{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 1.38264777048047,
  "eval_steps": 7000.0,
  "global_step": 7000,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.0,
      "learning_rate": 1.975503753457132e-07,
      "loss": 7.9807,
      "step": 10
    },
    {
      "epoch": 0.0,
      "learning_rate": 3.951007506914264e-07,
      "loss": 7.3025,
      "step": 20
    },
    {
      "epoch": 0.01,
      "learning_rate": 5.926511260371394e-07,
      "loss": 6.442,
      "step": 30
    },
    {
      "epoch": 0.01,
      "learning_rate": 7.902015013828528e-07,
      "loss": 4.9928,
      "step": 40
    },
    {
      "epoch": 0.01,
      "learning_rate": 9.877518767285659e-07,
      "loss": 3.0368,
      "step": 50
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.1853022520742789e-06,
      "loss": 1.5885,
      "step": 60
    },
    {
      "epoch": 0.01,
      "learning_rate": 1.382852627419992e-06,
      "loss": 1.118,
      "step": 70
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.5804030027657055e-06,
      "loss": 1.0564,
      "step": 80
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.7779533781114185e-06,
      "loss": 1.1117,
      "step": 90
    },
    {
      "epoch": 0.02,
      "learning_rate": 1.9755037534571318e-06,
      "loss": 1.1398,
      "step": 100
    },
    {
      "epoch": 0.02,
      "learning_rate": 2.173054128802845e-06,
      "loss": 0.9834,
      "step": 110
    },
    {
      "epoch": 0.02,
      "learning_rate": 2.3706045041485578e-06,
      "loss": 1.0677,
      "step": 120
    },
    {
      "epoch": 0.03,
      "learning_rate": 2.5681548794942714e-06,
      "loss": 0.9412,
      "step": 130
    },
    {
      "epoch": 0.03,
      "learning_rate": 2.765705254839984e-06,
      "loss": 0.9347,
      "step": 140
    },
    {
      "epoch": 0.03,
      "learning_rate": 2.9632556301856974e-06,
      "loss": 0.9574,
      "step": 150
    },
    {
      "epoch": 0.03,
      "learning_rate": 3.160806005531411e-06,
      "loss": 0.9889,
      "step": 160
    },
    {
      "epoch": 0.03,
      "learning_rate": 3.3583563808771234e-06,
      "loss": 0.9102,
      "step": 170
    },
    {
      "epoch": 0.04,
      "learning_rate": 3.555906756222837e-06,
      "loss": 0.9191,
      "step": 180
    },
    {
      "epoch": 0.04,
      "learning_rate": 3.7534571315685503e-06,
      "loss": 0.9766,
      "step": 190
    },
    {
      "epoch": 0.04,
      "learning_rate": 3.9510075069142635e-06,
      "loss": 0.9135,
      "step": 200
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.148557882259976e-06,
      "loss": 0.8757,
      "step": 210
    },
    {
      "epoch": 0.04,
      "learning_rate": 4.34610825760569e-06,
      "loss": 0.9067,
      "step": 220
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.543658632951403e-06,
      "loss": 0.9219,
      "step": 230
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.7412090082971155e-06,
      "loss": 0.8996,
      "step": 240
    },
    {
      "epoch": 0.05,
      "learning_rate": 4.938759383642829e-06,
      "loss": 0.9467,
      "step": 250
    },
    {
      "epoch": 0.05,
      "learning_rate": 5.136309758988543e-06,
      "loss": 0.8877,
      "step": 260
    },
    {
      "epoch": 0.05,
      "learning_rate": 5.333860134334256e-06,
      "loss": 0.8398,
      "step": 270
    },
    {
      "epoch": 0.06,
      "learning_rate": 5.531410509679968e-06,
      "loss": 0.8764,
      "step": 280
    },
    {
      "epoch": 0.06,
      "learning_rate": 5.728960885025682e-06,
      "loss": 0.8809,
      "step": 290
    },
    {
      "epoch": 0.06,
      "learning_rate": 5.926511260371395e-06,
      "loss": 0.8699,
      "step": 300
    },
    {
      "epoch": 0.06,
      "learning_rate": 6.124061635717108e-06,
      "loss": 0.8377,
      "step": 310
    },
    {
      "epoch": 0.06,
      "learning_rate": 6.321612011062822e-06,
      "loss": 0.8943,
      "step": 320
    },
    {
      "epoch": 0.07,
      "learning_rate": 6.519162386408534e-06,
      "loss": 0.8393,
      "step": 330
    },
    {
      "epoch": 0.07,
      "learning_rate": 6.716712761754247e-06,
      "loss": 0.8192,
      "step": 340
    },
    {
      "epoch": 0.07,
      "learning_rate": 6.9142631370999605e-06,
      "loss": 0.903,
      "step": 350
    },
    {
      "epoch": 0.07,
      "learning_rate": 7.111813512445674e-06,
      "loss": 0.8917,
      "step": 360
    },
    {
      "epoch": 0.07,
      "learning_rate": 7.309363887791387e-06,
      "loss": 0.916,
      "step": 370
    },
    {
      "epoch": 0.08,
      "learning_rate": 7.506914263137101e-06,
      "loss": 0.864,
      "step": 380
    },
    {
      "epoch": 0.08,
      "learning_rate": 7.704464638482814e-06,
      "loss": 0.7809,
      "step": 390
    },
    {
      "epoch": 0.08,
      "learning_rate": 7.902015013828527e-06,
      "loss": 0.885,
      "step": 400
    },
    {
      "epoch": 0.08,
      "learning_rate": 8.09956538917424e-06,
      "loss": 0.8183,
      "step": 410
    },
    {
      "epoch": 0.08,
      "learning_rate": 8.297115764519953e-06,
      "loss": 0.7897,
      "step": 420
    },
    {
      "epoch": 0.08,
      "learning_rate": 8.494666139865665e-06,
      "loss": 0.8315,
      "step": 430
    },
    {
      "epoch": 0.09,
      "learning_rate": 8.69221651521138e-06,
      "loss": 0.8266,
      "step": 440
    },
    {
      "epoch": 0.09,
      "learning_rate": 8.889766890557093e-06,
      "loss": 0.8755,
      "step": 450
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.087317265902805e-06,
      "loss": 0.8103,
      "step": 460
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.28486764124852e-06,
      "loss": 0.8226,
