451 + }, + { + "epoch": 0.34, + "learning_rate": 4.720360526923966e-05, + "loss": 0.1087, + "step": 1452 + }, + { + "epoch": 0.34, + "learning_rate": 4.720167937755181e-05, + "loss": 0.0053, + "step": 1453 + }, + { + "epoch": 0.34, + "learning_rate": 4.719975348586396e-05, + "loss": 0.0018, + "step": 1454 + }, + { + "epoch": 0.34, + "learning_rate": 4.719782759417611e-05, + "loss": 0.0036, + "step": 1455 + }, + { + "epoch": 0.34, + "learning_rate": 4.7195901702488255e-05, + "loss": 0.0058, + "step": 1456 + }, + { + "epoch": 0.34, + "learning_rate": 4.71939758108004e-05, + "loss": 0.0, + "step": 1457 + }, + { + "epoch": 0.34, + "learning_rate": 4.719204991911255e-05, + "loss": 0.0, + "step": 1458 + }, + { + "epoch": 0.34, + "learning_rate": 4.71901240274247e-05, + "loss": 0.0031, + "step": 1459 + }, + { + "epoch": 0.34, + "learning_rate": 4.718819813573685e-05, + "loss": 0.0024, + "step": 1460 + }, + { + "epoch": 0.34, + "learning_rate": 4.7186272244049e-05, + "loss": 0.0, + "step": 1461 + }, + { + "epoch": 0.34, + "learning_rate": 4.7184346352361146e-05, + "loss": 0.0003, + "step": 1462 + }, + { + "epoch": 0.34, + "learning_rate": 4.7182420460673294e-05, + "loss": 0.0, + "step": 1463 + }, + { + "epoch": 0.34, + "learning_rate": 4.718049456898545e-05, + "loss": 0.0125, + "step": 1464 + }, + { + "epoch": 0.34, + "learning_rate": 4.717856867729759e-05, + "loss": 0.0083, + "step": 1465 + }, + { + "epoch": 0.34, + "learning_rate": 4.7176642785609736e-05, + "loss": 0.016, + "step": 1466 + }, + { + "epoch": 0.34, + "learning_rate": 4.717471689392189e-05, + "loss": 0.0, + "step": 1467 + }, + { + "epoch": 0.34, + "learning_rate": 4.717279100223404e-05, + "loss": 0.0067, + "step": 1468 + }, + { + "epoch": 0.34, + "learning_rate": 4.7170865110546185e-05, + "loss": 0.0, + "step": 1469 + }, + { + "epoch": 0.34, + "learning_rate": 4.716893921885833e-05, + "loss": 0.0014, + "step": 1470 + }, + { + "epoch": 0.34, + "learning_rate": 4.716701332717049e-05, + "loss": 0.0253, + "step": 1471 + }, + { + "epoch": 0.34, + "learning_rate": 4.716508743548263e-05, + "loss": 0.0006, + "step": 1472 + }, + { + "epoch": 0.34, + "learning_rate": 4.716316154379478e-05, + "loss": 0.0004, + "step": 1473 + }, + { + "epoch": 0.34, + "learning_rate": 4.716123565210693e-05, + "loss": 0.083, + "step": 1474 + }, + { + "epoch": 0.34, + "learning_rate": 4.715930976041907e-05, + "loss": 0.0, + "step": 1475 + }, + { + "epoch": 0.34, + "learning_rate": 4.7157383868731225e-05, + "loss": 0.0023, + "step": 1476 + }, + { + "epoch": 0.34, + "learning_rate": 4.715545797704337e-05, + "loss": 0.0001, + "step": 1477 + }, + { + "epoch": 0.34, + "learning_rate": 4.715353208535552e-05, + "loss": 0.0002, + "step": 1478 + }, + { + "epoch": 0.34, + "learning_rate": 4.715160619366767e-05, + "loss": 0.0001, + "step": 1479 + }, + { + "epoch": 0.34, + "learning_rate": 4.714968030197982e-05, + "loss": 0.0007, + "step": 1480 + }, + { + "epoch": 0.34, + "learning_rate": 4.714775441029197e-05, + "loss": 0.0003, + "step": 1481 + }, + { + "epoch": 0.34, + "learning_rate": 4.7145828518604116e-05, + "loss": 0.0357, + "step": 1482 + }, + { + "epoch": 0.34, + "learning_rate": 4.7143902626916264e-05, + "loss": 0.0535, + "step": 1483 + }, + { + "epoch": 0.34, + "learning_rate": 4.714197673522841e-05, + "loss": 0.0334, + "step": 1484 + }, + { + "epoch": 0.34, + "learning_rate": 4.714005084354056e-05, + "loss": 0.0074, + "step": 1485 + }, + { + "epoch": 0.34, + "learning_rate": 4.7138124951852706e-05, + "loss": 0.0007, + "step": 1486 + }, + { + "epoch": 0.34, + "learning_rate": 4.713619906016486e-05, + "loss": 0.008, + "step": 1487 + }, + { + "epoch": 0.34, + "learning_rate": 4.713427316847701e-05, + "loss": 0.012, + "step": 1488 + }, + { + "epoch": 0.34, + "learning_rate": 4.7132347276789155e-05, + "loss": 0.0028, + "step": 1489 + }, + { + "epoch": 0.34, + "learning_rate": 4.71304213851013e-05, + "loss": 0.0406, + "step": 1490 + }, + { + "epoch": 0.34, + "learning_rate": 4.712849549341346e-05, + "loss": 0.021, + "step": 1491 + }, + { + "epoch": 0.34, + "learning_rate": 4.71265696017256e-05, + "loss": 0.0, + "step": 1492 + }, + { + "epoch": 0.35, + "learning_rate": 4.712464371003775e-05, + "loss": 0.0001, + "step": 1493 + }, + { + "epoch": 0.35, + "learning_rate": 4.71227178183499e-05, + "loss": 0.0195, + "step": 1494 + }, + { + "epoch": 0.35, + "learning_rate": 4.712079192666205e-05, + "loss": 0.0, + "step": 1495 + }, + { + "epoch": 0.35, + "learning_rate": 4.7118866034974194e-05, + "loss": 0.0001, + "step": 1496 + }, + { + "epoch": 0.35, + "learning_rate": 4.711694014328634e-05, + "loss": 0.0, + "step": 1497 + }, + { + "epoch": 0.35, + "learning_rate": 4.7115014251598496e-05, + "loss": 0.0, + "step": 1498 + }, + { + "epoch": 0.35, + "learning_rate": 4.711308835991064e-05, + "loss": 0.001, + "step": 1499 + }, + { + "epoch": 0.35, + "learning_rate": 4.711116246822279e-05, + "loss": 0.0089, + "step": 1500 + }, + { + "epoch": 0.35, + "learning_rate": 4.710923657653494e-05, + "loss": 0.0039, + "step": 1501 + }, + { + "epoch": 0.35, + "learning_rate": 4.7107310684847086e-05, + "loss": 0.0129, + "step": 1502 + }, + { + "epoch": 0.35, + "learning_rate": 4.7105384793159234e-05, + "loss": 0.0002, + "step": 1503 + }, + { + "epoch": 0.35, + "learning_rate": 4.710345890147138e-05, + "loss": 0.0, + "step": 1504 + }, + { + "epoch": 0.35, + "learning_rate": 4.7101533009783535e-05, + "loss": 0.0017, + "step": 1505 + }, + { + "epoch": 0.35, + "learning_rate": 4.7099607118095676e-05, + "loss": 0.0577, + "step": 1506 + }, + { + "epoch": 0.35, + "learning_rate": 4.709768122640783e-05, + "loss": 0.0043, + "step": 1507 + }, + { + "epoch": 0.35, + "learning_rate": 4.709575533471998e-05, + "loss": 0.0006, + "step": 1508 + }, + { + "epoch": 0.35, + "learning_rate": 4.7093829443032125e-05, + "loss": 0.0339, + "step": 1509 + }, + { + "epoch": 0.35, + "learning_rate": 4.709190355134427e-05, + "loss": 0.0084, + "step": 1510 + }, + { + "epoch": 0.35, + "learning_rate": 4.708997765965643e-05, + "loss": 0.0005, + "step": 1511 + }, + { + "epoch": 0.35, + "learning_rate": 4.7088051767968574e-05, + "loss": 0.0468, + "step": 1512 + }, + { + "epoch": 0.35, + "learning_rate": 4.7086125876280715e-05, + "loss": 0.0, + "step": 1513 + }, + { + "epoch": 0.35, + "learning_rate": 4.708419998459287e-05, + "loss": 0.0078, + "step": 1514 + }, + { + "epoch": 0.35, + "learning_rate": 4.708227409290502e-05, + "loss": 0.0794, + "step": 1515 + }, + { + "epoch": 0.35, + "learning_rate": 4.7080348201217164e-05, + "loss": 0.0001, + "step": 1516 + }, + { + "epoch": 0.35, + "learning_rate": 4.707842230952931e-05, + "loss": 0.0001, + "step": 1517 + }, + { + "epoch": 0.35, + "learning_rate": 4.7076496417841466e-05, + "loss": 0.0008, + "step": 1518 + }, + { + "epoch": 0.35, + "learning_rate": 4.7074570526153614e-05, + "loss": 0.0681, + "step": 1519 + }, + { + "epoch": 0.35, + "learning_rate": 4.707264463446576e-05, + "loss": 0.0058, + "step": 1520 + }, + { + "epoch": 0.35, + "learning_rate": 4.707071874277791e-05, + "loss": 0.0445, + "step": 1521 + }, + { + "epoch": 0.35, + "learning_rate": 4.7068792851090056e-05, + "loss": 0.0051, + "step": 1522 + }, + { + "epoch": 0.35, + "learning_rate": 4.7066866959402203e-05, + "loss": 0.0, + "step": 1523 + }, + { + "epoch": 0.35, + "learning_rate": 4.706494106771435e-05, + "loss": 0.0047, + "step": 1524 + }, + { + "epoch": 0.35, + "learning_rate": 4.7063015176026505e-05, + "loss": 0.0008, + "step": 1525 + }, + { + "epoch": 0.35, + "learning_rate": 4.706108928433865e-05, + "loss": 0.0001, + "step": 1526 + }, + { + "epoch": 0.35, + "learning_rate": 4.70591633926508e-05, + "loss": 0.0007, + "step": 1527 + }, + { + "epoch": 0.35, + "learning_rate": 4.705723750096295e-05, + "loss": 0.0003, + "step": 1528 + }, + { + "epoch": 0.35, + "learning_rate": 4.70553116092751e-05, + "loss": 0.0019, + "step": 1529 + }, + { + "epoch": 0.35, + "learning_rate": 4.705338571758724e-05, + "loss": 0.0198, + "step": 1530 + }, + { + "epoch": 0.35, + "learning_rate": 4.705145982589939e-05, + "loss": 0.0003, + "step": 1531 + }, + { + "epoch": 0.35, + "learning_rate": 4.7049533934211544e-05, + "loss": 0.0001, + "step": 1532 + }, + { + "epoch": 0.35, + "learning_rate": 4.704760804252369e-05, + "loss": 0.0, + "step": 1533 + }, + { + "epoch": 0.35, + "learning_rate": 4.704568215083584e-05, + "loss": 0.0037, + "step": 1534 + }, + { + "epoch": 0.35, + "learning_rate": 4.704375625914799e-05, + "loss": 0.0132, + "step": 1535 + }, + { + "epoch": 0.35, + "learning_rate": 4.704183036746014e-05, + "loss": 0.036, + "step": 1536 + }, + { + "epoch": 0.36, + "learning_rate": 4.703990447577228e-05, + "loss": 0.0, + "step": 1537 + }, + { + "epoch": 0.36, + "learning_rate": 4.7037978584084436e-05, + "loss": 0.067, + "step": 1538 + }, + { + "epoch": 0.36, + "learning_rate": 4.7036052692396583e-05, + "loss": 0.0053, + "step": 1539 + }, + { + "epoch": 0.36, + "learning_rate": 4.703412680070873e-05, + "loss": 0.0007, + "step": 1540 + }, + { + "epoch": 0.36, + "learning_rate": 4.703220090902088e-05, + "loss": 0.0432, + "step": 1541 + }, + { + "epoch": 0.36, + "learning_rate": 4.7030275017333026e-05, + "loss": 0.0, + "step": 1542 + }, + { + "epoch": 0.36, + "learning_rate": 4.702834912564518e-05, + "loss": 0.0634, + "step": 1543 + }, + { + "epoch": 0.36, + "learning_rate": 4.702642323395732e-05, + "loss": 0.0, + "step": 1544 + }, + { + "epoch": 0.36, + "learning_rate": 4.7024497342269475e-05, + "loss": 0.0, + "step": 1545 + }, + { + "epoch": 0.36, + "learning_rate": 4.702257145058162e-05, + "loss": 0.0017, + "step": 1546 + }, + { + "epoch": 0.36, + "learning_rate": 4.702064555889377e-05, + "loss": 0.0074, + "step": 1547 + }, + { + "epoch": 0.36, + "learning_rate": 4.701871966720592e-05, + "loss": 0.0214, + "step": 1548 + }, + { + "epoch": 0.36, + "learning_rate": 4.701679377551807e-05, + "loss": 0.0, + "step": 1549 + }, + { + "epoch": 0.36, + "learning_rate": 4.701486788383022e-05, + "loss": 0.0019, + "step": 1550 + }, + { + "epoch": 0.36, + "learning_rate": 4.701294199214236e-05, + "loss": 0.022, + "step": 1551 + }, + { + "epoch": 0.36, + "learning_rate": 4.7011016100454514e-05, + "loss": 0.0003, + "step": 1552 + }, + { + "epoch": 0.36, + "learning_rate": 4.700909020876666e-05, + "loss": 0.0001, + "step": 1553 + }, + { + "epoch": 0.36, + "learning_rate": 4.700716431707881e-05, + "loss": 0.0, + "step": 1554 + }, + { + "epoch": 0.36, + "learning_rate": 4.700523842539096e-05, + "loss": 0.0252, + "step": 1555 + }, + { + "epoch": 0.36, + "learning_rate": 4.700331253370311e-05, + "loss": 0.0004, + "step": 1556 + }, + { + "epoch": 0.36, + "learning_rate": 4.700138664201525e-05, + "loss": 0.074, + "step": 1557 + }, + { + "epoch": 0.36, + "learning_rate": 4.6999460750327406e-05, + "loss": 0.0116, + "step": 1558 + }, + { + "epoch": 0.36, + "learning_rate": 4.699753485863955e-05, + "loss": 0.0, + "step": 1559 + }, + { + "epoch": 0.36, + "learning_rate": 4.69956089669517e-05, + "loss": 0.0007, + "step": 1560 + }, + { + "epoch": 0.36, + "learning_rate": 4.699368307526385e-05, + "loss": 0.0, + "step": 1561 + }, + { + "epoch": 0.36, + "learning_rate": 4.6991757183575996e-05, + "loss": 0.0095, + "step": 1562 + }, + { + "epoch": 0.36, + "learning_rate": 4.698983129188815e-05, + "loss": 0.0147, + "step": 1563 + }, + { + "epoch": 0.36, + "learning_rate": 4.698790540020029e-05, + "loss": 0.0001, + "step": 1564 + }, + { + "epoch": 0.36, + "learning_rate": 4.6985979508512445e-05, + "loss": 0.0001, + "step": 1565 + }, + { + "epoch": 0.36, + "learning_rate": 4.698405361682459e-05, + "loss": 0.0, + "step": 1566 + }, + { + "epoch": 0.36, + "learning_rate": 4.698212772513674e-05, + "loss": 0.0, + "step": 1567 + }, + { + "epoch": 0.36, + "learning_rate": 4.698020183344889e-05, + "loss": 0.0643, + "step": 1568 + }, + { + "epoch": 0.36, + "learning_rate": 4.6978275941761035e-05, + "loss": 0.0002, + "step": 1569 + }, + { + "epoch": 0.36, + "learning_rate": 4.697635005007319e-05, + "loss": 0.0159, + "step": 1570 + }, + { + "epoch": 0.36, + "learning_rate": 4.697442415838533e-05, + "loss": 0.0018, + "step": 1571 + }, + { + "epoch": 0.36, + "learning_rate": 4.6972498266697484e-05, + "loss": 0.0366, + "step": 1572 + }, + { + "epoch": 0.36, + "learning_rate": 4.697057237500963e-05, + "loss": 0.0013, + "step": 1573 + }, + { + "epoch": 0.36, + "learning_rate": 4.696864648332178e-05, + "loss": 0.0167, + "step": 1574 + }, + { + "epoch": 0.36, + "learning_rate": 4.6966720591633927e-05, + "loss": 0.0001, + "step": 1575 + }, + { + "epoch": 0.36, + "learning_rate": 4.696479469994608e-05, + "loss": 0.0084, + "step": 1576 + }, + { + "epoch": 0.36, + "learning_rate": 4.696286880825823e-05, + "loss": 0.0038, + "step": 1577 + }, + { + "epoch": 0.36, + "learning_rate": 4.696094291657037e-05, + "loss": 0.0033, + "step": 1578 + }, + { + "epoch": 0.36, + "learning_rate": 4.695901702488252e-05, + "loss": 0.0048, + "step": 1579 + }, + { + "epoch": 0.37, + "learning_rate": 4.695709113319467e-05, + "loss": 0.0063, + "step": 1580 + }, + { + "epoch": 0.37, + "learning_rate": 4.695516524150682e-05, + "loss": 0.0374, + "step": 1581 + }, + { + "epoch": 0.37, + "learning_rate": 4.6953239349818966e-05, + "loss": 0.0005, + "step": 1582 + }, + { + "epoch": 0.37, + "learning_rate": 4.695131345813112e-05, + "loss": 0.0025, + "step": 1583 + }, + { + "epoch": 0.37, + "learning_rate": 4.694938756644327e-05, + "loss": 0.0288, + "step": 1584 + }, + { + "epoch": 0.37, + "learning_rate": 4.6947461674755415e-05, + "loss": 0.0062, + "step": 1585 + }, + { + "epoch": 0.37, + "learning_rate": 4.694553578306756e-05, + "loss": 0.0002, + "step": 1586 + }, + { + "epoch": 0.37, + "learning_rate": 4.694360989137971e-05, + "loss": 0.0105, + "step": 1587 + }, + { + "epoch": 0.37, + "learning_rate": 4.694168399969186e-05, + "loss": 0.0078, + "step": 1588 + }, + { + "epoch": 0.37, + "learning_rate": 4.6939758108004005e-05, + "loss": 0.0004, + "step": 1589 + }, + { + "epoch": 0.37, + "learning_rate": 4.693783221631616e-05, + "loss": 0.0046, + "step": 1590 + }, + { + "epoch": 0.37, + "learning_rate": 4.6935906324628306e-05, + "loss": 0.0045, + "step": 1591 + }, + { + "epoch": 0.37, + "learning_rate": 4.6933980432940454e-05, + "loss": 0.0175, + "step": 1592 + }, + { + "epoch": 0.37, + "learning_rate": 4.69320545412526e-05, + "loss": 0.0002, + "step": 1593 + }, + { + "epoch": 0.37, + "learning_rate": 4.6930128649564756e-05, + "loss": 0.0253, + "step": 1594 + }, + { + "epoch": 0.37, + "learning_rate": 4.6928202757876896e-05, + "loss": 0.0016, + "step": 1595 + }, + { + "epoch": 0.37, + "learning_rate": 4.6926276866189044e-05, + "loss": 0.0135, + "step": 1596 + }, + { + "epoch": 0.37, + "learning_rate": 4.69243509745012e-05, + "loss": 0.0309, + "step": 1597 + }, + { + "epoch": 0.37, + "learning_rate": 4.6922425082813346e-05, + "loss": 0.0, + "step": 1598 + }, + { + "epoch": 0.37, + "learning_rate": 4.692049919112549e-05, + "loss": 0.0024, + "step": 1599 + }, + { + "epoch": 0.37, + "learning_rate": 4.691857329943764e-05, + "loss": 0.0136, + "step": 1600 + }, + { + "epoch": 0.37, + "learning_rate": 4.6916647407749795e-05, + "loss": 0.0001, + "step": 1601 + }, + { + "epoch": 0.37, + "learning_rate": 4.6914721516061936e-05, + "loss": 0.0, + "step": 1602 + }, + { + "epoch": 0.37, + "learning_rate": 4.691279562437409e-05, + "loss": 0.0414, + "step": 1603 + }, + { + "epoch": 0.37, + "learning_rate": 4.691086973268624e-05, + "loss": 0.0293, + "step": 1604 + }, + { + "epoch": 0.37, + "learning_rate": 4.6908943840998385e-05, + "loss": 0.0006, + "step": 1605 + }, + { + "epoch": 0.37, + "learning_rate": 4.690701794931053e-05, + "loss": 0.0042, + "step": 1606 + }, + { + "epoch": 0.37, + "learning_rate": 4.690509205762268e-05, + "loss": 0.0279, + "step": 1607 + }, + { + "epoch": 0.37, + "learning_rate": 4.6903166165934834e-05, + "loss": 0.0163, + "step": 1608 + }, + { + "epoch": 0.37, + "learning_rate": 4.6901240274246975e-05, + "loss": 0.0008, + "step": 1609 + }, + { + "epoch": 0.37, + "learning_rate": 4.689931438255913e-05, + "loss": 0.0332, + "step": 1610 + }, + { + "epoch": 0.37, + "learning_rate": 4.6897388490871276e-05, + "loss": 0.0029, + "step": 1611 + }, + { + "epoch": 0.37, + "learning_rate": 4.6895462599183424e-05, + "loss": 0.0119, + "step": 1612 + }, + { + "epoch": 0.37, + "learning_rate": 4.689353670749557e-05, + "loss": 0.0298, + "step": 1613 + }, + { + "epoch": 0.37, + "learning_rate": 4.6891610815807726e-05, + "loss": 0.0012, + "step": 1614 + }, + { + "epoch": 0.37, + "learning_rate": 4.688968492411987e-05, + "loss": 0.0063, + "step": 1615 + }, + { + "epoch": 0.37, + "learning_rate": 4.6887759032432014e-05, + "loss": 0.0011, + "step": 1616 + }, + { + "epoch": 0.37, + "learning_rate": 4.688583314074417e-05, + "loss": 0.0081, + "step": 1617 + }, + { + "epoch": 0.37, + "learning_rate": 4.6883907249056316e-05, + "loss": 0.0159, + "step": 1618 + }, + { + "epoch": 0.37, + "learning_rate": 4.688198135736846e-05, + "loss": 0.0072, + "step": 1619 + }, + { + "epoch": 0.37, + "learning_rate": 4.688005546568061e-05, + "loss": 0.0013, + "step": 1620 + }, + { + "epoch": 0.37, + "learning_rate": 4.6878129573992765e-05, + "loss": 0.0008, + "step": 1621 + }, + { + "epoch": 0.37, + "learning_rate": 4.687620368230491e-05, + "loss": 0.0142, + "step": 1622 + }, + { + "epoch": 0.38, + "learning_rate": 4.687427779061706e-05, + "loss": 0.0, + "step": 1623 + }, + { + "epoch": 0.38, + "learning_rate": 4.687235189892921e-05, + "loss": 0.0001, + "step": 1624 + }, + { + "epoch": 0.38, + "learning_rate": 4.6870426007241355e-05, + "loss": 0.0, + "step": 1625 + }, + { + "epoch": 0.38, + "learning_rate": 4.68685001155535e-05, + "loss": 0.0001, + "step": 1626 + }, + { + "epoch": 0.38, + "learning_rate": 4.686657422386565e-05, + "loss": 0.009, + "step": 1627 + }, + { + "epoch": 0.38, + "learning_rate": 4.6864648332177804e-05, + "loss": 0.0029, + "step": 1628 + }, + { + "epoch": 0.38, + "learning_rate": 4.6862722440489945e-05, + "loss": 0.0, + "step": 1629 + }, + { + "epoch": 0.38, + "learning_rate": 4.68607965488021e-05, + "loss": 0.0097, + "step": 1630 + }, + { + "epoch": 0.38, + "learning_rate": 4.6858870657114246e-05, + "loss": 0.0055, + "step": 1631 + }, + { + "epoch": 0.38, + "learning_rate": 4.6856944765426394e-05, + "loss": 0.0024, + "step": 1632 + }, + { + "epoch": 0.38, + "learning_rate": 4.685501887373854e-05, + "loss": 0.0125, + "step": 1633 + }, + { + "epoch": 0.38, + "learning_rate": 4.685309298205069e-05, + "loss": 0.0003, + "step": 1634 + }, + { + "epoch": 0.38, + "learning_rate": 4.685116709036284e-05, + "loss": 0.0018, + "step": 1635 + }, + { + "epoch": 0.38, + "learning_rate": 4.6849241198674984e-05, + "loss": 0.0, + "step": 1636 + }, + { + "epoch": 0.38, + "learning_rate": 4.684731530698714e-05, + "loss": 0.006, + "step": 1637 + }, + { + "epoch": 0.38, + "learning_rate": 4.6845389415299285e-05, + "loss": 0.0029, + "step": 1638 + }, + { + "epoch": 0.38, + "learning_rate": 4.684346352361143e-05, + "loss": 0.0239, + "step": 1639 + }, + { + "epoch": 0.38, + "learning_rate": 4.684153763192358e-05, + "loss": 0.0049, + "step": 1640 + }, + { + "epoch": 0.38, + "learning_rate": 4.6839611740235735e-05, + "loss": 0.0299, + "step": 1641 + }, + { + "epoch": 0.38, + "learning_rate": 4.683768584854788e-05, + "loss": 0.0149, + "step": 1642 + }, + { + "epoch": 0.38, + "learning_rate": 4.683575995686002e-05, + "loss": 0.0, + "step": 1643 + }, + { + "epoch": 0.38, + "learning_rate": 4.683383406517218e-05, + "loss": 0.0002, + "step": 1644 + }, + { + "epoch": 0.38, + "learning_rate": 4.6831908173484325e-05, + "loss": 0.0118, + "step": 1645 + }, + { + "epoch": 0.38, + "learning_rate": 4.682998228179647e-05, + "loss": 0.001, + "step": 1646 + }, + { + "epoch": 0.38, + "learning_rate": 4.682805639010862e-05, + "loss": 0.0, + "step": 1647 + }, + { + "epoch": 0.38, + "learning_rate": 4.6826130498420774e-05, + "loss": 0.0056, + "step": 1648 + }, + { + "epoch": 0.38, + "learning_rate": 4.682420460673292e-05, + "loss": 0.0487, + "step": 1649 + }, + { + "epoch": 0.38, + "learning_rate": 4.682227871504507e-05, + "loss": 0.0078, + "step": 1650 + }, + { + "epoch": 0.38, + "learning_rate": 4.6820352823357216e-05, + "loss": 0.0, + "step": 1651 + }, + { + "epoch": 0.38, + "learning_rate": 4.6818426931669364e-05, + "loss": 0.0, + "step": 1652 + }, + { + "epoch": 0.38, + "learning_rate": 4.681650103998151e-05, + "loss": 0.0027, + "step": 1653 + }, + { + "epoch": 0.38, + "learning_rate": 4.681457514829366e-05, + "loss": 0.0023, + "step": 1654 + }, + { + "epoch": 0.38, + "learning_rate": 4.681264925660581e-05, + "loss": 0.0, + "step": 1655 + }, + { + "epoch": 0.38, + "learning_rate": 4.681072336491796e-05, + "loss": 0.0064, + "step": 1656 + }, + { + "epoch": 0.38, + "learning_rate": 4.680879747323011e-05, + "loss": 0.0035, + "step": 1657 + }, + { + "epoch": 0.38, + "learning_rate": 4.6806871581542255e-05, + "loss": 0.0094, + "step": 1658 + }, + { + "epoch": 0.38, + "learning_rate": 4.680494568985441e-05, + "loss": 0.0039, + "step": 1659 + }, + { + "epoch": 0.38, + "learning_rate": 4.680301979816655e-05, + "loss": 0.0126, + "step": 1660 + }, + { + "epoch": 0.38, + "learning_rate": 4.6801093906478704e-05, + "loss": 0.0224, + "step": 1661 + }, + { + "epoch": 0.38, + "learning_rate": 4.679916801479085e-05, + "loss": 0.0169, + "step": 1662 + }, + { + "epoch": 0.38, + "learning_rate": 4.6797242123103e-05, + "loss": 0.0006, + "step": 1663 + }, + { + "epoch": 0.38, + "learning_rate": 4.679531623141515e-05, + "loss": 0.0001, + "step": 1664 + }, + { + "epoch": 0.38, + "learning_rate": 4.6793390339727294e-05, + "loss": 0.0128, + "step": 1665 + }, + { + "epoch": 0.39, + "learning_rate": 4.679146444803945e-05, + "loss": 0.0061, + "step": 1666 + }, + { + "epoch": 0.39, + "learning_rate": 4.678953855635159e-05, + "loss": 0.0024, + "step": 1667 + }, + { + "epoch": 0.39, + "learning_rate": 4.6787612664663744e-05, + "loss": 0.0016, + "step": 1668 + }, + { + "epoch": 0.39, + "learning_rate": 4.678568677297589e-05, + "loss": 0.0039, + "step": 1669 + }, + { + "epoch": 0.39, + "learning_rate": 4.678376088128804e-05, + "loss": 0.0074, + "step": 1670 + }, + { + "epoch": 0.39, + "learning_rate": 4.6781834989600186e-05, + "loss": 0.0001, + "step": 1671 + }, + { + "epoch": 0.39, + "learning_rate": 4.6779909097912334e-05, + "loss": 0.0002, + "step": 1672 + }, + { + "epoch": 0.39, + "learning_rate": 4.677798320622449e-05, + "loss": 0.0085, + "step": 1673 + }, + { + "epoch": 0.39, + "learning_rate": 4.677605731453663e-05, + "loss": 0.0539, + "step": 1674 + }, + { + "epoch": 0.39, + "learning_rate": 4.677413142284878e-05, + "loss": 0.0042, + "step": 1675 + }, + { + "epoch": 0.39, + "learning_rate": 4.677220553116093e-05, + "loss": 0.0206, + "step": 1676 + }, + { + "epoch": 0.39, + "learning_rate": 4.677027963947308e-05, + "loss": 0.0, + "step": 1677 + }, + { + "epoch": 0.39, + "learning_rate": 4.6768353747785225e-05, + "loss": 0.0289, + "step": 1678 + }, + { + "epoch": 0.39, + "learning_rate": 4.676642785609738e-05, + "loss": 0.0121, + "step": 1679 + }, + { + "epoch": 0.39, + "learning_rate": 4.676450196440953e-05, + "loss": 0.0, + "step": 1680 + }, + { + "epoch": 0.39, + "learning_rate": 4.676257607272167e-05, + "loss": 0.0046, + "step": 1681 + }, + { + "epoch": 0.39, + "learning_rate": 4.676065018103382e-05, + "loss": 0.0015, + "step": 1682 + }, + { + "epoch": 0.39, + "learning_rate": 4.675872428934597e-05, + "loss": 0.006, + "step": 1683 + }, + { + "epoch": 0.39, + "learning_rate": 4.675679839765812e-05, + "loss": 0.0001, + "step": 1684 + }, + { + "epoch": 0.39, + "learning_rate": 4.6754872505970264e-05, + "loss": 0.0138, + "step": 1685 + }, + { + "epoch": 0.39, + "learning_rate": 4.675294661428242e-05, + "loss": 0.0031, + "step": 1686 + }, + { + "epoch": 0.39, + "learning_rate": 4.6751020722594566e-05, + "loss": 0.0086, + "step": 1687 + }, + { + "epoch": 0.39, + "learning_rate": 4.6749094830906713e-05, + "loss": 0.0005, + "step": 1688 + }, + { + "epoch": 0.39, + "learning_rate": 4.674716893921886e-05, + "loss": 0.0027, + "step": 1689 + }, + { + "epoch": 0.39, + "learning_rate": 4.674524304753101e-05, + "loss": 0.0005, + "step": 1690 + }, + { + "epoch": 0.39, + "learning_rate": 4.6743317155843156e-05, + "loss": 0.0, + "step": 1691 + }, + { + "epoch": 0.39, + "learning_rate": 4.6741391264155303e-05, + "loss": 0.0125, + "step": 1692 + }, + { + "epoch": 0.39, + "learning_rate": 4.673946537246746e-05, + "loss": 0.0004, + "step": 1693 + }, + { + "epoch": 0.39, + "learning_rate": 4.6737539480779605e-05, + "loss": 0.012, + "step": 1694 + }, + { + "epoch": 0.39, + "learning_rate": 4.673561358909175e-05, + "loss": 0.0, + "step": 1695 + }, + { + "epoch": 0.39, + "learning_rate": 4.67336876974039e-05, + "loss": 0.0018, + "step": 1696 + }, + { + "epoch": 0.39, + "learning_rate": 4.6731761805716054e-05, + "loss": 0.0, + "step": 1697 + }, + { + "epoch": 0.39, + "learning_rate": 4.6729835914028195e-05, + "loss": 0.0005, + "step": 1698 + }, + { + "epoch": 0.39, + "learning_rate": 4.672791002234034e-05, + "loss": 0.0, + "step": 1699 + }, + { + "epoch": 0.39, + "learning_rate": 4.67259841306525e-05, + "loss": 0.0002, + "step": 1700 + }, + { + "epoch": 0.39, + "learning_rate": 4.672405823896464e-05, + "loss": 0.0, + "step": 1701 + }, + { + "epoch": 0.39, + "learning_rate": 4.672213234727679e-05, + "loss": 0.0033, + "step": 1702 + }, + { + "epoch": 0.39, + "learning_rate": 4.672020645558894e-05, + "loss": 0.0018, + "step": 1703 + }, + { + "epoch": 0.39, + "learning_rate": 4.6718280563901093e-05, + "loss": 0.0121, + "step": 1704 + }, + { + "epoch": 0.39, + "learning_rate": 4.6716354672213234e-05, + "loss": 0.0045, + "step": 1705 + }, + { + "epoch": 0.39, + "learning_rate": 4.671442878052539e-05, + "loss": 0.0014, + "step": 1706 + }, + { + "epoch": 0.39, + "learning_rate": 4.6712502888837536e-05, + "loss": 0.0106, + "step": 1707 + }, + { + "epoch": 0.39, + "learning_rate": 4.6710576997149677e-05, + "loss": 0.0153, + "step": 1708 + }, + { + "epoch": 0.39, + "learning_rate": 4.670865110546183e-05, + "loss": 0.0, + "step": 1709 + }, + { + "epoch": 0.4, + "learning_rate": 4.670672521377398e-05, + "loss": 0.002, + "step": 1710 + }, + { + "epoch": 0.4, + "learning_rate": 4.6704799322086126e-05, + "loss": 0.0014, + "step": 1711 + }, + { + "epoch": 0.4, + "learning_rate": 4.670287343039827e-05, + "loss": 0.0002, + "step": 1712 + }, + { + "epoch": 0.4, + "learning_rate": 4.670094753871043e-05, + "loss": 0.0511, + "step": 1713 + }, + { + "epoch": 0.4, + "learning_rate": 4.6699021647022575e-05, + "loss": 0.0, + "step": 1714 + }, + { + "epoch": 0.4, + "learning_rate": 4.669709575533472e-05, + "loss": 0.0158, + "step": 1715 + }, + { + "epoch": 0.4, + "learning_rate": 4.669516986364687e-05, + "loss": 0.009, + "step": 1716 + }, + { + "epoch": 0.4, + "learning_rate": 4.6693243971959024e-05, + "loss": 0.0001, + "step": 1717 + }, + { + "epoch": 0.4, + "learning_rate": 4.6691318080271165e-05, + "loss": 0.0052, + "step": 1718 + }, + { + "epoch": 0.4, + "learning_rate": 4.668939218858331e-05, + "loss": 0.0121, + "step": 1719 + }, + { + "epoch": 0.4, + "learning_rate": 4.668746629689547e-05, + "loss": 0.0, + "step": 1720 + }, + { + "epoch": 0.4, + "learning_rate": 4.6685540405207614e-05, + "loss": 0.012, + "step": 1721 + }, + { + "epoch": 0.4, + "learning_rate": 4.668361451351976e-05, + "loss": 0.0004, + "step": 1722 + }, + { + "epoch": 0.4, + "learning_rate": 4.668168862183191e-05, + "loss": 0.0261, + "step": 1723 + }, + { + "epoch": 0.4, + "learning_rate": 4.667976273014406e-05, + "loss": 0.0, + "step": 1724 + }, + { + "epoch": 0.4, + "learning_rate": 4.6677836838456204e-05, + "loss": 0.0003, + "step": 1725 + }, + { + "epoch": 0.4, + "learning_rate": 4.667591094676836e-05, + "loss": 0.0001, + "step": 1726 + }, + { + "epoch": 0.4, + "learning_rate": 4.6673985055080506e-05, + "loss": 0.0003, + "step": 1727 + }, + { + "epoch": 0.4, + "learning_rate": 4.667205916339265e-05, + "loss": 0.0, + "step": 1728 + }, + { + "epoch": 0.4, + "learning_rate": 4.66701332717048e-05, + "loss": 0.0006, + "step": 1729 + }, + { + "epoch": 0.4, + "learning_rate": 4.666820738001695e-05, + "loss": 0.001, + "step": 1730 + }, + { + "epoch": 0.4, + "learning_rate": 4.66662814883291e-05, + "loss": 0.0367, + "step": 1731 + }, + { + "epoch": 0.4, + "learning_rate": 4.666435559664124e-05, + "loss": 0.002, + "step": 1732 + }, + { + "epoch": 0.4, + "learning_rate": 4.66624297049534e-05, + "loss": 0.0, + "step": 1733 + }, + { + "epoch": 0.4, + "learning_rate": 4.6660503813265545e-05, + "loss": 0.0, + "step": 1734 + }, + { + "epoch": 0.4, + "learning_rate": 4.665857792157769e-05, + "loss": 0.003, + "step": 1735 + }, + { + "epoch": 0.4, + "learning_rate": 4.665665202988984e-05, + "loss": 0.0023, + "step": 1736 + }, + { + "epoch": 0.4, + "learning_rate": 4.665472613820199e-05, + "loss": 0.0, + "step": 1737 + }, + { + "epoch": 0.4, + "learning_rate": 4.665280024651414e-05, + "loss": 0.0, + "step": 1738 + }, + { + "epoch": 0.4, + "learning_rate": 4.665087435482628e-05, + "loss": 0.0022, + "step": 1739 + }, + { + "epoch": 0.4, + "learning_rate": 4.6648948463138437e-05, + "loss": 0.0018, + "step": 1740 + }, + { + "epoch": 0.4, + "learning_rate": 4.6647022571450584e-05, + "loss": 0.0001, + "step": 1741 + }, + { + "epoch": 0.4, + "learning_rate": 4.664509667976273e-05, + "loss": 0.0074, + "step": 1742 + }, + { + "epoch": 0.4, + "learning_rate": 4.664317078807488e-05, + "loss": 0.0168, + "step": 1743 + }, + { + "epoch": 0.4, + "learning_rate": 4.664124489638703e-05, + "loss": 0.0013, + "step": 1744 + }, + { + "epoch": 0.4, + "learning_rate": 4.663931900469918e-05, + "loss": 0.0015, + "step": 1745 + }, + { + "epoch": 0.4, + "learning_rate": 4.663739311301132e-05, + "loss": 0.0014, + "step": 1746 + }, + { + "epoch": 0.4, + "learning_rate": 4.6635467221323476e-05, + "loss": 0.0, + "step": 1747 + }, + { + "epoch": 0.4, + "learning_rate": 4.663354132963562e-05, + "loss": 0.0006, + "step": 1748 + }, + { + "epoch": 0.4, + "learning_rate": 4.663161543794777e-05, + "loss": 0.0288, + "step": 1749 + }, + { + "epoch": 0.4, + "learning_rate": 4.662968954625992e-05, + "loss": 0.0005, + "step": 1750 + }, + { + "epoch": 0.4, + "learning_rate": 4.662776365457207e-05, + "loss": 0.0199, + "step": 1751 + }, + { + "epoch": 0.4, + "learning_rate": 4.662583776288422e-05, + "loss": 0.0009, + "step": 1752 + }, + { + "epoch": 0.41, + "learning_rate": 4.662391187119637e-05, + "loss": 0.0018, + "step": 1753 + }, + { + "epoch": 0.41, + "learning_rate": 4.6621985979508515e-05, + "loss": 0.0089, + "step": 1754 + }, + { + "epoch": 0.41, + "learning_rate": 4.662006008782066e-05, + "loss": 0.0014, + "step": 1755 + }, + { + "epoch": 0.41, + "learning_rate": 4.661813419613281e-05, + "loss": 0.0131, + "step": 1756 + }, + { + "epoch": 0.41, + "learning_rate": 4.661620830444496e-05, + "loss": 0.0007, + "step": 1757 + }, + { + "epoch": 0.41, + "learning_rate": 4.661428241275711e-05, + "loss": 0.0, + "step": 1758 + }, + { + "epoch": 0.41, + "learning_rate": 4.661235652106926e-05, + "loss": 0.0001, + "step": 1759 + }, + { + "epoch": 0.41, + "learning_rate": 4.6610430629381406e-05, + "loss": 0.0003, + "step": 1760 + }, + { + "epoch": 0.41, + "learning_rate": 4.6608504737693554e-05, + "loss": 0.0, + "step": 1761 + }, + { + "epoch": 0.41, + "learning_rate": 4.660657884600571e-05, + "loss": 0.0051, + "step": 1762 + }, + { + "epoch": 0.41, + "learning_rate": 4.660465295431785e-05, + "loss": 0.0317, + "step": 1763 + }, + { + "epoch": 0.41, + "learning_rate": 4.6602727062629996e-05, + "loss": 0.0001, + "step": 1764 + }, + { + "epoch": 0.41, + "learning_rate": 4.660080117094215e-05, + "loss": 0.0, + "step": 1765 + }, + { + "epoch": 0.41, + "learning_rate": 4.65988752792543e-05, + "loss": 0.0, + "step": 1766 + }, + { + "epoch": 0.41, + "learning_rate": 4.6596949387566446e-05, + "loss": 0.0, + "step": 1767 + }, + { + "epoch": 0.41, + "learning_rate": 4.659502349587859e-05, + "loss": 0.0003, + "step": 1768 + }, + { + "epoch": 0.41, + "learning_rate": 4.659309760419075e-05, + "loss": 0.054, + "step": 1769 + }, + { + "epoch": 0.41, + "learning_rate": 4.659117171250289e-05, + "loss": 0.0001, + "step": 1770 + }, + { + "epoch": 0.41, + "learning_rate": 4.658924582081504e-05, + "loss": 0.0015, + "step": 1771 + }, + { + "epoch": 0.41, + "learning_rate": 4.658731992912719e-05, + "loss": 0.0006, + "step": 1772 + }, + { + "epoch": 0.41, + "learning_rate": 4.658539403743934e-05, + "loss": 0.0178, + "step": 1773 + }, + { + "epoch": 0.41, + "learning_rate": 4.6583468145751485e-05, + "loss": 0.0015, + "step": 1774 + }, + { + "epoch": 0.41, + "learning_rate": 4.658154225406363e-05, + "loss": 0.0125, + "step": 1775 + }, + { + "epoch": 0.41, + "learning_rate": 4.6579616362375786e-05, + "loss": 0.0, + "step": 1776 + }, + { + "epoch": 0.41, + "learning_rate": 4.657769047068793e-05, + "loss": 0.0019, + "step": 1777 + }, + { + "epoch": 0.41, + "learning_rate": 4.657576457900008e-05, + "loss": 0.0, + "step": 1778 + }, + { + "epoch": 0.41, + "learning_rate": 4.657383868731223e-05, + "loss": 0.0, + "step": 1779 + }, + { + "epoch": 0.41, + "learning_rate": 4.6571912795624376e-05, + "loss": 0.0052, + "step": 1780 + }, + { + "epoch": 0.41, + "learning_rate": 4.6569986903936524e-05, + "loss": 0.0, + "step": 1781 + }, + { + "epoch": 0.41, + "learning_rate": 4.656806101224868e-05, + "loss": 0.0003, + "step": 1782 + }, + { + "epoch": 0.41, + "learning_rate": 4.656613512056082e-05, + "loss": 0.0075, + "step": 1783 + }, + { + "epoch": 0.41, + "learning_rate": 4.6564209228872966e-05, + "loss": 0.0029, + "step": 1784 + }, + { + "epoch": 0.41, + "learning_rate": 4.656228333718512e-05, + "loss": 0.0, + "step": 1785 + }, + { + "epoch": 0.41, + "learning_rate": 4.656035744549727e-05, + "loss": 0.0048, + "step": 1786 + }, + { + "epoch": 0.41, + "learning_rate": 4.6558431553809415e-05, + "loss": 0.0221, + "step": 1787 + }, + { + "epoch": 0.41, + "learning_rate": 4.655650566212156e-05, + "loss": 0.0113, + "step": 1788 + }, + { + "epoch": 0.41, + "learning_rate": 4.655457977043372e-05, + "loss": 0.0006, + "step": 1789 + }, + { + "epoch": 0.41, + "learning_rate": 4.655265387874586e-05, + "loss": 0.0061, + "step": 1790 + }, + { + "epoch": 0.41, + "learning_rate": 4.655072798705801e-05, + "loss": 0.0, + "step": 1791 + }, + { + "epoch": 0.41, + "learning_rate": 4.654880209537016e-05, + "loss": 0.018, + "step": 1792 + }, + { + "epoch": 0.41, + "learning_rate": 4.654687620368231e-05, + "loss": 0.0016, + "step": 1793 + }, + { + "epoch": 0.41, + "learning_rate": 4.6544950311994455e-05, + "loss": 0.0, + "step": 1794 + }, + { + "epoch": 0.41, + "learning_rate": 4.65430244203066e-05, + "loss": 0.0, + "step": 1795 + }, + { + "epoch": 0.42, + "learning_rate": 4.6541098528618756e-05, + "loss": 0.0217, + "step": 1796 + }, + { + "epoch": 0.42, + "learning_rate": 4.65391726369309e-05, + "loss": 0.0017, + "step": 1797 + }, + { + "epoch": 0.42, + "learning_rate": 4.653724674524305e-05, + "loss": 0.0328, + "step": 1798 + }, + { + "epoch": 0.42, + "learning_rate": 4.65353208535552e-05, + "loss": 0.0558, + "step": 1799 + }, + { + "epoch": 0.42, + "learning_rate": 4.6533394961867346e-05, + "loss": 0.0, + "step": 1800 + }, + { + "epoch": 0.42, + "learning_rate": 4.6531469070179494e-05, + "loss": 0.0004, + "step": 1801 + }, + { + "epoch": 0.42, + "learning_rate": 4.652954317849164e-05, + "loss": 0.001, + "step": 1802 + }, + { + "epoch": 0.42, + "learning_rate": 4.6527617286803795e-05, + "loss": 0.0, + "step": 1803 + }, + { + "epoch": 0.42, + "learning_rate": 4.6525691395115936e-05, + "loss": 0.0009, + "step": 1804 + }, + { + "epoch": 0.42, + "learning_rate": 4.652376550342809e-05, + "loss": 0.0, + "step": 1805 + }, + { + "epoch": 0.42, + "learning_rate": 4.652183961174024e-05, + "loss": 0.0064, + "step": 1806 + }, + { + "epoch": 0.42, + "learning_rate": 4.6519913720052385e-05, + "loss": 0.0093, + "step": 1807 + }, + { + "epoch": 0.42, + "learning_rate": 4.651798782836453e-05, + "loss": 0.0092, + "step": 1808 + }, + { + "epoch": 0.42, + "learning_rate": 4.651606193667669e-05, + "loss": 0.0014, + "step": 1809 + }, + { + "epoch": 0.42, + "learning_rate": 4.6514136044988835e-05, + "loss": 0.0, + "step": 1810 + }, + { + "epoch": 0.42, + "learning_rate": 4.6512210153300975e-05, + "loss": 0.001, + "step": 1811 + }, + { + "epoch": 0.42, + "learning_rate": 4.651028426161313e-05, + "loss": 0.0015, + "step": 1812 + }, + { + "epoch": 0.42, + "learning_rate": 4.650835836992528e-05, + "loss": 0.002, + "step": 1813 + }, + { + "epoch": 0.42, + "learning_rate": 4.6506432478237424e-05, + "loss": 0.0036, + "step": 1814 + }, + { + "epoch": 0.42, + "learning_rate": 4.650450658654957e-05, + "loss": 0.0003, + "step": 1815 + }, + { + "epoch": 0.42, + "learning_rate": 4.6502580694861726e-05, + "loss": 0.0003, + "step": 1816 + }, + { + "epoch": 0.42, + "learning_rate": 4.6500654803173874e-05, + "loss": 0.0008, + "step": 1817 + }, + { + "epoch": 0.42, + "learning_rate": 4.649872891148602e-05, + "loss": 0.0168, + "step": 1818 + }, + { + "epoch": 0.42, + "learning_rate": 4.649680301979817e-05, + "loss": 0.0298, + "step": 1819 + }, + { + "epoch": 0.42, + "learning_rate": 4.6494877128110316e-05, + "loss": 0.0375, + "step": 1820 + }, + { + "epoch": 0.42, + "learning_rate": 4.6492951236422464e-05, + "loss": 0.0032, + "step": 1821 + }, + { + "epoch": 0.42, + "learning_rate": 4.649102534473461e-05, + "loss": 0.0528, + "step": 1822 + }, + { + "epoch": 0.42, + "learning_rate": 4.6489099453046765e-05, + "loss": 0.0105, + "step": 1823 + }, + { + "epoch": 0.42, + "learning_rate": 4.648717356135891e-05, + "loss": 0.0004, + "step": 1824 + }, + { + "epoch": 0.42, + "learning_rate": 4.648524766967106e-05, + "loss": 0.0245, + "step": 1825 + }, + { + "epoch": 0.42, + "learning_rate": 4.648332177798321e-05, + "loss": 0.0159, + "step": 1826 + }, + { + "epoch": 0.42, + "learning_rate": 4.648139588629536e-05, + "loss": 0.0007, + "step": 1827 + }, + { + "epoch": 0.42, + "learning_rate": 4.64794699946075e-05, + "loss": 0.0, + "step": 1828 + }, + { + "epoch": 0.42, + "learning_rate": 4.647754410291966e-05, + "loss": 0.0069, + "step": 1829 + }, + { + "epoch": 0.42, + "learning_rate": 4.6475618211231804e-05, + "loss": 0.0006, + "step": 1830 + }, + { + "epoch": 0.42, + "learning_rate": 4.647369231954395e-05, + "loss": 0.0007, + "step": 1831 + }, + { + "epoch": 0.42, + "learning_rate": 4.64717664278561e-05, + "loss": 0.0011, + "step": 1832 + }, + { + "epoch": 0.42, + "learning_rate": 4.646984053616825e-05, + "loss": 0.0001, + "step": 1833 + }, + { + "epoch": 0.42, + "learning_rate": 4.64679146444804e-05, + "loss": 0.0008, + "step": 1834 + }, + { + "epoch": 0.42, + "learning_rate": 4.646598875279254e-05, + "loss": 0.0064, + "step": 1835 + }, + { + "epoch": 0.42, + "learning_rate": 4.6464062861104696e-05, + "loss": 0.0001, + "step": 1836 + }, + { + "epoch": 0.42, + "learning_rate": 4.6462136969416844e-05, + "loss": 0.0, + "step": 1837 + }, + { + "epoch": 0.42, + "learning_rate": 4.646021107772899e-05, + "loss": 0.0012, + "step": 1838 + }, + { + "epoch": 0.43, + "learning_rate": 4.645828518604114e-05, + "loss": 0.0011, + "step": 1839 + }, + { + "epoch": 0.43, + "learning_rate": 4.6456359294353286e-05, + "loss": 0.002, + "step": 1840 + }, + { + "epoch": 0.43, + "learning_rate": 4.645443340266544e-05, + "loss": 0.0006, + "step": 1841 + }, + { + "epoch": 0.43, + "learning_rate": 4.645250751097758e-05, + "loss": 0.0025, + "step": 1842 + }, + { + "epoch": 0.43, + "learning_rate": 4.6450581619289735e-05, + "loss": 0.0001, + "step": 1843 + }, + { + "epoch": 0.43, + "learning_rate": 4.644865572760188e-05, + "loss": 0.001, + "step": 1844 + }, + { + "epoch": 0.43, + "learning_rate": 4.644672983591403e-05, + "loss": 0.0093, + "step": 1845 + }, + { + "epoch": 0.43, + "learning_rate": 4.644480394422618e-05, + "loss": 0.0064, + "step": 1846 + }, + { + "epoch": 0.43, + "learning_rate": 4.644287805253833e-05, + "loss": 0.0016, + "step": 1847 + }, + { + "epoch": 0.43, + "learning_rate": 4.644095216085048e-05, + "loss": 0.0006, + "step": 1848 + }, + { + "epoch": 0.43, + "learning_rate": 4.643902626916262e-05, + "loss": 0.0058, + "step": 1849 + }, + { + "epoch": 0.43, + "learning_rate": 4.6437100377474774e-05, + "loss": 0.0, + "step": 1850 + }, + { + "epoch": 0.43, + "learning_rate": 4.643517448578692e-05, + "loss": 0.0295, + "step": 1851 + }, + { + "epoch": 0.43, + "learning_rate": 4.643324859409907e-05, + "loss": 0.0, + "step": 1852 + }, + { + "epoch": 0.43, + "learning_rate": 4.643132270241122e-05, + "loss": 0.0002, + "step": 1853 + }, + { + "epoch": 0.43, + "learning_rate": 4.642939681072337e-05, + "loss": 0.0003, + "step": 1854 + }, + { + "epoch": 0.43, + "learning_rate": 4.642747091903551e-05, + "loss": 0.001, + "step": 1855 + }, + { + "epoch": 0.43, + "learning_rate": 4.6425545027347666e-05, + "loss": 0.0092, + "step": 1856 + }, + { + "epoch": 0.43, + "learning_rate": 4.6423619135659813e-05, + "loss": 0.0003, + "step": 1857 + }, + { + "epoch": 0.43, + "learning_rate": 4.642169324397196e-05, + "loss": 0.0, + "step": 1858 + }, + { + "epoch": 0.43, + "learning_rate": 4.641976735228411e-05, + "loss": 0.0019, + "step": 1859 + }, + { + "epoch": 0.43, + "learning_rate": 4.6417841460596256e-05, + "loss": 0.022, + "step": 1860 + }, + { + "epoch": 0.43, + "learning_rate": 4.641591556890841e-05, + "loss": 0.0, + "step": 1861 + }, + { + "epoch": 0.43, + "learning_rate": 4.641398967722055e-05, + "loss": 0.0038, + "step": 1862 + }, + { + "epoch": 0.43, + "learning_rate": 4.6412063785532705e-05, + "loss": 0.0012, + "step": 1863 + }, + { + "epoch": 0.43, + "learning_rate": 4.641013789384485e-05, + "loss": 0.0001, + "step": 1864 + }, + { + "epoch": 0.43, + "learning_rate": 4.6408212002157e-05, + "loss": 0.0006, + "step": 1865 + }, + { + "epoch": 0.43, + "learning_rate": 4.640628611046915e-05, + "loss": 0.0122, + "step": 1866 + }, + { + "epoch": 0.43, + "learning_rate": 4.6404360218781295e-05, + "loss": 0.0276, + "step": 1867 + }, + { + "epoch": 0.43, + "learning_rate": 4.640243432709345e-05, + "loss": 0.007, + "step": 1868 + }, + { + "epoch": 0.43, + "learning_rate": 4.640050843540559e-05, + "loss": 0.0203, + "step": 1869 + }, + { + "epoch": 0.43, + "learning_rate": 4.6398582543717744e-05, + "loss": 0.0, + "step": 1870 + }, + { + "epoch": 0.43, + "learning_rate": 4.639665665202989e-05, + "loss": 0.0, + "step": 1871 + }, + { + "epoch": 0.43, + "learning_rate": 4.639473076034204e-05, + "loss": 0.0015, + "step": 1872 + }, + { + "epoch": 0.43, + "learning_rate": 4.6392804868654187e-05, + "loss": 0.0015, + "step": 1873 + }, + { + "epoch": 0.43, + "learning_rate": 4.639087897696634e-05, + "loss": 0.0036, + "step": 1874 + }, + { + "epoch": 0.43, + "learning_rate": 4.638895308527849e-05, + "loss": 0.0, + "step": 1875 + }, + { + "epoch": 0.43, + "learning_rate": 4.638702719359063e-05, + "loss": 0.0003, + "step": 1876 + }, + { + "epoch": 0.43, + "learning_rate": 4.638510130190278e-05, + "loss": 0.0062, + "step": 1877 + }, + { + "epoch": 0.43, + "learning_rate": 4.638317541021493e-05, + "loss": 0.0011, + "step": 1878 + }, + { + "epoch": 0.43, + "learning_rate": 4.638124951852708e-05, + "loss": 0.0205, + "step": 1879 + }, + { + "epoch": 0.43, + "learning_rate": 4.6379323626839226e-05, + "loss": 0.0004, + "step": 1880 + }, + { + "epoch": 0.43, + "learning_rate": 4.637739773515138e-05, + "loss": 0.0013, + "step": 1881 + }, + { + "epoch": 0.43, + "learning_rate": 4.637547184346353e-05, + "loss": 0.0019, + "step": 1882 + }, + { + "epoch": 0.44, + "learning_rate": 4.6373545951775675e-05, + "loss": 0.0, + "step": 1883 + }, + { + "epoch": 0.44, + "learning_rate": 4.637162006008782e-05, + "loss": 0.0115, + "step": 1884 + }, + { + "epoch": 0.44, + "learning_rate": 4.636969416839998e-05, + "loss": 0.0, + "step": 1885 + }, + { + "epoch": 0.44, + "learning_rate": 4.636776827671212e-05, + "loss": 0.0015, + "step": 1886 + }, + { + "epoch": 0.44, + "learning_rate": 4.6365842385024265e-05, + "loss": 0.0038, + "step": 1887 + }, + { + "epoch": 0.44, + "learning_rate": 4.636391649333642e-05, + "loss": 0.0, + "step": 1888 + }, + { + "epoch": 0.44, + "learning_rate": 4.6361990601648567e-05, + "loss": 0.0024, + "step": 1889 + }, + { + "epoch": 0.44, + "learning_rate": 4.6360064709960714e-05, + "loss": 0.0078, + "step": 1890 + }, + { + "epoch": 0.44, + "learning_rate": 4.635813881827286e-05, + "loss": 0.0002, + "step": 1891 + }, + { + "epoch": 0.44, + "learning_rate": 4.6356212926585016e-05, + "loss": 0.0, + "step": 1892 + }, + { + "epoch": 0.44, + "learning_rate": 4.6354287034897156e-05, + "loss": 0.0004, + "step": 1893 + }, + { + "epoch": 0.44, + "learning_rate": 4.635236114320931e-05, + "loss": 0.0004, + "step": 1894 + }, + { + "epoch": 0.44, + "learning_rate": 4.635043525152146e-05, + "loss": 0.0001, + "step": 1895 + }, + { + "epoch": 0.44, + "learning_rate": 4.6348509359833606e-05, + "loss": 0.0009, + "step": 1896 + }, + { + "epoch": 0.44, + "learning_rate": 4.634658346814575e-05, + "loss": 0.0148, + "step": 1897 + }, + { + "epoch": 0.44, + "learning_rate": 4.63446575764579e-05, + "loss": 0.0, + "step": 1898 + }, + { + "epoch": 0.44, + "learning_rate": 4.6342731684770055e-05, + "loss": 0.0409, + "step": 1899 + }, + { + "epoch": 0.44, + "learning_rate": 4.6340805793082196e-05, + "loss": 0.0, + "step": 1900 + }, + { + "epoch": 0.44, + "learning_rate": 4.633887990139435e-05, + "loss": 0.0032, + "step": 1901 + }, + { + "epoch": 0.44, + "learning_rate": 4.63369540097065e-05, + "loss": 0.0036, + "step": 1902 + }, + { + "epoch": 0.44, + "learning_rate": 4.6335028118018645e-05, + "loss": 0.0003, + "step": 1903 + }, + { + "epoch": 0.44, + "learning_rate": 4.633310222633079e-05, + "loss": 0.0016, + "step": 1904 + }, + { + "epoch": 0.44, + "learning_rate": 4.633117633464294e-05, + "loss": 0.0001, + "step": 1905 + }, + { + "epoch": 0.44, + "learning_rate": 4.6329250442955094e-05, + "loss": 0.0033, + "step": 1906 + }, + { + "epoch": 0.44, + "learning_rate": 4.6327324551267235e-05, + "loss": 0.0, + "step": 1907 + }, + { + "epoch": 0.44, + "learning_rate": 4.632539865957939e-05, + "loss": 0.0094, + "step": 1908 + }, + { + "epoch": 0.44, + "learning_rate": 4.6323472767891536e-05, + "loss": 0.0013, + "step": 1909 + }, + { + "epoch": 0.44, + "learning_rate": 4.6321546876203684e-05, + "loss": 0.0123, + "step": 1910 + }, + { + "epoch": 0.44, + "learning_rate": 4.631962098451583e-05, + "loss": 0.0, + "step": 1911 + }, + { + "epoch": 0.44, + "learning_rate": 4.6317695092827986e-05, + "loss": 0.0083, + "step": 1912 + }, + { + "epoch": 0.44, + "learning_rate": 4.631576920114013e-05, + "loss": 0.0003, + "step": 1913 + }, + { + "epoch": 0.44, + "learning_rate": 4.6313843309452274e-05, + "loss": 0.0056, + "step": 1914 + }, + { + "epoch": 0.44, + "learning_rate": 4.631191741776443e-05, + "loss": 0.0038, + "step": 1915 + }, + { + "epoch": 0.44, + "learning_rate": 4.6309991526076576e-05, + "loss": 0.0029, + "step": 1916 + }, + { + "epoch": 0.44, + "learning_rate": 4.630806563438872e-05, + "loss": 0.0433, + "step": 1917 + }, + { + "epoch": 0.44, + "learning_rate": 4.630613974270087e-05, + "loss": 0.0, + "step": 1918 + }, + { + "epoch": 0.44, + "learning_rate": 4.6304213851013025e-05, + "loss": 0.0, + "step": 1919 + }, + { + "epoch": 0.44, + "learning_rate": 4.630228795932517e-05, + "loss": 0.0022, + "step": 1920 + }, + { + "epoch": 0.44, + "learning_rate": 4.630036206763732e-05, + "loss": 0.0, + "step": 1921 + }, + { + "epoch": 0.44, + "learning_rate": 4.629843617594947e-05, + "loss": 0.0068, + "step": 1922 + }, + { + "epoch": 0.44, + "learning_rate": 4.6296510284261615e-05, + "loss": 0.0, + "step": 1923 + }, + { + "epoch": 0.44, + "learning_rate": 4.629458439257376e-05, + "loss": 0.0101, + "step": 1924 + }, + { + "epoch": 0.44, + "learning_rate": 4.629265850088591e-05, + "loss": 0.0006, + "step": 1925 + }, + { + "epoch": 0.45, + "learning_rate": 4.6290732609198064e-05, + "loss": 0.0001, + "step": 1926 + }, + { + "epoch": 0.45, + "learning_rate": 4.628880671751021e-05, + "loss": 0.0006, + "step": 1927 + }, + { + "epoch": 0.45, + "learning_rate": 4.628688082582236e-05, + "loss": 0.0324, + "step": 1928 + }, + { + "epoch": 0.45, + "learning_rate": 4.6284954934134506e-05, + "loss": 0.0025, + "step": 1929 + }, + { + "epoch": 0.45, + "learning_rate": 4.628302904244666e-05, + "loss": 0.0, + "step": 1930 + }, + { + "epoch": 0.45, + "learning_rate": 4.62811031507588e-05, + "loss": 0.0, + "step": 1931 + }, + { + "epoch": 0.45, + "learning_rate": 4.627917725907095e-05, + "loss": 0.0008, + "step": 1932 + }, + { + "epoch": 0.45, + "learning_rate": 4.62772513673831e-05, + "loss": 0.0042, + "step": 1933 + }, + { + "epoch": 0.45, + "learning_rate": 4.6275325475695244e-05, + "loss": 0.0, + "step": 1934 + }, + { + "epoch": 0.45, + "learning_rate": 4.62733995840074e-05, + "loss": 0.0104, + "step": 1935 + }, + { + "epoch": 0.45, + "learning_rate": 4.6271473692319545e-05, + "loss": 0.0, + "step": 1936 + }, + { + "epoch": 0.45, + "learning_rate": 4.626954780063169e-05, + "loss": 0.0022, + "step": 1937 + }, + { + "epoch": 0.45, + "learning_rate": 4.626762190894384e-05, + "loss": 0.0216, + "step": 1938 + }, + { + "epoch": 0.45, + "learning_rate": 4.6265696017255995e-05, + "loss": 0.0001, + "step": 1939 + }, + { + "epoch": 0.45, + "learning_rate": 4.626377012556814e-05, + "loss": 0.0002, + "step": 1940 + }, + { + "epoch": 0.45, + "learning_rate": 4.626184423388029e-05, + "loss": 0.0001, + "step": 1941 + }, + { + "epoch": 0.45, + "learning_rate": 4.625991834219244e-05, + "loss": 0.0001, + "step": 1942 + }, + { + "epoch": 0.45, + "learning_rate": 4.6257992450504585e-05, + "loss": 0.0056, + "step": 1943 + }, + { + "epoch": 0.45, + "learning_rate": 4.625606655881673e-05, + "loss": 0.0, + "step": 1944 + }, + { + "epoch": 0.45, + "learning_rate": 4.625414066712888e-05, + "loss": 0.002, + "step": 1945 + }, + { + "epoch": 0.45, + "learning_rate": 4.6252214775441034e-05, + "loss": 0.0017, + "step": 1946 + }, + { + "epoch": 0.45, + "learning_rate": 4.625028888375318e-05, + "loss": 0.0004, + "step": 1947 + }, + { + "epoch": 0.45, + "learning_rate": 4.624836299206533e-05, + "loss": 0.0008, + "step": 1948 + }, + { + "epoch": 0.45, + "learning_rate": 4.6246437100377476e-05, + "loss": 0.0014, + "step": 1949 + }, + { + "epoch": 0.45, + "learning_rate": 4.624451120868963e-05, + "loss": 0.0, + "step": 1950 + }, + { + "epoch": 0.45, + "learning_rate": 4.624258531700177e-05, + "loss": 0.0, + "step": 1951 + }, + { + "epoch": 0.45, + "learning_rate": 4.624065942531392e-05, + "loss": 0.0064, + "step": 1952 + }, + { + "epoch": 0.45, + "learning_rate": 4.623873353362607e-05, + "loss": 0.0, + "step": 1953 + }, + { + "epoch": 0.45, + "learning_rate": 4.623680764193822e-05, + "loss": 0.005, + "step": 1954 + }, + { + "epoch": 0.45, + "learning_rate": 4.623488175025037e-05, + "loss": 0.0002, + "step": 1955 + }, + { + "epoch": 0.45, + "learning_rate": 4.6232955858562515e-05, + "loss": 0.0204, + "step": 1956 + }, + { + "epoch": 0.45, + "learning_rate": 4.623102996687467e-05, + "loss": 0.0002, + "step": 1957 + }, + { + "epoch": 0.45, + "learning_rate": 4.622910407518681e-05, + "loss": 0.002, + "step": 1958 + }, + { + "epoch": 0.45, + "learning_rate": 4.6227178183498965e-05, + "loss": 0.0018, + "step": 1959 + }, + { + "epoch": 0.45, + "learning_rate": 4.622525229181111e-05, + "loss": 0.0007, + "step": 1960 + }, + { + "epoch": 0.45, + "learning_rate": 4.622332640012326e-05, + "loss": 0.0, + "step": 1961 + }, + { + "epoch": 0.45, + "learning_rate": 4.622140050843541e-05, + "loss": 0.0004, + "step": 1962 + }, + { + "epoch": 0.45, + "learning_rate": 4.6219474616747554e-05, + "loss": 0.0207, + "step": 1963 + }, + { + "epoch": 0.45, + "learning_rate": 4.621754872505971e-05, + "loss": 0.0, + "step": 1964 + }, + { + "epoch": 0.45, + "learning_rate": 4.621562283337185e-05, + "loss": 0.0002, + "step": 1965 + }, + { + "epoch": 0.45, + "learning_rate": 4.6213696941684004e-05, + "loss": 0.0965, + "step": 1966 + }, + { + "epoch": 0.45, + "learning_rate": 4.621177104999615e-05, + "loss": 0.0027, + "step": 1967 + }, + { + "epoch": 0.45, + "learning_rate": 4.62098451583083e-05, + "loss": 0.0004, + "step": 1968 + }, + { + "epoch": 0.46, + "learning_rate": 4.6207919266620446e-05, + "loss": 0.0003, + "step": 1969 + }, + { + "epoch": 0.46, + "learning_rate": 4.6205993374932594e-05, + "loss": 0.0085, + "step": 1970 + }, + { + "epoch": 0.46, + "learning_rate": 4.620406748324475e-05, + "loss": 0.0, + "step": 1971 + }, + { + "epoch": 0.46, + "learning_rate": 4.620214159155689e-05, + "loss": 0.0061, + "step": 1972 + }, + { + "epoch": 0.46, + "learning_rate": 4.620021569986904e-05, + "loss": 0.0002, + "step": 1973 + }, + { + "epoch": 0.46, + "learning_rate": 4.619828980818119e-05, + "loss": 0.0107, + "step": 1974 + }, + { + "epoch": 0.46, + "learning_rate": 4.619636391649334e-05, + "loss": 0.0031, + "step": 1975 + }, + { + "epoch": 0.46, + "learning_rate": 4.6194438024805485e-05, + "loss": 0.0006, + "step": 1976 + }, + { + "epoch": 0.46, + "learning_rate": 4.619251213311764e-05, + "loss": 0.0105, + "step": 1977 + }, + { + "epoch": 0.46, + "learning_rate": 4.619058624142979e-05, + "loss": 0.0056, + "step": 1978 + }, + { + "epoch": 0.46, + "learning_rate": 4.618866034974193e-05, + "loss": 0.0, + "step": 1979 + }, + { + "epoch": 0.46, + "learning_rate": 4.618673445805408e-05, + "loss": 0.0015, + "step": 1980 + }, + { + "epoch": 0.46, + "learning_rate": 4.618480856636623e-05, + "loss": 0.0013, + "step": 1981 + }, + { + "epoch": 0.46, + "learning_rate": 4.618288267467838e-05, + "loss": 0.0, + "step": 1982 + }, + { + "epoch": 0.46, + "learning_rate": 4.6180956782990524e-05, + "loss": 0.0009, + "step": 1983 + }, + { + "epoch": 0.46, + "learning_rate": 4.617903089130268e-05, + "loss": 0.0024, + "step": 1984 + }, + { + "epoch": 0.46, + "learning_rate": 4.6177104999614826e-05, + "loss": 0.0007, + "step": 1985 + }, + { + "epoch": 0.46, + "learning_rate": 4.6175179107926974e-05, + "loss": 0.0045, + "step": 1986 + }, + { + "epoch": 0.46, + "learning_rate": 4.617325321623912e-05, + "loss": 0.0506, + "step": 1987 + }, + { + "epoch": 0.46, + "learning_rate": 4.617132732455127e-05, + "loss": 0.0018, + "step": 1988 + }, + { + "epoch": 0.46, + "learning_rate": 4.6169401432863416e-05, + "loss": 0.0119, + "step": 1989 + }, + { + "epoch": 0.46, + "learning_rate": 4.6167475541175563e-05, + "loss": 0.0, + "step": 1990 + }, + { + "epoch": 0.46, + "learning_rate": 4.616554964948772e-05, + "loss": 0.0002, + "step": 1991 + }, + { + "epoch": 0.46, + "learning_rate": 4.6163623757799865e-05, + "loss": 0.0, + "step": 1992 + }, + { + "epoch": 0.46, + "learning_rate": 4.616169786611201e-05, + "loss": 0.0, + "step": 1993 + }, + { + "epoch": 0.46, + "learning_rate": 4.615977197442416e-05, + "loss": 0.0003, + "step": 1994 + }, + { + "epoch": 0.46, + "learning_rate": 4.6157846082736314e-05, + "loss": 0.0, + "step": 1995 + }, + { + "epoch": 0.46, + "learning_rate": 4.6155920191048455e-05, + "loss": 0.0012, + "step": 1996 + }, + { + "epoch": 0.46, + "learning_rate": 4.615399429936061e-05, + "loss": 0.017, + "step": 1997 + }, + { + "epoch": 0.46, + "learning_rate": 4.615206840767276e-05, + "loss": 0.0277, + "step": 1998 + }, + { + "epoch": 0.46, + "learning_rate": 4.6150142515984904e-05, + "loss": 0.0003, + "step": 1999 + }, + { + "epoch": 0.46, + "learning_rate": 4.614821662429705e-05, + "loss": 0.0201, + "step": 2000 + }, + { + "epoch": 0.46, + "learning_rate": 4.61462907326092e-05, + "loss": 0.0016, + "step": 2001 + }, + { + "epoch": 0.46, + "learning_rate": 4.6144364840921354e-05, + "loss": 0.0036, + "step": 2002 + }, + { + "epoch": 0.46, + "learning_rate": 4.6142438949233494e-05, + "loss": 0.0003, + "step": 2003 + }, + { + "epoch": 0.46, + "learning_rate": 4.614051305754565e-05, + "loss": 0.0005, + "step": 2004 + }, + { + "epoch": 0.46, + "learning_rate": 4.6138587165857796e-05, + "loss": 0.0084, + "step": 2005 + }, + { + "epoch": 0.46, + "learning_rate": 4.6136661274169943e-05, + "loss": 0.0, + "step": 2006 + }, + { + "epoch": 0.46, + "learning_rate": 4.613473538248209e-05, + "loss": 0.0004, + "step": 2007 + }, + { + "epoch": 0.46, + "learning_rate": 4.613280949079424e-05, + "loss": 0.0, + "step": 2008 + }, + { + "epoch": 0.46, + "learning_rate": 4.6130883599106386e-05, + "loss": 0.0001, + "step": 2009 + }, + { + "epoch": 0.46, + "learning_rate": 4.612895770741853e-05, + "loss": 0.0007, + "step": 2010 + }, + { + "epoch": 0.46, + "learning_rate": 4.612703181573069e-05, + "loss": 0.0051, + "step": 2011 + }, + { + "epoch": 0.46, + "learning_rate": 4.6125105924042835e-05, + "loss": 0.0132, + "step": 2012 + }, + { + "epoch": 0.47, + "learning_rate": 4.612318003235498e-05, + "loss": 0.0026, + "step": 2013 + }, + { + "epoch": 0.47, + "learning_rate": 4.612125414066713e-05, + "loss": 0.0, + "step": 2014 + }, + { + "epoch": 0.47, + "learning_rate": 4.6119328248979284e-05, + "loss": 0.0074, + "step": 2015 + }, + { + "epoch": 0.47, + "learning_rate": 4.6117402357291425e-05, + "loss": 0.0087, + "step": 2016 + }, + { + "epoch": 0.47, + "learning_rate": 4.611547646560357e-05, + "loss": 0.0003, + "step": 2017 + }, + { + "epoch": 0.47, + "learning_rate": 4.611355057391573e-05, + "loss": 0.0006, + "step": 2018 + }, + { + "epoch": 0.47, + "learning_rate": 4.6111624682227874e-05, + "loss": 0.0083, + "step": 2019 + }, + { + "epoch": 0.47, + "learning_rate": 4.610969879054002e-05, + "loss": 0.0007, + "step": 2020 + }, + { + "epoch": 0.47, + "learning_rate": 4.610777289885217e-05, + "loss": 0.0057, + "step": 2021 + }, + { + "epoch": 0.47, + "learning_rate": 4.6105847007164323e-05, + "loss": 0.0017, + "step": 2022 + }, + { + "epoch": 0.47, + "learning_rate": 4.6103921115476464e-05, + "loss": 0.001, + "step": 2023 + }, + { + "epoch": 0.47, + "learning_rate": 4.610199522378862e-05, + "loss": 0.0001, + "step": 2024 + }, + { + "epoch": 0.47, + "learning_rate": 4.6100069332100766e-05, + "loss": 0.0, + "step": 2025 + }, + { + "epoch": 0.47, + "learning_rate": 4.609814344041291e-05, + "loss": 0.0, + "step": 2026 + }, + { + "epoch": 0.47, + "learning_rate": 4.609621754872506e-05, + "loss": 0.0267, + "step": 2027 + }, + { + "epoch": 0.47, + "learning_rate": 4.609429165703721e-05, + "loss": 0.0015, + "step": 2028 + }, + { + "epoch": 0.47, + "learning_rate": 4.609236576534936e-05, + "loss": 0.0003, + "step": 2029 + }, + { + "epoch": 0.47, + "learning_rate": 4.60904398736615e-05, + "loss": 0.0006, + "step": 2030 + }, + { + "epoch": 0.47, + "learning_rate": 4.608851398197366e-05, + "loss": 0.0002, + "step": 2031 + }, + { + "epoch": 0.47, + "learning_rate": 4.6086588090285805e-05, + "loss": 0.0004, + "step": 2032 + }, + { + "epoch": 0.47, + "learning_rate": 4.608466219859795e-05, + "loss": 0.0229, + "step": 2033 + }, + { + "epoch": 0.47, + "learning_rate": 4.60827363069101e-05, + "loss": 0.0017, + "step": 2034 + }, + { + "epoch": 0.47, + "learning_rate": 4.608081041522225e-05, + "loss": 0.0244, + "step": 2035 + }, + { + "epoch": 0.47, + "learning_rate": 4.60788845235344e-05, + "loss": 0.0, + "step": 2036 + }, + { + "epoch": 0.47, + "learning_rate": 4.607695863184654e-05, + "loss": 0.0004, + "step": 2037 + }, + { + "epoch": 0.47, + "learning_rate": 4.60750327401587e-05, + "loss": 0.0005, + "step": 2038 + }, + { + "epoch": 0.47, + "learning_rate": 4.6073106848470844e-05, + "loss": 0.0018, + "step": 2039 + }, + { + "epoch": 0.47, + "learning_rate": 4.607118095678299e-05, + "loss": 0.0221, + "step": 2040 + }, + { + "epoch": 0.47, + "learning_rate": 4.606925506509514e-05, + "loss": 0.0097, + "step": 2041 + }, + { + "epoch": 0.47, + "learning_rate": 4.606732917340729e-05, + "loss": 0.0001, + "step": 2042 + }, + { + "epoch": 0.47, + "learning_rate": 4.606540328171944e-05, + "loss": 0.0, + "step": 2043 + }, + { + "epoch": 0.47, + "learning_rate": 4.606347739003158e-05, + "loss": 0.0209, + "step": 2044 + }, + { + "epoch": 0.47, + "learning_rate": 4.6061551498343736e-05, + "loss": 0.0198, + "step": 2045 + }, + { + "epoch": 0.47, + "learning_rate": 4.605962560665588e-05, + "loss": 0.0007, + "step": 2046 + }, + { + "epoch": 0.47, + "learning_rate": 4.605769971496803e-05, + "loss": 0.0, + "step": 2047 + }, + { + "epoch": 0.47, + "learning_rate": 4.605577382328018e-05, + "loss": 0.0036, + "step": 2048 + }, + { + "epoch": 0.47, + "learning_rate": 4.605384793159233e-05, + "loss": 0.0015, + "step": 2049 + }, + { + "epoch": 0.47, + "learning_rate": 4.605192203990448e-05, + "loss": 0.0, + "step": 2050 + }, + { + "epoch": 0.47, + "learning_rate": 4.604999614821663e-05, + "loss": 0.0073, + "step": 2051 + }, + { + "epoch": 0.47, + "learning_rate": 4.6048070256528775e-05, + "loss": 0.0007, + "step": 2052 + }, + { + "epoch": 0.47, + "learning_rate": 4.604614436484093e-05, + "loss": 0.0021, + "step": 2053 + }, + { + "epoch": 0.47, + "learning_rate": 4.604421847315307e-05, + "loss": 0.0, + "step": 2054 + }, + { + "epoch": 0.47, + "learning_rate": 4.604229258146522e-05, + "loss": 0.0003, + "step": 2055 + }, + { + "epoch": 0.48, + "learning_rate": 4.604036668977737e-05, + "loss": 0.0, + "step": 2056 + }, + { + "epoch": 0.48, + "learning_rate": 4.603844079808952e-05, + "loss": 0.0024, + "step": 2057 + }, + { + "epoch": 0.48, + "learning_rate": 4.6036514906401667e-05, + "loss": 0.0, + "step": 2058 + }, + { + "epoch": 0.48, + "learning_rate": 4.6034589014713814e-05, + "loss": 0.0006, + "step": 2059 + }, + { + "epoch": 0.48, + "learning_rate": 4.603266312302597e-05, + "loss": 0.0001, + "step": 2060 + }, + { + "epoch": 0.48, + "learning_rate": 4.603073723133811e-05, + "loss": 0.002, + "step": 2061 + }, + { + "epoch": 0.48, + "learning_rate": 4.602881133965026e-05, + "loss": 0.0484, + "step": 2062 + }, + { + "epoch": 0.48, + "learning_rate": 4.602688544796241e-05, + "loss": 0.0053, + "step": 2063 + }, + { + "epoch": 0.48, + "learning_rate": 4.602495955627456e-05, + "loss": 0.0001, + "step": 2064 + }, + { + "epoch": 0.48, + "learning_rate": 4.6023033664586706e-05, + "loss": 0.0057, + "step": 2065 + }, + { + "epoch": 0.48, + "learning_rate": 4.602110777289885e-05, + "loss": 0.0, + "step": 2066 + }, + { + "epoch": 0.48, + "learning_rate": 4.601918188121101e-05, + "loss": 0.0059, + "step": 2067 + }, + { + "epoch": 0.48, + "learning_rate": 4.601725598952315e-05, + "loss": 0.0021, + "step": 2068 + }, + { + "epoch": 0.48, + "learning_rate": 4.60153300978353e-05, + "loss": 0.0007, + "step": 2069 + }, + { + "epoch": 0.48, + "learning_rate": 4.601340420614745e-05, + "loss": 0.0059, + "step": 2070 + }, + { + "epoch": 0.48, + "learning_rate": 4.60114783144596e-05, + "loss": 0.0165, + "step": 2071 + }, + { + "epoch": 0.48, + "learning_rate": 4.6009552422771745e-05, + "loss": 0.0006, + "step": 2072 + }, + { + "epoch": 0.48, + "learning_rate": 4.600762653108389e-05, + "loss": 0.0003, + "step": 2073 + }, + { + "epoch": 0.48, + "learning_rate": 4.6005700639396046e-05, + "loss": 0.0002, + "step": 2074 + }, + { + "epoch": 0.48, + "learning_rate": 4.600377474770819e-05, + "loss": 0.0032, + "step": 2075 + }, + { + "epoch": 0.48, + "learning_rate": 4.600184885602034e-05, + "loss": 0.0001, + "step": 2076 + }, + { + "epoch": 0.48, + "learning_rate": 4.599992296433249e-05, + "loss": 0.0, + "step": 2077 + }, + { + "epoch": 0.48, + "learning_rate": 4.5997997072644636e-05, + "loss": 0.0, + "step": 2078 + }, + { + "epoch": 0.48, + "learning_rate": 4.5996071180956784e-05, + "loss": 0.0002, + "step": 2079 + }, + { + "epoch": 0.48, + "learning_rate": 4.599414528926894e-05, + "loss": 0.0165, + "step": 2080 + }, + { + "epoch": 0.48, + "learning_rate": 4.5992219397581086e-05, + "loss": 0.0001, + "step": 2081 + }, + { + "epoch": 0.48, + "learning_rate": 4.5990293505893226e-05, + "loss": 0.0054, + "step": 2082 + }, + { + "epoch": 0.48, + "learning_rate": 4.598836761420538e-05, + "loss": 0.0016, + "step": 2083 + }, + { + "epoch": 0.48, + "learning_rate": 4.598644172251753e-05, + "loss": 0.0, + "step": 2084 + }, + { + "epoch": 0.48, + "learning_rate": 4.5984515830829676e-05, + "loss": 0.0, + "step": 2085 + }, + { + "epoch": 0.48, + "learning_rate": 4.598258993914182e-05, + "loss": 0.0514, + "step": 2086 + }, + { + "epoch": 0.48, + "learning_rate": 4.598066404745398e-05, + "loss": 0.0047, + "step": 2087 + }, + { + "epoch": 0.48, + "learning_rate": 4.597873815576612e-05, + "loss": 0.0019, + "step": 2088 + }, + { + "epoch": 0.48, + "learning_rate": 4.597681226407827e-05, + "loss": 0.0058, + "step": 2089 + }, + { + "epoch": 0.48, + "learning_rate": 4.597488637239042e-05, + "loss": 0.0003, + "step": 2090 + }, + { + "epoch": 0.48, + "learning_rate": 4.597296048070257e-05, + "loss": 0.0059, + "step": 2091 + }, + { + "epoch": 0.48, + "learning_rate": 4.5971034589014715e-05, + "loss": 0.0165, + "step": 2092 + }, + { + "epoch": 0.48, + "learning_rate": 4.596910869732686e-05, + "loss": 0.0003, + "step": 2093 + }, + { + "epoch": 0.48, + "learning_rate": 4.5967182805639016e-05, + "loss": 0.0029, + "step": 2094 + }, + { + "epoch": 0.48, + "learning_rate": 4.596525691395116e-05, + "loss": 0.0012, + "step": 2095 + }, + { + "epoch": 0.48, + "learning_rate": 4.596333102226331e-05, + "loss": 0.0, + "step": 2096 + }, + { + "epoch": 0.48, + "learning_rate": 4.596140513057546e-05, + "loss": 0.0003, + "step": 2097 + }, + { + "epoch": 0.48, + "learning_rate": 4.5959479238887606e-05, + "loss": 0.0007, + "step": 2098 + }, + { + "epoch": 0.49, + "learning_rate": 4.5957553347199754e-05, + "loss": 0.0003, + "step": 2099 + }, + { + "epoch": 0.49, + "learning_rate": 4.59556274555119e-05, + "loss": 0.0008, + "step": 2100 + }, + { + "epoch": 0.49, + "learning_rate": 4.5953701563824055e-05, + "loss": 0.0, + "step": 2101 + }, + { + "epoch": 0.49, + "learning_rate": 4.5951775672136196e-05, + "loss": 0.0001, + "step": 2102 + }, + { + "epoch": 0.49, + "learning_rate": 4.594984978044835e-05, + "loss": 0.0006, + "step": 2103 + }, + { + "epoch": 0.49, + "learning_rate": 4.59479238887605e-05, + "loss": 0.0022, + "step": 2104 + }, + { + "epoch": 0.49, + "learning_rate": 4.5945997997072645e-05, + "loss": 0.0044, + "step": 2105 + }, + { + "epoch": 0.49, + "learning_rate": 4.594407210538479e-05, + "loss": 0.0082, + "step": 2106 + }, + { + "epoch": 0.49, + "learning_rate": 4.594214621369695e-05, + "loss": 0.0038, + "step": 2107 + }, + { + "epoch": 0.49, + "learning_rate": 4.5940220322009095e-05, + "loss": 0.0, + "step": 2108 + }, + { + "epoch": 0.49, + "learning_rate": 4.593829443032124e-05, + "loss": 0.0019, + "step": 2109 + }, + { + "epoch": 0.49, + "learning_rate": 4.593636853863339e-05, + "loss": 0.0019, + "step": 2110 + }, + { + "epoch": 0.49, + "learning_rate": 4.593444264694554e-05, + "loss": 0.0, + "step": 2111 + }, + { + "epoch": 0.49, + "learning_rate": 4.5932516755257685e-05, + "loss": 0.0, + "step": 2112 + }, + { + "epoch": 0.49, + "learning_rate": 4.593059086356983e-05, + "loss": 0.0001, + "step": 2113 + }, + { + "epoch": 0.49, + "learning_rate": 4.5928664971881986e-05, + "loss": 0.0001, + "step": 2114 + }, + { + "epoch": 0.49, + "learning_rate": 4.5926739080194134e-05, + "loss": 0.0006, + "step": 2115 + }, + { + "epoch": 0.49, + "learning_rate": 4.592481318850628e-05, + "loss": 0.0394, + "step": 2116 + }, + { + "epoch": 0.49, + "learning_rate": 4.592288729681843e-05, + "loss": 0.0153, + "step": 2117 + }, + { + "epoch": 0.49, + "learning_rate": 4.592096140513058e-05, + "loss": 0.0027, + "step": 2118 + }, + { + "epoch": 0.49, + "learning_rate": 4.5919035513442724e-05, + "loss": 0.0, + "step": 2119 + }, + { + "epoch": 0.49, + "learning_rate": 4.591710962175487e-05, + "loss": 0.0003, + "step": 2120 + }, + { + "epoch": 0.49, + "learning_rate": 4.5915183730067025e-05, + "loss": 0.0452, + "step": 2121 + }, + { + "epoch": 0.49, + "learning_rate": 4.591325783837917e-05, + "loss": 0.0017, + "step": 2122 + }, + { + "epoch": 0.49, + "learning_rate": 4.591133194669132e-05, + "loss": 0.0014, + "step": 2123 + }, + { + "epoch": 0.49, + "learning_rate": 4.590940605500347e-05, + "loss": 0.0017, + "step": 2124 + }, + { + "epoch": 0.49, + "learning_rate": 4.590748016331562e-05, + "loss": 0.0, + "step": 2125 + }, + { + "epoch": 0.49, + "learning_rate": 4.590555427162776e-05, + "loss": 0.0015, + "step": 2126 + }, + { + "epoch": 0.49, + "learning_rate": 4.590362837993992e-05, + "loss": 0.0029, + "step": 2127 + }, + { + "epoch": 0.49, + "learning_rate": 4.5901702488252064e-05, + "loss": 0.0003, + "step": 2128 + }, + { + "epoch": 0.49, + "learning_rate": 4.589977659656421e-05, + "loss": 0.0002, + "step": 2129 + }, + { + "epoch": 0.49, + "learning_rate": 4.589785070487636e-05, + "loss": 0.0098, + "step": 2130 + }, + { + "epoch": 0.49, + "learning_rate": 4.589592481318851e-05, + "loss": 0.0007, + "step": 2131 + }, + { + "epoch": 0.49, + "learning_rate": 4.589399892150066e-05, + "loss": 0.0, + "step": 2132 + }, + { + "epoch": 0.49, + "learning_rate": 4.58920730298128e-05, + "loss": 0.0034, + "step": 2133 + }, + { + "epoch": 0.49, + "learning_rate": 4.5890147138124956e-05, + "loss": 0.0, + "step": 2134 + }, + { + "epoch": 0.49, + "learning_rate": 4.5888221246437104e-05, + "loss": 0.0, + "step": 2135 + }, + { + "epoch": 0.49, + "learning_rate": 4.588629535474925e-05, + "loss": 0.0, + "step": 2136 + }, + { + "epoch": 0.49, + "learning_rate": 4.58843694630614e-05, + "loss": 0.0001, + "step": 2137 + }, + { + "epoch": 0.49, + "learning_rate": 4.5882443571373546e-05, + "loss": 0.0002, + "step": 2138 + }, + { + "epoch": 0.49, + "learning_rate": 4.58805176796857e-05, + "loss": 0.0037, + "step": 2139 + }, + { + "epoch": 0.49, + "learning_rate": 4.587859178799784e-05, + "loss": 0.0061, + "step": 2140 + }, + { + "epoch": 0.49, + "learning_rate": 4.5876665896309995e-05, + "loss": 0.0005, + "step": 2141 + }, + { + "epoch": 0.5, + "learning_rate": 4.587474000462214e-05, + "loss": 0.0034, + "step": 2142 + }, + { + "epoch": 0.5, + "learning_rate": 4.587281411293429e-05, + "loss": 0.0025, + "step": 2143 + }, + { + "epoch": 0.5, + "learning_rate": 4.587088822124644e-05, + "loss": 0.0047, + "step": 2144 + }, + { + "epoch": 0.5, + "learning_rate": 4.586896232955859e-05, + "loss": 0.0045, + "step": 2145 + }, + { + "epoch": 0.5, + "learning_rate": 4.586703643787074e-05, + "loss": 0.0077, + "step": 2146 + }, + { + "epoch": 0.5, + "learning_rate": 4.586511054618288e-05, + "loss": 0.006, + "step": 2147 + }, + { + "epoch": 0.5, + "learning_rate": 4.5863184654495034e-05, + "loss": 0.0001, + "step": 2148 + }, + { + "epoch": 0.5, + "learning_rate": 4.586125876280718e-05, + "loss": 0.0, + "step": 2149 + }, + { + "epoch": 0.5, + "learning_rate": 4.585933287111933e-05, + "loss": 0.0001, + "step": 2150 + }, + { + "epoch": 0.5, + "learning_rate": 4.585740697943148e-05, + "loss": 0.0009, + "step": 2151 + }, + { + "epoch": 0.5, + "learning_rate": 4.585548108774363e-05, + "loss": 0.009, + "step": 2152 + }, + { + "epoch": 0.5, + "learning_rate": 4.585355519605578e-05, + "loss": 0.0007, + "step": 2153 + }, + { + "epoch": 0.5, + "learning_rate": 4.5851629304367926e-05, + "loss": 0.0001, + "step": 2154 + }, + { + "epoch": 0.5, + "learning_rate": 4.5849703412680073e-05, + "loss": 0.0131, + "step": 2155 + }, + { + "epoch": 0.5, + "learning_rate": 4.584777752099222e-05, + "loss": 0.0124, + "step": 2156 + }, + { + "epoch": 0.5, + "learning_rate": 4.584585162930437e-05, + "loss": 0.0059, + "step": 2157 + }, + { + "epoch": 0.5, + "learning_rate": 4.5843925737616516e-05, + "loss": 0.0001, + "step": 2158 + }, + { + "epoch": 0.5, + "learning_rate": 4.584199984592867e-05, + "loss": 0.0001, + "step": 2159 + }, + { + "epoch": 0.5, + "learning_rate": 4.584007395424081e-05, + "loss": 0.001, + "step": 2160 + }, + { + "epoch": 0.5, + "learning_rate": 4.5838148062552965e-05, + "loss": 0.0004, + "step": 2161 + }, + { + "epoch": 0.5, + "learning_rate": 4.583622217086511e-05, + "loss": 0.0019, + "step": 2162 + }, + { + "epoch": 0.5, + "learning_rate": 4.583429627917726e-05, + "loss": 0.0001, + "step": 2163 + }, + { + "epoch": 0.5, + "learning_rate": 4.583237038748941e-05, + "loss": 0.0035, + "step": 2164 + }, + { + "epoch": 0.5, + "learning_rate": 4.583044449580156e-05, + "loss": 0.0462, + "step": 2165 + }, + { + "epoch": 0.5, + "learning_rate": 4.582851860411371e-05, + "loss": 0.0143, + "step": 2166 + }, + { + "epoch": 0.5, + "learning_rate": 4.582659271242585e-05, + "loss": 0.0001, + "step": 2167 + }, + { + "epoch": 0.5, + "learning_rate": 4.5824666820738004e-05, + "loss": 0.0024, + "step": 2168 + }, + { + "epoch": 0.5, + "learning_rate": 4.582274092905015e-05, + "loss": 0.0, + "step": 2169 + }, + { + "epoch": 0.5, + "learning_rate": 4.58208150373623e-05, + "loss": 0.0005, + "step": 2170 + }, + { + "epoch": 0.5, + "learning_rate": 4.581888914567445e-05, + "loss": 0.0, + "step": 2171 + }, + { + "epoch": 0.5, + "learning_rate": 4.58169632539866e-05, + "loss": 0.0007, + "step": 2172 + }, + { + "epoch": 0.5, + "learning_rate": 4.581503736229875e-05, + "loss": 0.0006, + "step": 2173 + }, + { + "epoch": 0.5, + "learning_rate": 4.5813111470610896e-05, + "loss": 0.0, + "step": 2174 + }, + { + "epoch": 0.5, + "learning_rate": 4.581118557892304e-05, + "loss": 0.0017, + "step": 2175 + }, + { + "epoch": 0.5, + "learning_rate": 4.580925968723519e-05, + "loss": 0.0, + "step": 2176 + }, + { + "epoch": 0.5, + "learning_rate": 4.580733379554734e-05, + "loss": 0.0015, + "step": 2177 + }, + { + "epoch": 0.5, + "learning_rate": 4.5805407903859486e-05, + "loss": 0.0001, + "step": 2178 + }, + { + "epoch": 0.5, + "learning_rate": 4.580348201217164e-05, + "loss": 0.0162, + "step": 2179 + }, + { + "epoch": 0.5, + "learning_rate": 4.580155612048379e-05, + "loss": 0.0049, + "step": 2180 + }, + { + "epoch": 0.5, + "learning_rate": 4.5799630228795935e-05, + "loss": 0.0004, + "step": 2181 + }, + { + "epoch": 0.5, + "learning_rate": 4.579770433710808e-05, + "loss": 0.0, + "step": 2182 + }, + { + "epoch": 0.5, + "learning_rate": 4.579577844542024e-05, + "loss": 0.0003, + "step": 2183 + }, + { + "epoch": 0.5, + "learning_rate": 4.579385255373238e-05, + "loss": 0.0167, + "step": 2184 + }, + { + "epoch": 0.5, + "learning_rate": 4.5791926662044525e-05, + "loss": 0.0, + "step": 2185 + }, + { + "epoch": 0.51, + "learning_rate": 4.579000077035668e-05, + "loss": 0.0046, + "step": 2186 + }, + { + "epoch": 0.51, + "learning_rate": 4.578807487866883e-05, + "loss": 0.0054, + "step": 2187 + }, + { + "epoch": 0.51, + "learning_rate": 4.5786148986980974e-05, + "loss": 0.0001, + "step": 2188 + }, + { + "epoch": 0.51, + "learning_rate": 4.578422309529312e-05, + "loss": 0.0001, + "step": 2189 + }, + { + "epoch": 0.51, + "learning_rate": 4.5782297203605276e-05, + "loss": 0.004, + "step": 2190 + }, + { + "epoch": 0.51, + "learning_rate": 4.5780371311917417e-05, + "loss": 0.0, + "step": 2191 + }, + { + "epoch": 0.51, + "learning_rate": 4.577844542022957e-05, + "loss": 0.0002, + "step": 2192 + }, + { + "epoch": 0.51, + "learning_rate": 4.577651952854172e-05, + "loss": 0.0, + "step": 2193 + }, + { + "epoch": 0.51, + "learning_rate": 4.5774593636853866e-05, + "loss": 0.003, + "step": 2194 + }, + { + "epoch": 0.51, + "learning_rate": 4.577266774516601e-05, + "loss": 0.0014, + "step": 2195 + }, + { + "epoch": 0.51, + "learning_rate": 4.577074185347816e-05, + "loss": 0.0011, + "step": 2196 + }, + { + "epoch": 0.51, + "learning_rate": 4.5768815961790315e-05, + "loss": 0.0, + "step": 2197 + }, + { + "epoch": 0.51, + "learning_rate": 4.5766890070102456e-05, + "loss": 0.0162, + "step": 2198 + }, + { + "epoch": 0.51, + "learning_rate": 4.576496417841461e-05, + "loss": 0.0, + "step": 2199 + }, + { + "epoch": 0.51, + "learning_rate": 4.576303828672676e-05, + "loss": 0.056, + "step": 2200 + }, + { + "epoch": 0.51, + "learning_rate": 4.5761112395038905e-05, + "loss": 0.0, + "step": 2201 + }, + { + "epoch": 0.51, + "learning_rate": 4.575918650335105e-05, + "loss": 0.0099, + "step": 2202 + }, + { + "epoch": 0.51, + "learning_rate": 4.57572606116632e-05, + "loss": 0.0001, + "step": 2203 + }, + { + "epoch": 0.51, + "learning_rate": 4.5755334719975354e-05, + "loss": 0.0005, + "step": 2204 + }, + { + "epoch": 0.51, + "learning_rate": 4.5753408828287495e-05, + "loss": 0.0017, + "step": 2205 + }, + { + "epoch": 0.51, + "learning_rate": 4.575148293659965e-05, + "loss": 0.0023, + "step": 2206 + }, + { + "epoch": 0.51, + "learning_rate": 4.5749557044911797e-05, + "loss": 0.0002, + "step": 2207 + }, + { + "epoch": 0.51, + "learning_rate": 4.5747631153223944e-05, + "loss": 0.0, + "step": 2208 + }, + { + "epoch": 0.51, + "learning_rate": 4.574570526153609e-05, + "loss": 0.0164, + "step": 2209 + }, + { + "epoch": 0.51, + "learning_rate": 4.5743779369848246e-05, + "loss": 0.0005, + "step": 2210 + }, + { + "epoch": 0.51, + "learning_rate": 4.574185347816039e-05, + "loss": 0.0021, + "step": 2211 + }, + { + "epoch": 0.51, + "learning_rate": 4.5739927586472534e-05, + "loss": 0.0006, + "step": 2212 + }, + { + "epoch": 0.51, + "learning_rate": 4.573800169478469e-05, + "loss": 0.0313, + "step": 2213 + }, + { + "epoch": 0.51, + "learning_rate": 4.5736075803096836e-05, + "loss": 0.0225, + "step": 2214 + }, + { + "epoch": 0.51, + "learning_rate": 4.573414991140898e-05, + "loss": 0.0004, + "step": 2215 + }, + { + "epoch": 0.51, + "learning_rate": 4.573222401972113e-05, + "loss": 0.0039, + "step": 2216 + }, + { + "epoch": 0.51, + "learning_rate": 4.5730298128033285e-05, + "loss": 0.0025, + "step": 2217 + }, + { + "epoch": 0.51, + "learning_rate": 4.572837223634543e-05, + "loss": 0.0011, + "step": 2218 + }, + { + "epoch": 0.51, + "learning_rate": 4.572644634465758e-05, + "loss": 0.0001, + "step": 2219 + }, + { + "epoch": 0.51, + "learning_rate": 4.572452045296973e-05, + "loss": 0.0001, + "step": 2220 + }, + { + "epoch": 0.51, + "learning_rate": 4.572259456128188e-05, + "loss": 0.0025, + "step": 2221 + }, + { + "epoch": 0.51, + "learning_rate": 4.572066866959402e-05, + "loss": 0.0057, + "step": 2222 + }, + { + "epoch": 0.51, + "learning_rate": 4.571874277790617e-05, + "loss": 0.0793, + "step": 2223 + }, + { + "epoch": 0.51, + "learning_rate": 4.5716816886218324e-05, + "loss": 0.0024, + "step": 2224 + }, + { + "epoch": 0.51, + "learning_rate": 4.571489099453047e-05, + "loss": 0.0012, + "step": 2225 + }, + { + "epoch": 0.51, + "learning_rate": 4.571296510284262e-05, + "loss": 0.0002, + "step": 2226 + }, + { + "epoch": 0.51, + "learning_rate": 4.5711039211154766e-05, + "loss": 0.0, + "step": 2227 + }, + { + "epoch": 0.51, + "learning_rate": 4.570911331946692e-05, + "loss": 0.0001, + "step": 2228 + }, + { + "epoch": 0.52, + "learning_rate": 4.570718742777906e-05, + "loss": 0.0003, + "step": 2229 + }, + { + "epoch": 0.52, + "learning_rate": 4.5705261536091216e-05, + "loss": 0.0003, + "step": 2230 + }, + { + "epoch": 0.52, + "learning_rate": 4.570333564440336e-05, + "loss": 0.0, + "step": 2231 + }, + { + "epoch": 0.52, + "learning_rate": 4.5701409752715504e-05, + "loss": 0.0053, + "step": 2232 + }, + { + "epoch": 0.52, + "learning_rate": 4.569948386102766e-05, + "loss": 0.0001, + "step": 2233 + }, + { + "epoch": 0.52, + "learning_rate": 4.5697557969339806e-05, + "loss": 0.0028, + "step": 2234 + }, + { + "epoch": 0.52, + "learning_rate": 4.569563207765195e-05, + "loss": 0.0132, + "step": 2235 + }, + { + "epoch": 0.52, + "learning_rate": 4.56937061859641e-05, + "loss": 0.0018, + "step": 2236 + }, + { + "epoch": 0.52, + "learning_rate": 4.5691780294276255e-05, + "loss": 0.0, + "step": 2237 + }, + { + "epoch": 0.52, + "learning_rate": 4.56898544025884e-05, + "loss": 0.0189, + "step": 2238 + }, + { + "epoch": 0.52, + "learning_rate": 4.568792851090055e-05, + "loss": 0.0008, + "step": 2239 + }, + { + "epoch": 0.52, + "learning_rate": 4.56860026192127e-05, + "loss": 0.0146, + "step": 2240 + }, + { + "epoch": 0.52, + "learning_rate": 4.5684076727524845e-05, + "loss": 0.0383, + "step": 2241 + }, + { + "epoch": 0.52, + "learning_rate": 4.568215083583699e-05, + "loss": 0.0243, + "step": 2242 + }, + { + "epoch": 0.52, + "learning_rate": 4.568022494414914e-05, + "loss": 0.0019, + "step": 2243 + }, + { + "epoch": 0.52, + "learning_rate": 4.5678299052461294e-05, + "loss": 0.002, + "step": 2244 + }, + { + "epoch": 0.52, + "learning_rate": 4.567637316077344e-05, + "loss": 0.002, + "step": 2245 + }, + { + "epoch": 0.52, + "learning_rate": 4.567444726908559e-05, + "loss": 0.0004, + "step": 2246 + }, + { + "epoch": 0.52, + "learning_rate": 4.5672521377397736e-05, + "loss": 0.0003, + "step": 2247 + }, + { + "epoch": 0.52, + "learning_rate": 4.567059548570989e-05, + "loss": 0.0001, + "step": 2248 + }, + { + "epoch": 0.52, + "learning_rate": 4.566866959402203e-05, + "loss": 0.0028, + "step": 2249 + }, + { + "epoch": 0.52, + "learning_rate": 4.566674370233418e-05, + "loss": 0.0115, + "step": 2250 + }, + { + "epoch": 0.52, + "learning_rate": 4.566481781064633e-05, + "loss": 0.0372, + "step": 2251 + }, + { + "epoch": 0.52, + "learning_rate": 4.566289191895848e-05, + "loss": 0.001, + "step": 2252 + }, + { + "epoch": 0.52, + "learning_rate": 4.566096602727063e-05, + "loss": 0.0, + "step": 2253 + }, + { + "epoch": 0.52, + "learning_rate": 4.5659040135582775e-05, + "loss": 0.0053, + "step": 2254 + }, + { + "epoch": 0.52, + "learning_rate": 4.565711424389493e-05, + "loss": 0.0014, + "step": 2255 + }, + { + "epoch": 0.52, + "learning_rate": 4.565518835220707e-05, + "loss": 0.0, + "step": 2256 + }, + { + "epoch": 0.52, + "learning_rate": 4.5653262460519225e-05, + "loss": 0.0003, + "step": 2257 + }, + { + "epoch": 0.52, + "learning_rate": 4.565133656883137e-05, + "loss": 0.0014, + "step": 2258 + }, + { + "epoch": 0.52, + "learning_rate": 4.564941067714352e-05, + "loss": 0.0001, + "step": 2259 + }, + { + "epoch": 0.52, + "learning_rate": 4.564748478545567e-05, + "loss": 0.0024, + "step": 2260 + }, + { + "epoch": 0.52, + "learning_rate": 4.5645558893767815e-05, + "loss": 0.0, + "step": 2261 + }, + { + "epoch": 0.52, + "learning_rate": 4.564363300207997e-05, + "loss": 0.007, + "step": 2262 + }, + { + "epoch": 0.52, + "learning_rate": 4.564170711039211e-05, + "loss": 0.0118, + "step": 2263 + }, + { + "epoch": 0.52, + "learning_rate": 4.5639781218704264e-05, + "loss": 0.0018, + "step": 2264 + }, + { + "epoch": 0.52, + "learning_rate": 4.563785532701641e-05, + "loss": 0.0027, + "step": 2265 + }, + { + "epoch": 0.52, + "learning_rate": 4.563592943532856e-05, + "loss": 0.0015, + "step": 2266 + }, + { + "epoch": 0.52, + "learning_rate": 4.5634003543640706e-05, + "loss": 0.0003, + "step": 2267 + }, + { + "epoch": 0.52, + "learning_rate": 4.5632077651952854e-05, + "loss": 0.0001, + "step": 2268 + }, + { + "epoch": 0.52, + "learning_rate": 4.563015176026501e-05, + "loss": 0.0007, + "step": 2269 + }, + { + "epoch": 0.52, + "learning_rate": 4.562822586857715e-05, + "loss": 0.0075, + "step": 2270 + }, + { + "epoch": 0.52, + "learning_rate": 4.56262999768893e-05, + "loss": 0.0008, + "step": 2271 + }, + { + "epoch": 0.53, + "learning_rate": 4.562437408520145e-05, + "loss": 0.0023, + "step": 2272 + }, + { + "epoch": 0.53, + "learning_rate": 4.56224481935136e-05, + "loss": 0.0049, + "step": 2273 + }, + { + "epoch": 0.53, + "learning_rate": 4.5620522301825745e-05, + "loss": 0.0004, + "step": 2274 + }, + { + "epoch": 0.53, + "learning_rate": 4.56185964101379e-05, + "loss": 0.0136, + "step": 2275 + }, + { + "epoch": 0.53, + "learning_rate": 4.561667051845005e-05, + "loss": 0.0001, + "step": 2276 + }, + { + "epoch": 0.53, + "learning_rate": 4.5614744626762195e-05, + "loss": 0.0006, + "step": 2277 + }, + { + "epoch": 0.53, + "learning_rate": 4.561281873507434e-05, + "loss": 0.0, + "step": 2278 + }, + { + "epoch": 0.53, + "learning_rate": 4.561089284338649e-05, + "loss": 0.0, + "step": 2279 + }, + { + "epoch": 0.53, + "learning_rate": 4.560896695169864e-05, + "loss": 0.0, + "step": 2280 + }, + { + "epoch": 0.53, + "learning_rate": 4.5607041060010784e-05, + "loss": 0.0, + "step": 2281 + }, + { + "epoch": 0.53, + "learning_rate": 4.560511516832294e-05, + "loss": 0.0003, + "step": 2282 + }, + { + "epoch": 0.53, + "learning_rate": 4.5603189276635086e-05, + "loss": 0.0, + "step": 2283 + }, + { + "epoch": 0.53, + "learning_rate": 4.5601263384947234e-05, + "loss": 0.0008, + "step": 2284 + }, + { + "epoch": 0.53, + "learning_rate": 4.559933749325938e-05, + "loss": 0.0, + "step": 2285 + }, + { + "epoch": 0.53, + "learning_rate": 4.5597411601571535e-05, + "loss": 0.0018, + "step": 2286 + }, + { + "epoch": 0.53, + "learning_rate": 4.5595485709883676e-05, + "loss": 0.0063, + "step": 2287 + }, + { + "epoch": 0.53, + "learning_rate": 4.5593559818195824e-05, + "loss": 0.0003, + "step": 2288 + }, + { + "epoch": 0.53, + "learning_rate": 4.559163392650798e-05, + "loss": 0.0035, + "step": 2289 + }, + { + "epoch": 0.53, + "learning_rate": 4.5589708034820125e-05, + "loss": 0.0, + "step": 2290 + }, + { + "epoch": 0.53, + "learning_rate": 4.558778214313227e-05, + "loss": 0.0, + "step": 2291 + }, + { + "epoch": 0.53, + "learning_rate": 4.558585625144442e-05, + "loss": 0.0, + "step": 2292 + }, + { + "epoch": 0.53, + "learning_rate": 4.5583930359756575e-05, + "loss": 0.0009, + "step": 2293 + }, + { + "epoch": 0.53, + "learning_rate": 4.5582004468068715e-05, + "loss": 0.0, + "step": 2294 + }, + { + "epoch": 0.53, + "learning_rate": 4.558007857638087e-05, + "loss": 0.0014, + "step": 2295 + }, + { + "epoch": 0.53, + "learning_rate": 4.557815268469302e-05, + "loss": 0.0167, + "step": 2296 + }, + { + "epoch": 0.53, + "learning_rate": 4.5576226793005164e-05, + "loss": 0.0, + "step": 2297 + }, + { + "epoch": 0.53, + "learning_rate": 4.557430090131731e-05, + "loss": 0.0031, + "step": 2298 + }, + { + "epoch": 0.53, + "learning_rate": 4.557237500962946e-05, + "loss": 0.002, + "step": 2299 + }, + { + "epoch": 0.53, + "learning_rate": 4.5570449117941614e-05, + "loss": 0.0001, + "step": 2300 + }, + { + "epoch": 0.53, + "learning_rate": 4.5568523226253754e-05, + "loss": 0.0002, + "step": 2301 + }, + { + "epoch": 0.53, + "learning_rate": 4.556659733456591e-05, + "loss": 0.0, + "step": 2302 + }, + { + "epoch": 0.53, + "learning_rate": 4.5564671442878056e-05, + "loss": 0.0004, + "step": 2303 + }, + { + "epoch": 0.53, + "learning_rate": 4.5562745551190204e-05, + "loss": 0.0009, + "step": 2304 + }, + { + "epoch": 0.53, + "learning_rate": 4.556081965950235e-05, + "loss": 0.0, + "step": 2305 + }, + { + "epoch": 0.53, + "learning_rate": 4.55588937678145e-05, + "loss": 0.0001, + "step": 2306 + }, + { + "epoch": 0.53, + "learning_rate": 4.555696787612665e-05, + "loss": 0.0063, + "step": 2307 + }, + { + "epoch": 0.53, + "learning_rate": 4.5555041984438793e-05, + "loss": 0.0, + "step": 2308 + }, + { + "epoch": 0.53, + "learning_rate": 4.555311609275095e-05, + "loss": 0.0003, + "step": 2309 + }, + { + "epoch": 0.53, + "learning_rate": 4.5551190201063095e-05, + "loss": 0.0042, + "step": 2310 + }, + { + "epoch": 0.53, + "learning_rate": 4.554926430937524e-05, + "loss": 0.0008, + "step": 2311 + }, + { + "epoch": 0.53, + "learning_rate": 4.554733841768739e-05, + "loss": 0.0, + "step": 2312 + }, + { + "epoch": 0.53, + "learning_rate": 4.5545412525999544e-05, + "loss": 0.0013, + "step": 2313 + }, + { + "epoch": 0.53, + "learning_rate": 4.5543486634311685e-05, + "loss": 0.003, + "step": 2314 + }, + { + "epoch": 0.54, + "learning_rate": 4.554156074262383e-05, + "loss": 0.0001, + "step": 2315 + }, + { + "epoch": 0.54, + "learning_rate": 4.553963485093599e-05, + "loss": 0.0158, + "step": 2316 + }, + { + "epoch": 0.54, + "learning_rate": 4.5537708959248134e-05, + "loss": 0.0349, + "step": 2317 + }, + { + "epoch": 0.54, + "learning_rate": 4.553578306756028e-05, + "loss": 0.0017, + "step": 2318 + }, + { + "epoch": 0.54, + "learning_rate": 4.553385717587243e-05, + "loss": 0.0002, + "step": 2319 + }, + { + "epoch": 0.54, + "learning_rate": 4.5531931284184584e-05, + "loss": 0.0001, + "step": 2320 + }, + { + "epoch": 0.54, + "learning_rate": 4.5530005392496724e-05, + "loss": 0.0057, + "step": 2321 + }, + { + "epoch": 0.54, + "learning_rate": 4.552807950080888e-05, + "loss": 0.0119, + "step": 2322 + }, + { + "epoch": 0.54, + "learning_rate": 4.5526153609121026e-05, + "loss": 0.0009, + "step": 2323 + }, + { + "epoch": 0.54, + "learning_rate": 4.5524227717433173e-05, + "loss": 0.0032, + "step": 2324 + }, + { + "epoch": 0.54, + "learning_rate": 4.552230182574532e-05, + "loss": 0.0, + "step": 2325 + }, + { + "epoch": 0.54, + "learning_rate": 4.552037593405747e-05, + "loss": 0.0005, + "step": 2326 + }, + { + "epoch": 0.54, + "learning_rate": 4.551845004236962e-05, + "loss": 0.0006, + "step": 2327 + }, + { + "epoch": 0.54, + "learning_rate": 4.551652415068176e-05, + "loss": 0.0001, + "step": 2328 + }, + { + "epoch": 0.54, + "learning_rate": 4.551459825899392e-05, + "loss": 0.0035, + "step": 2329 + }, + { + "epoch": 0.54, + "learning_rate": 4.5512672367306065e-05, + "loss": 0.0, + "step": 2330 + }, + { + "epoch": 0.54, + "learning_rate": 4.551074647561821e-05, + "loss": 0.0006, + "step": 2331 + }, + { + "epoch": 0.54, + "learning_rate": 4.550882058393036e-05, + "loss": 0.0024, + "step": 2332 + }, + { + "epoch": 0.54, + "learning_rate": 4.5506894692242514e-05, + "loss": 0.003, + "step": 2333 + }, + { + "epoch": 0.54, + "learning_rate": 4.550496880055466e-05, + "loss": 0.0, + "step": 2334 + }, + { + "epoch": 0.54, + "learning_rate": 4.55030429088668e-05, + "loss": 0.0068, + "step": 2335 + }, + { + "epoch": 0.54, + "learning_rate": 4.550111701717896e-05, + "loss": 0.0, + "step": 2336 + }, + { + "epoch": 0.54, + "learning_rate": 4.5499191125491104e-05, + "loss": 0.0, + "step": 2337 + }, + { + "epoch": 0.54, + "learning_rate": 4.549726523380325e-05, + "loss": 0.0048, + "step": 2338 + }, + { + "epoch": 0.54, + "learning_rate": 4.54953393421154e-05, + "loss": 0.0017, + "step": 2339 + }, + { + "epoch": 0.54, + "learning_rate": 4.5493413450427553e-05, + "loss": 0.0023, + "step": 2340 + }, + { + "epoch": 0.54, + "learning_rate": 4.54914875587397e-05, + "loss": 0.0013, + "step": 2341 + }, + { + "epoch": 0.54, + "learning_rate": 4.548956166705185e-05, + "loss": 0.0, + "step": 2342 + }, + { + "epoch": 0.54, + "learning_rate": 4.5487635775363996e-05, + "loss": 0.0, + "step": 2343 + }, + { + "epoch": 0.54, + "learning_rate": 4.548570988367614e-05, + "loss": 0.0017, + "step": 2344 + }, + { + "epoch": 0.54, + "learning_rate": 4.548378399198829e-05, + "loss": 0.0006, + "step": 2345 + }, + { + "epoch": 0.54, + "learning_rate": 4.548185810030044e-05, + "loss": 0.0006, + "step": 2346 + }, + { + "epoch": 0.54, + "learning_rate": 4.547993220861259e-05, + "loss": 0.0002, + "step": 2347 + }, + { + "epoch": 0.54, + "learning_rate": 4.547800631692474e-05, + "loss": 0.001, + "step": 2348 + }, + { + "epoch": 0.54, + "learning_rate": 4.547608042523689e-05, + "loss": 0.0, + "step": 2349 + }, + { + "epoch": 0.54, + "learning_rate": 4.5474154533549035e-05, + "loss": 0.0, + "step": 2350 + }, + { + "epoch": 0.54, + "learning_rate": 4.547222864186119e-05, + "loss": 0.0001, + "step": 2351 + }, + { + "epoch": 0.54, + "learning_rate": 4.547030275017333e-05, + "loss": 0.0078, + "step": 2352 + }, + { + "epoch": 0.54, + "learning_rate": 4.546837685848548e-05, + "loss": 0.0002, + "step": 2353 + }, + { + "epoch": 0.54, + "learning_rate": 4.546645096679763e-05, + "loss": 0.0002, + "step": 2354 + }, + { + "epoch": 0.54, + "learning_rate": 4.546452507510978e-05, + "loss": 0.0002, + "step": 2355 + }, + { + "epoch": 0.54, + "learning_rate": 4.5462599183421927e-05, + "loss": 0.0139, + "step": 2356 + }, + { + "epoch": 0.54, + "learning_rate": 4.5460673291734074e-05, + "loss": 0.0, + "step": 2357 + }, + { + "epoch": 0.54, + "learning_rate": 4.545874740004623e-05, + "loss": 0.0003, + "step": 2358 + }, + { + "epoch": 0.55, + "learning_rate": 4.545682150835837e-05, + "loss": 0.0068, + "step": 2359 + }, + { + "epoch": 0.55, + "learning_rate": 4.545489561667052e-05, + "loss": 0.0787, + "step": 2360 + }, + { + "epoch": 0.55, + "learning_rate": 4.545296972498267e-05, + "loss": 0.0003, + "step": 2361 + }, + { + "epoch": 0.55, + "learning_rate": 4.545104383329482e-05, + "loss": 0.0004, + "step": 2362 + }, + { + "epoch": 0.55, + "learning_rate": 4.5449117941606966e-05, + "loss": 0.0046, + "step": 2363 + }, + { + "epoch": 0.55, + "learning_rate": 4.544719204991911e-05, + "loss": 0.0003, + "step": 2364 + }, + { + "epoch": 0.55, + "learning_rate": 4.544526615823127e-05, + "loss": 0.0001, + "step": 2365 + }, + { + "epoch": 0.55, + "learning_rate": 4.544334026654341e-05, + "loss": 0.0048, + "step": 2366 + }, + { + "epoch": 0.55, + "learning_rate": 4.544141437485556e-05, + "loss": 0.0054, + "step": 2367 + }, + { + "epoch": 0.55, + "learning_rate": 4.543948848316771e-05, + "loss": 0.0002, + "step": 2368 + }, + { + "epoch": 0.55, + "learning_rate": 4.543756259147986e-05, + "loss": 0.0001, + "step": 2369 + }, + { + "epoch": 0.55, + "learning_rate": 4.5435636699792005e-05, + "loss": 0.0008, + "step": 2370 + }, + { + "epoch": 0.55, + "learning_rate": 4.543371080810415e-05, + "loss": 0.0017, + "step": 2371 + }, + { + "epoch": 0.55, + "learning_rate": 4.5431784916416307e-05, + "loss": 0.0052, + "step": 2372 + }, + { + "epoch": 0.55, + "learning_rate": 4.542985902472845e-05, + "loss": 0.0, + "step": 2373 + }, + { + "epoch": 0.55, + "learning_rate": 4.54279331330406e-05, + "loss": 0.0043, + "step": 2374 + }, + { + "epoch": 0.55, + "learning_rate": 4.542600724135275e-05, + "loss": 0.0023, + "step": 2375 + }, + { + "epoch": 0.55, + "learning_rate": 4.5424081349664896e-05, + "loss": 0.0002, + "step": 2376 + }, + { + "epoch": 0.55, + "learning_rate": 4.5422155457977044e-05, + "loss": 0.0129, + "step": 2377 + }, + { + "epoch": 0.55, + "learning_rate": 4.54202295662892e-05, + "loss": 0.0032, + "step": 2378 + }, + { + "epoch": 0.55, + "learning_rate": 4.5418303674601346e-05, + "loss": 0.0003, + "step": 2379 + }, + { + "epoch": 0.55, + "learning_rate": 4.5416377782913486e-05, + "loss": 0.0, + "step": 2380 + }, + { + "epoch": 0.55, + "learning_rate": 4.541445189122564e-05, + "loss": 0.0029, + "step": 2381 + }, + { + "epoch": 0.55, + "learning_rate": 4.541252599953779e-05, + "loss": 0.0024, + "step": 2382 + }, + { + "epoch": 0.55, + "learning_rate": 4.5410600107849936e-05, + "loss": 0.0029, + "step": 2383 + }, + { + "epoch": 0.55, + "learning_rate": 4.540867421616208e-05, + "loss": 0.012, + "step": 2384 + }, + { + "epoch": 0.55, + "learning_rate": 4.540674832447424e-05, + "loss": 0.0017, + "step": 2385 + }, + { + "epoch": 0.55, + "learning_rate": 4.540482243278638e-05, + "loss": 0.0013, + "step": 2386 + }, + { + "epoch": 0.55, + "learning_rate": 4.540289654109853e-05, + "loss": 0.0182, + "step": 2387 + }, + { + "epoch": 0.55, + "learning_rate": 4.540097064941068e-05, + "loss": 0.0, + "step": 2388 + }, + { + "epoch": 0.55, + "learning_rate": 4.539904475772283e-05, + "loss": 0.0002, + "step": 2389 + }, + { + "epoch": 0.55, + "learning_rate": 4.5397118866034975e-05, + "loss": 0.0, + "step": 2390 + }, + { + "epoch": 0.55, + "learning_rate": 4.539519297434712e-05, + "loss": 0.0002, + "step": 2391 + }, + { + "epoch": 0.55, + "learning_rate": 4.5393267082659276e-05, + "loss": 0.0006, + "step": 2392 + }, + { + "epoch": 0.55, + "learning_rate": 4.539134119097142e-05, + "loss": 0.0, + "step": 2393 + }, + { + "epoch": 0.55, + "learning_rate": 4.538941529928357e-05, + "loss": 0.0, + "step": 2394 + }, + { + "epoch": 0.55, + "learning_rate": 4.538748940759572e-05, + "loss": 0.0048, + "step": 2395 + }, + { + "epoch": 0.55, + "learning_rate": 4.5385563515907866e-05, + "loss": 0.0002, + "step": 2396 + }, + { + "epoch": 0.55, + "learning_rate": 4.5383637624220014e-05, + "loss": 0.0017, + "step": 2397 + }, + { + "epoch": 0.55, + "learning_rate": 4.538171173253217e-05, + "loss": 0.0, + "step": 2398 + }, + { + "epoch": 0.55, + "learning_rate": 4.5379785840844316e-05, + "loss": 0.0003, + "step": 2399 + }, + { + "epoch": 0.55, + "learning_rate": 4.5377859949156456e-05, + "loss": 0.0111, + "step": 2400 + }, + { + "epoch": 0.55, + "learning_rate": 4.537593405746861e-05, + "loss": 0.0002, + "step": 2401 + }, + { + "epoch": 0.56, + "learning_rate": 4.537400816578076e-05, + "loss": 0.0049, + "step": 2402 + }, + { + "epoch": 0.56, + "learning_rate": 4.5372082274092905e-05, + "loss": 0.0141, + "step": 2403 + }, + { + "epoch": 0.56, + "learning_rate": 4.537015638240505e-05, + "loss": 0.0009, + "step": 2404 + }, + { + "epoch": 0.56, + "learning_rate": 4.536823049071721e-05, + "loss": 0.0259, + "step": 2405 + }, + { + "epoch": 0.56, + "learning_rate": 4.5366304599029355e-05, + "loss": 0.001, + "step": 2406 + }, + { + "epoch": 0.56, + "learning_rate": 4.53643787073415e-05, + "loss": 0.0009, + "step": 2407 + }, + { + "epoch": 0.56, + "learning_rate": 4.536245281565365e-05, + "loss": 0.0027, + "step": 2408 + }, + { + "epoch": 0.56, + "learning_rate": 4.53605269239658e-05, + "loss": 0.0001, + "step": 2409 + }, + { + "epoch": 0.56, + "learning_rate": 4.5358601032277945e-05, + "loss": 0.0, + "step": 2410 + }, + { + "epoch": 0.56, + "learning_rate": 4.535667514059009e-05, + "loss": 0.0031, + "step": 2411 + }, + { + "epoch": 0.56, + "learning_rate": 4.5354749248902246e-05, + "loss": 0.0006, + "step": 2412 + }, + { + "epoch": 0.56, + "learning_rate": 4.5352823357214394e-05, + "loss": 0.0004, + "step": 2413 + }, + { + "epoch": 0.56, + "learning_rate": 4.535089746552654e-05, + "loss": 0.0019, + "step": 2414 + }, + { + "epoch": 0.56, + "learning_rate": 4.534897157383869e-05, + "loss": 0.0002, + "step": 2415 + }, + { + "epoch": 0.56, + "learning_rate": 4.534704568215084e-05, + "loss": 0.0088, + "step": 2416 + }, + { + "epoch": 0.56, + "learning_rate": 4.5345119790462984e-05, + "loss": 0.0002, + "step": 2417 + }, + { + "epoch": 0.56, + "learning_rate": 4.534319389877513e-05, + "loss": 0.0014, + "step": 2418 + }, + { + "epoch": 0.56, + "learning_rate": 4.5341268007087285e-05, + "loss": 0.0014, + "step": 2419 + }, + { + "epoch": 0.56, + "learning_rate": 4.533934211539943e-05, + "loss": 0.0, + "step": 2420 + }, + { + "epoch": 0.56, + "learning_rate": 4.533741622371158e-05, + "loss": 0.0, + "step": 2421 + }, + { + "epoch": 0.56, + "learning_rate": 4.533549033202373e-05, + "loss": 0.0139, + "step": 2422 + }, + { + "epoch": 0.56, + "learning_rate": 4.533356444033588e-05, + "loss": 0.0049, + "step": 2423 + }, + { + "epoch": 0.56, + "learning_rate": 4.533163854864802e-05, + "loss": 0.0003, + "step": 2424 + }, + { + "epoch": 0.56, + "learning_rate": 4.532971265696018e-05, + "loss": 0.0043, + "step": 2425 + }, + { + "epoch": 0.56, + "learning_rate": 4.5327786765272325e-05, + "loss": 0.0001, + "step": 2426 + }, + { + "epoch": 0.56, + "learning_rate": 4.532586087358447e-05, + "loss": 0.0004, + "step": 2427 + }, + { + "epoch": 0.56, + "learning_rate": 4.532393498189662e-05, + "loss": 0.0172, + "step": 2428 + }, + { + "epoch": 0.56, + "learning_rate": 4.532200909020877e-05, + "loss": 0.0, + "step": 2429 + }, + { + "epoch": 0.56, + "learning_rate": 4.532008319852092e-05, + "loss": 0.0, + "step": 2430 + }, + { + "epoch": 0.56, + "learning_rate": 4.531815730683306e-05, + "loss": 0.0024, + "step": 2431 + }, + { + "epoch": 0.56, + "learning_rate": 4.5316231415145216e-05, + "loss": 0.0, + "step": 2432 + }, + { + "epoch": 0.56, + "learning_rate": 4.5314305523457364e-05, + "loss": 0.0003, + "step": 2433 + }, + { + "epoch": 0.56, + "learning_rate": 4.531237963176951e-05, + "loss": 0.0002, + "step": 2434 + }, + { + "epoch": 0.56, + "learning_rate": 4.531045374008166e-05, + "loss": 0.0005, + "step": 2435 + }, + { + "epoch": 0.56, + "learning_rate": 4.5308527848393806e-05, + "loss": 0.0001, + "step": 2436 + }, + { + "epoch": 0.56, + "learning_rate": 4.530660195670596e-05, + "loss": 0.0, + "step": 2437 + }, + { + "epoch": 0.56, + "learning_rate": 4.53046760650181e-05, + "loss": 0.0029, + "step": 2438 + }, + { + "epoch": 0.56, + "learning_rate": 4.5302750173330255e-05, + "loss": 0.0041, + "step": 2439 + }, + { + "epoch": 0.56, + "learning_rate": 4.53008242816424e-05, + "loss": 0.0003, + "step": 2440 + }, + { + "epoch": 0.56, + "learning_rate": 4.529889838995455e-05, + "loss": 0.0, + "step": 2441 + }, + { + "epoch": 0.56, + "learning_rate": 4.52969724982667e-05, + "loss": 0.0003, + "step": 2442 + }, + { + "epoch": 0.56, + "learning_rate": 4.529504660657885e-05, + "loss": 0.0, + "step": 2443 + }, + { + "epoch": 0.56, + "learning_rate": 4.5293120714891e-05, + "loss": 0.0005, + "step": 2444 + }, + { + "epoch": 0.57, + "learning_rate": 4.529119482320315e-05, + "loss": 0.0, + "step": 2445 + }, + { + "epoch": 0.57, + "learning_rate": 4.5289268931515294e-05, + "loss": 0.0003, + "step": 2446 + }, + { + "epoch": 0.57, + "learning_rate": 4.528734303982744e-05, + "loss": 0.0087, + "step": 2447 + }, + { + "epoch": 0.57, + "learning_rate": 4.528541714813959e-05, + "loss": 0.0009, + "step": 2448 + }, + { + "epoch": 0.57, + "learning_rate": 4.528349125645174e-05, + "loss": 0.0122, + "step": 2449 + }, + { + "epoch": 0.57, + "learning_rate": 4.528156536476389e-05, + "loss": 0.0007, + "step": 2450 + }, + { + "epoch": 0.57, + "learning_rate": 4.527963947307604e-05, + "loss": 0.0, + "step": 2451 + }, + { + "epoch": 0.57, + "learning_rate": 4.5277713581388186e-05, + "loss": 0.0, + "step": 2452 + }, + { + "epoch": 0.57, + "learning_rate": 4.5275787689700334e-05, + "loss": 0.0, + "step": 2453 + }, + { + "epoch": 0.57, + "learning_rate": 4.527386179801249e-05, + "loss": 0.0022, + "step": 2454 + }, + { + "epoch": 0.57, + "learning_rate": 4.527193590632463e-05, + "loss": 0.0068, + "step": 2455 + }, + { + "epoch": 0.57, + "learning_rate": 4.5270010014636776e-05, + "loss": 0.0025, + "step": 2456 + }, + { + "epoch": 0.57, + "learning_rate": 4.526808412294893e-05, + "loss": 0.0016, + "step": 2457 + }, + { + "epoch": 0.57, + "learning_rate": 4.526615823126107e-05, + "loss": 0.0008, + "step": 2458 + }, + { + "epoch": 0.57, + "learning_rate": 4.5264232339573225e-05, + "loss": 0.0, + "step": 2459 + }, + { + "epoch": 0.57, + "learning_rate": 4.526230644788537e-05, + "loss": 0.0003, + "step": 2460 + }, + { + "epoch": 0.57, + "learning_rate": 4.526038055619753e-05, + "loss": 0.0, + "step": 2461 + }, + { + "epoch": 0.57, + "learning_rate": 4.525845466450967e-05, + "loss": 0.0038, + "step": 2462 + }, + { + "epoch": 0.57, + "learning_rate": 4.525652877282182e-05, + "loss": 0.0004, + "step": 2463 + }, + { + "epoch": 0.57, + "learning_rate": 4.525460288113397e-05, + "loss": 0.0, + "step": 2464 + }, + { + "epoch": 0.57, + "learning_rate": 4.525267698944611e-05, + "loss": 0.001, + "step": 2465 + }, + { + "epoch": 0.57, + "learning_rate": 4.5250751097758264e-05, + "loss": 0.0, + "step": 2466 + }, + { + "epoch": 0.57, + "learning_rate": 4.524882520607041e-05, + "loss": 0.0001, + "step": 2467 + }, + { + "epoch": 0.57, + "learning_rate": 4.524689931438256e-05, + "loss": 0.0, + "step": 2468 + }, + { + "epoch": 0.57, + "learning_rate": 4.524497342269471e-05, + "loss": 0.0006, + "step": 2469 + }, + { + "epoch": 0.57, + "learning_rate": 4.524304753100686e-05, + "loss": 0.0, + "step": 2470 + }, + { + "epoch": 0.57, + "learning_rate": 4.524112163931901e-05, + "loss": 0.0068, + "step": 2471 + }, + { + "epoch": 0.57, + "learning_rate": 4.5239195747631156e-05, + "loss": 0.0005, + "step": 2472 + }, + { + "epoch": 0.57, + "learning_rate": 4.5237269855943303e-05, + "loss": 0.0, + "step": 2473 + }, + { + "epoch": 0.57, + "learning_rate": 4.523534396425545e-05, + "loss": 0.0009, + "step": 2474 + }, + { + "epoch": 0.57, + "learning_rate": 4.52334180725676e-05, + "loss": 0.0, + "step": 2475 + }, + { + "epoch": 0.57, + "learning_rate": 4.5231492180879746e-05, + "loss": 0.0, + "step": 2476 + }, + { + "epoch": 0.57, + "learning_rate": 4.52295662891919e-05, + "loss": 0.0, + "step": 2477 + }, + { + "epoch": 0.57, + "learning_rate": 4.522764039750405e-05, + "loss": 0.0124, + "step": 2478 + }, + { + "epoch": 0.57, + "learning_rate": 4.5225714505816195e-05, + "loss": 0.0002, + "step": 2479 + }, + { + "epoch": 0.57, + "learning_rate": 4.522378861412834e-05, + "loss": 0.0298, + "step": 2480 + }, + { + "epoch": 0.57, + "learning_rate": 4.52218627224405e-05, + "loss": 0.0002, + "step": 2481 + }, + { + "epoch": 0.57, + "learning_rate": 4.521993683075264e-05, + "loss": 0.0022, + "step": 2482 + }, + { + "epoch": 0.57, + "learning_rate": 4.5218010939064785e-05, + "loss": 0.0025, + "step": 2483 + }, + { + "epoch": 0.57, + "learning_rate": 4.521608504737694e-05, + "loss": 0.0, + "step": 2484 + }, + { + "epoch": 0.57, + "learning_rate": 4.521415915568909e-05, + "loss": 0.0, + "step": 2485 + }, + { + "epoch": 0.57, + "learning_rate": 4.5212233264001234e-05, + "loss": 0.001, + "step": 2486 + }, + { + "epoch": 0.57, + "learning_rate": 4.521030737231338e-05, + "loss": 0.0, + "step": 2487 + }, + { + "epoch": 0.57, + "learning_rate": 4.5208381480625536e-05, + "loss": 0.0005, + "step": 2488 + }, + { + "epoch": 0.58, + "learning_rate": 4.520645558893768e-05, + "loss": 0.0001, + "step": 2489 + }, + { + "epoch": 0.58, + "learning_rate": 4.520452969724983e-05, + "loss": 0.0009, + "step": 2490 + }, + { + "epoch": 0.58, + "learning_rate": 4.520260380556198e-05, + "loss": 0.0053, + "step": 2491 + }, + { + "epoch": 0.58, + "learning_rate": 4.5200677913874126e-05, + "loss": 0.0, + "step": 2492 + }, + { + "epoch": 0.58, + "learning_rate": 4.519875202218627e-05, + "loss": 0.0002, + "step": 2493 + }, + { + "epoch": 0.58, + "learning_rate": 4.519682613049842e-05, + "loss": 0.0009, + "step": 2494 + }, + { + "epoch": 0.58, + "learning_rate": 4.5194900238810575e-05, + "loss": 0.0012, + "step": 2495 + }, + { + "epoch": 0.58, + "learning_rate": 4.5192974347122716e-05, + "loss": 0.0046, + "step": 2496 + }, + { + "epoch": 0.58, + "learning_rate": 4.519104845543487e-05, + "loss": 0.0001, + "step": 2497 + }, + { + "epoch": 0.58, + "learning_rate": 4.518912256374702e-05, + "loss": 0.0019, + "step": 2498 + }, + { + "epoch": 0.58, + "learning_rate": 4.5187196672059165e-05, + "loss": 0.0006, + "step": 2499 + }, + { + "epoch": 0.58, + "learning_rate": 4.518527078037131e-05, + "loss": 0.0003, + "step": 2500 + }, + { + "epoch": 0.58, + "learning_rate": 4.518334488868347e-05, + "loss": 0.0039, + "step": 2501 + }, + { + "epoch": 0.58, + "learning_rate": 4.5181418996995614e-05, + "loss": 0.0, + "step": 2502 + }, + { + "epoch": 0.58, + "learning_rate": 4.5179493105307755e-05, + "loss": 0.0, + "step": 2503 + }, + { + "epoch": 0.58, + "learning_rate": 4.517756721361991e-05, + "loss": 0.0065, + "step": 2504 + }, + { + "epoch": 0.58, + "learning_rate": 4.517564132193206e-05, + "loss": 0.027, + "step": 2505 + }, + { + "epoch": 0.58, + "learning_rate": 4.5173715430244204e-05, + "loss": 0.0016, + "step": 2506 + }, + { + "epoch": 0.58, + "learning_rate": 4.517178953855635e-05, + "loss": 0.0001, + "step": 2507 + }, + { + "epoch": 0.58, + "learning_rate": 4.5169863646868506e-05, + "loss": 0.0023, + "step": 2508 + }, + { + "epoch": 0.58, + "learning_rate": 4.516793775518065e-05, + "loss": 0.0132, + "step": 2509 + }, + { + "epoch": 0.58, + "learning_rate": 4.51660118634928e-05, + "loss": 0.0036, + "step": 2510 + }, + { + "epoch": 0.58, + "learning_rate": 4.516408597180495e-05, + "loss": 0.0013, + "step": 2511 + }, + { + "epoch": 0.58, + "learning_rate": 4.5162160080117096e-05, + "loss": 0.0001, + "step": 2512 + }, + { + "epoch": 0.58, + "learning_rate": 4.516023418842924e-05, + "loss": 0.006, + "step": 2513 + }, + { + "epoch": 0.58, + "learning_rate": 4.515830829674139e-05, + "loss": 0.0, + "step": 2514 + }, + { + "epoch": 0.58, + "learning_rate": 4.5156382405053545e-05, + "loss": 0.0011, + "step": 2515 + }, + { + "epoch": 0.58, + "learning_rate": 4.515445651336569e-05, + "loss": 0.0, + "step": 2516 + }, + { + "epoch": 0.58, + "learning_rate": 4.515253062167784e-05, + "loss": 0.0037, + "step": 2517 + }, + { + "epoch": 0.58, + "learning_rate": 4.515060472998999e-05, + "loss": 0.0322, + "step": 2518 + }, + { + "epoch": 0.58, + "learning_rate": 4.514867883830214e-05, + "loss": 0.0004, + "step": 2519 + }, + { + "epoch": 0.58, + "learning_rate": 4.514675294661428e-05, + "loss": 0.0, + "step": 2520 + }, + { + "epoch": 0.58, + "learning_rate": 4.514482705492643e-05, + "loss": 0.0, + "step": 2521 + }, + { + "epoch": 0.58, + "learning_rate": 4.5142901163238584e-05, + "loss": 0.0006, + "step": 2522 + }, + { + "epoch": 0.58, + "learning_rate": 4.514097527155073e-05, + "loss": 0.0003, + "step": 2523 + }, + { + "epoch": 0.58, + "learning_rate": 4.513904937986288e-05, + "loss": 0.0027, + "step": 2524 + }, + { + "epoch": 0.58, + "learning_rate": 4.5137123488175027e-05, + "loss": 0.0021, + "step": 2525 + }, + { + "epoch": 0.58, + "learning_rate": 4.513519759648718e-05, + "loss": 0.0, + "step": 2526 + }, + { + "epoch": 0.58, + "learning_rate": 4.513327170479932e-05, + "loss": 0.0004, + "step": 2527 + }, + { + "epoch": 0.58, + "learning_rate": 4.5131345813111476e-05, + "loss": 0.0, + "step": 2528 + }, + { + "epoch": 0.58, + "learning_rate": 4.512941992142362e-05, + "loss": 0.0007, + "step": 2529 + }, + { + "epoch": 0.58, + "learning_rate": 4.512749402973577e-05, + "loss": 0.0005, + "step": 2530 + }, + { + "epoch": 0.58, + "learning_rate": 4.512556813804792e-05, + "loss": 0.0164, + "step": 2531 + }, + { + "epoch": 0.59, + "learning_rate": 4.5123642246360066e-05, + "loss": 0.0, + "step": 2532 + }, + { + "epoch": 0.59, + "learning_rate": 4.512171635467222e-05, + "loss": 0.0031, + "step": 2533 + }, + { + "epoch": 0.59, + "learning_rate": 4.511979046298436e-05, + "loss": 0.0007, + "step": 2534 + }, + { + "epoch": 0.59, + "learning_rate": 4.5117864571296515e-05, + "loss": 0.0, + "step": 2535 + }, + { + "epoch": 0.59, + "learning_rate": 4.511593867960866e-05, + "loss": 0.0213, + "step": 2536 + }, + { + "epoch": 0.59, + "learning_rate": 4.511401278792081e-05, + "loss": 0.0, + "step": 2537 + }, + { + "epoch": 0.59, + "learning_rate": 4.511208689623296e-05, + "loss": 0.0, + "step": 2538 + }, + { + "epoch": 0.59, + "learning_rate": 4.5110161004545105e-05, + "loss": 0.0002, + "step": 2539 + }, + { + "epoch": 0.59, + "learning_rate": 4.510823511285725e-05, + "loss": 0.0, + "step": 2540 + }, + { + "epoch": 0.59, + "learning_rate": 4.51063092211694e-05, + "loss": 0.0, + "step": 2541 + }, + { + "epoch": 0.59, + "learning_rate": 4.5104383329481554e-05, + "loss": 0.0511, + "step": 2542 + }, + { + "epoch": 0.59, + "learning_rate": 4.51024574377937e-05, + "loss": 0.0001, + "step": 2543 + }, + { + "epoch": 0.59, + "learning_rate": 4.510053154610585e-05, + "loss": 0.0, + "step": 2544 + }, + { + "epoch": 0.59, + "learning_rate": 4.5098605654417996e-05, + "loss": 0.0002, + "step": 2545 + }, + { + "epoch": 0.59, + "learning_rate": 4.509667976273015e-05, + "loss": 0.0, + "step": 2546 + }, + { + "epoch": 0.59, + "learning_rate": 4.509475387104229e-05, + "loss": 0.0256, + "step": 2547 + }, + { + "epoch": 0.59, + "learning_rate": 4.509282797935444e-05, + "loss": 0.0062, + "step": 2548 + }, + { + "epoch": 0.59, + "learning_rate": 4.509090208766659e-05, + "loss": 0.0031, + "step": 2549 + }, + { + "epoch": 0.59, + "learning_rate": 4.508897619597874e-05, + "loss": 0.0063, + "step": 2550 + }, + { + "epoch": 0.59, + "learning_rate": 4.508705030429089e-05, + "loss": 0.0006, + "step": 2551 + }, + { + "epoch": 0.59, + "learning_rate": 4.5085124412603036e-05, + "loss": 0.0, + "step": 2552 + }, + { + "epoch": 0.59, + "learning_rate": 4.508319852091519e-05, + "loss": 0.039, + "step": 2553 + }, + { + "epoch": 0.59, + "learning_rate": 4.508127262922733e-05, + "loss": 0.0, + "step": 2554 + }, + { + "epoch": 0.59, + "learning_rate": 4.5079346737539485e-05, + "loss": 0.0015, + "step": 2555 + }, + { + "epoch": 0.59, + "learning_rate": 4.507742084585163e-05, + "loss": 0.0, + "step": 2556 + }, + { + "epoch": 0.59, + "learning_rate": 4.507549495416378e-05, + "loss": 0.0, + "step": 2557 + }, + { + "epoch": 0.59, + "learning_rate": 4.507356906247593e-05, + "loss": 0.0126, + "step": 2558 + }, + { + "epoch": 0.59, + "learning_rate": 4.5071643170788075e-05, + "loss": 0.0005, + "step": 2559 + }, + { + "epoch": 0.59, + "learning_rate": 4.506971727910023e-05, + "loss": 0.0013, + "step": 2560 + }, + { + "epoch": 0.59, + "learning_rate": 4.506779138741237e-05, + "loss": 0.0224, + "step": 2561 + }, + { + "epoch": 0.59, + "learning_rate": 4.5065865495724524e-05, + "loss": 0.0, + "step": 2562 + }, + { + "epoch": 0.59, + "learning_rate": 4.506393960403667e-05, + "loss": 0.0008, + "step": 2563 + }, + { + "epoch": 0.59, + "learning_rate": 4.506201371234882e-05, + "loss": 0.0007, + "step": 2564 + }, + { + "epoch": 0.59, + "learning_rate": 4.5060087820660966e-05, + "loss": 0.0009, + "step": 2565 + }, + { + "epoch": 0.59, + "learning_rate": 4.505816192897312e-05, + "loss": 0.001, + "step": 2566 + }, + { + "epoch": 0.59, + "learning_rate": 4.505623603728527e-05, + "loss": 0.0, + "step": 2567 + }, + { + "epoch": 0.59, + "learning_rate": 4.505431014559741e-05, + "loss": 0.0002, + "step": 2568 + }, + { + "epoch": 0.59, + "learning_rate": 4.505238425390956e-05, + "loss": 0.0082, + "step": 2569 + }, + { + "epoch": 0.59, + "learning_rate": 4.505045836222171e-05, + "loss": 0.0015, + "step": 2570 + }, + { + "epoch": 0.59, + "learning_rate": 4.504853247053386e-05, + "loss": 0.0006, + "step": 2571 + }, + { + "epoch": 0.59, + "learning_rate": 4.5046606578846005e-05, + "loss": 0.0007, + "step": 2572 + }, + { + "epoch": 0.59, + "learning_rate": 4.504468068715816e-05, + "loss": 0.0, + "step": 2573 + }, + { + "epoch": 0.59, + "learning_rate": 4.504275479547031e-05, + "loss": 0.0002, + "step": 2574 + }, + { + "epoch": 0.6, + "learning_rate": 4.5040828903782455e-05, + "loss": 0.0079, + "step": 2575 + }, + { + "epoch": 0.6, + "learning_rate": 4.50389030120946e-05, + "loss": 0.0, + "step": 2576 + }, + { + "epoch": 0.6, + "learning_rate": 4.503697712040675e-05, + "loss": 0.0, + "step": 2577 + }, + { + "epoch": 0.6, + "learning_rate": 4.50350512287189e-05, + "loss": 0.0, + "step": 2578 + }, + { + "epoch": 0.6, + "learning_rate": 4.5033125337031045e-05, + "loss": 0.0006, + "step": 2579 + }, + { + "epoch": 0.6, + "learning_rate": 4.50311994453432e-05, + "loss": 0.0, + "step": 2580 + }, + { + "epoch": 0.6, + "learning_rate": 4.5029273553655346e-05, + "loss": 0.0182, + "step": 2581 + }, + { + "epoch": 0.6, + "learning_rate": 4.5027347661967494e-05, + "loss": 0.0002, + "step": 2582 + }, + { + "epoch": 0.6, + "learning_rate": 4.502542177027964e-05, + "loss": 0.0, + "step": 2583 + }, + { + "epoch": 0.6, + "learning_rate": 4.5023495878591795e-05, + "loss": 0.0058, + "step": 2584 + }, + { + "epoch": 0.6, + "learning_rate": 4.5021569986903936e-05, + "loss": 0.0015, + "step": 2585 + }, + { + "epoch": 0.6, + "learning_rate": 4.5019644095216084e-05, + "loss": 0.0016, + "step": 2586 + }, + { + "epoch": 0.6, + "learning_rate": 4.501771820352824e-05, + "loss": 0.0, + "step": 2587 + }, + { + "epoch": 0.6, + "learning_rate": 4.5015792311840385e-05, + "loss": 0.0003, + "step": 2588 + }, + { + "epoch": 0.6, + "learning_rate": 4.501386642015253e-05, + "loss": 0.0005, + "step": 2589 + }, + { + "epoch": 0.6, + "learning_rate": 4.501194052846468e-05, + "loss": 0.0001, + "step": 2590 + }, + { + "epoch": 0.6, + "learning_rate": 4.5010014636776835e-05, + "loss": 0.0, + "step": 2591 + }, + { + "epoch": 0.6, + "learning_rate": 4.5008088745088975e-05, + "loss": 0.0024, + "step": 2592 + }, + { + "epoch": 0.6, + "learning_rate": 4.500616285340113e-05, + "loss": 0.0001, + "step": 2593 + }, + { + "epoch": 0.6, + "learning_rate": 4.500423696171328e-05, + "loss": 0.0022, + "step": 2594 + }, + { + "epoch": 0.6, + "learning_rate": 4.5002311070025425e-05, + "loss": 0.0005, + "step": 2595 + }, + { + "epoch": 0.6, + "learning_rate": 4.500038517833757e-05, + "loss": 0.001, + "step": 2596 + }, + { + "epoch": 0.6, + "learning_rate": 4.499845928664972e-05, + "loss": 0.0001, + "step": 2597 + }, + { + "epoch": 0.6, + "learning_rate": 4.4996533394961874e-05, + "loss": 0.0, + "step": 2598 + }, + { + "epoch": 0.6, + "learning_rate": 4.4994607503274014e-05, + "loss": 0.0, + "step": 2599 + }, + { + "epoch": 0.6, + "learning_rate": 4.499268161158617e-05, + "loss": 0.0066, + "step": 2600 + }, + { + "epoch": 0.6, + "learning_rate": 4.4990755719898316e-05, + "loss": 0.0001, + "step": 2601 + }, + { + "epoch": 0.6, + "learning_rate": 4.4988829828210464e-05, + "loss": 0.0063, + "step": 2602 + }, + { + "epoch": 0.6, + "learning_rate": 4.498690393652261e-05, + "loss": 0.0, + "step": 2603 + }, + { + "epoch": 0.6, + "learning_rate": 4.498497804483476e-05, + "loss": 0.0001, + "step": 2604 + }, + { + "epoch": 0.6, + "learning_rate": 4.498305215314691e-05, + "loss": 0.0001, + "step": 2605 + }, + { + "epoch": 0.6, + "learning_rate": 4.4981126261459054e-05, + "loss": 0.0, + "step": 2606 + }, + { + "epoch": 0.6, + "learning_rate": 4.497920036977121e-05, + "loss": 0.0, + "step": 2607 + }, + { + "epoch": 0.6, + "learning_rate": 4.4977274478083355e-05, + "loss": 0.0, + "step": 2608 + }, + { + "epoch": 0.6, + "learning_rate": 4.49753485863955e-05, + "loss": 0.0003, + "step": 2609 + }, + { + "epoch": 0.6, + "learning_rate": 4.497342269470765e-05, + "loss": 0.0, + "step": 2610 + }, + { + "epoch": 0.6, + "learning_rate": 4.4971496803019804e-05, + "loss": 0.0, + "step": 2611 + }, + { + "epoch": 0.6, + "learning_rate": 4.4969570911331945e-05, + "loss": 0.001, + "step": 2612 + }, + { + "epoch": 0.6, + "learning_rate": 4.496764501964409e-05, + "loss": 0.032, + "step": 2613 + }, + { + "epoch": 0.6, + "learning_rate": 4.496571912795625e-05, + "loss": 0.0001, + "step": 2614 + }, + { + "epoch": 0.6, + "learning_rate": 4.4963793236268394e-05, + "loss": 0.0022, + "step": 2615 + }, + { + "epoch": 0.6, + "learning_rate": 4.496186734458054e-05, + "loss": 0.0006, + "step": 2616 + }, + { + "epoch": 0.6, + "learning_rate": 4.495994145289269e-05, + "loss": 0.0009, + "step": 2617 + }, + { + "epoch": 0.61, + "learning_rate": 4.4958015561204844e-05, + "loss": 0.0002, + "step": 2618 + }, + { + "epoch": 0.61, + "learning_rate": 4.4956089669516984e-05, + "loss": 0.0003, + "step": 2619 + }, + { + "epoch": 0.61, + "learning_rate": 4.495416377782914e-05, + "loss": 0.0015, + "step": 2620 + }, + { + "epoch": 0.61, + "learning_rate": 4.4952237886141286e-05, + "loss": 0.1756, + "step": 2621 + }, + { + "epoch": 0.61, + "learning_rate": 4.4950311994453434e-05, + "loss": 0.0003, + "step": 2622 + }, + { + "epoch": 0.61, + "learning_rate": 4.494838610276558e-05, + "loss": 0.0007, + "step": 2623 + }, + { + "epoch": 0.61, + "learning_rate": 4.494646021107773e-05, + "loss": 0.0023, + "step": 2624 + }, + { + "epoch": 0.61, + "learning_rate": 4.494453431938988e-05, + "loss": 0.0, + "step": 2625 + }, + { + "epoch": 0.61, + "learning_rate": 4.4942608427702023e-05, + "loss": 0.0012, + "step": 2626 + }, + { + "epoch": 0.61, + "learning_rate": 4.494068253601418e-05, + "loss": 0.0004, + "step": 2627 + }, + { + "epoch": 0.61, + "learning_rate": 4.4938756644326325e-05, + "loss": 0.0071, + "step": 2628 + }, + { + "epoch": 0.61, + "learning_rate": 4.493683075263847e-05, + "loss": 0.0, + "step": 2629 + }, + { + "epoch": 0.61, + "learning_rate": 4.493490486095062e-05, + "loss": 0.0001, + "step": 2630 + }, + { + "epoch": 0.61, + "learning_rate": 4.4932978969262774e-05, + "loss": 0.0113, + "step": 2631 + }, + { + "epoch": 0.61, + "learning_rate": 4.493105307757492e-05, + "loss": 0.0004, + "step": 2632 + }, + { + "epoch": 0.61, + "learning_rate": 4.492912718588706e-05, + "loss": 0.0, + "step": 2633 + }, + { + "epoch": 0.61, + "learning_rate": 4.492720129419922e-05, + "loss": 0.0, + "step": 2634 + }, + { + "epoch": 0.61, + "learning_rate": 4.4925275402511364e-05, + "loss": 0.0, + "step": 2635 + }, + { + "epoch": 0.61, + "learning_rate": 4.492334951082351e-05, + "loss": 0.0005, + "step": 2636 + }, + { + "epoch": 0.61, + "learning_rate": 4.492142361913566e-05, + "loss": 0.0, + "step": 2637 + }, + { + "epoch": 0.61, + "learning_rate": 4.4919497727447813e-05, + "loss": 0.0004, + "step": 2638 + }, + { + "epoch": 0.61, + "learning_rate": 4.491757183575996e-05, + "loss": 0.0, + "step": 2639 + }, + { + "epoch": 0.61, + "learning_rate": 4.491564594407211e-05, + "loss": 0.0008, + "step": 2640 + }, + { + "epoch": 0.61, + "learning_rate": 4.4913720052384256e-05, + "loss": 0.0004, + "step": 2641 + }, + { + "epoch": 0.61, + "learning_rate": 4.4911794160696403e-05, + "loss": 0.0002, + "step": 2642 + }, + { + "epoch": 0.61, + "learning_rate": 4.490986826900855e-05, + "loss": 0.0, + "step": 2643 + }, + { + "epoch": 0.61, + "learning_rate": 4.49079423773207e-05, + "loss": 0.0029, + "step": 2644 + }, + { + "epoch": 0.61, + "learning_rate": 4.490601648563285e-05, + "loss": 0.0056, + "step": 2645 + }, + { + "epoch": 0.61, + "learning_rate": 4.4904090593945e-05, + "loss": 0.0004, + "step": 2646 + }, + { + "epoch": 0.61, + "learning_rate": 4.490216470225715e-05, + "loss": 0.0, + "step": 2647 + }, + { + "epoch": 0.61, + "learning_rate": 4.4900238810569295e-05, + "loss": 0.0003, + "step": 2648 + }, + { + "epoch": 0.61, + "learning_rate": 4.489831291888145e-05, + "loss": 0.0001, + "step": 2649 + }, + { + "epoch": 0.61, + "learning_rate": 4.489638702719359e-05, + "loss": 0.0002, + "step": 2650 + }, + { + "epoch": 0.61, + "learning_rate": 4.489446113550574e-05, + "loss": 0.0015, + "step": 2651 + }, + { + "epoch": 0.61, + "learning_rate": 4.489253524381789e-05, + "loss": 0.0301, + "step": 2652 + }, + { + "epoch": 0.61, + "learning_rate": 4.489060935213004e-05, + "loss": 0.0002, + "step": 2653 + }, + { + "epoch": 0.61, + "learning_rate": 4.488868346044219e-05, + "loss": 0.0001, + "step": 2654 + }, + { + "epoch": 0.61, + "learning_rate": 4.4886757568754334e-05, + "loss": 0.0065, + "step": 2655 + }, + { + "epoch": 0.61, + "learning_rate": 4.488483167706649e-05, + "loss": 0.0003, + "step": 2656 + }, + { + "epoch": 0.61, + "learning_rate": 4.488290578537863e-05, + "loss": 0.0001, + "step": 2657 + }, + { + "epoch": 0.61, + "learning_rate": 4.488097989369078e-05, + "loss": 0.0003, + "step": 2658 + }, + { + "epoch": 0.61, + "learning_rate": 4.487905400200293e-05, + "loss": 0.0006, + "step": 2659 + }, + { + "epoch": 0.61, + "learning_rate": 4.487712811031508e-05, + "loss": 0.0001, + "step": 2660 + }, + { + "epoch": 0.61, + "learning_rate": 4.4875202218627226e-05, + "loss": 0.0001, + "step": 2661 + }, + { + "epoch": 0.62, + "learning_rate": 4.487327632693937e-05, + "loss": 0.0, + "step": 2662 + }, + { + "epoch": 0.62, + "learning_rate": 4.487135043525153e-05, + "loss": 0.0, + "step": 2663 + }, + { + "epoch": 0.62, + "learning_rate": 4.486942454356367e-05, + "loss": 0.0022, + "step": 2664 + }, + { + "epoch": 0.62, + "learning_rate": 4.486749865187582e-05, + "loss": 0.0003, + "step": 2665 + }, + { + "epoch": 0.62, + "learning_rate": 4.486557276018797e-05, + "loss": 0.0011, + "step": 2666 + }, + { + "epoch": 0.62, + "learning_rate": 4.486364686850012e-05, + "loss": 0.0001, + "step": 2667 + }, + { + "epoch": 0.62, + "learning_rate": 4.4861720976812265e-05, + "loss": 0.0075, + "step": 2668 + }, + { + "epoch": 0.62, + "learning_rate": 4.485979508512441e-05, + "loss": 0.001, + "step": 2669 + }, + { + "epoch": 0.62, + "learning_rate": 4.485786919343657e-05, + "loss": 0.0001, + "step": 2670 + }, + { + "epoch": 0.62, + "learning_rate": 4.485594330174871e-05, + "loss": 0.0001, + "step": 2671 + }, + { + "epoch": 0.62, + "learning_rate": 4.485401741006086e-05, + "loss": 0.0002, + "step": 2672 + }, + { + "epoch": 0.62, + "learning_rate": 4.485209151837301e-05, + "loss": 0.0, + "step": 2673 + }, + { + "epoch": 0.62, + "learning_rate": 4.4850165626685157e-05, + "loss": 0.0002, + "step": 2674 + }, + { + "epoch": 0.62, + "learning_rate": 4.4848239734997304e-05, + "loss": 0.0049, + "step": 2675 + }, + { + "epoch": 0.62, + "learning_rate": 4.484631384330946e-05, + "loss": 0.0031, + "step": 2676 + }, + { + "epoch": 0.62, + "learning_rate": 4.4844387951621606e-05, + "loss": 0.0007, + "step": 2677 + }, + { + "epoch": 0.62, + "learning_rate": 4.484246205993375e-05, + "loss": 0.0008, + "step": 2678 + }, + { + "epoch": 0.62, + "learning_rate": 4.48405361682459e-05, + "loss": 0.0003, + "step": 2679 + }, + { + "epoch": 0.62, + "learning_rate": 4.483861027655805e-05, + "loss": 0.0002, + "step": 2680 + }, + { + "epoch": 0.62, + "learning_rate": 4.4836684384870196e-05, + "loss": 0.0, + "step": 2681 + }, + { + "epoch": 0.62, + "learning_rate": 4.483475849318234e-05, + "loss": 0.0028, + "step": 2682 + }, + { + "epoch": 0.62, + "learning_rate": 4.48328326014945e-05, + "loss": 0.0012, + "step": 2683 + }, + { + "epoch": 0.62, + "learning_rate": 4.4830906709806645e-05, + "loss": 0.1007, + "step": 2684 + }, + { + "epoch": 0.62, + "learning_rate": 4.482898081811879e-05, + "loss": 0.0006, + "step": 2685 + }, + { + "epoch": 0.62, + "learning_rate": 4.482705492643094e-05, + "loss": 0.0001, + "step": 2686 + }, + { + "epoch": 0.62, + "learning_rate": 4.4825129034743094e-05, + "loss": 0.0041, + "step": 2687 + }, + { + "epoch": 0.62, + "learning_rate": 4.4823203143055235e-05, + "loss": 0.0, + "step": 2688 + }, + { + "epoch": 0.62, + "learning_rate": 4.482127725136738e-05, + "loss": 0.0, + "step": 2689 + }, + { + "epoch": 0.62, + "learning_rate": 4.4819351359679537e-05, + "loss": 0.0014, + "step": 2690 + }, + { + "epoch": 0.62, + "learning_rate": 4.481742546799168e-05, + "loss": 0.0001, + "step": 2691 + }, + { + "epoch": 0.62, + "learning_rate": 4.481549957630383e-05, + "loss": 0.0006, + "step": 2692 + }, + { + "epoch": 0.62, + "learning_rate": 4.481357368461598e-05, + "loss": 0.0001, + "step": 2693 + }, + { + "epoch": 0.62, + "learning_rate": 4.4811647792928126e-05, + "loss": 0.0, + "step": 2694 + }, + { + "epoch": 0.62, + "learning_rate": 4.4809721901240274e-05, + "loss": 0.0002, + "step": 2695 + }, + { + "epoch": 0.62, + "learning_rate": 4.480779600955243e-05, + "loss": 0.0, + "step": 2696 + }, + { + "epoch": 0.62, + "learning_rate": 4.4805870117864576e-05, + "loss": 0.0004, + "step": 2697 + }, + { + "epoch": 0.62, + "learning_rate": 4.4803944226176716e-05, + "loss": 0.0016, + "step": 2698 + }, + { + "epoch": 0.62, + "learning_rate": 4.480201833448887e-05, + "loss": 0.0038, + "step": 2699 + }, + { + "epoch": 0.62, + "learning_rate": 4.480009244280102e-05, + "loss": 0.0, + "step": 2700 + }, + { + "epoch": 0.62, + "learning_rate": 4.4798166551113166e-05, + "loss": 0.0001, + "step": 2701 + }, + { + "epoch": 0.62, + "learning_rate": 4.479624065942531e-05, + "loss": 0.0, + "step": 2702 + }, + { + "epoch": 0.62, + "learning_rate": 4.479431476773747e-05, + "loss": 0.0013, + "step": 2703 + }, + { + "epoch": 0.62, + "learning_rate": 4.4792388876049615e-05, + "loss": 0.0005, + "step": 2704 + }, + { + "epoch": 0.63, + "learning_rate": 4.479046298436176e-05, + "loss": 0.0004, + "step": 2705 + }, + { + "epoch": 0.63, + "learning_rate": 4.478853709267391e-05, + "loss": 0.0, + "step": 2706 + }, + { + "epoch": 0.63, + "learning_rate": 4.478661120098606e-05, + "loss": 0.0, + "step": 2707 + }, + { + "epoch": 0.63, + "learning_rate": 4.4784685309298205e-05, + "loss": 0.0025, + "step": 2708 + }, + { + "epoch": 0.63, + "learning_rate": 4.478275941761035e-05, + "loss": 0.0164, + "step": 2709 + }, + { + "epoch": 0.63, + "learning_rate": 4.4780833525922506e-05, + "loss": 0.0009, + "step": 2710 + }, + { + "epoch": 0.63, + "learning_rate": 4.4778907634234654e-05, + "loss": 0.0006, + "step": 2711 + }, + { + "epoch": 0.63, + "learning_rate": 4.47769817425468e-05, + "loss": 0.0003, + "step": 2712 + }, + { + "epoch": 0.63, + "learning_rate": 4.477505585085895e-05, + "loss": 0.0, + "step": 2713 + }, + { + "epoch": 0.63, + "learning_rate": 4.47731299591711e-05, + "loss": 0.0001, + "step": 2714 + }, + { + "epoch": 0.63, + "learning_rate": 4.4771204067483244e-05, + "loss": 0.0142, + "step": 2715 + }, + { + "epoch": 0.63, + "learning_rate": 4.476927817579539e-05, + "loss": 0.0003, + "step": 2716 + }, + { + "epoch": 0.63, + "learning_rate": 4.4767352284107546e-05, + "loss": 0.0712, + "step": 2717 + }, + { + "epoch": 0.63, + "learning_rate": 4.476542639241969e-05, + "loss": 0.0029, + "step": 2718 + }, + { + "epoch": 0.63, + "learning_rate": 4.476350050073184e-05, + "loss": 0.0035, + "step": 2719 + }, + { + "epoch": 0.63, + "learning_rate": 4.476157460904399e-05, + "loss": 0.0, + "step": 2720 + }, + { + "epoch": 0.63, + "learning_rate": 4.475964871735614e-05, + "loss": 0.0182, + "step": 2721 + }, + { + "epoch": 0.63, + "learning_rate": 4.475772282566828e-05, + "loss": 0.0001, + "step": 2722 + }, + { + "epoch": 0.63, + "learning_rate": 4.475579693398044e-05, + "loss": 0.0009, + "step": 2723 + }, + { + "epoch": 0.63, + "learning_rate": 4.4753871042292585e-05, + "loss": 0.0, + "step": 2724 + }, + { + "epoch": 0.63, + "learning_rate": 4.475194515060473e-05, + "loss": 0.0007, + "step": 2725 + }, + { + "epoch": 0.63, + "learning_rate": 4.475001925891688e-05, + "loss": 0.0035, + "step": 2726 + }, + { + "epoch": 0.63, + "learning_rate": 4.474809336722903e-05, + "loss": 0.0107, + "step": 2727 + }, + { + "epoch": 0.63, + "learning_rate": 4.474616747554118e-05, + "loss": 0.0, + "step": 2728 + }, + { + "epoch": 0.63, + "learning_rate": 4.474424158385332e-05, + "loss": 0.0, + "step": 2729 + }, + { + "epoch": 0.63, + "learning_rate": 4.4742315692165476e-05, + "loss": 0.0, + "step": 2730 + }, + { + "epoch": 0.63, + "learning_rate": 4.4740389800477624e-05, + "loss": 0.0001, + "step": 2731 + }, + { + "epoch": 0.63, + "learning_rate": 4.473846390878977e-05, + "loss": 0.0002, + "step": 2732 + }, + { + "epoch": 0.63, + "learning_rate": 4.473653801710192e-05, + "loss": 0.0024, + "step": 2733 + }, + { + "epoch": 0.63, + "learning_rate": 4.473461212541407e-05, + "loss": 0.0008, + "step": 2734 + }, + { + "epoch": 0.63, + "learning_rate": 4.473268623372622e-05, + "loss": 0.0008, + "step": 2735 + }, + { + "epoch": 0.63, + "learning_rate": 4.473076034203836e-05, + "loss": 0.0023, + "step": 2736 + }, + { + "epoch": 0.63, + "learning_rate": 4.4728834450350515e-05, + "loss": 0.0, + "step": 2737 + }, + { + "epoch": 0.63, + "learning_rate": 4.472690855866266e-05, + "loss": 0.0178, + "step": 2738 + }, + { + "epoch": 0.63, + "learning_rate": 4.472498266697481e-05, + "loss": 0.1011, + "step": 2739 + }, + { + "epoch": 0.63, + "learning_rate": 4.472305677528696e-05, + "loss": 0.0165, + "step": 2740 + }, + { + "epoch": 0.63, + "learning_rate": 4.472113088359911e-05, + "loss": 0.0013, + "step": 2741 + }, + { + "epoch": 0.63, + "learning_rate": 4.471920499191126e-05, + "loss": 0.0044, + "step": 2742 + }, + { + "epoch": 0.63, + "learning_rate": 4.471727910022341e-05, + "loss": 0.0, + "step": 2743 + }, + { + "epoch": 0.63, + "learning_rate": 4.4715353208535555e-05, + "loss": 0.0, + "step": 2744 + }, + { + "epoch": 0.63, + "learning_rate": 4.47134273168477e-05, + "loss": 0.0001, + "step": 2745 + }, + { + "epoch": 0.63, + "learning_rate": 4.471150142515985e-05, + "loss": 0.0017, + "step": 2746 + }, + { + "epoch": 0.63, + "learning_rate": 4.4709575533472e-05, + "loss": 0.0001, + "step": 2747 + }, + { + "epoch": 0.64, + "learning_rate": 4.470764964178415e-05, + "loss": 0.0016, + "step": 2748 + }, + { + "epoch": 0.64, + "learning_rate": 4.47057237500963e-05, + "loss": 0.0029, + "step": 2749 + }, + { + "epoch": 0.64, + "learning_rate": 4.4703797858408446e-05, + "loss": 0.0001, + "step": 2750 + }, + { + "epoch": 0.64, + "learning_rate": 4.4701871966720594e-05, + "loss": 0.0003, + "step": 2751 + }, + { + "epoch": 0.64, + "learning_rate": 4.469994607503275e-05, + "loss": 0.0, + "step": 2752 + }, + { + "epoch": 0.64, + "learning_rate": 4.469802018334489e-05, + "loss": 0.0001, + "step": 2753 + }, + { + "epoch": 0.64, + "learning_rate": 4.4696094291657036e-05, + "loss": 0.0001, + "step": 2754 + }, + { + "epoch": 0.64, + "learning_rate": 4.469416839996919e-05, + "loss": 0.0011, + "step": 2755 + }, + { + "epoch": 0.64, + "learning_rate": 4.469224250828134e-05, + "loss": 0.001, + "step": 2756 + }, + { + "epoch": 0.64, + "learning_rate": 4.4690316616593485e-05, + "loss": 0.0003, + "step": 2757 + }, + { + "epoch": 0.64, + "learning_rate": 4.468839072490563e-05, + "loss": 0.0181, + "step": 2758 + }, + { + "epoch": 0.64, + "learning_rate": 4.468646483321779e-05, + "loss": 0.0, + "step": 2759 + }, + { + "epoch": 0.64, + "learning_rate": 4.468453894152993e-05, + "loss": 0.0001, + "step": 2760 + }, + { + "epoch": 0.64, + "learning_rate": 4.468261304984208e-05, + "loss": 0.0, + "step": 2761 + }, + { + "epoch": 0.64, + "learning_rate": 4.468068715815423e-05, + "loss": 0.0059, + "step": 2762 + }, + { + "epoch": 0.64, + "learning_rate": 4.467876126646637e-05, + "loss": 0.0051, + "step": 2763 + }, + { + "epoch": 0.64, + "learning_rate": 4.4676835374778524e-05, + "loss": 0.0, + "step": 2764 + }, + { + "epoch": 0.64, + "learning_rate": 4.467490948309067e-05, + "loss": 0.0, + "step": 2765 + }, + { + "epoch": 0.64, + "learning_rate": 4.467298359140282e-05, + "loss": 0.0014, + "step": 2766 + }, + { + "epoch": 0.64, + "learning_rate": 4.467105769971497e-05, + "loss": 0.0015, + "step": 2767 + }, + { + "epoch": 0.64, + "learning_rate": 4.466913180802712e-05, + "loss": 0.017, + "step": 2768 + }, + { + "epoch": 0.64, + "learning_rate": 4.466720591633927e-05, + "loss": 0.0003, + "step": 2769 + }, + { + "epoch": 0.64, + "learning_rate": 4.4665280024651416e-05, + "loss": 0.0065, + "step": 2770 + }, + { + "epoch": 0.64, + "learning_rate": 4.4663354132963564e-05, + "loss": 0.0111, + "step": 2771 + }, + { + "epoch": 0.64, + "learning_rate": 4.466142824127571e-05, + "loss": 0.0007, + "step": 2772 + }, + { + "epoch": 0.64, + "learning_rate": 4.465950234958786e-05, + "loss": 0.0, + "step": 2773 + }, + { + "epoch": 0.64, + "learning_rate": 4.4657576457900006e-05, + "loss": 0.0003, + "step": 2774 + }, + { + "epoch": 0.64, + "learning_rate": 4.465565056621216e-05, + "loss": 0.0, + "step": 2775 + }, + { + "epoch": 0.64, + "learning_rate": 4.465372467452431e-05, + "loss": 0.0243, + "step": 2776 + }, + { + "epoch": 0.64, + "learning_rate": 4.4651798782836455e-05, + "loss": 0.0002, + "step": 2777 + }, + { + "epoch": 0.64, + "learning_rate": 4.46498728911486e-05, + "loss": 0.0089, + "step": 2778 + }, + { + "epoch": 0.64, + "learning_rate": 4.464794699946076e-05, + "loss": 0.005, + "step": 2779 + }, + { + "epoch": 0.64, + "learning_rate": 4.46460211077729e-05, + "loss": 0.0011, + "step": 2780 + }, + { + "epoch": 0.64, + "learning_rate": 4.4644095216085045e-05, + "loss": 0.0025, + "step": 2781 + }, + { + "epoch": 0.64, + "learning_rate": 4.46421693243972e-05, + "loss": 0.0003, + "step": 2782 + }, + { + "epoch": 0.64, + "learning_rate": 4.464024343270935e-05, + "loss": 0.0001, + "step": 2783 + }, + { + "epoch": 0.64, + "learning_rate": 4.4638317541021494e-05, + "loss": 0.0004, + "step": 2784 + }, + { + "epoch": 0.64, + "learning_rate": 4.463639164933364e-05, + "loss": 0.0, + "step": 2785 + }, + { + "epoch": 0.64, + "learning_rate": 4.4634465757645796e-05, + "loss": 0.0026, + "step": 2786 + }, + { + "epoch": 0.64, + "learning_rate": 4.463253986595794e-05, + "loss": 0.0, + "step": 2787 + }, + { + "epoch": 0.64, + "learning_rate": 4.463061397427009e-05, + "loss": 0.0008, + "step": 2788 + }, + { + "epoch": 0.64, + "learning_rate": 4.462868808258224e-05, + "loss": 0.0011, + "step": 2789 + }, + { + "epoch": 0.64, + "learning_rate": 4.4626762190894386e-05, + "loss": 0.0, + "step": 2790 + }, + { + "epoch": 0.65, + "learning_rate": 4.4624836299206533e-05, + "loss": 0.0021, + "step": 2791 + }, + { + "epoch": 0.65, + "learning_rate": 4.462291040751868e-05, + "loss": 0.0013, + "step": 2792 + }, + { + "epoch": 0.65, + "learning_rate": 4.4620984515830835e-05, + "loss": 0.0038, + "step": 2793 + }, + { + "epoch": 0.65, + "learning_rate": 4.4619058624142976e-05, + "loss": 0.0011, + "step": 2794 + }, + { + "epoch": 0.65, + "learning_rate": 4.461713273245513e-05, + "loss": 0.0015, + "step": 2795 + }, + { + "epoch": 0.65, + "learning_rate": 4.461520684076728e-05, + "loss": 0.0001, + "step": 2796 + }, + { + "epoch": 0.65, + "learning_rate": 4.4613280949079425e-05, + "loss": 0.0088, + "step": 2797 + }, + { + "epoch": 0.65, + "learning_rate": 4.461135505739157e-05, + "loss": 0.0, + "step": 2798 + }, + { + "epoch": 0.65, + "learning_rate": 4.460942916570373e-05, + "loss": 0.0003, + "step": 2799 + }, + { + "epoch": 0.65, + "learning_rate": 4.4607503274015874e-05, + "loss": 0.0, + "step": 2800 + }, + { + "epoch": 0.65, + "learning_rate": 4.4605577382328015e-05, + "loss": 0.0009, + "step": 2801 + }, + { + "epoch": 0.65, + "learning_rate": 4.460365149064017e-05, + "loss": 0.0003, + "step": 2802 + }, + { + "epoch": 0.65, + "learning_rate": 4.460172559895232e-05, + "loss": 0.001, + "step": 2803 + }, + { + "epoch": 0.65, + "learning_rate": 4.4599799707264464e-05, + "loss": 0.0002, + "step": 2804 + }, + { + "epoch": 0.65, + "learning_rate": 4.459787381557661e-05, + "loss": 0.0002, + "step": 2805 + }, + { + "epoch": 0.65, + "learning_rate": 4.4595947923888766e-05, + "loss": 0.014, + "step": 2806 + }, + { + "epoch": 0.65, + "learning_rate": 4.4594022032200913e-05, + "loss": 0.0004, + "step": 2807 + }, + { + "epoch": 0.65, + "learning_rate": 4.459209614051306e-05, + "loss": 0.0018, + "step": 2808 + }, + { + "epoch": 0.65, + "learning_rate": 4.459017024882521e-05, + "loss": 0.0, + "step": 2809 + }, + { + "epoch": 0.65, + "learning_rate": 4.4588244357137356e-05, + "loss": 0.0008, + "step": 2810 + }, + { + "epoch": 0.65, + "learning_rate": 4.45863184654495e-05, + "loss": 0.0, + "step": 2811 + }, + { + "epoch": 0.65, + "learning_rate": 4.458439257376165e-05, + "loss": 0.0019, + "step": 2812 + }, + { + "epoch": 0.65, + "learning_rate": 4.4582466682073805e-05, + "loss": 0.0003, + "step": 2813 + }, + { + "epoch": 0.65, + "learning_rate": 4.458054079038595e-05, + "loss": 0.0315, + "step": 2814 + }, + { + "epoch": 0.65, + "learning_rate": 4.45786148986981e-05, + "loss": 0.0, + "step": 2815 + }, + { + "epoch": 0.65, + "learning_rate": 4.457668900701025e-05, + "loss": 0.0001, + "step": 2816 + }, + { + "epoch": 0.65, + "learning_rate": 4.45747631153224e-05, + "loss": 0.0077, + "step": 2817 + }, + { + "epoch": 0.65, + "learning_rate": 4.457283722363454e-05, + "loss": 0.0072, + "step": 2818 + }, + { + "epoch": 0.65, + "learning_rate": 4.457091133194669e-05, + "loss": 0.0023, + "step": 2819 + }, + { + "epoch": 0.65, + "learning_rate": 4.4568985440258844e-05, + "loss": 0.0001, + "step": 2820 + }, + { + "epoch": 0.65, + "learning_rate": 4.456705954857099e-05, + "loss": 0.0053, + "step": 2821 + }, + { + "epoch": 0.65, + "learning_rate": 4.456513365688314e-05, + "loss": 0.0002, + "step": 2822 + }, + { + "epoch": 0.65, + "learning_rate": 4.4563207765195287e-05, + "loss": 0.0008, + "step": 2823 + }, + { + "epoch": 0.65, + "learning_rate": 4.456128187350744e-05, + "loss": 0.0008, + "step": 2824 + }, + { + "epoch": 0.65, + "learning_rate": 4.455935598181958e-05, + "loss": 0.0009, + "step": 2825 + }, + { + "epoch": 0.65, + "learning_rate": 4.4557430090131736e-05, + "loss": 0.0043, + "step": 2826 + }, + { + "epoch": 0.65, + "learning_rate": 4.455550419844388e-05, + "loss": 0.0026, + "step": 2827 + }, + { + "epoch": 0.65, + "learning_rate": 4.455357830675603e-05, + "loss": 0.0019, + "step": 2828 + }, + { + "epoch": 0.65, + "learning_rate": 4.455165241506818e-05, + "loss": 0.0006, + "step": 2829 + }, + { + "epoch": 0.65, + "learning_rate": 4.4549726523380326e-05, + "loss": 0.0009, + "step": 2830 + }, + { + "epoch": 0.65, + "learning_rate": 4.454780063169248e-05, + "loss": 0.006, + "step": 2831 + }, + { + "epoch": 0.65, + "learning_rate": 4.454587474000462e-05, + "loss": 0.0, + "step": 2832 + }, + { + "epoch": 0.65, + "learning_rate": 4.4543948848316775e-05, + "loss": 0.0001, + "step": 2833 + }, + { + "epoch": 0.65, + "learning_rate": 4.454202295662892e-05, + "loss": 0.0024, + "step": 2834 + }, + { + "epoch": 0.66, + "learning_rate": 4.454009706494107e-05, + "loss": 0.0025, + "step": 2835 + }, + { + "epoch": 0.66, + "learning_rate": 4.453817117325322e-05, + "loss": 0.0153, + "step": 2836 + }, + { + "epoch": 0.66, + "learning_rate": 4.4536245281565365e-05, + "loss": 0.0, + "step": 2837 + }, + { + "epoch": 0.66, + "learning_rate": 4.453431938987752e-05, + "loss": 0.0033, + "step": 2838 + }, + { + "epoch": 0.66, + "learning_rate": 4.453239349818966e-05, + "loss": 0.0004, + "step": 2839 + }, + { + "epoch": 0.66, + "learning_rate": 4.4530467606501814e-05, + "loss": 0.0002, + "step": 2840 + }, + { + "epoch": 0.66, + "learning_rate": 4.452854171481396e-05, + "loss": 0.0004, + "step": 2841 + }, + { + "epoch": 0.66, + "learning_rate": 4.452661582312611e-05, + "loss": 0.0, + "step": 2842 + }, + { + "epoch": 0.66, + "learning_rate": 4.4524689931438256e-05, + "loss": 0.0017, + "step": 2843 + }, + { + "epoch": 0.66, + "learning_rate": 4.452276403975041e-05, + "loss": 0.0161, + "step": 2844 + }, + { + "epoch": 0.66, + "learning_rate": 4.452083814806255e-05, + "loss": 0.0023, + "step": 2845 + }, + { + "epoch": 0.66, + "learning_rate": 4.4518912256374706e-05, + "loss": 0.0, + "step": 2846 + }, + { + "epoch": 0.66, + "learning_rate": 4.451698636468685e-05, + "loss": 0.0102, + "step": 2847 + }, + { + "epoch": 0.66, + "learning_rate": 4.4515060472999e-05, + "loss": 0.0002, + "step": 2848 + }, + { + "epoch": 0.66, + "learning_rate": 4.451313458131115e-05, + "loss": 0.0077, + "step": 2849 + }, + { + "epoch": 0.66, + "learning_rate": 4.4511208689623296e-05, + "loss": 0.004, + "step": 2850 + }, + { + "epoch": 0.66, + "learning_rate": 4.450928279793545e-05, + "loss": 0.0001, + "step": 2851 + }, + { + "epoch": 0.66, + "learning_rate": 4.450735690624759e-05, + "loss": 0.0004, + "step": 2852 + }, + { + "epoch": 0.66, + "learning_rate": 4.4505431014559745e-05, + "loss": 0.0015, + "step": 2853 + }, + { + "epoch": 0.66, + "learning_rate": 4.450350512287189e-05, + "loss": 0.0012, + "step": 2854 + }, + { + "epoch": 0.66, + "learning_rate": 4.450157923118404e-05, + "loss": 0.0006, + "step": 2855 + }, + { + "epoch": 0.66, + "learning_rate": 4.449965333949619e-05, + "loss": 0.0001, + "step": 2856 + }, + { + "epoch": 0.66, + "learning_rate": 4.4497727447808335e-05, + "loss": 0.0018, + "step": 2857 + }, + { + "epoch": 0.66, + "learning_rate": 4.449580155612049e-05, + "loss": 0.0007, + "step": 2858 + }, + { + "epoch": 0.66, + "learning_rate": 4.449387566443263e-05, + "loss": 0.0003, + "step": 2859 + }, + { + "epoch": 0.66, + "learning_rate": 4.4491949772744784e-05, + "loss": 0.0019, + "step": 2860 + }, + { + "epoch": 0.66, + "learning_rate": 4.449002388105693e-05, + "loss": 0.0003, + "step": 2861 + }, + { + "epoch": 0.66, + "learning_rate": 4.448809798936908e-05, + "loss": 0.0034, + "step": 2862 + }, + { + "epoch": 0.66, + "learning_rate": 4.4486172097681226e-05, + "loss": 0.001, + "step": 2863 + }, + { + "epoch": 0.66, + "learning_rate": 4.448424620599338e-05, + "loss": 0.0001, + "step": 2864 + }, + { + "epoch": 0.66, + "learning_rate": 4.448232031430553e-05, + "loss": 0.0001, + "step": 2865 + }, + { + "epoch": 0.66, + "learning_rate": 4.448039442261767e-05, + "loss": 0.004, + "step": 2866 + }, + { + "epoch": 0.66, + "learning_rate": 4.447846853092982e-05, + "loss": 0.0, + "step": 2867 + }, + { + "epoch": 0.66, + "learning_rate": 4.447654263924197e-05, + "loss": 0.0004, + "step": 2868 + }, + { + "epoch": 0.66, + "learning_rate": 4.447461674755412e-05, + "loss": 0.0017, + "step": 2869 + }, + { + "epoch": 0.66, + "learning_rate": 4.4472690855866265e-05, + "loss": 0.0017, + "step": 2870 + }, + { + "epoch": 0.66, + "learning_rate": 4.447076496417842e-05, + "loss": 0.0001, + "step": 2871 + }, + { + "epoch": 0.66, + "learning_rate": 4.446883907249057e-05, + "loss": 0.0, + "step": 2872 + }, + { + "epoch": 0.66, + "learning_rate": 4.4466913180802715e-05, + "loss": 0.0027, + "step": 2873 + }, + { + "epoch": 0.66, + "learning_rate": 4.446498728911486e-05, + "loss": 0.0007, + "step": 2874 + }, + { + "epoch": 0.66, + "learning_rate": 4.446306139742701e-05, + "loss": 0.0002, + "step": 2875 + }, + { + "epoch": 0.66, + "learning_rate": 4.446113550573916e-05, + "loss": 0.0003, + "step": 2876 + }, + { + "epoch": 0.66, + "learning_rate": 4.4459209614051305e-05, + "loss": 0.0006, + "step": 2877 + }, + { + "epoch": 0.67, + "learning_rate": 4.445728372236346e-05, + "loss": 0.006, + "step": 2878 + }, + { + "epoch": 0.67, + "learning_rate": 4.4455357830675606e-05, + "loss": 0.0001, + "step": 2879 + }, + { + "epoch": 0.67, + "learning_rate": 4.4453431938987754e-05, + "loss": 0.001, + "step": 2880 + }, + { + "epoch": 0.67, + "learning_rate": 4.44515060472999e-05, + "loss": 0.0022, + "step": 2881 + }, + { + "epoch": 0.67, + "learning_rate": 4.4449580155612056e-05, + "loss": 0.0001, + "step": 2882 + }, + { + "epoch": 0.67, + "learning_rate": 4.4447654263924196e-05, + "loss": 0.0046, + "step": 2883 + }, + { + "epoch": 0.67, + "learning_rate": 4.4445728372236344e-05, + "loss": 0.0, + "step": 2884 + }, + { + "epoch": 0.67, + "learning_rate": 4.44438024805485e-05, + "loss": 0.0002, + "step": 2885 + }, + { + "epoch": 0.67, + "learning_rate": 4.4441876588860645e-05, + "loss": 0.0012, + "step": 2886 + }, + { + "epoch": 0.67, + "learning_rate": 4.443995069717279e-05, + "loss": 0.0007, + "step": 2887 + }, + { + "epoch": 0.67, + "learning_rate": 4.443802480548494e-05, + "loss": 0.0018, + "step": 2888 + }, + { + "epoch": 0.67, + "learning_rate": 4.4436098913797095e-05, + "loss": 0.001, + "step": 2889 + }, + { + "epoch": 0.67, + "learning_rate": 4.4434173022109235e-05, + "loss": 0.0001, + "step": 2890 + }, + { + "epoch": 0.67, + "learning_rate": 4.443224713042139e-05, + "loss": 0.0004, + "step": 2891 + }, + { + "epoch": 0.67, + "learning_rate": 4.443032123873354e-05, + "loss": 0.0012, + "step": 2892 + }, + { + "epoch": 0.67, + "learning_rate": 4.4428395347045685e-05, + "loss": 0.0006, + "step": 2893 + }, + { + "epoch": 0.67, + "learning_rate": 4.442646945535783e-05, + "loss": 0.0001, + "step": 2894 + }, + { + "epoch": 0.67, + "learning_rate": 4.442454356366998e-05, + "loss": 0.0036, + "step": 2895 + }, + { + "epoch": 0.67, + "learning_rate": 4.4422617671982134e-05, + "loss": 0.0004, + "step": 2896 + }, + { + "epoch": 0.67, + "learning_rate": 4.4420691780294275e-05, + "loss": 0.0, + "step": 2897 + }, + { + "epoch": 0.67, + "learning_rate": 4.441876588860643e-05, + "loss": 0.0047, + "step": 2898 + }, + { + "epoch": 0.67, + "learning_rate": 4.4416839996918576e-05, + "loss": 0.0, + "step": 2899 + }, + { + "epoch": 0.67, + "learning_rate": 4.4414914105230724e-05, + "loss": 0.0004, + "step": 2900 + }, + { + "epoch": 0.67, + "learning_rate": 4.441298821354287e-05, + "loss": 0.0025, + "step": 2901 + }, + { + "epoch": 0.67, + "learning_rate": 4.4411062321855025e-05, + "loss": 0.006, + "step": 2902 + }, + { + "epoch": 0.67, + "learning_rate": 4.440913643016717e-05, + "loss": 0.0001, + "step": 2903 + }, + { + "epoch": 0.67, + "learning_rate": 4.4407210538479314e-05, + "loss": 0.0069, + "step": 2904 + }, + { + "epoch": 0.67, + "learning_rate": 4.440528464679147e-05, + "loss": 0.0015, + "step": 2905 + }, + { + "epoch": 0.67, + "learning_rate": 4.4403358755103615e-05, + "loss": 0.0, + "step": 2906 + }, + { + "epoch": 0.67, + "learning_rate": 4.440143286341576e-05, + "loss": 0.0, + "step": 2907 + }, + { + "epoch": 0.67, + "learning_rate": 4.439950697172791e-05, + "loss": 0.0007, + "step": 2908 + }, + { + "epoch": 0.67, + "learning_rate": 4.4397581080040065e-05, + "loss": 0.0102, + "step": 2909 + }, + { + "epoch": 0.67, + "learning_rate": 4.439565518835221e-05, + "loss": 0.0286, + "step": 2910 + }, + { + "epoch": 0.67, + "learning_rate": 4.439372929666436e-05, + "loss": 0.0016, + "step": 2911 + }, + { + "epoch": 0.67, + "learning_rate": 4.439180340497651e-05, + "loss": 0.0004, + "step": 2912 + }, + { + "epoch": 0.67, + "learning_rate": 4.4389877513288654e-05, + "loss": 0.0005, + "step": 2913 + }, + { + "epoch": 0.67, + "learning_rate": 4.43879516216008e-05, + "loss": 0.0001, + "step": 2914 + }, + { + "epoch": 0.67, + "learning_rate": 4.438602572991295e-05, + "loss": 0.0049, + "step": 2915 + }, + { + "epoch": 0.67, + "learning_rate": 4.4384099838225104e-05, + "loss": 0.0013, + "step": 2916 + }, + { + "epoch": 0.67, + "learning_rate": 4.4382173946537244e-05, + "loss": 0.0017, + "step": 2917 + }, + { + "epoch": 0.67, + "learning_rate": 4.43802480548494e-05, + "loss": 0.0004, + "step": 2918 + }, + { + "epoch": 0.67, + "learning_rate": 4.4378322163161546e-05, + "loss": 0.0001, + "step": 2919 + }, + { + "epoch": 0.67, + "learning_rate": 4.4376396271473694e-05, + "loss": 0.0001, + "step": 2920 + }, + { + "epoch": 0.68, + "learning_rate": 4.437447037978584e-05, + "loss": 0.0, + "step": 2921 + }, + { + "epoch": 0.68, + "learning_rate": 4.437254448809799e-05, + "loss": 0.0, + "step": 2922 + }, + { + "epoch": 0.68, + "learning_rate": 4.437061859641014e-05, + "loss": 0.0001, + "step": 2923 + }, + { + "epoch": 0.68, + "learning_rate": 4.4368692704722284e-05, + "loss": 0.0001, + "step": 2924 + }, + { + "epoch": 0.68, + "learning_rate": 4.436676681303444e-05, + "loss": 0.0, + "step": 2925 + }, + { + "epoch": 0.68, + "learning_rate": 4.4364840921346585e-05, + "loss": 0.0003, + "step": 2926 + }, + { + "epoch": 0.68, + "learning_rate": 4.436291502965873e-05, + "loss": 0.0018, + "step": 2927 + }, + { + "epoch": 0.68, + "learning_rate": 4.436098913797088e-05, + "loss": 0.0, + "step": 2928 + }, + { + "epoch": 0.68, + "learning_rate": 4.4359063246283034e-05, + "loss": 0.0023, + "step": 2929 + }, + { + "epoch": 0.68, + "learning_rate": 4.435713735459518e-05, + "loss": 0.004, + "step": 2930 + }, + { + "epoch": 0.68, + "learning_rate": 4.435521146290732e-05, + "loss": 0.017, + "step": 2931 + }, + { + "epoch": 0.68, + "learning_rate": 4.435328557121948e-05, + "loss": 0.0028, + "step": 2932 + }, + { + "epoch": 0.68, + "learning_rate": 4.4351359679531624e-05, + "loss": 0.0001, + "step": 2933 + }, + { + "epoch": 0.68, + "learning_rate": 4.434943378784377e-05, + "loss": 0.0198, + "step": 2934 + }, + { + "epoch": 0.68, + "learning_rate": 4.434750789615592e-05, + "loss": 0.0003, + "step": 2935 + }, + { + "epoch": 0.68, + "learning_rate": 4.4345582004468074e-05, + "loss": 0.0002, + "step": 2936 + }, + { + "epoch": 0.68, + "learning_rate": 4.434365611278022e-05, + "loss": 0.0007, + "step": 2937 + }, + { + "epoch": 0.68, + "learning_rate": 4.434173022109237e-05, + "loss": 0.0008, + "step": 2938 + }, + { + "epoch": 0.68, + "learning_rate": 4.4339804329404516e-05, + "loss": 0.0001, + "step": 2939 + }, + { + "epoch": 0.68, + "learning_rate": 4.4337878437716663e-05, + "loss": 0.0014, + "step": 2940 + }, + { + "epoch": 0.68, + "learning_rate": 4.433595254602881e-05, + "loss": 0.0, + "step": 2941 + }, + { + "epoch": 0.68, + "learning_rate": 4.433402665434096e-05, + "loss": 0.0019, + "step": 2942 + }, + { + "epoch": 0.68, + "learning_rate": 4.433210076265311e-05, + "loss": 0.0003, + "step": 2943 + }, + { + "epoch": 0.68, + "learning_rate": 4.433017487096526e-05, + "loss": 0.0, + "step": 2944 + }, + { + "epoch": 0.68, + "learning_rate": 4.432824897927741e-05, + "loss": 0.0, + "step": 2945 + }, + { + "epoch": 0.68, + "learning_rate": 4.4326323087589555e-05, + "loss": 0.0, + "step": 2946 + }, + { + "epoch": 0.68, + "learning_rate": 4.432439719590171e-05, + "loss": 0.0, + "step": 2947 + }, + { + "epoch": 0.68, + "learning_rate": 4.432247130421385e-05, + "loss": 0.0079, + "step": 2948 + }, + { + "epoch": 0.68, + "learning_rate": 4.4320545412526e-05, + "loss": 0.0041, + "step": 2949 + }, + { + "epoch": 0.68, + "learning_rate": 4.431861952083815e-05, + "loss": 0.0, + "step": 2950 + }, + { + "epoch": 0.68, + "learning_rate": 4.43166936291503e-05, + "loss": 0.0, + "step": 2951 + }, + { + "epoch": 0.68, + "learning_rate": 4.431476773746245e-05, + "loss": 0.0035, + "step": 2952 + }, + { + "epoch": 0.68, + "learning_rate": 4.4312841845774594e-05, + "loss": 0.004, + "step": 2953 + }, + { + "epoch": 0.68, + "learning_rate": 4.431091595408675e-05, + "loss": 0.019, + "step": 2954 + }, + { + "epoch": 0.68, + "learning_rate": 4.430899006239889e-05, + "loss": 0.0, + "step": 2955 + }, + { + "epoch": 0.68, + "learning_rate": 4.4307064170711043e-05, + "loss": 0.0008, + "step": 2956 + }, + { + "epoch": 0.68, + "learning_rate": 4.430513827902319e-05, + "loss": 0.0, + "step": 2957 + }, + { + "epoch": 0.68, + "learning_rate": 4.430321238733534e-05, + "loss": 0.0001, + "step": 2958 + }, + { + "epoch": 0.68, + "learning_rate": 4.4301286495647486e-05, + "loss": 0.0002, + "step": 2959 + }, + { + "epoch": 0.68, + "learning_rate": 4.429936060395963e-05, + "loss": 0.0001, + "step": 2960 + }, + { + "epoch": 0.68, + "learning_rate": 4.429743471227179e-05, + "loss": 0.0025, + "step": 2961 + }, + { + "epoch": 0.68, + "learning_rate": 4.429550882058393e-05, + "loss": 0.0, + "step": 2962 + }, + { + "epoch": 0.68, + "learning_rate": 4.429358292889608e-05, + "loss": 0.0001, + "step": 2963 + }, + { + "epoch": 0.69, + "learning_rate": 4.429165703720823e-05, + "loss": 0.0, + "step": 2964 + }, + { + "epoch": 0.69, + "learning_rate": 4.428973114552038e-05, + "loss": 0.0, + "step": 2965 + }, + { + "epoch": 0.69, + "learning_rate": 4.4287805253832525e-05, + "loss": 0.0036, + "step": 2966 + }, + { + "epoch": 0.69, + "learning_rate": 4.428587936214468e-05, + "loss": 0.0002, + "step": 2967 + }, + { + "epoch": 0.69, + "learning_rate": 4.428395347045683e-05, + "loss": 0.0001, + "step": 2968 + }, + { + "epoch": 0.69, + "learning_rate": 4.428202757876897e-05, + "loss": 0.0006, + "step": 2969 + }, + { + "epoch": 0.69, + "learning_rate": 4.428010168708112e-05, + "loss": 0.0114, + "step": 2970 + }, + { + "epoch": 0.69, + "learning_rate": 4.427817579539327e-05, + "loss": 0.0008, + "step": 2971 + }, + { + "epoch": 0.69, + "learning_rate": 4.427624990370542e-05, + "loss": 0.0034, + "step": 2972 + }, + { + "epoch": 0.69, + "learning_rate": 4.4274324012017564e-05, + "loss": 0.0023, + "step": 2973 + }, + { + "epoch": 0.69, + "learning_rate": 4.427239812032972e-05, + "loss": 0.0, + "step": 2974 + }, + { + "epoch": 0.69, + "learning_rate": 4.4270472228641866e-05, + "loss": 0.0004, + "step": 2975 + }, + { + "epoch": 0.69, + "learning_rate": 4.426854633695401e-05, + "loss": 0.0, + "step": 2976 + }, + { + "epoch": 0.69, + "learning_rate": 4.426662044526616e-05, + "loss": 0.0008, + "step": 2977 + }, + { + "epoch": 0.69, + "learning_rate": 4.426469455357831e-05, + "loss": 0.0049, + "step": 2978 + }, + { + "epoch": 0.69, + "learning_rate": 4.4262768661890456e-05, + "loss": 0.0178, + "step": 2979 + }, + { + "epoch": 0.69, + "learning_rate": 4.42608427702026e-05, + "loss": 0.0003, + "step": 2980 + }, + { + "epoch": 0.69, + "learning_rate": 4.425891687851476e-05, + "loss": 0.0001, + "step": 2981 + }, + { + "epoch": 0.69, + "learning_rate": 4.4256990986826905e-05, + "loss": 0.0095, + "step": 2982 + }, + { + "epoch": 0.69, + "learning_rate": 4.425506509513905e-05, + "loss": 0.0125, + "step": 2983 + }, + { + "epoch": 0.69, + "learning_rate": 4.42531392034512e-05, + "loss": 0.0004, + "step": 2984 + }, + { + "epoch": 0.69, + "learning_rate": 4.4251213311763354e-05, + "loss": 0.0046, + "step": 2985 + }, + { + "epoch": 0.69, + "learning_rate": 4.4249287420075495e-05, + "loss": 0.004, + "step": 2986 + }, + { + "epoch": 0.69, + "learning_rate": 4.424736152838764e-05, + "loss": 0.0003, + "step": 2987 + }, + { + "epoch": 0.69, + "learning_rate": 4.42454356366998e-05, + "loss": 0.0001, + "step": 2988 + }, + { + "epoch": 0.69, + "learning_rate": 4.424350974501194e-05, + "loss": 0.0034, + "step": 2989 + }, + { + "epoch": 0.69, + "learning_rate": 4.424158385332409e-05, + "loss": 0.0, + "step": 2990 + }, + { + "epoch": 0.69, + "learning_rate": 4.423965796163624e-05, + "loss": 0.0, + "step": 2991 + }, + { + "epoch": 0.69, + "learning_rate": 4.423773206994839e-05, + "loss": 0.0001, + "step": 2992 + }, + { + "epoch": 0.69, + "learning_rate": 4.4235806178260534e-05, + "loss": 0.0, + "step": 2993 + }, + { + "epoch": 0.69, + "learning_rate": 4.423388028657269e-05, + "loss": 0.0007, + "step": 2994 + }, + { + "epoch": 0.69, + "learning_rate": 4.4231954394884836e-05, + "loss": 0.0, + "step": 2995 + }, + { + "epoch": 0.69, + "learning_rate": 4.4230028503196976e-05, + "loss": 0.0011, + "step": 2996 + }, + { + "epoch": 0.69, + "learning_rate": 4.422810261150913e-05, + "loss": 0.0014, + "step": 2997 + }, + { + "epoch": 0.69, + "learning_rate": 4.422617671982128e-05, + "loss": 0.0004, + "step": 2998 + }, + { + "epoch": 0.69, + "learning_rate": 4.4224250828133426e-05, + "loss": 0.0002, + "step": 2999 + }, + { + "epoch": 0.69, + "learning_rate": 4.422232493644557e-05, + "loss": 0.0003, + "step": 3000 + }, + { + "epoch": 0.69, + "learning_rate": 4.422039904475773e-05, + "loss": 0.0, + "step": 3001 + }, + { + "epoch": 0.69, + "learning_rate": 4.4218473153069875e-05, + "loss": 0.0642, + "step": 3002 + }, + { + "epoch": 0.69, + "learning_rate": 4.421654726138202e-05, + "loss": 0.0136, + "step": 3003 + }, + { + "epoch": 0.69, + "learning_rate": 4.421462136969417e-05, + "loss": 0.0, + "step": 3004 + }, + { + "epoch": 0.69, + "learning_rate": 4.421269547800632e-05, + "loss": 0.0, + "step": 3005 + }, + { + "epoch": 0.69, + "learning_rate": 4.4210769586318465e-05, + "loss": 0.0, + "step": 3006 + }, + { + "epoch": 0.69, + "learning_rate": 4.420884369463061e-05, + "loss": 0.0168, + "step": 3007 + }, + { + "epoch": 0.7, + "learning_rate": 4.4206917802942767e-05, + "loss": 0.002, + "step": 3008 + }, + { + "epoch": 0.7, + "learning_rate": 4.4204991911254914e-05, + "loss": 0.0062, + "step": 3009 + }, + { + "epoch": 0.7, + "learning_rate": 4.420306601956706e-05, + "loss": 0.0002, + "step": 3010 + }, + { + "epoch": 0.7, + "learning_rate": 4.420114012787921e-05, + "loss": 0.0004, + "step": 3011 + }, + { + "epoch": 0.7, + "learning_rate": 4.419921423619136e-05, + "loss": 0.0296, + "step": 3012 + }, + { + "epoch": 0.7, + "learning_rate": 4.4197288344503504e-05, + "loss": 0.0003, + "step": 3013 + }, + { + "epoch": 0.7, + "learning_rate": 4.419536245281566e-05, + "loss": 0.0003, + "step": 3014 + }, + { + "epoch": 0.7, + "learning_rate": 4.4193436561127806e-05, + "loss": 0.0007, + "step": 3015 + }, + { + "epoch": 0.7, + "learning_rate": 4.419151066943995e-05, + "loss": 0.0062, + "step": 3016 + }, + { + "epoch": 0.7, + "learning_rate": 4.41895847777521e-05, + "loss": 0.0, + "step": 3017 + }, + { + "epoch": 0.7, + "learning_rate": 4.418765888606425e-05, + "loss": 0.0033, + "step": 3018 + }, + { + "epoch": 0.7, + "learning_rate": 4.41857329943764e-05, + "loss": 0.0087, + "step": 3019 + }, + { + "epoch": 0.7, + "learning_rate": 4.418380710268854e-05, + "loss": 0.001, + "step": 3020 + }, + { + "epoch": 0.7, + "learning_rate": 4.41818812110007e-05, + "loss": 0.005, + "step": 3021 + }, + { + "epoch": 0.7, + "learning_rate": 4.4179955319312845e-05, + "loss": 0.0008, + "step": 3022 + }, + { + "epoch": 0.7, + "learning_rate": 4.417802942762499e-05, + "loss": 0.0, + "step": 3023 + }, + { + "epoch": 0.7, + "learning_rate": 4.417610353593714e-05, + "loss": 0.0, + "step": 3024 + }, + { + "epoch": 0.7, + "learning_rate": 4.417417764424929e-05, + "loss": 0.0006, + "step": 3025 + }, + { + "epoch": 0.7, + "learning_rate": 4.417225175256144e-05, + "loss": 0.0001, + "step": 3026 + }, + { + "epoch": 0.7, + "learning_rate": 4.417032586087358e-05, + "loss": 0.0147, + "step": 3027 + }, + { + "epoch": 0.7, + "learning_rate": 4.4168399969185736e-05, + "loss": 0.0, + "step": 3028 + }, + { + "epoch": 0.7, + "learning_rate": 4.4166474077497884e-05, + "loss": 0.0, + "step": 3029 + }, + { + "epoch": 0.7, + "learning_rate": 4.416454818581003e-05, + "loss": 0.0044, + "step": 3030 + }, + { + "epoch": 0.7, + "learning_rate": 4.416262229412218e-05, + "loss": 0.0017, + "step": 3031 + }, + { + "epoch": 0.7, + "learning_rate": 4.416069640243433e-05, + "loss": 0.0006, + "step": 3032 + }, + { + "epoch": 0.7, + "learning_rate": 4.415877051074648e-05, + "loss": 0.0001, + "step": 3033 + }, + { + "epoch": 0.7, + "learning_rate": 4.415684461905862e-05, + "loss": 0.0006, + "step": 3034 + }, + { + "epoch": 0.7, + "learning_rate": 4.4154918727370776e-05, + "loss": 0.007, + "step": 3035 + }, + { + "epoch": 0.7, + "learning_rate": 4.415299283568292e-05, + "loss": 0.0041, + "step": 3036 + }, + { + "epoch": 0.7, + "learning_rate": 4.415106694399507e-05, + "loss": 0.0003, + "step": 3037 + }, + { + "epoch": 0.7, + "learning_rate": 4.414914105230722e-05, + "loss": 0.0083, + "step": 3038 + }, + { + "epoch": 0.7, + "learning_rate": 4.414721516061937e-05, + "loss": 0.0, + "step": 3039 + }, + { + "epoch": 0.7, + "learning_rate": 4.414528926893152e-05, + "loss": 0.0003, + "step": 3040 + }, + { + "epoch": 0.7, + "learning_rate": 4.414336337724367e-05, + "loss": 0.0001, + "step": 3041 + }, + { + "epoch": 0.7, + "learning_rate": 4.4141437485555815e-05, + "loss": 0.0, + "step": 3042 + }, + { + "epoch": 0.7, + "learning_rate": 4.413951159386796e-05, + "loss": 0.0005, + "step": 3043 + }, + { + "epoch": 0.7, + "learning_rate": 4.413758570218011e-05, + "loss": 0.0002, + "step": 3044 + }, + { + "epoch": 0.7, + "learning_rate": 4.413565981049226e-05, + "loss": 0.0, + "step": 3045 + }, + { + "epoch": 0.7, + "learning_rate": 4.413373391880441e-05, + "loss": 0.0006, + "step": 3046 + }, + { + "epoch": 0.7, + "learning_rate": 4.413180802711656e-05, + "loss": 0.0002, + "step": 3047 + }, + { + "epoch": 0.7, + "learning_rate": 4.4129882135428706e-05, + "loss": 0.0, + "step": 3048 + }, + { + "epoch": 0.7, + "learning_rate": 4.4127956243740854e-05, + "loss": 0.0, + "step": 3049 + }, + { + "epoch": 0.7, + "learning_rate": 4.412603035205301e-05, + "loss": 0.0, + "step": 3050 + }, + { + "epoch": 0.71, + "learning_rate": 4.412410446036515e-05, + "loss": 0.0, + "step": 3051 + }, + { + "epoch": 0.71, + "learning_rate": 4.4122178568677296e-05, + "loss": 0.0104, + "step": 3052 + }, + { + "epoch": 0.71, + "learning_rate": 4.412025267698945e-05, + "loss": 0.0001, + "step": 3053 + }, + { + "epoch": 0.71, + "learning_rate": 4.41183267853016e-05, + "loss": 0.0016, + "step": 3054 + }, + { + "epoch": 0.71, + "learning_rate": 4.4116400893613745e-05, + "loss": 0.0, + "step": 3055 + }, + { + "epoch": 0.71, + "learning_rate": 4.411447500192589e-05, + "loss": 0.0006, + "step": 3056 + }, + { + "epoch": 0.71, + "learning_rate": 4.411254911023805e-05, + "loss": 0.0002, + "step": 3057 + }, + { + "epoch": 0.71, + "learning_rate": 4.411062321855019e-05, + "loss": 0.0002, + "step": 3058 + }, + { + "epoch": 0.71, + "learning_rate": 4.410869732686234e-05, + "loss": 0.0, + "step": 3059 + }, + { + "epoch": 0.71, + "learning_rate": 4.410677143517449e-05, + "loss": 0.0004, + "step": 3060 + }, + { + "epoch": 0.71, + "learning_rate": 4.410484554348664e-05, + "loss": 0.0011, + "step": 3061 + }, + { + "epoch": 0.71, + "learning_rate": 4.4102919651798785e-05, + "loss": 0.0001, + "step": 3062 + }, + { + "epoch": 0.71, + "learning_rate": 4.410099376011093e-05, + "loss": 0.0001, + "step": 3063 + }, + { + "epoch": 0.71, + "learning_rate": 4.4099067868423086e-05, + "loss": 0.0059, + "step": 3064 + }, + { + "epoch": 0.71, + "learning_rate": 4.409714197673523e-05, + "loss": 0.001, + "step": 3065 + }, + { + "epoch": 0.71, + "learning_rate": 4.409521608504738e-05, + "loss": 0.0, + "step": 3066 + }, + { + "epoch": 0.71, + "learning_rate": 4.409329019335953e-05, + "loss": 0.0004, + "step": 3067 + }, + { + "epoch": 0.71, + "learning_rate": 4.4091364301671676e-05, + "loss": 0.0, + "step": 3068 + }, + { + "epoch": 0.71, + "learning_rate": 4.4089438409983824e-05, + "loss": 0.0127, + "step": 3069 + }, + { + "epoch": 0.71, + "learning_rate": 4.408751251829598e-05, + "loss": 0.0004, + "step": 3070 + }, + { + "epoch": 0.71, + "learning_rate": 4.408558662660812e-05, + "loss": 0.0019, + "step": 3071 + }, + { + "epoch": 0.71, + "learning_rate": 4.4083660734920266e-05, + "loss": 0.0012, + "step": 3072 + }, + { + "epoch": 0.71, + "learning_rate": 4.408173484323242e-05, + "loss": 0.0014, + "step": 3073 + }, + { + "epoch": 0.71, + "learning_rate": 4.407980895154457e-05, + "loss": 0.0073, + "step": 3074 + }, + { + "epoch": 0.71, + "learning_rate": 4.4077883059856715e-05, + "loss": 0.0004, + "step": 3075 + }, + { + "epoch": 0.71, + "learning_rate": 4.407595716816886e-05, + "loss": 0.0002, + "step": 3076 + }, + { + "epoch": 0.71, + "learning_rate": 4.407403127648102e-05, + "loss": 0.0004, + "step": 3077 + }, + { + "epoch": 0.71, + "learning_rate": 4.407210538479316e-05, + "loss": 0.0, + "step": 3078 + }, + { + "epoch": 0.71, + "learning_rate": 4.407017949310531e-05, + "loss": 0.0092, + "step": 3079 + }, + { + "epoch": 0.71, + "learning_rate": 4.406825360141746e-05, + "loss": 0.0001, + "step": 3080 + }, + { + "epoch": 0.71, + "learning_rate": 4.406632770972961e-05, + "loss": 0.0008, + "step": 3081 + }, + { + "epoch": 0.71, + "learning_rate": 4.4064401818041754e-05, + "loss": 0.0002, + "step": 3082 + }, + { + "epoch": 0.71, + "learning_rate": 4.40624759263539e-05, + "loss": 0.0033, + "step": 3083 + }, + { + "epoch": 0.71, + "learning_rate": 4.4060550034666056e-05, + "loss": 0.0011, + "step": 3084 + }, + { + "epoch": 0.71, + "learning_rate": 4.40586241429782e-05, + "loss": 0.002, + "step": 3085 + }, + { + "epoch": 0.71, + "learning_rate": 4.405669825129035e-05, + "loss": 0.0028, + "step": 3086 + }, + { + "epoch": 0.71, + "learning_rate": 4.40547723596025e-05, + "loss": 0.0, + "step": 3087 + }, + { + "epoch": 0.71, + "learning_rate": 4.4052846467914646e-05, + "loss": 0.0007, + "step": 3088 + }, + { + "epoch": 0.71, + "learning_rate": 4.4050920576226794e-05, + "loss": 0.0002, + "step": 3089 + }, + { + "epoch": 0.71, + "learning_rate": 4.404899468453894e-05, + "loss": 0.0004, + "step": 3090 + }, + { + "epoch": 0.71, + "learning_rate": 4.4047068792851095e-05, + "loss": 0.0, + "step": 3091 + }, + { + "epoch": 0.71, + "learning_rate": 4.4045142901163236e-05, + "loss": 0.0006, + "step": 3092 + }, + { + "epoch": 0.71, + "learning_rate": 4.404321700947539e-05, + "loss": 0.0, + "step": 3093 + }, + { + "epoch": 0.72, + "learning_rate": 4.404129111778754e-05, + "loss": 0.0033, + "step": 3094 + }, + { + "epoch": 0.72, + "learning_rate": 4.4039365226099685e-05, + "loss": 0.0004, + "step": 3095 + }, + { + "epoch": 0.72, + "learning_rate": 4.403743933441183e-05, + "loss": 0.0131, + "step": 3096 + }, + { + "epoch": 0.72, + "learning_rate": 4.403551344272399e-05, + "loss": 0.0017, + "step": 3097 + }, + { + "epoch": 0.72, + "learning_rate": 4.4033587551036134e-05, + "loss": 0.0, + "step": 3098 + }, + { + "epoch": 0.72, + "learning_rate": 4.4031661659348275e-05, + "loss": 0.0001, + "step": 3099 + }, + { + "epoch": 0.72, + "learning_rate": 4.402973576766043e-05, + "loss": 0.0005, + "step": 3100 + }, + { + "epoch": 0.72, + "learning_rate": 4.402780987597258e-05, + "loss": 0.0, + "step": 3101 + }, + { + "epoch": 0.72, + "learning_rate": 4.4025883984284724e-05, + "loss": 0.0001, + "step": 3102 + }, + { + "epoch": 0.72, + "learning_rate": 4.402395809259687e-05, + "loss": 0.0008, + "step": 3103 + }, + { + "epoch": 0.72, + "learning_rate": 4.4022032200909026e-05, + "loss": 0.0002, + "step": 3104 + }, + { + "epoch": 0.72, + "learning_rate": 4.4020106309221173e-05, + "loss": 0.0045, + "step": 3105 + }, + { + "epoch": 0.72, + "learning_rate": 4.401818041753332e-05, + "loss": 0.0099, + "step": 3106 + }, + { + "epoch": 0.72, + "learning_rate": 4.401625452584547e-05, + "loss": 0.0001, + "step": 3107 + }, + { + "epoch": 0.72, + "learning_rate": 4.4014328634157616e-05, + "loss": 0.0007, + "step": 3108 + }, + { + "epoch": 0.72, + "learning_rate": 4.4012402742469763e-05, + "loss": 0.0409, + "step": 3109 + }, + { + "epoch": 0.72, + "learning_rate": 4.401047685078191e-05, + "loss": 0.0016, + "step": 3110 + }, + { + "epoch": 0.72, + "learning_rate": 4.4008550959094065e-05, + "loss": 0.0052, + "step": 3111 + }, + { + "epoch": 0.72, + "learning_rate": 4.400662506740621e-05, + "loss": 0.0, + "step": 3112 + }, + { + "epoch": 0.72, + "learning_rate": 4.400469917571836e-05, + "loss": 0.0007, + "step": 3113 + }, + { + "epoch": 0.72, + "learning_rate": 4.400277328403051e-05, + "loss": 0.0, + "step": 3114 + }, + { + "epoch": 0.72, + "learning_rate": 4.400084739234266e-05, + "loss": 0.0002, + "step": 3115 + }, + { + "epoch": 0.72, + "learning_rate": 4.39989215006548e-05, + "loss": 0.0005, + "step": 3116 + }, + { + "epoch": 0.72, + "learning_rate": 4.399699560896695e-05, + "loss": 0.0, + "step": 3117 + }, + { + "epoch": 0.72, + "learning_rate": 4.3995069717279104e-05, + "loss": 0.0004, + "step": 3118 + }, + { + "epoch": 0.72, + "learning_rate": 4.399314382559125e-05, + "loss": 0.0212, + "step": 3119 + }, + { + "epoch": 0.72, + "learning_rate": 4.39912179339034e-05, + "loss": 0.0025, + "step": 3120 + }, + { + "epoch": 0.72, + "learning_rate": 4.398929204221555e-05, + "loss": 0.0003, + "step": 3121 + }, + { + "epoch": 0.72, + "learning_rate": 4.39873661505277e-05, + "loss": 0.0028, + "step": 3122 + }, + { + "epoch": 0.72, + "learning_rate": 4.398544025883984e-05, + "loss": 0.0001, + "step": 3123 + }, + { + "epoch": 0.72, + "learning_rate": 4.3983514367151996e-05, + "loss": 0.0, + "step": 3124 + }, + { + "epoch": 0.72, + "learning_rate": 4.398158847546414e-05, + "loss": 0.0001, + "step": 3125 + }, + { + "epoch": 0.72, + "learning_rate": 4.397966258377629e-05, + "loss": 0.0002, + "step": 3126 + }, + { + "epoch": 0.72, + "learning_rate": 4.397773669208844e-05, + "loss": 0.0001, + "step": 3127 + }, + { + "epoch": 0.72, + "learning_rate": 4.3975810800400586e-05, + "loss": 0.0014, + "step": 3128 + }, + { + "epoch": 0.72, + "learning_rate": 4.397388490871274e-05, + "loss": 0.0001, + "step": 3129 + }, + { + "epoch": 0.72, + "learning_rate": 4.397195901702488e-05, + "loss": 0.0021, + "step": 3130 + }, + { + "epoch": 0.72, + "learning_rate": 4.3970033125337035e-05, + "loss": 0.0036, + "step": 3131 + }, + { + "epoch": 0.72, + "learning_rate": 4.396810723364918e-05, + "loss": 0.0001, + "step": 3132 + }, + { + "epoch": 0.72, + "learning_rate": 4.396618134196133e-05, + "loss": 0.0, + "step": 3133 + }, + { + "epoch": 0.72, + "learning_rate": 4.396425545027348e-05, + "loss": 0.0011, + "step": 3134 + }, + { + "epoch": 0.72, + "learning_rate": 4.396232955858563e-05, + "loss": 0.0008, + "step": 3135 + }, + { + "epoch": 0.72, + "learning_rate": 4.396040366689778e-05, + "loss": 0.0001, + "step": 3136 + }, + { + "epoch": 0.72, + "learning_rate": 4.395847777520992e-05, + "loss": 0.0003, + "step": 3137 + }, + { + "epoch": 0.73, + "learning_rate": 4.3956551883522074e-05, + "loss": 0.0, + "step": 3138 + }, + { + "epoch": 0.73, + "learning_rate": 4.395462599183422e-05, + "loss": 0.0146, + "step": 3139 + }, + { + "epoch": 0.73, + "learning_rate": 4.395270010014637e-05, + "loss": 0.0024, + "step": 3140 + }, + { + "epoch": 0.73, + "learning_rate": 4.3950774208458517e-05, + "loss": 0.0002, + "step": 3141 + }, + { + "epoch": 0.73, + "learning_rate": 4.394884831677067e-05, + "loss": 0.0106, + "step": 3142 + }, + { + "epoch": 0.73, + "learning_rate": 4.394692242508281e-05, + "loss": 0.0004, + "step": 3143 + }, + { + "epoch": 0.73, + "learning_rate": 4.3944996533394966e-05, + "loss": 0.0201, + "step": 3144 + }, + { + "epoch": 0.73, + "learning_rate": 4.394307064170711e-05, + "loss": 0.0007, + "step": 3145 + }, + { + "epoch": 0.73, + "learning_rate": 4.394114475001926e-05, + "loss": 0.0, + "step": 3146 + }, + { + "epoch": 0.73, + "learning_rate": 4.393921885833141e-05, + "loss": 0.0, + "step": 3147 + }, + { + "epoch": 0.73, + "learning_rate": 4.3937292966643556e-05, + "loss": 0.0017, + "step": 3148 + }, + { + "epoch": 0.73, + "learning_rate": 4.393536707495571e-05, + "loss": 0.0, + "step": 3149 + }, + { + "epoch": 0.73, + "learning_rate": 4.393344118326785e-05, + "loss": 0.0047, + "step": 3150 + }, + { + "epoch": 0.73, + "learning_rate": 4.3931515291580005e-05, + "loss": 0.0001, + "step": 3151 + }, + { + "epoch": 0.73, + "learning_rate": 4.392958939989215e-05, + "loss": 0.0017, + "step": 3152 + }, + { + "epoch": 0.73, + "learning_rate": 4.39276635082043e-05, + "loss": 0.0008, + "step": 3153 + }, + { + "epoch": 0.73, + "learning_rate": 4.392573761651645e-05, + "loss": 0.0001, + "step": 3154 + }, + { + "epoch": 0.73, + "learning_rate": 4.3923811724828595e-05, + "loss": 0.0003, + "step": 3155 + }, + { + "epoch": 0.73, + "learning_rate": 4.392188583314075e-05, + "loss": 0.0032, + "step": 3156 + }, + { + "epoch": 0.73, + "learning_rate": 4.391995994145289e-05, + "loss": 0.0025, + "step": 3157 + }, + { + "epoch": 0.73, + "learning_rate": 4.3918034049765044e-05, + "loss": 0.0, + "step": 3158 + }, + { + "epoch": 0.73, + "learning_rate": 4.391610815807719e-05, + "loss": 0.0001, + "step": 3159 + }, + { + "epoch": 0.73, + "learning_rate": 4.391418226638934e-05, + "loss": 0.0032, + "step": 3160 + }, + { + "epoch": 0.73, + "learning_rate": 4.3912256374701486e-05, + "loss": 0.0, + "step": 3161 + }, + { + "epoch": 0.73, + "learning_rate": 4.391033048301364e-05, + "loss": 0.0005, + "step": 3162 + }, + { + "epoch": 0.73, + "learning_rate": 4.390840459132579e-05, + "loss": 0.0002, + "step": 3163 + }, + { + "epoch": 0.73, + "learning_rate": 4.390647869963793e-05, + "loss": 0.0001, + "step": 3164 + }, + { + "epoch": 0.73, + "learning_rate": 4.390455280795008e-05, + "loss": 0.0, + "step": 3165 + }, + { + "epoch": 0.73, + "learning_rate": 4.390262691626223e-05, + "loss": 0.0048, + "step": 3166 + }, + { + "epoch": 0.73, + "learning_rate": 4.390070102457438e-05, + "loss": 0.0001, + "step": 3167 + }, + { + "epoch": 0.73, + "learning_rate": 4.3898775132886526e-05, + "loss": 0.0025, + "step": 3168 + }, + { + "epoch": 0.73, + "learning_rate": 4.389684924119868e-05, + "loss": 0.0, + "step": 3169 + }, + { + "epoch": 0.73, + "learning_rate": 4.389492334951083e-05, + "loss": 0.0001, + "step": 3170 + }, + { + "epoch": 0.73, + "learning_rate": 4.3892997457822975e-05, + "loss": 0.0, + "step": 3171 + }, + { + "epoch": 0.73, + "learning_rate": 4.389107156613512e-05, + "loss": 0.0001, + "step": 3172 + }, + { + "epoch": 0.73, + "learning_rate": 4.388914567444727e-05, + "loss": 0.0123, + "step": 3173 + }, + { + "epoch": 0.73, + "learning_rate": 4.388721978275942e-05, + "loss": 0.0088, + "step": 3174 + }, + { + "epoch": 0.73, + "learning_rate": 4.3885293891071565e-05, + "loss": 0.0, + "step": 3175 + }, + { + "epoch": 0.73, + "learning_rate": 4.388336799938372e-05, + "loss": 0.0002, + "step": 3176 + }, + { + "epoch": 0.73, + "learning_rate": 4.3881442107695866e-05, + "loss": 0.0024, + "step": 3177 + }, + { + "epoch": 0.73, + "learning_rate": 4.3879516216008014e-05, + "loss": 0.0002, + "step": 3178 + }, + { + "epoch": 0.73, + "learning_rate": 4.387759032432016e-05, + "loss": 0.0002, + "step": 3179 + }, + { + "epoch": 0.73, + "learning_rate": 4.3875664432632316e-05, + "loss": 0.0004, + "step": 3180 + }, + { + "epoch": 0.74, + "learning_rate": 4.3873738540944456e-05, + "loss": 0.0, + "step": 3181 + }, + { + "epoch": 0.74, + "learning_rate": 4.387181264925661e-05, + "loss": 0.0007, + "step": 3182 + }, + { + "epoch": 0.74, + "learning_rate": 4.386988675756876e-05, + "loss": 0.0012, + "step": 3183 + }, + { + "epoch": 0.74, + "learning_rate": 4.3867960865880906e-05, + "loss": 0.0, + "step": 3184 + }, + { + "epoch": 0.74, + "learning_rate": 4.386603497419305e-05, + "loss": 0.0, + "step": 3185 + }, + { + "epoch": 0.74, + "learning_rate": 4.38641090825052e-05, + "loss": 0.0, + "step": 3186 + }, + { + "epoch": 0.74, + "learning_rate": 4.3862183190817355e-05, + "loss": 0.0005, + "step": 3187 + }, + { + "epoch": 0.74, + "learning_rate": 4.3860257299129495e-05, + "loss": 0.0, + "step": 3188 + }, + { + "epoch": 0.74, + "learning_rate": 4.385833140744165e-05, + "loss": 0.0, + "step": 3189 + }, + { + "epoch": 0.74, + "learning_rate": 4.38564055157538e-05, + "loss": 0.0005, + "step": 3190 + }, + { + "epoch": 0.74, + "learning_rate": 4.3854479624065945e-05, + "loss": 0.0, + "step": 3191 + }, + { + "epoch": 0.74, + "learning_rate": 4.385255373237809e-05, + "loss": 0.0001, + "step": 3192 + }, + { + "epoch": 0.74, + "learning_rate": 4.385062784069024e-05, + "loss": 0.0001, + "step": 3193 + }, + { + "epoch": 0.74, + "learning_rate": 4.3848701949002394e-05, + "loss": 0.0017, + "step": 3194 + }, + { + "epoch": 0.74, + "learning_rate": 4.3846776057314535e-05, + "loss": 0.0068, + "step": 3195 + }, + { + "epoch": 0.74, + "learning_rate": 4.384485016562669e-05, + "loss": 0.0002, + "step": 3196 + }, + { + "epoch": 0.74, + "learning_rate": 4.3842924273938836e-05, + "loss": 0.001, + "step": 3197 + }, + { + "epoch": 0.74, + "learning_rate": 4.3840998382250984e-05, + "loss": 0.0, + "step": 3198 + }, + { + "epoch": 0.74, + "learning_rate": 4.383907249056313e-05, + "loss": 0.0, + "step": 3199 + }, + { + "epoch": 0.74, + "learning_rate": 4.3837146598875286e-05, + "loss": 0.0, + "step": 3200 + }, + { + "epoch": 0.74, + "learning_rate": 4.383522070718743e-05, + "loss": 0.0081, + "step": 3201 + }, + { + "epoch": 0.74, + "learning_rate": 4.3833294815499574e-05, + "loss": 0.0025, + "step": 3202 + }, + { + "epoch": 0.74, + "learning_rate": 4.383136892381173e-05, + "loss": 0.0001, + "step": 3203 + }, + { + "epoch": 0.74, + "learning_rate": 4.3829443032123875e-05, + "loss": 0.0001, + "step": 3204 + }, + { + "epoch": 0.74, + "learning_rate": 4.382751714043602e-05, + "loss": 0.0, + "step": 3205 + }, + { + "epoch": 0.74, + "learning_rate": 4.382559124874817e-05, + "loss": 0.0004, + "step": 3206 + }, + { + "epoch": 0.74, + "learning_rate": 4.3823665357060325e-05, + "loss": 0.0263, + "step": 3207 + }, + { + "epoch": 0.74, + "learning_rate": 4.382173946537247e-05, + "loss": 0.0, + "step": 3208 + }, + { + "epoch": 0.74, + "learning_rate": 4.381981357368462e-05, + "loss": 0.0002, + "step": 3209 + }, + { + "epoch": 0.74, + "learning_rate": 4.381788768199677e-05, + "loss": 0.0001, + "step": 3210 + }, + { + "epoch": 0.74, + "learning_rate": 4.3815961790308915e-05, + "loss": 0.0, + "step": 3211 + }, + { + "epoch": 0.74, + "learning_rate": 4.381403589862106e-05, + "loss": 0.0, + "step": 3212 + }, + { + "epoch": 0.74, + "learning_rate": 4.381211000693321e-05, + "loss": 0.0004, + "step": 3213 + }, + { + "epoch": 0.74, + "learning_rate": 4.3810184115245364e-05, + "loss": 0.0371, + "step": 3214 + }, + { + "epoch": 0.74, + "learning_rate": 4.380825822355751e-05, + "loss": 0.0012, + "step": 3215 + }, + { + "epoch": 0.74, + "learning_rate": 4.380633233186966e-05, + "loss": 0.0001, + "step": 3216 + }, + { + "epoch": 0.74, + "learning_rate": 4.3804406440181806e-05, + "loss": 0.0001, + "step": 3217 + }, + { + "epoch": 0.74, + "learning_rate": 4.380248054849396e-05, + "loss": 0.0, + "step": 3218 + }, + { + "epoch": 0.74, + "learning_rate": 4.38005546568061e-05, + "loss": 0.002, + "step": 3219 + }, + { + "epoch": 0.74, + "learning_rate": 4.379862876511825e-05, + "loss": 0.0031, + "step": 3220 + }, + { + "epoch": 0.74, + "learning_rate": 4.37967028734304e-05, + "loss": 0.0002, + "step": 3221 + }, + { + "epoch": 0.74, + "learning_rate": 4.3794776981742544e-05, + "loss": 0.0002, + "step": 3222 + }, + { + "epoch": 0.74, + "learning_rate": 4.37928510900547e-05, + "loss": 0.0002, + "step": 3223 + }, + { + "epoch": 0.75, + "learning_rate": 4.3790925198366845e-05, + "loss": 0.0015, + "step": 3224 + }, + { + "epoch": 0.75, + "learning_rate": 4.378899930667899e-05, + "loss": 0.0, + "step": 3225 + }, + { + "epoch": 0.75, + "learning_rate": 4.378707341499114e-05, + "loss": 0.0004, + "step": 3226 + }, + { + "epoch": 0.75, + "learning_rate": 4.3785147523303295e-05, + "loss": 0.0002, + "step": 3227 + }, + { + "epoch": 0.75, + "learning_rate": 4.378322163161544e-05, + "loss": 0.001, + "step": 3228 + }, + { + "epoch": 0.75, + "learning_rate": 4.378129573992758e-05, + "loss": 0.0037, + "step": 3229 + }, + { + "epoch": 0.75, + "learning_rate": 4.377936984823974e-05, + "loss": 0.0002, + "step": 3230 + }, + { + "epoch": 0.75, + "learning_rate": 4.3777443956551884e-05, + "loss": 0.0039, + "step": 3231 + }, + { + "epoch": 0.75, + "learning_rate": 4.377551806486403e-05, + "loss": 0.0002, + "step": 3232 + }, + { + "epoch": 0.75, + "learning_rate": 4.377359217317618e-05, + "loss": 0.0013, + "step": 3233 + }, + { + "epoch": 0.75, + "learning_rate": 4.3771666281488334e-05, + "loss": 0.0007, + "step": 3234 + }, + { + "epoch": 0.75, + "learning_rate": 4.376974038980048e-05, + "loss": 0.0004, + "step": 3235 + }, + { + "epoch": 0.75, + "learning_rate": 4.376781449811263e-05, + "loss": 0.0001, + "step": 3236 + }, + { + "epoch": 0.75, + "learning_rate": 4.3765888606424776e-05, + "loss": 0.0015, + "step": 3237 + }, + { + "epoch": 0.75, + "learning_rate": 4.376396271473693e-05, + "loss": 0.0, + "step": 3238 + }, + { + "epoch": 0.75, + "learning_rate": 4.376203682304907e-05, + "loss": 0.0011, + "step": 3239 + }, + { + "epoch": 0.75, + "learning_rate": 4.376011093136122e-05, + "loss": 0.0019, + "step": 3240 + }, + { + "epoch": 0.75, + "learning_rate": 4.375818503967337e-05, + "loss": 0.0009, + "step": 3241 + }, + { + "epoch": 0.75, + "learning_rate": 4.375625914798552e-05, + "loss": 0.0004, + "step": 3242 + }, + { + "epoch": 0.75, + "learning_rate": 4.375433325629767e-05, + "loss": 0.0075, + "step": 3243 + }, + { + "epoch": 0.75, + "learning_rate": 4.3752407364609815e-05, + "loss": 0.0004, + "step": 3244 + }, + { + "epoch": 0.75, + "learning_rate": 4.375048147292197e-05, + "loss": 0.0001, + "step": 3245 + }, + { + "epoch": 0.75, + "learning_rate": 4.374855558123411e-05, + "loss": 0.0, + "step": 3246 + }, + { + "epoch": 0.75, + "learning_rate": 4.3746629689546264e-05, + "loss": 0.0005, + "step": 3247 + }, + { + "epoch": 0.75, + "learning_rate": 4.374470379785841e-05, + "loss": 0.0115, + "step": 3248 + }, + { + "epoch": 0.75, + "learning_rate": 4.374277790617056e-05, + "loss": 0.0, + "step": 3249 + }, + { + "epoch": 0.75, + "learning_rate": 4.374085201448271e-05, + "loss": 0.0, + "step": 3250 + }, + { + "epoch": 0.75, + "learning_rate": 4.3738926122794854e-05, + "loss": 0.0002, + "step": 3251 + }, + { + "epoch": 0.75, + "learning_rate": 4.373700023110701e-05, + "loss": 0.0009, + "step": 3252 + }, + { + "epoch": 0.75, + "learning_rate": 4.373507433941915e-05, + "loss": 0.0097, + "step": 3253 + }, + { + "epoch": 0.75, + "learning_rate": 4.3733148447731304e-05, + "loss": 0.0011, + "step": 3254 + }, + { + "epoch": 0.75, + "learning_rate": 4.373122255604345e-05, + "loss": 0.0, + "step": 3255 + }, + { + "epoch": 0.75, + "learning_rate": 4.37292966643556e-05, + "loss": 0.0006, + "step": 3256 + }, + { + "epoch": 0.75, + "learning_rate": 4.3727370772667746e-05, + "loss": 0.0013, + "step": 3257 + }, + { + "epoch": 0.75, + "learning_rate": 4.3725444880979893e-05, + "loss": 0.0085, + "step": 3258 + }, + { + "epoch": 0.75, + "learning_rate": 4.372351898929205e-05, + "loss": 0.0, + "step": 3259 + }, + { + "epoch": 0.75, + "learning_rate": 4.372159309760419e-05, + "loss": 0.0001, + "step": 3260 + }, + { + "epoch": 0.75, + "learning_rate": 4.371966720591634e-05, + "loss": 0.001, + "step": 3261 + }, + { + "epoch": 0.75, + "learning_rate": 4.371774131422849e-05, + "loss": 0.0, + "step": 3262 + }, + { + "epoch": 0.75, + "learning_rate": 4.371581542254064e-05, + "loss": 0.0001, + "step": 3263 + }, + { + "epoch": 0.75, + "learning_rate": 4.3713889530852785e-05, + "loss": 0.0045, + "step": 3264 + }, + { + "epoch": 0.75, + "learning_rate": 4.371196363916494e-05, + "loss": 0.0018, + "step": 3265 + }, + { + "epoch": 0.75, + "learning_rate": 4.371003774747709e-05, + "loss": 0.0001, + "step": 3266 + }, + { + "epoch": 0.76, + "learning_rate": 4.370811185578923e-05, + "loss": 0.0001, + "step": 3267 + }, + { + "epoch": 0.76, + "learning_rate": 4.370618596410138e-05, + "loss": 0.0, + "step": 3268 + }, + { + "epoch": 0.76, + "learning_rate": 4.370426007241353e-05, + "loss": 0.0, + "step": 3269 + }, + { + "epoch": 0.76, + "learning_rate": 4.370233418072568e-05, + "loss": 0.0002, + "step": 3270 + }, + { + "epoch": 0.76, + "learning_rate": 4.3700408289037824e-05, + "loss": 0.002, + "step": 3271 + }, + { + "epoch": 0.76, + "learning_rate": 4.369848239734998e-05, + "loss": 0.0, + "step": 3272 + }, + { + "epoch": 0.76, + "learning_rate": 4.3696556505662126e-05, + "loss": 0.0004, + "step": 3273 + }, + { + "epoch": 0.76, + "learning_rate": 4.3694630613974273e-05, + "loss": 0.0002, + "step": 3274 + }, + { + "epoch": 0.76, + "learning_rate": 4.369270472228642e-05, + "loss": 0.0, + "step": 3275 + }, + { + "epoch": 0.76, + "learning_rate": 4.369077883059857e-05, + "loss": 0.0013, + "step": 3276 + }, + { + "epoch": 0.76, + "learning_rate": 4.3688852938910716e-05, + "loss": 0.0007, + "step": 3277 + }, + { + "epoch": 0.76, + "learning_rate": 4.368692704722286e-05, + "loss": 0.0001, + "step": 3278 + }, + { + "epoch": 0.76, + "learning_rate": 4.368500115553502e-05, + "loss": 0.0, + "step": 3279 + }, + { + "epoch": 0.76, + "learning_rate": 4.3683075263847165e-05, + "loss": 0.0001, + "step": 3280 + }, + { + "epoch": 0.76, + "learning_rate": 4.368114937215931e-05, + "loss": 0.0003, + "step": 3281 + }, + { + "epoch": 0.76, + "learning_rate": 4.367922348047146e-05, + "loss": 0.0242, + "step": 3282 + }, + { + "epoch": 0.76, + "learning_rate": 4.3677297588783614e-05, + "loss": 0.001, + "step": 3283 + }, + { + "epoch": 0.76, + "learning_rate": 4.3675371697095755e-05, + "loss": 0.0005, + "step": 3284 + }, + { + "epoch": 0.76, + "learning_rate": 4.36734458054079e-05, + "loss": 0.0, + "step": 3285 + }, + { + "epoch": 0.76, + "learning_rate": 4.367151991372006e-05, + "loss": 0.0332, + "step": 3286 + }, + { + "epoch": 0.76, + "learning_rate": 4.3669594022032204e-05, + "loss": 0.0192, + "step": 3287 + }, + { + "epoch": 0.76, + "learning_rate": 4.366766813034435e-05, + "loss": 0.0, + "step": 3288 + }, + { + "epoch": 0.76, + "learning_rate": 4.36657422386565e-05, + "loss": 0.0, + "step": 3289 + }, + { + "epoch": 0.76, + "learning_rate": 4.3663816346968653e-05, + "loss": 0.0003, + "step": 3290 + }, + { + "epoch": 0.76, + "learning_rate": 4.3661890455280794e-05, + "loss": 0.0, + "step": 3291 + }, + { + "epoch": 0.76, + "learning_rate": 4.365996456359295e-05, + "loss": 0.0, + "step": 3292 + }, + { + "epoch": 0.76, + "learning_rate": 4.3658038671905096e-05, + "loss": 0.0002, + "step": 3293 + }, + { + "epoch": 0.76, + "learning_rate": 4.365611278021724e-05, + "loss": 0.0042, + "step": 3294 + }, + { + "epoch": 0.76, + "learning_rate": 4.365418688852939e-05, + "loss": 0.0, + "step": 3295 + }, + { + "epoch": 0.76, + "learning_rate": 4.365226099684154e-05, + "loss": 0.0001, + "step": 3296 + }, + { + "epoch": 0.76, + "learning_rate": 4.3650335105153686e-05, + "loss": 0.0, + "step": 3297 + }, + { + "epoch": 0.76, + "learning_rate": 4.364840921346583e-05, + "loss": 0.0, + "step": 3298 + }, + { + "epoch": 0.76, + "learning_rate": 4.364648332177799e-05, + "loss": 0.0014, + "step": 3299 + }, + { + "epoch": 0.76, + "learning_rate": 4.3644557430090135e-05, + "loss": 0.0001, + "step": 3300 + }, + { + "epoch": 0.76, + "learning_rate": 4.364263153840228e-05, + "loss": 0.0, + "step": 3301 + }, + { + "epoch": 0.76, + "learning_rate": 4.364070564671443e-05, + "loss": 0.003, + "step": 3302 + }, + { + "epoch": 0.76, + "learning_rate": 4.3638779755026584e-05, + "loss": 0.0004, + "step": 3303 + }, + { + "epoch": 0.76, + "learning_rate": 4.3636853863338725e-05, + "loss": 0.0, + "step": 3304 + }, + { + "epoch": 0.76, + "learning_rate": 4.363492797165087e-05, + "loss": 0.003, + "step": 3305 + }, + { + "epoch": 0.76, + "learning_rate": 4.3633002079963027e-05, + "loss": 0.0001, + "step": 3306 + }, + { + "epoch": 0.76, + "learning_rate": 4.3631076188275174e-05, + "loss": 0.0844, + "step": 3307 + }, + { + "epoch": 0.76, + "learning_rate": 4.362915029658732e-05, + "loss": 0.0038, + "step": 3308 + }, + { + "epoch": 0.76, + "learning_rate": 4.362722440489947e-05, + "loss": 0.0058, + "step": 3309 + }, + { + "epoch": 0.76, + "learning_rate": 4.362529851321162e-05, + "loss": 0.0, + "step": 3310 + }, + { + "epoch": 0.77, + "learning_rate": 4.3623372621523764e-05, + "loss": 0.0, + "step": 3311 + }, + { + "epoch": 0.77, + "learning_rate": 4.362144672983592e-05, + "loss": 0.0001, + "step": 3312 + }, + { + "epoch": 0.77, + "learning_rate": 4.3619520838148066e-05, + "loss": 0.0, + "step": 3313 + }, + { + "epoch": 0.77, + "learning_rate": 4.361759494646021e-05, + "loss": 0.0, + "step": 3314 + }, + { + "epoch": 0.77, + "learning_rate": 4.361566905477236e-05, + "loss": 0.0012, + "step": 3315 + }, + { + "epoch": 0.77, + "learning_rate": 4.361374316308451e-05, + "loss": 0.0003, + "step": 3316 + }, + { + "epoch": 0.77, + "learning_rate": 4.361181727139666e-05, + "loss": 0.0039, + "step": 3317 + }, + { + "epoch": 0.77, + "learning_rate": 4.36098913797088e-05, + "loss": 0.0031, + "step": 3318 + }, + { + "epoch": 0.77, + "learning_rate": 4.360796548802096e-05, + "loss": 0.0006, + "step": 3319 + }, + { + "epoch": 0.77, + "learning_rate": 4.3606039596333105e-05, + "loss": 0.0, + "step": 3320 + }, + { + "epoch": 0.77, + "learning_rate": 4.360411370464525e-05, + "loss": 0.0075, + "step": 3321 + }, + { + "epoch": 0.77, + "learning_rate": 4.36021878129574e-05, + "loss": 0.0001, + "step": 3322 + }, + { + "epoch": 0.77, + "learning_rate": 4.360026192126955e-05, + "loss": 0.0, + "step": 3323 + }, + { + "epoch": 0.77, + "learning_rate": 4.35983360295817e-05, + "loss": 0.0001, + "step": 3324 + }, + { + "epoch": 0.77, + "learning_rate": 4.359641013789384e-05, + "loss": 0.0004, + "step": 3325 + }, + { + "epoch": 0.77, + "learning_rate": 4.3594484246205996e-05, + "loss": 0.0001, + "step": 3326 + }, + { + "epoch": 0.77, + "learning_rate": 4.3592558354518144e-05, + "loss": 0.0001, + "step": 3327 + }, + { + "epoch": 0.77, + "learning_rate": 4.359063246283029e-05, + "loss": 0.0002, + "step": 3328 + }, + { + "epoch": 0.77, + "learning_rate": 4.358870657114244e-05, + "loss": 0.0001, + "step": 3329 + }, + { + "epoch": 0.77, + "learning_rate": 4.358678067945459e-05, + "loss": 0.0046, + "step": 3330 + }, + { + "epoch": 0.77, + "learning_rate": 4.358485478776674e-05, + "loss": 0.0002, + "step": 3331 + }, + { + "epoch": 0.77, + "learning_rate": 4.358292889607888e-05, + "loss": 0.0017, + "step": 3332 + }, + { + "epoch": 0.77, + "learning_rate": 4.3581003004391036e-05, + "loss": 0.0001, + "step": 3333 + }, + { + "epoch": 0.77, + "learning_rate": 4.357907711270318e-05, + "loss": 0.0, + "step": 3334 + }, + { + "epoch": 0.77, + "learning_rate": 4.357715122101533e-05, + "loss": 0.0004, + "step": 3335 + }, + { + "epoch": 0.77, + "learning_rate": 4.357522532932748e-05, + "loss": 0.0023, + "step": 3336 + }, + { + "epoch": 0.77, + "learning_rate": 4.357329943763963e-05, + "loss": 0.0004, + "step": 3337 + }, + { + "epoch": 0.77, + "learning_rate": 4.357137354595178e-05, + "loss": 0.0, + "step": 3338 + }, + { + "epoch": 0.77, + "learning_rate": 4.356944765426393e-05, + "loss": 0.0, + "step": 3339 + }, + { + "epoch": 0.77, + "learning_rate": 4.3567521762576075e-05, + "loss": 0.0001, + "step": 3340 + }, + { + "epoch": 0.77, + "learning_rate": 4.356559587088822e-05, + "loss": 0.0, + "step": 3341 + }, + { + "epoch": 0.77, + "learning_rate": 4.356366997920037e-05, + "loss": 0.0004, + "step": 3342 + }, + { + "epoch": 0.77, + "learning_rate": 4.356174408751252e-05, + "loss": 0.0, + "step": 3343 + }, + { + "epoch": 0.77, + "learning_rate": 4.355981819582467e-05, + "loss": 0.0002, + "step": 3344 + }, + { + "epoch": 0.77, + "learning_rate": 4.355789230413682e-05, + "loss": 0.0, + "step": 3345 + }, + { + "epoch": 0.77, + "learning_rate": 4.3555966412448966e-05, + "loss": 0.0006, + "step": 3346 + }, + { + "epoch": 0.77, + "learning_rate": 4.3554040520761114e-05, + "loss": 0.0008, + "step": 3347 + }, + { + "epoch": 0.77, + "learning_rate": 4.355211462907327e-05, + "loss": 0.0086, + "step": 3348 + }, + { + "epoch": 0.77, + "learning_rate": 4.355018873738541e-05, + "loss": 0.0001, + "step": 3349 + }, + { + "epoch": 0.77, + "learning_rate": 4.354826284569756e-05, + "loss": 0.0002, + "step": 3350 + }, + { + "epoch": 0.77, + "learning_rate": 4.354633695400971e-05, + "loss": 0.0003, + "step": 3351 + }, + { + "epoch": 0.77, + "learning_rate": 4.354441106232186e-05, + "loss": 0.0005, + "step": 3352 + }, + { + "epoch": 0.77, + "learning_rate": 4.3542485170634005e-05, + "loss": 0.0014, + "step": 3353 + }, + { + "epoch": 0.78, + "learning_rate": 4.354055927894615e-05, + "loss": 0.0002, + "step": 3354 + }, + { + "epoch": 0.78, + "learning_rate": 4.353863338725831e-05, + "loss": 0.0022, + "step": 3355 + }, + { + "epoch": 0.78, + "learning_rate": 4.353670749557045e-05, + "loss": 0.001, + "step": 3356 + }, + { + "epoch": 0.78, + "learning_rate": 4.35347816038826e-05, + "loss": 0.0001, + "step": 3357 + }, + { + "epoch": 0.78, + "learning_rate": 4.353285571219475e-05, + "loss": 0.0044, + "step": 3358 + }, + { + "epoch": 0.78, + "learning_rate": 4.35309298205069e-05, + "loss": 0.0005, + "step": 3359 + }, + { + "epoch": 0.78, + "learning_rate": 4.3529003928819045e-05, + "loss": 0.0, + "step": 3360 + }, + { + "epoch": 0.78, + "learning_rate": 4.352707803713119e-05, + "loss": 0.0, + "step": 3361 + }, + { + "epoch": 0.78, + "learning_rate": 4.3525152145443346e-05, + "loss": 0.0, + "step": 3362 + }, + { + "epoch": 0.78, + "learning_rate": 4.352322625375549e-05, + "loss": 0.0004, + "step": 3363 + }, + { + "epoch": 0.78, + "learning_rate": 4.352130036206764e-05, + "loss": 0.0004, + "step": 3364 + }, + { + "epoch": 0.78, + "learning_rate": 4.351937447037979e-05, + "loss": 0.0001, + "step": 3365 + }, + { + "epoch": 0.78, + "learning_rate": 4.3517448578691936e-05, + "loss": 0.0008, + "step": 3366 + }, + { + "epoch": 0.78, + "learning_rate": 4.3515522687004084e-05, + "loss": 0.0004, + "step": 3367 + }, + { + "epoch": 0.78, + "learning_rate": 4.351359679531624e-05, + "loss": 0.0011, + "step": 3368 + }, + { + "epoch": 0.78, + "learning_rate": 4.351167090362838e-05, + "loss": 0.0004, + "step": 3369 + }, + { + "epoch": 0.78, + "learning_rate": 4.3509745011940526e-05, + "loss": 0.0004, + "step": 3370 + }, + { + "epoch": 0.78, + "learning_rate": 4.350781912025268e-05, + "loss": 0.0001, + "step": 3371 + }, + { + "epoch": 0.78, + "learning_rate": 4.350589322856483e-05, + "loss": 0.001, + "step": 3372 + }, + { + "epoch": 0.78, + "learning_rate": 4.3503967336876975e-05, + "loss": 0.0001, + "step": 3373 + }, + { + "epoch": 0.78, + "learning_rate": 4.350204144518912e-05, + "loss": 0.0002, + "step": 3374 + }, + { + "epoch": 0.78, + "learning_rate": 4.350011555350128e-05, + "loss": 0.0, + "step": 3375 + }, + { + "epoch": 0.78, + "learning_rate": 4.349818966181342e-05, + "loss": 0.0, + "step": 3376 + }, + { + "epoch": 0.78, + "learning_rate": 4.349626377012557e-05, + "loss": 0.0014, + "step": 3377 + }, + { + "epoch": 0.78, + "learning_rate": 4.349433787843772e-05, + "loss": 0.0001, + "step": 3378 + }, + { + "epoch": 0.78, + "learning_rate": 4.349241198674987e-05, + "loss": 0.0085, + "step": 3379 + }, + { + "epoch": 0.78, + "learning_rate": 4.3490486095062014e-05, + "loss": 0.0009, + "step": 3380 + }, + { + "epoch": 0.78, + "learning_rate": 4.348856020337416e-05, + "loss": 0.0024, + "step": 3381 + }, + { + "epoch": 0.78, + "learning_rate": 4.3486634311686316e-05, + "loss": 0.0007, + "step": 3382 + }, + { + "epoch": 0.78, + "learning_rate": 4.348470841999846e-05, + "loss": 0.0013, + "step": 3383 + }, + { + "epoch": 0.78, + "learning_rate": 4.348278252831061e-05, + "loss": 0.0003, + "step": 3384 + }, + { + "epoch": 0.78, + "learning_rate": 4.348085663662276e-05, + "loss": 0.0125, + "step": 3385 + }, + { + "epoch": 0.78, + "learning_rate": 4.3478930744934906e-05, + "loss": 0.0, + "step": 3386 + }, + { + "epoch": 0.78, + "learning_rate": 4.3477004853247054e-05, + "loss": 0.0, + "step": 3387 + }, + { + "epoch": 0.78, + "learning_rate": 4.34750789615592e-05, + "loss": 0.0001, + "step": 3388 + }, + { + "epoch": 0.78, + "learning_rate": 4.3473153069871355e-05, + "loss": 0.0, + "step": 3389 + }, + { + "epoch": 0.78, + "learning_rate": 4.3471227178183496e-05, + "loss": 0.0005, + "step": 3390 + }, + { + "epoch": 0.78, + "learning_rate": 4.346930128649565e-05, + "loss": 0.0005, + "step": 3391 + }, + { + "epoch": 0.78, + "learning_rate": 4.34673753948078e-05, + "loss": 0.0012, + "step": 3392 + }, + { + "epoch": 0.78, + "learning_rate": 4.3465449503119945e-05, + "loss": 0.0002, + "step": 3393 + }, + { + "epoch": 0.78, + "learning_rate": 4.346352361143209e-05, + "loss": 0.0001, + "step": 3394 + }, + { + "epoch": 0.78, + "learning_rate": 4.346159771974425e-05, + "loss": 0.0033, + "step": 3395 + }, + { + "epoch": 0.78, + "learning_rate": 4.3459671828056394e-05, + "loss": 0.0004, + "step": 3396 + }, + { + "epoch": 0.79, + "learning_rate": 4.3457745936368535e-05, + "loss": 0.0, + "step": 3397 + }, + { + "epoch": 0.79, + "learning_rate": 4.345582004468069e-05, + "loss": 0.0, + "step": 3398 + }, + { + "epoch": 0.79, + "learning_rate": 4.345389415299284e-05, + "loss": 0.0, + "step": 3399 + }, + { + "epoch": 0.79, + "learning_rate": 4.3451968261304984e-05, + "loss": 0.0, + "step": 3400 + }, + { + "epoch": 0.79, + "learning_rate": 4.345004236961713e-05, + "loss": 0.0001, + "step": 3401 + }, + { + "epoch": 0.79, + "learning_rate": 4.3448116477929286e-05, + "loss": 0.0, + "step": 3402 + }, + { + "epoch": 0.79, + "learning_rate": 4.3446190586241434e-05, + "loss": 0.0003, + "step": 3403 + }, + { + "epoch": 0.79, + "learning_rate": 4.344426469455358e-05, + "loss": 0.0, + "step": 3404 + }, + { + "epoch": 0.79, + "learning_rate": 4.344233880286573e-05, + "loss": 0.0026, + "step": 3405 + }, + { + "epoch": 0.79, + "learning_rate": 4.344041291117788e-05, + "loss": 0.0001, + "step": 3406 + }, + { + "epoch": 0.79, + "learning_rate": 4.3438487019490023e-05, + "loss": 0.0009, + "step": 3407 + }, + { + "epoch": 0.79, + "learning_rate": 4.343656112780217e-05, + "loss": 0.0, + "step": 3408 + }, + { + "epoch": 0.79, + "learning_rate": 4.3434635236114325e-05, + "loss": 0.0044, + "step": 3409 + }, + { + "epoch": 0.79, + "learning_rate": 4.343270934442647e-05, + "loss": 0.0004, + "step": 3410 + }, + { + "epoch": 0.79, + "learning_rate": 4.343078345273862e-05, + "loss": 0.0, + "step": 3411 + }, + { + "epoch": 0.79, + "learning_rate": 4.342885756105077e-05, + "loss": 0.0, + "step": 3412 + }, + { + "epoch": 0.79, + "learning_rate": 4.342693166936292e-05, + "loss": 0.0, + "step": 3413 + }, + { + "epoch": 0.79, + "learning_rate": 4.342500577767506e-05, + "loss": 0.0002, + "step": 3414 + }, + { + "epoch": 0.79, + "learning_rate": 4.342307988598722e-05, + "loss": 0.0003, + "step": 3415 + }, + { + "epoch": 0.79, + "learning_rate": 4.3421153994299364e-05, + "loss": 0.002, + "step": 3416 + }, + { + "epoch": 0.79, + "learning_rate": 4.341922810261151e-05, + "loss": 0.0001, + "step": 3417 + }, + { + "epoch": 0.79, + "learning_rate": 4.341730221092366e-05, + "loss": 0.0003, + "step": 3418 + }, + { + "epoch": 0.79, + "learning_rate": 4.341537631923581e-05, + "loss": 0.0, + "step": 3419 + }, + { + "epoch": 0.79, + "learning_rate": 4.341345042754796e-05, + "loss": 0.0012, + "step": 3420 + }, + { + "epoch": 0.79, + "learning_rate": 4.34115245358601e-05, + "loss": 0.0001, + "step": 3421 + }, + { + "epoch": 0.79, + "learning_rate": 4.3409598644172256e-05, + "loss": 0.0, + "step": 3422 + }, + { + "epoch": 0.79, + "learning_rate": 4.3407672752484403e-05, + "loss": 0.0003, + "step": 3423 + }, + { + "epoch": 0.79, + "learning_rate": 4.340574686079655e-05, + "loss": 0.0, + "step": 3424 + }, + { + "epoch": 0.79, + "learning_rate": 4.34038209691087e-05, + "loss": 0.0014, + "step": 3425 + }, + { + "epoch": 0.79, + "learning_rate": 4.3401895077420846e-05, + "loss": 0.0, + "step": 3426 + }, + { + "epoch": 0.79, + "learning_rate": 4.3399969185733e-05, + "loss": 0.0012, + "step": 3427 + }, + { + "epoch": 0.79, + "learning_rate": 4.339804329404514e-05, + "loss": 0.0012, + "step": 3428 + }, + { + "epoch": 0.79, + "learning_rate": 4.3396117402357295e-05, + "loss": 0.0003, + "step": 3429 + }, + { + "epoch": 0.79, + "learning_rate": 4.339419151066944e-05, + "loss": 0.0037, + "step": 3430 + }, + { + "epoch": 0.79, + "learning_rate": 4.339226561898159e-05, + "loss": 0.0007, + "step": 3431 + }, + { + "epoch": 0.79, + "learning_rate": 4.339033972729374e-05, + "loss": 0.0003, + "step": 3432 + }, + { + "epoch": 0.79, + "learning_rate": 4.338841383560589e-05, + "loss": 0.0026, + "step": 3433 + }, + { + "epoch": 0.79, + "learning_rate": 4.338648794391804e-05, + "loss": 0.0002, + "step": 3434 + }, + { + "epoch": 0.79, + "learning_rate": 4.338456205223018e-05, + "loss": 0.0, + "step": 3435 + }, + { + "epoch": 0.79, + "learning_rate": 4.3382636160542334e-05, + "loss": 0.0001, + "step": 3436 + }, + { + "epoch": 0.79, + "learning_rate": 4.338071026885448e-05, + "loss": 0.0, + "step": 3437 + }, + { + "epoch": 0.79, + "learning_rate": 4.337878437716663e-05, + "loss": 0.0044, + "step": 3438 + }, + { + "epoch": 0.79, + "learning_rate": 4.337685848547878e-05, + "loss": 0.0002, + "step": 3439 + }, + { + "epoch": 0.8, + "learning_rate": 4.337493259379093e-05, + "loss": 0.0002, + "step": 3440 + }, + { + "epoch": 0.8, + "learning_rate": 4.337300670210308e-05, + "loss": 0.0, + "step": 3441 + }, + { + "epoch": 0.8, + "learning_rate": 4.3371080810415226e-05, + "loss": 0.0005, + "step": 3442 + }, + { + "epoch": 0.8, + "learning_rate": 4.336915491872737e-05, + "loss": 0.0, + "step": 3443 + }, + { + "epoch": 0.8, + "learning_rate": 4.336722902703952e-05, + "loss": 0.0001, + "step": 3444 + }, + { + "epoch": 0.8, + "learning_rate": 4.336530313535167e-05, + "loss": 0.001, + "step": 3445 + }, + { + "epoch": 0.8, + "learning_rate": 4.3363377243663816e-05, + "loss": 0.0002, + "step": 3446 + }, + { + "epoch": 0.8, + "learning_rate": 4.336145135197597e-05, + "loss": 0.0009, + "step": 3447 + }, + { + "epoch": 0.8, + "learning_rate": 4.335952546028811e-05, + "loss": 0.0004, + "step": 3448 + }, + { + "epoch": 0.8, + "learning_rate": 4.3357599568600265e-05, + "loss": 0.0005, + "step": 3449 + }, + { + "epoch": 0.8, + "learning_rate": 4.335567367691241e-05, + "loss": 0.0002, + "step": 3450 + }, + { + "epoch": 0.8, + "learning_rate": 4.335374778522456e-05, + "loss": 0.0, + "step": 3451 + }, + { + "epoch": 0.8, + "learning_rate": 4.335182189353671e-05, + "loss": 0.0, + "step": 3452 + }, + { + "epoch": 0.8, + "learning_rate": 4.3349896001848855e-05, + "loss": 0.0, + "step": 3453 + }, + { + "epoch": 0.8, + "learning_rate": 4.334797011016101e-05, + "loss": 0.0019, + "step": 3454 + }, + { + "epoch": 0.8, + "learning_rate": 4.334604421847315e-05, + "loss": 0.0002, + "step": 3455 + }, + { + "epoch": 0.8, + "learning_rate": 4.3344118326785304e-05, + "loss": 0.0003, + "step": 3456 + }, + { + "epoch": 0.8, + "learning_rate": 4.334219243509745e-05, + "loss": 0.0016, + "step": 3457 + }, + { + "epoch": 0.8, + "learning_rate": 4.33402665434096e-05, + "loss": 0.0014, + "step": 3458 + }, + { + "epoch": 0.8, + "learning_rate": 4.3338340651721747e-05, + "loss": 0.0043, + "step": 3459 + }, + { + "epoch": 0.8, + "learning_rate": 4.33364147600339e-05, + "loss": 0.0026, + "step": 3460 + }, + { + "epoch": 0.8, + "learning_rate": 4.333448886834605e-05, + "loss": 0.0038, + "step": 3461 + }, + { + "epoch": 0.8, + "learning_rate": 4.3332562976658196e-05, + "loss": 0.006, + "step": 3462 + }, + { + "epoch": 0.8, + "learning_rate": 4.333063708497034e-05, + "loss": 0.0009, + "step": 3463 + }, + { + "epoch": 0.8, + "learning_rate": 4.332871119328249e-05, + "loss": 0.0007, + "step": 3464 + }, + { + "epoch": 0.8, + "learning_rate": 4.332678530159464e-05, + "loss": 0.0102, + "step": 3465 + }, + { + "epoch": 0.8, + "learning_rate": 4.3324859409906786e-05, + "loss": 0.0012, + "step": 3466 + }, + { + "epoch": 0.8, + "learning_rate": 4.332293351821894e-05, + "loss": 0.0003, + "step": 3467 + }, + { + "epoch": 0.8, + "learning_rate": 4.332100762653109e-05, + "loss": 0.0001, + "step": 3468 + }, + { + "epoch": 0.8, + "learning_rate": 4.3319081734843235e-05, + "loss": 0.0, + "step": 3469 + }, + { + "epoch": 0.8, + "learning_rate": 4.331715584315538e-05, + "loss": 0.0307, + "step": 3470 + }, + { + "epoch": 0.8, + "learning_rate": 4.3315229951467537e-05, + "loss": 0.0023, + "step": 3471 + }, + { + "epoch": 0.8, + "learning_rate": 4.331330405977968e-05, + "loss": 0.0005, + "step": 3472 + }, + { + "epoch": 0.8, + "learning_rate": 4.3311378168091825e-05, + "loss": 0.0, + "step": 3473 + }, + { + "epoch": 0.8, + "learning_rate": 4.330945227640398e-05, + "loss": 0.0001, + "step": 3474 + }, + { + "epoch": 0.8, + "learning_rate": 4.3307526384716127e-05, + "loss": 0.0003, + "step": 3475 + }, + { + "epoch": 0.8, + "learning_rate": 4.3305600493028274e-05, + "loss": 0.0015, + "step": 3476 + }, + { + "epoch": 0.8, + "learning_rate": 4.330367460134042e-05, + "loss": 0.0007, + "step": 3477 + }, + { + "epoch": 0.8, + "learning_rate": 4.3301748709652576e-05, + "loss": 0.0001, + "step": 3478 + }, + { + "epoch": 0.8, + "learning_rate": 4.3299822817964716e-05, + "loss": 0.0004, + "step": 3479 + }, + { + "epoch": 0.8, + "learning_rate": 4.329789692627687e-05, + "loss": 0.0106, + "step": 3480 + }, + { + "epoch": 0.8, + "learning_rate": 4.329597103458902e-05, + "loss": 0.0012, + "step": 3481 + }, + { + "epoch": 0.8, + "learning_rate": 4.3294045142901166e-05, + "loss": 0.0332, + "step": 3482 + }, + { + "epoch": 0.8, + "learning_rate": 4.329211925121331e-05, + "loss": 0.0, + "step": 3483 + }, + { + "epoch": 0.81, + "learning_rate": 4.329019335952546e-05, + "loss": 0.0001, + "step": 3484 + }, + { + "epoch": 0.81, + "learning_rate": 4.3288267467837615e-05, + "loss": 0.0036, + "step": 3485 + }, + { + "epoch": 0.81, + "learning_rate": 4.3286341576149756e-05, + "loss": 0.0003, + "step": 3486 + }, + { + "epoch": 0.81, + "learning_rate": 4.328441568446191e-05, + "loss": 0.0, + "step": 3487 + }, + { + "epoch": 0.81, + "learning_rate": 4.328248979277406e-05, + "loss": 0.0, + "step": 3488 + }, + { + "epoch": 0.81, + "learning_rate": 4.3280563901086205e-05, + "loss": 0.0, + "step": 3489 + }, + { + "epoch": 0.81, + "learning_rate": 4.327863800939835e-05, + "loss": 0.0, + "step": 3490 + }, + { + "epoch": 0.81, + "learning_rate": 4.32767121177105e-05, + "loss": 0.0004, + "step": 3491 + }, + { + "epoch": 0.81, + "learning_rate": 4.3274786226022654e-05, + "loss": 0.0002, + "step": 3492 + }, + { + "epoch": 0.81, + "learning_rate": 4.3272860334334795e-05, + "loss": 0.0, + "step": 3493 + }, + { + "epoch": 0.81, + "learning_rate": 4.327093444264695e-05, + "loss": 0.0, + "step": 3494 + }, + { + "epoch": 0.81, + "learning_rate": 4.3269008550959096e-05, + "loss": 0.0025, + "step": 3495 + }, + { + "epoch": 0.81, + "learning_rate": 4.3267082659271244e-05, + "loss": 0.0036, + "step": 3496 + }, + { + "epoch": 0.81, + "learning_rate": 4.326515676758339e-05, + "loss": 0.0061, + "step": 3497 + }, + { + "epoch": 0.81, + "learning_rate": 4.3263230875895546e-05, + "loss": 0.0003, + "step": 3498 + }, + { + "epoch": 0.81, + "learning_rate": 4.326130498420769e-05, + "loss": 0.0, + "step": 3499 + }, + { + "epoch": 0.81, + "learning_rate": 4.3259379092519834e-05, + "loss": 0.0006, + "step": 3500 + }, + { + "epoch": 0.81, + "learning_rate": 4.325745320083199e-05, + "loss": 0.0012, + "step": 3501 + }, + { + "epoch": 0.81, + "learning_rate": 4.3255527309144136e-05, + "loss": 0.0001, + "step": 3502 + }, + { + "epoch": 0.81, + "learning_rate": 4.325360141745628e-05, + "loss": 0.0733, + "step": 3503 + }, + { + "epoch": 0.81, + "learning_rate": 4.325167552576843e-05, + "loss": 0.0005, + "step": 3504 + }, + { + "epoch": 0.81, + "learning_rate": 4.3249749634080585e-05, + "loss": 0.0321, + "step": 3505 + }, + { + "epoch": 0.81, + "learning_rate": 4.324782374239273e-05, + "loss": 0.0003, + "step": 3506 + }, + { + "epoch": 0.81, + "learning_rate": 4.324589785070488e-05, + "loss": 0.0001, + "step": 3507 + }, + { + "epoch": 0.81, + "learning_rate": 4.324397195901703e-05, + "loss": 0.0066, + "step": 3508 + }, + { + "epoch": 0.81, + "learning_rate": 4.3242046067329175e-05, + "loss": 0.0, + "step": 3509 + }, + { + "epoch": 0.81, + "learning_rate": 4.324012017564132e-05, + "loss": 0.0001, + "step": 3510 + }, + { + "epoch": 0.81, + "learning_rate": 4.323819428395347e-05, + "loss": 0.0065, + "step": 3511 + }, + { + "epoch": 0.81, + "learning_rate": 4.3236268392265624e-05, + "loss": 0.0005, + "step": 3512 + }, + { + "epoch": 0.81, + "learning_rate": 4.323434250057777e-05, + "loss": 0.0001, + "step": 3513 + }, + { + "epoch": 0.81, + "learning_rate": 4.323241660888992e-05, + "loss": 0.0014, + "step": 3514 + }, + { + "epoch": 0.81, + "learning_rate": 4.3230490717202066e-05, + "loss": 0.0, + "step": 3515 + }, + { + "epoch": 0.81, + "learning_rate": 4.322856482551422e-05, + "loss": 0.0, + "step": 3516 + }, + { + "epoch": 0.81, + "learning_rate": 4.322663893382636e-05, + "loss": 0.0, + "step": 3517 + }, + { + "epoch": 0.81, + "learning_rate": 4.3224713042138515e-05, + "loss": 0.0003, + "step": 3518 + }, + { + "epoch": 0.81, + "learning_rate": 4.322278715045066e-05, + "loss": 0.0, + "step": 3519 + }, + { + "epoch": 0.81, + "learning_rate": 4.3220861258762804e-05, + "loss": 0.0059, + "step": 3520 + }, + { + "epoch": 0.81, + "learning_rate": 4.321893536707496e-05, + "loss": 0.0, + "step": 3521 + }, + { + "epoch": 0.81, + "learning_rate": 4.3217009475387105e-05, + "loss": 0.0008, + "step": 3522 + }, + { + "epoch": 0.81, + "learning_rate": 4.321508358369925e-05, + "loss": 0.0, + "step": 3523 + }, + { + "epoch": 0.81, + "learning_rate": 4.32131576920114e-05, + "loss": 0.001, + "step": 3524 + }, + { + "epoch": 0.81, + "learning_rate": 4.3211231800323555e-05, + "loss": 0.0004, + "step": 3525 + }, + { + "epoch": 0.81, + "learning_rate": 4.32093059086357e-05, + "loss": 0.0, + "step": 3526 + }, + { + "epoch": 0.82, + "learning_rate": 4.320738001694785e-05, + "loss": 0.0002, + "step": 3527 + }, + { + "epoch": 0.82, + "learning_rate": 4.320545412526e-05, + "loss": 0.0134, + "step": 3528 + }, + { + "epoch": 0.82, + "learning_rate": 4.3203528233572145e-05, + "loss": 0.0, + "step": 3529 + }, + { + "epoch": 0.82, + "learning_rate": 4.320160234188429e-05, + "loss": 0.0001, + "step": 3530 + }, + { + "epoch": 0.82, + "learning_rate": 4.319967645019644e-05, + "loss": 0.0, + "step": 3531 + }, + { + "epoch": 0.82, + "learning_rate": 4.3197750558508594e-05, + "loss": 0.0013, + "step": 3532 + }, + { + "epoch": 0.82, + "learning_rate": 4.319582466682074e-05, + "loss": 0.0227, + "step": 3533 + }, + { + "epoch": 0.82, + "learning_rate": 4.319389877513289e-05, + "loss": 0.0002, + "step": 3534 + }, + { + "epoch": 0.82, + "learning_rate": 4.3191972883445036e-05, + "loss": 0.0, + "step": 3535 + }, + { + "epoch": 0.82, + "learning_rate": 4.319004699175719e-05, + "loss": 0.0004, + "step": 3536 + }, + { + "epoch": 0.82, + "learning_rate": 4.318812110006933e-05, + "loss": 0.0012, + "step": 3537 + }, + { + "epoch": 0.82, + "learning_rate": 4.318619520838148e-05, + "loss": 0.0002, + "step": 3538 + }, + { + "epoch": 0.82, + "learning_rate": 4.318426931669363e-05, + "loss": 0.0, + "step": 3539 + }, + { + "epoch": 0.82, + "learning_rate": 4.318234342500578e-05, + "loss": 0.0, + "step": 3540 + }, + { + "epoch": 0.82, + "learning_rate": 4.318041753331793e-05, + "loss": 0.0006, + "step": 3541 + }, + { + "epoch": 0.82, + "learning_rate": 4.3178491641630075e-05, + "loss": 0.0001, + "step": 3542 + }, + { + "epoch": 0.82, + "learning_rate": 4.317656574994223e-05, + "loss": 0.0001, + "step": 3543 + }, + { + "epoch": 0.82, + "learning_rate": 4.317463985825437e-05, + "loss": 0.0002, + "step": 3544 + }, + { + "epoch": 0.82, + "learning_rate": 4.3172713966566525e-05, + "loss": 0.03, + "step": 3545 + }, + { + "epoch": 0.82, + "learning_rate": 4.317078807487867e-05, + "loss": 0.0177, + "step": 3546 + }, + { + "epoch": 0.82, + "learning_rate": 4.316886218319082e-05, + "loss": 0.0003, + "step": 3547 + }, + { + "epoch": 0.82, + "learning_rate": 4.316693629150297e-05, + "loss": 0.0058, + "step": 3548 + }, + { + "epoch": 0.82, + "learning_rate": 4.3165010399815114e-05, + "loss": 0.001, + "step": 3549 + }, + { + "epoch": 0.82, + "learning_rate": 4.316308450812727e-05, + "loss": 0.0, + "step": 3550 + }, + { + "epoch": 0.82, + "learning_rate": 4.316115861643941e-05, + "loss": 0.0001, + "step": 3551 + }, + { + "epoch": 0.82, + "learning_rate": 4.3159232724751564e-05, + "loss": 0.0003, + "step": 3552 + }, + { + "epoch": 0.82, + "learning_rate": 4.315730683306371e-05, + "loss": 0.0015, + "step": 3553 + }, + { + "epoch": 0.82, + "learning_rate": 4.315538094137586e-05, + "loss": 0.0, + "step": 3554 + }, + { + "epoch": 0.82, + "learning_rate": 4.3153455049688006e-05, + "loss": 0.0072, + "step": 3555 + }, + { + "epoch": 0.82, + "learning_rate": 4.3151529158000154e-05, + "loss": 0.0, + "step": 3556 + }, + { + "epoch": 0.82, + "learning_rate": 4.314960326631231e-05, + "loss": 0.0006, + "step": 3557 + }, + { + "epoch": 0.82, + "learning_rate": 4.314767737462445e-05, + "loss": 0.0001, + "step": 3558 + }, + { + "epoch": 0.82, + "learning_rate": 4.31457514829366e-05, + "loss": 0.0002, + "step": 3559 + }, + { + "epoch": 0.82, + "learning_rate": 4.314382559124875e-05, + "loss": 0.0031, + "step": 3560 + }, + { + "epoch": 0.82, + "learning_rate": 4.31418996995609e-05, + "loss": 0.0014, + "step": 3561 + }, + { + "epoch": 0.82, + "learning_rate": 4.3139973807873045e-05, + "loss": 0.0003, + "step": 3562 + }, + { + "epoch": 0.82, + "learning_rate": 4.31380479161852e-05, + "loss": 0.0024, + "step": 3563 + }, + { + "epoch": 0.82, + "learning_rate": 4.313612202449735e-05, + "loss": 0.0007, + "step": 3564 + }, + { + "epoch": 0.82, + "learning_rate": 4.313419613280949e-05, + "loss": 0.0019, + "step": 3565 + }, + { + "epoch": 0.82, + "learning_rate": 4.313227024112164e-05, + "loss": 0.0006, + "step": 3566 + }, + { + "epoch": 0.82, + "learning_rate": 4.313034434943379e-05, + "loss": 0.0, + "step": 3567 + }, + { + "epoch": 0.82, + "learning_rate": 4.312841845774594e-05, + "loss": 0.0038, + "step": 3568 + }, + { + "epoch": 0.82, + "learning_rate": 4.3126492566058084e-05, + "loss": 0.0, + "step": 3569 + }, + { + "epoch": 0.83, + "learning_rate": 4.312456667437024e-05, + "loss": 0.0012, + "step": 3570 + }, + { + "epoch": 0.83, + "learning_rate": 4.3122640782682386e-05, + "loss": 0.0001, + "step": 3571 + }, + { + "epoch": 0.83, + "learning_rate": 4.3120714890994534e-05, + "loss": 0.0075, + "step": 3572 + }, + { + "epoch": 0.83, + "learning_rate": 4.311878899930668e-05, + "loss": 0.0, + "step": 3573 + }, + { + "epoch": 0.83, + "learning_rate": 4.3116863107618835e-05, + "loss": 0.0015, + "step": 3574 + }, + { + "epoch": 0.83, + "learning_rate": 4.3114937215930976e-05, + "loss": 0.0, + "step": 3575 + }, + { + "epoch": 0.83, + "learning_rate": 4.3113011324243123e-05, + "loss": 0.0, + "step": 3576 + }, + { + "epoch": 0.83, + "learning_rate": 4.311108543255528e-05, + "loss": 0.0002, + "step": 3577 + }, + { + "epoch": 0.83, + "learning_rate": 4.3109159540867425e-05, + "loss": 0.0005, + "step": 3578 + }, + { + "epoch": 0.83, + "learning_rate": 4.310723364917957e-05, + "loss": 0.0001, + "step": 3579 + }, + { + "epoch": 0.83, + "learning_rate": 4.310530775749172e-05, + "loss": 0.0001, + "step": 3580 + }, + { + "epoch": 0.83, + "learning_rate": 4.3103381865803874e-05, + "loss": 0.0008, + "step": 3581 + }, + { + "epoch": 0.83, + "learning_rate": 4.3101455974116015e-05, + "loss": 0.0004, + "step": 3582 + }, + { + "epoch": 0.83, + "learning_rate": 4.309953008242817e-05, + "loss": 0.0002, + "step": 3583 + }, + { + "epoch": 0.83, + "learning_rate": 4.309760419074032e-05, + "loss": 0.0006, + "step": 3584 + }, + { + "epoch": 0.83, + "learning_rate": 4.3095678299052464e-05, + "loss": 0.0, + "step": 3585 + }, + { + "epoch": 0.83, + "learning_rate": 4.309375240736461e-05, + "loss": 0.0016, + "step": 3586 + }, + { + "epoch": 0.83, + "learning_rate": 4.309182651567676e-05, + "loss": 0.0, + "step": 3587 + }, + { + "epoch": 0.83, + "learning_rate": 4.3089900623988913e-05, + "loss": 0.0, + "step": 3588 + }, + { + "epoch": 0.83, + "learning_rate": 4.3087974732301054e-05, + "loss": 0.0004, + "step": 3589 + }, + { + "epoch": 0.83, + "learning_rate": 4.308604884061321e-05, + "loss": 0.0048, + "step": 3590 + }, + { + "epoch": 0.83, + "learning_rate": 4.3084122948925356e-05, + "loss": 0.0291, + "step": 3591 + }, + { + "epoch": 0.83, + "learning_rate": 4.3082197057237503e-05, + "loss": 0.0001, + "step": 3592 + }, + { + "epoch": 0.83, + "learning_rate": 4.308027116554965e-05, + "loss": 0.0131, + "step": 3593 + }, + { + "epoch": 0.83, + "learning_rate": 4.30783452738618e-05, + "loss": 0.0, + "step": 3594 + }, + { + "epoch": 0.83, + "learning_rate": 4.307641938217395e-05, + "loss": 0.0, + "step": 3595 + }, + { + "epoch": 0.83, + "learning_rate": 4.307449349048609e-05, + "loss": 0.0011, + "step": 3596 + }, + { + "epoch": 0.83, + "learning_rate": 4.307256759879825e-05, + "loss": 0.004, + "step": 3597 + }, + { + "epoch": 0.83, + "learning_rate": 4.3070641707110395e-05, + "loss": 0.0017, + "step": 3598 + }, + { + "epoch": 0.83, + "learning_rate": 4.306871581542254e-05, + "loss": 0.0007, + "step": 3599 + }, + { + "epoch": 0.83, + "learning_rate": 4.306678992373469e-05, + "loss": 0.0025, + "step": 3600 + }, + { + "epoch": 0.83, + "learning_rate": 4.3064864032046844e-05, + "loss": 0.0001, + "step": 3601 + }, + { + "epoch": 0.83, + "learning_rate": 4.3062938140358985e-05, + "loss": 0.0011, + "step": 3602 + }, + { + "epoch": 0.83, + "learning_rate": 4.306101224867113e-05, + "loss": 0.0, + "step": 3603 + }, + { + "epoch": 0.83, + "learning_rate": 4.305908635698329e-05, + "loss": 0.0, + "step": 3604 + }, + { + "epoch": 0.83, + "learning_rate": 4.3057160465295434e-05, + "loss": 0.0001, + "step": 3605 + }, + { + "epoch": 0.83, + "learning_rate": 4.305523457360758e-05, + "loss": 0.0, + "step": 3606 + }, + { + "epoch": 0.83, + "learning_rate": 4.305330868191973e-05, + "loss": 0.0012, + "step": 3607 + }, + { + "epoch": 0.83, + "learning_rate": 4.305138279023188e-05, + "loss": 0.0019, + "step": 3608 + }, + { + "epoch": 0.83, + "learning_rate": 4.3049456898544024e-05, + "loss": 0.0006, + "step": 3609 + }, + { + "epoch": 0.83, + "learning_rate": 4.304753100685618e-05, + "loss": 0.0002, + "step": 3610 + }, + { + "epoch": 0.83, + "learning_rate": 4.3045605115168326e-05, + "loss": 0.0014, + "step": 3611 + }, + { + "epoch": 0.83, + "learning_rate": 4.304367922348047e-05, + "loss": 0.0001, + "step": 3612 + }, + { + "epoch": 0.83, + "learning_rate": 4.304175333179262e-05, + "loss": 0.0, + "step": 3613 + }, + { + "epoch": 0.84, + "learning_rate": 4.303982744010477e-05, + "loss": 0.0035, + "step": 3614 + }, + { + "epoch": 0.84, + "learning_rate": 4.303790154841692e-05, + "loss": 0.0005, + "step": 3615 + }, + { + "epoch": 0.84, + "learning_rate": 4.303597565672906e-05, + "loss": 0.0033, + "step": 3616 + }, + { + "epoch": 0.84, + "learning_rate": 4.303404976504122e-05, + "loss": 0.0009, + "step": 3617 + }, + { + "epoch": 0.84, + "learning_rate": 4.3032123873353365e-05, + "loss": 0.0515, + "step": 3618 + }, + { + "epoch": 0.84, + "learning_rate": 4.303019798166551e-05, + "loss": 0.0, + "step": 3619 + }, + { + "epoch": 0.84, + "learning_rate": 4.302827208997766e-05, + "loss": 0.0001, + "step": 3620 + }, + { + "epoch": 0.84, + "learning_rate": 4.302634619828981e-05, + "loss": 0.0001, + "step": 3621 + }, + { + "epoch": 0.84, + "learning_rate": 4.302442030660196e-05, + "loss": 0.0, + "step": 3622 + }, + { + "epoch": 0.84, + "learning_rate": 4.30224944149141e-05, + "loss": 0.0003, + "step": 3623 + }, + { + "epoch": 0.84, + "learning_rate": 4.3020568523226257e-05, + "loss": 0.0007, + "step": 3624 + }, + { + "epoch": 0.84, + "learning_rate": 4.3018642631538404e-05, + "loss": 0.0176, + "step": 3625 + }, + { + "epoch": 0.84, + "learning_rate": 4.301671673985055e-05, + "loss": 0.0002, + "step": 3626 + }, + { + "epoch": 0.84, + "learning_rate": 4.30147908481627e-05, + "loss": 0.0006, + "step": 3627 + }, + { + "epoch": 0.84, + "learning_rate": 4.301286495647485e-05, + "loss": 0.0011, + "step": 3628 + }, + { + "epoch": 0.84, + "learning_rate": 4.3010939064787e-05, + "loss": 0.0011, + "step": 3629 + }, + { + "epoch": 0.84, + "learning_rate": 4.300901317309915e-05, + "loss": 0.0003, + "step": 3630 + }, + { + "epoch": 0.84, + "learning_rate": 4.3007087281411296e-05, + "loss": 0.0, + "step": 3631 + }, + { + "epoch": 0.84, + "learning_rate": 4.300516138972344e-05, + "loss": 0.0, + "step": 3632 + }, + { + "epoch": 0.84, + "learning_rate": 4.300323549803559e-05, + "loss": 0.0133, + "step": 3633 + }, + { + "epoch": 0.84, + "learning_rate": 4.300130960634774e-05, + "loss": 0.0007, + "step": 3634 + }, + { + "epoch": 0.84, + "learning_rate": 4.299938371465989e-05, + "loss": 0.0025, + "step": 3635 + }, + { + "epoch": 0.84, + "learning_rate": 4.299745782297204e-05, + "loss": 0.0004, + "step": 3636 + }, + { + "epoch": 0.84, + "learning_rate": 4.299553193128419e-05, + "loss": 0.0, + "step": 3637 + }, + { + "epoch": 0.84, + "learning_rate": 4.2993606039596335e-05, + "loss": 0.0006, + "step": 3638 + }, + { + "epoch": 0.84, + "learning_rate": 4.299168014790849e-05, + "loss": 0.0, + "step": 3639 + }, + { + "epoch": 0.84, + "learning_rate": 4.298975425622063e-05, + "loss": 0.0004, + "step": 3640 + }, + { + "epoch": 0.84, + "learning_rate": 4.298782836453278e-05, + "loss": 0.0001, + "step": 3641 + }, + { + "epoch": 0.84, + "learning_rate": 4.298590247284493e-05, + "loss": 0.0003, + "step": 3642 + }, + { + "epoch": 0.84, + "learning_rate": 4.298397658115708e-05, + "loss": 0.0, + "step": 3643 + }, + { + "epoch": 0.84, + "learning_rate": 4.2982050689469226e-05, + "loss": 0.0002, + "step": 3644 + }, + { + "epoch": 0.84, + "learning_rate": 4.2980124797781374e-05, + "loss": 0.0006, + "step": 3645 + }, + { + "epoch": 0.84, + "learning_rate": 4.297819890609353e-05, + "loss": 0.0, + "step": 3646 + }, + { + "epoch": 0.84, + "learning_rate": 4.297627301440567e-05, + "loss": 0.0005, + "step": 3647 + }, + { + "epoch": 0.84, + "learning_rate": 4.297434712271782e-05, + "loss": 0.0002, + "step": 3648 + }, + { + "epoch": 0.84, + "learning_rate": 4.297242123102997e-05, + "loss": 0.0002, + "step": 3649 + }, + { + "epoch": 0.84, + "learning_rate": 4.297049533934212e-05, + "loss": 0.0011, + "step": 3650 + }, + { + "epoch": 0.84, + "learning_rate": 4.2968569447654266e-05, + "loss": 0.0002, + "step": 3651 + }, + { + "epoch": 0.84, + "learning_rate": 4.296664355596641e-05, + "loss": 0.0, + "step": 3652 + }, + { + "epoch": 0.84, + "learning_rate": 4.296471766427857e-05, + "loss": 0.0, + "step": 3653 + }, + { + "epoch": 0.84, + "learning_rate": 4.296279177259071e-05, + "loss": 0.0, + "step": 3654 + }, + { + "epoch": 0.84, + "learning_rate": 4.296086588090286e-05, + "loss": 0.0019, + "step": 3655 + }, + { + "epoch": 0.84, + "learning_rate": 4.295893998921501e-05, + "loss": 0.0003, + "step": 3656 + }, + { + "epoch": 0.85, + "learning_rate": 4.295701409752716e-05, + "loss": 0.0071, + "step": 3657 + }, + { + "epoch": 0.85, + "learning_rate": 4.2955088205839305e-05, + "loss": 0.0117, + "step": 3658 + }, + { + "epoch": 0.85, + "learning_rate": 4.295316231415145e-05, + "loss": 0.0, + "step": 3659 + }, + { + "epoch": 0.85, + "learning_rate": 4.2951236422463606e-05, + "loss": 0.0002, + "step": 3660 + }, + { + "epoch": 0.85, + "learning_rate": 4.294931053077575e-05, + "loss": 0.0018, + "step": 3661 + }, + { + "epoch": 0.85, + "learning_rate": 4.29473846390879e-05, + "loss": 0.0048, + "step": 3662 + }, + { + "epoch": 0.85, + "learning_rate": 4.294545874740005e-05, + "loss": 0.0002, + "step": 3663 + }, + { + "epoch": 0.85, + "learning_rate": 4.2943532855712196e-05, + "loss": 0.0, + "step": 3664 + }, + { + "epoch": 0.85, + "learning_rate": 4.2941606964024344e-05, + "loss": 0.021, + "step": 3665 + }, + { + "epoch": 0.85, + "learning_rate": 4.29396810723365e-05, + "loss": 0.0004, + "step": 3666 + }, + { + "epoch": 0.85, + "learning_rate": 4.2937755180648646e-05, + "loss": 0.0252, + "step": 3667 + }, + { + "epoch": 0.85, + "learning_rate": 4.2935829288960786e-05, + "loss": 0.0094, + "step": 3668 + }, + { + "epoch": 0.85, + "learning_rate": 4.293390339727294e-05, + "loss": 0.0023, + "step": 3669 + }, + { + "epoch": 0.85, + "learning_rate": 4.293197750558509e-05, + "loss": 0.0, + "step": 3670 + }, + { + "epoch": 0.85, + "learning_rate": 4.2930051613897235e-05, + "loss": 0.0016, + "step": 3671 + }, + { + "epoch": 0.85, + "learning_rate": 4.292812572220938e-05, + "loss": 0.0027, + "step": 3672 + }, + { + "epoch": 0.85, + "learning_rate": 4.292619983052154e-05, + "loss": 0.0001, + "step": 3673 + }, + { + "epoch": 0.85, + "learning_rate": 4.292427393883368e-05, + "loss": 0.0, + "step": 3674 + }, + { + "epoch": 0.85, + "learning_rate": 4.292234804714583e-05, + "loss": 0.0034, + "step": 3675 + }, + { + "epoch": 0.85, + "learning_rate": 4.292042215545798e-05, + "loss": 0.0, + "step": 3676 + }, + { + "epoch": 0.85, + "learning_rate": 4.291849626377013e-05, + "loss": 0.0, + "step": 3677 + }, + { + "epoch": 0.85, + "learning_rate": 4.2916570372082275e-05, + "loss": 0.0001, + "step": 3678 + }, + { + "epoch": 0.85, + "learning_rate": 4.291464448039442e-05, + "loss": 0.0001, + "step": 3679 + }, + { + "epoch": 0.85, + "learning_rate": 4.2912718588706576e-05, + "loss": 0.0001, + "step": 3680 + }, + { + "epoch": 0.85, + "learning_rate": 4.291079269701872e-05, + "loss": 0.0008, + "step": 3681 + }, + { + "epoch": 0.85, + "learning_rate": 4.290886680533087e-05, + "loss": 0.0, + "step": 3682 + }, + { + "epoch": 0.85, + "learning_rate": 4.290694091364302e-05, + "loss": 0.0019, + "step": 3683 + }, + { + "epoch": 0.85, + "learning_rate": 4.2905015021955166e-05, + "loss": 0.0, + "step": 3684 + }, + { + "epoch": 0.85, + "learning_rate": 4.2903089130267314e-05, + "loss": 0.0, + "step": 3685 + }, + { + "epoch": 0.85, + "learning_rate": 4.290116323857947e-05, + "loss": 0.0012, + "step": 3686 + }, + { + "epoch": 0.85, + "learning_rate": 4.2899237346891615e-05, + "loss": 0.0001, + "step": 3687 + }, + { + "epoch": 0.85, + "learning_rate": 4.2897311455203756e-05, + "loss": 0.0009, + "step": 3688 + }, + { + "epoch": 0.85, + "learning_rate": 4.289538556351591e-05, + "loss": 0.0001, + "step": 3689 + }, + { + "epoch": 0.85, + "learning_rate": 4.289345967182806e-05, + "loss": 0.0, + "step": 3690 + }, + { + "epoch": 0.85, + "learning_rate": 4.2891533780140205e-05, + "loss": 0.0001, + "step": 3691 + }, + { + "epoch": 0.85, + "learning_rate": 4.288960788845235e-05, + "loss": 0.0001, + "step": 3692 + }, + { + "epoch": 0.85, + "learning_rate": 4.288768199676451e-05, + "loss": 0.0, + "step": 3693 + }, + { + "epoch": 0.85, + "learning_rate": 4.2885756105076655e-05, + "loss": 0.0115, + "step": 3694 + }, + { + "epoch": 0.85, + "learning_rate": 4.28838302133888e-05, + "loss": 0.0002, + "step": 3695 + }, + { + "epoch": 0.85, + "learning_rate": 4.288190432170095e-05, + "loss": 0.0012, + "step": 3696 + }, + { + "epoch": 0.85, + "learning_rate": 4.28799784300131e-05, + "loss": 0.001, + "step": 3697 + }, + { + "epoch": 0.85, + "learning_rate": 4.2878052538325244e-05, + "loss": 0.0, + "step": 3698 + }, + { + "epoch": 0.85, + "learning_rate": 4.287612664663739e-05, + "loss": 0.0, + "step": 3699 + }, + { + "epoch": 0.86, + "learning_rate": 4.2874200754949546e-05, + "loss": 0.0087, + "step": 3700 + }, + { + "epoch": 0.86, + "learning_rate": 4.2872274863261694e-05, + "loss": 0.001, + "step": 3701 + }, + { + "epoch": 0.86, + "learning_rate": 4.287034897157384e-05, + "loss": 0.0012, + "step": 3702 + }, + { + "epoch": 0.86, + "learning_rate": 4.286842307988599e-05, + "loss": 0.0, + "step": 3703 + }, + { + "epoch": 0.86, + "learning_rate": 4.286649718819814e-05, + "loss": 0.0002, + "step": 3704 + }, + { + "epoch": 0.86, + "learning_rate": 4.2864571296510284e-05, + "loss": 0.0006, + "step": 3705 + }, + { + "epoch": 0.86, + "learning_rate": 4.286264540482243e-05, + "loss": 0.0, + "step": 3706 + }, + { + "epoch": 0.86, + "learning_rate": 4.2860719513134585e-05, + "loss": 0.0, + "step": 3707 + }, + { + "epoch": 0.86, + "learning_rate": 4.285879362144673e-05, + "loss": 0.0, + "step": 3708 + }, + { + "epoch": 0.86, + "learning_rate": 4.285686772975888e-05, + "loss": 0.0001, + "step": 3709 + }, + { + "epoch": 0.86, + "learning_rate": 4.285494183807103e-05, + "loss": 0.0, + "step": 3710 + }, + { + "epoch": 0.86, + "learning_rate": 4.285301594638318e-05, + "loss": 0.0003, + "step": 3711 + }, + { + "epoch": 0.86, + "learning_rate": 4.285109005469532e-05, + "loss": 0.002, + "step": 3712 + }, + { + "epoch": 0.86, + "learning_rate": 4.284916416300748e-05, + "loss": 0.0002, + "step": 3713 + }, + { + "epoch": 0.86, + "learning_rate": 4.2847238271319624e-05, + "loss": 0.0008, + "step": 3714 + }, + { + "epoch": 0.86, + "learning_rate": 4.284531237963177e-05, + "loss": 0.0003, + "step": 3715 + }, + { + "epoch": 0.86, + "learning_rate": 4.284338648794392e-05, + "loss": 0.0, + "step": 3716 + }, + { + "epoch": 0.86, + "learning_rate": 4.284146059625607e-05, + "loss": 0.0, + "step": 3717 + }, + { + "epoch": 0.86, + "learning_rate": 4.283953470456822e-05, + "loss": 0.0002, + "step": 3718 + }, + { + "epoch": 0.86, + "learning_rate": 4.283760881288036e-05, + "loss": 0.0003, + "step": 3719 + }, + { + "epoch": 0.86, + "learning_rate": 4.2835682921192516e-05, + "loss": 0.0004, + "step": 3720 + }, + { + "epoch": 0.86, + "learning_rate": 4.2833757029504664e-05, + "loss": 0.0005, + "step": 3721 + }, + { + "epoch": 0.86, + "learning_rate": 4.283183113781681e-05, + "loss": 0.0, + "step": 3722 + }, + { + "epoch": 0.86, + "learning_rate": 4.282990524612896e-05, + "loss": 0.0003, + "step": 3723 + }, + { + "epoch": 0.86, + "learning_rate": 4.2827979354441106e-05, + "loss": 0.0003, + "step": 3724 + }, + { + "epoch": 0.86, + "learning_rate": 4.282605346275326e-05, + "loss": 0.0011, + "step": 3725 + }, + { + "epoch": 0.86, + "learning_rate": 4.28241275710654e-05, + "loss": 0.0, + "step": 3726 + }, + { + "epoch": 0.86, + "learning_rate": 4.2822201679377555e-05, + "loss": 0.0032, + "step": 3727 + }, + { + "epoch": 0.86, + "learning_rate": 4.28202757876897e-05, + "loss": 0.0, + "step": 3728 + }, + { + "epoch": 0.86, + "learning_rate": 4.281834989600185e-05, + "loss": 0.0008, + "step": 3729 + }, + { + "epoch": 0.86, + "learning_rate": 4.2816424004314e-05, + "loss": 0.0026, + "step": 3730 + }, + { + "epoch": 0.86, + "learning_rate": 4.281449811262615e-05, + "loss": 0.0001, + "step": 3731 + }, + { + "epoch": 0.86, + "learning_rate": 4.28125722209383e-05, + "loss": 0.0113, + "step": 3732 + }, + { + "epoch": 0.86, + "learning_rate": 4.281064632925044e-05, + "loss": 0.0001, + "step": 3733 + }, + { + "epoch": 0.86, + "learning_rate": 4.2808720437562594e-05, + "loss": 0.0, + "step": 3734 + }, + { + "epoch": 0.86, + "learning_rate": 4.280679454587474e-05, + "loss": 0.0023, + "step": 3735 + }, + { + "epoch": 0.86, + "learning_rate": 4.280486865418689e-05, + "loss": 0.0012, + "step": 3736 + }, + { + "epoch": 0.86, + "learning_rate": 4.280294276249904e-05, + "loss": 0.0001, + "step": 3737 + }, + { + "epoch": 0.86, + "learning_rate": 4.280101687081119e-05, + "loss": 0.0002, + "step": 3738 + }, + { + "epoch": 0.86, + "learning_rate": 4.279909097912334e-05, + "loss": 0.0006, + "step": 3739 + }, + { + "epoch": 0.86, + "learning_rate": 4.2797165087435486e-05, + "loss": 0.0013, + "step": 3740 + }, + { + "epoch": 0.86, + "learning_rate": 4.2795239195747633e-05, + "loss": 0.0003, + "step": 3741 + }, + { + "epoch": 0.86, + "learning_rate": 4.279331330405979e-05, + "loss": 0.0002, + "step": 3742 + }, + { + "epoch": 0.87, + "learning_rate": 4.279138741237193e-05, + "loss": 0.0004, + "step": 3743 + }, + { + "epoch": 0.87, + "learning_rate": 4.2789461520684076e-05, + "loss": 0.0, + "step": 3744 + }, + { + "epoch": 0.87, + "learning_rate": 4.278753562899623e-05, + "loss": 0.0006, + "step": 3745 + }, + { + "epoch": 0.87, + "learning_rate": 4.278560973730837e-05, + "loss": 0.0101, + "step": 3746 + }, + { + "epoch": 0.87, + "learning_rate": 4.2783683845620525e-05, + "loss": 0.0013, + "step": 3747 + }, + { + "epoch": 0.87, + "learning_rate": 4.278175795393267e-05, + "loss": 0.0011, + "step": 3748 + }, + { + "epoch": 0.87, + "learning_rate": 4.277983206224483e-05, + "loss": 0.0, + "step": 3749 + }, + { + "epoch": 0.87, + "learning_rate": 4.277790617055697e-05, + "loss": 0.0, + "step": 3750 + }, + { + "epoch": 0.87, + "learning_rate": 4.277598027886912e-05, + "loss": 0.0017, + "step": 3751 + }, + { + "epoch": 0.87, + "learning_rate": 4.277405438718127e-05, + "loss": 0.0002, + "step": 3752 + }, + { + "epoch": 0.87, + "learning_rate": 4.277212849549341e-05, + "loss": 0.0002, + "step": 3753 + }, + { + "epoch": 0.87, + "learning_rate": 4.2770202603805564e-05, + "loss": 0.0, + "step": 3754 + }, + { + "epoch": 0.87, + "learning_rate": 4.276827671211771e-05, + "loss": 0.0002, + "step": 3755 + }, + { + "epoch": 0.87, + "learning_rate": 4.276635082042986e-05, + "loss": 0.0003, + "step": 3756 + }, + { + "epoch": 0.87, + "learning_rate": 4.276442492874201e-05, + "loss": 0.0021, + "step": 3757 + }, + { + "epoch": 0.87, + "learning_rate": 4.276249903705416e-05, + "loss": 0.0001, + "step": 3758 + }, + { + "epoch": 0.87, + "learning_rate": 4.276057314536631e-05, + "loss": 0.0004, + "step": 3759 + }, + { + "epoch": 0.87, + "learning_rate": 4.2758647253678456e-05, + "loss": 0.001, + "step": 3760 + }, + { + "epoch": 0.87, + "learning_rate": 4.27567213619906e-05, + "loss": 0.0019, + "step": 3761 + }, + { + "epoch": 0.87, + "learning_rate": 4.275479547030275e-05, + "loss": 0.0, + "step": 3762 + }, + { + "epoch": 0.87, + "learning_rate": 4.27528695786149e-05, + "loss": 0.0, + "step": 3763 + }, + { + "epoch": 0.87, + "learning_rate": 4.2750943686927046e-05, + "loss": 0.0, + "step": 3764 + }, + { + "epoch": 0.87, + "learning_rate": 4.27490177952392e-05, + "loss": 0.0046, + "step": 3765 + }, + { + "epoch": 0.87, + "learning_rate": 4.274709190355135e-05, + "loss": 0.0001, + "step": 3766 + }, + { + "epoch": 0.87, + "learning_rate": 4.2745166011863495e-05, + "loss": 0.0107, + "step": 3767 + }, + { + "epoch": 0.87, + "learning_rate": 4.274324012017564e-05, + "loss": 0.0003, + "step": 3768 + }, + { + "epoch": 0.87, + "learning_rate": 4.27413142284878e-05, + "loss": 0.0008, + "step": 3769 + }, + { + "epoch": 0.87, + "learning_rate": 4.273938833679994e-05, + "loss": 0.0, + "step": 3770 + }, + { + "epoch": 0.87, + "learning_rate": 4.2737462445112085e-05, + "loss": 0.0005, + "step": 3771 + }, + { + "epoch": 0.87, + "learning_rate": 4.273553655342424e-05, + "loss": 0.0002, + "step": 3772 + }, + { + "epoch": 0.87, + "learning_rate": 4.2733610661736387e-05, + "loss": 0.0, + "step": 3773 + }, + { + "epoch": 0.87, + "learning_rate": 4.2731684770048534e-05, + "loss": 0.0001, + "step": 3774 + }, + { + "epoch": 0.87, + "learning_rate": 4.272975887836068e-05, + "loss": 0.0, + "step": 3775 + }, + { + "epoch": 0.87, + "learning_rate": 4.2727832986672836e-05, + "loss": 0.0001, + "step": 3776 + }, + { + "epoch": 0.87, + "learning_rate": 4.2725907094984977e-05, + "loss": 0.0004, + "step": 3777 + }, + { + "epoch": 0.87, + "learning_rate": 4.272398120329713e-05, + "loss": 0.0008, + "step": 3778 + }, + { + "epoch": 0.87, + "learning_rate": 4.272205531160928e-05, + "loss": 0.0001, + "step": 3779 + }, + { + "epoch": 0.87, + "learning_rate": 4.2720129419921426e-05, + "loss": 0.0001, + "step": 3780 + }, + { + "epoch": 0.87, + "learning_rate": 4.271820352823357e-05, + "loss": 0.013, + "step": 3781 + }, + { + "epoch": 0.87, + "learning_rate": 4.271627763654572e-05, + "loss": 0.0001, + "step": 3782 + }, + { + "epoch": 0.87, + "learning_rate": 4.2714351744857875e-05, + "loss": 0.0002, + "step": 3783 + }, + { + "epoch": 0.87, + "learning_rate": 4.2712425853170016e-05, + "loss": 0.0001, + "step": 3784 + }, + { + "epoch": 0.87, + "learning_rate": 4.271049996148217e-05, + "loss": 0.0003, + "step": 3785 + }, + { + "epoch": 0.87, + "learning_rate": 4.270857406979432e-05, + "loss": 0.0002, + "step": 3786 + }, + { + "epoch": 0.88, + "learning_rate": 4.2706648178106465e-05, + "loss": 0.0, + "step": 3787 + }, + { + "epoch": 0.88, + "learning_rate": 4.270472228641861e-05, + "loss": 0.0, + "step": 3788 + }, + { + "epoch": 0.88, + "learning_rate": 4.270279639473076e-05, + "loss": 0.0016, + "step": 3789 + }, + { + "epoch": 0.88, + "learning_rate": 4.2700870503042914e-05, + "loss": 0.0001, + "step": 3790 + }, + { + "epoch": 0.88, + "learning_rate": 4.2698944611355055e-05, + "loss": 0.0002, + "step": 3791 + }, + { + "epoch": 0.88, + "learning_rate": 4.269701871966721e-05, + "loss": 0.0, + "step": 3792 + }, + { + "epoch": 0.88, + "learning_rate": 4.2695092827979356e-05, + "loss": 0.0, + "step": 3793 + }, + { + "epoch": 0.88, + "learning_rate": 4.2693166936291504e-05, + "loss": 0.0, + "step": 3794 + }, + { + "epoch": 0.88, + "learning_rate": 4.269124104460365e-05, + "loss": 0.0002, + "step": 3795 + }, + { + "epoch": 0.88, + "learning_rate": 4.2689315152915806e-05, + "loss": 0.0005, + "step": 3796 + }, + { + "epoch": 0.88, + "learning_rate": 4.268738926122795e-05, + "loss": 0.0005, + "step": 3797 + }, + { + "epoch": 0.88, + "learning_rate": 4.26854633695401e-05, + "loss": 0.0009, + "step": 3798 + }, + { + "epoch": 0.88, + "learning_rate": 4.268353747785225e-05, + "loss": 0.0, + "step": 3799 + }, + { + "epoch": 0.88, + "learning_rate": 4.2681611586164396e-05, + "loss": 0.0017, + "step": 3800 + }, + { + "epoch": 0.88, + "learning_rate": 4.267968569447654e-05, + "loss": 0.0002, + "step": 3801 + }, + { + "epoch": 0.88, + "learning_rate": 4.267775980278869e-05, + "loss": 0.0, + "step": 3802 + }, + { + "epoch": 0.88, + "learning_rate": 4.2675833911100845e-05, + "loss": 0.0012, + "step": 3803 + }, + { + "epoch": 0.88, + "learning_rate": 4.267390801941299e-05, + "loss": 0.0013, + "step": 3804 + }, + { + "epoch": 0.88, + "learning_rate": 4.267198212772514e-05, + "loss": 0.0, + "step": 3805 + }, + { + "epoch": 0.88, + "learning_rate": 4.267005623603729e-05, + "loss": 0.0, + "step": 3806 + }, + { + "epoch": 0.88, + "learning_rate": 4.266813034434944e-05, + "loss": 0.0004, + "step": 3807 + }, + { + "epoch": 0.88, + "learning_rate": 4.266620445266158e-05, + "loss": 0.0004, + "step": 3808 + }, + { + "epoch": 0.88, + "learning_rate": 4.266427856097373e-05, + "loss": 0.0033, + "step": 3809 + }, + { + "epoch": 0.88, + "learning_rate": 4.2662352669285884e-05, + "loss": 0.0, + "step": 3810 + }, + { + "epoch": 0.88, + "learning_rate": 4.266042677759803e-05, + "loss": 0.0001, + "step": 3811 + }, + { + "epoch": 0.88, + "learning_rate": 4.265850088591018e-05, + "loss": 0.0001, + "step": 3812 + }, + { + "epoch": 0.88, + "learning_rate": 4.2656574994222326e-05, + "loss": 0.0003, + "step": 3813 + }, + { + "epoch": 0.88, + "learning_rate": 4.265464910253448e-05, + "loss": 0.0, + "step": 3814 + }, + { + "epoch": 0.88, + "learning_rate": 4.265272321084662e-05, + "loss": 0.0014, + "step": 3815 + }, + { + "epoch": 0.88, + "learning_rate": 4.2650797319158776e-05, + "loss": 0.0, + "step": 3816 + }, + { + "epoch": 0.88, + "learning_rate": 4.264887142747092e-05, + "loss": 0.0, + "step": 3817 + }, + { + "epoch": 0.88, + "learning_rate": 4.264694553578307e-05, + "loss": 0.0005, + "step": 3818 + }, + { + "epoch": 0.88, + "learning_rate": 4.264501964409522e-05, + "loss": 0.001, + "step": 3819 + }, + { + "epoch": 0.88, + "learning_rate": 4.2643093752407365e-05, + "loss": 0.0007, + "step": 3820 + }, + { + "epoch": 0.88, + "learning_rate": 4.264116786071952e-05, + "loss": 0.0002, + "step": 3821 + }, + { + "epoch": 0.88, + "learning_rate": 4.263924196903166e-05, + "loss": 0.0, + "step": 3822 + }, + { + "epoch": 0.88, + "learning_rate": 4.2637316077343815e-05, + "loss": 0.0003, + "step": 3823 + }, + { + "epoch": 0.88, + "learning_rate": 4.263539018565596e-05, + "loss": 0.0, + "step": 3824 + }, + { + "epoch": 0.88, + "learning_rate": 4.263346429396811e-05, + "loss": 0.0, + "step": 3825 + }, + { + "epoch": 0.88, + "learning_rate": 4.263153840228026e-05, + "loss": 0.0002, + "step": 3826 + }, + { + "epoch": 0.88, + "learning_rate": 4.2629612510592405e-05, + "loss": 0.0, + "step": 3827 + }, + { + "epoch": 0.88, + "learning_rate": 4.262768661890455e-05, + "loss": 0.0111, + "step": 3828 + }, + { + "epoch": 0.88, + "learning_rate": 4.26257607272167e-05, + "loss": 0.0, + "step": 3829 + }, + { + "epoch": 0.89, + "learning_rate": 4.2623834835528854e-05, + "loss": 0.0, + "step": 3830 + }, + { + "epoch": 0.89, + "learning_rate": 4.2621908943841e-05, + "loss": 0.0049, + "step": 3831 + }, + { + "epoch": 0.89, + "learning_rate": 4.261998305215315e-05, + "loss": 0.003, + "step": 3832 + }, + { + "epoch": 0.89, + "learning_rate": 4.2618057160465296e-05, + "loss": 0.0004, + "step": 3833 + }, + { + "epoch": 0.89, + "learning_rate": 4.261613126877745e-05, + "loss": 0.0001, + "step": 3834 + }, + { + "epoch": 0.89, + "learning_rate": 4.261420537708959e-05, + "loss": 0.0004, + "step": 3835 + }, + { + "epoch": 0.89, + "learning_rate": 4.261227948540174e-05, + "loss": 0.009, + "step": 3836 + }, + { + "epoch": 0.89, + "learning_rate": 4.261035359371389e-05, + "loss": 0.0001, + "step": 3837 + }, + { + "epoch": 0.89, + "learning_rate": 4.260842770202604e-05, + "loss": 0.0003, + "step": 3838 + }, + { + "epoch": 0.89, + "learning_rate": 4.260650181033819e-05, + "loss": 0.0016, + "step": 3839 + }, + { + "epoch": 0.89, + "learning_rate": 4.2604575918650335e-05, + "loss": 0.0, + "step": 3840 + }, + { + "epoch": 0.89, + "learning_rate": 4.260265002696249e-05, + "loss": 0.0004, + "step": 3841 + }, + { + "epoch": 0.89, + "learning_rate": 4.260072413527463e-05, + "loss": 0.0, + "step": 3842 + }, + { + "epoch": 0.89, + "learning_rate": 4.2598798243586785e-05, + "loss": 0.0009, + "step": 3843 + }, + { + "epoch": 0.89, + "learning_rate": 4.259687235189893e-05, + "loss": 0.0067, + "step": 3844 + }, + { + "epoch": 0.89, + "learning_rate": 4.259494646021108e-05, + "loss": 0.0, + "step": 3845 + }, + { + "epoch": 0.89, + "learning_rate": 4.259302056852323e-05, + "loss": 0.0001, + "step": 3846 + }, + { + "epoch": 0.89, + "learning_rate": 4.2591094676835374e-05, + "loss": 0.0001, + "step": 3847 + }, + { + "epoch": 0.89, + "learning_rate": 4.258916878514753e-05, + "loss": 0.0, + "step": 3848 + }, + { + "epoch": 0.89, + "learning_rate": 4.258724289345967e-05, + "loss": 0.0001, + "step": 3849 + }, + { + "epoch": 0.89, + "learning_rate": 4.2585317001771824e-05, + "loss": 0.0003, + "step": 3850 + }, + { + "epoch": 0.89, + "learning_rate": 4.258339111008397e-05, + "loss": 0.002, + "step": 3851 + }, + { + "epoch": 0.89, + "learning_rate": 4.258146521839612e-05, + "loss": 0.0002, + "step": 3852 + }, + { + "epoch": 0.89, + "learning_rate": 4.2579539326708266e-05, + "loss": 0.0, + "step": 3853 + }, + { + "epoch": 0.89, + "learning_rate": 4.257761343502042e-05, + "loss": 0.0, + "step": 3854 + }, + { + "epoch": 0.89, + "learning_rate": 4.257568754333257e-05, + "loss": 0.0, + "step": 3855 + }, + { + "epoch": 0.89, + "learning_rate": 4.257376165164471e-05, + "loss": 0.0001, + "step": 3856 + }, + { + "epoch": 0.89, + "learning_rate": 4.257183575995686e-05, + "loss": 0.0, + "step": 3857 + }, + { + "epoch": 0.89, + "learning_rate": 4.256990986826901e-05, + "loss": 0.0, + "step": 3858 + }, + { + "epoch": 0.89, + "learning_rate": 4.256798397658116e-05, + "loss": 0.0, + "step": 3859 + }, + { + "epoch": 0.89, + "learning_rate": 4.2566058084893305e-05, + "loss": 0.0, + "step": 3860 + }, + { + "epoch": 0.89, + "learning_rate": 4.256413219320546e-05, + "loss": 0.0017, + "step": 3861 + }, + { + "epoch": 0.89, + "learning_rate": 4.256220630151761e-05, + "loss": 0.0098, + "step": 3862 + }, + { + "epoch": 0.89, + "learning_rate": 4.2560280409829754e-05, + "loss": 0.0002, + "step": 3863 + }, + { + "epoch": 0.89, + "learning_rate": 4.25583545181419e-05, + "loss": 0.0, + "step": 3864 + }, + { + "epoch": 0.89, + "learning_rate": 4.255642862645405e-05, + "loss": 0.0001, + "step": 3865 + }, + { + "epoch": 0.89, + "learning_rate": 4.25545027347662e-05, + "loss": 0.0, + "step": 3866 + }, + { + "epoch": 0.89, + "learning_rate": 4.2552576843078344e-05, + "loss": 0.0007, + "step": 3867 + }, + { + "epoch": 0.89, + "learning_rate": 4.25506509513905e-05, + "loss": 0.0221, + "step": 3868 + }, + { + "epoch": 0.89, + "learning_rate": 4.2548725059702646e-05, + "loss": 0.0, + "step": 3869 + }, + { + "epoch": 0.89, + "learning_rate": 4.2546799168014794e-05, + "loss": 0.0005, + "step": 3870 + }, + { + "epoch": 0.89, + "learning_rate": 4.254487327632694e-05, + "loss": 0.0, + "step": 3871 + }, + { + "epoch": 0.89, + "learning_rate": 4.2542947384639095e-05, + "loss": 0.0028, + "step": 3872 + }, + { + "epoch": 0.9, + "learning_rate": 4.2541021492951236e-05, + "loss": 0.0002, + "step": 3873 + }, + { + "epoch": 0.9, + "learning_rate": 4.2539095601263384e-05, + "loss": 0.0001, + "step": 3874 + }, + { + "epoch": 0.9, + "learning_rate": 4.253716970957554e-05, + "loss": 0.0001, + "step": 3875 + }, + { + "epoch": 0.9, + "learning_rate": 4.2535243817887685e-05, + "loss": 0.0, + "step": 3876 + }, + { + "epoch": 0.9, + "learning_rate": 4.253331792619983e-05, + "loss": 0.0012, + "step": 3877 + }, + { + "epoch": 0.9, + "learning_rate": 4.253139203451198e-05, + "loss": 0.0, + "step": 3878 + }, + { + "epoch": 0.9, + "learning_rate": 4.2529466142824134e-05, + "loss": 0.0008, + "step": 3879 + }, + { + "epoch": 0.9, + "learning_rate": 4.2527540251136275e-05, + "loss": 0.0011, + "step": 3880 + }, + { + "epoch": 0.9, + "learning_rate": 4.252561435944843e-05, + "loss": 0.0002, + "step": 3881 + }, + { + "epoch": 0.9, + "learning_rate": 4.252368846776058e-05, + "loss": 0.0, + "step": 3882 + }, + { + "epoch": 0.9, + "learning_rate": 