{ "best_metric": null, "best_model_checkpoint": null, "epoch": 0.9999066417374873, "global_step": 2774000, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 1.9996395433879824e-05, "loss": 1.8135, "step": 500 }, { "epoch": 0.0, "learning_rate": 1.9992790867759642e-05, "loss": 1.718, "step": 1000 }, { "epoch": 0.0, "learning_rate": 1.9989186301639468e-05, "loss": 1.6146, "step": 1500 }, { "epoch": 0.0, "learning_rate": 1.998558173551929e-05, "loss": 1.641, "step": 2000 }, { "epoch": 0.0, "learning_rate": 1.998197716939911e-05, "loss": 1.6383, "step": 2500 }, { "epoch": 0.0, "learning_rate": 1.997837260327893e-05, "loss": 1.6062, "step": 3000 }, { "epoch": 0.0, "learning_rate": 1.9974768037158752e-05, "loss": 1.5478, "step": 3500 }, { "epoch": 0.0, "learning_rate": 1.9971163471038574e-05, "loss": 1.5395, "step": 4000 }, { "epoch": 0.0, "learning_rate": 1.9967558904918396e-05, "loss": 1.5065, "step": 4500 }, { "epoch": 0.0, "learning_rate": 1.996395433879822e-05, "loss": 1.5151, "step": 5000 }, { "epoch": 0.0, "learning_rate": 1.9960349772678037e-05, "loss": 1.4828, "step": 5500 }, { "epoch": 0.0, "learning_rate": 1.9956745206557862e-05, "loss": 1.5092, "step": 6000 }, { "epoch": 0.0, "learning_rate": 1.995314064043768e-05, "loss": 1.4928, "step": 6500 }, { "epoch": 0.0, "learning_rate": 1.9949536074317503e-05, "loss": 1.4528, "step": 7000 }, { "epoch": 0.0, "learning_rate": 1.9945931508197325e-05, "loss": 1.4359, "step": 7500 }, { "epoch": 0.0, "learning_rate": 1.9942326942077147e-05, "loss": 1.4415, "step": 8000 }, { "epoch": 0.0, "learning_rate": 1.993872237595697e-05, "loss": 1.4708, "step": 8500 }, { "epoch": 0.0, "learning_rate": 1.993511780983679e-05, "loss": 1.3905, "step": 9000 }, { "epoch": 0.0, "learning_rate": 1.9931513243716613e-05, "loss": 1.4172, "step": 9500 }, { "epoch": 0.0, "learning_rate": 1.9927908677596435e-05, "loss": 1.4055, "step": 10000 }, { "epoch": 0.0, "learning_rate": 1.9924304111476257e-05, "loss": 1.4127, "step": 10500 }, { "epoch": 0.0, "learning_rate": 1.9920699545356076e-05, "loss": 1.364, "step": 11000 }, { "epoch": 0.0, "learning_rate": 1.9917094979235898e-05, "loss": 1.4331, "step": 11500 }, { "epoch": 0.0, "learning_rate": 1.991349041311572e-05, "loss": 1.3773, "step": 12000 }, { "epoch": 0.0, "learning_rate": 1.9909885846995542e-05, "loss": 1.3516, "step": 12500 }, { "epoch": 0.0, "learning_rate": 1.9906281280875364e-05, "loss": 1.4033, "step": 13000 }, { "epoch": 0.0, "learning_rate": 1.9902676714755186e-05, "loss": 1.4004, "step": 13500 }, { "epoch": 0.01, "learning_rate": 1.9899072148635004e-05, "loss": 1.3363, "step": 14000 }, { "epoch": 0.01, "learning_rate": 1.989546758251483e-05, "loss": 1.3508, "step": 14500 }, { "epoch": 0.01, "learning_rate": 1.9891863016394652e-05, "loss": 1.3667, "step": 15000 }, { "epoch": 0.01, "learning_rate": 1.988825845027447e-05, "loss": 1.3719, "step": 15500 }, { "epoch": 0.01, "learning_rate": 1.9884653884154293e-05, "loss": 1.2866, "step": 16000 }, { "epoch": 0.01, "learning_rate": 1.9881049318034115e-05, "loss": 1.3132, "step": 16500 }, { "epoch": 0.01, "learning_rate": 1.9877444751913937e-05, "loss": 1.3323, "step": 17000 }, { "epoch": 0.01, "learning_rate": 1.987384018579376e-05, "loss": 1.3358, "step": 17500 }, { "epoch": 0.01, "learning_rate": 1.987023561967358e-05, "loss": 1.3277, "step": 18000 }, { "epoch": 0.01, "learning_rate": 1.98666310535534e-05, "loss": 1.3301, "step": 18500 }, { "epoch": 0.01, "learning_rate": 1.9863026487433225e-05, "loss": 1.3233, "step": 19000 }, { "epoch": 0.01, "learning_rate": 1.9859421921313043e-05, "loss": 1.2967, "step": 19500 }, { "epoch": 0.01, "learning_rate": 1.9855817355192865e-05, "loss": 1.3245, "step": 20000 }, { "epoch": 0.01, "learning_rate": 1.9852212789072687e-05, "loss": 1.317, "step": 20500 }, { "epoch": 0.01, "learning_rate": 1.984860822295251e-05, "loss": 1.2971, "step": 21000 }, { "epoch": 0.01, "learning_rate": 1.984500365683233e-05, "loss": 1.3023, "step": 21500 }, { "epoch": 0.01, "learning_rate": 1.9841399090712153e-05, "loss": 1.2846, "step": 22000 }, { "epoch": 0.01, "learning_rate": 1.9837794524591972e-05, "loss": 1.2756, "step": 22500 }, { "epoch": 0.01, "learning_rate": 1.9834189958471794e-05, "loss": 1.266, "step": 23000 }, { "epoch": 0.01, "learning_rate": 1.983058539235162e-05, "loss": 1.2877, "step": 23500 }, { "epoch": 0.01, "learning_rate": 1.9826980826231438e-05, "loss": 1.2718, "step": 24000 }, { "epoch": 0.01, "learning_rate": 1.982337626011126e-05, "loss": 1.2655, "step": 24500 }, { "epoch": 0.01, "learning_rate": 1.9819771693991082e-05, "loss": 1.2696, "step": 25000 }, { "epoch": 0.01, "learning_rate": 1.9816167127870904e-05, "loss": 1.2312, "step": 25500 }, { "epoch": 0.01, "learning_rate": 1.9812562561750726e-05, "loss": 1.2799, "step": 26000 }, { "epoch": 0.01, "learning_rate": 1.9808957995630548e-05, "loss": 1.2301, "step": 26500 }, { "epoch": 0.01, "learning_rate": 1.9805353429510367e-05, "loss": 1.2748, "step": 27000 }, { "epoch": 0.01, "learning_rate": 1.980174886339019e-05, "loss": 1.2477, "step": 27500 }, { "epoch": 0.01, "learning_rate": 1.979814429727001e-05, "loss": 1.2575, "step": 28000 }, { "epoch": 0.01, "learning_rate": 1.9794539731149833e-05, "loss": 1.2712, "step": 28500 }, { "epoch": 0.01, "learning_rate": 1.9790935165029655e-05, "loss": 1.2131, "step": 29000 }, { "epoch": 0.01, "learning_rate": 1.9787330598909477e-05, "loss": 1.2303, "step": 29500 }, { "epoch": 0.01, "learning_rate": 1.97837260327893e-05, "loss": 1.2314, "step": 30000 }, { "epoch": 0.01, "learning_rate": 1.978012146666912e-05, "loss": 1.2265, "step": 30500 }, { "epoch": 0.01, "learning_rate": 1.977651690054894e-05, "loss": 1.2312, "step": 31000 }, { "epoch": 0.01, "learning_rate": 1.977291233442876e-05, "loss": 1.2499, "step": 31500 }, { "epoch": 0.01, "learning_rate": 1.9769307768308583e-05, "loss": 1.2331, "step": 32000 }, { "epoch": 0.01, "learning_rate": 1.9765703202188405e-05, "loss": 1.2332, "step": 32500 }, { "epoch": 0.01, "learning_rate": 1.9762098636068227e-05, "loss": 1.2333, "step": 33000 }, { "epoch": 0.01, "learning_rate": 1.975849406994805e-05, "loss": 1.2194, "step": 33500 }, { "epoch": 0.01, "learning_rate": 1.975488950382787e-05, "loss": 1.2403, "step": 34000 }, { "epoch": 0.01, "learning_rate": 1.9751284937707693e-05, "loss": 1.2152, "step": 34500 }, { "epoch": 0.01, "learning_rate": 1.9747680371587515e-05, "loss": 1.2198, "step": 35000 }, { "epoch": 0.01, "learning_rate": 1.9744075805467334e-05, "loss": 1.2082, "step": 35500 }, { "epoch": 0.01, "learning_rate": 1.9740471239347156e-05, "loss": 1.2278, "step": 36000 }, { "epoch": 0.01, "learning_rate": 1.9736866673226978e-05, "loss": 1.2058, "step": 36500 }, { "epoch": 0.01, "learning_rate": 1.97332621071068e-05, "loss": 1.1993, "step": 37000 }, { "epoch": 0.01, "learning_rate": 1.9729657540986622e-05, "loss": 1.2101, "step": 37500 }, { "epoch": 0.01, "learning_rate": 1.9726052974866444e-05, "loss": 1.1926, "step": 38000 }, { "epoch": 0.01, "learning_rate": 1.9722448408746263e-05, "loss": 1.2015, "step": 38500 }, { "epoch": 0.01, "learning_rate": 1.9718843842626088e-05, "loss": 1.1912, "step": 39000 }, { "epoch": 0.01, "learning_rate": 1.971523927650591e-05, "loss": 1.1732, "step": 39500 }, { "epoch": 0.01, "learning_rate": 1.971163471038573e-05, "loss": 1.1782, "step": 40000 }, { "epoch": 0.01, "learning_rate": 1.970803014426555e-05, "loss": 1.1667, "step": 40500 }, { "epoch": 0.01, "learning_rate": 1.9704425578145373e-05, "loss": 1.1556, "step": 41000 }, { "epoch": 0.01, "learning_rate": 1.9700821012025195e-05, "loss": 1.1543, "step": 41500 }, { "epoch": 0.02, "learning_rate": 1.9697216445905017e-05, "loss": 1.1752, "step": 42000 }, { "epoch": 0.02, "learning_rate": 1.969361187978484e-05, "loss": 1.1643, "step": 42500 }, { "epoch": 0.02, "learning_rate": 1.9690007313664658e-05, "loss": 1.1842, "step": 43000 }, { "epoch": 0.02, "learning_rate": 1.9686402747544483e-05, "loss": 1.1373, "step": 43500 }, { "epoch": 0.02, "learning_rate": 1.96827981814243e-05, "loss": 1.1501, "step": 44000 }, { "epoch": 0.02, "learning_rate": 1.9679193615304124e-05, "loss": 1.1627, "step": 44500 }, { "epoch": 0.02, "learning_rate": 1.9675589049183946e-05, "loss": 1.1289, "step": 45000 }, { "epoch": 0.02, "learning_rate": 1.9671984483063768e-05, "loss": 1.1491, "step": 45500 }, { "epoch": 0.02, "learning_rate": 1.966837991694359e-05, "loss": 1.1644, "step": 46000 }, { "epoch": 0.02, "learning_rate": 1.966477535082341e-05, "loss": 1.1477, "step": 46500 }, { "epoch": 0.02, "learning_rate": 1.966117078470323e-05, "loss": 1.1317, "step": 47000 }, { "epoch": 0.02, "learning_rate": 1.9657566218583052e-05, "loss": 1.1458, "step": 47500 }, { "epoch": 0.02, "learning_rate": 1.9653961652462878e-05, "loss": 1.1383, "step": 48000 }, { "epoch": 0.02, "learning_rate": 1.9650357086342696e-05, "loss": 1.1336, "step": 48500 }, { "epoch": 0.02, "learning_rate": 1.964675252022252e-05, "loss": 1.116, "step": 49000 }, { "epoch": 0.02, "learning_rate": 1.964314795410234e-05, "loss": 1.1349, "step": 49500 }, { "epoch": 0.02, "learning_rate": 1.9639543387982162e-05, "loss": 1.1414, "step": 50000 }, { "epoch": 0.02, "learning_rate": 1.9635938821861984e-05, "loss": 1.1723, "step": 50500 }, { "epoch": 0.02, "learning_rate": 1.9632334255741806e-05, "loss": 1.1636, "step": 51000 }, { "epoch": 0.02, "learning_rate": 1.9628729689621625e-05, "loss": 1.1349, "step": 51500 }, { "epoch": 0.02, "learning_rate": 1.9625125123501447e-05, "loss": 1.122, "step": 52000 }, { "epoch": 0.02, "learning_rate": 1.962152055738127e-05, "loss": 1.1722, "step": 52500 }, { "epoch": 0.02, "learning_rate": 1.961791599126109e-05, "loss": 1.145, "step": 53000 }, { "epoch": 0.02, "learning_rate": 1.9614311425140913e-05, "loss": 1.1355, "step": 53500 }, { "epoch": 0.02, "learning_rate": 1.9610706859020735e-05, "loss": 1.1734, "step": 54000 }, { "epoch": 0.02, "learning_rate": 1.9607102292900557e-05, "loss": 1.1593, "step": 54500 }, { "epoch": 0.02, "learning_rate": 1.960349772678038e-05, "loss": 1.1722, "step": 55000 }, { "epoch": 0.02, "learning_rate": 1.95998931606602e-05, "loss": 1.1553, "step": 55500 }, { "epoch": 0.02, "learning_rate": 1.959628859454002e-05, "loss": 1.1524, "step": 56000 }, { "epoch": 0.02, "learning_rate": 1.9592684028419842e-05, "loss": 1.1237, "step": 56500 }, { "epoch": 0.02, "learning_rate": 1.9589079462299664e-05, "loss": 1.1625, "step": 57000 }, { "epoch": 0.02, "learning_rate": 1.9585474896179486e-05, "loss": 1.1507, "step": 57500 }, { "epoch": 0.02, "learning_rate": 1.9581870330059308e-05, "loss": 1.1378, "step": 58000 }, { "epoch": 0.02, "learning_rate": 1.957826576393913e-05, "loss": 1.1174, "step": 58500 }, { "epoch": 0.02, "learning_rate": 1.9574661197818952e-05, "loss": 1.1306, "step": 59000 }, { "epoch": 0.02, "learning_rate": 1.9571056631698774e-05, "loss": 1.123, "step": 59500 }, { "epoch": 0.02, "learning_rate": 1.9567452065578592e-05, "loss": 1.1698, "step": 60000 }, { "epoch": 0.02, "learning_rate": 1.9563847499458414e-05, "loss": 1.0951, "step": 60500 }, { "epoch": 0.02, "learning_rate": 1.956024293333824e-05, "loss": 1.1476, "step": 61000 }, { "epoch": 0.02, "learning_rate": 1.955663836721806e-05, "loss": 1.1055, "step": 61500 }, { "epoch": 0.02, "learning_rate": 1.955303380109788e-05, "loss": 1.1459, "step": 62000 }, { "epoch": 0.02, "learning_rate": 1.9549429234977703e-05, "loss": 1.1306, "step": 62500 }, { "epoch": 0.02, "learning_rate": 1.954582466885752e-05, "loss": 1.119, "step": 63000 }, { "epoch": 0.02, "learning_rate": 1.9542220102737347e-05, "loss": 1.1047, "step": 63500 }, { "epoch": 0.02, "learning_rate": 1.953861553661717e-05, "loss": 1.1048, "step": 64000 }, { "epoch": 0.02, "learning_rate": 1.9535010970496987e-05, "loss": 1.0746, "step": 64500 }, { "epoch": 0.02, "learning_rate": 1.953140640437681e-05, "loss": 1.0935, "step": 65000 }, { "epoch": 0.02, "learning_rate": 1.952780183825663e-05, "loss": 1.1134, "step": 65500 }, { "epoch": 0.02, "learning_rate": 1.9524197272136453e-05, "loss": 1.1032, "step": 66000 }, { "epoch": 0.02, "learning_rate": 1.9520592706016275e-05, "loss": 1.0926, "step": 66500 }, { "epoch": 0.02, "learning_rate": 1.9516988139896097e-05, "loss": 1.0877, "step": 67000 }, { "epoch": 0.02, "learning_rate": 1.9513383573775916e-05, "loss": 1.0836, "step": 67500 }, { "epoch": 0.02, "learning_rate": 1.950977900765574e-05, "loss": 1.0923, "step": 68000 }, { "epoch": 0.02, "learning_rate": 1.950617444153556e-05, "loss": 1.102, "step": 68500 }, { "epoch": 0.02, "learning_rate": 1.9502569875415382e-05, "loss": 1.0937, "step": 69000 }, { "epoch": 0.03, "learning_rate": 1.9498965309295204e-05, "loss": 1.128, "step": 69500 }, { "epoch": 0.03, "learning_rate": 1.9495360743175026e-05, "loss": 1.132, "step": 70000 }, { "epoch": 0.03, "learning_rate": 1.9491756177054848e-05, "loss": 1.0862, "step": 70500 }, { "epoch": 0.03, "learning_rate": 1.948815161093467e-05, "loss": 1.0815, "step": 71000 }, { "epoch": 0.03, "learning_rate": 1.948454704481449e-05, "loss": 1.1308, "step": 71500 }, { "epoch": 0.03, "learning_rate": 1.948094247869431e-05, "loss": 1.0572, "step": 72000 }, { "epoch": 0.03, "learning_rate": 1.9477337912574136e-05, "loss": 1.0837, "step": 72500 }, { "epoch": 0.03, "learning_rate": 1.9473733346453955e-05, "loss": 1.0809, "step": 73000 }, { "epoch": 0.03, "learning_rate": 1.9470128780333777e-05, "loss": 1.0687, "step": 73500 }, { "epoch": 0.03, "learning_rate": 1.94665242142136e-05, "loss": 1.0737, "step": 74000 }, { "epoch": 0.03, "learning_rate": 1.946291964809342e-05, "loss": 1.0723, "step": 74500 }, { "epoch": 0.03, "learning_rate": 1.9459315081973243e-05, "loss": 1.0779, "step": 75000 }, { "epoch": 0.03, "learning_rate": 1.9455710515853065e-05, "loss": 1.0642, "step": 75500 }, { "epoch": 0.03, "learning_rate": 1.9452105949732883e-05, "loss": 1.064, "step": 76000 }, { "epoch": 0.03, "learning_rate": 1.944850138361271e-05, "loss": 1.0576, "step": 76500 }, { "epoch": 0.03, "learning_rate": 1.9444896817492527e-05, "loss": 1.1038, "step": 77000 }, { "epoch": 0.03, "learning_rate": 1.944129225137235e-05, "loss": 1.065, "step": 77500 }, { "epoch": 0.03, "learning_rate": 1.943768768525217e-05, "loss": 1.0751, "step": 78000 }, { "epoch": 0.03, "learning_rate": 1.9434083119131993e-05, "loss": 1.0818, "step": 78500 }, { "epoch": 0.03, "learning_rate": 1.9430478553011815e-05, "loss": 1.0648, "step": 79000 }, { "epoch": 0.03, "learning_rate": 1.9426873986891637e-05, "loss": 1.0585, "step": 79500 }, { "epoch": 0.03, "learning_rate": 1.942326942077146e-05, "loss": 1.0964, "step": 80000 }, { "epoch": 0.03, "learning_rate": 1.9419664854651278e-05, "loss": 1.066, "step": 80500 }, { "epoch": 0.03, "learning_rate": 1.9416060288531103e-05, "loss": 1.0883, "step": 81000 }, { "epoch": 0.03, "learning_rate": 1.9412455722410922e-05, "loss": 1.0411, "step": 81500 }, { "epoch": 0.03, "learning_rate": 1.9408851156290744e-05, "loss": 1.0592, "step": 82000 }, { "epoch": 0.03, "learning_rate": 1.9405246590170566e-05, "loss": 1.098, "step": 82500 }, { "epoch": 0.03, "learning_rate": 1.9401642024050388e-05, "loss": 1.0668, "step": 83000 }, { "epoch": 0.03, "learning_rate": 1.939803745793021e-05, "loss": 1.0695, "step": 83500 }, { "epoch": 0.03, "learning_rate": 1.9394432891810032e-05, "loss": 1.0527, "step": 84000 }, { "epoch": 0.03, "learning_rate": 1.939082832568985e-05, "loss": 1.0561, "step": 84500 }, { "epoch": 0.03, "learning_rate": 1.9387223759569673e-05, "loss": 1.0546, "step": 85000 }, { "epoch": 0.03, "learning_rate": 1.9383619193449498e-05, "loss": 1.0605, "step": 85500 }, { "epoch": 0.03, "learning_rate": 1.9380014627329317e-05, "loss": 1.061, "step": 86000 }, { "epoch": 0.03, "learning_rate": 1.937641006120914e-05, "loss": 1.0368, "step": 86500 }, { "epoch": 0.03, "learning_rate": 1.937280549508896e-05, "loss": 1.0684, "step": 87000 }, { "epoch": 0.03, "learning_rate": 1.936920092896878e-05, "loss": 1.0929, "step": 87500 }, { "epoch": 0.03, "learning_rate": 1.9365596362848605e-05, "loss": 1.0282, "step": 88000 }, { "epoch": 0.03, "learning_rate": 1.9361991796728427e-05, "loss": 1.0465, "step": 88500 }, { "epoch": 0.03, "learning_rate": 1.9358387230608246e-05, "loss": 1.0669, "step": 89000 }, { "epoch": 0.03, "learning_rate": 1.9354782664488068e-05, "loss": 1.0563, "step": 89500 }, { "epoch": 0.03, "learning_rate": 1.935117809836789e-05, "loss": 1.0307, "step": 90000 }, { "epoch": 0.03, "learning_rate": 1.934757353224771e-05, "loss": 1.0275, "step": 90500 }, { "epoch": 0.03, "learning_rate": 1.9343968966127534e-05, "loss": 1.0532, "step": 91000 }, { "epoch": 0.03, "learning_rate": 1.9340364400007356e-05, "loss": 1.0682, "step": 91500 }, { "epoch": 0.03, "learning_rate": 1.9336759833887178e-05, "loss": 1.0665, "step": 92000 }, { "epoch": 0.03, "learning_rate": 1.9333155267767e-05, "loss": 1.0658, "step": 92500 }, { "epoch": 0.03, "learning_rate": 1.9329550701646818e-05, "loss": 1.0283, "step": 93000 }, { "epoch": 0.03, "learning_rate": 1.932594613552664e-05, "loss": 1.0346, "step": 93500 }, { "epoch": 0.03, "learning_rate": 1.9322341569406462e-05, "loss": 1.0563, "step": 94000 }, { "epoch": 0.03, "learning_rate": 1.9318737003286284e-05, "loss": 1.0334, "step": 94500 }, { "epoch": 0.03, "learning_rate": 1.9315132437166106e-05, "loss": 1.0557, "step": 95000 }, { "epoch": 0.03, "learning_rate": 1.931152787104593e-05, "loss": 1.0575, "step": 95500 }, { "epoch": 0.03, "learning_rate": 1.9307923304925747e-05, "loss": 1.0202, "step": 96000 }, { "epoch": 0.03, "learning_rate": 1.9304318738805572e-05, "loss": 1.0623, "step": 96500 }, { "epoch": 0.03, "learning_rate": 1.9300714172685394e-05, "loss": 1.0246, "step": 97000 }, { "epoch": 0.04, "learning_rate": 1.9297109606565213e-05, "loss": 1.0585, "step": 97500 }, { "epoch": 0.04, "learning_rate": 1.9293505040445035e-05, "loss": 1.0012, "step": 98000 }, { "epoch": 0.04, "learning_rate": 1.9289900474324857e-05, "loss": 1.065, "step": 98500 }, { "epoch": 0.04, "learning_rate": 1.928629590820468e-05, "loss": 1.0573, "step": 99000 }, { "epoch": 0.04, "learning_rate": 1.92826913420845e-05, "loss": 1.0359, "step": 99500 }, { "epoch": 0.04, "learning_rate": 1.9279086775964323e-05, "loss": 1.0547, "step": 100000 }, { "epoch": 0.04, "learning_rate": 1.927548220984414e-05, "loss": 1.0503, "step": 100500 }, { "epoch": 0.04, "learning_rate": 1.9271877643723967e-05, "loss": 1.0569, "step": 101000 }, { "epoch": 0.04, "learning_rate": 1.9268273077603786e-05, "loss": 1.0813, "step": 101500 }, { "epoch": 0.04, "learning_rate": 1.9264668511483608e-05, "loss": 1.0539, "step": 102000 }, { "epoch": 0.04, "learning_rate": 1.926106394536343e-05, "loss": 1.0207, "step": 102500 }, { "epoch": 0.04, "learning_rate": 1.9257459379243252e-05, "loss": 1.0356, "step": 103000 }, { "epoch": 0.04, "learning_rate": 1.9253854813123074e-05, "loss": 1.0276, "step": 103500 }, { "epoch": 0.04, "learning_rate": 1.9250250247002896e-05, "loss": 1.0149, "step": 104000 }, { "epoch": 0.04, "learning_rate": 1.9246645680882718e-05, "loss": 1.0465, "step": 104500 }, { "epoch": 0.04, "learning_rate": 1.9243041114762536e-05, "loss": 1.0613, "step": 105000 }, { "epoch": 0.04, "learning_rate": 1.9239436548642362e-05, "loss": 1.03, "step": 105500 }, { "epoch": 0.04, "learning_rate": 1.923583198252218e-05, "loss": 1.0615, "step": 106000 }, { "epoch": 0.04, "learning_rate": 1.9232227416402002e-05, "loss": 1.0117, "step": 106500 }, { "epoch": 0.04, "learning_rate": 1.9228622850281824e-05, "loss": 1.0217, "step": 107000 }, { "epoch": 0.04, "learning_rate": 1.9225018284161646e-05, "loss": 1.0425, "step": 107500 }, { "epoch": 0.04, "learning_rate": 1.922141371804147e-05, "loss": 1.0466, "step": 108000 }, { "epoch": 0.04, "learning_rate": 1.921780915192129e-05, "loss": 1.0264, "step": 108500 }, { "epoch": 0.04, "learning_rate": 1.921420458580111e-05, "loss": 1.0514, "step": 109000 }, { "epoch": 0.04, "learning_rate": 1.921060001968093e-05, "loss": 0.9921, "step": 109500 }, { "epoch": 0.04, "learning_rate": 1.9206995453560757e-05, "loss": 1.0158, "step": 110000 }, { "epoch": 0.04, "learning_rate": 1.9203390887440575e-05, "loss": 1.06, "step": 110500 }, { "epoch": 0.04, "learning_rate": 1.9199786321320397e-05, "loss": 1.0252, "step": 111000 }, { "epoch": 0.04, "learning_rate": 1.919618175520022e-05, "loss": 1.0177, "step": 111500 }, { "epoch": 0.04, "learning_rate": 1.919257718908004e-05, "loss": 0.998, "step": 112000 }, { "epoch": 0.04, "learning_rate": 1.9188972622959863e-05, "loss": 1.0424, "step": 112500 }, { "epoch": 0.04, "learning_rate": 1.9185368056839685e-05, "loss": 1.0104, "step": 113000 }, { "epoch": 0.04, "learning_rate": 1.9181763490719504e-05, "loss": 1.0145, "step": 113500 }, { "epoch": 0.04, "learning_rate": 1.9178158924599326e-05, "loss": 1.0339, "step": 114000 }, { "epoch": 0.04, "learning_rate": 1.9174554358479148e-05, "loss": 1.0311, "step": 114500 }, { "epoch": 0.04, "learning_rate": 1.917094979235897e-05, "loss": 1.0165, "step": 115000 }, { "epoch": 0.04, "learning_rate": 1.9167345226238792e-05, "loss": 1.0107, "step": 115500 }, { "epoch": 0.04, "learning_rate": 1.9163740660118614e-05, "loss": 1.0259, "step": 116000 }, { "epoch": 0.04, "learning_rate": 1.9160136093998436e-05, "loss": 0.9838, "step": 116500 }, { "epoch": 0.04, "learning_rate": 1.9156531527878258e-05, "loss": 1.0258, "step": 117000 }, { "epoch": 0.04, "learning_rate": 1.9152926961758077e-05, "loss": 1.0006, "step": 117500 }, { "epoch": 0.04, "learning_rate": 1.91493223956379e-05, "loss": 1.0121, "step": 118000 }, { "epoch": 0.04, "learning_rate": 1.914571782951772e-05, "loss": 0.9998, "step": 118500 }, { "epoch": 0.04, "learning_rate": 1.9142113263397543e-05, "loss": 1.0441, "step": 119000 }, { "epoch": 0.04, "learning_rate": 1.9138508697277365e-05, "loss": 1.0004, "step": 119500 }, { "epoch": 0.04, "learning_rate": 1.9134904131157187e-05, "loss": 0.9787, "step": 120000 }, { "epoch": 0.04, "learning_rate": 1.913129956503701e-05, "loss": 1.0043, "step": 120500 }, { "epoch": 0.04, "learning_rate": 1.912769499891683e-05, "loss": 1.0046, "step": 121000 }, { "epoch": 0.04, "learning_rate": 1.9124090432796653e-05, "loss": 1.0285, "step": 121500 }, { "epoch": 0.04, "learning_rate": 1.912048586667647e-05, "loss": 1.0232, "step": 122000 }, { "epoch": 0.04, "learning_rate": 1.9116881300556293e-05, "loss": 1.0173, "step": 122500 }, { "epoch": 0.04, "learning_rate": 1.9113276734436115e-05, "loss": 1.0243, "step": 123000 }, { "epoch": 0.04, "learning_rate": 1.9109672168315937e-05, "loss": 1.0227, "step": 123500 }, { "epoch": 0.04, "learning_rate": 1.910606760219576e-05, "loss": 1.0216, "step": 124000 }, { "epoch": 0.04, "learning_rate": 1.910246303607558e-05, "loss": 0.9835, "step": 124500 }, { "epoch": 0.05, "learning_rate": 1.90988584699554e-05, "loss": 1.0156, "step": 125000 }, { "epoch": 0.05, "learning_rate": 1.9095253903835225e-05, "loss": 1.0029, "step": 125500 }, { "epoch": 0.05, "learning_rate": 1.9091649337715047e-05, "loss": 1.0138, "step": 126000 }, { "epoch": 0.05, "learning_rate": 1.9088044771594866e-05, "loss": 1.0101, "step": 126500 }, { "epoch": 0.05, "learning_rate": 1.9084440205474688e-05, "loss": 0.9825, "step": 127000 }, { "epoch": 0.05, "learning_rate": 1.908083563935451e-05, "loss": 1.0028, "step": 127500 }, { "epoch": 0.05, "learning_rate": 1.9077231073234332e-05, "loss": 1.013, "step": 128000 }, { "epoch": 0.05, "learning_rate": 1.9073626507114154e-05, "loss": 1.018, "step": 128500 }, { "epoch": 0.05, "learning_rate": 1.9070021940993976e-05, "loss": 1.0087, "step": 129000 }, { "epoch": 0.05, "learning_rate": 1.9066417374873795e-05, "loss": 1.0106, "step": 129500 }, { "epoch": 0.05, "learning_rate": 1.906281280875362e-05, "loss": 1.0033, "step": 130000 }, { "epoch": 0.05, "learning_rate": 1.905920824263344e-05, "loss": 1.024, "step": 130500 }, { "epoch": 0.05, "learning_rate": 1.905560367651326e-05, "loss": 1.0284, "step": 131000 }, { "epoch": 0.05, "learning_rate": 1.9051999110393083e-05, "loss": 0.9808, "step": 131500 }, { "epoch": 0.05, "learning_rate": 1.9048394544272905e-05, "loss": 0.9862, "step": 132000 }, { "epoch": 0.05, "learning_rate": 1.9044789978152727e-05, "loss": 0.9713, "step": 132500 }, { "epoch": 0.05, "learning_rate": 1.904118541203255e-05, "loss": 1.0227, "step": 133000 }, { "epoch": 0.05, "learning_rate": 1.9037580845912367e-05, "loss": 1.0463, "step": 133500 }, { "epoch": 0.05, "learning_rate": 1.903397627979219e-05, "loss": 0.9701, "step": 134000 }, { "epoch": 0.05, "learning_rate": 1.9030371713672015e-05, "loss": 0.9966, "step": 134500 }, { "epoch": 0.05, "learning_rate": 1.9026767147551834e-05, "loss": 0.984, "step": 135000 }, { "epoch": 0.05, "learning_rate": 1.9023162581431656e-05, "loss": 0.9967, "step": 135500 }, { "epoch": 0.05, "learning_rate": 1.9019558015311478e-05, "loss": 0.9711, "step": 136000 }, { "epoch": 0.05, "learning_rate": 1.90159534491913e-05, "loss": 1.0305, "step": 136500 }, { "epoch": 0.05, "learning_rate": 1.901234888307112e-05, "loss": 0.9887, "step": 137000 }, { "epoch": 0.05, "learning_rate": 1.9008744316950944e-05, "loss": 1.0085, "step": 137500 }, { "epoch": 0.05, "learning_rate": 1.9005139750830762e-05, "loss": 0.9643, "step": 138000 }, { "epoch": 0.05, "learning_rate": 1.9001535184710584e-05, "loss": 0.9854, "step": 138500 }, { "epoch": 0.05, "learning_rate": 1.8997930618590406e-05, "loss": 0.9911, "step": 139000 }, { "epoch": 0.05, "learning_rate": 1.8994326052470228e-05, "loss": 1.0109, "step": 139500 }, { "epoch": 0.05, "learning_rate": 1.899072148635005e-05, "loss": 1.0044, "step": 140000 }, { "epoch": 0.05, "learning_rate": 1.8987116920229872e-05, "loss": 0.9709, "step": 140500 }, { "epoch": 0.05, "learning_rate": 1.8983512354109694e-05, "loss": 0.9578, "step": 141000 }, { "epoch": 0.05, "learning_rate": 1.8979907787989516e-05, "loss": 0.958, "step": 141500 }, { "epoch": 0.05, "learning_rate": 1.8976303221869335e-05, "loss": 0.9791, "step": 142000 }, { "epoch": 0.05, "learning_rate": 1.8972698655749157e-05, "loss": 0.9748, "step": 142500 }, { "epoch": 0.05, "learning_rate": 1.8969094089628982e-05, "loss": 0.9772, "step": 143000 }, { "epoch": 0.05, "learning_rate": 1.89654895235088e-05, "loss": 0.9828, "step": 143500 }, { "epoch": 0.05, "learning_rate": 1.8961884957388623e-05, "loss": 0.9936, "step": 144000 }, { "epoch": 0.05, "learning_rate": 1.8958280391268445e-05, "loss": 0.9877, "step": 144500 }, { "epoch": 0.05, "learning_rate": 1.8954675825148267e-05, "loss": 1.0036, "step": 145000 }, { "epoch": 0.05, "learning_rate": 1.895107125902809e-05, "loss": 0.98, "step": 145500 }, { "epoch": 0.05, "learning_rate": 1.894746669290791e-05, "loss": 1.0138, "step": 146000 }, { "epoch": 0.05, "learning_rate": 1.894386212678773e-05, "loss": 0.9828, "step": 146500 }, { "epoch": 0.05, "learning_rate": 1.8940257560667552e-05, "loss": 0.9645, "step": 147000 }, { "epoch": 0.05, "learning_rate": 1.8936652994547374e-05, "loss": 0.9901, "step": 147500 }, { "epoch": 0.05, "learning_rate": 1.8933048428427196e-05, "loss": 1.0043, "step": 148000 }, { "epoch": 0.05, "learning_rate": 1.8929443862307018e-05, "loss": 0.9831, "step": 148500 }, { "epoch": 0.05, "learning_rate": 1.892583929618684e-05, "loss": 0.9624, "step": 149000 }, { "epoch": 0.05, "learning_rate": 1.892223473006666e-05, "loss": 1.0206, "step": 149500 }, { "epoch": 0.05, "learning_rate": 1.8918630163946484e-05, "loss": 0.9742, "step": 150000 }, { "epoch": 0.05, "learning_rate": 1.8915025597826306e-05, "loss": 0.9783, "step": 150500 }, { "epoch": 0.05, "learning_rate": 1.8911421031706124e-05, "loss": 0.9921, "step": 151000 }, { "epoch": 0.05, "learning_rate": 1.8907816465585946e-05, "loss": 1.0055, "step": 151500 }, { "epoch": 0.05, "learning_rate": 1.890421189946577e-05, "loss": 0.9702, "step": 152000 }, { "epoch": 0.05, "learning_rate": 1.890060733334559e-05, "loss": 0.9944, "step": 152500 }, { "epoch": 0.06, "learning_rate": 1.8897002767225412e-05, "loss": 0.9677, "step": 153000 }, { "epoch": 0.06, "learning_rate": 1.8893398201105234e-05, "loss": 0.9684, "step": 153500 }, { "epoch": 0.06, "learning_rate": 1.8889793634985053e-05, "loss": 0.9762, "step": 154000 }, { "epoch": 0.06, "learning_rate": 1.888618906886488e-05, "loss": 0.9636, "step": 154500 }, { "epoch": 0.06, "learning_rate": 1.8882584502744697e-05, "loss": 0.9991, "step": 155000 }, { "epoch": 0.06, "learning_rate": 1.887897993662452e-05, "loss": 0.9634, "step": 155500 }, { "epoch": 0.06, "learning_rate": 1.887537537050434e-05, "loss": 0.9886, "step": 156000 }, { "epoch": 0.06, "learning_rate": 1.8871770804384163e-05, "loss": 1.001, "step": 156500 }, { "epoch": 0.06, "learning_rate": 1.8868166238263985e-05, "loss": 0.9641, "step": 157000 }, { "epoch": 0.06, "learning_rate": 1.8864561672143807e-05, "loss": 0.9689, "step": 157500 }, { "epoch": 0.06, "learning_rate": 1.8860957106023626e-05, "loss": 1.0067, "step": 158000 }, { "epoch": 0.06, "learning_rate": 1.885735253990345e-05, "loss": 1.009, "step": 158500 }, { "epoch": 0.06, "learning_rate": 1.8853747973783273e-05, "loss": 0.9777, "step": 159000 }, { "epoch": 0.06, "learning_rate": 1.8850143407663092e-05, "loss": 0.9592, "step": 159500 }, { "epoch": 0.06, "learning_rate": 1.8846538841542914e-05, "loss": 0.955, "step": 160000 }, { "epoch": 0.06, "learning_rate": 1.8842934275422736e-05, "loss": 0.9669, "step": 160500 }, { "epoch": 0.06, "learning_rate": 1.8839329709302558e-05, "loss": 0.9634, "step": 161000 }, { "epoch": 0.06, "learning_rate": 1.883572514318238e-05, "loss": 0.9575, "step": 161500 }, { "epoch": 0.06, "learning_rate": 1.8832120577062202e-05, "loss": 0.9495, "step": 162000 }, { "epoch": 0.06, "learning_rate": 1.882851601094202e-05, "loss": 0.9674, "step": 162500 }, { "epoch": 0.06, "learning_rate": 1.8824911444821846e-05, "loss": 0.9558, "step": 163000 }, { "epoch": 0.06, "learning_rate": 1.8821306878701665e-05, "loss": 0.9992, "step": 163500 }, { "epoch": 0.06, "learning_rate": 1.8817702312581487e-05, "loss": 0.9682, "step": 164000 }, { "epoch": 0.06, "learning_rate": 1.881409774646131e-05, "loss": 0.9721, "step": 164500 }, { "epoch": 0.06, "learning_rate": 1.881049318034113e-05, "loss": 0.9437, "step": 165000 }, { "epoch": 0.06, "learning_rate": 1.8806888614220953e-05, "loss": 0.9683, "step": 165500 }, { "epoch": 0.06, "learning_rate": 1.8803284048100775e-05, "loss": 0.9332, "step": 166000 }, { "epoch": 0.06, "learning_rate": 1.8799679481980597e-05, "loss": 0.9633, "step": 166500 }, { "epoch": 0.06, "learning_rate": 1.8796074915860415e-05, "loss": 0.9429, "step": 167000 }, { "epoch": 0.06, "learning_rate": 1.879247034974024e-05, "loss": 0.9673, "step": 167500 }, { "epoch": 0.06, "learning_rate": 1.878886578362006e-05, "loss": 0.9551, "step": 168000 }, { "epoch": 0.06, "learning_rate": 1.878526121749988e-05, "loss": 0.9523, "step": 168500 }, { "epoch": 0.06, "learning_rate": 1.8781656651379703e-05, "loss": 0.9663, "step": 169000 }, { "epoch": 0.06, "learning_rate": 1.8778052085259525e-05, "loss": 0.9409, "step": 169500 }, { "epoch": 0.06, "learning_rate": 1.8774447519139347e-05, "loss": 0.9777, "step": 170000 }, { "epoch": 0.06, "learning_rate": 1.877084295301917e-05, "loss": 0.9631, "step": 170500 }, { "epoch": 0.06, "learning_rate": 1.8767238386898988e-05, "loss": 0.9642, "step": 171000 }, { "epoch": 0.06, "learning_rate": 1.876363382077881e-05, "loss": 0.9391, "step": 171500 }, { "epoch": 0.06, "learning_rate": 1.8760029254658635e-05, "loss": 0.9326, "step": 172000 }, { "epoch": 0.06, "learning_rate": 1.8756424688538454e-05, "loss": 0.9774, "step": 172500 }, { "epoch": 0.06, "learning_rate": 1.8752820122418276e-05, "loss": 0.9893, "step": 173000 }, { "epoch": 0.06, "learning_rate": 1.8749215556298098e-05, "loss": 0.9462, "step": 173500 }, { "epoch": 0.06, "learning_rate": 1.874561099017792e-05, "loss": 0.9635, "step": 174000 }, { "epoch": 0.06, "learning_rate": 1.8742006424057742e-05, "loss": 0.9404, "step": 174500 }, { "epoch": 0.06, "learning_rate": 1.8738401857937564e-05, "loss": 0.9693, "step": 175000 }, { "epoch": 0.06, "learning_rate": 1.8734797291817383e-05, "loss": 0.9411, "step": 175500 }, { "epoch": 0.06, "learning_rate": 1.8731192725697205e-05, "loss": 0.917, "step": 176000 }, { "epoch": 0.06, "learning_rate": 1.8727588159577027e-05, "loss": 0.9582, "step": 176500 }, { "epoch": 0.06, "learning_rate": 1.872398359345685e-05, "loss": 0.9441, "step": 177000 }, { "epoch": 0.06, "learning_rate": 1.872037902733667e-05, "loss": 0.9767, "step": 177500 }, { "epoch": 0.06, "learning_rate": 1.8716774461216493e-05, "loss": 0.9528, "step": 178000 }, { "epoch": 0.06, "learning_rate": 1.8713169895096315e-05, "loss": 0.9488, "step": 178500 }, { "epoch": 0.06, "learning_rate": 1.8709565328976137e-05, "loss": 0.938, "step": 179000 }, { "epoch": 0.06, "learning_rate": 1.8705960762855955e-05, "loss": 0.9567, "step": 179500 }, { "epoch": 0.06, "learning_rate": 1.8702356196735777e-05, "loss": 0.9278, "step": 180000 }, { "epoch": 0.07, "learning_rate": 1.86987516306156e-05, "loss": 0.9444, "step": 180500 }, { "epoch": 0.07, "learning_rate": 1.869514706449542e-05, "loss": 0.955, "step": 181000 }, { "epoch": 0.07, "learning_rate": 1.8691542498375244e-05, "loss": 0.9333, "step": 181500 }, { "epoch": 0.07, "learning_rate": 1.8687937932255066e-05, "loss": 0.9397, "step": 182000 }, { "epoch": 0.07, "learning_rate": 1.8684333366134884e-05, "loss": 0.9512, "step": 182500 }, { "epoch": 0.07, "learning_rate": 1.868072880001471e-05, "loss": 0.9417, "step": 183000 }, { "epoch": 0.07, "learning_rate": 1.867712423389453e-05, "loss": 0.9384, "step": 183500 }, { "epoch": 0.07, "learning_rate": 1.867351966777435e-05, "loss": 0.9646, "step": 184000 }, { "epoch": 0.07, "learning_rate": 1.8669915101654172e-05, "loss": 0.9639, "step": 184500 }, { "epoch": 0.07, "learning_rate": 1.8666310535533994e-05, "loss": 0.9521, "step": 185000 }, { "epoch": 0.07, "learning_rate": 1.8662705969413816e-05, "loss": 0.966, "step": 185500 }, { "epoch": 0.07, "learning_rate": 1.8659101403293638e-05, "loss": 0.9439, "step": 186000 }, { "epoch": 0.07, "learning_rate": 1.865549683717346e-05, "loss": 0.9664, "step": 186500 }, { "epoch": 0.07, "learning_rate": 1.865189227105328e-05, "loss": 0.9762, "step": 187000 }, { "epoch": 0.07, "learning_rate": 1.8648287704933104e-05, "loss": 0.9652, "step": 187500 }, { "epoch": 0.07, "learning_rate": 1.8644683138812923e-05, "loss": 0.9611, "step": 188000 }, { "epoch": 0.07, "learning_rate": 1.8641078572692745e-05, "loss": 0.9403, "step": 188500 }, { "epoch": 0.07, "learning_rate": 1.8637474006572567e-05, "loss": 0.9762, "step": 189000 }, { "epoch": 0.07, "learning_rate": 1.863386944045239e-05, "loss": 0.9714, "step": 189500 }, { "epoch": 0.07, "learning_rate": 1.863026487433221e-05, "loss": 0.954, "step": 190000 }, { "epoch": 0.07, "learning_rate": 1.8626660308212033e-05, "loss": 0.9412, "step": 190500 }, { "epoch": 0.07, "learning_rate": 1.8623055742091855e-05, "loss": 0.9247, "step": 191000 }, { "epoch": 0.07, "learning_rate": 1.8619451175971674e-05, "loss": 0.9471, "step": 191500 }, { "epoch": 0.07, "learning_rate": 1.86158466098515e-05, "loss": 0.9749, "step": 192000 }, { "epoch": 0.07, "learning_rate": 1.8612242043731318e-05, "loss": 0.934, "step": 192500 }, { "epoch": 0.07, "learning_rate": 1.860863747761114e-05, "loss": 0.9476, "step": 193000 }, { "epoch": 0.07, "learning_rate": 1.8605032911490962e-05, "loss": 0.9282, "step": 193500 }, { "epoch": 0.07, "learning_rate": 1.8601428345370784e-05, "loss": 0.9277, "step": 194000 }, { "epoch": 0.07, "learning_rate": 1.8597823779250606e-05, "loss": 0.9223, "step": 194500 }, { "epoch": 0.07, "learning_rate": 1.8594219213130428e-05, "loss": 0.9287, "step": 195000 }, { "epoch": 0.07, "learning_rate": 1.8590614647010246e-05, "loss": 0.9576, "step": 195500 }, { "epoch": 0.07, "learning_rate": 1.858701008089007e-05, "loss": 0.9424, "step": 196000 }, { "epoch": 0.07, "learning_rate": 1.8583405514769894e-05, "loss": 0.9296, "step": 196500 }, { "epoch": 0.07, "learning_rate": 1.8579800948649712e-05, "loss": 0.9382, "step": 197000 }, { "epoch": 0.07, "learning_rate": 1.8576196382529534e-05, "loss": 0.9415, "step": 197500 }, { "epoch": 0.07, "learning_rate": 1.8572591816409356e-05, "loss": 0.9426, "step": 198000 }, { "epoch": 0.07, "learning_rate": 1.856898725028918e-05, "loss": 0.9226, "step": 198500 }, { "epoch": 0.07, "learning_rate": 1.8565382684169e-05, "loss": 0.9147, "step": 199000 }, { "epoch": 0.07, "learning_rate": 1.8561778118048822e-05, "loss": 0.9324, "step": 199500 }, { "epoch": 0.07, "learning_rate": 1.855817355192864e-05, "loss": 0.9193, "step": 200000 }, { "epoch": 0.07, "learning_rate": 1.8554568985808463e-05, "loss": 0.9432, "step": 200500 }, { "epoch": 0.07, "learning_rate": 1.8550964419688285e-05, "loss": 0.9756, "step": 201000 }, { "epoch": 0.07, "learning_rate": 1.8547359853568107e-05, "loss": 0.9378, "step": 201500 }, { "epoch": 0.07, "learning_rate": 1.854375528744793e-05, "loss": 0.9559, "step": 202000 }, { "epoch": 0.07, "learning_rate": 1.854015072132775e-05, "loss": 0.9615, "step": 202500 }, { "epoch": 0.07, "learning_rate": 1.8536546155207573e-05, "loss": 0.9753, "step": 203000 }, { "epoch": 0.07, "learning_rate": 1.8532941589087395e-05, "loss": 0.9324, "step": 203500 }, { "epoch": 0.07, "learning_rate": 1.8529337022967214e-05, "loss": 0.9351, "step": 204000 }, { "epoch": 0.07, "learning_rate": 1.8525732456847036e-05, "loss": 0.9456, "step": 204500 }, { "epoch": 0.07, "learning_rate": 1.8522127890726858e-05, "loss": 0.9216, "step": 205000 }, { "epoch": 0.07, "learning_rate": 1.851852332460668e-05, "loss": 0.9627, "step": 205500 }, { "epoch": 0.07, "learning_rate": 1.8514918758486502e-05, "loss": 0.9205, "step": 206000 }, { "epoch": 0.07, "learning_rate": 1.8511314192366324e-05, "loss": 0.9382, "step": 206500 }, { "epoch": 0.07, "learning_rate": 1.8507709626246143e-05, "loss": 0.9275, "step": 207000 }, { "epoch": 0.07, "learning_rate": 1.8504105060125968e-05, "loss": 0.9185, "step": 207500 }, { "epoch": 0.07, "learning_rate": 1.850050049400579e-05, "loss": 0.9503, "step": 208000 }, { "epoch": 0.08, "learning_rate": 1.849689592788561e-05, "loss": 0.9186, "step": 208500 }, { "epoch": 0.08, "learning_rate": 1.849329136176543e-05, "loss": 0.9649, "step": 209000 }, { "epoch": 0.08, "learning_rate": 1.8489686795645253e-05, "loss": 0.9012, "step": 209500 }, { "epoch": 0.08, "learning_rate": 1.8486082229525075e-05, "loss": 0.9385, "step": 210000 }, { "epoch": 0.08, "learning_rate": 1.8482477663404897e-05, "loss": 0.9219, "step": 210500 }, { "epoch": 0.08, "learning_rate": 1.847887309728472e-05, "loss": 0.9254, "step": 211000 }, { "epoch": 0.08, "learning_rate": 1.8475268531164537e-05, "loss": 0.9354, "step": 211500 }, { "epoch": 0.08, "learning_rate": 1.8471663965044363e-05, "loss": 0.9138, "step": 212000 }, { "epoch": 0.08, "learning_rate": 1.8468059398924185e-05, "loss": 0.9511, "step": 212500 }, { "epoch": 0.08, "learning_rate": 1.8464454832804003e-05, "loss": 0.9358, "step": 213000 }, { "epoch": 0.08, "learning_rate": 1.8460850266683825e-05, "loss": 0.9444, "step": 213500 }, { "epoch": 0.08, "learning_rate": 1.8457245700563647e-05, "loss": 0.9146, "step": 214000 }, { "epoch": 0.08, "learning_rate": 1.845364113444347e-05, "loss": 0.9127, "step": 214500 }, { "epoch": 0.08, "learning_rate": 1.845003656832329e-05, "loss": 0.9655, "step": 215000 }, { "epoch": 0.08, "learning_rate": 1.8446432002203113e-05, "loss": 0.8885, "step": 215500 }, { "epoch": 0.08, "learning_rate": 1.8442827436082932e-05, "loss": 0.9198, "step": 216000 }, { "epoch": 0.08, "learning_rate": 1.8439222869962757e-05, "loss": 0.9342, "step": 216500 }, { "epoch": 0.08, "learning_rate": 1.8435618303842576e-05, "loss": 0.908, "step": 217000 }, { "epoch": 0.08, "learning_rate": 1.8432013737722398e-05, "loss": 0.954, "step": 217500 }, { "epoch": 0.08, "learning_rate": 1.842840917160222e-05, "loss": 0.9286, "step": 218000 }, { "epoch": 0.08, "learning_rate": 1.8424804605482042e-05, "loss": 0.9307, "step": 218500 }, { "epoch": 0.08, "learning_rate": 1.8421200039361864e-05, "loss": 0.9651, "step": 219000 }, { "epoch": 0.08, "learning_rate": 1.8417595473241686e-05, "loss": 0.8838, "step": 219500 }, { "epoch": 0.08, "learning_rate": 1.8413990907121505e-05, "loss": 0.954, "step": 220000 }, { "epoch": 0.08, "learning_rate": 1.8410386341001327e-05, "loss": 0.9363, "step": 220500 }, { "epoch": 0.08, "learning_rate": 1.8406781774881152e-05, "loss": 0.9765, "step": 221000 }, { "epoch": 0.08, "learning_rate": 1.840317720876097e-05, "loss": 0.9159, "step": 221500 }, { "epoch": 0.08, "learning_rate": 1.8399572642640793e-05, "loss": 0.8963, "step": 222000 }, { "epoch": 0.08, "learning_rate": 1.8395968076520615e-05, "loss": 0.9495, "step": 222500 }, { "epoch": 0.08, "learning_rate": 1.8392363510400437e-05, "loss": 0.9163, "step": 223000 }, { "epoch": 0.08, "learning_rate": 1.838875894428026e-05, "loss": 0.8715, "step": 223500 }, { "epoch": 0.08, "learning_rate": 1.838515437816008e-05, "loss": 0.9254, "step": 224000 }, { "epoch": 0.08, "learning_rate": 1.83815498120399e-05, "loss": 0.9142, "step": 224500 }, { "epoch": 0.08, "learning_rate": 1.8377945245919725e-05, "loss": 0.9636, "step": 225000 }, { "epoch": 0.08, "learning_rate": 1.8374340679799543e-05, "loss": 0.9079, "step": 225500 }, { "epoch": 0.08, "learning_rate": 1.8370736113679365e-05, "loss": 0.8869, "step": 226000 }, { "epoch": 0.08, "learning_rate": 1.8367131547559187e-05, "loss": 0.9191, "step": 226500 }, { "epoch": 0.08, "learning_rate": 1.836352698143901e-05, "loss": 0.9259, "step": 227000 }, { "epoch": 0.08, "learning_rate": 1.835992241531883e-05, "loss": 0.9092, "step": 227500 }, { "epoch": 0.08, "learning_rate": 1.8356317849198654e-05, "loss": 0.8842, "step": 228000 }, { "epoch": 0.08, "learning_rate": 1.8352713283078472e-05, "loss": 0.9509, "step": 228500 }, { "epoch": 0.08, "learning_rate": 1.8349108716958294e-05, "loss": 0.9118, "step": 229000 }, { "epoch": 0.08, "learning_rate": 1.834550415083812e-05, "loss": 0.9391, "step": 229500 }, { "epoch": 0.08, "learning_rate": 1.8341899584717938e-05, "loss": 0.9577, "step": 230000 }, { "epoch": 0.08, "learning_rate": 1.833829501859776e-05, "loss": 0.9021, "step": 230500 }, { "epoch": 0.08, "learning_rate": 1.8334690452477582e-05, "loss": 0.9076, "step": 231000 }, { "epoch": 0.08, "learning_rate": 1.8331085886357404e-05, "loss": 0.9268, "step": 231500 }, { "epoch": 0.08, "learning_rate": 1.8327481320237226e-05, "loss": 0.9075, "step": 232000 }, { "epoch": 0.08, "learning_rate": 1.8323876754117048e-05, "loss": 0.902, "step": 232500 }, { "epoch": 0.08, "learning_rate": 1.8320272187996867e-05, "loss": 0.9148, "step": 233000 }, { "epoch": 0.08, "learning_rate": 1.831666762187669e-05, "loss": 0.9065, "step": 233500 }, { "epoch": 0.08, "learning_rate": 1.831306305575651e-05, "loss": 0.9263, "step": 234000 }, { "epoch": 0.08, "learning_rate": 1.8309458489636333e-05, "loss": 0.934, "step": 234500 }, { "epoch": 0.08, "learning_rate": 1.8305853923516155e-05, "loss": 0.9218, "step": 235000 }, { "epoch": 0.08, "learning_rate": 1.8302249357395977e-05, "loss": 0.9392, "step": 235500 }, { "epoch": 0.09, "learning_rate": 1.8298644791275796e-05, "loss": 0.9366, "step": 236000 }, { "epoch": 0.09, "learning_rate": 1.829504022515562e-05, "loss": 0.925, "step": 236500 }, { "epoch": 0.09, "learning_rate": 1.8291435659035443e-05, "loss": 0.9003, "step": 237000 }, { "epoch": 0.09, "learning_rate": 1.828783109291526e-05, "loss": 0.8924, "step": 237500 }, { "epoch": 0.09, "learning_rate": 1.8284226526795084e-05, "loss": 0.9149, "step": 238000 }, { "epoch": 0.09, "learning_rate": 1.8280621960674906e-05, "loss": 0.9238, "step": 238500 }, { "epoch": 0.09, "learning_rate": 1.8277017394554728e-05, "loss": 0.9455, "step": 239000 }, { "epoch": 0.09, "learning_rate": 1.827341282843455e-05, "loss": 0.9043, "step": 239500 }, { "epoch": 0.09, "learning_rate": 1.8269808262314372e-05, "loss": 0.9102, "step": 240000 }, { "epoch": 0.09, "learning_rate": 1.8266203696194194e-05, "loss": 0.9356, "step": 240500 }, { "epoch": 0.09, "learning_rate": 1.8262599130074016e-05, "loss": 0.8955, "step": 241000 }, { "epoch": 0.09, "learning_rate": 1.8258994563953834e-05, "loss": 0.8979, "step": 241500 }, { "epoch": 0.09, "learning_rate": 1.8255389997833656e-05, "loss": 0.9249, "step": 242000 }, { "epoch": 0.09, "learning_rate": 1.825178543171348e-05, "loss": 0.8999, "step": 242500 }, { "epoch": 0.09, "learning_rate": 1.82481808655933e-05, "loss": 0.9347, "step": 243000 }, { "epoch": 0.09, "learning_rate": 1.8244576299473122e-05, "loss": 0.9074, "step": 243500 }, { "epoch": 0.09, "learning_rate": 1.8240971733352944e-05, "loss": 0.9401, "step": 244000 }, { "epoch": 0.09, "learning_rate": 1.8237367167232763e-05, "loss": 0.9116, "step": 244500 }, { "epoch": 0.09, "learning_rate": 1.823376260111259e-05, "loss": 0.9174, "step": 245000 }, { "epoch": 0.09, "learning_rate": 1.823015803499241e-05, "loss": 0.8801, "step": 245500 }, { "epoch": 0.09, "learning_rate": 1.822655346887223e-05, "loss": 0.922, "step": 246000 }, { "epoch": 0.09, "learning_rate": 1.822294890275205e-05, "loss": 0.917, "step": 246500 }, { "epoch": 0.09, "learning_rate": 1.8219344336631873e-05, "loss": 0.8966, "step": 247000 }, { "epoch": 0.09, "learning_rate": 1.8215739770511695e-05, "loss": 0.8767, "step": 247500 }, { "epoch": 0.09, "learning_rate": 1.8212135204391517e-05, "loss": 0.9131, "step": 248000 }, { "epoch": 0.09, "learning_rate": 1.820853063827134e-05, "loss": 0.8907, "step": 248500 }, { "epoch": 0.09, "learning_rate": 1.8204926072151158e-05, "loss": 0.924, "step": 249000 }, { "epoch": 0.09, "learning_rate": 1.8201321506030983e-05, "loss": 0.8952, "step": 249500 }, { "epoch": 0.09, "learning_rate": 1.8197716939910802e-05, "loss": 0.9066, "step": 250000 }, { "epoch": 0.09, "learning_rate": 1.8194112373790624e-05, "loss": 0.9228, "step": 250500 }, { "epoch": 0.09, "learning_rate": 1.8190507807670446e-05, "loss": 0.9179, "step": 251000 }, { "epoch": 0.09, "learning_rate": 1.8186903241550268e-05, "loss": 0.9503, "step": 251500 }, { "epoch": 0.09, "learning_rate": 1.818329867543009e-05, "loss": 0.8989, "step": 252000 }, { "epoch": 0.09, "learning_rate": 1.8179694109309912e-05, "loss": 0.9082, "step": 252500 }, { "epoch": 0.09, "learning_rate": 1.817608954318973e-05, "loss": 0.905, "step": 253000 }, { "epoch": 0.09, "learning_rate": 1.8172484977069553e-05, "loss": 0.9223, "step": 253500 }, { "epoch": 0.09, "learning_rate": 1.8168880410949378e-05, "loss": 0.8908, "step": 254000 }, { "epoch": 0.09, "learning_rate": 1.8165275844829197e-05, "loss": 0.9015, "step": 254500 }, { "epoch": 0.09, "learning_rate": 1.816167127870902e-05, "loss": 0.8993, "step": 255000 }, { "epoch": 0.09, "learning_rate": 1.815806671258884e-05, "loss": 0.9224, "step": 255500 }, { "epoch": 0.09, "learning_rate": 1.8154462146468663e-05, "loss": 0.9017, "step": 256000 }, { "epoch": 0.09, "learning_rate": 1.8150857580348485e-05, "loss": 0.9149, "step": 256500 }, { "epoch": 0.09, "learning_rate": 1.8147253014228307e-05, "loss": 0.8939, "step": 257000 }, { "epoch": 0.09, "learning_rate": 1.8143648448108125e-05, "loss": 0.9126, "step": 257500 }, { "epoch": 0.09, "learning_rate": 1.8140043881987947e-05, "loss": 0.8938, "step": 258000 }, { "epoch": 0.09, "learning_rate": 1.813643931586777e-05, "loss": 0.8608, "step": 258500 }, { "epoch": 0.09, "learning_rate": 1.813283474974759e-05, "loss": 0.8965, "step": 259000 }, { "epoch": 0.09, "learning_rate": 1.8129230183627413e-05, "loss": 0.8874, "step": 259500 }, { "epoch": 0.09, "learning_rate": 1.8125625617507235e-05, "loss": 0.9208, "step": 260000 }, { "epoch": 0.09, "learning_rate": 1.8122021051387057e-05, "loss": 0.9213, "step": 260500 }, { "epoch": 0.09, "learning_rate": 1.811841648526688e-05, "loss": 0.884, "step": 261000 }, { "epoch": 0.09, "learning_rate": 1.81148119191467e-05, "loss": 0.9175, "step": 261500 }, { "epoch": 0.09, "learning_rate": 1.811120735302652e-05, "loss": 0.9067, "step": 262000 }, { "epoch": 0.09, "learning_rate": 1.8107602786906342e-05, "loss": 0.9296, "step": 262500 }, { "epoch": 0.09, "learning_rate": 1.8103998220786164e-05, "loss": 0.9063, "step": 263000 }, { "epoch": 0.09, "learning_rate": 1.8100393654665986e-05, "loss": 0.9062, "step": 263500 }, { "epoch": 0.1, "learning_rate": 1.8096789088545808e-05, "loss": 0.9025, "step": 264000 }, { "epoch": 0.1, "learning_rate": 1.809318452242563e-05, "loss": 0.8935, "step": 264500 }, { "epoch": 0.1, "learning_rate": 1.8089579956305452e-05, "loss": 0.905, "step": 265000 }, { "epoch": 0.1, "learning_rate": 1.8085975390185274e-05, "loss": 0.9047, "step": 265500 }, { "epoch": 0.1, "learning_rate": 1.8082370824065093e-05, "loss": 0.9192, "step": 266000 }, { "epoch": 0.1, "learning_rate": 1.8078766257944915e-05, "loss": 0.9441, "step": 266500 }, { "epoch": 0.1, "learning_rate": 1.8075161691824737e-05, "loss": 0.8962, "step": 267000 }, { "epoch": 0.1, "learning_rate": 1.807155712570456e-05, "loss": 0.888, "step": 267500 }, { "epoch": 0.1, "learning_rate": 1.806795255958438e-05, "loss": 0.8642, "step": 268000 }, { "epoch": 0.1, "learning_rate": 1.8064347993464203e-05, "loss": 0.8942, "step": 268500 }, { "epoch": 0.1, "learning_rate": 1.806074342734402e-05, "loss": 0.9032, "step": 269000 }, { "epoch": 0.1, "learning_rate": 1.8057138861223847e-05, "loss": 0.8965, "step": 269500 }, { "epoch": 0.1, "learning_rate": 1.805353429510367e-05, "loss": 0.9035, "step": 270000 }, { "epoch": 0.1, "learning_rate": 1.8049929728983487e-05, "loss": 0.8972, "step": 270500 }, { "epoch": 0.1, "learning_rate": 1.804632516286331e-05, "loss": 0.8823, "step": 271000 }, { "epoch": 0.1, "learning_rate": 1.804272059674313e-05, "loss": 0.8817, "step": 271500 }, { "epoch": 0.1, "learning_rate": 1.8039116030622953e-05, "loss": 0.8939, "step": 272000 }, { "epoch": 0.1, "learning_rate": 1.8035511464502775e-05, "loss": 0.9122, "step": 272500 }, { "epoch": 0.1, "learning_rate": 1.8031906898382598e-05, "loss": 0.9115, "step": 273000 }, { "epoch": 0.1, "learning_rate": 1.8028302332262416e-05, "loss": 0.8999, "step": 273500 }, { "epoch": 0.1, "learning_rate": 1.802469776614224e-05, "loss": 0.8963, "step": 274000 }, { "epoch": 0.1, "learning_rate": 1.802109320002206e-05, "loss": 0.9361, "step": 274500 }, { "epoch": 0.1, "learning_rate": 1.8017488633901882e-05, "loss": 0.8798, "step": 275000 }, { "epoch": 0.1, "learning_rate": 1.8013884067781704e-05, "loss": 0.8807, "step": 275500 }, { "epoch": 0.1, "learning_rate": 1.8010279501661526e-05, "loss": 0.9041, "step": 276000 }, { "epoch": 0.1, "learning_rate": 1.8006674935541348e-05, "loss": 0.877, "step": 276500 }, { "epoch": 0.1, "learning_rate": 1.800307036942117e-05, "loss": 0.8729, "step": 277000 }, { "epoch": 0.1, "learning_rate": 1.7999465803300992e-05, "loss": 0.8863, "step": 277500 }, { "epoch": 0.1, "learning_rate": 1.799586123718081e-05, "loss": 0.9174, "step": 278000 }, { "epoch": 0.1, "learning_rate": 1.7992256671060636e-05, "loss": 0.8749, "step": 278500 }, { "epoch": 0.1, "learning_rate": 1.7988652104940455e-05, "loss": 0.8967, "step": 279000 }, { "epoch": 0.1, "learning_rate": 1.7985047538820277e-05, "loss": 0.8616, "step": 279500 }, { "epoch": 0.1, "learning_rate": 1.79814429727001e-05, "loss": 0.8462, "step": 280000 }, { "epoch": 0.1, "learning_rate": 1.797783840657992e-05, "loss": 0.8347, "step": 280500 }, { "epoch": 0.1, "learning_rate": 1.7974233840459743e-05, "loss": 0.904, "step": 281000 }, { "epoch": 0.1, "learning_rate": 1.7970629274339565e-05, "loss": 0.8686, "step": 281500 }, { "epoch": 0.1, "learning_rate": 1.7967024708219384e-05, "loss": 0.9174, "step": 282000 }, { "epoch": 0.1, "learning_rate": 1.7963420142099206e-05, "loss": 0.869, "step": 282500 }, { "epoch": 0.1, "learning_rate": 1.795981557597903e-05, "loss": 0.9047, "step": 283000 }, { "epoch": 0.1, "learning_rate": 1.795621100985885e-05, "loss": 0.9001, "step": 283500 }, { "epoch": 0.1, "learning_rate": 1.795260644373867e-05, "loss": 0.8774, "step": 284000 }, { "epoch": 0.1, "learning_rate": 1.7949001877618494e-05, "loss": 0.8865, "step": 284500 }, { "epoch": 0.1, "learning_rate": 1.7945397311498316e-05, "loss": 0.8657, "step": 285000 }, { "epoch": 0.1, "learning_rate": 1.7941792745378138e-05, "loss": 0.9077, "step": 285500 }, { "epoch": 0.1, "learning_rate": 1.793818817925796e-05, "loss": 0.8887, "step": 286000 }, { "epoch": 0.1, "learning_rate": 1.793458361313778e-05, "loss": 0.8827, "step": 286500 }, { "epoch": 0.1, "learning_rate": 1.79309790470176e-05, "loss": 0.8946, "step": 287000 }, { "epoch": 0.1, "learning_rate": 1.7927374480897422e-05, "loss": 0.8768, "step": 287500 }, { "epoch": 0.1, "learning_rate": 1.7923769914777244e-05, "loss": 0.8706, "step": 288000 }, { "epoch": 0.1, "learning_rate": 1.7920165348657066e-05, "loss": 0.9003, "step": 288500 }, { "epoch": 0.1, "learning_rate": 1.791656078253689e-05, "loss": 0.8605, "step": 289000 }, { "epoch": 0.1, "learning_rate": 1.791295621641671e-05, "loss": 0.8974, "step": 289500 }, { "epoch": 0.1, "learning_rate": 1.7909351650296532e-05, "loss": 0.8761, "step": 290000 }, { "epoch": 0.1, "learning_rate": 1.790574708417635e-05, "loss": 0.8771, "step": 290500 }, { "epoch": 0.1, "learning_rate": 1.7902142518056173e-05, "loss": 0.8815, "step": 291000 }, { "epoch": 0.11, "learning_rate": 1.7898537951936e-05, "loss": 0.8829, "step": 291500 }, { "epoch": 0.11, "learning_rate": 1.7894933385815817e-05, "loss": 0.8649, "step": 292000 }, { "epoch": 0.11, "learning_rate": 1.789132881969564e-05, "loss": 0.8749, "step": 292500 }, { "epoch": 0.11, "learning_rate": 1.788772425357546e-05, "loss": 0.8849, "step": 293000 }, { "epoch": 0.11, "learning_rate": 1.788411968745528e-05, "loss": 0.8393, "step": 293500 }, { "epoch": 0.11, "learning_rate": 1.7880515121335105e-05, "loss": 0.8907, "step": 294000 }, { "epoch": 0.11, "learning_rate": 1.7876910555214927e-05, "loss": 0.8892, "step": 294500 }, { "epoch": 0.11, "learning_rate": 1.7873305989094746e-05, "loss": 0.8639, "step": 295000 }, { "epoch": 0.11, "learning_rate": 1.7869701422974568e-05, "loss": 0.9086, "step": 295500 }, { "epoch": 0.11, "learning_rate": 1.786609685685439e-05, "loss": 0.8922, "step": 296000 }, { "epoch": 0.11, "learning_rate": 1.7862492290734212e-05, "loss": 0.8786, "step": 296500 }, { "epoch": 0.11, "learning_rate": 1.7858887724614034e-05, "loss": 0.9119, "step": 297000 }, { "epoch": 0.11, "learning_rate": 1.7855283158493856e-05, "loss": 0.8669, "step": 297500 }, { "epoch": 0.11, "learning_rate": 1.7851678592373674e-05, "loss": 0.8817, "step": 298000 }, { "epoch": 0.11, "learning_rate": 1.78480740262535e-05, "loss": 0.8924, "step": 298500 }, { "epoch": 0.11, "learning_rate": 1.784446946013332e-05, "loss": 0.8883, "step": 299000 }, { "epoch": 0.11, "learning_rate": 1.784086489401314e-05, "loss": 0.894, "step": 299500 }, { "epoch": 0.11, "learning_rate": 1.7837260327892963e-05, "loss": 0.9032, "step": 300000 }, { "epoch": 0.11, "learning_rate": 1.7833655761772785e-05, "loss": 0.8864, "step": 300500 }, { "epoch": 0.11, "learning_rate": 1.7830051195652607e-05, "loss": 0.8789, "step": 301000 }, { "epoch": 0.11, "learning_rate": 1.782644662953243e-05, "loss": 0.9173, "step": 301500 }, { "epoch": 0.11, "learning_rate": 1.782284206341225e-05, "loss": 0.8823, "step": 302000 }, { "epoch": 0.11, "learning_rate": 1.781923749729207e-05, "loss": 0.9193, "step": 302500 }, { "epoch": 0.11, "learning_rate": 1.7815632931171895e-05, "loss": 0.8839, "step": 303000 }, { "epoch": 0.11, "learning_rate": 1.7812028365051713e-05, "loss": 0.8869, "step": 303500 }, { "epoch": 0.11, "learning_rate": 1.7808423798931535e-05, "loss": 0.8687, "step": 304000 }, { "epoch": 0.11, "learning_rate": 1.7804819232811357e-05, "loss": 0.8759, "step": 304500 }, { "epoch": 0.11, "learning_rate": 1.780121466669118e-05, "loss": 0.8821, "step": 305000 }, { "epoch": 0.11, "learning_rate": 1.7797610100571e-05, "loss": 0.8856, "step": 305500 }, { "epoch": 0.11, "learning_rate": 1.7794005534450823e-05, "loss": 0.8731, "step": 306000 }, { "epoch": 0.11, "learning_rate": 1.7790400968330642e-05, "loss": 0.8906, "step": 306500 }, { "epoch": 0.11, "learning_rate": 1.7786796402210467e-05, "loss": 0.8904, "step": 307000 }, { "epoch": 0.11, "learning_rate": 1.778319183609029e-05, "loss": 0.8519, "step": 307500 }, { "epoch": 0.11, "learning_rate": 1.7779587269970108e-05, "loss": 0.8794, "step": 308000 }, { "epoch": 0.11, "learning_rate": 1.777598270384993e-05, "loss": 0.8783, "step": 308500 }, { "epoch": 0.11, "learning_rate": 1.7772378137729752e-05, "loss": 0.8615, "step": 309000 }, { "epoch": 0.11, "learning_rate": 1.7768773571609574e-05, "loss": 0.8793, "step": 309500 }, { "epoch": 0.11, "learning_rate": 1.7765169005489396e-05, "loss": 0.8552, "step": 310000 }, { "epoch": 0.11, "learning_rate": 1.7761564439369218e-05, "loss": 0.9017, "step": 310500 }, { "epoch": 0.11, "learning_rate": 1.7757959873249037e-05, "loss": 0.8427, "step": 311000 }, { "epoch": 0.11, "learning_rate": 1.7754355307128862e-05, "loss": 0.9001, "step": 311500 }, { "epoch": 0.11, "learning_rate": 1.775075074100868e-05, "loss": 0.883, "step": 312000 }, { "epoch": 0.11, "learning_rate": 1.7747146174888503e-05, "loss": 0.8712, "step": 312500 }, { "epoch": 0.11, "learning_rate": 1.7743541608768325e-05, "loss": 0.8737, "step": 313000 }, { "epoch": 0.11, "learning_rate": 1.7739937042648147e-05, "loss": 0.9106, "step": 313500 }, { "epoch": 0.11, "learning_rate": 1.773633247652797e-05, "loss": 0.8705, "step": 314000 }, { "epoch": 0.11, "learning_rate": 1.773272791040779e-05, "loss": 0.8948, "step": 314500 }, { "epoch": 0.11, "learning_rate": 1.772912334428761e-05, "loss": 0.8595, "step": 315000 }, { "epoch": 0.11, "learning_rate": 1.772551877816743e-05, "loss": 0.8694, "step": 315500 }, { "epoch": 0.11, "learning_rate": 1.7721914212047257e-05, "loss": 0.8882, "step": 316000 }, { "epoch": 0.11, "learning_rate": 1.7718309645927075e-05, "loss": 0.852, "step": 316500 }, { "epoch": 0.11, "learning_rate": 1.7714705079806897e-05, "loss": 0.8647, "step": 317000 }, { "epoch": 0.11, "learning_rate": 1.771110051368672e-05, "loss": 0.8771, "step": 317500 }, { "epoch": 0.11, "learning_rate": 1.770749594756654e-05, "loss": 0.8851, "step": 318000 }, { "epoch": 0.11, "learning_rate": 1.7703891381446363e-05, "loss": 0.8985, "step": 318500 }, { "epoch": 0.11, "learning_rate": 1.7700286815326185e-05, "loss": 0.8712, "step": 319000 }, { "epoch": 0.12, "learning_rate": 1.7696682249206004e-05, "loss": 0.8985, "step": 319500 }, { "epoch": 0.12, "learning_rate": 1.7693077683085826e-05, "loss": 0.8603, "step": 320000 }, { "epoch": 0.12, "learning_rate": 1.7689473116965648e-05, "loss": 0.9056, "step": 320500 }, { "epoch": 0.12, "learning_rate": 1.768586855084547e-05, "loss": 0.8887, "step": 321000 }, { "epoch": 0.12, "learning_rate": 1.7682263984725292e-05, "loss": 0.8799, "step": 321500 }, { "epoch": 0.12, "learning_rate": 1.7678659418605114e-05, "loss": 0.8571, "step": 322000 }, { "epoch": 0.12, "learning_rate": 1.7675054852484936e-05, "loss": 0.877, "step": 322500 }, { "epoch": 0.12, "learning_rate": 1.7671450286364758e-05, "loss": 0.8798, "step": 323000 }, { "epoch": 0.12, "learning_rate": 1.766784572024458e-05, "loss": 0.868, "step": 323500 }, { "epoch": 0.12, "learning_rate": 1.76642411541244e-05, "loss": 0.8573, "step": 324000 }, { "epoch": 0.12, "learning_rate": 1.766063658800422e-05, "loss": 0.8514, "step": 324500 }, { "epoch": 0.12, "learning_rate": 1.7657032021884043e-05, "loss": 0.8449, "step": 325000 }, { "epoch": 0.12, "learning_rate": 1.7653427455763865e-05, "loss": 0.89, "step": 325500 }, { "epoch": 0.12, "learning_rate": 1.7649822889643687e-05, "loss": 0.8606, "step": 326000 }, { "epoch": 0.12, "learning_rate": 1.764621832352351e-05, "loss": 0.8826, "step": 326500 }, { "epoch": 0.12, "learning_rate": 1.764261375740333e-05, "loss": 0.8547, "step": 327000 }, { "epoch": 0.12, "learning_rate": 1.7639009191283153e-05, "loss": 0.8653, "step": 327500 }, { "epoch": 0.12, "learning_rate": 1.763540462516297e-05, "loss": 0.8469, "step": 328000 }, { "epoch": 0.12, "learning_rate": 1.7631800059042794e-05, "loss": 0.8702, "step": 328500 }, { "epoch": 0.12, "learning_rate": 1.7628195492922616e-05, "loss": 0.8564, "step": 329000 }, { "epoch": 0.12, "learning_rate": 1.7624590926802438e-05, "loss": 0.8829, "step": 329500 }, { "epoch": 0.12, "learning_rate": 1.762098636068226e-05, "loss": 0.84, "step": 330000 }, { "epoch": 0.12, "learning_rate": 1.761738179456208e-05, "loss": 0.8439, "step": 330500 }, { "epoch": 0.12, "learning_rate": 1.76137772284419e-05, "loss": 0.856, "step": 331000 }, { "epoch": 0.12, "learning_rate": 1.7610172662321726e-05, "loss": 0.8566, "step": 331500 }, { "epoch": 0.12, "learning_rate": 1.7606568096201548e-05, "loss": 0.8851, "step": 332000 }, { "epoch": 0.12, "learning_rate": 1.7602963530081366e-05, "loss": 0.8694, "step": 332500 }, { "epoch": 0.12, "learning_rate": 1.759935896396119e-05, "loss": 0.8383, "step": 333000 }, { "epoch": 0.12, "learning_rate": 1.759575439784101e-05, "loss": 0.8479, "step": 333500 }, { "epoch": 0.12, "learning_rate": 1.7592149831720832e-05, "loss": 0.8564, "step": 334000 }, { "epoch": 0.12, "learning_rate": 1.7588545265600654e-05, "loss": 0.8961, "step": 334500 }, { "epoch": 0.12, "learning_rate": 1.7584940699480476e-05, "loss": 0.8586, "step": 335000 }, { "epoch": 0.12, "learning_rate": 1.7581336133360295e-05, "loss": 0.853, "step": 335500 }, { "epoch": 0.12, "learning_rate": 1.757773156724012e-05, "loss": 0.8978, "step": 336000 }, { "epoch": 0.12, "learning_rate": 1.757412700111994e-05, "loss": 0.8346, "step": 336500 }, { "epoch": 0.12, "learning_rate": 1.757052243499976e-05, "loss": 0.8741, "step": 337000 }, { "epoch": 0.12, "learning_rate": 1.7566917868879583e-05, "loss": 0.8489, "step": 337500 }, { "epoch": 0.12, "learning_rate": 1.7563313302759405e-05, "loss": 0.8585, "step": 338000 }, { "epoch": 0.12, "learning_rate": 1.7559708736639227e-05, "loss": 0.9047, "step": 338500 }, { "epoch": 0.12, "learning_rate": 1.755610417051905e-05, "loss": 0.8765, "step": 339000 }, { "epoch": 0.12, "learning_rate": 1.7552499604398868e-05, "loss": 0.8706, "step": 339500 }, { "epoch": 0.12, "learning_rate": 1.754889503827869e-05, "loss": 0.8678, "step": 340000 }, { "epoch": 0.12, "learning_rate": 1.7545290472158515e-05, "loss": 0.8664, "step": 340500 }, { "epoch": 0.12, "learning_rate": 1.7541685906038334e-05, "loss": 0.8786, "step": 341000 }, { "epoch": 0.12, "learning_rate": 1.7538081339918156e-05, "loss": 0.861, "step": 341500 }, { "epoch": 0.12, "learning_rate": 1.7534476773797978e-05, "loss": 0.8583, "step": 342000 }, { "epoch": 0.12, "learning_rate": 1.75308722076778e-05, "loss": 0.8845, "step": 342500 }, { "epoch": 0.12, "learning_rate": 1.7527267641557622e-05, "loss": 0.8812, "step": 343000 }, { "epoch": 0.12, "learning_rate": 1.7523663075437444e-05, "loss": 0.878, "step": 343500 }, { "epoch": 0.12, "learning_rate": 1.7520058509317262e-05, "loss": 0.8458, "step": 344000 }, { "epoch": 0.12, "learning_rate": 1.7516453943197084e-05, "loss": 0.8416, "step": 344500 }, { "epoch": 0.12, "learning_rate": 1.7512849377076906e-05, "loss": 0.8629, "step": 345000 }, { "epoch": 0.12, "learning_rate": 1.750924481095673e-05, "loss": 0.8672, "step": 345500 }, { "epoch": 0.12, "learning_rate": 1.750564024483655e-05, "loss": 0.847, "step": 346000 }, { "epoch": 0.12, "learning_rate": 1.7502035678716373e-05, "loss": 0.865, "step": 346500 }, { "epoch": 0.13, "learning_rate": 1.7498431112596195e-05, "loss": 0.8453, "step": 347000 }, { "epoch": 0.13, "learning_rate": 1.7494826546476017e-05, "loss": 0.8636, "step": 347500 }, { "epoch": 0.13, "learning_rate": 1.749122198035584e-05, "loss": 0.8505, "step": 348000 }, { "epoch": 0.13, "learning_rate": 1.7487617414235657e-05, "loss": 0.8571, "step": 348500 }, { "epoch": 0.13, "learning_rate": 1.748401284811548e-05, "loss": 0.86, "step": 349000 }, { "epoch": 