      "step": 470
    },
    {
      "epoch": 0.09,
      "learning_rate": 9.482418016594231e-06,
      "loss": 0.8174,
      "step": 480
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.679968391939946e-06,
      "loss": 0.8336,
      "step": 490
    },
    {
      "epoch": 0.1,
      "learning_rate": 9.877518767285658e-06,
      "loss": 0.8214,
      "step": 500
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.0075069142631371e-05,
      "loss": 0.8534,
      "step": 510
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.0272619517977086e-05,
      "loss": 0.8148,
      "step": 520
    },
    {
      "epoch": 0.1,
      "learning_rate": 1.0470169893322798e-05,
      "loss": 0.8032,
      "step": 530
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.0667720268668511e-05,
      "loss": 0.7561,
      "step": 540
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.0865270644014224e-05,
      "loss": 0.8219,
      "step": 550
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.1062821019359937e-05,
      "loss": 0.8108,
      "step": 560
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.126037139470565e-05,
      "loss": 0.7673,
      "step": 570
    },
    {
      "epoch": 0.11,
      "learning_rate": 1.1457921770051364e-05,
      "loss": 0.7731,
      "step": 580
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.1655472145397077e-05,
      "loss": 0.8135,
      "step": 590
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.185302252074279e-05,
      "loss": 0.8525,
      "step": 600
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.2050572896088504e-05,
      "loss": 0.8473,
      "step": 610
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.2248123271434215e-05,
      "loss": 0.7937,
      "step": 620
    },
    {
      "epoch": 0.12,
      "learning_rate": 1.244567364677993e-05,
      "loss": 0.7924,
      "step": 630
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.2643224022125644e-05,
      "loss": 0.7778,
      "step": 640
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.2840774397471355e-05,
      "loss": 0.8282,
      "step": 650
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.3038324772817068e-05,
      "loss": 0.811,
      "step": 660
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.3235875148162783e-05,
      "loss": 0.7167,
      "step": 670
    },
    {
      "epoch": 0.13,
      "learning_rate": 1.3433425523508494e-05,
      "loss": 0.8321,
      "step": 680
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.363097589885421e-05,
      "loss": 0.7477,
      "step": 690
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.3828526274199921e-05,
      "loss": 0.8232,
      "step": 700
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.4026076649545635e-05,
      "loss": 0.799,
      "step": 710
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.4223627024891348e-05,
      "loss": 0.8554,
      "step": 720
    },
    {
      "epoch": 0.14,
      "learning_rate": 1.442117740023706e-05,
      "loss": 0.778,
      "step": 730
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.4618727775582774e-05,
      "loss": 0.8337,
      "step": 740
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.4816278150928487e-05,
      "loss": 0.7793,
      "step": 750
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.5013828526274201e-05,
      "loss": 0.7812,
      "step": 760
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.5211378901619914e-05,
      "loss": 0.7837,
      "step": 770
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.540892927696563e-05,
      "loss": 0.7714,
      "step": 780
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.560647965231134e-05,
      "loss": 0.74,
      "step": 790
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.5804030027657054e-05,
      "loss": 0.7725,
      "step": 800
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.6001580403002767e-05,
      "loss": 0.7618,
      "step": 810
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.619913077834848e-05,
      "loss": 0.767,
      "step": 820
    },
    {
      "epoch": 0.16,
      "learning_rate": 1.6396681153694192e-05,
      "loss": 0.8257,
      "step": 830
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.6594231529039905e-05,
      "loss": 0.8116,
      "step": 840
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.679178190438562e-05,
      "loss": 0.7649,
      "step": 850
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.698933227973133e-05,
      "loss": 0.7616,
      "step": 860
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.7186882655077047e-05,
      "loss": 0.7219,
      "step": 870
    },
    {
      "epoch": 0.17,
      "learning_rate": 1.738443303042276e-05,
      "loss": 0.7532,
      "step": 880
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.758198340576847e-05,
      "loss": 0.7122,
      "step": 890
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.7779533781114185e-05,
      "loss": 0.8242,
      "step": 900
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.7977084156459898e-05,
      "loss": 0.7554,
      "step": 910
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.817463453180561e-05,
      "loss": 0.7588,
      "step": 920
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.8372184907151324e-05,
      "loss": 0.7404,
      "step": 930
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.856973528249704e-05,
      "loss": 0.7907,
      "step": 940
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.876728565784275e-05,
      "loss": 0.7775,
      "step": 950
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.8964836033188462e-05,
      "loss": 0.7544,
      "step": 960