4.2521762576072724e-05, + "loss": 0.0004, + "step": 3883 + }, + { + "epoch": 0.9, + "learning_rate": 4.251983668438487e-05, + "loss": 0.0005, + "step": 3884 + }, + { + "epoch": 0.9, + "learning_rate": 4.251791079269702e-05, + "loss": 0.0001, + "step": 3885 + }, + { + "epoch": 0.9, + "learning_rate": 4.2515984901009174e-05, + "loss": 0.0, + "step": 3886 + }, + { + "epoch": 0.9, + "learning_rate": 4.2514059009321314e-05, + "loss": 0.0002, + "step": 3887 + }, + { + "epoch": 0.9, + "learning_rate": 4.251213311763347e-05, + "loss": 0.0015, + "step": 3888 + }, + { + "epoch": 0.9, + "learning_rate": 4.2510207225945616e-05, + "loss": 0.0, + "step": 3889 + }, + { + "epoch": 0.9, + "learning_rate": 4.2508281334257763e-05, + "loss": 0.0, + "step": 3890 + }, + { + "epoch": 0.9, + "learning_rate": 4.250635544256991e-05, + "loss": 0.0, + "step": 3891 + }, + { + "epoch": 0.9, + "learning_rate": 4.250442955088206e-05, + "loss": 0.0001, + "step": 3892 + }, + { + "epoch": 0.9, + "learning_rate": 4.250250365919421e-05, + "loss": 0.0001, + "step": 3893 + }, + { + "epoch": 0.9, + "learning_rate": 4.2500577767506353e-05, + "loss": 0.0, + "step": 3894 + }, + { + "epoch": 0.9, + "learning_rate": 4.249865187581851e-05, + "loss": 0.0, + "step": 3895 + }, + { + "epoch": 0.9, + "learning_rate": 4.2496725984130655e-05, + "loss": 0.0006, + "step": 3896 + }, + { + "epoch": 0.9, + "learning_rate": 4.24948000924428e-05, + "loss": 0.0, + "step": 3897 + }, + { + "epoch": 0.9, + "learning_rate": 4.249287420075495e-05, + "loss": 0.0123, + "step": 3898 + }, + { + "epoch": 0.9, + "learning_rate": 4.2490948309067104e-05, + "loss": 0.0, + "step": 3899 + }, + { + "epoch": 0.9, + "learning_rate": 4.2489022417379245e-05, + "loss": 0.0, + "step": 3900 + }, + { + "epoch": 0.9, + "learning_rate": 4.248709652569139e-05, + "loss": 0.0001, + "step": 3901 + }, + { + "epoch": 0.9, + "learning_rate": 4.248517063400355e-05, + "loss": 0.0001, + "step": 3902 + }, + { + "epoch": 0.9, + "learning_rate": 4.2483244742315694e-05, + "loss": 0.0007, + "step": 3903 + }, + { + "epoch": 0.9, + "learning_rate": 4.248131885062784e-05, + "loss": 0.0, + "step": 3904 + }, + { + "epoch": 0.9, + "learning_rate": 4.247939295893999e-05, + "loss": 0.0005, + "step": 3905 + }, + { + "epoch": 0.9, + "learning_rate": 4.2477467067252143e-05, + "loss": 0.0002, + "step": 3906 + }, + { + "epoch": 0.9, + "learning_rate": 4.2475541175564284e-05, + "loss": 0.0007, + "step": 3907 + }, + { + "epoch": 0.9, + "learning_rate": 4.247361528387644e-05, + "loss": 0.0, + "step": 3908 + }, + { + "epoch": 0.9, + "learning_rate": 4.2471689392188586e-05, + "loss": 0.0002, + "step": 3909 + }, + { + "epoch": 0.9, + "learning_rate": 4.246976350050073e-05, + "loss": 0.0011, + "step": 3910 + }, + { + "epoch": 0.9, + "learning_rate": 4.246783760881288e-05, + "loss": 0.0009, + "step": 3911 + }, + { + "epoch": 0.9, + "learning_rate": 4.246591171712503e-05, + "loss": 0.0005, + "step": 3912 + }, + { + "epoch": 0.9, + "learning_rate": 4.246398582543718e-05, + "loss": 0.0, + "step": 3913 + }, + { + "epoch": 0.9, + "learning_rate": 4.246205993374932e-05, + "loss": 0.0001, + "step": 3914 + }, + { + "epoch": 0.9, + "learning_rate": 4.246013404206148e-05, + "loss": 0.0005, + "step": 3915 + }, + { + "epoch": 0.91, + "learning_rate": 4.2458208150373625e-05, + "loss": 0.0, + "step": 3916 + }, + { + "epoch": 0.91, + "learning_rate": 4.245628225868577e-05, + "loss": 0.0203, + "step": 3917 + }, + { + "epoch": 0.91, + "learning_rate": 4.245435636699792e-05, + "loss": 0.0002, + "step": 3918 + }, + { + "epoch": 0.91, + "learning_rate": 4.2452430475310074e-05, + "loss": 0.0002, + "step": 3919 + }, + { + "epoch": 0.91, + "learning_rate": 4.245050458362222e-05, + "loss": 0.0009, + "step": 3920 + }, + { + "epoch": 0.91, + "learning_rate": 4.244857869193436e-05, + "loss": 0.0019, + "step": 3921 + }, + { + "epoch": 0.91, + "learning_rate": 4.244665280024652e-05, + "loss": 0.0001, + "step": 3922 + }, + { + "epoch": 0.91, + "learning_rate": 4.2444726908558664e-05, + "loss": 0.0004, + "step": 3923 + }, + { + "epoch": 0.91, + "learning_rate": 4.244280101687081e-05, + "loss": 0.0, + "step": 3924 + }, + { + "epoch": 0.91, + "learning_rate": 4.244087512518296e-05, + "loss": 0.0003, + "step": 3925 + }, + { + "epoch": 0.91, + "learning_rate": 4.243894923349511e-05, + "loss": 0.0016, + "step": 3926 + }, + { + "epoch": 0.91, + "learning_rate": 4.243702334180726e-05, + "loss": 0.0, + "step": 3927 + }, + { + "epoch": 0.91, + "learning_rate": 4.243509745011941e-05, + "loss": 0.0001, + "step": 3928 + }, + { + "epoch": 0.91, + "learning_rate": 4.2433171558431556e-05, + "loss": 0.0, + "step": 3929 + }, + { + "epoch": 0.91, + "learning_rate": 4.24312456667437e-05, + "loss": 0.0, + "step": 3930 + }, + { + "epoch": 0.91, + "learning_rate": 4.242931977505585e-05, + "loss": 0.0078, + "step": 3931 + }, + { + "epoch": 0.91, + "learning_rate": 4.2427393883368e-05, + "loss": 0.0006, + "step": 3932 + }, + { + "epoch": 0.91, + "learning_rate": 4.242546799168015e-05, + "loss": 0.0001, + "step": 3933 + }, + { + "epoch": 0.91, + "learning_rate": 4.24235420999923e-05, + "loss": 0.0001, + "step": 3934 + }, + { + "epoch": 0.91, + "learning_rate": 4.242161620830445e-05, + "loss": 0.0001, + "step": 3935 + }, + { + "epoch": 0.91, + "learning_rate": 4.2419690316616595e-05, + "loss": 0.0061, + "step": 3936 + }, + { + "epoch": 0.91, + "learning_rate": 4.241776442492875e-05, + "loss": 0.0, + "step": 3937 + }, + { + "epoch": 0.91, + "learning_rate": 4.241583853324089e-05, + "loss": 0.0058, + "step": 3938 + }, + { + "epoch": 0.91, + "learning_rate": 4.241391264155304e-05, + "loss": 0.0002, + "step": 3939 + }, + { + "epoch": 0.91, + "learning_rate": 4.241198674986519e-05, + "loss": 0.0011, + "step": 3940 + }, + { + "epoch": 0.91, + "learning_rate": 4.241006085817734e-05, + "loss": 0.0, + "step": 3941 + }, + { + "epoch": 0.91, + "learning_rate": 4.2408134966489487e-05, + "loss": 0.0043, + "step": 3942 + }, + { + "epoch": 0.91, + "learning_rate": 4.2406209074801634e-05, + "loss": 0.0001, + "step": 3943 + }, + { + "epoch": 0.91, + "learning_rate": 4.240428318311379e-05, + "loss": 0.0003, + "step": 3944 + }, + { + "epoch": 0.91, + "learning_rate": 4.240235729142593e-05, + "loss": 0.0001, + "step": 3945 + }, + { + "epoch": 0.91, + "learning_rate": 4.240043139973808e-05, + "loss": 0.0002, + "step": 3946 + }, + { + "epoch": 0.91, + "learning_rate": 4.239850550805023e-05, + "loss": 0.0003, + "step": 3947 + }, + { + "epoch": 0.91, + "learning_rate": 4.239657961636238e-05, + "loss": 0.0003, + "step": 3948 + }, + { + "epoch": 0.91, + "learning_rate": 4.2394653724674526e-05, + "loss": 0.0112, + "step": 3949 + }, + { + "epoch": 0.91, + "learning_rate": 4.239272783298667e-05, + "loss": 0.0002, + "step": 3950 + }, + { + "epoch": 0.91, + "learning_rate": 4.239080194129883e-05, + "loss": 0.0001, + "step": 3951 + }, + { + "epoch": 0.91, + "learning_rate": 4.238887604961097e-05, + "loss": 0.0006, + "step": 3952 + }, + { + "epoch": 0.91, + "learning_rate": 4.238695015792312e-05, + "loss": 0.0001, + "step": 3953 + }, + { + "epoch": 0.91, + "learning_rate": 4.238502426623527e-05, + "loss": 0.0002, + "step": 3954 + }, + { + "epoch": 0.91, + "learning_rate": 4.238309837454742e-05, + "loss": 0.0001, + "step": 3955 + }, + { + "epoch": 0.91, + "learning_rate": 4.2381172482859565e-05, + "loss": 0.0, + "step": 3956 + }, + { + "epoch": 0.91, + "learning_rate": 4.237924659117171e-05, + "loss": 0.0001, + "step": 3957 + }, + { + "epoch": 0.91, + "learning_rate": 4.2377320699483867e-05, + "loss": 0.0001, + "step": 3958 + }, + { + "epoch": 0.91, + "learning_rate": 4.237539480779601e-05, + "loss": 0.0001, + "step": 3959 + }, + { + "epoch": 0.92, + "learning_rate": 4.237346891610816e-05, + "loss": 0.0001, + "step": 3960 + }, + { + "epoch": 0.92, + "learning_rate": 4.237154302442031e-05, + "loss": 0.0004, + "step": 3961 + }, + { + "epoch": 0.92, + "learning_rate": 4.2369617132732456e-05, + "loss": 0.0001, + "step": 3962 + }, + { + "epoch": 0.92, + "learning_rate": 4.2367691241044604e-05, + "loss": 0.0005, + "step": 3963 + }, + { + "epoch": 0.92, + "learning_rate": 4.236576534935676e-05, + "loss": 0.0003, + "step": 3964 + }, + { + "epoch": 0.92, + "learning_rate": 4.2363839457668906e-05, + "loss": 0.0007, + "step": 3965 + }, + { + "epoch": 0.92, + "learning_rate": 4.236191356598105e-05, + "loss": 0.0008, + "step": 3966 + }, + { + "epoch": 0.92, + "learning_rate": 4.23599876742932e-05, + "loss": 0.0005, + "step": 3967 + }, + { + "epoch": 0.92, + "learning_rate": 4.235806178260535e-05, + "loss": 0.0005, + "step": 3968 + }, + { + "epoch": 0.92, + "learning_rate": 4.2356135890917496e-05, + "loss": 0.0, + "step": 3969 + }, + { + "epoch": 0.92, + "learning_rate": 4.235420999922964e-05, + "loss": 0.0002, + "step": 3970 + }, + { + "epoch": 0.92, + "learning_rate": 4.23522841075418e-05, + "loss": 0.0, + "step": 3971 + }, + { + "epoch": 0.92, + "learning_rate": 4.2350358215853945e-05, + "loss": 0.0, + "step": 3972 + }, + { + "epoch": 0.92, + "learning_rate": 4.234843232416609e-05, + "loss": 0.0001, + "step": 3973 + }, + { + "epoch": 0.92, + "learning_rate": 4.234650643247824e-05, + "loss": 0.0029, + "step": 3974 + }, + { + "epoch": 0.92, + "learning_rate": 4.2344580540790394e-05, + "loss": 0.0006, + "step": 3975 + }, + { + "epoch": 0.92, + "learning_rate": 4.2342654649102535e-05, + "loss": 0.0002, + "step": 3976 + }, + { + "epoch": 0.92, + "learning_rate": 4.234072875741468e-05, + "loss": 0.0001, + "step": 3977 + }, + { + "epoch": 0.92, + "learning_rate": 4.2338802865726836e-05, + "loss": 0.0001, + "step": 3978 + }, + { + "epoch": 0.92, + "learning_rate": 4.233687697403898e-05, + "loss": 0.0, + "step": 3979 + }, + { + "epoch": 0.92, + "learning_rate": 4.233495108235113e-05, + "loss": 0.0004, + "step": 3980 + }, + { + "epoch": 0.92, + "learning_rate": 4.233302519066328e-05, + "loss": 0.0, + "step": 3981 + }, + { + "epoch": 0.92, + "learning_rate": 4.2331099298975426e-05, + "loss": 0.0035, + "step": 3982 + }, + { + "epoch": 0.92, + "learning_rate": 4.2329173407287574e-05, + "loss": 0.0029, + "step": 3983 + }, + { + "epoch": 0.92, + "learning_rate": 4.232724751559973e-05, + "loss": 0.0001, + "step": 3984 + }, + { + "epoch": 0.92, + "learning_rate": 4.2325321623911876e-05, + "loss": 0.0007, + "step": 3985 + }, + { + "epoch": 0.92, + "learning_rate": 4.2323395732224016e-05, + "loss": 0.0002, + "step": 3986 + }, + { + "epoch": 0.92, + "learning_rate": 4.232146984053617e-05, + "loss": 0.0001, + "step": 3987 + }, + { + "epoch": 0.92, + "learning_rate": 4.231954394884832e-05, + "loss": 0.0002, + "step": 3988 + }, + { + "epoch": 0.92, + "learning_rate": 4.2317618057160465e-05, + "loss": 0.0, + "step": 3989 + }, + { + "epoch": 0.92, + "learning_rate": 4.231569216547261e-05, + "loss": 0.0, + "step": 3990 + }, + { + "epoch": 0.92, + "learning_rate": 4.231376627378477e-05, + "loss": 0.0, + "step": 3991 + }, + { + "epoch": 0.92, + "learning_rate": 4.2311840382096915e-05, + "loss": 0.0006, + "step": 3992 + }, + { + "epoch": 0.92, + "learning_rate": 4.230991449040906e-05, + "loss": 0.0, + "step": 3993 + }, + { + "epoch": 0.92, + "learning_rate": 4.230798859872121e-05, + "loss": 0.0001, + "step": 3994 + }, + { + "epoch": 0.92, + "learning_rate": 4.230606270703336e-05, + "loss": 0.0006, + "step": 3995 + }, + { + "epoch": 0.92, + "learning_rate": 4.2304136815345505e-05, + "loss": 0.0007, + "step": 3996 + }, + { + "epoch": 0.92, + "learning_rate": 4.230221092365765e-05, + "loss": 0.0, + "step": 3997 + }, + { + "epoch": 0.92, + "learning_rate": 4.2300285031969806e-05, + "loss": 0.0091, + "step": 3998 + }, + { + "epoch": 0.92, + "learning_rate": 4.2298359140281954e-05, + "loss": 0.0005, + "step": 3999 + }, + { + "epoch": 0.92, + "learning_rate": 4.22964332485941e-05, + "loss": 0.0003, + "step": 4000 + }, + { + "epoch": 0.92, + "learning_rate": 4.229450735690625e-05, + "loss": 0.0, + "step": 4001 + }, + { + "epoch": 0.92, + "learning_rate": 4.22925814652184e-05, + "loss": 0.0001, + "step": 4002 + }, + { + "epoch": 0.93, + "learning_rate": 4.2290655573530544e-05, + "loss": 0.0205, + "step": 4003 + }, + { + "epoch": 0.93, + "learning_rate": 4.228872968184269e-05, + "loss": 0.0015, + "step": 4004 + }, + { + "epoch": 0.93, + "learning_rate": 4.2286803790154845e-05, + "loss": 0.0001, + "step": 4005 + }, + { + "epoch": 0.93, + "learning_rate": 4.228487789846699e-05, + "loss": 0.0002, + "step": 4006 + }, + { + "epoch": 0.93, + "learning_rate": 4.228295200677914e-05, + "loss": 0.0007, + "step": 4007 + }, + { + "epoch": 0.93, + "learning_rate": 4.228102611509129e-05, + "loss": 0.0001, + "step": 4008 + }, + { + "epoch": 0.93, + "learning_rate": 4.227910022340344e-05, + "loss": 0.0001, + "step": 4009 + }, + { + "epoch": 0.93, + "learning_rate": 4.227717433171558e-05, + "loss": 0.0006, + "step": 4010 + }, + { + "epoch": 0.93, + "learning_rate": 4.227524844002774e-05, + "loss": 0.0, + "step": 4011 + }, + { + "epoch": 0.93, + "learning_rate": 4.2273322548339885e-05, + "loss": 0.0002, + "step": 4012 + }, + { + "epoch": 0.93, + "learning_rate": 4.227139665665203e-05, + "loss": 0.0001, + "step": 4013 + }, + { + "epoch": 0.93, + "learning_rate": 4.226947076496418e-05, + "loss": 0.0033, + "step": 4014 + }, + { + "epoch": 0.93, + "learning_rate": 4.226754487327633e-05, + "loss": 0.0009, + "step": 4015 + }, + { + "epoch": 0.93, + "learning_rate": 4.226561898158848e-05, + "loss": 0.0001, + "step": 4016 + }, + { + "epoch": 0.93, + "learning_rate": 4.226369308990062e-05, + "loss": 0.0, + "step": 4017 + }, + { + "epoch": 0.93, + "learning_rate": 4.2261767198212776e-05, + "loss": 0.0009, + "step": 4018 + }, + { + "epoch": 0.93, + "learning_rate": 4.2259841306524924e-05, + "loss": 0.0, + "step": 4019 + }, + { + "epoch": 0.93, + "learning_rate": 4.225791541483707e-05, + "loss": 0.0033, + "step": 4020 + }, + { + "epoch": 0.93, + "learning_rate": 4.225598952314922e-05, + "loss": 0.005, + "step": 4021 + }, + { + "epoch": 0.93, + "learning_rate": 4.225406363146137e-05, + "loss": 0.0001, + "step": 4022 + }, + { + "epoch": 0.93, + "learning_rate": 4.225213773977352e-05, + "loss": 0.002, + "step": 4023 + }, + { + "epoch": 0.93, + "learning_rate": 4.225021184808566e-05, + "loss": 0.0001, + "step": 4024 + }, + { + "epoch": 0.93, + "learning_rate": 4.2248285956397815e-05, + "loss": 0.0, + "step": 4025 + }, + { + "epoch": 0.93, + "learning_rate": 4.224636006470996e-05, + "loss": 0.0001, + "step": 4026 + }, + { + "epoch": 0.93, + "learning_rate": 4.224443417302211e-05, + "loss": 0.0011, + "step": 4027 + }, + { + "epoch": 0.93, + "learning_rate": 4.224250828133426e-05, + "loss": 0.0, + "step": 4028 + }, + { + "epoch": 0.93, + "learning_rate": 4.224058238964641e-05, + "loss": 0.0002, + "step": 4029 + }, + { + "epoch": 0.93, + "learning_rate": 4.223865649795856e-05, + "loss": 0.0003, + "step": 4030 + }, + { + "epoch": 0.93, + "learning_rate": 4.223673060627071e-05, + "loss": 0.0, + "step": 4031 + }, + { + "epoch": 0.93, + "learning_rate": 4.2234804714582854e-05, + "loss": 0.0055, + "step": 4032 + }, + { + "epoch": 0.93, + "learning_rate": 4.2232878822895e-05, + "loss": 0.0005, + "step": 4033 + }, + { + "epoch": 0.93, + "learning_rate": 4.223095293120715e-05, + "loss": 0.0, + "step": 4034 + }, + { + "epoch": 0.93, + "learning_rate": 4.22290270395193e-05, + "loss": 0.0005, + "step": 4035 + }, + { + "epoch": 0.93, + "learning_rate": 4.222710114783145e-05, + "loss": 0.0001, + "step": 4036 + }, + { + "epoch": 0.93, + "learning_rate": 4.22251752561436e-05, + "loss": 0.0005, + "step": 4037 + }, + { + "epoch": 0.93, + "learning_rate": 4.2223249364455746e-05, + "loss": 0.0003, + "step": 4038 + }, + { + "epoch": 0.93, + "learning_rate": 4.2221323472767894e-05, + "loss": 0.0001, + "step": 4039 + }, + { + "epoch": 0.93, + "learning_rate": 4.221939758108005e-05, + "loss": 0.0, + "step": 4040 + }, + { + "epoch": 0.93, + "learning_rate": 4.221747168939219e-05, + "loss": 0.0, + "step": 4041 + }, + { + "epoch": 0.93, + "learning_rate": 4.2215545797704336e-05, + "loss": 0.0002, + "step": 4042 + }, + { + "epoch": 0.93, + "learning_rate": 4.221361990601649e-05, + "loss": 0.0, + "step": 4043 + }, + { + "epoch": 0.93, + "learning_rate": 4.221169401432864e-05, + "loss": 0.0002, + "step": 4044 + }, + { + "epoch": 0.93, + "learning_rate": 4.2209768122640785e-05, + "loss": 0.0018, + "step": 4045 + }, + { + "epoch": 0.94, + "learning_rate": 4.220784223095293e-05, + "loss": 0.0001, + "step": 4046 + }, + { + "epoch": 0.94, + "learning_rate": 4.220591633926509e-05, + "loss": 0.0037, + "step": 4047 + }, + { + "epoch": 0.94, + "learning_rate": 4.220399044757723e-05, + "loss": 0.0003, + "step": 4048 + }, + { + "epoch": 0.94, + "learning_rate": 4.220206455588938e-05, + "loss": 0.0002, + "step": 4049 + }, + { + "epoch": 0.94, + "learning_rate": 4.220013866420153e-05, + "loss": 0.0189, + "step": 4050 + }, + { + "epoch": 0.94, + "learning_rate": 4.219821277251367e-05, + "loss": 0.0016, + "step": 4051 + }, + { + "epoch": 0.94, + "learning_rate": 4.2196286880825824e-05, + "loss": 0.0061, + "step": 4052 + }, + { + "epoch": 0.94, + "learning_rate": 4.219436098913797e-05, + "loss": 0.0007, + "step": 4053 + }, + { + "epoch": 0.94, + "learning_rate": 4.219243509745012e-05, + "loss": 0.0, + "step": 4054 + }, + { + "epoch": 0.94, + "learning_rate": 4.219050920576227e-05, + "loss": 0.0007, + "step": 4055 + }, + { + "epoch": 0.94, + "learning_rate": 4.218858331407442e-05, + "loss": 0.0, + "step": 4056 + }, + { + "epoch": 0.94, + "learning_rate": 4.218665742238657e-05, + "loss": 0.0148, + "step": 4057 + }, + { + "epoch": 0.94, + "learning_rate": 4.2184731530698716e-05, + "loss": 0.0, + "step": 4058 + }, + { + "epoch": 0.94, + "learning_rate": 4.2182805639010863e-05, + "loss": 0.0, + "step": 4059 + }, + { + "epoch": 0.94, + "learning_rate": 4.218087974732301e-05, + "loss": 0.0006, + "step": 4060 + }, + { + "epoch": 0.94, + "learning_rate": 4.217895385563516e-05, + "loss": 0.0, + "step": 4061 + }, + { + "epoch": 0.94, + "learning_rate": 4.2177027963947306e-05, + "loss": 0.0, + "step": 4062 + }, + { + "epoch": 0.94, + "learning_rate": 4.217510207225946e-05, + "loss": 0.0003, + "step": 4063 + }, + { + "epoch": 0.94, + "learning_rate": 4.217317618057161e-05, + "loss": 0.0001, + "step": 4064 + }, + { + "epoch": 0.94, + "learning_rate": 4.2171250288883755e-05, + "loss": 0.0009, + "step": 4065 + }, + { + "epoch": 0.94, + "learning_rate": 4.21693243971959e-05, + "loss": 0.0034, + "step": 4066 + }, + { + "epoch": 0.94, + "learning_rate": 4.216739850550806e-05, + "loss": 0.0001, + "step": 4067 + }, + { + "epoch": 0.94, + "learning_rate": 4.21654726138202e-05, + "loss": 0.0034, + "step": 4068 + }, + { + "epoch": 0.94, + "learning_rate": 4.2163546722132345e-05, + "loss": 0.0, + "step": 4069 + }, + { + "epoch": 0.94, + "learning_rate": 4.21616208304445e-05, + "loss": 0.0, + "step": 4070 + }, + { + "epoch": 0.94, + "learning_rate": 4.215969493875665e-05, + "loss": 0.0001, + "step": 4071 + }, + { + "epoch": 0.94, + "learning_rate": 4.2157769047068794e-05, + "loss": 0.0003, + "step": 4072 + }, + { + "epoch": 0.94, + "learning_rate": 4.215584315538094e-05, + "loss": 0.0002, + "step": 4073 + }, + { + "epoch": 0.94, + "learning_rate": 4.2153917263693096e-05, + "loss": 0.0012, + "step": 4074 + }, + { + "epoch": 0.94, + "learning_rate": 4.2151991372005237e-05, + "loss": 0.0001, + "step": 4075 + }, + { + "epoch": 0.94, + "learning_rate": 4.215006548031739e-05, + "loss": 0.0001, + "step": 4076 + }, + { + "epoch": 0.94, + "learning_rate": 4.214813958862954e-05, + "loss": 0.0, + "step": 4077 + }, + { + "epoch": 0.94, + "learning_rate": 4.2146213696941686e-05, + "loss": 0.0043, + "step": 4078 + }, + { + "epoch": 0.94, + "learning_rate": 4.214428780525383e-05, + "loss": 0.0121, + "step": 4079 + }, + { + "epoch": 0.94, + "learning_rate": 4.214236191356598e-05, + "loss": 0.0003, + "step": 4080 + }, + { + "epoch": 0.94, + "learning_rate": 4.2140436021878135e-05, + "loss": 0.0002, + "step": 4081 + }, + { + "epoch": 0.94, + "learning_rate": 4.2138510130190276e-05, + "loss": 0.0004, + "step": 4082 + }, + { + "epoch": 0.94, + "learning_rate": 4.213658423850243e-05, + "loss": 0.0007, + "step": 4083 + }, + { + "epoch": 0.94, + "learning_rate": 4.213465834681458e-05, + "loss": 0.0007, + "step": 4084 + }, + { + "epoch": 0.94, + "learning_rate": 4.2132732455126725e-05, + "loss": 0.0003, + "step": 4085 + }, + { + "epoch": 0.94, + "learning_rate": 4.213080656343887e-05, + "loss": 0.0, + "step": 4086 + }, + { + "epoch": 0.94, + "learning_rate": 4.212888067175103e-05, + "loss": 0.0, + "step": 4087 + }, + { + "epoch": 0.94, + "learning_rate": 4.2126954780063174e-05, + "loss": 0.0007, + "step": 4088 + }, + { + "epoch": 0.94, + "learning_rate": 4.2125028888375315e-05, + "loss": 0.0039, + "step": 4089 + }, + { + "epoch": 0.95, + "learning_rate": 4.212310299668747e-05, + "loss": 0.0, + "step": 4090 + }, + { + "epoch": 0.95, + "learning_rate": 4.2121177104999617e-05, + "loss": 0.0006, + "step": 4091 + }, + { + "epoch": 0.95, + "learning_rate": 4.2119251213311764e-05, + "loss": 0.0001, + "step": 4092 + }, + { + "epoch": 0.95, + "learning_rate": 4.211732532162391e-05, + "loss": 0.0, + "step": 4093 + }, + { + "epoch": 0.95, + "learning_rate": 4.2115399429936066e-05, + "loss": 0.0, + "step": 4094 + }, + { + "epoch": 0.95, + "learning_rate": 4.211347353824821e-05, + "loss": 0.0004, + "step": 4095 + }, + { + "epoch": 0.95, + "learning_rate": 4.211154764656036e-05, + "loss": 0.0, + "step": 4096 + }, + { + "epoch": 0.95, + "learning_rate": 4.210962175487251e-05, + "loss": 0.0034, + "step": 4097 + }, + { + "epoch": 0.95, + "learning_rate": 4.2107695863184656e-05, + "loss": 0.0005, + "step": 4098 + }, + { + "epoch": 0.95, + "learning_rate": 4.21057699714968e-05, + "loss": 0.0, + "step": 4099 + }, + { + "epoch": 0.95, + "learning_rate": 4.210384407980895e-05, + "loss": 0.0089, + "step": 4100 + }, + { + "epoch": 0.95, + "learning_rate": 4.2101918188121105e-05, + "loss": 0.0022, + "step": 4101 + }, + { + "epoch": 0.95, + "learning_rate": 4.209999229643325e-05, + "loss": 0.0002, + "step": 4102 + }, + { + "epoch": 0.95, + "learning_rate": 4.20980664047454e-05, + "loss": 0.0009, + "step": 4103 + }, + { + "epoch": 0.95, + "learning_rate": 4.209614051305755e-05, + "loss": 0.0003, + "step": 4104 + }, + { + "epoch": 0.95, + "learning_rate": 4.20942146213697e-05, + "loss": 0.001, + "step": 4105 + }, + { + "epoch": 0.95, + "learning_rate": 4.209228872968184e-05, + "loss": 0.0022, + "step": 4106 + }, + { + "epoch": 0.95, + "learning_rate": 4.209036283799399e-05, + "loss": 0.0074, + "step": 4107 + }, + { + "epoch": 0.95, + "learning_rate": 4.2088436946306144e-05, + "loss": 0.0, + "step": 4108 + }, + { + "epoch": 0.95, + "learning_rate": 4.208651105461829e-05, + "loss": 0.002, + "step": 4109 + }, + { + "epoch": 0.95, + "learning_rate": 4.208458516293044e-05, + "loss": 0.0085, + "step": 4110 + }, + { + "epoch": 0.95, + "learning_rate": 4.2082659271242586e-05, + "loss": 0.0002, + "step": 4111 + }, + { + "epoch": 0.95, + "learning_rate": 4.208073337955474e-05, + "loss": 0.0, + "step": 4112 + }, + { + "epoch": 0.95, + "learning_rate": 4.207880748786688e-05, + "loss": 0.0001, + "step": 4113 + }, + { + "epoch": 0.95, + "learning_rate": 4.2076881596179036e-05, + "loss": 0.0001, + "step": 4114 + }, + { + "epoch": 0.95, + "learning_rate": 4.207495570449118e-05, + "loss": 0.0001, + "step": 4115 + }, + { + "epoch": 0.95, + "learning_rate": 4.207302981280333e-05, + "loss": 0.0049, + "step": 4116 + }, + { + "epoch": 0.95, + "learning_rate": 4.207110392111548e-05, + "loss": 0.0003, + "step": 4117 + }, + { + "epoch": 0.95, + "learning_rate": 4.2069178029427626e-05, + "loss": 0.0267, + "step": 4118 + }, + { + "epoch": 0.95, + "learning_rate": 4.206725213773978e-05, + "loss": 0.0004, + "step": 4119 + }, + { + "epoch": 0.95, + "learning_rate": 4.206532624605192e-05, + "loss": 0.0006, + "step": 4120 + }, + { + "epoch": 0.95, + "learning_rate": 4.2063400354364075e-05, + "loss": 0.0001, + "step": 4121 + }, + { + "epoch": 0.95, + "learning_rate": 4.206147446267622e-05, + "loss": 0.0003, + "step": 4122 + }, + { + "epoch": 0.95, + "learning_rate": 4.205954857098837e-05, + "loss": 0.0002, + "step": 4123 + }, + { + "epoch": 0.95, + "learning_rate": 4.205762267930052e-05, + "loss": 0.0, + "step": 4124 + }, + { + "epoch": 0.95, + "learning_rate": 4.2055696787612665e-05, + "loss": 0.0021, + "step": 4125 + }, + { + "epoch": 0.95, + "learning_rate": 4.205377089592481e-05, + "loss": 0.0008, + "step": 4126 + }, + { + "epoch": 0.95, + "learning_rate": 4.205184500423696e-05, + "loss": 0.0003, + "step": 4127 + }, + { + "epoch": 0.95, + "learning_rate": 4.2049919112549114e-05, + "loss": 0.0, + "step": 4128 + }, + { + "epoch": 0.95, + "learning_rate": 4.204799322086126e-05, + "loss": 0.0, + "step": 4129 + }, + { + "epoch": 0.95, + "learning_rate": 4.204606732917341e-05, + "loss": 0.0, + "step": 4130 + }, + { + "epoch": 0.95, + "learning_rate": 4.2044141437485556e-05, + "loss": 0.0, + "step": 4131 + }, + { + "epoch": 0.95, + "learning_rate": 4.204221554579771e-05, + "loss": 0.0023, + "step": 4132 + }, + { + "epoch": 0.96, + "learning_rate": 4.204028965410985e-05, + "loss": 0.0, + "step": 4133 + }, + { + "epoch": 0.96, + "learning_rate": 4.2038363762422006e-05, + "loss": 0.0006, + "step": 4134 + }, + { + "epoch": 0.96, + "learning_rate": 4.203643787073415e-05, + "loss": 0.0001, + "step": 4135 + }, + { + "epoch": 0.96, + "learning_rate": 4.20345119790463e-05, + "loss": 0.0, + "step": 4136 + }, + { + "epoch": 0.96, + "learning_rate": 4.203258608735845e-05, + "loss": 0.0, + "step": 4137 + }, + { + "epoch": 0.96, + "learning_rate": 4.2030660195670595e-05, + "loss": 0.0001, + "step": 4138 + }, + { + "epoch": 0.96, + "learning_rate": 4.202873430398275e-05, + "loss": 0.0, + "step": 4139 + }, + { + "epoch": 0.96, + "learning_rate": 4.202680841229489e-05, + "loss": 0.0005, + "step": 4140 + }, + { + "epoch": 0.96, + "learning_rate": 4.2024882520607045e-05, + "loss": 0.0, + "step": 4141 + }, + { + "epoch": 0.96, + "learning_rate": 4.202295662891919e-05, + "loss": 0.0018, + "step": 4142 + }, + { + "epoch": 0.96, + "learning_rate": 4.202103073723134e-05, + "loss": 0.0097, + "step": 4143 + }, + { + "epoch": 0.96, + "learning_rate": 4.201910484554349e-05, + "loss": 0.0, + "step": 4144 + }, + { + "epoch": 0.96, + "learning_rate": 4.2017178953855635e-05, + "loss": 0.0024, + "step": 4145 + }, + { + "epoch": 0.96, + "learning_rate": 4.201525306216779e-05, + "loss": 0.0006, + "step": 4146 + }, + { + "epoch": 0.96, + "learning_rate": 4.201332717047993e-05, + "loss": 0.0003, + "step": 4147 + }, + { + "epoch": 0.96, + "learning_rate": 4.2011401278792084e-05, + "loss": 0.0006, + "step": 4148 + }, + { + "epoch": 0.96, + "learning_rate": 4.200947538710423e-05, + "loss": 0.0003, + "step": 4149 + }, + { + "epoch": 0.96, + "learning_rate": 4.200754949541638e-05, + "loss": 0.0011, + "step": 4150 + }, + { + "epoch": 0.96, + "learning_rate": 4.2005623603728526e-05, + "loss": 0.0006, + "step": 4151 + }, + { + "epoch": 0.96, + "learning_rate": 4.200369771204068e-05, + "loss": 0.0, + "step": 4152 + }, + { + "epoch": 0.96, + "learning_rate": 4.200177182035283e-05, + "loss": 0.0014, + "step": 4153 + }, + { + "epoch": 0.96, + "learning_rate": 4.199984592866497e-05, + "loss": 0.003, + "step": 4154 + }, + { + "epoch": 0.96, + "learning_rate": 4.199792003697712e-05, + "loss": 0.0005, + "step": 4155 + }, + { + "epoch": 0.96, + "learning_rate": 4.199599414528927e-05, + "loss": 0.0, + "step": 4156 + }, + { + "epoch": 0.96, + "learning_rate": 4.199406825360142e-05, + "loss": 0.0003, + "step": 4157 + }, + { + "epoch": 0.96, + "learning_rate": 4.1992142361913565e-05, + "loss": 0.0007, + "step": 4158 + }, + { + "epoch": 0.96, + "learning_rate": 4.199021647022572e-05, + "loss": 0.0004, + "step": 4159 + }, + { + "epoch": 0.96, + "learning_rate": 4.198829057853787e-05, + "loss": 0.0, + "step": 4160 + }, + { + "epoch": 0.96, + "learning_rate": 4.1986364686850015e-05, + "loss": 0.0, + "step": 4161 + }, + { + "epoch": 0.96, + "learning_rate": 4.198443879516216e-05, + "loss": 0.0, + "step": 4162 + }, + { + "epoch": 0.96, + "learning_rate": 4.198251290347431e-05, + "loss": 0.0004, + "step": 4163 + }, + { + "epoch": 0.96, + "learning_rate": 4.198058701178646e-05, + "loss": 0.0, + "step": 4164 + }, + { + "epoch": 0.96, + "learning_rate": 4.1978661120098604e-05, + "loss": 0.0, + "step": 4165 + }, + { + "epoch": 0.96, + "learning_rate": 4.197673522841076e-05, + "loss": 0.0013, + "step": 4166 + }, + { + "epoch": 0.96, + "learning_rate": 4.1974809336722906e-05, + "loss": 0.0089, + "step": 4167 + }, + { + "epoch": 0.96, + "learning_rate": 4.1972883445035054e-05, + "loss": 0.0027, + "step": 4168 + }, + { + "epoch": 0.96, + "learning_rate": 4.19709575533472e-05, + "loss": 0.0003, + "step": 4169 + }, + { + "epoch": 0.96, + "learning_rate": 4.1969031661659355e-05, + "loss": 0.0, + "step": 4170 + }, + { + "epoch": 0.96, + "learning_rate": 4.1967105769971496e-05, + "loss": 0.0, + "step": 4171 + }, + { + "epoch": 0.96, + "learning_rate": 4.1965179878283644e-05, + "loss": 0.0001, + "step": 4172 + }, + { + "epoch": 0.96, + "learning_rate": 4.19632539865958e-05, + "loss": 0.0004, + "step": 4173 + }, + { + "epoch": 0.96, + "learning_rate": 4.1961328094907945e-05, + "loss": 0.0001, + "step": 4174 + }, + { + "epoch": 0.96, + "learning_rate": 4.195940220322009e-05, + "loss": 0.001, + "step": 4175 + }, + { + "epoch": 0.97, + "learning_rate": 4.195747631153224e-05, + "loss": 0.0002, + "step": 4176 + }, + { + "epoch": 0.97, + "learning_rate": 4.1955550419844395e-05, + "loss": 0.0001, + "step": 4177 + }, + { + "epoch": 0.97, + "learning_rate": 4.1953624528156535e-05, + "loss": 0.0003, + "step": 4178 + }, + { + "epoch": 0.97, + "learning_rate": 4.195169863646869e-05, + "loss": 0.0, + "step": 4179 + }, + { + "epoch": 0.97, + "learning_rate": 4.194977274478084e-05, + "loss": 0.0003, + "step": 4180 + }, + { + "epoch": 0.97, + "learning_rate": 4.1947846853092984e-05, + "loss": 0.0002, + "step": 4181 + }, + { + "epoch": 0.97, + "learning_rate": 4.194592096140513e-05, + "loss": 0.0001, + "step": 4182 + }, + { + "epoch": 0.97, + "learning_rate": 4.194399506971728e-05, + "loss": 0.0008, + "step": 4183 + }, + { + "epoch": 0.97, + "learning_rate": 4.1942069178029434e-05, + "loss": 0.0002, + "step": 4184 + }, + { + "epoch": 0.97, + "learning_rate": 4.1940143286341574e-05, + "loss": 0.0004, + "step": 4185 + }, + { + "epoch": 0.97, + "learning_rate": 4.193821739465373e-05, + "loss": 0.0, + "step": 4186 + }, + { + "epoch": 0.97, + "learning_rate": 4.1936291502965876e-05, + "loss": 0.0, + "step": 4187 + }, + { + "epoch": 0.97, + "learning_rate": 4.1934365611278024e-05, + "loss": 0.0, + "step": 4188 + }, + { + "epoch": 0.97, + "learning_rate": 4.193243971959017e-05, + "loss": 0.0054, + "step": 4189 + }, + { + "epoch": 0.97, + "learning_rate": 4.1930513827902325e-05, + "loss": 0.0001, + "step": 4190 + }, + { + "epoch": 0.97, + "learning_rate": 4.192858793621447e-05, + "loss": 0.0003, + "step": 4191 + }, + { + "epoch": 0.97, + "learning_rate": 4.1926662044526613e-05, + "loss": 0.0002, + "step": 4192 + }, + { + "epoch": 0.97, + "learning_rate": 4.192473615283877e-05, + "loss": 0.0, + "step": 4193 + }, + { + "epoch": 0.97, + "learning_rate": 4.1922810261150915e-05, + "loss": 0.0003, + "step": 4194 + }, + { + "epoch": 0.97, + "learning_rate": 4.192088436946306e-05, + "loss": 0.0021, + "step": 4195 + }, + { + "epoch": 0.97, + "learning_rate": 4.191895847777521e-05, + "loss": 0.0, + "step": 4196 + }, + { + "epoch": 0.97, + "learning_rate": 4.1917032586087364e-05, + "loss": 0.0, + "step": 4197 + }, + { + "epoch": 0.97, + "learning_rate": 4.191510669439951e-05, + "loss": 0.0001, + "step": 4198 + }, + { + "epoch": 0.97, + "learning_rate": 4.191318080271166e-05, + "loss": 0.0001, + "step": 4199 + }, + { + "epoch": 0.97, + "learning_rate": 4.191125491102381e-05, + "loss": 0.0002, + "step": 4200 + }, + { + "epoch": 0.97, + "learning_rate": 4.1909329019335954e-05, + "loss": 0.0, + "step": 4201 + }, + { + "epoch": 0.97, + "learning_rate": 4.19074031276481e-05, + "loss": 0.0001, + "step": 4202 + }, + { + "epoch": 0.97, + "learning_rate": 4.190547723596025e-05, + "loss": 0.0001, + "step": 4203 + }, + { + "epoch": 0.97, + "learning_rate": 4.1903551344272404e-05, + "loss": 0.0003, + "step": 4204 + }, + { + "epoch": 0.97, + "learning_rate": 4.1901625452584544e-05, + "loss": 0.0, + "step": 4205 + }, + { + "epoch": 0.97, + "learning_rate": 4.18996995608967e-05, + "loss": 0.0, + "step": 4206 + }, + { + "epoch": 0.97, + "learning_rate": 4.1897773669208846e-05, + "loss": 0.0001, + "step": 4207 + }, + { + "epoch": 0.97, + "learning_rate": 4.1895847777520993e-05, + "loss": 0.0001, + "step": 4208 + }, + { + "epoch": 0.97, + "learning_rate": 4.189392188583314e-05, + "loss": 0.0024, + "step": 4209 + }, + { + "epoch": 0.97, + "learning_rate": 4.189199599414529e-05, + "loss": 0.0004, + "step": 4210 + }, + { + "epoch": 0.97, + "learning_rate": 4.189007010245744e-05, + "loss": 0.0003, + "step": 4211 + }, + { + "epoch": 0.97, + "learning_rate": 4.188814421076958e-05, + "loss": 0.0001, + "step": 4212 + }, + { + "epoch": 0.97, + "learning_rate": 4.188621831908174e-05, + "loss": 0.0001, + "step": 4213 + }, + { + "epoch": 0.97, + "learning_rate": 4.1884292427393885e-05, + "loss": 0.0007, + "step": 4214 + }, + { + "epoch": 0.97, + "learning_rate": 4.188236653570603e-05, + "loss": 0.0016, + "step": 4215 + }, + { + "epoch": 0.97, + "learning_rate": 4.188044064401818e-05, + "loss": 0.0, + "step": 4216 + }, + { + "epoch": 0.97, + "learning_rate": 4.1878514752330334e-05, + "loss": 0.0002, + "step": 4217 + }, + { + "epoch": 0.97, + "learning_rate": 4.187658886064248e-05, + "loss": 0.0001, + "step": 4218 + }, + { + "epoch": 0.98, + "learning_rate": 4.187466296895462e-05, + "loss": 0.0, + "step": 4219 + }, + { + "epoch": 0.98, + "learning_rate": 4.187273707726678e-05, + "loss": 0.0, + "step": 4220 + }, + { + "epoch": 0.98, + "learning_rate": 4.1870811185578924e-05, + "loss": 0.0001, + "step": 4221 + }, + { + "epoch": 0.98, + "learning_rate": 4.186888529389107e-05, + "loss": 0.0, + "step": 4222 + }, + { + "epoch": 0.98, + "learning_rate": 4.186695940220322e-05, + "loss": 0.0015, + "step": 4223 + }, + { + "epoch": 0.98, + "learning_rate": 4.1865033510515373e-05, + "loss": 0.0006, + "step": 4224 + }, + { + "epoch": 0.98, + "learning_rate": 4.186310761882752e-05, + "loss": 0.0, + "step": 4225 + }, + { + "epoch": 0.98, + "learning_rate": 4.186118172713967e-05, + "loss": 0.0, + "step": 4226 + }, + { + "epoch": 0.98, + "learning_rate": 4.1859255835451816e-05, + "loss": 0.0001, + "step": 4227 + }, + { + "epoch": 0.98, + "learning_rate": 4.185732994376396e-05, + "loss": 0.0, + "step": 4228 + }, + { + "epoch": 0.98, + "learning_rate": 4.185540405207611e-05, + "loss": 0.0008, + "step": 4229 + }, + { + "epoch": 0.98, + "learning_rate": 4.185347816038826e-05, + "loss": 0.0, + "step": 4230 + }, + { + "epoch": 0.98, + "learning_rate": 4.185155226870041e-05, + "loss": 0.0001, + "step": 4231 + }, + { + "epoch": 0.98, + "learning_rate": 4.184962637701256e-05, + "loss": 0.0, + "step": 4232 + }, + { + "epoch": 0.98, + "learning_rate": 4.184770048532471e-05, + "loss": 0.0002, + "step": 4233 + }, + { + "epoch": 0.98, + "learning_rate": 4.1845774593636855e-05, + "loss": 0.0, + "step": 4234 + }, + { + "epoch": 0.98, + "learning_rate": 4.184384870194901e-05, + "loss": 0.0, + "step": 4235 + }, + { + "epoch": 0.98, + "learning_rate": 4.184192281026115e-05, + "loss": 0.0, + "step": 4236 + }, + { + "epoch": 0.98, + "learning_rate": 4.18399969185733e-05, + "loss": 0.0002, + "step": 4237 + }, + { + "epoch": 0.98, + "learning_rate": 4.183807102688545e-05, + "loss": 0.0001, + "step": 4238 + }, + { + "epoch": 0.98, + "learning_rate": 4.18361451351976e-05, + "loss": 0.0, + "step": 4239 + }, + { + "epoch": 0.98, + "learning_rate": 4.183421924350975e-05, + "loss": 0.0, + "step": 4240 + }, + { + "epoch": 0.98, + "learning_rate": 4.1832293351821894e-05, + "loss": 0.0, + "step": 4241 + }, + { + "epoch": 0.98, + "learning_rate": 4.183036746013405e-05, + "loss": 0.0003, + "step": 4242 + }, + { + "epoch": 0.98, + "learning_rate": 4.182844156844619e-05, + "loss": 0.0271, + "step": 4243 + }, + { + "epoch": 0.98, + "learning_rate": 4.182651567675834e-05, + "loss": 0.0204, + "step": 4244 + }, + { + "epoch": 0.98, + "learning_rate": 4.182458978507049e-05, + "loss": 0.0007, + "step": 4245 + }, + { + "epoch": 0.98, + "learning_rate": 4.182266389338264e-05, + "loss": 0.0153, + "step": 4246 + }, + { + "epoch": 0.98, + "learning_rate": 4.1820738001694786e-05, + "loss": 0.0007, + "step": 4247 + }, + { + "epoch": 0.98, + "learning_rate": 4.181881211000693e-05, + "loss": 0.0, + "step": 4248 + }, + { + "epoch": 0.98, + "learning_rate": 4.181688621831909e-05, + "loss": 0.0, + "step": 4249 + }, + { + "epoch": 0.98, + "learning_rate": 4.181496032663123e-05, + "loss": 0.0001, + "step": 4250 + }, + { + "epoch": 0.98, + "learning_rate": 4.181303443494338e-05, + "loss": 0.0001, + "step": 4251 + }, + { + "epoch": 0.98, + "learning_rate": 4.181110854325553e-05, + "loss": 0.0, + "step": 4252 + }, + { + "epoch": 0.98, + "learning_rate": 4.180918265156768e-05, + "loss": 0.0, + "step": 4253 + }, + { + "epoch": 0.98, + "learning_rate": 4.1807256759879825e-05, + "loss": 0.0, + "step": 4254 + }, + { + "epoch": 0.98, + "learning_rate": 4.180533086819198e-05, + "loss": 0.0001, + "step": 4255 + }, + { + "epoch": 0.98, + "learning_rate": 4.1803404976504127e-05, + "loss": 0.0, + "step": 4256 + }, + { + "epoch": 0.98, + "learning_rate": 4.180147908481627e-05, + "loss": 0.0011, + "step": 4257 + }, + { + "epoch": 0.98, + "learning_rate": 4.179955319312842e-05, + "loss": 0.0, + "step": 4258 + }, + { + "epoch": 0.98, + "learning_rate": 4.179762730144057e-05, + "loss": 0.0001, + "step": 4259 + }, + { + "epoch": 0.98, + "learning_rate": 4.1795701409752717e-05, + "loss": 0.0007, + "step": 4260 + }, + { + "epoch": 0.98, + "learning_rate": 4.1793775518064864e-05, + "loss": 0.0, + "step": 4261 + }, + { + "epoch": 0.98, + "learning_rate": 4.179184962637702e-05, + "loss": 0.0, + "step": 4262 + }, + { + "epoch": 0.99, + "learning_rate": 4.1789923734689166e-05, + "loss": 0.0, + "step": 4263 + }, + { + "epoch": 0.99, + "learning_rate": 4.178799784300131e-05, + "loss": 0.0, + "step": 4264 + }, + { + "epoch": 0.99, + "learning_rate": 4.178607195131346e-05, + "loss": 0.0013, + "step": 4265 + }, + { + "epoch": 0.99, + "learning_rate": 4.178414605962561e-05, + "loss": 0.0004, + "step": 4266 + }, + { + "epoch": 0.99, + "learning_rate": 4.1782220167937756e-05, + "loss": 0.0, + "step": 4267 + }, + { + "epoch": 0.99, + "learning_rate": 4.17802942762499e-05, + "loss": 0.0, + "step": 4268 + }, + { + "epoch": 0.99, + "learning_rate": 4.177836838456206e-05, + "loss": 0.0, + "step": 4269 + }, + { + "epoch": 0.99, + "learning_rate": 4.1776442492874205e-05, + "loss": 0.0001, + "step": 4270 + }, + { + "epoch": 0.99, + "learning_rate": 4.177451660118635e-05, + "loss": 0.0003, + "step": 4271 + }, + { + "epoch": 0.99, + "learning_rate": 4.17725907094985e-05, + "loss": 0.0002, + "step": 4272 + }, + { + "epoch": 0.99, + "learning_rate": 4.1770664817810654e-05, + "loss": 0.0041, + "step": 4273 + }, + { + "epoch": 0.99, + "learning_rate": 4.1768738926122795e-05, + "loss": 0.0003, + "step": 4274 + }, + { + "epoch": 0.99, + "learning_rate": 4.176681303443494e-05, + "loss": 0.0003, + "step": 4275 + }, + { + "epoch": 0.99, + "learning_rate": 4.1764887142747096e-05, + "loss": 0.0, + "step": 4276 + }, + { + "epoch": 0.99, + "learning_rate": 4.176296125105924e-05, + "loss": 0.0008, + "step": 4277 + }, + { + "epoch": 0.99, + "learning_rate": 4.176103535937139e-05, + "loss": 0.0, + "step": 4278 + }, + { + "epoch": 0.99, + "learning_rate": 4.175910946768354e-05, + "loss": 0.0002, + "step": 4279 + }, + { + "epoch": 0.99, + "learning_rate": 4.1757183575995686e-05, + "loss": 0.006, + "step": 4280 + }, + { + "epoch": 0.99, + "learning_rate": 4.1755257684307834e-05, + "loss": 0.0, + "step": 4281 + }, + { + "epoch": 0.99, + "learning_rate": 4.175333179261999e-05, + "loss": 0.0, + "step": 4282 + }, + { + "epoch": 0.99, + "learning_rate": 4.1751405900932136e-05, + "loss": 0.0001, + "step": 4283 + }, + { + "epoch": 0.99, + "learning_rate": 4.1749480009244276e-05, + "loss": 0.0002, + "step": 4284 + }, + { + "epoch": 0.99, + "learning_rate": 4.174755411755643e-05, + "loss": 0.0005, + "step": 4285 + }, + { + "epoch": 0.99, + "learning_rate": 4.174562822586858e-05, + "loss": 0.0007, + "step": 4286 + }, + { + "epoch": 0.99, + "learning_rate": 4.1743702334180726e-05, + "loss": 0.0, + "step": 4287 + }, + { + "epoch": 0.99, + "learning_rate": 4.174177644249287e-05, + "loss": 0.0, + "step": 4288 + }, + { + "epoch": 0.99, + "learning_rate": 4.173985055080503e-05, + "loss": 0.0053, + "step": 4289 + }, + { + "epoch": 0.99, + "learning_rate": 4.1737924659117175e-05, + "loss": 0.0004, + "step": 4290 + }, + { + "epoch": 0.99, + "learning_rate": 4.173599876742932e-05, + "loss": 0.0005, + "step": 4291 + }, + { + "epoch": 0.99, + "learning_rate": 4.173407287574147e-05, + "loss": 0.0017, + "step": 4292 + }, + { + "epoch": 0.99, + "learning_rate": 4.173214698405362e-05, + "loss": 0.0084, + "step": 4293 + }, + { + "epoch": 0.99, + "learning_rate": 4.1730221092365765e-05, + "loss": 0.0, + "step": 4294 + }, + { + "epoch": 0.99, + "learning_rate": 4.172829520067791e-05, + "loss": 0.0, + "step": 4295 + }, + { + "epoch": 0.99, + "learning_rate": 4.1726369308990066e-05, + "loss": 0.0, + "step": 4296 + }, + { + "epoch": 0.99, + "learning_rate": 4.1724443417302214e-05, + "loss": 0.0, + "step": 4297 + }, + { + "epoch": 0.99, + "learning_rate": 4.172251752561436e-05, + "loss": 0.0, + "step": 4298 + }, + { + "epoch": 0.99, + "learning_rate": 4.172059163392651e-05, + "loss": 0.0023, + "step": 4299 + }, + { + "epoch": 0.99, + "learning_rate": 4.171866574223866e-05, + "loss": 0.0005, + "step": 4300 + }, + { + "epoch": 0.99, + "learning_rate": 4.1716739850550804e-05, + "loss": 0.0019, + "step": 4301 + }, + { + "epoch": 0.99, + "learning_rate": 4.171481395886296e-05, + "loss": 0.0001, + "step": 4302 + }, + { + "epoch": 0.99, + "learning_rate": 4.1712888067175105e-05, + "loss": 0.0004, + "step": 4303 + }, + { + "epoch": 0.99, + "learning_rate": 4.171096217548725e-05, + "loss": 0.0002, + "step": 4304 + }, + { + "epoch": 0.99, + "learning_rate": 4.17090362837994e-05, + "loss": 0.0027, + "step": 4305 + }, + { + "epoch": 1.0, + "learning_rate": 4.170711039211155e-05, + "loss": 0.0006, + "step": 4306 + }, + { + "epoch": 1.0, + "learning_rate": 4.17051845004237e-05, + "loss": 0.0004, + "step": 4307 + }, + { + "epoch": 1.0, + "learning_rate": 4.170325860873584e-05, + "loss": 0.0051, + "step": 4308 + }, + { + "epoch": 1.0, + "learning_rate": 4.1701332717048e-05, + "loss": 0.0001, + "step": 4309 + }, + { + "epoch": 1.0, + "learning_rate": 4.1699406825360145e-05, + "loss": 0.0, + "step": 4310 + }, + { + "epoch": 1.0, + "learning_rate": 4.169748093367229e-05, + "loss": 0.0036, + "step": 4311 + }, + { + "epoch": 1.0, + "learning_rate": 4.169555504198444e-05, + "loss": 0.0009, + "step": 4312 + }, + { + "epoch": 1.0, + "learning_rate": 4.169362915029659e-05, + "loss": 0.0, + "step": 4313 + }, + { + "epoch": 1.0, + "learning_rate": 4.169170325860874e-05, + "loss": 0.0017, + "step": 4314 + }, + { + "epoch": 1.0, + "learning_rate": 4.168977736692088e-05, + "loss": 0.0, + "step": 4315 + }, + { + "epoch": 1.0, + "learning_rate": 4.1687851475233036e-05, + "loss": 0.0095, + "step": 4316 + }, + { + "epoch": 1.0, + "learning_rate": 4.1685925583545184e-05, + "loss": 0.0032, + "step": 4317 + }, + { + "epoch": 1.0, + "learning_rate": 4.168399969185733e-05, + "loss": 0.0015, + "step": 4318 + }, + { + "epoch": 1.0, + "learning_rate": 4.168207380016948e-05, + "loss": 0.0004, + "step": 4319 + }, + { + "epoch": 1.0, + "learning_rate": 4.168014790848163e-05, + "loss": 0.0002, + "step": 4320 + }, + { + "epoch": 1.0, + "learning_rate": 4.167822201679378e-05, + "loss": 0.0001, + "step": 4321 + }, + { + "epoch": 1.0, + "learning_rate": 4.167629612510592e-05, + "loss": 0.0001, + "step": 4322 + }, + { + "epoch": 1.0, + "learning_rate": 4.1674370233418075e-05, + "loss": 0.0002, + "step": 4323 + }, + { + "epoch": 1.0, + "learning_rate": 4.167244434173022e-05, + "loss": 0.0, + "step": 4324 + }, + { + "epoch": 1.0, + "learning_rate": 4.167051845004237e-05, + "loss": 0.0, + "step": 4325 + }, + { + "epoch": 1.0, + "learning_rate": 4.166859255835452e-05, + "loss": 0.0002, + "step": 4326 + }, + { + "epoch": 1.0, + "learning_rate": 4.166666666666667e-05, + "loss": 0.0007, + "step": 4327 + }, + { + "epoch": 1.0, + "learning_rate": 4.166474077497882e-05, + "loss": 0.0005, + "step": 4328 + }, + { + "epoch": 1.0, + "learning_rate": 4.166281488329097e-05, + "loss": 0.0015, + "step": 4329 + }, + { + "epoch": 1.0, + "learning_rate": 4.1660888991603114e-05, + "loss": 0.0056, + "step": 4330 + }, + { + "epoch": 1.0, + "learning_rate": 4.165896309991526e-05, + "loss": 0.0, + "step": 4331 + }, + { + "epoch": 1.0, + "learning_rate": 4.165703720822741e-05, + "loss": 0.0, + "step": 4332 + }, + { + "epoch": 1.0, + "learning_rate": 4.165511131653956e-05, + "loss": 0.0, + "step": 4333 + }, + { + "epoch": 1.0, + "learning_rate": 4.165318542485171e-05, + "loss": 0.0, + "step": 4334 + }, + { + "epoch": 1.0, + "learning_rate": 4.165125953316386e-05, + "loss": 0.0, + "step": 4335 + }, + { + "epoch": 1.0, + "learning_rate": 4.1649333641476006e-05, + "loss": 0.0009, + "step": 4336 + }, + { + "epoch": 1.0, + "learning_rate": 4.1647407749788154e-05, + "loss": 0.0005, + "step": 4337 + }, + { + "epoch": 1.0, + "learning_rate": 4.164548185810031e-05, + "loss": 0.0, + "step": 4338 + }, + { + "epoch": 1.0, + "learning_rate": 4.164355596641245e-05, + "loss": 0.0, + "step": 4339 + }, + { + "epoch": 1.0, + "learning_rate": 4.1641630074724596e-05, + "loss": 0.0, + "step": 4340 + }, + { + "epoch": 1.0, + "learning_rate": 4.163970418303675e-05, + "loss": 0.0006, + "step": 4341 + }, + { + "epoch": 1.0, + "learning_rate": 4.16377782913489e-05, + "loss": 0.0001, + "step": 4342 + }, + { + "epoch": 1.0, + "learning_rate": 4.1635852399661045e-05, + "loss": 0.0, + "step": 4343 + }, + { + "epoch": 1.0, + "learning_rate": 4.163392650797319e-05, + "loss": 0.0004, + "step": 4344 + }, + { + "epoch": 1.0, + "learning_rate": 4.163200061628535e-05, + "loss": 0.0001, + "step": 4345 + }, + { + "epoch": 1.0, + "learning_rate": 4.163007472459749e-05, + "loss": 0.0001, + "step": 4346 + }, + { + "epoch": 1.0, + "learning_rate": 4.162814883290964e-05, + "loss": 0.0002, + "step": 4347 + }, + { + "epoch": 1.0, + "learning_rate": 4.162622294122179e-05, + "loss": 0.0, + "step": 4348 + }, + { + "epoch": 1.01, + "learning_rate": 4.162429704953393e-05, + "loss": 0.0004, + "step": 4349 + }, + { + "epoch": 1.01, + "learning_rate": 4.1622371157846084e-05, + "loss": 0.0002, + "step": 4350 + }, + { + "epoch": 1.01, + "learning_rate": 4.162044526615823e-05, + "loss": 0.0, + "step": 4351 + }, + { + "epoch": 1.01, + "learning_rate": 4.1618519374470386e-05, + "loss": 0.0002, + "step": 4352 + }, + { + "epoch": 1.01, + "learning_rate": 4.161659348278253e-05, + "loss": 0.0, + "step": 4353 + }, + { + "epoch": 1.01, + "learning_rate": 4.161466759109468e-05, + "loss": 0.0008, + "step": 4354 + }, + { + "epoch": 1.01, + "learning_rate": 4.161274169940683e-05, + "loss": 0.0, + "step": 4355 + }, + { + "epoch": 1.01, + "learning_rate": 4.1610815807718976e-05, + "loss": 0.0, + "step": 4356 + }, + { + "epoch": 1.01, + "learning_rate": 4.1608889916031123e-05, + "loss": 0.0001, + "step": 4357 + }, + { + "epoch": 1.01, + "learning_rate": 4.160696402434328e-05, + "loss": 0.0, + "step": 4358 + }, + { + "epoch": 1.01, + "learning_rate": 4.160503813265542e-05, + "loss": 0.0001, + "step": 4359 + }, + { + "epoch": 1.01, + "learning_rate": 4.1603112240967566e-05, + "loss": 0.0, + "step": 4360 + }, + { + "epoch": 1.01, + "learning_rate": 4.160118634927972e-05, + "loss": 0.0012, + "step": 4361 + }, + { + "epoch": 1.01, + "learning_rate": 4.159926045759187e-05, + "loss": 0.0075, + "step": 4362 + }, + { + "epoch": 1.01, + "learning_rate": 4.1597334565904015e-05, + "loss": 0.0068, + "step": 4363 + }, + { + "epoch": 1.01, + "learning_rate": 4.159540867421616e-05, + "loss": 0.0521, + "step": 4364 + }, + { + "epoch": 1.01, + "learning_rate": 4.159348278252832e-05, + "loss": 0.0001, + "step": 4365 + }, + { + "epoch": 1.01, + "learning_rate": 4.159155689084046e-05, + "loss": 0.0, + "step": 4366 + }, + { + "epoch": 1.01, + "learning_rate": 4.158963099915261e-05, + "loss": 0.0001, + "step": 4367 + }, + { + "epoch": 1.01, + "learning_rate": 4.158770510746476e-05, + "loss": 0.0002, + "step": 4368 + }, + { + "epoch": 1.01, + "learning_rate": 4.158577921577691e-05, + "loss": 0.0078, + "step": 4369 + }, + { + "epoch": 1.01, + "learning_rate": 4.1583853324089054e-05, + "loss": 0.0007, + "step": 4370 + }, + { + "epoch": 1.01, + "learning_rate": 4.15819274324012e-05, + "loss": 0.0, + "step": 4371 + }, + { + "epoch": 1.01, + "learning_rate": 4.1580001540713356e-05, + "loss": 0.0004, + "step": 4372 + }, + { + "epoch": 1.01, + "learning_rate": 4.15780756490255e-05, + "loss": 0.0035, + "step": 4373 + }, + { + "epoch": 1.01, + "learning_rate": 4.157614975733765e-05, + "loss": 0.0004, + "step": 4374 + }, + { + "epoch": 1.01, + "learning_rate": 4.15742238656498e-05, + "loss": 0.0002, + "step": 4375 + }, + { + "epoch": 1.01, + "learning_rate": 4.1572297973961946e-05, + "loss": 0.0002, + "step": 4376 + }, + { + "epoch": 1.01, + "learning_rate": 4.157037208227409e-05, + "loss": 0.0014, + "step": 4377 + }, + { + "epoch": 1.01, + "learning_rate": 4.156844619058624e-05, + "loss": 0.0004, + "step": 4378 + }, + { + "epoch": 1.01, + "learning_rate": 4.1566520298898395e-05, + "loss": 0.0004, + "step": 4379 + }, + { + "epoch": 1.01, + "learning_rate": 4.1564594407210536e-05, + "loss": 0.0003, + "step": 4380 + }, + { + "epoch": 1.01, + "learning_rate": 4.156266851552269e-05, + "loss": 0.0005, + "step": 4381 + }, + { + "epoch": 1.01, + "learning_rate": 4.156074262383484e-05, + "loss": 0.0, + "step": 4382 + }, + { + "epoch": 1.01, + "learning_rate": 4.1558816732146985e-05, + "loss": 0.0023, + "step": 4383 + }, + { + "epoch": 1.01, + "learning_rate": 4.155689084045913e-05, + "loss": 0.0001, + "step": 4384 + }, + { + "epoch": 1.01, + "learning_rate": 4.155496494877129e-05, + "loss": 0.0072, + "step": 4385 + }, + { + "epoch": 1.01, + "learning_rate": 4.1553039057083434e-05, + "loss": 0.0001, + "step": 4386 + }, + { + "epoch": 1.01, + "learning_rate": 4.1551113165395575e-05, + "loss": 0.0, + "step": 4387 + }, + { + "epoch": 1.01, + "learning_rate": 4.154918727370773e-05, + "loss": 0.0001, + "step": 4388 + }, + { + "epoch": 1.01, + "learning_rate": 4.154726138201988e-05, + "loss": 0.0, + "step": 4389 + }, + { + "epoch": 1.01, + "learning_rate": 4.1545335490332024e-05, + "loss": 0.0021, + "step": 4390 + }, + { + "epoch": 1.01, + "learning_rate": 4.154340959864417e-05, + "loss": 0.0, + "step": 4391 + }, + { + "epoch": 1.02, + "learning_rate": 4.1541483706956326e-05, + "loss": 0.0, + "step": 4392 + }, + { + "epoch": 1.02, + "learning_rate": 4.153955781526847e-05, + "loss": 0.0002, + "step": 4393 + }, + { + "epoch": 1.02, + "learning_rate": 4.153763192358062e-05, + "loss": 0.0002, + "step": 4394 + }, + { + "epoch": 1.02, + "learning_rate": 4.153570603189277e-05, + "loss": 0.0, + "step": 4395 + }, + { + "epoch": 1.02, + "learning_rate": 4.1533780140204916e-05, + "loss": 0.0125, + "step": 4396 + }, + { + "epoch": 1.02, + "learning_rate": 4.153185424851706e-05, + "loss": 0.0007, + "step": 4397 + }, + { + "epoch": 1.02, + "learning_rate": 4.152992835682921e-05, + "loss": 0.0001, + "step": 4398 + }, + { + "epoch": 1.02, + "learning_rate": 4.1528002465141365e-05, + "loss": 0.002, + "step": 4399 + }, + { + "epoch": 1.02, + "learning_rate": 4.152607657345351e-05, + "loss": 0.0003, + "step": 4400 + }, + { + "epoch": 1.02, + "learning_rate": 4.152415068176566e-05, + "loss": 0.0001, + "step": 4401 + }, + { + "epoch": 1.02, + "learning_rate": 4.152222479007781e-05, + "loss": 0.0001, + "step": 4402 + }, + { + "epoch": 1.02, + "learning_rate": 4.152029889838996e-05, + "loss": 0.0, + "step": 4403 + }, + { + "epoch": 1.02, + "learning_rate": 4.15183730067021e-05, + "loss": 0.0183, + "step": 4404 + }, + { + "epoch": 1.02, + "learning_rate": 4.151644711501425e-05, + "loss": 0.0001, + "step": 4405 + }, + { + "epoch": 1.02, + "learning_rate": 4.1514521223326404e-05, + "loss": 0.0, + "step": 4406 + }, + { + "epoch": 1.02, + "learning_rate": 4.151259533163855e-05, + "loss": 0.0011, + "step": 4407 + }, + { + "epoch": 1.02, + "learning_rate": 4.15106694399507e-05, + "loss": 0.0001, + "step": 4408 + }, + { + "epoch": 1.02, + "learning_rate": 4.1508743548262847e-05, + "loss": 0.0001, + "step": 4409 + }, + { + "epoch": 1.02, + "learning_rate": 4.1506817656575e-05, + "loss": 0.0, + "step": 4410 + }, + { + "epoch": 1.02, + "learning_rate": 4.150489176488714e-05, + "loss": 0.0, + "step": 4411 + }, + { + "epoch": 1.02, + "learning_rate": 4.1502965873199296e-05, + "loss": 0.0001, + "step": 4412 + }, + { + "epoch": 1.02, + "learning_rate": 4.150103998151144e-05, + "loss": 0.0009, + "step": 4413 + }, + { + "epoch": 1.02, + "learning_rate": 4.149911408982359e-05, + "loss": 0.0001, + "step": 4414 + }, + { + "epoch": 1.02, + "learning_rate": 4.149718819813574e-05, + "loss": 0.0022, + "step": 4415 + }, + { + "epoch": 1.02, + "learning_rate": 4.1495262306447886e-05, + "loss": 0.0001, + "step": 4416 + }, + { + "epoch": 1.02, + "learning_rate": 4.149333641476004e-05, + "loss": 0.0003, + "step": 4417 + }, + { + "epoch": 1.02, + "learning_rate": 4.149141052307218e-05, + "loss": 0.0011, + "step": 4418 + }, + { + "epoch": 1.02, + "learning_rate": 4.1489484631384335e-05, + "loss": 0.0, + "step": 4419 + }, + { + "epoch": 1.02, + "learning_rate": 4.148755873969648e-05, + "loss": 0.0, + "step": 4420 + }, + { + "epoch": 1.02, + "learning_rate": 4.148563284800863e-05, + "loss": 0.0023, + "step": 4421 + }, + { + "epoch": 1.02, + "learning_rate": 4.148370695632078e-05, + "loss": 0.0004, + "step": 4422 + }, + { + "epoch": 1.02, + "learning_rate": 4.148178106463293e-05, + "loss": 0.0, + "step": 4423 + }, + { + "epoch": 1.02, + "learning_rate": 4.147985517294508e-05, + "loss": 0.0001, + "step": 4424 + }, + { + "epoch": 1.02, + "learning_rate": 4.147792928125722e-05, + "loss": 0.0, + "step": 4425 + }, + { + "epoch": 1.02, + "learning_rate": 4.1476003389569374e-05, + "loss": 0.0, + "step": 4426 + }, + { + "epoch": 1.02, + "learning_rate": 4.147407749788152e-05, + "loss": 0.0001, + "step": 4427 + }, + { + "epoch": 1.02, + "learning_rate": 4.147215160619367e-05, + "loss": 0.0036, + "step": 4428 + }, + { + "epoch": 1.02, + "learning_rate": 4.1470225714505816e-05, + "loss": 0.0, + "step": 4429 + }, + { + "epoch": 1.02, + "learning_rate": 4.146829982281797e-05, + "loss": 0.0001, + "step": 4430 + }, + { + "epoch": 1.02, + "learning_rate": 4.146637393113011e-05, + "loss": 0.0, + "step": 4431 + }, + { + "epoch": 1.02, + "learning_rate": 4.1464448039442266e-05, + "loss": 0.0, + "step": 4432 + }, + { + "epoch": 1.02, + "learning_rate": 4.146252214775441e-05, + "loss": 0.0003, + "step": 4433 + }, + { + "epoch": 1.02, + "learning_rate": 4.146059625606656e-05, + "loss": 0.0016, + "step": 4434 + }, + { + "epoch": 1.02, + "learning_rate": 4.145867036437871e-05, + "loss": 0.0293, + "step": 4435 + }, + { + "epoch": 1.03, + "learning_rate": 4.1456744472690856e-05, + "loss": 0.0033, + "step": 4436 + }, + { + "epoch": 1.03, + "learning_rate": 4.145481858100301e-05, + "loss": 0.0015, + "step": 4437 + }, + { + "epoch": 1.03, + "learning_rate": 4.145289268931515e-05, + "loss": 0.0, + "step": 4438 + }, + { + "epoch": 1.03, + "learning_rate": 4.1450966797627305e-05, + "loss": 0.0002, + "step": 4439 + }, + { + "epoch": 1.03, + "learning_rate": 4.144904090593945e-05, + "loss": 0.002, + "step": 4440 + }, + { + "epoch": 1.03, + "learning_rate": 4.14471150142516e-05, + "loss": 0.0008, + "step": 4441 + }, + { + "epoch": 1.03, + "learning_rate": 4.144518912256375e-05, + "loss": 0.0004, + "step": 4442 + }, + { + "epoch": 1.03, + "learning_rate": 4.1443263230875895e-05, + "loss": 0.0, + "step": 4443 + }, + { + "epoch": 1.03, + "learning_rate": 4.144133733918805e-05, + "loss": 0.0002, + "step": 4444 + }, + { + "epoch": 1.03, + "learning_rate": 4.143941144750019e-05, + "loss": 0.0, + "step": 4445 + }, + { + "epoch": 1.03, + "learning_rate": 4.1437485555812344e-05, + "loss": 0.0, + "step": 4446 + }, + { + "epoch": 1.03, + "learning_rate": 4.143555966412449e-05, + "loss": 0.0, + "step": 4447 + }, + { + "epoch": 1.03, + "learning_rate": 4.143363377243664e-05, + "loss": 0.0002, + "step": 4448 + }, + { + "epoch": 1.03, + "learning_rate": 4.1431707880748786e-05, + "loss": 0.0, + "step": 4449 + }, + { + "epoch": 1.03, + "learning_rate": 4.142978198906094e-05, + "loss": 0.0001, + "step": 4450 + }, + { + "epoch": 1.03, + "learning_rate": 4.142785609737309e-05, + "loss": 0.001, + "step": 4451 + }, + { + "epoch": 1.03, + "learning_rate": 4.142593020568523e-05, + "loss": 0.0123, + "step": 4452 + }, + { + "epoch": 1.03, + "learning_rate": 4.142400431399738e-05, + "loss": 0.0001, + "step": 4453 + }, + { + "epoch": 1.03, + "learning_rate": 4.142207842230953e-05, + "loss": 0.0, + "step": 4454 + }, + { + "epoch": 1.03, + "learning_rate": 4.142015253062168e-05, + "loss": 0.0003, + "step": 4455 + }, + { + "epoch": 1.03, + "learning_rate": 4.1418226638933825e-05, + "loss": 0.0, + "step": 4456 + }, + { + "epoch": 1.03, + "learning_rate": 4.141630074724598e-05, + "loss": 0.0005, + "step": 4457 + }, + { + "epoch": 1.03, + "learning_rate": 4.141437485555813e-05, + "loss": 0.0004, + "step": 4458 + }, + { + "epoch": 1.03, + "learning_rate": 4.1412448963870275e-05, + "loss": 0.0001, + "step": 4459 + }, + { + "epoch": 1.03, + "learning_rate": 4.141052307218242e-05, + "loss": 0.0, + "step": 4460 + }, + { + "epoch": 1.03, + "learning_rate": 4.140859718049457e-05, + "loss": 0.0003, + "step": 4461 + }, + { + "epoch": 1.03, + "learning_rate": 4.140667128880672e-05, + "loss": 0.0, + "step": 4462 + }, + { + "epoch": 1.03, + "learning_rate": 4.1404745397118865e-05, + "loss": 0.0, + "step": 4463 + }, + { + "epoch": 1.03, + "learning_rate": 4.140281950543102e-05, + "loss": 0.0044, + "step": 4464 + }, + { + "epoch": 1.03, + "learning_rate": 4.1400893613743166e-05, + "loss": 0.0017, + "step": 4465 + }, + { + "epoch": 1.03, + "learning_rate": 4.1398967722055314e-05, + "loss": 0.0289, + "step": 4466 + }, + { + "epoch": 1.03, + "learning_rate": 4.139704183036746e-05, + "loss": 0.0, + "step": 4467 + }, + { + "epoch": 1.03, + "learning_rate": 4.1395115938679615e-05, + "loss": 0.0, + "step": 4468 + }, + { + "epoch": 1.03, + "learning_rate": 4.1393190046991756e-05, + "loss": 0.0, + "step": 4469 + }, + { + "epoch": 1.03, + "learning_rate": 4.139126415530391e-05, + "loss": 0.0, + "step": 4470 + }, + { + "epoch": 1.03, + "learning_rate": 4.138933826361606e-05, + "loss": 0.0012, + "step": 4471 + }, + { + "epoch": 1.03, + "learning_rate": 4.1387412371928205e-05, + "loss": 0.0, + "step": 4472 + }, + { + "epoch": 1.03, + "learning_rate": 4.138548648024035e-05, + "loss": 0.0002, + "step": 4473 + }, + { + "epoch": 1.03, + "learning_rate": 4.13835605885525e-05, + "loss": 0.0005, + "step": 4474 + }, + { + "epoch": 1.03, + "learning_rate": 4.1381634696864655e-05, + "loss": 0.0, + "step": 4475 + }, + { + "epoch": 1.03, + "learning_rate": 4.1379708805176795e-05, + "loss": 0.0, + "step": 4476 + }, + { + "epoch": 1.03, + "learning_rate": 4.137778291348895e-05, + "loss": 0.0135, + "step": 4477 + }, + { + "epoch": 1.03, + "learning_rate": 4.13758570218011e-05, + "loss": 0.0004, + "step": 4478 + }, + { + "epoch": 1.04, + "learning_rate": 4.1373931130113245e-05, + "loss": 0.0001, + "step": 4479 + }, + { + "epoch": 1.04, + "learning_rate": 4.137200523842539e-05, + "loss": 0.0, + "step": 4480 + }, + { + "epoch": 1.04, + "learning_rate": 4.137007934673754e-05, + "loss": 0.0007, + "step": 4481 + }, + { + "epoch": 1.04, + "learning_rate": 4.1368153455049694e-05, + "loss": 0.0, + "step": 4482 + }, + { + "epoch": 1.04, + "learning_rate": 4.1366227563361834e-05, + "loss": 0.0006, + "step": 4483 + }, + { + "epoch": 1.04, + "learning_rate": 4.136430167167399e-05, + "loss": 0.0002, + "step": 4484 + }, + { + "epoch": 1.04, + "learning_rate": 4.1362375779986136e-05, + "loss": 0.0003, + "step": 4485 + }, + { + "epoch": 1.04, + "learning_rate": 4.1360449888298284e-05, + "loss": 0.0, + "step": 4486 + }, + { + "epoch": 1.04, + "learning_rate": 4.135852399661043e-05, + "loss": 0.0, + "step": 4487 + }, + { + "epoch": 1.04, + "learning_rate": 4.1356598104922585e-05, + "loss": 0.0002, + "step": 4488 + }, + { + "epoch": 1.04, + "learning_rate": 4.135467221323473e-05, + "loss": 0.0001, + "step": 4489 + }, + { + "epoch": 1.04, + "learning_rate": 4.1352746321546874e-05, + "loss": 0.0001, + "step": 4490 + }, + { + "epoch": 1.04, + "learning_rate": 4.135082042985903e-05, + "loss": 0.0006, + "step": 4491 + }, + { + "epoch": 1.04, + "learning_rate": 4.1348894538171175e-05, + "loss": 0.0001, + "step": 4492 + }, + { + "epoch": 1.04, + "learning_rate": 4.134696864648332e-05, + "loss": 0.0493, + "step": 4493 + }, + { + "epoch": 1.04, + "learning_rate": 4.134504275479547e-05, + "loss": 0.0003, + "step": 4494 + }, + { + "epoch": 1.04, + "learning_rate": 4.1343116863107625e-05, + "loss": 0.0006, + "step": 4495 + }, + { + "epoch": 1.04, + "learning_rate": 4.134119097141977e-05, + "loss": 0.0019, + "step": 4496 + }, + { + "epoch": 1.04, + "learning_rate": 4.133926507973192e-05, + "loss": 0.0001, + "step": 4497 + }, + { + "epoch": 1.04, + "learning_rate": 4.133733918804407e-05, + "loss": 0.0005, + "step": 4498 + }, + { + "epoch": 1.04, + "learning_rate": 4.1335413296356214e-05, + "loss": 0.0, + "step": 4499 + }, + { + "epoch": 1.04, + "learning_rate": 4.133348740466836e-05, + "loss": 0.0, + "step": 4500 + }, + { + "epoch": 1.04, + "learning_rate": 4.133156151298051e-05, + "loss": 0.0011, + "step": 4501 + }, + { + "epoch": 1.04, + "learning_rate": 4.1329635621292664e-05, + "loss": 0.0004, + "step": 4502 + }, + { + "epoch": 1.04, + "learning_rate": 4.1327709729604804e-05, + "loss": 0.0002, + "step": 4503 + }, + { + "epoch": 1.04, + "learning_rate": 4.132578383791696e-05, + "loss": 0.0, + "step": 4504 + }, + { + "epoch": 1.04, + "learning_rate": 4.1323857946229106e-05, + "loss": 0.0006, + "step": 4505 + }, + { + "epoch": 1.04, + "learning_rate": 4.132193205454126e-05, + "loss": 0.0004, + "step": 4506 + }, + { + "epoch": 1.04, + "learning_rate": 4.13200061628534e-05, + "loss": 0.0, + "step": 4507 + }, + { + "epoch": 1.04, + "learning_rate": 4.131808027116555e-05, + "loss": 0.0, + "step": 4508 + }, + { + "epoch": 1.04, + "learning_rate": 4.13161543794777e-05, + "loss": 0.0, + "step": 4509 + }, + { + "epoch": 1.04, + "learning_rate": 4.1314228487789843e-05, + "loss": 0.0, + "step": 4510 + }, + { + "epoch": 1.04, + "learning_rate": 4.1312302596102e-05, + "loss": 0.0001, + "step": 4511 + }, + { + "epoch": 1.04, + "learning_rate": 4.1310376704414145e-05, + "loss": 0.0003, + "step": 4512 + }, + { + "epoch": 1.04, + "learning_rate": 4.130845081272629e-05, + "loss": 0.0019, + "step": 4513 + }, + { + "epoch": 1.04, + "learning_rate": 4.130652492103844e-05, + "loss": 0.0002, + "step": 4514 + }, + { + "epoch": 1.04, + "learning_rate": 4.1304599029350594e-05, + "loss": 0.0, + "step": 4515 + }, + { + "epoch": 1.04, + "learning_rate": 4.130267313766274e-05, + "loss": 0.0001, + "step": 4516 + }, + { + "epoch": 1.04, + "learning_rate": 4.130074724597488e-05, + "loss": 0.0011, + "step": 4517 + }, + { + "epoch": 1.04, + "learning_rate": 4.129882135428704e-05, + "loss": 0.0012, + "step": 4518 + }, + { + "epoch": 1.04, + "learning_rate": 4.1296895462599184e-05, + "loss": 0.0, + "step": 4519 + }, + { + "epoch": 1.04, + "learning_rate": 4.129496957091133e-05, + "loss": 0.0013, + "step": 4520 + }, + { + "epoch": 1.04, + "learning_rate": 4.129304367922348e-05, + "loss": 0.014, + "step": 4521 + }, + { + "epoch": 1.05, + "learning_rate": 4.1291117787535634e-05, + "loss": 0.0091, + "step": 4522 + }, + { + "epoch": 1.05, + "learning_rate": 4.128919189584778e-05, + "loss": 0.0, + "step": 4523 + }, + { + "epoch": 1.05, + "learning_rate": 4.128726600415993e-05, + "loss": 0.0001, + "step": 4524 + }, + { + "epoch": 1.05, + "learning_rate": 4.1285340112472076e-05, + "loss": 0.0021, + "step": 4525 + }, + { + "epoch": 1.05, + "learning_rate": 4.128341422078423e-05, + "loss": 0.0002, + "step": 4526 + }, + { + "epoch": 1.05, + "learning_rate": 4.128148832909637e-05, + "loss": 0.0, + "step": 4527 + }, + { + "epoch": 1.05, + "learning_rate": 4.127956243740852e-05, + "loss": 0.0004, + "step": 4528 + }, + { + "epoch": 1.05, + "learning_rate": 4.127763654572067e-05, + "loss": 0.0326, + "step": 4529 + }, + { + "epoch": 1.05, + "learning_rate": 4.127571065403282e-05, + "loss": 0.0001, + "step": 4530 + }, + { + "epoch": 1.05, + "learning_rate": 4.127378476234497e-05, + "loss": 0.0, + "step": 4531 + }, + { + "epoch": 1.05, + "learning_rate": 4.1271858870657115e-05, + "loss": 0.0001, + "step": 4532 + }, + { + "epoch": 1.05, + "learning_rate": 4.126993297896927e-05, + "loss": 0.0, + "step": 4533 + }, + { + "epoch": 1.05, + "learning_rate": 4.126800708728141e-05, + "loss": 0.0005, + "step": 4534 + }, + { + "epoch": 1.05, + "learning_rate": 4.1266081195593564e-05, + "loss": 0.0005, + "step": 4535 + }, + { + "epoch": 1.05, + "learning_rate": 4.126415530390571e-05, + "loss": 0.0002, + "step": 4536 + }, + { + "epoch": 1.05, + "learning_rate": 4.126222941221786e-05, + "loss": 0.0, + "step": 4537 + }, + { + "epoch": 1.05, + "learning_rate": 4.126030352053001e-05, + "loss": 0.0, + "step": 4538 + }, + { + "epoch": 1.05, + "learning_rate": 4.1258377628842154e-05, + "loss": 0.0025, + "step": 4539 + }, + { + "epoch": 1.05, + "learning_rate": 4.125645173715431e-05, + "loss": 0.0037, + "step": 4540 + }, + { + "epoch": 1.05, + "learning_rate": 4.125452584546645e-05, + "loss": 0.0008, + "step": 4541 + }, + { + "epoch": 1.05, + "learning_rate": 4.1252599953778603e-05, + "loss": 0.0001, + "step": 4542 + }, + { + "epoch": 1.05, + "learning_rate": 4.125067406209075e-05, + "loss": 0.0007, + "step": 4543 + }, + { + "epoch": 1.05, + "learning_rate": 4.12487481704029e-05, + "loss": 0.0001, + "step": 4544 + }, + { + "epoch": 1.05, + "learning_rate": 4.1246822278715046e-05, + "loss": 0.0, + "step": 4545 + }, + { + "epoch": 1.05, + "learning_rate": 4.124489638702719e-05, + "loss": 0.0037, + "step": 4546 + }, + { + "epoch": 1.05, + "learning_rate": 4.124297049533935e-05, + "loss": 0.0, + "step": 4547 + }, + { + "epoch": 1.05, + "learning_rate": 4.124104460365149e-05, + "loss": 0.0004, + "step": 4548 + }, + { + "epoch": 1.05, + "learning_rate": 4.123911871196364e-05, + "loss": 0.0, + "step": 4549 + }, + { + "epoch": 1.05, + "learning_rate": 4.123719282027579e-05, + "loss": 0.001, + "step": 4550 + }, + { + "epoch": 1.05, + "learning_rate": 4.123526692858794e-05, + "loss": 0.0, + "step": 4551 + }, + { + "epoch": 1.05, + "learning_rate": 4.1233341036900085e-05, + "loss": 0.0, + "step": 4552 + }, + { + "epoch": 1.05, + "learning_rate": 4.123141514521224e-05, + "loss": 0.0, + "step": 4553 + }, + { + "epoch": 1.05, + "learning_rate": 4.122948925352439e-05, + "loss": 0.0, + "step": 4554 + }, + { + "epoch": 1.05, + "learning_rate": 4.122756336183653e-05, + "loss": 0.0002, + "step": 4555 + }, + { + "epoch": 1.05, + "learning_rate": 4.122563747014868e-05, + "loss": 0.0, + "step": 4556 + }, + { + "epoch": 1.05, + "learning_rate": 4.122371157846083e-05, + "loss": 0.0004, + "step": 4557 + }, + { + "epoch": 1.05, + "learning_rate": 4.1221785686772977e-05, + "loss": 0.0007, + "step": 4558 + }, + { + "epoch": 1.05, + "learning_rate": 4.1219859795085124e-05, + "loss": 0.0015, + "step": 4559 + }, + { + "epoch": 1.05, + "learning_rate": 4.121793390339728e-05, + "loss": 0.0001, + "step": 4560 + }, + { + "epoch": 1.05, + "learning_rate": 4.1216008011709426e-05, + "loss": 0.0082, + "step": 4561 + }, + { + "epoch": 1.05, + "learning_rate": 4.121408212002157e-05, + "loss": 0.0003, + "step": 4562 + }, + { + "epoch": 1.05, + "learning_rate": 4.121215622833372e-05, + "loss": 0.0019, + "step": 4563 + }, + { + "epoch": 1.05, + "learning_rate": 4.121023033664587e-05, + "loss": 0.0003, + "step": 4564 + }, + { + "epoch": 1.06, + "learning_rate": 4.1208304444958016e-05, + "loss": 0.0, + "step": 4565 + }, + { + "epoch": 1.06, + "learning_rate": 4.120637855327016e-05, + "loss": 0.0, + "step": 4566 + }, + { + "epoch": 1.06, + "learning_rate": 4.120445266158232e-05, + "loss": 0.0001, + "step": 4567 + }, + { + "epoch": 1.06, + "learning_rate": 4.1202526769894465e-05, + "loss": 0.0001, + "step": 4568 + }, + { + "epoch": 1.06, + "learning_rate": 4.120060087820661e-05, + "loss": 0.0002, + "step": 4569 + }, + { + "epoch": 1.06, + "learning_rate": 4.119867498651876e-05, + "loss": 0.0004, + "step": 4570 + }, + { + "epoch": 1.06, + "learning_rate": 4.1196749094830914e-05, + "loss": 0.0, + "step": 4571 + }, + { + "epoch": 1.06, + "learning_rate": 4.1194823203143055e-05, + "loss": 0.0, + "step": 4572 + }, + { + "epoch": 1.06, + "learning_rate": 4.11928973114552e-05, + "loss": 0.0001, + "step": 4573 + }, + { + "epoch": 1.06, + "learning_rate": 4.1190971419767357e-05, + "loss": 0.0012, + "step": 4574 + }, + { + "epoch": 1.06, + "learning_rate": 4.1189045528079504e-05, + "loss": 0.0003, + "step": 4575 + }, + { + "epoch": 1.06, + "learning_rate": 4.118711963639165e-05, + "loss": 0.0003, + "step": 4576 + }, + { + "epoch": 1.06, + "learning_rate": 4.11851937447038e-05, + "loss": 0.0001, + "step": 4577 + }, + { + "epoch": 1.06, + "learning_rate": 4.118326785301595e-05, + "loss": 0.0003, + "step": 4578 + }, + { + "epoch": 1.06, + "learning_rate": 4.1181341961328094e-05, + "loss": 0.0, + "step": 4579 + }, + { + "epoch": 1.06, + "learning_rate": 4.117941606964025e-05, + "loss": 0.0021, + "step": 4580 + }, + { + "epoch": 1.06, + "learning_rate": 4.1177490177952396e-05, + "loss": 0.0002, + "step": 4581 + }, + { + "epoch": 1.06, + "learning_rate": 4.117556428626454e-05, + "loss": 0.0003, + "step": 4582 + }, + { + "epoch": 1.06, + "learning_rate": 4.117363839457669e-05, + "loss": 0.012, + "step": 4583 + }, + { + "epoch": 1.06, + "learning_rate": 4.117171250288884e-05, + "loss": 0.0001, + "step": 4584 + }, + { + "epoch": 1.06, + "learning_rate": 4.1169786611200986e-05, + "loss": 0.0027, + "step": 4585 + }, + { + "epoch": 1.06, + "learning_rate": 4.116786071951313e-05, + "loss": 0.0, + "step": 4586 + }, + { + "epoch": 1.06, + "learning_rate": 4.116593482782529e-05, + "loss": 0.0001, + "step": 4587 + }, + { + "epoch": 1.06, + "learning_rate": 4.1164008936137435e-05, + "loss": 0.0004, + "step": 4588 + }, + { + "epoch": 1.06, + "learning_rate": 4.116208304444958e-05, + "loss": 0.0001, + "step": 4589 + }, + { + "epoch": 1.06, + "learning_rate": 4.116015715276173e-05, + "loss": 0.0002, + "step": 4590 + }, + { + "epoch": 1.06, + "learning_rate": 4.1158231261073884e-05, + "loss": 0.0001, + "step": 4591 + }, + { + "epoch": 1.06, + "learning_rate": 4.1156305369386025e-05, + "loss": 0.0, + "step": 4592 + }, + { + "epoch": 1.06, + "learning_rate": 4.115437947769817e-05, + "loss": 0.0001, + "step": 4593 + }, + { + "epoch": 1.06, + "learning_rate": 4.1152453586010326e-05, + "loss": 0.0, + "step": 4594 + }, + { + "epoch": 1.06, + "learning_rate": 4.1150527694322474e-05, + "loss": 0.003, + "step": 4595 + }, + { + "epoch": 1.06, + "learning_rate": 4.114860180263462e-05, + "loss": 0.0016, + "step": 4596 + }, + { + "epoch": 1.06, + "learning_rate": 4.114667591094677e-05, + "loss": 0.0006, + "step": 4597 + }, + { + "epoch": 1.06, + "learning_rate": 4.114475001925892e-05, + "loss": 0.0003, + "step": 4598 + }, + { + "epoch": 1.06, + "learning_rate": 4.1142824127571064e-05, + "loss": 0.0003, + "step": 4599 + }, + { + "epoch": 1.06, + "learning_rate": 4.114089823588322e-05, + "loss": 0.0003, + "step": 4600 + }, + { + "epoch": 1.06, + "learning_rate": 4.1138972344195366e-05, + "loss": 0.0, + "step": 4601 + }, + { + "epoch": 1.06, + "learning_rate": 4.113704645250751e-05, + "loss": 0.0002, + "step": 4602 + }, + { + "epoch": 1.06, + "learning_rate": 4.113512056081966e-05, + "loss": 0.0, + "step": 4603 + }, + { + "epoch": 1.06, + "learning_rate": 4.113319466913181e-05, + "loss": 0.0, + "step": 4604 + }, + { + "epoch": 1.06, + "learning_rate": 4.113126877744396e-05, + "loss": 0.0001, + "step": 4605 + }, + { + "epoch": 1.06, + "learning_rate": 4.11293428857561e-05, + "loss": 0.0001, + "step": 4606 + }, + { + "epoch": 1.06, + "learning_rate": 4.112741699406826e-05, + "loss": 0.0009, + "step": 4607 + }, + { + "epoch": 1.06, + "learning_rate": 4.1125491102380405e-05, + "loss": 0.0002, + "step": 4608 + }, + { + "epoch": 1.07, + "learning_rate": 4.112356521069255e-05, + "loss": 0.0045, + "step": 4609 + }, + { + "epoch": 1.07, + "learning_rate": 4.11216393190047e-05, + "loss": 0.0, + "step": 4610 + }, + { + "epoch": 1.07, + "learning_rate": 4.111971342731685e-05, + "loss": 0.0043, + "step": 4611 + }, + { + "epoch": 1.07, + "learning_rate": 4.1117787535629e-05, + "loss": 0.0, + "step": 4612 + }, + { + "epoch": 1.07, + "learning_rate": 4.111586164394114e-05, + "loss": 0.0, + "step": 4613 + }, + { + "epoch": 1.07, + "learning_rate": 4.1113935752253296e-05, + "loss": 0.0001, + "step": 4614 + }, + { + "epoch": 1.07, + "learning_rate": 4.1112009860565444e-05, + "loss": 0.0, + "step": 4615 + }, + { + "epoch": 1.07, + "learning_rate": 4.111008396887759e-05, + "loss": 0.0, + "step": 4616 + }, + { + "epoch": 1.07, + "learning_rate": 4.110815807718974e-05, + "loss": 0.0016, + "step": 4617 + }, + { + "epoch": 1.07, + "learning_rate": 4.110623218550189e-05, + "loss": 0.0001, + "step": 4618 + }, + { + "epoch": 1.07, + "learning_rate": 4.110430629381404e-05, + "loss": 0.0, + "step": 4619 + }, + { + "epoch": 1.07, + "learning_rate": 4.110238040212618e-05, + "loss": 0.0001, + "step": 4620 + }, + { + "epoch": 1.07, + "learning_rate": 4.1100454510438335e-05, + "loss": 0.0, + "step": 4621 + }, + { + "epoch": 1.07, + "learning_rate": 4.109852861875048e-05, + "loss": 0.0002, + "step": 4622 + }, + { + "epoch": 1.07, + "learning_rate": 4.109660272706263e-05, + "loss": 0.0002, + "step": 4623 + }, + { + "epoch": 1.07, + "learning_rate": 4.109467683537478e-05, + "loss": 0.0005, + "step": 4624 + }, + { + "epoch": 1.07, + "learning_rate": 4.109275094368693e-05, + "loss": 0.0008, + "step": 4625 + }, + { + "epoch": 1.07, + "learning_rate": 4.109082505199908e-05, + "loss": 0.0, + "step": 4626 + }, + { + "epoch": 1.07, + "learning_rate": 4.108889916031123e-05, + "loss": 0.0, + "step": 4627 + }, + { + "epoch": 1.07, + "learning_rate": 4.1086973268623375e-05, + "loss": 0.0, + "step": 4628 + }, + { + "epoch": 1.07, + "learning_rate": 4.108504737693552e-05, + "loss": 0.0001, + "step": 4629 + }, + { + "epoch": 1.07, + "learning_rate": 4.108312148524767e-05, + "loss": 0.0086, + "step": 4630 + }, + { + "epoch": 1.07, + "learning_rate": 4.108119559355982e-05, + "loss": 0.0, + "step": 4631 + }, + { + "epoch": 1.07, + "learning_rate": 4.107926970187197e-05, + "loss": 0.0001, + "step": 4632 + }, + { + "epoch": 1.07, + "learning_rate": 4.107734381018412e-05, + "loss": 0.0004, + "step": 4633 + }, + { + "epoch": 1.07, + "learning_rate": 4.1075417918496266e-05, + "loss": 0.0003, + "step": 4634 + }, + { + "epoch": 1.07, + "learning_rate": 4.1073492026808414e-05, + "loss": 0.0002, + "step": 4635 + }, + { + "epoch": 1.07, + "learning_rate": 4.107156613512057e-05, + "loss": 0.0003, + "step": 4636 + }, + { + "epoch": 1.07, + "learning_rate": 4.106964024343271e-05, + "loss": 0.0, + "step": 4637 + }, + { + "epoch": 1.07, + "learning_rate": 4.106771435174486e-05, + "loss": 0.0005, + "step": 4638 + }, + { + "epoch": 1.07, + "learning_rate": 4.106578846005701e-05, + "loss": 0.0, + "step": 4639 + }, + { + "epoch": 1.07, + "learning_rate": 4.106386256836916e-05, + "loss": 0.0, + "step": 4640 + }, + { + "epoch": 1.07, + "learning_rate": 4.1061936676681305e-05, + "loss": 0.0001, + "step": 4641 + }, + { + "epoch": 1.07, + "learning_rate": 4.106001078499345e-05, + "loss": 0.0018, + "step": 4642 + }, + { + "epoch": 1.07, + "learning_rate": 4.105808489330561e-05, + "loss": 0.0012, + "step": 4643 + }, + { + "epoch": 1.07, + "learning_rate": 4.105615900161775e-05, + "loss": 0.0004, + "step": 4644 + }, + { + "epoch": 1.07, + "learning_rate": 4.10542331099299e-05, + "loss": 0.0014, + "step": 4645 + }, + { + "epoch": 1.07, + "learning_rate": 4.105230721824205e-05, + "loss": 0.0014, + "step": 4646 + }, + { + "epoch": 1.07, + "learning_rate": 4.10503813265542e-05, + "loss": 0.0061, + "step": 4647 + }, + { + "epoch": 1.07, + "learning_rate": 4.1048455434866344e-05, + "loss": 0.0005, + "step": 4648 + }, + { + "epoch": 1.07, + "learning_rate": 4.104652954317849e-05, + "loss": 0.0001, + "step": 4649 + }, + { + "epoch": 1.07, + "learning_rate": 4.1044603651490646e-05, + "loss": 0.0006, + "step": 4650 + }, + { + "epoch": 1.07, + "learning_rate": 4.104267775980279e-05, + "loss": 0.0218, + "step": 4651 + }, + { + "epoch": 1.08, + "learning_rate": 4.104075186811494e-05, + "loss": 0.0, + "step": 4652 + }, + { + "epoch": 1.08, + "learning_rate": 4.103882597642709e-05, + "loss": 0.0001, + "step": 4653 + }, + { + "epoch": 1.08, + "learning_rate": 4.1036900084739236e-05, + "loss": 0.0, + "step": 4654 + }, + { + "epoch": 1.08, + "learning_rate": 4.1034974193051384e-05, + "loss": 0.0, + "step": 4655 + }, + { + "epoch": 1.08, + "learning_rate": 4.103304830136354e-05, + "loss": 0.0001, + "step": 4656 + }, + { + "epoch": 1.08, + "learning_rate": 4.103112240967568e-05, + "loss": 0.0, + "step": 4657 + }, + { + "epoch": 1.08, + "learning_rate": 4.1029196517987826e-05, + "loss": 0.0, + "step": 4658 + }, + { + "epoch": 1.08, + "learning_rate": 4.102727062629998e-05, + "loss": 0.0005, + "step": 4659 + }, + { + "epoch": 1.08, + "learning_rate": 4.102534473461213e-05, + "loss": 0.0, + "step": 4660 + }, + { + "epoch": 1.08, + "learning_rate": 4.1023418842924275e-05, + "loss": 0.0006, + "step": 4661 + }, + { + "epoch": 1.08, + "learning_rate": 4.102149295123642e-05, + "loss": 0.0, + "step": 4662 + }, + { + "epoch": 1.08, + "learning_rate": 4.101956705954858e-05, + "loss": 0.0003, + "step": 4663 + }, + { + "epoch": 1.08, + "learning_rate": 4.101764116786072e-05, + "loss": 0.0001, + "step": 4664 + }, + { + "epoch": 1.08, + "learning_rate": 4.101571527617287e-05, + "loss": 0.0001, + "step": 4665 + }, + { + "epoch": 1.08, + "learning_rate": 4.101378938448502e-05, + "loss": 0.0001, + "step": 4666 + }, + { + "epoch": 1.08, + "learning_rate": 4.101186349279717e-05, + "loss": 0.0001, + "step": 4667 + }, + { + "epoch": 1.08, + "learning_rate": 4.1009937601109314e-05, + "loss": 0.0027, + "step": 4668 + }, + { + "epoch": 1.08, + "learning_rate": 4.100801170942146e-05, + "loss": 0.0006, + "step": 4669 + }, + { + "epoch": 1.08, + "learning_rate": 4.1006085817733616e-05, + "loss": 0.0001, + "step": 4670 + }, + { + "epoch": 1.08, + "learning_rate": 4.100415992604576e-05, + "loss": 0.0194, + "step": 4671 + }, + { + "epoch": 1.08, + "learning_rate": 4.100223403435791e-05, + "loss": 0.0002, + "step": 4672 + }, + { + "epoch": 1.08, + "learning_rate": 4.100030814267006e-05, + "loss": 0.0002, + "step": 4673 + }, + { + "epoch": 1.08, + "learning_rate": 4.0998382250982206e-05, + "loss": 0.0002, + "step": 4674 + }, + { + "epoch": 1.08, + "learning_rate": 4.0996456359294353e-05, + "loss": 0.0002, + "step": 4675 + }, + { + "epoch": 1.08, + "learning_rate": 4.09945304676065e-05, + "loss": 0.0001, + "step": 4676 + }, + { + "epoch": 1.08, + "learning_rate": 4.0992604575918655e-05, + "loss": 0.0011, + "step": 4677 + }, + { + "epoch": 1.08, + "learning_rate": 4.0990678684230796e-05, + "loss": 0.001, + "step": 4678 + }, + { + "epoch": 1.08, + "learning_rate": 4.098875279254295e-05, + "loss": 0.0, + "step": 4679 + }, + { + "epoch": 1.08, + "learning_rate": 4.09868269008551e-05, + "loss": 0.0, + "step": 4680 + }, + { + "epoch": 1.08, + "learning_rate": 4.0984901009167245e-05, + "loss": 0.0004, + "step": 4681 + }, + { + "epoch": 1.08, + "learning_rate": 4.098297511747939e-05, + "loss": 0.0001, + "step": 4682 + }, + { + "epoch": 1.08, + "learning_rate": 4.098104922579155e-05, + "loss": 0.0, + "step": 4683 + }, + { + "epoch": 1.08, + "learning_rate": 4.0979123334103694e-05, + "loss": 0.0001, + "step": 4684 + }, + { + "epoch": 1.08, + "learning_rate": 4.0977197442415835e-05, + "loss": 0.0, + "step": 4685 + }, + { + "epoch": 1.08, + "learning_rate": 4.097527155072799e-05, + "loss": 0.0011, + "step": 4686 + }, + { + "epoch": 1.08, + "learning_rate": 4.097334565904014e-05, + "loss": 0.0002, + "step": 4687 + }, + { + "epoch": 1.08, + "learning_rate": 4.0971419767352284e-05, + "loss": 0.0001, + "step": 4688 + }, + { + "epoch": 1.08, + "learning_rate": 4.096949387566443e-05, + "loss": 0.0, + "step": 4689 + }, + { + "epoch": 1.08, + "learning_rate": 4.0967567983976586e-05, + "loss": 0.0001, + "step": 4690 + }, + { + "epoch": 1.08, + "learning_rate": 4.0965642092288733e-05, + "loss": 0.0, + "step": 4691 + }, + { + "epoch": 1.08, + "learning_rate": 4.096371620060088e-05, + "loss": 0.0227, + "step": 4692 + }, + { + "epoch": 1.08, + "learning_rate": 4.096179030891303e-05, + "loss": 0.0002, + "step": 4693 + }, + { + "epoch": 1.08, + "learning_rate": 4.095986441722518e-05, + "loss": 0.0003, + "step": 4694 + }, + { + "epoch": 1.09, + "learning_rate": 4.095793852553732e-05, + "loss": 0.0, + "step": 4695 + }, + { + "epoch": 1.09, + "learning_rate": 4.095601263384947e-05, + "loss": 0.0005, + "step": 4696 + }, + { + "epoch": 1.09, + "learning_rate": 4.0954086742161625e-05, + "loss": 0.0002, + "step": 4697 + }, + { + "epoch": 1.09, + "learning_rate": 4.095216085047377e-05, + "loss": 0.0, + "step": 4698 + }, + { + "epoch": 1.09, + "learning_rate": 4.095023495878592e-05, + "loss": 0.0007, + "step": 4699 + }, + { + "epoch": 1.09, + "learning_rate": 4.094830906709807e-05, + "loss": 0.0001, + "step": 4700 + }, + { + "epoch": 1.09, + "learning_rate": 4.094638317541022e-05, + "loss": 0.0004, + "step": 4701 + }, + { + "epoch": 1.09, + "learning_rate": 4.094445728372236e-05, + "loss": 0.0006, + "step": 4702 + }, + { + "epoch": 1.09, + "learning_rate": 4.094253139203452e-05, + "loss": 0.0002, + "step": 4703 + }, + { + "epoch": 1.09, + "learning_rate": 4.0940605500346664e-05, + "loss": 0.0006, + "step": 4704 + }, + { + "epoch": 1.09, + "learning_rate": 4.093867960865881e-05, + "loss": 0.0, + "step": 4705 + }, + { + "epoch": 1.09, + "learning_rate": 4.093675371697096e-05, + "loss": 0.0005, + "step": 4706 + }, + { + "epoch": 1.09, + "learning_rate": 4.093482782528311e-05, + "loss": 0.0, + "step": 4707 + }, + { + "epoch": 1.09, + "learning_rate": 4.093290193359526e-05, + "loss": 0.0459, + "step": 4708 + }, + { + "epoch": 1.09, + "learning_rate": 4.09309760419074e-05, + "loss": 0.0, + "step": 4709 + }, + { + "epoch": 1.09, + "learning_rate": 4.0929050150219556e-05, + "loss": 0.0002, + "step": 4710 + }, + { + "epoch": 1.09, + "learning_rate": 4.09271242585317e-05, + "loss": 0.0, + "step": 4711 + }, + { + "epoch": 1.09, + "learning_rate": 4.092519836684385e-05, + "loss": 0.0, + "step": 4712 + }, + { + "epoch": 1.09, + "learning_rate": 4.0923272475156e-05, + "loss": 0.0012, + "step": 4713 + }, + { + "epoch": 1.09, + "learning_rate": 4.0921346583468146e-05, + "loss": 0.0002, + "step": 4714 + }, + { + "epoch": 1.09, + "learning_rate": 4.09194206917803e-05, + "loss": 0.0003, + "step": 4715 + }, + { + "epoch": 1.09, + "learning_rate": 4.091749480009244e-05, + "loss": 0.0002, + "step": 4716 + }, + { + "epoch": 1.09, + "learning_rate": 4.0915568908404595e-05, + "loss": 0.0, + "step": 4717 + }, + { + "epoch": 1.09, + "learning_rate": 4.091364301671674e-05, + "loss": 0.0, + "step": 4718 + }, + { + "epoch": 1.09, + "learning_rate": 4.091171712502889e-05, + "loss": 0.0002, + "step": 4719 + }, + { + "epoch": 1.09, + "learning_rate": 4.090979123334104e-05, + "loss": 0.0029, + "step": 4720 + }, + { + "epoch": 1.09, + "learning_rate": 4.090786534165319e-05, + "loss": 0.0, + "step": 4721 + }, + { + "epoch": 1.09, + "learning_rate": 4.090593944996534e-05, + "loss": 0.0047, + "step": 4722 + }, + { + "epoch": 1.09, + "learning_rate": 4.090401355827748e-05, + "loss": 0.0, + "step": 4723 + }, + { + "epoch": 1.09, + "learning_rate": 4.0902087666589634e-05, + "loss": 0.0, + "step": 4724 + }, + { + "epoch": 1.09, + "learning_rate": 4.090016177490178e-05, + "loss": 0.0009, + "step": 4725 + }, + { + "epoch": 1.09, + "learning_rate": 4.089823588321393e-05, + "loss": 0.0001, + "step": 4726 + }, + { + "epoch": 1.09, + "learning_rate": 4.0896309991526077e-05, + "loss": 0.0002, + "step": 4727 + }, + { + "epoch": 1.09, + "learning_rate": 4.089438409983823e-05, + "loss": 0.0002, + "step": 4728 + }, + { + "epoch": 1.09, + "learning_rate": 4.089245820815038e-05, + "loss": 0.0001, + "step": 4729 + }, + { + "epoch": 1.09, + "learning_rate": 4.0890532316462526e-05, + "loss": 0.0002, + "step": 4730 + }, + { + "epoch": 1.09, + "learning_rate": 4.088860642477467e-05, + "loss": 0.0, + "step": 4731 + }, + { + "epoch": 1.09, + "learning_rate": 4.088668053308682e-05, + "loss": 0.0, + "step": 4732 + }, + { + "epoch": 1.09, + "learning_rate": 4.088475464139897e-05, + "loss": 0.0001, + "step": 4733 + }, + { + "epoch": 1.09, + "learning_rate": 4.0882828749711116e-05, + "loss": 0.0, + "step": 4734 + }, + { + "epoch": 1.09, + "learning_rate": 4.088090285802327e-05, + "loss": 0.0, + "step": 4735 + }, + { + "epoch": 1.09, + "learning_rate": 4.087897696633541e-05, + "loss": 0.0001, + "step": 4736 + }, + { + "epoch": 1.09, + "learning_rate": 4.0877051074647565e-05, + "loss": 0.0025, + "step": 4737 + }, + { + "epoch": 1.09, + "learning_rate": 4.087512518295971e-05, + "loss": 0.0001, + "step": 4738 + }, + { + "epoch": 1.1, + "learning_rate": 4.087319929127186e-05, + "loss": 0.0001, + "step": 4739 + }, + { + "epoch": 1.1, + "learning_rate": 4.087127339958401e-05, + "loss": 0.0, + "step": 4740 + }, + { + "epoch": 1.1, + "learning_rate": 4.0869347507896155e-05, + "loss": 0.0, + "step": 4741 + }, + { + "epoch": 1.1, + "learning_rate": 4.086742161620831e-05, + "loss": 0.0, + "step": 4742 + }, + { + "epoch": 1.1, + "learning_rate": 4.086549572452045e-05, + "loss": 0.0, + "step": 4743 + }, + { + "epoch": 1.1, + "learning_rate": 4.0863569832832604e-05, + "loss": 0.0, + "step": 4744 + }, + { + "epoch": 1.1, + "learning_rate": 4.086164394114475e-05, + "loss": 0.0001, + "step": 4745 + }, + { + "epoch": 1.1, + "learning_rate": 4.08597180494569e-05, + "loss": 0.0002, + "step": 4746 + }, + { + "epoch": 1.1, + "learning_rate": 4.0857792157769046e-05, + "loss": 0.0001, + "step": 4747 + }, + { + "epoch": 1.1, + "learning_rate": 4.08558662660812e-05, + "loss": 0.0003, + "step": 4748 + }, + { + "epoch": 1.1, + "learning_rate": 4.085394037439335e-05, + "loss": 0.0, + "step": 4749 + }, + { + "epoch": 1.1, + "learning_rate": 4.0852014482705496e-05, + "loss": 0.0028, + "step": 4750 + }, + { + "epoch": 1.1, + "learning_rate": 4.085008859101764e-05, + "loss": 0.0002, + "step": 4751 + }, + { + "epoch": 1.1, + "learning_rate": 4.084816269932979e-05, + "loss": 0.0, + "step": 4752 + }, + { + "epoch": 1.1, + "learning_rate": 4.084623680764194e-05, + "loss": 0.0024, + "step": 4753 + }, + { + "epoch": 1.1, + "learning_rate": 4.0844310915954086e-05, + "loss": 0.0001, + "step": 4754 + }, + { + "epoch": 1.1, + "learning_rate": 4.084238502426624e-05, + "loss": 0.0, + "step": 4755 + }, + { + "epoch": 1.1, + "learning_rate": 4.084045913257839e-05, + "loss": 0.0002, + "step": 4756 + }, + { + "epoch": 1.1, + "learning_rate": 4.0838533240890535e-05, + "loss": 0.0026, + "step": 4757 + }, + { + "epoch": 1.1, + "learning_rate": 4.083660734920268e-05, + "loss": 0.0037, + "step": 4758 + }, + { + "epoch": 1.1, + "learning_rate": 4.0834681457514836e-05, + "loss": 0.0001, + "step": 4759 + }, + { + "epoch": 1.1, + "learning_rate": 4.083275556582698e-05, + "loss": 0.0, + "step": 4760 + }, + { + "epoch": 1.1, + "learning_rate": 4.0830829674139125e-05, + "loss": 0.0, + "step": 4761 + }, + { + "epoch": 1.1, + "learning_rate": 4.082890378245128e-05, + "loss": 0.0, + "step": 4762 + }, + { + "epoch": 1.1, + "learning_rate": 4.0826977890763426e-05, + "loss": 0.0001, + "step": 4763 + }, + { + "epoch": 1.1, + "learning_rate": 4.0825051999075574e-05, + "loss": 0.0005, + "step": 4764 + }, + { + "epoch": 1.1, + "learning_rate": 4.082312610738772e-05, + "loss": 0.0001, + "step": 4765 + }, + { + "epoch": 1.1, + "learning_rate": 4.0821200215699876e-05, + "loss": 0.0, + "step": 4766 + }, + { + "epoch": 1.1, + "learning_rate": 4.0819274324012016e-05, + "loss": 0.0001, + "step": 4767 + }, + { + "epoch": 1.1, + "learning_rate": 4.081734843232417e-05, + "loss": 0.0001, + "step": 4768 + }, + { + "epoch": 1.1, + "learning_rate": 4.081542254063632e-05, + "loss": 0.0008, + "step": 4769 + }, + { + "epoch": 1.1, + "learning_rate": 4.0813496648948465e-05, + "loss": 0.0148, + "step": 4770 + }, + { + "epoch": 1.1, + "learning_rate": 4.081157075726061e-05, + "loss": 0.0027, + "step": 4771 + }, + { + "epoch": 1.1, + "learning_rate": 4.080964486557276e-05, + "loss": 0.0227, + "step": 4772 + }, + { + "epoch": 1.1, + "learning_rate": 4.0807718973884915e-05, + "loss": 0.0003, + "step": 4773 + }, + { + "epoch": 1.1, + "learning_rate": 4.0805793082197055e-05, + "loss": 0.0, + "step": 4774 + }, + { + "epoch": 1.1, + "learning_rate": 4.080386719050921e-05, + "loss": 0.009, + "step": 4775 + }, + { + "epoch": 1.1, + "learning_rate": 4.080194129882136e-05, + "loss": 0.0027, + "step": 4776 + }, + { + "epoch": 1.1, + "learning_rate": 4.0800015407133505e-05, + "loss": 0.0002, + "step": 4777 + }, + { + "epoch": 1.1, + "learning_rate": 4.079808951544565e-05, + "loss": 0.001, + "step": 4778 + }, + { + "epoch": 1.1, + "learning_rate": 4.07961636237578e-05, + "loss": 0.0, + "step": 4779 + }, + { + "epoch": 1.1, + "learning_rate": 4.0794237732069954e-05, + "loss": 0.0, + "step": 4780 + }, + { + "epoch": 1.1, + "learning_rate": 4.0792311840382095e-05, + "loss": 0.0, + "step": 4781 + }, + { + "epoch": 1.11, + "learning_rate": 4.079038594869425e-05, + "loss": 0.0002, + "step": 4782 + }, + { + "epoch": 1.11, + "learning_rate": 4.0788460057006396e-05, + "loss": 0.0109, + "step": 4783 + }, + { + "epoch": 1.11, + "learning_rate": 4.0786534165318544e-05, + "loss": 0.0021, + "step": 4784 + }, + { + "epoch": 1.11, + "learning_rate": 4.078460827363069e-05, + "loss": 0.0002, + "step": 4785 + }, + { + "epoch": 1.11, + "learning_rate": 4.0782682381942845e-05, + "loss": 0.0, + "step": 4786 + }, + { + "epoch": 1.11, + "learning_rate": 4.078075649025499e-05, + "loss": 0.0007, + "step": 4787 + }, + { + "epoch": 1.11, + "learning_rate": 4.0778830598567134e-05, + "loss": 0.0001, + "step": 4788 + }, + { + "epoch": 1.11, + "learning_rate": 4.077690470687929e-05, + "loss": 0.0111, + "step": 4789 + }, + { + "epoch": 1.11, + "learning_rate": 4.0774978815191435e-05, + "loss": 0.0001, + "step": 4790 + }, + { + "epoch": 1.11, + "learning_rate": 4.077305292350358e-05, + "loss": 0.0, + "step": 4791 + }, + { + "epoch": 1.11, + "learning_rate": 4.077112703181573e-05, + "loss": 0.0001, + "step": 4792 + }, + { + "epoch": 1.11, + "learning_rate": 4.0769201140127885e-05, + "loss": 0.0022, + "step": 4793 + }, + { + "epoch": 1.11, + "learning_rate": 4.076727524844003e-05, + "loss": 0.0002, + "step": 4794 + }, + { + "epoch": 1.11, + "learning_rate": 4.076534935675218e-05, + "loss": 0.0001, + "step": 4795 + }, + { + "epoch": 1.11, + "learning_rate": 4.076342346506433e-05, + "loss": 0.0012, + "step": 4796 + }, + { + "epoch": 1.11, + "learning_rate": 4.0761497573376474e-05, + "loss": 0.0, + "step": 4797 + }, + { + "epoch": 1.11, + "learning_rate": 4.075957168168862e-05, + "loss": 0.0001, + "step": 4798 + }, + { + "epoch": 1.11, + "learning_rate": 4.075764579000077e-05, + "loss": 0.002, + "step": 4799 + }, + { + "epoch": 1.11, + "learning_rate": 4.0755719898312924e-05, + "loss": 0.0, + "step": 4800 + }, + { + "epoch": 1.11, + "learning_rate": 4.075379400662507e-05, + "loss": 0.0005, + "step": 4801 + }, + { + "epoch": 1.11, + "learning_rate": 4.075186811493722e-05, + "loss": 0.0002, + "step": 4802 + }, + { + "epoch": 1.11, + "learning_rate": 4.0749942223249366e-05, + "loss": 0.0001, + "step": 4803 + }, + { + "epoch": 1.11, + "learning_rate": 4.074801633156152e-05, + "loss": 0.0049, + "step": 4804 + }, + { + "epoch": 1.11, + "learning_rate": 4.074609043987366e-05, + "loss": 0.0, + "step": 4805 + }, + { + "epoch": 1.11, + "learning_rate": 4.0744164548185815e-05, + "loss": 0.0001, + "step": 4806 + }, + { + "epoch": 1.11, + "learning_rate": 4.074223865649796e-05, + "loss": 0.0, + "step": 4807 + }, + { + "epoch": 1.11, + "learning_rate": 4.0740312764810104e-05, + "loss": 0.0002, + "step": 4808 + }, + { + "epoch": 1.11, + "learning_rate": 4.073838687312226e-05, + "loss": 0.0, + "step": 4809 + }, + { + "epoch": 1.11, + "learning_rate": 4.0736460981434405e-05, + "loss": 0.0001, + "step": 4810 + }, + { + "epoch": 1.11, + "learning_rate": 4.073453508974655e-05, + "loss": 0.0, + "step": 4811 + }, + { + "epoch": 1.11, + "learning_rate": 4.07326091980587e-05, + "loss": 0.0, + "step": 4812 + }, + { + "epoch": 1.11, + "learning_rate": 4.0730683306370854e-05, + "loss": 0.0085, + "step": 4813 + }, + { + "epoch": 1.11, + "learning_rate": 4.0728757414683e-05, + "loss": 0.0003, + "step": 4814 + }, + { + "epoch": 1.11, + "learning_rate": 4.072683152299515e-05, + "loss": 0.0001, + "step": 4815 + }, + { + "epoch": 1.11, + "learning_rate": 4.07249056313073e-05, + "loss": 0.0001, + "step": 4816 + }, + { + "epoch": 1.11, + "learning_rate": 4.0722979739619444e-05, + "loss": 0.0, + "step": 4817 + }, + { + "epoch": 1.11, + "learning_rate": 4.072105384793159e-05, + "loss": 0.0, + "step": 4818 + }, + { + "epoch": 1.11, + "learning_rate": 4.071912795624374e-05, + "loss": 0.0, + "step": 4819 + }, + { + "epoch": 1.11, + "learning_rate": 4.0717202064555894e-05, + "loss": 0.0, + "step": 4820 + }, + { + "epoch": 1.11, + "learning_rate": 4.071527617286804e-05, + "loss": 0.0039, + "step": 4821 + }, + { + "epoch": 1.11, + "learning_rate": 4.071335028118019e-05, + "loss": 0.0001, + "step": 4822 + }, + { + "epoch": 1.11, + "learning_rate": 4.0711424389492336e-05, + "loss": 0.0001, + "step": 4823 + }, + { + "epoch": 1.11, + "learning_rate": 4.070949849780449e-05, + "loss": 0.0, + "step": 4824 + }, + { + "epoch": 1.12, + "learning_rate": 4.070757260611663e-05, + "loss": 0.0, + "step": 4825 + }, + { + "epoch": 1.12, + "learning_rate": 4.070564671442878e-05, + "loss": 0.0158, + "step": 4826 + }, + { + "epoch": 1.12, + "learning_rate": 4.070372082274093e-05, + "loss": 0.0, + "step": 4827 + }, + { + "epoch": 1.12, + "learning_rate": 4.070179493105308e-05, + "loss": 0.0001, + "step": 4828 + }, + { + "epoch": 1.12, + "learning_rate": 4.069986903936523e-05, + "loss": 0.0, + "step": 4829 + }, + { + "epoch": 1.12, + "learning_rate": 4.0697943147677375e-05, + "loss": 0.0004, + "step": 4830 + }, + { + "epoch": 1.12, + "learning_rate": 4.069601725598953e-05, + "loss": 0.0022, + "step": 4831 + }, + { + "epoch": 1.12, + "learning_rate": 4.069409136430167e-05, + "loss": 0.0014, + "step": 4832 + }, + { + "epoch": 1.12, + "learning_rate": 4.0692165472613824e-05, + "loss": 0.0005, + "step": 4833 + }, + { + "epoch": 1.12, + "learning_rate": 4.069023958092597e-05, + "loss": 0.0017, + "step": 4834 + }, + { + "epoch": 1.12, + "learning_rate": 4.068831368923812e-05, + "loss": 0.0001, + "step": 4835 + }, + { + "epoch": 1.12, + "learning_rate": 4.068638779755027e-05, + "loss": 0.0003, + "step": 4836 + }, + { + "epoch": 1.12, + "learning_rate": 4.0684461905862414e-05, + "loss": 0.0, + "step": 4837 + }, + { + "epoch": 1.12, + "learning_rate": 4.068253601417457e-05, + "loss": 0.004, + "step": 4838 + }, + { + "epoch": 1.12, + "learning_rate": 4.068061012248671e-05, + "loss": 0.0001, + "step": 4839 + }, + { + "epoch": 1.12, + "learning_rate": 4.0678684230798863e-05, + "loss": 0.001, + "step": 4840 + }, + { + "epoch": 1.12, + "learning_rate": 4.067675833911101e-05, + "loss": 0.0101, + "step": 4841 + }, + { + "epoch": 1.12, + "learning_rate": 4.067483244742316e-05, + "loss": 0.0003, + "step": 4842 + }, + { + "epoch": 1.12, + "learning_rate": 4.0672906555735306e-05, + "loss": 0.0004, + "step": 4843 + }, + { + "epoch": 1.12, + "learning_rate": 4.067098066404745e-05, + "loss": 0.0028, + "step": 4844 + }, + { + "epoch": 1.12, + "learning_rate": 4.066905477235961e-05, + "loss": 0.0, + "step": 4845 + }, + { + "epoch": 1.12, + "learning_rate": 4.066712888067175e-05, + "loss": 0.0002, + "step": 4846 + }, + { + "epoch": 1.12, + "learning_rate": 4.06652029889839e-05, + "loss": 0.0004, + "step": 4847 + }, + { + "epoch": 1.12, + "learning_rate": 4.066327709729605e-05, + "loss": 0.001, + "step": 4848 + }, + { + "epoch": 1.12, + "learning_rate": 4.06613512056082e-05, + "loss": 0.0, + "step": 4849 + }, + { + "epoch": 1.12, + "learning_rate": 4.0659425313920345e-05, + "loss": 0.0008, + "step": 4850 + }, + { + "epoch": 1.12, + "learning_rate": 4.06574994222325e-05, + "loss": 0.0, + "step": 4851 + }, + { + "epoch": 1.12, + "learning_rate": 4.065557353054465e-05, + "loss": 0.0006, + "step": 4852 + }, + { + "epoch": 1.12, + "learning_rate": 4.065364763885679e-05, + "loss": 0.0005, + "step": 4853 + }, + { + "epoch": 1.12, + "learning_rate": 4.065172174716894e-05, + "loss": 0.0021, + "step": 4854 + }, + { + "epoch": 1.12, + "learning_rate": 4.064979585548109e-05, + "loss": 0.0003, + "step": 4855 + }, + { + "epoch": 1.12, + "learning_rate": 4.064786996379324e-05, + "loss": 0.0, + "step": 4856 + }, + { + "epoch": 1.12, + "learning_rate": 4.0645944072105384e-05, + "loss": 0.0, + "step": 4857 + }, + { + "epoch": 1.12, + "learning_rate": 4.064401818041754e-05, + "loss": 0.0, + "step": 4858 + }, + { + "epoch": 1.12, + "learning_rate": 4.0642092288729686e-05, + "loss": 0.0001, + "step": 4859 + }, + { + "epoch": 1.12, + "learning_rate": 4.064016639704183e-05, + "loss": 0.0001, + "step": 4860 + }, + { + "epoch": 1.12, + "learning_rate": 4.063824050535398e-05, + "loss": 0.0, + "step": 4861 + }, + { + "epoch": 1.12, + "learning_rate": 4.0636314613666135e-05, + "loss": 0.0, + "step": 4862 + }, + { + "epoch": 1.12, + "learning_rate": 4.0634388721978276e-05, + "loss": 0.0001, + "step": 4863 + }, + { + "epoch": 1.12, + "learning_rate": 4.063246283029042e-05, + "loss": 0.0004, + "step": 4864 + }, + { + "epoch": 1.12, + "learning_rate": 4.063053693860258e-05, + "loss": 0.0, + "step": 4865 + }, + { + "epoch": 1.12, + "learning_rate": 4.0628611046914725e-05, + "loss": 0.0005, + "step": 4866 + }, + { + "epoch": 1.12, + "learning_rate": 4.062668515522687e-05, + "loss": 0.0001, + "step": 4867 + }, + { + "epoch": 1.13, + "learning_rate": 4.062475926353902e-05, + "loss": 0.0004, + "step": 4868 + }, + { + "epoch": 1.13, + "learning_rate": 4.0622833371851174e-05, + "loss": 0.0, + "step": 4869 + }, + { + "epoch": 1.13, + "learning_rate": 4.0620907480163315e-05, + "loss": 0.0, + "step": 4870 + }, + { + "epoch": 1.13, + "learning_rate": 4.061898158847547e-05, + "loss": 0.001, + "step": 4871 + }, + { + "epoch": 1.13, + "learning_rate": 4.061705569678762e-05, + "loss": 0.0004, + "step": 4872 + }, + { + "epoch": 1.13, + "learning_rate": 4.0615129805099764e-05, + "loss": 0.0001, + "step": 4873 + }, + { + "epoch": 1.13, + "learning_rate": 4.061320391341191e-05, + "loss": 0.0008, + "step": 4874 + }, + { + "epoch": 1.13, + "learning_rate": 4.061127802172406e-05, + "loss": 0.0, + "step": 4875 + }, + { + "epoch": 1.13, + "learning_rate": 4.060935213003621e-05, + "loss": 0.0001, + "step": 4876 + }, + { + "epoch": 1.13, + "learning_rate": 4.0607426238348354e-05, + "loss": 0.0001, + "step": 4877 + }, + { + "epoch": 1.13, + "learning_rate": 4.060550034666051e-05, + "loss": 0.0076, + "step": 4878 + }, + { + "epoch": 1.13, + "learning_rate": 4.0603574454972656e-05, + "loss": 0.0028, + "step": 4879 + }, + { + "epoch": 1.13, + "learning_rate": 4.06016485632848e-05, + "loss": 0.0, + "step": 4880 + }, + { + "epoch": 1.13, + "learning_rate": 4.059972267159695e-05, + "loss": 0.0001, + "step": 4881 + }, + { + "epoch": 1.13, + "learning_rate": 4.05977967799091e-05, + "loss": 0.0, + "step": 4882 + }, + { + "epoch": 1.13, + "learning_rate": 4.0595870888221246e-05, + "loss": 0.0004, + "step": 4883 + }, + { + "epoch": 1.13, + "learning_rate": 4.059394499653339e-05, + "loss": 0.0, + "step": 4884 + }, + { + "epoch": 1.13, + "learning_rate": 4.059201910484555e-05, + "loss": 0.0003, + "step": 4885 + }, + { + "epoch": 1.13, + "learning_rate": 4.0590093213157695e-05, + "loss": 0.0001, + "step": 4886 + }, + { + "epoch": 1.13, + "learning_rate": 4.058816732146984e-05, + "loss": 0.0, + "step": 4887 + }, + { + "epoch": 1.13, + "learning_rate": 4.058624142978199e-05, + "loss": 0.0, + "step": 4888 + }, + { + "epoch": 1.13, + "learning_rate": 4.0584315538094144e-05, + "loss": 0.0005, + "step": 4889 + }, + { + "epoch": 1.13, + "learning_rate": 4.0582389646406285e-05, + "loss": 0.0001, + "step": 4890 + }, + { + "epoch": 1.13, + "learning_rate": 4.058046375471843e-05, + "loss": 0.0001, + "step": 4891 + }, + { + "epoch": 1.13, + "learning_rate": 4.0578537863030587e-05, + "loss": 0.0, + "step": 4892 + }, + { + "epoch": 1.13, + "learning_rate": 4.0576611971342734e-05, + "loss": 0.0, + "step": 4893 + }, + { + "epoch": 1.13, + "learning_rate": 4.057468607965488e-05, + "loss": 0.0025, + "step": 4894 + }, + { + "epoch": 1.13, + "learning_rate": 4.057276018796703e-05, + "loss": 0.0001, + "step": 4895 + }, + { + "epoch": 1.13, + "learning_rate": 4.057083429627918e-05, + "loss": 0.0, + "step": 4896 + }, + { + "epoch": 1.13, + "learning_rate": 4.0568908404591324e-05, + "loss": 0.0019, + "step": 4897 + }, + { + "epoch": 1.13, + "learning_rate": 4.056698251290348e-05, + "loss": 0.0005, + "step": 4898 + }, + { + "epoch": 1.13, + "learning_rate": 4.0565056621215626e-05, + "loss": 0.0002, + "step": 4899 + }, + { + "epoch": 1.13, + "learning_rate": 4.056313072952777e-05, + "loss": 0.0, + "step": 4900 + }, + { + "epoch": 1.13, + "learning_rate": 4.056120483783992e-05, + "loss": 0.0, + "step": 4901 + }, + { + "epoch": 1.13, + "learning_rate": 4.055927894615207e-05, + "loss": 0.0002, + "step": 4902 + }, + { + "epoch": 1.13, + "learning_rate": 4.055735305446422e-05, + "loss": 0.0001, + "step": 4903 + }, + { + "epoch": 1.13, + "learning_rate": 4.055542716277636e-05, + "loss": 0.0005, + "step": 4904 + }, + { + "epoch": 1.13, + "learning_rate": 4.055350127108852e-05, + "loss": 0.0443, + "step": 4905 + }, + { + "epoch": 1.13, + "learning_rate": 4.0551575379400665e-05, + "loss": 0.0, + "step": 4906 + }, + { + "epoch": 1.13, + "learning_rate": 4.054964948771281e-05, + "loss": 0.0002, + "step": 4907 + }, + { + "epoch": 1.13, + "learning_rate": 4.054772359602496e-05, + "loss": 0.0, + "step": 4908 + }, + { + "epoch": 1.13, + "learning_rate": 4.054579770433711e-05, + "loss": 0.0001, + "step": 4909 + }, + { + "epoch": 1.13, + "learning_rate": 4.054387181264926e-05, + "loss": 0.0002, + "step": 4910 + }, + { + "epoch": 1.13, + "learning_rate": 4.05419459209614e-05, + "loss": 0.0033, + "step": 4911 + }, + { + "epoch": 1.14, + "learning_rate": 4.0540020029273556e-05, + "loss": 0.0002, + "step": 4912 + }, + { + "epoch": 1.14, + "learning_rate": 4.0538094137585704e-05, + "loss": 0.0008, + "step": 4913 + }, + { + "epoch": 1.14, + "learning_rate": 4.053616824589785e-05, + "loss": 0.0001, + "step": 4914 + }, + { + "epoch": 1.14, + "learning_rate": 4.053424235421e-05, + "loss": 0.0001, + "step": 4915 + }, + { + "epoch": 1.14, + "learning_rate": 4.053231646252215e-05, + "loss": 0.0, + "step": 4916 + }, + { + "epoch": 1.14, + "learning_rate": 4.05303905708343e-05, + "loss": 0.0001, + "step": 4917 + }, + { + "epoch": 1.14, + "learning_rate": 4.052846467914645e-05, + "loss": 0.0, + "step": 4918 + }, + { + "epoch": 1.14, + "learning_rate": 4.0526538787458596e-05, + "loss": 0.0001, + "step": 4919 + }, + { + "epoch": 1.14, + "learning_rate": 4.052461289577074e-05, + "loss": 0.0, + "step": 4920 + }, + { + "epoch": 1.14, + "learning_rate": 4.052268700408289e-05, + "loss": 0.0, + "step": 4921 + }, + { + "epoch": 1.14, + "learning_rate": 4.052076111239504e-05, + "loss": 0.0, + "step": 4922 + }, + { + "epoch": 1.14, + "learning_rate": 4.051883522070719e-05, + "loss": 0.0004, + "step": 4923 + }, + { + "epoch": 1.14, + "learning_rate": 4.051690932901934e-05, + "loss": 0.0, + "step": 4924 + }, + { + "epoch": 1.14, + "learning_rate": 4.051498343733149e-05, + "loss": 0.0, + "step": 4925 + }, + { + "epoch": 1.14, + "learning_rate": 4.0513057545643635e-05, + "loss": 0.0061, + "step": 4926 + }, + { + "epoch": 1.14, + "learning_rate": 4.051113165395579e-05, + "loss": 0.0001, + "step": 4927 + }, + { + "epoch": 1.14, + "learning_rate": 4.050920576226793e-05, + "loss": 0.0001, + "step": 4928 + }, + { + "epoch": 1.14, + "learning_rate": 4.050727987058008e-05, + "loss": 0.0, + "step": 4929 + }, + { + "epoch": 1.14, + "learning_rate": 4.050535397889223e-05, + "loss": 0.0001, + "step": 4930 + }, + { + "epoch": 1.14, + "learning_rate": 4.050342808720438e-05, + "loss": 0.0001, + "step": 4931 + }, + { + "epoch": 1.14, + "learning_rate": 4.0501502195516526e-05, + "loss": 0.0, + "step": 4932 + }, + { + "epoch": 1.14, + "learning_rate": 4.0499576303828674e-05, + "loss": 0.0, + "step": 4933 + }, + { + "epoch": 1.14, + "learning_rate": 4.049765041214083e-05, + "loss": 0.0, + "step": 4934 + }, + { + "epoch": 1.14, + "learning_rate": 4.049572452045297e-05, + "loss": 0.0001, + "step": 4935 + }, + { + "epoch": 1.14, + "learning_rate": 4.049379862876512e-05, + "loss": 0.0004, + "step": 4936 + }, + { + "epoch": 1.14, + "learning_rate": 4.049187273707727e-05, + "loss": 0.0, + "step": 4937 + }, + { + "epoch": 1.14, + "learning_rate": 4.048994684538942e-05, + "loss": 0.0, + "step": 4938 + }, + { + "epoch": 1.14, + "learning_rate": 4.0488020953701565e-05, + "loss": 0.0, + "step": 4939 + }, + { + "epoch": 1.14, + "learning_rate": 4.048609506201371e-05, + "loss": 0.0002, + "step": 4940 + }, + { + "epoch": 1.14, + "learning_rate": 4.048416917032587e-05, + "loss": 0.0001, + "step": 4941 + }, + { + "epoch": 1.14, + "learning_rate": 4.048224327863801e-05, + "loss": 0.0, + "step": 4942 + }, + { + "epoch": 1.14, + "learning_rate": 4.048031738695016e-05, + "loss": 0.0, + "step": 4943 + }, + { + "epoch": 1.14, + "learning_rate": 4.047839149526231e-05, + "loss": 0.0, + "step": 4944 + }, + { + "epoch": 1.14, + "learning_rate": 4.047646560357446e-05, + "loss": 0.0001, + "step": 4945 + }, + { + "epoch": 1.14, + "learning_rate": 4.0474539711886605e-05, + "loss": 0.0004, + "step": 4946 + }, + { + "epoch": 1.14, + "learning_rate": 4.047261382019875e-05, + "loss": 0.0002, + "step": 4947 + }, + { + "epoch": 1.14, + "learning_rate": 4.0470687928510906e-05, + "loss": 0.0, + "step": 4948 + }, + { + "epoch": 1.14, + "learning_rate": 4.046876203682305e-05, + "loss": 0.0001, + "step": 4949 + }, + { + "epoch": 1.14, + "learning_rate": 4.04668361451352e-05, + "loss": 0.0, + "step": 4950 + }, + { + "epoch": 1.14, + "learning_rate": 4.046491025344735e-05, + "loss": 0.0, + "step": 4951 + }, + { + "epoch": 1.14, + "learning_rate": 4.0462984361759496e-05, + "loss": 0.0, + "step": 4952 + }, + { + "epoch": 1.14, + "learning_rate": 4.0461058470071644e-05, + "loss": 0.0001, + "step": 4953 + }, + { + "epoch": 1.14, + "learning_rate": 4.04591325783838e-05, + "loss": 0.0001, + "step": 4954 + }, + { + "epoch": 1.15, + "learning_rate": 4.0457206686695945e-05, + "loss": 0.0008, + "step": 4955 + }, + { + "epoch": 1.15, + "learning_rate": 4.0455280795008086e-05, + "loss": 0.0, + "step": 4956 + }, + { + "epoch": 1.15, + "learning_rate": 4.045335490332024e-05, + "loss": 0.0, + "step": 4957 + }, + { + "epoch": 1.15, + "learning_rate": 4.045142901163239e-05, + "loss": 0.0002, + "step": 4958 + }, + { + "epoch": 1.15, + "learning_rate": 4.0449503119944535e-05, + "loss": 0.0, + "step": 4959 + }, + { + "epoch": 1.15, + "learning_rate": 4.044757722825668e-05, + "loss": 0.0, + "step": 4960 + }, + { + "epoch": 1.15, + "learning_rate": 4.044565133656884e-05, + "loss": 0.0001, + "step": 4961 + }, + { + "epoch": 1.15, + "learning_rate": 4.044372544488098e-05, + "loss": 0.0001, + "step": 4962 + }, + { + "epoch": 1.15, + "learning_rate": 4.044179955319313e-05, + "loss": 0.0, + "step": 4963 + }, + { + "epoch": 1.15, + "learning_rate": 4.043987366150528e-05, + "loss": 0.0001, + "step": 4964 + }, + { + "epoch": 1.15, + "learning_rate": 4.043794776981743e-05, + "loss": 0.0006, + "step": 4965 + }, + { + "epoch": 1.15, + "learning_rate": 4.0436021878129574e-05, + "loss": 0.0004, + "step": 4966 + }, + { + "epoch": 1.15, + "learning_rate": 4.043409598644172e-05, + "loss": 0.0004, + "step": 4967 + }, + { + "epoch": 1.15, + "learning_rate": 4.0432170094753876e-05, + "loss": 0.0, + "step": 4968 + }, + { + "epoch": 1.15, + "learning_rate": 4.043024420306602e-05, + "loss": 0.0009, + "step": 4969 + }, + { + "epoch": 1.15, + "learning_rate": 4.042831831137817e-05, + "loss": 0.0, + "step": 4970 + }, + { + "epoch": 1.15, + "learning_rate": 4.042639241969032e-05, + "loss": 0.0001, + "step": 4971 + }, + { + "epoch": 1.15, + "learning_rate": 4.0424466528002466e-05, + "loss": 0.009, + "step": 4972 + }, + { + "epoch": 1.15, + "learning_rate": 4.0422540636314614e-05, + "loss": 0.0001, + "step": 4973 + }, + { + "epoch": 1.15, + "learning_rate": 4.042061474462677e-05, + "loss": 0.0016, + "step": 4974 + }, + { + "epoch": 1.15, + "learning_rate": 4.0418688852938915e-05, + "loss": 0.0, + "step": 4975 + }, + { + "epoch": 1.15, + "learning_rate": 4.0416762961251056e-05, + "loss": 0.0085, + "step": 4976 + }, + { + "epoch": 1.15, + "learning_rate": 4.041483706956321e-05, + "loss": 0.0005, + "step": 4977 + }, + { + "epoch": 1.15, + "learning_rate": 4.041291117787536e-05, + "loss": 0.0, + "step": 4978 + }, + { + "epoch": 1.15, + "learning_rate": 4.0410985286187505e-05, + "loss": 0.0, + "step": 4979 + }, + { + "epoch": 1.15, + "learning_rate": 4.040905939449965e-05, + "loss": 0.0, + "step": 4980 + }, + { + "epoch": 1.15, + "learning_rate": 4.040713350281181e-05, + "loss": 0.0, + "step": 4981 + }, + { + "epoch": 1.15, + "learning_rate": 4.0405207611123954e-05, + "loss": 0.0, + "step": 4982 + }, + { + "epoch": 1.15, + "learning_rate": 4.04032817194361e-05, + "loss": 0.0, + "step": 4983 + }, + { + "epoch": 1.15, + "learning_rate": 4.040135582774825e-05, + "loss": 0.0058, + "step": 4984 + }, + { + "epoch": 1.15, + "learning_rate": 4.03994299360604e-05, + "loss": 0.0, + "step": 4985 + }, + { + "epoch": 1.15, + "learning_rate": 4.0397504044372544e-05, + "loss": 0.0001, + "step": 4986 + }, + { + "epoch": 1.15, + "learning_rate": 4.039557815268469e-05, + "loss": 0.0001, + "step": 4987 + }, + { + "epoch": 1.15, + "learning_rate": 4.0393652260996846e-05, + "loss": 0.0, + "step": 4988 + }, + { + "epoch": 1.15, + "learning_rate": 4.0391726369308994e-05, + "loss": 0.0003, + "step": 4989 + }, + { + "epoch": 1.15, + "learning_rate": 4.038980047762114e-05, + "loss": 0.0, + "step": 4990 + }, + { + "epoch": 1.15, + "learning_rate": 4.038787458593329e-05, + "loss": 0.0002, + "step": 4991 + }, + { + "epoch": 1.15, + "learning_rate": 4.038594869424544e-05, + "loss": 0.0, + "step": 4992 + }, + { + "epoch": 1.15, + "learning_rate": 4.0384022802557583e-05, + "loss": 0.0, + "step": 4993 + }, + { + "epoch": 1.15, + "learning_rate": 4.038209691086973e-05, + "loss": 0.0007, + "step": 4994 + }, + { + "epoch": 1.15, + "learning_rate": 4.0380171019181885e-05, + "loss": 0.0003, + "step": 4995 + }, + { + "epoch": 1.15, + "learning_rate": 4.037824512749403e-05, + "loss": 0.0004, + "step": 4996 + }, + { + "epoch": 1.15, + "learning_rate": 4.037631923580618e-05, + "loss": 0.0006, + "step": 4997 + }, + { + "epoch": 1.16, + "learning_rate": 4.037439334411833e-05, + "loss": 0.0, + "step": 4998 + }, + { + "epoch": 1.16, + "learning_rate": 4.037246745243048e-05, + "loss": 0.0, + "step": 4999 + }, + { + "epoch": 1.16, + "learning_rate": 4.037054156074262e-05, + "loss": 0.0, + "step": 5000 + }, + { + "epoch": 1.16, + "learning_rate": 4.036861566905478e-05, + "loss": 0.0, + "step": 5001 + }, + { + "epoch": 1.16, + "learning_rate": 4.0366689777366924e-05, + "loss": 0.0, + "step": 5002 + }, + { + "epoch": 1.16, + "learning_rate": 4.036476388567907e-05, + "loss": 0.0003, + "step": 5003 + }, + { + "epoch": 1.16, + "learning_rate": 4.036283799399122e-05, + "loss": 0.0001, + "step": 5004 + }, + { + "epoch": 1.16, + "learning_rate": 4.036091210230337e-05, + "loss": 0.0468, + "step": 5005 + }, + { + "epoch": 1.16, + "learning_rate": 4.035898621061552e-05, + "loss": 0.0001, + "step": 5006 + }, + { + "epoch": 1.16, + "learning_rate": 4.035706031892766e-05, + "loss": 0.0011, + "step": 5007 + }, + { + "epoch": 1.16, + "learning_rate": 4.0355134427239816e-05, + "loss": 0.0002, + "step": 5008 + }, + { + "epoch": 1.16, + "learning_rate": 4.0353208535551963e-05, + "loss": 0.0001, + "step": 5009 + }, + { + "epoch": 1.16, + "learning_rate": 4.035128264386411e-05, + "loss": 0.0001, + "step": 5010 + }, + { + "epoch": 1.16, + "learning_rate": 4.034935675217626e-05, + "loss": 0.0, + "step": 5011 + }, + { + "epoch": 1.16, + "learning_rate": 4.0347430860488406e-05, + "loss": 0.0003, + "step": 5012 + }, + { + "epoch": 1.16, + "learning_rate": 4.034550496880056e-05, + "loss": 0.0001, + "step": 5013 + }, + { + "epoch": 1.16, + "learning_rate": 4.03435790771127e-05, + "loss": 0.0001, + "step": 5014 + }, + { + "epoch": 1.16, + "learning_rate": 4.0341653185424855e-05, + "loss": 0.0, + "step": 5015 + }, + { + "epoch": 1.16, + "learning_rate": 4.0339727293737e-05, + "loss": 0.0051, + "step": 5016 + }, + { + "epoch": 1.16, + "learning_rate": 4.033780140204915e-05, + "loss": 0.0001, + "step": 5017 + }, + { + "epoch": 1.16, + "learning_rate": 4.03358755103613e-05, + "loss": 0.0008, + "step": 5018 + }, + { + "epoch": 1.16, + "learning_rate": 4.033394961867345e-05, + "loss": 0.0001, + "step": 5019 + }, + { + "epoch": 1.16, + "learning_rate": 4.03320237269856e-05, + "loss": 0.0005, + "step": 5020 + }, + { + "epoch": 1.16, + "learning_rate": 4.033009783529774e-05, + "loss": 0.0, + "step": 5021 + }, + { + "epoch": 1.16, + "learning_rate": 4.0328171943609894e-05, + "loss": 0.0, + "step": 5022 + }, + { + "epoch": 1.16, + "learning_rate": 4.032624605192204e-05, + "loss": 0.0001, + "step": 5023 + }, + { + "epoch": 1.16, + "learning_rate": 4.032432016023419e-05, + "loss": 0.0003, + "step": 5024 + }, + { + "epoch": 1.16, + "learning_rate": 4.0322394268546337e-05, + "loss": 0.0012, + "step": 5025 + }, + { + "epoch": 1.16, + "learning_rate": 4.032046837685849e-05, + "loss": 0.0002, + "step": 5026 + }, + { + "epoch": 1.16, + "learning_rate": 4.031854248517064e-05, + "loss": 0.0001, + "step": 5027 + }, + { + "epoch": 1.16, + "learning_rate": 4.0316616593482786e-05, + "loss": 0.0, + "step": 5028 + }, + { + "epoch": 1.16, + "learning_rate": 4.031469070179493e-05, + "loss": 0.0, + "step": 5029 + }, + { + "epoch": 1.16, + "learning_rate": 4.031276481010709e-05, + "loss": 0.0003, + "step": 5030 + }, + { + "epoch": 1.16, + "learning_rate": 4.031083891841923e-05, + "loss": 0.0, + "step": 5031 + }, + { + "epoch": 1.16, + "learning_rate": 4.0308913026731376e-05, + "loss": 0.0001, + "step": 5032 + }, + { + "epoch": 1.16, + "learning_rate": 4.030698713504353e-05, + "loss": 0.0007, + "step": 5033 + }, + { + "epoch": 1.16, + "learning_rate": 4.030506124335567e-05, + "loss": 0.0014, + "step": 5034 + }, + { + "epoch": 1.16, + "learning_rate": 4.0303135351667825e-05, + "loss": 0.0, + "step": 5035 + }, + { + "epoch": 1.16, + "learning_rate": 4.030120945997997e-05, + "loss": 0.0001, + "step": 5036 + }, + { + "epoch": 1.16, + "learning_rate": 4.029928356829212e-05, + "loss": 0.0, + "step": 5037 + }, + { + "epoch": 1.16, + "learning_rate": 4.029735767660427e-05, + "loss": 0.0, + "step": 5038 + }, + { + "epoch": 1.16, + "learning_rate": 4.029543178491642e-05, + "loss": 0.0011, + "step": 5039 + }, + { + "epoch": 1.16, + "learning_rate": 4.029350589322857e-05, + "loss": 0.0, + "step": 5040 + }, + { + "epoch": 1.17, + "learning_rate": 4.029158000154071e-05, + "loss": 0.0001, + "step": 5041 + }, + { + "epoch": 1.17, + "learning_rate": 4.0289654109852864e-05, + "loss": 0.0, + "step": 5042 + }, + { + "epoch": 1.17, + "learning_rate": 4.028772821816501e-05, + "loss": 0.0004, + "step": 5043 + }, + { + "epoch": 1.17, + "learning_rate": 4.028580232647716e-05, + "loss": 0.0, + "step": 5044 + }, + { + "epoch": 1.17, + "learning_rate": 4.0283876434789306e-05, + "loss": 0.0001, + "step": 5045 + }, + { + "epoch": 1.17, + "learning_rate": 4.028195054310146e-05, + "loss": 0.0061, + "step": 5046 + }, + { + "epoch": 1.17, + "learning_rate": 4.028002465141361e-05, + "loss": 0.0, + "step": 5047 + }, + { + "epoch": 1.17, + "learning_rate": 4.0278098759725756e-05, + "loss": 0.0002, + "step": 5048 + }, + { + "epoch": 1.17, + "learning_rate": 4.02761728680379e-05, + "loss": 0.0, + "step": 5049 + }, + { + "epoch": 1.17, + "learning_rate": 4.027424697635005e-05, + "loss": 0.0001, + "step": 5050 + }, + { + "epoch": 1.17, + "learning_rate": 4.02723210846622e-05, + "loss": 0.0002, + "step": 5051 + }, + { + "epoch": 1.17, + "learning_rate": 4.0270395192974346e-05, + "loss": 0.0005, + "step": 5052 + }, + { + "epoch": 1.17, + "learning_rate": 4.02684693012865e-05, + "loss": 0.0, + "step": 5053 + }, + { + "epoch": 1.17, + "learning_rate": 4.026654340959865e-05, + "loss": 0.0672, + "step": 5054 + }, + { + "epoch": 1.17, + "learning_rate": 4.0264617517910795e-05, + "loss": 0.0, + "step": 5055 + }, + { + "epoch": 1.17, + "learning_rate": 4.026269162622294e-05, + "loss": 0.0, + "step": 5056 + }, + { + "epoch": 1.17, + "learning_rate": 4.0260765734535097e-05, + "loss": 0.0008, + "step": 5057 + }, + { + "epoch": 1.17, + "learning_rate": 4.025883984284724e-05, + "loss": 0.003, + "step": 5058 + }, + { + "epoch": 1.17, + "learning_rate": 4.0256913951159385e-05, + "loss": 0.0, + "step": 5059 + }, + { + "epoch": 1.17, + "learning_rate": 4.025498805947154e-05, + "loss": 0.0001, + "step": 5060 + }, + { + "epoch": 1.17, + "learning_rate": 4.0253062167783686e-05, + "loss": 0.0004, + "step": 5061 + }, + { + "epoch": 1.17, + "learning_rate": 4.0251136276095834e-05, + "loss": 0.0001, + "step": 5062 + }, + { + "epoch": 1.17, + "learning_rate": 4.024921038440798e-05, + "loss": 0.0004, + "step": 5063 + }, + { + "epoch": 1.17, + "learning_rate": 4.0247284492720136e-05, + "loss": 0.0005, + "step": 5064 + }, + { + "epoch": 1.17, + "learning_rate": 4.0245358601032276e-05, + "loss": 0.0095, + "step": 5065 + }, + { + "epoch": 1.17, + "learning_rate": 4.024343270934443e-05, + "loss": 0.0, + "step": 5066 + }, + { + "epoch": 1.17, + "learning_rate": 4.024150681765658e-05, + "loss": 0.0002, + "step": 5067 + }, + { + "epoch": 1.17, + "learning_rate": 4.0239580925968726e-05, + "loss": 0.0025, + "step": 5068 + }, + { + "epoch": 1.17, + "learning_rate": 4.023765503428087e-05, + "loss": 0.0, + "step": 5069 + }, + { + "epoch": 1.17, + "learning_rate": 4.023572914259302e-05, + "loss": 0.0002, + "step": 5070 + }, + { + "epoch": 1.17, + "learning_rate": 4.0233803250905175e-05, + "loss": 0.0, + "step": 5071 + }, + { + "epoch": 1.17, + "learning_rate": 4.0231877359217315e-05, + "loss": 0.0001, + "step": 5072 + }, + { + "epoch": 1.17, + "learning_rate": 4.022995146752947e-05, + "loss": 0.0, + "step": 5073 + }, + { + "epoch": 1.17, + "learning_rate": 4.022802557584162e-05, + "loss": 0.0001, + "step": 5074 + }, + { + "epoch": 1.17, + "learning_rate": 4.0226099684153765e-05, + "loss": 0.0001, + "step": 5075 + }, + { + "epoch": 1.17, + "learning_rate": 4.022417379246591e-05, + "loss": 0.0, + "step": 5076 + }, + { + "epoch": 1.17, + "learning_rate": 4.022224790077806e-05, + "loss": 0.0, + "step": 5077 + }, + { + "epoch": 1.17, + "learning_rate": 4.0220322009090214e-05, + "loss": 0.0, + "step": 5078 + }, + { + "epoch": 1.17, + "learning_rate": 4.0218396117402355e-05, + "loss": 0.0027, + "step": 5079 + }, + { + "epoch": 1.17, + "learning_rate": 4.021647022571451e-05, + "loss": 0.0005, + "step": 5080 + }, + { + "epoch": 1.17, + "learning_rate": 4.0214544334026656e-05, + "loss": 0.0006, + "step": 5081 + }, + { + "epoch": 1.17, + "learning_rate": 4.0212618442338804e-05, + "loss": 0.0, + "step": 5082 + }, + { + "epoch": 1.17, + "learning_rate": 4.021069255065095e-05, + "loss": 0.0001, + "step": 5083 + }, + { + "epoch": 1.17, + "learning_rate": 4.0208766658963106e-05, + "loss": 0.0, + "step": 5084 + }, + { + "epoch": 1.18, + "learning_rate": 4.020684076727525e-05, + "loss": 0.0058, + "step": 5085 + }, + { + "epoch": 1.18, + "learning_rate": 4.02049148755874e-05, + "loss": 0.0009, + "step": 5086 + }, + { + "epoch": 1.18, + "learning_rate": 4.020298898389955e-05, + "loss": 0.0, + "step": 5087 + }, + { + "epoch": 1.18, + "learning_rate": 4.0201063092211695e-05, + "loss": 0.0001, + "step": 5088 + }, + { + "epoch": 1.18, + "learning_rate": 4.019913720052384e-05, + "loss": 0.0042, + "step": 5089 + }, + { + "epoch": 1.18, + "learning_rate": 4.019721130883599e-05, + "loss": 0.0, + "step": 5090 + }, + { + "epoch": 1.18, + "learning_rate": 4.0195285417148145e-05, + "loss": 0.0001, + "step": 5091 + }, + { + "epoch": 1.18, + "learning_rate": 4.019335952546029e-05, + "loss": 0.001, + "step": 5092 + }, + { + "epoch": 1.18, + "learning_rate": 4.019143363377244e-05, + "loss": 0.0001, + "step": 5093 + }, + { + "epoch": 1.18, + "learning_rate": 4.018950774208459e-05, + "loss": 0.0001, + "step": 5094 + }, + { + "epoch": 1.18, + "learning_rate": 4.018758185039674e-05, + "loss": 0.0, + "step": 5095 + }, + { + "epoch": 1.18, + "learning_rate": 4.018565595870888e-05, + "loss": 0.0009, + "step": 5096 + }, + { + "epoch": 1.18, + "learning_rate": 4.018373006702103e-05, + "loss": 0.001, + "step": 5097 + }, + { + "epoch": 1.18, + "learning_rate": 4.0181804175333184e-05, + "loss": 0.0005, + "step": 5098 + }, + { + "epoch": 1.18, + "learning_rate": 4.017987828364533e-05, + "loss": 0.0005, + "step": 5099 + }, + { + "epoch": 1.18, + "learning_rate": 4.017795239195748e-05, + "loss": 0.0001, + "step": 5100 + }, + { + "epoch": 1.18, + "learning_rate": 4.0176026500269626e-05, + "loss": 0.0001, + "step": 5101 + }, + { + "epoch": 1.18, + "learning_rate": 4.017410060858178e-05, + "loss": 0.001, + "step": 5102 + }, + { + "epoch": 1.18, + "learning_rate": 4.017217471689392e-05, + "loss": 0.0003, + "step": 5103 + }, + { + "epoch": 1.18, + "learning_rate": 4.0170248825206075e-05, + "loss": 0.0001, + "step": 5104 + }, + { + "epoch": 1.18, + "learning_rate": 4.016832293351822e-05, + "loss": 0.0001, + "step": 5105 + }, + { + "epoch": 1.18, + "learning_rate": 4.016639704183037e-05, + "loss": 0.0, + "step": 5106 + }, + { + "epoch": 1.18, + "learning_rate": 4.016447115014252e-05, + "loss": 0.0003, + "step": 5107 + }, + { + "epoch": 1.18, + "learning_rate": 4.0162545258454665e-05, + "loss": 0.0, + "step": 5108 + }, + { + "epoch": 1.18, + "learning_rate": 4.016061936676682e-05, + "loss": 0.001, + "step": 5109 + }, + { + "epoch": 1.18, + "learning_rate": 4.015869347507896e-05, + "loss": 0.0, + "step": 5110 + }, + { + "epoch": 1.18, + "learning_rate": 4.0156767583391115e-05, + "loss": 0.0, + "step": 5111 + }, + { + "epoch": 1.18, + "learning_rate": 4.015484169170326e-05, + "loss": 0.0001, + "step": 5112 + }, + { + "epoch": 1.18, + "learning_rate": 4.015291580001541e-05, + "loss": 0.0044, + "step": 5113 + }, + { + "epoch": 1.18, + "learning_rate": 4.015098990832756e-05, + "loss": 0.0, + "step": 5114 + }, + { + "epoch": 1.18, + "learning_rate": 4.0149064016639704e-05, + "loss": 0.0002, + "step": 5115 + }, + { + "epoch": 1.18, + "learning_rate": 4.014713812495185e-05, + "loss": 0.0001, + "step": 5116 + }, + { + "epoch": 1.18, + "learning_rate": 4.0145212233264e-05, + "loss": 0.0, + "step": 5117 + }, + { + "epoch": 1.18, + "learning_rate": 4.0143286341576154e-05, + "loss": 0.0033, + "step": 5118 + }, + { + "epoch": 1.18, + "learning_rate": 4.01413604498883e-05, + "loss": 0.0001, + "step": 5119 + }, + { + "epoch": 1.18, + "learning_rate": 4.013943455820045e-05, + "loss": 0.0002, + "step": 5120 + }, + { + "epoch": 1.18, + "learning_rate": 4.0137508666512596e-05, + "loss": 0.0, + "step": 5121 + }, + { + "epoch": 1.18, + "learning_rate": 4.013558277482475e-05, + "loss": 0.0, + "step": 5122 + }, + { + "epoch": 1.18, + "learning_rate": 4.013365688313689e-05, + "loss": 0.0, + "step": 5123 + }, + { + "epoch": 1.18, + "learning_rate": 4.013173099144904e-05, + "loss": 0.0001, + "step": 5124 + }, + { + "epoch": 1.18, + "learning_rate": 4.012980509976119e-05, + "loss": 0.0019, + "step": 5125 + }, + { + "epoch": 1.18, + "learning_rate": 4.012787920807334e-05, + "loss": 0.0004, + "step": 5126 + }, + { + "epoch": 1.18, + "learning_rate": 4.012595331638549e-05, + "loss": 0.0228, + "step": 5127 + }, + { + "epoch": 1.19, + "learning_rate": 4.0124027424697635e-05, + "loss": 0.0, + "step": 5128 + }, + { + "epoch": 1.19, + "learning_rate": 4.012210153300979e-05, + "loss": 0.0053, + "step": 5129 + }, + { + "epoch": 1.19, + "learning_rate": 4.012017564132193e-05, + "loss": 0.0012, + "step": 5130 + }, + { + "epoch": 1.19, + "learning_rate": 4.0118249749634084e-05, + "loss": 0.0001, + "step": 5131 + }, + { + "epoch": 1.19, + "learning_rate": 4.011632385794623e-05, + "loss": 0.0005, + "step": 5132 + }, + { + "epoch": 1.19, + "learning_rate": 4.011439796625838e-05, + "loss": 0.003, + "step": 5133 + }, + { + "epoch": 1.19, + "learning_rate": 4.011247207457053e-05, + "loss": 0.0001, + "step": 5134 + }, + { + "epoch": 1.19, + "learning_rate": 4.0110546182882674e-05, + "loss": 0.0001, + "step": 5135 + }, + { + "epoch": 1.19, + "learning_rate": 4.010862029119483e-05, + "loss": 0.0, + "step": 5136 + }, + { + "epoch": 1.19, + "learning_rate": 4.010669439950697e-05, + "loss": 0.0003, + "step": 5137 + }, + { + "epoch": 1.19, + "learning_rate": 4.0104768507819124e-05, + "loss": 0.0001, + "step": 5138 + }, + { + "epoch": 1.19, + "learning_rate": 4.010284261613127e-05, + "loss": 0.0001, + "step": 5139 + }, + { + "epoch": 1.19, + "learning_rate": 4.010091672444342e-05, + "loss": 0.0, + "step": 5140 + }, + { + "epoch": 1.19, + "learning_rate": 4.0098990832755566e-05, + "loss": 0.0103, + "step": 5141 + }, + { + "epoch": 1.19, + "learning_rate": 4.009706494106772e-05, + "loss": 0.0001, + "step": 5142 + }, + { + "epoch": 1.19, + "learning_rate": 4.009513904937987e-05, + "loss": 0.0004, + "step": 5143 + }, + { + "epoch": 1.19, + "learning_rate": 4.009321315769201e-05, + "loss": 0.0058, + "step": 5144 + }, + { + "epoch": 1.19, + "learning_rate": 4.009128726600416e-05, + "loss": 0.0003, + "step": 5145 + }, + { + "epoch": 1.19, + "learning_rate": 4.008936137431631e-05, + "loss": 0.0, + "step": 5146 + }, + { + "epoch": 1.19, + "learning_rate": 4.008743548262846e-05, + "loss": 0.0001, + "step": 5147 + }, + { + "epoch": 1.19, + "learning_rate": 4.0085509590940605e-05, + "loss": 0.0002, + "step": 5148 + }, + { + "epoch": 1.19, + "learning_rate": 4.008358369925276e-05, + "loss": 0.0, + "step": 5149 + }, + { + "epoch": 1.19, + "learning_rate": 4.008165780756491e-05, + "loss": 0.0, + "step": 5150 + }, + { + "epoch": 1.19, + "learning_rate": 4.0079731915877054e-05, + "loss": 0.0011, + "step": 5151 + }, + { + "epoch": 1.19, + "learning_rate": 4.00778060241892e-05, + "loss": 0.0001, + "step": 5152 + }, + { + "epoch": 1.19, + "learning_rate": 4.007588013250135e-05, + "loss": 0.0, + "step": 5153 + }, + { + "epoch": 1.19, + "learning_rate": 4.00739542408135e-05, + "loss": 0.0, + "step": 5154 + }, + { + "epoch": 1.19, + "learning_rate": 4.0072028349125644e-05, + "loss": 0.0003, + "step": 5155 + }, + { + "epoch": 1.19, + "learning_rate": 4.00701024574378e-05, + "loss": 0.0, + "step": 5156 + }, + { + "epoch": 1.19, + "learning_rate": 4.0068176565749946e-05, + "loss": 0.0002, + "step": 5157 + }, + { + "epoch": 1.19, + "learning_rate": 4.0066250674062093e-05, + "loss": 0.0004, + "step": 5158 + }, + { + "epoch": 1.19, + "learning_rate": 4.006432478237424e-05, + "loss": 0.0002, + "step": 5159 + }, + { + "epoch": 1.19, + "learning_rate": 4.0062398890686395e-05, + "loss": 0.0001, + "step": 5160 + }, + { + "epoch": 1.19, + "learning_rate": 4.0060472998998536e-05, + "loss": 0.0, + "step": 5161 + }, + { + "epoch": 1.19, + "learning_rate": 4.005854710731068e-05, + "loss": 0.0008, + "step": 5162 + }, + { + "epoch": 1.19, + "learning_rate": 4.005662121562284e-05, + "loss": 0.0006, + "step": 5163 + }, + { + "epoch": 1.19, + "learning_rate": 4.0054695323934985e-05, + "loss": 0.0001, + "step": 5164 + }, + { + "epoch": 1.19, + "learning_rate": 4.005276943224713e-05, + "loss": 0.0, + "step": 5165 + }, + { + "epoch": 1.19, + "learning_rate": 4.005084354055928e-05, + "loss": 0.0, + "step": 5166 + }, + { + "epoch": 1.19, + "learning_rate": 4.0048917648871434e-05, + "loss": 0.0002, + "step": 5167 + }, + { + "epoch": 1.19, + "learning_rate": 4.0046991757183575e-05, + "loss": 0.0005, + "step": 5168 + }, + { + "epoch": 1.19, + "learning_rate": 4.004506586549573e-05, + "loss": 0.0, + "step": 5169 + }, + { + "epoch": 1.19, + "learning_rate": 4.004313997380788e-05, + "loss": 0.0011, + "step": 5170 + }, + { + "epoch": 1.2, + "learning_rate": 4.0041214082120024e-05, + "loss": 0.0, + "step": 5171 + }, + { + "epoch": 1.2, + "learning_rate": 4.003928819043217e-05, + "loss": 0.0, + "step": 5172 + }, + { + "epoch": 1.2, + "learning_rate": 4.003736229874432e-05, + "loss": 0.0, + "step": 5173 + }, + { + "epoch": 1.2, + "learning_rate": 4.0035436407056473e-05, + "loss": 0.0, + "step": 5174 + }, + { + "epoch": 1.2, + "learning_rate": 4.0033510515368614e-05, + "loss": 0.0003, + "step": 5175 + }, + { + "epoch": 1.2, + "learning_rate": 4.003158462368077e-05, + "loss": 0.0, + "step": 5176 + }, + { + "epoch": 1.2, + "learning_rate": 4.0029658731992916e-05, + "loss": 0.0, + "step": 5177 + }, + { + "epoch": 1.2, + "learning_rate": 4.002773284030506e-05, + "loss": 0.0, + "step": 5178 + }, + { + "epoch": 1.2, + "learning_rate": 4.002580694861721e-05, + "loss": 0.0003, + "step": 5179 + }, + { + "epoch": 1.2, + "learning_rate": 4.002388105692936e-05, + "loss": 0.0, + "step": 5180 + }, + { + "epoch": 1.2, + "learning_rate": 4.002195516524151e-05, + "loss": 0.0001, + "step": 5181 + }, + { + "epoch": 1.2, + "learning_rate": 4.002002927355365e-05, + "loss": 0.0, + "step": 5182 + }, + { + "epoch": 1.2, + "learning_rate": 4.001810338186581e-05, + "loss": 0.0, + "step": 5183 + }, + { + "epoch": 1.2, + "learning_rate": 4.0016177490177955e-05, + "loss": 0.0002, + "step": 5184 + }, + { + "epoch": 1.2, + "learning_rate": 4.00142515984901e-05, + "loss": 0.0001, + "step": 5185 + }, + { + "epoch": 1.2, + "learning_rate": 4.001232570680225e-05, + "loss": 0.0001, + "step": 5186 + }, + { + "epoch": 1.2, + "learning_rate": 4.0010399815114404e-05, + "loss": 0.0005, + "step": 5187 + }, + { + "epoch": 1.2, + "learning_rate": 4.0008473923426545e-05, + "loss": 0.0, + "step": 5188 + }, + { + "epoch": 1.2, + "learning_rate": 4.000654803173869e-05, + "loss": 0.0002, + "step": 5189 + }, + { + "epoch": 1.2, + "learning_rate": 4.000462214005085e-05, + "loss": 0.0, + "step": 5190 + }, + { + "epoch": 1.2, + "learning_rate": 4.0002696248362994e-05, + "loss": 0.0045, + "step": 5191 + }, + { + "epoch": 1.2, + "learning_rate": 4.000077035667514e-05, + "loss": 0.0, + "step": 5192 + }, + { + "epoch": 1.2, + "learning_rate": 3.999884446498729e-05, + "loss": 0.0, + "step": 5193 + }, + { + "epoch": 1.2, + "learning_rate": 3.999691857329944e-05, + "loss": 0.0, + "step": 5194 + }, + { + "epoch": 1.2, + "learning_rate": 3.9994992681611584e-05, + "loss": 0.0002, + "step": 5195 + }, + { + "epoch": 1.2, + "learning_rate": 3.999306678992374e-05, + "loss": 0.0, + "step": 5196 + }, + { + "epoch": 1.2, + "learning_rate": 3.9991140898235886e-05, + "loss": 0.0004, + "step": 5197 + }, + { + "epoch": 1.2, + "learning_rate": 3.998921500654803e-05, + "loss": 0.0004, + "step": 5198 + }, + { + "epoch": 1.2, + "learning_rate": 3.998728911486018e-05, + "loss": 0.0, + "step": 5199 + }, + { + "epoch": 1.2, + "learning_rate": 3.998536322317233e-05, + "loss": 0.0001, + "step": 5200 + }, + { + "epoch": 1.2, + "learning_rate": 3.998343733148448e-05, + "loss": 0.0001, + "step": 5201 + }, + { + "epoch": 1.2, + "learning_rate": 3.998151143979662e-05, + "loss": 0.0009, + "step": 5202 + }, + { + "epoch": 1.2, + "learning_rate": 3.997958554810878e-05, + "loss": 0.0002, + "step": 5203 + }, + { + "epoch": 1.2, + "learning_rate": 3.9977659656420925e-05, + "loss": 0.0, + "step": 5204 + }, + { + "epoch": 1.2, + "learning_rate": 3.997573376473307e-05, + "loss": 0.0148, + "step": 5205 + }, + { + "epoch": 1.2, + "learning_rate": 3.997380787304522e-05, + "loss": 0.0002, + "step": 5206 + }, + { + "epoch": 1.2, + "learning_rate": 3.9971881981357374e-05, + "loss": 0.0014, + "step": 5207 + }, + { + "epoch": 1.2, + "learning_rate": 3.996995608966952e-05, + "loss": 0.0004, + "step": 5208 + }, + { + "epoch": 1.2, + "learning_rate": 3.996803019798166e-05, + "loss": 0.0, + "step": 5209 + }, + { + "epoch": 1.2, + "learning_rate": 3.9966104306293816e-05, + "loss": 0.0, + "step": 5210 + }, + { + "epoch": 1.2, + "learning_rate": 3.9964178414605964e-05, + "loss": 0.0001, + "step": 5211 + }, + { + "epoch": 1.2, + "learning_rate": 3.996225252291811e-05, + "loss": 0.0003, + "step": 5212 + }, + { + "epoch": 1.2, + "learning_rate": 3.996032663123026e-05, + "loss": 0.0001, + "step": 5213 + }, + { + "epoch": 1.2, + "learning_rate": 3.995840073954241e-05, + "loss": 0.0, + "step": 5214 + }, + { + "epoch": 1.21, + "learning_rate": 3.995647484785456e-05, + "loss": 0.0007, + "step": 5215 + }, + { + "epoch": 1.21, + "learning_rate": 3.995454895616671e-05, + "loss": 0.0002, + "step": 5216 + }, + { + "epoch": 1.21, + "learning_rate": 3.9952623064478856e-05, + "loss": 0.0001, + "step": 5217 + }, + { + "epoch": 1.21, + "learning_rate": 3.9950697172791e-05, + "loss": 0.0004, + "step": 5218 + }, + { + "epoch": 1.21, + "learning_rate": 3.994877128110315e-05, + "loss": 0.0006, + "step": 5219 + }, + { + "epoch": 1.21, + "learning_rate": 3.99468453894153e-05, + "loss": 0.0, + "step": 5220 + }, + { + "epoch": 1.21, + "learning_rate": 3.994491949772745e-05, + "loss": 0.0002, + "step": 5221 + }, + { + "epoch": 1.21, + "learning_rate": 3.99429936060396e-05, + "loss": 0.0, + "step": 5222 + }, + { + "epoch": 1.21, + "learning_rate": 3.994106771435175e-05, + "loss": 0.0002, + "step": 5223 + }, + { + "epoch": 1.21, + "learning_rate": 3.9939141822663895e-05, + "loss": 0.0002, + "step": 5224 + }, + { + "epoch": 1.21, + "learning_rate": 3.993721593097605e-05, + "loss": 0.0001, + "step": 5225 + }, + { + "epoch": 1.21, + "learning_rate": 3.993529003928819e-05, + "loss": 0.0, + "step": 5226 + }, + { + "epoch": 1.21, + "learning_rate": 3.993336414760034e-05, + "loss": 0.002, + "step": 5227 + }, + { + "epoch": 1.21, + "learning_rate": 3.993143825591249e-05, + "loss": 0.0, + "step": 5228 + }, + { + "epoch": 1.21, + "learning_rate": 3.992951236422464e-05, + "loss": 0.0, + "step": 5229 + }, + { + "epoch": 1.21, + "learning_rate": 3.9927586472536786e-05, + "loss": 0.0, + "step": 5230 + }, + { + "epoch": 1.21, + "learning_rate": 3.9925660580848934e-05, + "loss": 0.0, + "step": 5231 + }, + { + "epoch": 1.21, + "learning_rate": 3.992373468916109e-05, + "loss": 0.0, + "step": 5232 + }, + { + "epoch": 1.21, + "learning_rate": 3.992180879747323e-05, + "loss": 0.0, + "step": 5233 + }, + { + "epoch": 1.21, + "learning_rate": 3.991988290578538e-05, + "loss": 0.0001, + "step": 5234 + }, + { + "epoch": 1.21, + "learning_rate": 3.991795701409753e-05, + "loss": 0.0002, + "step": 5235 + }, + { + "epoch": 1.21, + "learning_rate": 3.991603112240968e-05, + "loss": 0.0003, + "step": 5236 + }, + { + "epoch": 1.21, + "learning_rate": 3.9914105230721826e-05, + "loss": 0.0, + "step": 5237 + }, + { + "epoch": 1.21, + "learning_rate": 3.991217933903397e-05, + "loss": 0.0, + "step": 5238 + }, + { + "epoch": 1.21, + "learning_rate": 3.991025344734613e-05, + "loss": 0.0, + "step": 5239 + }, + { + "epoch": 1.21, + "learning_rate": 3.990832755565827e-05, + "loss": 0.0001, + "step": 5240 + }, + { + "epoch": 1.21, + "learning_rate": 3.990640166397042e-05, + "loss": 0.0, + "step": 5241 + }, + { + "epoch": 1.21, + "learning_rate": 3.990447577228257e-05, + "loss": 0.0012, + "step": 5242 + }, + { + "epoch": 1.21, + "learning_rate": 3.990254988059472e-05, + "loss": 0.0, + "step": 5243 + }, + { + "epoch": 1.21, + "learning_rate": 3.9900623988906865e-05, + "loss": 0.0003, + "step": 5244 + }, + { + "epoch": 1.21, + "learning_rate": 3.989869809721901e-05, + "loss": 0.0, + "step": 5245 + }, + { + "epoch": 1.21, + "learning_rate": 3.9896772205531166e-05, + "loss": 0.0002, + "step": 5246 + }, + { + "epoch": 1.21, + "learning_rate": 3.989484631384331e-05, + "loss": 0.0, + "step": 5247 + }, + { + "epoch": 1.21, + "learning_rate": 3.989292042215546e-05, + "loss": 0.0, + "step": 5248 + }, + { + "epoch": 1.21, + "learning_rate": 3.989099453046761e-05, + "loss": 0.0002, + "step": 5249 + }, + { + "epoch": 1.21, + "learning_rate": 3.9889068638779756e-05, + "loss": 0.0001, + "step": 5250 + }, + { + "epoch": 1.21, + "learning_rate": 3.9887142747091904e-05, + "loss": 0.0, + "step": 5251 + }, + { + "epoch": 1.21, + "learning_rate": 3.988521685540406e-05, + "loss": 0.0, + "step": 5252 + }, + { + "epoch": 1.21, + "learning_rate": 3.9883290963716205e-05, + "loss": 0.0001, + "step": 5253 + }, + { + "epoch": 1.21, + "learning_rate": 3.9881365072028346e-05, + "loss": 0.0, + "step": 5254 + }, + { + "epoch": 1.21, + "learning_rate": 3.98794391803405e-05, + "loss": 0.0017, + "step": 5255 + }, + { + "epoch": 1.21, + "learning_rate": 3.987751328865265e-05, + "loss": 0.0, + "step": 5256 + }, + { + "epoch": 1.21, + "learning_rate": 3.9875587396964795e-05, + "loss": 0.0001, + "step": 5257 + }, + { + "epoch": 1.22, + "learning_rate": 3.987366150527694e-05, + "loss": 0.001, + "step": 5258 + }, + { + "epoch": 1.22, + "learning_rate": 3.98717356135891e-05, + "loss": 0.0002, + "step": 5259 + }, + { + "epoch": 1.22, + "learning_rate": 3.986980972190124e-05, + "loss": 0.0005, + "step": 5260 + }, + { + "epoch": 1.22, + "learning_rate": 3.986788383021339e-05, + "loss": 0.0003, + "step": 5261 + }, + { + "epoch": 1.22, + "learning_rate": 3.986595793852554e-05, + "loss": 0.0002, + "step": 5262 + }, + { + "epoch": 1.22, + "learning_rate": 3.9864032046837694e-05, + "loss": 0.0003, + "step": 5263 + }, + { + "epoch": 1.22, + "learning_rate": 3.9862106155149835e-05, + "loss": 0.0003, + "step": 5264 + }, + { + "epoch": 1.22, + "learning_rate": 3.986018026346198e-05, + "loss": 0.0003, + "step": 5265 + }, + { + "epoch": 1.22, + "learning_rate": 3.9858254371774136e-05, + "loss": 0.0002, + "step": 5266 + }, + { + "epoch": 1.22, + "learning_rate": 3.985632848008628e-05, + "loss": 0.0, + "step": 5267 + }, + { + "epoch": 1.22, + "learning_rate": 3.985440258839843e-05, + "loss": 0.0, + "step": 5268 + }, + { + "epoch": 1.22, + "learning_rate": 3.985247669671058e-05, + "loss": 0.0003, + "step": 5269 + }, + { + "epoch": 1.22, + "learning_rate": 3.9850550805022726e-05, + "loss": 0.0004, + "step": 5270 + }, + { + "epoch": 1.22, + "learning_rate": 3.9848624913334874e-05, + "loss": 0.0, + "step": 5271 + }, + { + "epoch": 1.22, + "learning_rate": 3.984669902164703e-05, + "loss": 0.0, + "step": 5272 + }, + { + "epoch": 1.22, + "learning_rate": 3.9844773129959175e-05, + "loss": 0.0001, + "step": 5273 + }, + { + "epoch": 1.22, + "learning_rate": 3.9842847238271316e-05, + "loss": 0.0, + "step": 5274 + }, + { + "epoch": 1.22, + "learning_rate": 3.984092134658347e-05, + "loss": 0.0003, + "step": 5275 + }, + { + "epoch": 1.22, + "learning_rate": 3.983899545489562e-05, + "loss": 0.0016, + "step": 5276 + }, + { + "epoch": 1.22, + "learning_rate": 3.9837069563207765e-05, + "loss": 0.0001, + "step": 5277 + }, + { + "epoch": 1.22, + "learning_rate": 3.983514367151991e-05, + "loss": 0.0003, + "step": 5278 + }, + { + "epoch": 1.22, + "learning_rate": 3.983321777983207e-05, + "loss": 0.0002, + "step": 5279 + }, + { + "epoch": 1.22, + "learning_rate": 3.9831291888144214e-05, + "loss": 0.0002, + "step": 5280 + }, + { + "epoch": 1.22, + "learning_rate": 3.982936599645636e-05, + "loss": 0.0, + "step": 5281 + }, + { + "epoch": 1.22, + "learning_rate": 3.982744010476851e-05, + "loss": 0.0, + "step": 5282 + }, + { + "epoch": 1.22, + "learning_rate": 3.982551421308066e-05, + "loss": 0.0007, + "step": 5283 + }, + { + "epoch": 1.22, + "learning_rate": 3.9823588321392804e-05, + "loss": 0.0, + "step": 5284 + }, + { + "epoch": 1.22, + "learning_rate": 3.982166242970495e-05, + "loss": 0.0157, + "step": 5285 + }, + { + "epoch": 1.22, + "learning_rate": 3.9819736538017106e-05, + "loss": 0.0022, + "step": 5286 + }, + { + "epoch": 1.22, + "learning_rate": 3.9817810646329254e-05, + "loss": 0.0002, + "step": 5287 + }, + { + "epoch": 1.22, + "learning_rate": 3.98158847546414e-05, + "loss": 0.0, + "step": 5288 + }, + { + "epoch": 1.22, + "learning_rate": 3.981395886295355e-05, + "loss": 0.0007, + "step": 5289 + }, + { + "epoch": 1.22, + "learning_rate": 3.98120329712657e-05, + "loss": 0.0001, + "step": 5290 + }, + { + "epoch": 1.22, + "learning_rate": 3.9810107079577844e-05, + "loss": 0.0001, + "step": 5291 + }, + { + "epoch": 1.22, + "learning_rate": 3.980818118788999e-05, + "loss": 0.0002, + "step": 5292 + }, + { + "epoch": 1.22, + "learning_rate": 3.9806255296202145e-05, + "loss": 0.0001, + "step": 5293 + }, + { + "epoch": 1.22, + "learning_rate": 3.980432940451429e-05, + "loss": 0.0, + "step": 5294 + }, + { + "epoch": 1.22, + "learning_rate": 3.980240351282644e-05, + "loss": 0.0026, + "step": 5295 + }, + { + "epoch": 1.22, + "learning_rate": 3.980047762113859e-05, + "loss": 0.0, + "step": 5296 + }, + { + "epoch": 1.22, + "learning_rate": 3.979855172945074e-05, + "loss": 0.0001, + "step": 5297 + }, + { + "epoch": 1.22, + "learning_rate": 3.979662583776288e-05, + "loss": 0.0107, + "step": 5298 + }, + { + "epoch": 1.22, + "learning_rate": 3.979469994607504e-05, + "loss": 0.0003, + "step": 5299 + }, + { + "epoch": 1.22, + "learning_rate": 3.9792774054387184e-05, + "loss": 0.0, + "step": 5300 + }, + { + "epoch": 1.23, + "learning_rate": 3.979084816269933e-05, + "loss": 0.0002, + "step": 5301 + }, + { + "epoch": 1.23, + "learning_rate": 3.978892227101148e-05, + "loss": 0.0002, + "step": 5302 + }, + { + "epoch": 1.23, + "learning_rate": 3.978699637932363e-05, + "loss": 0.0002, + "step": 5303 + }, + { + "epoch": 1.23, + "learning_rate": 3.978507048763578e-05, + "loss": 0.0, + "step": 5304 + }, + { + "epoch": 1.23, + "learning_rate": 3.978314459594792e-05, + "loss": 0.0, + "step": 5305 + }, + { + "epoch": 1.23, + "learning_rate": 3.9781218704260076e-05, + "loss": 0.0, + "step": 5306 + }, + { + "epoch": 1.23, + "learning_rate": 3.9779292812572223e-05, + "loss": 0.0, + "step": 5307 + }, + { + "epoch": 1.23, + "learning_rate": 3.977736692088437e-05, + "loss": 0.0, + "step": 5308 + }, + { + "epoch": 1.23, + "learning_rate": 3.977544102919652e-05, + "loss": 0.0, + "step": 5309 + }, + { + "epoch": 1.23, + "learning_rate": 3.9773515137508666e-05, + "loss": 0.0007, + "step": 5310 + }, + { + "epoch": 1.23, + "learning_rate": 3.977158924582082e-05, + "loss": 0.0, + "step": 5311 + }, + { + "epoch": 1.23, + "learning_rate": 3.976966335413296e-05, + "loss": 0.0, + "step": 5312 + }, + { + "epoch": 1.23, + "learning_rate": 3.9767737462445115e-05, + "loss": 0.0006, + "step": 5313 + }, + { + "epoch": 1.23, + "learning_rate": 3.976581157075726e-05, + "loss": 0.0, + "step": 5314 + }, + { + "epoch": 1.23, + "learning_rate": 3.976388567906941e-05, + "loss": 0.0003, + "step": 5315 + }, + { + "epoch": 1.23, + "learning_rate": 3.976195978738156e-05, + "loss": 0.0001, + "step": 5316 + }, + { + "epoch": 1.23, + "learning_rate": 3.976003389569371e-05, + "loss": 0.0, + "step": 5317 + }, + { + "epoch": 1.23, + "learning_rate": 3.975810800400586e-05, + "loss": 0.0, + "step": 5318 + }, + { + "epoch": 1.23, + "learning_rate": 3.975618211231801e-05, + "loss": 0.0, + "step": 5319 + }, + { + "epoch": 1.23, + "learning_rate": 3.9754256220630154e-05, + "loss": 0.0001, + "step": 5320 + }, + { + "epoch": 1.23, + "learning_rate": 3.97523303289423e-05, + "loss": 0.0001, + "step": 5321 + }, + { + "epoch": 1.23, + "learning_rate": 3.975040443725445e-05, + "loss": 0.001, + "step": 5322 + }, + { + "epoch": 1.23, + "learning_rate": 3.97484785455666e-05, + "loss": 0.0004, + "step": 5323 + }, + { + "epoch": 1.23, + "learning_rate": 3.974655265387875e-05, + "loss": 0.0, + "step": 5324 + }, + { + "epoch": 1.23, + "learning_rate": 3.97446267621909e-05, + "loss": 0.0001, + "step": 5325 + }, + { + "epoch": 1.23, + "learning_rate": 3.9742700870503046e-05, + "loss": 0.0001, + "step": 5326 + }, + { + "epoch": 1.23, + "learning_rate": 3.974077497881519e-05, + "loss": 0.0, + "step": 5327 + }, + { + "epoch": 1.23, + "learning_rate": 3.973884908712735e-05, + "loss": 0.001, + "step": 5328 + }, + { + "epoch": 1.23, + "learning_rate": 3.973692319543949e-05, + "loss": 0.0005, + "step": 5329 + }, + { + "epoch": 1.23, + "learning_rate": 3.9734997303751636e-05, + "loss": 0.0003, + "step": 5330 + }, + { + "epoch": 1.23, + "learning_rate": 3.973307141206379e-05, + "loss": 0.0003, + "step": 5331 + }, + { + "epoch": 1.23, + "learning_rate": 3.973114552037594e-05, + "loss": 0.0003, + "step": 5332 + }, + { + "epoch": 1.23, + "learning_rate": 3.9729219628688085e-05, + "loss": 0.0005, + "step": 5333 + }, + { + "epoch": 1.23, + "learning_rate": 3.972729373700023e-05, + "loss": 0.0002, + "step": 5334 + }, + { + "epoch": 1.23, + "learning_rate": 3.972536784531239e-05, + "loss": 0.0, + "step": 5335 + }, + { + "epoch": 1.23, + "learning_rate": 3.972344195362453e-05, + "loss": 0.0002, + "step": 5336 + }, + { + "epoch": 1.23, + "learning_rate": 3.972151606193668e-05, + "loss": 0.0, + "step": 5337 + }, + { + "epoch": 1.23, + "learning_rate": 3.971959017024883e-05, + "loss": 0.0, + "step": 5338 + }, + { + "epoch": 1.23, + "learning_rate": 3.971766427856097e-05, + "loss": 0.0001, + "step": 5339 + }, + { + "epoch": 1.23, + "learning_rate": 3.9715738386873124e-05, + "loss": 0.0005, + "step": 5340 + }, + { + "epoch": 1.23, + "learning_rate": 3.971381249518527e-05, + "loss": 0.0001, + "step": 5341 + }, + { + "epoch": 1.23, + "learning_rate": 3.971188660349742e-05, + "loss": 0.0, + "step": 5342 + }, + { + "epoch": 1.23, + "learning_rate": 3.9709960711809567e-05, + "loss": 0.0001, + "step": 5343 + }, + { + "epoch": 1.24, + "learning_rate": 3.970803482012172e-05, + "loss": 0.0, + "step": 5344 + }, + { + "epoch": 1.24, + "learning_rate": 3.970610892843387e-05, + "loss": 0.0035, + "step": 5345 + }, + { + "epoch": 1.24, + "learning_rate": 3.9704183036746016e-05, + "loss": 0.0005, + "step": 5346 + }, + { + "epoch": 1.24, + "learning_rate": 3.970225714505816e-05, + "loss": 0.0001, + "step": 5347 + }, + { + "epoch": 1.24, + "learning_rate": 3.970033125337031e-05, + "loss": 0.0007, + "step": 5348 + }, + { + "epoch": 1.24, + "learning_rate": 3.969840536168246e-05, + "loss": 0.0, + "step": 5349 + }, + { + "epoch": 1.24, + "learning_rate": 3.9696479469994606e-05, + "loss": 0.0018, + "step": 5350 + }, + { + "epoch": 1.24, + "learning_rate": 3.969455357830676e-05, + "loss": 0.0001, + "step": 5351 + }, + { + "epoch": 1.24, + "learning_rate": 3.969262768661891e-05, + "loss": 0.001, + "step": 5352 + }, + { + "epoch": 1.24, + "learning_rate": 3.9690701794931055e-05, + "loss": 0.0026, + "step": 5353 + }, + { + "epoch": 1.24, + "learning_rate": 3.96887759032432e-05, + "loss": 0.0, + "step": 5354 + }, + { + "epoch": 1.24, + "learning_rate": 3.968685001155536e-05, + "loss": 0.0001, + "step": 5355 + }, + { + "epoch": 1.24, + "learning_rate": 3.96849241198675e-05, + "loss": 0.0016, + "step": 5356 + }, + { + "epoch": 1.24, + "learning_rate": 3.9682998228179645e-05, + "loss": 0.0003, + "step": 5357 + }, + { + "epoch": 1.24, + "learning_rate": 3.96810723364918e-05, + "loss": 0.006, + "step": 5358 + }, + { + "epoch": 1.24, + "learning_rate": 3.9679146444803947e-05, + "loss": 0.0, + "step": 5359 + }, + { + "epoch": 1.24, + "learning_rate": 3.9677220553116094e-05, + "loss": 0.0005, + "step": 5360 + }, + { + "epoch": 1.24, + "learning_rate": 3.967529466142824e-05, + "loss": 0.0, + "step": 5361 + }, + { + "epoch": 1.24, + "learning_rate": 3.9673368769740396e-05, + "loss": 0.0, + "step": 5362 + }, + { + "epoch": 1.24, + "learning_rate": 3.9671442878052536e-05, + "loss": 0.0, + "step": 5363 + }, + { + "epoch": 1.24, + "learning_rate": 3.966951698636469e-05, + "loss": 0.001, + "step": 5364 + }, + { + "epoch": 1.24, + "learning_rate": 3.966759109467684e-05, + "loss": 0.0, + "step": 5365 + }, + { + "epoch": 1.24, + "learning_rate": 3.9665665202988986e-05, + "loss": 0.0003, + "step": 5366 + }, + { + "epoch": 1.24, + "learning_rate": 3.966373931130113e-05, + "loss": 0.0, + "step": 5367 + }, + { + "epoch": 1.24, + "learning_rate": 3.966181341961328e-05, + "loss": 0.0, + "step": 5368 + }, + { + "epoch": 1.24, + "learning_rate": 3.9659887527925435e-05, + "loss": 0.0, + "step": 5369 + }, + { + "epoch": 1.24, + "learning_rate": 3.9657961636237576e-05, + "loss": 0.0, + "step": 5370 + }, + { + "epoch": 1.24, + "learning_rate": 3.965603574454973e-05, + "loss": 0.0002, + "step": 5371 + }, + { + "epoch": 1.24, + "learning_rate": 3.965410985286188e-05, + "loss": 0.0, + "step": 5372 + }, + { + "epoch": 1.24, + "learning_rate": 3.9652183961174025e-05, + "loss": 0.0058, + "step": 5373 + }, + { + "epoch": 1.24, + "learning_rate": 3.965025806948617e-05, + "loss": 0.0, + "step": 5374 + }, + { + "epoch": 1.24, + "learning_rate": 3.9648332177798327e-05, + "loss": 0.0, + "step": 5375 + }, + { + "epoch": 1.24, + "learning_rate": 3.9646406286110474e-05, + "loss": 0.0, + "step": 5376 + }, + { + "epoch": 1.24, + "learning_rate": 3.9644480394422615e-05, + "loss": 0.0001, + "step": 5377 + }, + { + "epoch": 1.24, + "learning_rate": 3.964255450273477e-05, + "loss": 0.0, + "step": 5378 + }, + { + "epoch": 1.24, + "learning_rate": 3.9640628611046916e-05, + "loss": 0.0015, + "step": 5379 + }, + { + "epoch": 1.24, + "learning_rate": 3.9638702719359064e-05, + "loss": 0.0001, + "step": 5380 + }, + { + "epoch": 1.24, + "learning_rate": 3.963677682767121e-05, + "loss": 0.0002, + "step": 5381 + }, + { + "epoch": 1.24, + "learning_rate": 3.9634850935983366e-05, + "loss": 0.0001, + "step": 5382 + }, + { + "epoch": 1.24, + "learning_rate": 3.963292504429551e-05, + "loss": 0.0002, + "step": 5383 + }, + { + "epoch": 1.24, + "learning_rate": 3.963099915260766e-05, + "loss": 0.0002, + "step": 5384 + }, + { + "epoch": 1.24, + "learning_rate": 3.962907326091981e-05, + "loss": 0.0, + "step": 5385 + }, + { + "epoch": 1.24, + "learning_rate": 3.9627147369231956e-05, + "loss": 0.0, + "step": 5386 + }, + { + "epoch": 1.24, + "learning_rate": 3.96252214775441e-05, + "loss": 0.0, + "step": 5387 + }, + { + "epoch": 1.25, + "learning_rate": 3.962329558585625e-05, + "loss": 0.0, + "step": 5388 + }, + { + "epoch": 1.25, + "learning_rate": 3.9621369694168405e-05, + "loss": 0.0, + "step": 5389 + }, + { + "epoch": 1.25, + "learning_rate": 3.961944380248055e-05, + "loss": 0.0002, + "step": 5390 + }, + { + "epoch": 1.25, + "learning_rate": 3.96175179107927e-05, + "loss": 0.0026, + "step": 5391 + }, + { + "epoch": 1.25, + "learning_rate": 3.961559201910485e-05, + "loss": 0.0, + "step": 5392 + }, + { + "epoch": 1.25, + "learning_rate": 3.9613666127417e-05, + "loss": 0.0001, + "step": 5393 + }, + { + "epoch": 1.25, + "learning_rate": 3.961174023572914e-05, + "loss": 0.0009, + "step": 5394 + }, + { + "epoch": 1.25, + "learning_rate": 3.960981434404129e-05, + "loss": 0.0008, + "step": 5395 + }, + { + "epoch": 1.25, + "learning_rate": 3.9607888452353444e-05, + "loss": 0.0, + "step": 5396 + }, + { + "epoch": 1.25, + "learning_rate": 3.960596256066559e-05, + "loss": 0.0079, + "step": 5397 + }, + { + "epoch": 1.25, + "learning_rate": 3.960403666897774e-05, + "loss": 0.0, + "step": 5398 + }, + { + "epoch": 1.25, + "learning_rate": 3.9602110777289886e-05, + "loss": 0.0, + "step": 5399 + }, + { + "epoch": 1.25, + "learning_rate": 3.960018488560204e-05, + "loss": 0.0, + "step": 5400 + }, + { + "epoch": 1.25, + "learning_rate": 3.959825899391418e-05, + "loss": 0.0376, + "step": 5401 + }, + { + "epoch": 1.25, + "learning_rate": 3.9596333102226336e-05, + "loss": 0.0, + "step": 5402 + }, + { + "epoch": 1.25, + "learning_rate": 3.959440721053848e-05, + "loss": 0.0, + "step": 5403 + }, + { + "epoch": 1.25, + "learning_rate": 3.959248131885063e-05, + "loss": 0.0001, + "step": 5404 + }, + { + "epoch": 1.25, + "learning_rate": 3.959055542716278e-05, + "loss": 0.0001, + "step": 5405 + }, + { + "epoch": 1.25, + "learning_rate": 3.9588629535474925e-05, + "loss": 0.0, + "step": 5406 + }, + { + "epoch": 1.25, + "learning_rate": 3.958670364378708e-05, + "loss": 0.0001, + "step": 5407 + }, + { + "epoch": 1.25, + "learning_rate": 3.958477775209922e-05, + "loss": 0.0, + "step": 5408 + }, + { + "epoch": 1.25, + "learning_rate": 3.9582851860411375e-05, + "loss": 0.0001, + "step": 5409 + }, + { + "epoch": 1.25, + "learning_rate": 3.958092596872352e-05, + "loss": 0.0008, + "step": 5410 + }, + { + "epoch": 1.25, + "learning_rate": 3.957900007703567e-05, + "loss": 0.001, + "step": 5411 + }, + { + "epoch": 1.25, + "learning_rate": 3.957707418534782e-05, + "loss": 0.0006, + "step": 5412 + }, + { + "epoch": 1.25, + "learning_rate": 3.9575148293659965e-05, + "loss": 0.0, + "step": 5413 + }, + { + "epoch": 1.25, + "learning_rate": 3.957322240197211e-05, + "loss": 0.0, + "step": 5414 + }, + { + "epoch": 1.25, + "learning_rate": 3.957129651028426e-05, + "loss": 0.0001, + "step": 5415 + }, + { + "epoch": 1.25, + "learning_rate": 3.9569370618596414e-05, + "loss": 0.0, + "step": 5416 + }, + { + "epoch": 1.25, + "learning_rate": 3.956744472690856e-05, + "loss": 0.0, + "step": 5417 + }, + { + "epoch": 1.25, + "learning_rate": 3.956551883522071e-05, + "loss": 0.0, + "step": 5418 + }, + { + "epoch": 1.25, + "learning_rate": 3.9563592943532856e-05, + "loss": 0.0003, + "step": 5419 + }, + { + "epoch": 1.25, + "learning_rate": 3.956166705184501e-05, + "loss": 0.0003, + "step": 5420 + }, + { + "epoch": 1.25, + "learning_rate": 3.955974116015715e-05, + "loss": 0.0, + "step": 5421 + }, + { + "epoch": 1.25, + "learning_rate": 3.95578152684693e-05, + "loss": 0.0001, + "step": 5422 + }, + { + "epoch": 1.25, + "learning_rate": 3.955588937678145e-05, + "loss": 0.0118, + "step": 5423 + }, + { + "epoch": 1.25, + "learning_rate": 3.95539634850936e-05, + "loss": 0.0, + "step": 5424 + }, + { + "epoch": 1.25, + "learning_rate": 3.955203759340575e-05, + "loss": 0.0, + "step": 5425 + }, + { + "epoch": 1.25, + "learning_rate": 3.9550111701717895e-05, + "loss": 0.0002, + "step": 5426 + }, + { + "epoch": 1.25, + "learning_rate": 3.954818581003005e-05, + "loss": 0.0, + "step": 5427 + }, + { + "epoch": 1.25, + "learning_rate": 3.954625991834219e-05, + "loss": 0.0002, + "step": 5428 + }, + { + "epoch": 1.25, + "learning_rate": 3.9544334026654345e-05, + "loss": 0.0001, + "step": 5429 + }, + { + "epoch": 1.25, + "learning_rate": 3.954240813496649e-05, + "loss": 0.0006, + "step": 5430 + }, + { + "epoch": 1.26, + "learning_rate": 3.954048224327864e-05, + "loss": 0.0008, + "step": 5431 + }, + { + "epoch": 1.26, + "learning_rate": 3.953855635159079e-05, + "loss": 0.0006, + "step": 5432 + }, + { + "epoch": 1.26, + "learning_rate": 3.9536630459902934e-05, + "loss": 0.0001, + "step": 5433 + }, + { + "epoch": 1.26, + "learning_rate": 3.953470456821509e-05, + "loss": 0.0111, + "step": 5434 + }, + { + "epoch": 1.26, + "learning_rate": 3.953277867652723e-05, + "loss": 0.0, + "step": 5435 + }, + { + "epoch": 1.26, + "learning_rate": 3.9530852784839384e-05, + "loss": 0.0001, + "step": 5436 + }, + { + "epoch": 1.26, + "learning_rate": 3.952892689315153e-05, + "loss": 0.0001, + "step": 5437 + }, + { + "epoch": 1.26, + "learning_rate": 3.952700100146368e-05, + "loss": 0.0002, + "step": 5438 + }, + { + "epoch": 1.26, + "learning_rate": 3.9525075109775826e-05, + "loss": 0.0004, + "step": 5439 + }, + { + "epoch": 1.26, + "learning_rate": 3.952314921808798e-05, + "loss": 0.0, + "step": 5440 + }, + { + "epoch": 1.26, + "learning_rate": 3.952122332640013e-05, + "loss": 0.0, + "step": 5441 + }, + { + "epoch": 1.26, + "learning_rate": 3.951929743471227e-05, + "loss": 0.0058, + "step": 5442 + }, + { + "epoch": 1.26, + "learning_rate": 3.951737154302442e-05, + "loss": 0.0022, + "step": 5443 + }, + { + "epoch": 1.26, + "learning_rate": 3.951544565133657e-05, + "loss": 0.0, + "step": 5444 + }, + { + "epoch": 1.26, + "learning_rate": 3.951351975964872e-05, + "loss": 0.0, + "step": 5445 + }, + { + "epoch": 1.26, + "learning_rate": 3.9511593867960865e-05, + "loss": 0.0, + "step": 5446 + }, + { + "epoch": 1.26, + "learning_rate": 3.950966797627302e-05, + "loss": 0.0037, + "step": 5447 + }, + { + "epoch": 1.26, + "learning_rate": 3.950774208458517e-05, + "loss": 0.0001, + "step": 5448 + }, + { + "epoch": 1.26, + "learning_rate": 3.9505816192897314e-05, + "loss": 0.0011, + "step": 5449 + }, + { + "epoch": 1.26, + "learning_rate": 3.950389030120946e-05, + "loss": 0.0001, + "step": 5450 + }, + { + "epoch": 1.26, + "learning_rate": 3.950196440952161e-05, + "loss": 0.0, + "step": 5451 + }, + { + "epoch": 1.26, + "learning_rate": 3.950003851783376e-05, + "loss": 0.0, + "step": 5452 + }, + { + "epoch": 1.26, + "learning_rate": 3.9498112626145904e-05, + "loss": 0.0, + "step": 5453 + }, + { + "epoch": 1.26, + "learning_rate": 3.949618673445806e-05, + "loss": 0.0001, + "step": 5454 + }, + { + "epoch": 1.26, + "learning_rate": 3.9494260842770206e-05, + "loss": 0.0, + "step": 5455 + }, + { + "epoch": 1.26, + "learning_rate": 3.9492334951082354e-05, + "loss": 0.0003, + "step": 5456 + }, + { + "epoch": 1.26, + "learning_rate": 3.94904090593945e-05, + "loss": 0.0203, + "step": 5457 + }, + { + "epoch": 1.26, + "learning_rate": 3.9488483167706655e-05, + "loss": 0.0002, + "step": 5458 + }, + { + "epoch": 1.26, + "learning_rate": 3.9486557276018796e-05, + "loss": 0.0001, + "step": 5459 + }, + { + "epoch": 1.26, + "learning_rate": 3.9484631384330943e-05, + "loss": 0.0007, + "step": 5460 + }, + { + "epoch": 1.26, + "learning_rate": 3.94827054926431e-05, + "loss": 0.0002, + "step": 5461 + }, + { + "epoch": 1.26, + "learning_rate": 3.9480779600955245e-05, + "loss": 0.0003, + "step": 5462 + }, + { + "epoch": 1.26, + "learning_rate": 3.947885370926739e-05, + "loss": 0.0001, + "step": 5463 + }, + { + "epoch": 1.26, + "learning_rate": 3.947692781757954e-05, + "loss": 0.0067, + "step": 5464 + }, + { + "epoch": 1.26, + "learning_rate": 3.9475001925891694e-05, + "loss": 0.0, + "step": 5465 + }, + { + "epoch": 1.26, + "learning_rate": 3.9473076034203835e-05, + "loss": 0.0001, + "step": 5466 + }, + { + "epoch": 1.26, + "learning_rate": 3.947115014251599e-05, + "loss": 0.0, + "step": 5467 + }, + { + "epoch": 1.26, + "learning_rate": 3.946922425082814e-05, + "loss": 0.0005, + "step": 5468 + }, + { + "epoch": 1.26, + "learning_rate": 3.9467298359140284e-05, + "loss": 0.0011, + "step": 5469 + }, + { + "epoch": 1.26, + "learning_rate": 3.946537246745243e-05, + "loss": 0.0, + "step": 5470 + }, + { + "epoch": 1.26, + "learning_rate": 3.946344657576458e-05, + "loss": 0.0, + "step": 5471 + }, + { + "epoch": 1.26, + "learning_rate": 3.9461520684076734e-05, + "loss": 0.0001, + "step": 5472 + }, + { + "epoch": 1.26, + "learning_rate": 3.9459594792388874e-05, + "loss": 0.0, + "step": 5473 + }, + { + "epoch": 1.27, + "learning_rate": 3.945766890070103e-05, + "loss": 0.0, + "step": 5474 + }, + { + "epoch": 1.27, + "learning_rate": 3.9455743009013176e-05, + "loss": 0.0, + "step": 5475 + }, + { + "epoch": 1.27, + "learning_rate": 3.9453817117325323e-05, + "loss": 0.0011, + "step": 5476 + }, + { + "epoch": 1.27, + "learning_rate": 3.945189122563747e-05, + "loss": 0.0001, + "step": 5477 + }, + { + "epoch": 1.27, + "learning_rate": 3.944996533394962e-05, + "loss": 0.0001, + "step": 5478 + }, + { + "epoch": 1.27, + "learning_rate": 3.944803944226177e-05, + "loss": 0.0017, + "step": 5479 + }, + { + "epoch": 1.27, + "learning_rate": 3.944611355057391e-05, + "loss": 0.0, + "step": 5480 + }, + { + "epoch": 1.27, + "learning_rate": 3.944418765888607e-05, + "loss": 0.0002, + "step": 5481 + }, + { + "epoch": 1.27, + "learning_rate": 3.9442261767198215e-05, + "loss": 0.0001, + "step": 5482 + }, + { + "epoch": 1.27, + "learning_rate": 3.944033587551036e-05, + "loss": 0.0, + "step": 5483 + }, + { + "epoch": 1.27, + "learning_rate": 3.943840998382251e-05, + "loss": 0.0001, + "step": 5484 + }, + { + "epoch": 1.27, + "learning_rate": 3.9436484092134664e-05, + "loss": 0.0002, + "step": 5485 + }, + { + "epoch": 1.27, + "learning_rate": 3.943455820044681e-05, + "loss": 0.0001, + "step": 5486 + }, + { + "epoch": 1.27, + "learning_rate": 3.943263230875896e-05, + "loss": 0.0018, + "step": 5487 + }, + { + "epoch": 1.27, + "learning_rate": 3.943070641707111e-05, + "loss": 0.0008, + "step": 5488 + }, + { + "epoch": 1.27, + "learning_rate": 3.9428780525383254e-05, + "loss": 0.0002, + "step": 5489 + }, + { + "epoch": 1.27, + "learning_rate": 3.94268546336954e-05, + "loss": 0.0003, + "step": 5490 + }, + { + "epoch": 1.27, + "learning_rate": 3.942492874200755e-05, + "loss": 0.0, + "step": 5491 + }, + { + "epoch": 1.27, + "learning_rate": 3.94230028503197e-05, + "loss": 0.0001, + "step": 5492 + }, + { + "epoch": 1.27, + "learning_rate": 3.9421076958631844e-05, + "loss": 0.0001, + "step": 5493 + }, + { + "epoch": 1.27, + "learning_rate": 3.9419151066944e-05, + "loss": 0.0, + "step": 5494 + }, + { + "epoch": 1.27, + "learning_rate": 3.9417225175256146e-05, + "loss": 0.0008, + "step": 5495 + }, + { + "epoch": 1.27, + "learning_rate": 3.941529928356829e-05, + "loss": 0.0003, + "step": 5496 + }, + { + "epoch": 1.27, + "learning_rate": 3.941337339188044e-05, + "loss": 0.0001, + "step": 5497 + }, + { + "epoch": 1.27, + "learning_rate": 3.941144750019259e-05, + "loss": 0.0, + "step": 5498 + }, + { + "epoch": 1.27, + "learning_rate": 3.940952160850474e-05, + "loss": 0.0001, + "step": 5499 + }, + { + "epoch": 1.27, + "learning_rate": 3.940759571681688e-05, + "loss": 0.0001, + "step": 5500 + }, + { + "epoch": 1.27, + "learning_rate": 3.940566982512904e-05, + "loss": 0.0012, + "step": 5501 + }, + { + "epoch": 1.27, + "learning_rate": 3.9403743933441185e-05, + "loss": 0.0002, + "step": 5502 + }, + { + "epoch": 1.27, + "learning_rate": 3.940181804175333e-05, + "loss": 0.0049, + "step": 5503 + }, + { + "epoch": 1.27, + "learning_rate": 3.939989215006548e-05, + "loss": 0.0, + "step": 5504 + }, + { + "epoch": 1.27, + "learning_rate": 3.9397966258377634e-05, + "loss": 0.0001, + "step": 5505 + }, + { + "epoch": 1.27, + "learning_rate": 3.939604036668978e-05, + "loss": 0.0, + "step": 5506 + }, + { + "epoch": 1.27, + "learning_rate": 3.939411447500192e-05, + "loss": 0.0, + "step": 5507 + }, + { + "epoch": 1.27, + "learning_rate": 3.9392188583314077e-05, + "loss": 0.0002, + "step": 5508 + }, + { + "epoch": 1.27, + "learning_rate": 3.9390262691626224e-05, + "loss": 0.0, + "step": 5509 + }, + { + "epoch": 1.27, + "learning_rate": 3.938833679993837e-05, + "loss": 0.0001, + "step": 5510 + }, + { + "epoch": 1.27, + "learning_rate": 3.938641090825052e-05, + "loss": 0.0003, + "step": 5511 + }, + { + "epoch": 1.27, + "learning_rate": 3.938448501656267e-05, + "loss": 0.0001, + "step": 5512 + }, + { + "epoch": 1.27, + "learning_rate": 3.938255912487482e-05, + "loss": 0.0, + "step": 5513 + }, + { + "epoch": 1.27, + "learning_rate": 3.938063323318697e-05, + "loss": 0.0, + "step": 5514 + }, + { + "epoch": 1.27, + "learning_rate": 3.9378707341499116e-05, + "loss": 0.0, + "step": 5515 + }, + { + "epoch": 1.27, + "learning_rate": 3.937678144981126e-05, + "loss": 0.0, + "step": 5516 + }, + { + "epoch": 1.28, + "learning_rate": 3.937485555812341e-05, + "loss": 0.0001, + "step": 5517 + }, + { + "epoch": 1.28, + "learning_rate": 3.937292966643556e-05, + "loss": 0.0, + "step": 5518 + }, + { + "epoch": 1.28, + "learning_rate": 3.937100377474771e-05, + "loss": 0.0009, + "step": 5519 + }, + { + "epoch": 1.28, + "learning_rate": 3.936907788305986e-05, + "loss": 0.0, + "step": 5520 + }, + { + "epoch": 1.28, + "learning_rate": 3.936715199137201e-05, + "loss": 0.0, + "step": 5521 + }, + { + "epoch": 1.28, + "learning_rate": 3.9365226099684155e-05, + "loss": 0.0, + "step": 5522 + }, + { + "epoch": 1.28, + "learning_rate": 3.936330020799631e-05, + "loss": 0.0001, + "step": 5523 + }, + { + "epoch": 1.28, + "learning_rate": 3.936137431630845e-05, + "loss": 0.0002, + "step": 5524 + }, + { + "epoch": 1.28, + "learning_rate": 3.93594484246206e-05, + "loss": 0.0, + "step": 5525 + }, + { + "epoch": 1.28, + "learning_rate": 3.935752253293275e-05, + "loss": 0.0, + "step": 5526 + }, + { + "epoch": 1.28, + "learning_rate": 3.93555966412449e-05, + "loss": 0.0, + "step": 5527 + }, + { + "epoch": 1.28, + "learning_rate": 3.9353670749557046e-05, + "loss": 0.0, + "step": 5528 + }, + { + "epoch": 1.28, + "learning_rate": 3.9351744857869194e-05, + "loss": 0.0, + "step": 5529 + }, + { + "epoch": 1.28, + "learning_rate": 3.934981896618135e-05, + "loss": 0.0, + "step": 5530 + }, + { + "epoch": 1.28, + "learning_rate": 3.934789307449349e-05, + "loss": 0.0001, + "step": 5531 + }, + { + "epoch": 1.28, + "learning_rate": 3.934596718280564e-05, + "loss": 0.0001, + "step": 5532 + }, + { + "epoch": 1.28, + "learning_rate": 3.934404129111779e-05, + "loss": 0.0012, + "step": 5533 + }, + { + "epoch": 1.28, + "learning_rate": 3.934211539942994e-05, + "loss": 0.0, + "step": 5534 + }, + { + "epoch": 1.28, + "learning_rate": 3.9340189507742086e-05, + "loss": 0.0002, + "step": 5535 + }, + { + "epoch": 1.28, + "learning_rate": 3.933826361605423e-05, + "loss": 0.0, + "step": 5536 + }, + { + "epoch": 1.28, + "learning_rate": 3.933633772436639e-05, + "loss": 0.0, + "step": 5537 + }, + { + "epoch": 1.28, + "learning_rate": 3.933441183267853e-05, + "loss": 0.0, + "step": 5538 + }, + { + "epoch": 1.28, + "learning_rate": 3.933248594099068e-05, + "loss": 0.0, + "step": 5539 + }, + { + "epoch": 1.28, + "learning_rate": 3.933056004930283e-05, + "loss": 0.0, + "step": 5540 + }, + { + "epoch": 1.28, + "learning_rate": 3.932863415761498e-05, + "loss": 0.0005, + "step": 5541 + }, + { + "epoch": 1.28, + "learning_rate": 3.9326708265927125e-05, + "loss": 0.0001, + "step": 5542 + }, + { + "epoch": 1.28, + "learning_rate": 3.932478237423928e-05, + "loss": 0.0, + "step": 5543 + }, + { + "epoch": 1.28, + "learning_rate": 3.9322856482551426e-05, + "loss": 0.0027, + "step": 5544 + }, + { + "epoch": 1.28, + "learning_rate": 3.932093059086357e-05, + "loss": 0.0008, + "step": 5545 + }, + { + "epoch": 1.28, + "learning_rate": 3.931900469917572e-05, + "loss": 0.0, + "step": 5546 + }, + { + "epoch": 1.28, + "learning_rate": 3.931707880748787e-05, + "loss": 0.0004, + "step": 5547 + }, + { + "epoch": 1.28, + "learning_rate": 3.9315152915800016e-05, + "loss": 0.0001, + "step": 5548 + }, + { + "epoch": 1.28, + "learning_rate": 3.9313227024112164e-05, + "loss": 0.0, + "step": 5549 + }, + { + "epoch": 1.28, + "learning_rate": 3.931130113242432e-05, + "loss": 0.0003, + "step": 5550 + }, + { + "epoch": 1.28, + "learning_rate": 3.9309375240736466e-05, + "loss": 0.0377, + "step": 5551 + }, + { + "epoch": 1.28, + "learning_rate": 3.930744934904861e-05, + "loss": 0.0, + "step": 5552 + }, + { + "epoch": 1.28, + "learning_rate": 3.930552345736076e-05, + "loss": 0.0, + "step": 5553 + }, + { + "epoch": 1.28, + "learning_rate": 3.930359756567291e-05, + "loss": 0.0001, + "step": 5554 + }, + { + "epoch": 1.28, + "learning_rate": 3.9301671673985055e-05, + "loss": 0.0029, + "step": 5555 + }, + { + "epoch": 1.28, + "learning_rate": 3.92997457822972e-05, + "loss": 0.0007, + "step": 5556 + }, + { + "epoch": 1.28, + "learning_rate": 3.929781989060936e-05, + "loss": 0.0004, + "step": 5557 + }, + { + "epoch": 1.28, + "learning_rate": 3.9295893998921505e-05, + "loss": 0.0001, + "step": 5558 + }, + { + "epoch": 1.28, + "learning_rate": 3.929396810723365e-05, + "loss": 0.0012, + "step": 5559 + }, + { + "epoch": 1.28, + "learning_rate": 3.92920422155458e-05, + "loss": 0.0, + "step": 5560 + }, + { + "epoch": 1.29, + "learning_rate": 3.9290116323857954e-05, + "loss": 0.0, + "step": 5561 + }, + { + "epoch": 1.29, + "learning_rate": 3.9288190432170095e-05, + "loss": 0.0001, + "step": 5562 + }, + { + "epoch": 1.29, + "learning_rate": 3.928626454048224e-05, + "loss": 0.0, + "step": 5563 + }, + { + "epoch": 1.29, + "learning_rate": 3.9284338648794396e-05, + "loss": 0.0076, + "step": 5564 + }, + { + "epoch": 1.29, + "learning_rate": 3.928241275710654e-05, + "loss": 0.0005, + "step": 5565 + }, + { + "epoch": 1.29, + "learning_rate": 3.928048686541869e-05, + "loss": 0.0, + "step": 5566 + }, + { + "epoch": 1.29, + "learning_rate": 3.927856097373084e-05, + "loss": 0.0002, + "step": 5567 + }, + { + "epoch": 1.29, + "learning_rate": 3.9276635082042986e-05, + "loss": 0.0, + "step": 5568 + }, + { + "epoch": 1.29, + "learning_rate": 3.9274709190355134e-05, + "loss": 0.0002, + "step": 5569 + }, + { + "epoch": 1.29, + "learning_rate": 3.927278329866729e-05, + "loss": 0.0001, + "step": 5570 + }, + { + "epoch": 1.29, + "learning_rate": 3.9270857406979435e-05, + "loss": 0.0, + "step": 5571 + }, + { + "epoch": 1.29, + "learning_rate": 3.9268931515291576e-05, + "loss": 0.0002, + "step": 5572 + }, + { + "epoch": 1.29, + "learning_rate": 3.926700562360373e-05, + "loss": 0.0, + "step": 5573 + }, + { + "epoch": 1.29, + "learning_rate": 3.926507973191588e-05, + "loss": 0.0013, + "step": 5574 + }, + { + "epoch": 1.29, + "learning_rate": 3.9263153840228025e-05, + "loss": 0.0, + "step": 5575 + }, + { + "epoch": 1.29, + "learning_rate": 3.926122794854017e-05, + "loss": 0.0, + "step": 5576 + }, + { + "epoch": 1.29, + "learning_rate": 3.925930205685233e-05, + "loss": 0.0, + "step": 5577 + }, + { + "epoch": 1.29, + "learning_rate": 3.9257376165164475e-05, + "loss": 0.0008, + "step": 5578 + }, + { + "epoch": 1.29, + "learning_rate": 3.925545027347662e-05, + "loss": 0.0001, + "step": 5579 + }, + { + "epoch": 1.29, + "learning_rate": 3.925352438178877e-05, + "loss": 0.0, + "step": 5580 + }, + { + "epoch": 1.29, + "learning_rate": 3.925159849010092e-05, + "loss": 0.0, + "step": 5581 + }, + { + "epoch": 1.29, + "learning_rate": 3.9249672598413064e-05, + "loss": 0.0, + "step": 5582 + }, + { + "epoch": 1.29, + "learning_rate": 3.924774670672521e-05, + "loss": 0.0011, + "step": 5583 + }, + { + "epoch": 1.29, + "learning_rate": 3.9245820815037366e-05, + "loss": 0.0004, + "step": 5584 + }, + { + "epoch": 1.29, + "learning_rate": 3.9243894923349514e-05, + "loss": 0.0, + "step": 5585 + }, + { + "epoch": 1.29, + "learning_rate": 3.924196903166166e-05, + "loss": 0.0002, + "step": 5586 + }, + { + "epoch": 1.29, + "learning_rate": 3.924004313997381e-05, + "loss": 0.0001, + "step": 5587 + }, + { + "epoch": 1.29, + "learning_rate": 3.923811724828596e-05, + "loss": 0.0014, + "step": 5588 + }, + { + "epoch": 1.29, + "learning_rate": 3.9236191356598104e-05, + "loss": 0.0004, + "step": 5589 + }, + { + "epoch": 1.29, + "learning_rate": 3.923426546491025e-05, + "loss": 0.0034, + "step": 5590 + }, + { + "epoch": 1.29, + "learning_rate": 3.9232339573222405e-05, + "loss": 0.0008, + "step": 5591 + }, + { + "epoch": 1.29, + "learning_rate": 3.923041368153455e-05, + "loss": 0.0002, + "step": 5592 + }, + { + "epoch": 1.29, + "learning_rate": 3.92284877898467e-05, + "loss": 0.0001, + "step": 5593 + }, + { + "epoch": 1.29, + "learning_rate": 3.922656189815885e-05, + "loss": 0.0, + "step": 5594 + }, + { + "epoch": 1.29, + "learning_rate": 3.9224636006471e-05, + "loss": 0.0001, + "step": 5595 + }, + { + "epoch": 1.29, + "learning_rate": 3.922271011478314e-05, + "loss": 0.0, + "step": 5596 + }, + { + "epoch": 1.29, + "learning_rate": 3.92207842230953e-05, + "loss": 0.0002, + "step": 5597 + }, + { + "epoch": 1.29, + "learning_rate": 3.9218858331407444e-05, + "loss": 0.0, + "step": 5598 + }, + { + "epoch": 1.29, + "learning_rate": 3.921693243971959e-05, + "loss": 0.0, + "step": 5599 + }, + { + "epoch": 1.29, + "learning_rate": 3.921500654803174e-05, + "loss": 0.0001, + "step": 5600 + }, + { + "epoch": 1.29, + "learning_rate": 3.921308065634389e-05, + "loss": 0.0, + "step": 5601 + }, + { + "epoch": 1.29, + "learning_rate": 3.921115476465604e-05, + "loss": 0.0001, + "step": 5602 + }, + { + "epoch": 1.29, + "learning_rate": 3.920922887296818e-05, + "loss": 0.0, + "step": 5603 + }, + { + "epoch": 1.3, + "learning_rate": 3.9207302981280336e-05, + "loss": 0.0, + "step": 5604 + }, + { + "epoch": 1.3, + "learning_rate": 3.9205377089592484e-05, + "loss": 0.0001, + "step": 5605 + }, + { + "epoch": 1.3, + "learning_rate": 3.920345119790463e-05, + "loss": 0.0, + "step": 5606 + }, + { + "epoch": 1.3, + "learning_rate": 3.920152530621678e-05, + "loss": 0.0001, + "step": 5607 + }, + { + "epoch": 1.3, + "learning_rate": 3.919959941452893e-05, + "loss": 0.0007, + "step": 5608 + }, + { + "epoch": 1.3, + "learning_rate": 3.919767352284108e-05, + "loss": 0.0002, + "step": 5609 + }, + { + "epoch": 1.3, + "learning_rate": 3.919574763115322e-05, + "loss": 0.0, + "step": 5610 + }, + { + "epoch": 1.3, + "learning_rate": 3.9193821739465375e-05, + "loss": 0.0, + "step": 5611 + }, + { + "epoch": 1.3, + "learning_rate": 3.919189584777752e-05, + "loss": 0.0004, + "step": 5612 + }, + { + "epoch": 1.3, + "learning_rate": 3.918996995608967e-05, + "loss": 0.0, + "step": 5613 + }, + { + "epoch": 1.3, + "learning_rate": 3.918804406440182e-05, + "loss": 0.0007, + "step": 5614 + }, + { + "epoch": 1.3, + "learning_rate": 3.918611817271397e-05, + "loss": 0.0, + "step": 5615 + }, + { + "epoch": 1.3, + "learning_rate": 3.918419228102612e-05, + "loss": 0.0001, + "step": 5616 + }, + { + "epoch": 1.3, + "learning_rate": 3.918226638933827e-05, + "loss": 0.0012, + "step": 5617 + }, + { + "epoch": 1.3, + "learning_rate": 3.9180340497650414e-05, + "loss": 0.0, + "step": 5618 + }, + { + "epoch": 1.3, + "learning_rate": 3.917841460596256e-05, + "loss": 0.0002, + "step": 5619 + }, + { + "epoch": 1.3, + "learning_rate": 3.917648871427471e-05, + "loss": 0.0015, + "step": 5620 + }, + { + "epoch": 1.3, + "learning_rate": 3.917456282258686e-05, + "loss": 0.0, + "step": 5621 + }, + { + "epoch": 1.3, + "learning_rate": 3.917263693089901e-05, + "loss": 0.0, + "step": 5622 + }, + { + "epoch": 1.3, + "learning_rate": 3.917071103921116e-05, + "loss": 0.0001, + "step": 5623 + }, + { + "epoch": 1.3, + "learning_rate": 3.9168785147523306e-05, + "loss": 0.0002, + "step": 5624 + }, + { + "epoch": 1.3, + "learning_rate": 3.9166859255835453e-05, + "loss": 0.0001, + "step": 5625 + }, + { + "epoch": 1.3, + "learning_rate": 3.916493336414761e-05, + "loss": 0.0016, + "step": 5626 + }, + { + "epoch": 1.3, + "learning_rate": 3.916300747245975e-05, + "loss": 0.001, + "step": 5627 + }, + { + "epoch": 1.3, + "learning_rate": 3.9161081580771896e-05, + "loss": 0.0001, + "step": 5628 + }, + { + "epoch": 1.3, + "learning_rate": 3.915915568908405e-05, + "loss": 0.0002, + "step": 5629 + }, + { + "epoch": 1.3, + "learning_rate": 3.91572297973962e-05, + "loss": 0.0, + "step": 5630 + }, + { + "epoch": 1.3, + "learning_rate": 3.9155303905708345e-05, + "loss": 0.0, + "step": 5631 + }, + { + "epoch": 1.3, + "learning_rate": 3.915337801402049e-05, + "loss": 0.0, + "step": 5632 + }, + { + "epoch": 1.3, + "learning_rate": 3.915145212233265e-05, + "loss": 0.0, + "step": 5633 + }, + { + "epoch": 1.3, + "learning_rate": 3.914952623064479e-05, + "loss": 0.0006, + "step": 5634 + }, + { + "epoch": 1.3, + "learning_rate": 3.914760033895694e-05, + "loss": 0.002, + "step": 5635 + }, + { + "epoch": 1.3, + "learning_rate": 3.914567444726909e-05, + "loss": 0.0002, + "step": 5636 + }, + { + "epoch": 1.3, + "learning_rate": 3.914374855558123e-05, + "loss": 0.0, + "step": 5637 + }, + { + "epoch": 1.3, + "learning_rate": 3.9141822663893384e-05, + "loss": 0.0, + "step": 5638 + }, + { + "epoch": 1.3, + "learning_rate": 3.913989677220553e-05, + "loss": 0.0007, + "step": 5639 + }, + { + "epoch": 1.3, + "learning_rate": 3.9137970880517686e-05, + "loss": 0.0, + "step": 5640 + }, + { + "epoch": 1.3, + "learning_rate": 3.913604498882983e-05, + "loss": 0.0001, + "step": 5641 + }, + { + "epoch": 1.3, + "learning_rate": 3.913411909714198e-05, + "loss": 0.0, + "step": 5642 + }, + { + "epoch": 1.3, + "learning_rate": 3.913219320545413e-05, + "loss": 0.0, + "step": 5643 + }, + { + "epoch": 1.3, + "learning_rate": 3.9130267313766276e-05, + "loss": 0.0, + "step": 5644 + }, + { + "epoch": 1.3, + "learning_rate": 3.912834142207842e-05, + "loss": 0.0, + "step": 5645 + }, + { + "epoch": 1.3, + "learning_rate": 3.912641553039057e-05, + "loss": 0.0001, + "step": 5646 + }, + { + "epoch": 1.31, + "learning_rate": 3.912448963870272e-05, + "loss": 0.0002, + "step": 5647 + }, + { + "epoch": 1.31, + "learning_rate": 3.9122563747014866e-05, + "loss": 0.0, + "step": 5648 + }, + { + "epoch": 1.31, + "learning_rate": 3.912063785532702e-05, + "loss": 0.0013, + "step": 5649 + }, + { + "epoch": 1.31, + "learning_rate": 3.911871196363917e-05, + "loss": 0.0001, + "step": 5650 + }, + { + "epoch": 1.31, + "learning_rate": 3.9116786071951315e-05, + "loss": 0.0, + "step": 5651 + }, + { + "epoch": 1.31, + "learning_rate": 3.911486018026346e-05, + "loss": 0.0, + "step": 5652 + }, + { + "epoch": 1.31, + "learning_rate": 3.911293428857562e-05, + "loss": 0.0, + "step": 5653 + }, + { + "epoch": 1.31, + "learning_rate": 3.911100839688776e-05, + "loss": 0.0003, + "step": 5654 + }, + { + "epoch": 1.31, + "learning_rate": 3.910908250519991e-05, + "loss": 0.0, + "step": 5655 + }, + { + "epoch": 1.31, + "learning_rate": 3.910715661351206e-05, + "loss": 0.0002, + "step": 5656 + }, + { + "epoch": 1.31, + "learning_rate": 3.910523072182421e-05, + "loss": 0.0002, + "step": 5657 + }, + { + "epoch": 1.31, + "learning_rate": 3.9103304830136354e-05, + "loss": 0.0004, + "step": 5658 + }, + { + "epoch": 1.31, + "learning_rate": 3.91013789384485e-05, + "loss": 0.0001, + "step": 5659 + }, + { + "epoch": 1.31, + "learning_rate": 3.9099453046760656e-05, + "loss": 0.0, + "step": 5660 + }, + { + "epoch": 1.31, + "learning_rate": 3.9097527155072797e-05, + "loss": 0.0006, + "step": 5661 + }, + { + "epoch": 1.31, + "learning_rate": 3.909560126338495e-05, + "loss": 0.0, + "step": 5662 + }, + { + "epoch": 1.31, + "learning_rate": 3.90936753716971e-05, + "loss": 0.0, + "step": 5663 + }, + { + "epoch": 1.31, + "learning_rate": 3.9091749480009246e-05, + "loss": 0.0, + "step": 5664 + }, + { + "epoch": 1.31, + "learning_rate": 3.908982358832139e-05, + "loss": 0.0001, + "step": 5665 + }, + { + "epoch": 1.31, + "learning_rate": 3.908789769663354e-05, + "loss": 0.0001, + "step": 5666 + }, + { + "epoch": 1.31, + "learning_rate": 3.9085971804945695e-05, + "loss": 0.0005, + "step": 5667 + }, + { + "epoch": 1.31, + "learning_rate": 3.9084045913257836e-05, + "loss": 0.0002, + "step": 5668 + }, + { + "epoch": 1.31, + "learning_rate": 3.908212002156999e-05, + "loss": 0.0, + "step": 5669 + }, + { + "epoch": 1.31, + "learning_rate": 3.908019412988214e-05, + "loss": 0.0002, + "step": 5670 + }, + { + "epoch": 1.31, + "learning_rate": 3.9078268238194285e-05, + "loss": 0.0, + "step": 5671 + }, + { + "epoch": 1.31, + "learning_rate": 3.907634234650643e-05, + "loss": 0.0006, + "step": 5672 + }, + { + "epoch": 1.31, + "learning_rate": 3.9074416454818587e-05, + "loss": 0.0, + "step": 5673 + }, + { + "epoch": 1.31, + "learning_rate": 3.9072490563130734e-05, + "loss": 0.0, + "step": 5674 + }, + { + "epoch": 1.31, + "learning_rate": 3.9070564671442875e-05, + "loss": 0.0007, + "step": 5675 + }, + { + "epoch": 1.31, + "learning_rate": 3.906863877975503e-05, + "loss": 0.0003, + "step": 5676 + }, + { + "epoch": 1.31, + "learning_rate": 3.9066712888067177e-05, + "loss": 0.0, + "step": 5677 + }, + { + "epoch": 1.31, + "learning_rate": 3.9064786996379324e-05, + "loss": 0.0001, + "step": 5678 + }, + { + "epoch": 1.31, + "learning_rate": 3.906286110469147e-05, + "loss": 0.0, + "step": 5679 + }, + { + "epoch": 1.31, + "learning_rate": 3.9060935213003626e-05, + "loss": 0.0, + "step": 5680 + }, + { + "epoch": 1.31, + "learning_rate": 3.905900932131577e-05, + "loss": 0.0001, + "step": 5681 + }, + { + "epoch": 1.31, + "learning_rate": 3.905708342962792e-05, + "loss": 0.0011, + "step": 5682 + }, + { + "epoch": 1.31, + "learning_rate": 3.905515753794007e-05, + "loss": 0.0, + "step": 5683 + }, + { + "epoch": 1.31, + "learning_rate": 3.9053231646252216e-05, + "loss": 0.0002, + "step": 5684 + }, + { + "epoch": 1.31, + "learning_rate": 3.905130575456436e-05, + "loss": 0.0006, + "step": 5685 + }, + { + "epoch": 1.31, + "learning_rate": 3.904937986287651e-05, + "loss": 0.0004, + "step": 5686 + }, + { + "epoch": 1.31, + "learning_rate": 3.9047453971188665e-05, + "loss": 0.0011, + "step": 5687 + }, + { + "epoch": 1.31, + "learning_rate": 3.904552807950081e-05, + "loss": 0.0005, + "step": 5688 + }, + { + "epoch": 1.31, + "learning_rate": 3.904360218781296e-05, + "loss": 0.0001, + "step": 5689 + }, + { + "epoch": 1.31, + "learning_rate": 3.904167629612511e-05, + "loss": 0.0, + "step": 5690 + }, + { + "epoch": 1.32, + "learning_rate": 3.903975040443726e-05, + "loss": 0.0003, + "step": 5691 + }, + { + "epoch": 1.32, + "learning_rate": 3.90378245127494e-05, + "loss": 0.0012, + "step": 5692 + }, + { + "epoch": 1.32, + "learning_rate": 3.903589862106155e-05, + "loss": 0.0001, + "step": 5693 + }, + { + "epoch": 1.32, + "learning_rate": 3.9033972729373704e-05, + "loss": 0.0, + "step": 5694 + }, + { + "epoch": 1.32, + "learning_rate": 3.903204683768585e-05, + "loss": 0.0, + "step": 5695 + }, + { + "epoch": 1.32, + "learning_rate": 3.9030120945998e-05, + "loss": 0.0, + "step": 5696 + }, + { + "epoch": 1.32, + "learning_rate": 3.9028195054310146e-05, + "loss": 0.0, + "step": 5697 + }, + { + "epoch": 1.32, + "learning_rate": 3.90262691626223e-05, + "loss": 0.0005, + "step": 5698 + }, + { + "epoch": 1.32, + "learning_rate": 3.902434327093444e-05, + "loss": 0.0, + "step": 5699 + }, + { + "epoch": 1.32, + "learning_rate": 3.9022417379246596e-05, + "loss": 0.0003, + "step": 5700 + }, + { + "epoch": 1.32, + "learning_rate": 3.902049148755874e-05, + "loss": 0.0, + "step": 5701 + }, + { + "epoch": 1.32, + "learning_rate": 3.901856559587089e-05, + "loss": 0.0001, + "step": 5702 + }, + { + "epoch": 1.32, + "learning_rate": 3.901663970418304e-05, + "loss": 0.0, + "step": 5703 + }, + { + "epoch": 1.32, + "learning_rate": 3.9014713812495186e-05, + "loss": 0.0001, + "step": 5704 + }, + { + "epoch": 1.32, + "learning_rate": 3.901278792080734e-05, + "loss": 0.0002, + "step": 5705 + }, + { + "epoch": 1.32, + "learning_rate": 3.901086202911948e-05, + "loss": 0.0001, + "step": 5706 + }, + { + "epoch": 1.32, + "learning_rate": 3.9008936137431635e-05, + "loss": 0.0, + "step": 5707 + }, + { + "epoch": 1.32, + "learning_rate": 3.900701024574378e-05, + "loss": 0.0001, + "step": 5708 + }, + { + "epoch": 1.32, + "learning_rate": 3.900508435405593e-05, + "loss": 0.0, + "step": 5709 + }, + { + "epoch": 1.32, + "learning_rate": 3.900315846236808e-05, + "loss": 0.0003, + "step": 5710 + }, + { + "epoch": 1.32, + "learning_rate": 3.900123257068023e-05, + "loss": 0.0001, + "step": 5711 + }, + { + "epoch": 1.32, + "learning_rate": 3.899930667899238e-05, + "loss": 0.0002, + "step": 5712 + }, + { + "epoch": 1.32, + "learning_rate": 3.899738078730452e-05, + "loss": 0.0001, + "step": 5713 + }, + { + "epoch": 1.32, + "learning_rate": 3.8995454895616674e-05, + "loss": 0.0, + "step": 5714 + }, + { + "epoch": 1.32, + "learning_rate": 3.899352900392882e-05, + "loss": 0.0, + "step": 5715 + }, + { + "epoch": 1.32, + "learning_rate": 3.899160311224097e-05, + "loss": 0.0004, + "step": 5716 + }, + { + "epoch": 1.32, + "learning_rate": 3.8989677220553116e-05, + "loss": 0.0002, + "step": 5717 + }, + { + "epoch": 1.32, + "learning_rate": 3.898775132886527e-05, + "loss": 0.0001, + "step": 5718 + }, + { + "epoch": 1.32, + "learning_rate": 3.898582543717741e-05, + "loss": 0.0, + "step": 5719 + }, + { + "epoch": 1.32, + "learning_rate": 3.8983899545489565e-05, + "loss": 0.0002, + "step": 5720 + }, + { + "epoch": 1.32, + "learning_rate": 3.898197365380171e-05, + "loss": 0.0, + "step": 5721 + }, + { + "epoch": 1.32, + "learning_rate": 3.898004776211386e-05, + "loss": 0.0001, + "step": 5722 + }, + { + "epoch": 1.32, + "learning_rate": 3.897812187042601e-05, + "loss": 0.0, + "step": 5723 + }, + { + "epoch": 1.32, + "learning_rate": 3.8976195978738155e-05, + "loss": 0.0001, + "step": 5724 + }, + { + "epoch": 1.32, + "learning_rate": 3.897427008705031e-05, + "loss": 0.0, + "step": 5725 + }, + { + "epoch": 1.32, + "learning_rate": 3.897234419536245e-05, + "loss": 0.0, + "step": 5726 + }, + { + "epoch": 1.32, + "learning_rate": 3.8970418303674605e-05, + "loss": 0.0, + "step": 5727 + }, + { + "epoch": 1.32, + "learning_rate": 3.896849241198675e-05, + "loss": 0.0, + "step": 5728 + }, + { + "epoch": 1.32, + "learning_rate": 3.89665665202989e-05, + "loss": 0.0, + "step": 5729 + }, + { + "epoch": 1.32, + "learning_rate": 3.896464062861105e-05, + "loss": 0.0, + "step": 5730 + }, + { + "epoch": 1.32, + "learning_rate": 3.8962714736923195e-05, + "loss": 0.0002, + "step": 5731 + }, + { + "epoch": 1.32, + "learning_rate": 3.896078884523535e-05, + "loss": 0.0004, + "step": 5732 + }, + { + "epoch": 1.32, + "learning_rate": 3.895886295354749e-05, + "loss": 0.0009, + "step": 5733 + }, + { + "epoch": 1.33, + "learning_rate": 3.8956937061859644e-05, + "loss": 0.0, + "step": 5734 + }, + { + "epoch": 1.33, + "learning_rate": 3.895501117017179e-05, + "loss": 0.0, + "step": 5735 + }, + { + "epoch": 1.33, + "learning_rate": 3.895308527848394e-05, + "loss": 0.0002, + "step": 5736 + }, + { + "epoch": 1.33, + "learning_rate": 3.8951159386796086e-05, + "loss": 0.0, + "step": 5737 + }, + { + "epoch": 1.33, + "learning_rate": 3.894923349510824e-05, + "loss": 0.0, + "step": 5738 + }, + { + "epoch": 1.33, + "learning_rate": 3.894730760342039e-05, + "loss": 0.0003, + "step": 5739 + }, + { + "epoch": 1.33, + "learning_rate": 3.894538171173253e-05, + "loss": 0.0, + "step": 5740 + }, + { + "epoch": 1.33, + "learning_rate": 3.894345582004468e-05, + "loss": 0.0001, + "step": 5741 + }, + { + "epoch": 1.33, + "learning_rate": 3.894152992835683e-05, + "loss": 0.0, + "step": 5742 + }, + { + "epoch": 1.33, + "learning_rate": 3.893960403666898e-05, + "loss": 0.0, + "step": 5743 + }, + { + "epoch": 1.33, + "learning_rate": 3.8937678144981125e-05, + "loss": 0.0, + "step": 5744 + }, + { + "epoch": 1.33, + "learning_rate": 3.893575225329328e-05, + "loss": 0.0, + "step": 5745 + }, + { + "epoch": 1.33, + "learning_rate": 3.893382636160543e-05, + "loss": 0.0017, + "step": 5746 + }, + { + "epoch": 1.33, + "learning_rate": 3.8931900469917574e-05, + "loss": 0.0, + "step": 5747 + }, + { + "epoch": 1.33, + "learning_rate": 3.892997457822972e-05, + "loss": 0.0001, + "step": 5748 + }, + { + "epoch": 1.33, + "learning_rate": 3.892804868654187e-05, + "loss": 0.0002, + "step": 5749 + }, + { + "epoch": 1.33, + "learning_rate": 3.892612279485402e-05, + "loss": 0.0068, + "step": 5750 + }, + { + "epoch": 1.33, + "learning_rate": 3.8924196903166164e-05, + "loss": 0.0, + "step": 5751 + }, + { + "epoch": 1.33, + "learning_rate": 3.892227101147832e-05, + "loss": 0.0002, + "step": 5752 + }, + { + "epoch": 1.33, + "learning_rate": 3.8920345119790466e-05, + "loss": 0.0002, + "step": 5753 + }, + { + "epoch": 1.33, + "learning_rate": 3.8918419228102614e-05, + "loss": 0.0, + "step": 5754 + }, + { + "epoch": 1.33, + "learning_rate": 3.891649333641476e-05, + "loss": 0.0001, + "step": 5755 + }, + { + "epoch": 1.33, + "learning_rate": 3.8914567444726915e-05, + "loss": 0.0004, + "step": 5756 + }, + { + "epoch": 1.33, + "learning_rate": 3.8912641553039056e-05, + "loss": 0.0, + "step": 5757 + }, + { + "epoch": 1.33, + "learning_rate": 3.8910715661351204e-05, + "loss": 0.0, + "step": 5758 + }, + { + "epoch": 1.33, + "learning_rate": 3.890878976966336e-05, + "loss": 0.0001, + "step": 5759 + }, + { + "epoch": 1.33, + "learning_rate": 3.8906863877975505e-05, + "loss": 0.0001, + "step": 5760 + }, + { + "epoch": 1.33, + "learning_rate": 3.890493798628765e-05, + "loss": 0.0001, + "step": 5761 + }, + { + "epoch": 1.33, + "learning_rate": 3.89030120945998e-05, + "loss": 0.0008, + "step": 5762 + }, + { + "epoch": 1.33, + "learning_rate": 3.8901086202911954e-05, + "loss": 0.0, + "step": 5763 + }, + { + "epoch": 1.33, + "learning_rate": 3.8899160311224095e-05, + "loss": 0.0007, + "step": 5764 + }, + { + "epoch": 1.33, + "learning_rate": 3.889723441953625e-05, + "loss": 0.0002, + "step": 5765 + }, + { + "epoch": 1.33, + "learning_rate": 3.88953085278484e-05, + "loss": 0.0, + "step": 5766 + }, + { + "epoch": 1.33, + "learning_rate": 3.8893382636160544e-05, + "loss": 0.0, + "step": 5767 + }, + { + "epoch": 1.33, + "learning_rate": 3.889145674447269e-05, + "loss": 0.0001, + "step": 5768 + }, + { + "epoch": 1.33, + "learning_rate": 3.888953085278484e-05, + "loss": 0.0005, + "step": 5769 + }, + { + "epoch": 1.33, + "learning_rate": 3.8887604961096994e-05, + "loss": 0.0006, + "step": 5770 + }, + { + "epoch": 1.33, + "learning_rate": 3.8885679069409134e-05, + "loss": 0.0002, + "step": 5771 + }, + { + "epoch": 1.33, + "learning_rate": 3.888375317772129e-05, + "loss": 0.0, + "step": 5772 + }, + { + "epoch": 1.33, + "learning_rate": 3.8881827286033436e-05, + "loss": 0.0004, + "step": 5773 + }, + { + "epoch": 1.33, + "learning_rate": 3.8879901394345583e-05, + "loss": 0.0, + "step": 5774 + }, + { + "epoch": 1.33, + "learning_rate": 3.887797550265773e-05, + "loss": 0.0, + "step": 5775 + }, + { + "epoch": 1.33, + "learning_rate": 3.8876049610969885e-05, + "loss": 0.0, + "step": 5776 + }, + { + "epoch": 1.34, + "learning_rate": 3.887412371928203e-05, + "loss": 0.0022, + "step": 5777 + }, + { + "epoch": 1.34, + "learning_rate": 3.8872197827594173e-05, + "loss": 0.0, + "step": 5778 + }, + { + "epoch": 1.34, + "learning_rate": 3.887027193590633e-05, + "loss": 0.0, + "step": 5779 + }, + { + "epoch": 1.34, + "learning_rate": 3.8868346044218475e-05, + "loss": 0.0, + "step": 5780 + }, + { + "epoch": 1.34, + "learning_rate": 3.886642015253062e-05, + "loss": 0.0, + "step": 5781 + }, + { + "epoch": 1.34, + "learning_rate": 3.886449426084277e-05, + "loss": 0.0019, + "step": 5782 + }, + { + "epoch": 1.34, + "learning_rate": 3.8862568369154924e-05, + "loss": 0.0, + "step": 5783 + }, + { + "epoch": 1.34, + "learning_rate": 3.886064247746707e-05, + "loss": 0.0001, + "step": 5784 + }, + { + "epoch": 1.34, + "learning_rate": 3.885871658577922e-05, + "loss": 0.0001, + "step": 5785 + }, + { + "epoch": 1.34, + "learning_rate": 3.885679069409137e-05, + "loss": 0.0, + "step": 5786 + }, + { + "epoch": 1.34, + "learning_rate": 3.8854864802403514e-05, + "loss": 0.0, + "step": 5787 + }, + { + "epoch": 1.34, + "learning_rate": 3.885293891071566e-05, + "loss": 0.0002, + "step": 5788 + }, + { + "epoch": 1.34, + "learning_rate": 3.885101301902781e-05, + "loss": 0.0, + "step": 5789 + }, + { + "epoch": 1.34, + "learning_rate": 3.8849087127339963e-05, + "loss": 0.0, + "step": 5790 + }, + { + "epoch": 1.34, + "learning_rate": 3.8847161235652104e-05, + "loss": 0.001, + "step": 5791 + }, + { + "epoch": 1.34, + "learning_rate": 3.884523534396426e-05, + "loss": 0.0001, + "step": 5792 + }, + { + "epoch": 1.34, + "learning_rate": 3.8843309452276406e-05, + "loss": 0.0, + "step": 5793 + }, + { + "epoch": 1.34, + "learning_rate": 3.884138356058855e-05, + "loss": 0.0001, + "step": 5794 + }, + { + "epoch": 1.34, + "learning_rate": 3.88394576689007e-05, + "loss": 0.0, + "step": 5795 + }, + { + "epoch": 1.34, + "learning_rate": 3.883753177721285e-05, + "loss": 0.0001, + "step": 5796 + }, + { + "epoch": 1.34, + "learning_rate": 3.8835605885525e-05, + "loss": 0.0, + "step": 5797 + }, + { + "epoch": 1.34, + "learning_rate": 3.883367999383714e-05, + "loss": 0.0004, + "step": 5798 + }, + { + "epoch": 1.34, + "learning_rate": 3.88317541021493e-05, + "loss": 0.0002, + "step": 5799 + }, + { + "epoch": 1.34, + "learning_rate": 3.8829828210461445e-05, + "loss": 0.0002, + "step": 5800 + }, + { + "epoch": 1.34, + "learning_rate": 3.882790231877359e-05, + "loss": 0.0, + "step": 5801 + }, + { + "epoch": 1.34, + "learning_rate": 3.882597642708574e-05, + "loss": 0.0006, + "step": 5802 + }, + { + "epoch": 1.34, + "learning_rate": 3.8824050535397894e-05, + "loss": 0.0003, + "step": 5803 + }, + { + "epoch": 1.34, + "learning_rate": 3.882212464371004e-05, + "loss": 0.0001, + "step": 5804 + }, + { + "epoch": 1.34, + "learning_rate": 3.882019875202218e-05, + "loss": 0.0, + "step": 5805 + }, + { + "epoch": 1.34, + "learning_rate": 3.881827286033434e-05, + "loss": 0.0005, + "step": 5806 + }, + { + "epoch": 1.34, + "learning_rate": 3.8816346968646484e-05, + "loss": 0.0, + "step": 5807 + }, + { + "epoch": 1.34, + "learning_rate": 3.881442107695863e-05, + "loss": 0.0003, + "step": 5808 + }, + { + "epoch": 1.34, + "learning_rate": 3.881249518527078e-05, + "loss": 0.0013, + "step": 5809 + }, + { + "epoch": 1.34, + "learning_rate": 3.881056929358293e-05, + "loss": 0.0, + "step": 5810 + }, + { + "epoch": 1.34, + "learning_rate": 3.880864340189508e-05, + "loss": 0.0, + "step": 5811 + }, + { + "epoch": 1.34, + "learning_rate": 3.880671751020723e-05, + "loss": 0.0001, + "step": 5812 + }, + { + "epoch": 1.34, + "learning_rate": 3.8804791618519376e-05, + "loss": 0.0, + "step": 5813 + }, + { + "epoch": 1.34, + "learning_rate": 3.880286572683152e-05, + "loss": 0.0, + "step": 5814 + }, + { + "epoch": 1.34, + "learning_rate": 3.880093983514367e-05, + "loss": 0.0006, + "step": 5815 + }, + { + "epoch": 1.34, + "learning_rate": 3.879901394345582e-05, + "loss": 0.0, + "step": 5816 + }, + { + "epoch": 1.34, + "learning_rate": 3.879708805176797e-05, + "loss": 0.0036, + "step": 5817 + }, + { + "epoch": 1.34, + "learning_rate": 3.879516216008012e-05, + "loss": 0.0035, + "step": 5818 + }, + { + "epoch": 1.34, + "learning_rate": 3.879323626839227e-05, + "loss": 0.0001, + "step": 5819 + }, + { + "epoch": 1.35, + "learning_rate": 3.8791310376704415e-05, + "loss": 0.0003, + "step": 5820 + }, + { + "epoch": 1.35, + "learning_rate": 3.878938448501657e-05, + "loss": 0.0001, + "step": 5821 + }, + { + "epoch": 1.35, + "learning_rate": 3.878745859332871e-05, + "loss": 0.0, + "step": 5822 + }, + { + "epoch": 1.35, + "learning_rate": 3.8785532701640864e-05, + "loss": 0.0, + "step": 5823 + }, + { + "epoch": 1.35, + "learning_rate": 3.878360680995301e-05, + "loss": 0.0, + "step": 5824 + }, + { + "epoch": 1.35, + "learning_rate": 3.878168091826516e-05, + "loss": 0.0001, + "step": 5825 + }, + { + "epoch": 1.35, + "learning_rate": 3.8779755026577307e-05, + "loss": 0.0005, + "step": 5826 + }, + { + "epoch": 1.35, + "learning_rate": 3.8777829134889454e-05, + "loss": 0.0, + "step": 5827 + }, + { + "epoch": 1.35, + "learning_rate": 3.877590324320161e-05, + "loss": 0.0, + "step": 5828 + }, + { + "epoch": 1.35, + "learning_rate": 3.877397735151375e-05, + "loss": 0.0, + "step": 5829 + }, + { + "epoch": 1.35, + "learning_rate": 3.87720514598259e-05, + "loss": 0.0003, + "step": 5830 + }, + { + "epoch": 1.35, + "learning_rate": 3.877012556813805e-05, + "loss": 0.0, + "step": 5831 + }, + { + "epoch": 1.35, + "learning_rate": 3.87681996764502e-05, + "loss": 0.0001, + "step": 5832 + }, + { + "epoch": 1.35, + "learning_rate": 3.8766273784762346e-05, + "loss": 0.0001, + "step": 5833 + }, + { + "epoch": 1.35, + "learning_rate": 3.876434789307449e-05, + "loss": 0.0, + "step": 5834 + }, + { + "epoch": 1.35, + "learning_rate": 3.876242200138665e-05, + "loss": 0.0, + "step": 5835 + }, + { + "epoch": 1.35, + "learning_rate": 3.876049610969879e-05, + "loss": 0.0, + "step": 5836 + }, + { + "epoch": 1.35, + "learning_rate": 3.875857021801094e-05, + "loss": 0.0001, + "step": 5837 + }, + { + "epoch": 1.35, + "learning_rate": 3.875664432632309e-05, + "loss": 0.0, + "step": 5838 + }, + { + "epoch": 1.35, + "learning_rate": 3.875471843463524e-05, + "loss": 0.0001, + "step": 5839 + }, + { + "epoch": 1.35, + "learning_rate": 3.8752792542947385e-05, + "loss": 0.0, + "step": 5840 + }, + { + "epoch": 1.35, + "learning_rate": 3.875086665125954e-05, + "loss": 0.0001, + "step": 5841 + }, + { + "epoch": 1.35, + "learning_rate": 3.8748940759571687e-05, + "loss": 0.0001, + "step": 5842 + }, + { + "epoch": 1.35, + "learning_rate": 3.874701486788383e-05, + "loss": 0.0, + "step": 5843 + }, + { + "epoch": 1.35, + "learning_rate": 3.874508897619598e-05, + "loss": 0.0, + "step": 5844 + }, + { + "epoch": 1.35, + "learning_rate": 3.874316308450813e-05, + "loss": 0.0, + "step": 5845 + }, + { + "epoch": 1.35, + "learning_rate": 3.8741237192820276e-05, + "loss": 0.0, + "step": 5846 + }, + { + "epoch": 1.35, + "learning_rate": 3.8739311301132424e-05, + "loss": 0.0001, + "step": 5847 + }, + { + "epoch": 1.35, + "learning_rate": 3.873738540944458e-05, + "loss": 0.0, + "step": 5848 + }, + { + "epoch": 1.35, + "learning_rate": 3.8735459517756726e-05, + "loss": 0.0, + "step": 5849 + }, + { + "epoch": 1.35, + "learning_rate": 3.873353362606887e-05, + "loss": 0.0002, + "step": 5850 + }, + { + "epoch": 1.35, + "learning_rate": 3.873160773438102e-05, + "loss": 0.0, + "step": 5851 + }, + { + "epoch": 1.35, + "learning_rate": 3.872968184269317e-05, + "loss": 0.0001, + "step": 5852 + }, + { + "epoch": 1.35, + "learning_rate": 3.8727755951005316e-05, + "loss": 0.0, + "step": 5853 + }, + { + "epoch": 1.35, + "learning_rate": 3.872583005931746e-05, + "loss": 0.0002, + "step": 5854 + }, + { + "epoch": 1.35, + "learning_rate": 3.872390416762962e-05, + "loss": 0.0, + "step": 5855 + }, + { + "epoch": 1.35, + "learning_rate": 3.8721978275941765e-05, + "loss": 0.0, + "step": 5856 + }, + { + "epoch": 1.35, + "learning_rate": 3.872005238425391e-05, + "loss": 0.0001, + "step": 5857 + }, + { + "epoch": 1.35, + "learning_rate": 3.871812649256606e-05, + "loss": 0.0008, + "step": 5858 + }, + { + "epoch": 1.35, + "learning_rate": 3.8716200600878214e-05, + "loss": 0.0, + "step": 5859 + }, + { + "epoch": 1.35, + "learning_rate": 3.8714274709190355e-05, + "loss": 0.0, + "step": 5860 + }, + { + "epoch": 1.35, + "learning_rate": 3.87123488175025e-05, + "loss": 0.0002, + "step": 5861 + }, + { + "epoch": 1.35, + "learning_rate": 3.8710422925814656e-05, + "loss": 0.0006, + "step": 5862 + }, + { + "epoch": 1.35, + "learning_rate": 3.8708497034126804e-05, + "loss": 0.0, + "step": 5863 + }, + { + "epoch": 1.36, + "learning_rate": 3.870657114243895e-05, + "loss": 0.0001, + "step": 5864 + }, + { + "epoch": 1.36, + "learning_rate": 3.87046452507511e-05, + "loss": 0.0001, + "step": 5865 + }, + { + "epoch": 1.36, + "learning_rate": 3.870271935906325e-05, + "loss": 0.0014, + "step": 5866 + }, + { + "epoch": 1.36, + "learning_rate": 3.8700793467375394e-05, + "loss": 0.0, + "step": 5867 + }, + { + "epoch": 1.36, + "learning_rate": 3.869886757568755e-05, + "loss": 0.0, + "step": 5868 + }, + { + "epoch": 1.36, + "learning_rate": 3.8696941683999696e-05, + "loss": 0.0001, + "step": 5869 + }, + { + "epoch": 1.36, + "learning_rate": 3.8695015792311836e-05, + "loss": 0.0, + "step": 5870 + }, + { + "epoch": 1.36, + "learning_rate": 3.869308990062399e-05, + "loss": 0.0002, + "step": 5871 + }, + { + "epoch": 1.36, + "learning_rate": 3.869116400893614e-05, + "loss": 0.0, + "step": 5872 + }, + { + "epoch": 1.36, + "learning_rate": 3.8689238117248285e-05, + "loss": 0.0001, + "step": 5873 + }, + { + "epoch": 1.36, + "learning_rate": 3.868731222556043e-05, + "loss": 0.0001, + "step": 5874 + }, + { + "epoch": 1.36, + "learning_rate": 3.868538633387259e-05, + "loss": 0.0002, + "step": 5875 + }, + { + "epoch": 1.36, + "learning_rate": 3.8683460442184735e-05, + "loss": 0.0004, + "step": 5876 + }, + { + "epoch": 1.36, + "learning_rate": 3.868153455049688e-05, + "loss": 0.0, + "step": 5877 + }, + { + "epoch": 1.36, + "learning_rate": 3.867960865880903e-05, + "loss": 0.0022, + "step": 5878 + }, + { + "epoch": 1.36, + "learning_rate": 3.8677682767121184e-05, + "loss": 0.0, + "step": 5879 + }, + { + "epoch": 1.36, + "learning_rate": 3.8675756875433325e-05, + "loss": 0.0008, + "step": 5880 + }, + { + "epoch": 1.36, + "learning_rate": 3.867383098374547e-05, + "loss": 0.0, + "step": 5881 + }, + { + "epoch": 1.36, + "learning_rate": 3.8671905092057626e-05, + "loss": 0.0001, + "step": 5882 + }, + { + "epoch": 1.36, + "learning_rate": 3.8669979200369774e-05, + "loss": 0.0006, + "step": 5883 + }, + { + "epoch": 1.36, + "learning_rate": 3.866805330868192e-05, + "loss": 0.0104, + "step": 5884 + }, + { + "epoch": 1.36, + "learning_rate": 3.866612741699407e-05, + "loss": 0.0002, + "step": 5885 + }, + { + "epoch": 1.36, + "learning_rate": 3.866420152530622e-05, + "loss": 0.0, + "step": 5886 + }, + { + "epoch": 1.36, + "learning_rate": 3.8662275633618364e-05, + "loss": 0.0049, + "step": 5887 + }, + { + "epoch": 1.36, + "learning_rate": 3.866034974193052e-05, + "loss": 0.0, + "step": 5888 + }, + { + "epoch": 1.36, + "learning_rate": 3.8658423850242665e-05, + "loss": 0.0012, + "step": 5889 + }, + { + "epoch": 1.36, + "learning_rate": 3.865649795855481e-05, + "loss": 0.0, + "step": 5890 + }, + { + "epoch": 1.36, + "learning_rate": 3.865457206686696e-05, + "loss": 0.0001, + "step": 5891 + }, + { + "epoch": 1.36, + "learning_rate": 3.865264617517911e-05, + "loss": 0.0, + "step": 5892 + }, + { + "epoch": 1.36, + "learning_rate": 3.865072028349126e-05, + "loss": 0.0001, + "step": 5893 + }, + { + "epoch": 1.36, + "learning_rate": 3.86487943918034e-05, + "loss": 0.0001, + "step": 5894 + }, + { + "epoch": 1.36, + "learning_rate": 3.864686850011556e-05, + "loss": 0.0288, + "step": 5895 + }, + { + "epoch": 1.36, + "learning_rate": 3.8644942608427705e-05, + "loss": 0.0001, + "step": 5896 + }, + { + "epoch": 1.36, + "learning_rate": 3.864301671673985e-05, + "loss": 0.0, + "step": 5897 + }, + { + "epoch": 1.36, + "learning_rate": 3.8641090825052e-05, + "loss": 0.0003, + "step": 5898 + }, + { + "epoch": 1.36, + "learning_rate": 3.863916493336415e-05, + "loss": 0.0, + "step": 5899 + }, + { + "epoch": 1.36, + "learning_rate": 3.86372390416763e-05, + "loss": 0.0, + "step": 5900 + }, + { + "epoch": 1.36, + "learning_rate": 3.863531314998844e-05, + "loss": 0.0001, + "step": 5901 + }, + { + "epoch": 1.36, + "learning_rate": 3.8633387258300596e-05, + "loss": 0.0001, + "step": 5902 + }, + { + "epoch": 1.36, + "learning_rate": 3.8631461366612744e-05, + "loss": 0.0003, + "step": 5903 + }, + { + "epoch": 1.36, + "learning_rate": 3.862953547492489e-05, + "loss": 0.0, + "step": 5904 + }, + { + "epoch": 1.36, + "learning_rate": 3.862760958323704e-05, + "loss": 0.0, + "step": 5905 + }, + { + "epoch": 1.36, + "learning_rate": 3.862568369154919e-05, + "loss": 0.0, + "step": 5906 + }, + { + "epoch": 1.37, + "learning_rate": 3.862375779986134e-05, + "loss": 0.0, + "step": 5907 + }, + { + "epoch": 1.37, + "learning_rate": 3.862183190817348e-05, + "loss": 0.0, + "step": 5908 + }, + { + "epoch": 1.37, + "learning_rate": 3.8619906016485635e-05, + "loss": 0.0, + "step": 5909 + }, + { + "epoch": 1.37, + "learning_rate": 3.861798012479778e-05, + "loss": 0.0, + "step": 5910 + }, + { + "epoch": 1.37, + "learning_rate": 3.861605423310993e-05, + "loss": 0.0, + "step": 5911 + }, + { + "epoch": 1.37, + "learning_rate": 3.861412834142208e-05, + "loss": 0.0, + "step": 5912 + }, + { + "epoch": 1.37, + "learning_rate": 3.861220244973423e-05, + "loss": 0.0001, + "step": 5913 + }, + { + "epoch": 1.37, + "learning_rate": 3.861027655804638e-05, + "loss": 0.0018, + "step": 5914 + }, + { + "epoch": 1.37, + "learning_rate": 3.860835066635853e-05, + "loss": 0.0, + "step": 5915 + }, + { + "epoch": 1.37, + "learning_rate": 3.8606424774670674e-05, + "loss": 0.0005, + "step": 5916 + }, + { + "epoch": 1.37, + "learning_rate": 3.860449888298282e-05, + "loss": 0.0037, + "step": 5917 + }, + { + "epoch": 1.37, + "learning_rate": 3.860257299129497e-05, + "loss": 0.0, + "step": 5918 + }, + { + "epoch": 1.37, + "learning_rate": 3.860064709960712e-05, + "loss": 0.0, + "step": 5919 + }, + { + "epoch": 1.37, + "learning_rate": 3.859872120791927e-05, + "loss": 0.0002, + "step": 5920 + }, + { + "epoch": 1.37, + "learning_rate": 3.859679531623142e-05, + "loss": 0.0105, + "step": 5921 + }, + { + "epoch": 1.37, + "learning_rate": 3.8594869424543566e-05, + "loss": 0.0008, + "step": 5922 + }, + { + "epoch": 1.37, + "learning_rate": 3.8592943532855714e-05, + "loss": 0.0, + "step": 5923 + }, + { + "epoch": 1.37, + "learning_rate": 3.859101764116787e-05, + "loss": 0.0, + "step": 5924 + }, + { + "epoch": 1.37, + "learning_rate": 3.858909174948001e-05, + "loss": 0.0, + "step": 5925 + }, + { + "epoch": 1.37, + "learning_rate": 3.8587165857792156e-05, + "loss": 0.0008, + "step": 5926 + }, + { + "epoch": 1.37, + "learning_rate": 3.858523996610431e-05, + "loss": 0.0, + "step": 5927 + }, + { + "epoch": 1.37, + "learning_rate": 3.858331407441646e-05, + "loss": 0.0005, + "step": 5928 + }, + { + "epoch": 1.37, + "learning_rate": 3.8581388182728605e-05, + "loss": 0.0, + "step": 5929 + }, + { + "epoch": 1.37, + "learning_rate": 3.857946229104075e-05, + "loss": 0.0003, + "step": 5930 + }, + { + "epoch": 1.37, + "learning_rate": 3.857753639935291e-05, + "loss": 0.0, + "step": 5931 + }, + { + "epoch": 1.37, + "learning_rate": 3.857561050766505e-05, + "loss": 0.0, + "step": 5932 + }, + { + "epoch": 1.37, + "learning_rate": 3.85736846159772e-05, + "loss": 0.0042, + "step": 5933 + }, + { + "epoch": 1.37, + "learning_rate": 3.857175872428935e-05, + "loss": 0.0003, + "step": 5934 + }, + { + "epoch": 1.37, + "learning_rate": 3.85698328326015e-05, + "loss": 0.0002, + "step": 5935 + }, + { + "epoch": 1.37, + "learning_rate": 3.8567906940913644e-05, + "loss": 0.0, + "step": 5936 + }, + { + "epoch": 1.37, + "learning_rate": 3.856598104922579e-05, + "loss": 0.0, + "step": 5937 + }, + { + "epoch": 1.37, + "learning_rate": 3.8564055157537946e-05, + "loss": 0.0, + "step": 5938 + }, + { + "epoch": 1.37, + "learning_rate": 3.856212926585009e-05, + "loss": 0.0014, + "step": 5939 + }, + { + "epoch": 1.37, + "learning_rate": 3.856020337416224e-05, + "loss": 0.0046, + "step": 5940 + }, + { + "epoch": 1.37, + "learning_rate": 3.855827748247439e-05, + "loss": 0.0009, + "step": 5941 + }, + { + "epoch": 1.37, + "learning_rate": 3.8556351590786536e-05, + "loss": 0.0009, + "step": 5942 + }, + { + "epoch": 1.37, + "learning_rate": 3.8554425699098683e-05, + "loss": 0.0001, + "step": 5943 + }, + { + "epoch": 1.37, + "learning_rate": 3.855249980741084e-05, + "loss": 0.0001, + "step": 5944 + }, + { + "epoch": 1.37, + "learning_rate": 3.855057391572298e-05, + "loss": 0.0011, + "step": 5945 + }, + { + "epoch": 1.37, + "learning_rate": 3.8548648024035126e-05, + "loss": 0.0001, + "step": 5946 + }, + { + "epoch": 1.37, + "learning_rate": 3.854672213234728e-05, + "loss": 0.0, + "step": 5947 + }, + { + "epoch": 1.37, + "learning_rate": 3.854479624065943e-05, + "loss": 0.0001, + "step": 5948 + }, + { + "epoch": 1.37, + "learning_rate": 3.8542870348971575e-05, + "loss": 0.0001, + "step": 5949 + }, + { + "epoch": 1.38, + "learning_rate": 3.854094445728372e-05, + "loss": 0.0053, + "step": 5950 + }, + { + "epoch": 1.38, + "learning_rate": 3.853901856559588e-05, + "loss": 0.0, + "step": 5951 + }, + { + "epoch": 1.38, + "learning_rate": 3.853709267390802e-05, + "loss": 0.0005, + "step": 5952 + }, + { + "epoch": 1.38, + "learning_rate": 3.853516678222017e-05, + "loss": 0.0009, + "step": 5953 + }, + { + "epoch": 1.38, + "learning_rate": 3.853324089053232e-05, + "loss": 0.0001, + "step": 5954 + }, + { + "epoch": 1.38, + "learning_rate": 3.853131499884447e-05, + "loss": 0.1651, + "step": 5955 + }, + { + "epoch": 1.38, + "learning_rate": 3.8529389107156614e-05, + "loss": 0.0001, + "step": 5956 + }, + { + "epoch": 1.38, + "learning_rate": 3.852746321546876e-05, + "loss": 0.0, + "step": 5957 + }, + { + "epoch": 1.38, + "learning_rate": 3.8525537323780916e-05, + "loss": 0.0002, + "step": 5958 + }, + { + "epoch": 1.38, + "learning_rate": 3.852361143209306e-05, + "loss": 0.0001, + "step": 5959 + }, + { + "epoch": 1.38, + "learning_rate": 3.852168554040521e-05, + "loss": 0.0, + "step": 5960 + }, + { + "epoch": 1.38, + "learning_rate": 3.851975964871736e-05, + "loss": 0.0, + "step": 5961 + }, + { + "epoch": 1.38, + "learning_rate": 3.8517833757029506e-05, + "loss": 0.0, + "step": 5962 + }, + { + "epoch": 1.38, + "learning_rate": 3.851590786534165e-05, + "loss": 0.0007, + "step": 5963 + }, + { + "epoch": 1.38, + "learning_rate": 3.85139819736538e-05, + "loss": 0.0003, + "step": 5964 + }, + { + "epoch": 1.38, + "learning_rate": 3.8512056081965955e-05, + "loss": 0.0, + "step": 5965 + }, + { + "epoch": 1.38, + "learning_rate": 3.8510130190278096e-05, + "loss": 0.0, + "step": 5966 + }, + { + "epoch": 1.38, + "learning_rate": 3.850820429859025e-05, + "loss": 0.0004, + "step": 5967 + }, + { + "epoch": 1.38, + "learning_rate": 3.85062784069024e-05, + "loss": 0.0001, + "step": 5968 + }, + { + "epoch": 1.38, + "learning_rate": 3.8504352515214545e-05, + "loss": 0.0, + "step": 5969 + }, + { + "epoch": 1.38, + "learning_rate": 3.850242662352669e-05, + "loss": 0.0, + "step": 5970 + }, + { + "epoch": 1.38, + "learning_rate": 3.850050073183885e-05, + "loss": 0.0001, + "step": 5971 + }, + { + "epoch": 1.38, + "learning_rate": 3.8498574840150994e-05, + "loss": 0.0, + "step": 5972 + }, + { + "epoch": 1.38, + "learning_rate": 3.8496648948463135e-05, + "loss": 0.0001, + "step": 5973 + }, + { + "epoch": 1.38, + "learning_rate": 3.849472305677529e-05, + "loss": 0.0003, + "step": 5974 + }, + { + "epoch": 1.38, + "learning_rate": 3.8492797165087437e-05, + "loss": 0.0001, + "step": 5975 + }, + { + "epoch": 1.38, + "learning_rate": 3.8490871273399584e-05, + "loss": 0.0001, + "step": 5976 + }, + { + "epoch": 1.38, + "learning_rate": 3.848894538171173e-05, + "loss": 0.0002, + "step": 5977 + }, + { + "epoch": 1.38, + "learning_rate": 3.8487019490023886e-05, + "loss": 0.0, + "step": 5978 + }, + { + "epoch": 1.38, + "learning_rate": 3.848509359833603e-05, + "loss": 0.0001, + "step": 5979 + }, + { + "epoch": 1.38, + "learning_rate": 3.848316770664818e-05, + "loss": 0.0, + "step": 5980 + }, + { + "epoch": 1.38, + "learning_rate": 3.848124181496033e-05, + "loss": 0.0001, + "step": 5981 + }, + { + "epoch": 1.38, + "learning_rate": 3.8479315923272476e-05, + "loss": 0.0005, + "step": 5982 + }, + { + "epoch": 1.38, + "learning_rate": 3.847739003158462e-05, + "loss": 0.0, + "step": 5983 + }, + { + "epoch": 1.38, + "learning_rate": 3.847546413989677e-05, + "loss": 0.0, + "step": 5984 + }, + { + "epoch": 1.38, + "learning_rate": 3.8473538248208925e-05, + "loss": 0.0001, + "step": 5985 + }, + { + "epoch": 1.38, + "learning_rate": 3.847161235652107e-05, + "loss": 0.0001, + "step": 5986 + }, + { + "epoch": 1.38, + "learning_rate": 3.846968646483322e-05, + "loss": 0.0, + "step": 5987 + }, + { + "epoch": 1.38, + "learning_rate": 3.846776057314537e-05, + "loss": 0.0003, + "step": 5988 + }, + { + "epoch": 1.38, + "learning_rate": 3.846583468145752e-05, + "loss": 0.0002, + "step": 5989 + }, + { + "epoch": 1.38, + "learning_rate": 3.846390878976966e-05, + "loss": 0.0001, + "step": 5990 + }, + { + "epoch": 1.38, + "learning_rate": 3.8461982898081817e-05, + "loss": 0.0007, + "step": 5991 + }, + { + "epoch": 1.38, + "learning_rate": 3.8460057006393964e-05, + "loss": 0.0, + "step": 5992 + }, + { + "epoch": 1.39, + "learning_rate": 3.845813111470611e-05, + "loss": 0.0001, + "step": 5993 + }, + { + "epoch": 1.39, + "learning_rate": 3.845620522301826e-05, + "loss": 0.0001, + "step": 5994 + }, + { + "epoch": 1.39, + "learning_rate": 3.8454279331330406e-05, + "loss": 0.0003, + "step": 5995 + }, + { + "epoch": 1.39, + "learning_rate": 3.845235343964256e-05, + "loss": 0.0, + "step": 5996 + }, + { + "epoch": 1.39, + "learning_rate": 3.84504275479547e-05, + "loss": 0.0001, + "step": 5997 + }, + { + "epoch": 1.39, + "learning_rate": 3.8448501656266856e-05, + "loss": 0.0015, + "step": 5998 + }, + { + "epoch": 1.39, + "learning_rate": 3.8446575764579e-05, + "loss": 0.0013, + "step": 5999 + }, + { + "epoch": 1.39, + "learning_rate": 3.844464987289115e-05, + "loss": 0.0002, + "step": 6000 + }, + { + "epoch": 1.39, + "learning_rate": 3.84427239812033e-05, + "loss": 0.0, + "step": 6001 + }, + { + "epoch": 1.39, + "learning_rate": 3.8440798089515446e-05, + "loss": 0.001, + "step": 6002 + }, + { + "epoch": 1.39, + "learning_rate": 3.84388721978276e-05, + "loss": 0.0005, + "step": 6003 + }, + { + "epoch": 1.39, + "learning_rate": 3.843694630613974e-05, + "loss": 0.0002, + "step": 6004 + }, + { + "epoch": 1.39, + "learning_rate": 3.8435020414451895e-05, + "loss": 0.0003, + "step": 6005 + }, + { + "epoch": 1.39, + "learning_rate": 3.843309452276404e-05, + "loss": 0.0, + "step": 6006 + }, + { + "epoch": 1.39, + "learning_rate": 3.843116863107619e-05, + "loss": 0.0026, + "step": 6007 + }, + { + "epoch": 1.39, + "learning_rate": 3.842924273938834e-05, + "loss": 0.0004, + "step": 6008 + }, + { + "epoch": 1.39, + "learning_rate": 3.842731684770049e-05, + "loss": 0.0, + "step": 6009 + }, + { + "epoch": 1.39, + "learning_rate": 3.842539095601264e-05, + "loss": 0.0003, + "step": 6010 + }, + { + "epoch": 1.39, + "learning_rate": 3.842346506432478e-05, + "loss": 0.0004, + "step": 6011 + }, + { + "epoch": 1.39, + "learning_rate": 3.8421539172636934e-05, + "loss": 0.0001, + "step": 6012 + }, + { + "epoch": 1.39, + "learning_rate": 3.841961328094908e-05, + "loss": 0.0, + "step": 6013 + }, + { + "epoch": 1.39, + "learning_rate": 3.841768738926123e-05, + "loss": 0.0002, + "step": 6014 + }, + { + "epoch": 1.39, + "learning_rate": 3.8415761497573376e-05, + "loss": 0.0005, + "step": 6015 + }, + { + "epoch": 1.39, + "learning_rate": 3.841383560588553e-05, + "loss": 0.0, + "step": 6016 + }, + { + "epoch": 1.39, + "learning_rate": 3.841190971419767e-05, + "loss": 0.0001, + "step": 6017 + }, + { + "epoch": 1.39, + "learning_rate": 3.8409983822509826e-05, + "loss": 0.0001, + "step": 6018 + }, + { + "epoch": 1.39, + "learning_rate": 3.840805793082197e-05, + "loss": 0.0014, + "step": 6019 + }, + { + "epoch": 1.39, + "learning_rate": 3.840613203913412e-05, + "loss": 0.0, + "step": 6020 + }, + { + "epoch": 1.39, + "learning_rate": 3.840420614744627e-05, + "loss": 0.0003, + "step": 6021 + }, + { + "epoch": 1.39, + "learning_rate": 3.8402280255758415e-05, + "loss": 0.001, + "step": 6022 + }, + { + "epoch": 1.39, + "learning_rate": 3.840035436407057e-05, + "loss": 0.0002, + "step": 6023 + }, + { + "epoch": 1.39, + "learning_rate": 3.839842847238271e-05, + "loss": 0.0, + "step": 6024 + }, + { + "epoch": 1.39, + "learning_rate": 3.8396502580694865e-05, + "loss": 0.0006, + "step": 6025 + }, + { + "epoch": 1.39, + "learning_rate": 3.839457668900701e-05, + "loss": 0.0009, + "step": 6026 + }, + { + "epoch": 1.39, + "learning_rate": 3.839265079731916e-05, + "loss": 0.0, + "step": 6027 + }, + { + "epoch": 1.39, + "learning_rate": 3.839072490563131e-05, + "loss": 0.0005, + "step": 6028 + }, + { + "epoch": 1.39, + "learning_rate": 3.8388799013943455e-05, + "loss": 0.0008, + "step": 6029 + }, + { + "epoch": 1.39, + "learning_rate": 3.838687312225561e-05, + "loss": 0.0001, + "step": 6030 + }, + { + "epoch": 1.39, + "learning_rate": 3.838494723056775e-05, + "loss": 0.0, + "step": 6031 + }, + { + "epoch": 1.39, + "learning_rate": 3.8383021338879904e-05, + "loss": 0.0002, + "step": 6032 + }, + { + "epoch": 1.39, + "learning_rate": 3.838109544719205e-05, + "loss": 0.0, + "step": 6033 + }, + { + "epoch": 1.39, + "learning_rate": 3.83791695555042e-05, + "loss": 0.0001, + "step": 6034 + }, + { + "epoch": 1.39, + "learning_rate": 3.8377243663816346e-05, + "loss": 0.0002, + "step": 6035 + }, + { + "epoch": 1.39, + "learning_rate": 3.83753177721285e-05, + "loss": 0.0, + "step": 6036 + }, + { + "epoch": 1.4, + "learning_rate": 3.837339188044065e-05, + "loss": 0.0, + "step": 6037 + }, + { + "epoch": 1.4, + "learning_rate": 3.837146598875279e-05, + "loss": 0.0, + "step": 6038 + }, + { + "epoch": 1.4, + "learning_rate": 3.836954009706494e-05, + "loss": 0.0, + "step": 6039 + }, + { + "epoch": 1.4, + "learning_rate": 3.836761420537709e-05, + "loss": 0.0001, + "step": 6040 + }, + { + "epoch": 1.4, + "learning_rate": 3.836568831368924e-05, + "loss": 0.0001, + "step": 6041 + }, + { + "epoch": 1.4, + "learning_rate": 3.8363762422001385e-05, + "loss": 0.0, + "step": 6042 + }, + { + "epoch": 1.4, + "learning_rate": 3.836183653031354e-05, + "loss": 0.0, + "step": 6043 + }, + { + "epoch": 1.4, + "learning_rate": 3.835991063862569e-05, + "loss": 0.0, + "step": 6044 + }, + { + "epoch": 1.4, + "learning_rate": 3.8357984746937835e-05, + "loss": 0.0001, + "step": 6045 + }, + { + "epoch": 1.4, + "learning_rate": 3.835605885524998e-05, + "loss": 0.0, + "step": 6046 + }, + { + "epoch": 1.4, + "learning_rate": 3.8354132963562136e-05, + "loss": 0.0, + "step": 6047 + }, + { + "epoch": 1.4, + "learning_rate": 3.835220707187428e-05, + "loss": 0.0, + "step": 6048 + }, + { + "epoch": 1.4, + "learning_rate": 3.8350281180186424e-05, + "loss": 0.0009, + "step": 6049 + }, + { + "epoch": 1.4, + "learning_rate": 3.834835528849858e-05, + "loss": 0.0, + "step": 6050 + }, + { + "epoch": 1.4, + "learning_rate": 3.8346429396810726e-05, + "loss": 0.0004, + "step": 6051 + }, + { + "epoch": 1.4, + "learning_rate": 3.8344503505122874e-05, + "loss": 0.0, + "step": 6052 + }, + { + "epoch": 1.4, + "learning_rate": 3.834257761343502e-05, + "loss": 0.0002, + "step": 6053 + }, + { + "epoch": 1.4, + "learning_rate": 3.8340651721747175e-05, + "loss": 0.0, + "step": 6054 + }, + { + "epoch": 1.4, + "learning_rate": 3.8338725830059316e-05, + "loss": 0.0002, + "step": 6055 + }, + { + "epoch": 1.4, + "learning_rate": 3.833679993837147e-05, + "loss": 0.0012, + "step": 6056 + }, + { + "epoch": 1.4, + "learning_rate": 3.833487404668362e-05, + "loss": 0.0, + "step": 6057 + }, + { + "epoch": 1.4, + "learning_rate": 3.8332948154995765e-05, + "loss": 0.0006, + "step": 6058 + }, + { + "epoch": 1.4, + "learning_rate": 3.833102226330791e-05, + "loss": 0.0, + "step": 6059 + }, + { + "epoch": 1.4, + "learning_rate": 3.832909637162006e-05, + "loss": 0.0, + "step": 6060 + }, + { + "epoch": 1.4, + "learning_rate": 3.8327170479932215e-05, + "loss": 0.0002, + "step": 6061 + }, + { + "epoch": 1.4, + "learning_rate": 3.8325244588244355e-05, + "loss": 0.003, + "step": 6062 + }, + { + "epoch": 1.4, + "learning_rate": 3.832331869655651e-05, + "loss": 0.0, + "step": 6063 + }, + { + "epoch": 1.4, + "learning_rate": 3.832139280486866e-05, + "loss": 0.0, + "step": 6064 + }, + { + "epoch": 1.4, + "learning_rate": 3.8319466913180804e-05, + "loss": 0.0001, + "step": 6065 + }, + { + "epoch": 1.4, + "learning_rate": 3.831754102149295e-05, + "loss": 0.0, + "step": 6066 + }, + { + "epoch": 1.4, + "learning_rate": 3.83156151298051e-05, + "loss": 0.0001, + "step": 6067 + }, + { + "epoch": 1.4, + "learning_rate": 3.8313689238117254e-05, + "loss": 0.0003, + "step": 6068 + }, + { + "epoch": 1.4, + "learning_rate": 3.8311763346429394e-05, + "loss": 0.0, + "step": 6069 + }, + { + "epoch": 1.4, + "learning_rate": 3.830983745474155e-05, + "loss": 0.0001, + "step": 6070 + }, + { + "epoch": 1.4, + "learning_rate": 3.8307911563053696e-05, + "loss": 0.0, + "step": 6071 + }, + { + "epoch": 1.4, + "learning_rate": 3.8305985671365844e-05, + "loss": 0.0003, + "step": 6072 + }, + { + "epoch": 1.4, + "learning_rate": 3.830405977967799e-05, + "loss": 0.0002, + "step": 6073 + }, + { + "epoch": 1.4, + "learning_rate": 3.8302133887990145e-05, + "loss": 0.0, + "step": 6074 + }, + { + "epoch": 1.4, + "learning_rate": 3.830020799630229e-05, + "loss": 0.0002, + "step": 6075 + }, + { + "epoch": 1.4, + "learning_rate": 3.8298282104614433e-05, + "loss": 0.0001, + "step": 6076 + }, + { + "epoch": 1.4, + "learning_rate": 3.829635621292659e-05, + "loss": 0.0004, + "step": 6077 + }, + { + "epoch": 1.4, + "learning_rate": 3.8294430321238735e-05, + "loss": 0.0, + "step": 6078 + }, + { + "epoch": 1.4, + "learning_rate": 3.829250442955088e-05, + "loss": 0.0001, + "step": 6079 + }, + { + "epoch": 1.41, + "learning_rate": 3.829057853786303e-05, + "loss": 0.0001, + "step": 6080 + }, + { + "epoch": 1.41, + "learning_rate": 3.8288652646175184e-05, + "loss": 0.0001, + "step": 6081 + }, + { + "epoch": 1.41, + "learning_rate": 3.828672675448733e-05, + "loss": 0.0009, + "step": 6082 + }, + { + "epoch": 1.41, + "learning_rate": 3.828480086279948e-05, + "loss": 0.0007, + "step": 6083 + }, + { + "epoch": 1.41, + "learning_rate": 3.828287497111163e-05, + "loss": 0.0, + "step": 6084 + }, + { + "epoch": 1.41, + "learning_rate": 3.8280949079423774e-05, + "loss": 0.0002, + "step": 6085 + }, + { + "epoch": 1.41, + "learning_rate": 3.827902318773592e-05, + "loss": 0.0003, + "step": 6086 + }, + { + "epoch": 1.41, + "learning_rate": 3.827709729604807e-05, + "loss": 0.0, + "step": 6087 + }, + { + "epoch": 1.41, + "learning_rate": 3.8275171404360224e-05, + "loss": 0.0024, + "step": 6088 + }, + { + "epoch": 1.41, + "learning_rate": 3.827324551267237e-05, + "loss": 0.0001, + "step": 6089 + }, + { + "epoch": 1.41, + "learning_rate": 3.827131962098452e-05, + "loss": 0.0002, + "step": 6090 + }, + { + "epoch": 1.41, + "learning_rate": 3.8269393729296666e-05, + "loss": 0.0, + "step": 6091 + }, + { + "epoch": 1.41, + "learning_rate": 3.826746783760882e-05, + "loss": 0.0004, + "step": 6092 + }, + { + "epoch": 1.41, + "learning_rate": 3.826554194592096e-05, + "loss": 0.0002, + "step": 6093 + }, + { + "epoch": 1.41, + "learning_rate": 3.826361605423311e-05, + "loss": 0.0, + "step": 6094 + }, + { + "epoch": 1.41, + "learning_rate": 3.826169016254526e-05, + "loss": 0.0001, + "step": 6095 + }, + { + "epoch": 1.41, + "learning_rate": 3.82597642708574e-05, + "loss": 0.0, + "step": 6096 + }, + { + "epoch": 1.41, + "learning_rate": 3.825783837916956e-05, + "loss": 0.0006, + "step": 6097 + }, + { + "epoch": 1.41, + "learning_rate": 3.8255912487481705e-05, + "loss": 0.0002, + "step": 6098 + }, + { + "epoch": 1.41, + "learning_rate": 3.825398659579385e-05, + "loss": 0.0, + "step": 6099 + }, + { + "epoch": 1.41, + "learning_rate": 3.8252060704106e-05, + "loss": 0.0, + "step": 6100 + }, + { + "epoch": 1.41, + "learning_rate": 3.8250134812418154e-05, + "loss": 0.0001, + "step": 6101 + }, + { + "epoch": 1.41, + "learning_rate": 3.82482089207303e-05, + "loss": 0.0, + "step": 6102 + }, + { + "epoch": 1.41, + "learning_rate": 3.824628302904245e-05, + "loss": 0.0001, + "step": 6103 + }, + { + "epoch": 1.41, + "learning_rate": 3.82443571373546e-05, + "loss": 0.0, + "step": 6104 + }, + { + "epoch": 1.41, + "learning_rate": 3.8242431245666744e-05, + "loss": 0.0007, + "step": 6105 + }, + { + "epoch": 1.41, + "learning_rate": 3.824050535397889e-05, + "loss": 0.0003, + "step": 6106 + }, + { + "epoch": 1.41, + "learning_rate": 3.823857946229104e-05, + "loss": 0.0001, + "step": 6107 + }, + { + "epoch": 1.41, + "learning_rate": 3.8236653570603193e-05, + "loss": 0.0002, + "step": 6108 + }, + { + "epoch": 1.41, + "learning_rate": 3.823472767891534e-05, + "loss": 0.0, + "step": 6109 + }, + { + "epoch": 1.41, + "learning_rate": 3.823280178722749e-05, + "loss": 0.0002, + "step": 6110 + }, + { + "epoch": 1.41, + "learning_rate": 3.8230875895539636e-05, + "loss": 0.0001, + "step": 6111 + }, + { + "epoch": 1.41, + "learning_rate": 3.822895000385179e-05, + "loss": 0.0005, + "step": 6112 + }, + { + "epoch": 1.41, + "learning_rate": 3.822702411216393e-05, + "loss": 0.0001, + "step": 6113 + }, + { + "epoch": 1.41, + "learning_rate": 3.822509822047608e-05, + "loss": 0.0, + "step": 6114 + }, + { + "epoch": 1.41, + "learning_rate": 3.822317232878823e-05, + "loss": 0.0, + "step": 6115 + }, + { + "epoch": 1.41, + "learning_rate": 3.822124643710038e-05, + "loss": 0.0, + "step": 6116 + }, + { + "epoch": 1.41, + "learning_rate": 3.821932054541253e-05, + "loss": 0.0001, + "step": 6117 + }, + { + "epoch": 1.41, + "learning_rate": 3.8217394653724675e-05, + "loss": 0.0, + "step": 6118 + }, + { + "epoch": 1.41, + "learning_rate": 3.821546876203683e-05, + "loss": 0.001, + "step": 6119 + }, + { + "epoch": 1.41, + "learning_rate": 3.821354287034897e-05, + "loss": 0.0001, + "step": 6120 + }, + { + "epoch": 1.41, + "learning_rate": 3.8211616978661124e-05, + "loss": 0.0007, + "step": 6121 + }, + { + "epoch": 1.41, + "learning_rate": 3.820969108697327e-05, + "loss": 0.0, + "step": 6122 + }, + { + "epoch": 1.42, + "learning_rate": 3.820776519528542e-05, + "loss": 0.0, + "step": 6123 + }, + { + "epoch": 1.42, + "learning_rate": 3.820583930359757e-05, + "loss": 0.0, + "step": 6124 + }, + { + "epoch": 1.42, + "learning_rate": 3.8203913411909714e-05, + "loss": 0.0, + "step": 6125 + }, + { + "epoch": 1.42, + "learning_rate": 3.820198752022187e-05, + "loss": 0.0, + "step": 6126 + }, + { + "epoch": 1.42, + "learning_rate": 3.820006162853401e-05, + "loss": 0.0, + "step": 6127 + }, + { + "epoch": 1.42, + "learning_rate": 3.819813573684616e-05, + "loss": 0.0009, + "step": 6128 + }, + { + "epoch": 1.42, + "learning_rate": 3.819620984515831e-05, + "loss": 0.0, + "step": 6129 + }, + { + "epoch": 1.42, + "learning_rate": 3.819428395347046e-05, + "loss": 0.0, + "step": 6130 + }, + { + "epoch": 1.42, + "learning_rate": 3.8192358061782606e-05, + "loss": 0.0001, + "step": 6131 + }, + { + "epoch": 1.42, + "learning_rate": 3.819043217009475e-05, + "loss": 0.0, + "step": 6132 + }, + { + "epoch": 1.42, + "learning_rate": 3.818850627840691e-05, + "loss": 0.0, + "step": 6133 + }, + { + "epoch": 1.42, + "learning_rate": 3.818658038671905e-05, + "loss": 0.0, + "step": 6134 + }, + { + "epoch": 1.42, + "learning_rate": 3.81846544950312e-05, + "loss": 0.0, + "step": 6135 + }, + { + "epoch": 1.42, + "learning_rate": 3.818272860334335e-05, + "loss": 0.0, + "step": 6136 + }, + { + "epoch": 1.42, + "learning_rate": 3.81808027116555e-05, + "loss": 0.0, + "step": 6137 + }, + { + "epoch": 1.42, + "learning_rate": 3.8178876819967645e-05, + "loss": 0.0001, + "step": 6138 + }, + { + "epoch": 1.42, + "learning_rate": 3.81769509282798e-05, + "loss": 0.0001, + "step": 6139 + }, + { + "epoch": 1.42, + "learning_rate": 3.817502503659195e-05, + "loss": 0.0, + "step": 6140 + }, + { + "epoch": 1.42, + "learning_rate": 3.817309914490409e-05, + "loss": 0.0022, + "step": 6141 + }, + { + "epoch": 1.42, + "learning_rate": 3.817117325321624e-05, + "loss": 0.0, + "step": 6142 + }, + { + "epoch": 1.42, + "learning_rate": 3.816924736152839e-05, + "loss": 0.0002, + "step": 6143 + }, + { + "epoch": 1.42, + "learning_rate": 3.8167321469840537e-05, + "loss": 0.0001, + "step": 6144 + }, + { + "epoch": 1.42, + "learning_rate": 3.8165395578152684e-05, + "loss": 0.0, + "step": 6145 + }, + { + "epoch": 1.42, + "learning_rate": 3.816346968646484e-05, + "loss": 0.0001, + "step": 6146 + }, + { + "epoch": 1.42, + "learning_rate": 3.8161543794776986e-05, + "loss": 0.0, + "step": 6147 + }, + { + "epoch": 1.42, + "learning_rate": 3.815961790308913e-05, + "loss": 0.0001, + "step": 6148 + }, + { + "epoch": 1.42, + "learning_rate": 3.815769201140128e-05, + "loss": 0.0, + "step": 6149 + }, + { + "epoch": 1.42, + "learning_rate": 3.815576611971343e-05, + "loss": 0.0003, + "step": 6150 + }, + { + "epoch": 1.42, + "learning_rate": 3.8153840228025576e-05, + "loss": 0.0, + "step": 6151 + }, + { + "epoch": 1.42, + "learning_rate": 3.815191433633772e-05, + "loss": 0.0, + "step": 6152 + }, + { + "epoch": 1.42, + "learning_rate": 3.814998844464988e-05, + "loss": 0.0001, + "step": 6153 + }, + { + "epoch": 1.42, + "learning_rate": 3.8148062552962025e-05, + "loss": 0.0, + "step": 6154 + }, + { + "epoch": 1.42, + "learning_rate": 3.814613666127417e-05, + "loss": 0.0, + "step": 6155 + }, + { + "epoch": 1.42, + "learning_rate": 3.814421076958632e-05, + "loss": 0.0, + "step": 6156 + }, + { + "epoch": 1.42, + "learning_rate": 3.8142284877898474e-05, + "loss": 0.0, + "step": 6157 + }, + { + "epoch": 1.42, + "learning_rate": 3.8140358986210615e-05, + "loss": 0.0, + "step": 6158 + }, + { + "epoch": 1.42, + "learning_rate": 3.813843309452277e-05, + "loss": 0.0, + "step": 6159 + }, + { + "epoch": 1.42, + "learning_rate": 3.8136507202834916e-05, + "loss": 0.0001, + "step": 6160 + }, + { + "epoch": 1.42, + "learning_rate": 3.8134581311147064e-05, + "loss": 0.0, + "step": 6161 + }, + { + "epoch": 1.42, + "learning_rate": 3.813265541945921e-05, + "loss": 0.0, + "step": 6162 + }, + { + "epoch": 1.42, + "learning_rate": 3.813072952777136e-05, + "loss": 0.0, + "step": 6163 + }, + { + "epoch": 1.42, + "learning_rate": 3.812880363608351e-05, + "loss": 0.0, + "step": 6164 + }, + { + "epoch": 1.42, + "learning_rate": 3.8126877744395654e-05, + "loss": 0.0001, + "step": 6165 + }, + { + "epoch": 1.43, + "learning_rate": 3.812495185270781e-05, + "loss": 0.0, + "step": 6166 + }, + { + "epoch": 1.43, + "learning_rate": 3.8123025961019956e-05, + "loss": 0.0, + "step": 6167 + }, + { + "epoch": 1.43, + "learning_rate": 3.81211000693321e-05, + "loss": 0.0002, + "step": 6168 + }, + { + "epoch": 1.43, + "learning_rate": 3.811917417764425e-05, + "loss": 0.0007, + "step": 6169 + }, + { + "epoch": 1.43, + "learning_rate": 3.81172482859564e-05, + "loss": 0.0, + "step": 6170 + }, + { + "epoch": 1.43, + "learning_rate": 3.8115322394268546e-05, + "loss": 0.0, + "step": 6171 + }, + { + "epoch": 1.43, + "learning_rate": 3.811339650258069e-05, + "loss": 0.0005, + "step": 6172 + }, + { + "epoch": 1.43, + "learning_rate": 3.811147061089285e-05, + "loss": 0.0001, + "step": 6173 + }, + { + "epoch": 1.43, + "learning_rate": 3.8109544719204995e-05, + "loss": 0.0, + "step": 6174 + }, + { + "epoch": 1.43, + "learning_rate": 3.810761882751714e-05, + "loss": 0.0002, + "step": 6175 + }, + { + "epoch": 1.43, + "learning_rate": 3.810569293582929e-05, + "loss": 0.0025, + "step": 6176 + }, + { + "epoch": 1.43, + "learning_rate": 3.8103767044141444e-05, + "loss": 0.0001, + "step": 6177 + }, + { + "epoch": 1.43, + "learning_rate": 3.8101841152453585e-05, + "loss": 0.0003, + "step": 6178 + }, + { + "epoch": 1.43, + "learning_rate": 3.809991526076573e-05, + "loss": 0.0, + "step": 6179 + }, + { + "epoch": 1.43, + "learning_rate": 3.8097989369077886e-05, + "loss": 0.0012, + "step": 6180 + }, + { + "epoch": 1.43, + "learning_rate": 3.8096063477390034e-05, + "loss": 0.0, + "step": 6181 + }, + { + "epoch": 1.43, + "learning_rate": 3.809413758570218e-05, + "loss": 0.0, + "step": 6182 + }, + { + "epoch": 1.43, + "learning_rate": 3.809221169401433e-05, + "loss": 0.0, + "step": 6183 + }, + { + "epoch": 1.43, + "learning_rate": 3.809028580232648e-05, + "loss": 0.0004, + "step": 6184 + }, + { + "epoch": 1.43, + "learning_rate": 3.8088359910638624e-05, + "loss": 0.0, + "step": 6185 + }, + { + "epoch": 1.43, + "learning_rate": 3.808643401895078e-05, + "loss": 0.0, + "step": 6186 + }, + { + "epoch": 1.43, + "learning_rate": 3.8084508127262926e-05, + "loss": 0.0, + "step": 6187 + }, + { + "epoch": 1.43, + "learning_rate": 3.808258223557507e-05, + "loss": 0.0, + "step": 6188 + }, + { + "epoch": 1.43, + "learning_rate": 3.808065634388722e-05, + "loss": 0.0, + "step": 6189 + }, + { + "epoch": 1.43, + "learning_rate": 3.807873045219937e-05, + "loss": 0.0005, + "step": 6190 + }, + { + "epoch": 1.43, + "learning_rate": 3.807680456051152e-05, + "loss": 0.0003, + "step": 6191 + }, + { + "epoch": 1.43, + "learning_rate": 3.807487866882366e-05, + "loss": 0.0, + "step": 6192 + }, + { + "epoch": 1.43, + "learning_rate": 3.807295277713582e-05, + "loss": 0.0001, + "step": 6193 + }, + { + "epoch": 1.43, + "learning_rate": 3.8071026885447965e-05, + "loss": 0.0005, + "step": 6194 + }, + { + "epoch": 1.43, + "learning_rate": 3.806910099376011e-05, + "loss": 0.0007, + "step": 6195 + }, + { + "epoch": 1.43, + "learning_rate": 3.806717510207226e-05, + "loss": 0.0002, + "step": 6196 + }, + { + "epoch": 1.43, + "learning_rate": 3.806524921038441e-05, + "loss": 0.0, + "step": 6197 + }, + { + "epoch": 1.43, + "learning_rate": 3.806332331869656e-05, + "loss": 0.0001, + "step": 6198 + }, + { + "epoch": 1.43, + "learning_rate": 3.80613974270087e-05, + "loss": 0.0, + "step": 6199 + }, + { + "epoch": 1.43, + "learning_rate": 3.8059471535320856e-05, + "loss": 0.0001, + "step": 6200 + }, + { + "epoch": 1.43, + "learning_rate": 3.8057545643633004e-05, + "loss": 0.0004, + "step": 6201 + }, + { + "epoch": 1.43, + "learning_rate": 3.805561975194515e-05, + "loss": 0.0002, + "step": 6202 + }, + { + "epoch": 1.43, + "learning_rate": 3.80536938602573e-05, + "loss": 0.0, + "step": 6203 + }, + { + "epoch": 1.43, + "learning_rate": 3.805176796856945e-05, + "loss": 0.0003, + "step": 6204 + }, + { + "epoch": 1.43, + "learning_rate": 3.80498420768816e-05, + "loss": 0.0, + "step": 6205 + }, + { + "epoch": 1.43, + "learning_rate": 3.804791618519374e-05, + "loss": 0.0001, + "step": 6206 + }, + { + "epoch": 1.43, + "learning_rate": 3.8045990293505895e-05, + "loss": 0.0, + "step": 6207 + }, + { + "epoch": 1.43, + "learning_rate": 3.804406440181804e-05, + "loss": 0.0, + "step": 6208 + }, + { + "epoch": 1.43, + "learning_rate": 3.804213851013019e-05, + "loss": 0.0001, + "step": 6209 + }, + { + "epoch": 1.44, + "learning_rate": 3.804021261844234e-05, + "loss": 0.0, + "step": 6210 + }, + { + "epoch": 1.44, + "learning_rate": 3.803828672675449e-05, + "loss": 0.0003, + "step": 6211 + }, + { + "epoch": 1.44, + "learning_rate": 3.803636083506664e-05, + "loss": 0.0, + "step": 6212 + }, + { + "epoch": 1.44, + "learning_rate": 3.803443494337879e-05, + "loss": 0.0003, + "step": 6213 + }, + { + "epoch": 1.44, + "learning_rate": 3.8032509051690935e-05, + "loss": 0.0002, + "step": 6214 + }, + { + "epoch": 1.44, + "learning_rate": 3.803058316000309e-05, + "loss": 0.0002, + "step": 6215 + }, + { + "epoch": 1.44, + "learning_rate": 3.802865726831523e-05, + "loss": 0.0, + "step": 6216 + }, + { + "epoch": 1.44, + "learning_rate": 3.802673137662738e-05, + "loss": 0.0002, + "step": 6217 + }, + { + "epoch": 1.44, + "learning_rate": 3.802480548493953e-05, + "loss": 0.0001, + "step": 6218 + }, + { + "epoch": 1.44, + "learning_rate": 3.802287959325168e-05, + "loss": 0.0001, + "step": 6219 + }, + { + "epoch": 1.44, + "learning_rate": 3.8020953701563826e-05, + "loss": 0.0, + "step": 6220 + }, + { + "epoch": 1.44, + "learning_rate": 3.8019027809875974e-05, + "loss": 0.0, + "step": 6221 + }, + { + "epoch": 1.44, + "learning_rate": 3.801710191818813e-05, + "loss": 0.0, + "step": 6222 + }, + { + "epoch": 1.44, + "learning_rate": 3.801517602650027e-05, + "loss": 0.0002, + "step": 6223 + }, + { + "epoch": 1.44, + "learning_rate": 3.801325013481242e-05, + "loss": 0.0001, + "step": 6224 + }, + { + "epoch": 1.44, + "learning_rate": 3.801132424312457e-05, + "loss": 0.0006, + "step": 6225 + }, + { + "epoch": 1.44, + "learning_rate": 3.800939835143672e-05, + "loss": 0.0, + "step": 6226 + }, + { + "epoch": 1.44, + "learning_rate": 3.8007472459748865e-05, + "loss": 0.0001, + "step": 6227 + }, + { + "epoch": 1.44, + "learning_rate": 3.800554656806101e-05, + "loss": 0.0012, + "step": 6228 + }, + { + "epoch": 1.44, + "learning_rate": 3.800362067637317e-05, + "loss": 0.0, + "step": 6229 + }, + { + "epoch": 1.44, + "learning_rate": 3.800169478468531e-05, + "loss": 0.0, + "step": 6230 + }, + { + "epoch": 1.44, + "learning_rate": 3.799976889299746e-05, + "loss": 0.0012, + "step": 6231 + }, + { + "epoch": 1.44, + "learning_rate": 3.799784300130961e-05, + "loss": 0.0002, + "step": 6232 + }, + { + "epoch": 1.44, + "learning_rate": 3.799591710962176e-05, + "loss": 0.0006, + "step": 6233 + }, + { + "epoch": 1.44, + "learning_rate": 3.7993991217933904e-05, + "loss": 0.0, + "step": 6234 + }, + { + "epoch": 1.44, + "learning_rate": 3.799206532624605e-05, + "loss": 0.0001, + "step": 6235 + }, + { + "epoch": 1.44, + "learning_rate": 3.7990139434558206e-05, + "loss": 0.0, + "step": 6236 + }, + { + "epoch": 1.44, + "learning_rate": 3.798821354287035e-05, + "loss": 0.0003, + "step": 6237 + }, + { + "epoch": 1.44, + "learning_rate": 3.79862876511825e-05, + "loss": 0.0001, + "step": 6238 + }, + { + "epoch": 1.44, + "learning_rate": 3.798436175949465e-05, + "loss": 0.0, + "step": 6239 + }, + { + "epoch": 1.44, + "learning_rate": 3.7982435867806796e-05, + "loss": 0.0, + "step": 6240 + }, + { + "epoch": 1.44, + "learning_rate": 3.7980509976118944e-05, + "loss": 0.0, + "step": 6241 + }, + { + "epoch": 1.44, + "learning_rate": 3.79785840844311e-05, + "loss": 0.0001, + "step": 6242 + }, + { + "epoch": 1.44, + "learning_rate": 3.7976658192743245e-05, + "loss": 0.0072, + "step": 6243 + }, + { + "epoch": 1.44, + "learning_rate": 3.7974732301055386e-05, + "loss": 0.0, + "step": 6244 + }, + { + "epoch": 1.44, + "learning_rate": 3.797280640936754e-05, + "loss": 0.0022, + "step": 6245 + }, + { + "epoch": 1.44, + "learning_rate": 3.797088051767969e-05, + "loss": 0.0009, + "step": 6246 + }, + { + "epoch": 1.44, + "learning_rate": 3.7968954625991835e-05, + "loss": 0.0002, + "step": 6247 + }, + { + "epoch": 1.44, + "learning_rate": 3.796702873430398e-05, + "loss": 0.0001, + "step": 6248 + }, + { + "epoch": 1.44, + "learning_rate": 3.796510284261614e-05, + "loss": 0.0, + "step": 6249 + }, + { + "epoch": 1.44, + "learning_rate": 3.796317695092828e-05, + "loss": 0.0, + "step": 6250 + }, + { + "epoch": 1.44, + "learning_rate": 3.796125105924043e-05, + "loss": 0.0, + "step": 6251 + }, + { + "epoch": 1.44, + "learning_rate": 3.795932516755258e-05, + "loss": 0.0, + "step": 6252 + }, + { + "epoch": 1.45, + "learning_rate": 3.795739927586473e-05, + "loss": 0.0, + "step": 6253 + }, + { + "epoch": 1.45, + "learning_rate": 3.7955473384176874e-05, + "loss": 0.0004, + "step": 6254 + }, + { + "epoch": 1.45, + "learning_rate": 3.795354749248902e-05, + "loss": 0.0, + "step": 6255 + }, + { + "epoch": 1.45, + "learning_rate": 3.7951621600801176e-05, + "loss": 0.0007, + "step": 6256 + }, + { + "epoch": 1.45, + "learning_rate": 3.794969570911332e-05, + "loss": 0.0001, + "step": 6257 + }, + { + "epoch": 1.45, + "learning_rate": 3.794776981742547e-05, + "loss": 0.0, + "step": 6258 + }, + { + "epoch": 1.45, + "learning_rate": 3.794584392573762e-05, + "loss": 0.0009, + "step": 6259 + }, + { + "epoch": 1.45, + "learning_rate": 3.7943918034049766e-05, + "loss": 0.0, + "step": 6260 + }, + { + "epoch": 1.45, + "learning_rate": 3.7941992142361913e-05, + "loss": 0.0, + "step": 6261 + }, + { + "epoch": 1.45, + "learning_rate": 3.794006625067406e-05, + "loss": 0.0001, + "step": 6262 + }, + { + "epoch": 1.45, + "learning_rate": 3.7938140358986215e-05, + "loss": 0.0001, + "step": 6263 + }, + { + "epoch": 1.45, + "learning_rate": 3.7936214467298356e-05, + "loss": 0.0001, + "step": 6264 + }, + { + "epoch": 1.45, + "learning_rate": 3.793428857561051e-05, + "loss": 0.0, + "step": 6265 + }, + { + "epoch": 1.45, + "learning_rate": 3.793236268392266e-05, + "loss": 0.0, + "step": 6266 + }, + { + "epoch": 1.45, + "learning_rate": 3.7930436792234805e-05, + "loss": 0.0001, + "step": 6267 + }, + { + "epoch": 1.45, + "learning_rate": 3.792851090054695e-05, + "loss": 0.0, + "step": 6268 + }, + { + "epoch": 1.45, + "learning_rate": 3.792658500885911e-05, + "loss": 0.0002, + "step": 6269 + }, + { + "epoch": 1.45, + "learning_rate": 3.7924659117171254e-05, + "loss": 0.0003, + "step": 6270 + }, + { + "epoch": 1.45, + "learning_rate": 3.79227332254834e-05, + "loss": 0.0006, + "step": 6271 + }, + { + "epoch": 1.45, + "learning_rate": 3.792080733379555e-05, + "loss": 0.0006, + "step": 6272 + }, + { + "epoch": 1.45, + "learning_rate": 3.79188814421077e-05, + "loss": 0.0007, + "step": 6273 + }, + { + "epoch": 1.45, + "learning_rate": 3.7916955550419844e-05, + "loss": 0.0004, + "step": 6274 + }, + { + "epoch": 1.45, + "learning_rate": 3.791502965873199e-05, + "loss": 0.0, + "step": 6275 + }, + { + "epoch": 1.45, + "learning_rate": 3.7913103767044146e-05, + "loss": 0.0, + "step": 6276 + }, + { + "epoch": 1.45, + "learning_rate": 3.791117787535629e-05, + "loss": 0.0, + "step": 6277 + }, + { + "epoch": 1.45, + "learning_rate": 3.790925198366844e-05, + "loss": 0.0, + "step": 6278 + }, + { + "epoch": 1.45, + "learning_rate": 3.790732609198059e-05, + "loss": 0.0001, + "step": 6279 + }, + { + "epoch": 1.45, + "learning_rate": 3.790540020029274e-05, + "loss": 0.0, + "step": 6280 + }, + { + "epoch": 1.45, + "learning_rate": 3.790347430860488e-05, + "loss": 0.0, + "step": 6281 + }, + { + "epoch": 1.45, + "learning_rate": 3.790154841691703e-05, + "loss": 0.0, + "step": 6282 + }, + { + "epoch": 1.45, + "learning_rate": 3.7899622525229185e-05, + "loss": 0.0002, + "step": 6283 + }, + { + "epoch": 1.45, + "learning_rate": 3.789769663354133e-05, + "loss": 0.0002, + "step": 6284 + }, + { + "epoch": 1.45, + "learning_rate": 3.789577074185348e-05, + "loss": 0.0, + "step": 6285 + }, + { + "epoch": 1.45, + "learning_rate": 3.789384485016563e-05, + "loss": 0.0, + "step": 6286 + }, + { + "epoch": 1.45, + "learning_rate": 3.789191895847778e-05, + "loss": 0.0, + "step": 6287 + }, + { + "epoch": 1.45, + "learning_rate": 3.788999306678992e-05, + "loss": 0.0001, + "step": 6288 + }, + { + "epoch": 1.45, + "learning_rate": 3.788806717510208e-05, + "loss": 0.0002, + "step": 6289 + }, + { + "epoch": 1.45, + "learning_rate": 3.7886141283414224e-05, + "loss": 0.0007, + "step": 6290 + }, + { + "epoch": 1.45, + "learning_rate": 3.788421539172637e-05, + "loss": 0.0001, + "step": 6291 + }, + { + "epoch": 1.45, + "learning_rate": 3.788228950003852e-05, + "loss": 0.0, + "step": 6292 + }, + { + "epoch": 1.45, + "learning_rate": 3.7880363608350667e-05, + "loss": 0.0002, + "step": 6293 + }, + { + "epoch": 1.45, + "learning_rate": 3.787843771666282e-05, + "loss": 0.0001, + "step": 6294 + }, + { + "epoch": 1.45, + "learning_rate": 3.787651182497496e-05, + "loss": 0.0, + "step": 6295 + }, + { + "epoch": 1.46, + "learning_rate": 3.7874585933287116e-05, + "loss": 0.0001, + "step": 6296 + }, + { + "epoch": 1.46, + "learning_rate": 3.787266004159926e-05, + "loss": 0.0001, + "step": 6297 + }, + { + "epoch": 1.46, + "learning_rate": 3.787073414991141e-05, + "loss": 0.0, + "step": 6298 + }, + { + "epoch": 1.46, + "learning_rate": 3.786880825822356e-05, + "loss": 0.0001, + "step": 6299 + }, + { + "epoch": 1.46, + "learning_rate": 3.7866882366535706e-05, + "loss": 0.0, + "step": 6300 + }, + { + "epoch": 1.46, + "learning_rate": 3.786495647484786e-05, + "loss": 0.0002, + "step": 6301 + }, + { + "epoch": 1.46, + "learning_rate": 3.786303058316e-05, + "loss": 0.0002, + "step": 6302 + }, + { + "epoch": 1.46, + "learning_rate": 3.7861104691472155e-05, + "loss": 0.0001, + "step": 6303 + }, + { + "epoch": 1.46, + "learning_rate": 3.78591787997843e-05, + "loss": 0.0006, + "step": 6304 + }, + { + "epoch": 1.46, + "learning_rate": 3.785725290809645e-05, + "loss": 0.0, + "step": 6305 + }, + { + "epoch": 1.46, + "learning_rate": 3.78553270164086e-05, + "loss": 0.0, + "step": 6306 + }, + { + "epoch": 1.46, + "learning_rate": 3.785340112472075e-05, + "loss": 0.0, + "step": 6307 + }, + { + "epoch": 1.46, + "learning_rate": 3.78514752330329e-05, + "loss": 0.0001, + "step": 6308 + }, + { + "epoch": 1.46, + "learning_rate": 3.784954934134504e-05, + "loss": 0.0001, + "step": 6309 + }, + { + "epoch": 1.46, + "learning_rate": 3.7847623449657194e-05, + "loss": 0.0, + "step": 6310 + }, + { + "epoch": 1.46, + "learning_rate": 3.784569755796934e-05, + "loss": 0.0, + "step": 6311 + }, + { + "epoch": 1.46, + "learning_rate": 3.784377166628149e-05, + "loss": 0.0001, + "step": 6312 + }, + { + "epoch": 1.46, + "learning_rate": 3.7841845774593636e-05, + "loss": 0.0001, + "step": 6313 + }, + { + "epoch": 1.46, + "learning_rate": 3.783991988290579e-05, + "loss": 0.0001, + "step": 6314 + }, + { + "epoch": 1.46, + "learning_rate": 3.783799399121794e-05, + "loss": 0.0001, + "step": 6315 + }, + { + "epoch": 1.46, + "learning_rate": 3.7836068099530086e-05, + "loss": 0.0001, + "step": 6316 + }, + { + "epoch": 1.46, + "learning_rate": 3.783414220784223e-05, + "loss": 0.0002, + "step": 6317 + }, + { + "epoch": 1.46, + "learning_rate": 3.783221631615438e-05, + "loss": 0.0, + "step": 6318 + }, + { + "epoch": 1.46, + "learning_rate": 3.783029042446653e-05, + "loss": 0.0001, + "step": 6319 + }, + { + "epoch": 1.46, + "learning_rate": 3.7828364532778676e-05, + "loss": 0.0, + "step": 6320 + }, + { + "epoch": 1.46, + "learning_rate": 3.782643864109083e-05, + "loss": 0.0005, + "step": 6321 + }, + { + "epoch": 1.46, + "learning_rate": 3.782451274940297e-05, + "loss": 0.0, + "step": 6322 + }, + { + "epoch": 1.46, + "learning_rate": 3.7822586857715125e-05, + "loss": 0.0001, + "step": 6323 + }, + { + "epoch": 1.46, + "learning_rate": 3.782066096602727e-05, + "loss": 0.0, + "step": 6324 + }, + { + "epoch": 1.46, + "learning_rate": 