0.13, "learning_rate": 1.74804082819953e-05, "loss": 0.8564, "step": 349500 }, { "epoch": 0.13, "learning_rate": 1.7476803715875123e-05, "loss": 0.8774, "step": 350000 }, { "epoch": 0.13, "learning_rate": 1.7473199149754945e-05, "loss": 0.8511, "step": 350500 }, { "epoch": 0.13, "learning_rate": 1.7469594583634767e-05, "loss": 0.8884, "step": 351000 }, { "epoch": 0.13, "learning_rate": 1.746599001751459e-05, "loss": 0.8753, "step": 351500 }, { "epoch": 0.13, "learning_rate": 1.746238545139441e-05, "loss": 0.8339, "step": 352000 }, { "epoch": 0.13, "learning_rate": 1.745878088527423e-05, "loss": 0.8896, "step": 352500 }, { "epoch": 0.13, "learning_rate": 1.7455176319154052e-05, "loss": 0.9024, "step": 353000 }, { "epoch": 0.13, "learning_rate": 1.7451571753033874e-05, "loss": 0.844, "step": 353500 }, { "epoch": 0.13, "learning_rate": 1.7447967186913696e-05, "loss": 0.8594, "step": 354000 }, { "epoch": 0.13, "learning_rate": 1.7444362620793518e-05, "loss": 0.8699, "step": 354500 }, { "epoch": 0.13, "learning_rate": 1.744075805467334e-05, "loss": 0.8427, "step": 355000 }, { "epoch": 0.13, "learning_rate": 1.743715348855316e-05, "loss": 0.8572, "step": 355500 }, { "epoch": 0.13, "learning_rate": 1.7433548922432984e-05, "loss": 0.8786, "step": 356000 }, { "epoch": 0.13, "learning_rate": 1.7429944356312806e-05, "loss": 0.896, "step": 356500 }, { "epoch": 0.13, "learning_rate": 1.7426339790192625e-05, "loss": 0.8192, "step": 357000 }, { "epoch": 0.13, "learning_rate": 1.7422735224072447e-05, "loss": 0.8702, "step": 357500 }, { "epoch": 0.13, "learning_rate": 1.741913065795227e-05, "loss": 0.888, "step": 358000 }, { "epoch": 0.13, "learning_rate": 1.741552609183209e-05, "loss": 0.834, "step": 358500 }, { "epoch": 0.13, "learning_rate": 1.7411921525711913e-05, "loss": 0.8609, "step": 359000 }, { "epoch": 0.13, "learning_rate": 1.7408316959591735e-05, "loss": 0.8496, "step": 359500 }, { "epoch": 0.13, "learning_rate": 1.7404712393471553e-05, "loss": 0.8517, "step": 360000 }, { "epoch": 0.13, "learning_rate": 1.740110782735138e-05, "loss": 0.8672, "step": 360500 }, { "epoch": 0.13, "learning_rate": 1.7397503261231197e-05, "loss": 0.8208, "step": 361000 }, { "epoch": 0.13, "learning_rate": 1.739389869511102e-05, "loss": 0.8571, "step": 361500 }, { "epoch": 0.13, "learning_rate": 1.739029412899084e-05, "loss": 0.8303, "step": 362000 }, { "epoch": 0.13, "learning_rate": 1.7386689562870663e-05, "loss": 0.8389, "step": 362500 }, { "epoch": 0.13, "learning_rate": 1.7383084996750485e-05, "loss": 0.8751, "step": 363000 }, { "epoch": 0.13, "learning_rate": 1.7379480430630307e-05, "loss": 0.8488, "step": 363500 }, { "epoch": 0.13, "learning_rate": 1.7375875864510126e-05, "loss": 0.8565, "step": 364000 }, { "epoch": 0.13, "learning_rate": 1.7372271298389948e-05, "loss": 0.8516, "step": 364500 }, { "epoch": 0.13, "learning_rate": 1.7368666732269773e-05, "loss": 0.8488, "step": 365000 }, { "epoch": 0.13, "learning_rate": 1.7365062166149592e-05, "loss": 0.8898, "step": 365500 }, { "epoch": 0.13, "learning_rate": 1.7361457600029414e-05, "loss": 0.8203, "step": 366000 }, { "epoch": 0.13, "learning_rate": 1.7357853033909236e-05, "loss": 0.8526, "step": 366500 }, { "epoch": 0.13, "learning_rate": 1.7354248467789058e-05, "loss": 0.8465, "step": 367000 }, { "epoch": 0.13, "learning_rate": 1.735064390166888e-05, "loss": 0.8819, "step": 367500 }, { "epoch": 0.13, "learning_rate": 1.7347039335548702e-05, "loss": 0.8733, "step": 368000 }, { "epoch": 0.13, "learning_rate": 1.734343476942852e-05, "loss": 0.8293, "step": 368500 }, { "epoch": 0.13, "learning_rate": 1.7339830203308343e-05, "loss": 0.8598, "step": 369000 }, { "epoch": 0.13, "learning_rate": 1.7336225637188168e-05, "loss": 0.8139, "step": 369500 }, { "epoch": 0.13, "learning_rate": 1.7332621071067987e-05, "loss": 0.8654, "step": 370000 }, { "epoch": 0.13, "learning_rate": 1.732901650494781e-05, "loss": 0.8662, "step": 370500 }, { "epoch": 0.13, "learning_rate": 1.732541193882763e-05, "loss": 0.8556, "step": 371000 }, { "epoch": 0.13, "learning_rate": 1.7321807372707453e-05, "loss": 0.8538, "step": 371500 }, { "epoch": 0.13, "learning_rate": 1.7318202806587275e-05, "loss": 0.8648, "step": 372000 }, { "epoch": 0.13, "learning_rate": 1.7314598240467097e-05, "loss": 0.8327, "step": 372500 }, { "epoch": 0.13, "learning_rate": 1.7310993674346916e-05, "loss": 0.8686, "step": 373000 }, { "epoch": 0.13, "learning_rate": 1.730738910822674e-05, "loss": 0.8478, "step": 373500 }, { "epoch": 0.13, "learning_rate": 1.730378454210656e-05, "loss": 0.8358, "step": 374000 }, { "epoch": 0.13, "learning_rate": 1.730017997598638e-05, "loss": 0.8339, "step": 374500 }, { "epoch": 0.14, "learning_rate": 1.7296575409866204e-05, "loss": 0.8372, "step": 375000 }, { "epoch": 0.14, "learning_rate": 1.7292970843746026e-05, "loss": 0.8325, "step": 375500 }, { "epoch": 0.14, "learning_rate": 1.7289366277625848e-05, "loss": 0.8546, "step": 376000 }, { "epoch": 0.14, "learning_rate": 1.728576171150567e-05, "loss": 0.8654, "step": 376500 }, { "epoch": 0.14, "learning_rate": 1.7282157145385488e-05, "loss": 0.8471, "step": 377000 }, { "epoch": 0.14, "learning_rate": 1.727855257926531e-05, "loss": 0.843, "step": 377500 }, { "epoch": 0.14, "learning_rate": 1.7274948013145136e-05, "loss": 0.8578, "step": 378000 }, { "epoch": 0.14, "learning_rate": 1.7271343447024954e-05, "loss": 0.8682, "step": 378500 }, { "epoch": 0.14, "learning_rate": 1.7267738880904776e-05, "loss": 0.8306, "step": 379000 }, { "epoch": 0.14, "learning_rate": 1.72641343147846e-05, "loss": 0.8434, "step": 379500 }, { "epoch": 0.14, "learning_rate": 1.7260529748664417e-05, "loss": 0.8179, "step": 380000 }, { "epoch": 0.14, "learning_rate": 1.7256925182544242e-05, "loss": 0.8703, "step": 380500 }, { "epoch": 0.14, "learning_rate": 1.7253320616424064e-05, "loss": 0.8487, "step": 381000 }, { "epoch": 0.14, "learning_rate": 1.7249716050303883e-05, "loss": 0.8557, "step": 381500 }, { "epoch": 0.14, "learning_rate": 1.7246111484183705e-05, "loss": 0.8333, "step": 382000 }, { "epoch": 0.14, "learning_rate": 1.7242506918063527e-05, "loss": 0.8432, "step": 382500 }, { "epoch": 0.14, "learning_rate": 1.723890235194335e-05, "loss": 0.8537, "step": 383000 }, { "epoch": 0.14, "learning_rate": 1.723529778582317e-05, "loss": 0.852, "step": 383500 }, { "epoch": 0.14, "learning_rate": 1.7231693219702993e-05, "loss": 0.8728, "step": 384000 }, { "epoch": 0.14, "learning_rate": 1.7228088653582812e-05, "loss": 0.8766, "step": 384500 }, { "epoch": 0.14, "learning_rate": 1.7224484087462637e-05, "loss": 0.8359, "step": 385000 }, { "epoch": 0.14, "learning_rate": 1.7220879521342456e-05, "loss": 0.839, "step": 385500 }, { "epoch": 0.14, "learning_rate": 1.7217274955222278e-05, "loss": 0.9147, "step": 386000 }, { "epoch": 0.14, "learning_rate": 1.72136703891021e-05, "loss": 0.8625, "step": 386500 }, { "epoch": 0.14, "learning_rate": 1.7210065822981922e-05, "loss": 0.8804, "step": 387000 }, { "epoch": 0.14, "learning_rate": 1.7206461256861744e-05, "loss": 0.8547, "step": 387500 }, { "epoch": 0.14, "learning_rate": 1.7202856690741566e-05, "loss": 0.8426, "step": 388000 }, { "epoch": 0.14, "learning_rate": 1.7199252124621388e-05, "loss": 0.8094, "step": 388500 }, { "epoch": 0.14, "learning_rate": 1.719564755850121e-05, "loss": 0.856, "step": 389000 }, { "epoch": 0.14, "learning_rate": 1.7192042992381032e-05, "loss": 0.839, "step": 389500 }, { "epoch": 0.14, "learning_rate": 1.718843842626085e-05, "loss": 0.8262, "step": 390000 }, { "epoch": 0.14, "learning_rate": 1.7184833860140672e-05, "loss": 0.8363, "step": 390500 }, { "epoch": 0.14, "learning_rate": 1.7181229294020494e-05, "loss": 0.8134, "step": 391000 }, { "epoch": 0.14, "learning_rate": 1.7177624727900316e-05, "loss": 0.8685, "step": 391500 }, { "epoch": 0.14, "learning_rate": 1.717402016178014e-05, "loss": 0.8659, "step": 392000 }, { "epoch": 0.14, "learning_rate": 1.717041559565996e-05, "loss": 0.8771, "step": 392500 }, { "epoch": 0.14, "learning_rate": 1.716681102953978e-05, "loss": 0.8506, "step": 393000 }, { "epoch": 0.14, "learning_rate": 1.7163206463419605e-05, "loss": 0.8508, "step": 393500 }, { "epoch": 0.14, "learning_rate": 1.7159601897299427e-05, "loss": 0.8619, "step": 394000 }, { "epoch": 0.14, "learning_rate": 1.7155997331179245e-05, "loss": 0.8714, "step": 394500 }, { "epoch": 0.14, "learning_rate": 1.7152392765059067e-05, "loss": 0.8421, "step": 395000 }, { "epoch": 0.14, "learning_rate": 1.714878819893889e-05, "loss": 0.8741, "step": 395500 }, { "epoch": 0.14, "learning_rate": 1.714518363281871e-05, "loss": 0.8561, "step": 396000 }, { "epoch": 0.14, "learning_rate": 1.7141579066698533e-05, "loss": 0.8418, "step": 396500 }, { "epoch": 0.14, "learning_rate": 1.7137974500578355e-05, "loss": 0.8542, "step": 397000 }, { "epoch": 0.14, "learning_rate": 1.7134369934458174e-05, "loss": 0.8565, "step": 397500 }, { "epoch": 0.14, "learning_rate": 1.7130765368338e-05, "loss": 0.8396, "step": 398000 }, { "epoch": 0.14, "learning_rate": 1.7127160802217818e-05, "loss": 0.8036, "step": 398500 }, { "epoch": 0.14, "learning_rate": 1.712355623609764e-05, "loss": 0.8637, "step": 399000 }, { "epoch": 0.14, "learning_rate": 1.7119951669977462e-05, "loss": 0.8419, "step": 399500 }, { "epoch": 0.14, "learning_rate": 1.7116347103857284e-05, "loss": 0.8656, "step": 400000 }, { "epoch": 0.14, "learning_rate": 1.7112742537737106e-05, "loss": 0.8475, "step": 400500 }, { "epoch": 0.14, "learning_rate": 1.7109137971616928e-05, "loss": 0.8752, "step": 401000 }, { "epoch": 0.14, "learning_rate": 1.7105533405496747e-05, "loss": 0.8093, "step": 401500 }, { "epoch": 0.14, "learning_rate": 1.710192883937657e-05, "loss": 0.8152, "step": 402000 }, { "epoch": 0.15, "learning_rate": 1.7098324273256394e-05, "loss": 0.8354, "step": 402500 }, { "epoch": 0.15, "learning_rate": 1.7094719707136213e-05, "loss": 0.8376, "step": 403000 }, { "epoch": 0.15, "learning_rate": 1.7091115141016035e-05, "loss": 0.801, "step": 403500 }, { "epoch": 0.15, "learning_rate": 1.7087510574895857e-05, "loss": 0.8352, "step": 404000 }, { "epoch": 0.15, "learning_rate": 1.708390600877568e-05, "loss": 0.8699, "step": 404500 }, { "epoch": 0.15, "learning_rate": 1.70803014426555e-05, "loss": 0.8603, "step": 405000 }, { "epoch": 0.15, "learning_rate": 1.7076696876535323e-05, "loss": 0.8378, "step": 405500 }, { "epoch": 0.15, "learning_rate": 1.707309231041514e-05, "loss": 0.8454, "step": 406000 }, { "epoch": 0.15, "learning_rate": 1.7069487744294963e-05, "loss": 0.8237, "step": 406500 }, { "epoch": 0.15, "learning_rate": 1.7065883178174785e-05, "loss": 0.8373, "step": 407000 }, { "epoch": 0.15, "learning_rate": 1.7062278612054607e-05, "loss": 0.8757, "step": 407500 }, { "epoch": 0.15, "learning_rate": 1.705867404593443e-05, "loss": 0.8066, "step": 408000 }, { "epoch": 0.15, "learning_rate": 1.705506947981425e-05, "loss": 0.847, "step": 408500 }, { "epoch": 0.15, "learning_rate": 1.7051464913694073e-05, "loss": 0.8169, "step": 409000 }, { "epoch": 0.15, "learning_rate": 1.7047860347573895e-05, "loss": 0.8217, "step": 409500 }, { "epoch": 0.15, "learning_rate": 1.7044255781453714e-05, "loss": 0.8241, "step": 410000 }, { "epoch": 0.15, "learning_rate": 1.7040651215333536e-05, "loss": 0.8282, "step": 410500 }, { "epoch": 0.15, "learning_rate": 1.7037046649213358e-05, "loss": 0.8391, "step": 411000 }, { "epoch": 0.15, "learning_rate": 1.703344208309318e-05, "loss": 0.8791, "step": 411500 }, { "epoch": 0.15, "learning_rate": 1.7029837516973002e-05, "loss": 0.8271, "step": 412000 }, { "epoch": 0.15, "learning_rate": 1.7026232950852824e-05, "loss": 0.8427, "step": 412500 }, { "epoch": 0.15, "learning_rate": 1.7022628384732646e-05, "loss": 0.8448, "step": 413000 }, { "epoch": 0.15, "learning_rate": 1.7019023818612468e-05, "loss": 0.8184, "step": 413500 }, { "epoch": 0.15, "learning_rate": 1.701541925249229e-05, "loss": 0.8047, "step": 414000 }, { "epoch": 0.15, "learning_rate": 1.701181468637211e-05, "loss": 0.8223, "step": 414500 }, { "epoch": 0.15, "learning_rate": 1.700821012025193e-05, "loss": 0.8339, "step": 415000 }, { "epoch": 0.15, "learning_rate": 1.7004605554131753e-05, "loss": 0.8689, "step": 415500 }, { "epoch": 0.15, "learning_rate": 1.7001000988011575e-05, "loss": 0.8455, "step": 416000 }, { "epoch": 0.15, "learning_rate": 1.6997396421891397e-05, "loss": 0.8062, "step": 416500 }, { "epoch": 0.15, "learning_rate": 1.699379185577122e-05, "loss": 0.8431, "step": 417000 }, { "epoch": 0.15, "learning_rate": 1.6990187289651037e-05, "loss": 0.8327, "step": 417500 }, { "epoch": 0.15, "learning_rate": 1.6986582723530863e-05, "loss": 0.8703, "step": 418000 }, { "epoch": 0.15, "learning_rate": 1.6982978157410685e-05, "loss": 0.853, "step": 418500 }, { "epoch": 0.15, "learning_rate": 1.6979373591290504e-05, "loss": 0.837, "step": 419000 }, { "epoch": 0.15, "learning_rate": 1.6975769025170326e-05, "loss": 0.8271, "step": 419500 }, { "epoch": 0.15, "learning_rate": 1.6972164459050148e-05, "loss": 0.8534, "step": 420000 }, { "epoch": 0.15, "learning_rate": 1.696855989292997e-05, "loss": 0.8299, "step": 420500 }, { "epoch": 0.15, "learning_rate": 1.696495532680979e-05, "loss": 0.8094, "step": 421000 }, { "epoch": 0.15, "learning_rate": 1.6961350760689614e-05, "loss": 0.8275, "step": 421500 }, { "epoch": 0.15, "learning_rate": 1.6957746194569432e-05, "loss": 0.8516, "step": 422000 }, { "epoch": 0.15, "learning_rate": 1.6954141628449258e-05, "loss": 0.8011, "step": 422500 }, { "epoch": 0.15, "learning_rate": 1.6950537062329076e-05, "loss": 0.834, "step": 423000 }, { "epoch": 0.15, "learning_rate": 1.6946932496208898e-05, "loss": 0.843, "step": 423500 }, { "epoch": 0.15, "learning_rate": 1.694332793008872e-05, "loss": 0.8436, "step": 424000 }, { "epoch": 0.15, "learning_rate": 1.6939723363968542e-05, "loss": 0.8033, "step": 424500 }, { "epoch": 0.15, "learning_rate": 1.6936118797848364e-05, "loss": 0.83, "step": 425000 }, { "epoch": 0.15, "learning_rate": 1.6932514231728186e-05, "loss": 0.8157, "step": 425500 }, { "epoch": 0.15, "learning_rate": 1.6928909665608005e-05, "loss": 0.8619, "step": 426000 }, { "epoch": 0.15, "learning_rate": 1.6925305099487827e-05, "loss": 0.8281, "step": 426500 }, { "epoch": 0.15, "learning_rate": 1.6921700533367652e-05, "loss": 0.8485, "step": 427000 }, { "epoch": 0.15, "learning_rate": 1.691809596724747e-05, "loss": 0.8217, "step": 427500 }, { "epoch": 0.15, "learning_rate": 1.6914491401127293e-05, "loss": 0.8294, "step": 428000 }, { "epoch": 0.15, "learning_rate": 1.6910886835007115e-05, "loss": 0.8317, "step": 428500 }, { "epoch": 0.15, "learning_rate": 1.6907282268886937e-05, "loss": 0.857, "step": 429000 }, { "epoch": 0.15, "learning_rate": 1.690367770276676e-05, "loss": 0.8239, "step": 429500 }, { "epoch": 0.15, "learning_rate": 1.690007313664658e-05, "loss": 0.8229, "step": 430000 }, { "epoch": 0.16, "learning_rate": 1.68964685705264e-05, "loss": 0.8185, "step": 430500 }, { "epoch": 0.16, "learning_rate": 1.6892864004406222e-05, "loss": 0.8379, "step": 431000 }, { "epoch": 0.16, "learning_rate": 1.6889259438286044e-05, "loss": 0.813, "step": 431500 }, { "epoch": 0.16, "learning_rate": 1.6885654872165866e-05, "loss": 0.8469, "step": 432000 }, { "epoch": 0.16, "learning_rate": 1.6882050306045688e-05, "loss": 0.8396, "step": 432500 }, { "epoch": 0.16, "learning_rate": 1.687844573992551e-05, "loss": 0.8144, "step": 433000 }, { "epoch": 0.16, "learning_rate": 1.6874841173805332e-05, "loss": 0.8129, "step": 433500 }, { "epoch": 0.16, "learning_rate": 1.6871236607685154e-05, "loss": 0.8467, "step": 434000 }, { "epoch": 0.16, "learning_rate": 1.6867632041564976e-05, "loss": 0.8751, "step": 434500 }, { "epoch": 0.16, "learning_rate": 1.6864027475444794e-05, "loss": 0.8297, "step": 435000 }, { "epoch": 0.16, "learning_rate": 1.6860422909324616e-05, "loss": 0.8229, "step": 435500 }, { "epoch": 0.16, "learning_rate": 1.685681834320444e-05, "loss": 0.8398, "step": 436000 }, { "epoch": 0.16, "learning_rate": 1.685321377708426e-05, "loss": 0.8443, "step": 436500 }, { "epoch": 0.16, "learning_rate": 1.6849609210964082e-05, "loss": 0.824, "step": 437000 }, { "epoch": 0.16, "learning_rate": 1.6846004644843904e-05, "loss": 0.8211, "step": 437500 }, { "epoch": 0.16, "learning_rate": 1.6842400078723727e-05, "loss": 0.8332, "step": 438000 }, { "epoch": 0.16, "learning_rate": 1.683879551260355e-05, "loss": 0.8654, "step": 438500 }, { "epoch": 0.16, "learning_rate": 1.6835190946483367e-05, "loss": 0.8266, "step": 439000 }, { "epoch": 0.16, "learning_rate": 1.683158638036319e-05, "loss": 0.8301, "step": 439500 }, { "epoch": 0.16, "learning_rate": 1.6827981814243015e-05, "loss": 0.8394, "step": 440000 }, { "epoch": 0.16, "learning_rate": 1.6824377248122833e-05, "loss": 0.8555, "step": 440500 }, { "epoch": 0.16, "learning_rate": 1.6820772682002655e-05, "loss": 0.8295, "step": 441000 }, { "epoch": 0.16, "learning_rate": 1.6817168115882477e-05, "loss": 0.8458, "step": 441500 }, { "epoch": 0.16, "learning_rate": 1.6813563549762296e-05, "loss": 0.8142, "step": 442000 }, { "epoch": 0.16, "learning_rate": 1.680995898364212e-05, "loss": 0.8118, "step": 442500 }, { "epoch": 0.16, "learning_rate": 1.6806354417521943e-05, "loss": 0.8572, "step": 443000 }, { "epoch": 0.16, "learning_rate": 1.6802749851401762e-05, "loss": 0.814, "step": 443500 }, { "epoch": 0.16, "learning_rate": 1.6799145285281584e-05, "loss": 0.8259, "step": 444000 }, { "epoch": 0.16, "learning_rate": 1.6795540719161406e-05, "loss": 0.8571, "step": 444500 }, { "epoch": 0.16, "learning_rate": 1.6791936153041228e-05, "loss": 0.8216, "step": 445000 }, { "epoch": 0.16, "learning_rate": 1.678833158692105e-05, "loss": 0.8417, "step": 445500 }, { "epoch": 0.16, "learning_rate": 1.6784727020800872e-05, "loss": 0.8275, "step": 446000 }, { "epoch": 0.16, "learning_rate": 1.678112245468069e-05, "loss": 0.8176, "step": 446500 }, { "epoch": 0.16, "learning_rate": 1.6777517888560516e-05, "loss": 0.8229, "step": 447000 }, { "epoch": 0.16, "learning_rate": 1.6773913322440335e-05, "loss": 0.8178, "step": 447500 }, { "epoch": 0.16, "learning_rate": 1.6770308756320157e-05, "loss": 0.8214, "step": 448000 }, { "epoch": 0.16, "learning_rate": 1.676670419019998e-05, "loss": 0.8541, "step": 448500 }, { "epoch": 0.16, "learning_rate": 1.67630996240798e-05, "loss": 0.8458, "step": 449000 }, { "epoch": 0.16, "learning_rate": 1.6759495057959623e-05, "loss": 0.839, "step": 449500 }, { "epoch": 0.16, "learning_rate": 1.6755890491839445e-05, "loss": 0.8442, "step": 450000 }, { "epoch": 0.16, "learning_rate": 1.6752285925719263e-05, "loss": 0.8445, "step": 450500 }, { "epoch": 0.16, "learning_rate": 1.6748681359599085e-05, "loss": 0.8194, "step": 451000 }, { "epoch": 0.16, "learning_rate": 1.674507679347891e-05, "loss": 0.8173, "step": 451500 }, { "epoch": 0.16, "learning_rate": 1.674147222735873e-05, "loss": 0.8153, "step": 452000 }, { "epoch": 0.16, "learning_rate": 1.673786766123855e-05, "loss": 0.8301, "step": 452500 }, { "epoch": 0.16, "learning_rate": 1.6734263095118373e-05, "loss": 0.8245, "step": 453000 }, { "epoch": 0.16, "learning_rate": 1.6730658528998195e-05, "loss": 0.7968, "step": 453500 }, { "epoch": 0.16, "learning_rate": 1.6727053962878017e-05, "loss": 0.8291, "step": 454000 }, { "epoch": 0.16, "learning_rate": 1.672344939675784e-05, "loss": 0.8315, "step": 454500 }, { "epoch": 0.16, "learning_rate": 1.6719844830637658e-05, "loss": 0.824, "step": 455000 }, { "epoch": 0.16, "learning_rate": 1.6716240264517483e-05, "loss": 0.8212, "step": 455500 }, { "epoch": 0.16, "learning_rate": 1.6712635698397302e-05, "loss": 0.8121, "step": 456000 }, { "epoch": 0.16, "learning_rate": 1.6709031132277124e-05, "loss": 0.818, "step": 456500 }, { "epoch": 0.16, "learning_rate": 1.6705426566156946e-05, "loss": 0.8214, "step": 457000 }, { "epoch": 0.16, "learning_rate": 1.6701822000036768e-05, "loss": 0.8125, "step": 457500 }, { "epoch": 0.17, "learning_rate": 1.669821743391659e-05, "loss": 0.8522, "step": 458000 }, { "epoch": 0.17, "learning_rate": 1.6694612867796412e-05, "loss": 0.8398, "step": 458500 }, { "epoch": 0.17, "learning_rate": 1.6691008301676234e-05, "loss": 0.828, "step": 459000 }, { "epoch": 0.17, "learning_rate": 1.6687403735556053e-05, "loss": 0.8322, "step": 459500 }, { "epoch": 0.17, "learning_rate": 1.6683799169435878e-05, "loss": 0.8143, "step": 460000 }, { "epoch": 0.17, "learning_rate": 1.6680194603315697e-05, "loss": 0.8216, "step": 460500 }, { "epoch": 0.17, "learning_rate": 1.667659003719552e-05, "loss": 0.8578, "step": 461000 }, { "epoch": 0.17, "learning_rate": 1.667298547107534e-05, "loss": 0.8185, "step": 461500 }, { "epoch": 0.17, "learning_rate": 1.6669380904955163e-05, "loss": 0.8237, "step": 462000 }, { "epoch": 0.17, "learning_rate": 1.6665776338834985e-05, "loss": 0.8247, "step": 462500 }, { "epoch": 0.17, "learning_rate": 1.6662171772714807e-05, "loss": 0.8137, "step": 463000 }, { "epoch": 0.17, "learning_rate": 1.6658567206594625e-05, "loss": 0.8342, "step": 463500 }, { "epoch": 0.17, "learning_rate": 1.6654962640474448e-05, "loss": 0.8375, "step": 464000 }, { "epoch": 0.17, "learning_rate": 1.6651358074354273e-05, "loss": 0.8265, "step": 464500 }, { "epoch": 0.17, "learning_rate": 1.664775350823409e-05, "loss": 0.8138, "step": 465000 }, { "epoch": 0.17, "learning_rate": 1.6644148942113914e-05, "loss": 0.8633, "step": 465500 }, { "epoch": 0.17, "learning_rate": 1.6640544375993736e-05, "loss": 0.8314, "step": 466000 }, { "epoch": 0.17, "learning_rate": 1.6636939809873554e-05, "loss": 0.8198, "step": 466500 }, { "epoch": 0.17, "learning_rate": 1.663333524375338e-05, "loss": 0.8514, "step": 467000 }, { "epoch": 0.17, "learning_rate": 1.66297306776332e-05, "loss": 0.8145, "step": 467500 }, { "epoch": 0.17, "learning_rate": 1.662612611151302e-05, "loss": 0.8432, "step": 468000 }, { "epoch": 0.17, "learning_rate": 1.6622521545392842e-05, "loss": 0.8487, "step": 468500 }, { "epoch": 0.17, "learning_rate": 1.6618916979272664e-05, "loss": 0.8325, "step": 469000 }, { "epoch": 0.17, "learning_rate": 1.6615312413152486e-05, "loss": 0.8448, "step": 469500 }, { "epoch": 0.17, "learning_rate": 1.6611707847032308e-05, "loss": 0.8167, "step": 470000 }, { "epoch": 0.17, "learning_rate": 1.660810328091213e-05, "loss": 0.7967, "step": 470500 }, { "epoch": 0.17, "learning_rate": 1.6604498714791952e-05, "loss": 0.8447, "step": 471000 }, { "epoch": 0.17, "learning_rate": 1.6600894148671774e-05, "loss": 0.8472, "step": 471500 }, { "epoch": 0.17, "learning_rate": 1.6597289582551593e-05, "loss": 0.8269, "step": 472000 }, { "epoch": 0.17, "learning_rate": 1.6593685016431415e-05, "loss": 0.8309, "step": 472500 }, { "epoch": 0.17, "learning_rate": 1.6590080450311237e-05, "loss": 0.8048, "step": 473000 }, { "epoch": 0.17, "learning_rate": 1.658647588419106e-05, "loss": 0.838, "step": 473500 }, { "epoch": 0.17, "learning_rate": 1.658287131807088e-05, "loss": 0.8115, "step": 474000 }, { "epoch": 0.17, "learning_rate": 1.6579266751950703e-05, "loss": 0.7944, "step": 474500 }, { "epoch": 0.17, "learning_rate": 1.6575662185830525e-05, "loss": 0.8434, "step": 475000 }, { "epoch": 0.17, "learning_rate": 1.6572057619710347e-05, "loss": 0.8277, "step": 475500 }, { "epoch": 0.17, "learning_rate": 1.656845305359017e-05, "loss": 0.8102, "step": 476000 }, { "epoch": 0.17, "learning_rate": 1.6564848487469988e-05, "loss": 0.8206, "step": 476500 }, { "epoch": 0.17, "learning_rate": 1.656124392134981e-05, "loss": 0.8393, "step": 477000 }, { "epoch": 0.17, "learning_rate": 1.6557639355229632e-05, "loss": 0.7974, "step": 477500 }, { "epoch": 0.17, "learning_rate": 1.6554034789109454e-05, "loss": 0.8272, "step": 478000 }, { "epoch": 0.17, "learning_rate": 1.6550430222989276e-05, "loss": 0.8272, "step": 478500 }, { "epoch": 0.17, "learning_rate": 1.6546825656869098e-05, "loss": 0.8484, "step": 479000 }, { "epoch": 0.17, "learning_rate": 1.6543221090748916e-05, "loss": 0.838, "step": 479500 }, { "epoch": 0.17, "learning_rate": 1.6539616524628742e-05, "loss": 0.8246, "step": 480000 }, { "epoch": 0.17, "learning_rate": 1.6536011958508564e-05, "loss": 0.8294, "step": 480500 }, { "epoch": 0.17, "learning_rate": 1.6532407392388382e-05, "loss": 0.8343, "step": 481000 }, { "epoch": 0.17, "learning_rate": 1.6528802826268204e-05, "loss": 0.8582, "step": 481500 }, { "epoch": 0.17, "learning_rate": 1.6525198260148026e-05, "loss": 0.8248, "step": 482000 }, { "epoch": 0.17, "learning_rate": 1.652159369402785e-05, "loss": 0.8107, "step": 482500 }, { "epoch": 0.17, "learning_rate": 1.651798912790767e-05, "loss": 0.8361, "step": 483000 }, { "epoch": 0.17, "learning_rate": 1.6514384561787492e-05, "loss": 0.7743, "step": 483500 }, { "epoch": 0.17, "learning_rate": 1.651077999566731e-05, "loss": 0.8369, "step": 484000 }, { "epoch": 0.17, "learning_rate": 1.6507175429547137e-05, "loss": 0.8106, "step": 484500 }, { "epoch": 0.17, "learning_rate": 1.6503570863426955e-05, "loss": 0.8117, "step": 485000 }, { "epoch": 0.18, "learning_rate": 1.6499966297306777e-05, "loss": 0.8373, "step": 485500 }, { "epoch": 0.18, "learning_rate": 1.64963617311866e-05, "loss": 0.8175, "step": 486000 }, { "epoch": 0.18, "learning_rate": 1.649275716506642e-05, "loss": 0.7979, "step": 486500 }, { "epoch": 0.18, "learning_rate": 1.6489152598946243e-05, "loss": 0.839, "step": 487000 }, { "epoch": 0.18, "learning_rate": 1.6485548032826065e-05, "loss": 0.7952, "step": 487500 }, { "epoch": 0.18, "learning_rate": 1.6481943466705884e-05, "loss": 0.8172, "step": 488000 }, { "epoch": 0.18, "learning_rate": 1.6478338900585706e-05, "loss": 0.8145, "step": 488500 }, { "epoch": 0.18, "learning_rate": 1.647473433446553e-05, "loss": 0.8578, "step": 489000 }, { "epoch": 0.18, "learning_rate": 1.647112976834535e-05, "loss": 0.869, "step": 489500 }, { "epoch": 0.18, "learning_rate": 1.6467525202225172e-05, "loss": 0.8078, "step": 490000 }, { "epoch": 0.18, "learning_rate": 1.6463920636104994e-05, "loss": 0.8108, "step": 490500 }, { "epoch": 0.18, "learning_rate": 1.6460316069984816e-05, "loss": 0.7907, "step": 491000 }, { "epoch": 0.18, "learning_rate": 1.6456711503864638e-05, "loss": 0.7979, "step": 491500 }, { "epoch": 0.18, "learning_rate": 1.645310693774446e-05, "loss": 0.8243, "step": 492000 }, { "epoch": 0.18, "learning_rate": 1.644950237162428e-05, "loss": 0.8403, "step": 492500 }, { "epoch": 0.18, "learning_rate": 1.64458978055041e-05, "loss": 0.8106, "step": 493000 }, { "epoch": 0.18, "learning_rate": 1.6442293239383923e-05, "loss": 0.821, "step": 493500 }, { "epoch": 0.18, "learning_rate": 1.6438688673263745e-05, "loss": 0.8065, "step": 494000 }, { "epoch": 0.18, "learning_rate": 1.6435084107143567e-05, "loss": 0.801, "step": 494500 }, { "epoch": 0.18, "learning_rate": 1.643147954102339e-05, "loss": 0.814, "step": 495000 }, { "epoch": 0.18, "learning_rate": 1.642787497490321e-05, "loss": 0.8261, "step": 495500 }, { "epoch": 0.18, "learning_rate": 1.6424270408783033e-05, "loss": 0.8087, "step": 496000 }, { "epoch": 0.18, "learning_rate": 1.642066584266285e-05, "loss": 0.8306, "step": 496500 }, { "epoch": 0.18, "learning_rate": 1.6417061276542673e-05, "loss": 0.8147, "step": 497000 }, { "epoch": 0.18, "learning_rate": 1.6413456710422495e-05, "loss": 0.8273, "step": 497500 }, { "epoch": 0.18, "learning_rate": 1.6409852144302317e-05, "loss": 0.8257, "step": 498000 }, { "epoch": 0.18, "learning_rate": 1.640624757818214e-05, "loss": 0.8128, "step": 498500 }, { "epoch": 0.18, "learning_rate": 1.640264301206196e-05, "loss": 0.8177, "step": 499000 }, { "epoch": 0.18, "learning_rate": 1.6399038445941783e-05, "loss": 0.8372, "step": 499500 }, { "epoch": 0.18, "learning_rate": 1.6395433879821605e-05, "loss": 0.8068, "step": 500000 }, { "epoch": 0.18, "learning_rate": 1.6391829313701427e-05, "loss": 0.8158, "step": 500500 }, { "epoch": 0.18, "learning_rate": 1.6388224747581246e-05, "loss": 0.8197, "step": 501000 }, { "epoch": 0.18, "learning_rate": 1.6384620181461068e-05, "loss": 0.8129, "step": 501500 }, { "epoch": 0.18, "learning_rate": 1.638101561534089e-05, "loss": 0.8105, "step": 502000 }, { "epoch": 0.18, "learning_rate": 1.6377411049220712e-05, "loss": 0.792, "step": 502500 }, { "epoch": 0.18, "learning_rate": 1.6373806483100534e-05, "loss": 0.8464, "step": 503000 }, { "epoch": 0.18, "learning_rate": 1.6370201916980356e-05, "loss": 0.798, "step": 503500 }, { "epoch": 0.18, "learning_rate": 1.6366597350860175e-05, "loss": 0.8327, "step": 504000 }, { "epoch": 0.18, "learning_rate": 1.636299278474e-05, "loss": 0.7981, "step": 504500 }, { "epoch": 0.18, "learning_rate": 1.6359388218619822e-05, "loss": 0.7929, "step": 505000 }, { "epoch": 0.18, "learning_rate": 1.635578365249964e-05, "loss": 0.8582, "step": 505500 }, { "epoch": 0.18, "learning_rate": 1.6352179086379463e-05, "loss": 0.8021, "step": 506000 }, { "epoch": 0.18, "learning_rate": 1.6348574520259285e-05, "loss": 0.8072, "step": 506500 }, { "epoch": 0.18, "learning_rate": 1.6344969954139107e-05, "loss": 0.8347, "step": 507000 }, { "epoch": 0.18, "learning_rate": 1.634136538801893e-05, "loss": 0.7993, "step": 507500 }, { "epoch": 0.18, "learning_rate": 1.633776082189875e-05, "loss": 0.8307, "step": 508000 }, { "epoch": 0.18, "learning_rate": 1.633415625577857e-05, "loss": 0.8217, "step": 508500 }, { "epoch": 0.18, "learning_rate": 1.6330551689658395e-05, "loss": 0.8294, "step": 509000 }, { "epoch": 0.18, "learning_rate": 1.6326947123538213e-05, "loss": 0.7999, "step": 509500 }, { "epoch": 0.18, "learning_rate": 1.6323342557418035e-05, "loss": 0.792, "step": 510000 }, { "epoch": 0.18, "learning_rate": 1.6319737991297858e-05, "loss": 0.8127, "step": 510500 }, { "epoch": 0.18, "learning_rate": 1.631613342517768e-05, "loss": 0.8061, "step": 511000 }, { "epoch": 0.18, "learning_rate": 1.63125288590575e-05, "loss": 0.809, "step": 511500 }, { "epoch": 0.18, "learning_rate": 1.6308924292937324e-05, "loss": 0.8369, "step": 512000 }, { "epoch": 0.18, "learning_rate": 1.6305319726817142e-05, "loss": 0.8213, "step": 512500 }, { "epoch": 0.18, "learning_rate": 1.6301715160696964e-05, "loss": 0.816, "step": 513000 }, { "epoch": 0.19, "learning_rate": 1.629811059457679e-05, "loss": 0.7976, "step": 513500 }, { "epoch": 0.19, "learning_rate": 1.6294506028456608e-05, "loss": 0.8079, "step": 514000 }, { "epoch": 0.19, "learning_rate": 1.629090146233643e-05, "loss": 0.8362, "step": 514500 }, { "epoch": 0.19, "learning_rate": 1.6287296896216252e-05, "loss": 0.8466, "step": 515000 }, { "epoch": 0.19, "learning_rate": 1.6283692330096074e-05, "loss": 0.8382, "step": 515500 }, { "epoch": 0.19, "learning_rate": 1.6280087763975896e-05, "loss": 0.7879, "step": 516000 }, { "epoch": 0.19, "learning_rate": 1.6276483197855718e-05, "loss": 0.7871, "step": 516500 }, { "epoch": 0.19, "learning_rate": 1.6272878631735537e-05, "loss": 0.7969, "step": 517000 }, { "epoch": 0.19, "learning_rate": 1.626927406561536e-05, "loss": 0.7951, "step": 517500 }, { "epoch": 0.19, "learning_rate": 1.626566949949518e-05, "loss": 0.8134, "step": 518000 }, { "epoch": 0.19, "learning_rate": 1.6262064933375003e-05, "loss": 0.8543, "step": 518500 }, { "epoch": 0.19, "learning_rate": 1.6258460367254825e-05, "loss": 0.7851, "step": 519000 }, { "epoch": 0.19, "learning_rate": 1.6254855801134647e-05, "loss": 0.8184, "step": 519500 }, { "epoch": 0.19, "learning_rate": 1.625125123501447e-05, "loss": 0.7852, "step": 520000 }, { "epoch": 0.19, "learning_rate": 1.624764666889429e-05, "loss": 0.8136, "step": 520500 }, { "epoch": 0.19, "learning_rate": 1.624404210277411e-05, "loss": 0.7861, "step": 521000 }, { "epoch": 0.19, "learning_rate": 1.624043753665393e-05, "loss": 0.7762, "step": 521500 }, { "epoch": 0.19, "learning_rate": 1.6236832970533757e-05, "loss": 0.7982, "step": 522000 }, { "epoch": 0.19, "learning_rate": 1.6233228404413576e-05, "loss": 0.8048, "step": 522500 }, { "epoch": 0.19, "learning_rate": 1.6229623838293398e-05, "loss": 0.8208, "step": 523000 }, { "epoch": 0.19, "learning_rate": 1.622601927217322e-05, "loss": 0.7868, "step": 523500 }, { "epoch": 0.19, "learning_rate": 1.6222414706053042e-05, "loss": 0.8089, "step": 524000 }, { "epoch": 0.19, "learning_rate": 1.6218810139932864e-05, "loss": 0.7938, "step": 524500 }, { "epoch": 0.19, "learning_rate": 1.6215205573812686e-05, "loss": 0.7907, "step": 525000 }, { "epoch": 0.19, "learning_rate": 1.6211601007692504e-05, "loss": 0.8018, "step": 525500 }, { "epoch": 0.19, "learning_rate": 1.6207996441572326e-05, "loss": 0.828, "step": 526000 }, { "epoch": 0.19, "learning_rate": 1.6204391875452152e-05, "loss": 0.7714, "step": 526500 }, { "epoch": 0.19, "learning_rate": 1.620078730933197e-05, "loss": 0.8177, "step": 527000 }, { "epoch": 0.19, "learning_rate": 1.6197182743211792e-05, "loss": 0.8105, "step": 527500 }, { "epoch": 0.19, "learning_rate": 1.6193578177091614e-05, "loss": 0.8341, "step": 528000 }, { "epoch": 0.19, "learning_rate": 1.6189973610971433e-05, "loss": 0.7953, "step": 528500 }, { "epoch": 0.19, "learning_rate": 1.618636904485126e-05, "loss": 0.808, "step": 529000 }, { "epoch": 0.19, "learning_rate": 1.618276447873108e-05, "loss": 0.7921, "step": 529500 }, { "epoch": 0.19, "learning_rate": 1.61791599126109e-05, "loss": 0.7788, "step": 530000 }, { "epoch": 0.19, "learning_rate": 1.617555534649072e-05, "loss": 0.814, "step": 530500 }, { "epoch": 0.19, "learning_rate": 1.6171950780370543e-05, "loss": 0.8059, "step": 531000 }, { "epoch": 0.19, "learning_rate": 1.6168346214250365e-05, "loss": 0.7975, "step": 531500 }, { "epoch": 0.19, "learning_rate": 1.6164741648130187e-05, "loss": 0.808, "step": 532000 }, { "epoch": 0.19, "learning_rate": 1.616113708201001e-05, "loss": 0.7954, "step": 532500 }, { "epoch": 0.19, "learning_rate": 1.6157532515889828e-05, "loss": 0.7833, "step": 533000 }, { "epoch": 0.19, "learning_rate": 1.6153927949769653e-05, "loss": 0.8034, "step": 533500 }, { "epoch": 0.19, "learning_rate": 1.6150323383649472e-05, "loss": 0.7741, "step": 534000 }, { "epoch": 0.19, "learning_rate": 1.6146718817529294e-05, "loss": 0.8258, "step": 534500 }, { "epoch": 0.19, "learning_rate": 1.6143114251409116e-05, "loss": 0.798, "step": 535000 }, { "epoch": 0.19, "learning_rate": 1.6139509685288938e-05, "loss": 0.7879, "step": 535500 }, { "epoch": 0.19, "learning_rate": 1.613590511916876e-05, "loss": 0.7725, "step": 536000 }, { "epoch": 0.19, "learning_rate": 1.6132300553048582e-05, "loss": 0.7977, "step": 536500 }, { "epoch": 0.19, "learning_rate": 1.61286959869284e-05, "loss": 0.8393, "step": 537000 }, { "epoch": 0.19, "learning_rate": 1.6125091420808226e-05, "loss": 0.8031, "step": 537500 }, { "epoch": 0.19, "learning_rate": 1.6121486854688048e-05, "loss": 0.8348, "step": 538000 }, { "epoch": 0.19, "learning_rate": 1.6117882288567867e-05, "loss": 0.7989, "step": 538500 }, { "epoch": 0.19, "learning_rate": 1.611427772244769e-05, "loss": 0.7828, "step": 539000 }, { "epoch": 0.19, "learning_rate": 1.611067315632751e-05, "loss": 0.808, "step": 539500 }, { "epoch": 0.19, "learning_rate": 1.6107068590207333e-05, "loss": 0.7897, "step": 540000 }, { "epoch": 0.19, "learning_rate": 1.6103464024087155e-05, "loss": 0.8267, "step": 540500 }, { "epoch": 0.2, "learning_rate": 1.6099859457966977e-05, "loss": 0.831, "step": 541000 }, { "epoch": 0.2, "learning_rate": 1.6096254891846795e-05, "loss": 0.8017, "step": 541500 }, { "epoch": 0.2, "learning_rate": 1.609265032572662e-05, "loss": 0.8001, "step": 542000 }, { "epoch": 0.2, "learning_rate": 1.608904575960644e-05, "loss": 0.8096, "step": 542500 }, { "epoch": 0.2, "learning_rate": 1.608544119348626e-05, "loss": 0.8057, "step": 543000 }, { "epoch": 0.2, "learning_rate": 1.6081836627366083e-05, "loss": 0.8049, "step": 543500 }, { "epoch": 0.2, "learning_rate": 1.6078232061245905e-05, "loss": 0.835, "step": 544000 }, { "epoch": 0.2, "learning_rate": 1.6074627495125727e-05, "loss": 0.8109, "step": 544500 }, { "epoch": 0.2, "learning_rate": 1.607102292900555e-05, "loss": 0.8017, "step": 545000 }, { "epoch": 0.2, "learning_rate": 1.606741836288537e-05, "loss": 0.7894, "step": 545500 }, { "epoch": 0.2, "learning_rate": 1.606381379676519e-05, "loss": 0.8116, "step": 546000 }, { "epoch": 0.2, "learning_rate": 1.6060209230645015e-05, "loss": 0.7944, "step": 546500 }, { "epoch": 0.2, "learning_rate": 1.6056604664524834e-05, "loss": 0.7557, "step": 547000 }, { "epoch": 0.2, "learning_rate": 1.6053000098404656e-05, "loss": 0.7778, "step": 547500 }, { "epoch": 0.2, "learning_rate": 1.6049395532284478e-05, "loss": 0.8102, "step": 548000 }, { "epoch": 0.2, "learning_rate": 1.60457909661643e-05, "loss": 0.7879, "step": 548500 }, { "epoch": 0.2, "learning_rate": 1.6042186400044122e-05, "loss": 0.835, "step": 549000 }, { "epoch": 0.2, "learning_rate": 1.6038581833923944e-05, "loss": 0.8003, "step": 549500 }, { "epoch": 0.2, "learning_rate": 1.6034977267803763e-05, "loss": 0.8165, "step": 550000 }, { "epoch": 0.2, "learning_rate": 1.6031372701683585e-05, "loss": 0.7717, "step": 550500 }, { "epoch": 0.2, "learning_rate": 1.602776813556341e-05, "loss": 0.8118, "step": 551000 }, { "epoch": 0.2, "learning_rate": 1.602416356944323e-05, "loss": 0.7578, "step": 551500 }, { "epoch": 0.2, "learning_rate": 1.602055900332305e-05, "loss": 0.8173, "step": 552000 }, { "epoch": 0.2, "learning_rate": 1.6016954437202873e-05, "loss": 0.8048, "step": 552500 }, { "epoch": 0.2, "learning_rate": 1.6013349871082695e-05, "loss": 0.8139, "step": 553000 }, { "epoch": 0.2, "learning_rate": 1.6009745304962517e-05, "loss": 0.809, "step": 553500 }, { "epoch": 0.2, "learning_rate": 1.600614073884234e-05, "loss": 0.7888, "step": 554000 }, { "epoch": 0.2, "learning_rate": 1.6002536172722157e-05, "loss": 0.8225, "step": 554500 }, { "epoch": 0.2, "learning_rate": 1.599893160660198e-05, "loss": 0.7574, "step": 555000 }, { "epoch": 0.2, "learning_rate": 1.59953270404818e-05, "loss": 0.7699, "step": 555500 }, { "epoch": 0.2, "learning_rate": 1.5991722474361623e-05, "loss": 0.7778, "step": 556000 }, { "epoch": 0.2, "learning_rate": 1.5988117908241445e-05, "loss": 0.7988, "step": 556500 }, { "epoch": 0.2, "learning_rate": 1.5984513342121268e-05, "loss": 0.8014, "step": 557000 }, { "epoch": 0.2, "learning_rate": 1.598090877600109e-05, "loss": 0.8162, "step": 557500 }, { "epoch": 0.2, "learning_rate": 1.597730420988091e-05, "loss": 0.813, "step": 558000 }, { "epoch": 0.2, "learning_rate": 1.597369964376073e-05, "loss": 0.7831, "step": 558500 }, { "epoch": 0.2, "learning_rate": 1.5970095077640552e-05, "loss": 0.8056, "step": 559000 }, { "epoch": 0.2, "learning_rate": 1.5966490511520374e-05, "loss": 0.7801, "step": 559500 }, { "epoch": 0.2, "learning_rate": 1.5962885945400196e-05, "loss": 0.824, "step": 560000 }, { "epoch": 0.2, "learning_rate": 1.5959281379280018e-05, "loss": 0.8062, "step": 560500 }, { "epoch": 0.2, "learning_rate": 1.595567681315984e-05, "loss": 0.8037, "step": 561000 }, { "epoch": 0.2, "learning_rate": 1.595207224703966e-05, "loss": 0.8349, "step": 561500 }, { "epoch": 0.2, "learning_rate": 1.5948467680919484e-05, "loss": 0.8074, "step": 562000 }, { "epoch": 0.2, "learning_rate": 1.5944863114799306e-05, "loss": 0.827, "step": 562500 }, { "epoch": 0.2, "learning_rate": 1.5941258548679125e-05, "loss": 0.809, "step": 563000 }, { "epoch": 0.2, "learning_rate": 1.5937653982558947e-05, "loss": 0.7921, "step": 563500 }, { "epoch": 0.2, "learning_rate": 1.593404941643877e-05, "loss": 0.8016, "step": 564000 }, { "epoch": 0.2, "learning_rate": 1.593044485031859e-05, "loss": 0.7946, "step": 564500 }, { "epoch": 0.2, "learning_rate": 1.5926840284198413e-05, "loss": 0.8349, "step": 565000 }, { "epoch": 0.2, "learning_rate": 1.5923235718078235e-05, "loss": 0.8151, "step": 565500 }, { "epoch": 0.2, "learning_rate": 1.5919631151958054e-05, "loss": 0.804, "step": 566000 }, { "epoch": 0.2, "learning_rate": 1.591602658583788e-05, "loss": 0.7976, "step": 566500 }, { "epoch": 0.2, "learning_rate": 1.5912422019717698e-05, "loss": 0.8289, "step": 567000 }, { "epoch": 0.2, "learning_rate": 1.590881745359752e-05, "loss": 0.7996, "step": 567500 }, { "epoch": 0.2, "learning_rate": 1.590521288747734e-05, "loss": 0.8218, "step": 568000 }, { "epoch": 0.2, "learning_rate": 1.5901608321357164e-05, "loss": 0.7803, "step": 568500 }, { "epoch": 0.21, "learning_rate": 1.5898003755236986e-05, "loss": 0.8096, "step": 569000 }, { "epoch": 0.21, "learning_rate": 1.5894399189116808e-05, "loss": 0.8027, "step": 569500 }, { "epoch": 0.21, "learning_rate": 1.589079462299663e-05, "loss": 0.785, "step": 570000 }, { "epoch": 0.21, "learning_rate": 1.588719005687645e-05, "loss": 0.8006, "step": 570500 }, { "epoch": 0.21, "learning_rate": 1.5883585490756274e-05, "loss": 0.8056, "step": 571000 }, { "epoch": 0.21, "learning_rate": 1.5879980924636092e-05, "loss": 0.8416, "step": 571500 }, { "epoch": 0.21, "learning_rate": 1.5876376358515914e-05, "loss": 0.7975, "step": 572000 }, { "epoch": 0.21, "learning_rate": 1.5872771792395736e-05, "loss": 0.8156, "step": 572500 }, { "epoch": 0.21, "learning_rate": 1.586916722627556e-05, "loss": 0.7656, "step": 573000 }, { "epoch": 0.21, "learning_rate": 1.586556266015538e-05, "loss": 0.7891, "step": 573500 }, { "epoch": 0.21, "learning_rate": 1.5861958094035202e-05, "loss": 0.8096, "step": 574000 }, { "epoch": 0.21, "learning_rate": 1.585835352791502e-05, "loss": 0.7851, "step": 574500 }, { "epoch": 0.21, "learning_rate": 1.5854748961794843e-05, "loss": 0.8006, "step": 575000 }, { "epoch": 0.21, "learning_rate": 1.585114439567467e-05, "loss": 0.8082, "step": 575500 }, { "epoch": 0.21, "learning_rate": 1.5847539829554487e-05, "loss": 0.816, "step": 576000 }, { "epoch": 0.21, "learning_rate": 1.584393526343431e-05, "loss": 0.7686, "step": 576500 }, { "epoch": 0.21, "learning_rate": 1.584033069731413e-05, "loss": 0.8147, "step": 577000 }, { "epoch": 0.21, "learning_rate": 1.5836726131193953e-05, "loss": 0.7934, "step": 577500 }, { "epoch": 0.21, "learning_rate": 1.5833121565073775e-05, "loss": 0.7682, "step": 578000 }, { "epoch": 0.21, "learning_rate": 1.5829516998953597e-05, "loss": 0.7743, "step": 578500 }, { "epoch": 0.21, "learning_rate": 1.5825912432833416e-05, "loss": 0.7898, "step": 579000 }, { "epoch": 0.21, "learning_rate": 1.5822307866713238e-05, "loss": 0.7829, "step": 579500 }, { "epoch": 0.21, "learning_rate": 1.581870330059306e-05, "loss": 0.8132, "step": 580000 }, { "epoch": 0.21, "learning_rate": 1.5815098734472882e-05, "loss": 0.809, "step": 580500 }, { "epoch": 0.21, "learning_rate": 1.5811494168352704e-05, "loss": 0.7921, "step": 581000 }, { "epoch": 0.21, "learning_rate": 1.5807889602232526e-05, "loss": 0.7898, "step": 581500 }, { "epoch": 0.21, "learning_rate": 1.5804285036112348e-05, "loss": 0.7807, "step": 582000 }, { "epoch": 0.21, "learning_rate": 1.580068046999217e-05, "loss": 0.7815, "step": 582500 }, { "epoch": 0.21, "learning_rate": 1.579707590387199e-05, "loss": 0.8083, "step": 583000 }, { "epoch": 0.21, "learning_rate": 1.579347133775181e-05, "loss": 0.7837, "step": 583500 }, { "epoch": 0.21, "learning_rate": 1.5789866771631633e-05, "loss": 0.7893, "step": 584000 }, { "epoch": 0.21, "learning_rate": 1.5786262205511455e-05, "loss": 0.7862, "step": 584500 }, { "epoch": 0.21, "learning_rate": 1.5782657639391277e-05, "loss": 0.7853, "step": 585000 }, { "epoch": 0.21, "learning_rate": 1.57790530732711e-05, "loss": 0.7872, "step": 585500 }, { "epoch": 0.21, "learning_rate": 1.577544850715092e-05, "loss": 0.7943, "step": 586000 }, { "epoch": 0.21, "learning_rate": 1.5771843941030743e-05, "loss": 0.7858, "step": 586500 }, { "epoch": 0.21, "learning_rate": 1.5768239374910565e-05, "loss": 0.7947, "step": 587000 }, { "epoch": 0.21, "learning_rate": 1.5764634808790383e-05, "loss": 0.7756, "step": 587500 }, { "epoch": 0.21, "learning_rate": 1.5761030242670205e-05, "loss": 0.8207, "step": 588000 }, { "epoch": 0.21, "learning_rate": 1.5757425676550027e-05, "loss": 0.7847, "step": 588500 }, { "epoch": 0.21, "learning_rate": 1.575382111042985e-05, "loss": 0.8185, "step": 589000 }, { "epoch": 0.21, "learning_rate": 1.575021654430967e-05, "loss": 0.7883, "step": 589500 }, { "epoch": 0.21, "learning_rate": 1.5746611978189493e-05, "loss": 0.7879, "step": 590000 }, { "epoch": 0.21, "learning_rate": 1.5743007412069312e-05, "loss": 0.7867, "step": 590500 }, { "epoch": 0.21, "learning_rate": 1.5739402845949137e-05, "loss": 0.7934, "step": 591000 }, { "epoch": 0.21, "learning_rate": 1.573579827982896e-05, "loss": 0.7947, "step": 591500 }, { "epoch": 0.21, "learning_rate": 1.5732193713708778e-05, "loss": 0.7741, "step": 592000 }, { "epoch": 0.21, "learning_rate": 1.57285891475886e-05, "loss": 0.8238, "step": 592500 }, { "epoch": 0.21, "learning_rate": 1.5724984581468422e-05, "loss": 0.8144, "step": 593000 }, { "epoch": 0.21, "learning_rate": 1.5721380015348244e-05, "loss": 0.8018, "step": 593500 }, { "epoch": 0.21, "learning_rate": 1.5717775449228066e-05, "loss": 0.8043, "step": 594000 }, { "epoch": 0.21, "learning_rate": 1.5714170883107888e-05, "loss": 0.7818, "step": 594500 }, { "epoch": 0.21, "learning_rate": 1.5710566316987707e-05, "loss": 0.7822, "step": 595000 }, { "epoch": 0.21, "learning_rate": 1.5706961750867532e-05, "loss": 0.7781, "step": 595500 }, { "epoch": 0.21, "learning_rate": 1.570335718474735e-05, "loss": 0.8182, "step": 596000 }, { "epoch": 0.22, "learning_rate": 1.5699752618627173e-05, "loss": 0.8074, "step": 596500 }, { "epoch": 0.22, "learning_rate": 1.5696148052506995e-05, "loss": 0.7768, "step": 597000 }, { "epoch": 0.22, "learning_rate": 1.5692543486386817e-05, "loss": 0.7908, "step": 597500 }, { "epoch": 0.22, "learning_rate": 1.568893892026664e-05, "loss": 0.7943, "step": 598000 }, { "epoch": 0.22, "learning_rate": 1.568533435414646e-05, "loss": 0.7952, "step": 598500 }, { "epoch": 0.22, "learning_rate": 1.568172978802628e-05, "loss": 0.7784, "step": 599000 }, { "epoch": 0.22, "learning_rate": 1.56781252219061e-05, "loss": 0.7871, "step": 599500 }, { "epoch": 0.22, "learning_rate": 1.5674520655785927e-05, "loss": 0.7677, "step": 600000 }, { "epoch": 0.22, "learning_rate": 1.5670916089665745e-05, "loss": 0.7839, "step": 600500 }, { "epoch": 0.22, "learning_rate": 1.5667311523545567e-05, "loss": 0.7999, "step": 601000 }, { "epoch": 0.22, "learning_rate": 1.566370695742539e-05, "loss": 0.7964, "step": 601500 }, { "epoch": 0.22, "learning_rate": 1.566010239130521e-05, "loss": 0.7632, "step": 602000 }, { "epoch": 0.22, "learning_rate": 1.5656497825185033e-05, "loss": 0.812, "step": 602500 }, { "epoch": 0.22, "learning_rate": 1.5652893259064856e-05, "loss": 0.7873, "step": 603000 }, { "epoch": 0.22, "learning_rate": 1.5649288692944674e-05, "loss": 0.7887, "step": 603500 }, { "epoch": 0.22, "learning_rate": 1.56456841268245e-05, "loss": 0.7535, "step": 604000 }, { "epoch": 0.22, "learning_rate": 1.5642079560704318e-05, "loss": 0.7945, "step": 604500 }, { "epoch": 0.22, "learning_rate": 1.563847499458414e-05, "loss": 0.7731, "step": 605000 }, { "epoch": 0.22, "learning_rate": 1.5634870428463962e-05, "loss": 0.7729, "step": 605500 }, { "epoch": 0.22, "learning_rate": 1.5631265862343784e-05, "loss": 0.7821, "step": 606000 }, { "epoch": 0.22, "learning_rate": 1.5627661296223606e-05, "loss": 0.8183, "step": 606500 }, { "epoch": 0.22, "learning_rate": 1.5624056730103428e-05, "loss": 0.7777, "step": 607000 }, { "epoch": 0.22, "learning_rate": 1.5620452163983247e-05, "loss": 0.7857, "step": 607500 }, { "epoch": 0.22, "learning_rate": 1.561684759786307e-05, "loss": 0.7924, "step": 608000 }, { "epoch": 0.22, "learning_rate": 1.5613243031742894e-05, "loss": 0.7957, "step": 608500 }, { "epoch": 0.22, "learning_rate": 1.5609638465622713e-05, "loss": 0.7699, "step": 609000 }, { "epoch": 0.22, "learning_rate": 1.5606033899502535e-05, "loss": 0.8094, "step": 609500 }, { "epoch": 0.22, "learning_rate": 1.5602429333382357e-05, "loss": 0.7654, "step": 610000 }, { "epoch": 0.22, "learning_rate": 1.559882476726218e-05, "loss": 0.757, "step": 610500 }, { "epoch": 0.22, "learning_rate": 1.5595220201142e-05, "loss": 0.8075, "step": 611000 }, { "epoch": 0.22, "learning_rate": 1.5591615635021823e-05, "loss": 0.7922, "step": 611500 }, { "epoch": 0.22, "learning_rate": 1.558801106890164e-05, "loss": 0.7768, "step": 612000 }, { "epoch": 0.22, "learning_rate": 1.5584406502781464e-05, "loss": 0.7867, "step": 612500 }, { "epoch": 0.22, "learning_rate": 1.5580801936661286e-05, "loss": 0.7851, "step": 613000 }, { "epoch": 0.22, "learning_rate": 1.5577197370541108e-05, "loss": 0.7863, "step": 613500 }, { "epoch": 0.22, "learning_rate": 1.557359280442093e-05, "loss": 0.7925, "step": 614000 }, { "epoch": 0.22, "learning_rate": 1.556998823830075e-05, "loss": 0.7803, "step": 614500 }, { "epoch": 0.22, "learning_rate": 1.556638367218057e-05, "loss": 0.7969, "step": 615000 }, { "epoch": 0.22, "learning_rate": 1.5562779106060396e-05, "loss": 0.7938, "step": 615500 }, { "epoch": 0.22, "learning_rate": 1.5559174539940218e-05, "loss": 0.8077, "step": 616000 }, { "epoch": 0.22, "learning_rate": 1.5555569973820036e-05, "loss": 0.7865, "step": 616500 }, { "epoch": 0.22, "learning_rate": 1.555196540769986e-05, "loss": 0.7759, "step": 617000 }, { "epoch": 0.22, "learning_rate": 1.554836084157968e-05, "loss": 0.7848, "step": 617500 }, { "epoch": 0.22, "learning_rate": 1.5544756275459502e-05, "loss": 0.7812, "step": 618000 }, { "epoch": 0.22, "learning_rate": 1.5541151709339324e-05, "loss": 0.805, "step": 618500 }, { "epoch": 0.22, "learning_rate": 1.5537547143219146e-05, "loss": 0.8017, "step": 619000 }, { "epoch": 0.22, "learning_rate": 1.553394257709897e-05, "loss": 0.826, "step": 619500 }, { "epoch": 0.22, "learning_rate": 1.553033801097879e-05, "loss": 0.8026, "step": 620000 }, { "epoch": 0.22, "learning_rate": 1.552673344485861e-05, "loss": 0.7879, "step": 620500 }, { "epoch": 0.22, "learning_rate": 1.552312887873843e-05, "loss": 0.7912, "step": 621000 }, { "epoch": 0.22, "learning_rate": 1.5519524312618253e-05, "loss": 0.8041, "step": 621500 }, { "epoch": 0.22, "learning_rate": 1.5515919746498075e-05, "loss": 0.8189, "step": 622000 }, { "epoch": 0.22, "learning_rate": 1.5512315180377897e-05, "loss": 0.8019, "step": 622500 }, { "epoch": 0.22, "learning_rate": 1.550871061425772e-05, "loss": 0.7859, "step": 623000 }, { "epoch": 0.22, "learning_rate": 1.5505106048137538e-05, "loss": 0.7872, "step": 623500 }, { "epoch": 0.22, "learning_rate": 1.5501501482017363e-05, "loss": 0.7841, "step": 624000 }, { "epoch": 0.23, "learning_rate": 1.5497896915897185e-05, "loss": 0.7869, "step": 624500 }, { "epoch": 0.23, "learning_rate": 1.5494292349777004e-05, "loss": 0.8049, "step": 625000 }, { "epoch": 0.23, "learning_rate": 1.5490687783656826e-05, "loss": 0.7692, "step": 625500 }, { "epoch": 0.23, "learning_rate": 1.5487083217536648e-05, "loss": 0.7694, "step": 626000 }, { "epoch": 0.23, "learning_rate": 1.548347865141647e-05, "loss": 0.8004, "step": 626500 }, { "epoch": 0.23, "learning_rate": 1.5479874085296292e-05, "loss": 0.7541, "step": 627000 }, { "epoch": 0.23, "learning_rate": 1.5476269519176114e-05, "loss": 0.7692, "step": 627500 }, { "epoch": 0.23, "learning_rate": 1.5472664953055932e-05, "loss": 0.8097, "step": 628000 }, { "epoch": 0.23, "learning_rate": 1.5469060386935758e-05, "loss": 0.7746, "step": 628500 }, { "epoch": 0.23, "learning_rate": 1.5465455820815577e-05, "loss": 0.7739, "step": 629000 }, { "epoch": 0.23, "learning_rate": 1.54618512546954e-05, "loss": 0.7958, "step": 629500 }, { "epoch": 0.23, "learning_rate": 1.545824668857522e-05, "loss": 0.7675, "step": 630000 }, { "epoch": 0.23, "learning_rate": 1.5454642122455043e-05, "loss": 0.8152, "step": 630500 }, { "epoch": 0.23, "learning_rate": 1.5451037556334865e-05, "loss": 0.7815, "step": 631000 }, { "epoch": 0.23, "learning_rate": 1.5447432990214687e-05, "loss": 0.7928, "step": 631500 }, { "epoch": 0.23, "learning_rate": 1.5443828424094505e-05, "loss": 0.8195, "step": 632000 }, { "epoch": 0.23, "learning_rate": 1.5440223857974327e-05, "loss": 0.799, "step": 632500 }, { "epoch": 0.23, "learning_rate": 1.5436619291854153e-05, "loss": 0.7962, "step": 633000 }, { "epoch": 0.23, "learning_rate": 1.543301472573397e-05, "loss": 0.7644, "step": 633500 }, { "epoch": 0.23, "learning_rate": 1.5429410159613793e-05, "loss": 0.8092, "step": 634000 }, { "epoch": 0.23, "learning_rate": 1.5425805593493615e-05, "loss": 0.8012, "step": 634500 }, { "epoch": 0.23, "learning_rate": 1.5422201027373437e-05, "loss": 0.78, "step": 635000 }, { "epoch": 0.23, "learning_rate": 1.541859646125326e-05, "loss": 0.7796, "step": 635500 }, { "epoch": 0.23, "learning_rate": 1.541499189513308e-05, "loss": 0.8015, "step": 636000 }, { "epoch": 0.23, "learning_rate": 1.54113873290129e-05, "loss": 0.7934, "step": 636500 }, { "epoch": 0.23, "learning_rate": 1.5407782762892722e-05, "loss": 0.7879, "step": 637000 }, { "epoch": 0.23, "learning_rate": 1.5404178196772547e-05, "loss": 0.7477, "step": 637500 }, { "epoch": 0.23, "learning_rate": 1.5400573630652366e-05, "loss": 0.799, "step": 638000 }, { "epoch": 0.23, "learning_rate": 1.5396969064532188e-05, "loss": 0.7826, "step": 638500 }, { "epoch": 0.23, "learning_rate": 1.539336449841201e-05, "loss": 0.8188, "step": 639000 }, { "epoch": 0.23, "learning_rate": 1.5389759932291832e-05, "loss": 0.7773, "step": 639500 }, { "epoch": 0.23, "learning_rate": 1.5386155366171654e-05, "loss": 0.7729, "step": 640000 }, { "epoch": 0.23, "learning_rate": 1.5382550800051476e-05, "loss": 0.8174, "step": 640500 }, { "epoch": 0.23, "learning_rate": 1.5378946233931295e-05, "loss": 0.7705, "step": 641000 }, { "epoch": 0.23, "learning_rate": 1.5375341667811117e-05, "loss": 0.8065, "step": 641500 }, { "epoch": 0.23, "learning_rate": 1.537173710169094e-05, "loss": 0.7751, "step": 642000 }, { "epoch": 0.23, "learning_rate": 1.536813253557076e-05, "loss": 0.798, "step": 642500 }, { "epoch": 0.23, "learning_rate": 1.5364527969450583e-05, "loss": 0.7451, "step": 643000 }, { "epoch": 0.23, "learning_rate": 1.5360923403330405e-05, "loss": 0.7504, "step": 643500 }, { "epoch": 0.23, "learning_rate": 1.5357318837210227e-05, "loss": 0.7857, "step": 644000 }, { "epoch": 0.23, "learning_rate": 1.535371427109005e-05, "loss": 0.7971, "step": 644500 }, { "epoch": 0.23, "learning_rate": 1.5350109704969867e-05, "loss": 0.7967, "step": 645000 }, { "epoch": 0.23, "learning_rate": 1.534650513884969e-05, "loss": 0.8038, "step": 645500 }, { "epoch": 0.23, "learning_rate": 1.534290057272951e-05, "loss": 0.792, "step": 646000 }, { "epoch": 0.23, "learning_rate": 1.5339296006609333e-05, "loss": 0.7884, "step": 646500 }, { "epoch": 0.23, "learning_rate": 1.5335691440489155e-05, "loss": 0.7898, "step": 647000 }, { "epoch": 0.23, "learning_rate": 1.5332086874368977e-05, "loss": 0.7882, "step": 647500 }, { "epoch": 0.23, "learning_rate": 1.5328482308248796e-05, "loss": 0.7771, "step": 648000 }, { "epoch": 0.23, "learning_rate": 1.532487774212862e-05, "loss": 0.7961, "step": 648500 }, { "epoch": 0.23, "learning_rate": 1.5321273176008443e-05, "loss": 0.801, "step": 649000 }, { "epoch": 0.23, "learning_rate": 1.5317668609888262e-05, "loss": 0.7927, "step": 649500 }, { "epoch": 0.23, "learning_rate": 1.5314064043768084e-05, "loss": 0.7878, "step": 650000 }, { "epoch": 0.23, "learning_rate": 1.5310459477647906e-05, "loss": 0.8063, "step": 650500 }, { "epoch": 0.23, "learning_rate": 1.5306854911527728e-05, "loss": 0.7905, "step": 651000 }, { "epoch": 0.23, "learning_rate": 1.530325034540755e-05, "loss": 0.7452, "step": 651500 }, { "epoch": 0.24, "learning_rate": 1.5299645779287372e-05, "loss": 0.7589, "step": 652000 }, { "epoch": 0.24, "learning_rate": 1.529604121316719e-05, "loss": 0.7535, "step": 652500 }, { "epoch": 0.24, "learning_rate": 1.5292436647047016e-05, "loss": 0.7854, "step": 653000 }, { "epoch": 0.24, "learning_rate": 1.5288832080926835e-05, "loss": 0.7676, "step": 653500 }, { "epoch": 0.24, "learning_rate": 1.5285227514806657e-05, "loss": 0.7766, "step": 654000 }, { "epoch": 0.24, "learning_rate": 1.528162294868648e-05, "loss": 0.8005, "step": 654500 }, { "epoch": 0.24, "learning_rate": 1.52780183825663e-05, "loss": 0.7795, "step": 655000 }, { "epoch": 0.24, "learning_rate": 1.5274413816446123e-05, "loss": 0.7498, "step": 655500 }, { "epoch": 0.24, "learning_rate": 1.5270809250325945e-05, "loss": 0.7901, "step": 656000 }, { "epoch": 0.24, "learning_rate": 1.5267204684205767e-05, "loss": 0.7559, "step": 656500 }, { "epoch": 0.24, "learning_rate": 1.5263600118085586e-05, "loss": 0.8101, "step": 657000 }, { "epoch": 0.24, "learning_rate": 1.525999555196541e-05, "loss": 0.7852, "step": 657500 }, { "epoch": 0.24, "learning_rate": 1.5256390985845231e-05, "loss": 0.7388, "step": 658000 }, { "epoch": 0.24, "learning_rate": 1.5252786419725052e-05, "loss": 0.7806, "step": 658500 }, { "epoch": 0.24, "learning_rate": 1.5249181853604874e-05, "loss": 0.7962, "step": 659000 }, { "epoch": 0.24, "learning_rate": 1.5245577287484696e-05, "loss": 0.779, "step": 659500 }, { "epoch": 0.24, "learning_rate": 1.5241972721364518e-05, "loss": 0.7613, "step": 660000 }, { "epoch": 0.24, "learning_rate": 1.5238368155244338e-05, "loss": 0.7773, "step": 660500 }, { "epoch": 0.24, "learning_rate": 1.523476358912416e-05, "loss": 0.7614, "step": 661000 }, { "epoch": 0.24, "learning_rate": 1.523115902300398e-05, "loss": 0.7703, "step": 661500 }, { "epoch": 0.24, "learning_rate": 1.5227554456883804e-05, "loss": 0.793, "step": 662000 }, { "epoch": 0.24, "learning_rate": 1.5223949890763626e-05, "loss": 0.7688, "step": 662500 }, { "epoch": 0.24, "learning_rate": 1.5220345324643446e-05, "loss": 0.7882, "step": 663000 }, { "epoch": 0.24, "learning_rate": 1.5216740758523267e-05, "loss": 0.7733, "step": 663500 }, { "epoch": 0.24, "learning_rate": 1.521313619240309e-05, "loss": 0.7576, "step": 664000 }, { "epoch": 0.24, "learning_rate": 1.5209531626282912e-05, "loss": 0.7933, "step": 664500 }, { "epoch": 0.24, "learning_rate": 1.5205927060162733e-05, "loss": 0.791, "step": 665000 }, { "epoch": 0.24, "learning_rate": 1.5202322494042555e-05, "loss": 0.7815, "step": 665500 }, { "epoch": 0.24, "learning_rate": 1.5198717927922375e-05, "loss": 0.7789, "step": 666000 }, { "epoch": 0.24, "learning_rate": 1.5195113361802199e-05, "loss": 0.7981, "step": 666500 }, { "epoch": 0.24, "learning_rate": 1.5191508795682019e-05, "loss": 0.7924, "step": 667000 }, { "epoch": 0.24, "learning_rate": 1.5187904229561841e-05, "loss": 0.7936, "step": 667500 }, { "epoch": 0.24, "learning_rate": 1.5184299663441661e-05, "loss": 0.7929, "step": 668000 }, { "epoch": 0.24, "learning_rate": 1.5180695097321485e-05, "loss": 0.7837, "step": 668500 }, { "epoch": 0.24, "learning_rate": 1.5177090531201305e-05, "loss": 0.7909, "step": 669000 }, { "epoch": 0.24, "learning_rate": 1.5173485965081127e-05, "loss": 0.785, "step": 669500 }, { "epoch": 0.24, "learning_rate": 1.5169881398960948e-05, "loss": 0.7982, "step": 670000 }, { "epoch": 0.24, "learning_rate": 1.5166276832840771e-05, "loss": 0.7735, "step": 670500 }, { "epoch": 0.24, "learning_rate": 1.5162672266720593e-05, "loss": 0.7502, "step": 671000 }, { "epoch": 0.24, "learning_rate": 1.5159067700600414e-05, "loss": 0.7934, "step": 671500 }, { "epoch": 0.24, "learning_rate": 1.5155463134480236e-05, "loss": 0.7761, "step": 672000 }, { "epoch": 0.24, "learning_rate": 1.5151858568360056e-05, "loss": 0.7658, "step": 672500 }, { "epoch": 0.24, "learning_rate": 1.514825400223988e-05, "loss": 0.7587, "step": 673000 }, { "epoch": 0.24, "learning_rate": 1.51446494361197e-05, "loss": 0.7975, "step": 673500 }, { "epoch": 0.24, "learning_rate": 1.5141044869999522e-05, "loss": 0.8022, "step": 674000 }, { "epoch": 0.24, "learning_rate": 1.5137440303879342e-05, "loss": 0.7653, "step": 674500 }, { "epoch": 0.24, "learning_rate": 1.5133835737759166e-05, "loss": 0.7662, "step": 675000 }, { "epoch": 0.24, "learning_rate": 1.5130231171638987e-05, "loss": 0.7693, "step": 675500 }, { "epoch": 0.24, "learning_rate": 1.5126626605518809e-05, "loss": 0.7802, "step": 676000 }, { "epoch": 0.24, "learning_rate": 1.5123022039398629e-05, "loss": 0.7609, "step": 676500 }, { "epoch": 0.24, "learning_rate": 1.5119417473278451e-05, "loss": 0.7605, "step": 677000 }, { "epoch": 0.24, "learning_rate": 1.5115812907158275e-05, "loss": 0.7707, "step": 677500 }, { "epoch": 0.24, "learning_rate": 1.5112208341038095e-05, "loss": 0.7813, "step": 678000 }, { "epoch": 0.24, "learning_rate": 1.5108603774917915e-05, "loss": 0.7524, "step": 678500 }, { "epoch": 0.24, "learning_rate": 1.5104999208797737e-05, "loss": 0.7752, "step": 679000 }, { "epoch": 0.24, "learning_rate": 1.5101394642677561e-05, "loss": 0.7559, "step": 679500 }, { "epoch": 0.25, "learning_rate": 1.5097790076557381e-05, "loss": 0.7804, "step": 680000 }, { "epoch": 0.25, "learning_rate": 1.5094185510437203e-05, "loss": 0.7777, "step": 680500 }, { "epoch": 0.25, "learning_rate": 1.5090580944317024e-05, "loss": 0.8201, "step": 681000 }, { "epoch": 0.25, "learning_rate": 1.5086976378196846e-05, "loss": 0.7645, "step": 681500 }, { "epoch": 0.25, "learning_rate": 1.5083371812076668e-05, "loss": 0.7664, "step": 682000 }, { "epoch": 0.25, "learning_rate": 1.507976724595649e-05, "loss": 0.7711, "step": 682500 }, { "epoch": 0.25, "learning_rate": 1.507616267983631e-05, "loss": 0.8008, "step": 683000 }, { "epoch": 0.25, "learning_rate": 1.5072558113716132e-05, "loss": 0.7845, "step": 683500 }, { "epoch": 0.25, "learning_rate": 1.5068953547595954e-05, "loss": 0.7628, "step": 684000 }, { "epoch": 0.25, "learning_rate": 1.5065348981475776e-05, "loss": 0.7515, "step": 684500 }, { "epoch": 0.25, "learning_rate": 1.5061744415355596e-05, "loss": 0.7489, "step": 685000 }, { "epoch": 0.25, "learning_rate": 1.5058139849235418e-05, "loss": 0.7552, "step": 685500 }, { "epoch": 0.25, "learning_rate": 1.5054535283115242e-05, "loss": 0.7828, "step": 686000 }, { "epoch": 0.25, "learning_rate": 1.5050930716995062e-05, "loss": 0.799, "step": 686500 }, { "epoch": 0.25, "learning_rate": 1.5047326150874884e-05, "loss": 0.7463, "step": 687000 }, { "epoch": 0.25, "learning_rate": 1.5043721584754705e-05, "loss": 0.7517, "step": 687500 }, { "epoch": 0.25, "learning_rate": 1.5040117018634525e-05, "loss": 0.7671, "step": 688000 }, { "epoch": 0.25, "learning_rate": 1.5036512452514349e-05, "loss": 0.754, "step": 688500 }, { "epoch": 0.25, "learning_rate": 1.503290788639417e-05, "loss": 0.7481, "step": 689000 }, { "epoch": 0.25, "learning_rate": 1.5029303320273991e-05, "loss": 0.7829, "step": 689500 }, { "epoch": 0.25, "learning_rate": 1.5025698754153813e-05, "loss": 0.7926, "step": 690000 }, { "epoch": 0.25, "learning_rate": 1.5022094188033635e-05, "loss": 0.7931, "step": 690500 }, { "epoch": 0.25, "learning_rate": 1.5018489621913457e-05, "loss": 0.8153, "step": 691000 }, { "epoch": 0.25, "learning_rate": 1.5014885055793277e-05, "loss": 0.7948, "step": 691500 }, { "epoch": 0.25, "learning_rate": 1.50112804896731e-05, "loss": 0.766, "step": 692000 }, { "epoch": 0.25, "learning_rate": 1.500767592355292e-05, "loss": 0.7635, "step": 692500 }, { "epoch": 0.25, "learning_rate": 1.5004071357432743e-05, "loss": 0.7687, "step": 693000 }, { "epoch": 0.25, "learning_rate": 1.5000466791312564e-05, "loss": 0.7793, "step": 693500 }, { "epoch": 0.25, "learning_rate": 1.4996862225192386e-05, "loss": 0.7808, "step": 694000 }, { "epoch": 0.25, "learning_rate": 1.4993257659072206e-05, "loss": 0.7852, "step": 694500 }, { "epoch": 0.25, "learning_rate": 1.498965309295203e-05, "loss": 0.7655, "step": 695000 }, { "epoch": 0.25, "learning_rate": 1.4986048526831852e-05, "loss": 0.7917, "step": 695500 }, { "epoch": 0.25, "learning_rate": 1.4982443960711672e-05, "loss": 0.7279, "step": 696000 }, { "epoch": 0.25, "learning_rate": 1.4978839394591494e-05, "loss": 0.7684, "step": 696500 }, { "epoch": 0.25, "learning_rate": 1.4975234828471314e-05, "loss": 0.7554, "step": 697000 }, { "epoch": 0.25, "learning_rate": 1.4971630262351138e-05, "loss": 0.7851, "step": 