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.9162386408534178e-05,
      "loss": 0.8053,
      "step": 970
    },
    {
      "epoch": 0.19,
      "learning_rate": 1.935993678387989e-05,
      "loss": 0.7226,
      "step": 980
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9557487159225604e-05,
      "loss": 0.7482,
      "step": 990
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.9755037534571317e-05,
      "loss": 0.7171,
      "step": 1000
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.995258790991703e-05,
      "loss": 0.7421,
      "step": 1010
    },
    {
      "epoch": 0.2,
      "learning_rate": 2.0150138285262742e-05,
      "loss": 0.7356,
      "step": 1020
    },
    {
      "epoch": 0.2,
      "learning_rate": 2.0347688660608455e-05,
      "loss": 0.713,
      "step": 1030
    },
    {
      "epoch": 0.21,
      "learning_rate": 2.054523903595417e-05,
      "loss": 0.7705,
      "step": 1040
    },
    {
      "epoch": 0.21,
      "learning_rate": 2.074278941129988e-05,
      "loss": 0.7501,
      "step": 1050
    },
    {
      "epoch": 0.21,
      "learning_rate": 2.0940339786645597e-05,
      "loss": 0.7572,
      "step": 1060
    },
    {
      "epoch": 0.21,
      "learning_rate": 2.113789016199131e-05,
      "loss": 0.7382,
      "step": 1070
    },
    {
      "epoch": 0.21,
      "learning_rate": 2.1335440537337022e-05,
      "loss": 0.7596,
      "step": 1080
    },
    {
      "epoch": 0.22,
      "learning_rate": 2.1532990912682735e-05,
      "loss": 0.7063,
      "step": 1090
    },
    {
      "epoch": 0.22,
      "learning_rate": 2.1730541288028448e-05,
      "loss": 0.7239,
      "step": 1100
    },
    {
      "epoch": 0.22,
      "learning_rate": 2.192809166337416e-05,
      "loss": 0.7648,
      "step": 1110
    },
    {
      "epoch": 0.22,
      "learning_rate": 2.2125642038719874e-05,
      "loss": 0.7539,
      "step": 1120
    },
    {
      "epoch": 0.22,
      "learning_rate": 2.232319241406559e-05,
      "loss": 0.7498,
      "step": 1130
    },
    {
      "epoch": 0.23,
      "learning_rate": 2.25207427894113e-05,
      "loss": 0.7647,
      "step": 1140
    },
    {
      "epoch": 0.23,
      "learning_rate": 2.2718293164757015e-05,
      "loss": 0.7657,
      "step": 1150
    },
    {
      "epoch": 0.23,
      "learning_rate": 2.2915843540102728e-05,
      "loss": 0.7451,
      "step": 1160
    },
    {
      "epoch": 0.23,
      "learning_rate": 2.3113393915448438e-05,
      "loss": 0.7528,
      "step": 1170
    },
    {
      "epoch": 0.23,
      "learning_rate": 2.3310944290794154e-05,
      "loss": 0.7548,
      "step": 1180
    },
    {
      "epoch": 0.24,
      "learning_rate": 2.3508494666139867e-05,
      "loss": 0.7616,
      "step": 1190
    },
    {
      "epoch": 0.24,
      "learning_rate": 2.370604504148558e-05,
      "loss": 0.7654,
      "step": 1200
    },
    {
      "epoch": 0.24,
      "learning_rate": 2.3903595416831292e-05,
      "loss": 0.7681,
      "step": 1210
    },
    {
      "epoch": 0.24,
      "learning_rate": 2.410114579217701e-05,
      "loss": 0.7302,
      "step": 1220
    },
    {
      "epoch": 0.24,
      "learning_rate": 2.4298696167522718e-05,
      "loss": 0.7516,
      "step": 1230
    },
    {
      "epoch": 0.24,
      "learning_rate": 2.449624654286843e-05,
      "loss": 0.7331,
      "step": 1240
    },
    {
      "epoch": 0.25,
      "learning_rate": 2.4693796918214147e-05,
      "loss": 0.7709,
      "step": 1250
    },
    {
      "epoch": 0.25,
      "learning_rate": 2.489134729355986e-05,
      "loss": 0.696,
      "step": 1260
    },
    {
      "epoch": 0.25,
      "learning_rate": 2.508889766890557e-05,
      "loss": 0.7537,
      "step": 1270
    },
    {
      "epoch": 0.25,
      "learning_rate": 2.528644804425129e-05,
      "loss": 0.7789,
      "step": 1280
    },
    {
      "epoch": 0.25,
      "learning_rate": 2.5483998419596998e-05,
      "loss": 0.7277,
      "step": 1290
    },
    {
      "epoch": 0.26,
      "learning_rate": 2.568154879494271e-05,
      "loss": 0.6968,
      "step": 1300
    },
    {
      "epoch": 0.26,
      "learning_rate": 2.5879099170288423e-05,
      "loss": 0.7329,
      "step": 1310
    },
    {
      "epoch": 0.26,
      "learning_rate": 2.6076649545634136e-05,
      "loss": 0.7501,
      "step": 1320
    },
    {
      "epoch": 0.26,
      "learning_rate": 2.6274199920979852e-05,
      "loss": 0.7224,
      "step": 1330
    },
    {
      "epoch": 0.26,
      "learning_rate": 2.6471750296325565e-05,
      "loss": 0.7276,
      "step": 1340
    },
    {
      "epoch": 0.27,
      "learning_rate": 2.6669300671671278e-05,
      "loss": 0.6827,
      "step": 1350
    },
    {
      "epoch": 0.27,
      "learning_rate": 2.6866851047016987e-05,
      "loss": 0.7762,
      "step": 1360
    },
    {
      "epoch": 0.27,
      "learning_rate": 2.7064401422362707e-05,
      "loss": 0.7124,
      "step": 1370
    },
    {
      "epoch": 0.27,
      "learning_rate": 2.726195179770842e-05,
      "loss": 0.6812,
      "step": 1380
    },
    {
      "epoch": 0.27,
      "learning_rate": 2.745950217305413e-05,
      "loss": 0.6922,
      "step": 1390
    },
    {
      "epoch": 0.28,
      "learning_rate": 2.7657052548399842e-05,
      "loss": 0.7216,
      "step": 1400
    },
    {
      "epoch": 0.28,
      "learning_rate": 2.7854602923745555e-05,
      "loss": 0.6899,
      "step": 1410
    },
    {
      "epoch": 0.28,
      "learning_rate": 2.805215329909127e-05,
      "loss": 0.7433,
      "step": 1420
    },
    {
      "epoch": 0.28,
      "learning_rate": 2.8249703674436984e-05,
      "loss": 0.7103,
      "step": 1430
    },
    {
      "epoch": 0.28,