3.781873507433942e-05, + "loss": 0.0, + "step": 6325 + }, + { + "epoch": 1.46, + "learning_rate": 3.781680918265157e-05, + "loss": 0.0009, + "step": 6326 + }, + { + "epoch": 1.46, + "learning_rate": 3.781488329096372e-05, + "loss": 0.0, + "step": 6327 + }, + { + "epoch": 1.46, + "learning_rate": 3.781295739927587e-05, + "loss": 0.0002, + "step": 6328 + }, + { + "epoch": 1.46, + "learning_rate": 3.781103150758801e-05, + "loss": 0.0004, + "step": 6329 + }, + { + "epoch": 1.46, + "learning_rate": 3.7809105615900164e-05, + "loss": 0.0, + "step": 6330 + }, + { + "epoch": 1.46, + "learning_rate": 3.780717972421231e-05, + "loss": 0.0645, + "step": 6331 + }, + { + "epoch": 1.46, + "learning_rate": 3.780525383252446e-05, + "loss": 0.0001, + "step": 6332 + }, + { + "epoch": 1.46, + "learning_rate": 3.7803327940836606e-05, + "loss": 0.0001, + "step": 6333 + }, + { + "epoch": 1.46, + "learning_rate": 3.780140204914876e-05, + "loss": 0.0, + "step": 6334 + }, + { + "epoch": 1.46, + "learning_rate": 3.779947615746091e-05, + "loss": 0.0, + "step": 6335 + }, + { + "epoch": 1.46, + "learning_rate": 3.7797550265773056e-05, + "loss": 0.0, + "step": 6336 + }, + { + "epoch": 1.46, + "learning_rate": 3.77956243740852e-05, + "loss": 0.0, + "step": 6337 + }, + { + "epoch": 1.46, + "learning_rate": 3.779369848239735e-05, + "loss": 0.0, + "step": 6338 + }, + { + "epoch": 1.46, + "learning_rate": 3.77917725907095e-05, + "loss": 0.0, + "step": 6339 + }, + { + "epoch": 1.47, + "learning_rate": 3.7789846699021645e-05, + "loss": 0.0, + "step": 6340 + }, + { + "epoch": 1.47, + "learning_rate": 3.77879208073338e-05, + "loss": 0.0, + "step": 6341 + }, + { + "epoch": 1.47, + "learning_rate": 3.778599491564595e-05, + "loss": 0.0, + "step": 6342 + }, + { + "epoch": 1.47, + "learning_rate": 3.7784069023958095e-05, + "loss": 0.0001, + "step": 6343 + }, + { + "epoch": 1.47, + "learning_rate": 3.778214313227024e-05, + "loss": 0.0, + "step": 6344 + }, + { + "epoch": 1.47, + "learning_rate": 3.7780217240582396e-05, + "loss": 0.0002, + "step": 6345 + }, + { + "epoch": 1.47, + "learning_rate": 3.777829134889454e-05, + "loss": 0.0004, + "step": 6346 + }, + { + "epoch": 1.47, + "learning_rate": 3.7776365457206685e-05, + "loss": 0.0, + "step": 6347 + }, + { + "epoch": 1.47, + "learning_rate": 3.777443956551884e-05, + "loss": 0.0, + "step": 6348 + }, + { + "epoch": 1.47, + "learning_rate": 3.7772513673830986e-05, + "loss": 0.0045, + "step": 6349 + }, + { + "epoch": 1.47, + "learning_rate": 3.7770587782143134e-05, + "loss": 0.0, + "step": 6350 + }, + { + "epoch": 1.47, + "learning_rate": 3.776866189045528e-05, + "loss": 0.0, + "step": 6351 + }, + { + "epoch": 1.47, + "learning_rate": 3.7766735998767436e-05, + "loss": 0.0, + "step": 6352 + }, + { + "epoch": 1.47, + "learning_rate": 3.7764810107079576e-05, + "loss": 0.0, + "step": 6353 + }, + { + "epoch": 1.47, + "learning_rate": 3.776288421539173e-05, + "loss": 0.0, + "step": 6354 + }, + { + "epoch": 1.47, + "learning_rate": 3.776095832370388e-05, + "loss": 0.0001, + "step": 6355 + }, + { + "epoch": 1.47, + "learning_rate": 3.7759032432016025e-05, + "loss": 0.0, + "step": 6356 + }, + { + "epoch": 1.47, + "learning_rate": 3.775710654032817e-05, + "loss": 0.0001, + "step": 6357 + }, + { + "epoch": 1.47, + "learning_rate": 3.775518064864032e-05, + "loss": 0.0001, + "step": 6358 + }, + { + "epoch": 1.47, + "learning_rate": 3.7753254756952475e-05, + "loss": 0.0, + "step": 6359 + }, + { + "epoch": 1.47, + "learning_rate": 3.7751328865264615e-05, + "loss": 0.0, + "step": 6360 + }, + { + "epoch": 1.47, + "learning_rate": 3.774940297357677e-05, + "loss": 0.0, + "step": 6361 + }, + { + "epoch": 1.47, + "learning_rate": 3.774747708188892e-05, + "loss": 0.0002, + "step": 6362 + }, + { + "epoch": 1.47, + "learning_rate": 3.7745551190201065e-05, + "loss": 0.0002, + "step": 6363 + }, + { + "epoch": 1.47, + "learning_rate": 3.774362529851321e-05, + "loss": 0.0, + "step": 6364 + }, + { + "epoch": 1.47, + "learning_rate": 3.774169940682536e-05, + "loss": 0.0, + "step": 6365 + }, + { + "epoch": 1.47, + "learning_rate": 3.7739773515137514e-05, + "loss": 0.0006, + "step": 6366 + }, + { + "epoch": 1.47, + "learning_rate": 3.7737847623449654e-05, + "loss": 0.0, + "step": 6367 + }, + { + "epoch": 1.47, + "learning_rate": 3.773592173176181e-05, + "loss": 0.0005, + "step": 6368 + }, + { + "epoch": 1.47, + "learning_rate": 3.7733995840073956e-05, + "loss": 0.0002, + "step": 6369 + }, + { + "epoch": 1.47, + "learning_rate": 3.7732069948386104e-05, + "loss": 0.0, + "step": 6370 + }, + { + "epoch": 1.47, + "learning_rate": 3.773014405669825e-05, + "loss": 0.0, + "step": 6371 + }, + { + "epoch": 1.47, + "learning_rate": 3.7728218165010405e-05, + "loss": 0.0001, + "step": 6372 + }, + { + "epoch": 1.47, + "learning_rate": 3.772629227332255e-05, + "loss": 0.0001, + "step": 6373 + }, + { + "epoch": 1.47, + "learning_rate": 3.7724366381634694e-05, + "loss": 0.0001, + "step": 6374 + }, + { + "epoch": 1.47, + "learning_rate": 3.772244048994685e-05, + "loss": 0.0001, + "step": 6375 + }, + { + "epoch": 1.47, + "learning_rate": 3.7720514598258995e-05, + "loss": 0.0, + "step": 6376 + }, + { + "epoch": 1.47, + "learning_rate": 3.771858870657114e-05, + "loss": 0.0, + "step": 6377 + }, + { + "epoch": 1.47, + "learning_rate": 3.771666281488329e-05, + "loss": 0.0, + "step": 6378 + }, + { + "epoch": 1.47, + "learning_rate": 3.7714736923195445e-05, + "loss": 0.0001, + "step": 6379 + }, + { + "epoch": 1.47, + "learning_rate": 3.771281103150759e-05, + "loss": 0.0017, + "step": 6380 + }, + { + "epoch": 1.47, + "learning_rate": 3.771088513981974e-05, + "loss": 0.0001, + "step": 6381 + }, + { + "epoch": 1.47, + "learning_rate": 3.770895924813189e-05, + "loss": 0.0002, + "step": 6382 + }, + { + "epoch": 1.48, + "learning_rate": 3.770703335644404e-05, + "loss": 0.0008, + "step": 6383 + }, + { + "epoch": 1.48, + "learning_rate": 3.770510746475618e-05, + "loss": 0.0, + "step": 6384 + }, + { + "epoch": 1.48, + "learning_rate": 3.770318157306833e-05, + "loss": 0.0, + "step": 6385 + }, + { + "epoch": 1.48, + "learning_rate": 3.7701255681380484e-05, + "loss": 0.0, + "step": 6386 + }, + { + "epoch": 1.48, + "learning_rate": 3.769932978969263e-05, + "loss": 0.0034, + "step": 6387 + }, + { + "epoch": 1.48, + "learning_rate": 3.769740389800478e-05, + "loss": 0.0, + "step": 6388 + }, + { + "epoch": 1.48, + "learning_rate": 3.7695478006316926e-05, + "loss": 0.0001, + "step": 6389 + }, + { + "epoch": 1.48, + "learning_rate": 3.769355211462908e-05, + "loss": 0.0001, + "step": 6390 + }, + { + "epoch": 1.48, + "learning_rate": 3.769162622294122e-05, + "loss": 0.0001, + "step": 6391 + }, + { + "epoch": 1.48, + "learning_rate": 3.7689700331253375e-05, + "loss": 0.0, + "step": 6392 + }, + { + "epoch": 1.48, + "learning_rate": 3.768777443956552e-05, + "loss": 0.0003, + "step": 6393 + }, + { + "epoch": 1.48, + "learning_rate": 3.7685848547877663e-05, + "loss": 0.0001, + "step": 6394 + }, + { + "epoch": 1.48, + "learning_rate": 3.768392265618982e-05, + "loss": 0.0, + "step": 6395 + }, + { + "epoch": 1.48, + "learning_rate": 3.7681996764501965e-05, + "loss": 0.0001, + "step": 6396 + }, + { + "epoch": 1.48, + "learning_rate": 3.768007087281412e-05, + "loss": 0.0, + "step": 6397 + }, + { + "epoch": 1.48, + "learning_rate": 3.767814498112626e-05, + "loss": 0.0, + "step": 6398 + }, + { + "epoch": 1.48, + "learning_rate": 3.7676219089438414e-05, + "loss": 0.0004, + "step": 6399 + }, + { + "epoch": 1.48, + "learning_rate": 3.767429319775056e-05, + "loss": 0.0, + "step": 6400 + }, + { + "epoch": 1.48, + "learning_rate": 3.767236730606271e-05, + "loss": 0.0, + "step": 6401 + }, + { + "epoch": 1.48, + "learning_rate": 3.767044141437486e-05, + "loss": 0.0, + "step": 6402 + }, + { + "epoch": 1.48, + "learning_rate": 3.7668515522687004e-05, + "loss": 0.003, + "step": 6403 + }, + { + "epoch": 1.48, + "learning_rate": 3.766658963099915e-05, + "loss": 0.0001, + "step": 6404 + }, + { + "epoch": 1.48, + "learning_rate": 3.76646637393113e-05, + "loss": 0.0001, + "step": 6405 + }, + { + "epoch": 1.48, + "learning_rate": 3.7662737847623454e-05, + "loss": 0.0, + "step": 6406 + }, + { + "epoch": 1.48, + "learning_rate": 3.76608119559356e-05, + "loss": 0.0, + "step": 6407 + }, + { + "epoch": 1.48, + "learning_rate": 3.765888606424775e-05, + "loss": 0.0001, + "step": 6408 + }, + { + "epoch": 1.48, + "learning_rate": 3.7656960172559896e-05, + "loss": 0.0, + "step": 6409 + }, + { + "epoch": 1.48, + "learning_rate": 3.765503428087205e-05, + "loss": 0.0001, + "step": 6410 + }, + { + "epoch": 1.48, + "learning_rate": 3.765310838918419e-05, + "loss": 0.0006, + "step": 6411 + }, + { + "epoch": 1.48, + "learning_rate": 3.765118249749634e-05, + "loss": 0.0139, + "step": 6412 + }, + { + "epoch": 1.48, + "learning_rate": 3.764925660580849e-05, + "loss": 0.0004, + "step": 6413 + }, + { + "epoch": 1.48, + "learning_rate": 3.764733071412064e-05, + "loss": 0.0001, + "step": 6414 + }, + { + "epoch": 1.48, + "learning_rate": 3.764540482243279e-05, + "loss": 0.0003, + "step": 6415 + }, + { + "epoch": 1.48, + "learning_rate": 3.7643478930744935e-05, + "loss": 0.0, + "step": 6416 + }, + { + "epoch": 1.48, + "learning_rate": 3.764155303905709e-05, + "loss": 0.0, + "step": 6417 + }, + { + "epoch": 1.48, + "learning_rate": 3.763962714736923e-05, + "loss": 0.0004, + "step": 6418 + }, + { + "epoch": 1.48, + "learning_rate": 3.7637701255681384e-05, + "loss": 0.0002, + "step": 6419 + }, + { + "epoch": 1.48, + "learning_rate": 3.763577536399353e-05, + "loss": 0.0, + "step": 6420 + }, + { + "epoch": 1.48, + "learning_rate": 3.763384947230568e-05, + "loss": 0.0, + "step": 6421 + }, + { + "epoch": 1.48, + "learning_rate": 3.763192358061783e-05, + "loss": 0.0004, + "step": 6422 + }, + { + "epoch": 1.48, + "learning_rate": 3.7629997688929974e-05, + "loss": 0.0, + "step": 6423 + }, + { + "epoch": 1.48, + "learning_rate": 3.762807179724213e-05, + "loss": 0.0001, + "step": 6424 + }, + { + "epoch": 1.48, + "learning_rate": 3.762614590555427e-05, + "loss": 0.0, + "step": 6425 + }, + { + "epoch": 1.49, + "learning_rate": 3.7624220013866423e-05, + "loss": 0.0, + "step": 6426 + }, + { + "epoch": 1.49, + "learning_rate": 3.762229412217857e-05, + "loss": 0.0001, + "step": 6427 + }, + { + "epoch": 1.49, + "learning_rate": 3.762036823049072e-05, + "loss": 0.0, + "step": 6428 + }, + { + "epoch": 1.49, + "learning_rate": 3.7618442338802866e-05, + "loss": 0.0, + "step": 6429 + }, + { + "epoch": 1.49, + "learning_rate": 3.761651644711501e-05, + "loss": 0.0003, + "step": 6430 + }, + { + "epoch": 1.49, + "learning_rate": 3.761459055542717e-05, + "loss": 0.0, + "step": 6431 + }, + { + "epoch": 1.49, + "learning_rate": 3.761266466373931e-05, + "loss": 0.0008, + "step": 6432 + }, + { + "epoch": 1.49, + "learning_rate": 3.761073877205146e-05, + "loss": 0.0001, + "step": 6433 + }, + { + "epoch": 1.49, + "learning_rate": 3.760881288036361e-05, + "loss": 0.0114, + "step": 6434 + }, + { + "epoch": 1.49, + "learning_rate": 3.760688698867576e-05, + "loss": 0.0, + "step": 6435 + }, + { + "epoch": 1.49, + "learning_rate": 3.7604961096987905e-05, + "loss": 0.0, + "step": 6436 + }, + { + "epoch": 1.49, + "learning_rate": 3.760303520530006e-05, + "loss": 0.0001, + "step": 6437 + }, + { + "epoch": 1.49, + "learning_rate": 3.760110931361221e-05, + "loss": 0.0, + "step": 6438 + }, + { + "epoch": 1.49, + "learning_rate": 3.7599183421924354e-05, + "loss": 0.0, + "step": 6439 + }, + { + "epoch": 1.49, + "learning_rate": 3.75972575302365e-05, + "loss": 0.0, + "step": 6440 + }, + { + "epoch": 1.49, + "learning_rate": 3.759533163854865e-05, + "loss": 0.0004, + "step": 6441 + }, + { + "epoch": 1.49, + "learning_rate": 3.7593405746860797e-05, + "loss": 0.0001, + "step": 6442 + }, + { + "epoch": 1.49, + "learning_rate": 3.7591479855172944e-05, + "loss": 0.0, + "step": 6443 + }, + { + "epoch": 1.49, + "learning_rate": 3.75895539634851e-05, + "loss": 0.0, + "step": 6444 + }, + { + "epoch": 1.49, + "learning_rate": 3.7587628071797246e-05, + "loss": 0.0, + "step": 6445 + }, + { + "epoch": 1.49, + "learning_rate": 3.758570218010939e-05, + "loss": 0.0002, + "step": 6446 + }, + { + "epoch": 1.49, + "learning_rate": 3.758377628842154e-05, + "loss": 0.0, + "step": 6447 + }, + { + "epoch": 1.49, + "learning_rate": 3.7581850396733695e-05, + "loss": 0.0, + "step": 6448 + }, + { + "epoch": 1.49, + "learning_rate": 3.7579924505045836e-05, + "loss": 0.0, + "step": 6449 + }, + { + "epoch": 1.49, + "learning_rate": 3.757799861335798e-05, + "loss": 0.0, + "step": 6450 + }, + { + "epoch": 1.49, + "learning_rate": 3.757607272167014e-05, + "loss": 0.0005, + "step": 6451 + }, + { + "epoch": 1.49, + "learning_rate": 3.7574146829982285e-05, + "loss": 0.0, + "step": 6452 + }, + { + "epoch": 1.49, + "learning_rate": 3.757222093829443e-05, + "loss": 0.0002, + "step": 6453 + }, + { + "epoch": 1.49, + "learning_rate": 3.757029504660658e-05, + "loss": 0.0008, + "step": 6454 + }, + { + "epoch": 1.49, + "learning_rate": 3.7568369154918734e-05, + "loss": 0.0002, + "step": 6455 + }, + { + "epoch": 1.49, + "learning_rate": 3.7566443263230875e-05, + "loss": 0.0002, + "step": 6456 + }, + { + "epoch": 1.49, + "learning_rate": 3.756451737154303e-05, + "loss": 0.0, + "step": 6457 + }, + { + "epoch": 1.49, + "learning_rate": 3.7562591479855177e-05, + "loss": 0.0002, + "step": 6458 + }, + { + "epoch": 1.49, + "learning_rate": 3.7560665588167324e-05, + "loss": 0.0001, + "step": 6459 + }, + { + "epoch": 1.49, + "learning_rate": 3.755873969647947e-05, + "loss": 0.0001, + "step": 6460 + }, + { + "epoch": 1.49, + "learning_rate": 3.755681380479162e-05, + "loss": 0.0, + "step": 6461 + }, + { + "epoch": 1.49, + "learning_rate": 3.755488791310377e-05, + "loss": 0.0002, + "step": 6462 + }, + { + "epoch": 1.49, + "learning_rate": 3.7552962021415914e-05, + "loss": 0.0, + "step": 6463 + }, + { + "epoch": 1.49, + "learning_rate": 3.755103612972807e-05, + "loss": 0.0066, + "step": 6464 + }, + { + "epoch": 1.49, + "learning_rate": 3.7549110238040216e-05, + "loss": 0.0005, + "step": 6465 + }, + { + "epoch": 1.49, + "learning_rate": 3.754718434635236e-05, + "loss": 0.0002, + "step": 6466 + }, + { + "epoch": 1.49, + "learning_rate": 3.754525845466451e-05, + "loss": 0.0002, + "step": 6467 + }, + { + "epoch": 1.49, + "learning_rate": 3.754333256297666e-05, + "loss": 0.0, + "step": 6468 + }, + { + "epoch": 1.5, + "learning_rate": 3.754140667128881e-05, + "loss": 0.0, + "step": 6469 + }, + { + "epoch": 1.5, + "learning_rate": 3.753948077960095e-05, + "loss": 0.0003, + "step": 6470 + }, + { + "epoch": 1.5, + "learning_rate": 3.753755488791311e-05, + "loss": 0.0, + "step": 6471 + }, + { + "epoch": 1.5, + "learning_rate": 3.7535628996225255e-05, + "loss": 0.0028, + "step": 6472 + }, + { + "epoch": 1.5, + "learning_rate": 3.75337031045374e-05, + "loss": 0.0, + "step": 6473 + }, + { + "epoch": 1.5, + "learning_rate": 3.753177721284955e-05, + "loss": 0.0, + "step": 6474 + }, + { + "epoch": 1.5, + "learning_rate": 3.7529851321161704e-05, + "loss": 0.0, + "step": 6475 + }, + { + "epoch": 1.5, + "learning_rate": 3.7527925429473845e-05, + "loss": 0.0001, + "step": 6476 + }, + { + "epoch": 1.5, + "learning_rate": 3.752599953778599e-05, + "loss": 0.0, + "step": 6477 + }, + { + "epoch": 1.5, + "learning_rate": 3.7524073646098146e-05, + "loss": 0.0001, + "step": 6478 + }, + { + "epoch": 1.5, + "learning_rate": 3.7522147754410294e-05, + "loss": 0.0, + "step": 6479 + }, + { + "epoch": 1.5, + "learning_rate": 3.752022186272244e-05, + "loss": 0.0001, + "step": 6480 + }, + { + "epoch": 1.5, + "learning_rate": 3.751829597103459e-05, + "loss": 0.0, + "step": 6481 + }, + { + "epoch": 1.5, + "learning_rate": 3.751637007934674e-05, + "loss": 0.0, + "step": 6482 + }, + { + "epoch": 1.5, + "learning_rate": 3.7514444187658884e-05, + "loss": 0.0, + "step": 6483 + }, + { + "epoch": 1.5, + "learning_rate": 3.751251829597104e-05, + "loss": 0.0001, + "step": 6484 + }, + { + "epoch": 1.5, + "learning_rate": 3.7510592404283186e-05, + "loss": 0.0013, + "step": 6485 + }, + { + "epoch": 1.5, + "learning_rate": 3.750866651259533e-05, + "loss": 0.0, + "step": 6486 + }, + { + "epoch": 1.5, + "learning_rate": 3.750674062090748e-05, + "loss": 0.0001, + "step": 6487 + }, + { + "epoch": 1.5, + "learning_rate": 3.750481472921963e-05, + "loss": 0.0, + "step": 6488 + }, + { + "epoch": 1.5, + "learning_rate": 3.750288883753178e-05, + "loss": 0.0, + "step": 6489 + }, + { + "epoch": 1.5, + "learning_rate": 3.750096294584392e-05, + "loss": 0.001, + "step": 6490 + }, + { + "epoch": 1.5, + "learning_rate": 3.749903705415608e-05, + "loss": 0.0, + "step": 6491 + }, + { + "epoch": 1.5, + "learning_rate": 3.7497111162468225e-05, + "loss": 0.0, + "step": 6492 + }, + { + "epoch": 1.5, + "learning_rate": 3.749518527078037e-05, + "loss": 0.0004, + "step": 6493 + }, + { + "epoch": 1.5, + "learning_rate": 3.749325937909252e-05, + "loss": 0.0007, + "step": 6494 + }, + { + "epoch": 1.5, + "learning_rate": 3.7491333487404674e-05, + "loss": 0.0, + "step": 6495 + }, + { + "epoch": 1.5, + "learning_rate": 3.748940759571682e-05, + "loss": 0.0004, + "step": 6496 + }, + { + "epoch": 1.5, + "learning_rate": 3.748748170402896e-05, + "loss": 0.0005, + "step": 6497 + }, + { + "epoch": 1.5, + "learning_rate": 3.7485555812341116e-05, + "loss": 0.0006, + "step": 6498 + }, + { + "epoch": 1.5, + "learning_rate": 3.7483629920653264e-05, + "loss": 0.0, + "step": 6499 + }, + { + "epoch": 1.5, + "learning_rate": 3.748170402896541e-05, + "loss": 0.0, + "step": 6500 + }, + { + "epoch": 1.5, + "learning_rate": 3.747977813727756e-05, + "loss": 0.0, + "step": 6501 + }, + { + "epoch": 1.5, + "learning_rate": 3.747785224558971e-05, + "loss": 0.0009, + "step": 6502 + }, + { + "epoch": 1.5, + "learning_rate": 3.747592635390186e-05, + "loss": 0.0, + "step": 6503 + }, + { + "epoch": 1.5, + "learning_rate": 3.747400046221401e-05, + "loss": 0.0001, + "step": 6504 + }, + { + "epoch": 1.5, + "learning_rate": 3.7472074570526155e-05, + "loss": 0.0001, + "step": 6505 + }, + { + "epoch": 1.5, + "learning_rate": 3.74701486788383e-05, + "loss": 0.0, + "step": 6506 + }, + { + "epoch": 1.5, + "learning_rate": 3.746822278715045e-05, + "loss": 0.0016, + "step": 6507 + }, + { + "epoch": 1.5, + "learning_rate": 3.74662968954626e-05, + "loss": 0.0002, + "step": 6508 + }, + { + "epoch": 1.5, + "learning_rate": 3.746437100377475e-05, + "loss": 0.0, + "step": 6509 + }, + { + "epoch": 1.5, + "learning_rate": 3.74624451120869e-05, + "loss": 0.0007, + "step": 6510 + }, + { + "epoch": 1.5, + "learning_rate": 3.746051922039905e-05, + "loss": 0.0001, + "step": 6511 + }, + { + "epoch": 1.5, + "learning_rate": 3.7458593328711195e-05, + "loss": 0.0004, + "step": 6512 + }, + { + "epoch": 1.51, + "learning_rate": 3.745666743702335e-05, + "loss": 0.0017, + "step": 6513 + }, + { + "epoch": 1.51, + "learning_rate": 3.745474154533549e-05, + "loss": 0.0, + "step": 6514 + }, + { + "epoch": 1.51, + "learning_rate": 3.745281565364764e-05, + "loss": 0.0, + "step": 6515 + }, + { + "epoch": 1.51, + "learning_rate": 3.745088976195979e-05, + "loss": 0.0, + "step": 6516 + }, + { + "epoch": 1.51, + "learning_rate": 3.744896387027194e-05, + "loss": 0.0, + "step": 6517 + }, + { + "epoch": 1.51, + "learning_rate": 3.7447037978584086e-05, + "loss": 0.0, + "step": 6518 + }, + { + "epoch": 1.51, + "learning_rate": 3.7445112086896234e-05, + "loss": 0.0001, + "step": 6519 + }, + { + "epoch": 1.51, + "learning_rate": 3.744318619520839e-05, + "loss": 0.0, + "step": 6520 + }, + { + "epoch": 1.51, + "learning_rate": 3.744126030352053e-05, + "loss": 0.0, + "step": 6521 + }, + { + "epoch": 1.51, + "learning_rate": 3.743933441183268e-05, + "loss": 0.0, + "step": 6522 + }, + { + "epoch": 1.51, + "learning_rate": 3.743740852014483e-05, + "loss": 0.0, + "step": 6523 + }, + { + "epoch": 1.51, + "learning_rate": 3.743548262845698e-05, + "loss": 0.0, + "step": 6524 + }, + { + "epoch": 1.51, + "learning_rate": 3.7433556736769125e-05, + "loss": 0.0081, + "step": 6525 + }, + { + "epoch": 1.51, + "learning_rate": 3.743163084508127e-05, + "loss": 0.0, + "step": 6526 + }, + { + "epoch": 1.51, + "learning_rate": 3.742970495339343e-05, + "loss": 0.0002, + "step": 6527 + }, + { + "epoch": 1.51, + "learning_rate": 3.742777906170557e-05, + "loss": 0.0001, + "step": 6528 + }, + { + "epoch": 1.51, + "learning_rate": 3.742585317001772e-05, + "loss": 0.0001, + "step": 6529 + }, + { + "epoch": 1.51, + "learning_rate": 3.742392727832987e-05, + "loss": 0.0, + "step": 6530 + }, + { + "epoch": 1.51, + "learning_rate": 3.742200138664202e-05, + "loss": 0.0023, + "step": 6531 + }, + { + "epoch": 1.51, + "learning_rate": 3.7420075494954164e-05, + "loss": 0.0, + "step": 6532 + }, + { + "epoch": 1.51, + "learning_rate": 3.741814960326631e-05, + "loss": 0.0001, + "step": 6533 + }, + { + "epoch": 1.51, + "learning_rate": 3.7416223711578466e-05, + "loss": 0.0002, + "step": 6534 + }, + { + "epoch": 1.51, + "learning_rate": 3.741429781989061e-05, + "loss": 0.0001, + "step": 6535 + }, + { + "epoch": 1.51, + "learning_rate": 3.741237192820276e-05, + "loss": 0.0, + "step": 6536 + }, + { + "epoch": 1.51, + "learning_rate": 3.741044603651491e-05, + "loss": 0.0001, + "step": 6537 + }, + { + "epoch": 1.51, + "learning_rate": 3.7408520144827056e-05, + "loss": 0.0001, + "step": 6538 + }, + { + "epoch": 1.51, + "learning_rate": 3.7406594253139204e-05, + "loss": 0.0002, + "step": 6539 + }, + { + "epoch": 1.51, + "learning_rate": 3.740466836145136e-05, + "loss": 0.0, + "step": 6540 + }, + { + "epoch": 1.51, + "learning_rate": 3.7402742469763505e-05, + "loss": 0.0002, + "step": 6541 + }, + { + "epoch": 1.51, + "learning_rate": 3.7400816578075646e-05, + "loss": 0.0, + "step": 6542 + }, + { + "epoch": 1.51, + "learning_rate": 3.73988906863878e-05, + "loss": 0.0, + "step": 6543 + }, + { + "epoch": 1.51, + "learning_rate": 3.739696479469995e-05, + "loss": 0.0001, + "step": 6544 + }, + { + "epoch": 1.51, + "learning_rate": 3.7395038903012095e-05, + "loss": 0.0, + "step": 6545 + }, + { + "epoch": 1.51, + "learning_rate": 3.739311301132424e-05, + "loss": 0.0, + "step": 6546 + }, + { + "epoch": 1.51, + "learning_rate": 3.73911871196364e-05, + "loss": 0.0, + "step": 6547 + }, + { + "epoch": 1.51, + "learning_rate": 3.738926122794854e-05, + "loss": 0.0009, + "step": 6548 + }, + { + "epoch": 1.51, + "learning_rate": 3.738733533626069e-05, + "loss": 0.0001, + "step": 6549 + }, + { + "epoch": 1.51, + "learning_rate": 3.738540944457284e-05, + "loss": 0.1803, + "step": 6550 + }, + { + "epoch": 1.51, + "learning_rate": 3.738348355288499e-05, + "loss": 0.0, + "step": 6551 + }, + { + "epoch": 1.51, + "learning_rate": 3.7381557661197134e-05, + "loss": 0.0001, + "step": 6552 + }, + { + "epoch": 1.51, + "learning_rate": 3.737963176950928e-05, + "loss": 0.0, + "step": 6553 + }, + { + "epoch": 1.51, + "learning_rate": 3.7377705877821436e-05, + "loss": 0.0, + "step": 6554 + }, + { + "epoch": 1.51, + "learning_rate": 3.737577998613358e-05, + "loss": 0.0008, + "step": 6555 + }, + { + "epoch": 1.52, + "learning_rate": 3.737385409444573e-05, + "loss": 0.0001, + "step": 6556 + }, + { + "epoch": 1.52, + "learning_rate": 3.737192820275788e-05, + "loss": 0.0, + "step": 6557 + }, + { + "epoch": 1.52, + "learning_rate": 3.7370002311070026e-05, + "loss": 0.0002, + "step": 6558 + }, + { + "epoch": 1.52, + "learning_rate": 3.7368076419382173e-05, + "loss": 0.0, + "step": 6559 + }, + { + "epoch": 1.52, + "learning_rate": 3.736615052769433e-05, + "loss": 0.0006, + "step": 6560 + }, + { + "epoch": 1.52, + "learning_rate": 3.7364224636006475e-05, + "loss": 0.0, + "step": 6561 + }, + { + "epoch": 1.52, + "learning_rate": 3.7362298744318616e-05, + "loss": 0.0, + "step": 6562 + }, + { + "epoch": 1.52, + "learning_rate": 3.736037285263077e-05, + "loss": 0.0001, + "step": 6563 + }, + { + "epoch": 1.52, + "learning_rate": 3.735844696094292e-05, + "loss": 0.0061, + "step": 6564 + }, + { + "epoch": 1.52, + "learning_rate": 3.7356521069255065e-05, + "loss": 0.0105, + "step": 6565 + }, + { + "epoch": 1.52, + "learning_rate": 3.735459517756721e-05, + "loss": 0.0001, + "step": 6566 + }, + { + "epoch": 1.52, + "learning_rate": 3.735266928587937e-05, + "loss": 0.0, + "step": 6567 + }, + { + "epoch": 1.52, + "learning_rate": 3.7350743394191514e-05, + "loss": 0.0, + "step": 6568 + }, + { + "epoch": 1.52, + "learning_rate": 3.734881750250366e-05, + "loss": 0.0, + "step": 6569 + }, + { + "epoch": 1.52, + "learning_rate": 3.734689161081581e-05, + "loss": 0.0, + "step": 6570 + }, + { + "epoch": 1.52, + "learning_rate": 3.734496571912796e-05, + "loss": 0.0, + "step": 6571 + }, + { + "epoch": 1.52, + "learning_rate": 3.7343039827440104e-05, + "loss": 0.0001, + "step": 6572 + }, + { + "epoch": 1.52, + "learning_rate": 3.734111393575225e-05, + "loss": 0.0212, + "step": 6573 + }, + { + "epoch": 1.52, + "learning_rate": 3.7339188044064406e-05, + "loss": 0.0007, + "step": 6574 + }, + { + "epoch": 1.52, + "learning_rate": 3.7337262152376553e-05, + "loss": 0.0, + "step": 6575 + }, + { + "epoch": 1.52, + "learning_rate": 3.73353362606887e-05, + "loss": 0.0, + "step": 6576 + }, + { + "epoch": 1.52, + "learning_rate": 3.733341036900085e-05, + "loss": 0.0, + "step": 6577 + }, + { + "epoch": 1.52, + "learning_rate": 3.7331484477313e-05, + "loss": 0.0001, + "step": 6578 + }, + { + "epoch": 1.52, + "learning_rate": 3.732955858562514e-05, + "loss": 0.0001, + "step": 6579 + }, + { + "epoch": 1.52, + "learning_rate": 3.732763269393729e-05, + "loss": 0.0, + "step": 6580 + }, + { + "epoch": 1.52, + "learning_rate": 3.7325706802249445e-05, + "loss": 0.0, + "step": 6581 + }, + { + "epoch": 1.52, + "learning_rate": 3.732378091056159e-05, + "loss": 0.0001, + "step": 6582 + }, + { + "epoch": 1.52, + "learning_rate": 3.732185501887374e-05, + "loss": 0.0003, + "step": 6583 + }, + { + "epoch": 1.52, + "learning_rate": 3.731992912718589e-05, + "loss": 0.0, + "step": 6584 + }, + { + "epoch": 1.52, + "learning_rate": 3.731800323549804e-05, + "loss": 0.0, + "step": 6585 + }, + { + "epoch": 1.52, + "learning_rate": 3.731607734381018e-05, + "loss": 0.0, + "step": 6586 + }, + { + "epoch": 1.52, + "learning_rate": 3.731415145212234e-05, + "loss": 0.0001, + "step": 6587 + }, + { + "epoch": 1.52, + "learning_rate": 3.7312225560434484e-05, + "loss": 0.0005, + "step": 6588 + }, + { + "epoch": 1.52, + "learning_rate": 3.731029966874663e-05, + "loss": 0.0002, + "step": 6589 + }, + { + "epoch": 1.52, + "learning_rate": 3.730837377705878e-05, + "loss": 0.0, + "step": 6590 + }, + { + "epoch": 1.52, + "learning_rate": 3.730644788537093e-05, + "loss": 0.0001, + "step": 6591 + }, + { + "epoch": 1.52, + "learning_rate": 3.730452199368308e-05, + "loss": 0.0, + "step": 6592 + }, + { + "epoch": 1.52, + "learning_rate": 3.730259610199522e-05, + "loss": 0.0, + "step": 6593 + }, + { + "epoch": 1.52, + "learning_rate": 3.7300670210307376e-05, + "loss": 0.0, + "step": 6594 + }, + { + "epoch": 1.52, + "learning_rate": 3.729874431861952e-05, + "loss": 0.0, + "step": 6595 + }, + { + "epoch": 1.52, + "learning_rate": 3.729681842693167e-05, + "loss": 0.0003, + "step": 6596 + }, + { + "epoch": 1.52, + "learning_rate": 3.729489253524382e-05, + "loss": 0.0, + "step": 6597 + }, + { + "epoch": 1.52, + "learning_rate": 3.7292966643555966e-05, + "loss": 0.0001, + "step": 6598 + }, + { + "epoch": 1.53, + "learning_rate": 3.729104075186812e-05, + "loss": 0.0006, + "step": 6599 + }, + { + "epoch": 1.53, + "learning_rate": 3.728911486018026e-05, + "loss": 0.0005, + "step": 6600 + }, + { + "epoch": 1.53, + "learning_rate": 3.7287188968492415e-05, + "loss": 0.0023, + "step": 6601 + }, + { + "epoch": 1.53, + "learning_rate": 3.728526307680456e-05, + "loss": 0.0002, + "step": 6602 + }, + { + "epoch": 1.53, + "learning_rate": 3.728333718511671e-05, + "loss": 0.0002, + "step": 6603 + }, + { + "epoch": 1.53, + "learning_rate": 3.728141129342886e-05, + "loss": 0.0002, + "step": 6604 + }, + { + "epoch": 1.53, + "learning_rate": 3.727948540174101e-05, + "loss": 0.0, + "step": 6605 + }, + { + "epoch": 1.53, + "learning_rate": 3.727755951005316e-05, + "loss": 0.0003, + "step": 6606 + }, + { + "epoch": 1.53, + "learning_rate": 3.727563361836531e-05, + "loss": 0.0006, + "step": 6607 + }, + { + "epoch": 1.53, + "learning_rate": 3.7273707726677454e-05, + "loss": 0.0001, + "step": 6608 + }, + { + "epoch": 1.53, + "learning_rate": 3.72717818349896e-05, + "loss": 0.0, + "step": 6609 + }, + { + "epoch": 1.53, + "learning_rate": 3.726985594330175e-05, + "loss": 0.0, + "step": 6610 + }, + { + "epoch": 1.53, + "learning_rate": 3.7267930051613897e-05, + "loss": 0.0001, + "step": 6611 + }, + { + "epoch": 1.53, + "learning_rate": 3.726600415992605e-05, + "loss": 0.0006, + "step": 6612 + }, + { + "epoch": 1.53, + "learning_rate": 3.72640782682382e-05, + "loss": 0.0, + "step": 6613 + }, + { + "epoch": 1.53, + "learning_rate": 3.7262152376550346e-05, + "loss": 0.0005, + "step": 6614 + }, + { + "epoch": 1.53, + "learning_rate": 3.726022648486249e-05, + "loss": 0.0, + "step": 6615 + }, + { + "epoch": 1.53, + "learning_rate": 3.725830059317465e-05, + "loss": 0.0001, + "step": 6616 + }, + { + "epoch": 1.53, + "learning_rate": 3.725637470148679e-05, + "loss": 0.0006, + "step": 6617 + }, + { + "epoch": 1.53, + "learning_rate": 3.7254448809798936e-05, + "loss": 0.0, + "step": 6618 + }, + { + "epoch": 1.53, + "learning_rate": 3.725252291811109e-05, + "loss": 0.0, + "step": 6619 + }, + { + "epoch": 1.53, + "learning_rate": 3.725059702642324e-05, + "loss": 0.0, + "step": 6620 + }, + { + "epoch": 1.53, + "learning_rate": 3.7248671134735385e-05, + "loss": 0.0015, + "step": 6621 + }, + { + "epoch": 1.53, + "learning_rate": 3.724674524304753e-05, + "loss": 0.0, + "step": 6622 + }, + { + "epoch": 1.53, + "learning_rate": 3.7244819351359687e-05, + "loss": 0.0001, + "step": 6623 + }, + { + "epoch": 1.53, + "learning_rate": 3.724289345967183e-05, + "loss": 0.0001, + "step": 6624 + }, + { + "epoch": 1.53, + "learning_rate": 3.724096756798398e-05, + "loss": 0.0, + "step": 6625 + }, + { + "epoch": 1.53, + "learning_rate": 3.723904167629613e-05, + "loss": 0.0, + "step": 6626 + }, + { + "epoch": 1.53, + "learning_rate": 3.723711578460827e-05, + "loss": 0.0003, + "step": 6627 + }, + { + "epoch": 1.53, + "learning_rate": 3.7235189892920424e-05, + "loss": 0.0, + "step": 6628 + }, + { + "epoch": 1.53, + "learning_rate": 3.723326400123257e-05, + "loss": 0.0, + "step": 6629 + }, + { + "epoch": 1.53, + "learning_rate": 3.723133810954472e-05, + "loss": 0.0, + "step": 6630 + }, + { + "epoch": 1.53, + "learning_rate": 3.7229412217856866e-05, + "loss": 0.0, + "step": 6631 + }, + { + "epoch": 1.53, + "learning_rate": 3.722748632616902e-05, + "loss": 0.0, + "step": 6632 + }, + { + "epoch": 1.53, + "learning_rate": 3.722556043448117e-05, + "loss": 0.0008, + "step": 6633 + }, + { + "epoch": 1.53, + "learning_rate": 3.7223634542793316e-05, + "loss": 0.0002, + "step": 6634 + }, + { + "epoch": 1.53, + "learning_rate": 3.722170865110546e-05, + "loss": 0.0, + "step": 6635 + }, + { + "epoch": 1.53, + "learning_rate": 3.721978275941761e-05, + "loss": 0.0002, + "step": 6636 + }, + { + "epoch": 1.53, + "learning_rate": 3.721785686772976e-05, + "loss": 0.0, + "step": 6637 + }, + { + "epoch": 1.53, + "learning_rate": 3.7215930976041906e-05, + "loss": 0.0, + "step": 6638 + }, + { + "epoch": 1.53, + "learning_rate": 3.721400508435406e-05, + "loss": 0.0007, + "step": 6639 + }, + { + "epoch": 1.53, + "learning_rate": 3.721207919266621e-05, + "loss": 0.0, + "step": 6640 + }, + { + "epoch": 1.53, + "learning_rate": 3.7210153300978355e-05, + "loss": 0.0, + "step": 6641 + }, + { + "epoch": 1.54, + "learning_rate": 3.72082274092905e-05, + "loss": 0.0, + "step": 6642 + }, + { + "epoch": 1.54, + "learning_rate": 3.7206301517602656e-05, + "loss": 0.0001, + "step": 6643 + }, + { + "epoch": 1.54, + "learning_rate": 3.72043756259148e-05, + "loss": 0.0, + "step": 6644 + }, + { + "epoch": 1.54, + "learning_rate": 3.7202449734226945e-05, + "loss": 0.0001, + "step": 6645 + }, + { + "epoch": 1.54, + "learning_rate": 3.72005238425391e-05, + "loss": 0.0042, + "step": 6646 + }, + { + "epoch": 1.54, + "learning_rate": 3.7198597950851246e-05, + "loss": 0.0, + "step": 6647 + }, + { + "epoch": 1.54, + "learning_rate": 3.7196672059163394e-05, + "loss": 0.0001, + "step": 6648 + }, + { + "epoch": 1.54, + "learning_rate": 3.719474616747554e-05, + "loss": 0.0001, + "step": 6649 + }, + { + "epoch": 1.54, + "learning_rate": 3.7192820275787696e-05, + "loss": 0.0001, + "step": 6650 + }, + { + "epoch": 1.54, + "learning_rate": 3.7190894384099836e-05, + "loss": 0.0005, + "step": 6651 + }, + { + "epoch": 1.54, + "learning_rate": 3.718896849241199e-05, + "loss": 0.0, + "step": 6652 + }, + { + "epoch": 1.54, + "learning_rate": 3.718704260072414e-05, + "loss": 0.0, + "step": 6653 + }, + { + "epoch": 1.54, + "learning_rate": 3.7185116709036286e-05, + "loss": 0.0001, + "step": 6654 + }, + { + "epoch": 1.54, + "learning_rate": 3.718319081734843e-05, + "loss": 0.0, + "step": 6655 + }, + { + "epoch": 1.54, + "learning_rate": 3.718126492566058e-05, + "loss": 0.0008, + "step": 6656 + }, + { + "epoch": 1.54, + "learning_rate": 3.7179339033972735e-05, + "loss": 0.002, + "step": 6657 + }, + { + "epoch": 1.54, + "learning_rate": 3.7177413142284875e-05, + "loss": 0.0002, + "step": 6658 + }, + { + "epoch": 1.54, + "learning_rate": 3.717548725059703e-05, + "loss": 0.0, + "step": 6659 + }, + { + "epoch": 1.54, + "learning_rate": 3.717356135890918e-05, + "loss": 0.0, + "step": 6660 + }, + { + "epoch": 1.54, + "learning_rate": 3.7171635467221325e-05, + "loss": 0.0001, + "step": 6661 + }, + { + "epoch": 1.54, + "learning_rate": 3.716970957553347e-05, + "loss": 0.0, + "step": 6662 + }, + { + "epoch": 1.54, + "learning_rate": 3.7167783683845626e-05, + "loss": 0.0, + "step": 6663 + }, + { + "epoch": 1.54, + "learning_rate": 3.7165857792157774e-05, + "loss": 0.0, + "step": 6664 + }, + { + "epoch": 1.54, + "learning_rate": 3.7163931900469915e-05, + "loss": 0.0, + "step": 6665 + }, + { + "epoch": 1.54, + "learning_rate": 3.716200600878207e-05, + "loss": 0.0, + "step": 6666 + }, + { + "epoch": 1.54, + "learning_rate": 3.7160080117094216e-05, + "loss": 0.0014, + "step": 6667 + }, + { + "epoch": 1.54, + "learning_rate": 3.7158154225406364e-05, + "loss": 0.0017, + "step": 6668 + }, + { + "epoch": 1.54, + "learning_rate": 3.715622833371851e-05, + "loss": 0.0, + "step": 6669 + }, + { + "epoch": 1.54, + "learning_rate": 3.7154302442030665e-05, + "loss": 0.0007, + "step": 6670 + }, + { + "epoch": 1.54, + "learning_rate": 3.715237655034281e-05, + "loss": 0.0, + "step": 6671 + }, + { + "epoch": 1.54, + "learning_rate": 3.715045065865496e-05, + "loss": 0.0, + "step": 6672 + }, + { + "epoch": 1.54, + "learning_rate": 3.714852476696711e-05, + "loss": 0.0, + "step": 6673 + }, + { + "epoch": 1.54, + "learning_rate": 3.7146598875279255e-05, + "loss": 0.0, + "step": 6674 + }, + { + "epoch": 1.54, + "learning_rate": 3.71446729835914e-05, + "loss": 0.0002, + "step": 6675 + }, + { + "epoch": 1.54, + "learning_rate": 3.714274709190355e-05, + "loss": 0.0002, + "step": 6676 + }, + { + "epoch": 1.54, + "learning_rate": 3.7140821200215705e-05, + "loss": 0.0003, + "step": 6677 + }, + { + "epoch": 1.54, + "learning_rate": 3.713889530852785e-05, + "loss": 0.0003, + "step": 6678 + }, + { + "epoch": 1.54, + "learning_rate": 3.713696941684e-05, + "loss": 0.0001, + "step": 6679 + }, + { + "epoch": 1.54, + "learning_rate": 3.713504352515215e-05, + "loss": 0.0, + "step": 6680 + }, + { + "epoch": 1.54, + "learning_rate": 3.71331176334643e-05, + "loss": 0.0, + "step": 6681 + }, + { + "epoch": 1.54, + "learning_rate": 3.713119174177644e-05, + "loss": 0.0076, + "step": 6682 + }, + { + "epoch": 1.54, + "learning_rate": 3.712926585008859e-05, + "loss": 0.0, + "step": 6683 + }, + { + "epoch": 1.54, + "learning_rate": 3.7127339958400744e-05, + "loss": 0.0001, + "step": 6684 + }, + { + "epoch": 1.54, + "learning_rate": 3.712541406671289e-05, + "loss": 0.0, + "step": 6685 + }, + { + "epoch": 1.55, + "learning_rate": 3.712348817502504e-05, + "loss": 0.0, + "step": 6686 + }, + { + "epoch": 1.55, + "learning_rate": 3.7121562283337186e-05, + "loss": 0.0003, + "step": 6687 + }, + { + "epoch": 1.55, + "learning_rate": 3.711963639164934e-05, + "loss": 0.0018, + "step": 6688 + }, + { + "epoch": 1.55, + "learning_rate": 3.711771049996148e-05, + "loss": 0.0002, + "step": 6689 + }, + { + "epoch": 1.55, + "learning_rate": 3.7115784608273635e-05, + "loss": 0.001, + "step": 6690 + }, + { + "epoch": 1.55, + "learning_rate": 3.711385871658578e-05, + "loss": 0.0003, + "step": 6691 + }, + { + "epoch": 1.55, + "learning_rate": 3.711193282489793e-05, + "loss": 0.0, + "step": 6692 + }, + { + "epoch": 1.55, + "learning_rate": 3.711000693321008e-05, + "loss": 0.0002, + "step": 6693 + }, + { + "epoch": 1.55, + "learning_rate": 3.7108081041522225e-05, + "loss": 0.0, + "step": 6694 + }, + { + "epoch": 1.55, + "learning_rate": 3.710615514983438e-05, + "loss": 0.0, + "step": 6695 + }, + { + "epoch": 1.55, + "learning_rate": 3.710422925814652e-05, + "loss": 0.0, + "step": 6696 + }, + { + "epoch": 1.55, + "learning_rate": 3.7102303366458674e-05, + "loss": 0.0, + "step": 6697 + }, + { + "epoch": 1.55, + "learning_rate": 3.710037747477082e-05, + "loss": 0.0, + "step": 6698 + }, + { + "epoch": 1.55, + "learning_rate": 3.709845158308297e-05, + "loss": 0.0001, + "step": 6699 + }, + { + "epoch": 1.55, + "learning_rate": 3.709652569139512e-05, + "loss": 0.0002, + "step": 6700 + }, + { + "epoch": 1.55, + "learning_rate": 3.7094599799707264e-05, + "loss": 0.0, + "step": 6701 + }, + { + "epoch": 1.55, + "learning_rate": 3.709267390801941e-05, + "loss": 0.0001, + "step": 6702 + }, + { + "epoch": 1.55, + "learning_rate": 3.709074801633156e-05, + "loss": 0.0001, + "step": 6703 + }, + { + "epoch": 1.55, + "learning_rate": 3.7088822124643714e-05, + "loss": 0.0001, + "step": 6704 + }, + { + "epoch": 1.55, + "learning_rate": 3.708689623295586e-05, + "loss": 0.0003, + "step": 6705 + }, + { + "epoch": 1.55, + "learning_rate": 3.708497034126801e-05, + "loss": 0.0003, + "step": 6706 + }, + { + "epoch": 1.55, + "learning_rate": 3.7083044449580156e-05, + "loss": 0.0, + "step": 6707 + }, + { + "epoch": 1.55, + "learning_rate": 3.708111855789231e-05, + "loss": 0.0, + "step": 6708 + }, + { + "epoch": 1.55, + "learning_rate": 3.707919266620445e-05, + "loss": 0.0001, + "step": 6709 + }, + { + "epoch": 1.55, + "learning_rate": 3.70772667745166e-05, + "loss": 0.0007, + "step": 6710 + }, + { + "epoch": 1.55, + "learning_rate": 3.707534088282875e-05, + "loss": 0.0, + "step": 6711 + }, + { + "epoch": 1.55, + "learning_rate": 3.70734149911409e-05, + "loss": 0.0002, + "step": 6712 + }, + { + "epoch": 1.55, + "learning_rate": 3.707148909945305e-05, + "loss": 0.0, + "step": 6713 + }, + { + "epoch": 1.55, + "learning_rate": 3.7069563207765195e-05, + "loss": 0.0, + "step": 6714 + }, + { + "epoch": 1.55, + "learning_rate": 3.706763731607735e-05, + "loss": 0.0001, + "step": 6715 + }, + { + "epoch": 1.55, + "learning_rate": 3.706571142438949e-05, + "loss": 0.0001, + "step": 6716 + }, + { + "epoch": 1.55, + "learning_rate": 3.7063785532701644e-05, + "loss": 0.0, + "step": 6717 + }, + { + "epoch": 1.55, + "learning_rate": 3.706185964101379e-05, + "loss": 0.0, + "step": 6718 + }, + { + "epoch": 1.55, + "learning_rate": 3.705993374932594e-05, + "loss": 0.0004, + "step": 6719 + }, + { + "epoch": 1.55, + "learning_rate": 3.705800785763809e-05, + "loss": 0.0, + "step": 6720 + }, + { + "epoch": 1.55, + "learning_rate": 3.7056081965950234e-05, + "loss": 0.0, + "step": 6721 + }, + { + "epoch": 1.55, + "learning_rate": 3.705415607426239e-05, + "loss": 0.0, + "step": 6722 + }, + { + "epoch": 1.55, + "learning_rate": 3.705223018257453e-05, + "loss": 0.0, + "step": 6723 + }, + { + "epoch": 1.55, + "learning_rate": 3.7050304290886683e-05, + "loss": 0.0, + "step": 6724 + }, + { + "epoch": 1.55, + "learning_rate": 3.704837839919883e-05, + "loss": 0.0, + "step": 6725 + }, + { + "epoch": 1.55, + "learning_rate": 3.704645250751098e-05, + "loss": 0.0003, + "step": 6726 + }, + { + "epoch": 1.55, + "learning_rate": 3.7044526615823126e-05, + "loss": 0.0001, + "step": 6727 + }, + { + "epoch": 1.55, + "learning_rate": 3.704260072413528e-05, + "loss": 0.0001, + "step": 6728 + }, + { + "epoch": 1.56, + "learning_rate": 3.704067483244743e-05, + "loss": 0.0002, + "step": 6729 + }, + { + "epoch": 1.56, + "learning_rate": 3.703874894075957e-05, + "loss": 0.0003, + "step": 6730 + }, + { + "epoch": 1.56, + "learning_rate": 3.703682304907172e-05, + "loss": 0.0, + "step": 6731 + }, + { + "epoch": 1.56, + "learning_rate": 3.703489715738387e-05, + "loss": 0.0, + "step": 6732 + }, + { + "epoch": 1.56, + "learning_rate": 3.703297126569602e-05, + "loss": 0.0002, + "step": 6733 + }, + { + "epoch": 1.56, + "learning_rate": 3.7031045374008165e-05, + "loss": 0.0, + "step": 6734 + }, + { + "epoch": 1.56, + "learning_rate": 3.702911948232032e-05, + "loss": 0.0, + "step": 6735 + }, + { + "epoch": 1.56, + "learning_rate": 3.702719359063247e-05, + "loss": 0.0008, + "step": 6736 + }, + { + "epoch": 1.56, + "learning_rate": 3.7025267698944614e-05, + "loss": 0.0, + "step": 6737 + }, + { + "epoch": 1.56, + "learning_rate": 3.702334180725676e-05, + "loss": 0.0001, + "step": 6738 + }, + { + "epoch": 1.56, + "learning_rate": 3.702141591556891e-05, + "loss": 0.0, + "step": 6739 + }, + { + "epoch": 1.56, + "learning_rate": 3.701949002388106e-05, + "loss": 0.0008, + "step": 6740 + }, + { + "epoch": 1.56, + "learning_rate": 3.7017564132193204e-05, + "loss": 0.0051, + "step": 6741 + }, + { + "epoch": 1.56, + "learning_rate": 3.701563824050536e-05, + "loss": 0.0001, + "step": 6742 + }, + { + "epoch": 1.56, + "learning_rate": 3.7013712348817506e-05, + "loss": 0.0003, + "step": 6743 + }, + { + "epoch": 1.56, + "learning_rate": 3.701178645712965e-05, + "loss": 0.0, + "step": 6744 + }, + { + "epoch": 1.56, + "learning_rate": 3.70098605654418e-05, + "loss": 0.0001, + "step": 6745 + }, + { + "epoch": 1.56, + "learning_rate": 3.7007934673753955e-05, + "loss": 0.0006, + "step": 6746 + }, + { + "epoch": 1.56, + "learning_rate": 3.7006008782066096e-05, + "loss": 0.0, + "step": 6747 + }, + { + "epoch": 1.56, + "learning_rate": 3.700408289037824e-05, + "loss": 0.0, + "step": 6748 + }, + { + "epoch": 1.56, + "learning_rate": 3.70021569986904e-05, + "loss": 0.0003, + "step": 6749 + }, + { + "epoch": 1.56, + "learning_rate": 3.7000231107002545e-05, + "loss": 0.0, + "step": 6750 + }, + { + "epoch": 1.56, + "learning_rate": 3.699830521531469e-05, + "loss": 0.0002, + "step": 6751 + }, + { + "epoch": 1.56, + "learning_rate": 3.699637932362684e-05, + "loss": 0.0004, + "step": 6752 + }, + { + "epoch": 1.56, + "learning_rate": 3.6994453431938994e-05, + "loss": 0.0, + "step": 6753 + }, + { + "epoch": 1.56, + "learning_rate": 3.6992527540251135e-05, + "loss": 0.0002, + "step": 6754 + }, + { + "epoch": 1.56, + "learning_rate": 3.699060164856329e-05, + "loss": 0.0, + "step": 6755 + }, + { + "epoch": 1.56, + "learning_rate": 3.698867575687544e-05, + "loss": 0.0001, + "step": 6756 + }, + { + "epoch": 1.56, + "learning_rate": 3.6986749865187584e-05, + "loss": 0.0001, + "step": 6757 + }, + { + "epoch": 1.56, + "learning_rate": 3.698482397349973e-05, + "loss": 0.0, + "step": 6758 + }, + { + "epoch": 1.56, + "learning_rate": 3.698289808181188e-05, + "loss": 0.0022, + "step": 6759 + }, + { + "epoch": 1.56, + "learning_rate": 3.698097219012403e-05, + "loss": 0.0, + "step": 6760 + }, + { + "epoch": 1.56, + "learning_rate": 3.6979046298436174e-05, + "loss": 0.0, + "step": 6761 + }, + { + "epoch": 1.56, + "learning_rate": 3.697712040674833e-05, + "loss": 0.0005, + "step": 6762 + }, + { + "epoch": 1.56, + "learning_rate": 3.6975194515060476e-05, + "loss": 0.0001, + "step": 6763 + }, + { + "epoch": 1.56, + "learning_rate": 3.697326862337262e-05, + "loss": 0.0002, + "step": 6764 + }, + { + "epoch": 1.56, + "learning_rate": 3.697134273168477e-05, + "loss": 0.0, + "step": 6765 + }, + { + "epoch": 1.56, + "learning_rate": 3.696941683999692e-05, + "loss": 0.0, + "step": 6766 + }, + { + "epoch": 1.56, + "learning_rate": 3.696749094830907e-05, + "loss": 0.0006, + "step": 6767 + }, + { + "epoch": 1.56, + "learning_rate": 3.696556505662121e-05, + "loss": 0.0, + "step": 6768 + }, + { + "epoch": 1.56, + "learning_rate": 3.696363916493337e-05, + "loss": 0.0003, + "step": 6769 + }, + { + "epoch": 1.56, + "learning_rate": 3.6961713273245515e-05, + "loss": 0.0003, + "step": 6770 + }, + { + "epoch": 1.56, + "learning_rate": 3.695978738155766e-05, + "loss": 0.0001, + "step": 6771 + }, + { + "epoch": 1.57, + "learning_rate": 3.695786148986981e-05, + "loss": 0.0004, + "step": 6772 + }, + { + "epoch": 1.57, + "learning_rate": 3.6955935598181964e-05, + "loss": 0.0001, + "step": 6773 + }, + { + "epoch": 1.57, + "learning_rate": 3.6954009706494105e-05, + "loss": 0.0019, + "step": 6774 + }, + { + "epoch": 1.57, + "learning_rate": 3.695208381480626e-05, + "loss": 0.0, + "step": 6775 + }, + { + "epoch": 1.57, + "learning_rate": 3.6950157923118407e-05, + "loss": 0.0, + "step": 6776 + }, + { + "epoch": 1.57, + "learning_rate": 3.6948232031430554e-05, + "loss": 0.0, + "step": 6777 + }, + { + "epoch": 1.57, + "learning_rate": 3.69463061397427e-05, + "loss": 0.0002, + "step": 6778 + }, + { + "epoch": 1.57, + "learning_rate": 3.694438024805485e-05, + "loss": 0.0, + "step": 6779 + }, + { + "epoch": 1.57, + "learning_rate": 3.6942454356367e-05, + "loss": 0.0006, + "step": 6780 + }, + { + "epoch": 1.57, + "learning_rate": 3.6940528464679144e-05, + "loss": 0.0, + "step": 6781 + }, + { + "epoch": 1.57, + "learning_rate": 3.69386025729913e-05, + "loss": 0.0002, + "step": 6782 + }, + { + "epoch": 1.57, + "learning_rate": 3.6936676681303446e-05, + "loss": 0.0001, + "step": 6783 + }, + { + "epoch": 1.57, + "learning_rate": 3.693475078961559e-05, + "loss": 0.0001, + "step": 6784 + }, + { + "epoch": 1.57, + "learning_rate": 3.693282489792774e-05, + "loss": 0.0, + "step": 6785 + }, + { + "epoch": 1.57, + "learning_rate": 3.693089900623989e-05, + "loss": 0.0, + "step": 6786 + }, + { + "epoch": 1.57, + "learning_rate": 3.692897311455204e-05, + "loss": 0.0, + "step": 6787 + }, + { + "epoch": 1.57, + "learning_rate": 3.692704722286418e-05, + "loss": 0.001, + "step": 6788 + }, + { + "epoch": 1.57, + "learning_rate": 3.692512133117634e-05, + "loss": 0.0, + "step": 6789 + }, + { + "epoch": 1.57, + "learning_rate": 3.6923195439488485e-05, + "loss": 0.0, + "step": 6790 + }, + { + "epoch": 1.57, + "learning_rate": 3.692126954780063e-05, + "loss": 0.0001, + "step": 6791 + }, + { + "epoch": 1.57, + "learning_rate": 3.691934365611278e-05, + "loss": 0.0, + "step": 6792 + }, + { + "epoch": 1.57, + "learning_rate": 3.6917417764424934e-05, + "loss": 0.0, + "step": 6793 + }, + { + "epoch": 1.57, + "learning_rate": 3.691549187273708e-05, + "loss": 0.0001, + "step": 6794 + }, + { + "epoch": 1.57, + "learning_rate": 3.691356598104922e-05, + "loss": 0.0001, + "step": 6795 + }, + { + "epoch": 1.57, + "learning_rate": 3.6911640089361376e-05, + "loss": 0.0, + "step": 6796 + }, + { + "epoch": 1.57, + "learning_rate": 3.6909714197673524e-05, + "loss": 0.0007, + "step": 6797 + }, + { + "epoch": 1.57, + "learning_rate": 3.690778830598567e-05, + "loss": 0.0001, + "step": 6798 + }, + { + "epoch": 1.57, + "learning_rate": 3.690586241429782e-05, + "loss": 0.0001, + "step": 6799 + }, + { + "epoch": 1.57, + "learning_rate": 3.690393652260997e-05, + "loss": 0.0, + "step": 6800 + }, + { + "epoch": 1.57, + "learning_rate": 3.690201063092212e-05, + "loss": 0.0, + "step": 6801 + }, + { + "epoch": 1.57, + "learning_rate": 3.690008473923427e-05, + "loss": 0.0001, + "step": 6802 + }, + { + "epoch": 1.57, + "learning_rate": 3.6898158847546416e-05, + "loss": 0.0, + "step": 6803 + }, + { + "epoch": 1.57, + "learning_rate": 3.689623295585856e-05, + "loss": 0.0007, + "step": 6804 + }, + { + "epoch": 1.57, + "learning_rate": 3.689430706417071e-05, + "loss": 0.0001, + "step": 6805 + }, + { + "epoch": 1.57, + "learning_rate": 3.689238117248286e-05, + "loss": 0.0001, + "step": 6806 + }, + { + "epoch": 1.57, + "learning_rate": 3.689045528079501e-05, + "loss": 0.0, + "step": 6807 + }, + { + "epoch": 1.57, + "learning_rate": 3.688852938910716e-05, + "loss": 0.0, + "step": 6808 + }, + { + "epoch": 1.57, + "learning_rate": 3.688660349741931e-05, + "loss": 0.0001, + "step": 6809 + }, + { + "epoch": 1.57, + "learning_rate": 3.6884677605731455e-05, + "loss": 0.0, + "step": 6810 + }, + { + "epoch": 1.57, + "learning_rate": 3.688275171404361e-05, + "loss": 0.0008, + "step": 6811 + }, + { + "epoch": 1.57, + "learning_rate": 3.688082582235575e-05, + "loss": 0.0, + "step": 6812 + }, + { + "epoch": 1.57, + "learning_rate": 3.68788999306679e-05, + "loss": 0.0001, + "step": 6813 + }, + { + "epoch": 1.57, + "learning_rate": 3.687697403898005e-05, + "loss": 0.0, + "step": 6814 + }, + { + "epoch": 1.57, + "learning_rate": 3.68750481472922e-05, + "loss": 0.0, + "step": 6815 + }, + { + "epoch": 1.58, + "learning_rate": 3.6873122255604346e-05, + "loss": 0.0001, + "step": 6816 + }, + { + "epoch": 1.58, + "learning_rate": 3.6871196363916494e-05, + "loss": 0.0001, + "step": 6817 + }, + { + "epoch": 1.58, + "learning_rate": 3.686927047222865e-05, + "loss": 0.0, + "step": 6818 + }, + { + "epoch": 1.58, + "learning_rate": 3.686734458054079e-05, + "loss": 0.0, + "step": 6819 + }, + { + "epoch": 1.58, + "learning_rate": 3.686541868885294e-05, + "loss": 0.0001, + "step": 6820 + }, + { + "epoch": 1.58, + "learning_rate": 3.686349279716509e-05, + "loss": 0.0001, + "step": 6821 + }, + { + "epoch": 1.58, + "learning_rate": 3.686156690547724e-05, + "loss": 0.0, + "step": 6822 + }, + { + "epoch": 1.58, + "learning_rate": 3.6859641013789385e-05, + "loss": 0.0, + "step": 6823 + }, + { + "epoch": 1.58, + "learning_rate": 3.685771512210153e-05, + "loss": 0.0007, + "step": 6824 + }, + { + "epoch": 1.58, + "learning_rate": 3.685578923041369e-05, + "loss": 0.0001, + "step": 6825 + }, + { + "epoch": 1.58, + "learning_rate": 3.685386333872583e-05, + "loss": 0.0001, + "step": 6826 + }, + { + "epoch": 1.58, + "learning_rate": 3.685193744703798e-05, + "loss": 0.0, + "step": 6827 + }, + { + "epoch": 1.58, + "learning_rate": 3.685001155535013e-05, + "loss": 0.0001, + "step": 6828 + }, + { + "epoch": 1.58, + "learning_rate": 3.684808566366228e-05, + "loss": 0.0003, + "step": 6829 + }, + { + "epoch": 1.58, + "learning_rate": 3.6846159771974425e-05, + "loss": 0.0001, + "step": 6830 + }, + { + "epoch": 1.58, + "learning_rate": 3.684423388028658e-05, + "loss": 0.0006, + "step": 6831 + }, + { + "epoch": 1.58, + "learning_rate": 3.6842307988598726e-05, + "loss": 0.0, + "step": 6832 + }, + { + "epoch": 1.58, + "learning_rate": 3.684038209691087e-05, + "loss": 0.0001, + "step": 6833 + }, + { + "epoch": 1.58, + "learning_rate": 3.683845620522302e-05, + "loss": 0.0001, + "step": 6834 + }, + { + "epoch": 1.58, + "learning_rate": 3.683653031353517e-05, + "loss": 0.0, + "step": 6835 + }, + { + "epoch": 1.58, + "learning_rate": 3.6834604421847316e-05, + "loss": 0.0, + "step": 6836 + }, + { + "epoch": 1.58, + "learning_rate": 3.6832678530159464e-05, + "loss": 0.0001, + "step": 6837 + }, + { + "epoch": 1.58, + "learning_rate": 3.683075263847162e-05, + "loss": 0.0, + "step": 6838 + }, + { + "epoch": 1.58, + "learning_rate": 3.6828826746783765e-05, + "loss": 0.0, + "step": 6839 + }, + { + "epoch": 1.58, + "learning_rate": 3.682690085509591e-05, + "loss": 0.0, + "step": 6840 + }, + { + "epoch": 1.58, + "learning_rate": 3.682497496340806e-05, + "loss": 0.0, + "step": 6841 + }, + { + "epoch": 1.58, + "learning_rate": 3.682304907172021e-05, + "loss": 0.0, + "step": 6842 + }, + { + "epoch": 1.58, + "learning_rate": 3.6821123180032355e-05, + "loss": 0.0, + "step": 6843 + }, + { + "epoch": 1.58, + "learning_rate": 3.68191972883445e-05, + "loss": 0.0, + "step": 6844 + }, + { + "epoch": 1.58, + "learning_rate": 3.681727139665666e-05, + "loss": 0.0002, + "step": 6845 + }, + { + "epoch": 1.58, + "learning_rate": 3.6815345504968805e-05, + "loss": 0.0002, + "step": 6846 + }, + { + "epoch": 1.58, + "learning_rate": 3.681341961328095e-05, + "loss": 0.0, + "step": 6847 + }, + { + "epoch": 1.58, + "learning_rate": 3.68114937215931e-05, + "loss": 0.0, + "step": 6848 + }, + { + "epoch": 1.58, + "learning_rate": 3.6809567829905254e-05, + "loss": 0.0001, + "step": 6849 + }, + { + "epoch": 1.58, + "learning_rate": 3.6807641938217394e-05, + "loss": 0.0006, + "step": 6850 + }, + { + "epoch": 1.58, + "learning_rate": 3.680571604652954e-05, + "loss": 0.0001, + "step": 6851 + }, + { + "epoch": 1.58, + "learning_rate": 3.6803790154841696e-05, + "loss": 0.0, + "step": 6852 + }, + { + "epoch": 1.58, + "learning_rate": 3.680186426315384e-05, + "loss": 0.0004, + "step": 6853 + }, + { + "epoch": 1.58, + "learning_rate": 3.679993837146599e-05, + "loss": 0.0, + "step": 6854 + }, + { + "epoch": 1.58, + "learning_rate": 3.679801247977814e-05, + "loss": 0.0042, + "step": 6855 + }, + { + "epoch": 1.58, + "learning_rate": 3.6796086588090286e-05, + "loss": 0.0, + "step": 6856 + }, + { + "epoch": 1.58, + "learning_rate": 3.6794160696402434e-05, + "loss": 0.0005, + "step": 6857 + }, + { + "epoch": 1.58, + "learning_rate": 3.679223480471459e-05, + "loss": 0.0001, + "step": 6858 + }, + { + "epoch": 1.59, + "learning_rate": 3.6790308913026735e-05, + "loss": 0.0002, + "step": 6859 + }, + { + "epoch": 1.59, + "learning_rate": 3.6788383021338876e-05, + "loss": 0.0001, + "step": 6860 + }, + { + "epoch": 1.59, + "learning_rate": 3.678645712965103e-05, + "loss": 0.0, + "step": 6861 + }, + { + "epoch": 1.59, + "learning_rate": 3.678453123796318e-05, + "loss": 0.0002, + "step": 6862 + }, + { + "epoch": 1.59, + "learning_rate": 3.6782605346275325e-05, + "loss": 0.0004, + "step": 6863 + }, + { + "epoch": 1.59, + "learning_rate": 3.678067945458747e-05, + "loss": 0.0001, + "step": 6864 + }, + { + "epoch": 1.59, + "learning_rate": 3.677875356289963e-05, + "loss": 0.006, + "step": 6865 + }, + { + "epoch": 1.59, + "learning_rate": 3.6776827671211774e-05, + "loss": 0.0, + "step": 6866 + }, + { + "epoch": 1.59, + "learning_rate": 3.677490177952392e-05, + "loss": 0.0, + "step": 6867 + }, + { + "epoch": 1.59, + "learning_rate": 3.677297588783607e-05, + "loss": 0.0, + "step": 6868 + }, + { + "epoch": 1.59, + "learning_rate": 3.677104999614822e-05, + "loss": 0.0001, + "step": 6869 + }, + { + "epoch": 1.59, + "learning_rate": 3.6769124104460364e-05, + "loss": 0.0, + "step": 6870 + }, + { + "epoch": 1.59, + "learning_rate": 3.676719821277251e-05, + "loss": 0.0, + "step": 6871 + }, + { + "epoch": 1.59, + "learning_rate": 3.6765272321084666e-05, + "loss": 0.0001, + "step": 6872 + }, + { + "epoch": 1.59, + "learning_rate": 3.6763346429396814e-05, + "loss": 0.0, + "step": 6873 + }, + { + "epoch": 1.59, + "learning_rate": 3.676142053770896e-05, + "loss": 0.0, + "step": 6874 + }, + { + "epoch": 1.59, + "learning_rate": 3.675949464602111e-05, + "loss": 0.0, + "step": 6875 + }, + { + "epoch": 1.59, + "learning_rate": 3.675756875433326e-05, + "loss": 0.0003, + "step": 6876 + }, + { + "epoch": 1.59, + "learning_rate": 3.6755642862645403e-05, + "loss": 0.0003, + "step": 6877 + }, + { + "epoch": 1.59, + "learning_rate": 3.675371697095755e-05, + "loss": 0.0109, + "step": 6878 + }, + { + "epoch": 1.59, + "learning_rate": 3.6751791079269705e-05, + "loss": 0.0001, + "step": 6879 + }, + { + "epoch": 1.59, + "learning_rate": 3.674986518758185e-05, + "loss": 0.0, + "step": 6880 + }, + { + "epoch": 1.59, + "learning_rate": 3.6747939295894e-05, + "loss": 0.0006, + "step": 6881 + }, + { + "epoch": 1.59, + "learning_rate": 3.674601340420615e-05, + "loss": 0.0003, + "step": 6882 + }, + { + "epoch": 1.59, + "learning_rate": 3.67440875125183e-05, + "loss": 0.0053, + "step": 6883 + }, + { + "epoch": 1.59, + "learning_rate": 3.674216162083044e-05, + "loss": 0.0, + "step": 6884 + }, + { + "epoch": 1.59, + "learning_rate": 3.67402357291426e-05, + "loss": 0.0001, + "step": 6885 + }, + { + "epoch": 1.59, + "learning_rate": 3.6738309837454744e-05, + "loss": 0.0, + "step": 6886 + }, + { + "epoch": 1.59, + "learning_rate": 3.673638394576689e-05, + "loss": 0.0016, + "step": 6887 + }, + { + "epoch": 1.59, + "learning_rate": 3.673445805407904e-05, + "loss": 0.0002, + "step": 6888 + }, + { + "epoch": 1.59, + "learning_rate": 3.673253216239119e-05, + "loss": 0.0, + "step": 6889 + }, + { + "epoch": 1.59, + "learning_rate": 3.673060627070334e-05, + "loss": 0.0002, + "step": 6890 + }, + { + "epoch": 1.59, + "learning_rate": 3.672868037901548e-05, + "loss": 0.0001, + "step": 6891 + }, + { + "epoch": 1.59, + "learning_rate": 3.6726754487327636e-05, + "loss": 0.0001, + "step": 6892 + }, + { + "epoch": 1.59, + "learning_rate": 3.6724828595639783e-05, + "loss": 0.0, + "step": 6893 + }, + { + "epoch": 1.59, + "learning_rate": 3.672290270395193e-05, + "loss": 0.0, + "step": 6894 + }, + { + "epoch": 1.59, + "learning_rate": 3.672097681226408e-05, + "loss": 0.0001, + "step": 6895 + }, + { + "epoch": 1.59, + "learning_rate": 3.671905092057623e-05, + "loss": 0.0007, + "step": 6896 + }, + { + "epoch": 1.59, + "learning_rate": 3.671712502888838e-05, + "loss": 0.0, + "step": 6897 + }, + { + "epoch": 1.59, + "learning_rate": 3.671519913720052e-05, + "loss": 0.0, + "step": 6898 + }, + { + "epoch": 1.59, + "learning_rate": 3.6713273245512675e-05, + "loss": 0.0, + "step": 6899 + }, + { + "epoch": 1.59, + "learning_rate": 3.671134735382482e-05, + "loss": 0.0, + "step": 6900 + }, + { + "epoch": 1.59, + "learning_rate": 3.670942146213697e-05, + "loss": 0.0001, + "step": 6901 + }, + { + "epoch": 1.6, + "learning_rate": 3.670749557044912e-05, + "loss": 0.0, + "step": 6902 + }, + { + "epoch": 1.6, + "learning_rate": 3.670556967876127e-05, + "loss": 0.0, + "step": 6903 + }, + { + "epoch": 1.6, + "learning_rate": 3.670364378707342e-05, + "loss": 0.0029, + "step": 6904 + }, + { + "epoch": 1.6, + "learning_rate": 3.670171789538557e-05, + "loss": 0.0, + "step": 6905 + }, + { + "epoch": 1.6, + "learning_rate": 3.6699792003697714e-05, + "loss": 0.0, + "step": 6906 + }, + { + "epoch": 1.6, + "learning_rate": 3.669786611200986e-05, + "loss": 0.0003, + "step": 6907 + }, + { + "epoch": 1.6, + "learning_rate": 3.669594022032201e-05, + "loss": 0.0, + "step": 6908 + }, + { + "epoch": 1.6, + "learning_rate": 3.669401432863416e-05, + "loss": 0.0, + "step": 6909 + }, + { + "epoch": 1.6, + "learning_rate": 3.669208843694631e-05, + "loss": 0.0, + "step": 6910 + }, + { + "epoch": 1.6, + "learning_rate": 3.669016254525846e-05, + "loss": 0.0003, + "step": 6911 + }, + { + "epoch": 1.6, + "learning_rate": 3.6688236653570606e-05, + "loss": 0.0, + "step": 6912 + }, + { + "epoch": 1.6, + "learning_rate": 3.668631076188275e-05, + "loss": 0.0001, + "step": 6913 + }, + { + "epoch": 1.6, + "learning_rate": 3.668438487019491e-05, + "loss": 0.0, + "step": 6914 + }, + { + "epoch": 1.6, + "learning_rate": 3.668245897850705e-05, + "loss": 0.0, + "step": 6915 + }, + { + "epoch": 1.6, + "learning_rate": 3.6680533086819196e-05, + "loss": 0.0001, + "step": 6916 + }, + { + "epoch": 1.6, + "learning_rate": 3.667860719513135e-05, + "loss": 0.0001, + "step": 6917 + }, + { + "epoch": 1.6, + "learning_rate": 3.66766813034435e-05, + "loss": 0.0, + "step": 6918 + }, + { + "epoch": 1.6, + "learning_rate": 3.6674755411755645e-05, + "loss": 0.0003, + "step": 6919 + }, + { + "epoch": 1.6, + "learning_rate": 3.667282952006779e-05, + "loss": 0.0002, + "step": 6920 + }, + { + "epoch": 1.6, + "learning_rate": 3.667090362837995e-05, + "loss": 0.0001, + "step": 6921 + }, + { + "epoch": 1.6, + "learning_rate": 3.666897773669209e-05, + "loss": 0.0001, + "step": 6922 + }, + { + "epoch": 1.6, + "learning_rate": 3.666705184500424e-05, + "loss": 0.0, + "step": 6923 + }, + { + "epoch": 1.6, + "learning_rate": 3.666512595331639e-05, + "loss": 0.0, + "step": 6924 + }, + { + "epoch": 1.6, + "learning_rate": 3.666320006162853e-05, + "loss": 0.0, + "step": 6925 + }, + { + "epoch": 1.6, + "learning_rate": 3.6661274169940684e-05, + "loss": 0.0001, + "step": 6926 + }, + { + "epoch": 1.6, + "learning_rate": 3.665934827825283e-05, + "loss": 0.0003, + "step": 6927 + }, + { + "epoch": 1.6, + "learning_rate": 3.665742238656498e-05, + "loss": 0.0, + "step": 6928 + }, + { + "epoch": 1.6, + "learning_rate": 3.6655496494877127e-05, + "loss": 0.0, + "step": 6929 + }, + { + "epoch": 1.6, + "learning_rate": 3.665357060318928e-05, + "loss": 0.0001, + "step": 6930 + }, + { + "epoch": 1.6, + "learning_rate": 3.665164471150143e-05, + "loss": 0.0, + "step": 6931 + }, + { + "epoch": 1.6, + "learning_rate": 3.6649718819813576e-05, + "loss": 0.0001, + "step": 6932 + }, + { + "epoch": 1.6, + "learning_rate": 3.664779292812572e-05, + "loss": 0.0, + "step": 6933 + }, + { + "epoch": 1.6, + "learning_rate": 3.664586703643787e-05, + "loss": 0.0003, + "step": 6934 + }, + { + "epoch": 1.6, + "learning_rate": 3.664394114475002e-05, + "loss": 0.0, + "step": 6935 + }, + { + "epoch": 1.6, + "learning_rate": 3.6642015253062166e-05, + "loss": 0.0009, + "step": 6936 + }, + { + "epoch": 1.6, + "learning_rate": 3.664008936137432e-05, + "loss": 0.0, + "step": 6937 + }, + { + "epoch": 1.6, + "learning_rate": 3.663816346968647e-05, + "loss": 0.0001, + "step": 6938 + }, + { + "epoch": 1.6, + "learning_rate": 3.6636237577998615e-05, + "loss": 0.0, + "step": 6939 + }, + { + "epoch": 1.6, + "learning_rate": 3.663431168631076e-05, + "loss": 0.0001, + "step": 6940 + }, + { + "epoch": 1.6, + "learning_rate": 3.6632385794622917e-05, + "loss": 0.0, + "step": 6941 + }, + { + "epoch": 1.6, + "learning_rate": 3.663045990293506e-05, + "loss": 0.0001, + "step": 6942 + }, + { + "epoch": 1.6, + "learning_rate": 3.662853401124721e-05, + "loss": 0.0024, + "step": 6943 + }, + { + "epoch": 1.6, + "learning_rate": 3.662660811955936e-05, + "loss": 0.0001, + "step": 6944 + }, + { + "epoch": 1.61, + "learning_rate": 3.6624682227871506e-05, + "loss": 0.0001, + "step": 6945 + }, + { + "epoch": 1.61, + "learning_rate": 3.6622756336183654e-05, + "loss": 0.0, + "step": 6946 + }, + { + "epoch": 1.61, + "learning_rate": 3.66208304444958e-05, + "loss": 0.0, + "step": 6947 + }, + { + "epoch": 1.61, + "learning_rate": 3.6618904552807956e-05, + "loss": 0.0, + "step": 6948 + }, + { + "epoch": 1.61, + "learning_rate": 3.6616978661120096e-05, + "loss": 0.0002, + "step": 6949 + }, + { + "epoch": 1.61, + "learning_rate": 3.661505276943225e-05, + "loss": 0.0, + "step": 6950 + }, + { + "epoch": 1.61, + "learning_rate": 3.66131268777444e-05, + "loss": 0.0001, + "step": 6951 + }, + { + "epoch": 1.61, + "learning_rate": 3.6611200986056546e-05, + "loss": 0.0004, + "step": 6952 + }, + { + "epoch": 1.61, + "learning_rate": 3.660927509436869e-05, + "loss": 0.0002, + "step": 6953 + }, + { + "epoch": 1.61, + "learning_rate": 3.660734920268084e-05, + "loss": 0.0, + "step": 6954 + }, + { + "epoch": 1.61, + "learning_rate": 3.6605423310992995e-05, + "loss": 0.0, + "step": 6955 + }, + { + "epoch": 1.61, + "learning_rate": 3.6603497419305136e-05, + "loss": 0.0001, + "step": 6956 + }, + { + "epoch": 1.61, + "learning_rate": 3.660157152761729e-05, + "loss": 0.0, + "step": 6957 + }, + { + "epoch": 1.61, + "learning_rate": 3.659964563592944e-05, + "loss": 0.0131, + "step": 6958 + }, + { + "epoch": 1.61, + "learning_rate": 3.6597719744241585e-05, + "loss": 0.0001, + "step": 6959 + }, + { + "epoch": 1.61, + "learning_rate": 3.659579385255373e-05, + "loss": 0.0, + "step": 6960 + }, + { + "epoch": 1.61, + "learning_rate": 3.6593867960865886e-05, + "loss": 0.0002, + "step": 6961 + }, + { + "epoch": 1.61, + "learning_rate": 3.6591942069178034e-05, + "loss": 0.0003, + "step": 6962 + }, + { + "epoch": 1.61, + "learning_rate": 3.6590016177490175e-05, + "loss": 0.0, + "step": 6963 + }, + { + "epoch": 1.61, + "learning_rate": 3.658809028580233e-05, + "loss": 0.002, + "step": 6964 + }, + { + "epoch": 1.61, + "learning_rate": 3.6586164394114476e-05, + "loss": 0.0001, + "step": 6965 + }, + { + "epoch": 1.61, + "learning_rate": 3.6584238502426624e-05, + "loss": 0.0, + "step": 6966 + }, + { + "epoch": 1.61, + "learning_rate": 3.658231261073877e-05, + "loss": 0.0009, + "step": 6967 + }, + { + "epoch": 1.61, + "learning_rate": 3.6580386719050926e-05, + "loss": 0.0, + "step": 6968 + }, + { + "epoch": 1.61, + "learning_rate": 3.657846082736307e-05, + "loss": 0.0008, + "step": 6969 + }, + { + "epoch": 1.61, + "learning_rate": 3.657653493567522e-05, + "loss": 0.0, + "step": 6970 + }, + { + "epoch": 1.61, + "learning_rate": 3.657460904398737e-05, + "loss": 0.0, + "step": 6971 + }, + { + "epoch": 1.61, + "learning_rate": 3.6572683152299515e-05, + "loss": 0.0, + "step": 6972 + }, + { + "epoch": 1.61, + "learning_rate": 3.657075726061166e-05, + "loss": 0.0002, + "step": 6973 + }, + { + "epoch": 1.61, + "learning_rate": 3.656883136892381e-05, + "loss": 0.0, + "step": 6974 + }, + { + "epoch": 1.61, + "learning_rate": 3.6566905477235965e-05, + "loss": 0.0002, + "step": 6975 + }, + { + "epoch": 1.61, + "learning_rate": 3.656497958554811e-05, + "loss": 0.0001, + "step": 6976 + }, + { + "epoch": 1.61, + "learning_rate": 3.656305369386026e-05, + "loss": 0.0001, + "step": 6977 + }, + { + "epoch": 1.61, + "learning_rate": 3.656112780217241e-05, + "loss": 0.0, + "step": 6978 + }, + { + "epoch": 1.61, + "learning_rate": 3.655920191048456e-05, + "loss": 0.0002, + "step": 6979 + }, + { + "epoch": 1.61, + "learning_rate": 3.65572760187967e-05, + "loss": 0.0, + "step": 6980 + }, + { + "epoch": 1.61, + "learning_rate": 3.655535012710885e-05, + "loss": 0.0, + "step": 6981 + }, + { + "epoch": 1.61, + "learning_rate": 3.6553424235421004e-05, + "loss": 0.0005, + "step": 6982 + }, + { + "epoch": 1.61, + "learning_rate": 3.655149834373315e-05, + "loss": 0.0001, + "step": 6983 + }, + { + "epoch": 1.61, + "learning_rate": 3.65495724520453e-05, + "loss": 0.0, + "step": 6984 + }, + { + "epoch": 1.61, + "learning_rate": 3.6547646560357446e-05, + "loss": 0.0, + "step": 6985 + }, + { + "epoch": 1.61, + "learning_rate": 3.65457206686696e-05, + "loss": 0.0, + "step": 6986 + }, + { + "epoch": 1.61, + "learning_rate": 3.654379477698174e-05, + "loss": 0.0, + "step": 6987 + }, + { + "epoch": 1.61, + "learning_rate": 3.6541868885293895e-05, + "loss": 0.0, + "step": 6988 + }, + { + "epoch": 1.62, + "learning_rate": 3.653994299360604e-05, + "loss": 0.0001, + "step": 6989 + }, + { + "epoch": 1.62, + "learning_rate": 3.653801710191819e-05, + "loss": 0.0, + "step": 6990 + }, + { + "epoch": 1.62, + "learning_rate": 3.653609121023034e-05, + "loss": 0.0, + "step": 6991 + }, + { + "epoch": 1.62, + "learning_rate": 3.6534165318542485e-05, + "loss": 0.0001, + "step": 6992 + }, + { + "epoch": 1.62, + "learning_rate": 3.653223942685464e-05, + "loss": 0.0, + "step": 6993 + }, + { + "epoch": 1.62, + "learning_rate": 3.653031353516678e-05, + "loss": 0.0, + "step": 6994 + }, + { + "epoch": 1.62, + "learning_rate": 3.6528387643478935e-05, + "loss": 0.0, + "step": 6995 + }, + { + "epoch": 1.62, + "learning_rate": 3.652646175179108e-05, + "loss": 0.0004, + "step": 6996 + }, + { + "epoch": 1.62, + "learning_rate": 3.652453586010323e-05, + "loss": 0.0001, + "step": 6997 + }, + { + "epoch": 1.62, + "learning_rate": 3.652260996841538e-05, + "loss": 0.0084, + "step": 6998 + }, + { + "epoch": 1.62, + "learning_rate": 3.652068407672753e-05, + "loss": 0.0, + "step": 6999 + }, + { + "epoch": 1.62, + "learning_rate": 3.651875818503968e-05, + "loss": 0.0003, + "step": 7000 + }, + { + "epoch": 1.62, + "learning_rate": 3.651683229335182e-05, + "loss": 0.0, + "step": 7001 + }, + { + "epoch": 1.62, + "learning_rate": 3.6514906401663974e-05, + "loss": 0.0, + "step": 7002 + }, + { + "epoch": 1.62, + "learning_rate": 3.651298050997612e-05, + "loss": 0.0, + "step": 7003 + }, + { + "epoch": 1.62, + "learning_rate": 3.651105461828827e-05, + "loss": 0.0, + "step": 7004 + }, + { + "epoch": 1.62, + "learning_rate": 3.6509128726600416e-05, + "loss": 0.0001, + "step": 7005 + }, + { + "epoch": 1.62, + "learning_rate": 3.650720283491257e-05, + "loss": 0.0, + "step": 7006 + }, + { + "epoch": 1.62, + "learning_rate": 3.650527694322471e-05, + "loss": 0.0032, + "step": 7007 + }, + { + "epoch": 1.62, + "learning_rate": 3.6503351051536865e-05, + "loss": 0.0, + "step": 7008 + }, + { + "epoch": 1.62, + "learning_rate": 3.650142515984901e-05, + "loss": 0.0, + "step": 7009 + }, + { + "epoch": 1.62, + "learning_rate": 3.649949926816116e-05, + "loss": 0.0, + "step": 7010 + }, + { + "epoch": 1.62, + "learning_rate": 3.649757337647331e-05, + "loss": 0.0, + "step": 7011 + }, + { + "epoch": 1.62, + "learning_rate": 3.6495647484785455e-05, + "loss": 0.0031, + "step": 7012 + }, + { + "epoch": 1.62, + "learning_rate": 3.649372159309761e-05, + "loss": 0.0, + "step": 7013 + }, + { + "epoch": 1.62, + "learning_rate": 3.649179570140975e-05, + "loss": 0.0001, + "step": 7014 + }, + { + "epoch": 1.62, + "learning_rate": 3.6489869809721904e-05, + "loss": 0.0001, + "step": 7015 + }, + { + "epoch": 1.62, + "learning_rate": 3.648794391803405e-05, + "loss": 0.0, + "step": 7016 + }, + { + "epoch": 1.62, + "learning_rate": 3.64860180263462e-05, + "loss": 0.0009, + "step": 7017 + }, + { + "epoch": 1.62, + "learning_rate": 3.648409213465835e-05, + "loss": 0.0001, + "step": 7018 + }, + { + "epoch": 1.62, + "learning_rate": 3.6482166242970494e-05, + "loss": 0.0001, + "step": 7019 + }, + { + "epoch": 1.62, + "learning_rate": 3.648024035128265e-05, + "loss": 0.0004, + "step": 7020 + }, + { + "epoch": 1.62, + "learning_rate": 3.647831445959479e-05, + "loss": 0.0003, + "step": 7021 + }, + { + "epoch": 1.62, + "learning_rate": 3.6476388567906944e-05, + "loss": 0.0003, + "step": 7022 + }, + { + "epoch": 1.62, + "learning_rate": 3.647446267621909e-05, + "loss": 0.0, + "step": 7023 + }, + { + "epoch": 1.62, + "learning_rate": 3.647253678453124e-05, + "loss": 0.0076, + "step": 7024 + }, + { + "epoch": 1.62, + "learning_rate": 3.6470610892843386e-05, + "loss": 0.0, + "step": 7025 + }, + { + "epoch": 1.62, + "learning_rate": 3.646868500115554e-05, + "loss": 0.0, + "step": 7026 + }, + { + "epoch": 1.62, + "learning_rate": 3.646675910946769e-05, + "loss": 0.0, + "step": 7027 + }, + { + "epoch": 1.62, + "learning_rate": 3.646483321777983e-05, + "loss": 0.0, + "step": 7028 + }, + { + "epoch": 1.62, + "learning_rate": 3.646290732609198e-05, + "loss": 0.0, + "step": 7029 + }, + { + "epoch": 1.62, + "learning_rate": 3.646098143440413e-05, + "loss": 0.0006, + "step": 7030 + }, + { + "epoch": 1.62, + "learning_rate": 3.645905554271628e-05, + "loss": 0.0, + "step": 7031 + }, + { + "epoch": 1.63, + "learning_rate": 3.6457129651028425e-05, + "loss": 0.0002, + "step": 7032 + }, + { + "epoch": 1.63, + "learning_rate": 3.645520375934058e-05, + "loss": 0.0001, + "step": 7033 + }, + { + "epoch": 1.63, + "learning_rate": 3.645327786765273e-05, + "loss": 0.0, + "step": 7034 + }, + { + "epoch": 1.63, + "learning_rate": 3.6451351975964874e-05, + "loss": 0.0, + "step": 7035 + }, + { + "epoch": 1.63, + "learning_rate": 3.644942608427702e-05, + "loss": 0.0, + "step": 7036 + }, + { + "epoch": 1.63, + "learning_rate": 3.644750019258917e-05, + "loss": 0.0007, + "step": 7037 + }, + { + "epoch": 1.63, + "learning_rate": 3.644557430090132e-05, + "loss": 0.0, + "step": 7038 + }, + { + "epoch": 1.63, + "learning_rate": 3.6443648409213464e-05, + "loss": 0.0, + "step": 7039 + }, + { + "epoch": 1.63, + "learning_rate": 3.644172251752562e-05, + "loss": 0.0, + "step": 7040 + }, + { + "epoch": 1.63, + "learning_rate": 3.6439796625837766e-05, + "loss": 0.0, + "step": 7041 + }, + { + "epoch": 1.63, + "learning_rate": 3.6437870734149913e-05, + "loss": 0.0058, + "step": 7042 + }, + { + "epoch": 1.63, + "learning_rate": 3.643594484246206e-05, + "loss": 0.0, + "step": 7043 + }, + { + "epoch": 1.63, + "learning_rate": 3.6434018950774215e-05, + "loss": 0.0, + "step": 7044 + }, + { + "epoch": 1.63, + "learning_rate": 3.6432093059086356e-05, + "loss": 0.0, + "step": 7045 + }, + { + "epoch": 1.63, + "learning_rate": 3.64301671673985e-05, + "loss": 0.0002, + "step": 7046 + }, + { + "epoch": 1.63, + "learning_rate": 3.642824127571066e-05, + "loss": 0.0001, + "step": 7047 + }, + { + "epoch": 1.63, + "learning_rate": 3.6426315384022805e-05, + "loss": 0.0004, + "step": 7048 + }, + { + "epoch": 1.63, + "learning_rate": 3.642438949233495e-05, + "loss": 0.0149, + "step": 7049 + }, + { + "epoch": 1.63, + "learning_rate": 3.64224636006471e-05, + "loss": 0.0, + "step": 7050 + }, + { + "epoch": 1.63, + "learning_rate": 3.6420537708959254e-05, + "loss": 0.0001, + "step": 7051 + }, + { + "epoch": 1.63, + "learning_rate": 3.6418611817271395e-05, + "loss": 0.0002, + "step": 7052 + }, + { + "epoch": 1.63, + "learning_rate": 3.641668592558355e-05, + "loss": 0.0009, + "step": 7053 + }, + { + "epoch": 1.63, + "learning_rate": 3.64147600338957e-05, + "loss": 0.0003, + "step": 7054 + }, + { + "epoch": 1.63, + "learning_rate": 3.6412834142207844e-05, + "loss": 0.0, + "step": 7055 + }, + { + "epoch": 1.63, + "learning_rate": 3.641090825051999e-05, + "loss": 0.0, + "step": 7056 + }, + { + "epoch": 1.63, + "learning_rate": 3.640898235883214e-05, + "loss": 0.0002, + "step": 7057 + }, + { + "epoch": 1.63, + "learning_rate": 3.6407056467144293e-05, + "loss": 0.0001, + "step": 7058 + }, + { + "epoch": 1.63, + "learning_rate": 3.6405130575456434e-05, + "loss": 0.0, + "step": 7059 + }, + { + "epoch": 1.63, + "learning_rate": 3.640320468376859e-05, + "loss": 0.0001, + "step": 7060 + }, + { + "epoch": 1.63, + "learning_rate": 3.6401278792080736e-05, + "loss": 0.0, + "step": 7061 + }, + { + "epoch": 1.63, + "learning_rate": 3.639935290039288e-05, + "loss": 0.0, + "step": 7062 + }, + { + "epoch": 1.63, + "learning_rate": 3.639742700870503e-05, + "loss": 0.0012, + "step": 7063 + }, + { + "epoch": 1.63, + "learning_rate": 3.6395501117017185e-05, + "loss": 0.0, + "step": 7064 + }, + { + "epoch": 1.63, + "learning_rate": 3.639357522532933e-05, + "loss": 0.0, + "step": 7065 + }, + { + "epoch": 1.63, + "learning_rate": 3.639164933364147e-05, + "loss": 0.0, + "step": 7066 + }, + { + "epoch": 1.63, + "learning_rate": 3.638972344195363e-05, + "loss": 0.0, + "step": 7067 + }, + { + "epoch": 1.63, + "learning_rate": 3.6387797550265775e-05, + "loss": 0.0001, + "step": 7068 + }, + { + "epoch": 1.63, + "learning_rate": 3.638587165857792e-05, + "loss": 0.001, + "step": 7069 + }, + { + "epoch": 1.63, + "learning_rate": 3.638394576689007e-05, + "loss": 0.0001, + "step": 7070 + }, + { + "epoch": 1.63, + "learning_rate": 3.6382019875202224e-05, + "loss": 0.0019, + "step": 7071 + }, + { + "epoch": 1.63, + "learning_rate": 3.638009398351437e-05, + "loss": 0.0001, + "step": 7072 + }, + { + "epoch": 1.63, + "learning_rate": 3.637816809182652e-05, + "loss": 0.0, + "step": 7073 + }, + { + "epoch": 1.63, + "learning_rate": 3.637624220013867e-05, + "loss": 0.0074, + "step": 7074 + }, + { + "epoch": 1.64, + "learning_rate": 3.6374316308450814e-05, + "loss": 0.0005, + "step": 7075 + }, + { + "epoch": 1.64, + "learning_rate": 3.637239041676296e-05, + "loss": 0.0, + "step": 7076 + }, + { + "epoch": 1.64, + "learning_rate": 3.637046452507511e-05, + "loss": 0.0003, + "step": 7077 + }, + { + "epoch": 1.64, + "learning_rate": 3.636853863338726e-05, + "loss": 0.0002, + "step": 7078 + }, + { + "epoch": 1.64, + "learning_rate": 3.6366612741699404e-05, + "loss": 0.0, + "step": 7079 + }, + { + "epoch": 1.64, + "learning_rate": 3.636468685001156e-05, + "loss": 0.0, + "step": 7080 + }, + { + "epoch": 1.64, + "learning_rate": 3.6362760958323706e-05, + "loss": 0.0002, + "step": 7081 + }, + { + "epoch": 1.64, + "learning_rate": 3.636083506663585e-05, + "loss": 0.0, + "step": 7082 + }, + { + "epoch": 1.64, + "learning_rate": 3.6358909174948e-05, + "loss": 0.0, + "step": 7083 + }, + { + "epoch": 1.64, + "learning_rate": 3.635698328326015e-05, + "loss": 0.0, + "step": 7084 + }, + { + "epoch": 1.64, + "learning_rate": 3.63550573915723e-05, + "loss": 0.0001, + "step": 7085 + }, + { + "epoch": 1.64, + "learning_rate": 3.635313149988444e-05, + "loss": 0.0, + "step": 7086 + }, + { + "epoch": 1.64, + "learning_rate": 3.63512056081966e-05, + "loss": 0.0001, + "step": 7087 + }, + { + "epoch": 1.64, + "learning_rate": 3.6349279716508745e-05, + "loss": 0.0, + "step": 7088 + }, + { + "epoch": 1.64, + "learning_rate": 3.634735382482089e-05, + "loss": 0.0001, + "step": 7089 + }, + { + "epoch": 1.64, + "learning_rate": 3.634542793313304e-05, + "loss": 0.0, + "step": 7090 + }, + { + "epoch": 1.64, + "learning_rate": 3.6343502041445194e-05, + "loss": 0.0002, + "step": 7091 + }, + { + "epoch": 1.64, + "learning_rate": 3.634157614975734e-05, + "loss": 0.0, + "step": 7092 + }, + { + "epoch": 1.64, + "learning_rate": 3.633965025806948e-05, + "loss": 0.0023, + "step": 7093 + }, + { + "epoch": 1.64, + "learning_rate": 3.6337724366381637e-05, + "loss": 0.0, + "step": 7094 + }, + { + "epoch": 1.64, + "learning_rate": 3.6335798474693784e-05, + "loss": 0.0, + "step": 7095 + }, + { + "epoch": 1.64, + "learning_rate": 3.633387258300593e-05, + "loss": 0.0003, + "step": 7096 + }, + { + "epoch": 1.64, + "learning_rate": 3.633194669131808e-05, + "loss": 0.0, + "step": 7097 + }, + { + "epoch": 1.64, + "learning_rate": 3.633002079963023e-05, + "loss": 0.0, + "step": 7098 + }, + { + "epoch": 1.64, + "learning_rate": 3.632809490794238e-05, + "loss": 0.0, + "step": 7099 + }, + { + "epoch": 1.64, + "learning_rate": 3.632616901625453e-05, + "loss": 0.0002, + "step": 7100 + }, + { + "epoch": 1.64, + "learning_rate": 3.6324243124566676e-05, + "loss": 0.0001, + "step": 7101 + }, + { + "epoch": 1.64, + "learning_rate": 3.632231723287882e-05, + "loss": 0.0003, + "step": 7102 + }, + { + "epoch": 1.64, + "learning_rate": 3.632039134119097e-05, + "loss": 0.0, + "step": 7103 + }, + { + "epoch": 1.64, + "learning_rate": 3.631846544950312e-05, + "loss": 0.0, + "step": 7104 + }, + { + "epoch": 1.64, + "learning_rate": 3.631653955781527e-05, + "loss": 0.0, + "step": 7105 + }, + { + "epoch": 1.64, + "learning_rate": 3.631461366612742e-05, + "loss": 0.0002, + "step": 7106 + }, + { + "epoch": 1.64, + "learning_rate": 3.631268777443957e-05, + "loss": 0.0009, + "step": 7107 + }, + { + "epoch": 1.64, + "learning_rate": 3.6310761882751715e-05, + "loss": 0.0, + "step": 7108 + }, + { + "epoch": 1.64, + "learning_rate": 3.630883599106387e-05, + "loss": 0.0, + "step": 7109 + }, + { + "epoch": 1.64, + "learning_rate": 3.630691009937601e-05, + "loss": 0.0016, + "step": 7110 + }, + { + "epoch": 1.64, + "learning_rate": 3.6304984207688164e-05, + "loss": 0.0, + "step": 7111 + }, + { + "epoch": 1.64, + "learning_rate": 3.630305831600031e-05, + "loss": 0.0, + "step": 7112 + }, + { + "epoch": 1.64, + "learning_rate": 3.630113242431246e-05, + "loss": 0.0001, + "step": 7113 + }, + { + "epoch": 1.64, + "learning_rate": 3.6299206532624606e-05, + "loss": 0.0, + "step": 7114 + }, + { + "epoch": 1.64, + "learning_rate": 3.6297280640936754e-05, + "loss": 0.0, + "step": 7115 + }, + { + "epoch": 1.64, + "learning_rate": 3.629535474924891e-05, + "loss": 0.0, + "step": 7116 + }, + { + "epoch": 1.64, + "learning_rate": 3.629342885756105e-05, + "loss": 0.0002, + "step": 7117 + }, + { + "epoch": 1.65, + "learning_rate": 3.62915029658732e-05, + "loss": 0.0, + "step": 7118 + }, + { + "epoch": 1.65, + "learning_rate": 3.628957707418535e-05, + "loss": 0.0001, + "step": 7119 + }, + { + "epoch": 1.65, + "learning_rate": 3.62876511824975e-05, + "loss": 0.0, + "step": 7120 + }, + { + "epoch": 1.65, + "learning_rate": 3.6285725290809646e-05, + "loss": 0.0, + "step": 7121 + }, + { + "epoch": 1.65, + "learning_rate": 3.628379939912179e-05, + "loss": 0.0003, + "step": 7122 + }, + { + "epoch": 1.65, + "learning_rate": 3.628187350743395e-05, + "loss": 0.0006, + "step": 7123 + }, + { + "epoch": 1.65, + "learning_rate": 3.627994761574609e-05, + "loss": 0.0, + "step": 7124 + }, + { + "epoch": 1.65, + "learning_rate": 3.627802172405824e-05, + "loss": 0.0, + "step": 7125 + }, + { + "epoch": 1.65, + "learning_rate": 3.627609583237039e-05, + "loss": 0.0, + "step": 7126 + }, + { + "epoch": 1.65, + "learning_rate": 3.627416994068254e-05, + "loss": 0.0013, + "step": 7127 + }, + { + "epoch": 1.65, + "learning_rate": 3.6272244048994685e-05, + "loss": 0.0031, + "step": 7128 + }, + { + "epoch": 1.65, + "learning_rate": 3.627031815730684e-05, + "loss": 0.0001, + "step": 7129 + }, + { + "epoch": 1.65, + "learning_rate": 3.6268392265618986e-05, + "loss": 0.0, + "step": 7130 + }, + { + "epoch": 1.65, + "learning_rate": 3.626646637393113e-05, + "loss": 0.0, + "step": 7131 + }, + { + "epoch": 1.65, + "learning_rate": 3.626454048224328e-05, + "loss": 0.0006, + "step": 7132 + }, + { + "epoch": 1.65, + "learning_rate": 3.626261459055543e-05, + "loss": 0.0, + "step": 7133 + }, + { + "epoch": 1.65, + "learning_rate": 3.6260688698867576e-05, + "loss": 0.0, + "step": 7134 + }, + { + "epoch": 1.65, + "learning_rate": 3.6258762807179724e-05, + "loss": 0.0, + "step": 7135 + }, + { + "epoch": 1.65, + "learning_rate": 3.625683691549188e-05, + "loss": 0.0005, + "step": 7136 + }, + { + "epoch": 1.65, + "learning_rate": 3.6254911023804025e-05, + "loss": 0.0001, + "step": 7137 + }, + { + "epoch": 1.65, + "learning_rate": 3.625298513211617e-05, + "loss": 0.0001, + "step": 7138 + }, + { + "epoch": 1.65, + "learning_rate": 3.625105924042832e-05, + "loss": 0.0, + "step": 7139 + }, + { + "epoch": 1.65, + "learning_rate": 3.624913334874047e-05, + "loss": 0.0, + "step": 7140 + }, + { + "epoch": 1.65, + "learning_rate": 3.6247207457052615e-05, + "loss": 0.0, + "step": 7141 + }, + { + "epoch": 1.65, + "learning_rate": 3.624528156536476e-05, + "loss": 0.0, + "step": 7142 + }, + { + "epoch": 1.65, + "learning_rate": 3.624335567367692e-05, + "loss": 0.0, + "step": 7143 + }, + { + "epoch": 1.65, + "learning_rate": 3.6241429781989065e-05, + "loss": 0.0001, + "step": 7144 + }, + { + "epoch": 1.65, + "learning_rate": 3.623950389030121e-05, + "loss": 0.0008, + "step": 7145 + }, + { + "epoch": 1.65, + "learning_rate": 3.623757799861336e-05, + "loss": 0.0, + "step": 7146 + }, + { + "epoch": 1.65, + "learning_rate": 3.6235652106925514e-05, + "loss": 0.0, + "step": 7147 + }, + { + "epoch": 1.65, + "learning_rate": 3.6233726215237655e-05, + "loss": 0.0005, + "step": 7148 + }, + { + "epoch": 1.65, + "learning_rate": 3.62318003235498e-05, + "loss": 0.0, + "step": 7149 + }, + { + "epoch": 1.65, + "learning_rate": 3.6229874431861956e-05, + "loss": 0.0001, + "step": 7150 + }, + { + "epoch": 1.65, + "learning_rate": 3.62279485401741e-05, + "loss": 0.0, + "step": 7151 + }, + { + "epoch": 1.65, + "learning_rate": 3.622602264848625e-05, + "loss": 0.0004, + "step": 7152 + }, + { + "epoch": 1.65, + "learning_rate": 3.62240967567984e-05, + "loss": 0.0001, + "step": 7153 + }, + { + "epoch": 1.65, + "learning_rate": 3.622217086511055e-05, + "loss": 0.0, + "step": 7154 + }, + { + "epoch": 1.65, + "learning_rate": 3.6220244973422694e-05, + "loss": 0.0, + "step": 7155 + }, + { + "epoch": 1.65, + "learning_rate": 3.621831908173485e-05, + "loss": 0.0, + "step": 7156 + }, + { + "epoch": 1.65, + "learning_rate": 3.6216393190046995e-05, + "loss": 0.0, + "step": 7157 + }, + { + "epoch": 1.65, + "learning_rate": 3.6214467298359136e-05, + "loss": 0.0, + "step": 7158 + }, + { + "epoch": 1.65, + "learning_rate": 3.621254140667129e-05, + "loss": 0.0, + "step": 7159 + }, + { + "epoch": 1.65, + "learning_rate": 3.621061551498344e-05, + "loss": 0.0001, + "step": 7160 + }, + { + "epoch": 1.65, + "learning_rate": 3.6208689623295585e-05, + "loss": 0.0008, + "step": 7161 + }, + { + "epoch": 1.66, + "learning_rate": 3.620676373160773e-05, + "loss": 0.0002, + "step": 7162 + }, + { + "epoch": 1.66, + "learning_rate": 3.620483783991989e-05, + "loss": 0.0, + "step": 7163 + }, + { + "epoch": 1.66, + "learning_rate": 3.6202911948232035e-05, + "loss": 0.0002, + "step": 7164 + }, + { + "epoch": 1.66, + "learning_rate": 3.620098605654418e-05, + "loss": 0.0, + "step": 7165 + }, + { + "epoch": 1.66, + "learning_rate": 3.619906016485633e-05, + "loss": 0.0, + "step": 7166 + }, + { + "epoch": 1.66, + "learning_rate": 3.6197134273168484e-05, + "loss": 0.0, + "step": 7167 + }, + { + "epoch": 1.66, + "learning_rate": 3.6195208381480624e-05, + "loss": 0.0006, + "step": 7168 + }, + { + "epoch": 1.66, + "learning_rate": 3.619328248979277e-05, + "loss": 0.0, + "step": 7169 + }, + { + "epoch": 1.66, + "learning_rate": 3.6191356598104926e-05, + "loss": 0.0, + "step": 7170 + }, + { + "epoch": 1.66, + "learning_rate": 3.6189430706417074e-05, + "loss": 0.0001, + "step": 7171 + }, + { + "epoch": 1.66, + "learning_rate": 3.618750481472922e-05, + "loss": 0.0, + "step": 7172 + }, + { + "epoch": 1.66, + "learning_rate": 3.618557892304137e-05, + "loss": 0.0, + "step": 7173 + }, + { + "epoch": 1.66, + "learning_rate": 3.618365303135352e-05, + "loss": 0.0002, + "step": 7174 + }, + { + "epoch": 1.66, + "learning_rate": 3.6181727139665664e-05, + "loss": 0.0, + "step": 7175 + }, + { + "epoch": 1.66, + "learning_rate": 3.617980124797782e-05, + "loss": 0.0009, + "step": 7176 + }, + { + "epoch": 1.66, + "learning_rate": 3.6177875356289965e-05, + "loss": 0.0009, + "step": 7177 + }, + { + "epoch": 1.66, + "learning_rate": 3.617594946460211e-05, + "loss": 0.0001, + "step": 7178 + }, + { + "epoch": 1.66, + "learning_rate": 3.617402357291426e-05, + "loss": 0.0001, + "step": 7179 + }, + { + "epoch": 1.66, + "learning_rate": 3.617209768122641e-05, + "loss": 0.0097, + "step": 7180 + }, + { + "epoch": 1.66, + "learning_rate": 3.617017178953856e-05, + "loss": 0.0001, + "step": 7181 + }, + { + "epoch": 1.66, + "learning_rate": 3.61682458978507e-05, + "loss": 0.0, + "step": 7182 + }, + { + "epoch": 1.66, + "learning_rate": 3.616632000616286e-05, + "loss": 0.0, + "step": 7183 + }, + { + "epoch": 1.66, + "learning_rate": 3.6164394114475004e-05, + "loss": 0.0001, + "step": 7184 + }, + { + "epoch": 1.66, + "learning_rate": 3.616246822278715e-05, + "loss": 0.0, + "step": 7185 + }, + { + "epoch": 1.66, + "learning_rate": 3.61605423310993e-05, + "loss": 0.0, + "step": 7186 + }, + { + "epoch": 1.66, + "learning_rate": 3.615861643941145e-05, + "loss": 0.0038, + "step": 7187 + }, + { + "epoch": 1.66, + "learning_rate": 3.61566905477236e-05, + "loss": 0.0001, + "step": 7188 + }, + { + "epoch": 1.66, + "learning_rate": 3.615476465603574e-05, + "loss": 0.0, + "step": 7189 + }, + { + "epoch": 1.66, + "learning_rate": 3.6152838764347896e-05, + "loss": 0.0, + "step": 7190 + }, + { + "epoch": 1.66, + "learning_rate": 3.6150912872660044e-05, + "loss": 0.0, + "step": 7191 + }, + { + "epoch": 1.66, + "learning_rate": 3.614898698097219e-05, + "loss": 0.0001, + "step": 7192 + }, + { + "epoch": 1.66, + "learning_rate": 3.614706108928434e-05, + "loss": 0.0001, + "step": 7193 + }, + { + "epoch": 1.66, + "learning_rate": 3.614513519759649e-05, + "loss": 0.0, + "step": 7194 + }, + { + "epoch": 1.66, + "learning_rate": 3.614320930590864e-05, + "loss": 0.0, + "step": 7195 + }, + { + "epoch": 1.66, + "learning_rate": 3.614128341422078e-05, + "loss": 0.0002, + "step": 7196 + }, + { + "epoch": 1.66, + "learning_rate": 3.6139357522532935e-05, + "loss": 0.0003, + "step": 7197 + }, + { + "epoch": 1.66, + "learning_rate": 3.613743163084508e-05, + "loss": 0.0, + "step": 7198 + }, + { + "epoch": 1.66, + "learning_rate": 3.613550573915723e-05, + "loss": 0.0, + "step": 7199 + }, + { + "epoch": 1.66, + "learning_rate": 3.613357984746938e-05, + "loss": 0.0, + "step": 7200 + }, + { + "epoch": 1.66, + "learning_rate": 3.613165395578153e-05, + "loss": 0.0, + "step": 7201 + }, + { + "epoch": 1.66, + "learning_rate": 3.612972806409368e-05, + "loss": 0.0, + "step": 7202 + }, + { + "epoch": 1.66, + "learning_rate": 3.612780217240583e-05, + "loss": 0.0, + "step": 7203 + }, + { + "epoch": 1.66, + "learning_rate": 3.6125876280717974e-05, + "loss": 0.0005, + "step": 7204 + }, + { + "epoch": 1.67, + "learning_rate": 3.612395038903012e-05, + "loss": 0.0, + "step": 7205 + }, + { + "epoch": 1.67, + "learning_rate": 3.612202449734227e-05, + "loss": 0.0, + "step": 7206 + }, + { + "epoch": 1.67, + "learning_rate": 3.612009860565442e-05, + "loss": 0.0001, + "step": 7207 + }, + { + "epoch": 1.67, + "learning_rate": 3.611817271396657e-05, + "loss": 0.0, + "step": 7208 + }, + { + "epoch": 1.67, + "learning_rate": 3.611624682227872e-05, + "loss": 0.0001, + "step": 7209 + }, + { + "epoch": 1.67, + "learning_rate": 3.6114320930590866e-05, + "loss": 0.0009, + "step": 7210 + }, + { + "epoch": 1.67, + "learning_rate": 3.6112395038903013e-05, + "loss": 0.0001, + "step": 7211 + }, + { + "epoch": 1.67, + "learning_rate": 3.611046914721517e-05, + "loss": 0.0001, + "step": 7212 + }, + { + "epoch": 1.67, + "learning_rate": 3.610854325552731e-05, + "loss": 0.0, + "step": 7213 + }, + { + "epoch": 1.67, + "learning_rate": 3.6106617363839456e-05, + "loss": 0.0004, + "step": 7214 + }, + { + "epoch": 1.67, + "learning_rate": 3.610469147215161e-05, + "loss": 0.0008, + "step": 7215 + }, + { + "epoch": 1.67, + "learning_rate": 3.610276558046376e-05, + "loss": 0.0001, + "step": 7216 + }, + { + "epoch": 1.67, + "learning_rate": 3.6100839688775905e-05, + "loss": 0.0, + "step": 7217 + }, + { + "epoch": 1.67, + "learning_rate": 3.609891379708805e-05, + "loss": 0.0001, + "step": 7218 + }, + { + "epoch": 1.67, + "learning_rate": 3.609698790540021e-05, + "loss": 0.0021, + "step": 7219 + }, + { + "epoch": 1.67, + "learning_rate": 3.609506201371235e-05, + "loss": 0.0001, + "step": 7220 + }, + { + "epoch": 1.67, + "learning_rate": 3.60931361220245e-05, + "loss": 0.001, + "step": 7221 + }, + { + "epoch": 1.67, + "learning_rate": 3.609121023033665e-05, + "loss": 0.0002, + "step": 7222 + }, + { + "epoch": 1.67, + "learning_rate": 3.60892843386488e-05, + "loss": 0.0001, + "step": 7223 + }, + { + "epoch": 1.67, + "learning_rate": 3.6087358446960944e-05, + "loss": 0.0002, + "step": 7224 + }, + { + "epoch": 1.67, + "learning_rate": 3.608543255527309e-05, + "loss": 0.0, + "step": 7225 + }, + { + "epoch": 1.67, + "learning_rate": 3.6083506663585246e-05, + "loss": 0.0, + "step": 7226 + }, + { + "epoch": 1.67, + "learning_rate": 3.6081580771897387e-05, + "loss": 0.0, + "step": 7227 + }, + { + "epoch": 1.67, + "learning_rate": 3.607965488020954e-05, + "loss": 0.0001, + "step": 7228 + }, + { + "epoch": 1.67, + "learning_rate": 3.607772898852169e-05, + "loss": 0.0, + "step": 7229 + }, + { + "epoch": 1.67, + "learning_rate": 3.6075803096833836e-05, + "loss": 0.0, + "step": 7230 + }, + { + "epoch": 1.67, + "learning_rate": 3.607387720514598e-05, + "loss": 0.0082, + "step": 7231 + }, + { + "epoch": 1.67, + "learning_rate": 3.607195131345814e-05, + "loss": 0.0019, + "step": 7232 + }, + { + "epoch": 1.67, + "learning_rate": 3.607002542177028e-05, + "loss": 0.0, + "step": 7233 + }, + { + "epoch": 1.67, + "learning_rate": 3.6068099530082426e-05, + "loss": 0.0, + "step": 7234 + }, + { + "epoch": 1.67, + "learning_rate": 3.606617363839458e-05, + "loss": 0.0002, + "step": 7235 + }, + { + "epoch": 1.67, + "learning_rate": 3.606424774670673e-05, + "loss": 0.0001, + "step": 7236 + }, + { + "epoch": 1.67, + "learning_rate": 3.6062321855018875e-05, + "loss": 0.0001, + "step": 7237 + }, + { + "epoch": 1.67, + "learning_rate": 3.606039596333102e-05, + "loss": 0.0, + "step": 7238 + }, + { + "epoch": 1.67, + "learning_rate": 3.605847007164318e-05, + "loss": 0.0, + "step": 7239 + }, + { + "epoch": 1.67, + "learning_rate": 3.605654417995532e-05, + "loss": 0.0, + "step": 7240 + }, + { + "epoch": 1.67, + "learning_rate": 3.605461828826747e-05, + "loss": 0.0013, + "step": 7241 + }, + { + "epoch": 1.67, + "learning_rate": 3.605269239657962e-05, + "loss": 0.0001, + "step": 7242 + }, + { + "epoch": 1.67, + "learning_rate": 3.6050766504891767e-05, + "loss": 0.0001, + "step": 7243 + }, + { + "epoch": 1.67, + "learning_rate": 3.6048840613203914e-05, + "loss": 0.0001, + "step": 7244 + }, + { + "epoch": 1.67, + "learning_rate": 3.604691472151606e-05, + "loss": 0.0003, + "step": 7245 + }, + { + "epoch": 1.67, + "learning_rate": 3.6044988829828216e-05, + "loss": 0.0004, + "step": 7246 + }, + { + "epoch": 1.67, + "learning_rate": 3.6043062938140356e-05, + "loss": 0.0, + "step": 7247 + }, + { + "epoch": 1.68, + "learning_rate": 3.604113704645251e-05, + "loss": 0.0007, + "step": 7248 + }, + { + "epoch": 1.68, + "learning_rate": 3.603921115476466e-05, + "loss": 0.0, + "step": 7249 + }, + { + "epoch": 1.68, + "learning_rate": 3.6037285263076806e-05, + "loss": 0.0, + "step": 7250 + }, + { + "epoch": 1.68, + "learning_rate": 3.603535937138895e-05, + "loss": 0.0, + "step": 7251 + }, + { + "epoch": 1.68, + "learning_rate": 3.60334334797011e-05, + "loss": 0.0, + "step": 7252 + }, + { + "epoch": 1.68, + "learning_rate": 3.6031507588013255e-05, + "loss": 0.0001, + "step": 7253 + }, + { + "epoch": 1.68, + "learning_rate": 3.6029581696325396e-05, + "loss": 0.0, + "step": 7254 + }, + { + "epoch": 1.68, + "learning_rate": 3.602765580463755e-05, + "loss": 0.0011, + "step": 7255 + }, + { + "epoch": 1.68, + "learning_rate": 3.60257299129497e-05, + "loss": 0.0, + "step": 7256 + }, + { + "epoch": 1.68, + "learning_rate": 3.6023804021261845e-05, + "loss": 0.0002, + "step": 7257 + }, + { + "epoch": 1.68, + "learning_rate": 3.602187812957399e-05, + "loss": 0.0, + "step": 7258 + }, + { + "epoch": 1.68, + "learning_rate": 3.6019952237886147e-05, + "loss": 0.0, + "step": 7259 + }, + { + "epoch": 1.68, + "learning_rate": 3.6018026346198294e-05, + "loss": 0.0, + "step": 7260 + }, + { + "epoch": 1.68, + "learning_rate": 3.6016100454510435e-05, + "loss": 0.0001, + "step": 7261 + }, + { + "epoch": 1.68, + "learning_rate": 3.601417456282259e-05, + "loss": 0.0, + "step": 7262 + }, + { + "epoch": 1.68, + "learning_rate": 3.6012248671134736e-05, + "loss": 0.0, + "step": 7263 + }, + { + "epoch": 1.68, + "learning_rate": 3.6010322779446884e-05, + "loss": 0.0001, + "step": 7264 + }, + { + "epoch": 1.68, + "learning_rate": 3.600839688775903e-05, + "loss": 0.0, + "step": 7265 + }, + { + "epoch": 1.68, + "learning_rate": 3.6006470996071186e-05, + "loss": 0.0, + "step": 7266 + }, + { + "epoch": 1.68, + "learning_rate": 3.600454510438333e-05, + "loss": 0.0001, + "step": 7267 + }, + { + "epoch": 1.68, + "learning_rate": 3.600261921269548e-05, + "loss": 0.0, + "step": 7268 + }, + { + "epoch": 1.68, + "learning_rate": 3.600069332100763e-05, + "loss": 0.0, + "step": 7269 + }, + { + "epoch": 1.68, + "learning_rate": 3.5998767429319776e-05, + "loss": 0.0002, + "step": 7270 + }, + { + "epoch": 1.68, + "learning_rate": 3.599684153763192e-05, + "loss": 0.0004, + "step": 7271 + }, + { + "epoch": 1.68, + "learning_rate": 3.599491564594407e-05, + "loss": 0.0, + "step": 7272 + }, + { + "epoch": 1.68, + "learning_rate": 3.5992989754256225e-05, + "loss": 0.0, + "step": 7273 + }, + { + "epoch": 1.68, + "learning_rate": 3.599106386256837e-05, + "loss": 0.0002, + "step": 7274 + }, + { + "epoch": 1.68, + "learning_rate": 3.598913797088052e-05, + "loss": 0.0, + "step": 7275 + }, + { + "epoch": 1.68, + "learning_rate": 3.598721207919267e-05, + "loss": 0.0001, + "step": 7276 + }, + { + "epoch": 1.68, + "learning_rate": 3.598528618750482e-05, + "loss": 0.0033, + "step": 7277 + }, + { + "epoch": 1.68, + "learning_rate": 3.598336029581696e-05, + "loss": 0.0001, + "step": 7278 + }, + { + "epoch": 1.68, + "learning_rate": 3.5981434404129116e-05, + "loss": 0.0, + "step": 7279 + }, + { + "epoch": 1.68, + "learning_rate": 3.5979508512441264e-05, + "loss": 0.0, + "step": 7280 + }, + { + "epoch": 1.68, + "learning_rate": 3.597758262075341e-05, + "loss": 0.0, + "step": 7281 + }, + { + "epoch": 1.68, + "learning_rate": 3.597565672906556e-05, + "loss": 0.0001, + "step": 7282 + }, + { + "epoch": 1.68, + "learning_rate": 3.5973730837377706e-05, + "loss": 0.0, + "step": 7283 + }, + { + "epoch": 1.68, + "learning_rate": 3.597180494568986e-05, + "loss": 0.0, + "step": 7284 + }, + { + "epoch": 1.68, + "learning_rate": 3.5969879054002e-05, + "loss": 0.0, + "step": 7285 + }, + { + "epoch": 1.68, + "learning_rate": 3.5967953162314156e-05, + "loss": 0.0, + "step": 7286 + }, + { + "epoch": 1.68, + "learning_rate": 3.59660272706263e-05, + "loss": 0.0001, + "step": 7287 + }, + { + "epoch": 1.68, + "learning_rate": 3.596410137893845e-05, + "loss": 0.0032, + "step": 7288 + }, + { + "epoch": 1.68, + "learning_rate": 3.59621754872506e-05, + "loss": 0.0002, + "step": 7289 + }, + { + "epoch": 1.68, + "learning_rate": 3.5960249595562745e-05, + "loss": 0.0003, + "step": 7290 + }, + { + "epoch": 1.69, + "learning_rate": 3.59583237038749e-05, + "loss": 0.0, + "step": 7291 + }, + { + "epoch": 1.69, + "learning_rate": 3.595639781218704e-05, + "loss": 0.0003, + "step": 7292 + }, + { + "epoch": 1.69, + "learning_rate": 3.5954471920499195e-05, + "loss": 0.0003, + "step": 7293 + }, + { + "epoch": 1.69, + "learning_rate": 3.595254602881134e-05, + "loss": 0.0, + "step": 7294 + }, + { + "epoch": 1.69, + "learning_rate": 3.595062013712349e-05, + "loss": 0.0002, + "step": 7295 + }, + { + "epoch": 1.69, + "learning_rate": 3.594869424543564e-05, + "loss": 0.0001, + "step": 7296 + }, + { + "epoch": 1.69, + "learning_rate": 3.594676835374779e-05, + "loss": 0.0, + "step": 7297 + }, + { + "epoch": 1.69, + "learning_rate": 3.594484246205994e-05, + "loss": 0.0, + "step": 7298 + }, + { + "epoch": 1.69, + "learning_rate": 3.594291657037208e-05, + "loss": 0.0, + "step": 7299 + }, + { + "epoch": 1.69, + "learning_rate": 3.5940990678684234e-05, + "loss": 0.0, + "step": 7300 + }, + { + "epoch": 1.69, + "learning_rate": 3.593906478699638e-05, + "loss": 0.0026, + "step": 7301 + }, + { + "epoch": 1.69, + "learning_rate": 3.593713889530853e-05, + "loss": 0.0001, + "step": 7302 + }, + { + "epoch": 1.69, + "learning_rate": 3.5935213003620676e-05, + "loss": 0.0, + "step": 7303 + }, + { + "epoch": 1.69, + "learning_rate": 3.593328711193283e-05, + "loss": 0.0, + "step": 7304 + }, + { + "epoch": 1.69, + "learning_rate": 3.593136122024497e-05, + "loss": 0.0, + "step": 7305 + }, + { + "epoch": 1.69, + "learning_rate": 3.5929435328557125e-05, + "loss": 0.0017, + "step": 7306 + }, + { + "epoch": 1.69, + "learning_rate": 3.592750943686927e-05, + "loss": 0.0, + "step": 7307 + }, + { + "epoch": 1.69, + "learning_rate": 3.592558354518142e-05, + "loss": 0.0, + "step": 7308 + }, + { + "epoch": 1.69, + "learning_rate": 3.592365765349357e-05, + "loss": 0.0, + "step": 7309 + }, + { + "epoch": 1.69, + "learning_rate": 3.5921731761805715e-05, + "loss": 0.0001, + "step": 7310 + }, + { + "epoch": 1.69, + "learning_rate": 3.591980587011787e-05, + "loss": 0.0004, + "step": 7311 + }, + { + "epoch": 1.69, + "learning_rate": 3.591787997843001e-05, + "loss": 0.0, + "step": 7312 + }, + { + "epoch": 1.69, + "learning_rate": 3.5915954086742165e-05, + "loss": 0.0002, + "step": 7313 + }, + { + "epoch": 1.69, + "learning_rate": 3.591402819505431e-05, + "loss": 0.0, + "step": 7314 + }, + { + "epoch": 1.69, + "learning_rate": 3.591210230336646e-05, + "loss": 0.0, + "step": 7315 + }, + { + "epoch": 1.69, + "learning_rate": 3.591017641167861e-05, + "loss": 0.0001, + "step": 7316 + }, + { + "epoch": 1.69, + "learning_rate": 3.5908250519990754e-05, + "loss": 0.0003, + "step": 7317 + }, + { + "epoch": 1.69, + "learning_rate": 3.590632462830291e-05, + "loss": 0.0, + "step": 7318 + }, + { + "epoch": 1.69, + "learning_rate": 3.590439873661505e-05, + "loss": 0.0, + "step": 7319 + }, + { + "epoch": 1.69, + "learning_rate": 3.5902472844927204e-05, + "loss": 0.0, + "step": 7320 + }, + { + "epoch": 1.69, + "learning_rate": 3.590054695323935e-05, + "loss": 0.0, + "step": 7321 + }, + { + "epoch": 1.69, + "learning_rate": 3.58986210615515e-05, + "loss": 0.0005, + "step": 7322 + }, + { + "epoch": 1.69, + "learning_rate": 3.5896695169863646e-05, + "loss": 0.0, + "step": 7323 + }, + { + "epoch": 1.69, + "learning_rate": 3.58947692781758e-05, + "loss": 0.0001, + "step": 7324 + }, + { + "epoch": 1.69, + "learning_rate": 3.589284338648795e-05, + "loss": 0.0002, + "step": 7325 + }, + { + "epoch": 1.69, + "learning_rate": 3.589091749480009e-05, + "loss": 0.0, + "step": 7326 + }, + { + "epoch": 1.69, + "learning_rate": 3.588899160311224e-05, + "loss": 0.0029, + "step": 7327 + }, + { + "epoch": 1.69, + "learning_rate": 3.588706571142439e-05, + "loss": 0.0001, + "step": 7328 + }, + { + "epoch": 1.69, + "learning_rate": 3.588513981973654e-05, + "loss": 0.0002, + "step": 7329 + }, + { + "epoch": 1.69, + "learning_rate": 3.5883213928048685e-05, + "loss": 0.0, + "step": 7330 + }, + { + "epoch": 1.69, + "learning_rate": 3.588128803636084e-05, + "loss": 0.0, + "step": 7331 + }, + { + "epoch": 1.69, + "learning_rate": 3.587936214467299e-05, + "loss": 0.0, + "step": 7332 + }, + { + "epoch": 1.69, + "learning_rate": 3.5877436252985134e-05, + "loss": 0.0, + "step": 7333 + }, + { + "epoch": 1.69, + "learning_rate": 3.587551036129728e-05, + "loss": 0.0, + "step": 7334 + }, + { + "epoch": 1.7, + "learning_rate": 3.5873584469609436e-05, + "loss": 0.0, + "step": 7335 + }, + { + "epoch": 1.7, + "learning_rate": 3.587165857792158e-05, + "loss": 0.0001, + "step": 7336 + }, + { + "epoch": 1.7, + "learning_rate": 3.5869732686233724e-05, + "loss": 0.0001, + "step": 7337 + }, + { + "epoch": 1.7, + "learning_rate": 3.586780679454588e-05, + "loss": 0.0, + "step": 7338 + }, + { + "epoch": 1.7, + "learning_rate": 3.5865880902858026e-05, + "loss": 0.0, + "step": 7339 + }, + { + "epoch": 1.7, + "learning_rate": 3.5863955011170174e-05, + "loss": 0.0, + "step": 7340 + }, + { + "epoch": 1.7, + "learning_rate": 3.586202911948232e-05, + "loss": 0.0, + "step": 7341 + }, + { + "epoch": 1.7, + "learning_rate": 3.5860103227794475e-05, + "loss": 0.0, + "step": 7342 + }, + { + "epoch": 1.7, + "learning_rate": 3.5858177336106616e-05, + "loss": 0.0001, + "step": 7343 + }, + { + "epoch": 1.7, + "learning_rate": 3.585625144441877e-05, + "loss": 0.0, + "step": 7344 + }, + { + "epoch": 1.7, + "learning_rate": 3.585432555273092e-05, + "loss": 0.0, + "step": 7345 + }, + { + "epoch": 1.7, + "learning_rate": 3.5852399661043065e-05, + "loss": 0.0, + "step": 7346 + }, + { + "epoch": 1.7, + "learning_rate": 3.585047376935521e-05, + "loss": 0.0, + "step": 7347 + }, + { + "epoch": 1.7, + "learning_rate": 3.584854787766736e-05, + "loss": 0.0, + "step": 7348 + }, + { + "epoch": 1.7, + "learning_rate": 3.5846621985979514e-05, + "loss": 0.0, + "step": 7349 + }, + { + "epoch": 1.7, + "learning_rate": 3.5844696094291655e-05, + "loss": 0.001, + "step": 7350 + }, + { + "epoch": 1.7, + "learning_rate": 3.584277020260381e-05, + "loss": 0.0001, + "step": 7351 + }, + { + "epoch": 1.7, + "learning_rate": 3.584084431091596e-05, + "loss": 0.0004, + "step": 7352 + }, + { + "epoch": 1.7, + "learning_rate": 3.5838918419228104e-05, + "loss": 0.0001, + "step": 7353 + }, + { + "epoch": 1.7, + "learning_rate": 3.583699252754025e-05, + "loss": 0.0, + "step": 7354 + }, + { + "epoch": 1.7, + "learning_rate": 3.58350666358524e-05, + "loss": 0.0, + "step": 7355 + }, + { + "epoch": 1.7, + "learning_rate": 3.5833140744164554e-05, + "loss": 0.0, + "step": 7356 + }, + { + "epoch": 1.7, + "learning_rate": 3.5831214852476694e-05, + "loss": 0.0, + "step": 7357 + }, + { + "epoch": 1.7, + "learning_rate": 3.582928896078885e-05, + "loss": 0.0008, + "step": 7358 + }, + { + "epoch": 1.7, + "learning_rate": 3.5827363069100996e-05, + "loss": 0.0001, + "step": 7359 + }, + { + "epoch": 1.7, + "learning_rate": 3.5825437177413143e-05, + "loss": 0.0, + "step": 7360 + }, + { + "epoch": 1.7, + "learning_rate": 3.582351128572529e-05, + "loss": 0.0, + "step": 7361 + }, + { + "epoch": 1.7, + "learning_rate": 3.5821585394037445e-05, + "loss": 0.0001, + "step": 7362 + }, + { + "epoch": 1.7, + "learning_rate": 3.581965950234959e-05, + "loss": 0.0007, + "step": 7363 + }, + { + "epoch": 1.7, + "learning_rate": 3.581773361066173e-05, + "loss": 0.0, + "step": 7364 + }, + { + "epoch": 1.7, + "learning_rate": 3.581580771897389e-05, + "loss": 0.0, + "step": 7365 + }, + { + "epoch": 1.7, + "learning_rate": 3.5813881827286035e-05, + "loss": 0.0001, + "step": 7366 + }, + { + "epoch": 1.7, + "learning_rate": 3.581195593559818e-05, + "loss": 0.0, + "step": 7367 + }, + { + "epoch": 1.7, + "learning_rate": 3.581003004391033e-05, + "loss": 0.0005, + "step": 7368 + }, + { + "epoch": 1.7, + "learning_rate": 3.5808104152222484e-05, + "loss": 0.0, + "step": 7369 + }, + { + "epoch": 1.7, + "learning_rate": 3.580617826053463e-05, + "loss": 0.0, + "step": 7370 + }, + { + "epoch": 1.7, + "learning_rate": 3.580425236884678e-05, + "loss": 0.0001, + "step": 7371 + }, + { + "epoch": 1.7, + "learning_rate": 3.580232647715893e-05, + "loss": 0.001, + "step": 7372 + }, + { + "epoch": 1.7, + "learning_rate": 3.5800400585471074e-05, + "loss": 0.0, + "step": 7373 + }, + { + "epoch": 1.7, + "learning_rate": 3.579847469378322e-05, + "loss": 0.0001, + "step": 7374 + }, + { + "epoch": 1.7, + "learning_rate": 3.579654880209537e-05, + "loss": 0.0023, + "step": 7375 + }, + { + "epoch": 1.7, + "learning_rate": 3.5794622910407523e-05, + "loss": 0.0, + "step": 7376 + }, + { + "epoch": 1.7, + "learning_rate": 3.579269701871967e-05, + "loss": 0.0, + "step": 7377 + }, + { + "epoch": 1.71, + "learning_rate": 3.579077112703182e-05, + "loss": 0.0003, + "step": 7378 + }, + { + "epoch": 1.71, + "learning_rate": 3.5788845235343966e-05, + "loss": 0.0, + "step": 7379 + }, + { + "epoch": 1.71, + "learning_rate": 3.578691934365612e-05, + "loss": 0.0005, + "step": 7380 + }, + { + "epoch": 1.71, + "learning_rate": 3.578499345196826e-05, + "loss": 0.0001, + "step": 7381 + }, + { + "epoch": 1.71, + "learning_rate": 3.578306756028041e-05, + "loss": 0.0, + "step": 7382 + }, + { + "epoch": 1.71, + "learning_rate": 3.578114166859256e-05, + "loss": 0.0005, + "step": 7383 + }, + { + "epoch": 1.71, + "learning_rate": 3.57792157769047e-05, + "loss": 0.0, + "step": 7384 + }, + { + "epoch": 1.71, + "learning_rate": 3.577728988521686e-05, + "loss": 0.0001, + "step": 7385 + }, + { + "epoch": 1.71, + "learning_rate": 3.5775363993529005e-05, + "loss": 0.0001, + "step": 7386 + }, + { + "epoch": 1.71, + "learning_rate": 3.577343810184115e-05, + "loss": 0.0, + "step": 7387 + }, + { + "epoch": 1.71, + "learning_rate": 3.57715122101533e-05, + "loss": 0.0, + "step": 7388 + }, + { + "epoch": 1.71, + "learning_rate": 3.5769586318465454e-05, + "loss": 0.0, + "step": 7389 + }, + { + "epoch": 1.71, + "learning_rate": 3.57676604267776e-05, + "loss": 0.0, + "step": 7390 + }, + { + "epoch": 1.71, + "learning_rate": 3.576573453508975e-05, + "loss": 0.0, + "step": 7391 + }, + { + "epoch": 1.71, + "learning_rate": 3.5763808643401897e-05, + "loss": 0.0, + "step": 7392 + }, + { + "epoch": 1.71, + "learning_rate": 3.5761882751714044e-05, + "loss": 0.0001, + "step": 7393 + }, + { + "epoch": 1.71, + "learning_rate": 3.575995686002619e-05, + "loss": 0.0, + "step": 7394 + }, + { + "epoch": 1.71, + "learning_rate": 3.575803096833834e-05, + "loss": 0.0001, + "step": 7395 + }, + { + "epoch": 1.71, + "learning_rate": 3.575610507665049e-05, + "loss": 0.009, + "step": 7396 + }, + { + "epoch": 1.71, + "learning_rate": 3.575417918496264e-05, + "loss": 0.0001, + "step": 7397 + }, + { + "epoch": 1.71, + "learning_rate": 3.575225329327479e-05, + "loss": 0.0003, + "step": 7398 + }, + { + "epoch": 1.71, + "learning_rate": 3.5750327401586936e-05, + "loss": 0.0001, + "step": 7399 + }, + { + "epoch": 1.71, + "learning_rate": 3.574840150989909e-05, + "loss": 0.0, + "step": 7400 + }, + { + "epoch": 1.71, + "learning_rate": 3.574647561821123e-05, + "loss": 0.0, + "step": 7401 + }, + { + "epoch": 1.71, + "learning_rate": 3.574454972652338e-05, + "loss": 0.0, + "step": 7402 + }, + { + "epoch": 1.71, + "learning_rate": 3.574262383483553e-05, + "loss": 0.0012, + "step": 7403 + }, + { + "epoch": 1.71, + "learning_rate": 3.574069794314768e-05, + "loss": 0.0001, + "step": 7404 + }, + { + "epoch": 1.71, + "learning_rate": 3.573877205145983e-05, + "loss": 0.0, + "step": 7405 + }, + { + "epoch": 1.71, + "learning_rate": 3.5736846159771975e-05, + "loss": 0.0, + "step": 7406 + }, + { + "epoch": 1.71, + "learning_rate": 3.573492026808413e-05, + "loss": 0.0, + "step": 7407 + }, + { + "epoch": 1.71, + "learning_rate": 3.573299437639627e-05, + "loss": 0.0, + "step": 7408 + }, + { + "epoch": 1.71, + "learning_rate": 3.5731068484708424e-05, + "loss": 0.0, + "step": 7409 + }, + { + "epoch": 1.71, + "learning_rate": 3.572914259302057e-05, + "loss": 0.0001, + "step": 7410 + }, + { + "epoch": 1.71, + "learning_rate": 3.572721670133272e-05, + "loss": 0.0, + "step": 7411 + }, + { + "epoch": 1.71, + "learning_rate": 3.5725290809644866e-05, + "loss": 0.0, + "step": 7412 + }, + { + "epoch": 1.71, + "learning_rate": 3.5723364917957014e-05, + "loss": 0.0, + "step": 7413 + }, + { + "epoch": 1.71, + "learning_rate": 3.572143902626917e-05, + "loss": 0.0005, + "step": 7414 + }, + { + "epoch": 1.71, + "learning_rate": 3.571951313458131e-05, + "loss": 0.0, + "step": 7415 + }, + { + "epoch": 1.71, + "learning_rate": 3.571758724289346e-05, + "loss": 0.0, + "step": 7416 + }, + { + "epoch": 1.71, + "learning_rate": 3.571566135120561e-05, + "loss": 0.0, + "step": 7417 + }, + { + "epoch": 1.71, + "learning_rate": 3.571373545951776e-05, + "loss": 0.0006, + "step": 7418 + }, + { + "epoch": 1.71, + "learning_rate": 3.5711809567829906e-05, + "loss": 0.0, + "step": 7419 + }, + { + "epoch": 1.71, + "learning_rate": 3.570988367614205e-05, + "loss": 0.0001, + "step": 7420 + }, + { + "epoch": 1.72, + "learning_rate": 3.570795778445421e-05, + "loss": 0.0, + "step": 7421 + }, + { + "epoch": 1.72, + "learning_rate": 3.570603189276635e-05, + "loss": 0.0001, + "step": 7422 + }, + { + "epoch": 1.72, + "learning_rate": 3.57041060010785e-05, + "loss": 0.0, + "step": 7423 + }, + { + "epoch": 1.72, + "learning_rate": 3.570218010939065e-05, + "loss": 0.0, + "step": 7424 + }, + { + "epoch": 1.72, + "learning_rate": 3.57002542177028e-05, + "loss": 0.0002, + "step": 7425 + }, + { + "epoch": 1.72, + "learning_rate": 3.5698328326014945e-05, + "loss": 0.0, + "step": 7426 + }, + { + "epoch": 1.72, + "learning_rate": 3.56964024343271e-05, + "loss": 0.0027, + "step": 7427 + }, + { + "epoch": 1.72, + "learning_rate": 3.5694476542639246e-05, + "loss": 0.0, + "step": 7428 + }, + { + "epoch": 1.72, + "learning_rate": 3.569255065095139e-05, + "loss": 0.0, + "step": 7429 + }, + { + "epoch": 1.72, + "learning_rate": 3.569062475926354e-05, + "loss": 0.0017, + "step": 7430 + }, + { + "epoch": 1.72, + "learning_rate": 3.568869886757569e-05, + "loss": 0.0, + "step": 7431 + }, + { + "epoch": 1.72, + "learning_rate": 3.5686772975887836e-05, + "loss": 0.0, + "step": 7432 + }, + { + "epoch": 1.72, + "learning_rate": 3.5684847084199984e-05, + "loss": 0.0, + "step": 7433 + }, + { + "epoch": 1.72, + "learning_rate": 3.568292119251214e-05, + "loss": 0.0001, + "step": 7434 + }, + { + "epoch": 1.72, + "learning_rate": 3.5680995300824286e-05, + "loss": 0.0001, + "step": 7435 + }, + { + "epoch": 1.72, + "learning_rate": 3.567906940913643e-05, + "loss": 0.0, + "step": 7436 + }, + { + "epoch": 1.72, + "learning_rate": 3.567714351744858e-05, + "loss": 0.0, + "step": 7437 + }, + { + "epoch": 1.72, + "learning_rate": 3.567521762576073e-05, + "loss": 0.0, + "step": 7438 + }, + { + "epoch": 1.72, + "learning_rate": 3.5673291734072875e-05, + "loss": 0.0, + "step": 7439 + }, + { + "epoch": 1.72, + "learning_rate": 3.567136584238502e-05, + "loss": 0.0, + "step": 7440 + }, + { + "epoch": 1.72, + "learning_rate": 3.566943995069718e-05, + "loss": 0.0, + "step": 7441 + }, + { + "epoch": 1.72, + "learning_rate": 3.5667514059009325e-05, + "loss": 0.0, + "step": 7442 + }, + { + "epoch": 1.72, + "learning_rate": 3.566558816732147e-05, + "loss": 0.0001, + "step": 7443 + }, + { + "epoch": 1.72, + "learning_rate": 3.566366227563362e-05, + "loss": 0.0006, + "step": 7444 + }, + { + "epoch": 1.72, + "learning_rate": 3.5661736383945774e-05, + "loss": 0.0, + "step": 7445 + }, + { + "epoch": 1.72, + "learning_rate": 3.5659810492257915e-05, + "loss": 0.0001, + "step": 7446 + }, + { + "epoch": 1.72, + "learning_rate": 3.565788460057007e-05, + "loss": 0.0, + "step": 7447 + }, + { + "epoch": 1.72, + "learning_rate": 3.5655958708882216e-05, + "loss": 0.0, + "step": 7448 + }, + { + "epoch": 1.72, + "learning_rate": 3.5654032817194364e-05, + "loss": 0.0, + "step": 7449 + }, + { + "epoch": 1.72, + "learning_rate": 3.565210692550651e-05, + "loss": 0.0001, + "step": 7450 + }, + { + "epoch": 1.72, + "learning_rate": 3.565018103381866e-05, + "loss": 0.0002, + "step": 7451 + }, + { + "epoch": 1.72, + "learning_rate": 3.564825514213081e-05, + "loss": 0.0001, + "step": 7452 + }, + { + "epoch": 1.72, + "learning_rate": 3.5646329250442954e-05, + "loss": 0.0, + "step": 7453 + }, + { + "epoch": 1.72, + "learning_rate": 3.564440335875511e-05, + "loss": 0.0, + "step": 7454 + }, + { + "epoch": 1.72, + "learning_rate": 3.5642477467067255e-05, + "loss": 0.0001, + "step": 7455 + }, + { + "epoch": 1.72, + "learning_rate": 3.56405515753794e-05, + "loss": 0.0002, + "step": 7456 + }, + { + "epoch": 1.72, + "learning_rate": 3.563862568369155e-05, + "loss": 0.0001, + "step": 7457 + }, + { + "epoch": 1.72, + "learning_rate": 3.56366997920037e-05, + "loss": 0.0, + "step": 7458 + }, + { + "epoch": 1.72, + "learning_rate": 3.5634773900315845e-05, + "loss": 0.0, + "step": 7459 + }, + { + "epoch": 1.72, + "learning_rate": 3.563284800862799e-05, + "loss": 0.0, + "step": 7460 + }, + { + "epoch": 1.72, + "learning_rate": 3.563092211694015e-05, + "loss": 0.0007, + "step": 7461 + }, + { + "epoch": 1.72, + "learning_rate": 3.5628996225252295e-05, + "loss": 0.0001, + "step": 7462 + }, + { + "epoch": 1.72, + "learning_rate": 3.562707033356444e-05, + "loss": 0.0, + "step": 7463 + }, + { + "epoch": 1.72, + "learning_rate": 3.562514444187659e-05, + "loss": 0.0, + "step": 7464 + }, + { + "epoch": 1.73, + "learning_rate": 3.5623218550188744e-05, + "loss": 0.0, + "step": 7465 + }, + { + "epoch": 1.73, + "learning_rate": 3.5621292658500885e-05, + "loss": 0.0001, + "step": 7466 + }, + { + "epoch": 1.73, + "learning_rate": 3.561936676681303e-05, + "loss": 0.0039, + "step": 7467 + }, + { + "epoch": 1.73, + "learning_rate": 3.5617440875125186e-05, + "loss": 0.0001, + "step": 7468 + }, + { + "epoch": 1.73, + "learning_rate": 3.5615514983437334e-05, + "loss": 0.0, + "step": 7469 + }, + { + "epoch": 1.73, + "learning_rate": 3.561358909174948e-05, + "loss": 0.0007, + "step": 7470 + }, + { + "epoch": 1.73, + "learning_rate": 3.561166320006163e-05, + "loss": 0.0017, + "step": 7471 + }, + { + "epoch": 1.73, + "learning_rate": 3.560973730837378e-05, + "loss": 0.0, + "step": 7472 + }, + { + "epoch": 1.73, + "learning_rate": 3.5607811416685924e-05, + "loss": 0.0002, + "step": 7473 + }, + { + "epoch": 1.73, + "learning_rate": 3.560588552499808e-05, + "loss": 0.0001, + "step": 7474 + }, + { + "epoch": 1.73, + "learning_rate": 3.5603959633310225e-05, + "loss": 0.0002, + "step": 7475 + }, + { + "epoch": 1.73, + "learning_rate": 3.560203374162237e-05, + "loss": 0.0014, + "step": 7476 + }, + { + "epoch": 1.73, + "learning_rate": 3.560010784993452e-05, + "loss": 0.0001, + "step": 7477 + }, + { + "epoch": 1.73, + "learning_rate": 3.559818195824667e-05, + "loss": 0.0, + "step": 7478 + }, + { + "epoch": 1.73, + "learning_rate": 3.559625606655882e-05, + "loss": 0.0, + "step": 7479 + }, + { + "epoch": 1.73, + "learning_rate": 3.559433017487096e-05, + "loss": 0.0, + "step": 7480 + }, + { + "epoch": 1.73, + "learning_rate": 3.559240428318312e-05, + "loss": 0.0, + "step": 7481 + }, + { + "epoch": 1.73, + "learning_rate": 3.5590478391495264e-05, + "loss": 0.0, + "step": 7482 + }, + { + "epoch": 1.73, + "learning_rate": 3.558855249980741e-05, + "loss": 0.0001, + "step": 7483 + }, + { + "epoch": 1.73, + "learning_rate": 3.558662660811956e-05, + "loss": 0.0, + "step": 7484 + }, + { + "epoch": 1.73, + "learning_rate": 3.558470071643171e-05, + "loss": 0.0001, + "step": 7485 + }, + { + "epoch": 1.73, + "learning_rate": 3.558277482474386e-05, + "loss": 0.0061, + "step": 7486 + }, + { + "epoch": 1.73, + "learning_rate": 3.5580848933056e-05, + "loss": 0.0, + "step": 7487 + }, + { + "epoch": 1.73, + "learning_rate": 3.5578923041368156e-05, + "loss": 0.0001, + "step": 7488 + }, + { + "epoch": 1.73, + "learning_rate": 3.5576997149680304e-05, + "loss": 0.0, + "step": 7489 + }, + { + "epoch": 1.73, + "learning_rate": 3.557507125799245e-05, + "loss": 0.0, + "step": 7490 + }, + { + "epoch": 1.73, + "learning_rate": 3.55731453663046e-05, + "loss": 0.0001, + "step": 7491 + }, + { + "epoch": 1.73, + "learning_rate": 3.557121947461675e-05, + "loss": 0.0002, + "step": 7492 + }, + { + "epoch": 1.73, + "learning_rate": 3.55692935829289e-05, + "loss": 0.0, + "step": 7493 + }, + { + "epoch": 1.73, + "learning_rate": 3.556736769124104e-05, + "loss": 0.0001, + "step": 7494 + }, + { + "epoch": 1.73, + "learning_rate": 3.5565441799553195e-05, + "loss": 0.0, + "step": 7495 + }, + { + "epoch": 1.73, + "learning_rate": 3.556351590786534e-05, + "loss": 0.0001, + "step": 7496 + }, + { + "epoch": 1.73, + "learning_rate": 3.556159001617749e-05, + "loss": 0.0, + "step": 7497 + }, + { + "epoch": 1.73, + "learning_rate": 3.555966412448964e-05, + "loss": 0.0, + "step": 7498 + }, + { + "epoch": 1.73, + "learning_rate": 3.555773823280179e-05, + "loss": 0.0, + "step": 7499 + }, + { + "epoch": 1.73, + "learning_rate": 3.555581234111394e-05, + "loss": 0.0001, + "step": 7500 + }, + { + "epoch": 1.73, + "learning_rate": 3.555388644942609e-05, + "loss": 0.0, + "step": 7501 + }, + { + "epoch": 1.73, + "learning_rate": 3.5551960557738234e-05, + "loss": 0.0001, + "step": 7502 + }, + { + "epoch": 1.73, + "learning_rate": 3.555003466605039e-05, + "loss": 0.0001, + "step": 7503 + }, + { + "epoch": 1.73, + "learning_rate": 3.554810877436253e-05, + "loss": 0.0002, + "step": 7504 + }, + { + "epoch": 1.73, + "learning_rate": 3.554618288267468e-05, + "loss": 0.0, + "step": 7505 + }, + { + "epoch": 1.73, + "learning_rate": 3.554425699098683e-05, + "loss": 0.0, + "step": 7506 + }, + { + "epoch": 1.73, + "learning_rate": 3.554233109929898e-05, + "loss": 0.0001, + "step": 7507 + }, + { + "epoch": 1.74, + "learning_rate": 3.5540405207611126e-05, + "loss": 0.0, + "step": 7508 + }, + { + "epoch": 1.74, + "learning_rate": 3.5538479315923273e-05, + "loss": 0.0, + "step": 7509 + }, + { + "epoch": 1.74, + "learning_rate": 3.553655342423543e-05, + "loss": 0.0, + "step": 7510 + }, + { + "epoch": 1.74, + "learning_rate": 3.553462753254757e-05, + "loss": 0.0001, + "step": 7511 + }, + { + "epoch": 1.74, + "learning_rate": 3.553270164085972e-05, + "loss": 0.0, + "step": 7512 + }, + { + "epoch": 1.74, + "learning_rate": 3.553077574917187e-05, + "loss": 0.0001, + "step": 7513 + }, + { + "epoch": 1.74, + "learning_rate": 3.552884985748402e-05, + "loss": 0.0, + "step": 7514 + }, + { + "epoch": 1.74, + "learning_rate": 3.5526923965796165e-05, + "loss": 0.0, + "step": 7515 + }, + { + "epoch": 1.74, + "learning_rate": 3.552499807410831e-05, + "loss": 0.0, + "step": 7516 + }, + { + "epoch": 1.74, + "learning_rate": 3.552307218242047e-05, + "loss": 0.0003, + "step": 7517 + }, + { + "epoch": 1.74, + "learning_rate": 3.552114629073261e-05, + "loss": 0.0, + "step": 7518 + }, + { + "epoch": 1.74, + "learning_rate": 3.551922039904476e-05, + "loss": 0.0004, + "step": 7519 + }, + { + "epoch": 1.74, + "learning_rate": 3.551729450735691e-05, + "loss": 0.0, + "step": 7520 + }, + { + "epoch": 1.74, + "learning_rate": 3.551536861566906e-05, + "loss": 0.0, + "step": 7521 + }, + { + "epoch": 1.74, + "learning_rate": 3.5513442723981204e-05, + "loss": 0.0001, + "step": 7522 + }, + { + "epoch": 1.74, + "learning_rate": 3.551151683229335e-05, + "loss": 0.0001, + "step": 7523 + }, + { + "epoch": 1.74, + "learning_rate": 3.5509590940605506e-05, + "loss": 0.0002, + "step": 7524 + }, + { + "epoch": 1.74, + "learning_rate": 3.550766504891765e-05, + "loss": 0.0002, + "step": 7525 + }, + { + "epoch": 1.74, + "learning_rate": 3.55057391572298e-05, + "loss": 0.0003, + "step": 7526 + }, + { + "epoch": 1.74, + "learning_rate": 3.550381326554195e-05, + "loss": 0.0, + "step": 7527 + }, + { + "epoch": 1.74, + "learning_rate": 3.5501887373854096e-05, + "loss": 0.0001, + "step": 7528 + }, + { + "epoch": 1.74, + "learning_rate": 3.549996148216624e-05, + "loss": 0.0001, + "step": 7529 + }, + { + "epoch": 1.74, + "learning_rate": 3.54980355904784e-05, + "loss": 0.0, + "step": 7530 + }, + { + "epoch": 1.74, + "learning_rate": 3.5496109698790545e-05, + "loss": 0.0, + "step": 7531 + }, + { + "epoch": 1.74, + "learning_rate": 3.5494183807102686e-05, + "loss": 0.0003, + "step": 7532 + }, + { + "epoch": 1.74, + "learning_rate": 3.549225791541484e-05, + "loss": 0.0, + "step": 7533 + }, + { + "epoch": 1.74, + "learning_rate": 3.549033202372699e-05, + "loss": 0.0, + "step": 7534 + }, + { + "epoch": 1.74, + "learning_rate": 3.5488406132039135e-05, + "loss": 0.0, + "step": 7535 + }, + { + "epoch": 1.74, + "learning_rate": 3.548648024035128e-05, + "loss": 0.0003, + "step": 7536 + }, + { + "epoch": 1.74, + "learning_rate": 3.548455434866344e-05, + "loss": 0.0, + "step": 7537 + }, + { + "epoch": 1.74, + "learning_rate": 3.548262845697558e-05, + "loss": 0.0001, + "step": 7538 + }, + { + "epoch": 1.74, + "learning_rate": 3.548070256528773e-05, + "loss": 0.0002, + "step": 7539 + }, + { + "epoch": 1.74, + "learning_rate": 3.547877667359988e-05, + "loss": 0.0, + "step": 7540 + }, + { + "epoch": 1.74, + "learning_rate": 3.547685078191203e-05, + "loss": 0.0, + "step": 7541 + }, + { + "epoch": 1.74, + "learning_rate": 3.5474924890224174e-05, + "loss": 0.0001, + "step": 7542 + }, + { + "epoch": 1.74, + "learning_rate": 3.547299899853632e-05, + "loss": 0.0007, + "step": 7543 + }, + { + "epoch": 1.74, + "learning_rate": 3.5471073106848476e-05, + "loss": 0.0, + "step": 7544 + }, + { + "epoch": 1.74, + "learning_rate": 3.5469147215160617e-05, + "loss": 0.0001, + "step": 7545 + }, + { + "epoch": 1.74, + "learning_rate": 3.546722132347277e-05, + "loss": 0.0, + "step": 7546 + }, + { + "epoch": 1.74, + "learning_rate": 3.546529543178492e-05, + "loss": 0.0, + "step": 7547 + }, + { + "epoch": 1.74, + "learning_rate": 3.5463369540097066e-05, + "loss": 0.0001, + "step": 7548 + }, + { + "epoch": 1.74, + "learning_rate": 3.546144364840921e-05, + "loss": 0.0, + "step": 7549 + }, + { + "epoch": 1.74, + "learning_rate": 3.545951775672136e-05, + "loss": 0.0, + "step": 7550 + }, + { + "epoch": 1.75, + "learning_rate": 3.5457591865033515e-05, + "loss": 0.0, + "step": 7551 + }, + { + "epoch": 1.75, + "learning_rate": 3.5455665973345656e-05, + "loss": 0.0, + "step": 7552 + }, + { + "epoch": 1.75, + "learning_rate": 3.545374008165781e-05, + "loss": 0.0001, + "step": 7553 + }, + { + "epoch": 1.75, + "learning_rate": 3.545181418996996e-05, + "loss": 0.0, + "step": 7554 + }, + { + "epoch": 1.75, + "learning_rate": 3.5449888298282105e-05, + "loss": 0.0, + "step": 7555 + }, + { + "epoch": 1.75, + "learning_rate": 3.544796240659425e-05, + "loss": 0.0, + "step": 7556 + }, + { + "epoch": 1.75, + "learning_rate": 3.544603651490641e-05, + "loss": 0.0005, + "step": 7557 + }, + { + "epoch": 1.75, + "learning_rate": 3.5444110623218554e-05, + "loss": 0.0, + "step": 7558 + }, + { + "epoch": 1.75, + "learning_rate": 3.54421847315307e-05, + "loss": 0.0002, + "step": 7559 + }, + { + "epoch": 1.75, + "learning_rate": 3.544025883984285e-05, + "loss": 0.0, + "step": 7560 + }, + { + "epoch": 1.75, + "learning_rate": 3.5438332948154997e-05, + "loss": 0.0, + "step": 7561 + }, + { + "epoch": 1.75, + "learning_rate": 3.5436407056467144e-05, + "loss": 0.0001, + "step": 7562 + }, + { + "epoch": 1.75, + "learning_rate": 3.543448116477929e-05, + "loss": 0.0074, + "step": 7563 + }, + { + "epoch": 1.75, + "learning_rate": 3.5432555273091446e-05, + "loss": 0.0017, + "step": 7564 + }, + { + "epoch": 1.75, + "learning_rate": 3.543062938140359e-05, + "loss": 0.0001, + "step": 7565 + }, + { + "epoch": 1.75, + "learning_rate": 3.542870348971574e-05, + "loss": 0.0, + "step": 7566 + }, + { + "epoch": 1.75, + "learning_rate": 3.542677759802789e-05, + "loss": 0.0007, + "step": 7567 + }, + { + "epoch": 1.75, + "learning_rate": 3.542485170634004e-05, + "loss": 0.0, + "step": 7568 + }, + { + "epoch": 1.75, + "learning_rate": 3.542292581465218e-05, + "loss": 0.0001, + "step": 7569 + }, + { + "epoch": 1.75, + "learning_rate": 3.542099992296433e-05, + "loss": 0.0, + "step": 7570 + }, + { + "epoch": 1.75, + "learning_rate": 3.5419074031276485e-05, + "loss": 0.0001, + "step": 7571 + }, + { + "epoch": 1.75, + "learning_rate": 3.541714813958863e-05, + "loss": 0.0, + "step": 7572 + }, + { + "epoch": 1.75, + "learning_rate": 3.541522224790078e-05, + "loss": 0.0001, + "step": 7573 + }, + { + "epoch": 1.75, + "learning_rate": 3.541329635621293e-05, + "loss": 0.0, + "step": 7574 + }, + { + "epoch": 1.75, + "learning_rate": 3.541137046452508e-05, + "loss": 0.0, + "step": 7575 + }, + { + "epoch": 1.75, + "learning_rate": 3.540944457283722e-05, + "loss": 0.0021, + "step": 7576 + }, + { + "epoch": 1.75, + "learning_rate": 3.5407518681149377e-05, + "loss": 0.0, + "step": 7577 + }, + { + "epoch": 1.75, + "learning_rate": 3.5405592789461524e-05, + "loss": 0.0001, + "step": 7578 + }, + { + "epoch": 1.75, + "learning_rate": 3.540366689777367e-05, + "loss": 0.0, + "step": 7579 + }, + { + "epoch": 1.75, + "learning_rate": 3.540174100608582e-05, + "loss": 0.0, + "step": 7580 + }, + { + "epoch": 1.75, + "learning_rate": 3.5399815114397966e-05, + "loss": 0.0, + "step": 7581 + }, + { + "epoch": 1.75, + "learning_rate": 3.539788922271012e-05, + "loss": 0.0, + "step": 7582 + }, + { + "epoch": 1.75, + "learning_rate": 3.539596333102226e-05, + "loss": 0.0, + "step": 7583 + }, + { + "epoch": 1.75, + "learning_rate": 3.5394037439334416e-05, + "loss": 0.0004, + "step": 7584 + }, + { + "epoch": 1.75, + "learning_rate": 3.539211154764656e-05, + "loss": 0.0, + "step": 7585 + }, + { + "epoch": 1.75, + "learning_rate": 3.539018565595871e-05, + "loss": 0.0002, + "step": 7586 + }, + { + "epoch": 1.75, + "learning_rate": 3.538825976427086e-05, + "loss": 0.0, + "step": 7587 + }, + { + "epoch": 1.75, + "learning_rate": 3.5386333872583006e-05, + "loss": 0.0001, + "step": 7588 + }, + { + "epoch": 1.75, + "learning_rate": 3.538440798089516e-05, + "loss": 0.0001, + "step": 7589 + }, + { + "epoch": 1.75, + "learning_rate": 3.53824820892073e-05, + "loss": 0.001, + "step": 7590 + }, + { + "epoch": 1.75, + "learning_rate": 3.5380556197519455e-05, + "loss": 0.0031, + "step": 7591 + }, + { + "epoch": 1.75, + "learning_rate": 3.53786303058316e-05, + "loss": 0.0, + "step": 7592 + }, + { + "epoch": 1.75, + "learning_rate": 3.537670441414375e-05, + "loss": 0.0003, + "step": 7593 + }, + { + "epoch": 1.76, + "learning_rate": 3.53747785224559e-05, + "loss": 0.0, + "step": 7594 + }, + { + "epoch": 1.76, + "learning_rate": 3.537285263076805e-05, + "loss": 0.0, + "step": 7595 + }, + { + "epoch": 1.76, + "learning_rate": 3.53709267390802e-05, + "loss": 0.0, + "step": 7596 + }, + { + "epoch": 1.76, + "learning_rate": 3.536900084739234e-05, + "loss": 0.0008, + "step": 7597 + }, + { + "epoch": 1.76, + "learning_rate": 3.5367074955704494e-05, + "loss": 0.0001, + "step": 7598 + }, + { + "epoch": 1.76, + "learning_rate": 3.536514906401664e-05, + "loss": 0.0005, + "step": 7599 + }, + { + "epoch": 1.76, + "learning_rate": 3.536322317232879e-05, + "loss": 0.0003, + "step": 7600 + }, + { + "epoch": 1.76, + "learning_rate": 3.5361297280640936e-05, + "loss": 0.0, + "step": 7601 + }, + { + "epoch": 1.76, + "learning_rate": 3.535937138895309e-05, + "loss": 0.0001, + "step": 7602 + }, + { + "epoch": 1.76, + "learning_rate": 3.535744549726524e-05, + "loss": 0.0002, + "step": 7603 + }, + { + "epoch": 1.76, + "learning_rate": 3.5355519605577386e-05, + "loss": 0.0, + "step": 7604 + }, + { + "epoch": 1.76, + "learning_rate": 3.535359371388953e-05, + "loss": 0.0002, + "step": 7605 + }, + { + "epoch": 1.76, + "learning_rate": 3.535166782220168e-05, + "loss": 0.0, + "step": 7606 + }, + { + "epoch": 1.76, + "learning_rate": 3.534974193051383e-05, + "loss": 0.0002, + "step": 7607 + }, + { + "epoch": 1.76, + "learning_rate": 3.5347816038825975e-05, + "loss": 0.0001, + "step": 7608 + }, + { + "epoch": 1.76, + "learning_rate": 3.534589014713813e-05, + "loss": 0.0, + "step": 7609 + }, + { + "epoch": 1.76, + "learning_rate": 3.534396425545027e-05, + "loss": 0.0, + "step": 7610 + }, + { + "epoch": 1.76, + "learning_rate": 3.5342038363762425e-05, + "loss": 0.0, + "step": 7611 + }, + { + "epoch": 1.76, + "learning_rate": 3.534011247207457e-05, + "loss": 0.0002, + "step": 7612 + }, + { + "epoch": 1.76, + "learning_rate": 3.533818658038672e-05, + "loss": 0.0, + "step": 7613 + }, + { + "epoch": 1.76, + "learning_rate": 3.533626068869887e-05, + "loss": 0.0337, + "step": 7614 + }, + { + "epoch": 1.76, + "learning_rate": 3.533433479701102e-05, + "loss": 0.0, + "step": 7615 + }, + { + "epoch": 1.76, + "learning_rate": 3.533240890532317e-05, + "loss": 0.0, + "step": 7616 + }, + { + "epoch": 1.76, + "learning_rate": 3.533048301363531e-05, + "loss": 0.0, + "step": 7617 + }, + { + "epoch": 1.76, + "learning_rate": 3.5328557121947464e-05, + "loss": 0.0028, + "step": 7618 + }, + { + "epoch": 1.76, + "learning_rate": 3.532663123025961e-05, + "loss": 0.0, + "step": 7619 + }, + { + "epoch": 1.76, + "learning_rate": 3.532470533857176e-05, + "loss": 0.0, + "step": 7620 + }, + { + "epoch": 1.76, + "learning_rate": 3.5322779446883906e-05, + "loss": 0.0, + "step": 7621 + }, + { + "epoch": 1.76, + "learning_rate": 3.532085355519606e-05, + "loss": 0.0, + "step": 7622 + }, + { + "epoch": 1.76, + "learning_rate": 3.531892766350821e-05, + "loss": 0.0, + "step": 7623 + }, + { + "epoch": 1.76, + "learning_rate": 3.5317001771820355e-05, + "loss": 0.0004, + "step": 7624 + }, + { + "epoch": 1.76, + "learning_rate": 3.53150758801325e-05, + "loss": 0.0001, + "step": 7625 + }, + { + "epoch": 1.76, + "learning_rate": 3.531314998844465e-05, + "loss": 0.0002, + "step": 7626 + }, + { + "epoch": 1.76, + "learning_rate": 3.53112240967568e-05, + "loss": 0.0, + "step": 7627 + }, + { + "epoch": 1.76, + "learning_rate": 3.5309298205068945e-05, + "loss": 0.0, + "step": 7628 + }, + { + "epoch": 1.76, + "learning_rate": 3.53073723133811e-05, + "loss": 0.0, + "step": 7629 + }, + { + "epoch": 1.76, + "learning_rate": 3.530544642169325e-05, + "loss": 0.0, + "step": 7630 + }, + { + "epoch": 1.76, + "learning_rate": 3.5303520530005395e-05, + "loss": 0.0011, + "step": 7631 + }, + { + "epoch": 1.76, + "learning_rate": 3.530159463831754e-05, + "loss": 0.0001, + "step": 7632 + }, + { + "epoch": 1.76, + "learning_rate": 3.5299668746629696e-05, + "loss": 0.0, + "step": 7633 + }, + { + "epoch": 1.76, + "learning_rate": 3.529774285494184e-05, + "loss": 0.0, + "step": 7634 + }, + { + "epoch": 1.76, + "learning_rate": 3.5295816963253984e-05, + "loss": 0.0, + "step": 7635 + }, + { + "epoch": 1.76, + "learning_rate": 3.529389107156614e-05, + "loss": 0.0, + "step": 7636 + }, + { + "epoch": 1.76, + "learning_rate": 3.5291965179878286e-05, + "loss": 0.0, + "step": 7637 + }, + { + "epoch": 1.77, + "learning_rate": 3.5290039288190434e-05, + "loss": 0.0, + "step": 7638 + }, + { + "epoch": 1.77, + "learning_rate": 3.528811339650258e-05, + "loss": 0.0001, + "step": 7639 + }, + { + "epoch": 1.77, + "learning_rate": 3.5286187504814735e-05, + "loss": 0.0, + "step": 7640 + }, + { + "epoch": 1.77, + "learning_rate": 3.5284261613126876e-05, + "loss": 0.0, + "step": 7641 + }, + { + "epoch": 1.77, + "learning_rate": 3.528233572143903e-05, + "loss": 0.0, + "step": 7642 + }, + { + "epoch": 1.77, + "learning_rate": 3.528040982975118e-05, + "loss": 0.0001, + "step": 7643 + }, + { + "epoch": 1.77, + "learning_rate": 3.5278483938063325e-05, + "loss": 0.0, + "step": 7644 + }, + { + "epoch": 1.77, + "learning_rate": 3.527655804637547e-05, + "loss": 0.0, + "step": 7645 + }, + { + "epoch": 1.77, + "learning_rate": 3.527463215468762e-05, + "loss": 0.0, + "step": 7646 + }, + { + "epoch": 1.77, + "learning_rate": 3.5272706262999774e-05, + "loss": 0.0004, + "step": 7647 + }, + { + "epoch": 1.77, + "learning_rate": 3.5270780371311915e-05, + "loss": 0.0, + "step": 7648 + }, + { + "epoch": 1.77, + "learning_rate": 3.526885447962407e-05, + "loss": 0.0, + "step": 7649 + }, + { + "epoch": 1.77, + "learning_rate": 3.526692858793622e-05, + "loss": 0.0, + "step": 7650 + }, + { + "epoch": 1.77, + "learning_rate": 3.5265002696248364e-05, + "loss": 0.0001, + "step": 7651 + }, + { + "epoch": 1.77, + "learning_rate": 3.526307680456051e-05, + "loss": 0.0002, + "step": 7652 + }, + { + "epoch": 1.77, + "learning_rate": 3.526115091287266e-05, + "loss": 0.0005, + "step": 7653 + }, + { + "epoch": 1.77, + "learning_rate": 3.5259225021184814e-05, + "loss": 0.0, + "step": 7654 + }, + { + "epoch": 1.77, + "learning_rate": 3.5257299129496954e-05, + "loss": 0.0, + "step": 7655 + }, + { + "epoch": 1.77, + "learning_rate": 3.525537323780911e-05, + "loss": 0.0001, + "step": 7656 + }, + { + "epoch": 1.77, + "learning_rate": 3.5253447346121256e-05, + "loss": 0.0, + "step": 7657 + }, + { + "epoch": 1.77, + "learning_rate": 3.5251521454433404e-05, + "loss": 0.001, + "step": 7658 + }, + { + "epoch": 1.77, + "learning_rate": 3.524959556274555e-05, + "loss": 0.0005, + "step": 7659 + }, + { + "epoch": 1.77, + "learning_rate": 3.5247669671057705e-05, + "loss": 0.0002, + "step": 7660 + }, + { + "epoch": 1.77, + "learning_rate": 3.524574377936985e-05, + "loss": 0.0, + "step": 7661 + }, + { + "epoch": 1.77, + "learning_rate": 3.5243817887681993e-05, + "loss": 0.0001, + "step": 7662 + }, + { + "epoch": 1.77, + "learning_rate": 3.524189199599415e-05, + "loss": 0.0, + "step": 7663 + }, + { + "epoch": 1.77, + "learning_rate": 3.5239966104306295e-05, + "loss": 0.0, + "step": 7664 + }, + { + "epoch": 1.77, + "learning_rate": 3.523804021261844e-05, + "loss": 0.0, + "step": 7665 + }, + { + "epoch": 1.77, + "learning_rate": 3.523611432093059e-05, + "loss": 0.0001, + "step": 7666 + }, + { + "epoch": 1.77, + "learning_rate": 3.5234188429242744e-05, + "loss": 0.0, + "step": 7667 + }, + { + "epoch": 1.77, + "learning_rate": 3.523226253755489e-05, + "loss": 0.0002, + "step": 7668 + }, + { + "epoch": 1.77, + "learning_rate": 3.523033664586704e-05, + "loss": 0.0002, + "step": 7669 + }, + { + "epoch": 1.77, + "learning_rate": 3.522841075417919e-05, + "loss": 0.0, + "step": 7670 + }, + { + "epoch": 1.77, + "learning_rate": 3.522648486249134e-05, + "loss": 0.0, + "step": 7671 + }, + { + "epoch": 1.77, + "learning_rate": 3.522455897080348e-05, + "loss": 0.0, + "step": 7672 + }, + { + "epoch": 1.77, + "learning_rate": 3.522263307911563e-05, + "loss": 0.0, + "step": 7673 + }, + { + "epoch": 1.77, + "learning_rate": 3.5220707187427783e-05, + "loss": 0.0, + "step": 7674 + }, + { + "epoch": 1.77, + "learning_rate": 3.521878129573993e-05, + "loss": 0.0001, + "step": 7675 + }, + { + "epoch": 1.77, + "learning_rate": 3.521685540405208e-05, + "loss": 0.0, + "step": 7676 + }, + { + "epoch": 1.77, + "learning_rate": 3.5214929512364226e-05, + "loss": 0.0005, + "step": 7677 + }, + { + "epoch": 1.77, + "learning_rate": 3.521300362067638e-05, + "loss": 0.0, + "step": 7678 + }, + { + "epoch": 1.77, + "learning_rate": 3.521107772898852e-05, + "loss": 0.0, + "step": 7679 + }, + { + "epoch": 1.77, + "learning_rate": 3.5209151837300675e-05, + "loss": 0.0003, + "step": 7680 + }, + { + "epoch": 1.78, + "learning_rate": 3.520722594561282e-05, + "loss": 0.0, + "step": 7681 + }, + { + "epoch": 1.78, + "learning_rate": 3.520530005392496e-05, + "loss": 0.0021, + "step": 7682 + }, + { + "epoch": 1.78, + "learning_rate": 3.520337416223712e-05, + "loss": 0.0001, + "step": 7683 + }, + { + "epoch": 1.78, + "learning_rate": 3.5201448270549265e-05, + "loss": 0.0, + "step": 7684 + }, + { + "epoch": 1.78, + "learning_rate": 3.519952237886141e-05, + "loss": 0.0, + "step": 7685 + }, + { + "epoch": 1.78, + "learning_rate": 3.519759648717356e-05, + "loss": 0.0002, + "step": 7686 + }, + { + "epoch": 1.78, + "learning_rate": 3.5195670595485714e-05, + "loss": 0.0004, + "step": 7687 + }, + { + "epoch": 1.78, + "learning_rate": 3.519374470379786e-05, + "loss": 0.002, + "step": 7688 + }, + { + "epoch": 1.78, + "learning_rate": 3.519181881211001e-05, + "loss": 0.0, + "step": 7689 + }, + { + "epoch": 1.78, + "learning_rate": 3.518989292042216e-05, + "loss": 0.0, + "step": 7690 + }, + { + "epoch": 1.78, + "learning_rate": 3.5187967028734304e-05, + "loss": 0.0, + "step": 7691 + }, + { + "epoch": 1.78, + "learning_rate": 3.518604113704645e-05, + "loss": 0.0001, + "step": 7692 + }, + { + "epoch": 1.78, + "learning_rate": 3.51841152453586e-05, + "loss": 0.0002, + "step": 7693 + }, + { + "epoch": 1.78, + "learning_rate": 3.518218935367075e-05, + "loss": 0.0, + "step": 7694 + }, + { + "epoch": 1.78, + "learning_rate": 3.51802634619829e-05, + "loss": 0.0, + "step": 7695 + }, + { + "epoch": 1.78, + "learning_rate": 3.517833757029505e-05, + "loss": 0.0002, + "step": 7696 + }, + { + "epoch": 1.78, + "learning_rate": 3.5176411678607196e-05, + "loss": 0.0, + "step": 7697 + }, + { + "epoch": 1.78, + "learning_rate": 3.517448578691935e-05, + "loss": 0.0, + "step": 7698 + }, + { + "epoch": 1.78, + "learning_rate": 3.517255989523149e-05, + "loss": 0.0001, + "step": 7699 + }, + { + "epoch": 1.78, + "learning_rate": 3.517063400354364e-05, + "loss": 0.0, + "step": 7700 + }, + { + "epoch": 1.78, + "learning_rate": 3.516870811185579e-05, + "loss": 0.0, + "step": 7701 + }, + { + "epoch": 1.78, + "learning_rate": 3.516678222016794e-05, + "loss": 0.0, + "step": 7702 + }, + { + "epoch": 1.78, + "learning_rate": 3.516485632848009e-05, + "loss": 0.0, + "step": 7703 + }, + { + "epoch": 1.78, + "learning_rate": 3.5162930436792235e-05, + "loss": 0.0, + "step": 7704 + }, + { + "epoch": 1.78, + "learning_rate": 3.516100454510439e-05, + "loss": 0.0002, + "step": 7705 + }, + { + "epoch": 1.78, + "learning_rate": 3.515907865341653e-05, + "loss": 0.0005, + "step": 7706 + }, + { + "epoch": 1.78, + "learning_rate": 3.5157152761728684e-05, + "loss": 0.0, + "step": 7707 + }, + { + "epoch": 1.78, + "learning_rate": 3.515522687004083e-05, + "loss": 0.001, + "step": 7708 + }, + { + "epoch": 1.78, + "learning_rate": 3.515330097835298e-05, + "loss": 0.0002, + "step": 7709 + }, + { + "epoch": 1.78, + "learning_rate": 3.5151375086665127e-05, + "loss": 0.0, + "step": 7710 + }, + { + "epoch": 1.78, + "learning_rate": 3.5149449194977274e-05, + "loss": 0.0001, + "step": 7711 + }, + { + "epoch": 1.78, + "learning_rate": 3.514752330328943e-05, + "loss": 0.0, + "step": 7712 + }, + { + "epoch": 1.78, + "learning_rate": 3.514559741160157e-05, + "loss": 0.0006, + "step": 7713 + }, + { + "epoch": 1.78, + "learning_rate": 3.514367151991372e-05, + "loss": 0.0, + "step": 7714 + }, + { + "epoch": 1.78, + "learning_rate": 3.514174562822587e-05, + "loss": 0.0, + "step": 7715 + }, + { + "epoch": 1.78, + "learning_rate": 3.513981973653802e-05, + "loss": 0.0, + "step": 7716 + }, + { + "epoch": 1.78, + "learning_rate": 3.5137893844850166e-05, + "loss": 0.0032, + "step": 7717 + }, + { + "epoch": 1.78, + "learning_rate": 3.513596795316231e-05, + "loss": 0.0, + "step": 7718 + }, + { + "epoch": 1.78, + "learning_rate": 3.513404206147447e-05, + "loss": 0.0002, + "step": 7719 + }, + { + "epoch": 1.78, + "learning_rate": 3.513211616978661e-05, + "loss": 0.0, + "step": 7720 + }, + { + "epoch": 1.78, + "learning_rate": 3.513019027809876e-05, + "loss": 0.0012, + "step": 7721 + }, + { + "epoch": 1.78, + "learning_rate": 3.512826438641091e-05, + "loss": 0.0002, + "step": 7722 + }, + { + "epoch": 1.78, + "learning_rate": 3.512633849472306e-05, + "loss": 0.0, + "step": 7723 + }, + { + "epoch": 1.79, + "learning_rate": 3.5124412603035205e-05, + "loss": 0.0, + "step": 7724 + }, + { + "epoch": 1.79, + "learning_rate": 3.512248671134736e-05, + "loss": 0.0, + "step": 7725 + }, + { + "epoch": 1.79, + "learning_rate": 3.5120560819659507e-05, + "loss": 0.0, + "step": 7726 + }, + { + "epoch": 1.79, + "learning_rate": 3.5118634927971654e-05, + "loss": 0.0, + "step": 7727 + }, + { + "epoch": 1.79, + "learning_rate": 3.51167090362838e-05, + "loss": 0.0, + "step": 7728 + }, + { + "epoch": 1.79, + "learning_rate": 3.511478314459595e-05, + "loss": 0.0, + "step": 7729 + }, + { + "epoch": 1.79, + "learning_rate": 3.5112857252908096e-05, + "loss": 0.0, + "step": 7730 + }, + { + "epoch": 1.79, + "learning_rate": 3.5110931361220244e-05, + "loss": 0.0001, + "step": 7731 + }, + { + "epoch": 1.79, + "learning_rate": 3.51090054695324e-05, + "loss": 0.0, + "step": 7732 + }, + { + "epoch": 1.79, + "learning_rate": 3.5107079577844546e-05, + "loss": 0.0002, + "step": 7733 + }, + { + "epoch": 1.79, + "learning_rate": 3.510515368615669e-05, + "loss": 0.0002, + "step": 7734 + }, + { + "epoch": 1.79, + "learning_rate": 3.510322779446884e-05, + "loss": 0.0001, + "step": 7735 + }, + { + "epoch": 1.79, + "learning_rate": 3.5101301902780995e-05, + "loss": 0.0056, + "step": 7736 + }, + { + "epoch": 1.79, + "learning_rate": 3.5099376011093136e-05, + "loss": 0.0, + "step": 7737 + }, + { + "epoch": 1.79, + "learning_rate": 3.509745011940528e-05, + "loss": 0.0001, + "step": 7738 + }, + { + "epoch": 1.79, + "learning_rate": 3.509552422771744e-05, + "loss": 0.0, + "step": 7739 + }, + { + "epoch": 1.79, + "learning_rate": 3.5093598336029585e-05, + "loss": 0.0001, + "step": 7740 + }, + { + "epoch": 1.79, + "learning_rate": 3.509167244434173e-05, + "loss": 0.0002, + "step": 7741 + }, + { + "epoch": 1.79, + "learning_rate": 3.508974655265388e-05, + "loss": 0.0001, + "step": 7742 + }, + { + "epoch": 1.79, + "learning_rate": 3.5087820660966034e-05, + "loss": 0.0015, + "step": 7743 + }, + { + "epoch": 1.79, + "learning_rate": 3.5085894769278175e-05, + "loss": 0.0, + "step": 7744 + }, + { + "epoch": 1.79, + "learning_rate": 3.508396887759033e-05, + "loss": 0.0, + "step": 7745 + }, + { + "epoch": 1.79, + "learning_rate": 3.5082042985902476e-05, + "loss": 0.0001, + "step": 7746 + }, + { + "epoch": 1.79, + "learning_rate": 3.5080117094214624e-05, + "loss": 0.0, + "step": 7747 + }, + { + "epoch": 1.79, + "learning_rate": 3.507819120252677e-05, + "loss": 0.0, + "step": 7748 + }, + { + "epoch": 1.79, + "learning_rate": 3.507626531083892e-05, + "loss": 0.0002, + "step": 7749 + }, + { + "epoch": 1.79, + "learning_rate": 3.507433941915107e-05, + "loss": 0.0, + "step": 7750 + }, + { + "epoch": 1.79, + "learning_rate": 3.5072413527463214e-05, + "loss": 0.0, + "step": 7751 + }, + { + "epoch": 1.79, + "learning_rate": 3.507048763577537e-05, + "loss": 0.0, + "step": 7752 + }, + { + "epoch": 1.79, + "learning_rate": 3.5068561744087516e-05, + "loss": 0.0001, + "step": 7753 + }, + { + "epoch": 1.79, + "learning_rate": 3.506663585239966e-05, + "loss": 0.0, + "step": 7754 + }, + { + "epoch": 1.79, + "learning_rate": 3.506470996071181e-05, + "loss": 0.0001, + "step": 7755 + }, + { + "epoch": 1.79, + "learning_rate": 3.506278406902396e-05, + "loss": 0.0, + "step": 7756 + }, + { + "epoch": 1.79, + "learning_rate": 3.506085817733611e-05, + "loss": 0.0, + "step": 7757 + }, + { + "epoch": 1.79, + "learning_rate": 3.505893228564825e-05, + "loss": 0.0, + "step": 7758 + }, + { + "epoch": 1.79, + "learning_rate": 3.505700639396041e-05, + "loss": 0.0001, + "step": 7759 + }, + { + "epoch": 1.79, + "learning_rate": 3.5055080502272555e-05, + "loss": 0.0, + "step": 7760 + }, + { + "epoch": 1.79, + "learning_rate": 3.50531546105847e-05, + "loss": 0.0, + "step": 7761 + }, + { + "epoch": 1.79, + "learning_rate": 3.505122871889685e-05, + "loss": 0.0, + "step": 7762 + }, + { + "epoch": 1.79, + "learning_rate": 3.5049302827209004e-05, + "loss": 0.0003, + "step": 7763 + }, + { + "epoch": 1.79, + "learning_rate": 3.5047376935521145e-05, + "loss": 0.0, + "step": 7764 + }, + { + "epoch": 1.79, + "learning_rate": 3.504545104383329e-05, + "loss": 0.013, + "step": 7765 + }, + { + "epoch": 1.79, + "learning_rate": 3.5043525152145446e-05, + "loss": 0.0001, + "step": 7766 + }, + { + "epoch": 1.8, + "learning_rate": 3.5041599260457594e-05, + "loss": 0.0, + "step": 7767 + }, + { + "epoch": 1.8, + "learning_rate": 3.503967336876974e-05, + "loss": 0.0, + "step": 7768 + }, + { + "epoch": 1.8, + "learning_rate": 3.503774747708189e-05, + "loss": 0.0003, + "step": 7769 + }, + { + "epoch": 1.8, + "learning_rate": 3.503582158539404e-05, + "loss": 0.0, + "step": 7770 + }, + { + "epoch": 1.8, + "learning_rate": 3.5033895693706184e-05, + "loss": 0.0, + "step": 7771 + }, + { + "epoch": 1.8, + "learning_rate": 3.503196980201834e-05, + "loss": 0.0001, + "step": 7772 + }, + { + "epoch": 1.8, + "learning_rate": 3.5030043910330485e-05, + "loss": 0.0001, + "step": 7773 + }, + { + "epoch": 1.8, + "learning_rate": 3.502811801864263e-05, + "loss": 0.0, + "step": 7774 + }, + { + "epoch": 1.8, + "learning_rate": 3.502619212695478e-05, + "loss": 0.0, + "step": 7775 + }, + { + "epoch": 1.8, + "learning_rate": 3.502426623526693e-05, + "loss": 0.0, + "step": 7776 + }, + { + "epoch": 1.8, + "learning_rate": 3.502234034357908e-05, + "loss": 0.0, + "step": 7777 + }, + { + "epoch": 1.8, + "learning_rate": 3.502041445189122e-05, + "loss": 0.0001, + "step": 7778 + }, + { + "epoch": 1.8, + "learning_rate": 3.501848856020338e-05, + "loss": 0.0001, + "step": 7779 + }, + { + "epoch": 1.8, + "learning_rate": 3.5016562668515525e-05, + "loss": 0.0, + "step": 7780 + }, + { + "epoch": 1.8, + "learning_rate": 3.501463677682767e-05, + "loss": 0.0256, + "step": 7781 + }, + { + "epoch": 1.8, + "learning_rate": 3.501271088513982e-05, + "loss": 0.0001, + "step": 7782 + }, + { + "epoch": 1.8, + "learning_rate": 3.5010784993451974e-05, + "loss": 0.0, + "step": 7783 + }, + { + "epoch": 1.8, + "learning_rate": 3.500885910176412e-05, + "loss": 0.0001, + "step": 7784 + }, + { + "epoch": 1.8, + "learning_rate": 3.500693321007626e-05, + "loss": 0.0, + "step": 7785 + }, + { + "epoch": 1.8, + "learning_rate": 3.5005007318388416e-05, + "loss": 0.0, + "step": 7786 + }, + { + "epoch": 1.8, + "learning_rate": 3.5003081426700564e-05, + "loss": 0.0, + "step": 7787 + }, + { + "epoch": 1.8, + "learning_rate": 3.500115553501271e-05, + "loss": 0.0, + "step": 7788 + }, + { + "epoch": 1.8, + "learning_rate": 3.499922964332486e-05, + "loss": 0.0, + "step": 7789 + }, + { + "epoch": 1.8, + "learning_rate": 3.499730375163701e-05, + "loss": 0.0, + "step": 7790 + }, + { + "epoch": 1.8, + "learning_rate": 3.499537785994916e-05, + "loss": 0.0013, + "step": 7791 + }, + { + "epoch": 1.8, + "learning_rate": 3.499345196826131e-05, + "loss": 0.0, + "step": 7792 + }, + { + "epoch": 1.8, + "learning_rate": 3.4991526076573455e-05, + "loss": 0.0, + "step": 7793 + }, + { + "epoch": 1.8, + "learning_rate": 3.49896001848856e-05, + "loss": 0.0, + "step": 7794 + }, + { + "epoch": 1.8, + "learning_rate": 3.498767429319775e-05, + "loss": 0.0294, + "step": 7795 + }, + { + "epoch": 1.8, + "learning_rate": 3.49857484015099e-05, + "loss": 0.0, + "step": 7796 + }, + { + "epoch": 1.8, + "learning_rate": 3.498382250982205e-05, + "loss": 0.0, + "step": 7797 + }, + { + "epoch": 1.8, + "learning_rate": 3.49818966181342e-05, + "loss": 0.0, + "step": 7798 + }, + { + "epoch": 1.8, + "learning_rate": 3.497997072644635e-05, + "loss": 0.0, + "step": 7799 + }, + { + "epoch": 1.8, + "learning_rate": 3.4978044834758494e-05, + "loss": 0.0, + "step": 7800 + }, + { + "epoch": 1.8, + "learning_rate": 3.497611894307065e-05, + "loss": 0.0019, + "step": 7801 + }, + { + "epoch": 1.8, + "learning_rate": 3.497419305138279e-05, + "loss": 0.0, + "step": 7802 + }, + { + "epoch": 1.8, + "learning_rate": 3.497226715969494e-05, + "loss": 0.0002, + "step": 7803 + }, + { + "epoch": 1.8, + "learning_rate": 3.497034126800709e-05, + "loss": 0.0001, + "step": 7804 + }, + { + "epoch": 1.8, + "learning_rate": 3.496841537631924e-05, + "loss": 0.0001, + "step": 7805 + }, + { + "epoch": 1.8, + "learning_rate": 3.4966489484631386e-05, + "loss": 0.0002, + "step": 7806 + }, + { + "epoch": 1.8, + "learning_rate": 3.4964563592943534e-05, + "loss": 0.0, + "step": 7807 + }, + { + "epoch": 1.8, + "learning_rate": 3.496263770125569e-05, + "loss": 0.0001, + "step": 7808 + }, + { + "epoch": 1.8, + "learning_rate": 3.496071180956783e-05, + "loss": 0.0001, + "step": 7809 + }, + { + "epoch": 1.8, + "learning_rate": 3.495878591787998e-05, + "loss": 0.0002, + "step": 7810 + }, + { + "epoch": 1.81, + "learning_rate": 3.495686002619213e-05, + "loss": 0.0, + "step": 7811 + }, + { + "epoch": 1.81, + "learning_rate": 3.495493413450428e-05, + "loss": 0.0, + "step": 7812 + }, + { + "epoch": 1.81, + "learning_rate": 3.4953008242816425e-05, + "loss": 0.0002, + "step": 7813 + }, + { + "epoch": 1.81, + "learning_rate": 3.495108235112857e-05, + "loss": 0.0, + "step": 7814 + }, + { + "epoch": 1.81, + "learning_rate": 3.494915645944073e-05, + "loss": 0.0004, + "step": 7815 + }, + { + "epoch": 1.81, + "learning_rate": 3.494723056775287e-05, + "loss": 0.0, + "step": 7816 + }, + { + "epoch": 1.81, + "learning_rate": 3.494530467606502e-05, + "loss": 0.0, + "step": 7817 + }, + { + "epoch": 1.81, + "learning_rate": 3.494337878437717e-05, + "loss": 0.0007, + "step": 7818 + }, + { + "epoch": 1.81, + "learning_rate": 3.494145289268932e-05, + "loss": 0.0, + "step": 7819 + }, + { + "epoch": 1.81, + "learning_rate": 3.4939527001001464e-05, + "loss": 0.0, + "step": 7820 + }, + { + "epoch": 1.81, + "learning_rate": 3.493760110931361e-05, + "loss": 0.004, + "step": 7821 + }, + { + "epoch": 1.81, + "learning_rate": 3.4935675217625766e-05, + "loss": 0.0001, + "step": 7822 + }, + { + "epoch": 1.81, + "learning_rate": 3.493374932593791e-05, + "loss": 0.0, + "step": 7823 + }, + { + "epoch": 1.81, + "learning_rate": 3.493182343425006e-05, + "loss": 0.0, + "step": 7824 + }, + { + "epoch": 1.81, + "learning_rate": 3.492989754256221e-05, + "loss": 0.0001, + "step": 7825 + }, + { + "epoch": 1.81, + "learning_rate": 3.4927971650874356e-05, + "loss": 0.0, + "step": 7826 + }, + { + "epoch": 1.81, + "learning_rate": 3.4926045759186503e-05, + "loss": 0.0, + "step": 7827 + }, + { + "epoch": 1.81, + "learning_rate": 3.492411986749866e-05, + "loss": 0.0, + "step": 7828 + }, + { + "epoch": 1.81, + "learning_rate": 3.4922193975810805e-05, + "loss": 0.0, + "step": 7829 + }, + { + "epoch": 1.81, + "learning_rate": 3.4920268084122946e-05, + "loss": 0.0, + "step": 7830 + }, + { + "epoch": 1.81, + "learning_rate": 3.49183421924351e-05, + "loss": 0.0, + "step": 7831 + }, + { + "epoch": 1.81, + "learning_rate": 3.491641630074725e-05, + "loss": 0.0, + "step": 7832 + }, + { + "epoch": 1.81, + "learning_rate": 3.4914490409059395e-05, + "loss": 0.0002, + "step": 7833 + }, + { + "epoch": 1.81, + "learning_rate": 3.491256451737154e-05, + "loss": 0.0, + "step": 7834 + }, + { + "epoch": 1.81, + "learning_rate": 3.49106386256837e-05, + "loss": 0.0002, + "step": 7835 + }, + { + "epoch": 1.81, + "learning_rate": 3.490871273399584e-05, + "loss": 0.0, + "step": 7836 + }, + { + "epoch": 1.81, + "learning_rate": 3.490678684230799e-05, + "loss": 0.0, + "step": 7837 + }, + { + "epoch": 1.81, + "learning_rate": 3.490486095062014e-05, + "loss": 0.0, + "step": 7838 + }, + { + "epoch": 1.81, + "learning_rate": 3.490293505893229e-05, + "loss": 0.0, + "step": 7839 + }, + { + "epoch": 1.81, + "learning_rate": 3.4901009167244434e-05, + "loss": 0.0, + "step": 7840 + }, + { + "epoch": 1.81, + "learning_rate": 3.489908327555658e-05, + "loss": 0.0005, + "step": 7841 + }, + { + "epoch": 1.81, + "learning_rate": 3.4897157383868736e-05, + "loss": 0.0, + "step": 7842 + }, + { + "epoch": 1.81, + "learning_rate": 3.489523149218088e-05, + "loss": 0.0005, + "step": 7843 + }, + { + "epoch": 1.81, + "learning_rate": 3.489330560049303e-05, + "loss": 0.0002, + "step": 7844 + }, + { + "epoch": 1.81, + "learning_rate": 3.489137970880518e-05, + "loss": 0.0, + "step": 7845 + }, + { + "epoch": 1.81, + "learning_rate": 3.4889453817117326e-05, + "loss": 0.0001, + "step": 7846 + }, + { + "epoch": 1.81, + "learning_rate": 3.488752792542947e-05, + "loss": 0.0001, + "step": 7847 + }, + { + "epoch": 1.81, + "learning_rate": 3.488560203374163e-05, + "loss": 0.0, + "step": 7848 + }, + { + "epoch": 1.81, + "learning_rate": 3.4883676142053775e-05, + "loss": 0.0001, + "step": 7849 + }, + { + "epoch": 1.81, + "learning_rate": 3.4881750250365916e-05, + "loss": 0.0, + "step": 7850 + }, + { + "epoch": 1.81, + "learning_rate": 3.487982435867807e-05, + "loss": 0.0, + "step": 7851 + }, + { + "epoch": 1.81, + "learning_rate": 3.487789846699022e-05, + "loss": 0.0, + "step": 7852 + }, + { + "epoch": 1.81, + "learning_rate": 3.4875972575302365e-05, + "loss": 0.0, + "step": 7853 + }, + { + "epoch": 1.82, + "learning_rate": 3.487404668361451e-05, + "loss": 0.0, + "step": 7854 + }, + { + "epoch": 1.82, + "learning_rate": 3.487212079192667e-05, + "loss": 0.0, + "step": 7855 + }, + { + "epoch": 1.82, + "learning_rate": 3.4870194900238814e-05, + "loss": 0.0, + "step": 7856 + }, + { + "epoch": 1.82, + "learning_rate": 3.486826900855096e-05, + "loss": 0.0, + "step": 7857 + }, + { + "epoch": 1.82, + "learning_rate": 3.486634311686311e-05, + "loss": 0.0004, + "step": 7858 + }, + { + "epoch": 1.82, + "learning_rate": 3.486441722517526e-05, + "loss": 0.0001, + "step": 7859 + }, + { + "epoch": 1.82, + "learning_rate": 3.4862491333487404e-05, + "loss": 0.0, + "step": 7860 + }, + { + "epoch": 1.82, + "learning_rate": 3.486056544179955e-05, + "loss": 0.0, + "step": 7861 + }, + { + "epoch": 1.82, + "learning_rate": 3.4858639550111706e-05, + "loss": 0.0, + "step": 7862 + }, + { + "epoch": 1.82, + "learning_rate": 3.485671365842385e-05, + "loss": 0.0001, + "step": 7863 + }, + { + "epoch": 1.82, + "learning_rate": 3.4854787766736e-05, + "loss": 0.0, + "step": 7864 + }, + { + "epoch": 1.82, + "learning_rate": 3.485286187504815e-05, + "loss": 0.0, + "step": 7865 + }, + { + "epoch": 1.82, + "learning_rate": 3.48509359833603e-05, + "loss": 0.0001, + "step": 7866 + }, + { + "epoch": 1.82, + "learning_rate": 3.484901009167244e-05, + "loss": 0.0001, + "step": 7867 + }, + { + "epoch": 1.82, + "learning_rate": 3.484708419998459e-05, + "loss": 0.0, + "step": 7868 + }, + { + "epoch": 1.82, + "learning_rate": 3.4845158308296745e-05, + "loss": 0.0001, + "step": 7869 + }, + { + "epoch": 1.82, + "learning_rate": 3.484323241660889e-05, + "loss": 0.0, + "step": 7870 + }, + { + "epoch": 1.82, + "learning_rate": 3.484130652492104e-05, + "loss": 0.0, + "step": 7871 + }, + { + "epoch": 1.82, + "learning_rate": 3.483938063323319e-05, + "loss": 0.0001, + "step": 7872 + }, + { + "epoch": 1.82, + "learning_rate": 3.483745474154534e-05, + "loss": 0.0001, + "step": 7873 + }, + { + "epoch": 1.82, + "learning_rate": 3.483552884985748e-05, + "loss": 0.0, + "step": 7874 + }, + { + "epoch": 1.82, + "learning_rate": 3.4833602958169637e-05, + "loss": 0.0, + "step": 7875 + }, + { + "epoch": 1.82, + "learning_rate": 3.4831677066481784e-05, + "loss": 0.0001, + "step": 7876 + }, + { + "epoch": 1.82, + "learning_rate": 3.482975117479393e-05, + "loss": 0.0, + "step": 7877 + }, + { + "epoch": 1.82, + "learning_rate": 3.482782528310608e-05, + "loss": 0.0001, + "step": 7878 + }, + { + "epoch": 1.82, + "learning_rate": 3.4825899391418227e-05, + "loss": 0.0, + "step": 7879 + }, + { + "epoch": 1.82, + "learning_rate": 3.482397349973038e-05, + "loss": 0.0, + "step": 7880 + }, + { + "epoch": 1.82, + "learning_rate": 3.482204760804252e-05, + "loss": 0.0, + "step": 7881 + }, + { + "epoch": 1.82, + "learning_rate": 3.4820121716354676e-05, + "loss": 0.0, + "step": 7882 + }, + { + "epoch": 1.82, + "learning_rate": 3.481819582466682e-05, + "loss": 0.0001, + "step": 7883 + }, + { + "epoch": 1.82, + "learning_rate": 3.481626993297897e-05, + "loss": 0.0025, + "step": 7884 + }, + { + "epoch": 1.82, + "learning_rate": 3.481434404129112e-05, + "loss": 0.0, + "step": 7885 + }, + { + "epoch": 1.82, + "learning_rate": 3.4812418149603266e-05, + "loss": 0.0005, + "step": 7886 + }, + { + "epoch": 1.82, + "learning_rate": 3.481049225791542e-05, + "loss": 0.0, + "step": 7887 + }, + { + "epoch": 1.82, + "learning_rate": 3.480856636622756e-05, + "loss": 0.0001, + "step": 7888 + }, + { + "epoch": 1.82, + "learning_rate": 3.4806640474539715e-05, + "loss": 0.0, + "step": 7889 + }, + { + "epoch": 1.82, + "learning_rate": 3.480471458285186e-05, + "loss": 0.0005, + "step": 7890 + }, + { + "epoch": 1.82, + "learning_rate": 3.480278869116401e-05, + "loss": 0.0, + "step": 7891 + }, + { + "epoch": 1.82, + "learning_rate": 3.480086279947616e-05, + "loss": 0.0003, + "step": 7892 + }, + { + "epoch": 1.82, + "learning_rate": 3.479893690778831e-05, + "loss": 0.0005, + "step": 7893 + }, + { + "epoch": 1.82, + "learning_rate": 3.479701101610046e-05, + "loss": 0.0002, + "step": 7894 + }, + { + "epoch": 1.82, + "learning_rate": 3.47950851244126e-05, + "loss": 0.0, + "step": 7895 + }, + { + "epoch": 1.82, + "learning_rate": 3.4793159232724754e-05, + "loss": 0.0004, + "step": 7896 + }, + { + "epoch": 1.83, + "learning_rate": 3.47912333410369e-05, + "loss": 0.0, + "step": 7897 + }, + { + "epoch": 1.83, + "learning_rate": 3.478930744934905e-05, + "loss": 0.0, + "step": 7898 + }, + { + "epoch": 1.83, + "learning_rate": 3.4787381557661196e-05, + "loss": 0.0024, + "step": 7899 + }, + { + "epoch": 1.83, + "learning_rate": 3.478545566597335e-05, + "loss": 0.0, + "step": 7900 + }, + { + "epoch": 1.83, + "learning_rate": 3.47835297742855e-05, + "loss": 0.0001, + "step": 7901 + }, + { + "epoch": 1.83, + "learning_rate": 3.4781603882597646e-05, + "loss": 0.0, + "step": 7902 + }, + { + "epoch": 1.83, + "learning_rate": 3.477967799090979e-05, + "loss": 0.0, + "step": 7903 + }, + { + "epoch": 1.83, + "learning_rate": 3.477775209922195e-05, + "loss": 0.0003, + "step": 7904 + }, + { + "epoch": 1.83, + "learning_rate": 3.477582620753409e-05, + "loss": 0.0, + "step": 7905 + }, + { + "epoch": 1.83, + "learning_rate": 3.4773900315846236e-05, + "loss": 0.0, + "step": 7906 + }, + { + "epoch": 1.83, + "learning_rate": 3.477197442415839e-05, + "loss": 0.0, + "step": 7907 + }, + { + "epoch": 1.83, + "learning_rate": 3.477004853247053e-05, + "loss": 0.0014, + "step": 7908 + }, + { + "epoch": 1.83, + "learning_rate": 3.4768122640782685e-05, + "loss": 0.0, + "step": 7909 + }, + { + "epoch": 1.83, + "learning_rate": 3.476619674909483e-05, + "loss": 0.0, + "step": 7910 + }, + { + "epoch": 1.83, + "learning_rate": 3.4764270857406986e-05, + "loss": 0.0, + "step": 7911 + }, + { + "epoch": 1.83, + "learning_rate": 3.476234496571913e-05, + "loss": 0.0, + "step": 7912 + }, + { + "epoch": 1.83, + "learning_rate": 3.476041907403128e-05, + "loss": 0.0, + "step": 7913 + }, + { + "epoch": 1.83, + "learning_rate": 3.475849318234343e-05, + "loss": 0.0001, + "step": 7914 + }, + { + "epoch": 1.83, + "learning_rate": 3.475656729065557e-05, + "loss": 0.0, + "step": 7915 + }, + { + "epoch": 1.83, + "learning_rate": 3.4754641398967724e-05, + "loss": 0.0002, + "step": 7916 + }, + { + "epoch": 1.83, + "learning_rate": 3.475271550727987e-05, + "loss": 0.0, + "step": 7917 + }, + { + "epoch": 1.83, + "learning_rate": 3.475078961559202e-05, + "loss": 0.0001, + "step": 7918 + }, + { + "epoch": 1.83, + "learning_rate": 3.4748863723904166e-05, + "loss": 0.0, + "step": 7919 + }, + { + "epoch": 1.83, + "learning_rate": 3.474693783221632e-05, + "loss": 0.0, + "step": 7920 + }, + { + "epoch": 1.83, + "learning_rate": 3.474501194052847e-05, + "loss": 0.0001, + "step": 7921 + }, + { + "epoch": 1.83, + "learning_rate": 3.4743086048840615e-05, + "loss": 0.0001, + "step": 7922 + }, + { + "epoch": 1.83, + "learning_rate": 3.474116015715276e-05, + "loss": 0.0001, + "step": 7923 + }, + { + "epoch": 1.83, + "learning_rate": 3.473923426546491e-05, + "loss": 0.0, + "step": 7924 + }, + { + "epoch": 1.83, + "learning_rate": 3.473730837377706e-05, + "loss": 0.0, + "step": 7925 + }, + { + "epoch": 1.83, + "learning_rate": 3.4735382482089205e-05, + "loss": 0.0001, + "step": 7926 + }, + { + "epoch": 1.83, + "learning_rate": 3.473345659040136e-05, + "loss": 0.0, + "step": 7927 + }, + { + "epoch": 1.83, + "learning_rate": 3.473153069871351e-05, + "loss": 0.0004, + "step": 7928 + }, + { + "epoch": 1.83, + "learning_rate": 3.4729604807025655e-05, + "loss": 0.0, + "step": 7929 + }, + { + "epoch": 1.83, + "learning_rate": 3.47276789153378e-05, + "loss": 0.0103, + "step": 7930 + }, + { + "epoch": 1.83, + "learning_rate": 3.4725753023649956e-05, + "loss": 0.0008, + "step": 7931 + }, + { + "epoch": 1.83, + "learning_rate": 3.47238271319621e-05, + "loss": 0.0, + "step": 7932 + }, + { + "epoch": 1.83, + "learning_rate": 3.4721901240274245e-05, + "loss": 0.0, + "step": 7933 + }, + { + "epoch": 1.83, + "learning_rate": 3.47199753485864e-05, + "loss": 0.0001, + "step": 7934 + }, + { + "epoch": 1.83, + "learning_rate": 3.4718049456898546e-05, + "loss": 0.0001, + "step": 7935 + }, + { + "epoch": 1.83, + "learning_rate": 3.4716123565210694e-05, + "loss": 0.0, + "step": 7936 + }, + { + "epoch": 1.83, + "learning_rate": 3.471419767352284e-05, + "loss": 0.0003, + "step": 7937 + }, + { + "epoch": 1.83, + "learning_rate": 3.4712271781834995e-05, + "loss": 0.0007, + "step": 7938 + }, + { + "epoch": 1.83, + "learning_rate": 3.4710345890147136e-05, + "loss": 0.0, + "step": 7939 + }, + { + "epoch": 1.83, + "learning_rate": 3.470841999845929e-05, + "loss": 0.0, + "step": 7940 + }, + { + "epoch": 1.84, + "learning_rate": 3.470649410677144e-05, + "loss": 0.0082, + "step": 7941 + }, + { + "epoch": 1.84, + "learning_rate": 3.4704568215083585e-05, + "loss": 0.0, + "step": 7942 + }, + { + "epoch": 1.84, + "learning_rate": 3.470264232339573e-05, + "loss": 0.0, + "step": 7943 + }, + { + "epoch": 1.84, + "learning_rate": 3.470071643170788e-05, + "loss": 0.0001, + "step": 7944 + }, + { + "epoch": 1.84, + "learning_rate": 3.4698790540020035e-05, + "loss": 0.0, + "step": 7945 + }, + { + "epoch": 1.84, + "learning_rate": 3.4696864648332175e-05, + "loss": 0.0, + "step": 7946 + }, + { + "epoch": 1.84, + "learning_rate": 3.469493875664433e-05, + "loss": 0.0, + "step": 7947 + }, + { + "epoch": 1.84, + "learning_rate": 3.469301286495648e-05, + "loss": 0.0001, + "step": 7948 + }, + { + "epoch": 1.84, + "learning_rate": 3.4691086973268624e-05, + "loss": 0.0043, + "step": 7949 + }, + { + "epoch": 1.84, + "learning_rate": 3.468916108158077e-05, + "loss": 0.0, + "step": 7950 + }, + { + "epoch": 1.84, + "learning_rate": 3.468723518989292e-05, + "loss": 0.0, + "step": 7951 + }, + { + "epoch": 1.84, + "learning_rate": 3.4685309298205074e-05, + "loss": 0.0001, + "step": 7952 + }, + { + "epoch": 1.84, + "learning_rate": 3.4683383406517214e-05, + "loss": 0.0, + "step": 7953 + }, + { + "epoch": 1.84, + "learning_rate": 3.468145751482937e-05, + "loss": 0.0, + "step": 7954 + }, + { + "epoch": 1.84, + "learning_rate": 3.4679531623141516e-05, + "loss": 0.0001, + "step": 7955 + }, + { + "epoch": 1.84, + "learning_rate": 3.4677605731453664e-05, + "loss": 0.0002, + "step": 7956 + }, + { + "epoch": 1.84, + "learning_rate": 3.467567983976581e-05, + "loss": 0.0002, + "step": 7957 + }, + { + "epoch": 1.84, + "learning_rate": 3.4673753948077965e-05, + "loss": 0.0, + "step": 7958 + }, + { + "epoch": 1.84, + "learning_rate": 3.467182805639011e-05, + "loss": 0.0041, + "step": 7959 + }, + { + "epoch": 1.84, + "learning_rate": 3.466990216470226e-05, + "loss": 0.0001, + "step": 7960 + }, + { + "epoch": 1.84, + "learning_rate": 3.466797627301441e-05, + "loss": 0.0, + "step": 7961 + }, + { + "epoch": 1.84, + "learning_rate": 3.4666050381326555e-05, + "loss": 0.0001, + "step": 7962 + }, + { + "epoch": 1.84, + "learning_rate": 3.46641244896387e-05, + "loss": 0.0, + "step": 7963 + }, + { + "epoch": 1.84, + "learning_rate": 3.466219859795085e-05, + "loss": 0.0001, + "step": 7964 + }, + { + "epoch": 1.84, + "learning_rate": 3.4660272706263004e-05, + "loss": 0.0, + "step": 7965 + }, + { + "epoch": 1.84, + "learning_rate": 3.465834681457515e-05, + "loss": 0.0001, + "step": 7966 + }, + { + "epoch": 1.84, + "learning_rate": 3.46564209228873e-05, + "loss": 0.0, + "step": 7967 + }, + { + "epoch": 1.84, + "learning_rate": 3.465449503119945e-05, + "loss": 0.0, + "step": 7968 + }, + { + "epoch": 1.84, + "learning_rate": 3.46525691395116e-05, + "loss": 0.0, + "step": 7969 + }, + { + "epoch": 1.84, + "learning_rate": 3.465064324782374e-05, + "loss": 0.0002, + "step": 7970 + }, + { + "epoch": 1.84, + "learning_rate": 3.464871735613589e-05, + "loss": 0.0004, + "step": 7971 + }, + { + "epoch": 1.84, + "learning_rate": 3.4646791464448044e-05, + "loss": 0.0001, + "step": 7972 + }, + { + "epoch": 1.84, + "learning_rate": 3.464486557276019e-05, + "loss": 0.0001, + "step": 7973 + }, + { + "epoch": 1.84, + "learning_rate": 3.464293968107234e-05, + "loss": 0.0, + "step": 7974 + }, + { + "epoch": 1.84, + "learning_rate": 3.4641013789384486e-05, + "loss": 0.0, + "step": 7975 + }, + { + "epoch": 1.84, + "learning_rate": 3.463908789769664e-05, + "loss": 0.0, + "step": 7976 + }, + { + "epoch": 1.84, + "learning_rate": 3.463716200600878e-05, + "loss": 0.0, + "step": 7977 + }, + { + "epoch": 1.84, + "learning_rate": 3.4635236114320935e-05, + "loss": 0.0002, + "step": 7978 + }, + { + "epoch": 1.84, + "learning_rate": 3.463331022263308e-05, + "loss": 0.0, + "step": 7979 + }, + { + "epoch": 1.84, + "learning_rate": 3.463138433094523e-05, + "loss": 0.0, + "step": 7980 + }, + { + "epoch": 1.84, + "learning_rate": 3.462945843925738e-05, + "loss": 0.0, + "step": 7981 + }, + { + "epoch": 1.84, + "learning_rate": 3.4627532547569525e-05, + "loss": 0.0, + "step": 7982 + }, + { + "epoch": 1.84, + "learning_rate": 3.462560665588168e-05, + "loss": 0.0, + "step": 7983 + }, + { + "epoch": 1.85, + "learning_rate": 3.462368076419382e-05, + "loss": 0.0, + "step": 7984 + }, + { + "epoch": 1.85, + "learning_rate": 3.4621754872505974e-05, + "loss": 0.0001, + "step": 7985 + }, + { + "epoch": 1.85, + "learning_rate": 3.461982898081812e-05, + "loss": 0.0, + "step": 7986 + }, + { + "epoch": 1.85, + "learning_rate": 3.461790308913027e-05, + "loss": 0.0, + "step": 7987 + }, + { + "epoch": 1.85, + "learning_rate": 3.461597719744242e-05, + "loss": 0.0001, + "step": 7988 + }, + { + "epoch": 1.85, + "learning_rate": 3.4614051305754564e-05, + "loss": 0.0014, + "step": 7989 + }, + { + "epoch": 1.85, + "learning_rate": 3.461212541406671e-05, + "loss": 0.0002, + "step": 7990 + }, + { + "epoch": 1.85, + "learning_rate": 3.461019952237886e-05, + "loss": 0.0, + "step": 7991 + }, + { + "epoch": 1.85, + "learning_rate": 3.4608273630691013e-05, + "loss": 0.0, + "step": 7992 + }, + { + "epoch": 1.85, + "learning_rate": 3.460634773900316e-05, + "loss": 0.0001, + "step": 7993 + }, + { + "epoch": 1.85, + "learning_rate": 3.460442184731531e-05, + "loss": 0.0, + "step": 7994 + }, + { + "epoch": 1.85, + "learning_rate": 3.4602495955627456e-05, + "loss": 0.0, + "step": 7995 + }, + { + "epoch": 1.85, + "learning_rate": 3.460057006393961e-05, + "loss": 0.0, + "step": 7996 + }, + { + "epoch": 1.85, + "learning_rate": 3.459864417225175e-05, + "loss": 0.0, + "step": 7997 + }, + { + "epoch": 1.85, + "learning_rate": 3.45967182805639e-05, + "loss": 0.0, + "step": 7998 + }, + { + "epoch": 1.85, + "learning_rate": 3.459479238887605e-05, + "loss": 0.0, + "step": 7999 + }, + { + "epoch": 1.85, + "learning_rate": 3.45928664971882e-05, + "loss": 0.0, + "step": 8000 + }, + { + "epoch": 1.85, + "learning_rate": 3.459094060550035e-05, + "loss": 0.005, + "step": 8001 + }, + { + "epoch": 1.85, + "learning_rate": 3.4589014713812495e-05, + "loss": 0.0, + "step": 8002 + }, + { + "epoch": 1.85, + "learning_rate": 3.458708882212465e-05, + "loss": 0.0, + "step": 8003 + }, + { + "epoch": 1.85, + "learning_rate": 3.458516293043679e-05, + "loss": 0.0, + "step": 8004 + }, + { + "epoch": 1.85, + "learning_rate": 3.4583237038748944e-05, + "loss": 0.0, + "step": 8005 + }, + { + "epoch": 1.85, + "learning_rate": 3.458131114706109e-05, + "loss": 0.0, + "step": 8006 + }, + { + "epoch": 1.85, + "learning_rate": 3.457938525537324e-05, + "loss": 0.0008, + "step": 8007 + }, + { + "epoch": 1.85, + "learning_rate": 3.457745936368539e-05, + "loss": 0.0, + "step": 8008 + }, + { + "epoch": 1.85, + "learning_rate": 3.4575533471997534e-05, + "loss": 0.0, + "step": 8009 + }, + { + "epoch": 1.85, + "learning_rate": 3.457360758030969e-05, + "loss": 0.0001, + "step": 8010 + }, + { + "epoch": 1.85, + "learning_rate": 3.457168168862183e-05, + "loss": 0.0003, + "step": 8011 + }, + { + "epoch": 1.85, + "learning_rate": 3.456975579693398e-05, + "loss": 0.0001, + "step": 8012 + }, + { + "epoch": 1.85, + "learning_rate": 3.456782990524613e-05, + "loss": 0.0, + "step": 8013 + }, + { + "epoch": 1.85, + "learning_rate": 3.456590401355828e-05, + "loss": 0.0001, + "step": 8014 + }, + { + "epoch": 1.85, + "learning_rate": 3.4563978121870426e-05, + "loss": 0.0001, + "step": 8015 + }, + { + "epoch": 1.85, + "learning_rate": 3.456205223018258e-05, + "loss": 0.0, + "step": 8016 + }, + { + "epoch": 1.85, + "learning_rate": 3.456012633849473e-05, + "loss": 0.0008, + "step": 8017 + }, + { + "epoch": 1.85, + "learning_rate": 3.455820044680687e-05, + "loss": 0.0, + "step": 8018 + }, + { + "epoch": 1.85, + "learning_rate": 3.455627455511902e-05, + "loss": 0.0, + "step": 8019 + }, + { + "epoch": 1.85, + "learning_rate": 3.455434866343117e-05, + "loss": 0.0, + "step": 8020 + }, + { + "epoch": 1.85, + "learning_rate": 3.455242277174332e-05, + "loss": 0.0001, + "step": 8021 + }, + { + "epoch": 1.85, + "learning_rate": 3.4550496880055465e-05, + "loss": 0.0, + "step": 8022 + }, + { + "epoch": 1.85, + "learning_rate": 3.454857098836762e-05, + "loss": 0.0, + "step": 8023 + }, + { + "epoch": 1.85, + "learning_rate": 3.454664509667977e-05, + "loss": 0.0015, + "step": 8024 + }, + { + "epoch": 1.85, + "learning_rate": 3.4544719204991914e-05, + "loss": 0.0001, + "step": 8025 + }, + { + "epoch": 1.85, + "learning_rate": 3.454279331330406e-05, + "loss": 0.0004, + "step": 8026 + }, + { + "epoch": 1.86, + "learning_rate": 3.454086742161621e-05, + "loss": 0.0001, + "step": 8027 + }, + { + "epoch": 1.86, + "learning_rate": 3.4538941529928357e-05, + "loss": 0.0, + "step": 8028 + }, + { + "epoch": 1.86, + "learning_rate": 3.4537015638240504e-05, + "loss": 0.0001, + "step": 8029 + }, + { + "epoch": 1.86, + "learning_rate": 3.453508974655266e-05, + "loss": 0.0, + "step": 8030 + }, + { + "epoch": 1.86, + "learning_rate": 3.4533163854864806e-05, + "loss": 0.0004, + "step": 8031 + }, + { + "epoch": 1.86, + "learning_rate": 3.453123796317695e-05, + "loss": 0.0001, + "step": 8032 + }, + { + "epoch": 1.86, + "learning_rate": 3.45293120714891e-05, + "loss": 0.0004, + "step": 8033 + }, + { + "epoch": 1.86, + "learning_rate": 3.4527386179801255e-05, + "loss": 0.0001, + "step": 8034 + }, + { + "epoch": 1.86, + "learning_rate": 3.4525460288113396e-05, + "loss": 0.0, + "step": 8035 + }, + { + "epoch": 1.86, + "learning_rate": 3.452353439642554e-05, + "loss": 0.0001, + "step": 8036 + }, + { + "epoch": 1.86, + "learning_rate": 3.45216085047377e-05, + "loss": 0.0005, + "step": 8037 + }, + { + "epoch": 1.86, + "learning_rate": 3.4519682613049845e-05, + "loss": 0.0001, + "step": 8038 + }, + { + "epoch": 1.86, + "learning_rate": 3.451775672136199e-05, + "loss": 0.0, + "step": 8039 + }, + { + "epoch": 1.86, + "learning_rate": 3.451583082967414e-05, + "loss": 0.0, + "step": 8040 + }, + { + "epoch": 1.86, + "learning_rate": 3.4513904937986294e-05, + "loss": 0.0, + "step": 8041 + }, + { + "epoch": 1.86, + "learning_rate": 3.4511979046298435e-05, + "loss": 0.0001, + "step": 8042 + }, + { + "epoch": 1.86, + "learning_rate": 3.451005315461059e-05, + "loss": 0.0001, + "step": 8043 + }, + { + "epoch": 1.86, + "learning_rate": 3.4508127262922737e-05, + "loss": 0.0, + "step": 8044 + }, + { + "epoch": 1.86, + "learning_rate": 3.4506201371234884e-05, + "loss": 0.0004, + "step": 8045 + }, + { + "epoch": 1.86, + "learning_rate": 3.450427547954703e-05, + "loss": 0.0, + "step": 8046 + }, + { + "epoch": 1.86, + "learning_rate": 3.450234958785918e-05, + "loss": 0.0001, + "step": 8047 + }, + { + "epoch": 1.86, + "learning_rate": 3.450042369617133e-05, + "loss": 0.0001, + "step": 8048 + }, + { + "epoch": 1.86, + "learning_rate": 3.4498497804483474e-05, + "loss": 0.0001, + "step": 8049 + }, + { + "epoch": 1.86, + "learning_rate": 3.449657191279563e-05, + "loss": 0.0, + "step": 8050 + }, + { + "epoch": 1.86, + "learning_rate": 3.4494646021107776e-05, + "loss": 0.0, + "step": 8051 + }, + { + "epoch": 1.86, + "learning_rate": 3.449272012941992e-05, + "loss": 0.0, + "step": 8052 + }, + { + "epoch": 1.86, + "learning_rate": 3.449079423773207e-05, + "loss": 0.0, + "step": 8053 + }, + { + "epoch": 1.86, + "learning_rate": 3.448886834604422e-05, + "loss": 0.0, + "step": 8054 + }, + { + "epoch": 1.86, + "learning_rate": 3.448694245435637e-05, + "loss": 0.0, + "step": 8055 + }, + { + "epoch": 1.86, + "learning_rate": 3.448501656266851e-05, + "loss": 0.0, + "step": 8056 + }, + { + "epoch": 1.86, + "learning_rate": 3.448309067098067e-05, + "loss": 0.0001, + "step": 8057 + }, + { + "epoch": 1.86, + "learning_rate": 3.4481164779292815e-05, + "loss": 0.0, + "step": 8058 + }, + { + "epoch": 1.86, + "learning_rate": 3.447923888760496e-05, + "loss": 0.0002, + "step": 8059 + }, + { + "epoch": 1.86, + "learning_rate": 3.447731299591711e-05, + "loss": 0.0001, + "step": 8060 + }, + { + "epoch": 1.86, + "learning_rate": 3.4475387104229264e-05, + "loss": 0.0001, + "step": 8061 + }, + { + "epoch": 1.86, + "learning_rate": 3.4473461212541405e-05, + "loss": 0.0, + "step": 8062 + }, + { + "epoch": 1.86, + "learning_rate": 3.447153532085355e-05, + "loss": 0.0001, + "step": 8063 + }, + { + "epoch": 1.86, + "learning_rate": 3.4469609429165706e-05, + "loss": 0.0, + "step": 8064 + }, + { + "epoch": 1.86, + "learning_rate": 3.4467683537477854e-05, + "loss": 0.0, + "step": 8065 + }, + { + "epoch": 1.86, + "learning_rate": 3.446575764579e-05, + "loss": 0.0, + "step": 8066 + }, + { + "epoch": 1.86, + "learning_rate": 3.446383175410215e-05, + "loss": 0.0, + "step": 8067 + }, + { + "epoch": 1.86, + "learning_rate": 3.44619058624143e-05, + "loss": 0.0, + "step": 8068 + }, + { + "epoch": 1.86, + "learning_rate": 3.4459979970726444e-05, + "loss": 0.0, + "step": 8069 + }, + { + "epoch": 1.87, + "learning_rate": 3.44580540790386e-05, + "loss": 0.0001, + "step": 8070 + }, + { + "epoch": 1.87, + "learning_rate": 3.4456128187350746e-05, + "loss": 0.0001, + "step": 8071 + }, + { + "epoch": 1.87, + "learning_rate": 3.445420229566289e-05, + "loss": 0.0, + "step": 8072 + }, + { + "epoch": 1.87, + "learning_rate": 3.445227640397504e-05, + "loss": 0.0, + "step": 8073 + }, + { + "epoch": 1.87, + "learning_rate": 3.445035051228719e-05, + "loss": 0.0, + "step": 8074 + }, + { + "epoch": 1.87, + "learning_rate": 3.444842462059934e-05, + "loss": 0.0002, + "step": 8075 + }, + { + "epoch": 1.87, + "learning_rate": 3.444649872891148e-05, + "loss": 0.0, + "step": 8076 + }, + { + "epoch": 1.87, + "learning_rate": 3.444457283722364e-05, + "loss": 0.0001, + "step": 8077 + }, + { + "epoch": 1.87, + "learning_rate": 3.4442646945535785e-05, + "loss": 0.0, + "step": 8078 + }, + { + "epoch": 1.87, + "learning_rate": 3.444072105384793e-05, + "loss": 0.0, + "step": 8079 + }, + { + "epoch": 1.87, + "learning_rate": 3.443879516216008e-05, + "loss": 0.0, + "step": 8080 + }, + { + "epoch": 1.87, + "learning_rate": 3.4436869270472234e-05, + "loss": 0.0, + "step": 8081 + }, + { + "epoch": 1.87, + "learning_rate": 3.443494337878438e-05, + "loss": 0.0, + "step": 8082 + }, + { + "epoch": 1.87, + "learning_rate": 3.443301748709652e-05, + "loss": 0.0, + "step": 8083 + }, + { + "epoch": 1.87, + "learning_rate": 3.4431091595408676e-05, + "loss": 0.0005, + "step": 8084 + }, + { + "epoch": 1.87, + "learning_rate": 3.4429165703720824e-05, + "loss": 0.0, + "step": 8085 + }, + { + "epoch": 1.87, + "learning_rate": 3.442723981203297e-05, + "loss": 0.0002, + "step": 8086 + }, + { + "epoch": 1.87, + "learning_rate": 3.442531392034512e-05, + "loss": 0.0001, + "step": 8087 + }, + { + "epoch": 1.87, + "learning_rate": 3.442338802865727e-05, + "loss": 0.0, + "step": 8088 + }, + { + "epoch": 1.87, + "learning_rate": 3.442146213696942e-05, + "loss": 0.0, + "step": 8089 + }, + { + "epoch": 1.87, + "learning_rate": 3.441953624528157e-05, + "loss": 0.0004, + "step": 8090 + }, + { + "epoch": 1.87, + "learning_rate": 3.4417610353593715e-05, + "loss": 0.0001, + "step": 8091 + }, + { + "epoch": 1.87, + "learning_rate": 3.441568446190586e-05, + "loss": 0.0, + "step": 8092 + }, + { + "epoch": 1.87, + "learning_rate": 3.441375857021801e-05, + "loss": 0.0, + "step": 8093 + }, + { + "epoch": 1.87, + "learning_rate": 3.441183267853016e-05, + "loss": 0.0, + "step": 8094 + }, + { + "epoch": 1.87, + "learning_rate": 3.440990678684231e-05, + "loss": 0.0, + "step": 8095 + }, + { + "epoch": 1.87, + "learning_rate": 3.440798089515446e-05, + "loss": 0.0, + "step": 8096 + }, + { + "epoch": 1.87, + "learning_rate": 3.440605500346661e-05, + "loss": 0.0, + "step": 8097 + }, + { + "epoch": 1.87, + "learning_rate": 3.4404129111778755e-05, + "loss": 0.0, + "step": 8098 + }, + { + "epoch": 1.87, + "learning_rate": 3.440220322009091e-05, + "loss": 0.0096, + "step": 8099 + }, + { + "epoch": 1.87, + "learning_rate": 3.440027732840305e-05, + "loss": 0.0, + "step": 8100 + }, + { + "epoch": 1.87, + "learning_rate": 3.43983514367152e-05, + "loss": 0.0001, + "step": 8101 + }, + { + "epoch": 1.87, + "learning_rate": 3.439642554502735e-05, + "loss": 0.0004, + "step": 8102 + }, + { + "epoch": 1.87, + "learning_rate": 3.43944996533395e-05, + "loss": 0.0, + "step": 8103 + }, + { + "epoch": 1.87, + "learning_rate": 3.4392573761651646e-05, + "loss": 0.0, + "step": 8104 + }, + { + "epoch": 1.87, + "learning_rate": 3.4390647869963794e-05, + "loss": 0.0, + "step": 8105 + }, + { + "epoch": 1.87, + "learning_rate": 3.438872197827595e-05, + "loss": 0.0, + "step": 8106 + }, + { + "epoch": 1.87, + "learning_rate": 3.438679608658809e-05, + "loss": 0.0, + "step": 8107 + }, + { + "epoch": 1.87, + "learning_rate": 3.438487019490024e-05, + "loss": 0.0, + "step": 8108 + }, + { + "epoch": 1.87, + "learning_rate": 3.438294430321239e-05, + "loss": 0.0001, + "step": 8109 + }, + { + "epoch": 1.87, + "learning_rate": 3.438101841152454e-05, + "loss": 0.0, + "step": 8110 + }, + { + "epoch": 1.87, + "learning_rate": 3.4379092519836685e-05, + "loss": 0.0, + "step": 8111 + }, + { + "epoch": 1.87, + "learning_rate": 3.437716662814883e-05, + "loss": 0.0, + "step": 8112 + }, + { + "epoch": 1.87, + "learning_rate": 3.437524073646099e-05, + "loss": 0.0001, + "step": 8113 + }, + { + "epoch": 1.88, + "learning_rate": 3.437331484477313e-05, + "loss": 0.0001, + "step": 8114 + }, + { + "epoch": 1.88, + "learning_rate": 3.437138895308528e-05, + "loss": 0.0, + "step": 8115 + }, + { + "epoch": 1.88, + "learning_rate": 3.436946306139743e-05, + "loss": 0.0, + "step": 8116 + }, + { + "epoch": 1.88, + "learning_rate": 3.436753716970958e-05, + "loss": 0.0, + "step": 8117 + }, + { + "epoch": 1.88, + "learning_rate": 3.4365611278021724e-05, + "loss": 0.0, + "step": 8118 + }, + { + "epoch": 1.88, + "learning_rate": 3.436368538633387e-05, + "loss": 0.0, + "step": 8119 + }, + { + "epoch": 1.88, + "learning_rate": 3.4361759494646026e-05, + "loss": 0.0, + "step": 8120 + }, + { + "epoch": 1.88, + "learning_rate": 3.435983360295817e-05, + "loss": 0.0001, + "step": 8121 + }, + { + "epoch": 1.88, + "learning_rate": 3.435790771127032e-05, + "loss": 0.0002, + "step": 8122 + }, + { + "epoch": 1.88, + "learning_rate": 3.435598181958247e-05, + "loss": 0.0, + "step": 8123 + }, + { + "epoch": 1.88, + "learning_rate": 3.4354055927894616e-05, + "loss": 0.0001, + "step": 8124 + }, + { + "epoch": 1.88, + "learning_rate": 3.4352130036206764e-05, + "loss": 0.0001, + "step": 8125 + }, + { + "epoch": 1.88, + "learning_rate": 3.435020414451892e-05, + "loss": 0.001, + "step": 8126 + }, + { + "epoch": 1.88, + "learning_rate": 3.4348278252831065e-05, + "loss": 0.0, + "step": 8127 + }, + { + "epoch": 1.88, + "learning_rate": 3.434635236114321e-05, + "loss": 0.0, + "step": 8128 + }, + { + "epoch": 1.88, + "learning_rate": 3.434442646945536e-05, + "loss": 0.0002, + "step": 8129 + }, + { + "epoch": 1.88, + "learning_rate": 3.434250057776751e-05, + "loss": 0.0053, + "step": 8130 + }, + { + "epoch": 1.88, + "learning_rate": 3.4340574686079655e-05, + "loss": 0.0001, + "step": 8131 + }, + { + "epoch": 1.88, + "learning_rate": 3.43386487943918e-05, + "loss": 0.0, + "step": 8132 + }, + { + "epoch": 1.88, + "learning_rate": 3.433672290270396e-05, + "loss": 0.0, + "step": 8133 + }, + { + "epoch": 1.88, + "learning_rate": 3.4334797011016104e-05, + "loss": 0.0, + "step": 8134 + }, + { + "epoch": 1.88, + "learning_rate": 3.433287111932825e-05, + "loss": 0.0, + "step": 8135 + }, + { + "epoch": 1.88, + "learning_rate": 3.43309452276404e-05, + "loss": 0.0002, + "step": 8136 + }, + { + "epoch": 1.88, + "learning_rate": 3.4329019335952554e-05, + "loss": 0.0, + "step": 8137 + }, + { + "epoch": 1.88, + "learning_rate": 3.4327093444264694e-05, + "loss": 0.0001, + "step": 8138 + }, + { + "epoch": 1.88, + "learning_rate": 3.432516755257684e-05, + "loss": 0.0, + "step": 8139 + }, + { + "epoch": 1.88, + "learning_rate": 3.4323241660888996e-05, + "loss": 0.0, + "step": 8140 + }, + { + "epoch": 1.88, + "learning_rate": 3.432131576920114e-05, + "loss": 0.0, + "step": 8141 + }, + { + "epoch": 1.88, + "learning_rate": 3.431938987751329e-05, + "loss": 0.0003, + "step": 8142 + }, + { + "epoch": 1.88, + "learning_rate": 3.431746398582544e-05, + "loss": 0.0, + "step": 8143 + }, + { + "epoch": 1.88, + "learning_rate": 3.4315538094137586e-05, + "loss": 0.0, + "step": 8144 + }, + { + "epoch": 1.88, + "learning_rate": 3.4313612202449733e-05, + "loss": 0.0, + "step": 8145 + }, + { + "epoch": 1.88, + "learning_rate": 3.431168631076189e-05, + "loss": 0.0, + "step": 8146 + }, + { + "epoch": 1.88, + "learning_rate": 3.4309760419074035e-05, + "loss": 0.0001, + "step": 8147 + }, + { + "epoch": 1.88, + "learning_rate": 3.4307834527386176e-05, + "loss": 0.0001, + "step": 8148 + }, + { + "epoch": 1.88, + "learning_rate": 3.430590863569833e-05, + "loss": 0.0, + "step": 8149 + }, + { + "epoch": 1.88, + "learning_rate": 3.430398274401048e-05, + "loss": 0.0, + "step": 8150 + }, + { + "epoch": 1.88, + "learning_rate": 3.4302056852322625e-05, + "loss": 0.0007, + "step": 8151 + }, + { + "epoch": 1.88, + "learning_rate": 3.430013096063477e-05, + "loss": 0.0001, + "step": 8152 + }, + { + "epoch": 1.88, + "learning_rate": 3.429820506894693e-05, + "loss": 0.0, + "step": 8153 + }, + { + "epoch": 1.88, + "learning_rate": 3.4296279177259074e-05, + "loss": 0.0001, + "step": 8154 + }, + { + "epoch": 1.88, + "learning_rate": 3.429435328557122e-05, + "loss": 0.0, + "step": 8155 + }, + { + "epoch": 1.88, + "learning_rate": 3.429242739388337e-05, + "loss": 0.0, + "step": 8156 + }, + { + "epoch": 1.89, + "learning_rate": 3.429050150219552e-05, + "loss": 0.0001, + "step": 8157 + }, + { + "epoch": 1.89, + "learning_rate": 3.4288575610507664e-05, + "loss": 0.0, + "step": 8158 + }, + { + "epoch": 1.89, + "learning_rate": 3.428664971881981e-05, + "loss": 0.0, + "step": 8159 + }, + { + "epoch": 1.89, + "learning_rate": 3.4284723827131966e-05, + "loss": 0.0021, + "step": 8160 + }, + { + "epoch": 1.89, + "learning_rate": 3.4282797935444113e-05, + "loss": 0.0002, + "step": 8161 + }, + { + "epoch": 1.89, + "learning_rate": 3.428087204375626e-05, + "loss": 0.0003, + "step": 8162 + }, + { + "epoch": 1.89, + "learning_rate": 3.427894615206841e-05, + "loss": 0.0, + "step": 8163 + }, + { + "epoch": 1.89, + "learning_rate": 3.427702026038056e-05, + "loss": 0.0008, + "step": 8164 + }, + { + "epoch": 1.89, + "learning_rate": 3.42750943686927e-05, + "loss": 0.0, + "step": 8165 + }, + { + "epoch": 1.89, + "learning_rate": 3.427316847700485e-05, + "loss": 0.0, + "step": 8166 + }, + { + "epoch": 1.89, + "learning_rate": 3.4271242585317005e-05, + "loss": 0.0, + "step": 8167 + }, + { + "epoch": 1.89, + "learning_rate": 3.426931669362915e-05, + "loss": 0.0013, + "step": 8168 + }, + { + "epoch": 1.89, + "learning_rate": 3.42673908019413e-05, + "loss": 0.0002, + "step": 8169 + }, + { + "epoch": 1.89, + "learning_rate": 3.426546491025345e-05, + "loss": 0.0, + "step": 8170 + }, + { + "epoch": 1.89, + "learning_rate": 3.42635390185656e-05, + "loss": 0.0019, + "step": 8171 + }, + { + "epoch": 1.89, + "learning_rate": 3.426161312687774e-05, + "loss": 0.0006, + "step": 8172 + }, + { + "epoch": 1.89, + "learning_rate": 3.42596872351899e-05, + "loss": 0.0001, + "step": 8173 + }, + { + "epoch": 1.89, + "learning_rate": 3.4257761343502044e-05, + "loss": 0.0001, + "step": 8174 + }, + { + "epoch": 1.89, + "learning_rate": 3.425583545181419e-05, + "loss": 0.0, + "step": 8175 + }, + { + "epoch": 1.89, + "learning_rate": 3.425390956012634e-05, + "loss": 0.0001, + "step": 8176 + }, + { + "epoch": 1.89, + "learning_rate": 3.4251983668438487e-05, + "loss": 0.0, + "step": 8177 + }, + { + "epoch": 1.89, + "learning_rate": 3.425005777675064e-05, + "loss": 0.0, + "step": 8178 + }, + { + "epoch": 1.89, + "learning_rate": 3.424813188506278e-05, + "loss": 0.0, + "step": 8179 + }, + { + "epoch": 1.89, + "learning_rate": 3.4246205993374936e-05, + "loss": 0.0, + "step": 8180 + }, + { + "epoch": 1.89, + "learning_rate": 3.424428010168708e-05, + "loss": 0.0, + "step": 8181 + }, + { + "epoch": 1.89, + "learning_rate": 3.424235420999923e-05, + "loss": 0.0, + "step": 8182 + }, + { + "epoch": 1.89, + "learning_rate": 3.424042831831138e-05, + "loss": 0.0001, + "step": 8183 + }, + { + "epoch": 1.89, + "learning_rate": 3.423850242662353e-05, + "loss": 0.0001, + "step": 8184 + }, + { + "epoch": 1.89, + "learning_rate": 3.423657653493568e-05, + "loss": 0.0001, + "step": 8185 + }, + { + "epoch": 1.89, + "learning_rate": 3.423465064324782e-05, + "loss": 0.0001, + "step": 8186 + }, + { + "epoch": 1.89, + "learning_rate": 3.4232724751559975e-05, + "loss": 0.0001, + "step": 8187 + }, + { + "epoch": 1.89, + "learning_rate": 3.423079885987212e-05, + "loss": 0.0, + "step": 8188 + }, + { + "epoch": 1.89, + "learning_rate": 3.422887296818427e-05, + "loss": 0.0, + "step": 8189 + }, + { + "epoch": 1.89, + "learning_rate": 3.422694707649642e-05, + "loss": 0.0, + "step": 8190 + }, + { + "epoch": 1.89, + "learning_rate": 3.422502118480857e-05, + "loss": 0.0, + "step": 8191 + }, + { + "epoch": 1.89, + "learning_rate": 3.422309529312072e-05, + "loss": 0.0, + "step": 8192 + }, + { + "epoch": 1.89, + "learning_rate": 3.4221169401432867e-05, + "loss": 0.0, + "step": 8193 + }, + { + "epoch": 1.89, + "learning_rate": 3.4219243509745014e-05, + "loss": 0.0001, + "step": 8194 + }, + { + "epoch": 1.89, + "learning_rate": 3.421731761805716e-05, + "loss": 0.0, + "step": 8195 + }, + { + "epoch": 1.89, + "learning_rate": 3.421539172636931e-05, + "loss": 0.0008, + "step": 8196 + }, + { + "epoch": 1.89, + "learning_rate": 3.4213465834681456e-05, + "loss": 0.0002, + "step": 8197 + }, + { + "epoch": 1.89, + "learning_rate": 3.421153994299361e-05, + "loss": 0.0, + "step": 8198 + }, + { + "epoch": 1.89, + "learning_rate": 3.420961405130576e-05, + "loss": 0.0001, + "step": 8199 + }, + { + "epoch": 1.9, + "learning_rate": 3.4207688159617906e-05, + "loss": 0.0, + "step": 8200 + }, + { + "epoch": 1.9, + "learning_rate": 3.420576226793005e-05, + "loss": 0.0, + "step": 8201 + }, + { + "epoch": 1.9, + "learning_rate": 3.420383637624221e-05, + "loss": 0.0, + "step": 8202 + }, + { + "epoch": 1.9, + "learning_rate": 3.420191048455435e-05, + "loss": 0.0, + "step": 8203 + }, + { + "epoch": 1.9, + "learning_rate": 3.4199984592866496e-05, + "loss": 0.0, + "step": 8204 + }, + { + "epoch": 1.9, + "learning_rate": 3.419805870117865e-05, + "loss": 0.0001, + "step": 8205 + }, + { + "epoch": 1.9, + "learning_rate": 3.41961328094908e-05, + "loss": 0.0018, + "step": 8206 + }, + { + "epoch": 1.9, + "learning_rate": 3.4194206917802945e-05, + "loss": 0.0, + "step": 8207 + }, + { + "epoch": 1.9, + "learning_rate": 3.419228102611509e-05, + "loss": 0.0001, + "step": 8208 + }, + { + "epoch": 1.9, + "learning_rate": 3.4190355134427247e-05, + "loss": 0.0001, + "step": 8209 + }, + { + "epoch": 1.9, + "learning_rate": 3.418842924273939e-05, + "loss": 0.0, + "step": 8210 + }, + { + "epoch": 1.9, + "learning_rate": 3.418650335105154e-05, + "loss": 0.0, + "step": 8211 + }, + { + "epoch": 1.9, + "learning_rate": 3.418457745936369e-05, + "loss": 0.0002, + "step": 8212 + }, + { + "epoch": 1.9, + "learning_rate": 3.418265156767583e-05, + "loss": 0.0, + "step": 8213 + }, + { + "epoch": 1.9, + "learning_rate": 3.4180725675987984e-05, + "loss": 0.0005, + "step": 8214 + }, + { + "epoch": 1.9, + "learning_rate": 3.417879978430013e-05, + "loss": 0.0, + "step": 8215 + }, + { + "epoch": 1.9, + "learning_rate": 3.417687389261228e-05, + "loss": 0.0001, + "step": 8216 + }, + { + "epoch": 1.9, + "learning_rate": 3.4174948000924426e-05, + "loss": 0.0, + "step": 8217 + }, + { + "epoch": 1.9, + "learning_rate": 3.417302210923658e-05, + "loss": 0.0, + "step": 8218 + }, + { + "epoch": 1.9, + "learning_rate": 3.417109621754873e-05, + "loss": 0.0018, + "step": 8219 + }, + { + "epoch": 1.9, + "learning_rate": 3.4169170325860876e-05, + "loss": 0.0002, + "step": 8220 + }, + { + "epoch": 1.9, + "learning_rate": 3.416724443417302e-05, + "loss": 0.0, + "step": 8221 + }, + { + "epoch": 1.9, + "learning_rate": 3.416531854248517e-05, + "loss": 0.0, + "step": 8222 + }, + { + "epoch": 1.9, + "learning_rate": 3.416339265079732e-05, + "loss": 0.0, + "step": 8223 + }, + { + "epoch": 1.9, + "learning_rate": 3.4161466759109465e-05, + "loss": 0.0001, + "step": 8224 + }, + { + "epoch": 1.9, + "learning_rate": 3.415954086742162e-05, + "loss": 0.0002, + "step": 8225 + }, + { + "epoch": 1.9, + "learning_rate": 3.415761497573377e-05, + "loss": 0.0003, + "step": 8226 + }, + { + "epoch": 1.9, + "learning_rate": 3.4155689084045915e-05, + "loss": 0.0, + "step": 8227 + }, + { + "epoch": 1.9, + "learning_rate": 3.415376319235806e-05, + "loss": 0.0002, + "step": 8228 + }, + { + "epoch": 1.9, + "learning_rate": 3.4151837300670216e-05, + "loss": 0.0, + "step": 8229 + }, + { + "epoch": 1.9, + "learning_rate": 3.414991140898236e-05, + "loss": 0.0318, + "step": 8230 + }, + { + "epoch": 1.9, + "learning_rate": 3.4147985517294505e-05, + "loss": 0.0029, + "step": 8231 + }, + { + "epoch": 1.9, + "learning_rate": 3.414605962560666e-05, + "loss": 0.0001, + "step": 8232 + }, + { + "epoch": 1.9, + "learning_rate": 3.4144133733918806e-05, + "loss": 0.0001, + "step": 8233 + }, + { + "epoch": 1.9, + "learning_rate": 3.4142207842230954e-05, + "loss": 0.0, + "step": 8234 + }, + { + "epoch": 1.9, + "learning_rate": 3.41402819505431e-05, + "loss": 0.0, + "step": 8235 + }, + { + "epoch": 1.9, + "learning_rate": 3.4138356058855256e-05, + "loss": 0.0002, + "step": 8236 + }, + { + "epoch": 1.9, + "learning_rate": 3.4136430167167396e-05, + "loss": 0.0, + "step": 8237 + }, + { + "epoch": 1.9, + "learning_rate": 3.413450427547955e-05, + "loss": 0.0001, + "step": 8238 + }, + { + "epoch": 1.9, + "learning_rate": 3.41325783837917e-05, + "loss": 0.0, + "step": 8239 + }, + { + "epoch": 1.9, + "learning_rate": 3.4130652492103845e-05, + "loss": 0.0, + "step": 8240 + }, + { + "epoch": 1.9, + "learning_rate": 3.412872660041599e-05, + "loss": 0.0002, + "step": 8241 + }, + { + "epoch": 1.9, + "learning_rate": 3.412680070872814e-05, + "loss": 0.0001, + "step": 8242 + }, + { + "epoch": 1.91, + "learning_rate": 3.4124874817040295e-05, + "loss": 0.0, + "step": 8243 + }, + { + "epoch": 1.91, + "learning_rate": 3.4122948925352435e-05, + "loss": 0.0, + "step": 8244 + }, + { + "epoch": 1.91, + "learning_rate": 3.412102303366459e-05, + "loss": 0.0, + "step": 8245 + }, + { + "epoch": 1.91, + "learning_rate": 3.411909714197674e-05, + "loss": 0.0003, + "step": 8246 + }, + { + "epoch": 1.91, + "learning_rate": 3.4117171250288885e-05, + "loss": 0.0001, + "step": 8247 + }, + { + "epoch": 1.91, + "learning_rate": 3.411524535860103e-05, + "loss": 0.0, + "step": 8248 + }, + { + "epoch": 1.91, + "learning_rate": 3.4113319466913186e-05, + "loss": 0.0002, + "step": 8249 + }, + { + "epoch": 1.91, + "learning_rate": 3.4111393575225334e-05, + "loss": 0.0001, + "step": 8250 + }, + { + "epoch": 1.91, + "learning_rate": 3.4109467683537474e-05, + "loss": 0.0, + "step": 8251 + }, + { + "epoch": 1.91, + "learning_rate": 3.410754179184963e-05, + "loss": 0.0007, + "step": 8252 + }, + { + "epoch": 1.91, + "learning_rate": 3.4105615900161776e-05, + "loss": 0.0, + "step": 8253 + }, + { + "epoch": 1.91, + "learning_rate": 3.4103690008473924e-05, + "loss": 0.0004, + "step": 8254 + }, + { + "epoch": 1.91, + "learning_rate": 3.410176411678607e-05, + "loss": 0.0003, + "step": 8255 + }, + { + "epoch": 1.91, + "learning_rate": 3.4099838225098225e-05, + "loss": 0.0004, + "step": 8256 + }, + { + "epoch": 1.91, + "learning_rate": 3.409791233341037e-05, + "loss": 0.0005, + "step": 8257 + }, + { + "epoch": 1.91, + "learning_rate": 3.409598644172252e-05, + "loss": 0.0001, + "step": 8258 + }, + { + "epoch": 1.91, + "learning_rate": 3.409406055003467e-05, + "loss": 0.0, + "step": 8259 + }, + { + "epoch": 1.91, + "learning_rate": 3.4092134658346815e-05, + "loss": 0.0, + "step": 8260 + }, + { + "epoch": 1.91, + "learning_rate": 3.409020876665896e-05, + "loss": 0.0001, + "step": 8261 + }, + { + "epoch": 1.91, + "learning_rate": 3.408828287497111e-05, + "loss": 0.0, + "step": 8262 + }, + { + "epoch": 1.91, + "learning_rate": 3.4086356983283265e-05, + "loss": 0.0001, + "step": 8263 + }, + { + "epoch": 1.91, + "learning_rate": 3.408443109159541e-05, + "loss": 0.0001, + "step": 8264 + }, + { + "epoch": 1.91, + "learning_rate": 3.408250519990756e-05, + "loss": 0.0001, + "step": 8265 + }, + { + "epoch": 1.91, + "learning_rate": 3.408057930821971e-05, + "loss": 0.0006, + "step": 8266 + }, + { + "epoch": 1.91, + "learning_rate": 3.407865341653186e-05, + "loss": 0.0001, + "step": 8267 + }, + { + "epoch": 1.91, + "learning_rate": 3.4076727524844e-05, + "loss": 0.0, + "step": 8268 + }, + { + "epoch": 1.91, + "learning_rate": 3.407480163315615e-05, + "loss": 0.0, + "step": 8269 + }, + { + "epoch": 1.91, + "learning_rate": 3.4072875741468304e-05, + "loss": 0.0, + "step": 8270 + }, + { + "epoch": 1.91, + "learning_rate": 3.407094984978045e-05, + "loss": 0.0, + "step": 8271 + }, + { + "epoch": 1.91, + "learning_rate": 3.40690239580926e-05, + "loss": 0.0, + "step": 8272 + }, + { + "epoch": 1.91, + "learning_rate": 3.4067098066404746e-05, + "loss": 0.0, + "step": 8273 + }, + { + "epoch": 1.91, + "learning_rate": 3.40651721747169e-05, + "loss": 0.0, + "step": 8274 + }, + { + "epoch": 1.91, + "learning_rate": 3.406324628302904e-05, + "loss": 0.0481, + "step": 8275 + }, + { + "epoch": 1.91, + "learning_rate": 3.4061320391341195e-05, + "loss": 0.0, + "step": 8276 + }, + { + "epoch": 1.91, + "learning_rate": 3.405939449965334e-05, + "loss": 0.0, + "step": 8277 + }, + { + "epoch": 1.91, + "learning_rate": 3.405746860796549e-05, + "loss": 0.0008, + "step": 8278 + }, + { + "epoch": 1.91, + "learning_rate": 3.405554271627764e-05, + "loss": 0.0005, + "step": 8279 + }, + { + "epoch": 1.91, + "learning_rate": 3.4053616824589785e-05, + "loss": 0.0001, + "step": 8280 + }, + { + "epoch": 1.91, + "learning_rate": 3.405169093290194e-05, + "loss": 0.0, + "step": 8281 + }, + { + "epoch": 1.91, + "learning_rate": 3.404976504121408e-05, + "loss": 0.0, + "step": 8282 + }, + { + "epoch": 1.91, + "learning_rate": 3.4047839149526234e-05, + "loss": 0.0, + "step": 8283 + }, + { + "epoch": 1.91, + "learning_rate": 3.404591325783838e-05, + "loss": 0.0009, + "step": 8284 + }, + { + "epoch": 1.91, + "learning_rate": 3.404398736615053e-05, + "loss": 0.0, + "step": 8285 + }, + { + "epoch": 1.91, + "learning_rate": 3.404206147446268e-05, + "loss": 0.0001, + "step": 8286 + }, + { + "epoch": 1.92, + "learning_rate": 3.4040135582774824e-05, + "loss": 0.0, + "step": 8287 + }, + { + "epoch": 1.92, + "learning_rate": 3.403820969108698e-05, + "loss": 0.0002, + "step": 8288 + }, + { + "epoch": 1.92, + "learning_rate": 3.403628379939912e-05, + "loss": 0.0001, + "step": 8289 + }, + { + "epoch": 1.92, + "learning_rate": 3.4034357907711274e-05, + "loss": 0.0, + "step": 8290 + }, + { + "epoch": 1.92, + "learning_rate": 3.403243201602342e-05, + "loss": 0.0008, + "step": 8291 + }, + { + "epoch": 1.92, + "learning_rate": 3.403050612433557e-05, + "loss": 0.0, + "step": 8292 + }, + { + "epoch": 1.92, + "learning_rate": 3.4028580232647716e-05, + "loss": 0.0007, + "step": 8293 + }, + { + "epoch": 1.92, + "learning_rate": 3.402665434095987e-05, + "loss": 0.003, + "step": 8294 + }, + { + "epoch": 1.92, + "learning_rate": 3.402472844927201e-05, + "loss": 0.0, + "step": 8295 + }, + { + "epoch": 1.92, + "learning_rate": 3.4022802557584165e-05, + "loss": 0.0072, + "step": 8296 + }, + { + "epoch": 1.92, + "learning_rate": 3.402087666589631e-05, + "loss": 0.0, + "step": 8297 + }, + { + "epoch": 1.92, + "learning_rate": 3.401895077420846e-05, + "loss": 0.0001, + "step": 8298 + }, + { + "epoch": 1.92, + "learning_rate": 3.401702488252061e-05, + "loss": 0.0, + "step": 8299 + }, + { + "epoch": 1.92, + "learning_rate": 3.4015098990832755e-05, + "loss": 0.0, + "step": 8300 + }, + { + "epoch": 1.92, + "learning_rate": 3.401317309914491e-05, + "loss": 0.0, + "step": 8301 + }, + { + "epoch": 1.92, + "learning_rate": 3.401124720745705e-05, + "loss": 0.0, + "step": 8302 + }, + { + "epoch": 1.92, + "learning_rate": 3.4009321315769204e-05, + "loss": 0.0, + "step": 8303 + }, + { + "epoch": 1.92, + "learning_rate": 3.400739542408135e-05, + "loss": 0.0001, + "step": 8304 + }, + { + "epoch": 1.92, + "learning_rate": 3.40054695323935e-05, + "loss": 0.0, + "step": 8305 + }, + { + "epoch": 1.92, + "learning_rate": 3.400354364070565e-05, + "loss": 0.0, + "step": 8306 + }, + { + "epoch": 1.92, + "learning_rate": 3.4001617749017794e-05, + "loss": 0.0004, + "step": 8307 + }, + { + "epoch": 1.92, + "learning_rate": 3.399969185732995e-05, + "loss": 0.0, + "step": 8308 + }, + { + "epoch": 1.92, + "learning_rate": 3.399776596564209e-05, + "loss": 0.0, + "step": 8309 + }, + { + "epoch": 1.92, + "learning_rate": 3.3995840073954243e-05, + "loss": 0.0001, + "step": 8310 + }, + { + "epoch": 1.92, + "learning_rate": 3.399391418226639e-05, + "loss": 0.0, + "step": 8311 + }, + { + "epoch": 1.92, + "learning_rate": 3.399198829057854e-05, + "loss": 0.0, + "step": 8312 + }, + { + "epoch": 1.92, + "learning_rate": 3.3990062398890686e-05, + "loss": 0.0, + "step": 8313 + }, + { + "epoch": 1.92, + "learning_rate": 3.398813650720284e-05, + "loss": 0.0, + "step": 8314 + }, + { + "epoch": 1.92, + "learning_rate": 3.398621061551499e-05, + "loss": 0.0005, + "step": 8315 + }, + { + "epoch": 1.92, + "learning_rate": 3.398428472382713e-05, + "loss": 0.0001, + "step": 8316 + }, + { + "epoch": 1.92, + "learning_rate": 3.398235883213928e-05, + "loss": 0.0, + "step": 8317 + }, + { + "epoch": 1.92, + "learning_rate": 3.398043294045143e-05, + "loss": 0.0, + "step": 8318 + }, + { + "epoch": 1.92, + "learning_rate": 3.397850704876358e-05, + "loss": 0.0, + "step": 8319 + }, + { + "epoch": 1.92, + "learning_rate": 3.3976581157075725e-05, + "loss": 0.0, + "step": 8320 + }, + { + "epoch": 1.92, + "learning_rate": 3.397465526538788e-05, + "loss": 0.0, + "step": 8321 + }, + { + "epoch": 1.92, + "learning_rate": 3.397272937370003e-05, + "loss": 0.0, + "step": 8322 + }, + { + "epoch": 1.92, + "learning_rate": 3.3970803482012174e-05, + "loss": 0.0001, + "step": 8323 + }, + { + "epoch": 1.92, + "learning_rate": 3.396887759032432e-05, + "loss": 0.0, + "step": 8324 + }, + { + "epoch": 1.92, + "learning_rate": 3.396695169863647e-05, + "loss": 0.0, + "step": 8325 + }, + { + "epoch": 1.92, + "learning_rate": 3.396502580694862e-05, + "loss": 0.0001, + "step": 8326 + }, + { + "epoch": 1.92, + "learning_rate": 3.3963099915260764e-05, + "loss": 0.0026, + "step": 8327 + }, + { + "epoch": 1.92, + "learning_rate": 3.396117402357292e-05, + "loss": 0.0, + "step": 8328 + }, + { + "epoch": 1.92, + "learning_rate": 3.3959248131885066e-05, + "loss": 0.0008, + "step": 8329 + }, + { + "epoch": 1.93, + "learning_rate": 3.395732224019721e-05, + "loss": 0.0031, + "step": 8330 + }, + { + "epoch": 1.93, + "learning_rate": 3.395539634850936e-05, + "loss": 0.0001, + "step": 8331 + }, + { + "epoch": 1.93, + "learning_rate": 3.3953470456821515e-05, + "loss": 0.0, + "step": 8332 + }, + { + "epoch": 1.93, + "learning_rate": 3.3951544565133656e-05, + "loss": 0.0, + "step": 8333 + }, + { + "epoch": 1.93, + "learning_rate": 3.39496186734458e-05, + "loss": 0.0, + "step": 8334 + }, + { + "epoch": 1.93, + "learning_rate": 3.394769278175796e-05, + "loss": 0.0006, + "step": 8335 + }, + { + "epoch": 1.93, + "learning_rate": 3.3945766890070105e-05, + "loss": 0.0, + "step": 8336 + }, + { + "epoch": 1.93, + "learning_rate": 3.394384099838225e-05, + "loss": 0.0, + "step": 8337 + }, + { + "epoch": 1.93, + "learning_rate": 3.39419151066944e-05, + "loss": 0.0001, + "step": 8338 + }, + { + "epoch": 1.93, + "learning_rate": 3.3939989215006554e-05, + "loss": 0.0001, + "step": 8339 + }, + { + "epoch": 1.93, + "learning_rate": 3.3938063323318695e-05, + "loss": 0.0, + "step": 8340 + }, + { + "epoch": 1.93, + "learning_rate": 3.393613743163085e-05, + "loss": 0.0011, + "step": 8341 + }, + { + "epoch": 1.93, + "learning_rate": 3.3934211539942997e-05, + "loss": 0.0001, + "step": 8342 + }, + { + "epoch": 1.93, + "learning_rate": 3.3932285648255144e-05, + "loss": 0.0002, + "step": 8343 + }, + { + "epoch": 1.93, + "learning_rate": 3.393035975656729e-05, + "loss": 0.0, + "step": 8344 + }, + { + "epoch": 1.93, + "learning_rate": 3.392843386487944e-05, + "loss": 0.0, + "step": 8345 + }, + { + "epoch": 1.93, + "learning_rate": 3.392650797319159e-05, + "loss": 0.0002, + "step": 8346 + }, + { + "epoch": 1.93, + "learning_rate": 3.3924582081503734e-05, + "loss": 0.0, + "step": 8347 + }, + { + "epoch": 1.93, + "learning_rate": 3.392265618981589e-05, + "loss": 0.0, + "step": 8348 + }, + { + "epoch": 1.93, + "learning_rate": 3.3920730298128036e-05, + "loss": 0.0, + "step": 8349 + }, + { + "epoch": 1.93, + "learning_rate": 3.391880440644018e-05, + "loss": 0.0001, + "step": 8350 + }, + { + "epoch": 1.93, + "learning_rate": 3.391687851475233e-05, + "loss": 0.0002, + "step": 8351 + }, + { + "epoch": 1.93, + "learning_rate": 3.3914952623064485e-05, + "loss": 0.0, + "step": 8352 + }, + { + "epoch": 1.93, + "learning_rate": 3.391302673137663e-05, + "loss": 0.0, + "step": 8353 + }, + { + "epoch": 1.93, + "learning_rate": 3.391110083968877e-05, + "loss": 0.0, + "step": 8354 + }, + { + "epoch": 1.93, + "learning_rate": 3.390917494800093e-05, + "loss": 0.0, + "step": 8355 + }, + { + "epoch": 1.93, + "learning_rate": 3.3907249056313075e-05, + "loss": 0.0, + "step": 8356 + }, + { + "epoch": 1.93, + "learning_rate": 3.390532316462522e-05, + "loss": 0.0001, + "step": 8357 + }, + { + "epoch": 1.93, + "learning_rate": 3.390339727293737e-05, + "loss": 0.0001, + "step": 8358 + }, + { + "epoch": 1.93, + "learning_rate": 3.3901471381249524e-05, + "loss": 0.0, + "step": 8359 + }, + { + "epoch": 1.93, + "learning_rate": 3.389954548956167e-05, + "loss": 0.0, + "step": 8360 + }, + { + "epoch": 1.93, + "learning_rate": 3.389761959787382e-05, + "loss": 0.0002, + "step": 8361 + }, + { + "epoch": 1.93, + "learning_rate": 3.3895693706185966e-05, + "loss": 0.0, + "step": 8362 + }, + { + "epoch": 1.93, + "learning_rate": 3.3893767814498114e-05, + "loss": 0.0, + "step": 8363 + }, + { + "epoch": 1.93, + "learning_rate": 3.389184192281026e-05, + "loss": 0.0, + "step": 8364 + }, + { + "epoch": 1.93, + "learning_rate": 3.388991603112241e-05, + "loss": 0.0001, + "step": 8365 + }, + { + "epoch": 1.93, + "learning_rate": 3.388799013943456e-05, + "loss": 0.0003, + "step": 8366 + }, + { + "epoch": 1.93, + "learning_rate": 3.3886064247746704e-05, + "loss": 0.0, + "step": 8367 + }, + { + "epoch": 1.93, + "learning_rate": 3.388413835605886e-05, + "loss": 0.0, + "step": 8368 + }, + { + "epoch": 1.93, + "learning_rate": 3.3882212464371006e-05, + "loss": 0.0, + "step": 8369 + }, + { + "epoch": 1.93, + "learning_rate": 3.388028657268315e-05, + "loss": 0.0001, + "step": 8370 + }, + { + "epoch": 1.93, + "learning_rate": 3.38783606809953e-05, + "loss": 0.0001, + "step": 8371 + }, + { + "epoch": 1.93, + "learning_rate": 3.387643478930745e-05, + "loss": 0.0, + "step": 8372 + }, + { + "epoch": 1.94, + "learning_rate": 3.38745088976196e-05, + "loss": 0.0, + "step": 8373 + }, + { + "epoch": 1.94, + "learning_rate": 3.387258300593174e-05, + "loss": 0.0033, + "step": 8374 + }, + { + "epoch": 1.94, + "learning_rate": 3.38706571142439e-05, + "loss": 0.0, + "step": 8375 + }, + { + "epoch": 1.94, + "learning_rate": 3.3868731222556045e-05, + "loss": 0.0, + "step": 8376 + }, + { + "epoch": 1.94, + "learning_rate": 3.386680533086819e-05, + "loss": 0.0, + "step": 8377 + }, + { + "epoch": 1.94, + "learning_rate": 3.386487943918034e-05, + "loss": 0.0, + "step": 8378 + }, + { + "epoch": 1.94, + "learning_rate": 3.3862953547492494e-05, + "loss": 0.0001, + "step": 8379 + }, + { + "epoch": 1.94, + "learning_rate": 3.386102765580464e-05, + "loss": 0.0004, + "step": 8380 + }, + { + "epoch": 1.94, + "learning_rate": 3.385910176411678e-05, + "loss": 0.0, + "step": 8381 + }, + { + "epoch": 1.94, + "learning_rate": 3.3857175872428936e-05, + "loss": 0.0002, + "step": 8382 + }, + { + "epoch": 1.94, + "learning_rate": 3.3855249980741084e-05, + "loss": 0.0002, + "step": 8383 + }, + { + "epoch": 1.94, + "learning_rate": 3.385332408905323e-05, + "loss": 0.0001, + "step": 8384 + }, + { + "epoch": 1.94, + "learning_rate": 3.385139819736538e-05, + "loss": 0.0, + "step": 8385 + }, + { + "epoch": 1.94, + "learning_rate": 3.384947230567753e-05, + "loss": 0.0, + "step": 8386 + }, + { + "epoch": 1.94, + "learning_rate": 3.384754641398968e-05, + "loss": 0.0, + "step": 8387 + }, + { + "epoch": 1.94, + "learning_rate": 3.384562052230183e-05, + "loss": 0.0002, + "step": 8388 + }, + { + "epoch": 1.94, + "learning_rate": 3.3843694630613975e-05, + "loss": 0.0, + "step": 8389 + }, + { + "epoch": 1.94, + "learning_rate": 3.384176873892612e-05, + "loss": 0.0, + "step": 8390 + }, + { + "epoch": 1.94, + "learning_rate": 3.383984284723827e-05, + "loss": 0.0006, + "step": 8391 + }, + { + "epoch": 1.94, + "learning_rate": 3.383791695555042e-05, + "loss": 0.0002, + "step": 8392 + }, + { + "epoch": 1.94, + "learning_rate": 3.383599106386257e-05, + "loss": 0.0002, + "step": 8393 + }, + { + "epoch": 1.94, + "learning_rate": 3.383406517217472e-05, + "loss": 0.0013, + "step": 8394 + }, + { + "epoch": 1.94, + "learning_rate": 3.383213928048687e-05, + "loss": 0.0002, + "step": 8395 + }, + { + "epoch": 1.94, + "learning_rate": 3.3830213388799015e-05, + "loss": 0.0, + "step": 8396 + }, + { + "epoch": 1.94, + "learning_rate": 3.382828749711117e-05, + "loss": 0.0, + "step": 8397 + }, + { + "epoch": 1.94, + "learning_rate": 3.382636160542331e-05, + "loss": 0.0001, + "step": 8398 + }, + { + "epoch": 1.94, + "learning_rate": 3.382443571373546e-05, + "loss": 0.0001, + "step": 8399 + }, + { + "epoch": 1.94, + "learning_rate": 3.382250982204761e-05, + "loss": 0.0, + "step": 8400 + }, + { + "epoch": 1.94, + "learning_rate": 3.382058393035976e-05, + "loss": 0.0001, + "step": 8401 + }, + { + "epoch": 1.94, + "learning_rate": 3.3818658038671906e-05, + "loss": 0.0, + "step": 8402 + }, + { + "epoch": 1.94, + "learning_rate": 3.3816732146984054e-05, + "loss": 0.0, + "step": 8403 + }, + { + "epoch": 1.94, + "learning_rate": 3.381480625529621e-05, + "loss": 0.0, + "step": 8404 + }, + { + "epoch": 1.94, + "learning_rate": 3.381288036360835e-05, + "loss": 0.0, + "step": 8405 + }, + { + "epoch": 1.94, + "learning_rate": 3.38109544719205e-05, + "loss": 0.0001, + "step": 8406 + }, + { + "epoch": 1.94, + "learning_rate": 3.380902858023265e-05, + "loss": 0.0, + "step": 8407 + }, + { + "epoch": 1.94, + "learning_rate": 3.38071026885448e-05, + "loss": 0.0002, + "step": 8408 + }, + { + "epoch": 1.94, + "learning_rate": 3.3805176796856945e-05, + "loss": 0.0, + "step": 8409 + }, + { + "epoch": 1.94, + "learning_rate": 3.380325090516909e-05, + "loss": 0.0001, + "step": 8410 + }, + { + "epoch": 1.94, + "learning_rate": 3.380132501348125e-05, + "loss": 0.0, + "step": 8411 + }, + { + "epoch": 1.94, + "learning_rate": 3.379939912179339e-05, + "loss": 0.0, + "step": 8412 + }, + { + "epoch": 1.94, + "learning_rate": 3.379747323010554e-05, + "loss": 0.0, + "step": 8413 + }, + { + "epoch": 1.94, + "learning_rate": 3.379554733841769e-05, + "loss": 0.0001, + "step": 8414 + }, + { + "epoch": 1.94, + "learning_rate": 3.379362144672984e-05, + "loss": 0.0013, + "step": 8415 + }, + { + "epoch": 1.94, + "learning_rate": 3.3791695555041984e-05, + "loss": 0.0, + "step": 8416 + }, + { + "epoch": 1.95, + "learning_rate": 3.378976966335414e-05, + "loss": 0.0001, + "step": 8417 + }, + { + "epoch": 1.95, + "learning_rate": 3.3787843771666286e-05, + "loss": 0.0022, + "step": 8418 + }, + { + "epoch": 1.95, + "learning_rate": 3.378591787997843e-05, + "loss": 0.0, + "step": 8419 + }, + { + "epoch": 1.95, + "learning_rate": 3.378399198829058e-05, + "loss": 0.0001, + "step": 8420 + }, + { + "epoch": 1.95, + "learning_rate": 3.378206609660273e-05, + "loss": 0.0, + "step": 8421 + }, + { + "epoch": 1.95, + "learning_rate": 3.3780140204914876e-05, + "loss": 0.0, + "step": 8422 + }, + { + "epoch": 1.95, + "learning_rate": 3.3778214313227024e-05, + "loss": 0.0, + "step": 8423 + }, + { + "epoch": 1.95, + "learning_rate": 3.377628842153918e-05, + "loss": 0.0, + "step": 8424 + }, + { + "epoch": 1.95, + "learning_rate": 3.3774362529851325e-05, + "loss": 0.0, + "step": 8425 + }, + { + "epoch": 1.95, + "learning_rate": 3.377243663816347e-05, + "loss": 0.0001, + "step": 8426 + }, + { + "epoch": 1.95, + "learning_rate": 3.377051074647562e-05, + "loss": 0.0, + "step": 8427 + }, + { + "epoch": 1.95, + "learning_rate": 3.376858485478777e-05, + "loss": 0.0, + "step": 8428 + }, + { + "epoch": 1.95, + "learning_rate": 3.3766658963099915e-05, + "loss": 0.0, + "step": 8429 + }, + { + "epoch": 1.95, + "learning_rate": 3.376473307141206e-05, + "loss": 0.0, + "step": 8430 + }, + { + "epoch": 1.95, + "learning_rate": 3.376280717972422e-05, + "loss": 0.0, + "step": 8431 + }, + { + "epoch": 1.95, + "learning_rate": 3.3760881288036364e-05, + "loss": 0.0, + "step": 8432 + }, + { + "epoch": 1.95, + "learning_rate": 3.375895539634851e-05, + "loss": 0.0001, + "step": 8433 + }, + { + "epoch": 1.95, + "learning_rate": 3.375702950466066e-05, + "loss": 0.0, + "step": 8434 + }, + { + "epoch": 1.95, + "learning_rate": 3.3755103612972814e-05, + "loss": 0.0, + "step": 8435 + }, + { + "epoch": 1.95, + "learning_rate": 3.3753177721284954e-05, + "loss": 0.0, + "step": 8436 + }, + { + "epoch": 1.95, + "learning_rate": 3.37512518295971e-05, + "loss": 0.0, + "step": 8437 + }, + { + "epoch": 1.95, + "learning_rate": 3.3749325937909256e-05, + "loss": 0.0007, + "step": 8438 + }, + { + "epoch": 1.95, + "learning_rate": 3.37474000462214e-05, + "loss": 0.0, + "step": 8439 + }, + { + "epoch": 1.95, + "learning_rate": 3.374547415453355e-05, + "loss": 0.0, + "step": 8440 + }, + { + "epoch": 1.95, + "learning_rate": 3.37435482628457e-05, + "loss": 0.0, + "step": 8441 + }, + { + "epoch": 1.95, + "learning_rate": 3.3741622371157846e-05, + "loss": 0.0, + "step": 8442 + }, + { + "epoch": 1.95, + "learning_rate": 3.3739696479469994e-05, + "loss": 0.0001, + "step": 8443 + }, + { + "epoch": 1.95, + "learning_rate": 3.373777058778215e-05, + "loss": 0.0, + "step": 8444 + }, + { + "epoch": 1.95, + "learning_rate": 3.3735844696094295e-05, + "loss": 0.0, + "step": 8445 + }, + { + "epoch": 1.95, + "learning_rate": 3.3733918804406436e-05, + "loss": 0.0001, + "step": 8446 + }, + { + "epoch": 1.95, + "learning_rate": 3.373199291271859e-05, + "loss": 0.0001, + "step": 8447 + }, + { + "epoch": 1.95, + "learning_rate": 3.373006702103074e-05, + "loss": 0.0002, + "step": 8448 + }, + { + "epoch": 1.95, + "learning_rate": 3.3728141129342885e-05, + "loss": 0.0001, + "step": 8449 + }, + { + "epoch": 1.95, + "learning_rate": 3.372621523765503e-05, + "loss": 0.0002, + "step": 8450 + }, + { + "epoch": 1.95, + "learning_rate": 3.372428934596719e-05, + "loss": 0.0, + "step": 8451 + }, + { + "epoch": 1.95, + "learning_rate": 3.3722363454279334e-05, + "loss": 0.0003, + "step": 8452 + }, + { + "epoch": 1.95, + "learning_rate": 3.372043756259148e-05, + "loss": 0.0001, + "step": 8453 + }, + { + "epoch": 1.95, + "learning_rate": 3.371851167090363e-05, + "loss": 0.0, + "step": 8454 + }, + { + "epoch": 1.95, + "learning_rate": 3.371658577921578e-05, + "loss": 0.0, + "step": 8455 + }, + { + "epoch": 1.95, + "learning_rate": 3.3714659887527924e-05, + "loss": 0.0004, + "step": 8456 + }, + { + "epoch": 1.95, + "learning_rate": 3.371273399584007e-05, + "loss": 0.0007, + "step": 8457 + }, + { + "epoch": 1.95, + "learning_rate": 3.3710808104152226e-05, + "loss": 0.0, + "step": 8458 + }, + { + "epoch": 1.95, + "learning_rate": 3.3708882212464373e-05, + "loss": 0.0002, + "step": 8459 + }, + { + "epoch": 1.96, + "learning_rate": 3.370695632077652e-05, + "loss": 0.0007, + "step": 8460 + }, + { + "epoch": 1.96, + "learning_rate": 3.370503042908867e-05, + "loss": 0.0, + "step": 8461 + }, + { + "epoch": 1.96, + "learning_rate": 3.370310453740082e-05, + "loss": 0.0, + "step": 8462 + }, + { + "epoch": 1.96, + "learning_rate": 3.370117864571296e-05, + "loss": 0.0, + "step": 8463 + }, + { + "epoch": 1.96, + "learning_rate": 3.369925275402512e-05, + "loss": 0.0, + "step": 8464 + }, + { + "epoch": 1.96, + "learning_rate": 3.3697326862337265e-05, + "loss": 0.0, + "step": 8465 + }, + { + "epoch": 1.96, + "learning_rate": 3.369540097064941e-05, + "loss": 0.0004, + "step": 8466 + }, + { + "epoch": 1.96, + "learning_rate": 3.369347507896156e-05, + "loss": 0.0001, + "step": 8467 + }, + { + "epoch": 1.96, + "learning_rate": 3.369154918727371e-05, + "loss": 0.0, + "step": 8468 + }, + { + "epoch": 1.96, + "learning_rate": 3.368962329558586e-05, + "loss": 0.0, + "step": 8469 + }, + { + "epoch": 1.96, + "learning_rate": 3.3687697403898e-05, + "loss": 0.0, + "step": 8470 + }, + { + "epoch": 1.96, + "learning_rate": 3.368577151221016e-05, + "loss": 0.0, + "step": 8471 + }, + { + "epoch": 1.96, + "learning_rate": 3.3683845620522304e-05, + "loss": 0.0, + "step": 8472 + }, + { + "epoch": 1.96, + "learning_rate": 3.368191972883445e-05, + "loss": 0.0002, + "step": 8473 + }, + { + "epoch": 1.96, + "learning_rate": 3.36799938371466e-05, + "loss": 0.0001, + "step": 8474 + }, + { + "epoch": 1.96, + "learning_rate": 3.367806794545875e-05, + "loss": 0.0001, + "step": 8475 + }, + { + "epoch": 1.96, + "learning_rate": 3.36761420537709e-05, + "loss": 0.0, + "step": 8476 + }, + { + "epoch": 1.96, + "learning_rate": 3.367421616208304e-05, + "loss": 0.0001, + "step": 8477 + }, + { + "epoch": 1.96, + "learning_rate": 3.3672290270395196e-05, + "loss": 0.0, + "step": 8478 + }, + { + "epoch": 1.96, + "learning_rate": 3.367036437870734e-05, + "loss": 0.0001, + "step": 8479 + }, + { + "epoch": 1.96, + "learning_rate": 3.366843848701949e-05, + "loss": 0.0, + "step": 8480 + }, + { + "epoch": 1.96, + "learning_rate": 3.366651259533164e-05, + "loss": 0.0002, + "step": 8481 + }, + { + "epoch": 1.96, + "learning_rate": 3.366458670364379e-05, + "loss": 0.0, + "step": 8482 + }, + { + "epoch": 1.96, + "learning_rate": 3.366266081195594e-05, + "loss": 0.0001, + "step": 8483 + }, + { + "epoch": 1.96, + "learning_rate": 3.366073492026808e-05, + "loss": 0.0, + "step": 8484 + }, + { + "epoch": 1.96, + "learning_rate": 3.3658809028580235e-05, + "loss": 0.0, + "step": 8485 + }, + { + "epoch": 1.96, + "learning_rate": 3.365688313689238e-05, + "loss": 0.0001, + "step": 8486 + }, + { + "epoch": 1.96, + "learning_rate": 3.365495724520453e-05, + "loss": 0.0, + "step": 8487 + }, + { + "epoch": 1.96, + "learning_rate": 3.365303135351668e-05, + "loss": 0.0, + "step": 8488 + }, + { + "epoch": 1.96, + "learning_rate": 3.365110546182883e-05, + "loss": 0.0, + "step": 8489 + }, + { + "epoch": 1.96, + "learning_rate": 3.364917957014098e-05, + "loss": 0.0, + "step": 8490 + }, + { + "epoch": 1.96, + "learning_rate": 3.364725367845313e-05, + "loss": 0.0012, + "step": 8491 + }, + { + "epoch": 1.96, + "learning_rate": 3.3645327786765274e-05, + "loss": 0.0, + "step": 8492 + }, + { + "epoch": 1.96, + "learning_rate": 3.364340189507742e-05, + "loss": 0.0, + "step": 8493 + }, + { + "epoch": 1.96, + "learning_rate": 3.364147600338957e-05, + "loss": 0.0, + "step": 8494 + }, + { + "epoch": 1.96, + "learning_rate": 3.3639550111701717e-05, + "loss": 0.0, + "step": 8495 + }, + { + "epoch": 1.96, + "learning_rate": 3.363762422001387e-05, + "loss": 0.0001, + "step": 8496 + }, + { + "epoch": 1.96, + "learning_rate": 3.363569832832602e-05, + "loss": 0.0001, + "step": 8497 + }, + { + "epoch": 1.96, + "learning_rate": 3.3633772436638166e-05, + "loss": 0.0, + "step": 8498 + }, + { + "epoch": 1.96, + "learning_rate": 3.363184654495031e-05, + "loss": 0.0, + "step": 8499 + }, + { + "epoch": 1.96, + "learning_rate": 3.362992065326247e-05, + "loss": 0.0006, + "step": 8500 + }, + { + "epoch": 1.96, + "learning_rate": 3.362799476157461e-05, + "loss": 0.0001, + "step": 8501 + }, + { + "epoch": 1.96, + "learning_rate": 3.3626068869886756e-05, + "loss": 0.0, + "step": 8502 + }, + { + "epoch": 1.97, + "learning_rate": 3.362414297819891e-05, + "loss": 0.0001, + "step": 8503 + }, + { + "epoch": 1.97, + "learning_rate": 3.362221708651106e-05, + "loss": 0.0, + "step": 8504 + }, + { + "epoch": 1.97, + "learning_rate": 3.3620291194823205e-05, + "loss": 0.0001, + "step": 8505 + }, + { + "epoch": 1.97, + "learning_rate": 3.361836530313535e-05, + "loss": 0.0, + "step": 8506 + }, + { + "epoch": 1.97, + "learning_rate": 3.361643941144751e-05, + "loss": 0.0, + "step": 8507 + }, + { + "epoch": 1.97, + "learning_rate": 3.361451351975965e-05, + "loss": 0.0, + "step": 8508 + }, + { + "epoch": 1.97, + "learning_rate": 3.36125876280718e-05, + "loss": 0.0, + "step": 8509 + }, + { + "epoch": 1.97, + "learning_rate": 3.361066173638395e-05, + "loss": 0.0001, + "step": 8510 + }, + { + "epoch": 1.97, + "learning_rate": 3.3608735844696097e-05, + "loss": 0.0, + "step": 8511 + }, + { + "epoch": 1.97, + "learning_rate": 3.3606809953008244e-05, + "loss": 0.0, + "step": 8512 + }, + { + "epoch": 1.97, + "learning_rate": 3.360488406132039e-05, + "loss": 0.0002, + "step": 8513 + }, + { + "epoch": 1.97, + "learning_rate": 3.3602958169632546e-05, + "loss": 0.0002, + "step": 8514 + }, + { + "epoch": 1.97, + "learning_rate": 3.3601032277944686e-05, + "loss": 0.0, + "step": 8515 + }, + { + "epoch": 1.97, + "learning_rate": 3.359910638625684e-05, + "loss": 0.0, + "step": 8516 + }, + { + "epoch": 1.97, + "learning_rate": 3.359718049456899e-05, + "loss": 0.0, + "step": 8517 + }, + { + "epoch": 1.97, + "learning_rate": 3.3595254602881136e-05, + "loss": 0.0, + "step": 8518 + }, + { + "epoch": 1.97, + "learning_rate": 3.359332871119328e-05, + "loss": 0.0, + "step": 8519 + }, + { + "epoch": 1.97, + "learning_rate": 3.359140281950544e-05, + "loss": 0.0007, + "step": 8520 + }, + { + "epoch": 1.97, + "learning_rate": 3.358947692781758e-05, + "loss": 0.0001, + "step": 8521 + }, + { + "epoch": 1.97, + "learning_rate": 3.3587551036129726e-05, + "loss": 0.0, + "step": 8522 + }, + { + "epoch": 1.97, + "learning_rate": 3.358562514444188e-05, + "loss": 0.0, + "step": 8523 + }, + { + "epoch": 1.97, + "learning_rate": 3.358369925275403e-05, + "loss": 0.0, + "step": 8524 + }, + { + "epoch": 1.97, + "learning_rate": 3.3581773361066175e-05, + "loss": 0.0001, + "step": 8525 + }, + { + "epoch": 1.97, + "learning_rate": 3.357984746937832e-05, + "loss": 0.0001, + "step": 8526 + }, + { + "epoch": 1.97, + "learning_rate": 3.3577921577690477e-05, + "loss": 0.0001, + "step": 8527 + }, + { + "epoch": 1.97, + "learning_rate": 3.357599568600262e-05, + "loss": 0.0, + "step": 8528 + }, + { + "epoch": 1.97, + "learning_rate": 3.357406979431477e-05, + "loss": 0.0, + "step": 8529 + }, + { + "epoch": 1.97, + "learning_rate": 3.357214390262692e-05, + "loss": 0.0001, + "step": 8530 + }, + { + "epoch": 1.97, + "learning_rate": 3.3570218010939066e-05, + "loss": 0.0, + "step": 8531 + }, + { + "epoch": 1.97, + "learning_rate": 3.3568292119251214e-05, + "loss": 0.0002, + "step": 8532 + }, + { + "epoch": 1.97, + "learning_rate": 3.356636622756336e-05, + "loss": 0.0001, + "step": 8533 + }, + { + "epoch": 1.97, + "learning_rate": 3.3564440335875516e-05, + "loss": 0.0, + "step": 8534 + }, + { + "epoch": 1.97, + "learning_rate": 3.3562514444187656e-05, + "loss": 0.004, + "step": 8535 + }, + { + "epoch": 1.97, + "learning_rate": 3.356058855249981e-05, + "loss": 0.0, + "step": 8536 + }, + { + "epoch": 1.97, + "learning_rate": 3.355866266081196e-05, + "loss": 0.0, + "step": 8537 + }, + { + "epoch": 1.97, + "learning_rate": 3.3556736769124106e-05, + "loss": 0.0, + "step": 8538 + }, + { + "epoch": 1.97, + "learning_rate": 3.355481087743625e-05, + "loss": 0.0001, + "step": 8539 + }, + { + "epoch": 1.97, + "learning_rate": 3.35528849857484e-05, + "loss": 0.0, + "step": 8540 + }, + { + "epoch": 1.97, + "learning_rate": 3.3550959094060555e-05, + "loss": 0.0017, + "step": 8541 + }, + { + "epoch": 1.97, + "learning_rate": 3.3549033202372695e-05, + "loss": 0.0, + "step": 8542 + }, + { + "epoch": 1.97, + "learning_rate": 3.354710731068485e-05, + "loss": 0.0, + "step": 8543 + }, + { + "epoch": 1.97, + "learning_rate": 3.3545181418997e-05, + "loss": 0.0001, + "step": 8544 + }, + { + "epoch": 1.97, + "learning_rate": 3.3543255527309145e-05, + "loss": 0.0001, + "step": 8545 + }, + { + "epoch": 1.98, + "learning_rate": 3.354132963562129e-05, + "loss": 0.0172, + "step": 8546 + }, + { + "epoch": 1.98, + "learning_rate": 3.3539403743933446e-05, + "loss": 0.0001, + "step": 8547 + }, + { + "epoch": 1.98, + "learning_rate": 3.3537477852245594e-05, + "loss": 0.0, + "step": 8548 + }, + { + "epoch": 1.98, + "learning_rate": 3.3535551960557735e-05, + "loss": 0.0, + "step": 8549 + }, + { + "epoch": 1.98, + "learning_rate": 3.353362606886989e-05, + "loss": 0.0, + "step": 8550 + }, + { + "epoch": 1.98, + "learning_rate": 3.3531700177182036e-05, + "loss": 0.0, + "step": 8551 + }, + { + "epoch": 1.98, + "learning_rate": 3.3529774285494184e-05, + "loss": 0.0001, + "step": 8552 + }, + { + "epoch": 1.98, + "learning_rate": 3.352784839380633e-05, + "loss": 0.0, + "step": 8553 + }, + { + "epoch": 1.98, + "learning_rate": 3.3525922502118486e-05, + "loss": 0.0, + "step": 8554 + }, + { + "epoch": 1.98, + "learning_rate": 3.352399661043063e-05, + "loss": 0.0001, + "step": 8555 + }, + { + "epoch": 1.98, + "learning_rate": 3.352207071874278e-05, + "loss": 0.0, + "step": 8556 + }, + { + "epoch": 1.98, + "learning_rate": 3.352014482705493e-05, + "loss": 0.0, + "step": 8557 + }, + { + "epoch": 1.98, + "learning_rate": 3.3518218935367075e-05, + "loss": 0.0001, + "step": 8558 + }, + { + "epoch": 1.98, + "learning_rate": 3.351629304367922e-05, + "loss": 0.0, + "step": 8559 + }, + { + "epoch": 1.98, + "learning_rate": 3.351436715199137e-05, + "loss": 0.0, + "step": 8560 + }, + { + "epoch": 1.98, + "learning_rate": 3.3512441260303525e-05, + "loss": 0.001, + "step": 8561 + }, + { + "epoch": 1.98, + "learning_rate": 3.351051536861567e-05, + "loss": 0.0, + "step": 8562 + }, + { + "epoch": 1.98, + "learning_rate": 3.350858947692782e-05, + "loss": 0.0, + "step": 8563 + }, + { + "epoch": 1.98, + "learning_rate": 3.350666358523997e-05, + "loss": 0.0, + "step": 8564 + }, + { + "epoch": 1.98, + "learning_rate": 3.350473769355212e-05, + "loss": 0.0, + "step": 8565 + }, + { + "epoch": 1.98, + "learning_rate": 3.350281180186426e-05, + "loss": 0.0, + "step": 8566 + }, + { + "epoch": 1.98, + "learning_rate": 3.350088591017641e-05, + "loss": 0.0, + "step": 8567 + }, + { + "epoch": 1.98, + "learning_rate": 3.3498960018488564e-05, + "loss": 0.0001, + "step": 8568 + }, + { + "epoch": 1.98, + "learning_rate": 3.349703412680071e-05, + "loss": 0.0, + "step": 8569 + }, + { + "epoch": 1.98, + "learning_rate": 3.349510823511286e-05, + "loss": 0.0, + "step": 8570 + }, + { + "epoch": 1.98, + "learning_rate": 3.3493182343425006e-05, + "loss": 0.0, + "step": 8571 + }, + { + "epoch": 1.98, + "learning_rate": 3.349125645173716e-05, + "loss": 0.0003, + "step": 8572 + }, + { + "epoch": 1.98, + "learning_rate": 3.34893305600493e-05, + "loss": 0.0002, + "step": 8573 + }, + { + "epoch": 1.98, + "learning_rate": 3.3487404668361455e-05, + "loss": 0.0002, + "step": 8574 + }, + { + "epoch": 1.98, + "learning_rate": 3.34854787766736e-05, + "loss": 0.0012, + "step": 8575 + }, + { + "epoch": 1.98, + "learning_rate": 3.348355288498575e-05, + "loss": 0.0, + "step": 8576 + }, + { + "epoch": 1.98, + "learning_rate": 3.34816269932979e-05, + "loss": 0.0, + "step": 8577 + }, + { + "epoch": 1.98, + "learning_rate": 3.3479701101610045e-05, + "loss": 0.0021, + "step": 8578 + }, + { + "epoch": 1.98, + "learning_rate": 3.34777752099222e-05, + "loss": 0.0004, + "step": 8579 + }, + { + "epoch": 1.98, + "learning_rate": 3.347584931823434e-05, + "loss": 0.0, + "step": 8580 + }, + { + "epoch": 1.98, + "learning_rate": 3.3473923426546495e-05, + "loss": 0.0, + "step": 8581 + }, + { + "epoch": 1.98, + "learning_rate": 3.347199753485864e-05, + "loss": 0.0001, + "step": 8582 + }, + { + "epoch": 1.98, + "learning_rate": 3.347007164317079e-05, + "loss": 0.0004, + "step": 8583 + }, + { + "epoch": 1.98, + "learning_rate": 3.346814575148294e-05, + "loss": 0.0001, + "step": 8584 + }, + { + "epoch": 1.98, + "learning_rate": 3.346621985979509e-05, + "loss": 0.0001, + "step": 8585 + }, + { + "epoch": 1.98, + "learning_rate": 3.346429396810724e-05, + "loss": 0.0, + "step": 8586 + }, + { + "epoch": 1.98, + "learning_rate": 3.346236807641938e-05, + "loss": 0.0, + "step": 8587 + }, + { + "epoch": 1.98, + "learning_rate": 3.3460442184731534e-05, + "loss": 0.0009, + "step": 8588 + }, + { + "epoch": 1.98, + "learning_rate": 3.345851629304368e-05, + "loss": 0.0002, + "step": 8589 + }, + { + "epoch": 1.99, + "learning_rate": 3.345659040135583e-05, + "loss": 0.0, + "step": 8590 + }, + { + "epoch": 1.99, + "learning_rate": 3.3454664509667976e-05, + "loss": 0.0, + "step": 8591 + }, + { + "epoch": 1.99, + "learning_rate": 3.345273861798013e-05, + "loss": 0.0001, + "step": 8592 + }, + { + "epoch": 1.99, + "learning_rate": 3.345081272629227e-05, + "loss": 0.0, + "step": 8593 + }, + { + "epoch": 1.99, + "learning_rate": 3.3448886834604425e-05, + "loss": 0.0, + "step": 8594 + }, + { + "epoch": 1.99, + "learning_rate": 3.344696094291657e-05, + "loss": 0.0001, + "step": 8595 + }, + { + "epoch": 1.99, + "learning_rate": 3.344503505122872e-05, + "loss": 0.0, + "step": 8596 + }, + { + "epoch": 1.99, + "learning_rate": 3.344310915954087e-05, + "loss": 0.0002, + "step": 8597 + }, + { + "epoch": 1.99, + "learning_rate": 3.3441183267853015e-05, + "loss": 0.0001, + "step": 8598 + }, + { + "epoch": 1.99, + "learning_rate": 3.343925737616517e-05, + "loss": 0.0001, + "step": 8599 + }, + { + "epoch": 1.99, + "learning_rate": 3.343733148447731e-05, + "loss": 0.0006, + "step": 8600 + }, + { + "epoch": 1.99, + "learning_rate": 3.3435405592789464e-05, + "loss": 0.0001, + "step": 8601 + }, + { + "epoch": 1.99, + "learning_rate": 3.343347970110161e-05, + "loss": 0.0, + "step": 8602 + }, + { + "epoch": 1.99, + "learning_rate": 3.343155380941376e-05, + "loss": 0.0004, + "step": 8603 + }, + { + "epoch": 1.99, + "learning_rate": 3.342962791772591e-05, + "loss": 0.0, + "step": 8604 + }, + { + "epoch": 1.99, + "learning_rate": 3.3427702026038054e-05, + "loss": 0.0001, + "step": 8605 + }, + { + "epoch": 1.99, + "learning_rate": 3.342577613435021e-05, + "loss": 0.0, + "step": 8606 + }, + { + "epoch": 1.99, + "learning_rate": 3.342385024266235e-05, + "loss": 0.0, + "step": 8607 + }, + { + "epoch": 1.99, + "learning_rate": 3.3421924350974504e-05, + "loss": 0.0003, + "step": 8608 + }, + { + "epoch": 1.99, + "learning_rate": 3.341999845928665e-05, + "loss": 0.0, + "step": 8609 + }, + { + "epoch": 1.99, + "learning_rate": 3.34180725675988e-05, + "loss": 0.0, + "step": 8610 + }, + { + "epoch": 1.99, + "learning_rate": 3.3416146675910946e-05, + "loss": 0.0, + "step": 8611 + }, + { + "epoch": 1.99, + "learning_rate": 3.34142207842231e-05, + "loss": 0.0001, + "step": 8612 + }, + { + "epoch": 1.99, + "learning_rate": 3.341229489253525e-05, + "loss": 0.0, + "step": 8613 + }, + { + "epoch": 1.99, + "learning_rate": 3.341036900084739e-05, + "loss": 0.0, + "step": 8614 + }, + { + "epoch": 1.99, + "learning_rate": 3.340844310915954e-05, + "loss": 0.0004, + "step": 8615 + }, + { + "epoch": 1.99, + "learning_rate": 3.340651721747169e-05, + "loss": 0.0, + "step": 8616 + }, + { + "epoch": 1.99, + "learning_rate": 3.340459132578384e-05, + "loss": 0.0001, + "step": 8617 + }, + { + "epoch": 1.99, + "learning_rate": 3.3402665434095985e-05, + "loss": 0.0001, + "step": 8618 + }, + { + "epoch": 1.99, + "learning_rate": 3.340073954240814e-05, + "loss": 0.0001, + "step": 8619 + }, + { + "epoch": 1.99, + "learning_rate": 3.339881365072029e-05, + "loss": 0.0, + "step": 8620 + }, + { + "epoch": 1.99, + "learning_rate": 3.3396887759032434e-05, + "loss": 0.0, + "step": 8621 + }, + { + "epoch": 1.99, + "learning_rate": 3.339496186734458e-05, + "loss": 0.0004, + "step": 8622 + }, + { + "epoch": 1.99, + "learning_rate": 3.339303597565673e-05, + "loss": 0.0, + "step": 8623 + }, + { + "epoch": 1.99, + "learning_rate": 3.339111008396888e-05, + "loss": 0.0, + "step": 8624 + }, + { + "epoch": 1.99, + "learning_rate": 3.3389184192281024e-05, + "loss": 0.0, + "step": 8625 + }, + { + "epoch": 1.99, + "learning_rate": 3.338725830059318e-05, + "loss": 0.0, + "step": 8626 + }, + { + "epoch": 1.99, + "learning_rate": 3.3385332408905326e-05, + "loss": 0.0, + "step": 8627 + }, + { + "epoch": 1.99, + "learning_rate": 3.3383406517217473e-05, + "loss": 0.0, + "step": 8628 + }, + { + "epoch": 1.99, + "learning_rate": 3.338148062552962e-05, + "loss": 0.0003, + "step": 8629 + }, + { + "epoch": 1.99, + "learning_rate": 3.3379554733841775e-05, + "loss": 0.0006, + "step": 8630 + }, + { + "epoch": 1.99, + "learning_rate": 3.3377628842153916e-05, + "loss": 0.0, + "step": 8631 + }, + { + "epoch": 1.99, + "learning_rate": 3.337570295046607e-05, + "loss": 0.0, + "step": 8632 + }, + { + "epoch": 2.0, + "learning_rate": 3.337377705877822e-05, + "loss": 0.0, + "step": 8633 + }, + { + "epoch": 2.0, + "learning_rate": 3.3371851167090365e-05, + "loss": 0.0, + "step": 8634 + }, + { + "epoch": 2.0, + "learning_rate": 3.336992527540251e-05, + "loss": 0.0, + "step": 8635 + }, + { + "epoch": 2.0, + "learning_rate": 3.336799938371466e-05, + "loss": 0.0, + "step": 8636 + }, + { + "epoch": 2.0, + "learning_rate": 3.3366073492026814e-05, + "loss": 0.0, + "step": 8637 + }, + { + "epoch": 2.0, + "learning_rate": 3.3364147600338955e-05, + "loss": 0.0, + "step": 8638 + }, + { + "epoch": 2.0, + "learning_rate": 3.336222170865111e-05, + "loss": 0.0, + "step": 8639 + }, + { + "epoch": 2.0, + "learning_rate": 3.336029581696326e-05, + "loss": 0.0016, + "step": 8640 + }, + { + "epoch": 2.0, + "learning_rate": 3.3358369925275404e-05, + "loss": 0.0005, + "step": 8641 + }, + { + "epoch": 2.0, + "learning_rate": 3.335644403358755e-05, + "loss": 0.0001, + "step": 8642 + }, + { + "epoch": 2.0, + "learning_rate": 3.33545181418997e-05, + "loss": 0.0, + "step": 8643 + }, + { + "epoch": 2.0, + "learning_rate": 3.335259225021185e-05, + "loss": 0.0003, + "step": 8644 + }, + { + "epoch": 2.0, + "learning_rate": 3.3350666358523994e-05, + "loss": 0.0001, + "step": 8645 + }, + { + "epoch": 2.0, + "learning_rate": 3.334874046683615e-05, + "loss": 0.0, + "step": 8646 + }, + { + "epoch": 2.0, + "learning_rate": 3.3346814575148296e-05, + "loss": 0.0001, + "step": 8647 + }, + { + "epoch": 2.0, + "learning_rate": 3.334488868346044e-05, + "loss": 0.0, + "step": 8648 + }, + { + "epoch": 2.0, + "learning_rate": 3.334296279177259e-05, + "loss": 0.0001, + "step": 8649 + }, + { + "epoch": 2.0, + "learning_rate": 3.3341036900084745e-05, + "loss": 0.0, + "step": 8650 + }, + { + "epoch": 2.0, + "learning_rate": 3.333911100839689e-05, + "loss": 0.0001, + "step": 8651 + }, + { + "epoch": 2.0, + "learning_rate": 3.333718511670903e-05, + "loss": 0.0023, + "step": 8652 + }, + { + "epoch": 2.0, + "learning_rate": 3.333525922502119e-05, + "loss": 0.0, + "step": 8653 + }, + { + "epoch": 2.0, + "learning_rate": 3.3333333333333335e-05, + "loss": 0.0001, + "step": 8654 + }, + { + "epoch": 2.0, + "learning_rate": 3.333140744164548e-05, + "loss": 0.0, + "step": 8655 + }, + { + "epoch": 2.0, + "learning_rate": 3.332948154995763e-05, + "loss": 0.0, + "step": 8656 + }, + { + "epoch": 2.0, + "learning_rate": 3.3327555658269784e-05, + "loss": 0.0001, + "step": 8657 + }, + { + "epoch": 2.0, + "learning_rate": 3.332562976658193e-05, + "loss": 0.0, + "step": 8658 + }, + { + "epoch": 2.0, + "learning_rate": 3.332370387489408e-05, + "loss": 0.0001, + "step": 8659 + }, + { + "epoch": 2.0, + "learning_rate": 3.3321777983206227e-05, + "loss": 0.0001, + "step": 8660 + }, + { + "epoch": 2.0, + "learning_rate": 3.3319852091518374e-05, + "loss": 0.0, + "step": 8661 + }, + { + "epoch": 2.0, + "learning_rate": 3.331792619983052e-05, + "loss": 0.0, + "step": 8662 + }, + { + "epoch": 2.0, + "learning_rate": 3.331600030814267e-05, + "loss": 0.0001, + "step": 8663 + }, + { + "epoch": 2.0, + "learning_rate": 3.331407441645482e-05, + "loss": 0.0008, + "step": 8664 + }, + { + "epoch": 2.0, + "learning_rate": 3.3312148524766964e-05, + "loss": 0.0, + "step": 8665 + }, + { + "epoch": 2.0, + "learning_rate": 3.331022263307912e-05, + "loss": 0.0, + "step": 8666 + }, + { + "epoch": 2.0, + "learning_rate": 3.3308296741391266e-05, + "loss": 0.0001, + "step": 8667 + }, + { + "epoch": 2.0, + "learning_rate": 3.330637084970342e-05, + "loss": 0.0001, + "step": 8668 + }, + { + "epoch": 2.0, + "learning_rate": 3.330444495801556e-05, + "loss": 0.0019, + "step": 8669 + }, + { + "epoch": 2.0, + "learning_rate": 3.330251906632771e-05, + "loss": 0.0001, + "step": 8670 + }, + { + "epoch": 2.0, + "learning_rate": 3.330059317463986e-05, + "loss": 0.0001, + "step": 8671 + }, + { + "epoch": 2.0, + "learning_rate": 3.3298667282952e-05, + "loss": 0.0002, + "step": 8672 + }, + { + "epoch": 2.0, + "learning_rate": 3.329674139126416e-05, + "loss": 0.0, + "step": 8673 + }, + { + "epoch": 2.0, + "learning_rate": 3.3294815499576305e-05, + "loss": 0.0, + "step": 8674 + }, + { + "epoch": 2.0, + "learning_rate": 3.329288960788845e-05, + "loss": 0.0, + "step": 8675 + }, + { + "epoch": 2.01, + "learning_rate": 3.32909637162006e-05, + "loss": 0.0, + "step": 8676 + }, + { + "epoch": 2.01, + "learning_rate": 3.3289037824512754e-05, + "loss": 0.0001, + "step": 8677 + }, + { + "epoch": 2.01, + "learning_rate": 3.32871119328249e-05, + "loss": 0.0, + "step": 8678 + }, + { + "epoch": 2.01, + "learning_rate": 3.328518604113704e-05, + "loss": 0.0, + "step": 8679 + }, + { + "epoch": 2.01, + "learning_rate": 3.3283260149449196e-05, + "loss": 0.0, + "step": 8680 + }, + { + "epoch": 2.01, + "learning_rate": 3.3281334257761344e-05, + "loss": 0.0, + "step": 8681 + }, + { + "epoch": 2.01, + "learning_rate": 3.327940836607349e-05, + "loss": 0.0, + "step": 8682 + }, + { + "epoch": 2.01, + "learning_rate": 3.327748247438564e-05, + "loss": 0.0001, + "step": 8683 + }, + { + "epoch": 2.01, + "learning_rate": 3.327555658269779e-05, + "loss": 0.0, + "step": 8684 + }, + { + "epoch": 2.01, + "learning_rate": 3.327363069100994e-05, + "loss": 0.0004, + "step": 8685 + }, + { + "epoch": 2.01, + "learning_rate": 3.327170479932209e-05, + "loss": 0.0, + "step": 8686 + }, + { + "epoch": 2.01, + "learning_rate": 3.3269778907634236e-05, + "loss": 0.0, + "step": 8687 + }, + { + "epoch": 2.01, + "learning_rate": 3.326785301594639e-05, + "loss": 0.0, + "step": 8688 + }, + { + "epoch": 2.01, + "learning_rate": 3.326592712425853e-05, + "loss": 0.0, + "step": 8689 + }, + { + "epoch": 2.01, + "learning_rate": 3.326400123257068e-05, + "loss": 0.0, + "step": 8690 + }, + { + "epoch": 2.01, + "learning_rate": 3.326207534088283e-05, + "loss": 0.0, + "step": 8691 + }, + { + "epoch": 2.01, + "learning_rate": 3.326014944919498e-05, + "loss": 0.0, + "step": 8692 + }, + { + "epoch": 2.01, + "learning_rate": 3.325822355750713e-05, + "loss": 0.0, + "step": 8693 + }, + { + "epoch": 2.01, + "learning_rate": 3.3256297665819275e-05, + "loss": 0.0001, + "step": 8694 + }, + { + "epoch": 2.01, + "learning_rate": 3.325437177413143e-05, + "loss": 0.0001, + "step": 8695 + }, + { + "epoch": 2.01, + "learning_rate": 3.325244588244357e-05, + "loss": 0.0001, + "step": 8696 + }, + { + "epoch": 2.01, + "learning_rate": 3.3250519990755724e-05, + "loss": 0.0, + "step": 8697 + }, + { + "epoch": 2.01, + "learning_rate": 3.324859409906787e-05, + "loss": 0.0001, + "step": 8698 + }, + { + "epoch": 2.01, + "learning_rate": 3.324666820738002e-05, + "loss": 0.0, + "step": 8699 + }, + { + "epoch": 2.01, + "learning_rate": 3.3244742315692166e-05, + "loss": 0.0, + "step": 8700 + }, + { + "epoch": 2.01, + "learning_rate": 3.3242816424004314e-05, + "loss": 0.0, + "step": 8701 + }, + { + "epoch": 2.01, + "learning_rate": 3.324089053231647e-05, + "loss": 0.0, + "step": 8702 + }, + { + "epoch": 2.01, + "learning_rate": 3.323896464062861e-05, + "loss": 0.0, + "step": 8703 + }, + { + "epoch": 2.01, + "learning_rate": 3.323703874894076e-05, + "loss": 0.0, + "step": 8704 + }, + { + "epoch": 2.01, + "learning_rate": 3.323511285725291e-05, + "loss": 0.0002, + "step": 8705 + }, + { + "epoch": 2.01, + "learning_rate": 3.323318696556506e-05, + "loss": 0.0002, + "step": 8706 + }, + { + "epoch": 2.01, + "learning_rate": 3.3231261073877205e-05, + "loss": 0.0001, + "step": 8707 + }, + { + "epoch": 2.01, + "learning_rate": 3.322933518218935e-05, + "loss": 0.0, + "step": 8708 + }, + { + "epoch": 2.01, + "learning_rate": 3.322740929050151e-05, + "loss": 0.0, + "step": 8709 + }, + { + "epoch": 2.01, + "learning_rate": 3.322548339881365e-05, + "loss": 0.0, + "step": 8710 + }, + { + "epoch": 2.01, + "learning_rate": 3.32235575071258e-05, + "loss": 0.0001, + "step": 8711 + }, + { + "epoch": 2.01, + "learning_rate": 3.322163161543795e-05, + "loss": 0.0, + "step": 8712 + }, + { + "epoch": 2.01, + "learning_rate": 3.32197057237501e-05, + "loss": 0.0, + "step": 8713 + }, + { + "epoch": 2.01, + "learning_rate": 3.3217779832062245e-05, + "loss": 0.0, + "step": 8714 + }, + { + "epoch": 2.01, + "learning_rate": 3.32158539403744e-05, + "loss": 0.0, + "step": 8715 + }, + { + "epoch": 2.01, + "learning_rate": 3.3213928048686546e-05, + "loss": 0.0546, + "step": 8716 + }, + { + "epoch": 2.01, + "learning_rate": 3.321200215699869e-05, + "loss": 0.0, + "step": 8717 + }, + { + "epoch": 2.01, + "learning_rate": 3.321007626531084e-05, + "loss": 0.0, + "step": 8718 + }, + { + "epoch": 2.02, + "learning_rate": 3.320815037362299e-05, + "loss": 0.0, + "step": 8719 + }, + { + "epoch": 2.02, + "learning_rate": 3.3206224481935136e-05, + "loss": 0.0001, + "step": 8720 + }, + { + "epoch": 2.02, + "learning_rate": 3.3204298590247284e-05, + "loss": 0.0, + "step": 8721 + }, + { + "epoch": 2.02, + "learning_rate": 3.320237269855944e-05, + "loss": 0.0, + "step": 8722 + }, + { + "epoch": 2.02, + "learning_rate": 3.3200446806871585e-05, + "loss": 0.0, + "step": 8723 + }, + { + "epoch": 2.02, + "learning_rate": 3.319852091518373e-05, + "loss": 0.0003, + "step": 8724 + }, + { + "epoch": 2.02, + "learning_rate": 3.319659502349588e-05, + "loss": 0.0, + "step": 8725 + }, + { + "epoch": 2.02, + "learning_rate": 3.319466913180803e-05, + "loss": 0.0003, + "step": 8726 + }, + { + "epoch": 2.02, + "learning_rate": 3.3192743240120175e-05, + "loss": 0.0, + "step": 8727 + }, + { + "epoch": 2.02, + "learning_rate": 3.319081734843232e-05, + "loss": 0.0002, + "step": 8728 + }, + { + "epoch": 2.02, + "learning_rate": 3.318889145674448e-05, + "loss": 0.0007, + "step": 8729 + }, + { + "epoch": 2.02, + "learning_rate": 3.3186965565056625e-05, + "loss": 0.0, + "step": 8730 + }, + { + "epoch": 2.02, + "learning_rate": 3.318503967336877e-05, + "loss": 0.0007, + "step": 8731 + }, + { + "epoch": 2.02, + "learning_rate": 3.318311378168092e-05, + "loss": 0.0001, + "step": 8732 + }, + { + "epoch": 2.02, + "learning_rate": 3.3181187889993074e-05, + "loss": 0.0021, + "step": 8733 + }, + { + "epoch": 2.02, + "learning_rate": 3.3179261998305214e-05, + "loss": 0.0, + "step": 8734 + }, + { + "epoch": 2.02, + "learning_rate": 3.317733610661736e-05, + "loss": 0.0001, + "step": 8735 + }, + { + "epoch": 2.02, + "learning_rate": 3.3175410214929516e-05, + "loss": 0.0, + "step": 8736 + }, + { + "epoch": 2.02, + "learning_rate": 3.3173484323241664e-05, + "loss": 0.0001, + "step": 8737 + }, + { + "epoch": 2.02, + "learning_rate": 3.317155843155381e-05, + "loss": 0.0002, + "step": 8738 + }, + { + "epoch": 2.02, + "learning_rate": 3.316963253986596e-05, + "loss": 0.0001, + "step": 8739 + }, + { + "epoch": 2.02, + "learning_rate": 3.316770664817811e-05, + "loss": 0.0, + "step": 8740 + }, + { + "epoch": 2.02, + "learning_rate": 3.3165780756490254e-05, + "loss": 0.0, + "step": 8741 + }, + { + "epoch": 2.02, + "learning_rate": 3.316385486480241e-05, + "loss": 0.0, + "step": 8742 + }, + { + "epoch": 2.02, + "learning_rate": 3.3161928973114555e-05, + "loss": 0.0, + "step": 8743 + }, + { + "epoch": 2.02, + "learning_rate": 3.31600030814267e-05, + "loss": 0.0, + "step": 8744 + }, + { + "epoch": 2.02, + "learning_rate": 3.315807718973885e-05, + "loss": 0.0002, + "step": 8745 + }, + { + "epoch": 2.02, + "learning_rate": 3.3156151298051e-05, + "loss": 0.0, + "step": 8746 + }, + { + "epoch": 2.02, + "learning_rate": 3.3154225406363145e-05, + "loss": 0.0, + "step": 8747 + }, + { + "epoch": 2.02, + "learning_rate": 3.315229951467529e-05, + "loss": 0.0, + "step": 8748 + }, + { + "epoch": 2.02, + "learning_rate": 3.315037362298745e-05, + "loss": 0.0, + "step": 8749 + }, + { + "epoch": 2.02, + "learning_rate": 3.3148447731299594e-05, + "loss": 0.0, + "step": 8750 + }, + { + "epoch": 2.02, + "learning_rate": 3.314652183961174e-05, + "loss": 0.0, + "step": 8751 + }, + { + "epoch": 2.02, + "learning_rate": 3.314459594792389e-05, + "loss": 0.0, + "step": 8752 + }, + { + "epoch": 2.02, + "learning_rate": 3.3142670056236044e-05, + "loss": 0.0007, + "step": 8753 + }, + { + "epoch": 2.02, + "learning_rate": 3.3140744164548184e-05, + "loss": 0.0, + "step": 8754 + }, + { + "epoch": 2.02, + "learning_rate": 3.313881827286033e-05, + "loss": 0.0, + "step": 8755 + }, + { + "epoch": 2.02, + "learning_rate": 3.3136892381172486e-05, + "loss": 0.0004, + "step": 8756 + }, + { + "epoch": 2.02, + "learning_rate": 3.3134966489484634e-05, + "loss": 0.0, + "step": 8757 + }, + { + "epoch": 2.02, + "learning_rate": 3.313304059779678e-05, + "loss": 0.0014, + "step": 8758 + }, + { + "epoch": 2.02, + "learning_rate": 3.313111470610893e-05, + "loss": 0.0, + "step": 8759 + }, + { + "epoch": 2.02, + "learning_rate": 3.312918881442108e-05, + "loss": 0.0, + "step": 8760 + }, + { + "epoch": 2.02, + "learning_rate": 3.3127262922733223e-05, + "loss": 0.0, + "step": 8761 + }, + { + "epoch": 2.02, + "learning_rate": 3.312533703104538e-05, + "loss": 0.0001, + "step": 8762 + }, + { + "epoch": 2.03, + "learning_rate": 3.3123411139357525e-05, + "loss": 0.0, + "step": 8763 + }, + { + "epoch": 2.03, + "learning_rate": 3.312148524766967e-05, + "loss": 0.0, + "step": 8764 + }, + { + "epoch": 2.03, + "learning_rate": 3.311955935598182e-05, + "loss": 0.0, + "step": 8765 + }, + { + "epoch": 2.03, + "learning_rate": 3.311763346429397e-05, + "loss": 0.0, + "step": 8766 + }, + { + "epoch": 2.03, + "learning_rate": 3.311570757260612e-05, + "loss": 0.0001, + "step": 8767 + }, + { + "epoch": 2.03, + "learning_rate": 3.311378168091826e-05, + "loss": 0.0, + "step": 8768 + }, + { + "epoch": 2.03, + "learning_rate": 3.311185578923042e-05, + "loss": 0.0, + "step": 8769 + }, + { + "epoch": 2.03, + "learning_rate": 3.3109929897542564e-05, + "loss": 0.0, + "step": 8770 + }, + { + "epoch": 2.03, + "learning_rate": 3.310800400585471e-05, + "loss": 0.0, + "step": 8771 + }, + { + "epoch": 2.03, + "learning_rate": 3.310607811416686e-05, + "loss": 0.0001, + "step": 8772 + }, + { + "epoch": 2.03, + "learning_rate": 3.310415222247901e-05, + "loss": 0.0001, + "step": 8773 + }, + { + "epoch": 2.03, + "learning_rate": 3.310222633079116e-05, + "loss": 0.0, + "step": 8774 + }, + { + "epoch": 2.03, + "learning_rate": 3.31003004391033e-05, + "loss": 0.0059, + "step": 8775 + }, + { + "epoch": 2.03, + "learning_rate": 3.3098374547415456e-05, + "loss": 0.0129, + "step": 8776 + }, + { + "epoch": 2.03, + "learning_rate": 3.3096448655727603e-05, + "loss": 0.0, + "step": 8777 + }, + { + "epoch": 2.03, + "learning_rate": 3.309452276403975e-05, + "loss": 0.0, + "step": 8778 + }, + { + "epoch": 2.03, + "learning_rate": 3.30925968723519e-05, + "loss": 