697500 }, { "epoch": 0.25, "learning_rate": 1.4968025696230958e-05, "loss": 0.7822, "step": 698000 }, { "epoch": 0.25, "learning_rate": 1.496442113011078e-05, "loss": 0.7528, "step": 698500 }, { "epoch": 0.25, "learning_rate": 1.49608165639906e-05, "loss": 0.7787, "step": 699000 }, { "epoch": 0.25, "learning_rate": 1.4957211997870425e-05, "loss": 0.7515, "step": 699500 }, { "epoch": 0.25, "learning_rate": 1.4953607431750245e-05, "loss": 0.7607, "step": 700000 }, { "epoch": 0.25, "learning_rate": 1.4950002865630067e-05, "loss": 0.7678, "step": 700500 }, { "epoch": 0.25, "learning_rate": 1.4946398299509887e-05, "loss": 0.7991, "step": 701000 }, { "epoch": 0.25, "learning_rate": 1.4942793733389711e-05, "loss": 0.7845, "step": 701500 }, { "epoch": 0.25, "learning_rate": 1.4939189167269533e-05, "loss": 0.7862, "step": 702000 }, { "epoch": 0.25, "learning_rate": 1.4935584601149353e-05, "loss": 0.7822, "step": 702500 }, { "epoch": 0.25, "learning_rate": 1.4931980035029174e-05, "loss": 0.7965, "step": 703000 }, { "epoch": 0.25, "learning_rate": 1.4928375468908996e-05, "loss": 0.7956, "step": 703500 }, { "epoch": 0.25, "learning_rate": 1.492477090278882e-05, "loss": 0.7651, "step": 704000 }, { "epoch": 0.25, "learning_rate": 1.492116633666864e-05, "loss": 0.7849, "step": 704500 }, { "epoch": 0.25, "learning_rate": 1.4917561770548462e-05, "loss": 0.7575, "step": 705000 }, { "epoch": 0.25, "learning_rate": 1.4913957204428282e-05, "loss": 0.778, "step": 705500 }, { "epoch": 0.25, "learning_rate": 1.4910352638308106e-05, "loss": 0.7811, "step": 706000 }, { "epoch": 0.25, "learning_rate": 1.4906748072187926e-05, "loss": 0.746, "step": 706500 }, { "epoch": 0.25, "learning_rate": 1.4903143506067748e-05, "loss": 0.7624, "step": 707000 }, { "epoch": 0.26, "learning_rate": 1.4899538939947568e-05, "loss": 0.7823, "step": 707500 }, { "epoch": 0.26, "learning_rate": 1.489593437382739e-05, "loss": 0.7799, "step": 708000 }, { "epoch": 0.26, "learning_rate": 1.4892329807707212e-05, "loss": 0.7494, "step": 708500 }, { "epoch": 0.26, "learning_rate": 1.4888725241587034e-05, "loss": 0.7713, "step": 709000 }, { "epoch": 0.26, "learning_rate": 1.4885120675466855e-05, "loss": 0.7918, "step": 709500 }, { "epoch": 0.26, "learning_rate": 1.4881516109346677e-05, "loss": 0.7577, "step": 710000 }, { "epoch": 0.26, "learning_rate": 1.48779115432265e-05, "loss": 0.7627, "step": 710500 }, { "epoch": 0.26, "learning_rate": 1.487430697710632e-05, "loss": 0.7648, "step": 711000 }, { "epoch": 0.26, "learning_rate": 1.4870702410986143e-05, "loss": 0.8025, "step": 711500 }, { "epoch": 0.26, "learning_rate": 1.4867097844865963e-05, "loss": 0.789, "step": 712000 }, { "epoch": 0.26, "learning_rate": 1.4863493278745785e-05, "loss": 0.7934, "step": 712500 }, { "epoch": 0.26, "learning_rate": 1.4859888712625607e-05, "loss": 0.7738, "step": 713000 }, { "epoch": 0.26, "learning_rate": 1.4856284146505429e-05, "loss": 0.7761, "step": 713500 }, { "epoch": 0.26, "learning_rate": 1.485267958038525e-05, "loss": 0.7713, "step": 714000 }, { "epoch": 0.26, "learning_rate": 1.4849075014265071e-05, "loss": 0.7714, "step": 714500 }, { "epoch": 0.26, "learning_rate": 1.4845470448144893e-05, "loss": 0.786, "step": 715000 }, { "epoch": 0.26, "learning_rate": 1.4841865882024715e-05, "loss": 0.7706, "step": 715500 }, { "epoch": 0.26, "learning_rate": 1.4838261315904536e-05, "loss": 0.7788, "step": 716000 }, { "epoch": 0.26, "learning_rate": 1.4834656749784358e-05, "loss": 0.7629, "step": 716500 }, { "epoch": 0.26, "learning_rate": 1.4831052183664181e-05, "loss": 0.7656, "step": 717000 }, { "epoch": 0.26, "learning_rate": 1.4827447617544002e-05, "loss": 0.7843, "step": 717500 }, { "epoch": 0.26, "learning_rate": 1.4823843051423824e-05, "loss": 0.7743, "step": 718000 }, { "epoch": 0.26, "learning_rate": 1.4820238485303644e-05, "loss": 0.7965, "step": 718500 }, { "epoch": 0.26, "learning_rate": 1.4816633919183464e-05, "loss": 0.7654, "step": 719000 }, { "epoch": 0.26, "learning_rate": 1.4813029353063288e-05, "loss": 0.7814, "step": 719500 }, { "epoch": 0.26, "learning_rate": 1.480942478694311e-05, "loss": 0.7887, "step": 720000 }, { "epoch": 0.26, "learning_rate": 1.480582022082293e-05, "loss": 0.7684, "step": 720500 }, { "epoch": 0.26, "learning_rate": 1.4802215654702752e-05, "loss": 0.7733, "step": 721000 }, { "epoch": 0.26, "learning_rate": 1.4798611088582574e-05, "loss": 0.7716, "step": 721500 }, { "epoch": 0.26, "learning_rate": 1.4795006522462397e-05, "loss": 0.7702, "step": 722000 }, { "epoch": 0.26, "learning_rate": 1.4791401956342217e-05, "loss": 0.7837, "step": 722500 }, { "epoch": 0.26, "learning_rate": 1.4787797390222039e-05, "loss": 0.7785, "step": 723000 }, { "epoch": 0.26, "learning_rate": 1.478419282410186e-05, "loss": 0.7697, "step": 723500 }, { "epoch": 0.26, "learning_rate": 1.4780588257981683e-05, "loss": 0.7717, "step": 724000 }, { "epoch": 0.26, "learning_rate": 1.4776983691861503e-05, "loss": 0.7426, "step": 724500 }, { "epoch": 0.26, "learning_rate": 1.4773379125741325e-05, "loss": 0.7788, "step": 725000 }, { "epoch": 0.26, "learning_rate": 1.4769774559621146e-05, "loss": 0.7488, "step": 725500 }, { "epoch": 0.26, "learning_rate": 1.476616999350097e-05, "loss": 0.7726, "step": 726000 }, { "epoch": 0.26, "learning_rate": 1.4762565427380791e-05, "loss": 0.7852, "step": 726500 }, { "epoch": 0.26, "learning_rate": 1.4758960861260612e-05, "loss": 0.7663, "step": 727000 }, { "epoch": 0.26, "learning_rate": 1.4755356295140434e-05, "loss": 0.7773, "step": 727500 }, { "epoch": 0.26, "learning_rate": 1.4751751729020254e-05, "loss": 0.7457, "step": 728000 }, { "epoch": 0.26, "learning_rate": 1.4748147162900078e-05, "loss": 0.7698, "step": 728500 }, { "epoch": 0.26, "learning_rate": 1.4744542596779898e-05, "loss": 0.7562, "step": 729000 }, { "epoch": 0.26, "learning_rate": 1.474093803065972e-05, "loss": 0.7797, "step": 729500 }, { "epoch": 0.26, "learning_rate": 1.473733346453954e-05, "loss": 0.7633, "step": 730000 }, { "epoch": 0.26, "learning_rate": 1.4733728898419364e-05, "loss": 0.7606, "step": 730500 }, { "epoch": 0.26, "learning_rate": 1.4730124332299184e-05, "loss": 0.758, "step": 731000 }, { "epoch": 0.26, "learning_rate": 1.4726519766179006e-05, "loss": 0.7483, "step": 731500 }, { "epoch": 0.26, "learning_rate": 1.4722915200058827e-05, "loss": 0.7552, "step": 732000 }, { "epoch": 0.26, "learning_rate": 1.4719310633938649e-05, "loss": 0.7711, "step": 732500 }, { "epoch": 0.26, "learning_rate": 1.4715706067818472e-05, "loss": 0.7565, "step": 733000 }, { "epoch": 0.26, "learning_rate": 1.4712101501698293e-05, "loss": 0.7574, "step": 733500 }, { "epoch": 0.26, "learning_rate": 1.4708496935578113e-05, "loss": 0.7647, "step": 734000 }, { "epoch": 0.26, "learning_rate": 1.4704892369457935e-05, "loss": 0.7696, "step": 734500 }, { "epoch": 0.26, "learning_rate": 1.4701287803337759e-05, "loss": 0.7708, "step": 735000 }, { "epoch": 0.27, "learning_rate": 1.4697683237217579e-05, "loss": 0.7765, "step": 735500 }, { "epoch": 0.27, "learning_rate": 1.4694078671097401e-05, "loss": 0.7732, "step": 736000 }, { "epoch": 0.27, "learning_rate": 1.4690474104977221e-05, "loss": 0.7882, "step": 736500 }, { "epoch": 0.27, "learning_rate": 1.4686869538857045e-05, "loss": 0.7782, "step": 737000 }, { "epoch": 0.27, "learning_rate": 1.4683264972736865e-05, "loss": 0.7671, "step": 737500 }, { "epoch": 0.27, "learning_rate": 1.4679660406616687e-05, "loss": 0.7689, "step": 738000 }, { "epoch": 0.27, "learning_rate": 1.4676055840496508e-05, "loss": 0.7365, "step": 738500 }, { "epoch": 0.27, "learning_rate": 1.467245127437633e-05, "loss": 0.7594, "step": 739000 }, { "epoch": 0.27, "learning_rate": 1.4668846708256152e-05, "loss": 0.767, "step": 739500 }, { "epoch": 0.27, "learning_rate": 1.4665242142135974e-05, "loss": 0.7767, "step": 740000 }, { "epoch": 0.27, "learning_rate": 1.4661637576015794e-05, "loss": 0.7702, "step": 740500 }, { "epoch": 0.27, "learning_rate": 1.4658033009895616e-05, "loss": 0.7761, "step": 741000 }, { "epoch": 0.27, "learning_rate": 1.465442844377544e-05, "loss": 0.7734, "step": 741500 }, { "epoch": 0.27, "learning_rate": 1.465082387765526e-05, "loss": 0.7572, "step": 742000 }, { "epoch": 0.27, "learning_rate": 1.4647219311535082e-05, "loss": 0.7486, "step": 742500 }, { "epoch": 0.27, "learning_rate": 1.4643614745414902e-05, "loss": 0.7554, "step": 743000 }, { "epoch": 0.27, "learning_rate": 1.4640010179294723e-05, "loss": 0.7579, "step": 743500 }, { "epoch": 0.27, "learning_rate": 1.4636405613174546e-05, "loss": 0.7797, "step": 744000 }, { "epoch": 0.27, "learning_rate": 1.4632801047054368e-05, "loss": 0.7783, "step": 744500 }, { "epoch": 0.27, "learning_rate": 1.4629196480934189e-05, "loss": 0.7958, "step": 745000 }, { "epoch": 0.27, "learning_rate": 1.462559191481401e-05, "loss": 0.7889, "step": 745500 }, { "epoch": 0.27, "learning_rate": 1.4621987348693833e-05, "loss": 0.7833, "step": 746000 }, { "epoch": 0.27, "learning_rate": 1.4618382782573655e-05, "loss": 0.7716, "step": 746500 }, { "epoch": 0.27, "learning_rate": 1.4614778216453475e-05, "loss": 0.7637, "step": 747000 }, { "epoch": 0.27, "learning_rate": 1.4611173650333297e-05, "loss": 0.7625, "step": 747500 }, { "epoch": 0.27, "learning_rate": 1.4607569084213118e-05, "loss": 0.7999, "step": 748000 }, { "epoch": 0.27, "learning_rate": 1.4603964518092941e-05, "loss": 0.7646, "step": 748500 }, { "epoch": 0.27, "learning_rate": 1.4600359951972762e-05, "loss": 0.7431, "step": 749000 }, { "epoch": 0.27, "learning_rate": 1.4596755385852584e-05, "loss": 0.7604, "step": 749500 }, { "epoch": 0.27, "learning_rate": 1.4593150819732404e-05, "loss": 0.7327, "step": 750000 }, { "epoch": 0.27, "learning_rate": 1.4589546253612228e-05, "loss": 0.763, "step": 750500 }, { "epoch": 0.27, "learning_rate": 1.458594168749205e-05, "loss": 0.7555, "step": 751000 }, { "epoch": 0.27, "learning_rate": 1.458233712137187e-05, "loss": 0.7476, "step": 751500 }, { "epoch": 0.27, "learning_rate": 1.4578732555251692e-05, "loss": 0.7574, "step": 752000 }, { "epoch": 0.27, "learning_rate": 1.4575127989131514e-05, "loss": 0.7432, "step": 752500 }, { "epoch": 0.27, "learning_rate": 1.4571523423011336e-05, "loss": 0.7687, "step": 753000 }, { "epoch": 0.27, "learning_rate": 1.4567918856891156e-05, "loss": 0.7461, "step": 753500 }, { "epoch": 0.27, "learning_rate": 1.4564314290770978e-05, "loss": 0.7439, "step": 754000 }, { "epoch": 0.27, "learning_rate": 1.4560709724650799e-05, "loss": 0.774, "step": 754500 }, { "epoch": 0.27, "learning_rate": 1.4557105158530622e-05, "loss": 0.7483, "step": 755000 }, { "epoch": 0.27, "learning_rate": 1.4553500592410443e-05, "loss": 0.7452, "step": 755500 }, { "epoch": 0.27, "learning_rate": 1.4549896026290265e-05, "loss": 0.7877, "step": 756000 }, { "epoch": 0.27, "learning_rate": 1.4546291460170085e-05, "loss": 0.7852, "step": 756500 }, { "epoch": 0.27, "learning_rate": 1.4542686894049909e-05, "loss": 0.7645, "step": 757000 }, { "epoch": 0.27, "learning_rate": 1.453908232792973e-05, "loss": 0.7357, "step": 757500 }, { "epoch": 0.27, "learning_rate": 1.4535477761809551e-05, "loss": 0.7558, "step": 758000 }, { "epoch": 0.27, "learning_rate": 1.4531873195689371e-05, "loss": 0.7773, "step": 758500 }, { "epoch": 0.27, "learning_rate": 1.4528268629569193e-05, "loss": 0.7455, "step": 759000 }, { "epoch": 0.27, "learning_rate": 1.4524664063449017e-05, "loss": 0.7608, "step": 759500 }, { "epoch": 0.27, "learning_rate": 1.4521059497328837e-05, "loss": 0.7761, "step": 760000 }, { "epoch": 0.27, "learning_rate": 1.451745493120866e-05, "loss": 0.7385, "step": 760500 }, { "epoch": 0.27, "learning_rate": 1.451385036508848e-05, "loss": 0.7579, "step": 761000 }, { "epoch": 0.27, "learning_rate": 1.4510245798968303e-05, "loss": 0.7574, "step": 761500 }, { "epoch": 0.27, "learning_rate": 1.4506641232848124e-05, "loss": 0.7522, "step": 762000 }, { "epoch": 0.27, "learning_rate": 1.4503036666727946e-05, "loss": 0.7659, "step": 762500 }, { "epoch": 0.28, "learning_rate": 1.4499432100607766e-05, "loss": 0.7491, "step": 763000 }, { "epoch": 0.28, "learning_rate": 1.4495827534487588e-05, "loss": 0.7532, "step": 763500 }, { "epoch": 0.28, "learning_rate": 1.4492222968367412e-05, "loss": 0.7814, "step": 764000 }, { "epoch": 0.28, "learning_rate": 1.4488618402247232e-05, "loss": 0.7715, "step": 764500 }, { "epoch": 0.28, "learning_rate": 1.4485013836127052e-05, "loss": 0.745, "step": 765000 }, { "epoch": 0.28, "learning_rate": 1.4481409270006874e-05, "loss": 0.7467, "step": 765500 }, { "epoch": 0.28, "learning_rate": 1.4477804703886698e-05, "loss": 0.8059, "step": 766000 }, { "epoch": 0.28, "learning_rate": 1.4474200137766518e-05, "loss": 0.7683, "step": 766500 }, { "epoch": 0.28, "learning_rate": 1.447059557164634e-05, "loss": 0.7197, "step": 767000 }, { "epoch": 0.28, "learning_rate": 1.446699100552616e-05, "loss": 0.8072, "step": 767500 }, { "epoch": 0.28, "learning_rate": 1.4463386439405985e-05, "loss": 0.7753, "step": 768000 }, { "epoch": 0.28, "learning_rate": 1.4459781873285805e-05, "loss": 0.7618, "step": 768500 }, { "epoch": 0.28, "learning_rate": 1.4456177307165627e-05, "loss": 0.7821, "step": 769000 }, { "epoch": 0.28, "learning_rate": 1.4452572741045447e-05, "loss": 0.7711, "step": 769500 }, { "epoch": 0.28, "learning_rate": 1.444896817492527e-05, "loss": 0.7581, "step": 770000 }, { "epoch": 0.28, "learning_rate": 1.4445363608805091e-05, "loss": 0.7723, "step": 770500 }, { "epoch": 0.28, "learning_rate": 1.4441759042684913e-05, "loss": 0.7568, "step": 771000 }, { "epoch": 0.28, "learning_rate": 1.4438154476564734e-05, "loss": 0.728, "step": 771500 }, { "epoch": 0.28, "learning_rate": 1.4434549910444556e-05, "loss": 0.7776, "step": 772000 }, { "epoch": 0.28, "learning_rate": 1.443094534432438e-05, "loss": 0.7386, "step": 772500 }, { "epoch": 0.28, "learning_rate": 1.44273407782042e-05, "loss": 0.7706, "step": 773000 }, { "epoch": 0.28, "learning_rate": 1.4423736212084022e-05, "loss": 0.7821, "step": 773500 }, { "epoch": 0.28, "learning_rate": 1.4420131645963842e-05, "loss": 0.7561, "step": 774000 }, { "epoch": 0.28, "learning_rate": 1.4416527079843662e-05, "loss": 0.7742, "step": 774500 }, { "epoch": 0.28, "learning_rate": 1.4412922513723486e-05, "loss": 0.7513, "step": 775000 }, { "epoch": 0.28, "learning_rate": 1.4409317947603308e-05, "loss": 0.748, "step": 775500 }, { "epoch": 0.28, "learning_rate": 1.4405713381483128e-05, "loss": 0.7899, "step": 776000 }, { "epoch": 0.28, "learning_rate": 1.440210881536295e-05, "loss": 0.7359, "step": 776500 }, { "epoch": 0.28, "learning_rate": 1.4398504249242772e-05, "loss": 0.7685, "step": 777000 }, { "epoch": 0.28, "learning_rate": 1.4394899683122594e-05, "loss": 0.744, "step": 777500 }, { "epoch": 0.28, "learning_rate": 1.4391295117002415e-05, "loss": 0.7699, "step": 778000 }, { "epoch": 0.28, "learning_rate": 1.4387690550882237e-05, "loss": 0.7604, "step": 778500 }, { "epoch": 0.28, "learning_rate": 1.4384085984762057e-05, "loss": 0.7672, "step": 779000 }, { "epoch": 0.28, "learning_rate": 1.438048141864188e-05, "loss": 0.7406, "step": 779500 }, { "epoch": 0.28, "learning_rate": 1.4376876852521701e-05, "loss": 0.7585, "step": 780000 }, { "epoch": 0.28, "learning_rate": 1.4373272286401523e-05, "loss": 0.7844, "step": 780500 }, { "epoch": 0.28, "learning_rate": 1.4369667720281343e-05, "loss": 0.7658, "step": 781000 }, { "epoch": 0.28, "learning_rate": 1.4366063154161167e-05, "loss": 0.7602, "step": 781500 }, { "epoch": 0.28, "learning_rate": 1.4362458588040989e-05, "loss": 0.7518, "step": 782000 }, { "epoch": 0.28, "learning_rate": 1.435885402192081e-05, "loss": 0.7614, "step": 782500 }, { "epoch": 0.28, "learning_rate": 1.4355249455800631e-05, "loss": 0.7548, "step": 783000 }, { "epoch": 0.28, "learning_rate": 1.4351644889680453e-05, "loss": 0.7718, "step": 783500 }, { "epoch": 0.28, "learning_rate": 1.4348040323560275e-05, "loss": 0.7626, "step": 784000 }, { "epoch": 0.28, "learning_rate": 1.4344435757440096e-05, "loss": 0.761, "step": 784500 }, { "epoch": 0.28, "learning_rate": 1.4340831191319918e-05, "loss": 0.7538, "step": 785000 }, { "epoch": 0.28, "learning_rate": 1.4337226625199738e-05, "loss": 0.7539, "step": 785500 }, { "epoch": 0.28, "learning_rate": 1.4333622059079562e-05, "loss": 0.7152, "step": 786000 }, { "epoch": 0.28, "learning_rate": 1.4330017492959382e-05, "loss": 0.7488, "step": 786500 }, { "epoch": 0.28, "learning_rate": 1.4326412926839204e-05, "loss": 0.7792, "step": 787000 }, { "epoch": 0.28, "learning_rate": 1.4322808360719024e-05, "loss": 0.7694, "step": 787500 }, { "epoch": 0.28, "learning_rate": 1.4319203794598848e-05, "loss": 0.7836, "step": 788000 }, { "epoch": 0.28, "learning_rate": 1.431559922847867e-05, "loss": 0.7431, "step": 788500 }, { "epoch": 0.28, "learning_rate": 1.431199466235849e-05, "loss": 0.7609, "step": 789000 }, { "epoch": 0.28, "learning_rate": 1.430839009623831e-05, "loss": 0.7838, "step": 789500 }, { "epoch": 0.28, "learning_rate": 1.4304785530118133e-05, "loss": 0.7827, "step": 790000 }, { "epoch": 0.28, "learning_rate": 1.4301180963997956e-05, "loss": 0.7486, "step": 790500 }, { "epoch": 0.29, "learning_rate": 1.4297576397877777e-05, "loss": 0.7414, "step": 791000 }, { "epoch": 0.29, "learning_rate": 1.4293971831757599e-05, "loss": 0.7668, "step": 791500 }, { "epoch": 0.29, "learning_rate": 1.4290367265637419e-05, "loss": 0.7546, "step": 792000 }, { "epoch": 0.29, "learning_rate": 1.4286762699517243e-05, "loss": 0.7418, "step": 792500 }, { "epoch": 0.29, "learning_rate": 1.4283158133397063e-05, "loss": 0.7739, "step": 793000 }, { "epoch": 0.29, "learning_rate": 1.4279553567276885e-05, "loss": 0.7433, "step": 793500 }, { "epoch": 0.29, "learning_rate": 1.4275949001156705e-05, "loss": 0.7789, "step": 794000 }, { "epoch": 0.29, "learning_rate": 1.4272344435036528e-05, "loss": 0.7638, "step": 794500 }, { "epoch": 0.29, "learning_rate": 1.426873986891635e-05, "loss": 0.7766, "step": 795000 }, { "epoch": 0.29, "learning_rate": 1.4265135302796172e-05, "loss": 0.7722, "step": 795500 }, { "epoch": 0.29, "learning_rate": 1.4261530736675992e-05, "loss": 0.7378, "step": 796000 }, { "epoch": 0.29, "learning_rate": 1.4257926170555814e-05, "loss": 0.7774, "step": 796500 }, { "epoch": 0.29, "learning_rate": 1.4254321604435638e-05, "loss": 0.7433, "step": 797000 }, { "epoch": 0.29, "learning_rate": 1.4250717038315458e-05, "loss": 0.746, "step": 797500 }, { "epoch": 0.29, "learning_rate": 1.424711247219528e-05, "loss": 0.7661, "step": 798000 }, { "epoch": 0.29, "learning_rate": 1.42435079060751e-05, "loss": 0.7592, "step": 798500 }, { "epoch": 0.29, "learning_rate": 1.423990333995492e-05, "loss": 0.757, "step": 799000 }, { "epoch": 0.29, "learning_rate": 1.4236298773834744e-05, "loss": 0.767, "step": 799500 }, { "epoch": 0.29, "learning_rate": 1.4232694207714566e-05, "loss": 0.7493, "step": 800000 }, { "epoch": 0.29, "learning_rate": 1.4229089641594387e-05, "loss": 0.749, "step": 800500 }, { "epoch": 0.29, "learning_rate": 1.4225485075474209e-05, "loss": 0.7336, "step": 801000 }, { "epoch": 0.29, "learning_rate": 1.422188050935403e-05, "loss": 0.7851, "step": 801500 }, { "epoch": 0.29, "learning_rate": 1.4218275943233853e-05, "loss": 0.7782, "step": 802000 }, { "epoch": 0.29, "learning_rate": 1.4214671377113673e-05, "loss": 0.7587, "step": 802500 }, { "epoch": 0.29, "learning_rate": 1.4211066810993495e-05, "loss": 0.7811, "step": 803000 }, { "epoch": 0.29, "learning_rate": 1.4207462244873319e-05, "loss": 0.7365, "step": 803500 }, { "epoch": 0.29, "learning_rate": 1.4203857678753139e-05, "loss": 0.7584, "step": 804000 }, { "epoch": 0.29, "learning_rate": 1.420025311263296e-05, "loss": 0.7415, "step": 804500 }, { "epoch": 0.29, "learning_rate": 1.4196648546512781e-05, "loss": 0.7495, "step": 805000 }, { "epoch": 0.29, "learning_rate": 1.4193043980392602e-05, "loss": 0.7707, "step": 805500 }, { "epoch": 0.29, "learning_rate": 1.4189439414272425e-05, "loss": 0.7605, "step": 806000 }, { "epoch": 0.29, "learning_rate": 1.4185834848152247e-05, "loss": 0.7525, "step": 806500 }, { "epoch": 0.29, "learning_rate": 1.4182230282032068e-05, "loss": 0.7675, "step": 807000 }, { "epoch": 0.29, "learning_rate": 1.417862571591189e-05, "loss": 0.7407, "step": 807500 }, { "epoch": 0.29, "learning_rate": 1.4175021149791712e-05, "loss": 0.7358, "step": 808000 }, { "epoch": 0.29, "learning_rate": 1.4171416583671534e-05, "loss": 0.7903, "step": 808500 }, { "epoch": 0.29, "learning_rate": 1.4167812017551354e-05, "loss": 0.7255, "step": 809000 }, { "epoch": 0.29, "learning_rate": 1.4164207451431176e-05, "loss": 0.764, "step": 809500 }, { "epoch": 0.29, "learning_rate": 1.4160602885310996e-05, "loss": 0.7462, "step": 810000 }, { "epoch": 0.29, "learning_rate": 1.415699831919082e-05, "loss": 0.7583, "step": 810500 }, { "epoch": 0.29, "learning_rate": 1.415339375307064e-05, "loss": 0.7729, "step": 811000 }, { "epoch": 0.29, "learning_rate": 1.4149789186950462e-05, "loss": 0.7555, "step": 811500 }, { "epoch": 0.29, "learning_rate": 1.4146184620830283e-05, "loss": 0.7778, "step": 812000 }, { "epoch": 0.29, "learning_rate": 1.4142580054710106e-05, "loss": 0.7332, "step": 812500 }, { "epoch": 0.29, "learning_rate": 1.4138975488589928e-05, "loss": 0.7659, "step": 813000 }, { "epoch": 0.29, "learning_rate": 1.4135370922469749e-05, "loss": 0.7282, "step": 813500 }, { "epoch": 0.29, "learning_rate": 1.4131766356349569e-05, "loss": 0.7621, "step": 814000 }, { "epoch": 0.29, "learning_rate": 1.4128161790229391e-05, "loss": 0.7649, "step": 814500 }, { "epoch": 0.29, "learning_rate": 1.4124557224109215e-05, "loss": 0.7721, "step": 815000 }, { "epoch": 0.29, "learning_rate": 1.4120952657989035e-05, "loss": 0.7732, "step": 815500 }, { "epoch": 0.29, "learning_rate": 1.4117348091868857e-05, "loss": 0.7502, "step": 816000 }, { "epoch": 0.29, "learning_rate": 1.4113743525748677e-05, "loss": 0.7273, "step": 816500 }, { "epoch": 0.29, "learning_rate": 1.4110138959628501e-05, "loss": 0.752, "step": 817000 }, { "epoch": 0.29, "learning_rate": 1.4106534393508322e-05, "loss": 0.709, "step": 817500 }, { "epoch": 0.29, "learning_rate": 1.4102929827388144e-05, "loss": 0.7367, "step": 818000 }, { "epoch": 0.3, "learning_rate": 1.4099325261267964e-05, "loss": 0.7496, "step": 818500 }, { "epoch": 0.3, "learning_rate": 1.4095720695147788e-05, "loss": 0.7488, "step": 819000 }, { "epoch": 0.3, "learning_rate": 1.409211612902761e-05, "loss": 0.7473, "step": 819500 }, { "epoch": 0.3, "learning_rate": 1.408851156290743e-05, "loss": 0.7391, "step": 820000 }, { "epoch": 0.3, "learning_rate": 1.408490699678725e-05, "loss": 0.7498, "step": 820500 }, { "epoch": 0.3, "learning_rate": 1.4081302430667072e-05, "loss": 0.7519, "step": 821000 }, { "epoch": 0.3, "learning_rate": 1.4077697864546896e-05, "loss": 0.7693, "step": 821500 }, { "epoch": 0.3, "learning_rate": 1.4074093298426716e-05, "loss": 0.7478, "step": 822000 }, { "epoch": 0.3, "learning_rate": 1.4070488732306538e-05, "loss": 0.7572, "step": 822500 }, { "epoch": 0.3, "learning_rate": 1.4066884166186359e-05, "loss": 0.7569, "step": 823000 }, { "epoch": 0.3, "learning_rate": 1.4063279600066182e-05, "loss": 0.7399, "step": 823500 }, { "epoch": 0.3, "learning_rate": 1.4059675033946003e-05, "loss": 0.7441, "step": 824000 }, { "epoch": 0.3, "learning_rate": 1.4056070467825825e-05, "loss": 0.7615, "step": 824500 }, { "epoch": 0.3, "learning_rate": 1.4052465901705645e-05, "loss": 0.7327, "step": 825000 }, { "epoch": 0.3, "learning_rate": 1.4048861335585467e-05, "loss": 0.7436, "step": 825500 }, { "epoch": 0.3, "learning_rate": 1.4045256769465289e-05, "loss": 0.7467, "step": 826000 }, { "epoch": 0.3, "learning_rate": 1.4041652203345111e-05, "loss": 0.7312, "step": 826500 }, { "epoch": 0.3, "learning_rate": 1.4038047637224931e-05, "loss": 0.7271, "step": 827000 }, { "epoch": 0.3, "learning_rate": 1.4034443071104753e-05, "loss": 0.7312, "step": 827500 }, { "epoch": 0.3, "learning_rate": 1.4030838504984577e-05, "loss": 0.7344, "step": 828000 }, { "epoch": 0.3, "learning_rate": 1.4027233938864397e-05, "loss": 0.7625, "step": 828500 }, { "epoch": 0.3, "learning_rate": 1.402362937274422e-05, "loss": 0.7332, "step": 829000 }, { "epoch": 0.3, "learning_rate": 1.402002480662404e-05, "loss": 0.7533, "step": 829500 }, { "epoch": 0.3, "learning_rate": 1.401642024050386e-05, "loss": 0.7661, "step": 830000 }, { "epoch": 0.3, "learning_rate": 1.4012815674383684e-05, "loss": 0.751, "step": 830500 }, { "epoch": 0.3, "learning_rate": 1.4009211108263506e-05, "loss": 0.7448, "step": 831000 }, { "epoch": 0.3, "learning_rate": 1.4005606542143326e-05, "loss": 0.7502, "step": 831500 }, { "epoch": 0.3, "learning_rate": 1.4002001976023148e-05, "loss": 0.7338, "step": 832000 }, { "epoch": 0.3, "learning_rate": 1.399839740990297e-05, "loss": 0.7406, "step": 832500 }, { "epoch": 0.3, "learning_rate": 1.3994792843782792e-05, "loss": 0.7615, "step": 833000 }, { "epoch": 0.3, "learning_rate": 1.3991188277662612e-05, "loss": 0.7436, "step": 833500 }, { "epoch": 0.3, "learning_rate": 1.3987583711542434e-05, "loss": 0.7669, "step": 834000 }, { "epoch": 0.3, "learning_rate": 1.3983979145422258e-05, "loss": 0.7495, "step": 834500 }, { "epoch": 0.3, "learning_rate": 1.3980374579302078e-05, "loss": 0.7584, "step": 835000 }, { "epoch": 0.3, "learning_rate": 1.3976770013181899e-05, "loss": 0.746, "step": 835500 }, { "epoch": 0.3, "learning_rate": 1.397316544706172e-05, "loss": 0.7637, "step": 836000 }, { "epoch": 0.3, "learning_rate": 1.3969560880941541e-05, "loss": 0.742, "step": 836500 }, { "epoch": 0.3, "learning_rate": 1.3965956314821365e-05, "loss": 0.7454, "step": 837000 }, { "epoch": 0.3, "learning_rate": 1.3962351748701187e-05, "loss": 0.7696, "step": 837500 }, { "epoch": 0.3, "learning_rate": 1.3958747182581007e-05, "loss": 0.7551, "step": 838000 }, { "epoch": 0.3, "learning_rate": 1.3955142616460829e-05, "loss": 0.7737, "step": 838500 }, { "epoch": 0.3, "learning_rate": 1.3951538050340651e-05, "loss": 0.7424, "step": 839000 }, { "epoch": 0.3, "learning_rate": 1.3947933484220473e-05, "loss": 0.7577, "step": 839500 }, { "epoch": 0.3, "learning_rate": 1.3944328918100293e-05, "loss": 0.7804, "step": 840000 }, { "epoch": 0.3, "learning_rate": 1.3940724351980116e-05, "loss": 0.756, "step": 840500 }, { "epoch": 0.3, "learning_rate": 1.3937119785859936e-05, "loss": 0.7489, "step": 841000 }, { "epoch": 0.3, "learning_rate": 1.393351521973976e-05, "loss": 0.7455, "step": 841500 }, { "epoch": 0.3, "learning_rate": 1.392991065361958e-05, "loss": 0.7617, "step": 842000 }, { "epoch": 0.3, "learning_rate": 1.3926306087499402e-05, "loss": 0.7485, "step": 842500 }, { "epoch": 0.3, "learning_rate": 1.3922701521379222e-05, "loss": 0.7392, "step": 843000 }, { "epoch": 0.3, "learning_rate": 1.3919096955259046e-05, "loss": 0.7442, "step": 843500 }, { "epoch": 0.3, "learning_rate": 1.3915492389138868e-05, "loss": 0.7743, "step": 844000 }, { "epoch": 0.3, "learning_rate": 1.3911887823018688e-05, "loss": 0.7513, "step": 844500 }, { "epoch": 0.3, "learning_rate": 1.3908283256898509e-05, "loss": 0.7384, "step": 845000 }, { "epoch": 0.3, "learning_rate": 1.390467869077833e-05, "loss": 0.7596, "step": 845500 }, { "epoch": 0.3, "learning_rate": 1.3901074124658154e-05, "loss": 0.7649, "step": 846000 }, { "epoch": 0.31, "learning_rate": 1.3897469558537975e-05, "loss": 0.7425, "step": 846500 }, { "epoch": 0.31, "learning_rate": 1.3893864992417797e-05, "loss": 0.7469, "step": 847000 }, { "epoch": 0.31, "learning_rate": 1.3890260426297617e-05, "loss": 0.7308, "step": 847500 }, { "epoch": 0.31, "learning_rate": 1.388665586017744e-05, "loss": 0.7659, "step": 848000 }, { "epoch": 0.31, "learning_rate": 1.3883051294057261e-05, "loss": 0.7392, "step": 848500 }, { "epoch": 0.31, "learning_rate": 1.3879446727937083e-05, "loss": 0.7463, "step": 849000 }, { "epoch": 0.31, "learning_rate": 1.3875842161816903e-05, "loss": 0.75, "step": 849500 }, { "epoch": 0.31, "learning_rate": 1.3872237595696727e-05, "loss": 0.7446, "step": 850000 }, { "epoch": 0.31, "learning_rate": 1.3868633029576547e-05, "loss": 0.7884, "step": 850500 }, { "epoch": 0.31, "learning_rate": 1.386502846345637e-05, "loss": 0.783, "step": 851000 }, { "epoch": 0.31, "learning_rate": 1.386142389733619e-05, "loss": 0.7558, "step": 851500 }, { "epoch": 0.31, "learning_rate": 1.3857819331216012e-05, "loss": 0.7859, "step": 852000 }, { "epoch": 0.31, "learning_rate": 1.3854214765095835e-05, "loss": 0.7406, "step": 852500 }, { "epoch": 0.31, "learning_rate": 1.3850610198975656e-05, "loss": 0.7628, "step": 853000 }, { "epoch": 0.31, "learning_rate": 1.3847005632855478e-05, "loss": 0.7714, "step": 853500 }, { "epoch": 0.31, "learning_rate": 1.3843401066735298e-05, "loss": 0.7547, "step": 854000 }, { "epoch": 0.31, "learning_rate": 1.3839796500615122e-05, "loss": 0.7758, "step": 854500 }, { "epoch": 0.31, "learning_rate": 1.3836191934494942e-05, "loss": 0.7995, "step": 855000 }, { "epoch": 0.31, "learning_rate": 1.3832587368374764e-05, "loss": 0.7432, "step": 855500 }, { "epoch": 0.31, "learning_rate": 1.3828982802254584e-05, "loss": 0.7759, "step": 856000 }, { "epoch": 0.31, "learning_rate": 1.3825378236134406e-05, "loss": 0.7433, "step": 856500 }, { "epoch": 0.31, "learning_rate": 1.3821773670014228e-05, "loss": 0.7614, "step": 857000 }, { "epoch": 0.31, "learning_rate": 1.381816910389405e-05, "loss": 0.7599, "step": 857500 }, { "epoch": 0.31, "learning_rate": 1.381456453777387e-05, "loss": 0.7362, "step": 858000 }, { "epoch": 0.31, "learning_rate": 1.3810959971653693e-05, "loss": 0.737, "step": 858500 }, { "epoch": 0.31, "learning_rate": 1.3807355405533516e-05, "loss": 0.7501, "step": 859000 }, { "epoch": 0.31, "learning_rate": 1.3803750839413337e-05, "loss": 0.7507, "step": 859500 }, { "epoch": 0.31, "learning_rate": 1.3800146273293157e-05, "loss": 0.7616, "step": 860000 }, { "epoch": 0.31, "learning_rate": 1.3796541707172979e-05, "loss": 0.7502, "step": 860500 }, { "epoch": 0.31, "learning_rate": 1.37929371410528e-05, "loss": 0.7274, "step": 861000 }, { "epoch": 0.31, "learning_rate": 1.3789332574932623e-05, "loss": 0.7782, "step": 861500 }, { "epoch": 0.31, "learning_rate": 1.3785728008812445e-05, "loss": 0.7729, "step": 862000 }, { "epoch": 0.31, "learning_rate": 1.3782123442692265e-05, "loss": 0.7398, "step": 862500 }, { "epoch": 0.31, "learning_rate": 1.3778518876572087e-05, "loss": 0.7511, "step": 863000 }, { "epoch": 0.31, "learning_rate": 1.377491431045191e-05, "loss": 0.7468, "step": 863500 }, { "epoch": 0.31, "learning_rate": 1.3771309744331732e-05, "loss": 0.7508, "step": 864000 }, { "epoch": 0.31, "learning_rate": 1.3767705178211552e-05, "loss": 0.7399, "step": 864500 }, { "epoch": 0.31, "learning_rate": 1.3764100612091374e-05, "loss": 0.7317, "step": 865000 }, { "epoch": 0.31, "learning_rate": 1.3760496045971196e-05, "loss": 0.717, "step": 865500 }, { "epoch": 0.31, "learning_rate": 1.3756891479851018e-05, "loss": 0.7792, "step": 866000 }, { "epoch": 0.31, "learning_rate": 1.3753286913730838e-05, "loss": 0.7656, "step": 866500 }, { "epoch": 0.31, "learning_rate": 1.374968234761066e-05, "loss": 0.7449, "step": 867000 }, { "epoch": 0.31, "learning_rate": 1.374607778149048e-05, "loss": 0.7299, "step": 867500 }, { "epoch": 0.31, "learning_rate": 1.3742473215370304e-05, "loss": 0.7487, "step": 868000 }, { "epoch": 0.31, "learning_rate": 1.3738868649250126e-05, "loss": 0.7507, "step": 868500 }, { "epoch": 0.31, "learning_rate": 1.3735264083129947e-05, "loss": 0.7334, "step": 869000 }, { "epoch": 0.31, "learning_rate": 1.3731659517009769e-05, "loss": 0.7602, "step": 869500 }, { "epoch": 0.31, "learning_rate": 1.372805495088959e-05, "loss": 0.7687, "step": 870000 }, { "epoch": 0.31, "learning_rate": 1.3724450384769413e-05, "loss": 0.7433, "step": 870500 }, { "epoch": 0.31, "learning_rate": 1.3720845818649233e-05, "loss": 0.7709, "step": 871000 }, { "epoch": 0.31, "learning_rate": 1.3717241252529055e-05, "loss": 0.7162, "step": 871500 }, { "epoch": 0.31, "learning_rate": 1.3713636686408875e-05, "loss": 0.7352, "step": 872000 }, { "epoch": 0.31, "learning_rate": 1.3710032120288699e-05, "loss": 0.7424, "step": 872500 }, { "epoch": 0.31, "learning_rate": 1.370642755416852e-05, "loss": 0.7694, "step": 873000 }, { "epoch": 0.31, "learning_rate": 1.3702822988048341e-05, "loss": 0.7786, "step": 873500 }, { "epoch": 0.32, "learning_rate": 1.3699218421928162e-05, "loss": 0.761, "step": 874000 }, { "epoch": 0.32, "learning_rate": 1.3695613855807985e-05, "loss": 0.7491, "step": 874500 }, { "epoch": 0.32, "learning_rate": 1.3692009289687807e-05, "loss": 0.7405, "step": 875000 }, { "epoch": 0.32, "learning_rate": 1.3688404723567628e-05, "loss": 0.7562, "step": 875500 }, { "epoch": 0.32, "learning_rate": 1.3684800157447448e-05, "loss": 0.7676, "step": 876000 }, { "epoch": 0.32, "learning_rate": 1.368119559132727e-05, "loss": 0.7791, "step": 876500 }, { "epoch": 0.32, "learning_rate": 1.3677591025207094e-05, "loss": 0.7526, "step": 877000 }, { "epoch": 0.32, "learning_rate": 1.3673986459086914e-05, "loss": 0.7395, "step": 877500 }, { "epoch": 0.32, "learning_rate": 1.3670381892966736e-05, "loss": 0.7525, "step": 878000 }, { "epoch": 0.32, "learning_rate": 1.3666777326846556e-05, "loss": 0.7505, "step": 878500 }, { "epoch": 0.32, "learning_rate": 1.366317276072638e-05, "loss": 0.7396, "step": 879000 }, { "epoch": 0.32, "learning_rate": 1.36595681946062e-05, "loss": 0.7111, "step": 879500 }, { "epoch": 0.32, "learning_rate": 1.3655963628486022e-05, "loss": 0.7427, "step": 880000 }, { "epoch": 0.32, "learning_rate": 1.3652359062365843e-05, "loss": 0.7382, "step": 880500 }, { "epoch": 0.32, "learning_rate": 1.3648754496245665e-05, "loss": 0.7427, "step": 881000 }, { "epoch": 0.32, "learning_rate": 1.3645149930125487e-05, "loss": 0.741, "step": 881500 }, { "epoch": 0.32, "learning_rate": 1.3641545364005309e-05, "loss": 0.7534, "step": 882000 }, { "epoch": 0.32, "learning_rate": 1.3637940797885129e-05, "loss": 0.7381, "step": 882500 }, { "epoch": 0.32, "learning_rate": 1.3634336231764951e-05, "loss": 0.7494, "step": 883000 }, { "epoch": 0.32, "learning_rate": 1.3630731665644775e-05, "loss": 0.7366, "step": 883500 }, { "epoch": 0.32, "learning_rate": 1.3627127099524595e-05, "loss": 0.7439, "step": 884000 }, { "epoch": 0.32, "learning_rate": 1.3623522533404417e-05, "loss": 0.7311, "step": 884500 }, { "epoch": 0.32, "learning_rate": 1.3619917967284237e-05, "loss": 0.7533, "step": 885000 }, { "epoch": 0.32, "learning_rate": 1.3616313401164061e-05, "loss": 0.7405, "step": 885500 }, { "epoch": 0.32, "learning_rate": 1.3612708835043881e-05, "loss": 0.7801, "step": 886000 }, { "epoch": 0.32, "learning_rate": 1.3609104268923703e-05, "loss": 0.7256, "step": 886500 }, { "epoch": 0.32, "learning_rate": 1.3605499702803524e-05, "loss": 0.7575, "step": 887000 }, { "epoch": 0.32, "learning_rate": 1.3601895136683346e-05, "loss": 0.7493, "step": 887500 }, { "epoch": 0.32, "learning_rate": 1.3598290570563168e-05, "loss": 0.7318, "step": 888000 }, { "epoch": 0.32, "learning_rate": 1.359468600444299e-05, "loss": 0.7333, "step": 888500 }, { "epoch": 0.32, "learning_rate": 1.359108143832281e-05, "loss": 0.7344, "step": 889000 }, { "epoch": 0.32, "learning_rate": 1.3587476872202632e-05, "loss": 0.743, "step": 889500 }, { "epoch": 0.32, "learning_rate": 1.3583872306082456e-05, "loss": 0.7416, "step": 890000 }, { "epoch": 0.32, "learning_rate": 1.3580267739962276e-05, "loss": 0.7631, "step": 890500 }, { "epoch": 0.32, "learning_rate": 1.3576663173842097e-05, "loss": 0.749, "step": 891000 }, { "epoch": 0.32, "learning_rate": 1.3573058607721919e-05, "loss": 0.7416, "step": 891500 }, { "epoch": 0.32, "learning_rate": 1.3569454041601739e-05, "loss": 0.7327, "step": 892000 }, { "epoch": 0.32, "learning_rate": 1.3565849475481563e-05, "loss": 0.7588, "step": 892500 }, { "epoch": 0.32, "learning_rate": 1.3562244909361385e-05, "loss": 0.7493, "step": 893000 }, { "epoch": 0.32, "learning_rate": 1.3558640343241205e-05, "loss": 0.7306, "step": 893500 }, { "epoch": 0.32, "learning_rate": 1.3555035777121027e-05, "loss": 0.7653, "step": 894000 }, { "epoch": 0.32, "learning_rate": 1.3551431211000849e-05, "loss": 0.7169, "step": 894500 }, { "epoch": 0.32, "learning_rate": 1.3547826644880671e-05, "loss": 0.7471, "step": 895000 }, { "epoch": 0.32, "learning_rate": 1.3544222078760491e-05, "loss": 0.7511, "step": 895500 }, { "epoch": 0.32, "learning_rate": 1.3540617512640313e-05, "loss": 0.7404, "step": 896000 }, { "epoch": 0.32, "learning_rate": 1.3537012946520134e-05, "loss": 0.7541, "step": 896500 }, { "epoch": 0.32, "learning_rate": 1.3533408380399957e-05, "loss": 0.7382, "step": 897000 }, { "epoch": 0.32, "learning_rate": 1.3529803814279778e-05, "loss": 0.7302, "step": 897500 }, { "epoch": 0.32, "learning_rate": 1.35261992481596e-05, "loss": 0.7516, "step": 898000 }, { "epoch": 0.32, "learning_rate": 1.352259468203942e-05, "loss": 0.7572, "step": 898500 }, { "epoch": 0.32, "learning_rate": 1.3518990115919244e-05, "loss": 0.6987, "step": 899000 }, { "epoch": 0.32, "learning_rate": 1.3515385549799066e-05, "loss": 0.7368, "step": 899500 }, { "epoch": 0.32, "learning_rate": 1.3511780983678886e-05, "loss": 0.7307, "step": 900000 }, { "epoch": 0.32, "learning_rate": 1.3508176417558706e-05, "loss": 0.7475, "step": 900500 }, { "epoch": 0.32, "learning_rate": 1.350457185143853e-05, "loss": 0.7378, "step": 901000 }, { "epoch": 0.32, "learning_rate": 1.3500967285318352e-05, "loss": 0.7542, "step": 901500 }, { "epoch": 0.33, "learning_rate": 1.3497362719198172e-05, "loss": 0.7362, "step": 902000 }, { "epoch": 0.33, "learning_rate": 1.3493758153077994e-05, "loss": 0.72, "step": 902500 }, { "epoch": 0.33, "learning_rate": 1.3490153586957815e-05, "loss": 0.7618, "step": 903000 }, { "epoch": 0.33, "learning_rate": 1.3486549020837638e-05, "loss": 0.7837, "step": 903500 }, { "epoch": 0.33, "learning_rate": 1.3482944454717459e-05, "loss": 0.7342, "step": 904000 }, { "epoch": 0.33, "learning_rate": 1.347933988859728e-05, "loss": 0.7414, "step": 904500 }, { "epoch": 0.33, "learning_rate": 1.3475735322477101e-05, "loss": 0.7569, "step": 905000 }, { "epoch": 0.33, "learning_rate": 1.3472130756356925e-05, "loss": 0.7643, "step": 905500 }, { "epoch": 0.33, "learning_rate": 1.3468526190236745e-05, "loss": 0.7514, "step": 906000 }, { "epoch": 0.33, "learning_rate": 1.3464921624116567e-05, "loss": 0.7439, "step": 906500 }, { "epoch": 0.33, "learning_rate": 1.3461317057996387e-05, "loss": 0.7517, "step": 907000 }, { "epoch": 0.33, "learning_rate": 1.345771249187621e-05, "loss": 0.758, "step": 907500 }, { "epoch": 0.33, "learning_rate": 1.3454107925756033e-05, "loss": 0.7724, "step": 908000 }, { "epoch": 0.33, "learning_rate": 1.3450503359635853e-05, "loss": 0.7508, "step": 908500 }, { "epoch": 0.33, "learning_rate": 1.3446898793515675e-05, "loss": 0.7435, "step": 909000 }, { "epoch": 0.33, "learning_rate": 1.3443294227395496e-05, "loss": 0.7366, "step": 909500 }, { "epoch": 0.33, "learning_rate": 1.343968966127532e-05, "loss": 0.764, "step": 910000 }, { "epoch": 0.33, "learning_rate": 1.343608509515514e-05, "loss": 0.7443, "step": 910500 }, { "epoch": 0.33, "learning_rate": 1.3432480529034962e-05, "loss": 0.72, "step": 911000 }, { "epoch": 0.33, "learning_rate": 1.3428875962914782e-05, "loss": 0.7717, "step": 911500 }, { "epoch": 0.33, "learning_rate": 1.3425271396794604e-05, "loss": 0.7305, "step": 912000 }, { "epoch": 0.33, "learning_rate": 1.3421666830674426e-05, "loss": 0.7334, "step": 912500 }, { "epoch": 0.33, "learning_rate": 1.3418062264554248e-05, "loss": 0.7636, "step": 913000 }, { "epoch": 0.33, "learning_rate": 1.3414457698434069e-05, "loss": 0.7253, "step": 913500 }, { "epoch": 0.33, "learning_rate": 1.341085313231389e-05, "loss": 0.7359, "step": 914000 }, { "epoch": 0.33, "learning_rate": 1.3407248566193714e-05, "loss": 0.7274, "step": 914500 }, { "epoch": 0.33, "learning_rate": 1.3403644000073535e-05, "loss": 0.7312, "step": 915000 }, { "epoch": 0.33, "learning_rate": 1.3400039433953355e-05, "loss": 0.7297, "step": 915500 }, { "epoch": 0.33, "learning_rate": 1.3396434867833177e-05, "loss": 0.7612, "step": 916000 }, { "epoch": 0.33, "learning_rate": 1.3392830301713e-05, "loss": 0.7112, "step": 916500 }, { "epoch": 0.33, "learning_rate": 1.3389225735592821e-05, "loss": 0.7339, "step": 917000 }, { "epoch": 0.33, "learning_rate": 1.3385621169472643e-05, "loss": 0.736, "step": 917500 }, { "epoch": 0.33, "learning_rate": 1.3382016603352463e-05, "loss": 0.7631, "step": 918000 }, { "epoch": 0.33, "learning_rate": 1.3378412037232285e-05, "loss": 0.7355, "step": 918500 }, { "epoch": 0.33, "learning_rate": 1.3374807471112107e-05, "loss": 0.7333, "step": 919000 }, { "epoch": 0.33, "learning_rate": 1.337120290499193e-05, "loss": 0.7119, "step": 919500 }, { "epoch": 0.33, "learning_rate": 1.336759833887175e-05, "loss": 0.7245, "step": 920000 }, { "epoch": 0.33, "learning_rate": 1.3363993772751572e-05, "loss": 0.7544, "step": 920500 }, { "epoch": 0.33, "learning_rate": 1.3360389206631395e-05, "loss": 0.6946, "step": 921000 }, { "epoch": 0.33, "learning_rate": 1.3356784640511216e-05, "loss": 0.7636, "step": 921500 }, { "epoch": 0.33, "learning_rate": 1.3353180074391036e-05, "loss": 0.7131, "step": 922000 }, { "epoch": 0.33, "learning_rate": 1.3349575508270858e-05, "loss": 0.7298, "step": 922500 }, { "epoch": 0.33, "learning_rate": 1.3345970942150678e-05, "loss": 0.7718, "step": 923000 }, { "epoch": 0.33, "learning_rate": 1.3342366376030502e-05, "loss": 0.7656, "step": 923500 }, { "epoch": 0.33, "learning_rate": 1.3338761809910324e-05, "loss": 0.7608, "step": 924000 }, { "epoch": 0.33, "learning_rate": 1.3335157243790144e-05, "loss": 0.7098, "step": 924500 }, { "epoch": 0.33, "learning_rate": 1.3331552677669966e-05, "loss": 0.7556, "step": 925000 }, { "epoch": 0.33, "learning_rate": 1.3327948111549788e-05, "loss": 0.7505, "step": 925500 }, { "epoch": 0.33, "learning_rate": 1.332434354542961e-05, "loss": 0.7432, "step": 926000 }, { "epoch": 0.33, "learning_rate": 1.332073897930943e-05, "loss": 0.7318, "step": 926500 }, { "epoch": 0.33, "learning_rate": 1.3317134413189253e-05, "loss": 0.7331, "step": 927000 }, { "epoch": 0.33, "learning_rate": 1.3313529847069073e-05, "loss": 0.7838, "step": 927500 }, { "epoch": 0.33, "learning_rate": 1.3309925280948897e-05, "loss": 0.753, "step": 928000 }, { "epoch": 0.33, "learning_rate": 1.3306320714828717e-05, "loss": 0.7512, "step": 928500 }, { "epoch": 0.33, "learning_rate": 1.3302716148708539e-05, "loss": 0.7168, "step": 929000 }, { "epoch": 0.34, "learning_rate": 1.329911158258836e-05, "loss": 0.7408, "step": 929500 }, { "epoch": 0.34, "learning_rate": 1.3295507016468183e-05, "loss": 0.7664, "step": 930000 }, { "epoch": 0.34, "learning_rate": 1.3291902450348005e-05, "loss": 0.7914, "step": 930500 }, { "epoch": 0.34, "learning_rate": 1.3288297884227825e-05, "loss": 0.7593, "step": 931000 }, { "epoch": 0.34, "learning_rate": 1.3284693318107646e-05, "loss": 0.7589, "step": 931500 }, { "epoch": 0.34, "learning_rate": 1.328108875198747e-05, "loss": 0.7586, "step": 932000 }, { "epoch": 0.34, "learning_rate": 1.3277484185867291e-05, "loss": 0.7326, "step": 932500 }, { "epoch": 0.34, "learning_rate": 1.3273879619747112e-05, "loss": 0.7156, "step": 933000 }, { "epoch": 0.34, "learning_rate": 1.3270275053626934e-05, "loss": 0.7201, "step": 933500 }, { "epoch": 0.34, "learning_rate": 1.3266670487506754e-05, "loss": 0.7162, "step": 934000 }, { "epoch": 0.34, "learning_rate": 1.3263065921386578e-05, "loss": 0.718, "step": 934500 }, { "epoch": 0.34, "learning_rate": 1.3259461355266398e-05, "loss": 0.7345, "step": 935000 }, { "epoch": 0.34, "learning_rate": 1.325585678914622e-05, "loss": 0.7511, "step": 935500 }, { "epoch": 0.34, "learning_rate": 1.325225222302604e-05, "loss": 0.6961, "step": 936000 }, { "epoch": 0.34, "learning_rate": 1.3248647656905864e-05, "loss": 0.7563, "step": 936500 }, { "epoch": 0.34, "learning_rate": 1.3245043090785685e-05, "loss": 0.7116, "step": 937000 }, { "epoch": 0.34, "learning_rate": 1.3241438524665507e-05, "loss": 0.7388, "step": 937500 }, { "epoch": 0.34, "learning_rate": 1.3237833958545327e-05, "loss": 0.743, "step": 938000 }, { "epoch": 0.34, "learning_rate": 1.3234229392425149e-05, "loss": 0.7413, "step": 938500 }, { "epoch": 0.34, "learning_rate": 1.3230624826304973e-05, "loss": 0.7412, "step": 939000 }, { "epoch": 0.34, "learning_rate": 1.3227020260184793e-05, "loss": 0.7453, "step": 939500 }, { "epoch": 0.34, "learning_rate": 1.3223415694064615e-05, "loss": 0.7556, "step": 940000 }, { "epoch": 0.34, "learning_rate": 1.3219811127944435e-05, "loss": 0.7242, "step": 940500 }, { "epoch": 0.34, "learning_rate": 1.3216206561824259e-05, "loss": 0.7392, "step": 941000 }, { "epoch": 0.34, "learning_rate": 1.321260199570408e-05, "loss": 0.7433, "step": 941500 }, { "epoch": 0.34, "learning_rate": 1.3208997429583901e-05, "loss": 0.7283, "step": 942000 }, { "epoch": 0.34, "learning_rate": 1.3205392863463722e-05, "loss": 0.7503, "step": 942500 }, { "epoch": 0.34, "learning_rate": 1.3201788297343544e-05, "loss": 0.726, "step": 943000 }, { "epoch": 0.34, "learning_rate": 1.3198183731223366e-05, "loss": 0.7347, "step": 943500 }, { "epoch": 0.34, "learning_rate": 1.3194579165103188e-05, "loss": 0.7286, "step": 944000 }, { "epoch": 0.34, "learning_rate": 1.3190974598983008e-05, "loss": 0.7137, "step": 944500 }, { "epoch": 0.34, "learning_rate": 1.318737003286283e-05, "loss": 0.7518, "step": 945000 }, { "epoch": 0.34, "learning_rate": 1.3183765466742654e-05, "loss": 0.7175, "step": 945500 }, { "epoch": 0.34, "learning_rate": 1.3180160900622474e-05, "loss": 0.7352, "step": 946000 }, { "epoch": 0.34, "learning_rate": 1.3176556334502294e-05, "loss": 0.7219, "step": 946500 }, { "epoch": 0.34, "learning_rate": 1.3172951768382116e-05, "loss": 0.7354, "step": 947000 }, { "epoch": 0.34, "learning_rate": 1.3169347202261937e-05, "loss": 0.7262, "step": 947500 }, { "epoch": 0.34, "learning_rate": 1.316574263614176e-05, "loss": 0.7407, "step": 948000 }, { "epoch": 0.34, "learning_rate": 1.3162138070021582e-05, "loss": 0.7421, "step": 948500 }, { "epoch": 0.34, "learning_rate": 1.3158533503901403e-05, "loss": 0.7415, "step": 949000 }, { "epoch": 0.34, "learning_rate": 1.3154928937781225e-05, "loss": 0.7437, "step": 949500 }, { "epoch": 0.34, "learning_rate": 1.3151324371661047e-05, "loss": 0.7409, "step": 950000 }, { "epoch": 0.34, "learning_rate": 1.3147719805540869e-05, "loss": 0.7172, "step": 950500 }, { "epoch": 0.34, "learning_rate": 1.3144115239420689e-05, "loss": 0.7247, "step": 951000 }, { "epoch": 0.34, "learning_rate": 1.3140510673300511e-05, "loss": 0.7366, "step": 951500 }, { "epoch": 0.34, "learning_rate": 1.3136906107180333e-05, "loss": 0.6927, "step": 952000 }, { "epoch": 0.34, "learning_rate": 1.3133301541060155e-05, "loss": 0.7445, "step": 952500 }, { "epoch": 0.34, "learning_rate": 1.3129696974939975e-05, "loss": 0.7542, "step": 953000 }, { "epoch": 0.34, "learning_rate": 1.3126092408819797e-05, "loss": 0.7644, "step": 953500 }, { "epoch": 0.34, "learning_rate": 1.3122487842699618e-05, "loss": 0.7267, "step": 954000 }, { "epoch": 0.34, "learning_rate": 1.3118883276579441e-05, "loss": 0.7214, "step": 954500 }, { "epoch": 0.34, "learning_rate": 1.3115278710459263e-05, "loss": 0.7862, "step": 955000 }, { "epoch": 0.34, "learning_rate": 1.3111674144339084e-05, "loss": 0.7563, "step": 955500 }, { "epoch": 0.34, "learning_rate": 1.3108069578218904e-05, "loss": 0.7153, "step": 956000 }, { "epoch": 0.34, "learning_rate": 1.3104465012098728e-05, "loss": 0.7405, "step": 956500 }, { "epoch": 0.34, "learning_rate": 1.310086044597855e-05, "loss": 0.7795, "step": 957000 }, { "epoch": 0.35, "learning_rate": 1.309725587985837e-05, "loss": 0.7509, "step": 957500 }, { "epoch": 0.35, "learning_rate": 1.3093651313738192e-05, "loss": 0.7542, "step": 958000 }, { "epoch": 0.35, "learning_rate": 1.3090046747618012e-05, "loss": 0.7508, "step": 958500 }, { "epoch": 0.35, "learning_rate": 1.3086442181497836e-05, "loss": 0.7139, "step": 959000 }, { "epoch": 0.35, "learning_rate": 1.3082837615377657e-05, "loss": 0.7411, "step": 959500 }, { "epoch": 0.35, "learning_rate": 1.3079233049257479e-05, "loss": 0.7407, "step": 960000 }, { "epoch": 0.35, "learning_rate": 1.3075628483137299e-05, "loss": 0.7268, "step": 960500 }, { "epoch": 0.35, "learning_rate": 1.3072023917017123e-05, "loss": 0.7269, "step": 961000 }, { "epoch": 0.35, "learning_rate": 1.3068419350896943e-05, "loss": 0.7333, "step": 961500 }, { "epoch": 0.35, "learning_rate": 1.3064814784776765e-05, "loss": 0.7113, "step": 962000 }, { "epoch": 0.35, "learning_rate": 1.3061210218656585e-05, "loss": 0.7336, "step": 962500 }, { "epoch": 0.35, "learning_rate": 1.3057605652536407e-05, "loss": 0.7664, "step": 963000 }, { "epoch": 0.35, "learning_rate": 1.3054001086416231e-05, "loss": 0.7164, "step": 963500 }, { "epoch": 0.35, "learning_rate": 1.3050396520296051e-05, "loss": 0.731, "step": 964000 }, { "epoch": 0.35, "learning_rate": 1.3046791954175873e-05, "loss": 0.7152, "step": 964500 }, { "epoch": 0.35, "learning_rate": 1.3043187388055694e-05, "loss": 0.7274, "step": 965000 }, { "epoch": 0.35, "learning_rate": 1.3039582821935517e-05, "loss": 0.7771, "step": 965500 }, { "epoch": 0.35, "learning_rate": 1.3035978255815338e-05, "loss": 0.7607, "step": 966000 }, { "epoch": 0.35, "learning_rate": 1.303237368969516e-05, "loss": 0.75, "step": 966500 }, { "epoch": 0.35, "learning_rate": 1.302876912357498e-05, "loss": 0.737, "step": 967000 }, { "epoch": 0.35, "learning_rate": 1.3025164557454804e-05, "loss": 0.7052, "step": 967500 }, { "epoch": 0.35, "learning_rate": 1.3021559991334624e-05, "loss": 0.7205, "step": 968000 }, { "epoch": 0.35, "learning_rate": 1.3017955425214446e-05, "loss": 0.7153, "step": 968500 }, { "epoch": 0.35, "learning_rate": 1.3014350859094266e-05, "loss": 0.7091, "step": 969000 }, { "epoch": 0.35, "learning_rate": 1.3010746292974088e-05, "loss": 0.6886, "step": 969500 }, { "epoch": 0.35, "learning_rate": 1.3007141726853912e-05, "loss": 0.7397, "step": 970000 }, { "epoch": 0.35, "learning_rate": 1.3003537160733732e-05, "loss": 0.7531, "step": 970500 }, { "epoch": 0.35, "learning_rate": 1.2999932594613553e-05, "loss": 0.7515, "step": 971000 }, { "epoch": 0.35, "learning_rate": 1.2996328028493375e-05, "loss": 0.6988, "step": 971500 }, { "epoch": 0.35, "learning_rate": 1.2992723462373198e-05, "loss": 0.7128, "step": 972000 }, { "epoch": 0.35, "learning_rate": 1.2989118896253019e-05, "loss": 0.7542, "step": 972500 }, { "epoch": 0.35, "learning_rate": 1.298551433013284e-05, "loss": 0.7008, "step": 973000 }, { "epoch": 0.35, "learning_rate": 1.2981909764012661e-05, "loss": 0.7232, "step": 973500 }, { "epoch": 0.35, "learning_rate": 1.2978305197892483e-05, "loss": 0.7319, "step": 974000 }, { "epoch": 0.35, "learning_rate": 1.2974700631772305e-05, "loss": 0.7123, "step": 974500 }, { "epoch": 0.35, "learning_rate": 1.2971096065652127e-05, "loss": 0.7361, "step": 975000 }, { "epoch": 0.35, "learning_rate": 1.2967491499531947e-05, "loss": 0.7353, "step": 975500 }, { "epoch": 0.35, "learning_rate": 1.296388693341177e-05, "loss": 0.7269, "step": 976000 }, { "epoch": 0.35, "learning_rate": 1.2960282367291593e-05, "loss": 0.725, "step": 976500 }, { "epoch": 0.35, "learning_rate": 1.2956677801171413e-05, "loss": 0.7384, "step": 977000 }, { "epoch": 0.35, "learning_rate": 1.2953073235051234e-05, "loss": 0.723, "step": 977500 }, { "epoch": 0.35, "learning_rate": 1.2949468668931056e-05, "loss": 0.7341, "step": 978000 }, { "epoch": 0.35, "learning_rate": 1.2945864102810876e-05, "loss": 0.72, "step": 978500 }, { "epoch": 0.35, "learning_rate": 1.29422595366907e-05, "loss": 0.7468, "step": 979000 }, { "epoch": 0.35, "learning_rate": 1.2938654970570522e-05, "loss": 0.7514, "step": 979500 }, { "epoch": 0.35, "learning_rate": 1.2935050404450342e-05, "loss": 0.7273, "step": 980000 }, { "epoch": 0.35, "learning_rate": 1.2931445838330164e-05, "loss": 0.723, "step": 980500 }, { "epoch": 0.35, "learning_rate": 1.2927841272209986e-05, "loss": 0.7399, "step": 981000 }, { "epoch": 0.35, "learning_rate": 1.2924236706089808e-05, "loss": 0.7278, "step": 981500 }, { "epoch": 0.35, "learning_rate": 1.2920632139969628e-05, "loss": 0.7283, "step": 982000 }, { "epoch": 0.35, "learning_rate": 1.291702757384945e-05, "loss": 0.7381, "step": 982500 }, { "epoch": 0.35, "learning_rate": 1.2913423007729273e-05, "loss": 0.7051, "step": 983000 }, { "epoch": 0.35, "learning_rate": 1.2909818441609095e-05, "loss": 0.7266, "step": 983500 }, { "epoch": 0.35, "learning_rate": 1.2906213875488915e-05, "loss": 0.7661, "step": 984000 }, { "epoch": 0.35, "learning_rate": 1.2902609309368737e-05, "loss": 0.7049, "step": 984500 }, { "epoch": 0.36, "learning_rate": 1.2899004743248557e-05, "loss": 0.7383, "step": 985000 }, { "epoch": 0.36, "learning_rate": 1.2895400177128381e-05, "loss": 0.7007, "step": 985500 }, { "epoch": 0.36, "learning_rate": 1.2891795611008203e-05, "loss": 0.7215, "step": 986000 }, { "epoch": 0.36, "learning_rate": 1.2888191044888023e-05, "loss": 0.7193, "step": 986500 }, { "epoch": 0.36, "learning_rate": 1.2884586478767844e-05, "loss": 0.7593, "step": 987000 }, { "epoch": 0.36, "learning_rate": 1.2880981912647667e-05, "loss": 0.7505, "step": 987500 }, { "epoch": 0.36, "learning_rate": 1.287737734652749e-05, "loss": 0.7344, "step": 988000 }, { "epoch": 0.36, "learning_rate": 1.287377278040731e-05, "loss": 0.7211, "step": 988500 }, { "epoch": 0.36, "learning_rate": 1.2870168214287132e-05, "loss": 0.7381, "step": 989000 }, { "epoch": 0.36, "learning_rate": 1.2866563648166952e-05, "loss": 0.7296, "step": 989500 }, { "epoch": 0.36, "learning_rate": 1.2862959082046776e-05, "loss": 0.7296, "step": 990000 }, { "epoch": 0.36, "learning_rate": 1.2859354515926596e-05, "loss": 0.7093, "step": 990500 }, { "epoch": 0.36, "learning_rate": 1.2855749949806418e-05, "loss": 0.7498, "step": 991000 }, { "epoch": 0.36, "learning_rate": 1.2852145383686238e-05, "loss": 0.7254, "step": 991500 }, { "epoch": 0.36, "learning_rate": 1.2848540817566062e-05, "loss": 0.7231, "step": 992000 }, { "epoch": 0.36, "learning_rate": 1.2844936251445882e-05, "loss": 0.7581, "step": 992500 }, { "epoch": 0.36, "learning_rate": 1.2841331685325704e-05, "loss": 0.7141, "step": 993000 }, { "epoch": 0.36, "learning_rate": 1.2837727119205525e-05, "loss": 0.7045, "step": 993500 }, { "epoch": 0.36, "learning_rate": 1.2834122553085347e-05, "loss": 0.7146, "step": 994000 }, { "epoch": 0.36, "learning_rate": 1.283051798696517e-05, "loss": 0.7341, "step": 994500 }, { "epoch": 0.36, "learning_rate": 1.282691342084499e-05, "loss": 0.7151, "step": 995000 }, { "epoch": 0.36, "learning_rate": 1.2823308854724813e-05, "loss": 0.7381, "step": 995500 }, { "epoch": 0.36, "learning_rate": 1.2819704288604633e-05, "loss": 0.7151, "step": 996000 }, { "epoch": 0.36, "learning_rate": 1.2816099722484457e-05, "loss": 0.7505, "step": 996500 }, { "epoch": 0.36, "learning_rate": 1.2812495156364277e-05, "loss": 0.7423, "step": 997000 }, { "epoch": 0.36, "learning_rate": 1.2808890590244099e-05, "loss": 0.7314, "step": 997500 }, { "epoch": 0.36, "learning_rate": 1.280528602412392e-05, "loss": 0.713, "step": 998000 }, { "epoch": 0.36, "learning_rate": 1.2801681458003743e-05, "loss": 0.7365, "step": 998500 }, { "epoch": 0.36, "learning_rate": 1.2798076891883563e-05, "loss": 0.7294, "step": 999000 }, { "epoch": 0.36, "learning_rate": 1.2794472325763385e-05, "loss": 0.7523, "step": 999500 }, { "epoch": 0.36, "learning_rate": 1.2790867759643206e-05, "loss": 0.7442, "step": 1000000 }, { "epoch": 0.36, "learning_rate": 1.2787263193523028e-05, "loss": 0.7564, "step": 1000500 }, { "epoch": 0.36, "learning_rate": 1.2783658627402851e-05, "loss": 0.7353, "step": 1001000 }, { "epoch": 0.36, "learning_rate": 1.2780054061282672e-05, "loss": 0.7353, "step": 1001500 }, { "epoch": 0.36, "learning_rate": 1.2776449495162492e-05, "loss": 0.7212, "step": 1002000 }, { "epoch": 0.36, "learning_rate": 1.2772844929042314e-05, "loss": 0.7247, "step": 1002500 }, { "epoch": 0.36, "learning_rate": 1.2769240362922138e-05, "loss": 0.7464, "step": 1003000 }, { "epoch": 0.36, "learning_rate": 1.2765635796801958e-05, "loss": 0.7311, "step": 1003500 }, { "epoch": 0.36, "learning_rate": 1.276203123068178e-05, "loss": 0.7443, "step": 1004000 }, { "epoch": 0.36, "learning_rate": 1.27584266645616e-05, "loss": 0.7124, "step": 1004500 }, { "epoch": 0.36, "learning_rate": 1.2754822098441422e-05, "loss": 0.7226, "step": 1005000 }, { "epoch": 0.36, "learning_rate": 1.2751217532321245e-05, "loss": 0.7335, "step": 1005500 }, { "epoch": 0.36, "learning_rate": 1.2747612966201067e-05, "loss": 0.7128, "step": 1006000 }, { "epoch": 0.36, "learning_rate": 1.2744008400080887e-05, "loss": 0.7514, "step": 1006500 }, { "epoch": 0.36, "learning_rate": 1.2740403833960709e-05, "loss": 0.7634, "step": 1007000 }, { "epoch": 0.36, "learning_rate": 1.2736799267840531e-05, "loss": 0.7409, "step": 1007500 }, { "epoch": 0.36, "learning_rate": 1.2733194701720353e-05, "loss": 0.7408, "step": 1008000 }, { "epoch": 0.36, "learning_rate": 1.2729590135600173e-05, "loss": 0.6971, "step": 1008500 }, { "epoch": 0.36, "learning_rate": 1.2725985569479995e-05, "loss": 0.7361, "step": 1009000 }, { "epoch": 0.36, "learning_rate": 1.2722381003359816e-05, "loss": 0.7361, "step": 1009500 }, { "epoch": 0.36, "learning_rate": 1.271877643723964e-05, "loss": 0.7175, "step": 1010000 }, { "epoch": 0.36, "learning_rate": 1.2715171871119461e-05, "loss": 0.7213, "step": 1010500 }, { "epoch": 0.36, "learning_rate": 1.2711567304999282e-05, "loss": 0.7549, "step": 1011000 }, { "epoch": 0.36, "learning_rate": 1.2707962738879102e-05, "loss": 0.7286, "step": 1011500 }, { "epoch": 0.36, "learning_rate": 1.2704358172758926e-05, "loss": 0.7392, "step": 1012000 }, { "epoch": 0.36, "learning_rate": 1.2700753606638748e-05, "loss": 0.735, "step": 1012500 }, { "epoch": 0.37, "learning_rate": 1.2697149040518568e-05, "loss": 0.754, "step": 1013000 }, { "epoch": 0.37, "learning_rate": 1.269354447439839e-05, "loss": 0.7121, "step": 1013500 }, { "epoch": 0.37, "learning_rate": 1.2689939908278212e-05, "loss": 0.7412, "step": 1014000 }, { "epoch": 0.37, "learning_rate": 1.2686335342158034e-05, "loss": 0.7252, "step": 1014500 }, { "epoch": 0.37, "learning_rate": 1.2682730776037854e-05, "loss": 0.7501, "step": 1015000 }, { "epoch": 0.37, "learning_rate": 1.2679126209917676e-05, "loss": 0.7278, "step": 1015500 }, { "epoch": 0.37, "learning_rate": 1.2675521643797497e-05, "loss": 0.7201, "step": 1016000 }, { "epoch": 0.37, "learning_rate": 1.267191707767732e-05, "loss": 0.703, "step": 1016500 }, { "epoch": 0.37, "learning_rate": 1.266831251155714e-05, "loss": 0.7614, "step": 1017000 }, { "epoch": 0.37, "learning_rate": 1.2664707945436963e-05, "loss": 0.7268, "step": 1017500 }, { "epoch": 0.37, "learning_rate": 1.2661103379316783e-05, "loss": 0.7075, "step": 1018000 }, { "epoch": 0.37, "learning_rate": 1.2657498813196607e-05, "loss": 0.7147, "step": 1018500 }, { "epoch": 0.37, "learning_rate": 1.2653894247076429e-05, "loss": 0.7038, "step": 1019000 }, { "epoch": 0.37, "learning_rate": 1.2650289680956249e-05, "loss": 0.7607, "step": 1019500 }, { "epoch": 0.37, "learning_rate": 1.2646685114836071e-05, "loss": 0.7048, "step": 1020000 }, { "epoch": 0.37, "learning_rate": 1.2643080548715891e-05, "loss": 0.754, "step": 1020500 }, { "epoch": 0.37, "learning_rate": 1.2639475982595715e-05, "loss": 0.7465, "step": 1021000 }, { "epoch": 0.37, "learning_rate": 1.2635871416475535e-05, "loss": 0.7304, "step": 1021500 }, { "epoch": 0.37, "learning_rate": 1.2632266850355357e-05, "loss": 0.7199, "step": 1022000 }, { "epoch": 0.37, "learning_rate": 1.2628662284235178e-05, "loss": 0.7087, "step": 1022500 }, { "epoch": 0.37, "learning_rate": 1.2625057718115001e-05, "loss": 0.7018, "step": 1023000 }, { "epoch": 0.37, "learning_rate": 1.2621453151994822e-05, "loss": 0.7563, "step": 1023500 }, { "epoch": 0.37, "learning_rate": 1.2617848585874644e-05, "loss": 0.7351, "step": 1024000 }, { "epoch": 0.37, "learning_rate": 1.2614244019754464e-05, "loss": 0.7253, "step": 1024500 }, { "epoch": 0.37, "learning_rate": 1.2610639453634286e-05, "loss": 0.7393, "step": 1025000 }, { "epoch": 0.37, "learning_rate": 1.260703488751411e-05, "loss": 0.7245, "step": 1025500 }, { "epoch": 0.37, "learning_rate": 1.260343032139393e-05, "loss": 0.7546, "step": 1026000 }, { "epoch": 0.37, "learning_rate": 1.259982575527375e-05, "loss": 0.739, "step": 1026500 }, { "epoch": 0.37, "learning_rate": 1.2596221189153572e-05, "loss": 0.7294, "step": 1027000 }, { "epoch": 0.37, "learning_rate": 1.2592616623033396e-05, "loss": 0.7466, "step": 1027500 }, { "epoch": 0.37, "learning_rate": 1.2589012056913216e-05, "loss": 0.7059, "step": 1028000 }, { "epoch": 0.37, "learning_rate": 1.2585407490793039e-05, "loss": 0.6973, "step": 1028500 }, { "epoch": 0.37, "learning_rate": 1.2581802924672859e-05, "loss": 0.7152, "step": 1029000 }, { "epoch": 0.37, "learning_rate": 1.257819835855268e-05, "loss": 0.7656, "step": 1029500 }, { "epoch": 0.37, "learning_rate": 1.2574593792432503e-05, "loss": 0.7272, "step": 1030000 }, { "epoch": 0.37, "learning_rate": 1.2570989226312325e-05, "loss": 0.7319, "step": 1030500 }, { "epoch": 0.37, "learning_rate": 1.2567384660192145e-05, "loss": 0.7357, "step": 1031000 }, { "epoch": 0.37, "learning_rate": 1.2563780094071967e-05, "loss": 0.7273, "step": 1031500 }, { "epoch": 0.37, "learning_rate": 1.2560175527951791e-05, "loss": 0.7238, "step": 1032000 }, { "epoch": 0.37, "learning_rate": 1.2556570961831611e-05, "loss": 0.7324, "step": 1032500 }, { "epoch": 0.37, "learning_rate": 1.2552966395711432e-05, "loss": 0.7088, "step": 1033000 }, { "epoch": 0.37, "learning_rate": 1.2549361829591254e-05, "loss": 0.699, "step": 1033500 }, { "epoch": 0.37, "learning_rate": 1.2545757263471077e-05, "loss": 0.7267, "step": 1034000 }, { "epoch": 0.37, "learning_rate": 1.2542152697350898e-05, "loss": 0.7426, "step": 1034500 }, { "epoch": 0.37, "learning_rate": 1.253854813123072e-05, "loss": 0.7181, "step": 1035000 }, { "epoch": 0.37, "learning_rate": 1.253494356511054e-05, "loss": 0.745, "step": 1035500 }, { "epoch": 0.37, "learning_rate": 1.2531338998990362e-05, "loss": 0.7487, "step": 1036000 }, { "epoch": 0.37, "learning_rate": 1.2527734432870184e-05, "loss": 0.7253, "step": 1036500 }, { "epoch": 0.37, "learning_rate": 1.2524129866750006e-05, "loss": 0.7396, "step": 1037000 }, { "epoch": 0.37, "learning_rate": 1.2520525300629826e-05, "loss": 0.7497, "step": 1037500 }, { "epoch": 0.37, "learning_rate": 1.2516920734509648e-05, "loss": 0.7295, "step": 1038000 }, { "epoch": 0.37, "learning_rate": 1.251331616838947e-05, "loss": 0.7287, "step": 1038500 }, { "epoch": 0.37, "learning_rate": 1.2509711602269292e-05, "loss": 0.7256, "step": 1039000 }, { "epoch": 0.37, "learning_rate": 1.2506107036149113e-05, "loss": 0.7123, "step": 1039500 }, { "epoch": 0.37, "learning_rate": 1.2502502470028935e-05, "loss": 0.7378, "step": 1040000 }, { "epoch": 0.38, "learning_rate": 1.2498897903908755e-05, "loss": 0.7353, "step": 1040500 }, { "epoch": 0.38, "learning_rate": 1.2495293337788579e-05, "loss": 0.7193, "step": 1041000 }, { "epoch": 0.38, "learning_rate": 1.24916887716684e-05, "loss": 0.7248, "step": 1041500 }, { "epoch": 0.38, "learning_rate": 1.2488084205548221e-05, "loss": 0.7239, "step": 1042000 }, { "epoch": 0.38, "learning_rate": 1.2484479639428041e-05, "loss": 0.7295, "step": 1042500 }, { "epoch": 0.38, "learning_rate": 1.2480875073307865e-05, "loss": 0.7209, "step": 1043000 }, { "epoch": 0.38, "learning_rate": 1.2477270507187687e-05, "loss": 0.7253, "step": 1043500 }, { "epoch": 0.38, "learning_rate": 1.2473665941067507e-05, "loss": 0.7134, "step": 1044000 }, { "epoch": 0.38, "learning_rate": 1.247006137494733e-05, "loss": 0.76, "step": 1044500 }, { "epoch": 