{ "best_metric": null, "best_model_checkpoint": null, "epoch": 0.9999543982853756, "eval_steps": 500, "global_step": 8223, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 8.097165991902835e-08, "loss": 1.4174, "step": 1 }, { "epoch": 0.0, "learning_rate": 1.619433198380567e-07, "loss": 1.3504, "step": 2 }, { "epoch": 0.0, "learning_rate": 2.4291497975708504e-07, "loss": 1.3096, "step": 3 }, { "epoch": 0.0, "learning_rate": 3.238866396761134e-07, "loss": 1.3152, "step": 4 }, { "epoch": 0.0, "learning_rate": 4.048582995951417e-07, "loss": 1.3475, "step": 5 }, { "epoch": 0.0, "learning_rate": 4.858299595141701e-07, "loss": 1.3744, "step": 6 }, { "epoch": 0.0, "learning_rate": 5.668016194331984e-07, "loss": 1.2711, "step": 7 }, { "epoch": 0.0, "learning_rate": 6.477732793522268e-07, "loss": 1.3105, "step": 8 }, { "epoch": 0.0, "learning_rate": 7.287449392712551e-07, "loss": 1.3, "step": 9 }, { "epoch": 0.0, "learning_rate": 8.097165991902834e-07, "loss": 1.2946, "step": 10 }, { "epoch": 0.0, "learning_rate": 8.906882591093118e-07, "loss": 1.2633, "step": 11 }, { "epoch": 0.0, "learning_rate": 9.716599190283402e-07, "loss": 1.2974, "step": 12 }, { "epoch": 0.0, "learning_rate": 1.0526315789473685e-06, "loss": 1.2593, "step": 13 }, { "epoch": 0.0, "learning_rate": 1.133603238866397e-06, "loss": 1.2955, "step": 14 }, { "epoch": 0.0, "learning_rate": 1.214574898785425e-06, "loss": 1.1748, "step": 15 }, { "epoch": 0.0, "learning_rate": 1.2955465587044536e-06, "loss": 1.189, "step": 16 }, { "epoch": 0.0, "learning_rate": 1.3765182186234818e-06, "loss": 1.1961, "step": 17 }, { "epoch": 0.0, "learning_rate": 1.4574898785425101e-06, "loss": 1.188, "step": 18 }, { "epoch": 0.0, "learning_rate": 1.5384615384615387e-06, "loss": 1.2439, "step": 19 }, { "epoch": 0.0, "learning_rate": 1.6194331983805669e-06, "loss": 1.0984, "step": 20 }, { "epoch": 0.0, "learning_rate": 1.7004048582995952e-06, "loss": 1.1908, "step": 21 }, { "epoch": 0.0, "learning_rate": 1.7813765182186236e-06, "loss": 1.2113, "step": 22 }, { "epoch": 0.0, "learning_rate": 1.862348178137652e-06, "loss": 1.1703, "step": 23 }, { "epoch": 0.0, "learning_rate": 1.9433198380566803e-06, "loss": 1.1799, "step": 24 }, { "epoch": 0.0, "learning_rate": 2.0242914979757085e-06, "loss": 1.1877, "step": 25 }, { "epoch": 0.0, "learning_rate": 2.105263157894737e-06, "loss": 1.2464, "step": 26 }, { "epoch": 0.0, "learning_rate": 2.1862348178137656e-06, "loss": 1.0845, "step": 27 }, { "epoch": 0.0, "learning_rate": 2.267206477732794e-06, "loss": 1.1661, "step": 28 }, { "epoch": 0.0, "learning_rate": 2.348178137651822e-06, "loss": 1.1588, "step": 29 }, { "epoch": 0.0, "learning_rate": 2.42914979757085e-06, "loss": 1.1407, "step": 30 }, { "epoch": 0.0, "learning_rate": 2.5101214574898787e-06, "loss": 1.1641, "step": 31 }, { "epoch": 0.0, "learning_rate": 2.5910931174089072e-06, "loss": 1.0864, "step": 32 }, { "epoch": 0.0, "learning_rate": 2.672064777327936e-06, "loss": 1.1641, "step": 33 }, { "epoch": 0.0, "learning_rate": 2.7530364372469636e-06, "loss": 1.2039, "step": 34 }, { "epoch": 0.0, "learning_rate": 2.834008097165992e-06, "loss": 1.1002, "step": 35 }, { "epoch": 0.0, "learning_rate": 2.9149797570850203e-06, "loss": 1.0793, "step": 36 }, { "epoch": 0.0, "learning_rate": 2.995951417004049e-06, "loss": 1.2094, "step": 37 }, { "epoch": 0.0, "learning_rate": 3.0769230769230774e-06, "loss": 1.1693, "step": 38 }, { "epoch": 0.0, "learning_rate": 3.157894736842105e-06, "loss": 1.1509, "step": 39 }, { "epoch": 0.0, "learning_rate": 3.2388663967611337e-06, "loss": 1.0473, "step": 40 }, { "epoch": 0.0, "learning_rate": 3.3198380566801623e-06, "loss": 1.1329, "step": 41 }, { "epoch": 0.01, "learning_rate": 3.4008097165991905e-06, "loss": 1.0582, "step": 42 }, { "epoch": 0.01, "learning_rate": 3.481781376518219e-06, "loss": 1.0611, "step": 43 }, { "epoch": 0.01, "learning_rate": 3.562753036437247e-06, "loss": 1.1356, "step": 44 }, { "epoch": 0.01, "learning_rate": 3.6437246963562758e-06, "loss": 1.1055, "step": 45 }, { "epoch": 0.01, "learning_rate": 3.724696356275304e-06, "loss": 1.0601, "step": 46 }, { "epoch": 0.01, "learning_rate": 3.805668016194332e-06, "loss": 1.0448, "step": 47 }, { "epoch": 0.01, "learning_rate": 3.886639676113361e-06, "loss": 1.0654, "step": 48 }, { "epoch": 0.01, "learning_rate": 3.967611336032389e-06, "loss": 1.1115, "step": 49 }, { "epoch": 0.01, "learning_rate": 4.048582995951417e-06, "loss": 1.1049, "step": 50 }, { "epoch": 0.01, "learning_rate": 4.1295546558704455e-06, "loss": 1.0875, "step": 51 }, { "epoch": 0.01, "learning_rate": 4.210526315789474e-06, "loss": 1.0989, "step": 52 }, { "epoch": 0.01, "learning_rate": 4.291497975708503e-06, "loss": 1.0816, "step": 53 }, { "epoch": 0.01, "learning_rate": 4.372469635627531e-06, "loss": 1.0905, "step": 54 }, { "epoch": 0.01, "learning_rate": 4.453441295546559e-06, "loss": 1.0544, "step": 55 }, { "epoch": 0.01, "learning_rate": 4.534412955465588e-06, "loss": 1.0774, "step": 56 }, { "epoch": 0.01, "learning_rate": 4.615384615384616e-06, "loss": 1.0942, "step": 57 }, { "epoch": 0.01, "learning_rate": 4.696356275303644e-06, "loss": 1.0916, "step": 58 }, { "epoch": 0.01, "learning_rate": 4.7773279352226725e-06, "loss": 1.032, "step": 59 }, { "epoch": 0.01, "learning_rate": 4.8582995951417e-06, "loss": 0.9745, "step": 60 }, { "epoch": 0.01, "learning_rate": 4.939271255060729e-06, "loss": 1.0095, "step": 61 }, { "epoch": 0.01, "learning_rate": 5.020242914979757e-06, "loss": 1.0838, "step": 62 }, { "epoch": 0.01, "learning_rate": 5.101214574898786e-06, "loss": 1.0952, "step": 63 }, { "epoch": 0.01, "learning_rate": 5.1821862348178145e-06, "loss": 0.9778, "step": 64 }, { "epoch": 0.01, "learning_rate": 5.263157894736842e-06, "loss": 1.0018, "step": 65 }, { "epoch": 0.01, "learning_rate": 5.344129554655872e-06, "loss": 1.0088, "step": 66 }, { "epoch": 0.01, "learning_rate": 5.425101214574899e-06, "loss": 1.0437, "step": 67 }, { "epoch": 0.01, "learning_rate": 5.506072874493927e-06, "loss": 0.9948, "step": 68 }, { "epoch": 0.01, "learning_rate": 5.5870445344129565e-06, "loss": 1.0146, "step": 69 }, { "epoch": 0.01, "learning_rate": 5.668016194331984e-06, "loss": 1.1239, "step": 70 }, { "epoch": 0.01, "learning_rate": 5.748987854251013e-06, "loss": 1.0571, "step": 71 }, { "epoch": 0.01, "learning_rate": 5.8299595141700406e-06, "loss": 1.0255, "step": 72 }, { "epoch": 0.01, "learning_rate": 5.91093117408907e-06, "loss": 1.0425, "step": 73 }, { "epoch": 0.01, "learning_rate": 5.991902834008098e-06, "loss": 1.0134, "step": 74 }, { "epoch": 0.01, "learning_rate": 6.0728744939271254e-06, "loss": 1.0243, "step": 75 }, { "epoch": 0.01, "learning_rate": 6.153846153846155e-06, "loss": 1.07, "step": 76 }, { "epoch": 0.01, "learning_rate": 6.234817813765183e-06, "loss": 1.0681, "step": 77 }, { "epoch": 0.01, "learning_rate": 6.31578947368421e-06, "loss": 0.9945, "step": 78 }, { "epoch": 0.01, "learning_rate": 6.39676113360324e-06, "loss": 0.9479, "step": 79 }, { "epoch": 0.01, "learning_rate": 6.4777327935222675e-06, "loss": 1.0306, "step": 80 }, { "epoch": 0.01, "learning_rate": 6.558704453441296e-06, "loss": 1.1059, "step": 81 }, { "epoch": 0.01, "learning_rate": 6.639676113360325e-06, "loss": 0.9937, "step": 82 }, { "epoch": 0.01, "learning_rate": 6.720647773279353e-06, "loss": 0.9591, "step": 83 }, { "epoch": 0.01, "learning_rate": 6.801619433198381e-06, "loss": 0.993, "step": 84 }, { "epoch": 0.01, "learning_rate": 6.882591093117409e-06, "loss": 1.0419, "step": 85 }, { "epoch": 0.01, "learning_rate": 6.963562753036438e-06, "loss": 0.9818, "step": 86 }, { "epoch": 0.01, "learning_rate": 7.044534412955466e-06, "loss": 0.9987, "step": 87 }, { "epoch": 0.01, "learning_rate": 7.125506072874494e-06, "loss": 0.9047, "step": 88 }, { "epoch": 0.01, "learning_rate": 7.206477732793523e-06, "loss": 1.015, "step": 89 }, { "epoch": 0.01, "learning_rate": 7.2874493927125516e-06, "loss": 1.0272, "step": 90 }, { "epoch": 0.01, "learning_rate": 7.368421052631579e-06, "loss": 0.9604, "step": 91 }, { "epoch": 0.01, "learning_rate": 7.449392712550608e-06, "loss": 0.9399, "step": 92 }, { "epoch": 0.01, "learning_rate": 7.5303643724696364e-06, "loss": 1.0039, "step": 93 }, { "epoch": 0.01, "learning_rate": 7.611336032388664e-06, "loss": 1.0301, "step": 94 }, { "epoch": 0.01, "learning_rate": 7.692307692307694e-06, "loss": 1.0617, "step": 95 }, { "epoch": 0.01, "learning_rate": 7.773279352226721e-06, "loss": 1.0146, "step": 96 }, { "epoch": 0.01, "learning_rate": 7.854251012145749e-06, "loss": 0.9946, "step": 97 }, { "epoch": 0.01, "learning_rate": 7.935222672064778e-06, "loss": 1.0025, "step": 98 }, { "epoch": 0.01, "learning_rate": 8.016194331983806e-06, "loss": 1.0198, "step": 99 }, { "epoch": 0.01, "learning_rate": 8.097165991902834e-06, "loss": 0.9776, "step": 100 }, { "epoch": 0.01, "learning_rate": 8.178137651821862e-06, "loss": 0.987, "step": 101 }, { "epoch": 0.01, "learning_rate": 8.259109311740891e-06, "loss": 1.0206, "step": 102 }, { "epoch": 0.01, "learning_rate": 8.340080971659919e-06, "loss": 0.9864, "step": 103 }, { "epoch": 0.01, "learning_rate": 8.421052631578948e-06, "loss": 0.9756, "step": 104 }, { "epoch": 0.01, "learning_rate": 8.502024291497976e-06, "loss": 0.9213, "step": 105 }, { "epoch": 0.01, "learning_rate": 8.582995951417005e-06, "loss": 1.0238, "step": 106 }, { "epoch": 0.01, "learning_rate": 8.663967611336033e-06, "loss": 0.9609, "step": 107 }, { "epoch": 0.01, "learning_rate": 8.744939271255063e-06, "loss": 0.909, "step": 108 }, { "epoch": 0.01, "learning_rate": 8.82591093117409e-06, "loss": 0.9367, "step": 109 }, { "epoch": 0.01, "learning_rate": 8.906882591093118e-06, "loss": 1.0139, "step": 110 }, { "epoch": 0.01, "learning_rate": 8.987854251012147e-06, "loss": 0.9523, "step": 111 }, { "epoch": 0.01, "learning_rate": 9.068825910931175e-06, "loss": 0.8716, "step": 112 }, { "epoch": 0.01, "learning_rate": 9.149797570850203e-06, "loss": 0.9962, "step": 113 }, { "epoch": 0.01, "learning_rate": 9.230769230769232e-06, "loss": 0.9658, "step": 114 }, { "epoch": 0.01, "learning_rate": 9.31174089068826e-06, "loss": 0.9607, "step": 115 }, { "epoch": 0.01, "learning_rate": 9.392712550607288e-06, "loss": 0.9759, "step": 116 }, { "epoch": 0.01, "learning_rate": 9.473684210526315e-06, "loss": 0.9662, "step": 117 }, { "epoch": 0.01, "learning_rate": 9.554655870445345e-06, "loss": 0.9583, "step": 118 }, { "epoch": 0.01, "learning_rate": 9.635627530364373e-06, "loss": 0.9411, "step": 119 }, { "epoch": 0.01, "learning_rate": 9.7165991902834e-06, "loss": 0.9402, "step": 120 }, { "epoch": 0.01, "learning_rate": 9.79757085020243e-06, "loss": 0.9395, "step": 121 }, { "epoch": 0.01, "learning_rate": 9.878542510121458e-06, "loss": 0.897, "step": 122 }, { "epoch": 0.01, "learning_rate": 9.959514170040487e-06, "loss": 0.9025, "step": 123 }, { "epoch": 0.02, "learning_rate": 1.0040485829959515e-05, "loss": 0.8705, "step": 124 }, { "epoch": 0.02, "learning_rate": 1.0121457489878544e-05, "loss": 0.9021, "step": 125 }, { "epoch": 0.02, "learning_rate": 1.0202429149797572e-05, "loss": 0.9484, "step": 126 }, { "epoch": 0.02, "learning_rate": 1.02834008097166e-05, "loss": 0.9124, "step": 127 }, { "epoch": 0.02, "learning_rate": 1.0364372469635629e-05, "loss": 0.9555, "step": 128 }, { "epoch": 0.02, "learning_rate": 1.0445344129554658e-05, "loss": 0.8857, "step": 129 }, { "epoch": 0.02, "learning_rate": 1.0526315789473684e-05, "loss": 0.9616, "step": 130 }, { "epoch": 0.02, "learning_rate": 1.0607287449392714e-05, "loss": 0.9658, "step": 131 }, { "epoch": 0.02, "learning_rate": 1.0688259109311743e-05, "loss": 0.9294, "step": 132 }, { "epoch": 0.02, "learning_rate": 1.076923076923077e-05, "loss": 0.8978, "step": 133 }, { "epoch": 0.02, "learning_rate": 1.0850202429149799e-05, "loss": 0.9691, "step": 134 }, { "epoch": 0.02, "learning_rate": 1.0931174089068828e-05, "loss": 0.8902, "step": 135 }, { "epoch": 0.02, "learning_rate": 1.1012145748987854e-05, "loss": 0.9382, "step": 136 }, { "epoch": 0.02, "learning_rate": 1.1093117408906884e-05, "loss": 0.9961, "step": 137 }, { "epoch": 0.02, "learning_rate": 1.1174089068825913e-05, "loss": 0.9071, "step": 138 }, { "epoch": 0.02, "learning_rate": 1.1255060728744939e-05, "loss": 0.9573, "step": 139 }, { "epoch": 0.02, "learning_rate": 1.1336032388663969e-05, "loss": 0.9486, "step": 140 }, { "epoch": 0.02, "learning_rate": 1.1417004048582996e-05, "loss": 0.9131, "step": 141 }, { "epoch": 0.02, "learning_rate": 1.1497975708502026e-05, "loss": 0.9993, "step": 142 }, { "epoch": 0.02, "learning_rate": 1.1578947368421053e-05, "loss": 0.9482, "step": 143 }, { "epoch": 0.02, "learning_rate": 1.1659919028340081e-05, "loss": 0.881, "step": 144 }, { "epoch": 0.02, "learning_rate": 1.174089068825911e-05, "loss": 0.9403, "step": 145 }, { "epoch": 0.02, "learning_rate": 1.182186234817814e-05, "loss": 0.9189, "step": 146 }, { "epoch": 0.02, "learning_rate": 1.1902834008097166e-05, "loss": 1.0039, "step": 147 }, { "epoch": 0.02, "learning_rate": 1.1983805668016195e-05, "loss": 0.9288, "step": 148 }, { "epoch": 0.02, "learning_rate": 1.2064777327935225e-05, "loss": 0.8976, "step": 149 }, { "epoch": 0.02, "learning_rate": 1.2145748987854251e-05, "loss": 0.9111, "step": 150 }, { "epoch": 0.02, "learning_rate": 1.222672064777328e-05, "loss": 0.8571, "step": 151 }, { "epoch": 0.02, "learning_rate": 1.230769230769231e-05, "loss": 0.9883, "step": 152 }, { "epoch": 0.02, "learning_rate": 1.2388663967611336e-05, "loss": 0.93, "step": 153 }, { "epoch": 0.02, "learning_rate": 1.2469635627530365e-05, "loss": 0.8772, "step": 154 }, { "epoch": 0.02, "learning_rate": 1.2550607287449395e-05, "loss": 0.8855, "step": 155 }, { "epoch": 0.02, "learning_rate": 1.263157894736842e-05, "loss": 0.8718, "step": 156 }, { "epoch": 0.02, "learning_rate": 1.271255060728745e-05, "loss": 0.8557, "step": 157 }, { "epoch": 0.02, "learning_rate": 1.279352226720648e-05, "loss": 0.9323, "step": 158 }, { "epoch": 0.02, "learning_rate": 1.2874493927125507e-05, "loss": 0.8406, "step": 159 }, { "epoch": 0.02, "learning_rate": 1.2955465587044535e-05, "loss": 0.9287, "step": 160 }, { "epoch": 0.02, "learning_rate": 1.3036437246963564e-05, "loss": 0.891, "step": 161 }, { "epoch": 0.02, "learning_rate": 1.3117408906882592e-05, "loss": 1.0125, "step": 162 }, { "epoch": 0.02, "learning_rate": 1.3198380566801622e-05, "loss": 0.9397, "step": 163 }, { "epoch": 0.02, "learning_rate": 1.327935222672065e-05, "loss": 0.9045, "step": 164 }, { "epoch": 0.02, "learning_rate": 1.3360323886639677e-05, "loss": 0.8787, "step": 165 }, { "epoch": 0.02, "learning_rate": 1.3441295546558706e-05, "loss": 0.87, "step": 166 }, { "epoch": 0.02, "learning_rate": 1.3522267206477734e-05, "loss": 0.9395, "step": 167 }, { "epoch": 0.02, "learning_rate": 1.3603238866396762e-05, "loss": 0.9178, "step": 168 }, { "epoch": 0.02, "learning_rate": 1.3684210526315791e-05, "loss": 0.8713, "step": 169 }, { "epoch": 0.02, "learning_rate": 1.3765182186234817e-05, "loss": 0.9058, "step": 170 }, { "epoch": 0.02, "learning_rate": 1.3846153846153847e-05, "loss": 0.8679, "step": 171 }, { "epoch": 0.02, "learning_rate": 1.3927125506072876e-05, "loss": 0.9839, "step": 172 }, { "epoch": 0.02, "learning_rate": 1.4008097165991902e-05, "loss": 0.911, "step": 173 }, { "epoch": 0.02, "learning_rate": 1.4089068825910932e-05, "loss": 0.9285, "step": 174 }, { "epoch": 0.02, "learning_rate": 1.4170040485829961e-05, "loss": 0.8929, "step": 175 }, { "epoch": 0.02, "learning_rate": 1.4251012145748989e-05, "loss": 0.9524, "step": 176 }, { "epoch": 0.02, "learning_rate": 1.4331983805668017e-05, "loss": 0.8386, "step": 177 }, { "epoch": 0.02, "learning_rate": 1.4412955465587046e-05, "loss": 0.91, "step": 178 }, { "epoch": 0.02, "learning_rate": 1.4493927125506074e-05, "loss": 0.9068, "step": 179 }, { "epoch": 0.02, "learning_rate": 1.4574898785425103e-05, "loss": 0.8656, "step": 180 }, { "epoch": 0.02, "learning_rate": 1.465587044534413e-05, "loss": 0.9077, "step": 181 }, { "epoch": 0.02, "learning_rate": 1.4736842105263159e-05, "loss": 0.9116, "step": 182 }, { "epoch": 0.02, "learning_rate": 1.4817813765182188e-05, "loss": 0.9028, "step": 183 }, { "epoch": 0.02, "learning_rate": 1.4898785425101216e-05, "loss": 0.8775, "step": 184 }, { "epoch": 0.02, "learning_rate": 1.4979757085020243e-05, "loss": 0.9135, "step": 185 }, { "epoch": 0.02, "learning_rate": 1.5060728744939273e-05, "loss": 0.9077, "step": 186 }, { "epoch": 0.02, "learning_rate": 1.5141700404858302e-05, "loss": 0.9027, "step": 187 }, { "epoch": 0.02, "learning_rate": 1.5222672064777328e-05, "loss": 0.9058, "step": 188 }, { "epoch": 0.02, "learning_rate": 1.5303643724696356e-05, "loss": 0.8925, "step": 189 }, { "epoch": 0.02, "learning_rate": 1.5384615384615387e-05, "loss": 0.8957, "step": 190 }, { "epoch": 0.02, "learning_rate": 1.5465587044534415e-05, "loss": 0.9647, "step": 191 }, { "epoch": 0.02, "learning_rate": 1.5546558704453443e-05, "loss": 0.8421, "step": 192 }, { "epoch": 0.02, "learning_rate": 1.562753036437247e-05, "loss": 0.9678, "step": 193 }, { "epoch": 0.02, "learning_rate": 1.5708502024291498e-05, "loss": 0.8504, "step": 194 }, { "epoch": 0.02, "learning_rate": 1.578947368421053e-05, "loss": 0.9409, "step": 195 }, { "epoch": 0.02, "learning_rate": 1.5870445344129557e-05, "loss": 0.9358, "step": 196 }, { "epoch": 0.02, "learning_rate": 1.5951417004048585e-05, "loss": 0.9112, "step": 197 }, { "epoch": 0.02, "learning_rate": 1.6032388663967612e-05, "loss": 0.835, "step": 198 }, { "epoch": 0.02, "learning_rate": 1.6113360323886644e-05, "loss": 0.8832, "step": 199 }, { "epoch": 0.02, "learning_rate": 1.6194331983805668e-05, "loss": 0.8993, "step": 200 }, { "epoch": 0.02, "learning_rate": 1.62753036437247e-05, "loss": 0.9403, "step": 201 }, { "epoch": 0.02, "learning_rate": 1.6356275303643723e-05, "loss": 0.9026, "step": 202 }, { "epoch": 0.02, "learning_rate": 1.6437246963562754e-05, "loss": 0.8608, "step": 203 }, { "epoch": 0.02, "learning_rate": 1.6518218623481782e-05, "loss": 0.9153, "step": 204 }, { "epoch": 0.02, "learning_rate": 1.659919028340081e-05, "loss": 0.8355, "step": 205 }, { "epoch": 0.03, "learning_rate": 1.6680161943319838e-05, "loss": 0.8875, "step": 206 }, { "epoch": 0.03, "learning_rate": 1.676113360323887e-05, "loss": 0.8947, "step": 207 }, { "epoch": 0.03, "learning_rate": 1.6842105263157896e-05, "loss": 0.8666, "step": 208 }, { "epoch": 0.03, "learning_rate": 1.6923076923076924e-05, "loss": 0.8938, "step": 209 }, { "epoch": 0.03, "learning_rate": 1.7004048582995952e-05, "loss": 0.8616, "step": 210 }, { "epoch": 0.03, "learning_rate": 1.708502024291498e-05, "loss": 0.864, "step": 211 }, { "epoch": 0.03, "learning_rate": 1.716599190283401e-05, "loss": 0.792, "step": 212 }, { "epoch": 0.03, "learning_rate": 1.724696356275304e-05, "loss": 0.8408, "step": 213 }, { "epoch": 0.03, "learning_rate": 1.7327935222672066e-05, "loss": 0.8458, "step": 214 }, { "epoch": 0.03, "learning_rate": 1.7408906882591094e-05, "loss": 0.8747, "step": 215 }, { "epoch": 0.03, "learning_rate": 1.7489878542510125e-05, "loss": 0.9092, "step": 216 }, { "epoch": 0.03, "learning_rate": 1.757085020242915e-05, "loss": 0.9721, "step": 217 }, { "epoch": 0.03, "learning_rate": 1.765182186234818e-05, "loss": 0.9331, "step": 218 }, { "epoch": 0.03, "learning_rate": 1.7732793522267208e-05, "loss": 0.8436, "step": 219 }, { "epoch": 0.03, "learning_rate": 1.7813765182186236e-05, "loss": 0.9152, "step": 220 }, { "epoch": 0.03, "learning_rate": 1.7894736842105264e-05, "loss": 0.8917, "step": 221 }, { "epoch": 0.03, "learning_rate": 1.7975708502024295e-05, "loss": 0.9152, "step": 222 }, { "epoch": 0.03, "learning_rate": 1.805668016194332e-05, "loss": 0.8359, "step": 223 }, { "epoch": 0.03, "learning_rate": 1.813765182186235e-05, "loss": 0.8796, "step": 224 }, { "epoch": 0.03, "learning_rate": 1.8218623481781378e-05, "loss": 0.8914, "step": 225 }, { "epoch": 0.03, "learning_rate": 1.8299595141700406e-05, "loss": 0.8745, "step": 226 }, { "epoch": 0.03, "learning_rate": 1.8380566801619433e-05, "loss": 0.8549, "step": 227 }, { "epoch": 0.03, "learning_rate": 1.8461538461538465e-05, "loss": 0.8482, "step": 228 }, { "epoch": 0.03, "learning_rate": 1.8542510121457492e-05, "loss": 0.8851, "step": 229 }, { "epoch": 0.03, "learning_rate": 1.862348178137652e-05, "loss": 0.8435, "step": 230 }, { "epoch": 0.03, "learning_rate": 1.8704453441295548e-05, "loss": 0.8479, "step": 231 }, { "epoch": 0.03, "learning_rate": 1.8785425101214576e-05, "loss": 0.8758, "step": 232 }, { "epoch": 0.03, "learning_rate": 1.8866396761133607e-05, "loss": 0.8178, "step": 233 }, { "epoch": 0.03, "learning_rate": 1.894736842105263e-05, "loss": 0.7867, "step": 234 }, { "epoch": 0.03, "learning_rate": 1.9028340080971662e-05, "loss": 0.8478, "step": 235 }, { "epoch": 0.03, "learning_rate": 1.910931174089069e-05, "loss": 0.8745, "step": 236 }, { "epoch": 0.03, "learning_rate": 1.9190283400809718e-05, "loss": 0.8898, "step": 237 }, { "epoch": 0.03, "learning_rate": 1.9271255060728745e-05, "loss": 0.8362, "step": 238 }, { "epoch": 0.03, "learning_rate": 1.9352226720647776e-05, "loss": 0.7913, "step": 239 }, { "epoch": 0.03, "learning_rate": 1.94331983805668e-05, "loss": 0.7836, "step": 240 }, { "epoch": 0.03, "learning_rate": 1.9514170040485832e-05, "loss": 0.8927, "step": 241 }, { "epoch": 0.03, "learning_rate": 1.959514170040486e-05, "loss": 0.8241, "step": 242 }, { "epoch": 0.03, "learning_rate": 1.9676113360323887e-05, "loss": 0.8405, "step": 243 }, { "epoch": 0.03, "learning_rate": 1.9757085020242915e-05, "loss": 0.8984, "step": 244 }, { "epoch": 0.03, "learning_rate": 1.9838056680161946e-05, "loss": 0.8525, "step": 245 }, { "epoch": 0.03, "learning_rate": 1.9919028340080974e-05, "loss": 0.8313, "step": 246 }, { "epoch": 0.03, "learning_rate": 2e-05, "loss": 0.8707, "step": 247 }, { "epoch": 0.03, "learning_rate": 1.9999999224289888e-05, "loss": 0.8594, "step": 248 }, { "epoch": 0.03, "learning_rate": 1.999999689715967e-05, "loss": 0.882, "step": 249 }, { "epoch": 0.03, "learning_rate": 1.9999993018609706e-05, "loss": 0.8377, "step": 250 }, { "epoch": 0.03, "learning_rate": 1.99999875886406e-05, "loss": 0.7867, "step": 251 }, { "epoch": 0.03, "learning_rate": 1.999998060725319e-05, "loss": 0.8659, "step": 252 }, { "epoch": 0.03, "learning_rate": 1.9999972074448567e-05, "loss": 0.8576, "step": 253 }, { "epoch": 0.03, "learning_rate": 1.9999961990228046e-05, "loss": 0.8453, "step": 254 }, { "epoch": 0.03, "learning_rate": 1.9999950354593198e-05, "loss": 0.8371, "step": 255 }, { "epoch": 0.03, "learning_rate": 1.9999937167545823e-05, "loss": 0.8591, "step": 256 }, { "epoch": 0.03, "learning_rate": 1.999992242908797e-05, "loss": 0.85, "step": 257 }, { "epoch": 0.03, "learning_rate": 1.9999906139221928e-05, "loss": 0.824, "step": 258 }, { "epoch": 0.03, "learning_rate": 1.9999888297950218e-05, "loss": 0.8125, "step": 259 }, { "epoch": 0.03, "learning_rate": 1.9999868905275614e-05, "loss": 0.8701, "step": 260 }, { "epoch": 0.03, "learning_rate": 1.9999847961201123e-05, "loss": 0.8292, "step": 261 }, { "epoch": 0.03, "learning_rate": 1.9999825465729993e-05, "loss": 0.7963, "step": 262 }, { "epoch": 0.03, "learning_rate": 1.999980141886571e-05, "loss": 0.8452, "step": 263 }, { "epoch": 0.03, "learning_rate": 1.9999775820612014e-05, "loss": 0.8355, "step": 264 }, { "epoch": 0.03, "learning_rate": 1.999974867097287e-05, "loss": 0.8379, "step": 265 }, { "epoch": 0.03, "learning_rate": 1.9999719969952488e-05, "loss": 0.8732, "step": 266 }, { "epoch": 0.03, "learning_rate": 1.999968971755533e-05, "loss": 0.9029, "step": 267 }, { "epoch": 0.03, "learning_rate": 1.999965791378608e-05, "loss": 0.8918, "step": 268 }, { "epoch": 0.03, "learning_rate": 1.9999624558649674e-05, "loss": 0.8039, "step": 269 }, { "epoch": 0.03, "learning_rate": 1.9999589652151294e-05, "loss": 0.8199, "step": 270 }, { "epoch": 0.03, "learning_rate": 1.9999553194296343e-05, "loss": 0.8143, "step": 271 }, { "epoch": 0.03, "learning_rate": 1.9999515185090487e-05, "loss": 0.8757, "step": 272 }, { "epoch": 0.03, "learning_rate": 1.999947562453962e-05, "loss": 0.7519, "step": 273 }, { "epoch": 0.03, "learning_rate": 1.999943451264988e-05, "loss": 0.8694, "step": 274 }, { "epoch": 0.03, "learning_rate": 1.9999391849427644e-05, "loss": 0.816, "step": 275 }, { "epoch": 0.03, "learning_rate": 1.999934763487953e-05, "loss": 0.8926, "step": 276 }, { "epoch": 0.03, "learning_rate": 1.9999301869012405e-05, "loss": 0.8326, "step": 277 }, { "epoch": 0.03, "learning_rate": 1.9999254551833357e-05, "loss": 0.8686, "step": 278 }, { "epoch": 0.03, "learning_rate": 1.9999205683349733e-05, "loss": 0.8057, "step": 279 }, { "epoch": 0.03, "learning_rate": 1.9999155263569118e-05, "loss": 0.8287, "step": 280 }, { "epoch": 0.03, "learning_rate": 1.9999103292499326e-05, "loss": 0.9013, "step": 281 }, { "epoch": 0.03, "learning_rate": 1.9999049770148427e-05, "loss": 0.7721, "step": 282 }, { "epoch": 0.03, "learning_rate": 1.9998994696524726e-05, "loss": 0.8195, "step": 283 }, { "epoch": 0.03, "learning_rate": 1.999893807163676e-05, "loss": 0.8703, "step": 284 }, { "epoch": 0.03, "learning_rate": 1.9998879895493317e-05, "loss": 0.8147, "step": 285 }, { "epoch": 0.03, "learning_rate": 1.9998820168103426e-05, "loss": 0.8285, "step": 286 }, { "epoch": 0.03, "learning_rate": 1.9998758889476346e-05, "loss": 0.7439, "step": 287 }, { "epoch": 0.04, "learning_rate": 1.9998696059621593e-05, "loss": 0.8664, "step": 288 }, { "epoch": 0.04, "learning_rate": 1.9998631678548905e-05, "loss": 0.8374, "step": 289 }, { "epoch": 0.04, "learning_rate": 1.9998565746268278e-05, "loss": 0.8136, "step": 290 }, { "epoch": 0.04, "learning_rate": 1.9998498262789935e-05, "loss": 0.8649, "step": 291 }, { "epoch": 0.04, "learning_rate": 1.999842922812435e-05, "loss": 0.8222, "step": 292 }, { "epoch": 0.04, "learning_rate": 1.999835864228223e-05, "loss": 0.9027, "step": 293 }, { "epoch": 0.04, "learning_rate": 1.999828650527453e-05, "loss": 0.8327, "step": 294 }, { "epoch": 0.04, "learning_rate": 1.9998212817112437e-05, "loss": 0.8357, "step": 295 }, { "epoch": 0.04, "learning_rate": 1.9998137577807383e-05, "loss": 0.842, "step": 296 }, { "epoch": 0.04, "learning_rate": 1.9998060787371045e-05, "loss": 0.794, "step": 297 }, { "epoch": 0.04, "learning_rate": 1.9997982445815332e-05, "loss": 0.8327, "step": 298 }, { "epoch": 0.04, "learning_rate": 1.9997902553152406e-05, "loss": 0.7831, "step": 299 }, { "epoch": 0.04, "learning_rate": 1.9997821109394652e-05, "loss": 0.7906, "step": 300 }, { "epoch": 0.04, "learning_rate": 1.9997738114554705e-05, "loss": 0.7795, "step": 301 }, { "epoch": 0.04, "learning_rate": 1.999765356864545e-05, "loss": 0.8881, "step": 302 }, { "epoch": 0.04, "learning_rate": 1.9997567471679996e-05, "loss": 0.7869, "step": 303 }, { "epoch": 0.04, "learning_rate": 1.9997479823671706e-05, "loss": 0.8197, "step": 304 }, { "epoch": 0.04, "learning_rate": 1.9997390624634173e-05, "loss": 0.8274, "step": 305 }, { "epoch": 0.04, "learning_rate": 1.999729987458124e-05, "loss": 0.7367, "step": 306 }, { "epoch": 0.04, "learning_rate": 1.999720757352698e-05, "loss": 0.7844, "step": 307 }, { "epoch": 0.04, "learning_rate": 1.9997113721485718e-05, "loss": 0.8504, "step": 308 }, { "epoch": 0.04, "learning_rate": 1.9997018318472016e-05, "loss": 0.7747, "step": 309 }, { "epoch": 0.04, "learning_rate": 1.9996921364500665e-05, "loss": 0.7639, "step": 310 }, { "epoch": 0.04, "learning_rate": 1.999682285958672e-05, "loss": 0.7689, "step": 311 }, { "epoch": 0.04, "learning_rate": 1.9996722803745455e-05, "loss": 0.8802, "step": 312 }, { "epoch": 0.04, "learning_rate": 1.999662119699239e-05, "loss": 0.805, "step": 313 }, { "epoch": 0.04, "learning_rate": 1.9996518039343298e-05, "loss": 0.8259, "step": 314 }, { "epoch": 0.04, "learning_rate": 1.999641333081418e-05, "loss": 0.8076, "step": 315 }, { "epoch": 0.04, "learning_rate": 1.9996307071421272e-05, "loss": 0.8854, "step": 316 }, { "epoch": 0.04, "learning_rate": 1.999619926118107e-05, "loss": 0.8006, "step": 317 }, { "epoch": 0.04, "learning_rate": 1.9996089900110295e-05, "loss": 0.8093, "step": 318 }, { "epoch": 0.04, "learning_rate": 1.9995978988225912e-05, "loss": 0.7766, "step": 319 }, { "epoch": 0.04, "learning_rate": 1.9995866525545136e-05, "loss": 0.8409, "step": 320 }, { "epoch": 0.04, "learning_rate": 1.9995752512085408e-05, "loss": 0.8029, "step": 321 }, { "epoch": 0.04, "learning_rate": 1.9995636947864414e-05, "loss": 0.7538, "step": 322 }, { "epoch": 0.04, "learning_rate": 1.999551983290009e-05, "loss": 0.7758, "step": 323 }, { "epoch": 0.04, "learning_rate": 1.99954011672106e-05, "loss": 0.7762, "step": 324 }, { "epoch": 0.04, "learning_rate": 1.9995280950814355e-05, "loss": 0.788, "step": 325 }, { "epoch": 0.04, "learning_rate": 1.999515918373001e-05, "loss": 0.7808, "step": 326 }, { "epoch": 0.04, "learning_rate": 1.999503586597645e-05, "loss": 0.8115, "step": 327 }, { "epoch": 0.04, "learning_rate": 1.999491099757281e-05, "loss": 0.8136, "step": 328 }, { "epoch": 0.04, "learning_rate": 1.9994784578538465e-05, "loss": 0.8002, "step": 329 }, { "epoch": 0.04, "learning_rate": 1.999465660889302e-05, "loss": 0.7855, "step": 330 }, { "epoch": 0.04, "learning_rate": 1.9994527088656333e-05, "loss": 0.7545, "step": 331 }, { "epoch": 0.04, "learning_rate": 1.9994396017848504e-05, "loss": 0.8451, "step": 332 }, { "epoch": 0.04, "learning_rate": 1.9994263396489856e-05, "loss": 0.722, "step": 333 }, { "epoch": 0.04, "learning_rate": 1.9994129224600972e-05, "loss": 0.7856, "step": 334 }, { "epoch": 0.04, "learning_rate": 1.9993993502202666e-05, "loss": 0.8082, "step": 335 }, { "epoch": 0.04, "learning_rate": 1.9993856229315995e-05, "loss": 0.802, "step": 336 }, { "epoch": 0.04, "learning_rate": 1.9993717405962252e-05, "loss": 0.8631, "step": 337 }, { "epoch": 0.04, "learning_rate": 1.9993577032162978e-05, "loss": 0.7957, "step": 338 }, { "epoch": 0.04, "learning_rate": 1.999343510793995e-05, "loss": 0.7736, "step": 339 }, { "epoch": 0.04, "learning_rate": 1.9993291633315187e-05, "loss": 0.7917, "step": 340 }, { "epoch": 0.04, "learning_rate": 1.9993146608310947e-05, "loss": 0.8549, "step": 341 }, { "epoch": 0.04, "learning_rate": 1.9993000032949727e-05, "loss": 0.7881, "step": 342 }, { "epoch": 0.04, "learning_rate": 1.9992851907254275e-05, "loss": 0.7899, "step": 343 }, { "epoch": 0.04, "learning_rate": 1.999270223124756e-05, "loss": 0.7721, "step": 344 }, { "epoch": 0.04, "learning_rate": 1.9992551004952817e-05, "loss": 0.7397, "step": 345 }, { "epoch": 0.04, "learning_rate": 1.9992398228393496e-05, "loss": 0.7732, "step": 346 }, { "epoch": 0.04, "learning_rate": 1.9992243901593303e-05, "loss": 0.8498, "step": 347 }, { "epoch": 0.04, "learning_rate": 1.9992088024576178e-05, "loss": 0.8125, "step": 348 }, { "epoch": 0.04, "learning_rate": 1.999193059736631e-05, "loss": 0.7597, "step": 349 }, { "epoch": 0.04, "learning_rate": 1.999177161998812e-05, "loss": 0.7305, "step": 350 }, { "epoch": 0.04, "learning_rate": 1.9991611092466264e-05, "loss": 0.8083, "step": 351 }, { "epoch": 0.04, "learning_rate": 1.9991449014825662e-05, "loss": 0.7816, "step": 352 }, { "epoch": 0.04, "learning_rate": 1.9991285387091446e-05, "loss": 0.9272, "step": 353 }, { "epoch": 0.04, "learning_rate": 1.9991120209289007e-05, "loss": 0.7938, "step": 354 }, { "epoch": 0.04, "learning_rate": 1.9990953481443973e-05, "loss": 0.7829, "step": 355 }, { "epoch": 0.04, "learning_rate": 1.9990785203582205e-05, "loss": 0.8571, "step": 356 }, { "epoch": 0.04, "learning_rate": 1.9990615375729815e-05, "loss": 0.7553, "step": 357 }, { "epoch": 0.04, "learning_rate": 1.9990443997913146e-05, "loss": 0.7522, "step": 358 }, { "epoch": 0.04, "learning_rate": 1.9990271070158792e-05, "loss": 0.7994, "step": 359 }, { "epoch": 0.04, "learning_rate": 1.9990096592493576e-05, "loss": 0.7054, "step": 360 }, { "epoch": 0.04, "learning_rate": 1.9989920564944567e-05, "loss": 0.7204, "step": 361 }, { "epoch": 0.04, "learning_rate": 1.9989742987539075e-05, "loss": 0.8287, "step": 362 }, { "epoch": 0.04, "learning_rate": 1.9989563860304654e-05, "loss": 0.793, "step": 363 }, { "epoch": 0.04, "learning_rate": 1.9989383183269086e-05, "loss": 0.789, "step": 364 }, { "epoch": 0.04, "learning_rate": 1.9989200956460407e-05, "loss": 0.8232, "step": 365 }, { "epoch": 0.04, "learning_rate": 1.998901717990689e-05, "loss": 0.7775, "step": 366 }, { "epoch": 0.04, "learning_rate": 1.9988831853637042e-05, "loss": 0.7288, "step": 367 }, { "epoch": 0.04, "learning_rate": 1.9988644977679615e-05, "loss": 0.8113, "step": 368 }, { "epoch": 0.04, "learning_rate": 1.9988456552063603e-05, "loss": 0.7975, "step": 369 }, { "epoch": 0.04, "learning_rate": 1.998826657681824e-05, "loss": 0.7679, "step": 370 }, { "epoch": 0.05, "learning_rate": 1.9988075051973e-05, "loss": 0.7448, "step": 371 }, { "epoch": 0.05, "learning_rate": 1.9987881977557587e-05, "loss": 0.7501, "step": 372 }, { "epoch": 0.05, "learning_rate": 1.998768735360197e-05, "loss": 0.7183, "step": 373 }, { "epoch": 0.05, "learning_rate": 1.9987491180136335e-05, "loss": 0.786, "step": 374 }, { "epoch": 0.05, "learning_rate": 1.9987293457191114e-05, "loss": 0.7192, "step": 375 }, { "epoch": 0.05, "learning_rate": 1.9987094184796986e-05, "loss": 0.7755, "step": 376 }, { "epoch": 0.05, "learning_rate": 1.9986893362984864e-05, "loss": 0.8296, "step": 377 }, { "epoch": 0.05, "learning_rate": 1.998669099178591e-05, "loss": 0.7615, "step": 378 }, { "epoch": 0.05, "learning_rate": 1.998648707123151e-05, "loss": 0.7973, "step": 379 }, { "epoch": 0.05, "learning_rate": 1.9986281601353314e-05, "loss": 0.8101, "step": 380 }, { "epoch": 0.05, "learning_rate": 1.9986074582183187e-05, "loss": 0.7679, "step": 381 }, { "epoch": 0.05, "learning_rate": 1.9985866013753252e-05, "loss": 0.775, "step": 382 }, { "epoch": 0.05, "learning_rate": 1.9985655896095868e-05, "loss": 0.7339, "step": 383 }, { "epoch": 0.05, "learning_rate": 1.998544422924363e-05, "loss": 0.8115, "step": 384 }, { "epoch": 0.05, "learning_rate": 1.998523101322938e-05, "loss": 0.8391, "step": 385 }, { "epoch": 0.05, "learning_rate": 1.998501624808619e-05, "loss": 0.7934, "step": 386 }, { "epoch": 0.05, "learning_rate": 1.9984799933847383e-05, "loss": 0.837, "step": 387 }, { "epoch": 0.05, "learning_rate": 1.998458207054652e-05, "loss": 0.7736, "step": 388 }, { "epoch": 0.05, "learning_rate": 1.99843626582174e-05, "loss": 0.7373, "step": 389 }, { "epoch": 0.05, "learning_rate": 1.9984141696894063e-05, "loss": 0.7823, "step": 390 }, { "epoch": 0.05, "learning_rate": 1.9983919186610786e-05, "loss": 0.794, "step": 391 }, { "epoch": 0.05, "learning_rate": 1.99836951274021e-05, "loss": 0.7448, "step": 392 }, { "epoch": 0.05, "learning_rate": 1.998346951930275e-05, "loss": 0.761, "step": 393 }, { "epoch": 0.05, "learning_rate": 1.998324236234775e-05, "loss": 0.8399, "step": 394 }, { "epoch": 0.05, "learning_rate": 1.9983013656572338e-05, "loss": 0.7855, "step": 395 }, { "epoch": 0.05, "learning_rate": 1.9982783402011992e-05, "loss": 0.7119, "step": 396 }, { "epoch": 0.05, "learning_rate": 1.9982551598702443e-05, "loss": 0.7831, "step": 397 }, { "epoch": 0.05, "learning_rate": 1.9982318246679643e-05, "loss": 0.7776, "step": 398 }, { "epoch": 0.05, "learning_rate": 1.9982083345979803e-05, "loss": 0.7494, "step": 399 }, { "epoch": 0.05, "learning_rate": 1.9981846896639362e-05, "loss": 0.7679, "step": 400 }, { "epoch": 0.05, "learning_rate": 1.9981608898694998e-05, "loss": 0.8118, "step": 401 }, { "epoch": 0.05, "learning_rate": 1.9981369352183648e-05, "loss": 0.7852, "step": 402 }, { "epoch": 0.05, "learning_rate": 1.9981128257142463e-05, "loss": 0.7848, "step": 403 }, { "epoch": 0.05, "learning_rate": 1.9980885613608857e-05, "loss": 0.8317, "step": 404 }, { "epoch": 0.05, "learning_rate": 1.9980641421620467e-05, "loss": 0.7536, "step": 405 }, { "epoch": 0.05, "learning_rate": 1.9980395681215178e-05, "loss": 0.7777, "step": 406 }, { "epoch": 0.05, "learning_rate": 1.9980148392431117e-05, "loss": 0.7581, "step": 407 }, { "epoch": 0.05, "learning_rate": 1.997989955530665e-05, "loss": 0.7616, "step": 408 }, { "epoch": 0.05, "learning_rate": 1.9979649169880377e-05, "loss": 0.7926, "step": 409 }, { "epoch": 0.05, "learning_rate": 1.9979397236191152e-05, "loss": 0.7985, "step": 410 }, { "epoch": 0.05, "learning_rate": 1.9979143754278053e-05, "loss": 0.755, "step": 411 }, { "epoch": 0.05, "learning_rate": 1.9978888724180403e-05, "loss": 0.7764, "step": 412 }, { "epoch": 0.05, "learning_rate": 1.9978632145937776e-05, "loss": 0.7578, "step": 413 }, { "epoch": 0.05, "learning_rate": 1.997837401958998e-05, "loss": 0.7559, "step": 414 }, { "epoch": 0.05, "learning_rate": 1.997811434517705e-05, "loss": 0.7656, "step": 415 }, { "epoch": 0.05, "learning_rate": 1.997785312273928e-05, "loss": 0.7496, "step": 416 }, { "epoch": 0.05, "learning_rate": 1.9977590352317194e-05, "loss": 0.7277, "step": 417 }, { "epoch": 0.05, "learning_rate": 1.997732603395156e-05, "loss": 0.7961, "step": 418 }, { "epoch": 0.05, "learning_rate": 1.9977060167683385e-05, "loss": 0.7721, "step": 419 }, { "epoch": 0.05, "learning_rate": 1.9976792753553917e-05, "loss": 0.7792, "step": 420 }, { "epoch": 0.05, "learning_rate": 1.997652379160464e-05, "loss": 0.7372, "step": 421 }, { "epoch": 0.05, "learning_rate": 1.9976253281877286e-05, "loss": 0.7515, "step": 422 }, { "epoch": 0.05, "learning_rate": 1.9975981224413815e-05, "loss": 0.7676, "step": 423 }, { "epoch": 0.05, "learning_rate": 1.997570761925644e-05, "loss": 0.7875, "step": 424 }, { "epoch": 0.05, "learning_rate": 1.997543246644761e-05, "loss": 0.7827, "step": 425 }, { "epoch": 0.05, "learning_rate": 1.9975155766030013e-05, "loss": 0.7594, "step": 426 }, { "epoch": 0.05, "learning_rate": 1.9974877518046568e-05, "loss": 0.7261, "step": 427 }, { "epoch": 0.05, "learning_rate": 1.9974597722540455e-05, "loss": 0.698, "step": 428 }, { "epoch": 0.05, "learning_rate": 1.9974316379555074e-05, "loss": 0.817, "step": 429 }, { "epoch": 0.05, "learning_rate": 1.9974033489134076e-05, "loss": 0.7855, "step": 430 }, { "epoch": 0.05, "learning_rate": 1.997374905132135e-05, "loss": 0.7968, "step": 431 }, { "epoch": 0.05, "learning_rate": 1.997346306616102e-05, "loss": 0.8065, "step": 432 }, { "epoch": 0.05, "learning_rate": 1.9973175533697458e-05, "loss": 0.7407, "step": 433 }, { "epoch": 0.05, "learning_rate": 1.9972886453975272e-05, "loss": 0.7494, "step": 434 }, { "epoch": 0.05, "learning_rate": 1.9972595827039313e-05, "loss": 0.7922, "step": 435 }, { "epoch": 0.05, "learning_rate": 1.9972303652934663e-05, "loss": 0.7169, "step": 436 }, { "epoch": 0.05, "learning_rate": 1.9972009931706653e-05, "loss": 0.7107, "step": 437 }, { "epoch": 0.05, "learning_rate": 1.9971714663400855e-05, "loss": 0.7844, "step": 438 }, { "epoch": 0.05, "learning_rate": 1.997141784806307e-05, "loss": 0.762, "step": 439 }, { "epoch": 0.05, "learning_rate": 1.997111948573936e-05, "loss": 0.8319, "step": 440 }, { "epoch": 0.05, "learning_rate": 1.9970819576476e-05, "loss": 0.7598, "step": 441 }, { "epoch": 0.05, "learning_rate": 1.9970518120319523e-05, "loss": 0.733, "step": 442 }, { "epoch": 0.05, "learning_rate": 1.99702151173167e-05, "loss": 0.7783, "step": 443 }, { "epoch": 0.05, "learning_rate": 1.996991056751454e-05, "loss": 0.7635, "step": 444 }, { "epoch": 0.05, "learning_rate": 1.9969604470960283e-05, "loss": 0.8335, "step": 445 }, { "epoch": 0.05, "learning_rate": 1.996929682770143e-05, "loss": 0.8099, "step": 446 }, { "epoch": 0.05, "learning_rate": 1.9968987637785698e-05, "loss": 0.7036, "step": 447 }, { "epoch": 0.05, "learning_rate": 1.9968676901261062e-05, "loss": 0.7903, "step": 448 }, { "epoch": 0.05, "learning_rate": 1.996836461817573e-05, "loss": 0.7344, "step": 449 }, { "epoch": 0.05, "learning_rate": 1.996805078857815e-05, "loss": 0.7464, "step": 450 }, { "epoch": 0.05, "learning_rate": 1.9967735412517003e-05, "loss": 0.738, "step": 451 }, { "epoch": 0.05, "learning_rate": 1.996741849004123e-05, "loss": 0.7325, "step": 452 }, { "epoch": 0.06, "learning_rate": 1.996710002119999e-05, "loss": 0.7719, "step": 453 }, { "epoch": 0.06, "learning_rate": 1.996678000604269e-05, "loss": 0.7142, "step": 454 }, { "epoch": 0.06, "learning_rate": 1.9966458444618983e-05, "loss": 0.7869, "step": 455 }, { "epoch": 0.06, "learning_rate": 1.996613533697875e-05, "loss": 0.6969, "step": 456 }, { "epoch": 0.06, "learning_rate": 1.996581068317213e-05, "loss": 0.7312, "step": 457 }, { "epoch": 0.06, "learning_rate": 1.9965484483249478e-05, "loss": 0.7671, "step": 458 }, { "epoch": 0.06, "learning_rate": 1.9965156737261407e-05, "loss": 0.7543, "step": 459 }, { "epoch": 0.06, "learning_rate": 1.996482744525877e-05, "loss": 0.7872, "step": 460 }, { "epoch": 0.06, "learning_rate": 1.9964496607292642e-05, "loss": 0.7643, "step": 461 }, { "epoch": 0.06, "learning_rate": 1.9964164223414358e-05, "loss": 0.8375, "step": 462 }, { "epoch": 0.06, "learning_rate": 1.9963830293675483e-05, "loss": 0.7845, "step": 463 }, { "epoch": 0.06, "learning_rate": 1.9963494818127822e-05, "loss": 0.7487, "step": 464 }, { "epoch": 0.06, "learning_rate": 1.9963157796823424e-05, "loss": 0.7186, "step": 465 }, { "epoch": 0.06, "learning_rate": 1.996281922981457e-05, "loss": 0.7042, "step": 466 }, { "epoch": 0.06, "learning_rate": 1.9962479117153794e-05, "loss": 0.7696, "step": 467 }, { "epoch": 0.06, "learning_rate": 1.9962137458893856e-05, "loss": 0.8086, "step": 468 }, { "epoch": 0.06, "learning_rate": 1.996179425508776e-05, "loss": 0.732, "step": 469 }, { "epoch": 0.06, "learning_rate": 1.996144950578876e-05, "loss": 0.7846, "step": 470 }, { "epoch": 0.06, "learning_rate": 1.9961103211050333e-05, "loss": 0.7566, "step": 471 }, { "epoch": 0.06, "learning_rate": 1.99607553709262e-05, "loss": 0.7194, "step": 472 }, { "epoch": 0.06, "learning_rate": 1.996040598547034e-05, "loss": 0.745, "step": 473 }, { "epoch": 0.06, "learning_rate": 1.9960055054736945e-05, "loss": 0.8051, "step": 474 }, { "epoch": 0.06, "learning_rate": 1.9959702578780466e-05, "loss": 0.7943, "step": 475 }, { "epoch": 0.06, "learning_rate": 1.995934855765558e-05, "loss": 0.7582, "step": 476 }, { "epoch": 0.06, "learning_rate": 1.995899299141722e-05, "loss": 0.6714, "step": 477 }, { "epoch": 0.06, "learning_rate": 1.995863588012054e-05, "loss": 0.7336, "step": 478 }, { "epoch": 0.06, "learning_rate": 1.995827722382095e-05, "loss": 0.7681, "step": 479 }, { "epoch": 0.06, "learning_rate": 1.9957917022574086e-05, "loss": 0.8047, "step": 480 }, { "epoch": 0.06, "learning_rate": 1.9957555276435837e-05, "loss": 0.7573, "step": 481 }, { "epoch": 0.06, "learning_rate": 1.995719198546232e-05, "loss": 0.712, "step": 482 }, { "epoch": 0.06, "learning_rate": 1.9956827149709898e-05, "loss": 0.7467, "step": 483 }, { "epoch": 0.06, "learning_rate": 1.995646076923517e-05, "loss": 0.7302, "step": 484 }, { "epoch": 0.06, "learning_rate": 1.9956092844094988e-05, "loss": 0.7075, "step": 485 }, { "epoch": 0.06, "learning_rate": 1.9955723374346423e-05, "loss": 0.7544, "step": 486 }, { "epoch": 0.06, "learning_rate": 1.9955352360046795e-05, "loss": 0.7175, "step": 487 }, { "epoch": 0.06, "learning_rate": 1.9954979801253663e-05, "loss": 0.7005, "step": 488 }, { "epoch": 0.06, "learning_rate": 1.9954605698024833e-05, "loss": 0.7534, "step": 489 }, { "epoch": 0.06, "learning_rate": 1.995423005041834e-05, "loss": 0.7533, "step": 490 }, { "epoch": 0.06, "learning_rate": 1.9953852858492464e-05, "loss": 0.7562, "step": 491 }, { "epoch": 0.06, "learning_rate": 1.995347412230572e-05, "loss": 0.7719, "step": 492 }, { "epoch": 0.06, "learning_rate": 1.995309384191687e-05, "loss": 0.6621, "step": 493 }, { "epoch": 0.06, "learning_rate": 1.995271201738491e-05, "loss": 0.7222, "step": 494 }, { "epoch": 0.06, "learning_rate": 1.9952328648769076e-05, "loss": 0.7884, "step": 495 }, { "epoch": 0.06, "learning_rate": 1.995194373612885e-05, "loss": 0.8072, "step": 496 }, { "epoch": 0.06, "learning_rate": 1.9951557279523935e-05, "loss": 0.7692, "step": 497 }, { "epoch": 0.06, "learning_rate": 1.99511692790143e-05, "loss": 0.7273, "step": 498 }, { "epoch": 0.06, "learning_rate": 1.995077973466014e-05, "loss": 0.7424, "step": 499 }, { "epoch": 0.06, "learning_rate": 1.995038864652188e-05, "loss": 0.7795, "step": 500 }, { "epoch": 0.06, "learning_rate": 1.9949996014660202e-05, "loss": 0.7796, "step": 501 }, { "epoch": 0.06, "learning_rate": 1.9949601839136014e-05, "loss": 0.7725, "step": 502 }, { "epoch": 0.06, "learning_rate": 1.9949206120010472e-05, "loss": 0.7528, "step": 503 }, { "epoch": 0.06, "learning_rate": 1.9948808857344973e-05, "loss": 0.7044, "step": 504 }, { "epoch": 0.06, "learning_rate": 1.9948410051201144e-05, "loss": 0.7089, "step": 505 }, { "epoch": 0.06, "learning_rate": 1.9948009701640854e-05, "loss": 0.7908, "step": 506 }, { "epoch": 0.06, "learning_rate": 1.9947607808726222e-05, "loss": 0.709, "step": 507 }, { "epoch": 0.06, "learning_rate": 1.9947204372519593e-05, "loss": 0.7558, "step": 508 }, { "epoch": 0.06, "learning_rate": 1.9946799393083556e-05, "loss": 0.7729, "step": 509 }, { "epoch": 0.06, "learning_rate": 1.9946392870480944e-05, "loss": 0.7419, "step": 510 }, { "epoch": 0.06, "learning_rate": 1.9945984804774824e-05, "loss": 0.7772, "step": 511 }, { "epoch": 0.06, "learning_rate": 1.9945575196028506e-05, "loss": 0.7716, "step": 512 }, { "epoch": 0.06, "learning_rate": 1.9945164044305534e-05, "loss": 0.7021, "step": 513 }, { "epoch": 0.06, "learning_rate": 1.9944751349669697e-05, "loss": 0.7674, "step": 514 }, { "epoch": 0.06, "learning_rate": 1.994433711218502e-05, "loss": 0.7581, "step": 515 }, { "epoch": 0.06, "learning_rate": 1.9943921331915774e-05, "loss": 0.7718, "step": 516 }, { "epoch": 0.06, "learning_rate": 1.9943504008926454e-05, "loss": 0.7516, "step": 517 }, { "epoch": 0.06, "learning_rate": 1.9943085143281814e-05, "loss": 0.7614, "step": 518 }, { "epoch": 0.06, "learning_rate": 1.9942664735046833e-05, "loss": 0.7478, "step": 519 }, { "epoch": 0.06, "learning_rate": 1.9942242784286734e-05, "loss": 0.7193, "step": 520 }, { "epoch": 0.06, "learning_rate": 1.994181929106698e-05, "loss": 0.7023, "step": 521 }, { "epoch": 0.06, "learning_rate": 1.9941394255453274e-05, "loss": 0.6837, "step": 522 }, { "epoch": 0.06, "learning_rate": 1.9940967677511552e-05, "loss": 0.7791, "step": 523 }, { "epoch": 0.06, "learning_rate": 1.9940539557308e-05, "loss": 0.7629, "step": 524 }, { "epoch": 0.06, "learning_rate": 1.9940109894909035e-05, "loss": 0.7403, "step": 525 }, { "epoch": 0.06, "learning_rate": 1.9939678690381313e-05, "loss": 0.7105, "step": 526 }, { "epoch": 0.06, "learning_rate": 1.9939245943791737e-05, "loss": 0.7413, "step": 527 }, { "epoch": 0.06, "learning_rate": 1.9938811655207445e-05, "loss": 0.7192, "step": 528 }, { "epoch": 0.06, "learning_rate": 1.9938375824695808e-05, "loss": 0.72, "step": 529 }, { "epoch": 0.06, "learning_rate": 1.9937938452324442e-05, "loss": 0.7625, "step": 530 }, { "epoch": 0.06, "learning_rate": 1.9937499538161205e-05, "loss": 0.8188, "step": 531 }, { "epoch": 0.06, "learning_rate": 1.993705908227419e-05, "loss": 0.7336, "step": 532 }, { "epoch": 0.06, "learning_rate": 1.993661708473173e-05, "loss": 0.6845, "step": 533 }, { "epoch": 0.06, "learning_rate": 1.9936173545602395e-05, "loss": 0.7712, "step": 534 }, { "epoch": 0.07, "learning_rate": 1.9935728464955002e-05, "loss": 0.7508, "step": 535 }, { "epoch": 0.07, "learning_rate": 1.9935281842858596e-05, "loss": 0.7336, "step": 536 }, { "epoch": 0.07, "learning_rate": 1.9934833679382472e-05, "loss": 0.6709, "step": 537 }, { "epoch": 0.07, "learning_rate": 1.9934383974596153e-05, "loss": 0.6964, "step": 538 }, { "epoch": 0.07, "learning_rate": 1.9933932728569416e-05, "loss": 0.7609, "step": 539 }, { "epoch": 0.07, "learning_rate": 1.993347994137226e-05, "loss": 0.7092, "step": 540 }, { "epoch": 0.07, "learning_rate": 1.9933025613074935e-05, "loss": 0.7361, "step": 541 }, { "epoch": 0.07, "learning_rate": 1.9932569743747923e-05, "loss": 0.6938, "step": 542 }, { "epoch": 0.07, "learning_rate": 1.993211233346195e-05, "loss": 0.6929, "step": 543 }, { "epoch": 0.07, "learning_rate": 1.9931653382287988e-05, "loss": 0.7186, "step": 544 }, { "epoch": 0.07, "learning_rate": 1.9931192890297227e-05, "loss": 0.7298, "step": 545 }, { "epoch": 0.07, "learning_rate": 1.993073085756111e-05, "loss": 0.7253, "step": 546 }, { "epoch": 0.07, "learning_rate": 1.9930267284151328e-05, "loss": 0.691, "step": 547 }, { "epoch": 0.07, "learning_rate": 1.992980217013979e-05, "loss": 0.7218, "step": 548 }, { "epoch": 0.07, "learning_rate": 1.9929335515598664e-05, "loss": 0.7792, "step": 549 }, { "epoch": 0.07, "learning_rate": 1.992886732060034e-05, "loss": 0.7628, "step": 550 }, { "epoch": 0.07, "learning_rate": 1.9928397585217458e-05, "loss": 0.7017, "step": 551 }, { "epoch": 0.07, "learning_rate": 1.9927926309522892e-05, "loss": 0.7606, "step": 552 }, { "epoch": 0.07, "learning_rate": 1.9927453493589758e-05, "loss": 0.6782, "step": 553 }, { "epoch": 0.07, "learning_rate": 1.9926979137491408e-05, "loss": 0.7524, "step": 554 }, { "epoch": 0.07, "learning_rate": 1.992650324130144e-05, "loss": 0.6943, "step": 555 }, { "epoch": 0.07, "learning_rate": 1.9926025805093677e-05, "loss": 0.6973, "step": 556 }, { "epoch": 0.07, "learning_rate": 1.9925546828942197e-05, "loss": 0.6991, "step": 557 }, { "epoch": 0.07, "learning_rate": 1.9925066312921303e-05, "loss": 0.7329, "step": 558 }, { "epoch": 0.07, "learning_rate": 1.992458425710555e-05, "loss": 0.7482, "step": 559 }, { "epoch": 0.07, "learning_rate": 1.992410066156972e-05, "loss": 0.7326, "step": 560 }, { "epoch": 0.07, "learning_rate": 1.9923615526388843e-05, "loss": 0.7254, "step": 561 }, { "epoch": 0.07, "learning_rate": 1.992312885163818e-05, "loss": 0.7217, "step": 562 }, { "epoch": 0.07, "learning_rate": 1.9922640637393232e-05, "loss": 0.7237, "step": 563 }, { "epoch": 0.07, "learning_rate": 1.9922150883729752e-05, "loss": 0.7522, "step": 564 }, { "epoch": 0.07, "learning_rate": 1.992165959072371e-05, "loss": 0.785, "step": 565 }, { "epoch": 0.07, "learning_rate": 1.9921166758451335e-05, "loss": 0.7069, "step": 566 }, { "epoch": 0.07, "learning_rate": 1.992067238698908e-05, "loss": 0.7645, "step": 567 }, { "epoch": 0.07, "learning_rate": 1.9920176476413644e-05, "loss": 0.7024, "step": 568 }, { "epoch": 0.07, "learning_rate": 1.9919679026801967e-05, "loss": 0.665, "step": 569 }, { "epoch": 0.07, "learning_rate": 1.991918003823122e-05, "loss": 0.7144, "step": 570 }, { "epoch": 0.07, "learning_rate": 1.9918679510778823e-05, "loss": 0.7145, "step": 571 }, { "epoch": 0.07, "learning_rate": 1.991817744452242e-05, "loss": 0.7276, "step": 572 }, { "epoch": 0.07, "learning_rate": 1.991767383953991e-05, "loss": 0.7219, "step": 573 }, { "epoch": 0.07, "learning_rate": 1.991716869590942e-05, "loss": 0.7019, "step": 574 }, { "epoch": 0.07, "learning_rate": 1.991666201370932e-05, "loss": 0.6697, "step": 575 }, { "epoch": 0.07, "learning_rate": 1.991615379301822e-05, "loss": 0.7354, "step": 576 }, { "epoch": 0.07, "learning_rate": 1.9915644033914962e-05, "loss": 0.7201, "step": 577 }, { "epoch": 0.07, "learning_rate": 1.991513273647863e-05, "loss": 0.6851, "step": 578 }, { "epoch": 0.07, "learning_rate": 1.991461990078855e-05, "loss": 0.6742, "step": 579 }, { "epoch": 0.07, "learning_rate": 1.991410552692429e-05, "loss": 0.732, "step": 580 }, { "epoch": 0.07, "learning_rate": 1.9913589614965644e-05, "loss": 0.6856, "step": 581 }, { "epoch": 0.07, "learning_rate": 1.9913072164992654e-05, "loss": 0.7516, "step": 582 }, { "epoch": 0.07, "learning_rate": 1.9912553177085598e-05, "loss": 0.7016, "step": 583 }, { "epoch": 0.07, "learning_rate": 1.991203265132499e-05, "loss": 0.7742, "step": 584 }, { "epoch": 0.07, "learning_rate": 1.991151058779159e-05, "loss": 0.7607, "step": 585 }, { "epoch": 0.07, "learning_rate": 1.991098698656639e-05, "loss": 0.7408, "step": 586 }, { "epoch": 0.07, "learning_rate": 1.9910461847730623e-05, "loss": 0.7523, "step": 587 }, { "epoch": 0.07, "learning_rate": 1.990993517136576e-05, "loss": 0.6803, "step": 588 }, { "epoch": 0.07, "learning_rate": 1.990940695755351e-05, "loss": 0.7629, "step": 589 }, { "epoch": 0.07, "learning_rate": 1.990887720637582e-05, "loss": 0.7752, "step": 590 }, { "epoch": 0.07, "learning_rate": 1.990834591791488e-05, "loss": 0.7341, "step": 591 }, { "epoch": 0.07, "learning_rate": 1.9907813092253114e-05, "loss": 0.7097, "step": 592 }, { "epoch": 0.07, "learning_rate": 1.990727872947318e-05, "loss": 0.7717, "step": 593 }, { "epoch": 0.07, "learning_rate": 1.9906742829657993e-05, "loss": 0.6711, "step": 594 }, { "epoch": 0.07, "learning_rate": 1.990620539289068e-05, "loss": 0.7895, "step": 595 }, { "epoch": 0.07, "learning_rate": 1.990566641925463e-05, "loss": 0.6892, "step": 596 }, { "epoch": 0.07, "learning_rate": 1.9905125908833452e-05, "loss": 0.6578, "step": 597 }, { "epoch": 0.07, "learning_rate": 1.9904583861711008e-05, "loss": 0.659, "step": 598 }, { "epoch": 0.07, "learning_rate": 1.990404027797139e-05, "loss": 0.7575, "step": 599 }, { "epoch": 0.07, "learning_rate": 1.9903495157698934e-05, "loss": 0.7657, "step": 600 }, { "epoch": 0.07, "learning_rate": 1.9902948500978203e-05, "loss": 0.7698, "step": 601 }, { "epoch": 0.07, "learning_rate": 1.9902400307894014e-05, "loss": 0.7695, "step": 602 }, { "epoch": 0.07, "learning_rate": 1.990185057853141e-05, "loss": 0.7101, "step": 603 }, { "epoch": 0.07, "learning_rate": 1.9901299312975685e-05, "loss": 0.7203, "step": 604 }, { "epoch": 0.07, "learning_rate": 1.9900746511312353e-05, "loss": 0.6765, "step": 605 }, { "epoch": 0.07, "learning_rate": 1.9900192173627183e-05, "loss": 0.7313, "step": 606 }, { "epoch": 0.07, "learning_rate": 1.9899636300006175e-05, "loss": 0.7173, "step": 607 }, { "epoch": 0.07, "learning_rate": 1.9899078890535567e-05, "loss": 0.7292, "step": 608 }, { "epoch": 0.07, "learning_rate": 1.989851994530184e-05, "loss": 0.7305, "step": 609 }, { "epoch": 0.07, "learning_rate": 1.9897959464391708e-05, "loss": 0.7463, "step": 610 }, { "epoch": 0.07, "learning_rate": 1.9897397447892122e-05, "loss": 0.7072, "step": 611 }, { "epoch": 0.07, "learning_rate": 1.989683389589028e-05, "loss": 0.7491, "step": 612 }, { "epoch": 0.07, "learning_rate": 1.989626880847361e-05, "loss": 0.735, "step": 613 }, { "epoch": 0.07, "learning_rate": 1.9895702185729777e-05, "loss": 0.7307, "step": 614 }, { "epoch": 0.07, "learning_rate": 1.9895134027746694e-05, "loss": 0.7093, "step": 615 }, { "epoch": 0.07, "learning_rate": 1.9894564334612504e-05, "loss": 0.6207, "step": 616 }, { "epoch": 0.08, "learning_rate": 1.9893993106415585e-05, "loss": 0.6484, "step": 617 }, { "epoch": 0.08, "learning_rate": 1.989342034324457e-05, "loss": 0.6921, "step": 618 }, { "epoch": 0.08, "learning_rate": 1.989284604518831e-05, "loss": 0.7621, "step": 619 }, { "epoch": 0.08, "learning_rate": 1.9892270212335904e-05, "loss": 0.7074, "step": 620 }, { "epoch": 0.08, "learning_rate": 1.989169284477669e-05, "loss": 0.792, "step": 621 }, { "epoch": 0.08, "learning_rate": 1.989111394260024e-05, "loss": 0.7247, "step": 622 }, { "epoch": 0.08, "learning_rate": 1.9890533505896366e-05, "loss": 0.7318, "step": 623 }, { "epoch": 0.08, "learning_rate": 1.9889951534755122e-05, "loss": 0.7556, "step": 624 }, { "epoch": 0.08, "learning_rate": 1.9889368029266793e-05, "loss": 0.7584, "step": 625 }, { "epoch": 0.08, "learning_rate": 1.9888782989521903e-05, "loss": 0.7117, "step": 626 }, { "epoch": 0.08, "learning_rate": 1.988819641561122e-05, "loss": 0.752, "step": 627 }, { "epoch": 0.08, "learning_rate": 1.9887608307625745e-05, "loss": 0.7303, "step": 628 }, { "epoch": 0.08, "learning_rate": 1.9887018665656715e-05, "loss": 0.7205, "step": 629 }, { "epoch": 0.08, "learning_rate": 1.9886427489795616e-05, "loss": 0.7485, "step": 630 }, { "epoch": 0.08, "learning_rate": 1.9885834780134158e-05, "loss": 0.6865, "step": 631 }, { "epoch": 0.08, "learning_rate": 1.9885240536764298e-05, "loss": 0.7091, "step": 632 }, { "epoch": 0.08, "learning_rate": 1.9884644759778223e-05, "loss": 0.6609, "step": 633 }, { "epoch": 0.08, "learning_rate": 1.9884047449268372e-05, "loss": 0.7278, "step": 634 }, { "epoch": 0.08, "learning_rate": 1.9883448605327405e-05, "loss": 0.6559, "step": 635 }, { "epoch": 0.08, "learning_rate": 1.988284822804823e-05, "loss": 0.7005, "step": 636 }, { "epoch": 0.08, "learning_rate": 1.9882246317523992e-05, "loss": 0.6799, "step": 637 }, { "epoch": 0.08, "learning_rate": 1.9881642873848073e-05, "loss": 0.729, "step": 638 }, { "epoch": 0.08, "learning_rate": 1.988103789711409e-05, "loss": 0.7632, "step": 639 }, { "epoch": 0.08, "learning_rate": 1.9880431387415907e-05, "loss": 0.6492, "step": 640 }, { "epoch": 0.08, "learning_rate": 1.987982334484761e-05, "loss": 0.6907, "step": 641 }, { "epoch": 0.08, "learning_rate": 1.9879213769503537e-05, "loss": 0.7176, "step": 642 }, { "epoch": 0.08, "learning_rate": 1.9878602661478256e-05, "loss": 0.7365, "step": 643 }, { "epoch": 0.08, "learning_rate": 1.987799002086658e-05, "loss": 0.6238, "step": 644 }, { "epoch": 0.08, "learning_rate": 1.9877375847763553e-05, "loss": 0.8016, "step": 645 }, { "epoch": 0.08, "learning_rate": 1.9876760142264455e-05, "loss": 0.6907, "step": 646 }, { "epoch": 0.08, "learning_rate": 1.9876142904464816e-05, "loss": 0.7631, "step": 647 }, { "epoch": 0.08, "learning_rate": 1.9875524134460387e-05, "loss": 0.7666, "step": 648 }, { "epoch": 0.08, "learning_rate": 1.987490383234717e-05, "loss": 0.7979, "step": 649 }, { "epoch": 0.08, "learning_rate": 1.98742819982214e-05, "loss": 0.7219, "step": 650 }, { "epoch": 0.08, "learning_rate": 1.987365863217955e-05, "loss": 0.723, "step": 651 }, { "epoch": 0.08, "learning_rate": 1.987303373431833e-05, "loss": 0.6796, "step": 652 }, { "epoch": 0.08, "learning_rate": 1.9872407304734684e-05, "loss": 0.7117, "step": 653 }, { "epoch": 0.08, "learning_rate": 1.98717793435258e-05, "loss": 0.6986, "step": 654 }, { "epoch": 0.08, "learning_rate": 1.9871149850789102e-05, "loss": 0.6712, "step": 655 }, { "epoch": 0.08, "learning_rate": 1.9870518826622253e-05, "loss": 0.7087, "step": 656 }, { "epoch": 0.08, "learning_rate": 1.9869886271123147e-05, "loss": 0.7181, "step": 657 }, { "epoch": 0.08, "learning_rate": 1.9869252184389923e-05, "loss": 0.6129, "step": 658 }, { "epoch": 0.08, "learning_rate": 1.986861656652095e-05, "loss": 0.6962, "step": 659 }, { "epoch": 0.08, "learning_rate": 1.9867979417614846e-05, "loss": 0.7234, "step": 660 }, { "epoch": 0.08, "learning_rate": 1.9867340737770452e-05, "loss": 0.7141, "step": 661 }, { "epoch": 0.08, "learning_rate": 1.9866700527086862e-05, "loss": 0.7351, "step": 662 }, { "epoch": 0.08, "learning_rate": 1.9866058785663392e-05, "loss": 0.7537, "step": 663 }, { "epoch": 0.08, "learning_rate": 1.9865415513599612e-05, "loss": 0.7374, "step": 664 }, { "epoch": 0.08, "learning_rate": 1.986477071099531e-05, "loss": 0.7148, "step": 665 }, { "epoch": 0.08, "learning_rate": 1.986412437795053e-05, "loss": 0.7828, "step": 666 }, { "epoch": 0.08, "learning_rate": 1.986347651456554e-05, "loss": 0.6909, "step": 667 }, { "epoch": 0.08, "learning_rate": 1.9862827120940858e-05, "loss": 0.6827, "step": 668 }, { "epoch": 0.08, "learning_rate": 1.986217619717722e-05, "loss": 0.7605, "step": 669 }, { "epoch": 0.08, "learning_rate": 1.9861523743375626e-05, "loss": 0.7353, "step": 670 }, { "epoch": 0.08, "learning_rate": 1.986086975963729e-05, "loss": 0.7151, "step": 671 }, { "epoch": 0.08, "learning_rate": 1.9860214246063673e-05, "loss": 0.728, "step": 672 }, { "epoch": 0.08, "learning_rate": 1.9859557202756477e-05, "loss": 0.7595, "step": 673 }, { "epoch": 0.08, "learning_rate": 1.9858898629817635e-05, "loss": 0.7428, "step": 674 }, { "epoch": 0.08, "learning_rate": 1.9858238527349313e-05, "loss": 0.6901, "step": 675 }, { "epoch": 0.08, "learning_rate": 1.985757689545393e-05, "loss": 0.7461, "step": 676 }, { "epoch": 0.08, "learning_rate": 1.985691373423413e-05, "loss": 0.6807, "step": 677 }, { "epoch": 0.08, "learning_rate": 1.9856249043792798e-05, "loss": 0.6887, "step": 678 }, { "epoch": 0.08, "learning_rate": 1.9855582824233047e-05, "loss": 0.6745, "step": 679 }, { "epoch": 0.08, "learning_rate": 1.9854915075658246e-05, "loss": 0.7454, "step": 680 }, { "epoch": 0.08, "learning_rate": 1.985424579817199e-05, "loss": 0.6948, "step": 681 }, { "epoch": 0.08, "learning_rate": 1.9853574991878107e-05, "loss": 0.73, "step": 682 }, { "epoch": 0.08, "learning_rate": 1.985290265688067e-05, "loss": 0.7602, "step": 683 }, { "epoch": 0.08, "learning_rate": 1.9852228793283986e-05, "loss": 0.6771, "step": 684 }, { "epoch": 0.08, "learning_rate": 1.98515534011926e-05, "loss": 0.7538, "step": 685 }, { "epoch": 0.08, "learning_rate": 1.9850876480711294e-05, "loss": 0.7017, "step": 686 }, { "epoch": 0.08, "learning_rate": 1.985019803194508e-05, "loss": 0.7028, "step": 687 }, { "epoch": 0.08, "learning_rate": 1.984951805499923e-05, "loss": 0.7119, "step": 688 }, { "epoch": 0.08, "learning_rate": 1.984883654997922e-05, "loss": 0.7347, "step": 689 }, { "epoch": 0.08, "learning_rate": 1.984815351699079e-05, "loss": 0.7014, "step": 690 }, { "epoch": 0.08, "learning_rate": 1.9847468956139903e-05, "loss": 0.7169, "step": 691 }, { "epoch": 0.08, "learning_rate": 1.9846782867532766e-05, "loss": 0.7331, "step": 692 }, { "epoch": 0.08, "learning_rate": 1.9846095251275816e-05, "loss": 0.7235, "step": 693 }, { "epoch": 0.08, "learning_rate": 1.9845406107475736e-05, "loss": 0.7017, "step": 694 }, { "epoch": 0.08, "learning_rate": 1.9844715436239438e-05, "loss": 0.6643, "step": 695 }, { "epoch": 0.08, "learning_rate": 1.9844023237674075e-05, "loss": 0.7004, "step": 696 }, { "epoch": 0.08, "learning_rate": 1.9843329511887034e-05, "loss": 0.7037, "step": 697 }, { "epoch": 0.08, "learning_rate": 1.9842634258985946e-05, "loss": 0.7031, "step": 698 }, { "epoch": 0.09, "learning_rate": 1.984193747907867e-05, "loss": 0.7215, "step": 699 }, { "epoch": 0.09, "learning_rate": 1.9841239172273307e-05, "loss": 0.7646, "step": 700 }, { "epoch": 0.09, "learning_rate": 1.984053933867819e-05, "loss": 0.6985, "step": 701 }, { "epoch": 0.09, "learning_rate": 1.98398379784019e-05, "loss": 0.7085, "step": 702 }, { "epoch": 0.09, "learning_rate": 1.9839135091553247e-05, "loss": 0.7125, "step": 703 }, { "epoch": 0.09, "learning_rate": 1.9838430678241268e-05, "loss": 0.7367, "step": 704 }, { "epoch": 0.09, "learning_rate": 1.9837724738575258e-05, "loss": 0.594, "step": 705 }, { "epoch": 0.09, "learning_rate": 1.9837017272664734e-05, "loss": 0.7302, "step": 706 }, { "epoch": 0.09, "learning_rate": 1.983630828061945e-05, "loss": 0.6581, "step": 707 }, { "epoch": 0.09, "learning_rate": 1.983559776254941e-05, "loss": 0.7672, "step": 708 }, { "epoch": 0.09, "learning_rate": 1.9834885718564837e-05, "loss": 0.68, "step": 709 }, { "epoch": 0.09, "learning_rate": 1.98341721487762e-05, "loss": 0.7031, "step": 710 }, { "epoch": 0.09, "learning_rate": 1.9833457053294207e-05, "loss": 0.746, "step": 711 }, { "epoch": 0.09, "learning_rate": 1.9832740432229793e-05, "loss": 0.7185, "step": 712 }, { "epoch": 0.09, "learning_rate": 1.9832022285694146e-05, "loss": 0.717, "step": 713 }, { "epoch": 0.09, "learning_rate": 1.9831302613798674e-05, "loss": 0.7228, "step": 714 }, { "epoch": 0.09, "learning_rate": 1.9830581416655027e-05, "loss": 0.6415, "step": 715 }, { "epoch": 0.09, "learning_rate": 1.9829858694375096e-05, "loss": 0.6692, "step": 716 }, { "epoch": 0.09, "learning_rate": 1.9829134447071002e-05, "loss": 0.6713, "step": 717 }, { "epoch": 0.09, "learning_rate": 1.9828408674855117e-05, "loss": 0.6575, "step": 718 }, { "epoch": 0.09, "learning_rate": 1.9827681377840023e-05, "loss": 0.65, "step": 719 }, { "epoch": 0.09, "learning_rate": 1.9826952556138567e-05, "loss": 0.7028, "step": 720 }, { "epoch": 0.09, "learning_rate": 1.9826222209863814e-05, "loss": 0.6797, "step": 721 }, { "epoch": 0.09, "learning_rate": 1.982549033912907e-05, "loss": 0.7653, "step": 722 }, { "epoch": 0.09, "learning_rate": 1.9824756944047886e-05, "loss": 0.6921, "step": 723 }, { "epoch": 0.09, "learning_rate": 1.982402202473403e-05, "loss": 0.7015, "step": 724 }, { "epoch": 0.09, "learning_rate": 1.9823285581301537e-05, "loss": 0.7303, "step": 725 }, { "epoch": 0.09, "learning_rate": 1.9822547613864645e-05, "loss": 0.7372, "step": 726 }, { "epoch": 0.09, "learning_rate": 1.982180812253785e-05, "loss": 0.7494, "step": 727 }, { "epoch": 0.09, "learning_rate": 1.9821067107435876e-05, "loss": 0.6535, "step": 728 }, { "epoch": 0.09, "learning_rate": 1.982032456867368e-05, "loss": 0.6452, "step": 729 }, { "epoch": 0.09, "learning_rate": 1.9819580506366478e-05, "loss": 0.7076, "step": 730 }, { "epoch": 0.09, "learning_rate": 1.981883492062969e-05, "loss": 0.6776, "step": 731 }, { "epoch": 0.09, "learning_rate": 1.9818087811578994e-05, "loss": 0.7517, "step": 732 }, { "epoch": 0.09, "learning_rate": 1.9817339179330296e-05, "loss": 0.6327, "step": 733 }, { "epoch": 0.09, "learning_rate": 1.9816589023999743e-05, "loss": 0.6828, "step": 734 }, { "epoch": 0.09, "learning_rate": 1.981583734570371e-05, "loss": 0.7246, "step": 735 }, { "epoch": 0.09, "learning_rate": 1.981508414455882e-05, "loss": 0.6262, "step": 736 }, { "epoch": 0.09, "learning_rate": 1.9814329420681922e-05, "loss": 0.58, "step": 737 }, { "epoch": 0.09, "learning_rate": 1.9813573174190105e-05, "loss": 0.6681, "step": 738 }, { "epoch": 0.09, "learning_rate": 1.98128154052007e-05, "loss": 0.6704, "step": 739 }, { "epoch": 0.09, "learning_rate": 1.9812056113831263e-05, "loss": 0.6932, "step": 740 }, { "epoch": 0.09, "learning_rate": 1.9811295300199597e-05, "loss": 0.7132, "step": 741 }, { "epoch": 0.09, "learning_rate": 1.981053296442373e-05, "loss": 0.6924, "step": 742 }, { "epoch": 0.09, "learning_rate": 1.980976910662194e-05, "loss": 0.7506, "step": 743 }, { "epoch": 0.09, "learning_rate": 1.9809003726912726e-05, "loss": 0.7028, "step": 744 }, { "epoch": 0.09, "learning_rate": 1.9808236825414837e-05, "loss": 0.6308, "step": 745 }, { "epoch": 0.09, "learning_rate": 1.9807468402247245e-05, "loss": 0.7, "step": 746 }, { "epoch": 0.09, "learning_rate": 1.980669845752917e-05, "loss": 0.6501, "step": 747 }, { "epoch": 0.09, "learning_rate": 1.9805926991380057e-05, "loss": 0.7521, "step": 748 }, { "epoch": 0.09, "learning_rate": 1.9805154003919603e-05, "loss": 0.7295, "step": 749 }, { "epoch": 0.09, "learning_rate": 1.9804379495267722e-05, "loss": 0.6839, "step": 750 }, { "epoch": 0.09, "learning_rate": 1.9803603465544574e-05, "loss": 0.669, "step": 751 }, { "epoch": 0.09, "learning_rate": 1.9802825914870557e-05, "loss": 0.6285, "step": 752 }, { "epoch": 0.09, "learning_rate": 1.9802046843366298e-05, "loss": 0.6715, "step": 753 }, { "epoch": 0.09, "learning_rate": 1.9801266251152665e-05, "loss": 0.7219, "step": 754 }, { "epoch": 0.09, "learning_rate": 1.9800484138350768e-05, "loss": 0.6981, "step": 755 }, { "epoch": 0.09, "learning_rate": 1.9799700505081934e-05, "loss": 0.6851, "step": 756 }, { "epoch": 0.09, "learning_rate": 1.979891535146774e-05, "loss": 0.7274, "step": 757 }, { "epoch": 0.09, "learning_rate": 1.9798128677630006e-05, "loss": 0.6312, "step": 758 }, { "epoch": 0.09, "learning_rate": 1.9797340483690765e-05, "loss": 0.7454, "step": 759 }, { "epoch": 0.09, "learning_rate": 1.9796550769772307e-05, "loss": 0.6897, "step": 760 }, { "epoch": 0.09, "learning_rate": 1.979575953599715e-05, "loss": 0.7223, "step": 761 }, { "epoch": 0.09, "learning_rate": 1.979496678248804e-05, "loss": 0.6157, "step": 762 }, { "epoch": 0.09, "learning_rate": 1.9794172509367977e-05, "loss": 0.6757, "step": 763 }, { "epoch": 0.09, "learning_rate": 1.9793376716760178e-05, "loss": 0.6874, "step": 764 }, { "epoch": 0.09, "learning_rate": 1.9792579404788108e-05, "loss": 0.7102, "step": 765 }, { "epoch": 0.09, "learning_rate": 1.979178057357546e-05, "loss": 0.7398, "step": 766 }, { "epoch": 0.09, "learning_rate": 1.979098022324617e-05, "loss": 0.785, "step": 767 }, { "epoch": 0.09, "learning_rate": 1.9790178353924408e-05, "loss": 0.7172, "step": 768 }, { "epoch": 0.09, "learning_rate": 1.9789374965734572e-05, "loss": 0.6923, "step": 769 }, { "epoch": 0.09, "learning_rate": 1.9788570058801305e-05, "loss": 0.7006, "step": 770 }, { "epoch": 0.09, "learning_rate": 1.9787763633249474e-05, "loss": 0.7361, "step": 771 }, { "epoch": 0.09, "learning_rate": 1.9786955689204203e-05, "loss": 0.6977, "step": 772 }, { "epoch": 0.09, "learning_rate": 1.9786146226790828e-05, "loss": 0.6966, "step": 773 }, { "epoch": 0.09, "learning_rate": 1.9785335246134934e-05, "loss": 0.7488, "step": 774 }, { "epoch": 0.09, "learning_rate": 1.9784522747362337e-05, "loss": 0.6497, "step": 775 }, { "epoch": 0.09, "learning_rate": 1.978370873059909e-05, "loss": 0.716, "step": 776 }, { "epoch": 0.09, "learning_rate": 1.9782893195971485e-05, "loss": 0.7382, "step": 777 }, { "epoch": 0.09, "learning_rate": 1.978207614360604e-05, "loss": 0.7354, "step": 778 }, { "epoch": 0.09, "learning_rate": 1.9781257573629515e-05, "loss": 0.5992, "step": 779 }, { "epoch": 0.09, "learning_rate": 1.9780437486168908e-05, "loss": 0.7066, "step": 780 }, { "epoch": 0.09, "learning_rate": 1.977961588135145e-05, "loss": 0.7547, "step": 781 }, { "epoch": 0.1, "learning_rate": 1.97787927593046e-05, "loss": 0.6393, "step": 782 }, { "epoch": 0.1, "learning_rate": 1.977796812015606e-05, "loss": 0.7033, "step": 783 }, { "epoch": 0.1, "learning_rate": 1.9777141964033774e-05, "loss": 0.756, "step": 784 }, { "epoch": 0.1, "learning_rate": 1.9776314291065905e-05, "loss": 0.6595, "step": 785 }, { "epoch": 0.1, "learning_rate": 1.9775485101380868e-05, "loss": 0.6568, "step": 786 }, { "epoch": 0.1, "learning_rate": 1.9774654395107293e-05, "loss": 0.6764, "step": 787 }, { "epoch": 0.1, "learning_rate": 1.9773822172374068e-05, "loss": 0.6967, "step": 788 }, { "epoch": 0.1, "learning_rate": 1.97729884333103e-05, "loss": 0.7073, "step": 789 }, { "epoch": 0.1, "learning_rate": 1.9772153178045345e-05, "loss": 0.7431, "step": 790 }, { "epoch": 0.1, "learning_rate": 1.9771316406708777e-05, "loss": 0.7224, "step": 791 }, { "epoch": 0.1, "learning_rate": 1.9770478119430415e-05, "loss": 0.7288, "step": 792 }, { "epoch": 0.1, "learning_rate": 1.976963831634032e-05, "loss": 0.6409, "step": 793 }, { "epoch": 0.1, "learning_rate": 1.9768796997568776e-05, "loss": 0.7031, "step": 794 }, { "epoch": 0.1, "learning_rate": 1.9767954163246307e-05, "loss": 0.6837, "step": 795 }, { "epoch": 0.1, "learning_rate": 1.976710981350367e-05, "loss": 0.6812, "step": 796 }, { "epoch": 0.1, "learning_rate": 1.976626394847186e-05, "loss": 0.733, "step": 797 }, { "epoch": 0.1, "learning_rate": 1.976541656828211e-05, "loss": 0.6511, "step": 798 }, { "epoch": 0.1, "learning_rate": 1.9764567673065882e-05, "loss": 0.6738, "step": 799 }, { "epoch": 0.1, "learning_rate": 1.9763717262954875e-05, "loss": 0.6818, "step": 800 }, { "epoch": 0.1, "learning_rate": 1.976286533808102e-05, "loss": 0.6766, "step": 801 }, { "epoch": 0.1, "learning_rate": 1.9762011898576493e-05, "loss": 0.6851, "step": 802 }, { "epoch": 0.1, "learning_rate": 1.9761156944573697e-05, "loss": 0.7043, "step": 803 }, { "epoch": 0.1, "learning_rate": 1.9760300476205268e-05, "loss": 0.7178, "step": 804 }, { "epoch": 0.1, "learning_rate": 1.975944249360408e-05, "loss": 0.7117, "step": 805 }, { "epoch": 0.1, "learning_rate": 1.9758582996903245e-05, "loss": 0.6627, "step": 806 }, { "epoch": 0.1, "learning_rate": 1.9757721986236103e-05, "loss": 0.7063, "step": 807 }, { "epoch": 0.1, "learning_rate": 1.975685946173624e-05, "loss": 0.6901, "step": 808 }, { "epoch": 0.1, "learning_rate": 1.9755995423537465e-05, "loss": 0.6827, "step": 809 }, { "epoch": 0.1, "learning_rate": 1.9755129871773823e-05, "loss": 0.7208, "step": 810 }, { "epoch": 0.1, "learning_rate": 1.9754262806579607e-05, "loss": 0.6502, "step": 811 }, { "epoch": 0.1, "learning_rate": 1.975339422808933e-05, "loss": 0.7829, "step": 812 }, { "epoch": 0.1, "learning_rate": 1.975252413643774e-05, "loss": 0.6964, "step": 813 }, { "epoch": 0.1, "learning_rate": 1.9751652531759832e-05, "loss": 0.7444, "step": 814 }, { "epoch": 0.1, "learning_rate": 1.9750779414190827e-05, "loss": 0.6495, "step": 815 }, { "epoch": 0.1, "learning_rate": 1.974990478386618e-05, "loss": 0.7558, "step": 816 }, { "epoch": 0.1, "learning_rate": 1.9749028640921584e-05, "loss": 0.6669, "step": 817 }, { "epoch": 0.1, "learning_rate": 1.974815098549297e-05, "loss": 0.6804, "step": 818 }, { "epoch": 0.1, "learning_rate": 1.974727181771649e-05, "loss": 0.6979, "step": 819 }, { "epoch": 0.1, "learning_rate": 1.974639113772855e-05, "loss": 0.6217, "step": 820 }, { "epoch": 0.1, "learning_rate": 1.974550894566577e-05, "loss": 0.6527, "step": 821 }, { "epoch": 0.1, "learning_rate": 1.9744625241665028e-05, "loss": 0.6205, "step": 822 }, { "epoch": 0.1, "learning_rate": 1.974374002586341e-05, "loss": 0.6877, "step": 823 }, { "epoch": 0.1, "learning_rate": 1.974285329839826e-05, "loss": 0.7459, "step": 824 }, { "epoch": 0.1, "learning_rate": 1.974196505940714e-05, "loss": 0.7007, "step": 825 }, { "epoch": 0.1, "learning_rate": 1.974107530902786e-05, "loss": 0.7006, "step": 826 }, { "epoch": 0.1, "learning_rate": 1.974018404739845e-05, "loss": 0.7033, "step": 827 }, { "epoch": 0.1, "learning_rate": 1.9739291274657187e-05, "loss": 0.7206, "step": 828 }, { "epoch": 0.1, "learning_rate": 1.9738396990942577e-05, "loss": 0.7073, "step": 829 }, { "epoch": 0.1, "learning_rate": 1.9737501196393362e-05, "loss": 0.7291, "step": 830 }, { "epoch": 0.1, "learning_rate": 1.9736603891148515e-05, "loss": 0.7095, "step": 831 }, { "epoch": 0.1, "learning_rate": 1.973570507534725e-05, "loss": 0.7028, "step": 832 }, { "epoch": 0.1, "learning_rate": 1.9734804749129e-05, "loss": 0.6594, "step": 833 }, { "epoch": 0.1, "learning_rate": 1.9733902912633457e-05, "loss": 0.6893, "step": 834 }, { "epoch": 0.1, "learning_rate": 1.9732999566000527e-05, "loss": 0.6519, "step": 835 }, { "epoch": 0.1, "learning_rate": 1.9732094709370358e-05, "loss": 0.791, "step": 836 }, { "epoch": 0.1, "learning_rate": 1.973118834288333e-05, "loss": 0.6809, "step": 837 }, { "epoch": 0.1, "learning_rate": 1.973028046668006e-05, "loss": 0.6588, "step": 838 }, { "epoch": 0.1, "learning_rate": 1.9729371080901403e-05, "loss": 0.735, "step": 839 }, { "epoch": 0.1, "learning_rate": 1.9728460185688433e-05, "loss": 0.6583, "step": 840 }, { "epoch": 0.1, "learning_rate": 1.9727547781182474e-05, "loss": 0.7204, "step": 841 }, { "epoch": 0.1, "learning_rate": 1.9726633867525075e-05, "loss": 0.7359, "step": 842 }, { "epoch": 0.1, "learning_rate": 1.9725718444858027e-05, "loss": 0.6554, "step": 843 }, { "epoch": 0.1, "learning_rate": 1.9724801513323346e-05, "loss": 0.663, "step": 844 }, { "epoch": 0.1, "learning_rate": 1.972388307306329e-05, "loss": 0.7213, "step": 845 }, { "epoch": 0.1, "learning_rate": 1.9722963124220346e-05, "loss": 0.7131, "step": 846 }, { "epoch": 0.1, "learning_rate": 1.9722041666937236e-05, "loss": 0.6766, "step": 847 }, { "epoch": 0.1, "learning_rate": 1.9721118701356917e-05, "loss": 0.7284, "step": 848 }, { "epoch": 0.1, "learning_rate": 1.9720194227622583e-05, "loss": 0.6694, "step": 849 }, { "epoch": 0.1, "learning_rate": 1.971926824587765e-05, "loss": 0.6134, "step": 850 }, { "epoch": 0.1, "learning_rate": 1.971834075626579e-05, "loss": 0.7027, "step": 851 }, { "epoch": 0.1, "learning_rate": 1.9717411758930887e-05, "loss": 0.7101, "step": 852 }, { "epoch": 0.1, "learning_rate": 1.9716481254017064e-05, "loss": 0.6537, "step": 853 }, { "epoch": 0.1, "learning_rate": 1.9715549241668693e-05, "loss": 0.767, "step": 854 }, { "epoch": 0.1, "learning_rate": 1.971461572203036e-05, "loss": 0.7328, "step": 855 }, { "epoch": 0.1, "learning_rate": 1.9713680695246892e-05, "loss": 0.7133, "step": 856 }, { "epoch": 0.1, "learning_rate": 1.9712744161463357e-05, "loss": 0.6764, "step": 857 }, { "epoch": 0.1, "learning_rate": 1.9711806120825045e-05, "loss": 0.7489, "step": 858 }, { "epoch": 0.1, "learning_rate": 1.971086657347749e-05, "loss": 0.645, "step": 859 }, { "epoch": 0.1, "learning_rate": 1.9709925519566453e-05, "loss": 0.7161, "step": 860 }, { "epoch": 0.1, "learning_rate": 1.970898295923793e-05, "loss": 0.7002, "step": 861 }, { "epoch": 0.1, "learning_rate": 1.9708038892638152e-05, "loss": 0.6815, "step": 862 }, { "epoch": 0.1, "learning_rate": 1.9707093319913587e-05, "loss": 0.6837, "step": 863 }, { "epoch": 0.11, "learning_rate": 1.970614624121093e-05, "loss": 0.6153, "step": 864 }, { "epoch": 0.11, "learning_rate": 1.9705197656677113e-05, "loss": 0.7028, "step": 865 }, { "epoch": 0.11, "learning_rate": 1.97042475664593e-05, "loss": 0.7378, "step": 866 }, { "epoch": 0.11, "learning_rate": 1.9703295970704892e-05, "loss": 0.7301, "step": 867 }, { "epoch": 0.11, "learning_rate": 1.9702342869561524e-05, "loss": 0.73, "step": 868 }, { "epoch": 0.11, "learning_rate": 1.9701388263177055e-05, "loss": 0.6434, "step": 869 }, { "epoch": 0.11, "learning_rate": 1.9700432151699592e-05, "loss": 0.7104, "step": 870 }, { "epoch": 0.11, "learning_rate": 1.9699474535277462e-05, "loss": 0.6726, "step": 871 }, { "epoch": 0.11, "learning_rate": 1.9698515414059234e-05, "loss": 0.6743, "step": 872 }, { "epoch": 0.11, "learning_rate": 1.969755478819371e-05, "loss": 0.6136, "step": 873 }, { "epoch": 0.11, "learning_rate": 1.969659265782992e-05, "loss": 0.6956, "step": 874 }, { "epoch": 0.11, "learning_rate": 1.9695629023117133e-05, "loss": 0.6857, "step": 875 }, { "epoch": 0.11, "learning_rate": 1.969466388420485e-05, "loss": 0.6287, "step": 876 }, { "epoch": 0.11, "learning_rate": 1.9693697241242802e-05, "loss": 0.6857, "step": 877 }, { "epoch": 0.11, "learning_rate": 1.9692729094380956e-05, "loss": 0.6435, "step": 878 }, { "epoch": 0.11, "learning_rate": 1.9691759443769514e-05, "loss": 0.6606, "step": 879 }, { "epoch": 0.11, "learning_rate": 1.969078828955891e-05, "loss": 0.7433, "step": 880 }, { "epoch": 0.11, "learning_rate": 1.968981563189981e-05, "loss": 0.7161, "step": 881 }, { "epoch": 0.11, "learning_rate": 1.968884147094311e-05, "loss": 0.7235, "step": 882 }, { "epoch": 0.11, "learning_rate": 1.9687865806839956e-05, "loss": 0.6737, "step": 883 }, { "epoch": 0.11, "learning_rate": 1.9686888639741698e-05, "loss": 0.6412, "step": 884 }, { "epoch": 0.11, "learning_rate": 1.9685909969799946e-05, "loss": 0.6622, "step": 885 }, { "epoch": 0.11, "learning_rate": 1.968492979716653e-05, "loss": 0.6858, "step": 886 }, { "epoch": 0.11, "learning_rate": 1.9683948121993518e-05, "loss": 0.7304, "step": 887 }, { "epoch": 0.11, "learning_rate": 1.9682964944433204e-05, "loss": 0.7074, "step": 888 }, { "epoch": 0.11, "learning_rate": 1.9681980264638125e-05, "loss": 0.6777, "step": 889 }, { "epoch": 0.11, "learning_rate": 1.9680994082761045e-05, "loss": 0.6927, "step": 890 }, { "epoch": 0.11, "learning_rate": 1.9680006398954962e-05, "loss": 0.7321, "step": 891 }, { "epoch": 0.11, "learning_rate": 1.9679017213373107e-05, "loss": 0.7011, "step": 892 }, { "epoch": 0.11, "learning_rate": 1.9678026526168943e-05, "loss": 0.7548, "step": 893 }, { "epoch": 0.11, "learning_rate": 1.967703433749617e-05, "loss": 0.6656, "step": 894 }, { "epoch": 0.11, "learning_rate": 1.9676040647508718e-05, "loss": 0.6911, "step": 895 }, { "epoch": 0.11, "learning_rate": 1.967504545636075e-05, "loss": 0.6398, "step": 896 }, { "epoch": 0.11, "learning_rate": 1.9674048764206657e-05, "loss": 0.6934, "step": 897 }, { "epoch": 0.11, "learning_rate": 1.9673050571201072e-05, "loss": 0.6493, "step": 898 }, { "epoch": 0.11, "learning_rate": 1.967205087749886e-05, "loss": 0.6871, "step": 899 }, { "epoch": 0.11, "learning_rate": 1.9671049683255108e-05, "loss": 0.6569, "step": 900 }, { "epoch": 0.11, "learning_rate": 1.967004698862515e-05, "loss": 0.7171, "step": 901 }, { "epoch": 0.11, "learning_rate": 1.9669042793764542e-05, "loss": 0.6759, "step": 902 }, { "epoch": 0.11, "learning_rate": 1.9668037098829077e-05, "loss": 0.727, "step": 903 }, { "epoch": 0.11, "learning_rate": 1.966702990397478e-05, "loss": 0.7062, "step": 904 }, { "epoch": 0.11, "learning_rate": 1.9666021209357917e-05, "loss": 0.6832, "step": 905 }, { "epoch": 0.11, "learning_rate": 1.966501101513497e-05, "loss": 0.692, "step": 906 }, { "epoch": 0.11, "learning_rate": 1.9663999321462662e-05, "loss": 0.7335, "step": 907 }, { "epoch": 0.11, "learning_rate": 1.9662986128497957e-05, "loss": 0.7344, "step": 908 }, { "epoch": 0.11, "learning_rate": 1.9661971436398035e-05, "loss": 0.7054, "step": 909 }, { "epoch": 0.11, "learning_rate": 1.9660955245320325e-05, "loss": 0.6311, "step": 910 }, { "epoch": 0.11, "learning_rate": 1.9659937555422478e-05, "loss": 0.6769, "step": 911 }, { "epoch": 0.11, "learning_rate": 1.965891836686238e-05, "loss": 0.6892, "step": 912 }, { "epoch": 0.11, "learning_rate": 1.965789767979815e-05, "loss": 0.6687, "step": 913 }, { "epoch": 0.11, "learning_rate": 1.9656875494388137e-05, "loss": 0.6958, "step": 914 }, { "epoch": 0.11, "learning_rate": 1.965585181079093e-05, "loss": 0.7237, "step": 915 }, { "epoch": 0.11, "learning_rate": 1.965482662916534e-05, "loss": 0.7261, "step": 916 }, { "epoch": 0.11, "learning_rate": 1.9653799949670422e-05, "loss": 0.6971, "step": 917 }, { "epoch": 0.11, "learning_rate": 1.9652771772465455e-05, "loss": 0.7187, "step": 918 }, { "epoch": 0.11, "learning_rate": 1.9651742097709948e-05, "loss": 0.7358, "step": 919 }, { "epoch": 0.11, "learning_rate": 1.965071092556365e-05, "loss": 0.7433, "step": 920 }, { "epoch": 0.11, "learning_rate": 1.964967825618654e-05, "loss": 0.6243, "step": 921 }, { "epoch": 0.11, "learning_rate": 1.9648644089738828e-05, "loss": 0.7377, "step": 922 }, { "epoch": 0.11, "learning_rate": 1.9647608426380956e-05, "loss": 0.6851, "step": 923 }, { "epoch": 0.11, "learning_rate": 1.96465712662736e-05, "loss": 0.6704, "step": 924 }, { "epoch": 0.11, "learning_rate": 1.9645532609577667e-05, "loss": 0.7122, "step": 925 }, { "epoch": 0.11, "learning_rate": 1.964449245645429e-05, "loss": 0.7295, "step": 926 }, { "epoch": 0.11, "learning_rate": 1.9643450807064853e-05, "loss": 0.6381, "step": 927 }, { "epoch": 0.11, "learning_rate": 1.964240766157095e-05, "loss": 0.7198, "step": 928 }, { "epoch": 0.11, "learning_rate": 1.9641363020134417e-05, "loss": 0.6896, "step": 929 }, { "epoch": 0.11, "learning_rate": 1.9640316882917328e-05, "loss": 0.704, "step": 930 }, { "epoch": 0.11, "learning_rate": 1.9639269250081977e-05, "loss": 0.6838, "step": 931 }, { "epoch": 0.11, "learning_rate": 1.96382201217909e-05, "loss": 0.6804, "step": 932 }, { "epoch": 0.11, "learning_rate": 1.9637169498206854e-05, "loss": 0.7422, "step": 933 }, { "epoch": 0.11, "learning_rate": 1.9636117379492844e-05, "loss": 0.7228, "step": 934 }, { "epoch": 0.11, "learning_rate": 1.9635063765812093e-05, "loss": 0.7175, "step": 935 }, { "epoch": 0.11, "learning_rate": 1.9634008657328057e-05, "loss": 0.6606, "step": 936 }, { "epoch": 0.11, "learning_rate": 1.9632952054204435e-05, "loss": 0.6506, "step": 937 }, { "epoch": 0.11, "learning_rate": 1.9631893956605148e-05, "loss": 0.7046, "step": 938 }, { "epoch": 0.11, "learning_rate": 1.963083436469435e-05, "loss": 0.7295, "step": 939 }, { "epoch": 0.11, "learning_rate": 1.962977327863643e-05, "loss": 0.6944, "step": 940 }, { "epoch": 0.11, "learning_rate": 1.9628710698596005e-05, "loss": 0.6688, "step": 941 }, { "epoch": 0.11, "learning_rate": 1.962764662473793e-05, "loss": 0.6456, "step": 942 }, { "epoch": 0.11, "learning_rate": 1.9626581057227278e-05, "loss": 0.6322, "step": 943 }, { "epoch": 0.11, "learning_rate": 1.9625513996229374e-05, "loss": 0.6736, "step": 944 }, { "epoch": 0.11, "learning_rate": 1.9624445441909758e-05, "loss": 0.6961, "step": 945 }, { "epoch": 0.12, "learning_rate": 1.962337539443421e-05, "loss": 0.6287, "step": 946 }, { "epoch": 0.12, "learning_rate": 1.962230385396874e-05, "loss": 0.6277, "step": 947 }, { "epoch": 0.12, "learning_rate": 1.9621230820679587e-05, "loss": 0.6665, "step": 948 }, { "epoch": 0.12, "learning_rate": 1.9620156294733227e-05, "loss": 0.6635, "step": 949 }, { "epoch": 0.12, "learning_rate": 1.961908027629636e-05, "loss": 0.7558, "step": 950 }, { "epoch": 0.12, "learning_rate": 1.961800276553592e-05, "loss": 0.6714, "step": 951 }, { "epoch": 0.12, "learning_rate": 1.961692376261908e-05, "loss": 0.7088, "step": 952 }, { "epoch": 0.12, "learning_rate": 1.9615843267713234e-05, "loss": 0.6782, "step": 953 }, { "epoch": 0.12, "learning_rate": 1.9614761280986013e-05, "loss": 0.6679, "step": 954 }, { "epoch": 0.12, "learning_rate": 1.9613677802605285e-05, "loss": 0.6919, "step": 955 }, { "epoch": 0.12, "learning_rate": 1.9612592832739135e-05, "loss": 0.6031, "step": 956 }, { "epoch": 0.12, "learning_rate": 1.9611506371555888e-05, "loss": 0.7066, "step": 957 }, { "epoch": 0.12, "learning_rate": 1.9610418419224104e-05, "loss": 0.714, "step": 958 }, { "epoch": 0.12, "learning_rate": 1.9609328975912566e-05, "loss": 0.6828, "step": 959 }, { "epoch": 0.12, "learning_rate": 1.96082380417903e-05, "loss": 0.6534, "step": 960 }, { "epoch": 0.12, "learning_rate": 1.9607145617026545e-05, "loss": 0.7044, "step": 961 }, { "epoch": 0.12, "learning_rate": 1.9606051701790793e-05, "loss": 0.6618, "step": 962 }, { "epoch": 0.12, "learning_rate": 1.9604956296252744e-05, "loss": 0.7075, "step": 963 }, { "epoch": 0.12, "learning_rate": 1.960385940058235e-05, "loss": 0.6201, "step": 964 }, { "epoch": 0.12, "learning_rate": 1.9602761014949784e-05, "loss": 0.6742, "step": 965 }, { "epoch": 0.12, "learning_rate": 1.960166113952545e-05, "loss": 0.6853, "step": 966 }, { "epoch": 0.12, "learning_rate": 1.960055977447999e-05, "loss": 0.6381, "step": 967 }, { "epoch": 0.12, "learning_rate": 1.9599456919984267e-05, "loss": 0.7225, "step": 968 }, { "epoch": 0.12, "learning_rate": 1.959835257620938e-05, "loss": 0.7013, "step": 969 }, { "epoch": 0.12, "learning_rate": 1.959724674332666e-05, "loss": 0.7264, "step": 970 }, { "epoch": 0.12, "learning_rate": 1.959613942150767e-05, "loss": 0.6924, "step": 971 }, { "epoch": 0.12, "learning_rate": 1.9595030610924197e-05, "loss": 0.6508, "step": 972 }, { "epoch": 0.12, "learning_rate": 1.959392031174827e-05, "loss": 0.6748, "step": 973 }, { "epoch": 0.12, "learning_rate": 1.959280852415214e-05, "loss": 0.7127, "step": 974 }, { "epoch": 0.12, "learning_rate": 1.95916952483083e-05, "loss": 0.6892, "step": 975 }, { "epoch": 0.12, "learning_rate": 1.959058048438945e-05, "loss": 0.6663, "step": 976 }, { "epoch": 0.12, "learning_rate": 1.958946423256855e-05, "loss": 0.6538, "step": 977 }, { "epoch": 0.12, "learning_rate": 1.9588346493018772e-05, "loss": 0.6321, "step": 978 }, { "epoch": 0.12, "learning_rate": 1.9587227265913524e-05, "loss": 0.7183, "step": 979 }, { "epoch": 0.12, "learning_rate": 1.9586106551426443e-05, "loss": 0.7169, "step": 980 }, { "epoch": 0.12, "learning_rate": 1.958498434973141e-05, "loss": 0.7085, "step": 981 }, { "epoch": 0.12, "learning_rate": 1.9583860661002518e-05, "loss": 0.7027, "step": 982 }, { "epoch": 0.12, "learning_rate": 1.958273548541409e-05, "loss": 0.6633, "step": 983 }, { "epoch": 0.12, "learning_rate": 1.9581608823140703e-05, "loss": 0.6449, "step": 984 }, { "epoch": 0.12, "learning_rate": 1.9580480674357145e-05, "loss": 0.7094, "step": 985 }, { "epoch": 0.12, "learning_rate": 1.9579351039238436e-05, "loss": 0.6731, "step": 986 }, { "epoch": 0.12, "learning_rate": 1.9578219917959827e-05, "loss": 0.7372, "step": 987 }, { "epoch": 0.12, "learning_rate": 1.957708731069681e-05, "loss": 0.692, "step": 988 }, { "epoch": 0.12, "learning_rate": 1.9575953217625097e-05, "loss": 0.7244, "step": 989 }, { "epoch": 0.12, "learning_rate": 1.9574817638920632e-05, "loss": 0.6549, "step": 990 }, { "epoch": 0.12, "learning_rate": 1.957368057475959e-05, "loss": 0.6556, "step": 991 }, { "epoch": 0.12, "learning_rate": 1.9572542025318385e-05, "loss": 0.6451, "step": 992 }, { "epoch": 0.12, "learning_rate": 1.9571401990773645e-05, "loss": 0.6691, "step": 993 }, { "epoch": 0.12, "learning_rate": 1.957026047130224e-05, "loss": 0.6537, "step": 994 }, { "epoch": 0.12, "learning_rate": 1.956911746708127e-05, "loss": 0.6687, "step": 995 }, { "epoch": 0.12, "learning_rate": 1.956797297828806e-05, "loss": 0.6633, "step": 996 }, { "epoch": 0.12, "learning_rate": 1.956682700510017e-05, "loss": 0.6551, "step": 997 }, { "epoch": 0.12, "learning_rate": 1.9565679547695387e-05, "loss": 0.6547, "step": 998 }, { "epoch": 0.12, "learning_rate": 1.9564530606251737e-05, "loss": 0.7204, "step": 999 }, { "epoch": 0.12, "learning_rate": 1.956338018094746e-05, "loss": 0.7205, "step": 1000 }, { "epoch": 0.12, "learning_rate": 1.9562228271961036e-05, "loss": 0.6804, "step": 1001 }, { "epoch": 0.12, "learning_rate": 1.956107487947118e-05, "loss": 0.6831, "step": 1002 }, { "epoch": 0.12, "learning_rate": 1.9559920003656828e-05, "loss": 0.6626, "step": 1003 }, { "epoch": 0.12, "learning_rate": 1.955876364469715e-05, "loss": 0.65, "step": 1004 }, { "epoch": 0.12, "learning_rate": 1.9557605802771546e-05, "loss": 0.6228, "step": 1005 }, { "epoch": 0.12, "learning_rate": 1.9556446478059646e-05, "loss": 0.6971, "step": 1006 }, { "epoch": 0.12, "learning_rate": 1.955528567074131e-05, "loss": 0.6445, "step": 1007 }, { "epoch": 0.12, "learning_rate": 1.9554123380996632e-05, "loss": 0.6797, "step": 1008 }, { "epoch": 0.12, "learning_rate": 1.9552959609005923e-05, "loss": 0.6791, "step": 1009 }, { "epoch": 0.12, "learning_rate": 1.955179435494974e-05, "loss": 0.6834, "step": 1010 }, { "epoch": 0.12, "learning_rate": 1.955062761900886e-05, "loss": 0.7157, "step": 1011 }, { "epoch": 0.12, "learning_rate": 1.9549459401364296e-05, "loss": 0.6619, "step": 1012 }, { "epoch": 0.12, "learning_rate": 1.9548289702197286e-05, "loss": 0.7059, "step": 1013 }, { "epoch": 0.12, "learning_rate": 1.95471185216893e-05, "loss": 0.6426, "step": 1014 }, { "epoch": 0.12, "learning_rate": 1.954594586002203e-05, "loss": 0.7225, "step": 1015 }, { "epoch": 0.12, "learning_rate": 1.9544771717377418e-05, "loss": 0.7317, "step": 1016 }, { "epoch": 0.12, "learning_rate": 1.954359609393761e-05, "loss": 0.7555, "step": 1017 }, { "epoch": 0.12, "learning_rate": 1.9542418989885007e-05, "loss": 0.6595, "step": 1018 }, { "epoch": 0.12, "learning_rate": 1.9541240405402218e-05, "loss": 0.7091, "step": 1019 }, { "epoch": 0.12, "learning_rate": 1.9540060340672094e-05, "loss": 0.7167, "step": 1020 }, { "epoch": 0.12, "learning_rate": 1.9538878795877714e-05, "loss": 0.6804, "step": 1021 }, { "epoch": 0.12, "learning_rate": 1.9537695771202384e-05, "loss": 0.6672, "step": 1022 }, { "epoch": 0.12, "learning_rate": 1.9536511266829637e-05, "loss": 0.6723, "step": 1023 }, { "epoch": 0.12, "learning_rate": 1.9535325282943246e-05, "loss": 0.6562, "step": 1024 }, { "epoch": 0.12, "learning_rate": 1.9534137819727205e-05, "loss": 0.6768, "step": 1025 }, { "epoch": 0.12, "learning_rate": 1.9532948877365738e-05, "loss": 0.7162, "step": 1026 }, { "epoch": 0.12, "learning_rate": 1.95317584560433e-05, "loss": 0.7045, "step": 1027 }, { "epoch": 0.13, "learning_rate": 1.9530566555944575e-05, "loss": 0.6328, "step": 1028 }, { "epoch": 0.13, "learning_rate": 1.952937317725448e-05, "loss": 0.6856, "step": 1029 }, { "epoch": 0.13, "learning_rate": 1.9528178320158156e-05, "loss": 0.6643, "step": 1030 }, { "epoch": 0.13, "learning_rate": 1.9526981984840975e-05, "loss": 0.6252, "step": 1031 }, { "epoch": 0.13, "learning_rate": 1.9525784171488537e-05, "loss": 0.6172, "step": 1032 }, { "epoch": 0.13, "learning_rate": 1.952458488028668e-05, "loss": 0.6539, "step": 1033 }, { "epoch": 0.13, "learning_rate": 1.9523384111421455e-05, "loss": 0.6851, "step": 1034 }, { "epoch": 0.13, "learning_rate": 1.952218186507916e-05, "loss": 0.6717, "step": 1035 }, { "epoch": 0.13, "learning_rate": 1.952097814144631e-05, "loss": 0.694, "step": 1036 }, { "epoch": 0.13, "learning_rate": 1.9519772940709654e-05, "loss": 