0.0, + "step": 8779 + }, + { + "epoch": 2.03, + "learning_rate": 3.309067098066405e-05, + "loss": 0.0005, + "step": 8780 + }, + { + "epoch": 2.03, + "learning_rate": 3.30887450889762e-05, + "loss": 0.0, + "step": 8781 + }, + { + "epoch": 2.03, + "learning_rate": 3.308681919728834e-05, + "loss": 0.0, + "step": 8782 + }, + { + "epoch": 2.03, + "learning_rate": 3.3084893305600495e-05, + "loss": 0.0005, + "step": 8783 + }, + { + "epoch": 2.03, + "learning_rate": 3.308296741391264e-05, + "loss": 0.0, + "step": 8784 + }, + { + "epoch": 2.03, + "learning_rate": 3.308104152222479e-05, + "loss": 0.0, + "step": 8785 + }, + { + "epoch": 2.03, + "learning_rate": 3.307911563053694e-05, + "loss": 0.0001, + "step": 8786 + }, + { + "epoch": 2.03, + "learning_rate": 3.307718973884909e-05, + "loss": 0.0001, + "step": 8787 + }, + { + "epoch": 2.03, + "learning_rate": 3.307526384716124e-05, + "loss": 0.0, + "step": 8788 + }, + { + "epoch": 2.03, + "learning_rate": 3.307333795547339e-05, + "loss": 0.0, + "step": 8789 + }, + { + "epoch": 2.03, + "learning_rate": 3.3071412063785534e-05, + "loss": 0.0, + "step": 8790 + }, + { + "epoch": 2.03, + "learning_rate": 3.306948617209768e-05, + "loss": 0.0, + "step": 8791 + }, + { + "epoch": 2.03, + "learning_rate": 3.306756028040983e-05, + "loss": 0.0005, + "step": 8792 + }, + { + "epoch": 2.03, + "learning_rate": 3.306563438872198e-05, + "loss": 0.0003, + "step": 8793 + }, + { + "epoch": 2.03, + "learning_rate": 3.306370849703413e-05, + "loss": 0.0003, + "step": 8794 + }, + { + "epoch": 2.03, + "learning_rate": 3.306178260534628e-05, + "loss": 0.0, + "step": 8795 + }, + { + "epoch": 2.03, + "learning_rate": 3.3059856713658426e-05, + "loss": 0.0028, + "step": 8796 + }, + { + "epoch": 2.03, + "learning_rate": 3.305793082197057e-05, + "loss": 0.0005, + "step": 8797 + }, + { + "epoch": 2.03, + "learning_rate": 3.305600493028273e-05, + "loss": 0.0, + "step": 8798 + }, + { + "epoch": 2.03, + "learning_rate": 3.305407903859487e-05, + "loss": 0.0, + "step": 8799 + }, + { + "epoch": 2.03, + "learning_rate": 3.305215314690702e-05, + "loss": 0.0002, + "step": 8800 + }, + { + "epoch": 2.03, + "learning_rate": 3.305022725521917e-05, + "loss": 0.0, + "step": 8801 + }, + { + "epoch": 2.03, + "learning_rate": 3.304830136353132e-05, + "loss": 0.0001, + "step": 8802 + }, + { + "epoch": 2.03, + "learning_rate": 3.3046375471843465e-05, + "loss": 0.0, + "step": 8803 + }, + { + "epoch": 2.03, + "learning_rate": 3.304444958015561e-05, + "loss": 0.0, + "step": 8804 + }, + { + "epoch": 2.03, + "learning_rate": 3.304252368846777e-05, + "loss": 0.0, + "step": 8805 + }, + { + "epoch": 2.04, + "learning_rate": 3.304059779677991e-05, + "loss": 0.0, + "step": 8806 + }, + { + "epoch": 2.04, + "learning_rate": 3.303867190509206e-05, + "loss": 0.0001, + "step": 8807 + }, + { + "epoch": 2.04, + "learning_rate": 3.303674601340421e-05, + "loss": 0.0003, + "step": 8808 + }, + { + "epoch": 2.04, + "learning_rate": 3.303482012171636e-05, + "loss": 0.0009, + "step": 8809 + }, + { + "epoch": 2.04, + "learning_rate": 3.3032894230028504e-05, + "loss": 0.0, + "step": 8810 + }, + { + "epoch": 2.04, + "learning_rate": 3.303096833834065e-05, + "loss": 0.0, + "step": 8811 + }, + { + "epoch": 2.04, + "learning_rate": 3.3029042446652806e-05, + "loss": 0.0001, + "step": 8812 + }, + { + "epoch": 2.04, + "learning_rate": 3.3027116554964947e-05, + "loss": 0.0, + "step": 8813 + }, + { + "epoch": 2.04, + "learning_rate": 3.30251906632771e-05, + "loss": 0.0003, + "step": 8814 + }, + { + "epoch": 2.04, + "learning_rate": 3.302326477158925e-05, + "loss": 0.0, + "step": 8815 + }, + { + "epoch": 2.04, + "learning_rate": 3.3021338879901396e-05, + "loss": 0.0012, + "step": 8816 + }, + { + "epoch": 2.04, + "learning_rate": 3.301941298821354e-05, + "loss": 0.0001, + "step": 8817 + }, + { + "epoch": 2.04, + "learning_rate": 3.30174870965257e-05, + "loss": 0.0001, + "step": 8818 + }, + { + "epoch": 2.04, + "learning_rate": 3.301556120483784e-05, + "loss": 0.0001, + "step": 8819 + }, + { + "epoch": 2.04, + "learning_rate": 3.3013635313149986e-05, + "loss": 0.0004, + "step": 8820 + }, + { + "epoch": 2.04, + "learning_rate": 3.301170942146214e-05, + "loss": 0.0, + "step": 8821 + }, + { + "epoch": 2.04, + "learning_rate": 3.300978352977429e-05, + "loss": 0.0, + "step": 8822 + }, + { + "epoch": 2.04, + "learning_rate": 3.3007857638086435e-05, + "loss": 0.0, + "step": 8823 + }, + { + "epoch": 2.04, + "learning_rate": 3.300593174639858e-05, + "loss": 0.0, + "step": 8824 + }, + { + "epoch": 2.04, + "learning_rate": 3.3004005854710737e-05, + "loss": 0.0002, + "step": 8825 + }, + { + "epoch": 2.04, + "learning_rate": 3.300207996302288e-05, + "loss": 0.0, + "step": 8826 + }, + { + "epoch": 2.04, + "learning_rate": 3.300015407133503e-05, + "loss": 0.0002, + "step": 8827 + }, + { + "epoch": 2.04, + "learning_rate": 3.299822817964718e-05, + "loss": 0.0002, + "step": 8828 + }, + { + "epoch": 2.04, + "learning_rate": 3.2996302287959327e-05, + "loss": 0.0006, + "step": 8829 + }, + { + "epoch": 2.04, + "learning_rate": 3.2994376396271474e-05, + "loss": 0.0, + "step": 8830 + }, + { + "epoch": 2.04, + "learning_rate": 3.299245050458362e-05, + "loss": 0.0002, + "step": 8831 + }, + { + "epoch": 2.04, + "learning_rate": 3.2990524612895776e-05, + "loss": 0.0001, + "step": 8832 + }, + { + "epoch": 2.04, + "learning_rate": 3.2988598721207916e-05, + "loss": 0.0, + "step": 8833 + }, + { + "epoch": 2.04, + "learning_rate": 3.298667282952007e-05, + "loss": 0.0, + "step": 8834 + }, + { + "epoch": 2.04, + "learning_rate": 3.298474693783222e-05, + "loss": 0.0001, + "step": 8835 + }, + { + "epoch": 2.04, + "learning_rate": 3.2982821046144366e-05, + "loss": 0.0, + "step": 8836 + }, + { + "epoch": 2.04, + "learning_rate": 3.298089515445651e-05, + "loss": 0.0, + "step": 8837 + }, + { + "epoch": 2.04, + "learning_rate": 3.297896926276866e-05, + "loss": 0.0, + "step": 8838 + }, + { + "epoch": 2.04, + "learning_rate": 3.2977043371080815e-05, + "loss": 0.0007, + "step": 8839 + }, + { + "epoch": 2.04, + "learning_rate": 3.2975117479392956e-05, + "loss": 0.0, + "step": 8840 + }, + { + "epoch": 2.04, + "learning_rate": 3.297319158770511e-05, + "loss": 0.0001, + "step": 8841 + }, + { + "epoch": 2.04, + "learning_rate": 3.297126569601726e-05, + "loss": 0.0, + "step": 8842 + }, + { + "epoch": 2.04, + "learning_rate": 3.2969339804329405e-05, + "loss": 0.0002, + "step": 8843 + }, + { + "epoch": 2.04, + "learning_rate": 3.296741391264155e-05, + "loss": 0.0001, + "step": 8844 + }, + { + "epoch": 2.04, + "learning_rate": 3.2965488020953706e-05, + "loss": 0.0001, + "step": 8845 + }, + { + "epoch": 2.04, + "learning_rate": 3.2963562129265854e-05, + "loss": 0.0, + "step": 8846 + }, + { + "epoch": 2.04, + "learning_rate": 3.2961636237577995e-05, + "loss": 0.0, + "step": 8847 + }, + { + "epoch": 2.04, + "learning_rate": 3.295971034589015e-05, + "loss": 0.0, + "step": 8848 + }, + { + "epoch": 2.05, + "learning_rate": 3.2957784454202296e-05, + "loss": 0.0, + "step": 8849 + }, + { + "epoch": 2.05, + "learning_rate": 3.2955858562514444e-05, + "loss": 0.0003, + "step": 8850 + }, + { + "epoch": 2.05, + "learning_rate": 3.295393267082659e-05, + "loss": 0.0, + "step": 8851 + }, + { + "epoch": 2.05, + "learning_rate": 3.2952006779138746e-05, + "loss": 0.0, + "step": 8852 + }, + { + "epoch": 2.05, + "learning_rate": 3.295008088745089e-05, + "loss": 0.0001, + "step": 8853 + }, + { + "epoch": 2.05, + "learning_rate": 3.294815499576304e-05, + "loss": 0.0003, + "step": 8854 + }, + { + "epoch": 2.05, + "learning_rate": 3.294622910407519e-05, + "loss": 0.0, + "step": 8855 + }, + { + "epoch": 2.05, + "learning_rate": 3.294430321238734e-05, + "loss": 0.0, + "step": 8856 + }, + { + "epoch": 2.05, + "learning_rate": 3.294237732069948e-05, + "loss": 0.0, + "step": 8857 + }, + { + "epoch": 2.05, + "learning_rate": 3.294045142901163e-05, + "loss": 0.0, + "step": 8858 + }, + { + "epoch": 2.05, + "learning_rate": 3.2938525537323785e-05, + "loss": 0.0, + "step": 8859 + }, + { + "epoch": 2.05, + "learning_rate": 3.293659964563593e-05, + "loss": 0.0, + "step": 8860 + }, + { + "epoch": 2.05, + "learning_rate": 3.293467375394808e-05, + "loss": 0.0, + "step": 8861 + }, + { + "epoch": 2.05, + "learning_rate": 3.293274786226023e-05, + "loss": 0.0, + "step": 8862 + }, + { + "epoch": 2.05, + "learning_rate": 3.293082197057238e-05, + "loss": 0.0001, + "step": 8863 + }, + { + "epoch": 2.05, + "learning_rate": 3.292889607888452e-05, + "loss": 0.0, + "step": 8864 + }, + { + "epoch": 2.05, + "learning_rate": 3.2926970187196676e-05, + "loss": 0.0, + "step": 8865 + }, + { + "epoch": 2.05, + "learning_rate": 3.2925044295508824e-05, + "loss": 0.0, + "step": 8866 + }, + { + "epoch": 2.05, + "learning_rate": 3.292311840382097e-05, + "loss": 0.0, + "step": 8867 + }, + { + "epoch": 2.05, + "learning_rate": 3.292119251213312e-05, + "loss": 0.0001, + "step": 8868 + }, + { + "epoch": 2.05, + "learning_rate": 3.2919266620445266e-05, + "loss": 0.0053, + "step": 8869 + }, + { + "epoch": 2.05, + "learning_rate": 3.291734072875742e-05, + "loss": 0.0, + "step": 8870 + }, + { + "epoch": 2.05, + "learning_rate": 3.291541483706956e-05, + "loss": 0.0, + "step": 8871 + }, + { + "epoch": 2.05, + "learning_rate": 3.2913488945381715e-05, + "loss": 0.0, + "step": 8872 + }, + { + "epoch": 2.05, + "learning_rate": 3.291156305369386e-05, + "loss": 0.0, + "step": 8873 + }, + { + "epoch": 2.05, + "learning_rate": 3.290963716200601e-05, + "loss": 0.0, + "step": 8874 + }, + { + "epoch": 2.05, + "learning_rate": 3.290771127031816e-05, + "loss": 0.0, + "step": 8875 + }, + { + "epoch": 2.05, + "learning_rate": 3.2905785378630305e-05, + "loss": 0.0004, + "step": 8876 + }, + { + "epoch": 2.05, + "learning_rate": 3.290385948694246e-05, + "loss": 0.0009, + "step": 8877 + }, + { + "epoch": 2.05, + "learning_rate": 3.29019335952546e-05, + "loss": 0.0, + "step": 8878 + }, + { + "epoch": 2.05, + "learning_rate": 3.2900007703566755e-05, + "loss": 0.0007, + "step": 8879 + }, + { + "epoch": 2.05, + "learning_rate": 3.28980818118789e-05, + "loss": 0.0, + "step": 8880 + }, + { + "epoch": 2.05, + "learning_rate": 3.289615592019105e-05, + "loss": 0.0, + "step": 8881 + }, + { + "epoch": 2.05, + "learning_rate": 3.28942300285032e-05, + "loss": 0.0003, + "step": 8882 + }, + { + "epoch": 2.05, + "learning_rate": 3.289230413681535e-05, + "loss": 0.0, + "step": 8883 + }, + { + "epoch": 2.05, + "learning_rate": 3.28903782451275e-05, + "loss": 0.0, + "step": 8884 + }, + { + "epoch": 2.05, + "learning_rate": 3.288845235343964e-05, + "loss": 0.0, + "step": 8885 + }, + { + "epoch": 2.05, + "learning_rate": 3.2886526461751794e-05, + "loss": 0.0, + "step": 8886 + }, + { + "epoch": 2.05, + "learning_rate": 3.288460057006394e-05, + "loss": 0.0, + "step": 8887 + }, + { + "epoch": 2.05, + "learning_rate": 3.288267467837609e-05, + "loss": 0.0, + "step": 8888 + }, + { + "epoch": 2.05, + "learning_rate": 3.2880748786688236e-05, + "loss": 0.0003, + "step": 8889 + }, + { + "epoch": 2.05, + "learning_rate": 3.287882289500039e-05, + "loss": 0.0, + "step": 8890 + }, + { + "epoch": 2.05, + "learning_rate": 3.287689700331254e-05, + "loss": 0.0, + "step": 8891 + }, + { + "epoch": 2.06, + "learning_rate": 3.2874971111624685e-05, + "loss": 0.0, + "step": 8892 + }, + { + "epoch": 2.06, + "learning_rate": 3.287304521993683e-05, + "loss": 0.0, + "step": 8893 + }, + { + "epoch": 2.06, + "learning_rate": 3.287111932824898e-05, + "loss": 0.0006, + "step": 8894 + }, + { + "epoch": 2.06, + "learning_rate": 3.286919343656113e-05, + "loss": 0.0, + "step": 8895 + }, + { + "epoch": 2.06, + "learning_rate": 3.2867267544873275e-05, + "loss": 0.0, + "step": 8896 + }, + { + "epoch": 2.06, + "learning_rate": 3.286534165318543e-05, + "loss": 0.0001, + "step": 8897 + }, + { + "epoch": 2.06, + "learning_rate": 3.286341576149757e-05, + "loss": 0.0001, + "step": 8898 + }, + { + "epoch": 2.06, + "learning_rate": 3.2861489869809724e-05, + "loss": 0.0001, + "step": 8899 + }, + { + "epoch": 2.06, + "learning_rate": 3.285956397812187e-05, + "loss": 0.0001, + "step": 8900 + }, + { + "epoch": 2.06, + "learning_rate": 3.285763808643402e-05, + "loss": 0.0003, + "step": 8901 + }, + { + "epoch": 2.06, + "learning_rate": 3.285571219474617e-05, + "loss": 0.0007, + "step": 8902 + }, + { + "epoch": 2.06, + "learning_rate": 3.2853786303058314e-05, + "loss": 0.0001, + "step": 8903 + }, + { + "epoch": 2.06, + "learning_rate": 3.285186041137047e-05, + "loss": 0.0008, + "step": 8904 + }, + { + "epoch": 2.06, + "learning_rate": 3.284993451968261e-05, + "loss": 0.0, + "step": 8905 + }, + { + "epoch": 2.06, + "learning_rate": 3.2848008627994764e-05, + "loss": 0.0, + "step": 8906 + }, + { + "epoch": 2.06, + "learning_rate": 3.284608273630691e-05, + "loss": 0.0, + "step": 8907 + }, + { + "epoch": 2.06, + "learning_rate": 3.284415684461906e-05, + "loss": 0.0002, + "step": 8908 + }, + { + "epoch": 2.06, + "learning_rate": 3.2842230952931206e-05, + "loss": 0.0006, + "step": 8909 + }, + { + "epoch": 2.06, + "learning_rate": 3.284030506124336e-05, + "loss": 0.0001, + "step": 8910 + }, + { + "epoch": 2.06, + "learning_rate": 3.283837916955551e-05, + "loss": 0.0, + "step": 8911 + }, + { + "epoch": 2.06, + "learning_rate": 3.2836453277867655e-05, + "loss": 0.0001, + "step": 8912 + }, + { + "epoch": 2.06, + "learning_rate": 3.28345273861798e-05, + "loss": 0.0, + "step": 8913 + }, + { + "epoch": 2.06, + "learning_rate": 3.283260149449195e-05, + "loss": 0.0001, + "step": 8914 + }, + { + "epoch": 2.06, + "learning_rate": 3.28306756028041e-05, + "loss": 0.0, + "step": 8915 + }, + { + "epoch": 2.06, + "learning_rate": 3.2828749711116245e-05, + "loss": 0.0001, + "step": 8916 + }, + { + "epoch": 2.06, + "learning_rate": 3.28268238194284e-05, + "loss": 0.0, + "step": 8917 + }, + { + "epoch": 2.06, + "learning_rate": 3.282489792774055e-05, + "loss": 0.0003, + "step": 8918 + }, + { + "epoch": 2.06, + "learning_rate": 3.2822972036052694e-05, + "loss": 0.0, + "step": 8919 + }, + { + "epoch": 2.06, + "learning_rate": 3.282104614436484e-05, + "loss": 0.0001, + "step": 8920 + }, + { + "epoch": 2.06, + "learning_rate": 3.2819120252676996e-05, + "loss": 0.0012, + "step": 8921 + }, + { + "epoch": 2.06, + "learning_rate": 3.281719436098914e-05, + "loss": 0.0, + "step": 8922 + }, + { + "epoch": 2.06, + "learning_rate": 3.2815268469301284e-05, + "loss": 0.0, + "step": 8923 + }, + { + "epoch": 2.06, + "learning_rate": 3.281334257761344e-05, + "loss": 0.0, + "step": 8924 + }, + { + "epoch": 2.06, + "learning_rate": 3.2811416685925586e-05, + "loss": 0.0, + "step": 8925 + }, + { + "epoch": 2.06, + "learning_rate": 3.2809490794237733e-05, + "loss": 0.0, + "step": 8926 + }, + { + "epoch": 2.06, + "learning_rate": 3.280756490254988e-05, + "loss": 0.0, + "step": 8927 + }, + { + "epoch": 2.06, + "learning_rate": 3.2805639010862035e-05, + "loss": 0.0, + "step": 8928 + }, + { + "epoch": 2.06, + "learning_rate": 3.2803713119174176e-05, + "loss": 0.0006, + "step": 8929 + }, + { + "epoch": 2.06, + "learning_rate": 3.280178722748633e-05, + "loss": 0.0, + "step": 8930 + }, + { + "epoch": 2.06, + "learning_rate": 3.279986133579848e-05, + "loss": 0.0, + "step": 8931 + }, + { + "epoch": 2.06, + "learning_rate": 3.2797935444110625e-05, + "loss": 0.0, + "step": 8932 + }, + { + "epoch": 2.06, + "learning_rate": 3.279600955242277e-05, + "loss": 0.0, + "step": 8933 + }, + { + "epoch": 2.06, + "learning_rate": 3.279408366073492e-05, + "loss": 0.0, + "step": 8934 + }, + { + "epoch": 2.06, + "learning_rate": 3.2792157769047074e-05, + "loss": 0.0002, + "step": 8935 + }, + { + "epoch": 2.07, + "learning_rate": 3.2790231877359215e-05, + "loss": 0.0, + "step": 8936 + }, + { + "epoch": 2.07, + "learning_rate": 3.278830598567137e-05, + "loss": 0.0, + "step": 8937 + }, + { + "epoch": 2.07, + "learning_rate": 3.278638009398352e-05, + "loss": 0.0, + "step": 8938 + }, + { + "epoch": 2.07, + "learning_rate": 3.2784454202295664e-05, + "loss": 0.0001, + "step": 8939 + }, + { + "epoch": 2.07, + "learning_rate": 3.278252831060781e-05, + "loss": 0.0, + "step": 8940 + }, + { + "epoch": 2.07, + "learning_rate": 3.278060241891996e-05, + "loss": 0.0, + "step": 8941 + }, + { + "epoch": 2.07, + "learning_rate": 3.2778676527232113e-05, + "loss": 0.0, + "step": 8942 + }, + { + "epoch": 2.07, + "learning_rate": 3.2776750635544254e-05, + "loss": 0.0, + "step": 8943 + }, + { + "epoch": 2.07, + "learning_rate": 3.277482474385641e-05, + "loss": 0.0, + "step": 8944 + }, + { + "epoch": 2.07, + "learning_rate": 3.2772898852168556e-05, + "loss": 0.0, + "step": 8945 + }, + { + "epoch": 2.07, + "learning_rate": 3.27709729604807e-05, + "loss": 0.0, + "step": 8946 + }, + { + "epoch": 2.07, + "learning_rate": 3.276904706879285e-05, + "loss": 0.0001, + "step": 8947 + }, + { + "epoch": 2.07, + "learning_rate": 3.2767121177105005e-05, + "loss": 0.0001, + "step": 8948 + }, + { + "epoch": 2.07, + "learning_rate": 3.276519528541715e-05, + "loss": 0.0, + "step": 8949 + }, + { + "epoch": 2.07, + "learning_rate": 3.276326939372929e-05, + "loss": 0.0, + "step": 8950 + }, + { + "epoch": 2.07, + "learning_rate": 3.276134350204145e-05, + "loss": 0.0, + "step": 8951 + }, + { + "epoch": 2.07, + "learning_rate": 3.2759417610353595e-05, + "loss": 0.0, + "step": 8952 + }, + { + "epoch": 2.07, + "learning_rate": 3.275749171866574e-05, + "loss": 0.0, + "step": 8953 + }, + { + "epoch": 2.07, + "learning_rate": 3.275556582697789e-05, + "loss": 0.0004, + "step": 8954 + }, + { + "epoch": 2.07, + "learning_rate": 3.2753639935290044e-05, + "loss": 0.0, + "step": 8955 + }, + { + "epoch": 2.07, + "learning_rate": 3.275171404360219e-05, + "loss": 0.0, + "step": 8956 + }, + { + "epoch": 2.07, + "learning_rate": 3.274978815191434e-05, + "loss": 0.0009, + "step": 8957 + }, + { + "epoch": 2.07, + "learning_rate": 3.274786226022649e-05, + "loss": 0.0, + "step": 8958 + }, + { + "epoch": 2.07, + "learning_rate": 3.2745936368538634e-05, + "loss": 0.0, + "step": 8959 + }, + { + "epoch": 2.07, + "learning_rate": 3.274401047685078e-05, + "loss": 0.0001, + "step": 8960 + }, + { + "epoch": 2.07, + "learning_rate": 3.274208458516293e-05, + "loss": 0.0, + "step": 8961 + }, + { + "epoch": 2.07, + "learning_rate": 3.274015869347508e-05, + "loss": 0.0002, + "step": 8962 + }, + { + "epoch": 2.07, + "learning_rate": 3.273823280178723e-05, + "loss": 0.0001, + "step": 8963 + }, + { + "epoch": 2.07, + "learning_rate": 3.273630691009938e-05, + "loss": 0.0, + "step": 8964 + }, + { + "epoch": 2.07, + "learning_rate": 3.2734381018411526e-05, + "loss": 0.0, + "step": 8965 + }, + { + "epoch": 2.07, + "learning_rate": 3.273245512672368e-05, + "loss": 0.0002, + "step": 8966 + }, + { + "epoch": 2.07, + "learning_rate": 3.273052923503582e-05, + "loss": 0.0001, + "step": 8967 + }, + { + "epoch": 2.07, + "learning_rate": 3.2728603343347975e-05, + "loss": 0.0, + "step": 8968 + }, + { + "epoch": 2.07, + "learning_rate": 3.272667745166012e-05, + "loss": 0.0, + "step": 8969 + }, + { + "epoch": 2.07, + "learning_rate": 3.272475155997226e-05, + "loss": 0.0, + "step": 8970 + }, + { + "epoch": 2.07, + "learning_rate": 3.272282566828442e-05, + "loss": 0.0, + "step": 8971 + }, + { + "epoch": 2.07, + "learning_rate": 3.2720899776596565e-05, + "loss": 0.0, + "step": 8972 + }, + { + "epoch": 2.07, + "learning_rate": 3.271897388490871e-05, + "loss": 0.0, + "step": 8973 + }, + { + "epoch": 2.07, + "learning_rate": 3.271704799322086e-05, + "loss": 0.0042, + "step": 8974 + }, + { + "epoch": 2.07, + "learning_rate": 3.2715122101533014e-05, + "loss": 0.0001, + "step": 8975 + }, + { + "epoch": 2.07, + "learning_rate": 3.271319620984516e-05, + "loss": 0.0, + "step": 8976 + }, + { + "epoch": 2.07, + "learning_rate": 3.271127031815731e-05, + "loss": 0.0, + "step": 8977 + }, + { + "epoch": 2.07, + "learning_rate": 3.2709344426469457e-05, + "loss": 0.0, + "step": 8978 + }, + { + "epoch": 2.08, + "learning_rate": 3.2707418534781604e-05, + "loss": 0.0, + "step": 8979 + }, + { + "epoch": 2.08, + "learning_rate": 3.270549264309375e-05, + "loss": 0.0, + "step": 8980 + }, + { + "epoch": 2.08, + "learning_rate": 3.27035667514059e-05, + "loss": 0.0, + "step": 8981 + }, + { + "epoch": 2.08, + "learning_rate": 3.270164085971805e-05, + "loss": 0.005, + "step": 8982 + }, + { + "epoch": 2.08, + "learning_rate": 3.26997149680302e-05, + "loss": 0.0001, + "step": 8983 + }, + { + "epoch": 2.08, + "learning_rate": 3.269778907634235e-05, + "loss": 0.0, + "step": 8984 + }, + { + "epoch": 2.08, + "learning_rate": 3.2695863184654496e-05, + "loss": 0.0001, + "step": 8985 + }, + { + "epoch": 2.08, + "learning_rate": 3.269393729296665e-05, + "loss": 0.0001, + "step": 8986 + }, + { + "epoch": 2.08, + "learning_rate": 3.269201140127879e-05, + "loss": 0.0001, + "step": 8987 + }, + { + "epoch": 2.08, + "learning_rate": 3.269008550959094e-05, + "loss": 0.0, + "step": 8988 + }, + { + "epoch": 2.08, + "learning_rate": 3.268815961790309e-05, + "loss": 0.0, + "step": 8989 + }, + { + "epoch": 2.08, + "learning_rate": 3.268623372621524e-05, + "loss": 0.0, + "step": 8990 + }, + { + "epoch": 2.08, + "learning_rate": 3.268430783452739e-05, + "loss": 0.0, + "step": 8991 + }, + { + "epoch": 2.08, + "learning_rate": 3.2682381942839535e-05, + "loss": 0.0, + "step": 8992 + }, + { + "epoch": 2.08, + "learning_rate": 3.268045605115169e-05, + "loss": 0.0, + "step": 8993 + }, + { + "epoch": 2.08, + "learning_rate": 3.267853015946383e-05, + "loss": 0.0, + "step": 8994 + }, + { + "epoch": 2.08, + "learning_rate": 3.2676604267775984e-05, + "loss": 0.0, + "step": 8995 + }, + { + "epoch": 2.08, + "learning_rate": 3.267467837608813e-05, + "loss": 0.0001, + "step": 8996 + }, + { + "epoch": 2.08, + "learning_rate": 3.267275248440028e-05, + "loss": 0.0, + "step": 8997 + }, + { + "epoch": 2.08, + "learning_rate": 3.2670826592712426e-05, + "loss": 0.0, + "step": 8998 + }, + { + "epoch": 2.08, + "learning_rate": 3.2668900701024574e-05, + "loss": 0.0, + "step": 8999 + }, + { + "epoch": 2.08, + "learning_rate": 3.266697480933673e-05, + "loss": 0.0003, + "step": 9000 + }, + { + "epoch": 2.08, + "learning_rate": 3.266504891764887e-05, + "loss": 0.0, + "step": 9001 + }, + { + "epoch": 2.08, + "learning_rate": 3.266312302596102e-05, + "loss": 0.0, + "step": 9002 + }, + { + "epoch": 2.08, + "learning_rate": 3.266119713427317e-05, + "loss": 0.0, + "step": 9003 + }, + { + "epoch": 2.08, + "learning_rate": 3.265927124258532e-05, + "loss": 0.0012, + "step": 9004 + }, + { + "epoch": 2.08, + "learning_rate": 3.2657345350897466e-05, + "loss": 0.0001, + "step": 9005 + }, + { + "epoch": 2.08, + "learning_rate": 3.265541945920961e-05, + "loss": 0.0002, + "step": 9006 + }, + { + "epoch": 2.08, + "learning_rate": 3.265349356752177e-05, + "loss": 0.0, + "step": 9007 + }, + { + "epoch": 2.08, + "learning_rate": 3.265156767583391e-05, + "loss": 0.0, + "step": 9008 + }, + { + "epoch": 2.08, + "learning_rate": 3.264964178414606e-05, + "loss": 0.0, + "step": 9009 + }, + { + "epoch": 2.08, + "learning_rate": 3.264771589245821e-05, + "loss": 0.0, + "step": 9010 + }, + { + "epoch": 2.08, + "learning_rate": 3.264579000077036e-05, + "loss": 0.0, + "step": 9011 + }, + { + "epoch": 2.08, + "learning_rate": 3.2643864109082505e-05, + "loss": 0.0007, + "step": 9012 + }, + { + "epoch": 2.08, + "learning_rate": 3.264193821739466e-05, + "loss": 0.0, + "step": 9013 + }, + { + "epoch": 2.08, + "learning_rate": 3.2640012325706806e-05, + "loss": 0.0, + "step": 9014 + }, + { + "epoch": 2.08, + "learning_rate": 3.263808643401895e-05, + "loss": 0.0009, + "step": 9015 + }, + { + "epoch": 2.08, + "learning_rate": 3.26361605423311e-05, + "loss": 0.0001, + "step": 9016 + }, + { + "epoch": 2.08, + "learning_rate": 3.263423465064325e-05, + "loss": 0.0, + "step": 9017 + }, + { + "epoch": 2.08, + "learning_rate": 3.2632308758955396e-05, + "loss": 0.0001, + "step": 9018 + }, + { + "epoch": 2.08, + "learning_rate": 3.2630382867267544e-05, + "loss": 0.0, + "step": 9019 + }, + { + "epoch": 2.08, + "learning_rate": 3.26284569755797e-05, + "loss": 0.0004, + "step": 9020 + }, + { + "epoch": 2.08, + "learning_rate": 3.2626531083891846e-05, + "loss": 0.0, + "step": 9021 + }, + { + "epoch": 2.09, + "learning_rate": 3.262460519220399e-05, + "loss": 0.0, + "step": 9022 + }, + { + "epoch": 2.09, + "learning_rate": 3.262267930051614e-05, + "loss": 0.0037, + "step": 9023 + }, + { + "epoch": 2.09, + "learning_rate": 3.2620753408828295e-05, + "loss": 0.0, + "step": 9024 + }, + { + "epoch": 2.09, + "learning_rate": 3.2618827517140435e-05, + "loss": 0.0, + "step": 9025 + }, + { + "epoch": 2.09, + "learning_rate": 3.261690162545258e-05, + "loss": 0.0, + "step": 9026 + }, + { + "epoch": 2.09, + "learning_rate": 3.261497573376474e-05, + "loss": 0.0002, + "step": 9027 + }, + { + "epoch": 2.09, + "learning_rate": 3.2613049842076885e-05, + "loss": 0.0, + "step": 9028 + }, + { + "epoch": 2.09, + "learning_rate": 3.261112395038903e-05, + "loss": 0.0, + "step": 9029 + }, + { + "epoch": 2.09, + "learning_rate": 3.260919805870118e-05, + "loss": 0.0003, + "step": 9030 + }, + { + "epoch": 2.09, + "learning_rate": 3.2607272167013334e-05, + "loss": 0.0001, + "step": 9031 + }, + { + "epoch": 2.09, + "learning_rate": 3.2605346275325475e-05, + "loss": 0.0001, + "step": 9032 + }, + { + "epoch": 2.09, + "learning_rate": 3.260342038363763e-05, + "loss": 0.0, + "step": 9033 + }, + { + "epoch": 2.09, + "learning_rate": 3.2601494491949776e-05, + "loss": 0.0002, + "step": 9034 + }, + { + "epoch": 2.09, + "learning_rate": 3.2599568600261924e-05, + "loss": 0.0009, + "step": 9035 + }, + { + "epoch": 2.09, + "learning_rate": 3.259764270857407e-05, + "loss": 0.0, + "step": 9036 + }, + { + "epoch": 2.09, + "learning_rate": 3.259571681688622e-05, + "loss": 0.0001, + "step": 9037 + }, + { + "epoch": 2.09, + "learning_rate": 3.259379092519837e-05, + "loss": 0.0, + "step": 9038 + }, + { + "epoch": 2.09, + "learning_rate": 3.2591865033510514e-05, + "loss": 0.0, + "step": 9039 + }, + { + "epoch": 2.09, + "learning_rate": 3.258993914182267e-05, + "loss": 0.0, + "step": 9040 + }, + { + "epoch": 2.09, + "learning_rate": 3.2588013250134815e-05, + "loss": 0.0, + "step": 9041 + }, + { + "epoch": 2.09, + "learning_rate": 3.258608735844696e-05, + "loss": 0.0, + "step": 9042 + }, + { + "epoch": 2.09, + "learning_rate": 3.258416146675911e-05, + "loss": 0.0, + "step": 9043 + }, + { + "epoch": 2.09, + "learning_rate": 3.258223557507126e-05, + "loss": 0.0, + "step": 9044 + }, + { + "epoch": 2.09, + "learning_rate": 3.258030968338341e-05, + "loss": 0.0001, + "step": 9045 + }, + { + "epoch": 2.09, + "learning_rate": 3.257838379169555e-05, + "loss": 0.0002, + "step": 9046 + }, + { + "epoch": 2.09, + "learning_rate": 3.257645790000771e-05, + "loss": 0.0001, + "step": 9047 + }, + { + "epoch": 2.09, + "learning_rate": 3.2574532008319855e-05, + "loss": 0.0, + "step": 9048 + }, + { + "epoch": 2.09, + "learning_rate": 3.2572606116632e-05, + "loss": 0.0, + "step": 9049 + }, + { + "epoch": 2.09, + "learning_rate": 3.257068022494415e-05, + "loss": 0.0, + "step": 9050 + }, + { + "epoch": 2.09, + "learning_rate": 3.2568754333256304e-05, + "loss": 0.0003, + "step": 9051 + }, + { + "epoch": 2.09, + "learning_rate": 3.2566828441568444e-05, + "loss": 0.0, + "step": 9052 + }, + { + "epoch": 2.09, + "learning_rate": 3.256490254988059e-05, + "loss": 0.0, + "step": 9053 + }, + { + "epoch": 2.09, + "learning_rate": 3.2562976658192746e-05, + "loss": 0.0, + "step": 9054 + }, + { + "epoch": 2.09, + "learning_rate": 3.2561050766504894e-05, + "loss": 0.0, + "step": 9055 + }, + { + "epoch": 2.09, + "learning_rate": 3.255912487481704e-05, + "loss": 0.0003, + "step": 9056 + }, + { + "epoch": 2.09, + "learning_rate": 3.255719898312919e-05, + "loss": 0.0001, + "step": 9057 + }, + { + "epoch": 2.09, + "learning_rate": 3.255527309144134e-05, + "loss": 0.0, + "step": 9058 + }, + { + "epoch": 2.09, + "learning_rate": 3.2553347199753484e-05, + "loss": 0.0, + "step": 9059 + }, + { + "epoch": 2.09, + "learning_rate": 3.255142130806564e-05, + "loss": 0.0, + "step": 9060 + }, + { + "epoch": 2.09, + "learning_rate": 3.2549495416377785e-05, + "loss": 0.0, + "step": 9061 + }, + { + "epoch": 2.09, + "learning_rate": 3.254756952468993e-05, + "loss": 0.0028, + "step": 9062 + }, + { + "epoch": 2.09, + "learning_rate": 3.254564363300208e-05, + "loss": 0.0, + "step": 9063 + }, + { + "epoch": 2.09, + "learning_rate": 3.254371774131423e-05, + "loss": 0.0, + "step": 9064 + }, + { + "epoch": 2.09, + "learning_rate": 3.254179184962638e-05, + "loss": 0.0, + "step": 9065 + }, + { + "epoch": 2.1, + "learning_rate": 3.253986595793852e-05, + "loss": 0.0001, + "step": 9066 + }, + { + "epoch": 2.1, + "learning_rate": 3.253794006625068e-05, + "loss": 0.0, + "step": 9067 + }, + { + "epoch": 2.1, + "learning_rate": 3.2536014174562824e-05, + "loss": 0.0, + "step": 9068 + }, + { + "epoch": 2.1, + "learning_rate": 3.253408828287497e-05, + "loss": 0.0002, + "step": 9069 + }, + { + "epoch": 2.1, + "learning_rate": 3.253216239118712e-05, + "loss": 0.0001, + "step": 9070 + }, + { + "epoch": 2.1, + "learning_rate": 3.253023649949927e-05, + "loss": 0.0, + "step": 9071 + }, + { + "epoch": 2.1, + "learning_rate": 3.252831060781142e-05, + "loss": 0.0008, + "step": 9072 + }, + { + "epoch": 2.1, + "learning_rate": 3.252638471612356e-05, + "loss": 0.0, + "step": 9073 + }, + { + "epoch": 2.1, + "learning_rate": 3.2524458824435716e-05, + "loss": 0.0001, + "step": 9074 + }, + { + "epoch": 2.1, + "learning_rate": 3.2522532932747864e-05, + "loss": 0.0, + "step": 9075 + }, + { + "epoch": 2.1, + "learning_rate": 3.252060704106001e-05, + "loss": 0.0001, + "step": 9076 + }, + { + "epoch": 2.1, + "learning_rate": 3.251868114937216e-05, + "loss": 0.0013, + "step": 9077 + }, + { + "epoch": 2.1, + "learning_rate": 3.251675525768431e-05, + "loss": 0.0, + "step": 9078 + }, + { + "epoch": 2.1, + "learning_rate": 3.251482936599646e-05, + "loss": 0.0015, + "step": 9079 + }, + { + "epoch": 2.1, + "learning_rate": 3.251290347430861e-05, + "loss": 0.0, + "step": 9080 + }, + { + "epoch": 2.1, + "learning_rate": 3.2510977582620755e-05, + "loss": 0.0, + "step": 9081 + }, + { + "epoch": 2.1, + "learning_rate": 3.25090516909329e-05, + "loss": 0.0, + "step": 9082 + }, + { + "epoch": 2.1, + "learning_rate": 3.250712579924505e-05, + "loss": 0.0004, + "step": 9083 + }, + { + "epoch": 2.1, + "learning_rate": 3.25051999075572e-05, + "loss": 0.0, + "step": 9084 + }, + { + "epoch": 2.1, + "learning_rate": 3.250327401586935e-05, + "loss": 0.0, + "step": 9085 + }, + { + "epoch": 2.1, + "learning_rate": 3.25013481241815e-05, + "loss": 0.0, + "step": 9086 + }, + { + "epoch": 2.1, + "learning_rate": 3.249942223249365e-05, + "loss": 0.0, + "step": 9087 + }, + { + "epoch": 2.1, + "learning_rate": 3.2497496340805794e-05, + "loss": 0.0, + "step": 9088 + }, + { + "epoch": 2.1, + "learning_rate": 3.249557044911795e-05, + "loss": 0.0001, + "step": 9089 + }, + { + "epoch": 2.1, + "learning_rate": 3.249364455743009e-05, + "loss": 0.0, + "step": 9090 + }, + { + "epoch": 2.1, + "learning_rate": 3.249171866574224e-05, + "loss": 0.0, + "step": 9091 + }, + { + "epoch": 2.1, + "learning_rate": 3.248979277405439e-05, + "loss": 0.0001, + "step": 9092 + }, + { + "epoch": 2.1, + "learning_rate": 3.248786688236654e-05, + "loss": 0.0, + "step": 9093 + }, + { + "epoch": 2.1, + "learning_rate": 3.2485940990678686e-05, + "loss": 0.0, + "step": 9094 + }, + { + "epoch": 2.1, + "learning_rate": 3.2484015098990833e-05, + "loss": 0.0003, + "step": 9095 + }, + { + "epoch": 2.1, + "learning_rate": 3.248208920730299e-05, + "loss": 0.0003, + "step": 9096 + }, + { + "epoch": 2.1, + "learning_rate": 3.248016331561513e-05, + "loss": 0.0, + "step": 9097 + }, + { + "epoch": 2.1, + "learning_rate": 3.247823742392728e-05, + "loss": 0.0001, + "step": 9098 + }, + { + "epoch": 2.1, + "learning_rate": 3.247631153223943e-05, + "loss": 0.0001, + "step": 9099 + }, + { + "epoch": 2.1, + "learning_rate": 3.247438564055158e-05, + "loss": 0.0, + "step": 9100 + }, + { + "epoch": 2.1, + "learning_rate": 3.2472459748863725e-05, + "loss": 0.0001, + "step": 9101 + }, + { + "epoch": 2.1, + "learning_rate": 3.247053385717587e-05, + "loss": 0.0002, + "step": 9102 + }, + { + "epoch": 2.1, + "learning_rate": 3.246860796548803e-05, + "loss": 0.0, + "step": 9103 + }, + { + "epoch": 2.1, + "learning_rate": 3.246668207380017e-05, + "loss": 0.0, + "step": 9104 + }, + { + "epoch": 2.1, + "learning_rate": 3.246475618211232e-05, + "loss": 0.0003, + "step": 9105 + }, + { + "epoch": 2.1, + "learning_rate": 3.246283029042447e-05, + "loss": 0.0, + "step": 9106 + }, + { + "epoch": 2.1, + "learning_rate": 3.246090439873662e-05, + "loss": 0.0, + "step": 9107 + }, + { + "epoch": 2.1, + "learning_rate": 3.2458978507048764e-05, + "loss": 0.0001, + "step": 9108 + }, + { + "epoch": 2.11, + "learning_rate": 3.245705261536091e-05, + "loss": 0.0, + "step": 9109 + }, + { + "epoch": 2.11, + "learning_rate": 3.2455126723673066e-05, + "loss": 0.0036, + "step": 9110 + }, + { + "epoch": 2.11, + "learning_rate": 3.245320083198521e-05, + "loss": 0.0001, + "step": 9111 + }, + { + "epoch": 2.11, + "learning_rate": 3.245127494029736e-05, + "loss": 0.0, + "step": 9112 + }, + { + "epoch": 2.11, + "learning_rate": 3.244934904860951e-05, + "loss": 0.0, + "step": 9113 + }, + { + "epoch": 2.11, + "learning_rate": 3.2447423156921656e-05, + "loss": 0.0001, + "step": 9114 + }, + { + "epoch": 2.11, + "learning_rate": 3.24454972652338e-05, + "loss": 0.0, + "step": 9115 + }, + { + "epoch": 2.11, + "learning_rate": 3.244357137354596e-05, + "loss": 0.0003, + "step": 9116 + }, + { + "epoch": 2.11, + "learning_rate": 3.2441645481858105e-05, + "loss": 0.0005, + "step": 9117 + }, + { + "epoch": 2.11, + "learning_rate": 3.2439719590170246e-05, + "loss": 0.0006, + "step": 9118 + }, + { + "epoch": 2.11, + "learning_rate": 3.24377936984824e-05, + "loss": 0.0007, + "step": 9119 + }, + { + "epoch": 2.11, + "learning_rate": 3.243586780679455e-05, + "loss": 0.0001, + "step": 9120 + }, + { + "epoch": 2.11, + "learning_rate": 3.2433941915106695e-05, + "loss": 0.0, + "step": 9121 + }, + { + "epoch": 2.11, + "learning_rate": 3.243201602341884e-05, + "loss": 0.0, + "step": 9122 + }, + { + "epoch": 2.11, + "learning_rate": 3.2430090131731e-05, + "loss": 0.0, + "step": 9123 + }, + { + "epoch": 2.11, + "learning_rate": 3.242816424004314e-05, + "loss": 0.0, + "step": 9124 + }, + { + "epoch": 2.11, + "learning_rate": 3.242623834835529e-05, + "loss": 0.0, + "step": 9125 + }, + { + "epoch": 2.11, + "learning_rate": 3.242431245666744e-05, + "loss": 0.0, + "step": 9126 + }, + { + "epoch": 2.11, + "learning_rate": 3.2422386564979587e-05, + "loss": 0.0, + "step": 9127 + }, + { + "epoch": 2.11, + "learning_rate": 3.2420460673291734e-05, + "loss": 0.0, + "step": 9128 + }, + { + "epoch": 2.11, + "learning_rate": 3.241853478160388e-05, + "loss": 0.0113, + "step": 9129 + }, + { + "epoch": 2.11, + "learning_rate": 3.2416608889916036e-05, + "loss": 0.0004, + "step": 9130 + }, + { + "epoch": 2.11, + "learning_rate": 3.2414682998228176e-05, + "loss": 0.0, + "step": 9131 + }, + { + "epoch": 2.11, + "learning_rate": 3.241275710654033e-05, + "loss": 0.0, + "step": 9132 + }, + { + "epoch": 2.11, + "learning_rate": 3.241083121485248e-05, + "loss": 0.0, + "step": 9133 + }, + { + "epoch": 2.11, + "learning_rate": 3.2408905323164626e-05, + "loss": 0.0, + "step": 9134 + }, + { + "epoch": 2.11, + "learning_rate": 3.240697943147677e-05, + "loss": 0.0, + "step": 9135 + }, + { + "epoch": 2.11, + "learning_rate": 3.240505353978893e-05, + "loss": 0.0, + "step": 9136 + }, + { + "epoch": 2.11, + "learning_rate": 3.2403127648101075e-05, + "loss": 0.0, + "step": 9137 + }, + { + "epoch": 2.11, + "learning_rate": 3.2401201756413216e-05, + "loss": 0.0, + "step": 9138 + }, + { + "epoch": 2.11, + "learning_rate": 3.239927586472537e-05, + "loss": 0.0, + "step": 9139 + }, + { + "epoch": 2.11, + "learning_rate": 3.239734997303752e-05, + "loss": 0.0, + "step": 9140 + }, + { + "epoch": 2.11, + "learning_rate": 3.2395424081349665e-05, + "loss": 0.0, + "step": 9141 + }, + { + "epoch": 2.11, + "learning_rate": 3.239349818966181e-05, + "loss": 0.0003, + "step": 9142 + }, + { + "epoch": 2.11, + "learning_rate": 3.2391572297973967e-05, + "loss": 0.0002, + "step": 9143 + }, + { + "epoch": 2.11, + "learning_rate": 3.2389646406286114e-05, + "loss": 0.0, + "step": 9144 + }, + { + "epoch": 2.11, + "learning_rate": 3.238772051459826e-05, + "loss": 0.0005, + "step": 9145 + }, + { + "epoch": 2.11, + "learning_rate": 3.238579462291041e-05, + "loss": 0.0, + "step": 9146 + }, + { + "epoch": 2.11, + "learning_rate": 3.2383868731222556e-05, + "loss": 0.0, + "step": 9147 + }, + { + "epoch": 2.11, + "learning_rate": 3.2381942839534704e-05, + "loss": 0.0002, + "step": 9148 + }, + { + "epoch": 2.11, + "learning_rate": 3.238001694784685e-05, + "loss": 0.0002, + "step": 9149 + }, + { + "epoch": 2.11, + "learning_rate": 3.2378091056159006e-05, + "loss": 0.0, + "step": 9150 + }, + { + "epoch": 2.11, + "learning_rate": 3.237616516447115e-05, + "loss": 0.0001, + "step": 9151 + }, + { + "epoch": 2.12, + "learning_rate": 3.23742392727833e-05, + "loss": 0.0009, + "step": 9152 + }, + { + "epoch": 2.12, + "learning_rate": 3.237231338109545e-05, + "loss": 0.0, + "step": 9153 + }, + { + "epoch": 2.12, + "learning_rate": 3.23703874894076e-05, + "loss": 0.001, + "step": 9154 + }, + { + "epoch": 2.12, + "learning_rate": 3.236846159771974e-05, + "loss": 0.0, + "step": 9155 + }, + { + "epoch": 2.12, + "learning_rate": 3.236653570603189e-05, + "loss": 0.0001, + "step": 9156 + }, + { + "epoch": 2.12, + "learning_rate": 3.2364609814344045e-05, + "loss": 0.0, + "step": 9157 + }, + { + "epoch": 2.12, + "learning_rate": 3.236268392265619e-05, + "loss": 0.0, + "step": 9158 + }, + { + "epoch": 2.12, + "learning_rate": 3.236075803096834e-05, + "loss": 0.0, + "step": 9159 + }, + { + "epoch": 2.12, + "learning_rate": 3.235883213928049e-05, + "loss": 0.0, + "step": 9160 + }, + { + "epoch": 2.12, + "learning_rate": 3.235690624759264e-05, + "loss": 0.0, + "step": 9161 + }, + { + "epoch": 2.12, + "learning_rate": 3.235498035590478e-05, + "loss": 0.0, + "step": 9162 + }, + { + "epoch": 2.12, + "learning_rate": 3.2353054464216936e-05, + "loss": 0.0, + "step": 9163 + }, + { + "epoch": 2.12, + "learning_rate": 3.2351128572529084e-05, + "loss": 0.0, + "step": 9164 + }, + { + "epoch": 2.12, + "learning_rate": 3.234920268084123e-05, + "loss": 0.0001, + "step": 9165 + }, + { + "epoch": 2.12, + "learning_rate": 3.234727678915338e-05, + "loss": 0.0, + "step": 9166 + }, + { + "epoch": 2.12, + "learning_rate": 3.2345350897465526e-05, + "loss": 0.0001, + "step": 9167 + }, + { + "epoch": 2.12, + "learning_rate": 3.234342500577768e-05, + "loss": 0.0001, + "step": 9168 + }, + { + "epoch": 2.12, + "learning_rate": 3.234149911408982e-05, + "loss": 0.0, + "step": 9169 + }, + { + "epoch": 2.12, + "learning_rate": 3.2339573222401976e-05, + "loss": 0.0017, + "step": 9170 + }, + { + "epoch": 2.12, + "learning_rate": 3.233764733071412e-05, + "loss": 0.0001, + "step": 9171 + }, + { + "epoch": 2.12, + "learning_rate": 3.233572143902627e-05, + "loss": 0.0, + "step": 9172 + }, + { + "epoch": 2.12, + "learning_rate": 3.233379554733842e-05, + "loss": 0.0001, + "step": 9173 + }, + { + "epoch": 2.12, + "learning_rate": 3.2331869655650565e-05, + "loss": 0.0, + "step": 9174 + }, + { + "epoch": 2.12, + "learning_rate": 3.232994376396272e-05, + "loss": 0.0014, + "step": 9175 + }, + { + "epoch": 2.12, + "learning_rate": 3.232801787227486e-05, + "loss": 0.0, + "step": 9176 + }, + { + "epoch": 2.12, + "learning_rate": 3.2326091980587015e-05, + "loss": 0.0001, + "step": 9177 + }, + { + "epoch": 2.12, + "learning_rate": 3.232416608889916e-05, + "loss": 0.0004, + "step": 9178 + }, + { + "epoch": 2.12, + "learning_rate": 3.232224019721131e-05, + "loss": 0.0, + "step": 9179 + }, + { + "epoch": 2.12, + "learning_rate": 3.232031430552346e-05, + "loss": 0.0001, + "step": 9180 + }, + { + "epoch": 2.12, + "learning_rate": 3.231838841383561e-05, + "loss": 0.0, + "step": 9181 + }, + { + "epoch": 2.12, + "learning_rate": 3.231646252214776e-05, + "loss": 0.0002, + "step": 9182 + }, + { + "epoch": 2.12, + "learning_rate": 3.23145366304599e-05, + "loss": 0.0001, + "step": 9183 + }, + { + "epoch": 2.12, + "learning_rate": 3.2312610738772054e-05, + "loss": 0.0, + "step": 9184 + }, + { + "epoch": 2.12, + "learning_rate": 3.23106848470842e-05, + "loss": 0.0, + "step": 9185 + }, + { + "epoch": 2.12, + "learning_rate": 3.230875895539635e-05, + "loss": 0.0, + "step": 9186 + }, + { + "epoch": 2.12, + "learning_rate": 3.2306833063708496e-05, + "loss": 0.0001, + "step": 9187 + }, + { + "epoch": 2.12, + "learning_rate": 3.230490717202065e-05, + "loss": 0.0, + "step": 9188 + }, + { + "epoch": 2.12, + "learning_rate": 3.23029812803328e-05, + "loss": 0.0, + "step": 9189 + }, + { + "epoch": 2.12, + "learning_rate": 3.2301055388644945e-05, + "loss": 0.0, + "step": 9190 + }, + { + "epoch": 2.12, + "learning_rate": 3.229912949695709e-05, + "loss": 0.0001, + "step": 9191 + }, + { + "epoch": 2.12, + "learning_rate": 3.229720360526925e-05, + "loss": 0.0, + "step": 9192 + }, + { + "epoch": 2.12, + "learning_rate": 3.229527771358139e-05, + "loss": 0.0, + "step": 9193 + }, + { + "epoch": 2.12, + "learning_rate": 3.2293351821893535e-05, + "loss": 0.0013, + "step": 9194 + }, + { + "epoch": 2.13, + "learning_rate": 3.229142593020569e-05, + "loss": 0.0001, + "step": 9195 + }, + { + "epoch": 2.13, + "learning_rate": 3.228950003851783e-05, + "loss": 0.0, + "step": 9196 + }, + { + "epoch": 2.13, + "learning_rate": 3.2287574146829985e-05, + "loss": 0.0001, + "step": 9197 + }, + { + "epoch": 2.13, + "learning_rate": 3.228564825514213e-05, + "loss": 0.0001, + "step": 9198 + }, + { + "epoch": 2.13, + "learning_rate": 3.2283722363454286e-05, + "loss": 0.0, + "step": 9199 + }, + { + "epoch": 2.13, + "learning_rate": 3.228179647176643e-05, + "loss": 0.0001, + "step": 9200 + }, + { + "epoch": 2.13, + "learning_rate": 3.227987058007858e-05, + "loss": 0.0, + "step": 9201 + }, + { + "epoch": 2.13, + "learning_rate": 3.227794468839073e-05, + "loss": 0.0, + "step": 9202 + }, + { + "epoch": 2.13, + "learning_rate": 3.227601879670287e-05, + "loss": 0.0, + "step": 9203 + }, + { + "epoch": 2.13, + "learning_rate": 3.2274092905015024e-05, + "loss": 0.0, + "step": 9204 + }, + { + "epoch": 2.13, + "learning_rate": 3.227216701332717e-05, + "loss": 0.0, + "step": 9205 + }, + { + "epoch": 2.13, + "learning_rate": 3.227024112163932e-05, + "loss": 0.0, + "step": 9206 + }, + { + "epoch": 2.13, + "learning_rate": 3.2268315229951466e-05, + "loss": 0.0, + "step": 9207 + }, + { + "epoch": 2.13, + "learning_rate": 3.226638933826362e-05, + "loss": 0.0001, + "step": 9208 + }, + { + "epoch": 2.13, + "learning_rate": 3.226446344657577e-05, + "loss": 0.0, + "step": 9209 + }, + { + "epoch": 2.13, + "learning_rate": 3.2262537554887915e-05, + "loss": 0.0, + "step": 9210 + }, + { + "epoch": 2.13, + "learning_rate": 3.226061166320006e-05, + "loss": 0.0, + "step": 9211 + }, + { + "epoch": 2.13, + "learning_rate": 3.225868577151221e-05, + "loss": 0.0001, + "step": 9212 + }, + { + "epoch": 2.13, + "learning_rate": 3.225675987982436e-05, + "loss": 0.0, + "step": 9213 + }, + { + "epoch": 2.13, + "learning_rate": 3.2254833988136505e-05, + "loss": 0.0, + "step": 9214 + }, + { + "epoch": 2.13, + "learning_rate": 3.225290809644866e-05, + "loss": 0.0, + "step": 9215 + }, + { + "epoch": 2.13, + "learning_rate": 3.225098220476081e-05, + "loss": 0.0, + "step": 9216 + }, + { + "epoch": 2.13, + "learning_rate": 3.2249056313072954e-05, + "loss": 0.0, + "step": 9217 + }, + { + "epoch": 2.13, + "learning_rate": 3.22471304213851e-05, + "loss": 0.0003, + "step": 9218 + }, + { + "epoch": 2.13, + "learning_rate": 3.2245204529697256e-05, + "loss": 0.0001, + "step": 9219 + }, + { + "epoch": 2.13, + "learning_rate": 3.22432786380094e-05, + "loss": 0.0, + "step": 9220 + }, + { + "epoch": 2.13, + "learning_rate": 3.2241352746321544e-05, + "loss": 0.0, + "step": 9221 + }, + { + "epoch": 2.13, + "learning_rate": 3.22394268546337e-05, + "loss": 0.0007, + "step": 9222 + }, + { + "epoch": 2.13, + "learning_rate": 3.2237500962945846e-05, + "loss": 0.0, + "step": 9223 + }, + { + "epoch": 2.13, + "learning_rate": 3.2235575071257994e-05, + "loss": 0.0, + "step": 9224 + }, + { + "epoch": 2.13, + "learning_rate": 3.223364917957014e-05, + "loss": 0.0, + "step": 9225 + }, + { + "epoch": 2.13, + "learning_rate": 3.2231723287882295e-05, + "loss": 0.0, + "step": 9226 + }, + { + "epoch": 2.13, + "learning_rate": 3.2229797396194436e-05, + "loss": 0.0, + "step": 9227 + }, + { + "epoch": 2.13, + "learning_rate": 3.222787150450659e-05, + "loss": 0.0, + "step": 9228 + }, + { + "epoch": 2.13, + "learning_rate": 3.222594561281874e-05, + "loss": 0.0, + "step": 9229 + }, + { + "epoch": 2.13, + "learning_rate": 3.2224019721130885e-05, + "loss": 0.0001, + "step": 9230 + }, + { + "epoch": 2.13, + "learning_rate": 3.222209382944303e-05, + "loss": 0.0, + "step": 9231 + }, + { + "epoch": 2.13, + "learning_rate": 3.222016793775518e-05, + "loss": 0.0, + "step": 9232 + }, + { + "epoch": 2.13, + "learning_rate": 3.2218242046067334e-05, + "loss": 0.0, + "step": 9233 + }, + { + "epoch": 2.13, + "learning_rate": 3.2216316154379475e-05, + "loss": 0.0, + "step": 9234 + }, + { + "epoch": 2.13, + "learning_rate": 3.221439026269163e-05, + "loss": 0.0005, + "step": 9235 + }, + { + "epoch": 2.13, + "learning_rate": 3.221246437100378e-05, + "loss": 0.0, + "step": 9236 + }, + { + "epoch": 2.13, + "learning_rate": 3.2210538479315924e-05, + "loss": 0.0, + "step": 9237 + }, + { + "epoch": 2.13, + "learning_rate": 3.220861258762807e-05, + "loss": 0.0, + "step": 9238 + }, + { + "epoch": 2.14, + "learning_rate": 3.220668669594022e-05, + "loss": 0.0, + "step": 9239 + }, + { + "epoch": 2.14, + "learning_rate": 3.2204760804252374e-05, + "loss": 0.0002, + "step": 9240 + }, + { + "epoch": 2.14, + "learning_rate": 3.2202834912564514e-05, + "loss": 0.0, + "step": 9241 + }, + { + "epoch": 2.14, + "learning_rate": 3.220090902087667e-05, + "loss": 0.0, + "step": 9242 + }, + { + "epoch": 2.14, + "learning_rate": 3.2198983129188816e-05, + "loss": 0.0001, + "step": 9243 + }, + { + "epoch": 2.14, + "learning_rate": 3.2197057237500963e-05, + "loss": 0.0, + "step": 9244 + }, + { + "epoch": 2.14, + "learning_rate": 3.219513134581311e-05, + "loss": 0.0, + "step": 9245 + }, + { + "epoch": 2.14, + "learning_rate": 3.2193205454125265e-05, + "loss": 0.0, + "step": 9246 + }, + { + "epoch": 2.14, + "learning_rate": 3.219127956243741e-05, + "loss": 0.0, + "step": 9247 + }, + { + "epoch": 2.14, + "learning_rate": 3.218935367074956e-05, + "loss": 0.0007, + "step": 9248 + }, + { + "epoch": 2.14, + "learning_rate": 3.218742777906171e-05, + "loss": 0.0, + "step": 9249 + }, + { + "epoch": 2.14, + "learning_rate": 3.2185501887373855e-05, + "loss": 0.0001, + "step": 9250 + }, + { + "epoch": 2.14, + "learning_rate": 3.2183575995686e-05, + "loss": 0.0001, + "step": 9251 + }, + { + "epoch": 2.14, + "learning_rate": 3.218165010399815e-05, + "loss": 0.0, + "step": 9252 + }, + { + "epoch": 2.14, + "learning_rate": 3.2179724212310304e-05, + "loss": 0.0, + "step": 9253 + }, + { + "epoch": 2.14, + "learning_rate": 3.217779832062245e-05, + "loss": 0.0001, + "step": 9254 + }, + { + "epoch": 2.14, + "learning_rate": 3.21758724289346e-05, + "loss": 0.0, + "step": 9255 + }, + { + "epoch": 2.14, + "learning_rate": 3.217394653724675e-05, + "loss": 0.0, + "step": 9256 + }, + { + "epoch": 2.14, + "learning_rate": 3.21720206455589e-05, + "loss": 0.0, + "step": 9257 + }, + { + "epoch": 2.14, + "learning_rate": 3.217009475387104e-05, + "loss": 0.0, + "step": 9258 + }, + { + "epoch": 2.14, + "learning_rate": 3.216816886218319e-05, + "loss": 0.0, + "step": 9259 + }, + { + "epoch": 2.14, + "learning_rate": 3.2166242970495343e-05, + "loss": 0.0, + "step": 9260 + }, + { + "epoch": 2.14, + "learning_rate": 3.216431707880749e-05, + "loss": 0.0, + "step": 9261 + }, + { + "epoch": 2.14, + "learning_rate": 3.216239118711964e-05, + "loss": 0.0, + "step": 9262 + }, + { + "epoch": 2.14, + "learning_rate": 3.2160465295431786e-05, + "loss": 0.0017, + "step": 9263 + }, + { + "epoch": 2.14, + "learning_rate": 3.215853940374394e-05, + "loss": 0.0002, + "step": 9264 + }, + { + "epoch": 2.14, + "learning_rate": 3.215661351205608e-05, + "loss": 0.0, + "step": 9265 + }, + { + "epoch": 2.14, + "learning_rate": 3.2154687620368235e-05, + "loss": 0.0004, + "step": 9266 + }, + { + "epoch": 2.14, + "learning_rate": 3.215276172868038e-05, + "loss": 0.0, + "step": 9267 + }, + { + "epoch": 2.14, + "learning_rate": 3.215083583699253e-05, + "loss": 0.0002, + "step": 9268 + }, + { + "epoch": 2.14, + "learning_rate": 3.214890994530468e-05, + "loss": 0.0, + "step": 9269 + }, + { + "epoch": 2.14, + "learning_rate": 3.2146984053616825e-05, + "loss": 0.0, + "step": 9270 + }, + { + "epoch": 2.14, + "learning_rate": 3.214505816192898e-05, + "loss": 0.0, + "step": 9271 + }, + { + "epoch": 2.14, + "learning_rate": 3.214313227024112e-05, + "loss": 0.0, + "step": 9272 + }, + { + "epoch": 2.14, + "learning_rate": 3.2141206378553274e-05, + "loss": 0.0, + "step": 9273 + }, + { + "epoch": 2.14, + "learning_rate": 3.213928048686542e-05, + "loss": 0.0, + "step": 9274 + }, + { + "epoch": 2.14, + "learning_rate": 3.213735459517757e-05, + "loss": 0.0002, + "step": 9275 + }, + { + "epoch": 2.14, + "learning_rate": 3.213542870348972e-05, + "loss": 0.0002, + "step": 9276 + }, + { + "epoch": 2.14, + "learning_rate": 3.2133502811801864e-05, + "loss": 0.0001, + "step": 9277 + }, + { + "epoch": 2.14, + "learning_rate": 3.213157692011401e-05, + "loss": 0.0, + "step": 9278 + }, + { + "epoch": 2.14, + "learning_rate": 3.212965102842616e-05, + "loss": 0.0, + "step": 9279 + }, + { + "epoch": 2.14, + "learning_rate": 3.212772513673831e-05, + "loss": 0.0, + "step": 9280 + }, + { + "epoch": 2.14, + "learning_rate": 3.212579924505046e-05, + "loss": 0.0001, + "step": 9281 + }, + { + "epoch": 2.15, + "learning_rate": 3.212387335336261e-05, + "loss": 0.0, + "step": 9282 + }, + { + "epoch": 2.15, + "learning_rate": 3.2121947461674756e-05, + "loss": 0.0, + "step": 9283 + }, + { + "epoch": 2.15, + "learning_rate": 3.212002156998691e-05, + "loss": 0.0001, + "step": 9284 + }, + { + "epoch": 2.15, + "learning_rate": 3.211809567829905e-05, + "loss": 0.0004, + "step": 9285 + }, + { + "epoch": 2.15, + "learning_rate": 3.21161697866112e-05, + "loss": 0.0, + "step": 9286 + }, + { + "epoch": 2.15, + "learning_rate": 3.211424389492335e-05, + "loss": 0.0003, + "step": 9287 + }, + { + "epoch": 2.15, + "learning_rate": 3.21123180032355e-05, + "loss": 0.0001, + "step": 9288 + }, + { + "epoch": 2.15, + "learning_rate": 3.211039211154765e-05, + "loss": 0.0001, + "step": 9289 + }, + { + "epoch": 2.15, + "learning_rate": 3.2108466219859795e-05, + "loss": 0.0001, + "step": 9290 + }, + { + "epoch": 2.15, + "learning_rate": 3.210654032817195e-05, + "loss": 0.0, + "step": 9291 + }, + { + "epoch": 2.15, + "learning_rate": 3.210461443648409e-05, + "loss": 0.0, + "step": 9292 + }, + { + "epoch": 2.15, + "learning_rate": 3.2102688544796244e-05, + "loss": 0.0001, + "step": 9293 + }, + { + "epoch": 2.15, + "learning_rate": 3.210076265310839e-05, + "loss": 0.0, + "step": 9294 + }, + { + "epoch": 2.15, + "learning_rate": 3.209883676142054e-05, + "loss": 0.0004, + "step": 9295 + }, + { + "epoch": 2.15, + "learning_rate": 3.2096910869732687e-05, + "loss": 0.0, + "step": 9296 + }, + { + "epoch": 2.15, + "learning_rate": 3.2094984978044834e-05, + "loss": 0.0, + "step": 9297 + }, + { + "epoch": 2.15, + "learning_rate": 3.209305908635699e-05, + "loss": 0.0, + "step": 9298 + }, + { + "epoch": 2.15, + "learning_rate": 3.209113319466913e-05, + "loss": 0.0, + "step": 9299 + }, + { + "epoch": 2.15, + "learning_rate": 3.208920730298128e-05, + "loss": 0.0, + "step": 9300 + }, + { + "epoch": 2.15, + "learning_rate": 3.208728141129343e-05, + "loss": 0.0001, + "step": 9301 + }, + { + "epoch": 2.15, + "learning_rate": 3.208535551960558e-05, + "loss": 0.0001, + "step": 9302 + }, + { + "epoch": 2.15, + "learning_rate": 3.2083429627917726e-05, + "loss": 0.0018, + "step": 9303 + }, + { + "epoch": 2.15, + "learning_rate": 3.208150373622988e-05, + "loss": 0.0, + "step": 9304 + }, + { + "epoch": 2.15, + "learning_rate": 3.207957784454203e-05, + "loss": 0.0008, + "step": 9305 + }, + { + "epoch": 2.15, + "learning_rate": 3.207765195285417e-05, + "loss": 0.0, + "step": 9306 + }, + { + "epoch": 2.15, + "learning_rate": 3.207572606116632e-05, + "loss": 0.0, + "step": 9307 + }, + { + "epoch": 2.15, + "learning_rate": 3.207380016947847e-05, + "loss": 0.0003, + "step": 9308 + }, + { + "epoch": 2.15, + "learning_rate": 3.207187427779062e-05, + "loss": 0.0009, + "step": 9309 + }, + { + "epoch": 2.15, + "learning_rate": 3.2069948386102765e-05, + "loss": 0.0, + "step": 9310 + }, + { + "epoch": 2.15, + "learning_rate": 3.206802249441492e-05, + "loss": 0.0002, + "step": 9311 + }, + { + "epoch": 2.15, + "learning_rate": 3.2066096602727066e-05, + "loss": 0.0, + "step": 9312 + }, + { + "epoch": 2.15, + "learning_rate": 3.2064170711039214e-05, + "loss": 0.0, + "step": 9313 + }, + { + "epoch": 2.15, + "learning_rate": 3.206224481935136e-05, + "loss": 0.0001, + "step": 9314 + }, + { + "epoch": 2.15, + "learning_rate": 3.206031892766351e-05, + "loss": 0.0, + "step": 9315 + }, + { + "epoch": 2.15, + "learning_rate": 3.2058393035975656e-05, + "loss": 0.0001, + "step": 9316 + }, + { + "epoch": 2.15, + "learning_rate": 3.2056467144287804e-05, + "loss": 0.0, + "step": 9317 + }, + { + "epoch": 2.15, + "learning_rate": 3.205454125259996e-05, + "loss": 0.0, + "step": 9318 + }, + { + "epoch": 2.15, + "learning_rate": 3.2052615360912106e-05, + "loss": 0.0, + "step": 9319 + }, + { + "epoch": 2.15, + "learning_rate": 3.205068946922425e-05, + "loss": 0.0, + "step": 9320 + }, + { + "epoch": 2.15, + "learning_rate": 3.20487635775364e-05, + "loss": 0.0, + "step": 9321 + }, + { + "epoch": 2.15, + "learning_rate": 3.2046837685848555e-05, + "loss": 0.0003, + "step": 9322 + }, + { + "epoch": 2.15, + "learning_rate": 3.2044911794160696e-05, + "loss": 0.0, + "step": 9323 + }, + { + "epoch": 2.15, + "learning_rate": 3.204298590247284e-05, + "loss": 0.0, + "step": 9324 + }, + { + "epoch": 2.16, + "learning_rate": 3.2041060010785e-05, + "loss": 0.0, + "step": 9325 + }, + { + "epoch": 2.16, + "learning_rate": 3.2039134119097145e-05, + "loss": 0.0001, + "step": 9326 + }, + { + "epoch": 2.16, + "learning_rate": 3.203720822740929e-05, + "loss": 0.0, + "step": 9327 + }, + { + "epoch": 2.16, + "learning_rate": 3.203528233572144e-05, + "loss": 0.0001, + "step": 9328 + }, + { + "epoch": 2.16, + "learning_rate": 3.2033356444033594e-05, + "loss": 0.0, + "step": 9329 + }, + { + "epoch": 2.16, + "learning_rate": 3.2031430552345735e-05, + "loss": 0.0, + "step": 9330 + }, + { + "epoch": 2.16, + "learning_rate": 3.202950466065789e-05, + "loss": 0.0, + "step": 9331 + }, + { + "epoch": 2.16, + "learning_rate": 3.2027578768970036e-05, + "loss": 0.0, + "step": 9332 + }, + { + "epoch": 2.16, + "learning_rate": 3.2025652877282184e-05, + "loss": 0.0, + "step": 9333 + }, + { + "epoch": 2.16, + "learning_rate": 3.202372698559433e-05, + "loss": 0.0, + "step": 9334 + }, + { + "epoch": 2.16, + "learning_rate": 3.202180109390648e-05, + "loss": 0.0, + "step": 9335 + }, + { + "epoch": 2.16, + "learning_rate": 3.201987520221863e-05, + "loss": 0.0, + "step": 9336 + }, + { + "epoch": 2.16, + "learning_rate": 3.2017949310530774e-05, + "loss": 0.0, + "step": 9337 + }, + { + "epoch": 2.16, + "learning_rate": 3.201602341884293e-05, + "loss": 0.0001, + "step": 9338 + }, + { + "epoch": 2.16, + "learning_rate": 3.2014097527155075e-05, + "loss": 0.0, + "step": 9339 + }, + { + "epoch": 2.16, + "learning_rate": 3.201217163546722e-05, + "loss": 0.0, + "step": 9340 + }, + { + "epoch": 2.16, + "learning_rate": 3.201024574377937e-05, + "loss": 0.0001, + "step": 9341 + }, + { + "epoch": 2.16, + "learning_rate": 3.200831985209152e-05, + "loss": 0.0001, + "step": 9342 + }, + { + "epoch": 2.16, + "learning_rate": 3.200639396040367e-05, + "loss": 0.0, + "step": 9343 + }, + { + "epoch": 2.16, + "learning_rate": 3.200446806871581e-05, + "loss": 0.0001, + "step": 9344 + }, + { + "epoch": 2.16, + "learning_rate": 3.200254217702797e-05, + "loss": 0.0003, + "step": 9345 + }, + { + "epoch": 2.16, + "learning_rate": 3.2000616285340115e-05, + "loss": 0.0, + "step": 9346 + }, + { + "epoch": 2.16, + "learning_rate": 3.199869039365226e-05, + "loss": 0.0, + "step": 9347 + }, + { + "epoch": 2.16, + "learning_rate": 3.199676450196441e-05, + "loss": 0.0001, + "step": 9348 + }, + { + "epoch": 2.16, + "learning_rate": 3.1994838610276564e-05, + "loss": 0.0001, + "step": 9349 + }, + { + "epoch": 2.16, + "learning_rate": 3.1992912718588705e-05, + "loss": 0.0, + "step": 9350 + }, + { + "epoch": 2.16, + "learning_rate": 3.199098682690085e-05, + "loss": 0.0002, + "step": 9351 + }, + { + "epoch": 2.16, + "learning_rate": 3.1989060935213006e-05, + "loss": 0.0, + "step": 9352 + }, + { + "epoch": 2.16, + "learning_rate": 3.1987135043525154e-05, + "loss": 0.0003, + "step": 9353 + }, + { + "epoch": 2.16, + "learning_rate": 3.19852091518373e-05, + "loss": 0.0, + "step": 9354 + }, + { + "epoch": 2.16, + "learning_rate": 3.198328326014945e-05, + "loss": 0.0, + "step": 9355 + }, + { + "epoch": 2.16, + "learning_rate": 3.19813573684616e-05, + "loss": 0.0, + "step": 9356 + }, + { + "epoch": 2.16, + "learning_rate": 3.1979431476773744e-05, + "loss": 0.0008, + "step": 9357 + }, + { + "epoch": 2.16, + "learning_rate": 3.19775055850859e-05, + "loss": 0.0, + "step": 9358 + }, + { + "epoch": 2.16, + "learning_rate": 3.1975579693398045e-05, + "loss": 0.0001, + "step": 9359 + }, + { + "epoch": 2.16, + "learning_rate": 3.197365380171019e-05, + "loss": 0.0001, + "step": 9360 + }, + { + "epoch": 2.16, + "learning_rate": 3.197172791002234e-05, + "loss": 0.0, + "step": 9361 + }, + { + "epoch": 2.16, + "learning_rate": 3.196980201833449e-05, + "loss": 0.0, + "step": 9362 + }, + { + "epoch": 2.16, + "learning_rate": 3.196787612664664e-05, + "loss": 0.0001, + "step": 9363 + }, + { + "epoch": 2.16, + "learning_rate": 3.196595023495878e-05, + "loss": 0.0, + "step": 9364 + }, + { + "epoch": 2.16, + "learning_rate": 3.196402434327094e-05, + "loss": 0.0003, + "step": 9365 + }, + { + "epoch": 2.16, + "learning_rate": 3.1962098451583084e-05, + "loss": 0.0, + "step": 9366 + }, + { + "epoch": 2.16, + "learning_rate": 3.196017255989523e-05, + "loss": 0.0001, + "step": 9367 + }, + { + "epoch": 2.17, + "learning_rate": 3.195824666820738e-05, + "loss": 0.0003, + "step": 9368 + }, + { + "epoch": 2.17, + "learning_rate": 3.1956320776519534e-05, + "loss": 0.0, + "step": 9369 + }, + { + "epoch": 2.17, + "learning_rate": 3.195439488483168e-05, + "loss": 0.0, + "step": 9370 + }, + { + "epoch": 2.17, + "learning_rate": 3.195246899314382e-05, + "loss": 0.0, + "step": 9371 + }, + { + "epoch": 2.17, + "learning_rate": 3.1950543101455976e-05, + "loss": 0.0004, + "step": 9372 + }, + { + "epoch": 2.17, + "learning_rate": 3.1948617209768124e-05, + "loss": 0.0, + "step": 9373 + }, + { + "epoch": 2.17, + "learning_rate": 3.194669131808027e-05, + "loss": 0.0, + "step": 9374 + }, + { + "epoch": 2.17, + "learning_rate": 3.194476542639242e-05, + "loss": 0.0, + "step": 9375 + }, + { + "epoch": 2.17, + "learning_rate": 3.194283953470457e-05, + "loss": 0.0, + "step": 9376 + }, + { + "epoch": 2.17, + "learning_rate": 3.194091364301672e-05, + "loss": 0.0, + "step": 9377 + }, + { + "epoch": 2.17, + "learning_rate": 3.193898775132887e-05, + "loss": 0.0, + "step": 9378 + }, + { + "epoch": 2.17, + "learning_rate": 3.1937061859641015e-05, + "loss": 0.0, + "step": 9379 + }, + { + "epoch": 2.17, + "learning_rate": 3.193513596795316e-05, + "loss": 0.0, + "step": 9380 + }, + { + "epoch": 2.17, + "learning_rate": 3.193321007626531e-05, + "loss": 0.0, + "step": 9381 + }, + { + "epoch": 2.17, + "learning_rate": 3.193128418457746e-05, + "loss": 0.0001, + "step": 9382 + }, + { + "epoch": 2.17, + "learning_rate": 3.192935829288961e-05, + "loss": 0.0003, + "step": 9383 + }, + { + "epoch": 2.17, + "learning_rate": 3.192743240120176e-05, + "loss": 0.0, + "step": 9384 + }, + { + "epoch": 2.17, + "learning_rate": 3.192550650951391e-05, + "loss": 0.0021, + "step": 9385 + }, + { + "epoch": 2.17, + "learning_rate": 3.1923580617826054e-05, + "loss": 0.0006, + "step": 9386 + }, + { + "epoch": 2.17, + "learning_rate": 3.192165472613821e-05, + "loss": 0.0, + "step": 9387 + }, + { + "epoch": 2.17, + "learning_rate": 3.191972883445035e-05, + "loss": 0.0, + "step": 9388 + }, + { + "epoch": 2.17, + "learning_rate": 3.19178029427625e-05, + "loss": 0.0004, + "step": 9389 + }, + { + "epoch": 2.17, + "learning_rate": 3.191587705107465e-05, + "loss": 0.0, + "step": 9390 + }, + { + "epoch": 2.17, + "learning_rate": 3.19139511593868e-05, + "loss": 0.0003, + "step": 9391 + }, + { + "epoch": 2.17, + "learning_rate": 3.1912025267698946e-05, + "loss": 0.0, + "step": 9392 + }, + { + "epoch": 2.17, + "learning_rate": 3.1910099376011094e-05, + "loss": 0.0001, + "step": 9393 + }, + { + "epoch": 2.17, + "learning_rate": 3.190817348432325e-05, + "loss": 0.0, + "step": 9394 + }, + { + "epoch": 2.17, + "learning_rate": 3.190624759263539e-05, + "loss": 0.0001, + "step": 9395 + }, + { + "epoch": 2.17, + "learning_rate": 3.190432170094754e-05, + "loss": 0.0, + "step": 9396 + }, + { + "epoch": 2.17, + "learning_rate": 3.190239580925969e-05, + "loss": 0.0001, + "step": 9397 + }, + { + "epoch": 2.17, + "learning_rate": 3.190046991757184e-05, + "loss": 0.0, + "step": 9398 + }, + { + "epoch": 2.17, + "learning_rate": 3.1898544025883985e-05, + "loss": 0.0, + "step": 9399 + }, + { + "epoch": 2.17, + "learning_rate": 3.189661813419613e-05, + "loss": 0.0, + "step": 9400 + }, + { + "epoch": 2.17, + "learning_rate": 3.189469224250829e-05, + "loss": 0.0, + "step": 9401 + }, + { + "epoch": 2.17, + "learning_rate": 3.189276635082043e-05, + "loss": 0.0, + "step": 9402 + }, + { + "epoch": 2.17, + "learning_rate": 3.189084045913258e-05, + "loss": 0.0, + "step": 9403 + }, + { + "epoch": 2.17, + "learning_rate": 3.188891456744473e-05, + "loss": 0.0, + "step": 9404 + }, + { + "epoch": 2.17, + "learning_rate": 3.188698867575688e-05, + "loss": 0.0, + "step": 9405 + }, + { + "epoch": 2.17, + "learning_rate": 3.1885062784069024e-05, + "loss": 0.0, + "step": 9406 + }, + { + "epoch": 2.17, + "learning_rate": 3.188313689238117e-05, + "loss": 0.0, + "step": 9407 + }, + { + "epoch": 2.17, + "learning_rate": 3.1881211000693326e-05, + "loss": 0.0, + "step": 9408 + }, + { + "epoch": 2.17, + "learning_rate": 3.187928510900547e-05, + "loss": 0.0, + "step": 9409 + }, + { + "epoch": 2.17, + "learning_rate": 3.187735921731762e-05, + "loss": 0.0, + "step": 9410 + }, + { + "epoch": 2.17, + "learning_rate": 3.187543332562977e-05, + "loss": 0.0, + "step": 9411 + }, + { + "epoch": 2.18, + "learning_rate": 3.1873507433941916e-05, + "loss": 0.0, + "step": 9412 + }, + { + "epoch": 2.18, + "learning_rate": 3.187158154225406e-05, + "loss": 0.0001, + "step": 9413 + }, + { + "epoch": 2.18, + "learning_rate": 3.186965565056622e-05, + "loss": 0.0, + "step": 9414 + }, + { + "epoch": 2.18, + "learning_rate": 3.1867729758878365e-05, + "loss": 0.0, + "step": 9415 + }, + { + "epoch": 2.18, + "learning_rate": 3.186580386719051e-05, + "loss": 0.0, + "step": 9416 + }, + { + "epoch": 2.18, + "learning_rate": 3.186387797550266e-05, + "loss": 0.0, + "step": 9417 + }, + { + "epoch": 2.18, + "learning_rate": 3.186195208381481e-05, + "loss": 0.0001, + "step": 9418 + }, + { + "epoch": 2.18, + "learning_rate": 3.1860026192126955e-05, + "loss": 0.0, + "step": 9419 + }, + { + "epoch": 2.18, + "learning_rate": 3.18581003004391e-05, + "loss": 0.0004, + "step": 9420 + }, + { + "epoch": 2.18, + "learning_rate": 3.185617440875126e-05, + "loss": 0.0001, + "step": 9421 + }, + { + "epoch": 2.18, + "learning_rate": 3.1854248517063404e-05, + "loss": 0.0, + "step": 9422 + }, + { + "epoch": 2.18, + "learning_rate": 3.185232262537555e-05, + "loss": 0.0003, + "step": 9423 + }, + { + "epoch": 2.18, + "learning_rate": 3.18503967336877e-05, + "loss": 0.0, + "step": 9424 + }, + { + "epoch": 2.18, + "learning_rate": 3.1848470841999853e-05, + "loss": 0.0006, + "step": 9425 + }, + { + "epoch": 2.18, + "learning_rate": 3.1846544950311994e-05, + "loss": 0.0003, + "step": 9426 + }, + { + "epoch": 2.18, + "learning_rate": 3.184461905862414e-05, + "loss": 0.0001, + "step": 9427 + }, + { + "epoch": 2.18, + "learning_rate": 3.1842693166936296e-05, + "loss": 0.0001, + "step": 9428 + }, + { + "epoch": 2.18, + "learning_rate": 3.1840767275248437e-05, + "loss": 0.0, + "step": 9429 + }, + { + "epoch": 2.18, + "learning_rate": 3.183884138356059e-05, + "loss": 0.0, + "step": 9430 + }, + { + "epoch": 2.18, + "learning_rate": 3.183691549187274e-05, + "loss": 0.0, + "step": 9431 + }, + { + "epoch": 2.18, + "learning_rate": 3.1834989600184886e-05, + "loss": 0.0, + "step": 9432 + }, + { + "epoch": 2.18, + "learning_rate": 3.183306370849703e-05, + "loss": 0.0, + "step": 9433 + }, + { + "epoch": 2.18, + "learning_rate": 3.183113781680919e-05, + "loss": 0.0004, + "step": 9434 + }, + { + "epoch": 2.18, + "learning_rate": 3.1829211925121335e-05, + "loss": 0.0, + "step": 9435 + }, + { + "epoch": 2.18, + "learning_rate": 3.1827286033433476e-05, + "loss": 0.0001, + "step": 9436 + }, + { + "epoch": 2.18, + "learning_rate": 3.182536014174563e-05, + "loss": 0.0, + "step": 9437 + }, + { + "epoch": 2.18, + "learning_rate": 3.182343425005778e-05, + "loss": 0.0, + "step": 9438 + }, + { + "epoch": 2.18, + "learning_rate": 3.1821508358369925e-05, + "loss": 0.0001, + "step": 9439 + }, + { + "epoch": 2.18, + "learning_rate": 3.181958246668207e-05, + "loss": 0.0, + "step": 9440 + }, + { + "epoch": 2.18, + "learning_rate": 3.181765657499423e-05, + "loss": 0.0, + "step": 9441 + }, + { + "epoch": 2.18, + "learning_rate": 3.1815730683306374e-05, + "loss": 0.0004, + "step": 9442 + }, + { + "epoch": 2.18, + "learning_rate": 3.181380479161852e-05, + "loss": 0.0, + "step": 9443 + }, + { + "epoch": 2.18, + "learning_rate": 3.181187889993067e-05, + "loss": 0.0, + "step": 9444 + }, + { + "epoch": 2.18, + "learning_rate": 3.1809953008242817e-05, + "loss": 0.0001, + "step": 9445 + }, + { + "epoch": 2.18, + "learning_rate": 3.1808027116554964e-05, + "loss": 0.0001, + "step": 9446 + }, + { + "epoch": 2.18, + "learning_rate": 3.180610122486711e-05, + "loss": 0.0, + "step": 9447 + }, + { + "epoch": 2.18, + "learning_rate": 3.1804175333179266e-05, + "loss": 0.0, + "step": 9448 + }, + { + "epoch": 2.18, + "learning_rate": 3.180224944149141e-05, + "loss": 0.0, + "step": 9449 + }, + { + "epoch": 2.18, + "learning_rate": 3.180032354980356e-05, + "loss": 0.0, + "step": 9450 + }, + { + "epoch": 2.18, + "learning_rate": 3.179839765811571e-05, + "loss": 0.0007, + "step": 9451 + }, + { + "epoch": 2.18, + "learning_rate": 3.179647176642786e-05, + "loss": 0.0, + "step": 9452 + }, + { + "epoch": 2.18, + "learning_rate": 3.179454587474e-05, + "loss": 0.0, + "step": 9453 + }, + { + "epoch": 2.18, + "learning_rate": 3.179261998305215e-05, + "loss": 0.0002, + "step": 9454 + }, + { + "epoch": 2.19, + "learning_rate": 3.1790694091364305e-05, + "loss": 0.0003, + "step": 9455 + }, + { + "epoch": 2.19, + "learning_rate": 3.178876819967645e-05, + "loss": 0.0003, + "step": 9456 + }, + { + "epoch": 2.19, + "learning_rate": 3.17868423079886e-05, + "loss": 0.0, + "step": 9457 + }, + { + "epoch": 2.19, + "learning_rate": 3.178491641630075e-05, + "loss": 0.0, + "step": 9458 + }, + { + "epoch": 2.19, + "learning_rate": 3.17829905246129e-05, + "loss": 0.0001, + "step": 9459 + }, + { + "epoch": 2.19, + "learning_rate": 3.178106463292504e-05, + "loss": 0.0, + "step": 9460 + }, + { + "epoch": 2.19, + "learning_rate": 3.1779138741237197e-05, + "loss": 0.0, + "step": 9461 + }, + { + "epoch": 2.19, + "learning_rate": 3.1777212849549344e-05, + "loss": 0.0, + "step": 9462 + }, + { + "epoch": 2.19, + "learning_rate": 3.177528695786149e-05, + "loss": 0.0, + "step": 9463 + }, + { + "epoch": 2.19, + "learning_rate": 3.177336106617364e-05, + "loss": 0.0, + "step": 9464 + }, + { + "epoch": 2.19, + "learning_rate": 3.1771435174485786e-05, + "loss": 0.0, + "step": 9465 + }, + { + "epoch": 2.19, + "learning_rate": 3.176950928279794e-05, + "loss": 0.0001, + "step": 9466 + }, + { + "epoch": 2.19, + "learning_rate": 3.176758339111008e-05, + "loss": 0.0, + "step": 9467 + }, + { + "epoch": 2.19, + "learning_rate": 3.1765657499422236e-05, + "loss": 0.0001, + "step": 9468 + }, + { + "epoch": 2.19, + "learning_rate": 3.176373160773438e-05, + "loss": 0.0, + "step": 9469 + }, + { + "epoch": 2.19, + "learning_rate": 3.176180571604653e-05, + "loss": 0.0, + "step": 9470 + }, + { + "epoch": 2.19, + "learning_rate": 3.175987982435868e-05, + "loss": 0.0, + "step": 9471 + }, + { + "epoch": 2.19, + "learning_rate": 3.175795393267083e-05, + "loss": 0.0, + "step": 9472 + }, + { + "epoch": 2.19, + "learning_rate": 3.175602804098298e-05, + "loss": 0.0, + "step": 9473 + }, + { + "epoch": 2.19, + "learning_rate": 3.175410214929512e-05, + "loss": 0.0, + "step": 9474 + }, + { + "epoch": 2.19, + "learning_rate": 3.1752176257607275e-05, + "loss": 0.0001, + "step": 9475 + }, + { + "epoch": 2.19, + "learning_rate": 3.175025036591942e-05, + "loss": 0.0, + "step": 9476 + }, + { + "epoch": 2.19, + "learning_rate": 3.174832447423157e-05, + "loss": 0.0008, + "step": 9477 + }, + { + "epoch": 2.19, + "learning_rate": 3.174639858254372e-05, + "loss": 0.0, + "step": 9478 + }, + { + "epoch": 2.19, + "learning_rate": 3.174447269085587e-05, + "loss": 0.0, + "step": 9479 + }, + { + "epoch": 2.19, + "learning_rate": 3.174254679916802e-05, + "loss": 0.0, + "step": 9480 + }, + { + "epoch": 2.19, + "learning_rate": 3.1740620907480166e-05, + "loss": 0.0, + "step": 9481 + }, + { + "epoch": 2.19, + "learning_rate": 3.1738695015792314e-05, + "loss": 0.0, + "step": 9482 + }, + { + "epoch": 2.19, + "learning_rate": 3.173676912410446e-05, + "loss": 0.0, + "step": 9483 + }, + { + "epoch": 2.19, + "learning_rate": 3.173484323241661e-05, + "loss": 0.0001, + "step": 9484 + }, + { + "epoch": 2.19, + "learning_rate": 3.1732917340728756e-05, + "loss": 0.0, + "step": 9485 + }, + { + "epoch": 2.19, + "learning_rate": 3.173099144904091e-05, + "loss": 0.0001, + "step": 9486 + }, + { + "epoch": 2.19, + "learning_rate": 3.172906555735306e-05, + "loss": 0.0, + "step": 9487 + }, + { + "epoch": 2.19, + "learning_rate": 3.1727139665665206e-05, + "loss": 0.0, + "step": 9488 + }, + { + "epoch": 2.19, + "learning_rate": 3.172521377397735e-05, + "loss": 0.0, + "step": 9489 + }, + { + "epoch": 2.19, + "learning_rate": 3.172328788228951e-05, + "loss": 0.0001, + "step": 9490 + }, + { + "epoch": 2.19, + "learning_rate": 3.172136199060165e-05, + "loss": 0.0, + "step": 9491 + }, + { + "epoch": 2.19, + "learning_rate": 3.1719436098913795e-05, + "loss": 0.0, + "step": 9492 + }, + { + "epoch": 2.19, + "learning_rate": 3.171751020722595e-05, + "loss": 0.0004, + "step": 9493 + }, + { + "epoch": 2.19, + "learning_rate": 3.17155843155381e-05, + "loss": 0.0, + "step": 9494 + }, + { + "epoch": 2.19, + "learning_rate": 3.1713658423850245e-05, + "loss": 0.0, + "step": 9495 + }, + { + "epoch": 2.19, + "learning_rate": 3.171173253216239e-05, + "loss": 0.0, + "step": 9496 + }, + { + "epoch": 2.19, + "learning_rate": 3.1709806640474546e-05, + "loss": 0.0008, + "step": 9497 + }, + { + "epoch": 2.2, + "learning_rate": 3.170788074878669e-05, + "loss": 0.0, + "step": 9498 + }, + { + "epoch": 2.2, + "learning_rate": 3.170595485709884e-05, + "loss": 0.0, + "step": 9499 + }, + { + "epoch": 2.2, + "learning_rate": 3.170402896541099e-05, + "loss": 0.0, + "step": 9500 + }, + { + "epoch": 2.2, + "learning_rate": 3.170210307372313e-05, + "loss": 0.0002, + "step": 9501 + }, + { + "epoch": 2.2, + "learning_rate": 3.1700177182035284e-05, + "loss": 0.0, + "step": 9502 + }, + { + "epoch": 2.2, + "learning_rate": 3.169825129034743e-05, + "loss": 0.0, + "step": 9503 + }, + { + "epoch": 2.2, + "learning_rate": 3.169632539865958e-05, + "loss": 0.0, + "step": 9504 + }, + { + "epoch": 2.2, + "learning_rate": 3.1694399506971726e-05, + "loss": 0.0, + "step": 9505 + }, + { + "epoch": 2.2, + "learning_rate": 3.169247361528388e-05, + "loss": 0.0, + "step": 9506 + }, + { + "epoch": 2.2, + "learning_rate": 3.169054772359603e-05, + "loss": 0.0011, + "step": 9507 + }, + { + "epoch": 2.2, + "learning_rate": 3.1688621831908175e-05, + "loss": 0.0, + "step": 9508 + }, + { + "epoch": 2.2, + "learning_rate": 3.168669594022032e-05, + "loss": 0.0, + "step": 9509 + }, + { + "epoch": 2.2, + "learning_rate": 3.168477004853247e-05, + "loss": 0.0, + "step": 9510 + }, + { + "epoch": 2.2, + "learning_rate": 3.168284415684462e-05, + "loss": 0.0001, + "step": 9511 + }, + { + "epoch": 2.2, + "learning_rate": 3.1680918265156765e-05, + "loss": 0.0, + "step": 9512 + }, + { + "epoch": 2.2, + "learning_rate": 3.167899237346892e-05, + "loss": 0.0, + "step": 9513 + }, + { + "epoch": 2.2, + "learning_rate": 3.167706648178107e-05, + "loss": 0.0, + "step": 9514 + }, + { + "epoch": 2.2, + "learning_rate": 3.1675140590093215e-05, + "loss": 0.0005, + "step": 9515 + }, + { + "epoch": 2.2, + "learning_rate": 3.167321469840536e-05, + "loss": 0.0, + "step": 9516 + }, + { + "epoch": 2.2, + "learning_rate": 3.1671288806717516e-05, + "loss": 0.0001, + "step": 9517 + }, + { + "epoch": 2.2, + "learning_rate": 3.166936291502966e-05, + "loss": 0.0, + "step": 9518 + }, + { + "epoch": 2.2, + "learning_rate": 3.1667437023341804e-05, + "loss": 0.0, + "step": 9519 + }, + { + "epoch": 2.2, + "learning_rate": 3.166551113165396e-05, + "loss": 0.0, + "step": 9520 + }, + { + "epoch": 2.2, + "learning_rate": 3.1663585239966106e-05, + "loss": 0.0, + "step": 9521 + }, + { + "epoch": 2.2, + "learning_rate": 3.1661659348278254e-05, + "loss": 0.0, + "step": 9522 + }, + { + "epoch": 2.2, + "learning_rate": 3.16597334565904e-05, + "loss": 0.0, + "step": 9523 + }, + { + "epoch": 2.2, + "learning_rate": 3.1657807564902555e-05, + "loss": 0.0, + "step": 9524 + }, + { + "epoch": 2.2, + "learning_rate": 3.1655881673214696e-05, + "loss": 0.0, + "step": 9525 + }, + { + "epoch": 2.2, + "learning_rate": 3.165395578152685e-05, + "loss": 0.0001, + "step": 9526 + }, + { + "epoch": 2.2, + "learning_rate": 3.1652029889839e-05, + "loss": 0.0, + "step": 9527 + }, + { + "epoch": 2.2, + "learning_rate": 3.1650103998151145e-05, + "loss": 0.0001, + "step": 9528 + }, + { + "epoch": 2.2, + "learning_rate": 3.164817810646329e-05, + "loss": 0.0, + "step": 9529 + }, + { + "epoch": 2.2, + "learning_rate": 3.164625221477544e-05, + "loss": 0.0, + "step": 9530 + }, + { + "epoch": 2.2, + "learning_rate": 3.1644326323087595e-05, + "loss": 0.0001, + "step": 9531 + }, + { + "epoch": 2.2, + "learning_rate": 3.1642400431399735e-05, + "loss": 0.0001, + "step": 9532 + }, + { + "epoch": 2.2, + "learning_rate": 3.164047453971189e-05, + "loss": 0.0, + "step": 9533 + }, + { + "epoch": 2.2, + "learning_rate": 3.163854864802404e-05, + "loss": 0.0, + "step": 9534 + }, + { + "epoch": 2.2, + "learning_rate": 3.1636622756336184e-05, + "loss": 0.0, + "step": 9535 + }, + { + "epoch": 2.2, + "learning_rate": 3.163469686464833e-05, + "loss": 0.0, + "step": 9536 + }, + { + "epoch": 2.2, + "learning_rate": 3.1632770972960486e-05, + "loss": 0.0, + "step": 9537 + }, + { + "epoch": 2.2, + "learning_rate": 3.1630845081272634e-05, + "loss": 0.0, + "step": 9538 + }, + { + "epoch": 2.2, + "learning_rate": 3.1628919189584774e-05, + "loss": 0.0008, + "step": 9539 + }, + { + "epoch": 2.2, + "learning_rate": 3.162699329789693e-05, + "loss": 0.0001, + "step": 9540 + }, + { + "epoch": 2.2, + "learning_rate": 3.1625067406209076e-05, + "loss": 0.0, + "step": 9541 + }, + { + "epoch": 2.21, + "learning_rate": 3.1623141514521224e-05, + "loss": 0.0, + "step": 9542 + }, + { + "epoch": 2.21, + "learning_rate": 3.162121562283337e-05, + "loss": 0.0, + "step": 9543 + }, + { + "epoch": 2.21, + "learning_rate": 3.1619289731145525e-05, + "loss": 0.0, + "step": 9544 + }, + { + "epoch": 2.21, + "learning_rate": 3.161736383945767e-05, + "loss": 0.0, + "step": 9545 + }, + { + "epoch": 2.21, + "learning_rate": 3.161543794776982e-05, + "loss": 0.0, + "step": 9546 + }, + { + "epoch": 2.21, + "learning_rate": 3.161351205608197e-05, + "loss": 0.0, + "step": 9547 + }, + { + "epoch": 2.21, + "learning_rate": 3.1611586164394115e-05, + "loss": 0.0, + "step": 9548 + }, + { + "epoch": 2.21, + "learning_rate": 3.160966027270626e-05, + "loss": 0.0001, + "step": 9549 + }, + { + "epoch": 2.21, + "learning_rate": 3.160773438101841e-05, + "loss": 0.0001, + "step": 9550 + }, + { + "epoch": 2.21, + "learning_rate": 3.1605808489330564e-05, + "loss": 0.0001, + "step": 9551 + }, + { + "epoch": 2.21, + "learning_rate": 3.160388259764271e-05, + "loss": 0.0001, + "step": 9552 + }, + { + "epoch": 2.21, + "learning_rate": 3.160195670595486e-05, + "loss": 0.0, + "step": 9553 + }, + { + "epoch": 2.21, + "learning_rate": 3.160003081426701e-05, + "loss": 0.0001, + "step": 9554 + }, + { + "epoch": 2.21, + "learning_rate": 3.159810492257916e-05, + "loss": 0.0, + "step": 9555 + }, + { + "epoch": 2.21, + "learning_rate": 3.15961790308913e-05, + "loss": 0.0, + "step": 9556 + }, + { + "epoch": 2.21, + "learning_rate": 3.159425313920345e-05, + "loss": 0.0, + "step": 9557 + }, + { + "epoch": 2.21, + "learning_rate": 3.1592327247515604e-05, + "loss": 0.0001, + "step": 9558 + }, + { + "epoch": 2.21, + "learning_rate": 3.159040135582775e-05, + "loss": 0.0001, + "step": 9559 + }, + { + "epoch": 2.21, + "learning_rate": 3.15884754641399e-05, + "loss": 0.0001, + "step": 9560 + }, + { + "epoch": 2.21, + "learning_rate": 3.1586549572452046e-05, + "loss": 0.0, + "step": 9561 + }, + { + "epoch": 2.21, + "learning_rate": 3.15846236807642e-05, + "loss": 0.0, + "step": 9562 + }, + { + "epoch": 2.21, + "learning_rate": 3.158269778907634e-05, + "loss": 0.0, + "step": 9563 + }, + { + "epoch": 2.21, + "learning_rate": 3.1580771897388495e-05, + "loss": 0.0001, + "step": 9564 + }, + { + "epoch": 2.21, + "learning_rate": 3.157884600570064e-05, + "loss": 0.0, + "step": 9565 + }, + { + "epoch": 2.21, + "learning_rate": 3.157692011401279e-05, + "loss": 0.0004, + "step": 9566 + }, + { + "epoch": 2.21, + "learning_rate": 3.157499422232494e-05, + "loss": 0.0, + "step": 9567 + }, + { + "epoch": 2.21, + "learning_rate": 3.1573068330637085e-05, + "loss": 0.0, + "step": 9568 + }, + { + "epoch": 2.21, + "learning_rate": 3.157114243894924e-05, + "loss": 0.0, + "step": 9569 + }, + { + "epoch": 2.21, + "learning_rate": 3.156921654726138e-05, + "loss": 0.0001, + "step": 9570 + }, + { + "epoch": 2.21, + "learning_rate": 3.1567290655573534e-05, + "loss": 0.0, + "step": 9571 + }, + { + "epoch": 2.21, + "learning_rate": 3.156536476388568e-05, + "loss": 0.0, + "step": 9572 + }, + { + "epoch": 2.21, + "learning_rate": 3.156343887219783e-05, + "loss": 0.0, + "step": 9573 + }, + { + "epoch": 2.21, + "learning_rate": 3.156151298050998e-05, + "loss": 0.0001, + "step": 9574 + }, + { + "epoch": 2.21, + "learning_rate": 3.1559587088822124e-05, + "loss": 0.0001, + "step": 9575 + }, + { + "epoch": 2.21, + "learning_rate": 3.155766119713427e-05, + "loss": 0.0, + "step": 9576 + }, + { + "epoch": 2.21, + "learning_rate": 3.155573530544642e-05, + "loss": 0.0001, + "step": 9577 + }, + { + "epoch": 2.21, + "learning_rate": 3.1553809413758573e-05, + "loss": 0.0002, + "step": 9578 + }, + { + "epoch": 2.21, + "learning_rate": 3.155188352207072e-05, + "loss": 0.0, + "step": 9579 + }, + { + "epoch": 2.21, + "learning_rate": 3.154995763038287e-05, + "loss": 0.0, + "step": 9580 + }, + { + "epoch": 2.21, + "learning_rate": 3.1548031738695016e-05, + "loss": 0.0, + "step": 9581 + }, + { + "epoch": 2.21, + "learning_rate": 3.154610584700717e-05, + "loss": 0.0, + "step": 9582 + }, + { + "epoch": 2.21, + "learning_rate": 3.154417995531931e-05, + "loss": 0.0001, + "step": 9583 + }, + { + "epoch": 2.21, + "learning_rate": 3.1542254063631465e-05, + "loss": 0.0, + "step": 9584 + }, + { + "epoch": 2.22, + "learning_rate": 3.154032817194361e-05, + "loss": 0.0002, + "step": 9585 + }, + { + "epoch": 2.22, + "learning_rate": 3.153840228025576e-05, + "loss": 0.0, + "step": 9586 + }, + { + "epoch": 2.22, + "learning_rate": 3.153647638856791e-05, + "loss": 0.0005, + "step": 9587 + }, + { + "epoch": 2.22, + "learning_rate": 3.1534550496880055e-05, + "loss": 0.0001, + "step": 9588 + }, + { + "epoch": 2.22, + "learning_rate": 3.153262460519221e-05, + "loss": 0.0, + "step": 9589 + }, + { + "epoch": 2.22, + "learning_rate": 3.153069871350435e-05, + "loss": 0.0014, + "step": 9590 + }, + { + "epoch": 2.22, + "learning_rate": 3.1528772821816504e-05, + "loss": 0.0, + "step": 9591 + }, + { + "epoch": 2.22, + "learning_rate": 3.152684693012865e-05, + "loss": 0.0, + "step": 9592 + }, + { + "epoch": 2.22, + "learning_rate": 3.15249210384408e-05, + "loss": 0.0001, + "step": 9593 + }, + { + "epoch": 2.22, + "learning_rate": 3.1522995146752947e-05, + "loss": 0.0001, + "step": 9594 + }, + { + "epoch": 2.22, + "learning_rate": 3.1521069255065094e-05, + "loss": 0.002, + "step": 9595 + }, + { + "epoch": 2.22, + "learning_rate": 3.151914336337725e-05, + "loss": 0.0001, + "step": 9596 + }, + { + "epoch": 2.22, + "learning_rate": 3.151721747168939e-05, + "loss": 0.0, + "step": 9597 + }, + { + "epoch": 2.22, + "learning_rate": 3.151529158000154e-05, + "loss": 0.0008, + "step": 9598 + }, + { + "epoch": 2.22, + "learning_rate": 3.151336568831369e-05, + "loss": 0.0, + "step": 9599 + }, + { + "epoch": 2.22, + "learning_rate": 3.151143979662584e-05, + "loss": 0.0, + "step": 9600 + }, + { + "epoch": 2.22, + "learning_rate": 3.1509513904937986e-05, + "loss": 0.0, + "step": 9601 + }, + { + "epoch": 2.22, + "learning_rate": 3.150758801325014e-05, + "loss": 0.0002, + "step": 9602 + }, + { + "epoch": 2.22, + "learning_rate": 3.150566212156229e-05, + "loss": 0.0, + "step": 9603 + }, + { + "epoch": 2.22, + "learning_rate": 3.150373622987443e-05, + "loss": 0.0, + "step": 9604 + }, + { + "epoch": 2.22, + "learning_rate": 3.150181033818658e-05, + "loss": 0.0, + "step": 9605 + }, + { + "epoch": 2.22, + "learning_rate": 3.149988444649873e-05, + "loss": 0.0, + "step": 9606 + }, + { + "epoch": 2.22, + "learning_rate": 3.149795855481088e-05, + "loss": 0.0, + "step": 9607 + }, + { + "epoch": 2.22, + "learning_rate": 3.1496032663123025e-05, + "loss": 0.0, + "step": 9608 + }, + { + "epoch": 2.22, + "learning_rate": 3.149410677143518e-05, + "loss": 0.0, + "step": 9609 + }, + { + "epoch": 2.22, + "learning_rate": 3.1492180879747327e-05, + "loss": 0.0, + "step": 9610 + }, + { + "epoch": 2.22, + "learning_rate": 3.1490254988059474e-05, + "loss": 0.0, + "step": 9611 + }, + { + "epoch": 2.22, + "learning_rate": 3.148832909637162e-05, + "loss": 0.0066, + "step": 9612 + }, + { + "epoch": 2.22, + "learning_rate": 3.148640320468377e-05, + "loss": 0.0, + "step": 9613 + }, + { + "epoch": 2.22, + "learning_rate": 3.1484477312995916e-05, + "loss": 0.0, + "step": 9614 + }, + { + "epoch": 2.22, + "learning_rate": 3.1482551421308064e-05, + "loss": 0.0, + "step": 9615 + }, + { + "epoch": 2.22, + "learning_rate": 3.148062552962022e-05, + "loss": 0.0, + "step": 9616 + }, + { + "epoch": 2.22, + "learning_rate": 3.1478699637932366e-05, + "loss": 0.0, + "step": 9617 + }, + { + "epoch": 2.22, + "learning_rate": 3.147677374624451e-05, + "loss": 0.0, + "step": 9618 + }, + { + "epoch": 2.22, + "learning_rate": 3.147484785455666e-05, + "loss": 0.0003, + "step": 9619 + }, + { + "epoch": 2.22, + "learning_rate": 3.1472921962868815e-05, + "loss": 0.0, + "step": 9620 + }, + { + "epoch": 2.22, + "learning_rate": 3.1470996071180956e-05, + "loss": 0.0001, + "step": 9621 + }, + { + "epoch": 2.22, + "learning_rate": 3.14690701794931e-05, + "loss": 0.0001, + "step": 9622 + }, + { + "epoch": 2.22, + "learning_rate": 3.146714428780526e-05, + "loss": 0.0, + "step": 9623 + }, + { + "epoch": 2.22, + "learning_rate": 3.1465218396117405e-05, + "loss": 0.0001, + "step": 9624 + }, + { + "epoch": 2.22, + "learning_rate": 3.146329250442955e-05, + "loss": 0.0, + "step": 9625 + }, + { + "epoch": 2.22, + "learning_rate": 3.14613666127417e-05, + "loss": 0.0001, + "step": 9626 + }, + { + "epoch": 2.22, + "learning_rate": 3.1459440721053854e-05, + "loss": 0.0, + "step": 9627 + }, + { + "epoch": 2.23, + "learning_rate": 3.1457514829365995e-05, + "loss": 0.0, + "step": 9628 + }, + { + "epoch": 2.23, + "learning_rate": 3.145558893767815e-05, + "loss": 0.0001, + "step": 9629 + }, + { + "epoch": 2.23, + "learning_rate": 3.1453663045990296e-05, + "loss": 0.0001, + "step": 9630 + }, + { + "epoch": 2.23, + "learning_rate": 3.1451737154302444e-05, + "loss": 0.0, + "step": 9631 + }, + { + "epoch": 2.23, + "learning_rate": 3.144981126261459e-05, + "loss": 0.1132, + "step": 9632 + }, + { + "epoch": 2.23, + "learning_rate": 3.144788537092674e-05, + "loss": 0.0, + "step": 9633 + }, + { + "epoch": 2.23, + "learning_rate": 3.144595947923889e-05, + "loss": 0.0, + "step": 9634 + }, + { + "epoch": 2.23, + "learning_rate": 3.1444033587551034e-05, + "loss": 0.0, + "step": 9635 + }, + { + "epoch": 2.23, + "learning_rate": 3.144210769586319e-05, + "loss": 0.0, + "step": 9636 + }, + { + "epoch": 2.23, + "learning_rate": 3.1440181804175336e-05, + "loss": 0.0, + "step": 9637 + }, + { + "epoch": 2.23, + "learning_rate": 3.143825591248748e-05, + "loss": 0.0002, + "step": 9638 + }, + { + "epoch": 2.23, + "learning_rate": 3.143633002079963e-05, + "loss": 0.0, + "step": 9639 + }, + { + "epoch": 2.23, + "learning_rate": 3.1434404129111785e-05, + "loss": 0.0, + "step": 9640 + }, + { + "epoch": 2.23, + "learning_rate": 3.143247823742393e-05, + "loss": 0.0, + "step": 9641 + }, + { + "epoch": 2.23, + "learning_rate": 3.143055234573607e-05, + "loss": 0.0001, + "step": 9642 + }, + { + "epoch": 2.23, + "learning_rate": 3.142862645404823e-05, + "loss": 0.0, + "step": 9643 + }, + { + "epoch": 2.23, + "learning_rate": 3.1426700562360375e-05, + "loss": 0.0, + "step": 9644 + }, + { + "epoch": 2.23, + "learning_rate": 3.142477467067252e-05, + "loss": 0.0, + "step": 9645 + }, + { + "epoch": 2.23, + "learning_rate": 3.142284877898467e-05, + "loss": 0.0001, + "step": 9646 + }, + { + "epoch": 2.23, + "learning_rate": 3.1420922887296824e-05, + "loss": 0.0, + "step": 9647 + }, + { + "epoch": 2.23, + "learning_rate": 3.141899699560897e-05, + "loss": 0.0, + "step": 9648 + }, + { + "epoch": 2.23, + "learning_rate": 3.141707110392112e-05, + "loss": 0.0002, + "step": 9649 + }, + { + "epoch": 2.23, + "learning_rate": 3.1415145212233266e-05, + "loss": 0.0005, + "step": 9650 + }, + { + "epoch": 2.23, + "learning_rate": 3.1413219320545414e-05, + "loss": 0.0, + "step": 9651 + }, + { + "epoch": 2.23, + "learning_rate": 3.141129342885756e-05, + "loss": 0.0001, + "step": 9652 + }, + { + "epoch": 2.23, + "learning_rate": 3.140936753716971e-05, + "loss": 0.0, + "step": 9653 + }, + { + "epoch": 2.23, + "learning_rate": 3.140744164548186e-05, + "loss": 0.0, + "step": 9654 + }, + { + "epoch": 2.23, + "learning_rate": 3.1405515753794004e-05, + "loss": 0.0008, + "step": 9655 + }, + { + "epoch": 2.23, + "learning_rate": 3.140358986210616e-05, + "loss": 0.0003, + "step": 9656 + }, + { + "epoch": 2.23, + "learning_rate": 3.1401663970418305e-05, + "loss": 0.0, + "step": 9657 + }, + { + "epoch": 2.23, + "learning_rate": 3.139973807873045e-05, + "loss": 0.0005, + "step": 9658 + }, + { + "epoch": 2.23, + "learning_rate": 3.13978121870426e-05, + "loss": 0.0, + "step": 9659 + }, + { + "epoch": 2.23, + "learning_rate": 3.139588629535475e-05, + "loss": 0.0, + "step": 9660 + }, + { + "epoch": 2.23, + "learning_rate": 3.13939604036669e-05, + "loss": 0.0, + "step": 9661 + }, + { + "epoch": 2.23, + "learning_rate": 3.139203451197904e-05, + "loss": 0.0, + "step": 9662 + }, + { + "epoch": 2.23, + "learning_rate": 3.13901086202912e-05, + "loss": 0.0, + "step": 9663 + }, + { + "epoch": 2.23, + "learning_rate": 3.1388182728603345e-05, + "loss": 0.0, + "step": 9664 + }, + { + "epoch": 2.23, + "learning_rate": 3.138625683691549e-05, + "loss": 0.0001, + "step": 9665 + }, + { + "epoch": 2.23, + "learning_rate": 3.138433094522764e-05, + "loss": 0.0, + "step": 9666 + }, + { + "epoch": 2.23, + "learning_rate": 3.1382405053539794e-05, + "loss": 0.0, + "step": 9667 + }, + { + "epoch": 2.23, + "learning_rate": 3.138047916185194e-05, + "loss": 0.0001, + "step": 9668 + }, + { + "epoch": 2.23, + "learning_rate": 3.137855327016408e-05, + "loss": 0.0, + "step": 9669 + }, + { + "epoch": 2.23, + "learning_rate": 3.1376627378476236e-05, + "loss": 0.0001, + "step": 9670 + }, + { + "epoch": 2.24, + "learning_rate": 3.1374701486788384e-05, + "loss": 0.0, + "step": 9671 + }, + { + "epoch": 2.24, + "learning_rate": 3.137277559510053e-05, + "loss": 0.0, + "step": 9672 + }, + { + "epoch": 2.24, + "learning_rate": 3.137084970341268e-05, + "loss": 0.0, + "step": 9673 + }, + { + "epoch": 2.24, + "learning_rate": 3.136892381172483e-05, + "loss": 0.0001, + "step": 9674 + }, + { + "epoch": 2.24, + "learning_rate": 3.136699792003698e-05, + "loss": 0.0, + "step": 9675 + }, + { + "epoch": 2.24, + "learning_rate": 3.136507202834913e-05, + "loss": 0.0, + "step": 9676 + }, + { + "epoch": 2.24, + "learning_rate": 3.1363146136661275e-05, + "loss": 0.0, + "step": 9677 + }, + { + "epoch": 2.24, + "learning_rate": 3.136122024497342e-05, + "loss": 0.0001, + "step": 9678 + }, + { + "epoch": 2.24, + "learning_rate": 3.135929435328557e-05, + "loss": 0.0, + "step": 9679 + }, + { + "epoch": 2.24, + "learning_rate": 3.135736846159772e-05, + "loss": 0.0, + "step": 9680 + }, + { + "epoch": 2.24, + "learning_rate": 3.135544256990987e-05, + "loss": 0.0, + "step": 9681 + }, + { + "epoch": 2.24, + "learning_rate": 3.135351667822202e-05, + "loss": 0.0, + "step": 9682 + }, + { + "epoch": 2.24, + "learning_rate": 3.135159078653417e-05, + "loss": 0.0, + "step": 9683 + }, + { + "epoch": 2.24, + "learning_rate": 3.1349664894846314e-05, + "loss": 0.0017, + "step": 9684 + }, + { + "epoch": 2.24, + "learning_rate": 3.134773900315847e-05, + "loss": 0.0001, + "step": 9685 + }, + { + "epoch": 2.24, + "learning_rate": 3.134581311147061e-05, + "loss": 0.0, + "step": 9686 + }, + { + "epoch": 2.24, + "learning_rate": 3.134388721978276e-05, + "loss": 0.0, + "step": 9687 + }, + { + "epoch": 2.24, + "learning_rate": 3.134196132809491e-05, + "loss": 0.0, + "step": 9688 + }, + { + "epoch": 2.24, + "learning_rate": 3.134003543640706e-05, + "loss": 0.0, + "step": 9689 + }, + { + "epoch": 2.24, + "learning_rate": 3.1338109544719206e-05, + "loss": 0.0, + "step": 9690 + }, + { + "epoch": 2.24, + "learning_rate": 3.1336183653031354e-05, + "loss": 0.0, + "step": 9691 + }, + { + "epoch": 2.24, + "learning_rate": 3.133425776134351e-05, + "loss": 0.0, + "step": 9692 + }, + { + "epoch": 2.24, + "learning_rate": 3.133233186965565e-05, + "loss": 0.0, + "step": 9693 + }, + { + "epoch": 2.24, + "learning_rate": 3.13304059779678e-05, + "loss": 0.0, + "step": 9694 + }, + { + "epoch": 2.24, + "learning_rate": 3.132848008627995e-05, + "loss": 0.0, + "step": 9695 + }, + { + "epoch": 2.24, + "learning_rate": 3.13265541945921e-05, + "loss": 0.0, + "step": 9696 + }, + { + "epoch": 2.24, + "learning_rate": 3.1324628302904245e-05, + "loss": 0.0, + "step": 9697 + }, + { + "epoch": 2.24, + "learning_rate": 3.132270241121639e-05, + "loss": 0.0011, + "step": 9698 + }, + { + "epoch": 2.24, + "learning_rate": 3.132077651952855e-05, + "loss": 0.0002, + "step": 9699 + }, + { + "epoch": 2.24, + "learning_rate": 3.131885062784069e-05, + "loss": 0.0, + "step": 9700 + }, + { + "epoch": 2.24, + "learning_rate": 3.131692473615284e-05, + "loss": 0.0, + "step": 9701 + }, + { + "epoch": 2.24, + "learning_rate": 3.131499884446499e-05, + "loss": 0.0, + "step": 9702 + }, + { + "epoch": 2.24, + "learning_rate": 3.131307295277714e-05, + "loss": 0.0, + "step": 9703 + }, + { + "epoch": 2.24, + "learning_rate": 3.1311147061089284e-05, + "loss": 0.0, + "step": 9704 + }, + { + "epoch": 2.24, + "learning_rate": 3.130922116940144e-05, + "loss": 0.0, + "step": 9705 + }, + { + "epoch": 2.24, + "learning_rate": 3.1307295277713586e-05, + "loss": 0.0, + "step": 9706 + }, + { + "epoch": 2.24, + "learning_rate": 3.130536938602573e-05, + "loss": 0.0001, + "step": 9707 + }, + { + "epoch": 2.24, + "learning_rate": 3.130344349433788e-05, + "loss": 0.0, + "step": 9708 + }, + { + "epoch": 2.24, + "learning_rate": 3.130151760265003e-05, + "loss": 0.0, + "step": 9709 + }, + { + "epoch": 2.24, + "learning_rate": 3.1299591710962176e-05, + "loss": 0.0, + "step": 9710 + }, + { + "epoch": 2.24, + "learning_rate": 3.1297665819274323e-05, + "loss": 0.0, + "step": 9711 + }, + { + "epoch": 2.24, + "learning_rate": 3.129573992758648e-05, + "loss": 0.0, + "step": 9712 + }, + { + "epoch": 2.24, + "learning_rate": 3.1293814035898625e-05, + "loss": 0.0, + "step": 9713 + }, + { + "epoch": 2.24, + "learning_rate": 3.129188814421077e-05, + "loss": 0.0001, + "step": 9714 + }, + { + "epoch": 2.25, + "learning_rate": 3.128996225252292e-05, + "loss": 0.0001, + "step": 9715 + }, + { + "epoch": 2.25, + "learning_rate": 3.128803636083507e-05, + "loss": 0.0, + "step": 9716 + }, + { + "epoch": 2.25, + "learning_rate": 3.1286110469147215e-05, + "loss": 0.0, + "step": 9717 + }, + { + "epoch": 2.25, + "learning_rate": 3.128418457745936e-05, + "loss": 0.0, + "step": 9718 + }, + { + "epoch": 2.25, + "learning_rate": 3.128225868577152e-05, + "loss": 0.0, + "step": 9719 + }, + { + "epoch": 2.25, + "learning_rate": 3.1280332794083664e-05, + "loss": 0.0001, + "step": 9720 + }, + { + "epoch": 2.25, + "learning_rate": 3.127840690239581e-05, + "loss": 0.0, + "step": 9721 + }, + { + "epoch": 2.25, + "learning_rate": 3.127648101070796e-05, + "loss": 0.0, + "step": 9722 + }, + { + "epoch": 2.25, + "learning_rate": 3.1274555119020114e-05, + "loss": 0.0001, + "step": 9723 + }, + { + "epoch": 2.25, + "learning_rate": 3.1272629227332254e-05, + "loss": 0.0, + "step": 9724 + }, + { + "epoch": 2.25, + "learning_rate": 3.12707033356444e-05, + "loss": 0.0, + "step": 9725 + }, + { + "epoch": 2.25, + "learning_rate": 3.1268777443956556e-05, + "loss": 0.0, + "step": 9726 + }, + { + "epoch": 2.25, + "learning_rate": 3.12668515522687e-05, + "loss": 0.0006, + "step": 9727 + }, + { + "epoch": 2.25, + "learning_rate": 3.126492566058085e-05, + "loss": 0.0002, + "step": 9728 + }, + { + "epoch": 2.25, + "learning_rate": 3.1262999768893e-05, + "loss": 0.0218, + "step": 9729 + }, + { + "epoch": 2.25, + "learning_rate": 3.1261073877205146e-05, + "loss": 0.0, + "step": 9730 + }, + { + "epoch": 2.25, + "learning_rate": 3.125914798551729e-05, + "loss": 0.0, + "step": 9731 + }, + { + "epoch": 2.25, + "learning_rate": 3.125722209382945e-05, + "loss": 0.0, + "step": 9732 + }, + { + "epoch": 2.25, + "learning_rate": 3.1255296202141595e-05, + "loss": 0.0001, + "step": 9733 + }, + { + "epoch": 2.25, + "learning_rate": 3.1253370310453736e-05, + "loss": 0.0002, + "step": 9734 + }, + { + "epoch": 2.25, + "learning_rate": 3.125144441876589e-05, + "loss": 0.0, + "step": 9735 + }, + { + "epoch": 2.25, + "learning_rate": 3.124951852707804e-05, + "loss": 0.001, + "step": 9736 + }, + { + "epoch": 2.25, + "learning_rate": 3.1247592635390185e-05, + "loss": 0.0341, + "step": 9737 + }, + { + "epoch": 2.25, + "learning_rate": 3.124566674370233e-05, + "loss": 0.0, + "step": 9738 + }, + { + "epoch": 2.25, + "learning_rate": 3.124374085201449e-05, + "loss": 0.0, + "step": 9739 + }, + { + "epoch": 2.25, + "learning_rate": 3.1241814960326634e-05, + "loss": 0.0, + "step": 9740 + }, + { + "epoch": 2.25, + "learning_rate": 3.123988906863878e-05, + "loss": 0.0, + "step": 9741 + }, + { + "epoch": 2.25, + "learning_rate": 3.123796317695093e-05, + "loss": 0.0045, + "step": 9742 + }, + { + "epoch": 2.25, + "learning_rate": 3.123603728526308e-05, + "loss": 0.0, + "step": 9743 + }, + { + "epoch": 2.25, + "learning_rate": 3.1234111393575224e-05, + "loss": 0.0, + "step": 9744 + }, + { + "epoch": 2.25, + "learning_rate": 3.123218550188737e-05, + "loss": 0.0, + "step": 9745 + }, + { + "epoch": 2.25, + "learning_rate": 3.1230259610199526e-05, + "loss": 0.0002, + "step": 9746 + }, + { + "epoch": 2.25, + "learning_rate": 3.122833371851167e-05, + "loss": 0.0008, + "step": 9747 + }, + { + "epoch": 2.25, + "learning_rate": 3.122640782682382e-05, + "loss": 0.0, + "step": 9748 + }, + { + "epoch": 2.25, + "learning_rate": 3.122448193513597e-05, + "loss": 0.0, + "step": 9749 + }, + { + "epoch": 2.25, + "learning_rate": 3.122255604344812e-05, + "loss": 0.0, + "step": 9750 + }, + { + "epoch": 2.25, + "learning_rate": 3.122063015176026e-05, + "loss": 0.0001, + "step": 9751 + }, + { + "epoch": 2.25, + "learning_rate": 3.121870426007242e-05, + "loss": 0.0, + "step": 9752 + }, + { + "epoch": 2.25, + "learning_rate": 3.1216778368384565e-05, + "loss": 0.0, + "step": 9753 + }, + { + "epoch": 2.25, + "learning_rate": 3.121485247669671e-05, + "loss": 0.0, + "step": 9754 + }, + { + "epoch": 2.25, + "learning_rate": 3.121292658500886e-05, + "loss": 0.0001, + "step": 9755 + }, + { + "epoch": 2.25, + "learning_rate": 3.121100069332101e-05, + "loss": 0.0009, + "step": 9756 + }, + { + "epoch": 2.25, + "learning_rate": 3.120907480163316e-05, + "loss": 0.0007, + "step": 9757 + }, + { + "epoch": 2.26, + "learning_rate": 3.12071489099453e-05, + "loss": 0.0, + "step": 9758 + }, + { + "epoch": 2.26, + "learning_rate": 3.120522301825746e-05, + "loss": 0.0002, + "step": 9759 + }, + { + "epoch": 2.26, + "learning_rate": 3.1203297126569604e-05, + "loss": 0.0, + "step": 9760 + }, + { + "epoch": 2.26, + "learning_rate": 3.120137123488175e-05, + "loss": 0.0, + "step": 9761 + }, + { + "epoch": 2.26, + "learning_rate": 3.11994453431939e-05, + "loss": 0.0, + "step": 9762 + }, + { + "epoch": 2.26, + "learning_rate": 3.1197519451506047e-05, + "loss": 0.0006, + "step": 9763 + }, + { + "epoch": 2.26, + "learning_rate": 3.11955935598182e-05, + "loss": 0.0002, + "step": 9764 + }, + { + "epoch": 2.26, + "learning_rate": 3.119366766813034e-05, + "loss": 0.0, + "step": 9765 + }, + { + "epoch": 2.26, + "learning_rate": 3.1191741776442496e-05, + "loss": 0.0, + "step": 9766 + }, + { + "epoch": 2.26, + "learning_rate": 3.118981588475464e-05, + "loss": 0.0, + "step": 9767 + }, + { + "epoch": 2.26, + "learning_rate": 3.118788999306679e-05, + "loss": 0.0, + "step": 9768 + }, + { + "epoch": 2.26, + "learning_rate": 3.118596410137894e-05, + "loss": 0.0, + "step": 9769 + }, + { + "epoch": 2.26, + "learning_rate": 3.118403820969109e-05, + "loss": 0.0, + "step": 9770 + }, + { + "epoch": 2.26, + "learning_rate": 3.118211231800324e-05, + "loss": 0.0, + "step": 9771 + }, + { + "epoch": 2.26, + "learning_rate": 3.118018642631538e-05, + "loss": 0.0001, + "step": 9772 + }, + { + "epoch": 2.26, + "learning_rate": 3.1178260534627535e-05, + "loss": 0.0, + "step": 9773 + }, + { + "epoch": 2.26, + "learning_rate": 3.117633464293968e-05, + "loss": 0.0, + "step": 9774 + }, + { + "epoch": 2.26, + "learning_rate": 3.117440875125183e-05, + "loss": 0.0, + "step": 9775 + }, + { + "epoch": 2.26, + "learning_rate": 3.117248285956398e-05, + "loss": 0.0, + "step": 9776 + }, + { + "epoch": 2.26, + "learning_rate": 3.117055696787613e-05, + "loss": 0.0002, + "step": 9777 + }, + { + "epoch": 2.26, + "learning_rate": 3.116863107618828e-05, + "loss": 0.0, + "step": 9778 + }, + { + "epoch": 2.26, + "learning_rate": 3.1166705184500426e-05, + "loss": 0.0, + "step": 9779 + }, + { + "epoch": 2.26, + "learning_rate": 3.1164779292812574e-05, + "loss": 0.0, + "step": 9780 + }, + { + "epoch": 2.26, + "learning_rate": 3.116285340112472e-05, + "loss": 0.0, + "step": 9781 + }, + { + "epoch": 2.26, + "learning_rate": 3.116092750943687e-05, + "loss": 0.0, + "step": 9782 + }, + { + "epoch": 2.26, + "learning_rate": 3.1159001617749016e-05, + "loss": 0.0, + "step": 9783 + }, + { + "epoch": 2.26, + "learning_rate": 3.115707572606117e-05, + "loss": 0.0001, + "step": 9784 + }, + { + "epoch": 2.26, + "learning_rate": 3.115514983437332e-05, + "loss": 0.0004, + "step": 9785 + }, + { + "epoch": 2.26, + "learning_rate": 3.1153223942685466e-05, + "loss": 0.0, + "step": 9786 + }, + { + "epoch": 2.26, + "learning_rate": 3.115129805099761e-05, + "loss": 0.0, + "step": 9787 + }, + { + "epoch": 2.26, + "learning_rate": 3.114937215930977e-05, + "loss": 0.0, + "step": 9788 + }, + { + "epoch": 2.26, + "learning_rate": 3.114744626762191e-05, + "loss": 0.0001, + "step": 9789 + }, + { + "epoch": 2.26, + "learning_rate": 3.1145520375934056e-05, + "loss": 0.001, + "step": 9790 + }, + { + "epoch": 2.26, + "learning_rate": 3.114359448424621e-05, + "loss": 0.0, + "step": 9791 + }, + { + "epoch": 2.26, + "learning_rate": 3.114166859255836e-05, + "loss": 0.0, + "step": 9792 + }, + { + "epoch": 2.26, + "learning_rate": 3.1139742700870505e-05, + "loss": 0.0, + "step": 9793 + }, + { + "epoch": 2.26, + "learning_rate": 3.113781680918265e-05, + "loss": 0.0, + "step": 9794 + }, + { + "epoch": 2.26, + "learning_rate": 3.1135890917494806e-05, + "loss": 0.0001, + "step": 9795 + }, + { + "epoch": 2.26, + "learning_rate": 3.113396502580695e-05, + "loss": 0.0001, + "step": 9796 + }, + { + "epoch": 2.26, + "learning_rate": 3.11320391341191e-05, + "loss": 0.0, + "step": 9797 + }, + { + "epoch": 2.26, + "learning_rate": 3.113011324243125e-05, + "loss": 0.0, + "step": 9798 + }, + { + "epoch": 2.26, + "learning_rate": 3.112818735074339e-05, + "loss": 0.0001, + "step": 9799 + }, + { + "epoch": 2.26, + "learning_rate": 3.1126261459055544e-05, + "loss": 0.0, + "step": 9800 + }, + { + "epoch": 2.27, + "learning_rate": 3.112433556736769e-05, + "loss": 0.0001, + "step": 9801 + }, + { + "epoch": 2.27, + "learning_rate": 3.1122409675679846e-05, + "loss": 0.0001, + "step": 9802 + }, + { + "epoch": 2.27, + "learning_rate": 3.1120483783991986e-05, + "loss": 0.0, + "step": 9803 + }, + { + "epoch": 2.27, + "learning_rate": 3.111855789230414e-05, + "loss": 0.0, + "step": 9804 + }, + { + "epoch": 2.27, + "learning_rate": 3.111663200061629e-05, + "loss": 0.0, + "step": 9805 + }, + { + "epoch": 2.27, + "learning_rate": 3.1114706108928436e-05, + "loss": 0.0, + "step": 9806 + }, + { + "epoch": 2.27, + "learning_rate": 3.111278021724058e-05, + "loss": 0.0, + "step": 9807 + }, + { + "epoch": 2.27, + "learning_rate": 3.111085432555274e-05, + "loss": 0.0007, + "step": 9808 + }, + { + "epoch": 2.27, + "learning_rate": 3.110892843386488e-05, + "loss": 0.0004, + "step": 9809 + }, + { + "epoch": 2.27, + "learning_rate": 3.1107002542177025e-05, + "loss": 0.0, + "step": 9810 + }, + { + "epoch": 2.27, + "learning_rate": 3.110507665048918e-05, + "loss": 0.0, + "step": 9811 + }, + { + "epoch": 2.27, + "learning_rate": 3.110315075880133e-05, + "loss": 0.0001, + "step": 9812 + }, + { + "epoch": 2.27, + "learning_rate": 3.1101224867113475e-05, + "loss": 0.0, + "step": 9813 + }, + { + "epoch": 2.27, + "learning_rate": 3.109929897542562e-05, + "loss": 0.0, + "step": 9814 + }, + { + "epoch": 2.27, + "learning_rate": 3.1097373083737776e-05, + "loss": 0.0, + "step": 9815 + }, + { + "epoch": 2.27, + "learning_rate": 3.109544719204992e-05, + "loss": 0.0, + "step": 9816 + }, + { + "epoch": 2.27, + "learning_rate": 3.109352130036207e-05, + "loss": 0.0, + "step": 9817 + }, + { + "epoch": 2.27, + "learning_rate": 3.109159540867422e-05, + "loss": 0.0, + "step": 9818 + }, + { + "epoch": 2.27, + "learning_rate": 3.1089669516986366e-05, + "loss": 0.0001, + "step": 9819 + }, + { + "epoch": 2.27, + "learning_rate": 3.1087743625298514e-05, + "loss": 0.0, + "step": 9820 + }, + { + "epoch": 2.27, + "learning_rate": 3.108581773361066e-05, + "loss": 0.0, + "step": 9821 + }, + { + "epoch": 2.27, + "learning_rate": 3.1083891841922815e-05, + "loss": 0.0, + "step": 9822 + }, + { + "epoch": 2.27, + "learning_rate": 3.1081965950234956e-05, + "loss": 0.0, + "step": 9823 + }, + { + "epoch": 2.27, + "learning_rate": 3.108004005854711e-05, + "loss": 0.0, + "step": 9824 + }, + { + "epoch": 2.27, + "learning_rate": 3.107811416685926e-05, + "loss": 0.0, + "step": 9825 + }, + { + "epoch": 2.27, + "learning_rate": 3.1076188275171405e-05, + "loss": 0.0, + "step": 9826 + }, + { + "epoch": 2.27, + "learning_rate": 3.107426238348355e-05, + "loss": 0.0, + "step": 9827 + }, + { + "epoch": 2.27, + "learning_rate": 3.10723364917957e-05, + "loss": 0.001, + "step": 9828 + }, + { + "epoch": 2.27, + "learning_rate": 3.1070410600107855e-05, + "loss": 0.0, + "step": 9829 + }, + { + "epoch": 2.27, + "learning_rate": 3.1068484708419995e-05, + "loss": 0.0001, + "step": 9830 + }, + { + "epoch": 2.27, + "learning_rate": 3.106655881673215e-05, + "loss": 0.0, + "step": 9831 + }, + { + "epoch": 2.27, + "learning_rate": 3.10646329250443e-05, + "loss": 0.0, + "step": 9832 + }, + { + "epoch": 2.27, + "learning_rate": 3.1062707033356445e-05, + "loss": 0.0001, + "step": 9833 + }, + { + "epoch": 2.27, + "learning_rate": 3.106078114166859e-05, + "loss": 0.0003, + "step": 9834 + }, + { + "epoch": 2.27, + "learning_rate": 3.1058855249980746e-05, + "loss": 0.0, + "step": 9835 + }, + { + "epoch": 2.27, + "learning_rate": 3.1056929358292894e-05, + "loss": 0.0001, + "step": 9836 + }, + { + "epoch": 2.27, + "learning_rate": 3.1055003466605034e-05, + "loss": 0.0001, + "step": 9837 + }, + { + "epoch": 2.27, + "learning_rate": 3.105307757491719e-05, + "loss": 0.0, + "step": 9838 + }, + { + "epoch": 2.27, + "learning_rate": 3.1051151683229336e-05, + "loss": 0.0, + "step": 9839 + }, + { + "epoch": 2.27, + "learning_rate": 3.1049225791541484e-05, + "loss": 0.0, + "step": 9840 + }, + { + "epoch": 2.27, + "learning_rate": 3.104729989985363e-05, + "loss": 0.0, + "step": 9841 + }, + { + "epoch": 2.27, + "learning_rate": 3.1045374008165785e-05, + "loss": 0.0, + "step": 9842 + }, + { + "epoch": 2.27, + "learning_rate": 3.104344811647793e-05, + "loss": 0.0001, + "step": 9843 + }, + { + "epoch": 2.28, + "learning_rate": 3.104152222479008e-05, + "loss": 0.0001, + "step": 9844 + }, + { + "epoch": 2.28, + "learning_rate": 3.103959633310223e-05, + "loss": 0.0003, + "step": 9845 + }, + { + "epoch": 2.28, + "learning_rate": 3.1037670441414375e-05, + "loss": 0.0, + "step": 9846 + }, + { + "epoch": 2.28, + "learning_rate": 3.103574454972652e-05, + "loss": 0.0003, + "step": 9847 + }, + { + "epoch": 2.28, + "learning_rate": 3.103381865803867e-05, + "loss": 0.0, + "step": 9848 + }, + { + "epoch": 2.28, + "learning_rate": 3.1031892766350824e-05, + "loss": 0.0, + "step": 9849 + }, + { + "epoch": 2.28, + "learning_rate": 3.102996687466297e-05, + "loss": 0.0, + "step": 9850 + }, + { + "epoch": 2.28, + "learning_rate": 3.102804098297512e-05, + "loss": 0.0, + "step": 9851 + }, + { + "epoch": 2.28, + "learning_rate": 3.102611509128727e-05, + "loss": 0.0002, + "step": 9852 + }, + { + "epoch": 2.28, + "learning_rate": 3.102418919959942e-05, + "loss": 0.0, + "step": 9853 + }, + { + "epoch": 2.28, + "learning_rate": 3.102226330791156e-05, + "loss": 0.0, + "step": 9854 + }, + { + "epoch": 2.28, + "learning_rate": 3.102033741622371e-05, + "loss": 0.0, + "step": 9855 + }, + { + "epoch": 2.28, + "learning_rate": 3.1018411524535864e-05, + "loss": 0.0, + "step": 9856 + }, + { + "epoch": 2.28, + "learning_rate": 3.101648563284801e-05, + "loss": 0.0141, + "step": 9857 + }, + { + "epoch": 2.28, + "learning_rate": 3.101455974116016e-05, + "loss": 0.0, + "step": 9858 + }, + { + "epoch": 2.28, + "learning_rate": 3.1012633849472306e-05, + "loss": 0.0001, + "step": 9859 + }, + { + "epoch": 2.28, + "learning_rate": 3.101070795778446e-05, + "loss": 0.0, + "step": 9860 + }, + { + "epoch": 2.28, + "learning_rate": 3.10087820660966e-05, + "loss": 0.0, + "step": 9861 + }, + { + "epoch": 2.28, + "learning_rate": 3.1006856174408755e-05, + "loss": 0.0, + "step": 9862 + }, + { + "epoch": 2.28, + "learning_rate": 3.10049302827209e-05, + "loss": 0.0001, + "step": 9863 + }, + { + "epoch": 2.28, + "learning_rate": 3.100300439103305e-05, + "loss": 0.0, + "step": 9864 + }, + { + "epoch": 2.28, + "learning_rate": 3.10010784993452e-05, + "loss": 0.0, + "step": 9865 + }, + { + "epoch": 2.28, + "learning_rate": 3.0999152607657345e-05, + "loss": 0.0002, + "step": 9866 + }, + { + "epoch": 2.28, + "learning_rate": 3.09972267159695e-05, + "loss": 0.0, + "step": 9867 + }, + { + "epoch": 2.28, + "learning_rate": 3.099530082428164e-05, + "loss": 0.0, + "step": 9868 + }, + { + "epoch": 2.28, + "learning_rate": 3.0993374932593794e-05, + "loss": 0.0, + "step": 9869 + }, + { + "epoch": 2.28, + "learning_rate": 3.099144904090594e-05, + "loss": 0.0002, + "step": 9870 + }, + { + "epoch": 2.28, + "learning_rate": 3.098952314921809e-05, + "loss": 0.0, + "step": 9871 + }, + { + "epoch": 2.28, + "learning_rate": 3.098759725753024e-05, + "loss": 0.0, + "step": 9872 + }, + { + "epoch": 2.28, + "learning_rate": 3.098567136584239e-05, + "loss": 0.0, + "step": 9873 + }, + { + "epoch": 2.28, + "learning_rate": 3.098374547415454e-05, + "loss": 0.0, + "step": 9874 + }, + { + "epoch": 2.28, + "learning_rate": 3.098181958246668e-05, + "loss": 0.0, + "step": 9875 + }, + { + "epoch": 2.28, + "learning_rate": 3.0979893690778833e-05, + "loss": 0.0, + "step": 9876 + }, + { + "epoch": 2.28, + "learning_rate": 3.097796779909098e-05, + "loss": 0.0, + "step": 9877 + }, + { + "epoch": 2.28, + "learning_rate": 3.097604190740313e-05, + "loss": 0.0, + "step": 9878 + }, + { + "epoch": 2.28, + "learning_rate": 3.0974116015715276e-05, + "loss": 0.0, + "step": 9879 + }, + { + "epoch": 2.28, + "learning_rate": 3.097219012402743e-05, + "loss": 0.0, + "step": 9880 + }, + { + "epoch": 2.28, + "learning_rate": 3.097026423233957e-05, + "loss": 0.0, + "step": 9881 + }, + { + "epoch": 2.28, + "learning_rate": 3.0968338340651725e-05, + "loss": 0.0, + "step": 9882 + }, + { + "epoch": 2.28, + "learning_rate": 3.096641244896387e-05, + "loss": 0.0, + "step": 9883 + }, + { + "epoch": 2.28, + "learning_rate": 3.096448655727602e-05, + "loss": 0.0, + "step": 9884 + }, + { + "epoch": 2.28, + "learning_rate": 3.096256066558817e-05, + "loss": 0.0, + "step": 9885 + }, + { + "epoch": 2.28, + "learning_rate": 3.0960634773900315e-05, + "loss": 0.0, + "step": 9886 + }, + { + "epoch": 2.28, + "learning_rate": 3.095870888221247e-05, + "loss": 0.0, + "step": 9887 + }, + { + "epoch": 2.29, + "learning_rate": 3.095678299052461e-05, + "loss": 0.0, + "step": 9888 + }, + { + "epoch": 2.29, + "learning_rate": 3.0954857098836764e-05, + "loss": 0.0, + "step": 9889 + }, + { + "epoch": 2.29, + "learning_rate": 3.095293120714891e-05, + "loss": 0.0001, + "step": 9890 + }, + { + "epoch": 2.29, + "learning_rate": 3.095100531546106e-05, + "loss": 0.0, + "step": 9891 + }, + { + "epoch": 2.29, + "learning_rate": 3.094907942377321e-05, + "loss": 0.0001, + "step": 9892 + }, + { + "epoch": 2.29, + "learning_rate": 3.0947153532085354e-05, + "loss": 0.0001, + "step": 9893 + }, + { + "epoch": 2.29, + "learning_rate": 3.094522764039751e-05, + "loss": 0.0, + "step": 9894 + }, + { + "epoch": 2.29, + "learning_rate": 3.094330174870965e-05, + "loss": 0.0, + "step": 9895 + }, + { + "epoch": 2.29, + "learning_rate": 3.09413758570218e-05, + "loss": 0.0, + "step": 9896 + }, + { + "epoch": 2.29, + "learning_rate": 3.093944996533395e-05, + "loss": 0.0, + "step": 9897 + }, + { + "epoch": 2.29, + "learning_rate": 3.09375240736461e-05, + "loss": 0.0, + "step": 9898 + }, + { + "epoch": 2.29, + "learning_rate": 3.0935598181958246e-05, + "loss": 0.0, + "step": 9899 + }, + { + "epoch": 2.29, + "learning_rate": 3.09336722902704e-05, + "loss": 0.0002, + "step": 9900 + }, + { + "epoch": 2.29, + "learning_rate": 3.093174639858255e-05, + "loss": 0.0, + "step": 9901 + }, + { + "epoch": 2.29, + "learning_rate": 3.092982050689469e-05, + "loss": 0.0, + "step": 9902 + }, + { + "epoch": 2.29, + "learning_rate": 3.092789461520684e-05, + "loss": 0.0, + "step": 9903 + }, + { + "epoch": 2.29, + "learning_rate": 3.092596872351899e-05, + "loss": 0.0, + "step": 9904 + }, + { + "epoch": 2.29, + "learning_rate": 3.092404283183114e-05, + "loss": 0.0, + "step": 9905 + }, + { + "epoch": 2.29, + "learning_rate": 3.0922116940143285e-05, + "loss": 0.0, + "step": 9906 + }, + { + "epoch": 2.29, + "learning_rate": 3.092019104845544e-05, + "loss": 0.0001, + "step": 9907 + }, + { + "epoch": 2.29, + "learning_rate": 3.091826515676759e-05, + "loss": 0.0, + "step": 9908 + }, + { + "epoch": 2.29, + "learning_rate": 3.0916339265079734e-05, + "loss": 0.0, + "step": 9909 + }, + { + "epoch": 2.29, + "learning_rate": 3.091441337339188e-05, + "loss": 0.0, + "step": 9910 + }, + { + "epoch": 2.29, + "learning_rate": 3.091248748170403e-05, + "loss": 0.0, + "step": 9911 + }, + { + "epoch": 2.29, + "learning_rate": 3.0910561590016177e-05, + "loss": 0.0, + "step": 9912 + }, + { + "epoch": 2.29, + "learning_rate": 3.0908635698328324e-05, + "loss": 0.0, + "step": 9913 + }, + { + "epoch": 2.29, + "learning_rate": 3.090670980664048e-05, + "loss": 0.0, + "step": 9914 + }, + { + "epoch": 2.29, + "learning_rate": 3.0904783914952626e-05, + "loss": 0.0, + "step": 9915 + }, + { + "epoch": 2.29, + "learning_rate": 3.090285802326477e-05, + "loss": 0.0, + "step": 9916 + }, + { + "epoch": 2.29, + "learning_rate": 3.090093213157692e-05, + "loss": 0.0, + "step": 9917 + }, + { + "epoch": 2.29, + "learning_rate": 3.0899006239889075e-05, + "loss": 0.0, + "step": 9918 + }, + { + "epoch": 2.29, + "learning_rate": 3.0897080348201216e-05, + "loss": 0.0001, + "step": 9919 + }, + { + "epoch": 2.29, + "learning_rate": 3.089515445651337e-05, + "loss": 0.0003, + "step": 9920 + }, + { + "epoch": 2.29, + "learning_rate": 3.089322856482552e-05, + "loss": 0.0, + "step": 9921 + }, + { + "epoch": 2.29, + "learning_rate": 3.0891302673137665e-05, + "loss": 0.0, + "step": 9922 + }, + { + "epoch": 2.29, + "learning_rate": 3.088937678144981e-05, + "loss": 0.0, + "step": 9923 + }, + { + "epoch": 2.29, + "learning_rate": 3.088745088976196e-05, + "loss": 0.0, + "step": 9924 + }, + { + "epoch": 2.29, + "learning_rate": 3.0885524998074114e-05, + "loss": 0.0002, + "step": 9925 + }, + { + "epoch": 2.29, + "learning_rate": 3.0883599106386255e-05, + "loss": 0.0002, + "step": 9926 + }, + { + "epoch": 2.29, + "learning_rate": 3.088167321469841e-05, + "loss": 0.0, + "step": 9927 + }, + { + "epoch": 2.29, + "learning_rate": 3.0879747323010557e-05, + "loss": 0.0001, + "step": 9928 + }, + { + "epoch": 2.29, + "learning_rate": 3.0877821431322704e-05, + "loss": 0.0026, + "step": 9929 + }, + { + "epoch": 2.29, + "learning_rate": 3.087589553963485e-05, + "loss": 0.0003, + "step": 9930 + }, + { + "epoch": 2.3, + "learning_rate": 3.0873969647947e-05, + "loss": 0.0, + "step": 9931 + }, + { + "epoch": 2.3, + "learning_rate": 3.087204375625915e-05, + "loss": 0.0, + "step": 9932 + }, + { + "epoch": 2.3, + "learning_rate": 3.0870117864571294e-05, + "loss": 0.0001, + "step": 9933 + }, + { + "epoch": 2.3, + "learning_rate": 3.086819197288345e-05, + "loss": 0.001, + "step": 9934 + }, + { + "epoch": 2.3, + "learning_rate": 3.0866266081195596e-05, + "loss": 0.0, + "step": 9935 + }, + { + "epoch": 2.3, + "learning_rate": 3.086434018950774e-05, + "loss": 0.0001, + "step": 9936 + }, + { + "epoch": 2.3, + "learning_rate": 3.086241429781989e-05, + "loss": 0.0002, + "step": 9937 + }, + { + "epoch": 2.3, + "learning_rate": 3.0860488406132045e-05, + "loss": 0.0, + "step": 9938 + }, + { + "epoch": 2.3, + "learning_rate": 3.085856251444419e-05, + "loss": 0.0001, + "step": 9939 + }, + { + "epoch": 2.3, + "learning_rate": 3.085663662275633e-05, + "loss": 0.0, + "step": 9940 + }, + { + "epoch": 2.3, + "learning_rate": 3.085471073106849e-05, + "loss": 0.0, + "step": 9941 + }, + { + "epoch": 2.3, + "learning_rate": 3.0852784839380635e-05, + "loss": 0.0, + "step": 9942 + }, + { + "epoch": 2.3, + "learning_rate": 3.085085894769278e-05, + "loss": 0.0, + "step": 9943 + }, + { + "epoch": 2.3, + "learning_rate": 3.084893305600493e-05, + "loss": 0.0, + "step": 9944 + }, + { + "epoch": 2.3, + "learning_rate": 3.0847007164317084e-05, + "loss": 0.0, + "step": 9945 + }, + { + "epoch": 2.3, + "learning_rate": 3.084508127262923e-05, + "loss": 0.0, + "step": 9946 + }, + { + "epoch": 2.3, + "learning_rate": 3.084315538094138e-05, + "loss": 0.0, + "step": 9947 + }, + { + "epoch": 2.3, + "learning_rate": 3.0841229489253526e-05, + "loss": 0.0001, + "step": 9948 + }, + { + "epoch": 2.3, + "learning_rate": 3.0839303597565674e-05, + "loss": 0.0, + "step": 9949 + }, + { + "epoch": 2.3, + "learning_rate": 3.083737770587782e-05, + "loss": 0.0002, + "step": 9950 + }, + { + "epoch": 2.3, + "learning_rate": 3.083545181418997e-05, + "loss": 0.0001, + "step": 9951 + }, + { + "epoch": 2.3, + "learning_rate": 3.083352592250212e-05, + "loss": 0.0, + "step": 9952 + }, + { + "epoch": 2.3, + "learning_rate": 3.0831600030814264e-05, + "loss": 0.0, + "step": 9953 + }, + { + "epoch": 2.3, + "learning_rate": 3.082967413912642e-05, + "loss": 0.0, + "step": 9954 + }, + { + "epoch": 2.3, + "learning_rate": 3.0827748247438566e-05, + "loss": 0.0, + "step": 9955 + }, + { + "epoch": 2.3, + "learning_rate": 3.082582235575072e-05, + "loss": 0.0, + "step": 9956 + }, + { + "epoch": 2.3, + "learning_rate": 3.082389646406286e-05, + "loss": 0.0, + "step": 9957 + }, + { + "epoch": 2.3, + "learning_rate": 3.082197057237501e-05, + "loss": 0.0, + "step": 9958 + }, + { + "epoch": 2.3, + "learning_rate": 3.082004468068716e-05, + "loss": 0.0, + "step": 9959 + }, + { + "epoch": 2.3, + "learning_rate": 3.08181187889993e-05, + "loss": 0.0, + "step": 9960 + }, + { + "epoch": 2.3, + "learning_rate": 3.081619289731146e-05, + "loss": 0.0, + "step": 9961 + }, + { + "epoch": 2.3, + "learning_rate": 3.0814267005623605e-05, + "loss": 0.0, + "step": 9962 + }, + { + "epoch": 2.3, + "learning_rate": 3.081234111393575e-05, + "loss": 0.0, + "step": 9963 + }, + { + "epoch": 2.3, + "learning_rate": 3.08104152222479e-05, + "loss": 0.0, + "step": 9964 + }, + { + "epoch": 2.3, + "learning_rate": 3.0808489330560054e-05, + "loss": 0.0, + "step": 9965 + }, + { + "epoch": 2.3, + "learning_rate": 3.08065634388722e-05, + "loss": 0.0005, + "step": 9966 + }, + { + "epoch": 2.3, + "learning_rate": 3.080463754718434e-05, + "loss": 0.0, + "step": 9967 + }, + { + "epoch": 2.3, + "learning_rate": 3.0802711655496496e-05, + "loss": 0.0, + "step": 9968 + }, + { + "epoch": 2.3, + "learning_rate": 3.0800785763808644e-05, + "loss": 0.0047, + "step": 9969 + }, + { + "epoch": 2.3, + "learning_rate": 3.079885987212079e-05, + "loss": 0.0, + "step": 9970 + }, + { + "epoch": 2.3, + "learning_rate": 3.079693398043294e-05, + "loss": 0.0, + "step": 9971 + }, + { + "epoch": 2.3, + "learning_rate": 3.079500808874509e-05, + "loss": 0.0, + "step": 9972 + }, + { + "epoch": 2.3, + "learning_rate": 3.079308219705724e-05, + "loss": 0.0, + "step": 9973 + }, + { + "epoch": 2.31, + "learning_rate": 3.079115630536939e-05, + "loss": 0.0001, + "step": 9974 + }, + { + "epoch": 2.31, + "learning_rate": 3.0789230413681535e-05, + "loss": 0.0, + "step": 9975 + }, + { + "epoch": 2.31, + "learning_rate": 3.078730452199369e-05, + "loss": 0.0, + "step": 9976 + }, + { + "epoch": 2.31, + "learning_rate": 3.078537863030583e-05, + "loss": 0.0, + "step": 9977 + }, + { + "epoch": 2.31, + "learning_rate": 3.078345273861798e-05, + "loss": 0.0, + "step": 9978 + }, + { + "epoch": 2.31, + "learning_rate": 3.078152684693013e-05, + "loss": 0.0, + "step": 9979 + }, + { + "epoch": 2.31, + "learning_rate": 3.077960095524228e-05, + "loss": 0.0, + "step": 9980 + }, + { + "epoch": 2.31, + "learning_rate": 3.077767506355443e-05, + "loss": 0.0002, + "step": 9981 + }, + { + "epoch": 2.31, + "learning_rate": 3.0775749171866575e-05, + "loss": 0.0, + "step": 9982 + }, + { + "epoch": 2.31, + "learning_rate": 3.077382328017873e-05, + "loss": 0.0001, + "step": 9983 + }, + { + "epoch": 2.31, + "learning_rate": 3.077189738849087e-05, + "loss": 0.0, + "step": 9984 + }, + { + "epoch": 2.31, + "learning_rate": 3.0769971496803024e-05, + "loss": 0.0, + "step": 9985 + }, + { + "epoch": 2.31, + "learning_rate": 3.076804560511517e-05, + "loss": 0.0, + "step": 9986 + }, + { + "epoch": 2.31, + "learning_rate": 3.076611971342732e-05, + "loss": 0.0, + "step": 9987 + }, + { + "epoch": 2.31, + "learning_rate": 3.0764193821739466e-05, + "loss": 0.0, + "step": 9988 + }, + { + "epoch": 2.31, + "learning_rate": 3.0762267930051614e-05, + "loss": 0.0001, + "step": 9989 + }, + { + "epoch": 2.31, + "learning_rate": 3.076034203836377e-05, + "loss": 0.0, + "step": 9990 + }, + { + "epoch": 2.31, + "learning_rate": 3.075841614667591e-05, + "loss": 0.0, + "step": 9991 + }, + { + "epoch": 2.31, + "learning_rate": 3.075649025498806e-05, + "loss": 0.0, + "step": 9992 + }, + { + "epoch": 2.31, + "learning_rate": 3.075456436330021e-05, + "loss": 0.0001, + "step": 9993 + }, + { + "epoch": 2.31, + "learning_rate": 3.075263847161236e-05, + "loss": 0.0, + "step": 9994 + }, + { + "epoch": 2.31, + "learning_rate": 3.0750712579924505e-05, + "loss": 0.0, + "step": 9995 + }, + { + "epoch": 2.31, + "learning_rate": 3.074878668823665e-05, + "loss": 0.0, + "step": 9996 + }, + { + "epoch": 2.31, + "learning_rate": 3.074686079654881e-05, + "loss": 0.0068, + "step": 9997 + }, + { + "epoch": 2.31, + "learning_rate": 3.074493490486095e-05, + "loss": 0.0001, + "step": 9998 + }, + { + "epoch": 2.31, + "learning_rate": 3.07430090131731e-05, + "loss": 0.0, + "step": 9999 + }, + { + "epoch": 2.31, + "learning_rate": 3.074108312148525e-05, + "loss": 0.0, + "step": 10000 + }, + { + "epoch": 2.31, + "learning_rate": 3.07391572297974e-05, + "loss": 0.0, + "step": 10001 + }, + { + "epoch": 2.31, + "learning_rate": 3.0737231338109544e-05, + "loss": 0.0, + "step": 10002 + }, + { + "epoch": 2.31, + "learning_rate": 3.07353054464217e-05, + "loss": 0.0, + "step": 10003 + }, + { + "epoch": 2.31, + "learning_rate": 3.0733379554733846e-05, + "loss": 0.0, + "step": 10004 + }, + { + "epoch": 2.31, + "learning_rate": 3.073145366304599e-05, + "loss": 0.0, + "step": 10005 + }, + { + "epoch": 2.31, + "learning_rate": 3.072952777135814e-05, + "loss": 0.0, + "step": 10006 + }, + { + "epoch": 2.31, + "learning_rate": 3.072760187967029e-05, + "loss": 0.0, + "step": 10007 + }, + { + "epoch": 2.31, + "learning_rate": 3.0725675987982436e-05, + "loss": 0.0, + "step": 10008 + }, + { + "epoch": 2.31, + "learning_rate": 3.0723750096294584e-05, + "loss": 0.0, + "step": 10009 + }, + { + "epoch": 2.31, + "learning_rate": 3.072182420460674e-05, + "loss": 0.0, + "step": 10010 + }, + { + "epoch": 2.31, + "learning_rate": 3.0719898312918885e-05, + "loss": 0.0001, + "step": 10011 + }, + { + "epoch": 2.31, + "learning_rate": 3.071797242123103e-05, + "loss": 0.0, + "step": 10012 + }, + { + "epoch": 2.31, + "learning_rate": 3.071604652954318e-05, + "loss": 0.0, + "step": 10013 + }, + { + "epoch": 2.31, + "learning_rate": 3.071412063785533e-05, + "loss": 0.0, + "step": 10014 + }, + { + "epoch": 2.31, + "learning_rate": 3.0712194746167475e-05, + "loss": 0.0, + "step": 10015 + }, + { + "epoch": 2.31, + "learning_rate": 3.071026885447962e-05, + "loss": 0.0001, + "step": 10016 + }, + { + "epoch": 2.31, + "learning_rate": 3.070834296279178e-05, + "loss": 0.0, + "step": 10017 + }, + { + "epoch": 2.32, + "learning_rate": 3.0706417071103924e-05, + "loss": 0.0, + "step": 10018 + }, + { + "epoch": 2.32, + "learning_rate": 3.070449117941607e-05, + "loss": 0.0, + "step": 10019 + }, + { + "epoch": 2.32, + "learning_rate": 3.070256528772822e-05, + "loss": 0.0001, + "step": 10020 + }, + { + "epoch": 2.32, + "learning_rate": 3.0700639396040374e-05, + "loss": 0.0, + "step": 10021 + }, + { + "epoch": 2.32, + "learning_rate": 3.0698713504352514e-05, + "loss": 0.0003, + "step": 10022 + }, + { + "epoch": 2.32, + "learning_rate": 3.069678761266466e-05, + "loss": 0.0001, + "step": 10023 + }, + { + "epoch": 2.32, + "learning_rate": 3.0694861720976816e-05, + "loss": 0.0, + "step": 10024 + }, + { + "epoch": 2.32, + "learning_rate": 3.0692935829288964e-05, + "loss": 0.0, + "step": 10025 + }, + { + "epoch": 2.32, + "learning_rate": 3.069100993760111e-05, + "loss": 0.0, + "step": 10026 + }, + { + "epoch": 2.32, + "learning_rate": 3.068908404591326e-05, + "loss": 0.0001, + "step": 10027 + }, + { + "epoch": 2.32, + "learning_rate": 3.068715815422541e-05, + "loss": 0.0019, + "step": 10028 + }, + { + "epoch": 2.32, + "learning_rate": 3.0685232262537553e-05, + "loss": 0.0, + "step": 10029 + }, + { + "epoch": 2.32, + "learning_rate": 3.068330637084971e-05, + "loss": 0.0, + "step": 10030 + }, + { + "epoch": 2.32, + "learning_rate": 3.0681380479161855e-05, + "loss": 0.0005, + "step": 10031 + }, + { + "epoch": 2.32, + "learning_rate": 3.0679454587474e-05, + "loss": 0.0, + "step": 10032 + }, + { + "epoch": 2.32, + "learning_rate": 3.067752869578615e-05, + "loss": 0.0, + "step": 10033 + }, + { + "epoch": 2.32, + "learning_rate": 3.06756028040983e-05, + "loss": 0.0, + "step": 10034 + }, + { + "epoch": 2.32, + "learning_rate": 3.0673676912410445e-05, + "loss": 0.0, + "step": 10035 + }, + { + "epoch": 2.32, + "learning_rate": 3.067175102072259e-05, + "loss": 0.0, + "step": 10036 + }, + { + "epoch": 2.32, + "learning_rate": 3.066982512903475e-05, + "loss": 0.0, + "step": 10037 + }, + { + "epoch": 2.32, + "learning_rate": 3.0667899237346894e-05, + "loss": 0.0001, + "step": 10038 + }, + { + "epoch": 2.32, + "learning_rate": 3.066597334565904e-05, + "loss": 0.0, + "step": 10039 + }, + { + "epoch": 2.32, + "learning_rate": 3.066404745397119e-05, + "loss": 0.0, + "step": 10040 + }, + { + "epoch": 2.32, + "learning_rate": 3.0662121562283344e-05, + "loss": 0.0, + "step": 10041 + }, + { + "epoch": 2.32, + "learning_rate": 3.0660195670595484e-05, + "loss": 0.0, + "step": 10042 + }, + { + "epoch": 2.32, + "learning_rate": 3.065826977890763e-05, + "loss": 0.0, + "step": 10043 + }, + { + "epoch": 2.32, + "learning_rate": 3.0656343887219786e-05, + "loss": 0.0, + "step": 10044 + }, + { + "epoch": 2.32, + "learning_rate": 3.0654417995531933e-05, + "loss": 0.0, + "step": 10045 + }, + { + "epoch": 2.32, + "learning_rate": 3.065249210384408e-05, + "loss": 0.0, + "step": 10046 + }, + { + "epoch": 2.32, + "learning_rate": 3.065056621215623e-05, + "loss": 0.0, + "step": 10047 + }, + { + "epoch": 2.32, + "learning_rate": 3.064864032046838e-05, + "loss": 0.0001, + "step": 10048 + }, + { + "epoch": 2.32, + "learning_rate": 3.064671442878052e-05, + "loss": 0.0, + "step": 10049 + }, + { + "epoch": 2.32, + "learning_rate": 3.064478853709268e-05, + "loss": 0.0, + "step": 10050 + }, + { + "epoch": 2.32, + "learning_rate": 3.0642862645404825e-05, + "loss": 0.0003, + "step": 10051 + }, + { + "epoch": 2.32, + "learning_rate": 3.064093675371697e-05, + "loss": 0.0001, + "step": 10052 + }, + { + "epoch": 2.32, + "learning_rate": 3.063901086202912e-05, + "loss": 0.0, + "step": 10053 + }, + { + "epoch": 2.32, + "learning_rate": 3.063708497034127e-05, + "loss": 0.0, + "step": 10054 + }, + { + "epoch": 2.32, + "learning_rate": 3.063515907865342e-05, + "loss": 0.0002, + "step": 10055 + }, + { + "epoch": 2.32, + "learning_rate": 3.063323318696556e-05, + "loss": 0.0006, + "step": 10056 + }, + { + "epoch": 2.32, + "learning_rate": 3.063130729527772e-05, + "loss": 0.0, + "step": 10057 + }, + { + "epoch": 2.32, + "learning_rate": 3.0629381403589864e-05, + "loss": 0.0, + "step": 10058 + }, + { + "epoch": 2.32, + "learning_rate": 3.062745551190201e-05, + "loss": 0.0, + "step": 10059 + }, + { + "epoch": 2.32, + "learning_rate": 3.062552962021416e-05, + "loss": 0.0, + "step": 10060 + }, + { + "epoch": 2.33, + "learning_rate": 3.0623603728526307e-05, + "loss": 0.0109, + "step": 10061 + }, + { + "epoch": 2.33, + "learning_rate": 3.062167783683846e-05, + "loss": 0.0, + "step": 10062 + }, + { + "epoch": 2.33, + "learning_rate": 3.06197519451506e-05, + "loss": 0.0, + "step": 10063 + }, + { + "epoch": 2.33, + "learning_rate": 3.0617826053462756e-05, + "loss": 0.0, + "step": 10064 + }, + { + "epoch": 2.33, + "learning_rate": 3.06159001617749e-05, + "loss": 0.0003, + "step": 10065 + }, + { + "epoch": 2.33, + "learning_rate": 3.061397427008705e-05, + "loss": 0.0, + "step": 10066 + }, + { + "epoch": 2.33, + "learning_rate": 3.06120483783992e-05, + "loss": 0.0, + "step": 10067 + }, + { + "epoch": 2.33, + "learning_rate": 3.061012248671135e-05, + "loss": 0.0001, + "step": 10068 + }, + { + "epoch": 2.33, + "learning_rate": 3.06081965950235e-05, + "loss": 0.0001, + "step": 10069 + }, + { + "epoch": 2.33, + "learning_rate": 3.060627070333564e-05, + "loss": 0.0, + "step": 10070 + }, + { + "epoch": 2.33, + "learning_rate": 3.0604344811647795e-05, + "loss": 0.0, + "step": 10071 + }, + { + "epoch": 2.33, + "learning_rate": 3.060241891995994e-05, + "loss": 0.0, + "step": 10072 + }, + { + "epoch": 2.33, + "learning_rate": 3.060049302827209e-05, + "loss": 0.0, + "step": 10073 + }, + { + "epoch": 2.33, + "learning_rate": 3.059856713658424e-05, + "loss": 0.0001, + "step": 10074 + }, + { + "epoch": 2.33, + "learning_rate": 3.059664124489639e-05, + "loss": 0.0, + "step": 10075 + }, + { + "epoch": 2.33, + "learning_rate": 3.059471535320854e-05, + "loss": 0.0, + "step": 10076 + }, + { + "epoch": 2.33, + "learning_rate": 3.0592789461520687e-05, + "loss": 0.0, + "step": 10077 + }, + { + "epoch": 2.33, + "learning_rate": 3.0590863569832834e-05, + "loss": 0.0, + "step": 10078 + }, + { + "epoch": 2.33, + "learning_rate": 3.058893767814498e-05, + "loss": 0.0, + "step": 10079 + }, + { + "epoch": 2.33, + "learning_rate": 3.058701178645713e-05, + "loss": 0.0001, + "step": 10080 + }, + { + "epoch": 2.33, + "learning_rate": 3.0585085894769276e-05, + "loss": 0.0, + "step": 10081 + }, + { + "epoch": 2.33, + "learning_rate": 3.058316000308143e-05, + "loss": 0.0, + "step": 10082 + }, + { + "epoch": 2.33, + "learning_rate": 3.058123411139358e-05, + "loss": 0.0, + "step": 10083 + }, + { + "epoch": 2.33, + "learning_rate": 3.0579308219705726e-05, + "loss": 0.0001, + "step": 10084 + }, + { + "epoch": 2.33, + "learning_rate": 3.057738232801787e-05, + "loss": 0.0, + "step": 10085 + }, + { + "epoch": 2.33, + "learning_rate": 3.057545643633003e-05, + "loss": 0.0, + "step": 10086 + }, + { + "epoch": 2.33, + "learning_rate": 3.057353054464217e-05, + "loss": 0.0, + "step": 10087 + }, + { + "epoch": 2.33, + "learning_rate": 3.057160465295432e-05, + "loss": 0.0, + "step": 10088 + }, + { + "epoch": 2.33, + "learning_rate": 3.056967876126647e-05, + "loss": 0.0, + "step": 10089 + }, + { + "epoch": 2.33, + "learning_rate": 3.056775286957862e-05, + "loss": 0.0003, + "step": 10090 + }, + { + "epoch": 2.33, + "learning_rate": 3.0565826977890765e-05, + "loss": 0.0, + "step": 10091 + }, + { + "epoch": 2.33, + "learning_rate": 3.056390108620291e-05, + "loss": 0.0, + "step": 10092 + }, + { + "epoch": 2.33, + "learning_rate": 3.0561975194515067e-05, + "loss": 0.0002, + "step": 10093 + }, + { + "epoch": 2.33, + "learning_rate": 3.056004930282721e-05, + "loss": 0.0, + "step": 10094 + }, + { + "epoch": 2.33, + "learning_rate": 3.055812341113936e-05, + "loss": 0.0, + "step": 10095 + }, + { + "epoch": 2.33, + "learning_rate": 3.055619751945151e-05, + "loss": 0.0, + "step": 10096 + }, + { + "epoch": 2.33, + "learning_rate": 3.0554271627763656e-05, + "loss": 0.0001, + "step": 10097 + }, + { + "epoch": 2.33, + "learning_rate": 3.0552345736075804e-05, + "loss": 0.0002, + "step": 10098 + }, + { + "epoch": 2.33, + "learning_rate": 3.055041984438795e-05, + "loss": 0.0, + "step": 10099 + }, + { + "epoch": 2.33, + "learning_rate": 3.0548493952700106e-05, + "loss": 0.0, + "step": 10100 + }, + { + "epoch": 2.33, + "learning_rate": 3.0546568061012246e-05, + "loss": 0.0004, + "step": 10101 + }, + { + "epoch": 2.33, + "learning_rate": 3.05446421693244e-05, + "loss": 0.0, + "step": 10102 + }, + { + "epoch": 2.33, + "learning_rate": 3.054271627763655e-05, + "loss": 0.0, + "step": 10103 + }, + { + "epoch": 2.34, + "learning_rate": 3.0540790385948696e-05, + "loss": 0.0, + "step": 10104 + }, + { + "epoch": 2.34, + "learning_rate": 3.053886449426084e-05, + "loss": 0.0002, + "step": 10105 + }, + { + "epoch": 2.34, + "learning_rate": 3.0536938602573e-05, + "loss": 0.0, + "step": 10106 + }, + { + "epoch": 2.34, + "learning_rate": 3.053501271088514e-05, + "loss": 0.0001, + "step": 10107 + }, + { + "epoch": 2.34, + "learning_rate": 3.0533086819197285e-05, + "loss": 0.0003, + "step": 10108 + }, + { + "epoch": 2.34, + "learning_rate": 3.053116092750944e-05, + "loss": 0.0, + "step": 10109 + }, + { + "epoch": 2.34, + "learning_rate": 3.052923503582159e-05, + "loss": 0.0, + "step": 10110 + }, + { + "epoch": 2.34, + "learning_rate": 3.0527309144133735e-05, + "loss": 0.0001, + "step": 10111 + }, + { + "epoch": 2.34, + "learning_rate": 3.052538325244588e-05, + "loss": 0.0, + "step": 10112 + }, + { + "epoch": 2.34, + "learning_rate": 3.0523457360758036e-05, + "loss": 0.0, + "step": 10113 + }, + { + "epoch": 2.34, + "learning_rate": 3.052153146907018e-05, + "loss": 0.0, + "step": 10114 + }, + { + "epoch": 2.34, + "learning_rate": 3.051960557738233e-05, + "loss": 0.0, + "step": 10115 + }, + { + "epoch": 2.34, + "learning_rate": 3.051767968569448e-05, + "loss": 0.0002, + "step": 10116 + }, + { + "epoch": 2.34, + "learning_rate": 3.0515753794006623e-05, + "loss": 0.0001, + "step": 10117 + }, + { + "epoch": 2.34, + "learning_rate": 3.0513827902318777e-05, + "loss": 0.0, + "step": 10118 + }, + { + "epoch": 2.34, + "learning_rate": 3.051190201063092e-05, + "loss": 0.0002, + "step": 10119 + }, + { + "epoch": 2.34, + "learning_rate": 3.0509976118943072e-05, + "loss": 0.0, + "step": 10120 + }, + { + "epoch": 2.34, + "learning_rate": 3.050805022725522e-05, + "loss": 0.0, + "step": 10121 + }, + { + "epoch": 2.34, + "learning_rate": 3.050612433556737e-05, + "loss": 0.0, + "step": 10122 + }, + { + "epoch": 2.34, + "learning_rate": 3.0504198443879518e-05, + "loss": 0.0, + "step": 10123 + }, + { + "epoch": 2.34, + "learning_rate": 3.050227255219167e-05, + "loss": 0.0, + "step": 10124 + }, + { + "epoch": 2.34, + "learning_rate": 3.0500346660503813e-05, + "loss": 0.0, + "step": 10125 + }, + { + "epoch": 2.34, + "learning_rate": 3.049842076881596e-05, + "loss": 0.0, + "step": 10126 + }, + { + "epoch": 2.34, + "learning_rate": 3.049649487712811e-05, + "loss": 0.0001, + "step": 10127 + }, + { + "epoch": 2.34, + "learning_rate": 3.049456898544026e-05, + "loss": 0.0022, + "step": 10128 + }, + { + "epoch": 2.34, + "learning_rate": 3.049264309375241e-05, + "loss": 0.0, + "step": 10129 + }, + { + "epoch": 2.34, + "learning_rate": 3.0490717202064557e-05, + "loss": 0.0001, + "step": 10130 + }, + { + "epoch": 2.34, + "learning_rate": 3.0488791310376708e-05, + "loss": 0.0, + "step": 10131 + }, + { + "epoch": 2.34, + "learning_rate": 3.0486865418688852e-05, + "loss": 0.0, + "step": 10132 + }, + { + "epoch": 2.34, + "learning_rate": 3.0484939527001006e-05, + "loss": 0.0, + "step": 10133 + }, + { + "epoch": 2.34, + "learning_rate": 3.048301363531315e-05, + "loss": 0.0, + "step": 10134 + }, + { + "epoch": 2.34, + "learning_rate": 3.0481087743625298e-05, + "loss": 0.0, + "step": 10135 + }, + { + "epoch": 2.34, + "learning_rate": 3.047916185193745e-05, + "loss": 0.0001, + "step": 10136 + }, + { + "epoch": 2.34, + "learning_rate": 3.0477235960249596e-05, + "loss": 0.0, + "step": 10137 + }, + { + "epoch": 2.34, + "learning_rate": 3.0475310068561747e-05, + "loss": 0.0, + "step": 10138 + }, + { + "epoch": 2.34, + "learning_rate": 3.047338417687389e-05, + "loss": 0.0001, + "step": 10139 + }, + { + "epoch": 2.34, + "learning_rate": 3.0471458285186045e-05, + "loss": 0.0001, + "step": 10140 + }, + { + "epoch": 2.34, + "learning_rate": 3.046953239349819e-05, + "loss": 0.0, + "step": 10141 + }, + { + "epoch": 2.34, + "learning_rate": 3.046760650181034e-05, + "loss": 0.0, + "step": 10142 + }, + { + "epoch": 2.34, + "learning_rate": 3.0465680610122488e-05, + "loss": 0.0, + "step": 10143 + }, + { + "epoch": 2.34, + "learning_rate": 3.046375471843464e-05, + "loss": 0.0001, + "step": 10144 + }, + { + "epoch": 2.34, + "learning_rate": 3.0461828826746786e-05, + "loss": 0.0, + "step": 10145 + }, + { + "epoch": 2.34, + "learning_rate": 3.045990293505893e-05, + "loss": 0.0, + "step": 10146 + }, + { + "epoch": 2.35, + "learning_rate": 3.0457977043371085e-05, + "loss": 0.0, + "step": 10147 + }, + { + "epoch": 2.35, + "learning_rate": 3.045605115168323e-05, + "loss": 0.0002, + "step": 10148 + }, + { + "epoch": 2.35, + "learning_rate": 3.045412525999538e-05, + "loss": 0.0, + "step": 10149 + }, + { + "epoch": 2.35, + "learning_rate": 3.0452199368307527e-05, + "loss": 0.0001, + "step": 10150 + }, + { + "epoch": 2.35, + "learning_rate": 3.0450273476619678e-05, + "loss": 0.0003, + "step": 10151 + }, + { + "epoch": 2.35, + "learning_rate": 3.0448347584931825e-05, + "loss": 0.0, + "step": 10152 + }, + { + "epoch": 2.35, + "learning_rate": 3.0446421693243976e-05, + "loss": 0.0, + "step": 10153 + }, + { + "epoch": 2.35, + "learning_rate": 3.0444495801556124e-05, + "loss": 0.0, + "step": 10154 + }, + { + "epoch": 2.35, + "learning_rate": 3.0442569909868268e-05, + "loss": 0.0, + "step": 10155 + }, + { + "epoch": 2.35, + "learning_rate": 3.044064401818042e-05, + "loss": 0.0007, + "step": 10156 + }, + { + "epoch": 2.35, + "learning_rate": 3.0438718126492566e-05, + "loss": 0.0001, + "step": 10157 + }, + { + "epoch": 2.35, + "learning_rate": 3.0436792234804717e-05, + "loss": 0.0001, + "step": 10158 + }, + { + "epoch": 2.35, + "learning_rate": 3.0434866343116864e-05, + "loss": 0.0, + "step": 10159 + }, + { + "epoch": 2.35, + "learning_rate": 3.0432940451429015e-05, + "loss": 0.0, + "step": 10160 + }, + { + "epoch": 2.35, + "learning_rate": 3.043101455974116e-05, + "loss": 0.0, + "step": 10161 + }, + { + "epoch": 2.35, + "learning_rate": 3.0429088668053314e-05, + "loss": 0.0, + "step": 10162 + }, + { + "epoch": 2.35, + "learning_rate": 3.0427162776365458e-05, + "loss": 0.0, + "step": 10163 + }, + { + "epoch": 2.35, + "learning_rate": 3.0425236884677605e-05, + "loss": 0.0, + "step": 10164 + }, + { + "epoch": 2.35, + "learning_rate": 3.0423310992989756e-05, + "loss": 0.0, + "step": 10165 + }, + { + "epoch": 2.35, + "learning_rate": 3.0421385101301904e-05, + "loss": 0.0002, + "step": 10166 + }, + { + "epoch": 2.35, + "learning_rate": 3.0419459209614054e-05, + "loss": 0.0, + "step": 10167 + }, + { + "epoch": 2.35, + "learning_rate": 3.04175333179262e-05, + "loss": 0.0, + "step": 10168 + }, + { + "epoch": 2.35, + "learning_rate": 3.0415607426238353e-05, + "loss": 0.0001, + "step": 10169 + }, + { + "epoch": 2.35, + "learning_rate": 3.0413681534550497e-05, + "loss": 0.0, + "step": 10170 + }, + { + "epoch": 2.35, + "learning_rate": 3.0411755642862648e-05, + "loss": 0.0005, + "step": 10171 + }, + { + "epoch": 2.35, + "learning_rate": 3.0409829751174795e-05, + "loss": 0.0, + "step": 10172 + }, + { + "epoch": 2.35, + "learning_rate": 3.0407903859486943e-05, + "loss": 0.0, + "step": 10173 + }, + { + "epoch": 2.35, + "learning_rate": 3.0405977967799094e-05, + "loss": 0.0, + "step": 10174 + }, + { + "epoch": 2.35, + "learning_rate": 3.0404052076111238e-05, + "loss": 0.0, + "step": 10175 + }, + { + "epoch": 2.35, + "learning_rate": 3.0402126184423392e-05, + "loss": 0.0, + "step": 10176 + }, + { + "epoch": 2.35, + "learning_rate": 3.0400200292735536e-05, + "loss": 0.0, + "step": 10177 + }, + { + "epoch": 2.35, + "learning_rate": 3.0398274401047687e-05, + "loss": 0.0, + "step": 10178 + }, + { + "epoch": 2.35, + "learning_rate": 3.0396348509359834e-05, + "loss": 0.0001, + "step": 10179 + }, + { + "epoch": 2.35, + "learning_rate": 3.0394422617671985e-05, + "loss": 0.0, + "step": 10180 + }, + { + "epoch": 2.35, + "learning_rate": 3.0392496725984133e-05, + "loss": 0.0, + "step": 10181 + }, + { + "epoch": 2.35, + "learning_rate": 3.0390570834296277e-05, + "loss": 0.0, + "step": 10182 + }, + { + "epoch": 2.35, + "learning_rate": 3.038864494260843e-05, + "loss": 0.0004, + "step": 10183 + }, + { + "epoch": 2.35, + "learning_rate": 3.0386719050920575e-05, + "loss": 0.0, + "step": 10184 + }, + { + "epoch": 2.35, + "learning_rate": 3.0384793159232726e-05, + "loss": 0.0, + "step": 10185 + }, + { + "epoch": 2.35, + "learning_rate": 3.0382867267544873e-05, + "loss": 0.0, + "step": 10186 + }, + { + "epoch": 2.35, + "learning_rate": 3.0380941375857024e-05, + "loss": 0.0, + "step": 10187 + }, + { + "epoch": 2.35, + "learning_rate": 3.0379015484169172e-05, + "loss": 0.0, + "step": 10188 + }, + { + "epoch": 2.35, + "learning_rate": 3.0377089592481323e-05, + "loss": 0.0001, + "step": 10189 + }, + { + "epoch": 2.35, + "learning_rate": 3.037516370079347e-05, + "loss": 0.0001, + "step": 10190 + }, + { + "epoch": 2.36, + "learning_rate": 3.0373237809105614e-05, + "loss": 0.0, + "step": 10191 + }, + { + "epoch": 2.36, + "learning_rate": 3.0371311917417765e-05, + "loss": 0.0, + "step": 10192 + }, + { + "epoch": 2.36, + "learning_rate": 3.0369386025729913e-05, + "loss": 0.0, + "step": 10193 + }, + { + "epoch": 2.36, + "learning_rate": 3.0367460134042063e-05, + "loss": 0.0, + "step": 10194 + }, + { + "epoch": 2.36, + "learning_rate": 3.036553424235421e-05, + "loss": 0.0002, + "step": 10195 + }, + { + "epoch": 2.36, + "learning_rate": 3.0363608350666362e-05, + "loss": 0.0, + "step": 10196 + }, + { + "epoch": 2.36, + "learning_rate": 3.036168245897851e-05, + "loss": 0.0, + "step": 10197 + }, + { + "epoch": 2.36, + "learning_rate": 3.035975656729066e-05, + "loss": 0.0, + "step": 10198 + }, + { + "epoch": 2.36, + "learning_rate": 3.0357830675602804e-05, + "loss": 0.0, + "step": 10199 + }, + { + "epoch": 2.36, + "learning_rate": 3.035590478391496e-05, + "loss": 0.0001, + "step": 10200 + }, + { + "epoch": 2.36, + "learning_rate": 3.0353978892227103e-05, + "loss": 0.0, + "step": 10201 + }, + { + "epoch": 2.36, + "learning_rate": 3.035205300053925e-05, + "loss": 0.0, + "step": 10202 + }, + { + "epoch": 2.36, + "learning_rate": 3.03501271088514e-05, + "loss": 0.0001, + "step": 10203 + }, + { + "epoch": 2.36, + "learning_rate": 3.0348201217163545e-05, + "loss": 0.0, + "step": 10204 + }, + { + "epoch": 2.36, + "learning_rate": 3.03462753254757e-05, + "loss": 0.0, + "step": 10205 + }, + { + "epoch": 2.36, + "learning_rate": 3.0344349433787843e-05, + "loss": 0.0002, + "step": 10206 + }, + { + "epoch": 2.36, + "learning_rate": 3.0342423542099994e-05, + "loss": 0.0043, + "step": 10207 + }, + { + "epoch": 2.36, + "learning_rate": 3.0340497650412142e-05, + "loss": 0.0001, + "step": 10208 + }, + { + "epoch": 2.36, + "learning_rate": 3.0338571758724293e-05, + "loss": 0.0, + "step": 10209 + }, + { + "epoch": 2.36, + "learning_rate": 3.033664586703644e-05, + "loss": 0.0, + "step": 10210 + }, + { + "epoch": 2.36, + "learning_rate": 3.0334719975348584e-05, + "loss": 0.0, + "step": 10211 + }, + { + "epoch": 2.36, + "learning_rate": 3.033279408366074e-05, + "loss": 0.0, + "step": 10212 + }, + { + "epoch": 2.36, + "learning_rate": 3.0330868191972882e-05, + "loss": 0.0001, + "step": 10213 + }, + { + "epoch": 2.36, + "learning_rate": 3.0328942300285033e-05, + "loss": 0.0, + "step": 10214 + }, + { + "epoch": 2.36, + "learning_rate": 3.032701640859718e-05, + "loss": 0.0023, + "step": 10215 + }, + { + "epoch": 2.36, + "learning_rate": 3.032509051690933e-05, + "loss": 0.0, + "step": 10216 + }, + { + "epoch": 2.36, + "learning_rate": 3.032316462522148e-05, + "loss": 0.0, + "step": 10217 + }, + { + "epoch": 2.36, + "learning_rate": 3.032123873353363e-05, + "loss": 0.0, + "step": 10218 + }, + { + "epoch": 2.36, + "learning_rate": 3.0319312841845778e-05, + "loss": 0.0, + "step": 10219 + }, + { + "epoch": 2.36, + "learning_rate": 3.031738695015792e-05, + "loss": 0.0, + "step": 10220 + }, + { + "epoch": 2.36, + "learning_rate": 3.0315461058470072e-05, + "loss": 0.0, + "step": 10221 + }, + { + "epoch": 2.36, + "learning_rate": 3.031353516678222e-05, + "loss": 0.0, + "step": 10222 + }, + { + "epoch": 2.36, + "learning_rate": 3.031160927509437e-05, + "loss": 0.0, + "step": 10223 + }, + { + "epoch": 2.36, + "learning_rate": 3.0309683383406518e-05, + "loss": 0.0007, + "step": 10224 + }, + { + "epoch": 2.36, + "learning_rate": 3.030775749171867e-05, + "loss": 0.0001, + "step": 10225 + }, + { + "epoch": 2.36, + "learning_rate": 3.0305831600030817e-05, + "loss": 0.0, + "step": 10226 + }, + { + "epoch": 2.36, + "learning_rate": 3.0303905708342967e-05, + "loss": 0.0, + "step": 10227 + }, + { + "epoch": 2.36, + "learning_rate": 3.030197981665511e-05, + "loss": 0.0001, + "step": 10228 + }, + { + "epoch": 2.36, + "learning_rate": 3.030005392496726e-05, + "loss": 0.0, + "step": 10229 + }, + { + "epoch": 2.36, + "learning_rate": 3.029812803327941e-05, + "loss": 0.0, + "step": 10230 + }, + { + "epoch": 2.36, + "learning_rate": 3.0296202141591557e-05, + "loss": 0.0, + "step": 10231 + }, + { + "epoch": 2.36, + "learning_rate": 3.0294276249903708e-05, + "loss": 0.0, + "step": 10232 + }, + { + "epoch": 2.36, + "learning_rate": 3.0292350358215856e-05, + "loss": 0.0, + "step": 10233 + }, + { + "epoch": 2.37, + "learning_rate": 3.0290424466528007e-05, + "loss": 0.0, + "step": 10234 + }, + { + "epoch": 2.37, + "learning_rate": 3.028849857484015e-05, + "loss": 0.0, + "step": 10235 + }, + { + "epoch": 2.37, + "learning_rate": 3.0286572683152305e-05, + "loss": 0.0, + "step": 10236 + }, + { + "epoch": 2.37, + "learning_rate": 3.028464679146445e-05, + "loss": 0.0, + "step": 10237 + }, + { + "epoch": 2.37, + "learning_rate": 3.0282720899776597e-05, + "loss": 0.0, + "step": 10238 + }, + { + "epoch": 2.37, + "learning_rate": 3.0280795008088747e-05, + "loss": 0.0, + "step": 10239 + }, + { + "epoch": 2.37, + "learning_rate": 3.027886911640089e-05, + "loss": 0.0004, + "step": 10240 + }, + { + "epoch": 2.37, + "learning_rate": 3.0276943224713046e-05, + "loss": 0.0, + "step": 10241 + }, + { + "epoch": 2.37, + "learning_rate": 3.027501733302519e-05, + "loss": 0.0, + "step": 10242 + }, + { + "epoch": 2.37, + "learning_rate": 3.027309144133734e-05, + "loss": 0.0001, + "step": 10243 + }, + { + "epoch": 2.37, + "learning_rate": 3.0271165549649488e-05, + "loss": 0.0002, + "step": 10244 + }, + { + "epoch": 2.37, + "learning_rate": 3.026923965796164e-05, + "loss": 0.0001, + "step": 10245 + }, + { + "epoch": 2.37, + "learning_rate": 3.0267313766273787e-05, + "loss": 0.0003, + "step": 10246 + }, + { + "epoch": 2.37, + "learning_rate": 3.026538787458593e-05, + "loss": 0.0, + "step": 10247 + }, + { + "epoch": 2.37, + "learning_rate": 3.0263461982898085e-05, + "loss": 0.0, + "step": 10248 + }, + { + "epoch": 2.37, + "learning_rate": 3.026153609121023e-05, + "loss": 0.0002, + "step": 10249 + }, + { + "epoch": 2.37, + "learning_rate": 3.025961019952238e-05, + "loss": 0.0, + "step": 10250 + }, + { + "epoch": 2.37, + "learning_rate": 3.0257684307834527e-05, + "loss": 0.0, + "step": 10251 + }, + { + "epoch": 2.37, + "learning_rate": 3.0255758416146678e-05, + "loss": 0.0001, + "step": 10252 + }, + { + "epoch": 2.37, + "learning_rate": 3.0253832524458826e-05, + "loss": 0.0, + "step": 10253 + }, + { + "epoch": 2.37, + "learning_rate": 3.0251906632770976e-05, + "loss": 0.0, + "step": 10254 + }, + { + "epoch": 2.37, + "learning_rate": 3.0249980741083124e-05, + "loss": 0.0001, + "step": 10255 + }, + { + "epoch": 2.37, + "learning_rate": 3.0248054849395275e-05, + "loss": 0.0, + "step": 10256 + }, + { + "epoch": 2.37, + "learning_rate": 3.024612895770742e-05, + "loss": 0.0, + "step": 10257 + }, + { + "epoch": 2.37, + "learning_rate": 3.0244203066019566e-05, + "loss": 0.0003, + "step": 10258 + }, + { + "epoch": 2.37, + "learning_rate": 3.0242277174331717e-05, + "loss": 0.0, + "step": 10259 + }, + { + "epoch": 2.37, + "learning_rate": 3.0240351282643865e-05, + "loss": 0.0001, + "step": 10260 + }, + { + "epoch": 2.37, + "learning_rate": 3.0238425390956016e-05, + "loss": 0.0001, + "step": 10261 + }, + { + "epoch": 2.37, + "learning_rate": 3.0236499499268163e-05, + "loss": 0.0001, + "step": 10262 + }, + { + "epoch": 2.37, + "learning_rate": 3.0234573607580314e-05, + "loss": 0.0002, + "step": 10263 + }, + { + "epoch": 2.37, + "learning_rate": 3.0232647715892458e-05, + "loss": 0.0, + "step": 10264 + }, + { + "epoch": 2.37, + "learning_rate": 3.0230721824204612e-05, + "loss": 0.0, + "step": 10265 + }, + { + "epoch": 2.37, + "learning_rate": 3.0228795932516756e-05, + "loss": 0.0, + "step": 10266 + }, + { + "epoch": 2.37, + "learning_rate": 3.0226870040828904e-05, + "loss": 0.0, + "step": 10267 + }, + { + "epoch": 2.37, + "learning_rate": 3.0224944149141055e-05, + "loss": 0.0, + "step": 10268 + }, + { + "epoch": 2.37, + "learning_rate": 3.0223018257453202e-05, + "loss": 0.0114, + "step": 10269 + }, + { + "epoch": 2.37, + "learning_rate": 3.0221092365765353e-05, + "loss": 0.0002, + "step": 10270 + }, + { + "epoch": 2.37, + "learning_rate": 3.0219166474077497e-05, + "loss": 0.0, + "step": 10271 + }, + { + "epoch": 2.37, + "learning_rate": 3.021724058238965e-05, + "loss": 0.0002, + "step": 10272 + }, + { + "epoch": 2.37, + "learning_rate": 3.0215314690701796e-05, + "loss": 0.0003, + "step": 10273 + }, + { + "epoch": 2.37, + "learning_rate": 3.0213388799013946e-05, + "loss": 0.0001, + "step": 10274 + }, + { + "epoch": 2.37, + "learning_rate": 3.0211462907326094e-05, + "loss": 0.0, + "step": 10275 + }, + { + "epoch": 2.37, + "learning_rate": 3.0209537015638238e-05, + "loss": 0.0001, + "step": 10276 + }, + { + "epoch": 2.38, + "learning_rate": 3.0207611123950392e-05, + "loss": 0.0, + "step": 10277 + }, + { + "epoch": 2.38, + "learning_rate": 3.0205685232262536e-05, + "loss": 0.0003, + "step": 10278 + }, + { + "epoch": 2.38, + "learning_rate": 3.0203759340574687e-05, + "loss": 0.0, + "step": 10279 + }, + { + "epoch": 2.38, + "learning_rate": 3.0201833448886835e-05, + "loss": 0.0, + "step": 10280 + }, + { + "epoch": 2.38, + "learning_rate": 3.0199907557198986e-05, + "loss": 0.0, + "step": 10281 + }, + { + "epoch": 2.38, + "learning_rate": 3.0197981665511133e-05, + "loss": 0.0, + "step": 10282 + }, + { + "epoch": 2.38, + "learning_rate": 3.0196055773823284e-05, + "loss": 0.0002, + "step": 10283 + }, + { + "epoch": 2.38, + "learning_rate": 3.019412988213543e-05, + "loss": 0.0005, + "step": 10284 + }, + { + "epoch": 2.38, + "learning_rate": 3.0192203990447575e-05, + "loss": 0.0, + "step": 10285 + }, + { + "epoch": 2.38, + "learning_rate": 3.0190278098759726e-05, + "loss": 0.0, + "step": 10286 + }, + { + "epoch": 2.38, + "learning_rate": 3.0188352207071874e-05, + "loss": 0.0, + "step": 10287 + }, + { + "epoch": 2.38, + "learning_rate": 3.0186426315384025e-05, + "loss": 0.0001, + "step": 10288 + }, + { + "epoch": 2.38, + "learning_rate": 3.0184500423696172e-05, + "loss": 0.0, + "step": 10289 + }, + { + "epoch": 2.38, + "learning_rate": 3.0182574532008323e-05, + "loss": 0.0, + "step": 10290 + }, + { + "epoch": 2.38, + "learning_rate": 3.018064864032047e-05, + "loss": 0.0008, + "step": 10291 + }, + { + "epoch": 2.38, + "learning_rate": 3.017872274863262e-05, + "loss": 0.0, + "step": 10292 + }, + { + "epoch": 2.38, + "learning_rate": 3.0176796856944765e-05, + "loss": 0.0021, + "step": 10293 + }, + { + "epoch": 2.38, + "learning_rate": 3.0174870965256913e-05, + "loss": 0.0, + "step": 10294 + }, + { + "epoch": 2.38, + "learning_rate": 3.0172945073569064e-05, + "loss": 0.0005, + "step": 10295 + }, + { + "epoch": 2.38, + "learning_rate": 3.017101918188121e-05, + "loss": 0.0005, + "step": 10296 + }, + { + "epoch": 2.38, + "learning_rate": 3.0169093290193362e-05, + "loss": 0.0, + "step": 10297 + }, + { + "epoch": 2.38, + "learning_rate": 3.016716739850551e-05, + "loss": 0.0, + "step": 10298 + }, + { + "epoch": 2.38, + "learning_rate": 3.016524150681766e-05, + "loss": 0.0003, + "step": 10299 + }, + { + "epoch": 2.38, + "learning_rate": 3.0163315615129805e-05, + "loss": 0.0, + "step": 10300 + }, + { + "epoch": 2.38, + "learning_rate": 3.016138972344196e-05, + "loss": 0.0, + "step": 10301 + }, + { + "epoch": 2.38, + "learning_rate": 3.0159463831754103e-05, + "loss": 0.0, + "step": 10302 + }, + { + "epoch": 2.38, + "learning_rate": 3.015753794006625e-05, + "loss": 0.0002, + "step": 10303 + }, + { + "epoch": 2.38, + "learning_rate": 3.01556120483784e-05, + "loss": 0.0001, + "step": 10304 + }, + { + "epoch": 2.38, + "learning_rate": 3.015368615669055e-05, + "loss": 0.0, + "step": 10305 + }, + { + "epoch": 2.38, + "learning_rate": 3.01517602650027e-05, + "loss": 0.0, + "step": 10306 + }, + { + "epoch": 2.38, + "learning_rate": 3.0149834373314844e-05, + "loss": 0.0, + "step": 10307 + }, + { + "epoch": 2.38, + "learning_rate": 3.0147908481626998e-05, + "loss": 0.0, + "step": 10308 + }, + { + "epoch": 2.38, + "learning_rate": 3.0145982589939142e-05, + "loss": 0.0, + "step": 10309 + }, + { + "epoch": 2.38, + "learning_rate": 3.0144056698251293e-05, + "loss": 0.0, + "step": 10310 + }, + { + "epoch": 2.38, + "learning_rate": 3.014213080656344e-05, + "loss": 0.0001, + "step": 10311 + }, + { + "epoch": 2.38, + "learning_rate": 3.014020491487559e-05, + "loss": 0.0, + "step": 10312 + }, + { + "epoch": 2.38, + "learning_rate": 3.013827902318774e-05, + "loss": 0.0001, + "step": 10313 + }, + { + "epoch": 2.38, + "learning_rate": 3.0136353131499883e-05, + "loss": 0.0, + "step": 10314 + }, + { + "epoch": 2.38, + "learning_rate": 3.0134427239812034e-05, + "loss": 0.0001, + "step": 10315 + }, + { + "epoch": 2.38, + "learning_rate": 3.013250134812418e-05, + "loss": 0.0001, + "step": 10316 + }, + { + "epoch": 2.38, + "learning_rate": 3.0130575456436332e-05, + "loss": 0.0001, + "step": 10317 + }, + { + "epoch": 2.38, + "learning_rate": 3.012864956474848e-05, + "loss": 0.0001, + "step": 10318 + }, + { + "epoch": 2.38, + "learning_rate": 3.012672367306063e-05, + "loss": 0.0, + "step": 10319 + }, + { + "epoch": 2.39, + "learning_rate": 3.0124797781372778e-05, + "loss": 0.0, + "step": 10320 + }, + { + "epoch": 2.39, + "learning_rate": 3.012287188968493e-05, + "loss": 0.0001, + "step": 10321 + }, + { + "epoch": 2.39, + "learning_rate": 3.0120945997997073e-05, + "loss": 0.0001, + "step": 10322 + }, + { + "epoch": 2.39, + "learning_rate": 3.011902010630922e-05, + "loss": 0.0, + "step": 10323 + }, + { + "epoch": 2.39, + "learning_rate": 3.011709421462137e-05, + "loss": 0.0, + "step": 10324 + }, + { + "epoch": 2.39, + "learning_rate": 3.011516832293352e-05, + "loss": 0.0, + "step": 10325 + }, + { + "epoch": 2.39, + "learning_rate": 3.011324243124567e-05, + "loss": 0.0, + "step": 10326 + }, + { + "epoch": 2.39, + "learning_rate": 3.0111316539557817e-05, + "loss": 0.0, + "step": 10327 + }, + { + "epoch": 2.39, + "learning_rate": 3.0109390647869968e-05, + "loss": 0.0, + "step": 10328 + }, + { + "epoch": 2.39, + "learning_rate": 3.0107464756182112e-05, + "loss": 0.0, + "step": 10329 + }, + { + "epoch": 2.39, + "learning_rate": 3.0105538864494266e-05, + "loss": 0.0, + "step": 10330 + }, + { + "epoch": 2.39, + "learning_rate": 3.010361297280641e-05, + "loss": 0.0, + "step": 10331 + }, + { + "epoch": 2.39, + "learning_rate": 3.0101687081118558e-05, + "loss": 0.0, + "step": 10332 + }, + { + "epoch": 2.39, + "learning_rate": 3.009976118943071e-05, + "loss": 0.0, + "step": 10333 + }, + { + "epoch": 2.39, + "learning_rate": 3.0097835297742856e-05, + "loss": 0.0, + "step": 10334 + }, + { + "epoch": 2.39, + "learning_rate": 3.0095909406055007e-05, + "loss": 0.0001, + "step": 10335 + }, + { + "epoch": 2.39, + "learning_rate": 3.009398351436715e-05, + "loss": 0.0, + "step": 10336 + }, + { + "epoch": 2.39, + "learning_rate": 3.0092057622679305e-05, + "loss": 0.0001, + "step": 10337 + }, + { + "epoch": 2.39, + "learning_rate": 3.009013173099145e-05, + "loss": 0.0, + "step": 10338 + }, + { + "epoch": 2.39, + "learning_rate": 3.00882058393036e-05, + "loss": 0.0, + "step": 10339 + }, + { + "epoch": 2.39, + "learning_rate": 3.0086279947615748e-05, + "loss": 0.0, + "step": 10340 + }, + { + "epoch": 2.39, + "learning_rate": 3.0084354055927895e-05, + "loss": 0.0001, + "step": 10341 + }, + { + "epoch": 2.39, + "learning_rate": 3.0082428164240046e-05, + "loss": 0.0, + "step": 10342 + }, + { + "epoch": 2.39, + "learning_rate": 3.008050227255219e-05, + "loss": 0.0, + "step": 10343 + }, + { + "epoch": 2.39, + "learning_rate": 3.0078576380864344e-05, + "loss": 0.0, + "step": 10344 + }, + { + "epoch": 2.39, + "learning_rate": 3.007665048917649e-05, + "loss": 0.0, + "step": 10345 + }, + { + "epoch": 2.39, + "learning_rate": 3.007472459748864e-05, + "loss": 0.0, + "step": 10346 + }, + { + "epoch": 2.39, + "learning_rate": 3.0072798705800787e-05, + "loss": 0.0, + "step": 10347 + }, + { + "epoch": 2.39, + "learning_rate": 3.0070872814112938e-05, + "loss": 0.0, + "step": 10348 + }, + { + "epoch": 2.39, + "learning_rate": 3.0068946922425085e-05, + "loss": 0.0001, + "step": 10349 + }, + { + "epoch": 2.39, + "learning_rate": 3.006702103073723e-05, + "loss": 0.0, + "step": 10350 + }, + { + "epoch": 2.39, + "learning_rate": 3.0065095139049383e-05, + "loss": 0.0006, + "step": 10351 + }, + { + "epoch": 2.39, + "learning_rate": 3.0063169247361528e-05, + "loss": 0.0, + "step": 10352 + }, + { + "epoch": 2.39, + "learning_rate": 3.006124335567368e-05, + "loss": 0.0001, + "step": 10353 + }, + { + "epoch": 2.39, + "learning_rate": 3.0059317463985826e-05, + "loss": 0.0, + "step": 10354 + }, + { + "epoch": 2.39, + "learning_rate": 3.0057391572297977e-05, + "loss": 0.0, + "step": 10355 + }, + { + "epoch": 2.39, + "learning_rate": 3.0055465680610124e-05, + "loss": 0.0, + "step": 10356 + }, + { + "epoch": 2.39, + "learning_rate": 3.0053539788922275e-05, + "loss": 0.0131, + "step": 10357 + }, + { + "epoch": 2.39, + "learning_rate": 3.005161389723442e-05, + "loss": 0.0, + "step": 10358 + }, + { + "epoch": 2.39, + "learning_rate": 3.0049688005546567e-05, + "loss": 0.0, + "step": 10359 + }, + { + "epoch": 2.39, + "learning_rate": 3.0047762113858718e-05, + "loss": 0.0, + "step": 10360 + }, + { + "epoch": 2.39, + "learning_rate": 3.0045836222170865e-05, + "loss": 0.0, + "step": 10361 + }, + { + "epoch": 2.39, + "learning_rate": 3.0043910330483016e-05, + "loss": 0.0001, + "step": 10362 + }, + { + "epoch": 2.39, + "learning_rate": 3.0041984438795163e-05, + "loss": 0.0, + "step": 10363 + }, + { + "epoch": 2.4, + "learning_rate": 3.0040058547107314e-05, + "loss": 0.0, + "step": 10364 + }, + { + "epoch": 2.4, + "learning_rate": 3.003813265541946e-05, + "loss": 0.0, + "step": 10365 + }, + { + "epoch": 2.4, + "learning_rate": 3.0036206763731613e-05, + "loss": 0.0012, + "step": 10366 + }, + { + "epoch": 2.4, + "learning_rate": 3.0034280872043757e-05, + "loss": 0.0001, + "step": 10367 + }, + { + "epoch": 2.4, + "learning_rate": 3.0032354980355908e-05, + "loss": 0.0, + "step": 10368 + }, + { + "epoch": 2.4, + "learning_rate": 3.0030429088668055e-05, + "loss": 0.0, + "step": 10369 + }, + { + "epoch": 2.4, + "learning_rate": 3.0028503196980203e-05, + "loss": 0.0, + "step": 10370 + }, + { + "epoch": 2.4, + "learning_rate": 3.0026577305292353e-05, + "loss": 0.0, + "step": 10371 + }, + { + "epoch": 2.4, + "learning_rate": 3.0024651413604497e-05, + "loss": 0.0, + "step": 10372 + }, + { + "epoch": 2.4, + "learning_rate": 3.0022725521916652e-05, + "loss": 0.0, + "step": 10373 + }, + { + "epoch": 2.4, + "learning_rate": 3.0020799630228796e-05, + "loss": 0.0001, + "step": 10374 + }, + { + "epoch": 2.4, + "learning_rate": 3.0018873738540947e-05, + "loss": 0.0, + "step": 10375 + }, + { + "epoch": 2.4, + "learning_rate": 3.0016947846853094e-05, + "loss": 0.0, + "step": 10376 + }, + { + "epoch": 2.4, + "learning_rate": 3.0015021955165245e-05, + "loss": 0.0001, + "step": 10377 + }, + { + "epoch": 2.4, + "learning_rate": 3.0013096063477392e-05, + "loss": 0.0001, + "step": 10378 + }, + { + "epoch": 2.4, + "learning_rate": 3.0011170171789537e-05, + "loss": 0.0001, + "step": 10379 + }, + { + "epoch": 2.4, + "learning_rate": 3.000924428010169e-05, + "loss": 0.0, + "step": 10380 + }, + { + "epoch": 2.4, + "learning_rate": 3.0007318388413835e-05, + "loss": 0.0, + "step": 10381 + }, + { + "epoch": 2.4, + "learning_rate": 3.0005392496725986e-05, + "loss": 0.0, + "step": 10382 + }, + { + "epoch": 2.4, + "learning_rate": 3.0003466605038133e-05, + "loss": 0.0001, + "step": 10383 + }, + { + "epoch": 2.4, + "learning_rate": 3.0001540713350284e-05, + "loss": 0.0, + "step": 10384 + }, + { + "epoch": 2.4, + "learning_rate": 2.999961482166243e-05, + "loss": 0.0, + "step": 10385 + }, + { + "epoch": 2.4, + "learning_rate": 2.9997688929974582e-05, + "loss": 0.0, + "step": 10386 + }, + { + "epoch": 2.4, + "learning_rate": 2.999576303828673e-05, + "loss": 0.0, + "step": 10387 + }, + { + "epoch": 2.4, + "learning_rate": 2.9993837146598874e-05, + "loss": 0.0, + "step": 10388 + }, + { + "epoch": 2.4, + "learning_rate": 2.9991911254911025e-05, + "loss": 0.0, + "step": 10389 + }, + { + "epoch": 2.4, + "learning_rate": 2.9989985363223172e-05, + "loss": 0.0, + "step": 10390 + }, + { + "epoch": 2.4, + "learning_rate": 2.9988059471535323e-05, + "loss": 0.0, + "step": 10391 + }, + { + "epoch": 2.4, + "learning_rate": 2.998613357984747e-05, + "loss": 0.0004, + "step": 10392 + }, + { + "epoch": 2.4, + "learning_rate": 2.998420768815962e-05, + "loss": 0.0001, + "step": 10393 + }, + { + "epoch": 2.4, + "learning_rate": 2.9982281796471766e-05, + "loss": 0.0, + "step": 10394 + }, + { + "epoch": 2.4, + "learning_rate": 2.998035590478392e-05, + "loss": 0.0001, + "step": 10395 + }, + { + "epoch": 2.4, + "learning_rate": 2.9978430013096064e-05, + "loss": 0.0001, + "step": 10396 + }, + { + "epoch": 2.4, + "learning_rate": 2.997650412140821e-05, + "loss": 0.0, + "step": 10397 + }, + { + "epoch": 2.4, + "learning_rate": 2.9974578229720362e-05, + "loss": 0.0, + "step": 10398 + }, + { + "epoch": 2.4, + "learning_rate": 2.997265233803251e-05, + "loss": 0.0001, + "step": 10399 + }, + { + "epoch": 2.4, + "learning_rate": 2.997072644634466e-05, + "loss": 0.0, + "step": 10400 + }, + { + "epoch": 2.4, + "learning_rate": 2.9968800554656805e-05, + "loss": 0.0, + "step": 10401 + }, + { + "epoch": 2.4, + "learning_rate": 2.996687466296896e-05, + "loss": 0.0, + "step": 10402 + }, + { + "epoch": 2.4, + "learning_rate": 2.9964948771281103e-05, + "loss": 0.009, + "step": 10403 + }, + { + "epoch": 2.4, + "learning_rate": 2.9963022879593254e-05, + "loss": 0.0, + "step": 10404 + }, + { + "epoch": 2.4, + "learning_rate": 2.99610969879054e-05, + "loss": 0.0001, + "step": 10405 + }, + { + "epoch": 2.4, + "learning_rate": 2.995917109621755e-05, + "loss": 0.0001, + "step": 10406 + }, + { + "epoch": 2.41, + "learning_rate": 2.99572452045297e-05, + "loss": 0.0, + "step": 10407 + }, + { + "epoch": 2.41, + "learning_rate": 2.9955319312841844e-05, + "loss": 0.0, + "step": 10408 + }, + { + "epoch": 2.41, + "learning_rate": 2.9953393421153998e-05, + "loss": 0.0, + "step": 10409 + }, + { + "epoch": 2.41, + "learning_rate": 2.9951467529466142e-05, + "loss": 0.0, + "step": 10410 + }, + { + "epoch": 2.41, + "learning_rate": 2.9949541637778293e-05, + "loss": 0.0, + "step": 10411 + }, + { + "epoch": 2.41, + "learning_rate": 2.994761574609044e-05, + "loss": 0.0, + "step": 10412 + }, + { + "epoch": 2.41, + "learning_rate": 2.994568985440259e-05, + "loss": 0.0004, + "step": 10413 + }, + { + "epoch": 2.41, + "learning_rate": 2.994376396271474e-05, + "loss": 0.0, + "step": 10414 + }, + { + "epoch": 2.41, + "learning_rate": 2.9941838071026883e-05, + "loss": 0.0, + "step": 10415 + }, + { + "epoch": 2.41, + "learning_rate": 2.9939912179339037e-05, + "loss": 0.0, + "step": 10416 + }, + { + "epoch": 2.41, + "learning_rate": 2.993798628765118e-05, + "loss": 0.0, + "step": 10417 + }, + { + "epoch": 2.41, + "learning_rate": 2.9936060395963332e-05, + "loss": 0.0001, + "step": 10418 + }, + { + "epoch": 2.41, + "learning_rate": 2.993413450427548e-05, + "loss": 0.0, + "step": 10419 + }, + { + "epoch": 2.41, + "learning_rate": 2.993220861258763e-05, + "loss": 0.0001, + "step": 10420 + }, + { + "epoch": 2.41, + "learning_rate": 2.9930282720899778e-05, + "loss": 0.0001, + "step": 10421 + }, + { + "epoch": 2.41, + "learning_rate": 2.992835682921193e-05, + "loss": 0.0, + "step": 10422 + }, + { + "epoch": 2.41, + "learning_rate": 2.9926430937524076e-05, + "loss": 0.0004, + "step": 10423 + }, + { + "epoch": 2.41, + "learning_rate": 2.992450504583622e-05, + "loss": 0.0, + "step": 10424 + }, + { + "epoch": 2.41, + "learning_rate": 2.992257915414837e-05, + "loss": 0.0, + "step": 10425 + }, + { + "epoch": 2.41, + "learning_rate": 2.992065326246052e-05, + "loss": 0.0001, + "step": 10426 + }, + { + "epoch": 2.41, + "learning_rate": 2.991872737077267e-05, + "loss": 0.0, + "step": 10427 + }, + { + "epoch": 2.41, + "learning_rate": 2.9916801479084817e-05, + "loss": 0.0001, + "step": 10428 + }, + { + "epoch": 2.41, + "learning_rate": 2.9914875587396968e-05, + "loss": 0.0001, + "step": 10429 + }, + { + "epoch": 2.41, + "learning_rate": 2.9912949695709112e-05, + "loss": 0.0, + "step": 10430 + }, + { + "epoch": 2.41, + "learning_rate": 2.9911023804021266e-05, + "loss": 0.0, + "step": 10431 + }, + { + "epoch": 2.41, + "learning_rate": 2.990909791233341e-05, + "loss": 0.0001, + "step": 10432 + }, + { + "epoch": 2.41, + "learning_rate": 2.990717202064556e-05, + "loss": 0.0, + "step": 10433 + }, + { + "epoch": 2.41, + "learning_rate": 2.990524612895771e-05, + "loss": 0.0, + "step": 10434 + }, + { + "epoch": 2.41, + "learning_rate": 2.9903320237269856e-05, + "loss": 0.0, + "step": 10435 + }, + { + "epoch": 2.41, + "learning_rate": 2.9901394345582007e-05, + "loss": 0.0, + "step": 10436 + }, + { + "epoch": 2.41, + "learning_rate": 2.989946845389415e-05, + "loss": 0.0, + "step": 10437 + }, + { + "epoch": 2.41, + "learning_rate": 2.9897542562206306e-05, + "loss": 0.0, + "step": 10438 + }, + { + "epoch": 2.41, + "learning_rate": 2.989561667051845e-05, + "loss": 0.0, + "step": 10439 + }, + { + "epoch": 2.41, + "learning_rate": 2.98936907788306e-05, + "loss": 0.0, + "step": 10440 + }, + { + "epoch": 2.41, + "learning_rate": 2.9891764887142748e-05, + "loss": 0.0001, + "step": 10441 + }, + { + "epoch": 2.41, + "learning_rate": 2.98898389954549e-05, + "loss": 0.0, + "step": 10442 + }, + { + "epoch": 2.41, + "learning_rate": 2.9887913103767046e-05, + "loss": 0.0001, + "step": 10443 + }, + { + "epoch": 2.41, + "learning_rate": 2.988598721207919e-05, + "loss": 0.0001, + "step": 10444 + }, + { + "epoch": 2.41, + "learning_rate": 2.9884061320391345e-05, + "loss": 0.0, + "step": 10445 + }, + { + "epoch": 2.41, + "learning_rate": 2.988213542870349e-05, + "loss": 0.0001, + "step": 10446 + }, + { + "epoch": 2.41, + "learning_rate": 2.988020953701564e-05, + "loss": 0.0, + "step": 10447 + }, + { + "epoch": 2.41, + "learning_rate": 2.9878283645327787e-05, + "loss": 0.0, + "step": 10448 + }, + { + "epoch": 2.41, + "learning_rate": 2.9876357753639938e-05, + "loss": 0.0, + "step": 10449 + }, + { + "epoch": 2.42, + "learning_rate": 2.9874431861952085e-05, + "loss": 0.0, + "step": 10450 + }, + { + "epoch": 2.42, + "learning_rate": 2.9872505970264236e-05, + "loss": 0.0, + "step": 10451 + }, + { + "epoch": 2.42, + "learning_rate": 2.9870580078576384e-05, + "loss": 0.0, + "step": 10452 + }, + { + "epoch": 2.42, + "learning_rate": 2.9868654186888528e-05, + "loss": 0.0001, + "step": 10453 + }, + { + "epoch": 2.42, + "learning_rate": 2.986672829520068e-05, + "loss": 0.0, + "step": 10454 + }, + { + "epoch": 2.42, + "learning_rate": 2.9864802403512826e-05, + "loss": 0.0, + "step": 10455 + }, + { + "epoch": 2.42, + "learning_rate": 2.9862876511824977e-05, + "loss": 0.0001, + "step": 10456 + }, + { + "epoch": 2.42, + "learning_rate": 2.9860950620137125e-05, + "loss": 0.0, + "step": 10457 + }, + { + "epoch": 2.42, + "learning_rate": 2.9859024728449275e-05, + "loss": 0.0, + "step": 10458 + }, + { + "epoch": 2.42, + "learning_rate": 2.9857098836761423e-05, + "loss": 0.0, + "step": 10459 + }, + { + "epoch": 2.42, + "learning_rate": 2.9855172945073574e-05, + "loss": 0.0, + "step": 10460 + }, + { + "epoch": 2.42, + "learning_rate": 2.9853247053385718e-05, + "loss": 0.0, + "step": 10461 + }, + { + "epoch": 2.42, + "learning_rate": 2.9851321161697865e-05, + "loss": 0.0, + "step": 10462 + }, + { + "epoch": 2.42, + "learning_rate": 2.9849395270010016e-05, + "loss": 0.0, + "step": 10463 + }, + { + "epoch": 2.42, + "learning_rate": 2.9847469378322164e-05, + "loss": 0.0, + "step": 10464 + }, + { + "epoch": 2.42, + "learning_rate": 2.9845543486634315e-05, + "loss": 0.0, + "step": 10465 + }, + { + "epoch": 2.42, + "learning_rate": 2.984361759494646e-05, + "loss": 0.0, + "step": 10466 + }, + { + "epoch": 2.42, + "learning_rate": 2.9841691703258613e-05, + "loss": 0.0001, + "step": 10467 + }, + { + "epoch": 2.42, + "learning_rate": 2.9839765811570757e-05, + "loss": 0.0001, + "step": 10468 + }, + { + "epoch": 2.42, + "learning_rate": 2.9837839919882908e-05, + "loss": 0.0001, + "step": 10469 + }, + { + "epoch": 2.42, + "learning_rate": 2.9835914028195055e-05, + "loss": 0.0, + "step": 10470 + }, + { + "epoch": 2.42, + "learning_rate": 2.9833988136507203e-05, + "loss": 0.0, + "step": 10471 + }, + { + "epoch": 2.42, + "learning_rate": 2.9832062244819354e-05, + "loss": 0.0, + "step": 10472 + }, + { + "epoch": 2.42, + "learning_rate": 2.9830136353131498e-05, + "loss": 0.0, + "step": 10473 + }, + { + "epoch": 2.42, + "learning_rate": 2.9828210461443652e-05, + "loss": 0.0, + "step": 10474 + }, + { + "epoch": 2.42, + "learning_rate": 2.9826284569755796e-05, + "loss": 0.0, + "step": 10475 + }, + { + "epoch": 2.42, + "learning_rate": 2.9824358678067947e-05, + "loss": 0.0001, + "step": 10476 + }, + { + "epoch": 2.42, + "learning_rate": 2.9822432786380094e-05, + "loss": 0.0, + "step": 10477 + }, + { + "epoch": 2.42, + "learning_rate": 2.9820506894692245e-05, + "loss": 0.0, + "step": 10478 + }, + { + "epoch": 2.42, + "learning_rate": 2.9818581003004393e-05, + "loss": 0.0003, + "step": 10479 + }, + { + "epoch": 2.42, + "learning_rate": 2.9816655111316537e-05, + "loss": 0.0001, + "step": 10480 + }, + { + "epoch": 2.42, + "learning_rate": 2.981472921962869e-05, + "loss": 0.0, + "step": 10481 + }, + { + "epoch": 2.42, + "learning_rate": 2.9812803327940835e-05, + "loss": 0.0, + "step": 10482 + }, + { + "epoch": 2.42, + "learning_rate": 2.9810877436252986e-05, + "loss": 0.0, + "step": 10483 + }, + { + "epoch": 2.42, + "learning_rate": 2.9808951544565134e-05, + "loss": 0.0, + "step": 10484 + }, + { + "epoch": 2.42, + "learning_rate": 2.9807025652877284e-05, + "loss": 0.0, + "step": 10485 + }, + { + "epoch": 2.42, + "learning_rate": 2.9805099761189432e-05, + "loss": 0.0, + "step": 10486 + }, + { + "epoch": 2.42, + "learning_rate": 2.9803173869501583e-05, + "loss": 0.0, + "step": 10487 + }, + { + "epoch": 2.42, + "learning_rate": 2.980124797781373e-05, + "loss": 0.0, + "step": 10488 + }, + { + "epoch": 2.42, + "learning_rate": 2.979932208612588e-05, + "loss": 0.0, + "step": 10489 + }, + { + "epoch": 2.42, + "learning_rate": 2.9797396194438025e-05, + "loss": 0.0001, + "step": 10490 + }, + { + "epoch": 2.42, + "learning_rate": 2.9795470302750173e-05, + "loss": 0.0007, + "step": 10491 + }, + { + "epoch": 2.42, + "learning_rate": 2.9793544411062324e-05, + "loss": 0.0, + "step": 10492 + }, + { + "epoch": 2.43, + "learning_rate": 2.979161851937447e-05, + "loss": 0.0001, + "step": 10493 + }, + { + "epoch": 2.43, + "learning_rate": 2.9789692627686622e-05, + "loss": 0.0, + "step": 10494 + }, + { + "epoch": 2.43, + "learning_rate": 2.978776673599877e-05, + "loss": 0.0, + "step": 10495 + }, + { + "epoch": 2.43, + "learning_rate": 2.978584084431092e-05, + "loss": 0.0001, + "step": 10496 + }, + { + "epoch": 2.43, + "learning_rate": 2.9783914952623064e-05, + "loss": 0.0004, + "step": 10497 + }, + { + "epoch": 2.43, + "learning_rate": 2.978198906093522e-05, + "loss": 0.0, + "step": 10498 + }, + { + "epoch": 2.43, + "learning_rate": 2.9780063169247363e-05, + "loss": 0.0, + "step": 10499 + }, + { + "epoch": 2.43, + "learning_rate": 2.977813727755951e-05, + "loss": 0.0, + "step": 10500 + }, + { + "epoch": 2.43, + "learning_rate": 2.977621138587166e-05, + "loss": 0.0, + "step": 10501 + }, + { + "epoch": 2.43, + "learning_rate": 2.9774285494183805e-05, + "loss": 0.0065, + "step": 10502 + }, + { + "epoch": 2.43, + "learning_rate": 2.977235960249596e-05, + "loss": 0.0004, + "step": 10503 + }, + { + "epoch": 2.43, + "learning_rate": 2.9770433710808103e-05, + "loss": 0.0002, + "step": 10504 + }, + { + "epoch": 2.43, + "learning_rate": 2.9768507819120254e-05, + "loss": 0.0, + "step": 10505 + }, + { + "epoch": 2.43, + "learning_rate": 2.9766581927432402e-05, + "loss": 0.0013, + "step": 10506 + }, + { + "epoch": 2.43, + "learning_rate": 2.9764656035744553e-05, + "loss": 0.0002, + "step": 10507 + }, + { + "epoch": 2.43, + "learning_rate": 2.97627301440567e-05, + "loss": 0.0, + "step": 10508 + }, + { + "epoch": 2.43, + "learning_rate": 2.9760804252368844e-05, + "loss": 0.0, + "step": 10509 + }, + { + "epoch": 2.43, + "learning_rate": 2.9758878360681e-05, + "loss": 0.0, + "step": 10510 + }, + { + "epoch": 2.43, + "learning_rate": 2.9756952468993143e-05, + "loss": 0.0001, + "step": 10511 + }, + { + "epoch": 2.43, + "learning_rate": 2.9755026577305293e-05, + "loss": 0.0, + "step": 10512 + }, + { + "epoch": 2.43, + "learning_rate": 2.975310068561744e-05, + "loss": 0.0004, + "step": 10513 + }, + { + "epoch": 2.43, + "learning_rate": 2.9751174793929592e-05, + "loss": 0.0, + "step": 10514 + }, + { + "epoch": 2.43, + "learning_rate": 2.974924890224174e-05, + "loss": 0.0, + "step": 10515 + }, + { + "epoch": 2.43, + "learning_rate": 2.974732301055389e-05, + "loss": 0.0, + "step": 10516 + }, + { + "epoch": 2.43, + "learning_rate": 2.9745397118866038e-05, + "loss": 0.0, + "step": 10517 + }, + { + "epoch": 2.43, + "learning_rate": 2.974347122717818e-05, + "loss": 0.0, + "step": 10518 + }, + { + "epoch": 2.43, + "learning_rate": 2.9741545335490333e-05, + "loss": 0.0, + "step": 10519 + }, + { + "epoch": 2.43, + "learning_rate": 2.973961944380248e-05, + "loss": 0.0, + "step": 10520 + }, + { + "epoch": 2.43, + "learning_rate": 2.973769355211463e-05, + "loss": 0.0, + "step": 10521 + }, + { + "epoch": 2.43, + "learning_rate": 2.973576766042678e-05, + "loss": 0.0, + "step": 10522 + }, + { + "epoch": 2.43, + "learning_rate": 2.973384176873893e-05, + "loss": 0.0, + "step": 10523 + }, + { + "epoch": 2.43, + "learning_rate": 2.9731915877051077e-05, + "loss": 0.0, + "step": 10524 + }, + { + "epoch": 2.43, + "learning_rate": 2.9729989985363228e-05, + "loss": 0.0003, + "step": 10525 + }, + { + "epoch": 2.43, + "learning_rate": 2.972806409367537e-05, + "loss": 0.0, + "step": 10526 + }, + { + "epoch": 2.43, + "learning_rate": 2.972613820198752e-05, + "loss": 0.0, + "step": 10527 + }, + { + "epoch": 2.43, + "learning_rate": 2.972421231029967e-05, + "loss": 0.0, + "step": 10528 + }, + { + "epoch": 2.43, + "learning_rate": 2.9722286418611817e-05, + "loss": 0.0012, + "step": 10529 + }, + { + "epoch": 2.43, + "learning_rate": 2.972036052692397e-05, + "loss": 0.0, + "step": 10530 + }, + { + "epoch": 2.43, + "learning_rate": 2.9718434635236116e-05, + "loss": 0.0, + "step": 10531 + }, + { + "epoch": 2.43, + "learning_rate": 2.9716508743548267e-05, + "loss": 0.0, + "step": 10532 + }, + { + "epoch": 2.43, + "learning_rate": 2.971458285186041e-05, + "loss": 0.0, + "step": 10533 + }, + { + "epoch": 2.43, + "learning_rate": 2.9712656960172565e-05, + "loss": 0.0, + "step": 10534 + }, + { + "epoch": 2.43, + "learning_rate": 2.971073106848471e-05, + "loss": 0.0, + "step": 10535 + }, + { + "epoch": 2.43, + "learning_rate": 2.9708805176796857e-05, + "loss": 0.0, + "step": 10536 + }, + { + "epoch": 2.44, + "learning_rate": 2.9706879285109007e-05, + "loss": 0.0, + "step": 10537 + }, + { + "epoch": 2.44, + "learning_rate": 2.970495339342115e-05, + "loss": 0.0, + "step": 10538 + }, + { + "epoch": 2.44, + "learning_rate": 2.9703027501733306e-05, + "loss": 0.0001, + "step": 10539 + }, + { + "epoch": 2.44, + "learning_rate": 2.970110161004545e-05, + "loss": 0.0, + "step": 10540 + }, + { + "epoch": 2.44, + "learning_rate": 2.9699175718357604e-05, + "loss": 0.0, + "step": 10541 + }, + { + "epoch": 2.44, + "learning_rate": 2.9697249826669748e-05, + "loss": 0.0002, + "step": 10542 + }, + { + "epoch": 2.44, + "learning_rate": 2.96953239349819e-05, + "loss": 0.0001, + "step": 10543 + }, + { + "epoch": 2.44, + "learning_rate": 2.9693398043294047e-05, + "loss": 0.0001, + "step": 10544 + }, + { + "epoch": 2.44, + "learning_rate": 2.9691472151606197e-05, + "loss": 0.0, + "step": 10545 + }, + { + "epoch": 2.44, + "learning_rate": 2.9689546259918345e-05, + "loss": 0.0, + "step": 10546 + }, + { + "epoch": 2.44, + "learning_rate": 2.968762036823049e-05, + "loss": 0.0, + "step": 10547 + }, + { + "epoch": 2.44, + "learning_rate": 2.968569447654264e-05, + "loss": 0.0, + "step": 10548 + }, + { + "epoch": 2.44, + "learning_rate": 2.9683768584854787e-05, + "loss": 0.0, + "step": 10549 + }, + { + "epoch": 2.44, + "learning_rate": 2.9681842693166938e-05, + "loss": 0.0001, + "step": 10550 + }, + { + "epoch": 2.44, + "learning_rate": 2.9679916801479086e-05, + "loss": 0.0, + "step": 10551 + }, + { + "epoch": 2.44, + "learning_rate": 2.9677990909791237e-05, + "loss": 0.0001, + "step": 10552 + }, + { + "epoch": 2.44, + "learning_rate": 2.9676065018103384e-05, + "loss": 0.0003, + "step": 10553 + }, + { + "epoch": 2.44, + "learning_rate": 2.9674139126415535e-05, + "loss": 0.0003, + "step": 10554 + }, + { + "epoch": 2.44, + "learning_rate": 2.967221323472768e-05, + "loss": 0.0, + "step": 10555 + }, + { + "epoch": 2.44, + "learning_rate": 2.9670287343039826e-05, + "loss": 0.0001, + "step": 10556 + }, + { + "epoch": 2.44, + "learning_rate": 2.9668361451351977e-05, + "loss": 0.0008, + "step": 10557 + }, + { + "epoch": 2.44, + "learning_rate": 2.9666435559664125e-05, + "loss": 0.0004, + "step": 10558 + }, + { + "epoch": 2.44, + "learning_rate": 2.9664509667976276e-05, + "loss": 0.0, + "step": 10559 + }, + { + "epoch": 2.44, + "learning_rate": 2.9662583776288423e-05, + "loss": 0.0, + "step": 10560 + }, + { + "epoch": 2.44, + "learning_rate": 2.9660657884600574e-05, + "loss": 0.0, + "step": 10561 + }, + { + "epoch": 2.44, + "learning_rate": 2.9658731992912718e-05, + "loss": 0.0, + "step": 10562 + }, + { + "epoch": 2.44, + "learning_rate": 2.9656806101224872e-05, + "loss": 0.0, + "step": 10563 + }, + { + "epoch": 2.44, + "learning_rate": 2.9654880209537016e-05, + "loss": 0.0, + "step": 10564 + }, + { + "epoch": 2.44, + "learning_rate": 2.9652954317849164e-05, + "loss": 0.0, + "step": 10565 + }, + { + "epoch": 2.44, + "learning_rate": 2.9651028426161315e-05, + "loss": 0.0, + "step": 10566 + }, + { + "epoch": 2.44, + "learning_rate": 2.9649102534473462e-05, + "loss": 0.0, + "step": 10567 + }, + { + "epoch": 2.44, + "learning_rate": 2.9647176642785613e-05, + "loss": 0.0, + "step": 10568 + }, + { + "epoch": 2.44, + "learning_rate": 2.9645250751097757e-05, + "loss": 0.0, + "step": 10569 + }, + { + "epoch": 2.44, + "learning_rate": 2.964332485940991e-05, + "loss": 0.0003, + "step": 10570 + }, + { + "epoch": 2.44, + "learning_rate": 2.9641398967722056e-05, + "loss": 0.0, + "step": 10571 + }, + { + "epoch": 2.44, + "learning_rate": 2.9639473076034206e-05, + "loss": 0.0001, + "step": 10572 + }, + { + "epoch": 2.44, + "learning_rate": 2.9637547184346354e-05, + "loss": 0.0, + "step": 10573 + }, + { + "epoch": 2.44, + "learning_rate": 2.96356212926585e-05, + "loss": 0.0, + "step": 10574 + }, + { + "epoch": 2.44, + "learning_rate": 2.9633695400970652e-05, + "loss": 0.0111, + "step": 10575 + }, + { + "epoch": 2.44, + "learning_rate": 2.9631769509282796e-05, + "loss": 0.0, + "step": 10576 + }, + { + "epoch": 2.44, + "learning_rate": 2.962984361759495e-05, + "loss": 0.0005, + "step": 10577 + }, + { + "epoch": 2.44, + "learning_rate": 2.9627917725907095e-05, + "loss": 0.0001, + "step": 10578 + }, + { + "epoch": 2.44, + "learning_rate": 2.9625991834219246e-05, + "loss": 0.0, + "step": 10579 + }, + { + "epoch": 2.45, + "learning_rate": 2.9624065942531393e-05, + "loss": 0.0001, + "step": 10580 + }, + { + "epoch": 2.45, + "learning_rate": 2.9622140050843544e-05, + "loss": 0.0006, + "step": 10581 + }, + { + "epoch": 2.45, + "learning_rate": 2.962021415915569e-05, + "loss": 0.0, + "step": 10582 + }, + { + "epoch": 2.45, + "learning_rate": 2.9618288267467835e-05, + "loss": 0.0, + "step": 10583 + }, + { + "epoch": 2.45, + "learning_rate": 2.9616362375779986e-05, + "loss": 0.0, + "step": 10584 + }, + { + "epoch": 2.45, + "learning_rate": 2.9614436484092134e-05, + "loss": 0.0002, + "step": 10585 + }, + { + "epoch": 2.45, + "learning_rate": 2.9612510592404285e-05, + "loss": 0.0, + "step": 10586 + }, + { + "epoch": 2.45, + "learning_rate": 2.9610584700716432e-05, + "loss": 0.0002, + "step": 10587 + }, + { + "epoch": 2.45, + "learning_rate": 2.9608658809028583e-05, + "loss": 0.0001, + "step": 10588 + }, + { + "epoch": 2.45, + "learning_rate": 2.960673291734073e-05, + "loss": 0.0002, + "step": 10589 + }, + { + "epoch": 2.45, + "learning_rate": 2.960480702565288e-05, + "loss": 0.0, + "step": 10590 + }, + { + "epoch": 2.45, + "learning_rate": 2.9602881133965025e-05, + "loss": 0.0, + "step": 10591 + }, + { + "epoch": 2.45, + "learning_rate": 2.9600955242277173e-05, + "loss": 0.0, + "step": 10592 + }, + { + "epoch": 2.45, + "learning_rate": 2.9599029350589324e-05, + "loss": 0.0, + "step": 10593 + }, + { + "epoch": 2.45, + "learning_rate": 2.959710345890147e-05, + "loss": 0.0, + "step": 10594 + }, + { + "epoch": 2.45, + "learning_rate": 2.9595177567213622e-05, + "loss": 0.0, + "step": 10595 + }, + { + "epoch": 2.45, + "learning_rate": 2.959325167552577e-05, + "loss": 0.0, + "step": 10596 + }, + { + "epoch": 2.45, + "learning_rate": 2.959132578383792e-05, + "loss": 0.0001, + "step": 10597 + }, + { + "epoch": 2.45, + "learning_rate": 2.9589399892150065e-05, + "loss": 0.0001, + "step": 10598 + }, + { + "epoch": 2.45, + "learning_rate": 2.958747400046222e-05, + "loss": 0.0003, + "step": 10599 + }, + { + "epoch": 2.45, + "learning_rate": 2.9585548108774363e-05, + "loss": 0.0, + "step": 10600 + }, + { + "epoch": 2.45, + "learning_rate": 2.9583622217086514e-05, + "loss": 0.0, + "step": 10601 + }, + { + "epoch": 2.45, + "learning_rate": 2.958169632539866e-05, + "loss": 0.0, + "step": 10602 + }, + { + "epoch": 2.45, + "learning_rate": 2.957977043371081e-05, + "loss": 0.0, + "step": 10603 + }, + { + "epoch": 2.45, + "learning_rate": 2.957784454202296e-05, + "loss": 0.0, + "step": 10604 + }, + { + "epoch": 2.45, + "learning_rate": 2.9575918650335104e-05, + "loss": 0.0, + "step": 10605 + }, + { + "epoch": 2.45, + "learning_rate": 2.9573992758647258e-05, + "loss": 0.0001, + "step": 10606 + }, + { + "epoch": 2.45, + "learning_rate": 2.9572066866959402e-05, + "loss": 0.0, + "step": 10607 + }, + { + "epoch": 2.45, + "learning_rate": 2.9570140975271553e-05, + "loss": 0.0, + "step": 10608 + }, + { + "epoch": 2.45, + "learning_rate": 2.95682150835837e-05, + "loss": 0.0, + "step": 10609 + }, + { + "epoch": 2.45, + "learning_rate": 2.956628919189585e-05, + "loss": 0.0001, + "step": 10610 + }, + { + "epoch": 2.45, + "learning_rate": 2.9564363300208e-05, + "loss": 0.0, + "step": 10611 + }, + { + "epoch": 2.45, + "learning_rate": 2.9562437408520143e-05, + "loss": 0.0, + "step": 10612 + }, + { + "epoch": 2.45, + "learning_rate": 2.9560511516832297e-05, + "loss": 0.0, + "step": 10613 + }, + { + "epoch": 2.45, + "learning_rate": 2.955858562514444e-05, + "loss": 0.0, + "step": 10614 + }, + { + "epoch": 2.45, + "learning_rate": 2.9556659733456592e-05, + "loss": 0.0, + "step": 10615 + }, + { + "epoch": 2.45, + "learning_rate": 2.955473384176874e-05, + "loss": 0.0003, + "step": 10616 + }, + { + "epoch": 2.45, + "learning_rate": 2.955280795008089e-05, + "loss": 0.0004, + "step": 10617 + }, + { + "epoch": 2.45, + "learning_rate": 2.9550882058393038e-05, + "loss": 0.0, + "step": 10618 + }, + { + "epoch": 2.45, + "learning_rate": 2.954895616670519e-05, + "loss": 0.0, + "step": 10619 + }, + { + "epoch": 2.45, + "learning_rate": 2.9547030275017333e-05, + "loss": 0.0, + "step": 10620 + }, + { + "epoch": 2.45, + "learning_rate": 2.954510438332948e-05, + "loss": 0.0003, + "step": 10621 + }, + { + "epoch": 2.45, + "learning_rate": 2.954317849164163e-05, + "loss": 0.0001, + "step": 10622 + }, + { + "epoch": 2.46, + "learning_rate": 2.954125259995378e-05, + "loss": 0.0, + "step": 10623 + }, + { + "epoch": 2.46, + "learning_rate": 2.953932670826593e-05, + "loss": 0.0, + "step": 10624 + }, + { + "epoch": 2.46, + "learning_rate": 2.9537400816578077e-05, + "loss": 0.0, + "step": 10625 + }, + { + "epoch": 2.46, + "learning_rate": 2.9535474924890228e-05, + "loss": 0.0001, + "step": 10626 + }, + { + "epoch": 2.46, + "learning_rate": 2.9533549033202372e-05, + "loss": 0.0, + "step": 10627 + }, + { + "epoch": 2.46, + "learning_rate": 2.9531623141514526e-05, + "loss": 0.0, + "step": 10628 + }, + { + "epoch": 2.46, + "learning_rate": 2.952969724982667e-05, + "loss": 0.0, + "step": 10629 + }, + { + "epoch": 2.46, + "learning_rate": 2.9527771358138818e-05, + "loss": 0.0, + "step": 10630 + }, + { + "epoch": 2.46, + "learning_rate": 2.952584546645097e-05, + "loss": 0.0, + "step": 10631 + }, + { + "epoch": 2.46, + "learning_rate": 2.9523919574763116e-05, + "loss": 0.0, + "step": 10632 + }, + { + "epoch": 2.46, + "learning_rate": 2.9521993683075267e-05, + "loss": 0.0014, + "step": 10633 + }, + { + "epoch": 2.46, + "learning_rate": 2.952006779138741e-05, + "loss": 0.0, + "step": 10634 + }, + { + "epoch": 2.46, + "learning_rate": 2.9518141899699565e-05, + "loss": 0.0, + "step": 10635 + }, + { + "epoch": 2.46, + "learning_rate": 2.951621600801171e-05, + "loss": 0.0002, + "step": 10636 + }, + { + "epoch": 2.46, + "learning_rate": 2.951429011632386e-05, + "loss": 0.0001, + "step": 10637 + }, + { + "epoch": 2.46, + "learning_rate": 2.9512364224636008e-05, + "loss": 0.0002, + "step": 10638 + }, + { + "epoch": 2.46, + "learning_rate": 2.9510438332948155e-05, + "loss": 0.0, + "step": 10639 + }, + { + "epoch": 2.46, + "learning_rate": 2.9508512441260306e-05, + "loss": 0.0, + "step": 10640 + }, + { + "epoch": 2.46, + "learning_rate": 2.950658654957245e-05, + "loss": 0.0001, + "step": 10641 + }, + { + "epoch": 2.46, + "learning_rate": 2.9504660657884604e-05, + "loss": 0.0, + "step": 10642 + }, + { + "epoch": 2.46, + "learning_rate": 2.950273476619675e-05, + "loss": 0.0, + "step": 10643 + }, + { + "epoch": 2.46, + "learning_rate": 2.95008088745089e-05, + "loss": 0.0, + "step": 10644 + }, + { + "epoch": 2.46, + "learning_rate": 2.9498882982821047e-05, + "loss": 0.0, + "step": 10645 + }, + { + "epoch": 2.46, + "learning_rate": 2.9496957091133198e-05, + "loss": 0.0, + "step": 10646 + }, + { + "epoch": 2.46, + "learning_rate": 2.9495031199445345e-05, + "loss": 0.0001, + "step": 10647 + }, + { + "epoch": 2.46, + "learning_rate": 2.949310530775749e-05, + "loss": 0.0, + "step": 10648 + }, + { + "epoch": 2.46, + "learning_rate": 2.9491179416069644e-05, + "loss": 0.0003, + "step": 10649 + }, + { + "epoch": 2.46, + "learning_rate": 2.9489253524381788e-05, + "loss": 0.0001, + "step": 10650 + }, + { + "epoch": 2.46, + "learning_rate": 2.948732763269394e-05, + "loss": 0.0, + "step": 10651 + }, + { + "epoch": 2.46, + "learning_rate": 2.9485401741006086e-05, + "loss": 0.0, + "step": 10652 + }, + { + "epoch": 2.46, + "learning_rate": 2.9483475849318237e-05, + "loss": 0.0001, + "step": 10653 + }, + { + "epoch": 2.46, + "learning_rate": 2.9481549957630384e-05, + "loss": 0.0, + "step": 10654 + }, + { + "epoch": 2.46, + "learning_rate": 2.9479624065942535e-05, + "loss": 0.0, + "step": 10655 + }, + { + "epoch": 2.46, + "learning_rate": 2.947769817425468e-05, + "loss": 0.0, + "step": 10656 + }, + { + "epoch": 2.46, + "learning_rate": 2.9475772282566834e-05, + "loss": 0.0, + "step": 10657 + }, + { + "epoch": 2.46, + "learning_rate": 2.9473846390878978e-05, + "loss": 0.0, + "step": 10658 + }, + { + "epoch": 2.46, + "learning_rate": 2.9471920499191125e-05, + "loss": 0.0, + "step": 10659 + }, + { + "epoch": 2.46, + "learning_rate": 2.9469994607503276e-05, + "loss": 0.0, + "step": 10660 + }, + { + "epoch": 2.46, + "learning_rate": 2.9468068715815423e-05, + "loss": 0.0, + "step": 10661 + }, + { + "epoch": 2.46, + "learning_rate": 2.9466142824127574e-05, + "loss": 0.0, + "step": 10662 + }, + { + "epoch": 2.46, + "learning_rate": 2.946421693243972e-05, + "loss": 0.0, + "step": 10663 + }, + { + "epoch": 2.46, + "learning_rate": 2.9462291040751873e-05, + "loss": 0.0, + "step": 10664 + }, + { + "epoch": 2.46, + "learning_rate": 2.9460365149064017e-05, + "loss": 0.0, + "step": 10665 + }, + { + "epoch": 2.46, + "learning_rate": 2.9458439257376168e-05, + "loss": 0.0, + "step": 10666 + }, + { + "epoch": 2.47, + "learning_rate": 2.9456513365688315e-05, + "loss": 0.0001, + "step": 10667 + }, + { + "epoch": 2.47, + "learning_rate": 2.9454587474000463e-05, + "loss": 0.0, + "step": 10668 + }, + { + "epoch": 2.47, + "learning_rate": 2.9452661582312613e-05, + "loss": 0.0, + "step": 10669 + }, + { + "epoch": 2.47, + "learning_rate": 2.9450735690624758e-05, + "loss": 0.0001, + "step": 10670 + }, + { + "epoch": 2.47, + "learning_rate": 2.9448809798936912e-05, + "loss": 0.0, + "step": 10671 + }, + { + "epoch": 2.47, + "learning_rate": 2.9446883907249056e-05, + "loss": 0.0, + "step": 10672 + }, + { + "epoch": 2.47, + "learning_rate": 2.9444958015561207e-05, + "loss": 0.0, + "step": 10673 + }, + { + "epoch": 2.47, + "learning_rate": 2.9443032123873354e-05, + "loss": 0.0003, + "step": 10674 + }, + { + "epoch": 2.47, + "learning_rate": 2.9441106232185505e-05, + "loss": 0.0, + "step": 10675 + }, + { + "epoch": 2.47, + "learning_rate": 2.9439180340497653e-05, + "loss": 0.0, + "step": 10676 + }, + { + "epoch": 2.47, + "learning_rate": 2.9437254448809797e-05, + "loss": 0.0, + "step": 10677 + }, + { + "epoch": 2.47, + "learning_rate": 2.943532855712195e-05, + "loss": 0.0, + "step": 10678 + }, + { + "epoch": 2.47, + "learning_rate": 2.9433402665434095e-05, + "loss": 0.0, + "step": 10679 + }, + { + "epoch": 2.47, + "learning_rate": 2.9431476773746246e-05, + "loss": 0.0001, + "step": 10680 + }, + { + "epoch": 2.47, + "learning_rate": 2.9429550882058393e-05, + "loss": 0.0, + "step": 10681 + }, + { + "epoch": 2.47, + "learning_rate": 2.9427624990370544e-05, + "loss": 0.0, + "step": 10682 + }, + { + "epoch": 2.47, + "learning_rate": 2.9425699098682692e-05, + "loss": 0.0001, + "step": 10683 + }, + { + "epoch": 2.47, + "learning_rate": 2.9423773206994843e-05, + "loss": 0.0, + "step": 10684 + }, + { + "epoch": 2.47, + "learning_rate": 2.942184731530699e-05, + "loss": 0.0, + "step": 10685 + }, + { + "epoch": 2.47, + "learning_rate": 2.9419921423619134e-05, + "loss": 0.0, + "step": 10686 + }, + { + "epoch": 2.47, + "learning_rate": 2.9417995531931285e-05, + "loss": 0.0, + "step": 10687 + }, + { + "epoch": 2.47, + "learning_rate": 2.9416069640243432e-05, + "loss": 0.0, + "step": 10688 + }, + { + "epoch": 2.47, + "learning_rate": 2.9414143748555583e-05, + "loss": 0.0, + "step": 10689 + }, + { + "epoch": 2.47, + "learning_rate": 2.941221785686773e-05, + "loss": 0.0, + "step": 10690 + }, + { + "epoch": 2.47, + "learning_rate": 2.941029196517988e-05, + "loss": 0.0, + "step": 10691 + }, + { + "epoch": 2.47, + "learning_rate": 2.9408366073492026e-05, + "loss": 0.0001, + "step": 10692 + }, + { + "epoch": 2.47, + "learning_rate": 2.940644018180418e-05, + "loss": 0.0, + "step": 10693 + }, + { + "epoch": 2.47, + "learning_rate": 2.9404514290116324e-05, + "loss": 0.0, + "step": 10694 + }, + { + "epoch": 2.47, + "learning_rate": 2.940258839842847e-05, + "loss": 0.0, + "step": 10695 + }, + { + "epoch": 2.47, + "learning_rate": 2.9400662506740622e-05, + "loss": 0.0, + "step": 10696 + }, + { + "epoch": 2.47, + "learning_rate": 2.939873661505277e-05, + "loss": 0.0, + "step": 10697 + }, + { + "epoch": 2.47, + "learning_rate": 2.939681072336492e-05, + "loss": 0.0, + "step": 10698 + }, + { + "epoch": 2.47, + "learning_rate": 2.9394884831677065e-05, + "loss": 0.0001, + "step": 10699 + }, + { + "epoch": 2.47, + "learning_rate": 2.939295893998922e-05, + "loss": 0.0, + "step": 10700 + }, + { + "epoch": 2.47, + "learning_rate": 2.9391033048301363e-05, + "loss": 0.0, + "step": 10701 + }, + { + "epoch": 2.47, + "learning_rate": 2.9389107156613514e-05, + "loss": 0.0, + "step": 10702 + }, + { + "epoch": 2.47, + "learning_rate": 2.938718126492566e-05, + "loss": 0.002, + "step": 10703 + }, + { + "epoch": 2.47, + "learning_rate": 2.938525537323781e-05, + "loss": 0.0002, + "step": 10704 + }, + { + "epoch": 2.47, + "learning_rate": 2.938332948154996e-05, + "loss": 0.0, + "step": 10705 + }, + { + "epoch": 2.47, + "learning_rate": 2.9381403589862104e-05, + "loss": 0.0, + "step": 10706 + }, + { + "epoch": 2.47, + "learning_rate": 2.9379477698174258e-05, + "loss": 0.0, + "step": 10707 + }, + { + "epoch": 2.47, + "learning_rate": 2.9377551806486402e-05, + "loss": 0.0, + "step": 10708 + }, + { + "epoch": 2.47, + "learning_rate": 2.9375625914798553e-05, + "loss": 0.0001, + "step": 10709 + }, + { + "epoch": 2.48, + "learning_rate": 2.93737000231107e-05, + "loss": 0.0, + "step": 10710 + }, + { + "epoch": 2.48, + "learning_rate": 2.937177413142285e-05, + "loss": 0.0, + "step": 10711 + }, + { + "epoch": 2.48, + "learning_rate": 2.9369848239735e-05, + "loss": 0.0, + "step": 10712 + }, + { + "epoch": 2.48, + "learning_rate": 2.936792234804715e-05, + "loss": 0.0, + "step": 10713 + }, + { + "epoch": 2.48, + "learning_rate": 2.9365996456359297e-05, + "loss": 0.0001, + "step": 10714 + }, + { + "epoch": 2.48, + "learning_rate": 2.936407056467144e-05, + "loss": 0.0001, + "step": 10715 + }, + { + "epoch": 2.48, + "learning_rate": 2.9362144672983592e-05, + "loss": 0.0, + "step": 10716 + }, + { + "epoch": 2.48, + "learning_rate": 2.936021878129574e-05, + "loss": 0.0, + "step": 10717 + }, + { + "epoch": 2.48, + "learning_rate": 2.935829288960789e-05, + "loss": 0.0004, + "step": 10718 + }, + { + "epoch": 2.48, + "learning_rate": 2.9356366997920038e-05, + "loss": 0.0001, + "step": 10719 + }, + { + "epoch": 2.48, + "learning_rate": 2.935444110623219e-05, + "loss": 0.0, + "step": 10720 + }, + { + "epoch": 2.48, + "learning_rate": 2.9352515214544337e-05, + "loss": 0.0, + "step": 10721 + }, + { + "epoch": 2.48, + "learning_rate": 2.9350589322856487e-05, + "loss": 0.0001, + "step": 10722 + }, + { + "epoch": 2.48, + "learning_rate": 2.934866343116863e-05, + "loss": 0.0, + "step": 10723 + }, + { + "epoch": 2.48, + "learning_rate": 2.934673753948078e-05, + "loss": 0.0, + "step": 10724 + }, + { + "epoch": 2.48, + "learning_rate": 2.934481164779293e-05, + "loss": 0.0, + "step": 10725 + }, + { + "epoch": 2.48, + "learning_rate": 2.9342885756105077e-05, + "loss": 0.0, + "step": 10726 + }, + { + "epoch": 2.48, + "learning_rate": 2.9340959864417228e-05, + "loss": 0.0, + "step": 10727 + }, + { + "epoch": 2.48, + "learning_rate": 2.9339033972729372e-05, + "loss": 0.0, + "step": 10728 + }, + { + "epoch": 2.48, + "learning_rate": 2.9337108081041526e-05, + "loss": 0.0, + "step": 10729 + }, + { + "epoch": 2.48, + "learning_rate": 2.933518218935367e-05, + "loss": 0.0, + "step": 10730 + }, + { + "epoch": 2.48, + "learning_rate": 2.9333256297665825e-05, + "loss": 0.0001, + "step": 10731 + }, + { + "epoch": 2.48, + "learning_rate": 2.933133040597797e-05, + "loss": 0.0, + "step": 10732 + }, + { + "epoch": 2.48, + "learning_rate": 2.9329404514290116e-05, + "loss": 0.0, + "step": 10733 + }, + { + "epoch": 2.48, + "learning_rate": 2.9327478622602267e-05, + "loss": 0.0, + "step": 10734 + }, + { + "epoch": 2.48, + "learning_rate": 2.932555273091441e-05, + "loss": 0.0012, + "step": 10735 + }, + { + "epoch": 2.48, + "learning_rate": 2.9323626839226566e-05, + "loss": 0.0001, + "step": 10736 + }, + { + "epoch": 2.48, + "learning_rate": 2.932170094753871e-05, + "loss": 0.0, + "step": 10737 + }, + { + "epoch": 2.48, + "learning_rate": 2.931977505585086e-05, + "loss": 0.0001, + "step": 10738 + }, + { + "epoch": 2.48, + "learning_rate": 2.9317849164163008e-05, + "loss": 0.0, + "step": 10739 + }, + { + "epoch": 2.48, + "learning_rate": 2.931592327247516e-05, + "loss": 0.0005, + "step": 10740 + }, + { + "epoch": 2.48, + "learning_rate": 2.9313997380787306e-05, + "loss": 0.0, + "step": 10741 + }, + { + "epoch": 2.48, + "learning_rate": 2.931207148909945e-05, + "loss": 0.0, + "step": 10742 + }, + { + "epoch": 2.48, + "learning_rate": 2.9310145597411605e-05, + "loss": 0.0002, + "step": 10743 + }, + { + "epoch": 2.48, + "learning_rate": 2.930821970572375e-05, + "loss": 0.0, + "step": 10744 + }, + { + "epoch": 2.48, + "learning_rate": 2.93062938140359e-05, + "loss": 0.0001, + "step": 10745 + }, + { + "epoch": 2.48, + "learning_rate": 2.9304367922348047e-05, + "loss": 0.0, + "step": 10746 + }, + { + "epoch": 2.48, + "learning_rate": 2.9302442030660198e-05, + "loss": 0.0, + "step": 10747 + }, + { + "epoch": 2.48, + "learning_rate": 2.9300516138972346e-05, + "loss": 0.0, + "step": 10748 + }, + { + "epoch": 2.48, + "learning_rate": 2.9298590247284496e-05, + "loss": 0.0, + "step": 10749 + }, + { + "epoch": 2.48, + "learning_rate": 2.9296664355596644e-05, + "loss": 0.0002, + "step": 10750 + }, + { + "epoch": 2.48, + "learning_rate": 2.9294738463908788e-05, + "loss": 0.0001, + "step": 10751 + }, + { + "epoch": 2.48, + "learning_rate": 2.929281257222094e-05, + "loss": 0.0001, + "step": 10752 + }, + { + "epoch": 2.49, + "learning_rate": 2.9290886680533086e-05, + "loss": 0.0008, + "step": 10753 + }, + { + "epoch": 2.49, + "learning_rate": 2.9288960788845237e-05, + "loss": 0.0, + "step": 10754 + }, + { + "epoch": 2.49, + "learning_rate": 2.9287034897157385e-05, + "loss": 0.0, + "step": 10755 + }, + { + "epoch": 2.49, + "learning_rate": 2.9285109005469536e-05, + "loss": 0.0, + "step": 10756 + }, + { + "epoch": 2.49, + "learning_rate": 2.9283183113781683e-05, + "loss": 0.0002, + "step": 10757 + }, + { + "epoch": 2.49, + "learning_rate": 2.9281257222093834e-05, + "loss": 0.0, + "step": 10758 + }, + { + "epoch": 2.49, + "learning_rate": 2.9279331330405978e-05, + "loss": 0.0001, + "step": 10759 + }, + { + "epoch": 2.49, + "learning_rate": 2.9277405438718125e-05, + "loss": 0.0, + "step": 10760 + }, + { + "epoch": 2.49, + "learning_rate": 2.9275479547030276e-05, + "loss": 0.0, + "step": 10761 + }, + { + "epoch": 2.49, + "learning_rate": 2.9273553655342424e-05, + "loss": 0.0, + "step": 10762 + }, + { + "epoch": 2.49, + "learning_rate": 2.9271627763654575e-05, + "loss": 0.0, + "step": 10763 + }, + { + "epoch": 2.49, + "learning_rate": 2.9269701871966722e-05, + "loss": 0.0001, + "step": 10764 + }, + { + "epoch": 2.49, + "learning_rate": 2.9267775980278873e-05, + "loss": 0.0, + "step": 10765 + }, + { + "epoch": 2.49, + "learning_rate": 2.9265850088591017e-05, + "loss": 0.0, + "step": 10766 + }, + { + "epoch": 2.49, + "learning_rate": 2.926392419690317e-05, + "loss": 0.0, + "step": 10767 + }, + { + "epoch": 2.49, + "learning_rate": 2.9261998305215315e-05, + "loss": 0.0, + "step": 10768 + }, + { + "epoch": 2.49, + "learning_rate": 2.9260072413527466e-05, + "loss": 0.0, + "step": 10769 + }, + { + "epoch": 2.49, + "learning_rate": 2.9258146521839614e-05, + "loss": 0.0001, + "step": 10770 + }, + { + "epoch": 2.49, + "learning_rate": 2.9256220630151758e-05, + "loss": 0.0, + "step": 10771 + }, + { + "epoch": 2.49, + "learning_rate": 2.9254294738463912e-05, + "loss": 0.0, + "step": 10772 + }, + { + "epoch": 2.49, + "learning_rate": 2.9252368846776056e-05, + "loss": 0.0, + "step": 10773 + }, + { + "epoch": 2.49, + "learning_rate": 2.9250442955088207e-05, + "loss": 0.0004, + "step": 10774 + }, + { + "epoch": 2.49, + "learning_rate": 2.9248517063400355e-05, + "loss": 0.0, + "step": 10775 + }, + { + "epoch": 2.49, + "learning_rate": 2.9246591171712505e-05, + "loss": 0.0, + "step": 10776 + }, + { + "epoch": 2.49, + "learning_rate": 2.9244665280024653e-05, + "loss": 0.0, + "step": 10777 + }, + { + "epoch": 2.49, + "learning_rate": 2.9242739388336804e-05, + "loss": 0.0, + "step": 10778 + }, + { + "epoch": 2.49, + "learning_rate": 2.924081349664895e-05, + "loss": 0.0001, + "step": 10779 + }, + { + "epoch": 2.49, + "learning_rate": 2.9238887604961095e-05, + "loss": 0.0, + "step": 10780 + }, + { + "epoch": 2.49, + "learning_rate": 2.9236961713273246e-05, + "loss": 0.0001, + "step": 10781 + }, + { + "epoch": 2.49, + "learning_rate": 2.9235035821585394e-05, + "loss": 0.0, + "step": 10782 + }, + { + "epoch": 2.49, + "learning_rate": 2.9233109929897545e-05, + "loss": 0.0, + "step": 10783 + }, + { + "epoch": 2.49, + "learning_rate": 2.9231184038209692e-05, + "loss": 0.0, + "step": 10784 + }, + { + "epoch": 2.49, + "learning_rate": 2.9229258146521843e-05, + "loss": 0.0, + "step": 10785 + }, + { + "epoch": 2.49, + "learning_rate": 2.922733225483399e-05, + "loss": 0.0, + "step": 10786 + }, + { + "epoch": 2.49, + "learning_rate": 2.922540636314614e-05, + "loss": 0.0008, + "step": 10787 + }, + { + "epoch": 2.49, + "learning_rate": 2.9223480471458285e-05, + "loss": 0.0, + "step": 10788 + }, + { + "epoch": 2.49, + "learning_rate": 2.9221554579770433e-05, + "loss": 0.0, + "step": 10789 + }, + { + "epoch": 2.49, + "learning_rate": 2.9219628688082584e-05, + "loss": 0.0001, + "step": 10790 + }, + { + "epoch": 2.49, + "learning_rate": 2.921770279639473e-05, + "loss": 0.0005, + "step": 10791 + }, + { + "epoch": 2.49, + "learning_rate": 2.9215776904706882e-05, + "loss": 0.0, + "step": 10792 + }, + { + "epoch": 2.49, + "learning_rate": 2.921385101301903e-05, + "loss": 0.0, + "step": 10793 + }, + { + "epoch": 2.49, + "learning_rate": 2.921192512133118e-05, + "loss": 0.0001, + "step": 10794 + }, + { + "epoch": 2.49, + "learning_rate": 2.9209999229643324e-05, + "loss": 0.0, + "step": 10795 + }, + { + "epoch": 2.5, + "learning_rate": 2.920807333795548e-05, + "loss": 0.0001, + "step": 10796 + }, + { + "epoch": 2.5, + "learning_rate": 2.9206147446267623e-05, + "loss": 0.0, + "step": 10797 + }, + { + "epoch": 2.5, + "learning_rate": 2.920422155457977e-05, + "loss": 0.0, + "step": 10798 + }, + { + "epoch": 2.5, + "learning_rate": 2.920229566289192e-05, + "loss": 0.0001, + "step": 10799 + }, + { + "epoch": 2.5, + "learning_rate": 2.920036977120407e-05, + "loss": 0.0, + "step": 10800 + }, + { + "epoch": 2.5, + "learning_rate": 2.919844387951622e-05, + "loss": 0.0001, + "step": 10801 + }, + { + "epoch": 2.5, + "learning_rate": 2.9196517987828364e-05, + "loss": 0.0, + "step": 10802 + }, + { + "epoch": 2.5, + "learning_rate": 2.9194592096140518e-05, + "loss": 0.0001, + "step": 10803 + }, + { + "epoch": 2.5, + "learning_rate": 2.9192666204452662e-05, + "loss": 0.0, + "step": 10804 + }, + { + "epoch": 2.5, + "learning_rate": 2.9190740312764813e-05, + "loss": 0.0002, + "step": 10805 + }, + { + "epoch": 2.5, + "learning_rate": 2.918881442107696e-05, + "loss": 0.0001, + "step": 10806 + }, + { + "epoch": 2.5, + "learning_rate": 2.9186888529389104e-05, + "loss": 0.0, + "step": 10807 + }, + { + "epoch": 2.5, + "learning_rate": 2.918496263770126e-05, + "loss": 0.0, + "step": 10808 + }, + { + "epoch": 2.5, + "learning_rate": 2.9183036746013403e-05, + "loss": 0.0001, + "step": 10809 + }, + { + "epoch": 2.5, + "learning_rate": 2.9181110854325554e-05, + "loss": 0.0, + "step": 10810 + }, + { + "epoch": 2.5, + "learning_rate": 2.91791849626377e-05, + "loss": 0.0002, + "step": 10811 + }, + { + "epoch": 2.5, + "learning_rate": 2.9177259070949852e-05, + "loss": 0.0001, + "step": 10812 + }, + { + "epoch": 2.5, + "learning_rate": 2.9175333179262e-05, + "loss": 0.0002, + "step": 10813 + }, + { + "epoch": 2.5, + "learning_rate": 2.917340728757415e-05, + "loss": 0.0, + "step": 10814 + }, + { + "epoch": 2.5, + "learning_rate": 2.9171481395886298e-05, + "loss": 0.0, + "step": 10815 + }, + { + "epoch": 2.5, + "learning_rate": 2.9169555504198442e-05, + "loss": 0.0005, + "step": 10816 + }, + { + "epoch": 2.5, + "learning_rate": 2.9167629612510593e-05, + "loss": 0.0, + "step": 10817 + }, + { + "epoch": 2.5, + "learning_rate": 2.916570372082274e-05, + "loss": 0.0, + "step": 10818 + }, + { + "epoch": 2.5, + "learning_rate": 2.916377782913489e-05, + "loss": 0.0019, + "step": 10819 + }, + { + "epoch": 2.5, + "learning_rate": 2.916185193744704e-05, + "loss": 0.0, + "step": 10820 + }, + { + "epoch": 2.5, + "learning_rate": 2.915992604575919e-05, + "loss": 0.0002, + "step": 10821 + }, + { + "epoch": 2.5, + "learning_rate": 2.9158000154071337e-05, + "loss": 0.0, + "step": 10822 + }, + { + "epoch": 2.5, + "learning_rate": 2.9156074262383488e-05, + "loss": 0.0, + "step": 10823 + }, + { + "epoch": 2.5, + "learning_rate": 2.9154148370695632e-05, + "loss": 0.0, + "step": 10824 + }, + { + "epoch": 2.5, + "learning_rate": 2.9152222479007786e-05, + "loss": 0.0, + "step": 10825 + }, + { + "epoch": 2.5, + "learning_rate": 2.915029658731993e-05, + "loss": 0.0001, + "step": 10826 + }, + { + "epoch": 2.5, + "learning_rate": 2.9148370695632078e-05, + "loss": 0.0, + "step": 10827 + }, + { + "epoch": 2.5, + "learning_rate": 2.914644480394423e-05, + "loss": 0.0, + "step": 10828 + }, + { + "epoch": 2.5, + "learning_rate": 2.9144518912256376e-05, + "loss": 0.0, + "step": 10829 + }, + { + "epoch": 2.5, + "learning_rate": 2.9142593020568527e-05, + "loss": 0.0001, + "step": 10830 + }, + { + "epoch": 2.5, + "learning_rate": 2.914066712888067e-05, + "loss": 0.0003, + "step": 10831 + }, + { + "epoch": 2.5, + "learning_rate": 2.9138741237192825e-05, + "loss": 0.0, + "step": 10832 + }, + { + "epoch": 2.5, + "learning_rate": 2.913681534550497e-05, + "loss": 0.0, + "step": 10833 + }, + { + "epoch": 2.5, + "learning_rate": 2.913488945381712e-05, + "loss": 0.0, + "step": 10834 + }, + { + "epoch": 2.5, + "learning_rate": 2.9132963562129268e-05, + "loss": 0.0001, + "step": 10835 + }, + { + "epoch": 2.5, + "learning_rate": 2.9131037670441415e-05, + "loss": 0.0001, + "step": 10836 + }, + { + "epoch": 2.5, + "learning_rate": 2.9129111778753566e-05, + "loss": 0.0, + "step": 10837 + }, + { + "epoch": 2.5, + "learning_rate": 2.912718588706571e-05, + "loss": 0.0, + "step": 10838 + }, + { + "epoch": 2.5, + "learning_rate": 2.9125259995377864e-05, + "loss": 0.0001, + "step": 10839 + }, + { + "epoch": 2.51, + "learning_rate": 2.912333410369001e-05, + "loss": 0.0, + "step": 10840 + }, + { + "epoch": 2.51, + "learning_rate": 2.912140821200216e-05, + "loss": 0.0001, + "step": 10841 + }, + { + "epoch": 2.51, + "learning_rate": 2.9119482320314307e-05, + "loss": 0.0, + "step": 10842 + }, + { + "epoch": 2.51, + "learning_rate": 2.9117556428626458e-05, + "loss": 0.0, + "step": 10843 + }, + { + "epoch": 2.51, + "learning_rate": 2.9115630536938605e-05, + "loss": 0.0001, + "step": 10844 + }, + { + "epoch": 2.51, + "learning_rate": 2.911370464525075e-05, + "loss": 0.0002, + "step": 10845 + }, + { + "epoch": 2.51, + "learning_rate": 2.91117787535629e-05, + "loss": 0.0, + "step": 10846 + }, + { + "epoch": 2.51, + "learning_rate": 2.9109852861875047e-05, + "loss": 0.0, + "step": 10847 + }, + { + "epoch": 2.51, + "learning_rate": 2.91079269701872e-05, + "loss": 0.0, + "step": 10848 + }, + { + "epoch": 2.51, + "learning_rate": 2.9106001078499346e-05, + "loss": 0.0, + "step": 10849 + }, + { + "epoch": 2.51, + "learning_rate": 2.9104075186811497e-05, + "loss": 0.0001, + "step": 10850 + }, + { + "epoch": 2.51, + "learning_rate": 2.9102149295123644e-05, + "loss": 0.0, + "step": 10851 + }, + { + "epoch": 2.51, + "learning_rate": 2.9100223403435795e-05, + "loss": 0.0, + "step": 10852 + }, + { + "epoch": 2.51, + "learning_rate": 2.909829751174794e-05, + "loss": 0.0001, + "step": 10853 + }, + { + "epoch": 2.51, + "learning_rate": 2.9096371620060087e-05, + "loss": 0.0, + "step": 10854 + }, + { + "epoch": 2.51, + "learning_rate": 2.9094445728372237e-05, + "loss": 0.0, + "step": 10855 + }, + { + "epoch": 2.51, + "learning_rate": 2.9092519836684385e-05, + "loss": 0.0, + "step": 10856 + }, + { + "epoch": 2.51, + "learning_rate": 2.9090593944996536e-05, + "loss": 0.0002, + "step": 10857 + }, + { + "epoch": 2.51, + "learning_rate": 2.9088668053308683e-05, + "loss": 0.0, + "step": 10858 + }, + { + "epoch": 2.51, + "learning_rate": 2.9086742161620834e-05, + "loss": 0.0014, + "step": 10859 + }, + { + "epoch": 2.51, + "learning_rate": 2.9084816269932978e-05, + "loss": 0.0, + "step": 10860 + }, + { + "epoch": 2.51, + "learning_rate": 2.9082890378245132e-05, + "loss": 0.0, + "step": 10861 + }, + { + "epoch": 2.51, + "learning_rate": 2.9080964486557277e-05, + "loss": 0.0, + "step": 10862 + }, + { + "epoch": 2.51, + "learning_rate": 2.9079038594869424e-05, + "loss": 0.0, + "step": 10863 + }, + { + "epoch": 2.51, + "learning_rate": 2.9077112703181575e-05, + "loss": 0.0, + "step": 10864 + }, + { + "epoch": 2.51, + "learning_rate": 2.9075186811493722e-05, + "loss": 0.0, + "step": 10865 + }, + { + "epoch": 2.51, + "learning_rate": 2.9073260919805873e-05, + "loss": 0.0, + "step": 10866 + }, + { + "epoch": 2.51, + "learning_rate": 2.9071335028118017e-05, + "loss": 0.0001, + "step": 10867 + }, + { + "epoch": 2.51, + "learning_rate": 2.906940913643017e-05, + "loss": 0.0, + "step": 10868 + }, + { + "epoch": 2.51, + "learning_rate": 2.9067483244742316e-05, + "loss": 0.0, + "step": 10869 + }, + { + "epoch": 2.51, + "learning_rate": 2.9065557353054467e-05, + "loss": 0.0, + "step": 10870 + }, + { + "epoch": 2.51, + "learning_rate": 2.9063631461366614e-05, + "loss": 0.0, + "step": 10871 + }, + { + "epoch": 2.51, + "learning_rate": 2.906170556967876e-05, + "loss": 0.0, + "step": 10872 + }, + { + "epoch": 2.51, + "learning_rate": 2.9059779677990912e-05, + "loss": 0.0, + "step": 10873 + }, + { + "epoch": 2.51, + "learning_rate": 2.9057853786303056e-05, + "loss": 0.0, + "step": 10874 + }, + { + "epoch": 2.51, + "learning_rate": 2.905592789461521e-05, + "loss": 0.0, + "step": 10875 + }, + { + "epoch": 2.51, + "learning_rate": 2.9054002002927355e-05, + "loss": 0.0, + "step": 10876 + }, + { + "epoch": 2.51, + "learning_rate": 2.9052076111239506e-05, + "loss": 0.0, + "step": 10877 + }, + { + "epoch": 2.51, + "learning_rate": 2.9050150219551653e-05, + "loss": 0.0, + "step": 10878 + }, + { + "epoch": 2.51, + "learning_rate": 2.9048224327863804e-05, + "loss": 0.0, + "step": 10879 + }, + { + "epoch": 2.51, + "learning_rate": 2.904629843617595e-05, + "loss": 0.0, + "step": 10880 + }, + { + "epoch": 2.51, + "learning_rate": 2.9044372544488102e-05, + "loss": 0.0, + "step": 10881 + }, + { + "epoch": 2.51, + "learning_rate": 2.9042446652800246e-05, + "loss": 0.0, + "step": 10882 + }, + { + "epoch": 2.52, + "learning_rate": 2.9040520761112394e-05, + "loss": 0.0001, + "step": 10883 + }, + { + "epoch": 2.52, + "learning_rate": 2.9038594869424545e-05, + "loss": 0.0, + "step": 10884 + }, + { + "epoch": 2.52, + "learning_rate": 2.9036668977736692e-05, + "loss": 0.0, + "step": 10885 + }, + { + "epoch": 2.52, + "learning_rate": 2.9034743086048843e-05, + "loss": 0.0, + "step": 10886 + }, + { + "epoch": 2.52, + "learning_rate": 2.903281719436099e-05, + "loss": 0.0016, + "step": 10887 + }, + { + "epoch": 2.52, + "learning_rate": 2.903089130267314e-05, + "loss": 0.0, + "step": 10888 + }, + { + "epoch": 2.52, + "learning_rate": 2.9028965410985286e-05, + "loss": 0.0, + "step": 10889 + }, + { + "epoch": 2.52, + "learning_rate": 2.902703951929744e-05, + "loss": 0.0, + "step": 10890 + }, + { + "epoch": 2.52, + "learning_rate": 2.9025113627609584e-05, + "loss": 0.0, + "step": 10891 + }, + { + "epoch": 2.52, + "learning_rate": 2.902318773592173e-05, + "loss": 0.0, + "step": 10892 + }, + { + "epoch": 2.52, + "learning_rate": 2.9021261844233882e-05, + "loss": 0.0, + "step": 10893 + }, + { + "epoch": 2.52, + "learning_rate": 2.901933595254603e-05, + "loss": 0.0, + "step": 10894 + }, + { + "epoch": 2.52, + "learning_rate": 2.901741006085818e-05, + "loss": 0.0, + "step": 10895 + }, + { + "epoch": 2.52, + "learning_rate": 2.9015484169170325e-05, + "loss": 0.0, + "step": 10896 + }, + { + "epoch": 2.52, + "learning_rate": 2.901355827748248e-05, + "loss": 0.0, + "step": 10897 + }, + { + "epoch": 2.52, + "learning_rate": 2.9011632385794623e-05, + "loss": 0.0, + "step": 10898 + }, + { + "epoch": 2.52, + "learning_rate": 2.9009706494106774e-05, + "loss": 0.0, + "step": 10899 + }, + { + "epoch": 2.52, + "learning_rate": 2.900778060241892e-05, + "loss": 0.0, + "step": 10900 + }, + { + "epoch": 2.52, + "learning_rate": 2.900585471073107e-05, + "loss": 0.0001, + "step": 10901 + }, + { + "epoch": 2.52, + "learning_rate": 2.900392881904322e-05, + "loss": 0.0, + "step": 10902 + }, + { + "epoch": 2.52, + "learning_rate": 2.9002002927355364e-05, + "loss": 0.0, + "step": 10903 + }, + { + "epoch": 2.52, + "learning_rate": 2.9000077035667518e-05, + "loss": 0.0013, + "step": 10904 + }, + { + "epoch": 2.52, + "learning_rate": 2.8998151143979662e-05, + "loss": 0.0, + "step": 10905 + }, + { + "epoch": 2.52, + "learning_rate": 2.8996225252291813e-05, + "loss": 0.0, + "step": 10906 + }, + { + "epoch": 2.52, + "learning_rate": 2.899429936060396e-05, + "loss": 0.0, + "step": 10907 + }, + { + "epoch": 2.52, + "learning_rate": 2.899237346891611e-05, + "loss": 0.0, + "step": 10908 + }, + { + "epoch": 2.52, + "learning_rate": 2.899044757722826e-05, + "loss": 0.0, + "step": 10909 + }, + { + "epoch": 2.52, + "learning_rate": 2.8988521685540403e-05, + "loss": 0.0, + "step": 10910 + }, + { + "epoch": 2.52, + "learning_rate": 2.8986595793852557e-05, + "loss": 0.0, + "step": 10911 + }, + { + "epoch": 2.52, + "learning_rate": 2.89846699021647e-05, + "loss": 0.0, + "step": 10912 + }, + { + "epoch": 2.52, + "learning_rate": 2.8982744010476852e-05, + "loss": 0.0035, + "step": 10913 + }, + { + "epoch": 2.52, + "learning_rate": 2.8980818118789e-05, + "loss": 0.0, + "step": 10914 + }, + { + "epoch": 2.52, + "learning_rate": 2.897889222710115e-05, + "loss": 0.0, + "step": 10915 + }, + { + "epoch": 2.52, + "learning_rate": 2.8976966335413298e-05, + "loss": 0.0001, + "step": 10916 + }, + { + "epoch": 2.52, + "learning_rate": 2.897504044372545e-05, + "loss": 0.0, + "step": 10917 + }, + { + "epoch": 2.52, + "learning_rate": 2.8973114552037596e-05, + "loss": 0.0, + "step": 10918 + }, + { + "epoch": 2.52, + "learning_rate": 2.897118866034974e-05, + "loss": 0.0, + "step": 10919 + }, + { + "epoch": 2.52, + "learning_rate": 2.896926276866189e-05, + "loss": 0.0, + "step": 10920 + }, + { + "epoch": 2.52, + "learning_rate": 2.896733687697404e-05, + "loss": 0.0, + "step": 10921 + }, + { + "epoch": 2.52, + "learning_rate": 2.896541098528619e-05, + "loss": 0.0, + "step": 10922 + }, + { + "epoch": 2.52, + "learning_rate": 2.8963485093598337e-05, + "loss": 0.0, + "step": 10923 + }, + { + "epoch": 2.52, + "learning_rate": 2.8961559201910488e-05, + "loss": 0.0, + "step": 10924 + }, + { + "epoch": 2.52, + "learning_rate": 2.8959633310222632e-05, + "loss": 0.0, + "step": 10925 + }, + { + "epoch": 2.53, + "learning_rate": 2.8957707418534786e-05, + "loss": 0.0, + "step": 10926 + }, + { + "epoch": 2.53, + "learning_rate": 2.895578152684693e-05, + "loss": 0.0001, + "step": 10927 + }, + { + "epoch": 2.53, + "learning_rate": 2.8953855635159078e-05, + "loss": 0.0002, + "step": 10928 + }, + { + "epoch": 2.53, + "learning_rate": 2.895192974347123e-05, + "loss": 0.0, + "step": 10929 + }, + { + "epoch": 2.53, + "learning_rate": 2.8950003851783376e-05, + "loss": 0.0, + "step": 10930 + }, + { + "epoch": 2.53, + "learning_rate": 2.8948077960095527e-05, + "loss": 0.0, + "step": 10931 + }, + { + "epoch": 2.53, + "learning_rate": 2.894615206840767e-05, + "loss": 0.0, + "step": 10932 + }, + { + "epoch": 2.53, + "learning_rate": 2.8944226176719825e-05, + "loss": 0.0001, + "step": 10933 + }, + { + "epoch": 2.53, + "learning_rate": 2.894230028503197e-05, + "loss": 0.0001, + "step": 10934 + }, + { + "epoch": 2.53, + "learning_rate": 2.894037439334412e-05, + "loss": 0.0001, + "step": 10935 + }, + { + "epoch": 2.53, + "learning_rate": 2.8938448501656268e-05, + "loss": 0.0, + "step": 10936 + }, + { + "epoch": 2.53, + "learning_rate": 2.893652260996842e-05, + "loss": 0.0, + "step": 10937 + }, + { + "epoch": 2.53, + "learning_rate": 2.8934596718280566e-05, + "loss": 0.0001, + "step": 10938 + }, + { + "epoch": 2.53, + "learning_rate": 2.893267082659271e-05, + "loss": 0.0, + "step": 10939 + }, + { + "epoch": 2.53, + "learning_rate": 2.8930744934904865e-05, + "loss": 0.0001, + "step": 10940 + }, + { + "epoch": 2.53, + "learning_rate": 2.892881904321701e-05, + "loss": 0.0, + "step": 10941 + }, + { + "epoch": 2.53, + "learning_rate": 2.892689315152916e-05, + "loss": 0.0, + "step": 10942 + }, + { + "epoch": 2.53, + "learning_rate": 2.8924967259841307e-05, + "loss": 0.0, + "step": 10943 + }, + { + "epoch": 2.53, + "learning_rate": 2.8923041368153458e-05, + "loss": 0.0, + "step": 10944 + }, + { + "epoch": 2.53, + "learning_rate": 2.8921115476465605e-05, + "loss": 0.0, + "step": 10945 + }, + { + "epoch": 2.53, + "learning_rate": 2.8919189584777756e-05, + "loss": 0.0, + "step": 10946 + }, + { + "epoch": 2.53, + "learning_rate": 2.8917263693089904e-05, + "loss": 0.0001, + "step": 10947 + }, + { + "epoch": 2.53, + "learning_rate": 2.8915337801402048e-05, + "loss": 0.0, + "step": 10948 + }, + { + "epoch": 2.53, + "learning_rate": 2.89134119097142e-05, + "loss": 0.0, + "step": 10949 + }, + { + "epoch": 2.53, + "learning_rate": 2.8911486018026346e-05, + "loss": 0.0, + "step": 10950 + }, + { + "epoch": 2.53, + "learning_rate": 2.8909560126338497e-05, + "loss": 0.0, + "step": 10951 + }, + { + "epoch": 2.53, + "learning_rate": 2.8907634234650644e-05, + "loss": 0.0001, + "step": 10952 + }, + { + "epoch": 2.53, + "learning_rate": 2.8905708342962795e-05, + "loss": 0.0, + "step": 10953 + }, + { + "epoch": 2.53, + "learning_rate": 2.8903782451274943e-05, + "loss": 0.0, + "step": 10954 + }, + { + "epoch": 2.53, + "learning_rate": 2.8901856559587094e-05, + "loss": 0.0, + "step": 10955 + }, + { + "epoch": 2.53, + "learning_rate": 2.8899930667899238e-05, + "loss": 0.0, + "step": 10956 + }, + { + "epoch": 2.53, + "learning_rate": 2.8898004776211385e-05, + "loss": 0.0, + "step": 10957 + }, + { + "epoch": 2.53, + "learning_rate": 2.8896078884523536e-05, + "loss": 0.0001, + "step": 10958 + }, + { + "epoch": 2.53, + "learning_rate": 2.8894152992835684e-05, + "loss": 0.0, + "step": 10959 + }, + { + "epoch": 2.53, + "learning_rate": 2.8892227101147834e-05, + "loss": 0.0, + "step": 10960 + }, + { + "epoch": 2.53, + "learning_rate": 2.889030120945998e-05, + "loss": 0.0, + "step": 10961 + }, + { + "epoch": 2.53, + "learning_rate": 2.8888375317772133e-05, + "loss": 0.0, + "step": 10962 + }, + { + "epoch": 2.53, + "learning_rate": 2.8886449426084277e-05, + "loss": 0.0, + "step": 10963 + }, + { + "epoch": 2.53, + "learning_rate": 2.8884523534396428e-05, + "loss": 0.0, + "step": 10964 + }, + { + "epoch": 2.53, + "learning_rate": 2.8882597642708575e-05, + "loss": 0.0001, + "step": 10965 + }, + { + "epoch": 2.53, + "learning_rate": 2.8880671751020723e-05, + "loss": 0.0002, + "step": 10966 + }, + { + "epoch": 2.53, + "learning_rate": 2.8878745859332874e-05, + "loss": 0.0, + "step": 10967 + }, + { + "epoch": 2.53, + "learning_rate": 2.8876819967645018e-05, + "loss": 0.0, + "step": 10968 + }, + { + "epoch": 2.54, + "learning_rate": 2.8874894075957172e-05, + "loss": 0.0, + "step": 10969 + }, + { + "epoch": 2.54, + "learning_rate": 2.8872968184269316e-05, + "loss": 0.0, + "step": 10970 + }, + { + "epoch": 2.54, + "learning_rate": 2.8871042292581467e-05, + "loss": 0.0, + "step": 10971 + }, + { + "epoch": 2.54, + "learning_rate": 2.8869116400893614e-05, + "loss": 0.0, + "step": 10972 + }, + { + "epoch": 2.54, + "learning_rate": 2.8867190509205765e-05, + "loss": 0.0, + "step": 10973 + }, + { + "epoch": 2.54, + "learning_rate": 2.8865264617517913e-05, + "loss": 0.0002, + "step": 10974 + }, + { + "epoch": 2.54, + "learning_rate": 2.8863338725830057e-05, + "loss": 0.0001, + "step": 10975 + }, + { + "epoch": 2.54, + "learning_rate": 2.886141283414221e-05, + "loss": 0.0001, + "step": 10976 + }, + { + "epoch": 2.54, + "learning_rate": 2.8859486942454355e-05, + "loss": 0.0, + "step": 10977 + }, + { + "epoch": 2.54, + "learning_rate": 2.8857561050766506e-05, + "loss": 0.0, + "step": 10978 + }, + { + "epoch": 2.54, + "learning_rate": 2.8855635159078653e-05, + "loss": 0.0, + "step": 10979 + }, + { + "epoch": 2.54, + "learning_rate": 2.8853709267390804e-05, + "loss": 0.0, + "step": 10980 + }, + { + "epoch": 2.54, + "learning_rate": 2.8851783375702952e-05, + "loss": 0.0, + "step": 10981 + }, + { + "epoch": 2.54, + "learning_rate": 2.8849857484015103e-05, + "loss": 0.0, + "step": 10982 + }, + { + "epoch": 2.54, + "learning_rate": 2.884793159232725e-05, + "loss": 0.0, + "step": 10983 + }, + { + "epoch": 2.54, + "learning_rate": 2.8846005700639394e-05, + "loss": 0.0406, + "step": 10984 + }, + { + "epoch": 2.54, + "learning_rate": 2.8844079808951545e-05, + "loss": 0.0001, + "step": 10985 + }, + { + "epoch": 2.54, + "learning_rate": 2.8842153917263693e-05, + "loss": 0.0, + "step": 10986 + }, + { + "epoch": 2.54, + "learning_rate": 2.8840228025575843e-05, + "loss": 0.0001, + "step": 10987 + }, + { + "epoch": 2.54, + "learning_rate": 2.883830213388799e-05, + "loss": 0.0, + "step": 10988 + }, + { + "epoch": 2.54, + "learning_rate": 2.8836376242200142e-05, + "loss": 0.0001, + "step": 10989 + }, + { + "epoch": 2.54, + "learning_rate": 2.883445035051229e-05, + "loss": 0.0014, + "step": 10990 + }, + { + "epoch": 2.54, + "learning_rate": 2.883252445882444e-05, + "loss": 0.0, + "step": 10991 + }, + { + "epoch": 2.54, + "learning_rate": 2.8830598567136584e-05, + "loss": 0.0, + "step": 10992 + }, + { + "epoch": 2.54, + "learning_rate": 2.882867267544874e-05, + "loss": 0.0, + "step": 10993 + }, + { + "epoch": 2.54, + "learning_rate": 2.8826746783760883e-05, + "loss": 0.0001, + "step": 10994 + }, + { + "epoch": 2.54, + "learning_rate": 2.882482089207303e-05, + "loss": 0.0, + "step": 10995 + }, + { + "epoch": 2.54, + "learning_rate": 2.882289500038518e-05, + "loss": 0.0, + "step": 10996 + }, + { + "epoch": 2.54, + "learning_rate": 2.8820969108697325e-05, + "loss": 0.0, + "step": 10997 + }, + { + "epoch": 2.54, + "learning_rate": 2.881904321700948e-05, + "loss": 0.0, + "step": 10998 + }, + { + "epoch": 2.54, + "learning_rate": 2.8817117325321623e-05, + "loss": 0.0, + "step": 10999 + }, + { + "epoch": 2.54, + "learning_rate": 2.8815191433633774e-05, + "loss": 0.0001, + "step": 11000 + }, + { + "epoch": 2.54, + "learning_rate": 2.881326554194592e-05, + "loss": 0.0, + "step": 11001 + }, + { + "epoch": 2.54, + "learning_rate": 2.8811339650258073e-05, + "loss": 0.0001, + "step": 11002 + }, + { + "epoch": 2.54, + "learning_rate": 2.880941375857022e-05, + "loss": 0.0, + "step": 11003 + }, + { + "epoch": 2.54, + "learning_rate": 2.8807487866882364e-05, + "loss": 0.0008, + "step": 11004 + }, + { + "epoch": 2.54, + "learning_rate": 2.880556197519452e-05, + "loss": 0.0002, + "step": 11005 + }, + { + "epoch": 2.54, + "learning_rate": 2.8803636083506662e-05, + "loss": 0.0, + "step": 11006 + }, + { + "epoch": 2.54, + "learning_rate": 2.8801710191818813e-05, + "loss": 0.0, + "step": 11007 + }, + { + "epoch": 2.54, + "learning_rate": 2.879978430013096e-05, + "loss": 0.0, + "step": 11008 + }, + { + "epoch": 2.54, + "learning_rate": 2.879785840844311e-05, + "loss": 0.0, + "step": 11009 + }, + { + "epoch": 2.54, + "learning_rate": 2.879593251675526e-05, + "loss": 0.0, + "step": 11010 + }, + { + "epoch": 2.54, + "learning_rate": 2.879400662506741e-05, + "loss": 0.0, + "step": 11011 + }, + { + "epoch": 2.54, + "learning_rate": 2.8792080733379557e-05, + "loss": 0.0, + "step": 11012 + }, + { + "epoch": 2.55, + "learning_rate": 2.87901548416917e-05, + "loss": 0.0, + "step": 11013 + }, + { + "epoch": 2.55, + "learning_rate": 2.8788228950003852e-05, + "loss": 0.0001, + "step": 11014 + }, + { + "epoch": 2.55, + "learning_rate": 2.8786303058316e-05, + "loss": 0.0, + "step": 11015 + }, + { + "epoch": 2.55, + "learning_rate": 2.878437716662815e-05, + "loss": 0.0, + "step": 11016 + }, + { + "epoch": 2.55, + "learning_rate": 2.8782451274940298e-05, + "loss": 0.0, + "step": 11017 + }, + { + "epoch": 2.55, + "learning_rate": 2.878052538325245e-05, + "loss": 0.0, + "step": 11018 + }, + { + "epoch": 2.55, + "learning_rate": 2.8778599491564597e-05, + "loss": 0.0, + "step": 11019 + }, + { + "epoch": 2.55, + "learning_rate": 2.8776673599876747e-05, + "loss": 0.0, + "step": 11020 + }, + { + "epoch": 2.55, + "learning_rate": 2.877474770818889e-05, + "loss": 0.0, + "step": 11021 + }, + { + "epoch": 2.55, + "learning_rate": 2.877282181650104e-05, + "loss": 0.0, + "step": 11022 + }, + { + "epoch": 2.55, + "learning_rate": 2.877089592481319e-05, + "loss": 0.0, + "step": 11023 + }, + { + "epoch": 2.55, + "learning_rate": 2.8768970033125337e-05, + "loss": 0.0001, + "step": 11024 + }, + { + "epoch": 2.55, + "learning_rate": 2.8767044141437488e-05, + "loss": 0.0, + "step": 11025 + }, + { + "epoch": 2.55, + "learning_rate": 2.8765118249749636e-05, + "loss": 0.0, + "step": 11026 + }, + { + "epoch": 2.55, + "learning_rate": 2.8763192358061787e-05, + "loss": 0.0, + "step": 11027 + }, + { + "epoch": 2.55, + "learning_rate": 2.876126646637393e-05, + "loss": 0.0, + "step": 11028 + }, + { + "epoch": 2.55, + "learning_rate": 2.8759340574686085e-05, + "loss": 0.0, + "step": 11029 + }, + { + "epoch": 2.55, + "learning_rate": 2.875741468299823e-05, + "loss": 0.0, + "step": 11030 + }, + { + "epoch": 2.55, + "learning_rate": 2.8755488791310376e-05, + "loss": 0.0, + "step": 11031 + }, + { + "epoch": 2.55, + "learning_rate": 2.8753562899622527e-05, + "loss": 0.0, + "step": 11032 + }, + { + "epoch": 2.55, + "learning_rate": 2.875163700793467e-05, + "loss": 0.0, + "step": 11033 + }, + { + "epoch": 2.55, + "learning_rate": 2.8749711116246826e-05, + "loss": 0.0, + "step": 11034 + }, + { + "epoch": 2.55, + "learning_rate": 2.874778522455897e-05, + "loss": 0.0, + "step": 11035 + }, + { + "epoch": 2.55, + "learning_rate": 2.874585933287112e-05, + "loss": 0.0, + "step": 11036 + }, + { + "epoch": 2.55, + "learning_rate": 2.8743933441183268e-05, + "loss": 0.0001, + "step": 11037 + }, + { + "epoch": 2.55, + "learning_rate": 2.874200754949542e-05, + "loss": 0.0, + "step": 11038 + }, + { + "epoch": 2.55, + "learning_rate": 2.8740081657807566e-05, + "loss": 0.0001, + "step": 11039 + }, + { + "epoch": 2.55, + "learning_rate": 2.873815576611971e-05, + "loss": 0.0, + "step": 11040 + }, + { + "epoch": 2.55, + "learning_rate": 2.8736229874431865e-05, + "loss": 0.0, + "step": 11041 + }, + { + "epoch": 2.55, + "learning_rate": 2.873430398274401e-05, + "loss": 0.0, + "step": 11042 + }, + { + "epoch": 2.55, + "learning_rate": 2.873237809105616e-05, + "loss": 0.0, + "step": 11043 + }, + { + "epoch": 2.55, + "learning_rate": 2.8730452199368307e-05, + "loss": 0.0, + "step": 11044 + }, + { + "epoch": 2.55, + "learning_rate": 2.8728526307680458e-05, + "loss": 0.0002, + "step": 11045 + }, + { + "epoch": 2.55, + "learning_rate": 2.8726600415992606e-05, + "loss": 0.0, + "step": 11046 + }, + { + "epoch": 2.55, + "learning_rate": 2.8724674524304756e-05, + "loss": 0.0001, + "step": 11047 + }, + { + "epoch": 2.55, + "learning_rate": 2.8722748632616904e-05, + "loss": 0.0001, + "step": 11048 + }, + { + "epoch": 2.55, + "learning_rate": 2.8720822740929055e-05, + "loss": 0.0, + "step": 11049 + }, + { + "epoch": 2.55, + "learning_rate": 2.87188968492412e-05, + "loss": 0.0, + "step": 11050 + }, + { + "epoch": 2.55, + "learning_rate": 2.8716970957553346e-05, + "loss": 0.0, + "step": 11051 + }, + { + "epoch": 2.55, + "learning_rate": 2.8715045065865497e-05, + "loss": 0.0, + "step": 11052 + }, + { + "epoch": 2.55, + "learning_rate": 2.8713119174177645e-05, + "loss": 0.0, + "step": 11053 + }, + { + "epoch": 2.55, + "learning_rate": 2.8711193282489796e-05, + "loss": 0.0, + "step": 11054 + }, + { + "epoch": 2.55, + "learning_rate": 2.8709267390801943e-05, + "loss": 0.0, + "step": 11055 + }, + { + "epoch": 2.56, + "learning_rate": 2.8707341499114094e-05, + "loss": 0.0, + "step": 11056 + }, + { + "epoch": 2.56, + "learning_rate": 2.8705415607426238e-05, + "loss": 0.0001, + "step": 11057 + }, + { + "epoch": 2.56, + "learning_rate": 2.8703489715738392e-05, + "loss": 0.0003, + "step": 11058 + }, + { + "epoch": 2.56, + "learning_rate": 2.8701563824050536e-05, + "loss": 0.0, + "step": 11059 + }, + { + "epoch": 2.56, + "learning_rate": 2.8699637932362684e-05, + "loss": 0.0, + "step": 11060 + }, + { + "epoch": 2.56, + "learning_rate": 2.8697712040674835e-05, + "loss": 0.0, + "step": 11061 + }, + { + "epoch": 2.56, + "learning_rate": 2.8695786148986982e-05, + "loss": 0.0, + "step": 11062 + }, + { + "epoch": 2.56, + "learning_rate": 2.8693860257299133e-05, + "loss": 0.0, + "step": 11063 + }, + { + "epoch": 2.56, + "learning_rate": 2.8691934365611277e-05, + "loss": 0.0, + "step": 11064 + }, + { + "epoch": 2.56, + "learning_rate": 2.869000847392343e-05, + "loss": 0.0005, + "step": 11065 + }, + { + "epoch": 2.56, + "learning_rate": 2.8688082582235575e-05, + "loss": 0.0, + "step": 11066 + }, + { + "epoch": 2.56, + "learning_rate": 2.8686156690547726e-05, + "loss": 0.0, + "step": 11067 + }, + { + "epoch": 2.56, + "learning_rate": 2.8684230798859874e-05, + "loss": 0.0, + "step": 11068 + }, + { + "epoch": 2.56, + "learning_rate": 2.8682304907172018e-05, + "loss": 0.0, + "step": 11069 + }, + { + "epoch": 2.56, + "learning_rate": 2.8680379015484172e-05, + "loss": 0.0001, + "step": 11070 + }, + { + "epoch": 2.56, + "learning_rate": 2.8678453123796316e-05, + "loss": 0.0, + "step": 11071 + }, + { + "epoch": 2.56, + "learning_rate": 2.8676527232108467e-05, + "loss": 0.0, + "step": 11072 + }, + { + "epoch": 2.56, + "learning_rate": 2.8674601340420615e-05, + "loss": 0.0, + "step": 11073 + }, + { + "epoch": 2.56, + "learning_rate": 2.8672675448732765e-05, + "loss": 0.0, + "step": 11074 + }, + { + "epoch": 2.56, + "learning_rate": 2.8670749557044913e-05, + "loss": 0.0, + "step": 11075 + }, + { + "epoch": 2.56, + "learning_rate": 2.8668823665357064e-05, + "loss": 0.0, + "step": 11076 + }, + { + "epoch": 2.56, + "learning_rate": 2.866689777366921e-05, + "loss": 0.0, + "step": 11077 + }, + { + "epoch": 2.56, + "learning_rate": 2.8664971881981355e-05, + "loss": 0.0, + "step": 11078 + }, + { + "epoch": 2.56, + "learning_rate": 2.8663045990293506e-05, + "loss": 0.0, + "step": 11079 + }, + { + "epoch": 2.56, + "learning_rate": 2.8661120098605654e-05, + "loss": 0.0, + "step": 11080 + }, + { + "epoch": 2.56, + "learning_rate": 2.8659194206917805e-05, + "loss": 0.0, + "step": 11081 + }, + { + "epoch": 2.56, + "learning_rate": 2.8657268315229952e-05, + "loss": 0.0, + "step": 11082 + }, + { + "epoch": 2.56, + "learning_rate": 2.8655342423542103e-05, + "loss": 0.0, + "step": 11083 + }, + { + "epoch": 2.56, + "learning_rate": 2.865341653185425e-05, + "loss": 0.0, + "step": 11084 + }, + { + "epoch": 2.56, + "learning_rate": 2.86514906401664e-05, + "loss": 0.0, + "step": 11085 + }, + { + "epoch": 2.56, + "learning_rate": 2.8649564748478545e-05, + "loss": 0.0, + "step": 11086 + }, + { + "epoch": 2.56, + "learning_rate": 2.8647638856790693e-05, + "loss": 0.0007, + "step": 11087 + }, + { + "epoch": 2.56, + "learning_rate": 2.8645712965102844e-05, + "loss": 0.0, + "step": 11088 + }, + { + "epoch": 2.56, + "learning_rate": 2.864378707341499e-05, + "loss": 0.0001, + "step": 11089 + }, + { + "epoch": 2.56, + "learning_rate": 2.8641861181727142e-05, + "loss": 0.0, + "step": 11090 + }, + { + "epoch": 2.56, + "learning_rate": 2.863993529003929e-05, + "loss": 0.0002, + "step": 11091 + }, + { + "epoch": 2.56, + "learning_rate": 2.863800939835144e-05, + "loss": 0.0, + "step": 11092 + }, + { + "epoch": 2.56, + "learning_rate": 2.8636083506663584e-05, + "loss": 0.0, + "step": 11093 + }, + { + "epoch": 2.56, + "learning_rate": 2.863415761497574e-05, + "loss": 0.0, + "step": 11094 + }, + { + "epoch": 2.56, + "learning_rate": 2.8632231723287883e-05, + "loss": 0.0, + "step": 11095 + }, + { + "epoch": 2.56, + "learning_rate": 2.863030583160003e-05, + "loss": 0.0, + "step": 11096 + }, + { + "epoch": 2.56, + "learning_rate": 2.862837993991218e-05, + "loss": 0.0, + "step": 11097 + }, + { + "epoch": 2.56, + "learning_rate": 2.862645404822433e-05, + "loss": 0.0, + "step": 11098 + }, + { + "epoch": 2.57, + "learning_rate": 2.862452815653648e-05, + "loss": 0.0, + "step": 11099 + }, + { + "epoch": 2.57, + "learning_rate": 2.8622602264848624e-05, + "loss": 0.0, + "step": 11100 + }, + { + "epoch": 2.57, + "learning_rate": 2.8620676373160778e-05, + "loss": 0.0, + "step": 11101 + }, + { + "epoch": 2.57, + "learning_rate": 2.8618750481472922e-05, + "loss": 0.0, + "step": 11102 + }, + { + "epoch": 2.57, + "learning_rate": 2.8616824589785073e-05, + "loss": 0.0, + "step": 11103 + }, + { + "epoch": 2.57, + "learning_rate": 2.861489869809722e-05, + "loss": 0.0, + "step": 11104 + }, + { + "epoch": 2.57, + "learning_rate": 2.861297280640937e-05, + "loss": 0.0, + "step": 11105 + }, + { + "epoch": 2.57, + "learning_rate": 2.861104691472152e-05, + "loss": 0.0, + "step": 11106 + }, + { + "epoch": 2.57, + "learning_rate": 2.8609121023033663e-05, + "loss": 0.0, + "step": 11107 + }, + { + "epoch": 2.57, + "learning_rate": 2.8607195131345817e-05, + "loss": 0.0, + "step": 11108 + }, + { + "epoch": 2.57, + "learning_rate": 2.860526923965796e-05, + "loss": 0.0, + "step": 11109 + }, + { + "epoch": 2.57, + "learning_rate": 2.8603343347970112e-05, + "loss": 0.0, + "step": 11110 + }, + { + "epoch": 2.57, + "learning_rate": 2.860141745628226e-05, + "loss": 0.0, + "step": 11111 + }, + { + "epoch": 2.57, + "learning_rate": 2.859949156459441e-05, + "loss": 0.0, + "step": 11112 + }, + { + "epoch": 2.57, + "learning_rate": 2.8597565672906558e-05, + "loss": 0.0, + "step": 11113 + }, + { + "epoch": 2.57, + "learning_rate": 2.859563978121871e-05, + "loss": 0.0001, + "step": 11114 + }, + { + "epoch": 2.57, + "learning_rate": 2.8593713889530853e-05, + "loss": 0.0, + "step": 11115 + }, + { + "epoch": 2.57, + "learning_rate": 2.8591787997843e-05, + "loss": 0.0, + "step": 11116 + }, + { + "epoch": 2.57, + "learning_rate": 2.858986210615515e-05, + "loss": 0.0, + "step": 11117 + }, + { + "epoch": 2.57, + "learning_rate": 2.85879362144673e-05, + "loss": 0.0, + "step": 11118 + }, + { + "epoch": 2.57, + "learning_rate": 2.858601032277945e-05, + "loss": 0.0002, + "step": 11119 + }, + { + "epoch": 2.57, + "learning_rate": 2.8584084431091597e-05, + "loss": 0.0, + "step": 11120 + }, + { + "epoch": 2.57, + "learning_rate": 2.8582158539403748e-05, + "loss": 0.0, + "step": 11121 + }, + { + "epoch": 2.57, + "learning_rate": 2.8580232647715892e-05, + "loss": 0.0, + "step": 11122 + }, + { + "epoch": 2.57, + "learning_rate": 2.8578306756028046e-05, + "loss": 0.0, + "step": 11123 + }, + { + "epoch": 2.57, + "learning_rate": 2.857638086434019e-05, + "loss": 0.0001, + "step": 11124 + }, + { + "epoch": 2.57, + "learning_rate": 2.8574454972652338e-05, + "loss": 0.0, + "step": 11125 + }, + { + "epoch": 2.57, + "learning_rate": 2.857252908096449e-05, + "loss": 0.0, + "step": 11126 + }, + { + "epoch": 2.57, + "learning_rate": 2.8570603189276636e-05, + "loss": 0.0, + "step": 11127 + }, + { + "epoch": 2.57, + "learning_rate": 2.8568677297588787e-05, + "loss": 0.0, + "step": 11128 + }, + { + "epoch": 2.57, + "learning_rate": 2.856675140590093e-05, + "loss": 0.0, + "step": 11129 + }, + { + "epoch": 2.57, + "learning_rate": 2.8564825514213085e-05, + "loss": 0.0, + "step": 11130 + }, + { + "epoch": 2.57, + "learning_rate": 2.856289962252523e-05, + "loss": 0.0005, + "step": 11131 + }, + { + "epoch": 2.57, + "learning_rate": 2.856097373083738e-05, + "loss": 0.0001, + "step": 11132 + }, + { + "epoch": 2.57, + "learning_rate": 2.8559047839149528e-05, + "loss": 0.0, + "step": 11133 + }, + { + "epoch": 2.57, + "learning_rate": 2.8557121947461675e-05, + "loss": 0.0002, + "step": 11134 + }, + { + "epoch": 2.57, + "learning_rate": 2.8555196055773826e-05, + "loss": 0.0003, + "step": 11135 + }, + { + "epoch": 2.57, + "learning_rate": 2.855327016408597e-05, + "loss": 0.0, + "step": 11136 + }, + { + "epoch": 2.57, + "learning_rate": 2.8551344272398124e-05, + "loss": 0.0, + "step": 11137 + }, + { + "epoch": 2.57, + "learning_rate": 2.854941838071027e-05, + "loss": 0.0, + "step": 11138 + }, + { + "epoch": 2.57, + "learning_rate": 2.854749248902242e-05, + "loss": 0.0, + "step": 11139 + }, + { + "epoch": 2.57, + "learning_rate": 2.8545566597334567e-05, + "loss": 0.0, + "step": 11140 + }, + { + "epoch": 2.57, + "learning_rate": 2.8543640705646718e-05, + "loss": 0.0, + "step": 11141 + }, + { + "epoch": 2.57, + "learning_rate": 2.8541714813958865e-05, + "loss": 0.0001, + "step": 11142 + }, + { + "epoch": 2.58, + "learning_rate": 2.853978892227101e-05, + "loss": 0.0, + "step": 11143 + }, + { + "epoch": 2.58, + "learning_rate": 2.8537863030583163e-05, + "loss": 0.0001, + "step": 11144 + }, + { + "epoch": 2.58, + "learning_rate": 2.8535937138895308e-05, + "loss": 0.0001, + "step": 11145 + }, + { + "epoch": 2.58, + "learning_rate": 2.853401124720746e-05, + "loss": 0.0, + "step": 11146 + }, + { + "epoch": 2.58, + "learning_rate": 2.8532085355519606e-05, + "loss": 0.0, + "step": 11147 + }, + { + "epoch": 2.58, + "learning_rate": 2.8530159463831757e-05, + "loss": 0.0001, + "step": 11148 + }, + { + "epoch": 2.58, + "learning_rate": 2.8528233572143904e-05, + "loss": 0.0002, + "step": 11149 + }, + { + "epoch": 2.58, + "learning_rate": 2.8526307680456055e-05, + "loss": 0.0, + "step": 11150 + }, + { + "epoch": 2.58, + "learning_rate": 2.85243817887682e-05, + "loss": 0.0, + "step": 11151 + }, + { + "epoch": 2.58, + "learning_rate": 2.8522455897080347e-05, + "loss": 0.0, + "step": 11152 + }, + { + "epoch": 2.58, + "learning_rate": 2.8520530005392498e-05, + "loss": 0.0, + "step": 11153 + }, + { + "epoch": 2.58, + "learning_rate": 2.8518604113704645e-05, + "loss": 0.0, + "step": 11154 + }, + { + "epoch": 2.58, + "learning_rate": 2.8516678222016796e-05, + "loss": 0.0, + "step": 11155 + }, + { + "epoch": 2.58, + "learning_rate": 2.8514752330328943e-05, + "loss": 0.0, + "step": 11156 + }, + { + "epoch": 2.58, + "learning_rate": 2.8512826438641094e-05, + "loss": 0.0, + "step": 11157 + }, + { + "epoch": 2.58, + "learning_rate": 2.851090054695324e-05, + "loss": 0.0, + "step": 11158 + }, + { + "epoch": 2.58, + "learning_rate": 2.8508974655265393e-05, + "loss": 0.0001, + "step": 11159 + }, + { + "epoch": 2.58, + "learning_rate": 2.8507048763577537e-05, + "loss": 0.0001, + "step": 11160 + }, + { + "epoch": 2.58, + "learning_rate": 2.8505122871889688e-05, + "loss": 0.0, + "step": 11161 + }, + { + "epoch": 2.58, + "learning_rate": 2.8503196980201835e-05, + "loss": 0.0, + "step": 11162 + }, + { + "epoch": 2.58, + "learning_rate": 2.8501271088513982e-05, + "loss": 0.0, + "step": 11163 + }, + { + "epoch": 2.58, + "learning_rate": 2.8499345196826133e-05, + "loss": 0.0001, + "step": 11164 + }, + { + "epoch": 2.58, + "learning_rate": 2.8497419305138277e-05, + "loss": 0.0, + "step": 11165 + }, + { + "epoch": 2.58, + "learning_rate": 2.849549341345043e-05, + "loss": 0.0, + "step": 11166 + }, + { + "epoch": 2.58, + "learning_rate": 2.8493567521762576e-05, + "loss": 0.0, + "step": 11167 + }, + { + "epoch": 2.58, + "learning_rate": 2.8491641630074727e-05, + "loss": 0.0, + "step": 11168 + }, + { + "epoch": 2.58, + "learning_rate": 2.8489715738386874e-05, + "loss": 0.0, + "step": 11169 + }, + { + "epoch": 2.58, + "learning_rate": 2.8487789846699025e-05, + "loss": 0.0, + "step": 11170 + }, + { + "epoch": 2.58, + "learning_rate": 2.8485863955011172e-05, + "loss": 0.0, + "step": 11171 + }, + { + "epoch": 2.58, + "learning_rate": 2.8483938063323317e-05, + "loss": 0.0, + "step": 11172 + }, + { + "epoch": 2.58, + "learning_rate": 2.848201217163547e-05, + "loss": 0.0, + "step": 11173 + }, + { + "epoch": 2.58, + "learning_rate": 2.8480086279947615e-05, + "loss": 0.0, + "step": 11174 + }, + { + "epoch": 2.58, + "learning_rate": 2.8478160388259766e-05, + "loss": 0.0002, + "step": 11175 + }, + { + "epoch": 2.58, + "learning_rate": 2.8476234496571913e-05, + "loss": 0.0, + "step": 11176 + }, + { + "epoch": 2.58, + "learning_rate": 2.8474308604884064e-05, + "loss": 0.0, + "step": 11177 + }, + { + "epoch": 2.58, + "learning_rate": 2.847238271319621e-05, + "loss": 0.0, + "step": 11178 + }, + { + "epoch": 2.58, + "learning_rate": 2.8470456821508362e-05, + "loss": 0.0, + "step": 11179 + }, + { + "epoch": 2.58, + "learning_rate": 2.846853092982051e-05, + "loss": 0.0, + "step": 11180 + }, + { + "epoch": 2.58, + "learning_rate": 2.8466605038132654e-05, + "loss": 0.0001, + "step": 11181 + }, + { + "epoch": 2.58, + "learning_rate": 2.8464679146444805e-05, + "loss": 0.0003, + "step": 11182 + }, + { + "epoch": 2.58, + "learning_rate": 2.8462753254756952e-05, + "loss": 0.0, + "step": 11183 + }, + { + "epoch": 2.58, + "learning_rate": 2.8460827363069103e-05, + "loss": 0.0, + "step": 11184 + }, + { + "epoch": 2.58, + "learning_rate": 2.845890147138125e-05, + "loss": 0.0, + "step": 11185 + }, + { + "epoch": 2.59, + "learning_rate": 2.84569755796934e-05, + "loss": 0.0, + "step": 11186 + }, + { + "epoch": 2.59, + "learning_rate": 2.8455049688005546e-05, + "loss": 0.0, + "step": 11187 + }, + { + "epoch": 2.59, + "learning_rate": 2.84531237963177e-05, + "loss": 0.0001, + "step": 11188 + }, + { + "epoch": 2.59, + "learning_rate": 2.8451197904629844e-05, + "loss": 0.0, + "step": 11189 + }, + { + "epoch": 2.59, + "learning_rate": 2.844927201294199e-05, + "loss": 0.0, + "step": 11190 + }, + { + "epoch": 2.59, + "learning_rate": 2.8447346121254142e-05, + "loss": 0.0007, + "step": 11191 + }, + { + "epoch": 2.59, + "learning_rate": 2.844542022956629e-05, + "loss": 0.0001, + "step": 11192 + }, + { + "epoch": 2.59, + "learning_rate": 2.844349433787844e-05, + "loss": 0.0, + "step": 11193 + }, + { + "epoch": 2.59, + "learning_rate": 2.8441568446190585e-05, + "loss": 0.0, + "step": 11194 + }, + { + "epoch": 2.59, + "learning_rate": 2.843964255450274e-05, + "loss": 0.0, + "step": 11195 + }, + { + "epoch": 2.59, + "learning_rate": 2.8437716662814883e-05, + "loss": 0.0, + "step": 11196 + }, + { + "epoch": 2.59, + "learning_rate": 2.8435790771127034e-05, + "loss": 0.0, + "step": 11197 + }, + { + "epoch": 2.59, + "learning_rate": 2.843386487943918e-05, + "loss": 0.0, + "step": 11198 + }, + { + "epoch": 2.59, + "learning_rate": 2.843193898775133e-05, + "loss": 0.0, + "step": 11199 + }, + { + "epoch": 2.59, + "learning_rate": 2.843001309606348e-05, + "loss": 0.0, + "step": 11200 + }, + { + "epoch": 2.59, + "learning_rate": 2.8428087204375624e-05, + "loss": 0.0, + "step": 11201 + }, + { + "epoch": 2.59, + "learning_rate": 2.8426161312687778e-05, + "loss": 0.0004, + "step": 11202 + }, + { + "epoch": 2.59, + "learning_rate": 2.8424235420999922e-05, + "loss": 0.0, + "step": 11203 + }, + { + "epoch": 2.59, + "learning_rate": 2.8422309529312073e-05, + "loss": 0.0001, + "step": 11204 + }, + { + "epoch": 2.59, + "learning_rate": 2.842038363762422e-05, + "loss": 0.0, + "step": 11205 + }, + { + "epoch": 2.59, + "learning_rate": 2.841845774593637e-05, + "loss": 0.0002, + "step": 11206 + }, + { + "epoch": 2.59, + "learning_rate": 2.841653185424852e-05, + "loss": 0.0, + "step": 11207 + }, + { + "epoch": 2.59, + "learning_rate": 2.8414605962560663e-05, + "loss": 0.0, + "step": 11208 + }, + { + "epoch": 2.59, + "learning_rate": 2.8412680070872817e-05, + "loss": 0.0002, + "step": 11209 + }, + { + "epoch": 2.59, + "learning_rate": 2.841075417918496e-05, + "loss": 0.0, + "step": 11210 + }, + { + "epoch": 2.59, + "learning_rate": 2.8408828287497112e-05, + "loss": 0.0, + "step": 11211 + }, + { + "epoch": 2.59, + "learning_rate": 2.840690239580926e-05, + "loss": 0.0, + "step": 11212 + }, + { + "epoch": 2.59, + "learning_rate": 2.840497650412141e-05, + "loss": 0.0001, + "step": 11213 + }, + { + "epoch": 2.59, + "learning_rate": 2.8403050612433558e-05, + "loss": 0.0002, + "step": 11214 + }, + { + "epoch": 2.59, + "learning_rate": 2.840112472074571e-05, + "loss": 0.0, + "step": 11215 + }, + { + "epoch": 2.59, + "learning_rate": 2.8399198829057856e-05, + "loss": 0.0, + "step": 11216 + }, + { + "epoch": 2.59, + "learning_rate": 2.8397272937370007e-05, + "loss": 0.0, + "step": 11217 + }, + { + "epoch": 2.59, + "learning_rate": 2.839534704568215e-05, + "loss": 0.0, + "step": 11218 + }, + { + "epoch": 2.59, + "learning_rate": 2.83934211539943e-05, + "loss": 0.0001, + "step": 11219 + }, + { + "epoch": 2.59, + "learning_rate": 2.839149526230645e-05, + "loss": 0.0, + "step": 11220 + }, + { + "epoch": 2.59, + "learning_rate": 2.8389569370618597e-05, + "loss": 0.0, + "step": 11221 + }, + { + "epoch": 2.59, + "learning_rate": 2.8387643478930748e-05, + "loss": 0.0, + "step": 11222 + }, + { + "epoch": 2.59, + "learning_rate": 2.8385717587242892e-05, + "loss": 0.0, + "step": 11223 + }, + { + "epoch": 2.59, + "learning_rate": 2.8383791695555046e-05, + "loss": 0.0, + "step": 11224 + }, + { + "epoch": 2.59, + "learning_rate": 2.838186580386719e-05, + "loss": 0.0, + "step": 11225 + }, + { + "epoch": 2.59, + "learning_rate": 2.837993991217934e-05, + "loss": 0.0, + "step": 11226 + }, + { + "epoch": 2.59, + "learning_rate": 2.837801402049149e-05, + "loss": 