0.38, "learning_rate": 1.246645680882715e-05, "loss": 0.7335, "step": 1045000 }, { "epoch": 0.38, "learning_rate": 1.2462852242706973e-05, "loss": 0.7285, "step": 1045500 }, { "epoch": 0.38, "learning_rate": 1.2459247676586794e-05, "loss": 0.7344, "step": 1046000 }, { "epoch": 0.38, "learning_rate": 1.2455643110466616e-05, "loss": 0.7308, "step": 1046500 }, { "epoch": 0.38, "learning_rate": 1.2452038544346436e-05, "loss": 0.6914, "step": 1047000 }, { "epoch": 0.38, "learning_rate": 1.244843397822626e-05, "loss": 0.7575, "step": 1047500 }, { "epoch": 0.38, "learning_rate": 1.244482941210608e-05, "loss": 0.7073, "step": 1048000 }, { "epoch": 0.38, "learning_rate": 1.2441224845985902e-05, "loss": 0.7261, "step": 1048500 }, { "epoch": 0.38, "learning_rate": 1.2437620279865722e-05, "loss": 0.7316, "step": 1049000 }, { "epoch": 0.38, "learning_rate": 1.2434015713745546e-05, "loss": 0.7462, "step": 1049500 }, { "epoch": 0.38, "learning_rate": 1.2430411147625368e-05, "loss": 0.7477, "step": 1050000 }, { "epoch": 0.38, "learning_rate": 1.2426806581505188e-05, "loss": 0.7095, "step": 1050500 }, { "epoch": 0.38, "learning_rate": 1.242320201538501e-05, "loss": 0.7338, "step": 1051000 }, { "epoch": 0.38, "learning_rate": 1.241959744926483e-05, "loss": 0.7336, "step": 1051500 }, { "epoch": 0.38, "learning_rate": 1.2415992883144655e-05, "loss": 0.7171, "step": 1052000 }, { "epoch": 0.38, "learning_rate": 1.2412388317024475e-05, "loss": 0.7123, "step": 1052500 }, { "epoch": 0.38, "learning_rate": 1.2408783750904297e-05, "loss": 0.7225, "step": 1053000 }, { "epoch": 0.38, "learning_rate": 1.2405179184784117e-05, "loss": 0.702, "step": 1053500 }, { "epoch": 0.38, "learning_rate": 1.2401574618663941e-05, "loss": 0.7282, "step": 1054000 }, { "epoch": 0.38, "learning_rate": 1.2397970052543761e-05, "loss": 0.7358, "step": 1054500 }, { "epoch": 0.38, "learning_rate": 1.2394365486423583e-05, "loss": 0.7236, "step": 1055000 }, { "epoch": 0.38, "learning_rate": 1.2390760920303404e-05, "loss": 0.6972, "step": 1055500 }, { "epoch": 0.38, "learning_rate": 1.2387156354183226e-05, "loss": 0.733, "step": 1056000 }, { "epoch": 0.38, "learning_rate": 1.238355178806305e-05, "loss": 0.7536, "step": 1056500 }, { "epoch": 0.38, "learning_rate": 1.237994722194287e-05, "loss": 0.7132, "step": 1057000 }, { "epoch": 0.38, "learning_rate": 1.237634265582269e-05, "loss": 0.755, "step": 1057500 }, { "epoch": 0.38, "learning_rate": 1.2372738089702512e-05, "loss": 0.759, "step": 1058000 }, { "epoch": 0.38, "learning_rate": 1.2369133523582336e-05, "loss": 0.7529, "step": 1058500 }, { "epoch": 0.38, "learning_rate": 1.2365528957462156e-05, "loss": 0.7242, "step": 1059000 }, { "epoch": 0.38, "learning_rate": 1.2361924391341978e-05, "loss": 0.7228, "step": 1059500 }, { "epoch": 0.38, "learning_rate": 1.2358319825221798e-05, "loss": 0.7066, "step": 1060000 }, { "epoch": 0.38, "learning_rate": 1.235471525910162e-05, "loss": 0.7308, "step": 1060500 }, { "epoch": 0.38, "learning_rate": 1.2351110692981442e-05, "loss": 0.7372, "step": 1061000 }, { "epoch": 0.38, "learning_rate": 1.2347506126861264e-05, "loss": 0.7172, "step": 1061500 }, { "epoch": 0.38, "learning_rate": 1.2343901560741085e-05, "loss": 0.761, "step": 1062000 }, { "epoch": 0.38, "learning_rate": 1.2340296994620907e-05, "loss": 0.7345, "step": 1062500 }, { "epoch": 0.38, "learning_rate": 1.2336692428500729e-05, "loss": 0.7208, "step": 1063000 }, { "epoch": 0.38, "learning_rate": 1.233308786238055e-05, "loss": 0.7117, "step": 1063500 }, { "epoch": 0.38, "learning_rate": 1.2329483296260371e-05, "loss": 0.7209, "step": 1064000 }, { "epoch": 0.38, "learning_rate": 1.2325878730140193e-05, "loss": 0.715, "step": 1064500 }, { "epoch": 0.38, "learning_rate": 1.2322274164020017e-05, "loss": 0.7327, "step": 1065000 }, { "epoch": 0.38, "learning_rate": 1.2318669597899837e-05, "loss": 0.7161, "step": 1065500 }, { "epoch": 0.38, "learning_rate": 1.2315065031779659e-05, "loss": 0.6965, "step": 1066000 }, { "epoch": 0.38, "learning_rate": 1.231146046565948e-05, "loss": 0.7734, "step": 1066500 }, { "epoch": 0.38, "learning_rate": 1.23078558995393e-05, "loss": 0.7161, "step": 1067000 }, { "epoch": 0.38, "learning_rate": 1.2304251333419123e-05, "loss": 0.7377, "step": 1067500 }, { "epoch": 0.38, "learning_rate": 1.2300646767298945e-05, "loss": 0.7387, "step": 1068000 }, { "epoch": 0.39, "learning_rate": 1.2297042201178766e-05, "loss": 0.7168, "step": 1068500 }, { "epoch": 0.39, "learning_rate": 1.2293437635058588e-05, "loss": 0.7579, "step": 1069000 }, { "epoch": 0.39, "learning_rate": 1.228983306893841e-05, "loss": 0.7002, "step": 1069500 }, { "epoch": 0.39, "learning_rate": 1.2286228502818232e-05, "loss": 0.7236, "step": 1070000 }, { "epoch": 0.39, "learning_rate": 1.2282623936698052e-05, "loss": 0.732, "step": 1070500 }, { "epoch": 0.39, "learning_rate": 1.2279019370577874e-05, "loss": 0.7121, "step": 1071000 }, { "epoch": 0.39, "learning_rate": 1.2275414804457694e-05, "loss": 0.7002, "step": 1071500 }, { "epoch": 0.39, "learning_rate": 1.2271810238337518e-05, "loss": 0.7131, "step": 1072000 }, { "epoch": 0.39, "learning_rate": 1.2268205672217338e-05, "loss": 0.751, "step": 1072500 }, { "epoch": 0.39, "learning_rate": 1.226460110609716e-05, "loss": 0.7587, "step": 1073000 }, { "epoch": 0.39, "learning_rate": 1.226099653997698e-05, "loss": 0.7377, "step": 1073500 }, { "epoch": 0.39, "learning_rate": 1.2257391973856804e-05, "loss": 0.7303, "step": 1074000 }, { "epoch": 0.39, "learning_rate": 1.2253787407736626e-05, "loss": 0.7435, "step": 1074500 }, { "epoch": 0.39, "learning_rate": 1.2250182841616447e-05, "loss": 0.6799, "step": 1075000 }, { "epoch": 0.39, "learning_rate": 1.2246578275496269e-05, "loss": 0.7309, "step": 1075500 }, { "epoch": 0.39, "learning_rate": 1.2242973709376089e-05, "loss": 0.7729, "step": 1076000 }, { "epoch": 0.39, "learning_rate": 1.2239369143255913e-05, "loss": 0.7102, "step": 1076500 }, { "epoch": 0.39, "learning_rate": 1.2235764577135733e-05, "loss": 0.718, "step": 1077000 }, { "epoch": 0.39, "learning_rate": 1.2232160011015555e-05, "loss": 0.7404, "step": 1077500 }, { "epoch": 0.39, "learning_rate": 1.2228555444895376e-05, "loss": 0.7347, "step": 1078000 }, { "epoch": 0.39, "learning_rate": 1.22249508787752e-05, "loss": 0.7217, "step": 1078500 }, { "epoch": 0.39, "learning_rate": 1.222134631265502e-05, "loss": 0.7266, "step": 1079000 }, { "epoch": 0.39, "learning_rate": 1.2217741746534842e-05, "loss": 0.7344, "step": 1079500 }, { "epoch": 0.39, "learning_rate": 1.2214137180414662e-05, "loss": 0.7448, "step": 1080000 }, { "epoch": 0.39, "learning_rate": 1.2210532614294486e-05, "loss": 0.7161, "step": 1080500 }, { "epoch": 0.39, "learning_rate": 1.2206928048174308e-05, "loss": 0.7213, "step": 1081000 }, { "epoch": 0.39, "learning_rate": 1.2203323482054128e-05, "loss": 0.7196, "step": 1081500 }, { "epoch": 0.39, "learning_rate": 1.219971891593395e-05, "loss": 0.7359, "step": 1082000 }, { "epoch": 0.39, "learning_rate": 1.219611434981377e-05, "loss": 0.7199, "step": 1082500 }, { "epoch": 0.39, "learning_rate": 1.2192509783693594e-05, "loss": 0.7176, "step": 1083000 }, { "epoch": 0.39, "learning_rate": 1.2188905217573414e-05, "loss": 0.7351, "step": 1083500 }, { "epoch": 0.39, "learning_rate": 1.2185300651453236e-05, "loss": 0.7301, "step": 1084000 }, { "epoch": 0.39, "learning_rate": 1.2181696085333057e-05, "loss": 0.7325, "step": 1084500 }, { "epoch": 0.39, "learning_rate": 1.217809151921288e-05, "loss": 0.7169, "step": 1085000 }, { "epoch": 0.39, "learning_rate": 1.21744869530927e-05, "loss": 0.707, "step": 1085500 }, { "epoch": 0.39, "learning_rate": 1.2170882386972523e-05, "loss": 0.7029, "step": 1086000 }, { "epoch": 0.39, "learning_rate": 1.2167277820852343e-05, "loss": 0.6931, "step": 1086500 }, { "epoch": 0.39, "learning_rate": 1.2163673254732165e-05, "loss": 0.6774, "step": 1087000 }, { "epoch": 0.39, "learning_rate": 1.2160068688611989e-05, "loss": 0.725, "step": 1087500 }, { "epoch": 0.39, "learning_rate": 1.2156464122491809e-05, "loss": 0.6948, "step": 1088000 }, { "epoch": 0.39, "learning_rate": 1.215285955637163e-05, "loss": 0.7103, "step": 1088500 }, { "epoch": 0.39, "learning_rate": 1.2149254990251451e-05, "loss": 0.7076, "step": 1089000 }, { "epoch": 0.39, "learning_rate": 1.2145650424131275e-05, "loss": 0.7073, "step": 1089500 }, { "epoch": 0.39, "learning_rate": 1.2142045858011095e-05, "loss": 0.6979, "step": 1090000 }, { "epoch": 0.39, "learning_rate": 1.2138441291890917e-05, "loss": 0.6738, "step": 1090500 }, { "epoch": 0.39, "learning_rate": 1.2134836725770738e-05, "loss": 0.7144, "step": 1091000 }, { "epoch": 0.39, "learning_rate": 1.213123215965056e-05, "loss": 0.7489, "step": 1091500 }, { "epoch": 0.39, "learning_rate": 1.2127627593530382e-05, "loss": 0.7369, "step": 1092000 }, { "epoch": 0.39, "learning_rate": 1.2124023027410204e-05, "loss": 0.6934, "step": 1092500 }, { "epoch": 0.39, "learning_rate": 1.2120418461290024e-05, "loss": 0.7138, "step": 1093000 }, { "epoch": 0.39, "learning_rate": 1.2116813895169846e-05, "loss": 0.718, "step": 1093500 }, { "epoch": 0.39, "learning_rate": 1.2113209329049668e-05, "loss": 0.7443, "step": 1094000 }, { "epoch": 0.39, "learning_rate": 1.210960476292949e-05, "loss": 0.756, "step": 1094500 }, { "epoch": 0.39, "learning_rate": 1.210600019680931e-05, "loss": 0.7338, "step": 1095000 }, { "epoch": 0.39, "learning_rate": 1.2102395630689132e-05, "loss": 0.7442, "step": 1095500 }, { "epoch": 0.4, "learning_rate": 1.2098791064568953e-05, "loss": 0.7165, "step": 1096000 }, { "epoch": 0.4, "learning_rate": 1.2095186498448776e-05, "loss": 0.7174, "step": 1096500 }, { "epoch": 0.4, "learning_rate": 1.2091581932328598e-05, "loss": 0.7136, "step": 1097000 }, { "epoch": 0.4, "learning_rate": 1.2087977366208419e-05, "loss": 0.7042, "step": 1097500 }, { "epoch": 0.4, "learning_rate": 1.2084372800088239e-05, "loss": 0.7035, "step": 1098000 }, { "epoch": 0.4, "learning_rate": 1.2080768233968063e-05, "loss": 0.7603, "step": 1098500 }, { "epoch": 0.4, "learning_rate": 1.2077163667847885e-05, "loss": 0.7391, "step": 1099000 }, { "epoch": 0.4, "learning_rate": 1.2073559101727705e-05, "loss": 0.6979, "step": 1099500 }, { "epoch": 0.4, "learning_rate": 1.2069954535607527e-05, "loss": 0.7233, "step": 1100000 }, { "epoch": 0.4, "learning_rate": 1.206634996948735e-05, "loss": 0.7161, "step": 1100500 }, { "epoch": 0.4, "learning_rate": 1.2062745403367171e-05, "loss": 0.7465, "step": 1101000 }, { "epoch": 0.4, "learning_rate": 1.2059140837246992e-05, "loss": 0.7213, "step": 1101500 }, { "epoch": 0.4, "learning_rate": 1.2055536271126814e-05, "loss": 0.7226, "step": 1102000 }, { "epoch": 0.4, "learning_rate": 1.2051931705006634e-05, "loss": 0.7002, "step": 1102500 }, { "epoch": 0.4, "learning_rate": 1.2048327138886458e-05, "loss": 0.6744, "step": 1103000 }, { "epoch": 0.4, "learning_rate": 1.2044722572766278e-05, "loss": 0.724, "step": 1103500 }, { "epoch": 0.4, "learning_rate": 1.20411180066461e-05, "loss": 0.7443, "step": 1104000 }, { "epoch": 0.4, "learning_rate": 1.203751344052592e-05, "loss": 0.7231, "step": 1104500 }, { "epoch": 0.4, "learning_rate": 1.2033908874405744e-05, "loss": 0.7096, "step": 1105000 }, { "epoch": 0.4, "learning_rate": 1.2030304308285566e-05, "loss": 0.7146, "step": 1105500 }, { "epoch": 0.4, "learning_rate": 1.2026699742165386e-05, "loss": 0.7312, "step": 1106000 }, { "epoch": 0.4, "learning_rate": 1.2023095176045208e-05, "loss": 0.7461, "step": 1106500 }, { "epoch": 0.4, "learning_rate": 1.2019490609925029e-05, "loss": 0.6922, "step": 1107000 }, { "epoch": 0.4, "learning_rate": 1.2015886043804852e-05, "loss": 0.7339, "step": 1107500 }, { "epoch": 0.4, "learning_rate": 1.2012281477684673e-05, "loss": 0.7375, "step": 1108000 }, { "epoch": 0.4, "learning_rate": 1.2008676911564495e-05, "loss": 0.7193, "step": 1108500 }, { "epoch": 0.4, "learning_rate": 1.2005072345444315e-05, "loss": 0.7077, "step": 1109000 }, { "epoch": 0.4, "learning_rate": 1.2001467779324139e-05, "loss": 0.7506, "step": 1109500 }, { "epoch": 0.4, "learning_rate": 1.1997863213203959e-05, "loss": 0.7089, "step": 1110000 }, { "epoch": 0.4, "learning_rate": 1.1994258647083781e-05, "loss": 0.7112, "step": 1110500 }, { "epoch": 0.4, "learning_rate": 1.1990654080963601e-05, "loss": 0.6948, "step": 1111000 }, { "epoch": 0.4, "learning_rate": 1.1987049514843423e-05, "loss": 0.7175, "step": 1111500 }, { "epoch": 0.4, "learning_rate": 1.1983444948723247e-05, "loss": 0.7431, "step": 1112000 }, { "epoch": 0.4, "learning_rate": 1.1979840382603067e-05, "loss": 0.7503, "step": 1112500 }, { "epoch": 0.4, "learning_rate": 1.1976235816482888e-05, "loss": 0.7173, "step": 1113000 }, { "epoch": 0.4, "learning_rate": 1.197263125036271e-05, "loss": 0.7178, "step": 1113500 }, { "epoch": 0.4, "learning_rate": 1.1969026684242533e-05, "loss": 0.7221, "step": 1114000 }, { "epoch": 0.4, "learning_rate": 1.1965422118122354e-05, "loss": 0.7232, "step": 1114500 }, { "epoch": 0.4, "learning_rate": 1.1961817552002176e-05, "loss": 0.715, "step": 1115000 }, { "epoch": 0.4, "learning_rate": 1.1958212985881996e-05, "loss": 0.7501, "step": 1115500 }, { "epoch": 0.4, "learning_rate": 1.195460841976182e-05, "loss": 0.7276, "step": 1116000 }, { "epoch": 0.4, "learning_rate": 1.195100385364164e-05, "loss": 0.7139, "step": 1116500 }, { "epoch": 0.4, "learning_rate": 1.1947399287521462e-05, "loss": 0.7149, "step": 1117000 }, { "epoch": 0.4, "learning_rate": 1.1943794721401282e-05, "loss": 0.734, "step": 1117500 }, { "epoch": 0.4, "learning_rate": 1.1940190155281104e-05, "loss": 0.7308, "step": 1118000 }, { "epoch": 0.4, "learning_rate": 1.1936585589160926e-05, "loss": 0.6954, "step": 1118500 }, { "epoch": 0.4, "learning_rate": 1.1932981023040748e-05, "loss": 0.7089, "step": 1119000 }, { "epoch": 0.4, "learning_rate": 1.1929376456920569e-05, "loss": 0.7258, "step": 1119500 }, { "epoch": 0.4, "learning_rate": 1.192577189080039e-05, "loss": 0.7264, "step": 1120000 }, { "epoch": 0.4, "learning_rate": 1.1922167324680214e-05, "loss": 0.7109, "step": 1120500 }, { "epoch": 0.4, "learning_rate": 1.1918562758560035e-05, "loss": 0.7062, "step": 1121000 }, { "epoch": 0.4, "learning_rate": 1.1914958192439857e-05, "loss": 0.7416, "step": 1121500 }, { "epoch": 0.4, "learning_rate": 1.1911353626319677e-05, "loss": 0.7173, "step": 1122000 }, { "epoch": 0.4, "learning_rate": 1.1907749060199497e-05, "loss": 0.7135, "step": 1122500 }, { "epoch": 0.4, "learning_rate": 1.1904144494079321e-05, "loss": 0.727, "step": 1123000 }, { "epoch": 0.4, "learning_rate": 1.1900539927959143e-05, "loss": 0.7366, "step": 1123500 }, { "epoch": 0.41, "learning_rate": 1.1896935361838963e-05, "loss": 0.7125, "step": 1124000 }, { "epoch": 0.41, "learning_rate": 1.1893330795718786e-05, "loss": 0.7048, "step": 1124500 }, { "epoch": 0.41, "learning_rate": 1.1889726229598608e-05, "loss": 0.7269, "step": 1125000 }, { "epoch": 0.41, "learning_rate": 1.188612166347843e-05, "loss": 0.7022, "step": 1125500 }, { "epoch": 0.41, "learning_rate": 1.188251709735825e-05, "loss": 0.7154, "step": 1126000 }, { "epoch": 0.41, "learning_rate": 1.1878912531238072e-05, "loss": 0.7325, "step": 1126500 }, { "epoch": 0.41, "learning_rate": 1.1875307965117892e-05, "loss": 0.7176, "step": 1127000 }, { "epoch": 0.41, "learning_rate": 1.1871703398997716e-05, "loss": 0.7078, "step": 1127500 }, { "epoch": 0.41, "learning_rate": 1.1868098832877536e-05, "loss": 0.7363, "step": 1128000 }, { "epoch": 0.41, "learning_rate": 1.1864494266757358e-05, "loss": 0.7135, "step": 1128500 }, { "epoch": 0.41, "learning_rate": 1.1860889700637179e-05, "loss": 0.7085, "step": 1129000 }, { "epoch": 0.41, "learning_rate": 1.1857285134517002e-05, "loss": 0.7186, "step": 1129500 }, { "epoch": 0.41, "learning_rate": 1.1853680568396824e-05, "loss": 0.7176, "step": 1130000 }, { "epoch": 0.41, "learning_rate": 1.1850076002276645e-05, "loss": 0.6966, "step": 1130500 }, { "epoch": 0.41, "learning_rate": 1.1846471436156467e-05, "loss": 0.7065, "step": 1131000 }, { "epoch": 0.41, "learning_rate": 1.1842866870036289e-05, "loss": 0.7187, "step": 1131500 }, { "epoch": 0.41, "learning_rate": 1.183926230391611e-05, "loss": 0.7165, "step": 1132000 }, { "epoch": 0.41, "learning_rate": 1.1835657737795931e-05, "loss": 0.7303, "step": 1132500 }, { "epoch": 0.41, "learning_rate": 1.1832053171675753e-05, "loss": 0.7234, "step": 1133000 }, { "epoch": 0.41, "learning_rate": 1.1828448605555573e-05, "loss": 0.7477, "step": 1133500 }, { "epoch": 0.41, "learning_rate": 1.1824844039435397e-05, "loss": 0.7329, "step": 1134000 }, { "epoch": 0.41, "learning_rate": 1.1821239473315217e-05, "loss": 0.7138, "step": 1134500 }, { "epoch": 0.41, "learning_rate": 1.181763490719504e-05, "loss": 0.742, "step": 1135000 }, { "epoch": 0.41, "learning_rate": 1.181403034107486e-05, "loss": 0.7047, "step": 1135500 }, { "epoch": 0.41, "learning_rate": 1.1810425774954683e-05, "loss": 0.7375, "step": 1136000 }, { "epoch": 0.41, "learning_rate": 1.1806821208834505e-05, "loss": 0.737, "step": 1136500 }, { "epoch": 0.41, "learning_rate": 1.1803216642714326e-05, "loss": 0.7256, "step": 1137000 }, { "epoch": 0.41, "learning_rate": 1.1799612076594148e-05, "loss": 0.7197, "step": 1137500 }, { "epoch": 0.41, "learning_rate": 1.1796007510473968e-05, "loss": 0.7075, "step": 1138000 }, { "epoch": 0.41, "learning_rate": 1.1792402944353792e-05, "loss": 0.6927, "step": 1138500 }, { "epoch": 0.41, "learning_rate": 1.1788798378233612e-05, "loss": 0.7269, "step": 1139000 }, { "epoch": 0.41, "learning_rate": 1.1785193812113434e-05, "loss": 0.7181, "step": 1139500 }, { "epoch": 0.41, "learning_rate": 1.1781589245993254e-05, "loss": 0.7166, "step": 1140000 }, { "epoch": 0.41, "learning_rate": 1.1777984679873078e-05, "loss": 0.7195, "step": 1140500 }, { "epoch": 0.41, "learning_rate": 1.1774380113752898e-05, "loss": 0.745, "step": 1141000 }, { "epoch": 0.41, "learning_rate": 1.177077554763272e-05, "loss": 0.7288, "step": 1141500 }, { "epoch": 0.41, "learning_rate": 1.176717098151254e-05, "loss": 0.7211, "step": 1142000 }, { "epoch": 0.41, "learning_rate": 1.1763566415392363e-05, "loss": 0.712, "step": 1142500 }, { "epoch": 0.41, "learning_rate": 1.1759961849272186e-05, "loss": 0.7356, "step": 1143000 }, { "epoch": 0.41, "learning_rate": 1.1756357283152007e-05, "loss": 0.7172, "step": 1143500 }, { "epoch": 0.41, "learning_rate": 1.1752752717031827e-05, "loss": 0.704, "step": 1144000 }, { "epoch": 0.41, "learning_rate": 1.1749148150911649e-05, "loss": 0.7617, "step": 1144500 }, { "epoch": 0.41, "learning_rate": 1.1745543584791473e-05, "loss": 0.71, "step": 1145000 }, { "epoch": 0.41, "learning_rate": 1.1741939018671293e-05, "loss": 0.7047, "step": 1145500 }, { "epoch": 0.41, "learning_rate": 1.1738334452551115e-05, "loss": 0.722, "step": 1146000 }, { "epoch": 0.41, "learning_rate": 1.1734729886430935e-05, "loss": 0.7098, "step": 1146500 }, { "epoch": 0.41, "learning_rate": 1.173112532031076e-05, "loss": 0.7093, "step": 1147000 }, { "epoch": 0.41, "learning_rate": 1.172752075419058e-05, "loss": 0.7089, "step": 1147500 }, { "epoch": 0.41, "learning_rate": 1.1723916188070402e-05, "loss": 0.7217, "step": 1148000 }, { "epoch": 0.41, "learning_rate": 1.1720311621950222e-05, "loss": 0.7267, "step": 1148500 }, { "epoch": 0.41, "learning_rate": 1.1716707055830044e-05, "loss": 0.7177, "step": 1149000 }, { "epoch": 0.41, "learning_rate": 1.1713102489709866e-05, "loss": 0.7405, "step": 1149500 }, { "epoch": 0.41, "learning_rate": 1.1709497923589688e-05, "loss": 0.7571, "step": 1150000 }, { "epoch": 0.41, "learning_rate": 1.1705893357469508e-05, "loss": 0.7025, "step": 1150500 }, { "epoch": 0.41, "learning_rate": 1.170228879134933e-05, "loss": 0.7073, "step": 1151000 }, { "epoch": 0.42, "learning_rate": 1.1698684225229154e-05, "loss": 0.7251, "step": 1151500 }, { "epoch": 0.42, "learning_rate": 1.1695079659108974e-05, "loss": 0.7069, "step": 1152000 }, { "epoch": 0.42, "learning_rate": 1.1691475092988796e-05, "loss": 0.7148, "step": 1152500 }, { "epoch": 0.42, "learning_rate": 1.1687870526868617e-05, "loss": 0.7428, "step": 1153000 }, { "epoch": 0.42, "learning_rate": 1.1684265960748437e-05, "loss": 0.7277, "step": 1153500 }, { "epoch": 0.42, "learning_rate": 1.168066139462826e-05, "loss": 0.7102, "step": 1154000 }, { "epoch": 0.42, "learning_rate": 1.1677056828508083e-05, "loss": 0.681, "step": 1154500 }, { "epoch": 0.42, "learning_rate": 1.1673452262387903e-05, "loss": 0.7276, "step": 1155000 }, { "epoch": 0.42, "learning_rate": 1.1669847696267725e-05, "loss": 0.7174, "step": 1155500 }, { "epoch": 0.42, "learning_rate": 1.1666243130147547e-05, "loss": 0.6933, "step": 1156000 }, { "epoch": 0.42, "learning_rate": 1.1662638564027369e-05, "loss": 0.7038, "step": 1156500 }, { "epoch": 0.42, "learning_rate": 1.165903399790719e-05, "loss": 0.6948, "step": 1157000 }, { "epoch": 0.42, "learning_rate": 1.1655429431787011e-05, "loss": 0.7144, "step": 1157500 }, { "epoch": 0.42, "learning_rate": 1.1651824865666832e-05, "loss": 0.6954, "step": 1158000 }, { "epoch": 0.42, "learning_rate": 1.1648220299546655e-05, "loss": 0.7207, "step": 1158500 }, { "epoch": 0.42, "learning_rate": 1.1644615733426476e-05, "loss": 0.701, "step": 1159000 }, { "epoch": 0.42, "learning_rate": 1.1641011167306298e-05, "loss": 0.7311, "step": 1159500 }, { "epoch": 0.42, "learning_rate": 1.1637406601186118e-05, "loss": 0.6979, "step": 1160000 }, { "epoch": 0.42, "learning_rate": 1.1633802035065942e-05, "loss": 0.7149, "step": 1160500 }, { "epoch": 0.42, "learning_rate": 1.1630197468945764e-05, "loss": 0.74, "step": 1161000 }, { "epoch": 0.42, "learning_rate": 1.1626592902825584e-05, "loss": 0.7027, "step": 1161500 }, { "epoch": 0.42, "learning_rate": 1.1622988336705406e-05, "loss": 0.72, "step": 1162000 }, { "epoch": 0.42, "learning_rate": 1.1619383770585228e-05, "loss": 0.7196, "step": 1162500 }, { "epoch": 0.42, "learning_rate": 1.161577920446505e-05, "loss": 0.7504, "step": 1163000 }, { "epoch": 0.42, "learning_rate": 1.161217463834487e-05, "loss": 0.7072, "step": 1163500 }, { "epoch": 0.42, "learning_rate": 1.1608570072224692e-05, "loss": 0.7403, "step": 1164000 }, { "epoch": 0.42, "learning_rate": 1.1604965506104513e-05, "loss": 0.7146, "step": 1164500 }, { "epoch": 0.42, "learning_rate": 1.1601360939984336e-05, "loss": 0.7019, "step": 1165000 }, { "epoch": 0.42, "learning_rate": 1.1597756373864157e-05, "loss": 0.7174, "step": 1165500 }, { "epoch": 0.42, "learning_rate": 1.1594151807743979e-05, "loss": 0.713, "step": 1166000 }, { "epoch": 0.42, "learning_rate": 1.1590547241623799e-05, "loss": 0.712, "step": 1166500 }, { "epoch": 0.42, "learning_rate": 1.1586942675503623e-05, "loss": 0.6931, "step": 1167000 }, { "epoch": 0.42, "learning_rate": 1.1583338109383445e-05, "loss": 0.7023, "step": 1167500 }, { "epoch": 0.42, "learning_rate": 1.1579733543263265e-05, "loss": 0.7222, "step": 1168000 }, { "epoch": 0.42, "learning_rate": 1.1576128977143085e-05, "loss": 0.7155, "step": 1168500 }, { "epoch": 0.42, "learning_rate": 1.1572524411022907e-05, "loss": 0.7105, "step": 1169000 }, { "epoch": 0.42, "learning_rate": 1.1568919844902731e-05, "loss": 0.7083, "step": 1169500 }, { "epoch": 0.42, "learning_rate": 1.1565315278782551e-05, "loss": 0.6779, "step": 1170000 }, { "epoch": 0.42, "learning_rate": 1.1561710712662374e-05, "loss": 0.7213, "step": 1170500 }, { "epoch": 0.42, "learning_rate": 1.1558106146542194e-05, "loss": 0.6999, "step": 1171000 }, { "epoch": 0.42, "learning_rate": 1.1554501580422018e-05, "loss": 0.7243, "step": 1171500 }, { "epoch": 0.42, "learning_rate": 1.1550897014301838e-05, "loss": 0.735, "step": 1172000 }, { "epoch": 0.42, "learning_rate": 1.154729244818166e-05, "loss": 0.7229, "step": 1172500 }, { "epoch": 0.42, "learning_rate": 1.154368788206148e-05, "loss": 0.7121, "step": 1173000 }, { "epoch": 0.42, "learning_rate": 1.1540083315941302e-05, "loss": 0.684, "step": 1173500 }, { "epoch": 0.42, "learning_rate": 1.1536478749821124e-05, "loss": 0.7076, "step": 1174000 }, { "epoch": 0.42, "learning_rate": 1.1532874183700946e-05, "loss": 0.7106, "step": 1174500 }, { "epoch": 0.42, "learning_rate": 1.1529269617580767e-05, "loss": 0.7178, "step": 1175000 }, { "epoch": 0.42, "learning_rate": 1.1525665051460589e-05, "loss": 0.6993, "step": 1175500 }, { "epoch": 0.42, "learning_rate": 1.1522060485340412e-05, "loss": 0.7126, "step": 1176000 }, { "epoch": 0.42, "learning_rate": 1.1518455919220233e-05, "loss": 0.7038, "step": 1176500 }, { "epoch": 0.42, "learning_rate": 1.1514851353100055e-05, "loss": 0.7098, "step": 1177000 }, { "epoch": 0.42, "learning_rate": 1.1511246786979875e-05, "loss": 0.7069, "step": 1177500 }, { "epoch": 0.42, "learning_rate": 1.1507642220859695e-05, "loss": 0.7198, "step": 1178000 }, { "epoch": 0.42, "learning_rate": 1.1504037654739519e-05, "loss": 0.6971, "step": 1178500 }, { "epoch": 0.42, "learning_rate": 1.1500433088619341e-05, "loss": 0.723, "step": 1179000 }, { "epoch": 0.43, "learning_rate": 1.1496828522499161e-05, "loss": 0.72, "step": 1179500 }, { "epoch": 0.43, "learning_rate": 1.1493223956378983e-05, "loss": 0.7115, "step": 1180000 }, { "epoch": 0.43, "learning_rate": 1.1489619390258805e-05, "loss": 0.7152, "step": 1180500 }, { "epoch": 0.43, "learning_rate": 1.1486014824138627e-05, "loss": 0.7249, "step": 1181000 }, { "epoch": 0.43, "learning_rate": 1.1482410258018448e-05, "loss": 0.7186, "step": 1181500 }, { "epoch": 0.43, "learning_rate": 1.147880569189827e-05, "loss": 0.7118, "step": 1182000 }, { "epoch": 0.43, "learning_rate": 1.1475201125778093e-05, "loss": 0.7388, "step": 1182500 }, { "epoch": 0.43, "learning_rate": 1.1471596559657914e-05, "loss": 0.7094, "step": 1183000 }, { "epoch": 0.43, "learning_rate": 1.1467991993537734e-05, "loss": 0.7242, "step": 1183500 }, { "epoch": 0.43, "learning_rate": 1.1464387427417556e-05, "loss": 0.7431, "step": 1184000 }, { "epoch": 0.43, "learning_rate": 1.1460782861297376e-05, "loss": 0.7322, "step": 1184500 }, { "epoch": 0.43, "learning_rate": 1.14571782951772e-05, "loss": 0.7149, "step": 1185000 }, { "epoch": 0.43, "learning_rate": 1.1453573729057022e-05, "loss": 0.7116, "step": 1185500 }, { "epoch": 0.43, "learning_rate": 1.1449969162936842e-05, "loss": 0.7045, "step": 1186000 }, { "epoch": 0.43, "learning_rate": 1.1446364596816664e-05, "loss": 0.7161, "step": 1186500 }, { "epoch": 0.43, "learning_rate": 1.1442760030696486e-05, "loss": 0.7241, "step": 1187000 }, { "epoch": 0.43, "learning_rate": 1.1439155464576308e-05, "loss": 0.6797, "step": 1187500 }, { "epoch": 0.43, "learning_rate": 1.1435550898456129e-05, "loss": 0.6927, "step": 1188000 }, { "epoch": 0.43, "learning_rate": 1.143194633233595e-05, "loss": 0.6974, "step": 1188500 }, { "epoch": 0.43, "learning_rate": 1.1428341766215771e-05, "loss": 0.7226, "step": 1189000 }, { "epoch": 0.43, "learning_rate": 1.1424737200095595e-05, "loss": 0.7174, "step": 1189500 }, { "epoch": 0.43, "learning_rate": 1.1421132633975415e-05, "loss": 0.7055, "step": 1190000 }, { "epoch": 0.43, "learning_rate": 1.1417528067855237e-05, "loss": 0.742, "step": 1190500 }, { "epoch": 0.43, "learning_rate": 1.1413923501735057e-05, "loss": 0.6975, "step": 1191000 }, { "epoch": 0.43, "learning_rate": 1.1410318935614881e-05, "loss": 0.7058, "step": 1191500 }, { "epoch": 0.43, "learning_rate": 1.1406714369494703e-05, "loss": 0.7005, "step": 1192000 }, { "epoch": 0.43, "learning_rate": 1.1403109803374523e-05, "loss": 0.7095, "step": 1192500 }, { "epoch": 0.43, "learning_rate": 1.1399505237254345e-05, "loss": 0.6848, "step": 1193000 }, { "epoch": 0.43, "learning_rate": 1.1395900671134166e-05, "loss": 0.6775, "step": 1193500 }, { "epoch": 0.43, "learning_rate": 1.139229610501399e-05, "loss": 0.7483, "step": 1194000 }, { "epoch": 0.43, "learning_rate": 1.138869153889381e-05, "loss": 0.6997, "step": 1194500 }, { "epoch": 0.43, "learning_rate": 1.1385086972773632e-05, "loss": 0.7337, "step": 1195000 }, { "epoch": 0.43, "learning_rate": 1.1381482406653452e-05, "loss": 0.7377, "step": 1195500 }, { "epoch": 0.43, "learning_rate": 1.1377877840533276e-05, "loss": 0.7117, "step": 1196000 }, { "epoch": 0.43, "learning_rate": 1.1374273274413096e-05, "loss": 0.6704, "step": 1196500 }, { "epoch": 0.43, "learning_rate": 1.1370668708292918e-05, "loss": 0.7292, "step": 1197000 }, { "epoch": 0.43, "learning_rate": 1.1367064142172739e-05, "loss": 0.6796, "step": 1197500 }, { "epoch": 0.43, "learning_rate": 1.1363459576052562e-05, "loss": 0.7288, "step": 1198000 }, { "epoch": 0.43, "learning_rate": 1.1359855009932384e-05, "loss": 0.7294, "step": 1198500 }, { "epoch": 0.43, "learning_rate": 1.1356250443812205e-05, "loss": 0.7014, "step": 1199000 }, { "epoch": 0.43, "learning_rate": 1.1352645877692025e-05, "loss": 0.7181, "step": 1199500 }, { "epoch": 0.43, "learning_rate": 1.1349041311571847e-05, "loss": 0.6908, "step": 1200000 }, { "epoch": 0.43, "learning_rate": 1.134543674545167e-05, "loss": 0.7014, "step": 1200500 }, { "epoch": 0.43, "learning_rate": 1.1341832179331491e-05, "loss": 0.7405, "step": 1201000 }, { "epoch": 0.43, "learning_rate": 1.1338227613211313e-05, "loss": 0.7193, "step": 1201500 }, { "epoch": 0.43, "learning_rate": 1.1334623047091133e-05, "loss": 0.7224, "step": 1202000 }, { "epoch": 0.43, "learning_rate": 1.1331018480970957e-05, "loss": 0.7123, "step": 1202500 }, { "epoch": 0.43, "learning_rate": 1.1327413914850777e-05, "loss": 0.6933, "step": 1203000 }, { "epoch": 0.43, "learning_rate": 1.13238093487306e-05, "loss": 0.7171, "step": 1203500 }, { "epoch": 0.43, "learning_rate": 1.132020478261042e-05, "loss": 0.7312, "step": 1204000 }, { "epoch": 0.43, "learning_rate": 1.1316600216490242e-05, "loss": 0.6975, "step": 1204500 }, { "epoch": 0.43, "learning_rate": 1.1312995650370064e-05, "loss": 0.6956, "step": 1205000 }, { "epoch": 0.43, "learning_rate": 1.1309391084249886e-05, "loss": 0.7236, "step": 1205500 }, { "epoch": 0.43, "learning_rate": 1.1305786518129706e-05, "loss": 0.7122, "step": 1206000 }, { "epoch": 0.43, "learning_rate": 1.1302181952009528e-05, "loss": 0.7032, "step": 1206500 }, { "epoch": 0.44, "learning_rate": 1.1298577385889352e-05, "loss": 0.704, "step": 1207000 }, { "epoch": 0.44, "learning_rate": 1.1294972819769172e-05, "loss": 0.7477, "step": 1207500 }, { "epoch": 0.44, "learning_rate": 1.1291368253648994e-05, "loss": 0.7549, "step": 1208000 }, { "epoch": 0.44, "learning_rate": 1.1287763687528814e-05, "loss": 0.6979, "step": 1208500 }, { "epoch": 0.44, "learning_rate": 1.1284159121408635e-05, "loss": 0.7125, "step": 1209000 }, { "epoch": 0.44, "learning_rate": 1.1280554555288458e-05, "loss": 0.7295, "step": 1209500 }, { "epoch": 0.44, "learning_rate": 1.127694998916828e-05, "loss": 0.7079, "step": 1210000 }, { "epoch": 0.44, "learning_rate": 1.12733454230481e-05, "loss": 0.737, "step": 1210500 }, { "epoch": 0.44, "learning_rate": 1.1269740856927923e-05, "loss": 0.716, "step": 1211000 }, { "epoch": 0.44, "learning_rate": 1.1266136290807745e-05, "loss": 0.6828, "step": 1211500 }, { "epoch": 0.44, "learning_rate": 1.1262531724687567e-05, "loss": 0.6979, "step": 1212000 }, { "epoch": 0.44, "learning_rate": 1.1258927158567387e-05, "loss": 0.7024, "step": 1212500 }, { "epoch": 0.44, "learning_rate": 1.1255322592447209e-05, "loss": 0.7027, "step": 1213000 }, { "epoch": 0.44, "learning_rate": 1.1251718026327033e-05, "loss": 0.7116, "step": 1213500 }, { "epoch": 0.44, "learning_rate": 1.1248113460206853e-05, "loss": 0.7046, "step": 1214000 }, { "epoch": 0.44, "learning_rate": 1.1244508894086673e-05, "loss": 0.7317, "step": 1214500 }, { "epoch": 0.44, "learning_rate": 1.1240904327966495e-05, "loss": 0.6991, "step": 1215000 }, { "epoch": 0.44, "learning_rate": 1.1237299761846316e-05, "loss": 0.7101, "step": 1215500 }, { "epoch": 0.44, "learning_rate": 1.123369519572614e-05, "loss": 0.6803, "step": 1216000 }, { "epoch": 0.44, "learning_rate": 1.1230090629605961e-05, "loss": 0.6665, "step": 1216500 }, { "epoch": 0.44, "learning_rate": 1.1226486063485782e-05, "loss": 0.7189, "step": 1217000 }, { "epoch": 0.44, "learning_rate": 1.1222881497365604e-05, "loss": 0.7148, "step": 1217500 }, { "epoch": 0.44, "learning_rate": 1.1219276931245426e-05, "loss": 0.7156, "step": 1218000 }, { "epoch": 0.44, "learning_rate": 1.1215672365125248e-05, "loss": 0.7276, "step": 1218500 }, { "epoch": 0.44, "learning_rate": 1.1212067799005068e-05, "loss": 0.7, "step": 1219000 }, { "epoch": 0.44, "learning_rate": 1.120846323288489e-05, "loss": 0.7105, "step": 1219500 }, { "epoch": 0.44, "learning_rate": 1.120485866676471e-05, "loss": 0.747, "step": 1220000 }, { "epoch": 0.44, "learning_rate": 1.1201254100644534e-05, "loss": 0.7037, "step": 1220500 }, { "epoch": 0.44, "learning_rate": 1.1197649534524355e-05, "loss": 0.7125, "step": 1221000 }, { "epoch": 0.44, "learning_rate": 1.1194044968404177e-05, "loss": 0.7221, "step": 1221500 }, { "epoch": 0.44, "learning_rate": 1.1190440402283997e-05, "loss": 0.709, "step": 1222000 }, { "epoch": 0.44, "learning_rate": 1.118683583616382e-05, "loss": 0.6805, "step": 1222500 }, { "epoch": 0.44, "learning_rate": 1.1183231270043643e-05, "loss": 0.6866, "step": 1223000 }, { "epoch": 0.44, "learning_rate": 1.1179626703923463e-05, "loss": 0.7227, "step": 1223500 }, { "epoch": 0.44, "learning_rate": 1.1176022137803283e-05, "loss": 0.6994, "step": 1224000 }, { "epoch": 0.44, "learning_rate": 1.1172417571683105e-05, "loss": 0.6849, "step": 1224500 }, { "epoch": 0.44, "learning_rate": 1.1168813005562929e-05, "loss": 0.7146, "step": 1225000 }, { "epoch": 0.44, "learning_rate": 1.116520843944275e-05, "loss": 0.7264, "step": 1225500 }, { "epoch": 0.44, "learning_rate": 1.1161603873322571e-05, "loss": 0.6955, "step": 1226000 }, { "epoch": 0.44, "learning_rate": 1.1157999307202392e-05, "loss": 0.7091, "step": 1226500 }, { "epoch": 0.44, "learning_rate": 1.1154394741082215e-05, "loss": 0.7199, "step": 1227000 }, { "epoch": 0.44, "learning_rate": 1.1150790174962036e-05, "loss": 0.7303, "step": 1227500 }, { "epoch": 0.44, "learning_rate": 1.1147185608841858e-05, "loss": 0.7041, "step": 1228000 }, { "epoch": 0.44, "learning_rate": 1.1143581042721678e-05, "loss": 0.7066, "step": 1228500 }, { "epoch": 0.44, "learning_rate": 1.1139976476601502e-05, "loss": 0.7027, "step": 1229000 }, { "epoch": 0.44, "learning_rate": 1.1136371910481322e-05, "loss": 0.695, "step": 1229500 }, { "epoch": 0.44, "learning_rate": 1.1132767344361144e-05, "loss": 0.6989, "step": 1230000 }, { "epoch": 0.44, "learning_rate": 1.1129162778240964e-05, "loss": 0.7082, "step": 1230500 }, { "epoch": 0.44, "learning_rate": 1.1125558212120786e-05, "loss": 0.6705, "step": 1231000 }, { "epoch": 0.44, "learning_rate": 1.112195364600061e-05, "loss": 0.6852, "step": 1231500 }, { "epoch": 0.44, "learning_rate": 1.111834907988043e-05, "loss": 0.7148, "step": 1232000 }, { "epoch": 0.44, "learning_rate": 1.1114744513760252e-05, "loss": 0.7075, "step": 1232500 }, { "epoch": 0.44, "learning_rate": 1.1111139947640073e-05, "loss": 0.6968, "step": 1233000 }, { "epoch": 0.44, "learning_rate": 1.1107535381519896e-05, "loss": 0.719, "step": 1233500 }, { "epoch": 0.44, "learning_rate": 1.1103930815399717e-05, "loss": 0.6982, "step": 1234000 }, { "epoch": 0.44, "learning_rate": 1.1100326249279539e-05, "loss": 0.6837, "step": 1234500 }, { "epoch": 0.45, "learning_rate": 1.1096721683159359e-05, "loss": 0.686, "step": 1235000 }, { "epoch": 0.45, "learning_rate": 1.1093117117039181e-05, "loss": 0.7211, "step": 1235500 }, { "epoch": 0.45, "learning_rate": 1.1089512550919003e-05, "loss": 0.6955, "step": 1236000 }, { "epoch": 0.45, "learning_rate": 1.1085907984798825e-05, "loss": 0.7086, "step": 1236500 }, { "epoch": 0.45, "learning_rate": 1.1082303418678645e-05, "loss": 0.7279, "step": 1237000 }, { "epoch": 0.45, "learning_rate": 1.1078698852558467e-05, "loss": 0.7343, "step": 1237500 }, { "epoch": 0.45, "learning_rate": 1.1075094286438291e-05, "loss": 0.7181, "step": 1238000 }, { "epoch": 0.45, "learning_rate": 1.1071489720318111e-05, "loss": 0.7098, "step": 1238500 }, { "epoch": 0.45, "learning_rate": 1.1067885154197933e-05, "loss": 0.7128, "step": 1239000 }, { "epoch": 0.45, "learning_rate": 1.1064280588077754e-05, "loss": 0.7181, "step": 1239500 }, { "epoch": 0.45, "learning_rate": 1.1060676021957574e-05, "loss": 0.7064, "step": 1240000 }, { "epoch": 0.45, "learning_rate": 1.1057071455837398e-05, "loss": 0.7096, "step": 1240500 }, { "epoch": 0.45, "learning_rate": 1.105346688971722e-05, "loss": 0.7103, "step": 1241000 }, { "epoch": 0.45, "learning_rate": 1.104986232359704e-05, "loss": 0.6973, "step": 1241500 }, { "epoch": 0.45, "learning_rate": 1.1046257757476862e-05, "loss": 0.6993, "step": 1242000 }, { "epoch": 0.45, "learning_rate": 1.1042653191356684e-05, "loss": 0.7236, "step": 1242500 }, { "epoch": 0.45, "learning_rate": 1.1039048625236506e-05, "loss": 0.6864, "step": 1243000 }, { "epoch": 0.45, "learning_rate": 1.1035444059116327e-05, "loss": 0.7315, "step": 1243500 }, { "epoch": 0.45, "learning_rate": 1.1031839492996149e-05, "loss": 0.6839, "step": 1244000 }, { "epoch": 0.45, "learning_rate": 1.1028234926875972e-05, "loss": 0.6829, "step": 1244500 }, { "epoch": 0.45, "learning_rate": 1.1024630360755793e-05, "loss": 0.6949, "step": 1245000 }, { "epoch": 0.45, "learning_rate": 1.1021025794635613e-05, "loss": 0.6755, "step": 1245500 }, { "epoch": 0.45, "learning_rate": 1.1017421228515435e-05, "loss": 0.6912, "step": 1246000 }, { "epoch": 0.45, "learning_rate": 1.1013816662395255e-05, "loss": 0.6902, "step": 1246500 }, { "epoch": 0.45, "learning_rate": 1.1010212096275079e-05, "loss": 0.7155, "step": 1247000 }, { "epoch": 0.45, "learning_rate": 1.1006607530154901e-05, "loss": 0.7118, "step": 1247500 }, { "epoch": 0.45, "learning_rate": 1.1003002964034721e-05, "loss": 0.6739, "step": 1248000 }, { "epoch": 0.45, "learning_rate": 1.0999398397914543e-05, "loss": 0.7108, "step": 1248500 }, { "epoch": 0.45, "learning_rate": 1.0995793831794365e-05, "loss": 0.7327, "step": 1249000 }, { "epoch": 0.45, "learning_rate": 1.0992189265674187e-05, "loss": 0.7007, "step": 1249500 }, { "epoch": 0.45, "learning_rate": 1.0988584699554008e-05, "loss": 0.6995, "step": 1250000 }, { "epoch": 0.45, "learning_rate": 1.098498013343383e-05, "loss": 0.7248, "step": 1250500 }, { "epoch": 0.45, "learning_rate": 1.098137556731365e-05, "loss": 0.6581, "step": 1251000 }, { "epoch": 0.45, "learning_rate": 1.0977771001193474e-05, "loss": 0.6731, "step": 1251500 }, { "epoch": 0.45, "learning_rate": 1.0974166435073294e-05, "loss": 0.6869, "step": 1252000 }, { "epoch": 0.45, "learning_rate": 1.0970561868953116e-05, "loss": 0.7243, "step": 1252500 }, { "epoch": 0.45, "learning_rate": 1.0966957302832936e-05, "loss": 0.7207, "step": 1253000 }, { "epoch": 0.45, "learning_rate": 1.096335273671276e-05, "loss": 0.6875, "step": 1253500 }, { "epoch": 0.45, "learning_rate": 1.0959748170592582e-05, "loss": 0.6984, "step": 1254000 }, { "epoch": 0.45, "learning_rate": 1.0956143604472402e-05, "loss": 0.6936, "step": 1254500 }, { "epoch": 0.45, "learning_rate": 1.0952539038352223e-05, "loss": 0.7317, "step": 1255000 }, { "epoch": 0.45, "learning_rate": 1.0948934472232045e-05, "loss": 0.6793, "step": 1255500 }, { "epoch": 0.45, "learning_rate": 1.0945329906111868e-05, "loss": 0.6927, "step": 1256000 }, { "epoch": 0.45, "learning_rate": 1.0941725339991689e-05, "loss": 0.7026, "step": 1256500 }, { "epoch": 0.45, "learning_rate": 1.093812077387151e-05, "loss": 0.6982, "step": 1257000 }, { "epoch": 0.45, "learning_rate": 1.0934516207751331e-05, "loss": 0.6848, "step": 1257500 }, { "epoch": 0.45, "learning_rate": 1.0930911641631155e-05, "loss": 0.6842, "step": 1258000 }, { "epoch": 0.45, "learning_rate": 1.0927307075510975e-05, "loss": 0.7138, "step": 1258500 }, { "epoch": 0.45, "learning_rate": 1.0923702509390797e-05, "loss": 0.7073, "step": 1259000 }, { "epoch": 0.45, "learning_rate": 1.0920097943270617e-05, "loss": 0.7015, "step": 1259500 }, { "epoch": 0.45, "learning_rate": 1.091649337715044e-05, "loss": 0.7059, "step": 1260000 }, { "epoch": 0.45, "learning_rate": 1.0912888811030261e-05, "loss": 0.691, "step": 1260500 }, { "epoch": 0.45, "learning_rate": 1.0909284244910083e-05, "loss": 0.6756, "step": 1261000 }, { "epoch": 0.45, "learning_rate": 1.0905679678789904e-05, "loss": 0.7028, "step": 1261500 }, { "epoch": 0.45, "learning_rate": 1.0902075112669726e-05, "loss": 0.6995, "step": 1262000 }, { "epoch": 0.46, "learning_rate": 1.089847054654955e-05, "loss": 0.7196, "step": 1262500 }, { "epoch": 0.46, "learning_rate": 1.089486598042937e-05, "loss": 0.7364, "step": 1263000 }, { "epoch": 0.46, "learning_rate": 1.0891261414309192e-05, "loss": 0.7149, "step": 1263500 }, { "epoch": 0.46, "learning_rate": 1.0887656848189012e-05, "loss": 0.7016, "step": 1264000 }, { "epoch": 0.46, "learning_rate": 1.0884052282068836e-05, "loss": 0.6955, "step": 1264500 }, { "epoch": 0.46, "learning_rate": 1.0880447715948656e-05, "loss": 0.7141, "step": 1265000 }, { "epoch": 0.46, "learning_rate": 1.0876843149828478e-05, "loss": 0.6938, "step": 1265500 }, { "epoch": 0.46, "learning_rate": 1.0873238583708299e-05, "loss": 0.6897, "step": 1266000 }, { "epoch": 0.46, "learning_rate": 1.086963401758812e-05, "loss": 0.7265, "step": 1266500 }, { "epoch": 0.46, "learning_rate": 1.0866029451467943e-05, "loss": 0.7003, "step": 1267000 }, { "epoch": 0.46, "learning_rate": 1.0862424885347765e-05, "loss": 0.6869, "step": 1267500 }, { "epoch": 0.46, "learning_rate": 1.0858820319227585e-05, "loss": 0.6582, "step": 1268000 }, { "epoch": 0.46, "learning_rate": 1.0855215753107407e-05, "loss": 0.7125, "step": 1268500 }, { "epoch": 0.46, "learning_rate": 1.085161118698723e-05, "loss": 0.6769, "step": 1269000 }, { "epoch": 0.46, "learning_rate": 1.0848006620867051e-05, "loss": 0.7529, "step": 1269500 }, { "epoch": 0.46, "learning_rate": 1.0844402054746871e-05, "loss": 0.6814, "step": 1270000 }, { "epoch": 0.46, "learning_rate": 1.0840797488626693e-05, "loss": 0.7185, "step": 1270500 }, { "epoch": 0.46, "learning_rate": 1.0837192922506514e-05, "loss": 0.7153, "step": 1271000 }, { "epoch": 0.46, "learning_rate": 1.0833588356386337e-05, "loss": 0.7126, "step": 1271500 }, { "epoch": 0.46, "learning_rate": 1.082998379026616e-05, "loss": 0.714, "step": 1272000 }, { "epoch": 0.46, "learning_rate": 1.082637922414598e-05, "loss": 0.7229, "step": 1272500 }, { "epoch": 0.46, "learning_rate": 1.0822774658025802e-05, "loss": 0.7027, "step": 1273000 }, { "epoch": 0.46, "learning_rate": 1.0819170091905624e-05, "loss": 0.6965, "step": 1273500 }, { "epoch": 0.46, "learning_rate": 1.0815565525785446e-05, "loss": 0.696, "step": 1274000 }, { "epoch": 0.46, "learning_rate": 1.0811960959665266e-05, "loss": 0.6933, "step": 1274500 }, { "epoch": 0.46, "learning_rate": 1.0808356393545088e-05, "loss": 0.7079, "step": 1275000 }, { "epoch": 0.46, "learning_rate": 1.0804751827424908e-05, "loss": 0.6944, "step": 1275500 }, { "epoch": 0.46, "learning_rate": 1.0801147261304732e-05, "loss": 0.706, "step": 1276000 }, { "epoch": 0.46, "learning_rate": 1.0797542695184552e-05, "loss": 0.6979, "step": 1276500 }, { "epoch": 0.46, "learning_rate": 1.0793938129064374e-05, "loss": 0.7146, "step": 1277000 }, { "epoch": 0.46, "learning_rate": 1.0790333562944195e-05, "loss": 0.7231, "step": 1277500 }, { "epoch": 0.46, "learning_rate": 1.0786728996824018e-05, "loss": 0.7216, "step": 1278000 }, { "epoch": 0.46, "learning_rate": 1.078312443070384e-05, "loss": 0.7362, "step": 1278500 }, { "epoch": 0.46, "learning_rate": 1.077951986458366e-05, "loss": 0.6777, "step": 1279000 }, { "epoch": 0.46, "learning_rate": 1.0775915298463481e-05, "loss": 0.7136, "step": 1279500 }, { "epoch": 0.46, "learning_rate": 1.0772310732343305e-05, "loss": 0.7111, "step": 1280000 }, { "epoch": 0.46, "learning_rate": 1.0768706166223127e-05, "loss": 0.7099, "step": 1280500 }, { "epoch": 0.46, "learning_rate": 1.0765101600102947e-05, "loss": 0.6826, "step": 1281000 }, { "epoch": 0.46, "learning_rate": 1.0761497033982769e-05, "loss": 0.7227, "step": 1281500 }, { "epoch": 0.46, "learning_rate": 1.075789246786259e-05, "loss": 0.7059, "step": 1282000 }, { "epoch": 0.46, "learning_rate": 1.0754287901742413e-05, "loss": 0.7192, "step": 1282500 }, { "epoch": 0.46, "learning_rate": 1.0750683335622233e-05, "loss": 0.7068, "step": 1283000 }, { "epoch": 0.46, "learning_rate": 1.0747078769502055e-05, "loss": 0.7177, "step": 1283500 }, { "epoch": 0.46, "learning_rate": 1.0743474203381876e-05, "loss": 0.7178, "step": 1284000 }, { "epoch": 0.46, "learning_rate": 1.07398696372617e-05, "loss": 0.6819, "step": 1284500 }, { "epoch": 0.46, "learning_rate": 1.073626507114152e-05, "loss": 0.6888, "step": 1285000 }, { "epoch": 0.46, "learning_rate": 1.0732660505021342e-05, "loss": 0.7018, "step": 1285500 }, { "epoch": 0.46, "learning_rate": 1.0729055938901162e-05, "loss": 0.7066, "step": 1286000 }, { "epoch": 0.46, "learning_rate": 1.0725451372780984e-05, "loss": 0.6896, "step": 1286500 }, { "epoch": 0.46, "learning_rate": 1.0721846806660808e-05, "loss": 0.6899, "step": 1287000 }, { "epoch": 0.46, "learning_rate": 1.0718242240540628e-05, "loss": 0.6965, "step": 1287500 }, { "epoch": 0.46, "learning_rate": 1.071463767442045e-05, "loss": 0.7018, "step": 1288000 }, { "epoch": 0.46, "learning_rate": 1.071103310830027e-05, "loss": 0.6912, "step": 1288500 }, { "epoch": 0.46, "learning_rate": 1.0707428542180094e-05, "loss": 0.71, "step": 1289000 }, { "epoch": 0.46, "learning_rate": 1.0703823976059915e-05, "loss": 0.7261, "step": 1289500 }, { "epoch": 0.46, "learning_rate": 1.0700219409939737e-05, "loss": 0.6852, "step": 1290000 }, { "epoch": 0.47, "learning_rate": 1.0696614843819557e-05, "loss": 0.6934, "step": 1290500 }, { "epoch": 0.47, "learning_rate": 1.0693010277699379e-05, "loss": 0.708, "step": 1291000 }, { "epoch": 0.47, "learning_rate": 1.0689405711579201e-05, "loss": 0.707, "step": 1291500 }, { "epoch": 0.47, "learning_rate": 1.0685801145459023e-05, "loss": 0.7012, "step": 1292000 }, { "epoch": 0.47, "learning_rate": 1.0682196579338843e-05, "loss": 0.7016, "step": 1292500 }, { "epoch": 0.47, "learning_rate": 1.0678592013218665e-05, "loss": 0.6943, "step": 1293000 }, { "epoch": 0.47, "learning_rate": 1.0674987447098489e-05, "loss": 0.701, "step": 1293500 }, { "epoch": 0.47, "learning_rate": 1.067138288097831e-05, "loss": 0.7003, "step": 1294000 }, { "epoch": 0.47, "learning_rate": 1.0667778314858131e-05, "loss": 0.6869, "step": 1294500 }, { "epoch": 0.47, "learning_rate": 1.0664173748737952e-05, "loss": 0.6945, "step": 1295000 }, { "epoch": 0.47, "learning_rate": 1.0660569182617775e-05, "loss": 0.6766, "step": 1295500 }, { "epoch": 0.47, "learning_rate": 1.0656964616497596e-05, "loss": 0.6868, "step": 1296000 }, { "epoch": 0.47, "learning_rate": 1.0653360050377418e-05, "loss": 0.7213, "step": 1296500 }, { "epoch": 0.47, "learning_rate": 1.0649755484257238e-05, "loss": 0.6931, "step": 1297000 }, { "epoch": 0.47, "learning_rate": 1.064615091813706e-05, "loss": 0.6834, "step": 1297500 }, { "epoch": 0.47, "learning_rate": 1.0642546352016882e-05, "loss": 0.7161, "step": 1298000 }, { "epoch": 0.47, "learning_rate": 1.0638941785896704e-05, "loss": 0.7014, "step": 1298500 }, { "epoch": 0.47, "learning_rate": 1.0635337219776524e-05, "loss": 0.687, "step": 1299000 }, { "epoch": 0.47, "learning_rate": 1.0631732653656346e-05, "loss": 0.6967, "step": 1299500 }, { "epoch": 0.47, "learning_rate": 1.062812808753617e-05, "loss": 0.7034, "step": 1300000 }, { "epoch": 0.47, "learning_rate": 1.062452352141599e-05, "loss": 0.6998, "step": 1300500 }, { "epoch": 0.47, "learning_rate": 1.062091895529581e-05, "loss": 0.7142, "step": 1301000 }, { "epoch": 0.47, "learning_rate": 1.0617314389175633e-05, "loss": 0.7207, "step": 1301500 }, { "epoch": 0.47, "learning_rate": 1.0613709823055453e-05, "loss": 0.7123, "step": 1302000 }, { "epoch": 0.47, "learning_rate": 1.0610105256935277e-05, "loss": 0.7088, "step": 1302500 }, { "epoch": 0.47, "learning_rate": 1.0606500690815099e-05, "loss": 0.6877, "step": 1303000 }, { "epoch": 0.47, "learning_rate": 1.0602896124694919e-05, "loss": 0.6955, "step": 1303500 }, { "epoch": 0.47, "learning_rate": 1.0599291558574741e-05, "loss": 0.7035, "step": 1304000 }, { "epoch": 0.47, "learning_rate": 1.0595686992454563e-05, "loss": 0.7264, "step": 1304500 }, { "epoch": 0.47, "learning_rate": 1.0592082426334385e-05, "loss": 0.7071, "step": 1305000 }, { "epoch": 0.47, "learning_rate": 1.0588477860214205e-05, "loss": 0.6708, "step": 1305500 }, { "epoch": 0.47, "learning_rate": 1.0584873294094027e-05, "loss": 0.7424, "step": 1306000 }, { "epoch": 0.47, "learning_rate": 1.0581268727973848e-05, "loss": 0.6854, "step": 1306500 }, { "epoch": 0.47, "learning_rate": 1.0577664161853671e-05, "loss": 0.7023, "step": 1307000 }, { "epoch": 0.47, "learning_rate": 1.0574059595733492e-05, "loss": 0.6881, "step": 1307500 }, { "epoch": 0.47, "learning_rate": 1.0570455029613314e-05, "loss": 0.6949, "step": 1308000 }, { "epoch": 0.47, "learning_rate": 1.0566850463493134e-05, "loss": 0.7143, "step": 1308500 }, { "epoch": 0.47, "learning_rate": 1.0563245897372958e-05, "loss": 0.6574, "step": 1309000 }, { "epoch": 0.47, "learning_rate": 1.055964133125278e-05, "loss": 0.6858, "step": 1309500 }, { "epoch": 0.47, "learning_rate": 1.05560367651326e-05, "loss": 0.7084, "step": 1310000 }, { "epoch": 0.47, "learning_rate": 1.055243219901242e-05, "loss": 0.6921, "step": 1310500 }, { "epoch": 0.47, "learning_rate": 1.0548827632892244e-05, "loss": 0.6897, "step": 1311000 }, { "epoch": 0.47, "learning_rate": 1.0545223066772066e-05, "loss": 0.73, "step": 1311500 }, { "epoch": 0.47, "learning_rate": 1.0541618500651886e-05, "loss": 0.6929, "step": 1312000 }, { "epoch": 0.47, "learning_rate": 1.0538013934531709e-05, "loss": 0.7, "step": 1312500 }, { "epoch": 0.47, "learning_rate": 1.0534409368411529e-05, "loss": 0.7071, "step": 1313000 }, { "epoch": 0.47, "learning_rate": 1.0530804802291353e-05, "loss": 0.721, "step": 1313500 }, { "epoch": 0.47, "learning_rate": 1.0527200236171173e-05, "loss": 0.6879, "step": 1314000 }, { "epoch": 0.47, "learning_rate": 1.0523595670050995e-05, "loss": 0.6857, "step": 1314500 }, { "epoch": 0.47, "learning_rate": 1.0519991103930815e-05, "loss": 0.6802, "step": 1315000 }, { "epoch": 0.47, "learning_rate": 1.0516386537810639e-05, "loss": 0.695, "step": 1315500 }, { "epoch": 0.47, "learning_rate": 1.051278197169046e-05, "loss": 0.7169, "step": 1316000 }, { "epoch": 0.47, "learning_rate": 1.0509177405570281e-05, "loss": 0.6695, "step": 1316500 }, { "epoch": 0.47, "learning_rate": 1.0505572839450102e-05, "loss": 0.7142, "step": 1317000 }, { "epoch": 0.47, "learning_rate": 1.0501968273329924e-05, "loss": 0.6965, "step": 1317500 }, { "epoch": 0.48, "learning_rate": 1.0498363707209747e-05, "loss": 0.6999, "step": 1318000 }, { "epoch": 0.48, "learning_rate": 1.0494759141089568e-05, "loss": 0.7006, "step": 1318500 }, { "epoch": 0.48, "learning_rate": 1.049115457496939e-05, "loss": 0.7152, "step": 1319000 }, { "epoch": 0.48, "learning_rate": 1.048755000884921e-05, "loss": 0.7112, "step": 1319500 }, { "epoch": 0.48, "learning_rate": 1.0483945442729034e-05, "loss": 0.7078, "step": 1320000 }, { "epoch": 0.48, "learning_rate": 1.0480340876608854e-05, "loss": 0.6962, "step": 1320500 }, { "epoch": 0.48, "learning_rate": 1.0476736310488676e-05, "loss": 0.7101, "step": 1321000 }, { "epoch": 0.48, "learning_rate": 1.0473131744368496e-05, "loss": 0.693, "step": 1321500 }, { "epoch": 0.48, "learning_rate": 1.0469527178248318e-05, "loss": 0.6962, "step": 1322000 }, { "epoch": 0.48, "learning_rate": 1.046592261212814e-05, "loss": 0.6811, "step": 1322500 }, { "epoch": 0.48, "learning_rate": 1.0462318046007962e-05, "loss": 0.6836, "step": 1323000 }, { "epoch": 0.48, "learning_rate": 1.0458713479887783e-05, "loss": 0.7002, "step": 1323500 }, { "epoch": 0.48, "learning_rate": 1.0455108913767605e-05, "loss": 0.6802, "step": 1324000 }, { "epoch": 0.48, "learning_rate": 1.0451504347647428e-05, "loss": 0.7057, "step": 1324500 }, { "epoch": 0.48, "learning_rate": 1.0447899781527249e-05, "loss": 0.6857, "step": 1325000 }, { "epoch": 0.48, "learning_rate": 1.0444295215407069e-05, "loss": 0.6968, "step": 1325500 }, { "epoch": 0.48, "learning_rate": 1.0440690649286891e-05, "loss": 0.7096, "step": 1326000 }, { "epoch": 0.48, "learning_rate": 1.0437086083166711e-05, "loss": 0.7216, "step": 1326500 }, { "epoch": 0.48, "learning_rate": 1.0433481517046535e-05, "loss": 0.6943, "step": 1327000 }, { "epoch": 0.48, "learning_rate": 1.0429876950926357e-05, "loss": 0.7017, "step": 1327500 }, { "epoch": 0.48, "learning_rate": 1.0426272384806177e-05, "loss": 0.6795, "step": 1328000 }, { "epoch": 0.48, "learning_rate": 1.0422667818686e-05, "loss": 0.6561, "step": 1328500 }, { "epoch": 0.48, "learning_rate": 1.0419063252565821e-05, "loss": 0.6834, "step": 1329000 }, { "epoch": 0.48, "learning_rate": 1.0415458686445643e-05, "loss": 0.6925, "step": 1329500 }, { "epoch": 0.48, "learning_rate": 1.0411854120325464e-05, "loss": 0.6771, "step": 1330000 }, { "epoch": 0.48, "learning_rate": 1.0408249554205286e-05, "loss": 0.6939, "step": 1330500 }, { "epoch": 0.48, "learning_rate": 1.0404644988085108e-05, "loss": 0.713, "step": 1331000 }, { "epoch": 0.48, "learning_rate": 1.040104042196493e-05, "loss": 0.7162, "step": 1331500 }, { "epoch": 0.48, "learning_rate": 1.039743585584475e-05, "loss": 0.7152, "step": 1332000 }, { "epoch": 0.48, "learning_rate": 1.0393831289724572e-05, "loss": 0.7471, "step": 1332500 }, { "epoch": 0.48, "learning_rate": 1.0390226723604392e-05, "loss": 0.7071, "step": 1333000 }, { "epoch": 0.48, "learning_rate": 1.0386622157484216e-05, "loss": 0.7195, "step": 1333500 }, { "epoch": 0.48, "learning_rate": 1.0383017591364038e-05, "loss": 0.6798, "step": 1334000 }, { "epoch": 0.48, "learning_rate": 1.0379413025243858e-05, "loss": 0.6558, "step": 1334500 }, { "epoch": 0.48, "learning_rate": 1.0375808459123679e-05, "loss": 0.6913, "step": 1335000 }, { "epoch": 0.48, "learning_rate": 1.0372203893003503e-05, "loss": 0.706, "step": 1335500 }, { "epoch": 0.48, "learning_rate": 1.0368599326883325e-05, "loss": 0.6952, "step": 1336000 }, { "epoch": 0.48, "learning_rate": 1.0364994760763145e-05, "loss": 0.707, "step": 1336500 }, { "epoch": 0.48, "learning_rate": 1.0361390194642967e-05, "loss": 0.7158, "step": 1337000 }, { "epoch": 0.48, "learning_rate": 1.0357785628522787e-05, "loss": 0.7052, "step": 1337500 }, { "epoch": 0.48, "learning_rate": 1.0354181062402611e-05, "loss": 0.6806, "step": 1338000 }, { "epoch": 0.48, "learning_rate": 1.0350576496282431e-05, "loss": 0.6729, "step": 1338500 }, { "epoch": 0.48, "learning_rate": 1.0346971930162253e-05, "loss": 0.6878, "step": 1339000 }, { "epoch": 0.48, "learning_rate": 1.0343367364042074e-05, "loss": 0.7194, "step": 1339500 }, { "epoch": 0.48, "learning_rate": 1.0339762797921897e-05, "loss": 0.7043, "step": 1340000 }, { "epoch": 0.48, "learning_rate": 1.0336158231801718e-05, "loss": 0.7031, "step": 1340500 }, { "epoch": 0.48, "learning_rate": 1.033255366568154e-05, "loss": 0.7083, "step": 1341000 }, { "epoch": 0.48, "learning_rate": 1.032894909956136e-05, "loss": 0.702, "step": 1341500 }, { "epoch": 0.48, "learning_rate": 1.0325344533441182e-05, "loss": 0.6918, "step": 1342000 }, { "epoch": 0.48, "learning_rate": 1.0321739967321006e-05, "loss": 0.6677, "step": 1342500 }, { "epoch": 0.48, "learning_rate": 1.0318135401200826e-05, "loss": 0.7064, "step": 1343000 }, { "epoch": 0.48, "learning_rate": 1.0314530835080648e-05, "loss": 0.693, "step": 1343500 }, { "epoch": 0.48, "learning_rate": 1.0310926268960468e-05, "loss": 0.6894, "step": 1344000 }, { "epoch": 0.48, "learning_rate": 1.0307321702840292e-05, "loss": 0.7026, "step": 1344500 }, { "epoch": 0.48, "learning_rate": 1.0303717136720112e-05, "loss": 0.7198, "step": 1345000 }, { "epoch": 0.48, "learning_rate": 1.0300112570599934e-05, "loss": 0.7049, "step": 1345500 }, { "epoch": 0.49, "learning_rate": 1.0296508004479755e-05, "loss": 0.6992, "step": 1346000 }, { "epoch": 0.49, "learning_rate": 1.0292903438359578e-05, "loss": 0.7077, "step": 1346500 }, { "epoch": 0.49, "learning_rate": 1.0289298872239399e-05, "loss": 0.7078, "step": 1347000 }, { "epoch": 0.49, "learning_rate": 1.028569430611922e-05, "loss": 0.6577, "step": 1347500 }, { "epoch": 0.49, "learning_rate": 1.0282089739999041e-05, "loss": 0.719, "step": 1348000 }, { "epoch": 0.49, "learning_rate": 1.0278485173878863e-05, "loss": 0.7089, "step": 1348500 }, { "epoch": 0.49, "learning_rate": 1.0274880607758687e-05, "loss": 0.706, "step": 1349000 }, { "epoch": 0.49, "learning_rate": 1.0271276041638507e-05, "loss": 0.6972, "step": 1349500 }, { "epoch": 0.49, "learning_rate": 1.0267671475518329e-05, "loss": 0.6909, "step": 1350000 }, { "epoch": 0.49, "learning_rate": 1.026406690939815e-05, "loss": 0.6702, "step": 1350500 }, { "epoch": 0.49, "learning_rate": 1.0260462343277973e-05, "loss": 0.7322, "step": 1351000 }, { "epoch": 0.49, "learning_rate": 1.0256857777157793e-05, "loss": 0.7078, "step": 1351500 }, { "epoch": 0.49, "learning_rate": 1.0253253211037615e-05, "loss": 0.671, "step": 1352000 }, { "epoch": 0.49, "learning_rate": 1.0249648644917436e-05, "loss": 0.7147, "step": 1352500 }, { "epoch": 0.49, "learning_rate": 1.0246044078797258e-05, "loss": 0.6967, "step": 1353000 }, { "epoch": 0.49, "learning_rate": 1.024243951267708e-05, "loss": 0.6888, "step": 1353500 }, { "epoch": 0.49, "learning_rate": 1.0238834946556902e-05, "loss": 0.7259, "step": 1354000 }, { "epoch": 0.49, "learning_rate": 1.0235230380436722e-05, "loss": 0.7278, "step": 1354500 }, { "epoch": 0.49, "learning_rate": 1.0231625814316544e-05, "loss": 0.6896, "step": 1355000 }, { "epoch": 0.49, "learning_rate": 1.0228021248196368e-05, "loss": 0.7053, "step": 1355500 }, { "epoch": 0.49, "learning_rate": 1.0224416682076188e-05, "loss": 0.6793, "step": 1356000 }, { "epoch": 0.49, "learning_rate": 1.0220812115956008e-05, "loss": 0.7147, "step": 1356500 }, { "epoch": 0.49, "learning_rate": 1.021720754983583e-05, "loss": 0.7048, "step": 1357000 }, { "epoch": 0.49, "learning_rate": 1.021360298371565e-05, "loss": 0.6832, "step": 1357500 }, { "epoch": 0.49, "learning_rate": 1.0209998417595474e-05, "loss": 0.7149, "step": 1358000 }, { "epoch": 0.49, "learning_rate": 1.0206393851475296e-05, "loss": 0.7024, "step": 1358500 }, { "epoch": 0.49, "learning_rate": 1.0202789285355117e-05, "loss": 0.7048, "step": 1359000 }, { "epoch": 0.49, "learning_rate": 1.0199184719234939e-05, "loss": 0.7096, "step": 1359500 }, { "epoch": 0.49, "learning_rate": 1.019558015311476e-05, "loss": 0.7028, "step": 1360000 }, { "epoch": 0.49, "learning_rate": 1.0191975586994583e-05, "loss": 0.7018, "step": 1360500 }, { "epoch": 0.49, "learning_rate": 1.0188371020874403e-05, "loss": 0.6797, "step": 1361000 }, { "epoch": 0.49, "learning_rate": 1.0184766454754225e-05, "loss": 0.698, "step": 1361500 }, { "epoch": 0.49, "learning_rate": 1.0181161888634047e-05, "loss": 0.7225, "step": 1362000 }, { "epoch": 0.49, "learning_rate": 1.017755732251387e-05, "loss": 0.6867, "step": 1362500 }, { "epoch": 0.49, "learning_rate": 1.017395275639369e-05, "loss": 0.6949, "step": 1363000 }, { "epoch": 0.49, "learning_rate": 1.0170348190273512e-05, "loss": 0.7074, "step": 1363500 }, { "epoch": 0.49, "learning_rate": 1.0166743624153332e-05, "loss": 0.6877, "step": 1364000 }, { "epoch": 0.49, "learning_rate": 1.0163139058033156e-05, "loss": 0.6792, "step": 1364500 }, { "epoch": 0.49, "learning_rate": 1.0159534491912978e-05, "loss": 0.7315, "step": 1365000 }, { "epoch": 0.49, "learning_rate": 1.0155929925792798e-05, "loss": 0.6729, "step": 1365500 }, { "epoch": 0.49, "learning_rate": 1.0152325359672618e-05, "loss": 0.6914, "step": 1366000 }, { "epoch": 0.49, "learning_rate": 1.0148720793552442e-05, "loss": 0.6835, "step": 1366500 }, { "epoch": 0.49, "learning_rate": 1.0145116227432264e-05, "loss": 0.6577, "step": 1367000 }, { "epoch": 0.49, "learning_rate": 1.0141511661312084e-05, "loss": 0.685, "step": 1367500 }, { "epoch": 0.49, "learning_rate": 1.0137907095191906e-05, "loss": 0.7174, "step": 1368000 }, { "epoch": 0.49, "learning_rate": 1.0134302529071727e-05, "loss": 0.7176, "step": 1368500 }, { "epoch": 0.49, "learning_rate": 1.013069796295155e-05, "loss": 0.6974, "step": 1369000 }, { "epoch": 0.49, "learning_rate": 1.012709339683137e-05, "loss": 0.7033, "step": 1369500 }, { "epoch": 0.49, "learning_rate": 1.0123488830711193e-05, "loss": 0.6806, "step": 1370000 }, { "epoch": 0.49, "learning_rate": 1.0119884264591013e-05, "loss": 0.6841, "step": 1370500 }, { "epoch": 0.49, "learning_rate": 1.0116279698470837e-05, "loss": 0.7062, "step": 1371000 }, { "epoch": 0.49, "learning_rate": 1.0112675132350657e-05, "loss": 0.7011, "step": 1371500 }, { "epoch": 0.49, "learning_rate": 1.0109070566230479e-05, "loss": 0.7051, "step": 1372000 }, { "epoch": 0.49, "learning_rate": 1.01054660001103e-05, "loss": 0.7164, "step": 1372500 }, { "epoch": 0.49, "learning_rate": 1.0101861433990121e-05, "loss": 0.6732, "step": 1373000 }, { "epoch": 0.5, "learning_rate": 1.0098256867869945e-05, "loss": 0.6985, "step": 1373500 }, { "epoch": 0.5, "learning_rate": 1.0094652301749765e-05, "loss": 0.7094, "step": 1374000 }, { "epoch": 0.5, "learning_rate": 1.0091047735629587e-05, "loss": 0.7039, "step": 1374500 }, { "epoch": 0.5, "learning_rate": 1.0087443169509408e-05, "loss": 0.6952, "step": 1375000 }, { "epoch": 0.5, "learning_rate": 1.0083838603389231e-05, "loss": 0.7067, "step": 1375500 }, { "epoch": 0.5, "learning_rate": 1.0080234037269052e-05, "loss": 0.7087, "step": 1376000 }, { "epoch": 0.5, "learning_rate": 1.0076629471148874e-05, "loss": 0.704, "step": 1376500 }, { "epoch": 0.5, "learning_rate": 1.0073024905028694e-05, "loss": 0.6977, "step": 1377000 }, { "epoch": 0.5, "learning_rate": 1.0069420338908518e-05, "loss": 0.6982, "step": 1377500 }, { "epoch": 0.5, "learning_rate": 1.0065815772788338e-05, "loss": 0.7039, "step": 1378000 }, { "epoch": 0.5, "learning_rate": 1.006221120666816e-05, "loss": 0.7017, "step": 1378500 }, { "epoch": 0.5, "learning_rate": 1.005860664054798e-05, "loss": 0.7042, "step": 1379000 }, { "epoch": 0.5, "learning_rate": 1.0055002074427802e-05, "loss": 0.6767, "step": 1379500 }, { "epoch": 0.5, "learning_rate": 1.0051397508307626e-05, "loss": 0.7079, "step": 1380000 }, { "epoch": 0.5, "learning_rate": 1.0047792942187446e-05, "loss": 0.6932, "step": 1380500 }, { "epoch": 0.5, "learning_rate": 1.0044188376067267e-05, "loss": 0.7137, "step": 1381000 }, { "epoch": 0.5, "learning_rate": 1.0040583809947089e-05, "loss": 0.6885, "step": 1381500 }, { "epoch": 0.5, "learning_rate": 1.0036979243826913e-05, "loss": 0.7099, "step": 1382000 }, { "epoch": 0.5, "learning_rate": 1.0033374677706733e-05, "loss": 0.702, "step": 1382500 }, { "epoch": 0.5, "learning_rate": 1.0029770111586555e-05, "loss": 0.703, "step": 1383000 }, { "epoch": 0.5, "learning_rate": 1.0026165545466375e-05, "loss": 0.6907, "step": 1383500 }, { "epoch": 0.5, "learning_rate": 1.0022560979346197e-05, "loss": 0.7094, "step": 1384000 }, { "epoch": 0.5, "learning_rate": 1.001895641322602e-05, "loss": 0.7313, "step": 1384500 }, { "epoch": 0.5, "learning_rate": 1.0015351847105841e-05, "loss": 0.695, "step": 1385000 }, { "epoch": 0.5, "learning_rate": 1.0011747280985662e-05, "loss": 0.7046, "step": 1385500 }, { "epoch": 0.5, "learning_rate": 1.0008142714865484e-05, "loss": 0.6944, "step": 1386000 }, { "epoch": 0.5, "learning_rate": 1.0004538148745306e-05, "loss": 0.683, "step": 1386500 }, { "epoch": 0.5, "learning_rate": 1.0000933582625128e-05, "loss": 0.706, "step": 1387000 }, { "epoch": 0.5, "learning_rate": 9.997329016504948e-06, "loss": 0.6854, "step": 1387500 }, { "epoch": 0.5, "learning_rate": 9.99372445038477e-06, "loss": 0.6832, "step": 1388000 }, { "epoch": 0.5, "learning_rate": 9.990119884264592e-06, "loss": 0.7406, "step": 1388500 }, { "epoch": 0.5, "learning_rate": 9.986515318144412e-06, "loss": 