      "learning_rate": 2.8447254049782697e-05,
      "loss": 0.7348,
      "step": 1440
    },
    {
      "epoch": 0.29,
      "learning_rate": 2.8644804425128406e-05,
      "loss": 0.8088,
      "step": 1450
    },
    {
      "epoch": 0.29,
      "learning_rate": 2.884235480047412e-05,
      "loss": 0.7432,
      "step": 1460
    },
    {
      "epoch": 0.29,
      "learning_rate": 2.903990517581984e-05,
      "loss": 0.7381,
      "step": 1470
    },
    {
      "epoch": 0.29,
      "learning_rate": 2.9237455551165548e-05,
      "loss": 0.7195,
      "step": 1480
    },
    {
      "epoch": 0.29,
      "learning_rate": 2.943500592651126e-05,
      "loss": 0.7038,
      "step": 1490
    },
    {
      "epoch": 0.3,
      "learning_rate": 2.9632556301856973e-05,
      "loss": 0.6643,
      "step": 1500
    },
    {
      "epoch": 0.3,
      "learning_rate": 2.983010667720269e-05,
      "loss": 0.7159,
      "step": 1510
    },
    {
      "epoch": 0.3,
      "learning_rate": 3.0027657052548402e-05,
      "loss": 0.6977,
      "step": 1520
    },
    {
      "epoch": 0.3,
      "learning_rate": 3.0225207427894115e-05,
      "loss": 0.713,
      "step": 1530
    },
    {
      "epoch": 0.3,
      "learning_rate": 3.0422757803239828e-05,
      "loss": 0.6713,
      "step": 1540
    },
    {
      "epoch": 0.31,
      "learning_rate": 3.062030817858554e-05,
      "loss": 0.6948,
      "step": 1550
    },
    {
      "epoch": 0.31,
      "learning_rate": 3.081785855393126e-05,
      "loss": 0.6964,
      "step": 1560
    },
    {
      "epoch": 0.31,
      "learning_rate": 3.101540892927697e-05,
      "loss": 0.7344,
      "step": 1570
    },
    {
      "epoch": 0.31,
      "learning_rate": 3.121295930462268e-05,
      "loss": 0.6921,
      "step": 1580
    },
    {
      "epoch": 0.31,
      "learning_rate": 3.141050967996839e-05,
      "loss": 0.7013,
      "step": 1590
    },
    {
      "epoch": 0.32,
      "learning_rate": 3.160806005531411e-05,
      "loss": 0.6525,
      "step": 1600
    },
    {
      "epoch": 0.32,
      "learning_rate": 3.180561043065982e-05,
      "loss": 0.6412,
      "step": 1610
    },
    {
      "epoch": 0.32,
      "learning_rate": 3.2003160806005534e-05,
      "loss": 0.6697,
      "step": 1620
    },
    {
      "epoch": 0.32,
      "learning_rate": 3.2200711181351246e-05,
      "loss": 0.7215,
      "step": 1630
    },
    {
      "epoch": 0.32,
      "learning_rate": 3.239826155669696e-05,
      "loss": 0.6918,
      "step": 1640
    },
    {
      "epoch": 0.33,
      "learning_rate": 3.259581193204267e-05,
      "loss": 0.7667,
      "step": 1650
    },
    {
      "epoch": 0.33,
      "learning_rate": 3.2793362307388385e-05,
      "loss": 0.69,
      "step": 1660
    },
    {
      "epoch": 0.33,
      "learning_rate": 3.29909126827341e-05,
      "loss": 0.6899,
      "step": 1670
    },
    {
      "epoch": 0.33,
      "learning_rate": 3.318846305807981e-05,
      "loss": 0.663,
      "step": 1680
    },
    {
      "epoch": 0.33,
      "learning_rate": 3.338601343342552e-05,
      "loss": 0.7416,
      "step": 1690
    },
    {
      "epoch": 0.34,
      "learning_rate": 3.358356380877124e-05,
      "loss": 0.689,
      "step": 1700
    },
    {
      "epoch": 0.34,
      "learning_rate": 3.378111418411695e-05,
      "loss": 0.6992,
      "step": 1710
    },
    {
      "epoch": 0.34,
      "learning_rate": 3.397866455946266e-05,
      "loss": 0.6602,
      "step": 1720
    },
    {
      "epoch": 0.34,
      "learning_rate": 3.4176214934808374e-05,
      "loss": 0.6961,
      "step": 1730
    },
    {
      "epoch": 0.34,
      "learning_rate": 3.4373765310154094e-05,
      "loss": 0.7106,
      "step": 1740
    },
    {
      "epoch": 0.35,
      "learning_rate": 3.457131568549981e-05,
      "loss": 0.6957,
      "step": 1750
    },
    {
      "epoch": 0.35,
      "learning_rate": 3.476886606084552e-05,
      "loss": 0.6981,
      "step": 1760
    },
    {
      "epoch": 0.35,
      "learning_rate": 3.496641643619123e-05,
      "loss": 0.7585,
      "step": 1770
    },
    {
      "epoch": 0.35,
      "learning_rate": 3.516396681153694e-05,
      "loss": 0.6903,
      "step": 1780
    },
    {
      "epoch": 0.35,
      "learning_rate": 3.536151718688266e-05,
      "loss": 0.6592,
      "step": 1790
    },
    {
      "epoch": 0.36,
      "learning_rate": 3.555906756222837e-05,
      "loss": 0.732,
      "step": 1800
    },
    {
      "epoch": 0.36,
      "learning_rate": 3.5756617937574084e-05,
      "loss": 0.703,
      "step": 1810
    },
    {
      "epoch": 0.36,
      "learning_rate": 3.5954168312919796e-05,
      "loss": 0.7397,
      "step": 1820
    },
    {
      "epoch": 0.36,
      "learning_rate": 3.615171868826551e-05,
      "loss": 0.6665,
      "step": 1830
    },
    {
      "epoch": 0.36,
      "learning_rate": 3.634926906361122e-05,
      "loss": 0.7053,
      "step": 1840
    },
    {
      "epoch": 0.37,
      "learning_rate": 3.6546819438956935e-05,
      "loss": 0.7008,
      "step": 1850
    },
    {
      "epoch": 0.37,
      "learning_rate": 3.674436981430265e-05,
      "loss": 0.7181,
      "step": 1860
    },
    {
      "epoch": 0.37,
      "learning_rate": 3.694192018964836e-05,
      "loss": 0.7223,
      "step": 1870
    },
    {
      "epoch": 0.37,
      "learning_rate": 3.713947056499408e-05,
      "loss": 0.7027,
      "step": 1880
    },
    {
      "epoch": 0.37,
      "learning_rate": 3.733702094033979e-05,
      "loss": 0.6727,
      "step": 1890
    },
    {
      "epoch": 0.38,
      "learning_rate": 3.75345713156855e-05,
      "loss": 0.6756,
      "step": 1900
    },
    {
      "epoch": 0.38,
      "learning_rate": 3.773212169103121e-05,