0.0001, + "step": 11227 + }, + { + "epoch": 2.59, + "learning_rate": 2.8376088128803636e-05, + "loss": 0.0, + "step": 11228 + }, + { + "epoch": 2.6, + "learning_rate": 2.8374162237115787e-05, + "loss": 0.0003, + "step": 11229 + }, + { + "epoch": 2.6, + "learning_rate": 2.837223634542793e-05, + "loss": 0.0002, + "step": 11230 + }, + { + "epoch": 2.6, + "learning_rate": 2.8370310453740085e-05, + "loss": 0.0, + "step": 11231 + }, + { + "epoch": 2.6, + "learning_rate": 2.836838456205223e-05, + "loss": 0.0, + "step": 11232 + }, + { + "epoch": 2.6, + "learning_rate": 2.836645867036438e-05, + "loss": 0.0001, + "step": 11233 + }, + { + "epoch": 2.6, + "learning_rate": 2.8364532778676528e-05, + "loss": 0.0, + "step": 11234 + }, + { + "epoch": 2.6, + "learning_rate": 2.836260688698868e-05, + "loss": 0.0, + "step": 11235 + }, + { + "epoch": 2.6, + "learning_rate": 2.8360680995300826e-05, + "loss": 0.0004, + "step": 11236 + }, + { + "epoch": 2.6, + "learning_rate": 2.835875510361297e-05, + "loss": 0.0, + "step": 11237 + }, + { + "epoch": 2.6, + "learning_rate": 2.8356829211925125e-05, + "loss": 0.0, + "step": 11238 + }, + { + "epoch": 2.6, + "learning_rate": 2.835490332023727e-05, + "loss": 0.0, + "step": 11239 + }, + { + "epoch": 2.6, + "learning_rate": 2.835297742854942e-05, + "loss": 0.0001, + "step": 11240 + }, + { + "epoch": 2.6, + "learning_rate": 2.8351051536861567e-05, + "loss": 0.0, + "step": 11241 + }, + { + "epoch": 2.6, + "learning_rate": 2.8349125645173718e-05, + "loss": 0.0, + "step": 11242 + }, + { + "epoch": 2.6, + "learning_rate": 2.8347199753485865e-05, + "loss": 0.0, + "step": 11243 + }, + { + "epoch": 2.6, + "learning_rate": 2.8345273861798016e-05, + "loss": 0.0, + "step": 11244 + }, + { + "epoch": 2.6, + "learning_rate": 2.8343347970110164e-05, + "loss": 0.0, + "step": 11245 + }, + { + "epoch": 2.6, + "learning_rate": 2.8341422078422308e-05, + "loss": 0.0, + "step": 11246 + }, + { + "epoch": 2.6, + "learning_rate": 2.833949618673446e-05, + "loss": 0.0001, + "step": 11247 + }, + { + "epoch": 2.6, + "learning_rate": 2.8337570295046606e-05, + "loss": 0.0, + "step": 11248 + }, + { + "epoch": 2.6, + "learning_rate": 2.8335644403358757e-05, + "loss": 0.0, + "step": 11249 + }, + { + "epoch": 2.6, + "learning_rate": 2.8333718511670905e-05, + "loss": 0.0, + "step": 11250 + }, + { + "epoch": 2.6, + "learning_rate": 2.8331792619983055e-05, + "loss": 0.0, + "step": 11251 + }, + { + "epoch": 2.6, + "learning_rate": 2.8329866728295203e-05, + "loss": 0.0001, + "step": 11252 + }, + { + "epoch": 2.6, + "learning_rate": 2.8327940836607354e-05, + "loss": 0.0, + "step": 11253 + }, + { + "epoch": 2.6, + "learning_rate": 2.8326014944919498e-05, + "loss": 0.0, + "step": 11254 + }, + { + "epoch": 2.6, + "learning_rate": 2.8324089053231645e-05, + "loss": 0.0, + "step": 11255 + }, + { + "epoch": 2.6, + "learning_rate": 2.8322163161543796e-05, + "loss": 0.0, + "step": 11256 + }, + { + "epoch": 2.6, + "learning_rate": 2.8320237269855944e-05, + "loss": 0.0001, + "step": 11257 + }, + { + "epoch": 2.6, + "learning_rate": 2.8318311378168095e-05, + "loss": 0.0, + "step": 11258 + }, + { + "epoch": 2.6, + "learning_rate": 2.831638548648024e-05, + "loss": 0.0002, + "step": 11259 + }, + { + "epoch": 2.6, + "learning_rate": 2.8314459594792393e-05, + "loss": 0.0002, + "step": 11260 + }, + { + "epoch": 2.6, + "learning_rate": 2.8312533703104537e-05, + "loss": 0.0, + "step": 11261 + }, + { + "epoch": 2.6, + "learning_rate": 2.8310607811416688e-05, + "loss": 0.0001, + "step": 11262 + }, + { + "epoch": 2.6, + "learning_rate": 2.8308681919728835e-05, + "loss": 0.0, + "step": 11263 + }, + { + "epoch": 2.6, + "learning_rate": 2.8306756028040983e-05, + "loss": 0.0, + "step": 11264 + }, + { + "epoch": 2.6, + "learning_rate": 2.8304830136353134e-05, + "loss": 0.0, + "step": 11265 + }, + { + "epoch": 2.6, + "learning_rate": 2.8302904244665278e-05, + "loss": 0.0, + "step": 11266 + }, + { + "epoch": 2.6, + "learning_rate": 2.8300978352977432e-05, + "loss": 0.0, + "step": 11267 + }, + { + "epoch": 2.6, + "learning_rate": 2.8299052461289576e-05, + "loss": 0.0, + "step": 11268 + }, + { + "epoch": 2.6, + "learning_rate": 2.8297126569601727e-05, + "loss": 0.0, + "step": 11269 + }, + { + "epoch": 2.6, + "learning_rate": 2.8295200677913874e-05, + "loss": 0.0, + "step": 11270 + }, + { + "epoch": 2.6, + "learning_rate": 2.8293274786226025e-05, + "loss": 0.0, + "step": 11271 + }, + { + "epoch": 2.61, + "learning_rate": 2.8291348894538173e-05, + "loss": 0.0, + "step": 11272 + }, + { + "epoch": 2.61, + "learning_rate": 2.8289423002850324e-05, + "loss": 0.0, + "step": 11273 + }, + { + "epoch": 2.61, + "learning_rate": 2.828749711116247e-05, + "loss": 0.0, + "step": 11274 + }, + { + "epoch": 2.61, + "learning_rate": 2.8285571219474615e-05, + "loss": 0.0, + "step": 11275 + }, + { + "epoch": 2.61, + "learning_rate": 2.8283645327786766e-05, + "loss": 0.0001, + "step": 11276 + }, + { + "epoch": 2.61, + "learning_rate": 2.8281719436098914e-05, + "loss": 0.0, + "step": 11277 + }, + { + "epoch": 2.61, + "learning_rate": 2.8279793544411064e-05, + "loss": 0.0, + "step": 11278 + }, + { + "epoch": 2.61, + "learning_rate": 2.8277867652723212e-05, + "loss": 0.0, + "step": 11279 + }, + { + "epoch": 2.61, + "learning_rate": 2.8275941761035363e-05, + "loss": 0.0, + "step": 11280 + }, + { + "epoch": 2.61, + "learning_rate": 2.827401586934751e-05, + "loss": 0.0001, + "step": 11281 + }, + { + "epoch": 2.61, + "learning_rate": 2.827208997765966e-05, + "loss": 0.0, + "step": 11282 + }, + { + "epoch": 2.61, + "learning_rate": 2.8270164085971805e-05, + "loss": 0.0001, + "step": 11283 + }, + { + "epoch": 2.61, + "learning_rate": 2.8268238194283953e-05, + "loss": 0.0, + "step": 11284 + }, + { + "epoch": 2.61, + "learning_rate": 2.8266312302596104e-05, + "loss": 0.0, + "step": 11285 + }, + { + "epoch": 2.61, + "learning_rate": 2.826438641090825e-05, + "loss": 0.0001, + "step": 11286 + }, + { + "epoch": 2.61, + "learning_rate": 2.8262460519220402e-05, + "loss": 0.0, + "step": 11287 + }, + { + "epoch": 2.61, + "learning_rate": 2.826053462753255e-05, + "loss": 0.0003, + "step": 11288 + }, + { + "epoch": 2.61, + "learning_rate": 2.82586087358447e-05, + "loss": 0.0003, + "step": 11289 + }, + { + "epoch": 2.61, + "learning_rate": 2.8256682844156844e-05, + "loss": 0.0, + "step": 11290 + }, + { + "epoch": 2.61, + "learning_rate": 2.8254756952469e-05, + "loss": 0.0, + "step": 11291 + }, + { + "epoch": 2.61, + "learning_rate": 2.8252831060781143e-05, + "loss": 0.0, + "step": 11292 + }, + { + "epoch": 2.61, + "learning_rate": 2.825090516909329e-05, + "loss": 0.0, + "step": 11293 + }, + { + "epoch": 2.61, + "learning_rate": 2.824897927740544e-05, + "loss": 0.0001, + "step": 11294 + }, + { + "epoch": 2.61, + "learning_rate": 2.8247053385717585e-05, + "loss": 0.0001, + "step": 11295 + }, + { + "epoch": 2.61, + "learning_rate": 2.824512749402974e-05, + "loss": 0.0, + "step": 11296 + }, + { + "epoch": 2.61, + "learning_rate": 2.8243201602341883e-05, + "loss": 0.0, + "step": 11297 + }, + { + "epoch": 2.61, + "learning_rate": 2.8241275710654038e-05, + "loss": 0.0, + "step": 11298 + }, + { + "epoch": 2.61, + "learning_rate": 2.8239349818966182e-05, + "loss": 0.0002, + "step": 11299 + }, + { + "epoch": 2.61, + "learning_rate": 2.8237423927278333e-05, + "loss": 0.0, + "step": 11300 + }, + { + "epoch": 2.61, + "learning_rate": 2.823549803559048e-05, + "loss": 0.0, + "step": 11301 + }, + { + "epoch": 2.61, + "learning_rate": 2.8233572143902624e-05, + "loss": 0.0001, + "step": 11302 + }, + { + "epoch": 2.61, + "learning_rate": 2.823164625221478e-05, + "loss": 0.0, + "step": 11303 + }, + { + "epoch": 2.61, + "learning_rate": 2.8229720360526923e-05, + "loss": 0.0, + "step": 11304 + }, + { + "epoch": 2.61, + "learning_rate": 2.8227794468839073e-05, + "loss": 0.0, + "step": 11305 + }, + { + "epoch": 2.61, + "learning_rate": 2.822586857715122e-05, + "loss": 0.0, + "step": 11306 + }, + { + "epoch": 2.61, + "learning_rate": 2.8223942685463372e-05, + "loss": 0.0, + "step": 11307 + }, + { + "epoch": 2.61, + "learning_rate": 2.822201679377552e-05, + "loss": 0.0005, + "step": 11308 + }, + { + "epoch": 2.61, + "learning_rate": 2.822009090208767e-05, + "loss": 0.0, + "step": 11309 + }, + { + "epoch": 2.61, + "learning_rate": 2.8218165010399818e-05, + "loss": 0.0, + "step": 11310 + }, + { + "epoch": 2.61, + "learning_rate": 2.821623911871196e-05, + "loss": 0.0001, + "step": 11311 + }, + { + "epoch": 2.61, + "learning_rate": 2.8214313227024113e-05, + "loss": 0.0, + "step": 11312 + }, + { + "epoch": 2.61, + "learning_rate": 2.821238733533626e-05, + "loss": 0.0, + "step": 11313 + }, + { + "epoch": 2.61, + "learning_rate": 2.821046144364841e-05, + "loss": 0.0, + "step": 11314 + }, + { + "epoch": 2.61, + "learning_rate": 2.820853555196056e-05, + "loss": 0.0001, + "step": 11315 + }, + { + "epoch": 2.62, + "learning_rate": 2.820660966027271e-05, + "loss": 0.0001, + "step": 11316 + }, + { + "epoch": 2.62, + "learning_rate": 2.8204683768584857e-05, + "loss": 0.0, + "step": 11317 + }, + { + "epoch": 2.62, + "learning_rate": 2.8202757876897008e-05, + "loss": 0.0, + "step": 11318 + }, + { + "epoch": 2.62, + "learning_rate": 2.820083198520915e-05, + "loss": 0.0, + "step": 11319 + }, + { + "epoch": 2.62, + "learning_rate": 2.81989060935213e-05, + "loss": 0.0, + "step": 11320 + }, + { + "epoch": 2.62, + "learning_rate": 2.819698020183345e-05, + "loss": 0.0, + "step": 11321 + }, + { + "epoch": 2.62, + "learning_rate": 2.8195054310145597e-05, + "loss": 0.0, + "step": 11322 + }, + { + "epoch": 2.62, + "learning_rate": 2.819312841845775e-05, + "loss": 0.0005, + "step": 11323 + }, + { + "epoch": 2.62, + "learning_rate": 2.8191202526769896e-05, + "loss": 0.0, + "step": 11324 + }, + { + "epoch": 2.62, + "learning_rate": 2.8189276635082047e-05, + "loss": 0.0001, + "step": 11325 + }, + { + "epoch": 2.62, + "learning_rate": 2.818735074339419e-05, + "loss": 0.0004, + "step": 11326 + }, + { + "epoch": 2.62, + "learning_rate": 2.8185424851706345e-05, + "loss": 0.0002, + "step": 11327 + }, + { + "epoch": 2.62, + "learning_rate": 2.818349896001849e-05, + "loss": 0.0006, + "step": 11328 + }, + { + "epoch": 2.62, + "learning_rate": 2.818157306833064e-05, + "loss": 0.0001, + "step": 11329 + }, + { + "epoch": 2.62, + "learning_rate": 2.8179647176642787e-05, + "loss": 0.0002, + "step": 11330 + }, + { + "epoch": 2.62, + "learning_rate": 2.8177721284954935e-05, + "loss": 0.0, + "step": 11331 + }, + { + "epoch": 2.62, + "learning_rate": 2.8175795393267086e-05, + "loss": 0.0, + "step": 11332 + }, + { + "epoch": 2.62, + "learning_rate": 2.817386950157923e-05, + "loss": 0.0001, + "step": 11333 + }, + { + "epoch": 2.62, + "learning_rate": 2.8171943609891384e-05, + "loss": 0.0, + "step": 11334 + }, + { + "epoch": 2.62, + "learning_rate": 2.8170017718203528e-05, + "loss": 0.0, + "step": 11335 + }, + { + "epoch": 2.62, + "learning_rate": 2.816809182651568e-05, + "loss": 0.0002, + "step": 11336 + }, + { + "epoch": 2.62, + "learning_rate": 2.8166165934827827e-05, + "loss": 0.0, + "step": 11337 + }, + { + "epoch": 2.62, + "learning_rate": 2.8164240043139977e-05, + "loss": 0.0, + "step": 11338 + }, + { + "epoch": 2.62, + "learning_rate": 2.8162314151452125e-05, + "loss": 0.0, + "step": 11339 + }, + { + "epoch": 2.62, + "learning_rate": 2.816038825976427e-05, + "loss": 0.0, + "step": 11340 + }, + { + "epoch": 2.62, + "learning_rate": 2.815846236807642e-05, + "loss": 0.0, + "step": 11341 + }, + { + "epoch": 2.62, + "learning_rate": 2.8156536476388567e-05, + "loss": 0.0, + "step": 11342 + }, + { + "epoch": 2.62, + "learning_rate": 2.8154610584700718e-05, + "loss": 0.0, + "step": 11343 + }, + { + "epoch": 2.62, + "learning_rate": 2.8152684693012866e-05, + "loss": 0.0, + "step": 11344 + }, + { + "epoch": 2.62, + "learning_rate": 2.8150758801325017e-05, + "loss": 0.0, + "step": 11345 + }, + { + "epoch": 2.62, + "learning_rate": 2.8148832909637164e-05, + "loss": 0.0001, + "step": 11346 + }, + { + "epoch": 2.62, + "learning_rate": 2.8146907017949315e-05, + "loss": 0.0, + "step": 11347 + }, + { + "epoch": 2.62, + "learning_rate": 2.814498112626146e-05, + "loss": 0.0, + "step": 11348 + }, + { + "epoch": 2.62, + "learning_rate": 2.8143055234573606e-05, + "loss": 0.0, + "step": 11349 + }, + { + "epoch": 2.62, + "learning_rate": 2.8141129342885757e-05, + "loss": 0.0, + "step": 11350 + }, + { + "epoch": 2.62, + "learning_rate": 2.8139203451197905e-05, + "loss": 0.0, + "step": 11351 + }, + { + "epoch": 2.62, + "learning_rate": 2.8137277559510056e-05, + "loss": 0.0, + "step": 11352 + }, + { + "epoch": 2.62, + "learning_rate": 2.8135351667822203e-05, + "loss": 0.0, + "step": 11353 + }, + { + "epoch": 2.62, + "learning_rate": 2.8133425776134354e-05, + "loss": 0.0, + "step": 11354 + }, + { + "epoch": 2.62, + "learning_rate": 2.8131499884446498e-05, + "loss": 0.0, + "step": 11355 + }, + { + "epoch": 2.62, + "learning_rate": 2.8129573992758652e-05, + "loss": 0.0, + "step": 11356 + }, + { + "epoch": 2.62, + "learning_rate": 2.8127648101070796e-05, + "loss": 0.0001, + "step": 11357 + }, + { + "epoch": 2.62, + "learning_rate": 2.8125722209382944e-05, + "loss": 0.0, + "step": 11358 + }, + { + "epoch": 2.63, + "learning_rate": 2.8123796317695095e-05, + "loss": 0.0, + "step": 11359 + }, + { + "epoch": 2.63, + "learning_rate": 2.8121870426007242e-05, + "loss": 0.0, + "step": 11360 + }, + { + "epoch": 2.63, + "learning_rate": 2.8119944534319393e-05, + "loss": 0.0, + "step": 11361 + }, + { + "epoch": 2.63, + "learning_rate": 2.8118018642631537e-05, + "loss": 0.0, + "step": 11362 + }, + { + "epoch": 2.63, + "learning_rate": 2.811609275094369e-05, + "loss": 0.0, + "step": 11363 + }, + { + "epoch": 2.63, + "learning_rate": 2.8114166859255836e-05, + "loss": 0.0, + "step": 11364 + }, + { + "epoch": 2.63, + "learning_rate": 2.8112240967567986e-05, + "loss": 0.0, + "step": 11365 + }, + { + "epoch": 2.63, + "learning_rate": 2.8110315075880134e-05, + "loss": 0.0, + "step": 11366 + }, + { + "epoch": 2.63, + "learning_rate": 2.810838918419228e-05, + "loss": 0.0001, + "step": 11367 + }, + { + "epoch": 2.63, + "learning_rate": 2.8106463292504432e-05, + "loss": 0.0, + "step": 11368 + }, + { + "epoch": 2.63, + "learning_rate": 2.8104537400816576e-05, + "loss": 0.0001, + "step": 11369 + }, + { + "epoch": 2.63, + "learning_rate": 2.810261150912873e-05, + "loss": 0.0, + "step": 11370 + }, + { + "epoch": 2.63, + "learning_rate": 2.8100685617440875e-05, + "loss": 0.0, + "step": 11371 + }, + { + "epoch": 2.63, + "learning_rate": 2.8098759725753026e-05, + "loss": 0.0, + "step": 11372 + }, + { + "epoch": 2.63, + "learning_rate": 2.8096833834065173e-05, + "loss": 0.0, + "step": 11373 + }, + { + "epoch": 2.63, + "learning_rate": 2.8094907942377324e-05, + "loss": 0.0001, + "step": 11374 + }, + { + "epoch": 2.63, + "learning_rate": 2.809298205068947e-05, + "loss": 0.0, + "step": 11375 + }, + { + "epoch": 2.63, + "learning_rate": 2.8091056159001615e-05, + "loss": 0.0001, + "step": 11376 + }, + { + "epoch": 2.63, + "learning_rate": 2.8089130267313766e-05, + "loss": 0.0, + "step": 11377 + }, + { + "epoch": 2.63, + "learning_rate": 2.8087204375625914e-05, + "loss": 0.0003, + "step": 11378 + }, + { + "epoch": 2.63, + "learning_rate": 2.8085278483938065e-05, + "loss": 0.0002, + "step": 11379 + }, + { + "epoch": 2.63, + "learning_rate": 2.8083352592250212e-05, + "loss": 0.0, + "step": 11380 + }, + { + "epoch": 2.63, + "learning_rate": 2.8081426700562363e-05, + "loss": 0.0, + "step": 11381 + }, + { + "epoch": 2.63, + "learning_rate": 2.807950080887451e-05, + "loss": 0.0, + "step": 11382 + }, + { + "epoch": 2.63, + "learning_rate": 2.807757491718666e-05, + "loss": 0.0, + "step": 11383 + }, + { + "epoch": 2.63, + "learning_rate": 2.8075649025498805e-05, + "loss": 0.0, + "step": 11384 + }, + { + "epoch": 2.63, + "learning_rate": 2.807372313381096e-05, + "loss": 0.0, + "step": 11385 + }, + { + "epoch": 2.63, + "learning_rate": 2.8071797242123104e-05, + "loss": 0.0, + "step": 11386 + }, + { + "epoch": 2.63, + "learning_rate": 2.806987135043525e-05, + "loss": 0.0, + "step": 11387 + }, + { + "epoch": 2.63, + "learning_rate": 2.8067945458747402e-05, + "loss": 0.0, + "step": 11388 + }, + { + "epoch": 2.63, + "learning_rate": 2.806601956705955e-05, + "loss": 0.0001, + "step": 11389 + }, + { + "epoch": 2.63, + "learning_rate": 2.80640936753717e-05, + "loss": 0.0, + "step": 11390 + }, + { + "epoch": 2.63, + "learning_rate": 2.8062167783683845e-05, + "loss": 0.0, + "step": 11391 + }, + { + "epoch": 2.63, + "learning_rate": 2.8060241891996e-05, + "loss": 0.0, + "step": 11392 + }, + { + "epoch": 2.63, + "learning_rate": 2.8058316000308143e-05, + "loss": 0.0, + "step": 11393 + }, + { + "epoch": 2.63, + "learning_rate": 2.8056390108620294e-05, + "loss": 0.0, + "step": 11394 + }, + { + "epoch": 2.63, + "learning_rate": 2.805446421693244e-05, + "loss": 0.0, + "step": 11395 + }, + { + "epoch": 2.63, + "learning_rate": 2.805253832524459e-05, + "loss": 0.0002, + "step": 11396 + }, + { + "epoch": 2.63, + "learning_rate": 2.805061243355674e-05, + "loss": 0.0, + "step": 11397 + }, + { + "epoch": 2.63, + "learning_rate": 2.8048686541868884e-05, + "loss": 0.0, + "step": 11398 + }, + { + "epoch": 2.63, + "learning_rate": 2.8046760650181038e-05, + "loss": 0.0, + "step": 11399 + }, + { + "epoch": 2.63, + "learning_rate": 2.8044834758493182e-05, + "loss": 0.0, + "step": 11400 + }, + { + "epoch": 2.63, + "learning_rate": 2.8042908866805333e-05, + "loss": 0.0, + "step": 11401 + }, + { + "epoch": 2.64, + "learning_rate": 2.804098297511748e-05, + "loss": 0.0, + "step": 11402 + }, + { + "epoch": 2.64, + "learning_rate": 2.803905708342963e-05, + "loss": 0.0001, + "step": 11403 + }, + { + "epoch": 2.64, + "learning_rate": 2.803713119174178e-05, + "loss": 0.0001, + "step": 11404 + }, + { + "epoch": 2.64, + "learning_rate": 2.8035205300053923e-05, + "loss": 0.0, + "step": 11405 + }, + { + "epoch": 2.64, + "learning_rate": 2.8033279408366077e-05, + "loss": 0.0001, + "step": 11406 + }, + { + "epoch": 2.64, + "learning_rate": 2.803135351667822e-05, + "loss": 0.0, + "step": 11407 + }, + { + "epoch": 2.64, + "learning_rate": 2.8029427624990372e-05, + "loss": 0.0, + "step": 11408 + }, + { + "epoch": 2.64, + "learning_rate": 2.802750173330252e-05, + "loss": 0.0, + "step": 11409 + }, + { + "epoch": 2.64, + "learning_rate": 2.802557584161467e-05, + "loss": 0.0, + "step": 11410 + }, + { + "epoch": 2.64, + "learning_rate": 2.8023649949926818e-05, + "loss": 0.0, + "step": 11411 + }, + { + "epoch": 2.64, + "learning_rate": 2.802172405823897e-05, + "loss": 0.0, + "step": 11412 + }, + { + "epoch": 2.64, + "learning_rate": 2.8019798166551113e-05, + "loss": 0.0, + "step": 11413 + }, + { + "epoch": 2.64, + "learning_rate": 2.801787227486326e-05, + "loss": 0.0, + "step": 11414 + }, + { + "epoch": 2.64, + "learning_rate": 2.801594638317541e-05, + "loss": 0.0, + "step": 11415 + }, + { + "epoch": 2.64, + "learning_rate": 2.801402049148756e-05, + "loss": 0.0, + "step": 11416 + }, + { + "epoch": 2.64, + "learning_rate": 2.801209459979971e-05, + "loss": 0.0, + "step": 11417 + }, + { + "epoch": 2.64, + "learning_rate": 2.8010168708111857e-05, + "loss": 0.0, + "step": 11418 + }, + { + "epoch": 2.64, + "learning_rate": 2.8008242816424008e-05, + "loss": 0.0, + "step": 11419 + }, + { + "epoch": 2.64, + "learning_rate": 2.8006316924736152e-05, + "loss": 0.0, + "step": 11420 + }, + { + "epoch": 2.64, + "learning_rate": 2.8004391033048306e-05, + "loss": 0.0, + "step": 11421 + }, + { + "epoch": 2.64, + "learning_rate": 2.800246514136045e-05, + "loss": 0.0, + "step": 11422 + }, + { + "epoch": 2.64, + "learning_rate": 2.8000539249672598e-05, + "loss": 0.0001, + "step": 11423 + }, + { + "epoch": 2.64, + "learning_rate": 2.799861335798475e-05, + "loss": 0.0, + "step": 11424 + }, + { + "epoch": 2.64, + "learning_rate": 2.7996687466296896e-05, + "loss": 0.0003, + "step": 11425 + }, + { + "epoch": 2.64, + "learning_rate": 2.7994761574609047e-05, + "loss": 0.0001, + "step": 11426 + }, + { + "epoch": 2.64, + "learning_rate": 2.799283568292119e-05, + "loss": 0.0, + "step": 11427 + }, + { + "epoch": 2.64, + "learning_rate": 2.7990909791233345e-05, + "loss": 0.0, + "step": 11428 + }, + { + "epoch": 2.64, + "learning_rate": 2.798898389954549e-05, + "loss": 0.0, + "step": 11429 + }, + { + "epoch": 2.64, + "learning_rate": 2.798705800785764e-05, + "loss": 0.0, + "step": 11430 + }, + { + "epoch": 2.64, + "learning_rate": 2.7985132116169788e-05, + "loss": 0.0, + "step": 11431 + }, + { + "epoch": 2.64, + "learning_rate": 2.7983206224481935e-05, + "loss": 0.0, + "step": 11432 + }, + { + "epoch": 2.64, + "learning_rate": 2.7981280332794086e-05, + "loss": 0.0, + "step": 11433 + }, + { + "epoch": 2.64, + "learning_rate": 2.797935444110623e-05, + "loss": 0.0, + "step": 11434 + }, + { + "epoch": 2.64, + "learning_rate": 2.7977428549418384e-05, + "loss": 0.0, + "step": 11435 + }, + { + "epoch": 2.64, + "learning_rate": 2.797550265773053e-05, + "loss": 0.0, + "step": 11436 + }, + { + "epoch": 2.64, + "learning_rate": 2.797357676604268e-05, + "loss": 0.0, + "step": 11437 + }, + { + "epoch": 2.64, + "learning_rate": 2.7971650874354827e-05, + "loss": 0.0, + "step": 11438 + }, + { + "epoch": 2.64, + "learning_rate": 2.7969724982666978e-05, + "loss": 0.0, + "step": 11439 + }, + { + "epoch": 2.64, + "learning_rate": 2.7967799090979125e-05, + "loss": 0.0004, + "step": 11440 + }, + { + "epoch": 2.64, + "learning_rate": 2.7965873199291276e-05, + "loss": 0.0, + "step": 11441 + }, + { + "epoch": 2.64, + "learning_rate": 2.7963947307603424e-05, + "loss": 0.0, + "step": 11442 + }, + { + "epoch": 2.64, + "learning_rate": 2.7962021415915568e-05, + "loss": 0.0, + "step": 11443 + }, + { + "epoch": 2.64, + "learning_rate": 2.796009552422772e-05, + "loss": 0.0, + "step": 11444 + }, + { + "epoch": 2.65, + "learning_rate": 2.7958169632539866e-05, + "loss": 0.0, + "step": 11445 + }, + { + "epoch": 2.65, + "learning_rate": 2.7956243740852017e-05, + "loss": 0.0, + "step": 11446 + }, + { + "epoch": 2.65, + "learning_rate": 2.7954317849164164e-05, + "loss": 0.0001, + "step": 11447 + }, + { + "epoch": 2.65, + "learning_rate": 2.7952391957476315e-05, + "loss": 0.0, + "step": 11448 + }, + { + "epoch": 2.65, + "learning_rate": 2.795046606578846e-05, + "loss": 0.0, + "step": 11449 + }, + { + "epoch": 2.65, + "learning_rate": 2.7948540174100614e-05, + "loss": 0.0, + "step": 11450 + }, + { + "epoch": 2.65, + "learning_rate": 2.7946614282412758e-05, + "loss": 0.0001, + "step": 11451 + }, + { + "epoch": 2.65, + "learning_rate": 2.7944688390724905e-05, + "loss": 0.0001, + "step": 11452 + }, + { + "epoch": 2.65, + "learning_rate": 2.7942762499037056e-05, + "loss": 0.0, + "step": 11453 + }, + { + "epoch": 2.65, + "learning_rate": 2.7940836607349203e-05, + "loss": 0.0, + "step": 11454 + }, + { + "epoch": 2.65, + "learning_rate": 2.7938910715661354e-05, + "loss": 0.0, + "step": 11455 + }, + { + "epoch": 2.65, + "learning_rate": 2.79369848239735e-05, + "loss": 0.0, + "step": 11456 + }, + { + "epoch": 2.65, + "learning_rate": 2.7935058932285653e-05, + "loss": 0.0, + "step": 11457 + }, + { + "epoch": 2.65, + "learning_rate": 2.7933133040597797e-05, + "loss": 0.0, + "step": 11458 + }, + { + "epoch": 2.65, + "learning_rate": 2.7931207148909948e-05, + "loss": 0.0, + "step": 11459 + }, + { + "epoch": 2.65, + "learning_rate": 2.7929281257222095e-05, + "loss": 0.0002, + "step": 11460 + }, + { + "epoch": 2.65, + "learning_rate": 2.7927355365534243e-05, + "loss": 0.0, + "step": 11461 + }, + { + "epoch": 2.65, + "learning_rate": 2.7925429473846393e-05, + "loss": 0.0, + "step": 11462 + }, + { + "epoch": 2.65, + "learning_rate": 2.7923503582158538e-05, + "loss": 0.0, + "step": 11463 + }, + { + "epoch": 2.65, + "learning_rate": 2.7921577690470692e-05, + "loss": 0.0, + "step": 11464 + }, + { + "epoch": 2.65, + "learning_rate": 2.7919651798782836e-05, + "loss": 0.0, + "step": 11465 + }, + { + "epoch": 2.65, + "learning_rate": 2.7917725907094987e-05, + "loss": 0.0, + "step": 11466 + }, + { + "epoch": 2.65, + "learning_rate": 2.7915800015407134e-05, + "loss": 0.0, + "step": 11467 + }, + { + "epoch": 2.65, + "learning_rate": 2.7913874123719285e-05, + "loss": 0.0, + "step": 11468 + }, + { + "epoch": 2.65, + "learning_rate": 2.7911948232031433e-05, + "loss": 0.0003, + "step": 11469 + }, + { + "epoch": 2.65, + "learning_rate": 2.7910022340343577e-05, + "loss": 0.0001, + "step": 11470 + }, + { + "epoch": 2.65, + "learning_rate": 2.790809644865573e-05, + "loss": 0.0, + "step": 11471 + }, + { + "epoch": 2.65, + "learning_rate": 2.7906170556967875e-05, + "loss": 0.0, + "step": 11472 + }, + { + "epoch": 2.65, + "learning_rate": 2.7904244665280026e-05, + "loss": 0.0, + "step": 11473 + }, + { + "epoch": 2.65, + "learning_rate": 2.7902318773592173e-05, + "loss": 0.0, + "step": 11474 + }, + { + "epoch": 2.65, + "learning_rate": 2.7900392881904324e-05, + "loss": 0.0, + "step": 11475 + }, + { + "epoch": 2.65, + "learning_rate": 2.789846699021647e-05, + "loss": 0.0, + "step": 11476 + }, + { + "epoch": 2.65, + "learning_rate": 2.7896541098528623e-05, + "loss": 0.0, + "step": 11477 + }, + { + "epoch": 2.65, + "learning_rate": 2.789461520684077e-05, + "loss": 0.0001, + "step": 11478 + }, + { + "epoch": 2.65, + "learning_rate": 2.7892689315152914e-05, + "loss": 0.0002, + "step": 11479 + }, + { + "epoch": 2.65, + "learning_rate": 2.7890763423465065e-05, + "loss": 0.0, + "step": 11480 + }, + { + "epoch": 2.65, + "learning_rate": 2.7888837531777212e-05, + "loss": 0.0001, + "step": 11481 + }, + { + "epoch": 2.65, + "learning_rate": 2.7886911640089363e-05, + "loss": 0.0, + "step": 11482 + }, + { + "epoch": 2.65, + "learning_rate": 2.788498574840151e-05, + "loss": 0.0, + "step": 11483 + }, + { + "epoch": 2.65, + "learning_rate": 2.788305985671366e-05, + "loss": 0.0, + "step": 11484 + }, + { + "epoch": 2.65, + "learning_rate": 2.7881133965025806e-05, + "loss": 0.0, + "step": 11485 + }, + { + "epoch": 2.65, + "learning_rate": 2.787920807333796e-05, + "loss": 0.0, + "step": 11486 + }, + { + "epoch": 2.65, + "learning_rate": 2.7877282181650104e-05, + "loss": 0.0, + "step": 11487 + }, + { + "epoch": 2.65, + "learning_rate": 2.787535628996225e-05, + "loss": 0.0001, + "step": 11488 + }, + { + "epoch": 2.66, + "learning_rate": 2.7873430398274402e-05, + "loss": 0.0, + "step": 11489 + }, + { + "epoch": 2.66, + "learning_rate": 2.787150450658655e-05, + "loss": 0.0, + "step": 11490 + }, + { + "epoch": 2.66, + "learning_rate": 2.78695786148987e-05, + "loss": 0.0005, + "step": 11491 + }, + { + "epoch": 2.66, + "learning_rate": 2.7867652723210845e-05, + "loss": 0.0002, + "step": 11492 + }, + { + "epoch": 2.66, + "learning_rate": 2.7865726831523e-05, + "loss": 0.0008, + "step": 11493 + }, + { + "epoch": 2.66, + "learning_rate": 2.7863800939835143e-05, + "loss": 0.0001, + "step": 11494 + }, + { + "epoch": 2.66, + "learning_rate": 2.7861875048147294e-05, + "loss": 0.0, + "step": 11495 + }, + { + "epoch": 2.66, + "learning_rate": 2.785994915645944e-05, + "loss": 0.0, + "step": 11496 + }, + { + "epoch": 2.66, + "learning_rate": 2.7858023264771592e-05, + "loss": 0.0, + "step": 11497 + }, + { + "epoch": 2.66, + "learning_rate": 2.785609737308374e-05, + "loss": 0.0, + "step": 11498 + }, + { + "epoch": 2.66, + "learning_rate": 2.7854171481395884e-05, + "loss": 0.0, + "step": 11499 + }, + { + "epoch": 2.66, + "learning_rate": 2.7852245589708038e-05, + "loss": 0.0001, + "step": 11500 + }, + { + "epoch": 2.66, + "learning_rate": 2.7850319698020182e-05, + "loss": 0.0, + "step": 11501 + }, + { + "epoch": 2.66, + "learning_rate": 2.7848393806332333e-05, + "loss": 0.0, + "step": 11502 + }, + { + "epoch": 2.66, + "learning_rate": 2.784646791464448e-05, + "loss": 0.0005, + "step": 11503 + }, + { + "epoch": 2.66, + "learning_rate": 2.784454202295663e-05, + "loss": 0.0001, + "step": 11504 + }, + { + "epoch": 2.66, + "learning_rate": 2.784261613126878e-05, + "loss": 0.0, + "step": 11505 + }, + { + "epoch": 2.66, + "learning_rate": 2.784069023958093e-05, + "loss": 0.0, + "step": 11506 + }, + { + "epoch": 2.66, + "learning_rate": 2.7838764347893077e-05, + "loss": 0.0, + "step": 11507 + }, + { + "epoch": 2.66, + "learning_rate": 2.783683845620522e-05, + "loss": 0.0, + "step": 11508 + }, + { + "epoch": 2.66, + "learning_rate": 2.7834912564517372e-05, + "loss": 0.0, + "step": 11509 + }, + { + "epoch": 2.66, + "learning_rate": 2.783298667282952e-05, + "loss": 0.0002, + "step": 11510 + }, + { + "epoch": 2.66, + "learning_rate": 2.783106078114167e-05, + "loss": 0.0, + "step": 11511 + }, + { + "epoch": 2.66, + "learning_rate": 2.7829134889453818e-05, + "loss": 0.0, + "step": 11512 + }, + { + "epoch": 2.66, + "learning_rate": 2.782720899776597e-05, + "loss": 0.0, + "step": 11513 + }, + { + "epoch": 2.66, + "learning_rate": 2.7825283106078116e-05, + "loss": 0.0, + "step": 11514 + }, + { + "epoch": 2.66, + "learning_rate": 2.7823357214390267e-05, + "loss": 0.0, + "step": 11515 + }, + { + "epoch": 2.66, + "learning_rate": 2.782143132270241e-05, + "loss": 0.0, + "step": 11516 + }, + { + "epoch": 2.66, + "learning_rate": 2.781950543101456e-05, + "loss": 0.0, + "step": 11517 + }, + { + "epoch": 2.66, + "learning_rate": 2.781757953932671e-05, + "loss": 0.0, + "step": 11518 + }, + { + "epoch": 2.66, + "learning_rate": 2.7815653647638857e-05, + "loss": 0.0, + "step": 11519 + }, + { + "epoch": 2.66, + "learning_rate": 2.7813727755951008e-05, + "loss": 0.0, + "step": 11520 + }, + { + "epoch": 2.66, + "learning_rate": 2.7811801864263156e-05, + "loss": 0.0, + "step": 11521 + }, + { + "epoch": 2.66, + "learning_rate": 2.7809875972575306e-05, + "loss": 0.0001, + "step": 11522 + }, + { + "epoch": 2.66, + "learning_rate": 2.780795008088745e-05, + "loss": 0.0, + "step": 11523 + }, + { + "epoch": 2.66, + "learning_rate": 2.7806024189199605e-05, + "loss": 0.0, + "step": 11524 + }, + { + "epoch": 2.66, + "learning_rate": 2.780409829751175e-05, + "loss": 0.0, + "step": 11525 + }, + { + "epoch": 2.66, + "learning_rate": 2.7802172405823896e-05, + "loss": 0.0001, + "step": 11526 + }, + { + "epoch": 2.66, + "learning_rate": 2.7800246514136047e-05, + "loss": 0.0, + "step": 11527 + }, + { + "epoch": 2.66, + "learning_rate": 2.779832062244819e-05, + "loss": 0.0, + "step": 11528 + }, + { + "epoch": 2.66, + "learning_rate": 2.7796394730760346e-05, + "loss": 0.0, + "step": 11529 + }, + { + "epoch": 2.66, + "learning_rate": 2.779446883907249e-05, + "loss": 0.0, + "step": 11530 + }, + { + "epoch": 2.66, + "learning_rate": 2.779254294738464e-05, + "loss": 0.0, + "step": 11531 + }, + { + "epoch": 2.67, + "learning_rate": 2.7790617055696788e-05, + "loss": 0.0, + "step": 11532 + }, + { + "epoch": 2.67, + "learning_rate": 2.778869116400894e-05, + "loss": 0.0, + "step": 11533 + }, + { + "epoch": 2.67, + "learning_rate": 2.7786765272321086e-05, + "loss": 0.0, + "step": 11534 + }, + { + "epoch": 2.67, + "learning_rate": 2.778483938063323e-05, + "loss": 0.0004, + "step": 11535 + }, + { + "epoch": 2.67, + "learning_rate": 2.7782913488945385e-05, + "loss": 0.0, + "step": 11536 + }, + { + "epoch": 2.67, + "learning_rate": 2.778098759725753e-05, + "loss": 0.0, + "step": 11537 + }, + { + "epoch": 2.67, + "learning_rate": 2.777906170556968e-05, + "loss": 0.0, + "step": 11538 + }, + { + "epoch": 2.67, + "learning_rate": 2.7777135813881827e-05, + "loss": 0.0, + "step": 11539 + }, + { + "epoch": 2.67, + "learning_rate": 2.7775209922193978e-05, + "loss": 0.0, + "step": 11540 + }, + { + "epoch": 2.67, + "learning_rate": 2.7773284030506125e-05, + "loss": 0.0, + "step": 11541 + }, + { + "epoch": 2.67, + "learning_rate": 2.7771358138818276e-05, + "loss": 0.0, + "step": 11542 + }, + { + "epoch": 2.67, + "learning_rate": 2.7769432247130424e-05, + "loss": 0.0, + "step": 11543 + }, + { + "epoch": 2.67, + "learning_rate": 2.7767506355442568e-05, + "loss": 0.0001, + "step": 11544 + }, + { + "epoch": 2.67, + "learning_rate": 2.776558046375472e-05, + "loss": 0.0, + "step": 11545 + }, + { + "epoch": 2.67, + "learning_rate": 2.7763654572066866e-05, + "loss": 0.0001, + "step": 11546 + }, + { + "epoch": 2.67, + "learning_rate": 2.7761728680379017e-05, + "loss": 0.0, + "step": 11547 + }, + { + "epoch": 2.67, + "learning_rate": 2.7759802788691165e-05, + "loss": 0.0, + "step": 11548 + }, + { + "epoch": 2.67, + "learning_rate": 2.7757876897003315e-05, + "loss": 0.0001, + "step": 11549 + }, + { + "epoch": 2.67, + "learning_rate": 2.7755951005315463e-05, + "loss": 0.0, + "step": 11550 + }, + { + "epoch": 2.67, + "learning_rate": 2.7754025113627614e-05, + "loss": 0.0, + "step": 11551 + }, + { + "epoch": 2.67, + "learning_rate": 2.7752099221939758e-05, + "loss": 0.0, + "step": 11552 + }, + { + "epoch": 2.67, + "learning_rate": 2.7750173330251912e-05, + "loss": 0.0, + "step": 11553 + }, + { + "epoch": 2.67, + "learning_rate": 2.7748247438564056e-05, + "loss": 0.0, + "step": 11554 + }, + { + "epoch": 2.67, + "learning_rate": 2.7746321546876204e-05, + "loss": 0.0, + "step": 11555 + }, + { + "epoch": 2.67, + "learning_rate": 2.7744395655188355e-05, + "loss": 0.0, + "step": 11556 + }, + { + "epoch": 2.67, + "learning_rate": 2.7742469763500502e-05, + "loss": 0.0, + "step": 11557 + }, + { + "epoch": 2.67, + "learning_rate": 2.7740543871812653e-05, + "loss": 0.0, + "step": 11558 + }, + { + "epoch": 2.67, + "learning_rate": 2.7738617980124797e-05, + "loss": 0.0, + "step": 11559 + }, + { + "epoch": 2.67, + "learning_rate": 2.773669208843695e-05, + "loss": 0.0, + "step": 11560 + }, + { + "epoch": 2.67, + "learning_rate": 2.7734766196749095e-05, + "loss": 0.0, + "step": 11561 + }, + { + "epoch": 2.67, + "learning_rate": 2.7732840305061246e-05, + "loss": 0.0, + "step": 11562 + }, + { + "epoch": 2.67, + "learning_rate": 2.7730914413373394e-05, + "loss": 0.0009, + "step": 11563 + }, + { + "epoch": 2.67, + "learning_rate": 2.7728988521685538e-05, + "loss": 0.0, + "step": 11564 + }, + { + "epoch": 2.67, + "learning_rate": 2.7727062629997692e-05, + "loss": 0.0, + "step": 11565 + }, + { + "epoch": 2.67, + "learning_rate": 2.7725136738309836e-05, + "loss": 0.0, + "step": 11566 + }, + { + "epoch": 2.67, + "learning_rate": 2.7723210846621987e-05, + "loss": 0.0001, + "step": 11567 + }, + { + "epoch": 2.67, + "learning_rate": 2.7721284954934134e-05, + "loss": 0.0, + "step": 11568 + }, + { + "epoch": 2.67, + "learning_rate": 2.7719359063246285e-05, + "loss": 0.0, + "step": 11569 + }, + { + "epoch": 2.67, + "learning_rate": 2.7717433171558433e-05, + "loss": 0.0, + "step": 11570 + }, + { + "epoch": 2.67, + "learning_rate": 2.7715507279870584e-05, + "loss": 0.0, + "step": 11571 + }, + { + "epoch": 2.67, + "learning_rate": 2.771358138818273e-05, + "loss": 0.0, + "step": 11572 + }, + { + "epoch": 2.67, + "learning_rate": 2.7711655496494875e-05, + "loss": 0.0, + "step": 11573 + }, + { + "epoch": 2.67, + "learning_rate": 2.7709729604807026e-05, + "loss": 0.0, + "step": 11574 + }, + { + "epoch": 2.68, + "learning_rate": 2.7707803713119174e-05, + "loss": 0.0, + "step": 11575 + }, + { + "epoch": 2.68, + "learning_rate": 2.7705877821431324e-05, + "loss": 0.0, + "step": 11576 + }, + { + "epoch": 2.68, + "learning_rate": 2.7703951929743472e-05, + "loss": 0.0, + "step": 11577 + }, + { + "epoch": 2.68, + "learning_rate": 2.7702026038055623e-05, + "loss": 0.0, + "step": 11578 + }, + { + "epoch": 2.68, + "learning_rate": 2.770010014636777e-05, + "loss": 0.0001, + "step": 11579 + }, + { + "epoch": 2.68, + "learning_rate": 2.769817425467992e-05, + "loss": 0.0, + "step": 11580 + }, + { + "epoch": 2.68, + "learning_rate": 2.7696248362992065e-05, + "loss": 0.0, + "step": 11581 + }, + { + "epoch": 2.68, + "learning_rate": 2.7694322471304213e-05, + "loss": 0.0, + "step": 11582 + }, + { + "epoch": 2.68, + "learning_rate": 2.7692396579616364e-05, + "loss": 0.0003, + "step": 11583 + }, + { + "epoch": 2.68, + "learning_rate": 2.769047068792851e-05, + "loss": 0.0, + "step": 11584 + }, + { + "epoch": 2.68, + "learning_rate": 2.7688544796240662e-05, + "loss": 0.0001, + "step": 11585 + }, + { + "epoch": 2.68, + "learning_rate": 2.768661890455281e-05, + "loss": 0.0001, + "step": 11586 + }, + { + "epoch": 2.68, + "learning_rate": 2.768469301286496e-05, + "loss": 0.0, + "step": 11587 + }, + { + "epoch": 2.68, + "learning_rate": 2.7682767121177104e-05, + "loss": 0.0, + "step": 11588 + }, + { + "epoch": 2.68, + "learning_rate": 2.768084122948926e-05, + "loss": 0.0, + "step": 11589 + }, + { + "epoch": 2.68, + "learning_rate": 2.7678915337801403e-05, + "loss": 0.0, + "step": 11590 + }, + { + "epoch": 2.68, + "learning_rate": 2.767698944611355e-05, + "loss": 0.0, + "step": 11591 + }, + { + "epoch": 2.68, + "learning_rate": 2.76750635544257e-05, + "loss": 0.0, + "step": 11592 + }, + { + "epoch": 2.68, + "learning_rate": 2.767313766273785e-05, + "loss": 0.0, + "step": 11593 + }, + { + "epoch": 2.68, + "learning_rate": 2.767121177105e-05, + "loss": 0.0, + "step": 11594 + }, + { + "epoch": 2.68, + "learning_rate": 2.7669285879362143e-05, + "loss": 0.0, + "step": 11595 + }, + { + "epoch": 2.68, + "learning_rate": 2.7667359987674298e-05, + "loss": 0.0, + "step": 11596 + }, + { + "epoch": 2.68, + "learning_rate": 2.7665434095986442e-05, + "loss": 0.0, + "step": 11597 + }, + { + "epoch": 2.68, + "learning_rate": 2.7663508204298593e-05, + "loss": 0.0, + "step": 11598 + }, + { + "epoch": 2.68, + "learning_rate": 2.766158231261074e-05, + "loss": 0.0, + "step": 11599 + }, + { + "epoch": 2.68, + "learning_rate": 2.7659656420922884e-05, + "loss": 0.0, + "step": 11600 + }, + { + "epoch": 2.68, + "learning_rate": 2.765773052923504e-05, + "loss": 0.0, + "step": 11601 + }, + { + "epoch": 2.68, + "learning_rate": 2.7655804637547183e-05, + "loss": 0.0, + "step": 11602 + }, + { + "epoch": 2.68, + "learning_rate": 2.7653878745859333e-05, + "loss": 0.0002, + "step": 11603 + }, + { + "epoch": 2.68, + "learning_rate": 2.765195285417148e-05, + "loss": 0.0, + "step": 11604 + }, + { + "epoch": 2.68, + "learning_rate": 2.7650026962483632e-05, + "loss": 0.0, + "step": 11605 + }, + { + "epoch": 2.68, + "learning_rate": 2.764810107079578e-05, + "loss": 0.0, + "step": 11606 + }, + { + "epoch": 2.68, + "learning_rate": 2.764617517910793e-05, + "loss": 0.0001, + "step": 11607 + }, + { + "epoch": 2.68, + "learning_rate": 2.7644249287420078e-05, + "loss": 0.0, + "step": 11608 + }, + { + "epoch": 2.68, + "learning_rate": 2.764232339573223e-05, + "loss": 0.0015, + "step": 11609 + }, + { + "epoch": 2.68, + "learning_rate": 2.7640397504044373e-05, + "loss": 0.0005, + "step": 11610 + }, + { + "epoch": 2.68, + "learning_rate": 2.763847161235652e-05, + "loss": 0.0, + "step": 11611 + }, + { + "epoch": 2.68, + "learning_rate": 2.763654572066867e-05, + "loss": 0.0001, + "step": 11612 + }, + { + "epoch": 2.68, + "learning_rate": 2.763461982898082e-05, + "loss": 0.0, + "step": 11613 + }, + { + "epoch": 2.68, + "learning_rate": 2.763269393729297e-05, + "loss": 0.0, + "step": 11614 + }, + { + "epoch": 2.68, + "learning_rate": 2.7630768045605117e-05, + "loss": 0.0, + "step": 11615 + }, + { + "epoch": 2.68, + "learning_rate": 2.7628842153917268e-05, + "loss": 0.0, + "step": 11616 + }, + { + "epoch": 2.68, + "learning_rate": 2.7626916262229412e-05, + "loss": 0.0, + "step": 11617 + }, + { + "epoch": 2.69, + "learning_rate": 2.7624990370541566e-05, + "loss": 0.0, + "step": 11618 + }, + { + "epoch": 2.69, + "learning_rate": 2.762306447885371e-05, + "loss": 0.0001, + "step": 11619 + }, + { + "epoch": 2.69, + "learning_rate": 2.7621138587165858e-05, + "loss": 0.0, + "step": 11620 + }, + { + "epoch": 2.69, + "learning_rate": 2.761921269547801e-05, + "loss": 0.0, + "step": 11621 + }, + { + "epoch": 2.69, + "learning_rate": 2.7617286803790156e-05, + "loss": 0.0002, + "step": 11622 + }, + { + "epoch": 2.69, + "learning_rate": 2.7615360912102307e-05, + "loss": 0.0, + "step": 11623 + }, + { + "epoch": 2.69, + "learning_rate": 2.761343502041445e-05, + "loss": 0.0001, + "step": 11624 + }, + { + "epoch": 2.69, + "learning_rate": 2.7611509128726605e-05, + "loss": 0.0001, + "step": 11625 + }, + { + "epoch": 2.69, + "learning_rate": 2.760958323703875e-05, + "loss": 0.0001, + "step": 11626 + }, + { + "epoch": 2.69, + "learning_rate": 2.76076573453509e-05, + "loss": 0.0, + "step": 11627 + }, + { + "epoch": 2.69, + "learning_rate": 2.7605731453663048e-05, + "loss": 0.0, + "step": 11628 + }, + { + "epoch": 2.69, + "learning_rate": 2.7603805561975195e-05, + "loss": 0.0, + "step": 11629 + }, + { + "epoch": 2.69, + "learning_rate": 2.7601879670287346e-05, + "loss": 0.0, + "step": 11630 + }, + { + "epoch": 2.69, + "learning_rate": 2.759995377859949e-05, + "loss": 0.0, + "step": 11631 + }, + { + "epoch": 2.69, + "learning_rate": 2.7598027886911644e-05, + "loss": 0.0, + "step": 11632 + }, + { + "epoch": 2.69, + "learning_rate": 2.7596101995223788e-05, + "loss": 0.0, + "step": 11633 + }, + { + "epoch": 2.69, + "learning_rate": 2.759417610353594e-05, + "loss": 0.0001, + "step": 11634 + }, + { + "epoch": 2.69, + "learning_rate": 2.7592250211848087e-05, + "loss": 0.0, + "step": 11635 + }, + { + "epoch": 2.69, + "learning_rate": 2.7590324320160238e-05, + "loss": 0.0, + "step": 11636 + }, + { + "epoch": 2.69, + "learning_rate": 2.7588398428472385e-05, + "loss": 0.0, + "step": 11637 + }, + { + "epoch": 2.69, + "learning_rate": 2.758647253678453e-05, + "loss": 0.0, + "step": 11638 + }, + { + "epoch": 2.69, + "learning_rate": 2.758454664509668e-05, + "loss": 0.0001, + "step": 11639 + }, + { + "epoch": 2.69, + "learning_rate": 2.7582620753408827e-05, + "loss": 0.0, + "step": 11640 + }, + { + "epoch": 2.69, + "learning_rate": 2.7580694861720978e-05, + "loss": 0.0, + "step": 11641 + }, + { + "epoch": 2.69, + "learning_rate": 2.7578768970033126e-05, + "loss": 0.0, + "step": 11642 + }, + { + "epoch": 2.69, + "learning_rate": 2.7576843078345277e-05, + "loss": 0.0001, + "step": 11643 + }, + { + "epoch": 2.69, + "learning_rate": 2.7574917186657424e-05, + "loss": 0.0, + "step": 11644 + }, + { + "epoch": 2.69, + "learning_rate": 2.7572991294969575e-05, + "loss": 0.0, + "step": 11645 + }, + { + "epoch": 2.69, + "learning_rate": 2.757106540328172e-05, + "loss": 0.0, + "step": 11646 + }, + { + "epoch": 2.69, + "learning_rate": 2.7569139511593867e-05, + "loss": 0.0, + "step": 11647 + }, + { + "epoch": 2.69, + "learning_rate": 2.7567213619906017e-05, + "loss": 0.0, + "step": 11648 + }, + { + "epoch": 2.69, + "learning_rate": 2.7565287728218165e-05, + "loss": 0.0, + "step": 11649 + }, + { + "epoch": 2.69, + "learning_rate": 2.7563361836530316e-05, + "loss": 0.0, + "step": 11650 + }, + { + "epoch": 2.69, + "learning_rate": 2.7561435944842463e-05, + "loss": 0.0, + "step": 11651 + }, + { + "epoch": 2.69, + "learning_rate": 2.7559510053154614e-05, + "loss": 0.0, + "step": 11652 + }, + { + "epoch": 2.69, + "learning_rate": 2.7557584161466758e-05, + "loss": 0.0, + "step": 11653 + }, + { + "epoch": 2.69, + "learning_rate": 2.7555658269778912e-05, + "loss": 0.0, + "step": 11654 + }, + { + "epoch": 2.69, + "learning_rate": 2.7553732378091057e-05, + "loss": 0.0, + "step": 11655 + }, + { + "epoch": 2.69, + "learning_rate": 2.7551806486403204e-05, + "loss": 0.0, + "step": 11656 + }, + { + "epoch": 2.69, + "learning_rate": 2.7549880594715355e-05, + "loss": 0.0, + "step": 11657 + }, + { + "epoch": 2.69, + "learning_rate": 2.7547954703027502e-05, + "loss": 0.0, + "step": 11658 + }, + { + "epoch": 2.69, + "learning_rate": 2.7546028811339653e-05, + "loss": 0.0001, + "step": 11659 + }, + { + "epoch": 2.69, + "learning_rate": 2.7544102919651797e-05, + "loss": 0.0, + "step": 11660 + }, + { + "epoch": 2.69, + "learning_rate": 2.754217702796395e-05, + "loss": 0.0, + "step": 11661 + }, + { + "epoch": 2.7, + "learning_rate": 2.7540251136276096e-05, + "loss": 0.0, + "step": 11662 + }, + { + "epoch": 2.7, + "learning_rate": 2.7538325244588247e-05, + "loss": 0.0, + "step": 11663 + }, + { + "epoch": 2.7, + "learning_rate": 2.7536399352900394e-05, + "loss": 0.0, + "step": 11664 + }, + { + "epoch": 2.7, + "learning_rate": 2.7534473461212545e-05, + "loss": 0.0, + "step": 11665 + }, + { + "epoch": 2.7, + "learning_rate": 2.7532547569524692e-05, + "loss": 0.0, + "step": 11666 + }, + { + "epoch": 2.7, + "learning_rate": 2.7530621677836836e-05, + "loss": 0.0002, + "step": 11667 + }, + { + "epoch": 2.7, + "learning_rate": 2.752869578614899e-05, + "loss": 0.0, + "step": 11668 + }, + { + "epoch": 2.7, + "learning_rate": 2.7526769894461135e-05, + "loss": 0.0, + "step": 11669 + }, + { + "epoch": 2.7, + "learning_rate": 2.7524844002773286e-05, + "loss": 0.0001, + "step": 11670 + }, + { + "epoch": 2.7, + "learning_rate": 2.7522918111085433e-05, + "loss": 0.0, + "step": 11671 + }, + { + "epoch": 2.7, + "learning_rate": 2.7520992219397584e-05, + "loss": 0.0002, + "step": 11672 + }, + { + "epoch": 2.7, + "learning_rate": 2.751906632770973e-05, + "loss": 0.0, + "step": 11673 + }, + { + "epoch": 2.7, + "learning_rate": 2.7517140436021882e-05, + "loss": 0.0, + "step": 11674 + }, + { + "epoch": 2.7, + "learning_rate": 2.751521454433403e-05, + "loss": 0.0, + "step": 11675 + }, + { + "epoch": 2.7, + "learning_rate": 2.7513288652646174e-05, + "loss": 0.0, + "step": 11676 + }, + { + "epoch": 2.7, + "learning_rate": 2.7511362760958325e-05, + "loss": 0.0001, + "step": 11677 + }, + { + "epoch": 2.7, + "learning_rate": 2.7509436869270472e-05, + "loss": 0.0, + "step": 11678 + }, + { + "epoch": 2.7, + "learning_rate": 2.7507510977582623e-05, + "loss": 0.0, + "step": 11679 + }, + { + "epoch": 2.7, + "learning_rate": 2.750558508589477e-05, + "loss": 0.0, + "step": 11680 + }, + { + "epoch": 2.7, + "learning_rate": 2.750365919420692e-05, + "loss": 0.0142, + "step": 11681 + }, + { + "epoch": 2.7, + "learning_rate": 2.7501733302519066e-05, + "loss": 0.0003, + "step": 11682 + }, + { + "epoch": 2.7, + "learning_rate": 2.749980741083122e-05, + "loss": 0.0, + "step": 11683 + }, + { + "epoch": 2.7, + "learning_rate": 2.7497881519143364e-05, + "loss": 0.0001, + "step": 11684 + }, + { + "epoch": 2.7, + "learning_rate": 2.749595562745551e-05, + "loss": 0.0, + "step": 11685 + }, + { + "epoch": 2.7, + "learning_rate": 2.7494029735767662e-05, + "loss": 0.0, + "step": 11686 + }, + { + "epoch": 2.7, + "learning_rate": 2.749210384407981e-05, + "loss": 0.0, + "step": 11687 + }, + { + "epoch": 2.7, + "learning_rate": 2.749017795239196e-05, + "loss": 0.0005, + "step": 11688 + }, + { + "epoch": 2.7, + "learning_rate": 2.7488252060704105e-05, + "loss": 0.0, + "step": 11689 + }, + { + "epoch": 2.7, + "learning_rate": 2.748632616901626e-05, + "loss": 0.0036, + "step": 11690 + }, + { + "epoch": 2.7, + "learning_rate": 2.7484400277328403e-05, + "loss": 0.0, + "step": 11691 + }, + { + "epoch": 2.7, + "learning_rate": 2.7482474385640554e-05, + "loss": 0.0, + "step": 11692 + }, + { + "epoch": 2.7, + "learning_rate": 2.74805484939527e-05, + "loss": 0.0, + "step": 11693 + }, + { + "epoch": 2.7, + "learning_rate": 2.747862260226485e-05, + "loss": 0.0, + "step": 11694 + }, + { + "epoch": 2.7, + "learning_rate": 2.7476696710577e-05, + "loss": 0.0, + "step": 11695 + }, + { + "epoch": 2.7, + "learning_rate": 2.7474770818889144e-05, + "loss": 0.0, + "step": 11696 + }, + { + "epoch": 2.7, + "learning_rate": 2.7472844927201298e-05, + "loss": 0.0, + "step": 11697 + }, + { + "epoch": 2.7, + "learning_rate": 2.7470919035513442e-05, + "loss": 0.0, + "step": 11698 + }, + { + "epoch": 2.7, + "learning_rate": 2.7468993143825593e-05, + "loss": 0.0002, + "step": 11699 + }, + { + "epoch": 2.7, + "learning_rate": 2.746706725213774e-05, + "loss": 0.0, + "step": 11700 + }, + { + "epoch": 2.7, + "learning_rate": 2.746514136044989e-05, + "loss": 0.0, + "step": 11701 + }, + { + "epoch": 2.7, + "learning_rate": 2.746321546876204e-05, + "loss": 0.0, + "step": 11702 + }, + { + "epoch": 2.7, + "learning_rate": 2.7461289577074183e-05, + "loss": 0.0, + "step": 11703 + }, + { + "epoch": 2.7, + "learning_rate": 2.7459363685386337e-05, + "loss": 0.0347, + "step": 11704 + }, + { + "epoch": 2.71, + "learning_rate": 2.745743779369848e-05, + "loss": 0.01, + "step": 11705 + }, + { + "epoch": 2.71, + "learning_rate": 2.7455511902010632e-05, + "loss": 0.0, + "step": 11706 + }, + { + "epoch": 2.71, + "learning_rate": 2.745358601032278e-05, + "loss": 0.0, + "step": 11707 + }, + { + "epoch": 2.71, + "learning_rate": 2.745166011863493e-05, + "loss": 0.0, + "step": 11708 + }, + { + "epoch": 2.71, + "learning_rate": 2.7449734226947078e-05, + "loss": 0.0, + "step": 11709 + }, + { + "epoch": 2.71, + "learning_rate": 2.744780833525923e-05, + "loss": 0.0, + "step": 11710 + }, + { + "epoch": 2.71, + "learning_rate": 2.7445882443571376e-05, + "loss": 0.0, + "step": 11711 + }, + { + "epoch": 2.71, + "learning_rate": 2.744395655188352e-05, + "loss": 0.0001, + "step": 11712 + }, + { + "epoch": 2.71, + "learning_rate": 2.744203066019567e-05, + "loss": 0.0, + "step": 11713 + }, + { + "epoch": 2.71, + "learning_rate": 2.744010476850782e-05, + "loss": 0.0001, + "step": 11714 + }, + { + "epoch": 2.71, + "learning_rate": 2.743817887681997e-05, + "loss": 0.0, + "step": 11715 + }, + { + "epoch": 2.71, + "learning_rate": 2.7436252985132117e-05, + "loss": 0.0, + "step": 11716 + }, + { + "epoch": 2.71, + "learning_rate": 2.7434327093444268e-05, + "loss": 0.0001, + "step": 11717 + }, + { + "epoch": 2.71, + "learning_rate": 2.7432401201756412e-05, + "loss": 0.0, + "step": 11718 + }, + { + "epoch": 2.71, + "learning_rate": 2.7430475310068566e-05, + "loss": 0.0, + "step": 11719 + }, + { + "epoch": 2.71, + "learning_rate": 2.742854941838071e-05, + "loss": 0.0, + "step": 11720 + }, + { + "epoch": 2.71, + "learning_rate": 2.742662352669286e-05, + "loss": 0.0, + "step": 11721 + }, + { + "epoch": 2.71, + "learning_rate": 2.742469763500501e-05, + "loss": 0.0, + "step": 11722 + }, + { + "epoch": 2.71, + "learning_rate": 2.7422771743317156e-05, + "loss": 0.0, + "step": 11723 + }, + { + "epoch": 2.71, + "learning_rate": 2.7420845851629307e-05, + "loss": 0.0, + "step": 11724 + }, + { + "epoch": 2.71, + "learning_rate": 2.741891995994145e-05, + "loss": 0.0, + "step": 11725 + }, + { + "epoch": 2.71, + "learning_rate": 2.7416994068253605e-05, + "loss": 0.0, + "step": 11726 + }, + { + "epoch": 2.71, + "learning_rate": 2.741506817656575e-05, + "loss": 0.0004, + "step": 11727 + }, + { + "epoch": 2.71, + "learning_rate": 2.74131422848779e-05, + "loss": 0.0, + "step": 11728 + }, + { + "epoch": 2.71, + "learning_rate": 2.7411216393190048e-05, + "loss": 0.0, + "step": 11729 + }, + { + "epoch": 2.71, + "learning_rate": 2.74092905015022e-05, + "loss": 0.0, + "step": 11730 + }, + { + "epoch": 2.71, + "learning_rate": 2.7407364609814346e-05, + "loss": 0.0, + "step": 11731 + }, + { + "epoch": 2.71, + "learning_rate": 2.740543871812649e-05, + "loss": 0.0, + "step": 11732 + }, + { + "epoch": 2.71, + "learning_rate": 2.7403512826438645e-05, + "loss": 0.0, + "step": 11733 + }, + { + "epoch": 2.71, + "learning_rate": 2.740158693475079e-05, + "loss": 0.0001, + "step": 11734 + }, + { + "epoch": 2.71, + "learning_rate": 2.739966104306294e-05, + "loss": 0.0, + "step": 11735 + }, + { + "epoch": 2.71, + "learning_rate": 2.7397735151375087e-05, + "loss": 0.0, + "step": 11736 + }, + { + "epoch": 2.71, + "learning_rate": 2.7395809259687238e-05, + "loss": 0.0001, + "step": 11737 + }, + { + "epoch": 2.71, + "learning_rate": 2.7393883367999385e-05, + "loss": 0.0, + "step": 11738 + }, + { + "epoch": 2.71, + "learning_rate": 2.7391957476311536e-05, + "loss": 0.0, + "step": 11739 + }, + { + "epoch": 2.71, + "learning_rate": 2.7390031584623684e-05, + "loss": 0.0, + "step": 11740 + }, + { + "epoch": 2.71, + "learning_rate": 2.7388105692935828e-05, + "loss": 0.0, + "step": 11741 + }, + { + "epoch": 2.71, + "learning_rate": 2.738617980124798e-05, + "loss": 0.0, + "step": 11742 + }, + { + "epoch": 2.71, + "learning_rate": 2.7384253909560126e-05, + "loss": 0.0, + "step": 11743 + }, + { + "epoch": 2.71, + "learning_rate": 2.7382328017872277e-05, + "loss": 0.0, + "step": 11744 + }, + { + "epoch": 2.71, + "learning_rate": 2.7380402126184424e-05, + "loss": 0.0001, + "step": 11745 + }, + { + "epoch": 2.71, + "learning_rate": 2.7378476234496575e-05, + "loss": 0.0, + "step": 11746 + }, + { + "epoch": 2.71, + "learning_rate": 2.7376550342808723e-05, + "loss": 0.0001, + "step": 11747 + }, + { + "epoch": 2.72, + "learning_rate": 2.7374624451120874e-05, + "loss": 0.0, + "step": 11748 + }, + { + "epoch": 2.72, + "learning_rate": 2.7372698559433018e-05, + "loss": 0.0003, + "step": 11749 + }, + { + "epoch": 2.72, + "learning_rate": 2.7370772667745165e-05, + "loss": 0.0, + "step": 11750 + }, + { + "epoch": 2.72, + "learning_rate": 2.7368846776057316e-05, + "loss": 0.0, + "step": 11751 + }, + { + "epoch": 2.72, + "learning_rate": 2.7366920884369464e-05, + "loss": 0.0, + "step": 11752 + }, + { + "epoch": 2.72, + "learning_rate": 2.7364994992681614e-05, + "loss": 0.0001, + "step": 11753 + }, + { + "epoch": 2.72, + "learning_rate": 2.736306910099376e-05, + "loss": 0.0, + "step": 11754 + }, + { + "epoch": 2.72, + "learning_rate": 2.7361143209305913e-05, + "loss": 0.0, + "step": 11755 + }, + { + "epoch": 2.72, + "learning_rate": 2.7359217317618057e-05, + "loss": 0.0, + "step": 11756 + }, + { + "epoch": 2.72, + "learning_rate": 2.7357291425930208e-05, + "loss": 0.0, + "step": 11757 + }, + { + "epoch": 2.72, + "learning_rate": 2.7355365534242355e-05, + "loss": 0.0, + "step": 11758 + }, + { + "epoch": 2.72, + "learning_rate": 2.7353439642554503e-05, + "loss": 0.0, + "step": 11759 + }, + { + "epoch": 2.72, + "learning_rate": 2.7351513750866654e-05, + "loss": 0.0001, + "step": 11760 + }, + { + "epoch": 2.72, + "learning_rate": 2.7349587859178798e-05, + "loss": 0.0001, + "step": 11761 + }, + { + "epoch": 2.72, + "learning_rate": 2.7347661967490952e-05, + "loss": 0.0, + "step": 11762 + }, + { + "epoch": 2.72, + "learning_rate": 2.7345736075803096e-05, + "loss": 0.0, + "step": 11763 + }, + { + "epoch": 2.72, + "learning_rate": 2.7343810184115247e-05, + "loss": 0.0, + "step": 11764 + }, + { + "epoch": 2.72, + "learning_rate": 2.7341884292427394e-05, + "loss": 0.0001, + "step": 11765 + }, + { + "epoch": 2.72, + "learning_rate": 2.7339958400739545e-05, + "loss": 0.0, + "step": 11766 + }, + { + "epoch": 2.72, + "learning_rate": 2.7338032509051693e-05, + "loss": 0.0, + "step": 11767 + }, + { + "epoch": 2.72, + "learning_rate": 2.7336106617363837e-05, + "loss": 0.0001, + "step": 11768 + }, + { + "epoch": 2.72, + "learning_rate": 2.733418072567599e-05, + "loss": 0.0, + "step": 11769 + }, + { + "epoch": 2.72, + "learning_rate": 2.7332254833988135e-05, + "loss": 0.0, + "step": 11770 + }, + { + "epoch": 2.72, + "learning_rate": 2.7330328942300286e-05, + "loss": 0.0001, + "step": 11771 + }, + { + "epoch": 2.72, + "learning_rate": 2.7328403050612433e-05, + "loss": 0.0, + "step": 11772 + }, + { + "epoch": 2.72, + "learning_rate": 2.7326477158924584e-05, + "loss": 0.0004, + "step": 11773 + }, + { + "epoch": 2.72, + "learning_rate": 2.7324551267236732e-05, + "loss": 0.0, + "step": 11774 + }, + { + "epoch": 2.72, + "learning_rate": 2.7322625375548883e-05, + "loss": 0.0, + "step": 11775 + }, + { + "epoch": 2.72, + "learning_rate": 2.732069948386103e-05, + "loss": 0.0, + "step": 11776 + }, + { + "epoch": 2.72, + "learning_rate": 2.731877359217318e-05, + "loss": 0.0001, + "step": 11777 + }, + { + "epoch": 2.72, + "learning_rate": 2.7316847700485325e-05, + "loss": 0.0, + "step": 11778 + }, + { + "epoch": 2.72, + "learning_rate": 2.7314921808797473e-05, + "loss": 0.0, + "step": 11779 + }, + { + "epoch": 2.72, + "learning_rate": 2.7312995917109623e-05, + "loss": 0.0, + "step": 11780 + }, + { + "epoch": 2.72, + "learning_rate": 2.731107002542177e-05, + "loss": 0.0001, + "step": 11781 + }, + { + "epoch": 2.72, + "learning_rate": 2.7309144133733922e-05, + "loss": 0.0, + "step": 11782 + }, + { + "epoch": 2.72, + "learning_rate": 2.730721824204607e-05, + "loss": 0.0, + "step": 11783 + }, + { + "epoch": 2.72, + "learning_rate": 2.730529235035822e-05, + "loss": 0.0001, + "step": 11784 + }, + { + "epoch": 2.72, + "learning_rate": 2.7303366458670364e-05, + "loss": 0.0, + "step": 11785 + }, + { + "epoch": 2.72, + "learning_rate": 2.730144056698252e-05, + "loss": 0.0, + "step": 11786 + }, + { + "epoch": 2.72, + "learning_rate": 2.7299514675294663e-05, + "loss": 0.0, + "step": 11787 + }, + { + "epoch": 2.72, + "learning_rate": 2.729758878360681e-05, + "loss": 0.0, + "step": 11788 + }, + { + "epoch": 2.72, + "learning_rate": 2.729566289191896e-05, + "loss": 0.0003, + "step": 11789 + }, + { + "epoch": 2.72, + "learning_rate": 2.7293737000231105e-05, + "loss": 0.0002, + "step": 11790 + }, + { + "epoch": 2.72, + "learning_rate": 2.729181110854326e-05, + "loss": 0.0, + "step": 11791 + }, + { + "epoch": 2.73, + "learning_rate": 2.7289885216855403e-05, + "loss": 0.0, + "step": 11792 + }, + { + "epoch": 2.73, + "learning_rate": 2.7287959325167554e-05, + "loss": 0.0001, + "step": 11793 + }, + { + "epoch": 2.73, + "learning_rate": 2.72860334334797e-05, + "loss": 0.0, + "step": 11794 + }, + { + "epoch": 2.73, + "learning_rate": 2.7284107541791852e-05, + "loss": 0.0, + "step": 11795 + }, + { + "epoch": 2.73, + "learning_rate": 2.7282181650104e-05, + "loss": 0.0, + "step": 11796 + }, + { + "epoch": 2.73, + "learning_rate": 2.7280255758416144e-05, + "loss": 0.0, + "step": 11797 + }, + { + "epoch": 2.73, + "learning_rate": 2.72783298667283e-05, + "loss": 0.0, + "step": 11798 + }, + { + "epoch": 2.73, + "learning_rate": 2.7276403975040442e-05, + "loss": 0.0, + "step": 11799 + }, + { + "epoch": 2.73, + "learning_rate": 2.7274478083352593e-05, + "loss": 0.0, + "step": 11800 + }, + { + "epoch": 2.73, + "learning_rate": 2.727255219166474e-05, + "loss": 0.0, + "step": 11801 + }, + { + "epoch": 2.73, + "learning_rate": 2.727062629997689e-05, + "loss": 0.0, + "step": 11802 + }, + { + "epoch": 2.73, + "learning_rate": 2.726870040828904e-05, + "loss": 0.0, + "step": 11803 + }, + { + "epoch": 2.73, + "learning_rate": 2.726677451660119e-05, + "loss": 0.0, + "step": 11804 + }, + { + "epoch": 2.73, + "learning_rate": 2.7264848624913337e-05, + "loss": 0.0001, + "step": 11805 + }, + { + "epoch": 2.73, + "learning_rate": 2.726292273322548e-05, + "loss": 0.0, + "step": 11806 + }, + { + "epoch": 2.73, + "learning_rate": 2.7260996841537632e-05, + "loss": 0.0, + "step": 11807 + }, + { + "epoch": 2.73, + "learning_rate": 2.725907094984978e-05, + "loss": 0.0001, + "step": 11808 + }, + { + "epoch": 2.73, + "learning_rate": 2.725714505816193e-05, + "loss": 0.0002, + "step": 11809 + }, + { + "epoch": 2.73, + "learning_rate": 2.7255219166474078e-05, + "loss": 0.0, + "step": 11810 + }, + { + "epoch": 2.73, + "learning_rate": 2.725329327478623e-05, + "loss": 0.0, + "step": 11811 + }, + { + "epoch": 2.73, + "learning_rate": 2.7251367383098377e-05, + "loss": 0.0, + "step": 11812 + }, + { + "epoch": 2.73, + "learning_rate": 2.7249441491410527e-05, + "loss": 0.0, + "step": 11813 + }, + { + "epoch": 2.73, + "learning_rate": 2.724751559972267e-05, + "loss": 0.0, + "step": 11814 + }, + { + "epoch": 2.73, + "learning_rate": 2.724558970803482e-05, + "loss": 0.0001, + "step": 11815 + }, + { + "epoch": 2.73, + "learning_rate": 2.724366381634697e-05, + "loss": 0.0, + "step": 11816 + }, + { + "epoch": 2.73, + "learning_rate": 2.7241737924659117e-05, + "loss": 0.0, + "step": 11817 + }, + { + "epoch": 2.73, + "learning_rate": 2.7239812032971268e-05, + "loss": 0.0, + "step": 11818 + }, + { + "epoch": 2.73, + "learning_rate": 2.7237886141283416e-05, + "loss": 0.0, + "step": 11819 + }, + { + "epoch": 2.73, + "learning_rate": 2.7235960249595567e-05, + "loss": 0.0, + "step": 11820 + }, + { + "epoch": 2.73, + "learning_rate": 2.723403435790771e-05, + "loss": 0.0, + "step": 11821 + }, + { + "epoch": 2.73, + "learning_rate": 2.7232108466219865e-05, + "loss": 0.0, + "step": 11822 + }, + { + "epoch": 2.73, + "learning_rate": 2.723018257453201e-05, + "loss": 0.0001, + "step": 11823 + }, + { + "epoch": 2.73, + "learning_rate": 2.7228256682844156e-05, + "loss": 0.0, + "step": 11824 + }, + { + "epoch": 2.73, + "learning_rate": 2.7226330791156307e-05, + "loss": 0.0, + "step": 11825 + }, + { + "epoch": 2.73, + "learning_rate": 2.722440489946845e-05, + "loss": 0.0, + "step": 11826 + }, + { + "epoch": 2.73, + "learning_rate": 2.7222479007780606e-05, + "loss": 0.0, + "step": 11827 + }, + { + "epoch": 2.73, + "learning_rate": 2.722055311609275e-05, + "loss": 0.0, + "step": 11828 + }, + { + "epoch": 2.73, + "learning_rate": 2.72186272244049e-05, + "loss": 0.0, + "step": 11829 + }, + { + "epoch": 2.73, + "learning_rate": 2.7216701332717048e-05, + "loss": 0.0, + "step": 11830 + }, + { + "epoch": 2.73, + "learning_rate": 2.72147754410292e-05, + "loss": 0.0, + "step": 11831 + }, + { + "epoch": 2.73, + "learning_rate": 2.7212849549341346e-05, + "loss": 0.0, + "step": 11832 + }, + { + "epoch": 2.73, + "learning_rate": 2.7210923657653497e-05, + "loss": 0.0, + "step": 11833 + }, + { + "epoch": 2.73, + "learning_rate": 2.7208997765965645e-05, + "loss": 0.0001, + "step": 11834 + }, + { + "epoch": 2.74, + "learning_rate": 2.720707187427779e-05, + "loss": 0.0, + "step": 11835 + }, + { + "epoch": 2.74, + "learning_rate": 2.720514598258994e-05, + "loss": 0.0, + "step": 11836 + }, + { + "epoch": 2.74, + "learning_rate": 2.7203220090902087e-05, + "loss": 0.0, + "step": 11837 + }, + { + "epoch": 2.74, + "learning_rate": 2.7201294199214238e-05, + "loss": 0.0, + "step": 11838 + }, + { + "epoch": 2.74, + "learning_rate": 2.7199368307526386e-05, + "loss": 0.0, + "step": 11839 + }, + { + "epoch": 2.74, + "learning_rate": 2.7197442415838536e-05, + "loss": 0.0, + "step": 11840 + }, + { + "epoch": 2.74, + "learning_rate": 2.7195516524150684e-05, + "loss": 0.0, + "step": 11841 + }, + { + "epoch": 2.74, + "learning_rate": 2.7193590632462835e-05, + "loss": 0.0001, + "step": 11842 + }, + { + "epoch": 2.74, + "learning_rate": 2.719166474077498e-05, + "loss": 0.0001, + "step": 11843 + }, + { + "epoch": 2.74, + "learning_rate": 2.7189738849087126e-05, + "loss": 0.0, + "step": 11844 + }, + { + "epoch": 2.74, + "learning_rate": 2.7187812957399277e-05, + "loss": 0.0, + "step": 11845 + }, + { + "epoch": 2.74, + "learning_rate": 2.7185887065711425e-05, + "loss": 0.0, + "step": 11846 + }, + { + "epoch": 2.74, + "learning_rate": 2.7183961174023576e-05, + "loss": 0.0, + "step": 11847 + }, + { + "epoch": 2.74, + "learning_rate": 2.7182035282335723e-05, + "loss": 0.0, + "step": 11848 + }, + { + "epoch": 2.74, + "learning_rate": 2.7180109390647874e-05, + "loss": 0.0001, + "step": 11849 + }, + { + "epoch": 2.74, + "learning_rate": 2.7178183498960018e-05, + "loss": 0.0001, + "step": 11850 + }, + { + "epoch": 2.74, + "learning_rate": 2.7176257607272172e-05, + "loss": 0.0, + "step": 11851 + }, + { + "epoch": 2.74, + "learning_rate": 2.7174331715584316e-05, + "loss": 0.0, + "step": 11852 + }, + { + "epoch": 2.74, + "learning_rate": 2.7172405823896464e-05, + "loss": 0.0, + "step": 11853 + }, + { + "epoch": 2.74, + "learning_rate": 2.7170479932208615e-05, + "loss": 0.0, + "step": 11854 + }, + { + "epoch": 2.74, + "learning_rate": 2.7168554040520762e-05, + "loss": 0.0, + "step": 11855 + }, + { + "epoch": 2.74, + "learning_rate": 2.7166628148832913e-05, + "loss": 0.0, + "step": 11856 + }, + { + "epoch": 2.74, + "learning_rate": 2.7164702257145057e-05, + "loss": 0.0, + "step": 11857 + }, + { + "epoch": 2.74, + "learning_rate": 2.716277636545721e-05, + "loss": 0.0001, + "step": 11858 + }, + { + "epoch": 2.74, + "learning_rate": 2.7160850473769355e-05, + "loss": 0.0, + "step": 11859 + }, + { + "epoch": 2.74, + "learning_rate": 2.7158924582081506e-05, + "loss": 0.0, + "step": 11860 + }, + { + "epoch": 2.74, + "learning_rate": 2.7156998690393654e-05, + "loss": 0.0, + "step": 11861 + }, + { + "epoch": 2.74, + "learning_rate": 2.7155072798705798e-05, + "loss": 0.0, + "step": 11862 + }, + { + "epoch": 2.74, + "learning_rate": 2.7153146907017952e-05, + "loss": 0.0, + "step": 11863 + }, + { + "epoch": 2.74, + "learning_rate": 2.7151221015330096e-05, + "loss": 0.0, + "step": 11864 + }, + { + "epoch": 2.74, + "learning_rate": 2.714929512364225e-05, + "loss": 0.0, + "step": 11865 + }, + { + "epoch": 2.74, + "learning_rate": 2.7147369231954395e-05, + "loss": 0.0, + "step": 11866 + }, + { + "epoch": 2.74, + "learning_rate": 2.7145443340266545e-05, + "loss": 0.0, + "step": 11867 + }, + { + "epoch": 2.74, + "learning_rate": 2.7143517448578693e-05, + "loss": 0.0, + "step": 11868 + }, + { + "epoch": 2.74, + "learning_rate": 2.7141591556890844e-05, + "loss": 0.0, + "step": 11869 + }, + { + "epoch": 2.74, + "learning_rate": 2.713966566520299e-05, + "loss": 0.0, + "step": 11870 + }, + { + "epoch": 2.74, + "learning_rate": 2.7137739773515135e-05, + "loss": 0.0006, + "step": 11871 + }, + { + "epoch": 2.74, + "learning_rate": 2.7135813881827286e-05, + "loss": 0.0, + "step": 11872 + }, + { + "epoch": 2.74, + "learning_rate": 2.7133887990139434e-05, + "loss": 0.0, + "step": 11873 + }, + { + "epoch": 2.74, + "learning_rate": 2.7131962098451585e-05, + "loss": 0.0, + "step": 11874 + }, + { + "epoch": 2.74, + "learning_rate": 2.7130036206763732e-05, + "loss": 0.0, + "step": 11875 + }, + { + "epoch": 2.74, + "learning_rate": 2.7128110315075883e-05, + "loss": 0.0, + "step": 11876 + }, + { + "epoch": 2.74, + "learning_rate": 2.712618442338803e-05, + "loss": 0.0003, + "step": 11877 + }, + { + "epoch": 2.75, + "learning_rate": 2.712425853170018e-05, + "loss": 0.0003, + "step": 11878 + }, + { + "epoch": 2.75, + "learning_rate": 2.7122332640012325e-05, + "loss": 0.0, + "step": 11879 + }, + { + "epoch": 2.75, + "learning_rate": 2.7120406748324473e-05, + "loss": 0.0, + "step": 11880 + }, + { + "epoch": 2.75, + "learning_rate": 2.7118480856636624e-05, + "loss": 0.0, + "step": 11881 + }, + { + "epoch": 2.75, + "learning_rate": 2.711655496494877e-05, + "loss": 0.0, + "step": 11882 + }, + { + "epoch": 2.75, + "learning_rate": 2.7114629073260922e-05, + "loss": 0.0, + "step": 11883 + }, + { + "epoch": 2.75, + "learning_rate": 2.711270318157307e-05, + "loss": 0.0, + "step": 11884 + }, + { + "epoch": 2.75, + "learning_rate": 2.711077728988522e-05, + "loss": 0.0, + "step": 11885 + }, + { + "epoch": 2.75, + "learning_rate": 2.7108851398197364e-05, + "loss": 0.0, + "step": 11886 + }, + { + "epoch": 2.75, + "learning_rate": 2.710692550650952e-05, + "loss": 0.0, + "step": 11887 + }, + { + "epoch": 2.75, + "learning_rate": 2.7104999614821663e-05, + "loss": 0.0, + "step": 11888 + }, + { + "epoch": 2.75, + "learning_rate": 2.7103073723133814e-05, + "loss": 0.0001, + "step": 11889 + }, + { + "epoch": 2.75, + "learning_rate": 2.710114783144596e-05, + "loss": 0.0, + "step": 11890 + }, + { + "epoch": 2.75, + "learning_rate": 2.709922193975811e-05, + "loss": 0.0, + "step": 11891 + }, + { + "epoch": 2.75, + "learning_rate": 2.709729604807026e-05, + "loss": 0.0, + "step": 11892 + }, + { + "epoch": 2.75, + "learning_rate": 2.7095370156382404e-05, + "loss": 0.0001, + "step": 11893 + }, + { + "epoch": 2.75, + "learning_rate": 2.7093444264694558e-05, + "loss": 0.0, + "step": 11894 + }, + { + "epoch": 2.75, + "learning_rate": 2.7091518373006702e-05, + "loss": 0.0, + "step": 11895 + }, + { + "epoch": 2.75, + "learning_rate": 2.7089592481318853e-05, + "loss": 0.0, + "step": 11896 + }, + { + "epoch": 2.75, + "learning_rate": 2.7087666589631e-05, + "loss": 0.0001, + "step": 11897 + }, + { + "epoch": 2.75, + "learning_rate": 2.708574069794315e-05, + "loss": 0.0, + "step": 11898 + }, + { + "epoch": 2.75, + "learning_rate": 2.70838148062553e-05, + "loss": 0.0001, + "step": 11899 + }, + { + "epoch": 2.75, + "learning_rate": 2.7081888914567443e-05, + "loss": 0.0001, + "step": 11900 + }, + { + "epoch": 2.75, + "learning_rate": 2.7079963022879597e-05, + "loss": 0.0001, + "step": 11901 + }, + { + "epoch": 2.75, + "learning_rate": 2.707803713119174e-05, + "loss": 0.0, + "step": 11902 + }, + { + "epoch": 2.75, + "learning_rate": 2.7076111239503892e-05, + "loss": 0.0001, + "step": 11903 + }, + { + "epoch": 2.75, + "learning_rate": 2.707418534781604e-05, + "loss": 0.0001, + "step": 11904 + }, + { + "epoch": 2.75, + "learning_rate": 2.707225945612819e-05, + "loss": 0.0, + "step": 11905 + }, + { + "epoch": 2.75, + "learning_rate": 2.7070333564440338e-05, + "loss": 0.0, + "step": 11906 + }, + { + "epoch": 2.75, + "learning_rate": 2.706840767275249e-05, + "loss": 0.0, + "step": 11907 + }, + { + "epoch": 2.75, + "learning_rate": 2.7066481781064633e-05, + "loss": 0.0, + "step": 11908 + }, + { + "epoch": 2.75, + "learning_rate": 2.706455588937678e-05, + "loss": 0.0, + "step": 11909 + }, + { + "epoch": 2.75, + "learning_rate": 2.706262999768893e-05, + "loss": 0.0, + "step": 11910 + }, + { + "epoch": 2.75, + "learning_rate": 2.706070410600108e-05, + "loss": 0.0001, + "step": 11911 + }, + { + "epoch": 2.75, + "learning_rate": 2.705877821431323e-05, + "loss": 0.0, + "step": 11912 + }, + { + "epoch": 2.75, + "learning_rate": 2.7056852322625377e-05, + "loss": 0.0, + "step": 11913 + }, + { + "epoch": 2.75, + "learning_rate": 2.7054926430937528e-05, + "loss": 0.0, + "step": 11914 + }, + { + "epoch": 2.75, + "learning_rate": 2.7053000539249672e-05, + "loss": 0.0, + "step": 11915 + }, + { + "epoch": 2.75, + "learning_rate": 2.7051074647561826e-05, + "loss": 0.0003, + "step": 11916 + }, + { + "epoch": 2.75, + "learning_rate": 2.704914875587397e-05, + "loss": 0.0, + "step": 11917 + }, + { + "epoch": 2.75, + "learning_rate": 2.7047222864186118e-05, + "loss": 0.0001, + "step": 11918 + }, + { + "epoch": 2.75, + "learning_rate": 2.704529697249827e-05, + "loss": 0.0, + "step": 11919 + }, + { + "epoch": 2.75, + "learning_rate": 2.7043371080810416e-05, + "loss": 0.0, + "step": 11920 + }, + { + "epoch": 2.76, + "learning_rate": 2.7041445189122567e-05, + "loss": 0.0, + "step": 11921 + }, + { + "epoch": 2.76, + "learning_rate": 2.703951929743471e-05, + "loss": 0.0, + "step": 11922 + }, + { + "epoch": 2.76, + "learning_rate": 2.7037593405746865e-05, + "loss": 0.0, + "step": 11923 + }, + { + "epoch": 2.76, + "learning_rate": 2.703566751405901e-05, + "loss": 0.0, + "step": 11924 + }, + { + "epoch": 2.76, + "learning_rate": 2.703374162237116e-05, + "loss": 0.0, + "step": 11925 + }, + { + "epoch": 2.76, + "learning_rate": 2.7031815730683308e-05, + "loss": 0.0003, + "step": 11926 + }, + { + "epoch": 2.76, + "learning_rate": 2.7029889838995455e-05, + "loss": 0.0, + "step": 11927 + }, + { + "epoch": 2.76, + "learning_rate": 2.7027963947307606e-05, + "loss": 0.0, + "step": 11928 + }, + { + "epoch": 2.76, + "learning_rate": 2.702603805561975e-05, + "loss": 0.0, + "step": 11929 + }, + { + "epoch": 2.76, + "learning_rate": 2.7024112163931904e-05, + "loss": 0.0, + "step": 11930 + }, + { + "epoch": 2.76, + "learning_rate": 2.702218627224405e-05, + "loss": 0.0, + "step": 11931 + }, + { + "epoch": 2.76, + "learning_rate": 2.70202603805562e-05, + "loss": 0.0, + "step": 11932 + }, + { + "epoch": 2.76, + "learning_rate": 2.7018334488868347e-05, + "loss": 0.0001, + "step": 11933 + }, + { + "epoch": 2.76, + "learning_rate": 2.7016408597180498e-05, + "loss": 0.0, + "step": 11934 + }, + { + "epoch": 2.76, + "learning_rate": 2.7014482705492645e-05, + "loss": 0.0, + "step": 11935 + }, + { + "epoch": 2.76, + "learning_rate": 2.701255681380479e-05, + "loss": 0.0108, + "step": 11936 + }, + { + "epoch": 2.76, + "learning_rate": 2.7010630922116943e-05, + "loss": 0.0, + "step": 11937 + }, + { + "epoch": 2.76, + "learning_rate": 2.7008705030429088e-05, + "loss": 0.0, + "step": 11938 + }, + { + "epoch": 2.76, + "learning_rate": 2.700677913874124e-05, + "loss": 0.0, + "step": 11939 + }, + { + "epoch": 2.76, + "learning_rate": 2.7004853247053386e-05, + "loss": 0.0, + "step": 11940 + }, + { + "epoch": 2.76, + "learning_rate": 2.7002927355365537e-05, + "loss": 0.0003, + "step": 11941 + }, + { + "epoch": 2.76, + "learning_rate": 2.7001001463677684e-05, + "loss": 0.0001, + "step": 11942 + }, + { + "epoch": 2.76, + "learning_rate": 2.6999075571989835e-05, + "loss": 0.0005, + "step": 11943 + }, + { + "epoch": 2.76, + "learning_rate": 2.699714968030198e-05, + "loss": 0.0, + "step": 11944 + }, + { + "epoch": 2.76, + "learning_rate": 2.6995223788614133e-05, + "loss": 0.0, + "step": 11945 + }, + { + "epoch": 2.76, + "learning_rate": 2.6993297896926277e-05, + "loss": 0.0, + "step": 11946 + }, + { + "epoch": 2.76, + "learning_rate": 2.6991372005238425e-05, + "loss": 0.0001, + "step": 11947 + }, + { + "epoch": 2.76, + "learning_rate": 2.6989446113550576e-05, + "loss": 0.0001, + "step": 11948 + }, + { + "epoch": 2.76, + "learning_rate": 2.6987520221862723e-05, + "loss": 0.0, + "step": 11949 + }, + { + "epoch": 2.76, + "learning_rate": 2.6985594330174874e-05, + "loss": 0.0, + "step": 11950 + }, + { + "epoch": 2.76, + "learning_rate": 2.6983668438487018e-05, + "loss": 0.0, + "step": 11951 + }, + { + "epoch": 2.76, + "learning_rate": 2.6981742546799173e-05, + "loss": 0.0003, + "step": 11952 + }, + { + "epoch": 2.76, + "learning_rate": 2.6979816655111317e-05, + "loss": 0.0001, + "step": 11953 + }, + { + "epoch": 2.76, + "learning_rate": 2.6977890763423467e-05, + "loss": 0.0, + "step": 11954 + }, + { + "epoch": 2.76, + "learning_rate": 2.6975964871735615e-05, + "loss": 0.0, + "step": 11955 + }, + { + "epoch": 2.76, + "learning_rate": 2.6974038980047762e-05, + "loss": 0.0, + "step": 11956 + }, + { + "epoch": 2.76, + "learning_rate": 2.6972113088359913e-05, + "loss": 0.0, + "step": 11957 + }, + { + "epoch": 2.76, + "learning_rate": 2.6970187196672057e-05, + "loss": 0.0, + "step": 11958 + }, + { + "epoch": 2.76, + "learning_rate": 2.696826130498421e-05, + "loss": 0.0, + "step": 11959 + }, + { + "epoch": 2.76, + "learning_rate": 2.6966335413296356e-05, + "loss": 0.0, + "step": 11960 + }, + { + "epoch": 2.76, + "learning_rate": 2.6964409521608507e-05, + "loss": 0.0001, + "step": 11961 + }, + { + "epoch": 2.76, + "learning_rate": 2.6962483629920654e-05, + "loss": 0.0, + "step": 11962 + }, + { + "epoch": 2.76, + "learning_rate": 2.6960557738232805e-05, + "loss": 0.0, + "step": 11963 + }, + { + "epoch": 2.76, + "learning_rate": 2.6958631846544952e-05, + "loss": 0.0, + "step": 11964 + }, + { + "epoch": 2.77, + "learning_rate": 2.6956705954857097e-05, + "loss": 0.0, + "step": 11965 + }, + { + "epoch": 2.77, + "learning_rate": 2.695478006316925e-05, + "loss": 0.0001, + "step": 11966 + }, + { + "epoch": 2.77, + "learning_rate": 2.6952854171481395e-05, + "loss": 0.0, + "step": 11967 + }, + { + "epoch": 2.77, + "learning_rate": 2.6950928279793546e-05, + "loss": 0.0, + "step": 11968 + }, + { + "epoch": 2.77, + "learning_rate": 2.6949002388105693e-05, + "loss": 0.0, + "step": 11969 + }, + { + "epoch": 2.77, + "learning_rate": 2.6947076496417844e-05, + "loss": 0.0, + "step": 11970 + }, + { + "epoch": 2.77, + "learning_rate": 2.694515060472999e-05, + "loss": 0.0001, + "step": 11971 + }, + { + "epoch": 2.77, + "learning_rate": 2.6943224713042142e-05, + "loss": 0.0, + "step": 11972 + }, + { + "epoch": 2.77, + "learning_rate": 2.694129882135429e-05, + "loss": 0.0, + "step": 11973 + }, + { + "epoch": 2.77, + "learning_rate": 2.6939372929666434e-05, + "loss": 0.0002, + "step": 11974 + }, + { + "epoch": 2.77, + "learning_rate": 2.6937447037978585e-05, + "loss": 0.0001, + "step": 11975 + }, + { + "epoch": 2.77, + "learning_rate": 2.6935521146290732e-05, + "loss": 0.0, + "step": 11976 + }, + { + "epoch": 2.77, + "learning_rate": 2.6933595254602883e-05, + "loss": 0.0, + "step": 11977 + }, + { + "epoch": 2.77, + "learning_rate": 2.693166936291503e-05, + "loss": 0.0, + "step": 11978 + }, + { + "epoch": 2.77, + "learning_rate": 2.692974347122718e-05, + "loss": 0.0, + "step": 11979 + }, + { + "epoch": 2.77, + "learning_rate": 2.6927817579539326e-05, + "loss": 0.0001, + "step": 11980 + }, + { + "epoch": 2.77, + "learning_rate": 2.692589168785148e-05, + "loss": 0.0001, + "step": 11981 + }, + { + "epoch": 2.77, + "learning_rate": 2.6923965796163624e-05, + "loss": 0.0002, + "step": 11982 + }, + { + "epoch": 2.77, + "learning_rate": 2.692203990447577e-05, + "loss": 0.0, + "step": 11983 + }, + { + "epoch": 2.77, + "learning_rate": 2.6920114012787922e-05, + "loss": 0.0, + "step": 11984 + }, + { + "epoch": 2.77, + "learning_rate": 2.691818812110007e-05, + "loss": 0.0002, + "step": 11985 + }, + { + "epoch": 2.77, + "learning_rate": 2.691626222941222e-05, + "loss": 0.0, + "step": 11986 + }, + { + "epoch": 2.77, + "learning_rate": 2.6914336337724365e-05, + "loss": 0.0, + "step": 11987 + }, + { + "epoch": 2.77, + "learning_rate": 2.691241044603652e-05, + "loss": 0.0, + "step": 11988 + }, + { + "epoch": 2.77, + "learning_rate": 2.6910484554348663e-05, + "loss": 0.0, + "step": 11989 + }, + { + "epoch": 2.77, + "learning_rate": 2.6908558662660814e-05, + "loss": 0.0, + "step": 11990 + }, + { + "epoch": 2.77, + "learning_rate": 2.690663277097296e-05, + "loss": 0.0, + "step": 11991 + }, + { + "epoch": 2.77, + "learning_rate": 2.690470687928511e-05, + "loss": 0.0, + "step": 11992 + }, + { + "epoch": 2.77, + "learning_rate": 2.690278098759726e-05, + "loss": 0.0, + "step": 11993 + }, + { + "epoch": 2.77, + "learning_rate": 2.6900855095909404e-05, + "loss": 0.0, + "step": 11994 + }, + { + "epoch": 2.77, + "learning_rate": 2.6898929204221558e-05, + "loss": 0.0, + "step": 11995 + }, + { + "epoch": 2.77, + "learning_rate": 2.6897003312533702e-05, + "loss": 0.0, + "step": 11996 + }, + { + "epoch": 2.77, + "learning_rate": 2.6895077420845853e-05, + "loss": 0.0, + "step": 11997 + }, + { + "epoch": 2.77, + "learning_rate": 2.6893151529158e-05, + "loss": 0.0, + "step": 11998 + }, + { + "epoch": 2.77, + "learning_rate": 2.689122563747015e-05, + "loss": 0.0, + "step": 11999 + }, + { + "epoch": 2.77, + "learning_rate": 2.68892997457823e-05, + "loss": 0.0, + "step": 12000 + }, + { + "epoch": 2.77, + "learning_rate": 2.688737385409445e-05, + "loss": 0.0, + "step": 12001 + }, + { + "epoch": 2.77, + "learning_rate": 2.6885447962406597e-05, + "loss": 0.0005, + "step": 12002 + }, + { + "epoch": 2.77, + "learning_rate": 2.688352207071874e-05, + "loss": 0.0, + "step": 12003 + }, + { + "epoch": 2.77, + "learning_rate": 2.6881596179030892e-05, + "loss": 0.0, + "step": 12004 + }, + { + "epoch": 2.77, + "learning_rate": 2.687967028734304e-05, + "loss": 0.0, + "step": 12005 + }, + { + "epoch": 2.77, + "learning_rate": 2.687774439565519e-05, + "loss": 0.0, + "step": 12006 + }, + { + "epoch": 2.77, + "learning_rate": 2.6875818503967338e-05, + "loss": 0.0, + "step": 12007 + }, + { + "epoch": 2.78, + "learning_rate": 2.687389261227949e-05, + "loss": 0.0, + "step": 12008 + }, + { + "epoch": 2.78, + "learning_rate": 2.6871966720591636e-05, + "loss": 0.0001, + "step": 12009 + }, + { + "epoch": 2.78, + "learning_rate": 2.6870040828903787e-05, + "loss": 0.0, + "step": 12010 + }, + { + "epoch": 2.78, + "learning_rate": 2.686811493721593e-05, + "loss": 0.0002, + "step": 12011 + }, + { + "epoch": 2.78, + "learning_rate": 2.686618904552808e-05, + "loss": 0.0001, + "step": 12012 + }, + { + "epoch": 2.78, + "learning_rate": 2.686426315384023e-05, + "loss": 0.0, + "step": 12013 + }, + { + "epoch": 2.78, + "learning_rate": 2.6862337262152377e-05, + "loss": 0.0, + "step": 12014 + }, + { + "epoch": 2.78, + "learning_rate": 2.6860411370464528e-05, + "loss": 0.0061, + "step": 12015 + }, + { + "epoch": 2.78, + "learning_rate": 2.6858485478776672e-05, + "loss": 0.0, + "step": 12016 + }, + { + "epoch": 2.78, + "learning_rate": 2.6856559587088826e-05, + "loss": 0.0, + "step": 12017 + }, + { + "epoch": 2.78, + "learning_rate": 2.685463369540097e-05, + "loss": 0.0001, + "step": 12018 + }, + { + "epoch": 2.78, + "learning_rate": 2.6852707803713125e-05, + "loss": 0.0, + "step": 12019 + }, + { + "epoch": 2.78, + "learning_rate": 2.685078191202527e-05, + "loss": 0.0, + "step": 12020 + }, + { + "epoch": 2.78, + "learning_rate": 2.6848856020337416e-05, + "loss": 0.0, + "step": 12021 + }, + { + "epoch": 2.78, + "learning_rate": 2.6846930128649567e-05, + "loss": 0.0002, + "step": 12022 + }, + { + "epoch": 2.78, + "learning_rate": 2.684500423696171e-05, + "loss": 0.0, + "step": 12023 + }, + { + "epoch": 2.78, + "learning_rate": 2.6843078345273865e-05, + "loss": 0.0, + "step": 12024 + }, + { + "epoch": 2.78, + "learning_rate": 2.684115245358601e-05, + "loss": 0.0001, + "step": 12025 + }, + { + "epoch": 2.78, + "learning_rate": 2.683922656189816e-05, + "loss": 0.0, + "step": 12026 + }, + { + "epoch": 2.78, + "learning_rate": 2.6837300670210308e-05, + "loss": 0.0, + "step": 12027 + }, + { + "epoch": 2.78, + "learning_rate": 2.683537477852246e-05, + "loss": 0.0003, + "step": 12028 + }, + { + "epoch": 2.78, + "learning_rate": 2.6833448886834606e-05, + "loss": 0.0, + "step": 12029 + }, + { + "epoch": 2.78, + "learning_rate": 2.683152299514675e-05, + "loss": 0.0, + "step": 12030 + }, + { + "epoch": 2.78, + "learning_rate": 2.6829597103458905e-05, + "loss": 0.0, + "step": 12031 + }, + { + "epoch": 2.78, + "learning_rate": 2.682767121177105e-05, + "loss": 0.0, + "step": 12032 + }, + { + "epoch": 2.78, + "learning_rate": 2.68257453200832e-05, + "loss": 0.0, + "step": 12033 + }, + { + "epoch": 2.78, + "learning_rate": 2.6823819428395347e-05, + "loss": 0.0001, + "step": 12034 + }, + { + "epoch": 2.78, + "learning_rate": 2.6821893536707498e-05, + "loss": 0.0001, + "step": 12035 + }, + { + "epoch": 2.78, + "learning_rate": 2.6819967645019645e-05, + "loss": 0.0, + "step": 12036 + }, + { + "epoch": 2.78, + "learning_rate": 2.6818041753331796e-05, + "loss": 0.0, + "step": 12037 + }, + { + "epoch": 2.78, + "learning_rate": 2.6816115861643944e-05, + "loss": 0.0, + "step": 12038 + }, + { + "epoch": 2.78, + "learning_rate": 2.6814189969956088e-05, + "loss": 0.0001, + "step": 12039 + }, + { + "epoch": 2.78, + "learning_rate": 2.681226407826824e-05, + "loss": 0.0006, + "step": 12040 + }, + { + "epoch": 2.78, + "learning_rate": 2.6810338186580386e-05, + "loss": 0.0, + "step": 12041 + }, + { + "epoch": 2.78, + "learning_rate": 2.6808412294892537e-05, + "loss": 0.0, + "step": 12042 + }, + { + "epoch": 2.78, + "learning_rate": 2.6806486403204684e-05, + "loss": 0.0001, + "step": 12043 + }, + { + "epoch": 2.78, + "learning_rate": 2.6804560511516835e-05, + "loss": 0.0, + "step": 12044 + }, + { + "epoch": 2.78, + "learning_rate": 2.6802634619828983e-05, + "loss": 0.0, + "step": 12045 + }, + { + "epoch": 2.78, + "learning_rate": 2.6800708728141134e-05, + "loss": 0.0, + "step": 12046 + }, + { + "epoch": 2.78, + "learning_rate": 2.6798782836453278e-05, + "loss": 0.0, + "step": 12047 + }, + { + "epoch": 2.78, + "learning_rate": 2.6796856944765425e-05, + "loss": 0.0, + "step": 12048 + }, + { + "epoch": 2.78, + "learning_rate": 2.6794931053077576e-05, + "loss": 0.0, + "step": 12049 + }, + { + "epoch": 2.78, + "learning_rate": 2.6793005161389724e-05, + "loss": 0.0, + "step": 12050 + }, + { + "epoch": 2.79, + "learning_rate": 2.6791079269701874e-05, + "loss": 0.0, + "step": 12051 + }, + { + "epoch": 2.79, + "learning_rate": 2.678915337801402e-05, + "loss": 0.0, + "step": 12052 + }, + { + "epoch": 2.79, + "learning_rate": 2.6787227486326173e-05, + "loss": 0.0, + "step": 12053 + }, + { + "epoch": 2.79, + "learning_rate": 2.6785301594638317e-05, + "loss": 0.0, + "step": 12054 + }, + { + "epoch": 2.79, + "learning_rate": 2.678337570295047e-05, + "loss": 0.0, + "step": 12055 + }, + { + "epoch": 2.79, + "learning_rate": 2.6781449811262615e-05, + "loss": 0.0, + "step": 12056 + }, + { + "epoch": 2.79, + "learning_rate": 2.6779523919574766e-05, + "loss": 0.0003, + "step": 12057 + }, + { + "epoch": 2.79, + "learning_rate": 2.6777598027886914e-05, + "loss": 0.0, + "step": 12058 + }, + { + "epoch": 2.79, + "learning_rate": 2.6775672136199058e-05, + "loss": 0.0, + "step": 12059 + }, + { + "epoch": 2.79, + "learning_rate": 2.6773746244511212e-05, + "loss": 0.0, + "step": 12060 + }, + { + "epoch": 2.79, + "learning_rate": 2.6771820352823356e-05, + "loss": 0.0, + "step": 12061 + }, + { + "epoch": 2.79, + "learning_rate": 2.6769894461135507e-05, + "loss": 0.0, + "step": 12062 + }, + { + "epoch": 2.79, + "learning_rate": 2.6767968569447654e-05, + "loss": 0.0, + "step": 12063 + }, + { + "epoch": 2.79, + "learning_rate": 2.6766042677759805e-05, + "loss": 0.0, + "step": 12064 + }, + { + "epoch": 2.79, + "learning_rate": 2.6764116786071953e-05, + "loss": 0.0017, + "step": 12065 + }, + { + "epoch": 2.79, + "learning_rate": 2.6762190894384104e-05, + "loss": 0.0, + "step": 12066 + }, + { + "epoch": 2.79, + "learning_rate": 2.676026500269625e-05, + "loss": 0.0, + "step": 12067 + }, + { + "epoch": 2.79, + "learning_rate": 2.6758339111008395e-05, + "loss": 0.0005, + "step": 12068 + }, + { + "epoch": 2.79, + "learning_rate": 2.6756413219320546e-05, + "loss": 0.0, + "step": 12069 + }, + { + "epoch": 2.79, + "learning_rate": 2.6754487327632693e-05, + "loss": 0.0, + "step": 12070 + }, + { + "epoch": 2.79, + "learning_rate": 2.6752561435944844e-05, + "loss": 0.0001, + "step": 12071 + }, + { + "epoch": 2.79, + "learning_rate": 2.6750635544256992e-05, + "loss": 0.0, + "step": 12072 + }, + { + "epoch": 2.79, + "learning_rate": 2.6748709652569143e-05, + "loss": 0.0, + "step": 12073 + }, + { + "epoch": 2.79, + "learning_rate": 2.674678376088129e-05, + "loss": 0.0005, + "step": 12074 + }, + { + "epoch": 2.79, + "learning_rate": 2.674485786919344e-05, + "loss": 0.0002, + "step": 12075 + }, + { + "epoch": 2.79, + "learning_rate": 2.6742931977505585e-05, + "loss": 0.0, + "step": 12076 + }, + { + "epoch": 2.79, + "learning_rate": 2.6741006085817733e-05, + "loss": 0.0, + "step": 12077 + }, + { + "epoch": 2.79, + "learning_rate": 2.6739080194129883e-05, + "loss": 0.0011, + "step": 12078 + }, + { + "epoch": 2.79, + "learning_rate": 2.673715430244203e-05, + "loss": 0.0, + "step": 12079 + }, + { + "epoch": 2.79, + "learning_rate": 2.6735228410754182e-05, + "loss": 0.0001, + "step": 12080 + }, + { + "epoch": 2.79, + "learning_rate": 2.673330251906633e-05, + "loss": 0.0, + "step": 12081 + }, + { + "epoch": 2.79, + "learning_rate": 2.673137662737848e-05, + "loss": 0.0, + "step": 12082 + }, + { + "epoch": 2.79, + "learning_rate": 2.6729450735690624e-05, + "loss": 0.0, + "step": 12083 + }, + { + "epoch": 2.79, + "learning_rate": 2.672752484400278e-05, + "loss": 0.0, + "step": 12084 + }, + { + "epoch": 2.79, + "learning_rate": 2.6725598952314923e-05, + "loss": 0.0, + "step": 12085 + }, + { + "epoch": 2.79, + "learning_rate": 2.672367306062707e-05, + "loss": 0.0, + "step": 12086 + }, + { + "epoch": 2.79, + "learning_rate": 2.672174716893922e-05, + "loss": 0.0, + "step": 12087 + }, + { + "epoch": 2.79, + "learning_rate": 2.671982127725137e-05, + "loss": 0.0001, + "step": 12088 + }, + { + "epoch": 2.79, + "learning_rate": 2.671789538556352e-05, + "loss": 0.0, + "step": 12089 + }, + { + "epoch": 2.79, + "learning_rate": 2.6715969493875663e-05, + "loss": 0.0, + "step": 12090 + }, + { + "epoch": 2.79, + "learning_rate": 2.6714043602187818e-05, + "loss": 0.0092, + "step": 12091 + }, + { + "epoch": 2.79, + "learning_rate": 2.6712117710499962e-05, + "loss": 0.0, + "step": 12092 + }, + { + "epoch": 2.79, + "learning_rate": 2.6710191818812113e-05, + "loss": 0.0, + "step": 12093 + }, + { + "epoch": 2.8, + "learning_rate": 2.670826592712426e-05, + "loss": 0.0, + "step": 12094 + }, + { + "epoch": 2.8, + "learning_rate": 2.6706340035436404e-05, + "loss": 0.0001, + "step": 12095 + }, + { + "epoch": 2.8, + "learning_rate": 2.670441414374856e-05, + "loss": 0.0, + "step": 12096 + }, + { + "epoch": 2.8, + "learning_rate": 2.6702488252060702e-05, + "loss": 0.0, + "step": 12097 + }, + { + "epoch": 2.8, + "learning_rate": 2.6700562360372853e-05, + "loss": 0.0003, + "step": 12098 + }, + { + "epoch": 2.8, + "learning_rate": 2.6698636468685e-05, + "loss": 0.0, + "step": 12099 + }, + { + "epoch": 2.8, + "learning_rate": 2.6696710576997152e-05, + "loss": 0.0, + "step": 12100 + }, + { + "epoch": 2.8, + "learning_rate": 2.66947846853093e-05, + "loss": 0.0, + "step": 12101 + }, + { + "epoch": 2.8, + "learning_rate": 2.669285879362145e-05, + "loss": 0.0, + "step": 12102 + }, + { + "epoch": 2.8, + "learning_rate": 2.6690932901933598e-05, + "loss": 0.0, + "step": 12103 + }, + { + "epoch": 2.8, + "learning_rate": 2.668900701024574e-05, + "loss": 0.0001, + "step": 12104 + }, + { + "epoch": 2.8, + "learning_rate": 2.6687081118557892e-05, + "loss": 0.0, + "step": 12105 + }, + { + "epoch": 2.8, + "learning_rate": 2.668515522687004e-05, + "loss": 0.0, + "step": 12106 + }, + { + "epoch": 2.8, + "learning_rate": 2.668322933518219e-05, + "loss": 0.0, + "step": 12107 + }, + { + "epoch": 2.8, + "learning_rate": 2.6681303443494338e-05, + "loss": 0.0, + "step": 12108 + }, + { + "epoch": 2.8, + "learning_rate": 2.667937755180649e-05, + "loss": 0.0, + "step": 12109 + }, + { + "epoch": 2.8, + "learning_rate": 2.6677451660118637e-05, + "loss": 0.0, + "step": 12110 + }, + { + "epoch": 2.8, + "learning_rate": 2.6675525768430788e-05, + "loss": 0.0, + "step": 12111 + }, + { + "epoch": 2.8, + "learning_rate": 2.667359987674293e-05, + "loss": 0.0, + "step": 12112 + }, + { + "epoch": 2.8, + "learning_rate": 2.6671673985055086e-05, + "loss": 0.0, + "step": 12113 + }, + { + "epoch": 2.8, + "learning_rate": 2.666974809336723e-05, + "loss": 0.0, + "step": 12114 + }, + { + "epoch": 2.8, + "learning_rate": 2.6667822201679377e-05, + "loss": 0.0002, + "step": 12115 + }, + { + "epoch": 2.8, + "learning_rate": 2.6665896309991528e-05, + "loss": 0.0001, + "step": 12116 + }, + { + "epoch": 2.8, + "learning_rate": 2.6663970418303676e-05, + "loss": 0.0001, + "step": 12117 + }, + { + "epoch": 2.8, + "learning_rate": 2.6662044526615827e-05, + "loss": 0.0, + "step": 12118 + }, + { + "epoch": 2.8, + "learning_rate": 2.666011863492797e-05, + "loss": 0.0, + "step": 12119 + }, + { + "epoch": 2.8, + "learning_rate": 2.6658192743240125e-05, + "loss": 0.0002, + "step": 12120 + }, + { + "epoch": 2.8, + "learning_rate": 2.665626685155227e-05, + "loss": 0.0, + "step": 12121 + }, + { + "epoch": 2.8, + "learning_rate": 2.665434095986442e-05, + "loss": 0.0, + "step": 12122 + }, + { + "epoch": 2.8, + "learning_rate": 2.6652415068176567e-05, + "loss": 0.0, + "step": 12123 + }, + { + "epoch": 2.8, + "learning_rate": 2.6650489176488715e-05, + "loss": 0.0, + "step": 12124 + }, + { + "epoch": 2.8, + "learning_rate": 2.6648563284800866e-05, + "loss": 0.0, + "step": 12125 + }, + { + "epoch": 2.8, + "learning_rate": 2.664663739311301e-05, + "loss": 0.0, + "step": 12126 + }, + { + "epoch": 2.8, + "learning_rate": 2.6644711501425164e-05, + "loss": 0.0, + "step": 12127 + }, + { + "epoch": 2.8, + "learning_rate": 2.6642785609737308e-05, + "loss": 0.0, + "step": 12128 + }, + { + "epoch": 2.8, + "learning_rate": 2.664085971804946e-05, + "loss": 0.0, + "step": 12129 + }, + { + "epoch": 2.8, + "learning_rate": 2.6638933826361607e-05, + "loss": 0.0, + "step": 12130 + }, + { + "epoch": 2.8, + "learning_rate": 2.6637007934673757e-05, + "loss": 0.0, + "step": 12131 + }, + { + "epoch": 2.8, + "learning_rate": 2.6635082042985905e-05, + "loss": 0.0, + "step": 12132 + }, + { + "epoch": 2.8, + "learning_rate": 2.663315615129805e-05, + "loss": 0.0, + "step": 12133 + }, + { + "epoch": 2.8, + "learning_rate": 2.66312302596102e-05, + "loss": 0.0, + "step": 12134 + }, + { + "epoch": 2.8, + "learning_rate": 2.6629304367922347e-05, + "loss": 0.0, + "step": 12135 + }, + { + "epoch": 2.8, + "learning_rate": 2.6627378476234498e-05, + "loss": 0.0, + "step": 12136 + }, + { + "epoch": 2.8, + "learning_rate": 2.6625452584546646e-05, + "loss": 0.0, + "step": 12137 + }, + { + "epoch": 2.81, + "learning_rate": 2.6623526692858797e-05, + "loss": 0.0, + "step": 12138 + }, + { + "epoch": 2.81, + "learning_rate": 2.6621600801170944e-05, + "loss": 0.0, + "step": 12139 + }, + { + "epoch": 2.81, + "learning_rate": 2.6619674909483095e-05, + "loss": 0.0001, + "step": 12140 + }, + { + "epoch": 2.81, + "learning_rate": 2.661774901779524e-05, + "loss": 0.0, + "step": 12141 + }, + { + "epoch": 2.81, + "learning_rate": 2.6615823126107386e-05, + "loss": 0.0, + "step": 12142 + }, + { + "epoch": 2.81, + "learning_rate": 2.6613897234419537e-05, + "loss": 0.0001, + "step": 12143 + }, + { + "epoch": 2.81, + "learning_rate": 2.6611971342731685e-05, + "loss": 0.0, + "step": 12144 + }, + { + "epoch": 2.81, + "learning_rate": 2.6610045451043836e-05, + "loss": 0.0004, + "step": 12145 + }, + { + "epoch": 2.81, + "learning_rate": 2.6608119559355983e-05, + "loss": 0.0, + "step": 12146 + }, + { + "epoch": 2.81, + "learning_rate": 2.6606193667668134e-05, + "loss": 0.0001, + "step": 12147 + }, + { + "epoch": 2.81, + "learning_rate": 2.6604267775980278e-05, + "loss": 0.0, + "step": 12148 + }, + { + "epoch": 2.81, + "learning_rate": 2.6602341884292432e-05, + "loss": 0.0, + "step": 12149 + }, + { + "epoch": 2.81, + "learning_rate": 2.6600415992604576e-05, + "loss": 0.0, + "step": 12150 + }, + { + "epoch": 2.81, + "learning_rate": 2.6598490100916724e-05, + "loss": 0.0, + "step": 12151 + }, + { + "epoch": 2.81, + "learning_rate": 2.6596564209228875e-05, + "loss": 0.0, + "step": 12152 + }, + { + "epoch": 2.81, + "learning_rate": 2.6594638317541022e-05, + "loss": 0.0001, + "step": 12153 + }, + { + "epoch": 2.81, + "learning_rate": 2.6592712425853173e-05, + "loss": 0.0, + "step": 12154 + }, + { + "epoch": 2.81, + "learning_rate": 2.6590786534165317e-05, + "loss": 0.0, + "step": 12155 + }, + { + "epoch": 2.81, + "learning_rate": 2.658886064247747e-05, + "loss": 0.0, + "step": 12156 + }, + { + "epoch": 2.81, + "learning_rate": 2.6586934750789616e-05, + "loss": 0.0, + "step": 12157 + }, + { + "epoch": 2.81, + "learning_rate": 2.6585008859101766e-05, + "loss": 0.0, + "step": 12158 + }, + { + "epoch": 2.81, + "learning_rate": 2.6583082967413914e-05, + "loss": 0.0001, + "step": 12159 + }, + { + "epoch": 2.81, + "learning_rate": 2.658115707572606e-05, + "loss": 0.0, + "step": 12160 + }, + { + "epoch": 2.81, + "learning_rate": 2.6579231184038212e-05, + "loss": 0.0, + "step": 12161 + }, + { + "epoch": 2.81, + "learning_rate": 2.6577305292350356e-05, + "loss": 0.0, + "step": 12162 + }, + { + "epoch": 2.81, + "learning_rate": 2.657537940066251e-05, + "loss": 0.0, + "step": 12163 + }, + { + "epoch": 2.81, + "learning_rate": 2.6573453508974655e-05, + "loss": 0.0004, + "step": 12164 + }, + { + "epoch": 2.81, + "learning_rate": 2.6571527617286806e-05, + "loss": 0.0001, + "step": 12165 + }, + { + "epoch": 2.81, + "learning_rate": 2.6569601725598953e-05, + "loss": 0.0001, + "step": 12166 + }, + { + "epoch": 2.81, + "learning_rate": 2.6567675833911104e-05, + "loss": 0.0, + "step": 12167 + }, + { + "epoch": 2.81, + "learning_rate": 2.656574994222325e-05, + "loss": 0.0, + "step": 12168 + }, + { + "epoch": 2.81, + "learning_rate": 2.6563824050535402e-05, + "loss": 0.0001, + "step": 12169 + }, + { + "epoch": 2.81, + "learning_rate": 2.6561898158847546e-05, + "loss": 0.0, + "step": 12170 + }, + { + "epoch": 2.81, + "learning_rate": 2.6559972267159694e-05, + "loss": 0.0, + "step": 12171 + }, + { + "epoch": 2.81, + "learning_rate": 2.6558046375471845e-05, + "loss": 0.0, + "step": 12172 + }, + { + "epoch": 2.81, + "learning_rate": 2.6556120483783992e-05, + "loss": 0.0, + "step": 12173 + }, + { + "epoch": 2.81, + "learning_rate": 2.6554194592096143e-05, + "loss": 0.0001, + "step": 12174 + }, + { + "epoch": 2.81, + "learning_rate": 2.655226870040829e-05, + "loss": 0.0, + "step": 12175 + }, + { + "epoch": 2.81, + "learning_rate": 2.655034280872044e-05, + "loss": 0.0001, + "step": 12176 + }, + { + "epoch": 2.81, + "learning_rate": 2.6548416917032585e-05, + "loss": 0.0, + "step": 12177 + }, + { + "epoch": 2.81, + "learning_rate": 2.654649102534474e-05, + "loss": 0.0, + "step": 12178 + }, + { + "epoch": 2.81, + "learning_rate": 2.6544565133656884e-05, + "loss": 0.0, + "step": 12179 + }, + { + "epoch": 2.81, + "learning_rate": 2.654263924196903e-05, + "loss": 0.0001, + "step": 12180 + }, + { + "epoch": 2.82, + "learning_rate": 2.6540713350281182e-05, + "loss": 0.0001, + "step": 12181 + }, + { + "epoch": 2.82, + "learning_rate": 2.653878745859333e-05, + "loss": 0.0, + "step": 12182 + }, + { + "epoch": 2.82, + "learning_rate": 2.653686156690548e-05, + "loss": 0.0, + "step": 12183 + }, + { + "epoch": 2.82, + "learning_rate": 2.6534935675217625e-05, + "loss": 0.0, + "step": 12184 + }, + { + "epoch": 2.82, + "learning_rate": 2.653300978352978e-05, + "loss": 0.0, + "step": 12185 + }, + { + "epoch": 2.82, + "learning_rate": 2.6531083891841923e-05, + "loss": 0.0, + "step": 12186 + }, + { + "epoch": 2.82, + "learning_rate": 2.6529158000154074e-05, + "loss": 0.0002, + "step": 12187 + }, + { + "epoch": 2.82, + "learning_rate": 2.652723210846622e-05, + "loss": 0.0, + "step": 12188 + }, + { + "epoch": 2.82, + "learning_rate": 2.652530621677837e-05, + "loss": 0.0, + "step": 12189 + }, + { + "epoch": 2.82, + "learning_rate": 2.652338032509052e-05, + "loss": 0.0037, + "step": 12190 + }, + { + "epoch": 2.82, + "learning_rate": 2.6521454433402664e-05, + "loss": 0.0002, + "step": 12191 + }, + { + "epoch": 2.82, + "learning_rate": 2.6519528541714818e-05, + "loss": 0.0, + "step": 12192 + }, + { + "epoch": 2.82, + "learning_rate": 2.6517602650026962e-05, + "loss": 0.0, + "step": 12193 + }, + { + "epoch": 2.82, + "learning_rate": 2.6515676758339113e-05, + "loss": 0.0, + "step": 12194 + }, + { + "epoch": 2.82, + "learning_rate": 2.651375086665126e-05, + "loss": 0.0001, + "step": 12195 + }, + { + "epoch": 2.82, + "learning_rate": 2.651182497496341e-05, + "loss": 0.0, + "step": 12196 + }, + { + "epoch": 2.82, + "learning_rate": 2.650989908327556e-05, + "loss": 0.0, + "step": 12197 + }, + { + "epoch": 2.82, + "learning_rate": 2.6507973191587703e-05, + "loss": 0.0, + "step": 12198 + }, + { + "epoch": 2.82, + "learning_rate": 2.6506047299899857e-05, + "loss": 0.0, + "step": 12199 + }, + { + "epoch": 2.82, + "learning_rate": 2.6504121408212e-05, + "loss": 0.0, + "step": 12200 + }, + { + "epoch": 2.82, + "learning_rate": 2.6502195516524152e-05, + "loss": 0.0004, + "step": 12201 + }, + { + "epoch": 2.82, + "learning_rate": 2.65002696248363e-05, + "loss": 0.0, + "step": 12202 + }, + { + "epoch": 2.82, + "learning_rate": 2.649834373314845e-05, + "loss": 0.0, + "step": 12203 + }, + { + "epoch": 2.82, + "learning_rate": 2.6496417841460598e-05, + "loss": 0.0, + "step": 12204 + }, + { + "epoch": 2.82, + "learning_rate": 2.649449194977275e-05, + "loss": 0.0, + "step": 12205 + }, + { + "epoch": 2.82, + "learning_rate": 2.6492566058084893e-05, + "loss": 0.0, + "step": 12206 + }, + { + "epoch": 2.82, + "learning_rate": 2.649064016639704e-05, + "loss": 0.0, + "step": 12207 + }, + { + "epoch": 2.82, + "learning_rate": 2.648871427470919e-05, + "loss": 0.0, + "step": 12208 + }, + { + "epoch": 2.82, + "learning_rate": 2.648678838302134e-05, + "loss": 0.0, + "step": 12209 + }, + { + "epoch": 2.82, + "learning_rate": 2.648486249133349e-05, + "loss": 0.0, + "step": 12210 + }, + { + "epoch": 2.82, + "learning_rate": 2.6482936599645637e-05, + "loss": 0.0, + "step": 12211 + }, + { + "epoch": 2.82, + "learning_rate": 2.6481010707957788e-05, + "loss": 0.0, + "step": 12212 + }, + { + "epoch": 2.82, + "learning_rate": 2.6479084816269932e-05, + "loss": 0.0, + "step": 12213 + }, + { + "epoch": 2.82, + "learning_rate": 2.6477158924582086e-05, + "loss": 0.0, + "step": 12214 + }, + { + "epoch": 2.82, + "learning_rate": 2.647523303289423e-05, + "loss": 0.0, + "step": 12215 + }, + { + "epoch": 2.82, + "learning_rate": 2.6473307141206378e-05, + "loss": 0.0, + "step": 12216 + }, + { + "epoch": 2.82, + "learning_rate": 2.647138124951853e-05, + "loss": 0.0, + "step": 12217 + }, + { + "epoch": 2.82, + "learning_rate": 2.6469455357830676e-05, + "loss": 0.0001, + "step": 12218 + }, + { + "epoch": 2.82, + "learning_rate": 2.6467529466142827e-05, + "loss": 0.0, + "step": 12219 + }, + { + "epoch": 2.82, + "learning_rate": 2.646560357445497e-05, + "loss": 0.0, + "step": 12220 + }, + { + "epoch": 2.82, + "learning_rate": 2.6463677682767125e-05, + "loss": 0.0, + "step": 12221 + }, + { + "epoch": 2.82, + "learning_rate": 2.646175179107927e-05, + "loss": 0.0, + "step": 12222 + }, + { + "epoch": 2.82, + "learning_rate": 2.645982589939142e-05, + "loss": 0.0, + "step": 12223 + }, + { + "epoch": 2.83, + "learning_rate": 2.6457900007703568e-05, + "loss": 0.0, + "step": 12224 + }, + { + "epoch": 2.83, + "learning_rate": 2.645597411601572e-05, + "loss": 0.0, + "step": 12225 + }, + { + "epoch": 2.83, + "learning_rate": 2.6454048224327866e-05, + "loss": 0.0, + "step": 12226 + }, + { + "epoch": 2.83, + "learning_rate": 2.645212233264001e-05, + "loss": 0.0, + "step": 12227 + }, + { + "epoch": 2.83, + "learning_rate": 2.6450196440952164e-05, + "loss": 0.0001, + "step": 12228 + }, + { + "epoch": 2.83, + "learning_rate": 2.644827054926431e-05, + "loss": 0.0, + "step": 12229 + }, + { + "epoch": 2.83, + "learning_rate": 2.644634465757646e-05, + "loss": 0.0, + "step": 12230 + }, + { + "epoch": 2.83, + "learning_rate": 2.6444418765888607e-05, + "loss": 0.0, + "step": 12231 + }, + { + "epoch": 2.83, + "learning_rate": 2.6442492874200758e-05, + "loss": 0.0002, + "step": 12232 + }, + { + "epoch": 2.83, + "learning_rate": 2.6440566982512905e-05, + "loss": 0.0, + "step": 12233 + }, + { + "epoch": 2.83, + "learning_rate": 2.6438641090825056e-05, + "loss": 0.0003, + "step": 12234 + }, + { + "epoch": 2.83, + "learning_rate": 2.6436715199137204e-05, + "loss": 0.0, + "step": 12235 + }, + { + "epoch": 2.83, + "learning_rate": 2.6434789307449348e-05, + "loss": 0.0, + "step": 12236 + }, + { + "epoch": 2.83, + "learning_rate": 2.64328634157615e-05, + "loss": 0.0, + "step": 12237 + }, + { + "epoch": 2.83, + "learning_rate": 2.6430937524073646e-05, + "loss": 0.0, + "step": 12238 + }, + { + "epoch": 2.83, + "learning_rate": 2.6429011632385797e-05, + "loss": 0.0, + "step": 12239 + }, + { + "epoch": 2.83, + "learning_rate": 2.6427085740697944e-05, + "loss": 0.0, + "step": 12240 + }, + { + "epoch": 2.83, + "learning_rate": 2.6425159849010095e-05, + "loss": 0.0001, + "step": 12241 + }, + { + "epoch": 2.83, + "learning_rate": 2.6423233957322243e-05, + "loss": 0.0, + "step": 12242 + }, + { + "epoch": 2.83, + "learning_rate": 2.6421308065634393e-05, + "loss": 0.0, + "step": 12243 + }, + { + "epoch": 2.83, + "learning_rate": 2.6419382173946538e-05, + "loss": 0.0, + "step": 12244 + }, + { + "epoch": 2.83, + "learning_rate": 2.6417456282258685e-05, + "loss": 0.0, + "step": 12245 + }, + { + "epoch": 2.83, + "learning_rate": 2.6415530390570836e-05, + "loss": 0.0, + "step": 12246 + }, + { + "epoch": 2.83, + "learning_rate": 2.6413604498882983e-05, + "loss": 0.0001, + "step": 12247 + }, + { + "epoch": 2.83, + "learning_rate": 2.6411678607195134e-05, + "loss": 0.0, + "step": 12248 + }, + { + "epoch": 2.83, + "learning_rate": 2.640975271550728e-05, + "loss": 0.0, + "step": 12249 + }, + { + "epoch": 2.83, + "learning_rate": 2.6407826823819433e-05, + "loss": 0.0, + "step": 12250 + }, + { + "epoch": 2.83, + "learning_rate": 2.6405900932131577e-05, + "loss": 0.0, + "step": 12251 + }, + { + "epoch": 2.83, + "learning_rate": 2.6403975040443728e-05, + "loss": 0.0, + "step": 12252 + }, + { + "epoch": 2.83, + "learning_rate": 2.6402049148755875e-05, + "loss": 0.0002, + "step": 12253 + }, + { + "epoch": 2.83, + "learning_rate": 2.6400123257068023e-05, + "loss": 0.0, + "step": 12254 + }, + { + "epoch": 2.83, + "learning_rate": 2.6398197365380173e-05, + "loss": 0.0001, + "step": 12255 + }, + { + "epoch": 2.83, + "learning_rate": 2.6396271473692317e-05, + "loss": 0.0001, + "step": 12256 + }, + { + "epoch": 2.83, + "learning_rate": 2.6394345582004472e-05, + "loss": 0.0, + "step": 12257 + }, + { + "epoch": 2.83, + "learning_rate": 2.6392419690316616e-05, + "loss": 0.0, + "step": 12258 + }, + { + "epoch": 2.83, + "learning_rate": 2.6390493798628767e-05, + "loss": 0.0, + "step": 12259 + }, + { + "epoch": 2.83, + "learning_rate": 2.6388567906940914e-05, + "loss": 0.0, + "step": 12260 + }, + { + "epoch": 2.83, + "learning_rate": 2.6386642015253065e-05, + "loss": 0.0002, + "step": 12261 + }, + { + "epoch": 2.83, + "learning_rate": 2.6384716123565213e-05, + "loss": 0.0, + "step": 12262 + }, + { + "epoch": 2.83, + "learning_rate": 2.6382790231877357e-05, + "loss": 0.0, + "step": 12263 + }, + { + "epoch": 2.83, + "learning_rate": 2.638086434018951e-05, + "loss": 0.0, + "step": 12264 + }, + { + "epoch": 2.83, + "learning_rate": 2.6378938448501655e-05, + "loss": 0.0, + "step": 12265 + }, + { + "epoch": 2.83, + "learning_rate": 2.6377012556813806e-05, + "loss": 0.0, + "step": 12266 + }, + { + "epoch": 2.83, + "learning_rate": 2.6375086665125953e-05, + "loss": 0.0, + "step": 12267 + }, + { + "epoch": 2.84, + "learning_rate": 2.6373160773438104e-05, + "loss": 0.0, + "step": 12268 + }, + { + "epoch": 2.84, + "learning_rate": 2.637123488175025e-05, + "loss": 0.0, + "step": 12269 + }, + { + "epoch": 2.84, + "learning_rate": 2.6369308990062402e-05, + "loss": 0.0, + "step": 12270 + }, + { + "epoch": 2.84, + "learning_rate": 2.636738309837455e-05, + "loss": 0.0001, + "step": 12271 + }, + { + "epoch": 2.84, + "learning_rate": 2.6365457206686694e-05, + "loss": 0.0, + "step": 12272 + }, + { + "epoch": 2.84, + "learning_rate": 2.6363531314998845e-05, + "loss": 0.0, + "step": 12273 + }, + { + "epoch": 2.84, + "learning_rate": 2.6361605423310992e-05, + "loss": 0.0, + "step": 12274 + }, + { + "epoch": 2.84, + "learning_rate": 2.6359679531623143e-05, + "loss": 0.0, + "step": 12275 + }, + { + "epoch": 2.84, + "learning_rate": 2.635775363993529e-05, + "loss": 0.0, + "step": 12276 + }, + { + "epoch": 2.84, + "learning_rate": 2.635582774824744e-05, + "loss": 0.0001, + "step": 12277 + }, + { + "epoch": 2.84, + "learning_rate": 2.635390185655959e-05, + "loss": 0.0, + "step": 12278 + }, + { + "epoch": 2.84, + "learning_rate": 2.635197596487174e-05, + "loss": 0.0, + "step": 12279 + }, + { + "epoch": 2.84, + "learning_rate": 2.6350050073183884e-05, + "loss": 0.0, + "step": 12280 + }, + { + "epoch": 2.84, + "learning_rate": 2.634812418149604e-05, + "loss": 0.0, + "step": 12281 + }, + { + "epoch": 2.84, + "learning_rate": 2.6346198289808182e-05, + "loss": 0.0, + "step": 12282 + }, + { + "epoch": 2.84, + "learning_rate": 2.634427239812033e-05, + "loss": 0.0, + "step": 12283 + }, + { + "epoch": 2.84, + "learning_rate": 2.634234650643248e-05, + "loss": 0.0, + "step": 12284 + }, + { + "epoch": 2.84, + "learning_rate": 2.6340420614744625e-05, + "loss": 0.0, + "step": 12285 + }, + { + "epoch": 2.84, + "learning_rate": 2.633849472305678e-05, + "loss": 0.0, + "step": 12286 + }, + { + "epoch": 2.84, + "learning_rate": 2.6336568831368923e-05, + "loss": 0.0001, + "step": 12287 + }, + { + "epoch": 2.84, + "learning_rate": 2.6334642939681074e-05, + "loss": 0.0, + "step": 12288 + }, + { + "epoch": 2.84, + "learning_rate": 2.633271704799322e-05, + "loss": 0.0, + "step": 12289 + }, + { + "epoch": 2.84, + "learning_rate": 2.6330791156305372e-05, + "loss": 0.0001, + "step": 12290 + }, + { + "epoch": 2.84, + "learning_rate": 2.632886526461752e-05, + "loss": 0.0, + "step": 12291 + }, + { + "epoch": 2.84, + "learning_rate": 2.6326939372929664e-05, + "loss": 0.0, + "step": 12292 + }, + { + "epoch": 2.84, + "learning_rate": 2.6325013481241818e-05, + "loss": 0.0, + "step": 12293 + }, + { + "epoch": 2.84, + "learning_rate": 2.6323087589553962e-05, + "loss": 0.0001, + "step": 12294 + }, + { + "epoch": 2.84, + "learning_rate": 2.6321161697866113e-05, + "loss": 0.0, + "step": 12295 + }, + { + "epoch": 2.84, + "learning_rate": 2.631923580617826e-05, + "loss": 0.0, + "step": 12296 + }, + { + "epoch": 2.84, + "learning_rate": 2.631730991449041e-05, + "loss": 0.0, + "step": 12297 + }, + { + "epoch": 2.84, + "learning_rate": 2.631538402280256e-05, + "loss": 0.0, + "step": 12298 + }, + { + "epoch": 2.84, + "learning_rate": 2.631345813111471e-05, + "loss": 0.0001, + "step": 12299 + }, + { + "epoch": 2.84, + "learning_rate": 2.6311532239426857e-05, + "loss": 0.0005, + "step": 12300 + }, + { + "epoch": 2.84, + "learning_rate": 2.6309606347739e-05, + "loss": 0.0, + "step": 12301 + }, + { + "epoch": 2.84, + "learning_rate": 2.6307680456051152e-05, + "loss": 0.0, + "step": 12302 + }, + { + "epoch": 2.84, + "learning_rate": 2.63057545643633e-05, + "loss": 0.0, + "step": 12303 + }, + { + "epoch": 2.84, + "learning_rate": 2.630382867267545e-05, + "loss": 0.0, + "step": 12304 + }, + { + "epoch": 2.84, + "learning_rate": 2.6301902780987598e-05, + "loss": 0.0, + "step": 12305 + }, + { + "epoch": 2.84, + "learning_rate": 2.629997688929975e-05, + "loss": 0.0, + "step": 12306 + }, + { + "epoch": 2.84, + "learning_rate": 2.6298050997611896e-05, + "loss": 0.0001, + "step": 12307 + }, + { + "epoch": 2.84, + "learning_rate": 2.6296125105924047e-05, + "loss": 0.0, + "step": 12308 + }, + { + "epoch": 2.84, + "learning_rate": 2.629419921423619e-05, + "loss": 0.0, + "step": 12309 + }, + { + "epoch": 2.84, + "learning_rate": 2.629227332254834e-05, + "loss": 0.0, + "step": 12310 + }, + { + "epoch": 2.85, + "learning_rate": 2.629034743086049e-05, + "loss": 0.0, + "step": 12311 + }, + { + "epoch": 2.85, + "learning_rate": 2.6288421539172637e-05, + "loss": 0.0, + "step": 12312 + }, + { + "epoch": 2.85, + "learning_rate": 2.6286495647484788e-05, + "loss": 0.0001, + "step": 12313 + }, + { + "epoch": 2.85, + "learning_rate": 2.6284569755796936e-05, + "loss": 0.0, + "step": 12314 + }, + { + "epoch": 2.85, + "learning_rate": 2.6282643864109086e-05, + "loss": 0.0, + "step": 12315 + }, + { + "epoch": 2.85, + "learning_rate": 2.628071797242123e-05, + "loss": 0.0, + "step": 12316 + }, + { + "epoch": 2.85, + "learning_rate": 2.6278792080733385e-05, + "loss": 0.0, + "step": 12317 + }, + { + "epoch": 2.85, + "learning_rate": 2.627686618904553e-05, + "loss": 0.0001, + "step": 12318 + }, + { + "epoch": 2.85, + "learning_rate": 2.6274940297357676e-05, + "loss": 0.0, + "step": 12319 + }, + { + "epoch": 2.85, + "learning_rate": 2.6273014405669827e-05, + "loss": 0.0002, + "step": 12320 + }, + { + "epoch": 2.85, + "learning_rate": 2.627108851398197e-05, + "loss": 0.0, + "step": 12321 + }, + { + "epoch": 2.85, + "learning_rate": 2.6269162622294126e-05, + "loss": 0.0, + "step": 12322 + }, + { + "epoch": 2.85, + "learning_rate": 2.626723673060627e-05, + "loss": 0.0, + "step": 12323 + }, + { + "epoch": 2.85, + "learning_rate": 2.626531083891842e-05, + "loss": 0.0, + "step": 12324 + }, + { + "epoch": 2.85, + "learning_rate": 2.6263384947230568e-05, + "loss": 0.0, + "step": 12325 + }, + { + "epoch": 2.85, + "learning_rate": 2.626145905554272e-05, + "loss": 0.0, + "step": 12326 + }, + { + "epoch": 2.85, + "learning_rate": 2.6259533163854866e-05, + "loss": 0.0, + "step": 12327 + }, + { + "epoch": 2.85, + "learning_rate": 2.625760727216701e-05, + "loss": 0.0, + "step": 12328 + }, + { + "epoch": 2.85, + "learning_rate": 2.6255681380479165e-05, + "loss": 0.0001, + "step": 12329 + }, + { + "epoch": 2.85, + "learning_rate": 2.625375548879131e-05, + "loss": 0.0001, + "step": 12330 + }, + { + "epoch": 2.85, + "learning_rate": 2.625182959710346e-05, + "loss": 0.0, + "step": 12331 + }, + { + "epoch": 2.85, + "learning_rate": 2.6249903705415607e-05, + "loss": 0.0, + "step": 12332 + }, + { + "epoch": 2.85, + "learning_rate": 2.6247977813727758e-05, + "loss": 0.0, + "step": 12333 + }, + { + "epoch": 2.85, + "learning_rate": 2.6246051922039905e-05, + "loss": 0.0, + "step": 12334 + }, + { + "epoch": 2.85, + "learning_rate": 2.6244126030352056e-05, + "loss": 0.0036, + "step": 12335 + }, + { + "epoch": 2.85, + "learning_rate": 2.6242200138664204e-05, + "loss": 0.0001, + "step": 12336 + }, + { + "epoch": 2.85, + "learning_rate": 2.6240274246976355e-05, + "loss": 0.0, + "step": 12337 + }, + { + "epoch": 2.85, + "learning_rate": 2.62383483552885e-05, + "loss": 0.0, + "step": 12338 + }, + { + "epoch": 2.85, + "learning_rate": 2.6236422463600646e-05, + "loss": 0.0, + "step": 12339 + }, + { + "epoch": 2.85, + "learning_rate": 2.6234496571912797e-05, + "loss": 0.0001, + "step": 12340 + }, + { + "epoch": 2.85, + "learning_rate": 2.6232570680224945e-05, + "loss": 0.0, + "step": 12341 + }, + { + "epoch": 2.85, + "learning_rate": 2.6230644788537095e-05, + "loss": 0.0, + "step": 12342 + }, + { + "epoch": 2.85, + "learning_rate": 2.6228718896849243e-05, + "loss": 0.0, + "step": 12343 + }, + { + "epoch": 2.85, + "learning_rate": 2.6226793005161394e-05, + "loss": 0.0001, + "step": 12344 + }, + { + "epoch": 2.85, + "learning_rate": 2.6224867113473538e-05, + "loss": 0.0, + "step": 12345 + }, + { + "epoch": 2.85, + "learning_rate": 2.6222941221785692e-05, + "loss": 0.0001, + "step": 12346 + }, + { + "epoch": 2.85, + "learning_rate": 2.6221015330097836e-05, + "loss": 0.0, + "step": 12347 + }, + { + "epoch": 2.85, + "learning_rate": 2.6219089438409984e-05, + "loss": 0.0001, + "step": 12348 + }, + { + "epoch": 2.85, + "learning_rate": 2.6217163546722135e-05, + "loss": 0.0, + "step": 12349 + }, + { + "epoch": 2.85, + "learning_rate": 2.6215237655034282e-05, + "loss": 0.0, + "step": 12350 + }, + { + "epoch": 2.85, + "learning_rate": 2.6213311763346433e-05, + "loss": 0.0, + "step": 12351 + }, + { + "epoch": 2.85, + "learning_rate": 2.6211385871658577e-05, + "loss": 0.0, + "step": 12352 + }, + { + "epoch": 2.85, + "learning_rate": 2.620945997997073e-05, + "loss": 0.0001, + "step": 12353 + }, + { + "epoch": 2.86, + "learning_rate": 2.6207534088282875e-05, + "loss": 0.0001, + "step": 12354 + }, + { + "epoch": 2.86, + "learning_rate": 2.6205608196595026e-05, + "loss": 0.0, + "step": 12355 + }, + { + "epoch": 2.86, + "learning_rate": 2.6203682304907174e-05, + "loss": 0.0, + "step": 12356 + }, + { + "epoch": 2.86, + "learning_rate": 2.6201756413219318e-05, + "loss": 0.0, + "step": 12357 + }, + { + "epoch": 2.86, + "learning_rate": 2.6199830521531472e-05, + "loss": 0.0, + "step": 12358 + }, + { + "epoch": 2.86, + "learning_rate": 2.6197904629843616e-05, + "loss": 0.0, + "step": 12359 + }, + { + "epoch": 2.86, + "learning_rate": 2.6195978738155767e-05, + "loss": 0.0, + "step": 12360 + }, + { + "epoch": 2.86, + "learning_rate": 2.6194052846467914e-05, + "loss": 0.0001, + "step": 12361 + }, + { + "epoch": 2.86, + "learning_rate": 2.6192126954780065e-05, + "loss": 0.0, + "step": 12362 + }, + { + "epoch": 2.86, + "learning_rate": 2.6190201063092213e-05, + "loss": 0.0, + "step": 12363 + }, + { + "epoch": 2.86, + "learning_rate": 2.6188275171404364e-05, + "loss": 0.0, + "step": 12364 + }, + { + "epoch": 2.86, + "learning_rate": 2.618634927971651e-05, + "loss": 0.0, + "step": 12365 + }, + { + "epoch": 2.86, + "learning_rate": 2.6184423388028655e-05, + "loss": 0.0, + "step": 12366 + }, + { + "epoch": 2.86, + "learning_rate": 2.6182497496340806e-05, + "loss": 0.0, + "step": 12367 + }, + { + "epoch": 2.86, + "learning_rate": 2.6180571604652954e-05, + "loss": 0.0, + "step": 12368 + }, + { + "epoch": 2.86, + "learning_rate": 2.6178645712965104e-05, + "loss": 0.0, + "step": 12369 + }, + { + "epoch": 2.86, + "learning_rate": 2.6176719821277252e-05, + "loss": 0.0, + "step": 12370 + }, + { + "epoch": 2.86, + "learning_rate": 2.6174793929589403e-05, + "loss": 0.0, + "step": 12371 + }, + { + "epoch": 2.86, + "learning_rate": 2.617286803790155e-05, + "loss": 0.0, + "step": 12372 + }, + { + "epoch": 2.86, + "learning_rate": 2.61709421462137e-05, + "loss": 0.0, + "step": 12373 + }, + { + "epoch": 2.86, + "learning_rate": 2.6169016254525845e-05, + "loss": 0.0, + "step": 12374 + }, + { + "epoch": 2.86, + "learning_rate": 2.6167090362837993e-05, + "loss": 0.0, + "step": 12375 + }, + { + "epoch": 2.86, + "learning_rate": 2.6165164471150144e-05, + "loss": 0.0, + "step": 12376 + }, + { + "epoch": 2.86, + "learning_rate": 2.616323857946229e-05, + "loss": 0.0, + "step": 12377 + }, + { + "epoch": 2.86, + "learning_rate": 2.6161312687774442e-05, + "loss": 0.0, + "step": 12378 + }, + { + "epoch": 2.86, + "learning_rate": 2.615938679608659e-05, + "loss": 0.0, + "step": 12379 + }, + { + "epoch": 2.86, + "learning_rate": 2.615746090439874e-05, + "loss": 0.0001, + "step": 12380 + }, + { + "epoch": 2.86, + "learning_rate": 2.6155535012710884e-05, + "loss": 0.0, + "step": 12381 + }, + { + "epoch": 2.86, + "learning_rate": 2.615360912102304e-05, + "loss": 0.0, + "step": 12382 + }, + { + "epoch": 2.86, + "learning_rate": 2.6151683229335183e-05, + "loss": 0.0, + "step": 12383 + }, + { + "epoch": 2.86, + "learning_rate": 2.614975733764733e-05, + "loss": 0.0, + "step": 12384 + }, + { + "epoch": 2.86, + "learning_rate": 2.614783144595948e-05, + "loss": 0.0, + "step": 12385 + }, + { + "epoch": 2.86, + "learning_rate": 2.614590555427163e-05, + "loss": 0.0, + "step": 12386 + }, + { + "epoch": 2.86, + "learning_rate": 2.614397966258378e-05, + "loss": 0.0, + "step": 12387 + }, + { + "epoch": 2.86, + "learning_rate": 2.6142053770895923e-05, + "loss": 0.0, + "step": 12388 + }, + { + "epoch": 2.86, + "learning_rate": 2.6140127879208078e-05, + "loss": 0.0, + "step": 12389 + }, + { + "epoch": 2.86, + "learning_rate": 2.6138201987520222e-05, + "loss": 0.0001, + "step": 12390 + }, + { + "epoch": 2.86, + "learning_rate": 2.6136276095832373e-05, + "loss": 0.0, + "step": 12391 + }, + { + "epoch": 2.86, + "learning_rate": 2.613435020414452e-05, + "loss": 0.0001, + "step": 12392 + }, + { + "epoch": 2.86, + "learning_rate": 2.613242431245667e-05, + "loss": 0.0001, + "step": 12393 + }, + { + "epoch": 2.86, + "learning_rate": 2.613049842076882e-05, + "loss": 0.0, + "step": 12394 + }, + { + "epoch": 2.86, + "learning_rate": 2.6128572529080963e-05, + "loss": 0.0, + "step": 12395 + }, + { + "epoch": 2.86, + "learning_rate": 2.6126646637393113e-05, + "loss": 0.0, + "step": 12396 + }, + { + "epoch": 2.87, + "learning_rate": 2.612472074570526e-05, + "loss": 0.0003, + "step": 12397 + }, + { + "epoch": 2.87, + "learning_rate": 2.6122794854017412e-05, + "loss": 0.0001, + "step": 12398 + }, + { + "epoch": 2.87, + "learning_rate": 2.612086896232956e-05, + "loss": 0.0, + "step": 12399 + }, + { + "epoch": 2.87, + "learning_rate": 2.611894307064171e-05, + "loss": 0.0, + "step": 12400 + }, + { + "epoch": 2.87, + "learning_rate": 2.6117017178953858e-05, + "loss": 0.0, + "step": 12401 + }, + { + "epoch": 2.87, + "learning_rate": 2.611509128726601e-05, + "loss": 0.0, + "step": 12402 + }, + { + "epoch": 2.87, + "learning_rate": 2.6113165395578153e-05, + "loss": 0.0, + "step": 12403 + }, + { + "epoch": 2.87, + "learning_rate": 2.61112395038903e-05, + "loss": 0.0, + "step": 12404 + }, + { + "epoch": 2.87, + "learning_rate": 2.610931361220245e-05, + "loss": 0.0, + "step": 12405 + }, + { + "epoch": 2.87, + "learning_rate": 2.61073877205146e-05, + "loss": 0.0003, + "step": 12406 + }, + { + "epoch": 2.87, + "learning_rate": 2.610546182882675e-05, + "loss": 0.0001, + "step": 12407 + }, + { + "epoch": 2.87, + "learning_rate": 2.6103535937138897e-05, + "loss": 0.0, + "step": 12408 + }, + { + "epoch": 2.87, + "learning_rate": 2.6101610045451048e-05, + "loss": 0.0001, + "step": 12409 + }, + { + "epoch": 2.87, + "learning_rate": 2.609968415376319e-05, + "loss": 0.0, + "step": 12410 + }, + { + "epoch": 2.87, + "learning_rate": 2.6097758262075346e-05, + "loss": 0.0, + "step": 12411 + }, + { + "epoch": 2.87, + "learning_rate": 2.609583237038749e-05, + "loss": 0.0, + "step": 12412 + }, + { + "epoch": 2.87, + "learning_rate": 2.6093906478699638e-05, + "loss": 0.0, + "step": 12413 + }, + { + "epoch": 2.87, + "learning_rate": 2.609198058701179e-05, + "loss": 0.0, + "step": 12414 + }, + { + "epoch": 2.87, + "learning_rate": 2.6090054695323936e-05, + "loss": 0.0, + "step": 12415 + }, + { + "epoch": 2.87, + "learning_rate": 2.6088128803636087e-05, + "loss": 0.0, + "step": 12416 + }, + { + "epoch": 2.87, + "learning_rate": 2.608620291194823e-05, + "loss": 0.0, + "step": 12417 + }, + { + "epoch": 2.87, + "learning_rate": 2.6084277020260385e-05, + "loss": 0.0, + "step": 12418 + }, + { + "epoch": 2.87, + "learning_rate": 2.608235112857253e-05, + "loss": 0.0, + "step": 12419 + }, + { + "epoch": 2.87, + "learning_rate": 2.608042523688468e-05, + "loss": 0.0, + "step": 12420 + }, + { + "epoch": 2.87, + "learning_rate": 2.6078499345196827e-05, + "loss": 0.0, + "step": 12421 + }, + { + "epoch": 2.87, + "learning_rate": 2.6076573453508975e-05, + "loss": 0.0, + "step": 12422 + }, + { + "epoch": 2.87, + "learning_rate": 2.6074647561821126e-05, + "loss": 0.0, + "step": 12423 + }, + { + "epoch": 2.87, + "learning_rate": 2.607272167013327e-05, + "loss": 0.0, + "step": 12424 + }, + { + "epoch": 2.87, + "learning_rate": 2.6070795778445424e-05, + "loss": 0.0, + "step": 12425 + }, + { + "epoch": 2.87, + "learning_rate": 2.6068869886757568e-05, + "loss": 0.0, + "step": 12426 + }, + { + "epoch": 2.87, + "learning_rate": 2.606694399506972e-05, + "loss": 0.0, + "step": 12427 + }, + { + "epoch": 2.87, + "learning_rate": 2.6065018103381867e-05, + "loss": 0.0, + "step": 12428 + }, + { + "epoch": 2.87, + "learning_rate": 2.6063092211694017e-05, + "loss": 0.0001, + "step": 12429 + }, + { + "epoch": 2.87, + "learning_rate": 2.6061166320006165e-05, + "loss": 0.0, + "step": 12430 + }, + { + "epoch": 2.87, + "learning_rate": 2.605924042831831e-05, + "loss": 0.0003, + "step": 12431 + }, + { + "epoch": 2.87, + "learning_rate": 2.6057314536630463e-05, + "loss": 0.0001, + "step": 12432 + }, + { + "epoch": 2.87, + "learning_rate": 2.6055388644942607e-05, + "loss": 0.0, + "step": 12433 + }, + { + "epoch": 2.87, + "learning_rate": 2.6053462753254758e-05, + "loss": 0.0, + "step": 12434 + }, + { + "epoch": 2.87, + "learning_rate": 2.6051536861566906e-05, + "loss": 0.0001, + "step": 12435 + }, + { + "epoch": 2.87, + "learning_rate": 2.6049610969879057e-05, + "loss": 0.0, + "step": 12436 + }, + { + "epoch": 2.87, + "learning_rate": 2.6047685078191204e-05, + "loss": 0.0, + "step": 12437 + }, + { + "epoch": 2.87, + "learning_rate": 2.6045759186503355e-05, + "loss": 0.0, + "step": 12438 + }, + { + "epoch": 2.87, + "learning_rate": 2.60438332948155e-05, + "loss": 0.0003, + "step": 12439 + }, + { + "epoch": 2.87, + "learning_rate": 2.6041907403127647e-05, + "loss": 0.0001, + "step": 12440 + }, + { + "epoch": 2.88, + "learning_rate": 2.6039981511439797e-05, + "loss": 0.0, + "step": 12441 + }, + { + "epoch": 2.88, + "learning_rate": 2.6038055619751945e-05, + "loss": 0.0, + "step": 12442 + }, + { + "epoch": 2.88, + "learning_rate": 2.6036129728064096e-05, + "loss": 0.0001, + "step": 12443 + }, + { + "epoch": 2.88, + "learning_rate": 2.6034203836376243e-05, + "loss": 0.0, + "step": 12444 + }, + { + "epoch": 2.88, + "learning_rate": 2.6032277944688394e-05, + "loss": 0.0, + "step": 12445 + }, + { + "epoch": 2.88, + "learning_rate": 2.6030352053000538e-05, + "loss": 0.0, + "step": 12446 + }, + { + "epoch": 2.88, + "learning_rate": 2.6028426161312692e-05, + "loss": 0.0, + "step": 12447 + }, + { + "epoch": 2.88, + "learning_rate": 2.6026500269624837e-05, + "loss": 0.0, + "step": 12448 + }, + { + "epoch": 2.88, + "learning_rate": 2.6024574377936987e-05, + "loss": 0.0008, + "step": 12449 + }, + { + "epoch": 2.88, + "learning_rate": 2.6022648486249135e-05, + "loss": 0.0001, + "step": 12450 + }, + { + "epoch": 2.88, + "learning_rate": 2.6020722594561282e-05, + "loss": 0.0, + "step": 12451 + }, + { + "epoch": 2.88, + "learning_rate": 2.6018796702873433e-05, + "loss": 0.0001, + "step": 12452 + }, + { + "epoch": 2.88, + "learning_rate": 2.6016870811185577e-05, + "loss": 0.0, + "step": 12453 + }, + { + "epoch": 2.88, + "learning_rate": 2.601494491949773e-05, + "loss": 0.0, + "step": 12454 + }, + { + "epoch": 2.88, + "learning_rate": 2.6013019027809876e-05, + "loss": 0.0, + "step": 12455 + }, + { + "epoch": 2.88, + "learning_rate": 2.6011093136122026e-05, + "loss": 0.0, + "step": 12456 + }, + { + "epoch": 2.88, + "learning_rate": 2.6009167244434174e-05, + "loss": 0.0, + "step": 12457 + }, + { + "epoch": 2.88, + "learning_rate": 2.6007241352746325e-05, + "loss": 0.0, + "step": 12458 + }, + { + "epoch": 2.88, + "learning_rate": 2.6005315461058472e-05, + "loss": 0.0001, + "step": 12459 + }, + { + "epoch": 2.88, + "learning_rate": 2.6003389569370616e-05, + "loss": 0.0, + "step": 12460 + }, + { + "epoch": 2.88, + "learning_rate": 2.600146367768277e-05, + "loss": 0.0, + "step": 12461 + }, + { + "epoch": 2.88, + "learning_rate": 2.5999537785994915e-05, + "loss": 0.0, + "step": 12462 + }, + { + "epoch": 2.88, + "learning_rate": 2.5997611894307066e-05, + "loss": 0.0, + "step": 12463 + }, + { + "epoch": 2.88, + "learning_rate": 2.5995686002619213e-05, + "loss": 0.0, + "step": 12464 + }, + { + "epoch": 2.88, + "learning_rate": 2.5993760110931364e-05, + "loss": 0.0, + "step": 12465 + }, + { + "epoch": 2.88, + "learning_rate": 2.599183421924351e-05, + "loss": 0.0003, + "step": 12466 + }, + { + "epoch": 2.88, + "learning_rate": 2.5989908327555662e-05, + "loss": 0.0, + "step": 12467 + }, + { + "epoch": 2.88, + "learning_rate": 2.598798243586781e-05, + "loss": 0.0001, + "step": 12468 + }, + { + "epoch": 2.88, + "learning_rate": 2.5986056544179954e-05, + "loss": 0.0, + "step": 12469 + }, + { + "epoch": 2.88, + "learning_rate": 2.5984130652492105e-05, + "loss": 0.0, + "step": 12470 + }, + { + "epoch": 2.88, + "learning_rate": 2.5982204760804252e-05, + "loss": 0.0, + "step": 12471 + }, + { + "epoch": 2.88, + "learning_rate": 2.5980278869116403e-05, + "loss": 0.0, + "step": 12472 + }, + { + "epoch": 2.88, + "learning_rate": 2.597835297742855e-05, + "loss": 0.0008, + "step": 12473 + }, + { + "epoch": 2.88, + "learning_rate": 2.59764270857407e-05, + "loss": 0.0, + "step": 12474 + }, + { + "epoch": 2.88, + "learning_rate": 2.5974501194052846e-05, + "loss": 0.0, + "step": 12475 + }, + { + "epoch": 2.88, + "learning_rate": 2.5972575302365e-05, + "loss": 0.0, + "step": 12476 + }, + { + "epoch": 2.88, + "learning_rate": 2.5970649410677144e-05, + "loss": 0.0, + "step": 12477 + }, + { + "epoch": 2.88, + "learning_rate": 2.596872351898929e-05, + "loss": 0.0001, + "step": 12478 + }, + { + "epoch": 2.88, + "learning_rate": 2.5966797627301442e-05, + "loss": 0.0, + "step": 12479 + }, + { + "epoch": 2.88, + "learning_rate": 2.596487173561359e-05, + "loss": 0.0, + "step": 12480 + }, + { + "epoch": 2.88, + "learning_rate": 2.596294584392574e-05, + "loss": 0.0, + "step": 12481 + }, + { + "epoch": 2.88, + "learning_rate": 2.5961019952237885e-05, + "loss": 0.0, + "step": 12482 + }, + { + "epoch": 2.88, + "learning_rate": 2.595909406055004e-05, + "loss": 0.0041, + "step": 12483 + }, + { + "epoch": 2.89, + "learning_rate": 2.5957168168862183e-05, + "loss": 0.0, + "step": 12484 + }, + { + "epoch": 2.89, + "learning_rate": 2.5955242277174334e-05, + "loss": 0.0001, + "step": 12485 + }, + { + "epoch": 2.89, + "learning_rate": 2.595331638548648e-05, + "loss": 0.0, + "step": 12486 + }, + { + "epoch": 2.89, + "learning_rate": 2.595139049379863e-05, + "loss": 0.0001, + "step": 12487 + }, + { + "epoch": 2.89, + "learning_rate": 2.594946460211078e-05, + "loss": 0.0, + "step": 12488 + }, + { + "epoch": 2.89, + "learning_rate": 2.5947538710422924e-05, + "loss": 0.0, + "step": 12489 + }, + { + "epoch": 2.89, + "learning_rate": 2.5945612818735078e-05, + "loss": 0.0, + "step": 12490 + }, + { + "epoch": 2.89, + "learning_rate": 2.5943686927047222e-05, + "loss": 0.0001, + "step": 12491 + }, + { + "epoch": 2.89, + "learning_rate": 2.5941761035359373e-05, + "loss": 0.0, + "step": 12492 + }, + { + "epoch": 2.89, + "learning_rate": 2.593983514367152e-05, + "loss": 0.0, + "step": 12493 + }, + { + "epoch": 2.89, + "learning_rate": 2.593790925198367e-05, + "loss": 0.0, + "step": 12494 + }, + { + "epoch": 2.89, + "learning_rate": 2.593598336029582e-05, + "loss": 0.0, + "step": 12495 + }, + { + "epoch": 2.89, + "learning_rate": 2.5934057468607963e-05, + "loss": 0.0, + "step": 12496 + }, + { + "epoch": 2.89, + "learning_rate": 2.5932131576920117e-05, + "loss": 0.0, + "step": 12497 + }, + { + "epoch": 2.89, + "learning_rate": 2.593020568523226e-05, + "loss": 0.0, + "step": 12498 + }, + { + "epoch": 2.89, + "learning_rate": 2.5928279793544412e-05, + "loss": 0.0, + "step": 12499 + }, + { + "epoch": 2.89, + "learning_rate": 2.592635390185656e-05, + "loss": 0.0, + "step": 12500 + }, + { + "epoch": 2.89, + "learning_rate": 2.592442801016871e-05, + "loss": 0.0, + "step": 12501 + }, + { + "epoch": 2.89, + "learning_rate": 2.5922502118480858e-05, + "loss": 0.0, + "step": 12502 + }, + { + "epoch": 2.89, + "learning_rate": 2.592057622679301e-05, + "loss": 0.0, + "step": 12503 + }, + { + "epoch": 2.89, + "learning_rate": 2.5918650335105156e-05, + "loss": 0.0, + "step": 12504 + }, + { + "epoch": 2.89, + "learning_rate": 2.5916724443417307e-05, + "loss": 0.0001, + "step": 12505 + }, + { + "epoch": 2.89, + "learning_rate": 2.591479855172945e-05, + "loss": 0.0, + "step": 12506 + }, + { + "epoch": 2.89, + "learning_rate": 2.59128726600416e-05, + "loss": 0.0, + "step": 12507 + }, + { + "epoch": 2.89, + "learning_rate": 2.591094676835375e-05, + "loss": 0.0, + "step": 12508 + }, + { + "epoch": 2.89, + "learning_rate": 2.5909020876665897e-05, + "loss": 0.0, + "step": 12509 + }, + { + "epoch": 2.89, + "learning_rate": 2.5907094984978048e-05, + "loss": 0.0, + "step": 12510 + }, + { + "epoch": 2.89, + "learning_rate": 2.5905169093290192e-05, + "loss": 0.0, + "step": 12511 + }, + { + "epoch": 2.89, + "learning_rate": 2.5903243201602346e-05, + "loss": 0.0006, + "step": 12512 + }, + { + "epoch": 2.89, + "learning_rate": 2.590131730991449e-05, + "loss": 0.0, + "step": 12513 + }, + { + "epoch": 2.89, + "learning_rate": 2.589939141822664e-05, + "loss": 0.0, + "step": 12514 + }, + { + "epoch": 2.89, + "learning_rate": 2.589746552653879e-05, + "loss": 0.0, + "step": 12515 + }, + { + "epoch": 2.89, + "learning_rate": 2.5895539634850936e-05, + "loss": 0.0, + "step": 12516 + }, + { + "epoch": 2.89, + "learning_rate": 2.5893613743163087e-05, + "loss": 0.0, + "step": 12517 + }, + { + "epoch": 2.89, + "learning_rate": 2.589168785147523e-05, + "loss": 0.0, + "step": 12518 + }, + { + "epoch": 2.89, + "learning_rate": 2.5889761959787385e-05, + "loss": 0.0, + "step": 12519 + }, + { + "epoch": 2.89, + "learning_rate": 2.588783606809953e-05, + "loss": 0.0, + "step": 12520 + }, + { + "epoch": 2.89, + "learning_rate": 2.588591017641168e-05, + "loss": 0.0, + "step": 12521 + }, + { + "epoch": 2.89, + "learning_rate": 2.5883984284723828e-05, + "loss": 0.0001, + "step": 12522 + }, + { + "epoch": 2.89, + "learning_rate": 2.588205839303598e-05, + "loss": 0.0, + "step": 12523 + }, + { + "epoch": 2.89, + "learning_rate": 2.5880132501348126e-05, + "loss": 0.0, + "step": 12524 + }, + { + "epoch": 2.89, + "learning_rate": 2.587820660966027e-05, + "loss": 0.0001, + "step": 12525 + }, + { + "epoch": 2.89, + "learning_rate": 2.5876280717972424e-05, + "loss": 0.0, + "step": 12526 + }, + { + "epoch": 2.9, + "learning_rate": 2.587435482628457e-05, + "loss": 0.0, + "step": 12527 + }, + { + "epoch": 2.9, + "learning_rate": 2.587242893459672e-05, + "loss": 0.0, + "step": 12528 + }, + { + "epoch": 2.9, + "learning_rate": 2.5870503042908867e-05, + "loss": 0.0, + "step": 12529 + }, + { + "epoch": 2.9, + "learning_rate": 2.5868577151221018e-05, + "loss": 0.0001, + "step": 12530 + }, + { + "epoch": 2.9, + "learning_rate": 2.5866651259533165e-05, + "loss": 0.0, + "step": 12531 + }, + { + "epoch": 2.9, + "learning_rate": 2.5864725367845316e-05, + "loss": 0.0008, + "step": 12532 + }, + { + "epoch": 2.9, + "learning_rate": 2.5862799476157464e-05, + "loss": 0.0, + "step": 12533 + }, + { + "epoch": 2.9, + "learning_rate": 2.5860873584469608e-05, + "loss": 0.0, + "step": 12534 + }, + { + "epoch": 2.9, + "learning_rate": 2.585894769278176e-05, + "loss": 0.0001, + "step": 12535 + }, + { + "epoch": 2.9, + "learning_rate": 2.5857021801093906e-05, + "loss": 0.0, + "step": 12536 + }, + { + "epoch": 2.9, + "learning_rate": 2.5855095909406057e-05, + "loss": 0.0, + "step": 12537 + }, + { + "epoch": 2.9, + "learning_rate": 2.5853170017718204e-05, + "loss": 0.0, + "step": 12538 + }, + { + "epoch": 2.9, + "learning_rate": 2.5851244126030355e-05, + "loss": 0.0, + "step": 12539 + }, + { + "epoch": 2.9, + "learning_rate": 2.5849318234342503e-05, + "loss": 0.0, + "step": 12540 + }, + { + "epoch": 2.9, + "learning_rate": 2.5847392342654654e-05, + "loss": 0.0, + "step": 12541 + }, + { + "epoch": 2.9, + "learning_rate": 2.5845466450966798e-05, + "loss": 0.0, + "step": 12542 + }, + { + "epoch": 2.9, + "learning_rate": 2.5843540559278945e-05, + "loss": 0.0, + "step": 12543 + }, + { + "epoch": 2.9, + "learning_rate": 2.5841614667591096e-05, + "loss": 0.0001, + "step": 12544 + }, + { + "epoch": 2.9, + "learning_rate": 2.5839688775903243e-05, + "loss": 0.0, + "step": 12545 + }, + { + "epoch": 2.9, + "learning_rate": 2.5837762884215394e-05, + "loss": 0.0, + "step": 12546 + }, + { + "epoch": 2.9, + "learning_rate": 2.583583699252754e-05, + "loss": 0.0, + "step": 12547 + }, + { + "epoch": 2.9, + "learning_rate": 2.5833911100839693e-05, + "loss": 0.0001, + "step": 12548 + }, + { + "epoch": 2.9, + "learning_rate": 2.5831985209151837e-05, + "loss": 0.0001, + "step": 12549 + }, + { + "epoch": 2.9, + "learning_rate": 2.5830059317463988e-05, + "loss": 0.0, + "step": 12550 + }, + { + "epoch": 2.9, + "learning_rate": 2.5828133425776135e-05, + "loss": 0.0, + "step": 12551 + }, + { + "epoch": 2.9, + "learning_rate": 2.5826207534088283e-05, + "loss": 0.0, + "step": 12552 + }, + { + "epoch": 2.9, + "learning_rate": 2.5824281642400433e-05, + "loss": 0.0, + "step": 12553 + }, + { + "epoch": 2.9, + "learning_rate": 2.5822355750712578e-05, + "loss": 0.0, + "step": 12554 + }, + { + "epoch": 2.9, + "learning_rate": 2.5820429859024732e-05, + "loss": 0.0, + "step": 12555 + }, + { + "epoch": 2.9, + "learning_rate": 2.5818503967336876e-05, + "loss": 0.0, + "step": 12556 + }, + { + "epoch": 2.9, + "learning_rate": 2.5816578075649027e-05, + "loss": 0.0004, + "step": 12557 + }, + { + "epoch": 2.9, + "learning_rate": 2.5814652183961174e-05, + "loss": 0.0, + "step": 12558 + }, + { + "epoch": 2.9, + "learning_rate": 2.5812726292273325e-05, + "loss": 0.0, + "step": 12559 + }, + { + "epoch": 2.9, + "learning_rate": 2.5810800400585473e-05, + "loss": 0.0, + "step": 12560 + }, + { + "epoch": 2.9, + "learning_rate": 2.5808874508897623e-05, + "loss": 0.0, + "step": 12561 + }, + { + "epoch": 2.9, + "learning_rate": 2.580694861720977e-05, + "loss": 0.0017, + "step": 12562 + }, + { + "epoch": 2.9, + "learning_rate": 2.5805022725521915e-05, + "loss": 0.0, + "step": 12563 + }, + { + "epoch": 2.9, + "learning_rate": 2.5803096833834066e-05, + "loss": 0.0, + "step": 12564 + }, + { + "epoch": 2.9, + "learning_rate": 2.5801170942146213e-05, + "loss": 0.0, + "step": 12565 + }, + { + "epoch": 2.9, + "learning_rate": 2.5799245050458364e-05, + "loss": 0.0, + "step": 12566 + }, + { + "epoch": 2.9, + "learning_rate": 2.5797319158770512e-05, + "loss": 0.0, + "step": 12567 + }, + { + "epoch": 2.9, + "learning_rate": 2.5795393267082663e-05, + "loss": 0.0, + "step": 12568 + }, + { + "epoch": 2.9, + "learning_rate": 2.579346737539481e-05, + "loss": 0.0, + "step": 12569 + }, + { + "epoch": 2.91, + "learning_rate": 2.579154148370696e-05, + "loss": 0.0, + "step": 12570 + }, + { + "epoch": 2.91, + "learning_rate": 2.5789615592019105e-05, + "loss": 0.0002, + "step": 12571 + }, + { + "epoch": 2.91, + "learning_rate": 2.5787689700331252e-05, + "loss": 0.0, + "step": 12572 + }, + { + "epoch": 2.91, + "learning_rate": 2.5785763808643403e-05, + "loss": 0.0, + "step": 12573 + }, + { + "epoch": 2.91, + "learning_rate": 2.578383791695555e-05, + "loss": 0.0, + "step": 12574 + }, + { + "epoch": 2.91, + "learning_rate": 2.5781912025267702e-05, + "loss": 0.0, + "step": 12575 + }, + { + "epoch": 2.91, + "learning_rate": 2.577998613357985e-05, + "loss": 0.0, + "step": 12576 + }, + { + "epoch": 2.91, + "learning_rate": 2.5778060241892e-05, + "loss": 0.0, + "step": 12577 + }, + { + "epoch": 2.91, + "learning_rate": 2.5776134350204144e-05, + "loss": 0.0001, + "step": 12578 + }, + { + "epoch": 2.91, + "learning_rate": 2.57742084585163e-05, + "loss": 0.0, + "step": 12579 + }, + { + "epoch": 2.91, + "learning_rate": 2.5772282566828442e-05, + "loss": 0.0, + "step": 12580 + }, + { + "epoch": 2.91, + "learning_rate": 2.577035667514059e-05, + "loss": 0.0, + "step": 12581 + }, + { + "epoch": 2.91, + "learning_rate": 2.576843078345274e-05, + "loss": 0.0, + "step": 12582 + }, + { + "epoch": 2.91, + "learning_rate": 2.5766504891764885e-05, + "loss": 0.0, + "step": 12583 + }, + { + "epoch": 2.91, + "learning_rate": 2.576457900007704e-05, + "loss": 0.0, + "step": 12584 + }, + { + "epoch": 2.91, + "learning_rate": 2.5762653108389183e-05, + "loss": 0.0, + "step": 12585 + }, + { + "epoch": 2.91, + "learning_rate": 2.5760727216701338e-05, + "loss": 0.0, + "step": 12586 + }, + { + "epoch": 2.91, + "learning_rate": 2.575880132501348e-05, + "loss": 0.0, + "step": 12587 + }, + { + "epoch": 2.91, + "learning_rate": 2.5756875433325632e-05, + "loss": 0.0, + "step": 12588 + }, + { + "epoch": 2.91, + "learning_rate": 2.575494954163778e-05, + "loss": 0.0, + "step": 12589 + }, + { + "epoch": 2.91, + "learning_rate": 2.5753023649949924e-05, + "loss": 0.0, + "step": 12590 + }, + { + "epoch": 2.91, + "learning_rate": 2.5751097758262078e-05, + "loss": 0.0, + "step": 12591 + }, + { + "epoch": 2.91, + "learning_rate": 2.5749171866574222e-05, + "loss": 0.0, + "step": 12592 + }, + { + "epoch": 2.91, + "learning_rate": 2.5747245974886373e-05, + "loss": 0.0, + "step": 12593 + }, + { + "epoch": 2.91, + "learning_rate": 2.574532008319852e-05, + "loss": 0.0, + "step": 12594 + }, + { + "epoch": 2.91, + "learning_rate": 2.574339419151067e-05, + "loss": 0.0, + "step": 12595 + }, + { + "epoch": 2.91, + "learning_rate": 2.574146829982282e-05, + "loss": 0.0, + "step": 12596 + }, + { + "epoch": 2.91, + "learning_rate": 2.573954240813497e-05, + "loss": 0.0, + "step": 12597 + }, + { + "epoch": 2.91, + "learning_rate": 2.5737616516447117e-05, + "loss": 0.0002, + "step": 12598 + }, + { + "epoch": 2.91, + "learning_rate": 2.573569062475926e-05, + "loss": 0.0, + "step": 12599 + }, + { + "epoch": 2.91, + "learning_rate": 2.5733764733071412e-05, + "loss": 0.0, + "step": 12600 + }, + { + "epoch": 2.91, + "learning_rate": 2.573183884138356e-05, + "loss": 0.0, + "step": 12601 + }, + { + "epoch": 2.91, + "learning_rate": 2.572991294969571e-05, + "loss": 0.0002, + "step": 12602 + }, + { + "epoch": 2.91, + "learning_rate": 2.5727987058007858e-05, + "loss": 0.0, + "step": 12603 + }, + { + "epoch": 2.91, + "learning_rate": 2.572606116632001e-05, + "loss": 0.0, + "step": 12604 + }, + { + "epoch": 2.91, + "learning_rate": 2.5724135274632157e-05, + "loss": 0.0001, + "step": 12605 + }, + { + "epoch": 2.91, + "learning_rate": 2.5722209382944307e-05, + "loss": 0.0, + "step": 12606 + }, + { + "epoch": 2.91, + "learning_rate": 2.572028349125645e-05, + "loss": 0.0, + "step": 12607 + }, + { + "epoch": 2.91, + "learning_rate": 2.57183575995686e-05, + "loss": 0.0, + "step": 12608 + }, + { + "epoch": 2.91, + "learning_rate": 2.571643170788075e-05, + "loss": 0.0, + "step": 12609 + }, + { + "epoch": 2.91, + "learning_rate": 2.5714505816192897e-05, + "loss": 0.0, + "step": 12610 + }, + { + "epoch": 2.91, + "learning_rate": 2.5712579924505048e-05, + "loss": 0.0, + "step": 12611 + }, + { + "epoch": 2.91, + "learning_rate": 2.5710654032817196e-05, + "loss": 0.0, + "step": 12612 + }, + { + "epoch": 2.91, + "learning_rate": 2.5708728141129347e-05, + "loss": 0.0, + "step": 12613 + }, + { + "epoch": 2.92, + "learning_rate": 2.570680224944149e-05, + "loss": 0.0, + "step": 12614 + }, + { + "epoch": 2.92, + "learning_rate": 2.5704876357753645e-05, + "loss": 0.0, + "step": 12615 + }, + { + "epoch": 2.92, + "learning_rate": 2.570295046606579e-05, + "loss": 0.0, + "step": 12616 + }, + { + "epoch": 2.92, + "learning_rate": 2.570102457437794e-05, + "loss": 0.0, + "step": 12617 + }, + { + "epoch": 2.92, + "learning_rate": 2.5699098682690087e-05, + "loss": 0.0, + "step": 12618 + }, + { + "epoch": 2.92, + "learning_rate": 2.569717279100223e-05, + "loss": 0.0001, + "step": 12619 + }, + { + "epoch": 2.92, + "learning_rate": 2.5695246899314386e-05, + "loss": 0.0001, + "step": 12620 + }, + { + "epoch": 2.92, + "learning_rate": 2.569332100762653e-05, + "loss": 0.0, + "step": 12621 + }, + { + "epoch": 2.92, + "learning_rate": 2.5691395115938684e-05, + "loss": 0.0, + "step": 12622 + }, + { + "epoch": 2.92, + "learning_rate": 2.5689469224250828e-05, + "loss": 0.0, + "step": 12623 + }, + { + "epoch": 2.92, + "learning_rate": 2.568754333256298e-05, + "loss": 0.0, + "step": 12624 + }, + { + "epoch": 2.92, + "learning_rate": 2.5685617440875126e-05, + "loss": 0.0, + "step": 12625 + }, + { + "epoch": 2.92, + "learning_rate": 2.5683691549187277e-05, + "loss": 0.0, + "step": 12626 + }, + { + "epoch": 2.92, + "learning_rate": 2.5681765657499425e-05, + "loss": 0.0, + "step": 12627 + }, + { + "epoch": 2.92, + "learning_rate": 2.567983976581157e-05, + "loss": 0.0, + "step": 12628 + }, + { + "epoch": 2.92, + "learning_rate": 2.567791387412372e-05, + "loss": 0.0, + "step": 12629 + }, + { + "epoch": 2.92, + "learning_rate": 2.5675987982435867e-05, + "loss": 0.0, + "step": 12630 + }, + { + "epoch": 2.92, + "learning_rate": 2.5674062090748018e-05, + "loss": 0.0, + "step": 12631 + }, + { + "epoch": 2.92, + "learning_rate": 2.5672136199060166e-05, + "loss": 0.0001, + "step": 12632 + }, + { + "epoch": 2.92, + "learning_rate": 2.5670210307372316e-05, + "loss": 0.0001, + "step": 12633 + }, + { + "epoch": 2.92, + "learning_rate": 2.5668284415684464e-05, + "loss": 0.0, + "step": 12634 + }, + { + "epoch": 2.92, + "learning_rate": 2.5666358523996615e-05, + "loss": 0.0, + "step": 12635 + }, + { + "epoch": 2.92, + "learning_rate": 2.566443263230876e-05, + "loss": 0.0, + "step": 12636 + }, + { + "epoch": 2.92, + "learning_rate": 2.5662506740620906e-05, + "loss": 0.0, + "step": 12637 + }, + { + "epoch": 2.92, + "learning_rate": 2.5660580848933057e-05, + "loss": 0.0, + "step": 12638 + }, + { + "epoch": 2.92, + "learning_rate": 2.5658654957245205e-05, + "loss": 0.0001, + "step": 12639 + }, + { + "epoch": 2.92, + "learning_rate": 2.5656729065557356e-05, + "loss": 0.0001, + "step": 12640 + }, + { + "epoch": 2.92, + "learning_rate": 2.5654803173869503e-05, + "loss": 0.0, + "step": 12641 + }, + { + "epoch": 2.92, + "learning_rate": 2.5652877282181654e-05, + "loss": 0.0, + "step": 12642 + }, + { + "epoch": 2.92, + "learning_rate": 2.5650951390493798e-05, + "loss": 0.0, + "step": 12643 + }, + { + "epoch": 2.92, + "learning_rate": 2.5649025498805952e-05, + "loss": 0.0, + "step": 12644 + }, + { + "epoch": 2.92, + "learning_rate": 2.5647099607118096e-05, + "loss": 0.0, + "step": 12645 + }, + { + "epoch": 2.92, + "learning_rate": 2.5645173715430244e-05, + "loss": 0.0, + "step": 12646 + }, + { + "epoch": 2.92, + "learning_rate": 2.5643247823742395e-05, + "loss": 0.0, + "step": 12647 + }, + { + "epoch": 2.92, + "learning_rate": 2.5641321932054542e-05, + "loss": 0.0, + "step": 12648 + }, + { + "epoch": 2.92, + "learning_rate": 2.5639396040366693e-05, + "loss": 0.0, + "step": 12649 + }, + { + "epoch": 2.92, + "learning_rate": 2.5637470148678837e-05, + "loss": 0.0, + "step": 12650 + }, + { + "epoch": 2.92, + "learning_rate": 2.563554425699099e-05, + "loss": 0.0, + "step": 12651 + }, + { + "epoch": 2.92, + "learning_rate": 2.5633618365303135e-05, + "loss": 0.0008, + "step": 12652 + }, + { + "epoch": 2.92, + "learning_rate": 2.5631692473615286e-05, + "loss": 0.0, + "step": 12653 + }, + { + "epoch": 2.92, + "learning_rate": 2.5629766581927434e-05, + "loss": 0.0, + "step": 12654 + }, + { + "epoch": 2.92, + "learning_rate": 2.562784069023958e-05, + "loss": 0.0, + "step": 12655 + }, + { + "epoch": 2.92, + "learning_rate": 2.5625914798551732e-05, + "loss": 0.0001, + "step": 12656 + }, + { + "epoch": 2.93, + "learning_rate": 2.5623988906863876e-05, + "loss": 0.0003, + "step": 12657 + }, + { + "epoch": 2.93, + "learning_rate": 2.562206301517603e-05, + "loss": 0.0, + "step": 12658 + }, + { + "epoch": 2.93, + "learning_rate": 2.5620137123488175e-05, + "loss": 0.0, + "step": 12659 + }, + { + "epoch": 2.93, + "learning_rate": 2.5618211231800325e-05, + "loss": 0.0, + "step": 12660 + }, + { + "epoch": 2.93, + "learning_rate": 2.5616285340112473e-05, + "loss": 0.0, + "step": 12661 + }, + { + "epoch": 2.93, + "learning_rate": 2.5614359448424624e-05, + "loss": 0.0, + "step": 12662 + }, + { + "epoch": 2.93, + "learning_rate": 2.561243355673677e-05, + "loss": 0.0, + "step": 12663 + }, + { + "epoch": 2.93, + "learning_rate": 2.5610507665048915e-05, + "loss": 0.0, + "step": 12664 + }, + { + "epoch": 2.93, + "learning_rate": 2.5608581773361066e-05, + "loss": 0.0, + "step": 12665 + }, + { + "epoch": 2.93, + "learning_rate": 2.5606655881673214e-05, + "loss": 0.0, + "step": 12666 + }, + { + "epoch": 2.93, + "learning_rate": 2.5604729989985365e-05, + "loss": 0.0, + "step": 12667 + }, + { + "epoch": 2.93, + "learning_rate": 2.5602804098297512e-05, + "loss": 0.0023, + "step": 12668 + }, + { + "epoch": 2.93, + "learning_rate": 2.5600878206609663e-05, + "loss": 0.0, + "step": 12669 + }, + { + "epoch": 2.93, + "learning_rate": 2.559895231492181e-05, + "loss": 0.0001, + "step": 12670 + }, + { + "epoch": 2.93, + "learning_rate": 2.559702642323396e-05, + "loss": 0.0001, + "step": 12671 + }, + { + "epoch": 2.93, + "learning_rate": 2.5595100531546105e-05, + "loss": 0.0, + "step": 12672 + }, + { + "epoch": 2.93, + "learning_rate": 2.559317463985826e-05, + "loss": 0.0, + "step": 12673 + }, + { + "epoch": 2.93, + "learning_rate": 2.5591248748170404e-05, + "loss": 0.0, + "step": 12674 + }, + { + "epoch": 2.93, + "learning_rate": 2.558932285648255e-05, + "loss": 0.0004, + "step": 12675 + }, + { + "epoch": 2.93, + "learning_rate": 2.5587396964794702e-05, + "loss": 0.0002, + "step": 12676 + }, + { + "epoch": 2.93, + "learning_rate": 2.558547107310685e-05, + "loss": 0.0, + "step": 12677 + }, + { + "epoch": 2.93, + "learning_rate": 2.5583545181419e-05, + "loss": 0.0, + "step": 12678 + }, + { + "epoch": 2.93, + "learning_rate": 2.5581619289731144e-05, + "loss": 0.0, + "step": 12679 + }, + { + "epoch": 2.93, + "learning_rate": 2.55796933980433e-05, + "loss": 0.0, + "step": 12680 + }, + { + "epoch": 2.93, + "learning_rate": 2.5577767506355443e-05, + "loss": 0.0, + "step": 12681 + }, + { + "epoch": 2.93, + "learning_rate": 2.5575841614667594e-05, + "loss": 0.0, + "step": 12682 + }, + { + "epoch": 2.93, + "learning_rate": 2.557391572297974e-05, + "loss": 0.0, + "step": 12683 + }, + { + "epoch": 2.93, + "learning_rate": 2.557198983129189e-05, + "loss": 0.0, + "step": 12684 + }, + { + "epoch": 2.93, + "learning_rate": 2.557006393960404e-05, + "loss": 0.0002, + "step": 12685 + }, + { + "epoch": 2.93, + "learning_rate": 2.5568138047916184e-05, + "loss": 0.0003, + "step": 12686 + }, + { + "epoch": 2.93, + "learning_rate": 2.5566212156228338e-05, + "loss": 0.0001, + "step": 12687 + }, + { + "epoch": 2.93, + "learning_rate": 2.5564286264540482e-05, + "loss": 0.0, + "step": 12688 + }, + { + "epoch": 2.93, + "learning_rate": 2.5562360372852633e-05, + "loss": 0.0, + "step": 12689 + }, + { + "epoch": 2.93, + "learning_rate": 2.556043448116478e-05, + "loss": 0.0, + "step": 12690 + }, + { + "epoch": 2.93, + "learning_rate": 2.555850858947693e-05, + "loss": 0.0, + "step": 12691 + }, + { + "epoch": 2.93, + "learning_rate": 2.555658269778908e-05, + "loss": 0.0, + "step": 12692 + }, + { + "epoch": 2.93, + "learning_rate": 2.5554656806101223e-05, + "loss": 0.0, + "step": 12693 + }, + { + "epoch": 2.93, + "learning_rate": 2.5552730914413377e-05, + "loss": 0.0, + "step": 12694 + }, + { + "epoch": 2.93, + "learning_rate": 2.555080502272552e-05, + "loss": 0.0, + "step": 12695 + }, + { + "epoch": 2.93, + "learning_rate": 2.5548879131037672e-05, + "loss": 0.0, + "step": 12696 + }, + { + "epoch": 2.93, + "learning_rate": 2.554695323934982e-05, + "loss": 0.0, + "step": 12697 + }, + { + "epoch": 2.93, + "learning_rate": 2.554502734766197e-05, + "loss": 0.0, + "step": 12698 + }, + { + "epoch": 2.93, + "learning_rate": 2.5543101455974118e-05, + "loss": 0.0, + "step": 12699 + }, + { + "epoch": 2.94, + "learning_rate": 2.554117556428627e-05, + "loss": 0.0, + "step": 12700 + }, + { + "epoch": 2.94, + "learning_rate": 2.5539249672598413e-05, + "loss": 0.0, + "step": 12701 + }, + { + "epoch": 2.94, + "learning_rate": 2.553732378091056e-05, + "loss": 0.0, + "step": 12702 + }, + { + "epoch": 2.94, + "learning_rate": 2.553539788922271e-05, + "loss": 0.0, + "step": 12703 + }, + { + "epoch": 2.94, + "learning_rate": 2.553347199753486e-05, + "loss": 0.0, + "step": 12704 + }, + { + "epoch": 2.94, + "learning_rate": 2.553154610584701e-05, + "loss": 0.0017, + "step": 12705 + }, + { + "epoch": 2.94, + "learning_rate": 2.5529620214159157e-05, + "loss": 0.0001, + "step": 12706 + }, + { + "epoch": 2.94, + "learning_rate": 2.5527694322471308e-05, + "loss": 0.0, + "step": 12707 + }, + { + "epoch": 2.94, + "learning_rate": 2.5525768430783452e-05, + "loss": 0.0, + "step": 12708 + }, + { + "epoch": 2.94, + "learning_rate": 2.5523842539095606e-05, + "loss": 0.0, + "step": 12709 + }, + { + "epoch": 2.94, + "learning_rate": 2.552191664740775e-05, + "loss": 0.0, + "step": 12710 + }, + { + "epoch": 2.94, + "learning_rate": 2.5519990755719898e-05, + "loss": 0.0, + "step": 12711 + }, + { + "epoch": 2.94, + "learning_rate": 2.551806486403205e-05, + "loss": 0.0, + "step": 12712 + }, + { + "epoch": 2.94, + "learning_rate": 2.5516138972344196e-05, + "loss": 0.0, + "step": 12713 + }, + { + "epoch": 2.94, + "learning_rate": 2.5514213080656347e-05, + "loss": 0.0, + "step": 12714 + }, + { + "epoch": 2.94, + "learning_rate": 2.551228718896849e-05, + "loss": 0.0, + "step": 12715 + }, + { + "epoch": 2.94, + "learning_rate": 2.5510361297280645e-05, + "loss": 0.0001, + "step": 12716 + }, + { + "epoch": 2.94, + "learning_rate": 2.550843540559279e-05, + "loss": 0.0, + "step": 12717 + }, + { + "epoch": 2.94, + "learning_rate": 2.550650951390494e-05, + "loss": 0.0, + "step": 12718 + }, + { + "epoch": 2.94, + "learning_rate": 2.5504583622217088e-05, + "loss": 0.0001, + "step": 12719 + }, + { + "epoch": 2.94, + "learning_rate": 2.5502657730529235e-05, + "loss": 0.0, + "step": 12720 + }, + { + "epoch": 2.94, + "learning_rate": 2.5500731838841386e-05, + "loss": 0.0, + "step": 12721 + }, + { + "epoch": 2.94, + "learning_rate": 2.549880594715353e-05, + "loss": 0.0, + "step": 12722 + }, + { + "epoch": 2.94, + "learning_rate": 2.5496880055465684e-05, + "loss": 0.0001, + "step": 12723 + }, + { + "epoch": 2.94, + "learning_rate": 2.549495416377783e-05, + "loss": 0.0, + "step": 12724 + }, + { + "epoch": 2.94, + "learning_rate": 2.549302827208998e-05, + "loss": 0.0, + "step": 12725 + }, + { + "epoch": 2.94, + "learning_rate": 2.5491102380402127e-05, + "loss": 0.0003, + "step": 12726 + }, + { + "epoch": 2.94, + "learning_rate": 2.5489176488714278e-05, + "loss": 0.0, + "step": 12727 + }, + { + "epoch": 2.94, + "learning_rate": 2.5487250597026425e-05, + "loss": 0.0, + "step": 12728 + }, + { + "epoch": 2.94, + "learning_rate": 2.5485324705338576e-05, + "loss": 0.0, + "step": 12729 + }, + { + "epoch": 2.94, + "learning_rate": 2.5483398813650723e-05, + "loss": 0.0005, + "step": 12730 + }, + { + "epoch": 2.94, + "learning_rate": 2.5481472921962867e-05, + "loss": 0.0, + "step": 12731 + }, + { + "epoch": 2.94, + "learning_rate": 2.547954703027502e-05, + "loss": 0.0, + "step": 12732 + }, + { + "epoch": 2.94, + "learning_rate": 2.5477621138587166e-05, + "loss": 0.0002, + "step": 12733 + }, + { + "epoch": 2.94, + "learning_rate": 2.5475695246899317e-05, + "loss": 0.0, + "step": 12734 + }, + { + "epoch": 2.94, + "learning_rate": 2.5473769355211464e-05, + "loss": 0.0, + "step": 12735 + }, + { + "epoch": 2.94, + "learning_rate": 2.5471843463523615e-05, + "loss": 0.0, + "step": 12736 + }, + { + "epoch": 2.94, + "learning_rate": 2.546991757183576e-05, + "loss": 0.0, + "step": 12737 + }, + { + "epoch": 2.94, + "learning_rate": 2.5467991680147913e-05, + "loss": 0.0, + "step": 12738 + }, + { + "epoch": 2.94, + "learning_rate": 2.5466065788460057e-05, + "loss": 0.0, + "step": 12739 + }, + { + "epoch": 2.94, + "learning_rate": 2.5464139896772205e-05, + "loss": 0.0, + "step": 12740 + }, + { + "epoch": 2.94, + "learning_rate": 2.5462214005084356e-05, + "loss": 0.0, + "step": 12741 + }, + { + "epoch": 2.94, + "learning_rate": 2.5460288113396503e-05, + "loss": 0.0, + "step": 12742 + }, + { + "epoch": 2.94, + "learning_rate": 2.5458362221708654e-05, + "loss": 0.0, + "step": 12743 + }, + { + "epoch": 2.95, + "learning_rate": 2.5456436330020798e-05, + "loss": 0.0, + "step": 12744 + }, + { + "epoch": 2.95, + "learning_rate": 2.5454510438332952e-05, + "loss": 0.0, + "step": 12745 + }, + { + "epoch": 2.95, + "learning_rate": 2.5452584546645097e-05, + "loss": 0.0, + "step": 12746 + }, + { + "epoch": 2.95, + "learning_rate": 2.5450658654957247e-05, + "loss": 0.0, + "step": 12747 + }, + { + "epoch": 2.95, + "learning_rate": 2.5448732763269395e-05, + "loss": 0.0, + "step": 12748 + }, + { + "epoch": 2.95, + "learning_rate": 2.5446806871581542e-05, + "loss": 0.0, + "step": 12749 + }, + { + "epoch": 2.95, + "learning_rate": 2.5444880979893693e-05, + "loss": 0.0, + "step": 12750 + }, + { + "epoch": 2.95, + "learning_rate": 2.5442955088205837e-05, + "loss": 0.0, + "step": 12751 + }, + { + "epoch": 2.95, + "learning_rate": 2.544102919651799e-05, + "loss": 0.0, + "step": 12752 + }, + { + "epoch": 2.95, + "learning_rate": 2.5439103304830136e-05, + "loss": 0.0, + "step": 12753 + }, + { + "epoch": 2.95, + "learning_rate": 2.5437177413142287e-05, + "loss": 0.0, + "step": 12754 + }, + { + "epoch": 2.95, + "learning_rate": 2.5435251521454434e-05, + "loss": 0.0, + "step": 12755 + }, + { + "epoch": 2.95, + "learning_rate": 2.5433325629766585e-05, + "loss": 0.0001, + "step": 12756 + }, + { + "epoch": 2.95, + "learning_rate": 2.5431399738078732e-05, + "loss": 0.0003, + "step": 12757 + }, + { + "epoch": 2.95, + "learning_rate": 2.5429473846390876e-05, + "loss": 0.0, + "step": 12758 + }, + { + "epoch": 2.95, + "learning_rate": 2.542754795470303e-05, + "loss": 0.0015, + "step": 12759 + }, + { + "epoch": 2.95, + "learning_rate": 2.5425622063015175e-05, + "loss": 0.0, + "step": 12760 + }, + { + "epoch": 2.95, + "learning_rate": 2.5423696171327326e-05, + "loss": 0.0, + "step": 12761 + }, + { + "epoch": 2.95, + "learning_rate": 2.5421770279639473e-05, + "loss": 0.0, + "step": 12762 + }, + { + "epoch": 2.95, + "learning_rate": 2.5419844387951624e-05, + "loss": 0.0, + "step": 12763 + }, + { + "epoch": 2.95, + "learning_rate": 2.541791849626377e-05, + "loss": 0.0, + "step": 12764 + }, + { + "epoch": 2.95, + "learning_rate": 2.5415992604575922e-05, + "loss": 0.0002, + "step": 12765 + }, + { + "epoch": 2.95, + "learning_rate": 2.541406671288807e-05, + "loss": 0.0002, + "step": 12766 + }, + { + "epoch": 2.95, + "learning_rate": 2.5412140821200214e-05, + "loss": 0.0001, + "step": 12767 + }, + { + "epoch": 2.95, + "learning_rate": 2.5410214929512365e-05, + "loss": 0.0, + "step": 12768 + }, + { + "epoch": 2.95, + "learning_rate": 2.5408289037824512e-05, + "loss": 0.0, + "step": 12769 + }, + { + "epoch": 2.95, + "learning_rate": 2.5406363146136663e-05, + "loss": 0.0, + "step": 12770 + }, + { + "epoch": 2.95, + "learning_rate": 2.540443725444881e-05, + "loss": 0.0, + "step": 12771 + }, + { + "epoch": 2.95, + "learning_rate": 2.540251136276096e-05, + "loss": 0.0001, + "step": 12772 + }, + { + "epoch": 2.95, + "learning_rate": 2.5400585471073106e-05, + "loss": 0.0, + "step": 12773 + }, + { + "epoch": 2.95, + "learning_rate": 2.539865957938526e-05, + "loss": 0.0, + "step": 12774 + }, + { + "epoch": 2.95, + "learning_rate": 2.5396733687697404e-05, + "loss": 0.0, + "step": 12775 + }, + { + "epoch": 2.95, + "learning_rate": 2.539480779600955e-05, + "loss": 0.0, + "step": 12776 + }, + { + "epoch": 2.95, + "learning_rate": 2.5392881904321702e-05, + "loss": 0.0, + "step": 12777 + }, + { + "epoch": 2.95, + "learning_rate": 2.539095601263385e-05, + "loss": 0.0, + "step": 12778 + }, + { + "epoch": 2.95, + "learning_rate": 2.5389030120946e-05, + "loss": 0.0, + "step": 12779 + }, + { + "epoch": 2.95, + "learning_rate": 2.5387104229258145e-05, + "loss": 0.0, + "step": 12780 + }, + { + "epoch": 2.95, + "learning_rate": 2.53851783375703e-05, + "loss": 0.0001, + "step": 12781 + }, + { + "epoch": 2.95, + "learning_rate": 2.5383252445882443e-05, + "loss": 0.0, + "step": 12782 + }, + { + "epoch": 2.95, + "learning_rate": 2.5381326554194594e-05, + "loss": 0.0, + "step": 12783 + }, + { + "epoch": 2.95, + "learning_rate": 2.537940066250674e-05, + "loss": 0.0, + "step": 12784 + }, + { + "epoch": 2.95, + "learning_rate": 2.5377474770818892e-05, + "loss": 0.0, + "step": 12785 + }, + { + "epoch": 2.95, + "learning_rate": 2.537554887913104e-05, + "loss": 0.0, + "step": 12786 + }, + { + "epoch": 2.96, + "learning_rate": 2.5373622987443184e-05, + "loss": 0.0, + "step": 12787 + }, + { + "epoch": 2.96, + "learning_rate": 2.5371697095755338e-05, + "loss": 0.0, + "step": 12788 + }, + { + "epoch": 2.96, + "learning_rate": 2.5369771204067482e-05, + "loss": 0.0, + "step": 12789 + }, + { + "epoch": 2.96, + "learning_rate": 2.5367845312379633e-05, + "loss": 0.0001, + "step": 12790 + }, + { + "epoch": 2.96, + "learning_rate": 2.536591942069178e-05, + "loss": 0.0, + "step": 12791 + }, + { + "epoch": 2.96, + "learning_rate": 2.536399352900393e-05, + "loss": 0.0, + "step": 12792 + }, + { + "epoch": 2.96, + "learning_rate": 2.536206763731608e-05, + "loss": 0.0001, + "step": 12793 + }, + { + "epoch": 2.96, + "learning_rate": 2.536014174562823e-05, + "loss": 0.0, + "step": 12794 + }, + { + "epoch": 2.96, + "learning_rate": 2.5358215853940377e-05, + "loss": 0.0, + "step": 12795 + }, + { + "epoch": 2.96, + "learning_rate": 2.535628996225252e-05, + "loss": 0.0, + "step": 12796 + }, + { + "epoch": 2.96, + "learning_rate": 2.5354364070564672e-05, + "loss": 0.0, + "step": 12797 + }, + { + "epoch": 2.96, + "learning_rate": 2.535243817887682e-05, + "loss": 0.0, + "step": 12798 + }, + { + "epoch": 2.96, + "learning_rate": 2.535051228718897e-05, + "loss": 0.0, + "step": 12799 + }, + { + "epoch": 2.96, + "learning_rate": 2.5348586395501118e-05, + "loss": 0.0, + "step": 12800 + }, + { + "epoch": 2.96, + "learning_rate": 2.534666050381327e-05, + "loss": 0.0, + "step": 12801 + }, + { + "epoch": 2.96, + "learning_rate": 2.5344734612125416e-05, + "loss": 0.0, + "step": 12802 + }, + { + "epoch": 2.96, + "learning_rate": 2.5342808720437567e-05, + "loss": 0.0, + "step": 12803 + }, + { + "epoch": 2.96, + "learning_rate": 2.534088282874971e-05, + "loss": 0.0, + "step": 12804 + }, + { + "epoch": 2.96, + "learning_rate": 2.533895693706186e-05, + "loss": 0.0, + "step": 12805 + }, + { + "epoch": 2.96, + "learning_rate": 2.533703104537401e-05, + "loss": 0.0, + "step": 12806 + }, + { + "epoch": 2.96, + "learning_rate": 2.5335105153686157e-05, + "loss": 0.0, + "step": 12807 + }, + { + "epoch": 2.96, + "learning_rate": 2.5333179261998308e-05, + "loss": 0.0, + "step": 12808 + }, + { + "epoch": 2.96, + "learning_rate": 2.5331253370310455e-05, + "loss": 0.0, + "step": 12809 + }, + { + "epoch": 2.96, + "learning_rate": 2.5329327478622606e-05, + "loss": 0.0001, + "step": 12810 + }, + { + "epoch": 2.96, + "learning_rate": 2.532740158693475e-05, + "loss": 0.0, + "step": 12811 + }, + { + "epoch": 2.96, + "learning_rate": 2.5325475695246905e-05, + "loss": 0.0, + "step": 12812 + }, + { + "epoch": 2.96, + "learning_rate": 2.532354980355905e-05, + "loss": 0.0, + "step": 12813 + }, + { + "epoch": 2.96, + "learning_rate": 2.5321623911871196e-05, + "loss": 0.0, + "step": 12814 + }, + { + "epoch": 2.96, + "learning_rate": 2.5319698020183347e-05, + "loss": 0.0, + "step": 12815 + }, + { + "epoch": 2.96, + "learning_rate": 2.531777212849549e-05, + "loss": 0.0001, + "step": 12816 + }, + { + "epoch": 2.96, + "learning_rate": 2.5315846236807645e-05, + "loss": 0.0, + "step": 12817 + }, + { + "epoch": 2.96, + "learning_rate": 2.531392034511979e-05, + "loss": 0.0002, + "step": 12818 + }, + { + "epoch": 2.96, + "learning_rate": 2.531199445343194e-05, + "loss": 0.0001, + "step": 12819 + }, + { + "epoch": 2.96, + "learning_rate": 2.5310068561744088e-05, + "loss": 0.0, + "step": 12820 + }, + { + "epoch": 2.96, + "learning_rate": 2.530814267005624e-05, + "loss": 0.0, + "step": 12821 + }, + { + "epoch": 2.96, + "learning_rate": 2.5306216778368386e-05, + "loss": 0.0, + "step": 12822 + }, + { + "epoch": 2.96, + "learning_rate": 2.530429088668053e-05, + "loss": 0.0, + "step": 12823 + }, + { + "epoch": 2.96, + "learning_rate": 2.5302364994992685e-05, + "loss": 0.0, + "step": 12824 + }, + { + "epoch": 2.96, + "learning_rate": 2.530043910330483e-05, + "loss": 0.0, + "step": 12825 + }, + { + "epoch": 2.96, + "learning_rate": 2.529851321161698e-05, + "loss": 0.0, + "step": 12826 + }, + { + "epoch": 2.96, + "learning_rate": 2.5296587319929127e-05, + "loss": 0.0003, + "step": 12827 + }, + { + "epoch": 2.96, + "learning_rate": 2.5294661428241278e-05, + "loss": 0.0, + "step": 12828 + }, + { + "epoch": 2.96, + "learning_rate": 2.5292735536553425e-05, + "loss": 0.0, + "step": 12829 + }, + { + "epoch": 2.97, + "learning_rate": 2.5290809644865576e-05, + "loss": 0.0, + "step": 12830 + }, + { + "epoch": 2.97, + "learning_rate": 2.5288883753177724e-05, + "loss": 0.0, + "step": 12831 + }, + { + "epoch": 2.97, + "learning_rate": 2.5286957861489868e-05, + "loss": 0.0, + "step": 12832 + }, + { + "epoch": 2.97, + "learning_rate": 2.528503196980202e-05, + "loss": 0.0002, + "step": 12833 + }, + { + "epoch": 2.97, + "learning_rate": 2.5283106078114166e-05, + "loss": 0.0001, + "step": 12834 + }, + { + "epoch": 2.97, + "learning_rate": 2.5281180186426317e-05, + "loss": 0.0, + "step": 12835 + }, + { + "epoch": 2.97, + "learning_rate": 2.5279254294738464e-05, + "loss": 0.0, + "step": 12836 + }, + { + "epoch": 2.97, + "learning_rate": 2.5277328403050615e-05, + "loss": 0.0, + "step": 12837 + }, + { + "epoch": 2.97, + "learning_rate": 2.5275402511362763e-05, + "loss": 0.0, + "step": 12838 + }, + { + "epoch": 2.97, + "learning_rate": 2.5273476619674914e-05, + "loss": 0.0, + "step": 12839 + }, + { + "epoch": 2.97, + "learning_rate": 2.5271550727987058e-05, + "loss": 0.0, + "step": 12840 + }, + { + "epoch": 2.97, + "learning_rate": 2.5269624836299212e-05, + "loss": 0.0, + "step": 12841 + }, + { + "epoch": 2.97, + "learning_rate": 2.5267698944611356e-05, + "loss": 0.0, + "step": 12842 + }, + { + "epoch": 2.97, + "learning_rate": 2.5265773052923504e-05, + "loss": 0.0, + "step": 12843 + }, + { + "epoch": 2.97, + "learning_rate": 2.5263847161235654e-05, + "loss": 0.0, + "step": 12844 + }, + { + "epoch": 2.97, + "learning_rate": 2.5261921269547802e-05, + "loss": 0.0, + "step": 12845 + }, + { + "epoch": 2.97, + "learning_rate": 2.5259995377859953e-05, + "loss": 0.0, + "step": 12846 + }, + { + "epoch": 2.97, + "learning_rate": 2.5258069486172097e-05, + "loss": 0.0001, + "step": 12847 + }, + { + "epoch": 2.97, + "learning_rate": 2.525614359448425e-05, + "loss": 0.0, + "step": 12848 + }, + { + "epoch": 2.97, + "learning_rate": 2.5254217702796395e-05, + "loss": 0.0, + "step": 12849 + }, + { + "epoch": 2.97, + "learning_rate": 2.5252291811108546e-05, + "loss": 0.0, + "step": 12850 + }, + { + "epoch": 2.97, + "learning_rate": 2.5250365919420694e-05, + "loss": 0.0, + "step": 12851 + }, + { + "epoch": 2.97, + "learning_rate": 2.5248440027732838e-05, + "loss": 0.0, + "step": 12852 + }, + { + "epoch": 2.97, + "learning_rate": 2.5246514136044992e-05, + "loss": 0.0, + "step": 12853 + }, + { + "epoch": 2.97, + "learning_rate": 2.5244588244357136e-05, + "loss": 0.0, + "step": 12854 + }, + { + "epoch": 2.97, + "learning_rate": 2.5242662352669287e-05, + "loss": 0.0, + "step": 12855 + }, + { + "epoch": 2.97, + "learning_rate": 2.5240736460981434e-05, + "loss": 0.0, + "step": 12856 + }, + { + "epoch": 2.97, + "learning_rate": 2.5238810569293585e-05, + "loss": 0.0, + "step": 12857 + }, + { + "epoch": 2.97, + "learning_rate": 2.5236884677605733e-05, + "loss": 0.0, + "step": 12858 + }, + { + "epoch": 2.97, + "learning_rate": 2.5234958785917884e-05, + "loss": 0.0, + "step": 12859 + }, + { + "epoch": 2.97, + "learning_rate": 2.523303289423003e-05, + "loss": 0.0, + "step": 12860 + }, + { + "epoch": 2.97, + "learning_rate": 2.5231107002542175e-05, + "loss": 0.0007, + "step": 12861 + }, + { + "epoch": 2.97, + "learning_rate": 2.5229181110854326e-05, + "loss": 0.0, + "step": 12862 + }, + { + "epoch": 2.97, + "learning_rate": 2.5227255219166473e-05, + "loss": 0.0, + "step": 12863 + }, + { + "epoch": 2.97, + "learning_rate": 2.5225329327478624e-05, + "loss": 0.0, + "step": 12864 + }, + { + "epoch": 2.97, + "learning_rate": 2.5223403435790772e-05, + "loss": 0.0, + "step": 12865 + }, + { + "epoch": 2.97, + "learning_rate": 2.5221477544102923e-05, + "loss": 0.0, + "step": 12866 + }, + { + "epoch": 2.97, + "learning_rate": 2.521955165241507e-05, + "loss": 0.0001, + "step": 12867 + }, + { + "epoch": 2.97, + "learning_rate": 2.521762576072722e-05, + "loss": 0.0, + "step": 12868 + }, + { + "epoch": 2.97, + "learning_rate": 2.5215699869039365e-05, + "loss": 0.0, + "step": 12869 + }, + { + "epoch": 2.97, + "learning_rate": 2.5213773977351513e-05, + "loss": 0.0, + "step": 12870 + }, + { + "epoch": 2.97, + "learning_rate": 2.5211848085663663e-05, + "loss": 0.0, + "step": 12871 + }, + { + "epoch": 2.97, + "learning_rate": 2.520992219397581e-05, + "loss": 0.0, + "step": 12872 + }, + { + "epoch": 2.98, + "learning_rate": 2.5207996302287962e-05, + "loss": 0.0, + "step": 12873 + }, + { + "epoch": 2.98, + "learning_rate": 2.520607041060011e-05, + "loss": 0.0001, + "step": 12874 + }, + { + "epoch": 2.98, + "learning_rate": 2.520414451891226e-05, + "loss": 0.0, + "step": 12875 + }, + { + "epoch": 2.98, + "learning_rate": 2.5202218627224404e-05, + "loss": 0.0, + "step": 12876 + }, + { + "epoch": 2.98, + "learning_rate": 2.520029273553656e-05, + "loss": 0.0, + "step": 12877 + }, + { + "epoch": 2.98, + "learning_rate": 2.5198366843848703e-05, + "loss": 0.0, + "step": 12878 + }, + { + "epoch": 2.98, + "learning_rate": 2.519644095216085e-05, + "loss": 0.0, + "step": 12879 + }, + { + "epoch": 2.98, + "learning_rate": 2.5194515060473e-05, + "loss": 0.0, + "step": 12880 + }, + { + "epoch": 2.98, + "learning_rate": 2.519258916878515e-05, + "loss": 0.0, + "step": 12881 + }, + { + "epoch": 2.98, + "learning_rate": 2.51906632770973e-05, + "loss": 0.0, + "step": 12882 + }, + { + "epoch": 2.98, + "learning_rate": 2.5188737385409443e-05, + "loss": 0.0, + "step": 12883 + }, + { + "epoch": 2.98, + "learning_rate": 2.5186811493721598e-05, + "loss": 0.0, + "step": 12884 + }, + { + "epoch": 2.98, + "learning_rate": 2.518488560203374e-05, + "loss": 0.0, + "step": 12885 + }, + { + "epoch": 2.98, + "learning_rate": 2.5182959710345893e-05, + "loss": 0.0, + "step": 12886 + }, + { + "epoch": 2.98, + "learning_rate": 2.518103381865804e-05, + "loss": 0.0, + "step": 12887 + }, + { + "epoch": 2.98, + "learning_rate": 2.5179107926970184e-05, + "loss": 0.0, + "step": 12888 + }, + { + "epoch": 2.98, + "learning_rate": 2.517718203528234e-05, + "loss": 0.0, + "step": 12889 + }, + { + "epoch": 2.98, + "learning_rate": 2.5175256143594482e-05, + "loss": 0.0, + "step": 12890 + }, + { + "epoch": 2.98, + "learning_rate": 2.5173330251906633e-05, + "loss": 0.0, + "step": 12891 + }, + { + "epoch": 2.98, + "learning_rate": 2.517140436021878e-05, + "loss": 0.0, + "step": 12892 + }, + { + "epoch": 2.98, + "learning_rate": 2.516947846853093e-05, + "loss": 0.0, + "step": 12893 + }, + { + "epoch": 2.98, + "learning_rate": 2.516755257684308e-05, + "loss": 0.0001, + "step": 12894 + }, + { + "epoch": 2.98, + "learning_rate": 2.516562668515523e-05, + "loss": 0.0, + "step": 12895 + }, + { + "epoch": 2.98, + "learning_rate": 2.5163700793467377e-05, + "loss": 0.0, + "step": 12896 + }, + { + "epoch": 2.98, + "learning_rate": 2.516177490177953e-05, + "loss": 0.0, + "step": 12897 + }, + { + "epoch": 2.98, + "learning_rate": 2.5159849010091672e-05, + "loss": 0.0, + "step": 12898 + }, + { + "epoch": 2.98, + "learning_rate": 2.515792311840382e-05, + "loss": 0.0, + "step": 12899 + }, + { + "epoch": 2.98, + "learning_rate": 2.515599722671597e-05, + "loss": 0.0, + "step": 12900 + }, + { + "epoch": 2.98, + "learning_rate": 2.5154071335028118e-05, + "loss": 0.0, + "step": 12901 + }, + { + "epoch": 2.98, + "learning_rate": 2.515214544334027e-05, + "loss": 0.0, + "step": 12902 + }, + { + "epoch": 2.98, + "learning_rate": 2.5150219551652417e-05, + "loss": 0.0001, + "step": 12903 + }, + { + "epoch": 2.98, + "learning_rate": 2.5148293659964567e-05, + "loss": 0.0, + "step": 12904 + }, + { + "epoch": 2.98, + "learning_rate": 2.514636776827671e-05, + "loss": 0.0, + "step": 12905 + }, + { + "epoch": 2.98, + "learning_rate": 2.5144441876588866e-05, + "loss": 0.0, + "step": 12906 + }, + { + "epoch": 2.98, + "learning_rate": 2.514251598490101e-05, + "loss": 0.0, + "step": 12907 + }, + { + "epoch": 2.98, + "learning_rate": 2.5140590093213157e-05, + "loss": 0.0, + "step": 12908 + }, + { + "epoch": 2.98, + "learning_rate": 2.5138664201525308e-05, + "loss": 0.0, + "step": 12909 + }, + { + "epoch": 2.98, + "learning_rate": 2.5136738309837456e-05, + "loss": 0.0, + "step": 12910 + }, + { + "epoch": 2.98, + "learning_rate": 2.5134812418149607e-05, + "loss": 0.0, + "step": 12911 + }, + { + "epoch": 2.98, + "learning_rate": 2.513288652646175e-05, + "loss": 0.0, + "step": 12912 + }, + { + "epoch": 2.98, + "learning_rate": 2.5130960634773905e-05, + "loss": 0.0, + "step": 12913 + }, + { + "epoch": 2.98, + "learning_rate": 2.512903474308605e-05, + "loss": 0.0, + "step": 12914 + }, + { + "epoch": 2.98, + "learning_rate": 2.51271088513982e-05, + "loss": 0.0001, + "step": 12915 + }, + { + "epoch": 2.98, + "learning_rate": 2.5125182959710347e-05, + "loss": 0.0, + "step": 12916 + }, + { + "epoch": 2.99, + "learning_rate": 2.5123257068022495e-05, + "loss": 0.0, + "step": 12917 + }, + { + "epoch": 2.99, + "learning_rate": 2.5121331176334646e-05, + "loss": 0.0, + "step": 12918 + }, + { + "epoch": 2.99, + "learning_rate": 2.511940528464679e-05, + "loss": 0.0001, + "step": 12919 + }, + { + "epoch": 2.99, + "learning_rate": 2.5117479392958944e-05, + "loss": 0.0, + "step": 12920 + }, + { + "epoch": 2.99, + "learning_rate": 2.5115553501271088e-05, + "loss": 0.0, + "step": 12921 + }, + { + "epoch": 2.99, + "learning_rate": 2.511362760958324e-05, + "loss": 0.0, + "step": 12922 + }, + { + "epoch": 2.99, + "learning_rate": 2.5111701717895387e-05, + "loss": 0.0, + "step": 12923 + }, + { + "epoch": 2.99, + "learning_rate": 2.5109775826207537e-05, + "loss": 0.0, + "step": 12924 + }, + { + "epoch": 2.99, + "learning_rate": 2.5107849934519685e-05, + "loss": 0.0001, + "step": 12925 + }, + { + "epoch": 2.99, + "learning_rate": 2.510592404283183e-05, + "loss": 0.0, + "step": 12926 + }, + { + "epoch": 2.99, + "learning_rate": 2.510399815114398e-05, + "loss": 0.0, + "step": 12927 + }, + { + "epoch": 2.99, + "learning_rate": 2.5102072259456127e-05, + "loss": 0.0, + "step": 12928 + }, + { + "epoch": 2.99, + "learning_rate": 2.5100146367768278e-05, + "loss": 0.0002, + "step": 12929 + }, + { + "epoch": 2.99, + "learning_rate": 2.5098220476080426e-05, + "loss": 0.0, + "step": 12930 + }, + { + "epoch": 2.99, + "learning_rate": 2.5096294584392576e-05, + "loss": 0.0, + "step": 12931 + }, + { + "epoch": 2.99, + "learning_rate": 2.5094368692704724e-05, + "loss": 0.0, + "step": 12932 + }, + { + "epoch": 2.99, + "learning_rate": 2.5092442801016875e-05, + "loss": 0.0001, + "step": 12933 + }, + { + "epoch": 2.99, + "learning_rate": 2.509051690932902e-05, + "loss": 0.0, + "step": 12934 + }, + { + "epoch": 2.99, + "learning_rate": 2.5088591017641166e-05, + "loss": 0.0, + "step": 12935 + }, + { + "epoch": 2.99, + "learning_rate": 2.5086665125953317e-05, + "loss": 0.0004, + "step": 12936 + }, + { + "epoch": 2.99, + "learning_rate": 2.5084739234265465e-05, + "loss": 0.0, + "step": 12937 + }, + { + "epoch": 2.99, + "learning_rate": 2.5082813342577616e-05, + "loss": 0.0, + "step": 12938 + }, + { + "epoch": 2.99, + "learning_rate": 2.5080887450889763e-05, + "loss": 0.0, + "step": 12939 + }, + { + "epoch": 2.99, + "learning_rate": 2.5078961559201914e-05, + "loss": 0.0, + "step": 12940 + }, + { + "epoch": 2.99, + "learning_rate": 2.5077035667514058e-05, + "loss": 0.0001, + "step": 12941 + }, + { + "epoch": 2.99, + "learning_rate": 2.5075109775826212e-05, + "loss": 0.0, + "step": 12942 + }, + { + "epoch": 2.99, + "learning_rate": 2.5073183884138356e-05, + "loss": 0.0003, + "step": 12943 + }, + { + "epoch": 2.99, + "learning_rate": 2.5071257992450504e-05, + "loss": 0.0001, + "step": 12944 + }, + { + "epoch": 2.99, + "learning_rate": 2.5069332100762655e-05, + "loss": 0.0, + "step": 12945 + }, + { + "epoch": 2.99, + "learning_rate": 2.5067406209074802e-05, + "loss": 0.0, + "step": 12946 + }, + { + "epoch": 2.99, + "learning_rate": 2.5065480317386953e-05, + "loss": 0.0001, + "step": 12947 + }, + { + "epoch": 2.99, + "learning_rate": 2.5063554425699097e-05, + "loss": 0.0, + "step": 12948 + }, + { + "epoch": 2.99, + "learning_rate": 2.506162853401125e-05, + "loss": 0.0, + "step": 12949 + }, + { + "epoch": 2.99, + "learning_rate": 2.5059702642323396e-05, + "loss": 0.0, + "step": 12950 + }, + { + "epoch": 2.99, + "learning_rate": 2.5057776750635546e-05, + "loss": 0.0001, + "step": 12951 + }, + { + "epoch": 2.99, + "learning_rate": 2.5055850858947694e-05, + "loss": 0.0, + "step": 12952 + }, + { + "epoch": 2.99, + "learning_rate": 2.5053924967259845e-05, + "loss": 0.0, + "step": 12953 + }, + { + "epoch": 2.99, + "learning_rate": 2.5051999075571992e-05, + "loss": 0.0001, + "step": 12954 + }, + { + "epoch": 2.99, + "learning_rate": 2.5050073183884136e-05, + "loss": 0.0, + "step": 12955 + }, + { + "epoch": 2.99, + "learning_rate": 2.504814729219629e-05, + "loss": 0.0, + "step": 12956 + }, + { + "epoch": 2.99, + "learning_rate": 2.5046221400508435e-05, + "loss": 0.0003, + "step": 12957 + }, + { + "epoch": 2.99, + "learning_rate": 2.5044295508820585e-05, + "loss": 0.0, + "step": 12958 + }, + { + "epoch": 2.99, + "learning_rate": 2.5042369617132733e-05, + "loss": 0.0, + "step": 12959 + }, + { + "epoch": 3.0, + "learning_rate": 2.5040443725444884e-05, + "loss": 0.0, + "step": 12960 + }, + { + "epoch": 3.0, + "learning_rate": 2.503851783375703e-05, + "loss": 0.0, + "step": 12961 + }, + { + "epoch": 3.0, + "learning_rate": 2.5036591942069182e-05, + "loss": 0.0, + "step": 12962 + }, + { + "epoch": 3.0, + "learning_rate": 2.5034666050381326e-05, + "loss": 0.0001, + "step": 12963 + }, + { + "epoch": 3.0, + "learning_rate": 2.5032740158693474e-05, + "loss": 0.0, + "step": 12964 + }, + { + "epoch": 3.0, + "learning_rate": 2.5030814267005625e-05, + "loss": 0.0, + "step": 12965 + }, + { + "epoch": 3.0, + "learning_rate": 2.5028888375317772e-05, + "loss": 0.0, + "step": 12966 + }, + { + "epoch": 3.0, + "learning_rate": 2.5026962483629923e-05, + "loss": 0.0, + "step": 12967 + }, + { + "epoch": 3.0, + "learning_rate": 2.502503659194207e-05, + "loss": 0.0004, + "step": 12968 + }, + { + "epoch": 3.0, + "learning_rate": 2.502311070025422e-05, + "loss": 0.0, + "step": 12969 + }, + { + "epoch": 3.0, + "learning_rate": 2.5021184808566365e-05, + "loss": 0.0, + "step": 12970 + }, + { + "epoch": 3.0, + "learning_rate": 2.501925891687852e-05, + "loss": 0.0, + "step": 12971 + }, + { + "epoch": 3.0, + "learning_rate": 2.5017333025190664e-05, + "loss": 0.0, + "step": 12972 + }, + { + "epoch": 3.0, + "learning_rate": 2.501540713350281e-05, + "loss": 0.0, + "step": 12973 + }, + { + "epoch": 3.0, + "learning_rate": 2.5013481241814962e-05, + "loss": 0.0, + "step": 12974 + }, + { + "epoch": 3.0, + "learning_rate": 2.501155535012711e-05, + "loss": 0.0008, + "step": 12975 + }, + { + "epoch": 3.0, + "learning_rate": 2.500962945843926e-05, + "loss": 0.0, + "step": 12976 + }, + { + "epoch": 3.0, + "learning_rate": 2.5007703566751405e-05, + "loss": 0.0, + "step": 12977 + }, + { + "epoch": 3.0, + "learning_rate": 2.500577767506356e-05, + "loss": 0.0, + "step": 12978 + }, + { + "epoch": 3.0, + "learning_rate": 2.5003851783375703e-05, + "loss": 0.0, + "step": 12979 + }, + { + "epoch": 3.0, + "learning_rate": 2.5001925891687854e-05, + "loss": 0.0, + "step": 12980 + }, + { + "epoch": 3.0, + "learning_rate": 2.5e-05, + "loss": 0.0, + "step": 12981 + }, + { + "epoch": 3.0, + "learning_rate": 2.499807410831215e-05, + "loss": 0.0, + "step": 12982 + }, + { + "epoch": 3.0, + "learning_rate": 2.49961482166243e-05, + "loss": 0.0, + "step": 12983 + }, + { + "epoch": 3.0, + "learning_rate": 2.4994222324936447e-05, + "loss": 0.0, + "step": 12984 + }, + { + "epoch": 3.0, + "learning_rate": 2.4992296433248598e-05, + "loss": 0.0, + "step": 12985 + }, + { + "epoch": 3.0, + "learning_rate": 2.4990370541560742e-05, + "loss": 0.0, + "step": 12986 + }, + { + "epoch": 3.0, + "learning_rate": 2.4988444649872893e-05, + "loss": 0.0, + "step": 12987 + }, + { + "epoch": 3.0, + "learning_rate": 2.498651875818504e-05, + "loss": 0.0, + "step": 12988 + }, + { + "epoch": 3.0, + "learning_rate": 2.4984592866497188e-05, + "loss": 0.0, + "step": 12989 + }, + { + "epoch": 3.0, + "learning_rate": 2.498266697480934e-05, + "loss": 0.0, + "step": 12990 + }, + { + "epoch": 3.0, + "learning_rate": 2.4980741083121486e-05, + "loss": 0.0, + "step": 12991 + }, + { + "epoch": 3.0, + "learning_rate": 2.4978815191433637e-05, + "loss": 0.0, + "step": 12992 + }, + { + "epoch": 3.0, + "learning_rate": 2.4976889299745784e-05, + "loss": 0.0, + "step": 12993 + }, + { + "epoch": 3.0, + "learning_rate": 2.4974963408057932e-05, + "loss": 0.0, + "step": 12994 + }, + { + "epoch": 3.0, + "learning_rate": 2.497303751637008e-05, + "loss": 0.0, + "step": 12995 + }, + { + "epoch": 3.0, + "learning_rate": 2.4971111624682227e-05, + "loss": 0.0, + "step": 12996 + }, + { + "epoch": 3.0, + "learning_rate": 2.4969185732994378e-05, + "loss": 0.0, + "step": 12997 + }, + { + "epoch": 3.0, + "learning_rate": 2.4967259841306525e-05, + "loss": 0.0, + "step": 12998 + }, + { + "epoch": 3.0, + "learning_rate": 2.4965333949618676e-05, + "loss": 0.0, + "step": 12999 + }, + { + "epoch": 3.0, + "learning_rate": 2.4963408057930824e-05, + "loss": 0.0, + "step": 13000 + }, + { + "epoch": 3.0, + "learning_rate": 2.496148216624297e-05, + "loss": 0.0, + "step": 13001 + }, + { + "epoch": 3.0, + "learning_rate": 2.4959556274555122e-05, + "loss": 0.0001, + "step": 13002 + }, + { + "epoch": 3.01, + "learning_rate": 2.495763038286727e-05, + "loss": 0.0, + "step": 13003 + }, + { + "epoch": 3.01, + "learning_rate": 2.4955704491179417e-05, + "loss": 0.0006, + "step": 13004 + }, + { + "epoch": 3.01, + "learning_rate": 2.4953778599491564e-05, + "loss": 0.0, + "step": 13005 + }, + { + "epoch": 3.01, + "learning_rate": 2.4951852707803712e-05, + "loss": 0.0, + "step": 13006 + }, + { + "epoch": 3.01, + "learning_rate": 2.4949926816115863e-05, + "loss": 0.0, + "step": 13007 + }, + { + "epoch": 3.01, + "learning_rate": 2.494800092442801e-05, + "loss": 0.0, + "step": 13008 + }, + { + "epoch": 3.01, + "learning_rate": 2.494607503274016e-05, + "loss": 0.0, + "step": 13009 + }, + { + "epoch": 3.01, + "learning_rate": 2.494414914105231e-05, + "loss": 0.0, + "step": 13010 + }, + { + "epoch": 3.01, + "learning_rate": 2.4942223249364456e-05, + "loss": 0.0, + "step": 13011 + }, + { + "epoch": 3.01, + "learning_rate": 2.4940297357676607e-05, + "loss": 0.0, + "step": 13012 + }, + { + "epoch": 3.01, + "learning_rate": 2.4938371465988754e-05, + "loss": 0.0, + "step": 13013 + }, + { + "epoch": 3.01, + "learning_rate": 2.4936445574300902e-05, + "loss": 0.0, + "step": 13014 + }, + { + "epoch": 3.01, + "learning_rate": 2.493451968261305e-05, + "loss": 0.0, + "step": 13015 + }, + { + "epoch": 3.01, + "learning_rate": 2.49325937909252e-05, + "loss": 0.0, + "step": 13016 + }, + { + "epoch": 3.01, + "learning_rate": 2.4930667899237348e-05, + "loss": 0.0, + "step": 13017 + }, + { + "epoch": 3.01, + "learning_rate": 2.4928742007549495e-05, + "loss": 0.0003, + "step": 13018 + }, + { + "epoch": 3.01, + "learning_rate": 2.4926816115861646e-05, + "loss": 0.0, + "step": 13019 + }, + { + "epoch": 3.01, + "learning_rate": 2.4924890224173793e-05, + "loss": 0.0, + "step": 13020 + }, + { + "epoch": 3.01, + "learning_rate": 2.4922964332485944e-05, + "loss": 0.0001, + "step": 13021 + }, + { + "epoch": 3.01, + "learning_rate": 2.4921038440798092e-05, + "loss": 0.0, + "step": 13022 + }, + { + "epoch": 3.01, + "learning_rate": 2.491911254911024e-05, + "loss": 0.0, + "step": 13023 + }, + { + "epoch": 3.01, + "learning_rate": 2.4917186657422387e-05, + "loss": 0.0, + "step": 13024 + }, + { + "epoch": 3.01, + "learning_rate": 2.4915260765734534e-05, + "loss": 0.0001, + "step": 13025 + }, + { + "epoch": 3.01, + "learning_rate": 2.4913334874046685e-05, + "loss": 0.0, + "step": 13026 + }, + { + "epoch": 3.01, + "learning_rate": 2.4911408982358833e-05, + "loss": 0.0, + "step": 13027 + }, + { + "epoch": 3.01, + "learning_rate": 2.4909483090670983e-05, + "loss": 0.0, + "step": 13028 + }, + { + "epoch": 3.01, + "learning_rate": 2.490755719898313e-05, + "loss": 0.0, + "step": 13029 + }, + { + "epoch": 3.01, + "learning_rate": 2.490563130729528e-05, + "loss": 0.0, + "step": 13030 + }, + { + "epoch": 3.01, + "learning_rate": 2.490370541560743e-05, + "loss": 0.0, + "step": 13031 + }, + { + "epoch": 3.01, + "learning_rate": 2.4901779523919577e-05, + "loss": 0.0, + "step": 13032 + }, + { + "epoch": 3.01, + "learning_rate": 2.4899853632231724e-05, + "loss": 0.0, + "step": 13033 + }, + { + "epoch": 3.01, + "learning_rate": 2.4897927740543872e-05, + "loss": 0.0, + "step": 13034 + }, + { + "epoch": 3.01, + "learning_rate": 2.4896001848856023e-05, + "loss": 0.0, + "step": 13035 + }, + { + "epoch": 3.01, + "learning_rate": 2.489407595716817e-05, + "loss": 0.0, + "step": 13036 + }, + { + "epoch": 3.01, + "learning_rate": 2.4892150065480318e-05, + "loss": 0.0, + "step": 13037 + }, + { + "epoch": 3.01, + "learning_rate": 2.489022417379247e-05, + "loss": 0.0, + "step": 13038 + }, + { + "epoch": 3.01, + "learning_rate": 2.4888298282104616e-05, + "loss": 0.0, + "step": 13039 + }, + { + "epoch": 3.01, + "learning_rate": 2.4886372390416763e-05, + "loss": 0.0001, + "step": 13040 + }, + { + "epoch": 3.01, + "learning_rate": 2.4884446498728914e-05, + "loss": 0.0, + "step": 13041 + }, + { + "epoch": 3.01, + "learning_rate": 2.488252060704106e-05, + "loss": 0.0001, + "step": 13042 + }, + { + "epoch": 3.01, + "learning_rate": 2.488059471535321e-05, + "loss": 0.0, + "step": 13043 + }, + { + "epoch": 3.01, + "learning_rate": 2.4878668823665357e-05, + "loss": 0.0, + "step": 13044 + }, + { + "epoch": 3.01, + "learning_rate": 2.4876742931977508e-05, + "loss": 0.0, + "step": 13045 + }, + { + "epoch": 3.02, + "learning_rate": 2.4874817040289655e-05, + "loss": 0.0, + "step": 13046 + }, + { + "epoch": 3.02, + "learning_rate": 2.4872891148601802e-05, + "loss": 0.0, + "step": 13047 + }, + { + "epoch": 3.02, + "learning_rate": 2.4870965256913953e-05, + "loss": 0.0, + "step": 13048 + }, + { + "epoch": 3.02, + "learning_rate": 2.48690393652261e-05, + "loss": 0.0, + "step": 13049 + }, + { + "epoch": 3.02, + "learning_rate": 2.4867113473538252e-05, + "loss": 0.0006, + "step": 13050 + }, + { + "epoch": 3.02, + "learning_rate": 2.4865187581850396e-05, + "loss": 0.0, + "step": 13051 + }, + { + "epoch": 3.02, + "learning_rate": 2.4863261690162547e-05, + "loss": 0.0, + "step": 13052 + }, + { + "epoch": 3.02, + "learning_rate": 2.4861335798474694e-05, + "loss": 0.0, + "step": 13053 + }, + { + "epoch": 3.02, + "learning_rate": 2.485940990678684e-05, + "loss": 0.0, + "step": 13054 + }, + { + "epoch": 3.02, + "learning_rate": 2.4857484015098992e-05, + "loss": 0.0001, + "step": 13055 + }, + { + "epoch": 3.02, + "learning_rate": 2.485555812341114e-05, + "loss": 0.0, + "step": 13056 + }, + { + "epoch": 3.02, + "learning_rate": 2.485363223172329e-05, + "loss": 0.0, + "step": 13057 + }, + { + "epoch": 3.02, + "learning_rate": 2.4851706340035438e-05, + "loss": 0.0003, + "step": 13058 + }, + { + "epoch": 3.02, + "learning_rate": 2.4849780448347586e-05, + "loss": 0.0, + "step": 13059 + }, + { + "epoch": 3.02, + "learning_rate": 2.4847854556659737e-05, + "loss": 0.0, + "step": 13060 + }, + { + "epoch": 3.02, + "learning_rate": 2.484592866497188e-05, + "loss": 0.0, + "step": 13061 + }, + { + "epoch": 3.02, + "learning_rate": 2.484400277328403e-05, + "loss": 0.0, + "step": 13062 + }, + { + "epoch": 3.02, + "learning_rate": 2.484207688159618e-05, + "loss": 0.0, + "step": 13063 + }, + { + "epoch": 3.02, + "learning_rate": 2.484015098990833e-05, + "loss": 0.0, + "step": 13064 + }, + { + "epoch": 3.02, + "learning_rate": 2.4838225098220477e-05, + "loss": 0.0, + "step": 13065 + }, + { + "epoch": 3.02, + "learning_rate": 2.4836299206532625e-05, + "loss": 0.0, + "step": 13066 + }, + { + "epoch": 3.02, + "learning_rate": 2.4834373314844776e-05, + "loss": 0.0, + "step": 13067 + }, + { + "epoch": 3.02, + "learning_rate": 2.4832447423156923e-05, + "loss": 0.0, + "step": 13068 + }, + { + "epoch": 3.02, + "learning_rate": 2.4830521531469074e-05, + "loss": 0.0, + "step": 13069 + }, + { + "epoch": 3.02, + "learning_rate": 2.4828595639781218e-05, + "loss": 0.0, + "step": 13070 + }, + { + "epoch": 3.02, + "learning_rate": 2.482666974809337e-05, + "loss": 0.0, + "step": 13071 + }, + { + "epoch": 3.02, + "learning_rate": 2.4824743856405517e-05, + "loss": 0.0001, + "step": 13072 + }, + { + "epoch": 3.02, + "learning_rate": 2.4822817964717664e-05, + "loss": 0.0, + "step": 13073 + }, + { + "epoch": 3.02, + "learning_rate": 2.4820892073029815e-05, + "loss": 0.0, + "step": 13074 + }, + { + "epoch": 3.02, + "learning_rate": 2.4818966181341962e-05, + "loss": 0.0, + "step": 13075 + }, + { + "epoch": 3.02, + "learning_rate": 2.4817040289654113e-05, + "loss": 0.0, + "step": 13076 + }, + { + "epoch": 3.02, + "learning_rate": 2.481511439796626e-05, + "loss": 0.0, + "step": 13077 + }, + { + "epoch": 3.02, + "learning_rate": 2.4813188506278408e-05, + "loss": 0.0, + "step": 13078 + }, + { + "epoch": 3.02, + "learning_rate": 2.4811262614590556e-05, + "loss": 0.0001, + "step": 13079 + }, + { + "epoch": 3.02, + "learning_rate": 2.4809336722902703e-05, + "loss": 0.0, + "step": 13080 + }, + { + "epoch": 3.02, + "learning_rate": 2.4807410831214854e-05, + "loss": 0.0, + "step": 13081 + }, + { + "epoch": 3.02, + "learning_rate": 2.4805484939527e-05, + "loss": 0.0001, + "step": 13082 + }, + { + "epoch": 3.02, + "learning_rate": 2.480355904783915e-05, + "loss": 0.0, + "step": 13083 + }, + { + "epoch": 3.02, + "learning_rate": 2.48016331561513e-05, + "loss": 0.0, + "step": 13084 + }, + { + "epoch": 3.02, + "learning_rate": 2.4799707264463447e-05, + "loss": 0.0, + "step": 13085 + }, + { + "epoch": 3.02, + "learning_rate": 2.4797781372775598e-05, + "loss": 0.0002, + "step": 13086 + }, + { + "epoch": 3.02, + "learning_rate": 2.4795855481087746e-05, + "loss": 0.0, + "step": 13087 + }, + { + "epoch": 3.02, + "learning_rate": 2.4793929589399893e-05, + "loss": 0.0, + "step": 13088 + }, + { + "epoch": 3.02, + "learning_rate": 2.479200369771204e-05, + "loss": 0.0, + "step": 13089 + }, + { + "epoch": 3.03, + "learning_rate": 2.4790077806024188e-05, + "loss": 0.0, + "step": 13090 + }, + { + "epoch": 3.03, + "learning_rate": 2.478815191433634e-05, + "loss": 0.0, + "step": 13091 + }, + { + "epoch": 3.03, + "learning_rate": 2.4786226022648486e-05, + "loss": 0.0001, + "step": 13092 + }, + { + "epoch": 3.03, + "learning_rate": 2.4784300130960637e-05, + "loss": 0.0, + "step": 13093 + }, + { + "epoch": 3.03, + "learning_rate": 2.4782374239272785e-05, + "loss": 0.0, + "step": 13094 + }, + { + "epoch": 3.03, + "learning_rate": 2.4780448347584932e-05, + "loss": 0.0003, + "step": 13095 + }, + { + "epoch": 3.03, + "learning_rate": 2.4778522455897083e-05, + "loss": 0.0, + "step": 13096 + }, + { + "epoch": 3.03, + "learning_rate": 2.477659656420923e-05, + "loss": 0.0, + "step": 13097 + }, + { + "epoch": 3.03, + "learning_rate": 2.4774670672521378e-05, + "loss": 0.0, + "step": 13098 + }, + { + "epoch": 3.03, + "learning_rate": 2.4772744780833526e-05, + "loss": 0.0, + "step": 13099 + }, + { + "epoch": 3.03, + "learning_rate": 2.4770818889145676e-05, + "loss": 0.0, + "step": 13100 + }, + { + "epoch": 3.03, + "learning_rate": 2.4768892997457824e-05, + "loss": 0.0, + "step": 13101 + }, + { + "epoch": 3.03, + "learning_rate": 2.476696710576997e-05, + "loss": 0.0, + "step": 13102 + }, + { + "epoch": 3.03, + "learning_rate": 2.4765041214082122e-05, + "loss": 0.0001, + "step": 13103 + }, + { + "epoch": 3.03, + "learning_rate": 2.476311532239427e-05, + "loss": 0.0, + "step": 13104 + }, + { + "epoch": 3.03, + "learning_rate": 2.476118943070642e-05, + "loss": 0.0, + "step": 13105 + }, + { + "epoch": 3.03, + "learning_rate": 2.4759263539018568e-05, + "loss": 0.0001, + "step": 13106 + }, + { + "epoch": 3.03, + "learning_rate": 2.4757337647330716e-05, + "loss": 0.0, + "step": 13107 + }, + { + "epoch": 3.03, + "learning_rate": 2.4755411755642863e-05, + "loss": 0.0001, + "step": 13108 + }, + { + "epoch": 3.03, + "learning_rate": 2.475348586395501e-05, + "loss": 0.0, + "step": 13109 + }, + { + "epoch": 3.03, + "learning_rate": 2.475155997226716e-05, + "loss": 0.0, + "step": 13110 + }, + { + "epoch": 3.03, + "learning_rate": 2.474963408057931e-05, + "loss": 0.0, + "step": 13111 + }, + { + "epoch": 3.03, + "learning_rate": 2.474770818889146e-05, + "loss": 0.0, + "step": 13112 + }, + { + "epoch": 3.03, + "learning_rate": 2.4745782297203607e-05, + "loss": 0.0, + "step": 13113 + }, + { + "epoch": 3.03, + "learning_rate": 2.4743856405515755e-05, + "loss": 0.0, + "step": 13114 + }, + { + "epoch": 3.03, + "learning_rate": 2.4741930513827906e-05, + "loss": 0.0005, + "step": 13115 + }, + { + "epoch": 3.03, + "learning_rate": 2.4740004622140053e-05, + "loss": 0.0, + "step": 13116 + }, + { + "epoch": 3.03, + "learning_rate": 2.47380787304522e-05, + "loss": 0.0, + "step": 13117 + }, + { + "epoch": 3.03, + "learning_rate": 2.4736152838764348e-05, + "loss": 0.0, + "step": 13118 + }, + { + "epoch": 3.03, + "learning_rate": 2.4734226947076495e-05, + "loss": 0.0, + "step": 13119 + }, + { + "epoch": 3.03, + "learning_rate": 2.4732301055388646e-05, + "loss": 0.0, + "step": 13120 + }, + { + "epoch": 3.03, + "learning_rate": 2.4730375163700794e-05, + "loss": 0.0, + "step": 13121 + }, + { + "epoch": 3.03, + "learning_rate": 2.4728449272012945e-05, + "loss": 0.0, + "step": 13122 + }, + { + "epoch": 3.03, + "learning_rate": 2.4726523380325092e-05, + "loss": 0.0, + "step": 13123 + }, + { + "epoch": 3.03, + "learning_rate": 2.472459748863724e-05, + "loss": 0.0, + "step": 13124 + }, + { + "epoch": 3.03, + "learning_rate": 2.472267159694939e-05, + "loss": 0.0, + "step": 13125 + }, + { + "epoch": 3.03, + "learning_rate": 2.4720745705261535e-05, + "loss": 0.0, + "step": 13126 + }, + { + "epoch": 3.03, + "learning_rate": 2.4718819813573685e-05, + "loss": 0.0004, + "step": 13127 + }, + { + "epoch": 3.03, + "learning_rate": 2.4716893921885833e-05, + "loss": 0.0, + "step": 13128 + }, + { + "epoch": 3.03, + "learning_rate": 2.4714968030197984e-05, + "loss": 0.0003, + "step": 13129 + }, + { + "epoch": 3.03, + "learning_rate": 2.471304213851013e-05, + "loss": 0.0002, + "step": 13130 + }, + { + "epoch": 3.03, + "learning_rate": 2.471111624682228e-05, + "loss": 0.0001, + "step": 13131 + }, + { + "epoch": 3.03, + "learning_rate": 2.470919035513443e-05, + "loss": 0.0, + "step": 13132 + }, + { + "epoch": 3.04, + "learning_rate": 2.4707264463446577e-05, + "loss": 0.0, + "step": 13133 + }, + { + "epoch": 3.04, + "learning_rate": 2.4705338571758728e-05, + "loss": 0.0, + "step": 13134 + }, + { + "epoch": 3.04, + "learning_rate": 2.4703412680070872e-05, + "loss": 0.0, + "step": 13135 + }, + { + "epoch": 3.04, + "learning_rate": 2.4701486788383023e-05, + "loss": 0.0, + "step": 13136 + }, + { + "epoch": 3.04, + "learning_rate": 2.469956089669517e-05, + "loss": 0.0, + "step": 13137 + }, + { + "epoch": 3.04, + "learning_rate": 2.4697635005007318e-05, + "loss": 0.0, + "step": 13138 + }, + { + "epoch": 3.04, + "learning_rate": 2.469570911331947e-05, + "loss": 0.0, + "step": 13139 + }, + { + "epoch": 3.04, + "learning_rate": 2.4693783221631616e-05, + "loss": 0.0, + "step": 13140 + }, + { + "epoch": 3.04, + "learning_rate": 2.4691857329943767e-05, + "loss": 0.0, + "step": 13141 + }, + { + "epoch": 3.04, + "learning_rate": 2.4689931438255915e-05, + "loss": 0.0, + "step": 13142 + }, + { + "epoch": 3.04, + "learning_rate": 2.4688005546568062e-05, + "loss": 0.0, + "step": 13143 + }, + { + "epoch": 3.04, + "learning_rate": 2.4686079654880213e-05, + "loss": 0.0, + "step": 13144 + }, + { + "epoch": 3.04, + "learning_rate": 2.4684153763192357e-05, + "loss": 0.0001, + "step": 13145 + }, + { + "epoch": 3.04, + "learning_rate": 2.4682227871504508e-05, + "loss": 0.0, + "step": 13146 + }, + { + "epoch": 3.04, + "learning_rate": 2.4680301979816655e-05, + "loss": 0.0, + "step": 13147 + }, + { + "epoch": 3.04, + "learning_rate": 2.4678376088128806e-05, + "loss": 0.0, + "step": 13148 + }, + { + "epoch": 3.04, + "learning_rate": 2.4676450196440954e-05, + "loss": 0.0001, + "step": 13149 + }, + { + "epoch": 3.04, + "learning_rate": 2.46745243047531e-05, + "loss": 0.0, + "step": 13150 + }, + { + "epoch": 3.04, + "learning_rate": 2.4672598413065252e-05, + "loss": 0.0, + "step": 13151 + }, + { + "epoch": 3.04, + "learning_rate": 2.46706725213774e-05, + "loss": 0.0, + "step": 13152 + }, + { + "epoch": 3.04, + "learning_rate": 2.4668746629689547e-05, + "loss": 0.0, + "step": 13153 + }, + { + "epoch": 3.04, + "learning_rate": 2.4666820738001694e-05, + "loss": 0.0, + "step": 13154 + }, + { + "epoch": 3.04, + "learning_rate": 2.4664894846313842e-05, + "loss": 0.0, + "step": 13155 + }, + { + "epoch": 3.04, + "learning_rate": 2.4662968954625993e-05, + "loss": 0.0, + "step": 13156 + }, + { + "epoch": 3.04, + "learning_rate": 2.466104306293814e-05, + "loss": 0.0, + "step": 13157 + }, + { + "epoch": 3.04, + "learning_rate": 2.465911717125029e-05, + "loss": 0.0, + "step": 13158 + }, + { + "epoch": 3.04, + "learning_rate": 2.465719127956244e-05, + "loss": 0.0001, + "step": 13159 + }, + { + "epoch": 3.04, + "learning_rate": 2.4655265387874586e-05, + "loss": 0.0, + "step": 13160 + }, + { + "epoch": 3.04, + "learning_rate": 2.4653339496186737e-05, + "loss": 0.0001, + "step": 13161 + }, + { + "epoch": 3.04, + "learning_rate": 2.4651413604498884e-05, + "loss": 0.0, + "step": 13162 + }, + { + "epoch": 3.04, + "learning_rate": 2.4649487712811032e-05, + "loss": 0.0, + "step": 13163 + }, + { + "epoch": 3.04, + "learning_rate": 2.464756182112318e-05, + "loss": 0.0, + "step": 13164 + }, + { + "epoch": 3.04, + "learning_rate": 2.464563592943533e-05, + "loss": 0.0, + "step": 13165 + }, + { + "epoch": 3.04, + "learning_rate": 2.4643710037747478e-05, + "loss": 0.0001, + "step": 13166 + }, + { + "epoch": 3.04, + "learning_rate": 2.4641784146059625e-05, + "loss": 0.0, + "step": 13167 + }, + { + "epoch": 3.04, + "learning_rate": 2.4639858254371776e-05, + "loss": 0.0, + "step": 13168 + }, + { + "epoch": 3.04, + "learning_rate": 2.4637932362683924e-05, + "loss": 0.0, + "step": 13169 + }, + { + "epoch": 3.04, + "learning_rate": 2.4636006470996074e-05, + "loss": 0.0, + "step": 13170 + }, + { + "epoch": 3.04, + "learning_rate": 2.4634080579308222e-05, + "loss": 0.0, + "step": 13171 + }, + { + "epoch": 3.04, + "learning_rate": 2.463215468762037e-05, + "loss": 0.0, + "step": 13172 + }, + { + "epoch": 3.04, + "learning_rate": 2.4630228795932517e-05, + "loss": 0.0, + "step": 13173 + }, + { + "epoch": 3.04, + "learning_rate": 2.4628302904244664e-05, + "loss": 0.0, + "step": 13174 + }, + { + "epoch": 3.04, + "learning_rate": 2.4626377012556815e-05, + "loss": 0.0, + "step": 13175 + }, + { + "epoch": 3.05, + "learning_rate": 2.4624451120868963e-05, + "loss": 0.0, + "step": 13176 + }, + { + "epoch": 3.05, + "learning_rate": 2.4622525229181114e-05, + "loss": 0.0, + "step": 13177 + }, + { + "epoch": 3.05, + "learning_rate": 2.462059933749326e-05, + "loss": 0.0, + "step": 13178 + }, + { + "epoch": 3.05, + "learning_rate": 2.461867344580541e-05, + "loss": 0.0, + "step": 13179 + }, + { + "epoch": 3.05, + "learning_rate": 2.461674755411756e-05, + "loss": 0.0001, + "step": 13180 + }, + { + "epoch": 3.05, + "learning_rate": 2.4614821662429707e-05, + "loss": 0.0, + "step": 13181 + }, + { + "epoch": 3.05, + "learning_rate": 2.4612895770741854e-05, + "loss": 0.0, + "step": 13182 + }, + { + "epoch": 3.05, + "learning_rate": 2.4610969879054002e-05, + "loss": 0.0, + "step": 13183 + }, + { + "epoch": 3.05, + "learning_rate": 2.4609043987366153e-05, + "loss": 0.0, + "step": 13184 + }, + { + "epoch": 3.05, + "learning_rate": 2.46071180956783e-05, + "loss": 0.0, + "step": 13185 + }, + { + "epoch": 3.05, + "learning_rate": 2.4605192203990448e-05, + "loss": 0.0, + "step": 13186 + }, + { + "epoch": 3.05, + "learning_rate": 2.46032663123026e-05, + "loss": 0.0, + "step": 13187 + }, + { + "epoch": 3.05, + "learning_rate": 2.4601340420614746e-05, + "loss": 0.0, + "step": 13188 + }, + { + "epoch": 3.05, + "learning_rate": 2.4599414528926897e-05, + "loss": 0.0, + "step": 13189 + }, + { + "epoch": 3.05, + "learning_rate": 2.4597488637239044e-05, + "loss": 0.0, + "step": 13190 + }, + { + "epoch": 3.05, + "learning_rate": 2.459556274555119e-05, + "loss": 0.0, + "step": 13191 + }, + { + "epoch": 3.05, + "learning_rate": 2.459363685386334e-05, + "loss": 0.0, + "step": 13192 + }, + { + "epoch": 3.05, + "learning_rate": 2.4591710962175487e-05, + "loss": 0.0, + "step": 13193 + }, + { + "epoch": 3.05, + "learning_rate": 2.4589785070487638e-05, + "loss": 0.0, + "step": 13194 + }, + { + "epoch": 3.05, + "learning_rate": 2.4587859178799785e-05, + "loss": 0.0, + "step": 13195 + }, + { + "epoch": 3.05, + "learning_rate": 2.4585933287111933e-05, + "loss": 0.0, + "step": 13196 + }, + { + "epoch": 3.05, + "learning_rate": 2.4584007395424083e-05, + "loss": 0.0, + "step": 13197 + }, + { + "epoch": 3.05, + "learning_rate": 2.458208150373623e-05, + "loss": 0.0, + "step": 13198 + }, + { + "epoch": 3.05, + "learning_rate": 2.4580155612048382e-05, + "loss": 0.0, + "step": 13199 + }, + { + "epoch": 3.05, + "learning_rate": 2.457822972036053e-05, + "loss": 0.0001, + "step": 13200 + }, + { + "epoch": 3.05, + "learning_rate": 2.4576303828672677e-05, + "loss": 0.0, + "step": 13201 + }, + { + "epoch": 3.05, + "learning_rate": 2.4574377936984824e-05, + "loss": 0.0, + "step": 13202 + }, + { + "epoch": 3.05, + "learning_rate": 2.457245204529697e-05, + "loss": 0.0, + "step": 13203 + }, + { + "epoch": 3.05, + "learning_rate": 2.4570526153609123e-05, + "loss": 0.0, + "step": 13204 + }, + { + "epoch": 3.05, + "learning_rate": 2.456860026192127e-05, + "loss": 0.0, + "step": 13205 + }, + { + "epoch": 3.05, + "learning_rate": 2.456667437023342e-05, + "loss": 0.0, + "step": 13206 + }, + { + "epoch": 3.05, + "learning_rate": 2.456474847854557e-05, + "loss": 0.0, + "step": 13207 + }, + { + "epoch": 3.05, + "learning_rate": 2.4562822586857716e-05, + "loss": 0.0, + "step": 13208 + }, + { + "epoch": 3.05, + "learning_rate": 2.4560896695169867e-05, + "loss": 0.0001, + "step": 13209 + }, + { + "epoch": 3.05, + "learning_rate": 2.455897080348201e-05, + "loss": 0.0, + "step": 13210 + }, + { + "epoch": 3.05, + "learning_rate": 2.455704491179416e-05, + "loss": 0.0, + "step": 13211 + }, + { + "epoch": 3.05, + "learning_rate": 2.455511902010631e-05, + "loss": 0.0, + "step": 13212 + }, + { + "epoch": 3.05, + "learning_rate": 2.455319312841846e-05, + "loss": 0.0002, + "step": 13213 + }, + { + "epoch": 3.05, + "learning_rate": 2.4551267236730607e-05, + "loss": 0.0, + "step": 13214 + }, + { + "epoch": 3.05, + "learning_rate": 2.4549341345042755e-05, + "loss": 0.0, + "step": 13215 + }, + { + "epoch": 3.05, + "learning_rate": 2.4547415453354906e-05, + "loss": 0.0, + "step": 13216 + }, + { + "epoch": 3.05, + "learning_rate": 2.4545489561667053e-05, + "loss": 0.0, + "step": 13217 + }, + { + "epoch": 3.05, + "learning_rate": 2.4543563669979204e-05, + "loss": 0.0001, + "step": 13218 + }, + { + "epoch": 3.06, + "learning_rate": 2.4541637778291348e-05, + "loss": 0.0001, + "step": 13219 + }, + { + "epoch": 3.06, + "learning_rate": 2.45397118866035e-05, + "loss": 0.0001, + "step": 13220 + }, + { + "epoch": 3.06, + "learning_rate": 2.4537785994915647e-05, + "loss": 0.0, + "step": 13221 + }, + { + "epoch": 3.06, + "learning_rate": 2.4535860103227794e-05, + "loss": 0.0, + "step": 13222 + }, + { + "epoch": 3.06, + "learning_rate": 2.4533934211539945e-05, + "loss": 0.0, + "step": 13223 + }, + { + "epoch": 3.06, + "learning_rate": 2.4532008319852092e-05, + "loss": 0.0, + "step": 13224 + }, + { + "epoch": 3.06, + "learning_rate": 2.4530082428164243e-05, + "loss": 0.0, + "step": 13225 + }, + { + "epoch": 3.06, + "learning_rate": 2.452815653647639e-05, + "loss": 0.0, + "step": 13226 + }, + { + "epoch": 3.06, + "learning_rate": 2.4526230644788538e-05, + "loss": 0.0, + "step": 13227 + }, + { + "epoch": 3.06, + "learning_rate": 2.452430475310069e-05, + "loss": 0.0, + "step": 13228 + }, + { + "epoch": 3.06, + "learning_rate": 2.4522378861412833e-05, + "loss": 0.0, + "step": 13229 + }, + { + "epoch": 3.06, + "learning_rate": 2.4520452969724984e-05, + "loss": 0.0001, + "step": 13230 + }, + { + "epoch": 3.06, + "learning_rate": 2.451852707803713e-05, + "loss": 0.0, + "step": 13231 + }, + { + "epoch": 3.06, + "learning_rate": 2.451660118634928e-05, + "loss": 0.0, + "step": 13232 + }, + { + "epoch": 3.06, + "learning_rate": 2.451467529466143e-05, + "loss": 0.0, + "step": 13233 + }, + { + "epoch": 3.06, + "learning_rate": 2.4512749402973577e-05, + "loss": 0.0, + "step": 13234 + }, + { + "epoch": 3.06, + "learning_rate": 2.4510823511285728e-05, + "loss": 0.0, + "step": 13235 + }, + { + "epoch": 3.06, + "learning_rate": 2.4508897619597876e-05, + "loss": 0.0, + "step": 13236 + }, + { + "epoch": 3.06, + "learning_rate": 2.4506971727910023e-05, + "loss": 0.0, + "step": 13237 + }, + { + "epoch": 3.06, + "learning_rate": 2.450504583622217e-05, + "loss": 0.0, + "step": 13238 + }, + { + "epoch": 3.06, + "learning_rate": 2.4503119944534318e-05, + "loss": 0.0, + "step": 13239 + }, + { + "epoch": 3.06, + "learning_rate": 2.450119405284647e-05, + "loss": 0.0012, + "step": 13240 + }, + { + "epoch": 3.06, + "learning_rate": 2.4499268161158616e-05, + "loss": 0.0001, + "step": 13241 + }, + { + "epoch": 3.06, + "learning_rate": 2.4497342269470767e-05, + "loss": 0.0, + "step": 13242 + }, + { + "epoch": 3.06, + "learning_rate": 2.4495416377782915e-05, + "loss": 0.0, + "step": 13243 + }, + { + "epoch": 3.06, + "learning_rate": 2.4493490486095062e-05, + "loss": 0.0, + "step": 13244 + }, + { + "epoch": 3.06, + "learning_rate": 2.4491564594407213e-05, + "loss": 0.0001, + "step": 13245 + }, + { + "epoch": 3.06, + "learning_rate": 2.448963870271936e-05, + "loss": 0.0, + "step": 13246 + }, + { + "epoch": 3.06, + "learning_rate": 2.4487712811031508e-05, + "loss": 0.0, + "step": 13247 + }, + { + "epoch": 3.06, + "learning_rate": 2.4485786919343656e-05, + "loss": 0.0001, + "step": 13248 + }, + { + "epoch": 3.06, + "learning_rate": 2.4483861027655806e-05, + "loss": 0.0001, + "step": 13249 + }, + { + "epoch": 3.06, + "learning_rate": 2.4481935135967954e-05, + "loss": 0.0, + "step": 13250 + }, + { + "epoch": 3.06, + "learning_rate": 2.44800092442801e-05, + "loss": 0.0, + "step": 13251 + }, + { + "epoch": 3.06, + "learning_rate": 2.4478083352592252e-05, + "loss": 0.0001, + "step": 13252 + }, + { + "epoch": 3.06, + "learning_rate": 2.44761574609044e-05, + "loss": 0.0, + "step": 13253 + }, + { + "epoch": 3.06, + "learning_rate": 2.447423156921655e-05, + "loss": 0.0, + "step": 13254 + }, + { + "epoch": 3.06, + "learning_rate": 2.4472305677528698e-05, + "loss": 0.0, + "step": 13255 + }, + { + "epoch": 3.06, + "learning_rate": 2.4470379785840846e-05, + "loss": 0.0, + "step": 13256 + }, + { + "epoch": 3.06, + "learning_rate": 2.4468453894152993e-05, + "loss": 0.0, + "step": 13257 + }, + { + "epoch": 3.06, + "learning_rate": 2.446652800246514e-05, + "loss": 0.0, + "step": 13258 + }, + { + "epoch": 3.06, + "learning_rate": 2.446460211077729e-05, + "loss": 0.0, + "step": 13259 + }, + { + "epoch": 3.06, + "learning_rate": 2.446267621908944e-05, + "loss": 0.0, + "step": 13260 + }, + { + "epoch": 3.06, + "learning_rate": 2.446075032740159e-05, + "loss": 0.0, + "step": 13261 + }, + { + "epoch": 3.06, + "learning_rate": 2.4458824435713737e-05, + "loss": 0.0, + "step": 13262 + }, + { + "epoch": 3.07, + "learning_rate": 2.4456898544025885e-05, + "loss": 0.0002, + "step": 13263 + }, + { + "epoch": 3.07, + "learning_rate": 2.4454972652338036e-05, + "loss": 0.0, + "step": 13264 + }, + { + "epoch": 3.07, + "learning_rate": 2.4453046760650183e-05, + "loss": 0.0, + "step": 13265 + }, + { + "epoch": 3.07, + "learning_rate": 2.445112086896233e-05, + "loss": 0.0, + "step": 13266 + }, + { + "epoch": 3.07, + "learning_rate": 2.4449194977274478e-05, + "loss": 0.0, + "step": 13267 + }, + { + "epoch": 3.07, + "learning_rate": 2.4447269085586625e-05, + "loss": 0.0, + "step": 13268 + }, + { + "epoch": 3.07, + "learning_rate": 2.4445343193898776e-05, + "loss": 0.0, + "step": 13269 + }, + { + "epoch": 3.07, + "learning_rate": 2.4443417302210924e-05, + "loss": 0.0, + "step": 13270 + }, + { + "epoch": 3.07, + "learning_rate": 2.4441491410523075e-05, + "loss": 0.0, + "step": 13271 + }, + { + "epoch": 3.07, + "learning_rate": 2.4439565518835222e-05, + "loss": 0.0, + "step": 13272 + }, + { + "epoch": 3.07, + "learning_rate": 2.443763962714737e-05, + "loss": 0.0, + "step": 13273 + }, + { + "epoch": 3.07, + "learning_rate": 2.443571373545952e-05, + "loss": 0.0, + "step": 13274 + }, + { + "epoch": 3.07, + "learning_rate": 2.4433787843771665e-05, + "loss": 0.0, + "step": 13275 + }, + { + "epoch": 3.07, + "learning_rate": 2.4431861952083815e-05, + "loss": 0.0, + "step": 13276 + }, + { + "epoch": 3.07, + "learning_rate": 2.4429936060395963e-05, + "loss": 0.0, + "step": 13277 + }, + { + "epoch": 3.07, + "learning_rate": 2.4428010168708114e-05, + "loss": 0.0001, + "step": 13278 + }, + { + "epoch": 3.07, + "learning_rate": 2.442608427702026e-05, + "loss": 0.0, + "step": 13279 + }, + { + "epoch": 3.07, + "learning_rate": 2.442415838533241e-05, + "loss": 0.0, + "step": 13280 + }, + { + "epoch": 3.07, + "learning_rate": 2.442223249364456e-05, + "loss": 0.0, + "step": 13281 + }, + { + "epoch": 3.07, + "learning_rate": 2.4420306601956707e-05, + "loss": 0.0001, + "step": 13282 + }, + { + "epoch": 3.07, + "learning_rate": 2.4418380710268858e-05, + "loss": 0.0, + "step": 13283 + }, + { + "epoch": 3.07, + "learning_rate": 2.4416454818581005e-05, + "loss": 0.0001, + "step": 13284 + }, + { + "epoch": 3.07, + "learning_rate": 2.4414528926893153e-05, + "loss": 0.0, + "step": 13285 + }, + { + "epoch": 3.07, + "learning_rate": 2.44126030352053e-05, + "loss": 0.0001, + "step": 13286 + }, + { + "epoch": 3.07, + "learning_rate": 2.4410677143517448e-05, + "loss": 0.0, + "step": 13287 + }, + { + "epoch": 3.07, + "learning_rate": 2.44087512518296e-05, + "loss": 0.0, + "step": 13288 + }, + { + "epoch": 3.07, + "learning_rate": 2.4406825360141746e-05, + "loss": 0.0, + "step": 13289 + }, + { + "epoch": 3.07, + "learning_rate": 2.4404899468453897e-05, + "loss": 0.0, + "step": 13290 + }, + { + "epoch": 3.07, + "learning_rate": 2.4402973576766045e-05, + "loss": 0.0, + "step": 13291 + }, + { + "epoch": 3.07, + "learning_rate": 2.4401047685078192e-05, + "loss": 0.0, + "step": 13292 + }, + { + "epoch": 3.07, + "learning_rate": 2.4399121793390343e-05, + "loss": 0.0, + "step": 13293 + }, + { + "epoch": 3.07, + "learning_rate": 2.4397195901702487e-05, + "loss": 0.0, + "step": 13294 + }, + { + "epoch": 3.07, + "learning_rate": 2.4395270010014638e-05, + "loss": 0.0, + "step": 13295 + }, + { + "epoch": 3.07, + "learning_rate": 2.4393344118326785e-05, + "loss": 0.0001, + "step": 13296 + }, + { + "epoch": 3.07, + "learning_rate": 2.4391418226638936e-05, + "loss": 0.0, + "step": 13297 + }, + { + "epoch": 3.07, + "learning_rate": 2.4389492334951084e-05, + "loss": 0.0, + "step": 13298 + }, + { + "epoch": 3.07, + "learning_rate": 2.438756644326323e-05, + "loss": 0.0, + "step": 13299 + }, + { + "epoch": 3.07, + "learning_rate": 2.4385640551575382e-05, + "loss": 0.0, + "step": 13300 + }, + { + "epoch": 3.07, + "learning_rate": 2.438371465988753e-05, + "loss": 0.0, + "step": 13301 + }, + { + "epoch": 3.07, + "learning_rate": 2.438178876819968e-05, + "loss": 0.0, + "step": 13302 + }, + { + "epoch": 3.07, + "learning_rate": 2.4379862876511824e-05, + "loss": 0.0, + "step": 13303 + }, + { + "epoch": 3.07, + "learning_rate": 2.4377936984823972e-05, + "loss": 0.0002, + "step": 13304 + }, + { + "epoch": 3.07, + "learning_rate": 2.4376011093136123e-05, + "loss": 0.0, + "step": 13305 + }, + { + "epoch": 3.08, + "learning_rate": 2.437408520144827e-05, + "loss": 0.0, + "step": 13306 + }, + { + "epoch": 3.08, + "learning_rate": 2.437215930976042e-05, + "loss": 0.0, + "step": 13307 + }, + { + "epoch": 3.08, + "learning_rate": 2.437023341807257e-05, + "loss": 0.0, + "step": 13308 + }, + { + "epoch": 3.08, + "learning_rate": 2.4368307526384716e-05, + "loss": 0.0, + "step": 13309 + }, + { + "epoch": 3.08, + "learning_rate": 2.4366381634696867e-05, + "loss": 0.0, + "step": 13310 + }, + { + "epoch": 3.08, + "learning_rate": 2.4364455743009014e-05, + "loss": 0.0, + "step": 13311 + }, + { + "epoch": 3.08, + "learning_rate": 2.4362529851321165e-05, + "loss": 0.0001, + "step": 13312 + }, + { + "epoch": 3.08, + "learning_rate": 2.436060395963331e-05, + "loss": 0.0, + "step": 13313 + }, + { + "epoch": 3.08, + "learning_rate": 2.435867806794546e-05, + "loss": 0.0002, + "step": 13314 + }, + { + "epoch": 3.08, + "learning_rate": 2.4356752176257608e-05, + "loss": 0.0, + "step": 13315 + }, + { + "epoch": 3.08, + "learning_rate": 2.4354826284569755e-05, + "loss": 0.0, + "step": 13316 + }, + { + "epoch": 3.08, + "learning_rate": 2.4352900392881906e-05, + "loss": 0.0, + "step": 13317 + }, + { + "epoch": 3.08, + "learning_rate": 2.4350974501194054e-05, + "loss": 0.0001, + "step": 13318 + }, + { + "epoch": 3.08, + "learning_rate": 2.4349048609506204e-05, + "loss": 0.0, + "step": 13319 + }, + { + "epoch": 3.08, + "learning_rate": 2.4347122717818352e-05, + "loss": 0.0, + "step": 13320 + }, + { + "epoch": 3.08, + "learning_rate": 2.43451968261305e-05, + "loss": 0.0, + "step": 13321 + }, + { + "epoch": 3.08, + "learning_rate": 2.4343270934442647e-05, + "loss": 0.0, + "step": 13322 + }, + { + "epoch": 3.08, + "learning_rate": 2.4341345042754794e-05, + "loss": 0.0, + "step": 13323 + }, + { + "epoch": 3.08, + "learning_rate": 2.4339419151066945e-05, + "loss": 0.0, + "step": 13324 + }, + { + "epoch": 3.08, + "learning_rate": 2.4337493259379093e-05, + "loss": 0.0, + "step": 13325 + }, + { + "epoch": 3.08, + "learning_rate": 2.4335567367691244e-05, + "loss": 0.0001, + "step": 13326 + }, + { + "epoch": 3.08, + "learning_rate": 2.433364147600339e-05, + "loss": 0.0, + "step": 13327 + }, + { + "epoch": 3.08, + "learning_rate": 2.433171558431554e-05, + "loss": 0.0, + "step": 13328 + }, + { + "epoch": 3.08, + "learning_rate": 2.432978969262769e-05, + "loss": 0.0, + "step": 13329 + }, + { + "epoch": 3.08, + "learning_rate": 2.4327863800939837e-05, + "loss": 0.0, + "step": 13330 + }, + { + "epoch": 3.08, + "learning_rate": 2.4325937909251984e-05, + "loss": 0.0, + "step": 13331 + }, + { + "epoch": 3.08, + "learning_rate": 2.4324012017564132e-05, + "loss": 0.0, + "step": 13332 + }, + { + "epoch": 3.08, + "learning_rate": 2.4322086125876283e-05, + "loss": 0.0001, + "step": 13333 + }, + { + "epoch": 3.08, + "learning_rate": 2.432016023418843e-05, + "loss": 0.0, + "step": 13334 + }, + { + "epoch": 3.08, + "learning_rate": 2.4318234342500578e-05, + "loss": 0.0, + "step": 13335 + }, + { + "epoch": 3.08, + "learning_rate": 2.431630845081273e-05, + "loss": 0.0, + "step": 13336 + }, + { + "epoch": 3.08, + "learning_rate": 2.4314382559124876e-05, + "loss": 0.0, + "step": 13337 + }, + { + "epoch": 3.08, + "learning_rate": 2.4312456667437027e-05, + "loss": 0.0, + "step": 13338 + }, + { + "epoch": 3.08, + "learning_rate": 2.4310530775749174e-05, + "loss": 0.0, + "step": 13339 + }, + { + "epoch": 3.08, + "learning_rate": 2.4308604884061322e-05, + "loss": 0.0, + "step": 13340 + }, + { + "epoch": 3.08, + "learning_rate": 2.430667899237347e-05, + "loss": 0.0, + "step": 13341 + }, + { + "epoch": 3.08, + "learning_rate": 2.4304753100685617e-05, + "loss": 0.0, + "step": 13342 + }, + { + "epoch": 3.08, + "learning_rate": 2.4302827208997768e-05, + "loss": 0.0001, + "step": 13343 + }, + { + "epoch": 3.08, + "learning_rate": 2.4300901317309915e-05, + "loss": 0.0, + "step": 13344 + }, + { + "epoch": 3.08, + "learning_rate": 2.4298975425622063e-05, + "loss": 0.0, + "step": 13345 + }, + { + "epoch": 3.08, + "learning_rate": 2.4297049533934213e-05, + "loss": 0.0, + "step": 13346 + }, + { + "epoch": 3.08, + "learning_rate": 2.429512364224636e-05, + "loss": 0.0, + "step": 13347 + }, + { + "epoch": 3.08, + "learning_rate": 2.4293197750558512e-05, + "loss": 0.0, + "step": 13348 + }, + { + "epoch": 3.09, + "learning_rate": 2.429127185887066e-05, + "loss": 0.0, + "step": 13349 + }, + { + "epoch": 3.09, + "learning_rate": 2.4289345967182807e-05, + "loss": 0.0, + "step": 13350 + }, + { + "epoch": 3.09, + "learning_rate": 2.4287420075494954e-05, + "loss": 0.0, + "step": 13351 + }, + { + "epoch": 3.09, + "learning_rate": 2.4285494183807102e-05, + "loss": 0.0, + "step": 13352 + }, + { + "epoch": 3.09, + "learning_rate": 2.4283568292119253e-05, + "loss": 0.0, + "step": 13353 + }, + { + "epoch": 3.09, + "learning_rate": 2.42816424004314e-05, + "loss": 0.0, + "step": 13354 + }, + { + "epoch": 3.09, + "learning_rate": 2.427971650874355e-05, + "loss": 0.0, + "step": 13355 + }, + { + "epoch": 3.09, + "learning_rate": 2.42777906170557e-05, + "loss": 0.0, + "step": 13356 + }, + { + "epoch": 3.09, + "learning_rate": 2.4275864725367846e-05, + "loss": 0.0, + "step": 13357 + }, + { + "epoch": 3.09, + "learning_rate": 2.4273938833679997e-05, + "loss": 0.0, + "step": 13358 + }, + { + "epoch": 3.09, + "learning_rate": 2.427201294199214e-05, + "loss": 0.0, + "step": 13359 + }, + { + "epoch": 3.09, + "learning_rate": 2.427008705030429e-05, + "loss": 0.0, + "step": 13360 + }, + { + "epoch": 3.09, + "learning_rate": 2.426816115861644e-05, + "loss": 0.0076, + "step": 13361 + }, + { + "epoch": 3.09, + "learning_rate": 2.426623526692859e-05, + "loss": 0.0, + "step": 13362 + }, + { + "epoch": 3.09, + "learning_rate": 2.4264309375240738e-05, + "loss": 0.0, + "step": 13363 + }, + { + "epoch": 3.09, + "learning_rate": 2.4262383483552885e-05, + "loss": 0.0, + "step": 13364 + }, + { + "epoch": 3.09, + "learning_rate": 2.4260457591865036e-05, + "loss": 0.0, + "step": 13365 + }, + { + "epoch": 3.09, + "learning_rate": 2.4258531700177183e-05, + "loss": 0.0, + "step": 13366 + }, + { + "epoch": 3.09, + "learning_rate": 2.4256605808489334e-05, + "loss": 0.0, + "step": 13367 + }, + { + "epoch": 3.09, + "learning_rate": 2.425467991680148e-05, + "loss": 0.0, + "step": 13368 + }, + { + "epoch": 3.09, + "learning_rate": 2.425275402511363e-05, + "loss": 0.0, + "step": 13369 + }, + { + "epoch": 3.09, + "learning_rate": 2.4250828133425777e-05, + "loss": 0.0, + "step": 13370 + }, + { + "epoch": 3.09, + "learning_rate": 2.4248902241737924e-05, + "loss": 0.0, + "step": 13371 + }, + { + "epoch": 3.09, + "learning_rate": 2.4246976350050075e-05, + "loss": 0.0001, + "step": 13372 + }, + { + "epoch": 3.09, + "learning_rate": 2.4245050458362222e-05, + "loss": 0.0, + "step": 13373 + }, + { + "epoch": 3.09, + "learning_rate": 2.4243124566674373e-05, + "loss": 0.0, + "step": 13374 + }, + { + "epoch": 3.09, + "learning_rate": 2.424119867498652e-05, + "loss": 0.0, + "step": 13375 + }, + { + "epoch": 3.09, + "learning_rate": 2.4239272783298668e-05, + "loss": 0.0, + "step": 13376 + }, + { + "epoch": 3.09, + "learning_rate": 2.423734689161082e-05, + "loss": 0.0, + "step": 13377 + }, + { + "epoch": 3.09, + "learning_rate": 2.4235420999922963e-05, + "loss": 0.0, + "step": 13378 + }, + { + "epoch": 3.09, + "learning_rate": 2.4233495108235114e-05, + "loss": 0.0, + "step": 13379 + }, + { + "epoch": 3.09, + "learning_rate": 2.423156921654726e-05, + "loss": 0.0, + "step": 13380 + }, + { + "epoch": 3.09, + "learning_rate": 2.422964332485941e-05, + "loss": 0.0003, + "step": 13381 + }, + { + "epoch": 3.09, + "learning_rate": 2.422771743317156e-05, + "loss": 0.0, + "step": 13382 + }, + { + "epoch": 3.09, + "learning_rate": 2.4225791541483707e-05, + "loss": 0.0, + "step": 13383 + }, + { + "epoch": 3.09, + "learning_rate": 2.4223865649795858e-05, + "loss": 0.0, + "step": 13384 + }, + { + "epoch": 3.09, + "learning_rate": 2.4221939758108006e-05, + "loss": 0.0, + "step": 13385 + }, + { + "epoch": 3.09, + "learning_rate": 2.4220013866420153e-05, + "loss": 0.0, + "step": 13386 + }, + { + "epoch": 3.09, + "learning_rate": 2.42180879747323e-05, + "loss": 0.0, + "step": 13387 + }, + { + "epoch": 3.09, + "learning_rate": 2.4216162083044448e-05, + "loss": 0.0, + "step": 13388 + }, + { + "epoch": 3.09, + "learning_rate": 2.42142361913566e-05, + "loss": 0.0002, + "step": 13389 + }, + { + "epoch": 3.09, + "learning_rate": 2.4212310299668747e-05, + "loss": 0.0001, + "step": 13390 + }, + { + "epoch": 3.09, + "learning_rate": 2.4210384407980897e-05, + "loss": 0.0, + "step": 13391 + }, + { + "epoch": 3.09, + "learning_rate": 2.4208458516293045e-05, + "loss": 0.0, + "step": 13392 + }, + { + "epoch": 3.1, + "learning_rate": 2.4206532624605192e-05, + "loss": 0.0, + "step": 13393 + }, + { + "epoch": 3.1, + "learning_rate": 2.4204606732917343e-05, + "loss": 0.0, + "step": 13394 + }, + { + "epoch": 3.1, + "learning_rate": 2.420268084122949e-05, + "loss": 0.0, + "step": 13395 + }, + { + "epoch": 3.1, + "learning_rate": 2.420075494954164e-05, + "loss": 0.0, + "step": 13396 + }, + { + "epoch": 3.1, + "learning_rate": 2.4198829057853786e-05, + "loss": 0.0, + "step": 13397 + }, + { + "epoch": 3.1, + "learning_rate": 2.4196903166165936e-05, + "loss": 0.0, + "step": 13398 + }, + { + "epoch": 3.1, + "learning_rate": 2.4194977274478084e-05, + "loss": 0.0, + "step": 13399 + }, + { + "epoch": 3.1, + "learning_rate": 2.419305138279023e-05, + "loss": 0.0001, + "step": 13400 + }, + { + "epoch": 3.1, + "learning_rate": 2.4191125491102382e-05, + "loss": 0.0, + "step": 13401 + }, + { + "epoch": 3.1, + "learning_rate": 2.418919959941453e-05, + "loss": 0.0, + "step": 13402 + }, + { + "epoch": 3.1, + "learning_rate": 2.418727370772668e-05, + "loss": 0.0, + "step": 13403 + }, + { + "epoch": 3.1, + "learning_rate": 2.4185347816038828e-05, + "loss": 0.0, + "step": 13404 + }, + { + "epoch": 3.1, + "learning_rate": 2.4183421924350976e-05, + "loss": 0.0, + "step": 13405 + }, + { + "epoch": 3.1, + "learning_rate": 2.4181496032663123e-05, + "loss": 0.0, + "step": 13406 + }, + { + "epoch": 3.1, + "learning_rate": 2.417957014097527e-05, + "loss": 0.0, + "step": 13407 + }, + { + "epoch": 3.1, + "learning_rate": 2.417764424928742e-05, + "loss": 0.0, + "step": 13408 + }, + { + "epoch": 3.1, + "learning_rate": 2.417571835759957e-05, + "loss": 0.0, + "step": 13409 + }, + { + "epoch": 3.1, + "learning_rate": 2.417379246591172e-05, + "loss": 0.0001, + "step": 13410 + }, + { + "epoch": 3.1, + "learning_rate": 2.4171866574223867e-05, + "loss": 0.0027, + "step": 13411 + }, + { + "epoch": 3.1, + "learning_rate": 2.4169940682536015e-05, + "loss": 0.0, + "step": 13412 + }, + { + "epoch": 3.1, + "learning_rate": 2.4168014790848166e-05, + "loss": 0.0, + "step": 13413 + }, + { + "epoch": 3.1, + "learning_rate": 2.4166088899160313e-05, + "loss": 0.0, + "step": 13414 + }, + { + "epoch": 3.1, + "learning_rate": 2.416416300747246e-05, + "loss": 0.0, + "step": 13415 + }, + { + "epoch": 3.1, + "learning_rate": 2.4162237115784608e-05, + "loss": 0.0, + "step": 13416 + }, + { + "epoch": 3.1, + "learning_rate": 2.4160311224096756e-05, + "loss": 0.0, + "step": 13417 + }, + { + "epoch": 3.1, + "learning_rate": 2.4158385332408906e-05, + "loss": 0.0, + "step": 13418 + }, + { + "epoch": 3.1, + "learning_rate": 2.4156459440721054e-05, + "loss": 0.0, + "step": 13419 + }, + { + "epoch": 3.1, + "learning_rate": 2.4154533549033205e-05, + "loss": 0.0, + "step": 13420 + }, + { + "epoch": 3.1, + "learning_rate": 2.4152607657345352e-05, + "loss": 0.0, + "step": 13421 + }, + { + "epoch": 3.1, + "learning_rate": 2.41506817656575e-05, + "loss": 0.0, + "step": 13422 + }, + { + "epoch": 3.1, + "learning_rate": 2.414875587396965e-05, + "loss": 0.0, + "step": 13423 + }, + { + "epoch": 3.1, + "learning_rate": 2.4146829982281798e-05, + "loss": 0.0, + "step": 13424 + }, + { + "epoch": 3.1, + "learning_rate": 2.4144904090593946e-05, + "loss": 0.0, + "step": 13425 + }, + { + "epoch": 3.1, + "learning_rate": 2.4142978198906093e-05, + "loss": 0.0001, + "step": 13426 + }, + { + "epoch": 3.1, + "learning_rate": 2.4141052307218244e-05, + "loss": 0.0, + "step": 13427 + }, + { + "epoch": 3.1, + "learning_rate": 2.413912641553039e-05, + "loss": 0.0, + "step": 13428 + }, + { + "epoch": 3.1, + "learning_rate": 2.413720052384254e-05, + "loss": 0.0, + "step": 13429 + }, + { + "epoch": 3.1, + "learning_rate": 2.413527463215469e-05, + "loss": 0.0, + "step": 13430 + }, + { + "epoch": 3.1, + "learning_rate": 2.4133348740466837e-05, + "loss": 0.0, + "step": 13431 + }, + { + "epoch": 3.1, + "learning_rate": 2.4131422848778988e-05, + "loss": 0.0, + "step": 13432 + }, + { + "epoch": 3.1, + "learning_rate": 2.4129496957091135e-05, + "loss": 0.0, + "step": 13433 + }, + { + "epoch": 3.1, + "learning_rate": 2.4127571065403283e-05, + "loss": 0.0, + "step": 13434 + }, + { + "epoch": 3.1, + "learning_rate": 2.412564517371543e-05, + "loss": 0.0, + "step": 13435 + }, + { + "epoch": 3.11, + "learning_rate": 2.4123719282027578e-05, + "loss": 0.0, + "step": 13436 + }, + { + "epoch": 3.11, + "learning_rate": 2.412179339033973e-05, + "loss": 0.0001, + "step": 13437 + }, + { + "epoch": 3.11, + "learning_rate": 2.4119867498651876e-05, + "loss": 0.0, + "step": 13438 + }, + { + "epoch": 3.11, + "learning_rate": 2.4117941606964027e-05, + "loss": 0.0, + "step": 13439 + }, + { + "epoch": 3.11, + "learning_rate": 2.4116015715276175e-05, + "loss": 0.0001, + "step": 13440 + }, + { + "epoch": 3.11, + "learning_rate": 2.4114089823588322e-05, + "loss": 0.0, + "step": 13441 + }, + { + "epoch": 3.11, + "learning_rate": 2.4112163931900473e-05, + "loss": 0.0, + "step": 13442 + }, + { + "epoch": 3.11, + "learning_rate": 2.4110238040212617e-05, + "loss": 0.0, + "step": 13443 + }, + { + "epoch": 3.11, + "learning_rate": 2.4108312148524768e-05, + "loss": 0.0, + "step": 13444 + }, + { + "epoch": 3.11, + "learning_rate": 2.4106386256836915e-05, + "loss": 0.0, + "step": 13445 + }, + { + "epoch": 3.11, + "learning_rate": 2.4104460365149066e-05, + "loss": 0.0, + "step": 13446 + }, + { + "epoch": 3.11, + "learning_rate": 2.4102534473461214e-05, + "loss": 0.0, + "step": 13447 + }, + { + "epoch": 3.11, + "learning_rate": 2.410060858177336e-05, + "loss": 0.0052, + "step": 13448 + }, + { + "epoch": 3.11, + "learning_rate": 2.4098682690085512e-05, + "loss": 0.0, + "step": 13449 + }, + { + "epoch": 3.11, + "learning_rate": 2.409675679839766e-05, + "loss": 0.0, + "step": 13450 + }, + { + "epoch": 3.11, + "learning_rate": 2.409483090670981e-05, + "loss": 0.0002, + "step": 13451 + }, + { + "epoch": 3.11, + "learning_rate": 2.4092905015021958e-05, + "loss": 0.0, + "step": 13452 + }, + { + "epoch": 3.11, + "learning_rate": 2.4090979123334102e-05, + "loss": 0.0, + "step": 13453 + }, + { + "epoch": 3.11, + "learning_rate": 2.4089053231646253e-05, + "loss": 0.0, + "step": 13454 + }, + { + "epoch": 3.11, + "learning_rate": 2.40871273399584e-05, + "loss": 0.0, + "step": 13455 + }, + { + "epoch": 3.11, + "learning_rate": 2.408520144827055e-05, + "loss": 0.0, + "step": 13456 + }, + { + "epoch": 3.11, + "learning_rate": 2.40832755565827e-05, + "loss": 0.0, + "step": 13457 + }, + { + "epoch": 3.11, + "learning_rate": 2.4081349664894846e-05, + "loss": 0.0, + "step": 13458 + }, + { + "epoch": 3.11, + "learning_rate": 2.4079423773206997e-05, + "loss": 0.0001, + "step": 13459 + }, + { + "epoch": 3.11, + "learning_rate": 2.4077497881519144e-05, + "loss": 0.0, + "step": 13460 + }, + { + "epoch": 3.11, + "learning_rate": 2.4075571989831295e-05, + "loss": 0.0, + "step": 13461 + }, + { + "epoch": 3.11, + "learning_rate": 2.407364609814344e-05, + "loss": 0.0, + "step": 13462 + }, + { + "epoch": 3.11, + "learning_rate": 2.407172020645559e-05, + "loss": 0.0001, + "step": 13463 + }, + { + "epoch": 3.11, + "learning_rate": 2.4069794314767738e-05, + "loss": 0.0, + "step": 13464 + }, + { + "epoch": 3.11, + "learning_rate": 2.4067868423079885e-05, + "loss": 0.0, + "step": 13465 + }, + { + "epoch": 3.11, + "learning_rate": 2.4065942531392036e-05, + "loss": 0.0001, + "step": 13466 + }, + { + "epoch": 3.11, + "learning_rate": 2.4064016639704184e-05, + "loss": 0.0, + "step": 13467 + }, + { + "epoch": 3.11, + "learning_rate": 2.4062090748016334e-05, + "loss": 0.0, + "step": 13468 + }, + { + "epoch": 3.11, + "learning_rate": 2.4060164856328482e-05, + "loss": 0.0, + "step": 13469 + }, + { + "epoch": 3.11, + "learning_rate": 2.405823896464063e-05, + "loss": 0.0, + "step": 13470 + }, + { + "epoch": 3.11, + "learning_rate": 2.4056313072952777e-05, + "loss": 0.0, + "step": 13471 + }, + { + "epoch": 3.11, + "learning_rate": 2.4054387181264924e-05, + "loss": 0.0001, + "step": 13472 + }, + { + "epoch": 3.11, + "learning_rate": 2.4052461289577075e-05, + "loss": 0.0, + "step": 13473 + }, + { + "epoch": 3.11, + "learning_rate": 2.4050535397889223e-05, + "loss": 0.0, + "step": 13474 + }, + { + "epoch": 3.11, + "learning_rate": 2.4048609506201374e-05, + "loss": 0.0, + "step": 13475 + }, + { + "epoch": 3.11, + "learning_rate": 2.404668361451352e-05, + "loss": 0.0, + "step": 13476 + }, + { + "epoch": 3.11, + "learning_rate": 2.404475772282567e-05, + "loss": 0.0014, + "step": 13477 + }, + { + "epoch": 3.11, + "learning_rate": 2.404283183113782e-05, + "loss": 0.0, + "step": 13478 + }, + { + "epoch": 3.12, + "learning_rate": 2.4040905939449967e-05, + "loss": 0.0, + "step": 13479 + }, + { + "epoch": 3.12, + "learning_rate": 2.4038980047762118e-05, + "loss": 0.0, + "step": 13480 + }, + { + "epoch": 3.12, + "learning_rate": 2.4037054156074262e-05, + "loss": 0.0, + "step": 13481 + }, + { + "epoch": 3.12, + "learning_rate": 2.4035128264386413e-05, + "loss": 0.0, + "step": 13482 + }, + { + "epoch": 3.12, + "learning_rate": 2.403320237269856e-05, + "loss": 0.0, + "step": 13483 + }, + { + "epoch": 3.12, + "learning_rate": 2.4031276481010708e-05, + "loss": 0.0, + "step": 13484 + }, + { + "epoch": 3.12, + "learning_rate": 2.402935058932286e-05, + "loss": 0.0, + "step": 13485 + }, + { + "epoch": 3.12, + "learning_rate": 2.4027424697635006e-05, + "loss": 0.0, + "step": 13486 + }, + { + "epoch": 3.12, + "learning_rate": 2.4025498805947157e-05, + "loss": 0.0, + "step": 13487 + }, + { + "epoch": 3.12, + "learning_rate": 2.4023572914259304e-05, + "loss": 0.0, + "step": 13488 + }, + { + "epoch": 3.12, + "learning_rate": 2.4021647022571452e-05, + "loss": 0.0, + "step": 13489 + }, + { + "epoch": 3.12, + "learning_rate": 2.40197211308836e-05, + "loss": 0.0, + "step": 13490 + }, + { + "epoch": 3.12, + "learning_rate": 2.4017795239195747e-05, + "loss": 0.0001, + "step": 13491 + }, + { + "epoch": 3.12, + "learning_rate": 2.4015869347507898e-05, + "loss": 0.0004, + "step": 13492 + }, + { + "epoch": 3.12, + "learning_rate": 2.4013943455820045e-05, + "loss": 0.0, + "step": 13493 + }, + { + "epoch": 3.12, + "learning_rate": 2.4012017564132193e-05, + "loss": 0.0, + "step": 13494 + }, + { + "epoch": 3.12, + "learning_rate": 2.4010091672444343e-05, + "loss": 0.0002, + "step": 13495 + }, + { + "epoch": 3.12, + "learning_rate": 2.400816578075649e-05, + "loss": 0.0004, + "step": 13496 + }, + { + "epoch": 3.12, + "learning_rate": 2.4006239889068642e-05, + "loss": 0.0, + "step": 13497 + }, + { + "epoch": 3.12, + "learning_rate": 2.400431399738079e-05, + "loss": 0.0, + "step": 13498 + }, + { + "epoch": 3.12, + "learning_rate": 2.4002388105692937e-05, + "loss": 0.0, + "step": 13499 + }, + { + "epoch": 3.12, + "learning_rate": 2.4000462214005084e-05, + "loss": 0.0, + "step": 13500 + }, + { + "epoch": 3.12, + "learning_rate": 2.3998536322317232e-05, + "loss": 0.0, + "step": 13501 + }, + { + "epoch": 3.12, + "learning_rate": 2.3996610430629383e-05, + "loss": 0.0, + "step": 13502 + }, + { + "epoch": 3.12, + "learning_rate": 2.399468453894153e-05, + "loss": 0.0, + "step": 13503 + }, + { + "epoch": 3.12, + "learning_rate": 2.399275864725368e-05, + "loss": 0.0, + "step": 13504 + }, + { + "epoch": 3.12, + "learning_rate": 2.399083275556583e-05, + "loss": 0.0, + "step": 13505 + }, + { + "epoch": 3.12, + "learning_rate": 2.3988906863877976e-05, + "loss": 0.0, + "step": 13506 + }, + { + "epoch": 3.12, + "learning_rate": 2.3986980972190127e-05, + "loss": 0.0, + "step": 13507 + }, + { + "epoch": 3.12, + "learning_rate": 2.3985055080502274e-05, + "loss": 0.0, + "step": 13508 + }, + { + "epoch": 3.12, + "learning_rate": 2.3983129188814422e-05, + "loss": 0.0001, + "step": 13509 + }, + { + "epoch": 3.12, + "learning_rate": 2.398120329712657e-05, + "loss": 0.0, + "step": 13510 + }, + { + "epoch": 3.12, + "learning_rate": 2.397927740543872e-05, + "loss": 0.0, + "step": 13511 + }, + { + "epoch": 3.12, + "learning_rate": 2.3977351513750868e-05, + "loss": 0.0, + "step": 13512 + }, + { + "epoch": 3.12, + "learning_rate": 2.3975425622063015e-05, + "loss": 0.0001, + "step": 13513 + }, + { + "epoch": 3.12, + "learning_rate": 2.3973499730375166e-05, + "loss": 0.0, + "step": 13514 + }, + { + "epoch": 3.12, + "learning_rate": 2.3971573838687313e-05, + "loss": 0.0, + "step": 13515 + }, + { + "epoch": 3.12, + "learning_rate": 2.3969647946999464e-05, + "loss": 0.0, + "step": 13516 + }, + { + "epoch": 3.12, + "learning_rate": 2.3967722055311612e-05, + "loss": 0.0, + "step": 13517 + }, + { + "epoch": 3.12, + "learning_rate": 2.396579616362376e-05, + "loss": 0.0, + "step": 13518 + }, + { + "epoch": 3.12, + "learning_rate": 2.3963870271935907e-05, + "loss": 0.0, + "step": 13519 + }, + { + "epoch": 3.12, + "learning_rate": 2.3961944380248054e-05, + "loss": 0.0, + "step": 13520 + }, + { + "epoch": 3.12, + "learning_rate": 2.3960018488560205e-05, + "loss": 0.0, + "step": 13521 + }, + { + "epoch": 3.13, + "learning_rate": 2.3958092596872352e-05, + "loss": 0.0, + "step": 13522 + }, + { + "epoch": 3.13, + "learning_rate": 2.3956166705184503e-05, + "loss": 0.0, + "step": 13523 + }, + { + "epoch": 3.13, + "learning_rate": 2.395424081349665e-05, + "loss": 0.0, + "step": 13524 + }, + { + "epoch": 3.13, + "learning_rate": 2.39523149218088e-05, + "loss": 0.0, + "step": 13525 + }, + { + "epoch": 3.13, + "learning_rate": 2.395038903012095e-05, + "loss": 0.0, + "step": 13526 + }, + { + "epoch": 3.13, + "learning_rate": 2.3948463138433093e-05, + "loss": 0.0, + "step": 13527 + }, + { + "epoch": 3.13, + "learning_rate": 2.3946537246745244e-05, + "loss": 0.0002, + "step": 13528 + }, + { + "epoch": 3.13, + "learning_rate": 2.394461135505739e-05, + "loss": 0.0, + "step": 13529 + }, + { + "epoch": 3.13, + "learning_rate": 2.394268546336954e-05, + "loss": 0.0, + "step": 13530 + }, + { + "epoch": 3.13, + "learning_rate": 2.394075957168169e-05, + "loss": 0.0003, + "step": 13531 + }, + { + "epoch": 3.13, + "learning_rate": 2.3938833679993837e-05, + "loss": 0.0, + "step": 13532 + }, + { + "epoch": 3.13, + "learning_rate": 2.3936907788305988e-05, + "loss": 0.0002, + "step": 13533 + }, + { + "epoch": 3.13, + "learning_rate": 2.3934981896618136e-05, + "loss": 0.0, + "step": 13534 + }, + { + "epoch": 3.13, + "learning_rate": 2.3933056004930283e-05, + "loss": 0.0, + "step": 13535 + }, + { + "epoch": 3.13, + "learning_rate": 2.3931130113242434e-05, + "loss": 0.0, + "step": 13536 + }, + { + "epoch": 3.13, + "learning_rate": 2.3929204221554578e-05, + "loss": 0.0, + "step": 13537 + }, + { + "epoch": 3.13, + "learning_rate": 2.392727832986673e-05, + "loss": 0.0, + "step": 13538 + }, + { + "epoch": 3.13, + "learning_rate": 2.3925352438178877e-05, + "loss": 0.0001, + "step": 13539 + }, + { + "epoch": 3.13, + "learning_rate": 2.3923426546491027e-05, + "loss": 0.0, + "step": 13540 + }, + { + "epoch": 3.13, + "learning_rate": 2.3921500654803175e-05, + "loss": 0.0003, + "step": 13541 + }, + { + "epoch": 3.13, + "learning_rate": 2.3919574763115322e-05, + "loss": 0.0001, + "step": 13542 + }, + { + "epoch": 3.13, + "learning_rate": 2.3917648871427473e-05, + "loss": 0.0, + "step": 13543 + }, + { + "epoch": 3.13, + "learning_rate": 2.391572297973962e-05, + "loss": 0.0, + "step": 13544 + }, + { + "epoch": 3.13, + "learning_rate": 2.391379708805177e-05, + "loss": 0.0, + "step": 13545 + }, + { + "epoch": 3.13, + "learning_rate": 2.3911871196363916e-05, + "loss": 0.0005, + "step": 13546 + }, + { + "epoch": 3.13, + "learning_rate": 2.3909945304676067e-05, + "loss": 0.0054, + "step": 13547 + }, + { + "epoch": 3.13, + "learning_rate": 2.3908019412988214e-05, + "loss": 0.0, + "step": 13548 + }, + { + "epoch": 3.13, + "learning_rate": 2.390609352130036e-05, + "loss": 0.0001, + "step": 13549 + }, + { + "epoch": 3.13, + "learning_rate": 2.3904167629612512e-05, + "loss": 0.0, + "step": 13550 + }, + { + "epoch": 3.13, + "learning_rate": 2.390224173792466e-05, + "loss": 0.0013, + "step": 13551 + }, + { + "epoch": 3.13, + "learning_rate": 2.390031584623681e-05, + "loss": 0.0, + "step": 13552 + }, + { + "epoch": 3.13, + "learning_rate": 2.3898389954548958e-05, + "loss": 0.0, + "step": 13553 + }, + { + "epoch": 3.13, + "learning_rate": 2.3896464062861106e-05, + "loss": 0.0, + "step": 13554 + }, + { + "epoch": 3.13, + "learning_rate": 2.3894538171173253e-05, + "loss": 0.0, + "step": 13555 + }, + { + "epoch": 3.13, + "learning_rate": 2.38926122794854e-05, + "loss": 0.0, + "step": 13556 + }, + { + "epoch": 3.13, + "learning_rate": 2.389068638779755e-05, + "loss": 0.0, + "step": 13557 + }, + { + "epoch": 3.13, + "learning_rate": 2.38887604961097e-05, + "loss": 0.0, + "step": 13558 + }, + { + "epoch": 3.13, + "learning_rate": 2.388683460442185e-05, + "loss": 0.0, + "step": 13559 + }, + { + "epoch": 3.13, + "learning_rate": 2.3884908712733997e-05, + "loss": 0.0, + "step": 13560 + }, + { + "epoch": 3.13, + "learning_rate": 2.3882982821046145e-05, + "loss": 0.0, + "step": 13561 + }, + { + "epoch": 3.13, + "learning_rate": 2.3881056929358296e-05, + "loss": 0.0, + "step": 13562 + }, + { + "epoch": 3.13, + "learning_rate": 2.3879131037670443e-05, + "loss": 0.0, + "step": 13563 + }, + { + "epoch": 3.13, + "learning_rate": 2.3877205145982594e-05, + "loss": 0.0, + "step": 13564 + }, + { + "epoch": 3.13, + "learning_rate": 2.3875279254294738e-05, + "loss": 0.0, + "step": 13565 + }, + { + "epoch": 3.14, + "learning_rate": 2.387335336260689e-05, + "loss": 0.0, + "step": 13566 + }, + { + "epoch": 3.14, + "learning_rate": 2.3871427470919036e-05, + "loss": 0.0001, + "step": 13567 + }, + { + "epoch": 3.14, + "learning_rate": 2.3869501579231184e-05, + "loss": 0.0, + "step": 13568 + }, + { + "epoch": 3.14, + "learning_rate": 2.3867575687543335e-05, + "loss": 0.0, + "step": 13569 + }, + { + "epoch": 3.14, + "learning_rate": 2.3865649795855482e-05, + "loss": 0.0, + "step": 13570 + }, + { + "epoch": 3.14, + "learning_rate": 2.386372390416763e-05, + "loss": 0.0, + "step": 13571 + }, + { + "epoch": 3.14, + "learning_rate": 2.386179801247978e-05, + "loss": 0.001, + "step": 13572 + }, + { + "epoch": 3.14, + "learning_rate": 2.3859872120791928e-05, + "loss": 0.0, + "step": 13573 + }, + { + "epoch": 3.14, + "learning_rate": 2.3857946229104076e-05, + "loss": 0.0, + "step": 13574 + }, + { + "epoch": 3.14, + "learning_rate": 2.3856020337416223e-05, + "loss": 0.0, + "step": 13575 + }, + { + "epoch": 3.14, + "learning_rate": 2.3854094445728374e-05, + "loss": 0.0, + "step": 13576 + }, + { + "epoch": 3.14, + "learning_rate": 2.385216855404052e-05, + "loss": 0.0001, + "step": 13577 + }, + { + "epoch": 3.14, + "learning_rate": 2.385024266235267e-05, + "loss": 0.0, + "step": 13578 + }, + { + "epoch": 3.14, + "learning_rate": 2.384831677066482e-05, + "loss": 0.0, + "step": 13579 + }, + { + "epoch": 3.14, + "learning_rate": 2.3846390878976967e-05, + "loss": 0.0, + "step": 13580 + }, + { + "epoch": 3.14, + "learning_rate": 2.3844464987289118e-05, + "loss": 0.0, + "step": 13581 + }, + { + "epoch": 3.14, + "learning_rate": 2.3842539095601266e-05, + "loss": 0.0, + "step": 13582 + }, + { + "epoch": 3.14, + "learning_rate": 2.3840613203913413e-05, + "loss": 0.0, + "step": 13583 + }, + { + "epoch": 3.14, + "learning_rate": 2.383868731222556e-05, + "loss": 0.0, + "step": 13584 + }, + { + "epoch": 3.14, + "learning_rate": 2.3836761420537708e-05, + "loss": 0.0002, + "step": 13585 + }, + { + "epoch": 3.14, + "learning_rate": 2.383483552884986e-05, + "loss": 0.0, + "step": 13586 + }, + { + "epoch": 3.14, + "learning_rate": 2.3832909637162006e-05, + "loss": 0.0, + "step": 13587 + }, + { + "epoch": 3.14, + "learning_rate": 2.3830983745474157e-05, + "loss": 0.0, + "step": 13588 + }, + { + "epoch": 3.14, + "learning_rate": 2.3829057853786305e-05, + "loss": 0.0, + "step": 13589 + }, + { + "epoch": 3.14, + "learning_rate": 2.3827131962098452e-05, + "loss": 0.0, + "step": 13590 + }, + { + "epoch": 3.14, + "learning_rate": 2.3825206070410603e-05, + "loss": 0.0, + "step": 13591 + }, + { + "epoch": 3.14, + "learning_rate": 2.382328017872275e-05, + "loss": 0.0, + "step": 13592 + }, + { + "epoch": 3.14, + "learning_rate": 2.3821354287034898e-05, + "loss": 0.0, + "step": 13593 + }, + { + "epoch": 3.14, + "learning_rate": 2.3819428395347045e-05, + "loss": 0.0001, + "step": 13594 + }, + { + "epoch": 3.14, + "learning_rate": 2.3817502503659196e-05, + "loss": 0.0, + "step": 13595 + }, + { + "epoch": 3.14, + "learning_rate": 2.3815576611971344e-05, + "loss": 0.0001, + "step": 13596 + }, + { + "epoch": 3.14, + "learning_rate": 2.381365072028349e-05, + "loss": 0.0, + "step": 13597 + }, + { + "epoch": 3.14, + "learning_rate": 2.3811724828595642e-05, + "loss": 0.0, + "step": 13598 + }, + { + "epoch": 3.14, + "learning_rate": 2.380979893690779e-05, + "loss": 0.0, + "step": 13599 + }, + { + "epoch": 3.14, + "learning_rate": 2.380787304521994e-05, + "loss": 0.0, + "step": 13600 + }, + { + "epoch": 3.14, + "learning_rate": 2.3805947153532088e-05, + "loss": 0.0, + "step": 13601 + }, + { + "epoch": 3.14, + "learning_rate": 2.3804021261844235e-05, + "loss": 0.0, + "step": 13602 + }, + { + "epoch": 3.14, + "learning_rate": 2.3802095370156383e-05, + "loss": 0.0, + "step": 13603 + }, + { + "epoch": 3.14, + "learning_rate": 2.380016947846853e-05, + "loss": 0.0, + "step": 13604 + }, + { + "epoch": 3.14, + "learning_rate": 2.379824358678068e-05, + "loss": 0.0, + "step": 13605 + }, + { + "epoch": 3.14, + "learning_rate": 2.379631769509283e-05, + "loss": 0.0, + "step": 13606 + }, + { + "epoch": 3.14, + "learning_rate": 2.3794391803404976e-05, + "loss": 0.0, + "step": 13607 + }, + { + "epoch": 3.14, + "learning_rate": 2.3792465911717127e-05, + "loss": 0.0, + "step": 13608 + }, + { + "epoch": 3.15, + "learning_rate": 2.3790540020029275e-05, + "loss": 0.0, + "step": 13609 + }, + { + "epoch": 3.15, + "learning_rate": 2.3788614128341425e-05, + "loss": 0.0, + "step": 13610 + }, + { + "epoch": 3.15, + "learning_rate": 2.378668823665357e-05, + "loss": 0.0, + "step": 13611 + }, + { + "epoch": 3.15, + "learning_rate": 2.378476234496572e-05, + "loss": 0.0, + "step": 13612 + }, + { + "epoch": 3.15, + "learning_rate": 2.3782836453277868e-05, + "loss": 0.0001, + "step": 13613 + }, + { + "epoch": 3.15, + "learning_rate": 2.3780910561590015e-05, + "loss": 0.0, + "step": 13614 + }, + { + "epoch": 3.15, + "learning_rate": 2.3778984669902166e-05, + "loss": 0.0, + "step": 13615 + }, + { + "epoch": 3.15, + "learning_rate": 2.3777058778214314e-05, + "loss": 0.0001, + "step": 13616 + }, + { + "epoch": 3.15, + "learning_rate": 2.3775132886526465e-05, + "loss": 0.0, + "step": 13617 + }, + { + "epoch": 3.15, + "learning_rate": 2.3773206994838612e-05, + "loss": 0.0, + "step": 13618 + }, + { + "epoch": 3.15, + "learning_rate": 2.377128110315076e-05, + "loss": 0.0, + "step": 13619 + }, + { + "epoch": 3.15, + "learning_rate": 2.376935521146291e-05, + "loss": 0.0001, + "step": 13620 + }, + { + "epoch": 3.15, + "learning_rate": 2.3767429319775054e-05, + "loss": 0.0001, + "step": 13621 + }, + { + "epoch": 3.15, + "learning_rate": 2.3765503428087205e-05, + "loss": 0.0, + "step": 13622 + }, + { + "epoch": 3.15, + "learning_rate": 2.3763577536399353e-05, + "loss": 0.0, + "step": 13623 + }, + { + "epoch": 3.15, + "learning_rate": 2.3761651644711504e-05, + "loss": 0.0, + "step": 13624 + }, + { + "epoch": 3.15, + "learning_rate": 2.375972575302365e-05, + "loss": 0.0, + "step": 13625 + }, + { + "epoch": 3.15, + "learning_rate": 2.37577998613358e-05, + "loss": 0.0, + "step": 13626 + }, + { + "epoch": 3.15, + "learning_rate": 2.375587396964795e-05, + "loss": 0.0, + "step": 13627 + }, + { + "epoch": 3.15, + "learning_rate": 2.3753948077960097e-05, + "loss": 0.0, + "step": 13628 + }, + { + "epoch": 3.15, + "learning_rate": 2.3752022186272248e-05, + "loss": 0.0, + "step": 13629 + }, + { + "epoch": 3.15, + "learning_rate": 2.3750096294584392e-05, + "loss": 0.0001, + "step": 13630 + }, + { + "epoch": 3.15, + "learning_rate": 2.3748170402896543e-05, + "loss": 0.0, + "step": 13631 + }, + { + "epoch": 3.15, + "learning_rate": 2.374624451120869e-05, + "loss": 0.0, + "step": 13632 + }, + { + "epoch": 3.15, + "learning_rate": 2.3744318619520838e-05, + "loss": 0.0, + "step": 13633 + }, + { + "epoch": 3.15, + "learning_rate": 2.374239272783299e-05, + "loss": 0.0, + "step": 13634 + }, + { + "epoch": 3.15, + "learning_rate": 2.3740466836145136e-05, + "loss": 0.0, + "step": 13635 + }, + { + "epoch": 3.15, + "learning_rate": 2.3738540944457287e-05, + "loss": 0.0, + "step": 13636 + }, + { + "epoch": 3.15, + "learning_rate": 2.3736615052769434e-05, + "loss": 0.0, + "step": 13637 + }, + { + "epoch": 3.15, + "learning_rate": 2.3734689161081582e-05, + "loss": 0.0, + "step": 13638 + }, + { + "epoch": 3.15, + "learning_rate": 2.373276326939373e-05, + "loss": 0.0, + "step": 13639 + }, + { + "epoch": 3.15, + "learning_rate": 2.3730837377705877e-05, + "loss": 0.0, + "step": 13640 + }, + { + "epoch": 3.15, + "learning_rate": 2.3728911486018028e-05, + "loss": 0.0, + "step": 13641 + }, + { + "epoch": 3.15, + "learning_rate": 2.3726985594330175e-05, + "loss": 0.0, + "step": 13642 + }, + { + "epoch": 3.15, + "learning_rate": 2.3725059702642326e-05, + "loss": 0.0, + "step": 13643 + }, + { + "epoch": 3.15, + "learning_rate": 2.3723133810954474e-05, + "loss": 0.0, + "step": 13644 + }, + { + "epoch": 3.15, + "learning_rate": 2.372120791926662e-05, + "loss": 0.0, + "step": 13645 + }, + { + "epoch": 3.15, + "learning_rate": 2.3719282027578772e-05, + "loss": 0.0, + "step": 13646 + }, + { + "epoch": 3.15, + "learning_rate": 2.371735613589092e-05, + "loss": 0.0001, + "step": 13647 + }, + { + "epoch": 3.15, + "learning_rate": 2.3715430244203067e-05, + "loss": 0.0, + "step": 13648 + }, + { + "epoch": 3.15, + "learning_rate": 2.3713504352515214e-05, + "loss": 0.0, + "step": 13649 + }, + { + "epoch": 3.15, + "learning_rate": 2.3711578460827362e-05, + "loss": 0.0, + "step": 13650 + }, + { + "epoch": 3.15, + "learning_rate": 2.3709652569139513e-05, + "loss": 0.0, + "step": 13651 + }, + { + "epoch": 3.16, + "learning_rate": 2.370772667745166e-05, + "loss": 0.0, + "step": 13652 + }, + { + "epoch": 3.16, + "learning_rate": 2.370580078576381e-05, + "loss": 0.0, + "step": 13653 + }, + { + "epoch": 3.16, + "learning_rate": 2.370387489407596e-05, + "loss": 0.0, + "step": 13654 + }, + { + "epoch": 3.16, + "learning_rate": 2.3701949002388106e-05, + "loss": 0.0, + "step": 13655 + }, + { + "epoch": 3.16, + "learning_rate": 2.3700023110700257e-05, + "loss": 0.0, + "step": 13656 + }, + { + "epoch": 3.16, + "learning_rate": 2.3698097219012404e-05, + "loss": 0.0, + "step": 13657 + }, + { + "epoch": 3.16, + "learning_rate": 2.3696171327324552e-05, + "loss": 0.0002, + "step": 13658 + }, + { + "epoch": 3.16, + "learning_rate": 2.36942454356367e-05, + "loss": 0.0001, + "step": 13659 + }, + { + "epoch": 3.16, + "learning_rate": 2.369231954394885e-05, + "loss": 0.0, + "step": 13660 + }, + { + "epoch": 3.16, + "learning_rate": 2.3690393652260998e-05, + "loss": 0.0, + "step": 13661 + }, + { + "epoch": 3.16, + "learning_rate": 2.3688467760573145e-05, + "loss": 0.0, + "step": 13662 + }, + { + "epoch": 3.16, + "learning_rate": 2.3686541868885296e-05, + "loss": 0.0001, + "step": 13663 + }, + { + "epoch": 3.16, + "learning_rate": 2.3684615977197443e-05, + "loss": 0.0, + "step": 13664 + }, + { + "epoch": 3.16, + "learning_rate": 2.3682690085509594e-05, + "loss": 0.0, + "step": 13665 + }, + { + "epoch": 3.16, + "learning_rate": 2.3680764193821742e-05, + "loss": 