0.6821, "step": 1389000 }, { "epoch": 0.5, "learning_rate": 9.982910752024236e-06, "loss": 0.7106, "step": 1389500 }, { "epoch": 0.5, "learning_rate": 9.979306185904056e-06, "loss": 0.7243, "step": 1390000 }, { "epoch": 0.5, "learning_rate": 9.975701619783878e-06, "loss": 0.6988, "step": 1390500 }, { "epoch": 0.5, "learning_rate": 9.9720970536637e-06, "loss": 0.7002, "step": 1391000 }, { "epoch": 0.5, "learning_rate": 9.968492487543522e-06, "loss": 0.7297, "step": 1391500 }, { "epoch": 0.5, "learning_rate": 9.964887921423343e-06, "loss": 0.7365, "step": 1392000 }, { "epoch": 0.5, "learning_rate": 9.961283355303165e-06, "loss": 0.676, "step": 1392500 }, { "epoch": 0.5, "learning_rate": 9.957678789182987e-06, "loss": 0.7139, "step": 1393000 }, { "epoch": 0.5, "learning_rate": 9.954074223062807e-06, "loss": 0.7219, "step": 1393500 }, { "epoch": 0.5, "learning_rate": 9.950469656942629e-06, "loss": 0.6682, "step": 1394000 }, { "epoch": 0.5, "learning_rate": 9.946865090822451e-06, "loss": 0.7122, "step": 1394500 }, { "epoch": 0.5, "learning_rate": 9.943260524702273e-06, "loss": 0.7113, "step": 1395000 }, { "epoch": 0.5, "learning_rate": 9.939655958582093e-06, "loss": 0.6842, "step": 1395500 }, { "epoch": 0.5, "learning_rate": 9.936051392461917e-06, "loss": 0.6995, "step": 1396000 }, { "epoch": 0.5, "learning_rate": 9.932446826341737e-06, "loss": 0.6967, "step": 1396500 }, { "epoch": 0.5, "learning_rate": 9.92884226022156e-06, "loss": 0.7032, "step": 1397000 }, { "epoch": 0.5, "learning_rate": 9.925237694101381e-06, "loss": 0.6853, "step": 1397500 }, { "epoch": 0.5, "learning_rate": 9.921633127981203e-06, "loss": 0.6755, "step": 1398000 }, { "epoch": 0.5, "learning_rate": 9.918028561861024e-06, "loss": 0.7138, "step": 1398500 }, { "epoch": 0.5, "learning_rate": 9.914423995740846e-06, "loss": 0.7049, "step": 1399000 }, { "epoch": 0.5, "learning_rate": 9.910819429620668e-06, "loss": 0.6753, "step": 1399500 }, { "epoch": 0.5, "learning_rate": 9.907214863500488e-06, "loss": 0.7056, "step": 1400000 }, { "epoch": 0.5, "learning_rate": 9.90361029738031e-06, "loss": 0.703, "step": 1400500 }, { "epoch": 0.5, "learning_rate": 9.900005731260132e-06, "loss": 0.7055, "step": 1401000 }, { "epoch": 0.51, "learning_rate": 9.896401165139954e-06, "loss": 0.6794, "step": 1401500 }, { "epoch": 0.51, "learning_rate": 9.892796599019774e-06, "loss": 0.6656, "step": 1402000 }, { "epoch": 0.51, "learning_rate": 9.889192032899596e-06, "loss": 0.6705, "step": 1402500 }, { "epoch": 0.51, "learning_rate": 9.885587466779418e-06, "loss": 0.6872, "step": 1403000 }, { "epoch": 0.51, "learning_rate": 9.881982900659239e-06, "loss": 0.7126, "step": 1403500 }, { "epoch": 0.51, "learning_rate": 9.87837833453906e-06, "loss": 0.682, "step": 1404000 }, { "epoch": 0.51, "learning_rate": 9.874773768418883e-06, "loss": 0.697, "step": 1404500 }, { "epoch": 0.51, "learning_rate": 9.871169202298705e-06, "loss": 0.6968, "step": 1405000 }, { "epoch": 0.51, "learning_rate": 9.867564636178527e-06, "loss": 0.7064, "step": 1405500 }, { "epoch": 0.51, "learning_rate": 9.863960070058349e-06, "loss": 0.6929, "step": 1406000 }, { "epoch": 0.51, "learning_rate": 9.86035550393817e-06, "loss": 0.7044, "step": 1406500 }, { "epoch": 0.51, "learning_rate": 9.856750937817991e-06, "loss": 0.6757, "step": 1407000 }, { "epoch": 0.51, "learning_rate": 9.853146371697813e-06, "loss": 0.6799, "step": 1407500 }, { "epoch": 0.51, "learning_rate": 9.849541805577635e-06, "loss": 0.7139, "step": 1408000 }, { "epoch": 0.51, "learning_rate": 9.845937239457456e-06, "loss": 0.693, "step": 1408500 }, { "epoch": 0.51, "learning_rate": 9.842332673337278e-06, "loss": 0.6774, "step": 1409000 }, { "epoch": 0.51, "learning_rate": 9.8387281072171e-06, "loss": 0.6729, "step": 1409500 }, { "epoch": 0.51, "learning_rate": 9.83512354109692e-06, "loss": 0.6874, "step": 1410000 }, { "epoch": 0.51, "learning_rate": 9.831518974976742e-06, "loss": 0.7152, "step": 1410500 }, { "epoch": 0.51, "learning_rate": 9.827914408856564e-06, "loss": 0.688, "step": 1411000 }, { "epoch": 0.51, "learning_rate": 9.824309842736386e-06, "loss": 0.685, "step": 1411500 }, { "epoch": 0.51, "learning_rate": 9.820705276616206e-06, "loss": 0.7035, "step": 1412000 }, { "epoch": 0.51, "learning_rate": 9.81710071049603e-06, "loss": 0.6897, "step": 1412500 }, { "epoch": 0.51, "learning_rate": 9.81349614437585e-06, "loss": 0.7011, "step": 1413000 }, { "epoch": 0.51, "learning_rate": 9.809891578255672e-06, "loss": 0.698, "step": 1413500 }, { "epoch": 0.51, "learning_rate": 9.806287012135494e-06, "loss": 0.7033, "step": 1414000 }, { "epoch": 0.51, "learning_rate": 9.802682446015315e-06, "loss": 0.6825, "step": 1414500 }, { "epoch": 0.51, "learning_rate": 9.799077879895137e-06, "loss": 0.685, "step": 1415000 }, { "epoch": 0.51, "learning_rate": 9.795473313774959e-06, "loss": 0.6778, "step": 1415500 }, { "epoch": 0.51, "learning_rate": 9.79186874765478e-06, "loss": 0.6993, "step": 1416000 }, { "epoch": 0.51, "learning_rate": 9.788264181534601e-06, "loss": 0.6753, "step": 1416500 }, { "epoch": 0.51, "learning_rate": 9.784659615414423e-06, "loss": 0.703, "step": 1417000 }, { "epoch": 0.51, "learning_rate": 9.781055049294245e-06, "loss": 0.6705, "step": 1417500 }, { "epoch": 0.51, "learning_rate": 9.777450483174067e-06, "loss": 0.71, "step": 1418000 }, { "epoch": 0.51, "learning_rate": 9.773845917053887e-06, "loss": 0.7182, "step": 1418500 }, { "epoch": 0.51, "learning_rate": 9.77024135093371e-06, "loss": 0.7011, "step": 1419000 }, { "epoch": 0.51, "learning_rate": 9.766636784813531e-06, "loss": 0.6775, "step": 1419500 }, { "epoch": 0.51, "learning_rate": 9.763032218693352e-06, "loss": 0.696, "step": 1420000 }, { "epoch": 0.51, "learning_rate": 9.759427652573175e-06, "loss": 0.6786, "step": 1420500 }, { "epoch": 0.51, "learning_rate": 9.755823086452996e-06, "loss": 0.7173, "step": 1421000 }, { "epoch": 0.51, "learning_rate": 9.752218520332818e-06, "loss": 0.6809, "step": 1421500 }, { "epoch": 0.51, "learning_rate": 9.74861395421264e-06, "loss": 0.6912, "step": 1422000 }, { "epoch": 0.51, "learning_rate": 9.745009388092462e-06, "loss": 0.6938, "step": 1422500 }, { "epoch": 0.51, "learning_rate": 9.741404821972282e-06, "loss": 0.6957, "step": 1423000 }, { "epoch": 0.51, "learning_rate": 9.737800255852104e-06, "loss": 0.7234, "step": 1423500 }, { "epoch": 0.51, "learning_rate": 9.734195689731926e-06, "loss": 0.6911, "step": 1424000 }, { "epoch": 0.51, "learning_rate": 9.730591123611746e-06, "loss": 0.678, "step": 1424500 }, { "epoch": 0.51, "learning_rate": 9.726986557491568e-06, "loss": 0.6626, "step": 1425000 }, { "epoch": 0.51, "learning_rate": 9.72338199137139e-06, "loss": 0.7038, "step": 1425500 }, { "epoch": 0.51, "learning_rate": 9.719777425251212e-06, "loss": 0.6725, "step": 1426000 }, { "epoch": 0.51, "learning_rate": 9.716172859131033e-06, "loss": 0.7294, "step": 1426500 }, { "epoch": 0.51, "learning_rate": 9.712568293010855e-06, "loss": 0.6902, "step": 1427000 }, { "epoch": 0.51, "learning_rate": 9.708963726890677e-06, "loss": 0.6808, "step": 1427500 }, { "epoch": 0.51, "learning_rate": 9.705359160770499e-06, "loss": 0.684, "step": 1428000 }, { "epoch": 0.51, "learning_rate": 9.70175459465032e-06, "loss": 0.6735, "step": 1428500 }, { "epoch": 0.52, "learning_rate": 9.698150028530141e-06, "loss": 0.6888, "step": 1429000 }, { "epoch": 0.52, "learning_rate": 9.694545462409963e-06, "loss": 0.6829, "step": 1429500 }, { "epoch": 0.52, "learning_rate": 9.690940896289785e-06, "loss": 0.6993, "step": 1430000 }, { "epoch": 0.52, "learning_rate": 9.687336330169607e-06, "loss": 0.6873, "step": 1430500 }, { "epoch": 0.52, "learning_rate": 9.683731764049428e-06, "loss": 0.7136, "step": 1431000 }, { "epoch": 0.52, "learning_rate": 9.68012719792925e-06, "loss": 0.6877, "step": 1431500 }, { "epoch": 0.52, "learning_rate": 9.676522631809072e-06, "loss": 0.7092, "step": 1432000 }, { "epoch": 0.52, "learning_rate": 9.672918065688894e-06, "loss": 0.6823, "step": 1432500 }, { "epoch": 0.52, "learning_rate": 9.669313499568714e-06, "loss": 0.6899, "step": 1433000 }, { "epoch": 0.52, "learning_rate": 9.665708933448536e-06, "loss": 0.6919, "step": 1433500 }, { "epoch": 0.52, "learning_rate": 9.662104367328358e-06, "loss": 0.7098, "step": 1434000 }, { "epoch": 0.52, "learning_rate": 9.658499801208178e-06, "loss": 0.7363, "step": 1434500 }, { "epoch": 0.52, "learning_rate": 9.654895235088e-06, "loss": 0.6962, "step": 1435000 }, { "epoch": 0.52, "learning_rate": 9.651290668967822e-06, "loss": 0.6713, "step": 1435500 }, { "epoch": 0.52, "learning_rate": 9.647686102847644e-06, "loss": 0.7395, "step": 1436000 }, { "epoch": 0.52, "learning_rate": 9.644081536727465e-06, "loss": 0.67, "step": 1436500 }, { "epoch": 0.52, "learning_rate": 9.640476970607288e-06, "loss": 0.7092, "step": 1437000 }, { "epoch": 0.52, "learning_rate": 9.636872404487109e-06, "loss": 0.6485, "step": 1437500 }, { "epoch": 0.52, "learning_rate": 9.63326783836693e-06, "loss": 0.6763, "step": 1438000 }, { "epoch": 0.52, "learning_rate": 9.629663272246753e-06, "loss": 0.6859, "step": 1438500 }, { "epoch": 0.52, "learning_rate": 9.626058706126575e-06, "loss": 0.6902, "step": 1439000 }, { "epoch": 0.52, "learning_rate": 9.622454140006395e-06, "loss": 0.6735, "step": 1439500 }, { "epoch": 0.52, "learning_rate": 9.618849573886217e-06, "loss": 0.7246, "step": 1440000 }, { "epoch": 0.52, "learning_rate": 9.615245007766039e-06, "loss": 0.6937, "step": 1440500 }, { "epoch": 0.52, "learning_rate": 9.61164044164586e-06, "loss": 0.6946, "step": 1441000 }, { "epoch": 0.52, "learning_rate": 9.608035875525681e-06, "loss": 0.6723, "step": 1441500 }, { "epoch": 0.52, "learning_rate": 9.604431309405503e-06, "loss": 0.6941, "step": 1442000 }, { "epoch": 0.52, "learning_rate": 9.600826743285325e-06, "loss": 0.6538, "step": 1442500 }, { "epoch": 0.52, "learning_rate": 9.597222177165146e-06, "loss": 0.7074, "step": 1443000 }, { "epoch": 0.52, "learning_rate": 9.59361761104497e-06, "loss": 0.7188, "step": 1443500 }, { "epoch": 0.52, "learning_rate": 9.59001304492479e-06, "loss": 0.6937, "step": 1444000 }, { "epoch": 0.52, "learning_rate": 9.58640847880461e-06, "loss": 0.6747, "step": 1444500 }, { "epoch": 0.52, "learning_rate": 9.582803912684434e-06, "loss": 0.6971, "step": 1445000 }, { "epoch": 0.52, "learning_rate": 9.579199346564254e-06, "loss": 0.7007, "step": 1445500 }, { "epoch": 0.52, "learning_rate": 9.575594780444076e-06, "loss": 0.6733, "step": 1446000 }, { "epoch": 0.52, "learning_rate": 9.571990214323898e-06, "loss": 0.7072, "step": 1446500 }, { "epoch": 0.52, "learning_rate": 9.56838564820372e-06, "loss": 0.6647, "step": 1447000 }, { "epoch": 0.52, "learning_rate": 9.56478108208354e-06, "loss": 0.7099, "step": 1447500 }, { "epoch": 0.52, "learning_rate": 9.561176515963362e-06, "loss": 0.6907, "step": 1448000 }, { "epoch": 0.52, "learning_rate": 9.557571949843184e-06, "loss": 0.6808, "step": 1448500 }, { "epoch": 0.52, "learning_rate": 9.553967383723006e-06, "loss": 0.66, "step": 1449000 }, { "epoch": 0.52, "learning_rate": 9.550362817602827e-06, "loss": 0.7064, "step": 1449500 }, { "epoch": 0.52, "learning_rate": 9.546758251482649e-06, "loss": 0.6547, "step": 1450000 }, { "epoch": 0.52, "learning_rate": 9.54315368536247e-06, "loss": 0.6962, "step": 1450500 }, { "epoch": 0.52, "learning_rate": 9.539549119242291e-06, "loss": 0.7045, "step": 1451000 }, { "epoch": 0.52, "learning_rate": 9.535944553122115e-06, "loss": 0.6478, "step": 1451500 }, { "epoch": 0.52, "learning_rate": 9.532339987001935e-06, "loss": 0.713, "step": 1452000 }, { "epoch": 0.52, "learning_rate": 9.528735420881757e-06, "loss": 0.677, "step": 1452500 }, { "epoch": 0.52, "learning_rate": 9.52513085476158e-06, "loss": 0.6923, "step": 1453000 }, { "epoch": 0.52, "learning_rate": 9.521526288641401e-06, "loss": 0.6859, "step": 1453500 }, { "epoch": 0.52, "learning_rate": 9.517921722521221e-06, "loss": 0.6774, "step": 1454000 }, { "epoch": 0.52, "learning_rate": 9.514317156401044e-06, "loss": 0.7308, "step": 1454500 }, { "epoch": 0.52, "learning_rate": 9.510712590280866e-06, "loss": 0.6728, "step": 1455000 }, { "epoch": 0.52, "learning_rate": 9.507108024160686e-06, "loss": 0.6843, "step": 1455500 }, { "epoch": 0.52, "learning_rate": 9.503503458040508e-06, "loss": 0.697, "step": 1456000 }, { "epoch": 0.53, "learning_rate": 9.49989889192033e-06, "loss": 0.7059, "step": 1456500 }, { "epoch": 0.53, "learning_rate": 9.496294325800152e-06, "loss": 0.703, "step": 1457000 }, { "epoch": 0.53, "learning_rate": 9.492689759679972e-06, "loss": 0.6654, "step": 1457500 }, { "epoch": 0.53, "learning_rate": 9.489085193559794e-06, "loss": 0.6905, "step": 1458000 }, { "epoch": 0.53, "learning_rate": 9.485480627439616e-06, "loss": 0.7009, "step": 1458500 }, { "epoch": 0.53, "learning_rate": 9.481876061319438e-06, "loss": 0.7061, "step": 1459000 }, { "epoch": 0.53, "learning_rate": 9.478271495199259e-06, "loss": 0.6724, "step": 1459500 }, { "epoch": 0.53, "learning_rate": 9.47466692907908e-06, "loss": 0.6908, "step": 1460000 }, { "epoch": 0.53, "learning_rate": 9.471062362958903e-06, "loss": 0.6899, "step": 1460500 }, { "epoch": 0.53, "learning_rate": 9.467457796838725e-06, "loss": 0.6991, "step": 1461000 }, { "epoch": 0.53, "learning_rate": 9.463853230718547e-06, "loss": 0.6914, "step": 1461500 }, { "epoch": 0.53, "learning_rate": 9.460248664598367e-06, "loss": 0.6741, "step": 1462000 }, { "epoch": 0.53, "learning_rate": 9.456644098478189e-06, "loss": 0.7054, "step": 1462500 }, { "epoch": 0.53, "learning_rate": 9.453039532358011e-06, "loss": 0.6951, "step": 1463000 }, { "epoch": 0.53, "learning_rate": 9.449434966237833e-06, "loss": 0.6656, "step": 1463500 }, { "epoch": 0.53, "learning_rate": 9.445830400117653e-06, "loss": 0.6817, "step": 1464000 }, { "epoch": 0.53, "learning_rate": 9.442225833997475e-06, "loss": 0.679, "step": 1464500 }, { "epoch": 0.53, "learning_rate": 9.438621267877297e-06, "loss": 0.7081, "step": 1465000 }, { "epoch": 0.53, "learning_rate": 9.435016701757118e-06, "loss": 0.6733, "step": 1465500 }, { "epoch": 0.53, "learning_rate": 9.43141213563694e-06, "loss": 0.699, "step": 1466000 }, { "epoch": 0.53, "learning_rate": 9.427807569516762e-06, "loss": 0.7048, "step": 1466500 }, { "epoch": 0.53, "learning_rate": 9.424203003396584e-06, "loss": 0.6866, "step": 1467000 }, { "epoch": 0.53, "learning_rate": 9.420598437276404e-06, "loss": 0.6968, "step": 1467500 }, { "epoch": 0.53, "learning_rate": 9.416993871156228e-06, "loss": 0.6906, "step": 1468000 }, { "epoch": 0.53, "learning_rate": 9.413389305036048e-06, "loss": 0.6965, "step": 1468500 }, { "epoch": 0.53, "learning_rate": 9.40978473891587e-06, "loss": 0.6904, "step": 1469000 }, { "epoch": 0.53, "learning_rate": 9.406180172795692e-06, "loss": 0.672, "step": 1469500 }, { "epoch": 0.53, "learning_rate": 9.402575606675512e-06, "loss": 0.6995, "step": 1470000 }, { "epoch": 0.53, "learning_rate": 9.398971040555334e-06, "loss": 0.7147, "step": 1470500 }, { "epoch": 0.53, "learning_rate": 9.395366474435156e-06, "loss": 0.6907, "step": 1471000 }, { "epoch": 0.53, "learning_rate": 9.391761908314978e-06, "loss": 0.6912, "step": 1471500 }, { "epoch": 0.53, "learning_rate": 9.388157342194799e-06, "loss": 0.6976, "step": 1472000 }, { "epoch": 0.53, "learning_rate": 9.38455277607462e-06, "loss": 0.7059, "step": 1472500 }, { "epoch": 0.53, "learning_rate": 9.380948209954443e-06, "loss": 0.6832, "step": 1473000 }, { "epoch": 0.53, "learning_rate": 9.377343643834265e-06, "loss": 0.6978, "step": 1473500 }, { "epoch": 0.53, "learning_rate": 9.373739077714085e-06, "loss": 0.7014, "step": 1474000 }, { "epoch": 0.53, "learning_rate": 9.370134511593909e-06, "loss": 0.6914, "step": 1474500 }, { "epoch": 0.53, "learning_rate": 9.366529945473729e-06, "loss": 0.6844, "step": 1475000 }, { "epoch": 0.53, "learning_rate": 9.36292537935355e-06, "loss": 0.6728, "step": 1475500 }, { "epoch": 0.53, "learning_rate": 9.359320813233373e-06, "loss": 0.7041, "step": 1476000 }, { "epoch": 0.53, "learning_rate": 9.355716247113193e-06, "loss": 0.6941, "step": 1476500 }, { "epoch": 0.53, "learning_rate": 9.352111680993015e-06, "loss": 0.6894, "step": 1477000 }, { "epoch": 0.53, "learning_rate": 9.348507114872838e-06, "loss": 0.6824, "step": 1477500 }, { "epoch": 0.53, "learning_rate": 9.34490254875266e-06, "loss": 0.6994, "step": 1478000 }, { "epoch": 0.53, "learning_rate": 9.34129798263248e-06, "loss": 0.6765, "step": 1478500 }, { "epoch": 0.53, "learning_rate": 9.337693416512302e-06, "loss": 0.6682, "step": 1479000 }, { "epoch": 0.53, "learning_rate": 9.334088850392124e-06, "loss": 0.6657, "step": 1479500 }, { "epoch": 0.53, "learning_rate": 9.330484284271946e-06, "loss": 0.6845, "step": 1480000 }, { "epoch": 0.53, "learning_rate": 9.326879718151766e-06, "loss": 0.6948, "step": 1480500 }, { "epoch": 0.53, "learning_rate": 9.323275152031588e-06, "loss": 0.67, "step": 1481000 }, { "epoch": 0.53, "learning_rate": 9.31967058591141e-06, "loss": 0.6946, "step": 1481500 }, { "epoch": 0.53, "learning_rate": 9.31606601979123e-06, "loss": 0.6923, "step": 1482000 }, { "epoch": 0.53, "learning_rate": 9.312461453671053e-06, "loss": 0.6844, "step": 1482500 }, { "epoch": 0.53, "learning_rate": 9.308856887550875e-06, "loss": 0.6963, "step": 1483000 }, { "epoch": 0.53, "learning_rate": 9.305252321430697e-06, "loss": 0.6654, "step": 1483500 }, { "epoch": 0.53, "learning_rate": 9.301647755310519e-06, "loss": 0.6909, "step": 1484000 }, { "epoch": 0.54, "learning_rate": 9.29804318919034e-06, "loss": 0.673, "step": 1484500 }, { "epoch": 0.54, "learning_rate": 9.294438623070161e-06, "loss": 0.6916, "step": 1485000 }, { "epoch": 0.54, "learning_rate": 9.290834056949983e-06, "loss": 0.6757, "step": 1485500 }, { "epoch": 0.54, "learning_rate": 9.287229490829805e-06, "loss": 0.7092, "step": 1486000 }, { "epoch": 0.54, "learning_rate": 9.283624924709625e-06, "loss": 0.6862, "step": 1486500 }, { "epoch": 0.54, "learning_rate": 9.280020358589447e-06, "loss": 0.6701, "step": 1487000 }, { "epoch": 0.54, "learning_rate": 9.27641579246927e-06, "loss": 0.6718, "step": 1487500 }, { "epoch": 0.54, "learning_rate": 9.272811226349091e-06, "loss": 0.6868, "step": 1488000 }, { "epoch": 0.54, "learning_rate": 9.269206660228912e-06, "loss": 0.6777, "step": 1488500 }, { "epoch": 0.54, "learning_rate": 9.265602094108734e-06, "loss": 0.7123, "step": 1489000 }, { "epoch": 0.54, "learning_rate": 9.261997527988556e-06, "loss": 0.672, "step": 1489500 }, { "epoch": 0.54, "learning_rate": 9.258392961868378e-06, "loss": 0.7109, "step": 1490000 }, { "epoch": 0.54, "learning_rate": 9.254788395748198e-06, "loss": 0.6914, "step": 1490500 }, { "epoch": 0.54, "learning_rate": 9.25118382962802e-06, "loss": 0.7079, "step": 1491000 }, { "epoch": 0.54, "learning_rate": 9.247579263507842e-06, "loss": 0.6962, "step": 1491500 }, { "epoch": 0.54, "learning_rate": 9.243974697387662e-06, "loss": 0.6765, "step": 1492000 }, { "epoch": 0.54, "learning_rate": 9.240370131267486e-06, "loss": 0.6962, "step": 1492500 }, { "epoch": 0.54, "learning_rate": 9.236765565147306e-06, "loss": 0.692, "step": 1493000 }, { "epoch": 0.54, "learning_rate": 9.233160999027128e-06, "loss": 0.6726, "step": 1493500 }, { "epoch": 0.54, "learning_rate": 9.22955643290695e-06, "loss": 0.6702, "step": 1494000 }, { "epoch": 0.54, "learning_rate": 9.225951866786772e-06, "loss": 0.6847, "step": 1494500 }, { "epoch": 0.54, "learning_rate": 9.222347300666593e-06, "loss": 0.7141, "step": 1495000 }, { "epoch": 0.54, "learning_rate": 9.218742734546415e-06, "loss": 0.6769, "step": 1495500 }, { "epoch": 0.54, "learning_rate": 9.215138168426237e-06, "loss": 0.7128, "step": 1496000 }, { "epoch": 0.54, "learning_rate": 9.211533602306057e-06, "loss": 0.6543, "step": 1496500 }, { "epoch": 0.54, "learning_rate": 9.207929036185879e-06, "loss": 0.7096, "step": 1497000 }, { "epoch": 0.54, "learning_rate": 9.204324470065701e-06, "loss": 0.6949, "step": 1497500 }, { "epoch": 0.54, "learning_rate": 9.200719903945523e-06, "loss": 0.6808, "step": 1498000 }, { "epoch": 0.54, "learning_rate": 9.197115337825343e-06, "loss": 0.6895, "step": 1498500 }, { "epoch": 0.54, "learning_rate": 9.193510771705167e-06, "loss": 0.7145, "step": 1499000 }, { "epoch": 0.54, "learning_rate": 9.189906205584987e-06, "loss": 0.6796, "step": 1499500 }, { "epoch": 0.54, "learning_rate": 9.18630163946481e-06, "loss": 0.6708, "step": 1500000 }, { "epoch": 0.54, "learning_rate": 9.182697073344632e-06, "loss": 0.6761, "step": 1500500 }, { "epoch": 0.54, "learning_rate": 9.179092507224452e-06, "loss": 0.6636, "step": 1501000 }, { "epoch": 0.54, "learning_rate": 9.175487941104274e-06, "loss": 0.7076, "step": 1501500 }, { "epoch": 0.54, "learning_rate": 9.171883374984096e-06, "loss": 0.6913, "step": 1502000 }, { "epoch": 0.54, "learning_rate": 9.168278808863918e-06, "loss": 0.6845, "step": 1502500 }, { "epoch": 0.54, "learning_rate": 9.164674242743738e-06, "loss": 0.6921, "step": 1503000 }, { "epoch": 0.54, "learning_rate": 9.16106967662356e-06, "loss": 0.6882, "step": 1503500 }, { "epoch": 0.54, "learning_rate": 9.157465110503382e-06, "loss": 0.6797, "step": 1504000 }, { "epoch": 0.54, "learning_rate": 9.153860544383204e-06, "loss": 0.6581, "step": 1504500 }, { "epoch": 0.54, "learning_rate": 9.150255978263025e-06, "loss": 0.6871, "step": 1505000 }, { "epoch": 0.54, "learning_rate": 9.146651412142847e-06, "loss": 0.6862, "step": 1505500 }, { "epoch": 0.54, "learning_rate": 9.143046846022669e-06, "loss": 0.6278, "step": 1506000 }, { "epoch": 0.54, "learning_rate": 9.139442279902489e-06, "loss": 0.6617, "step": 1506500 }, { "epoch": 0.54, "learning_rate": 9.135837713782313e-06, "loss": 0.6652, "step": 1507000 }, { "epoch": 0.54, "learning_rate": 9.132233147662133e-06, "loss": 0.6881, "step": 1507500 }, { "epoch": 0.54, "learning_rate": 9.128628581541955e-06, "loss": 0.6983, "step": 1508000 }, { "epoch": 0.54, "learning_rate": 9.125024015421777e-06, "loss": 0.7109, "step": 1508500 }, { "epoch": 0.54, "learning_rate": 9.121419449301599e-06, "loss": 0.6871, "step": 1509000 }, { "epoch": 0.54, "learning_rate": 9.11781488318142e-06, "loss": 0.6861, "step": 1509500 }, { "epoch": 0.54, "learning_rate": 9.114210317061241e-06, "loss": 0.6993, "step": 1510000 }, { "epoch": 0.54, "learning_rate": 9.110605750941063e-06, "loss": 0.7179, "step": 1510500 }, { "epoch": 0.54, "learning_rate": 9.107001184820884e-06, "loss": 0.6845, "step": 1511000 }, { "epoch": 0.54, "learning_rate": 9.103396618700706e-06, "loss": 0.7, "step": 1511500 }, { "epoch": 0.55, "learning_rate": 9.099792052580528e-06, "loss": 0.6838, "step": 1512000 }, { "epoch": 0.55, "learning_rate": 9.09618748646035e-06, "loss": 0.7016, "step": 1512500 }, { "epoch": 0.55, "learning_rate": 9.09258292034017e-06, "loss": 0.6746, "step": 1513000 }, { "epoch": 0.55, "learning_rate": 9.088978354219992e-06, "loss": 0.711, "step": 1513500 }, { "epoch": 0.55, "learning_rate": 9.085373788099814e-06, "loss": 0.6691, "step": 1514000 }, { "epoch": 0.55, "learning_rate": 9.081769221979636e-06, "loss": 0.66, "step": 1514500 }, { "epoch": 0.55, "learning_rate": 9.078164655859456e-06, "loss": 0.6896, "step": 1515000 }, { "epoch": 0.55, "learning_rate": 9.07456008973928e-06, "loss": 0.674, "step": 1515500 }, { "epoch": 0.55, "learning_rate": 9.0709555236191e-06, "loss": 0.6685, "step": 1516000 }, { "epoch": 0.55, "learning_rate": 9.067350957498922e-06, "loss": 0.6967, "step": 1516500 }, { "epoch": 0.55, "learning_rate": 9.063746391378744e-06, "loss": 0.7119, "step": 1517000 }, { "epoch": 0.55, "learning_rate": 9.060141825258565e-06, "loss": 0.6918, "step": 1517500 }, { "epoch": 0.55, "learning_rate": 9.056537259138387e-06, "loss": 0.6879, "step": 1518000 }, { "epoch": 0.55, "learning_rate": 9.052932693018209e-06, "loss": 0.6845, "step": 1518500 }, { "epoch": 0.55, "learning_rate": 9.04932812689803e-06, "loss": 0.6745, "step": 1519000 }, { "epoch": 0.55, "learning_rate": 9.045723560777851e-06, "loss": 0.6994, "step": 1519500 }, { "epoch": 0.55, "learning_rate": 9.042118994657673e-06, "loss": 0.6806, "step": 1520000 }, { "epoch": 0.55, "learning_rate": 9.038514428537495e-06, "loss": 0.7047, "step": 1520500 }, { "epoch": 0.55, "learning_rate": 9.034909862417317e-06, "loss": 0.6903, "step": 1521000 }, { "epoch": 0.55, "learning_rate": 9.031305296297137e-06, "loss": 0.6764, "step": 1521500 }, { "epoch": 0.55, "learning_rate": 9.02770073017696e-06, "loss": 0.7002, "step": 1522000 }, { "epoch": 0.55, "learning_rate": 9.024096164056781e-06, "loss": 0.6732, "step": 1522500 }, { "epoch": 0.55, "learning_rate": 9.020491597936602e-06, "loss": 0.69, "step": 1523000 }, { "epoch": 0.55, "learning_rate": 9.016887031816425e-06, "loss": 0.6839, "step": 1523500 }, { "epoch": 0.55, "learning_rate": 9.013282465696246e-06, "loss": 0.6812, "step": 1524000 }, { "epoch": 0.55, "learning_rate": 9.009677899576068e-06, "loss": 0.6916, "step": 1524500 }, { "epoch": 0.55, "learning_rate": 9.00607333345589e-06, "loss": 0.6708, "step": 1525000 }, { "epoch": 0.55, "learning_rate": 9.002468767335712e-06, "loss": 0.6742, "step": 1525500 }, { "epoch": 0.55, "learning_rate": 8.998864201215532e-06, "loss": 0.6981, "step": 1526000 }, { "epoch": 0.55, "learning_rate": 8.995259635095354e-06, "loss": 0.6678, "step": 1526500 }, { "epoch": 0.55, "learning_rate": 8.991655068975176e-06, "loss": 0.67, "step": 1527000 }, { "epoch": 0.55, "learning_rate": 8.988050502854997e-06, "loss": 0.6878, "step": 1527500 }, { "epoch": 0.55, "learning_rate": 8.984445936734819e-06, "loss": 0.6941, "step": 1528000 }, { "epoch": 0.55, "learning_rate": 8.98084137061464e-06, "loss": 0.6733, "step": 1528500 }, { "epoch": 0.55, "learning_rate": 8.977236804494463e-06, "loss": 0.7078, "step": 1529000 }, { "epoch": 0.55, "learning_rate": 8.973632238374283e-06, "loss": 0.7015, "step": 1529500 }, { "epoch": 0.55, "learning_rate": 8.970027672254107e-06, "loss": 0.6631, "step": 1530000 }, { "epoch": 0.55, "learning_rate": 8.966423106133927e-06, "loss": 0.6681, "step": 1530500 }, { "epoch": 0.55, "learning_rate": 8.962818540013749e-06, "loss": 0.6899, "step": 1531000 }, { "epoch": 0.55, "learning_rate": 8.959213973893571e-06, "loss": 0.6835, "step": 1531500 }, { "epoch": 0.55, "learning_rate": 8.955609407773391e-06, "loss": 0.686, "step": 1532000 }, { "epoch": 0.55, "learning_rate": 8.952004841653213e-06, "loss": 0.6686, "step": 1532500 }, { "epoch": 0.55, "learning_rate": 8.948400275533035e-06, "loss": 0.6859, "step": 1533000 }, { "epoch": 0.55, "learning_rate": 8.944795709412857e-06, "loss": 0.6879, "step": 1533500 }, { "epoch": 0.55, "learning_rate": 8.941191143292678e-06, "loss": 0.6695, "step": 1534000 }, { "epoch": 0.55, "learning_rate": 8.9375865771725e-06, "loss": 0.6869, "step": 1534500 }, { "epoch": 0.55, "learning_rate": 8.933982011052322e-06, "loss": 0.6796, "step": 1535000 }, { "epoch": 0.55, "learning_rate": 8.930377444932144e-06, "loss": 0.6877, "step": 1535500 }, { "epoch": 0.55, "learning_rate": 8.926772878811964e-06, "loss": 0.6795, "step": 1536000 }, { "epoch": 0.55, "learning_rate": 8.923168312691786e-06, "loss": 0.6919, "step": 1536500 }, { "epoch": 0.55, "learning_rate": 8.919563746571608e-06, "loss": 0.6995, "step": 1537000 }, { "epoch": 0.55, "learning_rate": 8.915959180451428e-06, "loss": 0.684, "step": 1537500 }, { "epoch": 0.55, "learning_rate": 8.91235461433125e-06, "loss": 0.6812, "step": 1538000 }, { "epoch": 0.55, "learning_rate": 8.908750048211072e-06, "loss": 0.6923, "step": 1538500 }, { "epoch": 0.55, "learning_rate": 8.905145482090894e-06, "loss": 0.6885, "step": 1539000 }, { "epoch": 0.55, "learning_rate": 8.901540915970716e-06, "loss": 0.6828, "step": 1539500 }, { "epoch": 0.56, "learning_rate": 8.897936349850538e-06, "loss": 0.7031, "step": 1540000 }, { "epoch": 0.56, "learning_rate": 8.894331783730359e-06, "loss": 0.7018, "step": 1540500 }, { "epoch": 0.56, "learning_rate": 8.89072721761018e-06, "loss": 0.6888, "step": 1541000 }, { "epoch": 0.56, "learning_rate": 8.887122651490003e-06, "loss": 0.6967, "step": 1541500 }, { "epoch": 0.56, "learning_rate": 8.883518085369823e-06, "loss": 0.7062, "step": 1542000 }, { "epoch": 0.56, "learning_rate": 8.879913519249645e-06, "loss": 0.6927, "step": 1542500 }, { "epoch": 0.56, "learning_rate": 8.876308953129467e-06, "loss": 0.6925, "step": 1543000 }, { "epoch": 0.56, "learning_rate": 8.872704387009289e-06, "loss": 0.6771, "step": 1543500 }, { "epoch": 0.56, "learning_rate": 8.86909982088911e-06, "loss": 0.6837, "step": 1544000 }, { "epoch": 0.56, "learning_rate": 8.865495254768931e-06, "loss": 0.6694, "step": 1544500 }, { "epoch": 0.56, "learning_rate": 8.861890688648753e-06, "loss": 0.6887, "step": 1545000 }, { "epoch": 0.56, "learning_rate": 8.858286122528575e-06, "loss": 0.6741, "step": 1545500 }, { "epoch": 0.56, "learning_rate": 8.854681556408396e-06, "loss": 0.6908, "step": 1546000 }, { "epoch": 0.56, "learning_rate": 8.85107699028822e-06, "loss": 0.6934, "step": 1546500 }, { "epoch": 0.56, "learning_rate": 8.84747242416804e-06, "loss": 0.703, "step": 1547000 }, { "epoch": 0.56, "learning_rate": 8.84386785804786e-06, "loss": 0.6616, "step": 1547500 }, { "epoch": 0.56, "learning_rate": 8.840263291927684e-06, "loss": 0.6713, "step": 1548000 }, { "epoch": 0.56, "learning_rate": 8.836658725807504e-06, "loss": 0.6831, "step": 1548500 }, { "epoch": 0.56, "learning_rate": 8.833054159687326e-06, "loss": 0.7002, "step": 1549000 }, { "epoch": 0.56, "learning_rate": 8.829449593567148e-06, "loss": 0.6835, "step": 1549500 }, { "epoch": 0.56, "learning_rate": 8.82584502744697e-06, "loss": 0.6788, "step": 1550000 }, { "epoch": 0.56, "learning_rate": 8.82224046132679e-06, "loss": 0.717, "step": 1550500 }, { "epoch": 0.56, "learning_rate": 8.818635895206613e-06, "loss": 0.6821, "step": 1551000 }, { "epoch": 0.56, "learning_rate": 8.815031329086435e-06, "loss": 0.6653, "step": 1551500 }, { "epoch": 0.56, "learning_rate": 8.811426762966255e-06, "loss": 0.6666, "step": 1552000 }, { "epoch": 0.56, "learning_rate": 8.807822196846077e-06, "loss": 0.7118, "step": 1552500 }, { "epoch": 0.56, "learning_rate": 8.804217630725899e-06, "loss": 0.6746, "step": 1553000 }, { "epoch": 0.56, "learning_rate": 8.800613064605721e-06, "loss": 0.6852, "step": 1553500 }, { "epoch": 0.56, "learning_rate": 8.797008498485541e-06, "loss": 0.6893, "step": 1554000 }, { "epoch": 0.56, "learning_rate": 8.793403932365365e-06, "loss": 0.6965, "step": 1554500 }, { "epoch": 0.56, "learning_rate": 8.789799366245185e-06, "loss": 0.6835, "step": 1555000 }, { "epoch": 0.56, "learning_rate": 8.786194800125007e-06, "loss": 0.6705, "step": 1555500 }, { "epoch": 0.56, "learning_rate": 8.78259023400483e-06, "loss": 0.6815, "step": 1556000 }, { "epoch": 0.56, "learning_rate": 8.778985667884651e-06, "loss": 0.6741, "step": 1556500 }, { "epoch": 0.56, "learning_rate": 8.775381101764472e-06, "loss": 0.6681, "step": 1557000 }, { "epoch": 0.56, "learning_rate": 8.771776535644294e-06, "loss": 0.6757, "step": 1557500 }, { "epoch": 0.56, "learning_rate": 8.768171969524116e-06, "loss": 0.6617, "step": 1558000 }, { "epoch": 0.56, "learning_rate": 8.764567403403936e-06, "loss": 0.6641, "step": 1558500 }, { "epoch": 0.56, "learning_rate": 8.760962837283758e-06, "loss": 0.7002, "step": 1559000 }, { "epoch": 0.56, "learning_rate": 8.75735827116358e-06, "loss": 0.6707, "step": 1559500 }, { "epoch": 0.56, "learning_rate": 8.753753705043402e-06, "loss": 0.6727, "step": 1560000 }, { "epoch": 0.56, "learning_rate": 8.750149138923222e-06, "loss": 0.6714, "step": 1560500 }, { "epoch": 0.56, "learning_rate": 8.746544572803044e-06, "loss": 0.6742, "step": 1561000 }, { "epoch": 0.56, "learning_rate": 8.742940006682866e-06, "loss": 0.7021, "step": 1561500 }, { "epoch": 0.56, "learning_rate": 8.739335440562688e-06, "loss": 0.6695, "step": 1562000 }, { "epoch": 0.56, "learning_rate": 8.73573087444251e-06, "loss": 0.6646, "step": 1562500 }, { "epoch": 0.56, "learning_rate": 8.73212630832233e-06, "loss": 0.687, "step": 1563000 }, { "epoch": 0.56, "learning_rate": 8.728521742202153e-06, "loss": 0.6761, "step": 1563500 }, { "epoch": 0.56, "learning_rate": 8.724917176081975e-06, "loss": 0.6535, "step": 1564000 }, { "epoch": 0.56, "learning_rate": 8.721312609961797e-06, "loss": 0.6902, "step": 1564500 }, { "epoch": 0.56, "learning_rate": 8.717708043841617e-06, "loss": 0.6944, "step": 1565000 }, { "epoch": 0.56, "learning_rate": 8.714103477721439e-06, "loss": 0.7163, "step": 1565500 }, { "epoch": 0.56, "learning_rate": 8.710498911601261e-06, "loss": 0.7236, "step": 1566000 }, { "epoch": 0.56, "learning_rate": 8.706894345481083e-06, "loss": 0.6844, "step": 1566500 }, { "epoch": 0.56, "learning_rate": 8.703289779360903e-06, "loss": 0.6942, "step": 1567000 }, { "epoch": 0.57, "learning_rate": 8.699685213240725e-06, "loss": 0.7042, "step": 1567500 }, { "epoch": 0.57, "learning_rate": 8.696080647120547e-06, "loss": 0.6616, "step": 1568000 }, { "epoch": 0.57, "learning_rate": 8.692476081000368e-06, "loss": 0.6775, "step": 1568500 }, { "epoch": 0.57, "learning_rate": 8.68887151488019e-06, "loss": 0.6768, "step": 1569000 }, { "epoch": 0.57, "learning_rate": 8.685266948760012e-06, "loss": 0.6707, "step": 1569500 }, { "epoch": 0.57, "learning_rate": 8.681662382639834e-06, "loss": 0.6861, "step": 1570000 }, { "epoch": 0.57, "learning_rate": 8.678057816519654e-06, "loss": 0.693, "step": 1570500 }, { "epoch": 0.57, "learning_rate": 8.674453250399478e-06, "loss": 0.712, "step": 1571000 }, { "epoch": 0.57, "learning_rate": 8.670848684279298e-06, "loss": 0.6856, "step": 1571500 }, { "epoch": 0.57, "learning_rate": 8.66724411815912e-06, "loss": 0.6756, "step": 1572000 }, { "epoch": 0.57, "learning_rate": 8.663639552038942e-06, "loss": 0.6579, "step": 1572500 }, { "epoch": 0.57, "learning_rate": 8.660034985918763e-06, "loss": 0.6798, "step": 1573000 }, { "epoch": 0.57, "learning_rate": 8.656430419798585e-06, "loss": 0.6482, "step": 1573500 }, { "epoch": 0.57, "learning_rate": 8.652825853678407e-06, "loss": 0.6809, "step": 1574000 }, { "epoch": 0.57, "learning_rate": 8.649221287558229e-06, "loss": 0.6974, "step": 1574500 }, { "epoch": 0.57, "learning_rate": 8.645616721438049e-06, "loss": 0.6702, "step": 1575000 }, { "epoch": 0.57, "learning_rate": 8.642012155317871e-06, "loss": 0.7021, "step": 1575500 }, { "epoch": 0.57, "learning_rate": 8.638407589197693e-06, "loss": 0.7163, "step": 1576000 }, { "epoch": 0.57, "learning_rate": 8.634803023077515e-06, "loss": 0.6803, "step": 1576500 }, { "epoch": 0.57, "learning_rate": 8.631198456957335e-06, "loss": 0.6739, "step": 1577000 }, { "epoch": 0.57, "learning_rate": 8.627593890837157e-06, "loss": 0.685, "step": 1577500 }, { "epoch": 0.57, "learning_rate": 8.62398932471698e-06, "loss": 0.6974, "step": 1578000 }, { "epoch": 0.57, "learning_rate": 8.6203847585968e-06, "loss": 0.6913, "step": 1578500 }, { "epoch": 0.57, "learning_rate": 8.616780192476623e-06, "loss": 0.6512, "step": 1579000 }, { "epoch": 0.57, "learning_rate": 8.613175626356444e-06, "loss": 0.694, "step": 1579500 }, { "epoch": 0.57, "learning_rate": 8.609571060236266e-06, "loss": 0.6946, "step": 1580000 }, { "epoch": 0.57, "learning_rate": 8.605966494116088e-06, "loss": 0.691, "step": 1580500 }, { "epoch": 0.57, "learning_rate": 8.60236192799591e-06, "loss": 0.6814, "step": 1581000 }, { "epoch": 0.57, "learning_rate": 8.59875736187573e-06, "loss": 0.6852, "step": 1581500 }, { "epoch": 0.57, "learning_rate": 8.595152795755552e-06, "loss": 0.6552, "step": 1582000 }, { "epoch": 0.57, "learning_rate": 8.591548229635374e-06, "loss": 0.6765, "step": 1582500 }, { "epoch": 0.57, "learning_rate": 8.587943663515194e-06, "loss": 0.6658, "step": 1583000 }, { "epoch": 0.57, "learning_rate": 8.584339097395016e-06, "loss": 0.6656, "step": 1583500 }, { "epoch": 0.57, "learning_rate": 8.580734531274838e-06, "loss": 0.6794, "step": 1584000 }, { "epoch": 0.57, "learning_rate": 8.57712996515466e-06, "loss": 0.6769, "step": 1584500 }, { "epoch": 0.57, "learning_rate": 8.57352539903448e-06, "loss": 0.6651, "step": 1585000 }, { "epoch": 0.57, "learning_rate": 8.569920832914304e-06, "loss": 0.6632, "step": 1585500 }, { "epoch": 0.57, "learning_rate": 8.566316266794125e-06, "loss": 0.6837, "step": 1586000 }, { "epoch": 0.57, "learning_rate": 8.562711700673947e-06, "loss": 0.6707, "step": 1586500 }, { "epoch": 0.57, "learning_rate": 8.559107134553769e-06, "loss": 0.6747, "step": 1587000 }, { "epoch": 0.57, "learning_rate": 8.55550256843359e-06, "loss": 0.684, "step": 1587500 }, { "epoch": 0.57, "learning_rate": 8.551898002313411e-06, "loss": 0.684, "step": 1588000 }, { "epoch": 0.57, "learning_rate": 8.548293436193233e-06, "loss": 0.6497, "step": 1588500 }, { "epoch": 0.57, "learning_rate": 8.544688870073055e-06, "loss": 0.6748, "step": 1589000 }, { "epoch": 0.57, "learning_rate": 8.541084303952875e-06, "loss": 0.6785, "step": 1589500 }, { "epoch": 0.57, "learning_rate": 8.537479737832697e-06, "loss": 0.6651, "step": 1590000 }, { "epoch": 0.57, "learning_rate": 8.53387517171252e-06, "loss": 0.6814, "step": 1590500 }, { "epoch": 0.57, "learning_rate": 8.530270605592341e-06, "loss": 0.6456, "step": 1591000 }, { "epoch": 0.57, "learning_rate": 8.526666039472162e-06, "loss": 0.6874, "step": 1591500 }, { "epoch": 0.57, "learning_rate": 8.523061473351984e-06, "loss": 0.6764, "step": 1592000 }, { "epoch": 0.57, "learning_rate": 8.519456907231806e-06, "loss": 0.649, "step": 1592500 }, { "epoch": 0.57, "learning_rate": 8.515852341111626e-06, "loss": 0.6662, "step": 1593000 }, { "epoch": 0.57, "learning_rate": 8.512247774991448e-06, "loss": 0.668, "step": 1593500 }, { "epoch": 0.57, "learning_rate": 8.50864320887127e-06, "loss": 0.6802, "step": 1594000 }, { "epoch": 0.57, "learning_rate": 8.505038642751092e-06, "loss": 0.6659, "step": 1594500 }, { "epoch": 0.57, "learning_rate": 8.501434076630914e-06, "loss": 0.7095, "step": 1595000 }, { "epoch": 0.58, "learning_rate": 8.497829510510736e-06, "loss": 0.6977, "step": 1595500 }, { "epoch": 0.58, "learning_rate": 8.494224944390557e-06, "loss": 0.6859, "step": 1596000 }, { "epoch": 0.58, "learning_rate": 8.490620378270379e-06, "loss": 0.6506, "step": 1596500 }, { "epoch": 0.58, "learning_rate": 8.4870158121502e-06, "loss": 0.685, "step": 1597000 }, { "epoch": 0.58, "learning_rate": 8.483411246030023e-06, "loss": 0.6879, "step": 1597500 }, { "epoch": 0.58, "learning_rate": 8.479806679909843e-06, "loss": 0.6672, "step": 1598000 }, { "epoch": 0.58, "learning_rate": 8.476202113789665e-06, "loss": 0.6637, "step": 1598500 }, { "epoch": 0.58, "learning_rate": 8.472597547669487e-06, "loss": 0.7041, "step": 1599000 }, { "epoch": 0.58, "learning_rate": 8.468992981549307e-06, "loss": 0.7043, "step": 1599500 }, { "epoch": 0.58, "learning_rate": 8.46538841542913e-06, "loss": 0.6918, "step": 1600000 }, { "epoch": 0.58, "learning_rate": 8.461783849308951e-06, "loss": 0.6805, "step": 1600500 }, { "epoch": 0.58, "learning_rate": 8.458179283188773e-06, "loss": 0.6807, "step": 1601000 }, { "epoch": 0.58, "learning_rate": 8.454574717068594e-06, "loss": 0.6699, "step": 1601500 }, { "epoch": 0.58, "learning_rate": 8.450970150948417e-06, "loss": 0.6591, "step": 1602000 }, { "epoch": 0.58, "learning_rate": 8.447365584828238e-06, "loss": 0.6835, "step": 1602500 }, { "epoch": 0.58, "learning_rate": 8.44376101870806e-06, "loss": 0.682, "step": 1603000 }, { "epoch": 0.58, "learning_rate": 8.440156452587882e-06, "loss": 0.6702, "step": 1603500 }, { "epoch": 0.58, "learning_rate": 8.436551886467702e-06, "loss": 0.6655, "step": 1604000 }, { "epoch": 0.58, "learning_rate": 8.432947320347524e-06, "loss": 0.6709, "step": 1604500 }, { "epoch": 0.58, "learning_rate": 8.429342754227346e-06, "loss": 0.6827, "step": 1605000 }, { "epoch": 0.58, "learning_rate": 8.425738188107168e-06, "loss": 0.6856, "step": 1605500 }, { "epoch": 0.58, "learning_rate": 8.422133621986988e-06, "loss": 0.6989, "step": 1606000 }, { "epoch": 0.58, "learning_rate": 8.41852905586681e-06, "loss": 0.6849, "step": 1606500 }, { "epoch": 0.58, "learning_rate": 8.414924489746632e-06, "loss": 0.6741, "step": 1607000 }, { "epoch": 0.58, "learning_rate": 8.411319923626454e-06, "loss": 0.685, "step": 1607500 }, { "epoch": 0.58, "learning_rate": 8.407715357506275e-06, "loss": 0.6754, "step": 1608000 }, { "epoch": 0.58, "learning_rate": 8.404110791386097e-06, "loss": 0.6924, "step": 1608500 }, { "epoch": 0.58, "learning_rate": 8.400506225265919e-06, "loss": 0.6725, "step": 1609000 }, { "epoch": 0.58, "learning_rate": 8.396901659145739e-06, "loss": 0.6709, "step": 1609500 }, { "epoch": 0.58, "learning_rate": 8.393297093025563e-06, "loss": 0.6773, "step": 1610000 }, { "epoch": 0.58, "learning_rate": 8.389692526905383e-06, "loss": 0.6671, "step": 1610500 }, { "epoch": 0.58, "learning_rate": 8.386087960785205e-06, "loss": 0.694, "step": 1611000 }, { "epoch": 0.58, "learning_rate": 8.382483394665027e-06, "loss": 0.6744, "step": 1611500 }, { "epoch": 0.58, "learning_rate": 8.378878828544849e-06, "loss": 0.6768, "step": 1612000 }, { "epoch": 0.58, "learning_rate": 8.37527426242467e-06, "loss": 0.6774, "step": 1612500 }, { "epoch": 0.58, "learning_rate": 8.371669696304491e-06, "loss": 0.6758, "step": 1613000 }, { "epoch": 0.58, "learning_rate": 8.368065130184313e-06, "loss": 0.6846, "step": 1613500 }, { "epoch": 0.58, "learning_rate": 8.364460564064134e-06, "loss": 0.6607, "step": 1614000 }, { "epoch": 0.58, "learning_rate": 8.360855997943956e-06, "loss": 0.6638, "step": 1614500 }, { "epoch": 0.58, "learning_rate": 8.357251431823778e-06, "loss": 0.6799, "step": 1615000 }, { "epoch": 0.58, "learning_rate": 8.3536468657036e-06, "loss": 0.6794, "step": 1615500 }, { "epoch": 0.58, "learning_rate": 8.35004229958342e-06, "loss": 0.6892, "step": 1616000 }, { "epoch": 0.58, "learning_rate": 8.346437733463242e-06, "loss": 0.6731, "step": 1616500 }, { "epoch": 0.58, "learning_rate": 8.342833167343064e-06, "loss": 0.6958, "step": 1617000 }, { "epoch": 0.58, "learning_rate": 8.339228601222886e-06, "loss": 0.6488, "step": 1617500 }, { "epoch": 0.58, "learning_rate": 8.335624035102708e-06, "loss": 0.698, "step": 1618000 }, { "epoch": 0.58, "learning_rate": 8.332019468982528e-06, "loss": 0.6923, "step": 1618500 }, { "epoch": 0.58, "learning_rate": 8.32841490286235e-06, "loss": 0.6609, "step": 1619000 }, { "epoch": 0.58, "learning_rate": 8.324810336742173e-06, "loss": 0.6932, "step": 1619500 }, { "epoch": 0.58, "learning_rate": 8.321205770621995e-06, "loss": 0.6631, "step": 1620000 }, { "epoch": 0.58, "learning_rate": 8.317601204501815e-06, "loss": 0.6666, "step": 1620500 }, { "epoch": 0.58, "learning_rate": 8.313996638381637e-06, "loss": 0.6753, "step": 1621000 }, { "epoch": 0.58, "learning_rate": 8.310392072261459e-06, "loss": 0.683, "step": 1621500 }, { "epoch": 0.58, "learning_rate": 8.306787506141281e-06, "loss": 0.6749, "step": 1622000 }, { "epoch": 0.58, "learning_rate": 8.303182940021101e-06, "loss": 0.6746, "step": 1622500 }, { "epoch": 0.59, "learning_rate": 8.299578373900923e-06, "loss": 0.6711, "step": 1623000 }, { "epoch": 0.59, "learning_rate": 8.295973807780745e-06, "loss": 0.6719, "step": 1623500 }, { "epoch": 0.59, "learning_rate": 8.292369241660566e-06, "loss": 0.6826, "step": 1624000 }, { "epoch": 0.59, "learning_rate": 8.288764675540388e-06, "loss": 0.6917, "step": 1624500 }, { "epoch": 0.59, "learning_rate": 8.28516010942021e-06, "loss": 0.6523, "step": 1625000 }, { "epoch": 0.59, "learning_rate": 8.281555543300032e-06, "loss": 0.7016, "step": 1625500 }, { "epoch": 0.59, "learning_rate": 8.277950977179852e-06, "loss": 0.6774, "step": 1626000 }, { "epoch": 0.59, "learning_rate": 8.274346411059676e-06, "loss": 0.6952, "step": 1626500 }, { "epoch": 0.59, "learning_rate": 8.270741844939496e-06, "loss": 0.6734, "step": 1627000 }, { "epoch": 0.59, "learning_rate": 8.267137278819318e-06, "loss": 0.6729, "step": 1627500 }, { "epoch": 0.59, "learning_rate": 8.26353271269914e-06, "loss": 0.6912, "step": 1628000 }, { "epoch": 0.59, "learning_rate": 8.259928146578962e-06, "loss": 0.7124, "step": 1628500 }, { "epoch": 0.59, "learning_rate": 8.256323580458782e-06, "loss": 0.6722, "step": 1629000 }, { "epoch": 0.59, "learning_rate": 8.252719014338604e-06, "loss": 0.6819, "step": 1629500 }, { "epoch": 0.59, "learning_rate": 8.249114448218426e-06, "loss": 0.6948, "step": 1630000 }, { "epoch": 0.59, "learning_rate": 8.245509882098247e-06, "loss": 0.6679, "step": 1630500 }, { "epoch": 0.59, "learning_rate": 8.241905315978069e-06, "loss": 0.6833, "step": 1631000 }, { "epoch": 0.59, "learning_rate": 8.23830074985789e-06, "loss": 0.6644, "step": 1631500 }, { "epoch": 0.59, "learning_rate": 8.234696183737713e-06, "loss": 0.6845, "step": 1632000 }, { "epoch": 0.59, "learning_rate": 8.231091617617533e-06, "loss": 0.691, "step": 1632500 }, { "epoch": 0.59, "learning_rate": 8.227487051497357e-06, "loss": 0.6611, "step": 1633000 }, { "epoch": 0.59, "learning_rate": 8.223882485377177e-06, "loss": 0.6653, "step": 1633500 }, { "epoch": 0.59, "learning_rate": 8.220277919256997e-06, "loss": 0.6485, "step": 1634000 }, { "epoch": 0.59, "learning_rate": 8.216673353136821e-06, "loss": 0.6842, "step": 1634500 }, { "epoch": 0.59, "learning_rate": 8.213068787016641e-06, "loss": 0.6885, "step": 1635000 }, { "epoch": 0.59, "learning_rate": 8.209464220896463e-06, "loss": 0.6808, "step": 1635500 }, { "epoch": 0.59, "learning_rate": 8.205859654776285e-06, "loss": 0.6736, "step": 1636000 }, { "epoch": 0.59, "learning_rate": 8.202255088656107e-06, "loss": 0.6617, "step": 1636500 }, { "epoch": 0.59, "learning_rate": 8.198650522535928e-06, "loss": 0.6701, "step": 1637000 }, { "epoch": 0.59, "learning_rate": 8.19504595641575e-06, "loss": 0.6666, "step": 1637500 }, { "epoch": 0.59, "learning_rate": 8.191441390295572e-06, "loss": 0.6828, "step": 1638000 }, { "epoch": 0.59, "learning_rate": 8.187836824175394e-06, "loss": 0.6946, "step": 1638500 }, { "epoch": 0.59, "learning_rate": 8.184232258055214e-06, "loss": 0.6824, "step": 1639000 }, { "epoch": 0.59, "learning_rate": 8.180627691935036e-06, "loss": 0.6771, "step": 1639500 }, { "epoch": 0.59, "learning_rate": 8.177023125814858e-06, "loss": 0.6607, "step": 1640000 }, { "epoch": 0.59, "learning_rate": 8.173418559694678e-06, "loss": 0.6766, "step": 1640500 }, { "epoch": 0.59, "learning_rate": 8.169813993574502e-06, "loss": 0.6664, "step": 1641000 }, { "epoch": 0.59, "learning_rate": 8.166209427454322e-06, "loss": 0.6657, "step": 1641500 }, { "epoch": 0.59, "learning_rate": 8.162604861334144e-06, "loss": 0.7006, "step": 1642000 }, { "epoch": 0.59, "learning_rate": 8.159000295213967e-06, "loss": 0.6844, "step": 1642500 }, { "epoch": 0.59, "learning_rate": 8.155395729093789e-06, "loss": 0.6754, "step": 1643000 }, { "epoch": 0.59, "learning_rate": 8.151791162973609e-06, "loss": 0.6691, "step": 1643500 }, { "epoch": 0.59, "learning_rate": 8.148186596853431e-06, "loss": 0.6764, "step": 1644000 }, { "epoch": 0.59, "learning_rate": 8.144582030733253e-06, "loss": 0.6889, "step": 1644500 }, { "epoch": 0.59, "learning_rate": 8.140977464613073e-06, "loss": 0.6895, "step": 1645000 }, { "epoch": 0.59, "learning_rate": 8.137372898492895e-06, "loss": 0.6471, "step": 1645500 }, { "epoch": 0.59, "learning_rate": 8.133768332372717e-06, "loss": 0.6986, "step": 1646000 }, { "epoch": 0.59, "learning_rate": 8.13016376625254e-06, "loss": 0.6708, "step": 1646500 }, { "epoch": 0.59, "learning_rate": 8.12655920013236e-06, "loss": 0.6605, "step": 1647000 }, { "epoch": 0.59, "learning_rate": 8.122954634012182e-06, "loss": 0.6714, "step": 1647500 }, { "epoch": 0.59, "learning_rate": 8.119350067892004e-06, "loss": 0.6731, "step": 1648000 }, { "epoch": 0.59, "learning_rate": 8.115745501771826e-06, "loss": 0.6607, "step": 1648500 }, { "epoch": 0.59, "learning_rate": 8.112140935651646e-06, "loss": 0.6875, "step": 1649000 }, { "epoch": 0.59, "learning_rate": 8.108536369531468e-06, "loss": 0.676, "step": 1649500 }, { "epoch": 0.59, "learning_rate": 8.10493180341129e-06, "loss": 0.6452, "step": 1650000 }, { "epoch": 0.59, "learning_rate": 8.101327237291112e-06, "loss": 0.6804, "step": 1650500 }, { "epoch": 0.6, "learning_rate": 8.097722671170934e-06, "loss": 0.6804, "step": 1651000 }, { "epoch": 0.6, "learning_rate": 8.094118105050754e-06, "loss": 0.6786, "step": 1651500 }, { "epoch": 0.6, "learning_rate": 8.090513538930576e-06, "loss": 0.664, "step": 1652000 }, { "epoch": 0.6, "learning_rate": 8.086908972810398e-06, "loss": 0.6772, "step": 1652500 }, { "epoch": 0.6, "learning_rate": 8.08330440669022e-06, "loss": 0.6757, "step": 1653000 }, { "epoch": 0.6, "learning_rate": 8.07969984057004e-06, "loss": 0.6505, "step": 1653500 }, { "epoch": 0.6, "learning_rate": 8.076095274449863e-06, "loss": 0.6708, "step": 1654000 }, { "epoch": 0.6, "learning_rate": 8.072490708329685e-06, "loss": 0.6799, "step": 1654500 }, { "epoch": 0.6, "learning_rate": 8.068886142209505e-06, "loss": 0.6665, "step": 1655000 }, { "epoch": 0.6, "learning_rate": 8.065281576089327e-06, "loss": 0.6913, "step": 1655500 }, { "epoch": 0.6, "learning_rate": 8.061677009969149e-06, "loss": 0.6639, "step": 1656000 }, { "epoch": 0.6, "learning_rate": 8.058072443848971e-06, "loss": 0.6632, "step": 1656500 }, { "epoch": 0.6, "learning_rate": 8.054467877728791e-06, "loss": 0.6705, "step": 1657000 }, { "epoch": 0.6, "learning_rate": 8.050863311608615e-06, "loss": 0.6854, "step": 1657500 }, { "epoch": 0.6, "learning_rate": 8.047258745488435e-06, "loss": 0.6731, "step": 1658000 }, { "epoch": 0.6, "learning_rate": 8.043654179368257e-06, "loss": 0.6755, "step": 1658500 }, { "epoch": 0.6, "learning_rate": 8.04004961324808e-06, "loss": 0.7133, "step": 1659000 }, { "epoch": 0.6, "learning_rate": 8.0364450471279e-06, "loss": 0.6795, "step": 1659500 }, { "epoch": 0.6, "learning_rate": 8.032840481007722e-06, "loss": 0.696, "step": 1660000 }, { "epoch": 0.6, "learning_rate": 8.029235914887544e-06, "loss": 0.6794, "step": 1660500 }, { "epoch": 0.6, "learning_rate": 8.025631348767366e-06, "loss": 0.6539, "step": 1661000 }, { "epoch": 0.6, "learning_rate": 8.022026782647186e-06, "loss": 0.6987, "step": 1661500 }, { "epoch": 0.6, "learning_rate": 8.018422216527008e-06, "loss": 0.6681, "step": 1662000 }, { "epoch": 0.6, "learning_rate": 8.01481765040683e-06, "loss": 0.6682, "step": 1662500 }, { "epoch": 0.6, "learning_rate": 8.011213084286652e-06, "loss": 0.6715, "step": 1663000 }, { "epoch": 0.6, "learning_rate": 8.007608518166472e-06, "loss": 0.6846, "step": 1663500 }, { "epoch": 0.6, "learning_rate": 8.004003952046296e-06, "loss": 0.6802, "step": 1664000 }, { "epoch": 0.6, "learning_rate": 8.000399385926116e-06, "loss": 0.6517, "step": 1664500 }, { "epoch": 0.6, "learning_rate": 7.996794819805937e-06, "loss": 0.6871, "step": 1665000 }, { "epoch": 0.6, "learning_rate": 7.99319025368576e-06, "loss": 0.6707, "step": 1665500 }, { "epoch": 0.6, "learning_rate": 7.98958568756558e-06, "loss": 0.6775, "step": 1666000 }, { "epoch": 0.6, "learning_rate": 7.985981121445403e-06, "loss": 0.6838, "step": 1666500 }, { "epoch": 0.6, "learning_rate": 7.982376555325225e-06, "loss": 0.6905, "step": 1667000 }, { "epoch": 0.6, "learning_rate": 7.978771989205047e-06, "loss": 0.6745, "step": 1667500 }, { "epoch": 0.6, "learning_rate": 7.975167423084867e-06, "loss": 0.6727, "step": 1668000 }, { "epoch": 0.6, "learning_rate": 7.97156285696469e-06, "loss": 0.7191, "step": 1668500 }, { "epoch": 0.6, "learning_rate": 7.967958290844511e-06, "loss": 0.6885, "step": 1669000 }, { "epoch": 0.6, "learning_rate": 7.964353724724333e-06, "loss": 0.6557, "step": 1669500 }, { "epoch": 0.6, "learning_rate": 7.960749158604154e-06, "loss": 0.6636, "step": 1670000 }, { "epoch": 0.6, "learning_rate": 7.957144592483976e-06, "loss": 0.6885, "step": 1670500 }, { "epoch": 0.6, "learning_rate": 7.953540026363798e-06, "loss": 0.6914, "step": 1671000 }, { "epoch": 0.6, "learning_rate": 7.949935460243618e-06, "loss": 0.6895, "step": 1671500 }, { "epoch": 0.6, "learning_rate": 7.94633089412344e-06, "loss": 0.6899, "step": 1672000 }, { "epoch": 0.6, "learning_rate": 7.942726328003262e-06, "loss": 0.6864, "step": 1672500 }, { "epoch": 0.6, "learning_rate": 7.939121761883084e-06, "loss": 0.6822, "step": 1673000 }, { "epoch": 0.6, "learning_rate": 7.935517195762906e-06, "loss": 0.6698, "step": 1673500 }, { "epoch": 0.6, "learning_rate": 7.931912629642728e-06, "loss": 0.6884, "step": 1674000 }, { "epoch": 0.6, "learning_rate": 7.928308063522548e-06, "loss": 0.6719, "step": 1674500 }, { "epoch": 0.6, "learning_rate": 7.92470349740237e-06, "loss": 0.6791, "step": 1675000 }, { "epoch": 0.6, "learning_rate": 7.921098931282192e-06, "loss": 0.7027, "step": 1675500 }, { "epoch": 0.6, "learning_rate": 7.917494365162013e-06, "loss": 0.6794, "step": 1676000 }, { "epoch": 0.6, "learning_rate": 7.913889799041835e-06, "loss": 0.6554, "step": 1676500 }, { "epoch": 0.6, "learning_rate": 7.910285232921657e-06, "loss": 0.6601, "step": 1677000 }, { "epoch": 0.6, "learning_rate": 7.906680666801479e-06, "loss": 0.6968, "step": 1677500 }, { "epoch": 0.6, "learning_rate": 7.903076100681299e-06, "loss": 0.6621, "step": 1678000 }, { "epoch": 0.61, "learning_rate": 7.899471534561121e-06, "loss": 0.6755, "step": 1678500 }, { "epoch": 0.61, "learning_rate": 7.895866968440943e-06, "loss": 0.6709, "step": 1679000 }, { "epoch": 0.61, "learning_rate": 7.892262402320765e-06, "loss": 0.674, "step": 1679500 }, { "epoch": 0.61, "learning_rate": 7.888657836200585e-06, "loss": 0.67, "step": 1680000 }, { "epoch": 0.61, "learning_rate": 7.885053270080407e-06, "loss": 0.6698, "step": 1680500 }, { "epoch": 0.61, "learning_rate": 7.88144870396023e-06, "loss": 0.6867, "step": 1681000 }, { "epoch": 0.61, "learning_rate": 7.87784413784005e-06, "loss": 0.672, "step": 1681500 }, { "epoch": 0.61, "learning_rate": 7.874239571719873e-06, "loss": 0.6481, "step": 1682000 }, { "epoch": 0.61, "learning_rate": 7.870635005599694e-06, "loss": 0.684, "step": 1682500 }, { "epoch": 0.61, "learning_rate": 7.867030439479516e-06, "loss": 0.6803, "step": 1683000 }, { "epoch": 0.61, "learning_rate": 7.863425873359338e-06, "loss": 0.6654, "step": 1683500 }, { "epoch": 0.61, "learning_rate": 7.85982130723916e-06, "loss": 0.6632, "step": 1684000 }, { "epoch": 0.61, "learning_rate": 7.85621674111898e-06, "loss": 0.6612, "step": 1684500 }, { "epoch": 0.61, "learning_rate": 7.852612174998802e-06, "loss": 0.6691, "step": 1685000 }, { "epoch": 0.61, "learning_rate": 7.849007608878624e-06, "loss": 0.6683, "step": 1685500 }, { "epoch": 0.61, "learning_rate": 7.845403042758444e-06, "loss": 0.6712, "step": 1686000 }, { "epoch": 0.61, "learning_rate": 7.841798476638266e-06, "loss": 0.6612, "step": 1686500 }, { "epoch": 0.61, "learning_rate": 7.838193910518088e-06, "loss": 0.7006, "step": 1687000 }, { "epoch": 0.61, "learning_rate": 7.83458934439791e-06, "loss": 0.6768, "step": 1687500 }, { "epoch": 0.61, "learning_rate": 7.83098477827773e-06, "loss": 0.6722, "step": 1688000 }, { "epoch": 0.61, "learning_rate": 7.827380212157554e-06, "loss": 0.688, "step": 1688500 }, { "epoch": 0.61, "learning_rate": 7.823775646037375e-06, "loss": 0.7072, "step": 1689000 }, { "epoch": 0.61, "learning_rate": 7.820171079917197e-06, "loss": 0.6646, "step": 1689500 }, { "epoch": 0.61, "learning_rate": 7.816566513797019e-06, "loss": 0.6846, "step": 1690000 }, { "epoch": 0.61, "learning_rate": 7.81296194767684e-06, "loss": 0.6559, "step": 1690500 }, { "epoch": 0.61, "learning_rate": 7.809357381556661e-06, "loss": 0.6599, "step": 1691000 }, { "epoch": 0.61, "learning_rate": 7.805752815436483e-06, "loss": 0.6613, "step": 1691500 }, { "epoch": 0.61, "learning_rate": 7.802148249316305e-06, "loss": 0.6676, "step": 1692000 }, { "epoch": 0.61, "learning_rate": 7.798543683196126e-06, "loss": 0.6587, "step": 1692500 }, { "epoch": 0.61, "learning_rate": 7.794939117075948e-06, "loss": 0.6744, "step": 1693000 }, { "epoch": 0.61, "learning_rate": 7.79133455095577e-06, "loss": 0.6703, "step": 1693500 }, { "epoch": 0.61, "learning_rate": 7.787729984835592e-06, "loss": 0.6556, "step": 1694000 }, { "epoch": 0.61, "learning_rate": 7.784125418715412e-06, "loss": 0.6937, "step": 1694500 }, { "epoch": 0.61, "learning_rate": 7.780520852595234e-06, "loss": 0.6794, "step": 1695000 }, { "epoch": 0.61, "learning_rate": 7.776916286475056e-06, "loss": 0.6499, "step": 1695500 }, { "epoch": 0.61, "learning_rate": 7.773311720354876e-06, "loss": 0.6903, "step": 1696000 }, { "epoch": 0.61, "learning_rate": 7.7697071542347e-06, "loss": 0.6619, "step": 1696500 }, { "epoch": 0.61, "learning_rate": 7.76610258811452e-06, "loss": 0.646, "step": 1697000 }, { "epoch": 0.61, "learning_rate": 7.762498021994342e-06, "loss": 0.6987, "step": 1697500 }, { "epoch": 0.61, "learning_rate": 7.758893455874164e-06, "loss": 0.6802, "step": 1698000 }, { "epoch": 0.61, "learning_rate": 7.755288889753986e-06, "loss": 0.7138, "step": 1698500 }, { "epoch": 0.61, "learning_rate": 7.751684323633807e-06, "loss": 0.647, "step": 1699000 }, { "epoch": 0.61, "learning_rate": 7.748079757513629e-06, "loss": 0.6707, "step": 1699500 }, { "epoch": 0.61, "learning_rate": 7.74447519139345e-06, "loss": 0.6731, "step": 1700000 }, { "epoch": 0.61, "learning_rate": 7.740870625273271e-06, "loss": 0.6893, "step": 1700500 }, { "epoch": 0.61, "learning_rate": 7.737266059153093e-06, "loss": 0.682, "step": 1701000 }, { "epoch": 0.61, "learning_rate": 7.733661493032915e-06, "loss": 0.7139, "step": 1701500 }, { "epoch": 0.61, "learning_rate": 7.730056926912737e-06, "loss": 0.6666, "step": 1702000 }, { "epoch": 0.61, "learning_rate": 7.726452360792557e-06, "loss": 0.6618, "step": 1702500 }, { "epoch": 0.61, "learning_rate": 7.72284779467238e-06, "loss": 0.673, "step": 1703000 }, { "epoch": 0.61, "learning_rate": 7.719243228552201e-06, "loss": 0.6556, "step": 1703500 }, { "epoch": 0.61, "learning_rate": 7.715638662432023e-06, "loss": 0.6984, "step": 1704000 }, { "epoch": 0.61, "learning_rate": 7.712034096311844e-06, "loss": 0.6662, "step": 1704500 }, { "epoch": 0.61, "learning_rate": 7.708429530191667e-06, "loss": 0.6694, "step": 1705000 }, { "epoch": 0.61, "learning_rate": 7.704824964071488e-06, "loss": 0.6702, "step": 1705500 }, { "epoch": 0.61, "learning_rate": 7.70122039795131e-06, "loss": 0.6679, "step": 1706000 }, { "epoch": 0.62, "learning_rate": 7.697615831831132e-06, "loss": 0.6526, "step": 1706500 }, { "epoch": 0.62, "learning_rate": 7.694011265710952e-06, "loss": 0.6774, "step": 1707000 }, { "epoch": 0.62, "learning_rate": 7.690406699590774e-06, "loss": 0.6625, "step": 1707500 }, { "epoch": 0.62, "learning_rate": 7.686802133470596e-06, "loss": 0.6315, "step": 1708000 }, { "epoch": 0.62, "learning_rate": 7.683197567350418e-06, "loss": 0.6703, "step": 1708500 }, { "epoch": 0.62, "learning_rate": 7.679593001230238e-06, "loss": 0.6625, "step": 1709000 }, { "epoch": 0.62, "learning_rate": 7.67598843511006e-06, "loss": 0.6666, "step": 1709500 }, { "epoch": 0.62, "learning_rate": 7.672383868989882e-06, "loss": 0.6525, "step": 1710000 }, { "epoch": 0.62, "learning_rate": 7.668779302869704e-06, "loss": 0.6754, "step": 1710500 }, { "epoch": 0.62, "learning_rate": 7.665174736749525e-06, "loss": 0.6663, "step": 1711000 }, { "epoch": 0.62, "learning_rate": 7.661570170629347e-06, "loss": 0.7141, "step": 1711500 }, { "epoch": 0.62, "learning_rate": 7.657965604509169e-06, "loss": 0.6717, "step": 1712000 }, { "epoch": 0.62, "learning_rate": 7.654361038388989e-06, "loss": 0.6606, "step": 1712500 }, { "epoch": 0.62, "learning_rate": 7.650756472268813e-06, "loss": 0.6912, "step": 1713000 }, { "epoch": 0.62, "learning_rate": 7.647151906148633e-06, "loss": 0.6672, "step": 1713500 }, { "epoch": 0.62, "learning_rate": 7.643547340028455e-06, "loss": 0.6826, "step": 1714000 }, { "epoch": 0.62, "learning_rate": 7.639942773908277e-06, "loss": 0.6719, "step": 1714500 }, { "epoch": 0.62, "learning_rate": 7.6363382077881e-06, "loss": 0.6928, "step": 1715000 }, { "epoch": 0.62, "learning_rate": 7.63273364166792e-06, "loss": 0.6833, "step": 1715500 }, { "epoch": 0.62, "learning_rate": 7.629129075547741e-06, "loss": 0.6686, "step": 1716000 }, { "epoch": 0.62, "learning_rate": 7.6255245094275635e-06, "loss": 0.6822, "step": 1716500 }, { "epoch": 0.62, "learning_rate": 7.621919943307384e-06, "loss": 0.6666, "step": 1717000 }, { "epoch": 0.62, "learning_rate": 7.618315377187207e-06, "loss": 0.6747, "step": 1717500 }, { "epoch": 0.62, "learning_rate": 7.614710811067028e-06, "loss": 0.6673, "step": 1718000 }, { "epoch": 0.62, "learning_rate": 7.61110624494685e-06, "loss": 0.659, "step": 1718500 }, { "epoch": 0.62, "learning_rate": 7.607501678826671e-06, "loss": 0.6745, "step": 1719000 }, { "epoch": 0.62, "learning_rate": 7.603897112706493e-06, "loss": 0.6462, "step": 1719500 }, { "epoch": 0.62, "learning_rate": 7.600292546586314e-06, "loss": 0.6558, "step": 1720000 }, { "epoch": 0.62, "learning_rate": 7.596687980466136e-06, "loss": 0.6862, "step": 1720500 }, { "epoch": 0.62, "learning_rate": 7.5930834143459574e-06, "loss": 0.6659, "step": 1721000 }, { "epoch": 0.62, "learning_rate": 7.589478848225779e-06, "loss": 0.6778, "step": 1721500 }, { "epoch": 0.62, "learning_rate": 7.585874282105601e-06, "loss": 0.6634, "step": 1722000 }, { "epoch": 0.62, "learning_rate": 7.582269715985422e-06, "loss": 0.6677, "step": 1722500 }, { "epoch": 0.62, "learning_rate": 7.578665149865244e-06, "loss": 0.6543, "step": 1723000 }, { "epoch": 0.62, "learning_rate": 7.575060583745065e-06, "loss": 0.6667, "step": 1723500 }, { "epoch": 0.62, "learning_rate": 7.571456017624888e-06, "loss": 0.6719, "step": 1724000 }, { "epoch": 0.62, "learning_rate": 7.567851451504709e-06, "loss": 0.6763, "step": 1724500 }, { "epoch": 0.62, "learning_rate": 7.564246885384531e-06, "loss": 0.6796, "step": 1725000 }, { "epoch": 0.62, "learning_rate": 7.560642319264352e-06, "loss": 0.6692, "step": 1725500 }, { "epoch": 0.62, "learning_rate": 7.557037753144173e-06, "loss": 0.6482, "step": 1726000 }, { "epoch": 0.62, "learning_rate": 7.553433187023995e-06, "loss": 0.6903, "step": 1726500 }, { "epoch": 0.62, "learning_rate": 7.5498286209038165e-06, "loss": 0.6946, "step": 1727000 }, { "epoch": 0.62, "learning_rate": 7.5462240547836385e-06, "loss": 0.6884, "step": 1727500 }, { "epoch": 0.62, "learning_rate": 7.54261948866346e-06, "loss": 0.6621, "step": 1728000 }, { "epoch": 0.62, "learning_rate": 7.539014922543282e-06, "loss": 0.6533, "step": 1728500 }, { "epoch": 0.62, "learning_rate": 7.535410356423103e-06, "loss": 0.6805, "step": 1729000 }, { "epoch": 0.62, "learning_rate": 7.531805790302925e-06, "loss": 0.6709, "step": 1729500 }, { "epoch": 0.62, "learning_rate": 7.528201224182746e-06, "loss": 0.684, "step": 1730000 }, { "epoch": 0.62, "learning_rate": 7.524596658062568e-06, "loss": 0.6875, "step": 1730500 }, { "epoch": 0.62, "learning_rate": 7.520992091942389e-06, "loss": 0.6833, "step": 1731000 }, { "epoch": 0.62, "learning_rate": 7.51738752582221e-06, "loss": 0.6712, "step": 1731500 }, { "epoch": 0.62, "learning_rate": 7.513782959702033e-06, "loss": 0.6824, "step": 1732000 }, { "epoch": 0.62, "learning_rate": 7.510178393581854e-06, "loss": 0.6908, "step": 1732500 }, { "epoch": 0.62, "learning_rate": 7.5065738274616764e-06, "loss": 0.6588, "step": 1733000 }, { "epoch": 0.62, "learning_rate": 7.502969261341498e-06, "loss": 0.641, "step": 1733500 }, { "epoch": 0.63, "learning_rate": 7.49936469522132e-06, "loss": 0.6788, "step": 1734000 }, { "epoch": 0.63, "learning_rate": 7.495760129101141e-06, "loss": 0.649, "step": 1734500 }, { "epoch": 0.63, "learning_rate": 7.492155562980963e-06, "loss": 0.6563, "step": 1735000 }, { "epoch": 0.63, "learning_rate": 7.488550996860784e-06, "loss": 0.6658, "step": 1735500 }, { "epoch": 0.63, "learning_rate": 7.484946430740606e-06, "loss": 0.651, "step": 1736000 }, { "epoch": 0.63, "learning_rate": 7.481341864620427e-06, "loss": 0.6612, "step": 1736500 }, { "epoch": 0.63, "learning_rate": 