      "loss": 0.7386,
      "step": 1910
    },
    {
      "epoch": 0.38,
      "learning_rate": 3.7929672066376924e-05,
      "loss": 0.6824,
      "step": 1920
    },
    {
      "epoch": 0.38,
      "learning_rate": 3.8127222441722644e-05,
      "loss": 0.7292,
      "step": 1930
    },
    {
      "epoch": 0.38,
      "learning_rate": 3.8324772817068357e-05,
      "loss": 0.716,
      "step": 1940
    },
    {
      "epoch": 0.39,
      "learning_rate": 3.852232319241407e-05,
      "loss": 0.657,
      "step": 1950
    },
    {
      "epoch": 0.39,
      "learning_rate": 3.871987356775978e-05,
      "loss": 0.6604,
      "step": 1960
    },
    {
      "epoch": 0.39,
      "learning_rate": 3.891742394310549e-05,
      "loss": 0.679,
      "step": 1970
    },
    {
      "epoch": 0.39,
      "learning_rate": 3.911497431845121e-05,
      "loss": 0.6932,
      "step": 1980
    },
    {
      "epoch": 0.39,
      "learning_rate": 3.931252469379692e-05,
      "loss": 0.7374,
      "step": 1990
    },
    {
      "epoch": 0.4,
      "learning_rate": 3.951007506914263e-05,
      "loss": 0.7186,
      "step": 2000
    },
    {
      "epoch": 0.4,
      "learning_rate": 3.9707625444488346e-05,
      "loss": 0.7086,
      "step": 2010
    },
    {
      "epoch": 0.4,
      "learning_rate": 3.990517581983406e-05,
      "loss": 0.6731,
      "step": 2020
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.010272619517977e-05,
      "loss": 0.69,
      "step": 2030
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.0300276570525485e-05,
      "loss": 0.6567,
      "step": 2040
    },
    {
      "epoch": 0.4,
      "learning_rate": 4.04978269458712e-05,
      "loss": 0.6974,
      "step": 2050
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.069537732121691e-05,
      "loss": 0.7603,
      "step": 2060
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.089292769656263e-05,
      "loss": 0.7012,
      "step": 2070
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.109047807190834e-05,
      "loss": 0.7585,
      "step": 2080
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.128802844725405e-05,
      "loss": 0.7086,
      "step": 2090
    },
    {
      "epoch": 0.41,
      "learning_rate": 4.148557882259976e-05,
      "loss": 0.7345,
      "step": 2100
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.168312919794548e-05,
      "loss": 0.6888,
      "step": 2110
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.1880679573291194e-05,
      "loss": 0.6637,
      "step": 2120
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.2078229948636906e-05,
      "loss": 0.6936,
      "step": 2130
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.227578032398262e-05,
      "loss": 0.6844,
      "step": 2140
    },
    {
      "epoch": 0.42,
      "learning_rate": 4.2473330699328325e-05,
      "loss": 0.709,
      "step": 2150
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.2670881074674045e-05,
      "loss": 0.6244,
      "step": 2160
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.286843145001976e-05,
      "loss": 0.6936,
      "step": 2170
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.306598182536547e-05,
      "loss": 0.7073,
      "step": 2180
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.326353220071118e-05,
      "loss": 0.7202,
      "step": 2190
    },
    {
      "epoch": 0.43,
      "learning_rate": 4.3461082576056896e-05,
      "loss": 0.6506,
      "step": 2200
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.365863295140261e-05,
      "loss": 0.6339,
      "step": 2210
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.385618332674832e-05,
      "loss": 0.6635,
      "step": 2220
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.4053733702094034e-05,
      "loss": 0.6674,
      "step": 2230
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.425128407743975e-05,
      "loss": 0.6857,
      "step": 2240
    },
    {
      "epoch": 0.44,
      "learning_rate": 4.444883445278547e-05,
      "loss": 0.7223,
      "step": 2250
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.464638482813118e-05,
      "loss": 0.6716,
      "step": 2260
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.4843935203476886e-05,
      "loss": 0.7094,
      "step": 2270
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.50414855788226e-05,
      "loss": 0.6477,
      "step": 2280
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.523903595416831e-05,
      "loss": 0.6409,
      "step": 2290
    },
    {
      "epoch": 0.45,
      "learning_rate": 4.543658632951403e-05,
      "loss": 0.6682,
      "step": 2300
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.5634136704859744e-05,
      "loss": 0.6727,
      "step": 2310
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.5831687080205456e-05,
      "loss": 0.6348,
      "step": 2320
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.602923745555117e-05,
      "loss": 0.683,
      "step": 2330
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.6226787830896875e-05,
      "loss": 0.5872,
      "step": 2340
    },
    {
      "epoch": 0.46,
      "learning_rate": 4.6424338206242595e-05,
      "loss": 0.7156,
      "step": 2350
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.662188858158831e-05,
      "loss": 0.6441,
      "step": 2360
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.681943895693402e-05,
      "loss": 0.647,
      "step": 2370
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.701698933227973e-05,
      "loss": 0.7303,
      "step": 2380
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.7214539707625446e-05,
      "loss": 0.6304,
      "step": 2390
    },