0.0, + "step": 13666 + }, + { + "epoch": 3.16, + "learning_rate": 2.367883830213389e-05, + "loss": 0.0, + "step": 13667 + }, + { + "epoch": 3.16, + "learning_rate": 2.3676912410446037e-05, + "loss": 0.0, + "step": 13668 + }, + { + "epoch": 3.16, + "learning_rate": 2.3674986518758184e-05, + "loss": 0.0048, + "step": 13669 + }, + { + "epoch": 3.16, + "learning_rate": 2.3673060627070335e-05, + "loss": 0.0, + "step": 13670 + }, + { + "epoch": 3.16, + "learning_rate": 2.3671134735382483e-05, + "loss": 0.0002, + "step": 13671 + }, + { + "epoch": 3.16, + "learning_rate": 2.3669208843694633e-05, + "loss": 0.0, + "step": 13672 + }, + { + "epoch": 3.16, + "learning_rate": 2.366728295200678e-05, + "loss": 0.0, + "step": 13673 + }, + { + "epoch": 3.16, + "learning_rate": 2.366535706031893e-05, + "loss": 0.0, + "step": 13674 + }, + { + "epoch": 3.16, + "learning_rate": 2.366343116863108e-05, + "loss": 0.0, + "step": 13675 + }, + { + "epoch": 3.16, + "learning_rate": 2.3661505276943227e-05, + "loss": 0.0, + "step": 13676 + }, + { + "epoch": 3.16, + "learning_rate": 2.3659579385255374e-05, + "loss": 0.0002, + "step": 13677 + }, + { + "epoch": 3.16, + "learning_rate": 2.365765349356752e-05, + "loss": 0.0, + "step": 13678 + }, + { + "epoch": 3.16, + "learning_rate": 2.3655727601879673e-05, + "loss": 0.0, + "step": 13679 + }, + { + "epoch": 3.16, + "learning_rate": 2.365380171019182e-05, + "loss": 0.0001, + "step": 13680 + }, + { + "epoch": 3.16, + "learning_rate": 2.3651875818503967e-05, + "loss": 0.0001, + "step": 13681 + }, + { + "epoch": 3.16, + "learning_rate": 2.364994992681612e-05, + "loss": 0.0002, + "step": 13682 + }, + { + "epoch": 3.16, + "learning_rate": 2.3648024035128266e-05, + "loss": 0.0, + "step": 13683 + }, + { + "epoch": 3.16, + "learning_rate": 2.3646098143440413e-05, + "loss": 0.0, + "step": 13684 + }, + { + "epoch": 3.16, + "learning_rate": 2.3644172251752564e-05, + "loss": 0.0, + "step": 13685 + }, + { + "epoch": 3.16, + "learning_rate": 2.3642246360064708e-05, + "loss": 0.0, + "step": 13686 + }, + { + "epoch": 3.16, + "learning_rate": 2.364032046837686e-05, + "loss": 0.0, + "step": 13687 + }, + { + "epoch": 3.16, + "learning_rate": 2.3638394576689007e-05, + "loss": 0.0004, + "step": 13688 + }, + { + "epoch": 3.16, + "learning_rate": 2.3636468685001157e-05, + "loss": 0.0, + "step": 13689 + }, + { + "epoch": 3.16, + "learning_rate": 2.3634542793313305e-05, + "loss": 0.0, + "step": 13690 + }, + { + "epoch": 3.16, + "learning_rate": 2.3632616901625452e-05, + "loss": 0.0, + "step": 13691 + }, + { + "epoch": 3.16, + "learning_rate": 2.3630691009937603e-05, + "loss": 0.0, + "step": 13692 + }, + { + "epoch": 3.16, + "learning_rate": 2.362876511824975e-05, + "loss": 0.0, + "step": 13693 + }, + { + "epoch": 3.16, + "learning_rate": 2.36268392265619e-05, + "loss": 0.0, + "step": 13694 + }, + { + "epoch": 3.17, + "learning_rate": 2.3624913334874046e-05, + "loss": 0.0, + "step": 13695 + }, + { + "epoch": 3.17, + "learning_rate": 2.3622987443186197e-05, + "loss": 0.0, + "step": 13696 + }, + { + "epoch": 3.17, + "learning_rate": 2.3621061551498344e-05, + "loss": 0.0, + "step": 13697 + }, + { + "epoch": 3.17, + "learning_rate": 2.361913565981049e-05, + "loss": 0.0, + "step": 13698 + }, + { + "epoch": 3.17, + "learning_rate": 2.3617209768122642e-05, + "loss": 0.0, + "step": 13699 + }, + { + "epoch": 3.17, + "learning_rate": 2.361528387643479e-05, + "loss": 0.0, + "step": 13700 + }, + { + "epoch": 3.17, + "learning_rate": 2.361335798474694e-05, + "loss": 0.0, + "step": 13701 + }, + { + "epoch": 3.17, + "learning_rate": 2.3611432093059088e-05, + "loss": 0.0, + "step": 13702 + }, + { + "epoch": 3.17, + "learning_rate": 2.3609506201371236e-05, + "loss": 0.0, + "step": 13703 + }, + { + "epoch": 3.17, + "learning_rate": 2.3607580309683387e-05, + "loss": 0.0, + "step": 13704 + }, + { + "epoch": 3.17, + "learning_rate": 2.360565441799553e-05, + "loss": 0.0, + "step": 13705 + }, + { + "epoch": 3.17, + "learning_rate": 2.360372852630768e-05, + "loss": 0.0, + "step": 13706 + }, + { + "epoch": 3.17, + "learning_rate": 2.360180263461983e-05, + "loss": 0.0, + "step": 13707 + }, + { + "epoch": 3.17, + "learning_rate": 2.359987674293198e-05, + "loss": 0.0, + "step": 13708 + }, + { + "epoch": 3.17, + "learning_rate": 2.3597950851244127e-05, + "loss": 0.0, + "step": 13709 + }, + { + "epoch": 3.17, + "learning_rate": 2.3596024959556275e-05, + "loss": 0.0, + "step": 13710 + }, + { + "epoch": 3.17, + "learning_rate": 2.3594099067868426e-05, + "loss": 0.0, + "step": 13711 + }, + { + "epoch": 3.17, + "learning_rate": 2.3592173176180573e-05, + "loss": 0.0001, + "step": 13712 + }, + { + "epoch": 3.17, + "learning_rate": 2.3590247284492724e-05, + "loss": 0.0001, + "step": 13713 + }, + { + "epoch": 3.17, + "learning_rate": 2.3588321392804868e-05, + "loss": 0.0, + "step": 13714 + }, + { + "epoch": 3.17, + "learning_rate": 2.358639550111702e-05, + "loss": 0.0, + "step": 13715 + }, + { + "epoch": 3.17, + "learning_rate": 2.3584469609429166e-05, + "loss": 0.0, + "step": 13716 + }, + { + "epoch": 3.17, + "learning_rate": 2.3582543717741314e-05, + "loss": 0.0, + "step": 13717 + }, + { + "epoch": 3.17, + "learning_rate": 2.3580617826053465e-05, + "loss": 0.0, + "step": 13718 + }, + { + "epoch": 3.17, + "learning_rate": 2.3578691934365612e-05, + "loss": 0.0, + "step": 13719 + }, + { + "epoch": 3.17, + "learning_rate": 2.357676604267776e-05, + "loss": 0.0, + "step": 13720 + }, + { + "epoch": 3.17, + "learning_rate": 2.357484015098991e-05, + "loss": 0.0, + "step": 13721 + }, + { + "epoch": 3.17, + "learning_rate": 2.3572914259302058e-05, + "loss": 0.0, + "step": 13722 + }, + { + "epoch": 3.17, + "learning_rate": 2.3570988367614206e-05, + "loss": 0.0, + "step": 13723 + }, + { + "epoch": 3.17, + "learning_rate": 2.3569062475926353e-05, + "loss": 0.0, + "step": 13724 + }, + { + "epoch": 3.17, + "learning_rate": 2.3567136584238504e-05, + "loss": 0.0, + "step": 13725 + }, + { + "epoch": 3.17, + "learning_rate": 2.356521069255065e-05, + "loss": 0.0, + "step": 13726 + }, + { + "epoch": 3.17, + "learning_rate": 2.35632848008628e-05, + "loss": 0.0, + "step": 13727 + }, + { + "epoch": 3.17, + "learning_rate": 2.356135890917495e-05, + "loss": 0.0, + "step": 13728 + }, + { + "epoch": 3.17, + "learning_rate": 2.3559433017487097e-05, + "loss": 0.0, + "step": 13729 + }, + { + "epoch": 3.17, + "learning_rate": 2.3557507125799248e-05, + "loss": 0.0, + "step": 13730 + }, + { + "epoch": 3.17, + "learning_rate": 2.3555581234111396e-05, + "loss": 0.0, + "step": 13731 + }, + { + "epoch": 3.17, + "learning_rate": 2.3553655342423543e-05, + "loss": 0.0, + "step": 13732 + }, + { + "epoch": 3.17, + "learning_rate": 2.355172945073569e-05, + "loss": 0.0, + "step": 13733 + }, + { + "epoch": 3.17, + "learning_rate": 2.3549803559047838e-05, + "loss": 0.0, + "step": 13734 + }, + { + "epoch": 3.17, + "learning_rate": 2.354787766735999e-05, + "loss": 0.0, + "step": 13735 + }, + { + "epoch": 3.17, + "learning_rate": 2.3545951775672136e-05, + "loss": 0.0, + "step": 13736 + }, + { + "epoch": 3.17, + "learning_rate": 2.3544025883984287e-05, + "loss": 0.0001, + "step": 13737 + }, + { + "epoch": 3.17, + "learning_rate": 2.3542099992296435e-05, + "loss": 0.0, + "step": 13738 + }, + { + "epoch": 3.18, + "learning_rate": 2.3540174100608582e-05, + "loss": 0.0, + "step": 13739 + }, + { + "epoch": 3.18, + "learning_rate": 2.3538248208920733e-05, + "loss": 0.0, + "step": 13740 + }, + { + "epoch": 3.18, + "learning_rate": 2.353632231723288e-05, + "loss": 0.0, + "step": 13741 + }, + { + "epoch": 3.18, + "learning_rate": 2.3534396425545028e-05, + "loss": 0.0, + "step": 13742 + }, + { + "epoch": 3.18, + "learning_rate": 2.3532470533857175e-05, + "loss": 0.0, + "step": 13743 + }, + { + "epoch": 3.18, + "learning_rate": 2.3530544642169326e-05, + "loss": 0.0, + "step": 13744 + }, + { + "epoch": 3.18, + "learning_rate": 2.3528618750481474e-05, + "loss": 0.0, + "step": 13745 + }, + { + "epoch": 3.18, + "learning_rate": 2.352669285879362e-05, + "loss": 0.0, + "step": 13746 + }, + { + "epoch": 3.18, + "learning_rate": 2.3524766967105772e-05, + "loss": 0.0001, + "step": 13747 + }, + { + "epoch": 3.18, + "learning_rate": 2.352284107541792e-05, + "loss": 0.0003, + "step": 13748 + }, + { + "epoch": 3.18, + "learning_rate": 2.352091518373007e-05, + "loss": 0.0, + "step": 13749 + }, + { + "epoch": 3.18, + "learning_rate": 2.3518989292042218e-05, + "loss": 0.0001, + "step": 13750 + }, + { + "epoch": 3.18, + "learning_rate": 2.3517063400354365e-05, + "loss": 0.0, + "step": 13751 + }, + { + "epoch": 3.18, + "learning_rate": 2.3515137508666513e-05, + "loss": 0.0, + "step": 13752 + }, + { + "epoch": 3.18, + "learning_rate": 2.351321161697866e-05, + "loss": 0.0, + "step": 13753 + }, + { + "epoch": 3.18, + "learning_rate": 2.351128572529081e-05, + "loss": 0.0001, + "step": 13754 + }, + { + "epoch": 3.18, + "learning_rate": 2.350935983360296e-05, + "loss": 0.0, + "step": 13755 + }, + { + "epoch": 3.18, + "learning_rate": 2.350743394191511e-05, + "loss": 0.0, + "step": 13756 + }, + { + "epoch": 3.18, + "learning_rate": 2.3505508050227257e-05, + "loss": 0.0, + "step": 13757 + }, + { + "epoch": 3.18, + "learning_rate": 2.3503582158539405e-05, + "loss": 0.0, + "step": 13758 + }, + { + "epoch": 3.18, + "learning_rate": 2.3501656266851555e-05, + "loss": 0.0, + "step": 13759 + }, + { + "epoch": 3.18, + "learning_rate": 2.3499730375163703e-05, + "loss": 0.0, + "step": 13760 + }, + { + "epoch": 3.18, + "learning_rate": 2.349780448347585e-05, + "loss": 0.0683, + "step": 13761 + }, + { + "epoch": 3.18, + "learning_rate": 2.3495878591787998e-05, + "loss": 0.0, + "step": 13762 + }, + { + "epoch": 3.18, + "learning_rate": 2.3493952700100145e-05, + "loss": 0.0, + "step": 13763 + }, + { + "epoch": 3.18, + "learning_rate": 2.3492026808412296e-05, + "loss": 0.0, + "step": 13764 + }, + { + "epoch": 3.18, + "learning_rate": 2.3490100916724444e-05, + "loss": 0.0, + "step": 13765 + }, + { + "epoch": 3.18, + "learning_rate": 2.3488175025036595e-05, + "loss": 0.0, + "step": 13766 + }, + { + "epoch": 3.18, + "learning_rate": 2.3486249133348742e-05, + "loss": 0.0, + "step": 13767 + }, + { + "epoch": 3.18, + "learning_rate": 2.348432324166089e-05, + "loss": 0.0001, + "step": 13768 + }, + { + "epoch": 3.18, + "learning_rate": 2.348239734997304e-05, + "loss": 0.0, + "step": 13769 + }, + { + "epoch": 3.18, + "learning_rate": 2.3480471458285184e-05, + "loss": 0.0, + "step": 13770 + }, + { + "epoch": 3.18, + "learning_rate": 2.3478545566597335e-05, + "loss": 0.0, + "step": 13771 + }, + { + "epoch": 3.18, + "learning_rate": 2.3476619674909483e-05, + "loss": 0.0, + "step": 13772 + }, + { + "epoch": 3.18, + "learning_rate": 2.3474693783221634e-05, + "loss": 0.0, + "step": 13773 + }, + { + "epoch": 3.18, + "learning_rate": 2.347276789153378e-05, + "loss": 0.0, + "step": 13774 + }, + { + "epoch": 3.18, + "learning_rate": 2.347084199984593e-05, + "loss": 0.0, + "step": 13775 + }, + { + "epoch": 3.18, + "learning_rate": 2.346891610815808e-05, + "loss": 0.0, + "step": 13776 + }, + { + "epoch": 3.18, + "learning_rate": 2.3466990216470227e-05, + "loss": 0.0, + "step": 13777 + }, + { + "epoch": 3.18, + "learning_rate": 2.3465064324782378e-05, + "loss": 0.0001, + "step": 13778 + }, + { + "epoch": 3.18, + "learning_rate": 2.3463138433094522e-05, + "loss": 0.0, + "step": 13779 + }, + { + "epoch": 3.18, + "learning_rate": 2.3461212541406673e-05, + "loss": 0.0, + "step": 13780 + }, + { + "epoch": 3.18, + "learning_rate": 2.345928664971882e-05, + "loss": 0.0, + "step": 13781 + }, + { + "epoch": 3.19, + "learning_rate": 2.3457360758030968e-05, + "loss": 0.0, + "step": 13782 + }, + { + "epoch": 3.19, + "learning_rate": 2.345543486634312e-05, + "loss": 0.0, + "step": 13783 + }, + { + "epoch": 3.19, + "learning_rate": 2.3453508974655266e-05, + "loss": 0.0, + "step": 13784 + }, + { + "epoch": 3.19, + "learning_rate": 2.3451583082967417e-05, + "loss": 0.0, + "step": 13785 + }, + { + "epoch": 3.19, + "learning_rate": 2.3449657191279564e-05, + "loss": 0.0, + "step": 13786 + }, + { + "epoch": 3.19, + "learning_rate": 2.3447731299591712e-05, + "loss": 0.0, + "step": 13787 + }, + { + "epoch": 3.19, + "learning_rate": 2.3445805407903863e-05, + "loss": 0.0, + "step": 13788 + }, + { + "epoch": 3.19, + "learning_rate": 2.3443879516216007e-05, + "loss": 0.0, + "step": 13789 + }, + { + "epoch": 3.19, + "learning_rate": 2.3441953624528158e-05, + "loss": 0.0, + "step": 13790 + }, + { + "epoch": 3.19, + "learning_rate": 2.3440027732840305e-05, + "loss": 0.0001, + "step": 13791 + }, + { + "epoch": 3.19, + "learning_rate": 2.3438101841152456e-05, + "loss": 0.0, + "step": 13792 + }, + { + "epoch": 3.19, + "learning_rate": 2.3436175949464604e-05, + "loss": 0.0, + "step": 13793 + }, + { + "epoch": 3.19, + "learning_rate": 2.343425005777675e-05, + "loss": 0.0001, + "step": 13794 + }, + { + "epoch": 3.19, + "learning_rate": 2.3432324166088902e-05, + "loss": 0.0, + "step": 13795 + }, + { + "epoch": 3.19, + "learning_rate": 2.343039827440105e-05, + "loss": 0.0, + "step": 13796 + }, + { + "epoch": 3.19, + "learning_rate": 2.3428472382713197e-05, + "loss": 0.0, + "step": 13797 + }, + { + "epoch": 3.19, + "learning_rate": 2.3426546491025344e-05, + "loss": 0.0, + "step": 13798 + }, + { + "epoch": 3.19, + "learning_rate": 2.3424620599337492e-05, + "loss": 0.0001, + "step": 13799 + }, + { + "epoch": 3.19, + "learning_rate": 2.3422694707649643e-05, + "loss": 0.0, + "step": 13800 + }, + { + "epoch": 3.19, + "learning_rate": 2.342076881596179e-05, + "loss": 0.0, + "step": 13801 + }, + { + "epoch": 3.19, + "learning_rate": 2.341884292427394e-05, + "loss": 0.0, + "step": 13802 + }, + { + "epoch": 3.19, + "learning_rate": 2.341691703258609e-05, + "loss": 0.0001, + "step": 13803 + }, + { + "epoch": 3.19, + "learning_rate": 2.3414991140898236e-05, + "loss": 0.0, + "step": 13804 + }, + { + "epoch": 3.19, + "learning_rate": 2.3413065249210387e-05, + "loss": 0.0048, + "step": 13805 + }, + { + "epoch": 3.19, + "learning_rate": 2.3411139357522534e-05, + "loss": 0.0, + "step": 13806 + }, + { + "epoch": 3.19, + "learning_rate": 2.3409213465834682e-05, + "loss": 0.0, + "step": 13807 + }, + { + "epoch": 3.19, + "learning_rate": 2.340728757414683e-05, + "loss": 0.0, + "step": 13808 + }, + { + "epoch": 3.19, + "learning_rate": 2.340536168245898e-05, + "loss": 0.0, + "step": 13809 + }, + { + "epoch": 3.19, + "learning_rate": 2.3403435790771128e-05, + "loss": 0.0, + "step": 13810 + }, + { + "epoch": 3.19, + "learning_rate": 2.3401509899083275e-05, + "loss": 0.0, + "step": 13811 + }, + { + "epoch": 3.19, + "learning_rate": 2.3399584007395426e-05, + "loss": 0.0, + "step": 13812 + }, + { + "epoch": 3.19, + "learning_rate": 2.3397658115707573e-05, + "loss": 0.0, + "step": 13813 + }, + { + "epoch": 3.19, + "learning_rate": 2.3395732224019724e-05, + "loss": 0.0001, + "step": 13814 + }, + { + "epoch": 3.19, + "learning_rate": 2.3393806332331872e-05, + "loss": 0.0, + "step": 13815 + }, + { + "epoch": 3.19, + "learning_rate": 2.339188044064402e-05, + "loss": 0.0, + "step": 13816 + }, + { + "epoch": 3.19, + "learning_rate": 2.3389954548956167e-05, + "loss": 0.0, + "step": 13817 + }, + { + "epoch": 3.19, + "learning_rate": 2.3388028657268314e-05, + "loss": 0.0, + "step": 13818 + }, + { + "epoch": 3.19, + "learning_rate": 2.3386102765580465e-05, + "loss": 0.0, + "step": 13819 + }, + { + "epoch": 3.19, + "learning_rate": 2.3384176873892613e-05, + "loss": 0.0, + "step": 13820 + }, + { + "epoch": 3.19, + "learning_rate": 2.3382250982204763e-05, + "loss": 0.0, + "step": 13821 + }, + { + "epoch": 3.19, + "learning_rate": 2.338032509051691e-05, + "loss": 0.0, + "step": 13822 + }, + { + "epoch": 3.19, + "learning_rate": 2.337839919882906e-05, + "loss": 0.0, + "step": 13823 + }, + { + "epoch": 3.19, + "learning_rate": 2.337647330714121e-05, + "loss": 0.0, + "step": 13824 + }, + { + "epoch": 3.2, + "learning_rate": 2.3374547415453357e-05, + "loss": 0.0001, + "step": 13825 + }, + { + "epoch": 3.2, + "learning_rate": 2.3372621523765504e-05, + "loss": 0.0, + "step": 13826 + }, + { + "epoch": 3.2, + "learning_rate": 2.3370695632077652e-05, + "loss": 0.0, + "step": 13827 + }, + { + "epoch": 3.2, + "learning_rate": 2.3368769740389803e-05, + "loss": 0.0001, + "step": 13828 + }, + { + "epoch": 3.2, + "learning_rate": 2.336684384870195e-05, + "loss": 0.0, + "step": 13829 + }, + { + "epoch": 3.2, + "learning_rate": 2.3364917957014098e-05, + "loss": 0.0, + "step": 13830 + }, + { + "epoch": 3.2, + "learning_rate": 2.336299206532625e-05, + "loss": 0.0, + "step": 13831 + }, + { + "epoch": 3.2, + "learning_rate": 2.3361066173638396e-05, + "loss": 0.0, + "step": 13832 + }, + { + "epoch": 3.2, + "learning_rate": 2.3359140281950547e-05, + "loss": 0.0, + "step": 13833 + }, + { + "epoch": 3.2, + "learning_rate": 2.3357214390262694e-05, + "loss": 0.0, + "step": 13834 + }, + { + "epoch": 3.2, + "learning_rate": 2.3355288498574838e-05, + "loss": 0.0, + "step": 13835 + }, + { + "epoch": 3.2, + "learning_rate": 2.335336260688699e-05, + "loss": 0.0, + "step": 13836 + }, + { + "epoch": 3.2, + "learning_rate": 2.3351436715199137e-05, + "loss": 0.0, + "step": 13837 + }, + { + "epoch": 3.2, + "learning_rate": 2.3349510823511288e-05, + "loss": 0.0, + "step": 13838 + }, + { + "epoch": 3.2, + "learning_rate": 2.3347584931823435e-05, + "loss": 0.0002, + "step": 13839 + }, + { + "epoch": 3.2, + "learning_rate": 2.3345659040135582e-05, + "loss": 0.0, + "step": 13840 + }, + { + "epoch": 3.2, + "learning_rate": 2.3343733148447733e-05, + "loss": 0.0, + "step": 13841 + }, + { + "epoch": 3.2, + "learning_rate": 2.334180725675988e-05, + "loss": 0.0, + "step": 13842 + }, + { + "epoch": 3.2, + "learning_rate": 2.333988136507203e-05, + "loss": 0.0, + "step": 13843 + }, + { + "epoch": 3.2, + "learning_rate": 2.333795547338418e-05, + "loss": 0.0, + "step": 13844 + }, + { + "epoch": 3.2, + "learning_rate": 2.3336029581696327e-05, + "loss": 0.0002, + "step": 13845 + }, + { + "epoch": 3.2, + "learning_rate": 2.3334103690008474e-05, + "loss": 0.0, + "step": 13846 + }, + { + "epoch": 3.2, + "learning_rate": 2.333217779832062e-05, + "loss": 0.0, + "step": 13847 + }, + { + "epoch": 3.2, + "learning_rate": 2.3330251906632772e-05, + "loss": 0.0, + "step": 13848 + }, + { + "epoch": 3.2, + "learning_rate": 2.332832601494492e-05, + "loss": 0.0, + "step": 13849 + }, + { + "epoch": 3.2, + "learning_rate": 2.332640012325707e-05, + "loss": 0.0, + "step": 13850 + }, + { + "epoch": 3.2, + "learning_rate": 2.3324474231569218e-05, + "loss": 0.0, + "step": 13851 + }, + { + "epoch": 3.2, + "learning_rate": 2.3322548339881366e-05, + "loss": 0.0, + "step": 13852 + }, + { + "epoch": 3.2, + "learning_rate": 2.3320622448193517e-05, + "loss": 0.0, + "step": 13853 + }, + { + "epoch": 3.2, + "learning_rate": 2.331869655650566e-05, + "loss": 0.0, + "step": 13854 + }, + { + "epoch": 3.2, + "learning_rate": 2.331677066481781e-05, + "loss": 0.0, + "step": 13855 + }, + { + "epoch": 3.2, + "learning_rate": 2.331484477312996e-05, + "loss": 0.0, + "step": 13856 + }, + { + "epoch": 3.2, + "learning_rate": 2.331291888144211e-05, + "loss": 0.0001, + "step": 13857 + }, + { + "epoch": 3.2, + "learning_rate": 2.3310992989754257e-05, + "loss": 0.0, + "step": 13858 + }, + { + "epoch": 3.2, + "learning_rate": 2.3309067098066405e-05, + "loss": 0.0, + "step": 13859 + }, + { + "epoch": 3.2, + "learning_rate": 2.3307141206378556e-05, + "loss": 0.0, + "step": 13860 + }, + { + "epoch": 3.2, + "learning_rate": 2.3305215314690703e-05, + "loss": 0.0, + "step": 13861 + }, + { + "epoch": 3.2, + "learning_rate": 2.3303289423002854e-05, + "loss": 0.0, + "step": 13862 + }, + { + "epoch": 3.2, + "learning_rate": 2.3301363531314998e-05, + "loss": 0.0, + "step": 13863 + }, + { + "epoch": 3.2, + "learning_rate": 2.329943763962715e-05, + "loss": 0.0, + "step": 13864 + }, + { + "epoch": 3.2, + "learning_rate": 2.3297511747939297e-05, + "loss": 0.0, + "step": 13865 + }, + { + "epoch": 3.2, + "learning_rate": 2.3295585856251444e-05, + "loss": 0.0, + "step": 13866 + }, + { + "epoch": 3.2, + "learning_rate": 2.3293659964563595e-05, + "loss": 0.0, + "step": 13867 + }, + { + "epoch": 3.2, + "learning_rate": 2.3291734072875742e-05, + "loss": 0.0, + "step": 13868 + }, + { + "epoch": 3.21, + "learning_rate": 2.3289808181187893e-05, + "loss": 0.0, + "step": 13869 + }, + { + "epoch": 3.21, + "learning_rate": 2.328788228950004e-05, + "loss": 0.0, + "step": 13870 + }, + { + "epoch": 3.21, + "learning_rate": 2.3285956397812188e-05, + "loss": 0.0, + "step": 13871 + }, + { + "epoch": 3.21, + "learning_rate": 2.328403050612434e-05, + "loss": 0.0, + "step": 13872 + }, + { + "epoch": 3.21, + "learning_rate": 2.3282104614436483e-05, + "loss": 0.0, + "step": 13873 + }, + { + "epoch": 3.21, + "learning_rate": 2.3280178722748634e-05, + "loss": 0.0, + "step": 13874 + }, + { + "epoch": 3.21, + "learning_rate": 2.327825283106078e-05, + "loss": 0.0, + "step": 13875 + }, + { + "epoch": 3.21, + "learning_rate": 2.327632693937293e-05, + "loss": 0.0, + "step": 13876 + }, + { + "epoch": 3.21, + "learning_rate": 2.327440104768508e-05, + "loss": 0.0003, + "step": 13877 + }, + { + "epoch": 3.21, + "learning_rate": 2.3272475155997227e-05, + "loss": 0.0, + "step": 13878 + }, + { + "epoch": 3.21, + "learning_rate": 2.3270549264309378e-05, + "loss": 0.0, + "step": 13879 + }, + { + "epoch": 3.21, + "learning_rate": 2.3268623372621526e-05, + "loss": 0.0, + "step": 13880 + }, + { + "epoch": 3.21, + "learning_rate": 2.3266697480933673e-05, + "loss": 0.0, + "step": 13881 + }, + { + "epoch": 3.21, + "learning_rate": 2.326477158924582e-05, + "loss": 0.0, + "step": 13882 + }, + { + "epoch": 3.21, + "learning_rate": 2.3262845697557968e-05, + "loss": 0.0, + "step": 13883 + }, + { + "epoch": 3.21, + "learning_rate": 2.326091980587012e-05, + "loss": 0.0, + "step": 13884 + }, + { + "epoch": 3.21, + "learning_rate": 2.3258993914182266e-05, + "loss": 0.0, + "step": 13885 + }, + { + "epoch": 3.21, + "learning_rate": 2.3257068022494417e-05, + "loss": 0.0, + "step": 13886 + }, + { + "epoch": 3.21, + "learning_rate": 2.3255142130806565e-05, + "loss": 0.0, + "step": 13887 + }, + { + "epoch": 3.21, + "learning_rate": 2.3253216239118712e-05, + "loss": 0.0, + "step": 13888 + }, + { + "epoch": 3.21, + "learning_rate": 2.3251290347430863e-05, + "loss": 0.0002, + "step": 13889 + }, + { + "epoch": 3.21, + "learning_rate": 2.324936445574301e-05, + "loss": 0.0, + "step": 13890 + }, + { + "epoch": 3.21, + "learning_rate": 2.3247438564055158e-05, + "loss": 0.0, + "step": 13891 + }, + { + "epoch": 3.21, + "learning_rate": 2.3245512672367306e-05, + "loss": 0.0, + "step": 13892 + }, + { + "epoch": 3.21, + "learning_rate": 2.3243586780679456e-05, + "loss": 0.0006, + "step": 13893 + }, + { + "epoch": 3.21, + "learning_rate": 2.3241660888991604e-05, + "loss": 0.0, + "step": 13894 + }, + { + "epoch": 3.21, + "learning_rate": 2.323973499730375e-05, + "loss": 0.0004, + "step": 13895 + }, + { + "epoch": 3.21, + "learning_rate": 2.3237809105615902e-05, + "loss": 0.0, + "step": 13896 + }, + { + "epoch": 3.21, + "learning_rate": 2.323588321392805e-05, + "loss": 0.0, + "step": 13897 + }, + { + "epoch": 3.21, + "learning_rate": 2.32339573222402e-05, + "loss": 0.0, + "step": 13898 + }, + { + "epoch": 3.21, + "learning_rate": 2.3232031430552348e-05, + "loss": 0.0001, + "step": 13899 + }, + { + "epoch": 3.21, + "learning_rate": 2.3230105538864496e-05, + "loss": 0.0001, + "step": 13900 + }, + { + "epoch": 3.21, + "learning_rate": 2.3228179647176643e-05, + "loss": 0.0, + "step": 13901 + }, + { + "epoch": 3.21, + "learning_rate": 2.322625375548879e-05, + "loss": 0.0, + "step": 13902 + }, + { + "epoch": 3.21, + "learning_rate": 2.322432786380094e-05, + "loss": 0.0, + "step": 13903 + }, + { + "epoch": 3.21, + "learning_rate": 2.322240197211309e-05, + "loss": 0.0, + "step": 13904 + }, + { + "epoch": 3.21, + "learning_rate": 2.322047608042524e-05, + "loss": 0.0007, + "step": 13905 + }, + { + "epoch": 3.21, + "learning_rate": 2.3218550188737387e-05, + "loss": 0.0, + "step": 13906 + }, + { + "epoch": 3.21, + "learning_rate": 2.3216624297049535e-05, + "loss": 0.0, + "step": 13907 + }, + { + "epoch": 3.21, + "learning_rate": 2.3214698405361685e-05, + "loss": 0.0, + "step": 13908 + }, + { + "epoch": 3.21, + "learning_rate": 2.3212772513673833e-05, + "loss": 0.0, + "step": 13909 + }, + { + "epoch": 3.21, + "learning_rate": 2.321084662198598e-05, + "loss": 0.0, + "step": 13910 + }, + { + "epoch": 3.21, + "learning_rate": 2.3208920730298128e-05, + "loss": 0.0, + "step": 13911 + }, + { + "epoch": 3.22, + "learning_rate": 2.3206994838610275e-05, + "loss": 0.0, + "step": 13912 + }, + { + "epoch": 3.22, + "learning_rate": 2.3205068946922426e-05, + "loss": 0.0, + "step": 13913 + }, + { + "epoch": 3.22, + "learning_rate": 2.3203143055234574e-05, + "loss": 0.0004, + "step": 13914 + }, + { + "epoch": 3.22, + "learning_rate": 2.3201217163546725e-05, + "loss": 0.0, + "step": 13915 + }, + { + "epoch": 3.22, + "learning_rate": 2.3199291271858872e-05, + "loss": 0.0002, + "step": 13916 + }, + { + "epoch": 3.22, + "learning_rate": 2.319736538017102e-05, + "loss": 0.0001, + "step": 13917 + }, + { + "epoch": 3.22, + "learning_rate": 2.319543948848317e-05, + "loss": 0.0, + "step": 13918 + }, + { + "epoch": 3.22, + "learning_rate": 2.3193513596795315e-05, + "loss": 0.0, + "step": 13919 + }, + { + "epoch": 3.22, + "learning_rate": 2.3191587705107465e-05, + "loss": 0.0, + "step": 13920 + }, + { + "epoch": 3.22, + "learning_rate": 2.3189661813419613e-05, + "loss": 0.0001, + "step": 13921 + }, + { + "epoch": 3.22, + "learning_rate": 2.3187735921731764e-05, + "loss": 0.0001, + "step": 13922 + }, + { + "epoch": 3.22, + "learning_rate": 2.318581003004391e-05, + "loss": 0.0, + "step": 13923 + }, + { + "epoch": 3.22, + "learning_rate": 2.318388413835606e-05, + "loss": 0.0, + "step": 13924 + }, + { + "epoch": 3.22, + "learning_rate": 2.318195824666821e-05, + "loss": 0.0, + "step": 13925 + }, + { + "epoch": 3.22, + "learning_rate": 2.3180032354980357e-05, + "loss": 0.0003, + "step": 13926 + }, + { + "epoch": 3.22, + "learning_rate": 2.3178106463292508e-05, + "loss": 0.0, + "step": 13927 + }, + { + "epoch": 3.22, + "learning_rate": 2.3176180571604655e-05, + "loss": 0.0, + "step": 13928 + }, + { + "epoch": 3.22, + "learning_rate": 2.3174254679916803e-05, + "loss": 0.0, + "step": 13929 + }, + { + "epoch": 3.22, + "learning_rate": 2.317232878822895e-05, + "loss": 0.0, + "step": 13930 + }, + { + "epoch": 3.22, + "learning_rate": 2.3170402896541098e-05, + "loss": 0.0, + "step": 13931 + }, + { + "epoch": 3.22, + "learning_rate": 2.316847700485325e-05, + "loss": 0.0, + "step": 13932 + }, + { + "epoch": 3.22, + "learning_rate": 2.3166551113165396e-05, + "loss": 0.0, + "step": 13933 + }, + { + "epoch": 3.22, + "learning_rate": 2.3164625221477547e-05, + "loss": 0.0, + "step": 13934 + }, + { + "epoch": 3.22, + "learning_rate": 2.3162699329789694e-05, + "loss": 0.0, + "step": 13935 + }, + { + "epoch": 3.22, + "learning_rate": 2.3160773438101842e-05, + "loss": 0.0009, + "step": 13936 + }, + { + "epoch": 3.22, + "learning_rate": 2.3158847546413993e-05, + "loss": 0.0, + "step": 13937 + }, + { + "epoch": 3.22, + "learning_rate": 2.3156921654726137e-05, + "loss": 0.0, + "step": 13938 + }, + { + "epoch": 3.22, + "learning_rate": 2.3154995763038288e-05, + "loss": 0.0, + "step": 13939 + }, + { + "epoch": 3.22, + "learning_rate": 2.3153069871350435e-05, + "loss": 0.0, + "step": 13940 + }, + { + "epoch": 3.22, + "learning_rate": 2.3151143979662586e-05, + "loss": 0.0, + "step": 13941 + }, + { + "epoch": 3.22, + "learning_rate": 2.3149218087974734e-05, + "loss": 0.0, + "step": 13942 + }, + { + "epoch": 3.22, + "learning_rate": 2.314729219628688e-05, + "loss": 0.0, + "step": 13943 + }, + { + "epoch": 3.22, + "learning_rate": 2.3145366304599032e-05, + "loss": 0.0, + "step": 13944 + }, + { + "epoch": 3.22, + "learning_rate": 2.314344041291118e-05, + "loss": 0.0, + "step": 13945 + }, + { + "epoch": 3.22, + "learning_rate": 2.314151452122333e-05, + "loss": 0.0, + "step": 13946 + }, + { + "epoch": 3.22, + "learning_rate": 2.3139588629535474e-05, + "loss": 0.0, + "step": 13947 + }, + { + "epoch": 3.22, + "learning_rate": 2.3137662737847622e-05, + "loss": 0.0, + "step": 13948 + }, + { + "epoch": 3.22, + "learning_rate": 2.3135736846159773e-05, + "loss": 0.0, + "step": 13949 + }, + { + "epoch": 3.22, + "learning_rate": 2.313381095447192e-05, + "loss": 0.0, + "step": 13950 + }, + { + "epoch": 3.22, + "learning_rate": 2.313188506278407e-05, + "loss": 0.0, + "step": 13951 + }, + { + "epoch": 3.22, + "learning_rate": 2.312995917109622e-05, + "loss": 0.0, + "step": 13952 + }, + { + "epoch": 3.22, + "learning_rate": 2.3128033279408366e-05, + "loss": 0.0003, + "step": 13953 + }, + { + "epoch": 3.22, + "learning_rate": 2.3126107387720517e-05, + "loss": 0.0, + "step": 13954 + }, + { + "epoch": 3.23, + "learning_rate": 2.3124181496032664e-05, + "loss": 0.0, + "step": 13955 + }, + { + "epoch": 3.23, + "learning_rate": 2.3122255604344815e-05, + "loss": 0.0, + "step": 13956 + }, + { + "epoch": 3.23, + "learning_rate": 2.312032971265696e-05, + "loss": 0.0, + "step": 13957 + }, + { + "epoch": 3.23, + "learning_rate": 2.311840382096911e-05, + "loss": 0.0, + "step": 13958 + }, + { + "epoch": 3.23, + "learning_rate": 2.3116477929281258e-05, + "loss": 0.0, + "step": 13959 + }, + { + "epoch": 3.23, + "learning_rate": 2.3114552037593405e-05, + "loss": 0.0, + "step": 13960 + }, + { + "epoch": 3.23, + "learning_rate": 2.3112626145905556e-05, + "loss": 0.0001, + "step": 13961 + }, + { + "epoch": 3.23, + "learning_rate": 2.3110700254217703e-05, + "loss": 0.0, + "step": 13962 + }, + { + "epoch": 3.23, + "learning_rate": 2.3108774362529854e-05, + "loss": 0.0, + "step": 13963 + }, + { + "epoch": 3.23, + "learning_rate": 2.3106848470842002e-05, + "loss": 0.0, + "step": 13964 + }, + { + "epoch": 3.23, + "learning_rate": 2.310492257915415e-05, + "loss": 0.0, + "step": 13965 + }, + { + "epoch": 3.23, + "learning_rate": 2.3102996687466297e-05, + "loss": 0.0001, + "step": 13966 + }, + { + "epoch": 3.23, + "learning_rate": 2.3101070795778444e-05, + "loss": 0.0, + "step": 13967 + }, + { + "epoch": 3.23, + "learning_rate": 2.3099144904090595e-05, + "loss": 0.0, + "step": 13968 + }, + { + "epoch": 3.23, + "learning_rate": 2.3097219012402743e-05, + "loss": 0.0, + "step": 13969 + }, + { + "epoch": 3.23, + "learning_rate": 2.3095293120714893e-05, + "loss": 0.0006, + "step": 13970 + }, + { + "epoch": 3.23, + "learning_rate": 2.309336722902704e-05, + "loss": 0.0002, + "step": 13971 + }, + { + "epoch": 3.23, + "learning_rate": 2.309144133733919e-05, + "loss": 0.0, + "step": 13972 + }, + { + "epoch": 3.23, + "learning_rate": 2.308951544565134e-05, + "loss": 0.0, + "step": 13973 + }, + { + "epoch": 3.23, + "learning_rate": 2.3087589553963487e-05, + "loss": 0.0, + "step": 13974 + }, + { + "epoch": 3.23, + "learning_rate": 2.3085663662275634e-05, + "loss": 0.0, + "step": 13975 + }, + { + "epoch": 3.23, + "learning_rate": 2.3083737770587782e-05, + "loss": 0.0, + "step": 13976 + }, + { + "epoch": 3.23, + "learning_rate": 2.3081811878899933e-05, + "loss": 0.0, + "step": 13977 + }, + { + "epoch": 3.23, + "learning_rate": 2.307988598721208e-05, + "loss": 0.0, + "step": 13978 + }, + { + "epoch": 3.23, + "learning_rate": 2.3077960095524228e-05, + "loss": 0.0, + "step": 13979 + }, + { + "epoch": 3.23, + "learning_rate": 2.307603420383638e-05, + "loss": 0.0001, + "step": 13980 + }, + { + "epoch": 3.23, + "learning_rate": 2.3074108312148526e-05, + "loss": 0.0, + "step": 13981 + }, + { + "epoch": 3.23, + "learning_rate": 2.3072182420460677e-05, + "loss": 0.0001, + "step": 13982 + }, + { + "epoch": 3.23, + "learning_rate": 2.3070256528772824e-05, + "loss": 0.0, + "step": 13983 + }, + { + "epoch": 3.23, + "learning_rate": 2.3068330637084972e-05, + "loss": 0.0001, + "step": 13984 + }, + { + "epoch": 3.23, + "learning_rate": 2.306640474539712e-05, + "loss": 0.0, + "step": 13985 + }, + { + "epoch": 3.23, + "learning_rate": 2.3064478853709267e-05, + "loss": 0.0, + "step": 13986 + }, + { + "epoch": 3.23, + "learning_rate": 2.3062552962021418e-05, + "loss": 0.0, + "step": 13987 + }, + { + "epoch": 3.23, + "learning_rate": 2.3060627070333565e-05, + "loss": 0.0, + "step": 13988 + }, + { + "epoch": 3.23, + "learning_rate": 2.3058701178645713e-05, + "loss": 0.0002, + "step": 13989 + }, + { + "epoch": 3.23, + "learning_rate": 2.3056775286957863e-05, + "loss": 0.0001, + "step": 13990 + }, + { + "epoch": 3.23, + "learning_rate": 2.305484939527001e-05, + "loss": 0.0, + "step": 13991 + }, + { + "epoch": 3.23, + "learning_rate": 2.3052923503582162e-05, + "loss": 0.0, + "step": 13992 + }, + { + "epoch": 3.23, + "learning_rate": 2.305099761189431e-05, + "loss": 0.0, + "step": 13993 + }, + { + "epoch": 3.23, + "learning_rate": 2.3049071720206457e-05, + "loss": 0.0, + "step": 13994 + }, + { + "epoch": 3.23, + "learning_rate": 2.3047145828518604e-05, + "loss": 0.0, + "step": 13995 + }, + { + "epoch": 3.23, + "learning_rate": 2.304521993683075e-05, + "loss": 0.0, + "step": 13996 + }, + { + "epoch": 3.23, + "learning_rate": 2.3043294045142902e-05, + "loss": 0.0, + "step": 13997 + }, + { + "epoch": 3.24, + "learning_rate": 2.304136815345505e-05, + "loss": 0.0, + "step": 13998 + }, + { + "epoch": 3.24, + "learning_rate": 2.30394422617672e-05, + "loss": 0.0, + "step": 13999 + }, + { + "epoch": 3.24, + "learning_rate": 2.303751637007935e-05, + "loss": 0.0, + "step": 14000 + }, + { + "epoch": 3.24, + "learning_rate": 2.3035590478391496e-05, + "loss": 0.0, + "step": 14001 + }, + { + "epoch": 3.24, + "learning_rate": 2.3033664586703647e-05, + "loss": 0.0001, + "step": 14002 + }, + { + "epoch": 3.24, + "learning_rate": 2.303173869501579e-05, + "loss": 0.0, + "step": 14003 + }, + { + "epoch": 3.24, + "learning_rate": 2.302981280332794e-05, + "loss": 0.0, + "step": 14004 + }, + { + "epoch": 3.24, + "learning_rate": 2.302788691164009e-05, + "loss": 0.0, + "step": 14005 + }, + { + "epoch": 3.24, + "learning_rate": 2.302596101995224e-05, + "loss": 0.0, + "step": 14006 + }, + { + "epoch": 3.24, + "learning_rate": 2.3024035128264387e-05, + "loss": 0.0003, + "step": 14007 + }, + { + "epoch": 3.24, + "learning_rate": 2.3022109236576535e-05, + "loss": 0.0, + "step": 14008 + }, + { + "epoch": 3.24, + "learning_rate": 2.3020183344888686e-05, + "loss": 0.0001, + "step": 14009 + }, + { + "epoch": 3.24, + "learning_rate": 2.3018257453200833e-05, + "loss": 0.0002, + "step": 14010 + }, + { + "epoch": 3.24, + "learning_rate": 2.3016331561512984e-05, + "loss": 0.0, + "step": 14011 + }, + { + "epoch": 3.24, + "learning_rate": 2.301440566982513e-05, + "loss": 0.0001, + "step": 14012 + }, + { + "epoch": 3.24, + "learning_rate": 2.301247977813728e-05, + "loss": 0.0001, + "step": 14013 + }, + { + "epoch": 3.24, + "learning_rate": 2.3010553886449427e-05, + "loss": 0.0, + "step": 14014 + }, + { + "epoch": 3.24, + "learning_rate": 2.3008627994761574e-05, + "loss": 0.0001, + "step": 14015 + }, + { + "epoch": 3.24, + "learning_rate": 2.3006702103073725e-05, + "loss": 0.0, + "step": 14016 + }, + { + "epoch": 3.24, + "learning_rate": 2.3004776211385872e-05, + "loss": 0.0, + "step": 14017 + }, + { + "epoch": 3.24, + "learning_rate": 2.3002850319698023e-05, + "loss": 0.0, + "step": 14018 + }, + { + "epoch": 3.24, + "learning_rate": 2.300092442801017e-05, + "loss": 0.0001, + "step": 14019 + }, + { + "epoch": 3.24, + "learning_rate": 2.2998998536322318e-05, + "loss": 0.0, + "step": 14020 + }, + { + "epoch": 3.24, + "learning_rate": 2.299707264463447e-05, + "loss": 0.0001, + "step": 14021 + }, + { + "epoch": 3.24, + "learning_rate": 2.2995146752946613e-05, + "loss": 0.0, + "step": 14022 + }, + { + "epoch": 3.24, + "learning_rate": 2.2993220861258764e-05, + "loss": 0.0, + "step": 14023 + }, + { + "epoch": 3.24, + "learning_rate": 2.299129496957091e-05, + "loss": 0.0, + "step": 14024 + }, + { + "epoch": 3.24, + "learning_rate": 2.298936907788306e-05, + "loss": 0.0, + "step": 14025 + }, + { + "epoch": 3.24, + "learning_rate": 2.298744318619521e-05, + "loss": 0.0001, + "step": 14026 + }, + { + "epoch": 3.24, + "learning_rate": 2.2985517294507357e-05, + "loss": 0.0, + "step": 14027 + }, + { + "epoch": 3.24, + "learning_rate": 2.2983591402819508e-05, + "loss": 0.0, + "step": 14028 + }, + { + "epoch": 3.24, + "learning_rate": 2.2981665511131656e-05, + "loss": 0.0004, + "step": 14029 + }, + { + "epoch": 3.24, + "learning_rate": 2.2979739619443803e-05, + "loss": 0.0, + "step": 14030 + }, + { + "epoch": 3.24, + "learning_rate": 2.297781372775595e-05, + "loss": 0.0, + "step": 14031 + }, + { + "epoch": 3.24, + "learning_rate": 2.2975887836068098e-05, + "loss": 0.0001, + "step": 14032 + }, + { + "epoch": 3.24, + "learning_rate": 2.297396194438025e-05, + "loss": 0.0, + "step": 14033 + }, + { + "epoch": 3.24, + "learning_rate": 2.2972036052692396e-05, + "loss": 0.0, + "step": 14034 + }, + { + "epoch": 3.24, + "learning_rate": 2.2970110161004547e-05, + "loss": 0.0, + "step": 14035 + }, + { + "epoch": 3.24, + "learning_rate": 2.2968184269316695e-05, + "loss": 0.0, + "step": 14036 + }, + { + "epoch": 3.24, + "learning_rate": 2.2966258377628842e-05, + "loss": 0.0001, + "step": 14037 + }, + { + "epoch": 3.24, + "learning_rate": 2.2964332485940993e-05, + "loss": 0.0, + "step": 14038 + }, + { + "epoch": 3.24, + "learning_rate": 2.296240659425314e-05, + "loss": 0.0, + "step": 14039 + }, + { + "epoch": 3.24, + "learning_rate": 2.296048070256529e-05, + "loss": 0.0, + "step": 14040 + }, + { + "epoch": 3.24, + "learning_rate": 2.2958554810877436e-05, + "loss": 0.0, + "step": 14041 + }, + { + "epoch": 3.25, + "learning_rate": 2.2956628919189586e-05, + "loss": 0.0, + "step": 14042 + }, + { + "epoch": 3.25, + "learning_rate": 2.2954703027501734e-05, + "loss": 0.0, + "step": 14043 + }, + { + "epoch": 3.25, + "learning_rate": 2.295277713581388e-05, + "loss": 0.0009, + "step": 14044 + }, + { + "epoch": 3.25, + "learning_rate": 2.2950851244126032e-05, + "loss": 0.0, + "step": 14045 + }, + { + "epoch": 3.25, + "learning_rate": 2.294892535243818e-05, + "loss": 0.0, + "step": 14046 + }, + { + "epoch": 3.25, + "learning_rate": 2.294699946075033e-05, + "loss": 0.0, + "step": 14047 + }, + { + "epoch": 3.25, + "learning_rate": 2.2945073569062478e-05, + "loss": 0.0, + "step": 14048 + }, + { + "epoch": 3.25, + "learning_rate": 2.2943147677374626e-05, + "loss": 0.0, + "step": 14049 + }, + { + "epoch": 3.25, + "learning_rate": 2.2941221785686773e-05, + "loss": 0.0, + "step": 14050 + }, + { + "epoch": 3.25, + "learning_rate": 2.293929589399892e-05, + "loss": 0.0003, + "step": 14051 + }, + { + "epoch": 3.25, + "learning_rate": 2.293737000231107e-05, + "loss": 0.0, + "step": 14052 + }, + { + "epoch": 3.25, + "learning_rate": 2.293544411062322e-05, + "loss": 0.0002, + "step": 14053 + }, + { + "epoch": 3.25, + "learning_rate": 2.293351821893537e-05, + "loss": 0.0, + "step": 14054 + }, + { + "epoch": 3.25, + "learning_rate": 2.2931592327247517e-05, + "loss": 0.0, + "step": 14055 + }, + { + "epoch": 3.25, + "learning_rate": 2.2929666435559665e-05, + "loss": 0.0001, + "step": 14056 + }, + { + "epoch": 3.25, + "learning_rate": 2.2927740543871816e-05, + "loss": 0.0001, + "step": 14057 + }, + { + "epoch": 3.25, + "learning_rate": 2.2925814652183963e-05, + "loss": 0.0, + "step": 14058 + }, + { + "epoch": 3.25, + "learning_rate": 2.292388876049611e-05, + "loss": 0.0, + "step": 14059 + }, + { + "epoch": 3.25, + "learning_rate": 2.2921962868808258e-05, + "loss": 0.0, + "step": 14060 + }, + { + "epoch": 3.25, + "learning_rate": 2.2920036977120405e-05, + "loss": 0.0, + "step": 14061 + }, + { + "epoch": 3.25, + "learning_rate": 2.2918111085432556e-05, + "loss": 0.0, + "step": 14062 + }, + { + "epoch": 3.25, + "learning_rate": 2.2916185193744704e-05, + "loss": 0.0, + "step": 14063 + }, + { + "epoch": 3.25, + "learning_rate": 2.2914259302056855e-05, + "loss": 0.0, + "step": 14064 + }, + { + "epoch": 3.25, + "learning_rate": 2.2912333410369002e-05, + "loss": 0.0, + "step": 14065 + }, + { + "epoch": 3.25, + "learning_rate": 2.291040751868115e-05, + "loss": 0.0, + "step": 14066 + }, + { + "epoch": 3.25, + "learning_rate": 2.29084816269933e-05, + "loss": 0.0001, + "step": 14067 + }, + { + "epoch": 3.25, + "learning_rate": 2.2906555735305448e-05, + "loss": 0.0, + "step": 14068 + }, + { + "epoch": 3.25, + "learning_rate": 2.2904629843617595e-05, + "loss": 0.0001, + "step": 14069 + }, + { + "epoch": 3.25, + "learning_rate": 2.2902703951929743e-05, + "loss": 0.0, + "step": 14070 + }, + { + "epoch": 3.25, + "learning_rate": 2.2900778060241894e-05, + "loss": 0.0001, + "step": 14071 + }, + { + "epoch": 3.25, + "learning_rate": 2.289885216855404e-05, + "loss": 0.0, + "step": 14072 + }, + { + "epoch": 3.25, + "learning_rate": 2.289692627686619e-05, + "loss": 0.0, + "step": 14073 + }, + { + "epoch": 3.25, + "learning_rate": 2.289500038517834e-05, + "loss": 0.0, + "step": 14074 + }, + { + "epoch": 3.25, + "learning_rate": 2.2893074493490487e-05, + "loss": 0.0, + "step": 14075 + }, + { + "epoch": 3.25, + "learning_rate": 2.2891148601802638e-05, + "loss": 0.0, + "step": 14076 + }, + { + "epoch": 3.25, + "learning_rate": 2.2889222710114785e-05, + "loss": 0.0, + "step": 14077 + }, + { + "epoch": 3.25, + "learning_rate": 2.2887296818426933e-05, + "loss": 0.0, + "step": 14078 + }, + { + "epoch": 3.25, + "learning_rate": 2.288537092673908e-05, + "loss": 0.0001, + "step": 14079 + }, + { + "epoch": 3.25, + "learning_rate": 2.2883445035051228e-05, + "loss": 0.0, + "step": 14080 + }, + { + "epoch": 3.25, + "learning_rate": 2.288151914336338e-05, + "loss": 0.0, + "step": 14081 + }, + { + "epoch": 3.25, + "learning_rate": 2.2879593251675526e-05, + "loss": 0.0, + "step": 14082 + }, + { + "epoch": 3.25, + "learning_rate": 2.2877667359987677e-05, + "loss": 0.0, + "step": 14083 + }, + { + "epoch": 3.25, + "learning_rate": 2.2875741468299825e-05, + "loss": 0.0, + "step": 14084 + }, + { + "epoch": 3.26, + "learning_rate": 2.2873815576611972e-05, + "loss": 0.0, + "step": 14085 + }, + { + "epoch": 3.26, + "learning_rate": 2.2871889684924123e-05, + "loss": 0.0, + "step": 14086 + }, + { + "epoch": 3.26, + "learning_rate": 2.2869963793236267e-05, + "loss": 0.0, + "step": 14087 + }, + { + "epoch": 3.26, + "learning_rate": 2.2868037901548418e-05, + "loss": 0.0, + "step": 14088 + }, + { + "epoch": 3.26, + "learning_rate": 2.2866112009860565e-05, + "loss": 0.0, + "step": 14089 + }, + { + "epoch": 3.26, + "learning_rate": 2.2864186118172716e-05, + "loss": 0.0, + "step": 14090 + }, + { + "epoch": 3.26, + "learning_rate": 2.2862260226484864e-05, + "loss": 0.0, + "step": 14091 + }, + { + "epoch": 3.26, + "learning_rate": 2.286033433479701e-05, + "loss": 0.0004, + "step": 14092 + }, + { + "epoch": 3.26, + "learning_rate": 2.2858408443109162e-05, + "loss": 0.0, + "step": 14093 + }, + { + "epoch": 3.26, + "learning_rate": 2.285648255142131e-05, + "loss": 0.0, + "step": 14094 + }, + { + "epoch": 3.26, + "learning_rate": 2.285455665973346e-05, + "loss": 0.0, + "step": 14095 + }, + { + "epoch": 3.26, + "learning_rate": 2.2852630768045608e-05, + "loss": 0.0, + "step": 14096 + }, + { + "epoch": 3.26, + "learning_rate": 2.2850704876357752e-05, + "loss": 0.0, + "step": 14097 + }, + { + "epoch": 3.26, + "learning_rate": 2.2848778984669903e-05, + "loss": 0.0, + "step": 14098 + }, + { + "epoch": 3.26, + "learning_rate": 2.284685309298205e-05, + "loss": 0.0, + "step": 14099 + }, + { + "epoch": 3.26, + "learning_rate": 2.28449272012942e-05, + "loss": 0.0, + "step": 14100 + }, + { + "epoch": 3.26, + "learning_rate": 2.284300130960635e-05, + "loss": 0.0, + "step": 14101 + }, + { + "epoch": 3.26, + "learning_rate": 2.2841075417918496e-05, + "loss": 0.0, + "step": 14102 + }, + { + "epoch": 3.26, + "learning_rate": 2.2839149526230647e-05, + "loss": 0.0, + "step": 14103 + }, + { + "epoch": 3.26, + "learning_rate": 2.2837223634542794e-05, + "loss": 0.0, + "step": 14104 + }, + { + "epoch": 3.26, + "learning_rate": 2.2835297742854945e-05, + "loss": 0.0, + "step": 14105 + }, + { + "epoch": 3.26, + "learning_rate": 2.283337185116709e-05, + "loss": 0.0, + "step": 14106 + }, + { + "epoch": 3.26, + "learning_rate": 2.283144595947924e-05, + "loss": 0.0006, + "step": 14107 + }, + { + "epoch": 3.26, + "learning_rate": 2.2829520067791388e-05, + "loss": 0.0, + "step": 14108 + }, + { + "epoch": 3.26, + "learning_rate": 2.2827594176103535e-05, + "loss": 0.0, + "step": 14109 + }, + { + "epoch": 3.26, + "learning_rate": 2.2825668284415686e-05, + "loss": 0.0, + "step": 14110 + }, + { + "epoch": 3.26, + "learning_rate": 2.2823742392727834e-05, + "loss": 0.0, + "step": 14111 + }, + { + "epoch": 3.26, + "learning_rate": 2.2821816501039984e-05, + "loss": 0.0, + "step": 14112 + }, + { + "epoch": 3.26, + "learning_rate": 2.2819890609352132e-05, + "loss": 0.0, + "step": 14113 + }, + { + "epoch": 3.26, + "learning_rate": 2.281796471766428e-05, + "loss": 0.0, + "step": 14114 + }, + { + "epoch": 3.26, + "learning_rate": 2.2816038825976427e-05, + "loss": 0.0, + "step": 14115 + }, + { + "epoch": 3.26, + "learning_rate": 2.2814112934288574e-05, + "loss": 0.0, + "step": 14116 + }, + { + "epoch": 3.26, + "learning_rate": 2.2812187042600725e-05, + "loss": 0.0, + "step": 14117 + }, + { + "epoch": 3.26, + "learning_rate": 2.2810261150912873e-05, + "loss": 0.0, + "step": 14118 + }, + { + "epoch": 3.26, + "learning_rate": 2.2808335259225024e-05, + "loss": 0.0001, + "step": 14119 + }, + { + "epoch": 3.26, + "learning_rate": 2.280640936753717e-05, + "loss": 0.0, + "step": 14120 + }, + { + "epoch": 3.26, + "learning_rate": 2.280448347584932e-05, + "loss": 0.0, + "step": 14121 + }, + { + "epoch": 3.26, + "learning_rate": 2.280255758416147e-05, + "loss": 0.0, + "step": 14122 + }, + { + "epoch": 3.26, + "learning_rate": 2.2800631692473617e-05, + "loss": 0.0, + "step": 14123 + }, + { + "epoch": 3.26, + "learning_rate": 2.2798705800785768e-05, + "loss": 0.0, + "step": 14124 + }, + { + "epoch": 3.26, + "learning_rate": 2.2796779909097912e-05, + "loss": 0.0, + "step": 14125 + }, + { + "epoch": 3.26, + "learning_rate": 2.2794854017410063e-05, + "loss": 0.0, + "step": 14126 + }, + { + "epoch": 3.26, + "learning_rate": 2.279292812572221e-05, + "loss": 0.0, + "step": 14127 + }, + { + "epoch": 3.27, + "learning_rate": 2.2791002234034358e-05, + "loss": 0.0, + "step": 14128 + }, + { + "epoch": 3.27, + "learning_rate": 2.278907634234651e-05, + "loss": 0.0001, + "step": 14129 + }, + { + "epoch": 3.27, + "learning_rate": 2.2787150450658656e-05, + "loss": 0.0, + "step": 14130 + }, + { + "epoch": 3.27, + "learning_rate": 2.2785224558970807e-05, + "loss": 0.0001, + "step": 14131 + }, + { + "epoch": 3.27, + "learning_rate": 2.2783298667282954e-05, + "loss": 0.0, + "step": 14132 + }, + { + "epoch": 3.27, + "learning_rate": 2.2781372775595102e-05, + "loss": 0.0001, + "step": 14133 + }, + { + "epoch": 3.27, + "learning_rate": 2.277944688390725e-05, + "loss": 0.0, + "step": 14134 + }, + { + "epoch": 3.27, + "learning_rate": 2.2777520992219397e-05, + "loss": 0.0, + "step": 14135 + }, + { + "epoch": 3.27, + "learning_rate": 2.2775595100531548e-05, + "loss": 0.0001, + "step": 14136 + }, + { + "epoch": 3.27, + "learning_rate": 2.2773669208843695e-05, + "loss": 0.0, + "step": 14137 + }, + { + "epoch": 3.27, + "learning_rate": 2.2771743317155843e-05, + "loss": 0.0, + "step": 14138 + }, + { + "epoch": 3.27, + "learning_rate": 2.2769817425467993e-05, + "loss": 0.0, + "step": 14139 + }, + { + "epoch": 3.27, + "learning_rate": 2.276789153378014e-05, + "loss": 0.0, + "step": 14140 + }, + { + "epoch": 3.27, + "learning_rate": 2.2765965642092292e-05, + "loss": 0.0, + "step": 14141 + }, + { + "epoch": 3.27, + "learning_rate": 2.276403975040444e-05, + "loss": 0.0, + "step": 14142 + }, + { + "epoch": 3.27, + "learning_rate": 2.2762113858716587e-05, + "loss": 0.0, + "step": 14143 + }, + { + "epoch": 3.27, + "learning_rate": 2.2760187967028734e-05, + "loss": 0.0003, + "step": 14144 + }, + { + "epoch": 3.27, + "learning_rate": 2.275826207534088e-05, + "loss": 0.0, + "step": 14145 + }, + { + "epoch": 3.27, + "learning_rate": 2.2756336183653033e-05, + "loss": 0.0, + "step": 14146 + }, + { + "epoch": 3.27, + "learning_rate": 2.275441029196518e-05, + "loss": 0.0, + "step": 14147 + }, + { + "epoch": 3.27, + "learning_rate": 2.275248440027733e-05, + "loss": 0.0, + "step": 14148 + }, + { + "epoch": 3.27, + "learning_rate": 2.275055850858948e-05, + "loss": 0.0, + "step": 14149 + }, + { + "epoch": 3.27, + "learning_rate": 2.2748632616901626e-05, + "loss": 0.0, + "step": 14150 + }, + { + "epoch": 3.27, + "learning_rate": 2.2746706725213777e-05, + "loss": 0.0001, + "step": 14151 + }, + { + "epoch": 3.27, + "learning_rate": 2.2744780833525924e-05, + "loss": 0.0, + "step": 14152 + }, + { + "epoch": 3.27, + "learning_rate": 2.274285494183807e-05, + "loss": 0.0, + "step": 14153 + }, + { + "epoch": 3.27, + "learning_rate": 2.274092905015022e-05, + "loss": 0.016, + "step": 14154 + }, + { + "epoch": 3.27, + "learning_rate": 2.273900315846237e-05, + "loss": 0.0, + "step": 14155 + }, + { + "epoch": 3.27, + "learning_rate": 2.2737077266774517e-05, + "loss": 0.0, + "step": 14156 + }, + { + "epoch": 3.27, + "learning_rate": 2.2735151375086665e-05, + "loss": 0.0, + "step": 14157 + }, + { + "epoch": 3.27, + "learning_rate": 2.2733225483398816e-05, + "loss": 0.0, + "step": 14158 + }, + { + "epoch": 3.27, + "learning_rate": 2.2731299591710963e-05, + "loss": 0.0, + "step": 14159 + }, + { + "epoch": 3.27, + "learning_rate": 2.2729373700023114e-05, + "loss": 0.0, + "step": 14160 + }, + { + "epoch": 3.27, + "learning_rate": 2.272744780833526e-05, + "loss": 0.0, + "step": 14161 + }, + { + "epoch": 3.27, + "learning_rate": 2.272552191664741e-05, + "loss": 0.0, + "step": 14162 + }, + { + "epoch": 3.27, + "learning_rate": 2.2723596024959557e-05, + "loss": 0.0006, + "step": 14163 + }, + { + "epoch": 3.27, + "learning_rate": 2.2721670133271704e-05, + "loss": 0.0, + "step": 14164 + }, + { + "epoch": 3.27, + "learning_rate": 2.2719744241583855e-05, + "loss": 0.0003, + "step": 14165 + }, + { + "epoch": 3.27, + "learning_rate": 2.2717818349896002e-05, + "loss": 0.0, + "step": 14166 + }, + { + "epoch": 3.27, + "learning_rate": 2.2715892458208153e-05, + "loss": 0.0001, + "step": 14167 + }, + { + "epoch": 3.27, + "learning_rate": 2.27139665665203e-05, + "loss": 0.0, + "step": 14168 + }, + { + "epoch": 3.27, + "learning_rate": 2.2712040674832448e-05, + "loss": 0.0, + "step": 14169 + }, + { + "epoch": 3.27, + "learning_rate": 2.27101147831446e-05, + "loss": 0.0, + "step": 14170 + }, + { + "epoch": 3.28, + "learning_rate": 2.2708188891456743e-05, + "loss": 0.0, + "step": 14171 + }, + { + "epoch": 3.28, + "learning_rate": 2.2706262999768894e-05, + "loss": 0.0, + "step": 14172 + }, + { + "epoch": 3.28, + "learning_rate": 2.270433710808104e-05, + "loss": 0.0, + "step": 14173 + }, + { + "epoch": 3.28, + "learning_rate": 2.270241121639319e-05, + "loss": 0.0, + "step": 14174 + }, + { + "epoch": 3.28, + "learning_rate": 2.270048532470534e-05, + "loss": 0.0, + "step": 14175 + }, + { + "epoch": 3.28, + "learning_rate": 2.2698559433017487e-05, + "loss": 0.0, + "step": 14176 + }, + { + "epoch": 3.28, + "learning_rate": 2.2696633541329638e-05, + "loss": 0.0, + "step": 14177 + }, + { + "epoch": 3.28, + "learning_rate": 2.2694707649641786e-05, + "loss": 0.0, + "step": 14178 + }, + { + "epoch": 3.28, + "learning_rate": 2.2692781757953933e-05, + "loss": 0.0, + "step": 14179 + }, + { + "epoch": 3.28, + "learning_rate": 2.2690855866266084e-05, + "loss": 0.0, + "step": 14180 + }, + { + "epoch": 3.28, + "learning_rate": 2.2688929974578228e-05, + "loss": 0.0, + "step": 14181 + }, + { + "epoch": 3.28, + "learning_rate": 2.268700408289038e-05, + "loss": 0.0, + "step": 14182 + }, + { + "epoch": 3.28, + "learning_rate": 2.2685078191202526e-05, + "loss": 0.0, + "step": 14183 + }, + { + "epoch": 3.28, + "learning_rate": 2.2683152299514677e-05, + "loss": 0.0, + "step": 14184 + }, + { + "epoch": 3.28, + "learning_rate": 2.2681226407826825e-05, + "loss": 0.0, + "step": 14185 + }, + { + "epoch": 3.28, + "learning_rate": 2.2679300516138972e-05, + "loss": 0.0, + "step": 14186 + }, + { + "epoch": 3.28, + "learning_rate": 2.2677374624451123e-05, + "loss": 0.0002, + "step": 14187 + }, + { + "epoch": 3.28, + "learning_rate": 2.267544873276327e-05, + "loss": 0.0, + "step": 14188 + }, + { + "epoch": 3.28, + "learning_rate": 2.267352284107542e-05, + "loss": 0.0, + "step": 14189 + }, + { + "epoch": 3.28, + "learning_rate": 2.2671596949387566e-05, + "loss": 0.0, + "step": 14190 + }, + { + "epoch": 3.28, + "learning_rate": 2.2669671057699716e-05, + "loss": 0.0, + "step": 14191 + }, + { + "epoch": 3.28, + "learning_rate": 2.2667745166011864e-05, + "loss": 0.0, + "step": 14192 + }, + { + "epoch": 3.28, + "learning_rate": 2.266581927432401e-05, + "loss": 0.0001, + "step": 14193 + }, + { + "epoch": 3.28, + "learning_rate": 2.2663893382636162e-05, + "loss": 0.0, + "step": 14194 + }, + { + "epoch": 3.28, + "learning_rate": 2.266196749094831e-05, + "loss": 0.0, + "step": 14195 + }, + { + "epoch": 3.28, + "learning_rate": 2.266004159926046e-05, + "loss": 0.0, + "step": 14196 + }, + { + "epoch": 3.28, + "learning_rate": 2.2658115707572608e-05, + "loss": 0.0, + "step": 14197 + }, + { + "epoch": 3.28, + "learning_rate": 2.2656189815884756e-05, + "loss": 0.0, + "step": 14198 + }, + { + "epoch": 3.28, + "learning_rate": 2.2654263924196903e-05, + "loss": 0.0, + "step": 14199 + }, + { + "epoch": 3.28, + "learning_rate": 2.265233803250905e-05, + "loss": 0.0, + "step": 14200 + }, + { + "epoch": 3.28, + "learning_rate": 2.26504121408212e-05, + "loss": 0.0, + "step": 14201 + }, + { + "epoch": 3.28, + "learning_rate": 2.264848624913335e-05, + "loss": 0.0, + "step": 14202 + }, + { + "epoch": 3.28, + "learning_rate": 2.26465603574455e-05, + "loss": 0.0, + "step": 14203 + }, + { + "epoch": 3.28, + "learning_rate": 2.2644634465757647e-05, + "loss": 0.0, + "step": 14204 + }, + { + "epoch": 3.28, + "learning_rate": 2.2642708574069795e-05, + "loss": 0.0, + "step": 14205 + }, + { + "epoch": 3.28, + "learning_rate": 2.2640782682381946e-05, + "loss": 0.0, + "step": 14206 + }, + { + "epoch": 3.28, + "learning_rate": 2.2638856790694093e-05, + "loss": 0.0, + "step": 14207 + }, + { + "epoch": 3.28, + "learning_rate": 2.2636930899006244e-05, + "loss": 0.0004, + "step": 14208 + }, + { + "epoch": 3.28, + "learning_rate": 2.2635005007318388e-05, + "loss": 0.0, + "step": 14209 + }, + { + "epoch": 3.28, + "learning_rate": 2.2633079115630535e-05, + "loss": 0.0, + "step": 14210 + }, + { + "epoch": 3.28, + "learning_rate": 2.2631153223942686e-05, + "loss": 0.0002, + "step": 14211 + }, + { + "epoch": 3.28, + "learning_rate": 2.2629227332254834e-05, + "loss": 0.0, + "step": 14212 + }, + { + "epoch": 3.28, + "learning_rate": 2.2627301440566985e-05, + "loss": 0.0, + "step": 14213 + }, + { + "epoch": 3.28, + "learning_rate": 2.2625375548879132e-05, + "loss": 0.0, + "step": 14214 + }, + { + "epoch": 3.29, + "learning_rate": 2.262344965719128e-05, + "loss": 0.0, + "step": 14215 + }, + { + "epoch": 3.29, + "learning_rate": 2.262152376550343e-05, + "loss": 0.0, + "step": 14216 + }, + { + "epoch": 3.29, + "learning_rate": 2.2619597873815578e-05, + "loss": 0.0, + "step": 14217 + }, + { + "epoch": 3.29, + "learning_rate": 2.2617671982127725e-05, + "loss": 0.0001, + "step": 14218 + }, + { + "epoch": 3.29, + "learning_rate": 2.2615746090439873e-05, + "loss": 0.0097, + "step": 14219 + }, + { + "epoch": 3.29, + "learning_rate": 2.2613820198752024e-05, + "loss": 0.0, + "step": 14220 + }, + { + "epoch": 3.29, + "learning_rate": 2.261189430706417e-05, + "loss": 0.0, + "step": 14221 + }, + { + "epoch": 3.29, + "learning_rate": 2.260996841537632e-05, + "loss": 0.0, + "step": 14222 + }, + { + "epoch": 3.29, + "learning_rate": 2.260804252368847e-05, + "loss": 0.0, + "step": 14223 + }, + { + "epoch": 3.29, + "learning_rate": 2.2606116632000617e-05, + "loss": 0.0, + "step": 14224 + }, + { + "epoch": 3.29, + "learning_rate": 2.2604190740312768e-05, + "loss": 0.0, + "step": 14225 + }, + { + "epoch": 3.29, + "learning_rate": 2.2602264848624915e-05, + "loss": 0.0, + "step": 14226 + }, + { + "epoch": 3.29, + "learning_rate": 2.2600338956937063e-05, + "loss": 0.0, + "step": 14227 + }, + { + "epoch": 3.29, + "learning_rate": 2.259841306524921e-05, + "loss": 0.0, + "step": 14228 + }, + { + "epoch": 3.29, + "learning_rate": 2.2596487173561358e-05, + "loss": 0.0001, + "step": 14229 + }, + { + "epoch": 3.29, + "learning_rate": 2.259456128187351e-05, + "loss": 0.0, + "step": 14230 + }, + { + "epoch": 3.29, + "learning_rate": 2.2592635390185656e-05, + "loss": 0.0, + "step": 14231 + }, + { + "epoch": 3.29, + "learning_rate": 2.2590709498497807e-05, + "loss": 0.0, + "step": 14232 + }, + { + "epoch": 3.29, + "learning_rate": 2.2588783606809955e-05, + "loss": 0.0, + "step": 14233 + }, + { + "epoch": 3.29, + "learning_rate": 2.2586857715122102e-05, + "loss": 0.0004, + "step": 14234 + }, + { + "epoch": 3.29, + "learning_rate": 2.2584931823434253e-05, + "loss": 0.0, + "step": 14235 + }, + { + "epoch": 3.29, + "learning_rate": 2.25830059317464e-05, + "loss": 0.0, + "step": 14236 + }, + { + "epoch": 3.29, + "learning_rate": 2.2581080040058548e-05, + "loss": 0.0, + "step": 14237 + }, + { + "epoch": 3.29, + "learning_rate": 2.2579154148370695e-05, + "loss": 0.0, + "step": 14238 + }, + { + "epoch": 3.29, + "learning_rate": 2.2577228256682846e-05, + "loss": 0.0, + "step": 14239 + }, + { + "epoch": 3.29, + "learning_rate": 2.2575302364994994e-05, + "loss": 0.0, + "step": 14240 + }, + { + "epoch": 3.29, + "learning_rate": 2.257337647330714e-05, + "loss": 0.0, + "step": 14241 + }, + { + "epoch": 3.29, + "learning_rate": 2.2571450581619292e-05, + "loss": 0.0, + "step": 14242 + }, + { + "epoch": 3.29, + "learning_rate": 2.256952468993144e-05, + "loss": 0.0, + "step": 14243 + }, + { + "epoch": 3.29, + "learning_rate": 2.256759879824359e-05, + "loss": 0.0001, + "step": 14244 + }, + { + "epoch": 3.29, + "learning_rate": 2.2565672906555738e-05, + "loss": 0.0, + "step": 14245 + }, + { + "epoch": 3.29, + "learning_rate": 2.2563747014867885e-05, + "loss": 0.0, + "step": 14246 + }, + { + "epoch": 3.29, + "learning_rate": 2.2561821123180033e-05, + "loss": 0.0, + "step": 14247 + }, + { + "epoch": 3.29, + "learning_rate": 2.255989523149218e-05, + "loss": 0.0001, + "step": 14248 + }, + { + "epoch": 3.29, + "learning_rate": 2.255796933980433e-05, + "loss": 0.0, + "step": 14249 + }, + { + "epoch": 3.29, + "learning_rate": 2.255604344811648e-05, + "loss": 0.0, + "step": 14250 + }, + { + "epoch": 3.29, + "learning_rate": 2.2554117556428626e-05, + "loss": 0.0, + "step": 14251 + }, + { + "epoch": 3.29, + "learning_rate": 2.2552191664740777e-05, + "loss": 0.0, + "step": 14252 + }, + { + "epoch": 3.29, + "learning_rate": 2.2550265773052924e-05, + "loss": 0.0, + "step": 14253 + }, + { + "epoch": 3.29, + "learning_rate": 2.2548339881365075e-05, + "loss": 0.0, + "step": 14254 + }, + { + "epoch": 3.29, + "learning_rate": 2.254641398967722e-05, + "loss": 0.0, + "step": 14255 + }, + { + "epoch": 3.29, + "learning_rate": 2.254448809798937e-05, + "loss": 0.0, + "step": 14256 + }, + { + "epoch": 3.29, + "learning_rate": 2.2542562206301518e-05, + "loss": 0.0, + "step": 14257 + }, + { + "epoch": 3.3, + "learning_rate": 2.2540636314613665e-05, + "loss": 0.0, + "step": 14258 + }, + { + "epoch": 3.3, + "learning_rate": 2.2538710422925816e-05, + "loss": 0.0, + "step": 14259 + }, + { + "epoch": 3.3, + "learning_rate": 2.2536784531237964e-05, + "loss": 0.0, + "step": 14260 + }, + { + "epoch": 3.3, + "learning_rate": 2.2534858639550114e-05, + "loss": 0.0, + "step": 14261 + }, + { + "epoch": 3.3, + "learning_rate": 2.2532932747862262e-05, + "loss": 0.0, + "step": 14262 + }, + { + "epoch": 3.3, + "learning_rate": 2.253100685617441e-05, + "loss": 0.0, + "step": 14263 + }, + { + "epoch": 3.3, + "learning_rate": 2.252908096448656e-05, + "loss": 0.0, + "step": 14264 + }, + { + "epoch": 3.3, + "learning_rate": 2.2527155072798704e-05, + "loss": 0.0, + "step": 14265 + }, + { + "epoch": 3.3, + "learning_rate": 2.2525229181110855e-05, + "loss": 0.0, + "step": 14266 + }, + { + "epoch": 3.3, + "learning_rate": 2.2523303289423003e-05, + "loss": 0.0, + "step": 14267 + }, + { + "epoch": 3.3, + "learning_rate": 2.2521377397735154e-05, + "loss": 0.0001, + "step": 14268 + }, + { + "epoch": 3.3, + "learning_rate": 2.25194515060473e-05, + "loss": 0.0, + "step": 14269 + }, + { + "epoch": 3.3, + "learning_rate": 2.251752561435945e-05, + "loss": 0.0, + "step": 14270 + }, + { + "epoch": 3.3, + "learning_rate": 2.25155997226716e-05, + "loss": 0.0, + "step": 14271 + }, + { + "epoch": 3.3, + "learning_rate": 2.2513673830983747e-05, + "loss": 0.0, + "step": 14272 + }, + { + "epoch": 3.3, + "learning_rate": 2.2511747939295898e-05, + "loss": 0.0, + "step": 14273 + }, + { + "epoch": 3.3, + "learning_rate": 2.2509822047608042e-05, + "loss": 0.0, + "step": 14274 + }, + { + "epoch": 3.3, + "learning_rate": 2.2507896155920193e-05, + "loss": 0.0, + "step": 14275 + }, + { + "epoch": 3.3, + "learning_rate": 2.250597026423234e-05, + "loss": 0.0, + "step": 14276 + }, + { + "epoch": 3.3, + "learning_rate": 2.2504044372544488e-05, + "loss": 0.0, + "step": 14277 + }, + { + "epoch": 3.3, + "learning_rate": 2.250211848085664e-05, + "loss": 0.0, + "step": 14278 + }, + { + "epoch": 3.3, + "learning_rate": 2.2500192589168786e-05, + "loss": 0.0002, + "step": 14279 + }, + { + "epoch": 3.3, + "learning_rate": 2.2498266697480937e-05, + "loss": 0.0008, + "step": 14280 + }, + { + "epoch": 3.3, + "learning_rate": 2.2496340805793084e-05, + "loss": 0.0, + "step": 14281 + }, + { + "epoch": 3.3, + "learning_rate": 2.2494414914105232e-05, + "loss": 0.0, + "step": 14282 + }, + { + "epoch": 3.3, + "learning_rate": 2.249248902241738e-05, + "loss": 0.0, + "step": 14283 + }, + { + "epoch": 3.3, + "learning_rate": 2.2490563130729527e-05, + "loss": 0.0, + "step": 14284 + }, + { + "epoch": 3.3, + "learning_rate": 2.2488637239041678e-05, + "loss": 0.0001, + "step": 14285 + }, + { + "epoch": 3.3, + "learning_rate": 2.2486711347353825e-05, + "loss": 0.0, + "step": 14286 + }, + { + "epoch": 3.3, + "learning_rate": 2.2484785455665973e-05, + "loss": 0.0, + "step": 14287 + }, + { + "epoch": 3.3, + "learning_rate": 2.2482859563978123e-05, + "loss": 0.0, + "step": 14288 + }, + { + "epoch": 3.3, + "learning_rate": 2.248093367229027e-05, + "loss": 0.0001, + "step": 14289 + }, + { + "epoch": 3.3, + "learning_rate": 2.2479007780602422e-05, + "loss": 0.0, + "step": 14290 + }, + { + "epoch": 3.3, + "learning_rate": 2.247708188891457e-05, + "loss": 0.0, + "step": 14291 + }, + { + "epoch": 3.3, + "learning_rate": 2.2475155997226717e-05, + "loss": 0.0, + "step": 14292 + }, + { + "epoch": 3.3, + "learning_rate": 2.2473230105538864e-05, + "loss": 0.0, + "step": 14293 + }, + { + "epoch": 3.3, + "learning_rate": 2.2471304213851012e-05, + "loss": 0.0001, + "step": 14294 + }, + { + "epoch": 3.3, + "learning_rate": 2.2469378322163163e-05, + "loss": 0.0, + "step": 14295 + }, + { + "epoch": 3.3, + "learning_rate": 2.246745243047531e-05, + "loss": 0.0, + "step": 14296 + }, + { + "epoch": 3.3, + "learning_rate": 2.246552653878746e-05, + "loss": 0.0, + "step": 14297 + }, + { + "epoch": 3.3, + "learning_rate": 2.246360064709961e-05, + "loss": 0.0, + "step": 14298 + }, + { + "epoch": 3.3, + "learning_rate": 2.2461674755411756e-05, + "loss": 0.0, + "step": 14299 + }, + { + "epoch": 3.3, + "learning_rate": 2.2459748863723907e-05, + "loss": 0.0, + "step": 14300 + }, + { + "epoch": 3.31, + "learning_rate": 2.2457822972036054e-05, + "loss": 0.0, + "step": 14301 + }, + { + "epoch": 3.31, + "learning_rate": 2.2455897080348202e-05, + "loss": 0.0, + "step": 14302 + }, + { + "epoch": 3.31, + "learning_rate": 2.245397118866035e-05, + "loss": 0.0, + "step": 14303 + }, + { + "epoch": 3.31, + "learning_rate": 2.24520452969725e-05, + "loss": 0.0002, + "step": 14304 + }, + { + "epoch": 3.31, + "learning_rate": 2.2450119405284648e-05, + "loss": 0.0, + "step": 14305 + }, + { + "epoch": 3.31, + "learning_rate": 2.2448193513596795e-05, + "loss": 0.0, + "step": 14306 + }, + { + "epoch": 3.31, + "learning_rate": 2.2446267621908946e-05, + "loss": 0.0, + "step": 14307 + }, + { + "epoch": 3.31, + "learning_rate": 2.2444341730221093e-05, + "loss": 0.0, + "step": 14308 + }, + { + "epoch": 3.31, + "learning_rate": 2.2442415838533244e-05, + "loss": 0.0, + "step": 14309 + }, + { + "epoch": 3.31, + "learning_rate": 2.244048994684539e-05, + "loss": 0.0, + "step": 14310 + }, + { + "epoch": 3.31, + "learning_rate": 2.243856405515754e-05, + "loss": 0.0, + "step": 14311 + }, + { + "epoch": 3.31, + "learning_rate": 2.2436638163469687e-05, + "loss": 0.0002, + "step": 14312 + }, + { + "epoch": 3.31, + "learning_rate": 2.2434712271781834e-05, + "loss": 0.0, + "step": 14313 + }, + { + "epoch": 3.31, + "learning_rate": 2.2432786380093985e-05, + "loss": 0.0002, + "step": 14314 + }, + { + "epoch": 3.31, + "learning_rate": 2.2430860488406132e-05, + "loss": 0.0001, + "step": 14315 + }, + { + "epoch": 3.31, + "learning_rate": 2.2428934596718283e-05, + "loss": 0.0, + "step": 14316 + }, + { + "epoch": 3.31, + "learning_rate": 2.242700870503043e-05, + "loss": 0.0, + "step": 14317 + }, + { + "epoch": 3.31, + "learning_rate": 2.2425082813342578e-05, + "loss": 0.0, + "step": 14318 + }, + { + "epoch": 3.31, + "learning_rate": 2.242315692165473e-05, + "loss": 0.0, + "step": 14319 + }, + { + "epoch": 3.31, + "learning_rate": 2.2421231029966877e-05, + "loss": 0.0, + "step": 14320 + }, + { + "epoch": 3.31, + "learning_rate": 2.2419305138279024e-05, + "loss": 0.0, + "step": 14321 + }, + { + "epoch": 3.31, + "learning_rate": 2.241737924659117e-05, + "loss": 0.0, + "step": 14322 + }, + { + "epoch": 3.31, + "learning_rate": 2.2415453354903322e-05, + "loss": 0.0, + "step": 14323 + }, + { + "epoch": 3.31, + "learning_rate": 2.241352746321547e-05, + "loss": 0.0, + "step": 14324 + }, + { + "epoch": 3.31, + "learning_rate": 2.2411601571527617e-05, + "loss": 0.0, + "step": 14325 + }, + { + "epoch": 3.31, + "learning_rate": 2.2409675679839768e-05, + "loss": 0.0, + "step": 14326 + }, + { + "epoch": 3.31, + "learning_rate": 2.2407749788151916e-05, + "loss": 0.0, + "step": 14327 + }, + { + "epoch": 3.31, + "learning_rate": 2.2405823896464063e-05, + "loss": 0.0, + "step": 14328 + }, + { + "epoch": 3.31, + "learning_rate": 2.2403898004776214e-05, + "loss": 0.0, + "step": 14329 + }, + { + "epoch": 3.31, + "learning_rate": 2.2401972113088358e-05, + "loss": 0.0, + "step": 14330 + }, + { + "epoch": 3.31, + "learning_rate": 2.240004622140051e-05, + "loss": 0.0026, + "step": 14331 + }, + { + "epoch": 3.31, + "learning_rate": 2.2398120329712657e-05, + "loss": 0.0, + "step": 14332 + }, + { + "epoch": 3.31, + "learning_rate": 2.2396194438024807e-05, + "loss": 0.0, + "step": 14333 + }, + { + "epoch": 3.31, + "learning_rate": 2.2394268546336955e-05, + "loss": 0.0, + "step": 14334 + }, + { + "epoch": 3.31, + "learning_rate": 2.2392342654649102e-05, + "loss": 0.0, + "step": 14335 + }, + { + "epoch": 3.31, + "learning_rate": 2.2390416762961253e-05, + "loss": 0.0, + "step": 14336 + }, + { + "epoch": 3.31, + "learning_rate": 2.23884908712734e-05, + "loss": 0.0, + "step": 14337 + }, + { + "epoch": 3.31, + "learning_rate": 2.238656497958555e-05, + "loss": 0.0, + "step": 14338 + }, + { + "epoch": 3.31, + "learning_rate": 2.2384639087897696e-05, + "loss": 0.0, + "step": 14339 + }, + { + "epoch": 3.31, + "learning_rate": 2.2382713196209847e-05, + "loss": 0.0, + "step": 14340 + }, + { + "epoch": 3.31, + "learning_rate": 2.2380787304521994e-05, + "loss": 0.0, + "step": 14341 + }, + { + "epoch": 3.31, + "learning_rate": 2.237886141283414e-05, + "loss": 0.0001, + "step": 14342 + }, + { + "epoch": 3.31, + "learning_rate": 2.2376935521146292e-05, + "loss": 0.0, + "step": 14343 + }, + { + "epoch": 3.31, + "learning_rate": 2.237500962945844e-05, + "loss": 0.0, + "step": 14344 + }, + { + "epoch": 3.32, + "learning_rate": 2.237308373777059e-05, + "loss": 0.0, + "step": 14345 + }, + { + "epoch": 3.32, + "learning_rate": 2.2371157846082738e-05, + "loss": 0.0, + "step": 14346 + }, + { + "epoch": 3.32, + "learning_rate": 2.2369231954394886e-05, + "loss": 0.0, + "step": 14347 + }, + { + "epoch": 3.32, + "learning_rate": 2.2367306062707036e-05, + "loss": 0.0, + "step": 14348 + }, + { + "epoch": 3.32, + "learning_rate": 2.236538017101918e-05, + "loss": 0.0001, + "step": 14349 + }, + { + "epoch": 3.32, + "learning_rate": 2.236345427933133e-05, + "loss": 0.0002, + "step": 14350 + }, + { + "epoch": 3.32, + "learning_rate": 2.236152838764348e-05, + "loss": 0.0, + "step": 14351 + }, + { + "epoch": 3.32, + "learning_rate": 2.235960249595563e-05, + "loss": 0.0, + "step": 14352 + }, + { + "epoch": 3.32, + "learning_rate": 2.2357676604267777e-05, + "loss": 0.0, + "step": 14353 + }, + { + "epoch": 3.32, + "learning_rate": 2.2355750712579925e-05, + "loss": 0.0, + "step": 14354 + }, + { + "epoch": 3.32, + "learning_rate": 2.2353824820892076e-05, + "loss": 0.0001, + "step": 14355 + }, + { + "epoch": 3.32, + "learning_rate": 2.2351898929204223e-05, + "loss": 0.0, + "step": 14356 + }, + { + "epoch": 3.32, + "learning_rate": 2.2349973037516374e-05, + "loss": 0.0001, + "step": 14357 + }, + { + "epoch": 3.32, + "learning_rate": 2.2348047145828518e-05, + "loss": 0.0, + "step": 14358 + }, + { + "epoch": 3.32, + "learning_rate": 2.234612125414067e-05, + "loss": 0.0, + "step": 14359 + }, + { + "epoch": 3.32, + "learning_rate": 2.2344195362452816e-05, + "loss": 0.0, + "step": 14360 + }, + { + "epoch": 3.32, + "learning_rate": 2.2342269470764964e-05, + "loss": 0.0, + "step": 14361 + }, + { + "epoch": 3.32, + "learning_rate": 2.2340343579077115e-05, + "loss": 0.0, + "step": 14362 + }, + { + "epoch": 3.32, + "learning_rate": 2.2338417687389262e-05, + "loss": 0.0, + "step": 14363 + }, + { + "epoch": 3.32, + "learning_rate": 2.233649179570141e-05, + "loss": 0.0, + "step": 14364 + }, + { + "epoch": 3.32, + "learning_rate": 2.233456590401356e-05, + "loss": 0.0, + "step": 14365 + }, + { + "epoch": 3.32, + "learning_rate": 2.2332640012325708e-05, + "loss": 0.0, + "step": 14366 + }, + { + "epoch": 3.32, + "learning_rate": 2.2330714120637856e-05, + "loss": 0.0, + "step": 14367 + }, + { + "epoch": 3.32, + "learning_rate": 2.2328788228950003e-05, + "loss": 0.0, + "step": 14368 + }, + { + "epoch": 3.32, + "learning_rate": 2.2326862337262154e-05, + "loss": 0.0, + "step": 14369 + }, + { + "epoch": 3.32, + "learning_rate": 2.23249364455743e-05, + "loss": 0.0, + "step": 14370 + }, + { + "epoch": 3.32, + "learning_rate": 2.232301055388645e-05, + "loss": 0.0001, + "step": 14371 + }, + { + "epoch": 3.32, + "learning_rate": 2.23210846621986e-05, + "loss": 0.0, + "step": 14372 + }, + { + "epoch": 3.32, + "learning_rate": 2.2319158770510747e-05, + "loss": 0.0, + "step": 14373 + }, + { + "epoch": 3.32, + "learning_rate": 2.2317232878822898e-05, + "loss": 0.0, + "step": 14374 + }, + { + "epoch": 3.32, + "learning_rate": 2.2315306987135046e-05, + "loss": 0.0, + "step": 14375 + }, + { + "epoch": 3.32, + "learning_rate": 2.2313381095447193e-05, + "loss": 0.0001, + "step": 14376 + }, + { + "epoch": 3.32, + "learning_rate": 2.231145520375934e-05, + "loss": 0.0, + "step": 14377 + }, + { + "epoch": 3.32, + "learning_rate": 2.2309529312071488e-05, + "loss": 0.0001, + "step": 14378 + }, + { + "epoch": 3.32, + "learning_rate": 2.230760342038364e-05, + "loss": 0.0, + "step": 14379 + }, + { + "epoch": 3.32, + "learning_rate": 2.2305677528695786e-05, + "loss": 0.0, + "step": 14380 + }, + { + "epoch": 3.32, + "learning_rate": 2.2303751637007937e-05, + "loss": 0.0, + "step": 14381 + }, + { + "epoch": 3.32, + "learning_rate": 2.2301825745320085e-05, + "loss": 0.0, + "step": 14382 + }, + { + "epoch": 3.32, + "learning_rate": 2.2299899853632232e-05, + "loss": 0.0001, + "step": 14383 + }, + { + "epoch": 3.32, + "learning_rate": 2.2297973961944383e-05, + "loss": 0.0, + "step": 14384 + }, + { + "epoch": 3.32, + "learning_rate": 2.229604807025653e-05, + "loss": 0.0, + "step": 14385 + }, + { + "epoch": 3.32, + "learning_rate": 2.2294122178568678e-05, + "loss": 0.0001, + "step": 14386 + }, + { + "epoch": 3.32, + "learning_rate": 2.2292196286880825e-05, + "loss": 0.0, + "step": 14387 + }, + { + "epoch": 3.33, + "learning_rate": 2.2290270395192976e-05, + "loss": 0.0, + "step": 14388 + }, + { + "epoch": 3.33, + "learning_rate": 2.2288344503505124e-05, + "loss": 0.0, + "step": 14389 + }, + { + "epoch": 3.33, + "learning_rate": 2.228641861181727e-05, + "loss": 0.0001, + "step": 14390 + }, + { + "epoch": 3.33, + "learning_rate": 2.2284492720129422e-05, + "loss": 0.0, + "step": 14391 + }, + { + "epoch": 3.33, + "learning_rate": 2.228256682844157e-05, + "loss": 0.0, + "step": 14392 + }, + { + "epoch": 3.33, + "learning_rate": 2.228064093675372e-05, + "loss": 0.0, + "step": 14393 + }, + { + "epoch": 3.33, + "learning_rate": 2.2278715045065868e-05, + "loss": 0.0001, + "step": 14394 + }, + { + "epoch": 3.33, + "learning_rate": 2.2276789153378015e-05, + "loss": 0.0, + "step": 14395 + }, + { + "epoch": 3.33, + "learning_rate": 2.2274863261690163e-05, + "loss": 0.0005, + "step": 14396 + }, + { + "epoch": 3.33, + "learning_rate": 2.227293737000231e-05, + "loss": 0.0, + "step": 14397 + }, + { + "epoch": 3.33, + "learning_rate": 2.227101147831446e-05, + "loss": 0.0, + "step": 14398 + }, + { + "epoch": 3.33, + "learning_rate": 2.226908558662661e-05, + "loss": 0.0, + "step": 14399 + }, + { + "epoch": 3.33, + "learning_rate": 2.226715969493876e-05, + "loss": 0.0, + "step": 14400 + }, + { + "epoch": 3.33, + "learning_rate": 2.2265233803250907e-05, + "loss": 0.0002, + "step": 14401 + }, + { + "epoch": 3.33, + "learning_rate": 2.2263307911563055e-05, + "loss": 0.0, + "step": 14402 + }, + { + "epoch": 3.33, + "learning_rate": 2.2261382019875205e-05, + "loss": 0.0, + "step": 14403 + }, + { + "epoch": 3.33, + "learning_rate": 2.2259456128187353e-05, + "loss": 0.0, + "step": 14404 + }, + { + "epoch": 3.33, + "learning_rate": 2.22575302364995e-05, + "loss": 0.0, + "step": 14405 + }, + { + "epoch": 3.33, + "learning_rate": 2.2255604344811648e-05, + "loss": 0.0, + "step": 14406 + }, + { + "epoch": 3.33, + "learning_rate": 2.2253678453123795e-05, + "loss": 0.0, + "step": 14407 + }, + { + "epoch": 3.33, + "learning_rate": 2.2251752561435946e-05, + "loss": 0.0001, + "step": 14408 + }, + { + "epoch": 3.33, + "learning_rate": 2.2249826669748094e-05, + "loss": 0.0, + "step": 14409 + }, + { + "epoch": 3.33, + "learning_rate": 2.2247900778060244e-05, + "loss": 0.0, + "step": 14410 + }, + { + "epoch": 3.33, + "learning_rate": 2.2245974886372392e-05, + "loss": 0.0, + "step": 14411 + }, + { + "epoch": 3.33, + "learning_rate": 2.224404899468454e-05, + "loss": 0.0, + "step": 14412 + }, + { + "epoch": 3.33, + "learning_rate": 2.224212310299669e-05, + "loss": 0.0, + "step": 14413 + }, + { + "epoch": 3.33, + "learning_rate": 2.2240197211308834e-05, + "loss": 0.0001, + "step": 14414 + }, + { + "epoch": 3.33, + "learning_rate": 2.2238271319620985e-05, + "loss": 0.0, + "step": 14415 + }, + { + "epoch": 3.33, + "learning_rate": 2.2236345427933133e-05, + "loss": 0.0, + "step": 14416 + }, + { + "epoch": 3.33, + "learning_rate": 2.2234419536245284e-05, + "loss": 0.0, + "step": 14417 + }, + { + "epoch": 3.33, + "learning_rate": 2.223249364455743e-05, + "loss": 0.0, + "step": 14418 + }, + { + "epoch": 3.33, + "learning_rate": 2.223056775286958e-05, + "loss": 0.0, + "step": 14419 + }, + { + "epoch": 3.33, + "learning_rate": 2.222864186118173e-05, + "loss": 0.0, + "step": 14420 + }, + { + "epoch": 3.33, + "learning_rate": 2.2226715969493877e-05, + "loss": 0.0, + "step": 14421 + }, + { + "epoch": 3.33, + "learning_rate": 2.2224790077806028e-05, + "loss": 0.0, + "step": 14422 + }, + { + "epoch": 3.33, + "learning_rate": 2.2222864186118172e-05, + "loss": 0.0, + "step": 14423 + }, + { + "epoch": 3.33, + "learning_rate": 2.2220938294430323e-05, + "loss": 0.0, + "step": 14424 + }, + { + "epoch": 3.33, + "learning_rate": 2.221901240274247e-05, + "loss": 0.0002, + "step": 14425 + }, + { + "epoch": 3.33, + "learning_rate": 2.2217086511054618e-05, + "loss": 0.0, + "step": 14426 + }, + { + "epoch": 3.33, + "learning_rate": 2.221516061936677e-05, + "loss": 0.0001, + "step": 14427 + }, + { + "epoch": 3.33, + "learning_rate": 2.2213234727678916e-05, + "loss": 0.0, + "step": 14428 + }, + { + "epoch": 3.33, + "learning_rate": 2.2211308835991067e-05, + "loss": 0.0, + "step": 14429 + }, + { + "epoch": 3.33, + "learning_rate": 2.2209382944303214e-05, + "loss": 0.0, + "step": 14430 + }, + { + "epoch": 3.34, + "learning_rate": 2.2207457052615362e-05, + "loss": 0.0, + "step": 14431 + }, + { + "epoch": 3.34, + "learning_rate": 2.2205531160927513e-05, + "loss": 0.0, + "step": 14432 + }, + { + "epoch": 3.34, + "learning_rate": 2.2203605269239657e-05, + "loss": 0.0, + "step": 14433 + }, + { + "epoch": 3.34, + "learning_rate": 2.2201679377551808e-05, + "loss": 0.0, + "step": 14434 + }, + { + "epoch": 3.34, + "learning_rate": 2.2199753485863955e-05, + "loss": 0.0, + "step": 14435 + }, + { + "epoch": 3.34, + "learning_rate": 2.2197827594176106e-05, + "loss": 0.0, + "step": 14436 + }, + { + "epoch": 3.34, + "learning_rate": 2.2195901702488253e-05, + "loss": 0.0, + "step": 14437 + }, + { + "epoch": 3.34, + "learning_rate": 2.21939758108004e-05, + "loss": 0.0, + "step": 14438 + }, + { + "epoch": 3.34, + "learning_rate": 2.2192049919112552e-05, + "loss": 0.0, + "step": 14439 + }, + { + "epoch": 3.34, + "learning_rate": 2.21901240274247e-05, + "loss": 0.0, + "step": 14440 + }, + { + "epoch": 3.34, + "learning_rate": 2.2188198135736847e-05, + "loss": 0.0, + "step": 14441 + }, + { + "epoch": 3.34, + "learning_rate": 2.2186272244048994e-05, + "loss": 0.0, + "step": 14442 + }, + { + "epoch": 3.34, + "learning_rate": 2.2184346352361142e-05, + "loss": 0.0001, + "step": 14443 + }, + { + "epoch": 3.34, + "learning_rate": 2.2182420460673293e-05, + "loss": 0.0, + "step": 14444 + }, + { + "epoch": 3.34, + "learning_rate": 2.218049456898544e-05, + "loss": 0.0, + "step": 14445 + }, + { + "epoch": 3.34, + "learning_rate": 2.217856867729759e-05, + "loss": 0.0, + "step": 14446 + }, + { + "epoch": 3.34, + "learning_rate": 2.217664278560974e-05, + "loss": 0.0, + "step": 14447 + }, + { + "epoch": 3.34, + "learning_rate": 2.2174716893921886e-05, + "loss": 0.0, + "step": 14448 + }, + { + "epoch": 3.34, + "learning_rate": 2.2172791002234037e-05, + "loss": 0.0, + "step": 14449 + }, + { + "epoch": 3.34, + "learning_rate": 2.2170865110546184e-05, + "loss": 0.0001, + "step": 14450 + }, + { + "epoch": 3.34, + "learning_rate": 2.2168939218858332e-05, + "loss": 0.0, + "step": 14451 + }, + { + "epoch": 3.34, + "learning_rate": 2.216701332717048e-05, + "loss": 0.0, + "step": 14452 + }, + { + "epoch": 3.34, + "learning_rate": 2.216508743548263e-05, + "loss": 0.0, + "step": 14453 + }, + { + "epoch": 3.34, + "learning_rate": 2.2163161543794778e-05, + "loss": 0.0001, + "step": 14454 + }, + { + "epoch": 3.34, + "learning_rate": 2.2161235652106925e-05, + "loss": 0.0, + "step": 14455 + }, + { + "epoch": 3.34, + "learning_rate": 2.2159309760419076e-05, + "loss": 0.0, + "step": 14456 + }, + { + "epoch": 3.34, + "learning_rate": 2.2157383868731223e-05, + "loss": 0.0, + "step": 14457 + }, + { + "epoch": 3.34, + "learning_rate": 2.2155457977043374e-05, + "loss": 0.0, + "step": 14458 + }, + { + "epoch": 3.34, + "learning_rate": 2.2153532085355522e-05, + "loss": 0.0, + "step": 14459 + }, + { + "epoch": 3.34, + "learning_rate": 2.215160619366767e-05, + "loss": 0.0, + "step": 14460 + }, + { + "epoch": 3.34, + "learning_rate": 2.2149680301979817e-05, + "loss": 0.0001, + "step": 14461 + }, + { + "epoch": 3.34, + "learning_rate": 2.2147754410291964e-05, + "loss": 0.0, + "step": 14462 + }, + { + "epoch": 3.34, + "learning_rate": 2.2145828518604115e-05, + "loss": 0.0, + "step": 14463 + }, + { + "epoch": 3.34, + "learning_rate": 2.2143902626916263e-05, + "loss": 0.0001, + "step": 14464 + }, + { + "epoch": 3.34, + "learning_rate": 2.2141976735228413e-05, + "loss": 0.0, + "step": 14465 + }, + { + "epoch": 3.34, + "learning_rate": 2.214005084354056e-05, + "loss": 0.0, + "step": 14466 + }, + { + "epoch": 3.34, + "learning_rate": 2.213812495185271e-05, + "loss": 0.0, + "step": 14467 + }, + { + "epoch": 3.34, + "learning_rate": 2.213619906016486e-05, + "loss": 0.0001, + "step": 14468 + }, + { + "epoch": 3.34, + "learning_rate": 2.2134273168477007e-05, + "loss": 0.0, + "step": 14469 + }, + { + "epoch": 3.34, + "learning_rate": 2.2132347276789154e-05, + "loss": 0.0, + "step": 14470 + }, + { + "epoch": 3.34, + "learning_rate": 2.21304213851013e-05, + "loss": 0.0001, + "step": 14471 + }, + { + "epoch": 3.34, + "learning_rate": 2.2128495493413452e-05, + "loss": 0.0001, + "step": 14472 + }, + { + "epoch": 3.34, + "learning_rate": 2.21265696017256e-05, + "loss": 0.0, + "step": 14473 + }, + { + "epoch": 3.35, + "learning_rate": 2.2124643710037747e-05, + "loss": 0.0, + "step": 14474 + }, + { + "epoch": 3.35, + "learning_rate": 2.21227178183499e-05, + "loss": 0.0, + "step": 14475 + }, + { + "epoch": 3.35, + "learning_rate": 2.2120791926662046e-05, + "loss": 0.0, + "step": 14476 + }, + { + "epoch": 3.35, + "learning_rate": 2.2118866034974197e-05, + "loss": 0.0, + "step": 14477 + }, + { + "epoch": 3.35, + "learning_rate": 2.2116940143286344e-05, + "loss": 0.0, + "step": 14478 + }, + { + "epoch": 3.35, + "learning_rate": 2.2115014251598488e-05, + "loss": 0.0001, + "step": 14479 + }, + { + "epoch": 3.35, + "learning_rate": 2.211308835991064e-05, + "loss": 0.0, + "step": 14480 + }, + { + "epoch": 3.35, + "learning_rate": 2.2111162468222787e-05, + "loss": 0.0, + "step": 14481 + }, + { + "epoch": 3.35, + "learning_rate": 2.2109236576534937e-05, + "loss": 0.0, + "step": 14482 + }, + { + "epoch": 3.35, + "learning_rate": 2.2107310684847085e-05, + "loss": 0.0, + "step": 14483 + }, + { + "epoch": 3.35, + "learning_rate": 2.2105384793159232e-05, + "loss": 0.0, + "step": 14484 + }, + { + "epoch": 3.35, + "learning_rate": 2.2103458901471383e-05, + "loss": 0.0, + "step": 14485 + }, + { + "epoch": 3.35, + "learning_rate": 2.210153300978353e-05, + "loss": 0.0, + "step": 14486 + }, + { + "epoch": 3.35, + "learning_rate": 2.209960711809568e-05, + "loss": 0.0, + "step": 14487 + }, + { + "epoch": 3.35, + "learning_rate": 2.209768122640783e-05, + "loss": 0.0, + "step": 14488 + }, + { + "epoch": 3.35, + "learning_rate": 2.2095755334719977e-05, + "loss": 0.0, + "step": 14489 + }, + { + "epoch": 3.35, + "learning_rate": 2.2093829443032124e-05, + "loss": 0.0, + "step": 14490 + }, + { + "epoch": 3.35, + "learning_rate": 2.209190355134427e-05, + "loss": 0.0, + "step": 14491 + }, + { + "epoch": 3.35, + "learning_rate": 2.2089977659656422e-05, + "loss": 0.0, + "step": 14492 + }, + { + "epoch": 3.35, + "learning_rate": 2.208805176796857e-05, + "loss": 0.0, + "step": 14493 + }, + { + "epoch": 3.35, + "learning_rate": 2.208612587628072e-05, + "loss": 0.0001, + "step": 14494 + }, + { + "epoch": 3.35, + "learning_rate": 2.2084199984592868e-05, + "loss": 0.0, + "step": 14495 + }, + { + "epoch": 3.35, + "learning_rate": 2.2082274092905016e-05, + "loss": 0.0, + "step": 14496 + }, + { + "epoch": 3.35, + "learning_rate": 2.2080348201217167e-05, + "loss": 0.0, + "step": 14497 + }, + { + "epoch": 3.35, + "learning_rate": 2.207842230952931e-05, + "loss": 0.0, + "step": 14498 + }, + { + "epoch": 3.35, + "learning_rate": 2.207649641784146e-05, + "loss": 0.0, + "step": 14499 + }, + { + "epoch": 3.35, + "learning_rate": 2.207457052615361e-05, + "loss": 0.0, + "step": 14500 + }, + { + "epoch": 3.35, + "learning_rate": 2.207264463446576e-05, + "loss": 0.0, + "step": 14501 + }, + { + "epoch": 3.35, + "learning_rate": 2.2070718742777907e-05, + "loss": 0.002, + "step": 14502 + }, + { + "epoch": 3.35, + "learning_rate": 2.2068792851090055e-05, + "loss": 0.0001, + "step": 14503 + }, + { + "epoch": 3.35, + "learning_rate": 2.2066866959402206e-05, + "loss": 0.0, + "step": 14504 + }, + { + "epoch": 3.35, + "learning_rate": 2.2064941067714353e-05, + "loss": 0.0, + "step": 14505 + }, + { + "epoch": 3.35, + "learning_rate": 2.2063015176026504e-05, + "loss": 0.0, + "step": 14506 + }, + { + "epoch": 3.35, + "learning_rate": 2.2061089284338648e-05, + "loss": 0.0, + "step": 14507 + }, + { + "epoch": 3.35, + "learning_rate": 2.20591633926508e-05, + "loss": 0.0, + "step": 14508 + }, + { + "epoch": 3.35, + "learning_rate": 2.2057237500962946e-05, + "loss": 0.0, + "step": 14509 + }, + { + "epoch": 3.35, + "learning_rate": 2.2055311609275094e-05, + "loss": 0.0, + "step": 14510 + }, + { + "epoch": 3.35, + "learning_rate": 2.2053385717587245e-05, + "loss": 0.0, + "step": 14511 + }, + { + "epoch": 3.35, + "learning_rate": 2.2051459825899392e-05, + "loss": 0.0, + "step": 14512 + }, + { + "epoch": 3.35, + "learning_rate": 2.2049533934211543e-05, + "loss": 0.0, + "step": 14513 + }, + { + "epoch": 3.35, + "learning_rate": 2.204760804252369e-05, + "loss": 0.0, + "step": 14514 + }, + { + "epoch": 3.35, + "learning_rate": 2.2045682150835838e-05, + "loss": 0.0, + "step": 14515 + }, + { + "epoch": 3.35, + "learning_rate": 2.204375625914799e-05, + "loss": 0.0, + "step": 14516 + }, + { + "epoch": 3.35, + "learning_rate": 2.2041830367460133e-05, + "loss": 0.0, + "step": 14517 + }, + { + "epoch": 3.36, + "learning_rate": 2.2039904475772284e-05, + "loss": 0.0, + "step": 14518 + }, + { + "epoch": 3.36, + "learning_rate": 2.203797858408443e-05, + "loss": 0.0, + "step": 14519 + }, + { + "epoch": 3.36, + "learning_rate": 2.203605269239658e-05, + "loss": 0.0001, + "step": 14520 + }, + { + "epoch": 3.36, + "learning_rate": 2.203412680070873e-05, + "loss": 0.0, + "step": 14521 + }, + { + "epoch": 3.36, + "learning_rate": 2.2032200909020877e-05, + "loss": 0.0, + "step": 14522 + }, + { + "epoch": 3.36, + "learning_rate": 2.2030275017333028e-05, + "loss": 0.0, + "step": 14523 + }, + { + "epoch": 3.36, + "learning_rate": 2.2028349125645176e-05, + "loss": 0.0, + "step": 14524 + }, + { + "epoch": 3.36, + "learning_rate": 2.2026423233957323e-05, + "loss": 0.0, + "step": 14525 + }, + { + "epoch": 3.36, + "learning_rate": 2.202449734226947e-05, + "loss": 0.0, + "step": 14526 + }, + { + "epoch": 3.36, + "learning_rate": 2.2022571450581618e-05, + "loss": 0.0001, + "step": 14527 + }, + { + "epoch": 3.36, + "learning_rate": 2.202064555889377e-05, + "loss": 0.0, + "step": 14528 + }, + { + "epoch": 3.36, + "learning_rate": 2.2018719667205916e-05, + "loss": 0.0, + "step": 14529 + }, + { + "epoch": 3.36, + "learning_rate": 2.2016793775518067e-05, + "loss": 0.0001, + "step": 14530 + }, + { + "epoch": 3.36, + "learning_rate": 2.2014867883830215e-05, + "loss": 0.0, + "step": 14531 + }, + { + "epoch": 3.36, + "learning_rate": 2.2012941992142362e-05, + "loss": 0.0, + "step": 14532 + }, + { + "epoch": 3.36, + "learning_rate": 2.2011016100454513e-05, + "loss": 0.0, + "step": 14533 + }, + { + "epoch": 3.36, + "learning_rate": 2.200909020876666e-05, + "loss": 0.0, + "step": 14534 + }, + { + "epoch": 3.36, + "learning_rate": 2.2007164317078808e-05, + "loss": 0.0, + "step": 14535 + }, + { + "epoch": 3.36, + "learning_rate": 2.2005238425390955e-05, + "loss": 0.0, + "step": 14536 + }, + { + "epoch": 3.36, + "learning_rate": 2.2003312533703106e-05, + "loss": 0.0, + "step": 14537 + }, + { + "epoch": 3.36, + "learning_rate": 2.2001386642015254e-05, + "loss": 0.0, + "step": 14538 + }, + { + "epoch": 3.36, + "learning_rate": 2.19994607503274e-05, + "loss": 0.0, + "step": 14539 + }, + { + "epoch": 3.36, + "learning_rate": 2.1997534858639552e-05, + "loss": 0.0, + "step": 14540 + }, + { + "epoch": 3.36, + "learning_rate": 2.19956089669517e-05, + "loss": 0.0, + "step": 14541 + }, + { + "epoch": 3.36, + "learning_rate": 2.199368307526385e-05, + "loss": 0.0, + "step": 14542 + }, + { + "epoch": 3.36, + "learning_rate": 2.1991757183575998e-05, + "loss": 0.0133, + "step": 14543 + }, + { + "epoch": 3.36, + "learning_rate": 2.1989831291888145e-05, + "loss": 0.0, + "step": 14544 + }, + { + "epoch": 3.36, + "learning_rate": 2.1987905400200293e-05, + "loss": 0.0, + "step": 14545 + }, + { + "epoch": 3.36, + "learning_rate": 2.198597950851244e-05, + "loss": 0.0, + "step": 14546 + }, + { + "epoch": 3.36, + "learning_rate": 2.198405361682459e-05, + "loss": 0.0, + "step": 14547 + }, + { + "epoch": 3.36, + "learning_rate": 2.198212772513674e-05, + "loss": 0.0, + "step": 14548 + }, + { + "epoch": 3.36, + "learning_rate": 2.198020183344889e-05, + "loss": 0.0, + "step": 14549 + }, + { + "epoch": 3.36, + "learning_rate": 2.1978275941761037e-05, + "loss": 0.0, + "step": 14550 + }, + { + "epoch": 3.36, + "learning_rate": 2.1976350050073185e-05, + "loss": 0.0, + "step": 14551 + }, + { + "epoch": 3.36, + "learning_rate": 2.1974424158385335e-05, + "loss": 0.0, + "step": 14552 + }, + { + "epoch": 3.36, + "learning_rate": 2.1972498266697483e-05, + "loss": 0.0, + "step": 14553 + }, + { + "epoch": 3.36, + "learning_rate": 2.197057237500963e-05, + "loss": 0.0, + "step": 14554 + }, + { + "epoch": 3.36, + "learning_rate": 2.1968646483321778e-05, + "loss": 0.0, + "step": 14555 + }, + { + "epoch": 3.36, + "learning_rate": 2.1966720591633925e-05, + "loss": 0.0, + "step": 14556 + }, + { + "epoch": 3.36, + "learning_rate": 2.1964794699946076e-05, + "loss": 0.0, + "step": 14557 + }, + { + "epoch": 3.36, + "learning_rate": 2.1962868808258224e-05, + "loss": 0.0001, + "step": 14558 + }, + { + "epoch": 3.36, + "learning_rate": 2.1960942916570375e-05, + "loss": 0.0, + "step": 14559 + }, + { + "epoch": 3.36, + "learning_rate": 2.1959017024882522e-05, + "loss": 0.0, + "step": 14560 + }, + { + "epoch": 3.37, + "learning_rate": 2.195709113319467e-05, + "loss": 0.0, + "step": 14561 + }, + { + "epoch": 3.37, + "learning_rate": 2.195516524150682e-05, + "loss": 0.0, + "step": 14562 + }, + { + "epoch": 3.37, + "learning_rate": 2.1953239349818964e-05, + "loss": 0.0, + "step": 14563 + }, + { + "epoch": 3.37, + "learning_rate": 2.1951313458131115e-05, + "loss": 0.0, + "step": 14564 + }, + { + "epoch": 3.37, + "learning_rate": 2.1949387566443263e-05, + "loss": 0.0004, + "step": 14565 + }, + { + "epoch": 3.37, + "learning_rate": 2.1947461674755414e-05, + "loss": 0.0, + "step": 14566 + }, + { + "epoch": 3.37, + "learning_rate": 2.194553578306756e-05, + "loss": 0.0, + "step": 14567 + }, + { + "epoch": 3.37, + "learning_rate": 2.194360989137971e-05, + "loss": 0.0, + "step": 14568 + }, + { + "epoch": 3.37, + "learning_rate": 2.194168399969186e-05, + "loss": 0.0, + "step": 14569 + }, + { + "epoch": 3.37, + "learning_rate": 2.1939758108004007e-05, + "loss": 0.0, + "step": 14570 + }, + { + "epoch": 3.37, + "learning_rate": 2.1937832216316158e-05, + "loss": 0.0, + "step": 14571 + }, + { + "epoch": 3.37, + "learning_rate": 2.1935906324628305e-05, + "loss": 0.0001, + "step": 14572 + }, + { + "epoch": 3.37, + "learning_rate": 2.1933980432940453e-05, + "loss": 0.0, + "step": 14573 + }, + { + "epoch": 3.37, + "learning_rate": 2.19320545412526e-05, + "loss": 0.0011, + "step": 14574 + }, + { + "epoch": 3.37, + "learning_rate": 2.1930128649564748e-05, + "loss": 0.0, + "step": 14575 + }, + { + "epoch": 3.37, + "learning_rate": 2.19282027578769e-05, + "loss": 0.0, + "step": 14576 + }, + { + "epoch": 3.37, + "learning_rate": 2.1926276866189046e-05, + "loss": 0.0, + "step": 14577 + }, + { + "epoch": 3.37, + "learning_rate": 2.1924350974501197e-05, + "loss": 0.0001, + "step": 14578 + }, + { + "epoch": 3.37, + "learning_rate": 2.1922425082813344e-05, + "loss": 0.0, + "step": 14579 + }, + { + "epoch": 3.37, + "learning_rate": 2.1920499191125492e-05, + "loss": 0.0, + "step": 14580 + }, + { + "epoch": 3.37, + "learning_rate": 2.1918573299437643e-05, + "loss": 0.0, + "step": 14581 + }, + { + "epoch": 3.37, + "learning_rate": 2.1916647407749787e-05, + "loss": 0.0001, + "step": 14582 + }, + { + "epoch": 3.37, + "learning_rate": 2.1914721516061938e-05, + "loss": 0.0, + "step": 14583 + }, + { + "epoch": 3.37, + "learning_rate": 2.1912795624374085e-05, + "loss": 0.0, + "step": 14584 + }, + { + "epoch": 3.37, + "learning_rate": 2.1910869732686236e-05, + "loss": 0.0001, + "step": 14585 + }, + { + "epoch": 3.37, + "learning_rate": 2.1908943840998384e-05, + "loss": 0.0, + "step": 14586 + }, + { + "epoch": 3.37, + "learning_rate": 2.190701794931053e-05, + "loss": 0.0, + "step": 14587 + }, + { + "epoch": 3.37, + "learning_rate": 2.1905092057622682e-05, + "loss": 0.0001, + "step": 14588 + }, + { + "epoch": 3.37, + "learning_rate": 2.190316616593483e-05, + "loss": 0.0, + "step": 14589 + }, + { + "epoch": 3.37, + "learning_rate": 2.190124027424698e-05, + "loss": 0.0, + "step": 14590 + }, + { + "epoch": 3.37, + "learning_rate": 2.1899314382559124e-05, + "loss": 0.0, + "step": 14591 + }, + { + "epoch": 3.37, + "learning_rate": 2.1897388490871272e-05, + "loss": 0.0001, + "step": 14592 + }, + { + "epoch": 3.37, + "learning_rate": 2.1895462599183423e-05, + "loss": 0.0, + "step": 14593 + }, + { + "epoch": 3.37, + "learning_rate": 2.189353670749557e-05, + "loss": 0.0, + "step": 14594 + }, + { + "epoch": 3.37, + "learning_rate": 2.189161081580772e-05, + "loss": 0.0, + "step": 14595 + }, + { + "epoch": 3.37, + "learning_rate": 2.188968492411987e-05, + "loss": 0.0, + "step": 14596 + }, + { + "epoch": 3.37, + "learning_rate": 2.1887759032432016e-05, + "loss": 0.0, + "step": 14597 + }, + { + "epoch": 3.37, + "learning_rate": 2.1885833140744167e-05, + "loss": 0.0, + "step": 14598 + }, + { + "epoch": 3.37, + "learning_rate": 2.1883907249056314e-05, + "loss": 0.0, + "step": 14599 + }, + { + "epoch": 3.37, + "learning_rate": 2.1881981357368465e-05, + "loss": 0.0, + "step": 14600 + }, + { + "epoch": 3.37, + "learning_rate": 2.188005546568061e-05, + "loss": 0.0, + "step": 14601 + }, + { + "epoch": 3.37, + "learning_rate": 2.187812957399276e-05, + "loss": 0.0, + "step": 14602 + }, + { + "epoch": 3.37, + "learning_rate": 2.1876203682304908e-05, + "loss": 0.0001, + "step": 14603 + }, + { + "epoch": 3.38, + "learning_rate": 2.1874277790617055e-05, + "loss": 0.0, + "step": 14604 + }, + { + "epoch": 3.38, + "learning_rate": 2.1872351898929206e-05, + "loss": 0.0, + "step": 14605 + }, + { + "epoch": 3.38, + "learning_rate": 2.1870426007241353e-05, + "loss": 0.0, + "step": 14606 + }, + { + "epoch": 3.38, + "learning_rate": 2.1868500115553504e-05, + "loss": 0.0, + "step": 14607 + }, + { + "epoch": 3.38, + "learning_rate": 2.1866574223865652e-05, + "loss": 0.0001, + "step": 14608 + }, + { + "epoch": 3.38, + "learning_rate": 2.18646483321778e-05, + "loss": 0.0, + "step": 14609 + }, + { + "epoch": 3.38, + "learning_rate": 2.1862722440489947e-05, + "loss": 0.0001, + "step": 14610 + }, + { + "epoch": 3.38, + "learning_rate": 2.1860796548802094e-05, + "loss": 0.0, + "step": 14611 + }, + { + "epoch": 3.38, + "learning_rate": 2.1858870657114245e-05, + "loss": 0.0, + "step": 14612 + }, + { + "epoch": 3.38, + "learning_rate": 2.1856944765426393e-05, + "loss": 0.0, + "step": 14613 + }, + { + "epoch": 3.38, + "learning_rate": 2.1855018873738543e-05, + "loss": 0.0, + "step": 14614 + }, + { + "epoch": 3.38, + "learning_rate": 2.185309298205069e-05, + "loss": 0.0, + "step": 14615 + }, + { + "epoch": 3.38, + "learning_rate": 2.185116709036284e-05, + "loss": 0.0, + "step": 14616 + }, + { + "epoch": 3.38, + "learning_rate": 2.184924119867499e-05, + "loss": 0.0, + "step": 14617 + }, + { + "epoch": 3.38, + "learning_rate": 2.1847315306987137e-05, + "loss": 0.0, + "step": 14618 + }, + { + "epoch": 3.38, + "learning_rate": 2.1845389415299284e-05, + "loss": 0.0, + "step": 14619 + }, + { + "epoch": 3.38, + "learning_rate": 2.184346352361143e-05, + "loss": 0.0, + "step": 14620 + }, + { + "epoch": 3.38, + "learning_rate": 2.1841537631923583e-05, + "loss": 0.0, + "step": 14621 + }, + { + "epoch": 3.38, + "learning_rate": 2.183961174023573e-05, + "loss": 0.0, + "step": 14622 + }, + { + "epoch": 3.38, + "learning_rate": 2.1837685848547877e-05, + "loss": 0.0069, + "step": 14623 + }, + { + "epoch": 3.38, + "learning_rate": 2.183575995686003e-05, + "loss": 0.0, + "step": 14624 + }, + { + "epoch": 3.38, + "learning_rate": 2.1833834065172176e-05, + "loss": 0.0, + "step": 14625 + }, + { + "epoch": 3.38, + "learning_rate": 2.1831908173484327e-05, + "loss": 0.0001, + "step": 14626 + }, + { + "epoch": 3.38, + "learning_rate": 2.1829982281796474e-05, + "loss": 0.0, + "step": 14627 + }, + { + "epoch": 3.38, + "learning_rate": 2.182805639010862e-05, + "loss": 0.0, + "step": 14628 + }, + { + "epoch": 3.38, + "learning_rate": 2.182613049842077e-05, + "loss": 0.0, + "step": 14629 + }, + { + "epoch": 3.38, + "learning_rate": 2.1824204606732917e-05, + "loss": 0.0, + "step": 14630 + }, + { + "epoch": 3.38, + "learning_rate": 2.1822278715045067e-05, + "loss": 0.0, + "step": 14631 + }, + { + "epoch": 3.38, + "learning_rate": 2.1820352823357215e-05, + "loss": 0.0, + "step": 14632 + }, + { + "epoch": 3.38, + "learning_rate": 2.1818426931669362e-05, + "loss": 0.0, + "step": 14633 + }, + { + "epoch": 3.38, + "learning_rate": 2.1816501039981513e-05, + "loss": 0.0, + "step": 14634 + }, + { + "epoch": 3.38, + "learning_rate": 2.181457514829366e-05, + "loss": 0.0, + "step": 14635 + }, + { + "epoch": 3.38, + "learning_rate": 2.181264925660581e-05, + "loss": 0.0, + "step": 14636 + }, + { + "epoch": 3.38, + "learning_rate": 2.181072336491796e-05, + "loss": 0.0, + "step": 14637 + }, + { + "epoch": 3.38, + "learning_rate": 2.1808797473230107e-05, + "loss": 0.0, + "step": 14638 + }, + { + "epoch": 3.38, + "learning_rate": 2.1806871581542254e-05, + "loss": 0.0, + "step": 14639 + }, + { + "epoch": 3.38, + "learning_rate": 2.18049456898544e-05, + "loss": 0.0, + "step": 14640 + }, + { + "epoch": 3.38, + "learning_rate": 2.1803019798166552e-05, + "loss": 0.0, + "step": 14641 + }, + { + "epoch": 3.38, + "learning_rate": 2.18010939064787e-05, + "loss": 0.0, + "step": 14642 + }, + { + "epoch": 3.38, + "learning_rate": 2.179916801479085e-05, + "loss": 0.0, + "step": 14643 + }, + { + "epoch": 3.38, + "learning_rate": 2.1797242123102998e-05, + "loss": 0.0, + "step": 14644 + }, + { + "epoch": 3.38, + "learning_rate": 2.1795316231415146e-05, + "loss": 0.0, + "step": 14645 + }, + { + "epoch": 3.38, + "learning_rate": 2.1793390339727297e-05, + "loss": 0.0, + "step": 14646 + }, + { + "epoch": 3.39, + "learning_rate": 2.179146444803944e-05, + "loss": 0.0, + "step": 14647 + }, + { + "epoch": 3.39, + "learning_rate": 2.178953855635159e-05, + "loss": 0.0, + "step": 14648 + }, + { + "epoch": 3.39, + "learning_rate": 2.178761266466374e-05, + "loss": 0.0, + "step": 14649 + }, + { + "epoch": 3.39, + "learning_rate": 2.178568677297589e-05, + "loss": 0.0, + "step": 14650 + }, + { + "epoch": 3.39, + "learning_rate": 2.1783760881288037e-05, + "loss": 0.0, + "step": 14651 + }, + { + "epoch": 3.39, + "learning_rate": 2.1781834989600185e-05, + "loss": 0.0, + "step": 14652 + }, + { + "epoch": 3.39, + "learning_rate": 2.1779909097912336e-05, + "loss": 0.0, + "step": 14653 + }, + { + "epoch": 3.39, + "learning_rate": 2.1777983206224483e-05, + "loss": 0.0, + "step": 14654 + }, + { + "epoch": 3.39, + "learning_rate": 2.1776057314536634e-05, + "loss": 0.0, + "step": 14655 + }, + { + "epoch": 3.39, + "learning_rate": 2.177413142284878e-05, + "loss": 0.0, + "step": 14656 + }, + { + "epoch": 3.39, + "learning_rate": 2.177220553116093e-05, + "loss": 0.0, + "step": 14657 + }, + { + "epoch": 3.39, + "learning_rate": 2.1770279639473076e-05, + "loss": 0.0, + "step": 14658 + }, + { + "epoch": 3.39, + "learning_rate": 2.1768353747785224e-05, + "loss": 0.0, + "step": 14659 + }, + { + "epoch": 3.39, + "learning_rate": 2.1766427856097375e-05, + "loss": 0.0, + "step": 14660 + }, + { + "epoch": 3.39, + "learning_rate": 2.1764501964409522e-05, + "loss": 0.0, + "step": 14661 + }, + { + "epoch": 3.39, + "learning_rate": 2.1762576072721673e-05, + "loss": 0.0, + "step": 14662 + }, + { + "epoch": 3.39, + "learning_rate": 2.176065018103382e-05, + "loss": 0.0, + "step": 14663 + }, + { + "epoch": 3.39, + "learning_rate": 2.1758724289345968e-05, + "loss": 0.0, + "step": 14664 + }, + { + "epoch": 3.39, + "learning_rate": 2.175679839765812e-05, + "loss": 0.0, + "step": 14665 + }, + { + "epoch": 3.39, + "learning_rate": 2.1754872505970263e-05, + "loss": 0.0003, + "step": 14666 + }, + { + "epoch": 3.39, + "learning_rate": 2.1752946614282414e-05, + "loss": 0.0, + "step": 14667 + }, + { + "epoch": 3.39, + "learning_rate": 2.175102072259456e-05, + "loss": 0.0001, + "step": 14668 + }, + { + "epoch": 3.39, + "learning_rate": 2.174909483090671e-05, + "loss": 0.0, + "step": 14669 + }, + { + "epoch": 3.39, + "learning_rate": 2.174716893921886e-05, + "loss": 0.0, + "step": 14670 + }, + { + "epoch": 3.39, + "learning_rate": 2.1745243047531007e-05, + "loss": 0.0, + "step": 14671 + }, + { + "epoch": 3.39, + "learning_rate": 2.1743317155843158e-05, + "loss": 0.0, + "step": 14672 + }, + { + "epoch": 3.39, + "learning_rate": 2.1741391264155306e-05, + "loss": 0.0, + "step": 14673 + }, + { + "epoch": 3.39, + "learning_rate": 2.1739465372467453e-05, + "loss": 0.0, + "step": 14674 + }, + { + "epoch": 3.39, + "learning_rate": 2.17375394807796e-05, + "loss": 0.0, + "step": 14675 + }, + { + "epoch": 3.39, + "learning_rate": 2.1735613589091748e-05, + "loss": 0.0, + "step": 14676 + }, + { + "epoch": 3.39, + "learning_rate": 2.17336876974039e-05, + "loss": 0.0, + "step": 14677 + }, + { + "epoch": 3.39, + "learning_rate": 2.1731761805716046e-05, + "loss": 0.0, + "step": 14678 + }, + { + "epoch": 3.39, + "learning_rate": 2.1729835914028197e-05, + "loss": 0.0, + "step": 14679 + }, + { + "epoch": 3.39, + "learning_rate": 2.1727910022340345e-05, + "loss": 0.0, + "step": 14680 + }, + { + "epoch": 3.39, + "learning_rate": 2.1725984130652492e-05, + "loss": 0.0, + "step": 14681 + }, + { + "epoch": 3.39, + "learning_rate": 2.1724058238964643e-05, + "loss": 0.0, + "step": 14682 + }, + { + "epoch": 3.39, + "learning_rate": 2.172213234727679e-05, + "loss": 0.0, + "step": 14683 + }, + { + "epoch": 3.39, + "learning_rate": 2.172020645558894e-05, + "loss": 0.0, + "step": 14684 + }, + { + "epoch": 3.39, + "learning_rate": 2.1718280563901085e-05, + "loss": 0.0, + "step": 14685 + }, + { + "epoch": 3.39, + "learning_rate": 2.1716354672213236e-05, + "loss": 0.0, + "step": 14686 + }, + { + "epoch": 3.39, + "learning_rate": 2.1714428780525384e-05, + "loss": 0.0, + "step": 14687 + }, + { + "epoch": 3.39, + "learning_rate": 2.171250288883753e-05, + "loss": 0.0, + "step": 14688 + }, + { + "epoch": 3.39, + "learning_rate": 2.1710576997149682e-05, + "loss": 0.0, + "step": 14689 + }, + { + "epoch": 3.39, + "learning_rate": 2.170865110546183e-05, + "loss": 0.0001, + "step": 14690 + }, + { + "epoch": 3.4, + "learning_rate": 2.170672521377398e-05, + "loss": 0.0, + "step": 14691 + }, + { + "epoch": 3.4, + "learning_rate": 2.1704799322086128e-05, + "loss": 0.0001, + "step": 14692 + }, + { + "epoch": 3.4, + "learning_rate": 2.1702873430398275e-05, + "loss": 0.0, + "step": 14693 + }, + { + "epoch": 3.4, + "learning_rate": 2.1700947538710423e-05, + "loss": 0.0, + "step": 14694 + }, + { + "epoch": 3.4, + "learning_rate": 2.169902164702257e-05, + "loss": 0.0, + "step": 14695 + }, + { + "epoch": 3.4, + "learning_rate": 2.169709575533472e-05, + "loss": 0.0, + "step": 14696 + }, + { + "epoch": 3.4, + "learning_rate": 2.169516986364687e-05, + "loss": 0.0002, + "step": 14697 + }, + { + "epoch": 3.4, + "learning_rate": 2.169324397195902e-05, + "loss": 0.0, + "step": 14698 + }, + { + "epoch": 3.4, + "learning_rate": 2.1691318080271167e-05, + "loss": 0.0001, + "step": 14699 + }, + { + "epoch": 3.4, + "learning_rate": 2.1689392188583315e-05, + "loss": 0.0, + "step": 14700 + }, + { + "epoch": 3.4, + "learning_rate": 2.1687466296895465e-05, + "loss": 0.0001, + "step": 14701 + }, + { + "epoch": 3.4, + "learning_rate": 2.1685540405207613e-05, + "loss": 0.0001, + "step": 14702 + }, + { + "epoch": 3.4, + "learning_rate": 2.168361451351976e-05, + "loss": 0.0, + "step": 14703 + }, + { + "epoch": 3.4, + "learning_rate": 2.1681688621831908e-05, + "loss": 0.0, + "step": 14704 + }, + { + "epoch": 3.4, + "learning_rate": 2.1679762730144055e-05, + "loss": 0.0, + "step": 14705 + }, + { + "epoch": 3.4, + "learning_rate": 2.1677836838456206e-05, + "loss": 0.0, + "step": 14706 + }, + { + "epoch": 3.4, + "learning_rate": 2.1675910946768354e-05, + "loss": 0.0, + "step": 14707 + }, + { + "epoch": 3.4, + "learning_rate": 2.1673985055080505e-05, + "loss": 0.0, + "step": 14708 + }, + { + "epoch": 3.4, + "learning_rate": 2.1672059163392652e-05, + "loss": 0.0, + "step": 14709 + }, + { + "epoch": 3.4, + "learning_rate": 2.16701332717048e-05, + "loss": 0.0, + "step": 14710 + }, + { + "epoch": 3.4, + "learning_rate": 2.166820738001695e-05, + "loss": 0.0, + "step": 14711 + }, + { + "epoch": 3.4, + "learning_rate": 2.1666281488329098e-05, + "loss": 0.0, + "step": 14712 + }, + { + "epoch": 3.4, + "learning_rate": 2.1664355596641245e-05, + "loss": 0.0, + "step": 14713 + }, + { + "epoch": 3.4, + "learning_rate": 2.1662429704953393e-05, + "loss": 0.0, + "step": 14714 + }, + { + "epoch": 3.4, + "learning_rate": 2.1660503813265544e-05, + "loss": 0.0, + "step": 14715 + }, + { + "epoch": 3.4, + "learning_rate": 2.165857792157769e-05, + "loss": 0.0, + "step": 14716 + }, + { + "epoch": 3.4, + "learning_rate": 2.165665202988984e-05, + "loss": 0.0, + "step": 14717 + }, + { + "epoch": 3.4, + "learning_rate": 2.165472613820199e-05, + "loss": 0.0001, + "step": 14718 + }, + { + "epoch": 3.4, + "learning_rate": 2.1652800246514137e-05, + "loss": 0.0, + "step": 14719 + }, + { + "epoch": 3.4, + "learning_rate": 2.1650874354826288e-05, + "loss": 0.0001, + "step": 14720 + }, + { + "epoch": 3.4, + "learning_rate": 2.1648948463138435e-05, + "loss": 0.0, + "step": 14721 + }, + { + "epoch": 3.4, + "learning_rate": 2.1647022571450583e-05, + "loss": 0.0, + "step": 14722 + }, + { + "epoch": 3.4, + "learning_rate": 2.164509667976273e-05, + "loss": 0.0, + "step": 14723 + }, + { + "epoch": 3.4, + "learning_rate": 2.1643170788074878e-05, + "loss": 0.0, + "step": 14724 + }, + { + "epoch": 3.4, + "learning_rate": 2.164124489638703e-05, + "loss": 0.0, + "step": 14725 + }, + { + "epoch": 3.4, + "learning_rate": 2.1639319004699176e-05, + "loss": 0.0, + "step": 14726 + }, + { + "epoch": 3.4, + "learning_rate": 2.1637393113011327e-05, + "loss": 0.0, + "step": 14727 + }, + { + "epoch": 3.4, + "learning_rate": 2.1635467221323474e-05, + "loss": 0.0, + "step": 14728 + }, + { + "epoch": 3.4, + "learning_rate": 2.1633541329635622e-05, + "loss": 0.0, + "step": 14729 + }, + { + "epoch": 3.4, + "learning_rate": 2.1631615437947773e-05, + "loss": 0.0, + "step": 14730 + }, + { + "epoch": 3.4, + "learning_rate": 2.1629689546259917e-05, + "loss": 0.0, + "step": 14731 + }, + { + "epoch": 3.4, + "learning_rate": 2.1627763654572068e-05, + "loss": 0.0, + "step": 14732 + }, + { + "epoch": 3.4, + "learning_rate": 2.1625837762884215e-05, + "loss": 0.0, + "step": 14733 + }, + { + "epoch": 3.41, + "learning_rate": 2.1623911871196366e-05, + "loss": 0.0, + "step": 14734 + }, + { + "epoch": 3.41, + "learning_rate": 2.1621985979508514e-05, + "loss": 0.0, + "step": 14735 + }, + { + "epoch": 3.41, + "learning_rate": 2.162006008782066e-05, + "loss": 0.0, + "step": 14736 + }, + { + "epoch": 3.41, + "learning_rate": 2.1618134196132812e-05, + "loss": 0.0, + "step": 14737 + }, + { + "epoch": 3.41, + "learning_rate": 2.161620830444496e-05, + "loss": 0.0, + "step": 14738 + }, + { + "epoch": 3.41, + "learning_rate": 2.161428241275711e-05, + "loss": 0.0, + "step": 14739 + }, + { + "epoch": 3.41, + "learning_rate": 2.1612356521069258e-05, + "loss": 0.0, + "step": 14740 + }, + { + "epoch": 3.41, + "learning_rate": 2.1610430629381402e-05, + "loss": 0.0, + "step": 14741 + }, + { + "epoch": 3.41, + "learning_rate": 2.1608504737693553e-05, + "loss": 0.0, + "step": 14742 + }, + { + "epoch": 3.41, + "learning_rate": 2.16065788460057e-05, + "loss": 0.0, + "step": 14743 + }, + { + "epoch": 3.41, + "learning_rate": 2.160465295431785e-05, + "loss": 0.0, + "step": 14744 + }, + { + "epoch": 3.41, + "learning_rate": 2.160272706263e-05, + "loss": 0.0, + "step": 14745 + }, + { + "epoch": 3.41, + "learning_rate": 2.1600801170942146e-05, + "loss": 0.0, + "step": 14746 + }, + { + "epoch": 3.41, + "learning_rate": 2.1598875279254297e-05, + "loss": 0.0, + "step": 14747 + }, + { + "epoch": 3.41, + "learning_rate": 2.1596949387566444e-05, + "loss": 0.0, + "step": 14748 + }, + { + "epoch": 3.41, + "learning_rate": 2.1595023495878595e-05, + "loss": 0.0, + "step": 14749 + }, + { + "epoch": 3.41, + "learning_rate": 2.159309760419074e-05, + "loss": 0.0, + "step": 14750 + }, + { + "epoch": 3.41, + "learning_rate": 2.159117171250289e-05, + "loss": 0.0, + "step": 14751 + }, + { + "epoch": 3.41, + "learning_rate": 2.1589245820815038e-05, + "loss": 0.0, + "step": 14752 + }, + { + "epoch": 3.41, + "learning_rate": 2.1587319929127185e-05, + "loss": 0.0, + "step": 14753 + }, + { + "epoch": 3.41, + "learning_rate": 2.1585394037439336e-05, + "loss": 0.0, + "step": 14754 + }, + { + "epoch": 3.41, + "learning_rate": 2.1583468145751483e-05, + "loss": 0.0, + "step": 14755 + }, + { + "epoch": 3.41, + "learning_rate": 2.1581542254063634e-05, + "loss": 0.0, + "step": 14756 + }, + { + "epoch": 3.41, + "learning_rate": 2.1579616362375782e-05, + "loss": 0.0, + "step": 14757 + }, + { + "epoch": 3.41, + "learning_rate": 2.157769047068793e-05, + "loss": 0.0, + "step": 14758 + }, + { + "epoch": 3.41, + "learning_rate": 2.1575764579000077e-05, + "loss": 0.0, + "step": 14759 + }, + { + "epoch": 3.41, + "learning_rate": 2.1573838687312224e-05, + "loss": 0.0, + "step": 14760 + }, + { + "epoch": 3.41, + "learning_rate": 2.1571912795624375e-05, + "loss": 0.0, + "step": 14761 + }, + { + "epoch": 3.41, + "learning_rate": 2.1569986903936523e-05, + "loss": 0.0, + "step": 14762 + }, + { + "epoch": 3.41, + "learning_rate": 2.1568061012248673e-05, + "loss": 0.0, + "step": 14763 + }, + { + "epoch": 3.41, + "learning_rate": 2.156613512056082e-05, + "loss": 0.0001, + "step": 14764 + }, + { + "epoch": 3.41, + "learning_rate": 2.156420922887297e-05, + "loss": 0.0001, + "step": 14765 + }, + { + "epoch": 3.41, + "learning_rate": 2.156228333718512e-05, + "loss": 0.0, + "step": 14766 + }, + { + "epoch": 3.41, + "learning_rate": 2.1560357445497267e-05, + "loss": 0.0, + "step": 14767 + }, + { + "epoch": 3.41, + "learning_rate": 2.1558431553809418e-05, + "loss": 0.0, + "step": 14768 + }, + { + "epoch": 3.41, + "learning_rate": 2.1556505662121562e-05, + "loss": 0.0, + "step": 14769 + }, + { + "epoch": 3.41, + "learning_rate": 2.1554579770433713e-05, + "loss": 0.0, + "step": 14770 + }, + { + "epoch": 3.41, + "learning_rate": 2.155265387874586e-05, + "loss": 0.0, + "step": 14771 + }, + { + "epoch": 3.41, + "learning_rate": 2.1550727987058008e-05, + "loss": 0.0, + "step": 14772 + }, + { + "epoch": 3.41, + "learning_rate": 2.154880209537016e-05, + "loss": 0.0001, + "step": 14773 + }, + { + "epoch": 3.41, + "learning_rate": 2.1546876203682306e-05, + "loss": 0.0, + "step": 14774 + }, + { + "epoch": 3.41, + "learning_rate": 2.1544950311994457e-05, + "loss": 0.0, + "step": 14775 + }, + { + "epoch": 3.41, + "learning_rate": 2.1543024420306604e-05, + "loss": 0.0004, + "step": 14776 + }, + { + "epoch": 3.42, + "learning_rate": 2.1541098528618752e-05, + "loss": 0.0, + "step": 14777 + }, + { + "epoch": 3.42, + "learning_rate": 2.15391726369309e-05, + "loss": 0.0, + "step": 14778 + }, + { + "epoch": 3.42, + "learning_rate": 2.1537246745243047e-05, + "loss": 0.0, + "step": 14779 + }, + { + "epoch": 3.42, + "learning_rate": 2.1535320853555198e-05, + "loss": 0.0, + "step": 14780 + }, + { + "epoch": 3.42, + "learning_rate": 2.1533394961867345e-05, + "loss": 0.0, + "step": 14781 + }, + { + "epoch": 3.42, + "learning_rate": 2.1531469070179492e-05, + "loss": 0.0, + "step": 14782 + }, + { + "epoch": 3.42, + "learning_rate": 2.1529543178491643e-05, + "loss": 0.0, + "step": 14783 + }, + { + "epoch": 3.42, + "learning_rate": 2.152761728680379e-05, + "loss": 0.0, + "step": 14784 + }, + { + "epoch": 3.42, + "learning_rate": 2.152569139511594e-05, + "loss": 0.0001, + "step": 14785 + }, + { + "epoch": 3.42, + "learning_rate": 2.152376550342809e-05, + "loss": 0.0, + "step": 14786 + }, + { + "epoch": 3.42, + "learning_rate": 2.1521839611740237e-05, + "loss": 0.0, + "step": 14787 + }, + { + "epoch": 3.42, + "learning_rate": 2.1519913720052384e-05, + "loss": 0.0, + "step": 14788 + }, + { + "epoch": 3.42, + "learning_rate": 2.151798782836453e-05, + "loss": 0.0, + "step": 14789 + }, + { + "epoch": 3.42, + "learning_rate": 2.1516061936676682e-05, + "loss": 0.0, + "step": 14790 + }, + { + "epoch": 3.42, + "learning_rate": 2.151413604498883e-05, + "loss": 0.0, + "step": 14791 + }, + { + "epoch": 3.42, + "learning_rate": 2.151221015330098e-05, + "loss": 0.0002, + "step": 14792 + }, + { + "epoch": 3.42, + "learning_rate": 2.1510284261613128e-05, + "loss": 0.0, + "step": 14793 + }, + { + "epoch": 3.42, + "learning_rate": 2.1508358369925276e-05, + "loss": 0.0, + "step": 14794 + }, + { + "epoch": 3.42, + "learning_rate": 2.1506432478237427e-05, + "loss": 0.0, + "step": 14795 + }, + { + "epoch": 3.42, + "learning_rate": 2.1504506586549574e-05, + "loss": 0.0002, + "step": 14796 + }, + { + "epoch": 3.42, + "learning_rate": 2.150258069486172e-05, + "loss": 0.0, + "step": 14797 + }, + { + "epoch": 3.42, + "learning_rate": 2.150065480317387e-05, + "loss": 0.0, + "step": 14798 + }, + { + "epoch": 3.42, + "learning_rate": 2.149872891148602e-05, + "loss": 0.0, + "step": 14799 + }, + { + "epoch": 3.42, + "learning_rate": 2.1496803019798167e-05, + "loss": 0.0, + "step": 14800 + }, + { + "epoch": 3.42, + "learning_rate": 2.1494877128110315e-05, + "loss": 0.0, + "step": 14801 + }, + { + "epoch": 3.42, + "learning_rate": 2.1492951236422466e-05, + "loss": 0.0, + "step": 14802 + }, + { + "epoch": 3.42, + "learning_rate": 2.1491025344734613e-05, + "loss": 0.0, + "step": 14803 + }, + { + "epoch": 3.42, + "learning_rate": 2.1489099453046764e-05, + "loss": 0.0, + "step": 14804 + }, + { + "epoch": 3.42, + "learning_rate": 2.148717356135891e-05, + "loss": 0.0, + "step": 14805 + }, + { + "epoch": 3.42, + "learning_rate": 2.148524766967106e-05, + "loss": 0.0, + "step": 14806 + }, + { + "epoch": 3.42, + "learning_rate": 2.1483321777983207e-05, + "loss": 0.0, + "step": 14807 + }, + { + "epoch": 3.42, + "learning_rate": 2.1481395886295354e-05, + "loss": 0.0, + "step": 14808 + }, + { + "epoch": 3.42, + "learning_rate": 2.1479469994607505e-05, + "loss": 0.0002, + "step": 14809 + }, + { + "epoch": 3.42, + "learning_rate": 2.1477544102919652e-05, + "loss": 0.0, + "step": 14810 + }, + { + "epoch": 3.42, + "learning_rate": 2.1475618211231803e-05, + "loss": 0.0, + "step": 14811 + }, + { + "epoch": 3.42, + "learning_rate": 2.147369231954395e-05, + "loss": 0.0, + "step": 14812 + }, + { + "epoch": 3.42, + "learning_rate": 2.1471766427856098e-05, + "loss": 0.0, + "step": 14813 + }, + { + "epoch": 3.42, + "learning_rate": 2.146984053616825e-05, + "loss": 0.0001, + "step": 14814 + }, + { + "epoch": 3.42, + "learning_rate": 2.1467914644480393e-05, + "loss": 0.0, + "step": 14815 + }, + { + "epoch": 3.42, + "learning_rate": 2.1465988752792544e-05, + "loss": 0.0001, + "step": 14816 + }, + { + "epoch": 3.42, + "learning_rate": 2.146406286110469e-05, + "loss": 0.0001, + "step": 14817 + }, + { + "epoch": 3.42, + "learning_rate": 2.146213696941684e-05, + "loss": 0.0, + "step": 14818 + }, + { + "epoch": 3.42, + "learning_rate": 2.146021107772899e-05, + "loss": 0.0, + "step": 14819 + }, + { + "epoch": 3.43, + "learning_rate": 2.1458285186041137e-05, + "loss": 0.0, + "step": 14820 + }, + { + "epoch": 3.43, + "learning_rate": 2.1456359294353288e-05, + "loss": 0.0001, + "step": 14821 + }, + { + "epoch": 3.43, + "learning_rate": 2.1454433402665436e-05, + "loss": 0.0, + "step": 14822 + }, + { + "epoch": 3.43, + "learning_rate": 2.1452507510977583e-05, + "loss": 0.0001, + "step": 14823 + }, + { + "epoch": 3.43, + "learning_rate": 2.1450581619289734e-05, + "loss": 0.0, + "step": 14824 + }, + { + "epoch": 3.43, + "learning_rate": 2.1448655727601878e-05, + "loss": 0.0, + "step": 14825 + }, + { + "epoch": 3.43, + "learning_rate": 2.144672983591403e-05, + "loss": 0.0, + "step": 14826 + }, + { + "epoch": 3.43, + "learning_rate": 2.1444803944226176e-05, + "loss": 0.0, + "step": 14827 + }, + { + "epoch": 3.43, + "learning_rate": 2.1442878052538327e-05, + "loss": 0.0, + "step": 14828 + }, + { + "epoch": 3.43, + "learning_rate": 2.1440952160850475e-05, + "loss": 0.0, + "step": 14829 + }, + { + "epoch": 3.43, + "learning_rate": 2.1439026269162622e-05, + "loss": 0.0, + "step": 14830 + }, + { + "epoch": 3.43, + "learning_rate": 2.1437100377474773e-05, + "loss": 0.0001, + "step": 14831 + }, + { + "epoch": 3.43, + "learning_rate": 2.143517448578692e-05, + "loss": 0.0001, + "step": 14832 + }, + { + "epoch": 3.43, + "learning_rate": 2.143324859409907e-05, + "loss": 0.0, + "step": 14833 + }, + { + "epoch": 3.43, + "learning_rate": 2.1431322702411216e-05, + "loss": 0.0, + "step": 14834 + }, + { + "epoch": 3.43, + "learning_rate": 2.1429396810723366e-05, + "loss": 0.0, + "step": 14835 + }, + { + "epoch": 3.43, + "learning_rate": 2.1427470919035514e-05, + "loss": 0.0005, + "step": 14836 + }, + { + "epoch": 3.43, + "learning_rate": 2.142554502734766e-05, + "loss": 0.0, + "step": 14837 + }, + { + "epoch": 3.43, + "learning_rate": 2.1423619135659812e-05, + "loss": 0.0, + "step": 14838 + }, + { + "epoch": 3.43, + "learning_rate": 2.142169324397196e-05, + "loss": 0.0, + "step": 14839 + }, + { + "epoch": 3.43, + "learning_rate": 2.141976735228411e-05, + "loss": 0.0, + "step": 14840 + }, + { + "epoch": 3.43, + "learning_rate": 2.1417841460596258e-05, + "loss": 0.0, + "step": 14841 + }, + { + "epoch": 3.43, + "learning_rate": 2.1415915568908406e-05, + "loss": 0.0, + "step": 14842 + }, + { + "epoch": 3.43, + "learning_rate": 2.1413989677220553e-05, + "loss": 0.0, + "step": 14843 + }, + { + "epoch": 3.43, + "learning_rate": 2.14120637855327e-05, + "loss": 0.0, + "step": 14844 + }, + { + "epoch": 3.43, + "learning_rate": 2.141013789384485e-05, + "loss": 0.0, + "step": 14845 + }, + { + "epoch": 3.43, + "learning_rate": 2.1408212002157e-05, + "loss": 0.0, + "step": 14846 + }, + { + "epoch": 3.43, + "learning_rate": 2.140628611046915e-05, + "loss": 0.0, + "step": 14847 + }, + { + "epoch": 3.43, + "learning_rate": 2.1404360218781297e-05, + "loss": 0.0, + "step": 14848 + }, + { + "epoch": 3.43, + "learning_rate": 2.1402434327093445e-05, + "loss": 0.0, + "step": 14849 + }, + { + "epoch": 3.43, + "learning_rate": 2.1400508435405596e-05, + "loss": 0.0, + "step": 14850 + }, + { + "epoch": 3.43, + "learning_rate": 2.1398582543717743e-05, + "loss": 0.0, + "step": 14851 + }, + { + "epoch": 3.43, + "learning_rate": 2.1396656652029894e-05, + "loss": 0.0, + "step": 14852 + }, + { + "epoch": 3.43, + "learning_rate": 2.1394730760342038e-05, + "loss": 0.0, + "step": 14853 + }, + { + "epoch": 3.43, + "learning_rate": 2.1392804868654185e-05, + "loss": 0.0, + "step": 14854 + }, + { + "epoch": 3.43, + "learning_rate": 2.1390878976966336e-05, + "loss": 0.0, + "step": 14855 + }, + { + "epoch": 3.43, + "learning_rate": 2.1388953085278484e-05, + "loss": 0.0, + "step": 14856 + }, + { + "epoch": 3.43, + "learning_rate": 2.1387027193590635e-05, + "loss": 0.0, + "step": 14857 + }, + { + "epoch": 3.43, + "learning_rate": 2.1385101301902782e-05, + "loss": 0.0, + "step": 14858 + }, + { + "epoch": 3.43, + "learning_rate": 2.138317541021493e-05, + "loss": 0.0, + "step": 14859 + }, + { + "epoch": 3.43, + "learning_rate": 2.138124951852708e-05, + "loss": 0.0002, + "step": 14860 + }, + { + "epoch": 3.43, + "learning_rate": 2.1379323626839228e-05, + "loss": 0.0, + "step": 14861 + }, + { + "epoch": 3.43, + "learning_rate": 2.1377397735151375e-05, + "loss": 0.0, + "step": 14862 + }, + { + "epoch": 3.43, + "learning_rate": 2.1375471843463523e-05, + "loss": 0.0, + "step": 14863 + }, + { + "epoch": 3.44, + "learning_rate": 2.1373545951775674e-05, + "loss": 0.0, + "step": 14864 + }, + { + "epoch": 3.44, + "learning_rate": 2.137162006008782e-05, + "loss": 0.0, + "step": 14865 + }, + { + "epoch": 3.44, + "learning_rate": 2.136969416839997e-05, + "loss": 0.0, + "step": 14866 + }, + { + "epoch": 3.44, + "learning_rate": 2.136776827671212e-05, + "loss": 0.0, + "step": 14867 + }, + { + "epoch": 3.44, + "learning_rate": 2.1365842385024267e-05, + "loss": 0.0, + "step": 14868 + }, + { + "epoch": 3.44, + "learning_rate": 2.1363916493336418e-05, + "loss": 0.0, + "step": 14869 + }, + { + "epoch": 3.44, + "learning_rate": 2.1361990601648565e-05, + "loss": 0.0017, + "step": 14870 + }, + { + "epoch": 3.44, + "learning_rate": 2.1360064709960713e-05, + "loss": 0.0, + "step": 14871 + }, + { + "epoch": 3.44, + "learning_rate": 2.135813881827286e-05, + "loss": 0.0, + "step": 14872 + }, + { + "epoch": 3.44, + "learning_rate": 2.1356212926585008e-05, + "loss": 0.0, + "step": 14873 + }, + { + "epoch": 3.44, + "learning_rate": 2.135428703489716e-05, + "loss": 0.0, + "step": 14874 + }, + { + "epoch": 3.44, + "learning_rate": 2.1352361143209306e-05, + "loss": 0.0, + "step": 14875 + }, + { + "epoch": 3.44, + "learning_rate": 2.1350435251521457e-05, + "loss": 0.0, + "step": 14876 + }, + { + "epoch": 3.44, + "learning_rate": 2.1348509359833605e-05, + "loss": 0.0, + "step": 14877 + }, + { + "epoch": 3.44, + "learning_rate": 2.1346583468145752e-05, + "loss": 0.0, + "step": 14878 + }, + { + "epoch": 3.44, + "learning_rate": 2.1344657576457903e-05, + "loss": 0.0, + "step": 14879 + }, + { + "epoch": 3.44, + "learning_rate": 2.134273168477005e-05, + "loss": 0.0, + "step": 14880 + }, + { + "epoch": 3.44, + "learning_rate": 2.1340805793082198e-05, + "loss": 0.0, + "step": 14881 + }, + { + "epoch": 3.44, + "learning_rate": 2.1338879901394345e-05, + "loss": 0.0, + "step": 14882 + }, + { + "epoch": 3.44, + "learning_rate": 2.1336954009706496e-05, + "loss": 0.0, + "step": 14883 + }, + { + "epoch": 3.44, + "learning_rate": 2.1335028118018644e-05, + "loss": 0.0001, + "step": 14884 + }, + { + "epoch": 3.44, + "learning_rate": 2.133310222633079e-05, + "loss": 0.0001, + "step": 14885 + }, + { + "epoch": 3.44, + "learning_rate": 2.1331176334642942e-05, + "loss": 0.0, + "step": 14886 + }, + { + "epoch": 3.44, + "learning_rate": 2.132925044295509e-05, + "loss": 0.0, + "step": 14887 + }, + { + "epoch": 3.44, + "learning_rate": 2.132732455126724e-05, + "loss": 0.0, + "step": 14888 + }, + { + "epoch": 3.44, + "learning_rate": 2.1325398659579388e-05, + "loss": 0.0, + "step": 14889 + }, + { + "epoch": 3.44, + "learning_rate": 2.1323472767891535e-05, + "loss": 0.0, + "step": 14890 + }, + { + "epoch": 3.44, + "learning_rate": 2.1321546876203683e-05, + "loss": 0.0, + "step": 14891 + }, + { + "epoch": 3.44, + "learning_rate": 2.131962098451583e-05, + "loss": 0.0, + "step": 14892 + }, + { + "epoch": 3.44, + "learning_rate": 2.131769509282798e-05, + "loss": 0.0, + "step": 14893 + }, + { + "epoch": 3.44, + "learning_rate": 2.131576920114013e-05, + "loss": 0.0001, + "step": 14894 + }, + { + "epoch": 3.44, + "learning_rate": 2.1313843309452276e-05, + "loss": 0.0001, + "step": 14895 + }, + { + "epoch": 3.44, + "learning_rate": 2.1311917417764427e-05, + "loss": 0.0, + "step": 14896 + }, + { + "epoch": 3.44, + "learning_rate": 2.1309991526076574e-05, + "loss": 0.0, + "step": 14897 + }, + { + "epoch": 3.44, + "learning_rate": 2.1308065634388725e-05, + "loss": 0.0, + "step": 14898 + }, + { + "epoch": 3.44, + "learning_rate": 2.130613974270087e-05, + "loss": 0.0, + "step": 14899 + }, + { + "epoch": 3.44, + "learning_rate": 2.130421385101302e-05, + "loss": 0.0, + "step": 14900 + }, + { + "epoch": 3.44, + "learning_rate": 2.1302287959325168e-05, + "loss": 0.0001, + "step": 14901 + }, + { + "epoch": 3.44, + "learning_rate": 2.1300362067637315e-05, + "loss": 0.0, + "step": 14902 + }, + { + "epoch": 3.44, + "learning_rate": 2.1298436175949466e-05, + "loss": 0.0, + "step": 14903 + }, + { + "epoch": 3.44, + "learning_rate": 2.1296510284261614e-05, + "loss": 0.0, + "step": 14904 + }, + { + "epoch": 3.44, + "learning_rate": 2.1294584392573764e-05, + "loss": 0.0001, + "step": 14905 + }, + { + "epoch": 3.44, + "learning_rate": 2.1292658500885912e-05, + "loss": 0.0, + "step": 14906 + }, + { + "epoch": 3.45, + "learning_rate": 2.129073260919806e-05, + "loss": 0.0, + "step": 14907 + }, + { + "epoch": 3.45, + "learning_rate": 2.128880671751021e-05, + "loss": 0.0001, + "step": 14908 + }, + { + "epoch": 3.45, + "learning_rate": 2.1286880825822354e-05, + "loss": 0.0, + "step": 14909 + }, + { + "epoch": 3.45, + "learning_rate": 2.1284954934134505e-05, + "loss": 0.0, + "step": 14910 + }, + { + "epoch": 3.45, + "learning_rate": 2.1283029042446653e-05, + "loss": 0.0, + "step": 14911 + }, + { + "epoch": 3.45, + "learning_rate": 2.1281103150758803e-05, + "loss": 0.0, + "step": 14912 + }, + { + "epoch": 3.45, + "learning_rate": 2.127917725907095e-05, + "loss": 0.0, + "step": 14913 + }, + { + "epoch": 3.45, + "learning_rate": 2.12772513673831e-05, + "loss": 0.0, + "step": 14914 + }, + { + "epoch": 3.45, + "learning_rate": 2.127532547569525e-05, + "loss": 0.0006, + "step": 14915 + }, + { + "epoch": 3.45, + "learning_rate": 2.1273399584007397e-05, + "loss": 0.0, + "step": 14916 + }, + { + "epoch": 3.45, + "learning_rate": 2.1271473692319548e-05, + "loss": 0.0, + "step": 14917 + }, + { + "epoch": 3.45, + "learning_rate": 2.1269547800631692e-05, + "loss": 0.0001, + "step": 14918 + }, + { + "epoch": 3.45, + "learning_rate": 2.1267621908943843e-05, + "loss": 0.0, + "step": 14919 + }, + { + "epoch": 3.45, + "learning_rate": 2.126569601725599e-05, + "loss": 0.0, + "step": 14920 + }, + { + "epoch": 3.45, + "learning_rate": 2.1263770125568138e-05, + "loss": 0.0001, + "step": 14921 + }, + { + "epoch": 3.45, + "learning_rate": 2.126184423388029e-05, + "loss": 0.0, + "step": 14922 + }, + { + "epoch": 3.45, + "learning_rate": 2.1259918342192436e-05, + "loss": 0.0, + "step": 14923 + }, + { + "epoch": 3.45, + "learning_rate": 2.1257992450504587e-05, + "loss": 0.0, + "step": 14924 + }, + { + "epoch": 3.45, + "learning_rate": 2.1256066558816734e-05, + "loss": 0.0, + "step": 14925 + }, + { + "epoch": 3.45, + "learning_rate": 2.1254140667128882e-05, + "loss": 0.0, + "step": 14926 + }, + { + "epoch": 3.45, + "learning_rate": 2.125221477544103e-05, + "loss": 0.0, + "step": 14927 + }, + { + "epoch": 3.45, + "learning_rate": 2.1250288883753177e-05, + "loss": 0.0, + "step": 14928 + }, + { + "epoch": 3.45, + "learning_rate": 2.1248362992065328e-05, + "loss": 0.0, + "step": 14929 + }, + { + "epoch": 3.45, + "learning_rate": 2.1246437100377475e-05, + "loss": 0.0, + "step": 14930 + }, + { + "epoch": 3.45, + "learning_rate": 2.1244511208689623e-05, + "loss": 0.0, + "step": 14931 + }, + { + "epoch": 3.45, + "learning_rate": 2.1242585317001773e-05, + "loss": 0.0, + "step": 14932 + }, + { + "epoch": 3.45, + "learning_rate": 2.124065942531392e-05, + "loss": 0.0, + "step": 14933 + }, + { + "epoch": 3.45, + "learning_rate": 2.1238733533626072e-05, + "loss": 0.0, + "step": 14934 + }, + { + "epoch": 3.45, + "learning_rate": 2.123680764193822e-05, + "loss": 0.0, + "step": 14935 + }, + { + "epoch": 3.45, + "learning_rate": 2.1234881750250367e-05, + "loss": 0.0, + "step": 14936 + }, + { + "epoch": 3.45, + "learning_rate": 2.1232955858562514e-05, + "loss": 0.0, + "step": 14937 + }, + { + "epoch": 3.45, + "learning_rate": 2.123102996687466e-05, + "loss": 0.0, + "step": 14938 + }, + { + "epoch": 3.45, + "learning_rate": 2.1229104075186813e-05, + "loss": 0.0, + "step": 14939 + }, + { + "epoch": 3.45, + "learning_rate": 2.122717818349896e-05, + "loss": 0.0, + "step": 14940 + }, + { + "epoch": 3.45, + "learning_rate": 2.122525229181111e-05, + "loss": 0.0, + "step": 14941 + }, + { + "epoch": 3.45, + "learning_rate": 2.122332640012326e-05, + "loss": 0.0015, + "step": 14942 + }, + { + "epoch": 3.45, + "learning_rate": 2.1221400508435406e-05, + "loss": 0.0, + "step": 14943 + }, + { + "epoch": 3.45, + "learning_rate": 2.1219474616747557e-05, + "loss": 0.0, + "step": 14944 + }, + { + "epoch": 3.45, + "learning_rate": 2.1217548725059704e-05, + "loss": 0.0, + "step": 14945 + }, + { + "epoch": 3.45, + "learning_rate": 2.121562283337185e-05, + "loss": 0.0, + "step": 14946 + }, + { + "epoch": 3.45, + "learning_rate": 2.1213696941684e-05, + "loss": 0.0, + "step": 14947 + }, + { + "epoch": 3.45, + "learning_rate": 2.121177104999615e-05, + "loss": 0.0, + "step": 14948 + }, + { + "epoch": 3.45, + "learning_rate": 2.1209845158308297e-05, + "loss": 0.0637, + "step": 14949 + }, + { + "epoch": 3.46, + "learning_rate": 2.1207919266620445e-05, + "loss": 0.0, + "step": 14950 + }, + { + "epoch": 3.46, + "learning_rate": 2.1205993374932596e-05, + "loss": 0.0, + "step": 14951 + }, + { + "epoch": 3.46, + "learning_rate": 2.1204067483244743e-05, + "loss": 0.0, + "step": 14952 + }, + { + "epoch": 3.46, + "learning_rate": 2.1202141591556894e-05, + "loss": 0.0, + "step": 14953 + }, + { + "epoch": 3.46, + "learning_rate": 2.120021569986904e-05, + "loss": 0.0, + "step": 14954 + }, + { + "epoch": 3.46, + "learning_rate": 2.119828980818119e-05, + "loss": 0.0, + "step": 14955 + }, + { + "epoch": 3.46, + "learning_rate": 2.1196363916493337e-05, + "loss": 0.0, + "step": 14956 + }, + { + "epoch": 3.46, + "learning_rate": 2.1194438024805484e-05, + "loss": 0.0, + "step": 14957 + }, + { + "epoch": 3.46, + "learning_rate": 2.1192512133117635e-05, + "loss": 0.0001, + "step": 14958 + }, + { + "epoch": 3.46, + "learning_rate": 2.1190586241429782e-05, + "loss": 0.0, + "step": 14959 + }, + { + "epoch": 3.46, + "learning_rate": 2.1188660349741933e-05, + "loss": 0.0, + "step": 14960 + }, + { + "epoch": 3.46, + "learning_rate": 2.118673445805408e-05, + "loss": 0.0, + "step": 14961 + }, + { + "epoch": 3.46, + "learning_rate": 2.1184808566366228e-05, + "loss": 0.0, + "step": 14962 + }, + { + "epoch": 3.46, + "learning_rate": 2.118288267467838e-05, + "loss": 0.0, + "step": 14963 + }, + { + "epoch": 3.46, + "learning_rate": 2.1180956782990527e-05, + "loss": 0.0, + "step": 14964 + }, + { + "epoch": 3.46, + "learning_rate": 2.1179030891302674e-05, + "loss": 0.0, + "step": 14965 + }, + { + "epoch": 3.46, + "learning_rate": 2.117710499961482e-05, + "loss": 0.0, + "step": 14966 + }, + { + "epoch": 3.46, + "learning_rate": 2.1175179107926972e-05, + "loss": 0.0001, + "step": 14967 + }, + { + "epoch": 3.46, + "learning_rate": 2.117325321623912e-05, + "loss": 0.0009, + "step": 14968 + }, + { + "epoch": 3.46, + "learning_rate": 2.1171327324551267e-05, + "loss": 0.0, + "step": 14969 + }, + { + "epoch": 3.46, + "learning_rate": 2.1169401432863418e-05, + "loss": 0.0, + "step": 14970 + }, + { + "epoch": 3.46, + "learning_rate": 2.1167475541175566e-05, + "loss": 0.0001, + "step": 14971 + }, + { + "epoch": 3.46, + "learning_rate": 2.1165549649487713e-05, + "loss": 0.0, + "step": 14972 + }, + { + "epoch": 3.46, + "learning_rate": 2.1163623757799864e-05, + "loss": 0.0, + "step": 14973 + }, + { + "epoch": 3.46, + "learning_rate": 2.1161697866112008e-05, + "loss": 0.0, + "step": 14974 + }, + { + "epoch": 3.46, + "learning_rate": 2.115977197442416e-05, + "loss": 0.0, + "step": 14975 + }, + { + "epoch": 3.46, + "learning_rate": 2.1157846082736306e-05, + "loss": 0.0, + "step": 14976 + }, + { + "epoch": 3.46, + "learning_rate": 2.1155920191048457e-05, + "loss": 0.0, + "step": 14977 + }, + { + "epoch": 3.46, + "learning_rate": 2.1153994299360605e-05, + "loss": 0.0, + "step": 14978 + }, + { + "epoch": 3.46, + "learning_rate": 2.1152068407672752e-05, + "loss": 0.0, + "step": 14979 + }, + { + "epoch": 3.46, + "learning_rate": 2.1150142515984903e-05, + "loss": 0.0, + "step": 14980 + }, + { + "epoch": 3.46, + "learning_rate": 2.114821662429705e-05, + "loss": 0.01, + "step": 14981 + }, + { + "epoch": 3.46, + "learning_rate": 2.11462907326092e-05, + "loss": 0.0, + "step": 14982 + }, + { + "epoch": 3.46, + "learning_rate": 2.1144364840921346e-05, + "loss": 0.0, + "step": 14983 + }, + { + "epoch": 3.46, + "learning_rate": 2.1142438949233496e-05, + "loss": 0.0, + "step": 14984 + }, + { + "epoch": 3.46, + "learning_rate": 2.1140513057545644e-05, + "loss": 0.0, + "step": 14985 + }, + { + "epoch": 3.46, + "learning_rate": 2.113858716585779e-05, + "loss": 0.0, + "step": 14986 + }, + { + "epoch": 3.46, + "learning_rate": 2.1136661274169942e-05, + "loss": 0.0, + "step": 14987 + }, + { + "epoch": 3.46, + "learning_rate": 2.113473538248209e-05, + "loss": 0.0001, + "step": 14988 + }, + { + "epoch": 3.46, + "learning_rate": 2.113280949079424e-05, + "loss": 0.0, + "step": 14989 + }, + { + "epoch": 3.46, + "learning_rate": 2.1130883599106388e-05, + "loss": 0.0, + "step": 14990 + }, + { + "epoch": 3.46, + "learning_rate": 2.1128957707418536e-05, + "loss": 0.0, + "step": 14991 + }, + { + "epoch": 3.46, + "learning_rate": 2.1127031815730686e-05, + "loss": 0.0, + "step": 14992 + }, + { + "epoch": 3.46, + "learning_rate": 2.112510592404283e-05, + "loss": 0.0001, + "step": 14993 + }, + { + "epoch": 3.47, + "learning_rate": 2.112318003235498e-05, + "loss": 0.0, + "step": 14994 + }, + { + "epoch": 3.47, + "learning_rate": 2.112125414066713e-05, + "loss": 0.0, + "step": 14995 + }, + { + "epoch": 3.47, + "learning_rate": 2.111932824897928e-05, + "loss": 0.0, + "step": 14996 + }, + { + "epoch": 3.47, + "learning_rate": 2.1117402357291427e-05, + "loss": 0.0, + "step": 14997 + }, + { + "epoch": 3.47, + "learning_rate": 2.1115476465603575e-05, + "loss": 0.0, + "step": 14998 + }, + { + "epoch": 3.47, + "learning_rate": 2.1113550573915726e-05, + "loss": 0.0, + "step": 14999 + }, + { + "epoch": 3.47, + "learning_rate": 2.1111624682227873e-05, + "loss": 0.0, + "step": 15000 + }, + { + "epoch": 3.47, + "learning_rate": 2.1109698790540024e-05, + "loss": 0.0001, + "step": 15001 + }, + { + "epoch": 3.47, + "learning_rate": 2.1107772898852168e-05, + "loss": 0.0, + "step": 15002 + }, + { + "epoch": 3.47, + "learning_rate": 2.110584700716432e-05, + "loss": 0.0, + "step": 15003 + }, + { + "epoch": 3.47, + "learning_rate": 2.1103921115476466e-05, + "loss": 0.0, + "step": 15004 + }, + { + "epoch": 3.47, + "learning_rate": 2.1101995223788614e-05, + "loss": 0.0, + "step": 15005 + }, + { + "epoch": 3.47, + "learning_rate": 2.1100069332100765e-05, + "loss": 0.0, + "step": 15006 + }, + { + "epoch": 3.47, + "learning_rate": 2.1098143440412912e-05, + "loss": 0.0, + "step": 15007 + }, + { + "epoch": 3.47, + "learning_rate": 2.109621754872506e-05, + "loss": 0.0, + "step": 15008 + }, + { + "epoch": 3.47, + "learning_rate": 2.109429165703721e-05, + "loss": 0.0, + "step": 15009 + }, + { + "epoch": 3.47, + "learning_rate": 2.1092365765349358e-05, + "loss": 0.0, + "step": 15010 + }, + { + "epoch": 3.47, + "learning_rate": 2.1090439873661505e-05, + "loss": 0.0, + "step": 15011 + }, + { + "epoch": 3.47, + "learning_rate": 2.1088513981973653e-05, + "loss": 0.0001, + "step": 15012 + }, + { + "epoch": 3.47, + "learning_rate": 2.1086588090285804e-05, + "loss": 0.0, + "step": 15013 + }, + { + "epoch": 3.47, + "learning_rate": 2.108466219859795e-05, + "loss": 0.0, + "step": 15014 + }, + { + "epoch": 3.47, + "learning_rate": 2.10827363069101e-05, + "loss": 0.0, + "step": 15015 + }, + { + "epoch": 3.47, + "learning_rate": 2.108081041522225e-05, + "loss": 0.0, + "step": 15016 + }, + { + "epoch": 3.47, + "learning_rate": 2.1078884523534397e-05, + "loss": 0.0, + "step": 15017 + }, + { + "epoch": 3.47, + "learning_rate": 2.1076958631846548e-05, + "loss": 0.0, + "step": 15018 + }, + { + "epoch": 3.47, + "learning_rate": 2.1075032740158695e-05, + "loss": 0.0, + "step": 15019 + }, + { + "epoch": 3.47, + "learning_rate": 2.1073106848470843e-05, + "loss": 0.0, + "step": 15020 + }, + { + "epoch": 3.47, + "learning_rate": 2.107118095678299e-05, + "loss": 0.0, + "step": 15021 + }, + { + "epoch": 3.47, + "learning_rate": 2.1069255065095138e-05, + "loss": 0.0, + "step": 15022 + }, + { + "epoch": 3.47, + "learning_rate": 2.106732917340729e-05, + "loss": 0.0, + "step": 15023 + }, + { + "epoch": 3.47, + "learning_rate": 2.1065403281719436e-05, + "loss": 0.0, + "step": 15024 + }, + { + "epoch": 3.47, + "learning_rate": 2.1063477390031587e-05, + "loss": 0.0, + "step": 15025 + }, + { + "epoch": 3.47, + "learning_rate": 2.1061551498343735e-05, + "loss": 0.0, + "step": 15026 + }, + { + "epoch": 3.47, + "learning_rate": 2.1059625606655882e-05, + "loss": 0.0, + "step": 15027 + }, + { + "epoch": 3.47, + "learning_rate": 2.1057699714968033e-05, + "loss": 0.0, + "step": 15028 + }, + { + "epoch": 3.47, + "learning_rate": 2.105577382328018e-05, + "loss": 0.0, + "step": 15029 + }, + { + "epoch": 3.47, + "learning_rate": 2.1053847931592328e-05, + "loss": 0.0, + "step": 15030 + }, + { + "epoch": 3.47, + "learning_rate": 2.1051922039904475e-05, + "loss": 0.0, + "step": 15031 + }, + { + "epoch": 3.47, + "learning_rate": 2.1049996148216626e-05, + "loss": 0.0, + "step": 15032 + }, + { + "epoch": 3.47, + "learning_rate": 2.1048070256528774e-05, + "loss": 0.0, + "step": 15033 + }, + { + "epoch": 3.47, + "learning_rate": 2.104614436484092e-05, + "loss": 0.0, + "step": 15034 + }, + { + "epoch": 3.47, + "learning_rate": 2.1044218473153072e-05, + "loss": 0.0, + "step": 15035 + }, + { + "epoch": 3.47, + "learning_rate": 2.104229258146522e-05, + "loss": 0.0, + "step": 15036 + }, + { + "epoch": 3.48, + "learning_rate": 2.104036668977737e-05, + "loss": 0.0001, + "step": 15037 + }, + { + "epoch": 3.48, + "learning_rate": 2.1038440798089518e-05, + "loss": 0.0, + "step": 15038 + }, + { + "epoch": 3.48, + "learning_rate": 2.1036514906401665e-05, + "loss": 0.0002, + "step": 15039 + }, + { + "epoch": 3.48, + "learning_rate": 2.1034589014713813e-05, + "loss": 0.0, + "step": 15040 + }, + { + "epoch": 3.48, + "learning_rate": 2.103266312302596e-05, + "loss": 0.0, + "step": 15041 + }, + { + "epoch": 3.48, + "learning_rate": 2.103073723133811e-05, + "loss": 0.0, + "step": 15042 + }, + { + "epoch": 3.48, + "learning_rate": 2.102881133965026e-05, + "loss": 0.0, + "step": 15043 + }, + { + "epoch": 3.48, + "learning_rate": 2.1026885447962406e-05, + "loss": 0.0, + "step": 15044 + }, + { + "epoch": 3.48, + "learning_rate": 2.1024959556274557e-05, + "loss": 0.0, + "step": 15045 + }, + { + "epoch": 3.48, + "learning_rate": 2.1023033664586704e-05, + "loss": 0.0, + "step": 15046 + }, + { + "epoch": 3.48, + "learning_rate": 2.1021107772898855e-05, + "loss": 0.0, + "step": 15047 + }, + { + "epoch": 3.48, + "learning_rate": 2.1019181881211003e-05, + "loss": 0.0, + "step": 15048 + }, + { + "epoch": 3.48, + "learning_rate": 2.101725598952315e-05, + "loss": 0.0, + "step": 15049 + }, + { + "epoch": 3.48, + "learning_rate": 2.1015330097835298e-05, + "loss": 0.0, + "step": 15050 + }, + { + "epoch": 3.48, + "learning_rate": 2.1013404206147445e-05, + "loss": 0.0, + "step": 15051 + }, + { + "epoch": 3.48, + "learning_rate": 2.1011478314459596e-05, + "loss": 0.0, + "step": 15052 + }, + { + "epoch": 3.48, + "learning_rate": 2.1009552422771744e-05, + "loss": 0.0, + "step": 15053 + }, + { + "epoch": 3.48, + "learning_rate": 2.1007626531083894e-05, + "loss": 0.0, + "step": 15054 + }, + { + "epoch": 3.48, + "learning_rate": 2.1005700639396042e-05, + "loss": 0.0, + "step": 15055 + }, + { + "epoch": 3.48, + "learning_rate": 2.100377474770819e-05, + "loss": 0.0, + "step": 15056 + }, + { + "epoch": 3.48, + "learning_rate": 2.100184885602034e-05, + "loss": 0.0, + "step": 15057 + }, + { + "epoch": 3.48, + "learning_rate": 2.0999922964332484e-05, + "loss": 0.0, + "step": 15058 + }, + { + "epoch": 3.48, + "learning_rate": 2.0997997072644635e-05, + "loss": 0.0, + "step": 15059 + }, + { + "epoch": 3.48, + "learning_rate": 2.0996071180956783e-05, + "loss": 0.0, + "step": 15060 + }, + { + "epoch": 3.48, + "learning_rate": 2.0994145289268934e-05, + "loss": 0.0, + "step": 15061 + }, + { + "epoch": 3.48, + "learning_rate": 2.099221939758108e-05, + "loss": 0.0001, + "step": 15062 + }, + { + "epoch": 3.48, + "learning_rate": 2.099029350589323e-05, + "loss": 0.0, + "step": 15063 + }, + { + "epoch": 3.48, + "learning_rate": 2.098836761420538e-05, + "loss": 0.0, + "step": 15064 + }, + { + "epoch": 3.48, + "learning_rate": 2.0986441722517527e-05, + "loss": 0.0, + "step": 15065 + }, + { + "epoch": 3.48, + "learning_rate": 2.0984515830829678e-05, + "loss": 0.0, + "step": 15066 + }, + { + "epoch": 3.48, + "learning_rate": 2.0982589939141822e-05, + "loss": 0.0, + "step": 15067 + }, + { + "epoch": 3.48, + "learning_rate": 2.0980664047453973e-05, + "loss": 0.0, + "step": 15068 + }, + { + "epoch": 3.48, + "learning_rate": 2.097873815576612e-05, + "loss": 0.0004, + "step": 15069 + }, + { + "epoch": 3.48, + "learning_rate": 2.0976812264078268e-05, + "loss": 0.0, + "step": 15070 + }, + { + "epoch": 3.48, + "learning_rate": 2.097488637239042e-05, + "loss": 0.0, + "step": 15071 + }, + { + "epoch": 3.48, + "learning_rate": 2.0972960480702566e-05, + "loss": 0.0, + "step": 15072 + }, + { + "epoch": 3.48, + "learning_rate": 2.0971034589014717e-05, + "loss": 0.0001, + "step": 15073 + }, + { + "epoch": 3.48, + "learning_rate": 2.0969108697326864e-05, + "loss": 0.0, + "step": 15074 + }, + { + "epoch": 3.48, + "learning_rate": 2.0967182805639012e-05, + "loss": 0.0, + "step": 15075 + }, + { + "epoch": 3.48, + "learning_rate": 2.0965256913951163e-05, + "loss": 0.0, + "step": 15076 + }, + { + "epoch": 3.48, + "learning_rate": 2.0963331022263307e-05, + "loss": 0.0, + "step": 15077 + }, + { + "epoch": 3.48, + "learning_rate": 2.0961405130575458e-05, + "loss": 0.0, + "step": 15078 + }, + { + "epoch": 3.48, + "learning_rate": 2.0959479238887605e-05, + "loss": 0.0, + "step": 15079 + }, + { + "epoch": 3.49, + "learning_rate": 2.0957553347199756e-05, + "loss": 0.0, + "step": 15080 + }, + { + "epoch": 3.49, + "learning_rate": 2.0955627455511903e-05, + "loss": 0.0, + "step": 15081 + }, + { + "epoch": 3.49, + "learning_rate": 2.095370156382405e-05, + "loss": 0.0, + "step": 15082 + }, + { + "epoch": 3.49, + "learning_rate": 2.0951775672136202e-05, + "loss": 0.0, + "step": 15083 + }, + { + "epoch": 3.49, + "learning_rate": 2.094984978044835e-05, + "loss": 0.0, + "step": 15084 + }, + { + "epoch": 3.49, + "learning_rate": 2.0947923888760497e-05, + "loss": 0.0, + "step": 15085 + }, + { + "epoch": 3.49, + "learning_rate": 2.0945997997072644e-05, + "loss": 0.0, + "step": 15086 + }, + { + "epoch": 3.49, + "learning_rate": 2.094407210538479e-05, + "loss": 0.0, + "step": 15087 + }, + { + "epoch": 3.49, + "learning_rate": 2.0942146213696943e-05, + "loss": 0.0, + "step": 15088 + }, + { + "epoch": 3.49, + "learning_rate": 2.094022032200909e-05, + "loss": 0.0, + "step": 15089 + }, + { + "epoch": 3.49, + "learning_rate": 2.093829443032124e-05, + "loss": 0.0, + "step": 15090 + }, + { + "epoch": 3.49, + "learning_rate": 2.093636853863339e-05, + "loss": 0.0001, + "step": 15091 + }, + { + "epoch": 3.49, + "learning_rate": 2.0934442646945536e-05, + "loss": 0.0, + "step": 15092 + }, + { + "epoch": 3.49, + "learning_rate": 2.0932516755257687e-05, + "loss": 0.0, + "step": 15093 + }, + { + "epoch": 3.49, + "learning_rate": 2.0930590863569834e-05, + "loss": 0.0, + "step": 15094 + }, + { + "epoch": 3.49, + "learning_rate": 2.092866497188198e-05, + "loss": 0.0, + "step": 15095 + }, + { + "epoch": 3.49, + "learning_rate": 2.092673908019413e-05, + "loss": 0.0001, + "step": 15096 + }, + { + "epoch": 3.49, + "learning_rate": 2.092481318850628e-05, + "loss": 0.0, + "step": 15097 + }, + { + "epoch": 3.49, + "learning_rate": 2.0922887296818427e-05, + "loss": 0.0, + "step": 15098 + }, + { + "epoch": 3.49, + "learning_rate": 2.0920961405130575e-05, + "loss": 0.0, + "step": 15099 + }, + { + "epoch": 3.49, + "learning_rate": 2.0919035513442726e-05, + "loss": 0.0, + "step": 15100 + }, + { + "epoch": 3.49, + "learning_rate": 2.0917109621754873e-05, + "loss": 0.0002, + "step": 15101 + }, + { + "epoch": 3.49, + "learning_rate": 2.0915183730067024e-05, + "loss": 0.0001, + "step": 15102 + }, + { + "epoch": 3.49, + "learning_rate": 2.091325783837917e-05, + "loss": 0.0, + "step": 15103 + }, + { + "epoch": 3.49, + "learning_rate": 2.091133194669132e-05, + "loss": 0.0, + "step": 15104 + }, + { + "epoch": 3.49, + "learning_rate": 2.0909406055003467e-05, + "loss": 0.0, + "step": 15105 + }, + { + "epoch": 3.49, + "learning_rate": 2.0907480163315614e-05, + "loss": 0.0, + "step": 15106 + }, + { + "epoch": 3.49, + "learning_rate": 2.0905554271627765e-05, + "loss": 0.0, + "step": 15107 + }, + { + "epoch": 3.49, + "learning_rate": 2.0903628379939912e-05, + "loss": 0.0003, + "step": 15108 + }, + { + "epoch": 3.49, + "learning_rate": 2.0901702488252063e-05, + "loss": 0.0001, + "step": 15109 + }, + { + "epoch": 3.49, + "learning_rate": 2.089977659656421e-05, + "loss": 0.0, + "step": 15110 + }, + { + "epoch": 3.49, + "learning_rate": 2.0897850704876358e-05, + "loss": 0.0, + "step": 15111 + }, + { + "epoch": 3.49, + "learning_rate": 2.089592481318851e-05, + "loss": 0.0, + "step": 15112 + }, + { + "epoch": 3.49, + "learning_rate": 2.0893998921500657e-05, + "loss": 0.0, + "step": 15113 + }, + { + "epoch": 3.49, + "learning_rate": 2.0892073029812804e-05, + "loss": 0.0001, + "step": 15114 + }, + { + "epoch": 3.49, + "learning_rate": 2.089014713812495e-05, + "loss": 0.0, + "step": 15115 + }, + { + "epoch": 3.49, + "learning_rate": 2.0888221246437102e-05, + "loss": 0.0, + "step": 15116 + }, + { + "epoch": 3.49, + "learning_rate": 2.088629535474925e-05, + "loss": 0.0, + "step": 15117 + }, + { + "epoch": 3.49, + "learning_rate": 2.0884369463061397e-05, + "loss": 0.0002, + "step": 15118 + }, + { + "epoch": 3.49, + "learning_rate": 2.0882443571373548e-05, + "loss": 0.0, + "step": 15119 + }, + { + "epoch": 3.49, + "learning_rate": 2.0880517679685696e-05, + "loss": 0.0, + "step": 15120 + }, + { + "epoch": 3.49, + "learning_rate": 2.0878591787997843e-05, + "loss": 0.0, + "step": 15121 + }, + { + "epoch": 3.49, + "learning_rate": 2.0876665896309994e-05, + "loss": 0.0, + "step": 15122 + }, + { + "epoch": 3.5, + "learning_rate": 2.0874740004622138e-05, + "loss": 0.0, + "step": 15123 + }, + { + "epoch": 3.5, + "learning_rate": 2.087281411293429e-05, + "loss": 0.0, + "step": 15124 + }, + { + "epoch": 3.5, + "learning_rate": 2.0870888221246436e-05, + "loss": 0.0, + "step": 15125 + }, + { + "epoch": 3.5, + "learning_rate": 2.0868962329558587e-05, + "loss": 0.0, + "step": 15126 + }, + { + "epoch": 3.5, + "learning_rate": 2.0867036437870735e-05, + "loss": 0.0001, + "step": 15127 + }, + { + "epoch": 3.5, + "learning_rate": 2.0865110546182882e-05, + "loss": 0.0, + "step": 15128 + }, + { + "epoch": 3.5, + "learning_rate": 2.0863184654495033e-05, + "loss": 0.0, + "step": 15129 + }, + { + "epoch": 3.5, + "learning_rate": 2.086125876280718e-05, + "loss": 0.0, + "step": 15130 + }, + { + "epoch": 3.5, + "learning_rate": 2.085933287111933e-05, + "loss": 0.0, + "step": 15131 + }, + { + "epoch": 3.5, + "learning_rate": 2.085740697943148e-05, + "loss": 0.0, + "step": 15132 + }, + { + "epoch": 3.5, + "learning_rate": 2.0855481087743626e-05, + "loss": 0.0, + "step": 15133 + }, + { + "epoch": 3.5, + "learning_rate": 2.0853555196055774e-05, + "loss": 0.0, + "step": 15134 + }, + { + "epoch": 3.5, + "learning_rate": 2.085162930436792e-05, + "loss": 0.0001, + "step": 15135 + }, + { + "epoch": 3.5, + "learning_rate": 2.0849703412680072e-05, + "loss": 0.0, + "step": 15136 + }, + { + "epoch": 3.5, + "learning_rate": 2.084777752099222e-05, + "loss": 0.0001, + "step": 15137 + }, + { + "epoch": 3.5, + "learning_rate": 2.084585162930437e-05, + "loss": 0.0001, + "step": 15138 + }, + { + "epoch": 3.5, + "learning_rate": 2.0843925737616518e-05, + "loss": 0.0, + "step": 15139 + }, + { + "epoch": 3.5, + "learning_rate": 2.0841999845928666e-05, + "loss": 0.0, + "step": 15140 + }, + { + "epoch": 3.5, + "learning_rate": 2.0840073954240816e-05, + "loss": 0.0, + "step": 15141 + }, + { + "epoch": 3.5, + "learning_rate": 2.083814806255296e-05, + "loss": 0.0, + "step": 15142 + }, + { + "epoch": 3.5, + "learning_rate": 2.083622217086511e-05, + "loss": 0.0, + "step": 15143 + }, + { + "epoch": 3.5, + "learning_rate": 2.083429627917726e-05, + "loss": 0.0, + "step": 15144 + }, + { + "epoch": 3.5, + "learning_rate": 2.083237038748941e-05, + "loss": 0.0, + "step": 15145 + }, + { + "epoch": 3.5, + "learning_rate": 2.0830444495801557e-05, + "loss": 0.0, + "step": 15146 + }, + { + "epoch": 3.5, + "learning_rate": 2.0828518604113705e-05, + "loss": 0.0, + "step": 15147 + }, + { + "epoch": 3.5, + "learning_rate": 2.0826592712425856e-05, + "loss": 0.0, + "step": 15148 + }, + { + "epoch": 3.5, + "learning_rate": 2.0824666820738003e-05, + "loss": 0.0, + "step": 15149 + }, + { + "epoch": 3.5, + "learning_rate": 2.0822740929050154e-05, + "loss": 0.0, + "step": 15150 + }, + { + "epoch": 3.5, + "learning_rate": 2.0820815037362298e-05, + "loss": 0.0, + "step": 15151 + }, + { + "epoch": 3.5, + "learning_rate": 2.081888914567445e-05, + "loss": 0.0, + "step": 15152 + }, + { + "epoch": 3.5, + "learning_rate": 2.0816963253986596e-05, + "loss": 0.0, + "step": 15153 + }, + { + "epoch": 3.5, + "learning_rate": 2.0815037362298744e-05, + "loss": 0.0, + "step": 15154 + }, + { + "epoch": 3.5, + "learning_rate": 2.0813111470610895e-05, + "loss": 0.0, + "step": 15155 + }, + { + "epoch": 3.5, + "learning_rate": 2.0811185578923042e-05, + "loss": 0.0, + "step": 15156 + }, + { + "epoch": 3.5, + "learning_rate": 2.0809259687235193e-05, + "loss": 0.0, + "step": 15157 + }, + { + "epoch": 3.5, + "learning_rate": 2.080733379554734e-05, + "loss": 0.0, + "step": 15158 + }, + { + "epoch": 3.5, + "learning_rate": 2.0805407903859488e-05, + "loss": 0.0, + "step": 15159 + }, + { + "epoch": 3.5, + "learning_rate": 2.080348201217164e-05, + "loss": 0.0, + "step": 15160 + }, + { + "epoch": 3.5, + "learning_rate": 2.0801556120483783e-05, + "loss": 0.0, + "step": 15161 + }, + { + "epoch": 3.5, + "learning_rate": 2.0799630228795934e-05, + "loss": 0.0, + "step": 15162 + }, + { + "epoch": 3.5, + "learning_rate": 2.079770433710808e-05, + "loss": 0.0, + "step": 15163 + }, + { + "epoch": 3.5, + "learning_rate": 2.079577844542023e-05, + "loss": 0.0, + "step": 15164 + }, + { + "epoch": 3.5, + "learning_rate": 2.079385255373238e-05, + "loss": 0.0003, + "step": 15165 + }, + { + "epoch": 3.5, + "learning_rate": 2.0791926662044527e-05, + "loss": 0.0, + "step": 15166 + }, + { + "epoch": 3.51, + "learning_rate": 2.0790000770356678e-05, + "loss": 0.0, + "step": 15167 + }, + { + "epoch": 3.51, + "learning_rate": 2.0788074878668825e-05, + "loss": 0.0, + "step": 15168 + }, + { + "epoch": 3.51, + "learning_rate": 2.0786148986980973e-05, + "loss": 0.0, + "step": 15169 + }, + { + "epoch": 3.51, + "learning_rate": 2.078422309529312e-05, + "loss": 0.0, + "step": 15170 + }, + { + "epoch": 3.51, + "learning_rate": 2.0782297203605268e-05, + "loss": 0.0, + "step": 15171 + }, + { + "epoch": 3.51, + "learning_rate": 2.078037131191742e-05, + "loss": 0.0, + "step": 15172 + }, + { + "epoch": 3.51, + "learning_rate": 2.0778445420229566e-05, + "loss": 0.0001, + "step": 15173 + }, + { + "epoch": 3.51, + "learning_rate": 2.0776519528541717e-05, + "loss": 0.0, + "step": 15174 + }, + { + "epoch": 3.51, + "learning_rate": 2.0774593636853865e-05, + "loss": 0.0, + "step": 15175 + }, + { + "epoch": 3.51, + "learning_rate": 2.0772667745166012e-05, + "loss": 0.0, + "step": 15176 + }, + { + "epoch": 3.51, + "learning_rate": 2.0770741853478163e-05, + "loss": 0.0, + "step": 15177 + }, + { + "epoch": 3.51, + "learning_rate": 2.076881596179031e-05, + "loss": 0.0, + "step": 15178 + }, + { + "epoch": 3.51, + "learning_rate": 2.0766890070102458e-05, + "loss": 0.0, + "step": 15179 + }, + { + "epoch": 3.51, + "learning_rate": 2.0764964178414605e-05, + "loss": 0.0, + "step": 15180 + }, + { + "epoch": 3.51, + "learning_rate": 2.0763038286726756e-05, + "loss": 0.0, + "step": 15181 + }, + { + "epoch": 3.51, + "learning_rate": 2.0761112395038904e-05, + "loss": 0.0, + "step": 15182 + }, + { + "epoch": 3.51, + "learning_rate": 2.075918650335105e-05, + "loss": 0.0, + "step": 15183 + }, + { + "epoch": 3.51, + "learning_rate": 2.0757260611663202e-05, + "loss": 0.0, + "step": 15184 + }, + { + "epoch": 3.51, + "learning_rate": 2.075533471997535e-05, + "loss": 0.0, + "step": 15185 + }, + { + "epoch": 3.51, + "learning_rate": 2.07534088282875e-05, + "loss": 0.0, + "step": 15186 + }, + { + "epoch": 3.51, + "learning_rate": 2.0751482936599648e-05, + "loss": 0.0, + "step": 15187 + }, + { + "epoch": 3.51, + "learning_rate": 2.0749557044911795e-05, + "loss": 0.0, + "step": 15188 + }, + { + "epoch": 3.51, + "learning_rate": 2.0747631153223943e-05, + "loss": 0.0, + "step": 15189 + }, + { + "epoch": 3.51, + "learning_rate": 2.074570526153609e-05, + "loss": 0.0, + "step": 15190 + }, + { + "epoch": 3.51, + "learning_rate": 2.074377936984824e-05, + "loss": 0.0, + "step": 15191 + }, + { + "epoch": 3.51, + "learning_rate": 2.074185347816039e-05, + "loss": 0.0, + "step": 15192 + }, + { + "epoch": 3.51, + "learning_rate": 2.073992758647254e-05, + "loss": 0.0, + "step": 15193 + }, + { + "epoch": 3.51, + "learning_rate": 2.0738001694784687e-05, + "loss": 0.0, + "step": 15194 + }, + { + "epoch": 3.51, + "learning_rate": 2.0736075803096834e-05, + "loss": 0.0, + "step": 15195 + }, + { + "epoch": 3.51, + "learning_rate": 2.0734149911408985e-05, + "loss": 0.0, + "step": 15196 + }, + { + "epoch": 3.51, + "learning_rate": 2.0732224019721133e-05, + "loss": 0.0, + "step": 15197 + }, + { + "epoch": 3.51, + "learning_rate": 2.073029812803328e-05, + "loss": 0.0, + "step": 15198 + }, + { + "epoch": 3.51, + "learning_rate": 2.0728372236345428e-05, + "loss": 0.0, + "step": 15199 + }, + { + "epoch": 3.51, + "learning_rate": 2.0726446344657575e-05, + "loss": 0.0, + "step": 15200 + }, + { + "epoch": 3.51, + "learning_rate": 2.0724520452969726e-05, + "loss": 0.0, + "step": 15201 + }, + { + "epoch": 3.51, + "learning_rate": 2.0722594561281874e-05, + "loss": 0.0, + "step": 15202 + }, + { + "epoch": 3.51, + "learning_rate": 2.0720668669594024e-05, + "loss": 0.0, + "step": 15203 + }, + { + "epoch": 3.51, + "learning_rate": 2.0718742777906172e-05, + "loss": 0.0, + "step": 15204 + }, + { + "epoch": 3.51, + "learning_rate": 2.071681688621832e-05, + "loss": 0.0, + "step": 15205 + }, + { + "epoch": 3.51, + "learning_rate": 2.071489099453047e-05, + "loss": 0.0, + "step": 15206 + }, + { + "epoch": 3.51, + "learning_rate": 2.0712965102842614e-05, + "loss": 0.0002, + "step": 15207 + }, + { + "epoch": 3.51, + "learning_rate": 2.0711039211154765e-05, + "loss": 0.0, + "step": 15208 + }, + { + "epoch": 3.51, + "learning_rate": 2.0709113319466913e-05, + "loss": 0.0, + "step": 15209 + }, + { + "epoch": 3.52, + "learning_rate": 2.0707187427779064e-05, + "loss": 0.0, + "step": 15210 + }, + { + "epoch": 3.52, + "learning_rate": 2.070526153609121e-05, + "loss": 0.0, + "step": 15211 + }, + { + "epoch": 3.52, + "learning_rate": 2.070333564440336e-05, + "loss": 0.0, + "step": 15212 + }, + { + "epoch": 3.52, + "learning_rate": 2.070140975271551e-05, + "loss": 0.0002, + "step": 15213 + }, + { + "epoch": 3.52, + "learning_rate": 2.0699483861027657e-05, + "loss": 0.0, + "step": 15214 + }, + { + "epoch": 3.52, + "learning_rate": 2.0697557969339808e-05, + "loss": 0.0, + "step": 15215 + }, + { + "epoch": 3.52, + "learning_rate": 2.0695632077651955e-05, + "loss": 0.0, + "step": 15216 + }, + { + "epoch": 3.52, + "learning_rate": 2.0693706185964103e-05, + "loss": 0.0, + "step": 15217 + }, + { + "epoch": 3.52, + "learning_rate": 2.069178029427625e-05, + "loss": 0.0, + "step": 15218 + }, + { + "epoch": 3.52, + "learning_rate": 2.0689854402588398e-05, + "loss": 0.0, + "step": 15219 + }, + { + "epoch": 3.52, + "learning_rate": 2.068792851090055e-05, + "loss": 0.0, + "step": 15220 + }, + { + "epoch": 3.52, + "learning_rate": 2.0686002619212696e-05, + "loss": 0.0, + "step": 15221 + }, + { + "epoch": 3.52, + "learning_rate": 2.0684076727524847e-05, + "loss": 0.0, + "step": 15222 + }, + { + "epoch": 3.52, + "learning_rate": 2.0682150835836994e-05, + "loss": 0.0001, + "step": 15223 + }, + { + "epoch": 3.52, + "learning_rate": 2.0680224944149142e-05, + "loss": 0.0, + "step": 15224 + }, + { + "epoch": 3.52, + "learning_rate": 2.0678299052461293e-05, + "loss": 0.0, + "step": 15225 + }, + { + "epoch": 3.52, + "learning_rate": 2.0676373160773437e-05, + "loss": 0.0, + "step": 15226 + }, + { + "epoch": 3.52, + "learning_rate": 2.0674447269085588e-05, + "loss": 0.0, + "step": 15227 + }, + { + "epoch": 3.52, + "learning_rate": 2.0672521377397735e-05, + "loss": 0.0, + "step": 15228 + }, + { + "epoch": 3.52, + "learning_rate": 2.0670595485709886e-05, + "loss": 0.0, + "step": 15229 + }, + { + "epoch": 3.52, + "learning_rate": 2.0668669594022033e-05, + "loss": 0.0, + "step": 15230 + }, + { + "epoch": 3.52, + "learning_rate": 2.066674370233418e-05, + "loss": 0.0, + "step": 15231 + }, + { + "epoch": 3.52, + "learning_rate": 2.0664817810646332e-05, + "loss": 0.0, + "step": 15232 + }, + { + "epoch": 3.52, + "learning_rate": 2.066289191895848e-05, + "loss": 0.0, + "step": 15233 + }, + { + "epoch": 3.52, + "learning_rate": 2.066096602727063e-05, + "loss": 0.0, + "step": 15234 + }, + { + "epoch": 3.52, + "learning_rate": 2.0659040135582774e-05, + "loss": 0.0, + "step": 15235 + }, + { + "epoch": 3.52, + "learning_rate": 2.0657114243894922e-05, + "loss": 0.0, + "step": 15236 + }, + { + "epoch": 3.52, + "learning_rate": 2.0655188352207073e-05, + "loss": 0.0, + "step": 15237 + }, + { + "epoch": 3.52, + "learning_rate": 2.065326246051922e-05, + "loss": 0.0, + "step": 15238 + }, + { + "epoch": 3.52, + "learning_rate": 2.065133656883137e-05, + "loss": 0.0001, + "step": 15239 + }, + { + "epoch": 3.52, + "learning_rate": 2.064941067714352e-05, + "loss": 0.0001, + "step": 15240 + }, + { + "epoch": 3.52, + "learning_rate": 2.0647484785455666e-05, + "loss": 0.0, + "step": 15241 + }, + { + "epoch": 3.52, + "learning_rate": 2.0645558893767817e-05, + "loss": 0.0, + "step": 15242 + }, + { + "epoch": 3.52, + "learning_rate": 2.0643633002079964e-05, + "loss": 0.0, + "step": 15243 + }, + { + "epoch": 3.52, + "learning_rate": 2.0641707110392115e-05, + "loss": 0.0, + "step": 15244 + }, + { + "epoch": 3.52, + "learning_rate": 2.063978121870426e-05, + "loss": 0.0, + "step": 15245 + }, + { + "epoch": 3.52, + "learning_rate": 2.063785532701641e-05, + "loss": 0.0, + "step": 15246 + }, + { + "epoch": 3.52, + "learning_rate": 2.0635929435328558e-05, + "loss": 0.0, + "step": 15247 + }, + { + "epoch": 3.52, + "learning_rate": 2.0634003543640705e-05, + "loss": 0.0, + "step": 15248 + }, + { + "epoch": 3.52, + "learning_rate": 2.0632077651952856e-05, + "loss": 0.0, + "step": 15249 + }, + { + "epoch": 3.52, + "learning_rate": 2.0630151760265003e-05, + "loss": 0.0001, + "step": 15250 + }, + { + "epoch": 3.52, + "learning_rate": 2.0628225868577154e-05, + "loss": 0.0, + "step": 15251 + }, + { + "epoch": 3.52, + "learning_rate": 2.0626299976889302e-05, + "loss": 0.0, + "step": 15252 + }, + { + "epoch": 3.53, + "learning_rate": 2.062437408520145e-05, + "loss": 0.0, + "step": 15253 + }, + { + "epoch": 3.53, + "learning_rate": 2.0622448193513597e-05, + "loss": 0.0, + "step": 15254 + }, + { + "epoch": 3.53, + "learning_rate": 2.0620522301825744e-05, + "loss": 0.0, + "step": 15255 + }, + { + "epoch": 3.53, + "learning_rate": 2.0618596410137895e-05, + "loss": 0.0, + "step": 15256 + }, + { + "epoch": 3.53, + "learning_rate": 2.0616670518450042e-05, + "loss": 0.0, + "step": 15257 + }, + { + "epoch": 3.53, + "learning_rate": 2.0614744626762193e-05, + "loss": 0.0, + "step": 15258 + }, + { + "epoch": 3.53, + "learning_rate": 2.061281873507434e-05, + "loss": 0.0, + "step": 15259 + }, + { + "epoch": 3.53, + "learning_rate": 2.0610892843386488e-05, + "loss": 0.0001, + "step": 15260 + }, + { + "epoch": 3.53, + "learning_rate": 2.060896695169864e-05, + "loss": 0.0, + "step": 15261 + }, + { + "epoch": 3.53, + "learning_rate": 2.0607041060010787e-05, + "loss": 0.0, + "step": 15262 + }, + { + "epoch": 3.53, + "learning_rate": 2.0605115168322934e-05, + "loss": 0.0, + "step": 15263 + }, + { + "epoch": 3.53, + "learning_rate": 2.060318927663508e-05, + "loss": 0.0001, + "step": 15264 + }, + { + "epoch": 3.53, + "learning_rate": 2.0601263384947232e-05, + "loss": 0.0, + "step": 15265 + }, + { + "epoch": 3.53, + "learning_rate": 2.059933749325938e-05, + "loss": 0.0132, + "step": 15266 + }, + { + "epoch": 3.53, + "learning_rate": 2.0597411601571527e-05, + "loss": 0.0, + "step": 15267 + }, + { + "epoch": 3.53, + "learning_rate": 2.0595485709883678e-05, + "loss": 0.0, + "step": 15268 + }, + { + "epoch": 3.53, + "learning_rate": 2.0593559818195826e-05, + "loss": 0.0, + "step": 15269 + }, + { + "epoch": 3.53, + "learning_rate": 2.0591633926507977e-05, + "loss": 0.0001, + "step": 15270 + }, + { + "epoch": 3.53, + "learning_rate": 2.0589708034820124e-05, + "loss": 0.0, + "step": 15271 + }, + { + "epoch": 3.53, + "learning_rate": 2.058778214313227e-05, + "loss": 0.0, + "step": 15272 + }, + { + "epoch": 3.53, + "learning_rate": 2.058585625144442e-05, + "loss": 0.0, + "step": 15273 + }, + { + "epoch": 3.53, + "learning_rate": 2.0583930359756567e-05, + "loss": 0.0, + "step": 15274 + }, + { + "epoch": 3.53, + "learning_rate": 2.0582004468068717e-05, + "loss": 0.0, + "step": 15275 + }, + { + "epoch": 3.53, + "learning_rate": 2.0580078576380865e-05, + "loss": 0.0, + "step": 15276 + }, + { + "epoch": 3.53, + "learning_rate": 2.0578152684693012e-05, + "loss": 0.0, + "step": 15277 + }, + { + "epoch": 3.53, + "learning_rate": 2.0576226793005163e-05, + "loss": 0.0, + "step": 15278 + }, + { + "epoch": 3.53, + "learning_rate": 2.057430090131731e-05, + "loss": 0.0, + "step": 15279 + }, + { + "epoch": 3.53, + "learning_rate": 2.057237500962946e-05, + "loss": 0.0, + "step": 15280 + }, + { + "epoch": 3.53, + "learning_rate": 2.057044911794161e-05, + "loss": 0.0, + "step": 15281 + }, + { + "epoch": 3.53, + "learning_rate": 2.0568523226253757e-05, + "loss": 0.0, + "step": 15282 + }, + { + "epoch": 3.53, + "learning_rate": 2.0566597334565904e-05, + "loss": 0.0, + "step": 15283 + }, + { + "epoch": 3.53, + "learning_rate": 2.056467144287805e-05, + "loss": 0.0, + "step": 15284 + }, + { + "epoch": 3.53, + "learning_rate": 2.0562745551190202e-05, + "loss": 0.0, + "step": 15285 + }, + { + "epoch": 3.53, + "learning_rate": 2.056081965950235e-05, + "loss": 0.0, + "step": 15286 + }, + { + "epoch": 3.53, + "learning_rate": 2.05588937678145e-05, + "loss": 0.0, + "step": 15287 + }, + { + "epoch": 3.53, + "learning_rate": 2.0556967876126648e-05, + "loss": 0.0, + "step": 15288 + }, + { + "epoch": 3.53, + "learning_rate": 2.0555041984438796e-05, + "loss": 0.0, + "step": 15289 + }, + { + "epoch": 3.53, + "learning_rate": 2.0553116092750947e-05, + "loss": 0.0, + "step": 15290 + }, + { + "epoch": 3.53, + "learning_rate": 2.055119020106309e-05, + "loss": 0.0, + "step": 15291 + }, + { + "epoch": 3.53, + "learning_rate": 2.054926430937524e-05, + "loss": 0.0008, + "step": 15292 + }, + { + "epoch": 3.53, + "learning_rate": 2.054733841768739e-05, + "loss": 0.0, + "step": 15293 + }, + { + "epoch": 3.53, + "learning_rate": 2.054541252599954e-05, + "loss": 0.0, + "step": 15294 + }, + { + "epoch": 3.53, + "learning_rate": 2.0543486634311687e-05, + "loss": 0.0, + "step": 15295 + }, + { + "epoch": 3.54, + "learning_rate": 2.0541560742623835e-05, + "loss": 0.0, + "step": 15296 + }, + { + "epoch": 3.54, + "learning_rate": 2.0539634850935986e-05, + "loss": 0.0, + "step": 15297 + }, + { + "epoch": 3.54, + "learning_rate": 2.0537708959248133e-05, + "loss": 0.0, + "step": 15298 + }, + { + "epoch": 3.54, + "learning_rate": 2.0535783067560284e-05, + "loss": 0.0, + "step": 15299 + }, + { + "epoch": 3.54, + "learning_rate": 2.053385717587243e-05, + "loss": 0.0, + "step": 15300 + }, + { + "epoch": 3.54, + "learning_rate": 2.053193128418458e-05, + "loss": 0.0, + "step": 15301 + }, + { + "epoch": 3.54, + "learning_rate": 2.0530005392496726e-05, + "loss": 0.0, + "step": 15302 + }, + { + "epoch": 3.54, + "learning_rate": 2.0528079500808874e-05, + "loss": 0.0, + "step": 15303 + }, + { + "epoch": 3.54, + "learning_rate": 2.0526153609121025e-05, + "loss": 0.0, + "step": 15304 + }, + { + "epoch": 3.54, + "learning_rate": 2.0524227717433172e-05, + "loss": 0.0, + "step": 15305 + }, + { + "epoch": 3.54, + "learning_rate": 2.0522301825745323e-05, + "loss": 0.0, + "step": 15306 + }, + { + "epoch": 3.54, + "learning_rate": 2.052037593405747e-05, + "loss": 0.0, + "step": 15307 + }, + { + "epoch": 3.54, + "learning_rate": 2.0518450042369618e-05, + "loss": 0.0, + "step": 15308 + }, + { + "epoch": 3.54, + "learning_rate": 2.051652415068177e-05, + "loss": 0.0, + "step": 15309 + }, + { + "epoch": 3.54, + "learning_rate": 2.0514598258993913e-05, + "loss": 0.0, + "step": 15310 + }, + { + "epoch": 3.54, + "learning_rate": 2.0512672367306064e-05, + "loss": 0.0, + "step": 15311 + }, + { + "epoch": 3.54, + "learning_rate": 2.051074647561821e-05, + "loss": 0.0001, + "step": 15312 + }, + { + "epoch": 3.54, + "learning_rate": 2.050882058393036e-05, + "loss": 0.0001, + "step": 15313 + }, + { + "epoch": 3.54, + "learning_rate": 2.050689469224251e-05, + "loss": 0.0, + "step": 15314 + }, + { + "epoch": 3.54, + "learning_rate": 2.0504968800554657e-05, + "loss": 0.0, + "step": 15315 + }, + { + "epoch": 3.54, + "learning_rate": 2.0503042908866808e-05, + "loss": 0.0, + "step": 15316 + }, + { + "epoch": 3.54, + "learning_rate": 2.0501117017178956e-05, + "loss": 0.0, + "step": 15317 + }, + { + "epoch": 3.54, + "learning_rate": 2.0499191125491103e-05, + "loss": 0.0, + "step": 15318 + }, + { + "epoch": 3.54, + "learning_rate": 2.049726523380325e-05, + "loss": 0.0, + "step": 15319 + }, + { + "epoch": 3.54, + "learning_rate": 2.0495339342115398e-05, + "loss": 0.0, + "step": 15320 + }, + { + "epoch": 3.54, + "learning_rate": 2.049341345042755e-05, + "loss": 0.0, + "step": 15321 + }, + { + "epoch": 3.54, + "learning_rate": 2.0491487558739696e-05, + "loss": 0.0, + "step": 15322 + }, + { + "epoch": 3.54, + "learning_rate": 2.0489561667051847e-05, + "loss": 0.0, + "step": 15323 + }, + { + "epoch": 3.54, + "learning_rate": 2.0487635775363995e-05, + "loss": 0.0, + "step": 15324 + }, + { + "epoch": 3.54, + "learning_rate": 2.0485709883676142e-05, + "loss": 0.0001, + "step": 15325 + }, + { + "epoch": 3.54, + "learning_rate": 2.0483783991988293e-05, + "loss": 0.0, + "step": 15326 + }, + { + "epoch": 3.54, + "learning_rate": 2.048185810030044e-05, + "loss": 0.0001, + "step": 15327 + }, + { + "epoch": 3.54, + "learning_rate": 2.047993220861259e-05, + "loss": 0.0, + "step": 15328 + }, + { + "epoch": 3.54, + "learning_rate": 2.0478006316924735e-05, + "loss": 0.0, + "step": 15329 + }, + { + "epoch": 3.54, + "learning_rate": 2.0476080425236886e-05, + "loss": 0.0, + "step": 15330 + }, + { + "epoch": 3.54, + "learning_rate": 2.0474154533549034e-05, + "loss": 0.0, + "step": 15331 + }, + { + "epoch": 3.54, + "learning_rate": 2.047222864186118e-05, + "loss": 0.0001, + "step": 15332 + }, + { + "epoch": 3.54, + "learning_rate": 2.0470302750173332e-05, + "loss": 0.0, + "step": 15333 + }, + { + "epoch": 3.54, + "learning_rate": 2.046837685848548e-05, + "loss": 0.0, + "step": 15334 + }, + { + "epoch": 3.54, + "learning_rate": 2.046645096679763e-05, + "loss": 0.0, + "step": 15335 + }, + { + "epoch": 3.54, + "learning_rate": 2.0464525075109778e-05, + "loss": 0.0, + "step": 15336 + }, + { + "epoch": 3.54, + "learning_rate": 2.0462599183421925e-05, + "loss": 0.0, + "step": 15337 + }, + { + "epoch": 3.54, + "learning_rate": 2.0460673291734073e-05, + "loss": 0.0001, + "step": 15338 + }, + { + "epoch": 3.54, + "learning_rate": 2.045874740004622e-05, + "loss": 0.0, + "step": 15339 + }, + { + "epoch": 3.55, + "learning_rate": 2.045682150835837e-05, + "loss": 0.0, + "step": 15340 + }, + { + "epoch": 3.55, + "learning_rate": 2.045489561667052e-05, + "loss": 0.0, + "step": 15341 + }, + { + "epoch": 3.55, + "learning_rate": 2.045296972498267e-05, + "loss": 0.0, + "step": 15342 + }, + { + "epoch": 3.55, + "learning_rate": 2.0451043833294817e-05, + "loss": 0.0, + "step": 15343 + }, + { + "epoch": 3.55, + "learning_rate": 2.0449117941606965e-05, + "loss": 0.0, + "step": 15344 + }, + { + "epoch": 3.55, + "learning_rate": 2.0447192049919115e-05, + "loss": 0.0, + "step": 15345 + }, + { + "epoch": 3.55, + "learning_rate": 2.0445266158231263e-05, + "loss": 0.0, + "step": 15346 + }, + { + "epoch": 3.55, + "learning_rate": 2.044334026654341e-05, + "loss": 0.0, + "step": 15347 + }, + { + "epoch": 3.55, + "learning_rate": 2.0441414374855558e-05, + "loss": 0.0, + "step": 15348 + }, + { + "epoch": 3.55, + "learning_rate": 2.0439488483167705e-05, + "loss": 0.0, + "step": 15349 + }, + { + "epoch": 3.55, + "learning_rate": 2.0437562591479856e-05, + "loss": 0.0, + "step": 15350 + }, + { + "epoch": 3.55, + "learning_rate": 2.0435636699792004e-05, + "loss": 0.0001, + "step": 15351 + }, + { + "epoch": 3.55, + "learning_rate": 2.0433710808104155e-05, + "loss": 0.0002, + "step": 15352 + }, + { + "epoch": 3.55, + "learning_rate": 2.0431784916416302e-05, + "loss": 0.0, + "step": 15353 + }, + { + "epoch": 3.55, + "learning_rate": 2.042985902472845e-05, + "loss": 0.0, + "step": 15354 + }, + { + "epoch": 3.55, + "learning_rate": 2.04279331330406e-05, + "loss": 0.0, + "step": 15355 + }, + { + "epoch": 3.55, + "learning_rate": 2.0426007241352748e-05, + "loss": 0.0, + "step": 15356 + }, + { + "epoch": 3.55, + "learning_rate": 2.0424081349664895e-05, + "loss": 0.0, + "step": 15357 + }, + { + "epoch": 3.55, + "learning_rate": 2.0422155457977043e-05, + "loss": 0.0, + "step": 15358 + }, + { + "epoch": 3.55, + "learning_rate": 2.0420229566289194e-05, + "loss": 0.0, + "step": 15359 + }, + { + "epoch": 3.55, + "learning_rate": 2.041830367460134e-05, + "loss": 0.0, + "step": 15360 + }, + { + "epoch": 3.55, + "learning_rate": 2.041637778291349e-05, + "loss": 0.0, + "step": 15361 + }, + { + "epoch": 3.55, + "learning_rate": 2.041445189122564e-05, + "loss": 0.0, + "step": 15362 + }, + { + "epoch": 3.55, + "learning_rate": 2.0412525999537787e-05, + "loss": 0.0, + "step": 15363 + }, + { + "epoch": 3.55, + "learning_rate": 2.0410600107849938e-05, + "loss": 0.0, + "step": 15364 + }, + { + "epoch": 3.55, + "learning_rate": 2.0408674216162085e-05, + "loss": 0.0, + "step": 15365 + }, + { + "epoch": 3.55, + "learning_rate": 2.0406748324474233e-05, + "loss": 0.0, + "step": 15366 + }, + { + "epoch": 3.55, + "learning_rate": 2.040482243278638e-05, + "loss": 0.0, + "step": 15367 + }, + { + "epoch": 3.55, + "learning_rate": 2.0402896541098528e-05, + "loss": 0.0, + "step": 15368 + }, + { + "epoch": 3.55, + "learning_rate": 2.040097064941068e-05, + "loss": 0.0, + "step": 15369 + }, + { + "epoch": 3.55, + "learning_rate": 2.0399044757722826e-05, + "loss": 0.0, + "step": 15370 + }, + { + "epoch": 3.55, + "learning_rate": 2.0397118866034977e-05, + "loss": 0.0, + "step": 15371 + }, + { + "epoch": 3.55, + "learning_rate": 2.0395192974347124e-05, + "loss": 0.0001, + "step": 15372 + }, + { + "epoch": 3.55, + "learning_rate": 2.0393267082659272e-05, + "loss": 0.0, + "step": 15373 + }, + { + "epoch": 3.55, + "learning_rate": 2.0391341190971423e-05, + "loss": 0.0, + "step": 15374 + }, + { + "epoch": 3.55, + "learning_rate": 2.0389415299283567e-05, + "loss": 0.0, + "step": 15375 + }, + { + "epoch": 3.55, + "learning_rate": 2.0387489407595718e-05, + "loss": 0.0, + "step": 15376 + }, + { + "epoch": 3.55, + "learning_rate": 2.0385563515907865e-05, + "loss": 0.0001, + "step": 15377 + }, + { + "epoch": 3.55, + "learning_rate": 2.0383637624220016e-05, + "loss": 0.0, + "step": 15378 + }, + { + "epoch": 3.55, + "learning_rate": 2.0381711732532164e-05, + "loss": 0.0001, + "step": 15379 + }, + { + "epoch": 3.55, + "learning_rate": 2.037978584084431e-05, + "loss": 0.0, + "step": 15380 + }, + { + "epoch": 3.55, + "learning_rate": 2.0377859949156462e-05, + "loss": 0.0, + "step": 15381 + }, + { + "epoch": 3.55, + "learning_rate": 2.037593405746861e-05, + "loss": 0.0, + "step": 15382 + }, + { + "epoch": 3.56, + "learning_rate": 2.037400816578076e-05, + "loss": 0.0, + "step": 15383 + }, + { + "epoch": 3.56, + "learning_rate": 2.0372082274092908e-05, + "loss": 0.0, + "step": 15384 + }, + { + "epoch": 3.56, + "learning_rate": 2.0370156382405052e-05, + "loss": 0.0, + "step": 15385 + }, + { + "epoch": 3.56, + "learning_rate": 2.0368230490717203e-05, + "loss": 0.0, + "step": 15386 + }, + { + "epoch": 3.56, + "learning_rate": 2.036630459902935e-05, + "loss": 0.0, + "step": 15387 + }, + { + "epoch": 3.56, + "learning_rate": 2.03643787073415e-05, + "loss": 0.0, + "step": 15388 + }, + { + "epoch": 3.56, + "learning_rate": 2.036245281565365e-05, + "loss": 0.0, + "step": 15389 + }, + { + "epoch": 3.56, + "learning_rate": 2.0360526923965796e-05, + "loss": 0.0008, + "step": 15390 + }, + { + "epoch": 3.56, + "learning_rate": 2.0358601032277947e-05, + "loss": 0.0, + "step": 15391 + }, + { + "epoch": 3.56, + "learning_rate": 2.0356675140590094e-05, + "loss": 0.0, + "step": 15392 + }, + { + "epoch": 3.56, + "learning_rate": 2.0354749248902245e-05, + "loss": 0.0, + "step": 15393 + }, + { + "epoch": 3.56, + "learning_rate": 2.035282335721439e-05, + "loss": 0.0, + "step": 15394 + }, + { + "epoch": 3.56, + "learning_rate": 2.035089746552654e-05, + "loss": 0.0003, + "step": 15395 + }, + { + "epoch": 3.56, + "learning_rate": 2.0348971573838688e-05, + "loss": 0.0, + "step": 15396 + }, + { + "epoch": 3.56, + "learning_rate": 2.0347045682150835e-05, + "loss": 0.0, + "step": 15397 + }, + { + "epoch": 3.56, + "learning_rate": 2.0345119790462986e-05, + "loss": 0.0, + "step": 15398 + }, + { + "epoch": 3.56, + "learning_rate": 2.0343193898775133e-05, + "loss": 0.0, + "step": 15399 + }, + { + "epoch": 3.56, + "learning_rate": 2.0341268007087284e-05, + "loss": 0.0, + "step": 15400 + }, + { + "epoch": 3.56, + "learning_rate": 2.0339342115399432e-05, + "loss": 0.0, + "step": 15401 + }, + { + "epoch": 3.56, + "learning_rate": 2.033741622371158e-05, + "loss": 0.0, + "step": 15402 + }, + { + "epoch": 3.56, + "learning_rate": 2.0335490332023727e-05, + "loss": 0.0, + "step": 15403 + }, + { + "epoch": 3.56, + "learning_rate": 2.0333564440335874e-05, + "loss": 0.0, + "step": 15404 + }, + { + "epoch": 3.56, + "learning_rate": 2.0331638548648025e-05, + "loss": 0.0, + "step": 15405 + }, + { + "epoch": 3.56, + "learning_rate": 2.0329712656960173e-05, + "loss": 0.0, + "step": 15406 + }, + { + "epoch": 3.56, + "learning_rate": 2.0327786765272323e-05, + "loss": 0.0, + "step": 15407 + }, + { + "epoch": 3.56, + "learning_rate": 2.032586087358447e-05, + "loss": 0.0, + "step": 15408 + }, + { + "epoch": 3.56, + "learning_rate": 2.032393498189662e-05, + "loss": 0.0, + "step": 15409 + }, + { + "epoch": 3.56, + "learning_rate": 2.032200909020877e-05, + "loss": 0.0001, + "step": 15410 + }, + { + "epoch": 3.56, + "learning_rate": 2.0320083198520917e-05, + "loss": 0.0, + "step": 15411 + }, + { + "epoch": 3.56, + "learning_rate": 2.0318157306833068e-05, + "loss": 0.0, + "step": 15412 + }, + { + "epoch": 3.56, + "learning_rate": 2.031623141514521e-05, + "loss": 0.0001, + "step": 15413 + }, + { + "epoch": 3.56, + "learning_rate": 2.0314305523457363e-05, + "loss": 0.0, + "step": 15414 + }, + { + "epoch": 3.56, + "learning_rate": 2.031237963176951e-05, + "loss": 0.0, + "step": 15415 + }, + { + "epoch": 3.56, + "learning_rate": 2.0310453740081657e-05, + "loss": 0.0001, + "step": 15416 + }, + { + "epoch": 3.56, + "learning_rate": 2.030852784839381e-05, + "loss": 0.0, + "step": 15417 + }, + { + "epoch": 3.56, + "learning_rate": 2.0306601956705956e-05, + "loss": 0.0001, + "step": 15418 + }, + { + "epoch": 3.56, + "learning_rate": 2.0304676065018107e-05, + "loss": 0.0, + "step": 15419 + }, + { + "epoch": 3.56, + "learning_rate": 2.0302750173330254e-05, + "loss": 0.0, + "step": 15420 + }, + { + "epoch": 3.56, + "learning_rate": 2.03008242816424e-05, + "loss": 0.0, + "step": 15421 + }, + { + "epoch": 3.56, + "learning_rate": 2.029889838995455e-05, + "loss": 0.0, + "step": 15422 + }, + { + "epoch": 3.56, + "learning_rate": 2.0296972498266697e-05, + "loss": 0.0, + "step": 15423 + }, + { + "epoch": 3.56, + "learning_rate": 2.0295046606578847e-05, + "loss": 0.0, + "step": 15424 + }, + { + "epoch": 3.56, + "learning_rate": 2.0293120714890995e-05, + "loss": 0.0, + "step": 15425 + }, + { + "epoch": 3.57, + "learning_rate": 2.0291194823203142e-05, + "loss": 0.0, + "step": 15426 + }, + { + "epoch": 3.57, + "learning_rate": 2.0289268931515293e-05, + "loss": 0.0, + "step": 15427 + }, + { + "epoch": 3.57, + "learning_rate": 2.028734303982744e-05, + "loss": 0.0, + "step": 15428 + }, + { + "epoch": 3.57, + "learning_rate": 2.028541714813959e-05, + "loss": 0.0, + "step": 15429 + }, + { + "epoch": 3.57, + "learning_rate": 2.028349125645174e-05, + "loss": 0.0, + "step": 15430 + }, + { + "epoch": 3.57, + "learning_rate": 2.0281565364763887e-05, + "loss": 0.0, + "step": 15431 + }, + { + "epoch": 3.57, + "learning_rate": 2.0279639473076034e-05, + "loss": 0.0, + "step": 15432 + }, + { + "epoch": 3.57, + "learning_rate": 2.027771358138818e-05, + "loss": 0.0, + "step": 15433 + }, + { + "epoch": 3.57, + "learning_rate": 2.0275787689700332e-05, + "loss": 0.0, + "step": 15434 + }, + { + "epoch": 3.57, + "learning_rate": 2.027386179801248e-05, + "loss": 0.0, + "step": 15435 + }, + { + "epoch": 3.57, + "learning_rate": 2.027193590632463e-05, + "loss": 0.0, + "step": 15436 + }, + { + "epoch": 3.57, + "learning_rate": 2.0270010014636778e-05, + "loss": 0.0006, + "step": 15437 + }, + { + "epoch": 3.57, + "learning_rate": 2.0268084122948926e-05, + "loss": 0.0, + "step": 15438 + }, + { + "epoch": 3.57, + "learning_rate": 2.0266158231261077e-05, + "loss": 0.0022, + "step": 15439 + }, + { + "epoch": 3.57, + "learning_rate": 2.0264232339573224e-05, + "loss": 0.0001, + "step": 15440 + }, + { + "epoch": 3.57, + "learning_rate": 2.026230644788537e-05, + "loss": 0.0, + "step": 15441 + }, + { + "epoch": 3.57, + "learning_rate": 2.026038055619752e-05, + "loss": 0.0, + "step": 15442 + }, + { + "epoch": 3.57, + "learning_rate": 2.025845466450967e-05, + "loss": 0.0, + "step": 15443 + }, + { + "epoch": 3.57, + "learning_rate": 2.0256528772821817e-05, + "loss": 0.0, + "step": 15444 + }, + { + "epoch": 3.57, + "learning_rate": 2.0254602881133965e-05, + "loss": 0.0, + "step": 15445 + }, + { + "epoch": 3.57, + "learning_rate": 2.0252676989446116e-05, + "loss": 0.0, + "step": 15446 + }, + { + "epoch": 3.57, + "learning_rate": 2.0250751097758263e-05, + "loss": 0.0, + "step": 15447 + }, + { + "epoch": 3.57, + "learning_rate": 2.0248825206070414e-05, + "loss": 0.0, + "step": 15448 + }, + { + "epoch": 3.57, + "learning_rate": 2.024689931438256e-05, + "loss": 0.0, + "step": 15449 + }, + { + "epoch": 3.57, + "learning_rate": 2.024497342269471e-05, + "loss": 0.0005, + "step": 15450 + }, + { + "epoch": 3.57, + "learning_rate": 2.0243047531006856e-05, + "loss": 0.0, + "step": 15451 + }, + { + "epoch": 3.57, + "learning_rate": 2.0241121639319004e-05, + "loss": 0.0002, + "step": 15452 + }, + { + "epoch": 3.57, + "learning_rate": 2.0239195747631155e-05, + "loss": 0.0, + "step": 15453 + }, + { + "epoch": 3.57, + "learning_rate": 2.0237269855943302e-05, + "loss": 0.0, + "step": 15454 + }, + { + "epoch": 3.57, + "learning_rate": 2.0235343964255453e-05, + "loss": 0.0, + "step": 15455 + }, + { + "epoch": 3.57, + "learning_rate": 2.02334180725676e-05, + "loss": 0.0001, + "step": 15456 + }, + { + "epoch": 3.57, + "learning_rate": 2.0231492180879748e-05, + "loss": 0.0, + "step": 15457 + }, + { + "epoch": 3.57, + "learning_rate": 2.02295662891919e-05, + "loss": 0.0, + "step": 15458 + }, + { + "epoch": 3.57, + "learning_rate": 2.0227640397504043e-05, + "loss": 0.0, + "step": 15459 + }, + { + "epoch": 3.57, + "learning_rate": 2.0225714505816194e-05, + "loss": 0.0003, + "step": 15460 + }, + { + "epoch": 3.57, + "learning_rate": 2.022378861412834e-05, + "loss": 0.0, + "step": 15461 + }, + { + "epoch": 3.57, + "learning_rate": 2.022186272244049e-05, + "loss": 0.0, + "step": 15462 + }, + { + "epoch": 3.57, + "learning_rate": 2.021993683075264e-05, + "loss": 0.0003, + "step": 15463 + }, + { + "epoch": 3.57, + "learning_rate": 2.0218010939064787e-05, + "loss": 0.0, + "step": 15464 + }, + { + "epoch": 3.57, + "learning_rate": 2.0216085047376938e-05, + "loss": 0.0, + "step": 15465 + }, + { + "epoch": 3.57, + "learning_rate": 2.0214159155689086e-05, + "loss": 0.0, + "step": 15466 + }, + { + "epoch": 3.57, + "learning_rate": 2.0212233264001233e-05, + "loss": 0.0, + "step": 15467 + }, + { + "epoch": 3.57, + "learning_rate": 2.0210307372313384e-05, + "loss": 0.0, + "step": 15468 + }, + { + "epoch": 3.57, + "learning_rate": 2.0208381480625528e-05, + "loss": 0.0, + "step": 15469 + }, + { + "epoch": 3.58, + "learning_rate": 2.020645558893768e-05, + "loss": 0.0, + "step": 15470 + }, + { + "epoch": 3.58, + "learning_rate": 2.0204529697249826e-05, + "loss": 0.0, + "step": 15471 + }, + { + "epoch": 3.58, + "learning_rate": 2.0202603805561977e-05, + "loss": 0.0, + "step": 15472 + }, + { + "epoch": 3.58, + "learning_rate": 2.0200677913874125e-05, + "loss": 0.0, + "step": 15473 + }, + { + "epoch": 3.58, + "learning_rate": 2.0198752022186272e-05, + "loss": 0.0001, + "step": 15474 + }, + { + "epoch": 3.58, + "learning_rate": 2.0196826130498423e-05, + "loss": 0.0, + "step": 15475 + }, + { + "epoch": 3.58, + "learning_rate": 2.019490023881057e-05, + "loss": 0.0, + "step": 15476 + }, + { + "epoch": 3.58, + "learning_rate": 2.019297434712272e-05, + "loss": 0.0, + "step": 15477 + }, + { + "epoch": 3.58, + "learning_rate": 2.0191048455434865e-05, + "loss": 0.0, + "step": 15478 + }, + { + "epoch": 3.58, + "learning_rate": 2.0189122563747016e-05, + "loss": 0.0, + "step": 15479 + }, + { + "epoch": 3.58, + "learning_rate": 2.0187196672059164e-05, + "loss": 0.0, + "step": 15480 + }, + { + "epoch": 3.58, + "learning_rate": 2.018527078037131e-05, + "loss": 0.0, + "step": 15481 + }, + { + "epoch": 3.58, + "learning_rate": 2.0183344888683462e-05, + "loss": 0.0, + "step": 15482 + }, + { + "epoch": 3.58, + "learning_rate": 2.018141899699561e-05, + "loss": 0.0001, + "step": 15483 + }, + { + "epoch": 3.58, + "learning_rate": 2.017949310530776e-05, + "loss": 0.0, + "step": 15484 + }, + { + "epoch": 3.58, + "learning_rate": 2.0177567213619908e-05, + "loss": 0.0, + "step": 15485 + }, + { + "epoch": 3.58, + "learning_rate": 2.0175641321932055e-05, + "loss": 0.0017, + "step": 15486 + }, + { + "epoch": 3.58, + "learning_rate": 2.0173715430244203e-05, + "loss": 0.0, + "step": 15487 + }, + { + "epoch": 3.58, + "learning_rate": 2.017178953855635e-05, + "loss": 0.0, + "step": 15488 + }, + { + "epoch": 3.58, + "learning_rate": 2.01698636468685e-05, + "loss": 0.0, + "step": 15489 + }, + { + "epoch": 3.58, + "learning_rate": 2.016793775518065e-05, + "loss": 0.0, + "step": 15490 + }, + { + "epoch": 3.58, + "learning_rate": 2.01660118634928e-05, + "loss": 0.0, + "step": 15491 + }, + { + "epoch": 3.58, + "learning_rate": 2.0164085971804947e-05, + "loss": 0.0, + "step": 15492 + }, + { + "epoch": 3.58, + "learning_rate": 2.0162160080117095e-05, + "loss": 0.0, + "step": 15493 + }, + { + "epoch": 3.58, + "learning_rate": 2.0160234188429245e-05, + "loss": 0.0001, + "step": 15494 + }, + { + "epoch": 3.58, + "learning_rate": 2.0158308296741393e-05, + "loss": 0.0, + "step": 15495 + }, + { + "epoch": 3.58, + "learning_rate": 2.0156382405053544e-05, + "loss": 0.0, + "step": 15496 + }, + { + "epoch": 3.58, + "learning_rate": 2.0154456513365688e-05, + "loss": 0.0, + "step": 15497 + }, + { + "epoch": 3.58, + "learning_rate": 2.0152530621677835e-05, + "loss": 0.0, + "step": 15498 + }, + { + "epoch": 3.58, + "learning_rate": 2.0150604729989986e-05, + "loss": 0.0, + "step": 15499 + }, + { + "epoch": 3.58, + "learning_rate": 2.0148678838302134e-05, + "loss": 0.0, + "step": 15500 + }, + { + "epoch": 3.58, + "learning_rate": 2.0146752946614285e-05, + "loss": 0.0, + "step": 15501 + }, + { + "epoch": 3.58, + "learning_rate": 2.0144827054926432e-05, + "loss": 0.0, + "step": 15502 + }, + { + "epoch": 3.58, + "learning_rate": 2.014290116323858e-05, + "loss": 0.0, + "step": 15503 + }, + { + "epoch": 3.58, + "learning_rate": 2.014097527155073e-05, + "loss": 0.0001, + "step": 15504 + }, + { + "epoch": 3.58, + "learning_rate": 2.0139049379862878e-05, + "loss": 0.0, + "step": 15505 + }, + { + "epoch": 3.58, + "learning_rate": 2.0137123488175025e-05, + "loss": 0.0, + "step": 15506 + }, + { + "epoch": 3.58, + "learning_rate": 2.0135197596487173e-05, + "loss": 0.0, + "step": 15507 + }, + { + "epoch": 3.58, + "learning_rate": 2.0133271704799324e-05, + "loss": 0.0, + "step": 15508 + }, + { + "epoch": 3.58, + "learning_rate": 2.013134581311147e-05, + "loss": 0.0, + "step": 15509 + }, + { + "epoch": 3.58, + "learning_rate": 2.012941992142362e-05, + "loss": 0.0, + "step": 15510 + }, + { + "epoch": 3.58, + "learning_rate": 2.012749402973577e-05, + "loss": 0.0, + "step": 15511 + }, + { + "epoch": 3.58, + "learning_rate": 2.0125568138047917e-05, + "loss": 0.0001, + "step": 15512 + }, + { + "epoch": 3.59, + "learning_rate": 2.0123642246360068e-05, + "loss": 0.0, + "step": 15513 + }, + { + "epoch": 3.59, + "learning_rate": 2.0121716354672215e-05, + "loss": 0.0, + "step": 15514 + }, + { + "epoch": 3.59, + "learning_rate": 2.0119790462984363e-05, + "loss": 0.0, + "step": 15515 + }, + { + "epoch": 3.59, + "learning_rate": 2.011786457129651e-05, + "loss": 0.0, + "step": 15516 + }, + { + "epoch": 3.59, + "learning_rate": 2.0115938679608658e-05, + "loss": 0.0, + "step": 15517 + }, + { + "epoch": 3.59, + "learning_rate": 2.011401278792081e-05, + "loss": 0.0, + "step": 15518 + }, + { + "epoch": 3.59, + "learning_rate": 2.0112086896232956e-05, + "loss": 0.0, + "step": 15519 + }, + { + "epoch": 3.59, + "learning_rate": 2.0110161004545107e-05, + "loss": 0.0, + "step": 15520 + }, + { + "epoch": 3.59, + "learning_rate": 2.0108235112857254e-05, + "loss": 0.0, + "step": 15521 + }, + { + "epoch": 3.59, + "learning_rate": 2.0106309221169402e-05, + "loss": 0.0, + "step": 15522 + }, + { + "epoch": 3.59, + "learning_rate": 2.0104383329481553e-05, + "loss": 0.0, + "step": 15523 + }, + { + "epoch": 3.59, + "learning_rate": 2.01024574377937e-05, + "loss": 0.0, + "step": 15524 + }, + { + "epoch": 3.59, + "learning_rate": 2.0100531546105848e-05, + "loss": 0.0, + "step": 15525 + }, + { + "epoch": 3.59, + "learning_rate": 2.0098605654417995e-05, + "loss": 0.0, + "step": 15526 + }, + { + "epoch": 3.59, + "learning_rate": 2.0096679762730146e-05, + "loss": 0.0003, + "step": 15527 + }, + { + "epoch": 3.59, + "learning_rate": 2.0094753871042294e-05, + "loss": 0.0, + "step": 15528 + }, + { + "epoch": 3.59, + "learning_rate": 2.009282797935444e-05, + "loss": 0.0, + "step": 15529 + }, + { + "epoch": 3.59, + "learning_rate": 2.0090902087666592e-05, + "loss": 0.0, + "step": 15530 + }, + { + "epoch": 3.59, + "learning_rate": 2.008897619597874e-05, + "loss": 0.0, + "step": 15531 + }, + { + "epoch": 3.59, + "learning_rate": 2.008705030429089e-05, + "loss": 0.0, + "step": 15532 + }, + { + "epoch": 3.59, + "learning_rate": 2.0085124412603038e-05, + "loss": 0.0001, + "step": 15533 + }, + { + "epoch": 3.59, + "learning_rate": 2.0083198520915185e-05, + "loss": 0.0, + "step": 15534 + }, + { + "epoch": 3.59, + "learning_rate": 2.0081272629227333e-05, + "loss": 0.0, + "step": 15535 + }, + { + "epoch": 3.59, + "learning_rate": 2.007934673753948e-05, + "loss": 0.0, + "step": 15536 + }, + { + "epoch": 3.59, + "learning_rate": 2.007742084585163e-05, + "loss": 0.0, + "step": 15537 + }, + { + "epoch": 3.59, + "learning_rate": 2.007549495416378e-05, + "loss": 0.0, + "step": 15538 + }, + { + "epoch": 3.59, + "learning_rate": 2.0073569062475926e-05, + "loss": 0.0, + "step": 15539 + }, + { + "epoch": 3.59, + "learning_rate": 2.0071643170788077e-05, + "loss": 0.0, + "step": 15540 + }, + { + "epoch": 3.59, + "learning_rate": 2.0069717279100224e-05, + "loss": 0.0, + "step": 15541 + }, + { + "epoch": 3.59, + "learning_rate": 2.0067791387412375e-05, + "loss": 0.0, + "step": 15542 + }, + { + "epoch": 3.59, + "learning_rate": 2.006586549572452e-05, + "loss": 0.0001, + "step": 15543 + }, + { + "epoch": 3.59, + "learning_rate": 2.006393960403667e-05, + "loss": 0.0001, + "step": 15544 + }, + { + "epoch": 3.59, + "learning_rate": 2.0062013712348818e-05, + "loss": 0.0, + "step": 15545 + }, + { + "epoch": 3.59, + "learning_rate": 2.0060087820660965e-05, + "loss": 0.0, + "step": 15546 + }, + { + "epoch": 3.59, + "learning_rate": 2.0058161928973116e-05, + "loss": 0.0, + "step": 15547 + }, + { + "epoch": 3.59, + "learning_rate": 2.0056236037285263e-05, + "loss": 0.0, + "step": 15548 + }, + { + "epoch": 3.59, + "learning_rate": 2.0054310145597414e-05, + "loss": 0.0, + "step": 15549 + }, + { + "epoch": 3.59, + "learning_rate": 2.0052384253909562e-05, + "loss": 0.0, + "step": 15550 + }, + { + "epoch": 3.59, + "learning_rate": 2.005045836222171e-05, + "loss": 0.0, + "step": 15551 + }, + { + "epoch": 3.59, + "learning_rate": 2.004853247053386e-05, + "loss": 0.0, + "step": 15552 + }, + { + "epoch": 3.59, + "learning_rate": 2.0046606578846004e-05, + "loss": 0.0, + "step": 15553 + }, + { + "epoch": 3.59, + "learning_rate": 2.0044680687158155e-05, + "loss": 0.0, + "step": 15554 + }, + { + "epoch": 3.59, + "learning_rate": 2.0042754795470303e-05, + "loss": 0.0, + "step": 15555 + }, + { + "epoch": 3.6, + "learning_rate": 2.0040828903782453e-05, + "loss": 0.0, + "step": 15556 + }, + { + "epoch": 3.6, + "learning_rate": 2.00389030120946e-05, + "loss": 0.0001, + "step": 15557 + }, + { + "epoch": 3.6, + "learning_rate": 2.003697712040675e-05, + "loss": 0.0, + "step": 15558 + }, + { + "epoch": 3.6, + "learning_rate": 2.00350512287189e-05, + "loss": 0.0, + "step": 15559 + }, + { + "epoch": 3.6, + "learning_rate": 2.0033125337031047e-05, + "loss": 0.0, + "step": 15560 + }, + { + "epoch": 3.6, + "learning_rate": 2.0031199445343198e-05, + "loss": 0.0, + "step": 15561 + }, + { + "epoch": 3.6, + "learning_rate": 2.002927355365534e-05, + "loss": 0.0001, + "step": 15562 + }, + { + "epoch": 3.6, + "learning_rate": 2.0027347661967493e-05, + "loss": 0.0, + "step": 15563 + }, + { + "epoch": 3.6, + "learning_rate": 2.002542177027964e-05, + "loss": 0.0, + "step": 15564 + }, + { + "epoch": 3.6, + "learning_rate": 2.0023495878591787e-05, + "loss": 0.0, + "step": 15565 + }, + { + "epoch": 3.6, + "learning_rate": 2.002156998690394e-05, + "loss": 0.0, + "step": 15566 + }, + { + "epoch": 3.6, + "learning_rate": 2.0019644095216086e-05, + "loss": 0.0, + "step": 15567 + }, + { + "epoch": 3.6, + "learning_rate": 2.0017718203528237e-05, + "loss": 0.0, + "step": 15568 + }, + { + "epoch": 3.6, + "learning_rate": 2.0015792311840384e-05, + "loss": 0.0, + "step": 15569 + }, + { + "epoch": 3.6, + "learning_rate": 2.001386642015253e-05, + "loss": 0.0, + "step": 15570 + }, + { + "epoch": 3.6, + "learning_rate": 2.001194052846468e-05, + "loss": 0.0, + "step": 15571 + }, + { + "epoch": 3.6, + "learning_rate": 2.0010014636776827e-05, + "loss": 0.0, + "step": 15572 + }, + { + "epoch": 3.6, + "learning_rate": 2.0008088745088977e-05, + "loss": 0.0, + "step": 15573 + }, + { + "epoch": 3.6, + "learning_rate": 2.0006162853401125e-05, + "loss": 0.0, + "step": 15574 + }, + { + "epoch": 3.6, + "learning_rate": 2.0004236961713272e-05, + "loss": 0.0, + "step": 15575 + }, + { + "epoch": 3.6, + "learning_rate": 2.0002311070025423e-05, + "loss": 0.0, + "step": 15576 + }, + { + "epoch": 3.6, + "learning_rate": 2.000038517833757e-05, + "loss": 0.0, + "step": 15577 + }, + { + "epoch": 3.6, + "learning_rate": 1.999845928664972e-05, + "loss": 0.0, + "step": 15578 + }, + { + "epoch": 3.6, + "learning_rate": 1.999653339496187e-05, + "loss": 0.0161, + "step": 15579 + }, + { + "epoch": 3.6, + "learning_rate": 1.9994607503274017e-05, + "loss": 0.0, + "step": 15580 + }, + { + "epoch": 3.6, + "learning_rate": 1.9992681611586164e-05, + "loss": 0.0, + "step": 15581 + }, + { + "epoch": 3.6, + "learning_rate": 1.999075571989831e-05, + "loss": 0.0, + "step": 15582 + }, + { + "epoch": 3.6, + "learning_rate": 1.9988829828210462e-05, + "loss": 0.0016, + "step": 15583 + }, + { + "epoch": 3.6, + "learning_rate": 1.998690393652261e-05, + "loss": 0.0, + "step": 15584 + }, + { + "epoch": 3.6, + "learning_rate": 1.998497804483476e-05, + "loss": 0.0, + "step": 15585 + }, + { + "epoch": 3.6, + "learning_rate": 1.9983052153146908e-05, + "loss": 0.0, + "step": 15586 + }, + { + "epoch": 3.6, + "learning_rate": 1.9981126261459056e-05, + "loss": 0.0, + "step": 15587 + }, + { + "epoch": 3.6, + "learning_rate": 1.9979200369771207e-05, + "loss": 0.0, + "step": 15588 + }, + { + "epoch": 3.6, + "learning_rate": 1.9977274478083354e-05, + "loss": 0.0, + "step": 15589 + }, + { + "epoch": 3.6, + "learning_rate": 1.99753485863955e-05, + "loss": 0.0, + "step": 15590 + }, + { + "epoch": 3.6, + "learning_rate": 1.997342269470765e-05, + "loss": 0.0, + "step": 15591 + }, + { + "epoch": 3.6, + "learning_rate": 1.99714968030198e-05, + "loss": 0.0, + "step": 15592 + }, + { + "epoch": 3.6, + "learning_rate": 1.9969570911331947e-05, + "loss": 0.0, + "step": 15593 + }, + { + "epoch": 3.6, + "learning_rate": 1.9967645019644095e-05, + "loss": 0.0, + "step": 15594 + }, + { + "epoch": 3.6, + "learning_rate": 1.9965719127956246e-05, + "loss": 0.0, + "step": 15595 + }, + { + "epoch": 3.6, + "learning_rate": 1.9963793236268393e-05, + "loss": 0.0, + "step": 15596 + }, + { + "epoch": 3.6, + "learning_rate": 1.9961867344580544e-05, + "loss": 0.0, + "step": 15597 + }, + { + "epoch": 3.6, + "learning_rate": 1.995994145289269e-05, + "loss": 0.0, + "step": 15598 + }, + { + "epoch": 3.61, + "learning_rate": 1.995801556120484e-05, + "loss": 0.0, + "step": 15599 + }, + { + "epoch": 3.61, + "learning_rate": 1.9956089669516986e-05, + "loss": 0.0, + "step": 15600 + }, + { + "epoch": 3.61, + "learning_rate": 1.9954163777829134e-05, + "loss": 0.0, + "step": 15601 + }, + { + "epoch": 3.61, + "learning_rate": 1.9952237886141285e-05, + "loss": 0.0, + "step": 15602 + }, + { + "epoch": 3.61, + "learning_rate": 1.9950311994453432e-05, + "loss": 0.0, + "step": 15603 + }, + { + "epoch": 3.61, + "learning_rate": 1.9948386102765583e-05, + "loss": 0.0, + "step": 15604 + }, + { + "epoch": 3.61, + "learning_rate": 1.994646021107773e-05, + "loss": 0.0, + "step": 15605 + }, + { + "epoch": 3.61, + "learning_rate": 1.9944534319389878e-05, + "loss": 0.0001, + "step": 15606 + }, + { + "epoch": 3.61, + "learning_rate": 1.994260842770203e-05, + "loss": 0.0, + "step": 15607 + }, + { + "epoch": 3.61, + "learning_rate": 1.9940682536014173e-05, + "loss": 0.0, + "step": 15608 + }, + { + "epoch": 3.61, + "learning_rate": 1.9938756644326324e-05, + "loss": 0.0, + "step": 15609 + }, + { + "epoch": 3.61, + "learning_rate": 1.993683075263847e-05, + "loss": 0.0, + "step": 15610 + }, + { + "epoch": 3.61, + "learning_rate": 1.993490486095062e-05, + "loss": 0.0, + "step": 15611 + }, + { + "epoch": 3.61, + "learning_rate": 1.993297896926277e-05, + "loss": 0.0, + "step": 15612 + }, + { + "epoch": 3.61, + "learning_rate": 1.9931053077574917e-05, + "loss": 0.0, + "step": 15613 + }, + { + "epoch": 3.61, + "learning_rate": 1.9929127185887068e-05, + "loss": 0.0, + "step": 15614 + }, + { + "epoch": 3.61, + "learning_rate": 1.9927201294199216e-05, + "loss": 0.0, + "step": 15615 + }, + { + "epoch": 3.61, + "learning_rate": 1.9925275402511363e-05, + "loss": 0.0, + "step": 15616 + }, + { + "epoch": 3.61, + "learning_rate": 1.9923349510823514e-05, + "loss": 0.0, + "step": 15617 + }, + { + "epoch": 3.61, + "learning_rate": 1.9921423619135658e-05, + "loss": 0.0013, + "step": 15618 + }, + { + "epoch": 3.61, + "learning_rate": 1.991949772744781e-05, + "loss": 0.0, + "step": 15619 + }, + { + "epoch": 3.61, + "learning_rate": 1.9917571835759956e-05, + "loss": 0.0, + "step": 15620 + }, + { + "epoch": 3.61, + "learning_rate": 1.9915645944072107e-05, + "loss": 0.0, + "step": 15621 + }, + { + "epoch": 3.61, + "learning_rate": 1.9913720052384255e-05, + "loss": 0.0, + "step": 15622 + }, + { + "epoch": 3.61, + "learning_rate": 1.9911794160696402e-05, + "loss": 0.0, + "step": 15623 + }, + { + "epoch": 3.61, + "learning_rate": 1.9909868269008553e-05, + "loss": 0.0, + "step": 15624 + }, + { + "epoch": 3.61, + "learning_rate": 1.99079423773207e-05, + "loss": 0.0, + "step": 15625 + }, + { + "epoch": 3.61, + "learning_rate": 1.990601648563285e-05, + "loss": 0.0, + "step": 15626 + }, + { + "epoch": 3.61, + "learning_rate": 1.9904090593944995e-05, + "loss": 0.0, + "step": 15627 + }, + { + "epoch": 3.61, + "learning_rate": 1.9902164702257146e-05, + "loss": 0.0, + "step": 15628 + }, + { + "epoch": 3.61, + "learning_rate": 1.9900238810569294e-05, + "loss": 0.0, + "step": 15629 + }, + { + "epoch": 3.61, + "learning_rate": 1.989831291888144e-05, + "loss": 0.0, + "step": 15630 + }, + { + "epoch": 3.61, + "learning_rate": 1.9896387027193592e-05, + "loss": 0.0, + "step": 15631 + }, + { + "epoch": 3.61, + "learning_rate": 1.989446113550574e-05, + "loss": 0.0025, + "step": 15632 + }, + { + "epoch": 3.61, + "learning_rate": 1.989253524381789e-05, + "loss": 0.0, + "step": 15633 + }, + { + "epoch": 3.61, + "learning_rate": 1.9890609352130038e-05, + "loss": 0.0, + "step": 15634 + }, + { + "epoch": 3.61, + "learning_rate": 1.9888683460442185e-05, + "loss": 0.0, + "step": 15635 + }, + { + "epoch": 3.61, + "learning_rate": 1.9886757568754333e-05, + "loss": 0.0, + "step": 15636 + }, + { + "epoch": 3.61, + "learning_rate": 1.988483167706648e-05, + "loss": 0.0, + "step": 15637 + }, + { + "epoch": 3.61, + "learning_rate": 1.988290578537863e-05, + "loss": 0.0, + "step": 15638 + }, + { + "epoch": 3.61, + "learning_rate": 1.988097989369078e-05, + "loss": 0.0, + "step": 15639 + }, + { + "epoch": 3.61, + "learning_rate": 1.987905400200293e-05, + "loss": 0.0, + "step": 15640 + }, + { + "epoch": 3.61, + "learning_rate": 1.9877128110315077e-05, + "loss": 0.0, + "step": 15641 + }, + { + "epoch": 3.61, + "learning_rate": 1.9875202218627225e-05, + "loss": 0.0, + "step": 15642 + }, + { + "epoch": 3.62, + "learning_rate": 1.9873276326939375e-05, + "loss": 0.0, + "step": 15643 + }, + { + "epoch": 3.62, + "learning_rate": 1.9871350435251523e-05, + "loss": 0.0, + "step": 15644 + }, + { + "epoch": 3.62, + "learning_rate": 1.9869424543563674e-05, + "loss": 0.0, + "step": 15645 + }, + { + "epoch": 3.62, + "learning_rate": 1.9867498651875818e-05, + "loss": 0.0, + "step": 15646 + }, + { + "epoch": 3.62, + "learning_rate": 1.986557276018797e-05, + "loss": 0.0, + "step": 15647 + }, + { + "epoch": 3.62, + "learning_rate": 1.9863646868500116e-05, + "loss": 0.0, + "step": 15648 + }, + { + "epoch": 3.62, + "learning_rate": 1.9861720976812264e-05, + "loss": 0.0008, + "step": 15649 + }, + { + "epoch": 3.62, + "learning_rate": 1.9859795085124415e-05, + "loss": 0.0001, + "step": 15650 + }, + { + "epoch": 3.62, + "learning_rate": 1.9857869193436562e-05, + "loss": 0.0, + "step": 15651 + }, + { + "epoch": 3.62, + "learning_rate": 1.985594330174871e-05, + "loss": 0.0, + "step": 15652 + }, + { + "epoch": 3.62, + "learning_rate": 1.985401741006086e-05, + "loss": 0.0001, + "step": 15653 + }, + { + "epoch": 3.62, + "learning_rate": 1.9852091518373008e-05, + "loss": 0.0, + "step": 15654 + }, + { + "epoch": 3.62, + "learning_rate": 1.9850165626685155e-05, + "loss": 0.0, + "step": 15655 + }, + { + "epoch": 3.62, + "learning_rate": 1.9848239734997303e-05, + "loss": 0.0, + "step": 15656 + }, + { + "epoch": 3.62, + "learning_rate": 1.9846313843309454e-05, + "loss": 0.0001, + "step": 15657 + }, + { + "epoch": 3.62, + "learning_rate": 1.98443879516216e-05, + "loss": 0.0, + "step": 15658 + }, + { + "epoch": 3.62, + "learning_rate": 1.984246205993375e-05, + "loss": 0.0, + "step": 15659 + }, + { + "epoch": 3.62, + "learning_rate": 1.98405361682459e-05, + "loss": 0.0, + "step": 15660 + }, + { + "epoch": 3.62, + "learning_rate": 1.9838610276558047e-05, + "loss": 0.0, + "step": 15661 + }, + { + "epoch": 3.62, + "learning_rate": 1.9836684384870198e-05, + "loss": 0.0, + "step": 15662 + }, + { + "epoch": 3.62, + "learning_rate": 1.9834758493182345e-05, + "loss": 0.0001, + "step": 15663 + }, + { + "epoch": 3.62, + "learning_rate": 1.9832832601494493e-05, + "loss": 0.0, + "step": 15664 + }, + { + "epoch": 3.62, + "learning_rate": 1.983090670980664e-05, + "loss": 0.0, + "step": 15665 + }, + { + "epoch": 3.62, + "learning_rate": 1.9828980818118788e-05, + "loss": 0.0, + "step": 15666 + }, + { + "epoch": 3.62, + "learning_rate": 1.982705492643094e-05, + "loss": 0.0, + "step": 15667 + }, + { + "epoch": 3.62, + "learning_rate": 1.9825129034743086e-05, + "loss": 0.0, + "step": 15668 + }, + { + "epoch": 3.62, + "learning_rate": 1.9823203143055237e-05, + "loss": 0.0, + "step": 15669 + }, + { + "epoch": 3.62, + "learning_rate": 1.9821277251367384e-05, + "loss": 0.0, + "step": 15670 + }, + { + "epoch": 3.62, + "learning_rate": 1.9819351359679532e-05, + "loss": 0.0, + "step": 15671 + }, + { + "epoch": 3.62, + "learning_rate": 1.9817425467991683e-05, + "loss": 0.0002, + "step": 15672 + }, + { + "epoch": 3.62, + "learning_rate": 1.981549957630383e-05, + "loss": 0.0, + "step": 15673 + }, + { + "epoch": 3.62, + "learning_rate": 1.9813573684615978e-05, + "loss": 0.0, + "step": 15674 + }, + { + "epoch": 3.62, + "learning_rate": 1.9811647792928125e-05, + "loss": 0.0, + "step": 15675 + }, + { + "epoch": 3.62, + "learning_rate": 1.9809721901240276e-05, + "loss": 0.0, + "step": 15676 + }, + { + "epoch": 3.62, + "learning_rate": 1.9807796009552424e-05, + "loss": 0.0, + "step": 15677 + }, + { + "epoch": 3.62, + "learning_rate": 1.980587011786457e-05, + "loss": 0.0, + "step": 15678 + }, + { + "epoch": 3.62, + "learning_rate": 1.9803944226176722e-05, + "loss": 0.0, + "step": 15679 + }, + { + "epoch": 3.62, + "learning_rate": 1.980201833448887e-05, + "loss": 0.0, + "step": 15680 + }, + { + "epoch": 3.62, + "learning_rate": 1.980009244280102e-05, + "loss": 0.0, + "step": 15681 + }, + { + "epoch": 3.62, + "learning_rate": 1.9798166551113168e-05, + "loss": 0.0, + "step": 15682 + }, + { + "epoch": 3.62, + "learning_rate": 1.9796240659425315e-05, + "loss": 0.0, + "step": 15683 + }, + { + "epoch": 3.62, + "learning_rate": 1.9794314767737463e-05, + "loss": 0.0, + "step": 15684 + }, + { + "epoch": 3.62, + "learning_rate": 1.979238887604961e-05, + "loss": 0.0001, + "step": 15685 + }, + { + "epoch": 3.63, + "learning_rate": 1.979046298436176e-05, + "loss": 0.0, + "step": 15686 + }, + { + "epoch": 3.63, + "learning_rate": 1.978853709267391e-05, + "loss": 0.0, + "step": 15687 + }, + { + "epoch": 3.63, + "learning_rate": 1.9786611200986056e-05, + "loss": 0.0, + "step": 15688 + }, + { + "epoch": 3.63, + "learning_rate": 1.9784685309298207e-05, + "loss": 0.0, + "step": 15689 + }, + { + "epoch": 3.63, + "learning_rate": 1.9782759417610354e-05, + "loss": 0.0, + "step": 15690 + }, + { + "epoch": 3.63, + "learning_rate": 1.9780833525922505e-05, + "loss": 0.0001, + "step": 15691 + }, + { + "epoch": 3.63, + "learning_rate": 1.977890763423465e-05, + "loss": 0.0, + "step": 15692 + }, + { + "epoch": 3.63, + "learning_rate": 1.97769817425468e-05, + "loss": 0.0, + "step": 15693 + }, + { + "epoch": 3.63, + "learning_rate": 1.9775055850858948e-05, + "loss": 0.0, + "step": 15694 + }, + { + "epoch": 3.63, + "learning_rate": 1.9773129959171095e-05, + "loss": 0.0, + "step": 15695 + }, + { + "epoch": 3.63, + "learning_rate": 1.9771204067483246e-05, + "loss": 0.0001, + "step": 15696 + }, + { + "epoch": 3.63, + "learning_rate": 1.9769278175795393e-05, + "loss": 0.0, + "step": 15697 + }, + { + "epoch": 3.63, + "learning_rate": 1.9767352284107544e-05, + "loss": 0.0, + "step": 15698 + }, + { + "epoch": 3.63, + "learning_rate": 1.9765426392419692e-05, + "loss": 0.0, + "step": 15699 + }, + { + "epoch": 3.63, + "learning_rate": 1.976350050073184e-05, + "loss": 0.0, + "step": 15700 + }, + { + "epoch": 3.63, + "learning_rate": 1.976157460904399e-05, + "loss": 0.0, + "step": 15701 + }, + { + "epoch": 3.63, + "learning_rate": 1.9759648717356134e-05, + "loss": 0.0001, + "step": 15702 + }, + { + "epoch": 3.63, + "learning_rate": 1.9757722825668285e-05, + "loss": 0.0033, + "step": 15703 + }, + { + "epoch": 3.63, + "learning_rate": 1.9755796933980433e-05, + "loss": 0.0, + "step": 15704 + }, + { + "epoch": 3.63, + "learning_rate": 1.9753871042292583e-05, + "loss": 0.0, + "step": 15705 + }, + { + "epoch": 3.63, + "learning_rate": 1.975194515060473e-05, + "loss": 0.0, + "step": 15706 + }, + { + "epoch": 3.63, + "learning_rate": 1.975001925891688e-05, + "loss": 0.0, + "step": 15707 + }, + { + "epoch": 3.63, + "learning_rate": 1.974809336722903e-05, + "loss": 0.0, + "step": 15708 + }, + { + "epoch": 3.63, + "learning_rate": 1.9746167475541177e-05, + "loss": 0.0, + "step": 15709 + }, + { + "epoch": 3.63, + "learning_rate": 1.9744241583853328e-05, + "loss": 0.0, + "step": 15710 + }, + { + "epoch": 3.63, + "learning_rate": 1.9742315692165472e-05, + "loss": 0.0, + "step": 15711 + }, + { + "epoch": 3.63, + "learning_rate": 1.9740389800477623e-05, + "loss": 0.0, + "step": 15712 + }, + { + "epoch": 3.63, + "learning_rate": 1.973846390878977e-05, + "loss": 0.0, + "step": 15713 + }, + { + "epoch": 3.63, + "learning_rate": 1.9736538017101918e-05, + "loss": 0.0, + "step": 15714 + }, + { + "epoch": 3.63, + "learning_rate": 1.973461212541407e-05, + "loss": 0.0, + "step": 15715 + }, + { + "epoch": 3.63, + "learning_rate": 1.9732686233726216e-05, + "loss": 0.0, + "step": 15716 + }, + { + "epoch": 3.63, + "learning_rate": 1.9730760342038367e-05, + "loss": 0.0, + "step": 15717 + }, + { + "epoch": 3.63, + "learning_rate": 1.9728834450350514e-05, + "loss": 0.0, + "step": 15718 + }, + { + "epoch": 3.63, + "learning_rate": 1.9726908558662662e-05, + "loss": 0.0, + "step": 15719 + }, + { + "epoch": 3.63, + "learning_rate": 1.972498266697481e-05, + "loss": 0.0, + "step": 15720 + }, + { + "epoch": 3.63, + "learning_rate": 1.9723056775286957e-05, + "loss": 0.0, + "step": 15721 + }, + { + "epoch": 3.63, + "learning_rate": 1.9721130883599108e-05, + "loss": 0.0, + "step": 15722 + }, + { + "epoch": 3.63, + "learning_rate": 1.9719204991911255e-05, + "loss": 0.0, + "step": 15723 + }, + { + "epoch": 3.63, + "learning_rate": 1.9717279100223406e-05, + "loss": 0.0, + "step": 15724 + }, + { + "epoch": 3.63, + "learning_rate": 1.9715353208535553e-05, + "loss": 0.0, + "step": 15725 + }, + { + "epoch": 3.63, + "learning_rate": 1.97134273168477e-05, + "loss": 0.0, + "step": 15726 + }, + { + "epoch": 3.63, + "learning_rate": 1.971150142515985e-05, + "loss": 0.0, + "step": 15727 + }, + { + "epoch": 3.63, + "learning_rate": 1.9709575533472e-05, + "loss": 0.0, + "step": 15728 + }, + { + "epoch": 3.64, + "learning_rate": 1.9707649641784147e-05, + "loss": 0.0, + "step": 15729 + }, + { + "epoch": 3.64, + "learning_rate": 1.9705723750096294e-05, + "loss": 0.0, + "step": 15730 + }, + { + "epoch": 3.64, + "learning_rate": 1.970379785840844e-05, + "loss": 0.0, + "step": 15731 + }, + { + "epoch": 3.64, + "learning_rate": 1.9701871966720592e-05, + "loss": 0.0, + "step": 15732 + }, + { + "epoch": 3.64, + "learning_rate": 1.969994607503274e-05, + "loss": 0.0, + "step": 15733 + }, + { + "epoch": 3.64, + "learning_rate": 1.969802018334489e-05, + "loss": 0.0, + "step": 15734 + }, + { + "epoch": 3.64, + "learning_rate": 1.9696094291657038e-05, + "loss": 0.0373, + "step": 15735 + }, + { + "epoch": 3.64, + "learning_rate": 1.9694168399969186e-05, + "loss": 0.0004, + "step": 15736 + }, + { + "epoch": 3.64, + "learning_rate": 1.9692242508281337e-05, + "loss": 0.0, + "step": 15737 + }, + { + "epoch": 3.64, + "learning_rate": 1.9690316616593484e-05, + "loss": 0.0, + "step": 15738 + }, + { + "epoch": 3.64, + "learning_rate": 1.968839072490563e-05, + "loss": 0.0, + "step": 15739 + }, + { + "epoch": 3.64, + "learning_rate": 1.968646483321778e-05, + "loss": 0.0, + "step": 15740 + }, + { + "epoch": 3.64, + "learning_rate": 1.968453894152993e-05, + "loss": 0.0, + "step": 15741 + }, + { + "epoch": 3.64, + "learning_rate": 1.9682613049842077e-05, + "loss": 0.0, + "step": 15742 + }, + { + "epoch": 3.64, + "learning_rate": 1.9680687158154225e-05, + "loss": 0.0, + "step": 15743 + }, + { + "epoch": 3.64, + "learning_rate": 1.9678761266466376e-05, + "loss": 0.0, + "step": 15744 + }, + { + "epoch": 3.64, + "learning_rate": 1.9676835374778523e-05, + "loss": 0.0, + "step": 15745 + }, + { + "epoch": 3.64, + "learning_rate": 1.9674909483090674e-05, + "loss": 0.0001, + "step": 15746 + }, + { + "epoch": 3.64, + "learning_rate": 1.967298359140282e-05, + "loss": 0.0, + "step": 15747 + }, + { + "epoch": 3.64, + "learning_rate": 1.967105769971497e-05, + "loss": 0.0, + "step": 15748 + }, + { + "epoch": 3.64, + "learning_rate": 1.9669131808027117e-05, + "loss": 0.0, + "step": 15749 + }, + { + "epoch": 3.64, + "learning_rate": 1.9667205916339264e-05, + "loss": 0.0, + "step": 15750 + }, + { + "epoch": 3.64, + "learning_rate": 1.9665280024651415e-05, + "loss": 0.0001, + "step": 15751 + }, + { + "epoch": 3.64, + "learning_rate": 1.9663354132963562e-05, + "loss": 0.0, + "step": 15752 + }, + { + "epoch": 3.64, + "learning_rate": 1.9661428241275713e-05, + "loss": 0.0, + "step": 15753 + }, + { + "epoch": 3.64, + "learning_rate": 1.965950234958786e-05, + "loss": 0.0, + "step": 15754 + }, + { + "epoch": 3.64, + "learning_rate": 1.9657576457900008e-05, + "loss": 0.0, + "step": 15755 + }, + { + "epoch": 3.64, + "learning_rate": 1.965565056621216e-05, + "loss": 0.0007, + "step": 15756 + }, + { + "epoch": 3.64, + "learning_rate": 1.9653724674524307e-05, + "loss": 0.0, + "step": 15757 + }, + { + "epoch": 3.64, + "learning_rate": 1.9651798782836454e-05, + "loss": 0.0, + "step": 15758 + }, + { + "epoch": 3.64, + "learning_rate": 1.96498728911486e-05, + "loss": 0.0002, + "step": 15759 + }, + { + "epoch": 3.64, + "learning_rate": 1.9647946999460752e-05, + "loss": 0.0, + "step": 15760 + }, + { + "epoch": 3.64, + "learning_rate": 1.96460211077729e-05, + "loss": 0.0, + "step": 15761 + }, + { + "epoch": 3.64, + "learning_rate": 1.9644095216085047e-05, + "loss": 0.0, + "step": 15762 + }, + { + "epoch": 3.64, + "learning_rate": 1.9642169324397198e-05, + "loss": 0.0001, + "step": 15763 + }, + { + "epoch": 3.64, + "learning_rate": 1.9640243432709346e-05, + "loss": 0.0, + "step": 15764 + }, + { + "epoch": 3.64, + "learning_rate": 1.9638317541021493e-05, + "loss": 0.0, + "step": 15765 + }, + { + "epoch": 3.64, + "learning_rate": 1.9636391649333644e-05, + "loss": 0.0, + "step": 15766 + }, + { + "epoch": 3.64, + "learning_rate": 1.9634465757645788e-05, + "loss": 0.0, + "step": 15767 + }, + { + "epoch": 3.64, + "learning_rate": 1.963253986595794e-05, + "loss": 0.0, + "step": 15768 + }, + { + "epoch": 3.64, + "learning_rate": 1.9630613974270086e-05, + "loss": 0.0, + "step": 15769 + }, + { + "epoch": 3.64, + "learning_rate": 1.9628688082582237e-05, + "loss": 0.0, + "step": 15770 + }, + { + "epoch": 3.64, + "learning_rate": 1.9626762190894385e-05, + "loss": 0.0, + "step": 15771 + }, + { + "epoch": 3.65, + "learning_rate": 1.9624836299206532e-05, + "loss": 0.0, + "step": 15772 + }, + { + "epoch": 3.65, + "learning_rate": 1.9622910407518683e-05, + "loss": 0.0, + "step": 15773 + }, + { + "epoch": 3.65, + "learning_rate": 1.962098451583083e-05, + "loss": 0.0, + "step": 15774 + }, + { + "epoch": 3.65, + "learning_rate": 1.961905862414298e-05, + "loss": 0.0, + "step": 15775 + }, + { + "epoch": 3.65, + "learning_rate": 1.9617132732455126e-05, + "loss": 0.0, + "step": 15776 + }, + { + "epoch": 3.65, + "learning_rate": 1.9615206840767276e-05, + "loss": 0.0, + "step": 15777 + }, + { + "epoch": 3.65, + "learning_rate": 1.9613280949079424e-05, + "loss": 0.0, + "step": 15778 + }, + { + "epoch": 3.65, + "learning_rate": 1.961135505739157e-05, + "loss": 0.0, + "step": 15779 + }, + { + "epoch": 3.65, + "learning_rate": 1.9609429165703722e-05, + "loss": 0.0, + "step": 15780 + }, + { + "epoch": 3.65, + "learning_rate": 1.960750327401587e-05, + "loss": 0.0, + "step": 15781 + }, + { + "epoch": 3.65, + "learning_rate": 1.960557738232802e-05, + "loss": 0.0, + "step": 15782 + }, + { + "epoch": 3.65, + "learning_rate": 1.9603651490640168e-05, + "loss": 0.0, + "step": 15783 + }, + { + "epoch": 3.65, + "learning_rate": 1.9601725598952316e-05, + "loss": 0.0, + "step": 15784 + }, + { + "epoch": 3.65, + "learning_rate": 1.9599799707264466e-05, + "loss": 0.0, + "step": 15785 + }, + { + "epoch": 3.65, + "learning_rate": 1.959787381557661e-05, + "loss": 0.0016, + "step": 15786 + }, + { + "epoch": 3.65, + "learning_rate": 1.959594792388876e-05, + "loss": 0.0001, + "step": 15787 + }, + { + "epoch": 3.65, + "learning_rate": 1.959402203220091e-05, + "loss": 0.0, + "step": 15788 + }, + { + "epoch": 3.65, + "learning_rate": 1.959209614051306e-05, + "loss": 0.0007, + "step": 15789 + }, + { + "epoch": 3.65, + "learning_rate": 1.9590170248825207e-05, + "loss": 0.0, + "step": 15790 + }, + { + "epoch": 3.65, + "learning_rate": 1.9588244357137355e-05, + "loss": 0.0, + "step": 15791 + }, + { + "epoch": 3.65, + "learning_rate": 1.9586318465449506e-05, + "loss": 0.0, + "step": 15792 + }, + { + "epoch": 3.65, + "learning_rate": 1.9584392573761653e-05, + "loss": 0.0, + "step": 15793 + }, + { + "epoch": 3.65, + "learning_rate": 1.9582466682073804e-05, + "loss": 0.0, + "step": 15794 + }, + { + "epoch": 3.65, + "learning_rate": 1.9580540790385948e-05, + "loss": 0.0, + "step": 15795 + }, + { + "epoch": 3.65, + "learning_rate": 1.95786148986981e-05, + "loss": 0.0, + "step": 15796 + }, + { + "epoch": 3.65, + "learning_rate": 1.9576689007010246e-05, + "loss": 0.0, + "step": 15797 + }, + { + "epoch": 3.65, + "learning_rate": 1.9574763115322394e-05, + "loss": 0.0, + "step": 15798 + }, + { + "epoch": 3.65, + "learning_rate": 1.9572837223634545e-05, + "loss": 0.0, + "step": 15799 + }, + { + "epoch": 3.65, + "learning_rate": 1.9570911331946692e-05, + "loss": 0.0018, + "step": 15800 + }, + { + "epoch": 3.65, + "learning_rate": 1.9568985440258843e-05, + "loss": 0.0001, + "step": 15801 + }, + { + "epoch": 3.65, + "learning_rate": 1.956705954857099e-05, + "loss": 0.0, + "step": 15802 + }, + { + "epoch": 3.65, + "learning_rate": 1.9565133656883138e-05, + "loss": 0.0001, + "step": 15803 + }, + { + "epoch": 3.65, + "learning_rate": 1.9563207765195285e-05, + "loss": 0.0, + "step": 15804 + }, + { + "epoch": 3.65, + "learning_rate": 1.9561281873507433e-05, + "loss": 0.0, + "step": 15805 + }, + { + "epoch": 3.65, + "learning_rate": 1.9559355981819584e-05, + "loss": 0.0, + "step": 15806 + }, + { + "epoch": 3.65, + "learning_rate": 1.955743009013173e-05, + "loss": 0.0, + "step": 15807 + }, + { + "epoch": 3.65, + "learning_rate": 1.955550419844388e-05, + "loss": 0.0, + "step": 15808 + }, + { + "epoch": 3.65, + "learning_rate": 1.955357830675603e-05, + "loss": 0.0, + "step": 15809 + }, + { + "epoch": 3.65, + "learning_rate": 1.9551652415068177e-05, + "loss": 0.0, + "step": 15810 + }, + { + "epoch": 3.65, + "learning_rate": 1.9549726523380328e-05, + "loss": 0.0, + "step": 15811 + }, + { + "epoch": 3.65, + "learning_rate": 1.9547800631692475e-05, + "loss": 0.0, + "step": 15812 + }, + { + "epoch": 3.65, + "learning_rate": 1.9545874740004623e-05, + "loss": 0.0, + "step": 15813 + }, + { + "epoch": 3.65, + "learning_rate": 1.954394884831677e-05, + "loss": 0.0, + "step": 15814 + }, + { + "epoch": 3.65, + "learning_rate": 1.9542022956628918e-05, + "loss": 0.0, + "step": 15815 + }, + { + "epoch": 3.66, + "learning_rate": 1.954009706494107e-05, + "loss": 0.0, + "step": 15816 + }, + { + "epoch": 3.66, + "learning_rate": 1.9538171173253216e-05, + "loss": 0.0, + "step": 15817 + }, + { + "epoch": 3.66, + "learning_rate": 1.9536245281565367e-05, + "loss": 0.0, + "step": 15818 + }, + { + "epoch": 3.66, + "learning_rate": 1.9534319389877515e-05, + "loss": 0.0, + "step": 15819 + }, + { + "epoch": 3.66, + "learning_rate": 1.9532393498189662e-05, + "loss": 0.0, + "step": 15820 + }, + { + "epoch": 3.66, + "learning_rate": 1.9530467606501813e-05, + "loss": 0.0, + "step": 15821 + }, + { + "epoch": 3.66, + "learning_rate": 1.952854171481396e-05, + "loss": 0.0002, + "step": 15822 + }, + { + "epoch": 3.66, + "learning_rate": 1.9526615823126108e-05, + "loss": 0.0001, + "step": 15823 + }, + { + "epoch": 3.66, + "learning_rate": 1.9524689931438255e-05, + "loss": 0.0, + "step": 15824 + }, + { + "epoch": 3.66, + "learning_rate": 1.9522764039750406e-05, + "loss": 0.0, + "step": 15825 + }, + { + "epoch": 3.66, + "learning_rate": 1.9520838148062554e-05, + "loss": 0.0, + "step": 15826 + }, + { + "epoch": 3.66, + "learning_rate": 1.95189122563747e-05, + "loss": 0.0, + "step": 15827 + }, + { + "epoch": 3.66, + "learning_rate": 1.9516986364686852e-05, + "loss": 0.0, + "step": 15828 + }, + { + "epoch": 3.66, + "learning_rate": 1.9515060472999e-05, + "loss": 0.0, + "step": 15829 + }, + { + "epoch": 3.66, + "learning_rate": 1.951313458131115e-05, + "loss": 0.0, + "step": 15830 + }, + { + "epoch": 3.66, + "learning_rate": 1.9511208689623298e-05, + "loss": 0.0, + "step": 15831 + }, + { + "epoch": 3.66, + "learning_rate": 1.9509282797935445e-05, + "loss": 0.0, + "step": 15832 + }, + { + "epoch": 3.66, + "learning_rate": 1.9507356906247593e-05, + "loss": 0.0, + "step": 15833 + }, + { + "epoch": 3.66, + "learning_rate": 1.950543101455974e-05, + "loss": 0.0, + "step": 15834 + }, + { + "epoch": 3.66, + "learning_rate": 1.950350512287189e-05, + "loss": 0.0, + "step": 15835 + }, + { + "epoch": 3.66, + "learning_rate": 1.950157923118404e-05, + "loss": 0.0, + "step": 15836 + }, + { + "epoch": 3.66, + "learning_rate": 1.949965333949619e-05, + "loss": 0.0, + "step": 15837 + }, + { + "epoch": 3.66, + "learning_rate": 1.9497727447808337e-05, + "loss": 0.0006, + "step": 15838 + }, + { + "epoch": 3.66, + "learning_rate": 1.9495801556120484e-05, + "loss": 0.0, + "step": 15839 + }, + { + "epoch": 3.66, + "learning_rate": 1.9493875664432635e-05, + "loss": 0.0, + "step": 15840 + }, + { + "epoch": 3.66, + "learning_rate": 1.9491949772744783e-05, + "loss": 0.0, + "step": 15841 + }, + { + "epoch": 3.66, + "learning_rate": 1.949002388105693e-05, + "loss": 0.0, + "step": 15842 + }, + { + "epoch": 3.66, + "learning_rate": 1.9488097989369078e-05, + "loss": 0.0, + "step": 15843 + }, + { + "epoch": 3.66, + "learning_rate": 1.9486172097681225e-05, + "loss": 0.0, + "step": 15844 + }, + { + "epoch": 3.66, + "learning_rate": 1.9484246205993376e-05, + "loss": 0.0, + "step": 15845 + }, + { + "epoch": 3.66, + "learning_rate": 1.9482320314305524e-05, + "loss": 0.0001, + "step": 15846 + }, + { + "epoch": 3.66, + "learning_rate": 1.9480394422617674e-05, + "loss": 0.0001, + "step": 15847 + }, + { + "epoch": 3.66, + "learning_rate": 1.9478468530929822e-05, + "loss": 0.0, + "step": 15848 + }, + { + "epoch": 3.66, + "learning_rate": 1.947654263924197e-05, + "loss": 0.0, + "step": 15849 + }, + { + "epoch": 3.66, + "learning_rate": 1.947461674755412e-05, + "loss": 0.0, + "step": 15850 + }, + { + "epoch": 3.66, + "learning_rate": 1.9472690855866264e-05, + "loss": 0.0002, + "step": 15851 + }, + { + "epoch": 3.66, + "learning_rate": 1.9470764964178415e-05, + "loss": 0.0, + "step": 15852 + }, + { + "epoch": 3.66, + "learning_rate": 1.9468839072490563e-05, + "loss": 0.0, + "step": 15853 + }, + { + "epoch": 3.66, + "learning_rate": 1.9466913180802714e-05, + "loss": 0.0017, + "step": 15854 + }, + { + "epoch": 3.66, + "learning_rate": 1.946498728911486e-05, + "loss": 0.0, + "step": 15855 + }, + { + "epoch": 3.66, + "learning_rate": 1.946306139742701e-05, + "loss": 0.0007, + "step": 15856 + }, + { + "epoch": 3.66, + "learning_rate": 1.946113550573916e-05, + "loss": 0.0, + "step": 15857 + }, + { + "epoch": 3.66, + "learning_rate": 1.9459209614051307e-05, + "loss": 0.0001, + "step": 15858 + }, + { + "epoch": 3.67, + "learning_rate": 1.9457283722363458e-05, + "loss": 0.0, + "step": 15859 + }, + { + "epoch": 3.67, + "learning_rate": 1.9455357830675602e-05, + "loss": 0.0, + "step": 15860 + }, + { + "epoch": 3.67, + "learning_rate": 1.9453431938987753e-05, + "loss": 0.0, + "step": 15861 + }, + { + "epoch": 3.67, + "learning_rate": 1.94515060472999e-05, + "loss": 0.0, + "step": 15862 + }, + { + "epoch": 3.67, + "learning_rate": 1.9449580155612048e-05, + "loss": 0.0, + "step": 15863 + }, + { + "epoch": 3.67, + "learning_rate": 1.94476542639242e-05, + "loss": 0.0, + "step": 15864 + }, + { + "epoch": 3.67, + "learning_rate": 1.9445728372236346e-05, + "loss": 0.0, + "step": 15865 + }, + { + "epoch": 3.67, + "learning_rate": 1.9443802480548497e-05, + "loss": 0.0, + "step": 15866 + }, + { + "epoch": 3.67, + "learning_rate": 1.9441876588860644e-05, + "loss": 0.0, + "step": 15867 + }, + { + "epoch": 3.67, + "learning_rate": 1.9439950697172792e-05, + "loss": 0.0, + "step": 15868 + }, + { + "epoch": 3.67, + "learning_rate": 1.9438024805484943e-05, + "loss": 0.0, + "step": 15869 + }, + { + "epoch": 3.67, + "learning_rate": 1.9436098913797087e-05, + "loss": 0.0, + "step": 15870 + }, + { + "epoch": 3.67, + "learning_rate": 1.9434173022109238e-05, + "loss": 0.0, + "step": 15871 + }, + { + "epoch": 3.67, + "learning_rate": 1.9432247130421385e-05, + "loss": 0.0004, + "step": 15872 + }, + { + "epoch": 3.67, + "learning_rate": 1.9430321238733536e-05, + "loss": 0.0, + "step": 15873 + }, + { + "epoch": 3.67, + "learning_rate": 1.9428395347045683e-05, + "loss": 0.0, + "step": 15874 + }, + { + "epoch": 3.67, + "learning_rate": 1.942646945535783e-05, + "loss": 0.0, + "step": 15875 + }, + { + "epoch": 3.67, + "learning_rate": 1.9424543563669982e-05, + "loss": 0.0, + "step": 15876 + }, + { + "epoch": 3.67, + "learning_rate": 1.942261767198213e-05, + "loss": 0.0, + "step": 15877 + }, + { + "epoch": 3.67, + "learning_rate": 1.9420691780294277e-05, + "loss": 0.0, + "step": 15878 + }, + { + "epoch": 3.67, + "learning_rate": 1.9418765888606424e-05, + "loss": 0.0, + "step": 15879 + }, + { + "epoch": 3.67, + "learning_rate": 1.941683999691857e-05, + "loss": 0.0, + "step": 15880 + }, + { + "epoch": 3.67, + "learning_rate": 1.9414914105230723e-05, + "loss": 0.0, + "step": 15881 + }, + { + "epoch": 3.67, + "learning_rate": 1.941298821354287e-05, + "loss": 0.0, + "step": 15882 + }, + { + "epoch": 3.67, + "learning_rate": 1.941106232185502e-05, + "loss": 0.0, + "step": 15883 + }, + { + "epoch": 3.67, + "learning_rate": 1.940913643016717e-05, + "loss": 0.0, + "step": 15884 + }, + { + "epoch": 3.67, + "learning_rate": 1.9407210538479316e-05, + "loss": 0.0, + "step": 15885 + }, + { + "epoch": 3.67, + "learning_rate": 1.9405284646791467e-05, + "loss": 0.0, + "step": 15886 + }, + { + "epoch": 3.67, + "learning_rate": 1.9403358755103614e-05, + "loss": 0.0, + "step": 15887 + }, + { + "epoch": 3.67, + "learning_rate": 1.940143286341576e-05, + "loss": 0.0, + "step": 15888 + }, + { + "epoch": 3.67, + "learning_rate": 1.939950697172791e-05, + "loss": 0.0, + "step": 15889 + }, + { + "epoch": 3.67, + "learning_rate": 1.939758108004006e-05, + "loss": 0.0, + "step": 15890 + }, + { + "epoch": 3.67, + "learning_rate": 1.9395655188352207e-05, + "loss": 0.0, + "step": 15891 + }, + { + "epoch": 3.67, + "learning_rate": 1.9393729296664355e-05, + "loss": 0.0, + "step": 15892 + }, + { + "epoch": 3.67, + "learning_rate": 1.9391803404976506e-05, + "loss": 0.0, + "step": 15893 + }, + { + "epoch": 3.67, + "learning_rate": 1.9389877513288653e-05, + "loss": 0.0, + "step": 15894 + }, + { + "epoch": 3.67, + "learning_rate": 1.9387951621600804e-05, + "loss": 0.0, + "step": 15895 + }, + { + "epoch": 3.67, + "learning_rate": 1.938602572991295e-05, + "loss": 0.0, + "step": 15896 + }, + { + "epoch": 3.67, + "learning_rate": 1.93840998382251e-05, + "loss": 0.0, + "step": 15897 + }, + { + "epoch": 3.67, + "learning_rate": 1.9382173946537247e-05, + "loss": 0.0, + "step": 15898 + }, + { + "epoch": 3.67, + "learning_rate": 1.9380248054849394e-05, + "loss": 0.0, + "step": 15899 + }, + { + "epoch": 3.67, + "learning_rate": 1.9378322163161545e-05, + "loss": 0.0004, + "step": 15900 + }, + { + "epoch": 3.67, + "learning_rate": 1.9376396271473692e-05, + "loss": 0.0, + "step": 15901 + }, + { + "epoch": 3.68, + "learning_rate": 1.9374470379785843e-05, + "loss": 0.0, + "step": 15902 + }, + { + "epoch": 3.68, + "learning_rate": 1.937254448809799e-05, + "loss": 0.0, + "step": 15903 + }, + { + "epoch": 3.68, + "learning_rate": 1.9370618596410138e-05, + "loss": 0.0, + "step": 15904 + }, + { + "epoch": 3.68, + "learning_rate": 1.936869270472229e-05, + "loss": 0.0, + "step": 15905 + }, + { + "epoch": 3.68, + "learning_rate": 1.9366766813034437e-05, + "loss": 0.0, + "step": 15906 + }, + { + "epoch": 3.68, + "learning_rate": 1.9364840921346584e-05, + "loss": 0.0002, + "step": 15907 + }, + { + "epoch": 3.68, + "learning_rate": 1.936291502965873e-05, + "loss": 0.0, + "step": 15908 + }, + { + "epoch": 3.68, + "learning_rate": 1.9360989137970882e-05, + "loss": 0.0001, + "step": 15909 + }, + { + "epoch": 3.68, + "learning_rate": 1.935906324628303e-05, + "loss": 0.0, + "step": 15910 + }, + { + "epoch": 3.68, + "learning_rate": 1.9357137354595177e-05, + "loss": 0.0, + "step": 15911 + }, + { + "epoch": 3.68, + "learning_rate": 1.9355211462907328e-05, + "loss": 0.0, + "step": 15912 + }, + { + "epoch": 3.68, + "learning_rate": 1.9353285571219476e-05, + "loss": 0.0, + "step": 15913 + }, + { + "epoch": 3.68, + "learning_rate": 1.9351359679531627e-05, + "loss": 0.0, + "step": 15914 + }, + { + "epoch": 3.68, + "learning_rate": 1.9349433787843774e-05, + "loss": 0.0, + "step": 15915 + }, + { + "epoch": 3.68, + "learning_rate": 1.9347507896155918e-05, + "loss": 0.0, + "step": 15916 + }, + { + "epoch": 3.68, + "learning_rate": 1.934558200446807e-05, + "loss": 0.0, + "step": 15917 + }, + { + "epoch": 3.68, + "learning_rate": 1.9343656112780216e-05, + "loss": 0.0, + "step": 15918 + }, + { + "epoch": 3.68, + "learning_rate": 1.9341730221092367e-05, + "loss": 0.0, + "step": 15919 + }, + { + "epoch": 3.68, + "learning_rate": 1.9339804329404515e-05, + "loss": 0.0, + "step": 15920 + }, + { + "epoch": 3.68, + "learning_rate": 1.9337878437716662e-05, + "loss": 0.0, + "step": 15921 + }, + { + "epoch": 3.68, + "learning_rate": 1.9335952546028813e-05, + "loss": 0.0002, + "step": 15922 + }, + { + "epoch": 3.68, + "learning_rate": 1.933402665434096e-05, + "loss": 0.0, + "step": 15923 + }, + { + "epoch": 3.68, + "learning_rate": 1.933210076265311e-05, + "loss": 0.0, + "step": 15924 + }, + { + "epoch": 3.68, + "learning_rate": 1.933017487096526e-05, + "loss": 0.0, + "step": 15925 + }, + { + "epoch": 3.68, + "learning_rate": 1.9328248979277406e-05, + "loss": 0.0, + "step": 15926 + }, + { + "epoch": 3.68, + "learning_rate": 1.9326323087589554e-05, + "loss": 0.0, + "step": 15927 + }, + { + "epoch": 3.68, + "learning_rate": 1.93243971959017e-05, + "loss": 0.0, + "step": 15928 + }, + { + "epoch": 3.68, + "learning_rate": 1.9322471304213852e-05, + "loss": 0.0002, + "step": 15929 + }, + { + "epoch": 3.68, + "learning_rate": 1.9320545412526e-05, + "loss": 0.0, + "step": 15930 + }, + { + "epoch": 3.68, + "learning_rate": 1.931861952083815e-05, + "loss": 0.0, + "step": 15931 + }, + { + "epoch": 3.68, + "learning_rate": 1.9316693629150298e-05, + "loss": 0.0, + "step": 15932 + }, + { + "epoch": 3.68, + "learning_rate": 1.9314767737462446e-05, + "loss": 0.0, + "step": 15933 + }, + { + "epoch": 3.68, + "learning_rate": 1.9312841845774596e-05, + "loss": 0.0, + "step": 15934 + }, + { + "epoch": 3.68, + "learning_rate": 1.931091595408674e-05, + "loss": 0.0, + "step": 15935 + }, + { + "epoch": 3.68, + "learning_rate": 1.930899006239889e-05, + "loss": 0.0, + "step": 15936 + }, + { + "epoch": 3.68, + "learning_rate": 1.930706417071104e-05, + "loss": 0.0, + "step": 15937 + }, + { + "epoch": 3.68, + "learning_rate": 1.930513827902319e-05, + "loss": 0.0, + "step": 15938 + }, + { + "epoch": 3.68, + "learning_rate": 1.9303212387335337e-05, + "loss": 0.0, + "step": 15939 + }, + { + "epoch": 3.68, + "learning_rate": 1.9301286495647485e-05, + "loss": 0.0, + "step": 15940 + }, + { + "epoch": 3.68, + "learning_rate": 1.9299360603959636e-05, + "loss": 0.0, + "step": 15941 + }, + { + "epoch": 3.68, + "learning_rate": 1.9297434712271783e-05, + "loss": 0.0, + "step": 15942 + }, + { + "epoch": 3.68, + "learning_rate": 1.9295508820583934e-05, + "loss": 0.0, + "step": 15943 + }, + { + "epoch": 3.68, + "learning_rate": 1.9293582928896078e-05, + "loss": 0.0, + "step": 15944 + }, + { + "epoch": 3.69, + "learning_rate": 1.929165703720823e-05, + "loss": 0.0, + "step": 15945 + }, + { + "epoch": 3.69, + "learning_rate": 1.9289731145520376e-05, + "loss": 0.0, + "step": 15946 + }, + { + "epoch": 3.69, + "learning_rate": 1.9287805253832524e-05, + "loss": 0.0, + "step": 15947 + }, + { + "epoch": 3.69, + "learning_rate": 1.9285879362144675e-05, + "loss": 0.0, + "step": 15948 + }, + { + "epoch": 3.69, + "learning_rate": 1.9283953470456822e-05, + "loss": 0.0, + "step": 15949 + }, + { + "epoch": 3.69, + "learning_rate": 1.9282027578768973e-05, + "loss": 0.0, + "step": 15950 + }, + { + "epoch": 3.69, + "learning_rate": 1.928010168708112e-05, + "loss": 0.0, + "step": 15951 + }, + { + "epoch": 3.69, + "learning_rate": 1.9278175795393268e-05, + "loss": 0.0, + "step": 15952 + }, + { + "epoch": 3.69, + "learning_rate": 1.927624990370542e-05, + "loss": 0.0001, + "step": 15953 + }, + { + "epoch": 3.69, + "learning_rate": 1.9274324012017563e-05, + "loss": 0.0, + "step": 15954 + }, + { + "epoch": 3.69, + "learning_rate": 1.9272398120329714e-05, + "loss": 0.0, + "step": 15955 + }, + { + "epoch": 3.69, + "learning_rate": 1.927047222864186e-05, + "loss": 0.0, + "step": 15956 + }, + { + "epoch": 3.69, + "learning_rate": 1.926854633695401e-05, + "loss": 0.0, + "step": 15957 + }, + { + "epoch": 3.69, + "learning_rate": 1.926662044526616e-05, + "loss": 0.0, + "step": 15958 + }, + { + "epoch": 3.69, + "learning_rate": 1.9264694553578307e-05, + "loss": 0.0, + "step": 15959 + }, + { + "epoch": 3.69, + "learning_rate": 1.9262768661890458e-05, + "loss": 0.0, + "step": 15960 + }, + { + "epoch": 3.69, + "learning_rate": 1.9260842770202605e-05, + "loss": 0.0, + "step": 15961 + }, + { + "epoch": 3.69, + "learning_rate": 1.9258916878514753e-05, + "loss": 0.0, + "step": 15962 + }, + { + "epoch": 3.69, + "learning_rate": 1.92569909868269e-05, + "loss": 0.0001, + "step": 15963 + }, + { + "epoch": 3.69, + "learning_rate": 1.9255065095139048e-05, + "loss": 0.062, + "step": 15964 + }, + { + "epoch": 3.69, + "learning_rate": 1.92531392034512e-05, + "loss": 0.0001, + "step": 15965 + }, + { + "epoch": 3.69, + "learning_rate": 1.9251213311763346e-05, + "loss": 0.0, + "step": 15966 + }, + { + "epoch": 3.69, + "learning_rate": 1.9249287420075497e-05, + "loss": 0.0, + "step": 15967 + }, + { + "epoch": 3.69, + "learning_rate": 1.9247361528387645e-05, + "loss": 0.0, + "step": 15968 + }, + { + "epoch": 3.69, + "learning_rate": 1.9245435636699792e-05, + "loss": 0.0, + "step": 15969 + }, + { + "epoch": 3.69, + "learning_rate": 1.9243509745011943e-05, + "loss": 0.0, + "step": 15970 + }, + { + "epoch": 3.69, + "learning_rate": 1.924158385332409e-05, + "loss": 0.0, + "step": 15971 + }, + { + "epoch": 3.69, + "learning_rate": 1.9239657961636238e-05, + "loss": 0.0, + "step": 15972 + }, + { + "epoch": 3.69, + "learning_rate": 1.9237732069948385e-05, + "loss": 0.0, + "step": 15973 + }, + { + "epoch": 3.69, + "learning_rate": 1.9235806178260536e-05, + "loss": 0.0, + "step": 15974 + }, + { + "epoch": 3.69, + "learning_rate": 1.9233880286572684e-05, + "loss": 0.0, + "step": 15975 + }, + { + "epoch": 3.69, + "learning_rate": 1.923195439488483e-05, + "loss": 0.0, + "step": 15976 + }, + { + "epoch": 3.69, + "learning_rate": 1.9230028503196982e-05, + "loss": 0.0, + "step": 15977 + }, + { + "epoch": 3.69, + "learning_rate": 1.922810261150913e-05, + "loss": 0.0, + "step": 15978 + }, + { + "epoch": 3.69, + "learning_rate": 1.922617671982128e-05, + "loss": 0.0, + "step": 15979 + }, + { + "epoch": 3.69, + "learning_rate": 1.9224250828133428e-05, + "loss": 0.0, + "step": 15980 + }, + { + "epoch": 3.69, + "learning_rate": 1.9222324936445575e-05, + "loss": 0.0, + "step": 15981 + }, + { + "epoch": 3.69, + "learning_rate": 1.9220399044757723e-05, + "loss": 0.0, + "step": 15982 + }, + { + "epoch": 3.69, + "learning_rate": 1.921847315306987e-05, + "loss": 0.0, + "step": 15983 + }, + { + "epoch": 3.69, + "learning_rate": 1.921654726138202e-05, + "loss": 0.0, + "step": 15984 + }, + { + "epoch": 3.69, + "learning_rate": 1.921462136969417e-05, + "loss": 0.0, + "step": 15985 + }, + { + "epoch": 3.69, + "learning_rate": 1.921269547800632e-05, + "loss": 0.0, + "step": 15986 + }, + { + "epoch": 3.69, + "learning_rate": 1.9210769586318467e-05, + "loss": 0.0, + "step": 15987 + }, + { + "epoch": 3.69, + "learning_rate": 1.9208843694630614e-05, + "loss": 0.0, + "step": 15988 + }, + { + "epoch": 3.7, + "learning_rate": 1.9206917802942765e-05, + "loss": 0.0, + "step": 15989 + }, + { + "epoch": 3.7, + "learning_rate": 1.9204991911254913e-05, + "loss": 0.0, + "step": 15990 + }, + { + "epoch": 3.7, + "learning_rate": 1.920306601956706e-05, + "loss": 0.0, + "step": 15991 + }, + { + "epoch": 3.7, + "learning_rate": 1.9201140127879208e-05, + "loss": 0.0, + "step": 15992 + }, + { + "epoch": 3.7, + "learning_rate": 1.9199214236191355e-05, + "loss": 0.0, + "step": 15993 + }, + { + "epoch": 3.7, + "learning_rate": 1.9197288344503506e-05, + "loss": 0.0, + "step": 15994 + }, + { + "epoch": 3.7, + "learning_rate": 1.9195362452815654e-05, + "loss": 0.0, + "step": 15995 + }, + { + "epoch": 3.7, + "learning_rate": 1.9193436561127804e-05, + "loss": 0.0, + "step": 15996 + }, + { + "epoch": 3.7, + "learning_rate": 1.9191510669439952e-05, + "loss": 0.0, + "step": 15997 + }, + { + "epoch": 3.7, + "learning_rate": 1.91895847777521e-05, + "loss": 0.0, + "step": 15998 + }, + { + "epoch": 3.7, + "learning_rate": 1.918765888606425e-05, + "loss": 0.0, + "step": 15999 + }, + { + "epoch": 3.7, + "learning_rate": 1.9185732994376394e-05, + "loss": 0.0, + "step": 16000 + }, + { + "epoch": 3.7, + "learning_rate": 1.9183807102688545e-05, + "loss": 0.0, + "step": 16001 + }, + { + "epoch": 3.7, + "learning_rate": 1.9181881211000693e-05, + "loss": 0.0, + "step": 16002 + }, + { + "epoch": 3.7, + "learning_rate": 1.9179955319312844e-05, + "loss": 0.0, + "step": 16003 + }, + { + "epoch": 3.7, + "learning_rate": 1.917802942762499e-05, + "loss": 0.0, + "step": 16004 + }, + { + "epoch": 3.7, + "learning_rate": 1.917610353593714e-05, + "loss": 0.0, + "step": 16005 + }, + { + "epoch": 3.7, + "learning_rate": 1.917417764424929e-05, + "loss": 0.0, + "step": 16006 + }, + { + "epoch": 3.7, + "learning_rate": 1.9172251752561437e-05, + "loss": 0.0, + "step": 16007 + }, + { + "epoch": 3.7, + "learning_rate": 1.9170325860873588e-05, + "loss": 0.0, + "step": 16008 + }, + { + "epoch": 3.7, + "learning_rate": 1.9168399969185735e-05, + "loss": 0.0, + "step": 16009 + }, + { + "epoch": 3.7, + "learning_rate": 1.9166474077497883e-05, + "loss": 0.0, + "step": 16010 + }, + { + "epoch": 3.7, + "learning_rate": 1.916454818581003e-05, + "loss": 0.0, + "step": 16011 + }, + { + "epoch": 3.7, + "learning_rate": 1.9162622294122178e-05, + "loss": 0.0, + "step": 16012 + }, + { + "epoch": 3.7, + "learning_rate": 1.916069640243433e-05, + "loss": 0.0, + "step": 16013 + }, + { + "epoch": 3.7, + "learning_rate": 1.9158770510746476e-05, + "loss": 0.0, + "step": 16014 + }, + { + "epoch": 3.7, + "learning_rate": 1.9156844619058627e-05, + "loss": 0.0, + "step": 16015 + }, + { + "epoch": 3.7, + "learning_rate": 1.9154918727370774e-05, + "loss": 0.0002, + "step": 16016 + }, + { + "epoch": 3.7, + "learning_rate": 1.9152992835682922e-05, + "loss": 0.0, + "step": 16017 + }, + { + "epoch": 3.7, + "learning_rate": 1.9151066943995073e-05, + "loss": 0.0, + "step": 16018 + }, + { + "epoch": 3.7, + "learning_rate": 1.9149141052307217e-05, + "loss": 0.0, + "step": 16019 + }, + { + "epoch": 3.7, + "learning_rate": 1.9147215160619368e-05, + "loss": 0.0, + "step": 16020 + }, + { + "epoch": 3.7, + "learning_rate": 1.9145289268931515e-05, + "loss": 0.0, + "step": 16021 + }, + { + "epoch": 3.7, + "learning_rate": 1.9143363377243666e-05, + "loss": 0.0, + "step": 16022 + }, + { + "epoch": 3.7, + "learning_rate": 1.9141437485555813e-05, + "loss": 0.0, + "step": 16023 + }, + { + "epoch": 3.7, + "learning_rate": 1.913951159386796e-05, + "loss": 0.0, + "step": 16024 + }, + { + "epoch": 3.7, + "learning_rate": 1.9137585702180112e-05, + "loss": 0.0, + "step": 16025 + }, + { + "epoch": 3.7, + "learning_rate": 1.913565981049226e-05, + "loss": 0.0, + "step": 16026 + }, + { + "epoch": 3.7, + "learning_rate": 1.913373391880441e-05, + "loss": 0.0, + "step": 16027 + }, + { + "epoch": 3.7, + "learning_rate": 1.9131808027116554e-05, + "loss": 0.0, + "step": 16028 + }, + { + "epoch": 3.7, + "learning_rate": 1.91298821354287e-05, + "loss": 0.0, + "step": 16029 + }, + { + "epoch": 3.7, + "learning_rate": 1.9127956243740853e-05, + "loss": 0.0, + "step": 16030 + }, + { + "epoch": 3.7, + "learning_rate": 1.9126030352053e-05, + "loss": 0.0, + "step": 16031 + }, + { + "epoch": 3.71, + "learning_rate": 1.912410446036515e-05, + "loss": 0.0, + "step": 16032 + }, + { + "epoch": 3.71, + "learning_rate": 1.91221785686773e-05, + "loss": 0.0, + "step": 16033 + }, + { + "epoch": 3.71, + "learning_rate": 1.9120252676989446e-05, + "loss": 0.0, + "step": 16034 + }, + { + "epoch": 3.71, + "learning_rate": 1.9118326785301597e-05, + "loss": 0.0, + "step": 16035 + }, + { + "epoch": 3.71, + "learning_rate": 1.9116400893613744e-05, + "loss": 0.0, + "step": 16036 + }, + { + "epoch": 3.71, + "learning_rate": 1.9114475001925895e-05, + "loss": 0.0, + "step": 16037 + }, + { + "epoch": 3.71, + "learning_rate": 1.911254911023804e-05, + "loss": 0.0, + "step": 16038 + }, + { + "epoch": 3.71, + "learning_rate": 1.911062321855019e-05, + "loss": 0.0, + "step": 16039 + }, + { + "epoch": 3.71, + "learning_rate": 1.9108697326862337e-05, + "loss": 0.0, + "step": 16040 + }, + { + "epoch": 3.71, + "learning_rate": 1.9106771435174485e-05, + "loss": 0.0, + "step": 16041 + }, + { + "epoch": 3.71, + "learning_rate": 1.9104845543486636e-05, + "loss": 0.0, + "step": 16042 + }, + { + "epoch": 3.71, + "learning_rate": 1.9102919651798783e-05, + "loss": 0.0, + "step": 16043 + }, + { + "epoch": 3.71, + "learning_rate": 1.9100993760110934e-05, + "loss": 0.0, + "step": 16044 + }, + { + "epoch": 3.71, + "learning_rate": 1.909906786842308e-05, + "loss": 0.0, + "step": 16045 + }, + { + "epoch": 3.71, + "learning_rate": 1.909714197673523e-05, + "loss": 0.0, + "step": 16046 + }, + { + "epoch": 3.71, + "learning_rate": 1.9095216085047377e-05, + "loss": 0.0, + "step": 16047 + }, + { + "epoch": 3.71, + "learning_rate": 1.9093290193359524e-05, + "loss": 0.0, + "step": 16048 + }, + { + "epoch": 3.71, + "learning_rate": 1.9091364301671675e-05, + "loss": 0.0, + "step": 16049 + }, + { + "epoch": 3.71, + "learning_rate": 1.9089438409983822e-05, + "loss": 0.0001, + "step": 16050 + }, + { + "epoch": 3.71, + "learning_rate": 1.9087512518295973e-05, + "loss": 0.001, + "step": 16051 + }, + { + "epoch": 3.71, + "learning_rate": 1.908558662660812e-05, + "loss": 0.0, + "step": 16052 + }, + { + "epoch": 3.71, + "learning_rate": 1.9083660734920268e-05, + "loss": 0.0, + "step": 16053 + }, + { + "epoch": 3.71, + "learning_rate": 1.908173484323242e-05, + "loss": 0.0, + "step": 16054 + }, + { + "epoch": 3.71, + "learning_rate": 1.9079808951544567e-05, + "loss": 0.0, + "step": 16055 + }, + { + "epoch": 3.71, + "learning_rate": 1.9077883059856714e-05, + "loss": 0.0, + "step": 16056 + }, + { + "epoch": 3.71, + "learning_rate": 1.907595716816886e-05, + "loss": 0.0001, + "step": 16057 + }, + { + "epoch": 3.71, + "learning_rate": 1.9074031276481012e-05, + "loss": 0.0, + "step": 16058 + }, + { + "epoch": 3.71, + "learning_rate": 1.907210538479316e-05, + "loss": 0.0, + "step": 16059 + }, + { + "epoch": 3.71, + "learning_rate": 1.9070179493105307e-05, + "loss": 0.0, + "step": 16060 + }, + { + "epoch": 3.71, + "learning_rate": 1.9068253601417458e-05, + "loss": 0.0, + "step": 16061 + }, + { + "epoch": 3.71, + "learning_rate": 1.9066327709729606e-05, + "loss": 0.0, + "step": 16062 + }, + { + "epoch": 3.71, + "learning_rate": 1.9064401818041757e-05, + "loss": 0.0, + "step": 16063 + }, + { + "epoch": 3.71, + "learning_rate": 1.9062475926353904e-05, + "loss": 0.0, + "step": 16064 + }, + { + "epoch": 3.71, + "learning_rate": 1.906055003466605e-05, + "loss": 0.0, + "step": 16065 + }, + { + "epoch": 3.71, + "learning_rate": 1.90586241429782e-05, + "loss": 0.0001, + "step": 16066 + }, + { + "epoch": 3.71, + "learning_rate": 1.9056698251290347e-05, + "loss": 0.0, + "step": 16067 + }, + { + "epoch": 3.71, + "learning_rate": 1.9054772359602497e-05, + "loss": 0.0, + "step": 16068 + }, + { + "epoch": 3.71, + "learning_rate": 1.9052846467914645e-05, + "loss": 0.0, + "step": 16069 + }, + { + "epoch": 3.71, + "learning_rate": 1.9050920576226792e-05, + "loss": 0.0, + "step": 16070 + }, + { + "epoch": 3.71, + "learning_rate": 1.9048994684538943e-05, + "loss": 0.0, + "step": 16071 + }, + { + "epoch": 3.71, + "learning_rate": 1.904706879285109e-05, + "loss": 0.0, + "step": 16072 + }, + { + "epoch": 3.71, + "learning_rate": 1.904514290116324e-05, + "loss": 0.0, + "step": 16073 + }, + { + "epoch": 3.71, + "learning_rate": 1.904321700947539e-05, + "loss": 0.0, + "step": 16074 + }, + { + "epoch": 3.72, + "learning_rate": 1.9041291117787536e-05, + "loss": 0.0, + "step": 16075 + }, + { + "epoch": 3.72, + "learning_rate": 1.9039365226099684e-05, + "loss": 0.0, + "step": 16076 + }, + { + "epoch": 3.72, + "learning_rate": 1.903743933441183e-05, + "loss": 0.0, + "step": 16077 + }, + { + "epoch": 3.72, + "learning_rate": 1.9035513442723982e-05, + "loss": 0.0, + "step": 16078 + }, + { + "epoch": 3.72, + "learning_rate": 1.903358755103613e-05, + "loss": 0.0, + "step": 16079 + }, + { + "epoch": 3.72, + "learning_rate": 1.903166165934828e-05, + "loss": 0.0, + "step": 16080 + }, + { + "epoch": 3.72, + "learning_rate": 1.9029735767660428e-05, + "loss": 0.0, + "step": 16081 + }, + { + "epoch": 3.72, + "learning_rate": 1.9027809875972576e-05, + "loss": 0.0, + "step": 16082 + }, + { + "epoch": 3.72, + "learning_rate": 1.9025883984284726e-05, + "loss": 0.0, + "step": 16083 + }, + { + "epoch": 3.72, + "learning_rate": 1.902395809259687e-05, + "loss": 0.0, + "step": 16084 + }, + { + "epoch": 3.72, + "learning_rate": 1.902203220090902e-05, + "loss": 0.0, + "step": 16085 + }, + { + "epoch": 3.72, + "learning_rate": 1.902010630922117e-05, + "loss": 0.0001, + "step": 16086 + }, + { + "epoch": 3.72, + "learning_rate": 1.901818041753332e-05, + "loss": 0.0, + "step": 16087 + }, + { + "epoch": 3.72, + "learning_rate": 1.9016254525845467e-05, + "loss": 0.0, + "step": 16088 + }, + { + "epoch": 3.72, + "learning_rate": 1.9014328634157615e-05, + "loss": 0.0, + "step": 16089 + }, + { + "epoch": 3.72, + "learning_rate": 1.9012402742469766e-05, + "loss": 0.0, + "step": 16090 + }, + { + "epoch": 3.72, + "learning_rate": 1.9010476850781913e-05, + "loss": 0.0, + "step": 16091 + }, + { + "epoch": 3.72, + "learning_rate": 1.9008550959094064e-05, + "loss": 0.0, + "step": 16092 + }, + { + "epoch": 3.72, + "learning_rate": 1.900662506740621e-05, + "loss": 0.0, + "step": 16093 + }, + { + "epoch": 3.72, + "learning_rate": 1.900469917571836e-05, + "loss": 0.0, + "step": 16094 + }, + { + "epoch": 3.72, + "learning_rate": 1.9002773284030506e-05, + "loss": 0.0, + "step": 16095 + }, + { + "epoch": 3.72, + "learning_rate": 1.9000847392342654e-05, + "loss": 0.0, + "step": 16096 + }, + { + "epoch": 3.72, + "learning_rate": 1.8998921500654805e-05, + "loss": 0.0, + "step": 16097 + }, + { + "epoch": 3.72, + "learning_rate": 1.8996995608966952e-05, + "loss": 0.0, + "step": 16098 + }, + { + "epoch": 3.72, + "learning_rate": 1.8995069717279103e-05, + "loss": 0.0, + "step": 16099 + }, + { + "epoch": 3.72, + "learning_rate": 1.899314382559125e-05, + "loss": 0.0023, + "step": 16100 + }, + { + "epoch": 3.72, + "learning_rate": 1.8991217933903398e-05, + "loss": 0.0001, + "step": 16101 + }, + { + "epoch": 3.72, + "learning_rate": 1.898929204221555e-05, + "loss": 0.0, + "step": 16102 + }, + { + "epoch": 3.72, + "learning_rate": 1.8987366150527693e-05, + "loss": 0.0, + "step": 16103 + }, + { + "epoch": 3.72, + "learning_rate": 1.8985440258839844e-05, + "loss": 0.0, + "step": 16104 + }, + { + "epoch": 3.72, + "learning_rate": 1.898351436715199e-05, + "loss": 0.0008, + "step": 16105 + }, + { + "epoch": 3.72, + "learning_rate": 1.898158847546414e-05, + "loss": 0.0, + "step": 16106 + }, + { + "epoch": 3.72, + "learning_rate": 1.897966258377629e-05, + "loss": 0.0, + "step": 16107 + }, + { + "epoch": 3.72, + "learning_rate": 1.8977736692088437e-05, + "loss": 0.0, + "step": 16108 + }, + { + "epoch": 3.72, + "learning_rate": 1.8975810800400588e-05, + "loss": 0.0, + "step": 