7.477737298500248e-06, "loss": 0.683, "step": 1737000 }, { "epoch": 0.63, "learning_rate": 7.47413273238007e-06, "loss": 0.7018, "step": 1737500 }, { "epoch": 0.63, "learning_rate": 7.4705281662598915e-06, "loss": 0.6803, "step": 1738000 }, { "epoch": 0.63, "learning_rate": 7.4669236001397135e-06, "loss": 0.669, "step": 1738500 }, { "epoch": 0.63, "learning_rate": 7.463319034019535e-06, "loss": 0.6867, "step": 1739000 }, { "epoch": 0.63, "learning_rate": 7.4597144678993575e-06, "loss": 0.6878, "step": 1739500 }, { "epoch": 0.63, "learning_rate": 7.456109901779178e-06, "loss": 0.6512, "step": 1740000 }, { "epoch": 0.63, "learning_rate": 7.452505335659001e-06, "loss": 0.6595, "step": 1740500 }, { "epoch": 0.63, "learning_rate": 7.448900769538822e-06, "loss": 0.6795, "step": 1741000 }, { "epoch": 0.63, "learning_rate": 7.445296203418643e-06, "loss": 0.6916, "step": 1741500 }, { "epoch": 0.63, "learning_rate": 7.441691637298465e-06, "loss": 0.7061, "step": 1742000 }, { "epoch": 0.63, "learning_rate": 7.438087071178286e-06, "loss": 0.6482, "step": 1742500 }, { "epoch": 0.63, "learning_rate": 7.434482505058108e-06, "loss": 0.6858, "step": 1743000 }, { "epoch": 0.63, "learning_rate": 7.430877938937929e-06, "loss": 0.6639, "step": 1743500 }, { "epoch": 0.63, "learning_rate": 7.4272733728177514e-06, "loss": 0.6705, "step": 1744000 }, { "epoch": 0.63, "learning_rate": 7.423668806697573e-06, "loss": 0.6585, "step": 1744500 }, { "epoch": 0.63, "learning_rate": 7.420064240577395e-06, "loss": 0.6639, "step": 1745000 }, { "epoch": 0.63, "learning_rate": 7.416459674457216e-06, "loss": 0.6578, "step": 1745500 }, { "epoch": 0.63, "learning_rate": 7.412855108337038e-06, "loss": 0.6842, "step": 1746000 }, { "epoch": 0.63, "learning_rate": 7.409250542216859e-06, "loss": 0.6503, "step": 1746500 }, { "epoch": 0.63, "learning_rate": 7.40564597609668e-06, "loss": 0.6852, "step": 1747000 }, { "epoch": 0.63, "learning_rate": 7.402041409976502e-06, "loss": 0.6534, "step": 1747500 }, { "epoch": 0.63, "learning_rate": 7.398436843856323e-06, "loss": 0.6807, "step": 1748000 }, { "epoch": 0.63, "learning_rate": 7.394832277736146e-06, "loss": 0.6886, "step": 1748500 }, { "epoch": 0.63, "learning_rate": 7.391227711615967e-06, "loss": 0.6947, "step": 1749000 }, { "epoch": 0.63, "learning_rate": 7.387623145495789e-06, "loss": 0.6801, "step": 1749500 }, { "epoch": 0.63, "learning_rate": 7.3840185793756105e-06, "loss": 0.6707, "step": 1750000 }, { "epoch": 0.63, "learning_rate": 7.3804140132554325e-06, "loss": 0.644, "step": 1750500 }, { "epoch": 0.63, "learning_rate": 7.376809447135254e-06, "loss": 0.6645, "step": 1751000 }, { "epoch": 0.63, "learning_rate": 7.373204881015076e-06, "loss": 0.7022, "step": 1751500 }, { "epoch": 0.63, "learning_rate": 7.369600314894897e-06, "loss": 0.6569, "step": 1752000 }, { "epoch": 0.63, "learning_rate": 7.365995748774718e-06, "loss": 0.651, "step": 1752500 }, { "epoch": 0.63, "learning_rate": 7.36239118265454e-06, "loss": 0.6754, "step": 1753000 }, { "epoch": 0.63, "learning_rate": 7.358786616534361e-06, "loss": 0.6749, "step": 1753500 }, { "epoch": 0.63, "learning_rate": 7.355182050414183e-06, "loss": 0.6733, "step": 1754000 }, { "epoch": 0.63, "learning_rate": 7.351577484294004e-06, "loss": 0.679, "step": 1754500 }, { "epoch": 0.63, "learning_rate": 7.347972918173827e-06, "loss": 0.6648, "step": 1755000 }, { "epoch": 0.63, "learning_rate": 7.344368352053648e-06, "loss": 0.6374, "step": 1755500 }, { "epoch": 0.63, "learning_rate": 7.3407637859334704e-06, "loss": 0.687, "step": 1756000 }, { "epoch": 0.63, "learning_rate": 7.337159219813292e-06, "loss": 0.6376, "step": 1756500 }, { "epoch": 0.63, "learning_rate": 7.333554653693113e-06, "loss": 0.6668, "step": 1757000 }, { "epoch": 0.63, "learning_rate": 7.329950087572935e-06, "loss": 0.686, "step": 1757500 }, { "epoch": 0.63, "learning_rate": 7.326345521452756e-06, "loss": 0.6517, "step": 1758000 }, { "epoch": 0.63, "learning_rate": 7.322740955332578e-06, "loss": 0.6753, "step": 1758500 }, { "epoch": 0.63, "learning_rate": 7.319136389212399e-06, "loss": 0.6629, "step": 1759000 }, { "epoch": 0.63, "learning_rate": 7.315531823092221e-06, "loss": 0.6559, "step": 1759500 }, { "epoch": 0.63, "learning_rate": 7.311927256972042e-06, "loss": 0.6672, "step": 1760000 }, { "epoch": 0.63, "learning_rate": 7.308322690851864e-06, "loss": 0.6804, "step": 1760500 }, { "epoch": 0.63, "learning_rate": 7.3047181247316855e-06, "loss": 0.6782, "step": 1761000 }, { "epoch": 0.63, "learning_rate": 7.3011135586115075e-06, "loss": 0.6686, "step": 1761500 }, { "epoch": 0.64, "learning_rate": 7.297508992491329e-06, "loss": 0.652, "step": 1762000 }, { "epoch": 0.64, "learning_rate": 7.29390442637115e-06, "loss": 0.6532, "step": 1762500 }, { "epoch": 0.64, "learning_rate": 7.290299860250972e-06, "loss": 0.6489, "step": 1763000 }, { "epoch": 0.64, "learning_rate": 7.286695294130793e-06, "loss": 0.6507, "step": 1763500 }, { "epoch": 0.64, "learning_rate": 7.283090728010616e-06, "loss": 0.655, "step": 1764000 }, { "epoch": 0.64, "learning_rate": 7.279486161890437e-06, "loss": 0.6526, "step": 1764500 }, { "epoch": 0.64, "learning_rate": 7.275881595770259e-06, "loss": 0.6666, "step": 1765000 }, { "epoch": 0.64, "learning_rate": 7.27227702965008e-06, "loss": 0.6688, "step": 1765500 }, { "epoch": 0.64, "learning_rate": 7.268672463529902e-06, "loss": 0.6403, "step": 1766000 }, { "epoch": 0.64, "learning_rate": 7.265067897409723e-06, "loss": 0.6615, "step": 1766500 }, { "epoch": 0.64, "learning_rate": 7.261463331289545e-06, "loss": 0.6952, "step": 1767000 }, { "epoch": 0.64, "learning_rate": 7.257858765169367e-06, "loss": 0.6758, "step": 1767500 }, { "epoch": 0.64, "learning_rate": 7.254254199049188e-06, "loss": 0.675, "step": 1768000 }, { "epoch": 0.64, "learning_rate": 7.25064963292901e-06, "loss": 0.693, "step": 1768500 }, { "epoch": 0.64, "learning_rate": 7.247045066808831e-06, "loss": 0.6678, "step": 1769000 }, { "epoch": 0.64, "learning_rate": 7.243440500688653e-06, "loss": 0.6703, "step": 1769500 }, { "epoch": 0.64, "learning_rate": 7.239835934568474e-06, "loss": 0.6681, "step": 1770000 }, { "epoch": 0.64, "learning_rate": 7.236231368448296e-06, "loss": 0.6643, "step": 1770500 }, { "epoch": 0.64, "learning_rate": 7.232626802328117e-06, "loss": 0.6827, "step": 1771000 }, { "epoch": 0.64, "learning_rate": 7.22902223620794e-06, "loss": 0.6636, "step": 1771500 }, { "epoch": 0.64, "learning_rate": 7.225417670087761e-06, "loss": 0.6845, "step": 1772000 }, { "epoch": 0.64, "learning_rate": 7.221813103967582e-06, "loss": 0.663, "step": 1772500 }, { "epoch": 0.64, "learning_rate": 7.2182085378474045e-06, "loss": 0.6597, "step": 1773000 }, { "epoch": 0.64, "learning_rate": 7.214603971727226e-06, "loss": 0.6834, "step": 1773500 }, { "epoch": 0.64, "learning_rate": 7.210999405607048e-06, "loss": 0.6835, "step": 1774000 }, { "epoch": 0.64, "learning_rate": 7.207394839486869e-06, "loss": 0.6499, "step": 1774500 }, { "epoch": 0.64, "learning_rate": 7.203790273366691e-06, "loss": 0.6842, "step": 1775000 }, { "epoch": 0.64, "learning_rate": 7.200185707246512e-06, "loss": 0.6658, "step": 1775500 }, { "epoch": 0.64, "learning_rate": 7.196581141126334e-06, "loss": 0.649, "step": 1776000 }, { "epoch": 0.64, "learning_rate": 7.192976575006155e-06, "loss": 0.6831, "step": 1776500 }, { "epoch": 0.64, "learning_rate": 7.189372008885977e-06, "loss": 0.6241, "step": 1777000 }, { "epoch": 0.64, "learning_rate": 7.185767442765798e-06, "loss": 0.6682, "step": 1777500 }, { "epoch": 0.64, "learning_rate": 7.1821628766456196e-06, "loss": 0.6475, "step": 1778000 }, { "epoch": 0.64, "learning_rate": 7.178558310525442e-06, "loss": 0.6709, "step": 1778500 }, { "epoch": 0.64, "learning_rate": 7.174953744405263e-06, "loss": 0.6629, "step": 1779000 }, { "epoch": 0.64, "learning_rate": 7.171349178285086e-06, "loss": 0.6604, "step": 1779500 }, { "epoch": 0.64, "learning_rate": 7.167744612164907e-06, "loss": 0.6541, "step": 1780000 }, { "epoch": 0.64, "learning_rate": 7.164140046044729e-06, "loss": 0.6805, "step": 1780500 }, { "epoch": 0.64, "learning_rate": 7.16053547992455e-06, "loss": 0.6504, "step": 1781000 }, { "epoch": 0.64, "learning_rate": 7.156930913804372e-06, "loss": 0.6643, "step": 1781500 }, { "epoch": 0.64, "learning_rate": 7.153326347684193e-06, "loss": 0.6217, "step": 1782000 }, { "epoch": 0.64, "learning_rate": 7.149721781564014e-06, "loss": 0.6566, "step": 1782500 }, { "epoch": 0.64, "learning_rate": 7.146117215443836e-06, "loss": 0.7179, "step": 1783000 }, { "epoch": 0.64, "learning_rate": 7.1425126493236575e-06, "loss": 0.6495, "step": 1783500 }, { "epoch": 0.64, "learning_rate": 7.1389080832034795e-06, "loss": 0.6647, "step": 1784000 }, { "epoch": 0.64, "learning_rate": 7.135303517083301e-06, "loss": 0.6689, "step": 1784500 }, { "epoch": 0.64, "learning_rate": 7.131698950963123e-06, "loss": 0.6623, "step": 1785000 }, { "epoch": 0.64, "learning_rate": 7.128094384842944e-06, "loss": 0.6859, "step": 1785500 }, { "epoch": 0.64, "learning_rate": 7.124489818722766e-06, "loss": 0.6835, "step": 1786000 }, { "epoch": 0.64, "learning_rate": 7.120885252602587e-06, "loss": 0.6671, "step": 1786500 }, { "epoch": 0.64, "learning_rate": 7.11728068648241e-06, "loss": 0.6689, "step": 1787000 }, { "epoch": 0.64, "learning_rate": 7.113676120362231e-06, "loss": 0.6404, "step": 1787500 }, { "epoch": 0.64, "learning_rate": 7.110071554242051e-06, "loss": 0.6795, "step": 1788000 }, { "epoch": 0.64, "learning_rate": 7.106466988121874e-06, "loss": 0.6848, "step": 1788500 }, { "epoch": 0.64, "learning_rate": 7.102862422001695e-06, "loss": 0.6607, "step": 1789000 }, { "epoch": 0.65, "learning_rate": 7.099257855881517e-06, "loss": 0.6424, "step": 1789500 }, { "epoch": 0.65, "learning_rate": 7.095653289761339e-06, "loss": 0.6563, "step": 1790000 }, { "epoch": 0.65, "learning_rate": 7.092048723641161e-06, "loss": 0.6901, "step": 1790500 }, { "epoch": 0.65, "learning_rate": 7.088444157520982e-06, "loss": 0.6838, "step": 1791000 }, { "epoch": 0.65, "learning_rate": 7.084839591400804e-06, "loss": 0.6637, "step": 1791500 }, { "epoch": 0.65, "learning_rate": 7.081235025280625e-06, "loss": 0.623, "step": 1792000 }, { "epoch": 0.65, "learning_rate": 7.077630459160447e-06, "loss": 0.663, "step": 1792500 }, { "epoch": 0.65, "learning_rate": 7.074025893040268e-06, "loss": 0.6644, "step": 1793000 }, { "epoch": 0.65, "learning_rate": 7.070421326920089e-06, "loss": 0.6739, "step": 1793500 }, { "epoch": 0.65, "learning_rate": 7.066816760799911e-06, "loss": 0.6758, "step": 1794000 }, { "epoch": 0.65, "learning_rate": 7.0632121946797325e-06, "loss": 0.6881, "step": 1794500 }, { "epoch": 0.65, "learning_rate": 7.059607628559555e-06, "loss": 0.6572, "step": 1795000 }, { "epoch": 0.65, "learning_rate": 7.056003062439376e-06, "loss": 0.6371, "step": 1795500 }, { "epoch": 0.65, "learning_rate": 7.0523984963191985e-06, "loss": 0.695, "step": 1796000 }, { "epoch": 0.65, "learning_rate": 7.04879393019902e-06, "loss": 0.6845, "step": 1796500 }, { "epoch": 0.65, "learning_rate": 7.045189364078842e-06, "loss": 0.66, "step": 1797000 }, { "epoch": 0.65, "learning_rate": 7.041584797958663e-06, "loss": 0.6676, "step": 1797500 }, { "epoch": 0.65, "learning_rate": 7.037980231838484e-06, "loss": 0.6866, "step": 1798000 }, { "epoch": 0.65, "learning_rate": 7.034375665718306e-06, "loss": 0.6621, "step": 1798500 }, { "epoch": 0.65, "learning_rate": 7.030771099598127e-06, "loss": 0.6476, "step": 1799000 }, { "epoch": 0.65, "learning_rate": 7.027166533477949e-06, "loss": 0.6475, "step": 1799500 }, { "epoch": 0.65, "learning_rate": 7.02356196735777e-06, "loss": 0.6792, "step": 1800000 }, { "epoch": 0.65, "learning_rate": 7.019957401237592e-06, "loss": 0.6425, "step": 1800500 }, { "epoch": 0.65, "learning_rate": 7.0163528351174136e-06, "loss": 0.6604, "step": 1801000 }, { "epoch": 0.65, "learning_rate": 7.0127482689972356e-06, "loss": 0.6572, "step": 1801500 }, { "epoch": 0.65, "learning_rate": 7.009143702877057e-06, "loss": 0.6328, "step": 1802000 }, { "epoch": 0.65, "learning_rate": 7.00553913675688e-06, "loss": 0.6564, "step": 1802500 }, { "epoch": 0.65, "learning_rate": 7.0019345706367e-06, "loss": 0.6268, "step": 1803000 }, { "epoch": 0.65, "learning_rate": 6.998330004516521e-06, "loss": 0.6682, "step": 1803500 }, { "epoch": 0.65, "learning_rate": 6.994725438396344e-06, "loss": 0.6596, "step": 1804000 }, { "epoch": 0.65, "learning_rate": 6.991120872276165e-06, "loss": 0.7018, "step": 1804500 }, { "epoch": 0.65, "learning_rate": 6.987516306155987e-06, "loss": 0.6599, "step": 1805000 }, { "epoch": 0.65, "learning_rate": 6.983911740035808e-06, "loss": 0.6755, "step": 1805500 }, { "epoch": 0.65, "learning_rate": 6.98030717391563e-06, "loss": 0.6448, "step": 1806000 }, { "epoch": 0.65, "learning_rate": 6.9767026077954515e-06, "loss": 0.6799, "step": 1806500 }, { "epoch": 0.65, "learning_rate": 6.9730980416752735e-06, "loss": 0.6616, "step": 1807000 }, { "epoch": 0.65, "learning_rate": 6.969493475555095e-06, "loss": 0.6732, "step": 1807500 }, { "epoch": 0.65, "learning_rate": 6.965888909434916e-06, "loss": 0.6729, "step": 1808000 }, { "epoch": 0.65, "learning_rate": 6.962284343314738e-06, "loss": 0.6783, "step": 1808500 }, { "epoch": 0.65, "learning_rate": 6.958679777194559e-06, "loss": 0.6613, "step": 1809000 }, { "epoch": 0.65, "learning_rate": 6.955075211074381e-06, "loss": 0.6943, "step": 1809500 }, { "epoch": 0.65, "learning_rate": 6.951470644954202e-06, "loss": 0.6644, "step": 1810000 }, { "epoch": 0.65, "learning_rate": 6.947866078834025e-06, "loss": 0.6514, "step": 1810500 }, { "epoch": 0.65, "learning_rate": 6.944261512713845e-06, "loss": 0.6637, "step": 1811000 }, { "epoch": 0.65, "learning_rate": 6.940656946593668e-06, "loss": 0.6755, "step": 1811500 }, { "epoch": 0.65, "learning_rate": 6.937052380473489e-06, "loss": 0.6586, "step": 1812000 }, { "epoch": 0.65, "learning_rate": 6.933447814353311e-06, "loss": 0.6848, "step": 1812500 }, { "epoch": 0.65, "learning_rate": 6.9298432482331326e-06, "loss": 0.6498, "step": 1813000 }, { "epoch": 0.65, "learning_rate": 6.926238682112954e-06, "loss": 0.6686, "step": 1813500 }, { "epoch": 0.65, "learning_rate": 6.922634115992776e-06, "loss": 0.6594, "step": 1814000 }, { "epoch": 0.65, "learning_rate": 6.919029549872597e-06, "loss": 0.6627, "step": 1814500 }, { "epoch": 0.65, "learning_rate": 6.915424983752419e-06, "loss": 0.6982, "step": 1815000 }, { "epoch": 0.65, "learning_rate": 6.91182041763224e-06, "loss": 0.6448, "step": 1815500 }, { "epoch": 0.65, "learning_rate": 6.908215851512062e-06, "loss": 0.6752, "step": 1816000 }, { "epoch": 0.65, "learning_rate": 6.904611285391883e-06, "loss": 0.6723, "step": 1816500 }, { "epoch": 0.65, "learning_rate": 6.901006719271705e-06, "loss": 0.6753, "step": 1817000 }, { "epoch": 0.66, "learning_rate": 6.8974021531515265e-06, "loss": 0.6539, "step": 1817500 }, { "epoch": 0.66, "learning_rate": 6.893797587031349e-06, "loss": 0.647, "step": 1818000 }, { "epoch": 0.66, "learning_rate": 6.89019302091117e-06, "loss": 0.6795, "step": 1818500 }, { "epoch": 0.66, "learning_rate": 6.886588454790991e-06, "loss": 0.6567, "step": 1819000 }, { "epoch": 0.66, "learning_rate": 6.882983888670814e-06, "loss": 0.6637, "step": 1819500 }, { "epoch": 0.66, "learning_rate": 6.879379322550635e-06, "loss": 0.6395, "step": 1820000 }, { "epoch": 0.66, "learning_rate": 6.875774756430457e-06, "loss": 0.6488, "step": 1820500 }, { "epoch": 0.66, "learning_rate": 6.872170190310278e-06, "loss": 0.683, "step": 1821000 }, { "epoch": 0.66, "learning_rate": 6.8685656241901e-06, "loss": 0.6595, "step": 1821500 }, { "epoch": 0.66, "learning_rate": 6.864961058069921e-06, "loss": 0.6341, "step": 1822000 }, { "epoch": 0.66, "learning_rate": 6.861356491949743e-06, "loss": 0.6761, "step": 1822500 }, { "epoch": 0.66, "learning_rate": 6.857751925829564e-06, "loss": 0.6686, "step": 1823000 }, { "epoch": 0.66, "learning_rate": 6.8541473597093855e-06, "loss": 0.6345, "step": 1823500 }, { "epoch": 0.66, "learning_rate": 6.8505427935892076e-06, "loss": 0.6679, "step": 1824000 }, { "epoch": 0.66, "learning_rate": 6.846938227469029e-06, "loss": 0.6781, "step": 1824500 }, { "epoch": 0.66, "learning_rate": 6.843333661348851e-06, "loss": 0.6752, "step": 1825000 }, { "epoch": 0.66, "learning_rate": 6.839729095228672e-06, "loss": 0.6482, "step": 1825500 }, { "epoch": 0.66, "learning_rate": 6.836124529108494e-06, "loss": 0.6493, "step": 1826000 }, { "epoch": 0.66, "learning_rate": 6.832519962988315e-06, "loss": 0.6524, "step": 1826500 }, { "epoch": 0.66, "learning_rate": 6.828915396868138e-06, "loss": 0.6723, "step": 1827000 }, { "epoch": 0.66, "learning_rate": 6.825310830747959e-06, "loss": 0.6539, "step": 1827500 }, { "epoch": 0.66, "learning_rate": 6.821706264627781e-06, "loss": 0.6527, "step": 1828000 }, { "epoch": 0.66, "learning_rate": 6.818101698507602e-06, "loss": 0.6626, "step": 1828500 }, { "epoch": 0.66, "learning_rate": 6.8144971323874235e-06, "loss": 0.6633, "step": 1829000 }, { "epoch": 0.66, "learning_rate": 6.8108925662672455e-06, "loss": 0.6636, "step": 1829500 }, { "epoch": 0.66, "learning_rate": 6.807288000147067e-06, "loss": 0.6274, "step": 1830000 }, { "epoch": 0.66, "learning_rate": 6.803683434026889e-06, "loss": 0.6835, "step": 1830500 }, { "epoch": 0.66, "learning_rate": 6.80007886790671e-06, "loss": 0.6576, "step": 1831000 }, { "epoch": 0.66, "learning_rate": 6.796474301786532e-06, "loss": 0.662, "step": 1831500 }, { "epoch": 0.66, "learning_rate": 6.792869735666353e-06, "loss": 0.641, "step": 1832000 }, { "epoch": 0.66, "learning_rate": 6.789265169546175e-06, "loss": 0.6616, "step": 1832500 }, { "epoch": 0.66, "learning_rate": 6.785660603425996e-06, "loss": 0.6744, "step": 1833000 }, { "epoch": 0.66, "learning_rate": 6.782056037305817e-06, "loss": 0.7051, "step": 1833500 }, { "epoch": 0.66, "learning_rate": 6.778451471185639e-06, "loss": 0.6626, "step": 1834000 }, { "epoch": 0.66, "learning_rate": 6.7748469050654605e-06, "loss": 0.6752, "step": 1834500 }, { "epoch": 0.66, "learning_rate": 6.771242338945283e-06, "loss": 0.6481, "step": 1835000 }, { "epoch": 0.66, "learning_rate": 6.7676377728251046e-06, "loss": 0.6931, "step": 1835500 }, { "epoch": 0.66, "learning_rate": 6.7640332067049266e-06, "loss": 0.6682, "step": 1836000 }, { "epoch": 0.66, "learning_rate": 6.760428640584748e-06, "loss": 0.6737, "step": 1836500 }, { "epoch": 0.66, "learning_rate": 6.75682407446457e-06, "loss": 0.6747, "step": 1837000 }, { "epoch": 0.66, "learning_rate": 6.753219508344391e-06, "loss": 0.6544, "step": 1837500 }, { "epoch": 0.66, "learning_rate": 6.749614942224213e-06, "loss": 0.6618, "step": 1838000 }, { "epoch": 0.66, "learning_rate": 6.746010376104034e-06, "loss": 0.6633, "step": 1838500 }, { "epoch": 0.66, "learning_rate": 6.742405809983855e-06, "loss": 0.6813, "step": 1839000 }, { "epoch": 0.66, "learning_rate": 6.738801243863677e-06, "loss": 0.6623, "step": 1839500 }, { "epoch": 0.66, "learning_rate": 6.7351966777434984e-06, "loss": 0.6609, "step": 1840000 }, { "epoch": 0.66, "learning_rate": 6.7315921116233205e-06, "loss": 0.6661, "step": 1840500 }, { "epoch": 0.66, "learning_rate": 6.727987545503142e-06, "loss": 0.6447, "step": 1841000 }, { "epoch": 0.66, "learning_rate": 6.724382979382964e-06, "loss": 0.6643, "step": 1841500 }, { "epoch": 0.66, "learning_rate": 6.720778413262785e-06, "loss": 0.6706, "step": 1842000 }, { "epoch": 0.66, "learning_rate": 6.717173847142608e-06, "loss": 0.6637, "step": 1842500 }, { "epoch": 0.66, "learning_rate": 6.713569281022429e-06, "loss": 0.6651, "step": 1843000 }, { "epoch": 0.66, "learning_rate": 6.709964714902251e-06, "loss": 0.6581, "step": 1843500 }, { "epoch": 0.66, "learning_rate": 6.706360148782072e-06, "loss": 0.6681, "step": 1844000 }, { "epoch": 0.66, "learning_rate": 6.702755582661893e-06, "loss": 0.664, "step": 1844500 }, { "epoch": 0.67, "learning_rate": 6.699151016541715e-06, "loss": 0.6575, "step": 1845000 }, { "epoch": 0.67, "learning_rate": 6.695546450421536e-06, "loss": 0.6514, "step": 1845500 }, { "epoch": 0.67, "learning_rate": 6.691941884301358e-06, "loss": 0.6761, "step": 1846000 }, { "epoch": 0.67, "learning_rate": 6.6883373181811795e-06, "loss": 0.6656, "step": 1846500 }, { "epoch": 0.67, "learning_rate": 6.6847327520610016e-06, "loss": 0.6955, "step": 1847000 }, { "epoch": 0.67, "learning_rate": 6.681128185940823e-06, "loss": 0.6382, "step": 1847500 }, { "epoch": 0.67, "learning_rate": 6.677523619820645e-06, "loss": 0.6564, "step": 1848000 }, { "epoch": 0.67, "learning_rate": 6.673919053700466e-06, "loss": 0.6559, "step": 1848500 }, { "epoch": 0.67, "learning_rate": 6.670314487580287e-06, "loss": 0.6652, "step": 1849000 }, { "epoch": 0.67, "learning_rate": 6.666709921460109e-06, "loss": 0.6735, "step": 1849500 }, { "epoch": 0.67, "learning_rate": 6.66310535533993e-06, "loss": 0.6661, "step": 1850000 }, { "epoch": 0.67, "learning_rate": 6.659500789219753e-06, "loss": 0.645, "step": 1850500 }, { "epoch": 0.67, "learning_rate": 6.6558962230995734e-06, "loss": 0.6464, "step": 1851000 }, { "epoch": 0.67, "learning_rate": 6.652291656979396e-06, "loss": 0.6597, "step": 1851500 }, { "epoch": 0.67, "learning_rate": 6.6486870908592175e-06, "loss": 0.6849, "step": 1852000 }, { "epoch": 0.67, "learning_rate": 6.6450825247390395e-06, "loss": 0.6448, "step": 1852500 }, { "epoch": 0.67, "learning_rate": 6.641477958618861e-06, "loss": 0.6392, "step": 1853000 }, { "epoch": 0.67, "learning_rate": 6.637873392498683e-06, "loss": 0.6509, "step": 1853500 }, { "epoch": 0.67, "learning_rate": 6.634268826378504e-06, "loss": 0.68, "step": 1854000 }, { "epoch": 0.67, "learning_rate": 6.630664260258325e-06, "loss": 0.6454, "step": 1854500 }, { "epoch": 0.67, "learning_rate": 6.627059694138147e-06, "loss": 0.689, "step": 1855000 }, { "epoch": 0.67, "learning_rate": 6.623455128017968e-06, "loss": 0.6793, "step": 1855500 }, { "epoch": 0.67, "learning_rate": 6.61985056189779e-06, "loss": 0.6483, "step": 1856000 }, { "epoch": 0.67, "learning_rate": 6.616245995777611e-06, "loss": 0.6364, "step": 1856500 }, { "epoch": 0.67, "learning_rate": 6.612641429657433e-06, "loss": 0.6664, "step": 1857000 }, { "epoch": 0.67, "learning_rate": 6.6090368635372545e-06, "loss": 0.6585, "step": 1857500 }, { "epoch": 0.67, "learning_rate": 6.605432297417077e-06, "loss": 0.63, "step": 1858000 }, { "epoch": 0.67, "learning_rate": 6.6018277312968986e-06, "loss": 0.677, "step": 1858500 }, { "epoch": 0.67, "learning_rate": 6.5982231651767206e-06, "loss": 0.6674, "step": 1859000 }, { "epoch": 0.67, "learning_rate": 6.594618599056542e-06, "loss": 0.6637, "step": 1859500 }, { "epoch": 0.67, "learning_rate": 6.591014032936363e-06, "loss": 0.637, "step": 1860000 }, { "epoch": 0.67, "learning_rate": 6.587409466816185e-06, "loss": 0.6436, "step": 1860500 }, { "epoch": 0.67, "learning_rate": 6.583804900696006e-06, "loss": 0.6481, "step": 1861000 }, { "epoch": 0.67, "learning_rate": 6.580200334575828e-06, "loss": 0.6387, "step": 1861500 }, { "epoch": 0.67, "learning_rate": 6.576595768455649e-06, "loss": 0.6651, "step": 1862000 }, { "epoch": 0.67, "learning_rate": 6.572991202335471e-06, "loss": 0.649, "step": 1862500 }, { "epoch": 0.67, "learning_rate": 6.5693866362152924e-06, "loss": 0.6641, "step": 1863000 }, { "epoch": 0.67, "learning_rate": 6.5657820700951145e-06, "loss": 0.6757, "step": 1863500 }, { "epoch": 0.67, "learning_rate": 6.562177503974936e-06, "loss": 0.6613, "step": 1864000 }, { "epoch": 0.67, "learning_rate": 6.558572937854757e-06, "loss": 0.6666, "step": 1864500 }, { "epoch": 0.67, "learning_rate": 6.554968371734579e-06, "loss": 0.6701, "step": 1865000 }, { "epoch": 0.67, "learning_rate": 6.5513638056144e-06, "loss": 0.6689, "step": 1865500 }, { "epoch": 0.67, "learning_rate": 6.547759239494223e-06, "loss": 0.6558, "step": 1866000 }, { "epoch": 0.67, "learning_rate": 6.544154673374043e-06, "loss": 0.6584, "step": 1866500 }, { "epoch": 0.67, "learning_rate": 6.540550107253866e-06, "loss": 0.6565, "step": 1867000 }, { "epoch": 0.67, "learning_rate": 6.536945541133687e-06, "loss": 0.6661, "step": 1867500 }, { "epoch": 0.67, "learning_rate": 6.533340975013509e-06, "loss": 0.6626, "step": 1868000 }, { "epoch": 0.67, "learning_rate": 6.52973640889333e-06, "loss": 0.6842, "step": 1868500 }, { "epoch": 0.67, "learning_rate": 6.526131842773152e-06, "loss": 0.6525, "step": 1869000 }, { "epoch": 0.67, "learning_rate": 6.5225272766529735e-06, "loss": 0.6644, "step": 1869500 }, { "epoch": 0.67, "learning_rate": 6.518922710532795e-06, "loss": 0.6677, "step": 1870000 }, { "epoch": 0.67, "learning_rate": 6.515318144412617e-06, "loss": 0.6989, "step": 1870500 }, { "epoch": 0.67, "learning_rate": 6.511713578292438e-06, "loss": 0.6783, "step": 1871000 }, { "epoch": 0.67, "learning_rate": 6.50810901217226e-06, "loss": 0.6286, "step": 1871500 }, { "epoch": 0.67, "learning_rate": 6.504504446052081e-06, "loss": 0.6643, "step": 1872000 }, { "epoch": 0.67, "learning_rate": 6.500899879931903e-06, "loss": 0.6654, "step": 1872500 }, { "epoch": 0.68, "learning_rate": 6.497295313811724e-06, "loss": 0.6589, "step": 1873000 }, { "epoch": 0.68, "learning_rate": 6.493690747691547e-06, "loss": 0.6687, "step": 1873500 }, { "epoch": 0.68, "learning_rate": 6.4900861815713674e-06, "loss": 0.7137, "step": 1874000 }, { "epoch": 0.68, "learning_rate": 6.486481615451189e-06, "loss": 0.6381, "step": 1874500 }, { "epoch": 0.68, "learning_rate": 6.4828770493310115e-06, "loss": 0.683, "step": 1875000 }, { "epoch": 0.68, "learning_rate": 6.479272483210833e-06, "loss": 0.6468, "step": 1875500 }, { "epoch": 0.68, "learning_rate": 6.475667917090655e-06, "loss": 0.6896, "step": 1876000 }, { "epoch": 0.68, "learning_rate": 6.472063350970476e-06, "loss": 0.6763, "step": 1876500 }, { "epoch": 0.68, "learning_rate": 6.468458784850298e-06, "loss": 0.6572, "step": 1877000 }, { "epoch": 0.68, "learning_rate": 6.464854218730119e-06, "loss": 0.6557, "step": 1877500 }, { "epoch": 0.68, "learning_rate": 6.461249652609941e-06, "loss": 0.661, "step": 1878000 }, { "epoch": 0.68, "learning_rate": 6.457645086489762e-06, "loss": 0.6783, "step": 1878500 }, { "epoch": 0.68, "learning_rate": 6.454040520369584e-06, "loss": 0.6435, "step": 1879000 }, { "epoch": 0.68, "learning_rate": 6.450435954249405e-06, "loss": 0.6753, "step": 1879500 }, { "epoch": 0.68, "learning_rate": 6.4468313881292265e-06, "loss": 0.6777, "step": 1880000 }, { "epoch": 0.68, "learning_rate": 6.4432268220090485e-06, "loss": 0.6757, "step": 1880500 }, { "epoch": 0.68, "learning_rate": 6.43962225588887e-06, "loss": 0.6732, "step": 1881000 }, { "epoch": 0.68, "learning_rate": 6.436017689768692e-06, "loss": 0.6689, "step": 1881500 }, { "epoch": 0.68, "learning_rate": 6.432413123648513e-06, "loss": 0.654, "step": 1882000 }, { "epoch": 0.68, "learning_rate": 6.428808557528336e-06, "loss": 0.658, "step": 1882500 }, { "epoch": 0.68, "learning_rate": 6.425203991408157e-06, "loss": 0.6737, "step": 1883000 }, { "epoch": 0.68, "learning_rate": 6.421599425287979e-06, "loss": 0.6526, "step": 1883500 }, { "epoch": 0.68, "learning_rate": 6.4179948591678e-06, "loss": 0.6354, "step": 1884000 }, { "epoch": 0.68, "learning_rate": 6.414390293047622e-06, "loss": 0.6834, "step": 1884500 }, { "epoch": 0.68, "learning_rate": 6.410785726927443e-06, "loss": 0.6306, "step": 1885000 }, { "epoch": 0.68, "learning_rate": 6.407181160807264e-06, "loss": 0.6823, "step": 1885500 }, { "epoch": 0.68, "learning_rate": 6.4035765946870864e-06, "loss": 0.668, "step": 1886000 }, { "epoch": 0.68, "learning_rate": 6.399972028566908e-06, "loss": 0.6732, "step": 1886500 }, { "epoch": 0.68, "learning_rate": 6.39636746244673e-06, "loss": 0.6888, "step": 1887000 }, { "epoch": 0.68, "learning_rate": 6.392762896326551e-06, "loss": 0.6403, "step": 1887500 }, { "epoch": 0.68, "learning_rate": 6.389158330206373e-06, "loss": 0.6686, "step": 1888000 }, { "epoch": 0.68, "learning_rate": 6.385553764086194e-06, "loss": 0.6819, "step": 1888500 }, { "epoch": 0.68, "learning_rate": 6.381949197966017e-06, "loss": 0.6642, "step": 1889000 }, { "epoch": 0.68, "learning_rate": 6.378344631845837e-06, "loss": 0.663, "step": 1889500 }, { "epoch": 0.68, "learning_rate": 6.374740065725658e-06, "loss": 0.6713, "step": 1890000 }, { "epoch": 0.68, "learning_rate": 6.371135499605481e-06, "loss": 0.6433, "step": 1890500 }, { "epoch": 0.68, "learning_rate": 6.367530933485302e-06, "loss": 0.6419, "step": 1891000 }, { "epoch": 0.68, "learning_rate": 6.363926367365124e-06, "loss": 0.6348, "step": 1891500 }, { "epoch": 0.68, "learning_rate": 6.3603218012449455e-06, "loss": 0.6619, "step": 1892000 }, { "epoch": 0.68, "learning_rate": 6.3567172351247675e-06, "loss": 0.6608, "step": 1892500 }, { "epoch": 0.68, "learning_rate": 6.353112669004589e-06, "loss": 0.6523, "step": 1893000 }, { "epoch": 0.68, "learning_rate": 6.349508102884411e-06, "loss": 0.626, "step": 1893500 }, { "epoch": 0.68, "learning_rate": 6.345903536764232e-06, "loss": 0.6561, "step": 1894000 }, { "epoch": 0.68, "learning_rate": 6.342298970644054e-06, "loss": 0.657, "step": 1894500 }, { "epoch": 0.68, "learning_rate": 6.338694404523875e-06, "loss": 0.6455, "step": 1895000 }, { "epoch": 0.68, "learning_rate": 6.335089838403696e-06, "loss": 0.6675, "step": 1895500 }, { "epoch": 0.68, "learning_rate": 6.331485272283518e-06, "loss": 0.6608, "step": 1896000 }, { "epoch": 0.68, "learning_rate": 6.327880706163339e-06, "loss": 0.6418, "step": 1896500 }, { "epoch": 0.68, "learning_rate": 6.324276140043161e-06, "loss": 0.6498, "step": 1897000 }, { "epoch": 0.68, "learning_rate": 6.320671573922983e-06, "loss": 0.6481, "step": 1897500 }, { "epoch": 0.68, "learning_rate": 6.3170670078028054e-06, "loss": 0.6627, "step": 1898000 }, { "epoch": 0.68, "learning_rate": 6.313462441682627e-06, "loss": 0.6566, "step": 1898500 }, { "epoch": 0.68, "learning_rate": 6.309857875562449e-06, "loss": 0.6608, "step": 1899000 }, { "epoch": 0.68, "learning_rate": 6.30625330944227e-06, "loss": 0.657, "step": 1899500 }, { "epoch": 0.68, "learning_rate": 6.302648743322092e-06, "loss": 0.6562, "step": 1900000 }, { "epoch": 0.69, "learning_rate": 6.299044177201913e-06, "loss": 0.685, "step": 1900500 }, { "epoch": 0.69, "learning_rate": 6.295439611081734e-06, "loss": 0.6712, "step": 1901000 }, { "epoch": 0.69, "learning_rate": 6.291835044961556e-06, "loss": 0.6569, "step": 1901500 }, { "epoch": 0.69, "learning_rate": 6.288230478841377e-06, "loss": 0.6415, "step": 1902000 }, { "epoch": 0.69, "learning_rate": 6.284625912721199e-06, "loss": 0.6583, "step": 1902500 }, { "epoch": 0.69, "learning_rate": 6.2810213466010205e-06, "loss": 0.6326, "step": 1903000 }, { "epoch": 0.69, "learning_rate": 6.2774167804808425e-06, "loss": 0.6684, "step": 1903500 }, { "epoch": 0.69, "learning_rate": 6.273812214360664e-06, "loss": 0.66, "step": 1904000 }, { "epoch": 0.69, "learning_rate": 6.270207648240486e-06, "loss": 0.6444, "step": 1904500 }, { "epoch": 0.69, "learning_rate": 6.266603082120307e-06, "loss": 0.6322, "step": 1905000 }, { "epoch": 0.69, "learning_rate": 6.262998516000128e-06, "loss": 0.6619, "step": 1905500 }, { "epoch": 0.69, "learning_rate": 6.259393949879951e-06, "loss": 0.65, "step": 1906000 }, { "epoch": 0.69, "learning_rate": 6.255789383759771e-06, "loss": 0.6743, "step": 1906500 }, { "epoch": 0.69, "learning_rate": 6.252184817639594e-06, "loss": 0.6861, "step": 1907000 }, { "epoch": 0.69, "learning_rate": 6.248580251519415e-06, "loss": 0.6651, "step": 1907500 }, { "epoch": 0.69, "learning_rate": 6.244975685399237e-06, "loss": 0.6566, "step": 1908000 }, { "epoch": 0.69, "learning_rate": 6.241371119279058e-06, "loss": 0.6359, "step": 1908500 }, { "epoch": 0.69, "learning_rate": 6.2377665531588804e-06, "loss": 0.6748, "step": 1909000 }, { "epoch": 0.69, "learning_rate": 6.234161987038702e-06, "loss": 0.6586, "step": 1909500 }, { "epoch": 0.69, "learning_rate": 6.230557420918524e-06, "loss": 0.6505, "step": 1910000 }, { "epoch": 0.69, "learning_rate": 6.226952854798345e-06, "loss": 0.6919, "step": 1910500 }, { "epoch": 0.69, "learning_rate": 6.223348288678166e-06, "loss": 0.6579, "step": 1911000 }, { "epoch": 0.69, "learning_rate": 6.219743722557988e-06, "loss": 0.6679, "step": 1911500 }, { "epoch": 0.69, "learning_rate": 6.216139156437809e-06, "loss": 0.665, "step": 1912000 }, { "epoch": 0.69, "learning_rate": 6.212534590317631e-06, "loss": 0.6755, "step": 1912500 }, { "epoch": 0.69, "learning_rate": 6.208930024197452e-06, "loss": 0.6686, "step": 1913000 }, { "epoch": 0.69, "learning_rate": 6.205325458077275e-06, "loss": 0.6546, "step": 1913500 }, { "epoch": 0.69, "learning_rate": 6.201720891957096e-06, "loss": 0.696, "step": 1914000 }, { "epoch": 0.69, "learning_rate": 6.198116325836918e-06, "loss": 0.6689, "step": 1914500 }, { "epoch": 0.69, "learning_rate": 6.1945117597167395e-06, "loss": 0.6759, "step": 1915000 }, { "epoch": 0.69, "learning_rate": 6.190907193596561e-06, "loss": 0.6634, "step": 1915500 }, { "epoch": 0.69, "learning_rate": 6.187302627476383e-06, "loss": 0.6636, "step": 1916000 }, { "epoch": 0.69, "learning_rate": 6.183698061356204e-06, "loss": 0.6769, "step": 1916500 }, { "epoch": 0.69, "learning_rate": 6.180093495236026e-06, "loss": 0.6723, "step": 1917000 }, { "epoch": 0.69, "learning_rate": 6.176488929115847e-06, "loss": 0.652, "step": 1917500 }, { "epoch": 0.69, "learning_rate": 6.172884362995669e-06, "loss": 0.6874, "step": 1918000 }, { "epoch": 0.69, "learning_rate": 6.16927979687549e-06, "loss": 0.6811, "step": 1918500 }, { "epoch": 0.69, "learning_rate": 6.165675230755312e-06, "loss": 0.6624, "step": 1919000 }, { "epoch": 0.69, "learning_rate": 6.162070664635133e-06, "loss": 0.6739, "step": 1919500 }, { "epoch": 0.69, "learning_rate": 6.158466098514955e-06, "loss": 0.6599, "step": 1920000 }, { "epoch": 0.69, "learning_rate": 6.154861532394777e-06, "loss": 0.6676, "step": 1920500 }, { "epoch": 0.69, "learning_rate": 6.151256966274598e-06, "loss": 0.6709, "step": 1921000 }, { "epoch": 0.69, "learning_rate": 6.147652400154421e-06, "loss": 0.669, "step": 1921500 }, { "epoch": 0.69, "learning_rate": 6.144047834034241e-06, "loss": 0.6561, "step": 1922000 }, { "epoch": 0.69, "learning_rate": 6.140443267914064e-06, "loss": 0.6463, "step": 1922500 }, { "epoch": 0.69, "learning_rate": 6.136838701793885e-06, "loss": 0.6705, "step": 1923000 }, { "epoch": 0.69, "learning_rate": 6.133234135673707e-06, "loss": 0.6554, "step": 1923500 }, { "epoch": 0.69, "learning_rate": 6.129629569553528e-06, "loss": 0.6701, "step": 1924000 }, { "epoch": 0.69, "learning_rate": 6.12602500343335e-06, "loss": 0.6425, "step": 1924500 }, { "epoch": 0.69, "learning_rate": 6.122420437313171e-06, "loss": 0.6547, "step": 1925000 }, { "epoch": 0.69, "learning_rate": 6.118815871192993e-06, "loss": 0.6354, "step": 1925500 }, { "epoch": 0.69, "learning_rate": 6.1152113050728145e-06, "loss": 0.6432, "step": 1926000 }, { "epoch": 0.69, "learning_rate": 6.111606738952636e-06, "loss": 0.6355, "step": 1926500 }, { "epoch": 0.69, "learning_rate": 6.108002172832458e-06, "loss": 0.6744, "step": 1927000 }, { "epoch": 0.69, "learning_rate": 6.104397606712279e-06, "loss": 0.6509, "step": 1927500 }, { "epoch": 0.69, "learning_rate": 6.100793040592101e-06, "loss": 0.6859, "step": 1928000 }, { "epoch": 0.7, "learning_rate": 6.097188474471922e-06, "loss": 0.6603, "step": 1928500 }, { "epoch": 0.7, "learning_rate": 6.093583908351745e-06, "loss": 0.6595, "step": 1929000 }, { "epoch": 0.7, "learning_rate": 6.089979342231565e-06, "loss": 0.6631, "step": 1929500 }, { "epoch": 0.7, "learning_rate": 6.086374776111388e-06, "loss": 0.6507, "step": 1930000 }, { "epoch": 0.7, "learning_rate": 6.082770209991209e-06, "loss": 0.6687, "step": 1930500 }, { "epoch": 0.7, "learning_rate": 6.07916564387103e-06, "loss": 0.6713, "step": 1931000 }, { "epoch": 0.7, "learning_rate": 6.075561077750852e-06, "loss": 0.6278, "step": 1931500 }, { "epoch": 0.7, "learning_rate": 6.071956511630674e-06, "loss": 0.6763, "step": 1932000 }, { "epoch": 0.7, "learning_rate": 6.068351945510496e-06, "loss": 0.6462, "step": 1932500 }, { "epoch": 0.7, "learning_rate": 6.064747379390317e-06, "loss": 0.6584, "step": 1933000 }, { "epoch": 0.7, "learning_rate": 6.061142813270139e-06, "loss": 0.6988, "step": 1933500 }, { "epoch": 0.7, "learning_rate": 6.05753824714996e-06, "loss": 0.6566, "step": 1934000 }, { "epoch": 0.7, "learning_rate": 6.053933681029782e-06, "loss": 0.6581, "step": 1934500 }, { "epoch": 0.7, "learning_rate": 6.050329114909603e-06, "loss": 0.6265, "step": 1935000 }, { "epoch": 0.7, "learning_rate": 6.046724548789425e-06, "loss": 0.6547, "step": 1935500 }, { "epoch": 0.7, "learning_rate": 6.043119982669246e-06, "loss": 0.6513, "step": 1936000 }, { "epoch": 0.7, "learning_rate": 6.0395154165490675e-06, "loss": 0.689, "step": 1936500 }, { "epoch": 0.7, "learning_rate": 6.03591085042889e-06, "loss": 0.6403, "step": 1937000 }, { "epoch": 0.7, "learning_rate": 6.032306284308711e-06, "loss": 0.6688, "step": 1937500 }, { "epoch": 0.7, "learning_rate": 6.0287017181885335e-06, "loss": 0.6464, "step": 1938000 }, { "epoch": 0.7, "learning_rate": 6.025097152068355e-06, "loss": 0.6338, "step": 1938500 }, { "epoch": 0.7, "learning_rate": 6.021492585948177e-06, "loss": 0.6552, "step": 1939000 }, { "epoch": 0.7, "learning_rate": 6.017888019827998e-06, "loss": 0.6333, "step": 1939500 }, { "epoch": 0.7, "learning_rate": 6.01428345370782e-06, "loss": 0.6595, "step": 1940000 }, { "epoch": 0.7, "learning_rate": 6.010678887587641e-06, "loss": 0.6718, "step": 1940500 }, { "epoch": 0.7, "learning_rate": 6.007074321467463e-06, "loss": 0.6326, "step": 1941000 }, { "epoch": 0.7, "learning_rate": 6.003469755347284e-06, "loss": 0.6514, "step": 1941500 }, { "epoch": 0.7, "learning_rate": 5.999865189227105e-06, "loss": 0.7176, "step": 1942000 }, { "epoch": 0.7, "learning_rate": 5.996260623106927e-06, "loss": 0.6403, "step": 1942500 }, { "epoch": 0.7, "learning_rate": 5.9926560569867486e-06, "loss": 0.6403, "step": 1943000 }, { "epoch": 0.7, "learning_rate": 5.989051490866571e-06, "loss": 0.6632, "step": 1943500 }, { "epoch": 0.7, "learning_rate": 5.985446924746392e-06, "loss": 0.6522, "step": 1944000 }, { "epoch": 0.7, "learning_rate": 5.981842358626215e-06, "loss": 0.6268, "step": 1944500 }, { "epoch": 0.7, "learning_rate": 5.978237792506035e-06, "loss": 0.6564, "step": 1945000 }, { "epoch": 0.7, "learning_rate": 5.974633226385858e-06, "loss": 0.6797, "step": 1945500 }, { "epoch": 0.7, "learning_rate": 5.971028660265679e-06, "loss": 0.6558, "step": 1946000 }, { "epoch": 0.7, "learning_rate": 5.9674240941455e-06, "loss": 0.6453, "step": 1946500 }, { "epoch": 0.7, "learning_rate": 5.963819528025322e-06, "loss": 0.6542, "step": 1947000 }, { "epoch": 0.7, "learning_rate": 5.960214961905143e-06, "loss": 0.6882, "step": 1947500 }, { "epoch": 0.7, "learning_rate": 5.956610395784965e-06, "loss": 0.6539, "step": 1948000 }, { "epoch": 0.7, "learning_rate": 5.9530058296647865e-06, "loss": 0.6635, "step": 1948500 }, { "epoch": 0.7, "learning_rate": 5.9494012635446085e-06, "loss": 0.6445, "step": 1949000 }, { "epoch": 0.7, "learning_rate": 5.94579669742443e-06, "loss": 0.6501, "step": 1949500 }, { "epoch": 0.7, "learning_rate": 5.942192131304252e-06, "loss": 0.6635, "step": 1950000 }, { "epoch": 0.7, "learning_rate": 5.938587565184073e-06, "loss": 0.676, "step": 1950500 }, { "epoch": 0.7, "learning_rate": 5.934982999063895e-06, "loss": 0.6581, "step": 1951000 }, { "epoch": 0.7, "learning_rate": 5.931378432943716e-06, "loss": 0.654, "step": 1951500 }, { "epoch": 0.7, "learning_rate": 5.927773866823537e-06, "loss": 0.652, "step": 1952000 }, { "epoch": 0.7, "learning_rate": 5.924169300703359e-06, "loss": 0.6734, "step": 1952500 }, { "epoch": 0.7, "learning_rate": 5.92056473458318e-06, "loss": 0.661, "step": 1953000 }, { "epoch": 0.7, "learning_rate": 5.916960168463003e-06, "loss": 0.6546, "step": 1953500 }, { "epoch": 0.7, "learning_rate": 5.913355602342824e-06, "loss": 0.683, "step": 1954000 }, { "epoch": 0.7, "learning_rate": 5.909751036222646e-06, "loss": 0.6642, "step": 1954500 }, { "epoch": 0.7, "learning_rate": 5.906146470102468e-06, "loss": 0.6639, "step": 1955000 }, { "epoch": 0.7, "learning_rate": 5.90254190398229e-06, "loss": 0.6639, "step": 1955500 }, { "epoch": 0.71, "learning_rate": 5.898937337862111e-06, "loss": 0.6331, "step": 1956000 }, { "epoch": 0.71, "learning_rate": 5.895332771741932e-06, "loss": 0.6916, "step": 1956500 }, { "epoch": 0.71, "learning_rate": 5.891728205621754e-06, "loss": 0.6544, "step": 1957000 }, { "epoch": 0.71, "learning_rate": 5.888123639501575e-06, "loss": 0.6646, "step": 1957500 }, { "epoch": 0.71, "learning_rate": 5.884519073381397e-06, "loss": 0.6747, "step": 1958000 }, { "epoch": 0.71, "learning_rate": 5.880914507261218e-06, "loss": 0.6506, "step": 1958500 }, { "epoch": 0.71, "learning_rate": 5.87730994114104e-06, "loss": 0.6413, "step": 1959000 }, { "epoch": 0.71, "learning_rate": 5.8737053750208615e-06, "loss": 0.6489, "step": 1959500 }, { "epoch": 0.71, "learning_rate": 5.8701008089006835e-06, "loss": 0.6774, "step": 1960000 }, { "epoch": 0.71, "learning_rate": 5.866496242780505e-06, "loss": 0.6666, "step": 1960500 }, { "epoch": 0.71, "learning_rate": 5.8628916766603275e-06, "loss": 0.6466, "step": 1961000 }, { "epoch": 0.71, "learning_rate": 5.859287110540149e-06, "loss": 0.6579, "step": 1961500 }, { "epoch": 0.71, "learning_rate": 5.85568254441997e-06, "loss": 0.6441, "step": 1962000 }, { "epoch": 0.71, "learning_rate": 5.852077978299792e-06, "loss": 0.6472, "step": 1962500 }, { "epoch": 0.71, "learning_rate": 5.848473412179613e-06, "loss": 0.6447, "step": 1963000 }, { "epoch": 0.71, "learning_rate": 5.844868846059435e-06, "loss": 0.6606, "step": 1963500 }, { "epoch": 0.71, "learning_rate": 5.841264279939256e-06, "loss": 0.6751, "step": 1964000 }, { "epoch": 0.71, "learning_rate": 5.837659713819078e-06, "loss": 0.6385, "step": 1964500 }, { "epoch": 0.71, "learning_rate": 5.834055147698899e-06, "loss": 0.6778, "step": 1965000 }, { "epoch": 0.71, "learning_rate": 5.830450581578721e-06, "loss": 0.6619, "step": 1965500 }, { "epoch": 0.71, "learning_rate": 5.8268460154585426e-06, "loss": 0.6862, "step": 1966000 }, { "epoch": 0.71, "learning_rate": 5.8232414493383646e-06, "loss": 0.6587, "step": 1966500 }, { "epoch": 0.71, "learning_rate": 5.819636883218186e-06, "loss": 0.6688, "step": 1967000 }, { "epoch": 0.71, "learning_rate": 5.816032317098007e-06, "loss": 0.6727, "step": 1967500 }, { "epoch": 0.71, "learning_rate": 5.812427750977829e-06, "loss": 0.6748, "step": 1968000 }, { "epoch": 0.71, "learning_rate": 5.80882318485765e-06, "loss": 0.6222, "step": 1968500 }, { "epoch": 0.71, "learning_rate": 5.805218618737473e-06, "loss": 0.6624, "step": 1969000 }, { "epoch": 0.71, "learning_rate": 5.801614052617294e-06, "loss": 0.6263, "step": 1969500 }, { "epoch": 0.71, "learning_rate": 5.798009486497116e-06, "loss": 0.6603, "step": 1970000 }, { "epoch": 0.71, "learning_rate": 5.794404920376937e-06, "loss": 0.6509, "step": 1970500 }, { "epoch": 0.71, "learning_rate": 5.790800354256759e-06, "loss": 0.6702, "step": 1971000 }, { "epoch": 0.71, "learning_rate": 5.7871957881365805e-06, "loss": 0.6648, "step": 1971500 }, { "epoch": 0.71, "learning_rate": 5.783591222016402e-06, "loss": 0.6759, "step": 1972000 }, { "epoch": 0.71, "learning_rate": 5.779986655896224e-06, "loss": 0.6751, "step": 1972500 }, { "epoch": 0.71, "learning_rate": 5.776382089776045e-06, "loss": 0.6719, "step": 1973000 }, { "epoch": 0.71, "learning_rate": 5.772777523655867e-06, "loss": 0.6561, "step": 1973500 }, { "epoch": 0.71, "learning_rate": 5.769172957535688e-06, "loss": 0.6563, "step": 1974000 }, { "epoch": 0.71, "learning_rate": 5.76556839141551e-06, "loss": 0.6585, "step": 1974500 }, { "epoch": 0.71, "learning_rate": 5.761963825295331e-06, "loss": 0.67, "step": 1975000 }, { "epoch": 0.71, "learning_rate": 5.758359259175153e-06, "loss": 0.6616, "step": 1975500 }, { "epoch": 0.71, "learning_rate": 5.754754693054974e-06, "loss": 0.6876, "step": 1976000 }, { "epoch": 0.71, "learning_rate": 5.751150126934797e-06, "loss": 0.663, "step": 1976500 }, { "epoch": 0.71, "learning_rate": 5.747545560814618e-06, "loss": 0.6723, "step": 1977000 }, { "epoch": 0.71, "learning_rate": 5.743940994694439e-06, "loss": 0.6545, "step": 1977500 }, { "epoch": 0.71, "learning_rate": 5.7403364285742616e-06, "loss": 0.6238, "step": 1978000 }, { "epoch": 0.71, "learning_rate": 5.736731862454083e-06, "loss": 0.6788, "step": 1978500 }, { "epoch": 0.71, "learning_rate": 5.733127296333905e-06, "loss": 0.6546, "step": 1979000 }, { "epoch": 0.71, "learning_rate": 5.729522730213726e-06, "loss": 0.6793, "step": 1979500 }, { "epoch": 0.71, "learning_rate": 5.725918164093548e-06, "loss": 0.6681, "step": 1980000 }, { "epoch": 0.71, "learning_rate": 5.722313597973369e-06, "loss": 0.6459, "step": 1980500 }, { "epoch": 0.71, "learning_rate": 5.718709031853191e-06, "loss": 0.6361, "step": 1981000 }, { "epoch": 0.71, "learning_rate": 5.715104465733012e-06, "loss": 0.6643, "step": 1981500 }, { "epoch": 0.71, "learning_rate": 5.711499899612834e-06, "loss": 0.681, "step": 1982000 }, { "epoch": 0.71, "learning_rate": 5.7078953334926555e-06, "loss": 0.6595, "step": 1982500 }, { "epoch": 0.71, "learning_rate": 5.704290767372477e-06, "loss": 0.6468, "step": 1983000 }, { "epoch": 0.71, "learning_rate": 5.700686201252299e-06, "loss": 0.6564, "step": 1983500 }, { "epoch": 0.72, "learning_rate": 5.69708163513212e-06, "loss": 0.6502, "step": 1984000 }, { "epoch": 0.72, "learning_rate": 5.693477069011943e-06, "loss": 0.6496, "step": 1984500 }, { "epoch": 0.72, "learning_rate": 5.689872502891763e-06, "loss": 0.6535, "step": 1985000 }, { "epoch": 0.72, "learning_rate": 5.686267936771586e-06, "loss": 0.6666, "step": 1985500 }, { "epoch": 0.72, "learning_rate": 5.682663370651407e-06, "loss": 0.6668, "step": 1986000 }, { "epoch": 0.72, "learning_rate": 5.679058804531229e-06, "loss": 0.6856, "step": 1986500 }, { "epoch": 0.72, "learning_rate": 5.67545423841105e-06, "loss": 0.6531, "step": 1987000 }, { "epoch": 0.72, "learning_rate": 5.671849672290871e-06, "loss": 0.681, "step": 1987500 }, { "epoch": 0.72, "learning_rate": 5.668245106170693e-06, "loss": 0.6354, "step": 1988000 }, { "epoch": 0.72, "learning_rate": 5.6646405400505145e-06, "loss": 0.6747, "step": 1988500 }, { "epoch": 0.72, "learning_rate": 5.6610359739303366e-06, "loss": 0.6313, "step": 1989000 }, { "epoch": 0.72, "learning_rate": 5.657431407810158e-06, "loss": 0.6745, "step": 1989500 }, { "epoch": 0.72, "learning_rate": 5.65382684168998e-06, "loss": 0.6509, "step": 1990000 }, { "epoch": 0.72, "learning_rate": 5.650222275569801e-06, "loss": 0.6569, "step": 1990500 }, { "epoch": 0.72, "learning_rate": 5.646617709449623e-06, "loss": 0.6584, "step": 1991000 }, { "epoch": 0.72, "learning_rate": 5.643013143329444e-06, "loss": 0.6499, "step": 1991500 }, { "epoch": 0.72, "learning_rate": 5.639408577209267e-06, "loss": 0.6534, "step": 1992000 }, { "epoch": 0.72, "learning_rate": 5.635804011089088e-06, "loss": 0.6645, "step": 1992500 }, { "epoch": 0.72, "learning_rate": 5.6321994449689084e-06, "loss": 0.668, "step": 1993000 }, { "epoch": 0.72, "learning_rate": 5.628594878848731e-06, "loss": 0.6507, "step": 1993500 }, { "epoch": 0.72, "learning_rate": 5.6249903127285525e-06, "loss": 0.6457, "step": 1994000 }, { "epoch": 0.72, "learning_rate": 5.6213857466083745e-06, "loss": 0.6692, "step": 1994500 }, { "epoch": 0.72, "learning_rate": 5.617781180488196e-06, "loss": 0.6413, "step": 1995000 }, { "epoch": 0.72, "learning_rate": 5.614176614368018e-06, "loss": 0.673, "step": 1995500 }, { "epoch": 0.72, "learning_rate": 5.610572048247839e-06, "loss": 0.6697, "step": 1996000 }, { "epoch": 0.72, "learning_rate": 5.606967482127661e-06, "loss": 0.6802, "step": 1996500 }, { "epoch": 0.72, "learning_rate": 5.603362916007482e-06, "loss": 0.6656, "step": 1997000 }, { "epoch": 0.72, "learning_rate": 5.599758349887303e-06, "loss": 0.6566, "step": 1997500 }, { "epoch": 0.72, "learning_rate": 5.596153783767125e-06, "loss": 0.6461, "step": 1998000 }, { "epoch": 0.72, "learning_rate": 5.592549217646946e-06, "loss": 0.6584, "step": 1998500 }, { "epoch": 0.72, "learning_rate": 5.588944651526768e-06, "loss": 0.6767, "step": 1999000 }, { "epoch": 0.72, "learning_rate": 5.5853400854065895e-06, "loss": 0.6313, "step": 1999500 }, { "epoch": 0.72, "learning_rate": 5.581735519286412e-06, "loss": 0.6269, "step": 2000000 }, { "epoch": 0.72, "learning_rate": 5.578130953166233e-06, "loss": 0.6495, "step": 2000500 }, { "epoch": 0.72, "learning_rate": 5.5745263870460556e-06, "loss": 0.6551, "step": 2001000 }, { "epoch": 0.72, "learning_rate": 5.570921820925877e-06, "loss": 0.635, "step": 2001500 }, { "epoch": 0.72, "learning_rate": 5.567317254805699e-06, "loss": 0.665, "step": 2002000 }, { "epoch": 0.72, "learning_rate": 5.56371268868552e-06, "loss": 0.6236, "step": 2002500 }, { "epoch": 0.72, "learning_rate": 5.560108122565341e-06, "loss": 0.6429, "step": 2003000 }, { "epoch": 0.72, "learning_rate": 5.556503556445163e-06, "loss": 0.67, "step": 2003500 }, { "epoch": 0.72, "learning_rate": 5.552898990324984e-06, "loss": 0.6315, "step": 2004000 }, { "epoch": 0.72, "learning_rate": 5.549294424204806e-06, "loss": 0.6573, "step": 2004500 }, { "epoch": 0.72, "learning_rate": 5.5456898580846274e-06, "loss": 0.6447, "step": 2005000 }, { "epoch": 0.72, "learning_rate": 5.5420852919644495e-06, "loss": 0.6579, "step": 2005500 }, { "epoch": 0.72, "learning_rate": 5.538480725844271e-06, "loss": 0.6643, "step": 2006000 }, { "epoch": 0.72, "learning_rate": 5.534876159724093e-06, "loss": 0.65, "step": 2006500 }, { "epoch": 0.72, "learning_rate": 5.531271593603914e-06, "loss": 0.6612, "step": 2007000 }, { "epoch": 0.72, "learning_rate": 5.527667027483737e-06, "loss": 0.6391, "step": 2007500 }, { "epoch": 0.72, "learning_rate": 5.524062461363557e-06, "loss": 0.6535, "step": 2008000 }, { "epoch": 0.72, "learning_rate": 5.520457895243378e-06, "loss": 0.6691, "step": 2008500 }, { "epoch": 0.72, "learning_rate": 5.516853329123201e-06, "loss": 0.6847, "step": 2009000 }, { "epoch": 0.72, "learning_rate": 5.513248763003022e-06, "loss": 0.6502, "step": 2009500 }, { "epoch": 0.72, "learning_rate": 5.509644196882844e-06, "loss": 0.6732, "step": 2010000 }, { "epoch": 0.72, "learning_rate": 5.506039630762665e-06, "loss": 0.66, "step": 2010500 }, { "epoch": 0.72, "learning_rate": 5.502435064642487e-06, "loss": 0.6568, "step": 2011000 }, { "epoch": 0.73, "learning_rate": 5.4988304985223085e-06, "loss": 0.65, "step": 2011500 }, { "epoch": 0.73, "learning_rate": 5.4952259324021306e-06, "loss": 0.6625, "step": 2012000 }, { "epoch": 0.73, "learning_rate": 5.491621366281952e-06, "loss": 0.6271, "step": 2012500 }, { "epoch": 0.73, "learning_rate": 5.488016800161773e-06, "loss": 0.6496, "step": 2013000 }, { "epoch": 0.73, "learning_rate": 5.484412234041595e-06, "loss": 0.6724, "step": 2013500 }, { "epoch": 0.73, "learning_rate": 5.480807667921416e-06, "loss": 0.653, "step": 2014000 }, { "epoch": 0.73, "learning_rate": 5.477203101801238e-06, "loss": 0.6626, "step": 2014500 }, { "epoch": 0.73, "learning_rate": 5.473598535681059e-06, "loss": 0.6555, "step": 2015000 }, { "epoch": 0.73, "learning_rate": 5.469993969560882e-06, "loss": 0.6268, "step": 2015500 }, { "epoch": 0.73, "learning_rate": 5.4663894034407024e-06, "loss": 0.6467, "step": 2016000 }, { "epoch": 0.73, "learning_rate": 5.462784837320525e-06, "loss": 0.6618, "step": 2016500 }, { "epoch": 0.73, "learning_rate": 5.4591802712003465e-06, "loss": 0.6547, "step": 2017000 }, { "epoch": 0.73, "learning_rate": 5.4555757050801685e-06, "loss": 0.6387, "step": 2017500 }, { "epoch": 0.73, "learning_rate": 5.45197113895999e-06, "loss": 0.6511, "step": 2018000 }, { "epoch": 0.73, "learning_rate": 5.448366572839811e-06, "loss": 0.6368, "step": 2018500 }, { "epoch": 0.73, "learning_rate": 5.444762006719633e-06, "loss": 0.6672, "step": 2019000 }, { "epoch": 0.73, "learning_rate": 5.441157440599454e-06, "loss": 0.6552, "step": 2019500 }, { "epoch": 0.73, "learning_rate": 5.437552874479276e-06, "loss": 0.6374, "step": 2020000 }, { "epoch": 0.73, "learning_rate": 5.433948308359097e-06, "loss": 0.6704, "step": 2020500 }, { "epoch": 0.73, "learning_rate": 5.430343742238919e-06, "loss": 0.6688, "step": 2021000 }, { "epoch": 0.73, "learning_rate": 5.42673917611874e-06, "loss": 0.669, "step": 2021500 }, { "epoch": 0.73, "learning_rate": 5.423134609998562e-06, "loss": 0.693, "step": 2022000 }, { "epoch": 0.73, "learning_rate": 5.4195300438783835e-06, "loss": 0.6576, "step": 2022500 }, { "epoch": 0.73, "learning_rate": 5.415925477758205e-06, "loss": 0.6677, "step": 2023000 }, { "epoch": 0.73, "learning_rate": 5.412320911638027e-06, "loss": 0.6635, "step": 2023500 }, { "epoch": 0.73, "learning_rate": 5.408716345517848e-06, "loss": 0.6518, "step": 2024000 }, { "epoch": 0.73, "learning_rate": 5.405111779397671e-06, "loss": 0.6608, "step": 2024500 }, { "epoch": 0.73, "learning_rate": 5.401507213277492e-06, "loss": 0.6602, "step": 2025000 }, { "epoch": 0.73, "learning_rate": 5.397902647157314e-06, "loss": 0.6341, "step": 2025500 }, { "epoch": 0.73, "learning_rate": 5.394298081037135e-06, "loss": 0.6536, "step": 2026000 }, { "epoch": 0.73, "learning_rate": 5.390693514916957e-06, "loss": 0.6654, "step": 2026500 }, { "epoch": 0.73, "learning_rate": 5.387088948796778e-06, "loss": 0.6439, "step": 2027000 }, { "epoch": 0.73, "learning_rate": 5.3834843826766e-06, "loss": 0.6758, "step": 2027500 }, { "epoch": 0.73, "learning_rate": 5.3798798165564214e-06, "loss": 0.6122, "step": 2028000 }, { "epoch": 0.73, "learning_rate": 5.376275250436243e-06, "loss": 0.6665, "step": 2028500 }, { "epoch": 0.73, "learning_rate": 5.372670684316065e-06, "loss": 0.6631, "step": 2029000 }, { "epoch": 0.73, "learning_rate": 5.369066118195886e-06, "loss": 0.648, "step": 2029500 }, { "epoch": 0.73, "learning_rate": 5.365461552075708e-06, "loss": 0.6625, "step": 2030000 }, { "epoch": 0.73, "learning_rate": 5.361856985955529e-06, "loss": 0.661, "step": 2030500 }, { "epoch": 0.73, "learning_rate": 5.358252419835351e-06, "loss": 0.6462, "step": 2031000 }, { "epoch": 0.73, "learning_rate": 5.354647853715172e-06, "loss": 0.6219, "step": 2031500 }, { "epoch": 0.73, "learning_rate": 5.351043287594995e-06, "loss": 0.6476, "step": 2032000 }, { "epoch": 0.73, "learning_rate": 5.347438721474816e-06, "loss": 0.6591, "step": 2032500 }, { "epoch": 0.73, "learning_rate": 5.343834155354638e-06, "loss": 0.6251, "step": 2033000 }, { "epoch": 0.73, "learning_rate": 5.340229589234459e-06, "loss": 0.6581, "step": 2033500 }, { "epoch": 0.73, "learning_rate": 5.3366250231142805e-06, "loss": 0.6491, "step": 2034000 }, { "epoch": 0.73, "learning_rate": 5.3330204569941025e-06, "loss": 0.6491, "step": 2034500 }, { "epoch": 0.73, "learning_rate": 5.329415890873924e-06, "loss": 0.6618, "step": 2035000 }, { "epoch": 0.73, "learning_rate": 5.325811324753746e-06, "loss": 0.6377, "step": 2035500 }, { "epoch": 0.73, "learning_rate": 5.322206758633567e-06, "loss": 0.6348, "step": 2036000 }, { "epoch": 0.73, "learning_rate": 5.318602192513389e-06, "loss": 0.6533, "step": 2036500 }, { "epoch": 0.73, "learning_rate": 5.31499762639321e-06, "loss": 0.6417, "step": 2037000 }, { "epoch": 0.73, "learning_rate": 5.311393060273032e-06, "loss": 0.6282, "step": 2037500 }, { "epoch": 0.73, "learning_rate": 5.307788494152853e-06, "loss": 0.6605, "step": 2038000 }, { "epoch": 0.73, "learning_rate": 5.304183928032674e-06, "loss": 0.6709, "step": 2038500 }, { "epoch": 0.73, "learning_rate": 5.3005793619124964e-06, "loss": 0.623, "step": 2039000 }, { "epoch": 0.74, "learning_rate": 5.296974795792318e-06, "loss": 0.6474, "step": 2039500 }, { "epoch": 0.74, "learning_rate": 5.2933702296721405e-06, "loss": 0.6487, "step": 2040000 }, { "epoch": 0.74, "learning_rate": 5.289765663551961e-06, "loss": 0.6549, "step": 2040500 }, { "epoch": 0.74, "learning_rate": 5.286161097431784e-06, "loss": 0.6624, "step": 2041000 }, { "epoch": 0.74, "learning_rate": 5.282556531311605e-06, "loss": 0.6519, "step": 2041500 }, { "epoch": 0.74, "learning_rate": 5.278951965191427e-06, "loss": 0.678, "step": 2042000 }, { "epoch": 0.74, "learning_rate": 5.275347399071248e-06, "loss": 0.6355, "step": 2042500 }, { "epoch": 0.74, "learning_rate": 5.27174283295107e-06, "loss": 0.6256, "step": 2043000 }, { "epoch": 0.74, "learning_rate": 5.268138266830891e-06, "loss": 0.6748, "step": 2043500 }, { "epoch": 0.74, "learning_rate": 5.264533700710712e-06, "loss": 0.6541, "step": 2044000 }, { "epoch": 0.74, "learning_rate": 5.260929134590534e-06, "loss": 0.6413, "step": 2044500 }, { "epoch": 0.74, "learning_rate": 5.2573245684703555e-06, "loss": 0.6812, "step": 2045000 }, { "epoch": 0.74, "learning_rate": 5.2537200023501775e-06, "loss": 0.6567, "step": 2045500 }, { "epoch": 0.74, "learning_rate": 5.250115436229999e-06, "loss": 0.631, "step": 2046000 }, { "epoch": 0.74, "learning_rate": 5.246510870109821e-06, "loss": 0.6334, "step": 2046500 }, { "epoch": 0.74, "learning_rate": 5.242906303989642e-06, "loss": 0.6468, "step": 2047000 }, { "epoch": 0.74, "learning_rate": 5.239301737869465e-06, "loss": 0.6504, "step": 2047500 }, { "epoch": 0.74, "learning_rate": 5.235697171749286e-06, "loss": 0.6623, "step": 2048000 }, { "epoch": 0.74, "learning_rate": 5.232092605629108e-06, "loss": 0.6529, "step": 2048500 }, { "epoch": 0.74, "learning_rate": 5.228488039508929e-06, "loss": 0.6698, "step": 2049000 }, { "epoch": 0.74, "learning_rate": 5.22488347338875e-06, "loss": 0.6512, "step": 2049500 }, { "epoch": 0.74, "learning_rate": 5.221278907268572e-06, "loss": 0.6537, "step": 2050000 }, { "epoch": 0.74, "learning_rate": 5.217674341148393e-06, "loss": 0.6376, "step": 2050500 }, { "epoch": 0.74, "learning_rate": 5.2140697750282154e-06, "loss": 0.6076, "step": 2051000 }, { "epoch": 0.74, "learning_rate": 5.210465208908037e-06, "loss": 0.6555, "step": 2051500 }, { "epoch": 0.74, "learning_rate": 5.206860642787859e-06, "loss": 0.6537, "step": 2052000 }, { "epoch": 0.74, "learning_rate": 5.20325607666768e-06, "loss": 0.6408, "step": 2052500 }, { "epoch": 0.74, "learning_rate": 5.199651510547502e-06, "loss": 0.6588, "step": 2053000 }, { "epoch": 0.74, "learning_rate": 5.196046944427323e-06, "loss": 0.655, "step": 2053500 }, { "epoch": 0.74, "learning_rate": 5.192442378307144e-06, "loss": 0.6542, "step": 2054000 }, { "epoch": 0.74, "learning_rate": 5.188837812186966e-06, "loss": 0.6546, "step": 2054500 }, { "epoch": 0.74, "learning_rate": 5.185233246066787e-06, "loss": 0.6389, "step": 2055000 }, { "epoch": 0.74, "learning_rate": 5.18162867994661e-06, "loss": 0.6522, "step": 2055500 }, { "epoch": 0.74, "learning_rate": 5.1780241138264305e-06, "loss": 0.6284, "step": 2056000 }, { "epoch": 0.74, "learning_rate": 5.174419547706253e-06, "loss": 0.6726, "step": 2056500 }, { "epoch": 0.74, "learning_rate": 5.1708149815860745e-06, "loss": 0.6325, "step": 2057000 }, { "epoch": 0.74, "learning_rate": 5.1672104154658965e-06, "loss": 0.6592, "step": 2057500 }, { "epoch": 0.74, "learning_rate": 5.163605849345718e-06, "loss": 0.649, "step": 2058000 }, { "epoch": 0.74, "learning_rate": 5.16000128322554e-06, "loss": 0.6566, "step": 2058500 }, { "epoch": 0.74, "learning_rate": 5.156396717105361e-06, "loss": 0.648, "step": 2059000 }, { "epoch": 0.74, "learning_rate": 5.152792150985182e-06, "loss": 0.6584, "step": 2059500 }, { "epoch": 0.74, "learning_rate": 5.149187584865004e-06, "loss": 0.6569, "step": 2060000 }, { "epoch": 0.74, "learning_rate": 5.145583018744825e-06, "loss": 0.6751, "step": 2060500 }, { "epoch": 0.74, "learning_rate": 5.141978452624647e-06, "loss": 0.6404, "step": 2061000 }, { "epoch": 0.74, "learning_rate": 5.138373886504468e-06, "loss": 0.631, "step": 2061500 }, { "epoch": 0.74, "learning_rate": 5.13476932038429e-06, "loss": 0.644, "step": 2062000 }, { "epoch": 0.74, "learning_rate": 5.131164754264112e-06, "loss": 0.6461, "step": 2062500 }, { "epoch": 0.74, "learning_rate": 5.1275601881439344e-06, "loss": 0.6564, "step": 2063000 }, { "epoch": 0.74, "learning_rate": 5.123955622023755e-06, "loss": 0.6546, "step": 2063500 }, { "epoch": 0.74, "learning_rate": 5.120351055903576e-06, "loss": 0.6377, "step": 2064000 }, { "epoch": 0.74, "learning_rate": 5.116746489783399e-06, "loss": 0.6667, "step": 2064500 }, { "epoch": 0.74, "learning_rate": 5.11314192366322e-06, "loss": 0.6764, "step": 2065000 }, { "epoch": 0.74, "learning_rate": 5.109537357543042e-06, "loss": 0.6461, "step": 2065500 }, { "epoch": 0.74, "learning_rate": 5.105932791422863e-06, "loss": 0.6245, "step": 2066000 }, { "epoch": 0.74, "learning_rate": 5.102328225302685e-06, "loss": 0.6714, "step": 2066500 }, { "epoch": 0.75, "learning_rate": 5.098723659182506e-06, "loss": 0.6347, "step": 2067000 }, { "epoch": 0.75, "learning_rate": 5.095119093062328e-06, "loss": 0.6372, "step": 2067500 }, { "epoch": 0.75, "learning_rate": 5.0915145269421495e-06, "loss": 0.6591, "step": 2068000 }, { "epoch": 0.75, "learning_rate": 5.0879099608219715e-06, "loss": 0.6361, "step": 2068500 }, { "epoch": 0.75, "learning_rate": 5.084305394701793e-06, "loss": 0.646, "step": 2069000 }, { "epoch": 0.75, "learning_rate": 5.080700828581614e-06, "loss": 0.6512, "step": 2069500 }, { "epoch": 0.75, "learning_rate": 5.077096262461436e-06, "loss": 0.6439, "step": 2070000 }, { "epoch": 0.75, "learning_rate": 5.073491696341257e-06, "loss": 0.6374, "step": 2070500 }, { "epoch": 0.75, "learning_rate": 5.06988713022108e-06, "loss": 0.6593, "step": 2071000 }, { "epoch": 0.75, "learning_rate": 5.0662825641009e-06, "loss": 0.67, "step": 2071500 }, { "epoch": 0.75, "learning_rate": 5.062677997980723e-06, "loss": 0.643, "step": 2072000 }, { "epoch": 0.75, "learning_rate": 5.059073431860544e-06, "loss": 0.6489, "step": 2072500 }, { "epoch": 0.75, "learning_rate": 5.055468865740366e-06, "loss": 0.6598, "step": 2073000 }, { "epoch": 0.75, "learning_rate": 5.051864299620187e-06, "loss": 0.6631, "step": 2073500 }, { "epoch": 0.75, "learning_rate": 5.0482597335000094e-06, "loss": 0.6482, "step": 2074000 }, { "epoch": 0.75, "learning_rate": 5.044655167379831e-06, "loss": 0.6409, "step": 2074500 }, { "epoch": 0.75, "learning_rate": 5.041050601259652e-06, "loss": 0.6218, "step": 2075000 }, { "epoch": 0.75, "learning_rate": 5.037446035139474e-06, "loss": 0.6358, "step": 2075500 }, { "epoch": 0.75, "learning_rate": 5.033841469019295e-06, "loss": 0.6797, "step": 2076000 }, { "epoch": 0.75, "learning_rate": 5.030236902899117e-06, "loss": 0.6377, "step": 2076500 }, { "epoch": 0.75, "learning_rate": 5.026632336778938e-06, "loss": 0.649, "step": 2077000 }, { "epoch": 0.75, "learning_rate": 5.02302777065876e-06, "loss": 0.6931, "step": 2077500 }, { "epoch": 0.75, "learning_rate": 5.019423204538581e-06, "loss": 0.6697, "step": 2078000 }, { "epoch": 0.75, "learning_rate": 5.015818638418404e-06, "loss": 0.6332, "step": 2078500 }, { "epoch": 0.75, "learning_rate": 5.0122140722982245e-06, "loss": 0.6592, "step": 2079000 }, { "epoch": 0.75, "learning_rate": 5.008609506178046e-06, "loss": 0.6564, "step": 2079500 }, { "epoch": 0.75, "learning_rate": 5.0050049400578685e-06, "loss": 0.6618, "step": 2080000 }, { "epoch": 0.75, "learning_rate": 5.00140037393769e-06, "loss": 0.6813, "step": 2080500 }, { "epoch": 0.75, "learning_rate": 4.997795807817512e-06, "loss": 0.6396, "step": 2081000 }, { "epoch": 0.75, "learning_rate": 4.994191241697333e-06, "loss": 0.6451, "step": 2081500 }, { "epoch": 0.75, "learning_rate": 4.990586675577154e-06, "loss": 0.635, "step": 2082000 }, { "epoch": 0.75, "learning_rate": 4.986982109456976e-06, "loss": 0.6451, "step": 2082500 }, { "epoch": 0.75, "learning_rate": 4.983377543336797e-06, "loss": 0.6408, "step": 2083000 }, { "epoch": 0.75, "learning_rate": 4.979772977216619e-06, "loss": 0.6502, "step": 2083500 }, { "epoch": 