0.6975, "step": 1037 }, { "epoch": 0.13, "learning_rate": 1.951856626305617e-05, "loss": 0.7042, "step": 1038 }, { "epoch": 0.13, "learning_rate": 1.9517358108673056e-05, "loss": 0.6248, "step": 1039 }, { "epoch": 0.13, "learning_rate": 1.951614847774776e-05, "loss": 0.7417, "step": 1040 }, { "epoch": 0.13, "learning_rate": 1.9514937370467945e-05, "loss": 0.6017, "step": 1041 }, { "epoch": 0.13, "learning_rate": 1.9513724787021498e-05, "loss": 0.6904, "step": 1042 }, { "epoch": 0.13, "learning_rate": 1.9512510727596545e-05, "loss": 0.657, "step": 1043 }, { "epoch": 0.13, "learning_rate": 1.9511295192381437e-05, "loss": 0.6796, "step": 1044 }, { "epoch": 0.13, "learning_rate": 1.9510078181564753e-05, "loss": 0.64, "step": 1045 }, { "epoch": 0.13, "learning_rate": 1.950885969533531e-05, "loss": 0.6948, "step": 1046 }, { "epoch": 0.13, "learning_rate": 1.9507639733882135e-05, "loss": 0.6531, "step": 1047 }, { "epoch": 0.13, "learning_rate": 1.9506418297394504e-05, "loss": 0.6504, "step": 1048 }, { "epoch": 0.13, "learning_rate": 1.9505195386061914e-05, "loss": 0.6486, "step": 1049 }, { "epoch": 0.13, "learning_rate": 1.950397100007408e-05, "loss": 0.6122, "step": 1050 }, { "epoch": 0.13, "learning_rate": 1.9502745139620966e-05, "loss": 0.6505, "step": 1051 }, { "epoch": 0.13, "learning_rate": 1.9501517804892746e-05, "loss": 0.695, "step": 1052 }, { "epoch": 0.13, "learning_rate": 1.950028899607984e-05, "loss": 0.6504, "step": 1053 }, { "epoch": 0.13, "learning_rate": 1.9499058713372886e-05, "loss": 0.6557, "step": 1054 }, { "epoch": 0.13, "learning_rate": 1.9497826956962746e-05, "loss": 0.7072, "step": 1055 }, { "epoch": 0.13, "learning_rate": 1.9496593727040518e-05, "loss": 0.7031, "step": 1056 }, { "epoch": 0.13, "learning_rate": 1.9495359023797537e-05, "loss": 0.6436, "step": 1057 }, { "epoch": 0.13, "learning_rate": 1.9494122847425348e-05, "loss": 0.6686, "step": 1058 }, { "epoch": 0.13, "learning_rate": 1.949288519811574e-05, "loss": 0.7023, "step": 1059 }, { "epoch": 0.13, "learning_rate": 1.949164607606072e-05, "loss": 0.6663, "step": 1060 }, { "epoch": 0.13, "learning_rate": 1.949040548145253e-05, "loss": 0.6688, "step": 1061 }, { "epoch": 0.13, "learning_rate": 1.9489163414483638e-05, "loss": 0.6573, "step": 1062 }, { "epoch": 0.13, "learning_rate": 1.948791987534674e-05, "loss": 0.662, "step": 1063 }, { "epoch": 0.13, "learning_rate": 1.9486674864234763e-05, "loss": 0.7127, "step": 1064 }, { "epoch": 0.13, "learning_rate": 1.948542838134086e-05, "loss": 0.6956, "step": 1065 }, { "epoch": 0.13, "learning_rate": 1.9484180426858416e-05, "loss": 0.638, "step": 1066 }, { "epoch": 0.13, "learning_rate": 1.9482931000981028e-05, "loss": 0.6592, "step": 1067 }, { "epoch": 0.13, "learning_rate": 1.9481680103902552e-05, "loss": 0.6221, "step": 1068 }, { "epoch": 0.13, "learning_rate": 1.9480427735817045e-05, "loss": 0.6991, "step": 1069 }, { "epoch": 0.13, "learning_rate": 1.94791738969188e-05, "loss": 0.7294, "step": 1070 }, { "epoch": 0.13, "learning_rate": 1.9477918587402348e-05, "loss": 0.602, "step": 1071 }, { "epoch": 0.13, "learning_rate": 1.9476661807462437e-05, "loss": 0.6224, "step": 1072 }, { "epoch": 0.13, "learning_rate": 1.947540355729404e-05, "loss": 0.7151, "step": 1073 }, { "epoch": 0.13, "learning_rate": 1.9474143837092373e-05, "loss": 0.6302, "step": 1074 }, { "epoch": 0.13, "learning_rate": 1.947288264705287e-05, "loss": 0.6643, "step": 1075 }, { "epoch": 0.13, "learning_rate": 1.9471619987371193e-05, "loss": 0.6596, "step": 1076 }, { "epoch": 0.13, "learning_rate": 1.947035585824323e-05, "loss": 0.6327, "step": 1077 }, { "epoch": 0.13, "learning_rate": 1.9469090259865106e-05, "loss": 0.6639, "step": 1078 }, { "epoch": 0.13, "learning_rate": 1.9467823192433166e-05, "loss": 0.6821, "step": 1079 }, { "epoch": 0.13, "learning_rate": 1.9466554656143988e-05, "loss": 0.6804, "step": 1080 }, { "epoch": 0.13, "learning_rate": 1.946528465119437e-05, "loss": 0.6258, "step": 1081 }, { "epoch": 0.13, "learning_rate": 1.9464013177781347e-05, "loss": 0.6048, "step": 1082 }, { "epoch": 0.13, "learning_rate": 1.9462740236102183e-05, "loss": 0.6695, "step": 1083 }, { "epoch": 0.13, "learning_rate": 1.9461465826354353e-05, "loss": 0.6499, "step": 1084 }, { "epoch": 0.13, "learning_rate": 1.946018994873558e-05, "loss": 0.669, "step": 1085 }, { "epoch": 0.13, "learning_rate": 1.9458912603443806e-05, "loss": 0.6872, "step": 1086 }, { "epoch": 0.13, "learning_rate": 1.9457633790677194e-05, "loss": 0.6886, "step": 1087 }, { "epoch": 0.13, "learning_rate": 1.9456353510634148e-05, "loss": 0.6779, "step": 1088 }, { "epoch": 0.13, "learning_rate": 1.9455071763513293e-05, "loss": 0.5779, "step": 1089 }, { "epoch": 0.13, "learning_rate": 1.945378854951348e-05, "loss": 0.6539, "step": 1090 }, { "epoch": 0.13, "learning_rate": 1.945250386883379e-05, "loss": 0.7348, "step": 1091 }, { "epoch": 0.13, "learning_rate": 1.945121772167353e-05, "loss": 0.7027, "step": 1092 }, { "epoch": 0.13, "learning_rate": 1.9449930108232237e-05, "loss": 0.6807, "step": 1093 }, { "epoch": 0.13, "learning_rate": 1.9448641028709674e-05, "loss": 0.6045, "step": 1094 }, { "epoch": 0.13, "learning_rate": 1.944735048330583e-05, "loss": 0.6694, "step": 1095 }, { "epoch": 0.13, "learning_rate": 1.9446058472220924e-05, "loss": 0.637, "step": 1096 }, { "epoch": 0.13, "learning_rate": 1.94447649956554e-05, "loss": 0.5713, "step": 1097 }, { "epoch": 0.13, "learning_rate": 1.9443470053809932e-05, "loss": 0.6562, "step": 1098 }, { "epoch": 0.13, "learning_rate": 1.944217364688542e-05, "loss": 0.6164, "step": 1099 }, { "epoch": 0.13, "learning_rate": 1.9440875775082984e-05, "loss": 0.6025, "step": 1100 }, { "epoch": 0.13, "learning_rate": 1.943957643860399e-05, "loss": 0.663, "step": 1101 }, { "epoch": 0.13, "learning_rate": 1.9438275637650016e-05, "loss": 0.6675, "step": 1102 }, { "epoch": 0.13, "learning_rate": 1.943697337242287e-05, "loss": 0.6538, "step": 1103 }, { "epoch": 0.13, "learning_rate": 1.9435669643124584e-05, "loss": 0.697, "step": 1104 }, { "epoch": 0.13, "learning_rate": 1.9434364449957427e-05, "loss": 0.7378, "step": 1105 }, { "epoch": 0.13, "learning_rate": 1.9433057793123885e-05, "loss": 0.6665, "step": 1106 }, { "epoch": 0.13, "learning_rate": 1.943174967282668e-05, "loss": 0.6751, "step": 1107 }, { "epoch": 0.13, "learning_rate": 1.943044008926875e-05, "loss": 0.6643, "step": 1108 }, { "epoch": 0.13, "learning_rate": 1.9429129042653275e-05, "loss": 0.6603, "step": 1109 }, { "epoch": 0.13, "learning_rate": 1.9427816533183646e-05, "loss": 0.6733, "step": 1110 }, { "epoch": 0.14, "learning_rate": 1.942650256106349e-05, "loss": 0.6405, "step": 1111 }, { "epoch": 0.14, "learning_rate": 1.9425187126496662e-05, "loss": 0.6572, "step": 1112 }, { "epoch": 0.14, "learning_rate": 1.9423870229687238e-05, "loss": 0.6924, "step": 1113 }, { "epoch": 0.14, "learning_rate": 1.942255187083953e-05, "loss": 0.659, "step": 1114 }, { "epoch": 0.14, "learning_rate": 1.942123205015806e-05, "loss": 0.6906, "step": 1115 }, { "epoch": 0.14, "learning_rate": 1.9419910767847595e-05, "loss": 0.7288, "step": 1116 }, { "epoch": 0.14, "learning_rate": 1.9418588024113125e-05, "loss": 0.6279, "step": 1117 }, { "epoch": 0.14, "learning_rate": 1.9417263819159858e-05, "loss": 0.6799, "step": 1118 }, { "epoch": 0.14, "learning_rate": 1.941593815319323e-05, "loss": 0.6512, "step": 1119 }, { "epoch": 0.14, "learning_rate": 1.9414611026418916e-05, "loss": 0.6606, "step": 1120 }, { "epoch": 0.14, "learning_rate": 1.94132824390428e-05, "loss": 0.7256, "step": 1121 }, { "epoch": 0.14, "learning_rate": 1.9411952391271015e-05, "loss": 0.6721, "step": 1122 }, { "epoch": 0.14, "learning_rate": 1.9410620883309893e-05, "loss": 0.6009, "step": 1123 }, { "epoch": 0.14, "learning_rate": 1.9409287915366014e-05, "loss": 0.639, "step": 1124 }, { "epoch": 0.14, "learning_rate": 1.9407953487646178e-05, "loss": 0.7366, "step": 1125 }, { "epoch": 0.14, "learning_rate": 1.9406617600357408e-05, "loss": 0.6123, "step": 1126 }, { "epoch": 0.14, "learning_rate": 1.9405280253706955e-05, "loss": 0.6466, "step": 1127 }, { "epoch": 0.14, "learning_rate": 1.9403941447902303e-05, "loss": 0.7153, "step": 1128 }, { "epoch": 0.14, "learning_rate": 1.9402601183151155e-05, "loss": 0.5871, "step": 1129 }, { "epoch": 0.14, "learning_rate": 1.9401259459661438e-05, "loss": 0.6678, "step": 1130 }, { "epoch": 0.14, "learning_rate": 1.9399916277641314e-05, "loss": 0.6919, "step": 1131 }, { "epoch": 0.14, "learning_rate": 1.9398571637299168e-05, "loss": 0.6579, "step": 1132 }, { "epoch": 0.14, "learning_rate": 1.939722553884361e-05, "loss": 0.6657, "step": 1133 }, { "epoch": 0.14, "learning_rate": 1.9395877982483472e-05, "loss": 0.7024, "step": 1134 }, { "epoch": 0.14, "learning_rate": 1.9394528968427817e-05, "loss": 0.6443, "step": 1135 }, { "epoch": 0.14, "learning_rate": 1.939317849688594e-05, "loss": 0.6266, "step": 1136 }, { "epoch": 0.14, "learning_rate": 1.939182656806735e-05, "loss": 0.6468, "step": 1137 }, { "epoch": 0.14, "learning_rate": 1.9390473182181794e-05, "loss": 0.7064, "step": 1138 }, { "epoch": 0.14, "learning_rate": 1.938911833943923e-05, "loss": 0.6575, "step": 1139 }, { "epoch": 0.14, "learning_rate": 1.938776204004986e-05, "loss": 0.6207, "step": 1140 }, { "epoch": 0.14, "learning_rate": 1.9386404284224096e-05, "loss": 0.655, "step": 1141 }, { "epoch": 0.14, "learning_rate": 1.9385045072172585e-05, "loss": 0.7038, "step": 1142 }, { "epoch": 0.14, "learning_rate": 1.9383684404106204e-05, "loss": 0.6091, "step": 1143 }, { "epoch": 0.14, "learning_rate": 1.9382322280236042e-05, "loss": 0.6361, "step": 1144 }, { "epoch": 0.14, "learning_rate": 1.9380958700773425e-05, "loss": 0.6606, "step": 1145 }, { "epoch": 0.14, "learning_rate": 1.93795936659299e-05, "loss": 0.7237, "step": 1146 }, { "epoch": 0.14, "learning_rate": 1.937822717591724e-05, "loss": 0.6578, "step": 1147 }, { "epoch": 0.14, "learning_rate": 1.9376859230947455e-05, "loss": 0.7304, "step": 1148 }, { "epoch": 0.14, "learning_rate": 1.9375489831232755e-05, "loss": 0.6008, "step": 1149 }, { "epoch": 0.14, "learning_rate": 1.9374118976985603e-05, "loss": 0.6991, "step": 1150 }, { "epoch": 0.14, "learning_rate": 1.937274666841867e-05, "loss": 0.6981, "step": 1151 }, { "epoch": 0.14, "learning_rate": 1.9371372905744866e-05, "loss": 0.6125, "step": 1152 }, { "epoch": 0.14, "learning_rate": 1.936999768917731e-05, "loss": 0.6837, "step": 1153 }, { "epoch": 0.14, "learning_rate": 1.9368621018929364e-05, "loss": 0.6371, "step": 1154 }, { "epoch": 0.14, "learning_rate": 1.9367242895214602e-05, "loss": 0.7096, "step": 1155 }, { "epoch": 0.14, "learning_rate": 1.9365863318246832e-05, "loss": 0.6804, "step": 1156 }, { "epoch": 0.14, "learning_rate": 1.9364482288240082e-05, "loss": 0.6339, "step": 1157 }, { "epoch": 0.14, "learning_rate": 1.9363099805408608e-05, "loss": 0.677, "step": 1158 }, { "epoch": 0.14, "learning_rate": 1.9361715869966892e-05, "loss": 0.6932, "step": 1159 }, { "epoch": 0.14, "learning_rate": 1.9360330482129643e-05, "loss": 0.6741, "step": 1160 }, { "epoch": 0.14, "learning_rate": 1.935894364211179e-05, "loss": 0.6415, "step": 1161 }, { "epoch": 0.14, "learning_rate": 1.9357555350128488e-05, "loss": 0.6779, "step": 1162 }, { "epoch": 0.14, "learning_rate": 1.9356165606395122e-05, "loss": 0.6214, "step": 1163 }, { "epoch": 0.14, "learning_rate": 1.9354774411127304e-05, "loss": 0.6556, "step": 1164 }, { "epoch": 0.14, "learning_rate": 1.935338176454086e-05, "loss": 0.7268, "step": 1165 }, { "epoch": 0.14, "learning_rate": 1.935198766685185e-05, "loss": 0.6902, "step": 1166 }, { "epoch": 0.14, "learning_rate": 1.935059211827656e-05, "loss": 0.6364, "step": 1167 }, { "epoch": 0.14, "learning_rate": 1.9349195119031494e-05, "loss": 0.6249, "step": 1168 }, { "epoch": 0.14, "learning_rate": 1.9347796669333387e-05, "loss": 0.6782, "step": 1169 }, { "epoch": 0.14, "learning_rate": 1.93463967693992e-05, "loss": 0.6651, "step": 1170 }, { "epoch": 0.14, "learning_rate": 1.9344995419446113e-05, "loss": 0.67, "step": 1171 }, { "epoch": 0.14, "learning_rate": 1.9343592619691534e-05, "loss": 0.6788, "step": 1172 }, { "epoch": 0.14, "learning_rate": 1.9342188370353098e-05, "loss": 0.6131, "step": 1173 }, { "epoch": 0.14, "learning_rate": 1.934078267164866e-05, "loss": 0.6805, "step": 1174 }, { "epoch": 0.14, "learning_rate": 1.933937552379631e-05, "loss": 0.6179, "step": 1175 }, { "epoch": 0.14, "learning_rate": 1.9337966927014346e-05, "loss": 0.6635, "step": 1176 }, { "epoch": 0.14, "learning_rate": 1.933655688152131e-05, "loss": 0.6555, "step": 1177 }, { "epoch": 0.14, "learning_rate": 1.9335145387535954e-05, "loss": 0.5884, "step": 1178 }, { "epoch": 0.14, "learning_rate": 1.933373244527726e-05, "loss": 0.7101, "step": 1179 }, { "epoch": 0.14, "learning_rate": 1.9332318054964436e-05, "loss": 0.6808, "step": 1180 }, { "epoch": 0.14, "learning_rate": 1.9330902216816914e-05, "loss": 0.6123, "step": 1181 }, { "epoch": 0.14, "learning_rate": 1.932948493105435e-05, "loss": 0.6828, "step": 1182 }, { "epoch": 0.14, "learning_rate": 1.9328066197896623e-05, "loss": 0.6306, "step": 1183 }, { "epoch": 0.14, "learning_rate": 1.9326646017563836e-05, "loss": 0.6862, "step": 1184 }, { "epoch": 0.14, "learning_rate": 1.9325224390276324e-05, "loss": 0.6418, "step": 1185 }, { "epoch": 0.14, "learning_rate": 1.9323801316254637e-05, "loss": 0.6252, "step": 1186 }, { "epoch": 0.14, "learning_rate": 1.9322376795719557e-05, "loss": 0.7163, "step": 1187 }, { "epoch": 0.14, "learning_rate": 1.9320950828892083e-05, "loss": 0.674, "step": 1188 }, { "epoch": 0.14, "learning_rate": 1.9319523415993445e-05, "loss": 0.6407, "step": 1189 }, { "epoch": 0.14, "learning_rate": 1.9318094557245098e-05, "loss": 0.6626, "step": 1190 }, { "epoch": 0.14, "learning_rate": 1.9316664252868712e-05, "loss": 0.6453, "step": 1191 }, { "epoch": 0.14, "learning_rate": 1.9315232503086188e-05, "loss": 0.6646, "step": 1192 }, { "epoch": 0.15, "learning_rate": 1.931379930811965e-05, "loss": 0.6172, "step": 1193 }, { "epoch": 0.15, "learning_rate": 1.9312364668191454e-05, "loss": 0.6169, "step": 1194 }, { "epoch": 0.15, "learning_rate": 1.931092858352417e-05, "loss": 0.6559, "step": 1195 }, { "epoch": 0.15, "learning_rate": 1.9309491054340586e-05, "loss": 0.6818, "step": 1196 }, { "epoch": 0.15, "learning_rate": 1.9308052080863736e-05, "loss": 0.6733, "step": 1197 }, { "epoch": 0.15, "learning_rate": 1.9306611663316857e-05, "loss": 0.6533, "step": 1198 }, { "epoch": 0.15, "learning_rate": 1.930516980192342e-05, "loss": 0.6687, "step": 1199 }, { "epoch": 0.15, "learning_rate": 1.930372649690712e-05, "loss": 0.6449, "step": 1200 }, { "epoch": 0.15, "learning_rate": 1.930228174849187e-05, "loss": 0.6979, "step": 1201 }, { "epoch": 0.15, "learning_rate": 1.9300835556901817e-05, "loss": 0.7101, "step": 1202 }, { "epoch": 0.15, "learning_rate": 1.9299387922361327e-05, "loss": 0.6283, "step": 1203 }, { "epoch": 0.15, "learning_rate": 1.929793884509498e-05, "loss": 0.6332, "step": 1204 }, { "epoch": 0.15, "learning_rate": 1.9296488325327596e-05, "loss": 0.6563, "step": 1205 }, { "epoch": 0.15, "learning_rate": 1.929503636328421e-05, "loss": 0.6226, "step": 1206 }, { "epoch": 0.15, "learning_rate": 1.9293582959190086e-05, "loss": 0.6479, "step": 1207 }, { "epoch": 0.15, "learning_rate": 1.92921281132707e-05, "loss": 0.6264, "step": 1208 }, { "epoch": 0.15, "learning_rate": 1.9290671825751765e-05, "loss": 0.7158, "step": 1209 }, { "epoch": 0.15, "learning_rate": 1.928921409685921e-05, "loss": 0.6269, "step": 1210 }, { "epoch": 0.15, "learning_rate": 1.9287754926819195e-05, "loss": 0.5877, "step": 1211 }, { "epoch": 0.15, "learning_rate": 1.928629431585809e-05, "loss": 0.5957, "step": 1212 }, { "epoch": 0.15, "learning_rate": 1.9284832264202505e-05, "loss": 0.6025, "step": 1213 }, { "epoch": 0.15, "learning_rate": 1.9283368772079264e-05, "loss": 0.6488, "step": 1214 }, { "epoch": 0.15, "learning_rate": 1.9281903839715413e-05, "loss": 0.6894, "step": 1215 }, { "epoch": 0.15, "learning_rate": 1.9280437467338225e-05, "loss": 0.662, "step": 1216 }, { "epoch": 0.15, "learning_rate": 1.92789696551752e-05, "loss": 0.7008, "step": 1217 }, { "epoch": 0.15, "learning_rate": 1.9277500403454054e-05, "loss": 0.7198, "step": 1218 }, { "epoch": 0.15, "learning_rate": 1.927602971240273e-05, "loss": 0.6353, "step": 1219 }, { "epoch": 0.15, "learning_rate": 1.9274557582249393e-05, "loss": 0.5741, "step": 1220 }, { "epoch": 0.15, "learning_rate": 1.9273084013222435e-05, "loss": 0.6757, "step": 1221 }, { "epoch": 0.15, "learning_rate": 1.927160900555047e-05, "loss": 0.6354, "step": 1222 }, { "epoch": 0.15, "learning_rate": 1.9270132559462325e-05, "loss": 0.6112, "step": 1223 }, { "epoch": 0.15, "learning_rate": 1.9268654675187067e-05, "loss": 0.6689, "step": 1224 }, { "epoch": 0.15, "learning_rate": 1.926717535295398e-05, "loss": 0.63, "step": 1225 }, { "epoch": 0.15, "learning_rate": 1.926569459299256e-05, "loss": 0.635, "step": 1226 }, { "epoch": 0.15, "learning_rate": 1.9264212395532542e-05, "loss": 0.629, "step": 1227 }, { "epoch": 0.15, "learning_rate": 1.9262728760803876e-05, "loss": 0.6325, "step": 1228 }, { "epoch": 0.15, "learning_rate": 1.926124368903673e-05, "loss": 0.6356, "step": 1229 }, { "epoch": 0.15, "learning_rate": 1.925975718046151e-05, "loss": 0.724, "step": 1230 }, { "epoch": 0.15, "learning_rate": 1.925826923530883e-05, "loss": 0.6963, "step": 1231 }, { "epoch": 0.15, "learning_rate": 1.9256779853809537e-05, "loss": 0.5998, "step": 1232 }, { "epoch": 0.15, "learning_rate": 1.9255289036194692e-05, "loss": 0.6062, "step": 1233 }, { "epoch": 0.15, "learning_rate": 1.9253796782695588e-05, "loss": 0.6551, "step": 1234 }, { "epoch": 0.15, "learning_rate": 1.9252303093543734e-05, "loss": 0.6013, "step": 1235 }, { "epoch": 0.15, "learning_rate": 1.9250807968970863e-05, "loss": 0.6354, "step": 1236 }, { "epoch": 0.15, "learning_rate": 1.9249311409208934e-05, "loss": 0.6696, "step": 1237 }, { "epoch": 0.15, "learning_rate": 1.9247813414490122e-05, "loss": 0.6589, "step": 1238 }, { "epoch": 0.15, "learning_rate": 1.9246313985046833e-05, "loss": 0.7092, "step": 1239 }, { "epoch": 0.15, "learning_rate": 1.9244813121111693e-05, "loss": 0.7323, "step": 1240 }, { "epoch": 0.15, "learning_rate": 1.9243310822917543e-05, "loss": 0.6258, "step": 1241 }, { "epoch": 0.15, "learning_rate": 1.924180709069746e-05, "loss": 0.6912, "step": 1242 }, { "epoch": 0.15, "learning_rate": 1.924030192468473e-05, "loss": 0.5929, "step": 1243 }, { "epoch": 0.15, "learning_rate": 1.9238795325112867e-05, "loss": 0.7157, "step": 1244 }, { "epoch": 0.15, "learning_rate": 1.9237287292215617e-05, "loss": 0.6669, "step": 1245 }, { "epoch": 0.15, "learning_rate": 1.9235777826226928e-05, "loss": 0.7014, "step": 1246 }, { "epoch": 0.15, "learning_rate": 1.923426692738099e-05, "loss": 0.6579, "step": 1247 }, { "epoch": 0.15, "learning_rate": 1.92327545959122e-05, "loss": 0.5911, "step": 1248 }, { "epoch": 0.15, "learning_rate": 1.9231240832055188e-05, "loss": 0.6666, "step": 1249 }, { "epoch": 0.15, "learning_rate": 1.9229725636044804e-05, "loss": 0.6822, "step": 1250 }, { "epoch": 0.15, "learning_rate": 1.9228209008116115e-05, "loss": 0.6711, "step": 1251 }, { "epoch": 0.15, "learning_rate": 1.9226690948504414e-05, "loss": 0.6811, "step": 1252 }, { "epoch": 0.15, "learning_rate": 1.922517145744522e-05, "loss": 0.6462, "step": 1253 }, { "epoch": 0.15, "learning_rate": 1.9223650535174264e-05, "loss": 0.6202, "step": 1254 }, { "epoch": 0.15, "learning_rate": 1.9222128181927512e-05, "loss": 0.6302, "step": 1255 }, { "epoch": 0.15, "learning_rate": 1.9220604397941135e-05, "loss": 0.6251, "step": 1256 }, { "epoch": 0.15, "learning_rate": 1.9219079183451547e-05, "loss": 0.6115, "step": 1257 }, { "epoch": 0.15, "learning_rate": 1.9217552538695362e-05, "loss": 0.6637, "step": 1258 }, { "epoch": 0.15, "learning_rate": 1.9216024463909438e-05, "loss": 0.5842, "step": 1259 }, { "epoch": 0.15, "learning_rate": 1.9214494959330835e-05, "loss": 0.7308, "step": 1260 }, { "epoch": 0.15, "learning_rate": 1.921296402519685e-05, "loss": 0.6309, "step": 1261 }, { "epoch": 0.15, "learning_rate": 1.921143166174499e-05, "loss": 0.6698, "step": 1262 }, { "epoch": 0.15, "learning_rate": 1.920989786921299e-05, "loss": 0.6197, "step": 1263 }, { "epoch": 0.15, "learning_rate": 1.9208362647838804e-05, "loss": 0.702, "step": 1264 }, { "epoch": 0.15, "learning_rate": 1.9206825997860617e-05, "loss": 0.6775, "step": 1265 }, { "epoch": 0.15, "learning_rate": 1.920528791951682e-05, "loss": 0.6527, "step": 1266 }, { "epoch": 0.15, "learning_rate": 1.9203748413046033e-05, "loss": 0.6922, "step": 1267 }, { "epoch": 0.15, "learning_rate": 1.9202207478687107e-05, "loss": 0.7501, "step": 1268 }, { "epoch": 0.15, "learning_rate": 1.92006651166791e-05, "loss": 0.5973, "step": 1269 }, { "epoch": 0.15, "learning_rate": 1.919912132726129e-05, "loss": 0.6992, "step": 1270 }, { "epoch": 0.15, "learning_rate": 1.9197576110673198e-05, "loss": 0.5609, "step": 1271 }, { "epoch": 0.15, "learning_rate": 1.9196029467154544e-05, "loss": 0.5694, "step": 1272 }, { "epoch": 0.15, "learning_rate": 1.9194481396945276e-05, "loss": 0.6363, "step": 1273 }, { "epoch": 0.15, "learning_rate": 1.919293190028557e-05, "loss": 0.6575, "step": 1274 }, { "epoch": 0.16, "learning_rate": 1.919138097741581e-05, "loss": 0.6567, "step": 1275 }, { "epoch": 0.16, "learning_rate": 1.9189828628576618e-05, "loss": 0.7025, "step": 1276 }, { "epoch": 0.16, "learning_rate": 1.9188274854008822e-05, "loss": 0.7056, "step": 1277 }, { "epoch": 0.16, "learning_rate": 1.9186719653953483e-05, "loss": 0.6522, "step": 1278 }, { "epoch": 0.16, "learning_rate": 1.9185163028651878e-05, "loss": 0.6484, "step": 1279 }, { "epoch": 0.16, "learning_rate": 1.9183604978345502e-05, "loss": 0.6492, "step": 1280 }, { "epoch": 0.16, "learning_rate": 1.918204550327607e-05, "loss": 0.7159, "step": 1281 }, { "epoch": 0.16, "learning_rate": 1.9180484603685528e-05, "loss": 0.6365, "step": 1282 }, { "epoch": 0.16, "learning_rate": 1.917892227981604e-05, "loss": 0.7321, "step": 1283 }, { "epoch": 0.16, "learning_rate": 1.917735853190998e-05, "loss": 0.6375, "step": 1284 }, { "epoch": 0.16, "learning_rate": 1.9175793360209954e-05, "loss": 0.6792, "step": 1285 }, { "epoch": 0.16, "learning_rate": 1.9174226764958793e-05, "loss": 0.6772, "step": 1286 }, { "epoch": 0.16, "learning_rate": 1.917265874639953e-05, "loss": 0.641, "step": 1287 }, { "epoch": 0.16, "learning_rate": 1.9171089304775438e-05, "loss": 0.6927, "step": 1288 }, { "epoch": 0.16, "learning_rate": 1.9169518440330002e-05, "loss": 0.6326, "step": 1289 }, { "epoch": 0.16, "learning_rate": 1.9167946153306928e-05, "loss": 0.6455, "step": 1290 }, { "epoch": 0.16, "learning_rate": 1.9166372443950144e-05, "loss": 0.6705, "step": 1291 }, { "epoch": 0.16, "learning_rate": 1.91647973125038e-05, "loss": 0.5955, "step": 1292 }, { "epoch": 0.16, "learning_rate": 1.916322075921226e-05, "loss": 0.643, "step": 1293 }, { "epoch": 0.16, "learning_rate": 1.9161642784320123e-05, "loss": 0.662, "step": 1294 }, { "epoch": 0.16, "learning_rate": 1.9160063388072192e-05, "loss": 0.6492, "step": 1295 }, { "epoch": 0.16, "learning_rate": 1.91584825707135e-05, "loss": 0.6929, "step": 1296 }, { "epoch": 0.16, "learning_rate": 1.9156900332489298e-05, "loss": 0.6281, "step": 1297 }, { "epoch": 0.16, "learning_rate": 1.9155316673645052e-05, "loss": 0.6754, "step": 1298 }, { "epoch": 0.16, "learning_rate": 1.9153731594426466e-05, "loss": 0.6151, "step": 1299 }, { "epoch": 0.16, "learning_rate": 1.915214509507944e-05, "loss": 0.6952, "step": 1300 }, { "epoch": 0.16, "learning_rate": 1.915055717585012e-05, "loss": 0.6572, "step": 1301 }, { "epoch": 0.16, "learning_rate": 1.9148967836984845e-05, "loss": 0.6445, "step": 1302 }, { "epoch": 0.16, "learning_rate": 1.91473770787302e-05, "loss": 0.6324, "step": 1303 }, { "epoch": 0.16, "learning_rate": 1.914578490133297e-05, "loss": 0.636, "step": 1304 }, { "epoch": 0.16, "learning_rate": 1.914419130504017e-05, "loss": 0.6876, "step": 1305 }, { "epoch": 0.16, "learning_rate": 1.914259629009904e-05, "loss": 0.6269, "step": 1306 }, { "epoch": 0.16, "learning_rate": 1.9140999856757027e-05, "loss": 0.6062, "step": 1307 }, { "epoch": 0.16, "learning_rate": 1.913940200526181e-05, "loss": 0.6616, "step": 1308 }, { "epoch": 0.16, "learning_rate": 1.9137802735861276e-05, "loss": 0.6295, "step": 1309 }, { "epoch": 0.16, "learning_rate": 1.9136202048803545e-05, "loss": 0.7285, "step": 1310 }, { "epoch": 0.16, "learning_rate": 1.913459994433695e-05, "loss": 0.6632, "step": 1311 }, { "epoch": 0.16, "learning_rate": 1.9132996422710044e-05, "loss": 0.6658, "step": 1312 }, { "epoch": 0.16, "learning_rate": 1.9131391484171598e-05, "loss": 0.6767, "step": 1313 }, { "epoch": 0.16, "learning_rate": 1.9129785128970612e-05, "loss": 0.6508, "step": 1314 }, { "epoch": 0.16, "learning_rate": 1.912817735735629e-05, "loss": 0.7066, "step": 1315 }, { "epoch": 0.16, "learning_rate": 1.9126568169578073e-05, "loss": 0.6203, "step": 1316 }, { "epoch": 0.16, "learning_rate": 1.912495756588561e-05, "loss": 0.613, "step": 1317 }, { "epoch": 0.16, "learning_rate": 1.912334554652877e-05, "loss": 0.6286, "step": 1318 }, { "epoch": 0.16, "learning_rate": 1.9121732111757654e-05, "loss": 0.6555, "step": 1319 }, { "epoch": 0.16, "learning_rate": 1.9120117261822563e-05, "loss": 0.6227, "step": 1320 }, { "epoch": 0.16, "learning_rate": 1.911850099697404e-05, "loss": 0.6269, "step": 1321 }, { "epoch": 0.16, "learning_rate": 1.9116883317462826e-05, "loss": 0.6143, "step": 1322 }, { "epoch": 0.16, "learning_rate": 1.911526422353989e-05, "loss": 0.6145, "step": 1323 }, { "epoch": 0.16, "learning_rate": 1.9113643715456432e-05, "loss": 0.6394, "step": 1324 }, { "epoch": 0.16, "learning_rate": 1.911202179346385e-05, "loss": 0.65, "step": 1325 }, { "epoch": 0.16, "learning_rate": 1.911039845781378e-05, "loss": 0.6523, "step": 1326 }, { "epoch": 0.16, "learning_rate": 1.9108773708758063e-05, "loss": 0.6539, "step": 1327 }, { "epoch": 0.16, "learning_rate": 1.910714754654877e-05, "loss": 0.6224, "step": 1328 }, { "epoch": 0.16, "learning_rate": 1.9105519971438188e-05, "loss": 0.6281, "step": 1329 }, { "epoch": 0.16, "learning_rate": 1.9103890983678817e-05, "loss": 0.6712, "step": 1330 }, { "epoch": 0.16, "learning_rate": 1.910226058352339e-05, "loss": 0.6673, "step": 1331 }, { "epoch": 0.16, "learning_rate": 1.910062877122484e-05, "loss": 0.6216, "step": 1332 }, { "epoch": 0.16, "learning_rate": 1.909899554703634e-05, "loss": 0.6256, "step": 1333 }, { "epoch": 0.16, "learning_rate": 1.9097360911211263e-05, "loss": 0.6836, "step": 1334 }, { "epoch": 0.16, "learning_rate": 1.9095724864003218e-05, "loss": 0.6348, "step": 1335 }, { "epoch": 0.16, "learning_rate": 1.909408740566602e-05, "loss": 0.6996, "step": 1336 }, { "epoch": 0.16, "learning_rate": 1.9092448536453702e-05, "loss": 0.686, "step": 1337 }, { "epoch": 0.16, "learning_rate": 1.9090808256620536e-05, "loss": 0.6897, "step": 1338 }, { "epoch": 0.16, "learning_rate": 1.908916656642098e-05, "loss": 0.7066, "step": 1339 }, { "epoch": 0.16, "learning_rate": 1.9087523466109745e-05, "loss": 0.5999, "step": 1340 }, { "epoch": 0.16, "learning_rate": 1.908587895594174e-05, "loss": 0.7376, "step": 1341 }, { "epoch": 0.16, "learning_rate": 1.9084233036172094e-05, "loss": 0.6497, "step": 1342 }, { "epoch": 0.16, "learning_rate": 1.9082585707056157e-05, "loss": 0.6461, "step": 1343 }, { "epoch": 0.16, "learning_rate": 1.908093696884951e-05, "loss": 0.5593, "step": 1344 }, { "epoch": 0.16, "learning_rate": 1.907928682180793e-05, "loss": 0.6234, "step": 1345 }, { "epoch": 0.16, "learning_rate": 1.907763526618743e-05, "loss": 0.5637, "step": 1346 }, { "epoch": 0.16, "learning_rate": 1.9075982302244235e-05, "loss": 0.6836, "step": 1347 }, { "epoch": 0.16, "learning_rate": 1.9074327930234783e-05, "loss": 0.6188, "step": 1348 }, { "epoch": 0.16, "learning_rate": 1.9072672150415746e-05, "loss": 0.6319, "step": 1349 }, { "epoch": 0.16, "learning_rate": 1.9071014963044002e-05, "loss": 0.6967, "step": 1350 }, { "epoch": 0.16, "learning_rate": 1.9069356368376645e-05, "loss": 0.6579, "step": 1351 }, { "epoch": 0.16, "learning_rate": 1.9067696366671e-05, "loss": 0.6142, "step": 1352 }, { "epoch": 0.16, "learning_rate": 1.90660349581846e-05, "loss": 0.7081, "step": 1353 }, { "epoch": 0.16, "learning_rate": 1.90643721431752e-05, "loss": 0.6645, "step": 1354 }, { "epoch": 0.16, "learning_rate": 1.906270792190077e-05, "loss": 0.6684, "step": 1355 }, { "epoch": 0.16, "learning_rate": 1.9061042294619503e-05, "loss": 0.7828, "step": 1356 }, { "epoch": 0.17, "learning_rate": 1.9059375261589806e-05, "loss": 0.6959, "step": 1357 }, { "epoch": 0.17, "learning_rate": 1.905770682307031e-05, "loss": 0.6943, "step": 1358 }, { "epoch": 0.17, "learning_rate": 1.905603697931986e-05, "loss": 0.682, "step": 1359 }, { "epoch": 0.17, "learning_rate": 1.9054365730597507e-05, "loss": 0.5921, "step": 1360 }, { "epoch": 0.17, "learning_rate": 1.905269307716255e-05, "loss": 0.6333, "step": 1361 }, { "epoch": 0.17, "learning_rate": 1.905101901927447e-05, "loss": 0.6377, "step": 1362 }, { "epoch": 0.17, "learning_rate": 1.9049343557193e-05, "loss": 0.661, "step": 1363 }, { "epoch": 0.17, "learning_rate": 1.9047666691178063e-05, "loss": 0.6542, "step": 1364 }, { "epoch": 0.17, "learning_rate": 1.904598842148982e-05, "loss": 0.6123, "step": 1365 }, { "epoch": 0.17, "learning_rate": 1.904430874838863e-05, "loss": 0.6826, "step": 1366 }, { "epoch": 0.17, "learning_rate": 1.904262767213509e-05, "loss": 0.6588, "step": 1367 }, { "epoch": 0.17, "learning_rate": 1.9040945192990005e-05, "loss": 0.6486, "step": 1368 }, { "epoch": 0.17, "learning_rate": 1.9039261311214395e-05, "loss": 0.5914, "step": 1369 }, { "epoch": 0.17, "learning_rate": 1.9037576027069502e-05, "loss": 0.6878, "step": 1370 }, { "epoch": 0.17, "learning_rate": 1.9035889340816783e-05, "loss": 0.6503, "step": 1371 }, { "epoch": 0.17, "learning_rate": 1.903420125271792e-05, "loss": 0.6018, "step": 1372 }, { "epoch": 0.17, "learning_rate": 1.9032511763034795e-05, "loss": 0.6304, "step": 1373 }, { "epoch": 0.17, "learning_rate": 1.903082087202953e-05, "loss": 0.6854, "step": 1374 }, { "epoch": 0.17, "learning_rate": 1.902912857996445e-05, "loss": 0.5794, "step": 1375 }, { "epoch": 0.17, "learning_rate": 1.9027434887102095e-05, "loss": 0.5841, "step": 1376 }, { "epoch": 0.17, "learning_rate": 1.9025739793705237e-05, "loss": 0.6866, "step": 1377 }, { "epoch": 0.17, "learning_rate": 1.902404330003685e-05, "loss": 0.713, "step": 1378 }, { "epoch": 0.17, "learning_rate": 1.9022345406360135e-05, "loss": 0.6141, "step": 1379 }, { "epoch": 0.17, "learning_rate": 1.9020646112938502e-05, "loss": 0.6677, "step": 1380 }, { "epoch": 0.17, "learning_rate": 1.9018945420035587e-05, "loss": 0.633, "step": 1381 }, { "epoch": 0.17, "learning_rate": 1.9017243327915235e-05, "loss": 0.6163, "step": 1382 }, { "epoch": 0.17, "learning_rate": 1.9015539836841516e-05, "loss": 0.6902, "step": 1383 }, { "epoch": 0.17, "learning_rate": 1.9013834947078713e-05, "loss": 0.5944, "step": 1384 }, { "epoch": 0.17, "learning_rate": 1.901212865889132e-05, "loss": 0.6235, "step": 1385 }, { "epoch": 0.17, "learning_rate": 1.9010420972544065e-05, "loss": 0.6305, "step": 1386 }, { "epoch": 0.17, "learning_rate": 1.9008711888301873e-05, "loss": 0.6283, "step": 1387 }, { "epoch": 0.17, "learning_rate": 1.90070014064299e-05, "loss": 0.6185, "step": 1388 }, { "epoch": 0.17, "learning_rate": 1.9005289527193505e-05, "loss": 0.6737, "step": 1389 }, { "epoch": 0.17, "learning_rate": 1.9003576250858284e-05, "loss": 0.6274, "step": 1390 }, { "epoch": 0.17, "learning_rate": 1.9001861577690028e-05, "loss": 0.6758, "step": 1391 }, { "epoch": 0.17, "learning_rate": 1.900014550795476e-05, "loss": 0.6654, "step": 1392 }, { "epoch": 0.17, "learning_rate": 1.8998428041918716e-05, "loss": 0.6311, "step": 1393 }, { "epoch": 0.17, "learning_rate": 1.8996709179848346e-05, "loss": 0.6621, "step": 1394 }, { "epoch": 0.17, "learning_rate": 1.8994988922010316e-05, "loss": 0.6338, "step": 1395 }, { "epoch": 0.17, "learning_rate": 1.899326726867151e-05, "loss": 0.6268, "step": 1396 }, { "epoch": 0.17, "learning_rate": 1.899154422009903e-05, "loss": 0.6447, "step": 1397 }, { "epoch": 0.17, "learning_rate": 1.8989819776560193e-05, "loss": 0.6683, "step": 1398 }, { "epoch": 0.17, "learning_rate": 1.8988093938322533e-05, "loss": 0.5565, "step": 1399 }, { "epoch": 0.17, "learning_rate": 1.89863667056538e-05, "loss": 0.6271, "step": 1400 }, { "epoch": 0.17, "learning_rate": 1.898463807882196e-05, "loss": 0.6594, "step": 1401 }, { "epoch": 0.17, "learning_rate": 1.8982908058095197e-05, "loss": 0.6081, "step": 1402 }, { "epoch": 0.17, "learning_rate": 1.8981176643741907e-05, "loss": 0.6947, "step": 1403 }, { "epoch": 0.17, "learning_rate": 1.897944383603071e-05, "loss": 0.5954, "step": 1404 }, { "epoch": 0.17, "learning_rate": 1.8977709635230432e-05, "loss": 0.7283, "step": 1405 }, { "epoch": 0.17, "learning_rate": 1.8975974041610122e-05, "loss": 0.5676, "step": 1406 }, { "epoch": 0.17, "learning_rate": 1.8974237055439045e-05, "loss": 0.5951, "step": 1407 }, { "epoch": 0.17, "learning_rate": 1.8972498676986678e-05, "loss": 0.5949, "step": 1408 }, { "epoch": 0.17, "learning_rate": 1.897075890652272e-05, "loss": 0.5743, "step": 1409 }, { "epoch": 0.17, "learning_rate": 1.896901774431708e-05, "loss": 0.636, "step": 1410 }, { "epoch": 0.17, "learning_rate": 1.8967275190639888e-05, "loss": 0.6986, "step": 1411 }, { "epoch": 0.17, "learning_rate": 1.896553124576148e-05, "loss": 0.6873, "step": 1412 }, { "epoch": 0.17, "learning_rate": 1.8963785909952427e-05, "loss": 0.6231, "step": 1413 }, { "epoch": 0.17, "learning_rate": 1.8962039183483493e-05, "loss": 0.6207, "step": 1414 }, { "epoch": 0.17, "learning_rate": 1.8960291066625674e-05, "loss": 0.6777, "step": 1415 }, { "epoch": 0.17, "learning_rate": 1.895854155965018e-05, "loss": 0.637, "step": 1416 }, { "epoch": 0.17, "learning_rate": 1.8956790662828425e-05, "loss": 0.6326, "step": 1417 }, { "epoch": 0.17, "learning_rate": 1.8955038376432048e-05, "loss": 0.6358, "step": 1418 }, { "epoch": 0.17, "learning_rate": 1.8953284700732908e-05, "loss": 0.6247, "step": 1419 }, { "epoch": 0.17, "learning_rate": 1.8951529636003066e-05, "loss": 0.7099, "step": 1420 }, { "epoch": 0.17, "learning_rate": 1.8949773182514816e-05, "loss": 0.6456, "step": 1421 }, { "epoch": 0.17, "learning_rate": 1.894801534054065e-05, "loss": 0.6768, "step": 1422 }, { "epoch": 0.17, "learning_rate": 1.8946256110353287e-05, "loss": 0.5774, "step": 1423 }, { "epoch": 0.17, "learning_rate": 1.8944495492225656e-05, "loss": 0.6223, "step": 1424 }, { "epoch": 0.17, "learning_rate": 1.89427334864309e-05, "loss": 0.5689, "step": 1425 }, { "epoch": 0.17, "learning_rate": 1.8940970093242385e-05, "loss": 0.6338, "step": 1426 }, { "epoch": 0.17, "learning_rate": 1.8939205312933687e-05, "loss": 0.617, "step": 1427 }, { "epoch": 0.17, "learning_rate": 1.8937439145778595e-05, "loss": 0.627, "step": 1428 }, { "epoch": 0.17, "learning_rate": 1.8935671592051113e-05, "loss": 0.6814, "step": 1429 }, { "epoch": 0.17, "learning_rate": 1.8933902652025472e-05, "loss": 0.621, "step": 1430 }, { "epoch": 0.17, "learning_rate": 1.8932132325976102e-05, "loss": 0.6633, "step": 1431 }, { "epoch": 0.17, "learning_rate": 1.8930360614177656e-05, "loss": 0.6672, "step": 1432 }, { "epoch": 0.17, "learning_rate": 1.8928587516905002e-05, "loss": 0.6337, "step": 1433 }, { "epoch": 0.17, "learning_rate": 1.892681303443322e-05, "loss": 0.6699, "step": 1434 }, { "epoch": 0.17, "learning_rate": 1.8925037167037608e-05, "loss": 0.6365, "step": 1435 }, { "epoch": 0.17, "learning_rate": 1.892325991499368e-05, "loss": 0.5707, "step": 1436 }, { "epoch": 0.17, "learning_rate": 1.892148127857716e-05, "loss": 0.6812, "step": 1437 }, { "epoch": 0.17, "learning_rate": 1.8919701258063987e-05, "loss": 0.6435, "step": 1438 }, { "epoch": 0.17, "learning_rate": 1.891791985373032e-05, "loss": 0.6468, "step": 1439 }, { "epoch": 0.18, "learning_rate": 1.891613706585253e-05, "loss": 0.642, "step": 1440 }, { "epoch": 0.18, "learning_rate": 1.8914352894707206e-05, "loss": 0.6641, "step": 1441 }, { "epoch": 0.18, "learning_rate": 1.8912567340571137e-05, "loss": 0.6478, "step": 1442 }, { "epoch": 0.18, "learning_rate": 1.8910780403721348e-05, "loss": 0.5692, "step": 1443 }, { "epoch": 0.18, "learning_rate": 1.8908992084435063e-05, "loss": 0.622, "step": 1444 }, { "epoch": 0.18, "learning_rate": 1.8907202382989725e-05, "loss": 0.6371, "step": 1445 }, { "epoch": 0.18, "learning_rate": 1.8905411299662995e-05, "loss": 0.6509, "step": 1446 }, { "epoch": 0.18, "learning_rate": 1.8903618834732744e-05, "loss": 0.6121, "step": 1447 }, { "epoch": 0.18, "learning_rate": 1.8901824988477053e-05, "loss": 0.6602, "step": 1448 }, { "epoch": 0.18, "learning_rate": 1.8900029761174233e-05, "loss": 0.6161, "step": 1449 }, { "epoch": 0.18, "learning_rate": 1.8898233153102793e-05, "loss": 0.7003, "step": 1450 }, { "epoch": 0.18, "learning_rate": 1.8896435164541466e-05, "loss": 0.6279, "step": 1451 }, { "epoch": 0.18, "learning_rate": 1.889463579576919e-05, "loss": 0.6141, "step": 1452 }, { "epoch": 0.18, "learning_rate": 1.889283504706513e-05, "loss": 0.6034, "step": 1453 }, { "epoch": 0.18, "learning_rate": 1.8891032918708652e-05, "loss": 0.654, "step": 1454 }, { "epoch": 0.18, "learning_rate": 1.8889229410979344e-05, "loss": 0.6327, "step": 1455 }, { "epoch": 0.18, "learning_rate": 1.8887424524157002e-05, "loss": 0.7138, "step": 1456 }, { "epoch": 0.18, "learning_rate": 1.888561825852165e-05, "loss": 0.6563, "step": 1457 }, { "epoch": 0.18, "learning_rate": 1.8883810614353503e-05, "loss": 0.6298, "step": 1458 }, { "epoch": 0.18, "learning_rate": 1.8882001591933013e-05, "loss": 0.64, "step": 1459 }, { "epoch": 0.18, "learning_rate": 1.8880191191540825e-05, "loss": 0.6917, "step": 1460 }, { "epoch": 0.18, "learning_rate": 1.887837941345782e-05, "loss": 0.5908, "step": 1461 }, { "epoch": 0.18, "learning_rate": 1.8876566257965075e-05, "loss": 0.6796, "step": 1462 }, { "epoch": 0.18, "learning_rate": 1.8874751725343885e-05, "loss": 0.6801, "step": 1463 }, { "epoch": 0.18, "learning_rate": 1.887293581587576e-05, "loss": 0.6419, "step": 1464 }, { "epoch": 0.18, "learning_rate": 1.8871118529842428e-05, "loss": 0.5993, "step": 1465 }, { "epoch": 0.18, "learning_rate": 1.8869299867525825e-05, "loss": 0.6837, "step": 1466 }, { "epoch": 0.18, "learning_rate": 1.88674798292081e-05, "loss": 0.6183, "step": 1467 }, { "epoch": 0.18, "learning_rate": 1.886565841517162e-05, "loss": 0.6149, "step": 1468 }, { "epoch": 0.18, "learning_rate": 1.8863835625698962e-05, "loss": 0.6213, "step": 1469 }, { "epoch": 0.18, "learning_rate": 1.8862011461072915e-05, "loss": 0.6408, "step": 1470 }, { "epoch": 0.18, "learning_rate": 1.8860185921576483e-05, "loss": 0.7074, "step": 1471 }, { "epoch": 0.18, "learning_rate": 1.885835900749289e-05, "loss": 0.699, "step": 1472 }, { "epoch": 0.18, "learning_rate": 1.8856530719105565e-05, "loss": 0.6172, "step": 1473 }, { "epoch": 0.18, "learning_rate": 1.8854701056698146e-05, "loss": 0.6477, "step": 1474 }, { "epoch": 0.18, "learning_rate": 1.8852870020554496e-05, "loss": 0.6668, "step": 1475 }, { "epoch": 0.18, "learning_rate": 1.8851037610958686e-05, "loss": 0.5665, "step": 1476 }, { "epoch": 0.18, "learning_rate": 1.8849203828195e-05, "loss": 0.7171, "step": 1477 }, { "epoch": 0.18, "learning_rate": 1.8847368672547932e-05, "loss": 0.6705, "step": 1478 }, { "epoch": 0.18, "learning_rate": 1.8845532144302195e-05, "loss": 0.6919, "step": 1479 }, { "epoch": 0.18, "learning_rate": 1.8843694243742707e-05, "loss": 0.5591, "step": 1480 }, { "epoch": 0.18, "learning_rate": 1.8841854971154608e-05, "loss": 0.692, "step": 1481 }, { "epoch": 0.18, "learning_rate": 1.8840014326823245e-05, "loss": 0.6712, "step": 1482 }, { "epoch": 0.18, "learning_rate": 1.883817231103418e-05, "loss": 0.6668, "step": 1483 }, { "epoch": 0.18, "learning_rate": 1.8836328924073184e-05, "loss": 0.6825, "step": 1484 }, { "epoch": 0.18, "learning_rate": 1.883448416622625e-05, "loss": 0.5852, "step": 1485 }, { "epoch": 0.18, "learning_rate": 1.883263803777957e-05, "loss": 0.7207, "step": 1486 }, { "epoch": 0.18, "learning_rate": 1.883079053901956e-05, "loss": 0.6389, "step": 1487 }, { "epoch": 0.18, "learning_rate": 1.8828941670232845e-05, "loss": 0.6972, "step": 1488 }, { "epoch": 0.18, "learning_rate": 1.8827091431706262e-05, "loss": 0.6154, "step": 1489 }, { "epoch": 0.18, "learning_rate": 1.882523982372686e-05, "loss": 0.6247, "step": 1490 }, { "epoch": 0.18, "learning_rate": 1.88233868465819e-05, "loss": 0.6129, "step": 1491 }, { "epoch": 0.18, "learning_rate": 1.882153250055886e-05, "loss": 0.6902, "step": 1492 }, { "epoch": 0.18, "learning_rate": 1.8819676785945426e-05, "loss": 0.6802, "step": 1493 }, { "epoch": 0.18, "learning_rate": 1.8817819703029496e-05, "loss": 0.6077, "step": 1494 }, { "epoch": 0.18, "learning_rate": 1.881596125209918e-05, "loss": 0.6372, "step": 1495 }, { "epoch": 0.18, "learning_rate": 1.8814101433442805e-05, "loss": 0.5976, "step": 1496 }, { "epoch": 0.18, "learning_rate": 1.8812240247348906e-05, "loss": 0.6532, "step": 1497 }, { "epoch": 0.18, "learning_rate": 1.8810377694106233e-05, "loss": 0.5743, "step": 1498 }, { "epoch": 0.18, "learning_rate": 1.8808513774003742e-05, "loss": 0.6839, "step": 1499 }, { "epoch": 0.18, "learning_rate": 1.880664848733061e-05, "loss": 0.6702, "step": 1500 }, { "epoch": 0.18, "learning_rate": 1.8804781834376216e-05, "loss": 0.6199, "step": 1501 }, { "epoch": 0.18, "learning_rate": 1.8802913815430163e-05, "loss": 0.6005, "step": 1502 }, { "epoch": 0.18, "learning_rate": 1.880104443078225e-05, "loss": 0.6251, "step": 1503 }, { "epoch": 0.18, "learning_rate": 1.8799173680722508e-05, "loss": 0.5851, "step": 1504 }, { "epoch": 0.18, "learning_rate": 1.8797301565541164e-05, "loss": 0.6099, "step": 1505 }, { "epoch": 0.18, "learning_rate": 1.879542808552866e-05, "loss": 0.6644, "step": 1506 }, { "epoch": 0.18, "learning_rate": 1.879355324097565e-05, "loss": 0.594, "step": 1507 }, { "epoch": 0.18, "learning_rate": 1.8791677032173007e-05, "loss": 0.6221, "step": 1508 }, { "epoch": 0.18, "learning_rate": 1.8789799459411807e-05, "loss": 0.5735, "step": 1509 }, { "epoch": 0.18, "learning_rate": 1.878792052298334e-05, "loss": 0.6254, "step": 1510 }, { "epoch": 0.18, "learning_rate": 1.8786040223179106e-05, "loss": 0.6403, "step": 1511 }, { "epoch": 0.18, "learning_rate": 1.8784158560290825e-05, "loss": 0.6408, "step": 1512 }, { "epoch": 0.18, "learning_rate": 1.8782275534610414e-05, "loss": 0.6354, "step": 1513 }, { "epoch": 0.18, "learning_rate": 1.8780391146430017e-05, "loss": 0.6437, "step": 1514 }, { "epoch": 0.18, "learning_rate": 1.8778505396041974e-05, "loss": 0.625, "step": 1515 }, { "epoch": 0.18, "learning_rate": 1.877661828373885e-05, "loss": 0.6163, "step": 1516 }, { "epoch": 0.18, "learning_rate": 1.8774729809813416e-05, "loss": 0.6559, "step": 1517 }, { "epoch": 0.18, "learning_rate": 1.877283997455865e-05, "loss": 0.6235, "step": 1518 }, { "epoch": 0.18, "learning_rate": 1.8770948778267744e-05, "loss": 0.6195, "step": 1519 }, { "epoch": 0.18, "learning_rate": 1.8769056221234105e-05, "loss": 0.6765, "step": 1520 }, { "epoch": 0.18, "learning_rate": 1.8767162303751346e-05, "loss": 0.6237, "step": 1521 }, { "epoch": 0.19, "learning_rate": 1.87652670261133e-05, "loss": 0.6264, "step": 1522 }, { "epoch": 0.19, "learning_rate": 1.8763370388613995e-05, "loss": 0.594, "step": 1523 }, { "epoch": 0.19, "learning_rate": 1.876147239154768e-05, "loss": 0.6678, "step": 1524 }, { "epoch": 0.19, "learning_rate": 1.875957303520882e-05, "loss": 0.6514, "step": 1525 }, { "epoch": 0.19, "learning_rate": 1.875767231989208e-05, "loss": 0.7422, "step": 1526 }, { "epoch": 0.19, "learning_rate": 1.8755770245892345e-05, "loss": 0.615, "step": 1527 }, { "epoch": 0.19, "learning_rate": 1.8753866813504707e-05, "loss": 0.6198, "step": 1528 }, { "epoch": 0.19, "learning_rate": 1.8751962023024458e-05, "loss": 0.5908, "step": 1529 }, { "epoch": 0.19, "learning_rate": 1.8750055874747125e-05, "loss": 0.6101, "step": 1530 }, { "epoch": 0.19, "learning_rate": 1.8748148368968425e-05, "loss": 0.6567, "step": 1531 }, { "epoch": 0.19, "learning_rate": 1.8746239505984293e-05, "loss": 0.5958, "step": 1532 }, { "epoch": 0.19, "learning_rate": 1.874432928609087e-05, "loss": 0.6375, "step": 1533 }, { "epoch": 0.19, "learning_rate": 1.8742417709584516e-05, "loss": 0.5921, "step": 1534 }, { "epoch": 0.19, "learning_rate": 1.8740504776761795e-05, "loss": 0.6089, "step": 1535 }, { "epoch": 0.19, "learning_rate": 1.873859048791949e-05, "loss": 0.6154, "step": 1536 }, { "epoch": 0.19, "learning_rate": 1.8736674843354576e-05, "loss": 0.679, "step": 1537 }, { "epoch": 0.19, "learning_rate": 1.8734757843364256e-05, "loss": 0.634, "step": 1538 }, { "epoch": 0.19, "learning_rate": 1.873283948824594e-05, "loss": 0.654, "step": 1539 }, { "epoch": 0.19, "learning_rate": 1.873091977829724e-05, "loss": 0.6813, "step": 1540 }, { "epoch": 0.19, "learning_rate": 1.872899871381599e-05, "loss": 0.6159, "step": 1541 }, { "epoch": 0.19, "learning_rate": 1.872707629510022e-05, "loss": 0.6332, "step": 1542 }, { "epoch": 0.19, "learning_rate": 1.872515252244818e-05, "loss": 0.6241, "step": 1543 }, { "epoch": 0.19, "learning_rate": 1.8723227396158337e-05, "loss": 0.5508, "step": 1544 }, { "epoch": 0.19, "learning_rate": 1.8721300916529347e-05, "loss": 0.6301, "step": 1545 }, { "epoch": 0.19, "learning_rate": 1.8719373083860092e-05, "loss": 0.6516, "step": 1546 }, { "epoch": 0.19, "learning_rate": 1.8717443898449662e-05, "loss": 0.6331, "step": 1547 }, { "epoch": 0.19, "learning_rate": 1.8715513360597355e-05, "loss": 0.6, "step": 1548 }, { "epoch": 0.19, "learning_rate": 1.8713581470602674e-05, "loss": 0.6381, "step": 1549 }, { "epoch": 0.19, "learning_rate": 1.871164822876534e-05, "loss": 0.6668, "step": 1550 }, { "epoch": 0.19, "learning_rate": 1.8709713635385282e-05, "loss": 0.6534, "step": 1551 }, { "epoch": 0.19, "learning_rate": 1.870777769076263e-05, "loss": 0.7198, "step": 1552 }, { "epoch": 0.19, "learning_rate": 1.8705840395197742e-05, "loss": 0.685, "step": 1553 }, { "epoch": 0.19, "learning_rate": 1.870390174899116e-05, "loss": 0.6353, "step": 1554 }, { "epoch": 0.19, "learning_rate": 1.8701961752443658e-05, "loss": 0.6522, "step": 1555 }, { "epoch": 0.19, "learning_rate": 1.8700020405856208e-05, "loss": 0.6526, "step": 1556 }, { "epoch": 0.19, "learning_rate": 1.8698077709529996e-05, "loss": 0.6098, "step": 1557 }, { "epoch": 0.19, "learning_rate": 1.8696133663766414e-05, "loss": 0.6125, "step": 1558 }, { "epoch": 0.19, "learning_rate": 1.8694188268867063e-05, "loss": 0.6132, "step": 1559 }, { "epoch": 0.19, "learning_rate": 1.8692241525133765e-05, "loss": 0.5817, "step": 1560 }, { "epoch": 0.19, "learning_rate": 1.8690293432868534e-05, "loss": 0.6555, "step": 1561 }, { "epoch": 0.19, "learning_rate": 1.8688343992373602e-05, "loss": 0.6787, "step": 1562 }, { "epoch": 0.19, "learning_rate": 1.868639320395141e-05, "loss": 0.6322, "step": 1563 }, { "epoch": 0.19, "learning_rate": 1.8684441067904604e-05, "loss": 0.6991, "step": 1564 }, { "epoch": 0.19, "learning_rate": 1.868248758453605e-05, "loss": 0.672, "step": 1565 }, { "epoch": 0.19, "learning_rate": 1.8680532754148808e-05, "loss": 0.6943, "step": 1566 }, { "epoch": 0.19, "learning_rate": 1.8678576577046158e-05, "loss": 0.6571, "step": 1567 }, { "epoch": 0.19, "learning_rate": 1.8676619053531583e-05, "loss": 0.6358, "step": 1568 }, { "epoch": 0.19, "learning_rate": 1.867466018390878e-05, "loss": 0.7153, "step": 1569 }, { "epoch": 0.19, "learning_rate": 1.8672699968481648e-05, "loss": 0.6617, "step": 1570 }, { "epoch": 0.19, "learning_rate": 1.86707384075543e-05, "loss": 0.6727, "step": 1571 }, { "epoch": 0.19, "learning_rate": 1.8668775501431065e-05, "loss": 0.6422, "step": 1572 }, { "epoch": 0.19, "learning_rate": 1.866681125041646e-05, "loss": 0.6434, "step": 1573 }, { "epoch": 0.19, "learning_rate": 1.866484565481523e-05, "loss": 0.6634, "step": 1574 }, { "epoch": 0.19, "learning_rate": 1.8662878714932315e-05, "loss": 0.645, "step": 1575 }, { "epoch": 0.19, "learning_rate": 1.866091043107288e-05, "loss": 0.6548, "step": 1576 }, { "epoch": 0.19, "learning_rate": 1.865894080354228e-05, "loss": 0.6416, "step": 1577 }, { "epoch": 0.19, "learning_rate": 1.865696983264609e-05, "loss": 0.6089, "step": 1578 }, { "epoch": 0.19, "learning_rate": 1.8654997518690087e-05, "loss": 0.6271, "step": 1579 }, { "epoch": 0.19, "learning_rate": 1.865302386198027e-05, "loss": 0.5922, "step": 1580 }, { "epoch": 0.19, "learning_rate": 1.8651048862822823e-05, "loss": 0.6739, "step": 1581 }, { "epoch": 0.19, "learning_rate": 1.864907252152416e-05, "loss": 0.5993, "step": 1582 }, { "epoch": 0.19, "learning_rate": 1.8647094838390894e-05, "loss": 0.6211, "step": 1583 }, { "epoch": 0.19, "learning_rate": 1.864511581372984e-05, "loss": 0.6828, "step": 1584 }, { "epoch": 0.19, "learning_rate": 1.8643135447848037e-05, "loss": 0.5934, "step": 1585 }, { "epoch": 0.19, "learning_rate": 1.864115374105272e-05, "loss": 0.6323, "step": 1586 }, { "epoch": 0.19, "learning_rate": 1.8639170693651327e-05, "loss": 0.6081, "step": 1587 }, { "epoch": 0.19, "learning_rate": 1.8637186305951526e-05, "loss": 0.6314, "step": 1588 }, { "epoch": 0.19, "learning_rate": 1.8635200578261167e-05, "loss": 0.6252, "step": 1589 }, { "epoch": 0.19, "learning_rate": 1.8633213510888327e-05, "loss": 0.6211, "step": 1590 }, { "epoch": 0.19, "learning_rate": 1.863122510414128e-05, "loss": 0.6601, "step": 1591 }, { "epoch": 0.19, "learning_rate": 1.8629235358328513e-05, "loss": 0.668, "step": 1592 }, { "epoch": 0.19, "learning_rate": 1.8627244273758718e-05, "loss": 0.5636, "step": 1593 }, { "epoch": 0.19, "learning_rate": 1.86252518507408e-05, "loss": 0.6606, "step": 1594 }, { "epoch": 0.19, "learning_rate": 1.862325808958386e-05, "loss": 0.655, "step": 1595 }, { "epoch": 0.19, "learning_rate": 1.862126299059722e-05, "loss": 0.6316, "step": 1596 }, { "epoch": 0.19, "learning_rate": 1.86192665540904e-05, "loss": 0.6127, "step": 1597 }, { "epoch": 0.19, "learning_rate": 1.8617268780373136e-05, "loss": 0.6277, "step": 1598 }, { "epoch": 0.19, "learning_rate": 1.861526966975536e-05, "loss": 0.6793, "step": 1599 }, { "epoch": 0.19, "learning_rate": 1.8613269222547226e-05, "loss": 0.6187, "step": 1600 }, { "epoch": 0.19, "learning_rate": 1.8611267439059083e-05, "loss": 0.6713, "step": 1601 }, { "epoch": 0.19, "learning_rate": 1.8609264319601492e-05, "loss": 0.644, "step": 1602 }, { "epoch": 0.19, "learning_rate": 1.860725986448522e-05, "loss": 0.648, "step": 1603 }, { "epoch": 0.2, "learning_rate": 1.860525407402124e-05, "loss": 0.6371, "step": 1604 }, { "epoch": 0.2, "learning_rate": 1.8603246948520746e-05, "loss": 0.6777, "step": 1605 }, { "epoch": 0.2, "learning_rate": 1.8601238488295112e-05, "loss": 0.6212, "step": 1606 }, { "epoch": 0.2, "learning_rate": 1.8599228693655944e-05, "loss": 0.6078, "step": 1607 }, { "epoch": 0.2, "learning_rate": 1.8597217564915043e-05, "loss": 0.65, "step": 1608 }, { "epoch": 0.2, "learning_rate": 1.859520510238442e-05, "loss": 0.6087, "step": 1609 }, { "epoch": 0.2, "learning_rate": 1.8593191306376295e-05, "loss": 0.6568, "step": 1610 }, { "epoch": 0.2, "learning_rate": 1.8591176177203087e-05, "loss": 0.6719, "step": 1611 }, { "epoch": 0.2, "learning_rate": 1.8589159715177435e-05, "loss": 0.6421, "step": 1612 }, { "epoch": 0.2, "learning_rate": 1.8587141920612168e-05, "loss": 0.5687, "step": 1613 }, { "epoch": 0.2, "learning_rate": 1.8585122793820335e-05, "loss": 0.6674, "step": 1614 }, { "epoch": 0.2, "learning_rate": 1.8583102335115186e-05, "loss": 0.6042, "step": 1615 }, { "epoch": 0.2, "learning_rate": 1.858108054481018e-05, "loss": 0.6837, "step": 1616 }, { "epoch": 0.2, "learning_rate": 1.8579057423218984e-05, "loss": 0.6902, "step": 1617 }, { "epoch": 0.2, "learning_rate": 1.8577032970655467e-05, "loss": 0.5961, "step": 1618 }, { "epoch": 0.2, "learning_rate": 1.8575007187433707e-05, "loss": 0.6401, "step": 1619 }, { "epoch": 0.2, "learning_rate": 1.8572980073867987e-05, "loss": 0.6321, "step": 1620 }, { "epoch": 0.2, "learning_rate": 1.8570951630272796e-05, "loss": 0.6122, "step": 1621 }, { "epoch": 0.2, "learning_rate": 1.856892185696284e-05, "loss": 0.6685, "step": 1622 }, { "epoch": 0.2, "learning_rate": 1.856689075425301e-05, "loss": 0.6485, "step": 1623 }, { "epoch": 0.2, "learning_rate": 1.856485832245842e-05, "loss": 0.6536, "step": 1624 }, { "epoch": 0.2, "learning_rate": 1.856282456189439e-05, "loss": 0.6004, "step": 1625 }, { "epoch": 0.2, "learning_rate": 1.8560789472876434e-05, "loss": 0.6523, "step": 1626 }, { "epoch": 0.2, "learning_rate": 1.8558753055720288e-05, "loss": 0.6521, "step": 1627 }, { "epoch": 0.2, "learning_rate": 1.855671531074188e-05, "loss": 0.664, "step": 1628 }, { "epoch": 0.2, "learning_rate": 1.855467623825735e-05, "loss": 0.6772, "step": 1629 }, { "epoch": 0.2, "learning_rate": 1.855263583858305e-05, "loss": 0.6241, "step": 1630 }, { "epoch": 0.2, "learning_rate": 1.8550594112035522e-05, "loss": 0.5974, "step": 1631 }, { "epoch": 0.2, "learning_rate": 1.854855105893153e-05, "loss": 0.7122, "step": 1632 }, { "epoch": 0.2, "learning_rate": 1.8546506679588038e-05, "loss": 0.6648, "step": 1633 }, { "epoch": 0.2, "learning_rate": 1.8544460974322208e-05, "loss": 0.6289, "step": 1634 }, { "epoch": 0.2, "learning_rate": 1.854241394345143e-05, "loss": 0.609, "step": 1635 }, { "epoch": 0.2, "learning_rate": 1.8540365587293266e-05, "loss": 0.645, "step": 1636 }, { "epoch": 0.2, "learning_rate": 1.8538315906165513e-05, "loss": 0.6222, "step": 1637 }, { "epoch": 0.2, "learning_rate": 1.8536264900386162e-05, "loss": 0.5946, "step": 1638 }, { "epoch": 0.2, "learning_rate": 1.8534212570273408e-05, "loss": 0.6434, "step": 1639 }, { "epoch": 0.2, "learning_rate": 1.8532158916145656e-05, "loss": 0.6542, "step": 1640 }, { "epoch": 0.2, "learning_rate": 1.853010393832151e-05, "loss": 0.6344, "step": 1641 }, { "epoch": 0.2, "learning_rate": 1.8528047637119787e-05, "loss": 0.6393, "step": 1642 }, { "epoch": 0.2, "learning_rate": 1.8525990012859503e-05, "loss": 0.6216, "step": 1643 }, { "epoch": 0.2, "learning_rate": 1.8523931065859883e-05, "loss": 0.6383, "step": 1644 }, { "epoch": 0.2, "learning_rate": 1.8521870796440358e-05, "loss": 0.6653, "step": 1645 }, { "epoch": 0.2, "learning_rate": 1.8519809204920563e-05, "loss": 0.7117, "step": 1646 }, { "epoch": 0.2, "learning_rate": 1.851774629162033e-05, "loss": 0.6246, "step": 1647 }, { "epoch": 0.2, "learning_rate": 1.8515682056859713e-05, "loss": 0.6074, "step": 1648 }, { "epoch": 0.2, "learning_rate": 1.8513616500958957e-05, "loss": 0.6364, "step": 1649 }, { "epoch": 0.2, "learning_rate": 1.851154962423852e-05, "loss": 0.6208, "step": 1650 }, { "epoch": 0.2, "learning_rate": 1.8509481427019054e-05, "loss": 0.6414, "step": 1651 }, { "epoch": 0.2, "learning_rate": 1.850741190962143e-05, "loss": 0.5608, "step": 1652 }, { "epoch": 0.2, "learning_rate": 1.8505341072366717e-05, "loss": 0.6869, "step": 1653 }, { "epoch": 0.2, "learning_rate": 1.8503268915576184e-05, "loss": 0.563, "step": 1654 }, { "epoch": 0.2, "learning_rate": 1.8501195439571314e-05, "loss": 0.6341, "step": 1655 }, { "epoch": 0.2, "learning_rate": 1.8499120644673786e-05, "loss": 0.6339, "step": 1656 }, { "epoch": 0.2, "learning_rate": 1.849704453120549e-05, "loss": 0.6407, "step": 1657 }, { "epoch": 0.2, "learning_rate": 1.8494967099488524e-05, "loss": 0.6367, "step": 1658 }, { "epoch": 0.2, "learning_rate": 1.8492888349845178e-05, "loss": 0.6314, "step": 1659 }, { "epoch": 0.2, "learning_rate": 1.8490808282597955e-05, "loss": 0.6102, "step": 1660 }, { "epoch": 0.2, "learning_rate": 1.848872689806956e-05, "loss": 0.5944, "step": 1661 }, { "epoch": 0.2, "learning_rate": 1.8486644196582907e-05, "loss": 0.6268, "step": 1662 }, { "epoch": 0.2, "learning_rate": 1.8484560178461107e-05, "loss": 0.642, "step": 1663 }, { "epoch": 0.2, "learning_rate": 1.8482474844027478e-05, "loss": 0.6677, "step": 1664 }, { "epoch": 0.2, "learning_rate": 1.8480388193605546e-05, "loss": 0.6083, "step": 1665 }, { "epoch": 0.2, "learning_rate": 1.847830022751904e-05, "loss": 0.7219, "step": 1666 }, { "epoch": 0.2, "learning_rate": 1.8476210946091884e-05, "loss": 0.6501, "step": 1667 }, { "epoch": 0.2, "learning_rate": 1.8474120349648216e-05, "loss": 0.5736, "step": 1668 }, { "epoch": 0.2, "learning_rate": 1.847202843851238e-05, "loss": 0.6696, "step": 1669 }, { "epoch": 0.2, "learning_rate": 1.8469935213008918e-05, "loss": 0.6571, "step": 1670 }, { "epoch": 0.2, "learning_rate": 1.846784067346257e-05, "loss": 0.6562, "step": 1671 }, { "epoch": 0.2, "learning_rate": 1.84657448201983e-05, "loss": 0.5958, "step": 1672 }, { "epoch": 0.2, "learning_rate": 1.846364765354125e-05, "loss": 0.5829, "step": 1673 }, { "epoch": 0.2, "learning_rate": 1.8461549173816785e-05, "loss": 0.6045, "step": 1674 }, { "epoch": 0.2, "learning_rate": 1.845944938135047e-05, "loss": 0.6048, "step": 1675 }, { "epoch": 0.2, "learning_rate": 1.8457348276468066e-05, "loss": 0.616, "step": 1676 }, { "epoch": 0.2, "learning_rate": 1.8455245859495542e-05, "loss": 0.6221, "step": 1677 }, { "epoch": 0.2, "learning_rate": 1.8453142130759074e-05, "loss": 0.6509, "step": 1678 }, { "epoch": 0.2, "learning_rate": 1.845103709058504e-05, "loss": 0.572, "step": 1679 }, { "epoch": 0.2, "learning_rate": 1.8448930739300017e-05, "loss": 0.6693, "step": 1680 }, { "epoch": 0.2, "learning_rate": 1.8446823077230794e-05, "loss": 0.6239, "step": 1681 }, { "epoch": 0.2, "learning_rate": 1.844471410470435e-05, "loss": 0.5908, "step": 1682 }, { "epoch": 0.2, "learning_rate": 1.8442603822047878e-05, "loss": 0.6101, "step": 1683 }, { "epoch": 0.2, "learning_rate": 1.8440492229588777e-05, "loss": 0.6642, "step": 1684 }, { "epoch": 0.2, "learning_rate": 1.8438379327654633e-05, "loss": 0.6066, "step": 1685 }, { "epoch": 0.21, "learning_rate": 1.843626511657326e-05, "loss": 0.657, "step": 1686 }, { "epoch": 0.21, "learning_rate": 1.843414959667265e-05, "loss": 0.5936, "step": 1687 }, { "epoch": 0.21, "learning_rate": 1.843203276828101e-05, "loss": 0.6328, "step": 1688 }, { "epoch": 0.21, "learning_rate": 1.8429914631726757e-05, "loss": 0.6486, "step": 1689 }, { "epoch": 0.21, "learning_rate": 1.842779518733849e-05, "loss": 0.6425, "step": 1690 }, { "epoch": 0.21, "learning_rate": 1.8425674435445036e-05, "loss": 0.6601, "step": 1691 }, { "epoch": 0.21, "learning_rate": 1.842355237637541e-05, "loss": 0.6129, "step": 1692 }, { "epoch": 0.21, "learning_rate": 1.842142901045883e-05, "loss": 0.6502, "step": 1693 }, { "epoch": 0.21, "learning_rate": 1.8419304338024716e-05, "loss": 0.6893, "step": 1694 }, { "epoch": 0.21, "learning_rate": 1.84171783594027e-05, "loss": 0.675, "step": 1695 }, { "epoch": 0.21, "learning_rate": 1.8415051074922607e-05, "loss": 0.5859, "step": 1696 }, { "epoch": 0.21, "learning_rate": 1.841292248491447e-05, "loss": 0.5784, "step": 1697 }, { "epoch": 0.21, "learning_rate": 1.8410792589708527e-05, "loss": 0.5982, "step": 1698 }, { "epoch": 0.21, "learning_rate": 1.8408661389635205e-05, "loss": 0.5895, "step": 1699 }, { "epoch": 0.21, "learning_rate": 1.8406528885025147e-05, "loss": 0.6444, "step": 1700 }, { "epoch": 0.21, "learning_rate": 1.8404395076209197e-05, "loss": 0.6163, "step": 1701 }, { "epoch": 0.21, "learning_rate": 1.8402259963518395e-05, "loss": 0.6282, "step": 1702 }, { "epoch": 0.21, "learning_rate": 1.840012354728399e-05, "loss": 0.6234, "step": 1703 }, { "epoch": 0.21, "learning_rate": 1.8397985827837424e-05, "loss": 0.6326, "step": 1704 }, { "epoch": 0.21, "learning_rate": 1.8395846805510352e-05, "loss": 0.6596, "step": 1705 }, { "epoch": 0.21, "learning_rate": 1.8393706480634622e-05, "loss": 0.6466, "step": 1706 }, { "epoch": 0.21, "learning_rate": 1.8391564853542297e-05, "loss": 0.6064, "step": 1707 }, { "epoch": 0.21, "learning_rate": 1.838942192456562e-05, "loss": 0.6191, "step": 1708 }, { "epoch": 0.21, "learning_rate": 1.8387277694037065e-05, "loss": 0.6478, "step": 1709 }, { "epoch": 0.21, "learning_rate": 1.8385132162289282e-05, "loss": 0.6556, "step": 1710 }, { "epoch": 0.21, "learning_rate": 1.8382985329655132e-05, "loss": 0.6999, "step": 1711 }, { "epoch": 0.21, "learning_rate": 1.8380837196467686e-05, "loss": 0.6759, "step": 1712 }, { "epoch": 0.21, "learning_rate": 1.8378687763060203e-05, "loss": 0.5889, "step": 1713 }, { "epoch": 0.21, "learning_rate": 1.8376537029766155e-05, "loss": 0.6894, "step": 1714 }, { "epoch": 0.21, "learning_rate": 1.837438499691921e-05, "loss": 0.6668, "step": 1715 }, { "epoch": 0.21, "learning_rate": 1.8372231664853238e-05, "loss": 0.6199, "step": 1716 }, { "epoch": 0.21, "learning_rate": 1.837007703390231e-05, "loss": 0.6626, "step": 1717 }, { "epoch": 0.21, "learning_rate": 1.8367921104400702e-05, "loss": 0.6192, "step": 1718 }, { "epoch": 0.21, "learning_rate": 1.836576387668289e-05, "loss": 0.6443, "step": 1719 }, { "epoch": 0.21, "learning_rate": 1.8363605351083547e-05, "loss": 0.6283, "step": 1720 }, { "epoch": 0.21, "learning_rate": 1.8361445527937553e-05, "loss": 0.6997, "step": 1721 }, { "epoch": 0.21, "learning_rate": 1.835928440757999e-05, "loss": 0.6307, "step": 1722 }, { "epoch": 0.21, "learning_rate": 1.8357121990346136e-05, "loss": 0.6108, "step": 1723 }, { "epoch": 0.21, "learning_rate": 1.835495827657147e-05, "loss": 0.6793, "step": 1724 }, { "epoch": 0.21, "learning_rate": 1.8352793266591683e-05, "loss": 0.6296, "step": 1725 }, { "epoch": 0.21, "learning_rate": 1.8350626960742653e-05, "loss": 0.6619, "step": 1726 }, { "epoch": 0.21, "learning_rate": 1.8348459359360467e-05, "loss": 0.591, "step": 1727 }, { "epoch": 0.21, "learning_rate": 1.8346290462781408e-05, "loss": 0.6058, "step": 1728 }, { "epoch": 0.21, "learning_rate": 1.8344120271341964e-05, "loss": 0.6265, "step": 1729 }, { "epoch": 0.21, "learning_rate": 1.834194878537882e-05, "loss": 0.6179, "step": 1730 }, { "epoch": 0.21, "learning_rate": 1.8339776005228878e-05, "loss": 0.6274, "step": 1731 }, { "epoch": 0.21, "learning_rate": 1.8337601931229213e-05, "loss": 0.6169, "step": 1732 }, { "epoch": 0.21, "learning_rate": 1.833542656371712e-05, "loss": 0.6247, "step": 1733 }, { "epoch": 0.21, "learning_rate": 1.8333249903030095e-05, "loss": 0.5955, "step": 1734 }, { "epoch": 0.21, "learning_rate": 1.833107194950582e-05, "loss": 0.5814, "step": 1735 }, { "epoch": 0.21, "learning_rate": 1.8328892703482196e-05, "loss": 0.6054, "step": 1736 }, { "epoch": 0.21, "learning_rate": 1.8326712165297312e-05, "loss": 0.6119, "step": 1737 }, { "epoch": 0.21, "learning_rate": 1.8324530335289457e-05, "loss": 0.6298, "step": 1738 }, { "epoch": 0.21, "learning_rate": 1.832234721379713e-05, "loss": 0.6396, "step": 1739 }, { "epoch": 0.21, "learning_rate": 1.832016280115902e-05, "loss": 0.6138, "step": 1740 }, { "epoch": 0.21, "learning_rate": 1.831797709771403e-05, "loss": 0.6161, "step": 1741 }, { "epoch": 0.21, "learning_rate": 1.8315790103801246e-05, "loss": 0.5993, "step": 1742 }, { "epoch": 0.21, "learning_rate": 1.8313601819759967e-05, "loss": 0.6847, "step": 1743 }, { "epoch": 0.21, "learning_rate": 1.8311412245929686e-05, "loss": 0.6176, "step": 1744 }, { "epoch": 0.21, "learning_rate": 1.8309221382650097e-05, "loss": 0.5909, "step": 1745 }, { "epoch": 0.21, "learning_rate": 1.8307029230261094e-05, "loss": 0.6396, "step": 1746 }, { "epoch": 0.21, "learning_rate": 1.830483578910278e-05, "loss": 0.7081, "step": 1747 }, { "epoch": 0.21, "learning_rate": 1.8302641059515437e-05, "loss": 0.6682, "step": 1748 }, { "epoch": 0.21, "learning_rate": 1.8300445041839574e-05, "loss": 0.5924, "step": 1749 }, { "epoch": 0.21, "learning_rate": 1.8298247736415874e-05, "loss": 0.5736, "step": 1750 }, { "epoch": 0.21, "learning_rate": 1.8296049143585237e-05, "loss": 0.669, "step": 1751 }, { "epoch": 0.21, "learning_rate": 1.8293849263688758e-05, "loss": 0.6063, "step": 1752 }, { "epoch": 0.21, "learning_rate": 1.8291648097067723e-05, "loss": 0.622, "step": 1753 }, { "epoch": 0.21, "learning_rate": 1.8289445644063637e-05, "loss": 0.6199, "step": 1754 }, { "epoch": 0.21, "learning_rate": 1.8287241905018186e-05, "loss": 0.6387, "step": 1755 }, { "epoch": 0.21, "learning_rate": 1.8285036880273265e-05, "loss": 0.6059, "step": 1756 }, { "epoch": 0.21, "learning_rate": 1.828283057017096e-05, "loss": 0.6206, "step": 1757 }, { "epoch": 0.21, "learning_rate": 1.828062297505357e-05, "loss": 0.6338, "step": 1758 }, { "epoch": 0.21, "learning_rate": 1.8278414095263586e-05, "loss": 0.6319, "step": 1759 }, { "epoch": 0.21, "learning_rate": 1.8276203931143693e-05, "loss": 0.5829, "step": 1760 }, { "epoch": 0.21, "learning_rate": 1.827399248303678e-05, "loss": 0.6553, "step": 1761 }, { "epoch": 0.21, "learning_rate": 1.827177975128594e-05, "loss": 0.6215, "step": 1762 }, { "epoch": 0.21, "learning_rate": 1.8269565736234458e-05, "loss": 0.6661, "step": 1763 }, { "epoch": 0.21, "learning_rate": 1.8267350438225825e-05, "loss": 0.6012, "step": 1764 }, { "epoch": 0.21, "learning_rate": 1.8265133857603718e-05, "loss": 0.6346, "step": 1765 }, { "epoch": 0.21, "learning_rate": 1.826291599471203e-05, "loss": 0.5948, "step": 1766 }, { "epoch": 0.21, "learning_rate": 1.826069684989484e-05, "loss": 0.6478, "step": 1767 }, { "epoch": 0.21, "learning_rate": 1.8258476423496435e-05, "loss": 0.653, "step": 1768 }, { "epoch": 0.22, "learning_rate": 1.825625471586129e-05, "loss": 0.6182, "step": 1769 }, { "epoch": 0.22, "learning_rate": 1.8254031727334093e-05, "loss": 0.6039, "step": 1770 }, { "epoch": 0.22, "learning_rate": 1.8251807458259715e-05, "loss": 0.6004, "step": 1771 }, { "epoch": 0.22, "learning_rate": 1.824958190898324e-05, "loss": 0.6721, "step": 1772 }, { "epoch": 0.22, "learning_rate": 1.8247355079849943e-05, "loss": 0.6567, "step": 1773 }, { "epoch": 0.22, "learning_rate": 1.8245126971205294e-05, "loss": 0.6775, "step": 1774 }, { "epoch": 0.22, "learning_rate": 1.8242897583394975e-05, "loss": 0.5431, "step": 1775 }, { "epoch": 0.22, "learning_rate": 1.8240666916764847e-05, "loss": 0.5695, "step": 1776 }, { "epoch": 0.22, "learning_rate": 1.823843497166099e-05, "loss": 0.6498, "step": 1777 }, { "epoch": 0.22, "learning_rate": 1.8236201748429665e-05, "loss": 0.5796, "step": 1778 }, { "epoch": 0.22, "learning_rate": 1.823396724741734e-05, "loss": 0.6786, "step": 1779 }, { "epoch": 0.22, "learning_rate": 1.823173146897069e-05, "loss": 0.6512, "step": 1780 }, { "epoch": 0.22, "learning_rate": 1.8229494413436564e-05, "loss": 0.6675, "step": 1781 }, { "epoch": 0.22, "learning_rate": 1.822725608116203e-05, "loss": 0.577, "step": 1782 }, { "epoch": 0.22, "learning_rate": 1.8225016472494348e-05, "loss": 0.6421, "step": 1783 }, { "epoch": 0.22, "learning_rate": 1.8222775587780973e-05, "loss": 0.5963, "step": 1784 }, { "epoch": 0.22, "learning_rate": 1.8220533427369562e-05, "loss": 0.628, "step": 1785 }, { "epoch": 0.22, "learning_rate": 1.8218289991607968e-05, "loss": 0.6149, "step": 1786 }, { "epoch": 0.22, "learning_rate": 1.8216045280844246e-05, "loss": 0.6338, "step": 1787 }, { "epoch": 0.22, "learning_rate": 1.8213799295426636e-05, "loss": 0.6487, "step": 1788 }, { "epoch": 0.22, "learning_rate": 1.821155203570359e-05, "loss": 0.6312, "step": 1789 }, { "epoch": 0.22, "learning_rate": 1.8209303502023753e-05, "loss": 0.663, "step": 1790 }, { "epoch": 0.22, "learning_rate": 1.820705369473597e-05, "loss": 0.6274, "step": 1791 }, { "epoch": 0.22, "learning_rate": 1.8204802614189273e-05, "loss": 0.7114, "step": 1792 }, { "epoch": 0.22, "learning_rate": 1.8202550260732907e-05, "loss": 0.5734, "step": 1793 }, { "epoch": 0.22, "learning_rate": 1.8200296634716297e-05, "loss": 0.6253, "step": 1794 }, { "epoch": 0.22, "learning_rate": 1.8198041736489086e-05, "loss": 0.6695, "step": 1795 }, { "epoch": 0.22, "learning_rate": 1.8195785566401098e-05, "loss": 0.6278, "step": 1796 }, { "epoch": 0.22, "learning_rate": 1.8193528124802354e-05, "loss": 0.5948, "step": 1797 }, { "epoch": 0.22, "learning_rate": 1.8191269412043092e-05, "loss": 0.6333, "step": 1798 }, { "epoch": 0.22, "learning_rate": 1.818900942847372e-05, "loss": 0.6263, "step": 1799 }, { "epoch": 0.22, "learning_rate": 1.8186748174444865e-05, "loss": 0.5508, "step": 1800 }, { "epoch": 0.22, "learning_rate": 1.818448565030734e-05, "loss": 0.5634, "step": 1801 }, { "epoch": 0.22, "learning_rate": 1.8182221856412154e-05, "loss": 0.6436, "step": 1802 }, { "epoch": 0.22, "learning_rate": 1.8179956793110523e-05, "loss": 0.638, "step": 1803 }, { "epoch": 0.22, "learning_rate": 1.8177690460753848e-05, "loss": 0.662, "step": 1804 }, { "epoch": 0.22, "learning_rate": 1.8175422859693734e-05, "loss": 0.6523, "step": 1805 }, { "epoch": 0.22, "learning_rate": 1.8173153990281986e-05, "loss": 0.6313, "step": 1806 }, { "epoch": 0.22, "learning_rate": 1.8170883852870593e-05, "loss": 0.6013, "step": 1807 }, { "epoch": 0.22, "learning_rate": 1.816861244781175e-05, "loss": 0.5905, "step": 1808 }, { "epoch": 0.22, "learning_rate": 1.8166339775457853e-05, "loss": 0.6054, "step": 1809 }, { "epoch": 0.22, "learning_rate": 1.8164065836161485e-05, "loss": 0.5649, "step": 1810 }, { "epoch": 0.22, "learning_rate": 1.816179063027543e-05, "loss": 0.5791, "step": 1811 }, { "epoch": 0.22, "learning_rate": 1.815951415815267e-05, "loss": 0.603, "step": 1812 }, { "epoch": 0.22, "learning_rate": 1.8157236420146378e-05, "loss": 0.619, "step": 1813 }, { "epoch": 0.22, "learning_rate": 1.815495741660993e-05, "loss": 0.6584, "step": 1814 }, { "epoch": 0.22, "learning_rate": 1.8152677147896893e-05, "loss": 0.6067, "step": 1815 }, { "epoch": 0.22, "learning_rate": 1.815039561436103e-05, "loss": 0.641, "step": 1816 }, { "epoch": 0.22, "learning_rate": 1.814811281635631e-05, "loss": 0.6355, "step": 1817 }, { "epoch": 0.22, "learning_rate": 1.8145828754236888e-05, "loss": 0.6214, "step": 1818 }, { "epoch": 0.22, "learning_rate": 1.814354342835712e-05, "loss": 0.5644, "step": 1819 }, { "epoch": 0.22, "learning_rate": 1.8141256839071547e-05, "loss": 0.6199, "step": 1820 }, { "epoch": 0.22, "learning_rate": 1.8138968986734924e-05, "loss": 0.6285, "step": 1821 }, { "epoch": 0.22, "learning_rate": 1.8136679871702187e-05, "loss": 0.6137, "step": 1822 }, { "epoch": 0.22, "learning_rate": 1.8134389494328483e-05, "loss": 0.6915, "step": 1823 }, { "epoch": 0.22, "learning_rate": 1.8132097854969135e-05, "loss": 0.6838, "step": 1824 }, { "epoch": 0.22, "learning_rate": 1.8129804953979678e-05, "loss": 0.6408, "step": 1825 }, { "epoch": 0.22, "learning_rate": 1.8127510791715837e-05, "loss": 0.6306, "step": 1826 }, { "epoch": 0.22, "learning_rate": 1.8125215368533534e-05, "loss": 0.6692, "step": 1827 }, { "epoch": 0.22, "learning_rate": 1.8122918684788887e-05, "loss": 0.5967, "step": 1828 }, { "epoch": 0.22, "learning_rate": 1.8120620740838203e-05, "loss": 0.6636, "step": 1829 }, { "epoch": 0.22, "learning_rate": 1.811832153703799e-05, "loss": 0.632, "step": 1830 }, { "epoch": 0.22, "learning_rate": 1.8116021073744957e-05, "loss": 0.6085, "step": 1831 }, { "epoch": 0.22, "learning_rate": 1.8113719351315996e-05, "loss": 0.5362, "step": 1832 }, { "epoch": 0.22, "learning_rate": 1.8111416370108205e-05, "loss": 0.6576, "step": 1833 }, { "epoch": 0.22, "learning_rate": 1.8109112130478875e-05, "loss": 0.6955, "step": 1834 }, { "epoch": 0.22, "learning_rate": 1.8106806632785484e-05, "loss": 0.6662, "step": 1835 }, { "epoch": 0.22, "learning_rate": 1.8104499877385712e-05, "loss": 0.6242, "step": 1836 }, { "epoch": 0.22, "learning_rate": 1.8102191864637442e-05, "loss": 0.5852, "step": 1837 }, { "epoch": 0.22, "learning_rate": 1.8099882594898734e-05, "loss": 0.6104, "step": 1838 }, { "epoch": 0.22, "learning_rate": 1.8097572068527863e-05, "loss": 0.6192, "step": 1839 }, { "epoch": 0.22, "learning_rate": 1.809526028588328e-05, "loss": 0.6007, "step": 1840 }, { "epoch": 0.22, "learning_rate": 1.809294724732364e-05, "loss": 0.6006, "step": 1841 }, { "epoch": 0.22, "learning_rate": 1.8090632953207795e-05, "loss": 0.6036, "step": 1842 }, { "epoch": 0.22, "learning_rate": 1.8088317403894794e-05, "loss": 0.6279, "step": 1843 }, { "epoch": 0.22, "learning_rate": 1.8086000599743866e-05, "loss": 0.6835, "step": 1844 }, { "epoch": 0.22, "learning_rate": 1.8083682541114454e-05, "loss": 0.5763, "step": 1845 }, { "epoch": 0.22, "learning_rate": 1.8081363228366183e-05, "loss": 0.6379, "step": 1846 }, { "epoch": 0.22, "learning_rate": 1.807904266185887e-05, "loss": 0.652, "step": 1847 }, { "epoch": 0.22, "learning_rate": 1.8076720841952543e-05, "loss": 0.5878, "step": 1848 }, { "epoch": 0.22, "learning_rate": 1.8074397769007402e-05, "loss": 0.6433, "step": 1849 }, { "epoch": 0.22, "learning_rate": 1.8072073443383864e-05, "loss": 0.6385, "step": 1850 }, { "epoch": 0.23, "learning_rate": 1.8069747865442523e-05, "loss": 0.6151, "step": 1851 }, { "epoch": 0.23, "learning_rate": 1.806742103554418e-05, "loss": 0.6559, "step": 1852 }, { "epoch": 0.23, "learning_rate": 1.8065092954049816e-05, "loss": 0.6318, "step": 1853 }, { "epoch": 0.23, "learning_rate": 1.8062763621320623e-05, "loss": 0.643, "step": 1854 }, { "epoch": 0.23, "learning_rate": 1.806043303771797e-05, "loss": 0.6761, "step": 1855 }, { "epoch": 0.23, "learning_rate": 1.8058101203603432e-05, "loss": 0.6154, "step": 1856 }, { "epoch": 0.23, "learning_rate": 1.8055768119338777e-05, "loss": 0.6625, "step": 1857 }, { "epoch": 0.23, "learning_rate": 1.805343378528596e-05, "loss": 0.6087, "step": 1858 }, { "epoch": 0.23, "learning_rate": 1.8051098201807136e-05, "loss": 0.6398, "step": 1859 }, { "epoch": 0.23, "learning_rate": 1.8048761369264656e-05, "loss": 0.6285, "step": 1860 }, { "epoch": 0.23, "learning_rate": 1.8046423288021055e-05, "loss": 0.5971, "step": 1861 }, { "epoch": 0.23, "learning_rate": 1.804408395843907e-05, "loss": 0.633, "step": 1862 }, { "epoch": 0.23, "learning_rate": 1.804174338088163e-05, "loss": 0.6243, "step": 1863 }, { "epoch": 0.23, "learning_rate": 1.8039401555711857e-05, "loss": 0.5992, "step": 1864 }, { "epoch": 0.23, "learning_rate": 1.8037058483293065e-05, "loss": 0.6379, "step": 1865 }, { "epoch": 0.23, "learning_rate": 1.8034714163988764e-05, "loss": 0.6618, "step": 1866 }, { "epoch": 0.23, "learning_rate": 1.8032368598162656e-05, "loss": 0.6458, "step": 1867 }, { "epoch": 0.23, "learning_rate": 1.803002178617864e-05, "loss": 0.6219, "step": 1868 }, { "epoch": 0.23, "learning_rate": 1.80276737284008e-05, "loss": 0.5638, "step": 1869 }, { "epoch": 0.23, "learning_rate": 1.802532442519342e-05, "loss": 0.6415, "step": 1870 }, { "epoch": 0.23, "learning_rate": 1.802297387692098e-05, "loss": 0.6212, "step": 1871 }, { "epoch": 0.23, "learning_rate": 1.802062208394814e-05, "loss": 0.6275, "step": 1872 }, { "epoch": 0.23, "learning_rate": 1.801826904663977e-05, "loss": 0.6602, "step": 1873 }, { "epoch": 0.23, "learning_rate": 1.8015914765360922e-05, "loss": 0.6088, "step": 1874 }, { "epoch": 0.23, "learning_rate": 1.8013559240476842e-05, "loss": 0.6404, "step": 1875 }, { "epoch": 0.23, "learning_rate": 1.8011202472352975e-05, "loss": 0.604, "step": 1876 }, { "epoch": 0.23, "learning_rate": 1.8008844461354953e-05, "loss": 0.6165, "step": 1877 }, { "epoch": 0.23, "learning_rate": 1.80064852078486e-05, "loss": 0.6003, "step": 1878 }, { "epoch": 0.23, "learning_rate": 1.800412471219994e-05, "loss": 0.5931, "step": 1879 }, { "epoch": 0.23, "learning_rate": 1.8001762974775182e-05, "loss": 0.613, "step": 1880 }, { "epoch": 0.23, "learning_rate": 1.7999399995940733e-05, "loss": 0.6091, "step": 1881 }, { "epoch": 0.23, "learning_rate": 1.7997035776063187e-05, "loss": 0.6568, "step": 1882 }, { "epoch": 0.23, "learning_rate": 1.7994670315509334e-05, "loss": 0.636, "step": 1883 }, { "epoch": 0.23, "learning_rate": 1.7992303614646164e-05, "loss": 0.613, "step": 1884 }, { "epoch": 0.23, "learning_rate": 1.798993567384084e-05, "loss": 0.5944, "step": 1885 }, { "epoch": 0.23, "learning_rate": 1.798756649346074e-05, "loss": 0.6294, "step": 1886 }, { "epoch": 0.23, "learning_rate": 1.7985196073873414e-05, "loss": 0.7021, "step": 1887 }, { "epoch": 0.23, "learning_rate": 1.7982824415446617e-05, "loss": 0.6503, "step": 1888 }, { "epoch": 0.23, "learning_rate": 1.7980451518548296e-05, "loss": 0.5769, "step": 1889 }, { "epoch": 0.23, "learning_rate": 1.7978077383546585e-05, "loss": 0.5871, "step": 1890 }, { "epoch": 0.23, "learning_rate": 1.797570201080981e-05, "loss": 0.6279, "step": 1891 }, { "epoch": 0.23, "learning_rate": 1.7973325400706498e-05, "loss": 0.5757, "step": 1892 }, { "epoch": 0.23, "learning_rate": 1.7970947553605352e-05, "loss": 0.6032, "step": 1893 }, { "epoch": 0.23, "learning_rate": 1.7968568469875283e-05, "loss": 0.6384, "step": 1894 }, { "epoch": 0.23, "learning_rate": 1.796618814988538e-05, "loss": 0.6, "step": 1895 }, { "epoch": 0.23, "learning_rate": 1.796380659400494e-05, "loss": 0.5872, "step": 1896 }, { "epoch": 0.23, "learning_rate": 1.7961423802603433e-05, "loss": 0.5982, "step": 1897 }, { "epoch": 0.23, "learning_rate": 1.7959039776050535e-05, "loss": 0.5914, "step": 1898 }, { "epoch": 0.23, "learning_rate": 1.795665451471611e-05, "loss": 0.6265, "step": 1899 }, { "epoch": 0.23, "learning_rate": 1.795426801897021e-05, "loss": 0.6456, "step": 1900 }, { "epoch": 0.23, "learning_rate": 1.7951880289183078e-05, "loss": 0.5584, "step": 1901 }, { "epoch": 0.23, "learning_rate": 1.7949491325725158e-05, "loss": 0.6281, "step": 1902 }, { "epoch": 0.23, "learning_rate": 1.7947101128967074e-05, "loss": 0.6798, "step": 1903 }, { "epoch": 0.23, "learning_rate": 1.7944709699279643e-05, "loss": 0.6094, "step": 1904 }, { "epoch": 0.23, "learning_rate": 1.7942317037033886e-05, "loss": 0.592, "step": 1905 }, { "epoch": 0.23, "learning_rate": 1.7939923142600996e-05, "loss": 0.6254, "step": 1906 }, { "epoch": 0.23, "learning_rate": 1.793752801635237e-05, "loss": 0.6236, "step": 1907 }, { "epoch": 0.23, "learning_rate": 1.793513165865959e-05, "loss": 0.6842, "step": 1908 }, { "epoch": 0.23, "learning_rate": 1.7932734069894443e-05, "loss": 0.6407, "step": 1909 }, { "epoch": 0.23, "learning_rate": 1.7930335250428878e-05, "loss": 0.6345, "step": 1910 }, { "epoch": 0.23, "learning_rate": 1.792793520063507e-05, "loss": 0.6254, "step": 1911 }, { "epoch": 0.23, "learning_rate": 1.7925533920885354e-05, "loss": 0.6299, "step": 1912 }, { "epoch": 0.23, "learning_rate": 1.7923131411552278e-05, "loss": 0.5655, "step": 1913 }, { "epoch": 0.23, "learning_rate": 1.7920727673008568e-05, "loss": 0.628, "step": 1914 }, { "epoch": 0.23, "learning_rate": 1.7918322705627147e-05, "loss": 0.6116, "step": 1915 }, { "epoch": 0.23, "learning_rate": 1.7915916509781124e-05, "loss": 0.5929, "step": 1916 }, { "epoch": 0.23, "learning_rate": 1.7913509085843803e-05, "loss": 0.652, "step": 1917 }, { "epoch": 0.23, "learning_rate": 1.791110043418868e-05, "loss": 0.5728, "step": 1918 }, { "epoch": 0.23, "learning_rate": 1.790869055518943e-05, "loss": 0.6361, "step": 1919 }, { "epoch": 0.23, "learning_rate": 1.7906279449219933e-05, "loss": 0.6124, "step": 1920 }, { "epoch": 0.23, "learning_rate": 1.790386711665425e-05, "loss": 0.5957, "step": 1921 }, { "epoch": 0.23, "learning_rate": 1.7901453557866635e-05, "loss": 0.6564, "step": 1922 }, { "epoch": 0.23, "learning_rate": 1.7899038773231536e-05, "loss": 0.5855, "step": 1923 }, { "epoch": 0.23, "learning_rate": 1.7896622763123584e-05, "loss": 0.5946, "step": 1924 }, { "epoch": 0.23, "learning_rate": 1.7894205527917603e-05, "loss": 0.6682, "step": 1925 }, { "epoch": 0.23, "learning_rate": 1.7891787067988612e-05, "loss": 0.5964, "step": 1926 }, { "epoch": 0.23, "learning_rate": 1.788936738371181e-05, "loss": 0.6444, "step": 1927 }, { "epoch": 0.23, "learning_rate": 1.7886946475462594e-05, "loss": 0.6407, "step": 1928 }, { "epoch": 0.23, "learning_rate": 1.788452434361655e-05, "loss": 0.6489, "step": 1929 }, { "epoch": 0.23, "learning_rate": 1.7882100988549455e-05, "loss": 0.584, "step": 1930 }, { "epoch": 0.23, "learning_rate": 1.787967641063727e-05, "loss": 0.6053, "step": 1931 }, { "epoch": 0.23, "learning_rate": 1.7877250610256148e-05, "loss": 0.6233, "step": 1932 }, { "epoch": 0.24, "learning_rate": 1.7874823587782434e-05, "loss": 0.5972, "step": 1933 }, { "epoch": 0.24, "learning_rate": 1.787239534359266e-05, "loss": 0.6135, "step": 1934 }, { "epoch": 0.24, "learning_rate": 1.7869965878063546e-05, "loss": 0.5666, "step": 1935 }, { "epoch": 0.24, "learning_rate": 1.7867535191572016e-05, "loss": 0.609, "step": 1936 }, { "epoch": 0.24, "learning_rate": 1.7865103284495156e-05, "loss": 0.6611, "step": 1937 }, { "epoch": 0.24, "learning_rate": 1.7862670157210266e-05, "loss": 0.5837, "step": 1938 }, { "epoch": 0.24, "learning_rate": 1.7860235810094824e-05, "loss": 0.6337, "step": 1939 }, { "epoch": 0.24, "learning_rate": 1.7857800243526503e-05, "loss": 0.5914, "step": 1940 }, { "epoch": 0.24, "learning_rate": 1.7855363457883156e-05, "loss": 0.5582, "step": 1941 }, { "epoch": 0.24, "learning_rate": 1.7852925453542834e-05, "loss": 0.5972, "step": 1942 }, { "epoch": 0.24, "learning_rate": 1.7850486230883773e-05, "loss": 0.5801, "step": 1943 }, { "epoch": 0.24, "learning_rate": 1.7848045790284402e-05, "loss": 0.6158, "step": 1944 }, { "epoch": 0.24, "learning_rate": 1.784560413212333e-05, "loss": 0.6316, "step": 1945 }, { "epoch": 0.24, "learning_rate": 1.7843161256779365e-05, "loss": 0.5915, "step": 1946 }, { "epoch": 0.24, "learning_rate": 1.7840717164631502e-05, "loss": 0.6083, "step": 1947 }, { "epoch": 0.24, "learning_rate": 1.783827185605892e-05, "loss": 0.5923, "step": 1948 }, { "epoch": 0.24, "learning_rate": 1.7835825331440985e-05, "loss": 0.6413, "step": 1949 }, { "epoch": 0.24, "learning_rate": 1.7833377591157255e-05, "loss": 0.6131, "step": 1950 }, { "epoch": 0.24, "learning_rate": 1.7830928635587488e-05, "loss": 0.6618, "step": 1951 }, { "epoch": 0.24, "learning_rate": 1.7828478465111612e-05, "loss": 0.5686, "step": 1952 }, { "epoch": 0.24, "learning_rate": 1.7826027080109753e-05, "loss": 0.6175, "step": 1953 }, { "epoch": 0.24, "learning_rate": 1.782357448096222e-05, "loss": 0.6067, "step": 1954 }, { "epoch": 0.24, "learning_rate": 1.782112066804952e-05, "loss": 0.6338, "step": 1955 }, { "epoch": 0.24, "learning_rate": 1.781866564175234e-05, "loss": 0.6652, "step": 1956 }, { "epoch": 0.24, "learning_rate": 1.7816209402451562e-05, "loss": 0.6623, "step": 1957 }, { "epoch": 0.24, "learning_rate": 1.7813751950528245e-05, "loss": 0.6374, "step": 1958 }, { "epoch": 0.24, "learning_rate": 1.7811293286363646e-05, "loss": 0.6522, "step": 1959 }, { "epoch": 0.24, "learning_rate": 1.7808833410339207e-05, "loss": 0.6529, "step": 1960 }, { "epoch": 0.24, "learning_rate": 1.7806372322836555e-05, "loss": 0.6555, "step": 1961 }, { "epoch": 0.24, "learning_rate": 1.7803910024237517e-05, "loss": 0.7237, "step": 1962 }, { "epoch": 0.24, "learning_rate": 1.7801446514924093e-05, "loss": 0.6086, "step": 1963 }, { "epoch": 0.24, "learning_rate": 1.7798981795278477e-05, "loss": 0.6043, "step": 1964 }, { "epoch": 0.24, "learning_rate": 1.7796515865683047e-05, "loss": 0.5938, "step": 1965 }, { "epoch": 0.24, "learning_rate": 1.779404872652038e-05, "loss": 0.6061, "step": 1966 }, { "epoch": 0.24, "learning_rate": 1.7791580378173224e-05, "loss": 0.5534, "step": 1967 }, { "epoch": 0.24, "learning_rate": 1.7789110821024532e-05, "loss": 0.5947, "step": 1968 }, { "epoch": 0.24, "learning_rate": 1.7786640055457435e-05, "loss": 0.6621, "step": 1969 }, { "epoch": 0.24, "learning_rate": 1.778416808185525e-05, "loss": 0.5656, "step": 1970 }, { "epoch": 0.24, "learning_rate": 1.7781694900601482e-05, "loss": 0.6351, "step": 1971 }, { "epoch": 0.24, "learning_rate": 1.7779220512079825e-05, "loss": 0.7063, "step": 1972 }, { "epoch": 0.24, "learning_rate": 1.777674491667417e-05, "loss": 0.6538, "step": 1973 }, { "epoch": 0.24, "learning_rate": 1.7774268114768576e-05, "loss": 0.6284, "step": 1974 }, { "epoch": 0.24, "learning_rate": 1.77717901067473e-05, "loss": 0.6318, "step": 1975 }, { "epoch": 0.24, "learning_rate": 1.7769310892994788e-05, "loss": 0.5936, "step": 1976 }, { "epoch": 0.24, "learning_rate": 1.7766830473895672e-05, "loss": 0.6661, "step": 1977 }, { "epoch": 0.24, "learning_rate": 1.7764348849834766e-05, "loss": 0.6878, "step": 1978 }, { "epoch": 0.24, "learning_rate": 1.7761866021197075e-05, "loss": 0.6424, "step": 1979 }, { "epoch": 0.24, "learning_rate": 1.775938198836779e-05, "loss": 0.6657, "step": 1980 }, { "epoch": 0.24, "learning_rate": 1.775689675173229e-05, "loss": 0.6224, "step": 1981 }, { "epoch": 0.24, "learning_rate": 1.7754410311676135e-05, "loss": 0.646, "step": 1982 }, { "epoch": 0.24, "learning_rate": 1.7751922668585083e-05, "loss": 0.6654, "step": 1983 }, { "epoch": 0.24, "learning_rate": 1.7749433822845068e-05, "loss": 0.6025, "step": 1984 }, { "epoch": 0.24, "learning_rate": 1.7746943774842215e-05, "loss": 0.6231, "step": 1985 }, { "epoch": 0.24, "learning_rate": 1.7744452524962836e-05, "loss": 0.6558, "step": 1986 }, { "epoch": 0.24, "learning_rate": 1.774196007359343e-05, "loss": 0.5498, "step": 1987 }, { "epoch": 0.24, "learning_rate": 1.7739466421120677e-05, "loss": 0.5673, "step": 1988 }, { "epoch": 0.24, "learning_rate": 1.773697156793145e-05, "loss": 0.6107, "step": 1989 }, { "epoch": 0.24, "learning_rate": 1.7734475514412805e-05, "loss": 0.6349, "step": 1990 }, { "epoch": 0.24, "learning_rate": 1.7731978260951985e-05, "loss": 0.5965, "step": 1991 }, { "epoch": 0.24, "learning_rate": 1.772947980793642e-05, "loss": 0.7211, "step": 1992 }, { "epoch": 0.24, "learning_rate": 1.772698015575372e-05, "loss": 0.6634, "step": 1993 }, { "epoch": 0.24, "learning_rate": 1.772447930479169e-05, "loss": 0.5744, "step": 1994 }, { "epoch": 0.24, "learning_rate": 1.7721977255438318e-05, "loss": 0.6277, "step": 1995 }, { "epoch": 0.24, "learning_rate": 1.7719474008081776e-05, "loss": 0.5716, "step": 1996 }, { "epoch": 0.24, "learning_rate": 1.7716969563110426e-05, "loss": 0.5779, "step": 1997 }, { "epoch": 0.24, "learning_rate": 1.7714463920912804e-05, "loss": 0.598, "step": 1998 }, { "epoch": 0.24, "learning_rate": 1.771195708187765e-05, "loss": 0.6106, "step": 1999 }, { "epoch": 0.24, "learning_rate": 1.7709449046393872e-05, "loss": 0.5725, "step": 2000 }, { "epoch": 0.24, "learning_rate": 1.770693981485058e-05, "loss": 0.5767, "step": 2001 }, { "epoch": 0.24, "learning_rate": 1.770442938763705e-05, "loss": 0.6269, "step": 2002 }, { "epoch": 0.24, "learning_rate": 1.7701917765142768e-05, "loss": 0.6273, "step": 2003 }, { "epoch": 0.24, "learning_rate": 1.7699404947757385e-05, "loss": 0.6224, "step": 2004 }, { "epoch": 0.24, "learning_rate": 1.7696890935870745e-05, "loss": 0.6051, "step": 2005 }, { "epoch": 0.24, "learning_rate": 1.7694375729872874e-05, "loss": 0.6248, "step": 2006 }, { "epoch": 0.24, "learning_rate": 1.7691859330153992e-05, "loss": 0.5927, "step": 2007 }, { "epoch": 0.24, "learning_rate": 1.7689341737104495e-05, "loss": 0.5954, "step": 2008 }, { "epoch": 0.24, "learning_rate": 1.7686822951114972e-05, "loss": 0.6171, "step": 2009 }, { "epoch": 0.24, "learning_rate": 1.7684302972576184e-05, "loss": 0.6396, "step": 2010 }, { "epoch": 0.24, "learning_rate": 1.7681781801879093e-05, "loss": 0.5578, "step": 2011 }, { "epoch": 0.24, "learning_rate": 1.7679259439414835e-05, "loss": 0.6209, "step": 2012 }, { "epoch": 0.24, "learning_rate": 1.7676735885574734e-05, "loss": 0.6432, "step": 2013 }, { "epoch": 0.24, "learning_rate": 1.76742111407503e-05, "loss": 0.6483, "step": 2014 }, { "epoch": 0.25, "learning_rate": 1.7671685205333232e-05, "loss": 0.556, "step": 2015 }, { "epoch": 0.25, "learning_rate": 1.76691580797154e-05, "loss": 0.6254, "step": 2016 }, { "epoch": 0.25, "learning_rate": 1.7666629764288874e-05, "loss": 0.5746, "step": 2017 }, { "epoch": 0.25, "learning_rate": 1.7664100259445896e-05, "loss": 0.6121, "step": 2018 }, { "epoch": 0.25, "learning_rate": 1.7661569565578904e-05, "loss": 0.6118, "step": 2019 }, { "epoch": 0.25, "learning_rate": 1.7659037683080514e-05, "loss": 0.6225, "step": 2020 }, { "epoch": 0.25, "learning_rate": 1.7656504612343527e-05, "loss": 0.6386, "step": 2021 }, { "epoch": 0.25, "learning_rate": 1.7653970353760927e-05, "loss": 0.6584, "step": 2022 }, { "epoch": 0.25, "learning_rate": 1.7651434907725885e-05, "loss": 0.6281, "step": 2023 }, { "epoch": 0.25, "learning_rate": 1.7648898274631755e-05, "loss": 0.5899, "step": 2024 }, { "epoch": 0.25, "learning_rate": 1.7646360454872076e-05, "loss": 0.6197, "step": 2025 }, { "epoch": 0.25, "learning_rate": 1.7643821448840573e-05, "loss": 0.5735, "step": 2026 }, { "epoch": 0.25, "learning_rate": 1.7641281256931146e-05, "loss": 0.5392, "step": 2027 }, { "epoch": 0.25, "learning_rate": 1.7638739879537895e-05, "loss": 0.6132, "step": 2028 }, { "epoch": 0.25, "learning_rate": 1.7636197317055082e-05, "loss": 0.698, "step": 2029 }, { "epoch": 0.25, "learning_rate": 1.7633653569877175e-05, "loss": 0.6017, "step": 2030 }, { "epoch": 0.25, "learning_rate": 1.7631108638398812e-05, "loss": 0.6307, "step": 2031 }, { "epoch": 0.25, "learning_rate": 1.7628562523014824e-05, "loss": 0.6329, "step": 2032 }, { "epoch": 0.25, "learning_rate": 1.762601522412021e-05, "loss": 0.614, "step": 2033 }, { "epoch": 0.25, "learning_rate": 1.7623466742110175e-05, "loss": 0.6123, "step": 2034 }, { "epoch": 0.25, "learning_rate": 1.7620917077380088e-05, "loss": 0.6175, "step": 2035 }, { "epoch": 0.25, "learning_rate": 1.7618366230325508e-05, "loss": 0.5969, "step": 2036 }, { "epoch": 0.25, "learning_rate": 1.7615814201342188e-05, "loss": 0.6745, "step": 2037 }, { "epoch": 0.25, "learning_rate": 1.7613260990826043e-05, "loss": 0.6933, "step": 2038 }, { "epoch": 0.25, "learning_rate": 1.7610706599173194e-05, "loss": 0.6522, "step": 2039 }, { "epoch": 0.25, "learning_rate": 1.7608151026779927e-05, "loss": 0.676, "step": 2040 }, { "epoch": 0.25, "learning_rate": 1.7605594274042724e-05, "loss": 0.6467, "step": 2041 }, { "epoch": 0.25, "learning_rate": 1.7603036341358238e-05, "loss": 0.7077, "step": 2042 }, { "epoch": 0.25, "learning_rate": 1.7600477229123316e-05, "loss": 0.5736, "step": 2043 }, { "epoch": 0.25, "learning_rate": 1.7597916937734987e-05, "loss": 0.6431, "step": 2044 }, { "epoch": 0.25, "learning_rate": 1.7595355467590455e-05, "loss": 0.6008, "step": 2045 }, { "epoch": 0.25, "learning_rate": 1.759279281908711e-05, "loss": 0.5946, "step": 2046 }, { "epoch": 0.25, "learning_rate": 1.7590228992622535e-05, "loss": 0.6327, "step": 2047 }, { "epoch": 0.25, "learning_rate": 1.758766398859448e-05, "loss": 0.6946, "step": 2048 }, { "epoch": 0.25, "learning_rate": 1.7585097807400886e-05, "loss": 0.6567, "step": 2049 }, { "epoch": 0.25, "learning_rate": 1.7582530449439875e-05, "loss": 0.6133, "step": 2050 }, { "epoch": 0.25, "learning_rate": 1.7579961915109755e-05, "loss": 0.5545, "step": 2051 }, { "epoch": 0.25, "learning_rate": 1.7577392204809013e-05, "loss": 0.5976, "step": 2052 }, { "epoch": 0.25, "learning_rate": 1.7574821318936318e-05, "loss": 0.6309, "step": 2053 }, { "epoch": 0.25, "learning_rate": 1.7572249257890523e-05, "loss": 0.6401, "step": 2054 }, { "epoch": 0.25, "learning_rate": 1.756967602207066e-05, "loss": 0.6058, "step": 2055 }, { "epoch": 0.25, "learning_rate": 1.7567101611875946e-05, "loss": 0.5998, "step": 2056 }, { "epoch": 0.25, "learning_rate": 1.756452602770579e-05, "loss": 0.595, "step": 2057 }, { "epoch": 0.25, "learning_rate": 1.756194926995976e-05, "loss": 0.6264, "step": 2058 }, { "epoch": 0.25, "learning_rate": 1.7559371339037626e-05, "loss": 0.5636, "step": 2059 }, { "epoch": 0.25, "learning_rate": 1.755679223533933e-05, "loss": 0.6227, "step": 2060 }, { "epoch": 0.25, "learning_rate": 1.7554211959265008e-05, "loss": 0.6141, "step": 2061 }, { "epoch": 0.25, "learning_rate": 1.755163051121496e-05, "loss": 