    {
      "epoch": 0.47,
      "learning_rate": 4.741209008297116e-05,
      "loss": 0.6577,
      "step": 2400
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.760964045831687e-05,
      "loss": 0.6941,
      "step": 2410
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.7807190833662584e-05,
      "loss": 0.6309,
      "step": 2420
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.80047412090083e-05,
      "loss": 0.6584,
      "step": 2430
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.820229158435402e-05,
      "loss": 0.6387,
      "step": 2440
    },
    {
      "epoch": 0.48,
      "learning_rate": 4.839984195969973e-05,
      "loss": 0.6602,
      "step": 2450
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.8597392335045435e-05,
      "loss": 0.7043,
      "step": 2460
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.879494271039115e-05,
      "loss": 0.711,
      "step": 2470
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.899249308573686e-05,
      "loss": 0.6393,
      "step": 2480
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.919004346108258e-05,
      "loss": 0.7015,
      "step": 2490
    },
    {
      "epoch": 0.49,
      "learning_rate": 4.9387593836428293e-05,
      "loss": 0.6153,
      "step": 2500
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.9585144211774006e-05,
      "loss": 0.6787,
      "step": 2510
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.978269458711972e-05,
      "loss": 0.6316,
      "step": 2520
    },
    {
      "epoch": 0.5,
      "learning_rate": 4.998024496246543e-05,
      "loss": 0.6494,
      "step": 2530
    },
    {
      "epoch": 0.5,
      "learning_rate": 5.017779533781114e-05,
      "loss": 0.6925,
      "step": 2540
    },
    {
      "epoch": 0.5,
      "learning_rate": 5.0375345713156864e-05,
      "loss": 0.7249,
      "step": 2550
    },
    {
      "epoch": 0.51,
      "learning_rate": 5.057289608850258e-05,
      "loss": 0.6799,
      "step": 2560
    },
    {
      "epoch": 0.51,
      "learning_rate": 5.077044646384829e-05,
      "loss": 0.7094,
      "step": 2570
    },
    {
      "epoch": 0.51,
      "learning_rate": 5.0967996839193996e-05,
      "loss": 0.7007,
      "step": 2580
    },
    {
      "epoch": 0.51,
      "learning_rate": 5.116554721453971e-05,
      "loss": 0.7456,
      "step": 2590
    },
    {
      "epoch": 0.51,
      "learning_rate": 5.136309758988542e-05,
      "loss": 0.6767,
      "step": 2600
    },
    {
      "epoch": 0.52,
      "learning_rate": 5.1560647965231134e-05,
      "loss": 0.7366,
      "step": 2610
    },
    {
      "epoch": 0.52,
      "learning_rate": 5.175819834057685e-05,
      "loss": 0.6579,
      "step": 2620
    },
    {
      "epoch": 0.52,
      "learning_rate": 5.195574871592256e-05,
      "loss": 0.6985,
      "step": 2630
    },
    {
      "epoch": 0.52,
      "learning_rate": 5.215329909126827e-05,
      "loss": 0.7202,
      "step": 2640
    },
    {
      "epoch": 0.52,
      "learning_rate": 5.235084946661399e-05,
      "loss": 0.6233,
      "step": 2650
    },
    {
      "epoch": 0.53,
      "learning_rate": 5.2548399841959705e-05,
      "loss": 0.6271,
      "step": 2660
    },
    {
      "epoch": 0.53,
      "learning_rate": 5.274595021730542e-05,
      "loss": 0.697,
      "step": 2670
    },
    {
      "epoch": 0.53,
      "learning_rate": 5.294350059265113e-05,
      "loss": 0.6596,
      "step": 2680
    },
    {
      "epoch": 0.53,
      "learning_rate": 5.314105096799684e-05,
      "loss": 0.6474,
      "step": 2690
    },
    {
      "epoch": 0.53,
      "learning_rate": 5.3338601343342556e-05,
      "loss": 0.6291,
      "step": 2700
    },
    {
      "epoch": 0.54,
      "learning_rate": 5.353615171868827e-05,
      "loss": 0.634,
      "step": 2710
    },
    {
      "epoch": 0.54,
      "learning_rate": 5.3733702094033975e-05,
      "loss": 0.67,
      "step": 2720
    },
    {
      "epoch": 0.54,
      "learning_rate": 5.393125246937969e-05,
      "loss": 0.6784,
      "step": 2730
    },
    {
      "epoch": 0.54,
      "learning_rate": 5.4128802844725414e-05,
      "loss": 0.6826,
      "step": 2740
    },
    {
      "epoch": 0.54,
      "learning_rate": 5.432635322007113e-05,
      "loss": 0.6485,
      "step": 2750
    },
    {
      "epoch": 0.55,
      "learning_rate": 5.452390359541684e-05,
      "loss": 0.6364,
      "step": 2760
    },
    {
      "epoch": 0.55,
      "learning_rate": 5.4721453970762546e-05,
      "loss": 0.6196,
      "step": 2770
    },
    {
      "epoch": 0.55,
      "learning_rate": 5.491900434610826e-05,
      "loss": 0.6812,
      "step": 2780
    },
    {
      "epoch": 0.55,
      "learning_rate": 5.511655472145397e-05,
      "loss": 0.6404,
      "step": 2790
    },
    {
      "epoch": 0.55,
      "learning_rate": 5.5314105096799684e-05,
      "loss": 0.6725,
      "step": 2800
    },
    {
      "epoch": 0.56,
      "learning_rate": 5.55116554721454e-05,
      "loss": 0.6811,
      "step": 2810
    },
    {
      "epoch": 0.56,
      "learning_rate": 5.570920584749111e-05,
      "loss": 0.6282,
      "step": 2820
    },
    {
      "epoch": 0.56,
      "learning_rate": 5.590675622283683e-05,
      "loss": 0.6505,
      "step": 2830
    },
    {
      "epoch": 0.56,
      "learning_rate": 5.610430659818254e-05,
      "loss": 0.6352,
      "step": 2840
    },
    {
      "epoch": 0.56,
      "learning_rate": 5.6301856973528255e-05,
      "loss": 0.6836,
      "step": 2850
    },
    {
      "epoch": 0.56,
      "learning_rate": 5.649940734887397e-05,
      "loss": 0.5995,
      "step": 2860
    },
    {
      "epoch": 0.57,
      "learning_rate": 5.669695772421968e-05,