0.75, "learning_rate": 4.976168411096441e-06, "loss": 0.6483, "step": 2084000 }, { "epoch": 0.75, "learning_rate": 4.972563844976262e-06, "loss": 0.6396, "step": 2084500 }, { "epoch": 0.75, "learning_rate": 4.968959278856084e-06, "loss": 0.6358, "step": 2085000 }, { "epoch": 0.75, "learning_rate": 4.965354712735906e-06, "loss": 0.6322, "step": 2085500 }, { "epoch": 0.75, "learning_rate": 4.961750146615728e-06, "loss": 0.6457, "step": 2086000 }, { "epoch": 0.75, "learning_rate": 4.958145580495549e-06, "loss": 0.6466, "step": 2086500 }, { "epoch": 0.75, "learning_rate": 4.95454101437537e-06, "loss": 0.6754, "step": 2087000 }, { "epoch": 0.75, "learning_rate": 4.950936448255192e-06, "loss": 0.6375, "step": 2087500 }, { "epoch": 0.75, "learning_rate": 4.947331882135014e-06, "loss": 0.6509, "step": 2088000 }, { "epoch": 0.75, "learning_rate": 4.943727316014835e-06, "loss": 0.6419, "step": 2088500 }, { "epoch": 0.75, "learning_rate": 4.940122749894657e-06, "loss": 0.63, "step": 2089000 }, { "epoch": 0.75, "learning_rate": 4.936518183774478e-06, "loss": 0.6509, "step": 2089500 }, { "epoch": 0.75, "learning_rate": 4.9329136176543e-06, "loss": 0.6249, "step": 2090000 }, { "epoch": 0.75, "learning_rate": 4.9293090515341215e-06, "loss": 0.6138, "step": 2090500 }, { "epoch": 0.75, "learning_rate": 4.9257044854139435e-06, "loss": 0.6596, "step": 2091000 }, { "epoch": 0.75, "learning_rate": 4.9220999192937655e-06, "loss": 0.6655, "step": 2091500 }, { "epoch": 0.75, "learning_rate": 4.918495353173587e-06, "loss": 0.6568, "step": 2092000 }, { "epoch": 0.75, "learning_rate": 4.914890787053408e-06, "loss": 0.6264, "step": 2092500 }, { "epoch": 0.75, "learning_rate": 4.91128622093323e-06, "loss": 0.6463, "step": 2093000 }, { "epoch": 0.75, "learning_rate": 4.907681654813051e-06, "loss": 0.6465, "step": 2093500 }, { "epoch": 0.75, "learning_rate": 4.904077088692873e-06, "loss": 0.6571, "step": 2094000 }, { "epoch": 0.75, "learning_rate": 4.900472522572694e-06, "loss": 0.6494, "step": 2094500 }, { "epoch": 0.76, "learning_rate": 4.896867956452516e-06, "loss": 0.6615, "step": 2095000 }, { "epoch": 0.76, "learning_rate": 4.893263390332338e-06, "loss": 0.6376, "step": 2095500 }, { "epoch": 0.76, "learning_rate": 4.889658824212159e-06, "loss": 0.6448, "step": 2096000 }, { "epoch": 0.76, "learning_rate": 4.886054258091981e-06, "loss": 0.6667, "step": 2096500 }, { "epoch": 0.76, "learning_rate": 4.882449691971803e-06, "loss": 0.663, "step": 2097000 }, { "epoch": 0.76, "learning_rate": 4.878845125851624e-06, "loss": 0.6234, "step": 2097500 }, { "epoch": 0.76, "learning_rate": 4.875240559731446e-06, "loss": 0.6308, "step": 2098000 }, { "epoch": 0.76, "learning_rate": 4.871635993611267e-06, "loss": 0.6212, "step": 2098500 }, { "epoch": 0.76, "learning_rate": 4.868031427491089e-06, "loss": 0.6524, "step": 2099000 }, { "epoch": 0.76, "learning_rate": 4.864426861370911e-06, "loss": 0.6389, "step": 2099500 }, { "epoch": 0.76, "learning_rate": 4.860822295250732e-06, "loss": 0.6586, "step": 2100000 }, { "epoch": 0.76, "learning_rate": 4.857217729130554e-06, "loss": 0.6542, "step": 2100500 }, { "epoch": 0.76, "learning_rate": 4.853613163010375e-06, "loss": 0.6521, "step": 2101000 }, { "epoch": 0.76, "learning_rate": 4.850008596890197e-06, "loss": 0.6288, "step": 2101500 }, { "epoch": 0.76, "learning_rate": 4.8464040307700185e-06, "loss": 0.6397, "step": 2102000 }, { "epoch": 0.76, "learning_rate": 4.84279946464984e-06, "loss": 0.6593, "step": 2102500 }, { "epoch": 0.76, "learning_rate": 4.839194898529662e-06, "loss": 0.6497, "step": 2103000 }, { "epoch": 0.76, "learning_rate": 4.835590332409484e-06, "loss": 0.6563, "step": 2103500 }, { "epoch": 0.76, "learning_rate": 4.831985766289305e-06, "loss": 0.6485, "step": 2104000 }, { "epoch": 0.76, "learning_rate": 4.828381200169127e-06, "loss": 0.6424, "step": 2104500 }, { "epoch": 0.76, "learning_rate": 4.824776634048948e-06, "loss": 0.6265, "step": 2105000 }, { "epoch": 0.76, "learning_rate": 4.82117206792877e-06, "loss": 0.661, "step": 2105500 }, { "epoch": 0.76, "learning_rate": 4.817567501808591e-06, "loss": 0.6457, "step": 2106000 }, { "epoch": 0.76, "learning_rate": 4.813962935688413e-06, "loss": 0.6316, "step": 2106500 }, { "epoch": 0.76, "learning_rate": 4.810358369568235e-06, "loss": 0.6377, "step": 2107000 }, { "epoch": 0.76, "learning_rate": 4.806753803448056e-06, "loss": 0.6682, "step": 2107500 }, { "epoch": 0.76, "learning_rate": 4.8031492373278776e-06, "loss": 0.645, "step": 2108000 }, { "epoch": 0.76, "learning_rate": 4.7995446712077e-06, "loss": 0.6561, "step": 2108500 }, { "epoch": 0.76, "learning_rate": 4.795940105087521e-06, "loss": 0.6625, "step": 2109000 }, { "epoch": 0.76, "learning_rate": 4.792335538967343e-06, "loss": 0.6449, "step": 2109500 }, { "epoch": 0.76, "learning_rate": 4.788730972847164e-06, "loss": 0.6805, "step": 2110000 }, { "epoch": 0.76, "learning_rate": 4.785126406726986e-06, "loss": 0.6506, "step": 2110500 }, { "epoch": 0.76, "learning_rate": 4.781521840606808e-06, "loss": 0.613, "step": 2111000 }, { "epoch": 0.76, "learning_rate": 4.777917274486629e-06, "loss": 0.6586, "step": 2111500 }, { "epoch": 0.76, "learning_rate": 4.774312708366451e-06, "loss": 0.6098, "step": 2112000 }, { "epoch": 0.76, "learning_rate": 4.770708142246272e-06, "loss": 0.6369, "step": 2112500 }, { "epoch": 0.76, "learning_rate": 4.7671035761260935e-06, "loss": 0.6532, "step": 2113000 }, { "epoch": 0.76, "learning_rate": 4.7634990100059155e-06, "loss": 0.6755, "step": 2113500 }, { "epoch": 0.76, "learning_rate": 4.759894443885737e-06, "loss": 0.6345, "step": 2114000 }, { "epoch": 0.76, "learning_rate": 4.756289877765559e-06, "loss": 0.6477, "step": 2114500 }, { "epoch": 0.76, "learning_rate": 4.752685311645381e-06, "loss": 0.6252, "step": 2115000 }, { "epoch": 0.76, "learning_rate": 4.749080745525202e-06, "loss": 0.6644, "step": 2115500 }, { "epoch": 0.76, "learning_rate": 4.745476179405024e-06, "loss": 0.6652, "step": 2116000 }, { "epoch": 0.76, "learning_rate": 4.741871613284845e-06, "loss": 0.6427, "step": 2116500 }, { "epoch": 0.76, "learning_rate": 4.738267047164667e-06, "loss": 0.6438, "step": 2117000 }, { "epoch": 0.76, "learning_rate": 4.734662481044488e-06, "loss": 0.6445, "step": 2117500 }, { "epoch": 0.76, "learning_rate": 4.731057914924309e-06, "loss": 0.6617, "step": 2118000 }, { "epoch": 0.76, "learning_rate": 4.727453348804131e-06, "loss": 0.6531, "step": 2118500 }, { "epoch": 0.76, "learning_rate": 4.7238487826839525e-06, "loss": 0.6668, "step": 2119000 }, { "epoch": 0.76, "learning_rate": 4.7202442165637746e-06, "loss": 0.6422, "step": 2119500 }, { "epoch": 0.76, "learning_rate": 4.7166396504435966e-06, "loss": 0.6612, "step": 2120000 }, { "epoch": 0.76, "learning_rate": 4.713035084323418e-06, "loss": 0.6434, "step": 2120500 }, { "epoch": 0.76, "learning_rate": 4.70943051820324e-06, "loss": 0.6302, "step": 2121000 }, { "epoch": 0.76, "learning_rate": 4.705825952083061e-06, "loss": 0.6615, "step": 2121500 }, { "epoch": 0.76, "learning_rate": 4.702221385962883e-06, "loss": 0.6554, "step": 2122000 }, { "epoch": 0.77, "learning_rate": 4.698616819842705e-06, "loss": 0.653, "step": 2122500 }, { "epoch": 0.77, "learning_rate": 4.695012253722525e-06, "loss": 0.6389, "step": 2123000 }, { "epoch": 0.77, "learning_rate": 4.691407687602347e-06, "loss": 0.6667, "step": 2123500 }, { "epoch": 0.77, "learning_rate": 4.687803121482169e-06, "loss": 0.6421, "step": 2124000 }, { "epoch": 0.77, "learning_rate": 4.6841985553619905e-06, "loss": 0.6634, "step": 2124500 }, { "epoch": 0.77, "learning_rate": 4.6805939892418125e-06, "loss": 0.6424, "step": 2125000 }, { "epoch": 0.77, "learning_rate": 4.676989423121634e-06, "loss": 0.6429, "step": 2125500 }, { "epoch": 0.77, "learning_rate": 4.673384857001456e-06, "loss": 0.6406, "step": 2126000 }, { "epoch": 0.77, "learning_rate": 4.669780290881278e-06, "loss": 0.6482, "step": 2126500 }, { "epoch": 0.77, "learning_rate": 4.666175724761099e-06, "loss": 0.657, "step": 2127000 }, { "epoch": 0.77, "learning_rate": 4.662571158640921e-06, "loss": 0.6215, "step": 2127500 }, { "epoch": 0.77, "learning_rate": 4.658966592520742e-06, "loss": 0.6424, "step": 2128000 }, { "epoch": 0.77, "learning_rate": 4.655362026400563e-06, "loss": 0.6547, "step": 2128500 }, { "epoch": 0.77, "learning_rate": 4.651757460280385e-06, "loss": 0.6284, "step": 2129000 }, { "epoch": 0.77, "learning_rate": 4.648152894160206e-06, "loss": 0.6441, "step": 2129500 }, { "epoch": 0.77, "learning_rate": 4.644548328040028e-06, "loss": 0.6214, "step": 2130000 }, { "epoch": 0.77, "learning_rate": 4.6409437619198495e-06, "loss": 0.6411, "step": 2130500 }, { "epoch": 0.77, "learning_rate": 4.6373391957996716e-06, "loss": 0.6541, "step": 2131000 }, { "epoch": 0.77, "learning_rate": 4.6337346296794936e-06, "loss": 0.6286, "step": 2131500 }, { "epoch": 0.77, "learning_rate": 4.630130063559315e-06, "loss": 0.6621, "step": 2132000 }, { "epoch": 0.77, "learning_rate": 4.626525497439137e-06, "loss": 0.6629, "step": 2132500 }, { "epoch": 0.77, "learning_rate": 4.622920931318958e-06, "loss": 0.6566, "step": 2133000 }, { "epoch": 0.77, "learning_rate": 4.619316365198779e-06, "loss": 0.681, "step": 2133500 }, { "epoch": 0.77, "learning_rate": 4.615711799078601e-06, "loss": 0.6206, "step": 2134000 }, { "epoch": 0.77, "learning_rate": 4.612107232958422e-06, "loss": 0.6429, "step": 2134500 }, { "epoch": 0.77, "learning_rate": 4.608502666838244e-06, "loss": 0.6603, "step": 2135000 }, { "epoch": 0.77, "learning_rate": 4.604898100718066e-06, "loss": 0.6753, "step": 2135500 }, { "epoch": 0.77, "learning_rate": 4.6012935345978875e-06, "loss": 0.6481, "step": 2136000 }, { "epoch": 0.77, "learning_rate": 4.5976889684777095e-06, "loss": 0.6345, "step": 2136500 }, { "epoch": 0.77, "learning_rate": 4.594084402357531e-06, "loss": 0.6443, "step": 2137000 }, { "epoch": 0.77, "learning_rate": 4.590479836237353e-06, "loss": 0.6638, "step": 2137500 }, { "epoch": 0.77, "learning_rate": 4.586875270117175e-06, "loss": 0.6626, "step": 2138000 }, { "epoch": 0.77, "learning_rate": 4.583270703996995e-06, "loss": 0.6586, "step": 2138500 }, { "epoch": 0.77, "learning_rate": 4.579666137876817e-06, "loss": 0.6556, "step": 2139000 }, { "epoch": 0.77, "learning_rate": 4.576061571756639e-06, "loss": 0.6542, "step": 2139500 }, { "epoch": 0.77, "learning_rate": 4.57245700563646e-06, "loss": 0.6461, "step": 2140000 }, { "epoch": 0.77, "learning_rate": 4.568852439516282e-06, "loss": 0.669, "step": 2140500 }, { "epoch": 0.77, "learning_rate": 4.565247873396103e-06, "loss": 0.625, "step": 2141000 }, { "epoch": 0.77, "learning_rate": 4.561643307275925e-06, "loss": 0.6385, "step": 2141500 }, { "epoch": 0.77, "learning_rate": 4.5580387411557465e-06, "loss": 0.6325, "step": 2142000 }, { "epoch": 0.77, "learning_rate": 4.5544341750355686e-06, "loss": 0.6425, "step": 2142500 }, { "epoch": 0.77, "learning_rate": 4.5508296089153906e-06, "loss": 0.6373, "step": 2143000 }, { "epoch": 0.77, "learning_rate": 4.547225042795212e-06, "loss": 0.6579, "step": 2143500 }, { "epoch": 0.77, "learning_rate": 4.543620476675033e-06, "loss": 0.6512, "step": 2144000 }, { "epoch": 0.77, "learning_rate": 4.540015910554855e-06, "loss": 0.645, "step": 2144500 }, { "epoch": 0.77, "learning_rate": 4.536411344434676e-06, "loss": 0.6533, "step": 2145000 }, { "epoch": 0.77, "learning_rate": 4.532806778314498e-06, "loss": 0.6566, "step": 2145500 }, { "epoch": 0.77, "learning_rate": 4.529202212194319e-06, "loss": 0.6351, "step": 2146000 }, { "epoch": 0.77, "learning_rate": 4.525597646074141e-06, "loss": 0.6496, "step": 2146500 }, { "epoch": 0.77, "learning_rate": 4.521993079953963e-06, "loss": 0.6564, "step": 2147000 }, { "epoch": 0.77, "learning_rate": 4.5183885138337845e-06, "loss": 0.6619, "step": 2147500 }, { "epoch": 0.77, "learning_rate": 4.5147839477136065e-06, "loss": 0.6489, "step": 2148000 }, { "epoch": 0.77, "learning_rate": 4.511179381593428e-06, "loss": 0.64, "step": 2148500 }, { "epoch": 0.77, "learning_rate": 4.507574815473249e-06, "loss": 0.6449, "step": 2149000 }, { "epoch": 0.77, "learning_rate": 4.503970249353071e-06, "loss": 0.6293, "step": 2149500 }, { "epoch": 0.77, "learning_rate": 4.500365683232892e-06, "loss": 0.6352, "step": 2150000 }, { "epoch": 0.78, "learning_rate": 4.496761117112714e-06, "loss": 0.6513, "step": 2150500 }, { "epoch": 0.78, "learning_rate": 4.493156550992536e-06, "loss": 0.6325, "step": 2151000 }, { "epoch": 0.78, "learning_rate": 4.489551984872357e-06, "loss": 0.6295, "step": 2151500 }, { "epoch": 0.78, "learning_rate": 4.485947418752179e-06, "loss": 0.6464, "step": 2152000 }, { "epoch": 0.78, "learning_rate": 4.482342852632e-06, "loss": 0.6511, "step": 2152500 }, { "epoch": 0.78, "learning_rate": 4.478738286511822e-06, "loss": 0.6499, "step": 2153000 }, { "epoch": 0.78, "learning_rate": 4.4751337203916435e-06, "loss": 0.6368, "step": 2153500 }, { "epoch": 0.78, "learning_rate": 4.471529154271465e-06, "loss": 0.6707, "step": 2154000 }, { "epoch": 0.78, "learning_rate": 4.467924588151287e-06, "loss": 0.6495, "step": 2154500 }, { "epoch": 0.78, "learning_rate": 4.464320022031109e-06, "loss": 0.6315, "step": 2155000 }, { "epoch": 0.78, "learning_rate": 4.46071545591093e-06, "loss": 0.6541, "step": 2155500 }, { "epoch": 0.78, "learning_rate": 4.457110889790752e-06, "loss": 0.6594, "step": 2156000 }, { "epoch": 0.78, "learning_rate": 4.453506323670573e-06, "loss": 0.6544, "step": 2156500 }, { "epoch": 0.78, "learning_rate": 4.449901757550395e-06, "loss": 0.6389, "step": 2157000 }, { "epoch": 0.78, "learning_rate": 4.446297191430216e-06, "loss": 0.6658, "step": 2157500 }, { "epoch": 0.78, "learning_rate": 4.442692625310038e-06, "loss": 0.6257, "step": 2158000 }, { "epoch": 0.78, "learning_rate": 4.43908805918986e-06, "loss": 0.6435, "step": 2158500 }, { "epoch": 0.78, "learning_rate": 4.4354834930696815e-06, "loss": 0.652, "step": 2159000 }, { "epoch": 0.78, "learning_rate": 4.431878926949503e-06, "loss": 0.629, "step": 2159500 }, { "epoch": 0.78, "learning_rate": 4.428274360829325e-06, "loss": 0.6398, "step": 2160000 }, { "epoch": 0.78, "learning_rate": 4.424669794709146e-06, "loss": 0.6599, "step": 2160500 }, { "epoch": 0.78, "learning_rate": 4.421065228588968e-06, "loss": 0.6469, "step": 2161000 }, { "epoch": 0.78, "learning_rate": 4.417460662468789e-06, "loss": 0.6782, "step": 2161500 }, { "epoch": 0.78, "learning_rate": 4.413856096348611e-06, "loss": 0.6591, "step": 2162000 }, { "epoch": 0.78, "learning_rate": 4.410251530228433e-06, "loss": 0.6875, "step": 2162500 }, { "epoch": 0.78, "learning_rate": 4.406646964108254e-06, "loss": 0.6309, "step": 2163000 }, { "epoch": 0.78, "learning_rate": 4.403042397988076e-06, "loss": 0.6621, "step": 2163500 }, { "epoch": 0.78, "learning_rate": 4.399437831867897e-06, "loss": 0.6353, "step": 2164000 }, { "epoch": 0.78, "learning_rate": 4.3958332657477185e-06, "loss": 0.6524, "step": 2164500 }, { "epoch": 0.78, "learning_rate": 4.3922286996275405e-06, "loss": 0.6341, "step": 2165000 }, { "epoch": 0.78, "learning_rate": 4.388624133507362e-06, "loss": 0.6283, "step": 2165500 }, { "epoch": 0.78, "learning_rate": 4.385019567387184e-06, "loss": 0.6488, "step": 2166000 }, { "epoch": 0.78, "learning_rate": 4.381415001267006e-06, "loss": 0.6775, "step": 2166500 }, { "epoch": 0.78, "learning_rate": 4.377810435146827e-06, "loss": 0.632, "step": 2167000 }, { "epoch": 0.78, "learning_rate": 4.374205869026649e-06, "loss": 0.6482, "step": 2167500 }, { "epoch": 0.78, "learning_rate": 4.37060130290647e-06, "loss": 0.6552, "step": 2168000 }, { "epoch": 0.78, "learning_rate": 4.366996736786292e-06, "loss": 0.6555, "step": 2168500 }, { "epoch": 0.78, "learning_rate": 4.363392170666113e-06, "loss": 0.6697, "step": 2169000 }, { "epoch": 0.78, "learning_rate": 4.3597876045459344e-06, "loss": 0.6386, "step": 2169500 }, { "epoch": 0.78, "learning_rate": 4.3561830384257564e-06, "loss": 0.6575, "step": 2170000 }, { "epoch": 0.78, "learning_rate": 4.3525784723055785e-06, "loss": 0.6284, "step": 2170500 }, { "epoch": 0.78, "learning_rate": 4.3489739061854e-06, "loss": 0.6454, "step": 2171000 }, { "epoch": 0.78, "learning_rate": 4.345369340065222e-06, "loss": 0.6563, "step": 2171500 }, { "epoch": 0.78, "learning_rate": 4.341764773945043e-06, "loss": 0.6425, "step": 2172000 }, { "epoch": 0.78, "learning_rate": 4.338160207824865e-06, "loss": 0.622, "step": 2172500 }, { "epoch": 0.78, "learning_rate": 4.334555641704686e-06, "loss": 0.6412, "step": 2173000 }, { "epoch": 0.78, "learning_rate": 4.330951075584508e-06, "loss": 0.6342, "step": 2173500 }, { "epoch": 0.78, "learning_rate": 4.32734650946433e-06, "loss": 0.6627, "step": 2174000 }, { "epoch": 0.78, "learning_rate": 4.323741943344151e-06, "loss": 0.67, "step": 2174500 }, { "epoch": 0.78, "learning_rate": 4.320137377223972e-06, "loss": 0.6638, "step": 2175000 }, { "epoch": 0.78, "learning_rate": 4.316532811103794e-06, "loss": 0.6533, "step": 2175500 }, { "epoch": 0.78, "learning_rate": 4.3129282449836155e-06, "loss": 0.6438, "step": 2176000 }, { "epoch": 0.78, "learning_rate": 4.3093236788634375e-06, "loss": 0.6471, "step": 2176500 }, { "epoch": 0.78, "learning_rate": 4.305719112743259e-06, "loss": 0.6339, "step": 2177000 }, { "epoch": 0.78, "learning_rate": 4.302114546623081e-06, "loss": 0.6482, "step": 2177500 }, { "epoch": 0.79, "learning_rate": 4.298509980502903e-06, "loss": 0.6433, "step": 2178000 }, { "epoch": 0.79, "learning_rate": 4.294905414382724e-06, "loss": 0.6923, "step": 2178500 }, { "epoch": 0.79, "learning_rate": 4.291300848262546e-06, "loss": 0.6283, "step": 2179000 }, { "epoch": 0.79, "learning_rate": 4.287696282142367e-06, "loss": 0.6537, "step": 2179500 }, { "epoch": 0.79, "learning_rate": 4.284091716022188e-06, "loss": 0.6837, "step": 2180000 }, { "epoch": 0.79, "learning_rate": 4.28048714990201e-06, "loss": 0.6427, "step": 2180500 }, { "epoch": 0.79, "learning_rate": 4.2768825837818314e-06, "loss": 0.6522, "step": 2181000 }, { "epoch": 0.79, "learning_rate": 4.2732780176616534e-06, "loss": 0.6383, "step": 2181500 }, { "epoch": 0.79, "learning_rate": 4.2696734515414755e-06, "loss": 0.6641, "step": 2182000 }, { "epoch": 0.79, "learning_rate": 4.266068885421297e-06, "loss": 0.6399, "step": 2182500 }, { "epoch": 0.79, "learning_rate": 4.262464319301119e-06, "loss": 0.6391, "step": 2183000 }, { "epoch": 0.79, "learning_rate": 4.25885975318094e-06, "loss": 0.6204, "step": 2183500 }, { "epoch": 0.79, "learning_rate": 4.255255187060762e-06, "loss": 0.6616, "step": 2184000 }, { "epoch": 0.79, "learning_rate": 4.251650620940583e-06, "loss": 0.6572, "step": 2184500 }, { "epoch": 0.79, "learning_rate": 4.248046054820404e-06, "loss": 0.679, "step": 2185000 }, { "epoch": 0.79, "learning_rate": 4.244441488700226e-06, "loss": 0.6494, "step": 2185500 }, { "epoch": 0.79, "learning_rate": 4.240836922580048e-06, "loss": 0.6708, "step": 2186000 }, { "epoch": 0.79, "learning_rate": 4.237232356459869e-06, "loss": 0.6496, "step": 2186500 }, { "epoch": 0.79, "learning_rate": 4.233627790339691e-06, "loss": 0.6486, "step": 2187000 }, { "epoch": 0.79, "learning_rate": 4.2300232242195125e-06, "loss": 0.6336, "step": 2187500 }, { "epoch": 0.79, "learning_rate": 4.2264186580993345e-06, "loss": 0.6614, "step": 2188000 }, { "epoch": 0.79, "learning_rate": 4.222814091979156e-06, "loss": 0.643, "step": 2188500 }, { "epoch": 0.79, "learning_rate": 4.219209525858978e-06, "loss": 0.6506, "step": 2189000 }, { "epoch": 0.79, "learning_rate": 4.215604959738799e-06, "loss": 0.632, "step": 2189500 }, { "epoch": 0.79, "learning_rate": 4.21200039361862e-06, "loss": 0.6544, "step": 2190000 }, { "epoch": 0.79, "learning_rate": 4.208395827498442e-06, "loss": 0.6718, "step": 2190500 }, { "epoch": 0.79, "learning_rate": 4.204791261378264e-06, "loss": 0.6832, "step": 2191000 }, { "epoch": 0.79, "learning_rate": 4.201186695258085e-06, "loss": 0.6356, "step": 2191500 }, { "epoch": 0.79, "learning_rate": 4.197582129137907e-06, "loss": 0.6524, "step": 2192000 }, { "epoch": 0.79, "learning_rate": 4.1939775630177284e-06, "loss": 0.6429, "step": 2192500 }, { "epoch": 0.79, "learning_rate": 4.1903729968975504e-06, "loss": 0.6559, "step": 2193000 }, { "epoch": 0.79, "learning_rate": 4.1867684307773725e-06, "loss": 0.6354, "step": 2193500 }, { "epoch": 0.79, "learning_rate": 4.183163864657194e-06, "loss": 0.6416, "step": 2194000 }, { "epoch": 0.79, "learning_rate": 4.179559298537016e-06, "loss": 0.6481, "step": 2194500 }, { "epoch": 0.79, "learning_rate": 4.175954732416837e-06, "loss": 0.6409, "step": 2195000 }, { "epoch": 0.79, "learning_rate": 4.172350166296658e-06, "loss": 0.626, "step": 2195500 }, { "epoch": 0.79, "learning_rate": 4.16874560017648e-06, "loss": 0.6358, "step": 2196000 }, { "epoch": 0.79, "learning_rate": 4.165141034056301e-06, "loss": 0.6423, "step": 2196500 }, { "epoch": 0.79, "learning_rate": 4.161536467936123e-06, "loss": 0.6553, "step": 2197000 }, { "epoch": 0.79, "learning_rate": 4.157931901815944e-06, "loss": 0.651, "step": 2197500 }, { "epoch": 0.79, "learning_rate": 4.154327335695766e-06, "loss": 0.639, "step": 2198000 }, { "epoch": 0.79, "learning_rate": 4.150722769575588e-06, "loss": 0.632, "step": 2198500 }, { "epoch": 0.79, "learning_rate": 4.1471182034554095e-06, "loss": 0.6475, "step": 2199000 }, { "epoch": 0.79, "learning_rate": 4.1435136373352315e-06, "loss": 0.6502, "step": 2199500 }, { "epoch": 0.79, "learning_rate": 4.139909071215053e-06, "loss": 0.661, "step": 2200000 }, { "epoch": 0.79, "learning_rate": 4.136304505094874e-06, "loss": 0.6466, "step": 2200500 }, { "epoch": 0.79, "learning_rate": 4.132699938974696e-06, "loss": 0.6506, "step": 2201000 }, { "epoch": 0.79, "learning_rate": 4.129095372854517e-06, "loss": 0.6423, "step": 2201500 }, { "epoch": 0.79, "learning_rate": 4.125490806734339e-06, "loss": 0.6455, "step": 2202000 }, { "epoch": 0.79, "learning_rate": 4.121886240614161e-06, "loss": 0.6404, "step": 2202500 }, { "epoch": 0.79, "learning_rate": 4.118281674493982e-06, "loss": 0.6642, "step": 2203000 }, { "epoch": 0.79, "learning_rate": 4.114677108373804e-06, "loss": 0.6646, "step": 2203500 }, { "epoch": 0.79, "learning_rate": 4.111072542253625e-06, "loss": 0.6533, "step": 2204000 }, { "epoch": 0.79, "learning_rate": 4.1074679761334474e-06, "loss": 0.6551, "step": 2204500 }, { "epoch": 0.79, "learning_rate": 4.103863410013269e-06, "loss": 0.6833, "step": 2205000 }, { "epoch": 0.79, "learning_rate": 4.10025884389309e-06, "loss": 0.648, "step": 2205500 }, { "epoch": 0.8, "learning_rate": 4.096654277772912e-06, "loss": 0.6585, "step": 2206000 }, { "epoch": 0.8, "learning_rate": 4.093049711652734e-06, "loss": 0.6623, "step": 2206500 }, { "epoch": 0.8, "learning_rate": 4.089445145532555e-06, "loss": 0.6335, "step": 2207000 }, { "epoch": 0.8, "learning_rate": 4.085840579412377e-06, "loss": 0.6608, "step": 2207500 }, { "epoch": 0.8, "learning_rate": 4.082236013292198e-06, "loss": 0.62, "step": 2208000 }, { "epoch": 0.8, "learning_rate": 4.07863144717202e-06, "loss": 0.663, "step": 2208500 }, { "epoch": 0.8, "learning_rate": 4.075026881051841e-06, "loss": 0.6301, "step": 2209000 }, { "epoch": 0.8, "learning_rate": 4.071422314931663e-06, "loss": 0.6376, "step": 2209500 }, { "epoch": 0.8, "learning_rate": 4.0678177488114845e-06, "loss": 0.6047, "step": 2210000 }, { "epoch": 0.8, "learning_rate": 4.0642131826913065e-06, "loss": 0.6605, "step": 2210500 }, { "epoch": 0.8, "learning_rate": 4.060608616571128e-06, "loss": 0.6332, "step": 2211000 }, { "epoch": 0.8, "learning_rate": 4.05700405045095e-06, "loss": 0.6579, "step": 2211500 }, { "epoch": 0.8, "learning_rate": 4.053399484330771e-06, "loss": 0.6664, "step": 2212000 }, { "epoch": 0.8, "learning_rate": 4.049794918210593e-06, "loss": 0.6258, "step": 2212500 }, { "epoch": 0.8, "learning_rate": 4.046190352090414e-06, "loss": 0.6393, "step": 2213000 }, { "epoch": 0.8, "learning_rate": 4.042585785970236e-06, "loss": 0.6347, "step": 2213500 }, { "epoch": 0.8, "learning_rate": 4.038981219850058e-06, "loss": 0.681, "step": 2214000 }, { "epoch": 0.8, "learning_rate": 4.035376653729879e-06, "loss": 0.6481, "step": 2214500 }, { "epoch": 0.8, "learning_rate": 4.031772087609701e-06, "loss": 0.6544, "step": 2215000 }, { "epoch": 0.8, "learning_rate": 4.028167521489522e-06, "loss": 0.651, "step": 2215500 }, { "epoch": 0.8, "learning_rate": 4.024562955369344e-06, "loss": 0.6226, "step": 2216000 }, { "epoch": 0.8, "learning_rate": 4.020958389249166e-06, "loss": 0.6584, "step": 2216500 }, { "epoch": 0.8, "learning_rate": 4.017353823128987e-06, "loss": 0.6281, "step": 2217000 }, { "epoch": 0.8, "learning_rate": 4.013749257008809e-06, "loss": 0.6228, "step": 2217500 }, { "epoch": 0.8, "learning_rate": 4.010144690888631e-06, "loss": 0.6548, "step": 2218000 }, { "epoch": 0.8, "learning_rate": 4.006540124768452e-06, "loss": 0.641, "step": 2218500 }, { "epoch": 0.8, "learning_rate": 4.002935558648274e-06, "loss": 0.6641, "step": 2219000 }, { "epoch": 0.8, "learning_rate": 3.999330992528095e-06, "loss": 0.6407, "step": 2219500 }, { "epoch": 0.8, "learning_rate": 3.995726426407917e-06, "loss": 0.6563, "step": 2220000 }, { "epoch": 0.8, "learning_rate": 3.992121860287738e-06, "loss": 0.6465, "step": 2220500 }, { "epoch": 0.8, "learning_rate": 3.9885172941675595e-06, "loss": 0.6483, "step": 2221000 }, { "epoch": 0.8, "learning_rate": 3.9849127280473815e-06, "loss": 0.6418, "step": 2221500 }, { "epoch": 0.8, "learning_rate": 3.9813081619272035e-06, "loss": 0.6676, "step": 2222000 }, { "epoch": 0.8, "learning_rate": 3.977703595807025e-06, "loss": 0.6425, "step": 2222500 }, { "epoch": 0.8, "learning_rate": 3.974099029686847e-06, "loss": 0.6465, "step": 2223000 }, { "epoch": 0.8, "learning_rate": 3.970494463566668e-06, "loss": 0.6623, "step": 2223500 }, { "epoch": 0.8, "learning_rate": 3.96688989744649e-06, "loss": 0.6464, "step": 2224000 }, { "epoch": 0.8, "learning_rate": 3.963285331326311e-06, "loss": 0.631, "step": 2224500 }, { "epoch": 0.8, "learning_rate": 3.959680765206133e-06, "loss": 0.6164, "step": 2225000 }, { "epoch": 0.8, "learning_rate": 3.956076199085954e-06, "loss": 0.6242, "step": 2225500 }, { "epoch": 0.8, "learning_rate": 3.952471632965776e-06, "loss": 0.6448, "step": 2226000 }, { "epoch": 0.8, "learning_rate": 3.948867066845597e-06, "loss": 0.6678, "step": 2226500 }, { "epoch": 0.8, "learning_rate": 3.945262500725419e-06, "loss": 0.6401, "step": 2227000 }, { "epoch": 0.8, "learning_rate": 3.941657934605241e-06, "loss": 0.6673, "step": 2227500 }, { "epoch": 0.8, "learning_rate": 3.938053368485063e-06, "loss": 0.6678, "step": 2228000 }, { "epoch": 0.8, "learning_rate": 3.934448802364884e-06, "loss": 0.6414, "step": 2228500 }, { "epoch": 0.8, "learning_rate": 3.930844236244706e-06, "loss": 0.6693, "step": 2229000 }, { "epoch": 0.8, "learning_rate": 3.927239670124528e-06, "loss": 0.6355, "step": 2229500 }, { "epoch": 0.8, "learning_rate": 3.923635104004349e-06, "loss": 0.6391, "step": 2230000 }, { "epoch": 0.8, "learning_rate": 3.92003053788417e-06, "loss": 0.6836, "step": 2230500 }, { "epoch": 0.8, "learning_rate": 3.916425971763992e-06, "loss": 0.6305, "step": 2231000 }, { "epoch": 0.8, "learning_rate": 3.912821405643813e-06, "loss": 0.6628, "step": 2231500 }, { "epoch": 0.8, "learning_rate": 3.909216839523635e-06, "loss": 0.627, "step": 2232000 }, { "epoch": 0.8, "learning_rate": 3.9056122734034565e-06, "loss": 0.6355, "step": 2232500 }, { "epoch": 0.8, "learning_rate": 3.9020077072832785e-06, "loss": 0.6479, "step": 2233000 }, { "epoch": 0.81, "learning_rate": 3.8984031411631005e-06, "loss": 0.6639, "step": 2233500 }, { "epoch": 0.81, "learning_rate": 3.894798575042922e-06, "loss": 0.6576, "step": 2234000 }, { "epoch": 0.81, "learning_rate": 3.891194008922744e-06, "loss": 0.6376, "step": 2234500 }, { "epoch": 0.81, "learning_rate": 3.887589442802565e-06, "loss": 0.6429, "step": 2235000 }, { "epoch": 0.81, "learning_rate": 3.883984876682387e-06, "loss": 0.6562, "step": 2235500 }, { "epoch": 0.81, "learning_rate": 3.880380310562208e-06, "loss": 0.6492, "step": 2236000 }, { "epoch": 0.81, "learning_rate": 3.876775744442029e-06, "loss": 0.6698, "step": 2236500 }, { "epoch": 0.81, "learning_rate": 3.873171178321851e-06, "loss": 0.6603, "step": 2237000 }, { "epoch": 0.81, "learning_rate": 3.869566612201673e-06, "loss": 0.6317, "step": 2237500 }, { "epoch": 0.81, "learning_rate": 3.865962046081494e-06, "loss": 0.6295, "step": 2238000 }, { "epoch": 0.81, "learning_rate": 3.862357479961316e-06, "loss": 0.648, "step": 2238500 }, { "epoch": 0.81, "learning_rate": 3.858752913841138e-06, "loss": 0.6347, "step": 2239000 }, { "epoch": 0.81, "learning_rate": 3.85514834772096e-06, "loss": 0.6201, "step": 2239500 }, { "epoch": 0.81, "learning_rate": 3.851543781600781e-06, "loss": 0.656, "step": 2240000 }, { "epoch": 0.81, "learning_rate": 3.847939215480603e-06, "loss": 0.6528, "step": 2240500 }, { "epoch": 0.81, "learning_rate": 3.844334649360424e-06, "loss": 0.6465, "step": 2241000 }, { "epoch": 0.81, "learning_rate": 3.840730083240246e-06, "loss": 0.6498, "step": 2241500 }, { "epoch": 0.81, "learning_rate": 3.837125517120067e-06, "loss": 0.6588, "step": 2242000 }, { "epoch": 0.81, "learning_rate": 3.833520950999889e-06, "loss": 0.638, "step": 2242500 }, { "epoch": 0.81, "learning_rate": 3.82991638487971e-06, "loss": 0.631, "step": 2243000 }, { "epoch": 0.81, "learning_rate": 3.826311818759532e-06, "loss": 0.658, "step": 2243500 }, { "epoch": 0.81, "learning_rate": 3.8227072526393535e-06, "loss": 0.6191, "step": 2244000 }, { "epoch": 0.81, "learning_rate": 3.8191026865191755e-06, "loss": 0.6324, "step": 2244500 }, { "epoch": 0.81, "learning_rate": 3.8154981203989975e-06, "loss": 0.6419, "step": 2245000 }, { "epoch": 0.81, "learning_rate": 3.8118935542788187e-06, "loss": 0.6283, "step": 2245500 }, { "epoch": 0.81, "learning_rate": 3.80828898815864e-06, "loss": 0.6119, "step": 2246000 }, { "epoch": 0.81, "learning_rate": 3.8046844220384614e-06, "loss": 0.6377, "step": 2246500 }, { "epoch": 0.81, "learning_rate": 3.801079855918283e-06, "loss": 0.6187, "step": 2247000 }, { "epoch": 0.81, "learning_rate": 3.797475289798105e-06, "loss": 0.6608, "step": 2247500 }, { "epoch": 0.81, "learning_rate": 3.7938707236779266e-06, "loss": 0.6468, "step": 2248000 }, { "epoch": 0.81, "learning_rate": 3.7902661575577482e-06, "loss": 0.6428, "step": 2248500 }, { "epoch": 0.81, "learning_rate": 3.78666159143757e-06, "loss": 0.6378, "step": 2249000 }, { "epoch": 0.81, "learning_rate": 3.7830570253173914e-06, "loss": 0.6392, "step": 2249500 }, { "epoch": 0.81, "learning_rate": 3.779452459197213e-06, "loss": 0.6516, "step": 2250000 }, { "epoch": 0.81, "learning_rate": 3.775847893077035e-06, "loss": 0.6479, "step": 2250500 }, { "epoch": 0.81, "learning_rate": 3.7722433269568557e-06, "loss": 0.6521, "step": 2251000 }, { "epoch": 0.81, "learning_rate": 3.7686387608366778e-06, "loss": 0.6412, "step": 2251500 }, { "epoch": 0.81, "learning_rate": 3.7650341947164994e-06, "loss": 0.6439, "step": 2252000 }, { "epoch": 0.81, "learning_rate": 3.761429628596321e-06, "loss": 0.6191, "step": 2252500 }, { "epoch": 0.81, "learning_rate": 3.7578250624761425e-06, "loss": 0.6352, "step": 2253000 }, { "epoch": 0.81, "learning_rate": 3.754220496355964e-06, "loss": 0.6229, "step": 2253500 }, { "epoch": 0.81, "learning_rate": 3.7506159302357857e-06, "loss": 0.6263, "step": 2254000 }, { "epoch": 0.81, "learning_rate": 3.7470113641156073e-06, "loss": 0.613, "step": 2254500 }, { "epoch": 0.81, "learning_rate": 3.7434067979954293e-06, "loss": 0.6739, "step": 2255000 }, { "epoch": 0.81, "learning_rate": 3.739802231875251e-06, "loss": 0.6286, "step": 2255500 }, { "epoch": 0.81, "learning_rate": 3.736197665755072e-06, "loss": 0.6419, "step": 2256000 }, { "epoch": 0.81, "learning_rate": 3.7325930996348937e-06, "loss": 0.6536, "step": 2256500 }, { "epoch": 0.81, "learning_rate": 3.7289885335147153e-06, "loss": 0.6324, "step": 2257000 }, { "epoch": 0.81, "learning_rate": 3.725383967394537e-06, "loss": 0.6351, "step": 2257500 }, { "epoch": 0.81, "learning_rate": 3.7217794012743584e-06, "loss": 0.6404, "step": 2258000 }, { "epoch": 0.81, "learning_rate": 3.71817483515418e-06, "loss": 0.6471, "step": 2258500 }, { "epoch": 0.81, "learning_rate": 3.714570269034002e-06, "loss": 0.6196, "step": 2259000 }, { "epoch": 0.81, "learning_rate": 3.7109657029138236e-06, "loss": 0.6281, "step": 2259500 }, { "epoch": 0.81, "learning_rate": 3.7073611367936452e-06, "loss": 0.6512, "step": 2260000 }, { "epoch": 0.81, "learning_rate": 3.703756570673467e-06, "loss": 0.6408, "step": 2260500 }, { "epoch": 0.81, "learning_rate": 3.7001520045532884e-06, "loss": 0.6468, "step": 2261000 }, { "epoch": 0.82, "learning_rate": 3.6965474384331096e-06, "loss": 0.6833, "step": 2261500 }, { "epoch": 0.82, "learning_rate": 3.692942872312931e-06, "loss": 0.6696, "step": 2262000 }, { "epoch": 0.82, "learning_rate": 3.6893383061927527e-06, "loss": 0.6231, "step": 2262500 }, { "epoch": 0.82, "learning_rate": 3.6857337400725748e-06, "loss": 0.6221, "step": 2263000 }, { "epoch": 0.82, "learning_rate": 3.6821291739523964e-06, "loss": 0.636, "step": 2263500 }, { "epoch": 0.82, "learning_rate": 3.678524607832218e-06, "loss": 0.6129, "step": 2264000 }, { "epoch": 0.82, "learning_rate": 3.6749200417120395e-06, "loss": 0.634, "step": 2264500 }, { "epoch": 0.82, "learning_rate": 3.671315475591861e-06, "loss": 0.6443, "step": 2265000 }, { "epoch": 0.82, "learning_rate": 3.6677109094716827e-06, "loss": 0.6474, "step": 2265500 }, { "epoch": 0.82, "learning_rate": 3.6641063433515043e-06, "loss": 0.6487, "step": 2266000 }, { "epoch": 0.82, "learning_rate": 3.6605017772313255e-06, "loss": 0.681, "step": 2266500 }, { "epoch": 0.82, "learning_rate": 3.656897211111147e-06, "loss": 0.631, "step": 2267000 }, { "epoch": 0.82, "learning_rate": 3.653292644990969e-06, "loss": 0.6182, "step": 2267500 }, { "epoch": 0.82, "learning_rate": 3.6496880788707907e-06, "loss": 0.6278, "step": 2268000 }, { "epoch": 0.82, "learning_rate": 3.6460835127506123e-06, "loss": 0.6662, "step": 2268500 }, { "epoch": 0.82, "learning_rate": 3.642478946630434e-06, "loss": 0.6437, "step": 2269000 }, { "epoch": 0.82, "learning_rate": 3.6388743805102554e-06, "loss": 0.6246, "step": 2269500 }, { "epoch": 0.82, "learning_rate": 3.635269814390077e-06, "loss": 0.6362, "step": 2270000 }, { "epoch": 0.82, "learning_rate": 3.631665248269899e-06, "loss": 0.6449, "step": 2270500 }, { "epoch": 0.82, "learning_rate": 3.6280606821497206e-06, "loss": 0.6638, "step": 2271000 }, { "epoch": 0.82, "learning_rate": 3.624456116029542e-06, "loss": 0.6318, "step": 2271500 }, { "epoch": 0.82, "learning_rate": 3.6208515499093634e-06, "loss": 0.6511, "step": 2272000 }, { "epoch": 0.82, "learning_rate": 3.617246983789185e-06, "loss": 0.6364, "step": 2272500 }, { "epoch": 0.82, "learning_rate": 3.6136424176690066e-06, "loss": 0.6207, "step": 2273000 }, { "epoch": 0.82, "learning_rate": 3.610037851548828e-06, "loss": 0.6358, "step": 2273500 }, { "epoch": 0.82, "learning_rate": 3.6064332854286497e-06, "loss": 0.6382, "step": 2274000 }, { "epoch": 0.82, "learning_rate": 3.6028287193084718e-06, "loss": 0.6387, "step": 2274500 }, { "epoch": 0.82, "learning_rate": 3.5992241531882933e-06, "loss": 0.6236, "step": 2275000 }, { "epoch": 0.82, "learning_rate": 3.595619587068115e-06, "loss": 0.6445, "step": 2275500 }, { "epoch": 0.82, "learning_rate": 3.5920150209479365e-06, "loss": 0.6565, "step": 2276000 }, { "epoch": 0.82, "learning_rate": 3.5884104548277577e-06, "loss": 0.6446, "step": 2276500 }, { "epoch": 0.82, "learning_rate": 3.5848058887075793e-06, "loss": 0.6484, "step": 2277000 }, { "epoch": 0.82, "learning_rate": 3.581201322587401e-06, "loss": 0.6698, "step": 2277500 }, { "epoch": 0.82, "learning_rate": 3.5775967564672225e-06, "loss": 0.6528, "step": 2278000 }, { "epoch": 0.82, "learning_rate": 3.573992190347044e-06, "loss": 0.6315, "step": 2278500 }, { "epoch": 0.82, "learning_rate": 3.570387624226866e-06, "loss": 0.625, "step": 2279000 }, { "epoch": 0.82, "learning_rate": 3.5667830581066877e-06, "loss": 0.6355, "step": 2279500 }, { "epoch": 0.82, "learning_rate": 3.5631784919865093e-06, "loss": 0.6378, "step": 2280000 }, { "epoch": 0.82, "learning_rate": 3.559573925866331e-06, "loss": 0.6601, "step": 2280500 }, { "epoch": 0.82, "learning_rate": 3.5559693597461524e-06, "loss": 0.6645, "step": 2281000 }, { "epoch": 0.82, "learning_rate": 3.552364793625974e-06, "loss": 0.6521, "step": 2281500 }, { "epoch": 0.82, "learning_rate": 3.548760227505795e-06, "loss": 0.6454, "step": 2282000 }, { "epoch": 0.82, "learning_rate": 3.5451556613856168e-06, "loss": 0.6376, "step": 2282500 }, { "epoch": 0.82, "learning_rate": 3.541551095265439e-06, "loss": 0.6247, "step": 2283000 }, { "epoch": 0.82, "learning_rate": 3.5379465291452604e-06, "loss": 0.6614, "step": 2283500 }, { "epoch": 0.82, "learning_rate": 3.534341963025082e-06, "loss": 0.6267, "step": 2284000 }, { "epoch": 0.82, "learning_rate": 3.5307373969049036e-06, "loss": 0.6695, "step": 2284500 }, { "epoch": 0.82, "learning_rate": 3.527132830784725e-06, "loss": 0.6516, "step": 2285000 }, { "epoch": 0.82, "learning_rate": 3.5235282646645467e-06, "loss": 0.645, "step": 2285500 }, { "epoch": 0.82, "learning_rate": 3.5199236985443688e-06, "loss": 0.614, "step": 2286000 }, { "epoch": 0.82, "learning_rate": 3.5163191324241903e-06, "loss": 0.6282, "step": 2286500 }, { "epoch": 0.82, "learning_rate": 3.5127145663040115e-06, "loss": 0.6312, "step": 2287000 }, { "epoch": 0.82, "learning_rate": 3.509110000183833e-06, "loss": 0.656, "step": 2287500 }, { "epoch": 0.82, "learning_rate": 3.5055054340636547e-06, "loss": 0.6565, "step": 2288000 }, { "epoch": 0.82, "learning_rate": 3.5019008679434763e-06, "loss": 0.6607, "step": 2288500 }, { "epoch": 0.83, "learning_rate": 3.498296301823298e-06, "loss": 0.6257, "step": 2289000 }, { "epoch": 0.83, "learning_rate": 3.4946917357031195e-06, "loss": 0.6629, "step": 2289500 }, { "epoch": 0.83, "learning_rate": 3.491087169582941e-06, "loss": 0.6166, "step": 2290000 }, { "epoch": 0.83, "learning_rate": 3.487482603462763e-06, "loss": 0.6395, "step": 2290500 }, { "epoch": 0.83, "learning_rate": 3.4838780373425847e-06, "loss": 0.648, "step": 2291000 }, { "epoch": 0.83, "learning_rate": 3.4802734712224062e-06, "loss": 0.6592, "step": 2291500 }, { "epoch": 0.83, "learning_rate": 3.4766689051022274e-06, "loss": 0.634, "step": 2292000 }, { "epoch": 0.83, "learning_rate": 3.473064338982049e-06, "loss": 0.6277, "step": 2292500 }, { "epoch": 0.83, "learning_rate": 3.4694597728618706e-06, "loss": 0.6361, "step": 2293000 }, { "epoch": 0.83, "learning_rate": 3.465855206741692e-06, "loss": 0.6585, "step": 2293500 }, { "epoch": 0.83, "learning_rate": 3.4622506406215138e-06, "loss": 0.6569, "step": 2294000 }, { "epoch": 0.83, "learning_rate": 3.458646074501336e-06, "loss": 0.6567, "step": 2294500 }, { "epoch": 0.83, "learning_rate": 3.4550415083811574e-06, "loss": 0.6389, "step": 2295000 }, { "epoch": 0.83, "learning_rate": 3.451436942260979e-06, "loss": 0.6292, "step": 2295500 }, { "epoch": 0.83, "learning_rate": 3.4478323761408006e-06, "loss": 0.6452, "step": 2296000 }, { "epoch": 0.83, "learning_rate": 3.444227810020622e-06, "loss": 0.6381, "step": 2296500 }, { "epoch": 0.83, "learning_rate": 3.4406232439004433e-06, "loss": 0.621, "step": 2297000 }, { "epoch": 0.83, "learning_rate": 3.437018677780265e-06, "loss": 0.6385, "step": 2297500 }, { "epoch": 0.83, "learning_rate": 3.4334141116600865e-06, "loss": 0.6406, "step": 2298000 }, { "epoch": 0.83, "learning_rate": 3.4298095455399085e-06, "loss": 0.6557, "step": 2298500 }, { "epoch": 0.83, "learning_rate": 3.42620497941973e-06, "loss": 0.6226, "step": 2299000 }, { "epoch": 0.83, "learning_rate": 3.4226004132995517e-06, "loss": 0.6473, "step": 2299500 }, { "epoch": 0.83, "learning_rate": 3.4189958471793733e-06, "loss": 0.6046, "step": 2300000 }, { "epoch": 0.83, "learning_rate": 3.415391281059195e-06, "loss": 0.6234, "step": 2300500 }, { "epoch": 0.83, "learning_rate": 3.4117867149390165e-06, "loss": 0.6469, "step": 2301000 }, { "epoch": 0.83, "learning_rate": 3.408182148818838e-06, "loss": 0.6523, "step": 2301500 }, { "epoch": 0.83, "learning_rate": 3.40457758269866e-06, "loss": 0.6266, "step": 2302000 }, { "epoch": 0.83, "learning_rate": 3.400973016578481e-06, "loss": 0.6583, "step": 2302500 }, { "epoch": 0.83, "learning_rate": 3.397368450458303e-06, "loss": 0.6227, "step": 2303000 }, { "epoch": 0.83, "learning_rate": 3.3937638843381244e-06, "loss": 0.6443, "step": 2303500 }, { "epoch": 0.83, "learning_rate": 3.390159318217946e-06, "loss": 0.6122, "step": 2304000 }, { "epoch": 0.83, "learning_rate": 3.3865547520977676e-06, "loss": 0.6458, "step": 2304500 }, { "epoch": 0.83, "learning_rate": 3.382950185977589e-06, "loss": 0.6628, "step": 2305000 }, { "epoch": 0.83, "learning_rate": 3.3793456198574108e-06, "loss": 0.6257, "step": 2305500 }, { "epoch": 0.83, "learning_rate": 3.3757410537372328e-06, "loss": 0.6522, "step": 2306000 }, { "epoch": 0.83, "learning_rate": 3.3721364876170544e-06, "loss": 0.6278, "step": 2306500 }, { "epoch": 0.83, "learning_rate": 3.368531921496876e-06, "loss": 0.6447, "step": 2307000 }, { "epoch": 0.83, "learning_rate": 3.364927355376697e-06, "loss": 0.6702, "step": 2307500 }, { "epoch": 0.83, "learning_rate": 3.3613227892565187e-06, "loss": 0.6515, "step": 2308000 }, { "epoch": 0.83, "learning_rate": 3.3577182231363403e-06, "loss": 0.6536, "step": 2308500 }, { "epoch": 0.83, "learning_rate": 3.354113657016162e-06, "loss": 0.6224, "step": 2309000 }, { "epoch": 0.83, "learning_rate": 3.3505090908959835e-06, "loss": 0.6232, "step": 2309500 }, { "epoch": 0.83, "learning_rate": 3.346904524775805e-06, "loss": 0.6192, "step": 2310000 }, { "epoch": 0.83, "learning_rate": 3.343299958655627e-06, "loss": 0.6239, "step": 2310500 }, { "epoch": 0.83, "learning_rate": 3.3396953925354487e-06, "loss": 0.6298, "step": 2311000 }, { "epoch": 0.83, "learning_rate": 3.3360908264152703e-06, "loss": 0.6483, "step": 2311500 }, { "epoch": 0.83, "learning_rate": 3.332486260295092e-06, "loss": 0.6115, "step": 2312000 }, { "epoch": 0.83, "learning_rate": 3.328881694174913e-06, "loss": 0.6263, "step": 2312500 }, { "epoch": 0.83, "learning_rate": 3.3252771280547346e-06, "loss": 0.6326, "step": 2313000 }, { "epoch": 0.83, "learning_rate": 3.3216725619345562e-06, "loss": 0.6434, "step": 2313500 }, { "epoch": 0.83, "learning_rate": 3.318067995814378e-06, "loss": 0.6463, "step": 2314000 }, { "epoch": 0.83, "learning_rate": 3.3144634296942e-06, "loss": 0.6513, "step": 2314500 }, { "epoch": 0.83, "learning_rate": 3.3108588635740214e-06, "loss": 0.6435, "step": 2315000 }, { "epoch": 0.83, "learning_rate": 3.307254297453843e-06, "loss": 0.6618, "step": 2315500 }, { "epoch": 0.83, "learning_rate": 3.3036497313336646e-06, "loss": 0.6312, "step": 2316000 }, { "epoch": 0.83, "learning_rate": 3.300045165213486e-06, "loss": 0.6307, "step": 2316500 }, { "epoch": 0.84, "learning_rate": 3.2964405990933078e-06, "loss": 0.6564, "step": 2317000 }, { "epoch": 0.84, "learning_rate": 3.292836032973129e-06, "loss": 0.6212, "step": 2317500 }, { "epoch": 0.84, "learning_rate": 3.2892314668529505e-06, "loss": 0.6273, "step": 2318000 }, { "epoch": 0.84, "learning_rate": 3.2856269007327725e-06, "loss": 0.6359, "step": 2318500 }, { "epoch": 0.84, "learning_rate": 3.282022334612594e-06, "loss": 0.6206, "step": 2319000 }, { "epoch": 0.84, "learning_rate": 3.2784177684924157e-06, "loss": 0.6379, "step": 2319500 }, { "epoch": 0.84, "learning_rate": 3.2748132023722373e-06, "loss": 0.6409, "step": 2320000 }, { "epoch": 0.84, "learning_rate": 3.271208636252059e-06, "loss": 0.6351, "step": 2320500 }, { "epoch": 0.84, "learning_rate": 3.2676040701318805e-06, "loss": 0.6316, "step": 2321000 }, { "epoch": 0.84, "learning_rate": 3.263999504011702e-06, "loss": 0.6608, "step": 2321500 }, { "epoch": 0.84, "learning_rate": 3.260394937891524e-06, "loss": 0.653, "step": 2322000 }, { "epoch": 0.84, "learning_rate": 3.2567903717713457e-06, "loss": 0.6338, "step": 2322500 }, { "epoch": 0.84, "learning_rate": 3.253185805651167e-06, "loss": 0.6337, "step": 2323000 }, { "epoch": 0.84, "learning_rate": 3.2495812395309884e-06, "loss": 0.6217, "step": 2323500 }, { "epoch": 0.84, "learning_rate": 3.24597667341081e-06, "loss": 0.6146, "step": 2324000 }, { "epoch": 0.84, "learning_rate": 3.2423721072906316e-06, "loss": 0.6479, "step": 2324500 }, { "epoch": 0.84, "learning_rate": 3.2387675411704532e-06, "loss": 0.6144, "step": 2325000 }, { "epoch": 0.84, "learning_rate": 3.235162975050275e-06, "loss": 0.6181, "step": 2325500 }, { "epoch": 0.84, "learning_rate": 3.231558408930097e-06, "loss": 0.633, "step": 2326000 }, { "epoch": 0.84, "learning_rate": 3.2279538428099184e-06, "loss": 0.6392, "step": 2326500 }, { "epoch": 0.84, "learning_rate": 3.22434927668974e-06, "loss": 0.6317, "step": 2327000 }, { "epoch": 0.84, "learning_rate": 3.2207447105695616e-06, "loss": 0.6229, "step": 2327500 }, { "epoch": 0.84, "learning_rate": 3.2171401444493828e-06, "loss": 0.6695, "step": 2328000 }, { "epoch": 0.84, "learning_rate": 3.2135355783292043e-06, "loss": 0.6421, "step": 2328500 }, { "epoch": 0.84, "learning_rate": 3.209931012209026e-06, "loss": 0.6343, "step": 2329000 }, { "epoch": 0.84, "learning_rate": 3.2063264460888475e-06, "loss": 0.6386, "step": 2329500 }, { "epoch": 0.84, "learning_rate": 3.2027218799686695e-06, "loss": 0.6498, "step": 2330000 }, { "epoch": 0.84, "learning_rate": 3.199117313848491e-06, "loss": 0.6508, "step": 2330500 }, { "epoch": 0.84, "learning_rate": 3.1955127477283127e-06, "loss": 0.6205, "step": 2331000 }, { "epoch": 0.84, "learning_rate": 3.1919081816081343e-06, "loss": 0.6603, "step": 2331500 }, { "epoch": 0.84, "learning_rate": 3.188303615487956e-06, "loss": 0.6608, "step": 2332000 }, { "epoch": 0.84, "learning_rate": 3.1846990493677775e-06, "loss": 0.6336, "step": 2332500 }, { "epoch": 0.84, "learning_rate": 3.1810944832475987e-06, "loss": 0.6223, "step": 2333000 }, { "epoch": 0.84, "learning_rate": 3.1774899171274202e-06, "loss": 0.6366, "step": 2333500 }, { "epoch": 0.84, "learning_rate": 3.173885351007242e-06, "loss": 0.6222, "step": 2334000 }, { "epoch": 0.84, "learning_rate": 3.170280784887064e-06, "loss": 0.6633, "step": 2334500 }, { "epoch": 0.84, "learning_rate": 3.1666762187668854e-06, "loss": 0.6109, "step": 2335000 }, { "epoch": 0.84, "learning_rate": 3.163071652646707e-06, "loss": 0.637, "step": 2335500 }, { "epoch": 0.84, "learning_rate": 3.1594670865265286e-06, "loss": 0.6241, "step": 2336000 }, { "epoch": 0.84, "learning_rate": 3.1558625204063502e-06, "loss": 0.6622, "step": 2336500 }, { "epoch": 0.84, "learning_rate": 3.152257954286172e-06, "loss": 0.6455, "step": 2337000 }, { "epoch": 0.84, "learning_rate": 3.148653388165994e-06, "loss": 0.6193, "step": 2337500 }, { "epoch": 0.84, "learning_rate": 3.1450488220458146e-06, "loss": 0.6249, "step": 2338000 }, { "epoch": 0.84, "learning_rate": 3.1414442559256366e-06, "loss": 0.6137, "step": 2338500 }, { "epoch": 0.84, "learning_rate": 3.137839689805458e-06, "loss": 0.6625, "step": 2339000 }, { "epoch": 0.84, "learning_rate": 3.1342351236852798e-06, "loss": 0.6334, "step": 2339500 }, { "epoch": 0.84, "learning_rate": 3.1306305575651013e-06, "loss": 0.6418, "step": 2340000 }, { "epoch": 0.84, "learning_rate": 3.127025991444923e-06, "loss": 0.6298, "step": 2340500 }, { "epoch": 0.84, "learning_rate": 3.1234214253247445e-06, "loss": 0.6524, "step": 2341000 }, { "epoch": 0.84, "learning_rate": 3.1198168592045665e-06, "loss": 0.6582, "step": 2341500 }, { "epoch": 0.84, "learning_rate": 3.116212293084388e-06, "loss": 0.6341, "step": 2342000 }, { "epoch": 0.84, "learning_rate": 3.1126077269642097e-06, "loss": 0.6112, "step": 2342500 }, { "epoch": 0.84, "learning_rate": 3.1090031608440313e-06, "loss": 0.6259, "step": 2343000 }, { "epoch": 0.84, "learning_rate": 3.1053985947238525e-06, "loss": 0.6252, "step": 2343500 }, { "epoch": 0.84, "learning_rate": 3.101794028603674e-06, "loss": 0.6206, "step": 2344000 }, { "epoch": 0.85, "learning_rate": 3.0981894624834957e-06, "loss": 0.6326, "step": 2344500 }, { "epoch": 0.85, "learning_rate": 3.0945848963633172e-06, "loss": 0.6604, "step": 2345000 }, { "epoch": 0.85, "learning_rate": 3.090980330243139e-06, "loss": 0.6312, "step": 2345500 }, { "epoch": 0.85, "learning_rate": 3.087375764122961e-06, "loss": 0.6425, "step": 2346000 }, { "epoch": 0.85, "learning_rate": 3.0837711980027824e-06, "loss": 0.6546, "step": 2346500 }, { "epoch": 0.85, "learning_rate": 3.080166631882604e-06, "loss": 0.646, "step": 2347000 }, { "epoch": 0.85, "learning_rate": 3.0765620657624256e-06, "loss": 0.6259, "step": 2347500 }, { "epoch": 0.85, "learning_rate": 3.072957499642247e-06, "loss": 0.6137, "step": 2348000 }, { "epoch": 0.85, "learning_rate": 3.0693529335220684e-06, "loss": 0.6312, "step": 2348500 }, { "epoch": 0.85, "learning_rate": 3.06574836740189e-06, "loss": 0.6372, "step": 2349000 }, { "epoch": 0.85, "learning_rate": 3.0621438012817116e-06, "loss": 0.6197, "step": 2349500 }, { "epoch": 0.85, "learning_rate": 3.0585392351615336e-06, "loss": 0.6491, "step": 2350000 }, { "epoch": 0.85, "learning_rate": 3.054934669041355e-06, "loss": 0.648, "step": 2350500 }, { "epoch": 0.85, "learning_rate": 3.0513301029211768e-06, "loss": 0.628, "step": 2351000 }, { "epoch": 0.85, "learning_rate": 3.0477255368009983e-06, "loss": 0.6884, "step": 2351500 }, { "epoch": 0.85, "learning_rate": 3.04412097068082e-06, "loss": 0.6379, "step": 2352000 }, { "epoch": 0.85, "learning_rate": 3.0405164045606415e-06, "loss": 0.6593, "step": 2352500 }, { "epoch": 0.85, "learning_rate": 3.0369118384404635e-06, "loss": 0.6598, "step": 2353000 }, { "epoch": 0.85, "learning_rate": 3.0333072723202843e-06, "loss": 0.6269, "step": 2353500 }, { "epoch": 0.85, "learning_rate": 3.0297027062001063e-06, "loss": 0.6334, "step": 2354000 }, { "epoch": 0.85, "learning_rate": 3.026098140079928e-06, "loss": 0.6323, "step": 2354500 }, { "epoch": 0.85, "learning_rate": 3.0224935739597495e-06, "loss": 0.6493, "step": 2355000 }, { "epoch": 0.85, "learning_rate": 3.018889007839571e-06, "loss": 0.654, "step": 2355500 }, { "epoch": 0.85, "learning_rate": 3.0152844417193927e-06, "loss": 0.6619, "step": 2356000 }, { "epoch": 0.85, "learning_rate": 3.0116798755992142e-06, "loss": 0.6527, "step": 2356500 }, { "epoch": 0.85, "learning_rate": 3.008075309479036e-06, "loss": 0.6044, "step": 2357000 }, { "epoch": 0.85, "learning_rate": 3.004470743358858e-06, "loss": 0.6344, "step": 2357500 }, { "epoch": 0.85, "learning_rate": 3.0008661772386794e-06, "loss": 0.6422, "step": 2358000 }, { "epoch": 0.85, "learning_rate": 2.9972616111185006e-06, "loss": 0.6525, "step": 2358500 }, { "epoch": 0.85, "learning_rate": 2.993657044998322e-06, "loss": 0.6494, "step": 2359000 }, { "epoch": 0.85, "learning_rate": 2.9900524788781438e-06, "loss": 0.6508, "step": 2359500 }, { "epoch": 0.85, "learning_rate": 2.9864479127579654e-06, "loss": 0.6417, "step": 2360000 }, { "epoch": 0.85, "learning_rate": 2.982843346637787e-06, "loss": 0.6784, "step": 2360500 }, { "epoch": 0.85, "learning_rate": 2.9792387805176086e-06, "loss": 0.6456, "step": 2361000 }, { "epoch": 0.85, "learning_rate": 2.9756342143974306e-06, "loss": 0.6127, "step": 2361500 }, { "epoch": 0.85, "learning_rate": 2.972029648277252e-06, "loss": 0.6398, "step": 2362000 }, { "epoch": 0.85, "learning_rate": 2.9684250821570737e-06, "loss": 0.6192, "step": 2362500 }, { "epoch": 0.85, "learning_rate": 2.9648205160368953e-06, "loss": 0.64, "step": 2363000 }, { "epoch": 0.85, "learning_rate": 2.961215949916717e-06, "loss": 0.6513, "step": 2363500 }, { "epoch": 0.85, "learning_rate": 2.957611383796538e-06, "loss": 0.632, "step": 2364000 }, { "epoch": 0.85, "learning_rate": 2.9540068176763597e-06, "loss": 0.6114, "step": 2364500 }, { "epoch": 0.85, "learning_rate": 2.9504022515561813e-06, "loss": 0.6403, "step": 2365000 }, { "epoch": 0.85, "learning_rate": 2.9467976854360033e-06, "loss": 0.6507, "step": 2365500 }, { "epoch": 0.85, "learning_rate": 2.943193119315825e-06, "loss": 0.6293, "step": 2366000 }, { "epoch": 0.85, "learning_rate": 2.9395885531956465e-06, "loss": 0.6659, "step": 2366500 }, { "epoch": 0.85, "learning_rate": 2.935983987075468e-06, "loss": 0.644, "step": 2367000 }, { "epoch": 0.85, "learning_rate": 2.9323794209552897e-06, "loss": 0.6317, "step": 2367500 }, { "epoch": 0.85, "learning_rate": 2.9287748548351112e-06, "loss": 0.6631, "step": 2368000 }, { "epoch": 0.85, "learning_rate": 2.925170288714933e-06, "loss": 0.6572, "step": 2368500 }, { "epoch": 0.85, "learning_rate": 2.921565722594754e-06, "loss": 0.6342, "step": 2369000 }, { "epoch": 0.85, "learning_rate": 2.9179611564745756e-06, "loss": 0.6276, "step": 2369500 }, { "epoch": 0.85, "learning_rate": 2.9143565903543976e-06, "loss": 0.6615, "step": 2370000 }, { "epoch": 0.85, "learning_rate": 2.910752024234219e-06, "loss": 0.6419, "step": 2370500 }, { "epoch": 0.85, "learning_rate": 2.9071474581140408e-06, "loss": 0.6534, "step": 2371000 }, { "epoch": 0.85, "learning_rate": 2.9035428919938624e-06, "loss": 0.6639, "step": 2371500 }, { "epoch": 0.86, "learning_rate": 2.899938325873684e-06, "loss": 0.6338, "step": 2372000 }, { "epoch": 0.86, "learning_rate": 2.8963337597535056e-06, "loss": 0.6817, "step": 2372500 }, { "epoch": 0.86, "learning_rate": 2.8927291936333276e-06, "loss": 0.6443, "step": 2373000 }, { "epoch": 0.86, "learning_rate": 2.889124627513149e-06, "loss": 0.602, "step": 2373500 }, { "epoch": 0.86, "learning_rate": 2.8855200613929703e-06, "loss": 0.6297, "step": 2374000 }, { "epoch": 0.86, "learning_rate": 2.881915495272792e-06, "loss": 0.6523, "step": 2374500 }, { "epoch": 0.86, "learning_rate": 2.8783109291526135e-06, "loss": 0.6536, "step": 2375000 }, { "epoch": 0.86, "learning_rate": 2.874706363032435e-06, "loss": 0.6433, "step": 2375500 }, { "epoch": 0.86, "learning_rate": 2.8711017969122567e-06, "loss": 0.6602, "step": 2376000 }, { "epoch": 0.86, "learning_rate": 2.8674972307920783e-06, "loss": 0.6471, "step": 2376500 }, { "epoch": 0.86, "learning_rate": 2.8638926646719003e-06, "loss": 0.655, "step": 2377000 }, { "epoch": 0.86, "learning_rate": 2.860288098551722e-06, "loss": 0.6199, "step": 2377500 }, { "epoch": 0.86, "learning_rate": 2.8566835324315435e-06, "loss": 0.6573, "step": 2378000 }, { "epoch": 0.86, "learning_rate": 2.853078966311365e-06, "loss": 0.6226, "step": 2378500 }, { "epoch": 0.86, "learning_rate": 2.8494744001911862e-06, "loss": 0.643, "step": 2379000 }, { "epoch": 0.86, "learning_rate": 2.845869834071008e-06, "loss": 0.6444, "step": 2379500 }, { "epoch": 0.86, "learning_rate": 2.8422652679508294e-06, "loss": 0.6207, "step": 2380000 }, { "epoch": 0.86, "learning_rate": 2.838660701830651e-06, "loss": 0.6547, "step": 2380500 }, { "epoch": 0.86, "learning_rate": 2.8350561357104726e-06, "loss": 0.6381, "step": 2381000 }, { "epoch": 0.86, "learning_rate": 2.8314515695902946e-06, "loss": 0.651, "step": 2381500 }, { "epoch": 0.86, "learning_rate": 2.827847003470116e-06, "loss": 0.6226, "step": 2382000 }, { "epoch": 0.86, "learning_rate": 2.8242424373499378e-06, "loss": 0.6377, "step": 2382500 }, { "epoch": 0.86, "learning_rate": 2.8206378712297594e-06, "loss": 0.631, "step": 2383000 }, { "epoch": 0.86, "learning_rate": 2.817033305109581e-06, "loss": 0.6444, "step": 2383500 }, { "epoch": 0.86, "learning_rate": 2.8134287389894026e-06, "loss": 0.6288, "step": 2384000 }, { "epoch": 0.86, "learning_rate": 2.8098241728692237e-06, "loss": 0.6457, "step": 2384500 }, { "epoch": 0.86, "learning_rate": 2.8062196067490453e-06, "loss": 0.6297, "step": 2385000 }, { "epoch": 0.86, "learning_rate": 2.8026150406288673e-06, "loss": 0.6648, "step": 2385500 }, { "epoch": 0.86, "learning_rate": 2.799010474508689e-06, "loss": 0.6366, "step": 2386000 }, { "epoch": 0.86, "learning_rate": 2.7954059083885105e-06, "loss": 0.6156, "step": 2386500 }, { "epoch": 0.86, "learning_rate": 2.791801342268332e-06, "loss": 0.6655, "step": 2387000 }, { "epoch": 0.86, "learning_rate": 2.7881967761481537e-06, "loss": 0.6609, "step": 2387500 }, { "epoch": 0.86, "learning_rate": 2.7845922100279753e-06, "loss": 0.6199, "step": 2388000 }, { "epoch": 0.86, "learning_rate": 2.780987643907797e-06, "loss": 0.622, "step": 2388500 }, { "epoch": 0.86, "learning_rate": 2.777383077787619e-06, "loss": 0.6238, "step": 2389000 }, { "epoch": 0.86, "learning_rate": 2.77377851166744e-06, "loss": 0.613, "step": 2389500 }, { "epoch": 0.86, "learning_rate": 2.7701739455472616e-06, "loss": 0.6365, "step": 2390000 }, { "epoch": 0.86, "learning_rate": 2.7665693794270832e-06, "loss": 0.6331, "step": 2390500 }, { "epoch": 0.86, "learning_rate": 2.762964813306905e-06, "loss": 0.6454, "step": 2391000 }, { "epoch": 0.86, "learning_rate": 2.7593602471867264e-06, "loss": 0.6367, "step": 2391500 }, { "epoch": 0.86, "learning_rate": 2.755755681066548e-06, "loss": 0.657, "step": 2392000 }, { "epoch": 0.86, "learning_rate": 2.7521511149463696e-06, "loss": 0.6259, "step": 2392500 }, { "epoch": 0.86, "learning_rate": 2.7485465488261916e-06, "loss": 0.6323, "step": 2393000 }, { "epoch": 0.86, "learning_rate": 2.744941982706013e-06, "loss": 0.6551, "step": 2393500 }, { "epoch": 0.86, "learning_rate": 2.7413374165858348e-06, "loss": 0.6295, "step": 2394000 }, { "epoch": 0.86, "learning_rate": 2.737732850465656e-06, "loss": 0.629, "step": 2394500 }, { "epoch": 0.86, "learning_rate": 2.7341282843454775e-06, "loss": 0.635, "step": 2395000 }, { "epoch": 0.86, "learning_rate": 2.730523718225299e-06, "loss": 0.6354, "step": 2395500 }, { "epoch": 0.86, "learning_rate": 2.7269191521051207e-06, "loss": 0.6368, "step": 2396000 }, { "epoch": 0.86, "learning_rate": 2.7233145859849423e-06, "loss": 0.622, "step": 2396500 }, { "epoch": 0.86, "learning_rate": 2.7197100198647643e-06, "loss": 0.636, "step": 2397000 }, { "epoch": 0.86, "learning_rate": 2.716105453744586e-06, "loss": 0.6328, "step": 2397500 }, { "epoch": 0.86, "learning_rate": 2.7125008876244075e-06, "loss": 0.5999, "step": 2398000 }, { "epoch": 0.86, "learning_rate": 2.708896321504229e-06, "loss": 0.6488, "step": 2398500 }, { "epoch": 0.86, "learning_rate": 2.7052917553840507e-06, "loss": 0.6366, "step": 2399000 }, { "epoch": 0.86, "learning_rate": 2.701687189263872e-06, "loss": 0.6062, "step": 2399500 }, { "epoch": 0.87, "learning_rate": 2.6980826231436934e-06, "loss": 0.6248, "step": 2400000 }, { "epoch": 0.87, "learning_rate": 2.694478057023515e-06, "loss": 0.628, "step": 2400500 }, { "epoch": 0.87, "learning_rate": 2.6908734909033366e-06, "loss": 0.6398, "step": 2401000 }, { "epoch": 0.87, "learning_rate": 2.6872689247831586e-06, "loss": 0.6292, "step": 2401500 }, { "epoch": 0.87, "learning_rate": 2.6836643586629802e-06, "loss": 0.6452, "step": 2402000 }, { "epoch": 0.87, "learning_rate": 2.680059792542802e-06, "loss": 0.6278, "step": 2402500 }, { "epoch": 0.87, "learning_rate": 2.6764552264226234e-06, "loss": 0.6248, "step": 2403000 }, { "epoch": 0.87, "learning_rate": 2.672850660302445e-06, "loss": 0.6493, "step": 2403500 }, { "epoch": 0.87, "learning_rate": 2.6692460941822666e-06, "loss": 0.6395, "step": 2404000 }, { "epoch": 0.87, "learning_rate": 2.6656415280620886e-06, "loss": 0.6397, "step": 2404500 }, { "epoch": 0.87, "learning_rate": 2.6620369619419093e-06, "loss": 0.6408, "step": 2405000 }, { "epoch": 0.87, "learning_rate": 2.6584323958217314e-06, "loss": 0.645, "step": 2405500 }, { "epoch": 0.87, "learning_rate": 2.654827829701553e-06, "loss": 0.6565, "step": 2406000 }, { "epoch": 0.87, "learning_rate": 2.6512232635813745e-06, "loss": 0.6341, "step": 2406500 }, { "epoch": 0.87, "learning_rate": 2.647618697461196e-06, "loss": 0.6538, "step": 2407000 }, { "epoch": 0.87, "learning_rate": 2.6440141313410177e-06, "loss": 0.6298, "step": 2407500 }, { "epoch": 0.87, "learning_rate": 2.6404095652208393e-06, "loss": 0.6407, "step": 2408000 }, { "epoch": 0.87, "learning_rate": 2.6368049991006613e-06, "loss": 0.6479, "step": 2408500 }, { "epoch": 0.87, "learning_rate": 2.633200432980483e-06, "loss": 0.6397, "step": 2409000 }, { "epoch": 0.87, "learning_rate": 2.6295958668603045e-06, "loss": 0.6158, "step": 2409500 }, { "epoch": 0.87, "learning_rate": 2.6259913007401257e-06, "loss": 0.6454, "step": 2410000 }, { "epoch": 0.87, "learning_rate": 2.6223867346199473e-06, "loss": 0.6627, "step": 2410500 }, { "epoch": 0.87, "learning_rate": 2.618782168499769e-06, "loss": 0.628, "step": 2411000 }, { "epoch": 0.87, "learning_rate": 2.6151776023795904e-06, "loss": 0.6677, "step": 2411500 }, { "epoch": 0.87, "learning_rate": 2.611573036259412e-06, "loss": 0.6268, "step": 2412000 }, { "epoch": 0.87, "learning_rate": 2.6079684701392336e-06, "loss": 0.6255, "step": 2412500 }, { "epoch": 0.87, "learning_rate": 2.6043639040190556e-06, "loss": 0.6244, "step": 2413000 }, { "epoch": 0.87, "learning_rate": 2.6007593378988772e-06, "loss": 0.6377, "step": 2413500 }, { "epoch": 0.87, "learning_rate": 2.597154771778699e-06, "loss": 0.6321, "step": 2414000 }, { "epoch": 0.87, "learning_rate": 2.5935502056585204e-06, "loss": 0.6269, "step": 2414500 }, { "epoch": 0.87, "learning_rate": 2.5899456395383416e-06, "loss": 0.6369, "step": 2415000 }, { "epoch": 0.87, "learning_rate": 2.586341073418163e-06, "loss": 0.6351, "step": 2415500 }, { "epoch": 0.87, "learning_rate": 2.5827365072979847e-06, "loss": 0.6141, "step": 2416000 }, { "epoch": 0.87, "learning_rate": 2.5791319411778063e-06, "loss": 0.6341, "step": 2416500 }, { "epoch": 0.87, "learning_rate": 2.5755273750576284e-06, "loss": 0.6244, "step": 2417000 }, { "epoch": 0.87, "learning_rate": 2.57192280893745e-06, "loss": 0.6453, "step": 2417500 }, { "epoch": 0.87, "learning_rate": 2.5683182428172715e-06, "loss": 0.6443, "step": 2418000 }, { "epoch": 0.87, "learning_rate": 2.564713676697093e-06, "loss": 0.6281, "step": 2418500 }, { "epoch": 0.87, "learning_rate": 2.5611091105769147e-06, "loss": 0.6198, "step": 2419000 }, { "epoch": 0.87, "learning_rate": 2.5575045444567363e-06, "loss": 0.6551, "step": 2419500 }, { "epoch": 0.87, "learning_rate": 2.5538999783365575e-06, "loss": 0.6499, "step": 2420000 }, { "epoch": 0.87, "learning_rate": 2.550295412216379e-06, "loss": 0.6311, "step": 2420500 }, { "epoch": 0.87, "learning_rate": 2.546690846096201e-06, "loss": 0.6108, "step": 2421000 }, { "epoch": 0.87, "learning_rate": 2.5430862799760227e-06, "loss": 0.6406, "step": 2421500 }, { "epoch": 0.87, "learning_rate": 2.5394817138558443e-06, "loss": 0.6167, "step": 2422000 }, { "epoch": 0.87, "learning_rate": 2.535877147735666e-06, "loss": 0.6454, "step": 2422500 }, { "epoch": 0.87, "learning_rate": 2.5322725816154874e-06, "loss": 0.6263, "step": 2423000 }, { "epoch": 0.87, "learning_rate": 2.528668015495309e-06, "loss": 0.6143, "step": 2423500 }, { "epoch": 0.87, "learning_rate": 2.5250634493751306e-06, "loss": 0.6631, "step": 2424000 }, { "epoch": 0.87, "learning_rate": 2.5214588832549526e-06, "loss": 0.6321, "step": 2424500 }, { "epoch": 0.87, "learning_rate": 2.5178543171347742e-06, "loss": 0.6196, "step": 2425000 }, { "epoch": 0.87, "learning_rate": 2.5142497510145954e-06, "loss": 0.6193, "step": 2425500 }, { "epoch": 0.87, "learning_rate": 2.510645184894417e-06, "loss": 0.6435, "step": 2426000 }, { "epoch": 0.87, "learning_rate": 2.5070406187742386e-06, "loss": 0.6295, "step": 2426500 }, { "epoch": 0.87, "learning_rate": 2.50343605265406e-06, "loss": 0.6371, "step": 2427000 }, { "epoch": 0.88, "learning_rate": 2.4998314865338817e-06, "loss": 0.6112, "step": 2427500 }, { "epoch": 0.88, "learning_rate": 2.4962269204137033e-06, "loss": 0.6382, "step": 2428000 }, { "epoch": 0.88, "learning_rate": 2.4926223542935253e-06, "loss": 0.6252, "step": 2428500 }, { "epoch": 0.88, "learning_rate": 2.4890177881733465e-06, "loss": 0.6643, "step": 2429000 }, { "epoch": 0.88, "learning_rate": 2.485413222053168e-06, "loss": 