0.5761, "step": 2062 }, { "epoch": 0.25, "learning_rate": 1.754904789158968e-05, "loss": 0.5978, "step": 2063 }, { "epoch": 0.25, "learning_rate": 1.7546464100789843e-05, "loss": 0.6081, "step": 2064 }, { "epoch": 0.25, "learning_rate": 1.75438791392163e-05, "loss": 0.5474, "step": 2065 }, { "epoch": 0.25, "learning_rate": 1.754129300727009e-05, "loss": 0.6477, "step": 2066 }, { "epoch": 0.25, "learning_rate": 1.753870570535243e-05, "loss": 0.593, "step": 2067 }, { "epoch": 0.25, "learning_rate": 1.753611723386472e-05, "loss": 0.5768, "step": 2068 }, { "epoch": 0.25, "learning_rate": 1.7533527593208535e-05, "loss": 0.6229, "step": 2069 }, { "epoch": 0.25, "learning_rate": 1.7530936783785645e-05, "loss": 0.59, "step": 2070 }, { "epoch": 0.25, "learning_rate": 1.752834480599799e-05, "loss": 0.6333, "step": 2071 }, { "epoch": 0.25, "learning_rate": 1.7525751660247695e-05, "loss": 0.6101, "step": 2072 }, { "epoch": 0.25, "learning_rate": 1.7523157346937063e-05, "loss": 0.5926, "step": 2073 }, { "epoch": 0.25, "learning_rate": 1.7520561866468585e-05, "loss": 0.6623, "step": 2074 }, { "epoch": 0.25, "learning_rate": 1.751796521924493e-05, "loss": 0.5642, "step": 2075 }, { "epoch": 0.25, "learning_rate": 1.7515367405668942e-05, "loss": 0.5984, "step": 2076 }, { "epoch": 0.25, "learning_rate": 1.7512768426143653e-05, "loss": 0.6083, "step": 2077 }, { "epoch": 0.25, "learning_rate": 1.7510168281072277e-05, "loss": 0.5836, "step": 2078 }, { "epoch": 0.25, "learning_rate": 1.7507566970858202e-05, "loss": 0.6277, "step": 2079 }, { "epoch": 0.25, "learning_rate": 1.7504964495905004e-05, "loss": 0.6333, "step": 2080 }, { "epoch": 0.25, "learning_rate": 1.7502360856616435e-05, "loss": 0.5715, "step": 2081 }, { "epoch": 0.25, "learning_rate": 1.7499756053396426e-05, "loss": 0.6821, "step": 2082 }, { "epoch": 0.25, "learning_rate": 1.7497150086649094e-05, "loss": 0.6096, "step": 2083 }, { "epoch": 0.25, "learning_rate": 1.7494542956778735e-05, "loss": 0.6647, "step": 2084 }, { "epoch": 0.25, "learning_rate": 1.749193466418982e-05, "loss": 0.5834, "step": 2085 }, { "epoch": 0.25, "learning_rate": 1.7489325209287012e-05, "loss": 0.6078, "step": 2086 }, { "epoch": 0.25, "learning_rate": 1.7486714592475137e-05, "loss": 0.5962, "step": 2087 }, { "epoch": 0.25, "learning_rate": 1.7484102814159222e-05, "loss": 0.5554, "step": 2088 }, { "epoch": 0.25, "learning_rate": 1.748148987474446e-05, "loss": 0.6212, "step": 2089 }, { "epoch": 0.25, "learning_rate": 1.7478875774636223e-05, "loss": 0.6075, "step": 2090 }, { "epoch": 0.25, "learning_rate": 1.7476260514240073e-05, "loss": 0.6101, "step": 2091 }, { "epoch": 0.25, "learning_rate": 1.7473644093961743e-05, "loss": 0.5927, "step": 2092 }, { "epoch": 0.25, "learning_rate": 1.7471026514207153e-05, "loss": 0.6542, "step": 2093 }, { "epoch": 0.25, "learning_rate": 1.7468407775382403e-05, "loss": 0.5808, "step": 2094 }, { "epoch": 0.25, "learning_rate": 1.746578787789376e-05, "loss": 0.63, "step": 2095 }, { "epoch": 0.25, "learning_rate": 1.7463166822147685e-05, "loss": 0.5974, "step": 2096 }, { "epoch": 0.26, "learning_rate": 1.7460544608550814e-05, "loss": 0.5652, "step": 2097 }, { "epoch": 0.26, "learning_rate": 1.7457921237509965e-05, "loss": 0.5987, "step": 2098 }, { "epoch": 0.26, "learning_rate": 1.7455296709432128e-05, "loss": 0.561, "step": 2099 }, { "epoch": 0.26, "learning_rate": 1.7452671024724482e-05, "loss": 0.5543, "step": 2100 }, { "epoch": 0.26, "learning_rate": 1.7450044183794377e-05, "loss": 0.6122, "step": 2101 }, { "epoch": 0.26, "learning_rate": 1.744741618704935e-05, "loss": 0.6249, "step": 2102 }, { "epoch": 0.26, "learning_rate": 1.7444787034897114e-05, "loss": 0.5992, "step": 2103 }, { "epoch": 0.26, "learning_rate": 1.7442156727745558e-05, "loss": 0.5956, "step": 2104 }, { "epoch": 0.26, "learning_rate": 1.743952526600275e-05, "loss": 0.5835, "step": 2105 }, { "epoch": 0.26, "learning_rate": 1.7436892650076954e-05, "loss": 0.6586, "step": 2106 }, { "epoch": 0.26, "learning_rate": 1.7434258880376584e-05, "loss": 0.6031, "step": 2107 }, { "epoch": 0.26, "learning_rate": 1.7431623957310254e-05, "loss": 0.58, "step": 2108 }, { "epoch": 0.26, "learning_rate": 1.7428987881286756e-05, "loss": 0.528, "step": 2109 }, { "epoch": 0.26, "learning_rate": 1.7426350652715047e-05, "loss": 0.5788, "step": 2110 }, { "epoch": 0.26, "learning_rate": 1.7423712272004282e-05, "loss": 0.6223, "step": 2111 }, { "epoch": 0.26, "learning_rate": 1.7421072739563777e-05, "loss": 0.5995, "step": 2112 }, { "epoch": 0.26, "learning_rate": 1.7418432055803035e-05, "loss": 0.5964, "step": 2113 }, { "epoch": 0.26, "learning_rate": 1.7415790221131743e-05, "loss": 0.6449, "step": 2114 }, { "epoch": 0.26, "learning_rate": 1.7413147235959757e-05, "loss": 0.5953, "step": 2115 }, { "epoch": 0.26, "learning_rate": 1.741050310069711e-05, "loss": 0.5442, "step": 2116 }, { "epoch": 0.26, "learning_rate": 1.740785781575403e-05, "loss": 0.5638, "step": 2117 }, { "epoch": 0.26, "learning_rate": 1.74052113815409e-05, "loss": 0.5621, "step": 2118 }, { "epoch": 0.26, "learning_rate": 1.7402563798468297e-05, "loss": 0.6571, "step": 2119 }, { "epoch": 0.26, "learning_rate": 1.7399915066946977e-05, "loss": 0.6317, "step": 2120 }, { "epoch": 0.26, "learning_rate": 1.7397265187387867e-05, "loss": 0.5427, "step": 2121 }, { "epoch": 0.26, "learning_rate": 1.739461416020207e-05, "loss": 0.6128, "step": 2122 }, { "epoch": 0.26, "learning_rate": 1.7391961985800877e-05, "loss": 0.6098, "step": 2123 }, { "epoch": 0.26, "learning_rate": 1.7389308664595754e-05, "loss": 0.5948, "step": 2124 }, { "epoch": 0.26, "learning_rate": 1.7386654196998337e-05, "loss": 0.5924, "step": 2125 }, { "epoch": 0.26, "learning_rate": 1.7383998583420444e-05, "loss": 0.5799, "step": 2126 }, { "epoch": 0.26, "learning_rate": 1.738134182427408e-05, "loss": 0.5952, "step": 2127 }, { "epoch": 0.26, "learning_rate": 1.7378683919971412e-05, "loss": 0.6071, "step": 2128 }, { "epoch": 0.26, "learning_rate": 1.73760248709248e-05, "loss": 0.6517, "step": 2129 }, { "epoch": 0.26, "learning_rate": 1.7373364677546767e-05, "loss": 0.5819, "step": 2130 }, { "epoch": 0.26, "learning_rate": 1.737070334025003e-05, "loss": 0.5682, "step": 2131 }, { "epoch": 0.26, "learning_rate": 1.7368040859447465e-05, "loss": 0.5787, "step": 2132 }, { "epoch": 0.26, "learning_rate": 1.7365377235552133e-05, "loss": 0.5678, "step": 2133 }, { "epoch": 0.26, "learning_rate": 1.7362712468977286e-05, "loss": 0.6421, "step": 2134 }, { "epoch": 0.26, "learning_rate": 1.736004656013633e-05, "loss": 0.6031, "step": 2135 }, { "epoch": 0.26, "learning_rate": 1.7357379509442867e-05, "loss": 0.581, "step": 2136 }, { "epoch": 0.26, "learning_rate": 1.735471131731067e-05, "loss": 0.5742, "step": 2137 }, { "epoch": 0.26, "learning_rate": 1.7352041984153675e-05, "loss": 0.5699, "step": 2138 }, { "epoch": 0.26, "learning_rate": 1.7349371510386022e-05, "loss": 0.6347, "step": 2139 }, { "epoch": 0.26, "learning_rate": 1.7346699896422007e-05, "loss": 0.6299, "step": 2140 }, { "epoch": 0.26, "learning_rate": 1.734402714267611e-05, "loss": 0.5968, "step": 2141 }, { "epoch": 0.26, "learning_rate": 1.7341353249562982e-05, "loss": 0.594, "step": 2142 }, { "epoch": 0.26, "learning_rate": 1.7338678217497468e-05, "loss": 0.6103, "step": 2143 }, { "epoch": 0.26, "learning_rate": 1.733600204689457e-05, "loss": 0.5933, "step": 2144 }, { "epoch": 0.26, "learning_rate": 1.733332473816948e-05, "loss": 0.6373, "step": 2145 }, { "epoch": 0.26, "learning_rate": 1.7330646291737553e-05, "loss": 0.6284, "step": 2146 }, { "epoch": 0.26, "learning_rate": 1.7327966708014333e-05, "loss": 0.6308, "step": 2147 }, { "epoch": 0.26, "learning_rate": 1.732528598741554e-05, "loss": 0.6191, "step": 2148 }, { "epoch": 0.26, "learning_rate": 1.732260413035706e-05, "loss": 0.6336, "step": 2149 }, { "epoch": 0.26, "learning_rate": 1.7319921137254964e-05, "loss": 0.6014, "step": 2150 }, { "epoch": 0.26, "learning_rate": 1.73172370085255e-05, "loss": 0.5719, "step": 2151 }, { "epoch": 0.26, "learning_rate": 1.7314551744585083e-05, "loss": 0.627, "step": 2152 }, { "epoch": 0.26, "learning_rate": 1.7311865345850315e-05, "loss": 0.6105, "step": 2153 }, { "epoch": 0.26, "learning_rate": 1.7309177812737966e-05, "loss": 0.6215, "step": 2154 }, { "epoch": 0.26, "learning_rate": 1.7306489145664992e-05, "loss": 0.5955, "step": 2155 }, { "epoch": 0.26, "learning_rate": 1.730379934504851e-05, "loss": 0.617, "step": 2156 }, { "epoch": 0.26, "learning_rate": 1.7301108411305828e-05, "loss": 0.5772, "step": 2157 }, { "epoch": 0.26, "learning_rate": 1.7298416344854415e-05, "loss": 0.6076, "step": 2158 }, { "epoch": 0.26, "learning_rate": 1.7295723146111934e-05, "loss": 0.6708, "step": 2159 }, { "epoch": 0.26, "learning_rate": 1.72930288154962e-05, "loss": 0.6348, "step": 2160 }, { "epoch": 0.26, "learning_rate": 1.7290333353425232e-05, "loss": 0.6542, "step": 2161 }, { "epoch": 0.26, "learning_rate": 1.72876367603172e-05, "loss": 0.5803, "step": 2162 }, { "epoch": 0.26, "learning_rate": 1.728493903659046e-05, "loss": 0.5787, "step": 2163 }, { "epoch": 0.26, "learning_rate": 1.728224018266354e-05, "loss": 0.6256, "step": 2164 }, { "epoch": 0.26, "learning_rate": 1.7279540198955153e-05, "loss": 0.5791, "step": 2165 }, { "epoch": 0.26, "learning_rate": 1.7276839085884174e-05, "loss": 0.6233, "step": 2166 }, { "epoch": 0.26, "learning_rate": 1.7274136843869662e-05, "loss": 0.6098, "step": 2167 }, { "epoch": 0.26, "learning_rate": 1.7271433473330846e-05, "loss": 0.5736, "step": 2168 }, { "epoch": 0.26, "learning_rate": 1.726872897468713e-05, "loss": 0.5749, "step": 2169 }, { "epoch": 0.26, "learning_rate": 1.7266023348358105e-05, "loss": 0.6084, "step": 2170 }, { "epoch": 0.26, "learning_rate": 1.7263316594763518e-05, "loss": 0.6024, "step": 2171 }, { "epoch": 0.26, "learning_rate": 1.7260608714323305e-05, "loss": 0.6708, "step": 2172 }, { "epoch": 0.26, "learning_rate": 1.725789970745757e-05, "loss": 0.6127, "step": 2173 }, { "epoch": 0.26, "learning_rate": 1.7255189574586592e-05, "loss": 0.6067, "step": 2174 }, { "epoch": 0.26, "learning_rate": 1.7252478316130834e-05, "loss": 0.6056, "step": 2175 }, { "epoch": 0.26, "learning_rate": 1.7249765932510918e-05, "loss": 0.6117, "step": 2176 }, { "epoch": 0.26, "learning_rate": 1.724705242414765e-05, "loss": 0.6473, "step": 2177 }, { "epoch": 0.26, "learning_rate": 1.7244337791462014e-05, "loss": 0.643, "step": 2178 }, { "epoch": 0.26, "learning_rate": 1.7241622034875156e-05, "loss": 0.6076, "step": 2179 }, { "epoch": 0.27, "learning_rate": 1.7238905154808413e-05, "loss": 0.6131, "step": 2180 }, { "epoch": 0.27, "learning_rate": 1.7236187151683283e-05, "loss": 0.5997, "step": 2181 }, { "epoch": 0.27, "learning_rate": 1.7233468025921437e-05, "loss": 0.6645, "step": 2182 }, { "epoch": 0.27, "learning_rate": 1.7230747777944735e-05, "loss": 0.6511, "step": 2183 }, { "epoch": 0.27, "learning_rate": 1.7228026408175195e-05, "loss": 0.6014, "step": 2184 }, { "epoch": 0.27, "learning_rate": 1.722530391703502e-05, "loss": 0.6396, "step": 2185 }, { "epoch": 0.27, "learning_rate": 1.722258030494658e-05, "loss": 0.6347, "step": 2186 }, { "epoch": 0.27, "learning_rate": 1.7219855572332425e-05, "loss": 0.6679, "step": 2187 }, { "epoch": 0.27, "learning_rate": 1.7217129719615273e-05, "loss": 0.597, "step": 2188 }, { "epoch": 0.27, "learning_rate": 1.7214402747218017e-05, "loss": 0.5128, "step": 2189 }, { "epoch": 0.27, "learning_rate": 1.7211674655563727e-05, "loss": 0.6386, "step": 2190 }, { "epoch": 0.27, "learning_rate": 1.7208945445075648e-05, "loss": 0.577, "step": 2191 }, { "epoch": 0.27, "learning_rate": 1.7206215116177185e-05, "loss": 0.6421, "step": 2192 }, { "epoch": 0.27, "learning_rate": 1.720348366929194e-05, "loss": 0.5705, "step": 2193 }, { "epoch": 0.27, "learning_rate": 1.7200751104843664e-05, "loss": 0.5756, "step": 2194 }, { "epoch": 0.27, "learning_rate": 1.71980174232563e-05, "loss": 0.652, "step": 2195 }, { "epoch": 0.27, "learning_rate": 1.719528262495395e-05, "loss": 0.6025, "step": 2196 }, { "epoch": 0.27, "learning_rate": 1.7192546710360906e-05, "loss": 0.5986, "step": 2197 }, { "epoch": 0.27, "learning_rate": 1.718980967990161e-05, "loss": 0.6442, "step": 2198 }, { "epoch": 0.27, "learning_rate": 1.7187071534000704e-05, "loss": 0.5709, "step": 2199 }, { "epoch": 0.27, "learning_rate": 1.7184332273082983e-05, "loss": 0.5331, "step": 2200 }, { "epoch": 0.27, "learning_rate": 1.718159189757342e-05, "loss": 0.5878, "step": 2201 }, { "epoch": 0.27, "learning_rate": 1.7178850407897165e-05, "loss": 0.5819, "step": 2202 }, { "epoch": 0.27, "learning_rate": 1.717610780447954e-05, "loss": 0.6291, "step": 2203 }, { "epoch": 0.27, "learning_rate": 1.7173364087746032e-05, "loss": 0.5764, "step": 2204 }, { "epoch": 0.27, "learning_rate": 1.7170619258122313e-05, "loss": 0.5913, "step": 2205 }, { "epoch": 0.27, "learning_rate": 1.7167873316034215e-05, "loss": 0.6184, "step": 2206 }, { "epoch": 0.27, "learning_rate": 1.7165126261907755e-05, "loss": 0.6117, "step": 2207 }, { "epoch": 0.27, "learning_rate": 1.7162378096169116e-05, "loss": 0.6078, "step": 2208 }, { "epoch": 0.27, "learning_rate": 1.715962881924465e-05, "loss": 0.5621, "step": 2209 }, { "epoch": 0.27, "learning_rate": 1.715687843156089e-05, "loss": 0.6116, "step": 2210 }, { "epoch": 0.27, "learning_rate": 1.7154126933544532e-05, "loss": 0.6083, "step": 2211 }, { "epoch": 0.27, "learning_rate": 1.7151374325622452e-05, "loss": 0.6569, "step": 2212 }, { "epoch": 0.27, "learning_rate": 1.7148620608221693e-05, "loss": 0.5438, "step": 2213 }, { "epoch": 0.27, "learning_rate": 1.7145865781769478e-05, "loss": 0.6639, "step": 2214 }, { "epoch": 0.27, "learning_rate": 1.714310984669319e-05, "loss": 0.615, "step": 2215 }, { "epoch": 0.27, "learning_rate": 1.714035280342039e-05, "loss": 0.6212, "step": 2216 }, { "epoch": 0.27, "learning_rate": 1.7137594652378817e-05, "loss": 0.6347, "step": 2217 }, { "epoch": 0.27, "learning_rate": 1.7134835393996372e-05, "loss": 0.6143, "step": 2218 }, { "epoch": 0.27, "learning_rate": 1.7132075028701132e-05, "loss": 0.579, "step": 2219 }, { "epoch": 0.27, "learning_rate": 1.712931355692135e-05, "loss": 0.5818, "step": 2220 }, { "epoch": 0.27, "learning_rate": 1.7126550979085437e-05, "loss": 0.6277, "step": 2221 }, { "epoch": 0.27, "learning_rate": 1.7123787295621995e-05, "loss": 0.6702, "step": 2222 }, { "epoch": 0.27, "learning_rate": 1.712102250695978e-05, "loss": 0.5493, "step": 2223 }, { "epoch": 0.27, "learning_rate": 1.7118256613527734e-05, "loss": 0.6462, "step": 2224 }, { "epoch": 0.27, "learning_rate": 1.7115489615754954e-05, "loss": 0.5991, "step": 2225 }, { "epoch": 0.27, "learning_rate": 1.711272151407073e-05, "loss": 0.5532, "step": 2226 }, { "epoch": 0.27, "learning_rate": 1.71099523089045e-05, "loss": 0.6257, "step": 2227 }, { "epoch": 0.27, "learning_rate": 1.710718200068589e-05, "loss": 0.6285, "step": 2228 }, { "epoch": 0.27, "learning_rate": 1.7104410589844684e-05, "loss": 0.5698, "step": 2229 }, { "epoch": 0.27, "learning_rate": 1.7101638076810855e-05, "loss": 0.5488, "step": 2230 }, { "epoch": 0.27, "learning_rate": 1.709886446201453e-05, "loss": 0.5487, "step": 2231 }, { "epoch": 0.27, "learning_rate": 1.709608974588601e-05, "loss": 0.6299, "step": 2232 }, { "epoch": 0.27, "learning_rate": 1.7093313928855778e-05, "loss": 0.5861, "step": 2233 }, { "epoch": 0.27, "learning_rate": 1.7090537011354473e-05, "loss": 0.692, "step": 2234 }, { "epoch": 0.27, "learning_rate": 1.7087758993812916e-05, "loss": 0.677, "step": 2235 }, { "epoch": 0.27, "learning_rate": 1.7084979876662094e-05, "loss": 0.6304, "step": 2236 }, { "epoch": 0.27, "learning_rate": 1.708219966033316e-05, "loss": 0.6521, "step": 2237 }, { "epoch": 0.27, "learning_rate": 1.707941834525745e-05, "loss": 0.6317, "step": 2238 }, { "epoch": 0.27, "learning_rate": 1.707663593186646e-05, "loss": 0.576, "step": 2239 }, { "epoch": 0.27, "learning_rate": 1.7073852420591853e-05, "loss": 0.6242, "step": 2240 }, { "epoch": 0.27, "learning_rate": 1.7071067811865477e-05, "loss": 0.6287, "step": 2241 }, { "epoch": 0.27, "learning_rate": 1.7068282106119338e-05, "loss": 0.6393, "step": 2242 }, { "epoch": 0.27, "learning_rate": 1.7065495303785616e-05, "loss": 0.5491, "step": 2243 }, { "epoch": 0.27, "learning_rate": 1.706270740529666e-05, "loss": 0.5792, "step": 2244 }, { "epoch": 0.27, "learning_rate": 1.7059918411084996e-05, "loss": 0.6179, "step": 2245 }, { "epoch": 0.27, "learning_rate": 1.705712832158331e-05, "loss": 0.6414, "step": 2246 }, { "epoch": 0.27, "learning_rate": 1.705433713722446e-05, "loss": 0.6045, "step": 2247 }, { "epoch": 0.27, "learning_rate": 1.7051544858441476e-05, "loss": 0.6019, "step": 2248 }, { "epoch": 0.27, "learning_rate": 1.704875148566756e-05, "loss": 0.5716, "step": 2249 }, { "epoch": 0.27, "learning_rate": 1.7045957019336087e-05, "loss": 0.6412, "step": 2250 }, { "epoch": 0.27, "learning_rate": 1.7043161459880587e-05, "loss": 0.6299, "step": 2251 }, { "epoch": 0.27, "learning_rate": 1.7040364807734774e-05, "loss": 0.5994, "step": 2252 }, { "epoch": 0.27, "learning_rate": 1.7037567063332522e-05, "loss": 0.6087, "step": 2253 }, { "epoch": 0.27, "learning_rate": 1.7034768227107884e-05, "loss": 0.5934, "step": 2254 }, { "epoch": 0.27, "learning_rate": 1.7031968299495073e-05, "loss": 0.5876, "step": 2255 }, { "epoch": 0.27, "learning_rate": 1.7029167280928478e-05, "loss": 0.6198, "step": 2256 }, { "epoch": 0.27, "learning_rate": 1.7026365171842654e-05, "loss": 0.6179, "step": 2257 }, { "epoch": 0.27, "learning_rate": 1.7023561972672325e-05, "loss": 0.6266, "step": 2258 }, { "epoch": 0.27, "learning_rate": 1.7020757683852386e-05, "loss": 0.5599, "step": 2259 }, { "epoch": 0.27, "learning_rate": 1.7017952305817895e-05, "loss": 0.5685, "step": 2260 }, { "epoch": 0.27, "learning_rate": 1.7015145839004094e-05, "loss": 0.5496, "step": 2261 }, { "epoch": 0.28, "learning_rate": 1.7012338283846375e-05, "loss": 0.6139, "step": 2262 }, { "epoch": 0.28, "learning_rate": 1.7009529640780312e-05, "loss": 0.5662, "step": 2263 }, { "epoch": 0.28, "learning_rate": 1.7006719910241646e-05, "loss": 0.6798, "step": 2264 }, { "epoch": 0.28, "learning_rate": 1.7003909092666278e-05, "loss": 0.6404, "step": 2265 }, { "epoch": 0.28, "learning_rate": 1.7001097188490287e-05, "loss": 0.5624, "step": 2266 }, { "epoch": 0.28, "learning_rate": 1.699828419814992e-05, "loss": 0.6606, "step": 2267 }, { "epoch": 0.28, "learning_rate": 1.6995470122081586e-05, "loss": 0.629, "step": 2268 }, { "epoch": 0.28, "learning_rate": 1.6992654960721867e-05, "loss": 0.5827, "step": 2269 }, { "epoch": 0.28, "learning_rate": 1.6989838714507512e-05, "loss": 0.6229, "step": 2270 }, { "epoch": 0.28, "learning_rate": 1.6987021383875446e-05, "loss": 0.626, "step": 2271 }, { "epoch": 0.28, "learning_rate": 1.6984202969262747e-05, "loss": 0.5537, "step": 2272 }, { "epoch": 0.28, "learning_rate": 1.6981383471106677e-05, "loss": 0.6216, "step": 2273 }, { "epoch": 0.28, "learning_rate": 1.6978562889844652e-05, "loss": 0.5901, "step": 2274 }, { "epoch": 0.28, "learning_rate": 1.6975741225914264e-05, "loss": 0.5755, "step": 2275 }, { "epoch": 0.28, "learning_rate": 1.6972918479753272e-05, "loss": 0.624, "step": 2276 }, { "epoch": 0.28, "learning_rate": 1.6970094651799607e-05, "loss": 0.6365, "step": 2277 }, { "epoch": 0.28, "learning_rate": 1.6967269742491356e-05, "loss": 0.6019, "step": 2278 }, { "epoch": 0.28, "learning_rate": 1.6964443752266787e-05, "loss": 0.6425, "step": 2279 }, { "epoch": 0.28, "learning_rate": 1.696161668156433e-05, "loss": 0.5924, "step": 2280 }, { "epoch": 0.28, "learning_rate": 1.6958788530822573e-05, "loss": 0.6638, "step": 2281 }, { "epoch": 0.28, "learning_rate": 1.6955959300480296e-05, "loss": 0.5983, "step": 2282 }, { "epoch": 0.28, "learning_rate": 1.695312899097642e-05, "loss": 0.6314, "step": 2283 }, { "epoch": 0.28, "learning_rate": 1.6950297602750052e-05, "loss": 0.6276, "step": 2284 }, { "epoch": 0.28, "learning_rate": 1.6947465136240453e-05, "loss": 0.6301, "step": 2285 }, { "epoch": 0.28, "learning_rate": 1.6944631591887062e-05, "loss": 0.6077, "step": 2286 }, { "epoch": 0.28, "learning_rate": 1.6941796970129477e-05, "loss": 0.6224, "step": 2287 }, { "epoch": 0.28, "learning_rate": 1.693896127140747e-05, "loss": 0.5959, "step": 2288 }, { "epoch": 0.28, "learning_rate": 1.6936124496160978e-05, "loss": 0.5821, "step": 2289 }, { "epoch": 0.28, "learning_rate": 1.6933286644830104e-05, "loss": 0.6017, "step": 2290 }, { "epoch": 0.28, "learning_rate": 1.6930447717855112e-05, "loss": 0.5592, "step": 2291 }, { "epoch": 0.28, "learning_rate": 1.6927607715676446e-05, "loss": 0.6355, "step": 2292 }, { "epoch": 0.28, "learning_rate": 1.6924766638734706e-05, "loss": 0.6261, "step": 2293 }, { "epoch": 0.28, "learning_rate": 1.6921924487470663e-05, "loss": 0.5932, "step": 2294 }, { "epoch": 0.28, "learning_rate": 1.6919081262325257e-05, "loss": 0.601, "step": 2295 }, { "epoch": 0.28, "learning_rate": 1.6916236963739587e-05, "loss": 0.6207, "step": 2296 }, { "epoch": 0.28, "learning_rate": 1.6913391592154928e-05, "loss": 0.645, "step": 2297 }, { "epoch": 0.28, "learning_rate": 1.6910545148012712e-05, "loss": 0.6405, "step": 2298 }, { "epoch": 0.28, "learning_rate": 1.6907697631754542e-05, "loss": 0.5755, "step": 2299 }, { "epoch": 0.28, "learning_rate": 1.6904849043822193e-05, "loss": 0.6186, "step": 2300 }, { "epoch": 0.28, "learning_rate": 1.6901999384657596e-05, "loss": 0.6771, "step": 2301 }, { "epoch": 0.28, "learning_rate": 1.6899148654702858e-05, "loss": 0.5881, "step": 2302 }, { "epoch": 0.28, "learning_rate": 1.6896296854400242e-05, "loss": 0.6771, "step": 2303 }, { "epoch": 0.28, "learning_rate": 1.689344398419218e-05, "loss": 0.5963, "step": 2304 }, { "epoch": 0.28, "learning_rate": 1.6890590044521275e-05, "loss": 0.5882, "step": 2305 }, { "epoch": 0.28, "learning_rate": 1.6887735035830293e-05, "loss": 0.6088, "step": 2306 }, { "epoch": 0.28, "learning_rate": 1.6884878958562173e-05, "loss": 0.5996, "step": 2307 }, { "epoch": 0.28, "learning_rate": 1.688202181316e-05, "loss": 0.6229, "step": 2308 }, { "epoch": 0.28, "learning_rate": 1.6879163600067042e-05, "loss": 0.6222, "step": 2309 }, { "epoch": 0.28, "learning_rate": 1.6876304319726732e-05, "loss": 0.6151, "step": 2310 }, { "epoch": 0.28, "learning_rate": 1.6873443972582662e-05, "loss": 0.6029, "step": 2311 }, { "epoch": 0.28, "learning_rate": 1.687058255907859e-05, "loss": 0.5794, "step": 2312 }, { "epoch": 0.28, "learning_rate": 1.6867720079658444e-05, "loss": 0.5994, "step": 2313 }, { "epoch": 0.28, "learning_rate": 1.6864856534766314e-05, "loss": 0.5834, "step": 2314 }, { "epoch": 0.28, "learning_rate": 1.6861991924846457e-05, "loss": 0.5987, "step": 2315 }, { "epoch": 0.28, "learning_rate": 1.6859126250343295e-05, "loss": 0.5716, "step": 2316 }, { "epoch": 0.28, "learning_rate": 1.6856259511701407e-05, "loss": 0.5633, "step": 2317 }, { "epoch": 0.28, "learning_rate": 1.6853391709365555e-05, "loss": 0.6317, "step": 2318 }, { "epoch": 0.28, "learning_rate": 1.685052284378065e-05, "loss": 0.592, "step": 2319 }, { "epoch": 0.28, "learning_rate": 1.6847652915391776e-05, "loss": 0.6283, "step": 2320 }, { "epoch": 0.28, "learning_rate": 1.6844781924644177e-05, "loss": 0.6032, "step": 2321 }, { "epoch": 0.28, "learning_rate": 1.684190987198327e-05, "loss": 0.5553, "step": 2322 }, { "epoch": 0.28, "learning_rate": 1.683903675785462e-05, "loss": 0.6323, "step": 2323 }, { "epoch": 0.28, "learning_rate": 1.683616258270398e-05, "loss": 0.6238, "step": 2324 }, { "epoch": 0.28, "learning_rate": 1.6833287346977246e-05, "loss": 0.5975, "step": 2325 }, { "epoch": 0.28, "learning_rate": 1.6830411051120496e-05, "loss": 0.6185, "step": 2326 }, { "epoch": 0.28, "learning_rate": 1.6827533695579958e-05, "loss": 0.5798, "step": 2327 }, { "epoch": 0.28, "learning_rate": 1.6824655280802032e-05, "loss": 0.5843, "step": 2328 }, { "epoch": 0.28, "learning_rate": 1.6821775807233282e-05, "loss": 0.5951, "step": 2329 }, { "epoch": 0.28, "learning_rate": 1.6818895275320435e-05, "loss": 0.6259, "step": 2330 }, { "epoch": 0.28, "learning_rate": 1.6816013685510383e-05, "loss": 0.564, "step": 2331 }, { "epoch": 0.28, "learning_rate": 1.6813131038250183e-05, "loss": 0.6275, "step": 2332 }, { "epoch": 0.28, "learning_rate": 1.6810247333987052e-05, "loss": 0.5533, "step": 2333 }, { "epoch": 0.28, "learning_rate": 1.6807362573168378e-05, "loss": 0.632, "step": 2334 }, { "epoch": 0.28, "learning_rate": 1.68044767562417e-05, "loss": 0.6056, "step": 2335 }, { "epoch": 0.28, "learning_rate": 1.680158988365474e-05, "loss": 0.5962, "step": 2336 }, { "epoch": 0.28, "learning_rate": 1.6798701955855365e-05, "loss": 0.5855, "step": 2337 }, { "epoch": 0.28, "learning_rate": 1.679581297329162e-05, "loss": 0.6325, "step": 2338 }, { "epoch": 0.28, "learning_rate": 1.6792922936411705e-05, "loss": 0.6674, "step": 2339 }, { "epoch": 0.28, "learning_rate": 1.6790031845663984e-05, "loss": 0.5739, "step": 2340 }, { "epoch": 0.28, "learning_rate": 1.678713970149699e-05, "loss": 0.537, "step": 2341 }, { "epoch": 0.28, "learning_rate": 1.6784246504359412e-05, "loss": 0.5841, "step": 2342 }, { "epoch": 0.28, "learning_rate": 1.6781352254700114e-05, "loss": 0.5429, "step": 2343 }, { "epoch": 0.29, "learning_rate": 1.677845695296811e-05, "loss": 0.5866, "step": 2344 }, { "epoch": 0.29, "learning_rate": 1.6775560599612583e-05, "loss": 0.6291, "step": 2345 }, { "epoch": 0.29, "learning_rate": 1.677266319508288e-05, "loss": 0.5996, "step": 2346 }, { "epoch": 0.29, "learning_rate": 1.6769764739828512e-05, "loss": 0.5493, "step": 2347 }, { "epoch": 0.29, "learning_rate": 1.6766865234299147e-05, "loss": 0.574, "step": 2348 }, { "epoch": 0.29, "learning_rate": 1.6763964678944625e-05, "loss": 0.5864, "step": 2349 }, { "epoch": 0.29, "learning_rate": 1.676106307421494e-05, "loss": 0.6395, "step": 2350 }, { "epoch": 0.29, "learning_rate": 1.6758160420560253e-05, "loss": 0.6506, "step": 2351 }, { "epoch": 0.29, "learning_rate": 1.675525671843089e-05, "loss": 0.6021, "step": 2352 }, { "epoch": 0.29, "learning_rate": 1.6752351968277337e-05, "loss": 0.6186, "step": 2353 }, { "epoch": 0.29, "learning_rate": 1.674944617055024e-05, "loss": 0.5494, "step": 2354 }, { "epoch": 0.29, "learning_rate": 1.6746539325700417e-05, "loss": 0.6298, "step": 2355 }, { "epoch": 0.29, "learning_rate": 1.6743631434178834e-05, "loss": 0.6486, "step": 2356 }, { "epoch": 0.29, "learning_rate": 1.6740722496436633e-05, "loss": 0.5876, "step": 2357 }, { "epoch": 0.29, "learning_rate": 1.6737812512925108e-05, "loss": 0.6416, "step": 2358 }, { "epoch": 0.29, "learning_rate": 1.6734901484095722e-05, "loss": 0.5942, "step": 2359 }, { "epoch": 0.29, "learning_rate": 1.67319894104001e-05, "loss": 0.5363, "step": 2360 }, { "epoch": 0.29, "learning_rate": 1.672907629229002e-05, "loss": 0.5797, "step": 2361 }, { "epoch": 0.29, "learning_rate": 1.6726162130217436e-05, "loss": 0.606, "step": 2362 }, { "epoch": 0.29, "learning_rate": 1.6723246924634455e-05, "loss": 0.616, "step": 2363 }, { "epoch": 0.29, "learning_rate": 1.6720330675993345e-05, "loss": 0.539, "step": 2364 }, { "epoch": 0.29, "learning_rate": 1.6717413384746545e-05, "loss": 0.6325, "step": 2365 }, { "epoch": 0.29, "learning_rate": 1.6714495051346643e-05, "loss": 0.5909, "step": 2366 }, { "epoch": 0.29, "learning_rate": 1.6711575676246398e-05, "loss": 0.6326, "step": 2367 }, { "epoch": 0.29, "learning_rate": 1.6708655259898726e-05, "loss": 0.6068, "step": 2368 }, { "epoch": 0.29, "learning_rate": 1.670573380275671e-05, "loss": 0.6241, "step": 2369 }, { "epoch": 0.29, "learning_rate": 1.670281130527359e-05, "loss": 0.5647, "step": 2370 }, { "epoch": 0.29, "learning_rate": 1.6699887767902764e-05, "loss": 0.6538, "step": 2371 }, { "epoch": 0.29, "learning_rate": 1.66969631910978e-05, "loss": 0.6221, "step": 2372 }, { "epoch": 0.29, "learning_rate": 1.669403757531242e-05, "loss": 0.6595, "step": 2373 }, { "epoch": 0.29, "learning_rate": 1.6691110921000514e-05, "loss": 0.5998, "step": 2374 }, { "epoch": 0.29, "learning_rate": 1.668818322861612e-05, "loss": 0.5979, "step": 2375 }, { "epoch": 0.29, "learning_rate": 1.6685254498613458e-05, "loss": 0.5755, "step": 2376 }, { "epoch": 0.29, "learning_rate": 1.6682324731446886e-05, "loss": 0.5827, "step": 2377 }, { "epoch": 0.29, "learning_rate": 1.6679393927570943e-05, "loss": 0.5955, "step": 2378 }, { "epoch": 0.29, "learning_rate": 1.6676462087440314e-05, "loss": 0.5421, "step": 2379 }, { "epoch": 0.29, "learning_rate": 1.6673529211509854e-05, "loss": 0.633, "step": 2380 }, { "epoch": 0.29, "learning_rate": 1.6670595300234578e-05, "loss": 0.6157, "step": 2381 }, { "epoch": 0.29, "learning_rate": 1.666766035406965e-05, "loss": 0.6511, "step": 2382 }, { "epoch": 0.29, "learning_rate": 1.666472437347041e-05, "loss": 0.4701, "step": 2383 }, { "epoch": 0.29, "learning_rate": 1.666178735889235e-05, "loss": 0.652, "step": 2384 }, { "epoch": 0.29, "learning_rate": 1.6658849310791125e-05, "loss": 0.6203, "step": 2385 }, { "epoch": 0.29, "learning_rate": 1.665591022962255e-05, "loss": 0.5523, "step": 2386 }, { "epoch": 0.29, "learning_rate": 1.6652970115842597e-05, "loss": 0.598, "step": 2387 }, { "epoch": 0.29, "learning_rate": 1.6650028969907407e-05, "loss": 0.5937, "step": 2388 }, { "epoch": 0.29, "learning_rate": 1.6647086792273267e-05, "loss": 0.5924, "step": 2389 }, { "epoch": 0.29, "learning_rate": 1.6644143583396642e-05, "loss": 0.6178, "step": 2390 }, { "epoch": 0.29, "learning_rate": 1.6641199343734142e-05, "loss": 0.6588, "step": 2391 }, { "epoch": 0.29, "learning_rate": 1.6638254073742547e-05, "loss": 0.6306, "step": 2392 }, { "epoch": 0.29, "learning_rate": 1.6635307773878783e-05, "loss": 0.5789, "step": 2393 }, { "epoch": 0.29, "learning_rate": 1.663236044459995e-05, "loss": 0.5773, "step": 2394 }, { "epoch": 0.29, "learning_rate": 1.6629412086363303e-05, "loss": 0.5967, "step": 2395 }, { "epoch": 0.29, "learning_rate": 1.6626462699626257e-05, "loss": 0.5678, "step": 2396 }, { "epoch": 0.29, "learning_rate": 1.6623512284846386e-05, "loss": 0.6059, "step": 2397 }, { "epoch": 0.29, "learning_rate": 1.662056084248142e-05, "loss": 0.5912, "step": 2398 }, { "epoch": 0.29, "learning_rate": 1.6617608372989258e-05, "loss": 0.6161, "step": 2399 }, { "epoch": 0.29, "learning_rate": 1.6614654876827944e-05, "loss": 0.6057, "step": 2400 }, { "epoch": 0.29, "learning_rate": 1.6611700354455694e-05, "loss": 0.5133, "step": 2401 }, { "epoch": 0.29, "learning_rate": 1.6608744806330882e-05, "loss": 0.6196, "step": 2402 }, { "epoch": 0.29, "learning_rate": 1.660578823291203e-05, "loss": 0.6121, "step": 2403 }, { "epoch": 0.29, "learning_rate": 1.660283063465783e-05, "loss": 0.6676, "step": 2404 }, { "epoch": 0.29, "learning_rate": 1.6599872012027133e-05, "loss": 0.6049, "step": 2405 }, { "epoch": 0.29, "learning_rate": 1.6596912365478938e-05, "loss": 0.6183, "step": 2406 }, { "epoch": 0.29, "learning_rate": 1.659395169547242e-05, "loss": 0.5704, "step": 2407 }, { "epoch": 0.29, "learning_rate": 1.65909900024669e-05, "loss": 0.5966, "step": 2408 }, { "epoch": 0.29, "learning_rate": 1.6588027286921855e-05, "loss": 0.59, "step": 2409 }, { "epoch": 0.29, "learning_rate": 1.6585063549296933e-05, "loss": 0.5271, "step": 2410 }, { "epoch": 0.29, "learning_rate": 1.6582098790051936e-05, "loss": 0.5913, "step": 2411 }, { "epoch": 0.29, "learning_rate": 1.6579133009646816e-05, "loss": 0.664, "step": 2412 }, { "epoch": 0.29, "learning_rate": 1.6576166208541696e-05, "loss": 0.6021, "step": 2413 }, { "epoch": 0.29, "learning_rate": 1.6573198387196846e-05, "loss": 0.6452, "step": 2414 }, { "epoch": 0.29, "learning_rate": 1.6570229546072707e-05, "loss": 0.5739, "step": 2415 }, { "epoch": 0.29, "learning_rate": 1.6567259685629865e-05, "loss": 0.5951, "step": 2416 }, { "epoch": 0.29, "learning_rate": 1.656428880632907e-05, "loss": 0.5507, "step": 2417 }, { "epoch": 0.29, "learning_rate": 1.656131690863124e-05, "loss": 0.6043, "step": 2418 }, { "epoch": 0.29, "learning_rate": 1.6558343992997427e-05, "loss": 0.6526, "step": 2419 }, { "epoch": 0.29, "learning_rate": 1.6555370059888864e-05, "loss": 0.6285, "step": 2420 }, { "epoch": 0.29, "learning_rate": 1.655239510976693e-05, "loss": 0.6511, "step": 2421 }, { "epoch": 0.29, "learning_rate": 1.6549419143093165e-05, "loss": 0.6027, "step": 2422 }, { "epoch": 0.29, "learning_rate": 1.654644216032927e-05, "loss": 0.6108, "step": 2423 }, { "epoch": 0.29, "learning_rate": 1.6543464161937094e-05, "loss": 0.5628, "step": 2424 }, { "epoch": 0.29, "learning_rate": 1.6540485148378652e-05, "loss": 0.6159, "step": 2425 }, { "epoch": 0.3, "learning_rate": 1.653750512011612e-05, "loss": 0.6619, "step": 2426 }, { "epoch": 0.3, "learning_rate": 1.6534524077611818e-05, "loss": 0.628, "step": 2427 }, { "epoch": 0.3, "learning_rate": 1.6531542021328233e-05, "loss": 0.6139, "step": 2428 }, { "epoch": 0.3, "learning_rate": 1.652855895172801e-05, "loss": 0.5471, "step": 2429 }, { "epoch": 0.3, "learning_rate": 1.6525574869273944e-05, "loss": 0.6114, "step": 2430 }, { "epoch": 0.3, "learning_rate": 1.6522589774428993e-05, "loss": 0.6381, "step": 2431 }, { "epoch": 0.3, "learning_rate": 1.6519603667656274e-05, "loss": 0.5385, "step": 2432 }, { "epoch": 0.3, "learning_rate": 1.6516616549419055e-05, "loss": 0.6081, "step": 2433 }, { "epoch": 0.3, "learning_rate": 1.6513628420180764e-05, "loss": 0.5735, "step": 2434 }, { "epoch": 0.3, "learning_rate": 1.6510639280404986e-05, "loss": 0.5904, "step": 2435 }, { "epoch": 0.3, "learning_rate": 1.650764913055546e-05, "loss": 0.5685, "step": 2436 }, { "epoch": 0.3, "learning_rate": 1.6504657971096085e-05, "loss": 0.6362, "step": 2437 }, { "epoch": 0.3, "learning_rate": 1.6501665802490916e-05, "loss": 0.5892, "step": 2438 }, { "epoch": 0.3, "learning_rate": 1.6498672625204165e-05, "loss": 0.6314, "step": 2439 }, { "epoch": 0.3, "learning_rate": 1.64956784397002e-05, "loss": 0.5417, "step": 2440 }, { "epoch": 0.3, "learning_rate": 1.649268324644354e-05, "loss": 0.5953, "step": 2441 }, { "epoch": 0.3, "learning_rate": 1.648968704589887e-05, "loss": 0.6405, "step": 2442 }, { "epoch": 0.3, "learning_rate": 1.6486689838531026e-05, "loss": 0.6027, "step": 2443 }, { "epoch": 0.3, "learning_rate": 1.6483691624804995e-05, "loss": 0.6024, "step": 2444 }, { "epoch": 0.3, "learning_rate": 1.648069240518594e-05, "loss": 0.591, "step": 2445 }, { "epoch": 0.3, "learning_rate": 1.6477692180139152e-05, "loss": 0.5723, "step": 2446 }, { "epoch": 0.3, "learning_rate": 1.64746909501301e-05, "loss": 0.6144, "step": 2447 }, { "epoch": 0.3, "learning_rate": 1.6471688715624395e-05, "loss": 0.5702, "step": 2448 }, { "epoch": 0.3, "learning_rate": 1.6468685477087817e-05, "loss": 0.6159, "step": 2449 }, { "epoch": 0.3, "learning_rate": 1.6465681234986284e-05, "loss": 0.5864, "step": 2450 }, { "epoch": 0.3, "learning_rate": 1.6462675989785892e-05, "loss": 0.5783, "step": 2451 }, { "epoch": 0.3, "learning_rate": 1.6459669741952872e-05, "loss": 0.664, "step": 2452 }, { "epoch": 0.3, "learning_rate": 1.6456662491953626e-05, "loss": 0.6028, "step": 2453 }, { "epoch": 0.3, "learning_rate": 1.64536542402547e-05, "loss": 0.6036, "step": 2454 }, { "epoch": 0.3, "learning_rate": 1.64506449873228e-05, "loss": 0.6331, "step": 2455 }, { "epoch": 0.3, "learning_rate": 1.6447634733624795e-05, "loss": 0.5542, "step": 2456 }, { "epoch": 0.3, "learning_rate": 1.6444623479627688e-05, "loss": 0.6337, "step": 2457 }, { "epoch": 0.3, "learning_rate": 1.6441611225798667e-05, "loss": 0.6239, "step": 2458 }, { "epoch": 0.3, "learning_rate": 1.6438597972605047e-05, "loss": 0.5464, "step": 2459 }, { "epoch": 0.3, "learning_rate": 1.6435583720514315e-05, "loss": 0.5639, "step": 2460 }, { "epoch": 0.3, "learning_rate": 1.6432568469994108e-05, "loss": 0.6138, "step": 2461 }, { "epoch": 0.3, "learning_rate": 1.6429552221512216e-05, "loss": 0.63, "step": 2462 }, { "epoch": 0.3, "learning_rate": 1.6426534975536593e-05, "loss": 0.5924, "step": 2463 }, { "epoch": 0.3, "learning_rate": 1.6423516732535333e-05, "loss": 0.6428, "step": 2464 }, { "epoch": 0.3, "learning_rate": 1.6420497492976694e-05, "loss": 0.6159, "step": 2465 }, { "epoch": 0.3, "learning_rate": 1.6417477257329087e-05, "loss": 0.6332, "step": 2466 }, { "epoch": 0.3, "learning_rate": 1.641445602606108e-05, "loss": 0.6435, "step": 2467 }, { "epoch": 0.3, "learning_rate": 1.641143379964139e-05, "loss": 0.6318, "step": 2468 }, { "epoch": 0.3, "learning_rate": 1.6408410578538893e-05, "loss": 0.5762, "step": 2469 }, { "epoch": 0.3, "learning_rate": 1.6405386363222614e-05, "loss": 0.605, "step": 2470 }, { "epoch": 0.3, "learning_rate": 1.6402361154161742e-05, "loss": 0.6568, "step": 2471 }, { "epoch": 0.3, "learning_rate": 1.639933495182561e-05, "loss": 0.6102, "step": 2472 }, { "epoch": 0.3, "learning_rate": 1.639630775668371e-05, "loss": 0.5714, "step": 2473 }, { "epoch": 0.3, "learning_rate": 1.6393279569205687e-05, "loss": 0.641, "step": 2474 }, { "epoch": 0.3, "learning_rate": 1.639025038986134e-05, "loss": 0.6218, "step": 2475 }, { "epoch": 0.3, "learning_rate": 1.6387220219120626e-05, "loss": 0.5789, "step": 2476 }, { "epoch": 0.3, "learning_rate": 1.6384189057453643e-05, "loss": 0.6443, "step": 2477 }, { "epoch": 0.3, "learning_rate": 1.638115690533066e-05, "loss": 0.6485, "step": 2478 }, { "epoch": 0.3, "learning_rate": 1.6378123763222084e-05, "loss": 0.6738, "step": 2479 }, { "epoch": 0.3, "learning_rate": 1.6375089631598494e-05, "loss": 0.6107, "step": 2480 }, { "epoch": 0.3, "learning_rate": 1.63720545109306e-05, "loss": 0.6556, "step": 2481 }, { "epoch": 0.3, "learning_rate": 1.636901840168928e-05, "loss": 0.5786, "step": 2482 }, { "epoch": 0.3, "learning_rate": 1.6365981304345565e-05, "loss": 0.6302, "step": 2483 }, { "epoch": 0.3, "learning_rate": 1.6362943219370633e-05, "loss": 0.5856, "step": 2484 }, { "epoch": 0.3, "learning_rate": 1.635990414723582e-05, "loss": 0.5897, "step": 2485 }, { "epoch": 0.3, "learning_rate": 1.6356864088412612e-05, "loss": 0.6449, "step": 2486 }, { "epoch": 0.3, "learning_rate": 1.6353823043372656e-05, "loss": 0.5305, "step": 2487 }, { "epoch": 0.3, "learning_rate": 1.635078101258774e-05, "loss": 0.6234, "step": 2488 }, { "epoch": 0.3, "learning_rate": 1.634773799652981e-05, "loss": 0.6282, "step": 2489 }, { "epoch": 0.3, "learning_rate": 1.634469399567097e-05, "loss": 0.6295, "step": 2490 }, { "epoch": 0.3, "learning_rate": 1.634164901048347e-05, "loss": 0.5726, "step": 2491 }, { "epoch": 0.3, "learning_rate": 1.633860304143972e-05, "loss": 0.5458, "step": 2492 }, { "epoch": 0.3, "learning_rate": 1.6335556089012266e-05, "loss": 0.5743, "step": 2493 }, { "epoch": 0.3, "learning_rate": 1.633250815367383e-05, "loss": 0.6001, "step": 2494 }, { "epoch": 0.3, "learning_rate": 1.632945923589727e-05, "loss": 0.5949, "step": 2495 }, { "epoch": 0.3, "learning_rate": 1.6326409336155604e-05, "loss": 0.5817, "step": 2496 }, { "epoch": 0.3, "learning_rate": 1.6323358454921994e-05, "loss": 0.5897, "step": 2497 }, { "epoch": 0.3, "learning_rate": 1.6320306592669766e-05, "loss": 0.5757, "step": 2498 }, { "epoch": 0.3, "learning_rate": 1.6317253749872392e-05, "loss": 0.5332, "step": 2499 }, { "epoch": 0.3, "learning_rate": 1.631419992700349e-05, "loss": 0.6395, "step": 2500 }, { "epoch": 0.3, "learning_rate": 1.631114512453684e-05, "loss": 0.5706, "step": 2501 }, { "epoch": 0.3, "learning_rate": 1.6308089342946368e-05, "loss": 0.5707, "step": 2502 }, { "epoch": 0.3, "learning_rate": 1.6305032582706163e-05, "loss": 0.6343, "step": 2503 }, { "epoch": 0.3, "learning_rate": 1.6301974844290444e-05, "loss": 0.5541, "step": 2504 }, { "epoch": 0.3, "learning_rate": 1.6298916128173606e-05, "loss": 0.5942, "step": 2505 }, { "epoch": 0.3, "learning_rate": 1.6295856434830178e-05, "loss": 0.5658, "step": 2506 }, { "epoch": 0.3, "learning_rate": 1.6292795764734847e-05, "loss": 0.6403, "step": 2507 }, { "epoch": 0.3, "learning_rate": 1.6289734118362453e-05, "loss": 0.5544, "step": 2508 }, { "epoch": 0.31, "learning_rate": 1.628667149618799e-05, "loss": 0.5411, "step": 2509 }, { "epoch": 0.31, "learning_rate": 1.6283607898686592e-05, "loss": 0.5798, "step": 2510 }, { "epoch": 0.31, "learning_rate": 1.6280543326333553e-05, "loss": 0.5586, "step": 2511 }, { "epoch": 0.31, "learning_rate": 1.6277477779604325e-05, "loss": 0.5834, "step": 2512 }, { "epoch": 0.31, "learning_rate": 1.627441125897449e-05, "loss": 0.5565, "step": 2513 }, { "epoch": 0.31, "learning_rate": 1.6271343764919807e-05, "loss": 0.595, "step": 2514 }, { "epoch": 0.31, "learning_rate": 1.6268275297916165e-05, "loss": 0.6351, "step": 2515 }, { "epoch": 0.31, "learning_rate": 1.6265205858439616e-05, "loss": 0.6834, "step": 2516 }, { "epoch": 0.31, "learning_rate": 1.626213544696636e-05, "loss": 0.5722, "step": 2517 }, { "epoch": 0.31, "learning_rate": 1.6259064063972742e-05, "loss": 0.5815, "step": 2518 }, { "epoch": 0.31, "learning_rate": 1.6255991709935268e-05, "loss": 0.609, "step": 2519 }, { "epoch": 0.31, "learning_rate": 1.6252918385330586e-05, "loss": 0.6199, "step": 2520 }, { "epoch": 0.31, "learning_rate": 1.6249844090635495e-05, "loss": 0.6022, "step": 2521 }, { "epoch": 0.31, "learning_rate": 1.624676882632695e-05, "loss": 0.6978, "step": 2522 }, { "epoch": 0.31, "learning_rate": 1.624369259288206e-05, "loss": 0.6147, "step": 2523 }, { "epoch": 0.31, "learning_rate": 1.624061539077807e-05, "loss": 0.542, "step": 2524 }, { "epoch": 0.31, "learning_rate": 1.6237537220492386e-05, "loss": 0.6682, "step": 2525 }, { "epoch": 0.31, "learning_rate": 1.623445808250256e-05, "loss": 0.6176, "step": 2526 }, { "epoch": 0.31, "learning_rate": 1.6231377977286298e-05, "loss": 0.6308, "step": 2527 }, { "epoch": 0.31, "learning_rate": 1.6228296905321453e-05, "loss": 0.6952, "step": 2528 }, { "epoch": 0.31, "learning_rate": 1.622521486708603e-05, "loss": 0.62, "step": 2529 }, { "epoch": 0.31, "learning_rate": 1.6222131863058177e-05, "loss": 0.565, "step": 2530 }, { "epoch": 0.31, "learning_rate": 1.6219047893716203e-05, "loss": 0.6126, "step": 2531 }, { "epoch": 0.31, "learning_rate": 1.621596295953856e-05, "loss": 0.5647, "step": 2532 }, { "epoch": 0.31, "learning_rate": 1.6212877061003854e-05, "loss": 0.6362, "step": 2533 }, { "epoch": 0.31, "learning_rate": 1.620979019859083e-05, "loss": 0.6341, "step": 2534 }, { "epoch": 0.31, "learning_rate": 1.6206702372778398e-05, "loss": 0.5881, "step": 2535 }, { "epoch": 0.31, "learning_rate": 1.6203613584045603e-05, "loss": 0.5738, "step": 2536 }, { "epoch": 0.31, "learning_rate": 1.6200523832871652e-05, "loss": 0.6263, "step": 2537 }, { "epoch": 0.31, "learning_rate": 1.6197433119735892e-05, "loss": 0.5914, "step": 2538 }, { "epoch": 0.31, "learning_rate": 1.619434144511782e-05, "loss": 0.5762, "step": 2539 }, { "epoch": 0.31, "learning_rate": 1.6191248809497092e-05, "loss": 0.6323, "step": 2540 }, { "epoch": 0.31, "learning_rate": 1.6188155213353497e-05, "loss": 0.5832, "step": 2541 }, { "epoch": 0.31, "learning_rate": 1.6185060657166988e-05, "loss": 0.6258, "step": 2542 }, { "epoch": 0.31, "learning_rate": 1.6181965141417656e-05, "loss": 0.5616, "step": 2543 }, { "epoch": 0.31, "learning_rate": 1.617886866658575e-05, "loss": 0.5984, "step": 2544 }, { "epoch": 0.31, "learning_rate": 1.617577123315166e-05, "loss": 0.5572, "step": 2545 }, { "epoch": 0.31, "learning_rate": 1.617267284159593e-05, "loss": 0.6491, "step": 2546 }, { "epoch": 0.31, "learning_rate": 1.616957349239925e-05, "loss": 0.5841, "step": 2547 }, { "epoch": 0.31, "learning_rate": 1.616647318604246e-05, "loss": 0.6274, "step": 2548 }, { "epoch": 0.31, "learning_rate": 1.616337192300655e-05, "loss": 0.6461, "step": 2549 }, { "epoch": 0.31, "learning_rate": 1.6160269703772652e-05, "loss": 0.5506, "step": 2550 }, { "epoch": 0.31, "learning_rate": 1.615716652882205e-05, "loss": 0.5421, "step": 2551 }, { "epoch": 0.31, "learning_rate": 1.615406239863618e-05, "loss": 0.6368, "step": 2552 }, { "epoch": 0.31, "learning_rate": 1.6150957313696623e-05, "loss": 0.6061, "step": 2553 }, { "epoch": 0.31, "learning_rate": 1.614785127448511e-05, "loss": 0.6435, "step": 2554 }, { "epoch": 0.31, "learning_rate": 1.6144744281483514e-05, "loss": 0.6241, "step": 2555 }, { "epoch": 0.31, "learning_rate": 1.6141636335173858e-05, "loss": 0.6095, "step": 2556 }, { "epoch": 0.31, "learning_rate": 1.613852743603832e-05, "loss": 0.6221, "step": 2557 }, { "epoch": 0.31, "learning_rate": 1.6135417584559224e-05, "loss": 0.586, "step": 2558 }, { "epoch": 0.31, "learning_rate": 1.613230678121903e-05, "loss": 0.6393, "step": 2559 }, { "epoch": 0.31, "learning_rate": 1.612919502650036e-05, "loss": 0.5473, "step": 2560 }, { "epoch": 0.31, "learning_rate": 1.6126082320885973e-05, "loss": 0.5952, "step": 2561 }, { "epoch": 0.31, "learning_rate": 1.612296866485879e-05, "loss": 0.654, "step": 2562 }, { "epoch": 0.31, "learning_rate": 1.6119854058901857e-05, "loss": 0.5629, "step": 2563 }, { "epoch": 0.31, "learning_rate": 1.611673850349839e-05, "loss": 0.6145, "step": 2564 }, { "epoch": 0.31, "learning_rate": 1.611362199913174e-05, "loss": 0.6109, "step": 2565 }, { "epoch": 0.31, "learning_rate": 1.611050454628541e-05, "loss": 0.6437, "step": 2566 }, { "epoch": 0.31, "learning_rate": 1.6107386145443044e-05, "loss": 0.5542, "step": 2567 }, { "epoch": 0.31, "learning_rate": 1.610426679708844e-05, "loss": 0.5838, "step": 2568 }, { "epoch": 0.31, "learning_rate": 1.6101146501705533e-05, "loss": 0.6283, "step": 2569 }, { "epoch": 0.31, "learning_rate": 1.609802525977842e-05, "loss": 0.5855, "step": 2570 }, { "epoch": 0.31, "learning_rate": 1.6094903071791333e-05, "loss": 0.5604, "step": 2571 }, { "epoch": 0.31, "learning_rate": 1.6091779938228658e-05, "loss": 0.5839, "step": 2572 }, { "epoch": 0.31, "learning_rate": 1.608865585957492e-05, "loss": 0.5457, "step": 2573 }, { "epoch": 0.31, "learning_rate": 1.60855308363148e-05, "loss": 0.5202, "step": 2574 }, { "epoch": 0.31, "learning_rate": 1.6082404868933114e-05, "loss": 0.6361, "step": 2575 }, { "epoch": 0.31, "learning_rate": 1.6079277957914836e-05, "loss": 0.5609, "step": 2576 }, { "epoch": 0.31, "learning_rate": 1.607615010374508e-05, "loss": 0.6161, "step": 2577 }, { "epoch": 0.31, "learning_rate": 1.6073021306909108e-05, "loss": 0.6532, "step": 2578 }, { "epoch": 0.31, "learning_rate": 1.6069891567892323e-05, "loss": 0.5908, "step": 2579 }, { "epoch": 0.31, "learning_rate": 1.6066760887180287e-05, "loss": 0.5674, "step": 2580 }, { "epoch": 0.31, "learning_rate": 1.6063629265258694e-05, "loss": 0.5641, "step": 2581 }, { "epoch": 0.31, "learning_rate": 1.606049670261339e-05, "loss": 0.5953, "step": 2582 }, { "epoch": 0.31, "learning_rate": 1.6057363199730374e-05, "loss": 0.6969, "step": 2583 }, { "epoch": 0.31, "learning_rate": 1.6054228757095778e-05, "loss": 0.6567, "step": 2584 }, { "epoch": 0.31, "learning_rate": 1.6051093375195884e-05, "loss": 0.6829, "step": 2585 }, { "epoch": 0.31, "learning_rate": 1.604795705451713e-05, "loss": 0.6031, "step": 2586 }, { "epoch": 0.31, "learning_rate": 1.604481979554608e-05, "loss": 0.5342, "step": 2587 }, { "epoch": 0.31, "learning_rate": 1.6041681598769465e-05, "loss": 0.5776, "step": 2588 }, { "epoch": 0.31, "learning_rate": 1.6038542464674144e-05, "loss": 0.5843, "step": 2589 }, { "epoch": 0.31, "learning_rate": 1.6035402393747133e-05, "loss": 0.5956, "step": 2590 }, { "epoch": 0.32, "learning_rate": 1.6032261386475586e-05, "loss": 0.5421, "step": 2591 }, { "epoch": 0.32, "learning_rate": 1.6029119443346807e-05, "loss": 0.5665, "step": 2592 }, { "epoch": 0.32, "learning_rate": 1.6025976564848244e-05, "loss": 0.5261, "step": 2593 }, { "epoch": 0.32, "learning_rate": 1.6022832751467484e-05, "loss": 0.5966, "step": 2594 }, { "epoch": 0.32, "learning_rate": 1.6019688003692273e-05, "loss": 0.5653, "step": 2595 }, { "epoch": 0.32, "learning_rate": 1.601654232201049e-05, "loss": 0.5863, "step": 2596 }, { "epoch": 0.32, "learning_rate": 1.601339570691016e-05, "loss": 0.593, "step": 2597 }, { "epoch": 0.32, "learning_rate": 1.601024815887946e-05, "loss": 0.6534, "step": 2598 }, { "epoch": 0.32, "learning_rate": 1.60070996784067e-05, "loss": 0.5944, "step": 2599 }, { "epoch": 0.32, "learning_rate": 1.6003950265980347e-05, "loss": 0.6335, "step": 2600 }, { "epoch": 0.32, "learning_rate": 1.6000799922089006e-05, "loss": 0.6347, "step": 2601 }, { "epoch": 0.32, "learning_rate": 1.5997648647221428e-05, "loss": 0.5566, "step": 2602 }, { "epoch": 0.32, "learning_rate": 1.599449644186651e-05, "loss": 0.6179, "step": 2603 }, { "epoch": 0.32, "learning_rate": 1.599134330651329e-05, "loss": 0.5846, "step": 2604 }, { "epoch": 0.32, "learning_rate": 1.5988189241650945e-05, "loss": 0.5542, "step": 2605 }, { "epoch": 0.32, "learning_rate": 1.5985034247768813e-05, "loss": 0.5783, "step": 2606 }, { "epoch": 0.32, "learning_rate": 1.5981878325356363e-05, "loss": 0.6263, "step": 2607 }, { "epoch": 0.32, "learning_rate": 1.5978721474903212e-05, "loss": 0.6005, "step": 2608 }, { "epoch": 0.32, "learning_rate": 1.5975563696899117e-05, "loss": 0.5894, "step": 2609 }, { "epoch": 0.32, "learning_rate": 1.597240499183398e-05, "loss": 0.6257, "step": 2610 }, { "epoch": 0.32, "learning_rate": 1.5969245360197856e-05, "loss": 0.6391, "step": 2611 }, { "epoch": 0.32, "learning_rate": 1.596608480248093e-05, "loss": 0.6164, "step": 2612 }, { "epoch": 0.32, "learning_rate": 1.5962923319173547e-05, "loss": 0.5436, "step": 2613 }, { "epoch": 0.32, "learning_rate": 1.5959760910766176e-05, "loss": 0.6113, "step": 2614 }, { "epoch": 0.32, "learning_rate": 1.595659757774944e-05, "loss": 0.6357, "step": 2615 }, { "epoch": 0.32, "learning_rate": 1.5953433320614113e-05, "loss": 0.6262, "step": 2616 }, { "epoch": 0.32, "learning_rate": 1.5950268139851094e-05, "loss": 0.5945, "step": 2617 }, { "epoch": 0.32, "learning_rate": 1.5947102035951443e-05, "loss": 0.587, "step": 2618 }, { "epoch": 0.32, "learning_rate": 1.5943935009406356e-05, "loss": 0.5942, "step": 2619 }, { "epoch": 0.32, "learning_rate": 1.5940767060707164e-05, "loss": 0.633, "step": 2620 }, { "epoch": 0.32, "learning_rate": 1.5937598190345358e-05, "loss": 0.6412, "step": 2621 }, { "epoch": 0.32, "learning_rate": 1.593442839881256e-05, "loss": 0.5642, "step": 2622 }, { "epoch": 0.32, "learning_rate": 1.5931257686600533e-05, "loss": 0.5763, "step": 2623 }, { "epoch": 0.32, "learning_rate": 1.5928086054201194e-05, "loss": 0.653, "step": 2624 }, { "epoch": 0.32, "learning_rate": 1.5924913502106596e-05, "loss": 0.6014, "step": 2625 }, { "epoch": 0.32, "learning_rate": 1.592174003080893e-05, "loss": 0.5445, "step": 2626 }, { "epoch": 0.32, "learning_rate": 1.591856564080054e-05, "loss": 0.598, "step": 2627 }, { "epoch": 0.32, "learning_rate": 1.5915390332573907e-05, "loss": 0.542, "step": 2628 }, { "epoch": 0.32, "learning_rate": 1.591221410662165e-05, "loss": 0.6173, "step": 2629 }, { "epoch": 0.32, "learning_rate": 1.590903696343654e-05, "loss": 0.6181, "step": 2630 }, { "epoch": 0.32, "learning_rate": 1.5905858903511483e-05, "loss": 0.5763, "step": 2631 }, { "epoch": 0.32, "learning_rate": 1.590267992733953e-05, "loss": 0.5783, "step": 2632 }, { "epoch": 0.32, "learning_rate": 1.5899500035413874e-05, "loss": 0.6199, "step": 2633 }, { "epoch": 0.32, "learning_rate": 1.589631922822785e-05, "loss": 0.5711, "step": 2634 }, { "epoch": 0.32, "learning_rate": 1.5893137506274937e-05, "loss": 0.5892, "step": 2635 }, { "epoch": 0.32, "learning_rate": 1.588995487004875e-05, "loss": 0.6002, "step": 2636 }, { "epoch": 0.32, "learning_rate": 1.588677132004305e-05, "loss": 0.5735, "step": 2637 }, { "epoch": 0.32, "learning_rate": 1.5883586856751746e-05, "loss": 0.5834, "step": 2638 }, { "epoch": 0.32, "learning_rate": 1.5880401480668874e-05, "loss": 0.573, "step": 2639 }, { "epoch": 0.32, "learning_rate": 1.5877215192288617e-05, "loss": 0.5537, "step": 2640 }, { "epoch": 0.32, "learning_rate": 1.587402799210531e-05, "loss": 0.569, "step": 2641 }, { "epoch": 0.32, "learning_rate": 1.5870839880613423e-05, "loss": 0.624, "step": 2642 }, { "epoch": 0.32, "learning_rate": 1.5867650858307563e-05, "loss": 0.5937, "step": 2643 }, { "epoch": 0.32, "learning_rate": 1.5864460925682478e-05, "loss": 0.6348, "step": 2644 }, { "epoch": 0.32, "learning_rate": 1.5861270083233064e-05, "loss": 0.5873, "step": 2645 }, { "epoch": 0.32, "learning_rate": 1.5858078331454353e-05, "loss": 0.6456, "step": 2646 }, { "epoch": 0.32, "learning_rate": 1.585488567084152e-05, "loss": 0.622, "step": 2647 }, { "epoch": 0.32, "learning_rate": 1.5851692101889882e-05, "loss": 0.6276, "step": 2648 }, { "epoch": 0.32, "learning_rate": 1.58484976250949e-05, "loss": 0.5647, "step": 2649 }, { "epoch": 0.32, "learning_rate": 1.5845302240952164e-05, "loss": 0.6425, "step": 2650 }, { "epoch": 0.32, "learning_rate": 1.5842105949957414e-05, "loss": 0.5134, "step": 2651 }, { "epoch": 0.32, "learning_rate": 1.583890875260653e-05, "loss": 0.6087, "step": 2652 }, { "epoch": 0.32, "learning_rate": 1.5835710649395535e-05, "loss": 0.6021, "step": 2653 }, { "epoch": 0.32, "learning_rate": 1.5832511640820586e-05, "loss": 0.5941, "step": 2654 }, { "epoch": 0.32, "learning_rate": 1.5829311727377984e-05, "loss": 0.586, "step": 2655 }, { "epoch": 0.32, "learning_rate": 1.582611090956417e-05, "loss": 0.6487, "step": 2656 }, { "epoch": 0.32, "learning_rate": 1.5822909187875725e-05, "loss": 0.6035, "step": 2657 }, { "epoch": 0.32, "learning_rate": 1.5819706562809368e-05, "loss": 0.5748, "step": 2658 }, { "epoch": 0.32, "learning_rate": 1.581650303486197e-05, "loss": 0.5703, "step": 2659 }, { "epoch": 0.32, "learning_rate": 1.5813298604530522e-05, "loss": 0.5449, "step": 2660 }, { "epoch": 0.32, "learning_rate": 1.5810093272312172e-05, "loss": 0.5964, "step": 2661 }, { "epoch": 0.32, "learning_rate": 1.58068870387042e-05, "loss": 0.6455, "step": 2662 }, { "epoch": 0.32, "learning_rate": 1.5803679904204027e-05, "loss": 0.5426, "step": 2663 }, { "epoch": 0.32, "learning_rate": 1.5800471869309215e-05, "loss": 0.5475, "step": 2664 }, { "epoch": 0.32, "learning_rate": 1.5797262934517466e-05, "loss": 0.5586, "step": 2665 }, { "epoch": 0.32, "learning_rate": 1.5794053100326617e-05, "loss": 0.5852, "step": 2666 }, { "epoch": 0.32, "learning_rate": 1.5790842367234654e-05, "loss": 0.5775, "step": 2667 }, { "epoch": 0.32, "learning_rate": 1.5787630735739692e-05, "loss": 0.6378, "step": 2668 }, { "epoch": 0.32, "learning_rate": 1.578441820633999e-05, "loss": 0.574, "step": 2669 }, { "epoch": 0.32, "learning_rate": 1.578120477953395e-05, "loss": 0.577, "step": 2670 }, { "epoch": 0.32, "learning_rate": 1.5777990455820104e-05, "loss": 0.6068, "step": 2671 }, { "epoch": 0.32, "learning_rate": 1.5774775235697135e-05, "loss": 0.5716, "step": 2672 }, { "epoch": 0.33, "learning_rate": 1.5771559119663856e-05, "loss": 0.6545, "step": 2673 }, { "epoch": 0.33, "learning_rate": 1.576834210821922e-05, "loss": 0.5671, "step": 2674 }, { "epoch": 0.33, "learning_rate": 1.5765124201862322e-05, "loss": 0.5978, "step": 2675 }, { "epoch": 0.33, "learning_rate": 1.5761905401092397e-05, "loss": 0.589, "step": 2676 }, { "epoch": 0.33, "learning_rate": 1.5758685706408813e-05, "loss": 0.5973, "step": 2677 }, { "epoch": 0.33, "learning_rate": 1.5755465118311078e-05, "loss": 0.6049, "step": 2678 }, { "epoch": 0.33, "learning_rate": 1.5752243637298844e-05, "loss": 0.6289, "step": 2679 }, { "epoch": 0.33, "learning_rate": 1.57490212638719e-05, "loss": 0.551, "step": 2680 }, { "epoch": 0.33, "learning_rate": 1.5745797998530166e-05, "loss": 0.6211, "step": 2681 }, { "epoch": 0.33, "learning_rate": 1.5742573841773713e-05, "loss": 0.5994, "step": 2682 }, { "epoch": 0.33, "learning_rate": 1.5739348794102734e-05, "loss": 0.5823, "step": 2683 }, { "epoch": 0.33, "learning_rate": 1.573612285601758e-05, "loss": 0.6502, "step": 2684 }, { "epoch": 0.33, "learning_rate": 1.573289602801872e-05, "loss": 0.596, "step": 2685 }, { "epoch": 0.33, "learning_rate": 1.5729668310606777e-05, "loss": 0.5168, "step": 2686 }, { "epoch": 0.33, "learning_rate": 1.57264397042825e-05, "loss": 0.569, "step": 2687 }, { "epoch": 0.33, "learning_rate": 1.5723210209546793e-05, "loss": 0.5708, "step": 2688 }, { "epoch": 0.33, "learning_rate": 1.5719979826900666e-05, "loss": 0.5147, "step": 2689 }, { "epoch": 0.33, "learning_rate": 1.5716748556845307e-05, "loss": 0.6107, "step": 2690 }, { "epoch": 0.33, "learning_rate": 1.5713516399882013e-05, "loss": 0.6256, "step": 2691 }, { "epoch": 0.33, "learning_rate": 1.5710283356512225e-05, "loss": 0.5801, "step": 2692 }, { "epoch": 0.33, "learning_rate": 1.570704942723753e-05, "loss": 0.6131, "step": 2693 }, { "epoch": 0.33, "learning_rate": 1.5703814612559644e-05, "loss": 0.5517, "step": 2694 }, { "epoch": 0.33, "learning_rate": 1.5700578912980425e-05, "loss": 0.5568, "step": 2695 }, { "epoch": 0.33, "learning_rate": 1.569734232900186e-05, "loss": 0.5678, "step": 2696 }, { "epoch": 0.33, "learning_rate": 1.569410486112608e-05, "loss": 0.6121, "step": 2697 }, { "epoch": 0.33, "learning_rate": 1.5690866509855358e-05, "loss": 0.6126, "step": 2698 }, { "epoch": 0.33, "learning_rate": 1.5687627275692093e-05, "loss": 0.6235, "step": 2699 }, { "epoch": 0.33, "learning_rate": 1.5684387159138835e-05, "loss": 0.5998, "step": 2700 }, { "epoch": 0.33, "learning_rate": 1.5681146160698248e-05, "loss": 0.5486, "step": 2701 }, { "epoch": 0.33, "learning_rate": 1.5677904280873156e-05, "loss": 0.6237, "step": 2702 }, { "epoch": 0.33, "learning_rate": 1.5674661520166515e-05, "loss": 0.5545, "step": 2703 }, { "epoch": 0.33, "learning_rate": 1.56714178790814e-05, "loss": 0.5761, "step": 2704 }, { "epoch": 0.33, "learning_rate": 1.566817335812105e-05, "loss": 0.6603, "step": 2705 }, { "epoch": 0.33, "learning_rate": 1.5664927957788817e-05, "loss": 0.5324, "step": 2706 }, { "epoch": 0.33, "learning_rate": 1.5661681678588205e-05, "loss": 0.5506, "step": 2707 }, { "epoch": 0.33, "learning_rate": 1.5658434521022846e-05, "loss": 0.578, "step": 2708 }, { "epoch": 0.33, "learning_rate": 1.5655186485596508e-05, "loss": 0.5752, "step": 2709 }, { "epoch": 0.33, "learning_rate": 1.5651937572813105e-05, "loss": 0.5696, "step": 2710 }, { "epoch": 0.33, "learning_rate": 1.564868778317667e-05, "loss": 0.6692, "step": 2711 }, { "epoch": 0.33, "learning_rate": 1.564543711719139e-05, "loss": 0.581, "step": 2712 }, { "epoch": 0.33, "learning_rate": 1.5642185575361576e-05, "loss": 0.5997, "step": 2713 }, { "epoch": 0.33, "learning_rate": 1.563893315819168e-05, "loss": 0.5902, "step": 2714 }, { "epoch": 0.33, "learning_rate": 1.563567986618629e-05, "loss": 0.5718, "step": 2715 }, { "epoch": 0.33, "learning_rate": 1.5632425699850125e-05, "loss": 0.5893, "step": 2716 }, { "epoch": 0.33, "learning_rate": 1.5629170659688044e-05, "loss": 0.6116, "step": 2717 }, { "epoch": 0.33, "learning_rate": 1.5625914746205043e-05, "loss": 0.5821, "step": 2718 }, { "epoch": 0.33, "learning_rate": 1.5622657959906248e-05, "loss": 0.6245, "step": 2719 }, { "epoch": 0.33, "learning_rate": 1.5619400301296922e-05, "loss": 0.6412, "step": 2720 }, { "epoch": 0.33, "learning_rate": 1.561614177088247e-05, "loss": 0.5993, "step": 2721 }, { "epoch": 0.33, "learning_rate": 1.5612882369168426e-05, "loss": 0.5619, "step": 2722 }, { "epoch": 0.33, "learning_rate": 1.5609622096660456e-05, "loss": 0.6445, "step": 2723 }, { "epoch": 0.33, "learning_rate": 1.560636095386437e-05, "loss": 0.5882, "step": 2724 }, { "epoch": 0.33, "learning_rate": 1.5603098941286102e-05, "loss": 0.6179, "step": 2725 }, { "epoch": 0.33, "learning_rate": 1.5599836059431732e-05, "loss": 0.6207, "step": 2726 }, { "epoch": 0.33, "learning_rate": 1.5596572308807473e-05, "loss": 0.5916, "step": 2727 }, { "epoch": 0.33, "learning_rate": 1.5593307689919663e-05, "loss": 0.5894, "step": 2728 }, { "epoch": 0.33, "learning_rate": 1.5590042203274788e-05, "loss": 0.5937, "step": 2729 }, { "epoch": 0.33, "learning_rate": 1.5586775849379455e-05, "loss": 0.55, "step": 2730 }, { "epoch": 0.33, "learning_rate": 1.5583508628740416e-05, "loss": 0.6394, "step": 2731 }, { "epoch": 0.33, "learning_rate": 1.558024054186456e-05, "loss": 0.6202, "step": 2732 }, { "epoch": 0.33, "learning_rate": 1.5576971589258894e-05, "loss": 0.5247, "step": 2733 }, { "epoch": 0.33, "learning_rate": 1.5573701771430584e-05, "loss": 0.5698, "step": 2734 }, { "epoch": 0.33, "learning_rate": 1.55704310888869e-05, "loss": 0.5532, "step": 2735 }, { "epoch": 0.33, "learning_rate": 1.556715954213527e-05, "loss": 0.5975, "step": 2736 }, { "epoch": 0.33, "learning_rate": 1.5563887131683252e-05, "loss": 0.6805, "step": 2737 }, { "epoch": 0.33, "learning_rate": 1.5560613858038528e-05, "loss": 0.5988, "step": 2738 }, { "epoch": 0.33, "learning_rate": 1.5557339721708925e-05, "loss": 0.5804, "step": 2739 }, { "epoch": 0.33, "learning_rate": 1.5554064723202394e-05, "loss": 0.5689, "step": 2740 }, { "epoch": 0.33, "learning_rate": 1.5550788863027034e-05, "loss": 0.5932, "step": 2741 }, { "epoch": 0.33, "learning_rate": 1.5547512141691057e-05, "loss": 0.6935, "step": 2742 }, { "epoch": 0.33, "learning_rate": 1.554423455970283e-05, "loss": 0.5667, "step": 2743 }, { "epoch": 0.33, "learning_rate": 1.5540956117570835e-05, "loss": 0.5525, "step": 2744 }, { "epoch": 0.33, "learning_rate": 1.5537676815803703e-05, "loss": 0.5307, "step": 2745 }, { "epoch": 0.33, "learning_rate": 1.553439665491019e-05, "loss": 0.5944, "step": 2746 }, { "epoch": 0.33, "learning_rate": 1.5531115635399182e-05, "loss": 0.6108, "step": 2747 }, { "epoch": 0.33, "learning_rate": 1.5527833757779714e-05, "loss": 0.5771, "step": 2748 }, { "epoch": 0.33, "learning_rate": 1.552455102256093e-05, "loss": 0.6339, "step": 2749 }, { "epoch": 0.33, "learning_rate": 1.5521267430252123e-05, "loss": 0.5144, "step": 2750 }, { "epoch": 0.33, "learning_rate": 1.551798298136273e-05, "loss": 0.6245, "step": 2751 }, { "epoch": 0.33, "learning_rate": 1.5514697676402288e-05, "loss": 0.5204, "step": 2752 }, { "epoch": 0.33, "learning_rate": 1.55114115158805e-05, "loss": 0.5442, "step": 2753 }, { "epoch": 0.33, "learning_rate": 1.5508124500307177e-05, "loss": 0.574, "step": 2754 }, { "epoch": 0.34, "learning_rate": 1.550483663019228e-05, "loss": 0.5641, "step": 2755 }, { "epoch": 0.34, "learning_rate": 1.5501547906045894e-05, "loss": 0.6257, "step": 2756 }, { "epoch": 0.34, "learning_rate": 1.549825832837824e-05, "loss": 0.5813, "step": 2757 }, { "epoch": 0.34, "learning_rate": 1.5494967897699666e-05, "loss": 0.5486, "step": 2758 }, { "epoch": 0.34, "learning_rate": 1.549167661452066e-05, "loss": 0.6408, "step": 2759 }, { "epoch": 0.34, "learning_rate": 1.5488384479351834e-05, "loss": 0.5743, "step": 2760 }, { "epoch": 0.34, "learning_rate": 1.5485091492703938e-05, "loss": 0.5487, "step": 2761 }, { "epoch": 0.34, "learning_rate": 1.5481797655087854e-05, "loss": 0.6234, "step": 2762 }, { "epoch": 0.34, "learning_rate": 1.5478502967014595e-05, "loss": 0.6145, "step": 2763 }, { "epoch": 0.34, "learning_rate": 1.5475207428995306e-05, "loss": 0.5506, "step": 2764 }, { "epoch": 0.34, "learning_rate": 1.547191104154126e-05, "loss": 0.5817, "step": 2765 }, { "epoch": 0.34, "learning_rate": 1.546861380516387e-05, "loss": 0.5848, "step": 2766 }, { "epoch": 0.34, "learning_rate": 1.546531572037467e-05, "loss": 0.6751, "step": 2767 }, { "epoch": 0.34, "learning_rate": 1.5462016787685336e-05, "loss": 0.5505, "step": 2768 }, { "epoch": 0.34, "learning_rate": 1.5458717007607672e-05, "loss": 0.5823, "step": 2769 }, { "epoch": 0.34, "learning_rate": 1.545541638065361e-05, "loss": 0.6397, "step": 2770 }, { "epoch": 0.34, "learning_rate": 1.5452114907335217e-05, "loss": 0.6309, "step": 2771 }, { "epoch": 0.34, "learning_rate": 1.5448812588164686e-05, "loss": 0.5505, "step": 2772 }, { "epoch": 0.34, "learning_rate": 1.544550942365435e-05, "loss": 0.6799, "step": 2773 }, { "epoch": 0.34, "learning_rate": 1.544220541431667e-05, "loss": 0.5556, "step": 2774 }, { "epoch": 0.34, "learning_rate": 1.5438900560664232e-05, "loss": 0.5682, "step": 2775 }, { "epoch": 0.34, "learning_rate": 1.543559486320976e-05, "loss": 0.6077, "step": 2776 }, { "epoch": 0.34, "learning_rate": 1.5432288322466112e-05, "loss": 0.5907, "step": 2777 }, { "epoch": 0.34, "learning_rate": 1.542898093894626e-05, "loss": 0.5705, "step": 2778 }, { "epoch": 0.34, "learning_rate": 1.5425672713163327e-05, "loss": 0.566, "step": 2779 }, { "epoch": 0.34, "learning_rate": 1.5422363645630554e-05, "loss": 0.5361, "step": 2780 }, { "epoch": 0.34, "learning_rate": 1.5419053736861318e-05, "loss": 0.5689, "step": 2781 }, { "epoch": 0.34, "learning_rate": 1.5415742987369126e-05, "loss": 0.5591, "step": 2782 }, { "epoch": 0.34, "learning_rate": 1.541243139766761e-05, "loss": 0.5676, "step": 2783 }, { "epoch": 0.34, "learning_rate": 1.5409118968270545e-05, "loss": 0.5869, "step": 2784 }, { "epoch": 0.34, "learning_rate": 1.5405805699691815e-05, "loss": 0.6097, "step": 2785 }, { "epoch": 0.34, "learning_rate": 1.5402491592445464e-05, "loss": 0.572, "step": 2786 }, { "epoch": 0.34, "learning_rate": 1.5399176647045633e-05, "loss": 0.5434, "step": 2787 }, { "epoch": 0.34, "learning_rate": 1.5395860864006617e-05, "loss": 0.6068, "step": 2788 }, { "epoch": 0.34, "learning_rate": 1.5392544243842834e-05, "loss": 0.5543, "step": 2789 }, { "epoch": 0.34, "learning_rate": 1.538922678706883e-05, "loss": 0.5691, "step": 2790 }, { "epoch": 0.34, "learning_rate": 1.5385908494199282e-05, "loss": 0.5956, "step": 2791 }, { "epoch": 0.34, "learning_rate": 1.5382589365749e-05, "loss": 0.5674, "step": 2792 }, { "epoch": 0.34, "learning_rate": 1.537926940223291e-05, "loss": 0.5902, "step": 2793 }, { "epoch": 0.34, "learning_rate": 1.5375948604166088e-05, "loss": 0.5538, "step": 2794 }, { "epoch": 0.34, "learning_rate": 1.5372626972063725e-05, "loss": 0.5782, "step": 2795 }, { "epoch": 0.34, "learning_rate": 1.5369304506441148e-05, "loss": 0.6162, "step": 2796 }, { "epoch": 0.34, "learning_rate": 1.5365981207813807e-05, "loss": 0.5468, "step": 2797 }, { "epoch": 0.34, "learning_rate": 1.5362657076697292e-05, "loss": 0.5757, "step": 2798 }, { "epoch": 0.34, "learning_rate": 1.535933211360731e-05, "loss": 0.6188, "step": 2799 }, { "epoch": 0.34, "learning_rate": 1.5356006319059703e-05, "loss": 0.5722, "step": 2800 }, { "epoch": 0.34, "learning_rate": 1.5352679693570444e-05, "loss": 0.533, "step": 2801 }, { "epoch": 0.34, "learning_rate": 1.534935223765563e-05, "loss": 0.6247, "step": 2802 }, { "epoch": 0.34, "learning_rate": 1.534602395183149e-05, "loss": 0.6193, "step": 2803 }, { "epoch": 0.34, "learning_rate": 1.534269483661438e-05, "loss": 0.6233, "step": 2804 }, { "epoch": 0.34, "learning_rate": 1.5339364892520787e-05, "loss": 0.5923, "step": 2805 }, { "epoch": 0.34, "learning_rate": 1.5336034120067327e-05, "loss": 0.6591, "step": 2806 }, { "epoch": 0.34, "learning_rate": 1.5332702519770738e-05, "loss": 0.5888, "step": 2807 }, { "epoch": 0.34, "learning_rate": 1.53293700921479e-05, "loss": 0.6603, "step": 2808 }, { "epoch": 0.34, "learning_rate": 1.53260368377158e-05, "loss": 0.5964, "step": 2809 }, { "epoch": 0.34, "learning_rate": 1.5322702756991573e-05, "loss": 0.6265, "step": 2810 }, { "epoch": 0.34, "learning_rate": 1.5319367850492478e-05, "loss": 0.5704, "step": 2811 }, { "epoch": 0.34, "learning_rate": 1.5316032118735894e-05, "loss": 0.5767, "step": 2812 }, { "epoch": 0.34, "learning_rate": 1.531269556223933e-05, "loss": 0.6046, "step": 2813 }, { "epoch": 0.34, "learning_rate": 1.530935818152044e-05, "loss": 0.5744, "step": 2814 }, { "epoch": 0.34, "learning_rate": 1.5306019977096975e-05, "loss": 0.5754, "step": 2815 }, { "epoch": 0.34, "learning_rate": 1.5302680949486844e-05, "loss": 0.604, "step": 2816 }, { "epoch": 0.34, "learning_rate": 1.5299341099208063e-05, "loss": 0.5326, "step": 2817 }, { "epoch": 0.34, "learning_rate": 1.5296000426778784e-05, "loss": 0.5528, "step": 2818 }, { "epoch": 0.34, "learning_rate": 1.5292658932717288e-05, "loss": 0.6242, "step": 2819 }, { "epoch": 0.34, "learning_rate": 1.528931661754198e-05, "loss": 0.5452, "step": 2820 }, { "epoch": 0.34, "learning_rate": 1.5285973481771393e-05, "loss": 0.5854, "step": 2821 }, { "epoch": 0.34, "learning_rate": 1.528262952592419e-05, "loss": 0.6198, "step": 2822 }, { "epoch": 0.34, "learning_rate": 1.5279284750519158e-05, "loss": 0.5399, "step": 2823 }, { "epoch": 0.34, "learning_rate": 1.527593915607521e-05, "loss": 0.6192, "step": 2824 }, { "epoch": 0.34, "learning_rate": 1.5272592743111388e-05, "loss": 0.5943, "step": 2825 }, { "epoch": 0.34, "learning_rate": 1.5269245512146866e-05, "loss": 0.5371, "step": 2826 }, { "epoch": 0.34, "learning_rate": 1.5265897463700937e-05, "loss": 0.4974, "step": 2827 }, { "epoch": 0.34, "learning_rate": 1.5262548598293025e-05, "loss": 0.5457, "step": 2828 }, { "epoch": 0.34, "learning_rate": 1.5259198916442678e-05, "loss": 0.5755, "step": 2829 }, { "epoch": 0.34, "learning_rate": 1.5255848418669577e-05, "loss": 0.5899, "step": 2830 }, { "epoch": 0.34, "learning_rate": 1.525249710549352e-05, "loss": 0.563, "step": 2831 }, { "epoch": 0.34, "learning_rate": 1.5249144977434438e-05, "loss": 0.6248, "step": 2832 }, { "epoch": 0.34, "learning_rate": 1.5245792035012383e-05, "loss": 0.6138, "step": 2833 }, { "epoch": 0.34, "learning_rate": 1.5242438278747547e-05, "loss": 0.6373, "step": 2834 }, { "epoch": 0.34, "learning_rate": 1.5239083709160232e-05, "loss": 0.5886, "step": 2835 }, { "epoch": 0.34, "learning_rate": 1.5235728326770871e-05, "loss": 0.5197, "step": 2836 }, { "epoch": 0.34, "learning_rate": 1.5232372132100031e-05, "loss": 0.6096, "step": 2837 }, { "epoch": 0.35, "learning_rate": 1.5229015125668394e-05, "loss": 0.596, "step": 2838 }, { "epoch": 0.35, "learning_rate": 1.5225657307996771e-05, "loss": 0.5986, "step": 2839 }, { "epoch": 0.35, "learning_rate": 1.5222298679606107e-05, "loss": 0.5831, "step": 2840 }, { "epoch": 0.35, "learning_rate": 1.5218939241017461e-05, "loss": 0.602, "step": 2841 }, { "epoch": 0.35, "learning_rate": 1.5215578992752029e-05, "loss": 0.5436, "step": 2842 }, { "epoch": 0.35, "learning_rate": 1.5212217935331119e-05, "loss": 0.5708, "step": 2843 }, { "epoch": 0.35, "learning_rate": 1.5208856069276176e-05, "loss": 0.5124, "step": 2844 }, { "epoch": 0.35, "learning_rate": 1.5205493395108769e-05, "loss": 0.5984, "step": 2845 }, { "epoch": 0.35, "learning_rate": 1.5202129913350584e-05, "loss": 0.5914, "step": 2846 }, { "epoch": 0.35, "learning_rate": 1.5198765624523448e-05, "loss": 0.5646, "step": 2847 }, { "epoch": 0.35, "learning_rate": 1.5195400529149295e-05, "loss": 0.5844, "step": 2848 }, { "epoch": 0.35, "learning_rate": 1.51920346277502e-05, "loss": 0.5477, "step": 2849 }, { "epoch": 0.35, "learning_rate": 1.5188667920848344e-05, "loss": 0.5172, "step": 2850 }, { "epoch": 0.35, "learning_rate": 1.518530040896605e-05, "loss": 0.5745, "step": 2851 }, { "epoch": 0.35, "learning_rate": 1.5181932092625774e-05, "loss": 0.5651, "step": 2852 }, { "epoch": 0.35, "learning_rate": 1.5178562972350062e-05, "loss": 0.6093, "step": 2853 }, { "epoch": 0.35, "learning_rate": 1.5175193048661618e-05, "loss": 0.5867, "step": 2854 }, { "epoch": 0.35, "learning_rate": 1.5171822322083261e-05, "loss": 0.6004, "step": 2855 }, { "epoch": 0.35, "learning_rate": 1.5168450793137925e-05, "loss": 0.5703, "step": 2856 }, { "epoch": 0.35, "learning_rate": 1.5165078462348678e-05, "loss": 0.5572, "step": 2857 }, { "epoch": 0.35, "learning_rate": 1.5161705330238716e-05, "loss": 0.5954, "step": 2858 }, { "epoch": 0.35, "learning_rate": 1.5158331397331348e-05, "loss": 0.5591, "step": 2859 }, { "epoch": 0.35, "learning_rate": 1.515495666415001e-05, "loss": 0.6136, "step": 2860 }, { "epoch": 0.35, "learning_rate": 1.5151581131218269e-05, "loss": 0.5989, "step": 2861 }, { "epoch": 0.35, "learning_rate": 1.5148204799059816e-05, "loss": 0.4889, "step": 2862 }, { "epoch": 0.35, "learning_rate": 1.5144827668198452e-05, "loss": 0.6132, "step": 2863 }, { "epoch": 0.35, "learning_rate": 1.5141449739158119e-05, "loss": 0.6273, "step": 2864 }, { "epoch": 0.35, "learning_rate": 1.5138071012462876e-05, "loss": 0.6284, "step": 2865 }, { "epoch": 0.35, "learning_rate": 1.5134691488636903e-05, "loss": 0.5476, "step": 2866 }, { "epoch": 0.35, "learning_rate": 1.5131311168204508e-05, "loss": 0.5745, "step": 2867 }, { "epoch": 0.35, "learning_rate": 1.5127930051690116e-05, "loss": 0.6486, "step": 2868 }, { "epoch": 0.35, "learning_rate": 1.5124548139618287e-05, "loss": 0.5846, "step": 2869 }, { "epoch": 0.35, "learning_rate": 1.5121165432513691e-05, "loss": 0.5104, "step": 2870 }, { "epoch": 0.35, "learning_rate": 1.5117781930901134e-05, "loss": 0.6374, "step": 2871 }, { "epoch": 0.35, "learning_rate": 1.511439763530554e-05, "loss": 0.6427, "step": 2872 }, { "epoch": 0.35, "learning_rate": 1.5111012546251948e-05, "loss": 0.5525, "step": 2873 }, { "epoch": 0.35, "learning_rate": 1.5107626664265533e-05, "loss": 0.6084, "step": 2874 }, { "epoch": 0.35, "learning_rate": 1.5104239989871588e-05, "loss": 0.5588, "step": 2875 }, { "epoch": 0.35, "learning_rate": 1.5100852523595526e-05, "loss": 0.5983, "step": 2876 }, { "epoch": 0.35, "learning_rate": 1.5097464265962884e-05, "loss": 0.5861, "step": 2877 }, { "epoch": 0.35, "learning_rate": 1.5094075217499331e-05, "loss": 0.6089, "step": 2878 }, { "epoch": 0.35, "learning_rate": 1.509068537873064e-05, "loss": 0.6129, "step": 2879 }, { "epoch": 0.35, "learning_rate": 1.5087294750182726e-05, "loss": 0.6204, "step": 2880 }, { "epoch": 0.35, "learning_rate": 1.5083903332381612e-05, "loss": 0.5466, "step": 2881 }, { "epoch": 0.35, "learning_rate": 1.5080511125853454e-05, "loss": 0.6144, "step": 2882 }, { "epoch": 0.35, "learning_rate": 1.5077118131124521e-05, "loss": 0.6181, "step": 2883 }, { "epoch": 0.35, "learning_rate": 1.5073724348721216e-05, "loss": 0.5953, "step": 2884 }, { "epoch": 0.35, "learning_rate": 1.5070329779170055e-05, "loss": 0.5521, "step": 2885 }, { "epoch": 0.35, "learning_rate": 1.5066934422997672e-05, "loss": 0.6028, "step": 2886 }, { "epoch": 0.35, "learning_rate": 1.5063538280730833e-05, "loss": 0.5424, "step": 2887 }, { "epoch": 0.35, "learning_rate": 1.5060141352896427e-05, "loss": 0.5705, "step": 2888 }, { "epoch": 0.35, "learning_rate": 1.5056743640021456e-05, "loss": 0.6164, "step": 2889 }, { "epoch": 0.35, "learning_rate": 1.5053345142633049e-05, "loss": 0.5625, "step": 2890 }, { "epoch": 0.35, "learning_rate": 1.5049945861258453e-05, "loss": 0.575, "step": 2891 }, { "epoch": 0.35, "learning_rate": 1.5046545796425047e-05, "loss": 0.5695, "step": 2892 }, { "epoch": 0.35, "learning_rate": 1.5043144948660313e-05, "loss": 0.5381, "step": 2893 }, { "epoch": 0.35, "learning_rate": 1.5039743318491875e-05, "loss": 0.5468, "step": 2894 }, { "epoch": 0.35, "learning_rate": 1.5036340906447465e-05, "loss": 0.5564, "step": 2895 }, { "epoch": 0.35, "learning_rate": 1.5032937713054938e-05, "loss": 0.6303, "step": 2896 }, { "epoch": 0.35, "learning_rate": 1.5029533738842275e-05, "loss": 0.577, "step": 2897 }, { "epoch": 0.35, "learning_rate": 1.5026128984337577e-05, "loss": 0.6005, "step": 2898 }, { "epoch": 0.35, "learning_rate": 1.5022723450069058e-05, "loss": 0.6467, "step": 2899 }, { "epoch": 0.35, "learning_rate": 1.5019317136565067e-05, "loss": 0.5384, "step": 2900 }, { "epoch": 0.35, "learning_rate": 1.5015910044354062e-05, "loss": 0.6288, "step": 2901 }, { "epoch": 0.35, "learning_rate": 1.5012502173964629e-05, "loss": 0.5665, "step": 2902 }, { "epoch": 0.35, "learning_rate": 1.5009093525925468e-05, "loss": 0.6135, "step": 2903 }, { "epoch": 0.35, "learning_rate": 1.500568410076541e-05, "loss": 0.5919, "step": 2904 }, { "epoch": 0.35, "learning_rate": 1.500227389901339e-05, "loss": 0.5399, "step": 2905 }, { "epoch": 0.35, "learning_rate": 1.4998862921198485e-05, "loss": 0.5858, "step": 2906 }, { "epoch": 0.35, "learning_rate": 1.4995451167849873e-05, "loss": 0.5591, "step": 2907 }, { "epoch": 0.35, "learning_rate": 1.4992038639496863e-05, "loss": 0.5536, "step": 2908 }, { "epoch": 0.35, "learning_rate": 1.4988625336668882e-05, "loss": 0.585, "step": 2909 }, { "epoch": 0.35, "learning_rate": 1.4985211259895475e-05, "loss": 0.5755, "step": 2910 }, { "epoch": 0.35, "learning_rate": 1.4981796409706313e-05, "loss": 0.5711, "step": 2911 }, { "epoch": 0.35, "learning_rate": 1.4978380786631178e-05, "loss": 0.5937, "step": 2912 }, { "epoch": 0.35, "learning_rate": 1.4974964391199976e-05, "loss": 0.6307, "step": 2913 }, { "epoch": 0.35, "learning_rate": 1.4971547223942739e-05, "loss": 0.5954, "step": 2914 }, { "epoch": 0.35, "learning_rate": 1.4968129285389607e-05, "loss": 0.6648, "step": 2915 }, { "epoch": 0.35, "learning_rate": 1.4964710576070851e-05, "loss": 0.5689, "step": 2916 }, { "epoch": 0.35, "learning_rate": 1.4961291096516854e-05, "loss": 0.5575, "step": 2917 }, { "epoch": 0.35, "learning_rate": 1.4957870847258122e-05, "loss": 0.6221, "step": 2918 }, { "epoch": 0.35, "learning_rate": 1.4954449828825277e-05, "loss": 0.5947, "step": 2919 }, { "epoch": 0.36, "learning_rate": 1.4951028041749064e-05, "loss": 0.6004, "step": 2920 }, { "epoch": 0.36, "learning_rate": 1.4947605486560353e-05, "loss": 0.5811, "step": 2921 }, { "epoch": 0.36, "learning_rate": 1.4944182163790113e-05, "loss": 0.5985, "step": 2922 }, { "epoch": 0.36, "learning_rate": 1.4940758073969452e-05, "loss": 0.5935, "step": 2923 }, { "epoch": 0.36, "learning_rate": 1.4937333217629593e-05, "loss": 0.5224, "step": 2924 }, { "epoch": 0.36, "learning_rate": 1.493390759530187e-05, "loss": 0.6122, "step": 2925 }, { "epoch": 0.36, "learning_rate": 1.4930481207517744e-05, "loss": 0.6099, "step": 2926 }, { "epoch": 0.36, "learning_rate": 1.4927054054808792e-05, "loss": 0.5705, "step": 2927 }, { "epoch": 0.36, "learning_rate": 1.4923626137706708e-05, "loss": 0.5484, "step": 2928 }, { "epoch": 0.36, "learning_rate": 1.4920197456743304e-05, "loss": 0.6053, "step": 2929 }, { "epoch": 0.36, "learning_rate": 1.4916768012450517e-05, "loss": 0.5969, "step": 2930 }, { "epoch": 0.36, "learning_rate": 1.4913337805360398e-05, "loss": 0.5712, "step": 2931 }, { "epoch": 0.36, "learning_rate": 1.4909906836005111e-05, "loss": 0.5789, "step": 2932 }, { "epoch": 0.36, "learning_rate": 1.4906475104916947e-05, "loss": 0.5611, "step": 2933 }, { "epoch": 0.36, "learning_rate": 1.4903042612628312e-05, "loss": 0.5431, "step": 2934 }, { "epoch": 0.36, "learning_rate": 1.4899609359671724e-05, "loss": 0.6119, "step": 2935 }, { "epoch": 0.36, "learning_rate": 1.4896175346579835e-05, "loss": 0.5643, "step": 2936 }, { "epoch": 0.36, "learning_rate": 1.48927405738854e-05, "loss": 0.6061, "step": 2937 }, { "epoch": 0.36, "learning_rate": 1.4889305042121294e-05, "loss": 0.5999, "step": 2938 }, { "epoch": 0.36, "learning_rate": 1.4885868751820516e-05, "loss": 0.5308, "step": 2939 }, { "epoch": 0.36, "learning_rate": 1.4882431703516175e-05, "loss": 0.5783, "step": 2940 }, { "epoch": 0.36, "learning_rate": 1.4878993897741507e-05, "loss": 0.5446, "step": 2941 }, { "epoch": 0.36, "learning_rate": 1.4875555335029856e-05, "loss": 0.5843, "step": 2942 }, { "epoch": 0.36, "learning_rate": 1.4872116015914688e-05, "loss": 0.5764, "step": 2943 }, { "epoch": 0.36, "learning_rate": 1.486867594092959e-05, "loss": 0.5587, "step": 2944 }, { "epoch": 0.36, "learning_rate": 1.4865235110608258e-05, "loss": 0.6337, "step": 2945 }, { "epoch": 0.36, "learning_rate": 1.4861793525484509e-05, "loss": 0.6303, "step": 2946 }, { "epoch": 0.36, "learning_rate": 1.4858351186092277e-05, "loss": 0.5472, "step": 2947 }, { "epoch": 0.36, "learning_rate": 1.485490809296562e-05, "loss": 0.6224, "step": 2948 }, { "epoch": 0.36, "learning_rate": 1.4851464246638698e-05, "loss": 0.6618, "step": 2949 }, { "epoch": 0.36, "learning_rate": 1.4848019647645799e-05, "loss": 0.5516, "step": 2950 }, { "epoch": 0.36, "learning_rate": 1.484457429652133e-05, "loss": 0.6284, "step": 2951 }, { "epoch": 0.36, "learning_rate": 1.4841128193799806e-05, "loss": 0.6214, "step": 2952 }, { "epoch": 0.36, "learning_rate": 1.4837681340015857e-05, "loss": 0.5882, "step": 2953 }, { "epoch": 0.36, "learning_rate": 1.4834233735704247e-05, "loss": 0.6174, "step": 2954 }, { "epoch": 0.36, "learning_rate": 1.4830785381399833e-05, "loss": 0.5684, "step": 2955 }, { "epoch": 0.36, "learning_rate": 1.4827336277637603e-05, "loss": 0.5725, "step": 2956 }, { "epoch": 0.36, "learning_rate": 1.4823886424952663e-05, "loss": 0.5529, "step": 2957 }, { "epoch": 0.36, "learning_rate": 1.4820435823880224e-05, "loss": 0.6792, "step": 2958 }, { "epoch": 0.36, "learning_rate": 1.4816984474955617e-05, "loss": 0.5717, "step": 2959 }, { "epoch": 0.36, "learning_rate": 1.4813532378714303e-05, "loss": 0.5615, "step": 2960 }, { "epoch": 0.36, "learning_rate": 1.4810079535691837e-05, "loss": 0.5597, "step": 2961 }, { "epoch": 0.36, "learning_rate": 1.4806625946423903e-05, "loss": 0.6136, "step": 2962 }, { "epoch": 0.36, "learning_rate": 1.4803171611446296e-05, "loss": 0.6197, "step": 2963 }, { "epoch": 0.36, "learning_rate": 1.4799716531294932e-05, "loss": 0.5323, "step": 2964 }, { "epoch": 0.36, "learning_rate": 1.4796260706505834e-05, "loss": 0.5935, "step": 2965 }, { "epoch": 0.36, "learning_rate": 1.479280413761515e-05, "loss": 0.5994, "step": 2966 }, { "epoch": 0.36, "learning_rate": 1.4789346825159143e-05, "loss": 0.5577, "step": 2967 }, { "epoch": 0.36, "learning_rate": 1.478588876967418e-05, "loss": 0.5586, "step": 2968 }, { "epoch": 0.36, "learning_rate": 1.478242997169675e-05, "loss": 0.5492, "step": 2969 }, { "epoch": 0.36, "learning_rate": 1.4778970431763464e-05, "loss": 0.5638, "step": 2970 }, { "epoch": 0.36, "learning_rate": 1.4775510150411037e-05, "loss": 0.6363, "step": 2971 }, { "epoch": 0.36, "learning_rate": 1.477204912817631e-05, "loss": 0.6329, "step": 2972 }, { "epoch": 0.36, "learning_rate": 1.4768587365596222e-05, "loss": 0.5668, "step": 2973 }, { "epoch": 0.36, "learning_rate": 1.4765124863207854e-05, "loss": 0.5732, "step": 2974 }, { "epoch": 0.36, "learning_rate": 1.4761661621548373e-05, "loss": 0.5113, "step": 2975 }, { "epoch": 0.36, "learning_rate": 1.4758197641155075e-05, "loss": 0.6041, "step": 2976 }, { "epoch": 0.36, "learning_rate": 1.4754732922565371e-05, "loss": 0.5759, "step": 2977 }, { "epoch": 0.36, "learning_rate": 1.4751267466316786e-05, "loss": 0.4883, "step": 2978 }, { "epoch": 0.36, "learning_rate": 1.4747801272946956e-05, "loss": 0.5796, "step": 2979 }, { "epoch": 0.36, "learning_rate": 1.4744334342993634e-05, "loss": 0.6284, "step": 2980 }, { "epoch": 0.36, "learning_rate": 1.4740866676994685e-05, "loss": 0.5362, "step": 2981 }, { "epoch": 0.36, "learning_rate": 1.4737398275488091e-05, "loss": 0.5478, "step": 2982 }, { "epoch": 0.36, "learning_rate": 1.4733929139011944e-05, "loss": 0.5298, "step": 2983 }, { "epoch": 0.36, "learning_rate": 1.4730459268104458e-05, "loss": 0.6345, "step": 2984 }, { "epoch": 0.36, "learning_rate": 1.4726988663303954e-05, "loss": 0.5745, "step": 2985 }, { "epoch": 0.36, "learning_rate": 1.4723517325148866e-05, "loss": 0.6202, "step": 2986 }, { "epoch": 0.36, "learning_rate": 1.4720045254177748e-05, "loss": 0.5505, "step": 2987 }, { "epoch": 0.36, "learning_rate": 1.4716572450929261e-05, "loss": 0.6049, "step": 2988 }, { "epoch": 0.36, "learning_rate": 1.4713098915942184e-05, "loss": 0.5708, "step": 2989 }, { "epoch": 0.36, "learning_rate": 1.4709624649755409e-05, "loss": 0.5752, "step": 2990 }, { "epoch": 0.36, "learning_rate": 1.470614965290794e-05, "loss": 0.5725, "step": 2991 }, { "epoch": 0.36, "learning_rate": 1.4702673925938895e-05, "loss": 0.5267, "step": 2992 }, { "epoch": 0.36, "learning_rate": 1.4699197469387506e-05, "loss": 0.5694, "step": 2993 }, { "epoch": 0.36, "learning_rate": 1.4695720283793116e-05, "loss": 0.6028, "step": 2994 }, { "epoch": 0.36, "learning_rate": 1.4692242369695182e-05, "loss": 0.5303, "step": 2995 }, { "epoch": 0.36, "learning_rate": 1.468876372763328e-05, "loss": 0.6639, "step": 2996 }, { "epoch": 0.36, "learning_rate": 1.4685284358147085e-05, "loss": 0.6075, "step": 2997 }, { "epoch": 0.36, "learning_rate": 1.46818042617764e-05, "loss": 0.6205, "step": 2998 }, { "epoch": 0.36, "learning_rate": 1.4678323439061133e-05, "loss": 0.5549, "step": 2999 }, { "epoch": 0.36, "learning_rate": 1.4674841890541307e-05, "loss": 0.5665, "step": 3000 }, { "epoch": 0.36, "learning_rate": 1.4671359616757048e-05, "loss": 0.5501, "step": 3001 }, { "epoch": 0.37, "learning_rate": 1.4667876618248615e-05, "loss": 0.6241, "step": 3002 }, { "epoch": 0.37, "learning_rate": 1.4664392895556362e-05, "loss": 0.5604, "step": 3003 }, { "epoch": 0.37, "learning_rate": 1.4660908449220762e-05, "loss": 0.6329, "step": 3004 }, { "epoch": 0.37, "learning_rate": 1.4657423279782395e-05, "loss": 0.5701, "step": 3005 }, { "epoch": 0.37, "learning_rate": 1.465393738778196e-05, "loss": 0.6122, "step": 3006 }, { "epoch": 0.37, "learning_rate": 1.4650450773760265e-05, "loss": 0.6005, "step": 3007 }, { "epoch": 0.37, "learning_rate": 1.4646963438258233e-05, "loss": 0.6155, "step": 3008 }, { "epoch": 0.37, "learning_rate": 1.4643475381816892e-05, "loss": 0.5943, "step": 3009 }, { "epoch": 0.37, "learning_rate": 1.4639986604977393e-05, "loss": 0.5882, "step": 3010 }, { "epoch": 0.37, "learning_rate": 1.4636497108280983e-05, "loss": 0.5679, "step": 3011 }, { "epoch": 0.37, "learning_rate": 1.4633006892269033e-05, "loss": 0.6056, "step": 3012 }, { "epoch": 0.37, "learning_rate": 1.4629515957483022e-05, "loss": 0.5923, "step": 3013 }, { "epoch": 0.37, "learning_rate": 1.4626024304464545e-05, "loss": 0.6121, "step": 3014 }, { "epoch": 0.37, "learning_rate": 1.46225319337553e-05, "loss": 0.529, "step": 3015 }, { "epoch": 0.37, "learning_rate": 1.4619038845897102e-05, "loss": 0.6048, "step": 3016 }, { "epoch": 0.37, "learning_rate": 1.4615545041431874e-05, "loss": 0.5664, "step": 3017 }, { "epoch": 0.37, "learning_rate": 1.461205052090165e-05, "loss": 0.5569, "step": 3018 }, { "epoch": 0.37, "learning_rate": 1.4608555284848582e-05, "loss": 0.5728, "step": 3019 }, { "epoch": 0.37, "learning_rate": 1.4605059333814925e-05, "loss": 0.5582, "step": 3020 }, { "epoch": 0.37, "learning_rate": 1.460156266834305e-05, "loss": 0.6286, "step": 3021 }, { "epoch": 0.37, "learning_rate": 1.4598065288975433e-05, "loss": 0.5784, "step": 3022 }, { "epoch": 0.37, "learning_rate": 1.459456719625467e-05, "loss": 0.5947, "step": 3023 }, { "epoch": 0.37, "learning_rate": 1.4591068390723457e-05, "loss": 0.5663, "step": 3024 }, { "epoch": 0.37, "learning_rate": 1.4587568872924606e-05, "loss": 0.5942, "step": 3025 }, { "epoch": 0.37, "learning_rate": 1.4584068643401045e-05, "loss": 0.5555, "step": 3026 }, { "epoch": 0.37, "learning_rate": 1.4580567702695801e-05, "loss": 0.5766, "step": 3027 }, { "epoch": 0.37, "learning_rate": 1.457706605135202e-05, "loss": 0.5888, "step": 3028 }, { "epoch": 0.37, "learning_rate": 1.4573563689912953e-05, "loss": 0.5883, "step": 3029 }, { "epoch": 0.37, "learning_rate": 1.4570060618921964e-05, "loss": 0.5475, "step": 3030 }, { "epoch": 0.37, "learning_rate": 1.4566556838922526e-05, "loss": 0.5845, "step": 3031 }, { "epoch": 0.37, "learning_rate": 1.4563052350458225e-05, "loss": 0.6045, "step": 3032 }, { "epoch": 0.37, "learning_rate": 1.4559547154072756e-05, "loss": 0.5907, "step": 3033 }, { "epoch": 0.37, "learning_rate": 1.4556041250309916e-05, "loss": 0.6173, "step": 3034 }, { "epoch": 0.37, "learning_rate": 1.455253463971362e-05, "loss": 0.5744, "step": 3035 }, { "epoch": 0.37, "learning_rate": 1.4549027322827895e-05, "loss": 0.5988, "step": 3036 }, { "epoch": 0.37, "learning_rate": 1.4545519300196866e-05, "loss": 0.5277, "step": 3037 }, { "epoch": 0.37, "learning_rate": 1.4542010572364782e-05, "loss": 0.5554, "step": 3038 }, { "epoch": 0.37, "learning_rate": 1.4538501139875991e-05, "loss": 0.5277, "step": 3039 }, { "epoch": 0.37, "learning_rate": 1.4534991003274954e-05, "loss": 0.5862, "step": 3040 }, { "epoch": 0.37, "learning_rate": 1.4531480163106238e-05, "loss": 0.542, "step": 3041 }, { "epoch": 0.37, "learning_rate": 1.4527968619914525e-05, "loss": 0.5761, "step": 3042 }, { "epoch": 0.37, "learning_rate": 1.4524456374244602e-05, "loss": 0.541, "step": 3043 }, { "epoch": 0.37, "learning_rate": 1.4520943426641366e-05, "loss": 0.5777, "step": 3044 }, { "epoch": 0.37, "learning_rate": 1.451742977764982e-05, "loss": 0.642, "step": 3045 }, { "epoch": 0.37, "learning_rate": 1.4513915427815086e-05, "loss": 0.5563, "step": 3046 }, { "epoch": 0.37, "learning_rate": 1.4510400377682378e-05, "loss": 0.6036, "step": 3047 }, { "epoch": 0.37, "learning_rate": 1.4506884627797035e-05, "loss": 0.6014, "step": 3048 }, { "epoch": 0.37, "learning_rate": 1.4503368178704495e-05, "loss": 0.5804, "step": 3049 }, { "epoch": 0.37, "learning_rate": 1.4499851030950307e-05, "loss": 0.5458, "step": 3050 }, { "epoch": 0.37, "learning_rate": 1.4496333185080128e-05, "loss": 0.6423, "step": 3051 }, { "epoch": 0.37, "learning_rate": 1.4492814641639724e-05, "loss": 0.6176, "step": 3052 }, { "epoch": 0.37, "learning_rate": 1.4489295401174972e-05, "loss": 0.462, "step": 3053 }, { "epoch": 0.37, "learning_rate": 1.4485775464231848e-05, "loss": 0.5579, "step": 3054 }, { "epoch": 0.37, "learning_rate": 1.4482254831356444e-05, "loss": 0.5981, "step": 3055 }, { "epoch": 0.37, "learning_rate": 1.4478733503094964e-05, "loss": 0.5795, "step": 3056 }, { "epoch": 0.37, "learning_rate": 1.4475211479993707e-05, "loss": 0.5758, "step": 3057 }, { "epoch": 0.37, "learning_rate": 1.447168876259909e-05, "loss": 0.5415, "step": 3058 }, { "epoch": 0.37, "learning_rate": 1.4468165351457635e-05, "loss": 0.5302, "step": 3059 }, { "epoch": 0.37, "learning_rate": 1.446464124711597e-05, "loss": 0.561, "step": 3060 }, { "epoch": 0.37, "learning_rate": 1.4461116450120826e-05, "loss": 0.5798, "step": 3061 }, { "epoch": 0.37, "learning_rate": 1.4457590961019058e-05, "loss": 0.579, "step": 3062 }, { "epoch": 0.37, "learning_rate": 1.4454064780357611e-05, "loss": 0.55, "step": 3063 }, { "epoch": 0.37, "learning_rate": 1.4450537908683544e-05, "loss": 0.5273, "step": 3064 }, { "epoch": 0.37, "learning_rate": 1.4447010346544026e-05, "loss": 0.6262, "step": 3065 }, { "epoch": 0.37, "learning_rate": 1.444348209448633e-05, "loss": 0.5517, "step": 3066 }, { "epoch": 0.37, "learning_rate": 1.4439953153057829e-05, "loss": 0.5864, "step": 3067 }, { "epoch": 0.37, "learning_rate": 1.4436423522806019e-05, "loss": 0.5581, "step": 3068 }, { "epoch": 0.37, "learning_rate": 1.4432893204278488e-05, "loss": 0.5476, "step": 3069 }, { "epoch": 0.37, "learning_rate": 1.4429362198022942e-05, "loss": 0.5942, "step": 3070 }, { "epoch": 0.37, "learning_rate": 1.4425830504587183e-05, "loss": 0.5705, "step": 3071 }, { "epoch": 0.37, "learning_rate": 1.4422298124519129e-05, "loss": 0.5372, "step": 3072 }, { "epoch": 0.37, "learning_rate": 1.44187650583668e-05, "loss": 0.5798, "step": 3073 }, { "epoch": 0.37, "learning_rate": 1.4415231306678321e-05, "loss": 0.6319, "step": 3074 }, { "epoch": 0.37, "learning_rate": 1.4411696870001928e-05, "loss": 0.5444, "step": 3075 }, { "epoch": 0.37, "learning_rate": 1.4408161748885958e-05, "loss": 0.529, "step": 3076 }, { "epoch": 0.37, "learning_rate": 1.4404625943878858e-05, "loss": 0.5451, "step": 3077 }, { "epoch": 0.37, "learning_rate": 1.4401089455529183e-05, "loss": 0.5523, "step": 3078 }, { "epoch": 0.37, "learning_rate": 1.4397552284385586e-05, "loss": 0.521, "step": 3079 }, { "epoch": 0.37, "learning_rate": 1.4394014430996831e-05, "loss": 0.6018, "step": 3080 }, { "epoch": 0.37, "learning_rate": 1.4390475895911794e-05, "loss": 0.5638, "step": 3081 }, { "epoch": 0.37, "learning_rate": 1.4386936679679443e-05, "loss": 0.497, "step": 3082 }, { "epoch": 0.37, "learning_rate": 1.4383396782848864e-05, "loss": 0.5963, "step": 3083 }, { "epoch": 0.38, "learning_rate": 1.437985620596924e-05, "loss": 0.6069, "step": 3084 }, { "epoch": 0.38, "learning_rate": 1.4376314949589867e-05, "loss": 0.5852, "step": 3085 }, { "epoch": 0.38, "learning_rate": 