      "loss": 0.6773,
      "step": 2870
    },
    {
      "epoch": 0.57,
      "learning_rate": 5.689450809956539e-05,
      "loss": 0.6272,
      "step": 2880
    },
    {
      "epoch": 0.57,
      "learning_rate": 5.7092058474911106e-05,
      "loss": 0.681,
      "step": 2890
    },
    {
      "epoch": 0.57,
      "learning_rate": 5.728960885025681e-05,
      "loss": 0.6307,
      "step": 2900
    },
    {
      "epoch": 0.57,
      "learning_rate": 5.7487159225602525e-05,
      "loss": 0.6643,
      "step": 2910
    },
    {
      "epoch": 0.58,
      "learning_rate": 5.768470960094824e-05,
      "loss": 0.6184,
      "step": 2920
    },
    {
      "epoch": 0.58,
      "learning_rate": 5.7882259976293964e-05,
      "loss": 0.6572,
      "step": 2930
    },
    {
      "epoch": 0.58,
      "learning_rate": 5.807981035163968e-05,
      "loss": 0.6898,
      "step": 2940
    },
    {
      "epoch": 0.58,
      "learning_rate": 5.827736072698538e-05,
      "loss": 0.6442,
      "step": 2950
    },
    {
      "epoch": 0.58,
      "learning_rate": 5.8474911102331095e-05,
      "loss": 0.6782,
      "step": 2960
    },
    {
      "epoch": 0.59,
      "learning_rate": 5.867246147767681e-05,
      "loss": 0.6749,
      "step": 2970
    },
    {
      "epoch": 0.59,
      "learning_rate": 5.887001185302252e-05,
      "loss": 0.6217,
      "step": 2980
    },
    {
      "epoch": 0.59,
      "learning_rate": 5.9067562228368234e-05,
      "loss": 0.6644,
      "step": 2990
    },
    {
      "epoch": 0.59,
      "learning_rate": 5.926511260371395e-05,
      "loss": 0.6527,
      "step": 3000
    },
    {
      "epoch": 0.59,
      "learning_rate": 5.946266297905966e-05,
      "loss": 0.6961,
      "step": 3010
    },
    {
      "epoch": 0.6,
      "learning_rate": 5.966021335440538e-05,
      "loss": 0.6514,
      "step": 3020
    },
    {
      "epoch": 0.6,
      "learning_rate": 5.985776372975109e-05,
      "loss": 0.6066,
      "step": 3030
    },
    {
      "epoch": 0.6,
      "learning_rate": 6.0055314105096805e-05,
      "loss": 0.673,
      "step": 3040
    },
    {
      "epoch": 0.6,
      "learning_rate": 6.025286448044252e-05,
      "loss": 0.6685,
      "step": 3050
    },
    {
      "epoch": 0.6,
      "learning_rate": 6.045041485578823e-05,
      "loss": 0.6274,
      "step": 3060
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.064796523113394e-05,
      "loss": 0.7091,
      "step": 3070
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.0845515606479656e-05,
      "loss": 0.6791,
      "step": 3080
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.104306598182537e-05,
      "loss": 0.6348,
      "step": 3090
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.124061635717107e-05,
      "loss": 0.6043,
      "step": 3100
    },
    {
      "epoch": 0.61,
      "learning_rate": 6.14381667325168e-05,
      "loss": 0.6991,
      "step": 3110
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.163571710786251e-05,
      "loss": 0.6613,
      "step": 3120
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.183326748320822e-05,
      "loss": 0.6498,
      "step": 3130
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.203081785855394e-05,
      "loss": 0.6819,
      "step": 3140
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.222836823389965e-05,
      "loss": 0.6316,
      "step": 3150
    },
    {
      "epoch": 0.62,
      "learning_rate": 6.242591860924536e-05,
      "loss": 0.6672,
      "step": 3160
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.262346898459107e-05,
      "loss": 0.6602,
      "step": 3170
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.282101935993678e-05,
      "loss": 0.6463,
      "step": 3180
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.30185697352825e-05,
      "loss": 0.614,
      "step": 3190
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.321612011062822e-05,
      "loss": 0.6689,
      "step": 3200
    },
    {
      "epoch": 0.63,
      "learning_rate": 6.341367048597394e-05,
      "loss": 0.6224,
      "step": 3210
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.361122086131964e-05,
      "loss": 0.5965,
      "step": 3220
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.380877123666536e-05,
      "loss": 0.6219,
      "step": 3230
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.400632161201107e-05,
      "loss": 0.6675,
      "step": 3240
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.420387198735677e-05,
      "loss": 0.7074,
      "step": 3250
    },
    {
      "epoch": 0.64,
      "learning_rate": 6.440142236270249e-05,
      "loss": 0.6304,
      "step": 3260
    },
    {
      "epoch": 0.65,
      "learning_rate": 6.45989727380482e-05,
      "loss": 0.6729,
      "step": 3270
    },
    {
      "epoch": 0.65,
      "learning_rate": 6.479652311339392e-05,
      "loss": 0.6177,
      "step": 3280
    },
    {
      "epoch": 0.65,
      "learning_rate": 6.499407348873962e-05,
      "loss": 0.6709,
      "step": 3290
    },
    {
      "epoch": 0.65,
      "learning_rate": 6.519162386408534e-05,
      "loss": 0.7026,
      "step": 3300
    },
    {
      "epoch": 0.65,
      "learning_rate": 6.538917423943106e-05,
      "loss": 0.6838,
      "step": 3310
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.558672461477677e-05,
      "loss": 0.6294,
      "step": 3320
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.578427499012249e-05,
      "loss": 0.6447,
      "step": 3330
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.59818253654682e-05,
      "loss": 0.592,
      "step": 3340
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.617937574081391e-05,
      "loss": 0.6698,