0.6238, "step": 2429500 }, { "epoch": 0.88, "learning_rate": 2.4818086559329897e-06, "loss": 0.6422, "step": 2430000 }, { "epoch": 0.88, "learning_rate": 2.4782040898128117e-06, "loss": 0.6621, "step": 2430500 }, { "epoch": 0.88, "learning_rate": 2.4745995236926333e-06, "loss": 0.6279, "step": 2431000 }, { "epoch": 0.88, "learning_rate": 2.4709949575724545e-06, "loss": 0.643, "step": 2431500 }, { "epoch": 0.88, "learning_rate": 2.467390391452276e-06, "loss": 0.6296, "step": 2432000 }, { "epoch": 0.88, "learning_rate": 2.463785825332098e-06, "loss": 0.6364, "step": 2432500 }, { "epoch": 0.88, "learning_rate": 2.4601812592119197e-06, "loss": 0.6608, "step": 2433000 }, { "epoch": 0.88, "learning_rate": 2.4565766930917413e-06, "loss": 0.6307, "step": 2433500 }, { "epoch": 0.88, "learning_rate": 2.4529721269715624e-06, "loss": 0.6564, "step": 2434000 }, { "epoch": 0.88, "learning_rate": 2.449367560851384e-06, "loss": 0.6386, "step": 2434500 }, { "epoch": 0.88, "learning_rate": 2.445762994731206e-06, "loss": 0.6297, "step": 2435000 }, { "epoch": 0.88, "learning_rate": 2.4421584286110276e-06, "loss": 0.6419, "step": 2435500 }, { "epoch": 0.88, "learning_rate": 2.438553862490849e-06, "loss": 0.6273, "step": 2436000 }, { "epoch": 0.88, "learning_rate": 2.434949296370671e-06, "loss": 0.6443, "step": 2436500 }, { "epoch": 0.88, "learning_rate": 2.4313447302504924e-06, "loss": 0.6292, "step": 2437000 }, { "epoch": 0.88, "learning_rate": 2.427740164130314e-06, "loss": 0.6253, "step": 2437500 }, { "epoch": 0.88, "learning_rate": 2.4241355980101356e-06, "loss": 0.6363, "step": 2438000 }, { "epoch": 0.88, "learning_rate": 2.420531031889957e-06, "loss": 0.6198, "step": 2438500 }, { "epoch": 0.88, "learning_rate": 2.4169264657697787e-06, "loss": 0.6469, "step": 2439000 }, { "epoch": 0.88, "learning_rate": 2.4133218996496003e-06, "loss": 0.6361, "step": 2439500 }, { "epoch": 0.88, "learning_rate": 2.409717333529422e-06, "loss": 0.6225, "step": 2440000 }, { "epoch": 0.88, "learning_rate": 2.4061127674092435e-06, "loss": 0.6378, "step": 2440500 }, { "epoch": 0.88, "learning_rate": 2.402508201289065e-06, "loss": 0.6261, "step": 2441000 }, { "epoch": 0.88, "learning_rate": 2.3989036351688867e-06, "loss": 0.645, "step": 2441500 }, { "epoch": 0.88, "learning_rate": 2.3952990690487083e-06, "loss": 0.6413, "step": 2442000 }, { "epoch": 0.88, "learning_rate": 2.39169450292853e-06, "loss": 0.6277, "step": 2442500 }, { "epoch": 0.88, "learning_rate": 2.3880899368083515e-06, "loss": 0.6541, "step": 2443000 }, { "epoch": 0.88, "learning_rate": 2.384485370688173e-06, "loss": 0.6332, "step": 2443500 }, { "epoch": 0.88, "learning_rate": 2.380880804567995e-06, "loss": 0.6249, "step": 2444000 }, { "epoch": 0.88, "learning_rate": 2.3772762384478162e-06, "loss": 0.6156, "step": 2444500 }, { "epoch": 0.88, "learning_rate": 2.373671672327638e-06, "loss": 0.6279, "step": 2445000 }, { "epoch": 0.88, "learning_rate": 2.3700671062074594e-06, "loss": 0.6341, "step": 2445500 }, { "epoch": 0.88, "learning_rate": 2.366462540087281e-06, "loss": 0.6498, "step": 2446000 }, { "epoch": 0.88, "learning_rate": 2.362857973967103e-06, "loss": 0.6495, "step": 2446500 }, { "epoch": 0.88, "learning_rate": 2.359253407846924e-06, "loss": 0.6315, "step": 2447000 }, { "epoch": 0.88, "learning_rate": 2.3556488417267458e-06, "loss": 0.6309, "step": 2447500 }, { "epoch": 0.88, "learning_rate": 2.3520442756065674e-06, "loss": 0.6402, "step": 2448000 }, { "epoch": 0.88, "learning_rate": 2.3484397094863894e-06, "loss": 0.6428, "step": 2448500 }, { "epoch": 0.88, "learning_rate": 2.344835143366211e-06, "loss": 0.6307, "step": 2449000 }, { "epoch": 0.88, "learning_rate": 2.341230577246032e-06, "loss": 0.6128, "step": 2449500 }, { "epoch": 0.88, "learning_rate": 2.3376260111258537e-06, "loss": 0.6297, "step": 2450000 }, { "epoch": 0.88, "learning_rate": 2.3340214450056757e-06, "loss": 0.637, "step": 2450500 }, { "epoch": 0.88, "learning_rate": 2.3304168788854973e-06, "loss": 0.6383, "step": 2451000 }, { "epoch": 0.88, "learning_rate": 2.326812312765319e-06, "loss": 0.633, "step": 2451500 }, { "epoch": 0.88, "learning_rate": 2.32320774664514e-06, "loss": 0.6449, "step": 2452000 }, { "epoch": 0.88, "learning_rate": 2.319603180524962e-06, "loss": 0.6309, "step": 2452500 }, { "epoch": 0.88, "learning_rate": 2.3159986144047837e-06, "loss": 0.6436, "step": 2453000 }, { "epoch": 0.88, "learning_rate": 2.3123940482846053e-06, "loss": 0.6327, "step": 2453500 }, { "epoch": 0.88, "learning_rate": 2.308789482164427e-06, "loss": 0.627, "step": 2454000 }, { "epoch": 0.88, "learning_rate": 2.3051849160442485e-06, "loss": 0.6534, "step": 2454500 }, { "epoch": 0.88, "learning_rate": 2.30158034992407e-06, "loss": 0.6415, "step": 2455000 }, { "epoch": 0.89, "learning_rate": 2.2979757838038916e-06, "loss": 0.6086, "step": 2455500 }, { "epoch": 0.89, "learning_rate": 2.2943712176837132e-06, "loss": 0.6388, "step": 2456000 }, { "epoch": 0.89, "learning_rate": 2.290766651563535e-06, "loss": 0.6622, "step": 2456500 }, { "epoch": 0.89, "learning_rate": 2.2871620854433564e-06, "loss": 0.6336, "step": 2457000 }, { "epoch": 0.89, "learning_rate": 2.283557519323178e-06, "loss": 0.6062, "step": 2457500 }, { "epoch": 0.89, "learning_rate": 2.2799529532029996e-06, "loss": 0.6309, "step": 2458000 }, { "epoch": 0.89, "learning_rate": 2.276348387082821e-06, "loss": 0.6282, "step": 2458500 }, { "epoch": 0.89, "learning_rate": 2.2727438209626428e-06, "loss": 0.67, "step": 2459000 }, { "epoch": 0.89, "learning_rate": 2.2691392548424644e-06, "loss": 0.6023, "step": 2459500 }, { "epoch": 0.89, "learning_rate": 2.265534688722286e-06, "loss": 0.636, "step": 2460000 }, { "epoch": 0.89, "learning_rate": 2.2619301226021075e-06, "loss": 0.6454, "step": 2460500 }, { "epoch": 0.89, "learning_rate": 2.258325556481929e-06, "loss": 0.6474, "step": 2461000 }, { "epoch": 0.89, "learning_rate": 2.2547209903617507e-06, "loss": 0.6267, "step": 2461500 }, { "epoch": 0.89, "learning_rate": 2.2511164242415727e-06, "loss": 0.6219, "step": 2462000 }, { "epoch": 0.89, "learning_rate": 2.247511858121394e-06, "loss": 0.6183, "step": 2462500 }, { "epoch": 0.89, "learning_rate": 2.2439072920012155e-06, "loss": 0.6448, "step": 2463000 }, { "epoch": 0.89, "learning_rate": 2.240302725881037e-06, "loss": 0.6309, "step": 2463500 }, { "epoch": 0.89, "learning_rate": 2.236698159760859e-06, "loss": 0.6096, "step": 2464000 }, { "epoch": 0.89, "learning_rate": 2.2330935936406807e-06, "loss": 0.6534, "step": 2464500 }, { "epoch": 0.89, "learning_rate": 2.229489027520502e-06, "loss": 0.5911, "step": 2465000 }, { "epoch": 0.89, "learning_rate": 2.2258844614003234e-06, "loss": 0.6313, "step": 2465500 }, { "epoch": 0.89, "learning_rate": 2.2222798952801455e-06, "loss": 0.6455, "step": 2466000 }, { "epoch": 0.89, "learning_rate": 2.218675329159967e-06, "loss": 0.6373, "step": 2466500 }, { "epoch": 0.89, "learning_rate": 2.2150707630397886e-06, "loss": 0.6351, "step": 2467000 }, { "epoch": 0.89, "learning_rate": 2.21146619691961e-06, "loss": 0.6379, "step": 2467500 }, { "epoch": 0.89, "learning_rate": 2.2078616307994314e-06, "loss": 0.6504, "step": 2468000 }, { "epoch": 0.89, "learning_rate": 2.2042570646792534e-06, "loss": 0.6205, "step": 2468500 }, { "epoch": 0.89, "learning_rate": 2.200652498559075e-06, "loss": 0.6146, "step": 2469000 }, { "epoch": 0.89, "learning_rate": 2.1970479324388966e-06, "loss": 0.6518, "step": 2469500 }, { "epoch": 0.89, "learning_rate": 2.1934433663187178e-06, "loss": 0.6372, "step": 2470000 }, { "epoch": 0.89, "learning_rate": 2.1898388001985398e-06, "loss": 0.6392, "step": 2470500 }, { "epoch": 0.89, "learning_rate": 2.1862342340783614e-06, "loss": 0.6211, "step": 2471000 }, { "epoch": 0.89, "learning_rate": 2.182629667958183e-06, "loss": 0.6121, "step": 2471500 }, { "epoch": 0.89, "learning_rate": 2.1790251018380045e-06, "loss": 0.6237, "step": 2472000 }, { "epoch": 0.89, "learning_rate": 2.175420535717826e-06, "loss": 0.6545, "step": 2472500 }, { "epoch": 0.89, "learning_rate": 2.1718159695976477e-06, "loss": 0.612, "step": 2473000 }, { "epoch": 0.89, "learning_rate": 2.1682114034774693e-06, "loss": 0.6187, "step": 2473500 }, { "epoch": 0.89, "learning_rate": 2.164606837357291e-06, "loss": 0.6297, "step": 2474000 }, { "epoch": 0.89, "learning_rate": 2.1610022712371125e-06, "loss": 0.6541, "step": 2474500 }, { "epoch": 0.89, "learning_rate": 2.157397705116934e-06, "loss": 0.6402, "step": 2475000 }, { "epoch": 0.89, "learning_rate": 2.1537931389967557e-06, "loss": 0.6295, "step": 2475500 }, { "epoch": 0.89, "learning_rate": 2.1501885728765773e-06, "loss": 0.6442, "step": 2476000 }, { "epoch": 0.89, "learning_rate": 2.146584006756399e-06, "loss": 0.6555, "step": 2476500 }, { "epoch": 0.89, "learning_rate": 2.1429794406362204e-06, "loss": 0.5993, "step": 2477000 }, { "epoch": 0.89, "learning_rate": 2.1393748745160425e-06, "loss": 0.6478, "step": 2477500 }, { "epoch": 0.89, "learning_rate": 2.1357703083958636e-06, "loss": 0.6478, "step": 2478000 }, { "epoch": 0.89, "learning_rate": 2.1321657422756852e-06, "loss": 0.6551, "step": 2478500 }, { "epoch": 0.89, "learning_rate": 2.128561176155507e-06, "loss": 0.6069, "step": 2479000 }, { "epoch": 0.89, "learning_rate": 2.1249566100353284e-06, "loss": 0.6442, "step": 2479500 }, { "epoch": 0.89, "learning_rate": 2.1213520439151504e-06, "loss": 0.6096, "step": 2480000 }, { "epoch": 0.89, "learning_rate": 2.1177474777949716e-06, "loss": 0.6129, "step": 2480500 }, { "epoch": 0.89, "learning_rate": 2.114142911674793e-06, "loss": 0.6102, "step": 2481000 }, { "epoch": 0.89, "learning_rate": 2.1105383455546148e-06, "loss": 0.6551, "step": 2481500 }, { "epoch": 0.89, "learning_rate": 2.1069337794344368e-06, "loss": 0.6708, "step": 2482000 }, { "epoch": 0.89, "learning_rate": 2.1033292133142584e-06, "loss": 0.6251, "step": 2482500 }, { "epoch": 0.9, "learning_rate": 2.0997246471940795e-06, "loss": 0.6423, "step": 2483000 }, { "epoch": 0.9, "learning_rate": 2.096120081073901e-06, "loss": 0.6155, "step": 2483500 }, { "epoch": 0.9, "learning_rate": 2.092515514953723e-06, "loss": 0.6368, "step": 2484000 }, { "epoch": 0.9, "learning_rate": 2.0889109488335447e-06, "loss": 0.6428, "step": 2484500 }, { "epoch": 0.9, "learning_rate": 2.0853063827133663e-06, "loss": 0.6582, "step": 2485000 }, { "epoch": 0.9, "learning_rate": 2.0817018165931875e-06, "loss": 0.6584, "step": 2485500 }, { "epoch": 0.9, "learning_rate": 2.0780972504730095e-06, "loss": 0.6278, "step": 2486000 }, { "epoch": 0.9, "learning_rate": 2.074492684352831e-06, "loss": 0.604, "step": 2486500 }, { "epoch": 0.9, "learning_rate": 2.0708881182326527e-06, "loss": 0.6341, "step": 2487000 }, { "epoch": 0.9, "learning_rate": 2.0672835521124743e-06, "loss": 0.6139, "step": 2487500 }, { "epoch": 0.9, "learning_rate": 2.063678985992296e-06, "loss": 0.6218, "step": 2488000 }, { "epoch": 0.9, "learning_rate": 2.0600744198721174e-06, "loss": 0.6297, "step": 2488500 }, { "epoch": 0.9, "learning_rate": 2.056469853751939e-06, "loss": 0.6617, "step": 2489000 }, { "epoch": 0.9, "learning_rate": 2.0528652876317606e-06, "loss": 0.6302, "step": 2489500 }, { "epoch": 0.9, "learning_rate": 2.0492607215115822e-06, "loss": 0.6266, "step": 2490000 }, { "epoch": 0.9, "learning_rate": 2.045656155391404e-06, "loss": 0.6489, "step": 2490500 }, { "epoch": 0.9, "learning_rate": 2.0420515892712254e-06, "loss": 0.6021, "step": 2491000 }, { "epoch": 0.9, "learning_rate": 2.038447023151047e-06, "loss": 0.6348, "step": 2491500 }, { "epoch": 0.9, "learning_rate": 2.0348424570308686e-06, "loss": 0.6359, "step": 2492000 }, { "epoch": 0.9, "learning_rate": 2.03123789091069e-06, "loss": 0.6178, "step": 2492500 }, { "epoch": 0.9, "learning_rate": 2.0276333247905118e-06, "loss": 0.6367, "step": 2493000 }, { "epoch": 0.9, "learning_rate": 2.0240287586703333e-06, "loss": 0.6537, "step": 2493500 }, { "epoch": 0.9, "learning_rate": 2.020424192550155e-06, "loss": 0.6234, "step": 2494000 }, { "epoch": 0.9, "learning_rate": 2.0168196264299765e-06, "loss": 0.6186, "step": 2494500 }, { "epoch": 0.9, "learning_rate": 2.013215060309798e-06, "loss": 0.637, "step": 2495000 }, { "epoch": 0.9, "learning_rate": 2.0096104941896197e-06, "loss": 0.6222, "step": 2495500 }, { "epoch": 0.9, "learning_rate": 2.0060059280694413e-06, "loss": 0.656, "step": 2496000 }, { "epoch": 0.9, "learning_rate": 2.002401361949263e-06, "loss": 0.6383, "step": 2496500 }, { "epoch": 0.9, "learning_rate": 1.9987967958290845e-06, "loss": 0.6394, "step": 2497000 }, { "epoch": 0.9, "learning_rate": 1.9951922297089065e-06, "loss": 0.635, "step": 2497500 }, { "epoch": 0.9, "learning_rate": 1.9915876635887277e-06, "loss": 0.6378, "step": 2498000 }, { "epoch": 0.9, "learning_rate": 1.9879830974685492e-06, "loss": 0.6356, "step": 2498500 }, { "epoch": 0.9, "learning_rate": 1.984378531348371e-06, "loss": 0.6409, "step": 2499000 }, { "epoch": 0.9, "learning_rate": 1.980773965228193e-06, "loss": 0.6093, "step": 2499500 }, { "epoch": 0.9, "learning_rate": 1.9771693991080144e-06, "loss": 0.6643, "step": 2500000 }, { "epoch": 0.9, "learning_rate": 1.973564832987836e-06, "loss": 0.6406, "step": 2500500 }, { "epoch": 0.9, "learning_rate": 1.969960266867657e-06, "loss": 0.6427, "step": 2501000 }, { "epoch": 0.9, "learning_rate": 1.9663557007474792e-06, "loss": 0.6309, "step": 2501500 }, { "epoch": 0.9, "learning_rate": 1.962751134627301e-06, "loss": 0.6334, "step": 2502000 }, { "epoch": 0.9, "learning_rate": 1.9591465685071224e-06, "loss": 0.611, "step": 2502500 }, { "epoch": 0.9, "learning_rate": 1.955542002386944e-06, "loss": 0.6521, "step": 2503000 }, { "epoch": 0.9, "learning_rate": 1.951937436266765e-06, "loss": 0.6299, "step": 2503500 }, { "epoch": 0.9, "learning_rate": 1.948332870146587e-06, "loss": 0.6467, "step": 2504000 }, { "epoch": 0.9, "learning_rate": 1.9447283040264088e-06, "loss": 0.6157, "step": 2504500 }, { "epoch": 0.9, "learning_rate": 1.9411237379062303e-06, "loss": 0.6464, "step": 2505000 }, { "epoch": 0.9, "learning_rate": 1.937519171786052e-06, "loss": 0.637, "step": 2505500 }, { "epoch": 0.9, "learning_rate": 1.9339146056658735e-06, "loss": 0.6508, "step": 2506000 }, { "epoch": 0.9, "learning_rate": 1.930310039545695e-06, "loss": 0.62, "step": 2506500 }, { "epoch": 0.9, "learning_rate": 1.9267054734255167e-06, "loss": 0.6359, "step": 2507000 }, { "epoch": 0.9, "learning_rate": 1.9231009073053383e-06, "loss": 0.6334, "step": 2507500 }, { "epoch": 0.9, "learning_rate": 1.91949634118516e-06, "loss": 0.6425, "step": 2508000 }, { "epoch": 0.9, "learning_rate": 1.9158917750649815e-06, "loss": 0.6353, "step": 2508500 }, { "epoch": 0.9, "learning_rate": 1.912287208944803e-06, "loss": 0.64, "step": 2509000 }, { "epoch": 0.9, "learning_rate": 1.9086826428246247e-06, "loss": 0.6411, "step": 2509500 }, { "epoch": 0.9, "learning_rate": 1.9050780767044462e-06, "loss": 0.6229, "step": 2510000 }, { "epoch": 0.9, "learning_rate": 1.901473510584268e-06, "loss": 0.6434, "step": 2510500 }, { "epoch": 0.91, "learning_rate": 1.8978689444640894e-06, "loss": 0.6323, "step": 2511000 }, { "epoch": 0.91, "learning_rate": 1.894264378343911e-06, "loss": 0.6612, "step": 2511500 }, { "epoch": 0.91, "learning_rate": 1.8906598122237326e-06, "loss": 0.6144, "step": 2512000 }, { "epoch": 0.91, "learning_rate": 1.8870552461035544e-06, "loss": 0.6437, "step": 2512500 }, { "epoch": 0.91, "learning_rate": 1.883450679983376e-06, "loss": 0.6641, "step": 2513000 }, { "epoch": 0.91, "learning_rate": 1.8798461138631974e-06, "loss": 0.6409, "step": 2513500 }, { "epoch": 0.91, "learning_rate": 1.876241547743019e-06, "loss": 0.6292, "step": 2514000 }, { "epoch": 0.91, "learning_rate": 1.8726369816228408e-06, "loss": 0.6102, "step": 2514500 }, { "epoch": 0.91, "learning_rate": 1.8690324155026624e-06, "loss": 0.6441, "step": 2515000 }, { "epoch": 0.91, "learning_rate": 1.865427849382484e-06, "loss": 0.6296, "step": 2515500 }, { "epoch": 0.91, "learning_rate": 1.8618232832623053e-06, "loss": 0.6551, "step": 2516000 }, { "epoch": 0.91, "learning_rate": 1.8582187171421271e-06, "loss": 0.6195, "step": 2516500 }, { "epoch": 0.91, "learning_rate": 1.8546141510219487e-06, "loss": 0.6554, "step": 2517000 }, { "epoch": 0.91, "learning_rate": 1.8510095849017703e-06, "loss": 0.616, "step": 2517500 }, { "epoch": 0.91, "learning_rate": 1.8474050187815921e-06, "loss": 0.6273, "step": 2518000 }, { "epoch": 0.91, "learning_rate": 1.8438004526614135e-06, "loss": 0.6292, "step": 2518500 }, { "epoch": 0.91, "learning_rate": 1.840195886541235e-06, "loss": 0.642, "step": 2519000 }, { "epoch": 0.91, "learning_rate": 1.8365913204210567e-06, "loss": 0.6269, "step": 2519500 }, { "epoch": 0.91, "learning_rate": 1.8329867543008783e-06, "loss": 0.6491, "step": 2520000 }, { "epoch": 0.91, "learning_rate": 1.8293821881807e-06, "loss": 0.6481, "step": 2520500 }, { "epoch": 0.91, "learning_rate": 1.8257776220605217e-06, "loss": 0.6166, "step": 2521000 }, { "epoch": 0.91, "learning_rate": 1.822173055940343e-06, "loss": 0.6251, "step": 2521500 }, { "epoch": 0.91, "learning_rate": 1.8185684898201646e-06, "loss": 0.6255, "step": 2522000 }, { "epoch": 0.91, "learning_rate": 1.8149639236999864e-06, "loss": 0.6334, "step": 2522500 }, { "epoch": 0.91, "learning_rate": 1.811359357579808e-06, "loss": 0.6236, "step": 2523000 }, { "epoch": 0.91, "learning_rate": 1.8077547914596296e-06, "loss": 0.6597, "step": 2523500 }, { "epoch": 0.91, "learning_rate": 1.804150225339451e-06, "loss": 0.6328, "step": 2524000 }, { "epoch": 0.91, "learning_rate": 1.8005456592192728e-06, "loss": 0.6392, "step": 2524500 }, { "epoch": 0.91, "learning_rate": 1.7969410930990944e-06, "loss": 0.6387, "step": 2525000 }, { "epoch": 0.91, "learning_rate": 1.793336526978916e-06, "loss": 0.6495, "step": 2525500 }, { "epoch": 0.91, "learning_rate": 1.7897319608587378e-06, "loss": 0.6585, "step": 2526000 }, { "epoch": 0.91, "learning_rate": 1.7861273947385591e-06, "loss": 0.643, "step": 2526500 }, { "epoch": 0.91, "learning_rate": 1.7825228286183807e-06, "loss": 0.6337, "step": 2527000 }, { "epoch": 0.91, "learning_rate": 1.7789182624982023e-06, "loss": 0.6212, "step": 2527500 }, { "epoch": 0.91, "learning_rate": 1.7753136963780241e-06, "loss": 0.6543, "step": 2528000 }, { "epoch": 0.91, "learning_rate": 1.7717091302578457e-06, "loss": 0.644, "step": 2528500 }, { "epoch": 0.91, "learning_rate": 1.768104564137667e-06, "loss": 0.6032, "step": 2529000 }, { "epoch": 0.91, "learning_rate": 1.7644999980174887e-06, "loss": 0.6529, "step": 2529500 }, { "epoch": 0.91, "learning_rate": 1.7608954318973103e-06, "loss": 0.6663, "step": 2530000 }, { "epoch": 0.91, "learning_rate": 1.757290865777132e-06, "loss": 0.6373, "step": 2530500 }, { "epoch": 0.91, "learning_rate": 1.7536862996569537e-06, "loss": 0.6065, "step": 2531000 }, { "epoch": 0.91, "learning_rate": 1.750081733536775e-06, "loss": 0.6262, "step": 2531500 }, { "epoch": 0.91, "learning_rate": 1.7464771674165966e-06, "loss": 0.6095, "step": 2532000 }, { "epoch": 0.91, "learning_rate": 1.7428726012964184e-06, "loss": 0.6212, "step": 2532500 }, { "epoch": 0.91, "learning_rate": 1.73926803517624e-06, "loss": 0.6173, "step": 2533000 }, { "epoch": 0.91, "learning_rate": 1.7356634690560616e-06, "loss": 0.6397, "step": 2533500 }, { "epoch": 0.91, "learning_rate": 1.732058902935883e-06, "loss": 0.6299, "step": 2534000 }, { "epoch": 0.91, "learning_rate": 1.7284543368157048e-06, "loss": 0.6472, "step": 2534500 }, { "epoch": 0.91, "learning_rate": 1.7248497706955264e-06, "loss": 0.6466, "step": 2535000 }, { "epoch": 0.91, "learning_rate": 1.721245204575348e-06, "loss": 0.617, "step": 2535500 }, { "epoch": 0.91, "learning_rate": 1.7176406384551698e-06, "loss": 0.6414, "step": 2536000 }, { "epoch": 0.91, "learning_rate": 1.7140360723349912e-06, "loss": 0.6439, "step": 2536500 }, { "epoch": 0.91, "learning_rate": 1.7104315062148128e-06, "loss": 0.6334, "step": 2537000 }, { "epoch": 0.91, "learning_rate": 1.7068269400946343e-06, "loss": 0.6331, "step": 2537500 }, { "epoch": 0.91, "learning_rate": 1.7032223739744561e-06, "loss": 0.6188, "step": 2538000 }, { "epoch": 0.92, "learning_rate": 1.6996178078542777e-06, "loss": 0.6506, "step": 2538500 }, { "epoch": 0.92, "learning_rate": 1.6960132417340991e-06, "loss": 0.6153, "step": 2539000 }, { "epoch": 0.92, "learning_rate": 1.6924086756139207e-06, "loss": 0.6073, "step": 2539500 }, { "epoch": 0.92, "learning_rate": 1.6888041094937425e-06, "loss": 0.6313, "step": 2540000 }, { "epoch": 0.92, "learning_rate": 1.685199543373564e-06, "loss": 0.6264, "step": 2540500 }, { "epoch": 0.92, "learning_rate": 1.6815949772533857e-06, "loss": 0.6133, "step": 2541000 }, { "epoch": 0.92, "learning_rate": 1.6779904111332073e-06, "loss": 0.6323, "step": 2541500 }, { "epoch": 0.92, "learning_rate": 1.6743858450130287e-06, "loss": 0.6251, "step": 2542000 }, { "epoch": 0.92, "learning_rate": 1.6707812788928505e-06, "loss": 0.6419, "step": 2542500 }, { "epoch": 0.92, "learning_rate": 1.667176712772672e-06, "loss": 0.6467, "step": 2543000 }, { "epoch": 0.92, "learning_rate": 1.6635721466524936e-06, "loss": 0.6408, "step": 2543500 }, { "epoch": 0.92, "learning_rate": 1.6599675805323154e-06, "loss": 0.6445, "step": 2544000 }, { "epoch": 0.92, "learning_rate": 1.6563630144121368e-06, "loss": 0.6271, "step": 2544500 }, { "epoch": 0.92, "learning_rate": 1.6527584482919584e-06, "loss": 0.6007, "step": 2545000 }, { "epoch": 0.92, "learning_rate": 1.64915388217178e-06, "loss": 0.6161, "step": 2545500 }, { "epoch": 0.92, "learning_rate": 1.6455493160516018e-06, "loss": 0.5885, "step": 2546000 }, { "epoch": 0.92, "learning_rate": 1.6419447499314234e-06, "loss": 0.6329, "step": 2546500 }, { "epoch": 0.92, "learning_rate": 1.6383401838112448e-06, "loss": 0.6263, "step": 2547000 }, { "epoch": 0.92, "learning_rate": 1.6347356176910664e-06, "loss": 0.6482, "step": 2547500 }, { "epoch": 0.92, "learning_rate": 1.6311310515708882e-06, "loss": 0.6423, "step": 2548000 }, { "epoch": 0.92, "learning_rate": 1.6275264854507097e-06, "loss": 0.6314, "step": 2548500 }, { "epoch": 0.92, "learning_rate": 1.6239219193305313e-06, "loss": 0.6331, "step": 2549000 }, { "epoch": 0.92, "learning_rate": 1.6203173532103527e-06, "loss": 0.6357, "step": 2549500 }, { "epoch": 0.92, "learning_rate": 1.6167127870901745e-06, "loss": 0.6213, "step": 2550000 }, { "epoch": 0.92, "learning_rate": 1.6131082209699961e-06, "loss": 0.6263, "step": 2550500 }, { "epoch": 0.92, "learning_rate": 1.6095036548498177e-06, "loss": 0.6322, "step": 2551000 }, { "epoch": 0.92, "learning_rate": 1.6058990887296395e-06, "loss": 0.6233, "step": 2551500 }, { "epoch": 0.92, "learning_rate": 1.6022945226094609e-06, "loss": 0.6193, "step": 2552000 }, { "epoch": 0.92, "learning_rate": 1.5986899564892825e-06, "loss": 0.6409, "step": 2552500 }, { "epoch": 0.92, "learning_rate": 1.595085390369104e-06, "loss": 0.6441, "step": 2553000 }, { "epoch": 0.92, "learning_rate": 1.5914808242489257e-06, "loss": 0.608, "step": 2553500 }, { "epoch": 0.92, "learning_rate": 1.5878762581287475e-06, "loss": 0.6098, "step": 2554000 }, { "epoch": 0.92, "learning_rate": 1.5842716920085688e-06, "loss": 0.6132, "step": 2554500 }, { "epoch": 0.92, "learning_rate": 1.5806671258883904e-06, "loss": 0.6419, "step": 2555000 }, { "epoch": 0.92, "learning_rate": 1.577062559768212e-06, "loss": 0.6318, "step": 2555500 }, { "epoch": 0.92, "learning_rate": 1.5734579936480338e-06, "loss": 0.6193, "step": 2556000 }, { "epoch": 0.92, "learning_rate": 1.5698534275278554e-06, "loss": 0.6476, "step": 2556500 }, { "epoch": 0.92, "learning_rate": 1.5662488614076768e-06, "loss": 0.646, "step": 2557000 }, { "epoch": 0.92, "learning_rate": 1.5626442952874984e-06, "loss": 0.6425, "step": 2557500 }, { "epoch": 0.92, "learning_rate": 1.5590397291673202e-06, "loss": 0.6102, "step": 2558000 }, { "epoch": 0.92, "learning_rate": 1.5554351630471418e-06, "loss": 0.615, "step": 2558500 }, { "epoch": 0.92, "learning_rate": 1.5518305969269634e-06, "loss": 0.6293, "step": 2559000 }, { "epoch": 0.92, "learning_rate": 1.5482260308067847e-06, "loss": 0.6315, "step": 2559500 }, { "epoch": 0.92, "learning_rate": 1.5446214646866065e-06, "loss": 0.6453, "step": 2560000 }, { "epoch": 0.92, "learning_rate": 1.5410168985664281e-06, "loss": 0.6197, "step": 2560500 }, { "epoch": 0.92, "learning_rate": 1.5374123324462497e-06, "loss": 0.6317, "step": 2561000 }, { "epoch": 0.92, "learning_rate": 1.5338077663260715e-06, "loss": 0.6236, "step": 2561500 }, { "epoch": 0.92, "learning_rate": 1.5302032002058931e-06, "loss": 0.6625, "step": 2562000 }, { "epoch": 0.92, "learning_rate": 1.5265986340857145e-06, "loss": 0.6234, "step": 2562500 }, { "epoch": 0.92, "learning_rate": 1.522994067965536e-06, "loss": 0.6125, "step": 2563000 }, { "epoch": 0.92, "learning_rate": 1.5193895018453579e-06, "loss": 0.6181, "step": 2563500 }, { "epoch": 0.92, "learning_rate": 1.5157849357251795e-06, "loss": 0.6257, "step": 2564000 }, { "epoch": 0.92, "learning_rate": 1.512180369605001e-06, "loss": 0.6225, "step": 2564500 }, { "epoch": 0.92, "learning_rate": 1.5085758034848224e-06, "loss": 0.6331, "step": 2565000 }, { "epoch": 0.92, "learning_rate": 1.504971237364644e-06, "loss": 0.6311, "step": 2565500 }, { "epoch": 0.92, "learning_rate": 1.5013666712444658e-06, "loss": 0.6456, "step": 2566000 }, { "epoch": 0.93, "learning_rate": 1.4977621051242874e-06, "loss": 0.6074, "step": 2566500 }, { "epoch": 0.93, "learning_rate": 1.494157539004109e-06, "loss": 0.6251, "step": 2567000 }, { "epoch": 0.93, "learning_rate": 1.4905529728839304e-06, "loss": 0.6211, "step": 2567500 }, { "epoch": 0.93, "learning_rate": 1.4869484067637522e-06, "loss": 0.6385, "step": 2568000 }, { "epoch": 0.93, "learning_rate": 1.4833438406435738e-06, "loss": 0.6299, "step": 2568500 }, { "epoch": 0.93, "learning_rate": 1.4797392745233954e-06, "loss": 0.6287, "step": 2569000 }, { "epoch": 0.93, "learning_rate": 1.4761347084032172e-06, "loss": 0.6453, "step": 2569500 }, { "epoch": 0.93, "learning_rate": 1.4725301422830386e-06, "loss": 0.6251, "step": 2570000 }, { "epoch": 0.93, "learning_rate": 1.4689255761628601e-06, "loss": 0.6063, "step": 2570500 }, { "epoch": 0.93, "learning_rate": 1.4653210100426817e-06, "loss": 0.6353, "step": 2571000 }, { "epoch": 0.93, "learning_rate": 1.4617164439225035e-06, "loss": 0.622, "step": 2571500 }, { "epoch": 0.93, "learning_rate": 1.4581118778023251e-06, "loss": 0.6373, "step": 2572000 }, { "epoch": 0.93, "learning_rate": 1.4545073116821465e-06, "loss": 0.6319, "step": 2572500 }, { "epoch": 0.93, "learning_rate": 1.450902745561968e-06, "loss": 0.6359, "step": 2573000 }, { "epoch": 0.93, "learning_rate": 1.4472981794417899e-06, "loss": 0.6428, "step": 2573500 }, { "epoch": 0.93, "learning_rate": 1.4436936133216115e-06, "loss": 0.6511, "step": 2574000 }, { "epoch": 0.93, "learning_rate": 1.440089047201433e-06, "loss": 0.6447, "step": 2574500 }, { "epoch": 0.93, "learning_rate": 1.4364844810812545e-06, "loss": 0.6371, "step": 2575000 }, { "epoch": 0.93, "learning_rate": 1.432879914961076e-06, "loss": 0.6425, "step": 2575500 }, { "epoch": 0.93, "learning_rate": 1.4292753488408978e-06, "loss": 0.6358, "step": 2576000 }, { "epoch": 0.93, "learning_rate": 1.4256707827207194e-06, "loss": 0.6331, "step": 2576500 }, { "epoch": 0.93, "learning_rate": 1.422066216600541e-06, "loss": 0.6459, "step": 2577000 }, { "epoch": 0.93, "learning_rate": 1.4184616504803624e-06, "loss": 0.6364, "step": 2577500 }, { "epoch": 0.93, "learning_rate": 1.4148570843601842e-06, "loss": 0.6155, "step": 2578000 }, { "epoch": 0.93, "learning_rate": 1.4112525182400058e-06, "loss": 0.6292, "step": 2578500 }, { "epoch": 0.93, "learning_rate": 1.4076479521198274e-06, "loss": 0.611, "step": 2579000 }, { "epoch": 0.93, "learning_rate": 1.4040433859996492e-06, "loss": 0.6257, "step": 2579500 }, { "epoch": 0.93, "learning_rate": 1.4004388198794706e-06, "loss": 0.6437, "step": 2580000 }, { "epoch": 0.93, "learning_rate": 1.3968342537592922e-06, "loss": 0.6085, "step": 2580500 }, { "epoch": 0.93, "learning_rate": 1.3932296876391137e-06, "loss": 0.6244, "step": 2581000 }, { "epoch": 0.93, "learning_rate": 1.3896251215189355e-06, "loss": 0.636, "step": 2581500 }, { "epoch": 0.93, "learning_rate": 1.3860205553987571e-06, "loss": 0.6162, "step": 2582000 }, { "epoch": 0.93, "learning_rate": 1.3824159892785785e-06, "loss": 0.6288, "step": 2582500 }, { "epoch": 0.93, "learning_rate": 1.3788114231584001e-06, "loss": 0.6275, "step": 2583000 }, { "epoch": 0.93, "learning_rate": 1.375206857038222e-06, "loss": 0.6302, "step": 2583500 }, { "epoch": 0.93, "learning_rate": 1.3716022909180435e-06, "loss": 0.6206, "step": 2584000 }, { "epoch": 0.93, "learning_rate": 1.367997724797865e-06, "loss": 0.6337, "step": 2584500 }, { "epoch": 0.93, "learning_rate": 1.3643931586776869e-06, "loss": 0.638, "step": 2585000 }, { "epoch": 0.93, "learning_rate": 1.3607885925575083e-06, "loss": 0.6329, "step": 2585500 }, { "epoch": 0.93, "learning_rate": 1.3571840264373299e-06, "loss": 0.627, "step": 2586000 }, { "epoch": 0.93, "learning_rate": 1.3535794603171515e-06, "loss": 0.6184, "step": 2586500 }, { "epoch": 0.93, "learning_rate": 1.349974894196973e-06, "loss": 0.617, "step": 2587000 }, { "epoch": 0.93, "learning_rate": 1.3463703280767948e-06, "loss": 0.6473, "step": 2587500 }, { "epoch": 0.93, "learning_rate": 1.3427657619566162e-06, "loss": 0.6406, "step": 2588000 }, { "epoch": 0.93, "learning_rate": 1.3391611958364378e-06, "loss": 0.6332, "step": 2588500 }, { "epoch": 0.93, "learning_rate": 1.3355566297162594e-06, "loss": 0.6303, "step": 2589000 }, { "epoch": 0.93, "learning_rate": 1.3319520635960812e-06, "loss": 0.6359, "step": 2589500 }, { "epoch": 0.93, "learning_rate": 1.3283474974759028e-06, "loss": 0.6408, "step": 2590000 }, { "epoch": 0.93, "learning_rate": 1.3247429313557242e-06, "loss": 0.6293, "step": 2590500 }, { "epoch": 0.93, "learning_rate": 1.3211383652355458e-06, "loss": 0.6613, "step": 2591000 }, { "epoch": 0.93, "learning_rate": 1.3175337991153676e-06, "loss": 0.649, "step": 2591500 }, { "epoch": 0.93, "learning_rate": 1.3139292329951892e-06, "loss": 0.6416, "step": 2592000 }, { "epoch": 0.93, "learning_rate": 1.3103246668750107e-06, "loss": 0.6648, "step": 2592500 }, { "epoch": 0.93, "learning_rate": 1.3067201007548321e-06, "loss": 0.6288, "step": 2593000 }, { "epoch": 0.93, "learning_rate": 1.303115534634654e-06, "loss": 0.6573, "step": 2593500 }, { "epoch": 0.94, "learning_rate": 1.2995109685144755e-06, "loss": 0.6578, "step": 2594000 }, { "epoch": 0.94, "learning_rate": 1.295906402394297e-06, "loss": 0.6342, "step": 2594500 }, { "epoch": 0.94, "learning_rate": 1.292301836274119e-06, "loss": 0.6337, "step": 2595000 }, { "epoch": 0.94, "learning_rate": 1.2886972701539403e-06, "loss": 0.6484, "step": 2595500 }, { "epoch": 0.94, "learning_rate": 1.2850927040337619e-06, "loss": 0.6333, "step": 2596000 }, { "epoch": 0.94, "learning_rate": 1.2814881379135835e-06, "loss": 0.642, "step": 2596500 }, { "epoch": 0.94, "learning_rate": 1.2778835717934053e-06, "loss": 0.6422, "step": 2597000 }, { "epoch": 0.94, "learning_rate": 1.2742790056732269e-06, "loss": 0.6506, "step": 2597500 }, { "epoch": 0.94, "learning_rate": 1.2706744395530482e-06, "loss": 0.6324, "step": 2598000 }, { "epoch": 0.94, "learning_rate": 1.2670698734328698e-06, "loss": 0.6362, "step": 2598500 }, { "epoch": 0.94, "learning_rate": 1.2634653073126914e-06, "loss": 0.6298, "step": 2599000 }, { "epoch": 0.94, "learning_rate": 1.2598607411925132e-06, "loss": 0.623, "step": 2599500 }, { "epoch": 0.94, "learning_rate": 1.2562561750723348e-06, "loss": 0.662, "step": 2600000 }, { "epoch": 0.94, "learning_rate": 1.2526516089521562e-06, "loss": 0.6441, "step": 2600500 }, { "epoch": 0.94, "learning_rate": 1.249047042831978e-06, "loss": 0.6412, "step": 2601000 }, { "epoch": 0.94, "learning_rate": 1.2454424767117996e-06, "loss": 0.628, "step": 2601500 }, { "epoch": 0.94, "learning_rate": 1.2418379105916212e-06, "loss": 0.6165, "step": 2602000 }, { "epoch": 0.94, "learning_rate": 1.2382333444714428e-06, "loss": 0.6093, "step": 2602500 }, { "epoch": 0.94, "learning_rate": 1.2346287783512644e-06, "loss": 0.6351, "step": 2603000 }, { "epoch": 0.94, "learning_rate": 1.231024212231086e-06, "loss": 0.6189, "step": 2603500 }, { "epoch": 0.94, "learning_rate": 1.2274196461109075e-06, "loss": 0.6428, "step": 2604000 }, { "epoch": 0.94, "learning_rate": 1.2238150799907291e-06, "loss": 0.6106, "step": 2604500 }, { "epoch": 0.94, "learning_rate": 1.2202105138705507e-06, "loss": 0.6022, "step": 2605000 }, { "epoch": 0.94, "learning_rate": 1.2166059477503723e-06, "loss": 0.6538, "step": 2605500 }, { "epoch": 0.94, "learning_rate": 1.213001381630194e-06, "loss": 0.5985, "step": 2606000 }, { "epoch": 0.94, "learning_rate": 1.2093968155100155e-06, "loss": 0.6159, "step": 2606500 }, { "epoch": 0.94, "learning_rate": 1.2057922493898373e-06, "loss": 0.6474, "step": 2607000 }, { "epoch": 0.94, "learning_rate": 1.2021876832696587e-06, "loss": 0.6247, "step": 2607500 }, { "epoch": 0.94, "learning_rate": 1.1985831171494805e-06, "loss": 0.5978, "step": 2608000 }, { "epoch": 0.94, "learning_rate": 1.194978551029302e-06, "loss": 0.6081, "step": 2608500 }, { "epoch": 0.94, "learning_rate": 1.1913739849091234e-06, "loss": 0.6207, "step": 2609000 }, { "epoch": 0.94, "learning_rate": 1.1877694187889452e-06, "loss": 0.6411, "step": 2609500 }, { "epoch": 0.94, "learning_rate": 1.1841648526687668e-06, "loss": 0.6175, "step": 2610000 }, { "epoch": 0.94, "learning_rate": 1.1805602865485884e-06, "loss": 0.6007, "step": 2610500 }, { "epoch": 0.94, "learning_rate": 1.17695572042841e-06, "loss": 0.6344, "step": 2611000 }, { "epoch": 0.94, "learning_rate": 1.1733511543082316e-06, "loss": 0.625, "step": 2611500 }, { "epoch": 0.94, "learning_rate": 1.1697465881880532e-06, "loss": 0.5862, "step": 2612000 }, { "epoch": 0.94, "learning_rate": 1.1661420220678748e-06, "loss": 0.6295, "step": 2612500 }, { "epoch": 0.94, "learning_rate": 1.1625374559476964e-06, "loss": 0.6317, "step": 2613000 }, { "epoch": 0.94, "learning_rate": 1.158932889827518e-06, "loss": 0.6141, "step": 2613500 }, { "epoch": 0.94, "learning_rate": 1.1553283237073395e-06, "loss": 0.6239, "step": 2614000 }, { "epoch": 0.94, "learning_rate": 1.1517237575871611e-06, "loss": 0.6098, "step": 2614500 }, { "epoch": 0.94, "learning_rate": 1.148119191466983e-06, "loss": 0.6421, "step": 2615000 }, { "epoch": 0.94, "learning_rate": 1.1445146253468043e-06, "loss": 0.6287, "step": 2615500 }, { "epoch": 0.94, "learning_rate": 1.1409100592266261e-06, "loss": 0.6232, "step": 2616000 }, { "epoch": 0.94, "learning_rate": 1.1373054931064475e-06, "loss": 0.6072, "step": 2616500 }, { "epoch": 0.94, "learning_rate": 1.1337009269862693e-06, "loss": 0.6381, "step": 2617000 }, { "epoch": 0.94, "learning_rate": 1.1300963608660909e-06, "loss": 0.6277, "step": 2617500 }, { "epoch": 0.94, "learning_rate": 1.1264917947459125e-06, "loss": 0.6186, "step": 2618000 }, { "epoch": 0.94, "learning_rate": 1.122887228625734e-06, "loss": 0.6425, "step": 2618500 }, { "epoch": 0.94, "learning_rate": 1.1192826625055557e-06, "loss": 0.642, "step": 2619000 }, { "epoch": 0.94, "learning_rate": 1.1156780963853773e-06, "loss": 0.6532, "step": 2619500 }, { "epoch": 0.94, "learning_rate": 1.1120735302651988e-06, "loss": 0.606, "step": 2620000 }, { "epoch": 0.94, "learning_rate": 1.1084689641450204e-06, "loss": 0.607, "step": 2620500 }, { "epoch": 0.94, "learning_rate": 1.104864398024842e-06, "loss": 0.5978, "step": 2621000 }, { "epoch": 0.94, "learning_rate": 1.1012598319046636e-06, "loss": 0.6326, "step": 2621500 }, { "epoch": 0.95, "learning_rate": 1.0976552657844852e-06, "loss": 0.639, "step": 2622000 }, { "epoch": 0.95, "learning_rate": 1.0940506996643068e-06, "loss": 0.6183, "step": 2622500 }, { "epoch": 0.95, "learning_rate": 1.0904461335441284e-06, "loss": 0.6357, "step": 2623000 }, { "epoch": 0.95, "learning_rate": 1.08684156742395e-06, "loss": 0.6154, "step": 2623500 }, { "epoch": 0.95, "learning_rate": 1.0832370013037718e-06, "loss": 0.6616, "step": 2624000 }, { "epoch": 0.95, "learning_rate": 1.0796324351835932e-06, "loss": 0.6305, "step": 2624500 }, { "epoch": 0.95, "learning_rate": 1.076027869063415e-06, "loss": 0.6241, "step": 2625000 }, { "epoch": 0.95, "learning_rate": 1.0724233029432363e-06, "loss": 0.6301, "step": 2625500 }, { "epoch": 0.95, "learning_rate": 1.0688187368230581e-06, "loss": 0.6298, "step": 2626000 }, { "epoch": 0.95, "learning_rate": 1.0652141707028797e-06, "loss": 0.6246, "step": 2626500 }, { "epoch": 0.95, "learning_rate": 1.0616096045827013e-06, "loss": 0.6404, "step": 2627000 }, { "epoch": 0.95, "learning_rate": 1.058005038462523e-06, "loss": 0.6255, "step": 2627500 }, { "epoch": 0.95, "learning_rate": 1.0544004723423445e-06, "loss": 0.6506, "step": 2628000 }, { "epoch": 0.95, "learning_rate": 1.050795906222166e-06, "loss": 0.6127, "step": 2628500 }, { "epoch": 0.95, "learning_rate": 1.0471913401019877e-06, "loss": 0.6213, "step": 2629000 }, { "epoch": 0.95, "learning_rate": 1.0435867739818093e-06, "loss": 0.6155, "step": 2629500 }, { "epoch": 0.95, "learning_rate": 1.0399822078616309e-06, "loss": 0.6288, "step": 2630000 }, { "epoch": 0.95, "learning_rate": 1.0363776417414527e-06, "loss": 0.6245, "step": 2630500 }, { "epoch": 0.95, "learning_rate": 1.032773075621274e-06, "loss": 0.6307, "step": 2631000 }, { "epoch": 0.95, "learning_rate": 1.0291685095010956e-06, "loss": 0.6228, "step": 2631500 }, { "epoch": 0.95, "learning_rate": 1.0255639433809172e-06, "loss": 0.6311, "step": 2632000 }, { "epoch": 0.95, "learning_rate": 1.0219593772607388e-06, "loss": 0.6241, "step": 2632500 }, { "epoch": 0.95, "learning_rate": 1.0183548111405606e-06, "loss": 0.6059, "step": 2633000 }, { "epoch": 0.95, "learning_rate": 1.014750245020382e-06, "loss": 0.6415, "step": 2633500 }, { "epoch": 0.95, "learning_rate": 1.0111456789002038e-06, "loss": 0.6386, "step": 2634000 }, { "epoch": 0.95, "learning_rate": 1.0075411127800252e-06, "loss": 0.6108, "step": 2634500 }, { "epoch": 0.95, "learning_rate": 1.003936546659847e-06, "loss": 0.6131, "step": 2635000 }, { "epoch": 0.95, "learning_rate": 1.0003319805396686e-06, "loss": 0.6091, "step": 2635500 }, { "epoch": 0.95, "learning_rate": 9.967274144194902e-07, "loss": 0.6202, "step": 2636000 }, { "epoch": 0.95, "learning_rate": 9.931228482993117e-07, "loss": 0.6345, "step": 2636500 }, { "epoch": 0.95, "learning_rate": 9.895182821791333e-07, "loss": 0.638, "step": 2637000 }, { "epoch": 0.95, "learning_rate": 9.85913716058955e-07, "loss": 0.6056, "step": 2637500 }, { "epoch": 0.95, "learning_rate": 9.823091499387765e-07, "loss": 0.6264, "step": 2638000 }, { "epoch": 0.95, "learning_rate": 9.78704583818598e-07, "loss": 0.6236, "step": 2638500 }, { "epoch": 0.95, "learning_rate": 9.751000176984197e-07, "loss": 0.6204, "step": 2639000 }, { "epoch": 0.95, "learning_rate": 9.714954515782413e-07, "loss": 0.6293, "step": 2639500 }, { "epoch": 0.95, "learning_rate": 9.678908854580629e-07, "loss": 0.6139, "step": 2640000 }, { "epoch": 0.95, "learning_rate": 9.642863193378847e-07, "loss": 0.6019, "step": 2640500 }, { "epoch": 0.95, "learning_rate": 9.60681753217706e-07, "loss": 0.6273, "step": 2641000 }, { "epoch": 0.95, "learning_rate": 9.570771870975279e-07, "loss": 0.6311, "step": 2641500 }, { "epoch": 0.95, "learning_rate": 9.534726209773494e-07, "loss": 0.6064, "step": 2642000 }, { "epoch": 0.95, "learning_rate": 9.498680548571709e-07, "loss": 0.623, "step": 2642500 }, { "epoch": 0.95, "learning_rate": 9.462634887369926e-07, "loss": 0.5998, "step": 2643000 }, { "epoch": 0.95, "learning_rate": 9.426589226168141e-07, "loss": 0.6187, "step": 2643500 }, { "epoch": 0.95, "learning_rate": 9.390543564966358e-07, "loss": 0.6134, "step": 2644000 }, { "epoch": 0.95, "learning_rate": 9.354497903764574e-07, "loss": 0.6414, "step": 2644500 }, { "epoch": 0.95, "learning_rate": 9.31845224256279e-07, "loss": 0.6305, "step": 2645000 }, { "epoch": 0.95, "learning_rate": 9.282406581361006e-07, "loss": 0.6211, "step": 2645500 }, { "epoch": 0.95, "learning_rate": 9.246360920159221e-07, "loss": 0.6419, "step": 2646000 }, { "epoch": 0.95, "learning_rate": 9.210315258957438e-07, "loss": 0.6124, "step": 2646500 }, { "epoch": 0.95, "learning_rate": 9.174269597755655e-07, "loss": 0.652, "step": 2647000 }, { "epoch": 0.95, "learning_rate": 9.138223936553869e-07, "loss": 0.6533, "step": 2647500 }, { "epoch": 0.95, "learning_rate": 9.102178275352086e-07, "loss": 0.6191, "step": 2648000 }, { "epoch": 0.95, "learning_rate": 9.066132614150301e-07, "loss": 0.6088, "step": 2648500 }, { "epoch": 0.95, "learning_rate": 9.030086952948518e-07, "loss": 0.6387, "step": 2649000 }, { "epoch": 0.96, "learning_rate": 8.994041291746734e-07, "loss": 0.6247, "step": 2649500 }, { "epoch": 0.96, "learning_rate": 8.95799563054495e-07, "loss": 0.6273, "step": 2650000 }, { "epoch": 0.96, "learning_rate": 8.921949969343166e-07, "loss": 0.5963, "step": 2650500 }, { "epoch": 0.96, "learning_rate": 8.885904308141382e-07, "loss": 0.6257, "step": 2651000 }, { "epoch": 0.96, "learning_rate": 8.849858646939598e-07, "loss": 0.6063, "step": 2651500 }, { "epoch": 0.96, "learning_rate": 8.813812985737815e-07, "loss": 0.6082, "step": 2652000 }, { "epoch": 0.96, "learning_rate": 8.777767324536029e-07, "loss": 0.6071, "step": 2652500 }, { "epoch": 0.96, "learning_rate": 8.741721663334246e-07, "loss": 0.6314, "step": 2653000 }, { "epoch": 0.96, "learning_rate": 8.705676002132462e-07, "loss": 0.6071, "step": 2653500 }, { "epoch": 0.96, "learning_rate": 8.669630340930678e-07, "loss": 0.6294, "step": 2654000 }, { "epoch": 0.96, "learning_rate": 8.633584679728894e-07, "loss": 0.6309, "step": 2654500 }, { "epoch": 0.96, "learning_rate": 8.59753901852711e-07, "loss": 0.6202, "step": 2655000 }, { "epoch": 0.96, "learning_rate": 8.561493357325326e-07, "loss": 0.6338, "step": 2655500 }, { "epoch": 0.96, "learning_rate": 8.525447696123543e-07, "loss": 0.6549, "step": 2656000 }, { "epoch": 0.96, "learning_rate": 8.489402034921758e-07, "loss": 0.6321, "step": 2656500 }, { "epoch": 0.96, "learning_rate": 8.453356373719975e-07, "loss": 0.6351, "step": 2657000 }, { "epoch": 0.96, "learning_rate": 8.41731071251819e-07, "loss": 0.6195, "step": 2657500 }, { "epoch": 0.96, "learning_rate": 8.381265051316406e-07, "loss": 0.6364, "step": 2658000 }, { "epoch": 0.96, "learning_rate": 8.345219390114622e-07, "loss": 0.6118, "step": 2658500 }, { "epoch": 0.96, "learning_rate": 8.309173728912838e-07, "loss": 0.6647, "step": 2659000 }, { "epoch": 0.96, "learning_rate": 8.273128067711054e-07, "loss": 0.6222, "step": 2659500 }, { "epoch": 0.96, "learning_rate": 8.23708240650927e-07, "loss": 0.6145, "step": 2660000 }, { "epoch": 0.96, "learning_rate": 8.201036745307486e-07, "loss": 0.6278, "step": 2660500 }, { "epoch": 0.96, "learning_rate": 8.164991084105703e-07, "loss": 0.6206, "step": 2661000 }, { "epoch": 0.96, "learning_rate": 8.128945422903918e-07, "loss": 0.5964, "step": 2661500 }, { "epoch": 0.96, "learning_rate": 8.092899761702135e-07, "loss": 0.6518, "step": 2662000 }, { "epoch": 0.96, "learning_rate": 8.056854100500351e-07, "loss": 0.6317, "step": 2662500 }, { "epoch": 0.96, "learning_rate": 8.020808439298567e-07, "loss": 0.6427, "step": 2663000 }, { "epoch": 0.96, "learning_rate": 7.984762778096782e-07, "loss": 0.6515, "step": 2663500 }, { "epoch": 0.96, "learning_rate": 7.948717116894998e-07, "loss": 0.6284, "step": 2664000 }, { "epoch": 0.96, "learning_rate": 7.912671455693214e-07, "loss": 0.6424, "step": 2664500 }, { "epoch": 0.96, "learning_rate": 7.876625794491431e-07, "loss": 0.6409, "step": 2665000 }, { "epoch": 0.96, "learning_rate": 7.840580133289646e-07, "loss": 0.6036, "step": 2665500 }, { "epoch": 0.96, "learning_rate": 7.804534472087863e-07, "loss": 0.6264, "step": 2666000 }, { "epoch": 0.96, "learning_rate": 7.768488810886078e-07, "loss": 0.6279, "step": 2666500 }, { "epoch": 0.96, "learning_rate": 7.732443149684295e-07, "loss": 0.6243, "step": 2667000 }, { "epoch": 0.96, "learning_rate": 7.696397488482512e-07, "loss": 0.6061, "step": 2667500 }, { "epoch": 0.96, "learning_rate": 7.660351827280727e-07, "loss": 0.6046, "step": 2668000 }, { "epoch": 0.96, "learning_rate": 7.624306166078943e-07, "loss": 0.6468, "step": 2668500 }, { "epoch": 0.96, "learning_rate": 7.588260504877158e-07, "loss": 0.6338, "step": 2669000 }, { "epoch": 0.96, "learning_rate": 7.552214843675374e-07, "loss": 0.614, "step": 2669500 }, { "epoch": 0.96, "learning_rate": 7.516169182473591e-07, "loss": 0.6319, "step": 2670000 }, { "epoch": 0.96, "learning_rate": 7.480123521271806e-07, "loss": 0.6345, "step": 2670500 }, { "epoch": 0.96, "learning_rate": 7.444077860070023e-07, "loss": 0.629, "step": 2671000 }, { "epoch": 0.96, "learning_rate": 7.408032198868238e-07, "loss": 0.6429, "step": 2671500 }, { "epoch": 0.96, "learning_rate": 7.371986537666455e-07, "loss": 0.642, "step": 2672000 }, { "epoch": 0.96, "learning_rate": 7.335940876464672e-07, "loss": 0.6605, "step": 2672500 }, { "epoch": 0.96, "learning_rate": 7.299895215262887e-07, "loss": 0.6378, "step": 2673000 }, { "epoch": 0.96, "learning_rate": 7.263849554061103e-07, "loss": 0.5827, "step": 2673500 }, { "epoch": 0.96, "learning_rate": 7.22780389285932e-07, "loss": 0.6251, "step": 2674000 }, { "epoch": 0.96, "learning_rate": 7.191758231657534e-07, "loss": 0.6575, "step": 2674500 }, { "epoch": 0.96, "learning_rate": 7.155712570455751e-07, "loss": 0.6288, "step": 2675000 }, { "epoch": 0.96, "learning_rate": 7.119666909253966e-07, "loss": 0.6175, "step": 2675500 }, { "epoch": 0.96, "learning_rate": 7.083621248052183e-07, "loss": 0.5944, "step": 2676000 }, { "epoch": 0.96, "learning_rate": 7.0475755868504e-07, "loss": 0.6074, "step": 2676500 }, { "epoch": 0.96, "learning_rate": 7.011529925648615e-07, "loss": 0.6162, "step": 2677000 }, { "epoch": 0.97, "learning_rate": 6.975484264446832e-07, "loss": 0.6481, "step": 2677500 }, { "epoch": 0.97, "learning_rate": 6.939438603245047e-07, "loss": 0.6423, "step": 2678000 }, { "epoch": 0.97, "learning_rate": 6.903392942043264e-07, "loss": 0.6301, "step": 2678500 }, { "epoch": 0.97, "learning_rate": 6.86734728084148e-07, "loss": 0.6177, "step": 2679000 }, { "epoch": 0.97, "learning_rate": 6.831301619639695e-07, "loss": 0.6357, "step": 2679500 }, { "epoch": 0.97, "learning_rate": 6.795255958437911e-07, "loss": 0.6275, "step": 2680000 }, { "epoch": 0.97, "learning_rate": 6.759210297236126e-07, "loss": 0.6433, "step": 2680500 }, { "epoch": 0.97, "learning_rate": 6.723164636034343e-07, "loss": 0.6223, "step": 2681000 }, { "epoch": 0.97, "learning_rate": 6.68711897483256e-07, "loss": 0.6265, "step": 2681500 }, { "epoch": 0.97, "learning_rate": 6.651073313630775e-07, "loss": 0.639, "step": 2682000 }, { "epoch": 0.97, "learning_rate": 6.615027652428992e-07, "loss": 0.6394, "step": 2682500 }, { "epoch": 0.97, "learning_rate": 6.578981991227207e-07, "loss": 0.6289, "step": 2683000 }, { "epoch": 0.97, "learning_rate": 6.542936330025424e-07, "loss": 0.6006, "step": 2683500 }, { "epoch": 0.97, "learning_rate": 6.50689066882364e-07, "loss": 0.6449, "step": 2684000 }, { "epoch": 0.97, "learning_rate": 6.470845007621856e-07, "loss": 0.6153, "step": 2684500 }, { "epoch": 0.97, "learning_rate": 6.434799346420072e-07, "loss": 0.6214, "step": 2685000 }, { "epoch": 0.97, "learning_rate": 6.398753685218289e-07, "loss": 0.6307, "step": 2685500 }, { "epoch": 0.97, "learning_rate": 6.362708024016503e-07, "loss": 0.6228, "step": 2686000 }, { "epoch": 0.97, "learning_rate": 6.32666236281472e-07, "loss": 0.6344, "step": 2686500 }, { "epoch": 0.97, "learning_rate": 6.290616701612935e-07, "loss": 0.6479, "step": 2687000 }, { "epoch": 0.97, "learning_rate": 6.254571040411152e-07, "loss": 0.6418, "step": 2687500 }, { "epoch": 0.97, "learning_rate": 6.218525379209368e-07, "loss": 0.5979, "step": 2688000 }, { "epoch": 0.97, "learning_rate": 6.182479718007584e-07, "loss": 0.6222, "step": 2688500 }, { "epoch": 0.97, "learning_rate": 6.1464340568058e-07, "loss": 0.605, "step": 2689000 }, { "epoch": 0.97, "learning_rate": 6.110388395604016e-07, "loss": 0.618, "step": 2689500 }, { "epoch": 0.97, "learning_rate": 6.074342734402232e-07, "loss": 0.634, "step": 2690000 }, { "epoch": 0.97, "learning_rate": 6.038297073200448e-07, "loss": 0.6165, "step": 2690500 }, { "epoch": 0.97, "learning_rate": 6.002251411998664e-07, "loss": 0.6324, "step": 2691000 }, { "epoch": 0.97, "learning_rate": 5.96620575079688e-07, "loss": 0.6287, "step": 2691500 }, { "epoch": 0.97, "learning_rate": 5.930160089595096e-07, "loss": 0.6167, "step": 2692000 }, { "epoch": 0.97, "learning_rate": 5.894114428393312e-07, "loss": 0.6475, "step": 2692500 }, { "epoch": 0.97, "learning_rate": 5.858068767191528e-07, "loss": 0.6299, "step": 2693000 }, { "epoch": 0.97, "learning_rate": 5.822023105989744e-07, "loss": 0.6006, "step": 2693500 }, { "epoch": 0.97, "learning_rate": 5.78597744478796e-07, "loss": 0.6241, "step": 2694000 }, { "epoch": 0.97, "learning_rate": 5.749931783586176e-07, "loss": 0.6317, "step": 2694500 }, { "epoch": 0.97, "learning_rate": 5.713886122384392e-07, "loss": 0.6291, "step": 2695000 }, { "epoch": 0.97, "learning_rate": 5.677840461182608e-07, "loss": 0.6057, "step": 2695500 }, { "epoch": 0.97, "learning_rate": 5.641794799980825e-07, "loss": 0.6167, "step": 2696000 }, { "epoch": 0.97, "learning_rate": 5.60574913877904e-07, "loss": 0.6222, "step": 2696500 }, { "epoch": 0.97, "learning_rate": 5.569703477577256e-07, "loss": 0.6314, "step": 2697000 }, { "epoch": 0.97, "learning_rate": 5.533657816375472e-07, "loss": 0.6066, "step": 2697500 }, { "epoch": 0.97, "learning_rate": 5.497612155173688e-07, "loss": 0.6519, "step": 2698000 }, { "epoch": 0.97, "learning_rate": 5.461566493971904e-07, "loss": 0.6192, "step": 2698500 }, { "epoch": 0.97, "learning_rate": 5.42552083277012e-07, "loss": 0.622, "step": 2699000 }, { "epoch": 0.97, "learning_rate": 5.389475171568336e-07, "loss": 0.6234, "step": 2699500 }, { "epoch": 0.97, "learning_rate": 5.353429510366552e-07, "loss": 0.6234, "step": 2700000 }, { "epoch": 0.97, "learning_rate": 5.317383849164769e-07, "loss": 0.6542, "step": 2700500 }, { "epoch": 0.97, "learning_rate": 5.281338187962985e-07, "loss": 0.6254, "step": 2701000 }, { "epoch": 0.97, "learning_rate": 5.245292526761201e-07, "loss": 0.6396, "step": 2701500 }, { "epoch": 0.97, "learning_rate": 5.209246865559416e-07, "loss": 0.6289, "step": 2702000 }, { "epoch": 0.97, "learning_rate": 5.173201204357632e-07, "loss": 0.5921, "step": 2702500 }, { "epoch": 0.97, "learning_rate": 5.137155543155848e-07, "loss": 0.6269, "step": 2703000 }, { "epoch": 0.97, "learning_rate": 5.101109881954064e-07, "loss": 0.6025, "step": 2703500 }, { "epoch": 0.97, "learning_rate": 5.06506422075228e-07, "loss": 0.6259, "step": 2704000 }, { "epoch": 0.97, "learning_rate": 5.029018559550496e-07, "loss": 0.627, "step": 2704500 }, { "epoch": 0.98, "learning_rate": 4.992972898348713e-07, "loss": 0.6124, "step": 2705000 }, { "epoch": 0.98, "learning_rate": 4.956927237146929e-07, "loss": 0.6378, "step": 2705500 }, { "epoch": 0.98, "learning_rate": 4.920881575945145e-07, "loss": 0.6179, "step": 2706000 }, { "epoch": 0.98, "learning_rate": 4.884835914743361e-07, "loss": 0.633, "step": 2706500 }, { "epoch": 0.98, "learning_rate": 4.848790253541578e-07, "loss": 0.6615, "step": 2707000 }, { "epoch": 0.98, "learning_rate": 4.812744592339792e-07, "loss": 0.6343, "step": 2707500 }, { "epoch": 0.98, "learning_rate": 4.776698931138008e-07, "loss": 0.6073, "step": 2708000 }, { "epoch": 0.98, "learning_rate": 4.740653269936225e-07, "loss": 0.6216, "step": 2708500 }, { "epoch": 0.98, "learning_rate": 4.7046076087344407e-07, "loss": 0.6271, "step": 2709000 }, { "epoch": 0.98, "learning_rate": 4.668561947532657e-07, "loss": 0.6006, "step": 2709500 }, { "epoch": 0.98, "learning_rate": 4.632516286330873e-07, "loss": 0.6311, "step": 2710000 }, { "epoch": 0.98, "learning_rate": 4.596470625129089e-07, "loss": 0.6163, "step": 2710500 }, { "epoch": 0.98, "learning_rate": 4.560424963927305e-07, "loss": 0.5998, "step": 2711000 }, { "epoch": 0.98, "learning_rate": 4.5243793027255207e-07, "loss": 0.6639, "step": 2711500 }, { "epoch": 0.98, "learning_rate": 4.488333641523737e-07, "loss": 0.6355, "step": 2712000 }, { "epoch": 0.98, "learning_rate": 4.452287980321953e-07, "loss": 0.6103, "step": 2712500 }, { "epoch": 0.98, "learning_rate": 4.416242319120169e-07, "loss": 0.6172, "step": 2713000 }, { "epoch": 0.98, "learning_rate": 4.380196657918385e-07, "loss": 0.6279, "step": 2713500 }, { "epoch": 0.98, "learning_rate": 4.3441509967166013e-07, "loss": 0.6271, "step": 2714000 }, { "epoch": 0.98, "learning_rate": 4.308105335514817e-07, "loss": 0.6039, "step": 2714500 }, { "epoch": 0.98, "learning_rate": 4.272059674313033e-07, "loss": 0.6285, "step": 2715000 }, { "epoch": 0.98, "learning_rate": 4.236014013111249e-07, "loss": 0.6079, "step": 2715500 }, { "epoch": 0.98, "learning_rate": 4.199968351909465e-07, "loss": 0.6158, "step": 2716000 }, { "epoch": 0.98, "learning_rate": 4.1639226907076813e-07, "loss": 0.6244, "step": 2716500 }, { "epoch": 0.98, "learning_rate": 4.127877029505897e-07, "loss": 0.6267, "step": 2717000 }, { "epoch": 0.98, "learning_rate": 4.091831368304113e-07, "loss": 0.621, "step": 2717500 }, { "epoch": 0.98, "learning_rate": 4.055785707102329e-07, "loss": 0.6176, "step": 2718000 }, { "epoch": 0.98, "learning_rate": 4.0197400459005454e-07, "loss": 0.6594, "step": 2718500 }, { "epoch": 0.98, "learning_rate": 3.9836943846987613e-07, "loss": 0.6203, "step": 2719000 }, { "epoch": 0.98, "learning_rate": 3.947648723496977e-07, "loss": 0.6121, "step": 2719500 }, { "epoch": 0.98, "learning_rate": 3.911603062295193e-07, "loss": 0.6095, "step": 2720000 }, { "epoch": 0.98, "learning_rate": 3.875557401093409e-07, "loss": 0.6237, "step": 2720500 }, { "epoch": 0.98, "learning_rate": 3.8395117398916255e-07, "loss": 0.6223, "step": 2721000 }, { "epoch": 0.98, "learning_rate": 3.8034660786898414e-07, "loss": 0.6476, "step": 2721500 }, { "epoch": 0.98, "learning_rate": 3.7674204174880573e-07, "loss": 0.6131, "step": 2722000 }, { "epoch": 0.98, "learning_rate": 3.731374756286273e-07, "loss": 0.6359, "step": 2722500 }, { "epoch": 0.98, "learning_rate": 3.6953290950844896e-07, "loss": 0.6304, "step": 2723000 }, { "epoch": 0.98, "learning_rate": 3.6592834338827055e-07, "loss": 0.633, "step": 2723500 }, { "epoch": 0.98, "learning_rate": 3.6232377726809214e-07, "loss": 0.6172, "step": 2724000 }, { "epoch": 0.98, "learning_rate": 3.5871921114791373e-07, "loss": 0.6361, "step": 2724500 }, { "epoch": 0.98, "learning_rate": 3.551146450277353e-07, "loss": 0.583, "step": 2725000 }, { "epoch": 0.98, "learning_rate": 3.51510078907557e-07, "loss": 0.6281, "step": 2725500 }, { "epoch": 0.98, "learning_rate": 3.4790551278737856e-07, "loss": 0.6237, "step": 2726000 }, { "epoch": 0.98, "learning_rate": 3.4430094666720015e-07, "loss": 0.618, "step": 2726500 }, { "epoch": 0.98, "learning_rate": 3.4069638054702174e-07, "loss": 0.6164, "step": 2727000 }, { "epoch": 0.98, "learning_rate": 3.3709181442684333e-07, "loss": 0.6142, "step": 2727500 }, { "epoch": 0.98, "learning_rate": 3.33487248306665e-07, "loss": 0.586, "step": 2728000 }, { "epoch": 0.98, "learning_rate": 3.2988268218648656e-07, "loss": 0.6221, "step": 2728500 }, { "epoch": 0.98, "learning_rate": 3.2627811606630815e-07, "loss": 0.6464, "step": 2729000 }, { "epoch": 0.98, "learning_rate": 3.2267354994612974e-07, "loss": 0.6005, "step": 2729500 }, { "epoch": 0.98, "learning_rate": 3.1906898382595144e-07, "loss": 0.6451, "step": 2730000 }, { "epoch": 0.98, "learning_rate": 3.1546441770577303e-07, "loss": 0.6741, "step": 2730500 }, { "epoch": 0.98, "learning_rate": 3.118598515855946e-07, "loss": 0.6465, "step": 2731000 }, { "epoch": 0.98, "learning_rate": 3.0825528546541615e-07, "loss": 0.6428, "step": 2731500 }, { "epoch": 0.98, "learning_rate": 3.046507193452378e-07, "loss": 0.6392, "step": 2732000 }, { "epoch": 0.98, "learning_rate": 3.010461532250594e-07, "loss": 0.6214, "step": 2732500 }, { "epoch": 0.99, "learning_rate": 2.9744158710488103e-07, "loss": 0.5954, "step": 2733000 }, { "epoch": 0.99, "learning_rate": 2.938370209847026e-07, "loss": 0.6323, "step": 2733500 }, { "epoch": 0.99, "learning_rate": 2.902324548645242e-07, "loss": 0.6453, "step": 2734000 }, { "epoch": 0.99, "learning_rate": 2.866278887443458e-07, "loss": 0.6225, "step": 2734500 }, { "epoch": 0.99, "learning_rate": 2.8302332262416744e-07, "loss": 0.6251, "step": 2735000 }, { "epoch": 0.99, "learning_rate": 2.7941875650398903e-07, "loss": 0.6182, "step": 2735500 }, { "epoch": 0.99, "learning_rate": 2.758141903838106e-07, "loss": 0.6272, "step": 2736000 }, { "epoch": 0.99, "learning_rate": 2.722096242636322e-07, "loss": 0.6287, "step": 2736500 }, { "epoch": 0.99, "learning_rate": 2.686050581434538e-07, "loss": 0.5899, "step": 2737000 }, { "epoch": 0.99, "learning_rate": 2.6500049202327545e-07, "loss": 0.6059, "step": 2737500 }, { "epoch": 0.99, "learning_rate": 2.6139592590309704e-07, "loss": 0.6521, "step": 2738000 }, { "epoch": 0.99, "learning_rate": 2.5779135978291863e-07, "loss": 0.6329, "step": 2738500 }, { "epoch": 0.99, "learning_rate": 2.541867936627402e-07, "loss": 0.6462, "step": 2739000 }, { "epoch": 0.99, "learning_rate": 2.505822275425618e-07, "loss": 0.6225, "step": 2739500 }, { "epoch": 0.99, "learning_rate": 2.4697766142238345e-07, "loss": 0.6384, "step": 2740000 }, { "epoch": 0.99, "learning_rate": 2.4337309530220504e-07, "loss": 0.6489, "step": 2740500 }, { "epoch": 0.99, "learning_rate": 2.3976852918202663e-07, "loss": 0.6411, "step": 2741000 }, { "epoch": 0.99, "learning_rate": 2.3616396306184825e-07, "loss": 0.5927, "step": 2741500 }, { "epoch": 0.99, "learning_rate": 2.3255939694166987e-07, "loss": 0.6076, "step": 2742000 }, { "epoch": 0.99, "learning_rate": 2.2895483082149146e-07, "loss": 0.6309, "step": 2742500 }, { "epoch": 0.99, "learning_rate": 2.2535026470131307e-07, "loss": 0.6434, "step": 2743000 }, { "epoch": 0.99, "learning_rate": 2.2174569858113466e-07, "loss": 0.6309, "step": 2743500 }, { "epoch": 0.99, "learning_rate": 2.1814113246095625e-07, "loss": 0.6771, "step": 2744000 }, { "epoch": 0.99, "learning_rate": 2.1453656634077787e-07, "loss": 0.5981, "step": 2744500 }, { "epoch": 0.99, "learning_rate": 2.1093200022059946e-07, "loss": 0.6218, "step": 2745000 }, { "epoch": 0.99, "learning_rate": 2.0732743410042108e-07, "loss": 0.5953, "step": 2745500 }, { "epoch": 0.99, "learning_rate": 2.0372286798024267e-07, "loss": 0.6247, "step": 2746000 }, { "epoch": 0.99, "learning_rate": 2.0011830186006428e-07, "loss": 0.6318, "step": 2746500 }, { "epoch": 0.99, "learning_rate": 1.9651373573988587e-07, "loss": 0.6299, "step": 2747000 }, { "epoch": 0.99, "learning_rate": 1.9290916961970746e-07, "loss": 0.6241, "step": 2747500 }, { "epoch": 0.99, "learning_rate": 1.8930460349952908e-07, "loss": 0.6364, "step": 2748000 }, { "epoch": 0.99, "learning_rate": 1.8570003737935067e-07, "loss": 0.6129, "step": 2748500 }, { "epoch": 0.99, "learning_rate": 1.820954712591723e-07, "loss": 0.6096, "step": 2749000 }, { "epoch": 0.99, "learning_rate": 1.7849090513899388e-07, "loss": 0.6359, "step": 2749500 }, { "epoch": 0.99, "learning_rate": 1.748863390188155e-07, "loss": 0.619, "step": 2750000 }, { "epoch": 0.99, "learning_rate": 1.7128177289863708e-07, "loss": 0.6136, "step": 2750500 }, { "epoch": 0.99, "learning_rate": 1.676772067784587e-07, "loss": 0.6433, "step": 2751000 }, { "epoch": 0.99, "learning_rate": 1.640726406582803e-07, "loss": 0.6207, "step": 2751500 }, { "epoch": 0.99, "learning_rate": 1.6046807453810188e-07, "loss": 0.6292, "step": 2752000 }, { "epoch": 0.99, "learning_rate": 1.568635084179235e-07, "loss": 0.6357, "step": 2752500 }, { "epoch": 0.99, "learning_rate": 1.5325894229774511e-07, "loss": 0.6051, "step": 2753000 }, { "epoch": 0.99, "learning_rate": 1.496543761775667e-07, "loss": 0.6271, "step": 2753500 }, { "epoch": 0.99, "learning_rate": 1.460498100573883e-07, "loss": 0.621, "step": 2754000 }, { "epoch": 0.99, "learning_rate": 1.424452439372099e-07, "loss": 0.6309, "step": 2754500 }, { "epoch": 0.99, "learning_rate": 1.388406778170315e-07, "loss": 0.6445, "step": 2755000 }, { "epoch": 0.99, "learning_rate": 1.3523611169685312e-07, "loss": 0.644, "step": 2755500 }, { "epoch": 0.99, "learning_rate": 1.316315455766747e-07, "loss": 0.63, "step": 2756000 }, { "epoch": 0.99, "learning_rate": 1.2802697945649633e-07, "loss": 0.6481, "step": 2756500 }, { "epoch": 0.99, "learning_rate": 1.2442241333631794e-07, "loss": 0.6153, "step": 2757000 }, { "epoch": 0.99, "learning_rate": 1.2081784721613953e-07, "loss": 0.6387, "step": 2757500 }, { "epoch": 0.99, "learning_rate": 1.1721328109596112e-07, "loss": 0.6148, "step": 2758000 }, { "epoch": 0.99, "learning_rate": 1.1360871497578273e-07, "loss": 0.6323, "step": 2758500 }, { "epoch": 0.99, "learning_rate": 1.1000414885560433e-07, "loss": 0.6249, "step": 2759000 }, { "epoch": 0.99, "learning_rate": 1.0639958273542593e-07, "loss": 0.6195, "step": 2759500 }, { "epoch": 0.99, "learning_rate": 1.0279501661524754e-07, "loss": 0.6182, "step": 2760000 }, { "epoch": 1.0, "learning_rate": 9.919045049506914e-08, "loss": 0.617, "step": 2760500 }, { "epoch": 1.0, "learning_rate": 9.558588437489076e-08, "loss": 0.6281, "step": 2761000 }, { "epoch": 1.0, "learning_rate": 9.198131825471236e-08, "loss": 0.645, "step": 2761500 }, { "epoch": 1.0, "learning_rate": 8.837675213453394e-08, "loss": 0.6343, "step": 2762000 }, { "epoch": 1.0, "learning_rate": 8.477218601435555e-08, "loss": 0.6331, "step": 2762500 }, { "epoch": 1.0, "learning_rate": 8.116761989417716e-08, "loss": 0.6121, "step": 2763000 }, { "epoch": 1.0, "learning_rate": 7.756305377399876e-08, "loss": 0.6346, "step": 2763500 }, { "epoch": 1.0, "learning_rate": 7.395848765382036e-08, "loss": 0.6312, "step": 2764000 }, { "epoch": 1.0, "learning_rate": 7.035392153364197e-08, "loss": 0.5966, "step": 2764500 }, { "epoch": 1.0, "learning_rate": 6.674935541346356e-08, "loss": 0.6191, "step": 2765000 }, { "epoch": 1.0, "learning_rate": 6.314478929328516e-08, "loss": 0.5907, "step": 2765500 }, { "epoch": 1.0, "learning_rate": 5.954022317310677e-08, "loss": 0.6171, "step": 2766000 }, { "epoch": 1.0, "learning_rate": 5.5935657052928374e-08, "loss": 0.6346, "step": 2766500 }, { "epoch": 1.0, "learning_rate": 5.233109093274997e-08, "loss": 0.6252, "step": 2767000 }, { "epoch": 1.0, "learning_rate": 4.8726524812571575e-08, "loss": 0.6151, "step": 2767500 }, { "epoch": 1.0, "learning_rate": 4.5121958692393185e-08, "loss": 0.6131, "step": 2768000 }, { "epoch": 1.0, "learning_rate": 4.151739257221479e-08, "loss": 0.6405, "step": 2768500 }, { "epoch": 1.0, "learning_rate": 3.7912826452036385e-08, "loss": 0.6414, "step": 2769000 }, { "epoch": 1.0, "learning_rate": 3.430826033185799e-08, "loss": 0.6079, "step": 2769500 }, { "epoch": 1.0, "learning_rate": 3.070369421167959e-08, "loss": 0.6183, "step": 2770000 }, { "epoch": 1.0, "learning_rate": 2.7099128091501192e-08, "loss": 0.6306, "step": 2770500 }, { "epoch": 1.0, "learning_rate": 2.3494561971322795e-08, "loss": 0.6224, "step": 2771000 }, { "epoch": 1.0, "learning_rate": 1.9889995851144395e-08, "loss": 0.6215, "step": 2771500 }, { "epoch": 1.0, "learning_rate": 1.6285429730966e-08, "loss": 0.6208, "step": 2772000 }, { "epoch": 1.0, "learning_rate": 1.2680863610787602e-08, "loss": 0.6264, "step": 2772500 }, { "epoch": 1.0, "learning_rate": 9.076297490609206e-09, "loss": 0.6288, "step": 2773000 }, { "epoch": 1.0, "learning_rate": 5.471731370430808e-09, "loss": 0.6265, "step": 2773500 }, { "epoch": 1.0, "learning_rate": 1.8671652502524096e-09, "loss": 0.636, "step": 2774000 } ], "max_steps": 2774259, "num_train_epochs": 1, "total_flos": 1.8366895242215424e+17, "trial_name": null, "trial_params": null }