1.4372773014260142e-05, "loss": 0.5474, "step": 3086 }, { "epoch": 0.38, "learning_rate": 1.4369230400529563e-05, "loss": 0.5502, "step": 3087 }, { "epoch": 0.38, "learning_rate": 1.4365687108947746e-05, "loss": 0.5855, "step": 3088 }, { "epoch": 0.38, "learning_rate": 1.4362143140064401e-05, "loss": 0.5736, "step": 3089 }, { "epoch": 0.38, "learning_rate": 1.4358598494429345e-05, "loss": 0.554, "step": 3090 }, { "epoch": 0.38, "learning_rate": 1.4355053172592501e-05, "loss": 0.5768, "step": 3091 }, { "epoch": 0.38, "learning_rate": 1.4351507175103905e-05, "loss": 0.5148, "step": 3092 }, { "epoch": 0.38, "learning_rate": 1.434796050251368e-05, "loss": 0.5891, "step": 3093 }, { "epoch": 0.38, "learning_rate": 1.4344413155372068e-05, "loss": 0.6014, "step": 3094 }, { "epoch": 0.38, "learning_rate": 1.4340865134229412e-05, "loss": 0.6317, "step": 3095 }, { "epoch": 0.38, "learning_rate": 1.4337316439636162e-05, "loss": 0.5758, "step": 3096 }, { "epoch": 0.38, "learning_rate": 1.4333767072142861e-05, "loss": 0.6211, "step": 3097 }, { "epoch": 0.38, "learning_rate": 1.4330217032300177e-05, "loss": 0.5846, "step": 3098 }, { "epoch": 0.38, "learning_rate": 1.4326666320658862e-05, "loss": 0.6579, "step": 3099 }, { "epoch": 0.38, "learning_rate": 1.4323114937769781e-05, "loss": 0.6377, "step": 3100 }, { "epoch": 0.38, "learning_rate": 1.4319562884183906e-05, "loss": 0.5739, "step": 3101 }, { "epoch": 0.38, "learning_rate": 1.431601016045231e-05, "loss": 0.5729, "step": 3102 }, { "epoch": 0.38, "learning_rate": 1.4312456767126163e-05, "loss": 0.5922, "step": 3103 }, { "epoch": 0.38, "learning_rate": 1.4308902704756755e-05, "loss": 0.5277, "step": 3104 }, { "epoch": 0.38, "learning_rate": 1.4305347973895467e-05, "loss": 0.5754, "step": 3105 }, { "epoch": 0.38, "learning_rate": 1.4301792575093783e-05, "loss": 0.553, "step": 3106 }, { "epoch": 0.38, "learning_rate": 1.4298236508903298e-05, "loss": 0.5422, "step": 3107 }, { "epoch": 0.38, "learning_rate": 1.4294679775875708e-05, "loss": 0.5054, "step": 3108 }, { "epoch": 0.38, "learning_rate": 1.4291122376562811e-05, "loss": 0.6124, "step": 3109 }, { "epoch": 0.38, "learning_rate": 1.4287564311516509e-05, "loss": 0.5434, "step": 3110 }, { "epoch": 0.38, "learning_rate": 1.428400558128881e-05, "loss": 0.6366, "step": 3111 }, { "epoch": 0.38, "learning_rate": 1.4280446186431818e-05, "loss": 0.5729, "step": 3112 }, { "epoch": 0.38, "learning_rate": 1.4276886127497746e-05, "loss": 0.5532, "step": 3113 }, { "epoch": 0.38, "learning_rate": 1.4273325405038912e-05, "loss": 0.6078, "step": 3114 }, { "epoch": 0.38, "learning_rate": 1.4269764019607729e-05, "loss": 0.5578, "step": 3115 }, { "epoch": 0.38, "learning_rate": 1.426620197175672e-05, "loss": 0.5614, "step": 3116 }, { "epoch": 0.38, "learning_rate": 1.4262639262038509e-05, "loss": 0.5573, "step": 3117 }, { "epoch": 0.38, "learning_rate": 1.4259075891005823e-05, "loss": 0.5741, "step": 3118 }, { "epoch": 0.38, "learning_rate": 1.4255511859211486e-05, "loss": 0.566, "step": 3119 }, { "epoch": 0.38, "learning_rate": 1.425194716720843e-05, "loss": 0.5926, "step": 3120 }, { "epoch": 0.38, "learning_rate": 1.4248381815549695e-05, "loss": 0.5528, "step": 3121 }, { "epoch": 0.38, "learning_rate": 1.424481580478841e-05, "loss": 0.6035, "step": 3122 }, { "epoch": 0.38, "learning_rate": 1.4241249135477815e-05, "loss": 0.5896, "step": 3123 }, { "epoch": 0.38, "learning_rate": 1.423768180817125e-05, "loss": 0.5693, "step": 3124 }, { "epoch": 0.38, "learning_rate": 1.4234113823422162e-05, "loss": 0.6456, "step": 3125 }, { "epoch": 0.38, "learning_rate": 1.4230545181784087e-05, "loss": 0.5627, "step": 3126 }, { "epoch": 0.38, "learning_rate": 1.4226975883810677e-05, "loss": 0.502, "step": 3127 }, { "epoch": 0.38, "learning_rate": 1.422340593005568e-05, "loss": 0.5457, "step": 3128 }, { "epoch": 0.38, "learning_rate": 1.4219835321072943e-05, "loss": 0.6356, "step": 3129 }, { "epoch": 0.38, "learning_rate": 1.421626405741642e-05, "loss": 0.5593, "step": 3130 }, { "epoch": 0.38, "learning_rate": 1.4212692139640163e-05, "loss": 0.6003, "step": 3131 }, { "epoch": 0.38, "learning_rate": 1.4209119568298329e-05, "loss": 0.5461, "step": 3132 }, { "epoch": 0.38, "learning_rate": 1.4205546343945167e-05, "loss": 0.613, "step": 3133 }, { "epoch": 0.38, "learning_rate": 1.4201972467135039e-05, "loss": 0.5575, "step": 3134 }, { "epoch": 0.38, "learning_rate": 1.4198397938422408e-05, "loss": 0.6111, "step": 3135 }, { "epoch": 0.38, "learning_rate": 1.4194822758361829e-05, "loss": 0.6248, "step": 3136 }, { "epoch": 0.38, "learning_rate": 1.4191246927507958e-05, "loss": 0.6024, "step": 3137 }, { "epoch": 0.38, "learning_rate": 1.4187670446415567e-05, "loss": 0.576, "step": 3138 }, { "epoch": 0.38, "learning_rate": 1.4184093315639507e-05, "loss": 0.5454, "step": 3139 }, { "epoch": 0.38, "learning_rate": 1.418051553573475e-05, "loss": 0.5927, "step": 3140 }, { "epoch": 0.38, "learning_rate": 1.4176937107256356e-05, "loss": 0.5618, "step": 3141 }, { "epoch": 0.38, "learning_rate": 1.4173358030759492e-05, "loss": 0.5704, "step": 3142 }, { "epoch": 0.38, "learning_rate": 1.4169778306799422e-05, "loss": 0.5676, "step": 3143 }, { "epoch": 0.38, "learning_rate": 1.4166197935931512e-05, "loss": 0.5733, "step": 3144 }, { "epoch": 0.38, "learning_rate": 1.4162616918711224e-05, "loss": 0.6156, "step": 3145 }, { "epoch": 0.38, "learning_rate": 1.415903525569413e-05, "loss": 0.5599, "step": 3146 }, { "epoch": 0.38, "learning_rate": 1.4155452947435895e-05, "loss": 0.5436, "step": 3147 }, { "epoch": 0.38, "learning_rate": 1.4151869994492287e-05, "loss": 0.5803, "step": 3148 }, { "epoch": 0.38, "learning_rate": 1.4148286397419166e-05, "loss": 0.5577, "step": 3149 }, { "epoch": 0.38, "learning_rate": 1.4144702156772506e-05, "loss": 0.5626, "step": 3150 }, { "epoch": 0.38, "learning_rate": 1.4141117273108366e-05, "loss": 0.5858, "step": 3151 }, { "epoch": 0.38, "learning_rate": 1.4137531746982923e-05, "loss": 0.5461, "step": 3152 }, { "epoch": 0.38, "learning_rate": 1.413394557895243e-05, "loss": 0.5771, "step": 3153 }, { "epoch": 0.38, "learning_rate": 1.4130358769573261e-05, "loss": 0.5396, "step": 3154 }, { "epoch": 0.38, "learning_rate": 1.412677131940188e-05, "loss": 0.5927, "step": 3155 }, { "epoch": 0.38, "learning_rate": 1.4123183228994846e-05, "loss": 0.5102, "step": 3156 }, { "epoch": 0.38, "learning_rate": 1.4119594498908826e-05, "loss": 0.5798, "step": 3157 }, { "epoch": 0.38, "learning_rate": 1.4116005129700586e-05, "loss": 0.5731, "step": 3158 }, { "epoch": 0.38, "learning_rate": 1.4112415121926983e-05, "loss": 0.5936, "step": 3159 }, { "epoch": 0.38, "learning_rate": 1.4108824476144982e-05, "loss": 0.596, "step": 3160 }, { "epoch": 0.38, "learning_rate": 1.4105233192911642e-05, "loss": 0.6056, "step": 3161 }, { "epoch": 0.38, "learning_rate": 1.410164127278412e-05, "loss": 0.5575, "step": 3162 }, { "epoch": 0.38, "learning_rate": 1.409804871631967e-05, "loss": 0.5951, "step": 3163 }, { "epoch": 0.38, "learning_rate": 1.4094455524075658e-05, "loss": 0.5636, "step": 3164 }, { "epoch": 0.38, "learning_rate": 1.4090861696609534e-05, "loss": 0.6092, "step": 3165 }, { "epoch": 0.39, "learning_rate": 1.4087267234478852e-05, "loss": 0.661, "step": 3166 }, { "epoch": 0.39, "learning_rate": 1.4083672138241263e-05, "loss": 0.5928, "step": 3167 }, { "epoch": 0.39, "learning_rate": 1.4080076408454522e-05, "loss": 0.5719, "step": 3168 }, { "epoch": 0.39, "learning_rate": 1.4076480045676469e-05, "loss": 0.5858, "step": 3169 }, { "epoch": 0.39, "learning_rate": 1.407288305046506e-05, "loss": 0.5764, "step": 3170 }, { "epoch": 0.39, "learning_rate": 1.4069285423378338e-05, "loss": 0.597, "step": 3171 }, { "epoch": 0.39, "learning_rate": 1.4065687164974441e-05, "loss": 0.5796, "step": 3172 }, { "epoch": 0.39, "learning_rate": 1.4062088275811618e-05, "loss": 0.6456, "step": 3173 }, { "epoch": 0.39, "learning_rate": 1.4058488756448201e-05, "loss": 0.5831, "step": 3174 }, { "epoch": 0.39, "learning_rate": 1.405488860744263e-05, "loss": 0.594, "step": 3175 }, { "epoch": 0.39, "learning_rate": 1.4051287829353437e-05, "loss": 0.5609, "step": 3176 }, { "epoch": 0.39, "learning_rate": 1.404768642273926e-05, "loss": 0.524, "step": 3177 }, { "epoch": 0.39, "learning_rate": 1.4044084388158823e-05, "loss": 0.5669, "step": 3178 }, { "epoch": 0.39, "learning_rate": 1.4040481726170953e-05, "loss": 0.6147, "step": 3179 }, { "epoch": 0.39, "learning_rate": 1.4036878437334575e-05, "loss": 0.5824, "step": 3180 }, { "epoch": 0.39, "learning_rate": 1.4033274522208713e-05, "loss": 0.5818, "step": 3181 }, { "epoch": 0.39, "learning_rate": 1.4029669981352485e-05, "loss": 0.6263, "step": 3182 }, { "epoch": 0.39, "learning_rate": 1.4026064815325102e-05, "loss": 0.6227, "step": 3183 }, { "epoch": 0.39, "learning_rate": 1.4022459024685884e-05, "loss": 0.5557, "step": 3184 }, { "epoch": 0.39, "learning_rate": 1.4018852609994235e-05, "loss": 0.5612, "step": 3185 }, { "epoch": 0.39, "learning_rate": 1.4015245571809663e-05, "loss": 0.6467, "step": 3186 }, { "epoch": 0.39, "learning_rate": 1.401163791069177e-05, "loss": 0.598, "step": 3187 }, { "epoch": 0.39, "learning_rate": 1.4008029627200262e-05, "loss": 0.5982, "step": 3188 }, { "epoch": 0.39, "learning_rate": 1.4004420721894926e-05, "loss": 0.5479, "step": 3189 }, { "epoch": 0.39, "learning_rate": 1.4000811195335664e-05, "loss": 0.5971, "step": 3190 }, { "epoch": 0.39, "learning_rate": 1.399720104808246e-05, "loss": 0.5889, "step": 3191 }, { "epoch": 0.39, "learning_rate": 1.39935902806954e-05, "loss": 0.5377, "step": 3192 }, { "epoch": 0.39, "learning_rate": 1.3989978893734664e-05, "loss": 0.5092, "step": 3193 }, { "epoch": 0.39, "learning_rate": 1.3986366887760533e-05, "loss": 0.5404, "step": 3194 }, { "epoch": 0.39, "learning_rate": 1.3982754263333382e-05, "loss": 0.6017, "step": 3195 }, { "epoch": 0.39, "learning_rate": 1.397914102101368e-05, "loss": 0.6352, "step": 3196 }, { "epoch": 0.39, "learning_rate": 1.397552716136199e-05, "loss": 0.5041, "step": 3197 }, { "epoch": 0.39, "learning_rate": 1.3971912684938975e-05, "loss": 0.5396, "step": 3198 }, { "epoch": 0.39, "learning_rate": 1.3968297592305391e-05, "loss": 0.6123, "step": 3199 }, { "epoch": 0.39, "learning_rate": 1.3964681884022095e-05, "loss": 0.5814, "step": 3200 }, { "epoch": 0.39, "learning_rate": 1.396106556065003e-05, "loss": 0.5535, "step": 3201 }, { "epoch": 0.39, "learning_rate": 1.3957448622750243e-05, "loss": 0.5832, "step": 3202 }, { "epoch": 0.39, "learning_rate": 1.3953831070883872e-05, "loss": 0.5343, "step": 3203 }, { "epoch": 0.39, "learning_rate": 1.3950212905612155e-05, "loss": 0.6002, "step": 3204 }, { "epoch": 0.39, "learning_rate": 1.394659412749641e-05, "loss": 0.5473, "step": 3205 }, { "epoch": 0.39, "learning_rate": 1.3942974737098073e-05, "loss": 0.5758, "step": 3206 }, { "epoch": 0.39, "learning_rate": 1.3939354734978665e-05, "loss": 0.5581, "step": 3207 }, { "epoch": 0.39, "learning_rate": 1.393573412169979e-05, "loss": 0.5852, "step": 3208 }, { "epoch": 0.39, "learning_rate": 1.3932112897823167e-05, "loss": 0.5809, "step": 3209 }, { "epoch": 0.39, "learning_rate": 1.3928491063910592e-05, "loss": 0.5977, "step": 3210 }, { "epoch": 0.39, "learning_rate": 1.392486862052397e-05, "loss": 0.5283, "step": 3211 }, { "epoch": 0.39, "learning_rate": 1.392124556822529e-05, "loss": 0.5141, "step": 3212 }, { "epoch": 0.39, "learning_rate": 1.3917621907576641e-05, "loss": 0.6015, "step": 3213 }, { "epoch": 0.39, "learning_rate": 1.3913997639140209e-05, "loss": 0.6489, "step": 3214 }, { "epoch": 0.39, "learning_rate": 1.3910372763478263e-05, "loss": 0.5499, "step": 3215 }, { "epoch": 0.39, "learning_rate": 1.3906747281153178e-05, "loss": 0.5679, "step": 3216 }, { "epoch": 0.39, "learning_rate": 1.390312119272742e-05, "loss": 0.5493, "step": 3217 }, { "epoch": 0.39, "learning_rate": 1.3899494498763542e-05, "loss": 0.5914, "step": 3218 }, { "epoch": 0.39, "learning_rate": 1.3895867199824202e-05, "loss": 0.5322, "step": 3219 }, { "epoch": 0.39, "learning_rate": 1.3892239296472147e-05, "loss": 0.59, "step": 3220 }, { "epoch": 0.39, "learning_rate": 1.3888610789270213e-05, "loss": 0.5813, "step": 3221 }, { "epoch": 0.39, "learning_rate": 1.3884981678781334e-05, "loss": 0.5576, "step": 3222 }, { "epoch": 0.39, "learning_rate": 1.3881351965568542e-05, "loss": 0.5997, "step": 3223 }, { "epoch": 0.39, "learning_rate": 1.3877721650194955e-05, "loss": 0.5575, "step": 3224 }, { "epoch": 0.39, "learning_rate": 1.3874090733223787e-05, "loss": 0.54, "step": 3225 }, { "epoch": 0.39, "learning_rate": 1.3870459215218346e-05, "loss": 0.6191, "step": 3226 }, { "epoch": 0.39, "learning_rate": 1.386682709674204e-05, "loss": 0.5707, "step": 3227 }, { "epoch": 0.39, "learning_rate": 1.3863194378358352e-05, "loss": 0.6076, "step": 3228 }, { "epoch": 0.39, "learning_rate": 1.3859561060630871e-05, "loss": 0.5256, "step": 3229 }, { "epoch": 0.39, "learning_rate": 1.3855927144123284e-05, "loss": 0.5732, "step": 3230 }, { "epoch": 0.39, "learning_rate": 1.3852292629399362e-05, "loss": 0.611, "step": 3231 }, { "epoch": 0.39, "learning_rate": 1.3848657517022969e-05, "loss": 0.584, "step": 3232 }, { "epoch": 0.39, "learning_rate": 1.3845021807558064e-05, "loss": 0.579, "step": 3233 }, { "epoch": 0.39, "learning_rate": 1.38413855015687e-05, "loss": 0.5418, "step": 3234 }, { "epoch": 0.39, "learning_rate": 1.3837748599619017e-05, "loss": 0.5158, "step": 3235 }, { "epoch": 0.39, "learning_rate": 1.3834111102273254e-05, "loss": 0.5731, "step": 3236 }, { "epoch": 0.39, "learning_rate": 1.3830473010095742e-05, "loss": 0.5853, "step": 3237 }, { "epoch": 0.39, "learning_rate": 1.3826834323650899e-05, "loss": 0.4999, "step": 3238 }, { "epoch": 0.39, "learning_rate": 1.3823195043503237e-05, "loss": 0.6304, "step": 3239 }, { "epoch": 0.39, "learning_rate": 1.3819555170217367e-05, "loss": 0.5429, "step": 3240 }, { "epoch": 0.39, "learning_rate": 1.381591470435798e-05, "loss": 0.6021, "step": 3241 }, { "epoch": 0.39, "learning_rate": 1.3812273646489865e-05, "loss": 0.599, "step": 3242 }, { "epoch": 0.39, "learning_rate": 1.3808631997177911e-05, "loss": 0.6178, "step": 3243 }, { "epoch": 0.39, "learning_rate": 1.380498975698708e-05, "loss": 0.5656, "step": 3244 }, { "epoch": 0.39, "learning_rate": 1.3801346926482444e-05, "loss": 0.5269, "step": 3245 }, { "epoch": 0.39, "learning_rate": 1.3797703506229158e-05, "loss": 0.5887, "step": 3246 }, { "epoch": 0.39, "learning_rate": 1.3794059496792469e-05, "loss": 0.5322, "step": 3247 }, { "epoch": 0.39, "learning_rate": 1.3790414898737715e-05, "loss": 0.6079, "step": 3248 }, { "epoch": 0.4, "learning_rate": 1.3786769712630323e-05, "loss": 0.5708, "step": 3249 }, { "epoch": 0.4, "learning_rate": 1.3783123939035824e-05, "loss": 0.5295, "step": 3250 }, { "epoch": 0.4, "learning_rate": 1.3779477578519825e-05, "loss": 0.595, "step": 3251 }, { "epoch": 0.4, "learning_rate": 1.3775830631648029e-05, "loss": 0.5121, "step": 3252 }, { "epoch": 0.4, "learning_rate": 1.3772183098986232e-05, "loss": 0.599, "step": 3253 }, { "epoch": 0.4, "learning_rate": 1.3768534981100316e-05, "loss": 0.612, "step": 3254 }, { "epoch": 0.4, "learning_rate": 1.3764886278556264e-05, "loss": 0.568, "step": 3255 }, { "epoch": 0.4, "learning_rate": 1.376123699192014e-05, "loss": 0.6031, "step": 3256 }, { "epoch": 0.4, "learning_rate": 1.3757587121758103e-05, "loss": 0.5396, "step": 3257 }, { "epoch": 0.4, "learning_rate": 1.3753936668636395e-05, "loss": 0.4834, "step": 3258 }, { "epoch": 0.4, "learning_rate": 1.3750285633121361e-05, "loss": 0.6224, "step": 3259 }, { "epoch": 0.4, "learning_rate": 1.374663401577943e-05, "loss": 0.5473, "step": 3260 }, { "epoch": 0.4, "learning_rate": 1.3742981817177116e-05, "loss": 0.519, "step": 3261 }, { "epoch": 0.4, "learning_rate": 1.3739329037881038e-05, "loss": 0.5467, "step": 3262 }, { "epoch": 0.4, "learning_rate": 1.3735675678457886e-05, "loss": 0.5811, "step": 3263 }, { "epoch": 0.4, "learning_rate": 1.3732021739474455e-05, "loss": 0.5574, "step": 3264 }, { "epoch": 0.4, "learning_rate": 1.3728367221497624e-05, "loss": 0.5538, "step": 3265 }, { "epoch": 0.4, "learning_rate": 1.3724712125094358e-05, "loss": 0.5443, "step": 3266 }, { "epoch": 0.4, "learning_rate": 1.3721056450831719e-05, "loss": 0.5787, "step": 3267 }, { "epoch": 0.4, "learning_rate": 1.3717400199276858e-05, "loss": 0.5138, "step": 3268 }, { "epoch": 0.4, "learning_rate": 1.371374337099701e-05, "loss": 0.585, "step": 3269 }, { "epoch": 0.4, "learning_rate": 1.3710085966559505e-05, "loss": 0.5652, "step": 3270 }, { "epoch": 0.4, "learning_rate": 1.3706427986531759e-05, "loss": 0.5841, "step": 3271 }, { "epoch": 0.4, "learning_rate": 1.3702769431481272e-05, "loss": 0.5563, "step": 3272 }, { "epoch": 0.4, "learning_rate": 1.3699110301975655e-05, "loss": 0.5419, "step": 3273 }, { "epoch": 0.4, "learning_rate": 1.3695450598582581e-05, "loss": 0.622, "step": 3274 }, { "epoch": 0.4, "learning_rate": 1.3691790321869827e-05, "loss": 0.5998, "step": 3275 }, { "epoch": 0.4, "learning_rate": 1.3688129472405253e-05, "loss": 0.6113, "step": 3276 }, { "epoch": 0.4, "learning_rate": 1.3684468050756816e-05, "loss": 0.563, "step": 3277 }, { "epoch": 0.4, "learning_rate": 1.368080605749255e-05, "loss": 0.5857, "step": 3278 }, { "epoch": 0.4, "learning_rate": 1.3677143493180589e-05, "loss": 0.6182, "step": 3279 }, { "epoch": 0.4, "learning_rate": 1.3673480358389151e-05, "loss": 0.5404, "step": 3280 }, { "epoch": 0.4, "learning_rate": 1.3669816653686538e-05, "loss": 0.6311, "step": 3281 }, { "epoch": 0.4, "learning_rate": 1.3666152379641146e-05, "loss": 0.5845, "step": 3282 }, { "epoch": 0.4, "learning_rate": 1.3662487536821463e-05, "loss": 0.5907, "step": 3283 }, { "epoch": 0.4, "learning_rate": 1.3658822125796047e-05, "loss": 0.6079, "step": 3284 }, { "epoch": 0.4, "learning_rate": 1.3655156147133572e-05, "loss": 0.5466, "step": 3285 }, { "epoch": 0.4, "learning_rate": 1.365148960140278e-05, "loss": 0.5601, "step": 3286 }, { "epoch": 0.4, "learning_rate": 1.3647822489172505e-05, "loss": 0.5632, "step": 3287 }, { "epoch": 0.4, "learning_rate": 1.3644154811011666e-05, "loss": 0.5961, "step": 3288 }, { "epoch": 0.4, "learning_rate": 1.3640486567489284e-05, "loss": 0.5507, "step": 3289 }, { "epoch": 0.4, "learning_rate": 1.3636817759174453e-05, "loss": 0.565, "step": 3290 }, { "epoch": 0.4, "learning_rate": 1.3633148386636354e-05, "loss": 0.5798, "step": 3291 }, { "epoch": 0.4, "learning_rate": 1.3629478450444271e-05, "loss": 0.6042, "step": 3292 }, { "epoch": 0.4, "learning_rate": 1.3625807951167557e-05, "loss": 0.565, "step": 3293 }, { "epoch": 0.4, "learning_rate": 1.3622136889375665e-05, "loss": 0.5681, "step": 3294 }, { "epoch": 0.4, "learning_rate": 1.361846526563813e-05, "loss": 0.5532, "step": 3295 }, { "epoch": 0.4, "learning_rate": 1.3614793080524576e-05, "loss": 0.5526, "step": 3296 }, { "epoch": 0.4, "learning_rate": 1.3611120334604708e-05, "loss": 0.5408, "step": 3297 }, { "epoch": 0.4, "learning_rate": 1.3607447028448331e-05, "loss": 0.5582, "step": 3298 }, { "epoch": 0.4, "learning_rate": 1.3603773162625325e-05, "loss": 0.571, "step": 3299 }, { "epoch": 0.4, "learning_rate": 1.3600098737705661e-05, "loss": 0.6158, "step": 3300 }, { "epoch": 0.4, "learning_rate": 1.3596423754259396e-05, "loss": 0.573, "step": 3301 }, { "epoch": 0.4, "learning_rate": 1.3592748212856677e-05, "loss": 0.5611, "step": 3302 }, { "epoch": 0.4, "learning_rate": 1.3589072114067732e-05, "loss": 0.5967, "step": 3303 }, { "epoch": 0.4, "learning_rate": 1.358539545846288e-05, "loss": 0.5886, "step": 3304 }, { "epoch": 0.4, "learning_rate": 1.3581718246612524e-05, "loss": 0.579, "step": 3305 }, { "epoch": 0.4, "learning_rate": 1.357804047908716e-05, "loss": 0.54, "step": 3306 }, { "epoch": 0.4, "learning_rate": 1.3574362156457354e-05, "loss": 0.5741, "step": 3307 }, { "epoch": 0.4, "learning_rate": 1.3570683279293773e-05, "loss": 0.604, "step": 3308 }, { "epoch": 0.4, "learning_rate": 1.3567003848167169e-05, "loss": 0.599, "step": 3309 }, { "epoch": 0.4, "learning_rate": 1.356332386364837e-05, "loss": 0.5509, "step": 3310 }, { "epoch": 0.4, "learning_rate": 1.35596433263083e-05, "loss": 0.5219, "step": 3311 }, { "epoch": 0.4, "learning_rate": 1.3555962236717965e-05, "loss": 0.5683, "step": 3312 }, { "epoch": 0.4, "learning_rate": 1.3552280595448459e-05, "loss": 0.5431, "step": 3313 }, { "epoch": 0.4, "learning_rate": 1.3548598403070954e-05, "loss": 0.5739, "step": 3314 }, { "epoch": 0.4, "learning_rate": 1.3544915660156714e-05, "loss": 0.5393, "step": 3315 }, { "epoch": 0.4, "learning_rate": 1.3541232367277094e-05, "loss": 0.57, "step": 3316 }, { "epoch": 0.4, "learning_rate": 1.3537548525003516e-05, "loss": 0.5938, "step": 3317 }, { "epoch": 0.4, "learning_rate": 1.3533864133907506e-05, "loss": 0.5954, "step": 3318 }, { "epoch": 0.4, "learning_rate": 1.3530179194560668e-05, "loss": 0.5242, "step": 3319 }, { "epoch": 0.4, "learning_rate": 1.3526493707534687e-05, "loss": 0.6213, "step": 3320 }, { "epoch": 0.4, "learning_rate": 1.3522807673401342e-05, "loss": 0.6058, "step": 3321 }, { "epoch": 0.4, "learning_rate": 1.351912109273249e-05, "loss": 0.5358, "step": 3322 }, { "epoch": 0.4, "learning_rate": 1.3515433966100072e-05, "loss": 0.5449, "step": 3323 }, { "epoch": 0.4, "learning_rate": 1.3511746294076118e-05, "loss": 0.5786, "step": 3324 }, { "epoch": 0.4, "learning_rate": 1.3508058077232739e-05, "loss": 0.5759, "step": 3325 }, { "epoch": 0.4, "learning_rate": 1.3504369316142136e-05, "loss": 0.5492, "step": 3326 }, { "epoch": 0.4, "learning_rate": 1.350068001137659e-05, "loss": 0.6791, "step": 3327 }, { "epoch": 0.4, "learning_rate": 1.3496990163508468e-05, "loss": 0.5791, "step": 3328 }, { "epoch": 0.4, "learning_rate": 1.3493299773110218e-05, "loss": 0.5636, "step": 3329 }, { "epoch": 0.4, "learning_rate": 1.3489608840754374e-05, "loss": 0.5474, "step": 3330 }, { "epoch": 0.41, "learning_rate": 1.3485917367013558e-05, "loss": 0.5925, "step": 3331 }, { "epoch": 0.41, "learning_rate": 1.348222535246047e-05, "loss": 0.5547, "step": 3332 }, { "epoch": 0.41, "learning_rate": 1.3478532797667897e-05, "loss": 0.6091, "step": 3333 }, { "epoch": 0.41, "learning_rate": 1.3474839703208708e-05, "loss": 0.5886, "step": 3334 }, { "epoch": 0.41, "learning_rate": 1.3471146069655862e-05, "loss": 0.5757, "step": 3335 }, { "epoch": 0.41, "learning_rate": 1.3467451897582395e-05, "loss": 0.6412, "step": 3336 }, { "epoch": 0.41, "learning_rate": 1.3463757187561427e-05, "loss": 0.58, "step": 3337 }, { "epoch": 0.41, "learning_rate": 1.3460061940166158e-05, "loss": 0.6129, "step": 3338 }, { "epoch": 0.41, "learning_rate": 1.3456366155969886e-05, "loss": 0.5737, "step": 3339 }, { "epoch": 0.41, "learning_rate": 1.3452669835545977e-05, "loss": 0.6134, "step": 3340 }, { "epoch": 0.41, "learning_rate": 1.3448972979467886e-05, "loss": 0.5693, "step": 3341 }, { "epoch": 0.41, "learning_rate": 1.3445275588309152e-05, "loss": 0.5411, "step": 3342 }, { "epoch": 0.41, "learning_rate": 1.3441577662643393e-05, "loss": 0.5581, "step": 3343 }, { "epoch": 0.41, "learning_rate": 1.3437879203044312e-05, "loss": 0.5798, "step": 3344 }, { "epoch": 0.41, "learning_rate": 1.3434180210085704e-05, "loss": 0.6127, "step": 3345 }, { "epoch": 0.41, "learning_rate": 1.3430480684341427e-05, "loss": 0.5141, "step": 3346 }, { "epoch": 0.41, "learning_rate": 1.342678062638544e-05, "loss": 0.5605, "step": 3347 }, { "epoch": 0.41, "learning_rate": 1.3423080036791776e-05, "loss": 0.6099, "step": 3348 }, { "epoch": 0.41, "learning_rate": 1.3419378916134551e-05, "loss": 0.5054, "step": 3349 }, { "epoch": 0.41, "learning_rate": 1.3415677264987962e-05, "loss": 0.5626, "step": 3350 }, { "epoch": 0.41, "learning_rate": 1.3411975083926296e-05, "loss": 0.5367, "step": 3351 }, { "epoch": 0.41, "learning_rate": 1.3408272373523917e-05, "loss": 0.5326, "step": 3352 }, { "epoch": 0.41, "learning_rate": 1.3404569134355266e-05, "loss": 0.5087, "step": 3353 }, { "epoch": 0.41, "learning_rate": 1.3400865366994873e-05, "loss": 0.6319, "step": 3354 }, { "epoch": 0.41, "learning_rate": 1.3397161072017348e-05, "loss": 0.5572, "step": 3355 }, { "epoch": 0.41, "learning_rate": 1.3393456249997383e-05, "loss": 0.5577, "step": 3356 }, { "epoch": 0.41, "learning_rate": 1.3389750901509751e-05, "loss": 0.5215, "step": 3357 }, { "epoch": 0.41, "learning_rate": 1.3386045027129308e-05, "loss": 0.5429, "step": 3358 }, { "epoch": 0.41, "learning_rate": 1.3382338627430995e-05, "loss": 0.5699, "step": 3359 }, { "epoch": 0.41, "learning_rate": 1.3378631702989822e-05, "loss": 0.5559, "step": 3360 }, { "epoch": 0.41, "learning_rate": 1.3374924254380894e-05, "loss": 0.5296, "step": 3361 }, { "epoch": 0.41, "learning_rate": 1.337121628217939e-05, "loss": 0.6039, "step": 3362 }, { "epoch": 0.41, "learning_rate": 1.3367507786960574e-05, "loss": 0.5858, "step": 3363 }, { "epoch": 0.41, "learning_rate": 1.3363798769299787e-05, "loss": 0.5967, "step": 3364 }, { "epoch": 0.41, "learning_rate": 1.336008922977246e-05, "loss": 0.5663, "step": 3365 }, { "epoch": 0.41, "learning_rate": 1.335637916895409e-05, "loss": 0.5778, "step": 3366 }, { "epoch": 0.41, "learning_rate": 1.3352668587420267e-05, "loss": 0.575, "step": 3367 }, { "epoch": 0.41, "learning_rate": 1.3348957485746659e-05, "loss": 0.5823, "step": 3368 }, { "epoch": 0.41, "learning_rate": 1.3345245864509011e-05, "loss": 0.5576, "step": 3369 }, { "epoch": 0.41, "learning_rate": 1.3341533724283156e-05, "loss": 0.5468, "step": 3370 }, { "epoch": 0.41, "learning_rate": 1.3337821065644998e-05, "loss": 0.6059, "step": 3371 }, { "epoch": 0.41, "learning_rate": 1.3334107889170533e-05, "loss": 0.5147, "step": 3372 }, { "epoch": 0.41, "learning_rate": 1.3330394195435823e-05, "loss": 0.5713, "step": 3373 }, { "epoch": 0.41, "learning_rate": 1.332667998501702e-05, "loss": 0.5735, "step": 3374 }, { "epoch": 0.41, "learning_rate": 1.332296525849036e-05, "loss": 0.5515, "step": 3375 }, { "epoch": 0.41, "learning_rate": 1.3319250016432144e-05, "loss": 0.5935, "step": 3376 }, { "epoch": 0.41, "learning_rate": 1.3315534259418768e-05, "loss": 0.506, "step": 3377 }, { "epoch": 0.41, "learning_rate": 1.33118179880267e-05, "loss": 0.5974, "step": 3378 }, { "epoch": 0.41, "learning_rate": 1.3308101202832495e-05, "loss": 0.5384, "step": 3379 }, { "epoch": 0.41, "learning_rate": 1.3304383904412775e-05, "loss": 0.5534, "step": 3380 }, { "epoch": 0.41, "learning_rate": 1.3300666093344254e-05, "loss": 0.6207, "step": 3381 }, { "epoch": 0.41, "learning_rate": 1.3296947770203718e-05, "loss": 0.6173, "step": 3382 }, { "epoch": 0.41, "learning_rate": 1.3293228935568036e-05, "loss": 0.617, "step": 3383 }, { "epoch": 0.41, "learning_rate": 1.3289509590014157e-05, "loss": 0.6265, "step": 3384 }, { "epoch": 0.41, "learning_rate": 1.328578973411911e-05, "loss": 0.572, "step": 3385 }, { "epoch": 0.41, "learning_rate": 1.328206936845999e-05, "loss": 0.5347, "step": 3386 }, { "epoch": 0.41, "learning_rate": 1.3278348493613996e-05, "loss": 0.5875, "step": 3387 }, { "epoch": 0.41, "learning_rate": 1.3274627110158385e-05, "loss": 0.5783, "step": 3388 }, { "epoch": 0.41, "learning_rate": 1.3270905218670497e-05, "loss": 0.5887, "step": 3389 }, { "epoch": 0.41, "learning_rate": 1.3267182819727762e-05, "loss": 0.5573, "step": 3390 }, { "epoch": 0.41, "learning_rate": 1.3263459913907673e-05, "loss": 0.6049, "step": 3391 }, { "epoch": 0.41, "learning_rate": 1.3259736501787815e-05, "loss": 0.5725, "step": 3392 }, { "epoch": 0.41, "learning_rate": 1.3256012583945838e-05, "loss": 0.6072, "step": 3393 }, { "epoch": 0.41, "learning_rate": 1.3252288160959488e-05, "loss": 0.5603, "step": 3394 }, { "epoch": 0.41, "learning_rate": 1.3248563233406575e-05, "loss": 0.579, "step": 3395 }, { "epoch": 0.41, "learning_rate": 1.324483780186499e-05, "loss": 0.5138, "step": 3396 }, { "epoch": 0.41, "learning_rate": 1.3241111866912703e-05, "loss": 0.5774, "step": 3397 }, { "epoch": 0.41, "learning_rate": 1.3237385429127767e-05, "loss": 0.6052, "step": 3398 }, { "epoch": 0.41, "learning_rate": 1.3233658489088306e-05, "loss": 0.5928, "step": 3399 }, { "epoch": 0.41, "learning_rate": 1.3229931047372527e-05, "loss": 0.5559, "step": 3400 }, { "epoch": 0.41, "learning_rate": 1.3226203104558715e-05, "loss": 0.571, "step": 3401 }, { "epoch": 0.41, "learning_rate": 1.3222474661225226e-05, "loss": 0.5278, "step": 3402 }, { "epoch": 0.41, "learning_rate": 1.32187457179505e-05, "loss": 0.5731, "step": 3403 }, { "epoch": 0.41, "learning_rate": 1.3215016275313054e-05, "loss": 0.6102, "step": 3404 }, { "epoch": 0.41, "learning_rate": 1.3211286333891477e-05, "loss": 0.5361, "step": 3405 }, { "epoch": 0.41, "learning_rate": 1.3207555894264447e-05, "loss": 0.612, "step": 3406 }, { "epoch": 0.41, "learning_rate": 1.3203824957010707e-05, "loss": 0.5562, "step": 3407 }, { "epoch": 0.41, "learning_rate": 1.3200093522709084e-05, "loss": 0.571, "step": 3408 }, { "epoch": 0.41, "learning_rate": 1.3196361591938477e-05, "loss": 0.5466, "step": 3409 }, { "epoch": 0.41, "learning_rate": 1.3192629165277865e-05, "loss": 0.5782, "step": 3410 }, { "epoch": 0.41, "learning_rate": 1.3188896243306313e-05, "loss": 0.5237, "step": 3411 }, { "epoch": 0.41, "learning_rate": 1.3185162826602944e-05, "loss": 0.6041, "step": 3412 }, { "epoch": 0.42, "learning_rate": 1.3181428915746973e-05, "loss": 0.5926, "step": 3413 }, { "epoch": 0.42, "learning_rate": 1.3177694511317682e-05, "loss": 0.5943, "step": 3414 }, { "epoch": 0.42, "learning_rate": 1.3173959613894443e-05, "loss": 0.5555, "step": 3415 }, { "epoch": 0.42, "learning_rate": 1.3170224224056685e-05, "loss": 0.5617, "step": 3416 }, { "epoch": 0.42, "learning_rate": 1.316648834238393e-05, "loss": 0.609, "step": 3417 }, { "epoch": 0.42, "learning_rate": 1.3162751969455769e-05, "loss": 0.6013, "step": 3418 }, { "epoch": 0.42, "learning_rate": 1.3159015105851873e-05, "loss": 0.5473, "step": 3419 }, { "epoch": 0.42, "learning_rate": 1.315527775215198e-05, "loss": 0.5541, "step": 3420 }, { "epoch": 0.42, "learning_rate": 1.3151539908935917e-05, "loss": 0.527, "step": 3421 }, { "epoch": 0.42, "learning_rate": 1.3147801576783576e-05, "loss": 0.6281, "step": 3422 }, { "epoch": 0.42, "learning_rate": 1.3144062756274933e-05, "loss": 0.5895, "step": 3423 }, { "epoch": 0.42, "learning_rate": 1.3140323447990035e-05, "loss": 0.5956, "step": 3424 }, { "epoch": 0.42, "learning_rate": 1.3136583652509006e-05, "loss": 0.5916, "step": 3425 }, { "epoch": 0.42, "learning_rate": 1.3132843370412043e-05, "loss": 0.6202, "step": 3426 }, { "epoch": 0.42, "learning_rate": 1.3129102602279424e-05, "loss": 0.5379, "step": 3427 }, { "epoch": 0.42, "learning_rate": 1.3125361348691499e-05, "loss": 0.5457, "step": 3428 }, { "epoch": 0.42, "learning_rate": 1.312161961022869e-05, "loss": 0.5196, "step": 3429 }, { "epoch": 0.42, "learning_rate": 1.3117877387471506e-05, "loss": 0.5018, "step": 3430 }, { "epoch": 0.42, "learning_rate": 1.3114134681000514e-05, "loss": 0.5437, "step": 3431 }, { "epoch": 0.42, "learning_rate": 1.311039149139637e-05, "loss": 0.5737, "step": 3432 }, { "epoch": 0.42, "learning_rate": 1.31066478192398e-05, "loss": 0.5703, "step": 3433 }, { "epoch": 0.42, "learning_rate": 1.31029036651116e-05, "loss": 0.5662, "step": 3434 }, { "epoch": 0.42, "learning_rate": 1.3099159029592652e-05, "loss": 0.5897, "step": 3435 }, { "epoch": 0.42, "learning_rate": 1.3095413913263904e-05, "loss": 0.5436, "step": 3436 }, { "epoch": 0.42, "learning_rate": 1.3091668316706379e-05, "loss": 0.5879, "step": 3437 }, { "epoch": 0.42, "learning_rate": 1.308792224050118e-05, "loss": 0.5644, "step": 3438 }, { "epoch": 0.42, "learning_rate": 1.3084175685229477e-05, "loss": 0.5827, "step": 3439 }, { "epoch": 0.42, "learning_rate": 1.3080428651472519e-05, "loss": 0.5724, "step": 3440 }, { "epoch": 0.42, "learning_rate": 1.3076681139811634e-05, "loss": 0.5438, "step": 3441 }, { "epoch": 0.42, "learning_rate": 1.307293315082821e-05, "loss": 0.5224, "step": 3442 }, { "epoch": 0.42, "learning_rate": 1.3069184685103722e-05, "loss": 0.5713, "step": 3443 }, { "epoch": 0.42, "learning_rate": 1.3065435743219716e-05, "loss": 0.5492, "step": 3444 }, { "epoch": 0.42, "learning_rate": 1.3061686325757808e-05, "loss": 0.5365, "step": 3445 }, { "epoch": 0.42, "learning_rate": 1.305793643329969e-05, "loss": 0.5304, "step": 3446 }, { "epoch": 0.42, "learning_rate": 1.305418606642713e-05, "loss": 0.5872, "step": 3447 }, { "epoch": 0.42, "learning_rate": 1.3050435225721965e-05, "loss": 0.5897, "step": 3448 }, { "epoch": 0.42, "learning_rate": 1.3046683911766107e-05, "loss": 0.6107, "step": 3449 }, { "epoch": 0.42, "learning_rate": 1.3042932125141548e-05, "loss": 0.6047, "step": 3450 }, { "epoch": 0.42, "learning_rate": 1.3039179866430345e-05, "loss": 0.5595, "step": 3451 }, { "epoch": 0.42, "learning_rate": 1.3035427136214627e-05, "loss": 0.5538, "step": 3452 }, { "epoch": 0.42, "learning_rate": 1.3031673935076603e-05, "loss": 0.5792, "step": 3453 }, { "epoch": 0.42, "learning_rate": 1.3027920263598556e-05, "loss": 0.5229, "step": 3454 }, { "epoch": 0.42, "learning_rate": 1.3024166122362835e-05, "loss": 0.617, "step": 3455 }, { "epoch": 0.42, "learning_rate": 1.3020411511951863e-05, "loss": 0.5335, "step": 3456 }, { "epoch": 0.42, "learning_rate": 1.301665643294814e-05, "loss": 0.5587, "step": 3457 }, { "epoch": 0.42, "learning_rate": 1.3012900885934238e-05, "loss": 0.5644, "step": 3458 }, { "epoch": 0.42, "learning_rate": 1.3009144871492798e-05, "loss": 0.5254, "step": 3459 }, { "epoch": 0.42, "learning_rate": 1.3005388390206535e-05, "loss": 0.5655, "step": 3460 }, { "epoch": 0.42, "learning_rate": 1.3001631442658242e-05, "loss": 0.572, "step": 3461 }, { "epoch": 0.42, "learning_rate": 1.2997874029430773e-05, "loss": 0.5375, "step": 3462 }, { "epoch": 0.42, "learning_rate": 1.2994116151107061e-05, "loss": 0.6141, "step": 3463 }, { "epoch": 0.42, "learning_rate": 1.2990357808270118e-05, "loss": 0.625, "step": 3464 }, { "epoch": 0.42, "learning_rate": 1.2986599001503013e-05, "loss": 0.5368, "step": 3465 }, { "epoch": 0.42, "learning_rate": 1.2982839731388898e-05, "loss": 0.5228, "step": 3466 }, { "epoch": 0.42, "learning_rate": 1.2979079998510997e-05, "loss": 0.5452, "step": 3467 }, { "epoch": 0.42, "learning_rate": 1.29753198034526e-05, "loss": 0.5686, "step": 3468 }, { "epoch": 0.42, "learning_rate": 1.2971559146797065e-05, "loss": 0.6007, "step": 3469 }, { "epoch": 0.42, "learning_rate": 1.2967798029127838e-05, "loss": 0.5815, "step": 3470 }, { "epoch": 0.42, "learning_rate": 1.296403645102842e-05, "loss": 0.572, "step": 3471 }, { "epoch": 0.42, "learning_rate": 1.2960274413082394e-05, "loss": 0.5759, "step": 3472 }, { "epoch": 0.42, "learning_rate": 1.2956511915873406e-05, "loss": 0.5892, "step": 3473 }, { "epoch": 0.42, "learning_rate": 1.2952748959985182e-05, "loss": 0.5262, "step": 3474 }, { "epoch": 0.42, "learning_rate": 1.2948985546001508e-05, "loss": 0.5333, "step": 3475 }, { "epoch": 0.42, "learning_rate": 1.2945221674506253e-05, "loss": 0.6172, "step": 3476 }, { "epoch": 0.42, "learning_rate": 1.2941457346083353e-05, "loss": 0.5732, "step": 3477 }, { "epoch": 0.42, "learning_rate": 1.2937692561316811e-05, "loss": 0.5928, "step": 3478 }, { "epoch": 0.42, "learning_rate": 1.2933927320790698e-05, "loss": 0.6193, "step": 3479 }, { "epoch": 0.42, "learning_rate": 1.293016162508917e-05, "loss": 0.5889, "step": 3480 }, { "epoch": 0.42, "learning_rate": 1.2926395474796443e-05, "loss": 0.5463, "step": 3481 }, { "epoch": 0.42, "learning_rate": 1.2922628870496795e-05, "loss": 0.545, "step": 3482 }, { "epoch": 0.42, "learning_rate": 1.29188618127746e-05, "loss": 0.5723, "step": 3483 }, { "epoch": 0.42, "learning_rate": 1.2915094302214279e-05, "loss": 0.5588, "step": 3484 }, { "epoch": 0.42, "learning_rate": 1.2911326339400333e-05, "loss": 0.5683, "step": 3485 }, { "epoch": 0.42, "learning_rate": 1.2907557924917324e-05, "loss": 0.6609, "step": 3486 }, { "epoch": 0.42, "learning_rate": 1.2903789059349903e-05, "loss": 0.5992, "step": 3487 }, { "epoch": 0.42, "learning_rate": 1.290001974328277e-05, "loss": 0.6145, "step": 3488 }, { "epoch": 0.42, "learning_rate": 1.289624997730071e-05, "loss": 0.5289, "step": 3489 }, { "epoch": 0.42, "learning_rate": 1.2892479761988574e-05, "loss": 0.5632, "step": 3490 }, { "epoch": 0.42, "learning_rate": 1.2888709097931273e-05, "loss": 0.5968, "step": 3491 }, { "epoch": 0.42, "learning_rate": 1.2884937985713802e-05, "loss": 0.5743, "step": 3492 }, { "epoch": 0.42, "learning_rate": 1.2881166425921214e-05, "loss": 0.4904, "step": 3493 }, { "epoch": 0.42, "learning_rate": 1.287739441913864e-05, "loss": 0.5596, "step": 3494 }, { "epoch": 0.43, "learning_rate": 1.2873621965951278e-05, "loss": 0.5983, "step": 3495 }, { "epoch": 0.43, "learning_rate": 1.286984906694439e-05, "loss": 0.5878, "step": 3496 }, { "epoch": 0.43, "learning_rate": 1.2866075722703317e-05, "loss": 0.554, "step": 3497 }, { "epoch": 0.43, "learning_rate": 1.2862301933813455e-05, "loss": 0.5613, "step": 3498 }, { "epoch": 0.43, "learning_rate": 1.2858527700860281e-05, "loss": 0.6043, "step": 3499 }, { "epoch": 0.43, "learning_rate": 1.285475302442934e-05, "loss": 0.5829, "step": 3500 }, { "epoch": 0.43, "learning_rate": 1.2850977905106241e-05, "loss": 0.5566, "step": 3501 }, { "epoch": 0.43, "learning_rate": 1.2847202343476663e-05, "loss": 0.6064, "step": 3502 }, { "epoch": 0.43, "learning_rate": 1.2843426340126355e-05, "loss": 0.496, "step": 3503 }, { "epoch": 0.43, "learning_rate": 1.2839649895641133e-05, "loss": 0.6307, "step": 3504 }, { "epoch": 0.43, "learning_rate": 1.2835873010606883e-05, "loss": 0.6, "step": 3505 }, { "epoch": 0.43, "learning_rate": 1.2832095685609555e-05, "loss": 0.486, "step": 3506 }, { "epoch": 0.43, "learning_rate": 1.2828317921235178e-05, "loss": 0.5484, "step": 3507 }, { "epoch": 0.43, "learning_rate": 1.2824539718069835e-05, "loss": 0.5299, "step": 3508 }, { "epoch": 0.43, "learning_rate": 1.2820761076699688e-05, "loss": 0.5821, "step": 3509 }, { "epoch": 0.43, "learning_rate": 1.2816981997710967e-05, "loss": 0.5276, "step": 3510 }, { "epoch": 0.43, "learning_rate": 1.2813202481689954e-05, "loss": 0.5515, "step": 3511 }, { "epoch": 0.43, "learning_rate": 1.2809422529223019e-05, "loss": 0.6369, "step": 3512 }, { "epoch": 0.43, "learning_rate": 1.2805642140896595e-05, "loss": 0.506, "step": 3513 }, { "epoch": 0.43, "learning_rate": 1.280186131729717e-05, "loss": 0.5874, "step": 3514 }, { "epoch": 0.43, "learning_rate": 1.2798080059011317e-05, "loss": 0.5548, "step": 3515 }, { "epoch": 0.43, "learning_rate": 1.2794298366625662e-05, "loss": 0.5606, "step": 3516 }, { "epoch": 0.43, "learning_rate": 1.2790516240726907e-05, "loss": 0.5092, "step": 3517 }, { "epoch": 0.43, "learning_rate": 1.2786733681901815e-05, "loss": 0.553, "step": 3518 }, { "epoch": 0.43, "learning_rate": 1.2782950690737228e-05, "loss": 0.6057, "step": 3519 }, { "epoch": 0.43, "learning_rate": 1.277916726782004e-05, "loss": 0.6177, "step": 3520 }, { "epoch": 0.43, "learning_rate": 1.277538341373722e-05, "loss": 0.5453, "step": 3521 }, { "epoch": 0.43, "learning_rate": 1.2771599129075804e-05, "loss": 0.5611, "step": 3522 }, { "epoch": 0.43, "learning_rate": 1.2767814414422897e-05, "loss": 0.518, "step": 3523 }, { "epoch": 0.43, "learning_rate": 1.2764029270365657e-05, "loss": 0.5736, "step": 3524 }, { "epoch": 0.43, "learning_rate": 1.2760243697491328e-05, "loss": 0.564, "step": 3525 }, { "epoch": 0.43, "learning_rate": 1.2756457696387212e-05, "loss": 0.6011, "step": 3526 }, { "epoch": 0.43, "learning_rate": 1.2752671267640671e-05, "loss": 0.5546, "step": 3527 }, { "epoch": 0.43, "learning_rate": 1.2748884411839139e-05, "loss": 0.546, "step": 3528 }, { "epoch": 0.43, "learning_rate": 1.2745097129570123e-05, "loss": 0.5895, "step": 3529 }, { "epoch": 0.43, "learning_rate": 1.2741309421421186e-05, "loss": 0.5109, "step": 3530 }, { "epoch": 0.43, "learning_rate": 1.2737521287979959e-05, "loss": 0.6248, "step": 3531 }, { "epoch": 0.43, "learning_rate": 1.2733732729834142e-05, "loss": 0.6061, "step": 3532 }, { "epoch": 0.43, "learning_rate": 1.2729943747571505e-05, "loss": 0.531, "step": 3533 }, { "epoch": 0.43, "learning_rate": 1.2726154341779869e-05, "loss": 0.5734, "step": 3534 }, { "epoch": 0.43, "learning_rate": 1.2722364513047135e-05, "loss": 0.4969, "step": 3535 }, { "epoch": 0.43, "learning_rate": 1.2718574261961262e-05, "loss": 0.5284, "step": 3536 }, { "epoch": 0.43, "learning_rate": 1.2714783589110282e-05, "loss": 0.5758, "step": 3537 }, { "epoch": 0.43, "learning_rate": 1.2710992495082285e-05, "loss": 0.5291, "step": 3538 }, { "epoch": 0.43, "learning_rate": 1.2707200980465428e-05, "loss": 0.5738, "step": 3539 }, { "epoch": 0.43, "learning_rate": 1.2703409045847936e-05, "loss": 0.5746, "step": 3540 }, { "epoch": 0.43, "learning_rate": 1.2699616691818097e-05, "loss": 0.5238, "step": 3541 }, { "epoch": 0.43, "learning_rate": 1.269582391896426e-05, "loss": 0.5654, "step": 3542 }, { "epoch": 0.43, "learning_rate": 1.2692030727874853e-05, "loss": 0.5068, "step": 3543 }, { "epoch": 0.43, "learning_rate": 1.2688237119138351e-05, "loss": 0.5584, "step": 3544 }, { "epoch": 0.43, "learning_rate": 1.2684443093343306e-05, "loss": 0.5629, "step": 3545 }, { "epoch": 0.43, "learning_rate": 1.268064865107833e-05, "loss": 0.583, "step": 3546 }, { "epoch": 0.43, "learning_rate": 1.2676853792932103e-05, "loss": 0.5886, "step": 3547 }, { "epoch": 0.43, "learning_rate": 1.2673058519493358e-05, "loss": 0.547, "step": 3548 }, { "epoch": 0.43, "learning_rate": 1.2669262831350912e-05, "loss": 0.5737, "step": 3549 }, { "epoch": 0.43, "learning_rate": 1.266546672909363e-05, "loss": 0.5589, "step": 3550 }, { "epoch": 0.43, "learning_rate": 1.2661670213310449e-05, "loss": 0.5867, "step": 3551 }, { "epoch": 0.43, "learning_rate": 1.2657873284590367e-05, "loss": 0.5264, "step": 3552 }, { "epoch": 0.43, "learning_rate": 1.2654075943522451e-05, "loss": 0.5458, "step": 3553 }, { "epoch": 0.43, "learning_rate": 1.2650278190695817e-05, "loss": 0.6053, "step": 3554 }, { "epoch": 0.43, "learning_rate": 1.264648002669967e-05, "loss": 0.563, "step": 3555 }, { "epoch": 0.43, "learning_rate": 1.264268145212326e-05, "loss": 0.5965, "step": 3556 }, { "epoch": 0.43, "learning_rate": 1.26388824675559e-05, "loss": 0.624, "step": 3557 }, { "epoch": 0.43, "learning_rate": 1.263508307358698e-05, "loss": 0.5576, "step": 3558 }, { "epoch": 0.43, "learning_rate": 1.263128327080594e-05, "loss": 0.5497, "step": 3559 }, { "epoch": 0.43, "learning_rate": 1.262748305980229e-05, "loss": 0.5425, "step": 3560 }, { "epoch": 0.43, "learning_rate": 1.2623682441165608e-05, "loss": 0.5507, "step": 3561 }, { "epoch": 0.43, "learning_rate": 1.2619881415485523e-05, "loss": 0.594, "step": 3562 }, { "epoch": 0.43, "learning_rate": 1.2616079983351736e-05, "loss": 0.6334, "step": 3563 }, { "epoch": 0.43, "learning_rate": 1.2612278145354009e-05, "loss": 0.5676, "step": 3564 }, { "epoch": 0.43, "learning_rate": 1.2608475902082167e-05, "loss": 0.5378, "step": 3565 }, { "epoch": 0.43, "learning_rate": 1.2604673254126096e-05, "loss": 0.563, "step": 3566 }, { "epoch": 0.43, "learning_rate": 1.260087020207575e-05, "loss": 0.5655, "step": 3567 }, { "epoch": 0.43, "learning_rate": 1.259706674652114e-05, "loss": 0.5326, "step": 3568 }, { "epoch": 0.43, "learning_rate": 1.2593262888052342e-05, "loss": 0.5534, "step": 3569 }, { "epoch": 0.43, "learning_rate": 1.2589458627259497e-05, "loss": 0.5743, "step": 3570 }, { "epoch": 0.43, "learning_rate": 1.2585653964732799e-05, "loss": 0.5134, "step": 3571 }, { "epoch": 0.43, "learning_rate": 1.2581848901062514e-05, "loss": 0.5446, "step": 3572 }, { "epoch": 0.43, "learning_rate": 1.2578043436838972e-05, "loss": 0.5307, "step": 3573 }, { "epoch": 0.43, "learning_rate": 1.2574237572652555e-05, "loss": 0.5768, "step": 3574 }, { "epoch": 0.43, "learning_rate": 1.2570431309093714e-05, "loss": 0.5673, "step": 3575 }, { "epoch": 0.43, "learning_rate": 1.2566624646752966e-05, "loss": 0.5288, "step": 3576 }, { "epoch": 0.43, "learning_rate": 1.2562817586220872e-05, "loss": 0.5765, "step": 3577 }, { "epoch": 0.44, "learning_rate": 1.2559010128088073e-05, "loss": 0.6212, "step": 3578 }, { "epoch": 0.44, "learning_rate": 1.2555202272945273e-05, "loss": 0.5757, "step": 3579 }, { "epoch": 0.44, "learning_rate": 1.2551394021383221e-05, "loss": 0.5083, "step": 3580 }, { "epoch": 0.44, "learning_rate": 1.254758537399274e-05, "loss": 0.515, "step": 3581 }, { "epoch": 0.44, "learning_rate": 1.2543776331364713e-05, "loss": 0.5236, "step": 3582 }, { "epoch": 0.44, "learning_rate": 1.2539966894090082e-05, "loss": 0.6372, "step": 3583 }, { "epoch": 0.44, "learning_rate": 1.2536157062759842e-05, "loss": 0.5573, "step": 3584 }, { "epoch": 0.44, "learning_rate": 1.2532346837965073e-05, "loss": 0.5233, "step": 3585 }, { "epoch": 0.44, "learning_rate": 1.2528536220296892e-05, "loss": 0.5746, "step": 3586 }, { "epoch": 0.44, "learning_rate": 1.2524725210346484e-05, "loss": 0.4876, "step": 3587 }, { "epoch": 0.44, "learning_rate": 1.2520913808705104e-05, "loss": 0.5757, "step": 3588 }, { "epoch": 0.44, "learning_rate": 1.2517102015964058e-05, "loss": 0.4928, "step": 3589 }, { "epoch": 0.44, "learning_rate": 1.251328983271471e-05, "loss": 0.5482, "step": 3590 }, { "epoch": 0.44, "learning_rate": 1.2509477259548495e-05, "loss": 0.4955, "step": 3591 }, { "epoch": 0.44, "learning_rate": 1.2505664297056904e-05, "loss": 0.5829, "step": 3592 }, { "epoch": 0.44, "learning_rate": 1.2501850945831486e-05, "loss": 0.5813, "step": 3593 }, { "epoch": 0.44, "learning_rate": 1.2498037206463848e-05, "loss": 0.5842, "step": 3594 }, { "epoch": 0.44, "learning_rate": 1.2494223079545668e-05, "loss": 0.5389, "step": 3595 }, { "epoch": 0.44, "learning_rate": 1.2490408565668677e-05, "loss": 0.5835, "step": 3596 }, { "epoch": 0.44, "learning_rate": 1.248659366542466e-05, "loss": 0.5773, "step": 3597 }, { "epoch": 0.44, "learning_rate": 1.2482778379405473e-05, "loss": 0.582, "step": 3598 }, { "epoch": 0.44, "learning_rate": 1.247896270820303e-05, "loss": 0.577, "step": 3599 }, { "epoch": 0.44, "learning_rate": 1.2475146652409295e-05, "loss": 0.583, "step": 3600 }, { "epoch": 0.44, "learning_rate": 1.2471330212616302e-05, "loss": 0.5273, "step": 3601 }, { "epoch": 0.44, "learning_rate": 1.2467513389416141e-05, "loss": 0.5074, "step": 3602 }, { "epoch": 0.44, "learning_rate": 1.2463696183400964e-05, "loss": 0.5458, "step": 3603 }, { "epoch": 0.44, "learning_rate": 1.2459878595162978e-05, "loss": 0.532, "step": 3604 }, { "epoch": 0.44, "learning_rate": 1.245606062529445e-05, "loss": 0.5488, "step": 3605 }, { "epoch": 0.44, "learning_rate": 1.245224227438771e-05, "loss": 0.5935, "step": 3606 }, { "epoch": 0.44, "learning_rate": 1.2448423543035142e-05, "loss": 0.5164, "step": 3607 }, { "epoch": 0.44, "learning_rate": 1.2444604431829193e-05, "loss": 0.5192, "step": 3608 }, { "epoch": 0.44, "learning_rate": 1.2440784941362369e-05, "loss": 0.5618, "step": 3609 }, { "epoch": 0.44, "learning_rate": 1.2436965072227232e-05, "loss": 0.5823, "step": 3610 }, { "epoch": 0.44, "learning_rate": 1.2433144825016403e-05, "loss": 0.6465, "step": 3611 }, { "epoch": 0.44, "learning_rate": 1.242932420032257e-05, "loss": 0.564, "step": 3612 }, { "epoch": 0.44, "learning_rate": 1.2425503198738463e-05, "loss": 0.574, "step": 3613 }, { "epoch": 0.44, "learning_rate": 1.2421681820856881e-05, "loss": 0.5283, "step": 3614 }, { "epoch": 0.44, "learning_rate": 1.2417860067270688e-05, "loss": 0.5739, "step": 3615 }, { "epoch": 0.44, "learning_rate": 1.2414037938572792e-05, "loss": 0.5824, "step": 3616 }, { "epoch": 0.44, "learning_rate": 1.2410215435356166e-05, "loss": 0.5749, "step": 3617 }, { "epoch": 0.44, "learning_rate": 1.2406392558213845e-05, "loss": 0.5448, "step": 3618 }, { "epoch": 0.44, "learning_rate": 1.2402569307738914e-05, "loss": 0.5248, "step": 3619 }, { "epoch": 0.44, "learning_rate": 1.239874568452452e-05, "loss": 0.5513, "step": 3620 }, { "epoch": 0.44, "learning_rate": 1.239492168916387e-05, "loss": 0.5203, "step": 3621 }, { "epoch": 0.44, "learning_rate": 1.2391097322250228e-05, "loss": 0.5236, "step": 3622 }, { "epoch": 0.44, "learning_rate": 1.2387272584376906e-05, "loss": 0.5518, "step": 3623 }, { "epoch": 0.44, "learning_rate": 1.2383447476137289e-05, "loss": 0.5711, "step": 3624 }, { "epoch": 0.44, "learning_rate": 1.237962199812481e-05, "loss": 0.562, "step": 3625 }, { "epoch": 0.44, "learning_rate": 1.237579615093296e-05, "loss": 0.6263, "step": 3626 }, { "epoch": 0.44, "learning_rate": 1.237196993515529e-05, "loss": 0.5455, "step": 3627 }, { "epoch": 0.44, "learning_rate": 1.2368143351385408e-05, "loss": 0.6063, "step": 3628 }, { "epoch": 0.44, "learning_rate": 1.2364316400216976e-05, "loss": 0.5279, "step": 3629 }, { "epoch": 0.44, "learning_rate": 1.2360489082243714e-05, "loss": 0.5519, "step": 3630 }, { "epoch": 0.44, "learning_rate": 1.2356661398059405e-05, "loss": 0.545, "step": 3631 }, { "epoch": 0.44, "learning_rate": 1.2352833348257876e-05, "loss": 0.5599, "step": 3632 }, { "epoch": 0.44, "learning_rate": 1.2349004933433024e-05, "loss": 0.541, "step": 3633 }, { "epoch": 0.44, "learning_rate": 1.2345176154178795e-05, "loss": 0.5111, "step": 3634 }, { "epoch": 0.44, "learning_rate": 1.2341347011089195e-05, "loss": 0.5524, "step": 3635 }, { "epoch": 0.44, "learning_rate": 1.2337517504758285e-05, "loss": 0.6071, "step": 3636 }, { "epoch": 0.44, "learning_rate": 1.2333687635780179e-05, "loss": 0.5976, "step": 3637 }, { "epoch": 0.44, "learning_rate": 1.2329857404749052e-05, "loss": 0.5699, "step": 3638 }, { "epoch": 0.44, "learning_rate": 1.2326026812259138e-05, "loss": 0.5704, "step": 3639 }, { "epoch": 0.44, "learning_rate": 1.2322195858904717e-05, "loss": 0.5453, "step": 3640 }, { "epoch": 0.44, "learning_rate": 1.2318364545280135e-05, "loss": 0.6079, "step": 3641 }, { "epoch": 0.44, "learning_rate": 1.231453287197979e-05, "loss": 0.5787, "step": 3642 }, { "epoch": 0.44, "learning_rate": 1.2310700839598133e-05, "loss": 0.59, "step": 3643 }, { "epoch": 0.44, "learning_rate": 1.230686844872967e-05, "loss": 0.6034, "step": 3644 }, { "epoch": 0.44, "learning_rate": 1.2303035699968974e-05, "loss": 0.6115, "step": 3645 }, { "epoch": 0.44, "learning_rate": 1.2299202593910659e-05, "loss": 0.5267, "step": 3646 }, { "epoch": 0.44, "learning_rate": 1.2295369131149406e-05, "loss": 0.5706, "step": 3647 }, { "epoch": 0.44, "learning_rate": 1.2291535312279943e-05, "loss": 0.5352, "step": 3648 }, { "epoch": 0.44, "learning_rate": 1.2287701137897054e-05, "loss": 0.5637, "step": 3649 }, { "epoch": 0.44, "learning_rate": 1.2283866608595583e-05, "loss": 0.6423, "step": 3650 }, { "epoch": 0.44, "learning_rate": 1.2280031724970432e-05, "loss": 0.6237, "step": 3651 }, { "epoch": 0.44, "learning_rate": 1.2276196487616543e-05, "loss": 0.5947, "step": 3652 }, { "epoch": 0.44, "learning_rate": 1.2272360897128927e-05, "loss": 0.5919, "step": 3653 }, { "epoch": 0.44, "learning_rate": 1.2268524954102647e-05, "loss": 0.6015, "step": 3654 }, { "epoch": 0.44, "learning_rate": 1.2264688659132817e-05, "loss": 0.556, "step": 3655 }, { "epoch": 0.44, "learning_rate": 1.2260852012814606e-05, "loss": 0.5163, "step": 3656 }, { "epoch": 0.44, "learning_rate": 1.2257015015743243e-05, "loss": 0.6055, "step": 3657 }, { "epoch": 0.44, "learning_rate": 1.2253177668514006e-05, "loss": 0.5211, "step": 3658 }, { "epoch": 0.44, "learning_rate": 1.2249339971722228e-05, "loss": 0.4919, "step": 3659 }, { "epoch": 0.45, "learning_rate": 1.2245501925963295e-05, "loss": 0.5256, "step": 3660 }, { "epoch": 0.45, "learning_rate": 1.224166353183265e-05, "loss": 0.5307, "step": 3661 }, { "epoch": 0.45, "learning_rate": 1.2237824789925793e-05, "loss": 0.6353, "step": 3662 }, { "epoch": 0.45, "learning_rate": 1.2233985700838273e-05, "loss": 0.6084, "step": 3663 }, { "epoch": 0.45, "learning_rate": 1.223014626516569e-05, "loss": 0.5544, "step": 3664 }, { "epoch": 0.45, "learning_rate": 1.222630648350371e-05, "loss": 0.5061, "step": 3665 }, { "epoch": 0.45, "learning_rate": 1.2222466356448034e-05, "loss": 0.5345, "step": 3666 }, { "epoch": 0.45, "learning_rate": 1.2218625884594432e-05, "loss": 0.5893, "step": 3667 }, { "epoch": 0.45, "learning_rate": 1.2214785068538726e-05, "loss": 0.5896, "step": 3668 }, { "epoch": 0.45, "learning_rate": 1.2210943908876782e-05, "loss": 0.5968, "step": 3669 }, { "epoch": 0.45, "learning_rate": 1.2207102406204529e-05, "loss": 0.6299, "step": 3670 }, { "epoch": 0.45, "learning_rate": 1.220326056111795e-05, "loss": 0.6008, "step": 3671 }, { "epoch": 0.45, "learning_rate": 1.2199418374213064e-05, "loss": 0.5549, "step": 3672 }, { "epoch": 0.45, "learning_rate": 1.2195575846085963e-05, "loss": 0.5477, "step": 3673 }, { "epoch": 0.45, "learning_rate": 1.2191732977332786e-05, "loss": 0.5839, "step": 3674 }, { "epoch": 0.45, "learning_rate": 1.2187889768549723e-05, "loss": 0.6298, "step": 3675 }, { "epoch": 0.45, "learning_rate": 1.2184046220333015e-05, "loss": 0.5455, "step": 3676 }, { "epoch": 0.45, "learning_rate": 1.2180202333278958e-05, "loss": 0.5492, "step": 3677 }, { "epoch": 0.45, "learning_rate": 1.2176358107983903e-05, "loss": 0.4879, "step": 3678 }, { "epoch": 0.45, "learning_rate": 1.217251354504425e-05, "loss": 0.5833, "step": 3679 }, { "epoch": 0.45, "learning_rate": 1.2168668645056447e-05, "loss": 0.5901, "step": 3680 }, { "epoch": 0.45, "learning_rate": 1.2164823408617009e-05, "loss": 0.5613, "step": 3681 }, { "epoch": 0.45, "learning_rate": 1.2160977836322485e-05, "loss": 0.5922, "step": 3682 }, { "epoch": 0.45, "learning_rate": 1.215713192876949e-05, "loss": 0.526, "step": 3683 }, { "epoch": 0.45, "learning_rate": 1.2153285686554685e-05, "loss": 0.5612, "step": 3684 }, { "epoch": 0.45, "learning_rate": 1.2149439110274785e-05, "loss": 0.5683, "step": 3685 }, { "epoch": 0.45, "learning_rate": 1.2145592200526549e-05, "loss": 0.5902, "step": 3686 }, { "epoch": 0.45, "learning_rate": 1.2141744957906805e-05, "loss": 0.6158, "step": 3687 }, { "epoch": 0.45, "learning_rate": 1.2137897383012411e-05, "loss": 0.5788, "step": 3688 }, { "epoch": 0.45, "learning_rate": 1.2134049476440297e-05, "loss": 0.5498, "step": 3689 }, { "epoch": 0.45, "learning_rate": 1.2130201238787426e-05, "loss": 0.5964, "step": 3690 }, { "epoch": 0.45, "learning_rate": 1.2126352670650833e-05, "loss": 0.5958, "step": 3691 }, { "epoch": 0.45, "learning_rate": 1.212250377262758e-05, "loss": 0.5692, "step": 3692 }, { "epoch": 0.45, "learning_rate": 1.2118654545314801e-05, "loss": 0.6073, "step": 3693 }, { "epoch": 0.45, "learning_rate": 1.2114804989309672e-05, "loss": 0.5747, "step": 3694 }, { "epoch": 0.45, "learning_rate": 1.2110955105209417e-05, "loss": 0.5144, "step": 3695 }, { "epoch": 0.45, "learning_rate": 1.2107104893611318e-05, "loss": 0.5397, "step": 3696 }, { "epoch": 0.45, "learning_rate": 1.2103254355112704e-05, "loss": 0.5591, "step": 3697 }, { "epoch": 0.45, "learning_rate": 1.2099403490310955e-05, "loss": 0.5582, "step": 3698 }, { "epoch": 0.45, "learning_rate": 1.2095552299803502e-05, "loss": 0.5548, "step": 3699 }, { "epoch": 0.45, "learning_rate": 1.2091700784187825e-05, "loss": 0.5816, "step": 3700 }, { "epoch": 0.45, "learning_rate": 1.2087848944061462e-05, "loss": 0.657, "step": 3701 }, { "epoch": 0.45, "learning_rate": 1.208399678002199e-05, "loss": 0.5363, "step": 3702 }, { "epoch": 0.45, "learning_rate": 1.208014429266704e-05, "loss": 0.6292, "step": 3703 }, { "epoch": 0.45, "learning_rate": 1.2076291482594298e-05, "loss": 0.5885, "step": 3704 }, { "epoch": 0.45, "learning_rate": 1.2072438350401496e-05, "loss": 0.5297, "step": 3705 }, { "epoch": 0.45, "learning_rate": 1.2068584896686415e-05, "loss": 0.5955, "step": 3706 }, { "epoch": 0.45, "learning_rate": 1.2064731122046893e-05, "loss": 0.5959, "step": 3707 }, { "epoch": 0.45, "learning_rate": 1.2060877027080806e-05, "loss": 0.5509, "step": 3708 }, { "epoch": 0.45, "learning_rate": 1.205702261238609e-05, "loss": 0.5256, "step": 3709 }, { "epoch": 0.45, "learning_rate": 1.2053167878560723e-05, "loss": 0.5343, "step": 3710 }, { "epoch": 0.45, "learning_rate": 1.204931282620274e-05, "loss": 0.5416, "step": 3711 }, { "epoch": 0.45, "learning_rate": 1.204545745591022e-05, "loss": 0.5364, "step": 3712 }, { "epoch": 0.45, "learning_rate": 1.204160176828129e-05, "loss": 0.5768, "step": 3713 }, { "epoch": 0.45, "learning_rate": 1.2037745763914137e-05, "loss": 0.627, "step": 3714 }, { "epoch": 0.45, "learning_rate": 1.2033889443406982e-05, "loss": 0.5373, "step": 3715 }, { "epoch": 0.45, "learning_rate": 1.2030032807358104e-05, "loss": 0.5639, "step": 3716 }, { "epoch": 0.45, "learning_rate": 1.2026175856365834e-05, "loss": 0.5642, "step": 3717 }, { "epoch": 0.45, "learning_rate": 1.2022318591028541e-05, "loss": 0.5535, "step": 3718 }, { "epoch": 0.45, "learning_rate": 1.201846101194465e-05, "loss": 0.5653, "step": 3719 }, { "epoch": 0.45, "learning_rate": 1.2014603119712639e-05, "loss": 0.5949, "step": 3720 }, { "epoch": 0.45, "learning_rate": 1.2010744914931022e-05, "loss": 0.5401, "step": 3721 }, { "epoch": 0.45, "learning_rate": 1.2006886398198371e-05, "loss": 0.5927, "step": 3722 }, { "epoch": 0.45, "learning_rate": 1.2003027570113308e-05, "loss": 0.5704, "step": 3723 }, { "epoch": 0.45, "learning_rate": 1.1999168431274497e-05, "loss": 0.5053, "step": 3724 }, { "epoch": 0.45, "learning_rate": 1.1995308982280648e-05, "loss": 0.6083, "step": 3725 }, { "epoch": 0.45, "learning_rate": 1.1991449223730532e-05, "loss": 0.5445, "step": 3726 }, { "epoch": 0.45, "learning_rate": 1.1987589156222954e-05, "loss": 0.5345, "step": 3727 }, { "epoch": 0.45, "learning_rate": 1.1983728780356772e-05, "loss": 0.534, "step": 3728 }, { "epoch": 0.45, "learning_rate": 1.1979868096730896e-05, "loss": 0.5783, "step": 3729 }, { "epoch": 0.45, "learning_rate": 1.197600710594428e-05, "loss": 0.4947, "step": 3730 }, { "epoch": 0.45, "learning_rate": 1.1972145808595924e-05, "loss": 0.604, "step": 3731 }, { "epoch": 0.45, "learning_rate": 1.1968284205284877e-05, "loss": 0.5245, "step": 3732 }, { "epoch": 0.45, "learning_rate": 1.1964422296610239e-05, "loss": 0.6031, "step": 3733 }, { "epoch": 0.45, "learning_rate": 1.196056008317115e-05, "loss": 0.5145, "step": 3734 }, { "epoch": 0.45, "learning_rate": 1.1956697565566808e-05, "loss": 0.5687, "step": 3735 }, { "epoch": 0.45, "learning_rate": 1.1952834744396447e-05, "loss": 0.5411, "step": 3736 }, { "epoch": 0.45, "learning_rate": 1.1948971620259353e-05, "loss": 0.5343, "step": 3737 }, { "epoch": 0.45, "learning_rate": 1.194510819375486e-05, "loss": 0.5924, "step": 3738 }, { "epoch": 0.45, "learning_rate": 1.1941244465482349e-05, "loss": 0.5917, "step": 3739 }, { "epoch": 0.45, "learning_rate": 1.1937380436041242e-05, "loss": 0.5408, "step": 3740 }, { "epoch": 0.45, "learning_rate": 1.1933516106031017e-05, "loss": 0.5821, "step": 3741 }, { "epoch": 0.46, "learning_rate": 1.1929651476051195e-05, "loss": 0.5993, "step": 3742 }, { "epoch": 0.46, "learning_rate": 1.1925786546701336e-05, "loss": 0.6047, "step": 3743 }, { "epoch": 0.46, "learning_rate": 1.192192131858106e-05, "loss": 0.5265, "step": 3744 }, { "epoch": 0.46, "learning_rate": 1.1918055792290022e-05, "loss": 0.5681, "step": 3745 }, { "epoch": 0.46, "learning_rate": 1.1914189968427927e-05, "loss": 0.6039, "step": 3746 }, { "epoch": 0.46, "learning_rate": 1.1910323847594533e-05, "loss": 0.6098, "step": 3747 }, { "epoch": 0.46, "learning_rate": 1.190645743038963e-05, "loss": 0.6115, "step": 3748 }, { "epoch": 0.46, "learning_rate": 1.1902590717413067e-05, "loss": 0.6181, "step": 3749 }, { "epoch": 0.46, "learning_rate": 1.1898723709264732e-05, "loss": 0.5328, "step": 3750 }, { "epoch": 0.46, "learning_rate": 1.1894856406544559e-05, "loss": 0.5654, "step": 3751 }, { "epoch": 0.46, "learning_rate": 1.1890988809852528e-05, "loss": 0.5664, "step": 3752 }, { "epoch": 0.46, "learning_rate": 1.1887120919788674e-05, "loss": 0.5621, "step": 3753 }, { "epoch": 0.46, "learning_rate": 1.1883252736953061e-05, "loss": 0.5601, "step": 3754 }, { "epoch": 0.46, "learning_rate": 1.1879384261945808e-05, "loss": 0.5791, "step": 3755 }, { "epoch": 0.46, "learning_rate": 1.187551549536708e-05, "loss": 0.5495, "step": 3756 }, { "epoch": 0.46, "learning_rate": 1.1871646437817088e-05, "loss": 0.56, "step": 3757 }, { "epoch": 0.46, "learning_rate": 1.1867777089896075e-05, "loss": 0.5732, "step": 3758 }, { "epoch": 0.46, "learning_rate": 1.186390745220435e-05, "loss": 0.4985, "step": 3759 }, { "epoch": 0.46, "learning_rate": 1.1860037525342256e-05, "loss": 0.5639, "step": 3760 }, { "epoch": 0.46, "learning_rate": 1.1856167309910177e-05, "loss": 0.5548, "step": 3761 }, { "epoch": 0.46, "learning_rate": 1.1852296806508544e-05, "loss": 0.5931, "step": 3762 }, { "epoch": 0.46, "learning_rate": 1.184842601573784e-05, "loss": 0.5126, "step": 3763 }, { "epoch": 0.46, "learning_rate": 1.1844554938198583e-05, "loss": 0.5863, "step": 3764 }, { "epoch": 0.46, "learning_rate": 1.1840683574491343e-05, "loss": 0.5555, "step": 3765 }, { "epoch": 0.46, "learning_rate": 1.1836811925216731e-05, "loss": 0.6082, "step": 3766 }, { "epoch": 0.46, "learning_rate": 1.1832939990975404e-05, "loss": 0.5981, "step": 3767 }, { "epoch": 0.46, "learning_rate": 1.1829067772368057e-05, "loss": 0.5293, "step": 3768 }, { "epoch": 0.46, "learning_rate": 1.1825195269995436e-05, "loss": 0.5617, "step": 3769 }, { "epoch": 0.46, "learning_rate": 1.1821322484458329e-05, "loss": 0.5066, "step": 3770 }, { "epoch": 0.46, "learning_rate": 1.1817449416357567e-05, "loss": 0.5216, "step": 3771 }, { "epoch": 0.46, "learning_rate": 1.1813576066294027e-05, "loss": 0.5002, "step": 3772 }, { "epoch": 0.46, "learning_rate": 1.1809702434868633e-05, "loss": 0.5601, "step": 3773 }, { "epoch": 0.46, "learning_rate": 1.1805828522682339e-05, "loss": 0.5299, "step": 3774 }, { "epoch": 0.46, "learning_rate": 1.1801954330336153e-05, "loss": 0.5864, "step": 3775 }, { "epoch": 0.46, "learning_rate": 1.1798079858431131e-05, "loss": 0.5406, "step": 3776 }, { "epoch": 0.46, "learning_rate": 1.1794205107568361e-05, "loss": 0.5424, "step": 3777 }, { "epoch": 0.46, "learning_rate": 1.1790330078348982e-05, "loss": 0.5383, "step": 3778 }, { "epoch": 0.46, "learning_rate": 1.1786454771374174e-05, "loss": 0.5731, "step": 3779 }, { "epoch": 0.46, "learning_rate": 1.178257918724516e-05, "loss": 0.5603, "step": 3780 }, { "epoch": 0.46, "learning_rate": 1.1778703326563203e-05, "loss": 0.5739, "step": 3781 }, { "epoch": 0.46, "learning_rate": 1.1774827189929613e-05, "loss": 0.5125, "step": 3782 }, { "epoch": 0.46, "learning_rate": 1.1770950777945748e-05, "loss": 0.5674, "step": 3783 }, { "epoch": 0.46, "learning_rate": 1.1767074091212995e-05, "loss": 0.5482, "step": 3784 }, { "epoch": 0.46, "learning_rate": 1.1763197130332795e-05, "loss": 0.5554, "step": 3785 }, { "epoch": 0.46, "learning_rate": 1.1759319895906623e-05, "loss": 0.563, "step": 3786 }, { "epoch": 0.46, "learning_rate": 1.1755442388536008e-05, "loss": 0.5532, "step": 3787 }, { "epoch": 0.46, "learning_rate": 1.1751564608822503e-05, "loss": 0.5903, "step": 3788 }, { "epoch": 0.46, "learning_rate": 1.1747686557367728e-05, "loss": 0.5128, "step": 3789 }, { "epoch": 0.46, "learning_rate": 1.1743808234773324e-05, "loss": 0.5585, "step": 3790 }, { "epoch": 0.46, "learning_rate": 1.1739929641640983e-05, "loss": 0.5666, "step": 3791 }, { "epoch": 0.46, "learning_rate": 1.1736050778572439e-05, "loss": 0.5328, "step": 3792 }, { "epoch": 0.46, "learning_rate": 1.1732171646169465e-05, "loss": 0.5974, "step": 3793 }, { "epoch": 0.46, "learning_rate": 1.1728292245033876e-05, "loss": 0.5063, "step": 3794 }, { "epoch": 0.46, "learning_rate": 1.1724412575767536e-05, "loss": 0.55, "step": 3795 }, { "epoch": 0.46, "learning_rate": 1.1720532638972341e-05, "loss": 0.5621, "step": 3796 }, { "epoch": 0.46, "learning_rate": 1.1716652435250233e-05, "loss": 0.5766, "step": 3797 }, { "epoch": 0.46, "learning_rate": 1.1712771965203191e-05, "loss": 0.5655, "step": 3798 }, { "epoch": 0.46, "learning_rate": 1.1708891229433244e-05, "loss": 0.5326, "step": 3799 }, { "epoch": 0.46, "learning_rate": 1.1705010228542453e-05, "loss": 0.6009, "step": 3800 }, { "epoch": 0.46, "learning_rate": 1.1701128963132928e-05, "loss": 0.5459, "step": 3801 }, { "epoch": 0.46, "learning_rate": 1.1697247433806815e-05, "loss": 0.5684, "step": 3802 }, { "epoch": 0.46, "learning_rate": 1.1693365641166304e-05, "loss": 0.5761, "step": 3803 }, { "epoch": 0.46, "learning_rate": 1.1689483585813621e-05, "loss": 0.5586, "step": 3804 }, { "epoch": 0.46, "learning_rate": 1.1685601268351037e-05, "loss": 0.5477, "step": 3805 }, { "epoch": 0.46, "learning_rate": 1.168171868938086e-05, "loss": 0.5468, "step": 3806 }, { "epoch": 0.46, "learning_rate": 1.1677835849505448e-05, "loss": 0.5196, "step": 3807 }, { "epoch": 0.46, "learning_rate": 1.167395274932719e-05, "loss": 0.6021, "step": 3808 }, { "epoch": 0.46, "learning_rate": 1.1670069389448513e-05, "loss": 0.546, "step": 3809 }, { "epoch": 0.46, "learning_rate": 1.1666185770471893e-05, "loss": 0.5632, "step": 3810 }, { "epoch": 0.46, "learning_rate": 1.1662301892999846e-05, "loss": 0.5448, "step": 3811 }, { "epoch": 0.46, "learning_rate": 1.1658417757634916e-05, "loss": 0.5982, "step": 3812 }, { "epoch": 0.46, "learning_rate": 1.1654533364979706e-05, "loss": 0.5541, "step": 3813 }, { "epoch": 0.46, "learning_rate": 1.1650648715636839e-05, "loss": 0.5541, "step": 3814 }, { "epoch": 0.46, "learning_rate": 1.1646763810208993e-05, "loss": 0.5084, "step": 3815 }, { "epoch": 0.46, "learning_rate": 1.164287864929888e-05, "loss": 0.5845, "step": 3816 }, { "epoch": 0.46, "learning_rate": 1.163899323350925e-05, "loss": 0.5692, "step": 3817 }, { "epoch": 0.46, "learning_rate": 1.1635107563442892e-05, "loss": 0.5663, "step": 3818 }, { "epoch": 0.46, "learning_rate": 1.1631221639702643e-05, "loss": 0.5615, "step": 3819 }, { "epoch": 0.46, "learning_rate": 1.1627335462891366e-05, "loss": 0.5766, "step": 3820 }, { "epoch": 0.46, "learning_rate": 1.1623449033611977e-05, "loss": 0.5877, "step": 3821 }, { "epoch": 0.46, "learning_rate": 1.161956235246742e-05, "loss": 0.561, "step": 3822 }, { "epoch": 0.46, "learning_rate": 1.1615675420060685e-05, "loss": 0.5329, "step": 3823 }, { "epoch": 0.47, "learning_rate": 1.1611788236994794e-05, "loss": 0.5604, "step": 3824 }, { "epoch": 0.47, "learning_rate": 1.1607900803872818e-05, "loss": 0.5654, "step": 3825 }, { "epoch": 0.47, "learning_rate": 1.160401312129786e-05, "loss": 0.5493, "step": 3826 }, { "epoch": 0.47, "learning_rate": 1.1600125189873063e-05, "loss": 0.5611, "step": 3827 }, { "epoch": 0.47, "learning_rate": 1.1596237010201606e-05, "loss": 0.5249, "step": 3828 }, { "epoch": 0.47, "learning_rate": 1.1592348582886712e-05, "loss": 0.5678, "step": 3829 }, { "epoch": 0.47, "learning_rate": 1.1588459908531636e-05, "loss": 0.4881, "step": 3830 }, { "epoch": 0.47, "learning_rate": 1.1584570987739676e-05, "loss": 0.5296, "step": 3831 }, { "epoch": 0.47, "learning_rate": 1.158068182111417e-05, "loss": 0.588, "step": 3832 }, { "epoch": 0.47, "learning_rate": 1.1576792409258489e-05, "loss": 0.4988, "step": 3833 }, { "epoch": 0.47, "learning_rate": 1.1572902752776044e-05, "loss": 0.5749, "step": 3834 }, { "epoch": 0.47, "learning_rate": 1.1569012852270284e-05, "loss": 0.5554, "step": 3835 }, { "epoch": 0.47, "learning_rate": 1.1565122708344698e-05, "loss": 0.5518, "step": 3836 }, { "epoch": 0.47, "learning_rate": 1.1561232321602807e-05, "loss": 0.5424, "step": 3837 }, { "epoch": 0.47, "learning_rate": 1.1557341692648176e-05, "loss": 0.523, "step": 3838 }, { "epoch": 0.47, "learning_rate": 1.1553450822084409e-05, "loss": 0.5429, "step": 3839 }, { "epoch": 0.47, "learning_rate": 1.1549559710515138e-05, "loss": 0.6049, "step": 3840 }, { "epoch": 0.47, "learning_rate": 1.1545668358544035e-05, "loss": 0.5407, "step": 3841 }, { "epoch": 0.47, "learning_rate": 1.154177676677482e-05, "loss": 0.6156, "step": 3842 }, { "epoch": 0.47, "learning_rate": 1.153788493581124e-05, "loss": 0.6059, "step": 3843 }, { "epoch": 0.47, "learning_rate": 1.1533992866257078e-05, "loss": 0.5745, "step": 3844 }, { "epoch": 0.47, "learning_rate": 1.153010055871616e-05, "loss": 0.5603, "step": 3845 }, { "epoch": 0.47, "learning_rate": 1.1526208013792348e-05, "loss": 0.5405, "step": 3846 }, { "epoch": 0.47, "learning_rate": 1.1522315232089539e-05, "loss": 0.5954, "step": 3847 }, { "epoch": 0.47, "learning_rate": 1.151842221421166e-05, "loss": 0.5598, "step": 3848 }, { "epoch": 0.47, "learning_rate": 1.1514528960762694e-05, "loss": 0.5798, "step": 3849 }, { "epoch": 0.47, "learning_rate": 1.151063547234664e-05, "loss": 0.6041, "step": 3850 }, { "epoch": 0.47, "learning_rate": 1.1506741749567541e-05, "loss": 0.5465, "step": 3851 }, { "epoch": 0.47, "learning_rate": 1.1502847793029484e-05, "loss": 0.5476, "step": 3852 }, { "epoch": 0.47, "learning_rate": 1.1498953603336575e-05, "loss": 0.5365, "step": 3853 }, { "epoch": 0.47, "learning_rate": 1.1495059181092973e-05, "loss": 0.5785, "step": 3854 }, { "epoch": 0.47, "learning_rate": 1.149116452690287e-05, "loss": 0.5678, "step": 3855 }, { "epoch": 0.47, "learning_rate": 1.1487269641370484e-05, "loss": 0.5424, "step": 3856 }, { "epoch": 0.47, "learning_rate": 1.1483374525100074e-05, "loss": 0.5582, "step": 3857 }, { "epoch": 0.47, "learning_rate": 1.1479479178695946e-05, "loss": 0.6045, "step": 3858 }, { "epoch": 0.47, "learning_rate": 1.1475583602762424e-05, "loss": 0.6058, "step": 3859 }, { "epoch": 0.47, "learning_rate": 1.1471687797903874e-05, "loss": 0.5101, "step": 3860 }, { "epoch": 0.47, "learning_rate": 1.1467791764724705e-05, "loss": 0.6471, "step": 3861 }, { "epoch": 0.47, "learning_rate": 1.1463895503829354e-05, "loss": 0.5031, "step": 3862 }, { "epoch": 0.47, "learning_rate": 1.1459999015822292e-05, "loss": 0.5375, "step": 3863 }, { "epoch": 0.47, "learning_rate": 1.145610230130803e-05, "loss": 0.529, "step": 3864 }, { "epoch": 0.47, "learning_rate": 1.145220536089111e-05, "loss": 0.5602, "step": 3865 }, { "epoch": 0.47, "learning_rate": 1.1448308195176114e-05, "loss": 0.5368, "step": 3866 }, { "epoch": 0.47, "learning_rate": 1.1444410804767656e-05, "loss": 0.6214, "step": 3867 }, { "epoch": 0.47, "learning_rate": 1.1440513190270384e-05, "loss": 0.633, "step": 3868 }, { "epoch": 0.47, "learning_rate": 1.1436615352288983e-05, "loss": 0.5425, "step": 3869 }, { "epoch": 0.47, "learning_rate": 1.143271729142817e-05, "loss": 0.5086, "step": 3870 }, { "epoch": 0.47, "learning_rate": 1.1428819008292698e-05, "loss": 0.5321, "step": 3871 }, { "epoch": 0.47, "learning_rate": 1.1424920503487354e-05, "loss": 0.5984, "step": 3872 }, { "epoch": 0.47, "learning_rate": 1.1421021777616961e-05, "loss": 0.5231, "step": 3873 }, { "epoch": 0.47, "learning_rate": 1.1417122831286376e-05, "loss": 0.5501, "step": 3874 }, { "epoch": 0.47, "learning_rate": 1.1413223665100493e-05, "loss": 0.5474, "step": 3875 }, { "epoch": 0.47, "learning_rate": 1.1409324279664226e-05, "loss": 0.5151, "step": 3876 }, { "epoch": 0.47, "learning_rate": 1.1405424675582542e-05, "loss": 0.6097, "step": 3877 }, { "epoch": 0.47, "learning_rate": 1.1401524853460431e-05, "loss": 0.5588, "step": 3878 }, { "epoch": 0.47, "learning_rate": 1.1397624813902919e-05, "loss": 0.5775, "step": 3879 }, { "epoch": 0.47, "learning_rate": 1.1393724557515066e-05, "loss": 0.4962, "step": 3880 }, { "epoch": 0.47, "learning_rate": 1.1389824084901968e-05, "loss": 0.5649, "step": 3881 }, { "epoch": 0.47, "learning_rate": 1.1385923396668751e-05, "loss": 0.5663, "step": 3882 }, { "epoch": 0.47, "learning_rate": 1.1382022493420575e-05, "loss": 0.5009, "step": 3883 }, { "epoch": 0.47, "learning_rate": 1.137812137576263e-05, "loss": 0.5999, "step": 3884 }, { "epoch": 0.47, "learning_rate": 1.137422004430015e-05, "loss": 0.5599, "step": 3885 }, { "epoch": 0.47, "learning_rate": 1.1370318499638392e-05, "loss": 0.5312, "step": 3886 }, { "epoch": 0.47, "learning_rate": 1.1366416742382654e-05, "loss": 0.579, "step": 3887 }, { "epoch": 0.47, "learning_rate": 1.1362514773138254e-05, "loss": 0.5547, "step": 3888 }, { "epoch": 0.47, "learning_rate": 1.1358612592510561e-05, "loss": 0.5629, "step": 3889 }, { "epoch": 0.47, "learning_rate": 1.1354710201104955e-05, "loss": 0.6125, "step": 3890 }, { "epoch": 0.47, "learning_rate": 1.1350807599526876e-05, "loss": 0.4889, "step": 3891 }, { "epoch": 0.47, "learning_rate": 1.134690478838177e-05, "loss": 0.5337, "step": 3892 }, { "epoch": 0.47, "learning_rate": 1.134300176827513e-05, "loss": 0.545, "step": 3893 }, { "epoch": 0.47, "learning_rate": 1.133909853981248e-05, "loss": 0.5839, "step": 3894 }, { "epoch": 0.47, "learning_rate": 1.133519510359938e-05, "loss": 0.5776, "step": 3895 }, { "epoch": 0.47, "learning_rate": 1.1331291460241403e-05, "loss": 0.5539, "step": 3896 }, { "epoch": 0.47, "learning_rate": 1.1327387610344178e-05, "loss": 0.5487, "step": 3897 }, { "epoch": 0.47, "learning_rate": 1.1323483554513357e-05, "loss": 0.5624, "step": 3898 }, { "epoch": 0.47, "learning_rate": 1.1319579293354618e-05, "loss": 0.6297, "step": 3899 }, { "epoch": 0.47, "learning_rate": 1.1315674827473679e-05, "loss": 0.5775, "step": 3900 }, { "epoch": 0.47, "learning_rate": 1.1311770157476287e-05, "loss": 0.5174, "step": 3901 }, { "epoch": 0.47, "learning_rate": 1.1307865283968217e-05, "loss": 0.5526, "step": 3902 }, { "epoch": 0.47, "learning_rate": 1.1303960207555284e-05, "loss": 0.5571, "step": 3903 }, { "epoch": 0.47, "learning_rate": 1.1300054928843328e-05, "loss": 0.5863, "step": 3904 }, { "epoch": 0.47, "learning_rate": 1.1296149448438218e-05, "loss": 0.5712, "step": 3905 }, { "epoch": 0.47, "learning_rate": 1.1292243766945863e-05, "loss": 0.5861, "step": 3906 }, { "epoch": 0.48, "learning_rate": 1.1288337884972196e-05, "loss": 0.5104, "step": 3907 }, { "epoch": 0.48, "learning_rate": 1.128443180312318e-05, "loss": 0.5691, "step": 3908 }, { "epoch": 0.48, "learning_rate": 1.1280525522004821e-05, "loss": 0.5685, "step": 3909 }, { "epoch": 0.48, "learning_rate": 1.127661904222314e-05, "loss": 0.5745, "step": 3910 }, { "epoch": 0.48, "learning_rate": 1.1272712364384201e-05, "loss": 0.548, "step": 3911 }, { "epoch": 0.48, "learning_rate": 1.126880548909409e-05, "loss": 0.5598, "step": 3912 }, { "epoch": 0.48, "learning_rate": 1.126489841695893e-05, "loss": 0.5158, "step": 3913 }, { "epoch": 0.48, "learning_rate": 1.1260991148584868e-05, "loss": 0.5817, "step": 3914 }, { "epoch": 0.48, "learning_rate": 1.1257083684578092e-05, "loss": 0.5583, "step": 3915 }, { "epoch": 0.48, "learning_rate": 1.1253176025544808e-05, "loss": 0.5318, "step": 3916 }, { "epoch": 0.48, "learning_rate": 1.124926817209126e-05, "loss": 0.5513, "step": 3917 }, { "epoch": 0.48, "learning_rate": 1.1245360124823728e-05, "loss": 0.5968, "step": 3918 }, { "epoch": 0.48, "learning_rate": 1.12414518843485e-05, "loss": 0.5783, "step": 3919 }, { "epoch": 0.48, "learning_rate": 1.1237543451271917e-05, "loss": 0.5527, "step": 3920 }, { "epoch": 0.48, "learning_rate": 1.1233634826200341e-05, "loss": 0.5729, "step": 3921 }, { "epoch": 0.48, "learning_rate": 1.1229726009740162e-05, "loss": 0.5861, "step": 3922 }, { "epoch": 0.48, "learning_rate": 1.1225817002497805e-05, "loss": 0.5505, "step": 3923 }, { "epoch": 0.48, "learning_rate": 1.1221907805079716e-05, "loss": 0.4795, "step": 3924 }, { "epoch": 0.48, "learning_rate": 1.1217998418092382e-05, "loss": 0.5724, "step": 3925 }, { "epoch": 0.48, "learning_rate": 1.1214088842142305e-05, "loss": 0.5753, "step": 3926 }, { "epoch": 0.48, "learning_rate": 1.1210179077836032e-05, "loss": 0.4903, "step": 3927 }, { "epoch": 0.48, "learning_rate": 1.1206269125780132e-05, "loss": 0.5749, "step": 3928 }, { "epoch": 0.48, "learning_rate": 1.1202358986581197e-05, "loss": 0.607, "step": 3929 }, { "epoch": 0.48, "learning_rate": 1.1198448660845855e-05, "loss": 0.5729, "step": 3930 }, { "epoch": 0.48, "learning_rate": 1.1194538149180769e-05, "loss": 0.5586, "step": 3931 }, { "epoch": 0.48, "learning_rate": 1.1190627452192613e-05, "loss": 0.5785, "step": 3932 }, { "epoch": 0.48, "learning_rate": 1.1186716570488108e-05, "loss": 0.5544, "step": 3933 }, { "epoch": 0.48, "learning_rate": 1.1182805504673996e-05, "loss": 0.5893, "step": 3934 }, { "epoch": 0.48, "learning_rate": 1.1178894255357045e-05, "loss": 0.581, "step": 3935 }, { "epoch": 0.48, "learning_rate": 1.1174982823144051e-05, "loss": 0.5464, "step": 3936 }, { "epoch": 0.48, "learning_rate": 1.1171071208641848e-05, "loss": 0.5871, "step": 3937 }, { "epoch": 0.48, "learning_rate": 1.116715941245729e-05, "loss": 0.5593, "step": 3938 }, { "epoch": 0.48, "learning_rate": 1.1163247435197258e-05, "loss": 0.5382, "step": 3939 }, { "epoch": 0.48, "learning_rate": 1.1159335277468665e-05, "loss": 0.5342, "step": 3940 }, { "epoch": 0.48, "learning_rate": 1.1155422939878458e-05, "loss": 0.5301, "step": 3941 }, { "epoch": 0.48, "learning_rate": 1.1151510423033594e-05, "loss": 0.5435, "step": 3942 }, { "epoch": 0.48, "learning_rate": 1.1147597727541076e-05, "loss": 0.5755, "step": 3943 }, { "epoch": 0.48, "learning_rate": 1.1143684854007924e-05, "loss": 0.5899, "step": 3944 }, { "epoch": 0.48, "learning_rate": 1.1139771803041189e-05, "loss": 0.5892, "step": 3945 }, { "epoch": 0.48, "learning_rate": 1.1135858575247953e-05, "loss": 0.554, "step": 3946 }, { "epoch": 0.48, "learning_rate": 1.1131945171235319e-05, "loss": 0.5488, "step": 3947 }, { "epoch": 0.48, "learning_rate": 1.1128031591610427e-05, "loss": 0.5682, "step": 3948 }, { "epoch": 0.48, "learning_rate": 1.1124117836980425e-05, "loss": 0.57, "step": 3949 }, { "epoch": 0.48, "learning_rate": 1.1120203907952511e-05, "loss": 0.576, "step": 3950 }, { "epoch": 0.48, "learning_rate": 1.1116289805133897e-05, "loss": 0.5953, "step": 3951 }, { "epoch": 0.48, "learning_rate": 1.1112375529131826e-05, "loss": 0.5602, "step": 3952 }, { "epoch": 0.48, "learning_rate": 1.1108461080553566e-05, "loss": 0.5687, "step": 3953 }, { "epoch": 0.48, "learning_rate": 1.1104546460006412e-05, "loss": 0.5625, "step": 3954 }, { "epoch": 0.48, "learning_rate": 1.1100631668097683e-05, "loss": 0.5137, "step": 3955 }, { "epoch": 0.48, "learning_rate": 1.109671670543473e-05, "loss": 0.5661, "step": 3956 }, { "epoch": 0.48, "learning_rate": 1.1092801572624934e-05, "loss": 0.5638, "step": 3957 }, { "epoch": 0.48, "learning_rate": 1.1088886270275689e-05, "loss": 0.6084, "step": 3958 }, { "epoch": 0.48, "learning_rate": 1.1084970798994425e-05, "loss": 0.5849, "step": 3959 }, { "epoch": 0.48, "learning_rate": 1.1081055159388597e-05, "loss": 0.5858, "step": 3960 }, { "epoch": 0.48, "learning_rate": 1.1077139352065686e-05, "loss": 0.5589, "step": 3961 }, { "epoch": 0.48, "learning_rate": 1.1073223377633196e-05, "loss": 0.5678, "step": 3962 }, { "epoch": 0.48, "learning_rate": 1.1069307236698661e-05, "loss": 0.5649, "step": 3963 }, { "epoch": 0.48, "learning_rate": 1.1065390929869642e-05, "loss": 0.4977, "step": 3964 }, { "epoch": 0.48, "learning_rate": 1.1061474457753714e-05, "loss": 0.5421, "step": 3965 }, { "epoch": 0.48, "learning_rate": 1.1057557820958493e-05, "loss": 0.5552, "step": 3966 }, { "epoch": 0.48, "learning_rate": 1.1053641020091614e-05, "loss": 0.6063, "step": 3967 }, { "epoch": 0.48, "learning_rate": 1.1049724055760736e-05, "loss": 0.5601, "step": 3968 }, { "epoch": 0.48, "learning_rate": 1.1045806928573541e-05, "loss": 0.5322, "step": 3969 }, { "epoch": 0.48, "learning_rate": 1.1041889639137748e-05, "loss": 0.606, "step": 3970 }, { "epoch": 0.48, "learning_rate": 1.1037972188061091e-05, "loss": 0.5853, "step": 3971 }, { "epoch": 0.48, "learning_rate": 1.1034054575951326e-05, "loss": 0.5835, "step": 3972 }, { "epoch": 0.48, "learning_rate": 1.103013680341624e-05, "loss": 0.5761, "step": 3973 }, { "epoch": 0.48, "learning_rate": 1.102621887106365e-05, "loss": 0.5994, "step": 3974 }, { "epoch": 0.48, "learning_rate": 1.102230077950139e-05, "loss": 0.5753, "step": 3975 }, { "epoch": 0.48, "learning_rate": 1.1018382529337315e-05, "loss": 0.6037, "step": 3976 }, { "epoch": 0.48, "learning_rate": 1.101446412117932e-05, "loss": 0.5782, "step": 3977 }, { "epoch": 0.48, "learning_rate": 1.1010545555635306e-05, "loss": 0.6055, "step": 3978 }, { "epoch": 0.48, "learning_rate": 1.1006626833313209e-05, "loss": 0.5991, "step": 3979 }, { "epoch": 0.48, "learning_rate": 1.1002707954820993e-05, "loss": 0.5635, "step": 3980 }, { "epoch": 0.48, "learning_rate": 1.0998788920766634e-05, "loss": 0.5619, "step": 3981 }, { "epoch": 0.48, "learning_rate": 1.0994869731758142e-05, "loss": 0.5616, "step": 3982 }, { "epoch": 0.48, "learning_rate": 1.0990950388403548e-05, "loss": 0.5234, "step": 3983 }, { "epoch": 0.48, "learning_rate": 1.0987030891310909e-05, "loss": 0.5369, "step": 3984 }, { "epoch": 0.48, "learning_rate": 1.0983111241088298e-05, "loss": 0.5193, "step": 3985 }, { "epoch": 0.48, "learning_rate": 1.097919143834382e-05, "loss": 0.6024, "step": 3986 }, { "epoch": 0.48, "learning_rate": 1.0975271483685603e-05, "loss": 0.5621, "step": 3987 }, { "epoch": 0.48, "learning_rate": 1.0971351377721796e-05, "loss": 0.5532, "step": 3988 }, { "epoch": 0.49, "learning_rate": 1.096743112106057e-05, "loss": 0.5326, "step": 3989 }, { "epoch": 0.49, "learning_rate": 1.0963510714310123e-05, "loss": 0.5209, "step": 3990 }, { "epoch": 0.49, "learning_rate": 1.095959015807868e-05, "loss": 0.5658, "step": 3991 }, { "epoch": 0.49, "learning_rate": 1.0955669452974471e-05, "loss": 0.6201, "step": 3992 }, { "epoch": 0.49, "learning_rate": 1.0951748599605775e-05, "loss": 0.5881, "step": 3993 }, { "epoch": 0.49, "learning_rate": 1.0947827598580877e-05, "loss": 0.5197, "step": 3994 }, { "epoch": 0.49, "learning_rate": 1.0943906450508083e-05, "loss": 0.5535, "step": 3995 }, { "epoch": 0.49, "learning_rate": 1.0939985155995738e-05, "loss": 0.5549, "step": 3996 }, { "epoch": 0.49, "learning_rate": 1.0936063715652195e-05, "loss": 0.544, "step": 3997 }, { "epoch": 0.49, "learning_rate": 1.093214213008583e-05, "loss": 0.5337, "step": 3998 }, { "epoch": 0.49, "learning_rate": 1.0928220399905052e-05, "loss": 0.56, "step": 3999 }, { "epoch": 0.49, "learning_rate": 1.0924298525718284e-05, "loss": 0.5714, "step": 4000 }, { "epoch": 0.49, "learning_rate": 1.0920376508133971e-05, "loss": 0.5594, "step": 4001 }, { "epoch": 0.49, "learning_rate": 1.0916454347760588e-05, "loss": 0.5623, "step": 4002 }, { "epoch": 0.49, "learning_rate": 1.0912532045206621e-05, "loss": 0.543, "step": 4003 }, { "epoch": 0.49, "learning_rate": 1.0908609601080589e-05, "loss": 0.5304, "step": 4004 }, { "epoch": 0.49, "learning_rate": 1.0904687015991025e-05, "loss": 0.5586, "step": 4005 }, { "epoch": 0.49, "learning_rate": 1.0900764290546486e-05, "loss": 0.5756, "step": 4006 }, { "epoch": 0.49, "learning_rate": 1.0896841425355556e-05, "loss": 0.5382, "step": 4007 }, { "epoch": 0.49, "learning_rate": 1.0892918421026832e-05, "loss": 0.5686, "step": 4008 }, { "epoch": 0.49, "learning_rate": 1.0888995278168937e-05, "loss": 0.5592, "step": 4009 }, { "epoch": 0.49, "learning_rate": 1.0885071997390517e-05, "loss": 0.5783, "step": 4010 }, { "epoch": 0.49, "learning_rate": 1.0881148579300238e-05, "loss": 0.6229, "step": 4011 }, { "epoch": 0.49, "learning_rate": 1.0877225024506786e-05, "loss": 0.5359, "step": 4012 }, { "epoch": 0.49, "learning_rate": 1.0873301333618869e-05, "loss": 0.5673, "step": 4013 }, { "epoch": 0.49, "learning_rate": 1.0869377507245218e-05, "loss": 0.5324, "step": 4014 }, { "epoch": 0.49, "learning_rate": 1.086545354599458e-05, "loss": 0.568, "step": 4015 }, { "epoch": 0.49, "learning_rate": 1.0861529450475727e-05, "loss": 0.538, "step": 4016 }, { "epoch": 0.49, "learning_rate": 1.0857605221297455e-05, "loss": 0.5594, "step": 4017 }, { "epoch": 0.49, "learning_rate": 1.0853680859068573e-05, "loss": 0.54, "step": 4018 }, { "epoch": 0.49, "learning_rate": 1.0849756364397918e-05, "loss": 0.5157, "step": 4019 }, { "epoch": 0.49, "learning_rate": 1.0845831737894343e-05, "loss": 0.5381, "step": 4020 }, { "epoch": 0.49, "learning_rate": 1.0841906980166718e-05, "loss": 0.5796, "step": 4021 }, { "epoch": 0.49, "learning_rate": 1.083798209182394e-05, "loss": 0.5589, "step": 4022 }, { "epoch": 0.49, "learning_rate": 1.083405707347493e-05, "loss": 0.6442, "step": 4023 }, { "epoch": 0.49, "learning_rate": 1.0830131925728615e-05, "loss": 0.5487, "step": 4024 }, { "epoch": 0.49, "learning_rate": 1.0826206649193955e-05, "loss": 0.4704, "step": 4025 }, { "epoch": 0.49, "learning_rate": 1.0822281244479927e-05, "loss": 0.5746, "step": 4026 }, { "epoch": 0.49, "learning_rate": 1.081835571219552e-05, "loss": 0.5599, "step": 4027 }, { "epoch": 0.49, "learning_rate": 1.0814430052949754e-05, "loss": 0.5614, "step": 4028 }, { "epoch": 0.49, "learning_rate": 1.081050426735166e-05, "loss": 0.5783, "step": 4029 }, { "epoch": 0.49, "learning_rate": 1.08065783560103e-05, "loss": 0.5649, "step": 4030 }, { "epoch": 0.49, "learning_rate": 1.0802652319534739e-05, "loss": 0.5811, "step": 4031 }, { "epoch": 0.49, "learning_rate": 1.0798726158534071e-05, "loss": 0.5128, "step": 4032 }, { "epoch": 0.49, "learning_rate": 1.0794799873617418e-05, "loss": 0.5389, "step": 4033 }, { "epoch": 0.49, "learning_rate": 1.07908734653939e-05, "loss": 0.5531, "step": 4034 }, { "epoch": 0.49, "learning_rate": 1.0786946934472673e-05, "loss": 0.5506, "step": 4035 }, { "epoch": 0.49, "learning_rate": 1.0783020281462909e-05, "loss": 0.531, "step": 4036 }, { "epoch": 0.49, "learning_rate": 1.0779093506973792e-05, "loss": 0.5371, "step": 4037 }, { "epoch": 0.49, "learning_rate": 1.0775166611614534e-05, "loss": 0.559, "step": 4038 }, { "epoch": 0.49, "learning_rate": 1.077123959599436e-05, "loss": 0.5278, "step": 4039 }, { "epoch": 0.49, "learning_rate": 1.0767312460722517e-05, "loss": 0.5581, "step": 4040 }, { "epoch": 0.49, "learning_rate": 1.0763385206408263e-05, "loss": 0.5921, "step": 4041 }, { "epoch": 0.49, "learning_rate": 1.0759457833660887e-05, "loss": 0.5534, "step": 4042 }, { "epoch": 0.49, "learning_rate": 1.0755530343089687e-05, "loss": 0.59, "step": 4043 }, { "epoch": 0.49, "learning_rate": 1.0751602735303981e-05, "loss": 0.6048, "step": 4044 }, { "epoch": 0.49, "learning_rate": 1.0747675010913103e-05, "loss": 0.5079, "step": 4045 }, { "epoch": 0.49, "learning_rate": 1.0743747170526415e-05, "loss": 0.561, "step": 4046 }, { "epoch": 0.49, "learning_rate": 1.0739819214753286e-05, "loss": 0.6424, "step": 4047 }, { "epoch": 0.49, "learning_rate": 1.0735891144203108e-05, "loss": 0.5446, "step": 4048 }, { "epoch": 0.49, "learning_rate": 1.0731962959485286e-05, "loss": 0.5609, "step": 4049 }, { "epoch": 0.49, "learning_rate": 1.0728034661209256e-05, "loss": 0.6112, "step": 4050 }, { "epoch": 0.49, "learning_rate": 1.0724106249984454e-05, "loss": 0.5666, "step": 4051 }, { "epoch": 0.49, "learning_rate": 1.072017772642034e-05, "loss": 0.5179, "step": 4052 }, { "epoch": 0.49, "learning_rate": 1.0716249091126401e-05, "loss": 0.5584, "step": 4053 }, { "epoch": 0.49, "learning_rate": 1.0712320344712129e-05, "loss": 0.5109, "step": 4054 }, { "epoch": 0.49, "learning_rate": 1.0708391487787036e-05, "loss": 0.528, "step": 4055 }, { "epoch": 0.49, "learning_rate": 1.0704462520960657e-05, "loss": 0.581, "step": 4056 }, { "epoch": 0.49, "learning_rate": 1.0700533444842536e-05, "loss": 0.5827, "step": 4057 }, { "epoch": 0.49, "learning_rate": 1.0696604260042236e-05, "loss": 0.5921, "step": 4058 }, { "epoch": 0.49, "learning_rate": 1.0692674967169347e-05, "loss": 0.5234, "step": 4059 }, { "epoch": 0.49, "learning_rate": 1.0688745566833462e-05, "loss": 0.5635, "step": 4060 }, { "epoch": 0.49, "learning_rate": 1.0684816059644196e-05, "loss": 0.5296, "step": 4061 }, { "epoch": 0.49, "learning_rate": 1.0680886446211183e-05, "loss": 0.477, "step": 4062 }, { "epoch": 0.49, "learning_rate": 1.0676956727144067e-05, "loss": 0.5479, "step": 4063 }, { "epoch": 0.49, "learning_rate": 1.0673026903052515e-05, "loss": 0.6005, "step": 4064 }, { "epoch": 0.49, "learning_rate": 1.0669096974546209e-05, "loss": 0.5349, "step": 4065 }, { "epoch": 0.49, "learning_rate": 1.0665166942234849e-05, "loss": 0.5807, "step": 4066 }, { "epoch": 0.49, "learning_rate": 1.0661236806728139e-05, "loss": 0.4827, "step": 4067 }, { "epoch": 0.49, "learning_rate": 1.0657306568635815e-05, "loss": 0.6008, "step": 4068 }, { "epoch": 0.49, "learning_rate": 1.0653376228567617e-05, "loss": 0.5741, "step": 4069 }, { "epoch": 0.49, "learning_rate": 1.0649445787133315e-05, "loss": 0.5232, "step": 4070 }, { "epoch": 0.5, "learning_rate": 1.0645515244942676e-05, "loss": 0.5496, "step": 4071 }, { "epoch": 0.5, "learning_rate": 1.0641584602605496e-05, "loss": 0.5264, "step": 4072 }, { "epoch": 0.5, "learning_rate": 1.0637653860731589e-05, "loss": 0.5397, "step": 4073 }, { "epoch": 0.5, "learning_rate": 1.0633723019930766e-05, "loss": 0.5472, "step": 4074 }, { "epoch": 0.5, "learning_rate": 1.0629792080812875e-05, "loss": 0.5143, "step": 4075 }, { "epoch": 0.5, "learning_rate": 1.0625861043987767e-05, "loss": 0.5383, "step": 4076 }, { "epoch": 0.5, "learning_rate": 1.0621929910065307e-05, "loss": 0.5649, "step": 4077 }, { "epoch": 0.5, "learning_rate": 1.0617998679655386e-05, "loss": 0.4984, "step": 4078 }, { "epoch": 0.5, "learning_rate": 1.06140673533679e-05, "loss": 0.5509, "step": 4079 }, { "epoch": 0.5, "learning_rate": 1.0610135931812762e-05, "loss": 0.6183, "step": 4080 }, { "epoch": 0.5, "learning_rate": 1.0606204415599902e-05, "loss": 0.559, "step": 4081 }, { "epoch": 0.5, "learning_rate": 1.0602272805339262e-05, "loss": 0.5485, "step": 4082 }, { "epoch": 0.5, "learning_rate": 1.0598341101640801e-05, "loss": 0.4844, "step": 4083 }, { "epoch": 0.5, "learning_rate": 1.059440930511449e-05, "loss": 0.5545, "step": 4084 }, { "epoch": 0.5, "learning_rate": 1.059047741637032e-05, "loss": 0.5517, "step": 4085 }, { "epoch": 0.5, "learning_rate": 1.058654543601829e-05, "loss": 0.4976, "step": 4086 }, { "epoch": 0.5, "learning_rate": 1.058261336466841e-05, "loss": 0.5895, "step": 4087 }, { "epoch": 0.5, "learning_rate": 1.0578681202930714e-05, "loss": 0.5528, "step": 4088 }, { "epoch": 0.5, "learning_rate": 1.0574748951415252e-05, "loss": 0.5431, "step": 4089 }, { "epoch": 0.5, "learning_rate": 1.057081661073207e-05, "loss": 0.6056, "step": 4090 }, { "epoch": 0.5, "learning_rate": 1.0566884181491246e-05, "loss": 0.5842, "step": 4091 }, { "epoch": 0.5, "learning_rate": 1.0562951664302863e-05, "loss": 0.5514, "step": 4092 }, { "epoch": 0.5, "learning_rate": 1.0559019059777023e-05, "loss": 0.5866, "step": 4093 }, { "epoch": 0.5, "learning_rate": 1.055508636852383e-05, "loss": 0.593, "step": 4094 }, { "epoch": 0.5, "learning_rate": 1.055115359115342e-05, "loss": 0.5401, "step": 4095 }, { "epoch": 0.5, "learning_rate": 1.0547220728275927e-05, "loss": 0.5231, "step": 4096 }, { "epoch": 0.5, "learning_rate": 1.0543287780501502e-05, "loss": 0.5722, "step": 4097 }, { "epoch": 0.5, "learning_rate": 1.053935474844031e-05, "loss": 0.565, "step": 4098 }, { "epoch": 0.5, "learning_rate": 1.0535421632702534e-05, "loss": 0.5897, "step": 4099 }, { "epoch": 0.5, "learning_rate": 1.053148843389836e-05, "loss": 0.5762, "step": 4100 }, { "epoch": 0.5, "learning_rate": 1.0527555152637996e-05, "loss": 0.6062, "step": 4101 }, { "epoch": 0.5, "learning_rate": 1.0523621789531661e-05, "loss": 0.5724, "step": 4102 }, { "epoch": 0.5, "learning_rate": 1.0519688345189583e-05, "loss": 0.52, "step": 4103 }, { "epoch": 0.5, "learning_rate": 1.0515754820222002e-05, "loss": 0.5674, "step": 4104 }, { "epoch": 0.5, "learning_rate": 1.0511821215239174e-05, "loss": 0.4855, "step": 4105 }, { "epoch": 0.5, "learning_rate": 1.0507887530851368e-05, "loss": 0.5238, "step": 4106 }, { "epoch": 0.5, "learning_rate": 1.0503953767668865e-05, "loss": 0.5453, "step": 4107 }, { "epoch": 0.5, "learning_rate": 1.0500019926301953e-05, "loss": 0.5479, "step": 4108 }, { "epoch": 0.5, "learning_rate": 1.049608600736094e-05, "loss": 0.6017, "step": 4109 }, { "epoch": 