16109 + }, + { + "epoch": 3.72, + "learning_rate": 1.8973884908712735e-05, + "loss": 0.0, + "step": 16110 + }, + { + "epoch": 3.72, + "learning_rate": 1.8971959017024883e-05, + "loss": 0.0, + "step": 16111 + }, + { + "epoch": 3.72, + "learning_rate": 1.897003312533703e-05, + "loss": 0.0, + "step": 16112 + }, + { + "epoch": 3.72, + "learning_rate": 1.8968107233649178e-05, + "loss": 0.0, + "step": 16113 + }, + { + "epoch": 3.72, + "learning_rate": 1.896618134196133e-05, + "loss": 0.0, + "step": 16114 + }, + { + "epoch": 3.72, + "learning_rate": 1.8964255450273476e-05, + "loss": 0.0, + "step": 16115 + }, + { + "epoch": 3.72, + "learning_rate": 1.8962329558585627e-05, + "loss": 0.0, + "step": 16116 + }, + { + "epoch": 3.72, + "learning_rate": 1.8960403666897775e-05, + "loss": 0.0, + "step": 16117 + }, + { + "epoch": 3.72, + "learning_rate": 1.8958477775209922e-05, + "loss": 0.0, + "step": 16118 + }, + { + "epoch": 3.73, + "learning_rate": 1.8956551883522073e-05, + "loss": 0.0, + "step": 16119 + }, + { + "epoch": 3.73, + "learning_rate": 1.895462599183422e-05, + "loss": 0.0, + "step": 16120 + }, + { + "epoch": 3.73, + "learning_rate": 1.895270010014637e-05, + "loss": 0.0, + "step": 16121 + }, + { + "epoch": 3.73, + "learning_rate": 1.8950774208458515e-05, + "loss": 0.0, + "step": 16122 + }, + { + "epoch": 3.73, + "learning_rate": 1.8948848316770666e-05, + "loss": 0.0, + "step": 16123 + }, + { + "epoch": 3.73, + "learning_rate": 1.8946922425082814e-05, + "loss": 0.0, + "step": 16124 + }, + { + "epoch": 3.73, + "learning_rate": 1.894499653339496e-05, + "loss": 0.0001, + "step": 16125 + }, + { + "epoch": 3.73, + "learning_rate": 1.8943070641707112e-05, + "loss": 0.0, + "step": 16126 + }, + { + "epoch": 3.73, + "learning_rate": 1.894114475001926e-05, + "loss": 0.0, + "step": 16127 + }, + { + "epoch": 3.73, + "learning_rate": 1.893921885833141e-05, + "loss": 0.0, + "step": 16128 + }, + { + "epoch": 3.73, + "learning_rate": 1.8937292966643558e-05, + "loss": 0.0, + "step": 16129 + }, + { + "epoch": 3.73, + "learning_rate": 1.8935367074955705e-05, + "loss": 0.0, + "step": 16130 + }, + { + "epoch": 3.73, + "learning_rate": 1.8933441183267853e-05, + "loss": 0.0, + "step": 16131 + }, + { + "epoch": 3.73, + "learning_rate": 1.893151529158e-05, + "loss": 0.0, + "step": 16132 + }, + { + "epoch": 3.73, + "learning_rate": 1.892958939989215e-05, + "loss": 0.0, + "step": 16133 + }, + { + "epoch": 3.73, + "learning_rate": 1.89276635082043e-05, + "loss": 0.0, + "step": 16134 + }, + { + "epoch": 3.73, + "learning_rate": 1.892573761651645e-05, + "loss": 0.002, + "step": 16135 + }, + { + "epoch": 3.73, + "learning_rate": 1.8923811724828597e-05, + "loss": 0.0, + "step": 16136 + }, + { + "epoch": 3.73, + "learning_rate": 1.8921885833140744e-05, + "loss": 0.0, + "step": 16137 + }, + { + "epoch": 3.73, + "learning_rate": 1.8919959941452895e-05, + "loss": 0.0, + "step": 16138 + }, + { + "epoch": 3.73, + "learning_rate": 1.8918034049765043e-05, + "loss": 0.0, + "step": 16139 + }, + { + "epoch": 3.73, + "learning_rate": 1.891610815807719e-05, + "loss": 0.0, + "step": 16140 + }, + { + "epoch": 3.73, + "learning_rate": 1.8914182266389338e-05, + "loss": 0.0, + "step": 16141 + }, + { + "epoch": 3.73, + "learning_rate": 1.8912256374701485e-05, + "loss": 0.0, + "step": 16142 + }, + { + "epoch": 3.73, + "learning_rate": 1.8910330483013636e-05, + "loss": 0.0001, + "step": 16143 + }, + { + "epoch": 3.73, + "learning_rate": 1.8908404591325784e-05, + "loss": 0.0, + "step": 16144 + }, + { + "epoch": 3.73, + "learning_rate": 1.8906478699637934e-05, + "loss": 0.0, + "step": 16145 + }, + { + "epoch": 3.73, + "learning_rate": 1.8904552807950082e-05, + "loss": 0.0, + "step": 16146 + }, + { + "epoch": 3.73, + "learning_rate": 1.890262691626223e-05, + "loss": 0.0, + "step": 16147 + }, + { + "epoch": 3.73, + "learning_rate": 1.890070102457438e-05, + "loss": 0.0, + "step": 16148 + }, + { + "epoch": 3.73, + "learning_rate": 1.8898775132886528e-05, + "loss": 0.0, + "step": 16149 + }, + { + "epoch": 3.73, + "learning_rate": 1.8896849241198675e-05, + "loss": 0.0, + "step": 16150 + }, + { + "epoch": 3.73, + "learning_rate": 1.8894923349510823e-05, + "loss": 0.0, + "step": 16151 + }, + { + "epoch": 3.73, + "learning_rate": 1.8892997457822974e-05, + "loss": 0.0, + "step": 16152 + }, + { + "epoch": 3.73, + "learning_rate": 1.889107156613512e-05, + "loss": 0.0, + "step": 16153 + }, + { + "epoch": 3.73, + "learning_rate": 1.888914567444727e-05, + "loss": 0.0, + "step": 16154 + }, + { + "epoch": 3.73, + "learning_rate": 1.888721978275942e-05, + "loss": 0.0, + "step": 16155 + }, + { + "epoch": 3.73, + "learning_rate": 1.8885293891071567e-05, + "loss": 0.0, + "step": 16156 + }, + { + "epoch": 3.73, + "learning_rate": 1.8883367999383718e-05, + "loss": 0.0, + "step": 16157 + }, + { + "epoch": 3.73, + "learning_rate": 1.8881442107695865e-05, + "loss": 0.0, + "step": 16158 + }, + { + "epoch": 3.73, + "learning_rate": 1.8879516216008013e-05, + "loss": 0.0, + "step": 16159 + }, + { + "epoch": 3.73, + "learning_rate": 1.887759032432016e-05, + "loss": 0.0, + "step": 16160 + }, + { + "epoch": 3.73, + "learning_rate": 1.8875664432632308e-05, + "loss": 0.0, + "step": 16161 + }, + { + "epoch": 3.74, + "learning_rate": 1.887373854094446e-05, + "loss": 0.0001, + "step": 16162 + }, + { + "epoch": 3.74, + "learning_rate": 1.8871812649256606e-05, + "loss": 0.0001, + "step": 16163 + }, + { + "epoch": 3.74, + "learning_rate": 1.8869886757568757e-05, + "loss": 0.0, + "step": 16164 + }, + { + "epoch": 3.74, + "learning_rate": 1.8867960865880904e-05, + "loss": 0.0, + "step": 16165 + }, + { + "epoch": 3.74, + "learning_rate": 1.8866034974193052e-05, + "loss": 0.0, + "step": 16166 + }, + { + "epoch": 3.74, + "learning_rate": 1.8864109082505203e-05, + "loss": 0.0, + "step": 16167 + }, + { + "epoch": 3.74, + "learning_rate": 1.8862183190817347e-05, + "loss": 0.0, + "step": 16168 + }, + { + "epoch": 3.74, + "learning_rate": 1.8860257299129498e-05, + "loss": 0.0, + "step": 16169 + }, + { + "epoch": 3.74, + "learning_rate": 1.8858331407441645e-05, + "loss": 0.0, + "step": 16170 + }, + { + "epoch": 3.74, + "learning_rate": 1.8856405515753796e-05, + "loss": 0.0, + "step": 16171 + }, + { + "epoch": 3.74, + "learning_rate": 1.8854479624065943e-05, + "loss": 0.0, + "step": 16172 + }, + { + "epoch": 3.74, + "learning_rate": 1.885255373237809e-05, + "loss": 0.0, + "step": 16173 + }, + { + "epoch": 3.74, + "learning_rate": 1.8850627840690242e-05, + "loss": 0.0, + "step": 16174 + }, + { + "epoch": 3.74, + "learning_rate": 1.884870194900239e-05, + "loss": 0.0, + "step": 16175 + }, + { + "epoch": 3.74, + "learning_rate": 1.884677605731454e-05, + "loss": 0.0, + "step": 16176 + }, + { + "epoch": 3.74, + "learning_rate": 1.8844850165626688e-05, + "loss": 0.0, + "step": 16177 + }, + { + "epoch": 3.74, + "learning_rate": 1.8842924273938832e-05, + "loss": 0.0, + "step": 16178 + }, + { + "epoch": 3.74, + "learning_rate": 1.8840998382250983e-05, + "loss": 0.0, + "step": 16179 + }, + { + "epoch": 3.74, + "learning_rate": 1.883907249056313e-05, + "loss": 0.0001, + "step": 16180 + }, + { + "epoch": 3.74, + "learning_rate": 1.883714659887528e-05, + "loss": 0.0, + "step": 16181 + }, + { + "epoch": 3.74, + "learning_rate": 1.883522070718743e-05, + "loss": 0.0, + "step": 16182 + }, + { + "epoch": 3.74, + "learning_rate": 1.8833294815499576e-05, + "loss": 0.0, + "step": 16183 + }, + { + "epoch": 3.74, + "learning_rate": 1.8831368923811727e-05, + "loss": 0.0, + "step": 16184 + }, + { + "epoch": 3.74, + "learning_rate": 1.8829443032123874e-05, + "loss": 0.0, + "step": 16185 + }, + { + "epoch": 3.74, + "learning_rate": 1.8827517140436025e-05, + "loss": 0.0, + "step": 16186 + }, + { + "epoch": 3.74, + "learning_rate": 1.882559124874817e-05, + "loss": 0.0002, + "step": 16187 + }, + { + "epoch": 3.74, + "learning_rate": 1.882366535706032e-05, + "loss": 0.0, + "step": 16188 + }, + { + "epoch": 3.74, + "learning_rate": 1.8821739465372468e-05, + "loss": 0.0, + "step": 16189 + }, + { + "epoch": 3.74, + "learning_rate": 1.8819813573684615e-05, + "loss": 0.0, + "step": 16190 + }, + { + "epoch": 3.74, + "learning_rate": 1.8817887681996766e-05, + "loss": 0.0, + "step": 16191 + }, + { + "epoch": 3.74, + "learning_rate": 1.8815961790308913e-05, + "loss": 0.0, + "step": 16192 + }, + { + "epoch": 3.74, + "learning_rate": 1.8814035898621064e-05, + "loss": 0.0, + "step": 16193 + }, + { + "epoch": 3.74, + "learning_rate": 1.8812110006933212e-05, + "loss": 0.0, + "step": 16194 + }, + { + "epoch": 3.74, + "learning_rate": 1.881018411524536e-05, + "loss": 0.0, + "step": 16195 + }, + { + "epoch": 3.74, + "learning_rate": 1.8808258223557507e-05, + "loss": 0.0, + "step": 16196 + }, + { + "epoch": 3.74, + "learning_rate": 1.8806332331869654e-05, + "loss": 0.0, + "step": 16197 + }, + { + "epoch": 3.74, + "learning_rate": 1.8804406440181805e-05, + "loss": 0.0, + "step": 16198 + }, + { + "epoch": 3.74, + "learning_rate": 1.8802480548493952e-05, + "loss": 0.0, + "step": 16199 + }, + { + "epoch": 3.74, + "learning_rate": 1.8800554656806103e-05, + "loss": 0.0, + "step": 16200 + }, + { + "epoch": 3.74, + "learning_rate": 1.879862876511825e-05, + "loss": 0.0, + "step": 16201 + }, + { + "epoch": 3.74, + "learning_rate": 1.8796702873430398e-05, + "loss": 0.0, + "step": 16202 + }, + { + "epoch": 3.74, + "learning_rate": 1.879477698174255e-05, + "loss": 0.0, + "step": 16203 + }, + { + "epoch": 3.74, + "learning_rate": 1.8792851090054697e-05, + "loss": 0.0, + "step": 16204 + }, + { + "epoch": 3.75, + "learning_rate": 1.8790925198366848e-05, + "loss": 0.0, + "step": 16205 + }, + { + "epoch": 3.75, + "learning_rate": 1.878899930667899e-05, + "loss": 0.0, + "step": 16206 + }, + { + "epoch": 3.75, + "learning_rate": 1.8787073414991142e-05, + "loss": 0.0, + "step": 16207 + }, + { + "epoch": 3.75, + "learning_rate": 1.878514752330329e-05, + "loss": 0.0, + "step": 16208 + }, + { + "epoch": 3.75, + "learning_rate": 1.8783221631615437e-05, + "loss": 0.0, + "step": 16209 + }, + { + "epoch": 3.75, + "learning_rate": 1.8781295739927588e-05, + "loss": 0.0, + "step": 16210 + }, + { + "epoch": 3.75, + "learning_rate": 1.8779369848239736e-05, + "loss": 0.0, + "step": 16211 + }, + { + "epoch": 3.75, + "learning_rate": 1.8777443956551887e-05, + "loss": 0.0, + "step": 16212 + }, + { + "epoch": 3.75, + "learning_rate": 1.8775518064864034e-05, + "loss": 0.0, + "step": 16213 + }, + { + "epoch": 3.75, + "learning_rate": 1.877359217317618e-05, + "loss": 0.0, + "step": 16214 + }, + { + "epoch": 3.75, + "learning_rate": 1.877166628148833e-05, + "loss": 0.0, + "step": 16215 + }, + { + "epoch": 3.75, + "learning_rate": 1.8769740389800477e-05, + "loss": 0.0, + "step": 16216 + }, + { + "epoch": 3.75, + "learning_rate": 1.8767814498112627e-05, + "loss": 0.0, + "step": 16217 + }, + { + "epoch": 3.75, + "learning_rate": 1.8765888606424775e-05, + "loss": 0.0, + "step": 16218 + }, + { + "epoch": 3.75, + "learning_rate": 1.8763962714736922e-05, + "loss": 0.0, + "step": 16219 + }, + { + "epoch": 3.75, + "learning_rate": 1.8762036823049073e-05, + "loss": 0.0, + "step": 16220 + }, + { + "epoch": 3.75, + "learning_rate": 1.876011093136122e-05, + "loss": 0.0, + "step": 16221 + }, + { + "epoch": 3.75, + "learning_rate": 1.875818503967337e-05, + "loss": 0.0, + "step": 16222 + }, + { + "epoch": 3.75, + "learning_rate": 1.875625914798552e-05, + "loss": 0.0, + "step": 16223 + }, + { + "epoch": 3.75, + "learning_rate": 1.8754333256297667e-05, + "loss": 0.0, + "step": 16224 + }, + { + "epoch": 3.75, + "learning_rate": 1.8752407364609814e-05, + "loss": 0.0, + "step": 16225 + }, + { + "epoch": 3.75, + "learning_rate": 1.875048147292196e-05, + "loss": 0.0, + "step": 16226 + }, + { + "epoch": 3.75, + "learning_rate": 1.8748555581234112e-05, + "loss": 0.0001, + "step": 16227 + }, + { + "epoch": 3.75, + "learning_rate": 1.874662968954626e-05, + "loss": 0.0, + "step": 16228 + }, + { + "epoch": 3.75, + "learning_rate": 1.874470379785841e-05, + "loss": 0.0, + "step": 16229 + }, + { + "epoch": 3.75, + "learning_rate": 1.8742777906170558e-05, + "loss": 0.0, + "step": 16230 + }, + { + "epoch": 3.75, + "learning_rate": 1.8740852014482706e-05, + "loss": 0.0, + "step": 16231 + }, + { + "epoch": 3.75, + "learning_rate": 1.8738926122794857e-05, + "loss": 0.0, + "step": 16232 + }, + { + "epoch": 3.75, + "learning_rate": 1.8737000231107004e-05, + "loss": 0.0, + "step": 16233 + }, + { + "epoch": 3.75, + "learning_rate": 1.873507433941915e-05, + "loss": 0.0, + "step": 16234 + }, + { + "epoch": 3.75, + "learning_rate": 1.87331484477313e-05, + "loss": 0.0, + "step": 16235 + }, + { + "epoch": 3.75, + "learning_rate": 1.873122255604345e-05, + "loss": 0.0001, + "step": 16236 + }, + { + "epoch": 3.75, + "learning_rate": 1.8729296664355597e-05, + "loss": 0.0, + "step": 16237 + }, + { + "epoch": 3.75, + "learning_rate": 1.8727370772667745e-05, + "loss": 0.0, + "step": 16238 + }, + { + "epoch": 3.75, + "learning_rate": 1.8725444880979896e-05, + "loss": 0.0, + "step": 16239 + }, + { + "epoch": 3.75, + "learning_rate": 1.8723518989292043e-05, + "loss": 0.0, + "step": 16240 + }, + { + "epoch": 3.75, + "learning_rate": 1.8721593097604194e-05, + "loss": 0.0, + "step": 16241 + }, + { + "epoch": 3.75, + "learning_rate": 1.871966720591634e-05, + "loss": 0.0, + "step": 16242 + }, + { + "epoch": 3.75, + "learning_rate": 1.871774131422849e-05, + "loss": 0.0, + "step": 16243 + }, + { + "epoch": 3.75, + "learning_rate": 1.8715815422540636e-05, + "loss": 0.0, + "step": 16244 + }, + { + "epoch": 3.75, + "learning_rate": 1.8713889530852784e-05, + "loss": 0.0, + "step": 16245 + }, + { + "epoch": 3.75, + "learning_rate": 1.8711963639164935e-05, + "loss": 0.0, + "step": 16246 + }, + { + "epoch": 3.75, + "learning_rate": 1.8710037747477082e-05, + "loss": 0.0001, + "step": 16247 + }, + { + "epoch": 3.76, + "learning_rate": 1.8708111855789233e-05, + "loss": 0.0, + "step": 16248 + }, + { + "epoch": 3.76, + "learning_rate": 1.870618596410138e-05, + "loss": 0.0, + "step": 16249 + }, + { + "epoch": 3.76, + "learning_rate": 1.8704260072413528e-05, + "loss": 0.0, + "step": 16250 + }, + { + "epoch": 3.76, + "learning_rate": 1.870233418072568e-05, + "loss": 0.0, + "step": 16251 + }, + { + "epoch": 3.76, + "learning_rate": 1.8700408289037823e-05, + "loss": 0.0, + "step": 16252 + }, + { + "epoch": 3.76, + "learning_rate": 1.8698482397349974e-05, + "loss": 0.0, + "step": 16253 + }, + { + "epoch": 3.76, + "learning_rate": 1.869655650566212e-05, + "loss": 0.0, + "step": 16254 + }, + { + "epoch": 3.76, + "learning_rate": 1.869463061397427e-05, + "loss": 0.0, + "step": 16255 + }, + { + "epoch": 3.76, + "learning_rate": 1.869270472228642e-05, + "loss": 0.0, + "step": 16256 + }, + { + "epoch": 3.76, + "learning_rate": 1.8690778830598567e-05, + "loss": 0.0, + "step": 16257 + }, + { + "epoch": 3.76, + "learning_rate": 1.8688852938910718e-05, + "loss": 0.0, + "step": 16258 + }, + { + "epoch": 3.76, + "learning_rate": 1.8686927047222866e-05, + "loss": 0.0, + "step": 16259 + }, + { + "epoch": 3.76, + "learning_rate": 1.8685001155535013e-05, + "loss": 0.0, + "step": 16260 + }, + { + "epoch": 3.76, + "learning_rate": 1.8683075263847164e-05, + "loss": 0.0, + "step": 16261 + }, + { + "epoch": 3.76, + "learning_rate": 1.8681149372159308e-05, + "loss": 0.0, + "step": 16262 + }, + { + "epoch": 3.76, + "learning_rate": 1.867922348047146e-05, + "loss": 0.0, + "step": 16263 + }, + { + "epoch": 3.76, + "learning_rate": 1.8677297588783606e-05, + "loss": 0.0, + "step": 16264 + }, + { + "epoch": 3.76, + "learning_rate": 1.8675371697095757e-05, + "loss": 0.0, + "step": 16265 + }, + { + "epoch": 3.76, + "learning_rate": 1.8673445805407905e-05, + "loss": 0.0001, + "step": 16266 + }, + { + "epoch": 3.76, + "learning_rate": 1.8671519913720052e-05, + "loss": 0.0, + "step": 16267 + }, + { + "epoch": 3.76, + "learning_rate": 1.8669594022032203e-05, + "loss": 0.0, + "step": 16268 + }, + { + "epoch": 3.76, + "learning_rate": 1.866766813034435e-05, + "loss": 0.0001, + "step": 16269 + }, + { + "epoch": 3.76, + "learning_rate": 1.86657422386565e-05, + "loss": 0.0, + "step": 16270 + }, + { + "epoch": 3.76, + "learning_rate": 1.8663816346968645e-05, + "loss": 0.0, + "step": 16271 + }, + { + "epoch": 3.76, + "learning_rate": 1.8661890455280796e-05, + "loss": 0.0, + "step": 16272 + }, + { + "epoch": 3.76, + "learning_rate": 1.8659964563592944e-05, + "loss": 0.0, + "step": 16273 + }, + { + "epoch": 3.76, + "learning_rate": 1.865803867190509e-05, + "loss": 0.0, + "step": 16274 + }, + { + "epoch": 3.76, + "learning_rate": 1.8656112780217242e-05, + "loss": 0.0, + "step": 16275 + }, + { + "epoch": 3.76, + "learning_rate": 1.865418688852939e-05, + "loss": 0.0, + "step": 16276 + }, + { + "epoch": 3.76, + "learning_rate": 1.865226099684154e-05, + "loss": 0.0001, + "step": 16277 + }, + { + "epoch": 3.76, + "learning_rate": 1.8650335105153688e-05, + "loss": 0.0, + "step": 16278 + }, + { + "epoch": 3.76, + "learning_rate": 1.8648409213465835e-05, + "loss": 0.0, + "step": 16279 + }, + { + "epoch": 3.76, + "learning_rate": 1.8646483321777983e-05, + "loss": 0.0, + "step": 16280 + }, + { + "epoch": 3.76, + "learning_rate": 1.864455743009013e-05, + "loss": 0.0, + "step": 16281 + }, + { + "epoch": 3.76, + "learning_rate": 1.864263153840228e-05, + "loss": 0.0, + "step": 16282 + }, + { + "epoch": 3.76, + "learning_rate": 1.864070564671443e-05, + "loss": 0.0, + "step": 16283 + }, + { + "epoch": 3.76, + "learning_rate": 1.863877975502658e-05, + "loss": 0.0, + "step": 16284 + }, + { + "epoch": 3.76, + "learning_rate": 1.8636853863338727e-05, + "loss": 0.0, + "step": 16285 + }, + { + "epoch": 3.76, + "learning_rate": 1.8634927971650875e-05, + "loss": 0.0, + "step": 16286 + }, + { + "epoch": 3.76, + "learning_rate": 1.8633002079963025e-05, + "loss": 0.0, + "step": 16287 + }, + { + "epoch": 3.76, + "learning_rate": 1.8631076188275173e-05, + "loss": 0.0001, + "step": 16288 + }, + { + "epoch": 3.76, + "learning_rate": 1.8629150296587324e-05, + "loss": 0.0, + "step": 16289 + }, + { + "epoch": 3.76, + "learning_rate": 1.8627224404899468e-05, + "loss": 0.0, + "step": 16290 + }, + { + "epoch": 3.76, + "learning_rate": 1.862529851321162e-05, + "loss": 0.0, + "step": 16291 + }, + { + "epoch": 3.77, + "learning_rate": 1.8623372621523766e-05, + "loss": 0.0, + "step": 16292 + }, + { + "epoch": 3.77, + "learning_rate": 1.8621446729835914e-05, + "loss": 0.0, + "step": 16293 + }, + { + "epoch": 3.77, + "learning_rate": 1.8619520838148065e-05, + "loss": 0.0, + "step": 16294 + }, + { + "epoch": 3.77, + "learning_rate": 1.8617594946460212e-05, + "loss": 0.0, + "step": 16295 + }, + { + "epoch": 3.77, + "learning_rate": 1.861566905477236e-05, + "loss": 0.0, + "step": 16296 + }, + { + "epoch": 3.77, + "learning_rate": 1.861374316308451e-05, + "loss": 0.0, + "step": 16297 + }, + { + "epoch": 3.77, + "learning_rate": 1.8611817271396658e-05, + "loss": 0.0, + "step": 16298 + }, + { + "epoch": 3.77, + "learning_rate": 1.8609891379708805e-05, + "loss": 0.0, + "step": 16299 + }, + { + "epoch": 3.77, + "learning_rate": 1.8607965488020953e-05, + "loss": 0.0, + "step": 16300 + }, + { + "epoch": 3.77, + "learning_rate": 1.8606039596333104e-05, + "loss": 0.0001, + "step": 16301 + }, + { + "epoch": 3.77, + "learning_rate": 1.860411370464525e-05, + "loss": 0.0, + "step": 16302 + }, + { + "epoch": 3.77, + "learning_rate": 1.86021878129574e-05, + "loss": 0.0, + "step": 16303 + }, + { + "epoch": 3.77, + "learning_rate": 1.860026192126955e-05, + "loss": 0.0, + "step": 16304 + }, + { + "epoch": 3.77, + "learning_rate": 1.8598336029581697e-05, + "loss": 0.0, + "step": 16305 + }, + { + "epoch": 3.77, + "learning_rate": 1.8596410137893848e-05, + "loss": 0.0, + "step": 16306 + }, + { + "epoch": 3.77, + "learning_rate": 1.8594484246205995e-05, + "loss": 0.0005, + "step": 16307 + }, + { + "epoch": 3.77, + "learning_rate": 1.8592558354518143e-05, + "loss": 0.0, + "step": 16308 + }, + { + "epoch": 3.77, + "learning_rate": 1.859063246283029e-05, + "loss": 0.0, + "step": 16309 + }, + { + "epoch": 3.77, + "learning_rate": 1.8588706571142438e-05, + "loss": 0.0, + "step": 16310 + }, + { + "epoch": 3.77, + "learning_rate": 1.858678067945459e-05, + "loss": 0.0, + "step": 16311 + }, + { + "epoch": 3.77, + "learning_rate": 1.8584854787766736e-05, + "loss": 0.0, + "step": 16312 + }, + { + "epoch": 3.77, + "learning_rate": 1.8582928896078887e-05, + "loss": 0.0, + "step": 16313 + }, + { + "epoch": 3.77, + "learning_rate": 1.8581003004391034e-05, + "loss": 0.0, + "step": 16314 + }, + { + "epoch": 3.77, + "learning_rate": 1.8579077112703182e-05, + "loss": 0.0, + "step": 16315 + }, + { + "epoch": 3.77, + "learning_rate": 1.8577151221015333e-05, + "loss": 0.0, + "step": 16316 + }, + { + "epoch": 3.77, + "learning_rate": 1.857522532932748e-05, + "loss": 0.0, + "step": 16317 + }, + { + "epoch": 3.77, + "learning_rate": 1.8573299437639628e-05, + "loss": 0.0, + "step": 16318 + }, + { + "epoch": 3.77, + "learning_rate": 1.8571373545951775e-05, + "loss": 0.0, + "step": 16319 + }, + { + "epoch": 3.77, + "learning_rate": 1.8569447654263926e-05, + "loss": 0.0, + "step": 16320 + }, + { + "epoch": 3.77, + "learning_rate": 1.8567521762576074e-05, + "loss": 0.0, + "step": 16321 + }, + { + "epoch": 3.77, + "learning_rate": 1.856559587088822e-05, + "loss": 0.0, + "step": 16322 + }, + { + "epoch": 3.77, + "learning_rate": 1.8563669979200372e-05, + "loss": 0.0, + "step": 16323 + }, + { + "epoch": 3.77, + "learning_rate": 1.856174408751252e-05, + "loss": 0.0, + "step": 16324 + }, + { + "epoch": 3.77, + "learning_rate": 1.855981819582467e-05, + "loss": 0.0, + "step": 16325 + }, + { + "epoch": 3.77, + "learning_rate": 1.8557892304136818e-05, + "loss": 0.0, + "step": 16326 + }, + { + "epoch": 3.77, + "learning_rate": 1.8555966412448965e-05, + "loss": 0.0, + "step": 16327 + }, + { + "epoch": 3.77, + "learning_rate": 1.8554040520761113e-05, + "loss": 0.0, + "step": 16328 + }, + { + "epoch": 3.77, + "learning_rate": 1.855211462907326e-05, + "loss": 0.0, + "step": 16329 + }, + { + "epoch": 3.77, + "learning_rate": 1.855018873738541e-05, + "loss": 0.0, + "step": 16330 + }, + { + "epoch": 3.77, + "learning_rate": 1.854826284569756e-05, + "loss": 0.0, + "step": 16331 + }, + { + "epoch": 3.77, + "learning_rate": 1.8546336954009706e-05, + "loss": 0.0, + "step": 16332 + }, + { + "epoch": 3.77, + "learning_rate": 1.8544411062321857e-05, + "loss": 0.001, + "step": 16333 + }, + { + "epoch": 3.77, + "learning_rate": 1.8542485170634004e-05, + "loss": 0.0, + "step": 16334 + }, + { + "epoch": 3.78, + "learning_rate": 1.8540559278946155e-05, + "loss": 0.0, + "step": 16335 + }, + { + "epoch": 3.78, + "learning_rate": 1.85386333872583e-05, + "loss": 0.0, + "step": 16336 + }, + { + "epoch": 3.78, + "learning_rate": 1.853670749557045e-05, + "loss": 0.0, + "step": 16337 + }, + { + "epoch": 3.78, + "learning_rate": 1.8534781603882598e-05, + "loss": 0.0, + "step": 16338 + }, + { + "epoch": 3.78, + "learning_rate": 1.8532855712194745e-05, + "loss": 0.0001, + "step": 16339 + }, + { + "epoch": 3.78, + "learning_rate": 1.8530929820506896e-05, + "loss": 0.0, + "step": 16340 + }, + { + "epoch": 3.78, + "learning_rate": 1.8529003928819043e-05, + "loss": 0.0001, + "step": 16341 + }, + { + "epoch": 3.78, + "learning_rate": 1.8527078037131194e-05, + "loss": 0.0, + "step": 16342 + }, + { + "epoch": 3.78, + "learning_rate": 1.8525152145443342e-05, + "loss": 0.0, + "step": 16343 + }, + { + "epoch": 3.78, + "learning_rate": 1.852322625375549e-05, + "loss": 0.0, + "step": 16344 + }, + { + "epoch": 3.78, + "learning_rate": 1.852130036206764e-05, + "loss": 0.0, + "step": 16345 + }, + { + "epoch": 3.78, + "learning_rate": 1.8519374470379784e-05, + "loss": 0.0, + "step": 16346 + }, + { + "epoch": 3.78, + "learning_rate": 1.8517448578691935e-05, + "loss": 0.0, + "step": 16347 + }, + { + "epoch": 3.78, + "learning_rate": 1.8515522687004083e-05, + "loss": 0.0, + "step": 16348 + }, + { + "epoch": 3.78, + "learning_rate": 1.8513596795316233e-05, + "loss": 0.0, + "step": 16349 + }, + { + "epoch": 3.78, + "learning_rate": 1.851167090362838e-05, + "loss": 0.0, + "step": 16350 + }, + { + "epoch": 3.78, + "learning_rate": 1.850974501194053e-05, + "loss": 0.0, + "step": 16351 + }, + { + "epoch": 3.78, + "learning_rate": 1.850781912025268e-05, + "loss": 0.0, + "step": 16352 + }, + { + "epoch": 3.78, + "learning_rate": 1.8505893228564827e-05, + "loss": 0.0, + "step": 16353 + }, + { + "epoch": 3.78, + "learning_rate": 1.8503967336876978e-05, + "loss": 0.0, + "step": 16354 + }, + { + "epoch": 3.78, + "learning_rate": 1.850204144518912e-05, + "loss": 0.0, + "step": 16355 + }, + { + "epoch": 3.78, + "learning_rate": 1.8500115553501273e-05, + "loss": 0.0, + "step": 16356 + }, + { + "epoch": 3.78, + "learning_rate": 1.849818966181342e-05, + "loss": 0.0, + "step": 16357 + }, + { + "epoch": 3.78, + "learning_rate": 1.8496263770125567e-05, + "loss": 0.0, + "step": 16358 + }, + { + "epoch": 3.78, + "learning_rate": 1.849433787843772e-05, + "loss": 0.0, + "step": 16359 + }, + { + "epoch": 3.78, + "learning_rate": 1.8492411986749866e-05, + "loss": 0.0, + "step": 16360 + }, + { + "epoch": 3.78, + "learning_rate": 1.8490486095062017e-05, + "loss": 0.0, + "step": 16361 + }, + { + "epoch": 3.78, + "learning_rate": 1.8488560203374164e-05, + "loss": 0.0, + "step": 16362 + }, + { + "epoch": 3.78, + "learning_rate": 1.848663431168631e-05, + "loss": 0.0, + "step": 16363 + }, + { + "epoch": 3.78, + "learning_rate": 1.848470841999846e-05, + "loss": 0.0, + "step": 16364 + }, + { + "epoch": 3.78, + "learning_rate": 1.8482782528310607e-05, + "loss": 0.0, + "step": 16365 + }, + { + "epoch": 3.78, + "learning_rate": 1.8480856636622757e-05, + "loss": 0.0001, + "step": 16366 + }, + { + "epoch": 3.78, + "learning_rate": 1.8478930744934905e-05, + "loss": 0.0, + "step": 16367 + }, + { + "epoch": 3.78, + "learning_rate": 1.8477004853247052e-05, + "loss": 0.0, + "step": 16368 + }, + { + "epoch": 3.78, + "learning_rate": 1.8475078961559203e-05, + "loss": 0.0, + "step": 16369 + }, + { + "epoch": 3.78, + "learning_rate": 1.847315306987135e-05, + "loss": 0.0, + "step": 16370 + }, + { + "epoch": 3.78, + "learning_rate": 1.84712271781835e-05, + "loss": 0.0, + "step": 16371 + }, + { + "epoch": 3.78, + "learning_rate": 1.846930128649565e-05, + "loss": 0.0, + "step": 16372 + }, + { + "epoch": 3.78, + "learning_rate": 1.8467375394807797e-05, + "loss": 0.0, + "step": 16373 + }, + { + "epoch": 3.78, + "learning_rate": 1.8465449503119944e-05, + "loss": 0.0, + "step": 16374 + }, + { + "epoch": 3.78, + "learning_rate": 1.846352361143209e-05, + "loss": 0.0, + "step": 16375 + }, + { + "epoch": 3.78, + "learning_rate": 1.8461597719744242e-05, + "loss": 0.0, + "step": 16376 + }, + { + "epoch": 3.78, + "learning_rate": 1.845967182805639e-05, + "loss": 0.0, + "step": 16377 + }, + { + "epoch": 3.79, + "learning_rate": 1.845774593636854e-05, + "loss": 0.0, + "step": 16378 + }, + { + "epoch": 3.79, + "learning_rate": 1.8455820044680688e-05, + "loss": 0.0, + "step": 16379 + }, + { + "epoch": 3.79, + "learning_rate": 1.8453894152992836e-05, + "loss": 0.0004, + "step": 16380 + }, + { + "epoch": 3.79, + "learning_rate": 1.8451968261304987e-05, + "loss": 0.0, + "step": 16381 + }, + { + "epoch": 3.79, + "learning_rate": 1.8450042369617134e-05, + "loss": 0.0, + "step": 16382 + }, + { + "epoch": 3.79, + "learning_rate": 1.844811647792928e-05, + "loss": 0.0, + "step": 16383 + }, + { + "epoch": 3.79, + "learning_rate": 1.844619058624143e-05, + "loss": 0.0, + "step": 16384 + }, + { + "epoch": 3.79, + "learning_rate": 1.844426469455358e-05, + "loss": 0.0004, + "step": 16385 + }, + { + "epoch": 3.79, + "learning_rate": 1.8442338802865727e-05, + "loss": 0.0, + "step": 16386 + }, + { + "epoch": 3.79, + "learning_rate": 1.8440412911177875e-05, + "loss": 0.0, + "step": 16387 + }, + { + "epoch": 3.79, + "learning_rate": 1.8438487019490026e-05, + "loss": 0.0, + "step": 16388 + }, + { + "epoch": 3.79, + "learning_rate": 1.8436561127802173e-05, + "loss": 0.0, + "step": 16389 + }, + { + "epoch": 3.79, + "learning_rate": 1.8434635236114324e-05, + "loss": 0.0, + "step": 16390 + }, + { + "epoch": 3.79, + "learning_rate": 1.843270934442647e-05, + "loss": 0.0, + "step": 16391 + }, + { + "epoch": 3.79, + "learning_rate": 1.843078345273862e-05, + "loss": 0.0, + "step": 16392 + }, + { + "epoch": 3.79, + "learning_rate": 1.8428857561050766e-05, + "loss": 0.0001, + "step": 16393 + }, + { + "epoch": 3.79, + "learning_rate": 1.8426931669362914e-05, + "loss": 0.0004, + "step": 16394 + }, + { + "epoch": 3.79, + "learning_rate": 1.8425005777675065e-05, + "loss": 0.0, + "step": 16395 + }, + { + "epoch": 3.79, + "learning_rate": 1.8423079885987212e-05, + "loss": 0.0, + "step": 16396 + }, + { + "epoch": 3.79, + "learning_rate": 1.8421153994299363e-05, + "loss": 0.0, + "step": 16397 + }, + { + "epoch": 3.79, + "learning_rate": 1.841922810261151e-05, + "loss": 0.0, + "step": 16398 + }, + { + "epoch": 3.79, + "learning_rate": 1.8417302210923658e-05, + "loss": 0.0, + "step": 16399 + }, + { + "epoch": 3.79, + "learning_rate": 1.841537631923581e-05, + "loss": 0.0, + "step": 16400 + }, + { + "epoch": 3.79, + "learning_rate": 1.8413450427547956e-05, + "loss": 0.0, + "step": 16401 + }, + { + "epoch": 3.79, + "learning_rate": 1.8411524535860104e-05, + "loss": 0.0, + "step": 16402 + }, + { + "epoch": 3.79, + "learning_rate": 1.840959864417225e-05, + "loss": 0.0, + "step": 16403 + }, + { + "epoch": 3.79, + "learning_rate": 1.8407672752484402e-05, + "loss": 0.0, + "step": 16404 + }, + { + "epoch": 3.79, + "learning_rate": 1.840574686079655e-05, + "loss": 0.0, + "step": 16405 + }, + { + "epoch": 3.79, + "learning_rate": 1.8403820969108697e-05, + "loss": 0.0, + "step": 16406 + }, + { + "epoch": 3.79, + "learning_rate": 1.8401895077420848e-05, + "loss": 0.0002, + "step": 16407 + }, + { + "epoch": 3.79, + "learning_rate": 1.8399969185732996e-05, + "loss": 0.0, + "step": 16408 + }, + { + "epoch": 3.79, + "learning_rate": 1.8398043294045143e-05, + "loss": 0.0, + "step": 16409 + }, + { + "epoch": 3.79, + "learning_rate": 1.8396117402357294e-05, + "loss": 0.0, + "step": 16410 + }, + { + "epoch": 3.79, + "learning_rate": 1.8394191510669438e-05, + "loss": 0.0019, + "step": 16411 + }, + { + "epoch": 3.79, + "learning_rate": 1.839226561898159e-05, + "loss": 0.0, + "step": 16412 + }, + { + "epoch": 3.79, + "learning_rate": 1.8390339727293736e-05, + "loss": 0.0, + "step": 16413 + }, + { + "epoch": 3.79, + "learning_rate": 1.8388413835605887e-05, + "loss": 0.0, + "step": 16414 + }, + { + "epoch": 3.79, + "learning_rate": 1.8386487943918035e-05, + "loss": 0.0, + "step": 16415 + }, + { + "epoch": 3.79, + "learning_rate": 1.8384562052230182e-05, + "loss": 0.0, + "step": 16416 + }, + { + "epoch": 3.79, + "learning_rate": 1.8382636160542333e-05, + "loss": 0.0, + "step": 16417 + }, + { + "epoch": 3.79, + "learning_rate": 1.838071026885448e-05, + "loss": 0.0, + "step": 16418 + }, + { + "epoch": 3.79, + "learning_rate": 1.837878437716663e-05, + "loss": 0.0, + "step": 16419 + }, + { + "epoch": 3.79, + "learning_rate": 1.8376858485478775e-05, + "loss": 0.0, + "step": 16420 + }, + { + "epoch": 3.8, + "learning_rate": 1.8374932593790926e-05, + "loss": 0.0, + "step": 16421 + }, + { + "epoch": 3.8, + "learning_rate": 1.8373006702103074e-05, + "loss": 0.0, + "step": 16422 + }, + { + "epoch": 3.8, + "learning_rate": 1.837108081041522e-05, + "loss": 0.0003, + "step": 16423 + }, + { + "epoch": 3.8, + "learning_rate": 1.8369154918727372e-05, + "loss": 0.0, + "step": 16424 + }, + { + "epoch": 3.8, + "learning_rate": 1.836722902703952e-05, + "loss": 0.0, + "step": 16425 + }, + { + "epoch": 3.8, + "learning_rate": 1.836530313535167e-05, + "loss": 0.0, + "step": 16426 + }, + { + "epoch": 3.8, + "learning_rate": 1.8363377243663818e-05, + "loss": 0.0, + "step": 16427 + }, + { + "epoch": 3.8, + "learning_rate": 1.8361451351975965e-05, + "loss": 0.0, + "step": 16428 + }, + { + "epoch": 3.8, + "learning_rate": 1.8359525460288116e-05, + "loss": 0.0, + "step": 16429 + }, + { + "epoch": 3.8, + "learning_rate": 1.835759956860026e-05, + "loss": 0.0001, + "step": 16430 + }, + { + "epoch": 3.8, + "learning_rate": 1.835567367691241e-05, + "loss": 0.0, + "step": 16431 + }, + { + "epoch": 3.8, + "learning_rate": 1.835374778522456e-05, + "loss": 0.0, + "step": 16432 + }, + { + "epoch": 3.8, + "learning_rate": 1.835182189353671e-05, + "loss": 0.0, + "step": 16433 + }, + { + "epoch": 3.8, + "learning_rate": 1.8349896001848857e-05, + "loss": 0.0, + "step": 16434 + }, + { + "epoch": 3.8, + "learning_rate": 1.8347970110161005e-05, + "loss": 0.0, + "step": 16435 + }, + { + "epoch": 3.8, + "learning_rate": 1.8346044218473155e-05, + "loss": 0.0, + "step": 16436 + }, + { + "epoch": 3.8, + "learning_rate": 1.8344118326785303e-05, + "loss": 0.0, + "step": 16437 + }, + { + "epoch": 3.8, + "learning_rate": 1.8342192435097454e-05, + "loss": 0.0, + "step": 16438 + }, + { + "epoch": 3.8, + "learning_rate": 1.8340266543409598e-05, + "loss": 0.0, + "step": 16439 + }, + { + "epoch": 3.8, + "learning_rate": 1.833834065172175e-05, + "loss": 0.0001, + "step": 16440 + }, + { + "epoch": 3.8, + "learning_rate": 1.8336414760033896e-05, + "loss": 0.0, + "step": 16441 + }, + { + "epoch": 3.8, + "learning_rate": 1.8334488868346044e-05, + "loss": 0.0, + "step": 16442 + }, + { + "epoch": 3.8, + "learning_rate": 1.8332562976658195e-05, + "loss": 0.0, + "step": 16443 + }, + { + "epoch": 3.8, + "learning_rate": 1.8330637084970342e-05, + "loss": 0.0, + "step": 16444 + }, + { + "epoch": 3.8, + "learning_rate": 1.832871119328249e-05, + "loss": 0.0, + "step": 16445 + }, + { + "epoch": 3.8, + "learning_rate": 1.832678530159464e-05, + "loss": 0.0, + "step": 16446 + }, + { + "epoch": 3.8, + "learning_rate": 1.8324859409906788e-05, + "loss": 0.0, + "step": 16447 + }, + { + "epoch": 3.8, + "learning_rate": 1.8322933518218935e-05, + "loss": 0.0, + "step": 16448 + }, + { + "epoch": 3.8, + "learning_rate": 1.8321007626531083e-05, + "loss": 0.0, + "step": 16449 + }, + { + "epoch": 3.8, + "learning_rate": 1.8319081734843234e-05, + "loss": 0.0, + "step": 16450 + }, + { + "epoch": 3.8, + "learning_rate": 1.831715584315538e-05, + "loss": 0.0, + "step": 16451 + }, + { + "epoch": 3.8, + "learning_rate": 1.831522995146753e-05, + "loss": 0.0, + "step": 16452 + }, + { + "epoch": 3.8, + "learning_rate": 1.831330405977968e-05, + "loss": 0.0, + "step": 16453 + }, + { + "epoch": 3.8, + "learning_rate": 1.8311378168091827e-05, + "loss": 0.0, + "step": 16454 + }, + { + "epoch": 3.8, + "learning_rate": 1.8309452276403978e-05, + "loss": 0.0, + "step": 16455 + }, + { + "epoch": 3.8, + "learning_rate": 1.8307526384716125e-05, + "loss": 0.0, + "step": 16456 + }, + { + "epoch": 3.8, + "learning_rate": 1.8305600493028273e-05, + "loss": 0.0, + "step": 16457 + }, + { + "epoch": 3.8, + "learning_rate": 1.830367460134042e-05, + "loss": 0.0, + "step": 16458 + }, + { + "epoch": 3.8, + "learning_rate": 1.8301748709652568e-05, + "loss": 0.0001, + "step": 16459 + }, + { + "epoch": 3.8, + "learning_rate": 1.829982281796472e-05, + "loss": 0.0005, + "step": 16460 + }, + { + "epoch": 3.8, + "learning_rate": 1.8297896926276866e-05, + "loss": 0.0, + "step": 16461 + }, + { + "epoch": 3.8, + "learning_rate": 1.8295971034589017e-05, + "loss": 0.0, + "step": 16462 + }, + { + "epoch": 3.8, + "learning_rate": 1.8294045142901164e-05, + "loss": 0.0001, + "step": 16463 + }, + { + "epoch": 3.8, + "learning_rate": 1.8292119251213312e-05, + "loss": 0.0, + "step": 16464 + }, + { + "epoch": 3.81, + "learning_rate": 1.8290193359525463e-05, + "loss": 0.0, + "step": 16465 + }, + { + "epoch": 3.81, + "learning_rate": 1.828826746783761e-05, + "loss": 0.0, + "step": 16466 + }, + { + "epoch": 3.81, + "learning_rate": 1.8286341576149758e-05, + "loss": 0.0, + "step": 16467 + }, + { + "epoch": 3.81, + "learning_rate": 1.8284415684461905e-05, + "loss": 0.0, + "step": 16468 + }, + { + "epoch": 3.81, + "learning_rate": 1.8282489792774056e-05, + "loss": 0.0, + "step": 16469 + }, + { + "epoch": 3.81, + "learning_rate": 1.8280563901086204e-05, + "loss": 0.0, + "step": 16470 + }, + { + "epoch": 3.81, + "learning_rate": 1.827863800939835e-05, + "loss": 0.0, + "step": 16471 + }, + { + "epoch": 3.81, + "learning_rate": 1.8276712117710502e-05, + "loss": 0.0, + "step": 16472 + }, + { + "epoch": 3.81, + "learning_rate": 1.827478622602265e-05, + "loss": 0.0, + "step": 16473 + }, + { + "epoch": 3.81, + "learning_rate": 1.82728603343348e-05, + "loss": 0.0, + "step": 16474 + }, + { + "epoch": 3.81, + "learning_rate": 1.8270934442646948e-05, + "loss": 0.0, + "step": 16475 + }, + { + "epoch": 3.81, + "learning_rate": 1.8269008550959095e-05, + "loss": 0.0, + "step": 16476 + }, + { + "epoch": 3.81, + "learning_rate": 1.8267082659271243e-05, + "loss": 0.0, + "step": 16477 + }, + { + "epoch": 3.81, + "learning_rate": 1.826515676758339e-05, + "loss": 0.0, + "step": 16478 + }, + { + "epoch": 3.81, + "learning_rate": 1.826323087589554e-05, + "loss": 0.0, + "step": 16479 + }, + { + "epoch": 3.81, + "learning_rate": 1.826130498420769e-05, + "loss": 0.0, + "step": 16480 + }, + { + "epoch": 3.81, + "learning_rate": 1.825937909251984e-05, + "loss": 0.0, + "step": 16481 + }, + { + "epoch": 3.81, + "learning_rate": 1.8257453200831987e-05, + "loss": 0.0, + "step": 16482 + }, + { + "epoch": 3.81, + "learning_rate": 1.8255527309144134e-05, + "loss": 0.0, + "step": 16483 + }, + { + "epoch": 3.81, + "learning_rate": 1.8253601417456285e-05, + "loss": 0.0, + "step": 16484 + }, + { + "epoch": 3.81, + "learning_rate": 1.8251675525768433e-05, + "loss": 0.0, + "step": 16485 + }, + { + "epoch": 3.81, + "learning_rate": 1.824974963408058e-05, + "loss": 0.0, + "step": 16486 + }, + { + "epoch": 3.81, + "learning_rate": 1.8247823742392728e-05, + "loss": 0.0, + "step": 16487 + }, + { + "epoch": 3.81, + "learning_rate": 1.8245897850704875e-05, + "loss": 0.0, + "step": 16488 + }, + { + "epoch": 3.81, + "learning_rate": 1.8243971959017026e-05, + "loss": 0.0001, + "step": 16489 + }, + { + "epoch": 3.81, + "learning_rate": 1.8242046067329173e-05, + "loss": 0.0, + "step": 16490 + }, + { + "epoch": 3.81, + "learning_rate": 1.8240120175641324e-05, + "loss": 0.0001, + "step": 16491 + }, + { + "epoch": 3.81, + "learning_rate": 1.8238194283953472e-05, + "loss": 0.0, + "step": 16492 + }, + { + "epoch": 3.81, + "learning_rate": 1.823626839226562e-05, + "loss": 0.0001, + "step": 16493 + }, + { + "epoch": 3.81, + "learning_rate": 1.823434250057777e-05, + "loss": 0.0, + "step": 16494 + }, + { + "epoch": 3.81, + "learning_rate": 1.8232416608889914e-05, + "loss": 0.0, + "step": 16495 + }, + { + "epoch": 3.81, + "learning_rate": 1.8230490717202065e-05, + "loss": 0.0, + "step": 16496 + }, + { + "epoch": 3.81, + "learning_rate": 1.8228564825514213e-05, + "loss": 0.0, + "step": 16497 + }, + { + "epoch": 3.81, + "learning_rate": 1.8226638933826363e-05, + "loss": 0.0, + "step": 16498 + }, + { + "epoch": 3.81, + "learning_rate": 1.822471304213851e-05, + "loss": 0.0, + "step": 16499 + }, + { + "epoch": 3.81, + "learning_rate": 1.822278715045066e-05, + "loss": 0.0, + "step": 16500 + }, + { + "epoch": 3.81, + "learning_rate": 1.822086125876281e-05, + "loss": 0.0, + "step": 16501 + }, + { + "epoch": 3.81, + "learning_rate": 1.8218935367074957e-05, + "loss": 0.0, + "step": 16502 + }, + { + "epoch": 3.81, + "learning_rate": 1.8217009475387108e-05, + "loss": 0.0, + "step": 16503 + }, + { + "epoch": 3.81, + "learning_rate": 1.821508358369925e-05, + "loss": 0.0, + "step": 16504 + }, + { + "epoch": 3.81, + "learning_rate": 1.8213157692011403e-05, + "loss": 0.0, + "step": 16505 + }, + { + "epoch": 3.81, + "learning_rate": 1.821123180032355e-05, + "loss": 0.0, + "step": 16506 + }, + { + "epoch": 3.81, + "learning_rate": 1.8209305908635698e-05, + "loss": 0.0012, + "step": 16507 + }, + { + "epoch": 3.82, + "learning_rate": 1.820738001694785e-05, + "loss": 0.0, + "step": 16508 + }, + { + "epoch": 3.82, + "learning_rate": 1.8205454125259996e-05, + "loss": 0.0, + "step": 16509 + }, + { + "epoch": 3.82, + "learning_rate": 1.8203528233572147e-05, + "loss": 0.0, + "step": 16510 + }, + { + "epoch": 3.82, + "learning_rate": 1.8201602341884294e-05, + "loss": 0.0, + "step": 16511 + }, + { + "epoch": 3.82, + "learning_rate": 1.819967645019644e-05, + "loss": 0.0, + "step": 16512 + }, + { + "epoch": 3.82, + "learning_rate": 1.8197750558508593e-05, + "loss": 0.0, + "step": 16513 + }, + { + "epoch": 3.82, + "learning_rate": 1.8195824666820737e-05, + "loss": 0.001, + "step": 16514 + }, + { + "epoch": 3.82, + "learning_rate": 1.8193898775132887e-05, + "loss": 0.0, + "step": 16515 + }, + { + "epoch": 3.82, + "learning_rate": 1.8191972883445035e-05, + "loss": 0.0001, + "step": 16516 + }, + { + "epoch": 3.82, + "learning_rate": 1.8190046991757186e-05, + "loss": 0.0, + "step": 16517 + }, + { + "epoch": 3.82, + "learning_rate": 1.8188121100069333e-05, + "loss": 0.0001, + "step": 16518 + }, + { + "epoch": 3.82, + "learning_rate": 1.818619520838148e-05, + "loss": 0.0, + "step": 16519 + }, + { + "epoch": 3.82, + "learning_rate": 1.818426931669363e-05, + "loss": 0.0, + "step": 16520 + }, + { + "epoch": 3.82, + "learning_rate": 1.818234342500578e-05, + "loss": 0.0, + "step": 16521 + }, + { + "epoch": 3.82, + "learning_rate": 1.8180417533317927e-05, + "loss": 0.0, + "step": 16522 + }, + { + "epoch": 3.82, + "learning_rate": 1.8178491641630074e-05, + "loss": 0.0, + "step": 16523 + }, + { + "epoch": 3.82, + "learning_rate": 1.817656574994222e-05, + "loss": 0.0, + "step": 16524 + }, + { + "epoch": 3.82, + "learning_rate": 1.8174639858254372e-05, + "loss": 0.0, + "step": 16525 + }, + { + "epoch": 3.82, + "learning_rate": 1.817271396656652e-05, + "loss": 0.0, + "step": 16526 + }, + { + "epoch": 3.82, + "learning_rate": 1.817078807487867e-05, + "loss": 0.0, + "step": 16527 + }, + { + "epoch": 3.82, + "learning_rate": 1.8168862183190818e-05, + "loss": 0.0, + "step": 16528 + }, + { + "epoch": 3.82, + "learning_rate": 1.8166936291502966e-05, + "loss": 0.0, + "step": 16529 + }, + { + "epoch": 3.82, + "learning_rate": 1.8165010399815117e-05, + "loss": 0.0, + "step": 16530 + }, + { + "epoch": 3.82, + "learning_rate": 1.8163084508127264e-05, + "loss": 0.0, + "step": 16531 + }, + { + "epoch": 3.82, + "learning_rate": 1.816115861643941e-05, + "loss": 0.0, + "step": 16532 + }, + { + "epoch": 3.82, + "learning_rate": 1.815923272475156e-05, + "loss": 0.0, + "step": 16533 + }, + { + "epoch": 3.82, + "learning_rate": 1.815730683306371e-05, + "loss": 0.0, + "step": 16534 + }, + { + "epoch": 3.82, + "learning_rate": 1.8155380941375857e-05, + "loss": 0.0, + "step": 16535 + }, + { + "epoch": 3.82, + "learning_rate": 1.8153455049688005e-05, + "loss": 0.0, + "step": 16536 + }, + { + "epoch": 3.82, + "learning_rate": 1.8151529158000156e-05, + "loss": 0.0, + "step": 16537 + }, + { + "epoch": 3.82, + "learning_rate": 1.8149603266312303e-05, + "loss": 0.0001, + "step": 16538 + }, + { + "epoch": 3.82, + "learning_rate": 1.8147677374624454e-05, + "loss": 0.0, + "step": 16539 + }, + { + "epoch": 3.82, + "learning_rate": 1.81457514829366e-05, + "loss": 0.0, + "step": 16540 + }, + { + "epoch": 3.82, + "learning_rate": 1.814382559124875e-05, + "loss": 0.0, + "step": 16541 + }, + { + "epoch": 3.82, + "learning_rate": 1.8141899699560897e-05, + "loss": 0.0, + "step": 16542 + }, + { + "epoch": 3.82, + "learning_rate": 1.8139973807873044e-05, + "loss": 0.0, + "step": 16543 + }, + { + "epoch": 3.82, + "learning_rate": 1.8138047916185195e-05, + "loss": 0.0, + "step": 16544 + }, + { + "epoch": 3.82, + "learning_rate": 1.8136122024497342e-05, + "loss": 0.0001, + "step": 16545 + }, + { + "epoch": 3.82, + "learning_rate": 1.8134196132809493e-05, + "loss": 0.0, + "step": 16546 + }, + { + "epoch": 3.82, + "learning_rate": 1.813227024112164e-05, + "loss": 0.0, + "step": 16547 + }, + { + "epoch": 3.82, + "learning_rate": 1.8130344349433788e-05, + "loss": 0.0, + "step": 16548 + }, + { + "epoch": 3.82, + "learning_rate": 1.812841845774594e-05, + "loss": 0.0, + "step": 16549 + }, + { + "epoch": 3.82, + "learning_rate": 1.8126492566058086e-05, + "loss": 0.0, + "step": 16550 + }, + { + "epoch": 3.83, + "learning_rate": 1.8124566674370234e-05, + "loss": 0.0, + "step": 16551 + }, + { + "epoch": 3.83, + "learning_rate": 1.812264078268238e-05, + "loss": 0.0, + "step": 16552 + }, + { + "epoch": 3.83, + "learning_rate": 1.8120714890994532e-05, + "loss": 0.0, + "step": 16553 + }, + { + "epoch": 3.83, + "learning_rate": 1.811878899930668e-05, + "loss": 0.0, + "step": 16554 + }, + { + "epoch": 3.83, + "learning_rate": 1.8116863107618827e-05, + "loss": 0.0, + "step": 16555 + }, + { + "epoch": 3.83, + "learning_rate": 1.8114937215930978e-05, + "loss": 0.0, + "step": 16556 + }, + { + "epoch": 3.83, + "learning_rate": 1.8113011324243126e-05, + "loss": 0.0, + "step": 16557 + }, + { + "epoch": 3.83, + "learning_rate": 1.8111085432555276e-05, + "loss": 0.0, + "step": 16558 + }, + { + "epoch": 3.83, + "learning_rate": 1.8109159540867424e-05, + "loss": 0.0, + "step": 16559 + }, + { + "epoch": 3.83, + "learning_rate": 1.8107233649179568e-05, + "loss": 0.0, + "step": 16560 + }, + { + "epoch": 3.83, + "learning_rate": 1.810530775749172e-05, + "loss": 0.0, + "step": 16561 + }, + { + "epoch": 3.83, + "learning_rate": 1.8103381865803866e-05, + "loss": 0.0, + "step": 16562 + }, + { + "epoch": 3.83, + "learning_rate": 1.8101455974116017e-05, + "loss": 0.0, + "step": 16563 + }, + { + "epoch": 3.83, + "learning_rate": 1.8099530082428165e-05, + "loss": 0.0, + "step": 16564 + }, + { + "epoch": 3.83, + "learning_rate": 1.8097604190740312e-05, + "loss": 0.0, + "step": 16565 + }, + { + "epoch": 3.83, + "learning_rate": 1.8095678299052463e-05, + "loss": 0.0, + "step": 16566 + }, + { + "epoch": 3.83, + "learning_rate": 1.809375240736461e-05, + "loss": 0.0, + "step": 16567 + }, + { + "epoch": 3.83, + "learning_rate": 1.809182651567676e-05, + "loss": 0.0, + "step": 16568 + }, + { + "epoch": 3.83, + "learning_rate": 1.808990062398891e-05, + "loss": 0.0, + "step": 16569 + }, + { + "epoch": 3.83, + "learning_rate": 1.8087974732301056e-05, + "loss": 0.0, + "step": 16570 + }, + { + "epoch": 3.83, + "learning_rate": 1.8086048840613204e-05, + "loss": 0.0, + "step": 16571 + }, + { + "epoch": 3.83, + "learning_rate": 1.808412294892535e-05, + "loss": 0.0, + "step": 16572 + }, + { + "epoch": 3.83, + "learning_rate": 1.8082197057237502e-05, + "loss": 0.0, + "step": 16573 + }, + { + "epoch": 3.83, + "learning_rate": 1.808027116554965e-05, + "loss": 0.0, + "step": 16574 + }, + { + "epoch": 3.83, + "learning_rate": 1.80783452738618e-05, + "loss": 0.0, + "step": 16575 + }, + { + "epoch": 3.83, + "learning_rate": 1.8076419382173948e-05, + "loss": 0.0, + "step": 16576 + }, + { + "epoch": 3.83, + "learning_rate": 1.8074493490486095e-05, + "loss": 0.0, + "step": 16577 + }, + { + "epoch": 3.83, + "learning_rate": 1.8072567598798246e-05, + "loss": 0.0001, + "step": 16578 + }, + { + "epoch": 3.83, + "learning_rate": 1.807064170711039e-05, + "loss": 0.0, + "step": 16579 + }, + { + "epoch": 3.83, + "learning_rate": 1.806871581542254e-05, + "loss": 0.0, + "step": 16580 + }, + { + "epoch": 3.83, + "learning_rate": 1.806678992373469e-05, + "loss": 0.0, + "step": 16581 + }, + { + "epoch": 3.83, + "learning_rate": 1.806486403204684e-05, + "loss": 0.0, + "step": 16582 + }, + { + "epoch": 3.83, + "learning_rate": 1.8062938140358987e-05, + "loss": 0.0, + "step": 16583 + }, + { + "epoch": 3.83, + "learning_rate": 1.8061012248671135e-05, + "loss": 0.0, + "step": 16584 + }, + { + "epoch": 3.83, + "learning_rate": 1.8059086356983285e-05, + "loss": 0.0, + "step": 16585 + }, + { + "epoch": 3.83, + "learning_rate": 1.8057160465295433e-05, + "loss": 0.0, + "step": 16586 + }, + { + "epoch": 3.83, + "learning_rate": 1.8055234573607584e-05, + "loss": 0.0, + "step": 16587 + }, + { + "epoch": 3.83, + "learning_rate": 1.8053308681919728e-05, + "loss": 0.0001, + "step": 16588 + }, + { + "epoch": 3.83, + "learning_rate": 1.805138279023188e-05, + "loss": 0.0, + "step": 16589 + }, + { + "epoch": 3.83, + "learning_rate": 1.8049456898544026e-05, + "loss": 0.0, + "step": 16590 + }, + { + "epoch": 3.83, + "learning_rate": 1.8047531006856174e-05, + "loss": 0.0, + "step": 16591 + }, + { + "epoch": 3.83, + "learning_rate": 1.8045605115168325e-05, + "loss": 0.0, + "step": 16592 + }, + { + "epoch": 3.83, + "learning_rate": 1.8043679223480472e-05, + "loss": 0.0, + "step": 16593 + }, + { + "epoch": 3.83, + "learning_rate": 1.8041753331792623e-05, + "loss": 0.0, + "step": 16594 + }, + { + "epoch": 3.84, + "learning_rate": 1.803982744010477e-05, + "loss": 0.0, + "step": 16595 + }, + { + "epoch": 3.84, + "learning_rate": 1.8037901548416918e-05, + "loss": 0.0, + "step": 16596 + }, + { + "epoch": 3.84, + "learning_rate": 1.803597565672907e-05, + "loss": 0.0, + "step": 16597 + }, + { + "epoch": 3.84, + "learning_rate": 1.8034049765041213e-05, + "loss": 0.0, + "step": 16598 + }, + { + "epoch": 3.84, + "learning_rate": 1.8032123873353364e-05, + "loss": 0.0, + "step": 16599 + }, + { + "epoch": 3.84, + "learning_rate": 1.803019798166551e-05, + "loss": 0.0, + "step": 16600 + }, + { + "epoch": 3.84, + "learning_rate": 1.802827208997766e-05, + "loss": 0.0, + "step": 16601 + }, + { + "epoch": 3.84, + "learning_rate": 1.802634619828981e-05, + "loss": 0.0, + "step": 16602 + }, + { + "epoch": 3.84, + "learning_rate": 1.8024420306601957e-05, + "loss": 0.0, + "step": 16603 + }, + { + "epoch": 3.84, + "learning_rate": 1.8022494414914108e-05, + "loss": 0.0, + "step": 16604 + }, + { + "epoch": 3.84, + "learning_rate": 1.8020568523226255e-05, + "loss": 0.0, + "step": 16605 + }, + { + "epoch": 3.84, + "learning_rate": 1.8018642631538403e-05, + "loss": 0.0, + "step": 16606 + }, + { + "epoch": 3.84, + "learning_rate": 1.801671673985055e-05, + "loss": 0.0, + "step": 16607 + }, + { + "epoch": 3.84, + "learning_rate": 1.8014790848162698e-05, + "loss": 0.0003, + "step": 16608 + }, + { + "epoch": 3.84, + "learning_rate": 1.801286495647485e-05, + "loss": 0.0002, + "step": 16609 + }, + { + "epoch": 3.84, + "learning_rate": 1.8010939064786996e-05, + "loss": 0.0, + "step": 16610 + }, + { + "epoch": 3.84, + "learning_rate": 1.8009013173099147e-05, + "loss": 0.0, + "step": 16611 + }, + { + "epoch": 3.84, + "learning_rate": 1.8007087281411294e-05, + "loss": 0.0, + "step": 16612 + }, + { + "epoch": 3.84, + "learning_rate": 1.8005161389723442e-05, + "loss": 0.0, + "step": 16613 + }, + { + "epoch": 3.84, + "learning_rate": 1.8003235498035593e-05, + "loss": 0.0, + "step": 16614 + }, + { + "epoch": 3.84, + "learning_rate": 1.800130960634774e-05, + "loss": 0.0, + "step": 16615 + }, + { + "epoch": 3.84, + "learning_rate": 1.7999383714659888e-05, + "loss": 0.0, + "step": 16616 + }, + { + "epoch": 3.84, + "learning_rate": 1.7997457822972035e-05, + "loss": 0.0, + "step": 16617 + }, + { + "epoch": 3.84, + "learning_rate": 1.7995531931284186e-05, + "loss": 0.0, + "step": 16618 + }, + { + "epoch": 3.84, + "learning_rate": 1.7993606039596334e-05, + "loss": 0.0001, + "step": 16619 + }, + { + "epoch": 3.84, + "learning_rate": 1.799168014790848e-05, + "loss": 0.0, + "step": 16620 + }, + { + "epoch": 3.84, + "learning_rate": 1.7989754256220632e-05, + "loss": 0.0, + "step": 16621 + }, + { + "epoch": 3.84, + "learning_rate": 1.798782836453278e-05, + "loss": 0.0, + "step": 16622 + }, + { + "epoch": 3.84, + "learning_rate": 1.798590247284493e-05, + "loss": 0.0, + "step": 16623 + }, + { + "epoch": 3.84, + "learning_rate": 1.7983976581157078e-05, + "loss": 0.0, + "step": 16624 + }, + { + "epoch": 3.84, + "learning_rate": 1.7982050689469225e-05, + "loss": 0.0, + "step": 16625 + }, + { + "epoch": 3.84, + "learning_rate": 1.7980124797781373e-05, + "loss": 0.0, + "step": 16626 + }, + { + "epoch": 3.84, + "learning_rate": 1.797819890609352e-05, + "loss": 0.0, + "step": 16627 + }, + { + "epoch": 3.84, + "learning_rate": 1.797627301440567e-05, + "loss": 0.0, + "step": 16628 + }, + { + "epoch": 3.84, + "learning_rate": 1.797434712271782e-05, + "loss": 0.0, + "step": 16629 + }, + { + "epoch": 3.84, + "learning_rate": 1.797242123102997e-05, + "loss": 0.0, + "step": 16630 + }, + { + "epoch": 3.84, + "learning_rate": 1.7970495339342117e-05, + "loss": 0.0, + "step": 16631 + }, + { + "epoch": 3.84, + "learning_rate": 1.7968569447654264e-05, + "loss": 0.0, + "step": 16632 + }, + { + "epoch": 3.84, + "learning_rate": 1.7966643555966415e-05, + "loss": 0.0, + "step": 16633 + }, + { + "epoch": 3.84, + "learning_rate": 1.7964717664278563e-05, + "loss": 0.0, + "step": 16634 + }, + { + "epoch": 3.84, + "learning_rate": 1.796279177259071e-05, + "loss": 0.0, + "step": 16635 + }, + { + "epoch": 3.84, + "learning_rate": 1.7960865880902858e-05, + "loss": 0.0, + "step": 16636 + }, + { + "epoch": 3.84, + "learning_rate": 1.7958939989215005e-05, + "loss": 0.0, + "step": 16637 + }, + { + "epoch": 3.85, + "learning_rate": 1.7957014097527156e-05, + "loss": 0.0, + "step": 16638 + }, + { + "epoch": 3.85, + "learning_rate": 1.7955088205839303e-05, + "loss": 0.0001, + "step": 16639 + }, + { + "epoch": 3.85, + "learning_rate": 1.7953162314151454e-05, + "loss": 0.0, + "step": 16640 + }, + { + "epoch": 3.85, + "learning_rate": 1.7951236422463602e-05, + "loss": 0.0, + "step": 16641 + }, + { + "epoch": 3.85, + "learning_rate": 1.794931053077575e-05, + "loss": 0.0, + "step": 16642 + }, + { + "epoch": 3.85, + "learning_rate": 1.79473846390879e-05, + "loss": 0.0, + "step": 16643 + }, + { + "epoch": 3.85, + "learning_rate": 1.7945458747400044e-05, + "loss": 0.0, + "step": 16644 + }, + { + "epoch": 3.85, + "learning_rate": 1.7943532855712195e-05, + "loss": 0.0, + "step": 16645 + }, + { + "epoch": 3.85, + "learning_rate": 1.7941606964024343e-05, + "loss": 0.0, + "step": 16646 + }, + { + "epoch": 3.85, + "learning_rate": 1.7939681072336493e-05, + "loss": 0.0, + "step": 16647 + }, + { + "epoch": 3.85, + "learning_rate": 1.793775518064864e-05, + "loss": 0.0, + "step": 16648 + }, + { + "epoch": 3.85, + "learning_rate": 1.793582928896079e-05, + "loss": 0.0, + "step": 16649 + }, + { + "epoch": 3.85, + "learning_rate": 1.793390339727294e-05, + "loss": 0.0, + "step": 16650 + }, + { + "epoch": 3.85, + "learning_rate": 1.7931977505585087e-05, + "loss": 0.0, + "step": 16651 + }, + { + "epoch": 3.85, + "learning_rate": 1.7930051613897238e-05, + "loss": 0.0, + "step": 16652 + }, + { + "epoch": 3.85, + "learning_rate": 1.7928125722209385e-05, + "loss": 0.0, + "step": 16653 + }, + { + "epoch": 3.85, + "learning_rate": 1.7926199830521533e-05, + "loss": 0.0, + "step": 16654 + }, + { + "epoch": 3.85, + "learning_rate": 1.792427393883368e-05, + "loss": 0.0, + "step": 16655 + }, + { + "epoch": 3.85, + "learning_rate": 1.7922348047145828e-05, + "loss": 0.0, + "step": 16656 + }, + { + "epoch": 3.85, + "learning_rate": 1.792042215545798e-05, + "loss": 0.0, + "step": 16657 + }, + { + "epoch": 3.85, + "learning_rate": 1.7918496263770126e-05, + "loss": 0.0, + "step": 16658 + }, + { + "epoch": 3.85, + "learning_rate": 1.7916570372082277e-05, + "loss": 0.0, + "step": 16659 + }, + { + "epoch": 3.85, + "learning_rate": 1.7914644480394424e-05, + "loss": 0.0, + "step": 16660 + }, + { + "epoch": 3.85, + "learning_rate": 1.7912718588706572e-05, + "loss": 0.0, + "step": 16661 + }, + { + "epoch": 3.85, + "learning_rate": 1.7910792697018723e-05, + "loss": 0.0, + "step": 16662 + }, + { + "epoch": 3.85, + "learning_rate": 1.7908866805330867e-05, + "loss": 0.0, + "step": 16663 + }, + { + "epoch": 3.85, + "learning_rate": 1.7906940913643018e-05, + "loss": 0.0, + "step": 16664 + }, + { + "epoch": 3.85, + "learning_rate": 1.7905015021955165e-05, + "loss": 0.0002, + "step": 16665 + }, + { + "epoch": 3.85, + "learning_rate": 1.7903089130267316e-05, + "loss": 0.0001, + "step": 16666 + }, + { + "epoch": 3.85, + "learning_rate": 1.7901163238579463e-05, + "loss": 0.0, + "step": 16667 + }, + { + "epoch": 3.85, + "learning_rate": 1.789923734689161e-05, + "loss": 0.0, + "step": 16668 + }, + { + "epoch": 3.85, + "learning_rate": 1.7897311455203762e-05, + "loss": 0.0, + "step": 16669 + }, + { + "epoch": 3.85, + "learning_rate": 1.789538556351591e-05, + "loss": 0.0, + "step": 16670 + }, + { + "epoch": 3.85, + "learning_rate": 1.789345967182806e-05, + "loss": 0.0, + "step": 16671 + }, + { + "epoch": 3.85, + "learning_rate": 1.7891533780140204e-05, + "loss": 0.0, + "step": 16672 + }, + { + "epoch": 3.85, + "learning_rate": 1.788960788845235e-05, + "loss": 0.0, + "step": 16673 + }, + { + "epoch": 3.85, + "learning_rate": 1.7887681996764502e-05, + "loss": 0.0, + "step": 16674 + }, + { + "epoch": 3.85, + "learning_rate": 1.788575610507665e-05, + "loss": 0.0, + "step": 16675 + }, + { + "epoch": 3.85, + "learning_rate": 1.78838302133888e-05, + "loss": 0.0001, + "step": 16676 + }, + { + "epoch": 3.85, + "learning_rate": 1.7881904321700948e-05, + "loss": 0.0, + "step": 16677 + }, + { + "epoch": 3.85, + "learning_rate": 1.7879978430013096e-05, + "loss": 0.0, + "step": 16678 + }, + { + "epoch": 3.85, + "learning_rate": 1.7878052538325247e-05, + "loss": 0.0, + "step": 16679 + }, + { + "epoch": 3.85, + "learning_rate": 1.7876126646637394e-05, + "loss": 0.0, + "step": 16680 + }, + { + "epoch": 3.86, + "learning_rate": 1.7874200754949545e-05, + "loss": 0.0, + "step": 16681 + }, + { + "epoch": 3.86, + "learning_rate": 1.787227486326169e-05, + "loss": 0.0, + "step": 16682 + }, + { + "epoch": 3.86, + "learning_rate": 1.787034897157384e-05, + "loss": 0.0004, + "step": 16683 + }, + { + "epoch": 3.86, + "learning_rate": 1.7868423079885987e-05, + "loss": 0.0, + "step": 16684 + }, + { + "epoch": 3.86, + "learning_rate": 1.7866497188198135e-05, + "loss": 0.0, + "step": 16685 + }, + { + "epoch": 3.86, + "learning_rate": 1.7864571296510286e-05, + "loss": 0.0001, + "step": 16686 + }, + { + "epoch": 3.86, + "learning_rate": 1.7862645404822433e-05, + "loss": 0.0, + "step": 16687 + }, + { + "epoch": 3.86, + "learning_rate": 1.7860719513134584e-05, + "loss": 0.0, + "step": 16688 + }, + { + "epoch": 3.86, + "learning_rate": 1.785879362144673e-05, + "loss": 0.0, + "step": 16689 + }, + { + "epoch": 3.86, + "learning_rate": 1.785686772975888e-05, + "loss": 0.0, + "step": 16690 + }, + { + "epoch": 3.86, + "learning_rate": 1.7854941838071027e-05, + "loss": 0.0001, + "step": 16691 + }, + { + "epoch": 3.86, + "learning_rate": 1.7853015946383174e-05, + "loss": 0.0, + "step": 16692 + }, + { + "epoch": 3.86, + "learning_rate": 1.7851090054695325e-05, + "loss": 0.0, + "step": 16693 + }, + { + "epoch": 3.86, + "learning_rate": 1.7849164163007472e-05, + "loss": 0.0, + "step": 16694 + }, + { + "epoch": 3.86, + "learning_rate": 1.7847238271319623e-05, + "loss": 0.0, + "step": 16695 + }, + { + "epoch": 3.86, + "learning_rate": 1.784531237963177e-05, + "loss": 0.0, + "step": 16696 + }, + { + "epoch": 3.86, + "learning_rate": 1.7843386487943918e-05, + "loss": 0.0, + "step": 16697 + }, + { + "epoch": 3.86, + "learning_rate": 1.784146059625607e-05, + "loss": 0.0, + "step": 16698 + }, + { + "epoch": 3.86, + "learning_rate": 1.7839534704568217e-05, + "loss": 0.0, + "step": 16699 + }, + { + "epoch": 3.86, + "learning_rate": 1.7837608812880364e-05, + "loss": 0.0, + "step": 16700 + }, + { + "epoch": 3.86, + "learning_rate": 1.783568292119251e-05, + "loss": 0.0, + "step": 16701 + }, + { + "epoch": 3.86, + "learning_rate": 1.7833757029504662e-05, + "loss": 0.0, + "step": 16702 + }, + { + "epoch": 3.86, + "learning_rate": 1.783183113781681e-05, + "loss": 0.0, + "step": 16703 + }, + { + "epoch": 3.86, + "learning_rate": 1.7829905246128957e-05, + "loss": 0.0007, + "step": 16704 + }, + { + "epoch": 3.86, + "learning_rate": 1.7827979354441108e-05, + "loss": 0.0, + "step": 16705 + }, + { + "epoch": 3.86, + "learning_rate": 1.7826053462753256e-05, + "loss": 0.0, + "step": 16706 + }, + { + "epoch": 3.86, + "learning_rate": 1.7824127571065407e-05, + "loss": 0.0, + "step": 16707 + }, + { + "epoch": 3.86, + "learning_rate": 1.7822201679377554e-05, + "loss": 0.0, + "step": 16708 + }, + { + "epoch": 3.86, + "learning_rate": 1.78202757876897e-05, + "loss": 0.0, + "step": 16709 + }, + { + "epoch": 3.86, + "learning_rate": 1.781834989600185e-05, + "loss": 0.0, + "step": 16710 + }, + { + "epoch": 3.86, + "learning_rate": 1.7816424004313996e-05, + "loss": 0.0, + "step": 16711 + }, + { + "epoch": 3.86, + "learning_rate": 1.7814498112626147e-05, + "loss": 0.0, + "step": 16712 + }, + { + "epoch": 3.86, + "learning_rate": 1.7812572220938295e-05, + "loss": 0.0, + "step": 16713 + }, + { + "epoch": 3.86, + "learning_rate": 1.7810646329250442e-05, + "loss": 0.0, + "step": 16714 + }, + { + "epoch": 3.86, + "learning_rate": 1.7808720437562593e-05, + "loss": 0.0, + "step": 16715 + }, + { + "epoch": 3.86, + "learning_rate": 1.780679454587474e-05, + "loss": 0.0, + "step": 16716 + }, + { + "epoch": 3.86, + "learning_rate": 1.780486865418689e-05, + "loss": 0.0, + "step": 16717 + }, + { + "epoch": 3.86, + "learning_rate": 1.780294276249904e-05, + "loss": 0.0, + "step": 16718 + }, + { + "epoch": 3.86, + "learning_rate": 1.7801016870811186e-05, + "loss": 0.0, + "step": 16719 + }, + { + "epoch": 3.86, + "learning_rate": 1.7799090979123334e-05, + "loss": 0.0, + "step": 16720 + }, + { + "epoch": 3.86, + "learning_rate": 1.779716508743548e-05, + "loss": 0.0, + "step": 16721 + }, + { + "epoch": 3.86, + "learning_rate": 1.7795239195747632e-05, + "loss": 0.0, + "step": 16722 + }, + { + "epoch": 3.86, + "learning_rate": 1.779331330405978e-05, + "loss": 0.0, + "step": 16723 + }, + { + "epoch": 3.87, + "learning_rate": 1.779138741237193e-05, + "loss": 0.0, + "step": 16724 + }, + { + "epoch": 3.87, + "learning_rate": 1.7789461520684078e-05, + "loss": 0.0, + "step": 16725 + }, + { + "epoch": 3.87, + "learning_rate": 1.7787535628996226e-05, + "loss": 0.0, + "step": 16726 + }, + { + "epoch": 3.87, + "learning_rate": 1.7785609737308376e-05, + "loss": 0.0, + "step": 16727 + }, + { + "epoch": 3.87, + "learning_rate": 1.778368384562052e-05, + "loss": 0.0, + "step": 16728 + }, + { + "epoch": 3.87, + "learning_rate": 1.778175795393267e-05, + "loss": 0.0, + "step": 16729 + }, + { + "epoch": 3.87, + "learning_rate": 1.777983206224482e-05, + "loss": 0.0, + "step": 16730 + }, + { + "epoch": 3.87, + "learning_rate": 1.777790617055697e-05, + "loss": 0.0, + "step": 16731 + }, + { + "epoch": 3.87, + "learning_rate": 1.7775980278869117e-05, + "loss": 0.0, + "step": 16732 + }, + { + "epoch": 3.87, + "learning_rate": 1.7774054387181265e-05, + "loss": 0.0, + "step": 16733 + }, + { + "epoch": 3.87, + "learning_rate": 1.7772128495493416e-05, + "loss": 0.0, + "step": 16734 + }, + { + "epoch": 3.87, + "learning_rate": 1.7770202603805563e-05, + "loss": 0.0, + "step": 16735 + }, + { + "epoch": 3.87, + "learning_rate": 1.7768276712117714e-05, + "loss": 0.0, + "step": 16736 + }, + { + "epoch": 3.87, + "learning_rate": 1.776635082042986e-05, + "loss": 0.0, + "step": 16737 + }, + { + "epoch": 3.87, + "learning_rate": 1.776442492874201e-05, + "loss": 0.0001, + "step": 16738 + }, + { + "epoch": 3.87, + "learning_rate": 1.7762499037054156e-05, + "loss": 0.0, + "step": 16739 + }, + { + "epoch": 3.87, + "learning_rate": 1.7760573145366304e-05, + "loss": 0.0, + "step": 16740 + }, + { + "epoch": 3.87, + "learning_rate": 1.7758647253678455e-05, + "loss": 0.0, + "step": 16741 + }, + { + "epoch": 3.87, + "learning_rate": 1.7756721361990602e-05, + "loss": 0.0, + "step": 16742 + }, + { + "epoch": 3.87, + "learning_rate": 1.7754795470302753e-05, + "loss": 0.0, + "step": 16743 + }, + { + "epoch": 3.87, + "learning_rate": 1.77528695786149e-05, + "loss": 0.0, + "step": 16744 + }, + { + "epoch": 3.87, + "learning_rate": 1.7750943686927048e-05, + "loss": 0.0, + "step": 16745 + }, + { + "epoch": 3.87, + "learning_rate": 1.77490177952392e-05, + "loss": 0.0, + "step": 16746 + }, + { + "epoch": 3.87, + "learning_rate": 1.7747091903551343e-05, + "loss": 0.0, + "step": 16747 + }, + { + "epoch": 3.87, + "learning_rate": 1.7745166011863494e-05, + "loss": 0.0, + "step": 16748 + }, + { + "epoch": 3.87, + "learning_rate": 1.774324012017564e-05, + "loss": 0.0001, + "step": 16749 + }, + { + "epoch": 3.87, + "learning_rate": 1.774131422848779e-05, + "loss": 0.0, + "step": 16750 + }, + { + "epoch": 3.87, + "learning_rate": 1.773938833679994e-05, + "loss": 0.0001, + "step": 16751 + }, + { + "epoch": 3.87, + "learning_rate": 1.7737462445112087e-05, + "loss": 0.0002, + "step": 16752 + }, + { + "epoch": 3.87, + "learning_rate": 1.7735536553424238e-05, + "loss": 0.0, + "step": 16753 + }, + { + "epoch": 3.87, + "learning_rate": 1.7733610661736385e-05, + "loss": 0.0, + "step": 16754 + }, + { + "epoch": 3.87, + "learning_rate": 1.7731684770048533e-05, + "loss": 0.0, + "step": 16755 + }, + { + "epoch": 3.87, + "learning_rate": 1.772975887836068e-05, + "loss": 0.0, + "step": 16756 + }, + { + "epoch": 3.87, + "learning_rate": 1.7727832986672828e-05, + "loss": 0.0, + "step": 16757 + }, + { + "epoch": 3.87, + "learning_rate": 1.772590709498498e-05, + "loss": 0.0, + "step": 16758 + }, + { + "epoch": 3.87, + "learning_rate": 1.7723981203297126e-05, + "loss": 0.0, + "step": 16759 + }, + { + "epoch": 3.87, + "learning_rate": 1.7722055311609277e-05, + "loss": 0.0, + "step": 16760 + }, + { + "epoch": 3.87, + "learning_rate": 1.7720129419921425e-05, + "loss": 0.0, + "step": 16761 + }, + { + "epoch": 3.87, + "learning_rate": 1.7718203528233572e-05, + "loss": 0.0, + "step": 16762 + }, + { + "epoch": 3.87, + "learning_rate": 1.7716277636545723e-05, + "loss": 0.0, + "step": 16763 + }, + { + "epoch": 3.87, + "learning_rate": 1.771435174485787e-05, + "loss": 0.0, + "step": 16764 + }, + { + "epoch": 3.87, + "learning_rate": 1.771242585317002e-05, + "loss": 0.0, + "step": 16765 + }, + { + "epoch": 3.87, + "learning_rate": 1.7710499961482165e-05, + "loss": 0.0, + "step": 16766 + }, + { + "epoch": 3.87, + "learning_rate": 1.7708574069794316e-05, + "loss": 0.0001, + "step": 16767 + }, + { + "epoch": 3.88, + "learning_rate": 1.7706648178106464e-05, + "loss": 0.0, + "step": 16768 + }, + { + "epoch": 3.88, + "learning_rate": 1.770472228641861e-05, + "loss": 0.0, + "step": 16769 + }, + { + "epoch": 3.88, + "learning_rate": 1.7702796394730762e-05, + "loss": 0.0, + "step": 16770 + }, + { + "epoch": 3.88, + "learning_rate": 1.770087050304291e-05, + "loss": 0.0, + "step": 16771 + }, + { + "epoch": 3.88, + "learning_rate": 1.769894461135506e-05, + "loss": 0.0, + "step": 16772 + }, + { + "epoch": 3.88, + "learning_rate": 1.7697018719667208e-05, + "loss": 0.0, + "step": 16773 + }, + { + "epoch": 3.88, + "learning_rate": 1.7695092827979355e-05, + "loss": 0.0, + "step": 16774 + }, + { + "epoch": 3.88, + "learning_rate": 1.7693166936291503e-05, + "loss": 0.0, + "step": 16775 + }, + { + "epoch": 3.88, + "learning_rate": 1.769124104460365e-05, + "loss": 0.0, + "step": 16776 + }, + { + "epoch": 3.88, + "learning_rate": 1.76893151529158e-05, + "loss": 0.0, + "step": 16777 + }, + { + "epoch": 3.88, + "learning_rate": 1.768738926122795e-05, + "loss": 0.0, + "step": 16778 + }, + { + "epoch": 3.88, + "learning_rate": 1.76854633695401e-05, + "loss": 0.0, + "step": 16779 + }, + { + "epoch": 3.88, + "learning_rate": 1.7683537477852247e-05, + "loss": 0.0, + "step": 16780 + }, + { + "epoch": 3.88, + "learning_rate": 1.7681611586164394e-05, + "loss": 0.0, + "step": 16781 + }, + { + "epoch": 3.88, + "learning_rate": 1.7679685694476545e-05, + "loss": 0.0, + "step": 16782 + }, + { + "epoch": 3.88, + "learning_rate": 1.7677759802788693e-05, + "loss": 0.0, + "step": 16783 + }, + { + "epoch": 3.88, + "learning_rate": 1.767583391110084e-05, + "loss": 0.0, + "step": 16784 + }, + { + "epoch": 3.88, + "learning_rate": 1.7673908019412988e-05, + "loss": 0.0, + "step": 16785 + }, + { + "epoch": 3.88, + "learning_rate": 1.7671982127725135e-05, + "loss": 0.0, + "step": 16786 + }, + { + "epoch": 3.88, + "learning_rate": 1.7670056236037286e-05, + "loss": 0.0, + "step": 16787 + }, + { + "epoch": 3.88, + "learning_rate": 1.7668130344349434e-05, + "loss": 0.0, + "step": 16788 + }, + { + "epoch": 3.88, + "learning_rate": 1.7666204452661584e-05, + "loss": 0.0, + "step": 16789 + }, + { + "epoch": 3.88, + "learning_rate": 1.7664278560973732e-05, + "loss": 0.0, + "step": 16790 + }, + { + "epoch": 3.88, + "learning_rate": 1.766235266928588e-05, + "loss": 0.0, + "step": 16791 + }, + { + "epoch": 3.88, + "learning_rate": 1.766042677759803e-05, + "loss": 0.0002, + "step": 16792 + }, + { + "epoch": 3.88, + "learning_rate": 1.7658500885910178e-05, + "loss": 0.0001, + "step": 16793 + }, + { + "epoch": 3.88, + "learning_rate": 1.7656574994222325e-05, + "loss": 0.0, + "step": 16794 + }, + { + "epoch": 3.88, + "learning_rate": 1.7654649102534473e-05, + "loss": 0.0, + "step": 16795 + }, + { + "epoch": 3.88, + "learning_rate": 1.7652723210846624e-05, + "loss": 0.0, + "step": 16796 + }, + { + "epoch": 3.88, + "learning_rate": 1.765079731915877e-05, + "loss": 0.0, + "step": 16797 + }, + { + "epoch": 3.88, + "learning_rate": 1.764887142747092e-05, + "loss": 0.0, + "step": 16798 + }, + { + "epoch": 3.88, + "learning_rate": 1.764694553578307e-05, + "loss": 0.0, + "step": 16799 + }, + { + "epoch": 3.88, + "learning_rate": 1.7645019644095217e-05, + "loss": 0.0, + "step": 16800 + }, + { + "epoch": 3.88, + "learning_rate": 1.7643093752407368e-05, + "loss": 0.0, + "step": 16801 + }, + { + "epoch": 3.88, + "learning_rate": 1.7641167860719515e-05, + "loss": 0.0, + "step": 16802 + }, + { + "epoch": 3.88, + "learning_rate": 1.7639241969031663e-05, + "loss": 0.0, + "step": 16803 + }, + { + "epoch": 3.88, + "learning_rate": 1.763731607734381e-05, + "loss": 0.0, + "step": 16804 + }, + { + "epoch": 3.88, + "learning_rate": 1.7635390185655958e-05, + "loss": 0.0, + "step": 16805 + }, + { + "epoch": 3.88, + "learning_rate": 1.763346429396811e-05, + "loss": 0.0, + "step": 16806 + }, + { + "epoch": 3.88, + "learning_rate": 1.7631538402280256e-05, + "loss": 0.0001, + "step": 16807 + }, + { + "epoch": 3.88, + "learning_rate": 1.7629612510592407e-05, + "loss": 0.0, + "step": 16808 + }, + { + "epoch": 3.88, + "learning_rate": 1.7627686618904554e-05, + "loss": 0.0, + "step": 16809 + }, + { + "epoch": 3.88, + "learning_rate": 1.7625760727216702e-05, + "loss": 0.0, + "step": 16810 + }, + { + "epoch": 3.89, + "learning_rate": 1.7623834835528853e-05, + "loss": 0.0, + "step": 16811 + }, + { + "epoch": 3.89, + "learning_rate": 1.7621908943840997e-05, + "loss": 0.0, + "step": 16812 + }, + { + "epoch": 3.89, + "learning_rate": 1.7619983052153148e-05, + "loss": 0.0, + "step": 16813 + }, + { + "epoch": 3.89, + "learning_rate": 1.7618057160465295e-05, + "loss": 0.0, + "step": 16814 + }, + { + "epoch": 3.89, + "learning_rate": 1.7616131268777446e-05, + "loss": 0.0, + "step": 16815 + }, + { + "epoch": 3.89, + "learning_rate": 1.7614205377089593e-05, + "loss": 0.0004, + "step": 16816 + }, + { + "epoch": 3.89, + "learning_rate": 1.761227948540174e-05, + "loss": 0.0015, + "step": 16817 + }, + { + "epoch": 3.89, + "learning_rate": 1.7610353593713892e-05, + "loss": 0.0, + "step": 16818 + }, + { + "epoch": 3.89, + "learning_rate": 1.760842770202604e-05, + "loss": 0.0, + "step": 16819 + }, + { + "epoch": 3.89, + "learning_rate": 1.760650181033819e-05, + "loss": 0.0, + "step": 16820 + }, + { + "epoch": 3.89, + "learning_rate": 1.7604575918650338e-05, + "loss": 0.0, + "step": 16821 + }, + { + "epoch": 3.89, + "learning_rate": 1.760265002696248e-05, + "loss": 0.0, + "step": 16822 + }, + { + "epoch": 3.89, + "learning_rate": 1.7600724135274633e-05, + "loss": 0.0, + "step": 16823 + }, + { + "epoch": 3.89, + "learning_rate": 1.759879824358678e-05, + "loss": 0.0, + "step": 16824 + }, + { + "epoch": 3.89, + "learning_rate": 1.759687235189893e-05, + "loss": 0.0, + "step": 16825 + }, + { + "epoch": 3.89, + "learning_rate": 1.759494646021108e-05, + "loss": 0.0, + "step": 16826 + }, + { + "epoch": 3.89, + "learning_rate": 1.7593020568523226e-05, + "loss": 0.0, + "step": 16827 + }, + { + "epoch": 3.89, + "learning_rate": 1.7591094676835377e-05, + "loss": 0.0, + "step": 16828 + }, + { + "epoch": 3.89, + "learning_rate": 1.7589168785147524e-05, + "loss": 0.0, + "step": 16829 + }, + { + "epoch": 3.89, + "learning_rate": 1.7587242893459675e-05, + "loss": 0.0, + "step": 16830 + }, + { + "epoch": 3.89, + "learning_rate": 1.758531700177182e-05, + "loss": 0.0, + "step": 16831 + }, + { + "epoch": 3.89, + "learning_rate": 1.758339111008397e-05, + "loss": 0.0003, + "step": 16832 + }, + { + "epoch": 3.89, + "learning_rate": 1.7581465218396117e-05, + "loss": 0.0, + "step": 16833 + }, + { + "epoch": 3.89, + "learning_rate": 1.7579539326708265e-05, + "loss": 0.0, + "step": 16834 + }, + { + "epoch": 3.89, + "learning_rate": 1.7577613435020416e-05, + "loss": 0.0, + "step": 16835 + }, + { + "epoch": 3.89, + "learning_rate": 1.7575687543332563e-05, + "loss": 0.0, + "step": 16836 + }, + { + "epoch": 3.89, + "learning_rate": 1.7573761651644714e-05, + "loss": 0.0, + "step": 16837 + }, + { + "epoch": 3.89, + "learning_rate": 1.757183575995686e-05, + "loss": 0.0, + "step": 16838 + }, + { + "epoch": 3.89, + "learning_rate": 1.756990986826901e-05, + "loss": 0.0, + "step": 16839 + }, + { + "epoch": 3.89, + "learning_rate": 1.7567983976581157e-05, + "loss": 0.0, + "step": 16840 + }, + { + "epoch": 3.89, + "learning_rate": 1.7566058084893304e-05, + "loss": 0.0, + "step": 16841 + }, + { + "epoch": 3.89, + "learning_rate": 1.7564132193205455e-05, + "loss": 0.0, + "step": 16842 + }, + { + "epoch": 3.89, + "learning_rate": 1.7562206301517602e-05, + "loss": 0.0, + "step": 16843 + }, + { + "epoch": 3.89, + "learning_rate": 1.7560280409829753e-05, + "loss": 0.0, + "step": 16844 + }, + { + "epoch": 3.89, + "learning_rate": 1.75583545181419e-05, + "loss": 0.0, + "step": 16845 + }, + { + "epoch": 3.89, + "learning_rate": 1.7556428626454048e-05, + "loss": 0.0, + "step": 16846 + }, + { + "epoch": 3.89, + "learning_rate": 1.75545027347662e-05, + "loss": 0.0, + "step": 16847 + }, + { + "epoch": 3.89, + "learning_rate": 1.7552576843078347e-05, + "loss": 0.0, + "step": 16848 + }, + { + "epoch": 3.89, + "learning_rate": 1.7550650951390497e-05, + "loss": 0.0006, + "step": 16849 + }, + { + "epoch": 3.89, + "learning_rate": 1.754872505970264e-05, + "loss": 0.0, + "step": 16850 + }, + { + "epoch": 3.89, + "learning_rate": 1.7546799168014792e-05, + "loss": 0.0, + "step": 16851 + }, + { + "epoch": 3.89, + "learning_rate": 1.754487327632694e-05, + "loss": 0.0, + "step": 16852 + }, + { + "epoch": 3.89, + "learning_rate": 1.7542947384639087e-05, + "loss": 0.0, + "step": 16853 + }, + { + "epoch": 3.9, + "learning_rate": 1.7541021492951238e-05, + "loss": 0.0, + "step": 16854 + }, + { + "epoch": 3.9, + "learning_rate": 1.7539095601263386e-05, + "loss": 0.0, + "step": 16855 + }, + { + "epoch": 3.9, + "learning_rate": 1.7537169709575537e-05, + "loss": 0.0, + "step": 16856 + }, + { + "epoch": 3.9, + "learning_rate": 1.7535243817887684e-05, + "loss": 0.0, + "step": 16857 + }, + { + "epoch": 3.9, + "learning_rate": 1.753331792619983e-05, + "loss": 0.0, + "step": 16858 + }, + { + "epoch": 3.9, + "learning_rate": 1.753139203451198e-05, + "loss": 0.0, + "step": 16859 + }, + { + "epoch": 3.9, + "learning_rate": 1.7529466142824126e-05, + "loss": 0.0, + "step": 16860 + }, + { + "epoch": 3.9, + "learning_rate": 1.7527540251136277e-05, + "loss": 0.0003, + "step": 16861 + }, + { + "epoch": 3.9, + "learning_rate": 1.7525614359448425e-05, + "loss": 0.0, + "step": 16862 + }, + { + "epoch": 3.9, + "learning_rate": 1.7523688467760572e-05, + "loss": 0.0, + "step": 16863 + }, + { + "epoch": 3.9, + "learning_rate": 1.7521762576072723e-05, + "loss": 0.0, + "step": 16864 + }, + { + "epoch": 3.9, + "learning_rate": 1.751983668438487e-05, + "loss": 0.0, + "step": 16865 + }, + { + "epoch": 3.9, + "learning_rate": 1.751791079269702e-05, + "loss": 0.0, + "step": 16866 + }, + { + "epoch": 3.9, + "learning_rate": 1.751598490100917e-05, + "loss": 0.0, + "step": 16867 + }, + { + "epoch": 3.9, + "learning_rate": 1.7514059009321316e-05, + "loss": 0.0, + "step": 16868 + }, + { + "epoch": 3.9, + "learning_rate": 1.7512133117633464e-05, + "loss": 0.0, + "step": 16869 + }, + { + "epoch": 3.9, + "learning_rate": 1.751020722594561e-05, + "loss": 0.0, + "step": 16870 + }, + { + "epoch": 3.9, + "learning_rate": 1.7508281334257762e-05, + "loss": 0.0, + "step": 16871 + }, + { + "epoch": 3.9, + "learning_rate": 1.750635544256991e-05, + "loss": 0.0, + "step": 16872 + }, + { + "epoch": 3.9, + "learning_rate": 1.750442955088206e-05, + "loss": 0.0, + "step": 16873 + }, + { + "epoch": 3.9, + "learning_rate": 1.7502503659194208e-05, + "loss": 0.0, + "step": 16874 + }, + { + "epoch": 3.9, + "learning_rate": 1.7500577767506356e-05, + "loss": 0.0, + "step": 16875 + }, + { + "epoch": 3.9, + "learning_rate": 1.7498651875818506e-05, + "loss": 0.0, + "step": 16876 + }, + { + "epoch": 3.9, + "learning_rate": 1.7496725984130654e-05, + "loss": 0.0, + "step": 16877 + }, + { + "epoch": 3.9, + "learning_rate": 1.74948000924428e-05, + "loss": 0.0, + "step": 16878 + }, + { + "epoch": 3.9, + "learning_rate": 1.749287420075495e-05, + "loss": 0.0, + "step": 16879 + }, + { + "epoch": 3.9, + "learning_rate": 1.74909483090671e-05, + "loss": 0.0, + "step": 16880 + }, + { + "epoch": 3.9, + "learning_rate": 1.7489022417379247e-05, + "loss": 0.0001, + "step": 16881 + }, + { + "epoch": 3.9, + "learning_rate": 1.7487096525691395e-05, + "loss": 0.0, + "step": 16882 + }, + { + "epoch": 3.9, + "learning_rate": 1.7485170634003546e-05, + "loss": 0.0, + "step": 16883 + }, + { + "epoch": 3.9, + "learning_rate": 1.7483244742315693e-05, + "loss": 0.0, + "step": 16884 + }, + { + "epoch": 3.9, + "learning_rate": 1.7481318850627844e-05, + "loss": 0.0, + "step": 16885 + }, + { + "epoch": 3.9, + "learning_rate": 1.747939295893999e-05, + "loss": 0.0, + "step": 16886 + }, + { + "epoch": 3.9, + "learning_rate": 1.747746706725214e-05, + "loss": 0.0, + "step": 16887 + }, + { + "epoch": 3.9, + "learning_rate": 1.7475541175564286e-05, + "loss": 0.0, + "step": 16888 + }, + { + "epoch": 3.9, + "learning_rate": 1.7473615283876434e-05, + "loss": 0.0, + "step": 16889 + }, + { + "epoch": 3.9, + "learning_rate": 1.7471689392188585e-05, + "loss": 0.0, + "step": 16890 + }, + { + "epoch": 3.9, + "learning_rate": 1.7469763500500732e-05, + "loss": 0.0, + "step": 16891 + }, + { + "epoch": 3.9, + "learning_rate": 1.7467837608812883e-05, + "loss": 0.0, + "step": 16892 + }, + { + "epoch": 3.9, + "learning_rate": 1.746591171712503e-05, + "loss": 0.0, + "step": 16893 + }, + { + "epoch": 3.9, + "learning_rate": 1.7463985825437178e-05, + "loss": 0.0, + "step": 16894 + }, + { + "epoch": 3.9, + "learning_rate": 1.746205993374933e-05, + "loss": 0.0, + "step": 16895 + }, + { + "epoch": 3.9, + "learning_rate": 1.7460134042061473e-05, + "loss": 0.0, + "step": 16896 + }, + { + "epoch": 3.91, + "learning_rate": 1.7458208150373624e-05, + "loss": 0.0, + "step": 16897 + }, + { + "epoch": 3.91, + "learning_rate": 1.745628225868577e-05, + "loss": 0.0, + "step": 16898 + }, + { + "epoch": 3.91, + "learning_rate": 1.745435636699792e-05, + "loss": 0.0, + "step": 16899 + }, + { + "epoch": 3.91, + "learning_rate": 1.745243047531007e-05, + "loss": 0.0, + "step": 16900 + }, + { + "epoch": 3.91, + "learning_rate": 1.7450504583622217e-05, + "loss": 0.0001, + "step": 16901 + }, + { + "epoch": 3.91, + "learning_rate": 1.7448578691934368e-05, + "loss": 0.0, + "step": 16902 + }, + { + "epoch": 3.91, + "learning_rate": 1.7446652800246515e-05, + "loss": 0.0, + "step": 16903 + }, + { + "epoch": 3.91, + "learning_rate": 1.7444726908558663e-05, + "loss": 0.0001, + "step": 16904 + }, + { + "epoch": 3.91, + "learning_rate": 1.7442801016870814e-05, + "loss": 0.0, + "step": 16905 + }, + { + "epoch": 3.91, + "learning_rate": 1.7440875125182958e-05, + "loss": 0.0, + "step": 16906 + }, + { + "epoch": 3.91, + "learning_rate": 1.743894923349511e-05, + "loss": 0.0, + "step": 16907 + }, + { + "epoch": 3.91, + "learning_rate": 1.7437023341807256e-05, + "loss": 0.0, + "step": 16908 + }, + { + "epoch": 3.91, + "learning_rate": 1.7435097450119407e-05, + "loss": 0.0, + "step": 16909 + }, + { + "epoch": 3.91, + "learning_rate": 1.7433171558431555e-05, + "loss": 0.0, + "step": 16910 + }, + { + "epoch": 3.91, + "learning_rate": 1.7431245666743702e-05, + "loss": 0.0009, + "step": 16911 + }, + { + "epoch": 3.91, + "learning_rate": 1.7429319775055853e-05, + "loss": 0.0, + "step": 16912 + }, + { + "epoch": 3.91, + "learning_rate": 1.7427393883368e-05, + "loss": 0.0, + "step": 16913 + }, + { + "epoch": 3.91, + "learning_rate": 1.742546799168015e-05, + "loss": 0.0, + "step": 16914 + }, + { + "epoch": 3.91, + "learning_rate": 1.7423542099992295e-05, + "loss": 0.0001, + "step": 16915 + }, + { + "epoch": 3.91, + "learning_rate": 1.7421616208304446e-05, + "loss": 0.0, + "step": 16916 + }, + { + "epoch": 3.91, + "learning_rate": 1.7419690316616594e-05, + "loss": 0.0, + "step": 16917 + }, + { + "epoch": 3.91, + "learning_rate": 1.741776442492874e-05, + "loss": 0.0, + "step": 16918 + }, + { + "epoch": 3.91, + "learning_rate": 1.7415838533240892e-05, + "loss": 0.0, + "step": 16919 + }, + { + "epoch": 3.91, + "learning_rate": 1.741391264155304e-05, + "loss": 0.0, + "step": 16920 + }, + { + "epoch": 3.91, + "learning_rate": 1.741198674986519e-05, + "loss": 0.0, + "step": 16921 + }, + { + "epoch": 3.91, + "learning_rate": 1.7410060858177338e-05, + "loss": 0.0, + "step": 16922 + }, + { + "epoch": 3.91, + "learning_rate": 1.7408134966489485e-05, + "loss": 0.0003, + "step": 16923 + }, + { + "epoch": 3.91, + "learning_rate": 1.7406209074801633e-05, + "loss": 0.0, + "step": 16924 + }, + { + "epoch": 3.91, + "learning_rate": 1.740428318311378e-05, + "loss": 0.0, + "step": 16925 + }, + { + "epoch": 3.91, + "learning_rate": 1.740235729142593e-05, + "loss": 0.0, + "step": 16926 + }, + { + "epoch": 3.91, + "learning_rate": 1.740043139973808e-05, + "loss": 0.0, + "step": 16927 + }, + { + "epoch": 3.91, + "learning_rate": 1.739850550805023e-05, + "loss": 0.0, + "step": 16928 + }, + { + "epoch": 3.91, + "learning_rate": 1.7396579616362377e-05, + "loss": 0.0, + "step": 16929 + }, + { + "epoch": 3.91, + "learning_rate": 1.7394653724674524e-05, + "loss": 0.0, + "step": 16930 + }, + { + "epoch": 3.91, + "learning_rate": 1.7392727832986675e-05, + "loss": 0.0, + "step": 16931 + }, + { + "epoch": 3.91, + "learning_rate": 1.7390801941298823e-05, + "loss": 0.0, + "step": 16932 + }, + { + "epoch": 3.91, + "learning_rate": 1.7388876049610974e-05, + "loss": 0.0001, + "step": 16933 + }, + { + "epoch": 3.91, + "learning_rate": 1.7386950157923118e-05, + "loss": 0.0001, + "step": 16934 + }, + { + "epoch": 3.91, + "learning_rate": 1.7385024266235265e-05, + "loss": 0.0, + "step": 16935 + }, + { + "epoch": 3.91, + "learning_rate": 1.7383098374547416e-05, + "loss": 0.0, + "step": 16936 + }, + { + "epoch": 3.91, + "learning_rate": 1.7381172482859564e-05, + "loss": 0.0, + "step": 16937 + }, + { + "epoch": 3.91, + "learning_rate": 1.7379246591171714e-05, + "loss": 0.0, + "step": 16938 + }, + { + "epoch": 3.91, + "learning_rate": 1.7377320699483862e-05, + "loss": 0.0, + "step": 16939 + }, + { + "epoch": 3.91, + "learning_rate": 1.737539480779601e-05, + "loss": 0.0, + "step": 16940 + }, + { + "epoch": 3.92, + "learning_rate": 1.737346891610816e-05, + "loss": 0.0, + "step": 16941 + }, + { + "epoch": 3.92, + "learning_rate": 1.7371543024420308e-05, + "loss": 0.0, + "step": 16942 + }, + { + "epoch": 3.92, + "learning_rate": 1.7369617132732455e-05, + "loss": 0.0, + "step": 16943 + }, + { + "epoch": 3.92, + "learning_rate": 1.7367691241044603e-05, + "loss": 0.0, + "step": 16944 + }, + { + "epoch": 3.92, + "learning_rate": 1.7365765349356754e-05, + "loss": 0.0005, + "step": 16945 + }, + { + "epoch": 3.92, + "learning_rate": 1.73638394576689e-05, + "loss": 0.0, + "step": 16946 + }, + { + "epoch": 3.92, + "learning_rate": 1.736191356598105e-05, + "loss": 0.0, + "step": 16947 + }, + { + "epoch": 3.92, + "learning_rate": 1.73599876742932e-05, + "loss": 0.0, + "step": 16948 + }, + { + "epoch": 3.92, + "learning_rate": 1.7358061782605347e-05, + "loss": 0.0, + "step": 16949 + }, + { + "epoch": 3.92, + "learning_rate": 1.7356135890917498e-05, + "loss": 0.0, + "step": 16950 + }, + { + "epoch": 3.92, + "learning_rate": 1.7354209999229645e-05, + "loss": 0.0, + "step": 16951 + }, + { + "epoch": 3.92, + "learning_rate": 1.7352284107541793e-05, + "loss": 0.0, + "step": 16952 + }, + { + "epoch": 3.92, + "learning_rate": 1.735035821585394e-05, + "loss": 0.0, + "step": 16953 + }, + { + "epoch": 3.92, + "learning_rate": 1.7348432324166088e-05, + "loss": 0.0, + "step": 16954 + }, + { + "epoch": 3.92, + "learning_rate": 1.734650643247824e-05, + "loss": 0.0, + "step": 16955 + }, + { + "epoch": 3.92, + "learning_rate": 1.7344580540790386e-05, + "loss": 0.0, + "step": 16956 + }, + { + "epoch": 3.92, + "learning_rate": 1.7342654649102537e-05, + "loss": 0.0, + "step": 16957 + }, + { + "epoch": 3.92, + "learning_rate": 1.7340728757414684e-05, + "loss": 0.0, + "step": 16958 + }, + { + "epoch": 3.92, + "learning_rate": 1.7338802865726832e-05, + "loss": 0.0, + "step": 16959 + }, + { + "epoch": 3.92, + "learning_rate": 1.7336876974038983e-05, + "loss": 0.0, + "step": 16960 + }, + { + "epoch": 3.92, + "learning_rate": 1.733495108235113e-05, + "loss": 0.0, + "step": 16961 + }, + { + "epoch": 3.92, + "learning_rate": 1.7333025190663278e-05, + "loss": 0.0, + "step": 16962 + }, + { + "epoch": 3.92, + "learning_rate": 1.7331099298975425e-05, + "loss": 0.0, + "step": 16963 + }, + { + "epoch": 3.92, + "learning_rate": 1.7329173407287576e-05, + "loss": 0.0, + "step": 16964 + }, + { + "epoch": 3.92, + "learning_rate": 1.7327247515599723e-05, + "loss": 0.0, + "step": 16965 + }, + { + "epoch": 3.92, + "learning_rate": 1.732532162391187e-05, + "loss": 0.0, + "step": 16966 + }, + { + "epoch": 3.92, + "learning_rate": 1.7323395732224022e-05, + "loss": 0.0, + "step": 16967 + }, + { + "epoch": 3.92, + "learning_rate": 1.732146984053617e-05, + "loss": 0.0, + "step": 16968 + }, + { + "epoch": 3.92, + "learning_rate": 1.731954394884832e-05, + "loss": 0.0, + "step": 16969 + }, + { + "epoch": 3.92, + "learning_rate": 1.7317618057160468e-05, + "loss": 0.0, + "step": 16970 + }, + { + "epoch": 3.92, + "learning_rate": 1.7315692165472615e-05, + "loss": 0.0, + "step": 16971 + }, + { + "epoch": 3.92, + "learning_rate": 1.7313766273784763e-05, + "loss": 0.0, + "step": 16972 + }, + { + "epoch": 3.92, + "learning_rate": 1.731184038209691e-05, + "loss": 0.0001, + "step": 16973 + }, + { + "epoch": 3.92, + "learning_rate": 1.730991449040906e-05, + "loss": 0.0001, + "step": 16974 + }, + { + "epoch": 3.92, + "learning_rate": 1.730798859872121e-05, + "loss": 0.0, + "step": 16975 + }, + { + "epoch": 3.92, + "learning_rate": 1.7306062707033356e-05, + "loss": 0.0, + "step": 16976 + }, + { + "epoch": 3.92, + "learning_rate": 1.7304136815345507e-05, + "loss": 0.0, + "step": 16977 + }, + { + "epoch": 3.92, + "learning_rate": 1.7302210923657654e-05, + "loss": 0.0, + "step": 16978 + }, + { + "epoch": 3.92, + "learning_rate": 1.7300285031969805e-05, + "loss": 0.0, + "step": 16979 + }, + { + "epoch": 3.92, + "learning_rate": 1.729835914028195e-05, + "loss": 0.0, + "step": 16980 + }, + { + "epoch": 3.92, + "learning_rate": 1.72964332485941e-05, + "loss": 0.0, + "step": 16981 + }, + { + "epoch": 3.92, + "learning_rate": 1.7294507356906248e-05, + "loss": 0.0, + "step": 16982 + }, + { + "epoch": 3.92, + "learning_rate": 1.7292581465218395e-05, + "loss": 0.0, + "step": 16983 + }, + { + "epoch": 3.93, + "learning_rate": 1.7290655573530546e-05, + "loss": 0.0, + "step": 16984 + }, + { + "epoch": 3.93, + "learning_rate": 1.7288729681842693e-05, + "loss": 0.0, + "step": 16985 + }, + { + "epoch": 3.93, + "learning_rate": 1.7286803790154844e-05, + "loss": 0.0, + "step": 16986 + }, + { + "epoch": 3.93, + "learning_rate": 1.728487789846699e-05, + "loss": 0.0, + "step": 16987 + }, + { + "epoch": 3.93, + "learning_rate": 1.728295200677914e-05, + "loss": 0.0, + "step": 16988 + }, + { + "epoch": 3.93, + "learning_rate": 1.728102611509129e-05, + "loss": 0.0, + "step": 16989 + }, + { + "epoch": 3.93, + "learning_rate": 1.7279100223403434e-05, + "loss": 0.0, + "step": 16990 + }, + { + "epoch": 3.93, + "learning_rate": 1.7277174331715585e-05, + "loss": 0.0, + "step": 16991 + }, + { + "epoch": 3.93, + "learning_rate": 1.7275248440027732e-05, + "loss": 0.0, + "step": 16992 + }, + { + "epoch": 3.93, + "learning_rate": 1.7273322548339883e-05, + "loss": 0.0, + "step": 16993 + }, + { + "epoch": 3.93, + "learning_rate": 1.727139665665203e-05, + "loss": 0.0, + "step": 16994 + }, + { + "epoch": 3.93, + "learning_rate": 1.7269470764964178e-05, + "loss": 0.0, + "step": 16995 + }, + { + "epoch": 3.93, + "learning_rate": 1.726754487327633e-05, + "loss": 0.0, + "step": 16996 + }, + { + "epoch": 3.93, + "learning_rate": 1.7265618981588477e-05, + "loss": 0.0, + "step": 16997 + }, + { + "epoch": 3.93, + "learning_rate": 1.7263693089900627e-05, + "loss": 0.0, + "step": 16998 + }, + { + "epoch": 3.93, + "learning_rate": 1.726176719821277e-05, + "loss": 0.0003, + "step": 16999 + }, + { + "epoch": 3.93, + "learning_rate": 1.7259841306524922e-05, + "loss": 0.0001, + "step": 17000 + }, + { + "epoch": 3.93, + "learning_rate": 1.725791541483707e-05, + "loss": 0.0, + "step": 17001 + }, + { + "epoch": 3.93, + "learning_rate": 1.7255989523149217e-05, + "loss": 0.0, + "step": 17002 + }, + { + "epoch": 3.93, + "learning_rate": 1.7254063631461368e-05, + "loss": 0.0, + "step": 17003 + }, + { + "epoch": 3.93, + "learning_rate": 1.7252137739773516e-05, + "loss": 0.0, + "step": 17004 + }, + { + "epoch": 3.93, + "learning_rate": 1.7250211848085667e-05, + "loss": 0.0, + "step": 17005 + }, + { + "epoch": 3.93, + "learning_rate": 1.7248285956397814e-05, + "loss": 0.0, + "step": 17006 + }, + { + "epoch": 3.93, + "learning_rate": 1.724636006470996e-05, + "loss": 0.0, + "step": 17007 + }, + { + "epoch": 3.93, + "learning_rate": 1.724443417302211e-05, + "loss": 0.0, + "step": 17008 + }, + { + "epoch": 3.93, + "learning_rate": 1.7242508281334257e-05, + "loss": 0.0, + "step": 17009 + }, + { + "epoch": 3.93, + "learning_rate": 1.7240582389646407e-05, + "loss": 0.0, + "step": 17010 + }, + { + "epoch": 3.93, + "learning_rate": 1.7238656497958555e-05, + "loss": 0.0, + "step": 17011 + }, + { + "epoch": 3.93, + "learning_rate": 1.7236730606270702e-05, + "loss": 0.0, + "step": 17012 + }, + { + "epoch": 3.93, + "learning_rate": 1.7234804714582853e-05, + "loss": 0.0, + "step": 17013 + }, + { + "epoch": 3.93, + "learning_rate": 1.7232878822895e-05, + "loss": 0.0, + "step": 17014 + }, + { + "epoch": 3.93, + "learning_rate": 1.723095293120715e-05, + "loss": 0.0, + "step": 17015 + }, + { + "epoch": 3.93, + "learning_rate": 1.72290270395193e-05, + "loss": 0.0004, + "step": 17016 + }, + { + "epoch": 3.93, + "learning_rate": 1.7227101147831447e-05, + "loss": 0.0, + "step": 17017 + }, + { + "epoch": 3.93, + "learning_rate": 1.7225175256143594e-05, + "loss": 0.0, + "step": 17018 + }, + { + "epoch": 3.93, + "learning_rate": 1.722324936445574e-05, + "loss": 0.0, + "step": 17019 + }, + { + "epoch": 3.93, + "learning_rate": 1.7221323472767892e-05, + "loss": 0.0, + "step": 17020 + }, + { + "epoch": 3.93, + "learning_rate": 1.721939758108004e-05, + "loss": 0.0, + "step": 17021 + }, + { + "epoch": 3.93, + "learning_rate": 1.721747168939219e-05, + "loss": 0.0, + "step": 17022 + }, + { + "epoch": 3.93, + "learning_rate": 1.7215545797704338e-05, + "loss": 0.0, + "step": 17023 + }, + { + "epoch": 3.93, + "learning_rate": 1.7213619906016486e-05, + "loss": 0.0, + "step": 17024 + }, + { + "epoch": 3.93, + "learning_rate": 1.7211694014328636e-05, + "loss": 0.0, + "step": 17025 + }, + { + "epoch": 3.93, + "learning_rate": 1.7209768122640784e-05, + "loss": 0.0, + "step": 17026 + }, + { + "epoch": 3.94, + "learning_rate": 1.720784223095293e-05, + "loss": 0.0, + "step": 17027 + }, + { + "epoch": 3.94, + "learning_rate": 1.720591633926508e-05, + "loss": 0.0001, + "step": 17028 + }, + { + "epoch": 3.94, + "learning_rate": 1.720399044757723e-05, + "loss": 0.0, + "step": 17029 + }, + { + "epoch": 3.94, + "learning_rate": 1.7202064555889377e-05, + "loss": 0.0, + "step": 17030 + }, + { + "epoch": 3.94, + "learning_rate": 1.7200138664201525e-05, + "loss": 0.0, + "step": 17031 + }, + { + "epoch": 3.94, + "learning_rate": 1.7198212772513676e-05, + "loss": 0.0, + "step": 17032 + }, + { + "epoch": 3.94, + "learning_rate": 1.7196286880825823e-05, + "loss": 0.0, + "step": 17033 + }, + { + "epoch": 3.94, + "learning_rate": 1.7194360989137974e-05, + "loss": 0.0, + "step": 17034 + }, + { + "epoch": 3.94, + "learning_rate": 1.719243509745012e-05, + "loss": 0.0, + "step": 17035 + }, + { + "epoch": 3.94, + "learning_rate": 1.719050920576227e-05, + "loss": 0.0, + "step": 17036 + }, + { + "epoch": 3.94, + "learning_rate": 1.7188583314074416e-05, + "loss": 0.0, + "step": 17037 + }, + { + "epoch": 3.94, + "learning_rate": 1.7186657422386564e-05, + "loss": 0.0, + "step": 17038 + }, + { + "epoch": 3.94, + "learning_rate": 1.7184731530698715e-05, + "loss": 0.0002, + "step": 17039 + }, + { + "epoch": 3.94, + "learning_rate": 1.7182805639010862e-05, + "loss": 0.0, + "step": 17040 + }, + { + "epoch": 3.94, + "learning_rate": 1.7180879747323013e-05, + "loss": 0.0, + "step": 17041 + }, + { + "epoch": 3.94, + "learning_rate": 1.717895385563516e-05, + "loss": 0.0002, + "step": 17042 + }, + { + "epoch": 3.94, + "learning_rate": 1.7177027963947308e-05, + "loss": 0.0, + "step": 17043 + }, + { + "epoch": 3.94, + "learning_rate": 1.717510207225946e-05, + "loss": 0.0, + "step": 17044 + }, + { + "epoch": 3.94, + "learning_rate": 1.7173176180571606e-05, + "loss": 0.0, + "step": 17045 + }, + { + "epoch": 3.94, + "learning_rate": 1.7171250288883754e-05, + "loss": 0.0, + "step": 17046 + }, + { + "epoch": 3.94, + "learning_rate": 1.71693243971959e-05, + "loss": 0.0, + "step": 17047 + }, + { + "epoch": 3.94, + "learning_rate": 1.7167398505508052e-05, + "loss": 0.0, + "step": 17048 + }, + { + "epoch": 3.94, + "learning_rate": 1.71654726138202e-05, + "loss": 0.0, + "step": 17049 + }, + { + "epoch": 3.94, + "learning_rate": 1.7163546722132347e-05, + "loss": 0.0, + "step": 17050 + }, + { + "epoch": 3.94, + "learning_rate": 1.7161620830444498e-05, + "loss": 0.0, + "step": 17051 + }, + { + "epoch": 3.94, + "learning_rate": 1.7159694938756645e-05, + "loss": 0.0002, + "step": 17052 + }, + { + "epoch": 3.94, + "learning_rate": 1.7157769047068793e-05, + "loss": 0.0, + "step": 17053 + }, + { + "epoch": 3.94, + "learning_rate": 1.7155843155380944e-05, + "loss": 0.0, + "step": 17054 + }, + { + "epoch": 3.94, + "learning_rate": 1.7153917263693088e-05, + "loss": 0.0, + "step": 17055 + }, + { + "epoch": 3.94, + "learning_rate": 1.715199137200524e-05, + "loss": 0.0, + "step": 17056 + }, + { + "epoch": 3.94, + "learning_rate": 1.7150065480317386e-05, + "loss": 0.0, + "step": 17057 + }, + { + "epoch": 3.94, + "learning_rate": 1.7148139588629537e-05, + "loss": 0.0, + "step": 17058 + }, + { + "epoch": 3.94, + "learning_rate": 1.7146213696941685e-05, + "loss": 0.0, + "step": 17059 + }, + { + "epoch": 3.94, + "learning_rate": 1.7144287805253832e-05, + "loss": 0.0001, + "step": 17060 + }, + { + "epoch": 3.94, + "learning_rate": 1.7142361913565983e-05, + "loss": 0.0, + "step": 17061 + }, + { + "epoch": 3.94, + "learning_rate": 1.714043602187813e-05, + "loss": 0.0, + "step": 17062 + }, + { + "epoch": 3.94, + "learning_rate": 1.713851013019028e-05, + "loss": 0.0, + "step": 17063 + }, + { + "epoch": 3.94, + "learning_rate": 1.7136584238502425e-05, + "loss": 0.0, + "step": 17064 + }, + { + "epoch": 3.94, + "learning_rate": 1.7134658346814576e-05, + "loss": 0.0, + "step": 17065 + }, + { + "epoch": 3.94, + "learning_rate": 1.7132732455126724e-05, + "loss": 0.0, + "step": 17066 + }, + { + "epoch": 3.94, + "learning_rate": 1.713080656343887e-05, + "loss": 0.0, + "step": 17067 + }, + { + "epoch": 3.94, + "learning_rate": 1.7128880671751022e-05, + "loss": 0.0, + "step": 17068 + }, + { + "epoch": 3.94, + "learning_rate": 1.712695478006317e-05, + "loss": 0.0001, + "step": 17069 + }, + { + "epoch": 3.94, + "learning_rate": 1.712502888837532e-05, + "loss": 0.0, + "step": 17070 + }, + { + "epoch": 3.95, + "learning_rate": 1.7123102996687468e-05, + "loss": 0.0, + "step": 17071 + }, + { + "epoch": 3.95, + "learning_rate": 1.7121177104999615e-05, + "loss": 0.0, + "step": 17072 + }, + { + "epoch": 3.95, + "learning_rate": 1.7119251213311766e-05, + "loss": 0.0, + "step": 17073 + }, + { + "epoch": 3.95, + "learning_rate": 1.711732532162391e-05, + "loss": 0.0, + "step": 17074 + }, + { + "epoch": 3.95, + "learning_rate": 1.711539942993606e-05, + "loss": 0.0, + "step": 17075 + }, + { + "epoch": 3.95, + "learning_rate": 1.711347353824821e-05, + "loss": 0.0, + "step": 17076 + }, + { + "epoch": 3.95, + "learning_rate": 1.711154764656036e-05, + "loss": 0.0, + "step": 17077 + }, + { + "epoch": 3.95, + "learning_rate": 1.7109621754872507e-05, + "loss": 0.0, + "step": 17078 + }, + { + "epoch": 3.95, + "learning_rate": 1.7107695863184654e-05, + "loss": 0.0, + "step": 17079 + }, + { + "epoch": 3.95, + "learning_rate": 1.7105769971496805e-05, + "loss": 0.0, + "step": 17080 + }, + { + "epoch": 3.95, + "learning_rate": 1.7103844079808953e-05, + "loss": 0.0, + "step": 17081 + }, + { + "epoch": 3.95, + "learning_rate": 1.7101918188121104e-05, + "loss": 0.0, + "step": 17082 + }, + { + "epoch": 3.95, + "learning_rate": 1.7099992296433248e-05, + "loss": 0.0, + "step": 17083 + }, + { + "epoch": 3.95, + "learning_rate": 1.70980664047454e-05, + "loss": 0.0, + "step": 17084 + }, + { + "epoch": 3.95, + "learning_rate": 1.7096140513057546e-05, + "loss": 0.0, + "step": 17085 + }, + { + "epoch": 3.95, + "learning_rate": 1.7094214621369694e-05, + "loss": 0.0, + "step": 17086 + }, + { + "epoch": 3.95, + "learning_rate": 1.7092288729681844e-05, + "loss": 0.0, + "step": 17087 + }, + { + "epoch": 3.95, + "learning_rate": 1.7090362837993992e-05, + "loss": 0.0, + "step": 17088 + }, + { + "epoch": 3.95, + "learning_rate": 1.708843694630614e-05, + "loss": 0.0, + "step": 17089 + }, + { + "epoch": 3.95, + "learning_rate": 1.708651105461829e-05, + "loss": 0.0, + "step": 17090 + }, + { + "epoch": 3.95, + "learning_rate": 1.7084585162930438e-05, + "loss": 0.0, + "step": 17091 + }, + { + "epoch": 3.95, + "learning_rate": 1.7082659271242585e-05, + "loss": 0.0, + "step": 17092 + }, + { + "epoch": 3.95, + "learning_rate": 1.7080733379554733e-05, + "loss": 0.0, + "step": 17093 + }, + { + "epoch": 3.95, + "learning_rate": 1.7078807487866884e-05, + "loss": 0.0, + "step": 17094 + }, + { + "epoch": 3.95, + "learning_rate": 1.707688159617903e-05, + "loss": 0.0, + "step": 17095 + }, + { + "epoch": 3.95, + "learning_rate": 1.707495570449118e-05, + "loss": 0.0, + "step": 17096 + }, + { + "epoch": 3.95, + "learning_rate": 1.707302981280333e-05, + "loss": 0.0, + "step": 17097 + }, + { + "epoch": 3.95, + "learning_rate": 1.7071103921115477e-05, + "loss": 0.0, + "step": 17098 + }, + { + "epoch": 3.95, + "learning_rate": 1.7069178029427628e-05, + "loss": 0.0, + "step": 17099 + }, + { + "epoch": 3.95, + "learning_rate": 1.7067252137739775e-05, + "loss": 0.0001, + "step": 17100 + }, + { + "epoch": 3.95, + "learning_rate": 1.7065326246051923e-05, + "loss": 0.0, + "step": 17101 + }, + { + "epoch": 3.95, + "learning_rate": 1.706340035436407e-05, + "loss": 0.0, + "step": 17102 + }, + { + "epoch": 3.95, + "learning_rate": 1.7061474462676218e-05, + "loss": 0.0, + "step": 17103 + }, + { + "epoch": 3.95, + "learning_rate": 1.705954857098837e-05, + "loss": 0.0, + "step": 17104 + }, + { + "epoch": 3.95, + "learning_rate": 1.7057622679300516e-05, + "loss": 0.0, + "step": 17105 + }, + { + "epoch": 3.95, + "learning_rate": 1.7055696787612667e-05, + "loss": 0.0, + "step": 17106 + }, + { + "epoch": 3.95, + "learning_rate": 1.7053770895924814e-05, + "loss": 0.0, + "step": 17107 + }, + { + "epoch": 3.95, + "learning_rate": 1.7051845004236962e-05, + "loss": 0.0, + "step": 17108 + }, + { + "epoch": 3.95, + "learning_rate": 1.7049919112549113e-05, + "loss": 0.0001, + "step": 17109 + }, + { + "epoch": 3.95, + "learning_rate": 1.704799322086126e-05, + "loss": 0.0, + "step": 17110 + }, + { + "epoch": 3.95, + "learning_rate": 1.7046067329173408e-05, + "loss": 0.0, + "step": 17111 + }, + { + "epoch": 3.95, + "learning_rate": 1.7044141437485555e-05, + "loss": 0.0, + "step": 17112 + }, + { + "epoch": 3.95, + "learning_rate": 1.7042215545797706e-05, + "loss": 0.0, + "step": 17113 + }, + { + "epoch": 3.96, + "learning_rate": 1.7040289654109853e-05, + "loss": 0.0001, + "step": 17114 + }, + { + "epoch": 3.96, + "learning_rate": 1.7038363762422e-05, + "loss": 0.0, + "step": 17115 + }, + { + "epoch": 3.96, + "learning_rate": 1.7036437870734152e-05, + "loss": 0.0, + "step": 17116 + }, + { + "epoch": 3.96, + "learning_rate": 1.70345119790463e-05, + "loss": 0.0, + "step": 17117 + }, + { + "epoch": 3.96, + "learning_rate": 1.703258608735845e-05, + "loss": 0.0, + "step": 17118 + }, + { + "epoch": 3.96, + "learning_rate": 1.7030660195670598e-05, + "loss": 0.0, + "step": 17119 + }, + { + "epoch": 3.96, + "learning_rate": 1.7028734303982745e-05, + "loss": 0.0001, + "step": 17120 + }, + { + "epoch": 3.96, + "learning_rate": 1.7026808412294893e-05, + "loss": 0.0, + "step": 17121 + }, + { + "epoch": 3.96, + "learning_rate": 1.702488252060704e-05, + "loss": 0.0001, + "step": 17122 + }, + { + "epoch": 3.96, + "learning_rate": 1.702295662891919e-05, + "loss": 0.0, + "step": 17123 + }, + { + "epoch": 3.96, + "learning_rate": 1.702103073723134e-05, + "loss": 0.0, + "step": 17124 + }, + { + "epoch": 3.96, + "learning_rate": 1.701910484554349e-05, + "loss": 0.0, + "step": 17125 + }, + { + "epoch": 3.96, + "learning_rate": 1.7017178953855637e-05, + "loss": 0.0003, + "step": 17126 + }, + { + "epoch": 3.96, + "learning_rate": 1.7015253062167784e-05, + "loss": 0.0, + "step": 17127 + }, + { + "epoch": 3.96, + "learning_rate": 1.7013327170479935e-05, + "loss": 0.0, + "step": 17128 + }, + { + "epoch": 3.96, + "learning_rate": 1.7011401278792083e-05, + "loss": 0.0, + "step": 17129 + }, + { + "epoch": 3.96, + "learning_rate": 1.700947538710423e-05, + "loss": 0.0, + "step": 17130 + }, + { + "epoch": 3.96, + "learning_rate": 1.7007549495416378e-05, + "loss": 0.0, + "step": 17131 + }, + { + "epoch": 3.96, + "learning_rate": 1.7005623603728525e-05, + "loss": 0.0, + "step": 17132 + }, + { + "epoch": 3.96, + "learning_rate": 1.7003697712040676e-05, + "loss": 0.0, + "step": 17133 + }, + { + "epoch": 3.96, + "learning_rate": 1.7001771820352823e-05, + "loss": 0.0, + "step": 17134 + }, + { + "epoch": 3.96, + "learning_rate": 1.6999845928664974e-05, + "loss": 0.0, + "step": 17135 + }, + { + "epoch": 3.96, + "learning_rate": 1.6997920036977122e-05, + "loss": 0.0, + "step": 17136 + }, + { + "epoch": 3.96, + "learning_rate": 1.699599414528927e-05, + "loss": 0.0, + "step": 17137 + }, + { + "epoch": 3.96, + "learning_rate": 1.699406825360142e-05, + "loss": 0.0, + "step": 17138 + }, + { + "epoch": 3.96, + "learning_rate": 1.6992142361913564e-05, + "loss": 0.0, + "step": 17139 + }, + { + "epoch": 3.96, + "learning_rate": 1.6990216470225715e-05, + "loss": 0.0, + "step": 17140 + }, + { + "epoch": 3.96, + "learning_rate": 1.6988290578537862e-05, + "loss": 0.0, + "step": 17141 + }, + { + "epoch": 3.96, + "learning_rate": 1.6986364686850013e-05, + "loss": 0.0, + "step": 17142 + }, + { + "epoch": 3.96, + "learning_rate": 1.698443879516216e-05, + "loss": 0.0, + "step": 17143 + }, + { + "epoch": 3.96, + "learning_rate": 1.698251290347431e-05, + "loss": 0.0, + "step": 17144 + }, + { + "epoch": 3.96, + "learning_rate": 1.698058701178646e-05, + "loss": 0.0, + "step": 17145 + }, + { + "epoch": 3.96, + "learning_rate": 1.6978661120098607e-05, + "loss": 0.0, + "step": 17146 + }, + { + "epoch": 3.96, + "learning_rate": 1.6976735228410758e-05, + "loss": 0.0, + "step": 17147 + }, + { + "epoch": 3.96, + "learning_rate": 1.69748093367229e-05, + "loss": 0.0, + "step": 17148 + }, + { + "epoch": 3.96, + "learning_rate": 1.6972883445035052e-05, + "loss": 0.0, + "step": 17149 + }, + { + "epoch": 3.96, + "learning_rate": 1.69709575533472e-05, + "loss": 0.0, + "step": 17150 + }, + { + "epoch": 3.96, + "learning_rate": 1.6969031661659347e-05, + "loss": 0.0, + "step": 17151 + }, + { + "epoch": 3.96, + "learning_rate": 1.6967105769971498e-05, + "loss": 0.0, + "step": 17152 + }, + { + "epoch": 3.96, + "learning_rate": 1.6965179878283646e-05, + "loss": 0.0, + "step": 17153 + }, + { + "epoch": 3.96, + "learning_rate": 1.6963253986595797e-05, + "loss": 0.0, + "step": 17154 + }, + { + "epoch": 3.96, + "learning_rate": 1.6961328094907944e-05, + "loss": 0.0, + "step": 17155 + }, + { + "epoch": 3.96, + "learning_rate": 1.695940220322009e-05, + "loss": 0.0, + "step": 17156 + }, + { + "epoch": 3.97, + "learning_rate": 1.6957476311532242e-05, + "loss": 0.0, + "step": 17157 + }, + { + "epoch": 3.97, + "learning_rate": 1.6955550419844387e-05, + "loss": 0.0003, + "step": 17158 + }, + { + "epoch": 3.97, + "learning_rate": 1.6953624528156537e-05, + "loss": 0.0, + "step": 17159 + }, + { + "epoch": 3.97, + "learning_rate": 1.6951698636468685e-05, + "loss": 0.0, + "step": 17160 + }, + { + "epoch": 3.97, + "learning_rate": 1.6949772744780836e-05, + "loss": 0.0, + "step": 17161 + }, + { + "epoch": 3.97, + "learning_rate": 1.6947846853092983e-05, + "loss": 0.0, + "step": 17162 + }, + { + "epoch": 3.97, + "learning_rate": 1.694592096140513e-05, + "loss": 0.0, + "step": 17163 + }, + { + "epoch": 3.97, + "learning_rate": 1.694399506971728e-05, + "loss": 0.0, + "step": 17164 + }, + { + "epoch": 3.97, + "learning_rate": 1.694206917802943e-05, + "loss": 0.0, + "step": 17165 + }, + { + "epoch": 3.97, + "learning_rate": 1.6940143286341577e-05, + "loss": 0.0, + "step": 17166 + }, + { + "epoch": 3.97, + "learning_rate": 1.6938217394653724e-05, + "loss": 0.0, + "step": 17167 + }, + { + "epoch": 3.97, + "learning_rate": 1.693629150296587e-05, + "loss": 0.0, + "step": 17168 + }, + { + "epoch": 3.97, + "learning_rate": 1.6934365611278022e-05, + "loss": 0.0, + "step": 17169 + }, + { + "epoch": 3.97, + "learning_rate": 1.693243971959017e-05, + "loss": 0.0, + "step": 17170 + }, + { + "epoch": 3.97, + "learning_rate": 1.693051382790232e-05, + "loss": 0.0001, + "step": 17171 + }, + { + "epoch": 3.97, + "learning_rate": 1.6928587936214468e-05, + "loss": 0.0, + "step": 17172 + }, + { + "epoch": 3.97, + "learning_rate": 1.6926662044526616e-05, + "loss": 0.0, + "step": 17173 + }, + { + "epoch": 3.97, + "learning_rate": 1.6924736152838767e-05, + "loss": 0.0, + "step": 17174 + }, + { + "epoch": 3.97, + "learning_rate": 1.6922810261150914e-05, + "loss": 0.0, + "step": 17175 + }, + { + "epoch": 3.97, + "learning_rate": 1.692088436946306e-05, + "loss": 0.0, + "step": 17176 + }, + { + "epoch": 3.97, + "learning_rate": 1.691895847777521e-05, + "loss": 0.0, + "step": 17177 + }, + { + "epoch": 3.97, + "learning_rate": 1.691703258608736e-05, + "loss": 0.0, + "step": 17178 + }, + { + "epoch": 3.97, + "learning_rate": 1.6915106694399507e-05, + "loss": 0.0, + "step": 17179 + }, + { + "epoch": 3.97, + "learning_rate": 1.6913180802711655e-05, + "loss": 0.0, + "step": 17180 + }, + { + "epoch": 3.97, + "learning_rate": 1.6911254911023806e-05, + "loss": 0.0, + "step": 17181 + }, + { + "epoch": 3.97, + "learning_rate": 1.6909329019335953e-05, + "loss": 0.0, + "step": 17182 + }, + { + "epoch": 3.97, + "learning_rate": 1.6907403127648104e-05, + "loss": 0.0004, + "step": 17183 + }, + { + "epoch": 3.97, + "learning_rate": 1.690547723596025e-05, + "loss": 0.0, + "step": 17184 + }, + { + "epoch": 3.97, + "learning_rate": 1.69035513442724e-05, + "loss": 0.0, + "step": 17185 + }, + { + "epoch": 3.97, + "learning_rate": 1.6901625452584546e-05, + "loss": 0.0, + "step": 17186 + }, + { + "epoch": 3.97, + "learning_rate": 1.6899699560896694e-05, + "loss": 0.0, + "step": 17187 + }, + { + "epoch": 3.97, + "learning_rate": 1.6897773669208845e-05, + "loss": 0.0, + "step": 17188 + }, + { + "epoch": 3.97, + "learning_rate": 1.6895847777520992e-05, + "loss": 0.0, + "step": 17189 + }, + { + "epoch": 3.97, + "learning_rate": 1.6893921885833143e-05, + "loss": 0.0, + "step": 17190 + }, + { + "epoch": 3.97, + "learning_rate": 1.689199599414529e-05, + "loss": 0.0, + "step": 17191 + }, + { + "epoch": 3.97, + "learning_rate": 1.6890070102457438e-05, + "loss": 0.0, + "step": 17192 + }, + { + "epoch": 3.97, + "learning_rate": 1.688814421076959e-05, + "loss": 0.0, + "step": 17193 + }, + { + "epoch": 3.97, + "learning_rate": 1.6886218319081736e-05, + "loss": 0.0, + "step": 17194 + }, + { + "epoch": 3.97, + "learning_rate": 1.6884292427393884e-05, + "loss": 0.0, + "step": 17195 + }, + { + "epoch": 3.97, + "learning_rate": 1.688236653570603e-05, + "loss": 0.0, + "step": 17196 + }, + { + "epoch": 3.97, + "learning_rate": 1.6880440644018182e-05, + "loss": 0.0, + "step": 17197 + }, + { + "epoch": 3.97, + "learning_rate": 1.687851475233033e-05, + "loss": 0.0, + "step": 17198 + }, + { + "epoch": 3.97, + "learning_rate": 1.6876588860642477e-05, + "loss": 0.0, + "step": 17199 + }, + { + "epoch": 3.98, + "learning_rate": 1.6874662968954628e-05, + "loss": 0.0, + "step": 17200 + }, + { + "epoch": 3.98, + "learning_rate": 1.6872737077266776e-05, + "loss": 0.0, + "step": 17201 + }, + { + "epoch": 3.98, + "learning_rate": 1.6870811185578923e-05, + "loss": 0.0, + "step": 17202 + }, + { + "epoch": 3.98, + "learning_rate": 1.6868885293891074e-05, + "loss": 0.0, + "step": 17203 + }, + { + "epoch": 3.98, + "learning_rate": 1.6866959402203218e-05, + "loss": 0.0, + "step": 17204 + }, + { + "epoch": 3.98, + "learning_rate": 1.686503351051537e-05, + "loss": 0.0, + "step": 17205 + }, + { + "epoch": 3.98, + "learning_rate": 1.6863107618827516e-05, + "loss": 0.0, + "step": 17206 + }, + { + "epoch": 3.98, + "learning_rate": 1.6861181727139667e-05, + "loss": 0.0, + "step": 17207 + }, + { + "epoch": 3.98, + "learning_rate": 1.6859255835451815e-05, + "loss": 0.0, + "step": 17208 + }, + { + "epoch": 3.98, + "learning_rate": 1.6857329943763962e-05, + "loss": 0.0, + "step": 17209 + }, + { + "epoch": 3.98, + "learning_rate": 1.6855404052076113e-05, + "loss": 0.0, + "step": 17210 + }, + { + "epoch": 3.98, + "learning_rate": 1.685347816038826e-05, + "loss": 0.0, + "step": 17211 + }, + { + "epoch": 3.98, + "learning_rate": 1.685155226870041e-05, + "loss": 0.0006, + "step": 17212 + }, + { + "epoch": 3.98, + "learning_rate": 1.684962637701256e-05, + "loss": 0.0001, + "step": 17213 + }, + { + "epoch": 3.98, + "learning_rate": 1.6847700485324706e-05, + "loss": 0.0, + "step": 17214 + }, + { + "epoch": 3.98, + "learning_rate": 1.6845774593636854e-05, + "loss": 0.0, + "step": 17215 + }, + { + "epoch": 3.98, + "learning_rate": 1.6843848701949e-05, + "loss": 0.0, + "step": 17216 + }, + { + "epoch": 3.98, + "learning_rate": 1.6841922810261152e-05, + "loss": 0.0, + "step": 17217 + }, + { + "epoch": 3.98, + "learning_rate": 1.68399969185733e-05, + "loss": 0.0, + "step": 17218 + }, + { + "epoch": 3.98, + "learning_rate": 1.683807102688545e-05, + "loss": 0.0, + "step": 17219 + }, + { + "epoch": 3.98, + "learning_rate": 1.6836145135197598e-05, + "loss": 0.0, + "step": 17220 + }, + { + "epoch": 3.98, + "learning_rate": 1.6834219243509745e-05, + "loss": 0.0, + "step": 17221 + }, + { + "epoch": 3.98, + "learning_rate": 1.6832293351821896e-05, + "loss": 0.0, + "step": 17222 + }, + { + "epoch": 3.98, + "learning_rate": 1.683036746013404e-05, + "loss": 0.0, + "step": 17223 + }, + { + "epoch": 3.98, + "learning_rate": 1.682844156844619e-05, + "loss": 0.0, + "step": 17224 + }, + { + "epoch": 3.98, + "learning_rate": 1.682651567675834e-05, + "loss": 0.0, + "step": 17225 + }, + { + "epoch": 3.98, + "learning_rate": 1.682458978507049e-05, + "loss": 0.0, + "step": 17226 + }, + { + "epoch": 3.98, + "learning_rate": 1.6822663893382637e-05, + "loss": 0.0, + "step": 17227 + }, + { + "epoch": 3.98, + "learning_rate": 1.6820738001694785e-05, + "loss": 0.0, + "step": 17228 + }, + { + "epoch": 3.98, + "learning_rate": 1.6818812110006935e-05, + "loss": 0.0, + "step": 17229 + }, + { + "epoch": 3.98, + "learning_rate": 1.6816886218319083e-05, + "loss": 0.0, + "step": 17230 + }, + { + "epoch": 3.98, + "learning_rate": 1.6814960326631234e-05, + "loss": 0.0, + "step": 17231 + }, + { + "epoch": 3.98, + "learning_rate": 1.6813034434943378e-05, + "loss": 0.0, + "step": 17232 + }, + { + "epoch": 3.98, + "learning_rate": 1.681110854325553e-05, + "loss": 0.0, + "step": 17233 + }, + { + "epoch": 3.98, + "learning_rate": 1.6809182651567676e-05, + "loss": 0.0001, + "step": 17234 + }, + { + "epoch": 3.98, + "learning_rate": 1.6807256759879824e-05, + "loss": 0.0, + "step": 17235 + }, + { + "epoch": 3.98, + "learning_rate": 1.6805330868191975e-05, + "loss": 0.0, + "step": 17236 + }, + { + "epoch": 3.98, + "learning_rate": 1.6803404976504122e-05, + "loss": 0.0, + "step": 17237 + }, + { + "epoch": 3.98, + "learning_rate": 1.6801479084816273e-05, + "loss": 0.0, + "step": 17238 + }, + { + "epoch": 3.98, + "learning_rate": 1.679955319312842e-05, + "loss": 0.0, + "step": 17239 + }, + { + "epoch": 3.98, + "learning_rate": 1.6797627301440568e-05, + "loss": 0.0001, + "step": 17240 + }, + { + "epoch": 3.98, + "learning_rate": 1.679570140975272e-05, + "loss": 0.0, + "step": 17241 + }, + { + "epoch": 3.98, + "learning_rate": 1.6793775518064863e-05, + "loss": 0.0, + "step": 17242 + }, + { + "epoch": 3.98, + "learning_rate": 1.6791849626377014e-05, + "loss": 0.0, + "step": 17243 + }, + { + "epoch": 3.99, + "learning_rate": 1.678992373468916e-05, + "loss": 0.0, + "step": 17244 + }, + { + "epoch": 3.99, + "learning_rate": 1.678799784300131e-05, + "loss": 0.0, + "step": 17245 + }, + { + "epoch": 3.99, + "learning_rate": 1.678607195131346e-05, + "loss": 0.0, + "step": 17246 + }, + { + "epoch": 3.99, + "learning_rate": 1.6784146059625607e-05, + "loss": 0.0002, + "step": 17247 + }, + { + "epoch": 3.99, + "learning_rate": 1.6782220167937758e-05, + "loss": 0.0, + "step": 17248 + }, + { + "epoch": 3.99, + "learning_rate": 1.6780294276249905e-05, + "loss": 0.0, + "step": 17249 + }, + { + "epoch": 3.99, + "learning_rate": 1.6778368384562053e-05, + "loss": 0.0, + "step": 17250 + }, + { + "epoch": 3.99, + "learning_rate": 1.67764424928742e-05, + "loss": 0.0, + "step": 17251 + }, + { + "epoch": 3.99, + "learning_rate": 1.6774516601186348e-05, + "loss": 0.0, + "step": 17252 + }, + { + "epoch": 3.99, + "learning_rate": 1.67725907094985e-05, + "loss": 0.0, + "step": 17253 + }, + { + "epoch": 3.99, + "learning_rate": 1.6770664817810646e-05, + "loss": 0.0, + "step": 17254 + }, + { + "epoch": 3.99, + "learning_rate": 1.6768738926122797e-05, + "loss": 0.0, + "step": 17255 + }, + { + "epoch": 3.99, + "learning_rate": 1.6766813034434944e-05, + "loss": 0.0003, + "step": 17256 + }, + { + "epoch": 3.99, + "learning_rate": 1.6764887142747092e-05, + "loss": 0.0016, + "step": 17257 + }, + { + "epoch": 3.99, + "learning_rate": 1.6762961251059243e-05, + "loss": 0.0, + "step": 17258 + }, + { + "epoch": 3.99, + "learning_rate": 1.676103535937139e-05, + "loss": 0.0, + "step": 17259 + }, + { + "epoch": 3.99, + "learning_rate": 1.6759109467683538e-05, + "loss": 0.0, + "step": 17260 + }, + { + "epoch": 3.99, + "learning_rate": 1.6757183575995685e-05, + "loss": 0.0, + "step": 17261 + }, + { + "epoch": 3.99, + "learning_rate": 1.6755257684307836e-05, + "loss": 0.0, + "step": 17262 + }, + { + "epoch": 3.99, + "learning_rate": 1.6753331792619984e-05, + "loss": 0.0, + "step": 17263 + }, + { + "epoch": 3.99, + "learning_rate": 1.675140590093213e-05, + "loss": 0.0, + "step": 17264 + }, + { + "epoch": 3.99, + "learning_rate": 1.6749480009244282e-05, + "loss": 0.0, + "step": 17265 + }, + { + "epoch": 3.99, + "learning_rate": 1.674755411755643e-05, + "loss": 0.0, + "step": 17266 + }, + { + "epoch": 3.99, + "learning_rate": 1.674562822586858e-05, + "loss": 0.0001, + "step": 17267 + }, + { + "epoch": 3.99, + "learning_rate": 1.6743702334180728e-05, + "loss": 0.0, + "step": 17268 + }, + { + "epoch": 3.99, + "learning_rate": 1.6741776442492875e-05, + "loss": 0.0, + "step": 17269 + }, + { + "epoch": 3.99, + "learning_rate": 1.6739850550805023e-05, + "loss": 0.0, + "step": 17270 + }, + { + "epoch": 3.99, + "learning_rate": 1.673792465911717e-05, + "loss": 0.0, + "step": 17271 + }, + { + "epoch": 3.99, + "learning_rate": 1.673599876742932e-05, + "loss": 0.0, + "step": 17272 + }, + { + "epoch": 3.99, + "learning_rate": 1.673407287574147e-05, + "loss": 0.0, + "step": 17273 + }, + { + "epoch": 3.99, + "learning_rate": 1.673214698405362e-05, + "loss": 0.0003, + "step": 17274 + }, + { + "epoch": 3.99, + "learning_rate": 1.6730221092365767e-05, + "loss": 0.0, + "step": 17275 + }, + { + "epoch": 3.99, + "learning_rate": 1.6728295200677914e-05, + "loss": 0.0, + "step": 17276 + }, + { + "epoch": 3.99, + "learning_rate": 1.6726369308990065e-05, + "loss": 0.0, + "step": 17277 + }, + { + "epoch": 3.99, + "learning_rate": 1.6724443417302213e-05, + "loss": 0.0, + "step": 17278 + }, + { + "epoch": 3.99, + "learning_rate": 1.672251752561436e-05, + "loss": 0.0, + "step": 17279 + }, + { + "epoch": 3.99, + "learning_rate": 1.6720591633926508e-05, + "loss": 0.0, + "step": 17280 + }, + { + "epoch": 3.99, + "learning_rate": 1.6718665742238655e-05, + "loss": 0.0, + "step": 17281 + }, + { + "epoch": 3.99, + "learning_rate": 1.6716739850550806e-05, + "loss": 0.0, + "step": 17282 + }, + { + "epoch": 3.99, + "learning_rate": 1.6714813958862953e-05, + "loss": 0.0, + "step": 17283 + }, + { + "epoch": 3.99, + "learning_rate": 1.6712888067175104e-05, + "loss": 0.0004, + "step": 17284 + }, + { + "epoch": 3.99, + "learning_rate": 1.6710962175487252e-05, + "loss": 0.0001, + "step": 17285 + }, + { + "epoch": 3.99, + "learning_rate": 1.67090362837994e-05, + "loss": 0.0, + "step": 17286 + }, + { + "epoch": 4.0, + "learning_rate": 1.670711039211155e-05, + "loss": 0.0, + "step": 17287 + }, + { + "epoch": 4.0, + "learning_rate": 1.6705184500423694e-05, + "loss": 0.0, + "step": 17288 + }, + { + "epoch": 4.0, + "learning_rate": 1.6703258608735845e-05, + "loss": 0.0, + "step": 17289 + }, + { + "epoch": 4.0, + "learning_rate": 1.6701332717047993e-05, + "loss": 0.0, + "step": 17290 + }, + { + "epoch": 4.0, + "learning_rate": 1.6699406825360143e-05, + "loss": 0.0, + "step": 17291 + }, + { + "epoch": 4.0, + "learning_rate": 1.669748093367229e-05, + "loss": 0.0, + "step": 17292 + }, + { + "epoch": 4.0, + "learning_rate": 1.669555504198444e-05, + "loss": 0.0, + "step": 17293 + }, + { + "epoch": 4.0, + "learning_rate": 1.669362915029659e-05, + "loss": 0.0, + "step": 17294 + }, + { + "epoch": 4.0, + "learning_rate": 1.6691703258608737e-05, + "loss": 0.0, + "step": 17295 + }, + { + "epoch": 4.0, + "learning_rate": 1.6689777366920888e-05, + "loss": 0.0, + "step": 17296 + }, + { + "epoch": 4.0, + "learning_rate": 1.6687851475233035e-05, + "loss": 0.0, + "step": 17297 + }, + { + "epoch": 4.0, + "learning_rate": 1.6685925583545183e-05, + "loss": 0.0, + "step": 17298 + }, + { + "epoch": 4.0, + "learning_rate": 1.668399969185733e-05, + "loss": 0.0001, + "step": 17299 + }, + { + "epoch": 4.0, + "learning_rate": 1.6682073800169477e-05, + "loss": 0.0, + "step": 17300 + }, + { + "epoch": 4.0, + "learning_rate": 1.668014790848163e-05, + "loss": 0.0, + "step": 17301 + }, + { + "epoch": 4.0, + "learning_rate": 1.6678222016793776e-05, + "loss": 0.0, + "step": 17302 + }, + { + "epoch": 4.0, + "learning_rate": 1.6676296125105927e-05, + "loss": 0.0, + "step": 17303 + }, + { + "epoch": 4.0, + "learning_rate": 1.6674370233418074e-05, + "loss": 0.0, + "step": 17304 + }, + { + "epoch": 4.0, + "learning_rate": 1.667244434173022e-05, + "loss": 0.0, + "step": 17305 + }, + { + "epoch": 4.0, + "learning_rate": 1.6670518450042373e-05, + "loss": 0.0, + "step": 17306 + }, + { + "epoch": 4.0, + "learning_rate": 1.6668592558354517e-05, + "loss": 0.0, + "step": 17307 + }, + { + "epoch": 4.0, + "learning_rate": 1.6666666666666667e-05, + "loss": 0.0, + "step": 17308 + }, + { + "epoch": 4.0, + "learning_rate": 1.6664740774978815e-05, + "loss": 0.0, + "step": 17309 + }, + { + "epoch": 4.0, + "learning_rate": 1.6662814883290966e-05, + "loss": 0.0, + "step": 17310 + }, + { + "epoch": 4.0, + "learning_rate": 1.6660888991603113e-05, + "loss": 0.0, + "step": 17311 + }, + { + "epoch": 4.0, + "learning_rate": 1.665896309991526e-05, + "loss": 0.0, + "step": 17312 + }, + { + "epoch": 4.0, + "learning_rate": 1.665703720822741e-05, + "loss": 0.0, + "step": 17313 + }, + { + "epoch": 4.0, + "learning_rate": 1.665511131653956e-05, + "loss": 0.0, + "step": 17314 + }, + { + "epoch": 4.0, + "learning_rate": 1.665318542485171e-05, + "loss": 0.0, + "step": 17315 + }, + { + "epoch": 4.0, + "learning_rate": 1.6651259533163854e-05, + "loss": 0.0, + "step": 17316 + }, + { + "epoch": 4.0, + "learning_rate": 1.6649333641476e-05, + "loss": 0.0, + "step": 17317 + }, + { + "epoch": 4.0, + "learning_rate": 1.6647407749788152e-05, + "loss": 0.0, + "step": 17318 + }, + { + "epoch": 4.0, + "learning_rate": 1.66454818581003e-05, + "loss": 0.0, + "step": 17319 + }, + { + "epoch": 4.0, + "learning_rate": 1.664355596641245e-05, + "loss": 0.0, + "step": 17320 + }, + { + "epoch": 4.0, + "learning_rate": 1.6641630074724598e-05, + "loss": 0.0, + "step": 17321 + }, + { + "epoch": 4.0, + "learning_rate": 1.6639704183036746e-05, + "loss": 0.0, + "step": 17322 + }, + { + "epoch": 4.0, + "learning_rate": 1.6637778291348897e-05, + "loss": 0.0, + "step": 17323 + }, + { + "epoch": 4.0, + "learning_rate": 1.6635852399661044e-05, + "loss": 0.0, + "step": 17324 + }, + { + "epoch": 4.0, + "learning_rate": 1.6633926507973195e-05, + "loss": 0.0, + "step": 17325 + }, + { + "epoch": 4.0, + "learning_rate": 1.663200061628534e-05, + "loss": 0.0, + "step": 17326 + }, + { + "epoch": 4.0, + "learning_rate": 1.663007472459749e-05, + "loss": 0.0, + "step": 17327 + }, + { + "epoch": 4.0, + "learning_rate": 1.6628148832909637e-05, + "loss": 0.0, + "step": 17328 + }, + { + "epoch": 4.0, + "learning_rate": 1.6626222941221785e-05, + "loss": 0.0, + "step": 17329 + }, + { + "epoch": 4.01, + "learning_rate": 1.6624297049533936e-05, + "loss": 0.0, + "step": 17330 + }, + { + "epoch": 4.01, + "learning_rate": 1.6622371157846083e-05, + "loss": 0.0, + "step": 17331 + }, + { + "epoch": 4.01, + "learning_rate": 1.6620445266158234e-05, + "loss": 0.0, + "step": 17332 + }, + { + "epoch": 4.01, + "learning_rate": 1.661851937447038e-05, + "loss": 0.0, + "step": 17333 + }, + { + "epoch": 4.01, + "learning_rate": 1.661659348278253e-05, + "loss": 0.0001, + "step": 17334 + }, + { + "epoch": 4.01, + "learning_rate": 1.6614667591094676e-05, + "loss": 0.0, + "step": 17335 + }, + { + "epoch": 4.01, + "learning_rate": 1.6612741699406824e-05, + "loss": 0.0, + "step": 17336 + }, + { + "epoch": 4.01, + "learning_rate": 1.6610815807718975e-05, + "loss": 0.0, + "step": 17337 + }, + { + "epoch": 4.01, + "learning_rate": 1.6608889916031122e-05, + "loss": 0.0, + "step": 17338 + }, + { + "epoch": 4.01, + "learning_rate": 1.6606964024343273e-05, + "loss": 0.0, + "step": 17339 + }, + { + "epoch": 4.01, + "learning_rate": 1.660503813265542e-05, + "loss": 0.0, + "step": 17340 + }, + { + "epoch": 4.01, + "learning_rate": 1.6603112240967568e-05, + "loss": 0.0, + "step": 17341 + }, + { + "epoch": 4.01, + "learning_rate": 1.660118634927972e-05, + "loss": 0.0, + "step": 17342 + }, + { + "epoch": 4.01, + "learning_rate": 1.6599260457591866e-05, + "loss": 0.0, + "step": 17343 + }, + { + "epoch": 4.01, + "learning_rate": 1.6597334565904014e-05, + "loss": 0.0, + "step": 17344 + }, + { + "epoch": 4.01, + "learning_rate": 1.659540867421616e-05, + "loss": 0.0, + "step": 17345 + }, + { + "epoch": 4.01, + "learning_rate": 1.6593482782528312e-05, + "loss": 0.0, + "step": 17346 + }, + { + "epoch": 4.01, + "learning_rate": 1.659155689084046e-05, + "loss": 0.0, + "step": 17347 + }, + { + "epoch": 4.01, + "learning_rate": 1.6589630999152607e-05, + "loss": 0.0001, + "step": 17348 + }, + { + "epoch": 4.01, + "learning_rate": 1.6587705107464758e-05, + "loss": 0.0, + "step": 17349 + }, + { + "epoch": 4.01, + "learning_rate": 1.6585779215776906e-05, + "loss": 0.0, + "step": 17350 + }, + { + "epoch": 4.01, + "learning_rate": 1.6583853324089056e-05, + "loss": 0.0, + "step": 17351 + }, + { + "epoch": 4.01, + "learning_rate": 1.6581927432401204e-05, + "loss": 0.0, + "step": 17352 + }, + { + "epoch": 4.01, + "learning_rate": 1.658000154071335e-05, + "loss": 0.0, + "step": 17353 + }, + { + "epoch": 4.01, + "learning_rate": 1.65780756490255e-05, + "loss": 0.0, + "step": 17354 + }, + { + "epoch": 4.01, + "learning_rate": 1.6576149757337646e-05, + "loss": 0.0, + "step": 17355 + }, + { + "epoch": 4.01, + "learning_rate": 1.6574223865649797e-05, + "loss": 0.0, + "step": 17356 + }, + { + "epoch": 4.01, + "learning_rate": 1.6572297973961945e-05, + "loss": 0.0, + "step": 17357 + }, + { + "epoch": 4.01, + "learning_rate": 1.6570372082274092e-05, + "loss": 0.0, + "step": 17358 + }, + { + "epoch": 4.01, + "learning_rate": 1.6568446190586243e-05, + "loss": 0.0, + "step": 17359 + }, + { + "epoch": 4.01, + "learning_rate": 1.656652029889839e-05, + "loss": 0.0, + "step": 17360 + }, + { + "epoch": 4.01, + "learning_rate": 1.656459440721054e-05, + "loss": 0.0, + "step": 17361 + }, + { + "epoch": 4.01, + "learning_rate": 1.656266851552269e-05, + "loss": 0.0, + "step": 17362 + }, + { + "epoch": 4.01, + "learning_rate": 1.6560742623834836e-05, + "loss": 0.0, + "step": 17363 + }, + { + "epoch": 4.01, + "learning_rate": 1.6558816732146984e-05, + "loss": 0.0, + "step": 17364 + }, + { + "epoch": 4.01, + "learning_rate": 1.655689084045913e-05, + "loss": 0.0, + "step": 17365 + }, + { + "epoch": 4.01, + "learning_rate": 1.6554964948771282e-05, + "loss": 0.0, + "step": 17366 + }, + { + "epoch": 4.01, + "learning_rate": 1.655303905708343e-05, + "loss": 0.0, + "step": 17367 + }, + { + "epoch": 4.01, + "learning_rate": 1.655111316539558e-05, + "loss": 0.0, + "step": 17368 + }, + { + "epoch": 4.01, + "learning_rate": 1.6549187273707728e-05, + "loss": 0.0, + "step": 17369 + }, + { + "epoch": 4.01, + "learning_rate": 1.6547261382019875e-05, + "loss": 0.0, + "step": 17370 + }, + { + "epoch": 4.01, + "learning_rate": 1.6545335490332026e-05, + "loss": 0.0001, + "step": 17371 + }, + { + "epoch": 4.01, + "learning_rate": 1.654340959864417e-05, + "loss": 0.0, + "step": 17372 + }, + { + "epoch": 4.02, + "learning_rate": 1.654148370695632e-05, + "loss": 0.0, + "step": 17373 + }, + { + "epoch": 4.02, + "learning_rate": 1.653955781526847e-05, + "loss": 0.0, + "step": 17374 + }, + { + "epoch": 4.02, + "learning_rate": 1.653763192358062e-05, + "loss": 0.0001, + "step": 17375 + }, + { + "epoch": 4.02, + "learning_rate": 1.6535706031892767e-05, + "loss": 0.0, + "step": 17376 + }, + { + "epoch": 4.02, + "learning_rate": 1.6533780140204915e-05, + "loss": 0.0001, + "step": 17377 + }, + { + "epoch": 4.02, + "learning_rate": 1.6531854248517065e-05, + "loss": 0.0, + "step": 17378 + }, + { + "epoch": 4.02, + "learning_rate": 1.6529928356829213e-05, + "loss": 0.0, + "step": 17379 + }, + { + "epoch": 4.02, + "learning_rate": 1.6528002465141364e-05, + "loss": 0.0, + "step": 17380 + }, + { + "epoch": 4.02, + "learning_rate": 1.652607657345351e-05, + "loss": 0.0, + "step": 17381 + }, + { + "epoch": 4.02, + "learning_rate": 1.652415068176566e-05, + "loss": 0.0, + "step": 17382 + }, + { + "epoch": 4.02, + "learning_rate": 1.6522224790077806e-05, + "loss": 0.0, + "step": 17383 + }, + { + "epoch": 4.02, + "learning_rate": 1.6520298898389954e-05, + "loss": 0.0, + "step": 17384 + }, + { + "epoch": 4.02, + "learning_rate": 1.6518373006702105e-05, + "loss": 0.0, + "step": 17385 + }, + { + "epoch": 4.02, + "learning_rate": 1.6516447115014252e-05, + "loss": 0.0, + "step": 17386 + }, + { + "epoch": 4.02, + "learning_rate": 1.6514521223326403e-05, + "loss": 0.0001, + "step": 17387 + }, + { + "epoch": 4.02, + "learning_rate": 1.651259533163855e-05, + "loss": 0.0, + "step": 17388 + }, + { + "epoch": 4.02, + "learning_rate": 1.6510669439950698e-05, + "loss": 0.0, + "step": 17389 + }, + { + "epoch": 4.02, + "learning_rate": 1.650874354826285e-05, + "loss": 0.0, + "step": 17390 + }, + { + "epoch": 4.02, + "learning_rate": 1.6506817656574993e-05, + "loss": 0.0, + "step": 17391 + }, + { + "epoch": 4.02, + "learning_rate": 1.6504891764887144e-05, + "loss": 0.0, + "step": 17392 + }, + { + "epoch": 4.02, + "learning_rate": 1.650296587319929e-05, + "loss": 0.0, + "step": 17393 + }, + { + "epoch": 4.02, + "learning_rate": 1.650103998151144e-05, + "loss": 0.0, + "step": 17394 + }, + { + "epoch": 4.02, + "learning_rate": 1.649911408982359e-05, + "loss": 0.0, + "step": 17395 + }, + { + "epoch": 4.02, + "learning_rate": 1.6497188198135737e-05, + "loss": 0.0, + "step": 17396 + }, + { + "epoch": 4.02, + "learning_rate": 1.6495262306447888e-05, + "loss": 0.0, + "step": 17397 + }, + { + "epoch": 4.02, + "learning_rate": 1.6493336414760035e-05, + "loss": 0.0, + "step": 17398 + }, + { + "epoch": 4.02, + "learning_rate": 1.6491410523072183e-05, + "loss": 0.0, + "step": 17399 + }, + { + "epoch": 4.02, + "learning_rate": 1.648948463138433e-05, + "loss": 0.0, + "step": 17400 + }, + { + "epoch": 4.02, + "learning_rate": 1.6487558739696478e-05, + "loss": 0.0, + "step": 17401 + }, + { + "epoch": 4.02, + "learning_rate": 1.648563284800863e-05, + "loss": 0.0, + "step": 17402 + }, + { + "epoch": 4.02, + "learning_rate": 1.6483706956320776e-05, + "loss": 0.0, + "step": 17403 + }, + { + "epoch": 4.02, + "learning_rate": 1.6481781064632927e-05, + "loss": 0.0, + "step": 17404 + }, + { + "epoch": 4.02, + "learning_rate": 1.6479855172945074e-05, + "loss": 0.0024, + "step": 17405 + }, + { + "epoch": 4.02, + "learning_rate": 1.6477929281257222e-05, + "loss": 0.0, + "step": 17406 + }, + { + "epoch": 4.02, + "learning_rate": 1.6476003389569373e-05, + "loss": 0.0, + "step": 17407 + }, + { + "epoch": 4.02, + "learning_rate": 1.647407749788152e-05, + "loss": 0.0, + "step": 17408 + }, + { + "epoch": 4.02, + "learning_rate": 1.647215160619367e-05, + "loss": 0.0, + "step": 17409 + }, + { + "epoch": 4.02, + "learning_rate": 1.6470225714505815e-05, + "loss": 0.0, + "step": 17410 + }, + { + "epoch": 4.02, + "learning_rate": 1.6468299822817966e-05, + "loss": 0.0, + "step": 17411 + }, + { + "epoch": 4.02, + "learning_rate": 1.6466373931130114e-05, + "loss": 0.0, + "step": 17412 + }, + { + "epoch": 4.02, + "learning_rate": 1.646444803944226e-05, + "loss": 0.0, + "step": 17413 + }, + { + "epoch": 4.02, + "learning_rate": 1.6462522147754412e-05, + "loss": 0.0, + "step": 17414 + }, + { + "epoch": 4.02, + "learning_rate": 1.646059625606656e-05, + "loss": 0.0, + "step": 17415 + }, + { + "epoch": 4.02, + "learning_rate": 1.645867036437871e-05, + "loss": 0.0, + "step": 17416 + }, + { + "epoch": 4.03, + "learning_rate": 1.6456744472690858e-05, + "loss": 0.0, + "step": 17417 + }, + { + "epoch": 4.03, + "learning_rate": 1.6454818581003005e-05, + "loss": 0.0, + "step": 17418 + }, + { + "epoch": 4.03, + "learning_rate": 1.6452892689315153e-05, + "loss": 0.0, + "step": 17419 + }, + { + "epoch": 4.03, + "learning_rate": 1.64509667976273e-05, + "loss": 0.0, + "step": 17420 + }, + { + "epoch": 4.03, + "learning_rate": 1.644904090593945e-05, + "loss": 0.0, + "step": 17421 + }, + { + "epoch": 4.03, + "learning_rate": 1.64471150142516e-05, + "loss": 0.0, + "step": 17422 + }, + { + "epoch": 4.03, + "learning_rate": 1.644518912256375e-05, + "loss": 0.0, + "step": 17423 + }, + { + "epoch": 4.03, + "learning_rate": 1.6443263230875897e-05, + "loss": 0.0, + "step": 17424 + }, + { + "epoch": 4.03, + "learning_rate": 1.6441337339188044e-05, + "loss": 0.0, + "step": 17425 + }, + { + "epoch": 4.03, + "learning_rate": 1.6439411447500195e-05, + "loss": 0.0, + "step": 17426 + }, + { + "epoch": 4.03, + "learning_rate": 1.6437485555812343e-05, + "loss": 0.0, + "step": 17427 + }, + { + "epoch": 4.03, + "learning_rate": 1.643555966412449e-05, + "loss": 0.0, + "step": 17428 + }, + { + "epoch": 4.03, + "learning_rate": 1.6433633772436638e-05, + "loss": 0.0, + "step": 17429 + }, + { + "epoch": 4.03, + "learning_rate": 1.6431707880748785e-05, + "loss": 0.0, + "step": 17430 + }, + { + "epoch": 4.03, + "learning_rate": 1.6429781989060936e-05, + "loss": 0.0, + "step": 17431 + }, + { + "epoch": 4.03, + "learning_rate": 1.6427856097373083e-05, + "loss": 0.0, + "step": 17432 + }, + { + "epoch": 4.03, + "learning_rate": 1.6425930205685234e-05, + "loss": 0.0001, + "step": 17433 + }, + { + "epoch": 4.03, + "learning_rate": 1.6424004313997382e-05, + "loss": 0.0, + "step": 17434 + }, + { + "epoch": 4.03, + "learning_rate": 1.642207842230953e-05, + "loss": 0.0, + "step": 17435 + }, + { + "epoch": 4.03, + "learning_rate": 1.642015253062168e-05, + "loss": 0.0, + "step": 17436 + }, + { + "epoch": 4.03, + "learning_rate": 1.6418226638933828e-05, + "loss": 0.0001, + "step": 17437 + }, + { + "epoch": 4.03, + "learning_rate": 1.6416300747245975e-05, + "loss": 0.0, + "step": 17438 + }, + { + "epoch": 4.03, + "learning_rate": 1.6414374855558123e-05, + "loss": 0.0, + "step": 17439 + }, + { + "epoch": 4.03, + "learning_rate": 1.6412448963870273e-05, + "loss": 0.0, + "step": 17440 + }, + { + "epoch": 4.03, + "learning_rate": 1.641052307218242e-05, + "loss": 0.0, + "step": 17441 + }, + { + "epoch": 4.03, + "learning_rate": 1.640859718049457e-05, + "loss": 0.0, + "step": 17442 + }, + { + "epoch": 4.03, + "learning_rate": 1.640667128880672e-05, + "loss": 0.0001, + "step": 17443 + }, + { + "epoch": 4.03, + "learning_rate": 1.6404745397118867e-05, + "loss": 0.0, + "step": 17444 + }, + { + "epoch": 4.03, + "learning_rate": 1.6402819505431018e-05, + "loss": 0.0, + "step": 17445 + }, + { + "epoch": 4.03, + "learning_rate": 1.6400893613743165e-05, + "loss": 0.0, + "step": 17446 + }, + { + "epoch": 4.03, + "learning_rate": 1.6398967722055313e-05, + "loss": 0.0, + "step": 17447 + }, + { + "epoch": 4.03, + "learning_rate": 1.639704183036746e-05, + "loss": 0.0, + "step": 17448 + }, + { + "epoch": 4.03, + "learning_rate": 1.6395115938679608e-05, + "loss": 0.0, + "step": 17449 + }, + { + "epoch": 4.03, + "learning_rate": 1.639319004699176e-05, + "loss": 0.0, + "step": 17450 + }, + { + "epoch": 4.03, + "learning_rate": 1.6391264155303906e-05, + "loss": 0.0, + "step": 17451 + }, + { + "epoch": 4.03, + "learning_rate": 1.6389338263616057e-05, + "loss": 0.0, + "step": 17452 + }, + { + "epoch": 4.03, + "learning_rate": 1.6387412371928204e-05, + "loss": 0.0, + "step": 17453 + }, + { + "epoch": 4.03, + "learning_rate": 1.638548648024035e-05, + "loss": 0.0, + "step": 17454 + }, + { + "epoch": 4.03, + "learning_rate": 1.6383560588552503e-05, + "loss": 0.0, + "step": 17455 + }, + { + "epoch": 4.03, + "learning_rate": 1.6381634696864647e-05, + "loss": 0.0, + "step": 17456 + }, + { + "epoch": 4.03, + "learning_rate": 1.6379708805176798e-05, + "loss": 0.0, + "step": 17457 + }, + { + "epoch": 4.03, + "learning_rate": 1.6377782913488945e-05, + "loss": 0.0, + "step": 17458 + }, + { + "epoch": 4.03, + "learning_rate": 1.6375857021801096e-05, + "loss": 0.0, + "step": 17459 + }, + { + "epoch": 4.04, + "learning_rate": 1.6373931130113243e-05, + "loss": 0.0, + "step": 17460 + }, + { + "epoch": 4.04, + "learning_rate": 1.637200523842539e-05, + "loss": 0.0, + "step": 17461 + }, + { + "epoch": 4.04, + "learning_rate": 1.637007934673754e-05, + "loss": 0.0, + "step": 17462 + }, + { + "epoch": 4.04, + "learning_rate": 1.636815345504969e-05, + "loss": 0.0, + "step": 17463 + }, + { + "epoch": 4.04, + "learning_rate": 1.636622756336184e-05, + "loss": 0.0, + "step": 17464 + }, + { + "epoch": 4.04, + "learning_rate": 1.6364301671673987e-05, + "loss": 0.0, + "step": 17465 + }, + { + "epoch": 4.04, + "learning_rate": 1.636237577998613e-05, + "loss": 0.0, + "step": 17466 + }, + { + "epoch": 4.04, + "learning_rate": 1.6360449888298282e-05, + "loss": 0.0, + "step": 17467 + }, + { + "epoch": 4.04, + "learning_rate": 1.635852399661043e-05, + "loss": 0.0, + "step": 17468 + }, + { + "epoch": 4.04, + "learning_rate": 1.635659810492258e-05, + "loss": 0.0, + "step": 17469 + }, + { + "epoch": 4.04, + "learning_rate": 1.6354672213234728e-05, + "loss": 0.0, + "step": 17470 + }, + { + "epoch": 4.04, + "learning_rate": 1.6352746321546876e-05, + "loss": 0.0, + "step": 17471 + }, + { + "epoch": 4.04, + "learning_rate": 1.6350820429859027e-05, + "loss": 0.0, + "step": 17472 + }, + { + "epoch": 4.04, + "learning_rate": 1.6348894538171174e-05, + "loss": 0.0, + "step": 17473 + }, + { + "epoch": 4.04, + "learning_rate": 1.6346968646483325e-05, + "loss": 0.0, + "step": 17474 + }, + { + "epoch": 4.04, + "learning_rate": 1.634504275479547e-05, + "loss": 0.0, + "step": 17475 + }, + { + "epoch": 4.04, + "learning_rate": 1.634311686310762e-05, + "loss": 0.0, + "step": 17476 + }, + { + "epoch": 4.04, + "learning_rate": 1.6341190971419767e-05, + "loss": 0.0, + "step": 17477 + }, + { + "epoch": 4.04, + "learning_rate": 1.6339265079731915e-05, + "loss": 0.0, + "step": 17478 + }, + { + "epoch": 4.04, + "learning_rate": 1.6337339188044066e-05, + "loss": 0.0, + "step": 17479 + }, + { + "epoch": 4.04, + "learning_rate": 1.6335413296356213e-05, + "loss": 0.0001, + "step": 17480 + }, + { + "epoch": 4.04, + "learning_rate": 1.6333487404668364e-05, + "loss": 0.0, + "step": 17481 + }, + { + "epoch": 4.04, + "learning_rate": 1.633156151298051e-05, + "loss": 0.0, + "step": 17482 + }, + { + "epoch": 4.04, + "learning_rate": 1.632963562129266e-05, + "loss": 0.0, + "step": 17483 + }, + { + "epoch": 4.04, + "learning_rate": 1.6327709729604807e-05, + "loss": 0.0, + "step": 17484 + }, + { + "epoch": 4.04, + "learning_rate": 1.6325783837916954e-05, + "loss": 0.0, + "step": 17485 + }, + { + "epoch": 4.04, + "learning_rate": 1.6323857946229105e-05, + "loss": 0.0, + "step": 17486 + }, + { + "epoch": 4.04, + "learning_rate": 1.6321932054541252e-05, + "loss": 0.0, + "step": 17487 + }, + { + "epoch": 4.04, + "learning_rate": 1.6320006162853403e-05, + "loss": 0.0, + "step": 17488 + }, + { + "epoch": 4.04, + "learning_rate": 1.631808027116555e-05, + "loss": 0.0, + "step": 17489 + }, + { + "epoch": 4.04, + "learning_rate": 1.6316154379477698e-05, + "loss": 0.0, + "step": 17490 + }, + { + "epoch": 4.04, + "learning_rate": 1.631422848778985e-05, + "loss": 0.0, + "step": 17491 + }, + { + "epoch": 4.04, + "learning_rate": 1.6312302596101996e-05, + "loss": 0.0, + "step": 17492 + }, + { + "epoch": 4.04, + "learning_rate": 1.6310376704414147e-05, + "loss": 0.0, + "step": 17493 + }, + { + "epoch": 4.04, + "learning_rate": 1.630845081272629e-05, + "loss": 0.0, + "step": 17494 + }, + { + "epoch": 4.04, + "learning_rate": 1.6306524921038442e-05, + "loss": 0.0, + "step": 17495 + }, + { + "epoch": 4.04, + "learning_rate": 1.630459902935059e-05, + "loss": 0.0, + "step": 17496 + }, + { + "epoch": 4.04, + "learning_rate": 1.6302673137662737e-05, + "loss": 0.0, + "step": 17497 + }, + { + "epoch": 4.04, + "learning_rate": 1.6300747245974888e-05, + "loss": 0.0, + "step": 17498 + }, + { + "epoch": 4.04, + "learning_rate": 1.6298821354287036e-05, + "loss": 0.0, + "step": 17499 + }, + { + "epoch": 4.04, + "learning_rate": 1.6296895462599186e-05, + "loss": 0.0, + "step": 17500 + }, + { + "epoch": 4.04, + "learning_rate": 1.6294969570911334e-05, + "loss": 0.0, + "step": 17501 + }, + { + "epoch": 4.04, + "learning_rate": 1.629304367922348e-05, + "loss": 0.0, + "step": 17502 + }, + { + "epoch": 4.05, + "learning_rate": 1.629111778753563e-05, + "loss": 0.0, + "step": 17503 + }, + { + "epoch": 4.05, + "learning_rate": 1.6289191895847776e-05, + "loss": 0.0, + "step": 17504 + }, + { + "epoch": 4.05, + "learning_rate": 1.6287266004159927e-05, + "loss": 0.0, + "step": 17505 + }, + { + "epoch": 4.05, + "learning_rate": 1.6285340112472075e-05, + "loss": 0.0, + "step": 17506 + }, + { + "epoch": 4.05, + "learning_rate": 1.6283414220784222e-05, + "loss": 0.0, + "step": 17507 + }, + { + "epoch": 4.05, + "learning_rate": 1.6281488329096373e-05, + "loss": 0.0, + "step": 17508 + }, + { + "epoch": 4.05, + "learning_rate": 1.627956243740852e-05, + "loss": 0.0, + "step": 17509 + }, + { + "epoch": 4.05, + "learning_rate": 1.627763654572067e-05, + "loss": 0.0, + "step": 17510 + }, + { + "epoch": 4.05, + "learning_rate": 1.627571065403282e-05, + "loss": 0.0, + "step": 17511 + }, + { + "epoch": 4.05, + "learning_rate": 1.6273784762344966e-05, + "loss": 0.0, + "step": 17512 + }, + { + "epoch": 4.05, + "learning_rate": 1.6271858870657114e-05, + "loss": 0.0, + "step": 17513 + }, + { + "epoch": 4.05, + "learning_rate": 1.626993297896926e-05, + "loss": 0.0, + "step": 17514 + }, + { + "epoch": 4.05, + "learning_rate": 1.6268007087281412e-05, + "loss": 0.0, + "step": 17515 + }, + { + "epoch": 4.05, + "learning_rate": 1.626608119559356e-05, + "loss": 0.0, + "step": 17516 + }, + { + "epoch": 4.05, + "learning_rate": 1.626415530390571e-05, + "loss": 0.0, + "step": 17517 + }, + { + "epoch": 4.05, + "learning_rate": 1.6262229412217858e-05, + "loss": 0.0, + "step": 17518 + }, + { + "epoch": 4.05, + "learning_rate": 1.6260303520530006e-05, + "loss": 0.0, + "step": 17519 + }, + { + "epoch": 4.05, + "learning_rate": 1.6258377628842156e-05, + "loss": 0.0, + "step": 17520 + }, + { + "epoch": 4.05, + "learning_rate": 1.6256451737154304e-05, + "loss": 0.0, + "step": 17521 + }, + { + "epoch": 4.05, + "learning_rate": 1.625452584546645e-05, + "loss": 0.0, + "step": 17522 + }, + { + "epoch": 4.05, + "learning_rate": 1.62525999537786e-05, + "loss": 0.0, + "step": 17523 + }, + { + "epoch": 4.05, + "learning_rate": 1.625067406209075e-05, + "loss": 0.0, + "step": 17524 + }, + { + "epoch": 4.05, + "learning_rate": 1.6248748170402897e-05, + "loss": 0.0, + "step": 17525 + }, + { + "epoch": 4.05, + "learning_rate": 1.6246822278715045e-05, + "loss": 0.0, + "step": 17526 + }, + { + "epoch": 4.05, + "learning_rate": 1.6244896387027195e-05, + "loss": 0.0, + "step": 17527 + }, + { + "epoch": 4.05, + "learning_rate": 1.6242970495339343e-05, + "loss": 0.0, + "step": 17528 + }, + { + "epoch": 4.05, + "learning_rate": 1.6241044603651494e-05, + "loss": 0.0, + "step": 17529 + }, + { + "epoch": 4.05, + "learning_rate": 1.623911871196364e-05, + "loss": 0.0, + "step": 17530 + }, + { + "epoch": 4.05, + "learning_rate": 1.623719282027579e-05, + "loss": 0.0, + "step": 17531 + }, + { + "epoch": 4.05, + "learning_rate": 1.6235266928587936e-05, + "loss": 0.0, + "step": 17532 + }, + { + "epoch": 4.05, + "learning_rate": 1.6233341036900084e-05, + "loss": 0.0, + "step": 17533 + }, + { + "epoch": 4.05, + "learning_rate": 1.6231415145212235e-05, + "loss": 0.0, + "step": 17534 + }, + { + "epoch": 4.05, + "learning_rate": 1.6229489253524382e-05, + "loss": 0.0, + "step": 17535 + }, + { + "epoch": 4.05, + "learning_rate": 1.6227563361836533e-05, + "loss": 0.0, + "step": 17536 + }, + { + "epoch": 4.05, + "learning_rate": 1.622563747014868e-05, + "loss": 0.0, + "step": 17537 + }, + { + "epoch": 4.05, + "learning_rate": 1.6223711578460828e-05, + "loss": 0.0, + "step": 17538 + }, + { + "epoch": 4.05, + "learning_rate": 1.622178568677298e-05, + "loss": 0.0, + "step": 17539 + }, + { + "epoch": 4.05, + "learning_rate": 1.6219859795085123e-05, + "loss": 0.0, + "step": 17540 + }, + { + "epoch": 4.05, + "learning_rate": 1.6217933903397274e-05, + "loss": 0.0, + "step": 17541 + }, + { + "epoch": 4.05, + "learning_rate": 1.621600801170942e-05, + "loss": 0.0, + "step": 17542 + }, + { + "epoch": 4.05, + "learning_rate": 1.621408212002157e-05, + "loss": 0.0, + "step": 17543 + }, + { + "epoch": 4.05, + "learning_rate": 1.621215622833372e-05, + "loss": 0.0, + "step": 17544 + }, + { + "epoch": 4.05, + "learning_rate": 1.6210230336645867e-05, + "loss": 0.0, + "step": 17545 + }, + { + "epoch": 4.06, + "learning_rate": 1.6208304444958018e-05, + "loss": 0.0, + "step": 17546 + }, + { + "epoch": 4.06, + "learning_rate": 1.6206378553270165e-05, + "loss": 0.0, + "step": 17547 + }, + { + "epoch": 4.06, + "learning_rate": 1.6204452661582313e-05, + "loss": 0.0, + "step": 17548 + }, + { + "epoch": 4.06, + "learning_rate": 1.6202526769894464e-05, + "loss": 0.0, + "step": 17549 + }, + { + "epoch": 4.06, + "learning_rate": 1.6200600878206608e-05, + "loss": 0.0, + "step": 17550 + }, + { + "epoch": 4.06, + "learning_rate": 1.619867498651876e-05, + "loss": 0.0, + "step": 17551 + }, + { + "epoch": 4.06, + "learning_rate": 1.6196749094830906e-05, + "loss": 0.0, + "step": 17552 + }, + { + "epoch": 4.06, + "learning_rate": 1.6194823203143057e-05, + "loss": 0.0, + "step": 17553 + }, + { + "epoch": 4.06, + "learning_rate": 1.6192897311455204e-05, + "loss": 0.0, + "step": 17554 + }, + { + "epoch": 4.06, + "learning_rate": 1.6190971419767352e-05, + "loss": 0.0, + "step": 17555 + }, + { + "epoch": 4.06, + "learning_rate": 1.6189045528079503e-05, + "loss": 0.0003, + "step": 17556 + }, + { + "epoch": 4.06, + "learning_rate": 1.618711963639165e-05, + "loss": 0.0, + "step": 17557 + }, + { + "epoch": 4.06, + "learning_rate": 1.61851937447038e-05, + "loss": 0.0, + "step": 17558 + }, + { + "epoch": 4.06, + "learning_rate": 1.6183267853015945e-05, + "loss": 0.0, + "step": 17559 + }, + { + "epoch": 4.06, + "learning_rate": 1.6181341961328096e-05, + "loss": 0.0, + "step": 17560 + }, + { + "epoch": 4.06, + "learning_rate": 1.6179416069640244e-05, + "loss": 0.0001, + "step": 17561 + }, + { + "epoch": 4.06, + "learning_rate": 1.617749017795239e-05, + "loss": 0.0, + "step": 17562 + }, + { + "epoch": 4.06, + "learning_rate": 1.6175564286264542e-05, + "loss": 0.0, + "step": 17563 + }, + { + "epoch": 4.06, + "learning_rate": 1.617363839457669e-05, + "loss": 0.0, + "step": 17564 + }, + { + "epoch": 4.06, + "learning_rate": 1.617171250288884e-05, + "loss": 0.0, + "step": 17565 + }, + { + "epoch": 4.06, + "learning_rate": 1.6169786611200988e-05, + "loss": 0.0, + "step": 17566 + }, + { + "epoch": 4.06, + "learning_rate": 1.6167860719513135e-05, + "loss": 0.0, + "step": 17567 + }, + { + "epoch": 4.06, + "learning_rate": 1.6165934827825283e-05, + "loss": 0.0, + "step": 17568 + }, + { + "epoch": 4.06, + "learning_rate": 1.616400893613743e-05, + "loss": 0.0, + "step": 17569 + }, + { + "epoch": 4.06, + "learning_rate": 1.616208304444958e-05, + "loss": 0.0, + "step": 17570 + }, + { + "epoch": 4.06, + "learning_rate": 1.616015715276173e-05, + "loss": 0.0, + "step": 17571 + }, + { + "epoch": 4.06, + "learning_rate": 1.615823126107388e-05, + "loss": 0.0, + "step": 17572 + }, + { + "epoch": 4.06, + "learning_rate": 1.6156305369386027e-05, + "loss": 0.0, + "step": 17573 + }, + { + "epoch": 4.06, + "learning_rate": 1.6154379477698174e-05, + "loss": 0.0, + "step": 17574 + }, + { + "epoch": 4.06, + "learning_rate": 1.6152453586010325e-05, + "loss": 0.0, + "step": 17575 + }, + { + "epoch": 4.06, + "learning_rate": 1.6150527694322473e-05, + "loss": 0.0, + "step": 17576 + }, + { + "epoch": 4.06, + "learning_rate": 1.6148601802634624e-05, + "loss": 0.0, + "step": 17577 + }, + { + "epoch": 4.06, + "learning_rate": 1.6146675910946768e-05, + "loss": 0.0, + "step": 17578 + }, + { + "epoch": 4.06, + "learning_rate": 1.6144750019258915e-05, + "loss": 0.0, + "step": 17579 + }, + { + "epoch": 4.06, + "learning_rate": 1.6142824127571066e-05, + "loss": 0.0, + "step": 17580 + }, + { + "epoch": 4.06, + "learning_rate": 1.6140898235883214e-05, + "loss": 0.0, + "step": 17581 + }, + { + "epoch": 4.06, + "learning_rate": 1.6138972344195364e-05, + "loss": 0.0, + "step": 17582 + }, + { + "epoch": 4.06, + "learning_rate": 1.6137046452507512e-05, + "loss": 0.0, + "step": 17583 + }, + { + "epoch": 4.06, + "learning_rate": 1.613512056081966e-05, + "loss": 0.0, + "step": 17584 + }, + { + "epoch": 4.06, + "learning_rate": 1.613319466913181e-05, + "loss": 0.0, + "step": 17585 + }, + { + "epoch": 4.06, + "learning_rate": 1.6131268777443958e-05, + "loss": 0.0, + "step": 17586 + }, + { + "epoch": 4.06, + "learning_rate": 1.6129342885756105e-05, + "loss": 0.0, + "step": 17587 + }, + { + "epoch": 4.06, + "learning_rate": 1.6127416994068253e-05, + "loss": 0.0, + "step": 17588 + }, + { + "epoch": 4.06, + "learning_rate": 1.6125491102380403e-05, + "loss": 0.0, + "step": 17589 + }, + { + "epoch": 4.07, + "learning_rate": 1.612356521069255e-05, + "loss": 0.0, + "step": 17590 + }, + { + "epoch": 4.07, + "learning_rate": 1.61216393190047e-05, + "loss": 0.0, + "step": 17591 + }, + { + "epoch": 4.07, + "learning_rate": 1.611971342731685e-05, + "loss": 0.0, + "step": 17592 + }, + { + "epoch": 4.07, + "learning_rate": 1.6117787535628997e-05, + "loss": 0.0, + "step": 17593 + }, + { + "epoch": 4.07, + "learning_rate": 1.6115861643941148e-05, + "loss": 0.0, + "step": 17594 + }, + { + "epoch": 4.07, + "learning_rate": 1.6113935752253295e-05, + "loss": 0.0, + "step": 17595 + }, + { + "epoch": 4.07, + "learning_rate": 1.6112009860565443e-05, + "loss": 0.0, + "step": 17596 + }, + { + "epoch": 4.07, + "learning_rate": 1.611008396887759e-05, + "loss": 0.0, + "step": 17597 + }, + { + "epoch": 4.07, + "learning_rate": 1.6108158077189738e-05, + "loss": 0.0, + "step": 17598 + }, + { + "epoch": 4.07, + "learning_rate": 1.610623218550189e-05, + "loss": 0.0, + "step": 17599 + }, + { + "epoch": 4.07, + "learning_rate": 1.6104306293814036e-05, + "loss": 0.0, + "step": 17600 + }, + { + "epoch": 4.07, + "learning_rate": 1.6102380402126187e-05, + "loss": 0.0, + "step": 17601 + }, + { + "epoch": 4.07, + "learning_rate": 1.6100454510438334e-05, + "loss": 0.0, + "step": 17602 + }, + { + "epoch": 4.07, + "learning_rate": 1.6098528618750482e-05, + "loss": 0.0, + "step": 17603 + }, + { + "epoch": 4.07, + "learning_rate": 1.6096602727062633e-05, + "loss": 0.0, + "step": 17604 + }, + { + "epoch": 4.07, + "learning_rate": 1.609467683537478e-05, + "loss": 0.0, + "step": 17605 + }, + { + "epoch": 4.07, + "learning_rate": 1.6092750943686928e-05, + "loss": 0.0, + "step": 17606 + }, + { + "epoch": 4.07, + "learning_rate": 1.6090825051999075e-05, + "loss": 0.0001, + "step": 17607 + }, + { + "epoch": 4.07, + "learning_rate": 1.6088899160311226e-05, + "loss": 0.0, + "step": 17608 + }, + { + "epoch": 4.07, + "learning_rate": 1.6086973268623373e-05, + "loss": 0.0, + "step": 17609 + }, + { + "epoch": 4.07, + "learning_rate": 1.608504737693552e-05, + "loss": 0.0, + "step": 17610 + }, + { + "epoch": 4.07, + "learning_rate": 1.6083121485247672e-05, + "loss": 0.0009, + "step": 17611 + }, + { + "epoch": 4.07, + "learning_rate": 1.608119559355982e-05, + "loss": 0.0, + "step": 17612 + }, + { + "epoch": 4.07, + "learning_rate": 1.607926970187197e-05, + "loss": 0.0, + "step": 17613 + }, + { + "epoch": 4.07, + "learning_rate": 1.6077343810184118e-05, + "loss": 0.0, + "step": 17614 + }, + { + "epoch": 4.07, + "learning_rate": 1.6075417918496265e-05, + "loss": 0.0, + "step": 17615 + }, + { + "epoch": 4.07, + "learning_rate": 1.6073492026808412e-05, + "loss": 0.0, + "step": 17616 + }, + { + "epoch": 4.07, + "learning_rate": 1.607156613512056e-05, + "loss": 0.0, + "step": 17617 + }, + { + "epoch": 4.07, + "learning_rate": 1.606964024343271e-05, + "loss": 0.0, + "step": 17618 + }, + { + "epoch": 4.07, + "learning_rate": 1.606771435174486e-05, + "loss": 0.0, + "step": 17619 + }, + { + "epoch": 4.07, + "learning_rate": 1.6065788460057006e-05, + "loss": 0.0001, + "step": 17620 + }, + { + "epoch": 4.07, + "learning_rate": 1.6063862568369157e-05, + "loss": 0.0, + "step": 17621 + }, + { + "epoch": 4.07, + "learning_rate": 1.6061936676681304e-05, + "loss": 0.0, + "step": 17622 + }, + { + "epoch": 4.07, + "learning_rate": 1.6060010784993455e-05, + "loss": 0.0, + "step": 17623 + }, + { + "epoch": 4.07, + "learning_rate": 1.60580848933056e-05, + "loss": 0.0, + "step": 17624 + }, + { + "epoch": 4.07, + "learning_rate": 1.605615900161775e-05, + "loss": 0.0, + "step": 17625 + }, + { + "epoch": 4.07, + "learning_rate": 1.6054233109929897e-05, + "loss": 0.0, + "step": 17626 + }, + { + "epoch": 4.07, + "learning_rate": 1.6052307218242045e-05, + "loss": 0.0001, + "step": 17627 + }, + { + "epoch": 4.07, + "learning_rate": 1.6050381326554196e-05, + "loss": 0.0, + "step": 17628 + }, + { + "epoch": 4.07, + "learning_rate": 1.6048455434866343e-05, + "loss": 0.0, + "step": 17629 + }, + { + "epoch": 4.07, + "learning_rate": 1.6046529543178494e-05, + "loss": 0.0, + "step": 17630 + }, + { + "epoch": 4.07, + "learning_rate": 1.604460365149064e-05, + "loss": 0.0, + "step": 17631 + }, + { + "epoch": 4.07, + "learning_rate": 1.604267775980279e-05, + "loss": 0.0, + "step": 17632 + }, + { + "epoch": 4.08, + "learning_rate": 1.604075186811494e-05, + "loss": 0.0, + "step": 17633 + }, + { + "epoch": 4.08, + "learning_rate": 1.6038825976427084e-05, + "loss": 0.0, + "step": 17634 + }, + { + "epoch": 4.08, + "learning_rate": 1.6036900084739235e-05, + "loss": 0.0, + "step": 17635 + }, + { + "epoch": 4.08, + "learning_rate": 1.6034974193051382e-05, + "loss": 0.0, + "step": 17636 + }, + { + "epoch": 4.08, + "learning_rate": 1.6033048301363533e-05, + "loss": 0.0, + "step": 17637 + }, + { + "epoch": 4.08, + "learning_rate": 1.603112240967568e-05, + "loss": 0.0, + "step": 17638 + }, + { + "epoch": 4.08, + "learning_rate": 1.6029196517987828e-05, + "loss": 0.0, + "step": 17639 + }, + { + "epoch": 4.08, + "learning_rate": 1.602727062629998e-05, + "loss": 0.0, + "step": 17640 + }, + { + "epoch": 4.08, + "learning_rate": 1.6025344734612127e-05, + "loss": 0.0, + "step": 17641 + }, + { + "epoch": 4.08, + "learning_rate": 1.6023418842924277e-05, + "loss": 0.0, + "step": 17642 + }, + { + "epoch": 4.08, + "learning_rate": 1.602149295123642e-05, + "loss": 0.0, + "step": 17643 + }, + { + "epoch": 4.08, + "learning_rate": 1.6019567059548572e-05, + "loss": 0.0, + "step": 17644 + }, + { + "epoch": 4.08, + "learning_rate": 1.601764116786072e-05, + "loss": 0.0, + "step": 17645 + }, + { + "epoch": 4.08, + "learning_rate": 1.6015715276172867e-05, + "loss": 0.0, + "step": 17646 + }, + { + "epoch": 4.08, + "learning_rate": 1.6013789384485018e-05, + "loss": 0.0, + "step": 17647 + }, + { + "epoch": 4.08, + "learning_rate": 1.6011863492797166e-05, + "loss": 0.0001, + "step": 17648 + }, + { + "epoch": 4.08, + "learning_rate": 1.6009937601109317e-05, + "loss": 0.0001, + "step": 17649 + }, + { + "epoch": 4.08, + "learning_rate": 1.6008011709421464e-05, + "loss": 0.0, + "step": 17650 + }, + { + "epoch": 4.08, + "learning_rate": 1.600608581773361e-05, + "loss": 0.0, + "step": 17651 + }, + { + "epoch": 4.08, + "learning_rate": 1.600415992604576e-05, + "loss": 0.0, + "step": 17652 + }, + { + "epoch": 4.08, + "learning_rate": 1.6002234034357906e-05, + "loss": 0.0, + "step": 17653 + }, + { + "epoch": 4.08, + "learning_rate": 1.6000308142670057e-05, + "loss": 0.0004, + "step": 17654 + }, + { + "epoch": 4.08, + "learning_rate": 1.5998382250982205e-05, + "loss": 0.0, + "step": 17655 + }, + { + "epoch": 4.08, + "learning_rate": 1.5996456359294352e-05, + "loss": 0.0, + "step": 17656 + }, + { + "epoch": 4.08, + "learning_rate": 1.5994530467606503e-05, + "loss": 0.0002, + "step": 17657 + }, + { + "epoch": 4.08, + "learning_rate": 1.599260457591865e-05, + "loss": 0.0, + "step": 17658 + }, + { + "epoch": 4.08, + "learning_rate": 1.59906786842308e-05, + "loss": 0.0, + "step": 17659 + }, + { + "epoch": 4.08, + "learning_rate": 1.598875279254295e-05, + "loss": 0.0, + "step": 17660 + }, + { + "epoch": 4.08, + "learning_rate": 1.5986826900855096e-05, + "loss": 0.0, + "step": 17661 + }, + { + "epoch": 4.08, + "learning_rate": 1.5984901009167244e-05, + "loss": 0.0, + "step": 17662 + }, + { + "epoch": 4.08, + "learning_rate": 1.598297511747939e-05, + "loss": 0.0, + "step": 17663 + }, + { + "epoch": 4.08, + "learning_rate": 1.5981049225791542e-05, + "loss": 0.0001, + "step": 17664 + }, + { + "epoch": 4.08, + "learning_rate": 1.597912333410369e-05, + "loss": 0.0, + "step": 17665 + }, + { + "epoch": 4.08, + "learning_rate": 1.597719744241584e-05, + "loss": 0.0, + "step": 17666 + }, + { + "epoch": 4.08, + "learning_rate": 1.5975271550727988e-05, + "loss": 0.0001, + "step": 17667 + }, + { + "epoch": 4.08, + "learning_rate": 1.5973345659040136e-05, + "loss": 0.0, + "step": 17668 + }, + { + "epoch": 4.08, + "learning_rate": 1.5971419767352286e-05, + "loss": 0.0, + "step": 17669 + }, + { + "epoch": 4.08, + "learning_rate": 1.5969493875664434e-05, + "loss": 0.0, + "step": 17670 + }, + { + "epoch": 4.08, + "learning_rate": 1.596756798397658e-05, + "loss": 0.0, + "step": 17671 + }, + { + "epoch": 4.08, + "learning_rate": 1.596564209228873e-05, + "loss": 0.0, + "step": 17672 + }, + { + "epoch": 4.08, + "learning_rate": 1.596371620060088e-05, + "loss": 0.0, + "step": 17673 + }, + { + "epoch": 4.08, + "learning_rate": 1.5961790308913027e-05, + "loss": 0.0001, + "step": 17674 + }, + { + "epoch": 4.08, + "learning_rate": 1.5959864417225175e-05, + "loss": 0.0, + "step": 17675 + }, + { + "epoch": 4.09, + "learning_rate": 1.5957938525537326e-05, + "loss": 0.0, + "step": 17676 + }, + { + "epoch": 4.09, + "learning_rate": 1.5956012633849473e-05, + "loss": 0.0, + "step": 17677 + }, + { + "epoch": 4.09, + "learning_rate": 1.5954086742161624e-05, + "loss": 0.0, + "step": 17678 + }, + { + "epoch": 4.09, + "learning_rate": 1.595216085047377e-05, + "loss": 0.0, + "step": 17679 + }, + { + "epoch": 4.09, + "learning_rate": 1.595023495878592e-05, + "loss": 0.0, + "step": 17680 + }, + { + "epoch": 4.09, + "learning_rate": 1.5948309067098066e-05, + "loss": 0.0, + "step": 17681 + }, + { + "epoch": 4.09, + "learning_rate": 1.5946383175410214e-05, + "loss": 0.0, + "step": 17682 + }, + { + "epoch": 4.09, + "learning_rate": 1.5944457283722365e-05, + "loss": 0.0, + "step": 17683 + }, + { + "epoch": 4.09, + "learning_rate": 1.5942531392034512e-05, + "loss": 0.0, + "step": 17684 + }, + { + "epoch": 4.09, + "learning_rate": 1.5940605500346663e-05, + "loss": 0.0, + "step": 17685 + }, + { + "epoch": 4.09, + "learning_rate": 1.593867960865881e-05, + "loss": 0.0002, + "step": 17686 + }, + { + "epoch": 4.09, + "learning_rate": 1.5936753716970958e-05, + "loss": 0.0, + "step": 17687 + }, + { + "epoch": 4.09, + "learning_rate": 1.593482782528311e-05, + "loss": 0.0, + "step": 17688 + }, + { + "epoch": 4.09, + "learning_rate": 1.5932901933595256e-05, + "loss": 0.0, + "step": 17689 + }, + { + "epoch": 4.09, + "learning_rate": 1.5930976041907404e-05, + "loss": 0.0, + "step": 17690 + }, + { + "epoch": 4.09, + "learning_rate": 1.592905015021955e-05, + "loss": 0.0, + "step": 17691 + }, + { + "epoch": 4.09, + "learning_rate": 1.5927124258531702e-05, + "loss": 0.0, + "step": 17692 + }, + { + "epoch": 4.09, + "learning_rate": 1.592519836684385e-05, + "loss": 0.0, + "step": 17693 + }, + { + "epoch": 4.09, + "learning_rate": 1.5923272475155997e-05, + "loss": 0.0, + "step": 17694 + }, + { + "epoch": 4.09, + "learning_rate": 1.5921346583468148e-05, + "loss": 0.0, + "step": 17695 + }, + { + "epoch": 4.09, + "learning_rate": 1.5919420691780295e-05, + "loss": 0.0, + "step": 17696 + }, + { + "epoch": 4.09, + "learning_rate": 1.5917494800092443e-05, + "loss": 0.0, + "step": 17697 + }, + { + "epoch": 4.09, + "learning_rate": 1.5915568908404594e-05, + "loss": 0.0, + "step": 17698 + }, + { + "epoch": 4.09, + "learning_rate": 1.5913643016716738e-05, + "loss": 0.0, + "step": 17699 + }, + { + "epoch": 4.09, + "learning_rate": 1.591171712502889e-05, + "loss": 0.0, + "step": 17700 + }, + { + "epoch": 4.09, + "learning_rate": 1.5909791233341036e-05, + "loss": 0.0, + "step": 17701 + }, + { + "epoch": 4.09, + "learning_rate": 1.5907865341653187e-05, + "loss": 0.0, + "step": 17702 + }, + { + "epoch": 4.09, + "learning_rate": 1.5905939449965335e-05, + "loss": 0.0, + "step": 17703 + }, + { + "epoch": 4.09, + "learning_rate": 1.5904013558277482e-05, + "loss": 0.0, + "step": 17704 + }, + { + "epoch": 4.09, + "learning_rate": 1.5902087666589633e-05, + "loss": 0.0, + "step": 17705 + }, + { + "epoch": 4.09, + "learning_rate": 1.590016177490178e-05, + "loss": 0.0, + "step": 17706 + }, + { + "epoch": 4.09, + "learning_rate": 1.589823588321393e-05, + "loss": 0.0, + "step": 17707 + }, + { + "epoch": 4.09, + "learning_rate": 1.5896309991526075e-05, + "loss": 0.0, + "step": 17708 + }, + { + "epoch": 4.09, + "learning_rate": 1.5894384099838226e-05, + "loss": 0.0, + "step": 17709 + }, + { + "epoch": 4.09, + "learning_rate": 1.5892458208150374e-05, + "loss": 0.0, + "step": 17710 + }, + { + "epoch": 4.09, + "learning_rate": 1.589053231646252e-05, + "loss": 0.0, + "step": 17711 + }, + { + "epoch": 4.09, + "learning_rate": 1.5888606424774672e-05, + "loss": 0.0, + "step": 17712 + }, + { + "epoch": 4.09, + "learning_rate": 1.588668053308682e-05, + "loss": 0.0, + "step": 17713 + }, + { + "epoch": 4.09, + "learning_rate": 1.588475464139897e-05, + "loss": 0.0, + "step": 17714 + }, + { + "epoch": 4.09, + "learning_rate": 1.5882828749711118e-05, + "loss": 0.0, + "step": 17715 + }, + { + "epoch": 4.09, + "learning_rate": 1.5880902858023265e-05, + "loss": 0.0, + "step": 17716 + }, + { + "epoch": 4.09, + "learning_rate": 1.5878976966335416e-05, + "loss": 0.0011, + "step": 17717 + }, + { + "epoch": 4.09, + "learning_rate": 1.587705107464756e-05, + "loss": 0.0, + "step": 17718 + }, + { + "epoch": 4.09, + "learning_rate": 1.587512518295971e-05, + "loss": 0.0, + "step": 17719 + }, + { + "epoch": 4.1, + "learning_rate": 1.587319929127186e-05, + "loss": 0.0, + "step": 17720 + }, + { + "epoch": 4.1, + "learning_rate": 1.587127339958401e-05, + "loss": 0.0, + "step": 17721 + }, + { + "epoch": 4.1, + "learning_rate": 1.5869347507896157e-05, + "loss": 0.0, + "step": 17722 + }, + { + "epoch": 4.1, + "learning_rate": 1.5867421616208304e-05, + "loss": 0.0, + "step": 17723 + }, + { + "epoch": 4.1, + "learning_rate": 1.5865495724520455e-05, + "loss": 0.0, + "step": 17724 + }, + { + "epoch": 4.1, + "learning_rate": 1.5863569832832603e-05, + "loss": 0.0, + "step": 17725 + }, + { + "epoch": 4.1, + "learning_rate": 1.5861643941144754e-05, + "loss": 0.0001, + "step": 17726 + }, + { + "epoch": 4.1, + "learning_rate": 1.5859718049456898e-05, + "loss": 0.0, + "step": 17727 + }, + { + "epoch": 4.1, + "learning_rate": 1.585779215776905e-05, + "loss": 0.0, + "step": 17728 + }, + { + "epoch": 4.1, + "learning_rate": 1.5855866266081196e-05, + "loss": 0.0, + "step": 17729 + }, + { + "epoch": 4.1, + "learning_rate": 1.5853940374393344e-05, + "loss": 0.0, + "step": 17730 + }, + { + "epoch": 4.1, + "learning_rate": 1.5852014482705494e-05, + "loss": 0.0, + "step": 17731 + }, + { + "epoch": 4.1, + "learning_rate": 1.5850088591017642e-05, + "loss": 0.0, + "step": 17732 + }, + { + "epoch": 4.1, + "learning_rate": 1.584816269932979e-05, + "loss": 0.0, + "step": 17733 + }, + { + "epoch": 4.1, + "learning_rate": 1.584623680764194e-05, + "loss": 0.0, + "step": 17734 + }, + { + "epoch": 4.1, + "learning_rate": 1.5844310915954088e-05, + "loss": 0.0001, + "step": 17735 + }, + { + "epoch": 4.1, + "learning_rate": 1.5842385024266235e-05, + "loss": 0.0, + "step": 17736 + }, + { + "epoch": 4.1, + "learning_rate": 1.5840459132578383e-05, + "loss": 0.0, + "step": 17737 + }, + { + "epoch": 4.1, + "learning_rate": 1.5838533240890534e-05, + "loss": 0.0, + "step": 17738 + }, + { + "epoch": 4.1, + "learning_rate": 1.583660734920268e-05, + "loss": 0.0, + "step": 17739 + }, + { + "epoch": 4.1, + "learning_rate": 1.583468145751483e-05, + "loss": 0.0, + "step": 17740 + }, + { + "epoch": 4.1, + "learning_rate": 1.583275556582698e-05, + "loss": 0.0, + "step": 17741 + }, + { + "epoch": 4.1, + "learning_rate": 1.5830829674139127e-05, + "loss": 0.0, + "step": 17742 + }, + { + "epoch": 4.1, + "learning_rate": 1.5828903782451278e-05, + "loss": 0.0, + "step": 17743 + }, + { + "epoch": 4.1, + "learning_rate": 1.5826977890763425e-05, + "loss": 0.0, + "step": 17744 + }, + { + "epoch": 4.1, + "learning_rate": 1.5825051999075573e-05, + "loss": 0.0, + "step": 17745 + }, + { + "epoch": 4.1, + "learning_rate": 1.582312610738772e-05, + "loss": 0.0, + "step": 17746 + }, + { + "epoch": 4.1, + "learning_rate": 1.5821200215699868e-05, + "loss": 0.0, + "step": 17747 + }, + { + "epoch": 4.1, + "learning_rate": 1.581927432401202e-05, + "loss": 0.0, + "step": 17748 + }, + { + "epoch": 4.1, + "learning_rate": 1.5817348432324166e-05, + "loss": 0.0, + "step": 17749 + }, + { + "epoch": 4.1, + "learning_rate": 1.5815422540636317e-05, + "loss": 0.0, + "step": 17750 + }, + { + "epoch": 4.1, + "learning_rate": 1.5813496648948464e-05, + "loss": 0.0, + "step": 17751 + }, + { + "epoch": 4.1, + "learning_rate": 1.5811570757260612e-05, + "loss": 0.0, + "step": 17752 + }, + { + "epoch": 4.1, + "learning_rate": 1.5809644865572763e-05, + "loss": 0.0, + "step": 17753 + }, + { + "epoch": 4.1, + "learning_rate": 1.580771897388491e-05, + "loss": 0.0, + "step": 17754 + }, + { + "epoch": 4.1, + "learning_rate": 1.5805793082197058e-05, + "loss": 0.0, + "step": 17755 + }, + { + "epoch": 4.1, + "learning_rate": 1.5803867190509205e-05, + "loss": 0.0, + "step": 17756 + }, + { + "epoch": 4.1, + "learning_rate": 1.5801941298821356e-05, + "loss": 0.0, + "step": 17757 + }, + { + "epoch": 4.1, + "learning_rate": 1.5800015407133503e-05, + "loss": 0.0, + "step": 17758 + }, + { + "epoch": 4.1, + "learning_rate": 1.579808951544565e-05, + "loss": 0.0, + "step": 17759 + }, + { + "epoch": 4.1, + "learning_rate": 1.5796163623757802e-05, + "loss": 0.0001, + "step": 17760 + }, + { + "epoch": 4.1, + "learning_rate": 1.579423773206995e-05, + "loss": 0.0, + "step": 17761 + }, + { + "epoch": 4.1, + "learning_rate": 1.57923118403821e-05, + "loss": 0.0, + "step": 17762 + }, + { + "epoch": 4.11, + "learning_rate": 1.5790385948694248e-05, + "loss": 0.0, + "step": 17763 + }, + { + "epoch": 4.11, + "learning_rate": 1.5788460057006395e-05, + "loss": 0.0, + "step": 17764 + }, + { + "epoch": 4.11, + "learning_rate": 1.5786534165318543e-05, + "loss": 0.0001, + "step": 17765 + }, + { + "epoch": 4.11, + "learning_rate": 1.578460827363069e-05, + "loss": 0.0, + "step": 17766 + }, + { + "epoch": 4.11, + "learning_rate": 1.578268238194284e-05, + "loss": 0.0, + "step": 17767 + }, + { + "epoch": 4.11, + "learning_rate": 1.578075649025499e-05, + "loss": 0.0, + "step": 17768 + }, + { + "epoch": 4.11, + "learning_rate": 1.5778830598567136e-05, + "loss": 0.0, + "step": 17769 + }, + { + "epoch": 4.11, + "learning_rate": 1.5776904706879287e-05, + "loss": 0.0, + "step": 17770 + }, + { + "epoch": 4.11, + "learning_rate": 1.5774978815191434e-05, + "loss": 0.0, + "step": 17771 + }, + { + "epoch": 4.11, + "learning_rate": 1.5773052923503585e-05, + "loss": 0.0, + "step": 17772 + }, + { + "epoch": 4.11, + "learning_rate": 1.5771127031815733e-05, + "loss": 0.0, + "step": 17773 + }, + { + "epoch": 4.11, + "learning_rate": 1.576920114012788e-05, + "loss": 0.0, + "step": 17774 + }, + { + "epoch": 4.11, + "learning_rate": 1.5767275248440027e-05, + "loss": 0.0, + "step": 17775 + }, + { + "epoch": 4.11, + "learning_rate": 1.5765349356752175e-05, + "loss": 0.0, + "step": 17776 + }, + { + "epoch": 4.11, + "learning_rate": 1.5763423465064326e-05, + "loss": 0.0, + "step": 17777 + }, + { + "epoch": 4.11, + "learning_rate": 1.5761497573376473e-05, + "loss": 0.0, + "step": 17778 + }, + { + "epoch": 4.11, + "learning_rate": 1.5759571681688624e-05, + "loss": 0.0, + "step": 17779 + }, + { + "epoch": 4.11, + "learning_rate": 1.575764579000077e-05, + "loss": 0.0, + "step": 17780 + }, + { + "epoch": 4.11, + "learning_rate": 1.575571989831292e-05, + "loss": 0.0, + "step": 17781 + }, + { + "epoch": 4.11, + "learning_rate": 1.575379400662507e-05, + "loss": 0.0, + "step": 17782 + }, + { + "epoch": 4.11, + "learning_rate": 1.5751868114937214e-05, + "loss": 0.0, + "step": 17783 + }, + { + "epoch": 4.11, + "learning_rate": 1.5749942223249365e-05, + "loss": 0.0, + "step": 17784 + }, + { + "epoch": 4.11, + "learning_rate": 1.5748016331561512e-05, + "loss": 0.0, + "step": 17785 + }, + { + "epoch": 4.11, + "learning_rate": 1.5746090439873663e-05, + "loss": 0.001, + "step": 17786 + }, + { + "epoch": 4.11, + "learning_rate": 1.574416454818581e-05, + "loss": 0.0, + "step": 17787 + }, + { + "epoch": 4.11, + "learning_rate": 1.5742238656497958e-05, + "loss": 0.0, + "step": 17788 + }, + { + "epoch": 4.11, + "learning_rate": 1.574031276481011e-05, + "loss": 0.0001, + "step": 17789 + }, + { + "epoch": 4.11, + "learning_rate": 1.5738386873122257e-05, + "loss": 0.0, + "step": 17790 + }, + { + "epoch": 4.11, + "learning_rate": 1.5736460981434407e-05, + "loss": 0.0, + "step": 17791 + }, + { + "epoch": 4.11, + "learning_rate": 1.573453508974655e-05, + "loss": 0.0001, + "step": 17792 + }, + { + "epoch": 4.11, + "learning_rate": 1.5732609198058702e-05, + "loss": 0.0, + "step": 17793 + }, + { + "epoch": 4.11, + "learning_rate": 1.573068330637085e-05, + "loss": 0.0, + "step": 17794 + }, + { + "epoch": 4.11, + "learning_rate": 1.5728757414682997e-05, + "loss": 0.0, + "step": 17795 + }, + { + "epoch": 4.11, + "learning_rate": 1.5726831522995148e-05, + "loss": 0.0, + "step": 17796 + }, + { + "epoch": 4.11, + "learning_rate": 1.5724905631307296e-05, + "loss": 0.0, + "step": 17797 + }, + { + "epoch": 4.11, + "learning_rate": 1.5722979739619447e-05, + "loss": 0.0, + "step": 17798 + }, + { + "epoch": 4.11, + "learning_rate": 1.5721053847931594e-05, + "loss": 0.0, + "step": 17799 + }, + { + "epoch": 4.11, + "learning_rate": 1.571912795624374e-05, + "loss": 0.0, + "step": 17800 + }, + { + "epoch": 4.11, + "learning_rate": 1.5717202064555892e-05, + "loss": 0.0, + "step": 17801 + }, + { + "epoch": 4.11, + "learning_rate": 1.5715276172868036e-05, + "loss": 0.0002, + "step": 17802 + }, + { + "epoch": 4.11, + "learning_rate": 1.5713350281180187e-05, + "loss": 0.0, + "step": 17803 + }, + { + "epoch": 4.11, + "learning_rate": 1.5711424389492335e-05, + "loss": 0.0, + "step": 17804 + }, + { + "epoch": 4.11, + "learning_rate": 1.5709498497804486e-05, + "loss": 0.0, + "step": 17805 + }, + { + "epoch": 4.12, + "learning_rate": 1.5707572606116633e-05, + "loss": 0.0, + "step": 17806 + }, + { + "epoch": 4.12, + "learning_rate": 1.570564671442878e-05, + "loss": 0.0, + "step": 17807 + }, + { + "epoch": 4.12, + "learning_rate": 1.570372082274093e-05, + "loss": 0.0, + "step": 17808 + }, + { + "epoch": 4.12, + "learning_rate": 1.570179493105308e-05, + "loss": 0.0, + "step": 17809 + }, + { + "epoch": 4.12, + "learning_rate": 1.5699869039365226e-05, + "loss": 0.0002, + "step": 17810 + }, + { + "epoch": 4.12, + "learning_rate": 1.5697943147677374e-05, + "loss": 0.0, + "step": 17811 + }, + { + "epoch": 4.12, + "learning_rate": 1.569601725598952e-05, + "loss": 0.0, + "step": 17812 + }, + { + "epoch": 4.12, + "learning_rate": 1.5694091364301672e-05, + "loss": 0.0, + "step": 17813 + }, + { + "epoch": 4.12, + "learning_rate": 1.569216547261382e-05, + "loss": 0.0, + "step": 17814 + }, + { + "epoch": 4.12, + "learning_rate": 1.569023958092597e-05, + "loss": 0.0, + "step": 17815 + }, + { + "epoch": 4.12, + "learning_rate": 1.5688313689238118e-05, + "loss": 0.0, + "step": 17816 + }, + { + "epoch": 4.12, + "learning_rate": 1.5686387797550266e-05, + "loss": 0.0, + "step": 17817 + }, + { + "epoch": 4.12, + "learning_rate": 1.5684461905862416e-05, + "loss": 0.0001, + "step": 17818 + }, + { + "epoch": 4.12, + "learning_rate": 1.5682536014174564e-05, + "loss": 0.0001, + "step": 17819 + }, + { + "epoch": 4.12, + "learning_rate": 1.568061012248671e-05, + "loss": 0.0, + "step": 17820 + }, + { + "epoch": 4.12, + "learning_rate": 1.567868423079886e-05, + "loss": 0.0, + "step": 17821 + }, + { + "epoch": 4.12, + "learning_rate": 1.567675833911101e-05, + "loss": 0.0, + "step": 17822 + }, + { + "epoch": 4.12, + "learning_rate": 1.5674832447423157e-05, + "loss": 0.0, + "step": 17823 + }, + { + "epoch": 4.12, + "learning_rate": 1.5672906555735305e-05, + "loss": 0.0, + "step": 17824 + }, + { + "epoch": 4.12, + "learning_rate": 1.5670980664047456e-05, + "loss": 0.0, + "step": 17825 + }, + { + "epoch": 4.12, + "learning_rate": 1.5669054772359603e-05, + "loss": 0.0, + "step": 17826 + }, + { + "epoch": 4.12, + "learning_rate": 1.5667128880671754e-05, + "loss": 0.0, + "step": 17827 + }, + { + "epoch": 4.12, + "learning_rate": 1.56652029889839e-05, + "loss": 0.0, + "step": 17828 + }, + { + "epoch": 4.12, + "learning_rate": 1.566327709729605e-05, + "loss": 0.0, + "step": 17829 + }, + { + "epoch": 4.12, + "learning_rate": 1.5661351205608196e-05, + "loss": 0.0, + "step": 17830 + }, + { + "epoch": 4.12, + "learning_rate": 1.5659425313920344e-05, + "loss": 0.0, + "step": 17831 + }, + { + "epoch": 4.12, + "learning_rate": 1.5657499422232495e-05, + "loss": 0.0, + "step": 17832 + }, + { + "epoch": 4.12, + "learning_rate": 1.5655573530544642e-05, + "loss": 0.0, + "step": 17833 + }, + { + "epoch": 4.12, + "learning_rate": 1.5653647638856793e-05, + "loss": 0.0, + "step": 17834 + }, + { + "epoch": 4.12, + "learning_rate": 1.565172174716894e-05, + "loss": 0.0, + "step": 17835 + }, + { + "epoch": 4.12, + "learning_rate": 1.5649795855481088e-05, + "loss": 0.0, + "step": 17836 + }, + { + "epoch": 4.12, + "learning_rate": 1.564786996379324e-05, + "loss": 0.0, + "step": 17837 + }, + { + "epoch": 4.12, + "learning_rate": 1.5645944072105386e-05, + "loss": 0.0, + "step": 17838 + }, + { + "epoch": 4.12, + "learning_rate": 1.5644018180417534e-05, + "loss": 0.0, + "step": 17839 + }, + { + "epoch": 4.12, + "learning_rate": 1.564209228872968e-05, + "loss": 0.0, + "step": 17840 + }, + { + "epoch": 4.12, + "learning_rate": 1.5640166397041832e-05, + "loss": 0.0, + "step": 17841 + }, + { + "epoch": 4.12, + "learning_rate": 1.563824050535398e-05, + "loss": 0.0, + "step": 17842 + }, + { + "epoch": 4.12, + "learning_rate": 1.5636314613666127e-05, + "loss": 0.0, + "step": 17843 + }, + { + "epoch": 4.12, + "learning_rate": 1.5634388721978278e-05, + "loss": 0.0, + "step": 17844 + }, + { + "epoch": 4.12, + "learning_rate": 1.5632462830290425e-05, + "loss": 0.0, + "step": 17845 + }, + { + "epoch": 4.12, + "learning_rate": 1.5630536938602573e-05, + "loss": 0.0, + "step": 17846 + }, + { + "epoch": 4.12, + "learning_rate": 1.5628611046914724e-05, + "loss": 0.0, + "step": 17847 + }, + { + "epoch": 4.12, + "learning_rate": 1.5626685155226868e-05, + "loss": 0.0, + "step": 17848 + }, + { + "epoch": 4.13, + "learning_rate": 1.562475926353902e-05, + "loss": 0.0, + "step": 17849 + }, + { + "epoch": 4.13, + "learning_rate": 1.5622833371851166e-05, + "loss": 0.0, + "step": 17850 + }, + { + "epoch": 4.13, + "learning_rate": 1.5620907480163317e-05, + "loss": 0.0, + "step": 17851 + }, + { + "epoch": 4.13, + "learning_rate": 1.5618981588475465e-05, + "loss": 0.0, + "step": 17852 + }, + { + "epoch": 4.13, + "learning_rate": 1.5617055696787612e-05, + "loss": 0.0002, + "step": 17853 + }, + { + "epoch": 4.13, + "learning_rate": 1.5615129805099763e-05, + "loss": 0.0, + "step": 17854 + }, + { + "epoch": 4.13, + "learning_rate": 1.561320391341191e-05, + "loss": 0.0, + "step": 17855 + }, + { + "epoch": 4.13, + "learning_rate": 1.561127802172406e-05, + "loss": 0.0, + "step": 17856 + }, + { + "epoch": 4.13, + "learning_rate": 1.560935213003621e-05, + "loss": 0.0, + "step": 17857 + }, + { + "epoch": 4.13, + "learning_rate": 1.5607426238348356e-05, + "loss": 0.0, + "step": 17858 + }, + { + "epoch": 4.13, + "learning_rate": 1.5605500346660504e-05, + "loss": 0.0, + "step": 17859 + }, + { + "epoch": 4.13, + "learning_rate": 1.560357445497265e-05, + "loss": 0.0, + "step": 17860 + }, + { + "epoch": 4.13, + "learning_rate": 1.5601648563284802e-05, + "loss": 0.0, + "step": 17861 + }, + { + "epoch": 4.13, + "learning_rate": 1.559972267159695e-05, + "loss": 0.0, + "step": 17862 + }, + { + "epoch": 4.13, + "learning_rate": 1.55977967799091e-05, + "loss": 0.0, + "step": 17863 + }, + { + "epoch": 4.13, + "learning_rate": 1.5595870888221248e-05, + "loss": 0.0, + "step": 17864 + }, + { + "epoch": 4.13, + "learning_rate": 1.5593944996533395e-05, + "loss": 0.0, + "step": 17865 + }, + { + "epoch": 4.13, + "learning_rate": 1.5592019104845546e-05, + "loss": 0.0, + "step": 17866 + }, + { + "epoch": 4.13, + "learning_rate": 1.559009321315769e-05, + "loss": 0.0, + "step": 17867 + }, + { + "epoch": 4.13, + "learning_rate": 1.558816732146984e-05, + "loss": 0.0, + "step": 17868 + }, + { + "epoch": 4.13, + "learning_rate": 1.558624142978199e-05, + "loss": 0.0, + "step": 17869 + }, + { + "epoch": 4.13, + "learning_rate": 1.558431553809414e-05, + "loss": 0.0, + "step": 17870 + }, + { + "epoch": 4.13, + "learning_rate": 1.5582389646406287e-05, + "loss": 0.0, + "step": 17871 + }, + { + "epoch": 4.13, + "learning_rate": 1.5580463754718434e-05, + "loss": 0.0, + "step": 17872 + }, + { + "epoch": 4.13, + "learning_rate": 1.5578537863030585e-05, + "loss": 0.0, + "step": 17873 + }, + { + "epoch": 4.13, + "learning_rate": 1.5576611971342733e-05, + "loss": 0.0, + "step": 17874 + }, + { + "epoch": 4.13, + "learning_rate": 1.5574686079654884e-05, + "loss": 0.0, + "step": 17875 + }, + { + "epoch": 4.13, + "learning_rate": 1.5572760187967028e-05, + "loss": 0.0, + "step": 17876 + }, + { + "epoch": 4.13, + "learning_rate": 1.557083429627918e-05, + "loss": 0.0, + "step": 17877 + }, + { + "epoch": 4.13, + "learning_rate": 1.5568908404591326e-05, + "loss": 0.0, + "step": 17878 + }, + { + "epoch": 4.13, + "learning_rate": 1.5566982512903474e-05, + "loss": 0.0, + "step": 17879 + }, + { + "epoch": 4.13, + "learning_rate": 1.5565056621215624e-05, + "loss": 0.0, + "step": 17880 + }, + { + "epoch": 4.13, + "learning_rate": 1.5563130729527772e-05, + "loss": 0.0, + "step": 17881 + }, + { + "epoch": 4.13, + "learning_rate": 1.5561204837839923e-05, + "loss": 0.0, + "step": 17882 + }, + { + "epoch": 4.13, + "learning_rate": 1.555927894615207e-05, + "loss": 0.0, + "step": 17883 + }, + { + "epoch": 4.13, + "learning_rate": 1.5557353054464218e-05, + "loss": 0.0, + "step": 17884 + }, + { + "epoch": 4.13, + "learning_rate": 1.555542716277637e-05, + "loss": 0.0, + "step": 17885 + }, + { + "epoch": 4.13, + "learning_rate": 1.5553501271088513e-05, + "loss": 0.0, + "step": 17886 + }, + { + "epoch": 4.13, + "learning_rate": 1.5551575379400664e-05, + "loss": 0.0, + "step": 17887 + }, + { + "epoch": 4.13, + "learning_rate": 1.554964948771281e-05, + "loss": 0.0, + "step": 17888 + }, + { + "epoch": 4.13, + "learning_rate": 1.554772359602496e-05, + "loss": 0.0, + "step": 17889 + }, + { + "epoch": 4.13, + "learning_rate": 1.554579770433711e-05, + "loss": 0.0, + "step": 17890 + }, + { + "epoch": 4.13, + "learning_rate": 1.5543871812649257e-05, + "loss": 0.0, + "step": 17891 + }, + { + "epoch": 4.13, + "learning_rate": 1.5541945920961408e-05, + "loss": 0.0, + "step": 17892 + }, + { + "epoch": 4.14, + "learning_rate": 1.5540020029273555e-05, + "loss": 0.0, + "step": 17893 + }, + { + "epoch": 4.14, + "learning_rate": 1.5538094137585703e-05, + "loss": 0.0, + "step": 17894 + }, + { + "epoch": 4.14, + "learning_rate": 1.553616824589785e-05, + "loss": 0.0, + "step": 17895 + }, + { + "epoch": 4.14, + "learning_rate": 1.5534242354209998e-05, + "loss": 0.0, + "step": 17896 + }, + { + "epoch": 4.14, + "learning_rate": 1.553231646252215e-05, + "loss": 0.0, + "step": 17897 + }, + { + "epoch": 4.14, + "learning_rate": 1.5530390570834296e-05, + "loss": 0.0, + "step": 17898 + }, + { + "epoch": 4.14, + "learning_rate": 1.5528464679146447e-05, + "loss": 0.0, + "step": 17899 + }, + { + "epoch": 4.14, + "learning_rate": 1.5526538787458594e-05, + "loss": 0.0, + "step": 17900 + }, + { + "epoch": 4.14, + "learning_rate": 1.5524612895770742e-05, + "loss": 0.0, + "step": 17901 + }, + { + "epoch": 4.14, + "learning_rate": 1.5522687004082893e-05, + "loss": 0.0, + "step": 17902 + }, + { + "epoch": 4.14, + "learning_rate": 1.552076111239504e-05, + "loss": 0.0, + "step": 17903 + }, + { + "epoch": 4.14, + "learning_rate": 1.5518835220707188e-05, + "loss": 0.0, + "step": 17904 + }, + { + "epoch": 4.14, + "learning_rate": 1.5516909329019335e-05, + "loss": 0.0, + "step": 17905 + }, + { + "epoch": 4.14, + "learning_rate": 1.5514983437331486e-05, + "loss": 0.0, + "step": 17906 + }, + { + "epoch": 4.14, + "learning_rate": 1.5513057545643633e-05, + "loss": 0.0, + "step": 17907 + }, + { + "epoch": 4.14, + "learning_rate": 1.551113165395578e-05, + "loss": 0.0, + "step": 17908 + }, + { + "epoch": 4.14, + "learning_rate": 1.5509205762267932e-05, + "loss": 0.0, + "step": 17909 + }, + { + "epoch": 4.14, + "learning_rate": 1.550727987058008e-05, + "loss": 0.0, + "step": 17910 + }, + { + "epoch": 4.14, + "learning_rate": 1.550535397889223e-05, + "loss": 0.0, + "step": 17911 + }, + { + "epoch": 4.14, + "learning_rate": 1.5503428087204378e-05, + "loss": 0.0, + "step": 17912 + }, + { + "epoch": 4.14, + "learning_rate": 1.5501502195516525e-05, + "loss": 0.0, + "step": 17913 + }, + { + "epoch": 4.14, + "learning_rate": 1.5499576303828673e-05, + "loss": 0.0, + "step": 17914 + }, + { + "epoch": 4.14, + "learning_rate": 1.549765041214082e-05, + "loss": 0.0, + "step": 17915 + }, + { + "epoch": 4.14, + "learning_rate": 1.549572452045297e-05, + "loss": 0.0, + "step": 17916 + }, + { + "epoch": 4.14, + "learning_rate": 1.549379862876512e-05, + "loss": 0.002, + "step": 17917 + }, + { + "epoch": 4.14, + "learning_rate": 1.549187273707727e-05, + "loss": 0.0, + "step": 17918 + }, + { + "epoch": 4.14, + "learning_rate": 1.5489946845389417e-05, + "loss": 0.0003, + "step": 17919 + }, + { + "epoch": 4.14, + "learning_rate": 1.5488020953701564e-05, + "loss": 0.0, + "step": 17920 + }, + { + "epoch": 4.14, + "learning_rate": 1.5486095062013715e-05, + "loss": 0.0, + "step": 17921 + }, + { + "epoch": 4.14, + "learning_rate": 1.5484169170325863e-05, + "loss": 0.0001, + "step": 17922 + }, + { + "epoch": 4.14, + "learning_rate": 1.548224327863801e-05, + "loss": 0.0, + "step": 17923 + }, + { + "epoch": 4.14, + "learning_rate": 1.5480317386950158e-05, + "loss": 0.0, + "step": 17924 + }, + { + "epoch": 4.14, + "learning_rate": 1.5478391495262305e-05, + "loss": 0.0, + "step": 17925 + }, + { + "epoch": 4.14, + "learning_rate": 1.5476465603574456e-05, + "loss": 0.0, + "step": 17926 + }, + { + "epoch": 4.14, + "learning_rate": 1.5474539711886603e-05, + "loss": 0.0, + "step": 17927 + }, + { + "epoch": 4.14, + "learning_rate": 1.5472613820198754e-05, + "loss": 0.0, + "step": 17928 + }, + { + "epoch": 4.14, + "learning_rate": 1.54706879285109e-05, + "loss": 0.0, + "step": 17929 + }, + { + "epoch": 4.14, + "learning_rate": 1.546876203682305e-05, + "loss": 0.0, + "step": 17930 + }, + { + "epoch": 4.14, + "learning_rate": 1.54668361451352e-05, + "loss": 0.0, + "step": 17931 + }, + { + "epoch": 4.14, + "learning_rate": 1.5464910253447344e-05, + "loss": 0.0, + "step": 17932 + }, + { + "epoch": 4.14, + "learning_rate": 1.5462984361759495e-05, + "loss": 0.0, + "step": 17933 + }, + { + "epoch": 4.14, + "learning_rate": 1.5461058470071642e-05, + "loss": 0.0, + "step": 17934 + }, + { + "epoch": 4.14, + "learning_rate": 1.5459132578383793e-05, + "loss": 0.0001, + "step": 17935 + }, + { + "epoch": 4.15, + "learning_rate": 1.545720668669594e-05, + "loss": 0.0, + "step": 17936 + }, + { + "epoch": 4.15, + "learning_rate": 1.5455280795008088e-05, + "loss": 0.0, + "step": 17937 + }, + { + "epoch": 4.15, + "learning_rate": 1.545335490332024e-05, + "loss": 0.0, + "step": 17938 + }, + { + "epoch": 4.15, + "learning_rate": 1.5451429011632387e-05, + "loss": 0.0, + "step": 17939 + }, + { + "epoch": 4.15, + "learning_rate": 1.5449503119944537e-05, + "loss": 0.0, + "step": 17940 + }, + { + "epoch": 4.15, + "learning_rate": 1.5447577228256685e-05, + "loss": 0.0, + "step": 17941 + }, + { + "epoch": 4.15, + "learning_rate": 1.5445651336568832e-05, + "loss": 0.0, + "step": 17942 + }, + { + "epoch": 4.15, + "learning_rate": 1.544372544488098e-05, + "loss": 0.0, + "step": 17943 + }, + { + "epoch": 4.15, + "learning_rate": 1.5441799553193127e-05, + "loss": 0.0, + "step": 17944 + }, + { + "epoch": 4.15, + "learning_rate": 1.5439873661505278e-05, + "loss": 0.0, + "step": 17945 + }, + { + "epoch": 4.15, + "learning_rate": 1.5437947769817426e-05, + "loss": 0.0, + "step": 17946 + }, + { + "epoch": 4.15, + "learning_rate": 1.5436021878129577e-05, + "loss": 0.0, + "step": 17947 + }, + { + "epoch": 4.15, + "learning_rate": 1.5434095986441724e-05, + "loss": 0.0, + "step": 17948 + }, + { + "epoch": 4.15, + "learning_rate": 1.543217009475387e-05, + "loss": 0.0, + "step": 17949 + }, + { + "epoch": 4.15, + "learning_rate": 1.5430244203066022e-05, + "loss": 0.0, + "step": 17950 + }, + { + "epoch": 4.15, + "learning_rate": 1.5428318311378167e-05, + "loss": 0.0, + "step": 17951 + }, + { + "epoch": 4.15, + "learning_rate": 1.5426392419690317e-05, + "loss": 0.0, + "step": 17952 + }, + { + "epoch": 4.15, + "learning_rate": 1.5424466528002465e-05, + "loss": 0.0, + "step": 17953 + }, + { + "epoch": 4.15, + "learning_rate": 1.5422540636314616e-05, + "loss": 0.0, + "step": 17954 + }, + { + "epoch": 4.15, + "learning_rate": 1.5420614744626763e-05, + "loss": 0.0, + "step": 17955 + }, + { + "epoch": 4.15, + "learning_rate": 1.541868885293891e-05, + "loss": 0.0, + "step": 17956 + }, + { + "epoch": 4.15, + "learning_rate": 1.541676296125106e-05, + "loss": 0.0, + "step": 17957 + }, + { + "epoch": 4.15, + "learning_rate": 1.541483706956321e-05, + "loss": 0.0, + "step": 17958 + }, + { + "epoch": 4.15, + "learning_rate": 1.541291117787536e-05, + "loss": 0.0, + "step": 17959 + }, + { + "epoch": 4.15, + "learning_rate": 1.5410985286187504e-05, + "loss": 0.0, + "step": 17960 + }, + { + "epoch": 4.15, + "learning_rate": 1.540905939449965e-05, + "loss": 0.0, + "step": 17961 + }, + { + "epoch": 4.15, + "learning_rate": 1.5407133502811802e-05, + "loss": 0.0, + "step": 17962 + }, + { + "epoch": 4.15, + "learning_rate": 1.540520761112395e-05, + "loss": 0.0, + "step": 17963 + }, + { + "epoch": 4.15, + "learning_rate": 1.54032817194361e-05, + "loss": 0.0, + "step": 17964 + }, + { + "epoch": 4.15, + "learning_rate": 1.5401355827748248e-05, + "loss": 0.0, + "step": 17965 + }, + { + "epoch": 4.15, + "learning_rate": 1.5399429936060396e-05, + "loss": 0.0, + "step": 17966 + }, + { + "epoch": 4.15, + "learning_rate": 1.5397504044372546e-05, + "loss": 0.0, + "step": 17967 + }, + { + "epoch": 4.15, + "learning_rate": 1.5395578152684694e-05, + "loss": 0.0, + "step": 17968 + }, + { + "epoch": 4.15, + "learning_rate": 1.5393652260996845e-05, + "loss": 0.0001, + "step": 17969 + }, + { + "epoch": 4.15, + "learning_rate": 1.539172636930899e-05, + "loss": 0.0, + "step": 17970 + }, + { + "epoch": 4.15, + "learning_rate": 1.538980047762114e-05, + "loss": 0.0, + "step": 17971 + }, + { + "epoch": 4.15, + "learning_rate": 1.5387874585933287e-05, + "loss": 0.0, + "step": 17972 + }, + { + "epoch": 4.15, + "learning_rate": 1.5385948694245435e-05, + "loss": 0.0, + "step": 17973 + }, + { + "epoch": 4.15, + "learning_rate": 1.5384022802557586e-05, + "loss": 0.0, + "step": 17974 + }, + { + "epoch": 4.15, + "learning_rate": 1.5382096910869733e-05, + "loss": 0.0, + "step": 17975 + }, + { + "epoch": 4.15, + "learning_rate": 1.5380171019181884e-05, + "loss": 0.0, + "step": 17976 + }, + { + "epoch": 4.15, + "learning_rate": 1.537824512749403e-05, + "loss": 0.0, + "step": 17977 + }, + { + "epoch": 4.15, + "learning_rate": 1.537631923580618e-05, + "loss": 0.0, + "step": 17978 + }, + { + "epoch": 4.16, + "learning_rate": 1.5374393344118326e-05, + "loss": 0.0, + "step": 17979 + }, + { + "epoch": 4.16, + "learning_rate": 1.5372467452430474e-05, + "loss": 0.0, + "step": 17980 + }, + { + "epoch": 4.16, + "learning_rate": 1.5370541560742625e-05, + "loss": 0.0, + "step": 17981 + }, + { + "epoch": 4.16, + "learning_rate": 1.5368615669054772e-05, + "loss": 0.0, + "step": 17982 + }, + { + "epoch": 4.16, + "learning_rate": 1.5366689777366923e-05, + "loss": 0.0, + "step": 17983 + }, + { + "epoch": 4.16, + "learning_rate": 1.536476388567907e-05, + "loss": 0.0, + "step": 17984 + }, + { + "epoch": 4.16, + "learning_rate": 1.5362837993991218e-05, + "loss": 0.0, + "step": 17985 + }, + { + "epoch": 4.16, + "learning_rate": 1.536091210230337e-05, + "loss": 0.0, + "step": 17986 + }, + { + "epoch": 4.16, + "learning_rate": 1.5358986210615516e-05, + "loss": 0.0, + "step": 17987 + }, + { + "epoch": 4.16, + "learning_rate": 1.5357060318927664e-05, + "loss": 0.0, + "step": 17988 + }, + { + "epoch": 4.16, + "learning_rate": 1.535513442723981e-05, + "loss": 0.0, + "step": 17989 + }, + { + "epoch": 4.16, + "learning_rate": 1.5353208535551962e-05, + "loss": 0.0, + "step": 17990 + }, + { + "epoch": 4.16, + "learning_rate": 1.535128264386411e-05, + "loss": 0.0, + "step": 17991 + }, + { + "epoch": 4.16, + "learning_rate": 1.5349356752176257e-05, + "loss": 0.0, + "step": 17992 + }, + { + "epoch": 4.16, + "learning_rate": 1.5347430860488408e-05, + "loss": 0.0, + "step": 17993 + }, + { + "epoch": 4.16, + "learning_rate": 1.5345504968800556e-05, + "loss": 0.0, + "step": 17994 + }, + { + "epoch": 4.16, + "learning_rate": 1.5343579077112706e-05, + "loss": 0.0, + "step": 17995 + }, + { + "epoch": 4.16, + "learning_rate": 1.5341653185424854e-05, + "loss": 0.0, + "step": 17996 + }, + { + "epoch": 4.16, + "learning_rate": 1.5339727293737e-05, + "loss": 0.0, + "step": 17997 + }, + { + "epoch": 4.16, + "learning_rate": 1.533780140204915e-05, + "loss": 0.0, + "step": 17998 + }, + { + "epoch": 4.16, + "learning_rate": 1.5335875510361296e-05, + "loss": 0.0, + "step": 17999 + }, + { + "epoch": 4.16, + "learning_rate": 1.5333949618673447e-05, + "loss": 0.0, + "step": 18000 + }, + { + "epoch": 4.16, + "learning_rate": 1.5332023726985595e-05, + "loss": 0.0, + "step": 18001 + }, + { + "epoch": 4.16, + "learning_rate": 1.5330097835297742e-05, + "loss": 0.0, + "step": 18002 + }, + { + "epoch": 4.16, + "learning_rate": 1.5328171943609893e-05, + "loss": 0.0, + "step": 18003 + }, + { + "epoch": 4.16, + "learning_rate": 1.532624605192204e-05, + "loss": 0.0, + "step": 18004 + }, + { + "epoch": 4.16, + "learning_rate": 1.532432016023419e-05, + "loss": 0.0, + "step": 18005 + }, + { + "epoch": 4.16, + "learning_rate": 1.532239426854634e-05, + "loss": 0.0, + "step": 18006 + }, + { + "epoch": 4.16, + "learning_rate": 1.5320468376858486e-05, + "loss": 0.0, + "step": 18007 + }, + { + "epoch": 4.16, + "learning_rate": 1.5318542485170634e-05, + "loss": 0.0, + "step": 18008 + }, + { + "epoch": 4.16, + "learning_rate": 1.531661659348278e-05, + "loss": 0.0, + "step": 18009 + }, + { + "epoch": 4.16, + "learning_rate": 1.5314690701794932e-05, + "loss": 0.0, + "step": 18010 + }, + { + "epoch": 4.16, + "learning_rate": 1.531276481010708e-05, + "loss": 0.0, + "step": 18011 + }, + { + "epoch": 4.16, + "learning_rate": 1.531083891841923e-05, + "loss": 0.0001, + "step": 18012 + }, + { + "epoch": 4.16, + "learning_rate": 1.5308913026731378e-05, + "loss": 0.0, + "step": 18013 + }, + { + "epoch": 4.16, + "learning_rate": 1.5306987135043525e-05, + "loss": 0.0, + "step": 18014 + }, + { + "epoch": 4.16, + "learning_rate": 1.5305061243355676e-05, + "loss": 0.0, + "step": 18015 + }, + { + "epoch": 4.16, + "learning_rate": 1.530313535166782e-05, + "loss": 0.0, + "step": 18016 + }, + { + "epoch": 4.16, + "learning_rate": 1.530120945997997e-05, + "loss": 0.0, + "step": 18017 + }, + { + "epoch": 4.16, + "learning_rate": 1.529928356829212e-05, + "loss": 0.0, + "step": 18018 + }, + { + "epoch": 4.16, + "learning_rate": 1.529735767660427e-05, + "loss": 0.0, + "step": 18019 + }, + { + "epoch": 4.16, + "learning_rate": 1.5295431784916417e-05, + "loss": 0.0, + "step": 18020 + }, + { + "epoch": 4.16, + "learning_rate": 1.5293505893228565e-05, + "loss": 0.0, + "step": 18021 + }, + { + "epoch": 4.17, + "learning_rate": 1.5291580001540715e-05, + "loss": 0.0, + "step": 18022 + }, + { + "epoch": 4.17, + "learning_rate": 1.5289654109852863e-05, + "loss": 0.0, + "step": 18023 + }, + { + "epoch": 4.17, + "learning_rate": 1.5287728218165014e-05, + "loss": 0.0, + "step": 18024 + }, + { + "epoch": 4.17, + "learning_rate": 1.528580232647716e-05, + "loss": 0.0, + "step": 18025 + }, + { + "epoch": 4.17, + "learning_rate": 1.528387643478931e-05, + "loss": 0.0, + "step": 18026 + }, + { + "epoch": 4.17, + "learning_rate": 1.5281950543101456e-05, + "loss": 0.0, + "step": 18027 + }, + { + "epoch": 4.17, + "learning_rate": 1.5280024651413604e-05, + "loss": 0.0, + "step": 18028 + }, + { + "epoch": 4.17, + "learning_rate": 1.5278098759725754e-05, + "loss": 0.0, + "step": 18029 + }, + { + "epoch": 4.17, + "learning_rate": 1.5276172868037902e-05, + "loss": 0.0, + "step": 18030 + }, + { + "epoch": 4.17, + "learning_rate": 1.5274246976350053e-05, + "loss": 0.0001, + "step": 18031 + }, + { + "epoch": 4.17, + "learning_rate": 1.52723210846622e-05, + "loss": 0.0002, + "step": 18032 + }, + { + "epoch": 4.17, + "learning_rate": 1.5270395192974348e-05, + "loss": 0.0, + "step": 18033 + }, + { + "epoch": 4.17, + "learning_rate": 1.52684693012865e-05, + "loss": 0.0, + "step": 18034 + }, + { + "epoch": 4.17, + "learning_rate": 1.5266543409598643e-05, + "loss": 0.0, + "step": 18035 + }, + { + "epoch": 4.17, + "learning_rate": 1.5264617517910794e-05, + "loss": 0.0, + "step": 18036 + }, + { + "epoch": 4.17, + "learning_rate": 1.526269162622294e-05, + "loss": 0.0, + "step": 18037 + }, + { + "epoch": 4.17, + "learning_rate": 1.526076573453509e-05, + "loss": 0.0, + "step": 18038 + }, + { + "epoch": 4.17, + "learning_rate": 1.525883984284724e-05, + "loss": 0.0, + "step": 18039 + }, + { + "epoch": 4.17, + "learning_rate": 1.5256913951159389e-05, + "loss": 0.0, + "step": 18040 + }, + { + "epoch": 4.17, + "learning_rate": 1.5254988059471536e-05, + "loss": 0.0, + "step": 18041 + }, + { + "epoch": 4.17, + "learning_rate": 1.5253062167783685e-05, + "loss": 0.0, + "step": 18042 + }, + { + "epoch": 4.17, + "learning_rate": 1.5251136276095834e-05, + "loss": 0.0, + "step": 18043 + }, + { + "epoch": 4.17, + "learning_rate": 1.524921038440798e-05, + "loss": 0.0, + "step": 18044 + }, + { + "epoch": 4.17, + "learning_rate": 1.524728449272013e-05, + "loss": 0.0, + "step": 18045 + }, + { + "epoch": 4.17, + "learning_rate": 1.5245358601032279e-05, + "loss": 0.0, + "step": 18046 + }, + { + "epoch": 4.17, + "learning_rate": 1.5243432709344426e-05, + "loss": 0.0, + "step": 18047 + }, + { + "epoch": 4.17, + "learning_rate": 1.5241506817656575e-05, + "loss": 0.0, + "step": 18048 + }, + { + "epoch": 4.17, + "learning_rate": 1.5239580925968724e-05, + "loss": 0.0, + "step": 18049 + }, + { + "epoch": 4.17, + "learning_rate": 1.5237655034280874e-05, + "loss": 0.0, + "step": 18050 + }, + { + "epoch": 4.17, + "learning_rate": 1.5235729142593023e-05, + "loss": 0.0, + "step": 18051 + }, + { + "epoch": 4.17, + "learning_rate": 1.523380325090517e-05, + "loss": 0.0, + "step": 18052 + }, + { + "epoch": 4.17, + "learning_rate": 1.523187735921732e-05, + "loss": 0.0, + "step": 18053 + }, + { + "epoch": 4.17, + "learning_rate": 1.5229951467529465e-05, + "loss": 0.0, + "step": 18054 + }, + { + "epoch": 4.17, + "learning_rate": 1.5228025575841614e-05, + "loss": 0.0, + "step": 18055 + }, + { + "epoch": 4.17, + "learning_rate": 1.5226099684153763e-05, + "loss": 0.0, + "step": 18056 + }, + { + "epoch": 4.17, + "learning_rate": 1.5224173792465913e-05, + "loss": 0.0, + "step": 18057 + }, + { + "epoch": 4.17, + "learning_rate": 1.5222247900778062e-05, + "loss": 0.0001, + "step": 18058 + }, + { + "epoch": 4.17, + "learning_rate": 1.522032200909021e-05, + "loss": 0.0, + "step": 18059 + }, + { + "epoch": 4.17, + "learning_rate": 1.5218396117402358e-05, + "loss": 0.0, + "step": 18060 + }, + { + "epoch": 4.17, + "learning_rate": 1.5216470225714508e-05, + "loss": 0.0, + "step": 18061 + }, + { + "epoch": 4.17, + "learning_rate": 1.5214544334026657e-05, + "loss": 0.0001, + "step": 18062 + }, + { + "epoch": 4.17, + "learning_rate": 1.5212618442338803e-05, + "loss": 0.0, + "step": 18063 + }, + { + "epoch": 4.17, + "learning_rate": 1.5210692550650952e-05, + "loss": 0.0, + "step": 18064 + }, + { + "epoch": 4.17, + "learning_rate": 1.52087666589631e-05, + "loss": 0.0007, + "step": 18065 + }, + { + "epoch": 4.18, + "learning_rate": 1.5206840767275248e-05, + "loss": 0.0, + "step": 18066 + }, + { + "epoch": 4.18, + "learning_rate": 1.5204914875587398e-05, + "loss": 0.0, + "step": 18067 + }, + { + "epoch": 4.18, + "learning_rate": 1.5202988983899547e-05, + "loss": 0.0, + "step": 18068 + }, + { + "epoch": 4.18, + "learning_rate": 1.5201063092211696e-05, + "loss": 0.0, + "step": 18069 + }, + { + "epoch": 4.18, + "learning_rate": 1.5199137200523843e-05, + "loss": 0.0, + "step": 18070 + }, + { + "epoch": 4.18, + "learning_rate": 1.5197211308835993e-05, + "loss": 0.0, + "step": 18071 + }, + { + "epoch": 4.18, + "learning_rate": 1.5195285417148138e-05, + "loss": 0.0, + "step": 18072 + }, + { + "epoch": 4.18, + "learning_rate": 1.5193359525460288e-05, + "loss": 0.0, + "step": 18073 + }, + { + "epoch": 4.18, + "learning_rate": 1.5191433633772437e-05, + "loss": 0.0, + "step": 18074 + }, + { + "epoch": 4.18, + "learning_rate": 1.5189507742084586e-05, + "loss": 0.0, + "step": 18075 + }, + { + "epoch": 4.18, + "learning_rate": 1.5187581850396735e-05, + "loss": 0.0, + "step": 18076 + }, + { + "epoch": 4.18, + "learning_rate": 1.5185655958708883e-05, + "loss": 0.0, + "step": 18077 + }, + { + "epoch": 4.18, + "learning_rate": 1.5183730067021032e-05, + "loss": 0.0, + "step": 18078 + }, + { + "epoch": 4.18, + "learning_rate": 1.5181804175333181e-05, + "loss": 0.0003, + "step": 18079 + }, + { + "epoch": 4.18, + "learning_rate": 1.517987828364533e-05, + "loss": 0.0, + "step": 18080 + }, + { + "epoch": 4.18, + "learning_rate": 1.517795239195748e-05, + "loss": 0.0, + "step": 18081 + }, + { + "epoch": 4.18, + "learning_rate": 1.5176026500269625e-05, + "loss": 0.0, + "step": 18082 + }, + { + "epoch": 4.18, + "learning_rate": 1.5174100608581773e-05, + "loss": 0.0, + "step": 18083 + }, + { + "epoch": 4.18, + "learning_rate": 1.5172174716893922e-05, + "loss": 0.0, + "step": 18084 + }, + { + "epoch": 4.18, + "learning_rate": 1.5170248825206071e-05, + "loss": 0.0, + "step": 18085 + }, + { + "epoch": 4.18, + "learning_rate": 1.516832293351822e-05, + "loss": 0.0, + "step": 18086 + }, + { + "epoch": 4.18, + "learning_rate": 1.516639704183037e-05, + "loss": 0.0, + "step": 18087 + }, + { + "epoch": 4.18, + "learning_rate": 1.5164471150142517e-05, + "loss": 0.0, + "step": 18088 + }, + { + "epoch": 4.18, + "learning_rate": 1.5162545258454666e-05, + "loss": 0.0, + "step": 18089 + }, + { + "epoch": 4.18, + "learning_rate": 1.5160619366766815e-05, + "loss": 0.0, + "step": 18090 + }, + { + "epoch": 4.18, + "learning_rate": 1.515869347507896e-05, + "loss": 0.0, + "step": 18091 + }, + { + "epoch": 4.18, + "learning_rate": 1.515676758339111e-05, + "loss": 0.0, + "step": 18092 + }, + { + "epoch": 4.18, + "learning_rate": 1.5154841691703259e-05, + "loss": 0.0, + "step": 18093 + }, + { + "epoch": 4.18, + "learning_rate": 1.5152915800015408e-05, + "loss": 0.0, + "step": 18094 + }, + { + "epoch": 4.18, + "learning_rate": 1.5150989908327556e-05, + "loss": 0.0, + "step": 18095 + }, + { + "epoch": 4.18, + "learning_rate": 1.5149064016639705e-05, + "loss": 0.0, + "step": 18096 + }, + { + "epoch": 4.18, + "learning_rate": 1.5147138124951854e-05, + "loss": 0.0, + "step": 18097 + }, + { + "epoch": 4.18, + "learning_rate": 1.5145212233264003e-05, + "loss": 0.0, + "step": 18098 + }, + { + "epoch": 4.18, + "learning_rate": 1.5143286341576152e-05, + "loss": 0.0, + "step": 18099 + }, + { + "epoch": 4.18, + "learning_rate": 1.5141360449888298e-05, + "loss": 0.0, + "step": 18100 + }, + { + "epoch": 4.18, + "learning_rate": 1.5139434558200446e-05, + "loss": 0.0, + "step": 18101 + }, + { + "epoch": 4.18, + "learning_rate": 1.5137508666512595e-05, + "loss": 0.0, + "step": 18102 + }, + { + "epoch": 4.18, + "learning_rate": 1.5135582774824744e-05, + "loss": 0.0, + "step": 18103 + }, + { + "epoch": 4.18, + "learning_rate": 1.5133656883136893e-05, + "loss": 0.0, + "step": 18104 + }, + { + "epoch": 4.18, + "learning_rate": 1.5131730991449042e-05, + "loss": 0.0, + "step": 18105 + }, + { + "epoch": 4.18, + "learning_rate": 1.512980509976119e-05, + "loss": 0.0, + "step": 18106 + }, + { + "epoch": 4.18, + "learning_rate": 1.5127879208073339e-05, + "loss": 0.0, + "step": 18107 + }, + { + "epoch": 4.18, + "learning_rate": 1.5125953316385488e-05, + "loss": 0.0, + "step": 18108 + }, + { + "epoch": 4.19, + "learning_rate": 1.5124027424697637e-05, + "loss": 0.0, + "step": 18109 + }, + { + "epoch": 4.19, + "learning_rate": 1.5122101533009783e-05, + "loss": 0.0, + "step": 18110 + }, + { + "epoch": 4.19, + "learning_rate": 1.5120175641321932e-05, + "loss": 0.0, + "step": 18111 + }, + { + "epoch": 4.19, + "learning_rate": 1.5118249749634082e-05, + "loss": 0.0, + "step": 18112 + }, + { + "epoch": 4.19, + "learning_rate": 1.5116323857946229e-05, + "loss": 0.0, + "step": 18113 + }, + { + "epoch": 4.19, + "learning_rate": 1.5114397966258378e-05, + "loss": 0.0, + "step": 18114 + }, + { + "epoch": 4.19, + "learning_rate": 1.5112472074570527e-05, + "loss": 0.0, + "step": 18115 + }, + { + "epoch": 4.19, + "learning_rate": 1.5110546182882677e-05, + "loss": 0.0, + "step": 18116 + }, + { + "epoch": 4.19, + "learning_rate": 1.5108620291194826e-05, + "loss": 0.0, + "step": 18117 + }, + { + "epoch": 4.19, + "learning_rate": 1.5106694399506973e-05, + "loss": 0.0, + "step": 18118 + }, + { + "epoch": 4.19, + "learning_rate": 1.5104768507819119e-05, + "loss": 0.0, + "step": 18119 + }, + { + "epoch": 4.19, + "learning_rate": 1.5102842616131268e-05, + "loss": 0.0, + "step": 18120 + }, + { + "epoch": 4.19, + "learning_rate": 1.5100916724443417e-05, + "loss": 0.0001, + "step": 18121 + }, + { + "epoch": 4.19, + "learning_rate": 1.5098990832755566e-05, + "loss": 0.0, + "step": 18122 + }, + { + "epoch": 4.19, + "learning_rate": 1.5097064941067716e-05, + "loss": 0.0, + "step": 18123 + }, + { + "epoch": 4.19, + "learning_rate": 1.5095139049379863e-05, + "loss": 0.0, + "step": 18124 + }, + { + "epoch": 4.19, + "learning_rate": 1.5093213157692012e-05, + "loss": 0.0, + "step": 18125 + }, + { + "epoch": 4.19, + "learning_rate": 1.5091287266004161e-05, + "loss": 0.0, + "step": 18126 + }, + { + "epoch": 4.19, + "learning_rate": 1.508936137431631e-05, + "loss": 0.0, + "step": 18127 + }, + { + "epoch": 4.19, + "learning_rate": 1.5087435482628456e-05, + "loss": 0.0, + "step": 18128 + }, + { + "epoch": 4.19, + "learning_rate": 1.5085509590940606e-05, + "loss": 0.0, + "step": 18129 + }, + { + "epoch": 4.19, + "learning_rate": 1.5083583699252755e-05, + "loss": 0.0003, + "step": 18130 + }, + { + "epoch": 4.19, + "learning_rate": 1.5081657807564902e-05, + "loss": 0.0, + "step": 18131 + }, + { + "epoch": 4.19, + "learning_rate": 1.5079731915877051e-05, + "loss": 0.0, + "step": 18132 + }, + { + "epoch": 4.19, + "learning_rate": 1.50778060241892e-05, + "loss": 0.0, + "step": 18133 + }, + { + "epoch": 4.19, + "learning_rate": 1.507588013250135e-05, + "loss": 0.0, + "step": 18134 + }, + { + "epoch": 4.19, + "learning_rate": 1.5073954240813499e-05, + "loss": 0.0, + "step": 18135 + }, + { + "epoch": 4.19, + "learning_rate": 1.5072028349125646e-05, + "loss": 0.0, + "step": 18136 + }, + { + "epoch": 4.19, + "learning_rate": 1.5070102457437796e-05, + "loss": 0.0, + "step": 18137 + }, + { + "epoch": 4.19, + "learning_rate": 1.5068176565749941e-05, + "loss": 0.0, + "step": 18138 + }, + { + "epoch": 4.19, + "learning_rate": 1.506625067406209e-05, + "loss": 0.0, + "step": 18139 + }, + { + "epoch": 4.19, + "learning_rate": 1.506432478237424e-05, + "loss": 0.0, + "step": 18140 + }, + { + "epoch": 4.19, + "learning_rate": 1.5062398890686389e-05, + "loss": 0.0, + "step": 18141 + }, + { + "epoch": 4.19, + "learning_rate": 1.5060472998998536e-05, + "loss": 0.0, + "step": 18142 + }, + { + "epoch": 4.19, + "learning_rate": 1.5058547107310686e-05, + "loss": 0.0, + "step": 18143 + }, + { + "epoch": 4.19, + "learning_rate": 1.5056621215622835e-05, + "loss": 0.0, + "step": 18144 + }, + { + "epoch": 4.19, + "learning_rate": 1.5054695323934984e-05, + "loss": 0.0, + "step": 18145 + }, + { + "epoch": 4.19, + "learning_rate": 1.5052769432247133e-05, + "loss": 0.0, + "step": 18146 + }, + { + "epoch": 4.19, + "learning_rate": 1.5050843540559279e-05, + "loss": 0.0, + "step": 18147 + }, + { + "epoch": 4.19, + "learning_rate": 1.5048917648871428e-05, + "loss": 0.0, + "step": 18148 + }, + { + "epoch": 4.19, + "learning_rate": 1.5046991757183575e-05, + "loss": 0.0, + "step": 18149 + }, + { + "epoch": 4.19, + "learning_rate": 1.5045065865495725e-05, + "loss": 0.0, + "step": 18150 + }, + { + "epoch": 4.19, + "learning_rate": 1.5043139973807874e-05, + "loss": 0.0, + "step": 18151 + }, + { + "epoch": 4.2, + "learning_rate": 1.5041214082120023e-05, + "loss": 0.0, + "step": 18152 + }, + { + "epoch": 4.2, + "learning_rate": 1.5039288190432172e-05, + "loss": 0.0, + "step": 18153 + }, + { + "epoch": 4.2, + "learning_rate": 1.503736229874432e-05, + "loss": 0.0, + "step": 18154 + }, + { + "epoch": 4.2, + "learning_rate": 1.5035436407056469e-05, + "loss": 0.0, + "step": 18155 + }, + { + "epoch": 4.2, + "learning_rate": 1.5033510515368615e-05, + "loss": 0.0002, + "step": 18156 + }, + { + "epoch": 4.2, + "learning_rate": 1.5031584623680764e-05, + "loss": 0.0, + "step": 18157 + }, + { + "epoch": 4.2, + "learning_rate": 1.5029658731992913e-05, + "loss": 0.0, + "step": 18158 + }, + { + "epoch": 4.2, + "learning_rate": 1.5027732840305062e-05, + "loss": 0.0, + "step": 18159 + }, + { + "epoch": 4.2, + "learning_rate": 1.502580694861721e-05, + "loss": 0.0, + "step": 18160 + }, + { + "epoch": 4.2, + "learning_rate": 1.5023881056929359e-05, + "loss": 0.0, + "step": 18161 + }, + { + "epoch": 4.2, + "learning_rate": 1.5021955165241508e-05, + "loss": 0.0, + "step": 18162 + }, + { + "epoch": 4.2, + "learning_rate": 1.5020029273553657e-05, + "loss": 0.0, + "step": 18163 + }, + { + "epoch": 4.2, + "learning_rate": 1.5018103381865806e-05, + "loss": 0.0, + "step": 18164 + }, + { + "epoch": 4.2, + "learning_rate": 1.5016177490177954e-05, + "loss": 0.0, + "step": 18165 + }, + { + "epoch": 4.2, + "learning_rate": 1.5014251598490101e-05, + "loss": 0.0, + "step": 18166 + }, + { + "epoch": 4.2, + "learning_rate": 1.5012325706802249e-05, + "loss": 0.0, + "step": 18167 + }, + { + "epoch": 4.2, + "learning_rate": 1.5010399815114398e-05, + "loss": 0.0, + "step": 18168 + }, + { + "epoch": 4.2, + "learning_rate": 1.5008473923426547e-05, + "loss": 0.0, + "step": 18169 + }, + { + "epoch": 4.2, + "learning_rate": 1.5006548031738696e-05, + "loss": 0.0, + "step": 18170 + }, + { + "epoch": 4.2, + "learning_rate": 1.5004622140050845e-05, + "loss": 0.0, + "step": 18171 + }, + { + "epoch": 4.2, + "learning_rate": 1.5002696248362993e-05, + "loss": 0.0, + "step": 18172 + }, + { + "epoch": 4.2, + "learning_rate": 1.5000770356675142e-05, + "loss": 0.0, + "step": 18173 + }, + { + "epoch": 4.2, + "learning_rate": 1.4998844464987291e-05, + "loss": 0.0, + "step": 18174 + }, + { + "epoch": 4.2, + "learning_rate": 1.4996918573299437e-05, + "loss": 0.0, + "step": 18175 + }, + { + "epoch": 4.2, + "learning_rate": 1.4994992681611586e-05, + "loss": 0.0, + "step": 18176 + }, + { + "epoch": 4.2, + "learning_rate": 1.4993066789923735e-05, + "loss": 0.0, + "step": 18177 + }, + { + "epoch": 4.2, + "learning_rate": 1.4991140898235883e-05, + "loss": 0.0, + "step": 18178 + }, + { + "epoch": 4.2, + "learning_rate": 1.4989215006548032e-05, + "loss": 0.0, + "step": 18179 + }, + { + "epoch": 4.2, + "learning_rate": 1.4987289114860181e-05, + "loss": 0.0, + "step": 18180 + }, + { + "epoch": 4.2, + "learning_rate": 1.498536322317233e-05, + "loss": 0.0002, + "step": 18181 + }, + { + "epoch": 4.2, + "learning_rate": 1.498343733148448e-05, + "loss": 0.0, + "step": 18182 + }, + { + "epoch": 4.2, + "learning_rate": 1.4981511439796627e-05, + "loss": 0.0, + "step": 18183 + }, + { + "epoch": 4.2, + "learning_rate": 1.4979585548108774e-05, + "loss": 0.0, + "step": 18184 + }, + { + "epoch": 4.2, + "learning_rate": 1.4977659656420922e-05, + "loss": 0.0, + "step": 18185 + }, + { + "epoch": 4.2, + "learning_rate": 1.4975733764733071e-05, + "loss": 0.0, + "step": 18186 + }, + { + "epoch": 4.2, + "learning_rate": 1.497380787304522e-05, + "loss": 0.0001, + "step": 18187 + }, + { + "epoch": 4.2, + "learning_rate": 1.497188198135737e-05, + "loss": 0.0, + "step": 18188 + }, + { + "epoch": 4.2, + "learning_rate": 1.4969956089669519e-05, + "loss": 0.0, + "step": 18189 + }, + { + "epoch": 4.2, + "learning_rate": 1.4968030197981666e-05, + "loss": 0.0, + "step": 18190 + }, + { + "epoch": 4.2, + "learning_rate": 1.4966104306293815e-05, + "loss": 0.0, + "step": 18191 + }, + { + "epoch": 4.2, + "learning_rate": 1.4964178414605964e-05, + "loss": 0.0, + "step": 18192 + }, + { + "epoch": 4.2, + "learning_rate": 1.496225252291811e-05, + "loss": 0.0, + "step": 18193 + }, + { + "epoch": 4.2, + "learning_rate": 1.496032663123026e-05, + "loss": 0.0, + "step": 18194 + }, + { + "epoch": 4.2, + "learning_rate": 1.4958400739542409e-05, + "loss": 0.0, + "step": 18195 + }, + { + "epoch": 4.21, + "learning_rate": 1.4956474847854556e-05, + "loss": 0.0, + "step": 18196 + }, + { + "epoch": 4.21, + "learning_rate": 1.4954548956166705e-05, + "loss": 0.0, + "step": 18197 + }, + { + "epoch": 4.21, + "learning_rate": 1.4952623064478854e-05, + "loss": 0.0, + "step": 18198 + }, + { + "epoch": 4.21, + "learning_rate": 1.4950697172791004e-05, + "loss": 0.0, + "step": 18199 + }, + { + "epoch": 4.21, + "learning_rate": 1.4948771281103153e-05, + "loss": 0.0024, + "step": 18200 + }, + { + "epoch": 4.21, + "learning_rate": 1.49468453894153e-05, + "loss": 0.0, + "step": 18201 + }, + { + "epoch": 4.21, + "learning_rate": 1.494491949772745e-05, + "loss": 0.0, + "step": 18202 + }, + { + "epoch": 4.21, + "learning_rate": 1.4942993606039595e-05, + "loss": 0.0, + "step": 18203 + }, + { + "epoch": 4.21, + "learning_rate": 1.4941067714351744e-05, + "loss": 0.0, + "step": 18204 + }, + { + "epoch": 4.21, + "learning_rate": 1.4939141822663894e-05, + "loss": 0.0, + "step": 18205 + }, + { + "epoch": 4.21, + "learning_rate": 1.4937215930976043e-05, + "loss": 0.0002, + "step": 18206 + }, + { + "epoch": 4.21, + "learning_rate": 1.4935290039288192e-05, + "loss": 0.0, + "step": 18207 + }, + { + "epoch": 4.21, + "learning_rate": 1.493336414760034e-05, + "loss": 0.0, + "step": 18208 + }, + { + "epoch": 4.21, + "learning_rate": 1.4931438255912489e-05, + "loss": 0.0, + "step": 18209 + }, + { + "epoch": 4.21, + "learning_rate": 1.4929512364224638e-05, + "loss": 0.0, + "step": 18210 + }, + { + "epoch": 4.21, + "learning_rate": 1.4927586472536787e-05, + "loss": 0.0, + "step": 18211 + }, + { + "epoch": 4.21, + "learning_rate": 1.4925660580848933e-05, + "loss": 0.0002, + "step": 18212 + }, + { + "epoch": 4.21, + "learning_rate": 1.4923734689161082e-05, + "loss": 0.0, + "step": 18213 + }, + { + "epoch": 4.21, + "learning_rate": 1.492180879747323e-05, + "loss": 0.0, + "step": 18214 + }, + { + "epoch": 4.21, + "learning_rate": 1.4919882905785378e-05, + "loss": 0.0, + "step": 18215 + }, + { + "epoch": 4.21, + "learning_rate": 1.4917957014097528e-05, + "loss": 0.0, + "step": 18216 + }, + { + "epoch": 4.21, + "learning_rate": 1.4916031122409677e-05, + "loss": 0.0, + "step": 18217 + }, + { + "epoch": 4.21, + "learning_rate": 1.4914105230721826e-05, + "loss": 0.0, + "step": 18218 + }, + { + "epoch": 4.21, + "learning_rate": 1.4912179339033973e-05, + "loss": 0.0, + "step": 18219 + }, + { + "epoch": 4.21, + "learning_rate": 1.4910253447346123e-05, + "loss": 0.0, + "step": 18220 + }, + { + "epoch": 4.21, + "learning_rate": 1.4908327555658268e-05, + "loss": 0.0, + "step": 18221 + }, + { + "epoch": 4.21, + "learning_rate": 1.4906401663970418e-05, + "loss": 0.0, + "step": 18222 + }, + { + "epoch": 4.21, + "learning_rate": 1.4904475772282567e-05, + "loss": 0.0, + "step": 18223 + }, + { + "epoch": 4.21, + "learning_rate": 1.4902549880594716e-05, + "loss": 0.0001, + "step": 18224 + }, + { + "epoch": 4.21, + "learning_rate": 1.4900623988906865e-05, + "loss": 0.0, + "step": 18225 + }, + { + "epoch": 4.21, + "learning_rate": 1.4898698097219013e-05, + "loss": 0.0, + "step": 18226 + }, + { + "epoch": 4.21, + "learning_rate": 1.4896772205531162e-05, + "loss": 0.0, + "step": 18227 + }, + { + "epoch": 4.21, + "learning_rate": 1.4894846313843311e-05, + "loss": 0.0, + "step": 18228 + }, + { + "epoch": 4.21, + "learning_rate": 1.489292042215546e-05, + "loss": 0.0, + "step": 18229 + }, + { + "epoch": 4.21, + "learning_rate": 1.489099453046761e-05, + "loss": 0.0, + "step": 18230 + }, + { + "epoch": 4.21, + "learning_rate": 1.4889068638779755e-05, + "loss": 0.0, + "step": 18231 + }, + { + "epoch": 4.21, + "learning_rate": 1.4887142747091903e-05, + "loss": 0.0001, + "step": 18232 + }, + { + "epoch": 4.21, + "learning_rate": 1.4885216855404052e-05, + "loss": 0.0, + "step": 18233 + }, + { + "epoch": 4.21, + "learning_rate": 1.4883290963716201e-05, + "loss": 0.0, + "step": 18234 + }, + { + "epoch": 4.21, + "learning_rate": 1.488136507202835e-05, + "loss": 0.0, + "step": 18235 + }, + { + "epoch": 4.21, + "learning_rate": 1.48794391803405e-05, + "loss": 0.0, + "step": 18236 + }, + { + "epoch": 4.21, + "learning_rate": 1.4877513288652647e-05, + "loss": 0.0, + "step": 18237 + }, + { + "epoch": 4.21, + "learning_rate": 1.4875587396964796e-05, + "loss": 0.0, + "step": 18238 + }, + { + "epoch": 4.22, + "learning_rate": 1.4873661505276945e-05, + "loss": 0.0, + "step": 18239 + }, + { + "epoch": 4.22, + "learning_rate": 1.487173561358909e-05, + "loss": 0.0002, + "step": 18240 + }, + { + "epoch": 4.22, + "learning_rate": 1.486980972190124e-05, + "loss": 0.0, + "step": 18241 + }, + { + "epoch": 4.22, + "learning_rate": 1.486788383021339e-05, + "loss": 0.0, + "step": 18242 + }, + { + "epoch": 4.22, + "learning_rate": 1.4865957938525538e-05, + "loss": 0.0, + "step": 18243 + }, + { + "epoch": 4.22, + "learning_rate": 1.4864032046837686e-05, + "loss": 0.0, + "step": 18244 + }, + { + "epoch": 4.22, + "learning_rate": 1.4862106155149835e-05, + "loss": 0.0, + "step": 18245 + }, + { + "epoch": 4.22, + "learning_rate": 1.4860180263461984e-05, + "loss": 0.0, + "step": 18246 + }, + { + "epoch": 4.22, + "learning_rate": 1.4858254371774133e-05, + "loss": 0.0, + "step": 18247 + }, + { + "epoch": 4.22, + "learning_rate": 1.4856328480086283e-05, + "loss": 0.0, + "step": 18248 + }, + { + "epoch": 4.22, + "learning_rate": 1.4854402588398428e-05, + "loss": 0.0, + "step": 18249 + }, + { + "epoch": 4.22, + "learning_rate": 1.4852476696710576e-05, + "loss": 0.0, + "step": 18250 + }, + { + "epoch": 4.22, + "learning_rate": 1.4850550805022725e-05, + "loss": 0.0, + "step": 18251 + }, + { + "epoch": 4.22, + "learning_rate": 1.4848624913334874e-05, + "loss": 0.0, + "step": 18252 + }, + { + "epoch": 4.22, + "learning_rate": 1.4846699021647023e-05, + "loss": 0.0, + "step": 18253 + }, + { + "epoch": 4.22, + "learning_rate": 1.4844773129959172e-05, + "loss": 0.0, + "step": 18254 + }, + { + "epoch": 4.22, + "learning_rate": 1.484284723827132e-05, + "loss": 0.0, + "step": 18255 + }, + { + "epoch": 4.22, + "learning_rate": 1.4840921346583469e-05, + "loss": 0.0, + "step": 18256 + }, + { + "epoch": 4.22, + "learning_rate": 1.4838995454895618e-05, + "loss": 0.0, + "step": 18257 + }, + { + "epoch": 4.22, + "learning_rate": 1.4837069563207767e-05, + "loss": 0.0, + "step": 18258 + }, + { + "epoch": 4.22, + "learning_rate": 1.4835143671519913e-05, + "loss": 0.0, + "step": 18259 + }, + { + "epoch": 4.22, + "learning_rate": 1.4833217779832062e-05, + "loss": 0.0, + "step": 18260 + }, + { + "epoch": 4.22, + "learning_rate": 1.4831291888144212e-05, + "loss": 0.0, + "step": 18261 + }, + { + "epoch": 4.22, + "learning_rate": 1.4829365996456359e-05, + "loss": 0.0, + "step": 18262 + }, + { + "epoch": 4.22, + "learning_rate": 1.4827440104768508e-05, + "loss": 0.0, + "step": 18263 + }, + { + "epoch": 4.22, + "learning_rate": 1.4825514213080657e-05, + "loss": 0.0, + "step": 18264 + }, + { + "epoch": 4.22, + "learning_rate": 1.4823588321392807e-05, + "loss": 0.0, + "step": 18265 + }, + { + "epoch": 4.22, + "learning_rate": 1.4821662429704956e-05, + "loss": 0.0, + "step": 18266 + }, + { + "epoch": 4.22, + "learning_rate": 1.4819736538017103e-05, + "loss": 0.0, + "step": 18267 + }, + { + "epoch": 4.22, + "learning_rate": 1.481781064632925e-05, + "loss": 0.0, + "step": 18268 + }, + { + "epoch": 4.22, + "learning_rate": 1.4815884754641398e-05, + "loss": 0.0, + "step": 18269 + }, + { + "epoch": 4.22, + "learning_rate": 1.4813958862953547e-05, + "loss": 0.0, + "step": 18270 + }, + { + "epoch": 4.22, + "learning_rate": 1.4812032971265697e-05, + "loss": 0.0, + "step": 18271 + }, + { + "epoch": 4.22, + "learning_rate": 1.4810107079577846e-05, + "loss": 0.0, + "step": 18272 + }, + { + "epoch": 4.22, + "learning_rate": 1.4808181187889993e-05, + "loss": 0.0, + "step": 18273 + }, + { + "epoch": 4.22, + "learning_rate": 1.4806255296202142e-05, + "loss": 0.0, + "step": 18274 + }, + { + "epoch": 4.22, + "learning_rate": 1.4804329404514292e-05, + "loss": 0.0, + "step": 18275 + }, + { + "epoch": 4.22, + "learning_rate": 1.480240351282644e-05, + "loss": 0.0, + "step": 18276 + }, + { + "epoch": 4.22, + "learning_rate": 1.4800477621138586e-05, + "loss": 0.0, + "step": 18277 + }, + { + "epoch": 4.22, + "learning_rate": 1.4798551729450736e-05, + "loss": 0.0, + "step": 18278 + }, + { + "epoch": 4.22, + "learning_rate": 1.4796625837762885e-05, + "loss": 0.0, + "step": 18279 + }, + { + "epoch": 4.22, + "learning_rate": 1.4794699946075032e-05, + "loss": 0.0, + "step": 18280 + }, + { + "epoch": 4.22, + "learning_rate": 1.4792774054387181e-05, + "loss": 0.0, + "step": 18281 + }, + { + "epoch": 4.23, + "learning_rate": 1.479084816269933e-05, + "loss": 0.0, + "step": 18282 + }, + { + "epoch": 4.23, + "learning_rate": 1.478892227101148e-05, + "loss": 0.0, + "step": 18283 + }, + { + "epoch": 4.23, + "learning_rate": 1.4786996379323629e-05, + "loss": 0.0, + "step": 18284 + }, + { + "epoch": 4.23, + "learning_rate": 1.4785070487635776e-05, + "loss": 0.0, + "step": 18285 + }, + { + "epoch": 4.23, + "learning_rate": 1.4783144595947926e-05, + "loss": 0.0, + "step": 18286 + }, + { + "epoch": 4.23, + "learning_rate": 1.4781218704260071e-05, + "loss": 0.0, + "step": 18287 + }, + { + "epoch": 4.23, + "learning_rate": 1.477929281257222e-05, + "loss": 0.0, + "step": 18288 + }, + { + "epoch": 4.23, + "learning_rate": 1.477736692088437e-05, + "loss": 0.0, + "step": 18289 + }, + { + "epoch": 4.23, + "learning_rate": 1.4775441029196519e-05, + "loss": 0.0, + "step": 18290 + }, + { + "epoch": 4.23, + "learning_rate": 1.4773515137508666e-05, + "loss": 0.0, + "step": 18291 + }, + { + "epoch": 4.23, + "learning_rate": 1.4771589245820816e-05, + "loss": 0.0, + "step": 18292 + }, + { + "epoch": 4.23, + "learning_rate": 1.4769663354132965e-05, + "loss": 0.0, + "step": 18293 + }, + { + "epoch": 4.23, + "learning_rate": 1.4767737462445114e-05, + "loss": 0.0, + "step": 18294 + }, + { + "epoch": 4.23, + "learning_rate": 1.4765811570757263e-05, + "loss": 0.0, + "step": 18295 + }, + { + "epoch": 4.23, + "learning_rate": 1.4763885679069409e-05, + "loss": 0.0, + "step": 18296 + }, + { + "epoch": 4.23, + "learning_rate": 1.4761959787381558e-05, + "loss": 0.0, + "step": 18297 + }, + { + "epoch": 4.23, + "learning_rate": 1.4760033895693706e-05, + "loss": 0.0, + "step": 18298 + }, + { + "epoch": 4.23, + "learning_rate": 1.4758108004005855e-05, + "loss": 0.0, + "step": 18299 + }, + { + "epoch": 4.23, + "learning_rate": 1.4756182112318004e-05, + "loss": 0.0, + "step": 18300 + }, + { + "epoch": 4.23, + "learning_rate": 1.4754256220630153e-05, + "loss": 0.0, + "step": 18301 + }, + { + "epoch": 4.23, + "learning_rate": 1.4752330328942302e-05, + "loss": 0.0, + "step": 18302 + }, + { + "epoch": 4.23, + "learning_rate": 1.475040443725445e-05, + "loss": 0.0, + "step": 18303 + }, + { + "epoch": 4.23, + "learning_rate": 1.4748478545566599e-05, + "loss": 0.0, + "step": 18304 + }, + { + "epoch": 4.23, + "learning_rate": 1.4746552653878745e-05, + "loss": 0.0, + "step": 18305 + }, + { + "epoch": 4.23, + "learning_rate": 1.4744626762190894e-05, + "loss": 0.0, + "step": 18306 + }, + { + "epoch": 4.23, + "learning_rate": 1.4742700870503043e-05, + "loss": 0.0, + "step": 18307 + }, + { + "epoch": 4.23, + "learning_rate": 1.4740774978815192e-05, + "loss": 0.0, + "step": 18308 + }, + { + "epoch": 4.23, + "learning_rate": 1.473884908712734e-05, + "loss": 0.0, + "step": 18309 + }, + { + "epoch": 4.23, + "learning_rate": 1.4736923195439489e-05, + "loss": 0.0, + "step": 18310 + }, + { + "epoch": 4.23, + "learning_rate": 1.4734997303751638e-05, + "loss": 0.0, + "step": 18311 + }, + { + "epoch": 4.23, + "learning_rate": 1.4733071412063787e-05, + "loss": 0.0, + "step": 18312 + }, + { + "epoch": 4.23, + "learning_rate": 1.4731145520375936e-05, + "loss": 0.0, + "step": 18313 + }, + { + "epoch": 4.23, + "learning_rate": 1.4729219628688084e-05, + "loss": 0.0, + "step": 18314 + }, + { + "epoch": 4.23, + "learning_rate": 1.4727293737000231e-05, + "loss": 0.0, + "step": 18315 + }, + { + "epoch": 4.23, + "learning_rate": 1.4725367845312379e-05, + "loss": 0.0, + "step": 18316 + }, + { + "epoch": 4.23, + "learning_rate": 1.4723441953624528e-05, + "loss": 0.0, + "step": 18317 + }, + { + "epoch": 4.23, + "learning_rate": 1.4721516061936677e-05, + "loss": 0.0002, + "step": 18318 + }, + { + "epoch": 4.23, + "learning_rate": 1.4719590170248826e-05, + "loss": 0.0, + "step": 18319 + }, + { + "epoch": 4.23, + "learning_rate": 1.4717664278560975e-05, + "loss": 0.0, + "step": 18320 + }, + { + "epoch": 4.23, + "learning_rate": 1.4715738386873123e-05, + "loss": 0.0, + "step": 18321 + }, + { + "epoch": 4.23, + "learning_rate": 1.4713812495185272e-05, + "loss": 0.0, + "step": 18322 + }, + { + "epoch": 4.23, + "learning_rate": 1.4711886603497421e-05, + "loss": 0.0, + "step": 18323 + }, + { + "epoch": 4.23, + "learning_rate": 1.4709960711809567e-05, + "loss": 0.0002, + "step": 18324 + }, + { + "epoch": 4.24, + "learning_rate": 1.4708034820121716e-05, + "loss": 0.0, + "step": 18325 + }, + { + "epoch": 4.24, + "learning_rate": 1.4706108928433865e-05, + "loss": 0.0, + "step": 18326 + }, + { + "epoch": 4.24, + "learning_rate": 1.4704183036746013e-05, + "loss": 0.0, + "step": 18327 + }, + { + "epoch": 4.24, + "learning_rate": 1.4702257145058162e-05, + "loss": 0.0, + "step": 18328 + }, + { + "epoch": 4.24, + "learning_rate": 1.4700331253370311e-05, + "loss": 0.0, + "step": 18329 + }, + { + "epoch": 4.24, + "learning_rate": 1.469840536168246e-05, + "loss": 0.0, + "step": 18330 + }, + { + "epoch": 4.24, + "learning_rate": 1.469647946999461e-05, + "loss": 0.0, + "step": 18331 + }, + { + "epoch": 4.24, + "learning_rate": 1.4694553578306757e-05, + "loss": 0.0, + "step": 18332 + }, + { + "epoch": 4.24, + "learning_rate": 1.4692627686618905e-05, + "loss": 0.0001, + "step": 18333 + }, + { + "epoch": 4.24, + "learning_rate": 1.4690701794931052e-05, + "loss": 0.0, + "step": 18334 + }, + { + "epoch": 4.24, + "learning_rate": 1.4688775903243201e-05, + "loss": 0.0, + "step": 18335 + }, + { + "epoch": 4.24, + "learning_rate": 1.468685001155535e-05, + "loss": 0.0, + "step": 18336 + }, + { + "epoch": 4.24, + "learning_rate": 1.46849241198675e-05, + "loss": 0.0, + "step": 18337 + }, + { + "epoch": 4.24, + "learning_rate": 1.4682998228179649e-05, + "loss": 0.0, + "step": 18338 + }, + { + "epoch": 4.24, + "learning_rate": 1.4681072336491796e-05, + "loss": 0.0, + "step": 18339 + }, + { + "epoch": 4.24, + "learning_rate": 1.4679146444803945e-05, + "loss": 0.0, + "step": 18340 + }, + { + "epoch": 4.24, + "learning_rate": 1.4677220553116095e-05, + "loss": 0.0, + "step": 18341 + }, + { + "epoch": 4.24, + "learning_rate": 1.4675294661428244e-05, + "loss": 0.0, + "step": 18342 + }, + { + "epoch": 4.24, + "learning_rate": 1.467336876974039e-05, + "loss": 0.0, + "step": 18343 + }, + { + "epoch": 4.24, + "learning_rate": 1.4671442878052539e-05, + "loss": 0.0, + "step": 18344 + }, + { + "epoch": 4.24, + "learning_rate": 1.4669516986364686e-05, + "loss": 0.0001, + "step": 18345 + }, + { + "epoch": 4.24, + "learning_rate": 1.4667591094676835e-05, + "loss": 0.0, + "step": 18346 + }, + { + "epoch": 4.24, + "learning_rate": 1.4665665202988984e-05, + "loss": 0.0, + "step": 18347 + }, + { + "epoch": 4.24, + "learning_rate": 1.4663739311301134e-05, + "loss": 0.0002, + "step": 18348 + }, + { + "epoch": 4.24, + "learning_rate": 1.4661813419613283e-05, + "loss": 0.0, + "step": 18349 + }, + { + "epoch": 4.24, + "learning_rate": 1.465988752792543e-05, + "loss": 0.0, + "step": 18350 + }, + { + "epoch": 4.24, + "learning_rate": 1.465796163623758e-05, + "loss": 0.0, + "step": 18351 + }, + { + "epoch": 4.24, + "learning_rate": 1.4656035744549725e-05, + "loss": 0.0001, + "step": 18352 + }, + { + "epoch": 4.24, + "learning_rate": 1.4654109852861874e-05, + "loss": 0.0027, + "step": 18353 + }, + { + "epoch": 4.24, + "learning_rate": 1.4652183961174024e-05, + "loss": 0.0, + "step": 18354 + }, + { + "epoch": 4.24, + "learning_rate": 1.4650258069486173e-05, + "loss": 0.0, + "step": 18355 + }, + { + "epoch": 4.24, + "learning_rate": 1.4648332177798322e-05, + "loss": 0.0, + "step": 18356 + }, + { + "epoch": 4.24, + "learning_rate": 1.464640628611047e-05, + "loss": 0.0001, + "step": 18357 + }, + { + "epoch": 4.24, + "learning_rate": 1.4644480394422619e-05, + "loss": 0.0, + "step": 18358 + }, + { + "epoch": 4.24, + "learning_rate": 1.4642554502734768e-05, + "loss": 0.0, + "step": 18359 + }, + { + "epoch": 4.24, + "learning_rate": 1.4640628611046917e-05, + "loss": 0.0, + "step": 18360 + }, + { + "epoch": 4.24, + "learning_rate": 1.4638702719359063e-05, + "loss": 0.0, + "step": 18361 + }, + { + "epoch": 4.24, + "learning_rate": 1.4636776827671212e-05, + "loss": 0.0, + "step": 18362 + }, + { + "epoch": 4.24, + "learning_rate": 1.4634850935983361e-05, + "loss": 0.0, + "step": 18363 + }, + { + "epoch": 4.24, + "learning_rate": 1.4632925044295509e-05, + "loss": 0.0, + "step": 18364 + }, + { + "epoch": 4.24, + "learning_rate": 1.4630999152607658e-05, + "loss": 0.0, + "step": 18365 + }, + { + "epoch": 4.24, + "learning_rate": 1.4629073260919807e-05, + "loss": 0.0, + "step": 18366 + }, + { + "epoch": 4.24, + "learning_rate": 1.4627147369231956e-05, + "loss": 0.0, + "step": 18367 + }, + { + "epoch": 4.24, + "learning_rate": 1.4625221477544104e-05, + "loss": 0.0, + "step": 18368 + }, + { + "epoch": 4.25, + "learning_rate": 1.4623295585856253e-05, + "loss": 0.0, + "step": 18369 + }, + { + "epoch": 4.25, + "learning_rate": 1.4621369694168402e-05, + "loss": 0.0, + "step": 18370 + }, + { + "epoch": 4.25, + "learning_rate": 1.4619443802480548e-05, + "loss": 0.0, + "step": 18371 + }, + { + "epoch": 4.25, + "learning_rate": 1.4617517910792697e-05, + "loss": 0.0006, + "step": 18372 + }, + { + "epoch": 4.25, + "learning_rate": 1.4615592019104846e-05, + "loss": 0.0, + "step": 18373 + }, + { + "epoch": 4.25, + "learning_rate": 1.4613666127416995e-05, + "loss": 0.0, + "step": 18374 + }, + { + "epoch": 4.25, + "learning_rate": 1.4611740235729143e-05, + "loss": 0.0, + "step": 18375 + }, + { + "epoch": 4.25, + "learning_rate": 1.4609814344041292e-05, + "loss": 0.0, + "step": 18376 + }, + { + "epoch": 4.25, + "learning_rate": 1.4607888452353441e-05, + "loss": 0.0, + "step": 18377 + }, + { + "epoch": 4.25, + "learning_rate": 1.460596256066559e-05, + "loss": 0.0, + "step": 18378 + }, + { + "epoch": 4.25, + "learning_rate": 1.460403666897774e-05, + "loss": 0.0, + "step": 18379 + }, + { + "epoch": 4.25, + "learning_rate": 1.4602110777289885e-05, + "loss": 0.0, + "step": 18380 + }, + { + "epoch": 4.25, + "learning_rate": 1.4600184885602034e-05, + "loss": 0.0, + "step": 18381 + }, + { + "epoch": 4.25, + "learning_rate": 1.4598258993914182e-05, + "loss": 0.0, + "step": 18382 + }, + { + "epoch": 4.25, + "learning_rate": 1.4596333102226331e-05, + "loss": 0.0, + "step": 18383 + }, + { + "epoch": 4.25, + "learning_rate": 1.459440721053848e-05, + "loss": 0.0, + "step": 18384 + }, + { + "epoch": 4.25, + "learning_rate": 1.459248131885063e-05, + "loss": 0.0, + "step": 18385 + }, + { + "epoch": 4.25, + "learning_rate": 1.4590555427162777e-05, + "loss": 0.0, + "step": 18386 + }, + { + "epoch": 4.25, + "learning_rate": 1.4588629535474926e-05, + "loss": 0.0, + "step": 18387 + }, + { + "epoch": 4.25, + "learning_rate": 1.4586703643787075e-05, + "loss": 0.0, + "step": 18388 + }, + { + "epoch": 4.25, + "learning_rate": 1.4584777752099221e-05, + "loss": 0.0, + "step": 18389 + }, + { + "epoch": 4.25, + "learning_rate": 1.458285186041137e-05, + "loss": 0.0, + "step": 18390 + }, + { + "epoch": 4.25, + "learning_rate": 1.458092596872352e-05, + "loss": 0.0, + "step": 18391 + }, + { + "epoch": 4.25, + "learning_rate": 1.4579000077035668e-05, + "loss": 0.0, + "step": 18392 + }, + { + "epoch": 4.25, + "learning_rate": 1.4577074185347816e-05, + "loss": 0.0002, + "step": 18393 + }, + { + "epoch": 4.25, + "learning_rate": 1.4575148293659965e-05, + "loss": 0.0, + "step": 18394 + }, + { + "epoch": 4.25, + "learning_rate": 1.4573222401972114e-05, + "loss": 0.0, + "step": 18395 + }, + { + "epoch": 4.25, + "learning_rate": 1.4571296510284263e-05, + "loss": 0.0, + "step": 18396 + }, + { + "epoch": 4.25, + "learning_rate": 1.4569370618596413e-05, + "loss": 0.0, + "step": 18397 + }, + { + "epoch": 4.25, + "learning_rate": 1.456744472690856e-05, + "loss": 0.0, + "step": 18398 + }, + { + "epoch": 4.25, + "learning_rate": 1.4565518835220708e-05, + "loss": 0.0, + "step": 18399 + }, + { + "epoch": 4.25, + "learning_rate": 1.4563592943532855e-05, + "loss": 0.0, + "step": 18400 + }, + { + "epoch": 4.25, + "learning_rate": 1.4561667051845004e-05, + "loss": 0.0, + "step": 18401 + }, + { + "epoch": 4.25, + "learning_rate": 1.4559741160157153e-05, + "loss": 0.0, + "step": 18402 + }, + { + "epoch": 4.25, + "learning_rate": 1.4557815268469303e-05, + "loss": 0.0, + "step": 18403 + }, + { + "epoch": 4.25, + "learning_rate": 1.455588937678145e-05, + "loss": 0.0, + "step": 18404 + }, + { + "epoch": 4.25, + "learning_rate": 1.45539634850936e-05, + "loss": 0.0, + "step": 18405 + }, + { + "epoch": 4.25, + "learning_rate": 1.4552037593405748e-05, + "loss": 0.0, + "step": 18406 + }, + { + "epoch": 4.25, + "learning_rate": 1.4550111701717898e-05, + "loss": 0.0, + "step": 18407 + }, + { + "epoch": 4.25, + "learning_rate": 1.4548185810030043e-05, + "loss": 0.0, + "step": 18408 + }, + { + "epoch": 4.25, + "learning_rate": 1.4546259918342192e-05, + "loss": 0.0, + "step": 18409 + }, + { + "epoch": 4.25, + "learning_rate": 1.4544334026654342e-05, + "loss": 0.0, + "step": 18410 + }, + { + "epoch": 4.25, + "learning_rate": 1.4542408134966489e-05, + "loss": 0.0, + "step": 18411 + }, + { + "epoch": 4.26, + "learning_rate": 1.4540482243278638e-05, + "loss": 0.0, + "step": 18412 + }, + { + "epoch": 4.26, + "learning_rate": 1.4538556351590787e-05, + "loss": 0.0, + "step": 18413 + }, + { + "epoch": 4.26, + "learning_rate": 1.4536630459902937e-05, + "loss": 0.0, + "step": 18414 + }, + { + "epoch": 4.26, + "learning_rate": 1.4534704568215086e-05, + "loss": 0.0, + "step": 18415 + }, + { + "epoch": 4.26, + "learning_rate": 1.4532778676527233e-05, + "loss": 0.0, + "step": 18416 + }, + { + "epoch": 4.26, + "learning_rate": 1.453085278483938e-05, + "loss": 0.0001, + "step": 18417 + }, + { + "epoch": 4.26, + "learning_rate": 1.4528926893151528e-05, + "loss": 0.0, + "step": 18418 + }, + { + "epoch": 4.26, + "learning_rate": 1.4527001001463677e-05, + "loss": 0.0002, + "step": 18419 + }, + { + "epoch": 4.26, + "learning_rate": 1.4525075109775827e-05, + "loss": 0.0, + "step": 18420 + }, + { + "epoch": 4.26, + "learning_rate": 1.4523149218087976e-05, + "loss": 0.0001, + "step": 18421 + }, + { + "epoch": 4.26, + "learning_rate": 1.4521223326400123e-05, + "loss": 0.0, + "step": 18422 + }, + { + "epoch": 4.26, + "learning_rate": 1.4519297434712272e-05, + "loss": 0.0, + "step": 18423 + }, + { + "epoch": 4.26, + "learning_rate": 1.4517371543024422e-05, + "loss": 0.0, + "step": 18424 + }, + { + "epoch": 4.26, + "learning_rate": 1.451544565133657e-05, + "loss": 0.0, + "step": 18425 + }, + { + "epoch": 4.26, + "learning_rate": 1.451351975964872e-05, + "loss": 0.0, + "step": 18426 + }, + { + "epoch": 4.26, + "learning_rate": 1.4511593867960866e-05, + "loss": 0.0, + "step": 18427 + }, + { + "epoch": 4.26, + "learning_rate": 1.4509667976273015e-05, + "loss": 0.0, + "step": 18428 + }, + { + "epoch": 4.26, + "learning_rate": 1.4507742084585162e-05, + "loss": 0.0, + "step": 18429 + }, + { + "epoch": 4.26, + "learning_rate": 1.4505816192897312e-05, + "loss": 0.0001, + "step": 18430 + }, + { + "epoch": 4.26, + "learning_rate": 1.450389030120946e-05, + "loss": 0.0, + "step": 18431 + }, + { + "epoch": 4.26, + "learning_rate": 1.450196440952161e-05, + "loss": 0.0, + "step": 18432 + }, + { + "epoch": 4.26, + "learning_rate": 1.4500038517833759e-05, + "loss": 0.0, + "step": 18433 + }, + { + "epoch": 4.26, + "learning_rate": 1.4498112626145907e-05, + "loss": 0.0, + "step": 18434 + }, + { + "epoch": 4.26, + "learning_rate": 1.4496186734458056e-05, + "loss": 0.0, + "step": 18435 + }, + { + "epoch": 4.26, + "learning_rate": 1.4494260842770201e-05, + "loss": 0.0, + "step": 18436 + }, + { + "epoch": 4.26, + "learning_rate": 1.449233495108235e-05, + "loss": 0.0, + "step": 18437 + }, + { + "epoch": 4.26, + "learning_rate": 1.44904090593945e-05, + "loss": 0.0, + "step": 18438 + }, + { + "epoch": 4.26, + "learning_rate": 1.4488483167706649e-05, + "loss": 0.0, + "step": 18439 + }, + { + "epoch": 4.26, + "learning_rate": 1.4486557276018798e-05, + "loss": 0.0, + "step": 18440 + }, + { + "epoch": 4.26, + "learning_rate": 1.4484631384330946e-05, + "loss": 0.0002, + "step": 18441 + }, + { + "epoch": 4.26, + "learning_rate": 1.4482705492643095e-05, + "loss": 0.0, + "step": 18442 + }, + { + "epoch": 4.26, + "learning_rate": 1.4480779600955244e-05, + "loss": 0.0, + "step": 18443 + }, + { + "epoch": 4.26, + "learning_rate": 1.4478853709267393e-05, + "loss": 0.0, + "step": 18444 + }, + { + "epoch": 4.26, + "learning_rate": 1.4476927817579539e-05, + "loss": 0.0, + "step": 18445 + }, + { + "epoch": 4.26, + "learning_rate": 1.4475001925891688e-05, + "loss": 0.0, + "step": 18446 + }, + { + "epoch": 4.26, + "learning_rate": 1.4473076034203836e-05, + "loss": 0.0, + "step": 18447 + }, + { + "epoch": 4.26, + "learning_rate": 1.4471150142515985e-05, + "loss": 0.0, + "step": 18448 + }, + { + "epoch": 4.26, + "learning_rate": 1.4469224250828134e-05, + "loss": 0.0, + "step": 18449 + }, + { + "epoch": 4.26, + "learning_rate": 1.4467298359140283e-05, + "loss": 0.0, + "step": 18450 + }, + { + "epoch": 4.26, + "learning_rate": 1.4465372467452432e-05, + "loss": 0.0, + "step": 18451 + }, + { + "epoch": 4.26, + "learning_rate": 1.446344657576458e-05, + "loss": 0.0, + "step": 18452 + }, + { + "epoch": 4.26, + "learning_rate": 1.4461520684076729e-05, + "loss": 0.0, + "step": 18453 + }, + { + "epoch": 4.26, + "learning_rate": 1.4459594792388878e-05, + "loss": 0.0, + "step": 18454 + }, + { + "epoch": 4.27, + "learning_rate": 1.4457668900701024e-05, + "loss": 0.0, + "step": 18455 + }, + { + "epoch": 4.27, + "learning_rate": 1.4455743009013173e-05, + "loss": 0.0, + "step": 18456 + }, + { + "epoch": 4.27, + "learning_rate": 1.4453817117325322e-05, + "loss": 0.0, + "step": 18457 + }, + { + "epoch": 4.27, + "learning_rate": 1.4451891225637471e-05, + "loss": 0.0, + "step": 18458 + }, + { + "epoch": 4.27, + "learning_rate": 1.4449965333949619e-05, + "loss": 0.0, + "step": 18459 + }, + { + "epoch": 4.27, + "learning_rate": 1.4448039442261768e-05, + "loss": 0.0, + "step": 18460 + }, + { + "epoch": 4.27, + "learning_rate": 1.4446113550573917e-05, + "loss": 0.0, + "step": 18461 + }, + { + "epoch": 4.27, + "learning_rate": 1.4444187658886066e-05, + "loss": 0.0001, + "step": 18462 + }, + { + "epoch": 4.27, + "learning_rate": 1.4442261767198214e-05, + "loss": 0.0, + "step": 18463 + }, + { + "epoch": 4.27, + "learning_rate": 1.4440335875510361e-05, + "loss": 0.0, + "step": 18464 + }, + { + "epoch": 4.27, + "learning_rate": 1.4438409983822509e-05, + "loss": 0.0, + "step": 18465 + }, + { + "epoch": 4.27, + "learning_rate": 1.4436484092134658e-05, + "loss": 0.0, + "step": 18466 + }, + { + "epoch": 4.27, + "learning_rate": 1.4434558200446807e-05, + "loss": 0.0, + "step": 18467 + }, + { + "epoch": 4.27, + "learning_rate": 1.4432632308758956e-05, + "loss": 0.0, + "step": 18468 + }, + { + "epoch": 4.27, + "learning_rate": 1.4430706417071106e-05, + "loss": 0.0, + "step": 18469 + }, + { + "epoch": 4.27, + "learning_rate": 1.4428780525383253e-05, + "loss": 0.0, + "step": 18470 + }, + { + "epoch": 4.27, + "learning_rate": 1.4426854633695402e-05, + "loss": 0.0, + "step": 18471 + }, + { + "epoch": 4.27, + "learning_rate": 1.4424928742007551e-05, + "loss": 0.0, + "step": 18472 + }, + { + "epoch": 4.27, + "learning_rate": 1.4423002850319697e-05, + "loss": 0.0, + "step": 18473 + }, + { + "epoch": 4.27, + "learning_rate": 1.4421076958631846e-05, + "loss": 0.0, + "step": 18474 + }, + { + "epoch": 4.27, + "learning_rate": 1.4419151066943995e-05, + "loss": 0.0, + "step": 18475 + }, + { + "epoch": 4.27, + "learning_rate": 1.4417225175256145e-05, + "loss": 0.0, + "step": 18476 + }, + { + "epoch": 4.27, + "learning_rate": 1.4415299283568292e-05, + "loss": 0.0, + "step": 18477 + }, + { + "epoch": 4.27, + "learning_rate": 1.4413373391880441e-05, + "loss": 0.0, + "step": 18478 + }, + { + "epoch": 4.27, + "learning_rate": 1.441144750019259e-05, + "loss": 0.0, + "step": 18479 + }, + { + "epoch": 4.27, + "learning_rate": 1.440952160850474e-05, + "loss": 0.0, + "step": 18480 + }, + { + "epoch": 4.27, + "learning_rate": 1.4407595716816887e-05, + "loss": 0.0, + "step": 18481 + }, + { + "epoch": 4.27, + "learning_rate": 1.4405669825129036e-05, + "loss": 0.0, + "step": 18482 + }, + { + "epoch": 4.27, + "learning_rate": 1.4403743933441182e-05, + "loss": 0.0, + "step": 18483 + }, + { + "epoch": 4.27, + "learning_rate": 1.4401818041753331e-05, + "loss": 0.0, + "step": 18484 + }, + { + "epoch": 4.27, + "learning_rate": 1.439989215006548e-05, + "loss": 0.0, + "step": 18485 + }, + { + "epoch": 4.27, + "learning_rate": 1.439796625837763e-05, + "loss": 0.0, + "step": 18486 + }, + { + "epoch": 4.27, + "learning_rate": 1.4396040366689779e-05, + "loss": 0.0, + "step": 18487 + }, + { + "epoch": 4.27, + "learning_rate": 1.4394114475001926e-05, + "loss": 0.0, + "step": 18488 + }, + { + "epoch": 4.27, + "learning_rate": 1.4392188583314075e-05, + "loss": 0.0, + "step": 18489 + }, + { + "epoch": 4.27, + "learning_rate": 1.4390262691626225e-05, + "loss": 0.0, + "step": 18490 + }, + { + "epoch": 4.27, + "learning_rate": 1.4388336799938374e-05, + "loss": 0.0, + "step": 18491 + }, + { + "epoch": 4.27, + "learning_rate": 1.438641090825052e-05, + "loss": 0.0, + "step": 18492 + }, + { + "epoch": 4.27, + "learning_rate": 1.4384485016562669e-05, + "loss": 0.0, + "step": 18493 + }, + { + "epoch": 4.27, + "learning_rate": 1.4382559124874818e-05, + "loss": 0.0, + "step": 18494 + }, + { + "epoch": 4.27, + "learning_rate": 1.4380633233186965e-05, + "loss": 0.0, + "step": 18495 + }, + { + "epoch": 4.27, + "learning_rate": 1.4378707341499115e-05, + "loss": 0.0009, + "step": 18496 + }, + { + "epoch": 4.27, + "learning_rate": 1.4376781449811264e-05, + "loss": 0.0, + "step": 18497 + }, + { + "epoch": 4.28, + "learning_rate": 1.4374855558123413e-05, + "loss": 0.0, + "step": 18498 + }, + { + "epoch": 4.28, + "learning_rate": 1.437292966643556e-05, + "loss": 0.0, + "step": 18499 + }, + { + "epoch": 4.28, + "learning_rate": 1.437100377474771e-05, + "loss": 0.0, + "step": 18500 + }, + { + "epoch": 4.28, + "learning_rate": 1.4369077883059855e-05, + "loss": 0.0, + "step": 18501 + }, + { + "epoch": 4.28, + "learning_rate": 1.4367151991372004e-05, + "loss": 0.0, + "step": 18502 + }, + { + "epoch": 4.28, + "learning_rate": 1.4365226099684154e-05, + "loss": 0.0, + "step": 18503 + }, + { + "epoch": 4.28, + "learning_rate": 1.4363300207996303e-05, + "loss": 0.0, + "step": 18504 + }, + { + "epoch": 4.28, + "learning_rate": 1.4361374316308452e-05, + "loss": 0.0, + "step": 18505 + }, + { + "epoch": 4.28, + "learning_rate": 1.43594484246206e-05, + "loss": 0.0001, + "step": 18506 + }, + { + "epoch": 4.28, + "learning_rate": 1.4357522532932749e-05, + "loss": 0.0, + "step": 18507 + }, + { + "epoch": 4.28, + "learning_rate": 1.4355596641244898e-05, + "loss": 0.0, + "step": 18508 + }, + { + "epoch": 4.28, + "learning_rate": 1.4353670749557047e-05, + "loss": 0.0, + "step": 18509 + }, + { + "epoch": 4.28, + "learning_rate": 1.4351744857869196e-05, + "loss": 0.0, + "step": 18510 + }, + { + "epoch": 4.28, + "learning_rate": 1.4349818966181342e-05, + "loss": 0.0, + "step": 18511 + }, + { + "epoch": 4.28, + "learning_rate": 1.4347893074493491e-05, + "loss": 0.0, + "step": 18512 + }, + { + "epoch": 4.28, + "learning_rate": 1.4345967182805639e-05, + "loss": 0.0, + "step": 18513 + }, + { + "epoch": 4.28, + "learning_rate": 1.4344041291117788e-05, + "loss": 0.0, + "step": 18514 + }, + { + "epoch": 4.28, + "learning_rate": 1.4342115399429937e-05, + "loss": 0.0, + "step": 18515 + }, + { + "epoch": 4.28, + "learning_rate": 1.4340189507742086e-05, + "loss": 0.0, + "step": 18516 + }, + { + "epoch": 4.28, + "learning_rate": 1.4338263616054234e-05, + "loss": 0.0, + "step": 18517 + }, + { + "epoch": 4.28, + "learning_rate": 1.4336337724366383e-05, + "loss": 0.0, + "step": 18518 + }, + { + "epoch": 4.28, + "learning_rate": 1.4334411832678532e-05, + "loss": 0.0, + "step": 18519 + }, + { + "epoch": 4.28, + "learning_rate": 1.4332485940990678e-05, + "loss": 0.0, + "step": 18520 + }, + { + "epoch": 4.28, + "learning_rate": 1.4330560049302827e-05, + "loss": 0.0, + "step": 18521 + }, + { + "epoch": 4.28, + "learning_rate": 1.4328634157614976e-05, + "loss": 0.0, + "step": 18522 + }, + { + "epoch": 4.28, + "learning_rate": 1.4326708265927125e-05, + "loss": 0.0, + "step": 18523 + }, + { + "epoch": 4.28, + "learning_rate": 1.4324782374239273e-05, + "loss": 0.0, + "step": 18524 + }, + { + "epoch": 4.28, + "learning_rate": 1.4322856482551422e-05, + "loss": 0.0, + "step": 18525 + }, + { + "epoch": 4.28, + "learning_rate": 1.4320930590863571e-05, + "loss": 0.0, + "step": 18526 + }, + { + "epoch": 4.28, + "learning_rate": 1.431900469917572e-05, + "loss": 0.0, + "step": 18527 + }, + { + "epoch": 4.28, + "learning_rate": 1.431707880748787e-05, + "loss": 0.0, + "step": 18528 + }, + { + "epoch": 4.28, + "learning_rate": 1.4315152915800015e-05, + "loss": 0.0, + "step": 18529 + }, + { + "epoch": 4.28, + "learning_rate": 1.4313227024112164e-05, + "loss": 0.0, + "step": 18530 + }, + { + "epoch": 4.28, + "learning_rate": 1.4311301132424312e-05, + "loss": 0.0, + "step": 18531 + }, + { + "epoch": 4.28, + "learning_rate": 1.4309375240736461e-05, + "loss": 0.0, + "step": 18532 + }, + { + "epoch": 4.28, + "learning_rate": 1.430744934904861e-05, + "loss": 0.0, + "step": 18533 + }, + { + "epoch": 4.28, + "learning_rate": 1.430552345736076e-05, + "loss": 0.0, + "step": 18534 + }, + { + "epoch": 4.28, + "learning_rate": 1.4303597565672908e-05, + "loss": 0.0, + "step": 18535 + }, + { + "epoch": 4.28, + "learning_rate": 1.4301671673985056e-05, + "loss": 0.0, + "step": 18536 + }, + { + "epoch": 4.28, + "learning_rate": 1.4299745782297205e-05, + "loss": 0.0, + "step": 18537 + }, + { + "epoch": 4.28, + "learning_rate": 1.4297819890609354e-05, + "loss": 0.0, + "step": 18538 + }, + { + "epoch": 4.28, + "learning_rate": 1.42958939989215e-05, + "loss": 0.0, + "step": 18539 + }, + { + "epoch": 4.28, + "learning_rate": 1.429396810723365e-05, + "loss": 0.0005, + "step": 18540 + }, + { + "epoch": 4.28, + "learning_rate": 1.4292042215545798e-05, + "loss": 0.0003, + "step": 18541 + }, + { + "epoch": 4.29, + "learning_rate": 1.4290116323857946e-05, + "loss": 0.0, + "step": 18542 + }, + { + "epoch": 4.29, + "learning_rate": 1.4288190432170095e-05, + "loss": 0.0, + "step": 18543 + }, + { + "epoch": 4.29, + "learning_rate": 1.4286264540482244e-05, + "loss": 0.0, + "step": 18544 + }, + { + "epoch": 4.29, + "learning_rate": 1.4284338648794393e-05, + "loss": 0.0, + "step": 18545 + }, + { + "epoch": 4.29, + "learning_rate": 1.4282412757106543e-05, + "loss": 0.0, + "step": 18546 + }, + { + "epoch": 4.29, + "learning_rate": 1.428048686541869e-05, + "loss": 0.0, + "step": 18547 + }, + { + "epoch": 4.29, + "learning_rate": 1.4278560973730838e-05, + "loss": 0.0, + "step": 18548 + }, + { + "epoch": 4.29, + "learning_rate": 1.4276635082042985e-05, + "loss": 0.0, + "step": 18549 + }, + { + "epoch": 4.29, + "learning_rate": 1.4274709190355134e-05, + "loss": 0.0, + "step": 18550 + }, + { + "epoch": 4.29, + "learning_rate": 1.4272783298667283e-05, + "loss": 0.0005, + "step": 18551 + }, + { + "epoch": 4.29, + "learning_rate": 1.4270857406979433e-05, + "loss": 0.0, + "step": 18552 + }, + { + "epoch": 4.29, + "learning_rate": 1.4268931515291582e-05, + "loss": 0.0, + "step": 18553 + }, + { + "epoch": 4.29, + "learning_rate": 1.426700562360373e-05, + "loss": 0.0, + "step": 18554 + }, + { + "epoch": 4.29, + "learning_rate": 1.4265079731915878e-05, + "loss": 0.0, + "step": 18555 + }, + { + "epoch": 4.29, + "learning_rate": 1.4263153840228028e-05, + "loss": 0.0, + "step": 18556 + }, + { + "epoch": 4.29, + "learning_rate": 1.4261227948540173e-05, + "loss": 0.0, + "step": 18557 + }, + { + "epoch": 4.29, + "learning_rate": 1.4259302056852323e-05, + "loss": 0.0, + "step": 18558 + }, + { + "epoch": 4.29, + "learning_rate": 1.4257376165164472e-05, + "loss": 0.0001, + "step": 18559 + }, + { + "epoch": 4.29, + "learning_rate": 1.425545027347662e-05, + "loss": 0.0, + "step": 18560 + }, + { + "epoch": 4.29, + "learning_rate": 1.4253524381788768e-05, + "loss": 0.0, + "step": 18561 + }, + { + "epoch": 4.29, + "learning_rate": 1.4251598490100917e-05, + "loss": 0.0, + "step": 18562 + }, + { + "epoch": 4.29, + "learning_rate": 1.4249672598413067e-05, + "loss": 0.0, + "step": 18563 + }, + { + "epoch": 4.29, + "learning_rate": 1.4247746706725216e-05, + "loss": 0.0, + "step": 18564 + }, + { + "epoch": 4.29, + "learning_rate": 1.4245820815037363e-05, + "loss": 0.0, + "step": 18565 + }, + { + "epoch": 4.29, + "learning_rate": 1.4243894923349512e-05, + "loss": 0.0, + "step": 18566 + }, + { + "epoch": 4.29, + "learning_rate": 1.4241969031661658e-05, + "loss": 0.0, + "step": 18567 + }, + { + "epoch": 4.29, + "learning_rate": 1.4240043139973807e-05, + "loss": 0.0, + "step": 18568 + }, + { + "epoch": 4.29, + "learning_rate": 1.4238117248285957e-05, + "loss": 0.0, + "step": 18569 + }, + { + "epoch": 4.29, + "learning_rate": 1.4236191356598106e-05, + "loss": 0.0001, + "step": 18570 + }, + { + "epoch": 4.29, + "learning_rate": 1.4234265464910255e-05, + "loss": 0.0, + "step": 18571 + }, + { + "epoch": 4.29, + "learning_rate": 1.4232339573222402e-05, + "loss": 0.0, + "step": 18572 + }, + { + "epoch": 4.29, + "learning_rate": 1.4230413681534552e-05, + "loss": 0.0, + "step": 18573 + }, + { + "epoch": 4.29, + "learning_rate": 1.42284877898467e-05, + "loss": 0.0, + "step": 18574 + }, + { + "epoch": 4.29, + "learning_rate": 1.422656189815885e-05, + "loss": 0.0002, + "step": 18575 + }, + { + "epoch": 4.29, + "learning_rate": 1.4224636006470996e-05, + "loss": 0.0, + "step": 18576 + }, + { + "epoch": 4.29, + "learning_rate": 1.4222710114783145e-05, + "loss": 0.0, + "step": 18577 + }, + { + "epoch": 4.29, + "learning_rate": 1.4220784223095292e-05, + "loss": 0.0, + "step": 18578 + }, + { + "epoch": 4.29, + "learning_rate": 1.4218858331407442e-05, + "loss": 0.0, + "step": 18579 + }, + { + "epoch": 4.29, + "learning_rate": 1.421693243971959e-05, + "loss": 0.0, + "step": 18580 + }, + { + "epoch": 4.29, + "learning_rate": 1.421500654803174e-05, + "loss": 0.0, + "step": 18581 + }, + { + "epoch": 4.29, + "learning_rate": 1.4213080656343889e-05, + "loss": 0.0001, + "step": 18582 + }, + { + "epoch": 4.29, + "learning_rate": 1.4211154764656037e-05, + "loss": 0.0, + "step": 18583 + }, + { + "epoch": 4.29, + "learning_rate": 1.4209228872968186e-05, + "loss": 0.0, + "step": 18584 + }, + { + "epoch": 4.3, + "learning_rate": 1.4207302981280332e-05, + "loss": 0.0, + "step": 18585 + }, + { + "epoch": 4.3, + "learning_rate": 1.420537708959248e-05, + "loss": 0.0, + "step": 18586 + }, + { + "epoch": 4.3, + "learning_rate": 1.420345119790463e-05, + "loss": 0.0, + "step": 18587 + }, + { + "epoch": 4.3, + "learning_rate": 1.4201525306216779e-05, + "loss": 0.0, + "step": 18588 + }, + { + "epoch": 4.3, + "learning_rate": 1.4199599414528928e-05, + "loss": 0.0, + "step": 18589 + }, + { + "epoch": 4.3, + "learning_rate": 1.4197673522841076e-05, + "loss": 0.0005, + "step": 18590 + }, + { + "epoch": 4.3, + "learning_rate": 1.4195747631153225e-05, + "loss": 0.0, + "step": 18591 + }, + { + "epoch": 4.3, + "learning_rate": 1.4193821739465374e-05, + "loss": 0.0, + "step": 18592 + }, + { + "epoch": 4.3, + "learning_rate": 1.4191895847777523e-05, + "loss": 0.0, + "step": 18593 + }, + { + "epoch": 4.3, + "learning_rate": 1.418996995608967e-05, + "loss": 0.0, + "step": 18594 + }, + { + "epoch": 4.3, + "learning_rate": 1.4188044064401818e-05, + "loss": 0.0, + "step": 18595 + }, + { + "epoch": 4.3, + "learning_rate": 1.4186118172713966e-05, + "loss": 0.0, + "step": 18596 + }, + { + "epoch": 4.3, + "learning_rate": 1.4184192281026115e-05, + "loss": 0.0, + "step": 18597 + }, + { + "epoch": 4.3, + "learning_rate": 1.4182266389338264e-05, + "loss": 0.0, + "step": 18598 + }, + { + "epoch": 4.3, + "learning_rate": 1.4180340497650413e-05, + "loss": 0.0, + "step": 18599 + }, + { + "epoch": 4.3, + "learning_rate": 1.4178414605962562e-05, + "loss": 0.0001, + "step": 18600 + }, + { + "epoch": 4.3, + "learning_rate": 1.417648871427471e-05, + "loss": 0.0, + "step": 18601 + }, + { + "epoch": 4.3, + "learning_rate": 1.4174562822586859e-05, + "loss": 0.0, + "step": 18602 + }, + { + "epoch": 4.3, + "learning_rate": 1.4172636930899008e-05, + "loss": 0.0, + "step": 18603 + }, + { + "epoch": 4.3, + "learning_rate": 1.4170711039211154e-05, + "loss": 0.0, + "step": 18604 + }, + { + "epoch": 4.3, + "learning_rate": 1.4168785147523303e-05, + "loss": 0.0, + "step": 18605 + }, + { + "epoch": 4.3, + "learning_rate": 1.4166859255835452e-05, + "loss": 0.0, + "step": 18606 + }, + { + "epoch": 4.3, + "learning_rate": 1.4164933364147601e-05, + "loss": 0.0, + "step": 18607 + }, + { + "epoch": 4.3, + "learning_rate": 1.4163007472459749e-05, + "loss": 0.0, + "step": 18608 + }, + { + "epoch": 4.3, + "learning_rate": 1.4161081580771898e-05, + "loss": 0.0001, + "step": 18609 + }, + { + "epoch": 4.3, + "learning_rate": 1.4159155689084047e-05, + "loss": 0.0, + "step": 18610 + }, + { + "epoch": 4.3, + "learning_rate": 1.4157229797396196e-05, + "loss": 0.0, + "step": 18611 + }, + { + "epoch": 4.3, + "learning_rate": 1.4155303905708344e-05, + "loss": 0.0, + "step": 18612 + }, + { + "epoch": 4.3, + "learning_rate": 1.4153378014020491e-05, + "loss": 0.0, + "step": 18613 + }, + { + "epoch": 4.3, + "learning_rate": 1.4151452122332639e-05, + "loss": 0.0, + "step": 18614 + }, + { + "epoch": 4.3, + "learning_rate": 1.4149526230644788e-05, + "loss": 0.0, + "step": 18615 + }, + { + "epoch": 4.3, + "learning_rate": 1.4147600338956937e-05, + "loss": 0.0, + "step": 18616 + }, + { + "epoch": 4.3, + "learning_rate": 1.4145674447269086e-05, + "loss": 0.0, + "step": 18617 + }, + { + "epoch": 4.3, + "learning_rate": 1.4143748555581236e-05, + "loss": 0.0, + "step": 18618 + }, + { + "epoch": 4.3, + "learning_rate": 1.4141822663893383e-05, + "loss": 0.0, + "step": 18619 + }, + { + "epoch": 4.3, + "learning_rate": 1.4139896772205532e-05, + "loss": 0.0, + "step": 18620 + }, + { + "epoch": 4.3, + "learning_rate": 1.4137970880517681e-05, + "loss": 0.0, + "step": 18621 + }, + { + "epoch": 4.3, + "learning_rate": 1.413604498882983e-05, + "loss": 0.0, + "step": 18622 + }, + { + "epoch": 4.3, + "learning_rate": 1.4134119097141976e-05, + "loss": 0.0, + "step": 18623 + }, + { + "epoch": 4.3, + "learning_rate": 1.4132193205454125e-05, + "loss": 0.0, + "step": 18624 + }, + { + "epoch": 4.3, + "learning_rate": 1.4130267313766275e-05, + "loss": 0.0, + "step": 18625 + }, + { + "epoch": 4.3, + "learning_rate": 1.4128341422078422e-05, + "loss": 0.0, + "step": 18626 + }, + { + "epoch": 4.3, + "learning_rate": 1.4126415530390571e-05, + "loss": 0.0, + "step": 18627 + }, + { + "epoch": 4.31, + "learning_rate": 1.412448963870272e-05, + "loss": 0.0, + "step": 18628 + }, + { + "epoch": 4.31, + "learning_rate": 1.412256374701487e-05, + "loss": 0.0, + "step": 18629 + }, + { + "epoch": 4.31, + "learning_rate": 1.4120637855327019e-05, + "loss": 0.0, + "step": 18630 + }, + { + "epoch": 4.31, + "learning_rate": 1.4118711963639166e-05, + "loss": 0.0, + "step": 18631 + }, + { + "epoch": 4.31, + "learning_rate": 1.4116786071951312e-05, + "loss": 0.0, + "step": 18632 + }, + { + "epoch": 4.31, + "learning_rate": 1.4114860180263461e-05, + "loss": 0.0003, + "step": 18633 + }, + { + "epoch": 4.31, + "learning_rate": 1.411293428857561e-05, + "loss": 0.0, + "step": 18634 + }, + { + "epoch": 4.31, + "learning_rate": 1.411100839688776e-05, + "loss": 0.0, + "step": 18635 + }, + { + "epoch": 4.31, + "learning_rate": 1.4109082505199909e-05, + "loss": 0.0, + "step": 18636 + }, + { + "epoch": 4.31, + "learning_rate": 1.4107156613512056e-05, + "loss": 0.0, + "step": 18637 + }, + { + "epoch": 4.31, + "learning_rate": 1.4105230721824205e-05, + "loss": 0.0, + "step": 18638 + }, + { + "epoch": 4.31, + "learning_rate": 1.4103304830136355e-05, + "loss": 0.0001, + "step": 18639 + }, + { + "epoch": 4.31, + "learning_rate": 1.4101378938448504e-05, + "loss": 0.0, + "step": 18640 + }, + { + "epoch": 4.31, + "learning_rate": 1.409945304676065e-05, + "loss": 0.0, + "step": 18641 + }, + { + "epoch": 4.31, + "learning_rate": 1.4097527155072799e-05, + "loss": 0.0, + "step": 18642 + }, + { + "epoch": 4.31, + "learning_rate": 1.4095601263384948e-05, + "loss": 0.0, + "step": 18643 + }, + { + "epoch": 4.31, + "learning_rate": 1.4093675371697095e-05, + "loss": 0.0, + "step": 18644 + }, + { + "epoch": 4.31, + "learning_rate": 1.4091749480009245e-05, + "loss": 0.0, + "step": 18645 + }, + { + "epoch": 4.31, + "learning_rate": 1.4089823588321394e-05, + "loss": 0.0, + "step": 18646 + }, + { + "epoch": 4.31, + "learning_rate": 1.4087897696633543e-05, + "loss": 0.0, + "step": 18647 + }, + { + "epoch": 4.31, + "learning_rate": 1.4085971804945692e-05, + "loss": 0.0001, + "step": 18648 + }, + { + "epoch": 4.31, + "learning_rate": 1.408404591325784e-05, + "loss": 0.0, + "step": 18649 + }, + { + "epoch": 4.31, + "learning_rate": 1.4082120021569989e-05, + "loss": 0.0, + "step": 18650 + }, + { + "epoch": 4.31, + "learning_rate": 1.4080194129882134e-05, + "loss": 0.0, + "step": 18651 + }, + { + "epoch": 4.31, + "learning_rate": 1.4078268238194284e-05, + "loss": 0.0, + "step": 18652 + }, + { + "epoch": 4.31, + "learning_rate": 1.4076342346506433e-05, + "loss": 0.0, + "step": 18653 + }, + { + "epoch": 4.31, + "learning_rate": 1.4074416454818582e-05, + "loss": 0.0007, + "step": 18654 + }, + { + "epoch": 4.31, + "learning_rate": 1.407249056313073e-05, + "loss": 0.0, + "step": 18655 + }, + { + "epoch": 4.31, + "learning_rate": 1.4070564671442879e-05, + "loss": 0.0, + "step": 18656 + }, + { + "epoch": 4.31, + "learning_rate": 1.4068638779755028e-05, + "loss": 0.0001, + "step": 18657 + }, + { + "epoch": 4.31, + "learning_rate": 1.4066712888067177e-05, + "loss": 0.0, + "step": 18658 + }, + { + "epoch": 4.31, + "learning_rate": 1.4064786996379326e-05, + "loss": 0.0, + "step": 18659 + }, + { + "epoch": 4.31, + "learning_rate": 1.4062861104691472e-05, + "loss": 0.0, + "step": 18660 + }, + { + "epoch": 4.31, + "learning_rate": 1.4060935213003621e-05, + "loss": 0.0, + "step": 18661 + }, + { + "epoch": 4.31, + "learning_rate": 1.4059009321315769e-05, + "loss": 0.0, + "step": 18662 + }, + { + "epoch": 4.31, + "learning_rate": 1.4057083429627918e-05, + "loss": 0.0, + "step": 18663 + }, + { + "epoch": 4.31, + "learning_rate": 1.4055157537940067e-05, + "loss": 0.0, + "step": 18664 + }, + { + "epoch": 4.31, + "learning_rate": 1.4053231646252216e-05, + "loss": 0.0, + "step": 18665 + }, + { + "epoch": 4.31, + "learning_rate": 1.4051305754564365e-05, + "loss": 0.0, + "step": 18666 + }, + { + "epoch": 4.31, + "learning_rate": 1.4049379862876513e-05, + "loss": 0.0, + "step": 18667 + }, + { + "epoch": 4.31, + "learning_rate": 1.4047453971188662e-05, + "loss": 0.0, + "step": 18668 + }, + { + "epoch": 4.31, + "learning_rate": 1.4045528079500808e-05, + "loss": 0.0, + "step": 18669 + }, + { + "epoch": 4.31, + "learning_rate": 1.4043602187812957e-05, + "loss": 0.0, + "step": 18670 + }, + { + "epoch": 4.31, + "learning_rate": 1.4041676296125106e-05, + "loss": 0.0, + "step": 18671 + }, + { + "epoch": 4.32, + "learning_rate": 1.4039750404437255e-05, + "loss": 0.0, + "step": 18672 + }, + { + "epoch": 4.32, + "learning_rate": 1.4037824512749403e-05, + "loss": 0.0, + "step": 18673 + }, + { + "epoch": 4.32, + "learning_rate": 1.4035898621061552e-05, + "loss": 0.0, + "step": 18674 + }, + { + "epoch": 4.32, + "learning_rate": 1.4033972729373701e-05, + "loss": 0.0, + "step": 18675 + }, + { + "epoch": 4.32, + "learning_rate": 1.403204683768585e-05, + "loss": 0.0, + "step": 18676 + }, + { + "epoch": 4.32, + "learning_rate": 1.4030120945998e-05, + "loss": 0.0, + "step": 18677 + }, + { + "epoch": 4.32, + "learning_rate": 1.4028195054310147e-05, + "loss": 0.0, + "step": 18678 + }, + { + "epoch": 4.32, + "learning_rate": 1.4026269162622294e-05, + "loss": 0.0, + "step": 18679 + }, + { + "epoch": 4.32, + "learning_rate": 1.4024343270934442e-05, + "loss": 0.0, + "step": 18680 + }, + { + "epoch": 4.32, + "learning_rate": 1.4022417379246591e-05, + "loss": 0.0, + "step": 18681 + }, + { + "epoch": 4.32, + "learning_rate": 1.402049148755874e-05, + "loss": 0.0, + "step": 18682 + }, + { + "epoch": 4.32, + "learning_rate": 1.401856559587089e-05, + "loss": 0.0, + "step": 18683 + }, + { + "epoch": 4.32, + "learning_rate": 1.4016639704183039e-05, + "loss": 0.0, + "step": 18684 + }, + { + "epoch": 4.32, + "learning_rate": 1.4014713812495186e-05, + "loss": 0.0002, + "step": 18685 + }, + { + "epoch": 4.32, + "learning_rate": 1.4012787920807335e-05, + "loss": 0.0, + "step": 18686 + }, + { + "epoch": 4.32, + "learning_rate": 1.4010862029119484e-05, + "loss": 0.0, + "step": 18687 + }, + { + "epoch": 4.32, + "learning_rate": 1.400893613743163e-05, + "loss": 0.0, + "step": 18688 + }, + { + "epoch": 4.32, + "learning_rate": 1.400701024574378e-05, + "loss": 0.0, + "step": 18689 + }, + { + "epoch": 4.32, + "learning_rate": 1.4005084354055928e-05, + "loss": 0.0, + "step": 18690 + }, + { + "epoch": 4.32, + "learning_rate": 1.4003158462368076e-05, + "loss": 0.0, + "step": 18691 + }, + { + "epoch": 4.32, + "learning_rate": 1.4001232570680225e-05, + "loss": 0.0, + "step": 18692 + }, + { + "epoch": 4.32, + "learning_rate": 1.3999306678992374e-05, + "loss": 0.0, + "step": 18693 + }, + { + "epoch": 4.32, + "learning_rate": 1.3997380787304523e-05, + "loss": 0.0, + "step": 18694 + }, + { + "epoch": 4.32, + "learning_rate": 1.3995454895616673e-05, + "loss": 0.0, + "step": 18695 + }, + { + "epoch": 4.32, + "learning_rate": 1.399352900392882e-05, + "loss": 0.0, + "step": 18696 + }, + { + "epoch": 4.32, + "learning_rate": 1.3991603112240968e-05, + "loss": 0.0, + "step": 18697 + }, + { + "epoch": 4.32, + "learning_rate": 1.3989677220553115e-05, + "loss": 0.0, + "step": 18698 + }, + { + "epoch": 4.32, + "learning_rate": 1.3987751328865264e-05, + "loss": 0.0, + "step": 18699 + }, + { + "epoch": 4.32, + "learning_rate": 1.3985825437177413e-05, + "loss": 0.0, + "step": 18700 + }, + { + "epoch": 4.32, + "learning_rate": 1.3983899545489563e-05, + "loss": 0.0, + "step": 18701 + }, + { + "epoch": 4.32, + "learning_rate": 1.3981973653801712e-05, + "loss": 0.0, + "step": 18702 + }, + { + "epoch": 4.32, + "learning_rate": 1.398004776211386e-05, + "loss": 0.0006, + "step": 18703 + }, + { + "epoch": 4.32, + "learning_rate": 1.3978121870426008e-05, + "loss": 0.0, + "step": 18704 + }, + { + "epoch": 4.32, + "learning_rate": 1.3976195978738158e-05, + "loss": 0.0, + "step": 18705 + }, + { + "epoch": 4.32, + "learning_rate": 1.3974270087050307e-05, + "loss": 0.0, + "step": 18706 + }, + { + "epoch": 4.32, + "learning_rate": 1.3972344195362453e-05, + "loss": 0.0, + "step": 18707 + }, + { + "epoch": 4.32, + "learning_rate": 1.3970418303674602e-05, + "loss": 0.0, + "step": 18708 + }, + { + "epoch": 4.32, + "learning_rate": 1.396849241198675e-05, + "loss": 0.0, + "step": 18709 + }, + { + "epoch": 4.32, + "learning_rate": 1.3966566520298898e-05, + "loss": 0.0, + "step": 18710 + }, + { + "epoch": 4.32, + "learning_rate": 1.3964640628611048e-05, + "loss": 0.0, + "step": 18711 + }, + { + "epoch": 4.32, + "learning_rate": 1.3962714736923197e-05, + "loss": 0.0, + "step": 18712 + }, + { + "epoch": 4.32, + "learning_rate": 1.3960788845235346e-05, + "loss": 0.0, + "step": 18713 + }, + { + "epoch": 4.32, + "learning_rate": 1.3958862953547493e-05, + "loss": 0.0, + "step": 18714 + }, + { + "epoch": 4.33, + "learning_rate": 1.3956937061859643e-05, + "loss": 0.0, + "step": 18715 + }, + { + "epoch": 4.33, + "learning_rate": 1.3955011170171788e-05, + "loss": 0.0, + "step": 18716 + }, + { + "epoch": 4.33, + "learning_rate": 1.3953085278483937e-05, + "loss": 0.0, + "step": 18717 + }, + { + "epoch": 4.33, + "learning_rate": 1.3951159386796087e-05, + "loss": 0.0, + "step": 18718 + }, + { + "epoch": 4.33, + "learning_rate": 1.3949233495108236e-05, + "loss": 0.0, + "step": 18719 + }, + { + "epoch": 4.33, + "learning_rate": 1.3947307603420385e-05, + "loss": 0.0, + "step": 18720 + }, + { + "epoch": 4.33, + "learning_rate": 1.3945381711732532e-05, + "loss": 0.0, + "step": 18721 + }, + { + "epoch": 4.33, + "learning_rate": 1.3943455820044682e-05, + "loss": 0.0, + "step": 18722 + }, + { + "epoch": 4.33, + "learning_rate": 1.394152992835683e-05, + "loss": 0.0, + "step": 18723 + }, + { + "epoch": 4.33, + "learning_rate": 1.393960403666898e-05, + "loss": 0.0, + "step": 18724 + }, + { + "epoch": 4.33, + "learning_rate": 1.3937678144981126e-05, + "loss": 0.0, + "step": 18725 + }, + { + "epoch": 4.33, + "learning_rate": 1.3935752253293275e-05, + "loss": 0.0, + "step": 18726 + }, + { + "epoch": 4.33, + "learning_rate": 1.3933826361605422e-05, + "loss": 0.0, + "step": 18727 + }, + { + "epoch": 4.33, + "learning_rate": 1.3931900469917572e-05, + "loss": 0.0, + "step": 18728 + }, + { + "epoch": 4.33, + "learning_rate": 1.392997457822972e-05, + "loss": 0.0, + "step": 18729 + }, + { + "epoch": 4.33, + "learning_rate": 1.392804868654187e-05, + "loss": 0.0, + "step": 18730 + }, + { + "epoch": 4.33, + "learning_rate": 1.3926122794854019e-05, + "loss": 0.0, + "step": 18731 + }, + { + "epoch": 4.33, + "learning_rate": 1.3924196903166167e-05, + "loss": 0.0, + "step": 18732 + }, + { + "epoch": 4.33, + "learning_rate": 1.3922271011478316e-05, + "loss": 0.0, + "step": 18733 + }, + { + "epoch": 4.33, + "learning_rate": 1.3920345119790465e-05, + "loss": 0.0, + "step": 18734 + }, + { + "epoch": 4.33, + "learning_rate": 1.391841922810261e-05, + "loss": 0.0, + "step": 18735 + }, + { + "epoch": 4.33, + "learning_rate": 1.391649333641476e-05, + "loss": 0.0, + "step": 18736 + }, + { + "epoch": 4.33, + "learning_rate": 1.3914567444726909e-05, + "loss": 0.0, + "step": 18737 + }, + { + "epoch": 4.33, + "learning_rate": 1.3912641553039058e-05, + "loss": 0.0, + "step": 18738 + }, + { + "epoch": 4.33, + "learning_rate": 1.3910715661351206e-05, + "loss": 0.0, + "step": 18739 + }, + { + "epoch": 4.33, + "learning_rate": 1.3908789769663355e-05, + "loss": 0.0, + "step": 18740 + }, + { + "epoch": 4.33, + "learning_rate": 1.3906863877975504e-05, + "loss": 0.0, + "step": 18741 + }, + { + "epoch": 4.33, + "learning_rate": 1.3904937986287653e-05, + "loss": 0.0, + "step": 18742 + }, + { + "epoch": 4.33, + "learning_rate": 1.3903012094599802e-05, + "loss": 0.0, + "step": 18743 + }, + { + "epoch": 4.33, + "learning_rate": 1.3901086202911948e-05, + "loss": 0.0, + "step": 18744 + }, + { + "epoch": 4.33, + "learning_rate": 1.3899160311224096e-05, + "loss": 0.0001, + "step": 18745 + }, + { + "epoch": 4.33, + "learning_rate": 1.3897234419536245e-05, + "loss": 0.0, + "step": 18746 + }, + { + "epoch": 4.33, + "learning_rate": 1.3895308527848394e-05, + "loss": 0.0, + "step": 18747 + }, + { + "epoch": 4.33, + "learning_rate": 1.3893382636160543e-05, + "loss": 0.0, + "step": 18748 + }, + { + "epoch": 4.33, + "learning_rate": 1.3891456744472692e-05, + "loss": 0.0, + "step": 18749 + }, + { + "epoch": 4.33, + "learning_rate": 1.388953085278484e-05, + "loss": 0.0, + "step": 18750 + }, + { + "epoch": 4.33, + "learning_rate": 1.3887604961096989e-05, + "loss": 0.0003, + "step": 18751 + }, + { + "epoch": 4.33, + "learning_rate": 1.3885679069409138e-05, + "loss": 0.0, + "step": 18752 + }, + { + "epoch": 4.33, + "learning_rate": 1.3883753177721284e-05, + "loss": 0.0, + "step": 18753 + }, + { + "epoch": 4.33, + "learning_rate": 1.3881827286033433e-05, + "loss": 0.0, + "step": 18754 + }, + { + "epoch": 4.33, + "learning_rate": 1.3879901394345582e-05, + "loss": 0.0, + "step": 18755 + }, + { + "epoch": 4.33, + "learning_rate": 1.3877975502657731e-05, + "loss": 0.0, + "step": 18756 + }, + { + "epoch": 4.33, + "learning_rate": 1.3876049610969879e-05, + "loss": 0.0, + "step": 18757 + }, + { + "epoch": 4.34, + "learning_rate": 1.3874123719282028e-05, + "loss": 0.0, + "step": 18758 + }, + { + "epoch": 4.34, + "learning_rate": 1.3872197827594177e-05, + "loss": 0.0, + "step": 18759 + }, + { + "epoch": 4.34, + "learning_rate": 1.3870271935906326e-05, + "loss": 0.0, + "step": 18760 + }, + { + "epoch": 4.34, + "learning_rate": 1.3868346044218476e-05, + "loss": 0.0, + "step": 18761 + }, + { + "epoch": 4.34, + "learning_rate": 1.3866420152530623e-05, + "loss": 0.0, + "step": 18762 + }, + { + "epoch": 4.34, + "learning_rate": 1.3864494260842769e-05, + "loss": 0.0, + "step": 18763 + }, + { + "epoch": 4.34, + "learning_rate": 1.3862568369154918e-05, + "loss": 0.0, + "step": 18764 + }, + { + "epoch": 4.34, + "learning_rate": 1.3860642477467067e-05, + "loss": 0.0, + "step": 18765 + }, + { + "epoch": 4.34, + "learning_rate": 1.3858716585779216e-05, + "loss": 0.0, + "step": 18766 + }, + { + "epoch": 4.34, + "learning_rate": 1.3856790694091366e-05, + "loss": 0.0, + "step": 18767 + }, + { + "epoch": 4.34, + "learning_rate": 1.3854864802403513e-05, + "loss": 0.0, + "step": 18768 + }, + { + "epoch": 4.34, + "learning_rate": 1.3852938910715662e-05, + "loss": 0.0001, + "step": 18769 + }, + { + "epoch": 4.34, + "learning_rate": 1.3851013019027811e-05, + "loss": 0.0, + "step": 18770 + }, + { + "epoch": 4.34, + "learning_rate": 1.384908712733996e-05, + "loss": 0.0, + "step": 18771 + }, + { + "epoch": 4.34, + "learning_rate": 1.3847161235652106e-05, + "loss": 0.0, + "step": 18772 + }, + { + "epoch": 4.34, + "learning_rate": 1.3845235343964256e-05, + "loss": 0.0001, + "step": 18773 + }, + { + "epoch": 4.34, + "learning_rate": 1.3843309452276405e-05, + "loss": 0.0, + "step": 18774 + }, + { + "epoch": 4.34, + "learning_rate": 1.3841383560588552e-05, + "loss": 0.0, + "step": 18775 + }, + { + "epoch": 4.34, + "learning_rate": 1.3839457668900701e-05, + "loss": 0.0001, + "step": 18776 + }, + { + "epoch": 4.34, + "learning_rate": 1.383753177721285e-05, + "loss": 0.0, + "step": 18777 + }, + { + "epoch": 4.34, + "learning_rate": 1.3835605885525e-05, + "loss": 0.0, + "step": 18778 + }, + { + "epoch": 4.34, + "learning_rate": 1.3833679993837149e-05, + "loss": 0.0, + "step": 18779 + }, + { + "epoch": 4.34, + "learning_rate": 1.3831754102149296e-05, + "loss": 0.0, + "step": 18780 + }, + { + "epoch": 4.34, + "learning_rate": 1.3829828210461442e-05, + "loss": 0.0, + "step": 18781 + }, + { + "epoch": 4.34, + "learning_rate": 1.3827902318773591e-05, + "loss": 0.0, + "step": 18782 + }, + { + "epoch": 4.34, + "learning_rate": 1.382597642708574e-05, + "loss": 0.0, + "step": 18783 + }, + { + "epoch": 4.34, + "learning_rate": 1.382405053539789e-05, + "loss": 0.0, + "step": 18784 + }, + { + "epoch": 4.34, + "learning_rate": 1.3822124643710039e-05, + "loss": 0.0, + "step": 18785 + }, + { + "epoch": 4.34, + "learning_rate": 1.3820198752022186e-05, + "loss": 0.0, + "step": 18786 + }, + { + "epoch": 4.34, + "learning_rate": 1.3818272860334335e-05, + "loss": 0.0, + "step": 18787 + }, + { + "epoch": 4.34, + "learning_rate": 1.3816346968646485e-05, + "loss": 0.0, + "step": 18788 + }, + { + "epoch": 4.34, + "learning_rate": 1.3814421076958634e-05, + "loss": 0.0, + "step": 18789 + }, + { + "epoch": 4.34, + "learning_rate": 1.3812495185270783e-05, + "loss": 0.0, + "step": 18790 + }, + { + "epoch": 4.34, + "learning_rate": 1.3810569293582929e-05, + "loss": 0.0, + "step": 18791 + }, + { + "epoch": 4.34, + "learning_rate": 1.3808643401895078e-05, + "loss": 0.0, + "step": 18792 + }, + { + "epoch": 4.34, + "learning_rate": 1.3806717510207225e-05, + "loss": 0.0, + "step": 18793 + }, + { + "epoch": 4.34, + "learning_rate": 1.3804791618519375e-05, + "loss": 0.0, + "step": 18794 + }, + { + "epoch": 4.34, + "learning_rate": 1.3802865726831524e-05, + "loss": 0.0, + "step": 18795 + }, + { + "epoch": 4.34, + "learning_rate": 1.3800939835143673e-05, + "loss": 0.0, + "step": 18796 + }, + { + "epoch": 4.34, + "learning_rate": 1.3799013943455822e-05, + "loss": 0.0, + "step": 18797 + }, + { + "epoch": 4.34, + "learning_rate": 1.379708805176797e-05, + "loss": 0.0, + "step": 18798 + }, + { + "epoch": 4.34, + "learning_rate": 1.3795162160080119e-05, + "loss": 0.0, + "step": 18799 + }, + { + "epoch": 4.34, + "learning_rate": 1.3793236268392265e-05, + "loss": 0.0, + "step": 18800 + }, + { + "epoch": 4.35, + "learning_rate": 1.3791310376704414e-05, + "loss": 0.0, + "step": 18801 + }, + { + "epoch": 4.35, + "learning_rate": 1.3789384485016563e-05, + "loss": 0.0, + "step": 18802 + }, + { + "epoch": 4.35, + "learning_rate": 1.3787458593328712e-05, + "loss": 0.0, + "step": 18803 + }, + { + "epoch": 4.35, + "learning_rate": 1.378553270164086e-05, + "loss": 0.0, + "step": 18804 + }, + { + "epoch": 4.35, + "learning_rate": 1.3783606809953009e-05, + "loss": 0.0, + "step": 18805 + }, + { + "epoch": 4.35, + "learning_rate": 1.3781680918265158e-05, + "loss": 0.0, + "step": 18806 + }, + { + "epoch": 4.35, + "learning_rate": 1.3779755026577307e-05, + "loss": 0.0, + "step": 18807 + }, + { + "epoch": 4.35, + "learning_rate": 1.3777829134889456e-05, + "loss": 0.0, + "step": 18808 + }, + { + "epoch": 4.35, + "learning_rate": 1.3775903243201602e-05, + "loss": 0.0, + "step": 18809 + }, + { + "epoch": 4.35, + "learning_rate": 1.3773977351513751e-05, + "loss": 0.0, + "step": 18810 + }, + { + "epoch": 4.35, + "learning_rate": 1.3772051459825899e-05, + "loss": 0.0, + "step": 18811 + }, + { + "epoch": 4.35, + "learning_rate": 1.3770125568138048e-05, + "loss": 0.0, + "step": 18812 + }, + { + "epoch": 4.35, + "learning_rate": 1.3768199676450197e-05, + "loss": 0.0, + "step": 18813 + }, + { + "epoch": 4.35, + "learning_rate": 1.3766273784762346e-05, + "loss": 0.0, + "step": 18814 + }, + { + "epoch": 4.35, + "learning_rate": 1.3764347893074495e-05, + "loss": 0.0, + "step": 18815 + }, + { + "epoch": 4.35, + "learning_rate": 1.3762422001386643e-05, + "loss": 0.0, + "step": 18816 + }, + { + "epoch": 4.35, + "learning_rate": 1.3760496109698792e-05, + "loss": 0.0, + "step": 18817 + }, + { + "epoch": 4.35, + "learning_rate": 1.3758570218010941e-05, + "loss": 0.0, + "step": 18818 + }, + { + "epoch": 4.35, + "learning_rate": 1.3756644326323087e-05, + "loss": 0.0, + "step": 18819 + }, + { + "epoch": 4.35, + "learning_rate": 1.3754718434635236e-05, + "loss": 0.0, + "step": 18820 + }, + { + "epoch": 4.35, + "learning_rate": 1.3752792542947385e-05, + "loss": 0.0, + "step": 18821 + }, + { + "epoch": 4.35, + "learning_rate": 1.3750866651259533e-05, + "loss": 0.0, + "step": 18822 + }, + { + "epoch": 4.35, + "learning_rate": 1.3748940759571682e-05, + "loss": 0.0001, + "step": 18823 + }, + { + "epoch": 4.35, + "learning_rate": 1.3747014867883831e-05, + "loss": 0.0, + "step": 18824 + }, + { + "epoch": 4.35, + "learning_rate": 1.374508897619598e-05, + "loss": 0.0, + "step": 18825 + }, + { + "epoch": 4.35, + "learning_rate": 1.374316308450813e-05, + "loss": 0.0, + "step": 18826 + }, + { + "epoch": 4.35, + "learning_rate": 1.3741237192820277e-05, + "loss": 0.0, + "step": 18827 + }, + { + "epoch": 4.35, + "learning_rate": 1.3739311301132424e-05, + "loss": 0.0, + "step": 18828 + }, + { + "epoch": 4.35, + "learning_rate": 1.3737385409444572e-05, + "loss": 0.0, + "step": 18829 + }, + { + "epoch": 4.35, + "learning_rate": 1.3735459517756721e-05, + "loss": 0.0, + "step": 18830 + }, + { + "epoch": 4.35, + "learning_rate": 1.373353362606887e-05, + "loss": 0.0, + "step": 18831 + }, + { + "epoch": 4.35, + "learning_rate": 1.373160773438102e-05, + "loss": 0.0, + "step": 18832 + }, + { + "epoch": 4.35, + "learning_rate": 1.3729681842693169e-05, + "loss": 0.0, + "step": 18833 + }, + { + "epoch": 4.35, + "learning_rate": 1.3727755951005316e-05, + "loss": 0.0, + "step": 18834 + }, + { + "epoch": 4.35, + "learning_rate": 1.3725830059317465e-05, + "loss": 0.0, + "step": 18835 + }, + { + "epoch": 4.35, + "learning_rate": 1.3723904167629614e-05, + "loss": 0.0001, + "step": 18836 + }, + { + "epoch": 4.35, + "learning_rate": 1.372197827594176e-05, + "loss": 0.0, + "step": 18837 + }, + { + "epoch": 4.35, + "learning_rate": 1.372005238425391e-05, + "loss": 0.0, + "step": 18838 + }, + { + "epoch": 4.35, + "learning_rate": 1.3718126492566059e-05, + "loss": 0.0, + "step": 18839 + }, + { + "epoch": 4.35, + "learning_rate": 1.3716200600878206e-05, + "loss": 0.0, + "step": 18840 + }, + { + "epoch": 4.35, + "learning_rate": 1.3714274709190355e-05, + "loss": 0.0, + "step": 18841 + }, + { + "epoch": 4.35, + "learning_rate": 1.3712348817502504e-05, + "loss": 0.0, + "step": 18842 + }, + { + "epoch": 4.35, + "learning_rate": 1.3710422925814654e-05, + "loss": 0.0, + "step": 18843 + }, + { + "epoch": 4.35, + "learning_rate": 1.3708497034126803e-05, + "loss": 0.0, + "step": 18844 + }, + { + "epoch": 4.36, + "learning_rate": 1.370657114243895e-05, + "loss": 0.0, + "step": 18845 + }, + { + "epoch": 4.36, + "learning_rate": 1.37046452507511e-05, + "loss": 0.0, + "step": 18846 + }, + { + "epoch": 4.36, + "learning_rate": 1.3702719359063245e-05, + "loss": 0.0, + "step": 18847 + }, + { + "epoch": 4.36, + "learning_rate": 1.3700793467375394e-05, + "loss": 0.0, + "step": 18848 + }, + { + "epoch": 4.36, + "learning_rate": 1.3698867575687543e-05, + "loss": 0.0, + "step": 18849 + }, + { + "epoch": 4.36, + "learning_rate": 1.3696941683999693e-05, + "loss": 0.0, + "step": 18850 + }, + { + "epoch": 4.36, + "learning_rate": 1.3695015792311842e-05, + "loss": 0.0, + "step": 18851 + }, + { + "epoch": 4.36, + "learning_rate": 1.369308990062399e-05, + "loss": 0.0, + "step": 18852 + }, + { + "epoch": 4.36, + "learning_rate": 1.3691164008936138e-05, + "loss": 0.0, + "step": 18853 + }, + { + "epoch": 4.36, + "learning_rate": 1.3689238117248288e-05, + "loss": 0.0001, + "step": 18854 + }, + { + "epoch": 4.36, + "learning_rate": 1.3687312225560437e-05, + "loss": 0.0, + "step": 18855 + }, + { + "epoch": 4.36, + "learning_rate": 1.3685386333872583e-05, + "loss": 0.0, + "step": 18856 + }, + { + "epoch": 4.36, + "learning_rate": 1.3683460442184732e-05, + "loss": 0.0, + "step": 18857 + }, + { + "epoch": 4.36, + "learning_rate": 1.368153455049688e-05, + "loss": 0.0, + "step": 18858 + }, + { + "epoch": 4.36, + "learning_rate": 1.3679608658809028e-05, + "loss": 0.0, + "step": 18859 + }, + { + "epoch": 4.36, + "learning_rate": 1.3677682767121178e-05, + "loss": 0.0, + "step": 18860 + }, + { + "epoch": 4.36, + "learning_rate": 1.3675756875433327e-05, + "loss": 0.0, + "step": 18861 + }, + { + "epoch": 4.36, + "learning_rate": 1.3673830983745476e-05, + "loss": 0.0, + "step": 18862 + }, + { + "epoch": 4.36, + "learning_rate": 1.3671905092057623e-05, + "loss": 0.0, + "step": 18863 + }, + { + "epoch": 4.36, + "learning_rate": 1.3669979200369773e-05, + "loss": 0.0, + "step": 18864 + }, + { + "epoch": 4.36, + "learning_rate": 1.3668053308681918e-05, + "loss": 0.0, + "step": 18865 + }, + { + "epoch": 4.36, + "learning_rate": 1.3666127416994068e-05, + "loss": 0.0, + "step": 18866 + }, + { + "epoch": 4.36, + "learning_rate": 1.3664201525306217e-05, + "loss": 0.0, + "step": 18867 + }, + { + "epoch": 4.36, + "learning_rate": 1.3662275633618366e-05, + "loss": 0.0001, + "step": 18868 + }, + { + "epoch": 4.36, + "learning_rate": 1.3660349741930515e-05, + "loss": 0.0, + "step": 18869 + }, + { + "epoch": 4.36, + "learning_rate": 1.3658423850242663e-05, + "loss": 0.0, + "step": 18870 + }, + { + "epoch": 4.36, + "learning_rate": 1.3656497958554812e-05, + "loss": 0.0, + "step": 18871 + }, + { + "epoch": 4.36, + "learning_rate": 1.3654572066866961e-05, + "loss": 0.0, + "step": 18872 + }, + { + "epoch": 4.36, + "learning_rate": 1.365264617517911e-05, + "loss": 0.0, + "step": 18873 + }, + { + "epoch": 4.36, + "learning_rate": 1.365072028349126e-05, + "loss": 0.0, + "step": 18874 + }, + { + "epoch": 4.36, + "learning_rate": 1.3648794391803405e-05, + "loss": 0.0001, + "step": 18875 + }, + { + "epoch": 4.36, + "learning_rate": 1.3646868500115552e-05, + "loss": 0.0, + "step": 18876 + }, + { + "epoch": 4.36, + "learning_rate": 1.3644942608427702e-05, + "loss": 0.0, + "step": 18877 + }, + { + "epoch": 4.36, + "learning_rate": 1.364301671673985e-05, + "loss": 0.0, + "step": 18878 + }, + { + "epoch": 4.36, + "learning_rate": 1.3641090825052e-05, + "loss": 0.0, + "step": 18879 + }, + { + "epoch": 4.36, + "learning_rate": 1.363916493336415e-05, + "loss": 0.0, + "step": 18880 + }, + { + "epoch": 4.36, + "learning_rate": 1.3637239041676297e-05, + "loss": 0.0, + "step": 18881 + }, + { + "epoch": 4.36, + "learning_rate": 1.3635313149988446e-05, + "loss": 0.0, + "step": 18882 + }, + { + "epoch": 4.36, + "learning_rate": 1.3633387258300595e-05, + "loss": 0.0, + "step": 18883 + }, + { + "epoch": 4.36, + "learning_rate": 1.363146136661274e-05, + "loss": 0.0, + "step": 18884 + }, + { + "epoch": 4.36, + "learning_rate": 1.362953547492489e-05, + "loss": 0.0, + "step": 18885 + }, + { + "epoch": 4.36, + "learning_rate": 1.3627609583237039e-05, + "loss": 0.0, + "step": 18886 + }, + { + "epoch": 4.36, + "learning_rate": 1.3625683691549188e-05, + "loss": 0.0, + "step": 18887 + }, + { + "epoch": 4.37, + "learning_rate": 1.3623757799861336e-05, + "loss": 0.0, + "step": 18888 + }, + { + "epoch": 4.37, + "learning_rate": 1.3621831908173485e-05, + "loss": 0.0, + "step": 18889 + }, + { + "epoch": 4.37, + "learning_rate": 1.3619906016485634e-05, + "loss": 0.0, + "step": 18890 + }, + { + "epoch": 4.37, + "learning_rate": 1.3617980124797783e-05, + "loss": 0.0, + "step": 18891 + }, + { + "epoch": 4.37, + "learning_rate": 1.3616054233109932e-05, + "loss": 0.0, + "step": 18892 + }, + { + "epoch": 4.37, + "learning_rate": 1.3614128341422078e-05, + "loss": 0.0, + "step": 18893 + }, + { + "epoch": 4.37, + "learning_rate": 1.3612202449734226e-05, + "loss": 0.0, + "step": 18894 + }, + { + "epoch": 4.37, + "learning_rate": 1.3610276558046375e-05, + "loss": 0.0, + "step": 18895 + }, + { + "epoch": 4.37, + "learning_rate": 1.3608350666358524e-05, + "loss": 0.0, + "step": 18896 + }, + { + "epoch": 4.37, + "learning_rate": 1.3606424774670673e-05, + "loss": 0.0, + "step": 18897 + }, + { + "epoch": 4.37, + "learning_rate": 1.3604498882982822e-05, + "loss": 0.0, + "step": 18898 + }, + { + "epoch": 4.37, + "learning_rate": 1.360257299129497e-05, + "loss": 0.0, + "step": 18899 + }, + { + "epoch": 4.37, + "learning_rate": 1.3600647099607119e-05, + "loss": 0.0, + "step": 18900 + }, + { + "epoch": 4.37, + "learning_rate": 1.3598721207919268e-05, + "loss": 0.0, + "step": 18901 + }, + { + "epoch": 4.37, + "learning_rate": 1.3596795316231417e-05, + "loss": 0.0, + "step": 18902 + }, + { + "epoch": 4.37, + "learning_rate": 1.3594869424543563e-05, + "loss": 0.0, + "step": 18903 + }, + { + "epoch": 4.37, + "learning_rate": 1.3592943532855712e-05, + "loss": 0.0, + "step": 18904 + }, + { + "epoch": 4.37, + "learning_rate": 1.3591017641167862e-05, + "loss": 0.0, + "step": 18905 + }, + { + "epoch": 4.37, + "learning_rate": 1.3589091749480009e-05, + "loss": 0.0, + "step": 18906 + }, + { + "epoch": 4.37, + "learning_rate": 1.3587165857792158e-05, + "loss": 0.0, + "step": 18907 + }, + { + "epoch": 4.37, + "learning_rate": 1.3585239966104307e-05, + "loss": 0.0, + "step": 18908 + }, + { + "epoch": 4.37, + "learning_rate": 1.3583314074416457e-05, + "loss": 0.0, + "step": 18909 + }, + { + "epoch": 4.37, + "learning_rate": 1.3581388182728606e-05, + "loss": 0.0, + "step": 18910 + }, + { + "epoch": 4.37, + "learning_rate": 1.3579462291040753e-05, + "loss": 0.0, + "step": 18911 + }, + { + "epoch": 4.37, + "learning_rate": 1.3577536399352899e-05, + "loss": 0.0, + "step": 18912 + }, + { + "epoch": 4.37, + "learning_rate": 1.3575610507665048e-05, + "loss": 0.0001, + "step": 18913 + }, + { + "epoch": 4.37, + "learning_rate": 1.3573684615977197e-05, + "loss": 0.0001, + "step": 18914 + }, + { + "epoch": 4.37, + "learning_rate": 1.3571758724289346e-05, + "loss": 0.0, + "step": 18915 + }, + { + "epoch": 4.37, + "learning_rate": 1.3569832832601496e-05, + "loss": 0.0, + "step": 18916 + }, + { + "epoch": 4.37, + "learning_rate": 1.3567906940913643e-05, + "loss": 0.0, + "step": 18917 + }, + { + "epoch": 4.37, + "learning_rate": 1.3565981049225792e-05, + "loss": 0.0, + "step": 18918 + }, + { + "epoch": 4.37, + "learning_rate": 1.3564055157537941e-05, + "loss": 0.0, + "step": 18919 + }, + { + "epoch": 4.37, + "learning_rate": 1.356212926585009e-05, + "loss": 0.0002, + "step": 18920 + }, + { + "epoch": 4.37, + "learning_rate": 1.3560203374162236e-05, + "loss": 0.0, + "step": 18921 + }, + { + "epoch": 4.37, + "learning_rate": 1.3558277482474386e-05, + "loss": 0.0, + "step": 18922 + }, + { + "epoch": 4.37, + "learning_rate": 1.3556351590786535e-05, + "loss": 0.0, + "step": 18923 + }, + { + "epoch": 4.37, + "learning_rate": 1.3554425699098682e-05, + "loss": 0.0, + "step": 18924 + }, + { + "epoch": 4.37, + "learning_rate": 1.3552499807410831e-05, + "loss": 0.0, + "step": 18925 + }, + { + "epoch": 4.37, + "learning_rate": 1.355057391572298e-05, + "loss": 0.0, + "step": 18926 + }, + { + "epoch": 4.37, + "learning_rate": 1.354864802403513e-05, + "loss": 0.0, + "step": 18927 + }, + { + "epoch": 4.37, + "learning_rate": 1.3546722132347279e-05, + "loss": 0.0, + "step": 18928 + }, + { + "epoch": 4.37, + "learning_rate": 1.3544796240659426e-05, + "loss": 0.0, + "step": 18929 + }, + { + "epoch": 4.37, + "learning_rate": 1.3542870348971576e-05, + "loss": 0.0, + "step": 18930 + }, + { + "epoch": 4.38, + "learning_rate": 1.3540944457283721e-05, + "loss": 0.0, + "step": 18931 + }, + { + "epoch": 4.38, + "learning_rate": 1.353901856559587e-05, + "loss": 0.0, + "step": 18932 + }, + { + "epoch": 4.38, + "learning_rate": 1.353709267390802e-05, + "loss": 0.0, + "step": 18933 + }, + { + "epoch": 4.38, + "learning_rate": 1.3535166782220169e-05, + "loss": 0.0, + "step": 18934 + }, + { + "epoch": 4.38, + "learning_rate": 1.3533240890532316e-05, + "loss": 0.0, + "step": 18935 + }, + { + "epoch": 4.38, + "learning_rate": 1.3531314998844466e-05, + "loss": 0.0, + "step": 18936 + }, + { + "epoch": 4.38, + "learning_rate": 1.3529389107156615e-05, + "loss": 0.0, + "step": 18937 + }, + { + "epoch": 4.38, + "learning_rate": 1.3527463215468764e-05, + "loss": 0.0, + "step": 18938 + }, + { + "epoch": 4.38, + "learning_rate": 1.3525537323780913e-05, + "loss": 0.0, + "step": 18939 + }, + { + "epoch": 4.38, + "learning_rate": 1.3523611432093059e-05, + "loss": 0.0, + "step": 18940 + }, + { + "epoch": 4.38, + "learning_rate": 1.3521685540405208e-05, + "loss": 0.0, + "step": 18941 + }, + { + "epoch": 4.38, + "learning_rate": 1.3519759648717355e-05, + "loss": 0.0, + "step": 18942 + }, + { + "epoch": 4.38, + "learning_rate": 1.3517833757029505e-05, + "loss": 0.0, + "step": 18943 + }, + { + "epoch": 4.38, + "learning_rate": 1.3515907865341654e-05, + "loss": 0.0, + "step": 18944 + }, + { + "epoch": 4.38, + "learning_rate": 1.3513981973653803e-05, + "loss": 0.0, + "step": 18945 + }, + { + "epoch": 4.38, + "learning_rate": 1.3512056081965952e-05, + "loss": 0.0, + "step": 18946 + }, + { + "epoch": 4.38, + "learning_rate": 1.35101301902781e-05, + "loss": 0.0, + "step": 18947 + }, + { + "epoch": 4.38, + "learning_rate": 1.3508204298590249e-05, + "loss": 0.0, + "step": 18948 + }, + { + "epoch": 4.38, + "learning_rate": 1.3506278406902395e-05, + "loss": 0.0, + "step": 18949 + }, + { + "epoch": 4.38, + "learning_rate": 1.3504352515214544e-05, + "loss": 0.0, + "step": 18950 + }, + { + "epoch": 4.38, + "learning_rate": 1.3502426623526693e-05, + "loss": 0.0, + "step": 18951 + }, + { + "epoch": 4.38, + "learning_rate": 1.3500500731838842e-05, + "loss": 0.0, + "step": 18952 + }, + { + "epoch": 4.38, + "learning_rate": 1.349857484015099e-05, + "loss": 0.0, + "step": 18953 + }, + { + "epoch": 4.38, + "learning_rate": 1.3496648948463139e-05, + "loss": 0.0, + "step": 18954 + }, + { + "epoch": 4.38, + "learning_rate": 1.3494723056775288e-05, + "loss": 0.0, + "step": 18955 + }, + { + "epoch": 4.38, + "learning_rate": 1.3492797165087437e-05, + "loss": 0.0, + "step": 18956 + }, + { + "epoch": 4.38, + "learning_rate": 1.3490871273399586e-05, + "loss": 0.0, + "step": 18957 + }, + { + "epoch": 4.38, + "learning_rate": 1.3488945381711734e-05, + "loss": 0.0, + "step": 18958 + }, + { + "epoch": 4.38, + "learning_rate": 1.3487019490023881e-05, + "loss": 0.0, + "step": 18959 + }, + { + "epoch": 4.38, + "learning_rate": 1.3485093598336029e-05, + "loss": 0.0, + "step": 18960 + }, + { + "epoch": 4.38, + "learning_rate": 1.3483167706648178e-05, + "loss": 0.0, + "step": 18961 + }, + { + "epoch": 4.38, + "learning_rate": 1.3481241814960327e-05, + "loss": 0.0, + "step": 18962 + }, + { + "epoch": 4.38, + "learning_rate": 1.3479315923272476e-05, + "loss": 0.0, + "step": 18963 + }, + { + "epoch": 4.38, + "learning_rate": 1.3477390031584625e-05, + "loss": 0.0, + "step": 18964 + }, + { + "epoch": 4.38, + "learning_rate": 1.3475464139896773e-05, + "loss": 0.0001, + "step": 18965 + }, + { + "epoch": 4.38, + "learning_rate": 1.3473538248208922e-05, + "loss": 0.0, + "step": 18966 + }, + { + "epoch": 4.38, + "learning_rate": 1.3471612356521071e-05, + "loss": 0.0, + "step": 18967 + }, + { + "epoch": 4.38, + "learning_rate": 1.3469686464833217e-05, + "loss": 0.0, + "step": 18968 + }, + { + "epoch": 4.38, + "learning_rate": 1.3467760573145366e-05, + "loss": 0.0, + "step": 18969 + }, + { + "epoch": 4.38, + "learning_rate": 1.3465834681457515e-05, + "loss": 0.0, + "step": 18970 + }, + { + "epoch": 4.38, + "learning_rate": 1.3463908789769663e-05, + "loss": 0.0, + "step": 18971 + }, + { + "epoch": 4.38, + "learning_rate": 1.3461982898081812e-05, + "loss": 0.0, + "step": 18972 + }, + { + "epoch": 4.38, + "learning_rate": 1.3460057006393961e-05, + "loss": 0.0001, + "step": 18973 + }, + { + "epoch": 4.39, + "learning_rate": 1.345813111470611e-05, + "loss": 0.0, + "step": 18974 + }, + { + "epoch": 4.39, + "learning_rate": 1.345620522301826e-05, + "loss": 0.0, + "step": 18975 + }, + { + "epoch": 4.39, + "learning_rate": 1.3454279331330407e-05, + "loss": 0.0, + "step": 18976 + }, + { + "epoch": 4.39, + "learning_rate": 1.3452353439642554e-05, + "loss": 0.0, + "step": 18977 + }, + { + "epoch": 4.39, + "learning_rate": 1.3450427547954702e-05, + "loss": 0.0, + "step": 18978 + }, + { + "epoch": 4.39, + "learning_rate": 1.3448501656266851e-05, + "loss": 0.0, + "step": 18979 + }, + { + "epoch": 4.39, + "learning_rate": 1.3446575764579e-05, + "loss": 0.0, + "step": 18980 + }, + { + "epoch": 4.39, + "learning_rate": 1.344464987289115e-05, + "loss": 0.0, + "step": 18981 + }, + { + "epoch": 4.39, + "learning_rate": 1.3442723981203299e-05, + "loss": 0.0, + "step": 18982 + }, + { + "epoch": 4.39, + "learning_rate": 1.3440798089515446e-05, + "loss": 0.0, + "step": 18983 + }, + { + "epoch": 4.39, + "learning_rate": 1.3438872197827595e-05, + "loss": 0.0, + "step": 18984 + }, + { + "epoch": 4.39, + "learning_rate": 1.3436946306139744e-05, + "loss": 0.0, + "step": 18985 + }, + { + "epoch": 4.39, + "learning_rate": 1.3435020414451894e-05, + "loss": 0.0, + "step": 18986 + }, + { + "epoch": 4.39, + "learning_rate": 1.343309452276404e-05, + "loss": 0.0, + "step": 18987 + }, + { + "epoch": 4.39, + "learning_rate": 1.3431168631076189e-05, + "loss": 0.0, + "step": 18988 + }, + { + "epoch": 4.39, + "learning_rate": 1.3429242739388336e-05, + "loss": 0.0, + "step": 18989 + }, + { + "epoch": 4.39, + "learning_rate": 1.3427316847700485e-05, + "loss": 0.0, + "step": 18990 + }, + { + "epoch": 4.39, + "learning_rate": 1.3425390956012634e-05, + "loss": 0.0, + "step": 18991 + }, + { + "epoch": 4.39, + "learning_rate": 1.3423465064324784e-05, + "loss": 0.0, + "step": 18992 + }, + { + "epoch": 4.39, + "learning_rate": 1.3421539172636933e-05, + "loss": 0.0, + "step": 18993 + }, + { + "epoch": 4.39, + "learning_rate": 1.341961328094908e-05, + "loss": 0.0, + "step": 18994 + }, + { + "epoch": 4.39, + "learning_rate": 1.341768738926123e-05, + "loss": 0.0, + "step": 18995 + }, + { + "epoch": 4.39, + "learning_rate": 1.3415761497573375e-05, + "loss": 0.0, + "step": 18996 + }, + { + "epoch": 4.39, + "learning_rate": 1.3413835605885524e-05, + "loss": 0.0, + "step": 18997 + }, + { + "epoch": 4.39, + "learning_rate": 1.3411909714197674e-05, + "loss": 0.0, + "step": 18998 + }, + { + "epoch": 4.39, + "learning_rate": 1.3409983822509823e-05, + "loss": 0.0, + "step": 18999 + }, + { + "epoch": 4.39, + "learning_rate": 1.3408057930821972e-05, + "loss": 0.0, + "step": 19000 + }, + { + "epoch": 4.39, + "learning_rate": 1.340613203913412e-05, + "loss": 0.0, + "step": 19001 + }, + { + "epoch": 4.39, + "learning_rate": 1.3404206147446269e-05, + "loss": 0.0, + "step": 19002 + }, + { + "epoch": 4.39, + "learning_rate": 1.3402280255758418e-05, + "loss": 0.0, + "step": 19003 + }, + { + "epoch": 4.39, + "learning_rate": 1.3400354364070567e-05, + "loss": 0.0, + "step": 19004 + }, + { + "epoch": 4.39, + "learning_rate": 1.3398428472382713e-05, + "loss": 0.0, + "step": 19005 + }, + { + "epoch": 4.39, + "learning_rate": 1.3396502580694862e-05, + "loss": 0.0, + "step": 19006 + }, + { + "epoch": 4.39, + "learning_rate": 1.339457668900701e-05, + "loss": 0.0, + "step": 19007 + }, + { + "epoch": 4.39, + "learning_rate": 1.3392650797319158e-05, + "loss": 0.0, + "step": 19008 + }, + { + "epoch": 4.39, + "learning_rate": 1.3390724905631308e-05, + "loss": 0.0, + "step": 19009 + }, + { + "epoch": 4.39, + "learning_rate": 1.3388799013943457e-05, + "loss": 0.0, + "step": 19010 + }, + { + "epoch": 4.39, + "learning_rate": 1.3386873122255606e-05, + "loss": 0.0, + "step": 19011 + }, + { + "epoch": 4.39, + "learning_rate": 1.3384947230567753e-05, + "loss": 0.0, + "step": 19012 + }, + { + "epoch": 4.39, + "learning_rate": 1.3383021338879903e-05, + "loss": 0.0, + "step": 19013 + }, + { + "epoch": 4.39, + "learning_rate": 1.3381095447192052e-05, + "loss": 0.0, + "step": 19014 + }, + { + "epoch": 4.39, + "learning_rate": 1.3379169555504198e-05, + "loss": 0.0, + "step": 19015 + }, + { + "epoch": 4.39, + "learning_rate": 1.3377243663816347e-05, + "loss": 0.0, + "step": 19016 + }, + { + "epoch": 4.39, + "learning_rate": 1.3375317772128496e-05, + "loss": 0.0, + "step": 19017 + }, + { + "epoch": 4.4, + "learning_rate": 1.3373391880440645e-05, + "loss": 0.0, + "step": 19018 + }, + { + "epoch": 4.4, + "learning_rate": 1.3371465988752793e-05, + "loss": 0.0, + "step": 19019 + }, + { + "epoch": 4.4, + "learning_rate": 1.3369540097064942e-05, + "loss": 0.0, + "step": 19020 + }, + { + "epoch": 4.4, + "learning_rate": 1.3367614205377091e-05, + "loss": 0.0, + "step": 19021 + }, + { + "epoch": 4.4, + "learning_rate": 1.336568831368924e-05, + "loss": 0.0, + "step": 19022 + }, + { + "epoch": 4.4, + "learning_rate": 1.336376242200139e-05, + "loss": 0.0, + "step": 19023 + }, + { + "epoch": 4.4, + "learning_rate": 1.3361836530313535e-05, + "loss": 0.0, + "step": 19024 + }, + { + "epoch": 4.4, + "learning_rate": 1.3359910638625684e-05, + "loss": 0.0, + "step": 19025 + }, + { + "epoch": 4.4, + "learning_rate": 1.3357984746937832e-05, + "loss": 0.0, + "step": 19026 + }, + { + "epoch": 4.4, + "learning_rate": 1.3356058855249981e-05, + "loss": 0.0, + "step": 19027 + }, + { + "epoch": 4.4, + "learning_rate": 1.335413296356213e-05, + "loss": 0.0, + "step": 19028 + }, + { + "epoch": 4.4, + "learning_rate": 1.335220707187428e-05, + "loss": 0.0, + "step": 19029 + }, + { + "epoch": 4.4, + "learning_rate": 1.3350281180186427e-05, + "loss": 0.0, + "step": 19030 + }, + { + "epoch": 4.4, + "learning_rate": 1.3348355288498576e-05, + "loss": 0.0, + "step": 19031 + }, + { + "epoch": 4.4, + "learning_rate": 1.3346429396810725e-05, + "loss": 0.0, + "step": 19032 + }, + { + "epoch": 4.4, + "learning_rate": 1.334450350512287e-05, + "loss": 0.0, + "step": 19033 + }, + { + "epoch": 4.4, + "learning_rate": 1.334257761343502e-05, + "loss": 0.0, + "step": 19034 + }, + { + "epoch": 4.4, + "learning_rate": 1.3340651721747169e-05, + "loss": 0.0, + "step": 19035 + }, + { + "epoch": 4.4, + "learning_rate": 1.3338725830059318e-05, + "loss": 0.0, + "step": 19036 + }, + { + "epoch": 4.4, + "learning_rate": 1.3336799938371466e-05, + "loss": 0.0, + "step": 19037 + }, + { + "epoch": 4.4, + "learning_rate": 1.3334874046683615e-05, + "loss": 0.0, + "step": 19038 + }, + { + "epoch": 4.4, + "learning_rate": 1.3332948154995764e-05, + "loss": 0.0, + "step": 19039 + }, + { + "epoch": 4.4, + "learning_rate": 1.3331022263307913e-05, + "loss": 0.0, + "step": 19040 + }, + { + "epoch": 4.4, + "learning_rate": 1.3329096371620062e-05, + "loss": 0.0, + "step": 19041 + }, + { + "epoch": 4.4, + "learning_rate": 1.332717047993221e-05, + "loss": 0.0, + "step": 19042 + }, + { + "epoch": 4.4, + "learning_rate": 1.3325244588244357e-05, + "loss": 0.0, + "step": 19043 + }, + { + "epoch": 4.4, + "learning_rate": 1.3323318696556505e-05, + "loss": 0.0, + "step": 19044 + }, + { + "epoch": 4.4, + "learning_rate": 1.3321392804868654e-05, + "loss": 0.0, + "step": 19045 + }, + { + "epoch": 4.4, + "learning_rate": 1.3319466913180803e-05, + "loss": 0.0, + "step": 19046 + }, + { + "epoch": 4.4, + "learning_rate": 1.3317541021492952e-05, + "loss": 0.0, + "step": 19047 + }, + { + "epoch": 4.4, + "learning_rate": 1.33156151298051e-05, + "loss": 0.0, + "step": 19048 + }, + { + "epoch": 4.4, + "learning_rate": 1.3313689238117249e-05, + "loss": 0.0, + "step": 19049 + }, + { + "epoch": 4.4, + "learning_rate": 1.3311763346429398e-05, + "loss": 0.0, + "step": 19050 + }, + { + "epoch": 4.4, + "learning_rate": 1.3309837454741547e-05, + "loss": 0.0, + "step": 19051 + }, + { + "epoch": 4.4, + "learning_rate": 1.3307911563053693e-05, + "loss": 0.0, + "step": 19052 + }, + { + "epoch": 4.4, + "learning_rate": 1.3305985671365842e-05, + "loss": 0.0, + "step": 19053 + }, + { + "epoch": 4.4, + "learning_rate": 1.3304059779677992e-05, + "loss": 0.0, + "step": 19054 + }, + { + "epoch": 4.4, + "learning_rate": 1.3302133887990139e-05, + "loss": 0.0, + "step": 19055 + }, + { + "epoch": 4.4, + "learning_rate": 1.3300207996302288e-05, + "loss": 0.0, + "step": 19056 + }, + { + "epoch": 4.4, + "learning_rate": 1.3298282104614437e-05, + "loss": 0.0, + "step": 19057 + }, + { + "epoch": 4.4, + "learning_rate": 1.3296356212926587e-05, + "loss": 0.0001, + "step": 19058 + }, + { + "epoch": 4.4, + "learning_rate": 1.3294430321238736e-05, + "loss": 0.0, + "step": 19059 + }, + { + "epoch": 4.4, + "learning_rate": 1.3292504429550883e-05, + "loss": 0.0, + "step": 19060 + }, + { + "epoch": 4.41, + "learning_rate": 1.329057853786303e-05, + "loss": 0.0, + "step": 19061 + }, + { + "epoch": 4.41, + "learning_rate": 1.3288652646175178e-05, + "loss": 0.0, + "step": 19062 + }, + { + "epoch": 4.41, + "learning_rate": 1.3286726754487327e-05, + "loss": 0.0, + "step": 19063 + }, + { + "epoch": 4.41, + "learning_rate": 1.3284800862799477e-05, + "loss": 0.0, + "step": 19064 + }, + { + "epoch": 4.41, + "learning_rate": 1.3282874971111626e-05, + "loss": 0.0, + "step": 19065 + }, + { + "epoch": 4.41, + "learning_rate": 1.3280949079423773e-05, + "loss": 0.0, + "step": 19066 + }, + { + "epoch": 4.41, + "learning_rate": 1.3279023187735922e-05, + "loss": 0.0, + "step": 19067 + }, + { + "epoch": 4.41, + "learning_rate": 1.3277097296048071e-05, + "loss": 0.0, + "step": 19068 + }, + { + "epoch": 4.41, + "learning_rate": 1.327517140436022e-05, + "loss": 0.0, + "step": 19069 + }, + { + "epoch": 4.41, + "learning_rate": 1.327324551267237e-05, + "loss": 0.0, + "step": 19070 + }, + { + "epoch": 4.41, + "learning_rate": 1.3271319620984516e-05, + "loss": 0.0, + "step": 19071 + }, + { + "epoch": 4.41, + "learning_rate": 1.3269393729296665e-05, + "loss": 0.0, + "step": 19072 + }, + { + "epoch": 4.41, + "learning_rate": 1.3267467837608812e-05, + "loss": 0.0, + "step": 19073 + }, + { + "epoch": 4.41, + "learning_rate": 1.3265541945920961e-05, + "loss": 0.0, + "step": 19074 + }, + { + "epoch": 4.41, + "learning_rate": 1.326361605423311e-05, + "loss": 0.0, + "step": 19075 + }, + { + "epoch": 4.41, + "learning_rate": 1.326169016254526e-05, + "loss": 0.0, + "step": 19076 + }, + { + "epoch": 4.41, + "learning_rate": 1.3259764270857409e-05, + "loss": 0.0, + "step": 19077 + }, + { + "epoch": 4.41, + "learning_rate": 1.3257838379169556e-05, + "loss": 0.0, + "step": 19078 + }, + { + "epoch": 4.41, + "learning_rate": 1.3255912487481706e-05, + "loss": 0.0001, + "step": 19079 + }, + { + "epoch": 4.41, + "learning_rate": 1.3253986595793851e-05, + "loss": 0.0, + "step": 19080 + }, + { + "epoch": 4.41, + "learning_rate": 1.3252060704106e-05, + "loss": 0.0, + "step": 19081 + }, + { + "epoch": 4.41, + "learning_rate": 1.325013481241815e-05, + "loss": 0.0, + "step": 19082 + }, + { + "epoch": 4.41, + "learning_rate": 1.3248208920730299e-05, + "loss": 0.0001, + "step": 19083 + }, + { + "epoch": 4.41, + "learning_rate": 1.3246283029042446e-05, + "loss": 0.0, + "step": 19084 + }, + { + "epoch": 4.41, + "learning_rate": 1.3244357137354596e-05, + "loss": 0.0, + "step": 19085 + }, + { + "epoch": 4.41, + "learning_rate": 1.3242431245666745e-05, + "loss": 0.0, + "step": 19086 + }, + { + "epoch": 4.41, + "learning_rate": 1.3240505353978894e-05, + "loss": 0.0, + "step": 19087 + }, + { + "epoch": 4.41, + "learning_rate": 1.3238579462291043e-05, + "loss": 0.0, + "step": 19088 + }, + { + "epoch": 4.41, + "learning_rate": 1.3236653570603189e-05, + "loss": 0.0, + "step": 19089 + }, + { + "epoch": 4.41, + "learning_rate": 1.3234727678915338e-05, + "loss": 0.0, + "step": 19090 + }, + { + "epoch": 4.41, + "learning_rate": 1.3232801787227486e-05, + "loss": 0.0, + "step": 19091 + }, + { + "epoch": 4.41, + "learning_rate": 1.3230875895539635e-05, + "loss": 0.0, + "step": 19092 + }, + { + "epoch": 4.41, + "learning_rate": 1.3228950003851784e-05, + "loss": 0.0, + "step": 19093 + }, + { + "epoch": 4.41, + "learning_rate": 1.3227024112163933e-05, + "loss": 0.0, + "step": 19094 + }, + { + "epoch": 4.41, + "learning_rate": 1.3225098220476082e-05, + "loss": 0.0, + "step": 19095 + }, + { + "epoch": 4.41, + "learning_rate": 1.322317232878823e-05, + "loss": 0.0, + "step": 19096 + }, + { + "epoch": 4.41, + "learning_rate": 1.3221246437100379e-05, + "loss": 0.0, + "step": 19097 + }, + { + "epoch": 4.41, + "learning_rate": 1.3219320545412528e-05, + "loss": 0.0, + "step": 19098 + }, + { + "epoch": 4.41, + "learning_rate": 1.3217394653724674e-05, + "loss": 0.0, + "step": 19099 + }, + { + "epoch": 4.41, + "learning_rate": 1.3215468762036823e-05, + "loss": 0.0, + "step": 19100 + }, + { + "epoch": 4.41, + "learning_rate": 1.3213542870348972e-05, + "loss": 0.0, + "step": 19101 + }, + { + "epoch": 4.41, + "learning_rate": 1.3211616978661121e-05, + "loss": 0.0, + "step": 19102 + }, + { + "epoch": 4.41, + "learning_rate": 1.3209691086973269e-05, + "loss": 0.0, + "step": 19103 + }, + { + "epoch": 4.42, + "learning_rate": 1.3207765195285418e-05, + "loss": 0.0, + "step": 19104 + }, + { + "epoch": 4.42, + "learning_rate": 1.3205839303597567e-05, + "loss": 0.0, + "step": 19105 + }, + { + "epoch": 4.42, + "learning_rate": 1.3203913411909716e-05, + "loss": 0.0, + "step": 19106 + }, + { + "epoch": 4.42, + "learning_rate": 1.3201987520221864e-05, + "loss": 0.0, + "step": 19107 + }, + { + "epoch": 4.42, + "learning_rate": 1.3200061628534011e-05, + "loss": 0.0, + "step": 19108 + }, + { + "epoch": 4.42, + "learning_rate": 1.3198135736846159e-05, + "loss": 0.0, + "step": 19109 + }, + { + "epoch": 4.42, + "learning_rate": 1.3196209845158308e-05, + "loss": 0.0, + "step": 19110 + }, + { + "epoch": 4.42, + "learning_rate": 1.3194283953470457e-05, + "loss": 0.0, + "step": 19111 + }, + { + "epoch": 4.42, + "learning_rate": 1.3192358061782606e-05, + "loss": 0.0, + "step": 19112 + }, + { + "epoch": 4.42, + "learning_rate": 1.3190432170094755e-05, + "loss": 0.0, + "step": 19113 + }, + { + "epoch": 4.42, + "learning_rate": 1.3188506278406903e-05, + "loss": 0.0, + "step": 19114 + }, + { + "epoch": 4.42, + "learning_rate": 1.3186580386719052e-05, + "loss": 0.0, + "step": 19115 + }, + { + "epoch": 4.42, + "learning_rate": 1.3184654495031201e-05, + "loss": 0.0001, + "step": 19116 + }, + { + "epoch": 4.42, + "learning_rate": 1.3182728603343347e-05, + "loss": 0.0, + "step": 19117 + }, + { + "epoch": 4.42, + "learning_rate": 1.3180802711655496e-05, + "loss": 0.0, + "step": 19118 + }, + { + "epoch": 4.42, + "learning_rate": 1.3178876819967645e-05, + "loss": 0.0, + "step": 19119 + }, + { + "epoch": 4.42, + "learning_rate": 1.3176950928279795e-05, + "loss": 0.0, + "step": 19120 + }, + { + "epoch": 4.42, + "learning_rate": 1.3175025036591942e-05, + "loss": 0.0, + "step": 19121 + }, + { + "epoch": 4.42, + "learning_rate": 1.3173099144904091e-05, + "loss": 0.0, + "step": 19122 + }, + { + "epoch": 4.42, + "learning_rate": 1.317117325321624e-05, + "loss": 0.0, + "step": 19123 + }, + { + "epoch": 4.42, + "learning_rate": 1.316924736152839e-05, + "loss": 0.0, + "step": 19124 + }, + { + "epoch": 4.42, + "learning_rate": 1.3167321469840537e-05, + "loss": 0.0, + "step": 19125 + }, + { + "epoch": 4.42, + "learning_rate": 1.3165395578152686e-05, + "loss": 0.0, + "step": 19126 + }, + { + "epoch": 4.42, + "learning_rate": 1.3163469686464832e-05, + "loss": 0.0, + "step": 19127 + }, + { + "epoch": 4.42, + "learning_rate": 1.3161543794776981e-05, + "loss": 0.0, + "step": 19128 + }, + { + "epoch": 4.42, + "learning_rate": 1.315961790308913e-05, + "loss": 0.0, + "step": 19129 + }, + { + "epoch": 4.42, + "learning_rate": 1.315769201140128e-05, + "loss": 0.0, + "step": 19130 + }, + { + "epoch": 4.42, + "learning_rate": 1.3155766119713429e-05, + "loss": 0.0, + "step": 19131 + }, + { + "epoch": 4.42, + "learning_rate": 1.3153840228025576e-05, + "loss": 0.0, + "step": 19132 + }, + { + "epoch": 4.42, + "learning_rate": 1.3151914336337725e-05, + "loss": 0.0, + "step": 19133 + }, + { + "epoch": 4.42, + "learning_rate": 1.3149988444649874e-05, + "loss": 0.0, + "step": 19134 + }, + { + "epoch": 4.42, + "learning_rate": 1.3148062552962024e-05, + "loss": 0.0, + "step": 19135 + }, + { + "epoch": 4.42, + "learning_rate": 1.314613666127417e-05, + "loss": 0.0, + "step": 19136 + }, + { + "epoch": 4.42, + "learning_rate": 1.3144210769586319e-05, + "loss": 0.0, + "step": 19137 + }, + { + "epoch": 4.42, + "learning_rate": 1.3142284877898468e-05, + "loss": 0.0, + "step": 19138 + }, + { + "epoch": 4.42, + "learning_rate": 1.3140358986210615e-05, + "loss": 0.0, + "step": 19139 + }, + { + "epoch": 4.42, + "learning_rate": 1.3138433094522764e-05, + "loss": 0.0, + "step": 19140 + }, + { + "epoch": 4.42, + "learning_rate": 1.3136507202834914e-05, + "loss": 0.0, + "step": 19141 + }, + { + "epoch": 4.42, + "learning_rate": 1.3134581311147063e-05, + "loss": 0.0, + "step": 19142 + }, + { + "epoch": 4.42, + "learning_rate": 1.313265541945921e-05, + "loss": 0.0, + "step": 19143 + }, + { + "epoch": 4.42, + "learning_rate": 1.313072952777136e-05, + "loss": 0.0, + "step": 19144 + }, + { + "epoch": 4.42, + "learning_rate": 1.3128803636083505e-05, + "loss": 0.0, + "step": 19145 + }, + { + "epoch": 4.42, + "learning_rate": 1.3126877744395654e-05, + "loss": 0.0, + "step": 19146 + }, + { + "epoch": 4.43, + "learning_rate": 1.3124951852707804e-05, + "loss": 0.0, + "step": 19147 + }, + { + "epoch": 4.43, + "learning_rate": 1.3123025961019953e-05, + "loss": 0.0, + "step": 19148 + }, + { + "epoch": 4.43, + "learning_rate": 1.3121100069332102e-05, + "loss": 0.0, + "step": 19149 + }, + { + "epoch": 4.43, + "learning_rate": 1.311917417764425e-05, + "loss": 0.0, + "step": 19150 + }, + { + "epoch": 4.43, + "learning_rate": 1.3117248285956399e-05, + "loss": 0.0, + "step": 19151 + }, + { + "epoch": 4.43, + "learning_rate": 1.3115322394268548e-05, + "loss": 0.0, + "step": 19152 + }, + { + "epoch": 4.43, + "learning_rate": 1.3113396502580697e-05, + "loss": 0.0, + "step": 19153 + }, + { + "epoch": 4.43, + "learning_rate": 1.3111470610892846e-05, + "loss": 0.0, + "step": 19154 + }, + { + "epoch": 4.43, + "learning_rate": 1.3109544719204992e-05, + "loss": 0.0002, + "step": 19155 + }, + { + "epoch": 4.43, + "learning_rate": 1.3107618827517141e-05, + "loss": 0.0, + "step": 19156 + }, + { + "epoch": 4.43, + "learning_rate": 1.3105692935829288e-05, + "loss": 0.0, + "step": 19157 + }, + { + "epoch": 4.43, + "learning_rate": 1.3103767044141438e-05, + "loss": 0.0, + "step": 19158 + }, + { + "epoch": 4.43, + "learning_rate": 1.3101841152453587e-05, + "loss": 0.0, + "step": 19159 + }, + { + "epoch": 4.43, + "learning_rate": 1.3099915260765736e-05, + "loss": 0.0, + "step": 19160 + }, + { + "epoch": 4.43, + "learning_rate": 1.3097989369077883e-05, + "loss": 0.0, + "step": 19161 + }, + { + "epoch": 4.43, + "learning_rate": 1.3096063477390033e-05, + "loss": 0.0, + "step": 19162 + }, + { + "epoch": 4.43, + "learning_rate": 1.3094137585702182e-05, + "loss": 0.0, + "step": 19163 + }, + { + "epoch": 4.43, + "learning_rate": 1.3092211694014328e-05, + "loss": 0.0, + "step": 19164 + }, + { + "epoch": 4.43, + "learning_rate": 1.3090285802326477e-05, + "loss": 0.0, + "step": 19165 + }, + { + "epoch": 4.43, + "learning_rate": 1.3088359910638626e-05, + "loss": 0.0, + "step": 19166 + }, + { + "epoch": 4.43, + "learning_rate": 1.3086434018950775e-05, + "loss": 0.0, + "step": 19167 + }, + { + "epoch": 4.43, + "learning_rate": 1.3084508127262923e-05, + "loss": 0.0, + "step": 19168 + }, + { + "epoch": 4.43, + "learning_rate": 1.3082582235575072e-05, + "loss": 0.0, + "step": 19169 + }, + { + "epoch": 4.43, + "learning_rate": 1.3080656343887221e-05, + "loss": 0.0, + "step": 19170 + }, + { + "epoch": 4.43, + "learning_rate": 1.307873045219937e-05, + "loss": 0.0, + "step": 19171 + }, + { + "epoch": 4.43, + "learning_rate": 1.307680456051152e-05, + "loss": 0.0, + "step": 19172 + }, + { + "epoch": 4.43, + "learning_rate": 1.3074878668823665e-05, + "loss": 0.0, + "step": 19173 + }, + { + "epoch": 4.43, + "learning_rate": 1.3072952777135814e-05, + "loss": 0.0, + "step": 19174 + }, + { + "epoch": 4.43, + "learning_rate": 1.3071026885447962e-05, + "loss": 0.0, + "step": 19175 + }, + { + "epoch": 4.43, + "learning_rate": 1.3069100993760111e-05, + "loss": 0.0, + "step": 19176 + }, + { + "epoch": 4.43, + "learning_rate": 1.306717510207226e-05, + "loss": 0.0, + "step": 19177 + }, + { + "epoch": 4.43, + "learning_rate": 1.306524921038441e-05, + "loss": 0.0, + "step": 19178 + }, + { + "epoch": 4.43, + "learning_rate": 1.3063323318696557e-05, + "loss": 0.0, + "step": 19179 + }, + { + "epoch": 4.43, + "learning_rate": 1.3061397427008706e-05, + "loss": 0.0, + "step": 19180 + }, + { + "epoch": 4.43, + "learning_rate": 1.3059471535320855e-05, + "loss": 0.0, + "step": 19181 + }, + { + "epoch": 4.43, + "learning_rate": 1.3057545643633004e-05, + "loss": 0.0, + "step": 19182 + }, + { + "epoch": 4.43, + "learning_rate": 1.305561975194515e-05, + "loss": 0.0, + "step": 19183 + }, + { + "epoch": 4.43, + "learning_rate": 1.30536938602573e-05, + "loss": 0.0, + "step": 19184 + }, + { + "epoch": 4.43, + "learning_rate": 1.3051767968569448e-05, + "loss": 0.0, + "step": 19185 + }, + { + "epoch": 4.43, + "learning_rate": 1.3049842076881596e-05, + "loss": 0.0, + "step": 19186 + }, + { + "epoch": 4.43, + "learning_rate": 1.3047916185193745e-05, + "loss": 0.0, + "step": 19187 + }, + { + "epoch": 4.43, + "learning_rate": 1.3045990293505894e-05, + "loss": 0.0, + "step": 19188 + }, + { + "epoch": 4.43, + "learning_rate": 1.3044064401818043e-05, + "loss": 0.0, + "step": 19189 + }, + { + "epoch": 4.43, + "learning_rate": 1.3042138510130193e-05, + "loss": 0.0, + "step": 19190 + }, + { + "epoch": 4.44, + "learning_rate": 1.304021261844234e-05, + "loss": 0.0, + "step": 19191 + }, + { + "epoch": 4.44, + "learning_rate": 1.3038286726754487e-05, + "loss": 0.0, + "step": 19192 + }, + { + "epoch": 4.44, + "learning_rate": 1.3036360835066635e-05, + "loss": 0.0, + "step": 19193 + }, + { + "epoch": 4.44, + "learning_rate": 1.3034434943378784e-05, + "loss": 0.0, + "step": 19194 + }, + { + "epoch": 4.44, + "learning_rate": 1.3032509051690933e-05, + "loss": 0.0, + "step": 19195 + }, + { + "epoch": 4.44, + "learning_rate": 1.3030583160003082e-05, + "loss": 0.0004, + "step": 19196 + }, + { + "epoch": 4.44, + "learning_rate": 1.3028657268315232e-05, + "loss": 0.0, + "step": 19197 + }, + { + "epoch": 4.44, + "learning_rate": 1.3026731376627379e-05, + "loss": 0.0, + "step": 19198 + }, + { + "epoch": 4.44, + "learning_rate": 1.3024805484939528e-05, + "loss": 0.0, + "step": 19199 + }, + { + "epoch": 4.44, + "learning_rate": 1.3022879593251677e-05, + "loss": 0.0, + "step": 19200 + }, + { + "epoch": 4.44, + "learning_rate": 1.3020953701563823e-05, + "loss": 0.0, + "step": 19201 + }, + { + "epoch": 4.44, + "learning_rate": 1.3019027809875972e-05, + "loss": 0.0, + "step": 19202 + }, + { + "epoch": 4.44, + "learning_rate": 1.3017101918188122e-05, + "loss": 0.0, + "step": 19203 + }, + { + "epoch": 4.44, + "learning_rate": 1.3015176026500269e-05, + "loss": 0.0, + "step": 19204 + }, + { + "epoch": 4.44, + "learning_rate": 1.3013250134812418e-05, + "loss": 0.0, + "step": 19205 + }, + { + "epoch": 4.44, + "learning_rate": 1.3011324243124567e-05, + "loss": 0.0, + "step": 19206 + }, + { + "epoch": 4.44, + "learning_rate": 1.3009398351436717e-05, + "loss": 0.0, + "step": 19207 + }, + { + "epoch": 4.44, + "learning_rate": 1.3007472459748866e-05, + "loss": 0.0, + "step": 19208 + }, + { + "epoch": 4.44, + "learning_rate": 1.3005546568061013e-05, + "loss": 0.0, + "step": 19209 + }, + { + "epoch": 4.44, + "learning_rate": 1.3003620676373162e-05, + "loss": 0.0, + "step": 19210 + }, + { + "epoch": 4.44, + "learning_rate": 1.3001694784685308e-05, + "loss": 0.0, + "step": 19211 + }, + { + "epoch": 4.44, + "learning_rate": 1.2999768892997457e-05, + "loss": 0.0, + "step": 19212 + }, + { + "epoch": 4.44, + "learning_rate": 1.2997843001309607e-05, + "loss": 0.0, + "step": 19213 + }, + { + "epoch": 4.44, + "learning_rate": 1.2995917109621756e-05, + "loss": 0.0, + "step": 19214 + }, + { + "epoch": 4.44, + "learning_rate": 1.2993991217933905e-05, + "loss": 0.0, + "step": 19215 + }, + { + "epoch": 4.44, + "learning_rate": 1.2992065326246052e-05, + "loss": 0.0, + "step": 19216 + }, + { + "epoch": 4.44, + "learning_rate": 1.2990139434558202e-05, + "loss": 0.0, + "step": 19217 + }, + { + "epoch": 4.44, + "learning_rate": 1.298821354287035e-05, + "loss": 0.0, + "step": 19218 + }, + { + "epoch": 4.44, + "learning_rate": 1.29862876511825e-05, + "loss": 0.0, + "step": 19219 + }, + { + "epoch": 4.44, + "learning_rate": 1.2984361759494646e-05, + "loss": 0.0, + "step": 19220 + }, + { + "epoch": 4.44, + "learning_rate": 1.2982435867806795e-05, + "loss": 0.0, + "step": 19221 + }, + { + "epoch": 4.44, + "learning_rate": 1.2980509976118942e-05, + "loss": 0.0, + "step": 19222 + }, + { + "epoch": 4.44, + "learning_rate": 1.2978584084431091e-05, + "loss": 0.0, + "step": 19223 + }, + { + "epoch": 4.44, + "learning_rate": 1.297665819274324e-05, + "loss": 0.0, + "step": 19224 + }, + { + "epoch": 4.44, + "learning_rate": 1.297473230105539e-05, + "loss": 0.0, + "step": 19225 + }, + { + "epoch": 4.44, + "learning_rate": 1.2972806409367539e-05, + "loss": 0.0, + "step": 19226 + }, + { + "epoch": 4.44, + "learning_rate": 1.2970880517679686e-05, + "loss": 0.0, + "step": 19227 + }, + { + "epoch": 4.44, + "learning_rate": 1.2968954625991836e-05, + "loss": 0.0, + "step": 19228 + }, + { + "epoch": 4.44, + "learning_rate": 1.2967028734303981e-05, + "loss": 0.0, + "step": 19229 + }, + { + "epoch": 4.44, + "learning_rate": 1.296510284261613e-05, + "loss": 0.0, + "step": 19230 + }, + { + "epoch": 4.44, + "learning_rate": 1.296317695092828e-05, + "loss": 0.0, + "step": 19231 + }, + { + "epoch": 4.44, + "learning_rate": 1.2961251059240429e-05, + "loss": 0.0, + "step": 19232 + }, + { + "epoch": 4.44, + "learning_rate": 1.2959325167552578e-05, + "loss": 0.0, + "step": 19233 + }, + { + "epoch": 4.45, + "learning_rate": 1.2957399275864726e-05, + "loss": 0.0, + "step": 19234 + }, + { + "epoch": 4.45, + "learning_rate": 1.2955473384176875e-05, + "loss": 0.0, + "step": 19235 + }, + { + "epoch": 4.45, + "learning_rate": 1.2953547492489024e-05, + "loss": 0.0, + "step": 19236 + }, + { + "epoch": 4.45, + "learning_rate": 1.2951621600801173e-05, + "loss": 0.0062, + "step": 19237 + }, + { + "epoch": 4.45, + "learning_rate": 1.294969570911332e-05, + "loss": 0.0, + "step": 19238 + }, + { + "epoch": 4.45, + "learning_rate": 1.2947769817425468e-05, + "loss": 0.0, + "step": 19239 + }, + { + "epoch": 4.45, + "learning_rate": 1.2945843925737616e-05, + "loss": 0.0, + "step": 19240 + }, + { + "epoch": 4.45, + "learning_rate": 1.2943918034049765e-05, + "loss": 0.0, + "step": 19241 + }, + { + "epoch": 4.45, + "learning_rate": 1.2941992142361914e-05, + "loss": 0.0, + "step": 19242 + }, + { + "epoch": 4.45, + "learning_rate": 1.2940066250674063e-05, + "loss": 0.0, + "step": 19243 + }, + { + "epoch": 4.45, + "learning_rate": 1.2938140358986212e-05, + "loss": 0.0, + "step": 19244 + }, + { + "epoch": 4.45, + "learning_rate": 1.293621446729836e-05, + "loss": 0.0, + "step": 19245 + }, + { + "epoch": 4.45, + "learning_rate": 1.2934288575610509e-05, + "loss": 0.0, + "step": 19246 + }, + { + "epoch": 4.45, + "learning_rate": 1.2932362683922658e-05, + "loss": 0.0, + "step": 19247 + }, + { + "epoch": 4.45, + "learning_rate": 1.2930436792234804e-05, + "loss": 0.0, + "step": 19248 + }, + { + "epoch": 4.45, + "learning_rate": 1.2928510900546953e-05, + "loss": 0.0, + "step": 19249 + }, + { + "epoch": 4.45, + "learning_rate": 1.2926585008859102e-05, + "loss": 0.0, + "step": 19250 + }, + { + "epoch": 4.45, + "learning_rate": 1.2924659117171251e-05, + "loss": 0.0, + "step": 19251 + }, + { + "epoch": 4.45, + "learning_rate": 1.2922733225483399e-05, + "loss": 0.0, + "step": 19252 + }, + { + "epoch": 4.45, + "learning_rate": 1.2920807333795548e-05, + "loss": 0.0, + "step": 19253 + }, + { + "epoch": 4.45, + "learning_rate": 1.2918881442107697e-05, + "loss": 0.0, + "step": 19254 + }, + { + "epoch": 4.45, + "learning_rate": 1.2916955550419846e-05, + "loss": 0.0, + "step": 19255 + }, + { + "epoch": 4.45, + "learning_rate": 1.2915029658731994e-05, + "loss": 0.0, + "step": 19256 + }, + { + "epoch": 4.45, + "learning_rate": 1.2913103767044141e-05, + "loss": 0.0, + "step": 19257 + }, + { + "epoch": 4.45, + "learning_rate": 1.2911177875356289e-05, + "loss": 0.0, + "step": 19258 + }, + { + "epoch": 4.45, + "learning_rate": 1.2909251983668438e-05, + "loss": 0.0, + "step": 19259 + }, + { + "epoch": 4.45, + "learning_rate": 1.2907326091980587e-05, + "loss": 0.0, + "step": 19260 + }, + { + "epoch": 4.45, + "learning_rate": 1.2905400200292736e-05, + "loss": 0.0, + "step": 19261 + }, + { + "epoch": 4.45, + "learning_rate": 1.2903474308604885e-05, + "loss": 0.0, + "step": 19262 + }, + { + "epoch": 4.45, + "learning_rate": 1.2901548416917033e-05, + "loss": 0.0, + "step": 19263 + }, + { + "epoch": 4.45, + "learning_rate": 1.2899622525229182e-05, + "loss": 0.0, + "step": 19264 + }, + { + "epoch": 4.45, + "learning_rate": 1.2897696633541331e-05, + "loss": 0.0, + "step": 19265 + }, + { + "epoch": 4.45, + "learning_rate": 1.289577074185348e-05, + "loss": 0.0, + "step": 19266 + }, + { + "epoch": 4.45, + "learning_rate": 1.2893844850165626e-05, + "loss": 0.0, + "step": 19267 + }, + { + "epoch": 4.45, + "learning_rate": 1.2891918958477775e-05, + "loss": 0.0001, + "step": 19268 + }, + { + "epoch": 4.45, + "learning_rate": 1.2889993066789925e-05, + "loss": 0.0, + "step": 19269 + }, + { + "epoch": 4.45, + "learning_rate": 1.2888067175102072e-05, + "loss": 0.0, + "step": 19270 + }, + { + "epoch": 4.45, + "learning_rate": 1.2886141283414221e-05, + "loss": 0.0, + "step": 19271 + }, + { + "epoch": 4.45, + "learning_rate": 1.288421539172637e-05, + "loss": 0.0001, + "step": 19272 + }, + { + "epoch": 4.45, + "learning_rate": 1.288228950003852e-05, + "loss": 0.0, + "step": 19273 + }, + { + "epoch": 4.45, + "learning_rate": 1.2880363608350669e-05, + "loss": 0.0, + "step": 19274 + }, + { + "epoch": 4.45, + "learning_rate": 1.2878437716662816e-05, + "loss": 0.0, + "step": 19275 + }, + { + "epoch": 4.45, + "learning_rate": 1.2876511824974962e-05, + "loss": 0.0, + "step": 19276 + }, + { + "epoch": 4.46, + "learning_rate": 1.2874585933287111e-05, + "loss": 0.0, + "step": 19277 + }, + { + "epoch": 4.46, + "learning_rate": 1.287266004159926e-05, + "loss": 0.0, + "step": 19278 + }, + { + "epoch": 4.46, + "learning_rate": 1.287073414991141e-05, + "loss": 0.0, + "step": 19279 + }, + { + "epoch": 4.46, + "learning_rate": 1.2868808258223559e-05, + "loss": 0.0, + "step": 19280 + }, + { + "epoch": 4.46, + "learning_rate": 1.2866882366535706e-05, + "loss": 0.0, + "step": 19281 + }, + { + "epoch": 4.46, + "learning_rate": 1.2864956474847855e-05, + "loss": 0.0, + "step": 19282 + }, + { + "epoch": 4.46, + "learning_rate": 1.2863030583160005e-05, + "loss": 0.0, + "step": 19283 + }, + { + "epoch": 4.46, + "learning_rate": 1.2861104691472154e-05, + "loss": 0.0, + "step": 19284 + }, + { + "epoch": 4.46, + "learning_rate": 1.28591787997843e-05, + "loss": 0.0, + "step": 19285 + }, + { + "epoch": 4.46, + "learning_rate": 1.2857252908096449e-05, + "loss": 0.0, + "step": 19286 + }, + { + "epoch": 4.46, + "learning_rate": 1.2855327016408598e-05, + "loss": 0.0, + "step": 19287 + }, + { + "epoch": 4.46, + "learning_rate": 1.2853401124720745e-05, + "loss": 0.0, + "step": 19288 + }, + { + "epoch": 4.46, + "learning_rate": 1.2851475233032894e-05, + "loss": 0.0, + "step": 19289 + }, + { + "epoch": 4.46, + "learning_rate": 1.2849549341345044e-05, + "loss": 0.0, + "step": 19290 + }, + { + "epoch": 4.46, + "learning_rate": 1.2847623449657193e-05, + "loss": 0.0, + "step": 19291 + }, + { + "epoch": 4.46, + "learning_rate": 1.2845697557969342e-05, + "loss": 0.0, + "step": 19292 + }, + { + "epoch": 4.46, + "learning_rate": 1.284377166628149e-05, + "loss": 0.0, + "step": 19293 + }, + { + "epoch": 4.46, + "learning_rate": 1.2841845774593639e-05, + "loss": 0.0, + "step": 19294 + }, + { + "epoch": 4.46, + "learning_rate": 1.2839919882905784e-05, + "loss": 0.0, + "step": 19295 + }, + { + "epoch": 4.46, + "learning_rate": 1.2837993991217934e-05, + "loss": 0.0, + "step": 19296 + }, + { + "epoch": 4.46, + "learning_rate": 1.2836068099530083e-05, + "loss": 0.0, + "step": 19297 + }, + { + "epoch": 4.46, + "learning_rate": 1.2834142207842232e-05, + "loss": 0.0, + "step": 19298 + }, + { + "epoch": 4.46, + "learning_rate": 1.283221631615438e-05, + "loss": 0.0, + "step": 19299 + }, + { + "epoch": 4.46, + "learning_rate": 1.2830290424466529e-05, + "loss": 0.0, + "step": 19300 + }, + { + "epoch": 4.46, + "learning_rate": 1.2828364532778678e-05, + "loss": 0.0, + "step": 19301 + }, + { + "epoch": 4.46, + "learning_rate": 1.2826438641090827e-05, + "loss": 0.0, + "step": 19302 + }, + { + "epoch": 4.46, + "learning_rate": 1.2824512749402976e-05, + "loss": 0.0, + "step": 19303 + }, + { + "epoch": 4.46, + "learning_rate": 1.2822586857715122e-05, + "loss": 0.0, + "step": 19304 + }, + { + "epoch": 4.46, + "learning_rate": 1.2820660966027271e-05, + "loss": 0.0, + "step": 19305 + }, + { + "epoch": 4.46, + "learning_rate": 1.2818735074339419e-05, + "loss": 0.0001, + "step": 19306 + }, + { + "epoch": 4.46, + "learning_rate": 1.2816809182651568e-05, + "loss": 0.0, + "step": 19307 + }, + { + "epoch": 4.46, + "learning_rate": 1.2814883290963717e-05, + "loss": 0.0, + "step": 19308 + }, + { + "epoch": 4.46, + "learning_rate": 1.2812957399275866e-05, + "loss": 0.0, + "step": 19309 + }, + { + "epoch": 4.46, + "learning_rate": 1.2811031507588015e-05, + "loss": 0.0001, + "step": 19310 + }, + { + "epoch": 4.46, + "learning_rate": 1.2809105615900163e-05, + "loss": 0.0, + "step": 19311 + }, + { + "epoch": 4.46, + "learning_rate": 1.2807179724212312e-05, + "loss": 0.0, + "step": 19312 + }, + { + "epoch": 4.46, + "learning_rate": 1.2805253832524458e-05, + "loss": 0.0, + "step": 19313 + }, + { + "epoch": 4.46, + "learning_rate": 1.2803327940836607e-05, + "loss": 0.0, + "step": 19314 + }, + { + "epoch": 4.46, + "learning_rate": 1.2801402049148756e-05, + "loss": 0.0, + "step": 19315 + }, + { + "epoch": 4.46, + "learning_rate": 1.2799476157460905e-05, + "loss": 0.0, + "step": 19316 + }, + { + "epoch": 4.46, + "learning_rate": 1.2797550265773053e-05, + "loss": 0.0, + "step": 19317 + }, + { + "epoch": 4.46, + "learning_rate": 1.2795624374085202e-05, + "loss": 0.0, + "step": 19318 + }, + { + "epoch": 4.46, + "learning_rate": 1.2793698482397351e-05, + "loss": 0.0002, + "step": 19319 + }, + { + "epoch": 4.46, + "learning_rate": 1.27917725907095e-05, + "loss": 0.0, + "step": 19320 + }, + { + "epoch": 4.47, + "learning_rate": 1.278984669902165e-05, + "loss": 0.0, + "step": 19321 + }, + { + "epoch": 4.47, + "learning_rate": 1.2787920807333797e-05, + "loss": 0.0, + "step": 19322 + }, + { + "epoch": 4.47, + "learning_rate": 1.2785994915645944e-05, + "loss": 0.0, + "step": 19323 + }, + { + "epoch": 4.47, + "learning_rate": 1.2784069023958092e-05, + "loss": 0.0, + "step": 19324 + }, + { + "epoch": 4.47, + "learning_rate": 1.2782143132270241e-05, + "loss": 0.0, + "step": 19325 + }, + { + "epoch": 4.47, + "learning_rate": 1.278021724058239e-05, + "loss": 0.0, + "step": 19326 + }, + { + "epoch": 4.47, + "learning_rate": 1.277829134889454e-05, + "loss": 0.0, + "step": 19327 + }, + { + "epoch": 4.47, + "learning_rate": 1.2776365457206688e-05, + "loss": 0.0, + "step": 19328 + }, + { + "epoch": 4.47, + "learning_rate": 1.2774439565518836e-05, + "loss": 0.0, + "step": 19329 + }, + { + "epoch": 4.47, + "learning_rate": 1.2772513673830985e-05, + "loss": 0.0, + "step": 19330 + }, + { + "epoch": 4.47, + "learning_rate": 1.2770587782143134e-05, + "loss": 0.0, + "step": 19331 + }, + { + "epoch": 4.47, + "learning_rate": 1.276866189045528e-05, + "loss": 0.0, + "step": 19332 + }, + { + "epoch": 4.47, + "learning_rate": 1.276673599876743e-05, + "loss": 0.0, + "step": 19333 + }, + { + "epoch": 4.47, + "learning_rate": 1.2764810107079578e-05, + "loss": 0.0001, + "step": 19334 + }, + { + "epoch": 4.47, + "learning_rate": 1.2762884215391726e-05, + "loss": 0.0, + "step": 19335 + }, + { + "epoch": 4.47, + "learning_rate": 1.2760958323703875e-05, + "loss": 0.0, + "step": 19336 + }, + { + "epoch": 4.47, + "learning_rate": 1.2759032432016024e-05, + "loss": 0.0, + "step": 19337 + }, + { + "epoch": 4.47, + "learning_rate": 1.2757106540328173e-05, + "loss": 0.0, + "step": 19338 + }, + { + "epoch": 4.47, + "learning_rate": 1.2755180648640323e-05, + "loss": 0.0, + "step": 19339 + }, + { + "epoch": 4.47, + "learning_rate": 1.275325475695247e-05, + "loss": 0.0, + "step": 19340 + }, + { + "epoch": 4.47, + "learning_rate": 1.2751328865264618e-05, + "loss": 0.0001, + "step": 19341 + }, + { + "epoch": 4.47, + "learning_rate": 1.2749402973576765e-05, + "loss": 0.0, + "step": 19342 + }, + { + "epoch": 4.47, + "learning_rate": 1.2747477081888914e-05, + "loss": 0.0, + "step": 19343 + }, + { + "epoch": 4.47, + "learning_rate": 1.2745551190201063e-05, + "loss": 0.0, + "step": 19344 + }, + { + "epoch": 4.47, + "learning_rate": 1.2743625298513213e-05, + "loss": 0.0, + "step": 19345 + }, + { + "epoch": 4.47, + "learning_rate": 1.2741699406825362e-05, + "loss": 0.0, + "step": 19346 + }, + { + "epoch": 4.47, + "learning_rate": 1.273977351513751e-05, + "loss": 0.0, + "step": 19347 + }, + { + "epoch": 4.47, + "learning_rate": 1.2737847623449658e-05, + "loss": 0.0, + "step": 19348 + }, + { + "epoch": 4.47, + "learning_rate": 1.2735921731761808e-05, + "loss": 0.0, + "step": 19349 + }, + { + "epoch": 4.47, + "learning_rate": 1.2733995840073957e-05, + "loss": 0.0, + "step": 19350 + }, + { + "epoch": 4.47, + "learning_rate": 1.2732069948386102e-05, + "loss": 0.0, + "step": 19351 + }, + { + "epoch": 4.47, + "learning_rate": 1.2730144056698252e-05, + "loss": 0.0, + "step": 19352 + }, + { + "epoch": 4.47, + "learning_rate": 1.2728218165010399e-05, + "loss": 0.0001, + "step": 19353 + }, + { + "epoch": 4.47, + "learning_rate": 1.2726292273322548e-05, + "loss": 0.0001, + "step": 19354 + }, + { + "epoch": 4.47, + "learning_rate": 1.2724366381634697e-05, + "loss": 0.0, + "step": 19355 + }, + { + "epoch": 4.47, + "learning_rate": 1.2722440489946847e-05, + "loss": 0.0, + "step": 19356 + }, + { + "epoch": 4.47, + "learning_rate": 1.2720514598258996e-05, + "loss": 0.0, + "step": 19357 + }, + { + "epoch": 4.47, + "learning_rate": 1.2718588706571143e-05, + "loss": 0.0, + "step": 19358 + }, + { + "epoch": 4.47, + "learning_rate": 1.2716662814883292e-05, + "loss": 0.0, + "step": 19359 + }, + { + "epoch": 4.47, + "learning_rate": 1.2714736923195438e-05, + "loss": 0.0001, + "step": 19360 + }, + { + "epoch": 4.47, + "learning_rate": 1.2712811031507587e-05, + "loss": 0.0, + "step": 19361 + }, + { + "epoch": 4.47, + "learning_rate": 1.2710885139819737e-05, + "loss": 0.0002, + "step": 19362 + }, + { + "epoch": 4.47, + "learning_rate": 1.2708959248131886e-05, + "loss": 0.0, + "step": 19363 + }, + { + "epoch": 4.48, + "learning_rate": 1.2707033356444035e-05, + "loss": 0.0, + "step": 19364 + }, + { + "epoch": 4.48, + "learning_rate": 1.2705107464756182e-05, + "loss": 0.0, + "step": 19365 + }, + { + "epoch": 4.48, + "learning_rate": 1.2703181573068332e-05, + "loss": 0.0, + "step": 19366 + }, + { + "epoch": 4.48, + "learning_rate": 1.270125568138048e-05, + "loss": 0.0, + "step": 19367 + }, + { + "epoch": 4.48, + "learning_rate": 1.269932978969263e-05, + "loss": 0.0, + "step": 19368 + }, + { + "epoch": 4.48, + "learning_rate": 1.2697403898004776e-05, + "loss": 0.0, + "step": 19369 + }, + { + "epoch": 4.48, + "learning_rate": 1.2695478006316925e-05, + "loss": 0.0, + "step": 19370 + }, + { + "epoch": 4.48, + "learning_rate": 1.2693552114629072e-05, + "loss": 0.0, + "step": 19371 + }, + { + "epoch": 4.48, + "learning_rate": 1.2691626222941222e-05, + "loss": 0.0, + "step": 19372 + }, + { + "epoch": 4.48, + "learning_rate": 1.268970033125337e-05, + "loss": 0.0, + "step": 19373 + }, + { + "epoch": 4.48, + "learning_rate": 1.268777443956552e-05, + "loss": 0.0, + "step": 19374 + }, + { + "epoch": 4.48, + "learning_rate": 1.2685848547877669e-05, + "loss": 0.0, + "step": 19375 + }, + { + "epoch": 4.48, + "learning_rate": 1.2683922656189817e-05, + "loss": 0.0001, + "step": 19376 + }, + { + "epoch": 4.48, + "learning_rate": 1.2681996764501966e-05, + "loss": 0.0, + "step": 19377 + }, + { + "epoch": 4.48, + "learning_rate": 1.2680070872814115e-05, + "loss": 0.0, + "step": 19378 + }, + { + "epoch": 4.48, + "learning_rate": 1.267814498112626e-05, + "loss": 0.0, + "step": 19379 + }, + { + "epoch": 4.48, + "learning_rate": 1.267621908943841e-05, + "loss": 0.0, + "step": 19380 + }, + { + "epoch": 4.48, + "learning_rate": 1.2674293197750559e-05, + "loss": 0.0, + "step": 19381 + }, + { + "epoch": 4.48, + "learning_rate": 1.2672367306062708e-05, + "loss": 0.0, + "step": 19382 + }, + { + "epoch": 4.48, + "learning_rate": 1.2670441414374856e-05, + "loss": 0.0, + "step": 19383 + }, + { + "epoch": 4.48, + "learning_rate": 1.2668515522687005e-05, + "loss": 0.0, + "step": 19384 + }, + { + "epoch": 4.48, + "learning_rate": 1.2666589630999154e-05, + "loss": 0.0, + "step": 19385 + }, + { + "epoch": 4.48, + "learning_rate": 1.2664663739311303e-05, + "loss": 0.0, + "step": 19386 + }, + { + "epoch": 4.48, + "learning_rate": 1.2662737847623452e-05, + "loss": 0.0, + "step": 19387 + }, + { + "epoch": 4.48, + "learning_rate": 1.2660811955935598e-05, + "loss": 0.0, + "step": 19388 + }, + { + "epoch": 4.48, + "learning_rate": 1.2658886064247746e-05, + "loss": 0.0, + "step": 19389 + }, + { + "epoch": 4.48, + "learning_rate": 1.2656960172559895e-05, + "loss": 0.0, + "step": 19390 + }, + { + "epoch": 4.48, + "learning_rate": 1.2655034280872044e-05, + "loss": 0.0, + "step": 19391 + }, + { + "epoch": 4.48, + "learning_rate": 1.2653108389184193e-05, + "loss": 0.0, + "step": 19392 + }, + { + "epoch": 4.48, + "learning_rate": 1.2651182497496342e-05, + "loss": 0.0, + "step": 19393 + }, + { + "epoch": 4.48, + "learning_rate": 1.264925660580849e-05, + "loss": 0.0, + "step": 19394 + }, + { + "epoch": 4.48, + "learning_rate": 1.2647330714120639e-05, + "loss": 0.0, + "step": 19395 + }, + { + "epoch": 4.48, + "learning_rate": 1.2645404822432788e-05, + "loss": 0.0, + "step": 19396 + }, + { + "epoch": 4.48, + "learning_rate": 1.2643478930744934e-05, + "loss": 0.0, + "step": 19397 + }, + { + "epoch": 4.48, + "learning_rate": 1.2641553039057083e-05, + "loss": 0.0, + "step": 19398 + }, + { + "epoch": 4.48, + "learning_rate": 1.2639627147369232e-05, + "loss": 0.0, + "step": 19399 + }, + { + "epoch": 4.48, + "learning_rate": 1.2637701255681381e-05, + "loss": 0.0, + "step": 19400 + }, + { + "epoch": 4.48, + "learning_rate": 1.2635775363993529e-05, + "loss": 0.0, + "step": 19401 + }, + { + "epoch": 4.48, + "learning_rate": 1.2633849472305678e-05, + "loss": 0.0, + "step": 19402 + }, + { + "epoch": 4.48, + "learning_rate": 1.2631923580617827e-05, + "loss": 0.0, + "step": 19403 + }, + { + "epoch": 4.48, + "learning_rate": 1.2629997688929976e-05, + "loss": 0.0, + "step": 19404 + }, + { + "epoch": 4.48, + "learning_rate": 1.2628071797242126e-05, + "loss": 0.0, + "step": 19405 + }, + { + "epoch": 4.48, + "learning_rate": 1.2626145905554273e-05, + "loss": 0.0001, + "step": 19406 + }, + { + "epoch": 4.49, + "learning_rate": 1.2624220013866419e-05, + "loss": 0.0, + "step": 19407 + }, + { + "epoch": 4.49, + "learning_rate": 1.2622294122178568e-05, + "loss": 0.0, + "step": 19408 + }, + { + "epoch": 4.49, + "learning_rate": 1.2620368230490717e-05, + "loss": 0.0, + "step": 19409 + }, + { + "epoch": 4.49, + "learning_rate": 1.2618442338802866e-05, + "loss": 0.0, + "step": 19410 + }, + { + "epoch": 4.49, + "learning_rate": 1.2616516447115016e-05, + "loss": 0.0, + "step": 19411 + }, + { + "epoch": 4.49, + "learning_rate": 1.2614590555427163e-05, + "loss": 0.0, + "step": 19412 + }, + { + "epoch": 4.49, + "learning_rate": 1.2612664663739312e-05, + "loss": 0.0, + "step": 19413 + }, + { + "epoch": 4.49, + "learning_rate": 1.2610738772051461e-05, + "loss": 0.0, + "step": 19414 + }, + { + "epoch": 4.49, + "learning_rate": 1.260881288036361e-05, + "loss": 0.0, + "step": 19415 + }, + { + "epoch": 4.49, + "learning_rate": 1.2606886988675756e-05, + "loss": 0.0, + "step": 19416 + }, + { + "epoch": 4.49, + "learning_rate": 1.2604961096987905e-05, + "loss": 0.0, + "step": 19417 + }, + { + "epoch": 4.49, + "learning_rate": 1.2603035205300055e-05, + "loss": 0.0, + "step": 19418 + }, + { + "epoch": 4.49, + "learning_rate": 1.2601109313612202e-05, + "loss": 0.0, + "step": 19419 + }, + { + "epoch": 4.49, + "learning_rate": 1.2599183421924351e-05, + "loss": 0.0, + "step": 19420 + }, + { + "epoch": 4.49, + "learning_rate": 1.25972575302365e-05, + "loss": 0.0, + "step": 19421 + }, + { + "epoch": 4.49, + "learning_rate": 1.259533163854865e-05, + "loss": 0.0, + "step": 19422 + }, + { + "epoch": 4.49, + "learning_rate": 1.2593405746860799e-05, + "loss": 0.0, + "step": 19423 + }, + { + "epoch": 4.49, + "learning_rate": 1.2591479855172946e-05, + "loss": 0.0, + "step": 19424 + }, + { + "epoch": 4.49, + "learning_rate": 1.2589553963485092e-05, + "loss": 0.0, + "step": 19425 + }, + { + "epoch": 4.49, + "learning_rate": 1.2587628071797241e-05, + "loss": 0.0, + "step": 19426 + }, + { + "epoch": 4.49, + "learning_rate": 1.258570218010939e-05, + "loss": 0.0, + "step": 19427 + }, + { + "epoch": 4.49, + "learning_rate": 1.258377628842154e-05, + "loss": 0.0, + "step": 19428 + }, + { + "epoch": 4.49, + "learning_rate": 1.2581850396733689e-05, + "loss": 0.0, + "step": 19429 + }, + { + "epoch": 4.49, + "learning_rate": 1.2579924505045836e-05, + "loss": 0.0, + "step": 19430 + }, + { + "epoch": 4.49, + "learning_rate": 1.2577998613357985e-05, + "loss": 0.0, + "step": 19431 + }, + { + "epoch": 4.49, + "learning_rate": 1.2576072721670135e-05, + "loss": 0.0, + "step": 19432 + }, + { + "epoch": 4.49, + "learning_rate": 1.2574146829982284e-05, + "loss": 0.0, + "step": 19433 + }, + { + "epoch": 4.49, + "learning_rate": 1.2572220938294433e-05, + "loss": 0.0, + "step": 19434 + }, + { + "epoch": 4.49, + "learning_rate": 1.2570295046606579e-05, + "loss": 0.0, + "step": 19435 + }, + { + "epoch": 4.49, + "learning_rate": 1.2568369154918728e-05, + "loss": 0.0, + "step": 19436 + }, + { + "epoch": 4.49, + "learning_rate": 1.2566443263230875e-05, + "loss": 0.0, + "step": 19437 + }, + { + "epoch": 4.49, + "learning_rate": 1.2564517371543025e-05, + "loss": 0.0, + "step": 19438 + }, + { + "epoch": 4.49, + "learning_rate": 1.2562591479855174e-05, + "loss": 0.0, + "step": 19439 + }, + { + "epoch": 4.49, + "learning_rate": 1.2560665588167323e-05, + "loss": 0.0, + "step": 19440 + }, + { + "epoch": 4.49, + "learning_rate": 1.2558739696479472e-05, + "loss": 0.0, + "step": 19441 + }, + { + "epoch": 4.49, + "learning_rate": 1.255681380479162e-05, + "loss": 0.0, + "step": 19442 + }, + { + "epoch": 4.49, + "learning_rate": 1.2554887913103769e-05, + "loss": 0.0, + "step": 19443 + }, + { + "epoch": 4.49, + "learning_rate": 1.2552962021415914e-05, + "loss": 0.0, + "step": 19444 + }, + { + "epoch": 4.49, + "learning_rate": 1.2551036129728064e-05, + "loss": 0.0, + "step": 19445 + }, + { + "epoch": 4.49, + "learning_rate": 1.2549110238040213e-05, + "loss": 0.0, + "step": 19446 + }, + { + "epoch": 4.49, + "learning_rate": 1.2547184346352362e-05, + "loss": 0.0, + "step": 19447 + }, + { + "epoch": 4.49, + "learning_rate": 1.254525845466451e-05, + "loss": 0.0, + "step": 19448 + }, + { + "epoch": 4.49, + "learning_rate": 1.2543332562976659e-05, + "loss": 0.0, + "step": 19449 + }, + { + "epoch": 4.5, + "learning_rate": 1.2541406671288808e-05, + "loss": 0.0, + "step": 19450 + }, + { + "epoch": 4.5, + "learning_rate": 1.2539480779600957e-05, + "loss": 0.0, + "step": 19451 + }, + { + "epoch": 4.5, + "learning_rate": 1.2537554887913106e-05, + "loss": 0.0, + "step": 19452 + }, + { + "epoch": 4.5, + "learning_rate": 1.2535628996225252e-05, + "loss": 0.0, + "step": 19453 + }, + { + "epoch": 4.5, + "learning_rate": 1.2533703104537401e-05, + "loss": 0.0, + "step": 19454 + }, + { + "epoch": 4.5, + "learning_rate": 1.2531777212849549e-05, + "loss": 0.0, + "step": 19455 + }, + { + "epoch": 4.5, + "learning_rate": 1.2529851321161698e-05, + "loss": 0.0, + "step": 19456 + }, + { + "epoch": 4.5, + "learning_rate": 1.2527925429473847e-05, + "loss": 0.0, + "step": 19457 + }, + { + "epoch": 4.5, + "learning_rate": 1.2525999537785996e-05, + "loss": 0.0, + "step": 19458 + }, + { + "epoch": 4.5, + "learning_rate": 1.2524073646098145e-05, + "loss": 0.0, + "step": 19459 + }, + { + "epoch": 4.5, + "learning_rate": 1.2522147754410293e-05, + "loss": 0.0, + "step": 19460 + }, + { + "epoch": 4.5, + "learning_rate": 1.2520221862722442e-05, + "loss": 0.0, + "step": 19461 + }, + { + "epoch": 4.5, + "learning_rate": 1.2518295971034591e-05, + "loss": 0.0, + "step": 19462 + }, + { + "epoch": 4.5, + "learning_rate": 1.2516370079346737e-05, + "loss": 0.0, + "step": 19463 + }, + { + "epoch": 4.5, + "learning_rate": 1.2514444187658886e-05, + "loss": 0.0, + "step": 19464 + }, + { + "epoch": 4.5, + "learning_rate": 1.2512518295971035e-05, + "loss": 0.0, + "step": 19465 + }, + { + "epoch": 4.5, + "learning_rate": 1.2510592404283183e-05, + "loss": 0.0, + "step": 19466 + }, + { + "epoch": 4.5, + "learning_rate": 1.2508666512595332e-05, + "loss": 0.0, + "step": 19467 + }, + { + "epoch": 4.5, + "learning_rate": 1.2506740620907481e-05, + "loss": 0.0, + "step": 19468 + }, + { + "epoch": 4.5, + "learning_rate": 1.250481472921963e-05, + "loss": 0.0, + "step": 19469 + }, + { + "epoch": 4.5, + "learning_rate": 1.250288883753178e-05, + "loss": 0.0, + "step": 19470 + }, + { + "epoch": 4.5, + "learning_rate": 1.2500962945843927e-05, + "loss": 0.0, + "step": 19471 + }, + { + "epoch": 4.5, + "learning_rate": 1.2499037054156074e-05, + "loss": 0.0, + "step": 19472 + }, + { + "epoch": 4.5, + "learning_rate": 1.2497111162468224e-05, + "loss": 0.0, + "step": 19473 + }, + { + "epoch": 4.5, + "learning_rate": 1.2495185270780371e-05, + "loss": 0.0, + "step": 19474 + }, + { + "epoch": 4.5, + "learning_rate": 1.249325937909252e-05, + "loss": 0.0, + "step": 19475 + }, + { + "epoch": 4.5, + "learning_rate": 1.249133348740467e-05, + "loss": 0.0, + "step": 19476 + }, + { + "epoch": 4.5, + "learning_rate": 1.2489407595716819e-05, + "loss": 0.0, + "step": 19477 + }, + { + "epoch": 4.5, + "learning_rate": 1.2487481704028966e-05, + "loss": 0.0, + "step": 19478 + }, + { + "epoch": 4.5, + "learning_rate": 1.2485555812341113e-05, + "loss": 0.0, + "step": 19479 + }, + { + "epoch": 4.5, + "learning_rate": 1.2483629920653263e-05, + "loss": 0.0, + "step": 19480 + }, + { + "epoch": 4.5, + "learning_rate": 1.2481704028965412e-05, + "loss": 0.0, + "step": 19481 + }, + { + "epoch": 4.5, + "learning_rate": 1.2479778137277561e-05, + "loss": 0.0, + "step": 19482 + }, + { + "epoch": 4.5, + "learning_rate": 1.2477852245589708e-05, + "loss": 0.0, + "step": 19483 + }, + { + "epoch": 4.5, + "learning_rate": 1.2475926353901856e-05, + "loss": 0.001, + "step": 19484 + }, + { + "epoch": 4.5, + "learning_rate": 1.2474000462214005e-05, + "loss": 0.0, + "step": 19485 + }, + { + "epoch": 4.5, + "learning_rate": 1.2472074570526154e-05, + "loss": 0.0, + "step": 19486 + }, + { + "epoch": 4.5, + "learning_rate": 1.2470148678838303e-05, + "loss": 0.0, + "step": 19487 + }, + { + "epoch": 4.5, + "learning_rate": 1.2468222787150451e-05, + "loss": 0.0, + "step": 19488 + }, + { + "epoch": 4.5, + "learning_rate": 1.24662968954626e-05, + "loss": 0.0, + "step": 19489 + }, + { + "epoch": 4.5, + "learning_rate": 1.2464371003774748e-05, + "loss": 0.0, + "step": 19490 + }, + { + "epoch": 4.5, + "learning_rate": 1.2462445112086897e-05, + "loss": 0.0, + "step": 19491 + }, + { + "epoch": 4.5, + "learning_rate": 1.2460519220399046e-05, + "loss": 0.0, + "step": 19492 + }, + { + "epoch": 4.5, + "learning_rate": 1.2458593328711193e-05, + "loss": 0.0, + "step": 19493 + }, + { + "epoch": 4.51, + "learning_rate": 1.2456667437023343e-05, + "loss": 0.0, + "step": 19494 + }, + { + "epoch": 4.51, + "learning_rate": 1.2454741545335492e-05, + "loss": 0.0, + "step": 19495 + }, + { + "epoch": 4.51, + "learning_rate": 1.245281565364764e-05, + "loss": 0.0, + "step": 19496 + }, + { + "epoch": 4.51, + "learning_rate": 1.2450889761959788e-05, + "loss": 0.0, + "step": 19497 + }, + { + "epoch": 4.51, + "learning_rate": 1.2448963870271936e-05, + "loss": 0.0, + "step": 19498 + }, + { + "epoch": 4.51, + "learning_rate": 1.2447037978584085e-05, + "loss": 0.0, + "step": 19499 + }, + { + "epoch": 4.51, + "learning_rate": 1.2445112086896234e-05, + "loss": 0.0, + "step": 19500 + }, + { + "epoch": 4.51, + "learning_rate": 1.2443186195208382e-05, + "loss": 0.0, + "step": 19501 + }, + { + "epoch": 4.51, + "learning_rate": 1.244126030352053e-05, + "loss": 0.0, + "step": 19502 + }, + { + "epoch": 4.51, + "learning_rate": 1.2439334411832678e-05, + "loss": 0.0, + "step": 19503 + }, + { + "epoch": 4.51, + "learning_rate": 1.2437408520144828e-05, + "loss": 0.0, + "step": 19504 + }, + { + "epoch": 4.51, + "learning_rate": 1.2435482628456977e-05, + "loss": 0.0, + "step": 19505 + }, + { + "epoch": 4.51, + "learning_rate": 1.2433556736769126e-05, + "loss": 0.0, + "step": 19506 + }, + { + "epoch": 4.51, + "learning_rate": 1.2431630845081273e-05, + "loss": 0.0, + "step": 19507 + }, + { + "epoch": 4.51, + "learning_rate": 1.242970495339342e-05, + "loss": 0.0, + "step": 19508 + }, + { + "epoch": 4.51, + "learning_rate": 1.242777906170557e-05, + "loss": 0.0003, + "step": 19509 + }, + { + "epoch": 4.51, + "learning_rate": 1.2425853170017719e-05, + "loss": 0.0, + "step": 19510 + }, + { + "epoch": 4.51, + "learning_rate": 1.2423927278329868e-05, + "loss": 0.0, + "step": 19511 + }, + { + "epoch": 4.51, + "learning_rate": 1.2422001386642016e-05, + "loss": 0.0, + "step": 19512 + }, + { + "epoch": 4.51, + "learning_rate": 1.2420075494954165e-05, + "loss": 0.0, + "step": 19513 + }, + { + "epoch": 4.51, + "learning_rate": 1.2418149603266312e-05, + "loss": 0.0, + "step": 19514 + }, + { + "epoch": 4.51, + "learning_rate": 1.2416223711578462e-05, + "loss": 0.0, + "step": 19515 + }, + { + "epoch": 4.51, + "learning_rate": 1.2414297819890609e-05, + "loss": 0.0, + "step": 19516 + }, + { + "epoch": 4.51, + "learning_rate": 1.2412371928202758e-05, + "loss": 0.0, + "step": 19517 + }, + { + "epoch": 4.51, + "learning_rate": 1.2410446036514907e-05, + "loss": 0.0, + "step": 19518 + }, + { + "epoch": 4.51, + "learning_rate": 1.2408520144827057e-05, + "loss": 0.0, + "step": 19519 + }, + { + "epoch": 4.51, + "learning_rate": 1.2406594253139204e-05, + "loss": 0.0001, + "step": 19520 + }, + { + "epoch": 4.51, + "learning_rate": 1.2404668361451352e-05, + "loss": 0.0, + "step": 19521 + }, + { + "epoch": 4.51, + "learning_rate": 1.24027424697635e-05, + "loss": 0.0, + "step": 19522 + }, + { + "epoch": 4.51, + "learning_rate": 1.240081657807565e-05, + "loss": 0.0, + "step": 19523 + }, + { + "epoch": 4.51, + "learning_rate": 1.2398890686387799e-05, + "loss": 0.0, + "step": 19524 + }, + { + "epoch": 4.51, + "learning_rate": 1.2396964794699947e-05, + "loss": 0.0, + "step": 19525 + }, + { + "epoch": 4.51, + "learning_rate": 1.2395038903012094e-05, + "loss": 0.0, + "step": 19526 + }, + { + "epoch": 4.51, + "learning_rate": 1.2393113011324243e-05, + "loss": 0.0, + "step": 19527 + }, + { + "epoch": 4.51, + "learning_rate": 1.2391187119636392e-05, + "loss": 0.0, + "step": 19528 + }, + { + "epoch": 4.51, + "learning_rate": 1.2389261227948542e-05, + "loss": 0.0, + "step": 19529 + }, + { + "epoch": 4.51, + "learning_rate": 1.2387335336260689e-05, + "loss": 0.0, + "step": 19530 + }, + { + "epoch": 4.51, + "learning_rate": 1.2385409444572838e-05, + "loss": 0.0, + "step": 19531 + }, + { + "epoch": 4.51, + "learning_rate": 1.2383483552884986e-05, + "loss": 0.0, + "step": 19532 + }, + { + "epoch": 4.51, + "learning_rate": 1.2381557661197135e-05, + "loss": 0.0, + "step": 19533 + }, + { + "epoch": 4.51, + "learning_rate": 1.2379631769509284e-05, + "loss": 0.0, + "step": 19534 + }, + { + "epoch": 4.51, + "learning_rate": 1.2377705877821432e-05, + "loss": 0.0, + "step": 19535 + }, + { + "epoch": 4.51, + "learning_rate": 1.237577998613358e-05, + "loss": 0.0, + "step": 19536 + }, + { + "epoch": 4.52, + "learning_rate": 1.237385409444573e-05, + "loss": 0.0, + "step": 19537 + }, + { + "epoch": 4.52, + "learning_rate": 1.2371928202757877e-05, + "loss": 0.0, + "step": 19538 + }, + { + "epoch": 4.52, + "learning_rate": 1.2370002311070026e-05, + "loss": 0.0, + "step": 19539 + }, + { + "epoch": 4.52, + "learning_rate": 1.2368076419382174e-05, + "loss": 0.0, + "step": 19540 + }, + { + "epoch": 4.52, + "learning_rate": 1.2366150527694323e-05, + "loss": 0.0, + "step": 19541 + }, + { + "epoch": 4.52, + "learning_rate": 1.2364224636006472e-05, + "loss": 0.0, + "step": 19542 + }, + { + "epoch": 4.52, + "learning_rate": 1.236229874431862e-05, + "loss": 0.0, + "step": 19543 + }, + { + "epoch": 4.52, + "learning_rate": 1.2360372852630767e-05, + "loss": 0.0, + "step": 19544 + }, + { + "epoch": 4.52, + "learning_rate": 1.2358446960942916e-05, + "loss": 0.0, + "step": 19545 + }, + { + "epoch": 4.52, + "learning_rate": 1.2356521069255066e-05, + "loss": 0.0, + "step": 19546 + }, + { + "epoch": 4.52, + "learning_rate": 1.2354595177567215e-05, + "loss": 0.0, + "step": 19547 + }, + { + "epoch": 4.52, + "learning_rate": 1.2352669285879364e-05, + "loss": 0.0, + "step": 19548 + }, + { + "epoch": 4.52, + "learning_rate": 1.2350743394191511e-05, + "loss": 0.0, + "step": 19549 + }, + { + "epoch": 4.52, + "learning_rate": 1.2348817502503659e-05, + "loss": 0.0, + "step": 19550 + }, + { + "epoch": 4.52, + "learning_rate": 1.2346891610815808e-05, + "loss": 0.0, + "step": 19551 + }, + { + "epoch": 4.52, + "learning_rate": 1.2344965719127957e-05, + "loss": 0.0, + "step": 19552 + }, + { + "epoch": 4.52, + "learning_rate": 1.2343039827440106e-05, + "loss": 0.0, + "step": 19553 + }, + { + "epoch": 4.52, + "learning_rate": 1.2341113935752254e-05, + "loss": 0.0, + "step": 19554 + }, + { + "epoch": 4.52, + "learning_rate": 1.2339188044064403e-05, + "loss": 0.0, + "step": 19555 + }, + { + "epoch": 4.52, + "learning_rate": 1.233726215237655e-05, + "loss": 0.0, + "step": 19556 + }, + { + "epoch": 4.52, + "learning_rate": 1.23353362606887e-05, + "loss": 0.0, + "step": 19557 + }, + { + "epoch": 4.52, + "learning_rate": 1.2333410369000847e-05, + "loss": 0.0, + "step": 19558 + }, + { + "epoch": 4.52, + "learning_rate": 1.2331484477312996e-05, + "loss": 0.0, + "step": 19559 + }, + { + "epoch": 4.52, + "learning_rate": 1.2329558585625146e-05, + "loss": 0.0, + "step": 19560 + }, + { + "epoch": 4.52, + "learning_rate": 1.2327632693937293e-05, + "loss": 0.0, + "step": 19561 + }, + { + "epoch": 4.52, + "learning_rate": 1.2325706802249442e-05, + "loss": 0.0, + "step": 19562 + }, + { + "epoch": 4.52, + "learning_rate": 1.232378091056159e-05, + "loss": 0.0, + "step": 19563 + }, + { + "epoch": 4.52, + "learning_rate": 1.2321855018873739e-05, + "loss": 0.0, + "step": 19564 + }, + { + "epoch": 4.52, + "learning_rate": 1.2319929127185888e-05, + "loss": 0.0, + "step": 19565 + }, + { + "epoch": 4.52, + "learning_rate": 1.2318003235498037e-05, + "loss": 0.0, + "step": 19566 + }, + { + "epoch": 4.52, + "learning_rate": 1.2316077343810185e-05, + "loss": 0.0, + "step": 19567 + }, + { + "epoch": 4.52, + "learning_rate": 1.2314151452122332e-05, + "loss": 0.0, + "step": 19568 + }, + { + "epoch": 4.52, + "learning_rate": 1.2312225560434481e-05, + "loss": 0.0, + "step": 19569 + }, + { + "epoch": 4.52, + "learning_rate": 1.231029966874663e-05, + "loss": 0.0, + "step": 19570 + }, + { + "epoch": 4.52, + "learning_rate": 1.230837377705878e-05, + "loss": 0.0, + "step": 19571 + }, + { + "epoch": 4.52, + "learning_rate": 1.2306447885370927e-05, + "loss": 0.0, + "step": 19572 + }, + { + "epoch": 4.52, + "learning_rate": 1.2304521993683076e-05, + "loss": 0.0, + "step": 19573 + }, + { + "epoch": 4.52, + "learning_rate": 1.2302596101995224e-05, + "loss": 0.0, + "step": 19574 + }, + { + "epoch": 4.52, + "learning_rate": 1.2300670210307373e-05, + "loss": 0.0, + "step": 19575 + }, + { + "epoch": 4.52, + "learning_rate": 1.2298744318619522e-05, + "loss": 0.0, + "step": 19576 + }, + { + "epoch": 4.52, + "learning_rate": 1.229681842693167e-05, + "loss": 0.0, + "step": 19577 + }, + { + "epoch": 4.52, + "learning_rate": 1.2294892535243819e-05, + "loss": 0.0, + "step": 19578 + }, + { + "epoch": 4.52, + "learning_rate": 1.2292966643555966e-05, + "loss": 0.0, + "step": 19579 + }, + { + "epoch": 4.53, + "learning_rate": 1.2291040751868115e-05, + "loss": 0.0, + "step": 19580 + }, + { + "epoch": 4.53, + "learning_rate": 1.2289114860180265e-05, + "loss": 0.0, + "step": 19581 + }, + { + "epoch": 4.53, + "learning_rate": 1.2287188968492412e-05, + "loss": 0.0, + "step": 19582 + }, + { + "epoch": 4.53, + "learning_rate": 1.2285263076804561e-05, + "loss": 0.0, + "step": 19583 + }, + { + "epoch": 4.53, + "learning_rate": 1.228333718511671e-05, + "loss": 0.0, + "step": 19584 + }, + { + "epoch": 4.53, + "learning_rate": 1.2281411293428858e-05, + "loss": 0.0, + "step": 19585 + }, + { + "epoch": 4.53, + "learning_rate": 1.2279485401741005e-05, + "loss": 0.0, + "step": 19586 + }, + { + "epoch": 4.53, + "learning_rate": 1.2277559510053155e-05, + "loss": 0.0, + "step": 19587 + }, + { + "epoch": 4.53, + "learning_rate": 1.2275633618365304e-05, + "loss": 0.0, + "step": 19588 + }, + { + "epoch": 4.53, + "learning_rate": 1.2273707726677453e-05, + "loss": 0.0, + "step": 19589 + }, + { + "epoch": 4.53, + "learning_rate": 1.2271781834989602e-05, + "loss": 0.0, + "step": 19590 + }, + { + "epoch": 4.53, + "learning_rate": 1.226985594330175e-05, + "loss": 0.0, + "step": 19591 + }, + { + "epoch": 4.53, + "learning_rate": 1.2267930051613897e-05, + "loss": 0.0, + "step": 19592 + }, + { + "epoch": 4.53, + "learning_rate": 1.2266004159926046e-05, + "loss": 0.0003, + "step": 19593 + }, + { + "epoch": 4.53, + "learning_rate": 1.2264078268238195e-05, + "loss": 0.0, + "step": 19594 + }, + { + "epoch": 4.53, + "learning_rate": 1.2262152376550345e-05, + "loss": 0.0, + "step": 19595 + }, + { + "epoch": 4.53, + "learning_rate": 1.2260226484862492e-05, + "loss": 0.0, + "step": 19596 + }, + { + "epoch": 4.53, + "learning_rate": 1.225830059317464e-05, + "loss": 0.0004, + "step": 19597 + }, + { + "epoch": 4.53, + "learning_rate": 1.2256374701486789e-05, + "loss": 0.0, + "step": 19598 + }, + { + "epoch": 4.53, + "learning_rate": 1.2254448809798938e-05, + "loss": 0.0001, + "step": 19599 + }, + { + "epoch": 4.53, + "learning_rate": 1.2252522918111085e-05, + "loss": 0.0, + "step": 19600 + }, + { + "epoch": 4.53, + "learning_rate": 1.2250597026423234e-05, + "loss": 0.0, + "step": 19601 + }, + { + "epoch": 4.53, + "learning_rate": 1.2248671134735384e-05, + "loss": 0.0, + "step": 19602 + }, + { + "epoch": 4.53, + "learning_rate": 1.2246745243047531e-05, + "loss": 0.0, + "step": 19603 + }, + { + "epoch": 4.53, + "learning_rate": 1.224481935135968e-05, + "loss": 0.0, + "step": 19604 + }, + { + "epoch": 4.53, + "learning_rate": 1.2242893459671828e-05, + "loss": 0.0, + "step": 19605 + }, + { + "epoch": 4.53, + "learning_rate": 1.2240967567983977e-05, + "loss": 0.0, + "step": 19606 + }, + { + "epoch": 4.53, + "learning_rate": 1.2239041676296126e-05, + "loss": 0.0, + "step": 19607 + }, + { + "epoch": 4.53, + "learning_rate": 1.2237115784608275e-05, + "loss": 0.0, + "step": 19608 + }, + { + "epoch": 4.53, + "learning_rate": 1.2235189892920423e-05, + "loss": 0.0, + "step": 19609 + }, + { + "epoch": 4.53, + "learning_rate": 1.223326400123257e-05, + "loss": 0.0, + "step": 19610 + }, + { + "epoch": 4.53, + "learning_rate": 1.223133810954472e-05, + "loss": 0.0, + "step": 19611 + }, + { + "epoch": 4.53, + "learning_rate": 1.2229412217856869e-05, + "loss": 0.0, + "step": 19612 + }, + { + "epoch": 4.53, + "learning_rate": 1.2227486326169018e-05, + "loss": 0.0, + "step": 19613 + }, + { + "epoch": 4.53, + "learning_rate": 1.2225560434481165e-05, + "loss": 0.0, + "step": 19614 + }, + { + "epoch": 4.53, + "learning_rate": 1.2223634542793313e-05, + "loss": 0.0, + "step": 19615 + }, + { + "epoch": 4.53, + "learning_rate": 1.2221708651105462e-05, + "loss": 0.0, + "step": 19616 + }, + { + "epoch": 4.53, + "learning_rate": 1.2219782759417611e-05, + "loss": 0.0, + "step": 19617 + }, + { + "epoch": 4.53, + "learning_rate": 1.221785686772976e-05, + "loss": 0.0, + "step": 19618 + }, + { + "epoch": 4.53, + "learning_rate": 1.2215930976041908e-05, + "loss": 0.0, + "step": 19619 + }, + { + "epoch": 4.53, + "learning_rate": 1.2214005084354057e-05, + "loss": 0.0, + "step": 19620 + }, + { + "epoch": 4.53, + "learning_rate": 1.2212079192666204e-05, + "loss": 0.0, + "step": 19621 + }, + { + "epoch": 4.53, + "learning_rate": 1.2210153300978354e-05, + "loss": 0.0, + "step": 19622 + }, + { + "epoch": 4.54, + "learning_rate": 1.2208227409290503e-05, + "loss": 0.0, + "step": 19623 + }, + { + "epoch": 4.54, + "learning_rate": 1.220630151760265e-05, + "loss": 0.0, + "step": 19624 + }, + { + "epoch": 4.54, + "learning_rate": 1.22043756259148e-05, + "loss": 0.0, + "step": 19625 + }, + { + "epoch": 4.54, + "learning_rate": 1.2202449734226949e-05, + "loss": 0.0, + "step": 19626 + }, + { + "epoch": 4.54, + "learning_rate": 1.2200523842539096e-05, + "loss": 0.0, + "step": 19627 + }, + { + "epoch": 4.54, + "learning_rate": 1.2198597950851244e-05, + "loss": 0.0, + "step": 19628 + }, + { + "epoch": 4.54, + "learning_rate": 1.2196672059163393e-05, + "loss": 0.0, + "step": 19629 + }, + { + "epoch": 4.54, + "learning_rate": 1.2194746167475542e-05, + "loss": 0.0001, + "step": 19630 + }, + { + "epoch": 4.54, + "learning_rate": 1.2192820275787691e-05, + "loss": 0.0, + "step": 19631 + }, + { + "epoch": 4.54, + "learning_rate": 1.219089438409984e-05, + "loss": 0.0, + "step": 19632 + }, + { + "epoch": 4.54, + "learning_rate": 1.2188968492411986e-05, + "loss": 0.0, + "step": 19633 + }, + { + "epoch": 4.54, + "learning_rate": 1.2187042600724135e-05, + "loss": 0.0, + "step": 19634 + }, + { + "epoch": 4.54, + "learning_rate": 1.2185116709036284e-05, + "loss": 0.0, + "step": 19635 + }, + { + "epoch": 4.54, + "learning_rate": 1.2183190817348433e-05, + "loss": 0.0, + "step": 19636 + }, + { + "epoch": 4.54, + "learning_rate": 1.2181264925660583e-05, + "loss": 0.0, + "step": 19637 + }, + { + "epoch": 4.54, + "learning_rate": 1.217933903397273e-05, + "loss": 0.0, + "step": 19638 + }, + { + "epoch": 4.54, + "learning_rate": 1.2177413142284878e-05, + "loss": 0.0, + "step": 19639 + }, + { + "epoch": 4.54, + "learning_rate": 1.2175487250597027e-05, + "loss": 0.0, + "step": 19640 + }, + { + "epoch": 4.54, + "learning_rate": 1.2173561358909176e-05, + "loss": 0.0, + "step": 19641 + }, + { + "epoch": 4.54, + "learning_rate": 1.2171635467221323e-05, + "loss": 0.0, + "step": 19642 + }, + { + "epoch": 4.54, + "learning_rate": 1.2169709575533473e-05, + "loss": 0.0, + "step": 19643 + }, + { + "epoch": 4.54, + "learning_rate": 1.2167783683845622e-05, + "loss": 0.0, + "step": 19644 + }, + { + "epoch": 4.54, + "learning_rate": 1.216585779215777e-05, + "loss": 0.0, + "step": 19645 + }, + { + "epoch": 4.54, + "learning_rate": 1.2163931900469918e-05, + "loss": 0.0, + "step": 19646 + }, + { + "epoch": 4.54, + "learning_rate": 1.2162006008782066e-05, + "loss": 0.0, + "step": 19647 + }, + { + "epoch": 4.54, + "learning_rate": 1.2160080117094215e-05, + "loss": 0.0, + "step": 19648 + }, + { + "epoch": 4.54, + "learning_rate": 1.2158154225406364e-05, + "loss": 0.0, + "step": 19649 + }, + { + "epoch": 4.54, + "learning_rate": 1.2156228333718513e-05, + "loss": 0.0, + "step": 19650 + }, + { + "epoch": 4.54, + "learning_rate": 1.2154302442030661e-05, + "loss": 0.0, + "step": 19651 + }, + { + "epoch": 4.54, + "learning_rate": 1.2152376550342808e-05, + "loss": 0.0001, + "step": 19652 + }, + { + "epoch": 4.54, + "learning_rate": 1.2150450658654958e-05, + "loss": 0.0, + "step": 19653 + }, + { + "epoch": 4.54, + "learning_rate": 1.2148524766967107e-05, + "loss": 0.0, + "step": 19654 + }, + { + "epoch": 4.54, + "learning_rate": 1.2146598875279256e-05, + "loss": 0.0, + "step": 19655 + }, + { + "epoch": 4.54, + "learning_rate": 1.2144672983591403e-05, + "loss": 0.0, + "step": 19656 + }, + { + "epoch": 4.54, + "learning_rate": 1.2142747091903551e-05, + "loss": 0.0, + "step": 19657 + }, + { + "epoch": 4.54, + "learning_rate": 1.21408212002157e-05, + "loss": 0.0, + "step": 19658 + }, + { + "epoch": 4.54, + "learning_rate": 1.213889530852785e-05, + "loss": 0.0, + "step": 19659 + }, + { + "epoch": 4.54, + "learning_rate": 1.2136969416839998e-05, + "loss": 0.0, + "step": 19660 + }, + { + "epoch": 4.54, + "learning_rate": 1.2135043525152146e-05, + "loss": 0.0, + "step": 19661 + }, + { + "epoch": 4.54, + "learning_rate": 1.2133117633464295e-05, + "loss": 0.0, + "step": 19662 + }, + { + "epoch": 4.54, + "learning_rate": 1.2131191741776442e-05, + "loss": 0.0, + "step": 19663 + }, + { + "epoch": 4.54, + "learning_rate": 1.2129265850088592e-05, + "loss": 0.0, + "step": 19664 + }, + { + "epoch": 4.54, + "learning_rate": 1.212733995840074e-05, + "loss": 0.0, + "step": 19665 + }, + { + "epoch": 4.54, + "learning_rate": 1.2125414066712888e-05, + "loss": 0.0, + "step": 19666 + }, + { + "epoch": 4.55, + "learning_rate": 1.2123488175025037e-05, + "loss": 0.0, + "step": 19667 + }, + { + "epoch": 4.55, + "learning_rate": 1.2121562283337187e-05, + "loss": 0.0, + "step": 19668 + }, + { + "epoch": 4.55, + "learning_rate": 1.2119636391649334e-05, + "loss": 0.0, + "step": 19669 + }, + { + "epoch": 4.55, + "learning_rate": 1.2117710499961482e-05, + "loss": 0.0, + "step": 19670 + }, + { + "epoch": 4.55, + "learning_rate": 1.211578460827363e-05, + "loss": 0.0, + "step": 19671 + }, + { + "epoch": 4.55, + "learning_rate": 1.211385871658578e-05, + "loss": 0.0, + "step": 19672 + }, + { + "epoch": 4.55, + "learning_rate": 1.2111932824897929e-05, + "loss": 0.0, + "step": 19673 + }, + { + "epoch": 4.55, + "learning_rate": 1.2110006933210077e-05, + "loss": 0.0, + "step": 19674 + }, + { + "epoch": 4.55, + "learning_rate": 1.2108081041522224e-05, + "loss": 0.0, + "step": 19675 + }, + { + "epoch": 4.55, + "learning_rate": 1.2106155149834373e-05, + "loss": 0.0, + "step": 19676 + }, + { + "epoch": 4.55, + "learning_rate": 1.2104229258146522e-05, + "loss": 0.0002, + "step": 19677 + }, + { + "epoch": 4.55, + "learning_rate": 1.2102303366458672e-05, + "loss": 0.0, + "step": 19678 + }, + { + "epoch": 4.55, + "learning_rate": 1.210037747477082e-05, + "loss": 0.0, + "step": 19679 + }, + { + "epoch": 4.55, + "learning_rate": 1.2098451583082968e-05, + "loss": 0.0, + "step": 19680 + }, + { + "epoch": 4.55, + "learning_rate": 1.2096525691395116e-05, + "loss": 0.0, + "step": 19681 + }, + { + "epoch": 4.55, + "learning_rate": 1.2094599799707265e-05, + "loss": 0.0, + "step": 19682 + }, + { + "epoch": 4.55, + "learning_rate": 1.2092673908019414e-05, + "loss": 0.0, + "step": 19683 + }, + { + "epoch": 4.55, + "learning_rate": 1.2090748016331562e-05, + "loss": 0.0, + "step": 19684 + }, + { + "epoch": 4.55, + "learning_rate": 1.208882212464371e-05, + "loss": 0.0, + "step": 19685 + }, + { + "epoch": 4.55, + "learning_rate": 1.208689623295586e-05, + "loss": 0.0, + "step": 19686 + }, + { + "epoch": 4.55, + "learning_rate": 1.2084970341268007e-05, + "loss": 0.0, + "step": 19687 + }, + { + "epoch": 4.55, + "learning_rate": 1.2083044449580157e-05, + "loss": 0.0, + "step": 19688 + }, + { + "epoch": 4.55, + "learning_rate": 1.2081118557892304e-05, + "loss": 0.0, + "step": 19689 + }, + { + "epoch": 4.55, + "learning_rate": 1.2079192666204453e-05, + "loss": 0.0001, + "step": 19690 + }, + { + "epoch": 4.55, + "learning_rate": 1.2077266774516602e-05, + "loss": 0.0, + "step": 19691 + }, + { + "epoch": 4.55, + "learning_rate": 1.207534088282875e-05, + "loss": 0.0, + "step": 19692 + }, + { + "epoch": 4.55, + "learning_rate": 1.2073414991140899e-05, + "loss": 0.0, + "step": 19693 + }, + { + "epoch": 4.55, + "learning_rate": 1.2071489099453046e-05, + "loss": 0.0, + "step": 19694 + }, + { + "epoch": 4.55, + "learning_rate": 1.2069563207765196e-05, + "loss": 0.0, + "step": 19695 + }, + { + "epoch": 4.55, + "learning_rate": 1.2067637316077345e-05, + "loss": 0.0, + "step": 19696 + }, + { + "epoch": 4.55, + "learning_rate": 1.2065711424389494e-05, + "loss": 0.0, + "step": 19697 + }, + { + "epoch": 4.55, + "learning_rate": 1.2063785532701641e-05, + "loss": 0.0, + "step": 19698 + }, + { + "epoch": 4.55, + "learning_rate": 1.2061859641013789e-05, + "loss": 0.0, + "step": 19699 + }, + { + "epoch": 4.55, + "learning_rate": 1.2059933749325938e-05, + "loss": 0.0002, + "step": 19700 + }, + { + "epoch": 4.55, + "learning_rate": 1.2058007857638087e-05, + "loss": 0.0, + "step": 19701 + }, + { + "epoch": 4.55, + "learning_rate": 1.2056081965950236e-05, + "loss": 0.0001, + "step": 19702 + }, + { + "epoch": 4.55, + "learning_rate": 1.2054156074262384e-05, + "loss": 0.0, + "step": 19703 + }, + { + "epoch": 4.55, + "learning_rate": 1.2052230182574533e-05, + "loss": 0.0, + "step": 19704 + }, + { + "epoch": 4.55, + "learning_rate": 1.205030429088668e-05, + "loss": 0.0, + "step": 19705 + }, + { + "epoch": 4.55, + "learning_rate": 1.204837839919883e-05, + "loss": 0.0, + "step": 19706 + }, + { + "epoch": 4.55, + "learning_rate": 1.2046452507510979e-05, + "loss": 0.0, + "step": 19707 + }, + { + "epoch": 4.55, + "learning_rate": 1.2044526615823126e-05, + "loss": 0.0, + "step": 19708 + }, + { + "epoch": 4.55, + "learning_rate": 1.2042600724135276e-05, + "loss": 0.0, + "step": 19709 + }, + { + "epoch": 4.56, + "learning_rate": 1.2040674832447423e-05, + "loss": 0.0, + "step": 19710 + }, + { + "epoch": 4.56, + "learning_rate": 1.2038748940759572e-05, + "loss": 0.0, + "step": 19711 + }, + { + "epoch": 4.56, + "learning_rate": 1.203682304907172e-05, + "loss": 0.0, + "step": 19712 + }, + { + "epoch": 4.56, + "learning_rate": 1.2034897157383869e-05, + "loss": 0.0, + "step": 19713 + }, + { + "epoch": 4.56, + "learning_rate": 1.2032971265696018e-05, + "loss": 0.0, + "step": 19714 + }, + { + "epoch": 4.56, + "learning_rate": 1.2031045374008167e-05, + "loss": 0.0, + "step": 19715 + }, + { + "epoch": 4.56, + "learning_rate": 1.2029119482320315e-05, + "loss": 0.0, + "step": 19716 + }, + { + "epoch": 4.56, + "learning_rate": 1.2027193590632462e-05, + "loss": 0.0, + "step": 19717 + }, + { + "epoch": 4.56, + "learning_rate": 1.2025267698944611e-05, + "loss": 0.0, + "step": 19718 + }, + { + "epoch": 4.56, + "learning_rate": 1.202334180725676e-05, + "loss": 0.0, + "step": 19719 + }, + { + "epoch": 4.56, + "learning_rate": 1.202141591556891e-05, + "loss": 0.0, + "step": 19720 + }, + { + "epoch": 4.56, + "learning_rate": 1.2019490023881059e-05, + "loss": 0.0, + "step": 19721 + }, + { + "epoch": 4.56, + "learning_rate": 1.2017564132193206e-05, + "loss": 0.0, + "step": 19722 + }, + { + "epoch": 4.56, + "learning_rate": 1.2015638240505354e-05, + "loss": 0.0, + "step": 19723 + }, + { + "epoch": 4.56, + "learning_rate": 1.2013712348817503e-05, + "loss": 0.0, + "step": 19724 + }, + { + "epoch": 4.56, + "learning_rate": 1.2011786457129652e-05, + "loss": 0.0, + "step": 19725 + }, + { + "epoch": 4.56, + "learning_rate": 1.20098605654418e-05, + "loss": 0.0, + "step": 19726 + }, + { + "epoch": 4.56, + "learning_rate": 1.2007934673753949e-05, + "loss": 0.0, + "step": 19727 + }, + { + "epoch": 4.56, + "learning_rate": 1.2006008782066096e-05, + "loss": 0.0, + "step": 19728 + }, + { + "epoch": 4.56, + "learning_rate": 1.2004082890378245e-05, + "loss": 0.0, + "step": 19729 + }, + { + "epoch": 4.56, + "learning_rate": 1.2002156998690395e-05, + "loss": 0.0, + "step": 19730 + }, + { + "epoch": 4.56, + "learning_rate": 1.2000231107002542e-05, + "loss": 0.0001, + "step": 19731 + }, + { + "epoch": 4.56, + "learning_rate": 1.1998305215314691e-05, + "loss": 0.0, + "step": 19732 + }, + { + "epoch": 4.56, + "learning_rate": 1.199637932362684e-05, + "loss": 0.0, + "step": 19733 + }, + { + "epoch": 4.56, + "learning_rate": 1.1994453431938988e-05, + "loss": 0.0, + "step": 19734 + }, + { + "epoch": 4.56, + "learning_rate": 1.1992527540251137e-05, + "loss": 0.0, + "step": 19735 + }, + { + "epoch": 4.56, + "learning_rate": 1.1990601648563285e-05, + "loss": 0.0, + "step": 19736 + }, + { + "epoch": 4.56, + "learning_rate": 1.1988675756875434e-05, + "loss": 0.0, + "step": 19737 + }, + { + "epoch": 4.56, + "learning_rate": 1.1986749865187583e-05, + "loss": 0.0, + "step": 19738 + }, + { + "epoch": 4.56, + "learning_rate": 1.1984823973499732e-05, + "loss": 0.0, + "step": 19739 + }, + { + "epoch": 4.56, + "learning_rate": 1.198289808181188e-05, + "loss": 0.0, + "step": 19740 + }, + { + "epoch": 4.56, + "learning_rate": 1.1980972190124027e-05, + "loss": 0.0, + "step": 19741 + }, + { + "epoch": 4.56, + "learning_rate": 1.1979046298436176e-05, + "loss": 0.0, + "step": 19742 + }, + { + "epoch": 4.56, + "learning_rate": 1.1977120406748325e-05, + "loss": 0.0, + "step": 19743 + }, + { + "epoch": 4.56, + "learning_rate": 1.1975194515060475e-05, + "loss": 0.0, + "step": 19744 + }, + { + "epoch": 4.56, + "learning_rate": 1.1973268623372622e-05, + "loss": 0.0, + "step": 19745 + }, + { + "epoch": 4.56, + "learning_rate": 1.197134273168477e-05, + "loss": 0.0, + "step": 19746 + }, + { + "epoch": 4.56, + "learning_rate": 1.1969416839996919e-05, + "loss": 0.0, + "step": 19747 + }, + { + "epoch": 4.56, + "learning_rate": 1.1967490948309068e-05, + "loss": 0.0, + "step": 19748 + }, + { + "epoch": 4.56, + "learning_rate": 1.1965565056621217e-05, + "loss": 0.0, + "step": 19749 + }, + { + "epoch": 4.56, + "learning_rate": 1.1963639164933365e-05, + "loss": 0.0, + "step": 19750 + }, + { + "epoch": 4.56, + "learning_rate": 1.1961713273245514e-05, + "loss": 0.0, + "step": 19751 + }, + { + "epoch": 4.56, + "learning_rate": 1.1959787381557661e-05, + "loss": 0.0, + "step": 19752 + }, + { + "epoch": 4.57, + "learning_rate": 1.195786148986981e-05, + "loss": 0.0, + "step": 19753 + }, + { + "epoch": 4.57, + "learning_rate": 1.1955935598181958e-05, + "loss": 0.0, + "step": 19754 + }, + { + "epoch": 4.57, + "learning_rate": 1.1954009706494107e-05, + "loss": 0.0, + "step": 19755 + }, + { + "epoch": 4.57, + "learning_rate": 1.1952083814806256e-05, + "loss": 0.0, + "step": 19756 + }, + { + "epoch": 4.57, + "learning_rate": 1.1950157923118405e-05, + "loss": 0.0, + "step": 19757 + }, + { + "epoch": 4.57, + "learning_rate": 1.1948232031430553e-05, + "loss": 0.0, + "step": 19758 + }, + { + "epoch": 4.57, + "learning_rate": 1.19463061397427e-05, + "loss": 0.0, + "step": 19759 + }, + { + "epoch": 4.57, + "learning_rate": 1.194438024805485e-05, + "loss": 0.0, + "step": 19760 + }, + { + "epoch": 4.57, + "learning_rate": 1.1942454356366999e-05, + "loss": 0.0003, + "step": 19761 + }, + { + "epoch": 4.57, + "learning_rate": 1.1940528464679148e-05, + "loss": 0.0001, + "step": 19762 + }, + { + "epoch": 4.57, + "learning_rate": 1.1938602572991297e-05, + "loss": 0.0, + "step": 19763 + }, + { + "epoch": 4.57, + "learning_rate": 1.1936676681303444e-05, + "loss": 0.0, + "step": 19764 + }, + { + "epoch": 4.57, + "learning_rate": 1.1934750789615592e-05, + "loss": 0.0, + "step": 19765 + }, + { + "epoch": 4.57, + "learning_rate": 1.1932824897927741e-05, + "loss": 0.0, + "step": 19766 + }, + { + "epoch": 4.57, + "learning_rate": 1.193089900623989e-05, + "loss": 0.0, + "step": 19767 + }, + { + "epoch": 4.57, + "learning_rate": 1.1928973114552038e-05, + "loss": 0.0, + "step": 19768 + }, + { + "epoch": 4.57, + "learning_rate": 1.1927047222864187e-05, + "loss": 0.0, + "step": 19769 + }, + { + "epoch": 4.57, + "learning_rate": 1.1925121331176334e-05, + "loss": 0.0, + "step": 19770 + }, + { + "epoch": 4.57, + "learning_rate": 1.1923195439488484e-05, + "loss": 0.015, + "step": 19771 + }, + { + "epoch": 4.57, + "learning_rate": 1.1921269547800633e-05, + "loss": 0.0004, + "step": 19772 + }, + { + "epoch": 4.57, + "learning_rate": 1.191934365611278e-05, + "loss": 0.0, + "step": 19773 + }, + { + "epoch": 4.57, + "learning_rate": 1.191741776442493e-05, + "loss": 0.0001, + "step": 19774 + }, + { + "epoch": 4.57, + "learning_rate": 1.1915491872737079e-05, + "loss": 0.0, + "step": 19775 + }, + { + "epoch": 4.57, + "learning_rate": 1.1913565981049226e-05, + "loss": 0.0, + "step": 19776 + }, + { + "epoch": 4.57, + "learning_rate": 1.1911640089361375e-05, + "loss": 0.0, + "step": 19777 + }, + { + "epoch": 4.57, + "learning_rate": 1.1909714197673523e-05, + "loss": 0.0, + "step": 19778 + }, + { + "epoch": 4.57, + "learning_rate": 1.1907788305985672e-05, + "loss": 0.0, + "step": 19779 + }, + { + "epoch": 4.57, + "learning_rate": 1.1905862414297821e-05, + "loss": 0.0, + "step": 19780 + }, + { + "epoch": 4.57, + "learning_rate": 1.190393652260997e-05, + "loss": 0.0, + "step": 19781 + }, + { + "epoch": 4.57, + "learning_rate": 1.1902010630922118e-05, + "loss": 0.0, + "step": 19782 + }, + { + "epoch": 4.57, + "learning_rate": 1.1900084739234265e-05, + "loss": 0.0, + "step": 19783 + }, + { + "epoch": 4.57, + "learning_rate": 1.1898158847546414e-05, + "loss": 0.0, + "step": 19784 + }, + { + "epoch": 4.57, + "learning_rate": 1.1896232955858564e-05, + "loss": 0.0, + "step": 19785 + }, + { + "epoch": 4.57, + "learning_rate": 1.1894307064170713e-05, + "loss": 0.0, + "step": 19786 + }, + { + "epoch": 4.57, + "learning_rate": 1.189238117248286e-05, + "loss": 0.0, + "step": 19787 + }, + { + "epoch": 4.57, + "learning_rate": 1.1890455280795008e-05, + "loss": 0.0, + "step": 19788 + }, + { + "epoch": 4.57, + "learning_rate": 1.1888529389107157e-05, + "loss": 0.0, + "step": 19789 + }, + { + "epoch": 4.57, + "learning_rate": 1.1886603497419306e-05, + "loss": 0.0, + "step": 19790 + }, + { + "epoch": 4.57, + "learning_rate": 1.1884677605731455e-05, + "loss": 0.0, + "step": 19791 + }, + { + "epoch": 4.57, + "learning_rate": 1.1882751714043603e-05, + "loss": 0.0, + "step": 19792 + }, + { + "epoch": 4.57, + "learning_rate": 1.1880825822355752e-05, + "loss": 0.0, + "step": 19793 + }, + { + "epoch": 4.57, + "learning_rate": 1.18788999306679e-05, + "loss": 0.0, + "step": 19794 + }, + { + "epoch": 4.57, + "learning_rate": 1.1876974038980048e-05, + "loss": 0.0, + "step": 19795 + }, + { + "epoch": 4.57, + "learning_rate": 1.1875048147292196e-05, + "loss": 0.0, + "step": 19796 + }, + { + "epoch": 4.58, + "learning_rate": 1.1873122255604345e-05, + "loss": 0.0, + "step": 19797 + }, + { + "epoch": 4.58, + "learning_rate": 1.1871196363916494e-05, + "loss": 0.0, + "step": 19798 + }, + { + "epoch": 4.58, + "learning_rate": 1.1869270472228643e-05, + "loss": 0.0, + "step": 19799 + }, + { + "epoch": 4.58, + "learning_rate": 1.1867344580540791e-05, + "loss": 0.0, + "step": 19800 + }, + { + "epoch": 4.58, + "learning_rate": 1.1865418688852938e-05, + "loss": 0.0, + "step": 19801 + }, + { + "epoch": 4.58, + "learning_rate": 1.1863492797165088e-05, + "loss": 0.0, + "step": 19802 + }, + { + "epoch": 4.58, + "learning_rate": 1.1861566905477237e-05, + "loss": 0.0, + "step": 19803 + }, + { + "epoch": 4.58, + "learning_rate": 1.1859641013789386e-05, + "loss": 0.0, + "step": 19804 + }, + { + "epoch": 4.58, + "learning_rate": 1.1857715122101533e-05, + "loss": 0.0, + "step": 19805 + }, + { + "epoch": 4.58, + "learning_rate": 1.1855789230413681e-05, + "loss": 0.0, + "step": 19806 + }, + { + "epoch": 4.58, + "learning_rate": 1.185386333872583e-05, + "loss": 0.0, + "step": 19807 + }, + { + "epoch": 4.58, + "learning_rate": 1.185193744703798e-05, + "loss": 0.0001, + "step": 19808 + }, + { + "epoch": 4.58, + "learning_rate": 1.1850011555350128e-05, + "loss": 0.0, + "step": 19809 + }, + { + "epoch": 4.58, + "learning_rate": 1.1848085663662276e-05, + "loss": 0.0, + "step": 19810 + }, + { + "epoch": 4.58, + "learning_rate": 1.1846159771974425e-05, + "loss": 0.0, + "step": 19811 + }, + { + "epoch": 4.58, + "learning_rate": 1.1844233880286573e-05, + "loss": 0.0, + "step": 19812 + }, + { + "epoch": 4.58, + "learning_rate": 1.1842307988598722e-05, + "loss": 0.0, + "step": 19813 + }, + { + "epoch": 4.58, + "learning_rate": 1.1840382096910871e-05, + "loss": 0.0, + "step": 19814 + }, + { + "epoch": 4.58, + "learning_rate": 1.1838456205223018e-05, + "loss": 0.0, + "step": 19815 + }, + { + "epoch": 4.58, + "learning_rate": 1.1836530313535168e-05, + "loss": 0.0, + "step": 19816 + }, + { + "epoch": 4.58, + "learning_rate": 1.1834604421847317e-05, + "loss": 0.0, + "step": 19817 + }, + { + "epoch": 4.58, + "learning_rate": 1.1832678530159464e-05, + "loss": 0.0, + "step": 19818 + }, + { + "epoch": 4.58, + "learning_rate": 1.1830752638471613e-05, + "loss": 0.0, + "step": 19819 + }, + { + "epoch": 4.58, + "learning_rate": 1.182882674678376e-05, + "loss": 0.0, + "step": 19820 + }, + { + "epoch": 4.58, + "learning_rate": 1.182690085509591e-05, + "loss": 0.0, + "step": 19821 + }, + { + "epoch": 4.58, + "learning_rate": 1.182497496340806e-05, + "loss": 0.0, + "step": 19822 + }, + { + "epoch": 4.58, + "learning_rate": 1.1823049071720207e-05, + "loss": 0.0, + "step": 19823 + }, + { + "epoch": 4.58, + "learning_rate": 1.1821123180032354e-05, + "loss": 0.0, + "step": 19824 + }, + { + "epoch": 4.58, + "learning_rate": 1.1819197288344503e-05, + "loss": 0.0, + "step": 19825 + }, + { + "epoch": 4.58, + "learning_rate": 1.1817271396656652e-05, + "loss": 0.0, + "step": 19826 + }, + { + "epoch": 4.58, + "learning_rate": 1.1815345504968802e-05, + "loss": 0.0, + "step": 19827 + }, + { + "epoch": 4.58, + "learning_rate": 1.181341961328095e-05, + "loss": 0.0, + "step": 19828 + }, + { + "epoch": 4.58, + "learning_rate": 1.1811493721593098e-05, + "loss": 0.0, + "step": 19829 + }, + { + "epoch": 4.58, + "learning_rate": 1.1809567829905246e-05, + "loss": 0.0, + "step": 19830 + }, + { + "epoch": 4.58, + "learning_rate": 1.1807641938217395e-05, + "loss": 0.0, + "step": 19831 + }, + { + "epoch": 4.58, + "learning_rate": 1.1805716046529544e-05, + "loss": 0.0, + "step": 19832 + }, + { + "epoch": 4.58, + "learning_rate": 1.1803790154841693e-05, + "loss": 0.0001, + "step": 19833 + }, + { + "epoch": 4.58, + "learning_rate": 1.180186426315384e-05, + "loss": 0.0, + "step": 19834 + }, + { + "epoch": 4.58, + "learning_rate": 1.179993837146599e-05, + "loss": 0.0, + "step": 19835 + }, + { + "epoch": 4.58, + "learning_rate": 1.1798012479778137e-05, + "loss": 0.0, + "step": 19836 + }, + { + "epoch": 4.58, + "learning_rate": 1.1796086588090287e-05, + "loss": 0.0, + "step": 19837 + }, + { + "epoch": 4.58, + "learning_rate": 1.1794160696402434e-05, + "loss": 0.0, + "step": 19838 + }, + { + "epoch": 4.58, + "learning_rate": 1.1792234804714583e-05, + "loss": 0.0, + "step": 19839 + }, + { + "epoch": 4.59, + "learning_rate": 1.1790308913026732e-05, + "loss": 0.0, + "step": 19840 + }, + { + "epoch": 4.59, + "learning_rate": 1.178838302133888e-05, + "loss": 0.0, + "step": 19841 + }, + { + "epoch": 4.59, + "learning_rate": 1.1786457129651029e-05, + "loss": 0.0, + "step": 19842 + }, + { + "epoch": 4.59, + "learning_rate": 1.1784531237963177e-05, + "loss": 0.0, + "step": 19843 + }, + { + "epoch": 4.59, + "learning_rate": 1.1782605346275326e-05, + "loss": 0.0, + "step": 19844 + }, + { + "epoch": 4.59, + "learning_rate": 1.1780679454587475e-05, + "loss": 0.0, + "step": 19845 + }, + { + "epoch": 4.59, + "learning_rate": 1.1778753562899624e-05, + "loss": 0.0, + "step": 19846 + }, + { + "epoch": 4.59, + "learning_rate": 1.1776827671211772e-05, + "loss": 0.0, + "step": 19847 + }, + { + "epoch": 4.59, + "learning_rate": 1.1774901779523919e-05, + "loss": 0.0, + "step": 19848 + }, + { + "epoch": 4.59, + "learning_rate": 1.1772975887836068e-05, + "loss": 0.0, + "step": 19849 + }, + { + "epoch": 4.59, + "learning_rate": 1.1771049996148217e-05, + "loss": 0.0, + "step": 19850 + }, + { + "epoch": 4.59, + "learning_rate": 1.1769124104460367e-05, + "loss": 0.0, + "step": 19851 + }, + { + "epoch": 4.59, + "learning_rate": 1.1767198212772514e-05, + "loss": 0.0, + "step": 19852 + }, + { + "epoch": 4.59, + "learning_rate": 1.1765272321084663e-05, + "loss": 0.0, + "step": 19853 + }, + { + "epoch": 4.59, + "learning_rate": 1.176334642939681e-05, + "loss": 0.0, + "step": 19854 + }, + { + "epoch": 4.59, + "learning_rate": 1.176142053770896e-05, + "loss": 0.0001, + "step": 19855 + }, + { + "epoch": 4.59, + "learning_rate": 1.1759494646021109e-05, + "loss": 0.0, + "step": 19856 + }, + { + "epoch": 4.59, + "learning_rate": 1.1757568754333256e-05, + "loss": 0.0, + "step": 19857 + }, + { + "epoch": 4.59, + "learning_rate": 1.1755642862645406e-05, + "loss": 0.0, + "step": 19858 + }, + { + "epoch": 4.59, + "learning_rate": 1.1753716970957555e-05, + "loss": 0.0, + "step": 19859 + }, + { + "epoch": 4.59, + "learning_rate": 1.1751791079269702e-05, + "loss": 0.0, + "step": 19860 + }, + { + "epoch": 4.59, + "learning_rate": 1.1749865187581851e-05, + "loss": 0.0, + "step": 19861 + }, + { + "epoch": 4.59, + "learning_rate": 1.1747939295893999e-05, + "loss": 0.0, + "step": 19862 + }, + { + "epoch": 4.59, + "learning_rate": 1.1746013404206148e-05, + "loss": 0.0, + "step": 19863 + }, + { + "epoch": 4.59, + "learning_rate": 1.1744087512518297e-05, + "loss": 0.0, + "step": 19864 + }, + { + "epoch": 4.59, + "learning_rate": 1.1742161620830445e-05, + "loss": 0.0, + "step": 19865 + }, + { + "epoch": 4.59, + "learning_rate": 1.1740235729142592e-05, + "loss": 0.0, + "step": 19866 + }, + { + "epoch": 4.59, + "learning_rate": 1.1738309837454741e-05, + "loss": 0.0, + "step": 19867 + }, + { + "epoch": 4.59, + "learning_rate": 1.173638394576689e-05, + "loss": 0.0, + "step": 19868 + }, + { + "epoch": 4.59, + "learning_rate": 1.173445805407904e-05, + "loss": 0.0, + "step": 19869 + }, + { + "epoch": 4.59, + "learning_rate": 1.1732532162391189e-05, + "loss": 0.0, + "step": 19870 + }, + { + "epoch": 4.59, + "learning_rate": 1.1730606270703336e-05, + "loss": 0.0, + "step": 19871 + }, + { + "epoch": 4.59, + "learning_rate": 1.1728680379015484e-05, + "loss": 0.0, + "step": 19872 + }, + { + "epoch": 4.59, + "learning_rate": 1.1726754487327633e-05, + "loss": 0.0, + "step": 19873 + }, + { + "epoch": 4.59, + "learning_rate": 1.1724828595639782e-05, + "loss": 0.0, + "step": 19874 + }, + { + "epoch": 4.59, + "learning_rate": 1.1722902703951931e-05, + "loss": 0.0, + "step": 19875 + }, + { + "epoch": 4.59, + "learning_rate": 1.1720976812264079e-05, + "loss": 0.0, + "step": 19876 + }, + { + "epoch": 4.59, + "learning_rate": 1.1719050920576228e-05, + "loss": 0.0, + "step": 19877 + }, + { + "epoch": 4.59, + "learning_rate": 1.1717125028888376e-05, + "loss": 0.0, + "step": 19878 + }, + { + "epoch": 4.59, + "learning_rate": 1.1715199137200525e-05, + "loss": 0.0, + "step": 19879 + }, + { + "epoch": 4.59, + "learning_rate": 1.1713273245512672e-05, + "loss": 0.0, + "step": 19880 + }, + { + "epoch": 4.59, + "learning_rate": 1.1711347353824821e-05, + "loss": 0.0, + "step": 19881 + }, + { + "epoch": 4.59, + "learning_rate": 1.170942146213697e-05, + "loss": 0.0, + "step": 19882 + }, + { + "epoch": 4.6, + "learning_rate": 1.1707495570449118e-05, + "loss": 0.0, + "step": 19883 + }, + { + "epoch": 4.6, + "learning_rate": 1.1705569678761267e-05, + "loss": 0.0, + "step": 19884 + }, + { + "epoch": 4.6, + "learning_rate": 1.1703643787073415e-05, + "loss": 0.0, + "step": 19885 + }, + { + "epoch": 4.6, + "learning_rate": 1.1701717895385564e-05, + "loss": 0.0, + "step": 19886 + }, + { + "epoch": 4.6, + "learning_rate": 1.1699792003697713e-05, + "loss": 0.0, + "step": 19887 + }, + { + "epoch": 4.6, + "learning_rate": 1.1697866112009862e-05, + "loss": 0.0, + "step": 19888 + }, + { + "epoch": 4.6, + "learning_rate": 1.169594022032201e-05, + "loss": 0.0001, + "step": 19889 + }, + { + "epoch": 4.6, + "learning_rate": 1.1694014328634157e-05, + "loss": 0.0, + "step": 19890 + }, + { + "epoch": 4.6, + "learning_rate": 1.1692088436946306e-05, + "loss": 0.0, + "step": 19891 + }, + { + "epoch": 4.6, + "learning_rate": 1.1690162545258455e-05, + "loss": 0.0, + "step": 19892 + }, + { + "epoch": 4.6, + "learning_rate": 1.1688236653570605e-05, + "loss": 0.0, + "step": 19893 + }, + { + "epoch": 4.6, + "learning_rate": 1.1686310761882752e-05, + "loss": 0.0, + "step": 19894 + }, + { + "epoch": 4.6, + "learning_rate": 1.1684384870194901e-05, + "loss": 0.0, + "step": 19895 + }, + { + "epoch": 4.6, + "learning_rate": 1.1682458978507049e-05, + "loss": 0.0, + "step": 19896 + }, + { + "epoch": 4.6, + "learning_rate": 1.1680533086819198e-05, + "loss": 0.0, + "step": 19897 + }, + { + "epoch": 4.6, + "learning_rate": 1.1678607195131347e-05, + "loss": 0.0, + "step": 19898 + }, + { + "epoch": 4.6, + "learning_rate": 1.1676681303443495e-05, + "loss": 0.0001, + "step": 19899 + }, + { + "epoch": 4.6, + "learning_rate": 1.1674755411755644e-05, + "loss": 0.0, + "step": 19900 + }, + { + "epoch": 4.6, + "learning_rate": 1.1672829520067791e-05, + "loss": 0.0, + "step": 19901 + }, + { + "epoch": 4.6, + "learning_rate": 1.167090362837994e-05, + "loss": 0.0, + "step": 19902 + }, + { + "epoch": 4.6, + "learning_rate": 1.166897773669209e-05, + "loss": 0.0, + "step": 19903 + }, + { + "epoch": 4.6, + "learning_rate": 1.1667051845004237e-05, + "loss": 0.0001, + "step": 19904 + }, + { + "epoch": 4.6, + "learning_rate": 1.1665125953316386e-05, + "loss": 0.0, + "step": 19905 + }, + { + "epoch": 4.6, + "learning_rate": 1.1663200061628535e-05, + "loss": 0.0, + "step": 19906 + }, + { + "epoch": 4.6, + "learning_rate": 1.1661274169940683e-05, + "loss": 0.0, + "step": 19907 + }, + { + "epoch": 4.6, + "learning_rate": 1.165934827825283e-05, + "loss": 0.0, + "step": 19908 + }, + { + "epoch": 4.6, + "learning_rate": 1.165742238656498e-05, + "loss": 0.0, + "step": 19909 + }, + { + "epoch": 4.6, + "learning_rate": 1.1655496494877129e-05, + "loss": 0.0, + "step": 19910 + }, + { + "epoch": 4.6, + "learning_rate": 1.1653570603189278e-05, + "loss": 0.0, + "step": 19911 + }, + { + "epoch": 4.6, + "learning_rate": 1.1651644711501427e-05, + "loss": 0.0, + "step": 19912 + }, + { + "epoch": 4.6, + "learning_rate": 1.1649718819813575e-05, + "loss": 0.0, + "step": 19913 + }, + { + "epoch": 4.6, + "learning_rate": 1.1647792928125722e-05, + "loss": 0.0, + "step": 19914 + }, + { + "epoch": 4.6, + "learning_rate": 1.1645867036437871e-05, + "loss": 0.0, + "step": 19915 + }, + { + "epoch": 4.6, + "learning_rate": 1.164394114475002e-05, + "loss": 0.0, + "step": 19916 + }, + { + "epoch": 4.6, + "learning_rate": 1.164201525306217e-05, + "loss": 0.0, + "step": 19917 + }, + { + "epoch": 4.6, + "learning_rate": 1.1640089361374317e-05, + "loss": 0.0, + "step": 19918 + }, + { + "epoch": 4.6, + "learning_rate": 1.1638163469686464e-05, + "loss": 0.0139, + "step": 19919 + }, + { + "epoch": 4.6, + "learning_rate": 1.1636237577998614e-05, + "loss": 0.0, + "step": 19920 + }, + { + "epoch": 4.6, + "learning_rate": 1.1634311686310763e-05, + "loss": 0.0, + "step": 19921 + }, + { + "epoch": 4.6, + "learning_rate": 1.163238579462291e-05, + "loss": 0.0, + "step": 19922 + }, + { + "epoch": 4.6, + "learning_rate": 1.163045990293506e-05, + "loss": 0.0, + "step": 19923 + }, + { + "epoch": 4.6, + "learning_rate": 1.1628534011247209e-05, + "loss": 0.0, + "step": 19924 + }, + { + "epoch": 4.6, + "learning_rate": 1.1626608119559356e-05, + "loss": 0.0, + "step": 19925 + }, + { + "epoch": 4.61, + "learning_rate": 1.1624682227871505e-05, + "loss": 0.0, + "step": 19926 + }, + { + "epoch": 4.61, + "learning_rate": 1.1622756336183653e-05, + "loss": 0.0, + "step": 19927 + }, + { + "epoch": 4.61, + "learning_rate": 1.1620830444495802e-05, + "loss": 0.0, + "step": 19928 + }, + { + "epoch": 4.61, + "learning_rate": 1.1618904552807951e-05, + "loss": 0.0001, + "step": 19929 + }, + { + "epoch": 4.61, + "learning_rate": 1.16169786611201e-05, + "loss": 0.0, + "step": 19930 + }, + { + "epoch": 4.61, + "learning_rate": 1.1615052769432248e-05, + "loss": 0.0, + "step": 19931 + }, + { + "epoch": 4.61, + "learning_rate": 1.1613126877744395e-05, + "loss": 0.0, + "step": 19932 + }, + { + "epoch": 4.61, + "learning_rate": 1.1611200986056544e-05, + "loss": 0.0, + "step": 19933 + }, + { + "epoch": 4.61, + "learning_rate": 1.1609275094368694e-05, + "loss": 0.0, + "step": 19934 + }, + { + "epoch": 4.61, + "learning_rate": 1.1607349202680843e-05, + "loss": 0.0, + "step": 19935 + }, + { + "epoch": 4.61, + "learning_rate": 1.160542331099299e-05, + "loss": 0.0, + "step": 19936 + }, + { + "epoch": 4.61, + "learning_rate": 1.1603497419305138e-05, + "loss": 0.0, + "step": 19937 + }, + { + "epoch": 4.61, + "learning_rate": 1.1601571527617287e-05, + "loss": 0.0, + "step": 19938 + }, + { + "epoch": 4.61, + "learning_rate": 1.1599645635929436e-05, + "loss": 0.0, + "step": 19939 + }, + { + "epoch": 4.61, + "learning_rate": 1.1597719744241585e-05, + "loss": 0.0, + "step": 19940 + }, + { + "epoch": 4.61, + "learning_rate": 1.1595793852553733e-05, + "loss": 0.0, + "step": 19941 + }, + { + "epoch": 4.61, + "learning_rate": 1.1593867960865882e-05, + "loss": 0.0, + "step": 19942 + }, + { + "epoch": 4.61, + "learning_rate": 1.159194206917803e-05, + "loss": 0.0, + "step": 19943 + }, + { + "epoch": 4.61, + "learning_rate": 1.1590016177490179e-05, + "loss": 0.0, + "step": 19944 + }, + { + "epoch": 4.61, + "learning_rate": 1.1588090285802328e-05, + "loss": 0.0, + "step": 19945 + }, + { + "epoch": 4.61, + "learning_rate": 1.1586164394114475e-05, + "loss": 0.0, + "step": 19946 + }, + { + "epoch": 4.61, + "learning_rate": 1.1584238502426624e-05, + "loss": 0.0, + "step": 19947 + }, + { + "epoch": 4.61, + "learning_rate": 1.1582312610738774e-05, + "loss": 0.0, + "step": 19948 + }, + { + "epoch": 4.61, + "learning_rate": 1.1580386719050921e-05, + "loss": 0.0001, + "step": 19949 + }, + { + "epoch": 4.61, + "learning_rate": 1.1578460827363068e-05, + "loss": 0.0, + "step": 19950 + }, + { + "epoch": 4.61, + "learning_rate": 1.1576534935675218e-05, + "loss": 0.0, + "step": 19951 + }, + { + "epoch": 4.61, + "learning_rate": 1.1574609043987367e-05, + "loss": 0.0, + "step": 19952 + }, + { + "epoch": 4.61, + "learning_rate": 1.1572683152299516e-05, + "loss": 0.0001, + "step": 19953 + }, + { + "epoch": 4.61, + "learning_rate": 1.1570757260611665e-05, + "loss": 0.0, + "step": 19954 + }, + { + "epoch": 4.61, + "learning_rate": 1.1568831368923811e-05, + "loss": 0.0, + "step": 19955 + }, + { + "epoch": 4.61, + "learning_rate": 1.156690547723596e-05, + "loss": 0.0, + "step": 19956 + }, + { + "epoch": 4.61, + "learning_rate": 1.156497958554811e-05, + "loss": 0.0, + "step": 19957 + }, + { + "epoch": 4.61, + "learning_rate": 1.1563053693860258e-05, + "loss": 0.0, + "step": 19958 + }, + { + "epoch": 4.61, + "learning_rate": 1.1561127802172408e-05, + "loss": 0.0, + "step": 19959 + }, + { + "epoch": 4.61, + "learning_rate": 1.1559201910484555e-05, + "loss": 0.0, + "step": 19960 + }, + { + "epoch": 4.61, + "learning_rate": 1.1557276018796703e-05, + "loss": 0.0, + "step": 19961 + }, + { + "epoch": 4.61, + "learning_rate": 1.1555350127108852e-05, + "loss": 0.0001, + "step": 19962 + }, + { + "epoch": 4.61, + "learning_rate": 1.1553424235421001e-05, + "loss": 0.0, + "step": 19963 + }, + { + "epoch": 4.61, + "learning_rate": 1.1551498343733148e-05, + "loss": 0.0, + "step": 19964 + }, + { + "epoch": 4.61, + "learning_rate": 1.1549572452045298e-05, + "loss": 0.0, + "step": 19965 + }, + { + "epoch": 4.61, + "learning_rate": 1.1547646560357447e-05, + "loss": 0.0001, + "step": 19966 + }, + { + "epoch": 4.61, + "learning_rate": 1.1545720668669594e-05, + "loss": 0.0, + "step": 19967 + }, + { + "epoch": 4.61, + "learning_rate": 1.1543794776981743e-05, + "loss": 0.0, + "step": 19968 + }, + { + "epoch": 4.61, + "learning_rate": 1.1541868885293891e-05, + "loss": 0.0, + "step": 19969 + }, + { + "epoch": 4.62, + "learning_rate": 1.153994299360604e-05, + "loss": 0.0, + "step": 19970 + }, + { + "epoch": 4.62, + "learning_rate": 1.153801710191819e-05, + "loss": 0.0, + "step": 19971 + }, + { + "epoch": 4.62, + "learning_rate": 1.1536091210230338e-05, + "loss": 0.0, + "step": 19972 + }, + { + "epoch": 4.62, + "learning_rate": 1.1534165318542486e-05, + "loss": 0.0, + "step": 19973 + }, + { + "epoch": 4.62, + "learning_rate": 1.1532239426854633e-05, + "loss": 0.0, + "step": 19974 + }, + { + "epoch": 4.62, + "learning_rate": 1.1530313535166783e-05, + "loss": 0.0, + "step": 19975 + }, + { + "epoch": 4.62, + "learning_rate": 1.1528387643478932e-05, + "loss": 0.0, + "step": 19976 + }, + { + "epoch": 4.62, + "learning_rate": 1.1526461751791081e-05, + "loss": 0.0, + "step": 19977 + }, + { + "epoch": 4.62, + "learning_rate": 1.1524535860103228e-05, + "loss": 0.0, + "step": 19978 + }, + { + "epoch": 4.62, + "learning_rate": 1.1522609968415376e-05, + "loss": 0.0, + "step": 19979 + }, + { + "epoch": 4.62, + "learning_rate": 1.1520684076727525e-05, + "loss": 0.0, + "step": 19980 + }, + { + "epoch": 4.62, + "learning_rate": 1.1518758185039674e-05, + "loss": 0.0, + "step": 19981 + }, + { + "epoch": 4.62, + "learning_rate": 1.1516832293351823e-05, + "loss": 0.0, + "step": 19982 + }, + { + "epoch": 4.62, + "learning_rate": 1.151490640166397e-05, + "loss": 0.0, + "step": 19983 + }, + { + "epoch": 4.62, + "learning_rate": 1.151298050997612e-05, + "loss": 0.0, + "step": 19984 + }, + { + "epoch": 4.62, + "learning_rate": 1.1511054618288267e-05, + "loss": 0.0, + "step": 19985 + }, + { + "epoch": 4.62, + "learning_rate": 1.1509128726600417e-05, + "loss": 0.0, + "step": 19986 + }, + { + "epoch": 4.62, + "learning_rate": 1.1507202834912566e-05, + "loss": 0.0, + "step": 19987 + }, + { + "epoch": 4.62, + "learning_rate": 1.1505276943224713e-05, + "loss": 0.0, + "step": 19988 + }, + { + "epoch": 4.62, + "learning_rate": 1.1503351051536862e-05, + "loss": 0.0, + "step": 19989 + }, + { + "epoch": 4.62, + "learning_rate": 1.1501425159849012e-05, + "loss": 0.0, + "step": 19990 + }, + { + "epoch": 4.62, + "learning_rate": 1.1499499268161159e-05, + "loss": 0.0, + "step": 19991 + }, + { + "epoch": 4.62, + "learning_rate": 1.1497573376473307e-05, + "loss": 0.0, + "step": 19992 + }, + { + "epoch": 4.62, + "learning_rate": 1.1495647484785456e-05, + "loss": 0.0, + "step": 19993 + }, + { + "epoch": 4.62, + "learning_rate": 1.1493721593097605e-05, + "loss": 0.0001, + "step": 19994 + }, + { + "epoch": 4.62, + "learning_rate": 1.1491795701409754e-05, + "loss": 0.0001, + "step": 19995 + }, + { + "epoch": 4.62, + "learning_rate": 1.1489869809721902e-05, + "loss": 0.0, + "step": 19996 + }, + { + "epoch": 4.62, + "learning_rate": 1.1487943918034049e-05, + "loss": 0.0, + "step": 19997 + }, + { + "epoch": 4.62, + "learning_rate": 1.1486018026346198e-05, + "loss": 0.0, + "step": 19998 + }, + { + "epoch": 4.62, + "learning_rate": 1.1484092134658347e-05, + "loss": 0.0, + "step": 19999 + }, + { + "epoch": 4.62, + "learning_rate": 1.1482166242970497e-05, + "loss": 0.0, + "step": 20000 + }, + { + "epoch": 4.62, + "learning_rate": 1.1480240351282646e-05, + "loss": 0.0, + "step": 20001 + }, + { + "epoch": 4.62, + "learning_rate": 1.1478314459594793e-05, + "loss": 0.0, + "step": 20002 + }, + { + "epoch": 4.62, + "learning_rate": 1.147638856790694e-05, + "loss": 0.0, + "step": 20003 + }, + { + "epoch": 4.62, + "learning_rate": 1.147446267621909e-05, + "loss": 0.0, + "step": 20004 + }, + { + "epoch": 4.62, + "learning_rate": 1.1472536784531239e-05, + "loss": 0.0, + "step": 20005 + }, + { + "epoch": 4.62, + "learning_rate": 1.1470610892843387e-05, + "loss": 0.0, + "step": 20006 + }, + { + "epoch": 4.62, + "learning_rate": 1.1468685001155536e-05, + "loss": 0.0, + "step": 20007 + }, + { + "epoch": 4.62, + "learning_rate": 1.1466759109467685e-05, + "loss": 0.0, + "step": 20008 + }, + { + "epoch": 4.62, + "learning_rate": 1.1464833217779832e-05, + "loss": 0.0, + "step": 20009 + }, + { + "epoch": 4.62, + "learning_rate": 1.1462907326091982e-05, + "loss": 0.0, + "step": 20010 + }, + { + "epoch": 4.62, + "learning_rate": 1.1460981434404129e-05, + "loss": 0.0004, + "step": 20011 + }, + { + "epoch": 4.62, + "learning_rate": 1.1459055542716278e-05, + "loss": 0.0, + "step": 20012 + }, + { + "epoch": 4.63, + "learning_rate": 1.1457129651028427e-05, + "loss": 0.0, + "step": 20013 + }, + { + "epoch": 4.63, + "learning_rate": 1.1455203759340575e-05, + "loss": 0.0001, + "step": 20014 + }, + { + "epoch": 4.63, + "learning_rate": 1.1453277867652724e-05, + "loss": 0.0, + "step": 20015 + }, + { + "epoch": 4.63, + "learning_rate": 1.1451351975964871e-05, + "loss": 0.0, + "step": 20016 + }, + { + "epoch": 4.63, + "learning_rate": 1.144942608427702e-05, + "loss": 0.0, + "step": 20017 + }, + { + "epoch": 4.63, + "learning_rate": 1.144750019258917e-05, + "loss": 0.0, + "step": 20018 + }, + { + "epoch": 4.63, + "learning_rate": 1.1445574300901319e-05, + "loss": 0.0, + "step": 20019 + }, + { + "epoch": 4.63, + "learning_rate": 1.1443648409213466e-05, + "loss": 0.0, + "step": 20020 + }, + { + "epoch": 4.63, + "learning_rate": 1.1441722517525614e-05, + "loss": 0.0, + "step": 20021 + }, + { + "epoch": 4.63, + "learning_rate": 1.1439796625837763e-05, + "loss": 0.0, + "step": 20022 + }, + { + "epoch": 4.63, + "learning_rate": 1.1437870734149912e-05, + "loss": 0.0, + "step": 20023 + }, + { + "epoch": 4.63, + "learning_rate": 1.1435944842462061e-05, + "loss": 0.0191, + "step": 20024 + }, + { + "epoch": 4.63, + "learning_rate": 1.1434018950774209e-05, + "loss": 0.0, + "step": 20025 + }, + { + "epoch": 4.63, + "learning_rate": 1.1432093059086358e-05, + "loss": 0.0, + "step": 20026 + }, + { + "epoch": 4.63, + "learning_rate": 1.1430167167398506e-05, + "loss": 0.0, + "step": 20027 + }, + { + "epoch": 4.63, + "learning_rate": 1.1428241275710655e-05, + "loss": 0.0, + "step": 20028 + }, + { + "epoch": 4.63, + "learning_rate": 1.1426315384022804e-05, + "loss": 0.0, + "step": 20029 + }, + { + "epoch": 4.63, + "learning_rate": 1.1424389492334951e-05, + "loss": 0.0, + "step": 20030 + }, + { + "epoch": 4.63, + "learning_rate": 1.14224636006471e-05, + "loss": 0.0, + "step": 20031 + }, + { + "epoch": 4.63, + "learning_rate": 1.1420537708959248e-05, + "loss": 0.0, + "step": 20032 + }, + { + "epoch": 4.63, + "learning_rate": 1.1418611817271397e-05, + "loss": 0.0, + "step": 20033 + }, + { + "epoch": 4.63, + "learning_rate": 1.1416685925583545e-05, + "loss": 0.0, + "step": 20034 + }, + { + "epoch": 4.63, + "learning_rate": 1.1414760033895694e-05, + "loss": 0.0, + "step": 20035 + }, + { + "epoch": 4.63, + "learning_rate": 1.1412834142207843e-05, + "loss": 0.0, + "step": 20036 + }, + { + "epoch": 4.63, + "learning_rate": 1.1410908250519992e-05, + "loss": 0.0, + "step": 20037 + }, + { + "epoch": 4.63, + "learning_rate": 1.140898235883214e-05, + "loss": 0.0001, + "step": 20038 + }, + { + "epoch": 4.63, + "learning_rate": 1.1407056467144287e-05, + "loss": 0.0, + "step": 20039 + }, + { + "epoch": 4.63, + "learning_rate": 1.1405130575456436e-05, + "loss": 0.0, + "step": 20040 + }, + { + "epoch": 4.63, + "learning_rate": 1.1403204683768586e-05, + "loss": 0.0, + "step": 20041 + }, + { + "epoch": 4.63, + "learning_rate": 1.1401278792080735e-05, + "loss": 0.0, + "step": 20042 + }, + { + "epoch": 4.63, + "learning_rate": 1.1399352900392884e-05, + "loss": 0.0, + "step": 20043 + }, + { + "epoch": 4.63, + "learning_rate": 1.1397427008705031e-05, + "loss": 0.0, + "step": 20044 + }, + { + "epoch": 4.63, + "learning_rate": 1.1395501117017179e-05, + "loss": 0.0, + "step": 20045 + }, + { + "epoch": 4.63, + "learning_rate": 1.1393575225329328e-05, + "loss": 0.0, + "step": 20046 + }, + { + "epoch": 4.63, + "learning_rate": 1.1391649333641477e-05, + "loss": 0.0, + "step": 20047 + }, + { + "epoch": 4.63, + "learning_rate": 1.1389723441953625e-05, + "loss": 0.0, + "step": 20048 + }, + { + "epoch": 4.63, + "learning_rate": 1.1387797550265774e-05, + "loss": 0.0, + "step": 20049 + }, + { + "epoch": 4.63, + "learning_rate": 1.1385871658577921e-05, + "loss": 0.0001, + "step": 20050 + }, + { + "epoch": 4.63, + "learning_rate": 1.138394576689007e-05, + "loss": 0.0, + "step": 20051 + }, + { + "epoch": 4.63, + "learning_rate": 1.138201987520222e-05, + "loss": 0.0, + "step": 20052 + }, + { + "epoch": 4.63, + "learning_rate": 1.1380093983514367e-05, + "loss": 0.0, + "step": 20053 + }, + { + "epoch": 4.63, + "learning_rate": 1.1378168091826516e-05, + "loss": 0.0, + "step": 20054 + }, + { + "epoch": 4.63, + "learning_rate": 1.1376242200138665e-05, + "loss": 0.0, + "step": 20055 + }, + { + "epoch": 4.64, + "learning_rate": 1.1374316308450813e-05, + "loss": 0.0, + "step": 20056 + }, + { + "epoch": 4.64, + "learning_rate": 1.1372390416762962e-05, + "loss": 0.0, + "step": 20057 + }, + { + "epoch": 4.64, + "learning_rate": 1.137046452507511e-05, + "loss": 0.0, + "step": 20058 + }, + { + "epoch": 4.64, + "learning_rate": 1.1368538633387259e-05, + "loss": 0.0, + "step": 20059 + }, + { + "epoch": 4.64, + "learning_rate": 1.1366612741699408e-05, + "loss": 0.0, + "step": 20060 + }, + { + "epoch": 4.64, + "learning_rate": 1.1364686850011557e-05, + "loss": 0.0, + "step": 20061 + }, + { + "epoch": 4.64, + "learning_rate": 1.1362760958323705e-05, + "loss": 0.0, + "step": 20062 + }, + { + "epoch": 4.64, + "learning_rate": 1.1360835066635852e-05, + "loss": 0.0, + "step": 20063 + }, + { + "epoch": 4.64, + "learning_rate": 1.1358909174948001e-05, + "loss": 0.0, + "step": 20064 + }, + { + "epoch": 4.64, + "learning_rate": 1.135698328326015e-05, + "loss": 0.0, + "step": 20065 + }, + { + "epoch": 4.64, + "learning_rate": 1.13550573915723e-05, + "loss": 0.0001, + "step": 20066 + }, + { + "epoch": 4.64, + "learning_rate": 1.1353131499884447e-05, + "loss": 0.0, + "step": 20067 + }, + { + "epoch": 4.64, + "learning_rate": 1.1351205608196595e-05, + "loss": 0.0, + "step": 20068 + }, + { + "epoch": 4.64, + "learning_rate": 1.1349279716508744e-05, + "loss": 0.0, + "step": 20069 + }, + { + "epoch": 4.64, + "learning_rate": 1.1347353824820893e-05, + "loss": 0.0, + "step": 20070 + }, + { + "epoch": 4.64, + "learning_rate": 1.1345427933133042e-05, + "loss": 0.0, + "step": 20071 + }, + { + "epoch": 4.64, + "learning_rate": 1.134350204144519e-05, + "loss": 0.0, + "step": 20072 + }, + { + "epoch": 4.64, + "learning_rate": 1.1341576149757339e-05, + "loss": 0.0, + "step": 20073 + }, + { + "epoch": 4.64, + "learning_rate": 1.1339650258069486e-05, + "loss": 0.0, + "step": 20074 + }, + { + "epoch": 4.64, + "learning_rate": 1.1337724366381635e-05, + "loss": 0.0, + "step": 20075 + }, + { + "epoch": 4.64, + "learning_rate": 1.1335798474693783e-05, + "loss": 0.0, + "step": 20076 + }, + { + "epoch": 4.64, + "learning_rate": 1.1333872583005932e-05, + "loss": 0.0009, + "step": 20077 + }, + { + "epoch": 4.64, + "learning_rate": 1.1331946691318081e-05, + "loss": 0.0, + "step": 20078 + }, + { + "epoch": 4.64, + "learning_rate": 1.133002079963023e-05, + "loss": 0.0, + "step": 20079 + }, + { + "epoch": 4.64, + "learning_rate": 1.1328094907942378e-05, + "loss": 0.0, + "step": 20080 + }, + { + "epoch": 4.64, + "learning_rate": 1.1326169016254525e-05, + "loss": 0.0, + "step": 20081 + }, + { + "epoch": 4.64, + "learning_rate": 1.1324243124566674e-05, + "loss": 0.0, + "step": 20082 + }, + { + "epoch": 4.64, + "learning_rate": 1.1322317232878824e-05, + "loss": 0.0, + "step": 20083 + }, + { + "epoch": 4.64, + "learning_rate": 1.1320391341190973e-05, + "loss": 0.0, + "step": 20084 + }, + { + "epoch": 4.64, + "learning_rate": 1.1318465449503122e-05, + "loss": 0.0, + "step": 20085 + }, + { + "epoch": 4.64, + "learning_rate": 1.1316539557815268e-05, + "loss": 0.0, + "step": 20086 + }, + { + "epoch": 4.64, + "learning_rate": 1.1314613666127417e-05, + "loss": 0.0254, + "step": 20087 + }, + { + "epoch": 4.64, + "learning_rate": 1.1312687774439566e-05, + "loss": 0.0, + "step": 20088 + }, + { + "epoch": 4.64, + "learning_rate": 1.1310761882751715e-05, + "loss": 0.0, + "step": 20089 + }, + { + "epoch": 4.64, + "learning_rate": 1.1308835991063863e-05, + "loss": 0.0, + "step": 20090 + }, + { + "epoch": 4.64, + "learning_rate": 1.1306910099376012e-05, + "loss": 0.0, + "step": 20091 + }, + { + "epoch": 4.64, + "learning_rate": 1.130498420768816e-05, + "loss": 0.0, + "step": 20092 + }, + { + "epoch": 4.64, + "learning_rate": 1.1303058316000309e-05, + "loss": 0.0, + "step": 20093 + }, + { + "epoch": 4.64, + "learning_rate": 1.1301132424312458e-05, + "loss": 0.0, + "step": 20094 + }, + { + "epoch": 4.64, + "learning_rate": 1.1299206532624605e-05, + "loss": 0.0, + "step": 20095 + }, + { + "epoch": 4.64, + "learning_rate": 1.1297280640936754e-05, + "loss": 0.0, + "step": 20096 + }, + { + "epoch": 4.64, + "learning_rate": 1.1295354749248904e-05, + "loss": 0.0, + "step": 20097 + }, + { + "epoch": 4.64, + "learning_rate": 1.1293428857561051e-05, + "loss": 0.0, + "step": 20098 + }, + { + "epoch": 4.65, + "learning_rate": 1.12915029658732e-05, + "loss": 0.0, + "step": 20099 + }, + { + "epoch": 4.65, + "learning_rate": 1.1289577074185348e-05, + "loss": 0.0, + "step": 20100 + }, + { + "epoch": 4.65, + "learning_rate": 1.1287651182497497e-05, + "loss": 0.0195, + "step": 20101 + }, + { + "epoch": 4.65, + "learning_rate": 1.1285725290809646e-05, + "loss": 0.0, + "step": 20102 + }, + { + "epoch": 4.65, + "learning_rate": 1.1283799399121795e-05, + "loss": 0.0, + "step": 20103 + }, + { + "epoch": 4.65, + "learning_rate": 1.1281873507433943e-05, + "loss": 0.0, + "step": 20104 + }, + { + "epoch": 4.65, + "learning_rate": 1.127994761574609e-05, + "loss": 0.0, + "step": 20105 + }, + { + "epoch": 4.65, + "learning_rate": 1.127802172405824e-05, + "loss": 0.0, + "step": 20106 + }, + { + "epoch": 4.65, + "learning_rate": 1.1276095832370388e-05, + "loss": 0.0, + "step": 20107 + }, + { + "epoch": 4.65, + "learning_rate": 1.1274169940682538e-05, + "loss": 0.0, + "step": 20108 + }, + { + "epoch": 4.65, + "learning_rate": 1.1272244048994685e-05, + "loss": 0.0, + "step": 20109 + }, + { + "epoch": 4.65, + "learning_rate": 1.1270318157306833e-05, + "loss": 0.0, + "step": 20110 + }, + { + "epoch": 4.65, + "learning_rate": 1.1268392265618982e-05, + "loss": 0.0013, + "step": 20111 + }, + { + "epoch": 4.65, + "learning_rate": 1.1266466373931131e-05, + "loss": 0.0, + "step": 20112 + }, + { + "epoch": 4.65, + "learning_rate": 1.126454048224328e-05, + "loss": 0.0, + "step": 20113 + }, + { + "epoch": 4.65, + "learning_rate": 1.1262614590555428e-05, + "loss": 0.0, + "step": 20114 + }, + { + "epoch": 4.65, + "learning_rate": 1.1260688698867577e-05, + "loss": 0.0, + "step": 20115 + }, + { + "epoch": 4.65, + "learning_rate": 1.1258762807179724e-05, + "loss": 0.0, + "step": 20116 + }, + { + "epoch": 4.65, + "learning_rate": 1.1256836915491873e-05, + "loss": 0.0, + "step": 20117 + }, + { + "epoch": 4.65, + "learning_rate": 1.1254911023804021e-05, + "loss": 0.0, + "step": 20118 + }, + { + "epoch": 4.65, + "learning_rate": 1.125298513211617e-05, + "loss": 0.0001, + "step": 20119 + }, + { + "epoch": 4.65, + "learning_rate": 1.125105924042832e-05, + "loss": 0.0, + "step": 20120 + }, + { + "epoch": 4.65, + "learning_rate": 1.1249133348740468e-05, + "loss": 0.0, + "step": 20121 + }, + { + "epoch": 4.65, + "learning_rate": 1.1247207457052616e-05, + "loss": 0.0, + "step": 20122 + }, + { + "epoch": 4.65, + "learning_rate": 1.1245281565364763e-05, + "loss": 0.0, + "step": 20123 + }, + { + "epoch": 4.65, + "learning_rate": 1.1243355673676913e-05, + "loss": 0.0, + "step": 20124 + }, + { + "epoch": 4.65, + "learning_rate": 1.1241429781989062e-05, + "loss": 0.0, + "step": 20125 + }, + { + "epoch": 4.65, + "learning_rate": 1.1239503890301211e-05, + "loss": 0.0, + "step": 20126 + }, + { + "epoch": 4.65, + "learning_rate": 1.1237577998613358e-05, + "loss": 0.0, + "step": 20127 + }, + { + "epoch": 4.65, + "learning_rate": 1.1235652106925506e-05, + "loss": 0.0, + "step": 20128 + }, + { + "epoch": 4.65, + "learning_rate": 1.1233726215237655e-05, + "loss": 0.0, + "step": 20129 + }, + { + "epoch": 4.65, + "learning_rate": 1.1231800323549804e-05, + "loss": 0.0, + "step": 20130 + }, + { + "epoch": 4.65, + "learning_rate": 1.1229874431861953e-05, + "loss": 0.0, + "step": 20131 + }, + { + "epoch": 4.65, + "learning_rate": 1.1227948540174101e-05, + "loss": 0.0001, + "step": 20132 + }, + { + "epoch": 4.65, + "learning_rate": 1.122602264848625e-05, + "loss": 0.0, + "step": 20133 + }, + { + "epoch": 4.65, + "learning_rate": 1.1224096756798397e-05, + "loss": 0.0, + "step": 20134 + }, + { + "epoch": 4.65, + "learning_rate": 1.1222170865110547e-05, + "loss": 0.0, + "step": 20135 + }, + { + "epoch": 4.65, + "learning_rate": 1.1220244973422696e-05, + "loss": 0.0, + "step": 20136 + }, + { + "epoch": 4.65, + "learning_rate": 1.1218319081734843e-05, + "loss": 0.0, + "step": 20137 + }, + { + "epoch": 4.65, + "learning_rate": 1.1216393190046992e-05, + "loss": 0.0001, + "step": 20138 + }, + { + "epoch": 4.65, + "learning_rate": 1.1214467298359142e-05, + "loss": 0.0, + "step": 20139 + }, + { + "epoch": 4.65, + "learning_rate": 1.1212541406671289e-05, + "loss": 0.0, + "step": 20140 + }, + { + "epoch": 4.65, + "learning_rate": 1.1210615514983438e-05, + "loss": 0.0, + "step": 20141 + }, + { + "epoch": 4.65, + "learning_rate": 1.1208689623295586e-05, + "loss": 0.0, + "step": 20142 + }, + { + "epoch": 4.66, + "learning_rate": 1.1206763731607735e-05, + "loss": 0.0, + "step": 20143 + }, + { + "epoch": 4.66, + "learning_rate": 1.1204837839919884e-05, + "loss": 0.0, + "step": 20144 + }, + { + "epoch": 4.66, + "learning_rate": 1.1202911948232032e-05, + "loss": 0.0, + "step": 20145 + }, + { + "epoch": 4.66, + "learning_rate": 1.1200986056544179e-05, + "loss": 0.0, + "step": 20146 + }, + { + "epoch": 4.66, + "learning_rate": 1.1199060164856328e-05, + "loss": 0.0, + "step": 20147 + }, + { + "epoch": 4.66, + "learning_rate": 1.1197134273168477e-05, + "loss": 0.0, + "step": 20148 + }, + { + "epoch": 4.66, + "learning_rate": 1.1195208381480627e-05, + "loss": 0.0, + "step": 20149 + }, + { + "epoch": 4.66, + "learning_rate": 1.1193282489792776e-05, + "loss": 0.0, + "step": 20150 + }, + { + "epoch": 4.66, + "learning_rate": 1.1191356598104923e-05, + "loss": 0.0, + "step": 20151 + }, + { + "epoch": 4.66, + "learning_rate": 1.118943070641707e-05, + "loss": 0.0, + "step": 20152 + }, + { + "epoch": 4.66, + "learning_rate": 1.118750481472922e-05, + "loss": 0.0, + "step": 20153 + }, + { + "epoch": 4.66, + "learning_rate": 1.1185578923041369e-05, + "loss": 0.0, + "step": 20154 + }, + { + "epoch": 4.66, + "learning_rate": 1.1183653031353518e-05, + "loss": 0.0, + "step": 20155 + }, + { + "epoch": 4.66, + "learning_rate": 1.1181727139665666e-05, + "loss": 0.0, + "step": 20156 + }, + { + "epoch": 4.66, + "learning_rate": 1.1179801247977815e-05, + "loss": 0.0, + "step": 20157 + }, + { + "epoch": 4.66, + "learning_rate": 1.1177875356289962e-05, + "loss": 0.0001, + "step": 20158 + }, + { + "epoch": 4.66, + "learning_rate": 1.1175949464602112e-05, + "loss": 0.0, + "step": 20159 + }, + { + "epoch": 4.66, + "learning_rate": 1.1174023572914259e-05, + "loss": 0.0, + "step": 20160 + }, + { + "epoch": 4.66, + "learning_rate": 1.1172097681226408e-05, + "loss": 0.0, + "step": 20161 + }, + { + "epoch": 4.66, + "learning_rate": 1.1170171789538557e-05, + "loss": 0.0, + "step": 20162 + }, + { + "epoch": 4.66, + "learning_rate": 1.1168245897850705e-05, + "loss": 0.0, + "step": 20163 + }, + { + "epoch": 4.66, + "learning_rate": 1.1166320006162854e-05, + "loss": 0.0028, + "step": 20164 + }, + { + "epoch": 4.66, + "learning_rate": 1.1164394114475001e-05, + "loss": 0.0, + "step": 20165 + }, + { + "epoch": 4.66, + "learning_rate": 1.116246822278715e-05, + "loss": 0.0, + "step": 20166 + }, + { + "epoch": 4.66, + "learning_rate": 1.11605423310993e-05, + "loss": 0.0, + "step": 20167 + }, + { + "epoch": 4.66, + "learning_rate": 1.1158616439411449e-05, + "loss": 0.0, + "step": 20168 + }, + { + "epoch": 4.66, + "learning_rate": 1.1156690547723596e-05, + "loss": 0.0, + "step": 20169 + }, + { + "epoch": 4.66, + "learning_rate": 1.1154764656035744e-05, + "loss": 0.0, + "step": 20170 + }, + { + "epoch": 4.66, + "learning_rate": 1.1152838764347893e-05, + "loss": 0.0, + "step": 20171 + }, + { + "epoch": 4.66, + "learning_rate": 1.1150912872660042e-05, + "loss": 0.0, + "step": 20172 + }, + { + "epoch": 4.66, + "learning_rate": 1.1148986980972191e-05, + "loss": 0.0, + "step": 20173 + }, + { + "epoch": 4.66, + "learning_rate": 1.1147061089284339e-05, + "loss": 0.0, + "step": 20174 + }, + { + "epoch": 4.66, + "learning_rate": 1.1145135197596488e-05, + "loss": 0.0, + "step": 20175 + }, + { + "epoch": 4.66, + "learning_rate": 1.1143209305908636e-05, + "loss": 0.0, + "step": 20176 + }, + { + "epoch": 4.66, + "learning_rate": 1.1141283414220785e-05, + "loss": 0.0, + "step": 20177 + }, + { + "epoch": 4.66, + "learning_rate": 1.1139357522532934e-05, + "loss": 0.0, + "step": 20178 + }, + { + "epoch": 4.66, + "learning_rate": 1.1137431630845081e-05, + "loss": 0.0, + "step": 20179 + }, + { + "epoch": 4.66, + "learning_rate": 1.113550573915723e-05, + "loss": 0.0, + "step": 20180 + }, + { + "epoch": 4.66, + "learning_rate": 1.113357984746938e-05, + "loss": 0.0, + "step": 20181 + }, + { + "epoch": 4.66, + "learning_rate": 1.1131653955781527e-05, + "loss": 0.0, + "step": 20182 + }, + { + "epoch": 4.66, + "learning_rate": 1.1129728064093676e-05, + "loss": 0.0, + "step": 20183 + }, + { + "epoch": 4.66, + "learning_rate": 1.1127802172405824e-05, + "loss": 0.0, + "step": 20184 + }, + { + "epoch": 4.66, + "learning_rate": 1.1125876280717973e-05, + "loss": 0.0, + "step": 20185 + }, + { + "epoch": 4.67, + "learning_rate": 1.1123950389030122e-05, + "loss": 0.0, + "step": 20186 + }, + { + "epoch": 4.67, + "learning_rate": 1.112202449734227e-05, + "loss": 0.0, + "step": 20187 + }, + { + "epoch": 4.67, + "learning_rate": 1.1120098605654417e-05, + "loss": 0.0, + "step": 20188 + }, + { + "epoch": 4.67, + "learning_rate": 1.1118172713966566e-05, + "loss": 0.0, + "step": 20189 + }, + { + "epoch": 4.67, + "learning_rate": 1.1116246822278716e-05, + "loss": 0.0, + "step": 20190 + }, + { + "epoch": 4.67, + "learning_rate": 1.1114320930590865e-05, + "loss": 0.0, + "step": 20191 + }, + { + "epoch": 4.67, + "learning_rate": 1.1112395038903014e-05, + "loss": 0.0, + "step": 20192 + }, + { + "epoch": 4.67, + "learning_rate": 1.1110469147215161e-05, + "loss": 0.0, + "step": 20193 + }, + { + "epoch": 4.67, + "learning_rate": 1.1108543255527309e-05, + "loss": 0.0, + "step": 20194 + }, + { + "epoch": 4.67, + "learning_rate": 1.1106617363839458e-05, + "loss": 0.0, + "step": 20195 + }, + { + "epoch": 4.67, + "learning_rate": 1.1104691472151607e-05, + "loss": 0.0, + "step": 20196 + }, + { + "epoch": 4.67, + "learning_rate": 1.1102765580463756e-05, + "loss": 0.0, + "step": 20197 + }, + { + "epoch": 4.67, + "learning_rate": 1.1100839688775904e-05, + "loss": 0.0, + "step": 20198 + }, + { + "epoch": 4.67, + "learning_rate": 1.1098913797088053e-05, + "loss": 0.0, + "step": 20199 + }, + { + "epoch": 4.67, + "learning_rate": 1.10969879054002e-05, + "loss": 0.0, + "step": 20200 + }, + { + "epoch": 4.67, + "learning_rate": 1.109506201371235e-05, + "loss": 0.0, + "step": 20201 + }, + { + "epoch": 4.67, + "learning_rate": 1.1093136122024497e-05, + "loss": 0.0, + "step": 20202 + }, + { + "epoch": 4.67, + "learning_rate": 1.1091210230336646e-05, + "loss": 0.0, + "step": 20203 + }, + { + "epoch": 4.67, + "learning_rate": 1.1089284338648795e-05, + "loss": 0.0, + "step": 20204 + }, + { + "epoch": 4.67, + "learning_rate": 1.1087358446960943e-05, + "loss": 0.0, + "step": 20205 + }, + { + "epoch": 4.67, + "learning_rate": 1.1085432555273092e-05, + "loss": 0.0, + "step": 20206 + }, + { + "epoch": 4.67, + "learning_rate": 1.108350666358524e-05, + "loss": 0.0, + "step": 20207 + }, + { + "epoch": 4.67, + "learning_rate": 1.1081580771897389e-05, + "loss": 0.0, + "step": 20208 + }, + { + "epoch": 4.67, + "learning_rate": 1.1079654880209538e-05, + "loss": 0.0, + "step": 20209 + }, + { + "epoch": 4.67, + "learning_rate": 1.1077728988521687e-05, + "loss": 0.0, + "step": 20210 + }, + { + "epoch": 4.67, + "learning_rate": 1.1075803096833835e-05, + "loss": 0.0, + "step": 20211 + }, + { + "epoch": 4.67, + "learning_rate": 1.1073877205145982e-05, + "loss": 0.0, + "step": 20212 + }, + { + "epoch": 4.67, + "learning_rate": 1.1071951313458131e-05, + "loss": 0.0, + "step": 20213 + }, + { + "epoch": 4.67, + "learning_rate": 1.107002542177028e-05, + "loss": 0.0, + "step": 20214 + }, + { + "epoch": 4.67, + "learning_rate": 1.106809953008243e-05, + "loss": 0.0, + "step": 20215 + }, + { + "epoch": 4.67, + "learning_rate": 1.1066173638394577e-05, + "loss": 0.0, + "step": 20216 + }, + { + "epoch": 4.67, + "learning_rate": 1.1064247746706726e-05, + "loss": 0.0, + "step": 20217 + }, + { + "epoch": 4.67, + "learning_rate": 1.1062321855018874e-05, + "loss": 0.0, + "step": 20218 + }, + { + "epoch": 4.67, + "learning_rate": 1.1060395963331023e-05, + "loss": 0.0, + "step": 20219 + }, + { + "epoch": 4.67, + "learning_rate": 1.1058470071643172e-05, + "loss": 0.0, + "step": 20220 + }, + { + "epoch": 4.67, + "learning_rate": 1.105654417995532e-05, + "loss": 0.0, + "step": 20221 + }, + { + "epoch": 4.67, + "learning_rate": 1.1054618288267469e-05, + "loss": 0.0001, + "step": 20222 + }, + { + "epoch": 4.67, + "learning_rate": 1.1052692396579616e-05, + "loss": 0.0, + "step": 20223 + }, + { + "epoch": 4.67, + "learning_rate": 1.1050766504891765e-05, + "loss": 0.0, + "step": 20224 + }, + { + "epoch": 4.67, + "learning_rate": 1.1048840613203915e-05, + "loss": 0.0, + "step": 20225 + }, + { + "epoch": 4.67, + "learning_rate": 1.1046914721516062e-05, + "loss": 0.0, + "step": 20226 + }, + { + "epoch": 4.67, + "learning_rate": 1.1044988829828211e-05, + "loss": 0.0, + "step": 20227 + }, + { + "epoch": 4.67, + "learning_rate": 1.104306293814036e-05, + "loss": 0.0, + "step": 20228 + }, + { + "epoch": 4.68, + "learning_rate": 1.1041137046452508e-05, + "loss": 0.0001, + "step": 20229 + }, + { + "epoch": 4.68, + "learning_rate": 1.1039211154764655e-05, + "loss": 0.0, + "step": 20230 + }, + { + "epoch": 4.68, + "learning_rate": 1.1037285263076804e-05, + "loss": 0.0, + "step": 20231 + }, + { + "epoch": 4.68, + "learning_rate": 1.1035359371388954e-05, + "loss": 0.0, + "step": 20232 + }, + { + "epoch": 4.68, + "learning_rate": 1.1033433479701103e-05, + "loss": 0.0, + "step": 20233 + }, + { + "epoch": 4.68, + "learning_rate": 1.1031507588013252e-05, + "loss": 0.0, + "step": 20234 + }, + { + "epoch": 4.68, + "learning_rate": 1.10295816963254e-05, + "loss": 0.0, + "step": 20235 + }, + { + "epoch": 4.68, + "learning_rate": 1.1027655804637547e-05, + "loss": 0.0, + "step": 20236 + }, + { + "epoch": 4.68, + "learning_rate": 1.1025729912949696e-05, + "loss": 0.0, + "step": 20237 + }, + { + "epoch": 4.68, + "learning_rate": 1.1023804021261845e-05, + "loss": 0.0, + "step": 20238 + }, + { + "epoch": 4.68, + "learning_rate": 1.1021878129573994e-05, + "loss": 0.0, + "step": 20239 + }, + { + "epoch": 4.68, + "learning_rate": 1.1019952237886142e-05, + "loss": 0.0, + "step": 20240 + }, + { + "epoch": 4.68, + "learning_rate": 1.101802634619829e-05, + "loss": 0.0, + "step": 20241 + }, + { + "epoch": 4.68, + "learning_rate": 1.1016100454510439e-05, + "loss": 0.0, + "step": 20242 + }, + { + "epoch": 4.68, + "learning_rate": 1.1014174562822588e-05, + "loss": 0.0, + "step": 20243 + }, + { + "epoch": 4.68, + "learning_rate": 1.1012248671134735e-05, + "loss": 0.0, + "step": 20244 + }, + { + "epoch": 4.68, + "learning_rate": 1.1010322779446884e-05, + "loss": 0.0, + "step": 20245 + }, + { + "epoch": 4.68, + "learning_rate": 1.1008396887759034e-05, + "loss": 0.0, + "step": 20246 + }, + { + "epoch": 4.68, + "learning_rate": 1.1006470996071181e-05, + "loss": 0.0, + "step": 20247 + }, + { + "epoch": 4.68, + "learning_rate": 1.100454510438333e-05, + "loss": 0.0, + "step": 20248 + }, + { + "epoch": 4.68, + "learning_rate": 1.1002619212695478e-05, + "loss": 0.0, + "step": 20249 + }, + { + "epoch": 4.68, + "learning_rate": 1.1000693321007627e-05, + "loss": 0.0, + "step": 20250 + }, + { + "epoch": 4.68, + "learning_rate": 1.0998767429319776e-05, + "loss": 0.0, + "step": 20251 + }, + { + "epoch": 4.68, + "learning_rate": 1.0996841537631925e-05, + "loss": 0.0, + "step": 20252 + }, + { + "epoch": 4.68, + "learning_rate": 1.0994915645944073e-05, + "loss": 0.0, + "step": 20253 + }, + { + "epoch": 4.68, + "learning_rate": 1.099298975425622e-05, + "loss": 0.0, + "step": 20254 + }, + { + "epoch": 4.68, + "learning_rate": 1.099106386256837e-05, + "loss": 0.0, + "step": 20255 + }, + { + "epoch": 4.68, + "learning_rate": 1.0989137970880519e-05, + "loss": 0.0, + "step": 20256 + }, + { + "epoch": 4.68, + "learning_rate": 1.0987212079192668e-05, + "loss": 0.0, + "step": 20257 + }, + { + "epoch": 4.68, + "learning_rate": 1.0985286187504815e-05, + "loss": 0.0, + "step": 20258 + }, + { + "epoch": 4.68, + "learning_rate": 1.0983360295816963e-05, + "loss": 0.0, + "step": 20259 + }, + { + "epoch": 4.68, + "learning_rate": 1.0981434404129112e-05, + "loss": 0.0, + "step": 20260 + }, + { + "epoch": 4.68, + "learning_rate": 1.0979508512441261e-05, + "loss": 0.0, + "step": 20261 + }, + { + "epoch": 4.68, + "learning_rate": 1.097758262075341e-05, + "loss": 0.0015, + "step": 20262 + }, + { + "epoch": 4.68, + "learning_rate": 1.0975656729065558e-05, + "loss": 0.0, + "step": 20263 + }, + { + "epoch": 4.68, + "learning_rate": 1.0973730837377707e-05, + "loss": 0.0, + "step": 20264 + }, + { + "epoch": 4.68, + "learning_rate": 1.0971804945689854e-05, + "loss": 0.0, + "step": 20265 + }, + { + "epoch": 4.68, + "learning_rate": 1.0969879054002003e-05, + "loss": 0.0003, + "step": 20266 + }, + { + "epoch": 4.68, + "learning_rate": 1.0967953162314153e-05, + "loss": 0.0, + "step": 20267 + }, + { + "epoch": 4.68, + "learning_rate": 1.09660272706263e-05, + "loss": 0.0, + "step": 20268 + }, + { + "epoch": 4.68, + "learning_rate": 1.096410137893845e-05, + "loss": 0.0, + "step": 20269 + }, + { + "epoch": 4.68, + "learning_rate": 1.0962175487250598e-05, + "loss": 0.0, + "step": 20270 + }, + { + "epoch": 4.68, + "learning_rate": 1.0960249595562746e-05, + "loss": 0.0, + "step": 20271 + }, + { + "epoch": 4.69, + "learning_rate": 1.0958323703874893e-05, + "loss": 0.0, + "step": 20272 + }, + { + "epoch": 4.69, + "learning_rate": 1.0956397812187043e-05, + "loss": 0.0, + "step": 20273 + }, + { + "epoch": 4.69, + "learning_rate": 1.0954471920499192e-05, + "loss": 0.0, + "step": 20274 + }, + { + "epoch": 4.69, + "learning_rate": 1.0952546028811341e-05, + "loss": 0.0, + "step": 20275 + }, + { + "epoch": 4.69, + "learning_rate": 1.095062013712349e-05, + "loss": 0.0, + "step": 20276 + }, + { + "epoch": 4.69, + "learning_rate": 1.0948694245435636e-05, + "loss": 0.0, + "step": 20277 + }, + { + "epoch": 4.69, + "learning_rate": 1.0946768353747785e-05, + "loss": 0.0, + "step": 20278 + }, + { + "epoch": 4.69, + "learning_rate": 1.0944842462059934e-05, + "loss": 0.0, + "step": 20279 + }, + { + "epoch": 4.69, + "learning_rate": 1.0942916570372083e-05, + "loss": 0.0, + "step": 20280 + }, + { + "epoch": 4.69, + "learning_rate": 1.0940990678684233e-05, + "loss": 0.0, + "step": 20281 + }, + { + "epoch": 4.69, + "learning_rate": 1.093906478699638e-05, + "loss": 0.0, + "step": 20282 + }, + { + "epoch": 4.69, + "learning_rate": 1.0937138895308528e-05, + "loss": 0.0, + "step": 20283 + }, + { + "epoch": 4.69, + "learning_rate": 1.0935213003620677e-05, + "loss": 0.0, + "step": 20284 + }, + { + "epoch": 4.69, + "learning_rate": 1.0933287111932826e-05, + "loss": 0.0, + "step": 20285 + }, + { + "epoch": 4.69, + "learning_rate": 1.0931361220244973e-05, + "loss": 0.0, + "step": 20286 + }, + { + "epoch": 4.69, + "learning_rate": 1.0929435328557123e-05, + "loss": 0.0, + "step": 20287 + }, + { + "epoch": 4.69, + "learning_rate": 1.0927509436869272e-05, + "loss": 0.0, + "step": 20288 + }, + { + "epoch": 4.69, + "learning_rate": 1.092558354518142e-05, + "loss": 0.0, + "step": 20289 + }, + { + "epoch": 4.69, + "learning_rate": 1.0923657653493568e-05, + "loss": 0.0, + "step": 20290 + }, + { + "epoch": 4.69, + "learning_rate": 1.0921731761805716e-05, + "loss": 0.0, + "step": 20291 + }, + { + "epoch": 4.69, + "learning_rate": 1.0919805870117865e-05, + "loss": 0.0, + "step": 20292 + }, + { + "epoch": 4.69, + "learning_rate": 1.0917879978430014e-05, + "loss": 0.0, + "step": 20293 + }, + { + "epoch": 4.69, + "learning_rate": 1.0915954086742163e-05, + "loss": 0.0, + "step": 20294 + }, + { + "epoch": 4.69, + "learning_rate": 1.091402819505431e-05, + "loss": 0.0, + "step": 20295 + }, + { + "epoch": 4.69, + "learning_rate": 1.0912102303366458e-05, + "loss": 0.0, + "step": 20296 + }, + { + "epoch": 4.69, + "learning_rate": 1.0910176411678607e-05, + "loss": 0.0, + "step": 20297 + }, + { + "epoch": 4.69, + "learning_rate": 1.0908250519990757e-05, + "loss": 0.0, + "step": 20298 + }, + { + "epoch": 4.69, + "learning_rate": 1.0906324628302906e-05, + "loss": 0.0, + "step": 20299 + }, + { + "epoch": 4.69, + "learning_rate": 1.0904398736615053e-05, + "loss": 0.0, + "step": 20300 + }, + { + "epoch": 4.69, + "learning_rate": 1.09024728449272e-05, + "loss": 0.0, + "step": 20301 + }, + { + "epoch": 4.69, + "learning_rate": 1.090054695323935e-05, + "loss": 0.0, + "step": 20302 + }, + { + "epoch": 4.69, + "learning_rate": 1.0898621061551499e-05, + "loss": 0.0, + "step": 20303 + }, + { + "epoch": 4.69, + "learning_rate": 1.0896695169863648e-05, + "loss": 0.0, + "step": 20304 + }, + { + "epoch": 4.69, + "learning_rate": 1.0894769278175796e-05, + "loss": 0.0, + "step": 20305 + }, + { + "epoch": 4.69, + "learning_rate": 1.0892843386487945e-05, + "loss": 0.0001, + "step": 20306 + }, + { + "epoch": 4.69, + "learning_rate": 1.0890917494800092e-05, + "loss": 0.0, + "step": 20307 + }, + { + "epoch": 4.69, + "learning_rate": 1.0888991603112242e-05, + "loss": 0.0, + "step": 20308 + }, + { + "epoch": 4.69, + "learning_rate": 1.088706571142439e-05, + "loss": 0.0, + "step": 20309 + }, + { + "epoch": 4.69, + "learning_rate": 1.0885139819736538e-05, + "loss": 0.0, + "step": 20310 + }, + { + "epoch": 4.69, + "learning_rate": 1.0883213928048687e-05, + "loss": 0.0, + "step": 20311 + }, + { + "epoch": 4.69, + "learning_rate": 1.0881288036360837e-05, + "loss": 0.0, + "step": 20312 + }, + { + "epoch": 4.69, + "learning_rate": 1.0879362144672984e-05, + "loss": 0.0, + "step": 20313 + }, + { + "epoch": 4.69, + "learning_rate": 1.0877436252985132e-05, + "loss": 0.0, + "step": 20314 + }, + { + "epoch": 4.69, + "learning_rate": 1.087551036129728e-05, + "loss": 0.0, + "step": 20315 + }, + { + "epoch": 4.7, + "learning_rate": 1.087358446960943e-05, + "loss": 0.0, + "step": 20316 + }, + { + "epoch": 4.7, + "learning_rate": 1.0871658577921579e-05, + "loss": 0.0, + "step": 20317 + }, + { + "epoch": 4.7, + "learning_rate": 1.0869732686233727e-05, + "loss": 0.0, + "step": 20318 + }, + { + "epoch": 4.7, + "learning_rate": 1.0867806794545874e-05, + "loss": 0.0, + "step": 20319 + }, + { + "epoch": 4.7, + "learning_rate": 1.0865880902858023e-05, + "loss": 0.0, + "step": 20320 + }, + { + "epoch": 4.7, + "learning_rate": 1.0863955011170172e-05, + "loss": 0.0, + "step": 20321 + }, + { + "epoch": 4.7, + "learning_rate": 1.0862029119482322e-05, + "loss": 0.0, + "step": 20322 + }, + { + "epoch": 4.7, + "learning_rate": 1.086010322779447e-05, + "loss": 0.0, + "step": 20323 + }, + { + "epoch": 4.7, + "learning_rate": 1.0858177336106618e-05, + "loss": 0.0, + "step": 20324 + }, + { + "epoch": 4.7, + "learning_rate": 1.0856251444418766e-05, + "loss": 0.0, + "step": 20325 + }, + { + "epoch": 4.7, + "learning_rate": 1.0854325552730915e-05, + "loss": 0.0, + "step": 20326 + }, + { + "epoch": 4.7, + "learning_rate": 1.0852399661043064e-05, + "loss": 0.0, + "step": 20327 + }, + { + "epoch": 4.7, + "learning_rate": 1.0850473769355211e-05, + "loss": 0.0, + "step": 20328 + }, + { + "epoch": 4.7, + "learning_rate": 1.084854787766736e-05, + "loss": 0.0, + "step": 20329 + }, + { + "epoch": 4.7, + "learning_rate": 1.084662198597951e-05, + "loss": 0.0, + "step": 20330 + }, + { + "epoch": 4.7, + "learning_rate": 1.0844696094291657e-05, + "loss": 0.0, + "step": 20331 + }, + { + "epoch": 4.7, + "learning_rate": 1.0842770202603806e-05, + "loss": 0.0, + "step": 20332 + }, + { + "epoch": 4.7, + "learning_rate": 1.0840844310915954e-05, + "loss": 0.0, + "step": 20333 + }, + { + "epoch": 4.7, + "learning_rate": 1.0838918419228103e-05, + "loss": 0.0, + "step": 20334 + }, + { + "epoch": 4.7, + "learning_rate": 1.0836992527540252e-05, + "loss": 0.0, + "step": 20335 + }, + { + "epoch": 4.7, + "learning_rate": 1.08350666358524e-05, + "loss": 0.0, + "step": 20336 + }, + { + "epoch": 4.7, + "learning_rate": 1.0833140744164549e-05, + "loss": 0.0, + "step": 20337 + }, + { + "epoch": 4.7, + "learning_rate": 1.0831214852476696e-05, + "loss": 0.0, + "step": 20338 + }, + { + "epoch": 4.7, + "learning_rate": 1.0829288960788846e-05, + "loss": 0.0, + "step": 20339 + }, + { + "epoch": 4.7, + "learning_rate": 1.0827363069100995e-05, + "loss": 0.0, + "step": 20340 + }, + { + "epoch": 4.7, + "learning_rate": 1.0825437177413144e-05, + "loss": 0.0, + "step": 20341 + }, + { + "epoch": 4.7, + "learning_rate": 1.0823511285725291e-05, + "loss": 0.0, + "step": 20342 + }, + { + "epoch": 4.7, + "learning_rate": 1.0821585394037439e-05, + "loss": 0.0, + "step": 20343 + }, + { + "epoch": 4.7, + "learning_rate": 1.0819659502349588e-05, + "loss": 0.0, + "step": 20344 + }, + { + "epoch": 4.7, + "learning_rate": 1.0817733610661737e-05, + "loss": 0.0, + "step": 20345 + }, + { + "epoch": 4.7, + "learning_rate": 1.0815807718973886e-05, + "loss": 0.0, + "step": 20346 + }, + { + "epoch": 4.7, + "learning_rate": 1.0813881827286034e-05, + "loss": 0.0, + "step": 20347 + }, + { + "epoch": 4.7, + "learning_rate": 1.0811955935598183e-05, + "loss": 0.0, + "step": 20348 + }, + { + "epoch": 4.7, + "learning_rate": 1.081003004391033e-05, + "loss": 0.0, + "step": 20349 + }, + { + "epoch": 4.7, + "learning_rate": 1.080810415222248e-05, + "loss": 0.0001, + "step": 20350 + }, + { + "epoch": 4.7, + "learning_rate": 1.0806178260534629e-05, + "loss": 0.0, + "step": 20351 + }, + { + "epoch": 4.7, + "learning_rate": 1.0804252368846776e-05, + "loss": 0.0, + "step": 20352 + }, + { + "epoch": 4.7, + "learning_rate": 1.0802326477158926e-05, + "loss": 0.0, + "step": 20353 + }, + { + "epoch": 4.7, + "learning_rate": 1.0800400585471073e-05, + "loss": 0.0, + "step": 20354 + }, + { + "epoch": 4.7, + "learning_rate": 1.0798474693783222e-05, + "loss": 0.0, + "step": 20355 + }, + { + "epoch": 4.7, + "learning_rate": 1.079654880209537e-05, + "loss": 0.0, + "step": 20356 + }, + { + "epoch": 4.7, + "learning_rate": 1.0794622910407519e-05, + "loss": 0.0, + "step": 20357 + }, + { + "epoch": 4.7, + "learning_rate": 1.0792697018719668e-05, + "loss": 0.0, + "step": 20358 + }, + { + "epoch": 4.71, + "learning_rate": 1.0790771127031817e-05, + "loss": 0.0, + "step": 20359 + }, + { + "epoch": 4.71, + "learning_rate": 1.0788845235343965e-05, + "loss": 0.0, + "step": 20360 + }, + { + "epoch": 4.71, + "learning_rate": 1.0786919343656112e-05, + "loss": 0.0, + "step": 20361 + }, + { + "epoch": 4.71, + "learning_rate": 1.0784993451968261e-05, + "loss": 0.0, + "step": 20362 + }, + { + "epoch": 4.71, + "learning_rate": 1.078306756028041e-05, + "loss": 0.0001, + "step": 20363 + }, + { + "epoch": 4.71, + "learning_rate": 1.078114166859256e-05, + "loss": 0.0, + "step": 20364 + }, + { + "epoch": 4.71, + "learning_rate": 1.0779215776904709e-05, + "loss": 0.0, + "step": 20365 + }, + { + "epoch": 4.71, + "learning_rate": 1.0777289885216856e-05, + "loss": 0.0, + "step": 20366 + }, + { + "epoch": 4.71, + "learning_rate": 1.0775363993529004e-05, + "loss": 0.0, + "step": 20367 + }, + { + "epoch": 4.71, + "learning_rate": 1.0773438101841153e-05, + "loss": 0.0, + "step": 20368 + }, + { + "epoch": 4.71, + "learning_rate": 1.0771512210153302e-05, + "loss": 0.0, + "step": 20369 + }, + { + "epoch": 4.71, + "learning_rate": 1.076958631846545e-05, + "loss": 0.0, + "step": 20370 + }, + { + "epoch": 4.71, + "learning_rate": 1.0767660426777599e-05, + "loss": 0.0, + "step": 20371 + }, + { + "epoch": 4.71, + "learning_rate": 1.0765734535089746e-05, + "loss": 0.0, + "step": 20372 + }, + { + "epoch": 4.71, + "learning_rate": 1.0763808643401895e-05, + "loss": 0.0004, + "step": 20373 + }, + { + "epoch": 4.71, + "learning_rate": 1.0761882751714045e-05, + "loss": 0.0, + "step": 20374 + }, + { + "epoch": 4.71, + "learning_rate": 1.0759956860026192e-05, + "loss": 0.0, + "step": 20375 + }, + { + "epoch": 4.71, + "learning_rate": 1.0758030968338341e-05, + "loss": 0.0, + "step": 20376 + }, + { + "epoch": 4.71, + "learning_rate": 1.075610507665049e-05, + "loss": 0.0, + "step": 20377 + }, + { + "epoch": 4.71, + "learning_rate": 1.0754179184962638e-05, + "loss": 0.0, + "step": 20378 + }, + { + "epoch": 4.71, + "learning_rate": 1.0752253293274787e-05, + "loss": 0.0, + "step": 20379 + }, + { + "epoch": 4.71, + "learning_rate": 1.0750327401586935e-05, + "loss": 0.0, + "step": 20380 + }, + { + "epoch": 4.71, + "learning_rate": 1.0748401509899084e-05, + "loss": 0.0, + "step": 20381 + }, + { + "epoch": 4.71, + "learning_rate": 1.0746475618211233e-05, + "loss": 0.0, + "step": 20382 + }, + { + "epoch": 4.71, + "learning_rate": 1.0744549726523382e-05, + "loss": 0.0, + "step": 20383 + }, + { + "epoch": 4.71, + "learning_rate": 1.074262383483553e-05, + "loss": 0.0, + "step": 20384 + }, + { + "epoch": 4.71, + "learning_rate": 1.0740697943147677e-05, + "loss": 0.0001, + "step": 20385 + }, + { + "epoch": 4.71, + "learning_rate": 1.0738772051459826e-05, + "loss": 0.0, + "step": 20386 + }, + { + "epoch": 4.71, + "learning_rate": 1.0736846159771975e-05, + "loss": 0.0, + "step": 20387 + }, + { + "epoch": 4.71, + "learning_rate": 1.0734920268084125e-05, + "loss": 0.0, + "step": 20388 + }, + { + "epoch": 4.71, + "learning_rate": 1.0732994376396272e-05, + "loss": 0.0, + "step": 20389 + }, + { + "epoch": 4.71, + "learning_rate": 1.073106848470842e-05, + "loss": 0.0, + "step": 20390 + }, + { + "epoch": 4.71, + "learning_rate": 1.0729142593020569e-05, + "loss": 0.0, + "step": 20391 + }, + { + "epoch": 4.71, + "learning_rate": 1.0727216701332718e-05, + "loss": 0.0, + "step": 20392 + }, + { + "epoch": 4.71, + "learning_rate": 1.0725290809644867e-05, + "loss": 0.0, + "step": 20393 + }, + { + "epoch": 4.71, + "learning_rate": 1.0723364917957014e-05, + "loss": 0.0, + "step": 20394 + }, + { + "epoch": 4.71, + "learning_rate": 1.0721439026269164e-05, + "loss": 0.0, + "step": 20395 + }, + { + "epoch": 4.71, + "learning_rate": 1.0719513134581311e-05, + "loss": 0.0, + "step": 20396 + }, + { + "epoch": 4.71, + "learning_rate": 1.071758724289346e-05, + "loss": 0.0, + "step": 20397 + }, + { + "epoch": 4.71, + "learning_rate": 1.0715661351205608e-05, + "loss": 0.0, + "step": 20398 + }, + { + "epoch": 4.71, + "learning_rate": 1.0713735459517757e-05, + "loss": 0.0, + "step": 20399 + }, + { + "epoch": 4.71, + "learning_rate": 1.0711809567829906e-05, + "loss": 0.0, + "step": 20400 + }, + { + "epoch": 4.71, + "learning_rate": 1.0709883676142055e-05, + "loss": 0.0, + "step": 20401 + }, + { + "epoch": 4.72, + "learning_rate": 1.0707957784454203e-05, + "loss": 0.0, + "step": 20402 + }, + { + "epoch": 4.72, + "learning_rate": 1.070603189276635e-05, + "loss": 0.0, + "step": 20403 + }, + { + "epoch": 4.72, + "learning_rate": 1.07041060010785e-05, + "loss": 0.0, + "step": 20404 + }, + { + "epoch": 4.72, + "learning_rate": 1.0702180109390649e-05, + "loss": 0.0, + "step": 20405 + }, + { + "epoch": 4.72, + "learning_rate": 1.0700254217702798e-05, + "loss": 0.0, + "step": 20406 + }, + { + "epoch": 4.72, + "learning_rate": 1.0698328326014947e-05, + "loss": 0.0, + "step": 20407 + }, + { + "epoch": 4.72, + "learning_rate": 1.0696402434327093e-05, + "loss": 0.0001, + "step": 20408 + }, + { + "epoch": 4.72, + "learning_rate": 1.0694476542639242e-05, + "loss": 0.0, + "step": 20409 + }, + { + "epoch": 4.72, + "learning_rate": 1.0692550650951391e-05, + "loss": 0.0, + "step": 20410 + }, + { + "epoch": 4.72, + "learning_rate": 1.069062475926354e-05, + "loss": 0.0, + "step": 20411 + }, + { + "epoch": 4.72, + "learning_rate": 1.0688698867575688e-05, + "loss": 0.0, + "step": 20412 + }, + { + "epoch": 4.72, + "learning_rate": 1.0686772975887837e-05, + "loss": 0.0, + "step": 20413 + }, + { + "epoch": 4.72, + "learning_rate": 1.0684847084199984e-05, + "loss": 0.0, + "step": 20414 + }, + { + "epoch": 4.72, + "learning_rate": 1.0682921192512134e-05, + "loss": 0.0, + "step": 20415 + }, + { + "epoch": 4.72, + "learning_rate": 1.0680995300824283e-05, + "loss": 0.0, + "step": 20416 + }, + { + "epoch": 4.72, + "learning_rate": 1.067906940913643e-05, + "loss": 0.0001, + "step": 20417 + }, + { + "epoch": 4.72, + "learning_rate": 1.067714351744858e-05, + "loss": 0.0, + "step": 20418 + }, + { + "epoch": 4.72, + "learning_rate": 1.0675217625760729e-05, + "loss": 0.0, + "step": 20419 + }, + { + "epoch": 4.72, + "learning_rate": 1.0673291734072876e-05, + "loss": 0.0, + "step": 20420 + }, + { + "epoch": 4.72, + "learning_rate": 1.0671365842385025e-05, + "loss": 0.0001, + "step": 20421 + }, + { + "epoch": 4.72, + "learning_rate": 1.0669439950697173e-05, + "loss": 0.0, + "step": 20422 + }, + { + "epoch": 4.72, + "learning_rate": 1.0667514059009322e-05, + "loss": 0.0, + "step": 20423 + }, + { + "epoch": 4.72, + "learning_rate": 1.0665588167321471e-05, + "loss": 0.0, + "step": 20424 + }, + { + "epoch": 4.72, + "learning_rate": 1.066366227563362e-05, + "loss": 0.0, + "step": 20425 + }, + { + "epoch": 4.72, + "learning_rate": 1.0661736383945768e-05, + "loss": 0.0, + "step": 20426 + }, + { + "epoch": 4.72, + "learning_rate": 1.0659810492257915e-05, + "loss": 0.0, + "step": 20427 + }, + { + "epoch": 4.72, + "learning_rate": 1.0657884600570064e-05, + "loss": 0.0, + "step": 20428 + }, + { + "epoch": 4.72, + "learning_rate": 1.0655958708882213e-05, + "loss": 0.0, + "step": 20429 + }, + { + "epoch": 4.72, + "learning_rate": 1.0654032817194363e-05, + "loss": 0.0, + "step": 20430 + }, + { + "epoch": 4.72, + "learning_rate": 1.065210692550651e-05, + "loss": 0.0, + "step": 20431 + }, + { + "epoch": 4.72, + "learning_rate": 1.0650181033818658e-05, + "loss": 0.0, + "step": 20432 + }, + { + "epoch": 4.72, + "learning_rate": 1.0648255142130807e-05, + "loss": 0.0, + "step": 20433 + }, + { + "epoch": 4.72, + "learning_rate": 1.0646329250442956e-05, + "loss": 0.0, + "step": 20434 + }, + { + "epoch": 4.72, + "learning_rate": 1.0644403358755105e-05, + "loss": 0.0, + "step": 20435 + }, + { + "epoch": 4.72, + "learning_rate": 1.0642477467067253e-05, + "loss": 0.0, + "step": 20436 + }, + { + "epoch": 4.72, + "learning_rate": 1.0640551575379402e-05, + "loss": 0.0002, + "step": 20437 + }, + { + "epoch": 4.72, + "learning_rate": 1.063862568369155e-05, + "loss": 0.0, + "step": 20438 + }, + { + "epoch": 4.72, + "learning_rate": 1.0636699792003698e-05, + "loss": 0.0, + "step": 20439 + }, + { + "epoch": 4.72, + "learning_rate": 1.0634773900315846e-05, + "loss": 0.0, + "step": 20440 + }, + { + "epoch": 4.72, + "learning_rate": 1.0632848008627995e-05, + "loss": 0.0, + "step": 20441 + }, + { + "epoch": 4.72, + "learning_rate": 1.0630922116940144e-05, + "loss": 0.0, + "step": 20442 + }, + { + "epoch": 4.72, + "learning_rate": 1.0628996225252293e-05, + "loss": 0.0, + "step": 20443 + }, + { + "epoch": 4.72, + "learning_rate": 1.0627070333564441e-05, + "loss": 0.0, + "step": 20444 + }, + { + "epoch": 4.72, + "learning_rate": 1.0625144441876588e-05, + "loss": 0.0014, + "step": 20445 + }, + { + "epoch": 4.73, + "learning_rate": 1.0623218550188738e-05, + "loss": 0.0, + "step": 20446 + }, + { + "epoch": 4.73, + "learning_rate": 1.0621292658500887e-05, + "loss": 0.0, + "step": 20447 + }, + { + "epoch": 4.73, + "learning_rate": 1.0619366766813036e-05, + "loss": 0.0, + "step": 20448 + }, + { + "epoch": 4.73, + "learning_rate": 1.0617440875125183e-05, + "loss": 0.0, + "step": 20449 + }, + { + "epoch": 4.73, + "learning_rate": 1.061551498343733e-05, + "loss": 0.0, + "step": 20450 + }, + { + "epoch": 4.73, + "learning_rate": 1.061358909174948e-05, + "loss": 0.0, + "step": 20451 + }, + { + "epoch": 4.73, + "learning_rate": 1.061166320006163e-05, + "loss": 0.0, + "step": 20452 + }, + { + "epoch": 4.73, + "learning_rate": 1.0609737308373778e-05, + "loss": 0.0, + "step": 20453 + }, + { + "epoch": 4.73, + "learning_rate": 1.0607811416685926e-05, + "loss": 0.0, + "step": 20454 + }, + { + "epoch": 4.73, + "learning_rate": 1.0605885524998075e-05, + "loss": 0.0, + "step": 20455 + }, + { + "epoch": 4.73, + "learning_rate": 1.0603959633310222e-05, + "loss": 0.0001, + "step": 20456 + }, + { + "epoch": 4.73, + "learning_rate": 1.0602033741622372e-05, + "loss": 0.0, + "step": 20457 + }, + { + "epoch": 4.73, + "learning_rate": 1.060010784993452e-05, + "loss": 0.0, + "step": 20458 + }, + { + "epoch": 4.73, + "learning_rate": 1.0598181958246668e-05, + "loss": 0.0, + "step": 20459 + }, + { + "epoch": 4.73, + "learning_rate": 1.0596256066558817e-05, + "loss": 0.0, + "step": 20460 + }, + { + "epoch": 4.73, + "learning_rate": 1.0594330174870967e-05, + "loss": 0.0, + "step": 20461 + }, + { + "epoch": 4.73, + "learning_rate": 1.0592404283183114e-05, + "loss": 0.0, + "step": 20462 + }, + { + "epoch": 4.73, + "learning_rate": 1.0590478391495263e-05, + "loss": 0.0, + "step": 20463 + }, + { + "epoch": 4.73, + "learning_rate": 1.058855249980741e-05, + "loss": 0.0008, + "step": 20464 + }, + { + "epoch": 4.73, + "learning_rate": 1.058662660811956e-05, + "loss": 0.0, + "step": 20465 + }, + { + "epoch": 4.73, + "learning_rate": 1.0584700716431709e-05, + "loss": 0.0, + "step": 20466 + }, + { + "epoch": 4.73, + "learning_rate": 1.0582774824743857e-05, + "loss": 0.0, + "step": 20467 + }, + { + "epoch": 4.73, + "learning_rate": 1.0580848933056004e-05, + "loss": 0.0, + "step": 20468 + }, + { + "epoch": 4.73, + "learning_rate": 1.0578923041368153e-05, + "loss": 0.0, + "step": 20469 + }, + { + "epoch": 4.73, + "learning_rate": 1.0576997149680302e-05, + "loss": 0.0, + "step": 20470 + }, + { + "epoch": 4.73, + "learning_rate": 1.0575071257992452e-05, + "loss": 0.0, + "step": 20471 + }, + { + "epoch": 4.73, + "learning_rate": 1.05731453663046e-05, + "loss": 0.0, + "step": 20472 + }, + { + "epoch": 4.73, + "learning_rate": 1.0571219474616748e-05, + "loss": 0.0, + "step": 20473 + }, + { + "epoch": 4.73, + "learning_rate": 1.0569293582928896e-05, + "loss": 0.0, + "step": 20474 + }, + { + "epoch": 4.73, + "learning_rate": 1.0567367691241045e-05, + "loss": 0.0, + "step": 20475 + }, + { + "epoch": 4.73, + "learning_rate": 1.0565441799553194e-05, + "loss": 0.0, + "step": 20476 + }, + { + "epoch": 4.73, + "learning_rate": 1.0563515907865343e-05, + "loss": 0.0, + "step": 20477 + }, + { + "epoch": 4.73, + "learning_rate": 1.056159001617749e-05, + "loss": 0.0, + "step": 20478 + }, + { + "epoch": 4.73, + "learning_rate": 1.055966412448964e-05, + "loss": 0.0, + "step": 20479 + }, + { + "epoch": 4.73, + "learning_rate": 1.0557738232801787e-05, + "loss": 0.0, + "step": 20480 + }, + { + "epoch": 4.73, + "learning_rate": 1.0555812341113937e-05, + "loss": 0.0, + "step": 20481 + }, + { + "epoch": 4.73, + "learning_rate": 1.0553886449426084e-05, + "loss": 0.0, + "step": 20482 + }, + { + "epoch": 4.73, + "learning_rate": 1.0551960557738233e-05, + "loss": 0.0, + "step": 20483 + }, + { + "epoch": 4.73, + "learning_rate": 1.0550034666050382e-05, + "loss": 0.0, + "step": 20484 + }, + { + "epoch": 4.73, + "learning_rate": 1.054810877436253e-05, + "loss": 0.0, + "step": 20485 + }, + { + "epoch": 4.73, + "learning_rate": 1.0546182882674679e-05, + "loss": 0.0001, + "step": 20486 + }, + { + "epoch": 4.73, + "learning_rate": 1.0544256990986826e-05, + "loss": 0.0, + "step": 20487 + }, + { + "epoch": 4.73, + "learning_rate": 1.0542331099298976e-05, + "loss": 0.0, + "step": 20488 + }, + { + "epoch": 4.74, + "learning_rate": 1.0540405207611125e-05, + "loss": 0.0001, + "step": 20489 + }, + { + "epoch": 4.74, + "learning_rate": 1.0538479315923274e-05, + "loss": 0.0, + "step": 20490 + }, + { + "epoch": 4.74, + "learning_rate": 1.0536553424235421e-05, + "loss": 0.0, + "step": 20491 + }, + { + "epoch": 4.74, + "learning_rate": 1.0534627532547569e-05, + "loss": 0.0, + "step": 20492 + }, + { + "epoch": 4.74, + "learning_rate": 1.0532701640859718e-05, + "loss": 0.0, + "step": 20493 + }, + { + "epoch": 4.74, + "learning_rate": 1.0530775749171867e-05, + "loss": 0.0, + "step": 20494 + }, + { + "epoch": 4.74, + "learning_rate": 1.0528849857484016e-05, + "loss": 0.0, + "step": 20495 + }, + { + "epoch": 4.74, + "learning_rate": 1.0526923965796164e-05, + "loss": 0.0, + "step": 20496 + }, + { + "epoch": 4.74, + "learning_rate": 1.0524998074108313e-05, + "loss": 0.0, + "step": 20497 + }, + { + "epoch": 4.74, + "learning_rate": 1.052307218242046e-05, + "loss": 0.0001, + "step": 20498 + }, + { + "epoch": 4.74, + "learning_rate": 1.052114629073261e-05, + "loss": 0.0, + "step": 20499 + }, + { + "epoch": 4.74, + "learning_rate": 1.0519220399044759e-05, + "loss": 0.0, + "step": 20500 + }, + { + "epoch": 4.74, + "learning_rate": 1.0517294507356906e-05, + "loss": 0.0, + "step": 20501 + }, + { + "epoch": 4.74, + "learning_rate": 1.0515368615669056e-05, + "loss": 0.0, + "step": 20502 + }, + { + "epoch": 4.74, + "learning_rate": 1.0513442723981203e-05, + "loss": 0.0, + "step": 20503 + }, + { + "epoch": 4.74, + "learning_rate": 1.0511516832293352e-05, + "loss": 0.0, + "step": 20504 + }, + { + "epoch": 4.74, + "learning_rate": 1.0509590940605501e-05, + "loss": 0.0, + "step": 20505 + }, + { + "epoch": 4.74, + "learning_rate": 1.0507665048917649e-05, + "loss": 0.0, + "step": 20506 + }, + { + "epoch": 4.74, + "learning_rate": 1.0505739157229798e-05, + "loss": 0.0, + "step": 20507 + }, + { + "epoch": 4.74, + "learning_rate": 1.0503813265541947e-05, + "loss": 0.0, + "step": 20508 + }, + { + "epoch": 4.74, + "learning_rate": 1.0501887373854095e-05, + "loss": 0.0, + "step": 20509 + }, + { + "epoch": 4.74, + "learning_rate": 1.0499961482166242e-05, + "loss": 0.0, + "step": 20510 + }, + { + "epoch": 4.74, + "learning_rate": 1.0498035590478391e-05, + "loss": 0.0, + "step": 20511 + }, + { + "epoch": 4.74, + "learning_rate": 1.049610969879054e-05, + "loss": 0.0, + "step": 20512 + }, + { + "epoch": 4.74, + "learning_rate": 1.049418380710269e-05, + "loss": 0.0, + "step": 20513 + }, + { + "epoch": 4.74, + "learning_rate": 1.0492257915414839e-05, + "loss": 0.0, + "step": 20514 + }, + { + "epoch": 4.74, + "learning_rate": 1.0490332023726986e-05, + "loss": 0.0, + "step": 20515 + }, + { + "epoch": 4.74, + "learning_rate": 1.0488406132039134e-05, + "loss": 0.0, + "step": 20516 + }, + { + "epoch": 4.74, + "learning_rate": 1.0486480240351283e-05, + "loss": 0.0, + "step": 20517 + }, + { + "epoch": 4.74, + "learning_rate": 1.0484554348663432e-05, + "loss": 0.0, + "step": 20518 + }, + { + "epoch": 4.74, + "learning_rate": 1.0482628456975581e-05, + "loss": 0.0, + "step": 20519 + }, + { + "epoch": 4.74, + "learning_rate": 1.0480702565287729e-05, + "loss": 0.0, + "step": 20520 + }, + { + "epoch": 4.74, + "learning_rate": 1.0478776673599878e-05, + "loss": 0.0, + "step": 20521 + }, + { + "epoch": 4.74, + "learning_rate": 1.0476850781912025e-05, + "loss": 0.0, + "step": 20522 + }, + { + "epoch": 4.74, + "learning_rate": 1.0474924890224175e-05, + "loss": 0.0, + "step": 20523 + }, + { + "epoch": 4.74, + "learning_rate": 1.0472998998536322e-05, + "loss": 0.0, + "step": 20524 + }, + { + "epoch": 4.74, + "learning_rate": 1.0471073106848471e-05, + "loss": 0.0, + "step": 20525 + }, + { + "epoch": 4.74, + "learning_rate": 1.046914721516062e-05, + "loss": 0.0, + "step": 20526 + }, + { + "epoch": 4.74, + "learning_rate": 1.0467221323472768e-05, + "loss": 0.0, + "step": 20527 + }, + { + "epoch": 4.74, + "learning_rate": 1.0465295431784917e-05, + "loss": 0.0, + "step": 20528 + }, + { + "epoch": 4.74, + "learning_rate": 1.0463369540097065e-05, + "loss": 0.0, + "step": 20529 + }, + { + "epoch": 4.74, + "learning_rate": 1.0461443648409214e-05, + "loss": 0.0, + "step": 20530 + }, + { + "epoch": 4.74, + "learning_rate": 1.0459517756721363e-05, + "loss": 0.0, + "step": 20531 + }, + { + "epoch": 4.75, + "learning_rate": 1.0457591865033512e-05, + "loss": 0.0, + "step": 20532 + }, + { + "epoch": 4.75, + "learning_rate": 1.045566597334566e-05, + "loss": 0.0, + "step": 20533 + }, + { + "epoch": 4.75, + "learning_rate": 1.0453740081657807e-05, + "loss": 0.0, + "step": 20534 + }, + { + "epoch": 4.75, + "learning_rate": 1.0451814189969956e-05, + "loss": 0.0001, + "step": 20535 + }, + { + "epoch": 4.75, + "learning_rate": 1.0449888298282105e-05, + "loss": 0.0, + "step": 20536 + }, + { + "epoch": 4.75, + "learning_rate": 1.0447962406594255e-05, + "loss": 0.0, + "step": 20537 + }, + { + "epoch": 4.75, + "learning_rate": 1.0446036514906402e-05, + "loss": 0.0, + "step": 20538 + }, + { + "epoch": 4.75, + "learning_rate": 1.0444110623218551e-05, + "loss": 0.0, + "step": 20539 + }, + { + "epoch": 4.75, + "learning_rate": 1.0442184731530699e-05, + "loss": 0.0, + "step": 20540 + }, + { + "epoch": 4.75, + "learning_rate": 1.0440258839842848e-05, + "loss": 0.0, + "step": 20541 + }, + { + "epoch": 4.75, + "learning_rate": 1.0438332948154997e-05, + "loss": 0.0, + "step": 20542 + }, + { + "epoch": 4.75, + "learning_rate": 1.0436407056467145e-05, + "loss": 0.0, + "step": 20543 + }, + { + "epoch": 4.75, + "learning_rate": 1.0434481164779294e-05, + "loss": 0.0, + "step": 20544 + }, + { + "epoch": 4.75, + "learning_rate": 1.0432555273091441e-05, + "loss": 0.0, + "step": 20545 + }, + { + "epoch": 4.75, + "learning_rate": 1.043062938140359e-05, + "loss": 0.0, + "step": 20546 + }, + { + "epoch": 4.75, + "learning_rate": 1.042870348971574e-05, + "loss": 0.0047, + "step": 20547 + }, + { + "epoch": 4.75, + "learning_rate": 1.0426777598027887e-05, + "loss": 0.0, + "step": 20548 + }, + { + "epoch": 4.75, + "learning_rate": 1.0424851706340036e-05, + "loss": 0.0, + "step": 20549 + }, + { + "epoch": 4.75, + "learning_rate": 1.0422925814652185e-05, + "loss": 0.0, + "step": 20550 + }, + { + "epoch": 4.75, + "learning_rate": 1.0420999922964333e-05, + "loss": 0.0, + "step": 20551 + }, + { + "epoch": 4.75, + "learning_rate": 1.041907403127648e-05, + "loss": 0.0001, + "step": 20552 + }, + { + "epoch": 4.75, + "learning_rate": 1.041714813958863e-05, + "loss": 0.0, + "step": 20553 + }, + { + "epoch": 4.75, + "learning_rate": 1.0415222247900779e-05, + "loss": 0.0, + "step": 20554 + }, + { + "epoch": 4.75, + "learning_rate": 1.0413296356212928e-05, + "loss": 0.0, + "step": 20555 + }, + { + "epoch": 4.75, + "learning_rate": 1.0411370464525077e-05, + "loss": 0.0, + "step": 20556 + }, + { + "epoch": 4.75, + "learning_rate": 1.0409444572837224e-05, + "loss": 0.0, + "step": 20557 + }, + { + "epoch": 4.75, + "learning_rate": 1.0407518681149372e-05, + "loss": 0.0, + "step": 20558 + }, + { + "epoch": 4.75, + "learning_rate": 1.0405592789461521e-05, + "loss": 0.0, + "step": 20559 + }, + { + "epoch": 4.75, + "learning_rate": 1.040366689777367e-05, + "loss": 0.0, + "step": 20560 + }, + { + "epoch": 4.75, + "learning_rate": 1.040174100608582e-05, + "loss": 0.0, + "step": 20561 + }, + { + "epoch": 4.75, + "learning_rate": 1.0399815114397967e-05, + "loss": 0.0, + "step": 20562 + }, + { + "epoch": 4.75, + "learning_rate": 1.0397889222710114e-05, + "loss": 0.0, + "step": 20563 + }, + { + "epoch": 4.75, + "learning_rate": 1.0395963331022264e-05, + "loss": 0.0, + "step": 20564 + }, + { + "epoch": 4.75, + "learning_rate": 1.0394037439334413e-05, + "loss": 0.0, + "step": 20565 + }, + { + "epoch": 4.75, + "learning_rate": 1.039211154764656e-05, + "loss": 0.0, + "step": 20566 + }, + { + "epoch": 4.75, + "learning_rate": 1.039018565595871e-05, + "loss": 0.0, + "step": 20567 + }, + { + "epoch": 4.75, + "learning_rate": 1.0388259764270859e-05, + "loss": 0.0, + "step": 20568 + }, + { + "epoch": 4.75, + "learning_rate": 1.0386333872583006e-05, + "loss": 0.0, + "step": 20569 + }, + { + "epoch": 4.75, + "learning_rate": 1.0384407980895155e-05, + "loss": 0.0, + "step": 20570 + }, + { + "epoch": 4.75, + "learning_rate": 1.0382482089207303e-05, + "loss": 0.0, + "step": 20571 + }, + { + "epoch": 4.75, + "learning_rate": 1.0380556197519452e-05, + "loss": 0.0, + "step": 20572 + }, + { + "epoch": 4.75, + "learning_rate": 1.0378630305831601e-05, + "loss": 0.0, + "step": 20573 + }, + { + "epoch": 4.75, + "learning_rate": 1.037670441414375e-05, + "loss": 0.0, + "step": 20574 + }, + { + "epoch": 4.76, + "learning_rate": 1.0374778522455898e-05, + "loss": 0.0, + "step": 20575 + }, + { + "epoch": 4.76, + "learning_rate": 1.0372852630768045e-05, + "loss": 0.0, + "step": 20576 + }, + { + "epoch": 4.76, + "learning_rate": 1.0370926739080194e-05, + "loss": 0.0, + "step": 20577 + }, + { + "epoch": 4.76, + "learning_rate": 1.0369000847392343e-05, + "loss": 0.0, + "step": 20578 + }, + { + "epoch": 4.76, + "learning_rate": 1.0367074955704493e-05, + "loss": 0.0, + "step": 20579 + }, + { + "epoch": 4.76, + "learning_rate": 1.036514906401664e-05, + "loss": 0.0, + "step": 20580 + }, + { + "epoch": 4.76, + "learning_rate": 1.0363223172328788e-05, + "loss": 0.0, + "step": 20581 + }, + { + "epoch": 4.76, + "learning_rate": 1.0361297280640937e-05, + "loss": 0.0009, + "step": 20582 + }, + { + "epoch": 4.76, + "learning_rate": 1.0359371388953086e-05, + "loss": 0.0, + "step": 20583 + }, + { + "epoch": 4.76, + "learning_rate": 1.0357445497265235e-05, + "loss": 0.0, + "step": 20584 + }, + { + "epoch": 4.76, + "learning_rate": 1.0355519605577383e-05, + "loss": 0.0, + "step": 20585 + }, + { + "epoch": 4.76, + "learning_rate": 1.0353593713889532e-05, + "loss": 0.0, + "step": 20586 + }, + { + "epoch": 4.76, + "learning_rate": 1.035166782220168e-05, + "loss": 0.0, + "step": 20587 + }, + { + "epoch": 4.76, + "learning_rate": 1.0349741930513828e-05, + "loss": 0.0, + "step": 20588 + }, + { + "epoch": 4.76, + "learning_rate": 1.0347816038825978e-05, + "loss": 0.0, + "step": 20589 + }, + { + "epoch": 4.76, + "learning_rate": 1.0345890147138125e-05, + "loss": 0.0, + "step": 20590 + }, + { + "epoch": 4.76, + "learning_rate": 1.0343964255450274e-05, + "loss": 0.0, + "step": 20591 + }, + { + "epoch": 4.76, + "learning_rate": 1.0342038363762423e-05, + "loss": 0.0, + "step": 20592 + }, + { + "epoch": 4.76, + "learning_rate": 1.0340112472074571e-05, + "loss": 0.0, + "step": 20593 + }, + { + "epoch": 4.76, + "learning_rate": 1.0338186580386718e-05, + "loss": 0.0, + "step": 20594 + }, + { + "epoch": 4.76, + "learning_rate": 1.0336260688698868e-05, + "loss": 0.0, + "step": 20595 + }, + { + "epoch": 4.76, + "learning_rate": 1.0334334797011017e-05, + "loss": 0.0, + "step": 20596 + }, + { + "epoch": 4.76, + "learning_rate": 1.0332408905323166e-05, + "loss": 0.0, + "step": 20597 + }, + { + "epoch": 4.76, + "learning_rate": 1.0330483013635315e-05, + "loss": 0.0, + "step": 20598 + }, + { + "epoch": 4.76, + "learning_rate": 1.0328557121947461e-05, + "loss": 0.0, + "step": 20599 + }, + { + "epoch": 4.76, + "learning_rate": 1.032663123025961e-05, + "loss": 0.0, + "step": 20600 + }, + { + "epoch": 4.76, + "learning_rate": 1.032470533857176e-05, + "loss": 0.0, + "step": 20601 + }, + { + "epoch": 4.76, + "learning_rate": 1.0322779446883908e-05, + "loss": 0.0, + "step": 20602 + }, + { + "epoch": 4.76, + "learning_rate": 1.0320853555196058e-05, + "loss": 0.0, + "step": 20603 + }, + { + "epoch": 4.76, + "learning_rate": 1.0318927663508205e-05, + "loss": 0.0, + "step": 20604 + }, + { + "epoch": 4.76, + "learning_rate": 1.0317001771820353e-05, + "loss": 0.0, + "step": 20605 + }, + { + "epoch": 4.76, + "learning_rate": 1.0315075880132502e-05, + "loss": 0.0, + "step": 20606 + }, + { + "epoch": 4.76, + "learning_rate": 1.0313149988444651e-05, + "loss": 0.0, + "step": 20607 + }, + { + "epoch": 4.76, + "learning_rate": 1.0311224096756798e-05, + "loss": 0.0, + "step": 20608 + }, + { + "epoch": 4.76, + "learning_rate": 1.0309298205068947e-05, + "loss": 0.0, + "step": 20609 + }, + { + "epoch": 4.76, + "learning_rate": 1.0307372313381097e-05, + "loss": 0.0, + "step": 20610 + }, + { + "epoch": 4.76, + "learning_rate": 1.0305446421693244e-05, + "loss": 0.0, + "step": 20611 + }, + { + "epoch": 4.76, + "learning_rate": 1.0303520530005393e-05, + "loss": 0.0, + "step": 20612 + }, + { + "epoch": 4.76, + "learning_rate": 1.030159463831754e-05, + "loss": 0.0, + "step": 20613 + }, + { + "epoch": 4.76, + "learning_rate": 1.029966874662969e-05, + "loss": 0.0001, + "step": 20614 + }, + { + "epoch": 4.76, + "learning_rate": 1.0297742854941839e-05, + "loss": 0.0, + "step": 20615 + }, + { + "epoch": 4.76, + "learning_rate": 1.0295816963253988e-05, + "loss": 0.0, + "step": 20616 + }, + { + "epoch": 4.76, + "learning_rate": 1.0293891071566136e-05, + "loss": 0.0, + "step": 20617 + }, + { + "epoch": 4.76, + "learning_rate": 1.0291965179878283e-05, + "loss": 0.0, + "step": 20618 + }, + { + "epoch": 4.77, + "learning_rate": 1.0290039288190432e-05, + "loss": 0.0, + "step": 20619 + }, + { + "epoch": 4.77, + "learning_rate": 1.0288113396502582e-05, + "loss": 0.0, + "step": 20620 + }, + { + "epoch": 4.77, + "learning_rate": 1.028618750481473e-05, + "loss": 0.0, + "step": 20621 + }, + { + "epoch": 4.77, + "learning_rate": 1.0284261613126878e-05, + "loss": 0.0, + "step": 20622 + }, + { + "epoch": 4.77, + "learning_rate": 1.0282335721439026e-05, + "loss": 0.0, + "step": 20623 + }, + { + "epoch": 4.77, + "learning_rate": 1.0280409829751175e-05, + "loss": 0.0, + "step": 20624 + }, + { + "epoch": 4.77, + "learning_rate": 1.0278483938063324e-05, + "loss": 0.0, + "step": 20625 + }, + { + "epoch": 4.77, + "learning_rate": 1.0276558046375473e-05, + "loss": 0.0, + "step": 20626 + }, + { + "epoch": 4.77, + "learning_rate": 1.027463215468762e-05, + "loss": 0.0, + "step": 20627 + }, + { + "epoch": 4.77, + "learning_rate": 1.027270626299977e-05, + "loss": 0.0, + "step": 20628 + }, + { + "epoch": 4.77, + "learning_rate": 1.0270780371311917e-05, + "loss": 0.0, + "step": 20629 + }, + { + "epoch": 4.77, + "learning_rate": 1.0268854479624067e-05, + "loss": 0.0, + "step": 20630 + }, + { + "epoch": 4.77, + "learning_rate": 1.0266928587936216e-05, + "loss": 0.0, + "step": 20631 + }, + { + "epoch": 4.77, + "learning_rate": 1.0265002696248363e-05, + "loss": 0.0, + "step": 20632 + }, + { + "epoch": 4.77, + "learning_rate": 1.0263076804560512e-05, + "loss": 0.0, + "step": 20633 + }, + { + "epoch": 4.77, + "learning_rate": 1.0261150912872662e-05, + "loss": 0.0, + "step": 20634 + }, + { + "epoch": 4.77, + "learning_rate": 1.0259225021184809e-05, + "loss": 0.0001, + "step": 20635 + }, + { + "epoch": 4.77, + "learning_rate": 1.0257299129496957e-05, + "loss": 0.0, + "step": 20636 + }, + { + "epoch": 4.77, + "learning_rate": 1.0255373237809106e-05, + "loss": 0.0, + "step": 20637 + }, + { + "epoch": 4.77, + "learning_rate": 1.0253447346121255e-05, + "loss": 0.0, + "step": 20638 + }, + { + "epoch": 4.77, + "learning_rate": 1.0251521454433404e-05, + "loss": 0.0, + "step": 20639 + }, + { + "epoch": 4.77, + "learning_rate": 1.0249595562745551e-05, + "loss": 0.0, + "step": 20640 + }, + { + "epoch": 4.77, + "learning_rate": 1.0247669671057699e-05, + "loss": 0.0, + "step": 20641 + }, + { + "epoch": 4.77, + "learning_rate": 1.0245743779369848e-05, + "loss": 0.0, + "step": 20642 + }, + { + "epoch": 4.77, + "learning_rate": 1.0243817887681997e-05, + "loss": 0.0, + "step": 20643 + }, + { + "epoch": 4.77, + "learning_rate": 1.0241891995994146e-05, + "loss": 0.0, + "step": 20644 + }, + { + "epoch": 4.77, + "learning_rate": 1.0239966104306296e-05, + "loss": 0.0, + "step": 20645 + }, + { + "epoch": 4.77, + "learning_rate": 1.0238040212618443e-05, + "loss": 0.0, + "step": 20646 + }, + { + "epoch": 4.77, + "learning_rate": 1.023611432093059e-05, + "loss": 0.0002, + "step": 20647 + }, + { + "epoch": 4.77, + "learning_rate": 1.023418842924274e-05, + "loss": 0.0, + "step": 20648 + }, + { + "epoch": 4.77, + "learning_rate": 1.0232262537554889e-05, + "loss": 0.0, + "step": 20649 + }, + { + "epoch": 4.77, + "learning_rate": 1.0230336645867036e-05, + "loss": 0.0, + "step": 20650 + }, + { + "epoch": 4.77, + "learning_rate": 1.0228410754179186e-05, + "loss": 0.0, + "step": 20651 + }, + { + "epoch": 4.77, + "learning_rate": 1.0226484862491335e-05, + "loss": 0.0002, + "step": 20652 + }, + { + "epoch": 4.77, + "learning_rate": 1.0224558970803482e-05, + "loss": 0.0, + "step": 20653 + }, + { + "epoch": 4.77, + "learning_rate": 1.0222633079115631e-05, + "loss": 0.0, + "step": 20654 + }, + { + "epoch": 4.77, + "learning_rate": 1.0220707187427779e-05, + "loss": 0.0, + "step": 20655 + }, + { + "epoch": 4.77, + "learning_rate": 1.0218781295739928e-05, + "loss": 0.0, + "step": 20656 + }, + { + "epoch": 4.77, + "learning_rate": 1.0216855404052077e-05, + "loss": 0.0, + "step": 20657 + }, + { + "epoch": 4.77, + "learning_rate": 1.0214929512364225e-05, + "loss": 0.0, + "step": 20658 + }, + { + "epoch": 4.77, + "learning_rate": 1.0213003620676374e-05, + "loss": 0.0, + "step": 20659 + }, + { + "epoch": 4.77, + "learning_rate": 1.0211077728988521e-05, + "loss": 0.0, + "step": 20660 + }, + { + "epoch": 4.77, + "learning_rate": 1.020915183730067e-05, + "loss": 0.0, + "step": 20661 + }, + { + "epoch": 4.78, + "learning_rate": 1.020722594561282e-05, + "loss": 0.0, + "step": 20662 + }, + { + "epoch": 4.78, + "learning_rate": 1.0205300053924969e-05, + "loss": 0.0, + "step": 20663 + }, + { + "epoch": 4.78, + "learning_rate": 1.0203374162237116e-05, + "loss": 0.0, + "step": 20664 + }, + { + "epoch": 4.78, + "learning_rate": 1.0201448270549264e-05, + "loss": 0.0, + "step": 20665 + }, + { + "epoch": 4.78, + "learning_rate": 1.0199522378861413e-05, + "loss": 0.0, + "step": 20666 + }, + { + "epoch": 4.78, + "learning_rate": 1.0197596487173562e-05, + "loss": 0.0, + "step": 20667 + }, + { + "epoch": 4.78, + "learning_rate": 1.0195670595485711e-05, + "loss": 0.0, + "step": 20668 + }, + { + "epoch": 4.78, + "learning_rate": 1.0193744703797859e-05, + "loss": 0.0, + "step": 20669 + }, + { + "epoch": 4.78, + "learning_rate": 1.0191818812110008e-05, + "loss": 0.0, + "step": 20670 + }, + { + "epoch": 4.78, + "learning_rate": 1.0189892920422155e-05, + "loss": 0.0, + "step": 20671 + }, + { + "epoch": 4.78, + "learning_rate": 1.0187967028734305e-05, + "loss": 0.0, + "step": 20672 + }, + { + "epoch": 4.78, + "learning_rate": 1.0186041137046454e-05, + "loss": 0.0, + "step": 20673 + }, + { + "epoch": 4.78, + "learning_rate": 1.0184115245358601e-05, + "loss": 0.0, + "step": 20674 + }, + { + "epoch": 4.78, + "learning_rate": 1.018218935367075e-05, + "loss": 0.0, + "step": 20675 + }, + { + "epoch": 4.78, + "learning_rate": 1.0180263461982898e-05, + "loss": 0.0, + "step": 20676 + }, + { + "epoch": 4.78, + "learning_rate": 1.0178337570295047e-05, + "loss": 0.0, + "step": 20677 + }, + { + "epoch": 4.78, + "learning_rate": 1.0176411678607195e-05, + "loss": 0.0, + "step": 20678 + }, + { + "epoch": 4.78, + "learning_rate": 1.0174485786919344e-05, + "loss": 0.0, + "step": 20679 + }, + { + "epoch": 4.78, + "learning_rate": 1.0172559895231493e-05, + "loss": 0.0009, + "step": 20680 + }, + { + "epoch": 4.78, + "learning_rate": 1.0170634003543642e-05, + "loss": 0.0, + "step": 20681 + }, + { + "epoch": 4.78, + "learning_rate": 1.016870811185579e-05, + "loss": 0.0, + "step": 20682 + }, + { + "epoch": 4.78, + "learning_rate": 1.0166782220167937e-05, + "loss": 0.0, + "step": 20683 + }, + { + "epoch": 4.78, + "learning_rate": 1.0164856328480086e-05, + "loss": 0.0, + "step": 20684 + }, + { + "epoch": 4.78, + "learning_rate": 1.0162930436792235e-05, + "loss": 0.0, + "step": 20685 + }, + { + "epoch": 4.78, + "learning_rate": 1.0161004545104385e-05, + "loss": 0.0, + "step": 20686 + }, + { + "epoch": 4.78, + "learning_rate": 1.0159078653416534e-05, + "loss": 0.0, + "step": 20687 + }, + { + "epoch": 4.78, + "learning_rate": 1.0157152761728681e-05, + "loss": 0.0, + "step": 20688 + }, + { + "epoch": 4.78, + "learning_rate": 1.0155226870040829e-05, + "loss": 0.0, + "step": 20689 + }, + { + "epoch": 4.78, + "learning_rate": 1.0153300978352978e-05, + "loss": 0.0, + "step": 20690 + }, + { + "epoch": 4.78, + "learning_rate": 1.0151375086665127e-05, + "loss": 0.0001, + "step": 20691 + }, + { + "epoch": 4.78, + "learning_rate": 1.0149449194977275e-05, + "loss": 0.0, + "step": 20692 + }, + { + "epoch": 4.78, + "learning_rate": 1.0147523303289424e-05, + "loss": 0.0, + "step": 20693 + }, + { + "epoch": 4.78, + "learning_rate": 1.0145597411601571e-05, + "loss": 0.0, + "step": 20694 + }, + { + "epoch": 4.78, + "learning_rate": 1.014367151991372e-05, + "loss": 0.0, + "step": 20695 + }, + { + "epoch": 4.78, + "learning_rate": 1.014174562822587e-05, + "loss": 0.0, + "step": 20696 + }, + { + "epoch": 4.78, + "learning_rate": 1.0139819736538017e-05, + "loss": 0.0, + "step": 20697 + }, + { + "epoch": 4.78, + "learning_rate": 1.0137893844850166e-05, + "loss": 0.0, + "step": 20698 + }, + { + "epoch": 4.78, + "learning_rate": 1.0135967953162315e-05, + "loss": 0.0, + "step": 20699 + }, + { + "epoch": 4.78, + "learning_rate": 1.0134042061474463e-05, + "loss": 0.0, + "step": 20700 + }, + { + "epoch": 4.78, + "learning_rate": 1.0132116169786612e-05, + "loss": 0.0, + "step": 20701 + }, + { + "epoch": 4.78, + "learning_rate": 1.013019027809876e-05, + "loss": 0.0, + "step": 20702 + }, + { + "epoch": 4.78, + "learning_rate": 1.0128264386410909e-05, + "loss": 0.0, + "step": 20703 + }, + { + "epoch": 4.78, + "learning_rate": 1.0126338494723058e-05, + "loss": 0.0, + "step": 20704 + }, + { + "epoch": 4.79, + "learning_rate": 1.0124412603035207e-05, + "loss": 0.0, + "step": 20705 + }, + { + "epoch": 4.79, + "learning_rate": 1.0122486711347354e-05, + "loss": 0.0, + "step": 20706 + }, + { + "epoch": 4.79, + "learning_rate": 1.0120560819659502e-05, + "loss": 0.0, + "step": 20707 + }, + { + "epoch": 4.79, + "learning_rate": 1.0118634927971651e-05, + "loss": 0.0, + "step": 20708 + }, + { + "epoch": 4.79, + "learning_rate": 1.01167090362838e-05, + "loss": 0.0, + "step": 20709 + }, + { + "epoch": 4.79, + "learning_rate": 1.011478314459595e-05, + "loss": 0.0, + "step": 20710 + }, + { + "epoch": 4.79, + "learning_rate": 1.0112857252908097e-05, + "loss": 0.0, + "step": 20711 + }, + { + "epoch": 4.79, + "learning_rate": 1.0110931361220244e-05, + "loss": 0.0, + "step": 20712 + }, + { + "epoch": 4.79, + "learning_rate": 1.0109005469532394e-05, + "loss": 0.0, + "step": 20713 + }, + { + "epoch": 4.79, + "learning_rate": 1.0107079577844543e-05, + "loss": 0.0, + "step": 20714 + }, + { + "epoch": 4.79, + "learning_rate": 1.0105153686156692e-05, + "loss": 0.0, + "step": 20715 + }, + { + "epoch": 4.79, + "learning_rate": 1.010322779446884e-05, + "loss": 0.0, + "step": 20716 + }, + { + "epoch": 4.79, + "learning_rate": 1.0101301902780989e-05, + "loss": 0.0, + "step": 20717 + }, + { + "epoch": 4.79, + "learning_rate": 1.0099376011093136e-05, + "loss": 0.0001, + "step": 20718 + }, + { + "epoch": 4.79, + "learning_rate": 1.0097450119405285e-05, + "loss": 0.0, + "step": 20719 + }, + { + "epoch": 4.79, + "learning_rate": 1.0095524227717433e-05, + "loss": 0.0, + "step": 20720 + }, + { + "epoch": 4.79, + "learning_rate": 1.0093598336029582e-05, + "loss": 0.0, + "step": 20721 + }, + { + "epoch": 4.79, + "learning_rate": 1.0091672444341731e-05, + "loss": 0.0, + "step": 20722 + }, + { + "epoch": 4.79, + "learning_rate": 1.008974655265388e-05, + "loss": 0.0, + "step": 20723 + }, + { + "epoch": 4.79, + "learning_rate": 1.0087820660966028e-05, + "loss": 0.0, + "step": 20724 + }, + { + "epoch": 4.79, + "learning_rate": 1.0085894769278175e-05, + "loss": 0.0, + "step": 20725 + }, + { + "epoch": 4.79, + "learning_rate": 1.0083968877590324e-05, + "loss": 0.0, + "step": 20726 + }, + { + "epoch": 4.79, + "learning_rate": 1.0082042985902474e-05, + "loss": 0.0, + "step": 20727 + }, + { + "epoch": 4.79, + "learning_rate": 1.0080117094214623e-05, + "loss": 0.0, + "step": 20728 + }, + { + "epoch": 4.79, + "learning_rate": 1.0078191202526772e-05, + "loss": 0.0, + "step": 20729 + }, + { + "epoch": 4.79, + "learning_rate": 1.0076265310838918e-05, + "loss": 0.0, + "step": 20730 + }, + { + "epoch": 4.79, + "learning_rate": 1.0074339419151067e-05, + "loss": 0.0, + "step": 20731 + }, + { + "epoch": 4.79, + "learning_rate": 1.0072413527463216e-05, + "loss": 0.0001, + "step": 20732 + }, + { + "epoch": 4.79, + "learning_rate": 1.0070487635775365e-05, + "loss": 0.0, + "step": 20733 + }, + { + "epoch": 4.79, + "learning_rate": 1.0068561744087513e-05, + "loss": 0.0, + "step": 20734 + }, + { + "epoch": 4.79, + "learning_rate": 1.0066635852399662e-05, + "loss": 0.0, + "step": 20735 + }, + { + "epoch": 4.79, + "learning_rate": 1.006470996071181e-05, + "loss": 0.0, + "step": 20736 + }, + { + "epoch": 4.79, + "learning_rate": 1.0062784069023958e-05, + "loss": 0.0, + "step": 20737 + }, + { + "epoch": 4.79, + "learning_rate": 1.0060858177336108e-05, + "loss": 0.0, + "step": 20738 + }, + { + "epoch": 4.79, + "learning_rate": 1.0058932285648255e-05, + "loss": 0.0, + "step": 20739 + }, + { + "epoch": 4.79, + "learning_rate": 1.0057006393960404e-05, + "loss": 0.0, + "step": 20740 + }, + { + "epoch": 4.79, + "learning_rate": 1.0055080502272553e-05, + "loss": 0.0, + "step": 20741 + }, + { + "epoch": 4.79, + "learning_rate": 1.0053154610584701e-05, + "loss": 0.0, + "step": 20742 + }, + { + "epoch": 4.79, + "learning_rate": 1.005122871889685e-05, + "loss": 0.0, + "step": 20743 + }, + { + "epoch": 4.79, + "learning_rate": 1.0049302827208998e-05, + "loss": 0.0, + "step": 20744 + }, + { + "epoch": 4.79, + "learning_rate": 1.0047376935521147e-05, + "loss": 0.0, + "step": 20745 + }, + { + "epoch": 4.79, + "learning_rate": 1.0045451043833296e-05, + "loss": 0.0, + "step": 20746 + }, + { + "epoch": 4.79, + "learning_rate": 1.0043525152145445e-05, + "loss": 0.0, + "step": 20747 + }, + { + "epoch": 4.8, + "learning_rate": 1.0041599260457593e-05, + "loss": 0.0, + "step": 20748 + }, + { + "epoch": 4.8, + "learning_rate": 1.003967336876974e-05, + "loss": 0.0, + "step": 20749 + }, + { + "epoch": 4.8, + "learning_rate": 1.003774747708189e-05, + "loss": 0.0, + "step": 20750 + }, + { + "epoch": 4.8, + "learning_rate": 1.0035821585394038e-05, + "loss": 0.0, + "step": 20751 + }, + { + "epoch": 4.8, + "learning_rate": 1.0033895693706188e-05, + "loss": 0.0, + "step": 20752 + }, + { + "epoch": 4.8, + "learning_rate": 1.0031969802018335e-05, + "loss": 0.0, + "step": 20753 + }, + { + "epoch": 4.8, + "learning_rate": 1.0030043910330483e-05, + "loss": 0.0, + "step": 20754 + }, + { + "epoch": 4.8, + "learning_rate": 1.0028118018642632e-05, + "loss": 0.0, + "step": 20755 + }, + { + "epoch": 4.8, + "learning_rate": 1.0026192126954781e-05, + "loss": 0.0, + "step": 20756 + }, + { + "epoch": 4.8, + "learning_rate": 1.002426623526693e-05, + "loss": 0.0, + "step": 20757 + }, + { + "epoch": 4.8, + "learning_rate": 1.0022340343579078e-05, + "loss": 0.0, + "step": 20758 + }, + { + "epoch": 4.8, + "learning_rate": 1.0020414451891227e-05, + "loss": 0.0, + "step": 20759 + }, + { + "epoch": 4.8, + "learning_rate": 1.0018488560203374e-05, + "loss": 0.0, + "step": 20760 + }, + { + "epoch": 4.8, + "learning_rate": 1.0016562668515523e-05, + "loss": 0.0, + "step": 20761 + }, + { + "epoch": 4.8, + "learning_rate": 1.001463677682767e-05, + "loss": 0.0, + "step": 20762 + }, + { + "epoch": 4.8, + "learning_rate": 1.001271088513982e-05, + "loss": 0.0, + "step": 20763 + }, + { + "epoch": 4.8, + "learning_rate": 1.001078499345197e-05, + "loss": 0.0, + "step": 20764 + }, + { + "epoch": 4.8, + "learning_rate": 1.0008859101764118e-05, + "loss": 0.0, + "step": 20765 + }, + { + "epoch": 4.8, + "learning_rate": 1.0006933210076266e-05, + "loss": 0.0, + "step": 20766 + }, + { + "epoch": 4.8, + "learning_rate": 1.0005007318388413e-05, + "loss": 0.0, + "step": 20767 + }, + { + "epoch": 4.8, + "learning_rate": 1.0003081426700562e-05, + "loss": 0.0, + "step": 20768 + }, + { + "epoch": 4.8, + "learning_rate": 1.0001155535012712e-05, + "loss": 0.0, + "step": 20769 + }, + { + "epoch": 4.8, + "learning_rate": 9.99922964332486e-06, + "loss": 0.0, + "step": 20770 + }, + { + "epoch": 4.8, + "learning_rate": 9.997303751637008e-06, + "loss": 0.0, + "step": 20771 + }, + { + "epoch": 4.8, + "learning_rate": 9.995377859949156e-06, + "loss": 0.0, + "step": 20772 + }, + { + "epoch": 4.8, + "learning_rate": 9.993451968261305e-06, + "loss": 0.0, + "step": 20773 + }, + { + "epoch": 4.8, + "learning_rate": 9.991526076573454e-06, + "loss": 0.0, + "step": 20774 + }, + { + "epoch": 4.8, + "learning_rate": 9.989600184885603e-06, + "loss": 0.0, + "step": 20775 + }, + { + "epoch": 4.8, + "learning_rate": 9.98767429319775e-06, + "loss": 0.0, + "step": 20776 + }, + { + "epoch": 4.8, + "learning_rate": 9.9857484015099e-06, + "loss": 0.0, + "step": 20777 + }, + { + "epoch": 4.8, + "learning_rate": 9.983822509822047e-06, + "loss": 0.0, + "step": 20778 + }, + { + "epoch": 4.8, + "learning_rate": 9.981896618134197e-06, + "loss": 0.0, + "step": 20779 + }, + { + "epoch": 4.8, + "learning_rate": 9.979970726446346e-06, + "loss": 0.0, + "step": 20780 + }, + { + "epoch": 4.8, + "learning_rate": 9.978044834758493e-06, + "loss": 0.0, + "step": 20781 + }, + { + "epoch": 4.8, + "learning_rate": 9.976118943070642e-06, + "loss": 0.0, + "step": 20782 + }, + { + "epoch": 4.8, + "learning_rate": 9.974193051382792e-06, + "loss": 0.0, + "step": 20783 + }, + { + "epoch": 4.8, + "learning_rate": 9.972267159694939e-06, + "loss": 0.0, + "step": 20784 + }, + { + "epoch": 4.8, + "learning_rate": 9.970341268007087e-06, + "loss": 0.0, + "step": 20785 + }, + { + "epoch": 4.8, + "learning_rate": 9.968415376319236e-06, + "loss": 0.0, + "step": 20786 + }, + { + "epoch": 4.8, + "learning_rate": 9.966489484631385e-06, + "loss": 0.0, + "step": 20787 + }, + { + "epoch": 4.8, + "learning_rate": 9.964563592943534e-06, + "loss": 0.0, + "step": 20788 + }, + { + "epoch": 4.8, + "learning_rate": 9.962637701255682e-06, + "loss": 0.0, + "step": 20789 + }, + { + "epoch": 4.8, + "learning_rate": 9.960711809567829e-06, + "loss": 0.0, + "step": 20790 + }, + { + "epoch": 4.8, + "learning_rate": 9.958785917879978e-06, + "loss": 0.0, + "step": 20791 + }, + { + "epoch": 4.81, + "learning_rate": 9.956860026192127e-06, + "loss": 0.0, + "step": 20792 + }, + { + "epoch": 4.81, + "learning_rate": 9.954934134504277e-06, + "loss": 0.0, + "step": 20793 + }, + { + "epoch": 4.81, + "learning_rate": 9.953008242816426e-06, + "loss": 0.0, + "step": 20794 + }, + { + "epoch": 4.81, + "learning_rate": 9.951082351128573e-06, + "loss": 0.0, + "step": 20795 + }, + { + "epoch": 4.81, + "learning_rate": 9.94915645944072e-06, + "loss": 0.0, + "step": 20796 + }, + { + "epoch": 4.81, + "learning_rate": 9.94723056775287e-06, + "loss": 0.0, + "step": 20797 + }, + { + "epoch": 4.81, + "learning_rate": 9.945304676065019e-06, + "loss": 0.0, + "step": 20798 + }, + { + "epoch": 4.81, + "learning_rate": 9.943378784377166e-06, + "loss": 0.0, + "step": 20799 + }, + { + "epoch": 4.81, + "learning_rate": 9.941452892689316e-06, + "loss": 0.0, + "step": 20800 + }, + { + "epoch": 4.81, + "learning_rate": 9.939527001001465e-06, + "loss": 0.0001, + "step": 20801 + }, + { + "epoch": 4.81, + "learning_rate": 9.937601109313612e-06, + "loss": 0.0, + "step": 20802 + }, + { + "epoch": 4.81, + "learning_rate": 9.935675217625761e-06, + "loss": 0.0, + "step": 20803 + }, + { + "epoch": 4.81, + "learning_rate": 9.933749325937909e-06, + "loss": 0.0, + "step": 20804 + }, + { + "epoch": 4.81, + "learning_rate": 9.931823434250058e-06, + "loss": 0.0, + "step": 20805 + }, + { + "epoch": 4.81, + "learning_rate": 9.929897542562207e-06, + "loss": 0.0, + "step": 20806 + }, + { + "epoch": 4.81, + "learning_rate": 9.927971650874355e-06, + "loss": 0.0, + "step": 20807 + }, + { + "epoch": 4.81, + "learning_rate": 9.926045759186504e-06, + "loss": 0.0, + "step": 20808 + }, + { + "epoch": 4.81, + "learning_rate": 9.924119867498651e-06, + "loss": 0.0, + "step": 20809 + }, + { + "epoch": 4.81, + "learning_rate": 9.9221939758108e-06, + "loss": 0.0, + "step": 20810 + }, + { + "epoch": 4.81, + "learning_rate": 9.92026808412295e-06, + "loss": 0.0, + "step": 20811 + }, + { + "epoch": 4.81, + "learning_rate": 9.918342192435099e-06, + "loss": 0.0, + "step": 20812 + }, + { + "epoch": 4.81, + "learning_rate": 9.916416300747246e-06, + "loss": 0.0, + "step": 20813 + }, + { + "epoch": 4.81, + "learning_rate": 9.914490409059394e-06, + "loss": 0.0, + "step": 20814 + }, + { + "epoch": 4.81, + "learning_rate": 9.912564517371543e-06, + "loss": 0.0002, + "step": 20815 + }, + { + "epoch": 4.81, + "learning_rate": 9.910638625683692e-06, + "loss": 0.0, + "step": 20816 + }, + { + "epoch": 4.81, + "learning_rate": 9.908712733995841e-06, + "loss": 0.0, + "step": 20817 + }, + { + "epoch": 4.81, + "learning_rate": 9.906786842307989e-06, + "loss": 0.0, + "step": 20818 + }, + { + "epoch": 4.81, + "learning_rate": 9.904860950620138e-06, + "loss": 0.0, + "step": 20819 + }, + { + "epoch": 4.81, + "learning_rate": 9.902935058932286e-06, + "loss": 0.0001, + "step": 20820 + }, + { + "epoch": 4.81, + "learning_rate": 9.901009167244435e-06, + "loss": 0.0, + "step": 20821 + }, + { + "epoch": 4.81, + "learning_rate": 9.899083275556584e-06, + "loss": 0.0, + "step": 20822 + }, + { + "epoch": 4.81, + "learning_rate": 9.897157383868731e-06, + "loss": 0.0001, + "step": 20823 + }, + { + "epoch": 4.81, + "learning_rate": 9.89523149218088e-06, + "loss": 0.0001, + "step": 20824 + }, + { + "epoch": 4.81, + "learning_rate": 9.893305600493028e-06, + "loss": 0.0, + "step": 20825 + }, + { + "epoch": 4.81, + "learning_rate": 9.891379708805177e-06, + "loss": 0.0, + "step": 20826 + }, + { + "epoch": 4.81, + "learning_rate": 9.889453817117325e-06, + "loss": 0.0, + "step": 20827 + }, + { + "epoch": 4.81, + "learning_rate": 9.887527925429474e-06, + "loss": 0.0, + "step": 20828 + }, + { + "epoch": 4.81, + "learning_rate": 9.885602033741623e-06, + "loss": 0.0, + "step": 20829 + }, + { + "epoch": 4.81, + "learning_rate": 9.883676142053772e-06, + "loss": 0.0, + "step": 20830 + }, + { + "epoch": 4.81, + "learning_rate": 9.88175025036592e-06, + "loss": 0.0, + "step": 20831 + }, + { + "epoch": 4.81, + "learning_rate": 9.879824358678067e-06, + "loss": 0.0, + "step": 20832 + }, + { + "epoch": 4.81, + "learning_rate": 9.877898466990216e-06, + "loss": 0.0001, + "step": 20833 + }, + { + "epoch": 4.81, + "learning_rate": 9.875972575302365e-06, + "loss": 0.0, + "step": 20834 + }, + { + "epoch": 4.82, + "learning_rate": 9.874046683614515e-06, + "loss": 0.0, + "step": 20835 + }, + { + "epoch": 4.82, + "learning_rate": 9.872120791926664e-06, + "loss": 0.0, + "step": 20836 + }, + { + "epoch": 4.82, + "learning_rate": 9.870194900238811e-06, + "loss": 0.0, + "step": 20837 + }, + { + "epoch": 4.82, + "learning_rate": 9.868269008550959e-06, + "loss": 0.0, + "step": 20838 + }, + { + "epoch": 4.82, + "learning_rate": 9.866343116863108e-06, + "loss": 0.0, + "step": 20839 + }, + { + "epoch": 4.82, + "learning_rate": 9.864417225175257e-06, + "loss": 0.0, + "step": 20840 + }, + { + "epoch": 4.82, + "learning_rate": 9.862491333487405e-06, + "loss": 0.0, + "step": 20841 + }, + { + "epoch": 4.82, + "learning_rate": 9.860565441799554e-06, + "loss": 0.0, + "step": 20842 + }, + { + "epoch": 4.82, + "learning_rate": 9.858639550111703e-06, + "loss": 0.0, + "step": 20843 + }, + { + "epoch": 4.82, + "learning_rate": 9.85671365842385e-06, + "loss": 0.0, + "step": 20844 + }, + { + "epoch": 4.82, + "learning_rate": 9.854787766736e-06, + "loss": 0.0, + "step": 20845 + }, + { + "epoch": 4.82, + "learning_rate": 9.852861875048147e-06, + "loss": 0.0, + "step": 20846 + }, + { + "epoch": 4.82, + "learning_rate": 9.850935983360296e-06, + "loss": 0.0, + "step": 20847 + }, + { + "epoch": 4.82, + "learning_rate": 9.849010091672445e-06, + "loss": 0.0001, + "step": 20848 + }, + { + "epoch": 4.82, + "learning_rate": 9.847084199984593e-06, + "loss": 0.0, + "step": 20849 + }, + { + "epoch": 4.82, + "learning_rate": 9.845158308296742e-06, + "loss": 0.0, + "step": 20850 + }, + { + "epoch": 4.82, + "learning_rate": 9.84323241660889e-06, + "loss": 0.0, + "step": 20851 + }, + { + "epoch": 4.82, + "learning_rate": 9.841306524921039e-06, + "loss": 0.0, + "step": 20852 + }, + { + "epoch": 4.82, + "learning_rate": 9.839380633233188e-06, + "loss": 0.0, + "step": 20853 + }, + { + "epoch": 4.82, + "learning_rate": 9.837454741545337e-06, + "loss": 0.0, + "step": 20854 + }, + { + "epoch": 4.82, + "learning_rate": 9.835528849857485e-06, + "loss": 0.0, + "step": 20855 + }, + { + "epoch": 4.82, + "learning_rate": 9.833602958169632e-06, + "loss": 0.0, + "step": 20856 + }, + { + "epoch": 4.82, + "learning_rate": 9.831677066481781e-06, + "loss": 0.0, + "step": 20857 + }, + { + "epoch": 4.82, + "learning_rate": 9.82975117479393e-06, + "loss": 0.0, + "step": 20858 + }, + { + "epoch": 4.82, + "learning_rate": 9.82782528310608e-06, + "loss": 0.0, + "step": 20859 + }, + { + "epoch": 4.82, + "learning_rate": 9.825899391418227e-06, + "loss": 0.0, + "step": 20860 + }, + { + "epoch": 4.82, + "learning_rate": 9.823973499730376e-06, + "loss": 0.0, + "step": 20861 + }, + { + "epoch": 4.82, + "learning_rate": 9.822047608042524e-06, + "loss": 0.0, + "step": 20862 + }, + { + "epoch": 4.82, + "learning_rate": 9.820121716354673e-06, + "loss": 0.0, + "step": 20863 + }, + { + "epoch": 4.82, + "learning_rate": 9.818195824666822e-06, + "loss": 0.0, + "step": 20864 + }, + { + "epoch": 4.82, + "learning_rate": 9.81626993297897e-06, + "loss": 0.0, + "step": 20865 + }, + { + "epoch": 4.82, + "learning_rate": 9.814344041291119e-06, + "loss": 0.0, + "step": 20866 + }, + { + "epoch": 4.82, + "learning_rate": 9.812418149603266e-06, + "loss": 0.0, + "step": 20867 + }, + { + "epoch": 4.82, + "learning_rate": 9.810492257915415e-06, + "loss": 0.0, + "step": 20868 + }, + { + "epoch": 4.82, + "learning_rate": 9.808566366227563e-06, + "loss": 0.0, + "step": 20869 + }, + { + "epoch": 4.82, + "learning_rate": 9.806640474539712e-06, + "loss": 0.0, + "step": 20870 + }, + { + "epoch": 4.82, + "learning_rate": 9.804714582851861e-06, + "loss": 0.0, + "step": 20871 + }, + { + "epoch": 4.82, + "learning_rate": 9.80278869116401e-06, + "loss": 0.0, + "step": 20872 + }, + { + "epoch": 4.82, + "learning_rate": 9.800862799476158e-06, + "loss": 0.0, + "step": 20873 + }, + { + "epoch": 4.82, + "learning_rate": 9.798936907788305e-06, + "loss": 0.0, + "step": 20874 + }, + { + "epoch": 4.82, + "learning_rate": 9.797011016100454e-06, + "loss": 0.0, + "step": 20875 + }, + { + "epoch": 4.82, + "learning_rate": 9.795085124412604e-06, + "loss": 0.0, + "step": 20876 + }, + { + "epoch": 4.82, + "learning_rate": 9.793159232724753e-06, + "loss": 0.0, + "step": 20877 + }, + { + "epoch": 4.83, + "learning_rate": 9.791233341036902e-06, + "loss": 0.0, + "step": 20878 + }, + { + "epoch": 4.83, + "learning_rate": 9.78930744934905e-06, + "loss": 0.0, + "step": 20879 + }, + { + "epoch": 4.83, + "learning_rate": 9.787381557661197e-06, + "loss": 0.0, + "step": 20880 + }, + { + "epoch": 4.83, + "learning_rate": 9.785455665973346e-06, + "loss": 0.0, + "step": 20881 + }, + { + "epoch": 4.83, + "learning_rate": 9.783529774285495e-06, + "loss": 0.0, + "step": 20882 + }, + { + "epoch": 4.83, + "learning_rate": 9.781603882597643e-06, + "loss": 0.0001, + "step": 20883 + }, + { + "epoch": 4.83, + "learning_rate": 9.779677990909792e-06, + "loss": 0.0, + "step": 20884 + }, + { + "epoch": 4.83, + "learning_rate": 9.77775209922194e-06, + "loss": 0.0, + "step": 20885 + }, + { + "epoch": 4.83, + "learning_rate": 9.775826207534089e-06, + "loss": 0.0, + "step": 20886 + }, + { + "epoch": 4.83, + "learning_rate": 9.773900315846238e-06, + "loss": 0.0, + "step": 20887 + }, + { + "epoch": 4.83, + "learning_rate": 9.771974424158385e-06, + "loss": 0.0, + "step": 20888 + }, + { + "epoch": 4.83, + "learning_rate": 9.770048532470534e-06, + "loss": 0.0001, + "step": 20889 + }, + { + "epoch": 4.83, + "learning_rate": 9.768122640782684e-06, + "loss": 0.0, + "step": 20890 + }, + { + "epoch": 4.83, + "learning_rate": 9.766196749094831e-06, + "loss": 0.0, + "step": 20891 + }, + { + "epoch": 4.83, + "learning_rate": 9.76427085740698e-06, + "loss": 0.0, + "step": 20892 + }, + { + "epoch": 4.83, + "learning_rate": 9.762344965719128e-06, + "loss": 0.0, + "step": 20893 + }, + { + "epoch": 4.83, + "learning_rate": 9.760419074031277e-06, + "loss": 0.0, + "step": 20894 + }, + { + "epoch": 4.83, + "learning_rate": 9.758493182343426e-06, + "loss": 0.0, + "step": 20895 + }, + { + "epoch": 4.83, + "learning_rate": 9.756567290655575e-06, + "loss": 0.0001, + "step": 20896 + }, + { + "epoch": 4.83, + "learning_rate": 9.754641398967723e-06, + "loss": 0.0, + "step": 20897 + }, + { + "epoch": 4.83, + "learning_rate": 9.75271550727987e-06, + "loss": 0.0, + "step": 20898 + }, + { + "epoch": 4.83, + "learning_rate": 9.75078961559202e-06, + "loss": 0.0, + "step": 20899 + }, + { + "epoch": 4.83, + "learning_rate": 9.748863723904168e-06, + "loss": 0.0001, + "step": 20900 + }, + { + "epoch": 4.83, + "learning_rate": 9.746937832216318e-06, + "loss": 0.0, + "step": 20901 + }, + { + "epoch": 4.83, + "learning_rate": 9.745011940528465e-06, + "loss": 0.0, + "step": 20902 + }, + { + "epoch": 4.83, + "learning_rate": 9.743086048840613e-06, + "loss": 0.0, + "step": 20903 + }, + { + "epoch": 4.83, + "learning_rate": 9.741160157152762e-06, + "loss": 0.0001, + "step": 20904 + }, + { + "epoch": 4.83, + "learning_rate": 9.739234265464911e-06, + "loss": 0.0, + "step": 20905 + }, + { + "epoch": 4.83, + "learning_rate": 9.73730837377706e-06, + "loss": 0.0001, + "step": 20906 + }, + { + "epoch": 4.83, + "learning_rate": 9.735382482089208e-06, + "loss": 0.0, + "step": 20907 + }, + { + "epoch": 4.83, + "learning_rate": 9.733456590401357e-06, + "loss": 0.0, + "step": 20908 + }, + { + "epoch": 4.83, + "learning_rate": 9.731530698713504e-06, + "loss": 0.0, + "step": 20909 + }, + { + "epoch": 4.83, + "learning_rate": 9.729604807025653e-06, + "loss": 0.0, + "step": 20910 + }, + { + "epoch": 4.83, + "learning_rate": 9.727678915337801e-06, + "loss": 0.0, + "step": 20911 + }, + { + "epoch": 4.83, + "learning_rate": 9.72575302364995e-06, + "loss": 0.0, + "step": 20912 + }, + { + "epoch": 4.83, + "learning_rate": 9.7238271319621e-06, + "loss": 0.0, + "step": 20913 + }, + { + "epoch": 4.83, + "learning_rate": 9.721901240274248e-06, + "loss": 0.0, + "step": 20914 + }, + { + "epoch": 4.83, + "learning_rate": 9.719975348586396e-06, + "loss": 0.0, + "step": 20915 + }, + { + "epoch": 4.83, + "learning_rate": 9.718049456898543e-06, + "loss": 0.0, + "step": 20916 + }, + { + "epoch": 4.83, + "learning_rate": 9.716123565210693e-06, + "loss": 0.0, + "step": 20917 + }, + { + "epoch": 4.83, + "learning_rate": 9.714197673522842e-06, + "loss": 0.0, + "step": 20918 + }, + { + "epoch": 4.83, + "learning_rate": 9.712271781834991e-06, + "loss": 0.0, + "step": 20919 + }, + { + "epoch": 4.83, + "learning_rate": 9.710345890147138e-06, + "loss": 0.0, + "step": 20920 + }, + { + "epoch": 4.83, + "learning_rate": 9.708419998459286e-06, + "loss": 0.0, + "step": 20921 + }, + { + "epoch": 4.84, + "learning_rate": 9.706494106771435e-06, + "loss": 0.0, + "step": 20922 + }, + { + "epoch": 4.84, + "learning_rate": 9.704568215083584e-06, + "loss": 0.0, + "step": 20923 + }, + { + "epoch": 4.84, + "learning_rate": 9.702642323395733e-06, + "loss": 0.0, + "step": 20924 + }, + { + "epoch": 4.84, + "learning_rate": 9.70071643170788e-06, + "loss": 0.0, + "step": 20925 + }, + { + "epoch": 4.84, + "learning_rate": 9.69879054002003e-06, + "loss": 0.0, + "step": 20926 + }, + { + "epoch": 4.84, + "learning_rate": 9.696864648332177e-06, + "loss": 0.0, + "step": 20927 + }, + { + "epoch": 4.84, + "learning_rate": 9.694938756644327e-06, + "loss": 0.0, + "step": 20928 + }, + { + "epoch": 4.84, + "learning_rate": 9.693012864956476e-06, + "loss": 0.0, + "step": 20929 + }, + { + "epoch": 4.84, + "learning_rate": 9.691086973268623e-06, + "loss": 0.0, + "step": 20930 + }, + { + "epoch": 4.84, + "learning_rate": 9.689161081580772e-06, + "loss": 0.0, + "step": 20931 + }, + { + "epoch": 4.84, + "learning_rate": 9.687235189892922e-06, + "loss": 0.0, + "step": 20932 + }, + { + "epoch": 4.84, + "learning_rate": 9.685309298205069e-06, + "loss": 0.0, + "step": 20933 + }, + { + "epoch": 4.84, + "learning_rate": 9.683383406517218e-06, + "loss": 0.0, + "step": 20934 + }, + { + "epoch": 4.84, + "learning_rate": 9.681457514829366e-06, + "loss": 0.0, + "step": 20935 + }, + { + "epoch": 4.84, + "learning_rate": 9.679531623141515e-06, + "loss": 0.0, + "step": 20936 + }, + { + "epoch": 4.84, + "learning_rate": 9.677605731453664e-06, + "loss": 0.0, + "step": 20937 + }, + { + "epoch": 4.84, + "learning_rate": 9.675679839765813e-06, + "loss": 0.0, + "step": 20938 + }, + { + "epoch": 4.84, + "learning_rate": 9.673753948077959e-06, + "loss": 0.0, + "step": 20939 + }, + { + "epoch": 4.84, + "learning_rate": 9.671828056390108e-06, + "loss": 0.0, + "step": 20940 + }, + { + "epoch": 4.84, + "learning_rate": 9.669902164702257e-06, + "loss": 0.0, + "step": 20941 + }, + { + "epoch": 4.84, + "learning_rate": 9.667976273014407e-06, + "loss": 0.0, + "step": 20942 + }, + { + "epoch": 4.84, + "learning_rate": 9.666050381326556e-06, + "loss": 0.0, + "step": 20943 + }, + { + "epoch": 4.84, + "learning_rate": 9.664124489638703e-06, + "loss": 0.0, + "step": 20944 + }, + { + "epoch": 4.84, + "learning_rate": 9.66219859795085e-06, + "loss": 0.0, + "step": 20945 + }, + { + "epoch": 4.84, + "learning_rate": 9.660272706263e-06, + "loss": 0.0, + "step": 20946 + }, + { + "epoch": 4.84, + "learning_rate": 9.658346814575149e-06, + "loss": 0.0, + "step": 20947 + }, + { + "epoch": 4.84, + "learning_rate": 9.656420922887298e-06, + "loss": 0.0003, + "step": 20948 + }, + { + "epoch": 4.84, + "learning_rate": 9.654495031199446e-06, + "loss": 0.0, + "step": 20949 + }, + { + "epoch": 4.84, + "learning_rate": 9.652569139511595e-06, + "loss": 0.0, + "step": 20950 + }, + { + "epoch": 4.84, + "learning_rate": 9.650643247823742e-06, + "loss": 0.0, + "step": 20951 + }, + { + "epoch": 4.84, + "learning_rate": 9.648717356135892e-06, + "loss": 0.0, + "step": 20952 + }, + { + "epoch": 4.84, + "learning_rate": 9.646791464448039e-06, + "loss": 0.0, + "step": 20953 + }, + { + "epoch": 4.84, + "learning_rate": 9.644865572760188e-06, + "loss": 0.0, + "step": 20954 + }, + { + "epoch": 4.84, + "learning_rate": 9.642939681072337e-06, + "loss": 0.0, + "step": 20955 + }, + { + "epoch": 4.84, + "learning_rate": 9.641013789384487e-06, + "loss": 0.0, + "step": 20956 + }, + { + "epoch": 4.84, + "learning_rate": 9.639087897696634e-06, + "loss": 0.0, + "step": 20957 + }, + { + "epoch": 4.84, + "learning_rate": 9.637162006008781e-06, + "loss": 0.0, + "step": 20958 + }, + { + "epoch": 4.84, + "learning_rate": 9.63523611432093e-06, + "loss": 0.0, + "step": 20959 + }, + { + "epoch": 4.84, + "learning_rate": 9.63331022263308e-06, + "loss": 0.0, + "step": 20960 + }, + { + "epoch": 4.84, + "learning_rate": 9.631384330945229e-06, + "loss": 0.0, + "step": 20961 + }, + { + "epoch": 4.84, + "learning_rate": 9.629458439257376e-06, + "loss": 0.0, + "step": 20962 + }, + { + "epoch": 4.84, + "learning_rate": 9.627532547569524e-06, + "loss": 0.0, + "step": 20963 + }, + { + "epoch": 4.84, + "learning_rate": 9.625606655881673e-06, + "loss": 0.0, + "step": 20964 + }, + { + "epoch": 4.85, + "learning_rate": 9.623680764193822e-06, + "loss": 0.0, + "step": 20965 + }, + { + "epoch": 4.85, + "learning_rate": 9.621754872505971e-06, + "loss": 0.0, + "step": 20966 + }, + { + "epoch": 4.85, + "learning_rate": 9.619828980818119e-06, + "loss": 0.0, + "step": 20967 + }, + { + "epoch": 4.85, + "learning_rate": 9.617903089130268e-06, + "loss": 0.0, + "step": 20968 + }, + { + "epoch": 4.85, + "learning_rate": 9.615977197442416e-06, + "loss": 0.0, + "step": 20969 + }, + { + "epoch": 4.85, + "learning_rate": 9.614051305754565e-06, + "loss": 0.0, + "step": 20970 + }, + { + "epoch": 4.85, + "learning_rate": 9.612125414066714e-06, + "loss": 0.0, + "step": 20971 + }, + { + "epoch": 4.85, + "learning_rate": 9.610199522378861e-06, + "loss": 0.0, + "step": 20972 + }, + { + "epoch": 4.85, + "learning_rate": 9.60827363069101e-06, + "loss": 0.0, + "step": 20973 + }, + { + "epoch": 4.85, + "learning_rate": 9.60634773900316e-06, + "loss": 0.0, + "step": 20974 + }, + { + "epoch": 4.85, + "learning_rate": 9.604421847315307e-06, + "loss": 0.0, + "step": 20975 + }, + { + "epoch": 4.85, + "learning_rate": 9.602495955627456e-06, + "loss": 0.0, + "step": 20976 + }, + { + "epoch": 4.85, + "learning_rate": 9.600570063939604e-06, + "loss": 0.0, + "step": 20977 + }, + { + "epoch": 4.85, + "learning_rate": 9.598644172251753e-06, + "loss": 0.0, + "step": 20978 + }, + { + "epoch": 4.85, + "learning_rate": 9.596718280563902e-06, + "loss": 0.0, + "step": 20979 + }, + { + "epoch": 4.85, + "learning_rate": 9.59479238887605e-06, + "loss": 0.0, + "step": 20980 + }, + { + "epoch": 4.85, + "learning_rate": 9.592866497188197e-06, + "loss": 0.0, + "step": 20981 + }, + { + "epoch": 4.85, + "learning_rate": 9.590940605500346e-06, + "loss": 0.0, + "step": 20982 + }, + { + "epoch": 4.85, + "learning_rate": 9.589014713812496e-06, + "loss": 0.0, + "step": 20983 + }, + { + "epoch": 4.85, + "learning_rate": 9.587088822124645e-06, + "loss": 0.0, + "step": 20984 + }, + { + "epoch": 4.85, + "learning_rate": 9.585162930436794e-06, + "loss": 0.0, + "step": 20985 + }, + { + "epoch": 4.85, + "learning_rate": 9.583237038748941e-06, + "loss": 0.0, + "step": 20986 + }, + { + "epoch": 4.85, + "learning_rate": 9.581311147061089e-06, + "loss": 0.0, + "step": 20987 + }, + { + "epoch": 4.85, + "learning_rate": 9.579385255373238e-06, + "loss": 0.0, + "step": 20988 + }, + { + "epoch": 4.85, + "learning_rate": 9.577459363685387e-06, + "loss": 0.0, + "step": 20989 + }, + { + "epoch": 4.85, + "learning_rate": 9.575533471997536e-06, + "loss": 0.0, + "step": 20990 + }, + { + "epoch": 4.85, + "learning_rate": 9.573607580309684e-06, + "loss": 0.0, + "step": 20991 + }, + { + "epoch": 4.85, + "learning_rate": 9.571681688621833e-06, + "loss": 0.0, + "step": 20992 + }, + { + "epoch": 4.85, + "learning_rate": 9.56975579693398e-06, + "loss": 0.0, + "step": 20993 + }, + { + "epoch": 4.85, + "learning_rate": 9.56782990524613e-06, + "loss": 0.0, + "step": 20994 + }, + { + "epoch": 4.85, + "learning_rate": 9.565904013558277e-06, + "loss": 0.0, + "step": 20995 + }, + { + "epoch": 4.85, + "learning_rate": 9.563978121870426e-06, + "loss": 0.0, + "step": 20996 + }, + { + "epoch": 4.85, + "learning_rate": 9.562052230182575e-06, + "loss": 0.0, + "step": 20997 + }, + { + "epoch": 4.85, + "learning_rate": 9.560126338494723e-06, + "loss": 0.0, + "step": 20998 + }, + { + "epoch": 4.85, + "learning_rate": 9.558200446806872e-06, + "loss": 0.0, + "step": 20999 + }, + { + "epoch": 4.85, + "learning_rate": 9.55627455511902e-06, + "loss": 0.0, + "step": 21000 + } + ], + "max_steps": 25962, + "num_train_epochs": 6, + "total_flos": 5.350753132904448e+16, + "trial_name": null, + "trial_params": null +}