0.5, "learning_rate": 1.049215201145614e-05, "loss": 0.5513, "step": 4110 }, { "epoch": 0.5, "learning_rate": 1.048821793919788e-05, "loss": 0.5849, "step": 4111 }, { "epoch": 0.5, "learning_rate": 1.0484283791196504e-05, "loss": 0.5702, "step": 4112 }, { "epoch": 0.5, "learning_rate": 1.0480349568062357e-05, "loss": 0.4896, "step": 4113 }, { "epoch": 0.5, "learning_rate": 1.0476415270405812e-05, "loss": 0.6029, "step": 4114 }, { "epoch": 0.5, "learning_rate": 1.0472480898837235e-05, "loss": 0.5806, "step": 4115 }, { "epoch": 0.5, "learning_rate": 1.0468546453967017e-05, "loss": 0.5437, "step": 4116 }, { "epoch": 0.5, "learning_rate": 1.0464611936405553e-05, "loss": 0.5451, "step": 4117 }, { "epoch": 0.5, "learning_rate": 1.0460677346763256e-05, "loss": 0.5841, "step": 4118 }, { "epoch": 0.5, "learning_rate": 1.045674268565054e-05, "loss": 0.5264, "step": 4119 }, { "epoch": 0.5, "learning_rate": 1.0452807953677841e-05, "loss": 0.5269, "step": 4120 }, { "epoch": 0.5, "learning_rate": 1.04488731514556e-05, "loss": 0.5964, "step": 4121 }, { "epoch": 0.5, "learning_rate": 1.0444938279594274e-05, "loss": 0.552, "step": 4122 }, { "epoch": 0.5, "learning_rate": 1.044100333870432e-05, "loss": 0.5134, "step": 4123 }, { "epoch": 0.5, "learning_rate": 1.0437068329396213e-05, "loss": 0.6109, "step": 4124 }, { "epoch": 0.5, "learning_rate": 1.0433133252280448e-05, "loss": 0.5842, "step": 4125 }, { "epoch": 0.5, "learning_rate": 1.0429198107967512e-05, "loss": 0.5407, "step": 4126 }, { "epoch": 0.5, "learning_rate": 1.0425262897067911e-05, "loss": 0.5414, "step": 4127 }, { "epoch": 0.5, "learning_rate": 1.0421327620192166e-05, "loss": 0.5221, "step": 4128 }, { "epoch": 0.5, "learning_rate": 1.0417392277950802e-05, "loss": 0.5484, "step": 4129 }, { "epoch": 0.5, "learning_rate": 1.0413456870954353e-05, "loss": 0.5567, "step": 4130 }, { "epoch": 0.5, "learning_rate": 1.0409521399813371e-05, "loss": 0.5613, "step": 4131 }, { "epoch": 0.5, "learning_rate": 1.0405585865138412e-05, "loss": 0.5751, "step": 4132 }, { "epoch": 0.5, "learning_rate": 1.040165026754004e-05, "loss": 0.554, "step": 4133 }, { "epoch": 0.5, "learning_rate": 1.0397714607628834e-05, "loss": 0.5108, "step": 4134 }, { "epoch": 0.5, "learning_rate": 1.039377888601538e-05, "loss": 0.543, "step": 4135 }, { "epoch": 0.5, "learning_rate": 1.0389843103310268e-05, "loss": 0.5274, "step": 4136 }, { "epoch": 0.5, "learning_rate": 1.0385907260124114e-05, "loss": 0.5841, "step": 4137 }, { "epoch": 0.5, "learning_rate": 1.0381971357067526e-05, "loss": 0.5094, "step": 4138 }, { "epoch": 0.5, "learning_rate": 1.037803539475113e-05, "loss": 0.5594, "step": 4139 }, { "epoch": 0.5, "learning_rate": 1.0374099373785555e-05, "loss": 0.5308, "step": 4140 }, { "epoch": 0.5, "learning_rate": 1.0370163294781448e-05, "loss": 0.5382, "step": 4141 }, { "epoch": 0.5, "learning_rate": 1.036622715834946e-05, "loss": 0.5255, "step": 4142 }, { "epoch": 0.5, "learning_rate": 1.0362290965100248e-05, "loss": 0.568, "step": 4143 }, { "epoch": 0.5, "learning_rate": 1.0358354715644483e-05, "loss": 0.5727, "step": 4144 }, { "epoch": 0.5, "learning_rate": 1.0354418410592846e-05, "loss": 0.5287, "step": 4145 }, { "epoch": 0.5, "learning_rate": 1.0350482050556016e-05, "loss": 0.5439, "step": 4146 }, { "epoch": 0.5, "learning_rate": 1.0346545636144694e-05, "loss": 0.4952, "step": 4147 }, { "epoch": 0.5, "learning_rate": 1.0342609167969579e-05, "loss": 0.4948, "step": 4148 }, { "epoch": 0.5, "learning_rate": 1.0338672646641386e-05, "loss": 0.5977, "step": 4149 }, { "epoch": 0.5, "learning_rate": 1.0334736072770833e-05, "loss": 0.5466, "step": 4150 }, { "epoch": 0.5, "learning_rate": 1.033079944696865e-05, "loss": 0.5202, "step": 4151 }, { "epoch": 0.5, "learning_rate": 1.0326862769845571e-05, "loss": 0.5199, "step": 4152 }, { "epoch": 0.51, "learning_rate": 1.032292604201234e-05, "loss": 0.4733, "step": 4153 }, { "epoch": 0.51, "learning_rate": 1.0318989264079708e-05, "loss": 0.5879, "step": 4154 }, { "epoch": 0.51, "learning_rate": 1.031505243665844e-05, "loss": 0.5605, "step": 4155 }, { "epoch": 0.51, "learning_rate": 1.0311115560359298e-05, "loss": 0.6202, "step": 4156 }, { "epoch": 0.51, "learning_rate": 1.0307178635793057e-05, "loss": 0.5664, "step": 4157 }, { "epoch": 0.51, "learning_rate": 1.0303241663570507e-05, "loss": 0.6002, "step": 4158 }, { "epoch": 0.51, "learning_rate": 1.0299304644302427e-05, "loss": 0.5387, "step": 4159 }, { "epoch": 0.51, "learning_rate": 1.0295367578599618e-05, "loss": 0.5048, "step": 4160 }, { "epoch": 0.51, "learning_rate": 1.029143046707289e-05, "loss": 0.5685, "step": 4161 }, { "epoch": 0.51, "learning_rate": 1.0287493310333048e-05, "loss": 0.5924, "step": 4162 }, { "epoch": 0.51, "learning_rate": 1.0283556108990911e-05, "loss": 0.4989, "step": 4163 }, { "epoch": 0.51, "learning_rate": 1.0279618863657305e-05, "loss": 0.5429, "step": 4164 }, { "epoch": 0.51, "learning_rate": 1.0275681574943068e-05, "loss": 0.6408, "step": 4165 }, { "epoch": 0.51, "learning_rate": 1.0271744243459028e-05, "loss": 0.5895, "step": 4166 }, { "epoch": 0.51, "learning_rate": 1.0267806869816038e-05, "loss": 0.5296, "step": 4167 }, { "epoch": 0.51, "learning_rate": 1.026386945462495e-05, "loss": 0.5398, "step": 4168 }, { "epoch": 0.51, "learning_rate": 1.0259931998496622e-05, "loss": 0.5872, "step": 4169 }, { "epoch": 0.51, "learning_rate": 1.0255994502041915e-05, "loss": 0.5759, "step": 4170 }, { "epoch": 0.51, "learning_rate": 1.0252056965871705e-05, "loss": 0.6176, "step": 4171 }, { "epoch": 0.51, "learning_rate": 1.0248119390596866e-05, "loss": 0.5326, "step": 4172 }, { "epoch": 0.51, "learning_rate": 1.0244181776828284e-05, "loss": 0.5579, "step": 4173 }, { "epoch": 0.51, "learning_rate": 1.0240244125176846e-05, "loss": 0.5572, "step": 4174 }, { "epoch": 0.51, "learning_rate": 1.0236306436253454e-05, "loss": 0.6169, "step": 4175 }, { "epoch": 0.51, "learning_rate": 1.0232368710669e-05, "loss": 0.5545, "step": 4176 }, { "epoch": 0.51, "learning_rate": 1.0228430949034392e-05, "loss": 0.5341, "step": 4177 }, { "epoch": 0.51, "learning_rate": 1.0224493151960544e-05, "loss": 0.4834, "step": 4178 }, { "epoch": 0.51, "learning_rate": 1.0220555320058376e-05, "loss": 0.5598, "step": 4179 }, { "epoch": 0.51, "learning_rate": 1.021661745393881e-05, "loss": 0.518, "step": 4180 }, { "epoch": 0.51, "learning_rate": 1.0212679554212777e-05, "loss": 0.5923, "step": 4181 }, { "epoch": 0.51, "learning_rate": 1.0208741621491202e-05, "loss": 0.5245, "step": 4182 }, { "epoch": 0.51, "learning_rate": 1.0204803656385034e-05, "loss": 0.545, "step": 4183 }, { "epoch": 0.51, "learning_rate": 1.020086565950521e-05, "loss": 0.574, "step": 4184 }, { "epoch": 0.51, "learning_rate": 1.0196927631462684e-05, "loss": 0.571, "step": 4185 }, { "epoch": 0.51, "learning_rate": 1.0192989572868403e-05, "loss": 0.5646, "step": 4186 }, { "epoch": 0.51, "learning_rate": 1.0189051484333332e-05, "loss": 0.5606, "step": 4187 }, { "epoch": 0.51, "learning_rate": 1.0185113366468431e-05, "loss": 0.5172, "step": 4188 }, { "epoch": 0.51, "learning_rate": 1.018117521988467e-05, "loss": 0.5453, "step": 4189 }, { "epoch": 0.51, "learning_rate": 1.0177237045193012e-05, "loss": 0.5031, "step": 4190 }, { "epoch": 0.51, "learning_rate": 1.0173298843004446e-05, "loss": 0.5655, "step": 4191 }, { "epoch": 0.51, "learning_rate": 1.0169360613929946e-05, "loss": 0.5698, "step": 4192 }, { "epoch": 0.51, "learning_rate": 1.01654223585805e-05, "loss": 0.5863, "step": 4193 }, { "epoch": 0.51, "learning_rate": 1.0161484077567092e-05, "loss": 0.5907, "step": 4194 }, { "epoch": 0.51, "learning_rate": 1.0157545771500719e-05, "loss": 0.5365, "step": 4195 }, { "epoch": 0.51, "learning_rate": 1.0153607440992372e-05, "loss": 0.5846, "step": 4196 }, { "epoch": 0.51, "learning_rate": 1.0149669086653062e-05, "loss": 0.5575, "step": 4197 }, { "epoch": 0.51, "learning_rate": 1.0145730709093785e-05, "loss": 0.59, "step": 4198 }, { "epoch": 0.51, "learning_rate": 1.0141792308925548e-05, "loss": 0.5758, "step": 4199 }, { "epoch": 0.51, "learning_rate": 1.0137853886759368e-05, "loss": 0.547, "step": 4200 }, { "epoch": 0.51, "learning_rate": 1.013391544320626e-05, "loss": 0.5297, "step": 4201 }, { "epoch": 0.51, "learning_rate": 1.0129976978877234e-05, "loss": 0.5415, "step": 4202 }, { "epoch": 0.51, "learning_rate": 1.0126038494383319e-05, "loss": 0.5206, "step": 4203 }, { "epoch": 0.51, "learning_rate": 1.012209999033554e-05, "loss": 0.5717, "step": 4204 }, { "epoch": 0.51, "learning_rate": 1.0118161467344915e-05, "loss": 0.5692, "step": 4205 }, { "epoch": 0.51, "learning_rate": 1.0114222926022485e-05, "loss": 0.5943, "step": 4206 }, { "epoch": 0.51, "learning_rate": 1.011028436697928e-05, "loss": 0.5919, "step": 4207 }, { "epoch": 0.51, "learning_rate": 1.0106345790826333e-05, "loss": 0.5923, "step": 4208 }, { "epoch": 0.51, "learning_rate": 1.0102407198174687e-05, "loss": 0.5549, "step": 4209 }, { "epoch": 0.51, "learning_rate": 1.0098468589635377e-05, "loss": 0.51, "step": 4210 }, { "epoch": 0.51, "learning_rate": 1.0094529965819458e-05, "loss": 0.5334, "step": 4211 }, { "epoch": 0.51, "learning_rate": 1.0090591327337964e-05, "loss": 0.5608, "step": 4212 }, { "epoch": 0.51, "learning_rate": 1.0086652674801948e-05, "loss": 0.5794, "step": 4213 }, { "epoch": 0.51, "learning_rate": 1.008271400882246e-05, "loss": 0.5003, "step": 4214 }, { "epoch": 0.51, "learning_rate": 1.0078775330010557e-05, "loss": 0.5938, "step": 4215 }, { "epoch": 0.51, "learning_rate": 1.0074836638977289e-05, "loss": 0.5202, "step": 4216 }, { "epoch": 0.51, "learning_rate": 1.0070897936333712e-05, "loss": 0.5736, "step": 4217 }, { "epoch": 0.51, "learning_rate": 1.0066959222690886e-05, "loss": 0.5278, "step": 4218 }, { "epoch": 0.51, "learning_rate": 1.0063020498659873e-05, "loss": 0.5589, "step": 4219 }, { "epoch": 0.51, "learning_rate": 1.0059081764851728e-05, "loss": 0.5826, "step": 4220 }, { "epoch": 0.51, "learning_rate": 1.0055143021877523e-05, "loss": 0.5654, "step": 4221 }, { "epoch": 0.51, "learning_rate": 1.0051204270348314e-05, "loss": 0.556, "step": 4222 }, { "epoch": 0.51, "learning_rate": 1.0047265510875173e-05, "loss": 0.5452, "step": 4223 }, { "epoch": 0.51, "learning_rate": 1.0043326744069166e-05, "loss": 0.5187, "step": 4224 }, { "epoch": 0.51, "learning_rate": 1.003938797054136e-05, "loss": 0.5284, "step": 4225 }, { "epoch": 0.51, "learning_rate": 1.0035449190902822e-05, "loss": 0.611, "step": 4226 }, { "epoch": 0.51, "learning_rate": 1.0031510405764628e-05, "loss": 0.5515, "step": 4227 }, { "epoch": 0.51, "learning_rate": 1.0027571615737844e-05, "loss": 0.5073, "step": 4228 }, { "epoch": 0.51, "learning_rate": 1.0023632821433544e-05, "loss": 0.5355, "step": 4229 }, { "epoch": 0.51, "learning_rate": 1.0019694023462801e-05, "loss": 0.5428, "step": 4230 }, { "epoch": 0.51, "learning_rate": 1.001575522243669e-05, "loss": 0.4915, "step": 4231 }, { "epoch": 0.51, "learning_rate": 1.0011816418966275e-05, "loss": 0.5917, "step": 4232 }, { "epoch": 0.51, "learning_rate": 1.000787761366264e-05, "loss": 0.562, "step": 4233 }, { "epoch": 0.51, "learning_rate": 1.000393880713686e-05, "loss": 0.5333, "step": 4234 }, { "epoch": 0.51, "learning_rate": 1e-05, "loss": 0.5507, "step": 4235 }, { "epoch": 0.52, "learning_rate": 9.996061192863143e-06, "loss": 0.559, "step": 4236 }, { "epoch": 0.52, "learning_rate": 9.992122386337362e-06, "loss": 0.5552, "step": 4237 }, { "epoch": 0.52, "learning_rate": 9.988183581033728e-06, "loss": 0.5404, "step": 4238 }, { "epoch": 0.52, "learning_rate": 9.984244777563317e-06, "loss": 0.5248, "step": 4239 }, { "epoch": 0.52, "learning_rate": 9.980305976537202e-06, "loss": 0.5357, "step": 4240 }, { "epoch": 0.52, "learning_rate": 9.976367178566461e-06, "loss": 0.5248, "step": 4241 }, { "epoch": 0.52, "learning_rate": 9.972428384262156e-06, "loss": 0.5557, "step": 4242 }, { "epoch": 0.52, "learning_rate": 9.968489594235376e-06, "loss": 0.6033, "step": 4243 }, { "epoch": 0.52, "learning_rate": 9.96455080909718e-06, "loss": 0.5031, "step": 4244 }, { "epoch": 0.52, "learning_rate": 9.960612029458642e-06, "loss": 0.5129, "step": 4245 }, { "epoch": 0.52, "learning_rate": 9.956673255930836e-06, "loss": 0.5929, "step": 4246 }, { "epoch": 0.52, "learning_rate": 9.95273448912483e-06, "loss": 0.5219, "step": 4247 }, { "epoch": 0.52, "learning_rate": 9.948795729651689e-06, "loss": 0.5143, "step": 4248 }, { "epoch": 0.52, "learning_rate": 9.944856978122482e-06, "loss": 0.563, "step": 4249 }, { "epoch": 0.52, "learning_rate": 9.940918235148273e-06, "loss": 0.5718, "step": 4250 }, { "epoch": 0.52, "learning_rate": 9.936979501340134e-06, "loss": 0.5433, "step": 4251 }, { "epoch": 0.52, "learning_rate": 9.93304077730912e-06, "loss": 0.5522, "step": 4252 }, { "epoch": 0.52, "learning_rate": 9.929102063666293e-06, "loss": 0.5246, "step": 4253 }, { "epoch": 0.52, "learning_rate": 9.925163361022713e-06, "loss": 0.5717, "step": 4254 }, { "epoch": 0.52, "learning_rate": 9.921224669989444e-06, "loss": 0.5369, "step": 4255 }, { "epoch": 0.52, "learning_rate": 9.91728599117754e-06, "loss": 0.6069, "step": 4256 }, { "epoch": 0.52, "learning_rate": 9.913347325198055e-06, "loss": 0.5823, "step": 4257 }, { "epoch": 0.52, "learning_rate": 9.90940867266204e-06, "loss": 0.5526, "step": 4258 }, { "epoch": 0.52, "learning_rate": 9.905470034180546e-06, "loss": 0.5365, "step": 4259 }, { "epoch": 0.52, "learning_rate": 9.901531410364625e-06, "loss": 0.6075, "step": 4260 }, { "epoch": 0.52, "learning_rate": 9.897592801825318e-06, "loss": 0.5523, "step": 4261 }, { "epoch": 0.52, "learning_rate": 9.89365420917367e-06, "loss": 0.5988, "step": 4262 }, { "epoch": 0.52, "learning_rate": 9.889715633020724e-06, "loss": 0.5635, "step": 4263 }, { "epoch": 0.52, "learning_rate": 9.88577707397752e-06, "loss": 0.5592, "step": 4264 }, { "epoch": 0.52, "learning_rate": 9.88183853265509e-06, "loss": 0.5667, "step": 4265 }, { "epoch": 0.52, "learning_rate": 9.877900009664465e-06, "loss": 0.5091, "step": 4266 }, { "epoch": 0.52, "learning_rate": 9.873961505616683e-06, "loss": 0.5444, "step": 4267 }, { "epoch": 0.52, "learning_rate": 9.870023021122767e-06, "loss": 0.5108, "step": 4268 }, { "epoch": 0.52, "learning_rate": 9.866084556793743e-06, "loss": 0.5331, "step": 4269 }, { "epoch": 0.52, "learning_rate": 9.862146113240633e-06, "loss": 0.5243, "step": 4270 }, { "epoch": 0.52, "learning_rate": 9.858207691074453e-06, "loss": 0.6031, "step": 4271 }, { "epoch": 0.52, "learning_rate": 9.854269290906219e-06, "loss": 0.5201, "step": 4272 }, { "epoch": 0.52, "learning_rate": 9.850330913346941e-06, "loss": 0.5663, "step": 4273 }, { "epoch": 0.52, "learning_rate": 9.846392559007631e-06, "loss": 0.4744, "step": 4274 }, { "epoch": 0.52, "learning_rate": 9.842454228499286e-06, "loss": 0.5135, "step": 4275 }, { "epoch": 0.52, "learning_rate": 9.838515922432912e-06, "loss": 0.5441, "step": 4276 }, { "epoch": 0.52, "learning_rate": 9.834577641419507e-06, "loss": 0.5328, "step": 4277 }, { "epoch": 0.52, "learning_rate": 9.830639386070054e-06, "loss": 0.5656, "step": 4278 }, { "epoch": 0.52, "learning_rate": 9.826701156995556e-06, "loss": 0.5298, "step": 4279 }, { "epoch": 0.52, "learning_rate": 9.822762954806988e-06, "loss": 0.5577, "step": 4280 }, { "epoch": 0.52, "learning_rate": 9.818824780115334e-06, "loss": 0.5311, "step": 4281 }, { "epoch": 0.52, "learning_rate": 9.81488663353157e-06, "loss": 0.5385, "step": 4282 }, { "epoch": 0.52, "learning_rate": 9.810948515666671e-06, "loss": 0.5317, "step": 4283 }, { "epoch": 0.52, "learning_rate": 9.8070104271316e-06, "loss": 0.5854, "step": 4284 }, { "epoch": 0.52, "learning_rate": 9.803072368537321e-06, "loss": 0.5605, "step": 4285 }, { "epoch": 0.52, "learning_rate": 9.799134340494792e-06, "loss": 0.52, "step": 4286 }, { "epoch": 0.52, "learning_rate": 9.795196343614971e-06, "loss": 0.6072, "step": 4287 }, { "epoch": 0.52, "learning_rate": 9.791258378508801e-06, "loss": 0.5262, "step": 4288 }, { "epoch": 0.52, "learning_rate": 9.78732044578723e-06, "loss": 0.588, "step": 4289 }, { "epoch": 0.52, "learning_rate": 9.783382546061192e-06, "loss": 0.525, "step": 4290 }, { "epoch": 0.52, "learning_rate": 9.779444679941624e-06, "loss": 0.5378, "step": 4291 }, { "epoch": 0.52, "learning_rate": 9.775506848039456e-06, "loss": 0.5584, "step": 4292 }, { "epoch": 0.52, "learning_rate": 9.771569050965613e-06, "loss": 0.6196, "step": 4293 }, { "epoch": 0.52, "learning_rate": 9.767631289331006e-06, "loss": 0.5401, "step": 4294 }, { "epoch": 0.52, "learning_rate": 9.763693563746551e-06, "loss": 0.5443, "step": 4295 }, { "epoch": 0.52, "learning_rate": 9.759755874823156e-06, "loss": 0.5239, "step": 4296 }, { "epoch": 0.52, "learning_rate": 9.75581822317172e-06, "loss": 0.532, "step": 4297 }, { "epoch": 0.52, "learning_rate": 9.751880609403137e-06, "loss": 0.5263, "step": 4298 }, { "epoch": 0.52, "learning_rate": 9.747943034128299e-06, "loss": 0.4949, "step": 4299 }, { "epoch": 0.52, "learning_rate": 9.74400549795809e-06, "loss": 0.6238, "step": 4300 }, { "epoch": 0.52, "learning_rate": 9.740068001503384e-06, "loss": 0.523, "step": 4301 }, { "epoch": 0.52, "learning_rate": 9.736130545375052e-06, "loss": 0.5385, "step": 4302 }, { "epoch": 0.52, "learning_rate": 9.732193130183964e-06, "loss": 0.5191, "step": 4303 }, { "epoch": 0.52, "learning_rate": 9.728255756540973e-06, "loss": 0.5512, "step": 4304 }, { "epoch": 0.52, "learning_rate": 9.724318425056935e-06, "loss": 0.5265, "step": 4305 }, { "epoch": 0.52, "learning_rate": 9.720381136342696e-06, "loss": 0.5449, "step": 4306 }, { "epoch": 0.52, "learning_rate": 9.716443891009092e-06, "loss": 0.5687, "step": 4307 }, { "epoch": 0.52, "learning_rate": 9.712506689666957e-06, "loss": 0.5912, "step": 4308 }, { "epoch": 0.52, "learning_rate": 9.708569532927114e-06, "loss": 0.5836, "step": 4309 }, { "epoch": 0.52, "learning_rate": 9.704632421400385e-06, "loss": 0.5606, "step": 4310 }, { "epoch": 0.52, "learning_rate": 9.700695355697578e-06, "loss": 0.5412, "step": 4311 }, { "epoch": 0.52, "learning_rate": 9.696758336429498e-06, "loss": 0.5393, "step": 4312 }, { "epoch": 0.52, "learning_rate": 9.692821364206944e-06, "loss": 0.5303, "step": 4313 }, { "epoch": 0.52, "learning_rate": 9.688884439640702e-06, "loss": 0.5191, "step": 4314 }, { "epoch": 0.52, "learning_rate": 9.68494756334156e-06, "loss": 0.5769, "step": 4315 }, { "epoch": 0.52, "learning_rate": 9.681010735920293e-06, "loss": 0.5705, "step": 4316 }, { "epoch": 0.52, "learning_rate": 9.677073957987662e-06, "loss": 0.5666, "step": 4317 }, { "epoch": 0.53, "learning_rate": 9.673137230154432e-06, "loss": 0.5686, "step": 4318 }, { "epoch": 0.53, "learning_rate": 9.669200553031354e-06, "loss": 0.5856, "step": 4319 }, { "epoch": 0.53, "learning_rate": 9.66526392722917e-06, "loss": 0.5241, "step": 4320 }, { "epoch": 0.53, "learning_rate": 9.661327353358617e-06, "loss": 0.5665, "step": 4321 }, { "epoch": 0.53, "learning_rate": 9.657390832030424e-06, "loss": 0.586, "step": 4322 }, { "epoch": 0.53, "learning_rate": 9.653454363855312e-06, "loss": 0.4972, "step": 4323 }, { "epoch": 0.53, "learning_rate": 9.649517949443989e-06, "loss": 0.5548, "step": 4324 }, { "epoch": 0.53, "learning_rate": 9.64558158940716e-06, "loss": 0.5533, "step": 4325 }, { "epoch": 0.53, "learning_rate": 9.641645284355517e-06, "loss": 0.4918, "step": 4326 }, { "epoch": 0.53, "learning_rate": 9.637709034899752e-06, "loss": 0.5575, "step": 4327 }, { "epoch": 0.53, "learning_rate": 9.633772841650542e-06, "loss": 0.5932, "step": 4328 }, { "epoch": 0.53, "learning_rate": 9.629836705218555e-06, "loss": 0.5036, "step": 4329 }, { "epoch": 0.53, "learning_rate": 9.625900626214447e-06, "loss": 0.4901, "step": 4330 }, { "epoch": 0.53, "learning_rate": 9.621964605248873e-06, "loss": 0.5274, "step": 4331 }, { "epoch": 0.53, "learning_rate": 9.618028642932476e-06, "loss": 0.5929, "step": 4332 }, { "epoch": 0.53, "learning_rate": 9.614092739875888e-06, "loss": 0.5573, "step": 4333 }, { "epoch": 0.53, "learning_rate": 9.610156896689733e-06, "loss": 0.5854, "step": 4334 }, { "epoch": 0.53, "learning_rate": 9.606221113984624e-06, "loss": 0.577, "step": 4335 }, { "epoch": 0.53, "learning_rate": 9.60228539237117e-06, "loss": 0.4595, "step": 4336 }, { "epoch": 0.53, "learning_rate": 9.598349732459964e-06, "loss": 0.5896, "step": 4337 }, { "epoch": 0.53, "learning_rate": 9.594414134861588e-06, "loss": 0.5709, "step": 4338 }, { "epoch": 0.53, "learning_rate": 9.590478600186629e-06, "loss": 0.5264, "step": 4339 }, { "epoch": 0.53, "learning_rate": 9.586543129045647e-06, "loss": 0.5808, "step": 4340 }, { "epoch": 0.53, "learning_rate": 9.5826077220492e-06, "loss": 0.5657, "step": 4341 }, { "epoch": 0.53, "learning_rate": 9.578672379807838e-06, "loss": 0.5585, "step": 4342 }, { "epoch": 0.53, "learning_rate": 9.57473710293209e-06, "loss": 0.5779, "step": 4343 }, { "epoch": 0.53, "learning_rate": 9.570801892032492e-06, "loss": 0.586, "step": 4344 }, { "epoch": 0.53, "learning_rate": 9.566866747719554e-06, "loss": 0.6265, "step": 4345 }, { "epoch": 0.53, "learning_rate": 9.562931670603788e-06, "loss": 0.5488, "step": 4346 }, { "epoch": 0.53, "learning_rate": 9.558996661295685e-06, "loss": 0.52, "step": 4347 }, { "epoch": 0.53, "learning_rate": 9.555061720405732e-06, "loss": 0.4797, "step": 4348 }, { "epoch": 0.53, "learning_rate": 9.551126848544402e-06, "loss": 0.5274, "step": 4349 }, { "epoch": 0.53, "learning_rate": 9.54719204632216e-06, "loss": 0.5967, "step": 4350 }, { "epoch": 0.53, "learning_rate": 9.543257314349461e-06, "loss": 0.5681, "step": 4351 }, { "epoch": 0.53, "learning_rate": 9.53932265323675e-06, "loss": 0.5739, "step": 4352 }, { "epoch": 0.53, "learning_rate": 9.53538806359445e-06, "loss": 0.4778, "step": 4353 }, { "epoch": 0.53, "learning_rate": 9.531453546032986e-06, "loss": 0.4926, "step": 4354 }, { "epoch": 0.53, "learning_rate": 9.527519101162766e-06, "loss": 0.536, "step": 4355 }, { "epoch": 0.53, "learning_rate": 9.523584729594193e-06, "loss": 0.5719, "step": 4356 }, { "epoch": 0.53, "learning_rate": 9.519650431937644e-06, "loss": 0.5757, "step": 4357 }, { "epoch": 0.53, "learning_rate": 9.5157162088035e-06, "loss": 0.5746, "step": 4358 }, { "epoch": 0.53, "learning_rate": 9.511782060802124e-06, "loss": 0.5953, "step": 4359 }, { "epoch": 0.53, "learning_rate": 9.507847988543866e-06, "loss": 0.5206, "step": 4360 }, { "epoch": 0.53, "learning_rate": 9.503913992639064e-06, "loss": 0.5919, "step": 4361 }, { "epoch": 0.53, "learning_rate": 9.499980073698049e-06, "loss": 0.5283, "step": 4362 }, { "epoch": 0.53, "learning_rate": 9.496046232331137e-06, "loss": 0.5241, "step": 4363 }, { "epoch": 0.53, "learning_rate": 9.492112469148633e-06, "loss": 0.507, "step": 4364 }, { "epoch": 0.53, "learning_rate": 9.48817878476083e-06, "loss": 0.5814, "step": 4365 }, { "epoch": 0.53, "learning_rate": 9.484245179778003e-06, "loss": 0.5297, "step": 4366 }, { "epoch": 0.53, "learning_rate": 9.48031165481042e-06, "loss": 0.5105, "step": 4367 }, { "epoch": 0.53, "learning_rate": 9.47637821046834e-06, "loss": 0.5906, "step": 4368 }, { "epoch": 0.53, "learning_rate": 9.472444847362007e-06, "loss": 0.5318, "step": 4369 }, { "epoch": 0.53, "learning_rate": 9.468511566101644e-06, "loss": 0.5398, "step": 4370 }, { "epoch": 0.53, "learning_rate": 9.464578367297471e-06, "loss": 0.5943, "step": 4371 }, { "epoch": 0.53, "learning_rate": 9.460645251559694e-06, "loss": 0.5443, "step": 4372 }, { "epoch": 0.53, "learning_rate": 9.456712219498505e-06, "loss": 0.6003, "step": 4373 }, { "epoch": 0.53, "learning_rate": 9.452779271724074e-06, "loss": 0.5773, "step": 4374 }, { "epoch": 0.53, "learning_rate": 9.448846408846582e-06, "loss": 0.5666, "step": 4375 }, { "epoch": 0.53, "learning_rate": 9.44491363147617e-06, "loss": 0.4848, "step": 4376 }, { "epoch": 0.53, "learning_rate": 9.44098094022298e-06, "loss": 0.57, "step": 4377 }, { "epoch": 0.53, "learning_rate": 9.437048335697139e-06, "loss": 0.5485, "step": 4378 }, { "epoch": 0.53, "learning_rate": 9.433115818508758e-06, "loss": 0.5866, "step": 4379 }, { "epoch": 0.53, "learning_rate": 9.429183389267933e-06, "loss": 0.5739, "step": 4380 }, { "epoch": 0.53, "learning_rate": 9.425251048584751e-06, "loss": 0.5262, "step": 4381 }, { "epoch": 0.53, "learning_rate": 9.421318797069288e-06, "loss": 0.4714, "step": 4382 }, { "epoch": 0.53, "learning_rate": 9.417386635331594e-06, "loss": 0.6037, "step": 4383 }, { "epoch": 0.53, "learning_rate": 9.413454563981717e-06, "loss": 0.6088, "step": 4384 }, { "epoch": 0.53, "learning_rate": 9.409522583629684e-06, "loss": 0.5301, "step": 4385 }, { "epoch": 0.53, "learning_rate": 9.40559069488551e-06, "loss": 0.604, "step": 4386 }, { "epoch": 0.53, "learning_rate": 9.4016588983592e-06, "loss": 0.5294, "step": 4387 }, { "epoch": 0.53, "learning_rate": 9.39772719466074e-06, "loss": 0.6274, "step": 4388 }, { "epoch": 0.53, "learning_rate": 9.3937955844001e-06, "loss": 0.5567, "step": 4389 }, { "epoch": 0.53, "learning_rate": 9.38986406818724e-06, "loss": 0.5294, "step": 4390 }, { "epoch": 0.53, "learning_rate": 9.385932646632103e-06, "loss": 0.571, "step": 4391 }, { "epoch": 0.53, "learning_rate": 9.382001320344617e-06, "loss": 0.5227, "step": 4392 }, { "epoch": 0.53, "learning_rate": 9.378070089934696e-06, "loss": 0.5245, "step": 4393 }, { "epoch": 0.53, "learning_rate": 9.374138956012238e-06, "loss": 0.6012, "step": 4394 }, { "epoch": 0.53, "learning_rate": 9.370207919187128e-06, "loss": 0.5721, "step": 4395 }, { "epoch": 0.53, "learning_rate": 9.366276980069238e-06, "loss": 0.5275, "step": 4396 }, { "epoch": 0.53, "learning_rate": 9.362346139268416e-06, "loss": 0.6294, "step": 4397 }, { "epoch": 0.53, "learning_rate": 9.358415397394504e-06, "loss": 0.4931, "step": 4398 }, { "epoch": 0.53, "learning_rate": 9.354484755057325e-06, "loss": 0.5723, "step": 4399 }, { "epoch": 0.54, "learning_rate": 9.350554212866687e-06, "loss": 0.5002, "step": 4400 }, { "epoch": 0.54, "learning_rate": 9.346623771432384e-06, "loss": 0.5131, "step": 4401 }, { "epoch": 0.54, "learning_rate": 9.342693431364189e-06, "loss": 0.5498, "step": 4402 }, { "epoch": 0.54, "learning_rate": 9.338763193271865e-06, "loss": 0.5553, "step": 4403 }, { "epoch": 0.54, "learning_rate": 9.334833057765156e-06, "loss": 0.547, "step": 4404 }, { "epoch": 0.54, "learning_rate": 9.330903025453793e-06, "loss": 0.5389, "step": 4405 }, { "epoch": 0.54, "learning_rate": 9.326973096947486e-06, "loss": 0.5479, "step": 4406 }, { "epoch": 0.54, "learning_rate": 9.323043272855936e-06, "loss": 0.5331, "step": 4407 }, { "epoch": 0.54, "learning_rate": 9.31911355378882e-06, "loss": 0.5345, "step": 4408 }, { "epoch": 0.54, "learning_rate": 9.315183940355809e-06, "loss": 0.5985, "step": 4409 }, { "epoch": 0.54, "learning_rate": 9.311254433166538e-06, "loss": 0.6345, "step": 4410 }, { "epoch": 0.54, "learning_rate": 9.307325032830655e-06, "loss": 0.5855, "step": 4411 }, { "epoch": 0.54, "learning_rate": 9.303395739957763e-06, "loss": 0.503, "step": 4412 }, { "epoch": 0.54, "learning_rate": 9.299466555157467e-06, "loss": 0.5129, "step": 4413 }, { "epoch": 0.54, "learning_rate": 9.295537479039347e-06, "loss": 0.559, "step": 4414 }, { "epoch": 0.54, "learning_rate": 9.291608512212969e-06, "loss": 0.5922, "step": 4415 }, { "epoch": 0.54, "learning_rate": 9.287679655287876e-06, "loss": 0.581, "step": 4416 }, { "epoch": 0.54, "learning_rate": 9.283750908873602e-06, "loss": 0.6057, "step": 4417 }, { "epoch": 0.54, "learning_rate": 9.279822273579664e-06, "loss": 0.4798, "step": 4418 }, { "epoch": 0.54, "learning_rate": 9.275893750015552e-06, "loss": 0.4629, "step": 4419 }, { "epoch": 0.54, "learning_rate": 9.271965338790749e-06, "loss": 0.5332, "step": 4420 }, { "epoch": 0.54, "learning_rate": 9.268037040514715e-06, "loss": 0.5605, "step": 4421 }, { "epoch": 0.54, "learning_rate": 9.264108855796894e-06, "loss": 0.5512, "step": 4422 }, { "epoch": 0.54, "learning_rate": 9.260180785246715e-06, "loss": 0.6012, "step": 4423 }, { "epoch": 0.54, "learning_rate": 9.256252829473587e-06, "loss": 0.5504, "step": 4424 }, { "epoch": 0.54, "learning_rate": 9.252324989086898e-06, "loss": 0.5523, "step": 4425 }, { "epoch": 0.54, "learning_rate": 9.248397264696024e-06, "loss": 0.5952, "step": 4426 }, { "epoch": 0.54, "learning_rate": 9.244469656910315e-06, "loss": 0.5414, "step": 4427 }, { "epoch": 0.54, "learning_rate": 9.240542166339116e-06, "loss": 0.6217, "step": 4428 }, { "epoch": 0.54, "learning_rate": 9.23661479359174e-06, "loss": 0.5443, "step": 4429 }, { "epoch": 0.54, "learning_rate": 9.232687539277487e-06, "loss": 0.5208, "step": 4430 }, { "epoch": 0.54, "learning_rate": 9.228760404005642e-06, "loss": 0.6209, "step": 4431 }, { "epoch": 0.54, "learning_rate": 9.224833388385471e-06, "loss": 0.5058, "step": 4432 }, { "epoch": 0.54, "learning_rate": 9.220906493026211e-06, "loss": 0.5099, "step": 4433 }, { "epoch": 0.54, "learning_rate": 9.216979718537095e-06, "loss": 0.5168, "step": 4434 }, { "epoch": 0.54, "learning_rate": 9.213053065527329e-06, "loss": 0.5305, "step": 4435 }, { "epoch": 0.54, "learning_rate": 9.209126534606102e-06, "loss": 0.5558, "step": 4436 }, { "epoch": 0.54, "learning_rate": 9.205200126382586e-06, "loss": 0.552, "step": 4437 }, { "epoch": 0.54, "learning_rate": 9.20127384146593e-06, "loss": 0.5649, "step": 4438 }, { "epoch": 0.54, "learning_rate": 9.197347680465264e-06, "loss": 0.5535, "step": 4439 }, { "epoch": 0.54, "learning_rate": 9.193421643989704e-06, "loss": 0.455, "step": 4440 }, { "epoch": 0.54, "learning_rate": 9.189495732648343e-06, "loss": 0.569, "step": 4441 }, { "epoch": 0.54, "learning_rate": 9.18556994705025e-06, "loss": 0.5042, "step": 4442 }, { "epoch": 0.54, "learning_rate": 9.181644287804483e-06, "loss": 0.5642, "step": 4443 }, { "epoch": 0.54, "learning_rate": 9.177718755520078e-06, "loss": 0.5113, "step": 4444 }, { "epoch": 0.54, "learning_rate": 9.17379335080605e-06, "loss": 0.5298, "step": 4445 }, { "epoch": 0.54, "learning_rate": 9.169868074271384e-06, "loss": 0.55, "step": 4446 }, { "epoch": 0.54, "learning_rate": 9.165942926525073e-06, "loss": 0.5326, "step": 4447 }, { "epoch": 0.54, "learning_rate": 9.16201790817606e-06, "loss": 0.5899, "step": 4448 }, { "epoch": 0.54, "learning_rate": 9.158093019833284e-06, "loss": 0.5313, "step": 4449 }, { "epoch": 0.54, "learning_rate": 9.15416826210566e-06, "loss": 0.5344, "step": 4450 }, { "epoch": 0.54, "learning_rate": 9.150243635602085e-06, "loss": 0.5156, "step": 4451 }, { "epoch": 0.54, "learning_rate": 9.146319140931428e-06, "loss": 0.5424, "step": 4452 }, { "epoch": 0.54, "learning_rate": 9.142394778702546e-06, "loss": 0.5728, "step": 4453 }, { "epoch": 0.54, "learning_rate": 9.138470549524275e-06, "loss": 0.5395, "step": 4454 }, { "epoch": 0.54, "learning_rate": 9.134546454005426e-06, "loss": 0.6066, "step": 4455 }, { "epoch": 0.54, "learning_rate": 9.130622492754789e-06, "loss": 0.6005, "step": 4456 }, { "epoch": 0.54, "learning_rate": 9.126698666381135e-06, "loss": 0.5385, "step": 4457 }, { "epoch": 0.54, "learning_rate": 9.122774975493216e-06, "loss": 0.5724, "step": 4458 }, { "epoch": 0.54, "learning_rate": 9.118851420699763e-06, "loss": 0.5247, "step": 4459 }, { "epoch": 0.54, "learning_rate": 9.114928002609483e-06, "loss": 0.5696, "step": 4460 }, { "epoch": 0.54, "learning_rate": 9.111004721831065e-06, "loss": 0.4857, "step": 4461 }, { "epoch": 0.54, "learning_rate": 9.107081578973172e-06, "loss": 0.5026, "step": 4462 }, { "epoch": 0.54, "learning_rate": 9.103158574644446e-06, "loss": 0.6107, "step": 4463 }, { "epoch": 0.54, "learning_rate": 9.099235709453517e-06, "loss": 0.5354, "step": 4464 }, { "epoch": 0.54, "learning_rate": 9.095312984008978e-06, "loss": 0.5451, "step": 4465 }, { "epoch": 0.54, "learning_rate": 9.091390398919414e-06, "loss": 0.5085, "step": 4466 }, { "epoch": 0.54, "learning_rate": 9.08746795479338e-06, "loss": 0.5761, "step": 4467 }, { "epoch": 0.54, "learning_rate": 9.083545652239417e-06, "loss": 0.5251, "step": 4468 }, { "epoch": 0.54, "learning_rate": 9.079623491866032e-06, "loss": 0.5479, "step": 4469 }, { "epoch": 0.54, "learning_rate": 9.07570147428172e-06, "loss": 0.5509, "step": 4470 }, { "epoch": 0.54, "learning_rate": 9.07177960009495e-06, "loss": 0.5654, "step": 4471 }, { "epoch": 0.54, "learning_rate": 9.067857869914171e-06, "loss": 0.5401, "step": 4472 }, { "epoch": 0.54, "learning_rate": 9.063936284347807e-06, "loss": 0.5481, "step": 4473 }, { "epoch": 0.54, "learning_rate": 9.060014844004265e-06, "loss": 0.514, "step": 4474 }, { "epoch": 0.54, "learning_rate": 9.056093549491918e-06, "loss": 0.4885, "step": 4475 }, { "epoch": 0.54, "learning_rate": 9.052172401419128e-06, "loss": 0.5363, "step": 4476 }, { "epoch": 0.54, "learning_rate": 9.048251400394228e-06, "loss": 0.5488, "step": 4477 }, { "epoch": 0.54, "learning_rate": 9.044330547025532e-06, "loss": 0.5783, "step": 4478 }, { "epoch": 0.54, "learning_rate": 9.040409841921327e-06, "loss": 0.5224, "step": 4479 }, { "epoch": 0.54, "learning_rate": 9.03648928568988e-06, "loss": 0.5456, "step": 4480 }, { "epoch": 0.54, "learning_rate": 9.032568878939436e-06, "loss": 0.4574, "step": 4481 }, { "epoch": 0.55, "learning_rate": 9.028648622278206e-06, "loss": 0.5662, "step": 4482 }, { "epoch": 0.55, "learning_rate": 9.0247285163144e-06, "loss": 0.5265, "step": 4483 }, { "epoch": 0.55, "learning_rate": 9.020808561656182e-06, "loss": 0.5492, "step": 4484 }, { "epoch": 0.55, "learning_rate": 9.016888758911706e-06, "loss": 0.504, "step": 4485 }, { "epoch": 0.55, "learning_rate": 9.012969108689094e-06, "loss": 0.4952, "step": 4486 }, { "epoch": 0.55, "learning_rate": 9.009049611596455e-06, "loss": 0.5942, "step": 4487 }, { "epoch": 0.55, "learning_rate": 9.005130268241861e-06, "loss": 0.5109, "step": 4488 }, { "epoch": 0.55, "learning_rate": 9.001211079233369e-06, "loss": 0.549, "step": 4489 }, { "epoch": 0.55, "learning_rate": 8.99729204517901e-06, "loss": 0.5469, "step": 4490 }, { "epoch": 0.55, "learning_rate": 8.993373166686794e-06, "loss": 0.4767, "step": 4491 }, { "epoch": 0.55, "learning_rate": 8.989454444364699e-06, "loss": 0.5631, "step": 4492 }, { "epoch": 0.55, "learning_rate": 8.985535878820686e-06, "loss": 0.5555, "step": 4493 }, { "epoch": 0.55, "learning_rate": 8.981617470662685e-06, "loss": 0.5389, "step": 4494 }, { "epoch": 0.55, "learning_rate": 8.977699220498612e-06, "loss": 0.5123, "step": 4495 }, { "epoch": 0.55, "learning_rate": 8.97378112893635e-06, "loss": 0.5347, "step": 4496 }, { "epoch": 0.55, "learning_rate": 8.969863196583761e-06, "loss": 0.544, "step": 4497 }, { "epoch": 0.55, "learning_rate": 8.96594542404868e-06, "loss": 0.5712, "step": 4498 }, { "epoch": 0.55, "learning_rate": 8.962027811938914e-06, "loss": 0.5048, "step": 4499 }, { "epoch": 0.55, "learning_rate": 8.958110360862255e-06, "loss": 0.4696, "step": 4500 }, { "epoch": 0.55, "learning_rate": 8.95419307142646e-06, "loss": 0.4941, "step": 4501 }, { "epoch": 0.55, "learning_rate": 8.950275944239268e-06, "loss": 0.563, "step": 4502 }, { "epoch": 0.55, "learning_rate": 8.94635897990839e-06, "loss": 0.5875, "step": 4503 }, { "epoch": 0.55, "learning_rate": 8.942442179041512e-06, "loss": 0.4957, "step": 4504 }, { "epoch": 0.55, "learning_rate": 8.93852554224629e-06, "loss": 0.5534, "step": 4505 }, { "epoch": 0.55, "learning_rate": 8.934609070130363e-06, "loss": 0.5316, "step": 4506 }, { "epoch": 0.55, "learning_rate": 8.93069276330134e-06, "loss": 0.4866, "step": 4507 }, { "epoch": 0.55, "learning_rate": 8.926776622366806e-06, "loss": 0.5508, "step": 4508 }, { "epoch": 0.55, "learning_rate": 8.922860647934315e-06, "loss": 0.5699, "step": 4509 }, { "epoch": 0.55, "learning_rate": 8.918944840611405e-06, "loss": 0.5332, "step": 4510 }, { "epoch": 0.55, "learning_rate": 8.915029201005578e-06, "loss": 0.5938, "step": 4511 }, { "epoch": 0.55, "learning_rate": 8.911113729724314e-06, "loss": 0.6235, "step": 4512 }, { "epoch": 0.55, "learning_rate": 8.907198427375069e-06, "loss": 0.5161, "step": 4513 }, { "epoch": 0.55, "learning_rate": 8.903283294565272e-06, "loss": 0.5809, "step": 4514 }, { "epoch": 0.55, "learning_rate": 8.899368331902322e-06, "loss": 0.5418, "step": 4515 }, { "epoch": 0.55, "learning_rate": 8.895453539993593e-06, "loss": 0.5653, "step": 4516 }, { "epoch": 0.55, "learning_rate": 8.891538919446438e-06, "loss": 0.5071, "step": 4517 }, { "epoch": 0.55, "learning_rate": 8.887624470868175e-06, "loss": 0.5108, "step": 4518 }, { "epoch": 0.55, "learning_rate": 8.883710194866103e-06, "loss": 0.5425, "step": 4519 }, { "epoch": 0.55, "learning_rate": 8.87979609204749e-06, "loss": 0.5643, "step": 4520 }, { "epoch": 0.55, "learning_rate": 8.875882163019576e-06, "loss": 0.5942, "step": 4521 }, { "epoch": 0.55, "learning_rate": 8.871968408389578e-06, "loss": 0.497, "step": 4522 }, { "epoch": 0.55, "learning_rate": 8.868054828764683e-06, "loss": 0.5414, "step": 4523 }, { "epoch": 0.55, "learning_rate": 8.86414142475205e-06, "loss": 0.5775, "step": 4524 }, { "epoch": 0.55, "learning_rate": 8.860228196958813e-06, "loss": 0.5703, "step": 4525 }, { "epoch": 0.55, "learning_rate": 8.856315145992081e-06, "loss": 0.5222, "step": 4526 }, { "epoch": 0.55, "learning_rate": 8.85240227245893e-06, "loss": 0.5611, "step": 4527 }, { "epoch": 0.55, "learning_rate": 8.84848957696641e-06, "loss": 0.5018, "step": 4528 }, { "epoch": 0.55, "learning_rate": 8.844577060121549e-06, "loss": 0.5361, "step": 4529 }, { "epoch": 0.55, "learning_rate": 8.840664722531335e-06, "loss": 0.4944, "step": 4530 }, { "epoch": 0.55, "learning_rate": 8.836752564802743e-06, "loss": 0.5114, "step": 4531 }, { "epoch": 0.55, "learning_rate": 8.832840587542712e-06, "loss": 0.5268, "step": 4532 }, { "epoch": 0.55, "learning_rate": 8.828928791358154e-06, "loss": 0.5148, "step": 4533 }, { "epoch": 0.55, "learning_rate": 8.82501717685595e-06, "loss": 0.5494, "step": 4534 }, { "epoch": 0.55, "learning_rate": 8.821105744642959e-06, "loss": 0.5934, "step": 4535 }, { "epoch": 0.55, "learning_rate": 8.817194495326007e-06, "loss": 0.5051, "step": 4536 }, { "epoch": 0.55, "learning_rate": 8.813283429511896e-06, "loss": 0.5323, "step": 4537 }, { "epoch": 0.55, "learning_rate": 8.80937254780739e-06, "loss": 0.5422, "step": 4538 }, { "epoch": 0.55, "learning_rate": 8.805461850819236e-06, "loss": 0.6164, "step": 4539 }, { "epoch": 0.55, "learning_rate": 8.801551339154148e-06, "loss": 0.5885, "step": 4540 }, { "epoch": 0.55, "learning_rate": 8.797641013418809e-06, "loss": 0.5879, "step": 4541 }, { "epoch": 0.55, "learning_rate": 8.79373087421987e-06, "loss": 0.5225, "step": 4542 }, { "epoch": 0.55, "learning_rate": 8.789820922163968e-06, "loss": 0.534, "step": 4543 }, { "epoch": 0.55, "learning_rate": 8.785911157857695e-06, "loss": 0.4824, "step": 4544 }, { "epoch": 0.55, "learning_rate": 8.78200158190762e-06, "loss": 0.5849, "step": 4545 }, { "epoch": 0.55, "learning_rate": 8.778092194920285e-06, "loss": 0.5412, "step": 4546 }, { "epoch": 0.55, "learning_rate": 8.774182997502198e-06, "loss": 0.5756, "step": 4547 }, { "epoch": 0.55, "learning_rate": 8.77027399025984e-06, "loss": 0.5339, "step": 4548 }, { "epoch": 0.55, "learning_rate": 8.76636517379966e-06, "loss": 0.5167, "step": 4549 }, { "epoch": 0.55, "learning_rate": 8.762456548728087e-06, "loss": 0.5562, "step": 4550 }, { "epoch": 0.55, "learning_rate": 8.758548115651504e-06, "loss": 0.5687, "step": 4551 }, { "epoch": 0.55, "learning_rate": 8.754639875176278e-06, "loss": 0.5177, "step": 4552 }, { "epoch": 0.55, "learning_rate": 8.750731827908741e-06, "loss": 0.4968, "step": 4553 }, { "epoch": 0.55, "learning_rate": 8.746823974455194e-06, "loss": 0.5148, "step": 4554 }, { "epoch": 0.55, "learning_rate": 8.74291631542191e-06, "loss": 0.5777, "step": 4555 }, { "epoch": 0.55, "learning_rate": 8.739008851415134e-06, "loss": 0.5432, "step": 4556 }, { "epoch": 0.55, "learning_rate": 8.735101583041075e-06, "loss": 0.544, "step": 4557 }, { "epoch": 0.55, "learning_rate": 8.731194510905914e-06, "loss": 0.5342, "step": 4558 }, { "epoch": 0.55, "learning_rate": 8.727287635615802e-06, "loss": 0.5402, "step": 4559 }, { "epoch": 0.55, "learning_rate": 8.723380957776864e-06, "loss": 0.5725, "step": 4560 }, { "epoch": 0.55, "learning_rate": 8.719474477995182e-06, "loss": 0.5342, "step": 4561 }, { "epoch": 0.55, "learning_rate": 8.71556819687682e-06, "loss": 0.5868, "step": 4562 }, { "epoch": 0.55, "learning_rate": 8.71166211502781e-06, "loss": 0.5632, "step": 4563 }, { "epoch": 0.56, "learning_rate": 8.707756233054142e-06, "loss": 0.5229, "step": 4564 }, { "epoch": 0.56, "learning_rate": 8.703850551561785e-06, "loss": 0.579, "step": 4565 }, { "epoch": 0.56, "learning_rate": 8.699945071156674e-06, "loss": 0.5579, "step": 4566 }, { "epoch": 0.56, "learning_rate": 8.696039792444717e-06, "loss": 0.5706, "step": 4567 }, { "epoch": 0.56, "learning_rate": 8.692134716031783e-06, "loss": 0.5415, "step": 4568 }, { "epoch": 0.56, "learning_rate": 8.688229842523717e-06, "loss": 0.5582, "step": 4569 }, { "epoch": 0.56, "learning_rate": 8.684325172526323e-06, "loss": 0.5472, "step": 4570 }, { "epoch": 0.56, "learning_rate": 8.680420706645385e-06, "loss": 0.5768, "step": 4571 }, { "epoch": 0.56, "learning_rate": 8.676516445486646e-06, "loss": 0.5358, "step": 4572 }, { "epoch": 0.56, "learning_rate": 8.672612389655825e-06, "loss": 0.5123, "step": 4573 }, { "epoch": 0.56, "learning_rate": 8.6687085397586e-06, "loss": 0.5738, "step": 4574 }, { "epoch": 0.56, "learning_rate": 8.664804896400628e-06, "loss": 0.5376, "step": 4575 }, { "epoch": 0.56, "learning_rate": 8.660901460187521e-06, "loss": 0.6034, "step": 4576 }, { "epoch": 0.56, "learning_rate": 8.656998231724873e-06, "loss": 0.4944, "step": 4577 }, { "epoch": 0.56, "learning_rate": 8.65309521161823e-06, "loss": 0.5048, "step": 4578 }, { "epoch": 0.56, "learning_rate": 8.649192400473127e-06, "loss": 0.5728, "step": 4579 }, { "epoch": 0.56, "learning_rate": 8.645289798895045e-06, "loss": 0.5816, "step": 4580 }, { "epoch": 0.56, "learning_rate": 8.641387407489442e-06, "loss": 0.5512, "step": 4581 }, { "epoch": 0.56, "learning_rate": 8.63748522686175e-06, "loss": 0.5297, "step": 4582 }, { "epoch": 0.56, "learning_rate": 8.633583257617351e-06, "loss": 0.539, "step": 4583 }, { "epoch": 0.56, "learning_rate": 8.62968150036161e-06, "loss": 0.56, "step": 4584 }, { "epoch": 0.56, "learning_rate": 8.625779955699852e-06, "loss": 0.5208, "step": 4585 }, { "epoch": 0.56, "learning_rate": 8.621878624237374e-06, "loss": 0.5611, "step": 4586 }, { "epoch": 0.56, "learning_rate": 8.617977506579431e-06, "loss": 0.5014, "step": 4587 }, { "epoch": 0.56, "learning_rate": 8.614076603331254e-06, "loss": 0.5423, "step": 4588 }, { "epoch": 0.56, "learning_rate": 8.610175915098034e-06, "loss": 0.5188, "step": 4589 }, { "epoch": 0.56, "learning_rate": 8.606275442484934e-06, "loss": 0.5531, "step": 4590 }, { "epoch": 0.56, "learning_rate": 8.602375186097081e-06, "loss": 0.567, "step": 4591 }, { "epoch": 0.56, "learning_rate": 8.59847514653957e-06, "loss": 0.5331, "step": 4592 }, { "epoch": 0.56, "learning_rate": 8.59457532441746e-06, "loss": 0.4971, "step": 4593 }, { "epoch": 0.56, "learning_rate": 8.590675720335775e-06, "loss": 0.4925, "step": 4594 }, { "epoch": 0.56, "learning_rate": 8.586776334899512e-06, "loss": 0.5127, "step": 4595 }, { "epoch": 0.56, "learning_rate": 8.582877168713626e-06, "loss": 0.5566, "step": 4596 }, { "epoch": 0.56, "learning_rate": 8.57897822238304e-06, "loss": 0.6074, "step": 4597 }, { "epoch": 0.56, "learning_rate": 8.575079496512648e-06, "loss": 0.5411, "step": 4598 }, { "epoch": 0.56, "learning_rate": 8.571180991707307e-06, "loss": 0.5086, "step": 4599 }, { "epoch": 0.56, "learning_rate": 8.567282708571835e-06, "loss": 0.59, "step": 4600 }, { "epoch": 0.56, "learning_rate": 8.563384647711022e-06, "loss": 0.5758, "step": 4601 }, { "epoch": 0.56, "learning_rate": 8.559486809729618e-06, "loss": 0.5757, "step": 4602 }, { "epoch": 0.56, "learning_rate": 8.555589195232346e-06, "loss": 0.549, "step": 4603 }, { "epoch": 0.56, "learning_rate": 8.551691804823888e-06, "loss": 0.563, "step": 4604 }, { "epoch": 0.56, "learning_rate": 8.547794639108893e-06, "loss": 0.5731, "step": 4605 }, { "epoch": 0.56, "learning_rate": 8.543897698691974e-06, "loss": 0.5375, "step": 4606 }, { "epoch": 0.56, "learning_rate": 8.540000984177711e-06, "loss": 0.5687, "step": 4607 }, { "epoch": 0.56, "learning_rate": 8.536104496170649e-06, "loss": 0.5379, "step": 4608 }, { "epoch": 0.56, "learning_rate": 8.532208235275298e-06, "loss": 0.624, "step": 4609 }, { "epoch": 0.56, "learning_rate": 8.528312202096129e-06, "loss": 0.4937, "step": 4610 }, { "epoch": 0.56, "learning_rate": 8.524416397237581e-06, "loss": 0.582, "step": 4611 }, { "epoch": 0.56, "learning_rate": 8.520520821304058e-06, "loss": 0.6003, "step": 4612 }, { "epoch": 0.56, "learning_rate": 8.516625474899929e-06, "loss": 0.5687, "step": 4613 }, { "epoch": 0.56, "learning_rate": 8.512730358629517e-06, "loss": 0.5344, "step": 4614 }, { "epoch": 0.56, "learning_rate": 8.508835473097133e-06, "loss": 0.4982, "step": 4615 }, { "epoch": 0.56, "learning_rate": 8.504940818907027e-06, "loss": 0.6127, "step": 4616 }, { "epoch": 0.56, "learning_rate": 8.501046396663426e-06, "loss": 0.5395, "step": 4617 }, { "epoch": 0.56, "learning_rate": 8.49715220697052e-06, "loss": 0.4957, "step": 4618 }, { "epoch": 0.56, "learning_rate": 8.49325825043246e-06, "loss": 0.5397, "step": 4619 }, { "epoch": 0.56, "learning_rate": 8.489364527653364e-06, "loss": 0.5468, "step": 4620 }, { "epoch": 0.56, "learning_rate": 8.48547103923731e-06, "loss": 0.5784, "step": 4621 }, { "epoch": 0.56, "learning_rate": 8.481577785788342e-06, "loss": 0.5261, "step": 4622 }, { "epoch": 0.56, "learning_rate": 8.477684767910466e-06, "loss": 0.5886, "step": 4623 }, { "epoch": 0.56, "learning_rate": 8.473791986207656e-06, "loss": 0.5299, "step": 4624 }, { "epoch": 0.56, "learning_rate": 8.469899441283842e-06, "loss": 0.5563, "step": 4625 }, { "epoch": 0.56, "learning_rate": 8.466007133742924e-06, "loss": 0.5469, "step": 4626 }, { "epoch": 0.56, "learning_rate": 8.462115064188762e-06, "loss": 0.5317, "step": 4627 }, { "epoch": 0.56, "learning_rate": 8.458223233225182e-06, "loss": 0.5894, "step": 4628 }, { "epoch": 0.56, "learning_rate": 8.454331641455967e-06, "loss": 0.5418, "step": 4629 }, { "epoch": 0.56, "learning_rate": 8.450440289484867e-06, "loss": 0.604, "step": 4630 }, { "epoch": 0.56, "learning_rate": 8.446549177915595e-06, "loss": 0.5656, "step": 4631 }, { "epoch": 0.56, "learning_rate": 8.442658307351825e-06, "loss": 0.5506, "step": 4632 }, { "epoch": 0.56, "learning_rate": 8.438767678397194e-06, "loss": 0.5128, "step": 4633 }, { "epoch": 0.56, "learning_rate": 8.434877291655306e-06, "loss": 0.5629, "step": 4634 }, { "epoch": 0.56, "learning_rate": 8.43098714772972e-06, "loss": 0.558, "step": 4635 }, { "epoch": 0.56, "learning_rate": 8.427097247223961e-06, "loss": 0.5477, "step": 4636 }, { "epoch": 0.56, "learning_rate": 8.423207590741516e-06, "loss": 0.528, "step": 4637 }, { "epoch": 0.56, "learning_rate": 8.419318178885832e-06, "loss": 0.5457, "step": 4638 }, { "epoch": 0.56, "learning_rate": 8.415429012260326e-06, "loss": 0.5251, "step": 4639 }, { "epoch": 0.56, "learning_rate": 8.411540091468366e-06, "loss": 0.5567, "step": 4640 }, { "epoch": 0.56, "learning_rate": 8.407651417113291e-06, "loss": 0.5503, "step": 4641 }, { "epoch": 0.56, "learning_rate": 8.403762989798397e-06, "loss": 0.5644, "step": 4642 }, { "epoch": 0.56, "learning_rate": 8.39987481012694e-06, "loss": 0.5274, "step": 4643 }, { "epoch": 0.56, "learning_rate": 8.395986878702141e-06, "loss": 0.5285, "step": 4644 }, { "epoch": 0.56, "learning_rate": 8.392099196127183e-06, "loss": 0.5304, "step": 4645 }, { "epoch": 0.56, "learning_rate": 8.38821176300521e-06, "loss": 0.5297, "step": 4646 }, { "epoch": 0.57, "learning_rate": 8.38432457993932e-06, "loss": 0.5865, "step": 4647 }, { "epoch": 0.57, "learning_rate": 8.380437647532581e-06, "loss": 0.5935, "step": 4648 }, { "epoch": 0.57, "learning_rate": 8.376550966388028e-06, "loss": 0.5347, "step": 4649 }, { "epoch": 0.57, "learning_rate": 8.372664537108632e-06, "loss": 0.5163, "step": 4650 }, { "epoch": 0.57, "learning_rate": 8.36877836029736e-06, "loss": 0.5472, "step": 4651 }, { "epoch": 0.57, "learning_rate": 8.364892436557108e-06, "loss": 0.5145, "step": 4652 }, { "epoch": 0.57, "learning_rate": 8.361006766490753e-06, "loss": 0.5386, "step": 4653 }, { "epoch": 0.57, "learning_rate": 8.357121350701121e-06, "loss": 0.5491, "step": 4654 }, { "epoch": 0.57, "learning_rate": 8.35323618979101e-06, "loss": 0.4902, "step": 4655 }, { "epoch": 0.57, "learning_rate": 8.349351284363164e-06, "loss": 0.5441, "step": 4656 }, { "epoch": 0.57, "learning_rate": 8.345466635020298e-06, "loss": 0.5459, "step": 4657 }, { "epoch": 0.57, "learning_rate": 8.341582242365085e-06, "loss": 0.5064, "step": 4658 }, { "epoch": 0.57, "learning_rate": 8.33769810700016e-06, "loss": 0.5926, "step": 4659 }, { "epoch": 0.57, "learning_rate": 8.33381422952811e-06, "loss": 0.5301, "step": 4660 }, { "epoch": 0.57, "learning_rate": 8.32993061055149e-06, "loss": 0.516, "step": 4661 }, { "epoch": 0.57, "learning_rate": 8.326047250672812e-06, "loss": 0.5225, "step": 4662 }, { "epoch": 0.57, "learning_rate": 8.322164150494552e-06, "loss": 0.562, "step": 4663 }, { "epoch": 0.57, "learning_rate": 8.318281310619141e-06, "loss": 0.6048, "step": 4664 }, { "epoch": 0.57, "learning_rate": 8.314398731648968e-06, "loss": 0.5307, "step": 4665 }, { "epoch": 0.57, "learning_rate": 8.310516414186382e-06, "loss": 0.545, "step": 4666 }, { "epoch": 0.57, "learning_rate": 8.306634358833698e-06, "loss": 0.5387, "step": 4667 }, { "epoch": 0.57, "learning_rate": 8.302752566193187e-06, "loss": 0.5381, "step": 4668 }, { "epoch": 0.57, "learning_rate": 8.298871036867074e-06, "loss": 0.4854, "step": 4669 }, { "epoch": 0.57, "learning_rate": 8.294989771457548e-06, "loss": 0.5015, "step": 4670 }, { "epoch": 0.57, "learning_rate": 8.291108770566761e-06, "loss": 0.5489, "step": 4671 }, { "epoch": 0.57, "learning_rate": 8.287228034796814e-06, "loss": 0.596, "step": 4672 }, { "epoch": 0.57, "learning_rate": 8.283347564749773e-06, "loss": 0.5443, "step": 4673 }, { "epoch": 0.57, "learning_rate": 8.27946736102766e-06, "loss": 0.4959, "step": 4674 }, { "epoch": 0.57, "learning_rate": 8.275587424232466e-06, "loss": 0.578, "step": 4675 }, { "epoch": 0.57, "learning_rate": 8.271707754966122e-06, "loss": 0.5414, "step": 4676 }, { "epoch": 0.57, "learning_rate": 8.267828353830537e-06, "loss": 0.5329, "step": 4677 }, { "epoch": 0.57, "learning_rate": 8.263949221427565e-06, "loss": 0.5535, "step": 4678 }, { "epoch": 0.57, "learning_rate": 8.260070358359018e-06, "loss": 0.5186, "step": 4679 }, { "epoch": 0.57, "learning_rate": 8.256191765226677e-06, "loss": 0.5344, "step": 4680 }, { "epoch": 0.57, "learning_rate": 8.252313442632274e-06, "loss": 0.4845, "step": 4681 }, { "epoch": 0.57, "learning_rate": 8.248435391177499e-06, "loss": 0.534, "step": 4682 }, { "epoch": 0.57, "learning_rate": 8.244557611463997e-06, "loss": 0.5002, "step": 4683 }, { "epoch": 0.57, "learning_rate": 8.240680104093379e-06, "loss": 0.5144, "step": 4684 }, { "epoch": 0.57, "learning_rate": 8.23680286966721e-06, "loss": 0.5226, "step": 4685 }, { "epoch": 0.57, "learning_rate": 8.232925908787005e-06, "loss": 0.5202, "step": 4686 }, { "epoch": 0.57, "learning_rate": 8.229049222054254e-06, "loss": 0.5405, "step": 4687 }, { "epoch": 0.57, "learning_rate": 8.225172810070387e-06, "loss": 0.5559, "step": 4688 }, { "epoch": 0.57, "learning_rate": 8.221296673436798e-06, "loss": 0.5692, "step": 4689 }, { "epoch": 0.57, "learning_rate": 8.217420812754842e-06, "loss": 0.5209, "step": 4690 }, { "epoch": 0.57, "learning_rate": 8.21354522862583e-06, "loss": 0.5895, "step": 4691 }, { "epoch": 0.57, "learning_rate": 8.209669921651021e-06, "loss": 0.5326, "step": 4692 }, { "epoch": 0.57, "learning_rate": 8.205794892431642e-06, "loss": 0.497, "step": 4693 }, { "epoch": 0.57, "learning_rate": 8.201920141568872e-06, "loss": 0.5454, "step": 4694 }, { "epoch": 0.57, "learning_rate": 8.198045669663852e-06, "loss": 0.5223, "step": 4695 }, { "epoch": 0.57, "learning_rate": 8.194171477317668e-06, "loss": 0.5257, "step": 4696 }, { "epoch": 0.57, "learning_rate": 8.190297565131372e-06, "loss": 0.4934, "step": 4697 }, { "epoch": 0.57, "learning_rate": 8.186423933705971e-06, "loss": 0.5323, "step": 4698 }, { "epoch": 0.57, "learning_rate": 8.182550583642433e-06, "loss": 0.5657, "step": 4699 }, { "epoch": 0.57, "learning_rate": 8.17867751554167e-06, "loss": 0.5402, "step": 4700 }, { "epoch": 0.57, "learning_rate": 8.174804730004566e-06, "loss": 0.5302, "step": 4701 }, { "epoch": 0.57, "learning_rate": 8.170932227631945e-06, "loss": 0.5636, "step": 4702 }, { "epoch": 0.57, "learning_rate": 8.167060009024598e-06, "loss": 0.5196, "step": 4703 }, { "epoch": 0.57, "learning_rate": 8.163188074783272e-06, "loss": 0.5602, "step": 4704 }, { "epoch": 0.57, "learning_rate": 8.159316425508658e-06, "loss": 0.5379, "step": 4705 }, { "epoch": 0.57, "learning_rate": 8.155445061801419e-06, "loss": 0.52, "step": 4706 }, { "epoch": 0.57, "learning_rate": 8.151573984262164e-06, "loss": 0.5474, "step": 4707 }, { "epoch": 0.57, "learning_rate": 8.147703193491461e-06, "loss": 0.5544, "step": 4708 }, { "epoch": 0.57, "learning_rate": 8.14383269008983e-06, "loss": 0.6073, "step": 4709 }, { "epoch": 0.57, "learning_rate": 8.139962474657747e-06, "loss": 0.5154, "step": 4710 }, { "epoch": 0.57, "learning_rate": 8.13609254779565e-06, "loss": 0.5501, "step": 4711 }, { "epoch": 0.57, "learning_rate": 8.132222910103925e-06, "loss": 0.5258, "step": 4712 }, { "epoch": 0.57, "learning_rate": 8.128353562182916e-06, "loss": 0.5474, "step": 4713 }, { "epoch": 0.57, "learning_rate": 8.124484504632922e-06, "loss": 0.5695, "step": 4714 }, { "epoch": 0.57, "learning_rate": 8.120615738054194e-06, "loss": 0.5674, "step": 4715 }, { "epoch": 0.57, "learning_rate": 8.116747263046942e-06, "loss": 0.5576, "step": 4716 }, { "epoch": 0.57, "learning_rate": 8.112879080211329e-06, "loss": 0.5467, "step": 4717 }, { "epoch": 0.57, "learning_rate": 8.109011190147473e-06, "loss": 0.6155, "step": 4718 }, { "epoch": 0.57, "learning_rate": 8.105143593455446e-06, "loss": 0.5534, "step": 4719 }, { "epoch": 0.57, "learning_rate": 8.101276290735273e-06, "loss": 0.5793, "step": 4720 }, { "epoch": 0.57, "learning_rate": 8.097409282586938e-06, "loss": 0.5162, "step": 4721 }, { "epoch": 0.57, "learning_rate": 8.09354256961037e-06, "loss": 0.5113, "step": 4722 }, { "epoch": 0.57, "learning_rate": 8.089676152405468e-06, "loss": 0.5269, "step": 4723 }, { "epoch": 0.57, "learning_rate": 8.085810031572073e-06, "loss": 0.5455, "step": 4724 }, { "epoch": 0.57, "learning_rate": 8.08194420770998e-06, "loss": 0.5352, "step": 4725 }, { "epoch": 0.57, "learning_rate": 8.078078681418942e-06, "loss": 0.5431, "step": 4726 }, { "epoch": 0.57, "learning_rate": 8.074213453298667e-06, "loss": 0.5676, "step": 4727 }, { "epoch": 0.57, "learning_rate": 8.07034852394881e-06, "loss": 0.6008, "step": 4728 }, { "epoch": 0.58, "learning_rate": 8.066483893968984e-06, "loss": 0.4936, "step": 4729 }, { "epoch": 0.58, "learning_rate": 8.06261956395876e-06, "loss": 0.4854, "step": 4730 }, { "epoch": 0.58, "learning_rate": 8.058755534517658e-06, "loss": 0.566, "step": 4731 }, { "epoch": 0.58, "learning_rate": 8.054891806245145e-06, "loss": 0.5408, "step": 4732 }, { "epoch": 0.58, "learning_rate": 8.05102837974065e-06, "loss": 0.5462, "step": 4733 }, { "epoch": 0.58, "learning_rate": 8.047165255603555e-06, "loss": 0.5612, "step": 4734 }, { "epoch": 0.58, "learning_rate": 8.043302434433194e-06, "loss": 0.5236, "step": 4735 }, { "epoch": 0.58, "learning_rate": 8.039439916828848e-06, "loss": 0.4797, "step": 4736 }, { "epoch": 0.58, "learning_rate": 8.035577703389765e-06, "loss": 0.52, "step": 4737 }, { "epoch": 0.58, "learning_rate": 8.031715794715125e-06, "loss": 0.5485, "step": 4738 }, { "epoch": 0.58, "learning_rate": 8.027854191404078e-06, "loss": 0.5725, "step": 4739 }, { "epoch": 0.58, "learning_rate": 8.023992894055723e-06, "loss": 0.5645, "step": 4740 }, { "epoch": 0.58, "learning_rate": 8.020131903269109e-06, "loss": 0.5333, "step": 4741 }, { "epoch": 0.58, "learning_rate": 8.016271219643233e-06, "loss": 0.5688, "step": 4742 }, { "epoch": 0.58, "learning_rate": 8.012410843777051e-06, "loss": 0.5094, "step": 4743 }, { "epoch": 0.58, "learning_rate": 8.008550776269475e-06, "loss": 0.5416, "step": 4744 }, { "epoch": 0.58, "learning_rate": 8.004691017719355e-06, "loss": 0.5286, "step": 4745 }, { "epoch": 0.58, "learning_rate": 8.000831568725505e-06, "loss": 0.538, "step": 4746 }, { "epoch": 0.58, "learning_rate": 7.996972429886693e-06, "loss": 0.591, "step": 4747 }, { "epoch": 0.58, "learning_rate": 7.993113601801629e-06, "loss": 0.5778, "step": 4748 }, { "epoch": 0.58, "learning_rate": 7.98925508506898e-06, "loss": 0.4733, "step": 4749 }, { "epoch": 0.58, "learning_rate": 7.985396880287365e-06, "loss": 0.514, "step": 4750 }, { "epoch": 0.58, "learning_rate": 7.981538988055351e-06, "loss": 0.4782, "step": 4751 }, { "epoch": 0.58, "learning_rate": 7.977681408971462e-06, "loss": 0.5342, "step": 4752 }, { "epoch": 0.58, "learning_rate": 7.973824143634169e-06, "loss": 0.5202, "step": 4753 }, { "epoch": 0.58, "learning_rate": 7.969967192641899e-06, "loss": 0.5509, "step": 4754 }, { "epoch": 0.58, "learning_rate": 7.966110556593023e-06, "loss": 0.501, "step": 4755 }, { "epoch": 0.58, "learning_rate": 7.962254236085868e-06, "loss": 0.5177, "step": 4756 }, { "epoch": 0.58, "learning_rate": 7.958398231718713e-06, "loss": 0.5628, "step": 4757 }, { "epoch": 0.58, "learning_rate": 7.954542544089782e-06, "loss": 0.5372, "step": 4758 }, { "epoch": 0.58, "learning_rate": 7.950687173797262e-06, "loss": 0.5847, "step": 4759 }, { "epoch": 0.58, "learning_rate": 7.94683212143928e-06, "loss": 0.5168, "step": 4760 }, { "epoch": 0.58, "learning_rate": 7.942977387613915e-06, "loss": 0.5136, "step": 4761 }, { "epoch": 0.58, "learning_rate": 7.939122972919196e-06, "loss": 0.6059, "step": 4762 }, { "epoch": 0.58, "learning_rate": 7.93526887795311e-06, "loss": 0.5595, "step": 4763 }, { "epoch": 0.58, "learning_rate": 7.931415103313587e-06, "loss": 0.5491, "step": 4764 }, { "epoch": 0.58, "learning_rate": 7.927561649598507e-06, "loss": 0.5417, "step": 4765 }, { "epoch": 0.58, "learning_rate": 7.923708517405705e-06, "loss": 0.5033, "step": 4766 }, { "epoch": 0.58, "learning_rate": 7.919855707332966e-06, "loss": 0.5598, "step": 4767 }, { "epoch": 0.58, "learning_rate": 7.916003219978016e-06, "loss": 0.5723, "step": 4768 }, { "epoch": 0.58, "learning_rate": 7.912151055938542e-06, "loss": 0.516, "step": 4769 }, { "epoch": 0.58, "learning_rate": 7.908299215812175e-06, "loss": 0.5942, "step": 4770 }, { "epoch": 0.58, "learning_rate": 7.9044477001965e-06, "loss": 0.504, "step": 4771 }, { "epoch": 0.58, "learning_rate": 7.900596509689046e-06, "loss": 0.5602, "step": 4772 }, { "epoch": 0.58, "learning_rate": 7.8967456448873e-06, "loss": 0.5242, "step": 4773 }, { "epoch": 0.58, "learning_rate": 7.892895106388685e-06, "loss": 0.509, "step": 4774 }, { "epoch": 0.58, "learning_rate": 7.889044894790586e-06, "loss": 0.546, "step": 4775 }, { "epoch": 0.58, "learning_rate": 7.885195010690332e-06, "loss": 0.5397, "step": 4776 }, { "epoch": 0.58, "learning_rate": 7.881345454685202e-06, "loss": 0.5602, "step": 4777 }, { "epoch": 0.58, "learning_rate": 7.877496227372423e-06, "loss": 0.5729, "step": 4778 }, { "epoch": 0.58, "learning_rate": 7.873647329349172e-06, "loss": 0.5348, "step": 4779 }, { "epoch": 0.58, "learning_rate": 7.869798761212575e-06, "loss": 0.5488, "step": 4780 }, { "epoch": 0.58, "learning_rate": 7.865950523559708e-06, "loss": 0.5418, "step": 4781 }, { "epoch": 0.58, "learning_rate": 7.862102616987589e-06, "loss": 0.4947, "step": 4782 }, { "epoch": 0.58, "learning_rate": 7.8582550420932e-06, "loss": 0.5098, "step": 4783 }, { "epoch": 0.58, "learning_rate": 7.854407799473451e-06, "loss": 0.5098, "step": 4784 }, { "epoch": 0.58, "learning_rate": 7.850560889725218e-06, "loss": 0.4907, "step": 4785 }, { "epoch": 0.58, "learning_rate": 7.846714313445319e-06, "loss": 0.5631, "step": 4786 }, { "epoch": 0.58, "learning_rate": 7.842868071230512e-06, "loss": 0.5091, "step": 4787 }, { "epoch": 0.58, "learning_rate": 7.839022163677517e-06, "loss": 0.5541, "step": 4788 }, { "epoch": 0.58, "learning_rate": 7.835176591382995e-06, "loss": 0.5354, "step": 4789 }, { "epoch": 0.58, "learning_rate": 7.831331354943556e-06, "loss": 0.5357, "step": 4790 }, { "epoch": 0.58, "learning_rate": 7.827486454955756e-06, "loss": 0.5566, "step": 4791 }, { "epoch": 0.58, "learning_rate": 7.8236418920161e-06, "loss": 0.5015, "step": 4792 }, { "epoch": 0.58, "learning_rate": 7.819797666721045e-06, "loss": 0.56, "step": 4793 }, { "epoch": 0.58, "learning_rate": 7.815953779666986e-06, "loss": 0.5099, "step": 4794 }, { "epoch": 0.58, "learning_rate": 7.812110231450279e-06, "loss": 0.4774, "step": 4795 }, { "epoch": 0.58, "learning_rate": 7.808267022667216e-06, "loss": 0.5272, "step": 4796 }, { "epoch": 0.58, "learning_rate": 7.804424153914038e-06, "loss": 0.5198, "step": 4797 }, { "epoch": 0.58, "learning_rate": 7.80058162578694e-06, "loss": 0.5259, "step": 4798 }, { "epoch": 0.58, "learning_rate": 7.796739438882056e-06, "loss": 0.5003, "step": 4799 }, { "epoch": 0.58, "learning_rate": 7.792897593795473e-06, "loss": 0.5657, "step": 4800 }, { "epoch": 0.58, "learning_rate": 7.78905609112322e-06, "loss": 0.5773, "step": 4801 }, { "epoch": 0.58, "learning_rate": 7.785214931461276e-06, "loss": 0.5509, "step": 4802 }, { "epoch": 0.58, "learning_rate": 7.78137411540557e-06, "loss": 0.4738, "step": 4803 }, { "epoch": 0.58, "learning_rate": 7.77753364355197e-06, "loss": 0.5278, "step": 4804 }, { "epoch": 0.58, "learning_rate": 7.773693516496296e-06, "loss": 0.5147, "step": 4805 }, { "epoch": 0.58, "learning_rate": 7.769853734834309e-06, "loss": 0.5629, "step": 4806 }, { "epoch": 0.58, "learning_rate": 7.766014299161728e-06, "loss": 0.5163, "step": 4807 }, { "epoch": 0.58, "learning_rate": 7.762175210074206e-06, "loss": 0.5181, "step": 4808 }, { "epoch": 0.58, "learning_rate": 7.758336468167351e-06, "loss": 0.5057, "step": 4809 }, { "epoch": 0.58, "learning_rate": 7.75449807403671e-06, "loss": 0.4837, "step": 4810 }, { "epoch": 0.59, "learning_rate": 7.750660028277777e-06, "loss": 0.5718, "step": 4811 }, { "epoch": 0.59, "learning_rate": 7.746822331485997e-06, "loss": 0.5453, "step": 4812 }, { "epoch": 0.59, "learning_rate": 7.74298498425676e-06, "loss": 0.5512, "step": 4813 }, { "epoch": 0.59, "learning_rate": 7.739147987185396e-06, "loss": 0.5285, "step": 4814 }, { "epoch": 0.59, "learning_rate": 7.735311340867187e-06, "loss": 0.5299, "step": 4815 }, { "epoch": 0.59, "learning_rate": 7.731475045897356e-06, "loss": 0.5551, "step": 4816 }, { "epoch": 0.59, "learning_rate": 7.727639102871078e-06, "loss": 0.5649, "step": 4817 }, { "epoch": 0.59, "learning_rate": 7.723803512383457e-06, "loss": 0.5186, "step": 4818 }, { "epoch": 0.59, "learning_rate": 7.719968275029572e-06, "loss": 0.5556, "step": 4819 }, { "epoch": 0.59, "learning_rate": 7.716133391404417e-06, "loss": 0.5254, "step": 4820 }, { "epoch": 0.59, "learning_rate": 7.712298862102948e-06, "loss": 0.5978, "step": 4821 }, { "epoch": 0.59, "learning_rate": 7.70846468772006e-06, "loss": 0.5236, "step": 4822 }, { "epoch": 0.59, "learning_rate": 7.704630868850598e-06, "loss": 0.4757, "step": 4823 }, { "epoch": 0.59, "learning_rate": 7.700797406089343e-06, "loss": 0.5426, "step": 4824 }, { "epoch": 0.59, "learning_rate": 7.696964300031028e-06, "loss": 0.4987, "step": 4825 }, { "epoch": 0.59, "learning_rate": 7.693131551270334e-06, "loss": 0.5421, "step": 4826 }, { "epoch": 0.59, "learning_rate": 7.689299160401874e-06, "loss": 0.5148, "step": 4827 }, { "epoch": 0.59, "learning_rate": 7.685467128020215e-06, "loss": 0.539, "step": 4828 }, { "epoch": 0.59, "learning_rate": 7.681635454719867e-06, "loss": 0.517, "step": 4829 }, { "epoch": 0.59, "learning_rate": 7.677804141095281e-06, "loss": 0.5202, "step": 4830 }, { "epoch": 0.59, "learning_rate": 7.673973187740862e-06, "loss": 0.5187, "step": 4831 }, { "epoch": 0.59, "learning_rate": 7.67014259525095e-06, "loss": 0.5297, "step": 4832 }, { "epoch": 0.59, "learning_rate": 7.666312364219823e-06, "loss": 0.5112, "step": 4833 }, { "epoch": 0.59, "learning_rate": 7.662482495241718e-06, "loss": 0.5379, "step": 4834 }, { "epoch": 0.59, "learning_rate": 7.658652988910807e-06, "loss": 0.5385, "step": 4835 }, { "epoch": 0.59, "learning_rate": 7.654823845821209e-06, "loss": 0.5928, "step": 4836 }, { "epoch": 0.59, "learning_rate": 7.650995066566979e-06, "loss": 0.4537, "step": 4837 }, { "epoch": 0.59, "learning_rate": 7.647166651742128e-06, "loss": 0.5095, "step": 4838 }, { "epoch": 0.59, "learning_rate": 7.6433386019406e-06, "loss": 0.5151, "step": 4839 }, { "epoch": 0.59, "learning_rate": 7.63951091775629e-06, "loss": 0.5246, "step": 4840 }, { "epoch": 0.59, "learning_rate": 7.63568359978303e-06, "loss": 0.525, "step": 4841 }, { "epoch": 0.59, "learning_rate": 7.631856648614594e-06, "loss": 0.514, "step": 4842 }, { "epoch": 0.59, "learning_rate": 7.6280300648447115e-06, "loss": 0.5491, "step": 4843 }, { "epoch": 0.59, "learning_rate": 7.6242038490670415e-06, "loss": 0.4956, "step": 4844 }, { "epoch": 0.59, "learning_rate": 7.620378001875192e-06, "loss": 0.5505, "step": 4845 }, { "epoch": 0.59, "learning_rate": 7.616552523862714e-06, "loss": 0.5175, "step": 4846 }, { "epoch": 0.59, "learning_rate": 7.612727415623097e-06, "loss": 0.5235, "step": 4847 }, { "epoch": 0.59, "learning_rate": 7.608902677749776e-06, "loss": 0.5297, "step": 4848 }, { "epoch": 0.59, "learning_rate": 7.605078310836134e-06, "loss": 0.5781, "step": 4849 }, { "epoch": 0.59, "learning_rate": 7.601254315475483e-06, "loss": 0.6264, "step": 4850 }, { "epoch": 0.59, "learning_rate": 7.59743069226109e-06, "loss": 0.4882, "step": 4851 }, { "epoch": 0.59, "learning_rate": 7.593607441786158e-06, "loss": 0.5302, "step": 4852 }, { "epoch": 0.59, "learning_rate": 7.589784564643838e-06, "loss": 0.5049, "step": 4853 }, { "epoch": 0.59, "learning_rate": 7.585962061427209e-06, "loss": 0.5274, "step": 4854 }, { "epoch": 0.59, "learning_rate": 7.582139932729315e-06, "loss": 0.5234, "step": 4855 }, { "epoch": 0.59, "learning_rate": 7.5783181791431205e-06, "loss": 0.5572, "step": 4856 }, { "epoch": 0.59, "learning_rate": 7.574496801261541e-06, "loss": 0.5065, "step": 4857 }, { "epoch": 0.59, "learning_rate": 7.5706757996774334e-06, "loss": 0.548, "step": 4858 }, { "epoch": 0.59, "learning_rate": 7.566855174983598e-06, "loss": 0.5031, "step": 4859 }, { "epoch": 0.59, "learning_rate": 7.563034927772771e-06, "loss": 0.5641, "step": 4860 }, { "epoch": 0.59, "learning_rate": 7.559215058637634e-06, "loss": 0.5804, "step": 4861 }, { "epoch": 0.59, "learning_rate": 7.555395568170809e-06, "loss": 0.4745, "step": 4862 }, { "epoch": 0.59, "learning_rate": 7.5515764569648645e-06, "loss": 0.5599, "step": 4863 }, { "epoch": 0.59, "learning_rate": 7.547757725612296e-06, "loss": 0.5788, "step": 4864 }, { "epoch": 0.59, "learning_rate": 7.5439393747055545e-06, "loss": 0.5146, "step": 4865 }, { "epoch": 0.59, "learning_rate": 7.5401214048370246e-06, "loss": 0.5792, "step": 4866 }, { "epoch": 0.59, "learning_rate": 7.5363038165990366e-06, "loss": 0.5827, "step": 4867 }, { "epoch": 0.59, "learning_rate": 7.5324866105838605e-06, "loss": 0.5113, "step": 4868 }, { "epoch": 0.59, "learning_rate": 7.5286697873837e-06, "loss": 0.5121, "step": 4869 }, { "epoch": 0.59, "learning_rate": 7.524853347590708e-06, "loss": 0.616, "step": 4870 }, { "epoch": 0.59, "learning_rate": 7.521037291796973e-06, "loss": 0.5624, "step": 4871 }, { "epoch": 0.59, "learning_rate": 7.517221620594529e-06, "loss": 0.5182, "step": 4872 }, { "epoch": 0.59, "learning_rate": 7.513406334575343e-06, "loss": 0.4994, "step": 4873 }, { "epoch": 0.59, "learning_rate": 7.509591434331327e-06, "loss": 0.5395, "step": 4874 }, { "epoch": 0.59, "learning_rate": 7.505776920454333e-06, "loss": 0.5195, "step": 4875 }, { "epoch": 0.59, "learning_rate": 7.501962793536155e-06, "loss": 0.5637, "step": 4876 }, { "epoch": 0.59, "learning_rate": 7.4981490541685194e-06, "loss": 0.5014, "step": 4877 }, { "epoch": 0.59, "learning_rate": 7.494335702943099e-06, "loss": 0.5365, "step": 4878 }, { "epoch": 0.59, "learning_rate": 7.490522740451507e-06, "loss": 0.51, "step": 4879 }, { "epoch": 0.59, "learning_rate": 7.486710167285292e-06, "loss": 0.5303, "step": 4880 }, { "epoch": 0.59, "learning_rate": 7.482897984035945e-06, "loss": 0.5118, "step": 4881 }, { "epoch": 0.59, "learning_rate": 7.479086191294898e-06, "loss": 0.5238, "step": 4882 }, { "epoch": 0.59, "learning_rate": 7.475274789653517e-06, "loss": 0.5687, "step": 4883 }, { "epoch": 0.59, "learning_rate": 7.471463779703112e-06, "loss": 0.5618, "step": 4884 }, { "epoch": 0.59, "learning_rate": 7.467653162034929e-06, "loss": 0.5143, "step": 4885 }, { "epoch": 0.59, "learning_rate": 7.46384293724016e-06, "loss": 0.5207, "step": 4886 }, { "epoch": 0.59, "learning_rate": 7.460033105909925e-06, "loss": 0.5268, "step": 4887 }, { "epoch": 0.59, "learning_rate": 7.456223668635291e-06, "loss": 0.4979, "step": 4888 }, { "epoch": 0.59, "learning_rate": 7.452414626007265e-06, "loss": 0.4625, "step": 4889 }, { "epoch": 0.59, "learning_rate": 7.448605978616779e-06, "loss": 0.5154, "step": 4890 }, { "epoch": 0.59, "learning_rate": 7.44479772705473e-06, "loss": 0.525, "step": 4891 }, { "epoch": 0.59, "learning_rate": 7.4409898719119265e-06, "loss": 0.5445, "step": 4892 }, { "epoch": 0.6, "learning_rate": 7.437182413779132e-06, "loss": 0.549, "step": 4893 }, { "epoch": 0.6, "learning_rate": 7.4333753532470385e-06, "loss": 0.5314, "step": 4894 }, { "epoch": 0.6, "learning_rate": 7.429568690906288e-06, "loss": 0.5281, "step": 4895 }, { "epoch": 0.6, "learning_rate": 7.425762427347447e-06, "loss": 0.5341, "step": 4896 }, { "epoch": 0.6, "learning_rate": 7.42195656316103e-06, "loss": 0.6065, "step": 4897 }, { "epoch": 0.6, "learning_rate": 7.418151098937487e-06, "loss": 0.505, "step": 4898 }, { "epoch": 0.6, "learning_rate": 7.4143460352672056e-06, "loss": 0.5107, "step": 4899 }, { "epoch": 0.6, "learning_rate": 7.410541372740509e-06, "loss": 0.5433, "step": 4900 }, { "epoch": 0.6, "learning_rate": 7.4067371119476615e-06, "loss": 0.5589, "step": 4901 }, { "epoch": 0.6, "learning_rate": 7.402933253478861e-06, "loss": 0.5578, "step": 4902 }, { "epoch": 0.6, "learning_rate": 7.399129797924251e-06, "loss": 0.553, "step": 4903 }, { "epoch": 0.6, "learning_rate": 7.3953267458739035e-06, "loss": 0.5168, "step": 4904 }, { "epoch": 0.6, "learning_rate": 7.391524097917836e-06, "loss": 0.5093, "step": 4905 }, { "epoch": 0.6, "learning_rate": 7.387721854645994e-06, "loss": 0.5214, "step": 4906 }, { "epoch": 0.6, "learning_rate": 7.383920016648267e-06, "loss": 0.5448, "step": 4907 }, { "epoch": 0.6, "learning_rate": 7.380118584514481e-06, "loss": 0.5074, "step": 4908 }, { "epoch": 0.6, "learning_rate": 7.3763175588343964e-06, "loss": 0.5301, "step": 4909 }, { "epoch": 0.6, "learning_rate": 7.372516940197711e-06, "loss": 0.5153, "step": 4910 }, { "epoch": 0.6, "learning_rate": 7.368716729194063e-06, "loss": 0.5381, "step": 4911 }, { "epoch": 0.6, "learning_rate": 7.364916926413026e-06, "loss": 0.5699, "step": 4912 }, { "epoch": 0.6, "learning_rate": 7.361117532444104e-06, "loss": 0.4987, "step": 4913 }, { "epoch": 0.6, "learning_rate": 7.357318547876744e-06, "loss": 0.5442, "step": 4914 }, { "epoch": 0.6, "learning_rate": 7.353519973300331e-06, "loss": 0.5056, "step": 4915 }, { "epoch": 0.6, "learning_rate": 7.349721809304182e-06, "loss": 0.5516, "step": 4916 }, { "epoch": 0.6, "learning_rate": 7.345924056477553e-06, "loss": 0.549, "step": 4917 }, { "epoch": 0.6, "learning_rate": 7.342126715409636e-06, "loss": 0.4786, "step": 4918 }, { "epoch": 0.6, "learning_rate": 7.338329786689554e-06, "loss": 0.567, "step": 4919 }, { "epoch": 0.6, "learning_rate": 7.334533270906372e-06, "loss": 0.5637, "step": 4920 }, { "epoch": 0.6, "learning_rate": 7.330737168649092e-06, "loss": 0.5399, "step": 4921 }, { "epoch": 0.6, "learning_rate": 7.326941480506645e-06, "loss": 0.5825, "step": 4922 }, { "epoch": 0.6, "learning_rate": 7.323146207067904e-06, "loss": 0.5534, "step": 4923 }, { "epoch": 0.6, "learning_rate": 7.3193513489216725e-06, "loss": 0.5566, "step": 4924 }, { "epoch": 0.6, "learning_rate": 7.315556906656699e-06, "loss": 0.5313, "step": 4925 }, { "epoch": 0.6, "learning_rate": 7.311762880861648e-06, "loss": 0.4694, "step": 4926 }, { "epoch": 0.6, "learning_rate": 7.307969272125147e-06, "loss": 0.5739, "step": 4927 }, { "epoch": 0.6, "learning_rate": 7.304176081035741e-06, "loss": 0.5589, "step": 4928 }, { "epoch": 0.6, "learning_rate": 7.300383308181907e-06, "loss": 0.4578, "step": 4929 }, { "epoch": 0.6, "learning_rate": 7.296590954152067e-06, "loss": 0.5502, "step": 4930 }, { "epoch": 0.6, "learning_rate": 7.292799019534576e-06, "loss": 0.6128, "step": 4931 }, { "epoch": 0.6, "learning_rate": 7.289007504917719e-06, "loss": 0.5807, "step": 4932 }, { "epoch": 0.6, "learning_rate": 7.285216410889722e-06, "loss": 0.5401, "step": 4933 }, { "epoch": 0.6, "learning_rate": 7.28142573803874e-06, "loss": 0.552, "step": 4934 }, { "epoch": 0.6, "learning_rate": 7.277635486952871e-06, "loss": 0.5077, "step": 4935 }, { "epoch": 0.6, "learning_rate": 7.273845658220137e-06, "loss": 0.5833, "step": 4936 }, { "epoch": 0.6, "learning_rate": 7.2700562524285015e-06, "loss": 0.4895, "step": 4937 }, { "epoch": 0.6, "learning_rate": 7.266267270165858e-06, "loss": 0.5605, "step": 4938 }, { "epoch": 0.6, "learning_rate": 7.262478712020043e-06, "loss": 0.5369, "step": 4939 }, { "epoch": 0.6, "learning_rate": 7.258690578578816e-06, "loss": 0.498, "step": 4940 }, { "epoch": 0.6, "learning_rate": 7.25490287042988e-06, "loss": 0.4536, "step": 4941 }, { "epoch": 0.6, "learning_rate": 7.251115588160862e-06, "loss": 0.5377, "step": 4942 }, { "epoch": 0.6, "learning_rate": 7.247328732359333e-06, "loss": 0.5849, "step": 4943 }, { "epoch": 0.6, "learning_rate": 7.243542303612792e-06, "loss": 0.5397, "step": 4944 }, { "epoch": 0.6, "learning_rate": 7.239756302508674e-06, "loss": 0.5516, "step": 4945 }, { "epoch": 0.6, "learning_rate": 7.235970729634345e-06, "loss": 0.4726, "step": 4946 }, { "epoch": 0.6, "learning_rate": 7.232185585577109e-06, "loss": 0.5039, "step": 4947 }, { "epoch": 0.6, "learning_rate": 7.2284008709241994e-06, "loss": 0.5101, "step": 4948 }, { "epoch": 0.6, "learning_rate": 7.224616586262784e-06, "loss": 0.5758, "step": 4949 }, { "epoch": 0.6, "learning_rate": 7.220832732179963e-06, "loss": 0.5069, "step": 4950 }, { "epoch": 0.6, "learning_rate": 7.217049309262774e-06, "loss": 0.4771, "step": 4951 }, { "epoch": 0.6, "learning_rate": 7.213266318098185e-06, "loss": 0.5145, "step": 4952 }, { "epoch": 0.6, "learning_rate": 7.209483759273095e-06, "loss": 0.5066, "step": 4953 }, { "epoch": 0.6, "learning_rate": 7.205701633374343e-06, "loss": 0.5433, "step": 4954 }, { "epoch": 0.6, "learning_rate": 7.201919940988686e-06, "loss": 0.5124, "step": 4955 }, { "epoch": 0.6, "learning_rate": 7.198138682702831e-06, "loss": 0.5038, "step": 4956 }, { "epoch": 0.6, "learning_rate": 7.194357859103408e-06, "loss": 0.5152, "step": 4957 }, { "epoch": 0.6, "learning_rate": 7.1905774707769825e-06, "loss": 0.5625, "step": 4958 }, { "epoch": 0.6, "learning_rate": 7.186797518310049e-06, "loss": 0.5369, "step": 4959 }, { "epoch": 0.6, "learning_rate": 7.183018002289039e-06, "loss": 0.5829, "step": 4960 }, { "epoch": 0.6, "learning_rate": 7.1792389233003135e-06, "loss": 0.5057, "step": 4961 }, { "epoch": 0.6, "learning_rate": 7.1754602819301645e-06, "loss": 0.4641, "step": 4962 }, { "epoch": 0.6, "learning_rate": 7.171682078764823e-06, "loss": 0.5269, "step": 4963 }, { "epoch": 0.6, "learning_rate": 7.167904314390447e-06, "loss": 0.5426, "step": 4964 }, { "epoch": 0.6, "learning_rate": 7.164126989393121e-06, "loss": 0.562, "step": 4965 }, { "epoch": 0.6, "learning_rate": 7.16035010435887e-06, "loss": 0.5091, "step": 4966 }, { "epoch": 0.6, "learning_rate": 7.156573659873647e-06, "loss": 0.5246, "step": 4967 }, { "epoch": 0.6, "learning_rate": 7.1527976565233405e-06, "loss": 0.5667, "step": 4968 }, { "epoch": 0.6, "learning_rate": 7.149022094893762e-06, "loss": 0.4996, "step": 4969 }, { "epoch": 0.6, "learning_rate": 7.145246975570661e-06, "loss": 0.5377, "step": 4970 }, { "epoch": 0.6, "learning_rate": 7.141472299139722e-06, "loss": 0.5717, "step": 4971 }, { "epoch": 0.6, "learning_rate": 7.1376980661865495e-06, "loss": 0.5681, "step": 4972 }, { "epoch": 0.6, "learning_rate": 7.133924277296689e-06, "loss": 0.5835, "step": 4973 }, { "epoch": 0.6, "learning_rate": 7.130150933055611e-06, "loss": 0.5242, "step": 4974 }, { "epoch": 0.6, "learning_rate": 7.126378034048722e-06, "loss": 0.4895, "step": 4975 }, { "epoch": 0.61, "learning_rate": 7.122605580861359e-06, "loss": 0.5724, "step": 4976 }, { "epoch": 0.61, "learning_rate": 7.118833574078787e-06, "loss": 0.5346, "step": 4977 }, { "epoch": 0.61, "learning_rate": 7.115062014286202e-06, "loss": 0.5396, "step": 4978 }, { "epoch": 0.61, "learning_rate": 7.111290902068729e-06, "loss": 0.471, "step": 4979 }, { "epoch": 0.61, "learning_rate": 7.107520238011429e-06, "loss": 0.5007, "step": 4980 }, { "epoch": 0.61, "learning_rate": 7.103750022699292e-06, "loss": 0.5334, "step": 4981 }, { "epoch": 0.61, "learning_rate": 7.099980256717231e-06, "loss": 0.6155, "step": 4982 }, { "epoch": 0.61, "learning_rate": 7.096210940650101e-06, "loss": 0.4948, "step": 4983 }, { "epoch": 0.61, "learning_rate": 7.092442075082678e-06, "loss": 0.5756, "step": 4984 }, { "epoch": 0.61, "learning_rate": 7.088673660599674e-06, "loss": 0.4644, "step": 4985 }, { "epoch": 0.61, "learning_rate": 7.084905697785721e-06, "loss": 0.504, "step": 4986 }, { "epoch": 0.61, "learning_rate": 7.0811381872254e-06, "loss": 0.5782, "step": 4987 }, { "epoch": 0.61, "learning_rate": 7.077371129503204e-06, "loss": 0.5886, "step": 4988 }, { "epoch": 0.61, "learning_rate": 7.073604525203561e-06, "loss": 0.51, "step": 4989 }, { "epoch": 0.61, "learning_rate": 7.0698383749108324e-06, "loss": 0.5375, "step": 4990 }, { "epoch": 0.61, "learning_rate": 7.066072679209304e-06, "loss": 0.5459, "step": 4991 }, { "epoch": 0.61, "learning_rate": 7.062307438683193e-06, "loss": 0.4769, "step": 4992 }, { "epoch": 0.61, "learning_rate": 7.058542653916649e-06, "loss": 0.5564, "step": 4993 }, { "epoch": 0.61, "learning_rate": 7.05477832549375e-06, "loss": 0.5423, "step": 4994 }, { "epoch": 0.61, "learning_rate": 7.051014453998495e-06, "loss": 0.5458, "step": 4995 }, { "epoch": 0.61, "learning_rate": 7.047251040014823e-06, "loss": 0.5237, "step": 4996 }, { "epoch": 0.61, "learning_rate": 7.0434880841265975e-06, "loss": 0.4707, "step": 4997 }, { "epoch": 0.61, "learning_rate": 7.039725586917608e-06, "loss": 0.5767, "step": 4998 }, { "epoch": 0.61, "learning_rate": 7.03596354897158e-06, "loss": 0.6088, "step": 4999 }, { "epoch": 0.61, "learning_rate": 7.032201970872166e-06, "loss": 0.5877, "step": 5000 }, { "epoch": 0.61, "learning_rate": 7.028440853202936e-06, "loss": 0.5575, "step": 5001 }, { "epoch": 0.61, "learning_rate": 7.024680196547404e-06, "loss": 0.5636, "step": 5002 }, { "epoch": 0.61, "learning_rate": 7.020920001489005e-06, "loss": 0.4928, "step": 5003 }, { "epoch": 0.61, "learning_rate": 7.017160268611104e-06, "loss": 0.4559, "step": 5004 }, { "epoch": 0.61, "learning_rate": 7.01340099849699e-06, "loss": 0.5219, "step": 5005 }, { "epoch": 0.61, "learning_rate": 7.009642191729887e-06, "loss": 0.541, "step": 5006 }, { "epoch": 0.61, "learning_rate": 7.005883848892943e-06, "loss": 0.5062, "step": 5007 }, { "epoch": 0.61, "learning_rate": 7.0021259705692335e-06, "loss": 0.5474, "step": 5008 }, { "epoch": 0.61, "learning_rate": 6.998368557341763e-06, "loss": 0.5504, "step": 5009 }, { "epoch": 0.61, "learning_rate": 6.994611609793466e-06, "loss": 0.5832, "step": 5010 }, { "epoch": 0.61, "learning_rate": 6.9908551285072035e-06, "loss": 0.539, "step": 5011 }, { "epoch": 0.61, "learning_rate": 6.987099114065763e-06, "loss": 0.5378, "step": 5012 }, { "epoch": 0.61, "learning_rate": 6.983343567051861e-06, "loss": 0.4925, "step": 5013 }, { "epoch": 0.61, "learning_rate": 6.979588488048139e-06, "loss": 0.5587, "step": 5014 }, { "epoch": 0.61, "learning_rate": 6.975833877637168e-06, "loss": 0.5323, "step": 5015 }, { "epoch": 0.61, "learning_rate": 6.972079736401444e-06, "loss": 0.5217, "step": 5016 }, { "epoch": 0.61, "learning_rate": 6.968326064923398e-06, "loss": 0.54, "step": 5017 }, { "epoch": 0.61, "learning_rate": 6.964572863785377e-06, "loss": 0.5512, "step": 5018 }, { "epoch": 0.61, "learning_rate": 6.96082013356966e-06, "loss": 0.5433, "step": 5019 }, { "epoch": 0.61, "learning_rate": 6.957067874858455e-06, "loss": 0.4887, "step": 5020 }, { "epoch": 0.61, "learning_rate": 6.953316088233897e-06, "loss": 0.5239, "step": 5021 }, { "epoch": 0.61, "learning_rate": 6.949564774278037e-06, "loss": 0.5383, "step": 5022 }, { "epoch": 0.61, "learning_rate": 6.945813933572874e-06, "loss": 0.4929, "step": 5023 }, { "epoch": 0.61, "learning_rate": 6.942063566700313e-06, "loss": 0.5014, "step": 5024 }, { "epoch": 0.61, "learning_rate": 6.938313674242194e-06, "loss": 0.5826, "step": 5025 }, { "epoch": 0.61, "learning_rate": 6.9345642567802855e-06, "loss": 0.5546, "step": 5026 }, { "epoch": 0.61, "learning_rate": 6.9308153148962795e-06, "loss": 0.5085, "step": 5027 }, { "epoch": 0.61, "learning_rate": 6.927066849171793e-06, "loss": 0.4975, "step": 5028 }, { "epoch": 0.61, "learning_rate": 6.9233188601883705e-06, "loss": 0.5335, "step": 5029 }, { "epoch": 0.61, "learning_rate": 6.919571348527484e-06, "loss": 0.5015, "step": 5030 }, { "epoch": 0.61, "learning_rate": 6.915824314770527e-06, "loss": 0.483, "step": 5031 }, { "epoch": 0.61, "learning_rate": 6.912077759498825e-06, "loss": 0.5581, "step": 5032 }, { "epoch": 0.61, "learning_rate": 6.908331683293625e-06, "loss": 0.5631, "step": 5033 }, { "epoch": 0.61, "learning_rate": 6.904586086736098e-06, "loss": 0.5403, "step": 5034 }, { "epoch": 0.61, "learning_rate": 6.900840970407349e-06, "loss": 0.5449, "step": 5035 }, { "epoch": 0.61, "learning_rate": 6.897096334888402e-06, "loss": 0.5661, "step": 5036 }, { "epoch": 0.61, "learning_rate": 6.893352180760204e-06, "loss": 0.526, "step": 5037 }, { "epoch": 0.61, "learning_rate": 6.889608508603634e-06, "loss": 0.5686, "step": 5038 }, { "epoch": 0.61, "learning_rate": 6.885865318999487e-06, "loss": 0.5391, "step": 5039 }, { "epoch": 0.61, "learning_rate": 6.882122612528499e-06, "loss": 0.5998, "step": 5040 }, { "epoch": 0.61, "learning_rate": 6.878380389771311e-06, "loss": 0.5535, "step": 5041 }, { "epoch": 0.61, "learning_rate": 6.874638651308505e-06, "loss": 0.4887, "step": 5042 }, { "epoch": 0.61, "learning_rate": 6.870897397720578e-06, "loss": 0.5058, "step": 5043 }, { "epoch": 0.61, "learning_rate": 6.867156629587961e-06, "loss": 0.5027, "step": 5044 }, { "epoch": 0.61, "learning_rate": 6.863416347490999e-06, "loss": 0.4894, "step": 5045 }, { "epoch": 0.61, "learning_rate": 6.859676552009966e-06, "loss": 0.5308, "step": 5046 }, { "epoch": 0.61, "learning_rate": 6.855937243725068e-06, "loss": 0.5829, "step": 5047 }, { "epoch": 0.61, "learning_rate": 6.8521984232164255e-06, "loss": 0.4859, "step": 5048 }, { "epoch": 0.61, "learning_rate": 6.848460091064084e-06, "loss": 0.5025, "step": 5049 }, { "epoch": 0.61, "learning_rate": 6.844722247848022e-06, "loss": 0.5338, "step": 5050 }, { "epoch": 0.61, "learning_rate": 6.840984894148131e-06, "loss": 0.5448, "step": 5051 }, { "epoch": 0.61, "learning_rate": 6.8372480305442325e-06, "loss": 0.5567, "step": 5052 }, { "epoch": 0.61, "learning_rate": 6.8335116576160735e-06, "loss": 0.5516, "step": 5053 }, { "epoch": 0.61, "learning_rate": 6.829775775943318e-06, "loss": 0.5084, "step": 5054 }, { "epoch": 0.61, "learning_rate": 6.826040386105562e-06, "loss": 0.5278, "step": 5055 }, { "epoch": 0.61, "learning_rate": 6.822305488682319e-06, "loss": 0.5579, "step": 5056 }, { "epoch": 0.61, "learning_rate": 6.818571084253032e-06, "loss": 0.5141, "step": 5057 }, { "epoch": 0.62, "learning_rate": 6.814837173397057e-06, "loss": 0.4939, "step": 5058 }, { "epoch": 0.62, "learning_rate": 6.811103756693691e-06, "loss": 0.5308, "step": 5059 }, { "epoch": 0.62, "learning_rate": 6.8073708347221355e-06, "loss": 0.5695, "step": 5060 }, { "epoch": 0.62, "learning_rate": 6.803638408061526e-06, "loss": 0.5524, "step": 5061 }, { "epoch": 0.62, "learning_rate": 6.79990647729092e-06, "loss": 0.5095, "step": 5062 }, { "epoch": 0.62, "learning_rate": 6.796175042989297e-06, "loss": 0.5586, "step": 5063 }, { "epoch": 0.62, "learning_rate": 6.792444105735556e-06, "loss": 0.5376, "step": 5064 }, { "epoch": 0.62, "learning_rate": 6.788713666108524e-06, "loss": 0.5522, "step": 5065 }, { "epoch": 0.62, "learning_rate": 6.784983724686949e-06, "loss": 0.5217, "step": 5066 }, { "epoch": 0.62, "learning_rate": 6.7812542820495046e-06, "loss": 0.4829, "step": 5067 }, { "epoch": 0.62, "learning_rate": 6.777525338774779e-06, "loss": 0.5462, "step": 5068 }, { "epoch": 0.62, "learning_rate": 6.7737968954412895e-06, "loss": 0.475, "step": 5069 }, { "epoch": 0.62, "learning_rate": 6.770068952627473e-06, "loss": 0.4758, "step": 5070 }, { "epoch": 0.62, "learning_rate": 6.7663415109116945e-06, "loss": 0.5189, "step": 5071 }, { "epoch": 0.62, "learning_rate": 6.762614570872237e-06, "loss": 0.5323, "step": 5072 }, { "epoch": 0.62, "learning_rate": 6.7588881330873e-06, "loss": 0.5332, "step": 5073 }, { "epoch": 0.62, "learning_rate": 6.755162198135015e-06, "loss": 0.5417, "step": 5074 }, { "epoch": 0.62, "learning_rate": 6.751436766593429e-06, "loss": 0.5275, "step": 5075 }, { "epoch": 0.62, "learning_rate": 6.747711839040515e-06, "loss": 0.5581, "step": 5076 }, { "epoch": 0.62, "learning_rate": 6.743987416054163e-06, "loss": 0.5176, "step": 5077 }, { "epoch": 0.62, "learning_rate": 6.740263498212189e-06, "loss": 0.5113, "step": 5078 }, { "epoch": 0.62, "learning_rate": 6.736540086092329e-06, "loss": 0.5469, "step": 5079 }, { "epoch": 0.62, "learning_rate": 6.732817180272244e-06, "loss": 0.5875, "step": 5080 }, { "epoch": 0.62, "learning_rate": 6.729094781329506e-06, "loss": 0.5416, "step": 5081 }, { "epoch": 0.62, "learning_rate": 6.725372889841619e-06, "loss": 0.5388, "step": 5082 }, { "epoch": 0.62, "learning_rate": 6.721651506386007e-06, "loss": 0.5308, "step": 5083 }, { "epoch": 0.62, "learning_rate": 6.717930631540011e-06, "loss": 0.589, "step": 5084 }, { "epoch": 0.62, "learning_rate": 6.714210265880894e-06, "loss": 0.5395, "step": 5085 }, { "epoch": 0.62, "learning_rate": 6.710490409985846e-06, "loss": 0.5511, "step": 5086 }, { "epoch": 0.62, "learning_rate": 6.7067710644319664e-06, "loss": 0.5909, "step": 5087 }, { "epoch": 0.62, "learning_rate": 6.703052229796285e-06, "loss": 0.5125, "step": 5088 }, { "epoch": 0.62, "learning_rate": 6.6993339066557495e-06, "loss": 0.5557, "step": 5089 }, { "epoch": 0.62, "learning_rate": 6.69561609558723e-06, "loss": 0.5611, "step": 5090 }, { "epoch": 0.62, "learning_rate": 6.691898797167508e-06, "loss": 0.5142, "step": 5091 }, { "epoch": 0.62, "learning_rate": 6.688182011973301e-06, "loss": 0.5326, "step": 5092 }, { "epoch": 0.62, "learning_rate": 6.6844657405812365e-06, "loss": 0.5546, "step": 5093 }, { "epoch": 0.62, "learning_rate": 6.680749983567858e-06, "loss": 0.5602, "step": 5094 }, { "epoch": 0.62, "learning_rate": 6.677034741509645e-06, "loss": 0.5175, "step": 5095 }, { "epoch": 0.62, "learning_rate": 6.673320014982983e-06, "loss": 0.503, "step": 5096 }, { "epoch": 0.62, "learning_rate": 6.669605804564181e-06, "loss": 0.5635, "step": 5097 }, { "epoch": 0.62, "learning_rate": 6.66589211082947e-06, "loss": 0.5225, "step": 5098 }, { "epoch": 0.62, "learning_rate": 6.662178934355005e-06, "loss": 0.5303, "step": 5099 }, { "epoch": 0.62, "learning_rate": 6.658466275716849e-06, "loss": 0.5468, "step": 5100 }, { "epoch": 0.62, "learning_rate": 6.654754135490991e-06, "loss": 0.5305, "step": 5101 }, { "epoch": 0.62, "learning_rate": 6.651042514253345e-06, "loss": 0.5193, "step": 5102 }, { "epoch": 0.62, "learning_rate": 6.647331412579738e-06, "loss": 0.4966, "step": 5103 }, { "epoch": 0.62, "learning_rate": 6.643620831045916e-06, "loss": 0.5544, "step": 5104 }, { "epoch": 0.62, "learning_rate": 6.639910770227546e-06, "loss": 0.5752, "step": 5105 }, { "epoch": 0.62, "learning_rate": 6.636201230700212e-06, "loss": 0.5294, "step": 5106 }, { "epoch": 0.62, "learning_rate": 6.632492213039428e-06, "loss": 0.5435, "step": 5107 }, { "epoch": 0.62, "learning_rate": 6.628783717820611e-06, "loss": 0.5253, "step": 5108 }, { "epoch": 0.62, "learning_rate": 6.62507574561911e-06, "loss": 0.5093, "step": 5109 }, { "epoch": 0.62, "learning_rate": 6.621368297010181e-06, "loss": 0.4738, "step": 5110 }, { "epoch": 0.62, "learning_rate": 6.617661372569009e-06, "loss": 0.4938, "step": 5111 }, { "epoch": 0.62, "learning_rate": 6.613954972870694e-06, "loss": 0.5093, "step": 5112 }, { "epoch": 0.62, "learning_rate": 6.6102490984902515e-06, "loss": 0.5349, "step": 5113 }, { "epoch": 0.62, "learning_rate": 6.606543750002621e-06, "loss": 0.5475, "step": 5114 }, { "epoch": 0.62, "learning_rate": 6.602838927982656e-06, "loss": 0.5693, "step": 5115 }, { "epoch": 0.62, "learning_rate": 6.599134633005133e-06, "loss": 0.5427, "step": 5116 }, { "epoch": 0.62, "learning_rate": 6.595430865644739e-06, "loss": 0.5252, "step": 5117 }, { "epoch": 0.62, "learning_rate": 6.591727626476086e-06, "loss": 0.5494, "step": 5118 }, { "epoch": 0.62, "learning_rate": 6.588024916073703e-06, "loss": 0.5316, "step": 5119 }, { "epoch": 0.62, "learning_rate": 6.584322735012038e-06, "loss": 0.5256, "step": 5120 }, { "epoch": 0.62, "learning_rate": 6.5806210838654505e-06, "loss": 0.5169, "step": 5121 }, { "epoch": 0.62, "learning_rate": 6.576919963208228e-06, "loss": 0.5311, "step": 5122 }, { "epoch": 0.62, "learning_rate": 6.573219373614563e-06, "loss": 0.5571, "step": 5123 }, { "epoch": 0.62, "learning_rate": 6.569519315658574e-06, "loss": 0.5411, "step": 5124 }, { "epoch": 0.62, "learning_rate": 6.5658197899143e-06, "loss": 0.4996, "step": 5125 }, { "epoch": 0.62, "learning_rate": 6.56212079695569e-06, "loss": 0.5329, "step": 5126 }, { "epoch": 0.62, "learning_rate": 6.558422337356612e-06, "loss": 0.5182, "step": 5127 }, { "epoch": 0.62, "learning_rate": 6.554724411690853e-06, "loss": 0.5567, "step": 5128 }, { "epoch": 0.62, "learning_rate": 6.5510270205321195e-06, "loss": 0.5087, "step": 5129 }, { "epoch": 0.62, "learning_rate": 6.547330164454025e-06, "loss": 0.5055, "step": 5130 }, { "epoch": 0.62, "learning_rate": 6.5436338440301136e-06, "loss": 0.5656, "step": 5131 }, { "epoch": 0.62, "learning_rate": 6.539938059833842e-06, "loss": 0.5582, "step": 5132 }, { "epoch": 0.62, "learning_rate": 6.536242812438577e-06, "loss": 0.543, "step": 5133 }, { "epoch": 0.62, "learning_rate": 6.5325481024176075e-06, "loss": 0.5068, "step": 5134 }, { "epoch": 0.62, "learning_rate": 6.528853930344141e-06, "loss": 0.512, "step": 5135 }, { "epoch": 0.62, "learning_rate": 6.5251602967912946e-06, "loss": 0.5443, "step": 5136 }, { "epoch": 0.62, "learning_rate": 6.521467202332108e-06, "loss": 0.5098, "step": 5137 }, { "epoch": 0.62, "learning_rate": 6.517774647539534e-06, "loss": 0.5243, "step": 5138 }, { "epoch": 0.62, "learning_rate": 6.514082632986448e-06, "loss": 0.5492, "step": 5139 }, { "epoch": 0.63, "learning_rate": 6.51039115924563e-06, "loss": 0.5149, "step": 5140 }, { "epoch": 0.63, "learning_rate": 6.5067002268897874e-06, "loss": 0.5362, "step": 5141 }, { "epoch": 0.63, "learning_rate": 6.503009836491534e-06, "loss": 0.5542, "step": 5142 }, { "epoch": 0.63, "learning_rate": 6.499319988623409e-06, "loss": 0.5166, "step": 5143 }, { "epoch": 0.63, "learning_rate": 6.495630683857863e-06, "loss": 0.5359, "step": 5144 }, { "epoch": 0.63, "learning_rate": 6.491941922767262e-06, "loss": 0.5676, "step": 5145 }, { "epoch": 0.63, "learning_rate": 6.488253705923886e-06, "loss": 0.5064, "step": 5146 }, { "epoch": 0.63, "learning_rate": 6.4845660338999315e-06, "loss": 0.5303, "step": 5147 }, { "epoch": 0.63, "learning_rate": 6.480878907267514e-06, "loss": 0.4774, "step": 5148 }, { "epoch": 0.63, "learning_rate": 6.477192326598662e-06, "loss": 0.5187, "step": 5149 }, { "epoch": 0.63, "learning_rate": 6.4735062924653145e-06, "loss": 0.5902, "step": 5150 }, { "epoch": 0.63, "learning_rate": 6.469820805439335e-06, "loss": 0.5424, "step": 5151 }, { "epoch": 0.63, "learning_rate": 6.466135866092498e-06, "loss": 0.5282, "step": 5152 }, { "epoch": 0.63, "learning_rate": 6.46245147499649e-06, "loss": 0.4599, "step": 5153 }, { "epoch": 0.63, "learning_rate": 6.458767632722911e-06, "loss": 0.5219, "step": 5154 }, { "epoch": 0.63, "learning_rate": 6.455084339843286e-06, "loss": 0.5478, "step": 5155 }, { "epoch": 0.63, "learning_rate": 6.451401596929048e-06, "loss": 0.5638, "step": 5156 }, { "epoch": 0.63, "learning_rate": 6.447719404551543e-06, "loss": 0.4966, "step": 5157 }, { "epoch": 0.63, "learning_rate": 6.444037763282036e-06, "loss": 0.5702, "step": 5158 }, { "epoch": 0.63, "learning_rate": 6.4403566736917016e-06, "loss": 0.4933, "step": 5159 }, { "epoch": 0.63, "learning_rate": 6.436676136351632e-06, "loss": 0.4922, "step": 5160 }, { "epoch": 0.63, "learning_rate": 6.432996151832835e-06, "loss": 0.504, "step": 5161 }, { "epoch": 0.63, "learning_rate": 6.429316720706232e-06, "loss": 0.5261, "step": 5162 }, { "epoch": 0.63, "learning_rate": 6.425637843542652e-06, "loss": 0.5783, "step": 5163 }, { "epoch": 0.63, "learning_rate": 6.421959520912846e-06, "loss": 0.4948, "step": 5164 }, { "epoch": 0.63, "learning_rate": 6.418281753387477e-06, "loss": 0.4796, "step": 5165 }, { "epoch": 0.63, "learning_rate": 6.41460454153712e-06, "loss": 0.535, "step": 5166 }, { "epoch": 0.63, "learning_rate": 6.41092788593227e-06, "loss": 0.5169, "step": 5167 }, { "epoch": 0.63, "learning_rate": 6.4072517871433275e-06, "loss": 0.5381, "step": 5168 }, { "epoch": 0.63, "learning_rate": 6.403576245740607e-06, "loss": 0.5763, "step": 5169 }, { "epoch": 0.63, "learning_rate": 6.399901262294342e-06, "loss": 0.5275, "step": 5170 }, { "epoch": 0.63, "learning_rate": 6.396226837374677e-06, "loss": 0.536, "step": 5171 }, { "epoch": 0.63, "learning_rate": 6.392552971551674e-06, "loss": 0.4841, "step": 5172 }, { "epoch": 0.63, "learning_rate": 6.388879665395294e-06, "loss": 0.539, "step": 5173 }, { "epoch": 0.63, "learning_rate": 6.385206919475428e-06, "loss": 0.5276, "step": 5174 }, { "epoch": 0.63, "learning_rate": 6.3815347343618735e-06, "loss": 0.5118, "step": 5175 }, { "epoch": 0.63, "learning_rate": 6.3778631106243384e-06, "loss": 0.5312, "step": 5176 }, { "epoch": 0.63, "learning_rate": 6.374192048832446e-06, "loss": 0.5211, "step": 5177 }, { "epoch": 0.63, "learning_rate": 6.37052154955573e-06, "loss": 0.5636, "step": 5178 }, { "epoch": 0.63, "learning_rate": 6.3668516133636445e-06, "loss": 0.5277, "step": 5179 }, { "epoch": 0.63, "learning_rate": 6.363182240825549e-06, "loss": 0.4967, "step": 5180 }, { "epoch": 0.63, "learning_rate": 6.359513432510718e-06, "loss": 0.5187, "step": 5181 }, { "epoch": 0.63, "learning_rate": 6.355845188988334e-06, "loss": 0.5436, "step": 5182 }, { "epoch": 0.63, "learning_rate": 6.352177510827499e-06, "loss": 0.5219, "step": 5183 }, { "epoch": 0.63, "learning_rate": 6.348510398597223e-06, "loss": 0.5534, "step": 5184 }, { "epoch": 0.63, "learning_rate": 6.344843852866432e-06, "loss": 0.5351, "step": 5185 }, { "epoch": 0.63, "learning_rate": 6.341177874203954e-06, "loss": 0.513, "step": 5186 }, { "epoch": 0.63, "learning_rate": 6.337512463178543e-06, "loss": 0.529, "step": 5187 }, { "epoch": 0.63, "learning_rate": 6.333847620358856e-06, "loss": 0.5545, "step": 5188 }, { "epoch": 0.63, "learning_rate": 6.330183346313468e-06, "loss": 0.5475, "step": 5189 }, { "epoch": 0.63, "learning_rate": 6.3265196416108496e-06, "loss": 0.5055, "step": 5190 }, { "epoch": 0.63, "learning_rate": 6.322856506819412e-06, "loss": 0.526, "step": 5191 }, { "epoch": 0.63, "learning_rate": 6.319193942507451e-06, "loss": 0.5311, "step": 5192 }, { "epoch": 0.63, "learning_rate": 6.315531949243185e-06, "loss": 0.5367, "step": 5193 }, { "epoch": 0.63, "learning_rate": 6.31187052759475e-06, "loss": 0.5397, "step": 5194 }, { "epoch": 0.63, "learning_rate": 6.308209678130177e-06, "loss": 0.5374, "step": 5195 }, { "epoch": 0.63, "learning_rate": 6.3045494014174215e-06, "loss": 0.5008, "step": 5196 }, { "epoch": 0.63, "learning_rate": 6.300889698024346e-06, "loss": 0.5001, "step": 5197 }, { "epoch": 0.63, "learning_rate": 6.2972305685187284e-06, "loss": 0.5305, "step": 5198 }, { "epoch": 0.63, "learning_rate": 6.293572013468248e-06, "loss": 0.5498, "step": 5199 }, { "epoch": 0.63, "learning_rate": 6.2899140334405004e-06, "loss": 0.4713, "step": 5200 }, { "epoch": 0.63, "learning_rate": 6.286256629002994e-06, "loss": 0.5484, "step": 5201 }, { "epoch": 0.63, "learning_rate": 6.282599800723143e-06, "loss": 0.5308, "step": 5202 }, { "epoch": 0.63, "learning_rate": 6.278943549168281e-06, "loss": 0.6045, "step": 5203 }, { "epoch": 0.63, "learning_rate": 6.2752878749056465e-06, "loss": 0.4834, "step": 5204 }, { "epoch": 0.63, "learning_rate": 6.2716327785023815e-06, "loss": 0.5698, "step": 5205 }, { "epoch": 0.63, "learning_rate": 6.267978260525547e-06, "loss": 0.4939, "step": 5206 }, { "epoch": 0.63, "learning_rate": 6.264324321542115e-06, "loss": 0.5414, "step": 5207 }, { "epoch": 0.63, "learning_rate": 6.260670962118966e-06, "loss": 0.5159, "step": 5208 }, { "epoch": 0.63, "learning_rate": 6.2570181828228845e-06, "loss": 0.5373, "step": 5209 }, { "epoch": 0.63, "learning_rate": 6.253365984220574e-06, "loss": 0.5146, "step": 5210 }, { "epoch": 0.63, "learning_rate": 6.249714366878643e-06, "loss": 0.5618, "step": 5211 }, { "epoch": 0.63, "learning_rate": 6.2460633313636095e-06, "loss": 0.5289, "step": 5212 }, { "epoch": 0.63, "learning_rate": 6.242412878241903e-06, "loss": 0.474, "step": 5213 }, { "epoch": 0.63, "learning_rate": 6.2387630080798624e-06, "loss": 0.4264, "step": 5214 }, { "epoch": 0.63, "learning_rate": 6.235113721443738e-06, "loss": 0.5057, "step": 5215 }, { "epoch": 0.63, "learning_rate": 6.231465018899684e-06, "loss": 0.4417, "step": 5216 }, { "epoch": 0.63, "learning_rate": 6.227816901013772e-06, "loss": 0.5614, "step": 5217 }, { "epoch": 0.63, "learning_rate": 6.224169368351975e-06, "loss": 0.5696, "step": 5218 }, { "epoch": 0.63, "learning_rate": 6.220522421480178e-06, "loss": 0.5151, "step": 5219 }, { "epoch": 0.63, "learning_rate": 6.2168760609641765e-06, "loss": 0.5506, "step": 5220 }, { "epoch": 0.63, "learning_rate": 6.2132302873696786e-06, "loss": 0.5247, "step": 5221 }, { "epoch": 0.64, "learning_rate": 6.2095851012622895e-06, "loss": 0.5535, "step": 5222 }, { "epoch": 0.64, "learning_rate": 6.205940503207535e-06, "loss": 0.5743, "step": 5223 }, { "epoch": 0.64, "learning_rate": 6.2022964937708455e-06, "loss": 0.5846, "step": 5224 }, { "epoch": 0.64, "learning_rate": 6.19865307351756e-06, "loss": 0.5438, "step": 5225 }, { "epoch": 0.64, "learning_rate": 6.19501024301292e-06, "loss": 0.5506, "step": 5226 }, { "epoch": 0.64, "learning_rate": 6.191368002822093e-06, "loss": 0.4971, "step": 5227 }, { "epoch": 0.64, "learning_rate": 6.187726353510136e-06, "loss": 0.5059, "step": 5228 }, { "epoch": 0.64, "learning_rate": 6.184085295642024e-06, "loss": 0.56, "step": 5229 }, { "epoch": 0.64, "learning_rate": 6.180444829782635e-06, "loss": 0.5238, "step": 5230 }, { "epoch": 0.64, "learning_rate": 6.176804956496765e-06, "loss": 0.5242, "step": 5231 }, { "epoch": 0.64, "learning_rate": 6.173165676349103e-06, "loss": 0.5251, "step": 5232 }, { "epoch": 0.64, "learning_rate": 6.16952698990426e-06, "loss": 0.5402, "step": 5233 }, { "epoch": 0.64, "learning_rate": 6.165888897726749e-06, "loss": 0.5424, "step": 5234 }, { "epoch": 0.64, "learning_rate": 6.162251400380986e-06, "loss": 0.5234, "step": 5235 }, { "epoch": 0.64, "learning_rate": 6.158614498431305e-06, "loss": 0.5507, "step": 5236 }, { "epoch": 0.64, "learning_rate": 6.1549781924419395e-06, "loss": 0.5204, "step": 5237 }, { "epoch": 0.64, "learning_rate": 6.151342482977036e-06, "loss": 0.5136, "step": 5238 }, { "epoch": 0.64, "learning_rate": 6.1477073706006386e-06, "loss": 0.5885, "step": 5239 }, { "epoch": 0.64, "learning_rate": 6.144072855876717e-06, "loss": 0.4546, "step": 5240 }, { "epoch": 0.64, "learning_rate": 6.14043893936913e-06, "loss": 0.5401, "step": 5241 }, { "epoch": 0.64, "learning_rate": 6.136805621641653e-06, "loss": 0.5639, "step": 5242 }, { "epoch": 0.64, "learning_rate": 6.133172903257964e-06, "loss": 0.5719, "step": 5243 }, { "epoch": 0.64, "learning_rate": 6.129540784781656e-06, "loss": 0.5004, "step": 5244 }, { "epoch": 0.64, "learning_rate": 6.1259092667762155e-06, "loss": 0.5503, "step": 5245 }, { "epoch": 0.64, "learning_rate": 6.122278349805049e-06, "loss": 0.4796, "step": 5246 }, { "epoch": 0.64, "learning_rate": 6.1186480344314605e-06, "loss": 0.5822, "step": 5247 }, { "epoch": 0.64, "learning_rate": 6.11501832121867e-06, "loss": 0.5281, "step": 5248 }, { "epoch": 0.64, "learning_rate": 6.111389210729793e-06, "loss": 0.5189, "step": 5249 }, { "epoch": 0.64, "learning_rate": 6.107760703527857e-06, "loss": 0.5952, "step": 5250 }, { "epoch": 0.64, "learning_rate": 6.104132800175798e-06, "loss": 0.5442, "step": 5251 }, { "epoch": 0.64, "learning_rate": 6.1005055012364575e-06, "loss": 0.5264, "step": 5252 }, { "epoch": 0.64, "learning_rate": 6.096878807272582e-06, "loss": 0.5692, "step": 5253 }, { "epoch": 0.64, "learning_rate": 6.0932527188468224e-06, "loss": 0.4957, "step": 5254 }, { "epoch": 0.64, "learning_rate": 6.089627236521738e-06, "loss": 0.5901, "step": 5255 }, { "epoch": 0.64, "learning_rate": 6.086002360859793e-06, "loss": 0.4974, "step": 5256 }, { "epoch": 0.64, "learning_rate": 6.08237809242336e-06, "loss": 0.5324, "step": 5257 }, { "epoch": 0.64, "learning_rate": 6.078754431774711e-06, "loss": 0.5162, "step": 5258 }, { "epoch": 0.64, "learning_rate": 6.0751313794760334e-06, "loss": 0.5184, "step": 5259 }, { "epoch": 0.64, "learning_rate": 6.07150893608941e-06, "loss": 0.5226, "step": 5260 }, { "epoch": 0.64, "learning_rate": 6.0678871021768384e-06, "loss": 0.5488, "step": 5261 }, { "epoch": 0.64, "learning_rate": 6.064265878300213e-06, "loss": 0.5242, "step": 5262 }, { "epoch": 0.64, "learning_rate": 6.060645265021336e-06, "loss": 0.5271, "step": 5263 }, { "epoch": 0.64, "learning_rate": 6.057025262901926e-06, "loss": 0.5722, "step": 5264 }, { "epoch": 0.64, "learning_rate": 6.0534058725035895e-06, "loss": 0.5546, "step": 5265 }, { "epoch": 0.64, "learning_rate": 6.049787094387849e-06, "loss": 0.5555, "step": 5266 }, { "epoch": 0.64, "learning_rate": 6.046168929116131e-06, "loss": 0.5326, "step": 5267 }, { "epoch": 0.64, "learning_rate": 6.042551377249759e-06, "loss": 0.5801, "step": 5268 }, { "epoch": 0.64, "learning_rate": 6.038934439349973e-06, "loss": 0.5182, "step": 5269 }, { "epoch": 0.64, "learning_rate": 6.035318115977908e-06, "loss": 0.5449, "step": 5270 }, { "epoch": 0.64, "learning_rate": 6.031702407694613e-06, "loss": 0.4971, "step": 5271 }, { "epoch": 0.64, "learning_rate": 6.028087315061029e-06, "loss": 0.532, "step": 5272 }, { "epoch": 0.64, "learning_rate": 6.024472838638014e-06, "loss": 0.5272, "step": 5273 }, { "epoch": 0.64, "learning_rate": 6.020858978986326e-06, "loss": 0.5052, "step": 5274 }, { "epoch": 0.64, "learning_rate": 6.017245736666617e-06, "loss": 0.5423, "step": 5275 }, { "epoch": 0.64, "learning_rate": 6.013633112239467e-06, "loss": 0.5463, "step": 5276 }, { "epoch": 0.64, "learning_rate": 6.010021106265338e-06, "loss": 0.5565, "step": 5277 }, { "epoch": 0.64, "learning_rate": 6.006409719304603e-06, "loss": 0.5404, "step": 5278 }, { "epoch": 0.64, "learning_rate": 6.002798951917543e-06, "loss": 0.513, "step": 5279 }, { "epoch": 0.64, "learning_rate": 5.9991888046643395e-06, "loss": 0.4997, "step": 5280 }, { "epoch": 0.64, "learning_rate": 5.995579278105075e-06, "loss": 0.4769, "step": 5281 }, { "epoch": 0.64, "learning_rate": 5.991970372799742e-06, "loss": 0.4617, "step": 5282 }, { "epoch": 0.64, "learning_rate": 5.98836208930823e-06, "loss": 0.5134, "step": 5283 }, { "epoch": 0.64, "learning_rate": 5.984754428190343e-06, "loss": 0.5161, "step": 5284 }, { "epoch": 0.64, "learning_rate": 5.98114739000577e-06, "loss": 0.515, "step": 5285 }, { "epoch": 0.64, "learning_rate": 5.97754097531412e-06, "loss": 0.497, "step": 5286 }, { "epoch": 0.64, "learning_rate": 5.973935184674898e-06, "loss": 0.5109, "step": 5287 }, { "epoch": 0.64, "learning_rate": 5.9703300186475165e-06, "loss": 0.5354, "step": 5288 }, { "epoch": 0.64, "learning_rate": 5.966725477791286e-06, "loss": 0.5087, "step": 5289 }, { "epoch": 0.64, "learning_rate": 5.963121562665426e-06, "loss": 0.4918, "step": 5290 }, { "epoch": 0.64, "learning_rate": 5.959518273829049e-06, "loss": 0.5555, "step": 5291 }, { "epoch": 0.64, "learning_rate": 5.955915611841179e-06, "loss": 0.5114, "step": 5292 }, { "epoch": 0.64, "learning_rate": 5.952313577260744e-06, "loss": 0.5854, "step": 5293 }, { "epoch": 0.64, "learning_rate": 5.948712170646564e-06, "loss": 0.5772, "step": 5294 }, { "epoch": 0.64, "learning_rate": 5.945111392557373e-06, "loss": 0.4701, "step": 5295 }, { "epoch": 0.64, "learning_rate": 5.941511243551802e-06, "loss": 0.5244, "step": 5296 }, { "epoch": 0.64, "learning_rate": 5.937911724188388e-06, "loss": 0.5384, "step": 5297 }, { "epoch": 0.64, "learning_rate": 5.934312835025562e-06, "loss": 0.505, "step": 5298 }, { "epoch": 0.64, "learning_rate": 5.930714576621665e-06, "loss": 0.4846, "step": 5299 }, { "epoch": 0.64, "learning_rate": 5.927116949534941e-06, "loss": 0.5134, "step": 5300 }, { "epoch": 0.64, "learning_rate": 5.923519954323531e-06, "loss": 0.4904, "step": 5301 }, { "epoch": 0.64, "learning_rate": 5.919923591545481e-06, "loss": 0.4976, "step": 5302 }, { "epoch": 0.64, "learning_rate": 5.91632786175874e-06, "loss": 0.529, "step": 5303 }, { "epoch": 0.64, "learning_rate": 5.912732765521152e-06, "loss": 0.5265, "step": 5304 }, { "epoch": 0.65, "learning_rate": 5.909138303390469e-06, "loss": 0.4931, "step": 5305 }, { "epoch": 0.65, "learning_rate": 5.905544475924344e-06, "loss": 0.5181, "step": 5306 }, { "epoch": 0.65, "learning_rate": 5.901951283680333e-06, "loss": 0.4806, "step": 5307 }, { "epoch": 0.65, "learning_rate": 5.898358727215887e-06, "loss": 0.5335, "step": 5308 }, { "epoch": 0.65, "learning_rate": 5.894766807088364e-06, "loss": 0.5467, "step": 5309 }, { "epoch": 0.65, "learning_rate": 5.891175523855022e-06, "loss": 0.5609, "step": 5310 }, { "epoch": 0.65, "learning_rate": 5.887584878073016e-06, "loss": 0.4993, "step": 5311 }, { "epoch": 0.65, "learning_rate": 5.8839948702994144e-06, "loss": 0.5509, "step": 5312 }, { "epoch": 0.65, "learning_rate": 5.8804055010911755e-06, "loss": 0.549, "step": 5313 }, { "epoch": 0.65, "learning_rate": 5.876816771005156e-06, "loss": 0.4842, "step": 5314 }, { "epoch": 0.65, "learning_rate": 5.8732286805981245e-06, "loss": 0.511, "step": 5315 }, { "epoch": 0.65, "learning_rate": 5.869641230426742e-06, "loss": 0.5106, "step": 5316 }, { "epoch": 0.65, "learning_rate": 5.866054421047574e-06, "loss": 0.5279, "step": 5317 }, { "epoch": 0.65, "learning_rate": 5.862468253017081e-06, "loss": 0.5465, "step": 5318 }, { "epoch": 0.65, "learning_rate": 5.8588827268916345e-06, "loss": 0.5011, "step": 5319 }, { "epoch": 0.65, "learning_rate": 5.855297843227499e-06, "loss": 0.5205, "step": 5320 }, { "epoch": 0.65, "learning_rate": 5.851713602580838e-06, "loss": 0.4861, "step": 5321 }, { "epoch": 0.65, "learning_rate": 5.848130005507719e-06, "loss": 0.5871, "step": 5322 }, { "epoch": 0.65, "learning_rate": 5.844547052564106e-06, "loss": 0.5339, "step": 5323 }, { "epoch": 0.65, "learning_rate": 5.84096474430587e-06, "loss": 0.526, "step": 5324 }, { "epoch": 0.65, "learning_rate": 5.837383081288779e-06, "loss": 0.5432, "step": 5325 }, { "epoch": 0.65, "learning_rate": 5.83380206406849e-06, "loss": 0.5252, "step": 5326 }, { "epoch": 0.65, "learning_rate": 5.830221693200582e-06, "loss": 0.4471, "step": 5327 }, { "epoch": 0.65, "learning_rate": 5.826641969240513e-06, "loss": 0.5324, "step": 5328 }, { "epoch": 0.65, "learning_rate": 5.8230628927436465e-06, "loss": 0.4934, "step": 5329 }, { "epoch": 0.65, "learning_rate": 5.819484464265255e-06, "loss": 0.513, "step": 5330 }, { "epoch": 0.65, "learning_rate": 5.815906684360495e-06, "loss": 0.504, "step": 5331 }, { "epoch": 0.65, "learning_rate": 5.812329553584439e-06, "loss": 0.602, "step": 5332 }, { "epoch": 0.65, "learning_rate": 5.808753072492047e-06, "loss": 0.5752, "step": 5333 }, { "epoch": 0.65, "learning_rate": 5.805177241638176e-06, "loss": 0.5876, "step": 5334 }, { "epoch": 0.65, "learning_rate": 5.801602061577593e-06, "loss": 0.5899, "step": 5335 }, { "epoch": 0.65, "learning_rate": 5.79802753286496e-06, "loss": 0.5551, "step": 5336 }, { "epoch": 0.65, "learning_rate": 5.794453656054837e-06, "loss": 0.4872, "step": 5337 }, { "epoch": 0.65, "learning_rate": 5.790880431701675e-06, "loss": 0.5013, "step": 5338 }, { "epoch": 0.65, "learning_rate": 5.7873078603598405e-06, "loss": 0.4694, "step": 5339 }, { "epoch": 0.65, "learning_rate": 5.7837359425835815e-06, "loss": 0.5319, "step": 5340 }, { "epoch": 0.65, "learning_rate": 5.780164678927059e-06, "loss": 0.5739, "step": 5341 }, { "epoch": 0.65, "learning_rate": 5.776594069944326e-06, "loss": 0.5555, "step": 5342 }, { "epoch": 0.65, "learning_rate": 5.773024116189325e-06, "loss": 0.5208, "step": 5343 }, { "epoch": 0.65, "learning_rate": 5.769454818215916e-06, "loss": 0.5352, "step": 5344 }, { "epoch": 0.65, "learning_rate": 5.765886176577845e-06, "loss": 0.535, "step": 5345 }, { "epoch": 0.65, "learning_rate": 5.762318191828752e-06, "loss": 0.5708, "step": 5346 }, { "epoch": 0.65, "learning_rate": 5.758750864522186e-06, "loss": 0.5543, "step": 5347 }, { "epoch": 0.65, "learning_rate": 5.755184195211593e-06, "loss": 0.5185, "step": 5348 }, { "epoch": 0.65, "learning_rate": 5.7516181844503065e-06, "loss": 0.5821, "step": 5349 }, { "epoch": 0.65, "learning_rate": 5.748052832791571e-06, "loss": 0.545, "step": 5350 }, { "epoch": 0.65, "learning_rate": 5.744488140788519e-06, "loss": 0.5212, "step": 5351 }, { "epoch": 0.65, "learning_rate": 5.7409241089941805e-06, "loss": 0.56, "step": 5352 }, { "epoch": 0.65, "learning_rate": 5.737360737961495e-06, "loss": 0.5454, "step": 5353 }, { "epoch": 0.65, "learning_rate": 5.733798028243285e-06, "loss": 0.555, "step": 5354 }, { "epoch": 0.65, "learning_rate": 5.730235980392274e-06, "loss": 0.4658, "step": 5355 }, { "epoch": 0.65, "learning_rate": 5.726674594961095e-06, "loss": 0.5053, "step": 5356 }, { "epoch": 0.65, "learning_rate": 5.7231138725022565e-06, "loss": 0.5499, "step": 5357 }, { "epoch": 0.65, "learning_rate": 5.7195538135681875e-06, "loss": 0.4838, "step": 5358 }, { "epoch": 0.65, "learning_rate": 5.7159944187111924e-06, "loss": 0.4725, "step": 5359 }, { "epoch": 0.65, "learning_rate": 5.712435688483493e-06, "loss": 0.5409, "step": 5360 }, { "epoch": 0.65, "learning_rate": 5.7088776234371905e-06, "loss": 0.48, "step": 5361 }, { "epoch": 0.65, "learning_rate": 5.705320224124294e-06, "loss": 0.5663, "step": 5362 }, { "epoch": 0.65, "learning_rate": 5.701763491096702e-06, "loss": 0.508, "step": 5363 }, { "epoch": 0.65, "learning_rate": 5.69820742490622e-06, "loss": 0.5366, "step": 5364 }, { "epoch": 0.65, "learning_rate": 5.694652026104539e-06, "loss": 0.5271, "step": 5365 }, { "epoch": 0.65, "learning_rate": 5.691097295243246e-06, "loss": 0.4999, "step": 5366 }, { "epoch": 0.65, "learning_rate": 5.687543232873839e-06, "loss": 0.4933, "step": 5367 }, { "epoch": 0.65, "learning_rate": 5.683989839547697e-06, "loss": 0.5188, "step": 5368 }, { "epoch": 0.65, "learning_rate": 5.6804371158160975e-06, "loss": 0.4879, "step": 5369 }, { "epoch": 0.65, "learning_rate": 5.6768850622302244e-06, "loss": 0.529, "step": 5370 }, { "epoch": 0.65, "learning_rate": 5.673333679341142e-06, "loss": 0.5386, "step": 5371 }, { "epoch": 0.65, "learning_rate": 5.669782967699824e-06, "loss": 0.5594, "step": 5372 }, { "epoch": 0.65, "learning_rate": 5.666232927857138e-06, "loss": 0.5401, "step": 5373 }, { "epoch": 0.65, "learning_rate": 5.662683560363843e-06, "loss": 0.587, "step": 5374 }, { "epoch": 0.65, "learning_rate": 5.659134865770588e-06, "loss": 0.5138, "step": 5375 }, { "epoch": 0.65, "learning_rate": 5.655586844627935e-06, "loss": 0.541, "step": 5376 }, { "epoch": 0.65, "learning_rate": 5.6520394974863255e-06, "loss": 0.5078, "step": 5377 }, { "epoch": 0.65, "learning_rate": 5.6484928248960995e-06, "loss": 0.5327, "step": 5378 }, { "epoch": 0.65, "learning_rate": 5.6449468274075004e-06, "loss": 0.4981, "step": 5379 }, { "epoch": 0.65, "learning_rate": 5.641401505570656e-06, "loss": 0.5339, "step": 5380 }, { "epoch": 0.65, "learning_rate": 5.637856859935602e-06, "loss": 0.5884, "step": 5381 }, { "epoch": 0.65, "learning_rate": 5.634312891052258e-06, "loss": 0.5426, "step": 5382 }, { "epoch": 0.65, "learning_rate": 5.630769599470438e-06, "loss": 0.5614, "step": 5383 }, { "epoch": 0.65, "learning_rate": 5.62722698573986e-06, "loss": 0.5025, "step": 5384 }, { "epoch": 0.65, "learning_rate": 5.623685050410135e-06, "loss": 0.5266, "step": 5385 }, { "epoch": 0.65, "learning_rate": 5.620143794030759e-06, "loss": 0.5302, "step": 5386 }, { "epoch": 0.66, "learning_rate": 5.6166032171511385e-06, "loss": 0.5286, "step": 5387 }, { "epoch": 0.66, "learning_rate": 5.613063320320561e-06, "loss": 0.5639, "step": 5388 }, { "epoch": 0.66, "learning_rate": 5.609524104088209e-06, "loss": 0.4987, "step": 5389 }, { "epoch": 0.66, "learning_rate": 5.605985569003172e-06, "loss": 0.4884, "step": 5390 }, { "epoch": 0.66, "learning_rate": 5.602447715614419e-06, "loss": 0.5852, "step": 5391 }, { "epoch": 0.66, "learning_rate": 5.598910544470822e-06, "loss": 0.5368, "step": 5392 }, { "epoch": 0.66, "learning_rate": 5.595374056121146e-06, "loss": 0.5689, "step": 5393 }, { "epoch": 0.66, "learning_rate": 5.591838251114048e-06, "loss": 0.5391, "step": 5394 }, { "epoch": 0.66, "learning_rate": 5.588303129998072e-06, "loss": 0.5542, "step": 5395 }, { "epoch": 0.66, "learning_rate": 5.58476869332168e-06, "loss": 0.5894, "step": 5396 }, { "epoch": 0.66, "learning_rate": 5.581234941633203e-06, "loss": 0.6014, "step": 5397 }, { "epoch": 0.66, "learning_rate": 5.577701875480872e-06, "loss": 0.5047, "step": 5398 }, { "epoch": 0.66, "learning_rate": 5.57416949541282e-06, "loss": 0.518, "step": 5399 }, { "epoch": 0.66, "learning_rate": 5.570637801977063e-06, "loss": 0.5333, "step": 5400 }, { "epoch": 0.66, "learning_rate": 5.567106795721513e-06, "loss": 0.4618, "step": 5401 }, { "epoch": 0.66, "learning_rate": 5.563576477193986e-06, "loss": 0.4732, "step": 5402 }, { "epoch": 0.66, "learning_rate": 5.560046846942173e-06, "loss": 0.5736, "step": 5403 }, { "epoch": 0.66, "learning_rate": 5.5565179055136765e-06, "loss": 0.5724, "step": 5404 }, { "epoch": 0.66, "learning_rate": 5.552989653455979e-06, "loss": 0.5625, "step": 5405 }, { "epoch": 0.66, "learning_rate": 5.549462091316457e-06, "loss": 0.5231, "step": 5406 }, { "epoch": 0.66, "learning_rate": 5.54593521964239e-06, "loss": 0.4825, "step": 5407 }, { "epoch": 0.66, "learning_rate": 5.542409038980945e-06, "loss": 0.5214, "step": 5408 }, { "epoch": 0.66, "learning_rate": 5.538883549879173e-06, "loss": 0.5449, "step": 5409 }, { "epoch": 0.66, "learning_rate": 5.535358752884035e-06, "loss": 0.5173, "step": 5410 }, { "epoch": 0.66, "learning_rate": 5.53183464854237e-06, "loss": 0.4707, "step": 5411 }, { "epoch": 0.66, "learning_rate": 5.528311237400912e-06, "loss": 0.5591, "step": 5412 }, { "epoch": 0.66, "learning_rate": 5.524788520006296e-06, "loss": 0.5349, "step": 5413 }, { "epoch": 0.66, "learning_rate": 5.52126649690504e-06, "loss": 0.5088, "step": 5414 }, { "epoch": 0.66, "learning_rate": 5.517745168643555e-06, "loss": 0.5735, "step": 5415 }, { "epoch": 0.66, "learning_rate": 5.514224535768156e-06, "loss": 0.4883, "step": 5416 }, { "epoch": 0.66, "learning_rate": 5.510704598825035e-06, "loss": 0.5044, "step": 5417 }, { "epoch": 0.66, "learning_rate": 5.507185358360279e-06, "loss": 0.4896, "step": 5418 }, { "epoch": 0.66, "learning_rate": 5.503666814919872e-06, "loss": 0.5067, "step": 5419 }, { "epoch": 0.66, "learning_rate": 5.5001489690496956e-06, "loss": 0.5468, "step": 5420 }, { "epoch": 0.66, "learning_rate": 5.496631821295506e-06, "loss": 0.468, "step": 5421 }, { "epoch": 0.66, "learning_rate": 5.493115372202966e-06, "loss": 0.4539, "step": 5422 }, { "epoch": 0.66, "learning_rate": 5.489599622317625e-06, "loss": 0.4826, "step": 5423 }, { "epoch": 0.66, "learning_rate": 5.486084572184917e-06, "loss": 0.5467, "step": 5424 }, { "epoch": 0.66, "learning_rate": 5.482570222350182e-06, "loss": 0.5538, "step": 5425 }, { "epoch": 0.66, "learning_rate": 5.479056573358637e-06, "loss": 0.5405, "step": 5426 }, { "epoch": 0.66, "learning_rate": 5.4755436257554014e-06, "loss": 0.541, "step": 5427 }, { "epoch": 0.66, "learning_rate": 5.472031380085481e-06, "loss": 0.5792, "step": 5428 }, { "epoch": 0.66, "learning_rate": 5.468519836893764e-06, "loss": 0.5133, "step": 5429 }, { "epoch": 0.66, "learning_rate": 5.465008996725052e-06, "loss": 0.5403, "step": 5430 }, { "epoch": 0.66, "learning_rate": 5.4614988601240105e-06, "loss": 0.4686, "step": 5431 }, { "epoch": 0.66, "learning_rate": 5.457989427635217e-06, "loss": 0.4718, "step": 5432 }, { "epoch": 0.66, "learning_rate": 5.454480699803134e-06, "loss": 0.5124, "step": 5433 }, { "epoch": 0.66, "learning_rate": 5.450972677172109e-06, "loss": 0.5286, "step": 5434 }, { "epoch": 0.66, "learning_rate": 5.4474653602863815e-06, "loss": 0.5131, "step": 5435 }, { "epoch": 0.66, "learning_rate": 5.443958749690088e-06, "loss": 0.4858, "step": 5436 }, { "epoch": 0.66, "learning_rate": 5.44045284592725e-06, "loss": 0.534, "step": 5437 }, { "epoch": 0.66, "learning_rate": 5.436947649541776e-06, "loss": 0.5762, "step": 5438 }, { "epoch": 0.66, "learning_rate": 5.433443161077476e-06, "loss": 0.4828, "step": 5439 }, { "epoch": 0.66, "learning_rate": 5.429939381078042e-06, "loss": 0.6069, "step": 5440 }, { "epoch": 0.66, "learning_rate": 5.426436310087051e-06, "loss": 0.5146, "step": 5441 }, { "epoch": 0.66, "learning_rate": 5.422933948647986e-06, "loss": 0.5839, "step": 5442 }, { "epoch": 0.66, "learning_rate": 5.419432297304202e-06, "loss": 0.5357, "step": 5443 }, { "epoch": 0.66, "learning_rate": 5.4159313565989556e-06, "loss": 0.4964, "step": 5444 }, { "epoch": 0.66, "learning_rate": 5.412431127075395e-06, "loss": 0.4994, "step": 5445 }, { "epoch": 0.66, "learning_rate": 5.4089316092765476e-06, "loss": 0.4907, "step": 5446 }, { "epoch": 0.66, "learning_rate": 5.405432803745332e-06, "loss": 0.502, "step": 5447 }, { "epoch": 0.66, "learning_rate": 5.40193471102457e-06, "loss": 0.5442, "step": 5448 }, { "epoch": 0.66, "learning_rate": 5.398437331656952e-06, "loss": 0.5342, "step": 5449 }, { "epoch": 0.66, "learning_rate": 5.394940666185078e-06, "loss": 0.5956, "step": 5450 }, { "epoch": 0.66, "learning_rate": 5.391444715151423e-06, "loss": 0.5114, "step": 5451 }, { "epoch": 0.66, "learning_rate": 5.3879494790983525e-06, "loss": 0.4892, "step": 5452 }, { "epoch": 0.66, "learning_rate": 5.384454958568133e-06, "loss": 0.5244, "step": 5453 }, { "epoch": 0.66, "learning_rate": 5.380961154102905e-06, "loss": 0.5533, "step": 5454 }, { "epoch": 0.66, "learning_rate": 5.3774680662447e-06, "loss": 0.5046, "step": 5455 }, { "epoch": 0.66, "learning_rate": 5.373975695535456e-06, "loss": 0.5325, "step": 5456 }, { "epoch": 0.66, "learning_rate": 5.370484042516979e-06, "loss": 0.5622, "step": 5457 }, { "epoch": 0.66, "learning_rate": 5.3669931077309675e-06, "loss": 0.5924, "step": 5458 }, { "epoch": 0.66, "learning_rate": 5.363502891719022e-06, "loss": 0.5233, "step": 5459 }, { "epoch": 0.66, "learning_rate": 5.360013395022613e-06, "loss": 0.5523, "step": 5460 }, { "epoch": 0.66, "learning_rate": 5.356524618183108e-06, "loss": 0.5492, "step": 5461 }, { "epoch": 0.66, "learning_rate": 5.35303656174177e-06, "loss": 0.5675, "step": 5462 }, { "epoch": 0.66, "learning_rate": 5.349549226239739e-06, "loss": 0.5163, "step": 5463 }, { "epoch": 0.66, "learning_rate": 5.346062612218044e-06, "loss": 0.4872, "step": 5464 }, { "epoch": 0.66, "learning_rate": 5.34257672021761e-06, "loss": 0.556, "step": 5465 }, { "epoch": 0.66, "learning_rate": 5.339091550779243e-06, "loss": 0.4989, "step": 5466 }, { "epoch": 0.66, "learning_rate": 5.335607104443638e-06, "loss": 0.5179, "step": 5467 }, { "epoch": 0.66, "learning_rate": 5.332123381751386e-06, "loss": 0.5436, "step": 5468 }, { "epoch": 0.67, "learning_rate": 5.328640383242953e-06, "loss": 0.5695, "step": 5469 }, { "epoch": 0.67, "learning_rate": 5.325158109458697e-06, "loss": 0.4787, "step": 5470 }, { "epoch": 0.67, "learning_rate": 5.32167656093887e-06, "loss": 0.5451, "step": 5471 }, { "epoch": 0.67, "learning_rate": 5.3181957382235995e-06, "loss": 0.5074, "step": 5472 }, { "epoch": 0.67, "learning_rate": 5.314715641852917e-06, "loss": 0.5131, "step": 5473 }, { "epoch": 0.67, "learning_rate": 5.311236272366727e-06, "loss": 0.5181, "step": 5474 }, { "epoch": 0.67, "learning_rate": 5.307757630304819e-06, "loss": 0.4437, "step": 5475 }, { "epoch": 0.67, "learning_rate": 5.3042797162068885e-06, "loss": 0.4692, "step": 5476 }, { "epoch": 0.67, "learning_rate": 5.3008025306125e-06, "loss": 0.5059, "step": 5477 }, { "epoch": 0.67, "learning_rate": 5.297326074061109e-06, "loss": 0.5284, "step": 5478 }, { "epoch": 0.67, "learning_rate": 5.29385034709206e-06, "loss": 0.5588, "step": 5479 }, { "epoch": 0.67, "learning_rate": 5.290375350244593e-06, "loss": 0.5264, "step": 5480 }, { "epoch": 0.67, "learning_rate": 5.286901084057816e-06, "loss": 0.545, "step": 5481 }, { "epoch": 0.67, "learning_rate": 5.283427549070741e-06, "loss": 0.5809, "step": 5482 }, { "epoch": 0.67, "learning_rate": 5.2799547458222554e-06, "loss": 0.5175, "step": 5483 }, { "epoch": 0.67, "learning_rate": 5.276482674851134e-06, "loss": 0.5157, "step": 5484 }, { "epoch": 0.67, "learning_rate": 5.273011336696049e-06, "loss": 0.5302, "step": 5485 }, { "epoch": 0.67, "learning_rate": 5.269540731895546e-06, "loss": 0.5583, "step": 5486 }, { "epoch": 0.67, "learning_rate": 5.266070860988057e-06, "loss": 0.5532, "step": 5487 }, { "epoch": 0.67, "learning_rate": 5.262601724511915e-06, "loss": 0.5404, "step": 5488 }, { "epoch": 0.67, "learning_rate": 5.259133323005318e-06, "loss": 0.5937, "step": 5489 }, { "epoch": 0.67, "learning_rate": 5.25566565700637e-06, "loss": 0.4846, "step": 5490 }, { "epoch": 0.67, "learning_rate": 5.252198727053046e-06, "loss": 0.5371, "step": 5491 }, { "epoch": 0.67, "learning_rate": 5.248732533683216e-06, "loss": 0.5061, "step": 5492 }, { "epoch": 0.67, "learning_rate": 5.245267077434629e-06, "loss": 0.5393, "step": 5493 }, { "epoch": 0.67, "learning_rate": 5.241802358844927e-06, "loss": 0.4514, "step": 5494 }, { "epoch": 0.67, "learning_rate": 5.238338378451628e-06, "loss": 0.5332, "step": 5495 }, { "epoch": 0.67, "learning_rate": 5.234875136792149e-06, "loss": 0.4765, "step": 5496 }, { "epoch": 0.67, "learning_rate": 5.231412634403778e-06, "loss": 0.5333, "step": 5497 }, { "epoch": 0.67, "learning_rate": 5.227950871823693e-06, "loss": 0.5878, "step": 5498 }, { "epoch": 0.67, "learning_rate": 5.224489849588965e-06, "loss": 0.4911, "step": 5499 }, { "epoch": 0.67, "learning_rate": 5.221029568236541e-06, "loss": 0.5761, "step": 5500 }, { "epoch": 0.67, "learning_rate": 5.217570028303251e-06, "loss": 0.4953, "step": 5501 }, { "epoch": 0.67, "learning_rate": 5.214111230325826e-06, "loss": 0.4765, "step": 5502 }, { "epoch": 0.67, "learning_rate": 5.2106531748408595e-06, "loss": 0.5254, "step": 5503 }, { "epoch": 0.67, "learning_rate": 5.207195862384846e-06, "loss": 0.5243, "step": 5504 }, { "epoch": 0.67, "learning_rate": 5.203739293494166e-06, "loss": 0.5244, "step": 5505 }, { "epoch": 0.67, "learning_rate": 5.200283468705072e-06, "loss": 0.5329, "step": 5506 }, { "epoch": 0.67, "learning_rate": 5.1968283885537054e-06, "loss": 0.5679, "step": 5507 }, { "epoch": 0.67, "learning_rate": 5.193374053576101e-06, "loss": 0.5246, "step": 5508 }, { "epoch": 0.67, "learning_rate": 5.189920464308168e-06, "loss": 0.5223, "step": 5509 }, { "epoch": 0.67, "learning_rate": 5.186467621285699e-06, "loss": 0.5257, "step": 5510 }, { "epoch": 0.67, "learning_rate": 5.183015525044383e-06, "loss": 0.5708, "step": 5511 }, { "epoch": 0.67, "learning_rate": 5.179564176119779e-06, "loss": 0.4918, "step": 5512 }, { "epoch": 0.67, "learning_rate": 5.176113575047342e-06, "loss": 0.4849, "step": 5513 }, { "epoch": 0.67, "learning_rate": 5.172663722362401e-06, "loss": 0.503, "step": 5514 }, { "epoch": 0.67, "learning_rate": 5.169214618600171e-06, "loss": 0.4993, "step": 5515 }, { "epoch": 0.67, "learning_rate": 5.165766264295755e-06, "loss": 0.5312, "step": 5516 }, { "epoch": 0.67, "learning_rate": 5.162318659984144e-06, "loss": 0.4907, "step": 5517 }, { "epoch": 0.67, "learning_rate": 5.1588718062002e-06, "loss": 0.4697, "step": 5518 }, { "epoch": 0.67, "learning_rate": 5.155425703478671e-06, "loss": 0.4912, "step": 5519 }, { "epoch": 0.67, "learning_rate": 5.151980352354202e-06, "loss": 0.4791, "step": 5520 }, { "epoch": 0.67, "learning_rate": 5.148535753361303e-06, "loss": 0.572, "step": 5521 }, { "epoch": 0.67, "learning_rate": 5.145091907034385e-06, "loss": 0.5138, "step": 5522 }, { "epoch": 0.67, "learning_rate": 5.141648813907727e-06, "loss": 0.4923, "step": 5523 }, { "epoch": 0.67, "learning_rate": 5.138206474515495e-06, "loss": 0.5017, "step": 5524 }, { "epoch": 0.67, "learning_rate": 5.134764889391748e-06, "loss": 0.5346, "step": 5525 }, { "epoch": 0.67, "learning_rate": 5.131324059070416e-06, "loss": 0.5837, "step": 5526 }, { "epoch": 0.67, "learning_rate": 5.12788398408531e-06, "loss": 0.5331, "step": 5527 }, { "epoch": 0.67, "learning_rate": 5.124444664970145e-06, "loss": 0.5163, "step": 5528 }, { "epoch": 0.67, "learning_rate": 5.121006102258496e-06, "loss": 0.5063, "step": 5529 }, { "epoch": 0.67, "learning_rate": 5.1175682964838245e-06, "loss": 0.4938, "step": 5530 }, { "epoch": 0.67, "learning_rate": 5.114131248179487e-06, "loss": 0.5307, "step": 5531 }, { "epoch": 0.67, "learning_rate": 5.1106949578787106e-06, "loss": 0.4702, "step": 5532 }, { "epoch": 0.67, "learning_rate": 5.107259426114602e-06, "loss": 0.5854, "step": 5533 }, { "epoch": 0.67, "learning_rate": 5.103824653420167e-06, "loss": 0.4661, "step": 5534 }, { "epoch": 0.67, "learning_rate": 5.1003906403282755e-06, "loss": 0.5388, "step": 5535 }, { "epoch": 0.67, "learning_rate": 5.096957387371693e-06, "loss": 0.4786, "step": 5536 }, { "epoch": 0.67, "learning_rate": 5.09352489508306e-06, "loss": 0.5387, "step": 5537 }, { "epoch": 0.67, "learning_rate": 5.090093163994892e-06, "loss": 0.5567, "step": 5538 }, { "epoch": 0.67, "learning_rate": 5.0866621946396045e-06, "loss": 0.5554, "step": 5539 }, { "epoch": 0.67, "learning_rate": 5.083231987549484e-06, "loss": 0.5338, "step": 5540 }, { "epoch": 0.67, "learning_rate": 5.079802543256698e-06, "loss": 0.4958, "step": 5541 }, { "epoch": 0.67, "learning_rate": 5.076373862293293e-06, "loss": 0.5029, "step": 5542 }, { "epoch": 0.67, "learning_rate": 5.072945945191211e-06, "loss": 0.5153, "step": 5543 }, { "epoch": 0.67, "learning_rate": 5.0695187924822556e-06, "loss": 0.5447, "step": 5544 }, { "epoch": 0.67, "learning_rate": 5.066092404698132e-06, "loss": 0.5457, "step": 5545 }, { "epoch": 0.67, "learning_rate": 5.0626667823704116e-06, "loss": 0.4806, "step": 5546 }, { "epoch": 0.67, "learning_rate": 5.0592419260305495e-06, "loss": 0.5612, "step": 5547 }, { "epoch": 0.67, "learning_rate": 5.055817836209891e-06, "loss": 0.4619, "step": 5548 }, { "epoch": 0.67, "learning_rate": 5.052394513439656e-06, "loss": 0.5572, "step": 5549 }, { "epoch": 0.67, "learning_rate": 5.0489719582509365e-06, "loss": 0.4938, "step": 5550 }, { "epoch": 0.68, "learning_rate": 5.045550171174724e-06, "loss": 0.5605, "step": 5551 }, { "epoch": 0.68, "learning_rate": 5.042129152741882e-06, "loss": 0.5102, "step": 5552 }, { "epoch": 0.68, "learning_rate": 5.038708903483147e-06, "loss": 0.5974, "step": 5553 }, { "epoch": 0.68, "learning_rate": 5.03528942392915e-06, "loss": 0.4697, "step": 5554 }, { "epoch": 0.68, "learning_rate": 5.031870714610396e-06, "loss": 0.5312, "step": 5555 }, { "epoch": 0.68, "learning_rate": 5.0284527760572645e-06, "loss": 0.5716, "step": 5556 }, { "epoch": 0.68, "learning_rate": 5.025035608800027e-06, "loss": 0.548, "step": 5557 }, { "epoch": 0.68, "learning_rate": 5.021619213368829e-06, "loss": 0.5238, "step": 5558 }, { "epoch": 0.68, "learning_rate": 5.018203590293691e-06, "loss": 0.5171, "step": 5559 }, { "epoch": 0.68, "learning_rate": 5.014788740104528e-06, "loss": 0.5397, "step": 5560 }, { "epoch": 0.68, "learning_rate": 5.011374663331119e-06, "loss": 0.5754, "step": 5561 }, { "epoch": 0.68, "learning_rate": 5.00796136050314e-06, "loss": 0.5459, "step": 5562 }, { "epoch": 0.68, "learning_rate": 5.0045488321501275e-06, "loss": 0.5425, "step": 5563 }, { "epoch": 0.68, "learning_rate": 5.001137078801518e-06, "loss": 0.5622, "step": 5564 }, { "epoch": 0.68, "learning_rate": 4.9977261009866086e-06, "loss": 0.5816, "step": 5565 }, { "epoch": 0.68, "learning_rate": 4.994315899234593e-06, "loss": 0.5544, "step": 5566 }, { "epoch": 0.68, "learning_rate": 4.990906474074531e-06, "loss": 0.5239, "step": 5567 }, { "epoch": 0.68, "learning_rate": 4.9874978260353726e-06, "loss": 0.5054, "step": 5568 }, { "epoch": 0.68, "learning_rate": 4.98408995564594e-06, "loss": 0.5164, "step": 5569 }, { "epoch": 0.68, "learning_rate": 4.980682863434933e-06, "loss": 0.4208, "step": 5570 }, { "epoch": 0.68, "learning_rate": 4.9772765499309436e-06, "loss": 0.4897, "step": 5571 }, { "epoch": 0.68, "learning_rate": 4.973871015662429e-06, "loss": 0.5093, "step": 5572 }, { "epoch": 0.68, "learning_rate": 4.970466261157727e-06, "loss": 0.5359, "step": 5573 }, { "epoch": 0.68, "learning_rate": 4.967062286945067e-06, "loss": 0.5283, "step": 5574 }, { "epoch": 0.68, "learning_rate": 4.963659093552538e-06, "loss": 0.5429, "step": 5575 }, { "epoch": 0.68, "learning_rate": 4.960256681508125e-06, "loss": 0.5187, "step": 5576 }, { "epoch": 0.68, "learning_rate": 4.956855051339689e-06, "loss": 0.5062, "step": 5577 }, { "epoch": 0.68, "learning_rate": 4.953454203574958e-06, "loss": 0.5453, "step": 5578 }, { "epoch": 0.68, "learning_rate": 4.9500541387415466e-06, "loss": 0.5817, "step": 5579 }, { "epoch": 0.68, "learning_rate": 4.946654857366955e-06, "loss": 0.5516, "step": 5580 }, { "epoch": 0.68, "learning_rate": 4.943256359978549e-06, "loss": 0.5949, "step": 5581 }, { "epoch": 0.68, "learning_rate": 4.939858647103575e-06, "loss": 0.5154, "step": 5582 }, { "epoch": 0.68, "learning_rate": 4.93646171926917e-06, "loss": 0.507, "step": 5583 }, { "epoch": 0.68, "learning_rate": 4.93306557700233e-06, "loss": 0.5728, "step": 5584 }, { "epoch": 0.68, "learning_rate": 4.92967022082995e-06, "loss": 0.5349, "step": 5585 }, { "epoch": 0.68, "learning_rate": 4.926275651278788e-06, "loss": 0.5585, "step": 5586 }, { "epoch": 0.68, "learning_rate": 4.9228818688754795e-06, "loss": 0.5224, "step": 5587 }, { "epoch": 0.68, "learning_rate": 4.919488874146547e-06, "loss": 0.5225, "step": 5588 }, { "epoch": 0.68, "learning_rate": 4.91609666761839e-06, "loss": 0.5316, "step": 5589 }, { "epoch": 0.68, "learning_rate": 4.912705249817275e-06, "loss": 0.5484, "step": 5590 }, { "epoch": 0.68, "learning_rate": 4.909314621269362e-06, "loss": 0.5065, "step": 5591 }, { "epoch": 0.68, "learning_rate": 4.905924782500674e-06, "loss": 0.5292, "step": 5592 }, { "epoch": 0.68, "learning_rate": 4.902535734037116e-06, "loss": 0.5034, "step": 5593 }, { "epoch": 0.68, "learning_rate": 4.899147476404478e-06, "loss": 0.4953, "step": 5594 }, { "epoch": 0.68, "learning_rate": 4.895760010128416e-06, "loss": 0.5241, "step": 5595 }, { "epoch": 0.68, "learning_rate": 4.8923733357344685e-06, "loss": 0.5357, "step": 5596 }, { "epoch": 0.68, "learning_rate": 4.888987453748056e-06, "loss": 0.4806, "step": 5597 }, { "epoch": 0.68, "learning_rate": 4.885602364694466e-06, "loss": 0.5059, "step": 5598 }, { "epoch": 0.68, "learning_rate": 4.882218069098863e-06, "loss": 0.5615, "step": 5599 }, { "epoch": 0.68, "learning_rate": 4.878834567486308e-06, "loss": 0.5113, "step": 5600 }, { "epoch": 0.68, "learning_rate": 4.875451860381717e-06, "loss": 0.5443, "step": 5601 }, { "epoch": 0.68, "learning_rate": 4.872069948309885e-06, "loss": 0.5543, "step": 5602 }, { "epoch": 0.68, "learning_rate": 4.868688831795496e-06, "loss": 0.4829, "step": 5603 }, { "epoch": 0.68, "learning_rate": 4.865308511363102e-06, "loss": 0.508, "step": 5604 }, { "epoch": 0.68, "learning_rate": 4.861928987537126e-06, "loss": 0.5173, "step": 5605 }, { "epoch": 0.68, "learning_rate": 4.858550260841883e-06, "loss": 0.5702, "step": 5606 }, { "epoch": 0.68, "learning_rate": 4.85517233180155e-06, "loss": 0.5744, "step": 5607 }, { "epoch": 0.68, "learning_rate": 4.851795200940189e-06, "loss": 0.5569, "step": 5608 }, { "epoch": 0.68, "learning_rate": 4.848418868781735e-06, "loss": 0.5265, "step": 5609 }, { "epoch": 0.68, "learning_rate": 4.845043335849993e-06, "loss": 0.5812, "step": 5610 }, { "epoch": 0.68, "learning_rate": 4.8416686026686546e-06, "loss": 0.5912, "step": 5611 }, { "epoch": 0.68, "learning_rate": 4.838294669761287e-06, "loss": 0.5196, "step": 5612 }, { "epoch": 0.68, "learning_rate": 4.834921537651319e-06, "loss": 0.4827, "step": 5613 }, { "epoch": 0.68, "learning_rate": 4.8315492068620765e-06, "loss": 0.538, "step": 5614 }, { "epoch": 0.68, "learning_rate": 4.828177677916742e-06, "loss": 0.581, "step": 5615 }, { "epoch": 0.68, "learning_rate": 4.8248069513383815e-06, "loss": 0.4766, "step": 5616 }, { "epoch": 0.68, "learning_rate": 4.82143702764994e-06, "loss": 0.4824, "step": 5617 }, { "epoch": 0.68, "learning_rate": 4.818067907374233e-06, "loss": 0.5438, "step": 5618 }, { "epoch": 0.68, "learning_rate": 4.814699591033949e-06, "loss": 0.4914, "step": 5619 }, { "epoch": 0.68, "learning_rate": 4.811332079151661e-06, "loss": 0.4809, "step": 5620 }, { "epoch": 0.68, "learning_rate": 4.80796537224981e-06, "loss": 0.5123, "step": 5621 }, { "epoch": 0.68, "learning_rate": 4.804599470850708e-06, "loss": 0.5476, "step": 5622 }, { "epoch": 0.68, "learning_rate": 4.801234375476553e-06, "loss": 0.4715, "step": 5623 }, { "epoch": 0.68, "learning_rate": 4.7978700866494155e-06, "loss": 0.4723, "step": 5624 }, { "epoch": 0.68, "learning_rate": 4.794506604891231e-06, "loss": 0.4972, "step": 5625 }, { "epoch": 0.68, "learning_rate": 4.791143930723825e-06, "loss": 0.539, "step": 5626 }, { "epoch": 0.68, "learning_rate": 4.787782064668885e-06, "loss": 0.4908, "step": 5627 }, { "epoch": 0.68, "learning_rate": 4.784421007247975e-06, "loss": 0.6114, "step": 5628 }, { "epoch": 0.68, "learning_rate": 4.7810607589825404e-06, "loss": 0.5161, "step": 5629 }, { "epoch": 0.68, "learning_rate": 4.777701320393895e-06, "loss": 0.4793, "step": 5630 }, { "epoch": 0.68, "learning_rate": 4.774342692003231e-06, "loss": 0.5314, "step": 5631 }, { "epoch": 0.68, "learning_rate": 4.770984874331612e-06, "loss": 0.5252, "step": 5632 }, { "epoch": 0.68, "learning_rate": 4.767627867899972e-06, "loss": 0.4984, "step": 5633 }, { "epoch": 0.69, "learning_rate": 4.764271673229132e-06, "loss": 0.5246, "step": 5634 }, { "epoch": 0.69, "learning_rate": 4.760916290839771e-06, "loss": 0.5569, "step": 5635 }, { "epoch": 0.69, "learning_rate": 4.757561721252453e-06, "loss": 0.5252, "step": 5636 }, { "epoch": 0.69, "learning_rate": 4.754207964987616e-06, "loss": 0.4883, "step": 5637 }, { "epoch": 0.69, "learning_rate": 4.750855022565567e-06, "loss": 0.5226, "step": 5638 }, { "epoch": 0.69, "learning_rate": 4.747502894506481e-06, "loss": 0.5172, "step": 5639 }, { "epoch": 0.69, "learning_rate": 4.7441515813304265e-06, "loss": 0.4863, "step": 5640 }, { "epoch": 0.69, "learning_rate": 4.740801083557325e-06, "loss": 0.5494, "step": 5641 }, { "epoch": 0.69, "learning_rate": 4.737451401706976e-06, "loss": 0.5515, "step": 5642 }, { "epoch": 0.69, "learning_rate": 4.734102536299065e-06, "loss": 0.5426, "step": 5643 }, { "epoch": 0.69, "learning_rate": 4.730754487853139e-06, "loss": 0.5132, "step": 5644 }, { "epoch": 0.69, "learning_rate": 4.7274072568886145e-06, "loss": 0.4788, "step": 5645 }, { "epoch": 0.69, "learning_rate": 4.724060843924796e-06, "loss": 0.533, "step": 5646 }, { "epoch": 0.69, "learning_rate": 4.720715249480846e-06, "loss": 0.5031, "step": 5647 }, { "epoch": 0.69, "learning_rate": 4.7173704740758105e-06, "loss": 0.5289, "step": 5648 }, { "epoch": 0.69, "learning_rate": 4.714026518228609e-06, "loss": 0.5172, "step": 5649 }, { "epoch": 0.69, "learning_rate": 4.710683382458024e-06, "loss": 0.497, "step": 5650 }, { "epoch": 0.69, "learning_rate": 4.707341067282713e-06, "loss": 0.494, "step": 5651 }, { "epoch": 0.69, "learning_rate": 4.70399957322122e-06, "loss": 0.586, "step": 5652 }, { "epoch": 0.69, "learning_rate": 4.700658900791939e-06, "loss": 0.5295, "step": 5653 }, { "epoch": 0.69, "learning_rate": 4.697319050513159e-06, "loss": 0.4894, "step": 5654 }, { "epoch": 0.69, "learning_rate": 4.693980022903028e-06, "loss": 0.5548, "step": 5655 }, { "epoch": 0.69, "learning_rate": 4.690641818479563e-06, "loss": 0.4996, "step": 5656 }, { "epoch": 0.69, "learning_rate": 4.68730443776067e-06, "loss": 0.525, "step": 5657 }, { "epoch": 0.69, "learning_rate": 4.683967881264112e-06, "loss": 0.531, "step": 5658 }, { "epoch": 0.69, "learning_rate": 4.680632149507523e-06, "loss": 0.5305, "step": 5659 }, { "epoch": 0.69, "learning_rate": 4.677297243008427e-06, "loss": 0.5132, "step": 5660 }, { "epoch": 0.69, "learning_rate": 4.673963162284203e-06, "loss": 0.5186, "step": 5661 }, { "epoch": 0.69, "learning_rate": 4.670629907852102e-06, "loss": 0.5253, "step": 5662 }, { "epoch": 0.69, "learning_rate": 4.6672974802292625e-06, "loss": 0.5273, "step": 5663 }, { "epoch": 0.69, "learning_rate": 4.663965879932676e-06, "loss": 0.5953, "step": 5664 }, { "epoch": 0.69, "learning_rate": 4.6606351074792134e-06, "loss": 0.4972, "step": 5665 }, { "epoch": 0.69, "learning_rate": 4.657305163385622e-06, "loss": 0.5218, "step": 5666 }, { "epoch": 0.69, "learning_rate": 4.653976048168515e-06, "loss": 0.5346, "step": 5667 }, { "epoch": 0.69, "learning_rate": 4.6506477623443725e-06, "loss": 0.4736, "step": 5668 }, { "epoch": 0.69, "learning_rate": 4.64732030642956e-06, "loss": 0.5894, "step": 5669 }, { "epoch": 0.69, "learning_rate": 4.643993680940299e-06, "loss": 0.4904, "step": 5670 }, { "epoch": 0.69, "learning_rate": 4.6406678863926905e-06, "loss": 0.5243, "step": 5671 }, { "epoch": 0.69, "learning_rate": 4.637342923302709e-06, "loss": 0.4785, "step": 5672 }, { "epoch": 0.69, "learning_rate": 4.634018792186194e-06, "loss": 0.5465, "step": 5673 }, { "epoch": 0.69, "learning_rate": 4.630695493558853e-06, "loss": 0.5132, "step": 5674 }, { "epoch": 0.69, "learning_rate": 4.627373027936277e-06, "loss": 0.5227, "step": 5675 }, { "epoch": 0.69, "learning_rate": 4.624051395833913e-06, "loss": 0.5296, "step": 5676 }, { "epoch": 0.69, "learning_rate": 4.620730597767092e-06, "loss": 0.5188, "step": 5677 }, { "epoch": 0.69, "learning_rate": 4.6174106342510075e-06, "loss": 0.4879, "step": 5678 }, { "epoch": 0.69, "learning_rate": 4.614091505800719e-06, "loss": 0.5515, "step": 5679 }, { "epoch": 0.69, "learning_rate": 4.610773212931172e-06, "loss": 0.5553, "step": 5680 }, { "epoch": 0.69, "learning_rate": 4.60745575615717e-06, "loss": 0.5228, "step": 5681 }, { "epoch": 0.69, "learning_rate": 4.604139135993385e-06, "loss": 0.5578, "step": 5682 }, { "epoch": 0.69, "learning_rate": 4.600823352954368e-06, "loss": 0.5157, "step": 5683 }, { "epoch": 0.69, "learning_rate": 4.597508407554541e-06, "loss": 0.4902, "step": 5684 }, { "epoch": 0.69, "learning_rate": 4.594194300308182e-06, "loss": 0.4867, "step": 5685 }, { "epoch": 0.69, "learning_rate": 4.590881031729459e-06, "loss": 0.5284, "step": 5686 }, { "epoch": 0.69, "learning_rate": 4.58756860233239e-06, "loss": 0.5525, "step": 5687 }, { "epoch": 0.69, "learning_rate": 4.584257012630875e-06, "loss": 0.5624, "step": 5688 }, { "epoch": 0.69, "learning_rate": 4.580946263138683e-06, "loss": 0.5279, "step": 5689 }, { "epoch": 0.69, "learning_rate": 4.577636354369449e-06, "loss": 0.5915, "step": 5690 }, { "epoch": 0.69, "learning_rate": 4.574327286836675e-06, "loss": 0.5086, "step": 5691 }, { "epoch": 0.69, "learning_rate": 4.571019061053743e-06, "loss": 0.5508, "step": 5692 }, { "epoch": 0.69, "learning_rate": 4.567711677533891e-06, "loss": 0.5431, "step": 5693 }, { "epoch": 0.69, "learning_rate": 4.564405136790241e-06, "loss": 0.584, "step": 5694 }, { "epoch": 0.69, "learning_rate": 4.5610994393357685e-06, "loss": 0.5744, "step": 5695 }, { "epoch": 0.69, "learning_rate": 4.557794585683334e-06, "loss": 0.4918, "step": 5696 }, { "epoch": 0.69, "learning_rate": 4.55449057634565e-06, "loss": 0.5108, "step": 5697 }, { "epoch": 0.69, "learning_rate": 4.5511874118353175e-06, "loss": 0.5327, "step": 5698 }, { "epoch": 0.69, "learning_rate": 4.547885092664787e-06, "loss": 0.4986, "step": 5699 }, { "epoch": 0.69, "learning_rate": 4.544583619346393e-06, "loss": 0.5725, "step": 5700 }, { "epoch": 0.69, "learning_rate": 4.5412829923923314e-06, "loss": 0.5215, "step": 5701 }, { "epoch": 0.69, "learning_rate": 4.5379832123146635e-06, "loss": 0.5143, "step": 5702 }, { "epoch": 0.69, "learning_rate": 4.5346842796253325e-06, "loss": 0.5516, "step": 5703 }, { "epoch": 0.69, "learning_rate": 4.531386194836136e-06, "loss": 0.5319, "step": 5704 }, { "epoch": 0.69, "learning_rate": 4.528088958458742e-06, "loss": 0.5402, "step": 5705 }, { "epoch": 0.69, "learning_rate": 4.524792571004698e-06, "loss": 0.5249, "step": 5706 }, { "epoch": 0.69, "learning_rate": 4.521497032985406e-06, "loss": 0.5218, "step": 5707 }, { "epoch": 0.69, "learning_rate": 4.518202344912146e-06, "loss": 0.5047, "step": 5708 }, { "epoch": 0.69, "learning_rate": 4.5149085072960645e-06, "loss": 0.5023, "step": 5709 }, { "epoch": 0.69, "learning_rate": 4.511615520648171e-06, "loss": 0.5403, "step": 5710 }, { "epoch": 0.69, "learning_rate": 4.508323385479344e-06, "loss": 0.4968, "step": 5711 }, { "epoch": 0.69, "learning_rate": 4.505032102300337e-06, "loss": 0.4764, "step": 5712 }, { "epoch": 0.69, "learning_rate": 4.501741671621765e-06, "loss": 0.5348, "step": 5713 }, { "epoch": 0.69, "learning_rate": 4.498452093954108e-06, "loss": 0.543, "step": 5714 }, { "epoch": 0.69, "learning_rate": 4.495163369807723e-06, "loss": 0.4596, "step": 5715 }, { "epoch": 0.7, "learning_rate": 4.491875499692825e-06, "loss": 0.5218, "step": 5716 }, { "epoch": 0.7, "learning_rate": 4.488588484119505e-06, "loss": 0.5413, "step": 5717 }, { "epoch": 0.7, "learning_rate": 4.485302323597716e-06, "loss": 0.5829, "step": 5718 }, { "epoch": 0.7, "learning_rate": 4.482017018637275e-06, "loss": 0.5332, "step": 5719 }, { "epoch": 0.7, "learning_rate": 4.478732569747873e-06, "loss": 0.5081, "step": 5720 }, { "epoch": 0.7, "learning_rate": 4.475448977439073e-06, "loss": 0.5136, "step": 5721 }, { "epoch": 0.7, "learning_rate": 4.472166242220292e-06, "loss": 0.5091, "step": 5722 }, { "epoch": 0.7, "learning_rate": 4.468884364600818e-06, "loss": 0.5518, "step": 5723 }, { "epoch": 0.7, "learning_rate": 4.465603345089814e-06, "loss": 0.4784, "step": 5724 }, { "epoch": 0.7, "learning_rate": 4.462323184196298e-06, "loss": 0.5018, "step": 5725 }, { "epoch": 0.7, "learning_rate": 4.459043882429167e-06, "loss": 0.5146, "step": 5726 }, { "epoch": 0.7, "learning_rate": 4.455765440297176e-06, "loss": 0.5012, "step": 5727 }, { "epoch": 0.7, "learning_rate": 4.452487858308945e-06, "loss": 0.5102, "step": 5728 }, { "epoch": 0.7, "learning_rate": 4.449211136972972e-06, "loss": 0.5499, "step": 5729 }, { "epoch": 0.7, "learning_rate": 4.445935276797609e-06, "loss": 0.5019, "step": 5730 }, { "epoch": 0.7, "learning_rate": 4.442660278291074e-06, "loss": 0.5615, "step": 5731 }, { "epoch": 0.7, "learning_rate": 4.439386141961473e-06, "loss": 0.5328, "step": 5732 }, { "epoch": 0.7, "learning_rate": 4.436112868316751e-06, "loss": 0.5114, "step": 5733 }, { "epoch": 0.7, "learning_rate": 4.43284045786473e-06, "loss": 0.5117, "step": 5734 }, { "epoch": 0.7, "learning_rate": 4.429568911113104e-06, "loss": 0.551, "step": 5735 }, { "epoch": 0.7, "learning_rate": 4.426298228569423e-06, "loss": 0.5303, "step": 5736 }, { "epoch": 0.7, "learning_rate": 4.423028410741106e-06, "loss": 0.5293, "step": 5737 }, { "epoch": 0.7, "learning_rate": 4.4197594581354445e-06, "loss": 0.5308, "step": 5738 }, { "epoch": 0.7, "learning_rate": 4.4164913712595835e-06, "loss": 0.5298, "step": 5739 }, { "epoch": 0.7, "learning_rate": 4.413224150620548e-06, "loss": 0.5333, "step": 5740 }, { "epoch": 0.7, "learning_rate": 4.409957796725219e-06, "loss": 0.5064, "step": 5741 }, { "epoch": 0.7, "learning_rate": 4.406692310080339e-06, "loss": 0.5101, "step": 5742 }, { "epoch": 0.7, "learning_rate": 4.403427691192529e-06, "loss": 0.5009, "step": 5743 }, { "epoch": 0.7, "learning_rate": 4.400163940568269e-06, "loss": 0.5171, "step": 5744 }, { "epoch": 0.7, "learning_rate": 4.396901058713902e-06, "loss": 0.5384, "step": 5745 }, { "epoch": 0.7, "learning_rate": 4.393639046135634e-06, "loss": 0.5577, "step": 5746 }, { "epoch": 0.7, "learning_rate": 4.390377903339548e-06, "loss": 0.5771, "step": 5747 }, { "epoch": 0.7, "learning_rate": 4.387117630831577e-06, "loss": 0.5273, "step": 5748 }, { "epoch": 0.7, "learning_rate": 4.383858229117532e-06, "loss": 0.5729, "step": 5749 }, { "epoch": 0.7, "learning_rate": 4.380599698703081e-06, "loss": 0.5693, "step": 5750 }, { "epoch": 0.7, "learning_rate": 4.377342040093756e-06, "loss": 0.4976, "step": 5751 }, { "epoch": 0.7, "learning_rate": 4.3740852537949615e-06, "loss": 0.5511, "step": 5752 }, { "epoch": 0.7, "learning_rate": 4.370829340311962e-06, "loss": 0.548, "step": 5753 }, { "epoch": 0.7, "learning_rate": 4.36757430014988e-06, "loss": 0.5318, "step": 5754 }, { "epoch": 0.7, "learning_rate": 4.364320133813712e-06, "loss": 0.5424, "step": 5755 }, { "epoch": 0.7, "learning_rate": 4.361066841808323e-06, "loss": 0.512, "step": 5756 }, { "epoch": 0.7, "learning_rate": 4.357814424638424e-06, "loss": 0.5343, "step": 5757 }, { "epoch": 0.7, "learning_rate": 4.354562882808612e-06, "loss": 0.4774, "step": 5758 }, { "epoch": 0.7, "learning_rate": 4.3513122168233326e-06, "loss": 0.487, "step": 5759 }, { "epoch": 0.7, "learning_rate": 4.348062427186898e-06, "loss": 0.5483, "step": 5760 }, { "epoch": 0.7, "learning_rate": 4.3448135144034935e-06, "loss": 0.5374, "step": 5761 }, { "epoch": 0.7, "learning_rate": 4.341565478977159e-06, "loss": 0.559, "step": 5762 }, { "epoch": 0.7, "learning_rate": 4.338318321411796e-06, "loss": 0.5246, "step": 5763 }, { "epoch": 0.7, "learning_rate": 4.335072042211186e-06, "loss": 0.5303, "step": 5764 }, { "epoch": 0.7, "learning_rate": 4.331826641878953e-06, "loss": 0.5098, "step": 5765 }, { "epoch": 0.7, "learning_rate": 4.328582120918602e-06, "loss": 0.5171, "step": 5766 }, { "epoch": 0.7, "learning_rate": 4.325338479833488e-06, "loss": 0.5555, "step": 5767 }, { "epoch": 0.7, "learning_rate": 4.322095719126846e-06, "loss": 0.489, "step": 5768 }, { "epoch": 0.7, "learning_rate": 4.318853839301753e-06, "loss": 0.4848, "step": 5769 }, { "epoch": 0.7, "learning_rate": 4.31561284086117e-06, "loss": 0.5321, "step": 5770 }, { "epoch": 0.7, "learning_rate": 4.312372724307905e-06, "loss": 0.5649, "step": 5771 }, { "epoch": 0.7, "learning_rate": 4.309133490144644e-06, "loss": 0.5652, "step": 5772 }, { "epoch": 0.7, "learning_rate": 4.3058951388739225e-06, "loss": 0.4936, "step": 5773 }, { "epoch": 0.7, "learning_rate": 4.302657670998143e-06, "loss": 0.5427, "step": 5774 }, { "epoch": 0.7, "learning_rate": 4.29942108701958e-06, "loss": 0.5442, "step": 5775 }, { "epoch": 0.7, "learning_rate": 4.296185387440359e-06, "loss": 0.5596, "step": 5776 }, { "epoch": 0.7, "learning_rate": 4.292950572762471e-06, "loss": 0.5781, "step": 5777 }, { "epoch": 0.7, "learning_rate": 4.289716643487777e-06, "loss": 0.5373, "step": 5778 }, { "epoch": 0.7, "learning_rate": 4.28648360011799e-06, "loss": 0.4902, "step": 5779 }, { "epoch": 0.7, "learning_rate": 4.283251443154693e-06, "loss": 0.4742, "step": 5780 }, { "epoch": 0.7, "learning_rate": 4.280020173099334e-06, "loss": 0.5575, "step": 5781 }, { "epoch": 0.7, "learning_rate": 4.276789790453215e-06, "loss": 0.5106, "step": 5782 }, { "epoch": 0.7, "learning_rate": 4.273560295717499e-06, "loss": 0.5449, "step": 5783 }, { "epoch": 0.7, "learning_rate": 4.270331689393227e-06, "loss": 0.5158, "step": 5784 }, { "epoch": 0.7, "learning_rate": 4.267103971981284e-06, "loss": 0.4828, "step": 5785 }, { "epoch": 0.7, "learning_rate": 4.263877143982423e-06, "loss": 0.504, "step": 5786 }, { "epoch": 0.7, "learning_rate": 4.260651205897267e-06, "loss": 0.5257, "step": 5787 }, { "epoch": 0.7, "learning_rate": 4.2574261582262886e-06, "loss": 0.5688, "step": 5788 }, { "epoch": 0.7, "learning_rate": 4.254202001469835e-06, "loss": 0.4867, "step": 5789 }, { "epoch": 0.7, "learning_rate": 4.250978736128104e-06, "loss": 0.481, "step": 5790 }, { "epoch": 0.7, "learning_rate": 4.247756362701156e-06, "loss": 0.5393, "step": 5791 }, { "epoch": 0.7, "learning_rate": 4.244534881688922e-06, "loss": 0.5204, "step": 5792 }, { "epoch": 0.7, "learning_rate": 4.241314293591191e-06, "loss": 0.5716, "step": 5793 }, { "epoch": 0.7, "learning_rate": 4.238094598907604e-06, "loss": 0.5299, "step": 5794 }, { "epoch": 0.7, "learning_rate": 4.234875798137679e-06, "loss": 0.4878, "step": 5795 }, { "epoch": 0.7, "learning_rate": 4.231657891780783e-06, "loss": 0.5721, "step": 5796 }, { "epoch": 0.7, "learning_rate": 4.228440880336144e-06, "loss": 0.5262, "step": 5797 }, { "epoch": 0.71, "learning_rate": 4.225224764302866e-06, "loss": 0.5441, "step": 5798 }, { "epoch": 0.71, "learning_rate": 4.222009544179898e-06, "loss": 0.4942, "step": 5799 }, { "epoch": 0.71, "learning_rate": 4.218795220466053e-06, "loss": 0.5407, "step": 5800 }, { "epoch": 0.71, "learning_rate": 4.2155817936600126e-06, "loss": 0.5424, "step": 5801 }, { "epoch": 0.71, "learning_rate": 4.212369264260314e-06, "loss": 0.5013, "step": 5802 }, { "epoch": 0.71, "learning_rate": 4.2091576327653475e-06, "loss": 0.5147, "step": 5803 }, { "epoch": 0.71, "learning_rate": 4.205946899673384e-06, "loss": 0.6042, "step": 5804 }, { "epoch": 0.71, "learning_rate": 4.202737065482538e-06, "loss": 0.5409, "step": 5805 }, { "epoch": 0.71, "learning_rate": 4.199528130690786e-06, "loss": 0.5483, "step": 5806 }, { "epoch": 0.71, "learning_rate": 4.196320095795976e-06, "loss": 0.5021, "step": 5807 }, { "epoch": 0.71, "learning_rate": 4.193112961295805e-06, "loss": 0.4812, "step": 5808 }, { "epoch": 0.71, "learning_rate": 4.18990672768783e-06, "loss": 0.4538, "step": 5809 }, { "epoch": 0.71, "learning_rate": 4.186701395469481e-06, "loss": 0.5278, "step": 5810 }, { "epoch": 0.71, "learning_rate": 4.183496965138033e-06, "loss": 0.5353, "step": 5811 }, { "epoch": 0.71, "learning_rate": 4.1802934371906335e-06, "loss": 0.529, "step": 5812 }, { "epoch": 0.71, "learning_rate": 4.17709081212428e-06, "loss": 0.5405, "step": 5813 }, { "epoch": 0.71, "learning_rate": 4.173889090435833e-06, "loss": 0.5739, "step": 5814 }, { "epoch": 0.71, "learning_rate": 4.1706882726220165e-06, "loss": 0.5374, "step": 5815 }, { "epoch": 0.71, "learning_rate": 4.167488359179416e-06, "loss": 0.4782, "step": 5816 }, { "epoch": 0.71, "learning_rate": 4.1642893506044645e-06, "loss": 0.4755, "step": 5817 }, { "epoch": 0.71, "learning_rate": 4.16109124739347e-06, "loss": 0.5394, "step": 5818 }, { "epoch": 0.71, "learning_rate": 4.157894050042589e-06, "loss": 0.479, "step": 5819 }, { "epoch": 0.71, "learning_rate": 4.154697759047839e-06, "loss": 0.5857, "step": 5820 }, { "epoch": 0.71, "learning_rate": 4.151502374905104e-06, "loss": 0.5048, "step": 5821 }, { "epoch": 0.71, "learning_rate": 4.148307898110121e-06, "loss": 0.5161, "step": 5822 }, { "epoch": 0.71, "learning_rate": 4.145114329158482e-06, "loss": 0.5091, "step": 5823 }, { "epoch": 0.71, "learning_rate": 4.141921668545652e-06, "loss": 0.5476, "step": 5824 }, { "epoch": 0.71, "learning_rate": 4.138729916766944e-06, "loss": 0.6025, "step": 5825 }, { "epoch": 0.71, "learning_rate": 4.135539074317527e-06, "loss": 0.4702, "step": 5826 }, { "epoch": 0.71, "learning_rate": 4.13234914169244e-06, "loss": 0.5078, "step": 5827 }, { "epoch": 0.71, "learning_rate": 4.129160119386579e-06, "loss": 0.4867, "step": 5828 }, { "epoch": 0.71, "learning_rate": 4.1259720078946885e-06, "loss": 0.5416, "step": 5829 }, { "epoch": 0.71, "learning_rate": 4.122784807711385e-06, "loss": 0.5257, "step": 5830 }, { "epoch": 0.71, "learning_rate": 4.119598519331134e-06, "loss": 0.5281, "step": 5831 }, { "epoch": 0.71, "learning_rate": 4.116413143248258e-06, "loss": 0.5338, "step": 5832 }, { "epoch": 0.71, "learning_rate": 4.113228679956952e-06, "loss": 0.6135, "step": 5833 }, { "epoch": 0.71, "learning_rate": 4.110045129951251e-06, "loss": 0.4979, "step": 5834 }, { "epoch": 0.71, "learning_rate": 4.106862493725066e-06, "loss": 0.5289, "step": 5835 }, { "epoch": 0.71, "learning_rate": 4.103680771772153e-06, "loss": 0.5235, "step": 5836 }, { "epoch": 0.71, "learning_rate": 4.1004999645861276e-06, "loss": 0.5032, "step": 5837 }, { "epoch": 0.71, "learning_rate": 4.097320072660474e-06, "loss": 0.4568, "step": 5838 }, { "epoch": 0.71, "learning_rate": 4.094141096488519e-06, "loss": 0.4821, "step": 5839 }, { "epoch": 0.71, "learning_rate": 4.09096303656346e-06, "loss": 0.5466, "step": 5840 }, { "epoch": 0.71, "learning_rate": 4.087785893378351e-06, "loss": 0.5465, "step": 5841 }, { "epoch": 0.71, "learning_rate": 4.084609667426097e-06, "loss": 0.5309, "step": 5842 }, { "epoch": 0.71, "learning_rate": 4.0814343591994596e-06, "loss": 0.5198, "step": 5843 }, { "epoch": 0.71, "learning_rate": 4.0782599691910715e-06, "loss": 0.5295, "step": 5844 }, { "epoch": 0.71, "learning_rate": 4.075086497893408e-06, "loss": 0.4895, "step": 5845 }, { "epoch": 0.71, "learning_rate": 4.071913945798806e-06, "loss": 0.4697, "step": 5846 }, { "epoch": 0.71, "learning_rate": 4.0687423133994685e-06, "loss": 0.4843, "step": 5847 }, { "epoch": 0.71, "learning_rate": 4.065571601187446e-06, "loss": 0.4925, "step": 5848 }, { "epoch": 0.71, "learning_rate": 4.062401809654643e-06, "loss": 0.514, "step": 5849 }, { "epoch": 0.71, "learning_rate": 4.059232939292839e-06, "loss": 0.5061, "step": 5850 }, { "epoch": 0.71, "learning_rate": 4.056064990593647e-06, "loss": 0.4552, "step": 5851 }, { "epoch": 0.71, "learning_rate": 4.052897964048556e-06, "loss": 0.5095, "step": 5852 }, { "epoch": 0.71, "learning_rate": 4.049731860148907e-06, "loss": 0.4976, "step": 5853 }, { "epoch": 0.71, "learning_rate": 4.0465666793858916e-06, "loss": 0.4627, "step": 5854 }, { "epoch": 0.71, "learning_rate": 4.04340242225056e-06, "loss": 0.5247, "step": 5855 }, { "epoch": 0.71, "learning_rate": 4.040239089233828e-06, "loss": 0.5249, "step": 5856 }, { "epoch": 0.71, "learning_rate": 4.037076680826455e-06, "loss": 0.5494, "step": 5857 }, { "epoch": 0.71, "learning_rate": 4.033915197519069e-06, "loss": 0.5346, "step": 5858 }, { "epoch": 0.71, "learning_rate": 4.030754639802148e-06, "loss": 0.461, "step": 5859 }, { "epoch": 0.71, "learning_rate": 4.027595008166022e-06, "loss": 0.509, "step": 5860 }, { "epoch": 0.71, "learning_rate": 4.024436303100889e-06, "loss": 0.4927, "step": 5861 }, { "epoch": 0.71, "learning_rate": 4.0212785250967945e-06, "loss": 0.5876, "step": 5862 }, { "epoch": 0.71, "learning_rate": 4.018121674643636e-06, "loss": 0.5417, "step": 5863 }, { "epoch": 0.71, "learning_rate": 4.014965752231187e-06, "loss": 0.5154, "step": 5864 }, { "epoch": 0.71, "learning_rate": 4.011810758349056e-06, "loss": 0.4867, "step": 5865 }, { "epoch": 0.71, "learning_rate": 4.0086566934867125e-06, "loss": 0.6008, "step": 5866 }, { "epoch": 0.71, "learning_rate": 4.005503558133491e-06, "loss": 0.5741, "step": 5867 }, { "epoch": 0.71, "learning_rate": 4.002351352778573e-06, "loss": 0.5275, "step": 5868 }, { "epoch": 0.71, "learning_rate": 3.999200077910994e-06, "loss": 0.4761, "step": 5869 }, { "epoch": 0.71, "learning_rate": 3.996049734019654e-06, "loss": 0.5971, "step": 5870 }, { "epoch": 0.71, "learning_rate": 3.9929003215933035e-06, "loss": 0.5232, "step": 5871 }, { "epoch": 0.71, "learning_rate": 3.989751841120544e-06, "loss": 0.5326, "step": 5872 }, { "epoch": 0.71, "learning_rate": 3.986604293089843e-06, "loss": 0.5403, "step": 5873 }, { "epoch": 0.71, "learning_rate": 3.983457677989512e-06, "loss": 0.5518, "step": 5874 }, { "epoch": 0.71, "learning_rate": 3.980311996307725e-06, "loss": 0.5035, "step": 5875 }, { "epoch": 0.71, "learning_rate": 3.977167248532516e-06, "loss": 0.507, "step": 5876 }, { "epoch": 0.71, "learning_rate": 3.97402343515176e-06, "loss": 0.475, "step": 5877 }, { "epoch": 0.71, "learning_rate": 3.9708805566531934e-06, "loss": 0.5562, "step": 5878 }, { "epoch": 0.71, "learning_rate": 3.967738613524417e-06, "loss": 0.4688, "step": 5879 }, { "epoch": 0.72, "learning_rate": 3.964597606252868e-06, "loss": 0.5016, "step": 5880 }, { "epoch": 0.72, "learning_rate": 3.961457535325858e-06, "loss": 0.5348, "step": 5881 }, { "epoch": 0.72, "learning_rate": 3.9583184012305395e-06, "loss": 0.5553, "step": 5882 }, { "epoch": 0.72, "learning_rate": 3.955180204453921e-06, "loss": 0.4996, "step": 5883 }, { "epoch": 0.72, "learning_rate": 3.9520429454828755e-06, "loss": 0.5122, "step": 5884 }, { "epoch": 0.72, "learning_rate": 3.94890662480412e-06, "loss": 0.5299, "step": 5885 }, { "epoch": 0.72, "learning_rate": 3.945771242904226e-06, "loss": 0.5269, "step": 5886 }, { "epoch": 0.72, "learning_rate": 3.942636800269628e-06, "loss": 0.503, "step": 5887 }, { "epoch": 0.72, "learning_rate": 3.939503297386611e-06, "loss": 0.5347, "step": 5888 }, { "epoch": 0.72, "learning_rate": 3.936370734741307e-06, "loss": 0.522, "step": 5889 }, { "epoch": 0.72, "learning_rate": 3.9332391128197164e-06, "loss": 0.5403, "step": 5890 }, { "epoch": 0.72, "learning_rate": 3.93010843210768e-06, "loss": 0.5164, "step": 5891 }, { "epoch": 0.72, "learning_rate": 3.926978693090895e-06, "loss": 0.534, "step": 5892 }, { "epoch": 0.72, "learning_rate": 3.923849896254922e-06, "loss": 0.5049, "step": 5893 }, { "epoch": 0.72, "learning_rate": 3.920722042085166e-06, "loss": 0.5879, "step": 5894 }, { "epoch": 0.72, "learning_rate": 3.917595131066887e-06, "loss": 0.489, "step": 5895 }, { "epoch": 0.72, "learning_rate": 3.9144691636852035e-06, "loss": 0.5548, "step": 5896 }, { "epoch": 0.72, "learning_rate": 3.91134414042508e-06, "loss": 0.5334, "step": 5897 }, { "epoch": 0.72, "learning_rate": 3.908220061771344e-06, "loss": 0.4921, "step": 5898 }, { "epoch": 0.72, "learning_rate": 3.905096928208666e-06, "loss": 0.5796, "step": 5899 }, { "epoch": 0.72, "learning_rate": 3.9019747402215825e-06, "loss": 0.5203, "step": 5900 }, { "epoch": 0.72, "learning_rate": 3.898853498294468e-06, "loss": 0.5289, "step": 5901 }, { "epoch": 0.72, "learning_rate": 3.895733202911566e-06, "loss": 0.5295, "step": 5902 }, { "epoch": 0.72, "learning_rate": 3.892613854556957e-06, "loss": 0.5284, "step": 5903 }, { "epoch": 0.72, "learning_rate": 3.889495453714593e-06, "loss": 0.4932, "step": 5904 }, { "epoch": 0.72, "learning_rate": 3.886378000868262e-06, "loss": 0.4717, "step": 5905 }, { "epoch": 0.72, "learning_rate": 3.883261496501609e-06, "loss": 0.5443, "step": 5906 }, { "epoch": 0.72, "learning_rate": 3.880145941098145e-06, "loss": 0.5115, "step": 5907 }, { "epoch": 0.72, "learning_rate": 3.877031335141217e-06, "loss": 0.5099, "step": 5908 }, { "epoch": 0.72, "learning_rate": 3.873917679114029e-06, "loss": 0.4818, "step": 5909 }, { "epoch": 0.72, "learning_rate": 3.870804973499646e-06, "loss": 0.5473, "step": 5910 }, { "epoch": 0.72, "learning_rate": 3.867693218780974e-06, "loss": 0.5164, "step": 5911 }, { "epoch": 0.72, "learning_rate": 3.864582415440779e-06, "loss": 0.5596, "step": 5912 }, { "epoch": 0.72, "learning_rate": 3.8614725639616804e-06, "loss": 0.5056, "step": 5913 }, { "epoch": 0.72, "learning_rate": 3.858363664826145e-06, "loss": 0.5331, "step": 5914 }, { "epoch": 0.72, "learning_rate": 3.855255718516488e-06, "loss": 0.5079, "step": 5915 }, { "epoch": 0.72, "learning_rate": 3.852148725514892e-06, "loss": 0.5636, "step": 5916 }, { "epoch": 0.72, "learning_rate": 3.849042686303379e-06, "loss": 0.5374, "step": 5917 }, { "epoch": 0.72, "learning_rate": 3.8459376013638195e-06, "loss": 0.4787, "step": 5918 }, { "epoch": 0.72, "learning_rate": 3.842833471177952e-06, "loss": 0.5449, "step": 5919 }, { "epoch": 0.72, "learning_rate": 3.83973029622735e-06, "loss": 0.5216, "step": 5920 }, { "epoch": 0.72, "learning_rate": 3.836628076993452e-06, "loss": 0.5495, "step": 5921 }, { "epoch": 0.72, "learning_rate": 3.833526813957542e-06, "loss": 0.5616, "step": 5922 }, { "epoch": 0.72, "learning_rate": 3.830426507600751e-06, "loss": 0.5509, "step": 5923 }, { "epoch": 0.72, "learning_rate": 3.8273271584040705e-06, "loss": 0.5054, "step": 5924 }, { "epoch": 0.72, "learning_rate": 3.8242287668483426e-06, "loss": 0.5939, "step": 5925 }, { "epoch": 0.72, "learning_rate": 3.821131333414256e-06, "loss": 0.4936, "step": 5926 }, { "epoch": 0.72, "learning_rate": 3.818034858582347e-06, "loss": 0.5662, "step": 5927 }, { "epoch": 0.72, "learning_rate": 3.814939342833017e-06, "loss": 0.5047, "step": 5928 }, { "epoch": 0.72, "learning_rate": 3.811844786646505e-06, "loss": 0.5348, "step": 5929 }, { "epoch": 0.72, "learning_rate": 3.8087511905029127e-06, "loss": 0.5384, "step": 5930 }, { "epoch": 0.72, "learning_rate": 3.8056585548821836e-06, "loss": 0.5558, "step": 5931 }, { "epoch": 0.72, "learning_rate": 3.802566880264111e-06, "loss": 0.4799, "step": 5932 }, { "epoch": 0.72, "learning_rate": 3.7994761671283507e-06, "loss": 0.5558, "step": 5933 }, { "epoch": 0.72, "learning_rate": 3.7963864159544004e-06, "loss": 0.5148, "step": 5934 }, { "epoch": 0.72, "learning_rate": 3.7932976272216017e-06, "loss": 0.535, "step": 5935 }, { "epoch": 0.72, "learning_rate": 3.7902098014091703e-06, "loss": 0.4994, "step": 5936 }, { "epoch": 0.72, "learning_rate": 3.7871229389961494e-06, "loss": 0.5725, "step": 5937 }, { "epoch": 0.72, "learning_rate": 3.78403704046144e-06, "loss": 0.5648, "step": 5938 }, { "epoch": 0.72, "learning_rate": 3.7809521062837997e-06, "loss": 0.5012, "step": 5939 }, { "epoch": 0.72, "learning_rate": 3.777868136941828e-06, "loss": 0.5518, "step": 5940 }, { "epoch": 0.72, "learning_rate": 3.774785132913975e-06, "loss": 0.4642, "step": 5941 }, { "epoch": 0.72, "learning_rate": 3.7717030946785516e-06, "loss": 0.5244, "step": 5942 }, { "epoch": 0.72, "learning_rate": 3.7686220227137047e-06, "loss": 0.5015, "step": 5943 }, { "epoch": 0.72, "learning_rate": 3.765541917497444e-06, "loss": 0.4875, "step": 5944 }, { "epoch": 0.72, "learning_rate": 3.76246277950762e-06, "loss": 0.5407, "step": 5945 }, { "epoch": 0.72, "learning_rate": 3.759384609221933e-06, "loss": 0.5167, "step": 5946 }, { "epoch": 0.72, "learning_rate": 3.7563074071179407e-06, "loss": 0.5215, "step": 5947 }, { "epoch": 0.72, "learning_rate": 3.7532311736730496e-06, "loss": 0.467, "step": 5948 }, { "epoch": 0.72, "learning_rate": 3.750155909364509e-06, "loss": 0.4631, "step": 5949 }, { "epoch": 0.72, "learning_rate": 3.747081614669418e-06, "loss": 0.5674, "step": 5950 }, { "epoch": 0.72, "learning_rate": 3.7440082900647356e-06, "loss": 0.5126, "step": 5951 }, { "epoch": 0.72, "learning_rate": 3.740935936027258e-06, "loss": 0.5012, "step": 5952 }, { "epoch": 0.72, "learning_rate": 3.7378645530336423e-06, "loss": 0.4863, "step": 5953 }, { "epoch": 0.72, "learning_rate": 3.734794141560386e-06, "loss": 0.5685, "step": 5954 }, { "epoch": 0.72, "learning_rate": 3.7317247020838353e-06, "loss": 0.533, "step": 5955 }, { "epoch": 0.72, "learning_rate": 3.728656235080196e-06, "loss": 0.5282, "step": 5956 }, { "epoch": 0.72, "learning_rate": 3.725588741025513e-06, "loss": 0.4986, "step": 5957 }, { "epoch": 0.72, "learning_rate": 3.72252222039568e-06, "loss": 0.5207, "step": 5958 }, { "epoch": 0.72, "learning_rate": 3.719456673666447e-06, "loss": 0.4958, "step": 5959 }, { "epoch": 0.72, "learning_rate": 3.716392101313412e-06, "loss": 0.4874, "step": 5960 }, { "epoch": 0.72, "learning_rate": 3.7133285038120126e-06, "loss": 0.5162, "step": 5961 }, { "epoch": 0.73, "learning_rate": 3.7102658816375482e-06, "loss": 0.5377, "step": 5962 }, { "epoch": 0.73, "learning_rate": 3.707204235265156e-06, "loss": 0.538, "step": 5963 }, { "epoch": 0.73, "learning_rate": 3.7041435651698244e-06, "loss": 0.4717, "step": 5964 }, { "epoch": 0.73, "learning_rate": 3.7010838718263976e-06, "loss": 0.4927, "step": 5965 }, { "epoch": 0.73, "learning_rate": 3.6980251557095594e-06, "loss": 0.5082, "step": 5966 }, { "epoch": 0.73, "learning_rate": 3.694967417293841e-06, "loss": 0.5119, "step": 5967 }, { "epoch": 0.73, "learning_rate": 3.6919106570536335e-06, "loss": 0.5378, "step": 5968 }, { "epoch": 0.73, "learning_rate": 3.6888548754631625e-06, "loss": 0.5, "step": 5969 }, { "epoch": 0.73, "learning_rate": 3.685800072996515e-06, "loss": 0.4703, "step": 5970 }, { "epoch": 0.73, "learning_rate": 3.6827462501276125e-06, "loss": 0.5735, "step": 5971 }, { "epoch": 0.73, "learning_rate": 3.679693407330236e-06, "loss": 0.4896, "step": 5972 }, { "epoch": 0.73, "learning_rate": 3.6766415450780057e-06, "loss": 0.5451, "step": 5973 }, { "epoch": 0.73, "learning_rate": 3.6735906638443995e-06, "loss": 0.5112, "step": 5974 }, { "epoch": 0.73, "learning_rate": 3.670540764102729e-06, "loss": 0.461, "step": 5975 }, { "epoch": 0.73, "learning_rate": 3.667491846326171e-06, "loss": 0.5098, "step": 5976 }, { "epoch": 0.73, "learning_rate": 3.664443910987736e-06, "loss": 0.5497, "step": 5977 }, { "epoch": 0.73, "learning_rate": 3.6613969585602847e-06, "loss": 0.5128, "step": 5978 }, { "epoch": 0.73, "learning_rate": 3.6583509895165325e-06, "loss": 0.5375, "step": 5979 }, { "epoch": 0.73, "learning_rate": 3.6553060043290343e-06, "loss": 0.5549, "step": 5980 }, { "epoch": 0.73, "learning_rate": 3.6522620034701927e-06, "loss": 0.5671, "step": 5981 }, { "epoch": 0.73, "learning_rate": 3.6492189874122663e-06, "loss": 0.4906, "step": 5982 }, { "epoch": 0.73, "learning_rate": 3.6461769566273475e-06, "loss": 0.5383, "step": 5983 }, { "epoch": 0.73, "learning_rate": 3.6431359115873875e-06, "loss": 0.4737, "step": 5984 }, { "epoch": 0.73, "learning_rate": 3.640095852764183e-06, "loss": 0.4743, "step": 5985 }, { "epoch": 0.73, "learning_rate": 3.6370567806293724e-06, "loss": 0.5233, "step": 5986 }, { "epoch": 0.73, "learning_rate": 3.634018695654439e-06, "loss": 0.4647, "step": 5987 }, { "epoch": 0.73, "learning_rate": 3.6309815983107234e-06, "loss": 0.5136, "step": 5988 }, { "epoch": 0.73, "learning_rate": 3.6279454890694065e-06, "loss": 0.5547, "step": 5989 }, { "epoch": 0.73, "learning_rate": 3.62491036840151e-06, "loss": 0.5379, "step": 5990 }, { "epoch": 0.73, "learning_rate": 3.6218762367779182e-06, "loss": 0.4417, "step": 5991 }, { "epoch": 0.73, "learning_rate": 3.618843094669343e-06, "loss": 0.4747, "step": 5992 }, { "epoch": 0.73, "learning_rate": 3.6158109425463607e-06, "loss": 0.5555, "step": 5993 }, { "epoch": 0.73, "learning_rate": 3.6127797808793806e-06, "loss": 0.5062, "step": 5994 }, { "epoch": 0.73, "learning_rate": 3.609749610138662e-06, "loss": 0.5416, "step": 5995 }, { "epoch": 0.73, "learning_rate": 3.606720430794315e-06, "loss": 0.4993, "step": 5996 }, { "epoch": 0.73, "learning_rate": 3.6036922433162934e-06, "loss": 0.496, "step": 5997 }, { "epoch": 0.73, "learning_rate": 3.600665048174391e-06, "loss": 0.5217, "step": 5998 }, { "epoch": 0.73, "learning_rate": 3.5976388458382605e-06, "loss": 0.5257, "step": 5999 }, { "epoch": 0.73, "learning_rate": 3.5946136367773897e-06, "loss": 0.5352, "step": 6000 }, { "epoch": 0.73, "learning_rate": 3.5915894214611113e-06, "loss": 0.4934, "step": 6001 }, { "epoch": 0.73, "learning_rate": 3.5885662003586143e-06, "loss": 0.4916, "step": 6002 }, { "epoch": 0.73, "learning_rate": 3.5855439739389263e-06, "loss": 0.4886, "step": 6003 }, { "epoch": 0.73, "learning_rate": 3.5825227426709154e-06, "loss": 0.5239, "step": 6004 }, { "epoch": 0.73, "learning_rate": 3.5795025070233113e-06, "loss": 0.4997, "step": 6005 }, { "epoch": 0.73, "learning_rate": 3.5764832674646733e-06, "loss": 0.4864, "step": 6006 }, { "epoch": 0.73, "learning_rate": 3.5734650244634074e-06, "loss": 0.5043, "step": 6007 }, { "epoch": 0.73, "learning_rate": 3.5704477784877834e-06, "loss": 0.4809, "step": 6008 }, { "epoch": 0.73, "learning_rate": 3.5674315300058947e-06, "loss": 0.4897, "step": 6009 }, { "epoch": 0.73, "learning_rate": 3.564416279485686e-06, "loss": 0.559, "step": 6010 }, { "epoch": 0.73, "learning_rate": 3.561402027394957e-06, "loss": 0.5475, "step": 6011 }, { "epoch": 0.73, "learning_rate": 3.558388774201339e-06, "loss": 0.5426, "step": 6012 }, { "epoch": 0.73, "learning_rate": 3.5553765203723123e-06, "loss": 0.5235, "step": 6013 }, { "epoch": 0.73, "learning_rate": 3.5523652663752107e-06, "loss": 0.4675, "step": 6014 }, { "epoch": 0.73, "learning_rate": 3.5493550126772004e-06, "loss": 0.4772, "step": 6015 }, { "epoch": 0.73, "learning_rate": 3.5463457597453033e-06, "loss": 0.5209, "step": 6016 }, { "epoch": 0.73, "learning_rate": 3.543337508046378e-06, "loss": 0.6009, "step": 6017 }, { "epoch": 0.73, "learning_rate": 3.540330258047129e-06, "loss": 0.5477, "step": 6018 }, { "epoch": 0.73, "learning_rate": 3.5373240102141084e-06, "loss": 0.5447, "step": 6019 }, { "epoch": 0.73, "learning_rate": 3.534318765013717e-06, "loss": 0.472, "step": 6020 }, { "epoch": 0.73, "learning_rate": 3.531314522912185e-06, "loss": 0.5396, "step": 6021 }, { "epoch": 0.73, "learning_rate": 3.528311284375606e-06, "loss": 0.5193, "step": 6022 }, { "epoch": 0.73, "learning_rate": 3.5253090498699037e-06, "loss": 0.4654, "step": 6023 }, { "epoch": 0.73, "learning_rate": 3.5223078198608483e-06, "loss": 0.5196, "step": 6024 }, { "epoch": 0.73, "learning_rate": 3.519307594814063e-06, "loss": 0.5948, "step": 6025 }, { "epoch": 0.73, "learning_rate": 3.5163083751950057e-06, "loss": 0.5193, "step": 6026 }, { "epoch": 0.73, "learning_rate": 3.513310161468978e-06, "loss": 0.4488, "step": 6027 }, { "epoch": 0.73, "learning_rate": 3.5103129541011347e-06, "loss": 0.4976, "step": 6028 }, { "epoch": 0.73, "learning_rate": 3.5073167535564667e-06, "loss": 0.486, "step": 6029 }, { "epoch": 0.73, "learning_rate": 3.504321560299806e-06, "loss": 0.4984, "step": 6030 }, { "epoch": 0.73, "learning_rate": 3.5013273747958364e-06, "loss": 0.5514, "step": 6031 }, { "epoch": 0.73, "learning_rate": 3.4983341975090866e-06, "loss": 0.5106, "step": 6032 }, { "epoch": 0.73, "learning_rate": 3.4953420289039165e-06, "loss": 0.5403, "step": 6033 }, { "epoch": 0.73, "learning_rate": 3.492350869444543e-06, "loss": 0.5137, "step": 6034 }, { "epoch": 0.73, "learning_rate": 3.4893607195950185e-06, "loss": 0.505, "step": 6035 }, { "epoch": 0.73, "learning_rate": 3.4863715798192377e-06, "loss": 0.4599, "step": 6036 }, { "epoch": 0.73, "learning_rate": 3.483383450580947e-06, "loss": 0.5218, "step": 6037 }, { "epoch": 0.73, "learning_rate": 3.4803963323437264e-06, "loss": 0.5156, "step": 6038 }, { "epoch": 0.73, "learning_rate": 3.4774102255710083e-06, "loss": 0.5273, "step": 6039 }, { "epoch": 0.73, "learning_rate": 3.47442513072606e-06, "loss": 0.5136, "step": 6040 }, { "epoch": 0.73, "learning_rate": 3.471441048271994e-06, "loss": 0.4826, "step": 6041 }, { "epoch": 0.73, "learning_rate": 3.468457978671771e-06, "loss": 0.5512, "step": 6042 }, { "epoch": 0.73, "learning_rate": 3.465475922388184e-06, "loss": 0.5581, "step": 6043 }, { "epoch": 0.73, "learning_rate": 3.462494879883881e-06, "loss": 0.5286, "step": 6044 }, { "epoch": 0.74, "learning_rate": 3.4595148516213472e-06, "loss": 0.4881, "step": 6045 }, { "epoch": 0.74, "learning_rate": 3.4565358380629088e-06, "loss": 0.5741, "step": 6046 }, { "epoch": 0.74, "learning_rate": 3.4535578396707327e-06, "loss": 0.541, "step": 6047 }, { "epoch": 0.74, "learning_rate": 3.4505808569068365e-06, "loss": 0.5411, "step": 6048 }, { "epoch": 0.74, "learning_rate": 3.447604890233074e-06, "loss": 0.4682, "step": 6049 }, { "epoch": 0.74, "learning_rate": 3.444629940111138e-06, "loss": 0.4821, "step": 6050 }, { "epoch": 0.74, "learning_rate": 3.441656007002576e-06, "loss": 0.5367, "step": 6051 }, { "epoch": 0.74, "learning_rate": 3.4386830913687676e-06, "loss": 0.5247, "step": 6052 }, { "epoch": 0.74, "learning_rate": 3.43571119367093e-06, "loss": 0.4944, "step": 6053 }, { "epoch": 0.74, "learning_rate": 3.4327403143701387e-06, "loss": 0.5186, "step": 6054 }, { "epoch": 0.74, "learning_rate": 3.4297704539272957e-06, "loss": 0.5107, "step": 6055 }, { "epoch": 0.74, "learning_rate": 3.4268016128031535e-06, "loss": 0.476, "step": 6056 }, { "epoch": 0.74, "learning_rate": 3.4238337914583062e-06, "loss": 0.56, "step": 6057 }, { "epoch": 0.74, "learning_rate": 3.420866990353188e-06, "loss": 0.4594, "step": 6058 }, { "epoch": 0.74, "learning_rate": 3.417901209948067e-06, "loss": 0.5212, "step": 6059 }, { "epoch": 0.74, "learning_rate": 3.414936450703069e-06, "loss": 0.4806, "step": 6060 }, { "epoch": 0.74, "learning_rate": 3.4119727130781457e-06, "loss": 0.5031, "step": 6061 }, { "epoch": 0.74, "learning_rate": 3.409009997533105e-06, "loss": 0.5094, "step": 6062 }, { "epoch": 0.74, "learning_rate": 3.4060483045275838e-06, "loss": 0.5519, "step": 6063 }, { "epoch": 0.74, "learning_rate": 3.403087634521063e-06, "loss": 0.5587, "step": 6064 }, { "epoch": 0.74, "learning_rate": 3.4001279879728722e-06, "loss": 0.5321, "step": 6065 }, { "epoch": 0.74, "learning_rate": 3.3971693653421746e-06, "loss": 0.4989, "step": 6066 }, { "epoch": 0.74, "learning_rate": 3.3942117670879704e-06, "loss": 0.5039, "step": 6067 }, { "epoch": 0.74, "learning_rate": 3.391255193669121e-06, "loss": 0.5473, "step": 6068 }, { "epoch": 0.74, "learning_rate": 3.388299645544306e-06, "loss": 0.5613, "step": 6069 }, { "epoch": 0.74, "learning_rate": 3.3853451231720556e-06, "loss": 0.5146, "step": 6070 }, { "epoch": 0.74, "learning_rate": 3.382391627010745e-06, "loss": 0.5065, "step": 6071 }, { "epoch": 0.74, "learning_rate": 3.3794391575185814e-06, "loss": 0.5196, "step": 6072 }, { "epoch": 0.74, "learning_rate": 3.3764877151536147e-06, "loss": 0.5106, "step": 6073 }, { "epoch": 0.74, "learning_rate": 3.373537300373745e-06, "loss": 0.5309, "step": 6074 }, { "epoch": 0.74, "learning_rate": 3.3705879136367003e-06, "loss": 0.4947, "step": 6075 }, { "epoch": 0.74, "learning_rate": 3.367639555400053e-06, "loss": 0.4973, "step": 6076 }, { "epoch": 0.74, "learning_rate": 3.3646922261212224e-06, "loss": 0.5301, "step": 6077 }, { "epoch": 0.74, "learning_rate": 3.361745926257458e-06, "loss": 0.5367, "step": 6078 }, { "epoch": 0.74, "learning_rate": 3.358800656265857e-06, "loss": 0.5171, "step": 6079 }, { "epoch": 0.74, "learning_rate": 3.355856416603358e-06, "loss": 0.534, "step": 6080 }, { "epoch": 0.74, "learning_rate": 3.3529132077267334e-06, "loss": 0.5002, "step": 6081 }, { "epoch": 0.74, "learning_rate": 3.349971030092596e-06, "loss": 0.4711, "step": 6082 }, { "epoch": 0.74, "learning_rate": 3.3470298841574057e-06, "loss": 0.502, "step": 6083 }, { "epoch": 0.74, "learning_rate": 3.344089770377452e-06, "loss": 0.539, "step": 6084 }, { "epoch": 0.74, "learning_rate": 3.3411506892088784e-06, "loss": 0.502, "step": 6085 }, { "epoch": 0.74, "learning_rate": 3.338212641107654e-06, "loss": 0.4835, "step": 6086 }, { "epoch": 0.74, "learning_rate": 3.3352756265295925e-06, "loss": 0.514, "step": 6087 }, { "epoch": 0.74, "learning_rate": 3.3323396459303534e-06, "loss": 0.4844, "step": 6088 }, { "epoch": 0.74, "learning_rate": 3.329404699765428e-06, "loss": 0.4785, "step": 6089 }, { "epoch": 0.74, "learning_rate": 3.326470788490146e-06, "loss": 0.5389, "step": 6090 }, { "epoch": 0.74, "learning_rate": 3.3235379125596846e-06, "loss": 0.5292, "step": 6091 }, { "epoch": 0.74, "learning_rate": 3.320606072429059e-06, "loss": 0.582, "step": 6092 }, { "epoch": 0.74, "learning_rate": 3.3176752685531134e-06, "loss": 0.5425, "step": 6093 }, { "epoch": 0.74, "learning_rate": 3.3147455013865447e-06, "loss": 0.5262, "step": 6094 }, { "epoch": 0.74, "learning_rate": 3.3118167713838823e-06, "loss": 0.5134, "step": 6095 }, { "epoch": 0.74, "learning_rate": 3.3088890789994897e-06, "loss": 0.4569, "step": 6096 }, { "epoch": 0.74, "learning_rate": 3.305962424687583e-06, "loss": 0.4955, "step": 6097 }, { "epoch": 0.74, "learning_rate": 3.3030368089022045e-06, "loss": 0.5172, "step": 6098 }, { "epoch": 0.74, "learning_rate": 3.300112232097238e-06, "loss": 0.5341, "step": 6099 }, { "epoch": 0.74, "learning_rate": 3.2971886947264144e-06, "loss": 0.5106, "step": 6100 }, { "epoch": 0.74, "learning_rate": 3.294266197243291e-06, "loss": 0.5278, "step": 6101 }, { "epoch": 0.74, "learning_rate": 3.2913447401012766e-06, "loss": 0.6003, "step": 6102 }, { "epoch": 0.74, "learning_rate": 3.288424323753604e-06, "loss": 0.5147, "step": 6103 }, { "epoch": 0.74, "learning_rate": 3.285504948653361e-06, "loss": 0.5512, "step": 6104 }, { "epoch": 0.74, "learning_rate": 3.2825866152534568e-06, "loss": 0.4653, "step": 6105 }, { "epoch": 0.74, "learning_rate": 3.2796693240066566e-06, "loss": 0.5073, "step": 6106 }, { "epoch": 0.74, "learning_rate": 3.2767530753655462e-06, "loss": 0.4807, "step": 6107 }, { "epoch": 0.74, "learning_rate": 3.2738378697825656e-06, "loss": 0.5, "step": 6108 }, { "epoch": 0.74, "learning_rate": 3.270923707709983e-06, "loss": 0.4702, "step": 6109 }, { "epoch": 0.74, "learning_rate": 3.2680105895999035e-06, "loss": 0.5062, "step": 6110 }, { "epoch": 0.74, "learning_rate": 3.26509851590428e-06, "loss": 0.4753, "step": 6111 }, { "epoch": 0.74, "learning_rate": 3.262187487074896e-06, "loss": 0.497, "step": 6112 }, { "epoch": 0.74, "learning_rate": 3.2592775035633693e-06, "loss": 0.5307, "step": 6113 }, { "epoch": 0.74, "learning_rate": 3.2563685658211676e-06, "loss": 0.5034, "step": 6114 }, { "epoch": 0.74, "learning_rate": 3.2534606742995843e-06, "loss": 0.4904, "step": 6115 }, { "epoch": 0.74, "learning_rate": 3.2505538294497575e-06, "loss": 0.4891, "step": 6116 }, { "epoch": 0.74, "learning_rate": 3.2476480317226634e-06, "loss": 0.4757, "step": 6117 }, { "epoch": 0.74, "learning_rate": 3.2447432815691128e-06, "loss": 0.5521, "step": 6118 }, { "epoch": 0.74, "learning_rate": 3.2418395794397483e-06, "loss": 0.5363, "step": 6119 }, { "epoch": 0.74, "learning_rate": 3.2389369257850645e-06, "loss": 0.5065, "step": 6120 }, { "epoch": 0.74, "learning_rate": 3.236035321055381e-06, "loss": 0.5223, "step": 6121 }, { "epoch": 0.74, "learning_rate": 3.233134765700856e-06, "loss": 0.5213, "step": 6122 }, { "epoch": 0.74, "learning_rate": 3.2302352601714924e-06, "loss": 0.5152, "step": 6123 }, { "epoch": 0.74, "learning_rate": 3.2273368049171216e-06, "loss": 0.4549, "step": 6124 }, { "epoch": 0.74, "learning_rate": 3.2244394003874204e-06, "loss": 0.5056, "step": 6125 }, { "epoch": 0.74, "learning_rate": 3.221543047031894e-06, "loss": 0.5465, "step": 6126 }, { "epoch": 0.75, "learning_rate": 3.2186477452998876e-06, "loss": 0.5626, "step": 6127 }, { "epoch": 0.75, "learning_rate": 3.2157534956405858e-06, "loss": 0.4989, "step": 6128 }, { "epoch": 0.75, "learning_rate": 3.212860298503011e-06, "loss": 0.5183, "step": 6129 }, { "epoch": 0.75, "learning_rate": 3.2099681543360185e-06, "loss": 0.4894, "step": 6130 }, { "epoch": 0.75, "learning_rate": 3.2070770635882962e-06, "loss": 0.524, "step": 6131 }, { "epoch": 0.75, "learning_rate": 3.204187026708382e-06, "loss": 0.515, "step": 6132 }, { "epoch": 0.75, "learning_rate": 3.201298044144634e-06, "loss": 0.5324, "step": 6133 }, { "epoch": 0.75, "learning_rate": 3.198410116345262e-06, "loss": 0.5124, "step": 6134 }, { "epoch": 0.75, "learning_rate": 3.195523243758302e-06, "loss": 0.4661, "step": 6135 }, { "epoch": 0.75, "learning_rate": 3.192637426831625e-06, "loss": 0.5695, "step": 6136 }, { "epoch": 0.75, "learning_rate": 3.1897526660129507e-06, "loss": 0.518, "step": 6137 }, { "epoch": 0.75, "learning_rate": 3.186868961749822e-06, "loss": 0.5996, "step": 6138 }, { "epoch": 0.75, "learning_rate": 3.1839863144896156e-06, "loss": 0.5032, "step": 6139 }, { "epoch": 0.75, "learning_rate": 3.1811047246795658e-06, "loss": 0.471, "step": 6140 }, { "epoch": 0.75, "learning_rate": 3.178224192766721e-06, "loss": 0.5258, "step": 6141 }, { "epoch": 0.75, "learning_rate": 3.1753447191979703e-06, "loss": 0.5241, "step": 6142 }, { "epoch": 0.75, "learning_rate": 3.1724663044200465e-06, "loss": 0.4813, "step": 6143 }, { "epoch": 0.75, "learning_rate": 3.1695889488795097e-06, "loss": 0.4938, "step": 6144 }, { "epoch": 0.75, "learning_rate": 3.1667126530227553e-06, "loss": 0.497, "step": 6145 }, { "epoch": 0.75, "learning_rate": 3.163837417296024e-06, "loss": 0.532, "step": 6146 }, { "epoch": 0.75, "learning_rate": 3.160963242145384e-06, "loss": 0.5204, "step": 6147 }, { "epoch": 0.75, "learning_rate": 3.1580901280167353e-06, "loss": 0.5564, "step": 6148 }, { "epoch": 0.75, "learning_rate": 3.1552180753558257e-06, "loss": 0.5188, "step": 6149 }, { "epoch": 0.75, "learning_rate": 3.152347084608226e-06, "loss": 0.4656, "step": 6150 }, { "epoch": 0.75, "learning_rate": 3.1494771562193504e-06, "loss": 0.4524, "step": 6151 }, { "epoch": 0.75, "learning_rate": 3.1466082906344475e-06, "loss": 0.4963, "step": 6152 }, { "epoch": 0.75, "learning_rate": 3.1437404882985956e-06, "loss": 0.5124, "step": 6153 }, { "epoch": 0.75, "learning_rate": 3.1408737496567087e-06, "loss": 0.5017, "step": 6154 }, { "epoch": 0.75, "learning_rate": 3.138008075153546e-06, "loss": 0.5307, "step": 6155 }, { "epoch": 0.75, "learning_rate": 3.1351434652336865e-06, "loss": 0.6043, "step": 6156 }, { "epoch": 0.75, "learning_rate": 3.1322799203415576e-06, "loss": 0.5141, "step": 6157 }, { "epoch": 0.75, "learning_rate": 3.129417440921413e-06, "loss": 0.4764, "step": 6158 }, { "epoch": 0.75, "learning_rate": 3.12655602741734e-06, "loss": 0.4927, "step": 6159 }, { "epoch": 0.75, "learning_rate": 3.12369568027327e-06, "loss": 0.547, "step": 6160 }, { "epoch": 0.75, "learning_rate": 3.1208363999329617e-06, "loss": 0.5409, "step": 6161 }, { "epoch": 0.75, "learning_rate": 3.1179781868400037e-06, "loss": 0.5329, "step": 6162 }, { "epoch": 0.75, "learning_rate": 3.1151210414378297e-06, "loss": 0.5088, "step": 6163 }, { "epoch": 0.75, "learning_rate": 3.1122649641697067e-06, "loss": 0.5696, "step": 6164 }, { "epoch": 0.75, "learning_rate": 3.109409955478726e-06, "loss": 0.4931, "step": 6165 }, { "epoch": 0.75, "learning_rate": 3.106556015807823e-06, "loss": 0.5109, "step": 6166 }, { "epoch": 0.75, "learning_rate": 3.1037031455997644e-06, "loss": 0.5318, "step": 6167 }, { "epoch": 0.75, "learning_rate": 3.1008513452971458e-06, "loss": 0.5069, "step": 6168 }, { "epoch": 0.75, "learning_rate": 3.098000615342406e-06, "loss": 0.4882, "step": 6169 }, { "epoch": 0.75, "learning_rate": 3.0951509561778104e-06, "loss": 0.4888, "step": 6170 }, { "epoch": 0.75, "learning_rate": 3.092302368245459e-06, "loss": 0.5465, "step": 6171 }, { "epoch": 0.75, "learning_rate": 3.0894548519872923e-06, "loss": 0.4904, "step": 6172 }, { "epoch": 0.75, "learning_rate": 3.0866084078450754e-06, "loss": 0.5188, "step": 6173 }, { "epoch": 0.75, "learning_rate": 3.0837630362604154e-06, "loss": 0.5719, "step": 6174 }, { "epoch": 0.75, "learning_rate": 3.0809187376747453e-06, "loss": 0.5345, "step": 6175 }, { "epoch": 0.75, "learning_rate": 3.0780755125293383e-06, "loss": 0.5238, "step": 6176 }, { "epoch": 0.75, "learning_rate": 3.075233361265295e-06, "loss": 0.5057, "step": 6177 }, { "epoch": 0.75, "learning_rate": 3.0723922843235564e-06, "loss": 0.5442, "step": 6178 }, { "epoch": 0.75, "learning_rate": 3.0695522821448885e-06, "loss": 0.5169, "step": 6179 }, { "epoch": 0.75, "learning_rate": 3.0667133551699e-06, "loss": 0.5605, "step": 6180 }, { "epoch": 0.75, "learning_rate": 3.0638755038390257e-06, "loss": 0.5023, "step": 6181 }, { "epoch": 0.75, "learning_rate": 3.0610387285925303e-06, "loss": 0.5212, "step": 6182 }, { "epoch": 0.75, "learning_rate": 3.0582030298705255e-06, "loss": 0.545, "step": 6183 }, { "epoch": 0.75, "learning_rate": 3.0553684081129444e-06, "loss": 0.5353, "step": 6184 }, { "epoch": 0.75, "learning_rate": 3.052534863759551e-06, "loss": 0.5139, "step": 6185 }, { "epoch": 0.75, "learning_rate": 3.049702397249953e-06, "loss": 0.5028, "step": 6186 }, { "epoch": 0.75, "learning_rate": 3.0468710090235807e-06, "loss": 0.5008, "step": 6187 }, { "epoch": 0.75, "learning_rate": 3.0440406995197038e-06, "loss": 0.5132, "step": 6188 }, { "epoch": 0.75, "learning_rate": 3.0412114691774252e-06, "loss": 0.5143, "step": 6189 }, { "epoch": 0.75, "learning_rate": 3.0383833184356746e-06, "loss": 0.559, "step": 6190 }, { "epoch": 0.75, "learning_rate": 3.0355562477332133e-06, "loss": 0.5074, "step": 6191 }, { "epoch": 0.75, "learning_rate": 3.032730257508646e-06, "loss": 0.4796, "step": 6192 }, { "epoch": 0.75, "learning_rate": 3.029905348200398e-06, "loss": 0.5894, "step": 6193 }, { "epoch": 0.75, "learning_rate": 3.0270815202467285e-06, "loss": 0.476, "step": 6194 }, { "epoch": 0.75, "learning_rate": 3.02425877408574e-06, "loss": 0.4987, "step": 6195 }, { "epoch": 0.75, "learning_rate": 3.0214371101553506e-06, "loss": 0.5538, "step": 6196 }, { "epoch": 0.75, "learning_rate": 3.0186165288933267e-06, "loss": 0.5035, "step": 6197 }, { "epoch": 0.75, "learning_rate": 3.0157970307372554e-06, "loss": 0.505, "step": 6198 }, { "epoch": 0.75, "learning_rate": 3.0129786161245555e-06, "loss": 0.4284, "step": 6199 }, { "epoch": 0.75, "learning_rate": 3.0101612854924853e-06, "loss": 0.5276, "step": 6200 }, { "epoch": 0.75, "learning_rate": 3.0073450392781343e-06, "loss": 0.5387, "step": 6201 }, { "epoch": 0.75, "learning_rate": 3.004529877918415e-06, "loss": 0.5102, "step": 6202 }, { "epoch": 0.75, "learning_rate": 3.0017158018500823e-06, "loss": 0.5183, "step": 6203 }, { "epoch": 0.75, "learning_rate": 2.998902811509715e-06, "loss": 0.5266, "step": 6204 }, { "epoch": 0.75, "learning_rate": 2.996090907333723e-06, "loss": 0.4914, "step": 6205 }, { "epoch": 0.75, "learning_rate": 2.993280089758357e-06, "loss": 0.474, "step": 6206 }, { "epoch": 0.75, "learning_rate": 2.9904703592196904e-06, "loss": 0.4434, "step": 6207 }, { "epoch": 0.75, "learning_rate": 2.987661716153627e-06, "loss": 0.5224, "step": 6208 }, { "epoch": 0.76, "learning_rate": 2.98485416099591e-06, "loss": 0.5389, "step": 6209 }, { "epoch": 0.76, "learning_rate": 2.9820476941821087e-06, "loss": 0.4988, "step": 6210 }, { "epoch": 0.76, "learning_rate": 2.979242316147617e-06, "loss": 0.5569, "step": 6211 }, { "epoch": 0.76, "learning_rate": 2.9764380273276785e-06, "loss": 0.5264, "step": 6212 }, { "epoch": 0.76, "learning_rate": 2.9736348281573503e-06, "loss": 0.5032, "step": 6213 }, { "epoch": 0.76, "learning_rate": 2.9708327190715237e-06, "loss": 0.49, "step": 6214 }, { "epoch": 0.76, "learning_rate": 2.9680317005049286e-06, "loss": 0.5843, "step": 6215 }, { "epoch": 0.76, "learning_rate": 2.9652317728921197e-06, "loss": 0.53, "step": 6216 }, { "epoch": 0.76, "learning_rate": 2.9624329366674787e-06, "loss": 0.5273, "step": 6217 }, { "epoch": 0.76, "learning_rate": 2.95963519226523e-06, "loss": 0.548, "step": 6218 }, { "epoch": 0.76, "learning_rate": 2.9568385401194145e-06, "loss": 0.547, "step": 6219 }, { "epoch": 0.76, "learning_rate": 2.954042980663916e-06, "loss": 0.5323, "step": 6220 }, { "epoch": 0.76, "learning_rate": 2.9512485143324408e-06, "loss": 0.5196, "step": 6221 }, { "epoch": 0.76, "learning_rate": 2.948455141558526e-06, "loss": 0.5253, "step": 6222 }, { "epoch": 0.76, "learning_rate": 2.9456628627755423e-06, "loss": 0.4998, "step": 6223 }, { "epoch": 0.76, "learning_rate": 2.9428716784166942e-06, "loss": 0.4648, "step": 6224 }, { "epoch": 0.76, "learning_rate": 2.9400815889150045e-06, "loss": 0.5421, "step": 6225 }, { "epoch": 0.76, "learning_rate": 2.937292594703339e-06, "loss": 0.5225, "step": 6226 }, { "epoch": 0.76, "learning_rate": 2.9345046962143875e-06, "loss": 0.5449, "step": 6227 }, { "epoch": 0.76, "learning_rate": 2.9317178938806634e-06, "loss": 0.5644, "step": 6228 }, { "epoch": 0.76, "learning_rate": 2.9289321881345257e-06, "loss": 0.5283, "step": 6229 }, { "epoch": 0.76, "learning_rate": 2.9261475794081506e-06, "loss": 0.4552, "step": 6230 }, { "epoch": 0.76, "learning_rate": 2.923364068133544e-06, "loss": 0.4787, "step": 6231 }, { "epoch": 0.76, "learning_rate": 2.9205816547425525e-06, "loss": 0.4772, "step": 6232 }, { "epoch": 0.76, "learning_rate": 2.917800339666843e-06, "loss": 0.5243, "step": 6233 }, { "epoch": 0.76, "learning_rate": 2.9150201233379093e-06, "loss": 0.5253, "step": 6234 }, { "epoch": 0.76, "learning_rate": 2.9122410061870877e-06, "loss": 0.5561, "step": 6235 }, { "epoch": 0.76, "learning_rate": 2.9094629886455296e-06, "loss": 0.5652, "step": 6236 }, { "epoch": 0.76, "learning_rate": 2.9066860711442248e-06, "loss": 0.5505, "step": 6237 }, { "epoch": 0.76, "learning_rate": 2.903910254113992e-06, "loss": 0.5471, "step": 6238 }, { "epoch": 0.76, "learning_rate": 2.9011355379854768e-06, "loss": 0.4933, "step": 6239 }, { "epoch": 0.76, "learning_rate": 2.898361923189148e-06, "loss": 0.5634, "step": 6240 }, { "epoch": 0.76, "learning_rate": 2.8955894101553185e-06, "loss": 0.577, "step": 6241 }, { "epoch": 0.76, "learning_rate": 2.892817999314115e-06, "loss": 0.5164, "step": 6242 }, { "epoch": 0.76, "learning_rate": 2.8900476910955044e-06, "loss": 0.5181, "step": 6243 }, { "epoch": 0.76, "learning_rate": 2.8872784859292757e-06, "loss": 0.4953, "step": 6244 }, { "epoch": 0.76, "learning_rate": 2.8845103842450463e-06, "loss": 0.5204, "step": 6245 }, { "epoch": 0.76, "learning_rate": 2.881743386472271e-06, "loss": 0.4954, "step": 6246 }, { "epoch": 0.76, "learning_rate": 2.8789774930402238e-06, "loss": 0.5218, "step": 6247 }, { "epoch": 0.76, "learning_rate": 2.8762127043780063e-06, "loss": 0.5405, "step": 6248 }, { "epoch": 0.76, "learning_rate": 2.8734490209145637e-06, "loss": 0.5038, "step": 6249 }, { "epoch": 0.76, "learning_rate": 2.870686443078655e-06, "loss": 0.5233, "step": 6250 }, { "epoch": 0.76, "learning_rate": 2.867924971298869e-06, "loss": 0.5422, "step": 6251 }, { "epoch": 0.76, "learning_rate": 2.865164606003631e-06, "loss": 0.494, "step": 6252 }, { "epoch": 0.76, "learning_rate": 2.8624053476211876e-06, "loss": 0.5452, "step": 6253 }, { "epoch": 0.76, "learning_rate": 2.8596471965796113e-06, "loss": 0.4943, "step": 6254 }, { "epoch": 0.76, "learning_rate": 2.856890153306815e-06, "loss": 0.5066, "step": 6255 }, { "epoch": 0.76, "learning_rate": 2.854134218230528e-06, "loss": 0.5604, "step": 6256 }, { "epoch": 0.76, "learning_rate": 2.8513793917783083e-06, "loss": 0.51, "step": 6257 }, { "epoch": 0.76, "learning_rate": 2.848625674377553e-06, "loss": 0.4718, "step": 6258 }, { "epoch": 0.76, "learning_rate": 2.8458730664554713e-06, "loss": 0.574, "step": 6259 }, { "epoch": 0.76, "learning_rate": 2.8431215684391123e-06, "loss": 0.5189, "step": 6260 }, { "epoch": 0.76, "learning_rate": 2.8403711807553527e-06, "loss": 0.5193, "step": 6261 }, { "epoch": 0.76, "learning_rate": 2.837621903830888e-06, "loss": 0.4668, "step": 6262 }, { "epoch": 0.76, "learning_rate": 2.8348737380922453e-06, "loss": 0.5363, "step": 6263 }, { "epoch": 0.76, "learning_rate": 2.832126683965787e-06, "loss": 0.4981, "step": 6264 }, { "epoch": 0.76, "learning_rate": 2.82938074187769e-06, "loss": 0.4814, "step": 6265 }, { "epoch": 0.76, "learning_rate": 2.82663591225397e-06, "loss": 0.5083, "step": 6266 }, { "epoch": 0.76, "learning_rate": 2.8238921955204644e-06, "loss": 0.5394, "step": 6267 }, { "epoch": 0.76, "learning_rate": 2.821149592102835e-06, "loss": 0.4895, "step": 6268 }, { "epoch": 0.76, "learning_rate": 2.818408102426582e-06, "loss": 0.4957, "step": 6269 }, { "epoch": 0.76, "learning_rate": 2.8156677269170208e-06, "loss": 0.5355, "step": 6270 }, { "epoch": 0.76, "learning_rate": 2.8129284659992973e-06, "loss": 0.5288, "step": 6271 }, { "epoch": 0.76, "learning_rate": 2.810190320098387e-06, "loss": 0.5452, "step": 6272 }, { "epoch": 0.76, "learning_rate": 2.807453289639097e-06, "loss": 0.4928, "step": 6273 }, { "epoch": 0.76, "learning_rate": 2.804717375046049e-06, "loss": 0.4644, "step": 6274 }, { "epoch": 0.76, "learning_rate": 2.8019825767437025e-06, "loss": 0.4813, "step": 6275 }, { "epoch": 0.76, "learning_rate": 2.799248895156339e-06, "loss": 0.5119, "step": 6276 }, { "epoch": 0.76, "learning_rate": 2.7965163307080623e-06, "loss": 0.5256, "step": 6277 }, { "epoch": 0.76, "learning_rate": 2.7937848838228156e-06, "loss": 0.5288, "step": 6278 }, { "epoch": 0.76, "learning_rate": 2.7910545549243573e-06, "loss": 0.4836, "step": 6279 }, { "epoch": 0.76, "learning_rate": 2.7883253444362746e-06, "loss": 0.5506, "step": 6280 }, { "epoch": 0.76, "learning_rate": 2.785597252781986e-06, "loss": 0.4932, "step": 6281 }, { "epoch": 0.76, "learning_rate": 2.78287028038473e-06, "loss": 0.4858, "step": 6282 }, { "epoch": 0.76, "learning_rate": 2.780144427667578e-06, "loss": 0.4958, "step": 6283 }, { "epoch": 0.76, "learning_rate": 2.7774196950534205e-06, "loss": 0.4474, "step": 6284 }, { "epoch": 0.76, "learning_rate": 2.7746960829649826e-06, "loss": 0.4911, "step": 6285 }, { "epoch": 0.76, "learning_rate": 2.771973591824806e-06, "loss": 0.4596, "step": 6286 }, { "epoch": 0.76, "learning_rate": 2.7692522220552685e-06, "loss": 0.4928, "step": 6287 }, { "epoch": 0.76, "learning_rate": 2.7665319740785634e-06, "loss": 0.5903, "step": 6288 }, { "epoch": 0.76, "learning_rate": 2.7638128483167214e-06, "loss": 0.5497, "step": 6289 }, { "epoch": 0.76, "learning_rate": 2.7610948451915908e-06, "loss": 0.548, "step": 6290 }, { "epoch": 0.77, "learning_rate": 2.7583779651248443e-06, "loss": 0.4935, "step": 6291 }, { "epoch": 0.77, "learning_rate": 2.7556622085379903e-06, "loss": 0.5032, "step": 6292 }, { "epoch": 0.77, "learning_rate": 2.752947575852354e-06, "loss": 0.4979, "step": 6293 }, { "epoch": 0.77, "learning_rate": 2.750234067489086e-06, "loss": 0.5048, "step": 6294 }, { "epoch": 0.77, "learning_rate": 2.747521683869171e-06, "loss": 0.5109, "step": 6295 }, { "epoch": 0.77, "learning_rate": 2.744810425413408e-06, "loss": 0.5225, "step": 6296 }, { "epoch": 0.77, "learning_rate": 2.74210029254243e-06, "loss": 0.5207, "step": 6297 }, { "epoch": 0.77, "learning_rate": 2.7393912856766968e-06, "loss": 0.5055, "step": 6298 }, { "epoch": 0.77, "learning_rate": 2.7366834052364842e-06, "loss": 0.5601, "step": 6299 }, { "epoch": 0.77, "learning_rate": 2.7339766516418965e-06, "loss": 0.5704, "step": 6300 }, { "epoch": 0.77, "learning_rate": 2.7312710253128705e-06, "loss": 0.5452, "step": 6301 }, { "epoch": 0.77, "learning_rate": 2.7285665266691598e-06, "loss": 0.4608, "step": 6302 }, { "epoch": 0.77, "learning_rate": 2.7258631561303416e-06, "loss": 0.546, "step": 6303 }, { "epoch": 0.77, "learning_rate": 2.72316091411583e-06, "loss": 0.5142, "step": 6304 }, { "epoch": 0.77, "learning_rate": 2.720459801044849e-06, "loss": 0.5081, "step": 6305 }, { "epoch": 0.77, "learning_rate": 2.7177598173364617e-06, "loss": 0.5267, "step": 6306 }, { "epoch": 0.77, "learning_rate": 2.715060963409546e-06, "loss": 0.5537, "step": 6307 }, { "epoch": 0.77, "learning_rate": 2.7123632396828037e-06, "loss": 0.5507, "step": 6308 }, { "epoch": 0.77, "learning_rate": 2.709666646574769e-06, "loss": 0.5509, "step": 6309 }, { "epoch": 0.77, "learning_rate": 2.7069711845037993e-06, "loss": 0.5158, "step": 6310 }, { "epoch": 0.77, "learning_rate": 2.704276853888067e-06, "loss": 0.5561, "step": 6311 }, { "epoch": 0.77, "learning_rate": 2.7015836551455844e-06, "loss": 0.5626, "step": 6312 }, { "epoch": 0.77, "learning_rate": 2.6988915886941758e-06, "loss": 0.506, "step": 6313 }, { "epoch": 0.77, "learning_rate": 2.6962006549514896e-06, "loss": 0.5198, "step": 6314 }, { "epoch": 0.77, "learning_rate": 2.69351085433501e-06, "loss": 0.4939, "step": 6315 }, { "epoch": 0.77, "learning_rate": 2.6908221872620354e-06, "loss": 0.4667, "step": 6316 }, { "epoch": 0.77, "learning_rate": 2.6881346541496867e-06, "loss": 0.5242, "step": 6317 }, { "epoch": 0.77, "learning_rate": 2.68544825541492e-06, "loss": 0.5251, "step": 6318 }, { "epoch": 0.77, "learning_rate": 2.682762991474506e-06, "loss": 0.494, "step": 6319 }, { "epoch": 0.77, "learning_rate": 2.6800788627450347e-06, "loss": 0.5233, "step": 6320 }, { "epoch": 0.77, "learning_rate": 2.6773958696429413e-06, "loss": 0.5608, "step": 6321 }, { "epoch": 0.77, "learning_rate": 2.674714012584463e-06, "loss": 0.5221, "step": 6322 }, { "epoch": 0.77, "learning_rate": 2.672033291985666e-06, "loss": 0.5331, "step": 6323 }, { "epoch": 0.77, "learning_rate": 2.6693537082624498e-06, "loss": 0.5192, "step": 6324 }, { "epoch": 0.77, "learning_rate": 2.6666752618305257e-06, "loss": 0.5508, "step": 6325 }, { "epoch": 0.77, "learning_rate": 2.6639979531054305e-06, "loss": 0.4895, "step": 6326 }, { "epoch": 0.77, "learning_rate": 2.661321782502535e-06, "loss": 0.6, "step": 6327 }, { "epoch": 0.77, "learning_rate": 2.6586467504370184e-06, "loss": 0.4659, "step": 6328 }, { "epoch": 0.77, "learning_rate": 2.6559728573238963e-06, "loss": 0.5306, "step": 6329 }, { "epoch": 0.77, "learning_rate": 2.6533001035779993e-06, "loss": 0.5593, "step": 6330 }, { "epoch": 0.77, "learning_rate": 2.650628489613981e-06, "loss": 0.5331, "step": 6331 }, { "epoch": 0.77, "learning_rate": 2.6479580158463246e-06, "loss": 0.4982, "step": 6332 }, { "epoch": 0.77, "learning_rate": 2.645288682689334e-06, "loss": 0.6158, "step": 6333 }, { "epoch": 0.77, "learning_rate": 2.642620490557134e-06, "loss": 0.4904, "step": 6334 }, { "epoch": 0.77, "learning_rate": 2.6399534398636686e-06, "loss": 0.5477, "step": 6335 }, { "epoch": 0.77, "learning_rate": 2.6372875310227165e-06, "loss": 0.5562, "step": 6336 }, { "epoch": 0.77, "learning_rate": 2.6346227644478663e-06, "loss": 0.5283, "step": 6337 }, { "epoch": 0.77, "learning_rate": 2.6319591405525403e-06, "loss": 0.4574, "step": 6338 }, { "epoch": 0.77, "learning_rate": 2.629296659749976e-06, "loss": 0.4726, "step": 6339 }, { "epoch": 0.77, "learning_rate": 2.6266353224532338e-06, "loss": 0.4835, "step": 6340 }, { "epoch": 0.77, "learning_rate": 2.6239751290752035e-06, "loss": 0.5234, "step": 6341 }, { "epoch": 0.77, "learning_rate": 2.621316080028591e-06, "loss": 0.5275, "step": 6342 }, { "epoch": 0.77, "learning_rate": 2.6186581757259234e-06, "loss": 0.5377, "step": 6343 }, { "epoch": 0.77, "learning_rate": 2.6160014165795557e-06, "loss": 0.468, "step": 6344 }, { "epoch": 0.77, "learning_rate": 2.613345803001667e-06, "loss": 0.508, "step": 6345 }, { "epoch": 0.77, "learning_rate": 2.610691335404247e-06, "loss": 0.5188, "step": 6346 }, { "epoch": 0.77, "learning_rate": 2.6080380141991225e-06, "loss": 0.5378, "step": 6347 }, { "epoch": 0.77, "learning_rate": 2.6053858397979326e-06, "loss": 0.4579, "step": 6348 }, { "epoch": 0.77, "learning_rate": 2.6027348126121356e-06, "loss": 0.5001, "step": 6349 }, { "epoch": 0.77, "learning_rate": 2.6000849330530253e-06, "loss": 0.5155, "step": 6350 }, { "epoch": 0.77, "learning_rate": 2.5974362015317057e-06, "loss": 0.4981, "step": 6351 }, { "epoch": 0.77, "learning_rate": 2.5947886184591043e-06, "loss": 0.5023, "step": 6352 }, { "epoch": 0.77, "learning_rate": 2.592142184245976e-06, "loss": 0.4867, "step": 6353 }, { "epoch": 0.77, "learning_rate": 2.589496899302891e-06, "loss": 0.5317, "step": 6354 }, { "epoch": 0.77, "learning_rate": 2.5868527640402484e-06, "loss": 0.5251, "step": 6355 }, { "epoch": 0.77, "learning_rate": 2.5842097788682586e-06, "loss": 0.473, "step": 6356 }, { "epoch": 0.77, "learning_rate": 2.5815679441969667e-06, "loss": 0.5348, "step": 6357 }, { "epoch": 0.77, "learning_rate": 2.5789272604362247e-06, "loss": 0.4698, "step": 6358 }, { "epoch": 0.77, "learning_rate": 2.5762877279957222e-06, "loss": 0.5633, "step": 6359 }, { "epoch": 0.77, "learning_rate": 2.5736493472849535e-06, "loss": 0.5653, "step": 6360 }, { "epoch": 0.77, "learning_rate": 2.5710121187132476e-06, "loss": 0.5631, "step": 6361 }, { "epoch": 0.77, "learning_rate": 2.5683760426897486e-06, "loss": 0.5049, "step": 6362 }, { "epoch": 0.77, "learning_rate": 2.5657411196234184e-06, "loss": 0.4926, "step": 6363 }, { "epoch": 0.77, "learning_rate": 2.563107349923052e-06, "loss": 0.4969, "step": 6364 }, { "epoch": 0.77, "learning_rate": 2.5604747339972513e-06, "loss": 0.5237, "step": 6365 }, { "epoch": 0.77, "learning_rate": 2.5578432722544467e-06, "loss": 0.5904, "step": 6366 }, { "epoch": 0.77, "learning_rate": 2.5552129651028912e-06, "loss": 0.5186, "step": 6367 }, { "epoch": 0.77, "learning_rate": 2.552583812950652e-06, "loss": 0.5191, "step": 6368 }, { "epoch": 0.77, "learning_rate": 2.5499558162056236e-06, "loss": 0.4727, "step": 6369 }, { "epoch": 0.77, "learning_rate": 2.5473289752755215e-06, "loss": 0.535, "step": 6370 }, { "epoch": 0.77, "learning_rate": 2.5447032905678757e-06, "loss": 0.4793, "step": 6371 }, { "epoch": 0.77, "learning_rate": 2.542078762490038e-06, "loss": 0.5318, "step": 6372 }, { "epoch": 0.77, "learning_rate": 2.539455391449188e-06, "loss": 0.4913, "step": 6373 }, { "epoch": 0.78, "learning_rate": 2.53683317785232e-06, "loss": 0.474, "step": 6374 }, { "epoch": 0.78, "learning_rate": 2.534212122106244e-06, "loss": 0.5681, "step": 6375 }, { "epoch": 0.78, "learning_rate": 2.531592224617603e-06, "loss": 0.5513, "step": 6376 }, { "epoch": 0.78, "learning_rate": 2.528973485792847e-06, "loss": 0.5645, "step": 6377 }, { "epoch": 0.78, "learning_rate": 2.526355906038259e-06, "loss": 0.4765, "step": 6378 }, { "epoch": 0.78, "learning_rate": 2.5237394857599318e-06, "loss": 0.5067, "step": 6379 }, { "epoch": 0.78, "learning_rate": 2.5211242253637803e-06, "loss": 0.5404, "step": 6380 }, { "epoch": 0.78, "learning_rate": 2.518510125255542e-06, "loss": 0.4754, "step": 6381 }, { "epoch": 0.78, "learning_rate": 2.51589718584078e-06, "loss": 0.5777, "step": 6382 }, { "epoch": 0.78, "learning_rate": 2.5132854075248613e-06, "loss": 0.4807, "step": 6383 }, { "epoch": 0.78, "learning_rate": 2.5106747907129914e-06, "loss": 0.4714, "step": 6384 }, { "epoch": 0.78, "learning_rate": 2.5080653358101827e-06, "loss": 0.4872, "step": 6385 }, { "epoch": 0.78, "learning_rate": 2.5054570432212676e-06, "loss": 0.5517, "step": 6386 }, { "epoch": 0.78, "learning_rate": 2.502849913350909e-06, "loss": 0.5665, "step": 6387 }, { "epoch": 0.78, "learning_rate": 2.500243946603579e-06, "loss": 0.5637, "step": 6388 }, { "epoch": 0.78, "learning_rate": 2.497639143383569e-06, "loss": 0.5382, "step": 6389 }, { "epoch": 0.78, "learning_rate": 2.4950355040949993e-06, "loss": 0.5633, "step": 6390 }, { "epoch": 0.78, "learning_rate": 2.492433029141802e-06, "loss": 0.5415, "step": 6391 }, { "epoch": 0.78, "learning_rate": 2.4898317189277234e-06, "loss": 0.4943, "step": 6392 }, { "epoch": 0.78, "learning_rate": 2.487231573856348e-06, "loss": 0.499, "step": 6393 }, { "epoch": 0.78, "learning_rate": 2.484632594331062e-06, "loss": 0.4469, "step": 6394 }, { "epoch": 0.78, "learning_rate": 2.4820347807550728e-06, "loss": 0.4501, "step": 6395 }, { "epoch": 0.78, "learning_rate": 2.4794381335314175e-06, "loss": 0.5181, "step": 6396 }, { "epoch": 0.78, "learning_rate": 2.4768426530629407e-06, "loss": 0.5292, "step": 6397 }, { "epoch": 0.78, "learning_rate": 2.474248339752309e-06, "loss": 0.5537, "step": 6398 }, { "epoch": 0.78, "learning_rate": 2.471655194002014e-06, "loss": 0.5652, "step": 6399 }, { "epoch": 0.78, "learning_rate": 2.4690632162143567e-06, "loss": 0.4443, "step": 6400 }, { "epoch": 0.78, "learning_rate": 2.466472406791467e-06, "loss": 0.5303, "step": 6401 }, { "epoch": 0.78, "learning_rate": 2.463882766135286e-06, "loss": 0.5161, "step": 6402 }, { "epoch": 0.78, "learning_rate": 2.4612942946475727e-06, "loss": 0.4767, "step": 6403 }, { "epoch": 0.78, "learning_rate": 2.45870699272991e-06, "loss": 0.4964, "step": 6404 }, { "epoch": 0.78, "learning_rate": 2.4561208607837007e-06, "loss": 0.5168, "step": 6405 }, { "epoch": 0.78, "learning_rate": 2.4535358992101567e-06, "loss": 0.4883, "step": 6406 }, { "epoch": 0.78, "learning_rate": 2.4509521084103204e-06, "loss": 0.4869, "step": 6407 }, { "epoch": 0.78, "learning_rate": 2.448369488785043e-06, "loss": 0.5325, "step": 6408 }, { "epoch": 0.78, "learning_rate": 2.445788040734993e-06, "loss": 0.4623, "step": 6409 }, { "epoch": 0.78, "learning_rate": 2.44320776466067e-06, "loss": 0.5538, "step": 6410 }, { "epoch": 0.78, "learning_rate": 2.4406286609623787e-06, "loss": 0.5784, "step": 6411 }, { "epoch": 0.78, "learning_rate": 2.4380507300402436e-06, "loss": 0.5518, "step": 6412 }, { "epoch": 0.78, "learning_rate": 2.4354739722942155e-06, "loss": 0.5255, "step": 6413 }, { "epoch": 0.78, "learning_rate": 2.4328983881240566e-06, "loss": 0.4788, "step": 6414 }, { "epoch": 0.78, "learning_rate": 2.430323977929344e-06, "loss": 0.5966, "step": 6415 }, { "epoch": 0.78, "learning_rate": 2.4277507421094804e-06, "loss": 0.4817, "step": 6416 }, { "epoch": 0.78, "learning_rate": 2.4251786810636856e-06, "loss": 0.566, "step": 6417 }, { "epoch": 0.78, "learning_rate": 2.4226077951909876e-06, "loss": 0.5346, "step": 6418 }, { "epoch": 0.78, "learning_rate": 2.4200380848902462e-06, "loss": 0.4931, "step": 6419 }, { "epoch": 0.78, "learning_rate": 2.4174695505601277e-06, "loss": 0.4313, "step": 6420 }, { "epoch": 0.78, "learning_rate": 2.4149021925991156e-06, "loss": 0.5258, "step": 6421 }, { "epoch": 0.78, "learning_rate": 2.412336011405524e-06, "loss": 0.4948, "step": 6422 }, { "epoch": 0.78, "learning_rate": 2.409771007377466e-06, "loss": 0.5269, "step": 6423 }, { "epoch": 0.78, "learning_rate": 2.4072071809128904e-06, "loss": 0.5675, "step": 6424 }, { "epoch": 0.78, "learning_rate": 2.404644532409549e-06, "loss": 0.5124, "step": 6425 }, { "epoch": 0.78, "learning_rate": 2.402083062265016e-06, "loss": 0.5162, "step": 6426 }, { "epoch": 0.78, "learning_rate": 2.3995227708766866e-06, "loss": 0.4942, "step": 6427 }, { "epoch": 0.78, "learning_rate": 2.3969636586417644e-06, "loss": 0.5053, "step": 6428 }, { "epoch": 0.78, "learning_rate": 2.3944057259572783e-06, "loss": 0.5584, "step": 6429 }, { "epoch": 0.78, "learning_rate": 2.3918489732200746e-06, "loss": 0.5495, "step": 6430 }, { "epoch": 0.78, "learning_rate": 2.3892934008268087e-06, "loss": 0.4766, "step": 6431 }, { "epoch": 0.78, "learning_rate": 2.3867390091739563e-06, "loss": 0.5139, "step": 6432 }, { "epoch": 0.78, "learning_rate": 2.3841857986578154e-06, "loss": 0.6086, "step": 6433 }, { "epoch": 0.78, "learning_rate": 2.381633769674494e-06, "loss": 0.4561, "step": 6434 }, { "epoch": 0.78, "learning_rate": 2.3790829226199154e-06, "loss": 0.4937, "step": 6435 }, { "epoch": 0.78, "learning_rate": 2.376533257889829e-06, "loss": 0.5762, "step": 6436 }, { "epoch": 0.78, "learning_rate": 2.3739847758797928e-06, "loss": 0.4883, "step": 6437 }, { "epoch": 0.78, "learning_rate": 2.371437476985181e-06, "loss": 0.4511, "step": 6438 }, { "epoch": 0.78, "learning_rate": 2.368891361601191e-06, "loss": 0.4614, "step": 6439 }, { "epoch": 0.78, "learning_rate": 2.3663464301228267e-06, "loss": 0.5297, "step": 6440 }, { "epoch": 0.78, "learning_rate": 2.3638026829449177e-06, "loss": 0.51, "step": 6441 }, { "epoch": 0.78, "learning_rate": 2.361260120462109e-06, "loss": 0.5285, "step": 6442 }, { "epoch": 0.78, "learning_rate": 2.358718743068855e-06, "loss": 0.5607, "step": 6443 }, { "epoch": 0.78, "learning_rate": 2.3561785511594283e-06, "loss": 0.493, "step": 6444 }, { "epoch": 0.78, "learning_rate": 2.353639545127925e-06, "loss": 0.5453, "step": 6445 }, { "epoch": 0.78, "learning_rate": 2.3511017253682442e-06, "loss": 0.5072, "step": 6446 }, { "epoch": 0.78, "learning_rate": 2.3485650922741166e-06, "loss": 0.5292, "step": 6447 }, { "epoch": 0.78, "learning_rate": 2.3460296462390766e-06, "loss": 0.5126, "step": 6448 }, { "epoch": 0.78, "learning_rate": 2.3434953876564747e-06, "loss": 0.4547, "step": 6449 }, { "epoch": 0.78, "learning_rate": 2.3409623169194884e-06, "loss": 0.4685, "step": 6450 }, { "epoch": 0.78, "learning_rate": 2.3384304344210995e-06, "loss": 0.4577, "step": 6451 }, { "epoch": 0.78, "learning_rate": 2.335899740554104e-06, "loss": 0.5357, "step": 6452 }, { "epoch": 0.78, "learning_rate": 2.3333702357111287e-06, "loss": 0.4634, "step": 6453 }, { "epoch": 0.78, "learning_rate": 2.3308419202846033e-06, "loss": 0.5267, "step": 6454 }, { "epoch": 0.78, "learning_rate": 2.3283147946667707e-06, "loss": 0.4884, "step": 6455 }, { "epoch": 0.79, "learning_rate": 2.3257888592497012e-06, "loss": 0.536, "step": 6456 }, { "epoch": 0.79, "learning_rate": 2.3232641144252697e-06, "loss": 0.5443, "step": 6457 }, { "epoch": 0.79, "learning_rate": 2.3207405605851683e-06, "loss": 0.5701, "step": 6458 }, { "epoch": 0.79, "learning_rate": 2.3182181981209117e-06, "loss": 0.5297, "step": 6459 }, { "epoch": 0.79, "learning_rate": 2.31569702742382e-06, "loss": 0.5531, "step": 6460 }, { "epoch": 0.79, "learning_rate": 2.3131770488850324e-06, "loss": 0.5017, "step": 6461 }, { "epoch": 0.79, "learning_rate": 2.3106582628955075e-06, "loss": 0.5138, "step": 6462 }, { "epoch": 0.79, "learning_rate": 2.3081406698460097e-06, "loss": 0.4833, "step": 6463 }, { "epoch": 0.79, "learning_rate": 2.305624270127126e-06, "loss": 0.5285, "step": 6464 }, { "epoch": 0.79, "learning_rate": 2.303109064129259e-06, "loss": 0.5347, "step": 6465 }, { "epoch": 0.79, "learning_rate": 2.300595052242619e-06, "loss": 0.564, "step": 6466 }, { "epoch": 0.79, "learning_rate": 2.298082234857233e-06, "loss": 0.5011, "step": 6467 }, { "epoch": 0.79, "learning_rate": 2.2955706123629504e-06, "loss": 0.4822, "step": 6468 }, { "epoch": 0.79, "learning_rate": 2.2930601851494225e-06, "loss": 0.4961, "step": 6469 }, { "epoch": 0.79, "learning_rate": 2.29055095360613e-06, "loss": 0.4679, "step": 6470 }, { "epoch": 0.79, "learning_rate": 2.288042918122355e-06, "loss": 0.4953, "step": 6471 }, { "epoch": 0.79, "learning_rate": 2.2855360790871983e-06, "loss": 0.4979, "step": 6472 }, { "epoch": 0.79, "learning_rate": 2.2830304368895793e-06, "loss": 0.5519, "step": 6473 }, { "epoch": 0.79, "learning_rate": 2.2805259919182277e-06, "loss": 0.4951, "step": 6474 }, { "epoch": 0.79, "learning_rate": 2.2780227445616844e-06, "loss": 0.4915, "step": 6475 }, { "epoch": 0.79, "learning_rate": 2.2755206952083107e-06, "loss": 0.5735, "step": 6476 }, { "epoch": 0.79, "learning_rate": 2.273019844246284e-06, "loss": 0.5014, "step": 6477 }, { "epoch": 0.79, "learning_rate": 2.2705201920635843e-06, "loss": 0.5135, "step": 6478 }, { "epoch": 0.79, "learning_rate": 2.2680217390480186e-06, "loss": 0.5079, "step": 6479 }, { "epoch": 0.79, "learning_rate": 2.2655244855871996e-06, "loss": 0.5134, "step": 6480 }, { "epoch": 0.79, "learning_rate": 2.2630284320685525e-06, "loss": 0.5558, "step": 6481 }, { "epoch": 0.79, "learning_rate": 2.260533578879326e-06, "loss": 0.5062, "step": 6482 }, { "epoch": 0.79, "learning_rate": 2.2580399264065746e-06, "loss": 0.54, "step": 6483 }, { "epoch": 0.79, "learning_rate": 2.2555474750371655e-06, "loss": 0.5745, "step": 6484 }, { "epoch": 0.79, "learning_rate": 2.2530562251577883e-06, "loss": 0.447, "step": 6485 }, { "epoch": 0.79, "learning_rate": 2.2505661771549335e-06, "loss": 0.5526, "step": 6486 }, { "epoch": 0.79, "learning_rate": 2.2480773314149205e-06, "loss": 0.5023, "step": 6487 }, { "epoch": 0.79, "learning_rate": 2.245589688323866e-06, "loss": 0.5123, "step": 6488 }, { "epoch": 0.79, "learning_rate": 2.243103248267714e-06, "loss": 0.472, "step": 6489 }, { "epoch": 0.79, "learning_rate": 2.240618011632212e-06, "loss": 0.5593, "step": 6490 }, { "epoch": 0.79, "learning_rate": 2.2381339788029275e-06, "loss": 0.5205, "step": 6491 }, { "epoch": 0.79, "learning_rate": 2.2356511501652346e-06, "loss": 0.5431, "step": 6492 }, { "epoch": 0.79, "learning_rate": 2.2331695261043297e-06, "loss": 0.4793, "step": 6493 }, { "epoch": 0.79, "learning_rate": 2.230689107005214e-06, "loss": 0.5412, "step": 6494 }, { "epoch": 0.79, "learning_rate": 2.2282098932527007e-06, "loss": 0.4928, "step": 6495 }, { "epoch": 0.79, "learning_rate": 2.225731885231428e-06, "loss": 0.4556, "step": 6496 }, { "epoch": 0.79, "learning_rate": 2.2232550833258347e-06, "loss": 0.4953, "step": 6497 }, { "epoch": 0.79, "learning_rate": 2.220779487920175e-06, "loss": 0.5164, "step": 6498 }, { "epoch": 0.79, "learning_rate": 2.218305099398522e-06, "loss": 0.4921, "step": 6499 }, { "epoch": 0.79, "learning_rate": 2.215831918144753e-06, "loss": 0.5112, "step": 6500 }, { "epoch": 0.79, "learning_rate": 2.2133599445425646e-06, "loss": 0.4955, "step": 6501 }, { "epoch": 0.79, "learning_rate": 2.210889178975467e-06, "loss": 0.4797, "step": 6502 }, { "epoch": 0.79, "learning_rate": 2.208419621826776e-06, "loss": 0.5085, "step": 6503 }, { "epoch": 0.79, "learning_rate": 2.205951273479623e-06, "loss": 0.533, "step": 6504 }, { "epoch": 0.79, "learning_rate": 2.2034841343169555e-06, "loss": 0.5266, "step": 6505 }, { "epoch": 0.79, "learning_rate": 2.201018204721529e-06, "loss": 0.4711, "step": 6506 }, { "epoch": 0.79, "learning_rate": 2.1985534850759104e-06, "loss": 0.4605, "step": 6507 }, { "epoch": 0.79, "learning_rate": 2.1960899757624853e-06, "loss": 0.5251, "step": 6508 }, { "epoch": 0.79, "learning_rate": 2.1936276771634447e-06, "loss": 0.4908, "step": 6509 }, { "epoch": 0.79, "learning_rate": 2.1911665896607968e-06, "loss": 0.4596, "step": 6510 }, { "epoch": 0.79, "learning_rate": 2.1887067136363593e-06, "loss": 0.483, "step": 6511 }, { "epoch": 0.79, "learning_rate": 2.1862480494717584e-06, "loss": 0.4979, "step": 6512 }, { "epoch": 0.79, "learning_rate": 2.1837905975484397e-06, "loss": 0.5104, "step": 6513 }, { "epoch": 0.79, "learning_rate": 2.18133435824766e-06, "loss": 0.4701, "step": 6514 }, { "epoch": 0.79, "learning_rate": 2.178879331950482e-06, "loss": 0.5116, "step": 6515 }, { "epoch": 0.79, "learning_rate": 2.1764255190377814e-06, "loss": 0.5087, "step": 6516 }, { "epoch": 0.79, "learning_rate": 2.1739729198902528e-06, "loss": 0.4968, "step": 6517 }, { "epoch": 0.79, "learning_rate": 2.1715215348883913e-06, "loss": 0.5256, "step": 6518 }, { "epoch": 0.79, "learning_rate": 2.169071364412515e-06, "loss": 0.546, "step": 6519 }, { "epoch": 0.79, "learning_rate": 2.166622408842748e-06, "loss": 0.5048, "step": 6520 }, { "epoch": 0.79, "learning_rate": 2.1641746685590205e-06, "loss": 0.5098, "step": 6521 }, { "epoch": 0.79, "learning_rate": 2.1617281439410876e-06, "loss": 0.524, "step": 6522 }, { "epoch": 0.79, "learning_rate": 2.1592828353685037e-06, "loss": 0.5036, "step": 6523 }, { "epoch": 0.79, "learning_rate": 2.1568387432206337e-06, "loss": 0.4888, "step": 6524 }, { "epoch": 0.79, "learning_rate": 2.1543958678766706e-06, "loss": 0.4986, "step": 6525 }, { "epoch": 0.79, "learning_rate": 2.151954209715602e-06, "loss": 0.4823, "step": 6526 }, { "epoch": 0.79, "learning_rate": 2.1495137691162284e-06, "loss": 0.5052, "step": 6527 }, { "epoch": 0.79, "learning_rate": 2.1470745464571695e-06, "loss": 0.5225, "step": 6528 }, { "epoch": 0.79, "learning_rate": 2.144636542116849e-06, "loss": 0.5213, "step": 6529 }, { "epoch": 0.79, "learning_rate": 2.1421997564735007e-06, "loss": 0.4987, "step": 6530 }, { "epoch": 0.79, "learning_rate": 2.139764189905179e-06, "loss": 0.4729, "step": 6531 }, { "epoch": 0.79, "learning_rate": 2.137329842789736e-06, "loss": 0.483, "step": 6532 }, { "epoch": 0.79, "learning_rate": 2.1348967155048483e-06, "loss": 0.51, "step": 6533 }, { "epoch": 0.79, "learning_rate": 2.1324648084279907e-06, "loss": 0.5079, "step": 6534 }, { "epoch": 0.79, "learning_rate": 2.130034121936455e-06, "loss": 0.4891, "step": 6535 }, { "epoch": 0.79, "learning_rate": 2.1276046564073425e-06, "loss": 0.5419, "step": 6536 }, { "epoch": 0.79, "learning_rate": 2.125176412217569e-06, "loss": 0.5015, "step": 6537 }, { "epoch": 0.8, "learning_rate": 2.122749389743856e-06, "loss": 0.4873, "step": 6538 }, { "epoch": 0.8, "learning_rate": 2.1203235893627314e-06, "loss": 0.5016, "step": 6539 }, { "epoch": 0.8, "learning_rate": 2.117899011450546e-06, "loss": 0.5339, "step": 6540 }, { "epoch": 0.8, "learning_rate": 2.1154756563834477e-06, "loss": 0.5249, "step": 6541 }, { "epoch": 0.8, "learning_rate": 2.113053524537406e-06, "loss": 0.529, "step": 6542 }, { "epoch": 0.8, "learning_rate": 2.110632616288194e-06, "loss": 0.5179, "step": 6543 }, { "epoch": 0.8, "learning_rate": 2.108212932011392e-06, "loss": 0.4968, "step": 6544 }, { "epoch": 0.8, "learning_rate": 2.1057944720824007e-06, "loss": 0.5097, "step": 6545 }, { "epoch": 0.8, "learning_rate": 2.1033772368764217e-06, "loss": 0.4926, "step": 6546 }, { "epoch": 0.8, "learning_rate": 2.1009612267684677e-06, "loss": 0.515, "step": 6547 }, { "epoch": 0.8, "learning_rate": 2.0985464421333658e-06, "loss": 0.5299, "step": 6548 }, { "epoch": 0.8, "learning_rate": 2.096132883345753e-06, "loss": 0.5755, "step": 6549 }, { "epoch": 0.8, "learning_rate": 2.0937205507800684e-06, "loss": 0.5314, "step": 6550 }, { "epoch": 0.8, "learning_rate": 2.091309444810572e-06, "loss": 0.51, "step": 6551 }, { "epoch": 0.8, "learning_rate": 2.088899565811324e-06, "loss": 0.5307, "step": 6552 }, { "epoch": 0.8, "learning_rate": 2.0864909141561972e-06, "loss": 0.5202, "step": 6553 }, { "epoch": 0.8, "learning_rate": 2.084083490218879e-06, "loss": 0.5413, "step": 6554 }, { "epoch": 0.8, "learning_rate": 2.081677294372858e-06, "loss": 0.5032, "step": 6555 }, { "epoch": 0.8, "learning_rate": 2.079272326991435e-06, "loss": 0.5062, "step": 6556 }, { "epoch": 0.8, "learning_rate": 2.0768685884477267e-06, "loss": 0.5017, "step": 6557 }, { "epoch": 0.8, "learning_rate": 2.0744660791146477e-06, "loss": 0.4364, "step": 6558 }, { "epoch": 0.8, "learning_rate": 2.0720647993649356e-06, "loss": 0.5184, "step": 6559 }, { "epoch": 0.8, "learning_rate": 2.069664749571122e-06, "loss": 0.5443, "step": 6560 }, { "epoch": 0.8, "learning_rate": 2.067265930105563e-06, "loss": 0.5661, "step": 6561 }, { "epoch": 0.8, "learning_rate": 2.064868341340409e-06, "loss": 0.5038, "step": 6562 }, { "epoch": 0.8, "learning_rate": 2.0624719836476336e-06, "loss": 0.5554, "step": 6563 }, { "epoch": 0.8, "learning_rate": 2.060076857399007e-06, "loss": 0.5225, "step": 6564 }, { "epoch": 0.8, "learning_rate": 2.0576829629661178e-06, "loss": 0.4375, "step": 6565 }, { "epoch": 0.8, "learning_rate": 2.0552903007203596e-06, "loss": 0.5242, "step": 6566 }, { "epoch": 0.8, "learning_rate": 2.052898871032929e-06, "loss": 0.5074, "step": 6567 }, { "epoch": 0.8, "learning_rate": 2.0505086742748457e-06, "loss": 0.491, "step": 6568 }, { "epoch": 0.8, "learning_rate": 2.048119710816925e-06, "loss": 0.5188, "step": 6569 }, { "epoch": 0.8, "learning_rate": 2.0457319810297936e-06, "loss": 0.5241, "step": 6570 }, { "epoch": 0.8, "learning_rate": 2.043345485283893e-06, "loss": 0.5135, "step": 6571 }, { "epoch": 0.8, "learning_rate": 2.0409602239494653e-06, "loss": 0.5486, "step": 6572 }, { "epoch": 0.8, "learning_rate": 2.0385761973965666e-06, "loss": 0.5324, "step": 6573 }, { "epoch": 0.8, "learning_rate": 2.0361934059950627e-06, "loss": 0.4523, "step": 6574 }, { "epoch": 0.8, "learning_rate": 2.033811850114621e-06, "loss": 0.5064, "step": 6575 }, { "epoch": 0.8, "learning_rate": 2.031431530124719e-06, "loss": 0.5138, "step": 6576 }, { "epoch": 0.8, "learning_rate": 2.0290524463946503e-06, "loss": 0.5454, "step": 6577 }, { "epoch": 0.8, "learning_rate": 2.0266745992935065e-06, "loss": 0.4338, "step": 6578 }, { "epoch": 0.8, "learning_rate": 2.0242979891901894e-06, "loss": 0.5412, "step": 6579 }, { "epoch": 0.8, "learning_rate": 2.021922616453418e-06, "loss": 0.5187, "step": 6580 }, { "epoch": 0.8, "learning_rate": 2.019548481451705e-06, "loss": 0.4561, "step": 6581 }, { "epoch": 0.8, "learning_rate": 2.0171755845533847e-06, "loss": 0.5088, "step": 6582 }, { "epoch": 0.8, "learning_rate": 2.014803926126592e-06, "loss": 0.5228, "step": 6583 }, { "epoch": 0.8, "learning_rate": 2.0124335065392654e-06, "loss": 0.5702, "step": 6584 }, { "epoch": 0.8, "learning_rate": 2.01006432615916e-06, "loss": 0.535, "step": 6585 }, { "epoch": 0.8, "learning_rate": 2.007696385353839e-06, "loss": 0.5114, "step": 6586 }, { "epoch": 0.8, "learning_rate": 2.0053296844906644e-06, "loss": 0.4885, "step": 6587 }, { "epoch": 0.8, "learning_rate": 2.0029642239368143e-06, "loss": 0.5174, "step": 6588 }, { "epoch": 0.8, "learning_rate": 2.00060000405927e-06, "loss": 0.55, "step": 6589 }, { "epoch": 0.8, "learning_rate": 1.9982370252248184e-06, "loss": 0.5415, "step": 6590 }, { "epoch": 0.8, "learning_rate": 1.995875287800062e-06, "loss": 0.4925, "step": 6591 }, { "epoch": 0.8, "learning_rate": 1.993514792151402e-06, "loss": 0.54, "step": 6592 }, { "epoch": 0.8, "learning_rate": 1.99115553864505e-06, "loss": 0.4727, "step": 6593 }, { "epoch": 0.8, "learning_rate": 1.9887975276470285e-06, "loss": 0.5174, "step": 6594 }, { "epoch": 0.8, "learning_rate": 1.9864407595231617e-06, "loss": 0.4704, "step": 6595 }, { "epoch": 0.8, "learning_rate": 1.9840852346390794e-06, "loss": 0.5291, "step": 6596 }, { "epoch": 0.8, "learning_rate": 1.9817309533602325e-06, "loss": 0.4685, "step": 6597 }, { "epoch": 0.8, "learning_rate": 1.9793779160518634e-06, "loss": 0.4393, "step": 6598 }, { "epoch": 0.8, "learning_rate": 1.9770261230790242e-06, "loss": 0.499, "step": 6599 }, { "epoch": 0.8, "learning_rate": 1.9746755748065825e-06, "loss": 0.5015, "step": 6600 }, { "epoch": 0.8, "learning_rate": 1.972326271599204e-06, "loss": 0.5715, "step": 6601 }, { "epoch": 0.8, "learning_rate": 1.969978213821363e-06, "loss": 0.5317, "step": 6602 }, { "epoch": 0.8, "learning_rate": 1.9676314018373466e-06, "loss": 0.5256, "step": 6603 }, { "epoch": 0.8, "learning_rate": 1.965285836011237e-06, "loss": 0.4804, "step": 6604 }, { "epoch": 0.8, "learning_rate": 1.9629415167069376e-06, "loss": 0.5063, "step": 6605 }, { "epoch": 0.8, "learning_rate": 1.9605984442881477e-06, "loss": 0.4692, "step": 6606 }, { "epoch": 0.8, "learning_rate": 1.958256619118372e-06, "loss": 0.5365, "step": 6607 }, { "epoch": 0.8, "learning_rate": 1.95591604156093e-06, "loss": 0.5519, "step": 6608 }, { "epoch": 0.8, "learning_rate": 1.9535767119789475e-06, "loss": 0.545, "step": 6609 }, { "epoch": 0.8, "learning_rate": 1.9512386307353447e-06, "loss": 0.5351, "step": 6610 }, { "epoch": 0.8, "learning_rate": 1.948901798192864e-06, "loss": 0.5231, "step": 6611 }, { "epoch": 0.8, "learning_rate": 1.946566214714043e-06, "loss": 0.4871, "step": 6612 }, { "epoch": 0.8, "learning_rate": 1.944231880661225e-06, "loss": 0.5045, "step": 6613 }, { "epoch": 0.8, "learning_rate": 1.9418987963965697e-06, "loss": 0.51, "step": 6614 }, { "epoch": 0.8, "learning_rate": 1.9395669622820347e-06, "loss": 0.5081, "step": 6615 }, { "epoch": 0.8, "learning_rate": 1.93723637867938e-06, "loss": 0.4469, "step": 6616 }, { "epoch": 0.8, "learning_rate": 1.9349070459501862e-06, "loss": 0.5813, "step": 6617 }, { "epoch": 0.8, "learning_rate": 1.9325789644558247e-06, "loss": 0.538, "step": 6618 }, { "epoch": 0.8, "learning_rate": 1.9302521345574775e-06, "loss": 0.4562, "step": 6619 }, { "epoch": 0.81, "learning_rate": 1.9279265566161364e-06, "loss": 0.4854, "step": 6620 }, { "epoch": 0.81, "learning_rate": 1.9256022309925995e-06, "loss": 0.558, "step": 6621 }, { "epoch": 0.81, "learning_rate": 1.923279158047461e-06, "loss": 0.4776, "step": 6622 }, { "epoch": 0.81, "learning_rate": 1.9209573381411317e-06, "loss": 0.5223, "step": 6623 }, { "epoch": 0.81, "learning_rate": 1.9186367716338227e-06, "loss": 0.5395, "step": 6624 }, { "epoch": 0.81, "learning_rate": 1.916317458885548e-06, "loss": 0.5119, "step": 6625 }, { "epoch": 0.81, "learning_rate": 1.9139994002561357e-06, "loss": 0.5048, "step": 6626 }, { "epoch": 0.81, "learning_rate": 1.911682596105209e-06, "loss": 0.4725, "step": 6627 }, { "epoch": 0.81, "learning_rate": 1.9093670467922066e-06, "loss": 0.4352, "step": 6628 }, { "epoch": 0.81, "learning_rate": 1.907052752676364e-06, "loss": 0.5395, "step": 6629 }, { "epoch": 0.81, "learning_rate": 1.9047397141167245e-06, "loss": 0.4824, "step": 6630 }, { "epoch": 0.81, "learning_rate": 1.9024279314721417e-06, "loss": 0.5226, "step": 6631 }, { "epoch": 0.81, "learning_rate": 1.9001174051012661e-06, "loss": 0.5295, "step": 6632 }, { "epoch": 0.81, "learning_rate": 1.8978081353625587e-06, "loss": 0.4369, "step": 6633 }, { "epoch": 0.81, "learning_rate": 1.8955001226142877e-06, "loss": 0.4896, "step": 6634 }, { "epoch": 0.81, "learning_rate": 1.893193367214521e-06, "loss": 0.5008, "step": 6635 }, { "epoch": 0.81, "learning_rate": 1.8908878695211287e-06, "loss": 0.528, "step": 6636 }, { "epoch": 0.81, "learning_rate": 1.888583629891797e-06, "loss": 0.5174, "step": 6637 }, { "epoch": 0.81, "learning_rate": 1.8862806486840068e-06, "loss": 0.5545, "step": 6638 }, { "epoch": 0.81, "learning_rate": 1.8839789262550457e-06, "loss": 0.4568, "step": 6639 }, { "epoch": 0.81, "learning_rate": 1.881678462962012e-06, "loss": 0.4684, "step": 6640 }, { "epoch": 0.81, "learning_rate": 1.8793792591618032e-06, "loss": 0.5179, "step": 6641 }, { "epoch": 0.81, "learning_rate": 1.8770813152111167e-06, "loss": 0.4769, "step": 6642 }, { "epoch": 0.81, "learning_rate": 1.8747846314664685e-06, "loss": 0.4739, "step": 6643 }, { "epoch": 0.81, "learning_rate": 1.872489208284164e-06, "loss": 0.5217, "step": 6644 }, { "epoch": 0.81, "learning_rate": 1.8701950460203222e-06, "loss": 0.5298, "step": 6645 }, { "epoch": 0.81, "learning_rate": 1.8679021450308676e-06, "loss": 0.5282, "step": 6646 }, { "epoch": 0.81, "learning_rate": 1.8656105056715223e-06, "loss": 0.5151, "step": 6647 }, { "epoch": 0.81, "learning_rate": 1.8633201282978142e-06, "loss": 0.4958, "step": 6648 }, { "epoch": 0.81, "learning_rate": 1.861031013265081e-06, "loss": 0.5165, "step": 6649 }, { "epoch": 0.81, "learning_rate": 1.8587431609284556e-06, "loss": 0.4567, "step": 6650 }, { "epoch": 0.81, "learning_rate": 1.8564565716428851e-06, "loss": 0.5193, "step": 6651 }, { "epoch": 0.81, "learning_rate": 1.8541712457631144e-06, "loss": 0.4976, "step": 6652 }, { "epoch": 0.81, "learning_rate": 1.8518871836436902e-06, "loss": 0.4575, "step": 6653 }, { "epoch": 0.81, "learning_rate": 1.84960438563897e-06, "loss": 0.5475, "step": 6654 }, { "epoch": 0.81, "learning_rate": 1.8473228521031116e-06, "loss": 0.5289, "step": 6655 }, { "epoch": 0.81, "learning_rate": 1.8450425833900709e-06, "loss": 0.5137, "step": 6656 }, { "epoch": 0.81, "learning_rate": 1.8427635798536236e-06, "loss": 0.4942, "step": 6657 }, { "epoch": 0.81, "learning_rate": 1.8404858418473336e-06, "loss": 0.5118, "step": 6658 }, { "epoch": 0.81, "learning_rate": 1.8382093697245707e-06, "loss": 0.5136, "step": 6659 }, { "epoch": 0.81, "learning_rate": 1.8359341638385175e-06, "loss": 0.5143, "step": 6660 }, { "epoch": 0.81, "learning_rate": 1.8336602245421509e-06, "loss": 0.4764, "step": 6661 }, { "epoch": 0.81, "learning_rate": 1.831387552188252e-06, "loss": 0.5557, "step": 6662 }, { "epoch": 0.81, "learning_rate": 1.8291161471294116e-06, "loss": 0.5272, "step": 6663 }, { "epoch": 0.81, "learning_rate": 1.8268460097180207e-06, "loss": 0.465, "step": 6664 }, { "epoch": 0.81, "learning_rate": 1.8245771403062674e-06, "loss": 0.5076, "step": 6665 }, { "epoch": 0.81, "learning_rate": 1.8223095392461555e-06, "loss": 0.505, "step": 6666 }, { "epoch": 0.81, "learning_rate": 1.820043206889479e-06, "loss": 0.5597, "step": 6667 }, { "epoch": 0.81, "learning_rate": 1.8177781435878449e-06, "loss": 0.5502, "step": 6668 }, { "epoch": 0.81, "learning_rate": 1.815514349692662e-06, "loss": 0.5057, "step": 6669 }, { "epoch": 0.81, "learning_rate": 1.8132518255551368e-06, "loss": 0.5139, "step": 6670 }, { "epoch": 0.81, "learning_rate": 1.8109905715262799e-06, "loss": 0.529, "step": 6671 }, { "epoch": 0.81, "learning_rate": 1.8087305879569118e-06, "loss": 0.4878, "step": 6672 }, { "epoch": 0.81, "learning_rate": 1.8064718751976451e-06, "loss": 0.4585, "step": 6673 }, { "epoch": 0.81, "learning_rate": 1.8042144335989065e-06, "loss": 0.6064, "step": 6674 }, { "epoch": 0.81, "learning_rate": 1.8019582635109177e-06, "loss": 0.5051, "step": 6675 }, { "epoch": 0.81, "learning_rate": 1.7997033652837037e-06, "loss": 0.5311, "step": 6676 }, { "epoch": 0.81, "learning_rate": 1.7974497392670975e-06, "loss": 0.5331, "step": 6677 }, { "epoch": 0.81, "learning_rate": 1.7951973858107297e-06, "loss": 0.4863, "step": 6678 }, { "epoch": 0.81, "learning_rate": 1.7929463052640328e-06, "loss": 0.4588, "step": 6679 }, { "epoch": 0.81, "learning_rate": 1.7906964979762452e-06, "loss": 0.5341, "step": 6680 }, { "epoch": 0.81, "learning_rate": 1.7884479642964103e-06, "loss": 0.5146, "step": 6681 }, { "epoch": 0.81, "learning_rate": 1.7862007045733642e-06, "loss": 0.481, "step": 6682 }, { "epoch": 0.81, "learning_rate": 1.783954719155757e-06, "loss": 0.4788, "step": 6683 }, { "epoch": 0.81, "learning_rate": 1.7817100083920336e-06, "loss": 0.5248, "step": 6684 }, { "epoch": 0.81, "learning_rate": 1.7794665726304383e-06, "loss": 0.5435, "step": 6685 }, { "epoch": 0.81, "learning_rate": 1.7772244122190297e-06, "loss": 0.5187, "step": 6686 }, { "epoch": 0.81, "learning_rate": 1.774983527505656e-06, "loss": 0.4966, "step": 6687 }, { "epoch": 0.81, "learning_rate": 1.7727439188379725e-06, "loss": 0.5488, "step": 6688 }, { "epoch": 0.81, "learning_rate": 1.7705055865634401e-06, "loss": 0.5386, "step": 6689 }, { "epoch": 0.81, "learning_rate": 1.7682685310293147e-06, "loss": 0.5231, "step": 6690 }, { "epoch": 0.81, "learning_rate": 1.7660327525826603e-06, "loss": 0.4805, "step": 6691 }, { "epoch": 0.81, "learning_rate": 1.7637982515703378e-06, "loss": 0.5159, "step": 6692 }, { "epoch": 0.81, "learning_rate": 1.7615650283390152e-06, "loss": 0.5073, "step": 6693 }, { "epoch": 0.81, "learning_rate": 1.759333083235154e-06, "loss": 0.4552, "step": 6694 }, { "epoch": 0.81, "learning_rate": 1.7571024166050299e-06, "loss": 0.4729, "step": 6695 }, { "epoch": 0.81, "learning_rate": 1.7548730287947057e-06, "loss": 0.5279, "step": 6696 }, { "epoch": 0.81, "learning_rate": 1.7526449201500606e-06, "loss": 0.5316, "step": 6697 }, { "epoch": 0.81, "learning_rate": 1.7504180910167622e-06, "loss": 0.4682, "step": 6698 }, { "epoch": 0.81, "learning_rate": 1.7481925417402856e-06, "loss": 0.5102, "step": 6699 }, { "epoch": 0.81, "learning_rate": 1.745968272665911e-06, "loss": 0.5804, "step": 6700 }, { "epoch": 0.81, "learning_rate": 1.7437452841387138e-06, "loss": 0.5302, "step": 6701 }, { "epoch": 0.81, "learning_rate": 1.7415235765035687e-06, "loss": 0.4987, "step": 6702 }, { "epoch": 0.82, "learning_rate": 1.7393031501051638e-06, "loss": 0.4925, "step": 6703 }, { "epoch": 0.82, "learning_rate": 1.7370840052879723e-06, "loss": 0.5215, "step": 6704 }, { "epoch": 0.82, "learning_rate": 1.734866142396282e-06, "loss": 0.5314, "step": 6705 }, { "epoch": 0.82, "learning_rate": 1.732649561774179e-06, "loss": 0.4188, "step": 6706 }, { "epoch": 0.82, "learning_rate": 1.7304342637655436e-06, "loss": 0.5072, "step": 6707 }, { "epoch": 0.82, "learning_rate": 1.7282202487140608e-06, "loss": 0.524, "step": 6708 }, { "epoch": 0.82, "learning_rate": 1.7260075169632218e-06, "loss": 0.5058, "step": 6709 }, { "epoch": 0.82, "learning_rate": 1.7237960688563116e-06, "loss": 0.4605, "step": 6710 }, { "epoch": 0.82, "learning_rate": 1.7215859047364169e-06, "loss": 0.5327, "step": 6711 }, { "epoch": 0.82, "learning_rate": 1.7193770249464315e-06, "loss": 0.4552, "step": 6712 }, { "epoch": 0.82, "learning_rate": 1.7171694298290399e-06, "loss": 0.5461, "step": 6713 }, { "epoch": 0.82, "learning_rate": 1.714963119726739e-06, "loss": 0.51, "step": 6714 }, { "epoch": 0.82, "learning_rate": 1.7127580949818179e-06, "loss": 0.5341, "step": 6715 }, { "epoch": 0.82, "learning_rate": 1.7105543559363647e-06, "loss": 0.5697, "step": 6716 }, { "epoch": 0.82, "learning_rate": 1.7083519029322758e-06, "loss": 0.5445, "step": 6717 }, { "epoch": 0.82, "learning_rate": 1.7061507363112462e-06, "loss": 0.5041, "step": 6718 }, { "epoch": 0.82, "learning_rate": 1.7039508564147666e-06, "loss": 0.5107, "step": 6719 }, { "epoch": 0.82, "learning_rate": 1.7017522635841278e-06, "loss": 0.4872, "step": 6720 }, { "epoch": 0.82, "learning_rate": 1.6995549581604298e-06, "loss": 0.5573, "step": 6721 }, { "epoch": 0.82, "learning_rate": 1.6973589404845625e-06, "loss": 0.507, "step": 6722 }, { "epoch": 0.82, "learning_rate": 1.6951642108972244e-06, "loss": 0.5646, "step": 6723 }, { "epoch": 0.82, "learning_rate": 1.6929707697389086e-06, "loss": 0.4633, "step": 6724 }, { "epoch": 0.82, "learning_rate": 1.6907786173499064e-06, "loss": 0.5269, "step": 6725 }, { "epoch": 0.82, "learning_rate": 1.6885877540703188e-06, "loss": 0.5007, "step": 6726 }, { "epoch": 0.82, "learning_rate": 1.6863981802400386e-06, "loss": 0.5167, "step": 6727 }, { "epoch": 0.82, "learning_rate": 1.6842098961987552e-06, "loss": 0.5508, "step": 6728 }, { "epoch": 0.82, "learning_rate": 1.6820229022859725e-06, "loss": 0.4825, "step": 6729 }, { "epoch": 0.82, "learning_rate": 1.6798371988409811e-06, "loss": 0.5175, "step": 6730 }, { "epoch": 0.82, "learning_rate": 1.6776527862028723e-06, "loss": 0.5202, "step": 6731 }, { "epoch": 0.82, "learning_rate": 1.6754696647105462e-06, "loss": 0.4892, "step": 6732 }, { "epoch": 0.82, "learning_rate": 1.6732878347026937e-06, "loss": 0.5148, "step": 6733 }, { "epoch": 0.82, "learning_rate": 1.6711072965178067e-06, "loss": 0.5109, "step": 6734 }, { "epoch": 0.82, "learning_rate": 1.6689280504941818e-06, "loss": 0.4951, "step": 6735 }, { "epoch": 0.82, "learning_rate": 1.6667500969699068e-06, "loss": 0.4797, "step": 6736 }, { "epoch": 0.82, "learning_rate": 1.6645734362828803e-06, "loss": 0.51, "step": 6737 }, { "epoch": 0.82, "learning_rate": 1.6623980687707898e-06, "loss": 0.5118, "step": 6738 }, { "epoch": 0.82, "learning_rate": 1.660223994771124e-06, "loss": 0.4981, "step": 6739 }, { "epoch": 0.82, "learning_rate": 1.6580512146211768e-06, "loss": 0.5027, "step": 6740 }, { "epoch": 0.82, "learning_rate": 1.6558797286580386e-06, "loss": 0.5351, "step": 6741 }, { "epoch": 0.82, "learning_rate": 1.6537095372185973e-06, "loss": 0.4516, "step": 6742 }, { "epoch": 0.82, "learning_rate": 1.6515406406395362e-06, "loss": 0.4865, "step": 6743 }, { "epoch": 0.82, "learning_rate": 1.6493730392573493e-06, "loss": 0.5182, "step": 6744 }, { "epoch": 0.82, "learning_rate": 1.6472067334083163e-06, "loss": 0.5149, "step": 6745 }, { "epoch": 0.82, "learning_rate": 1.6450417234285288e-06, "loss": 0.5017, "step": 6746 }, { "epoch": 0.82, "learning_rate": 1.6428780096538666e-06, "loss": 0.485, "step": 6747 }, { "epoch": 0.82, "learning_rate": 1.6407155924200102e-06, "loss": 0.5349, "step": 6748 }, { "epoch": 0.82, "learning_rate": 1.638554472062448e-06, "loss": 0.5049, "step": 6749 }, { "epoch": 0.82, "learning_rate": 1.6363946489164572e-06, "loss": 0.5049, "step": 6750 }, { "epoch": 0.82, "learning_rate": 1.6342361233171145e-06, "loss": 0.508, "step": 6751 }, { "epoch": 0.82, "learning_rate": 1.6320788955992995e-06, "loss": 0.5315, "step": 6752 }, { "epoch": 0.82, "learning_rate": 1.6299229660976933e-06, "loss": 0.531, "step": 6753 }, { "epoch": 0.82, "learning_rate": 1.6277683351467643e-06, "loss": 0.5486, "step": 6754 }, { "epoch": 0.82, "learning_rate": 1.6256150030807939e-06, "loss": 0.5508, "step": 6755 }, { "epoch": 0.82, "learning_rate": 1.623462970233849e-06, "loss": 0.5785, "step": 6756 }, { "epoch": 0.82, "learning_rate": 1.6213122369397983e-06, "loss": 0.5371, "step": 6757 }, { "epoch": 0.82, "learning_rate": 1.6191628035323182e-06, "loss": 0.4699, "step": 6758 }, { "epoch": 0.82, "learning_rate": 1.6170146703448707e-06, "loss": 0.5111, "step": 6759 }, { "epoch": 0.82, "learning_rate": 1.614867837710722e-06, "loss": 0.5203, "step": 6760 }, { "epoch": 0.82, "learning_rate": 1.6127223059629381e-06, "loss": 0.4748, "step": 6761 }, { "epoch": 0.82, "learning_rate": 1.610578075434378e-06, "loss": 0.494, "step": 6762 }, { "epoch": 0.82, "learning_rate": 1.6084351464577065e-06, "loss": 0.4853, "step": 6763 }, { "epoch": 0.82, "learning_rate": 1.6062935193653773e-06, "loss": 0.5113, "step": 6764 }, { "epoch": 0.82, "learning_rate": 1.6041531944896516e-06, "loss": 0.5069, "step": 6765 }, { "epoch": 0.82, "learning_rate": 1.602014172162577e-06, "loss": 0.5665, "step": 6766 }, { "epoch": 0.82, "learning_rate": 1.5998764527160137e-06, "loss": 0.479, "step": 6767 }, { "epoch": 0.82, "learning_rate": 1.5977400364816053e-06, "loss": 0.5636, "step": 6768 }, { "epoch": 0.82, "learning_rate": 1.5956049237908043e-06, "loss": 0.4901, "step": 6769 }, { "epoch": 0.82, "learning_rate": 1.593471114974854e-06, "loss": 0.5642, "step": 6770 }, { "epoch": 0.82, "learning_rate": 1.5913386103647965e-06, "loss": 0.4999, "step": 6771 }, { "epoch": 0.82, "learning_rate": 1.5892074102914767e-06, "loss": 0.4932, "step": 6772 }, { "epoch": 0.82, "learning_rate": 1.5870775150855311e-06, "loss": 0.5103, "step": 6773 }, { "epoch": 0.82, "learning_rate": 1.5849489250773942e-06, "loss": 0.5183, "step": 6774 }, { "epoch": 0.82, "learning_rate": 1.582821640597304e-06, "loss": 0.4681, "step": 6775 }, { "epoch": 0.82, "learning_rate": 1.5806956619752866e-06, "loss": 0.504, "step": 6776 }, { "epoch": 0.82, "learning_rate": 1.5785709895411727e-06, "loss": 0.4921, "step": 6777 }, { "epoch": 0.82, "learning_rate": 1.5764476236245918e-06, "loss": 0.5152, "step": 6778 }, { "epoch": 0.82, "learning_rate": 1.5743255645549649e-06, "loss": 0.5503, "step": 6779 }, { "epoch": 0.82, "learning_rate": 1.5722048126615086e-06, "loss": 0.508, "step": 6780 }, { "epoch": 0.82, "learning_rate": 1.5700853682732476e-06, "loss": 0.5136, "step": 6781 }, { "epoch": 0.82, "learning_rate": 1.567967231718992e-06, "loss": 0.4722, "step": 6782 }, { "epoch": 0.82, "learning_rate": 1.5658504033273536e-06, "loss": 0.5366, "step": 6783 }, { "epoch": 0.82, "learning_rate": 1.563734883426744e-06, "loss": 0.4922, "step": 6784 }, { "epoch": 0.83, "learning_rate": 1.5616206723453665e-06, "loss": 0.5448, "step": 6785 }, { "epoch": 0.83, "learning_rate": 1.5595077704112272e-06, "loss": 0.5197, "step": 6786 }, { "epoch": 0.83, "learning_rate": 1.557396177952125e-06, "loss": 0.5278, "step": 6787 }, { "epoch": 0.83, "learning_rate": 1.555285895295654e-06, "loss": 0.515, "step": 6788 }, { "epoch": 0.83, "learning_rate": 1.5531769227692094e-06, "loss": 0.4285, "step": 6789 }, { "epoch": 0.83, "learning_rate": 1.5510692606999844e-06, "loss": 0.5207, "step": 6790 }, { "epoch": 0.83, "learning_rate": 1.5489629094149605e-06, "loss": 0.5441, "step": 6791 }, { "epoch": 0.83, "learning_rate": 1.5468578692409264e-06, "loss": 0.5045, "step": 6792 }, { "epoch": 0.83, "learning_rate": 1.5447541405044607e-06, "loss": 0.5634, "step": 6793 }, { "epoch": 0.83, "learning_rate": 1.542651723531937e-06, "loss": 0.5265, "step": 6794 }, { "epoch": 0.83, "learning_rate": 1.5405506186495334e-06, "loss": 0.4985, "step": 6795 }, { "epoch": 0.83, "learning_rate": 1.5384508261832165e-06, "loss": 0.4543, "step": 6796 }, { "epoch": 0.83, "learning_rate": 1.5363523464587515e-06, "loss": 0.5187, "step": 6797 }, { "epoch": 0.83, "learning_rate": 1.5342551798017036e-06, "loss": 0.5222, "step": 6798 }, { "epoch": 0.83, "learning_rate": 1.5321593265374314e-06, "loss": 0.4904, "step": 6799 }, { "epoch": 0.83, "learning_rate": 1.5300647869910833e-06, "loss": 0.524, "step": 6800 }, { "epoch": 0.83, "learning_rate": 1.52797156148762e-06, "loss": 0.4815, "step": 6801 }, { "epoch": 0.83, "learning_rate": 1.5258796503517848e-06, "loss": 0.4757, "step": 6802 }, { "epoch": 0.83, "learning_rate": 1.523789053908119e-06, "loss": 0.4985, "step": 6803 }, { "epoch": 0.83, "learning_rate": 1.5216997724809646e-06, "loss": 0.4995, "step": 6804 }, { "epoch": 0.83, "learning_rate": 1.5196118063944576e-06, "loss": 0.5107, "step": 6805 }, { "epoch": 0.83, "learning_rate": 1.5175251559725246e-06, "loss": 0.5116, "step": 6806 }, { "epoch": 0.83, "learning_rate": 1.5154398215388977e-06, "loss": 0.6089, "step": 6807 }, { "epoch": 0.83, "learning_rate": 1.5133558034170958e-06, "loss": 0.4679, "step": 6808 }, { "epoch": 0.83, "learning_rate": 1.5112731019304428e-06, "loss": 0.5176, "step": 6809 }, { "epoch": 0.83, "learning_rate": 1.50919171740205e-06, "loss": 0.5172, "step": 6810 }, { "epoch": 0.83, "learning_rate": 1.5071116501548256e-06, "loss": 0.4696, "step": 6811 }, { "epoch": 0.83, "learning_rate": 1.5050329005114771e-06, "loss": 0.478, "step": 6812 }, { "epoch": 0.83, "learning_rate": 1.5029554687945092e-06, "loss": 0.5558, "step": 6813 }, { "epoch": 0.83, "learning_rate": 1.5008793553262147e-06, "loss": 0.512, "step": 6814 }, { "epoch": 0.83, "learning_rate": 1.4988045604286893e-06, "loss": 0.4861, "step": 6815 }, { "epoch": 0.83, "learning_rate": 1.4967310844238192e-06, "loss": 0.5296, "step": 6816 }, { "epoch": 0.83, "learning_rate": 1.4946589276332857e-06, "loss": 0.5344, "step": 6817 }, { "epoch": 0.83, "learning_rate": 1.4925880903785717e-06, "loss": 0.4695, "step": 6818 }, { "epoch": 0.83, "learning_rate": 1.490518572980949e-06, "loss": 0.5152, "step": 6819 }, { "epoch": 0.83, "learning_rate": 1.4884503757614832e-06, "loss": 0.5462, "step": 6820 }, { "epoch": 0.83, "learning_rate": 1.486383499041044e-06, "loss": 0.5143, "step": 6821 }, { "epoch": 0.83, "learning_rate": 1.48431794314029e-06, "loss": 0.5296, "step": 6822 }, { "epoch": 0.83, "learning_rate": 1.4822537083796706e-06, "loss": 0.4644, "step": 6823 }, { "epoch": 0.83, "learning_rate": 1.4801907950794402e-06, "loss": 0.5366, "step": 6824 }, { "epoch": 0.83, "learning_rate": 1.478129203559644e-06, "loss": 0.5564, "step": 6825 }, { "epoch": 0.83, "learning_rate": 1.476068934140118e-06, "loss": 0.4664, "step": 6826 }, { "epoch": 0.83, "learning_rate": 1.4740099871405e-06, "loss": 0.526, "step": 6827 }, { "epoch": 0.83, "learning_rate": 1.4719523628802178e-06, "loss": 0.4793, "step": 6828 }, { "epoch": 0.83, "learning_rate": 1.469896061678493e-06, "loss": 0.4967, "step": 6829 }, { "epoch": 0.83, "learning_rate": 1.4678410838543477e-06, "loss": 0.4848, "step": 6830 }, { "epoch": 0.83, "learning_rate": 1.4657874297265927e-06, "loss": 0.5088, "step": 6831 }, { "epoch": 0.83, "learning_rate": 1.46373509961384e-06, "loss": 0.4753, "step": 6832 }, { "epoch": 0.83, "learning_rate": 1.461684093834489e-06, "loss": 0.4687, "step": 6833 }, { "epoch": 0.83, "learning_rate": 1.4596344127067351e-06, "loss": 0.4658, "step": 6834 }, { "epoch": 0.83, "learning_rate": 1.4575860565485754e-06, "loss": 0.4647, "step": 6835 }, { "epoch": 0.83, "learning_rate": 1.4555390256777912e-06, "loss": 0.5543, "step": 6836 }, { "epoch": 0.83, "learning_rate": 1.4534933204119638e-06, "loss": 0.5768, "step": 6837 }, { "epoch": 0.83, "learning_rate": 1.4514489410684718e-06, "loss": 0.5511, "step": 6838 }, { "epoch": 0.83, "learning_rate": 1.4494058879644812e-06, "loss": 0.4297, "step": 6839 }, { "epoch": 0.83, "learning_rate": 1.4473641614169543e-06, "loss": 0.4733, "step": 6840 }, { "epoch": 0.83, "learning_rate": 1.4453237617426519e-06, "loss": 0.4937, "step": 6841 }, { "epoch": 0.83, "learning_rate": 1.4432846892581242e-06, "loss": 0.4864, "step": 6842 }, { "epoch": 0.83, "learning_rate": 1.441246944279715e-06, "loss": 0.4911, "step": 6843 }, { "epoch": 0.83, "learning_rate": 1.4392105271235678e-06, "loss": 0.5535, "step": 6844 }, { "epoch": 0.83, "learning_rate": 1.4371754381056147e-06, "loss": 0.5295, "step": 6845 }, { "epoch": 0.83, "learning_rate": 1.4351416775415816e-06, "loss": 0.5198, "step": 6846 }, { "epoch": 0.83, "learning_rate": 1.4331092457469942e-06, "loss": 0.4819, "step": 6847 }, { "epoch": 0.83, "learning_rate": 1.4310781430371646e-06, "loss": 0.4934, "step": 6848 }, { "epoch": 0.83, "learning_rate": 1.4290483697272028e-06, "loss": 0.4752, "step": 6849 }, { "epoch": 0.83, "learning_rate": 1.427019926132015e-06, "loss": 0.5094, "step": 6850 }, { "epoch": 0.83, "learning_rate": 1.424992812566297e-06, "loss": 0.5436, "step": 6851 }, { "epoch": 0.83, "learning_rate": 1.4229670293445342e-06, "loss": 0.56, "step": 6852 }, { "epoch": 0.83, "learning_rate": 1.420942576781018e-06, "loss": 0.5349, "step": 6853 }, { "epoch": 0.83, "learning_rate": 1.41891945518982e-06, "loss": 0.4966, "step": 6854 }, { "epoch": 0.83, "learning_rate": 1.4168976648848165e-06, "loss": 0.5145, "step": 6855 }, { "epoch": 0.83, "learning_rate": 1.41487720617967e-06, "loss": 0.4776, "step": 6856 }, { "epoch": 0.83, "learning_rate": 1.4128580793878356e-06, "loss": 0.4942, "step": 6857 }, { "epoch": 0.83, "learning_rate": 1.41084028482257e-06, "loss": 0.526, "step": 6858 }, { "epoch": 0.83, "learning_rate": 1.4088238227969153e-06, "loss": 0.5405, "step": 6859 }, { "epoch": 0.83, "learning_rate": 1.4068086936237046e-06, "loss": 0.5537, "step": 6860 }, { "epoch": 0.83, "learning_rate": 1.404794897615579e-06, "loss": 0.5308, "step": 6861 }, { "epoch": 0.83, "learning_rate": 1.4027824350849573e-06, "loss": 0.5754, "step": 6862 }, { "epoch": 0.83, "learning_rate": 1.4007713063440553e-06, "loss": 0.4477, "step": 6863 }, { "epoch": 0.83, "learning_rate": 1.3987615117048893e-06, "loss": 0.5302, "step": 6864 }, { "epoch": 0.83, "learning_rate": 1.3967530514792582e-06, "loss": 0.508, "step": 6865 }, { "epoch": 0.83, "learning_rate": 1.3947459259787589e-06, "loss": 0.5448, "step": 6866 }, { "epoch": 0.84, "learning_rate": 1.392740135514783e-06, "loss": 0.5593, "step": 6867 }, { "epoch": 0.84, "learning_rate": 1.3907356803985118e-06, "loss": 0.493, "step": 6868 }, { "epoch": 0.84, "learning_rate": 1.3887325609409197e-06, "loss": 0.4995, "step": 6869 }, { "epoch": 0.84, "learning_rate": 1.386730777452776e-06, "loss": 0.5224, "step": 6870 }, { "epoch": 0.84, "learning_rate": 1.3847303302446392e-06, "loss": 0.4626, "step": 6871 }, { "epoch": 0.84, "learning_rate": 1.382731219626865e-06, "loss": 0.4996, "step": 6872 }, { "epoch": 0.84, "learning_rate": 1.3807334459096012e-06, "loss": 0.5219, "step": 6873 }, { "epoch": 0.84, "learning_rate": 1.3787370094027842e-06, "loss": 0.5251, "step": 6874 }, { "epoch": 0.84, "learning_rate": 1.3767419104161416e-06, "loss": 0.5121, "step": 6875 }, { "epoch": 0.84, "learning_rate": 1.3747481492592042e-06, "loss": 0.5241, "step": 6876 }, { "epoch": 0.84, "learning_rate": 1.3727557262412828e-06, "loss": 0.4999, "step": 6877 }, { "epoch": 0.84, "learning_rate": 1.3707646416714892e-06, "loss": 0.5259, "step": 6878 }, { "epoch": 0.84, "learning_rate": 1.3687748958587233e-06, "loss": 0.4637, "step": 6879 }, { "epoch": 0.84, "learning_rate": 1.3667864891116745e-06, "loss": 0.5261, "step": 6880 }, { "epoch": 0.84, "learning_rate": 1.364799421738835e-06, "loss": 0.5663, "step": 6881 }, { "epoch": 0.84, "learning_rate": 1.3628136940484793e-06, "loss": 0.5064, "step": 6882 }, { "epoch": 0.84, "learning_rate": 1.3608293063486732e-06, "loss": 0.5391, "step": 6883 }, { "epoch": 0.84, "learning_rate": 1.3588462589472838e-06, "loss": 0.4809, "step": 6884 }, { "epoch": 0.84, "learning_rate": 1.3568645521519652e-06, "loss": 0.4816, "step": 6885 }, { "epoch": 0.84, "learning_rate": 1.35488418627016e-06, "loss": 0.5349, "step": 6886 }, { "epoch": 0.84, "learning_rate": 1.3529051616091094e-06, "loss": 0.5448, "step": 6887 }, { "epoch": 0.84, "learning_rate": 1.350927478475843e-06, "loss": 0.5156, "step": 6888 }, { "epoch": 0.84, "learning_rate": 1.348951137177179e-06, "loss": 0.5623, "step": 6889 }, { "epoch": 0.84, "learning_rate": 1.3469761380197355e-06, "loss": 0.4732, "step": 6890 }, { "epoch": 0.84, "learning_rate": 1.345002481309915e-06, "loss": 0.5426, "step": 6891 }, { "epoch": 0.84, "learning_rate": 1.3430301673539149e-06, "loss": 0.4643, "step": 6892 }, { "epoch": 0.84, "learning_rate": 1.3410591964577247e-06, "loss": 0.4792, "step": 6893 }, { "epoch": 0.84, "learning_rate": 1.3390895689271244e-06, "loss": 0.4875, "step": 6894 }, { "epoch": 0.84, "learning_rate": 1.3371212850676874e-06, "loss": 0.5221, "step": 6895 }, { "epoch": 0.84, "learning_rate": 1.3351543451847737e-06, "loss": 0.4525, "step": 6896 }, { "epoch": 0.84, "learning_rate": 1.3331887495835428e-06, "loss": 0.4724, "step": 6897 }, { "epoch": 0.84, "learning_rate": 1.3312244985689372e-06, "loss": 0.5417, "step": 6898 }, { "epoch": 0.84, "learning_rate": 1.3292615924456987e-06, "loss": 0.5471, "step": 6899 }, { "epoch": 0.84, "learning_rate": 1.3273000315183526e-06, "loss": 0.5519, "step": 6900 }, { "epoch": 0.84, "learning_rate": 1.3253398160912223e-06, "loss": 0.4653, "step": 6901 }, { "epoch": 0.84, "learning_rate": 1.32338094646842e-06, "loss": 0.5403, "step": 6902 }, { "epoch": 0.84, "learning_rate": 1.3214234229538447e-06, "loss": 0.55, "step": 6903 }, { "epoch": 0.84, "learning_rate": 1.3194672458511947e-06, "loss": 0.5171, "step": 6904 }, { "epoch": 0.84, "learning_rate": 1.317512415463954e-06, "loss": 0.4656, "step": 6905 }, { "epoch": 0.84, "learning_rate": 1.3155589320953965e-06, "loss": 0.4805, "step": 6906 }, { "epoch": 0.84, "learning_rate": 1.3136067960485944e-06, "loss": 0.4862, "step": 6907 }, { "epoch": 0.84, "learning_rate": 1.3116560076264006e-06, "loss": 0.4652, "step": 6908 }, { "epoch": 0.84, "learning_rate": 1.3097065671314669e-06, "loss": 0.5117, "step": 6909 }, { "epoch": 0.84, "learning_rate": 1.3077584748662364e-06, "loss": 0.4822, "step": 6910 }, { "epoch": 0.84, "learning_rate": 1.3058117311329367e-06, "loss": 0.508, "step": 6911 }, { "epoch": 0.84, "learning_rate": 1.303866336233589e-06, "loss": 0.4788, "step": 6912 }, { "epoch": 0.84, "learning_rate": 1.301922290470008e-06, "loss": 0.5295, "step": 6913 }, { "epoch": 0.84, "learning_rate": 1.2999795941437965e-06, "loss": 0.5234, "step": 6914 }, { "epoch": 0.84, "learning_rate": 1.2980382475563446e-06, "loss": 0.5007, "step": 6915 }, { "epoch": 0.84, "learning_rate": 1.2960982510088427e-06, "loss": 0.5103, "step": 6916 }, { "epoch": 0.84, "learning_rate": 1.294159604802261e-06, "loss": 0.5229, "step": 6917 }, { "epoch": 0.84, "learning_rate": 1.2922223092373687e-06, "loss": 0.558, "step": 6918 }, { "epoch": 0.84, "learning_rate": 1.2902863646147202e-06, "loss": 0.4676, "step": 6919 }, { "epoch": 0.84, "learning_rate": 1.2883517712346593e-06, "loss": 0.4256, "step": 6920 }, { "epoch": 0.84, "learning_rate": 1.2864185293973252e-06, "loss": 0.4962, "step": 6921 }, { "epoch": 0.84, "learning_rate": 1.2844866394026467e-06, "loss": 0.4811, "step": 6922 }, { "epoch": 0.84, "learning_rate": 1.2825561015503397e-06, "loss": 0.507, "step": 6923 }, { "epoch": 0.84, "learning_rate": 1.2806269161399088e-06, "loss": 0.5388, "step": 6924 }, { "epoch": 0.84, "learning_rate": 1.278699083470657e-06, "loss": 0.5028, "step": 6925 }, { "epoch": 0.84, "learning_rate": 1.2767726038416662e-06, "loss": 0.5363, "step": 6926 }, { "epoch": 0.84, "learning_rate": 1.2748474775518204e-06, "loss": 0.4441, "step": 6927 }, { "epoch": 0.84, "learning_rate": 1.2729237048997833e-06, "loss": 0.4497, "step": 6928 }, { "epoch": 0.84, "learning_rate": 1.2710012861840137e-06, "loss": 0.5183, "step": 6929 }, { "epoch": 0.84, "learning_rate": 1.269080221702762e-06, "loss": 0.4931, "step": 6930 }, { "epoch": 0.84, "learning_rate": 1.2671605117540642e-06, "loss": 0.4888, "step": 6931 }, { "epoch": 0.84, "learning_rate": 1.2652421566357431e-06, "loss": 0.5331, "step": 6932 }, { "epoch": 0.84, "learning_rate": 1.2633251566454252e-06, "loss": 0.4981, "step": 6933 }, { "epoch": 0.84, "learning_rate": 1.2614095120805138e-06, "loss": 0.5331, "step": 6934 }, { "epoch": 0.84, "learning_rate": 1.259495223238204e-06, "loss": 0.5048, "step": 6935 }, { "epoch": 0.84, "learning_rate": 1.2575822904154855e-06, "loss": 0.589, "step": 6936 }, { "epoch": 0.84, "learning_rate": 1.2556707139091328e-06, "loss": 0.4599, "step": 6937 }, { "epoch": 0.84, "learning_rate": 1.253760494015711e-06, "loss": 0.5395, "step": 6938 }, { "epoch": 0.84, "learning_rate": 1.2518516310315787e-06, "loss": 0.4739, "step": 6939 }, { "epoch": 0.84, "learning_rate": 1.2499441252528755e-06, "loss": 0.5041, "step": 6940 }, { "epoch": 0.84, "learning_rate": 1.2480379769755413e-06, "loss": 0.5264, "step": 6941 }, { "epoch": 0.84, "learning_rate": 1.2461331864952986e-06, "loss": 0.5199, "step": 6942 }, { "epoch": 0.84, "learning_rate": 1.2442297541076554e-06, "loss": 0.481, "step": 6943 }, { "epoch": 0.84, "learning_rate": 1.2423276801079188e-06, "loss": 0.5131, "step": 6944 }, { "epoch": 0.84, "learning_rate": 1.2404269647911815e-06, "loss": 0.514, "step": 6945 }, { "epoch": 0.84, "learning_rate": 1.2385276084523212e-06, "loss": 0.5136, "step": 6946 }, { "epoch": 0.84, "learning_rate": 1.2366296113860076e-06, "loss": 0.4926, "step": 6947 }, { "epoch": 0.84, "learning_rate": 1.2347329738867042e-06, "loss": 0.5248, "step": 6948 }, { "epoch": 0.85, "learning_rate": 1.2328376962486533e-06, "loss": 0.4346, "step": 6949 }, { "epoch": 0.85, "learning_rate": 1.2309437787658962e-06, "loss": 0.4765, "step": 6950 }, { "epoch": 0.85, "learning_rate": 1.2290512217322591e-06, "loss": 0.5134, "step": 6951 }, { "epoch": 0.85, "learning_rate": 1.2271600254413529e-06, "loss": 0.5735, "step": 6952 }, { "epoch": 0.85, "learning_rate": 1.2252701901865872e-06, "loss": 0.511, "step": 6953 }, { "epoch": 0.85, "learning_rate": 1.2233817162611517e-06, "loss": 0.5923, "step": 6954 }, { "epoch": 0.85, "learning_rate": 1.2214946039580268e-06, "loss": 0.5259, "step": 6955 }, { "epoch": 0.85, "learning_rate": 1.2196088535699845e-06, "loss": 0.4935, "step": 6956 }, { "epoch": 0.85, "learning_rate": 1.217724465389587e-06, "loss": 0.5304, "step": 6957 }, { "epoch": 0.85, "learning_rate": 1.2158414397091766e-06, "loss": 0.5294, "step": 6958 }, { "epoch": 0.85, "learning_rate": 1.2139597768208943e-06, "loss": 0.5541, "step": 6959 }, { "epoch": 0.85, "learning_rate": 1.2120794770166632e-06, "loss": 0.5272, "step": 6960 }, { "epoch": 0.85, "learning_rate": 1.2102005405881944e-06, "loss": 0.4743, "step": 6961 }, { "epoch": 0.85, "learning_rate": 1.2083229678269948e-06, "loss": 0.4969, "step": 6962 }, { "epoch": 0.85, "learning_rate": 1.2064467590243523e-06, "loss": 0.4828, "step": 6963 }, { "epoch": 0.85, "learning_rate": 1.2045719144713442e-06, "loss": 0.5033, "step": 6964 }, { "epoch": 0.85, "learning_rate": 1.2026984344588398e-06, "loss": 0.4379, "step": 6965 }, { "epoch": 0.85, "learning_rate": 1.2008263192774927e-06, "loss": 0.5463, "step": 6966 }, { "epoch": 0.85, "learning_rate": 1.19895556921775e-06, "loss": 0.5328, "step": 6967 }, { "epoch": 0.85, "learning_rate": 1.1970861845698402e-06, "loss": 0.482, "step": 6968 }, { "epoch": 0.85, "learning_rate": 1.1952181656237849e-06, "loss": 0.493, "step": 6969 }, { "epoch": 0.85, "learning_rate": 1.1933515126693917e-06, "loss": 0.4697, "step": 6970 }, { "epoch": 0.85, "learning_rate": 1.191486225996259e-06, "loss": 0.5212, "step": 6971 }, { "epoch": 0.85, "learning_rate": 1.1896223058937683e-06, "loss": 0.5492, "step": 6972 }, { "epoch": 0.85, "learning_rate": 1.1877597526510943e-06, "loss": 0.5006, "step": 6973 }, { "epoch": 0.85, "learning_rate": 1.1858985665571965e-06, "loss": 0.533, "step": 6974 }, { "epoch": 0.85, "learning_rate": 1.1840387479008209e-06, "loss": 0.4812, "step": 6975 }, { "epoch": 0.85, "learning_rate": 1.1821802969705066e-06, "loss": 0.5429, "step": 6976 }, { "epoch": 0.85, "learning_rate": 1.1803232140545774e-06, "loss": 0.5359, "step": 6977 }, { "epoch": 0.85, "learning_rate": 1.1784674994411404e-06, "loss": 0.487, "step": 6978 }, { "epoch": 0.85, "learning_rate": 1.1766131534181012e-06, "loss": 0.4423, "step": 6979 }, { "epoch": 0.85, "learning_rate": 1.174760176273142e-06, "loss": 0.4937, "step": 6980 }, { "epoch": 0.85, "learning_rate": 1.1729085682937391e-06, "loss": 0.5203, "step": 6981 }, { "epoch": 0.85, "learning_rate": 1.1710583297671563e-06, "loss": 0.5455, "step": 6982 }, { "epoch": 0.85, "learning_rate": 1.1692094609804417e-06, "loss": 0.5097, "step": 6983 }, { "epoch": 0.85, "learning_rate": 1.1673619622204313e-06, "loss": 0.4364, "step": 6984 }, { "epoch": 0.85, "learning_rate": 1.1655158337737538e-06, "loss": 0.5587, "step": 6985 }, { "epoch": 0.85, "learning_rate": 1.1636710759268177e-06, "loss": 0.5032, "step": 6986 }, { "epoch": 0.85, "learning_rate": 1.161827688965822e-06, "loss": 0.4665, "step": 6987 }, { "epoch": 0.85, "learning_rate": 1.159985673176758e-06, "loss": 0.458, "step": 6988 }, { "epoch": 0.85, "learning_rate": 1.1581450288453934e-06, "loss": 0.5256, "step": 6989 }, { "epoch": 0.85, "learning_rate": 1.1563057562572943e-06, "loss": 0.4993, "step": 6990 }, { "epoch": 0.85, "learning_rate": 1.1544678556978096e-06, "loss": 0.5361, "step": 6991 }, { "epoch": 0.85, "learning_rate": 1.1526313274520695e-06, "loss": 0.5081, "step": 6992 }, { "epoch": 0.85, "learning_rate": 1.1507961718050009e-06, "loss": 0.5325, "step": 6993 }, { "epoch": 0.85, "learning_rate": 1.1489623890413148e-06, "loss": 0.4745, "step": 6994 }, { "epoch": 0.85, "learning_rate": 1.1471299794455038e-06, "loss": 0.496, "step": 6995 }, { "epoch": 0.85, "learning_rate": 1.1452989433018558e-06, "loss": 0.556, "step": 6996 }, { "epoch": 0.85, "learning_rate": 1.1434692808944404e-06, "loss": 0.5484, "step": 6997 }, { "epoch": 0.85, "learning_rate": 1.1416409925071115e-06, "loss": 0.4979, "step": 6998 }, { "epoch": 0.85, "learning_rate": 1.1398140784235179e-06, "loss": 0.5643, "step": 6999 }, { "epoch": 0.85, "learning_rate": 1.1379885389270895e-06, "loss": 0.5148, "step": 7000 }, { "epoch": 0.85, "learning_rate": 1.136164374301042e-06, "loss": 0.5267, "step": 7001 }, { "epoch": 0.85, "learning_rate": 1.1343415848283834e-06, "loss": 0.4517, "step": 7002 }, { "epoch": 0.85, "learning_rate": 1.132520170791903e-06, "loss": 0.4821, "step": 7003 }, { "epoch": 0.85, "learning_rate": 1.130700132474175e-06, "loss": 0.4418, "step": 7004 }, { "epoch": 0.85, "learning_rate": 1.1288814701575723e-06, "loss": 0.4997, "step": 7005 }, { "epoch": 0.85, "learning_rate": 1.1270641841242414e-06, "loss": 0.4839, "step": 7006 }, { "epoch": 0.85, "learning_rate": 1.1252482746561166e-06, "loss": 0.4507, "step": 7007 }, { "epoch": 0.85, "learning_rate": 1.1234337420349284e-06, "loss": 0.5222, "step": 7008 }, { "epoch": 0.85, "learning_rate": 1.1216205865421825e-06, "loss": 0.487, "step": 7009 }, { "epoch": 0.85, "learning_rate": 1.119808808459174e-06, "loss": 0.5283, "step": 7010 }, { "epoch": 0.85, "learning_rate": 1.1179984080669915e-06, "loss": 0.5287, "step": 7011 }, { "epoch": 0.85, "learning_rate": 1.1161893856464978e-06, "loss": 0.5633, "step": 7012 }, { "epoch": 0.85, "learning_rate": 1.1143817414783543e-06, "loss": 0.4698, "step": 7013 }, { "epoch": 0.85, "learning_rate": 1.1125754758429996e-06, "loss": 0.4976, "step": 7014 }, { "epoch": 0.85, "learning_rate": 1.11077058902066e-06, "loss": 0.5085, "step": 7015 }, { "epoch": 0.85, "learning_rate": 1.10896708129135e-06, "loss": 0.5397, "step": 7016 }, { "epoch": 0.85, "learning_rate": 1.1071649529348727e-06, "loss": 0.533, "step": 7017 }, { "epoch": 0.85, "learning_rate": 1.1053642042308089e-06, "loss": 0.472, "step": 7018 }, { "epoch": 0.85, "learning_rate": 1.103564835458536e-06, "loss": 0.5152, "step": 7019 }, { "epoch": 0.85, "learning_rate": 1.1017668468972076e-06, "loss": 0.5014, "step": 7020 }, { "epoch": 0.85, "learning_rate": 1.0999702388257672e-06, "loss": 0.4818, "step": 7021 }, { "epoch": 0.85, "learning_rate": 1.0981750115229474e-06, "loss": 0.471, "step": 7022 }, { "epoch": 0.85, "learning_rate": 1.0963811652672617e-06, "loss": 0.4688, "step": 7023 }, { "epoch": 0.85, "learning_rate": 1.094588700337008e-06, "loss": 0.59, "step": 7024 }, { "epoch": 0.85, "learning_rate": 1.0927976170102784e-06, "loss": 0.4396, "step": 7025 }, { "epoch": 0.85, "learning_rate": 1.0910079155649422e-06, "loss": 0.4959, "step": 7026 }, { "epoch": 0.85, "learning_rate": 1.0892195962786557e-06, "loss": 0.4798, "step": 7027 }, { "epoch": 0.85, "learning_rate": 1.0874326594288642e-06, "loss": 0.4737, "step": 7028 }, { "epoch": 0.85, "learning_rate": 1.0856471052927987e-06, "loss": 0.5458, "step": 7029 }, { "epoch": 0.85, "learning_rate": 1.0838629341474694e-06, "loss": 0.473, "step": 7030 }, { "epoch": 0.86, "learning_rate": 1.0820801462696806e-06, "loss": 0.4783, "step": 7031 }, { "epoch": 0.86, "learning_rate": 1.0802987419360155e-06, "loss": 0.5272, "step": 7032 }, { "epoch": 0.86, "learning_rate": 1.0785187214228433e-06, "loss": 0.5412, "step": 7033 }, { "epoch": 0.86, "learning_rate": 1.0767400850063236e-06, "loss": 0.4677, "step": 7034 }, { "epoch": 0.86, "learning_rate": 1.0749628329623928e-06, "loss": 0.5681, "step": 7035 }, { "epoch": 0.86, "learning_rate": 1.0731869655667836e-06, "loss": 0.4613, "step": 7036 }, { "epoch": 0.86, "learning_rate": 1.0714124830950035e-06, "loss": 0.4916, "step": 7037 }, { "epoch": 0.86, "learning_rate": 1.0696393858223476e-06, "loss": 0.4251, "step": 7038 }, { "epoch": 0.86, "learning_rate": 1.0678676740239024e-06, "loss": 0.544, "step": 7039 }, { "epoch": 0.86, "learning_rate": 1.0660973479745307e-06, "loss": 0.5531, "step": 7040 }, { "epoch": 0.86, "learning_rate": 1.0643284079488858e-06, "loss": 0.4751, "step": 7041 }, { "epoch": 0.86, "learning_rate": 1.0625608542214084e-06, "loss": 0.4846, "step": 7042 }, { "epoch": 0.86, "learning_rate": 1.0607946870663165e-06, "loss": 0.5027, "step": 7043 }, { "epoch": 0.86, "learning_rate": 1.059029906757616e-06, "loss": 0.478, "step": 7044 }, { "epoch": 0.86, "learning_rate": 1.0572665135691017e-06, "loss": 0.5181, "step": 7045 }, { "epoch": 0.86, "learning_rate": 1.0555045077743476e-06, "loss": 0.5515, "step": 7046 }, { "epoch": 0.86, "learning_rate": 1.0537438896467146e-06, "loss": 0.5581, "step": 7047 }, { "epoch": 0.86, "learning_rate": 1.0519846594593519e-06, "loss": 0.5278, "step": 7048 }, { "epoch": 0.86, "learning_rate": 1.0502268174851871e-06, "loss": 0.5579, "step": 7049 }, { "epoch": 0.86, "learning_rate": 1.048470363996933e-06, "loss": 0.5232, "step": 7050 }, { "epoch": 0.86, "learning_rate": 1.046715299267096e-06, "loss": 0.5147, "step": 7051 }, { "epoch": 0.86, "learning_rate": 1.0449616235679538e-06, "loss": 0.4966, "step": 7052 }, { "epoch": 0.86, "learning_rate": 1.0432093371715779e-06, "loss": 0.537, "step": 7053 }, { "epoch": 0.86, "learning_rate": 1.041458440349824e-06, "loss": 0.5264, "step": 7054 }, { "epoch": 0.86, "learning_rate": 1.039708933374327e-06, "loss": 0.5121, "step": 7055 }, { "epoch": 0.86, "learning_rate": 1.0379608165165078e-06, "loss": 0.4865, "step": 7056 }, { "epoch": 0.86, "learning_rate": 1.036214090047577e-06, "loss": 0.4101, "step": 7057 }, { "epoch": 0.86, "learning_rate": 1.0344687542385202e-06, "loss": 0.5101, "step": 7058 }, { "epoch": 0.86, "learning_rate": 1.032724809360116e-06, "loss": 0.5076, "step": 7059 }, { "epoch": 0.86, "learning_rate": 1.030982255682923e-06, "loss": 0.5137, "step": 7060 }, { "epoch": 0.86, "learning_rate": 1.0292410934772822e-06, "loss": 0.522, "step": 7061 }, { "epoch": 0.86, "learning_rate": 1.0275013230133246e-06, "loss": 0.4939, "step": 7062 }, { "epoch": 0.86, "learning_rate": 1.0257629445609597e-06, "loss": 0.5222, "step": 7063 }, { "epoch": 0.86, "learning_rate": 1.0240259583898793e-06, "loss": 0.5173, "step": 7064 }, { "epoch": 0.86, "learning_rate": 1.0222903647695704e-06, "loss": 0.4715, "step": 7065 }, { "epoch": 0.86, "learning_rate": 1.020556163969293e-06, "loss": 0.4782, "step": 7066 }, { "epoch": 0.86, "learning_rate": 1.018823356258093e-06, "loss": 0.5613, "step": 7067 }, { "epoch": 0.86, "learning_rate": 1.0170919419048043e-06, "loss": 0.5017, "step": 7068 }, { "epoch": 0.86, "learning_rate": 1.015361921178042e-06, "loss": 0.5468, "step": 7069 }, { "epoch": 0.86, "learning_rate": 1.0136332943462001e-06, "loss": 0.5456, "step": 7070 }, { "epoch": 0.86, "learning_rate": 1.0119060616774689e-06, "loss": 0.5505, "step": 7071 }, { "epoch": 0.86, "learning_rate": 1.0101802234398094e-06, "loss": 0.5713, "step": 7072 }, { "epoch": 0.86, "learning_rate": 1.008455779900972e-06, "loss": 0.5422, "step": 7073 }, { "epoch": 0.86, "learning_rate": 1.0067327313284936e-06, "loss": 0.4533, "step": 7074 }, { "epoch": 0.86, "learning_rate": 1.005011077989686e-06, "loss": 0.5178, "step": 7075 }, { "epoch": 0.86, "learning_rate": 1.0032908201516544e-06, "loss": 0.4756, "step": 7076 }, { "epoch": 0.86, "learning_rate": 1.0015719580812845e-06, "loss": 0.525, "step": 7077 }, { "epoch": 0.86, "learning_rate": 9.998544920452403e-07, "loss": 0.5316, "step": 7078 }, { "epoch": 0.86, "learning_rate": 9.981384223099733e-07, "loss": 0.5674, "step": 7079 }, { "epoch": 0.86, "learning_rate": 9.964237491417194e-07, "loss": 0.4793, "step": 7080 }, { "epoch": 0.86, "learning_rate": 9.947104728064949e-07, "loss": 0.5007, "step": 7081 }, { "epoch": 0.86, "learning_rate": 9.929985935701048e-07, "loss": 0.5089, "step": 7082 }, { "epoch": 0.86, "learning_rate": 9.9128811169813e-07, "loss": 0.5907, "step": 7083 }, { "epoch": 0.86, "learning_rate": 9.895790274559359e-07, "loss": 0.4901, "step": 7084 }, { "epoch": 0.86, "learning_rate": 9.878713411086794e-07, "loss": 0.4604, "step": 7085 }, { "epoch": 0.86, "learning_rate": 9.861650529212907e-07, "loss": 0.5487, "step": 7086 }, { "epoch": 0.86, "learning_rate": 9.84460163158486e-07, "loss": 0.4283, "step": 7087 }, { "epoch": 0.86, "learning_rate": 9.827566720847659e-07, "loss": 0.5431, "step": 7088 }, { "epoch": 0.86, "learning_rate": 9.810545799644166e-07, "loss": 0.4672, "step": 7089 }, { "epoch": 0.86, "learning_rate": 9.793538870615005e-07, "loss": 0.4716, "step": 7090 }, { "epoch": 0.86, "learning_rate": 9.776545936398685e-07, "loss": 0.4851, "step": 7091 }, { "epoch": 0.86, "learning_rate": 9.759566999631531e-07, "loss": 0.4767, "step": 7092 }, { "epoch": 0.86, "learning_rate": 9.742602062947647e-07, "loss": 0.5185, "step": 7093 }, { "epoch": 0.86, "learning_rate": 9.725651128979052e-07, "loss": 0.4787, "step": 7094 }, { "epoch": 0.86, "learning_rate": 9.70871420035554e-07, "loss": 0.4492, "step": 7095 }, { "epoch": 0.86, "learning_rate": 9.69179127970471e-07, "loss": 0.508, "step": 7096 }, { "epoch": 0.86, "learning_rate": 9.67488236965206e-07, "loss": 0.495, "step": 7097 }, { "epoch": 0.86, "learning_rate": 9.65798747282084e-07, "loss": 0.4904, "step": 7098 }, { "epoch": 0.86, "learning_rate": 9.641106591832183e-07, "loss": 0.5094, "step": 7099 }, { "epoch": 0.86, "learning_rate": 9.624239729304996e-07, "loss": 0.4847, "step": 7100 }, { "epoch": 0.86, "learning_rate": 9.607386887856073e-07, "loss": 0.4833, "step": 7101 }, { "epoch": 0.86, "learning_rate": 9.59054807009997e-07, "loss": 0.5384, "step": 7102 }, { "epoch": 0.86, "learning_rate": 9.573723278649105e-07, "loss": 0.5074, "step": 7103 }, { "epoch": 0.86, "learning_rate": 9.556912516113703e-07, "loss": 0.5023, "step": 7104 }, { "epoch": 0.86, "learning_rate": 9.540115785101844e-07, "loss": 0.5158, "step": 7105 }, { "epoch": 0.86, "learning_rate": 9.523333088219388e-07, "loss": 0.4769, "step": 7106 }, { "epoch": 0.86, "learning_rate": 9.50656442807002e-07, "loss": 0.5353, "step": 7107 }, { "epoch": 0.86, "learning_rate": 9.489809807255302e-07, "loss": 0.5723, "step": 7108 }, { "epoch": 0.86, "learning_rate": 9.473069228374554e-07, "loss": 0.5194, "step": 7109 }, { "epoch": 0.86, "learning_rate": 9.456342694024933e-07, "loss": 0.5043, "step": 7110 }, { "epoch": 0.86, "learning_rate": 9.439630206801453e-07, "loss": 0.513, "step": 7111 }, { "epoch": 0.86, "learning_rate": 9.422931769296905e-07, "loss": 0.5974, "step": 7112 }, { "epoch": 0.86, "learning_rate": 9.406247384101919e-07, "loss": 0.5233, "step": 7113 }, { "epoch": 0.87, "learning_rate": 9.389577053804977e-07, "loss": 0.5508, "step": 7114 }, { "epoch": 0.87, "learning_rate": 9.372920780992323e-07, "loss": 0.5211, "step": 7115 }, { "epoch": 0.87, "learning_rate": 9.356278568248012e-07, "loss": 0.5735, "step": 7116 }, { "epoch": 0.87, "learning_rate": 9.339650418154011e-07, "loss": 0.5231, "step": 7117 }, { "epoch": 0.87, "learning_rate": 9.323036333290025e-07, "loss": 0.5196, "step": 7118 }, { "epoch": 0.87, "learning_rate": 9.306436316233558e-07, "loss": 0.521, "step": 7119 }, { "epoch": 0.87, "learning_rate": 9.289850369560016e-07, "loss": 0.5192, "step": 7120 }, { "epoch": 0.87, "learning_rate": 9.273278495842553e-07, "loss": 0.5029, "step": 7121 }, { "epoch": 0.87, "learning_rate": 9.256720697652189e-07, "loss": 0.4759, "step": 7122 }, { "epoch": 0.87, "learning_rate": 9.240176977557713e-07, "loss": 0.5244, "step": 7123 }, { "epoch": 0.87, "learning_rate": 9.223647338125729e-07, "loss": 0.5362, "step": 7124 }, { "epoch": 0.87, "learning_rate": 9.207131781920708e-07, "loss": 0.4887, "step": 7125 }, { "epoch": 0.87, "learning_rate": 9.190630311504922e-07, "loss": 0.4556, "step": 7126 }, { "epoch": 0.87, "learning_rate": 9.174142929438423e-07, "loss": 0.5251, "step": 7127 }, { "epoch": 0.87, "learning_rate": 9.157669638279087e-07, "loss": 0.4678, "step": 7128 }, { "epoch": 0.87, "learning_rate": 9.141210440582638e-07, "loss": 0.5297, "step": 7129 }, { "epoch": 0.87, "learning_rate": 9.124765338902553e-07, "loss": 0.5204, "step": 7130 }, { "epoch": 0.87, "learning_rate": 9.108334335790192e-07, "loss": 0.5122, "step": 7131 }, { "epoch": 0.87, "learning_rate": 9.091917433794695e-07, "loss": 0.5061, "step": 7132 }, { "epoch": 0.87, "learning_rate": 9.075514635462978e-07, "loss": 0.5351, "step": 7133 }, { "epoch": 0.87, "learning_rate": 9.059125943339842e-07, "loss": 0.5042, "step": 7134 }, { "epoch": 0.87, "learning_rate": 9.04275135996785e-07, "loss": 0.476, "step": 7135 }, { "epoch": 0.87, "learning_rate": 9.026390887887348e-07, "loss": 0.5332, "step": 7136 }, { "epoch": 0.87, "learning_rate": 9.010044529636608e-07, "loss": 0.477, "step": 7137 }, { "epoch": 0.87, "learning_rate": 8.993712287751599e-07, "loss": 0.4663, "step": 7138 }, { "epoch": 0.87, "learning_rate": 8.977394164766118e-07, "loss": 0.4537, "step": 7139 }, { "epoch": 0.87, "learning_rate": 8.961090163211828e-07, "loss": 0.4952, "step": 7140 }, { "epoch": 0.87, "learning_rate": 8.94480028561815e-07, "loss": 0.5051, "step": 7141 }, { "epoch": 0.87, "learning_rate": 8.928524534512306e-07, "loss": 0.5522, "step": 7142 }, { "epoch": 0.87, "learning_rate": 8.91226291241939e-07, "loss": 0.5123, "step": 7143 }, { "epoch": 0.87, "learning_rate": 8.896015421862248e-07, "loss": 0.5093, "step": 7144 }, { "epoch": 0.87, "learning_rate": 8.87978206536152e-07, "loss": 0.5533, "step": 7145 }, { "epoch": 0.87, "learning_rate": 8.863562845435725e-07, "loss": 0.475, "step": 7146 }, { "epoch": 0.87, "learning_rate": 8.847357764601106e-07, "loss": 0.4492, "step": 7147 }, { "epoch": 0.87, "learning_rate": 8.831166825371762e-07, "loss": 0.481, "step": 7148 }, { "epoch": 0.87, "learning_rate": 8.814990030259629e-07, "loss": 0.4667, "step": 7149 }, { "epoch": 0.87, "learning_rate": 8.798827381774367e-07, "loss": 0.5109, "step": 7150 }, { "epoch": 0.87, "learning_rate": 8.782678882423479e-07, "loss": 0.5588, "step": 7151 }, { "epoch": 0.87, "learning_rate": 8.766544534712306e-07, "loss": 0.457, "step": 7152 }, { "epoch": 0.87, "learning_rate": 8.750424341143926e-07, "loss": 0.4544, "step": 7153 }, { "epoch": 0.87, "learning_rate": 8.734318304219302e-07, "loss": 0.4993, "step": 7154 }, { "epoch": 0.87, "learning_rate": 8.718226426437127e-07, "loss": 0.5274, "step": 7155 }, { "epoch": 0.87, "learning_rate": 8.702148710293912e-07, "loss": 0.5481, "step": 7156 }, { "epoch": 0.87, "learning_rate": 8.68608515828403e-07, "loss": 0.528, "step": 7157 }, { "epoch": 0.87, "learning_rate": 8.670035772899599e-07, "loss": 0.5227, "step": 7158 }, { "epoch": 0.87, "learning_rate": 8.654000556630516e-07, "loss": 0.469, "step": 7159 }, { "epoch": 0.87, "learning_rate": 8.637979511964544e-07, "loss": 0.5012, "step": 7160 }, { "epoch": 0.87, "learning_rate": 8.621972641387244e-07, "loss": 0.5013, "step": 7161 }, { "epoch": 0.87, "learning_rate": 8.605979947381915e-07, "loss": 0.5452, "step": 7162 }, { "epoch": 0.87, "learning_rate": 8.59000143242974e-07, "loss": 0.52, "step": 7163 }, { "epoch": 0.87, "learning_rate": 8.574037099009624e-07, "loss": 0.4975, "step": 7164 }, { "epoch": 0.87, "learning_rate": 8.558086949598287e-07, "loss": 0.4393, "step": 7165 }, { "epoch": 0.87, "learning_rate": 8.542150986670328e-07, "loss": 0.4785, "step": 7166 }, { "epoch": 0.87, "learning_rate": 8.526229212698034e-07, "loss": 0.5178, "step": 7167 }, { "epoch": 0.87, "learning_rate": 8.510321630151552e-07, "loss": 0.5237, "step": 7168 }, { "epoch": 0.87, "learning_rate": 8.494428241498831e-07, "loss": 0.5214, "step": 7169 }, { "epoch": 0.87, "learning_rate": 8.478549049205586e-07, "loss": 0.5323, "step": 7170 }, { "epoch": 0.87, "learning_rate": 8.46268405573537e-07, "loss": 0.5092, "step": 7171 }, { "epoch": 0.87, "learning_rate": 8.446833263549481e-07, "loss": 0.4994, "step": 7172 }, { "epoch": 0.87, "learning_rate": 8.430996675107061e-07, "loss": 0.4701, "step": 7173 }, { "epoch": 0.87, "learning_rate": 8.415174292865025e-07, "loss": 0.5645, "step": 7174 }, { "epoch": 0.87, "learning_rate": 8.399366119278107e-07, "loss": 0.5462, "step": 7175 }, { "epoch": 0.87, "learning_rate": 8.383572156798781e-07, "loss": 0.5695, "step": 7176 }, { "epoch": 0.87, "learning_rate": 8.367792407877395e-07, "loss": 0.5183, "step": 7177 }, { "epoch": 0.87, "learning_rate": 8.352026874962038e-07, "loss": 0.5015, "step": 7178 }, { "epoch": 0.87, "learning_rate": 8.336275560498575e-07, "loss": 0.4976, "step": 7179 }, { "epoch": 0.87, "learning_rate": 8.32053846693075e-07, "loss": 0.5253, "step": 7180 }, { "epoch": 0.87, "learning_rate": 8.304815596700022e-07, "loss": 0.4731, "step": 7181 }, { "epoch": 0.87, "learning_rate": 8.289106952245641e-07, "loss": 0.4899, "step": 7182 }, { "epoch": 0.87, "learning_rate": 8.273412536004722e-07, "loss": 0.4764, "step": 7183 }, { "epoch": 0.87, "learning_rate": 8.257732350412107e-07, "loss": 0.4926, "step": 7184 }, { "epoch": 0.87, "learning_rate": 8.242066397900439e-07, "loss": 0.5394, "step": 7185 }, { "epoch": 0.87, "learning_rate": 8.226414680900208e-07, "loss": 0.5336, "step": 7186 }, { "epoch": 0.87, "learning_rate": 8.210777201839637e-07, "loss": 0.509, "step": 7187 }, { "epoch": 0.87, "learning_rate": 8.19515396314472e-07, "loss": 0.4586, "step": 7188 }, { "epoch": 0.87, "learning_rate": 8.179544967239317e-07, "loss": 0.5046, "step": 7189 }, { "epoch": 0.87, "learning_rate": 8.163950216545036e-07, "loss": 0.4493, "step": 7190 }, { "epoch": 0.87, "learning_rate": 8.148369713481242e-07, "loss": 0.5208, "step": 7191 }, { "epoch": 0.87, "learning_rate": 8.132803460465178e-07, "loss": 0.5187, "step": 7192 }, { "epoch": 0.87, "learning_rate": 8.117251459911779e-07, "loss": 0.4537, "step": 7193 }, { "epoch": 0.87, "learning_rate": 8.10171371423385e-07, "loss": 0.4983, "step": 7194 }, { "epoch": 0.87, "learning_rate": 8.086190225841927e-07, "loss": 0.4831, "step": 7195 }, { "epoch": 0.88, "learning_rate": 8.070680997144353e-07, "loss": 0.5196, "step": 7196 }, { "epoch": 0.88, "learning_rate": 8.055186030547257e-07, "loss": 0.4625, "step": 7197 }, { "epoch": 0.88, "learning_rate": 8.039705328454595e-07, "loss": 0.5234, "step": 7198 }, { "epoch": 0.88, "learning_rate": 8.024238893268033e-07, "loss": 0.4896, "step": 7199 }, { "epoch": 0.88, "learning_rate": 8.008786727387097e-07, "loss": 0.5276, "step": 7200 }, { "epoch": 0.88, "learning_rate": 7.993348833209058e-07, "loss": 0.5437, "step": 7201 }, { "epoch": 0.88, "learning_rate": 7.977925213128945e-07, "loss": 0.4779, "step": 7202 }, { "epoch": 0.88, "learning_rate": 7.962515869539677e-07, "loss": 0.5198, "step": 7203 }, { "epoch": 0.88, "learning_rate": 7.947120804831843e-07, "loss": 0.5381, "step": 7204 }, { "epoch": 0.88, "learning_rate": 7.931740021393864e-07, "loss": 0.6089, "step": 7205 }, { "epoch": 0.88, "learning_rate": 7.916373521611964e-07, "loss": 0.5341, "step": 7206 }, { "epoch": 0.88, "learning_rate": 7.901021307870138e-07, "loss": 0.4968, "step": 7207 }, { "epoch": 0.88, "learning_rate": 7.885683382550113e-07, "loss": 0.5017, "step": 7208 }, { "epoch": 0.88, "learning_rate": 7.87035974803152e-07, "loss": 0.5192, "step": 7209 }, { "epoch": 0.88, "learning_rate": 7.855050406691656e-07, "loss": 0.5088, "step": 7210 }, { "epoch": 0.88, "learning_rate": 7.839755360905621e-07, "loss": 0.5424, "step": 7211 }, { "epoch": 0.88, "learning_rate": 7.824474613046373e-07, "loss": 0.536, "step": 7212 }, { "epoch": 0.88, "learning_rate": 7.809208165484571e-07, "loss": 0.525, "step": 7213 }, { "epoch": 0.88, "learning_rate": 7.793956020588667e-07, "loss": 0.5233, "step": 7214 }, { "epoch": 0.88, "learning_rate": 7.778718180724931e-07, "loss": 0.4758, "step": 7215 }, { "epoch": 0.88, "learning_rate": 7.763494648257374e-07, "loss": 0.5297, "step": 7216 }, { "epoch": 0.88, "learning_rate": 7.748285425547841e-07, "loss": 0.5229, "step": 7217 }, { "epoch": 0.88, "learning_rate": 7.733090514955888e-07, "loss": 0.4944, "step": 7218 }, { "epoch": 0.88, "learning_rate": 7.717909918838884e-07, "loss": 0.469, "step": 7219 }, { "epoch": 0.88, "learning_rate": 7.702743639551979e-07, "loss": 0.4893, "step": 7220 }, { "epoch": 0.88, "learning_rate": 7.687591679448136e-07, "loss": 0.4943, "step": 7221 }, { "epoch": 0.88, "learning_rate": 7.67245404087802e-07, "loss": 0.5263, "step": 7222 }, { "epoch": 0.88, "learning_rate": 7.657330726190137e-07, "loss": 0.5222, "step": 7223 }, { "epoch": 0.88, "learning_rate": 7.642221737730737e-07, "loss": 0.5085, "step": 7224 }, { "epoch": 0.88, "learning_rate": 7.627127077843855e-07, "loss": 0.4971, "step": 7225 }, { "epoch": 0.88, "learning_rate": 7.612046748871327e-07, "loss": 0.5229, "step": 7226 }, { "epoch": 0.88, "learning_rate": 7.596980753152738e-07, "loss": 0.5123, "step": 7227 }, { "epoch": 0.88, "learning_rate": 7.581929093025431e-07, "loss": 0.4542, "step": 7228 }, { "epoch": 0.88, "learning_rate": 7.566891770824591e-07, "loss": 0.5471, "step": 7229 }, { "epoch": 0.88, "learning_rate": 7.55186878888311e-07, "loss": 0.4961, "step": 7230 }, { "epoch": 0.88, "learning_rate": 7.536860149531689e-07, "loss": 0.5278, "step": 7231 }, { "epoch": 0.88, "learning_rate": 7.521865855098809e-07, "loss": 0.4782, "step": 7232 }, { "epoch": 0.88, "learning_rate": 7.506885907910699e-07, "loss": 0.5176, "step": 7233 }, { "epoch": 0.88, "learning_rate": 7.49192031029139e-07, "loss": 0.5744, "step": 7234 }, { "epoch": 0.88, "learning_rate": 7.476969064562689e-07, "loss": 0.5144, "step": 7235 }, { "epoch": 0.88, "learning_rate": 7.462032173044154e-07, "loss": 0.5008, "step": 7236 }, { "epoch": 0.88, "learning_rate": 7.447109638053096e-07, "loss": 0.5107, "step": 7237 }, { "epoch": 0.88, "learning_rate": 7.432201461904664e-07, "loss": 0.511, "step": 7238 }, { "epoch": 0.88, "learning_rate": 7.417307646911709e-07, "loss": 0.4141, "step": 7239 }, { "epoch": 0.88, "learning_rate": 7.402428195384925e-07, "loss": 0.505, "step": 7240 }, { "epoch": 0.88, "learning_rate": 7.387563109632723e-07, "loss": 0.5265, "step": 7241 }, { "epoch": 0.88, "learning_rate": 7.372712391961279e-07, "loss": 0.474, "step": 7242 }, { "epoch": 0.88, "learning_rate": 7.357876044674606e-07, "loss": 0.5071, "step": 7243 }, { "epoch": 0.88, "learning_rate": 7.343054070074429e-07, "loss": 0.4666, "step": 7244 }, { "epoch": 0.88, "learning_rate": 7.328246470460209e-07, "loss": 0.4861, "step": 7245 }, { "epoch": 0.88, "learning_rate": 7.31345324812931e-07, "loss": 0.5402, "step": 7246 }, { "epoch": 0.88, "learning_rate": 7.298674405376749e-07, "loss": 0.5445, "step": 7247 }, { "epoch": 0.88, "learning_rate": 7.283909944495327e-07, "loss": 0.5532, "step": 7248 }, { "epoch": 0.88, "learning_rate": 7.269159867775655e-07, "loss": 0.5038, "step": 7249 }, { "epoch": 0.88, "learning_rate": 7.254424177506091e-07, "loss": 0.4622, "step": 7250 }, { "epoch": 0.88, "learning_rate": 7.239702875972732e-07, "loss": 0.5453, "step": 7251 }, { "epoch": 0.88, "learning_rate": 7.224995965459503e-07, "loss": 0.4962, "step": 7252 }, { "epoch": 0.88, "learning_rate": 7.210303448248046e-07, "loss": 0.4734, "step": 7253 }, { "epoch": 0.88, "learning_rate": 7.195625326617772e-07, "loss": 0.5357, "step": 7254 }, { "epoch": 0.88, "learning_rate": 7.180961602845915e-07, "loss": 0.568, "step": 7255 }, { "epoch": 0.88, "learning_rate": 7.166312279207399e-07, "loss": 0.5091, "step": 7256 }, { "epoch": 0.88, "learning_rate": 7.151677357974962e-07, "loss": 0.5259, "step": 7257 }, { "epoch": 0.88, "learning_rate": 7.137056841419121e-07, "loss": 0.5743, "step": 7258 }, { "epoch": 0.88, "learning_rate": 7.122450731808095e-07, "loss": 0.5341, "step": 7259 }, { "epoch": 0.88, "learning_rate": 7.107859031407915e-07, "loss": 0.5338, "step": 7260 }, { "epoch": 0.88, "learning_rate": 7.093281742482383e-07, "loss": 0.5104, "step": 7261 }, { "epoch": 0.88, "learning_rate": 7.078718867293022e-07, "loss": 0.5052, "step": 7262 }, { "epoch": 0.88, "learning_rate": 7.06417040809918e-07, "loss": 0.4908, "step": 7263 }, { "epoch": 0.88, "learning_rate": 7.049636367157908e-07, "loss": 0.4694, "step": 7264 }, { "epoch": 0.88, "learning_rate": 7.035116746724046e-07, "loss": 0.5376, "step": 7265 }, { "epoch": 0.88, "learning_rate": 7.020611549050216e-07, "loss": 0.5177, "step": 7266 }, { "epoch": 0.88, "learning_rate": 7.006120776386782e-07, "loss": 0.5211, "step": 7267 }, { "epoch": 0.88, "learning_rate": 6.991644430981836e-07, "loss": 0.539, "step": 7268 }, { "epoch": 0.88, "learning_rate": 6.977182515081304e-07, "loss": 0.498, "step": 7269 }, { "epoch": 0.88, "learning_rate": 6.962735030928835e-07, "loss": 0.5122, "step": 7270 }, { "epoch": 0.88, "learning_rate": 6.948301980765825e-07, "loss": 0.5223, "step": 7271 }, { "epoch": 0.88, "learning_rate": 6.933883366831462e-07, "loss": 0.487, "step": 7272 }, { "epoch": 0.88, "learning_rate": 6.919479191362677e-07, "loss": 0.5285, "step": 7273 }, { "epoch": 0.88, "learning_rate": 6.90508945659415e-07, "loss": 0.4935, "step": 7274 }, { "epoch": 0.88, "learning_rate": 6.89071416475835e-07, "loss": 0.4495, "step": 7275 }, { "epoch": 0.88, "learning_rate": 6.876353318085471e-07, "loss": 0.4726, "step": 7276 }, { "epoch": 0.88, "learning_rate": 6.862006918803487e-07, "loss": 0.4812, "step": 7277 }, { "epoch": 0.89, "learning_rate": 6.847674969138152e-07, "loss": 0.4666, "step": 7278 }, { "epoch": 0.89, "learning_rate": 6.833357471312918e-07, "loss": 0.4644, "step": 7279 }, { "epoch": 0.89, "learning_rate": 6.819054427549054e-07, "loss": 0.483, "step": 7280 }, { "epoch": 0.89, "learning_rate": 6.804765840065553e-07, "loss": 0.5041, "step": 7281 }, { "epoch": 0.89, "learning_rate": 6.790491711079184e-07, "loss": 0.5188, "step": 7282 }, { "epoch": 0.89, "learning_rate": 6.776232042804454e-07, "loss": 0.4783, "step": 7283 }, { "epoch": 0.89, "learning_rate": 6.761986837453649e-07, "loss": 0.552, "step": 7284 }, { "epoch": 0.89, "learning_rate": 6.74775609723678e-07, "loss": 0.4718, "step": 7285 }, { "epoch": 0.89, "learning_rate": 6.733539824361668e-07, "loss": 0.5076, "step": 7286 }, { "epoch": 0.89, "learning_rate": 6.719338021033816e-07, "loss": 0.4742, "step": 7287 }, { "epoch": 0.89, "learning_rate": 6.70515068945653e-07, "loss": 0.513, "step": 7288 }, { "epoch": 0.89, "learning_rate": 6.690977831830881e-07, "loss": 0.464, "step": 7289 }, { "epoch": 0.89, "learning_rate": 6.676819450355665e-07, "loss": 0.5146, "step": 7290 }, { "epoch": 0.89, "learning_rate": 6.662675547227415e-07, "loss": 0.489, "step": 7291 }, { "epoch": 0.89, "learning_rate": 6.648546124640487e-07, "loss": 0.4721, "step": 7292 }, { "epoch": 0.89, "learning_rate": 6.634431184786916e-07, "loss": 0.4719, "step": 7293 }, { "epoch": 0.89, "learning_rate": 6.620330729856528e-07, "loss": 0.4735, "step": 7294 }, { "epoch": 0.89, "learning_rate": 6.606244762036928e-07, "loss": 0.5431, "step": 7295 }, { "epoch": 0.89, "learning_rate": 6.592173283513403e-07, "loss": 0.4994, "step": 7296 }, { "epoch": 0.89, "learning_rate": 6.57811629646904e-07, "loss": 0.4902, "step": 7297 }, { "epoch": 0.89, "learning_rate": 6.564073803084691e-07, "loss": 0.4895, "step": 7298 }, { "epoch": 0.89, "learning_rate": 6.550045805538907e-07, "loss": 0.5287, "step": 7299 }, { "epoch": 0.89, "learning_rate": 6.536032306008022e-07, "loss": 0.5162, "step": 7300 }, { "epoch": 0.89, "learning_rate": 6.522033306666143e-07, "loss": 0.5309, "step": 7301 }, { "epoch": 0.89, "learning_rate": 6.508048809685074e-07, "loss": 0.473, "step": 7302 }, { "epoch": 0.89, "learning_rate": 6.494078817234428e-07, "loss": 0.4882, "step": 7303 }, { "epoch": 0.89, "learning_rate": 6.480123331481525e-07, "loss": 0.4837, "step": 7304 }, { "epoch": 0.89, "learning_rate": 6.466182354591421e-07, "loss": 0.4819, "step": 7305 }, { "epoch": 0.89, "learning_rate": 6.452255888726977e-07, "loss": 0.4811, "step": 7306 }, { "epoch": 0.89, "learning_rate": 6.438343936048775e-07, "loss": 0.5088, "step": 7307 }, { "epoch": 0.89, "learning_rate": 6.424446498715131e-07, "loss": 0.4767, "step": 7308 }, { "epoch": 0.89, "learning_rate": 6.410563578882134e-07, "loss": 0.4965, "step": 7309 }, { "epoch": 0.89, "learning_rate": 6.396695178703605e-07, "loss": 0.5237, "step": 7310 }, { "epoch": 0.89, "learning_rate": 6.382841300331088e-07, "loss": 0.4911, "step": 7311 }, { "epoch": 0.89, "learning_rate": 6.369001945913944e-07, "loss": 0.498, "step": 7312 }, { "epoch": 0.89, "learning_rate": 6.35517711759922e-07, "loss": 0.5366, "step": 7313 }, { "epoch": 0.89, "learning_rate": 6.341366817531713e-07, "loss": 0.4932, "step": 7314 }, { "epoch": 0.89, "learning_rate": 6.327571047854009e-07, "loss": 0.526, "step": 7315 }, { "epoch": 0.89, "learning_rate": 6.313789810706395e-07, "loss": 0.4482, "step": 7316 }, { "epoch": 0.89, "learning_rate": 6.300023108226893e-07, "loss": 0.5078, "step": 7317 }, { "epoch": 0.89, "learning_rate": 6.286270942551364e-07, "loss": 0.5555, "step": 7318 }, { "epoch": 0.89, "learning_rate": 6.272533315813301e-07, "loss": 0.4829, "step": 7319 }, { "epoch": 0.89, "learning_rate": 6.258810230143986e-07, "loss": 0.4849, "step": 7320 }, { "epoch": 0.89, "learning_rate": 6.245101687672461e-07, "loss": 0.5064, "step": 7321 }, { "epoch": 0.89, "learning_rate": 6.231407690525503e-07, "loss": 0.5408, "step": 7322 }, { "epoch": 0.89, "learning_rate": 6.217728240827592e-07, "loss": 0.5228, "step": 7323 }, { "epoch": 0.89, "learning_rate": 6.204063340701028e-07, "loss": 0.4759, "step": 7324 }, { "epoch": 0.89, "learning_rate": 6.190412992265771e-07, "loss": 0.5239, "step": 7325 }, { "epoch": 0.89, "learning_rate": 6.176777197639605e-07, "loss": 0.5472, "step": 7326 }, { "epoch": 0.89, "learning_rate": 6.163155958937994e-07, "loss": 0.5276, "step": 7327 }, { "epoch": 0.89, "learning_rate": 6.149549278274147e-07, "loss": 0.5015, "step": 7328 }, { "epoch": 0.89, "learning_rate": 6.135957157759053e-07, "loss": 0.5075, "step": 7329 }, { "epoch": 0.89, "learning_rate": 6.122379599501437e-07, "loss": 0.4842, "step": 7330 }, { "epoch": 0.89, "learning_rate": 6.108816605607726e-07, "loss": 0.5302, "step": 7331 }, { "epoch": 0.89, "learning_rate": 6.09526817818209e-07, "loss": 0.4547, "step": 7332 }, { "epoch": 0.89, "learning_rate": 6.081734319326516e-07, "loss": 0.5095, "step": 7333 }, { "epoch": 0.89, "learning_rate": 6.068215031140612e-07, "loss": 0.5382, "step": 7334 }, { "epoch": 0.89, "learning_rate": 6.054710315721835e-07, "loss": 0.4984, "step": 7335 }, { "epoch": 0.89, "learning_rate": 6.041220175165319e-07, "loss": 0.5218, "step": 7336 }, { "epoch": 0.89, "learning_rate": 6.027744611563935e-07, "loss": 0.5645, "step": 7337 }, { "epoch": 0.89, "learning_rate": 6.014283627008333e-07, "loss": 0.5014, "step": 7338 }, { "epoch": 0.89, "learning_rate": 6.000837223586875e-07, "loss": 0.5784, "step": 7339 }, { "epoch": 0.89, "learning_rate": 5.987405403385638e-07, "loss": 0.4904, "step": 7340 }, { "epoch": 0.89, "learning_rate": 5.973988168488476e-07, "loss": 0.5393, "step": 7341 }, { "epoch": 0.89, "learning_rate": 5.96058552097698e-07, "loss": 0.5033, "step": 7342 }, { "epoch": 0.89, "learning_rate": 5.947197462930454e-07, "loss": 0.4905, "step": 7343 }, { "epoch": 0.89, "learning_rate": 5.933823996425947e-07, "loss": 0.5283, "step": 7344 }, { "epoch": 0.89, "learning_rate": 5.920465123538255e-07, "loss": 0.4479, "step": 7345 }, { "epoch": 0.89, "learning_rate": 5.907120846339876e-07, "loss": 0.5035, "step": 7346 }, { "epoch": 0.89, "learning_rate": 5.893791166901097e-07, "loss": 0.5269, "step": 7347 }, { "epoch": 0.89, "learning_rate": 5.880476087289899e-07, "loss": 0.5082, "step": 7348 }, { "epoch": 0.89, "learning_rate": 5.867175609571995e-07, "loss": 0.458, "step": 7349 }, { "epoch": 0.89, "learning_rate": 5.853889735810881e-07, "loss": 0.4942, "step": 7350 }, { "epoch": 0.89, "learning_rate": 5.840618468067716e-07, "loss": 0.5533, "step": 7351 }, { "epoch": 0.89, "learning_rate": 5.827361808401466e-07, "loss": 0.4142, "step": 7352 }, { "epoch": 0.89, "learning_rate": 5.814119758868775e-07, "loss": 0.5324, "step": 7353 }, { "epoch": 0.89, "learning_rate": 5.800892321524054e-07, "loss": 0.5398, "step": 7354 }, { "epoch": 0.89, "learning_rate": 5.787679498419418e-07, "loss": 0.4979, "step": 7355 }, { "epoch": 0.89, "learning_rate": 5.77448129160475e-07, "loss": 0.4416, "step": 7356 }, { "epoch": 0.89, "learning_rate": 5.761297703127633e-07, "loss": 0.5734, "step": 7357 }, { "epoch": 0.89, "learning_rate": 5.74812873503341e-07, "loss": 0.5094, "step": 7358 }, { "epoch": 0.89, "learning_rate": 5.734974389365133e-07, "loss": 0.531, "step": 7359 }, { "epoch": 0.9, "learning_rate": 5.721834668163573e-07, "loss": 0.4786, "step": 7360 }, { "epoch": 0.9, "learning_rate": 5.708709573467297e-07, "loss": 0.5335, "step": 7361 }, { "epoch": 0.9, "learning_rate": 5.695599107312522e-07, "loss": 0.5258, "step": 7362 }, { "epoch": 0.9, "learning_rate": 5.682503271733242e-07, "loss": 0.5173, "step": 7363 }, { "epoch": 0.9, "learning_rate": 5.669422068761177e-07, "loss": 0.4547, "step": 7364 }, { "epoch": 0.9, "learning_rate": 5.656355500425758e-07, "loss": 0.4594, "step": 7365 }, { "epoch": 0.9, "learning_rate": 5.643303568754177e-07, "loss": 0.5374, "step": 7366 }, { "epoch": 0.9, "learning_rate": 5.630266275771335e-07, "loss": 0.5555, "step": 7367 }, { "epoch": 0.9, "learning_rate": 5.617243623499858e-07, "loss": 0.4812, "step": 7368 }, { "epoch": 0.9, "learning_rate": 5.604235613960096e-07, "loss": 0.5561, "step": 7369 }, { "epoch": 0.9, "learning_rate": 5.591242249170159e-07, "loss": 0.487, "step": 7370 }, { "epoch": 0.9, "learning_rate": 5.578263531145855e-07, "loss": 0.523, "step": 7371 }, { "epoch": 0.9, "learning_rate": 5.56529946190072e-07, "loss": 0.5117, "step": 7372 }, { "epoch": 0.9, "learning_rate": 5.552350043446031e-07, "loss": 0.5524, "step": 7373 }, { "epoch": 0.9, "learning_rate": 5.539415277790794e-07, "loss": 0.5219, "step": 7374 }, { "epoch": 0.9, "learning_rate": 5.526495166941726e-07, "loss": 0.5588, "step": 7375 }, { "epoch": 0.9, "learning_rate": 5.5135897129033e-07, "loss": 0.5102, "step": 7376 }, { "epoch": 0.9, "learning_rate": 5.500698917677649e-07, "loss": 0.4845, "step": 7377 }, { "epoch": 0.9, "learning_rate": 5.487822783264707e-07, "loss": 0.4896, "step": 7378 }, { "epoch": 0.9, "learning_rate": 5.474961311662119e-07, "loss": 0.493, "step": 7379 }, { "epoch": 0.9, "learning_rate": 5.462114504865212e-07, "loss": 0.5095, "step": 7380 }, { "epoch": 0.9, "learning_rate": 5.449282364867093e-07, "loss": 0.5133, "step": 7381 }, { "epoch": 0.9, "learning_rate": 5.436464893658533e-07, "loss": 0.4429, "step": 7382 }, { "epoch": 0.9, "learning_rate": 5.423662093228077e-07, "loss": 0.5171, "step": 7383 }, { "epoch": 0.9, "learning_rate": 5.410873965561981e-07, "loss": 0.5766, "step": 7384 }, { "epoch": 0.9, "learning_rate": 5.398100512644222e-07, "loss": 0.4972, "step": 7385 }, { "epoch": 0.9, "learning_rate": 5.385341736456484e-07, "loss": 0.5213, "step": 7386 }, { "epoch": 0.9, "learning_rate": 5.372597638978205e-07, "loss": 0.5428, "step": 7387 }, { "epoch": 0.9, "learning_rate": 5.359868222186537e-07, "loss": 0.531, "step": 7388 }, { "epoch": 0.9, "learning_rate": 5.3471534880563e-07, "loss": 0.5636, "step": 7389 }, { "epoch": 0.9, "learning_rate": 5.334453438560138e-07, "loss": 0.5253, "step": 7390 }, { "epoch": 0.9, "learning_rate": 5.321768075668354e-07, "loss": 0.5222, "step": 7391 }, { "epoch": 0.9, "learning_rate": 5.30909740134895e-07, "loss": 0.509, "step": 7392 }, { "epoch": 0.9, "learning_rate": 5.296441417567722e-07, "loss": 0.5367, "step": 7393 }, { "epoch": 0.9, "learning_rate": 5.283800126288119e-07, "loss": 0.4739, "step": 7394 }, { "epoch": 0.9, "learning_rate": 5.271173529471319e-07, "loss": 0.4889, "step": 7395 }, { "epoch": 0.9, "learning_rate": 5.258561629076287e-07, "loss": 0.5705, "step": 7396 }, { "epoch": 0.9, "learning_rate": 5.245964427059602e-07, "loss": 0.5341, "step": 7397 }, { "epoch": 0.9, "learning_rate": 5.233381925375669e-07, "loss": 0.4953, "step": 7398 }, { "epoch": 0.9, "learning_rate": 5.220814125976548e-07, "loss": 0.4921, "step": 7399 }, { "epoch": 0.9, "learning_rate": 5.208261030812001e-07, "loss": 0.4726, "step": 7400 }, { "epoch": 0.9, "learning_rate": 5.19572264182957e-07, "loss": 0.5095, "step": 7401 }, { "epoch": 0.9, "learning_rate": 5.1831989609745e-07, "loss": 0.5308, "step": 7402 }, { "epoch": 0.9, "learning_rate": 5.170689990189703e-07, "loss": 0.5189, "step": 7403 }, { "epoch": 0.9, "learning_rate": 5.158195731415883e-07, "loss": 0.485, "step": 7404 }, { "epoch": 0.9, "learning_rate": 5.14571618659141e-07, "loss": 0.4894, "step": 7405 }, { "epoch": 0.9, "learning_rate": 5.133251357652369e-07, "loss": 0.576, "step": 7406 }, { "epoch": 0.9, "learning_rate": 5.120801246532603e-07, "loss": 0.5366, "step": 7407 }, { "epoch": 0.9, "learning_rate": 5.108365855163644e-07, "loss": 0.4502, "step": 7408 }, { "epoch": 0.9, "learning_rate": 5.095945185474716e-07, "loss": 0.4883, "step": 7409 }, { "epoch": 0.9, "learning_rate": 5.08353923939282e-07, "loss": 0.5185, "step": 7410 }, { "epoch": 0.9, "learning_rate": 5.071148018842631e-07, "loss": 0.4709, "step": 7411 }, { "epoch": 0.9, "learning_rate": 5.058771525746531e-07, "loss": 0.4867, "step": 7412 }, { "epoch": 0.9, "learning_rate": 5.046409762024651e-07, "loss": 0.5579, "step": 7413 }, { "epoch": 0.9, "learning_rate": 5.034062729594824e-07, "loss": 0.5032, "step": 7414 }, { "epoch": 0.9, "learning_rate": 5.021730430372574e-07, "loss": 0.45, "step": 7415 }, { "epoch": 0.9, "learning_rate": 5.00941286627119e-07, "loss": 0.5084, "step": 7416 }, { "epoch": 0.9, "learning_rate": 4.997110039201614e-07, "loss": 0.5574, "step": 7417 }, { "epoch": 0.9, "learning_rate": 4.984821951072538e-07, "loss": 0.5137, "step": 7418 }, { "epoch": 0.9, "learning_rate": 4.972548603790372e-07, "loss": 0.5339, "step": 7419 }, { "epoch": 0.9, "learning_rate": 4.960289999259216e-07, "loss": 0.4592, "step": 7420 }, { "epoch": 0.9, "learning_rate": 4.948046139380902e-07, "loss": 0.5365, "step": 7421 }, { "epoch": 0.9, "learning_rate": 4.935817026054978e-07, "loss": 0.4974, "step": 7422 }, { "epoch": 0.9, "learning_rate": 4.923602661178661e-07, "loss": 0.5341, "step": 7423 }, { "epoch": 0.9, "learning_rate": 4.911403046646945e-07, "loss": 0.4667, "step": 7424 }, { "epoch": 0.9, "learning_rate": 4.899218184352473e-07, "loss": 0.5502, "step": 7425 }, { "epoch": 0.9, "learning_rate": 4.887048076185652e-07, "loss": 0.4747, "step": 7426 }, { "epoch": 0.9, "learning_rate": 4.874892724034574e-07, "loss": 0.4592, "step": 7427 }, { "epoch": 0.9, "learning_rate": 4.862752129785054e-07, "loss": 0.5382, "step": 7428 }, { "epoch": 0.9, "learning_rate": 4.850626295320571e-07, "loss": 0.5604, "step": 7429 }, { "epoch": 0.9, "learning_rate": 4.8385152225224e-07, "loss": 0.5685, "step": 7430 }, { "epoch": 0.9, "learning_rate": 4.826418913269448e-07, "loss": 0.4864, "step": 7431 }, { "epoch": 0.9, "learning_rate": 4.81433736943836e-07, "loss": 0.4653, "step": 7432 }, { "epoch": 0.9, "learning_rate": 4.802270592903502e-07, "loss": 0.5044, "step": 7433 }, { "epoch": 0.9, "learning_rate": 4.790218585536943e-07, "loss": 0.5118, "step": 7434 }, { "epoch": 0.9, "learning_rate": 4.778181349208422e-07, "loss": 0.5362, "step": 7435 }, { "epoch": 0.9, "learning_rate": 4.766158885785466e-07, "loss": 0.5004, "step": 7436 }, { "epoch": 0.9, "learning_rate": 4.754151197133228e-07, "loss": 0.4834, "step": 7437 }, { "epoch": 0.9, "learning_rate": 4.7421582851146287e-07, "loss": 0.525, "step": 7438 }, { "epoch": 0.9, "learning_rate": 4.730180151590269e-07, "loss": 0.522, "step": 7439 }, { "epoch": 0.9, "learning_rate": 4.7182167984184625e-07, "loss": 0.5007, "step": 7440 }, { "epoch": 0.9, "learning_rate": 4.7062682274552016e-07, "loss": 0.5406, "step": 7441 }, { "epoch": 0.9, "learning_rate": 4.694334440554249e-07, "loss": 0.5043, "step": 7442 }, { "epoch": 0.91, "learning_rate": 4.6824154395670005e-07, "loss": 0.4879, "step": 7443 }, { "epoch": 0.91, "learning_rate": 4.6705112263426334e-07, "loss": 0.4861, "step": 7444 }, { "epoch": 0.91, "learning_rate": 4.6586218027279695e-07, "loss": 0.514, "step": 7445 }, { "epoch": 0.91, "learning_rate": 4.646747170567545e-07, "loss": 0.4629, "step": 7446 }, { "epoch": 0.91, "learning_rate": 4.6348873317036415e-07, "loss": 0.5319, "step": 7447 }, { "epoch": 0.91, "learning_rate": 4.6230422879762095e-07, "loss": 0.5193, "step": 7448 }, { "epoch": 0.91, "learning_rate": 4.611212041222879e-07, "loss": 0.567, "step": 7449 }, { "epoch": 0.91, "learning_rate": 4.5993965932790706e-07, "loss": 0.4718, "step": 7450 }, { "epoch": 0.91, "learning_rate": 4.5875959459778405e-07, "loss": 0.4916, "step": 7451 }, { "epoch": 0.91, "learning_rate": 4.575810101149947e-07, "loss": 0.4761, "step": 7452 }, { "epoch": 0.91, "learning_rate": 4.564039060623904e-07, "loss": 0.5592, "step": 7453 }, { "epoch": 0.91, "learning_rate": 4.552282826225862e-07, "loss": 0.4927, "step": 7454 }, { "epoch": 0.91, "learning_rate": 4.5405413997797074e-07, "loss": 0.4909, "step": 7455 }, { "epoch": 0.91, "learning_rate": 4.5288147831070495e-07, "loss": 0.5639, "step": 7456 }, { "epoch": 0.91, "learning_rate": 4.517102978027166e-07, "loss": 0.5327, "step": 7457 }, { "epoch": 0.91, "learning_rate": 4.5054059863570475e-07, "loss": 0.569, "step": 7458 }, { "epoch": 0.91, "learning_rate": 4.49372380991141e-07, "loss": 0.5111, "step": 7459 }, { "epoch": 0.91, "learning_rate": 4.4820564505026145e-07, "loss": 0.5272, "step": 7460 }, { "epoch": 0.91, "learning_rate": 4.470403909940779e-07, "loss": 0.5021, "step": 7461 }, { "epoch": 0.91, "learning_rate": 4.4587661900337144e-07, "loss": 0.5293, "step": 7462 }, { "epoch": 0.91, "learning_rate": 4.4471432925869085e-07, "loss": 0.4679, "step": 7463 }, { "epoch": 0.91, "learning_rate": 4.4355352194035526e-07, "loss": 0.5226, "step": 7464 }, { "epoch": 0.91, "learning_rate": 4.4239419722845623e-07, "loss": 0.5511, "step": 7465 }, { "epoch": 0.91, "learning_rate": 4.4123635530285204e-07, "loss": 0.4988, "step": 7466 }, { "epoch": 0.91, "learning_rate": 4.4007999634317454e-07, "loss": 0.5068, "step": 7467 }, { "epoch": 0.91, "learning_rate": 4.3892512052882253e-07, "loss": 0.544, "step": 7468 }, { "epoch": 0.91, "learning_rate": 4.3777172803896484e-07, "loss": 0.4986, "step": 7469 }, { "epoch": 0.91, "learning_rate": 4.366198190525439e-07, "loss": 0.5485, "step": 7470 }, { "epoch": 0.91, "learning_rate": 4.354693937482668e-07, "loss": 0.5017, "step": 7471 }, { "epoch": 0.91, "learning_rate": 4.3432045230461183e-07, "loss": 0.4462, "step": 7472 }, { "epoch": 0.91, "learning_rate": 4.331729948998298e-07, "loss": 0.4894, "step": 7473 }, { "epoch": 0.91, "learning_rate": 4.3202702171194155e-07, "loss": 0.4453, "step": 7474 }, { "epoch": 0.91, "learning_rate": 4.308825329187316e-07, "loss": 0.4777, "step": 7475 }, { "epoch": 0.91, "learning_rate": 4.2973952869776236e-07, "loss": 0.47, "step": 7476 }, { "epoch": 0.91, "learning_rate": 4.2859800922635864e-07, "loss": 0.4623, "step": 7477 }, { "epoch": 0.91, "learning_rate": 4.2745797468161874e-07, "loss": 0.4883, "step": 7478 }, { "epoch": 0.91, "learning_rate": 4.2631942524041124e-07, "loss": 0.4509, "step": 7479 }, { "epoch": 0.91, "learning_rate": 4.251823610793726e-07, "loss": 0.5418, "step": 7480 }, { "epoch": 0.91, "learning_rate": 4.240467823749073e-07, "loss": 0.5332, "step": 7481 }, { "epoch": 0.91, "learning_rate": 4.229126893031932e-07, "loss": 0.5471, "step": 7482 }, { "epoch": 0.91, "learning_rate": 4.217800820401752e-07, "loss": 0.4882, "step": 7483 }, { "epoch": 0.91, "learning_rate": 4.2064896076156937e-07, "loss": 0.5866, "step": 7484 }, { "epoch": 0.91, "learning_rate": 4.195193256428576e-07, "loss": 0.4944, "step": 7485 }, { "epoch": 0.91, "learning_rate": 4.183911768592974e-07, "loss": 0.5161, "step": 7486 }, { "epoch": 0.91, "learning_rate": 4.1726451458590887e-07, "loss": 0.5138, "step": 7487 }, { "epoch": 0.91, "learning_rate": 4.161393389974866e-07, "loss": 0.5127, "step": 7488 }, { "epoch": 0.91, "learning_rate": 4.1501565026859093e-07, "loss": 0.5467, "step": 7489 }, { "epoch": 0.91, "learning_rate": 4.1389344857355573e-07, "loss": 0.5523, "step": 7490 }, { "epoch": 0.91, "learning_rate": 4.1277273408647955e-07, "loss": 0.386, "step": 7491 }, { "epoch": 0.91, "learning_rate": 4.116535069812311e-07, "loss": 0.5593, "step": 7492 }, { "epoch": 0.91, "learning_rate": 4.105357674314536e-07, "loss": 0.5243, "step": 7493 }, { "epoch": 0.91, "learning_rate": 4.0941951561055295e-07, "loss": 0.4694, "step": 7494 }, { "epoch": 0.91, "learning_rate": 4.083047516917049e-07, "loss": 0.5115, "step": 7495 }, { "epoch": 0.91, "learning_rate": 4.071914758478601e-07, "loss": 0.4926, "step": 7496 }, { "epoch": 0.91, "learning_rate": 4.060796882517304e-07, "loss": 0.4697, "step": 7497 }, { "epoch": 0.91, "learning_rate": 4.049693890758033e-07, "loss": 0.5161, "step": 7498 }, { "epoch": 0.91, "learning_rate": 4.038605784923333e-07, "loss": 0.4689, "step": 7499 }, { "epoch": 0.91, "learning_rate": 4.0275325667334274e-07, "loss": 0.478, "step": 7500 }, { "epoch": 0.91, "learning_rate": 4.0164742379062317e-07, "loss": 0.5597, "step": 7501 }, { "epoch": 0.91, "learning_rate": 4.0054308001573614e-07, "loss": 0.5075, "step": 7502 }, { "epoch": 0.91, "learning_rate": 3.994402255200125e-07, "loss": 0.5211, "step": 7503 }, { "epoch": 0.91, "learning_rate": 3.9833886047454973e-07, "loss": 0.462, "step": 7504 }, { "epoch": 0.91, "learning_rate": 3.9723898505021783e-07, "loss": 0.4684, "step": 7505 }, { "epoch": 0.91, "learning_rate": 3.9614059941765146e-07, "loss": 0.4984, "step": 7506 }, { "epoch": 0.91, "learning_rate": 3.950437037472588e-07, "loss": 0.5598, "step": 7507 }, { "epoch": 0.91, "learning_rate": 3.9394829820921245e-07, "loss": 0.4887, "step": 7508 }, { "epoch": 0.91, "learning_rate": 3.9285438297345657e-07, "loss": 0.5458, "step": 7509 }, { "epoch": 0.91, "learning_rate": 3.917619582097021e-07, "loss": 0.4954, "step": 7510 }, { "epoch": 0.91, "learning_rate": 3.906710240874334e-07, "loss": 0.4771, "step": 7511 }, { "epoch": 0.91, "learning_rate": 3.8958158077589847e-07, "loss": 0.5542, "step": 7512 }, { "epoch": 0.91, "learning_rate": 3.884936284441132e-07, "loss": 0.4782, "step": 7513 }, { "epoch": 0.91, "learning_rate": 3.8740716726086925e-07, "loss": 0.4571, "step": 7514 }, { "epoch": 0.91, "learning_rate": 3.863221973947184e-07, "loss": 0.4658, "step": 7515 }, { "epoch": 0.91, "learning_rate": 3.8523871901398725e-07, "loss": 0.4758, "step": 7516 }, { "epoch": 0.91, "learning_rate": 3.8415673228676896e-07, "loss": 0.4937, "step": 7517 }, { "epoch": 0.91, "learning_rate": 3.830762373809238e-07, "loss": 0.4995, "step": 7518 }, { "epoch": 0.91, "learning_rate": 3.819972344640832e-07, "loss": 0.5121, "step": 7519 }, { "epoch": 0.91, "learning_rate": 3.809197237036455e-07, "loss": 0.4986, "step": 7520 }, { "epoch": 0.91, "learning_rate": 3.798437052667758e-07, "loss": 0.5101, "step": 7521 }, { "epoch": 0.91, "learning_rate": 3.78769179320414e-07, "loss": 0.5485, "step": 7522 }, { "epoch": 0.91, "learning_rate": 3.776961460312611e-07, "loss": 0.505, "step": 7523 }, { "epoch": 0.91, "learning_rate": 3.766246055657896e-07, "loss": 0.5298, "step": 7524 }, { "epoch": 0.92, "learning_rate": 3.755545580902431e-07, "loss": 0.4692, "step": 7525 }, { "epoch": 0.92, "learning_rate": 3.7448600377062885e-07, "loss": 0.545, "step": 7526 }, { "epoch": 0.92, "learning_rate": 3.7341894277272306e-07, "loss": 0.5258, "step": 7527 }, { "epoch": 0.92, "learning_rate": 3.723533752620756e-07, "loss": 0.5071, "step": 7528 }, { "epoch": 0.92, "learning_rate": 3.7128930140399643e-07, "loss": 0.4602, "step": 7529 }, { "epoch": 0.92, "learning_rate": 3.702267213635724e-07, "loss": 0.5023, "step": 7530 }, { "epoch": 0.92, "learning_rate": 3.6916563530565275e-07, "loss": 0.5865, "step": 7531 }, { "epoch": 0.92, "learning_rate": 3.6810604339485354e-07, "loss": 0.4467, "step": 7532 }, { "epoch": 0.92, "learning_rate": 3.670479457955656e-07, "loss": 0.5108, "step": 7533 }, { "epoch": 0.92, "learning_rate": 3.6599134267194415e-07, "loss": 0.5432, "step": 7534 }, { "epoch": 0.92, "learning_rate": 3.649362341879115e-07, "loss": 0.5276, "step": 7535 }, { "epoch": 0.92, "learning_rate": 3.6388262050715883e-07, "loss": 0.5499, "step": 7536 }, { "epoch": 0.92, "learning_rate": 3.628305017931477e-07, "loss": 0.4397, "step": 7537 }, { "epoch": 0.92, "learning_rate": 3.617798782091042e-07, "loss": 0.4921, "step": 7538 }, { "epoch": 0.92, "learning_rate": 3.607307499180257e-07, "loss": 0.4679, "step": 7539 }, { "epoch": 0.92, "learning_rate": 3.596831170826753e-07, "loss": 0.5198, "step": 7540 }, { "epoch": 0.92, "learning_rate": 3.5863697986558423e-07, "loss": 0.4757, "step": 7541 }, { "epoch": 0.92, "learning_rate": 3.575923384290536e-07, "loss": 0.4633, "step": 7542 }, { "epoch": 0.92, "learning_rate": 3.5654919293515056e-07, "loss": 0.4803, "step": 7543 }, { "epoch": 0.92, "learning_rate": 3.5550754354571005e-07, "loss": 0.5707, "step": 7544 }, { "epoch": 0.92, "learning_rate": 3.544673904223361e-07, "loss": 0.5147, "step": 7545 }, { "epoch": 0.92, "learning_rate": 3.53428733726402e-07, "loss": 0.4589, "step": 7546 }, { "epoch": 0.92, "learning_rate": 3.523915736190453e-07, "loss": 0.466, "step": 7547 }, { "epoch": 0.92, "learning_rate": 3.51355910261173e-07, "loss": 0.4829, "step": 7548 }, { "epoch": 0.92, "learning_rate": 3.5032174381346095e-07, "loss": 0.4979, "step": 7549 }, { "epoch": 0.92, "learning_rate": 3.492890744363508e-07, "loss": 0.4828, "step": 7550 }, { "epoch": 0.92, "learning_rate": 3.482579022900545e-07, "loss": 0.5348, "step": 7551 }, { "epoch": 0.92, "learning_rate": 3.4722822753454846e-07, "loss": 0.5132, "step": 7552 }, { "epoch": 0.92, "learning_rate": 3.462000503295782e-07, "loss": 0.501, "step": 7553 }, { "epoch": 0.92, "learning_rate": 3.451733708346605e-07, "loss": 0.5394, "step": 7554 }, { "epoch": 0.92, "learning_rate": 3.4414818920907146e-07, "loss": 0.5005, "step": 7555 }, { "epoch": 0.92, "learning_rate": 3.4312450561186484e-07, "loss": 0.4781, "step": 7556 }, { "epoch": 0.92, "learning_rate": 3.4210232020185253e-07, "loss": 0.5033, "step": 7557 }, { "epoch": 0.92, "learning_rate": 3.410816331376232e-07, "loss": 0.4758, "step": 7558 }, { "epoch": 0.92, "learning_rate": 3.400624445775236e-07, "loss": 0.505, "step": 7559 }, { "epoch": 0.92, "learning_rate": 3.39044754679676e-07, "loss": 0.4483, "step": 7560 }, { "epoch": 0.92, "learning_rate": 3.3802856360196423e-07, "loss": 0.5477, "step": 7561 }, { "epoch": 0.92, "learning_rate": 3.3701387150204547e-07, "loss": 0.4679, "step": 7562 }, { "epoch": 0.92, "learning_rate": 3.360006785373393e-07, "loss": 0.4707, "step": 7563 }, { "epoch": 0.92, "learning_rate": 3.349889848650334e-07, "loss": 0.4752, "step": 7564 }, { "epoch": 0.92, "learning_rate": 3.339787906420866e-07, "loss": 0.4872, "step": 7565 }, { "epoch": 0.92, "learning_rate": 3.329700960252202e-07, "loss": 0.5566, "step": 7566 }, { "epoch": 0.92, "learning_rate": 3.319629011709258e-07, "loss": 0.5112, "step": 7567 }, { "epoch": 0.92, "learning_rate": 3.309572062354616e-07, "loss": 0.5213, "step": 7568 }, { "epoch": 0.92, "learning_rate": 3.299530113748528e-07, "loss": 0.5005, "step": 7569 }, { "epoch": 0.92, "learning_rate": 3.2895031674489266e-07, "loss": 0.4477, "step": 7570 }, { "epoch": 0.92, "learning_rate": 3.279491225011433e-07, "loss": 0.5314, "step": 7571 }, { "epoch": 0.92, "learning_rate": 3.2694942879892946e-07, "loss": 0.5341, "step": 7572 }, { "epoch": 0.92, "learning_rate": 3.259512357933448e-07, "loss": 0.4963, "step": 7573 }, { "epoch": 0.92, "learning_rate": 3.2495454363925425e-07, "loss": 0.5038, "step": 7574 }, { "epoch": 0.92, "learning_rate": 3.2395935249128427e-07, "loss": 0.5053, "step": 7575 }, { "epoch": 0.92, "learning_rate": 3.229656625038302e-07, "loss": 0.5276, "step": 7576 }, { "epoch": 0.92, "learning_rate": 3.219734738310576e-07, "loss": 0.4668, "step": 7577 }, { "epoch": 0.92, "learning_rate": 3.2098278662689443e-07, "loss": 0.4583, "step": 7578 }, { "epoch": 0.92, "learning_rate": 3.1999360104503995e-07, "loss": 0.4813, "step": 7579 }, { "epoch": 0.92, "learning_rate": 3.19005917238957e-07, "loss": 0.5477, "step": 7580 }, { "epoch": 0.92, "learning_rate": 3.180197353618764e-07, "loss": 0.5419, "step": 7581 }, { "epoch": 0.92, "learning_rate": 3.170350555667967e-07, "loss": 0.4636, "step": 7582 }, { "epoch": 0.92, "learning_rate": 3.1605187800648585e-07, "loss": 0.4668, "step": 7583 }, { "epoch": 0.92, "learning_rate": 3.150702028334707e-07, "loss": 0.5748, "step": 7584 }, { "epoch": 0.92, "learning_rate": 3.1409003020005603e-07, "loss": 0.4331, "step": 7585 }, { "epoch": 0.92, "learning_rate": 3.131113602583047e-07, "loss": 0.5231, "step": 7586 }, { "epoch": 0.92, "learning_rate": 3.1213419316004856e-07, "loss": 0.4731, "step": 7587 }, { "epoch": 0.92, "learning_rate": 3.111585290568897e-07, "loss": 0.4846, "step": 7588 }, { "epoch": 0.92, "learning_rate": 3.101843681001937e-07, "loss": 0.4623, "step": 7589 }, { "epoch": 0.92, "learning_rate": 3.092117104410919e-07, "loss": 0.4988, "step": 7590 }, { "epoch": 0.92, "learning_rate": 3.08240556230488e-07, "loss": 0.5056, "step": 7591 }, { "epoch": 0.92, "learning_rate": 3.0727090561904707e-07, "loss": 0.475, "step": 7592 }, { "epoch": 0.92, "learning_rate": 3.0630275875720096e-07, "loss": 0.4655, "step": 7593 }, { "epoch": 0.92, "learning_rate": 3.0533611579515286e-07, "loss": 0.5024, "step": 7594 }, { "epoch": 0.92, "learning_rate": 3.0437097688286955e-07, "loss": 0.4873, "step": 7595 }, { "epoch": 0.92, "learning_rate": 3.034073421700812e-07, "loss": 0.5064, "step": 7596 }, { "epoch": 0.92, "learning_rate": 3.024452118062926e-07, "loss": 0.4229, "step": 7597 }, { "epoch": 0.92, "learning_rate": 3.014845859407678e-07, "loss": 0.4306, "step": 7598 }, { "epoch": 0.92, "learning_rate": 3.0052546472253977e-07, "loss": 0.4769, "step": 7599 }, { "epoch": 0.92, "learning_rate": 2.995678483004116e-07, "loss": 0.4715, "step": 7600 }, { "epoch": 0.92, "learning_rate": 2.9861173682294576e-07, "loss": 0.5539, "step": 7601 }, { "epoch": 0.92, "learning_rate": 2.97657130438479e-07, "loss": 0.5302, "step": 7602 }, { "epoch": 0.92, "learning_rate": 2.9670402929510846e-07, "loss": 0.5445, "step": 7603 }, { "epoch": 0.92, "learning_rate": 2.957524335407014e-07, "loss": 0.473, "step": 7604 }, { "epoch": 0.92, "learning_rate": 2.948023433228886e-07, "loss": 0.5182, "step": 7605 }, { "epoch": 0.92, "learning_rate": 2.938537587890722e-07, "loss": 0.4647, "step": 7606 }, { "epoch": 0.93, "learning_rate": 2.929066800864133e-07, "loss": 0.5216, "step": 7607 }, { "epoch": 0.93, "learning_rate": 2.919611073618478e-07, "loss": 0.4907, "step": 7608 }, { "epoch": 0.93, "learning_rate": 2.910170407620727e-07, "loss": 0.489, "step": 7609 }, { "epoch": 0.93, "learning_rate": 2.900744804335498e-07, "loss": 0.4813, "step": 7610 }, { "epoch": 0.93, "learning_rate": 2.891334265225121e-07, "loss": 0.5004, "step": 7611 }, { "epoch": 0.93, "learning_rate": 2.881938791749561e-07, "loss": 0.4638, "step": 7612 }, { "epoch": 0.93, "learning_rate": 2.872558385366453e-07, "loss": 0.5331, "step": 7613 }, { "epoch": 0.93, "learning_rate": 2.8631930475310875e-07, "loss": 0.4606, "step": 7614 }, { "epoch": 0.93, "learning_rate": 2.8538427796964365e-07, "loss": 0.5167, "step": 7615 }, { "epoch": 0.93, "learning_rate": 2.844507583313083e-07, "loss": 0.5321, "step": 7616 }, { "epoch": 0.93, "learning_rate": 2.835187459829347e-07, "loss": 0.4921, "step": 7617 }, { "epoch": 0.93, "learning_rate": 2.825882410691161e-07, "loss": 0.5414, "step": 7618 }, { "epoch": 0.93, "learning_rate": 2.816592437342114e-07, "loss": 0.4486, "step": 7619 }, { "epoch": 0.93, "learning_rate": 2.807317541223487e-07, "loss": 0.4904, "step": 7620 }, { "epoch": 0.93, "learning_rate": 2.7980577237742055e-07, "loss": 0.5531, "step": 7621 }, { "epoch": 0.93, "learning_rate": 2.788812986430844e-07, "loss": 0.4913, "step": 7622 }, { "epoch": 0.93, "learning_rate": 2.7795833306276665e-07, "loss": 0.5533, "step": 7623 }, { "epoch": 0.93, "learning_rate": 2.770368757796571e-07, "loss": 0.5057, "step": 7624 }, { "epoch": 0.93, "learning_rate": 2.7611692693671254e-07, "loss": 0.5341, "step": 7625 }, { "epoch": 0.93, "learning_rate": 2.751984866766566e-07, "loss": 0.475, "step": 7626 }, { "epoch": 0.93, "learning_rate": 2.742815551419753e-07, "loss": 0.4736, "step": 7627 }, { "epoch": 0.93, "learning_rate": 2.7336613247492594e-07, "loss": 0.5586, "step": 7628 }, { "epoch": 0.93, "learning_rate": 2.724522188175283e-07, "loss": 0.5002, "step": 7629 }, { "epoch": 0.93, "learning_rate": 2.7153981431156775e-07, "loss": 0.5119, "step": 7630 }, { "epoch": 0.93, "learning_rate": 2.706289190985989e-07, "loss": 0.5455, "step": 7631 }, { "epoch": 0.93, "learning_rate": 2.697195333199387e-07, "loss": 0.4812, "step": 7632 }, { "epoch": 0.93, "learning_rate": 2.688116571166688e-07, "loss": 0.5546, "step": 7633 }, { "epoch": 0.93, "learning_rate": 2.679052906296431e-07, "loss": 0.5666, "step": 7634 }, { "epoch": 0.93, "learning_rate": 2.670004339994747e-07, "loss": 0.5586, "step": 7635 }, { "epoch": 0.93, "learning_rate": 2.660970873665447e-07, "loss": 0.4951, "step": 7636 }, { "epoch": 0.93, "learning_rate": 2.6519525087100096e-07, "loss": 0.5262, "step": 7637 }, { "epoch": 0.93, "learning_rate": 2.6429492465275595e-07, "loss": 0.4306, "step": 7638 }, { "epoch": 0.93, "learning_rate": 2.633961088514869e-07, "loss": 0.4593, "step": 7639 }, { "epoch": 0.93, "learning_rate": 2.6249880360664116e-07, "loss": 0.4759, "step": 7640 }, { "epoch": 0.93, "learning_rate": 2.61603009057424e-07, "loss": 0.5158, "step": 7641 }, { "epoch": 0.93, "learning_rate": 2.6070872534281424e-07, "loss": 0.5358, "step": 7642 }, { "epoch": 0.93, "learning_rate": 2.5981595260155204e-07, "loss": 0.5174, "step": 7643 }, { "epoch": 0.93, "learning_rate": 2.5892469097214433e-07, "loss": 0.5188, "step": 7644 }, { "epoch": 0.93, "learning_rate": 2.580349405928617e-07, "loss": 0.5189, "step": 7645 }, { "epoch": 0.93, "learning_rate": 2.571467016017437e-07, "loss": 0.503, "step": 7646 }, { "epoch": 0.93, "learning_rate": 2.562599741365912e-07, "loss": 0.5497, "step": 7647 }, { "epoch": 0.93, "learning_rate": 2.5537475833497526e-07, "loss": 0.4617, "step": 7648 }, { "epoch": 0.93, "learning_rate": 2.5449105433422936e-07, "loss": 0.5117, "step": 7649 }, { "epoch": 0.93, "learning_rate": 2.5360886227145167e-07, "loss": 0.4643, "step": 7650 }, { "epoch": 0.93, "learning_rate": 2.527281822835104e-07, "loss": 0.4986, "step": 7651 }, { "epoch": 0.93, "learning_rate": 2.518490145070329e-07, "loss": 0.5304, "step": 7652 }, { "epoch": 0.93, "learning_rate": 2.5097135907841464e-07, "loss": 0.4835, "step": 7653 }, { "epoch": 0.93, "learning_rate": 2.5009521613382105e-07, "loss": 0.512, "step": 7654 }, { "epoch": 0.93, "learning_rate": 2.492205858091745e-07, "loss": 0.4746, "step": 7655 }, { "epoch": 0.93, "learning_rate": 2.483474682401688e-07, "loss": 0.5646, "step": 7656 }, { "epoch": 0.93, "learning_rate": 2.474758635622621e-07, "loss": 0.4735, "step": 7657 }, { "epoch": 0.93, "learning_rate": 2.4660577191067516e-07, "loss": 0.5112, "step": 7658 }, { "epoch": 0.93, "learning_rate": 2.457371934203945e-07, "loss": 0.4583, "step": 7659 }, { "epoch": 0.93, "learning_rate": 2.4487012822617673e-07, "loss": 0.4696, "step": 7660 }, { "epoch": 0.93, "learning_rate": 2.4400457646253874e-07, "loss": 0.5366, "step": 7661 }, { "epoch": 0.93, "learning_rate": 2.431405382637619e-07, "loss": 0.5538, "step": 7662 }, { "epoch": 0.93, "learning_rate": 2.4227801376389694e-07, "loss": 0.4616, "step": 7663 }, { "epoch": 0.93, "learning_rate": 2.4141700309675777e-07, "loss": 0.5239, "step": 7664 }, { "epoch": 0.93, "learning_rate": 2.405575063959209e-07, "loss": 0.4413, "step": 7665 }, { "epoch": 0.93, "learning_rate": 2.3969952379473414e-07, "loss": 0.4923, "step": 7666 }, { "epoch": 0.93, "learning_rate": 2.388430554263055e-07, "loss": 0.5052, "step": 7667 }, { "epoch": 0.93, "learning_rate": 2.3798810142350636e-07, "loss": 0.5142, "step": 7668 }, { "epoch": 0.93, "learning_rate": 2.371346619189796e-07, "loss": 0.5581, "step": 7669 }, { "epoch": 0.93, "learning_rate": 2.3628273704512705e-07, "loss": 0.4703, "step": 7670 }, { "epoch": 0.93, "learning_rate": 2.3543232693411967e-07, "loss": 0.4853, "step": 7671 }, { "epoch": 0.93, "learning_rate": 2.3458343171789187e-07, "loss": 0.4107, "step": 7672 }, { "epoch": 0.93, "learning_rate": 2.337360515281406e-07, "loss": 0.5844, "step": 7673 }, { "epoch": 0.93, "learning_rate": 2.3289018649633288e-07, "loss": 0.4655, "step": 7674 }, { "epoch": 0.93, "learning_rate": 2.3204583675369706e-07, "loss": 0.4421, "step": 7675 }, { "epoch": 0.93, "learning_rate": 2.3120300243122618e-07, "loss": 0.4841, "step": 7676 }, { "epoch": 0.93, "learning_rate": 2.3036168365968114e-07, "loss": 0.507, "step": 7677 }, { "epoch": 0.93, "learning_rate": 2.2952188056958423e-07, "loss": 0.4847, "step": 7678 }, { "epoch": 0.93, "learning_rate": 2.2868359329122458e-07, "loss": 0.5319, "step": 7679 }, { "epoch": 0.93, "learning_rate": 2.27846821954657e-07, "loss": 0.5117, "step": 7680 }, { "epoch": 0.93, "learning_rate": 2.270115666896988e-07, "loss": 0.5312, "step": 7681 }, { "epoch": 0.93, "learning_rate": 2.2617782762593298e-07, "loss": 0.4834, "step": 7682 }, { "epoch": 0.93, "learning_rate": 2.253456048927083e-07, "loss": 0.5209, "step": 7683 }, { "epoch": 0.93, "learning_rate": 2.24514898619137e-07, "loss": 0.5599, "step": 7684 }, { "epoch": 0.93, "learning_rate": 2.2368570893409492e-07, "loss": 0.4501, "step": 7685 }, { "epoch": 0.93, "learning_rate": 2.2285803596622802e-07, "loss": 0.5114, "step": 7686 }, { "epoch": 0.93, "learning_rate": 2.2203187984393916e-07, "loss": 0.525, "step": 7687 }, { "epoch": 0.93, "learning_rate": 2.2120724069540246e-07, "loss": 0.4848, "step": 7688 }, { "epoch": 0.94, "learning_rate": 2.2038411864855224e-07, "loss": 0.5213, "step": 7689 }, { "epoch": 0.94, "learning_rate": 2.1956251383109195e-07, "loss": 0.4992, "step": 7690 }, { "epoch": 0.94, "learning_rate": 2.1874242637048514e-07, "loss": 0.4582, "step": 7691 }, { "epoch": 0.94, "learning_rate": 2.179238563939623e-07, "loss": 0.4267, "step": 7692 }, { "epoch": 0.94, "learning_rate": 2.1710680402851737e-07, "loss": 0.526, "step": 7693 }, { "epoch": 0.94, "learning_rate": 2.1629126940091005e-07, "loss": 0.5363, "step": 7694 }, { "epoch": 0.94, "learning_rate": 2.1547725263766473e-07, "loss": 0.5075, "step": 7695 }, { "epoch": 0.94, "learning_rate": 2.1466475386506813e-07, "loss": 0.5335, "step": 7696 }, { "epoch": 0.94, "learning_rate": 2.138537732091739e-07, "loss": 0.5434, "step": 7697 }, { "epoch": 0.94, "learning_rate": 2.1304431079580024e-07, "loss": 0.524, "step": 7698 }, { "epoch": 0.94, "learning_rate": 2.122363667505256e-07, "loss": 0.4729, "step": 7699 }, { "epoch": 0.94, "learning_rate": 2.1142994119869975e-07, "loss": 0.508, "step": 7700 }, { "epoch": 0.94, "learning_rate": 2.1062503426543036e-07, "loss": 0.4856, "step": 7701 }, { "epoch": 0.94, "learning_rate": 2.0982164607559308e-07, "loss": 0.5178, "step": 7702 }, { "epoch": 0.94, "learning_rate": 2.0901977675382934e-07, "loss": 0.5287, "step": 7703 }, { "epoch": 0.94, "learning_rate": 2.0821942642453964e-07, "loss": 0.5214, "step": 7704 }, { "epoch": 0.94, "learning_rate": 2.074205952118935e-07, "loss": 0.4658, "step": 7705 }, { "epoch": 0.94, "learning_rate": 2.0662328323982296e-07, "loss": 0.4809, "step": 7706 }, { "epoch": 0.94, "learning_rate": 2.0582749063202566e-07, "loss": 0.5287, "step": 7707 }, { "epoch": 0.94, "learning_rate": 2.050332175119596e-07, "loss": 0.5297, "step": 7708 }, { "epoch": 0.94, "learning_rate": 2.0424046400285392e-07, "loss": 0.5406, "step": 7709 }, { "epoch": 0.94, "learning_rate": 2.0344923022769357e-07, "loss": 0.494, "step": 7710 }, { "epoch": 0.94, "learning_rate": 2.0265951630923596e-07, "loss": 0.5198, "step": 7711 }, { "epoch": 0.94, "learning_rate": 2.0187132236999752e-07, "loss": 0.4851, "step": 7712 }, { "epoch": 0.94, "learning_rate": 2.0108464853225928e-07, "loss": 0.5091, "step": 7713 }, { "epoch": 0.94, "learning_rate": 2.002994949180681e-07, "loss": 0.476, "step": 7714 }, { "epoch": 0.94, "learning_rate": 1.9951586164923542e-07, "loss": 0.503, "step": 7715 }, { "epoch": 0.94, "learning_rate": 1.9873374884733399e-07, "loss": 0.4678, "step": 7716 }, { "epoch": 0.94, "learning_rate": 1.979531566337034e-07, "loss": 0.5139, "step": 7717 }, { "epoch": 0.94, "learning_rate": 1.971740851294457e-07, "loss": 0.4803, "step": 7718 }, { "epoch": 0.94, "learning_rate": 1.963965344554286e-07, "loss": 0.5725, "step": 7719 }, { "epoch": 0.94, "learning_rate": 1.9562050473228121e-07, "loss": 0.5002, "step": 7720 }, { "epoch": 0.94, "learning_rate": 1.9484599608040056e-07, "loss": 0.4877, "step": 7721 }, { "epoch": 0.94, "learning_rate": 1.9407300861994383e-07, "loss": 0.5568, "step": 7722 }, { "epoch": 0.94, "learning_rate": 1.9330154247083398e-07, "loss": 0.5041, "step": 7723 }, { "epoch": 0.94, "learning_rate": 1.9253159775275864e-07, "loss": 0.5166, "step": 7724 }, { "epoch": 0.94, "learning_rate": 1.917631745851667e-07, "loss": 0.5218, "step": 7725 }, { "epoch": 0.94, "learning_rate": 1.909962730872761e-07, "loss": 0.5108, "step": 7726 }, { "epoch": 0.94, "learning_rate": 1.9023089337806277e-07, "loss": 0.504, "step": 7727 }, { "epoch": 0.94, "learning_rate": 1.8946703557627067e-07, "loss": 0.5037, "step": 7728 }, { "epoch": 0.94, "learning_rate": 1.8870469980040496e-07, "loss": 0.5572, "step": 7729 }, { "epoch": 0.94, "learning_rate": 1.8794388616873771e-07, "loss": 0.473, "step": 7730 }, { "epoch": 0.94, "learning_rate": 1.8718459479930119e-07, "loss": 0.5159, "step": 7731 }, { "epoch": 0.94, "learning_rate": 1.8642682580989556e-07, "loss": 0.5302, "step": 7732 }, { "epoch": 0.94, "learning_rate": 1.856705793180802e-07, "loss": 0.5316, "step": 7733 }, { "epoch": 0.94, "learning_rate": 1.849158554411834e-07, "loss": 0.4423, "step": 7734 }, { "epoch": 0.94, "learning_rate": 1.8416265429629265e-07, "loss": 0.5107, "step": 7735 }, { "epoch": 0.94, "learning_rate": 1.8341097600026003e-07, "loss": 0.4825, "step": 7736 }, { "epoch": 0.94, "learning_rate": 1.826608206697045e-07, "loss": 0.5342, "step": 7737 }, { "epoch": 0.94, "learning_rate": 1.8191218842100623e-07, "loss": 0.5102, "step": 7738 }, { "epoch": 0.94, "learning_rate": 1.8116507937031014e-07, "loss": 0.5369, "step": 7739 }, { "epoch": 0.94, "learning_rate": 1.804194936335235e-07, "loss": 0.5332, "step": 7740 }, { "epoch": 0.94, "learning_rate": 1.7967543132631714e-07, "loss": 0.5073, "step": 7741 }, { "epoch": 0.94, "learning_rate": 1.789328925641276e-07, "loss": 0.5137, "step": 7742 }, { "epoch": 0.94, "learning_rate": 1.7819187746215382e-07, "loss": 0.4922, "step": 7743 }, { "epoch": 0.94, "learning_rate": 1.7745238613535832e-07, "loss": 0.5557, "step": 7744 }, { "epoch": 0.94, "learning_rate": 1.7671441869846595e-07, "loss": 0.4368, "step": 7745 }, { "epoch": 0.94, "learning_rate": 1.7597797526596848e-07, "loss": 0.4991, "step": 7746 }, { "epoch": 0.94, "learning_rate": 1.7524305595211788e-07, "loss": 0.5375, "step": 7747 }, { "epoch": 0.94, "learning_rate": 1.745096608709318e-07, "loss": 0.5348, "step": 7748 }, { "epoch": 0.94, "learning_rate": 1.7377779013618923e-07, "loss": 0.4502, "step": 7749 }, { "epoch": 0.94, "learning_rate": 1.7304744386143602e-07, "loss": 0.5022, "step": 7750 }, { "epoch": 0.94, "learning_rate": 1.723186221599782e-07, "loss": 0.5036, "step": 7751 }, { "epoch": 0.94, "learning_rate": 1.7159132514488753e-07, "loss": 0.4816, "step": 7752 }, { "epoch": 0.94, "learning_rate": 1.7086555292899821e-07, "loss": 0.5253, "step": 7753 }, { "epoch": 0.94, "learning_rate": 1.7014130562490682e-07, "loss": 0.4759, "step": 7754 }, { "epoch": 0.94, "learning_rate": 1.6941858334497684e-07, "loss": 0.4866, "step": 7755 }, { "epoch": 0.94, "learning_rate": 1.686973862013308e-07, "loss": 0.4577, "step": 7756 }, { "epoch": 0.94, "learning_rate": 1.6797771430585697e-07, "loss": 0.5474, "step": 7757 }, { "epoch": 0.94, "learning_rate": 1.6725956777020713e-07, "loss": 0.5581, "step": 7758 }, { "epoch": 0.94, "learning_rate": 1.6654294670579552e-07, "loss": 0.5074, "step": 7759 }, { "epoch": 0.94, "learning_rate": 1.6582785122380208e-07, "loss": 0.5163, "step": 7760 }, { "epoch": 0.94, "learning_rate": 1.6511428143516473e-07, "loss": 0.5152, "step": 7761 }, { "epoch": 0.94, "learning_rate": 1.644022374505916e-07, "loss": 0.544, "step": 7762 }, { "epoch": 0.94, "learning_rate": 1.6369171938054874e-07, "loss": 0.5223, "step": 7763 }, { "epoch": 0.94, "learning_rate": 1.62982727335268e-07, "loss": 0.5312, "step": 7764 }, { "epoch": 0.94, "learning_rate": 1.6227526142474248e-07, "loss": 0.4722, "step": 7765 }, { "epoch": 0.94, "learning_rate": 1.6156932175873218e-07, "loss": 0.4881, "step": 7766 }, { "epoch": 0.94, "learning_rate": 1.6086490844675727e-07, "loss": 0.4847, "step": 7767 }, { "epoch": 0.94, "learning_rate": 1.6016202159810034e-07, "loss": 0.4727, "step": 7768 }, { "epoch": 0.94, "learning_rate": 1.594606613218097e-07, "loss": 0.5309, "step": 7769 }, { "epoch": 0.94, "learning_rate": 1.587608277266961e-07, "loss": 0.5496, "step": 7770 }, { "epoch": 0.94, "learning_rate": 1.5806252092133268e-07, "loss": 0.4902, "step": 7771 }, { "epoch": 0.95, "learning_rate": 1.5736574101405722e-07, "loss": 0.4694, "step": 7772 }, { "epoch": 0.95, "learning_rate": 1.566704881129677e-07, "loss": 0.5441, "step": 7773 }, { "epoch": 0.95, "learning_rate": 1.5597676232592785e-07, "loss": 0.5262, "step": 7774 }, { "epoch": 0.95, "learning_rate": 1.552845637605649e-07, "loss": 0.5153, "step": 7775 }, { "epoch": 0.95, "learning_rate": 1.5459389252426738e-07, "loss": 0.5067, "step": 7776 }, { "epoch": 0.95, "learning_rate": 1.5390474872418515e-07, "loss": 0.5447, "step": 7777 }, { "epoch": 0.95, "learning_rate": 1.5321713246723713e-07, "loss": 0.4554, "step": 7778 }, { "epoch": 0.95, "learning_rate": 1.525310438600991e-07, "loss": 0.4902, "step": 7779 }, { "epoch": 0.95, "learning_rate": 1.5184648300921146e-07, "loss": 0.4555, "step": 7780 }, { "epoch": 0.95, "learning_rate": 1.5116345002078148e-07, "loss": 0.4736, "step": 7781 }, { "epoch": 0.95, "learning_rate": 1.5048194500077328e-07, "loss": 0.4346, "step": 7782 }, { "epoch": 0.95, "learning_rate": 1.4980196805491788e-07, "loss": 0.5306, "step": 7783 }, { "epoch": 0.95, "learning_rate": 1.491235192887097e-07, "loss": 0.4754, "step": 7784 }, { "epoch": 0.95, "learning_rate": 1.484465988074013e-07, "loss": 0.489, "step": 7785 }, { "epoch": 0.95, "learning_rate": 1.477712067160153e-07, "loss": 0.5543, "step": 7786 }, { "epoch": 0.95, "learning_rate": 1.470973431193312e-07, "loss": 0.5185, "step": 7787 }, { "epoch": 0.95, "learning_rate": 1.464250081218943e-07, "loss": 0.4587, "step": 7788 }, { "epoch": 0.95, "learning_rate": 1.4575420182801114e-07, "loss": 0.51, "step": 7789 }, { "epoch": 0.95, "learning_rate": 1.4508492434175402e-07, "loss": 0.5535, "step": 7790 }, { "epoch": 0.95, "learning_rate": 1.4441717576695325e-07, "loss": 0.4878, "step": 7791 }, { "epoch": 0.95, "learning_rate": 1.4375095620720702e-07, "loss": 0.4401, "step": 7792 }, { "epoch": 0.95, "learning_rate": 1.430862657658727e-07, "loss": 0.4941, "step": 7793 }, { "epoch": 0.95, "learning_rate": 1.4242310454607111e-07, "loss": 0.4998, "step": 7794 }, { "epoch": 0.95, "learning_rate": 1.4176147265068884e-07, "loss": 0.4918, "step": 7795 }, { "epoch": 0.95, "learning_rate": 1.4110137018237046e-07, "loss": 0.5362, "step": 7796 }, { "epoch": 0.95, "learning_rate": 1.4044279724352516e-07, "loss": 0.5399, "step": 7797 }, { "epoch": 0.95, "learning_rate": 1.3978575393632788e-07, "loss": 0.4696, "step": 7798 }, { "epoch": 0.95, "learning_rate": 1.391302403627126e-07, "loss": 0.5401, "step": 7799 }, { "epoch": 0.95, "learning_rate": 1.384762566243758e-07, "loss": 0.522, "step": 7800 }, { "epoch": 0.95, "learning_rate": 1.3782380282277964e-07, "loss": 0.4749, "step": 7801 }, { "epoch": 0.95, "learning_rate": 1.3717287905914645e-07, "loss": 0.5608, "step": 7802 }, { "epoch": 0.95, "learning_rate": 1.3652348543446104e-07, "loss": 0.4894, "step": 7803 }, { "epoch": 0.95, "learning_rate": 1.358756220494728e-07, "loss": 0.4787, "step": 7804 }, { "epoch": 0.95, "learning_rate": 1.3522928900469022e-07, "loss": 0.5128, "step": 7805 }, { "epoch": 0.95, "learning_rate": 1.345844864003909e-07, "loss": 0.4764, "step": 7806 }, { "epoch": 0.95, "learning_rate": 1.3394121433660812e-07, "loss": 0.4952, "step": 7807 }, { "epoch": 0.95, "learning_rate": 1.332994729131387e-07, "loss": 0.5292, "step": 7808 }, { "epoch": 0.95, "learning_rate": 1.326592622295475e-07, "loss": 0.4984, "step": 7809 }, { "epoch": 0.95, "learning_rate": 1.3202058238515613e-07, "loss": 0.5196, "step": 7810 }, { "epoch": 0.95, "learning_rate": 1.313834334790498e-07, "loss": 0.4961, "step": 7811 }, { "epoch": 0.95, "learning_rate": 1.3074781561007944e-07, "loss": 0.4741, "step": 7812 }, { "epoch": 0.95, "learning_rate": 1.3011372887685503e-07, "loss": 0.5071, "step": 7813 }, { "epoch": 0.95, "learning_rate": 1.2948117337774902e-07, "loss": 0.531, "step": 7814 }, { "epoch": 0.95, "learning_rate": 1.2885014921089845e-07, "loss": 0.5458, "step": 7815 }, { "epoch": 0.95, "learning_rate": 1.2822065647420167e-07, "loss": 0.4862, "step": 7816 }, { "epoch": 0.95, "learning_rate": 1.2759269526531947e-07, "loss": 0.4708, "step": 7817 }, { "epoch": 0.95, "learning_rate": 1.2696626568167391e-07, "loss": 0.547, "step": 7818 }, { "epoch": 0.95, "learning_rate": 1.263413678204528e-07, "loss": 0.4929, "step": 7819 }, { "epoch": 0.95, "learning_rate": 1.2571800177860083e-07, "loss": 0.5377, "step": 7820 }, { "epoch": 0.95, "learning_rate": 1.250961676528306e-07, "loss": 0.4978, "step": 7821 }, { "epoch": 0.95, "learning_rate": 1.24475865539615e-07, "loss": 0.5329, "step": 7822 }, { "epoch": 0.95, "learning_rate": 1.23857095535187e-07, "loss": 0.4826, "step": 7823 }, { "epoch": 0.95, "learning_rate": 1.2323985773554535e-07, "loss": 0.5283, "step": 7824 }, { "epoch": 0.95, "learning_rate": 1.226241522364502e-07, "loss": 0.4726, "step": 7825 }, { "epoch": 0.95, "learning_rate": 1.2200997913342171e-07, "loss": 0.4982, "step": 7826 }, { "epoch": 0.95, "learning_rate": 1.2139733852174484e-07, "loss": 0.5074, "step": 7827 }, { "epoch": 0.95, "learning_rate": 1.2078623049646464e-07, "loss": 0.4831, "step": 7828 }, { "epoch": 0.95, "learning_rate": 1.2017665515239197e-07, "loss": 0.495, "step": 7829 }, { "epoch": 0.95, "learning_rate": 1.1956861258409558e-07, "loss": 0.492, "step": 7830 }, { "epoch": 0.95, "learning_rate": 1.1896210288590893e-07, "loss": 0.4829, "step": 7831 }, { "epoch": 0.95, "learning_rate": 1.1835712615192785e-07, "loss": 0.4984, "step": 7832 }, { "epoch": 0.95, "learning_rate": 1.1775368247600838e-07, "loss": 0.5066, "step": 7833 }, { "epoch": 0.95, "learning_rate": 1.1715177195177007e-07, "loss": 0.5428, "step": 7834 }, { "epoch": 0.95, "learning_rate": 1.1655139467259714e-07, "loss": 0.486, "step": 7835 }, { "epoch": 0.95, "learning_rate": 1.1595255073163059e-07, "loss": 0.4962, "step": 7836 }, { "epoch": 0.95, "learning_rate": 1.1535524022177725e-07, "loss": 0.4665, "step": 7837 }, { "epoch": 0.95, "learning_rate": 1.147594632357052e-07, "loss": 0.6119, "step": 7838 }, { "epoch": 0.95, "learning_rate": 1.1416521986584383e-07, "loss": 0.4355, "step": 7839 }, { "epoch": 0.95, "learning_rate": 1.1357251020438498e-07, "loss": 0.4797, "step": 7840 }, { "epoch": 0.95, "learning_rate": 1.1298133434328507e-07, "loss": 0.5528, "step": 7841 }, { "epoch": 0.95, "learning_rate": 1.123916923742574e-07, "loss": 0.5345, "step": 7842 }, { "epoch": 0.95, "learning_rate": 1.1180358438878214e-07, "loss": 0.5164, "step": 7843 }, { "epoch": 0.95, "learning_rate": 1.1121701047809851e-07, "loss": 0.4624, "step": 7844 }, { "epoch": 0.95, "learning_rate": 1.106319707332093e-07, "loss": 0.4515, "step": 7845 }, { "epoch": 0.95, "learning_rate": 1.1004846524487856e-07, "loss": 0.4958, "step": 7846 }, { "epoch": 0.95, "learning_rate": 1.0946649410363386e-07, "loss": 0.5435, "step": 7847 }, { "epoch": 0.95, "learning_rate": 1.0888605739976077e-07, "loss": 0.5326, "step": 7848 }, { "epoch": 0.95, "learning_rate": 1.083071552233117e-07, "loss": 0.5231, "step": 7849 }, { "epoch": 0.95, "learning_rate": 1.0772978766409703e-07, "loss": 0.5256, "step": 7850 }, { "epoch": 0.95, "learning_rate": 1.0715395481169177e-07, "loss": 0.4969, "step": 7851 }, { "epoch": 0.95, "learning_rate": 1.0657965675543225e-07, "loss": 0.5202, "step": 7852 }, { "epoch": 0.95, "learning_rate": 1.0600689358441496e-07, "loss": 0.4908, "step": 7853 }, { "epoch": 0.96, "learning_rate": 1.0543566538749994e-07, "loss": 0.4934, "step": 7854 }, { "epoch": 0.96, "learning_rate": 1.0486597225330964e-07, "loss": 0.514, "step": 7855 }, { "epoch": 0.96, "learning_rate": 1.0429781427022556e-07, "loss": 0.5131, "step": 7856 }, { "epoch": 0.96, "learning_rate": 1.0373119152639389e-07, "loss": 0.504, "step": 7857 }, { "epoch": 0.96, "learning_rate": 1.0316610410972206e-07, "loss": 0.4305, "step": 7858 }, { "epoch": 0.96, "learning_rate": 1.0260255210787884e-07, "loss": 0.4435, "step": 7859 }, { "epoch": 0.96, "learning_rate": 1.020405356082943e-07, "loss": 0.5361, "step": 7860 }, { "epoch": 0.96, "learning_rate": 1.0148005469816091e-07, "loss": 0.5181, "step": 7861 }, { "epoch": 0.96, "learning_rate": 1.0092110946443246e-07, "loss": 0.5165, "step": 7862 }, { "epoch": 0.96, "learning_rate": 1.0036369999382511e-07, "loss": 0.4786, "step": 7863 }, { "epoch": 0.96, "learning_rate": 9.98078263728175e-08, "loss": 0.5622, "step": 7864 }, { "epoch": 0.96, "learning_rate": 9.925348868764839e-08, "loss": 0.5088, "step": 7865 }, { "epoch": 0.96, "learning_rate": 9.87006870243179e-08, "loss": 0.475, "step": 7866 }, { "epoch": 0.96, "learning_rate": 9.814942146858963e-08, "loss": 0.4851, "step": 7867 }, { "epoch": 0.96, "learning_rate": 9.759969210598741e-08, "loss": 0.5093, "step": 7868 }, { "epoch": 0.96, "learning_rate": 9.705149902179856e-08, "loss": 0.4878, "step": 7869 }, { "epoch": 0.96, "learning_rate": 9.65048423010695e-08, "loss": 0.5509, "step": 7870 }, { "epoch": 0.96, "learning_rate": 9.595972202861126e-08, "loss": 0.477, "step": 7871 }, { "epoch": 0.96, "learning_rate": 9.541613828899288e-08, "loss": 0.5116, "step": 7872 }, { "epoch": 0.96, "learning_rate": 9.48740911665491e-08, "loss": 0.5437, "step": 7873 }, { "epoch": 0.96, "learning_rate": 9.433358074537269e-08, "loss": 0.5098, "step": 7874 }, { "epoch": 0.96, "learning_rate": 9.379460710932097e-08, "loss": 0.4793, "step": 7875 }, { "epoch": 0.96, "learning_rate": 9.325717034200932e-08, "loss": 0.4805, "step": 7876 }, { "epoch": 0.96, "learning_rate": 9.272127052681879e-08, "loss": 0.5428, "step": 7877 }, { "epoch": 0.96, "learning_rate": 9.218690774688844e-08, "loss": 0.5566, "step": 7878 }, { "epoch": 0.96, "learning_rate": 9.165408208512194e-08, "loss": 0.5206, "step": 7879 }, { "epoch": 0.96, "learning_rate": 9.112279362418097e-08, "loss": 0.4849, "step": 7880 }, { "epoch": 0.96, "learning_rate": 9.059304244649181e-08, "loss": 0.4652, "step": 7881 }, { "epoch": 0.96, "learning_rate": 9.006482863424204e-08, "loss": 0.4924, "step": 7882 }, { "epoch": 0.96, "learning_rate": 8.953815226937834e-08, "loss": 0.5378, "step": 7883 }, { "epoch": 0.96, "learning_rate": 8.901301343361091e-08, "loss": 0.5156, "step": 7884 }, { "epoch": 0.96, "learning_rate": 8.848941220841123e-08, "loss": 0.5256, "step": 7885 }, { "epoch": 0.96, "learning_rate": 8.7967348675011e-08, "loss": 0.5185, "step": 7886 }, { "epoch": 0.96, "learning_rate": 8.744682291440431e-08, "loss": 0.5235, "step": 7887 }, { "epoch": 0.96, "learning_rate": 8.69278350073488e-08, "loss": 0.4713, "step": 7888 }, { "epoch": 0.96, "learning_rate": 8.641038503435784e-08, "loss": 0.4569, "step": 7889 }, { "epoch": 0.96, "learning_rate": 8.589447307571164e-08, "loss": 0.548, "step": 7890 }, { "epoch": 0.96, "learning_rate": 8.538009921144841e-08, "loss": 0.4852, "step": 7891 }, { "epoch": 0.96, "learning_rate": 8.486726352137209e-08, "loss": 0.497, "step": 7892 }, { "epoch": 0.96, "learning_rate": 8.435596608504126e-08, "loss": 0.4811, "step": 7893 }, { "epoch": 0.96, "learning_rate": 8.384620698178358e-08, "loss": 0.4741, "step": 7894 }, { "epoch": 0.96, "learning_rate": 8.333798629068024e-08, "loss": 0.5434, "step": 7895 }, { "epoch": 0.96, "learning_rate": 8.283130409058038e-08, "loss": 0.4702, "step": 7896 }, { "epoch": 0.96, "learning_rate": 8.232616046009112e-08, "loss": 0.4523, "step": 7897 }, { "epoch": 0.96, "learning_rate": 8.182255547758088e-08, "loss": 0.5476, "step": 7898 }, { "epoch": 0.96, "learning_rate": 8.132048922117942e-08, "loss": 0.506, "step": 7899 }, { "epoch": 0.96, "learning_rate": 8.081996176877994e-08, "loss": 0.446, "step": 7900 }, { "epoch": 0.96, "learning_rate": 8.032097319803367e-08, "loss": 0.4806, "step": 7901 }, { "epoch": 0.96, "learning_rate": 7.982352358635647e-08, "loss": 0.5081, "step": 7902 }, { "epoch": 0.96, "learning_rate": 7.932761301092218e-08, "loss": 0.5585, "step": 7903 }, { "epoch": 0.96, "learning_rate": 7.883324154866812e-08, "loss": 0.4871, "step": 7904 }, { "epoch": 0.96, "learning_rate": 7.834040927629071e-08, "loss": 0.5375, "step": 7905 }, { "epoch": 0.96, "learning_rate": 7.784911627025105e-08, "loss": 0.4669, "step": 7906 }, { "epoch": 0.96, "learning_rate": 7.735936260676814e-08, "loss": 0.4787, "step": 7907 }, { "epoch": 0.96, "learning_rate": 7.68711483618234e-08, "loss": 0.5156, "step": 7908 }, { "epoch": 0.96, "learning_rate": 7.638447361115964e-08, "loss": 0.5221, "step": 7909 }, { "epoch": 0.96, "learning_rate": 7.589933843028086e-08, "loss": 0.5516, "step": 7910 }, { "epoch": 0.96, "learning_rate": 7.54157428944513e-08, "loss": 0.4686, "step": 7911 }, { "epoch": 0.96, "learning_rate": 7.493368707869652e-08, "loss": 0.4736, "step": 7912 }, { "epoch": 0.96, "learning_rate": 7.445317105780559e-08, "loss": 0.4833, "step": 7913 }, { "epoch": 0.96, "learning_rate": 7.397419490632441e-08, "loss": 0.4867, "step": 7914 }, { "epoch": 0.96, "learning_rate": 7.349675869856354e-08, "loss": 0.5389, "step": 7915 }, { "epoch": 0.96, "learning_rate": 7.302086250859375e-08, "loss": 0.514, "step": 7916 }, { "epoch": 0.96, "learning_rate": 7.254650641024486e-08, "loss": 0.5002, "step": 7917 }, { "epoch": 0.96, "learning_rate": 7.207369047711021e-08, "loss": 0.5388, "step": 7918 }, { "epoch": 0.96, "learning_rate": 7.160241478254559e-08, "loss": 0.501, "step": 7919 }, { "epoch": 0.96, "learning_rate": 7.113267939966251e-08, "loss": 0.4399, "step": 7920 }, { "epoch": 0.96, "learning_rate": 7.066448440133821e-08, "loss": 0.5325, "step": 7921 }, { "epoch": 0.96, "learning_rate": 7.019782986020907e-08, "loss": 0.531, "step": 7922 }, { "epoch": 0.96, "learning_rate": 6.973271584867269e-08, "loss": 0.5259, "step": 7923 }, { "epoch": 0.96, "learning_rate": 6.926914243888916e-08, "loss": 0.5113, "step": 7924 }, { "epoch": 0.96, "learning_rate": 6.880710970277649e-08, "loss": 0.4639, "step": 7925 }, { "epoch": 0.96, "learning_rate": 6.834661771201622e-08, "loss": 0.4709, "step": 7926 }, { "epoch": 0.96, "learning_rate": 6.788766653804901e-08, "loss": 0.4582, "step": 7927 }, { "epoch": 0.96, "learning_rate": 6.7430256252079e-08, "loss": 0.4745, "step": 7928 }, { "epoch": 0.96, "learning_rate": 6.697438692506719e-08, "loss": 0.5205, "step": 7929 }, { "epoch": 0.96, "learning_rate": 6.652005862774147e-08, "loss": 0.5089, "step": 7930 }, { "epoch": 0.96, "learning_rate": 6.606727143058545e-08, "loss": 0.4949, "step": 7931 }, { "epoch": 0.96, "learning_rate": 6.561602540384626e-08, "loss": 0.4503, "step": 7932 }, { "epoch": 0.96, "learning_rate": 6.516632061752904e-08, "loss": 0.4895, "step": 7933 }, { "epoch": 0.96, "learning_rate": 6.471815714140462e-08, "loss": 0.51, "step": 7934 }, { "epoch": 0.96, "learning_rate": 6.427153504499961e-08, "loss": 0.5389, "step": 7935 }, { "epoch": 0.97, "learning_rate": 6.382645439760637e-08, "loss": 0.4954, "step": 7936 }, { "epoch": 0.97, "learning_rate": 6.3382915268273e-08, "loss": 0.5344, "step": 7937 }, { "epoch": 0.97, "learning_rate": 6.294091772581223e-08, "loss": 0.47, "step": 7938 }, { "epoch": 0.97, "learning_rate": 6.250046183879698e-08, "loss": 0.5303, "step": 7939 }, { "epoch": 0.97, "learning_rate": 6.206154767555928e-08, "loss": 0.4899, "step": 7940 }, { "epoch": 0.97, "learning_rate": 6.162417530419462e-08, "loss": 0.4979, "step": 7941 }, { "epoch": 0.97, "learning_rate": 6.118834479255653e-08, "loss": 0.479, "step": 7942 }, { "epoch": 0.97, "learning_rate": 6.075405620826202e-08, "loss": 0.4959, "step": 7943 }, { "epoch": 0.97, "learning_rate": 6.03213096186861e-08, "loss": 0.4927, "step": 7944 }, { "epoch": 0.97, "learning_rate": 5.989010509096726e-08, "loss": 0.5151, "step": 7945 }, { "epoch": 0.97, "learning_rate": 5.946044269200091e-08, "loss": 0.5297, "step": 7946 }, { "epoch": 0.97, "learning_rate": 5.903232248844926e-08, "loss": 0.5036, "step": 7947 }, { "epoch": 0.97, "learning_rate": 5.860574454672918e-08, "loss": 0.4647, "step": 7948 }, { "epoch": 0.97, "learning_rate": 5.8180708933022186e-08, "loss": 0.4944, "step": 7949 }, { "epoch": 0.97, "learning_rate": 5.7757215713267755e-08, "loss": 0.5025, "step": 7950 }, { "epoch": 0.97, "learning_rate": 5.733526495316888e-08, "loss": 0.4853, "step": 7951 }, { "epoch": 0.97, "learning_rate": 5.691485671818764e-08, "loss": 0.4797, "step": 7952 }, { "epoch": 0.97, "learning_rate": 5.649599107354631e-08, "loss": 0.515, "step": 7953 }, { "epoch": 0.97, "learning_rate": 5.607866808422846e-08, "loss": 0.5299, "step": 7954 }, { "epoch": 0.97, "learning_rate": 5.566288781498008e-08, "loss": 0.5786, "step": 7955 }, { "epoch": 0.97, "learning_rate": 5.524865033030402e-08, "loss": 0.5386, "step": 7956 }, { "epoch": 0.97, "learning_rate": 5.483595569446776e-08, "loss": 0.5218, "step": 7957 }, { "epoch": 0.97, "learning_rate": 5.442480397149563e-08, "loss": 0.4633, "step": 7958 }, { "epoch": 0.97, "learning_rate": 5.401519522517662e-08, "loss": 0.5088, "step": 7959 }, { "epoch": 0.97, "learning_rate": 5.3607129519057664e-08, "loss": 0.5321, "step": 7960 }, { "epoch": 0.97, "learning_rate": 5.320060691644591e-08, "loss": 0.5454, "step": 7961 }, { "epoch": 0.97, "learning_rate": 5.279562748041089e-08, "loss": 0.4551, "step": 7962 }, { "epoch": 0.97, "learning_rate": 5.239219127378015e-08, "loss": 0.5049, "step": 7963 }, { "epoch": 0.97, "learning_rate": 5.1990298359146926e-08, "loss": 0.5196, "step": 7964 }, { "epoch": 0.97, "learning_rate": 5.1589948798859144e-08, "loss": 0.4601, "step": 7965 }, { "epoch": 0.97, "learning_rate": 5.1191142655029336e-08, "loss": 0.535, "step": 7966 }, { "epoch": 0.97, "learning_rate": 5.0793879989528005e-08, "loss": 0.5071, "step": 7967 }, { "epoch": 0.97, "learning_rate": 5.0398160863988076e-08, "loss": 0.4907, "step": 7968 }, { "epoch": 0.97, "learning_rate": 5.0003985339801555e-08, "loss": 0.4673, "step": 7969 }, { "epoch": 0.97, "learning_rate": 4.9611353478122845e-08, "loss": 0.502, "step": 7970 }, { "epoch": 0.97, "learning_rate": 4.922026533986324e-08, "loss": 0.5269, "step": 7971 }, { "epoch": 0.97, "learning_rate": 4.8830720985699744e-08, "loss": 0.5136, "step": 7972 }, { "epoch": 0.97, "learning_rate": 4.844272047606513e-08, "loss": 0.4832, "step": 7973 }, { "epoch": 0.97, "learning_rate": 4.8056263871154584e-08, "loss": 0.545, "step": 7974 }, { "epoch": 0.97, "learning_rate": 4.767135123092459e-08, "loss": 0.489, "step": 7975 }, { "epoch": 0.97, "learning_rate": 4.728798261509182e-08, "loss": 0.4854, "step": 7976 }, { "epoch": 0.97, "learning_rate": 4.6906158083130924e-08, "loss": 0.4584, "step": 7977 }, { "epoch": 0.97, "learning_rate": 4.652587769428118e-08, "loss": 0.5087, "step": 7978 }, { "epoch": 0.97, "learning_rate": 4.614714150753874e-08, "loss": 0.4916, "step": 7979 }, { "epoch": 0.97, "learning_rate": 4.576994958166214e-08, "loss": 0.4831, "step": 7980 }, { "epoch": 0.97, "learning_rate": 4.539430197516792e-08, "loss": 0.4576, "step": 7981 }, { "epoch": 0.97, "learning_rate": 4.5020198746337226e-08, "loss": 0.5272, "step": 7982 }, { "epoch": 0.97, "learning_rate": 4.464763995320809e-08, "loss": 0.4786, "step": 7983 }, { "epoch": 0.97, "learning_rate": 4.4276625653579817e-08, "loss": 0.5242, "step": 7984 }, { "epoch": 0.97, "learning_rate": 4.3907155905013044e-08, "loss": 0.5835, "step": 7985 }, { "epoch": 0.97, "learning_rate": 4.3539230764827464e-08, "loss": 0.5453, "step": 7986 }, { "epoch": 0.97, "learning_rate": 4.3172850290102985e-08, "loss": 0.5002, "step": 7987 }, { "epoch": 0.97, "learning_rate": 4.2808014537683015e-08, "loss": 0.5376, "step": 7988 }, { "epoch": 0.97, "learning_rate": 4.244472356416562e-08, "loss": 0.5136, "step": 7989 }, { "epoch": 0.97, "learning_rate": 4.208297742591572e-08, "loss": 0.5668, "step": 7990 }, { "epoch": 0.97, "learning_rate": 4.172277617905285e-08, "loss": 0.5013, "step": 7991 }, { "epoch": 0.97, "learning_rate": 4.136411987946121e-08, "loss": 0.5029, "step": 7992 }, { "epoch": 0.97, "learning_rate": 4.100700858278295e-08, "loss": 0.5171, "step": 7993 }, { "epoch": 0.97, "learning_rate": 4.0651442344420424e-08, "loss": 0.548, "step": 7994 }, { "epoch": 0.97, "learning_rate": 4.029742121953617e-08, "loss": 0.4902, "step": 7995 }, { "epoch": 0.97, "learning_rate": 3.9944945263056256e-08, "loss": 0.4612, "step": 7996 }, { "epoch": 0.97, "learning_rate": 3.9594014529662495e-08, "loss": 0.425, "step": 7997 }, { "epoch": 0.97, "learning_rate": 3.924462907379911e-08, "loss": 0.4928, "step": 7998 }, { "epoch": 0.97, "learning_rate": 3.8896788949671635e-08, "loss": 0.4557, "step": 7999 }, { "epoch": 0.97, "learning_rate": 3.8550494211243575e-08, "loss": 0.4998, "step": 8000 }, { "epoch": 0.97, "learning_rate": 3.8205744912239717e-08, "loss": 0.5414, "step": 8001 }, { "epoch": 0.97, "learning_rate": 3.786254110614618e-08, "loss": 0.4762, "step": 8002 }, { "epoch": 0.97, "learning_rate": 3.752088284620703e-08, "loss": 0.4858, "step": 8003 }, { "epoch": 0.97, "learning_rate": 3.718077018542987e-08, "loss": 0.4313, "step": 8004 }, { "epoch": 0.97, "learning_rate": 3.684220317657805e-08, "loss": 0.5136, "step": 8005 }, { "epoch": 0.97, "learning_rate": 3.650518187217955e-08, "loss": 0.4888, "step": 8006 }, { "epoch": 0.97, "learning_rate": 3.6169706324518106e-08, "loss": 0.5153, "step": 8007 }, { "epoch": 0.97, "learning_rate": 3.583577658564319e-08, "loss": 0.5491, "step": 8008 }, { "epoch": 0.97, "learning_rate": 3.550339270735892e-08, "loss": 0.5272, "step": 8009 }, { "epoch": 0.97, "learning_rate": 3.517255474123293e-08, "loss": 0.4688, "step": 8010 }, { "epoch": 0.97, "learning_rate": 3.484326273859195e-08, "loss": 0.5116, "step": 8011 }, { "epoch": 0.97, "learning_rate": 3.4515516750522894e-08, "loss": 0.5803, "step": 8012 }, { "epoch": 0.97, "learning_rate": 3.4189316827872854e-08, "loss": 0.5257, "step": 8013 }, { "epoch": 0.97, "learning_rate": 3.3864663021250244e-08, "loss": 0.5004, "step": 8014 }, { "epoch": 0.97, "learning_rate": 3.354155538102033e-08, "loss": 0.5257, "step": 8015 }, { "epoch": 0.97, "learning_rate": 3.3219993957313014e-08, "loss": 0.4932, "step": 8016 }, { "epoch": 0.97, "learning_rate": 3.289997880001505e-08, "loss": 0.4579, "step": 8017 }, { "epoch": 0.98, "learning_rate": 3.258150995877341e-08, "loss": 0.5223, "step": 8018 }, { "epoch": 0.98, "learning_rate": 3.226458748299743e-08, "loss": 0.5323, "step": 8019 }, { "epoch": 0.98, "learning_rate": 3.1949211421853366e-08, "loss": 0.4742, "step": 8020 }, { "epoch": 0.98, "learning_rate": 3.163538182427095e-08, "loss": 0.5183, "step": 8021 }, { "epoch": 0.98, "learning_rate": 3.1323098738937905e-08, "loss": 0.4909, "step": 8022 }, { "epoch": 0.98, "learning_rate": 3.101236221430215e-08, "loss": 0.5136, "step": 8023 }, { "epoch": 0.98, "learning_rate": 3.070317229857289e-08, "loss": 0.5356, "step": 8024 }, { "epoch": 0.98, "learning_rate": 3.039552903971621e-08, "loss": 0.5222, "step": 8025 }, { "epoch": 0.98, "learning_rate": 3.008943248546281e-08, "loss": 0.5298, "step": 8026 }, { "epoch": 0.98, "learning_rate": 2.9784882683300265e-08, "loss": 0.4826, "step": 8027 }, { "epoch": 0.98, "learning_rate": 2.9481879680476333e-08, "loss": 0.4649, "step": 8028 }, { "epoch": 0.98, "learning_rate": 2.918042352400119e-08, "loss": 0.525, "step": 8029 }, { "epoch": 0.98, "learning_rate": 2.8880514260641868e-08, "loss": 0.5157, "step": 8030 }, { "epoch": 0.98, "learning_rate": 2.8582151936927815e-08, "loss": 0.4619, "step": 8031 }, { "epoch": 0.98, "learning_rate": 2.828533659914756e-08, "loss": 0.4793, "step": 8032 }, { "epoch": 0.98, "learning_rate": 2.7990068293348715e-08, "loss": 0.506, "step": 8033 }, { "epoch": 0.98, "learning_rate": 2.76963470653413e-08, "loss": 0.5157, "step": 8034 }, { "epoch": 0.98, "learning_rate": 2.7404172960691088e-08, "loss": 0.4811, "step": 8035 }, { "epoch": 0.98, "learning_rate": 2.7113546024729598e-08, "loss": 0.4531, "step": 8036 }, { "epoch": 0.98, "learning_rate": 2.6824466302544094e-08, "loss": 0.4981, "step": 8037 }, { "epoch": 0.98, "learning_rate": 2.6536933838982038e-08, "loss": 0.5083, "step": 8038 }, { "epoch": 0.98, "learning_rate": 2.6250948678653297e-08, "loss": 0.4699, "step": 8039 }, { "epoch": 0.98, "learning_rate": 2.596651086592572e-08, "loss": 0.4552, "step": 8040 }, { "epoch": 0.98, "learning_rate": 2.5683620444927337e-08, "loss": 0.5172, "step": 8041 }, { "epoch": 0.98, "learning_rate": 2.5402277459546377e-08, "loss": 0.4714, "step": 8042 }, { "epoch": 0.98, "learning_rate": 2.5122481953431253e-08, "loss": 0.5111, "step": 8043 }, { "epoch": 0.98, "learning_rate": 2.4844233969989473e-08, "loss": 0.5205, "step": 8044 }, { "epoch": 0.98, "learning_rate": 2.4567533552389832e-08, "loss": 0.48, "step": 8045 }, { "epoch": 0.98, "learning_rate": 2.4292380743559107e-08, "loss": 0.5226, "step": 8046 }, { "epoch": 0.98, "learning_rate": 2.4018775586185372e-08, "loss": 0.5323, "step": 8047 }, { "epoch": 0.98, "learning_rate": 2.3746718122718005e-08, "loss": 0.5618, "step": 8048 }, { "epoch": 0.98, "learning_rate": 2.347620839536102e-08, "loss": 0.514, "step": 8049 }, { "epoch": 0.98, "learning_rate": 2.3207246446085296e-08, "loss": 0.4517, "step": 8050 }, { "epoch": 0.98, "learning_rate": 2.2939832316616338e-08, "loss": 0.5061, "step": 8051 }, { "epoch": 0.98, "learning_rate": 2.2673966048440965e-08, "loss": 0.5217, "step": 8052 }, { "epoch": 0.98, "learning_rate": 2.24096476828084e-08, "loss": 0.4848, "step": 8053 }, { "epoch": 0.98, "learning_rate": 2.2146877260722512e-08, "loss": 0.5189, "step": 8054 }, { "epoch": 0.98, "learning_rate": 2.1885654822952906e-08, "loss": 0.4985, "step": 8055 }, { "epoch": 0.98, "learning_rate": 2.162598041002384e-08, "loss": 0.5567, "step": 8056 }, { "epoch": 0.98, "learning_rate": 2.136785406222308e-08, "loss": 0.5301, "step": 8057 }, { "epoch": 0.98, "learning_rate": 2.111127581959749e-08, "loss": 0.5477, "step": 8058 }, { "epoch": 0.98, "learning_rate": 2.0856245721950775e-08, "loss": 0.5179, "step": 8059 }, { "epoch": 0.98, "learning_rate": 2.060276380885129e-08, "loss": 0.5298, "step": 8060 }, { "epoch": 0.98, "learning_rate": 2.035083011962313e-08, "loss": 0.4978, "step": 8061 }, { "epoch": 0.98, "learning_rate": 2.0100444693351707e-08, "loss": 0.5084, "step": 8062 }, { "epoch": 0.98, "learning_rate": 1.9851607568883714e-08, "loss": 0.5208, "step": 8063 }, { "epoch": 0.98, "learning_rate": 1.9604318784823827e-08, "loss": 0.4267, "step": 8064 }, { "epoch": 0.98, "learning_rate": 1.935857837953581e-08, "loss": 0.5163, "step": 8065 }, { "epoch": 0.98, "learning_rate": 1.911438639114582e-08, "loss": 0.5251, "step": 8066 }, { "epoch": 0.98, "learning_rate": 1.8871742857536903e-08, "loss": 0.5421, "step": 8067 }, { "epoch": 0.98, "learning_rate": 1.8630647816354486e-08, "loss": 0.4354, "step": 8068 }, { "epoch": 0.98, "learning_rate": 1.839110130500199e-08, "loss": 0.4592, "step": 8069 }, { "epoch": 0.98, "learning_rate": 1.815310336064191e-08, "loss": 0.5196, "step": 8070 }, { "epoch": 0.98, "learning_rate": 1.7916654020200263e-08, "loss": 0.5109, "step": 8071 }, { "epoch": 0.98, "learning_rate": 1.768175332035882e-08, "loss": 0.5365, "step": 8072 }, { "epoch": 0.98, "learning_rate": 1.7448401297560647e-08, "loss": 0.5301, "step": 8073 }, { "epoch": 0.98, "learning_rate": 1.7216597988007898e-08, "loss": 0.4818, "step": 8074 }, { "epoch": 0.98, "learning_rate": 1.698634342766403e-08, "loss": 0.4751, "step": 8075 }, { "epoch": 0.98, "learning_rate": 1.675763765225158e-08, "loss": 0.5049, "step": 8076 }, { "epoch": 0.98, "learning_rate": 1.653048069725105e-08, "loss": 0.4645, "step": 8077 }, { "epoch": 0.98, "learning_rate": 1.6304872597904254e-08, "loss": 0.5796, "step": 8078 }, { "epoch": 0.98, "learning_rate": 1.608081338921319e-08, "loss": 0.5172, "step": 8079 }, { "epoch": 0.98, "learning_rate": 1.5858303105937834e-08, "loss": 0.5034, "step": 8080 }, { "epoch": 0.98, "learning_rate": 1.563734178260057e-08, "loss": 0.4966, "step": 8081 }, { "epoch": 0.98, "learning_rate": 1.541792945348064e-08, "loss": 0.4973, "step": 8082 }, { "epoch": 0.98, "learning_rate": 1.5200066152617488e-08, "loss": 0.4869, "step": 8083 }, { "epoch": 0.98, "learning_rate": 1.498375191381296e-08, "loss": 0.4778, "step": 8084 }, { "epoch": 0.98, "learning_rate": 1.476898677062355e-08, "loss": 0.5304, "step": 8085 }, { "epoch": 0.98, "learning_rate": 1.455577075637149e-08, "loss": 0.4697, "step": 8086 }, { "epoch": 0.98, "learning_rate": 1.4344103904132544e-08, "loss": 0.484, "step": 8087 }, { "epoch": 0.98, "learning_rate": 1.4133986246748221e-08, "loss": 0.4787, "step": 8088 }, { "epoch": 0.98, "learning_rate": 1.3925417816813557e-08, "loss": 0.5353, "step": 8089 }, { "epoch": 0.98, "learning_rate": 1.3718398646688225e-08, "loss": 0.4912, "step": 8090 }, { "epoch": 0.98, "learning_rate": 1.3512928768489864e-08, "loss": 0.5285, "step": 8091 }, { "epoch": 0.98, "learning_rate": 1.3309008214094088e-08, "loss": 0.4746, "step": 8092 }, { "epoch": 0.98, "learning_rate": 1.3106637015137813e-08, "loss": 0.5219, "step": 8093 }, { "epoch": 0.98, "learning_rate": 1.2905815203018146e-08, "loss": 0.5088, "step": 8094 }, { "epoch": 0.98, "learning_rate": 1.2706542808890166e-08, "loss": 0.5306, "step": 8095 }, { "epoch": 0.98, "learning_rate": 1.250881986366914e-08, "loss": 0.4651, "step": 8096 }, { "epoch": 0.98, "learning_rate": 1.2312646398031647e-08, "loss": 0.5239, "step": 8097 }, { "epoch": 0.98, "learning_rate": 1.2118022442412225e-08, "loss": 0.4509, "step": 8098 }, { "epoch": 0.98, "learning_rate": 1.1924948027003391e-08, "loss": 0.4942, "step": 8099 }, { "epoch": 0.98, "learning_rate": 1.1733423181761183e-08, "loss": 0.5052, "step": 8100 }, { "epoch": 0.99, "learning_rate": 1.154344793639739e-08, "loss": 0.4921, "step": 8101 }, { "epoch": 0.99, "learning_rate": 1.135502232038732e-08, "loss": 0.5197, "step": 8102 }, { "epoch": 0.99, "learning_rate": 1.116814636296093e-08, "loss": 0.497, "step": 8103 }, { "epoch": 0.99, "learning_rate": 1.09828200931128e-08, "loss": 0.508, "step": 8104 }, { "epoch": 0.99, "learning_rate": 1.0799043539594378e-08, "loss": 0.5709, "step": 8105 }, { "epoch": 0.99, "learning_rate": 1.0616816730916191e-08, "loss": 0.5023, "step": 8106 }, { "epoch": 0.99, "learning_rate": 1.0436139695350067e-08, "loss": 0.5339, "step": 8107 }, { "epoch": 0.99, "learning_rate": 1.0257012460926919e-08, "loss": 0.5086, "step": 8108 }, { "epoch": 0.99, "learning_rate": 1.0079435055435626e-08, "loss": 0.4792, "step": 8109 }, { "epoch": 0.99, "learning_rate": 9.903407506427486e-09, "loss": 0.5046, "step": 8110 }, { "epoch": 0.99, "learning_rate": 9.728929841210654e-09, "loss": 0.4818, "step": 8111 }, { "epoch": 0.99, "learning_rate": 9.556002086854588e-09, "loss": 0.5068, "step": 8112 }, { "epoch": 0.99, "learning_rate": 9.384624270186715e-09, "loss": 0.5067, "step": 8113 }, { "epoch": 0.99, "learning_rate": 9.214796417795768e-09, "loss": 0.5174, "step": 8114 }, { "epoch": 0.99, "learning_rate": 9.04651855602845e-09, "loss": 0.5491, "step": 8115 }, { "epoch": 0.99, "learning_rate": 8.879790710992764e-09, "loss": 0.4989, "step": 8116 }, { "epoch": 0.99, "learning_rate": 8.714612908555797e-09, "loss": 0.4836, "step": 8117 }, { "epoch": 0.99, "learning_rate": 8.550985174340388e-09, "loss": 0.5235, "step": 8118 }, { "epoch": 0.99, "learning_rate": 8.388907533736224e-09, "loss": 0.5764, "step": 8119 }, { "epoch": 0.99, "learning_rate": 8.228380011884307e-09, "loss": 0.5406, "step": 8120 }, { "epoch": 0.99, "learning_rate": 8.069402633692491e-09, "loss": 0.4927, "step": 8121 }, { "epoch": 0.99, "learning_rate": 7.91197542382327e-09, "loss": 0.4442, "step": 8122 }, { "epoch": 0.99, "learning_rate": 7.756098406700441e-09, "loss": 0.4712, "step": 8123 }, { "epoch": 0.99, "learning_rate": 7.60177160650688e-09, "loss": 0.4958, "step": 8124 }, { "epoch": 0.99, "learning_rate": 7.448995047185659e-09, "loss": 0.5027, "step": 8125 }, { "epoch": 0.99, "learning_rate": 7.297768752438927e-09, "loss": 0.5055, "step": 8126 }, { "epoch": 0.99, "learning_rate": 7.148092745726809e-09, "loss": 0.4143, "step": 8127 }, { "epoch": 0.99, "learning_rate": 6.999967050272949e-09, "loss": 0.4775, "step": 8128 }, { "epoch": 0.99, "learning_rate": 6.853391689055633e-09, "loss": 0.4809, "step": 8129 }, { "epoch": 0.99, "learning_rate": 6.7083666848144494e-09, "loss": 0.5371, "step": 8130 }, { "epoch": 0.99, "learning_rate": 6.5648920600513974e-09, "loss": 0.5746, "step": 8131 }, { "epoch": 0.99, "learning_rate": 6.42296783702423e-09, "loss": 0.5364, "step": 8132 }, { "epoch": 0.99, "learning_rate": 6.282594037749778e-09, "loss": 0.5067, "step": 8133 }, { "epoch": 0.99, "learning_rate": 6.1437706840083984e-09, "loss": 0.5022, "step": 8134 }, { "epoch": 0.99, "learning_rate": 6.006497797336197e-09, "loss": 0.4932, "step": 8135 }, { "epoch": 0.99, "learning_rate": 5.870775399029471e-09, "loss": 0.4807, "step": 8136 }, { "epoch": 0.99, "learning_rate": 5.7366035101458215e-09, "loss": 0.5402, "step": 8137 }, { "epoch": 0.99, "learning_rate": 5.603982151499709e-09, "loss": 0.5362, "step": 8138 }, { "epoch": 0.99, "learning_rate": 5.472911343666898e-09, "loss": 0.486, "step": 8139 }, { "epoch": 0.99, "learning_rate": 5.3433911069822316e-09, "loss": 0.4316, "step": 8140 }, { "epoch": 0.99, "learning_rate": 5.215421461538528e-09, "loss": 0.503, "step": 8141 }, { "epoch": 0.99, "learning_rate": 5.089002427191014e-09, "loss": 0.4964, "step": 8142 }, { "epoch": 0.99, "learning_rate": 4.964134023551781e-09, "loss": 0.4902, "step": 8143 }, { "epoch": 0.99, "learning_rate": 4.8408162699919995e-09, "loss": 0.4547, "step": 8144 }, { "epoch": 0.99, "learning_rate": 4.719049185645252e-09, "loss": 0.488, "step": 8145 }, { "epoch": 0.99, "learning_rate": 4.598832789401986e-09, "loss": 0.4977, "step": 8146 }, { "epoch": 0.99, "learning_rate": 4.480167099912836e-09, "loss": 0.5274, "step": 8147 }, { "epoch": 0.99, "learning_rate": 4.36305213558752e-09, "loss": 0.4834, "step": 8148 }, { "epoch": 0.99, "learning_rate": 4.247487914594839e-09, "loss": 0.4875, "step": 8149 }, { "epoch": 0.99, "learning_rate": 4.1334744548660045e-09, "loss": 0.5182, "step": 8150 }, { "epoch": 0.99, "learning_rate": 4.02101177408798e-09, "loss": 0.4923, "step": 8151 }, { "epoch": 0.99, "learning_rate": 3.910099889707919e-09, "loss": 0.4993, "step": 8152 }, { "epoch": 0.99, "learning_rate": 3.80073881893317e-09, "loss": 0.54, "step": 8153 }, { "epoch": 0.99, "learning_rate": 3.6929285787301593e-09, "loss": 0.4698, "step": 8154 }, { "epoch": 0.99, "learning_rate": 3.586669185825509e-09, "loss": 0.4819, "step": 8155 }, { "epoch": 0.99, "learning_rate": 3.481960656703809e-09, "loss": 0.5735, "step": 8156 }, { "epoch": 0.99, "learning_rate": 3.3788030076109534e-09, "loss": 0.4613, "step": 8157 }, { "epoch": 0.99, "learning_rate": 3.277196254548587e-09, "loss": 0.5412, "step": 8158 }, { "epoch": 0.99, "learning_rate": 3.1771404132829864e-09, "loss": 0.4604, "step": 8159 }, { "epoch": 0.99, "learning_rate": 3.07863549933507e-09, "loss": 0.5213, "step": 8160 }, { "epoch": 0.99, "learning_rate": 2.9816815279881673e-09, "loss": 0.562, "step": 8161 }, { "epoch": 0.99, "learning_rate": 2.8862785142824702e-09, "loss": 0.4722, "step": 8162 }, { "epoch": 0.99, "learning_rate": 2.7924264730205818e-09, "loss": 0.4609, "step": 8163 }, { "epoch": 0.99, "learning_rate": 2.7001254187619676e-09, "loss": 0.4821, "step": 8164 }, { "epoch": 0.99, "learning_rate": 2.6093753658273935e-09, "loss": 0.5302, "step": 8165 }, { "epoch": 0.99, "learning_rate": 2.5201763282955983e-09, "loss": 0.4982, "step": 8166 }, { "epoch": 0.99, "learning_rate": 2.4325283200044015e-09, "loss": 0.4841, "step": 8167 }, { "epoch": 0.99, "learning_rate": 2.3464313545518146e-09, "loss": 0.4838, "step": 8168 }, { "epoch": 0.99, "learning_rate": 2.2618854452960415e-09, "loss": 0.5531, "step": 8169 }, { "epoch": 0.99, "learning_rate": 2.1788906053521463e-09, "loss": 0.5097, "step": 8170 }, { "epoch": 0.99, "learning_rate": 2.0974468475976064e-09, "loss": 0.4799, "step": 8171 }, { "epoch": 0.99, "learning_rate": 2.017554184667869e-09, "loss": 0.5205, "step": 8172 }, { "epoch": 0.99, "learning_rate": 1.9392126289563552e-09, "loss": 0.5268, "step": 8173 }, { "epoch": 0.99, "learning_rate": 1.8624221926177854e-09, "loss": 0.4655, "step": 8174 }, { "epoch": 0.99, "learning_rate": 1.7871828875659636e-09, "loss": 0.4944, "step": 8175 }, { "epoch": 0.99, "learning_rate": 1.7134947254726641e-09, "loss": 0.4897, "step": 8176 }, { "epoch": 0.99, "learning_rate": 1.641357717770964e-09, "loss": 0.5335, "step": 8177 }, { "epoch": 0.99, "learning_rate": 1.5707718756519109e-09, "loss": 0.5554, "step": 8178 }, { "epoch": 0.99, "learning_rate": 1.5017372100667448e-09, "loss": 0.5168, "step": 8179 }, { "epoch": 0.99, "learning_rate": 1.4342537317246773e-09, "loss": 0.4834, "step": 8180 }, { "epoch": 0.99, "learning_rate": 1.3683214510962217e-09, "loss": 0.5261, "step": 8181 }, { "epoch": 0.99, "learning_rate": 1.3039403784098625e-09, "loss": 0.5371, "step": 8182 }, { "epoch": 1.0, "learning_rate": 1.2411105236542764e-09, "loss": 0.5346, "step": 8183 }, { "epoch": 1.0, "learning_rate": 1.1798318965761113e-09, "loss": 0.5351, "step": 8184 }, { "epoch": 1.0, "learning_rate": 1.120104506683317e-09, "loss": 0.454, "step": 8185 }, { "epoch": 1.0, "learning_rate": 1.0619283632418153e-09, "loss": 0.4978, "step": 8186 }, { "epoch": 1.0, "learning_rate": 1.0053034752754986e-09, "loss": 0.5057, "step": 8187 }, { "epoch": 1.0, "learning_rate": 9.502298515717822e-10, "loss": 0.5427, "step": 8188 }, { "epoch": 1.0, "learning_rate": 8.967075006738324e-10, "loss": 0.497, "step": 8189 }, { "epoch": 1.0, "learning_rate": 8.44736430885007e-10, "loss": 0.5395, "step": 8190 }, { "epoch": 1.0, "learning_rate": 7.943166502677457e-10, "loss": 0.5232, "step": 8191 }, { "epoch": 1.0, "learning_rate": 7.454481666457902e-10, "loss": 0.4537, "step": 8192 }, { "epoch": 1.0, "learning_rate": 6.981309875986331e-10, "loss": 0.4805, "step": 8193 }, { "epoch": 1.0, "learning_rate": 6.523651204692893e-10, "loss": 0.4994, "step": 8194 }, { "epoch": 1.0, "learning_rate": 6.08150572356525e-10, "loss": 0.3942, "step": 8195 }, { "epoch": 1.0, "learning_rate": 5.654873501215186e-10, "loss": 0.5165, "step": 8196 }, { "epoch": 1.0, "learning_rate": 5.243754603800888e-10, "loss": 0.5172, "step": 8197 }, { "epoch": 1.0, "learning_rate": 4.848149095137978e-10, "loss": 0.4796, "step": 8198 }, { "epoch": 1.0, "learning_rate": 4.4680570365773777e-10, "loss": 0.5361, "step": 8199 }, { "epoch": 1.0, "learning_rate": 4.1034784871052393e-10, "loss": 0.5232, "step": 8200 }, { "epoch": 1.0, "learning_rate": 3.7544135032652196e-10, "loss": 0.4516, "step": 8201 }, { "epoch": 1.0, "learning_rate": 3.420862139225101e-10, "loss": 0.5043, "step": 8202 }, { "epoch": 1.0, "learning_rate": 3.102824446732378e-10, "loss": 0.5142, "step": 8203 }, { "epoch": 1.0, "learning_rate": 2.8003004751253615e-10, "loss": 0.464, "step": 8204 }, { "epoch": 1.0, "learning_rate": 2.51329027133318e-10, "loss": 0.5167, "step": 8205 }, { "epoch": 1.0, "learning_rate": 2.2417938798868777e-10, "loss": 0.4623, "step": 8206 }, { "epoch": 1.0, "learning_rate": 1.985811342908317e-10, "loss": 0.5431, "step": 8207 }, { "epoch": 1.0, "learning_rate": 1.745342700099073e-10, "loss": 0.4895, "step": 8208 }, { "epoch": 1.0, "learning_rate": 1.5203879887959462e-10, "loss": 0.4939, "step": 8209 }, { "epoch": 1.0, "learning_rate": 1.310947243859939e-10, "loss": 0.5221, "step": 8210 }, { "epoch": 1.0, "learning_rate": 1.1170204978205867e-10, "loss": 0.5676, "step": 8211 }, { "epoch": 1.0, "learning_rate": 9.386077807427285e-11, "loss": 0.512, "step": 8212 }, { "epoch": 1.0, "learning_rate": 7.757091203042244e-11, "loss": 0.497, "step": 8213 }, { "epoch": 1.0, "learning_rate": 6.283245417959549e-11, "loss": 0.5184, "step": 8214 }, { "epoch": 1.0, "learning_rate": 4.964540680552077e-11, "loss": 0.5089, "step": 8215 }, { "epoch": 1.0, "learning_rate": 3.800977195655975e-11, "loss": 0.4708, "step": 8216 }, { "epoch": 1.0, "learning_rate": 2.7925551436824848e-11, "loss": 0.4957, "step": 8217 }, { "epoch": 1.0, "learning_rate": 1.9392746811730535e-11, "loss": 0.5123, "step": 8218 }, { "epoch": 1.0, "learning_rate": 1.2411359403552426e-11, "loss": 0.5467, "step": 8219 }, { "epoch": 1.0, "learning_rate": 6.981390295868196e-12, "loss": 0.5023, "step": 8220 }, { "epoch": 1.0, "learning_rate": 3.1028403324473433e-12, "loss": 0.5471, "step": 8221 }, { "epoch": 1.0, "learning_rate": 7.757101128103018e-13, "loss": 0.5277, "step": 8222 }, { "epoch": 1.0, "learning_rate": 0.0, "loss": 0.4914, "step": 8223 }, { "epoch": 1.0, "step": 8223, "total_flos": 1.9950922907516928e+17, "train_loss": 0.23422772625523966, "train_runtime": 192520.3494, "train_samples_per_second": 19.136, "train_steps_per_second": 0.043 } ], "logging_steps": 1.0, "max_steps": 8223, "num_train_epochs": 1, "save_steps": 500, "total_flos": 1.9950922907516928e+17, "trial_name": null, "trial_params": null }