      "step": 3350
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.637692611615962e-05,
      "loss": 0.6528,
      "step": 3360
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.657447649150533e-05,
      "loss": 0.6679,
      "step": 3370
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.677202686685105e-05,
      "loss": 0.6379,
      "step": 3380
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.696957724219677e-05,
      "loss": 0.6213,
      "step": 3390
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.716712761754249e-05,
      "loss": 0.6696,
      "step": 3400
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.736467799288819e-05,
      "loss": 0.6174,
      "step": 3410
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.75622283682339e-05,
      "loss": 0.6211,
      "step": 3420
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.775977874357962e-05,
      "loss": 0.6016,
      "step": 3430
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.795732911892532e-05,
      "loss": 0.6407,
      "step": 3440
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.815487949427104e-05,
      "loss": 0.6349,
      "step": 3450
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.835242986961675e-05,
      "loss": 0.6478,
      "step": 3460
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.854998024496247e-05,
      "loss": 0.6733,
      "step": 3470
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.874753062030819e-05,
      "loss": 0.6037,
      "step": 3480
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.89450809956539e-05,
      "loss": 0.6621,
      "step": 3490
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.914263137099961e-05,
      "loss": 0.6122,
      "step": 3500
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.934018174634532e-05,
      "loss": 0.594,
      "step": 3510
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.953773212169104e-05,
      "loss": 0.6445,
      "step": 3520
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.973528249703675e-05,
      "loss": 0.6363,
      "step": 3530
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.993283287238246e-05,
      "loss": 0.635,
      "step": 3540
    },
    {
      "epoch": 0.7,
      "learning_rate": 7.013038324772817e-05,
      "loss": 0.6506,
      "step": 3550
    },
    {
      "epoch": 0.7,
      "learning_rate": 7.032793362307388e-05,
      "loss": 0.6546,
      "step": 3560
    },
    {
      "epoch": 0.71,
      "learning_rate": 7.052548399841961e-05,
      "loss": 0.6486,
      "step": 3570
    },
    {
      "epoch": 0.71,
      "learning_rate": 7.072303437376532e-05,
      "loss": 0.6543,
      "step": 3580
    },
    {
      "epoch": 0.71,
      "learning_rate": 7.092058474911104e-05,
      "loss": 0.6799,
      "step": 3590
    },
    {
      "epoch": 0.71,
      "learning_rate": 7.111813512445674e-05,
      "loss": 0.665,
      "step": 3600
    },
    {
      "epoch": 0.71,
      "learning_rate": 7.131568549980245e-05,
      "loss": 0.6728,
      "step": 3610
    },
    {
      "epoch": 0.72,
      "learning_rate": 7.151323587514817e-05,
      "loss": 0.6574,
      "step": 3620
    },
    {
      "epoch": 0.72,
      "learning_rate": 7.171078625049387e-05,
      "loss": 0.6618,
      "step": 3630
    },
    {
      "epoch": 0.72,
      "learning_rate": 7.190833662583959e-05,
      "loss": 0.5915,
      "step": 3640
    },
    {
      "epoch": 0.72,
      "learning_rate": 7.21058870011853e-05,
      "loss": 0.5632,
      "step": 3650
    },
    {
      "epoch": 0.72,
      "learning_rate": 7.230343737653102e-05,
      "loss": 0.6802,
      "step": 3660
    },
    {
      "epoch": 0.72,
      "learning_rate": 7.250098775187674e-05,
      "loss": 0.6749,
      "step": 3670
    },
    {
      "epoch": 0.73,
      "learning_rate": 7.269853812722244e-05,
      "loss": 0.665,
      "step": 3680
    },
    {
      "epoch": 0.73,
      "learning_rate": 7.289608850256816e-05,
      "loss": 0.7057,
      "step": 3690
    },
    {
      "epoch": 0.73,
      "learning_rate": 7.309363887791387e-05,
      "loss": 0.6284,
      "step": 3700
    },
    {
      "epoch": 0.73,
      "learning_rate": 7.329118925325959e-05,
      "loss": 0.6445,
      "step": 3710
    },
    {
      "epoch": 0.73,
      "learning_rate": 7.34887396286053e-05,
      "loss": 0.6352,
      "step": 3720
    },
    {
      "epoch": 0.74,
      "learning_rate": 7.368629000395101e-05,
      "loss": 0.6989,
      "step": 3730
    },
    {
      "epoch": 0.74,
      "learning_rate": 7.388384037929672e-05,
      "loss": 0.6579,
      "step": 3740
    },
    {
      "epoch": 0.74,
      "learning_rate": 7.408139075464243e-05,
      "loss": 0.6215,
      "step": 3750
    },
    {
      "epoch": 0.74,
      "learning_rate": 7.427894112998816e-05,
      "loss": 0.5917,
      "step": 3760
    },
    {
      "epoch": 0.74,
      "learning_rate": 7.447649150533387e-05,
      "loss": 0.6685,
      "step": 3770
    },
    {
      "epoch": 0.75,
      "learning_rate": 7.467404188067959e-05,
      "loss": 0.6314,
      "step": 3780
    },
    {
      "epoch": 0.75,
      "learning_rate": 7.487159225602529e-05,
      "loss": 0.6184,
      "step": 3790
    },
    {
      "epoch": 0.75,
      "learning_rate": 7.5069142631371e-05,
      "loss": 0.6256,
      "step": 3800
    },
    {
      "epoch": 0.75,
      "learning_rate": 7.526669300671672e-05,
      "loss": 0.6055,
      "step": 3810
    },
    {
      "epoch": 0.75,
      "learning_rate": 7.546424338206242e-05,
      "loss": 0.6527,
      "step": 3820
    },
    {
      "epoch": 0.76,
      "learning_rate": 7.566179375740814e-05,
      "loss": 0.6632,
      "step": 3830
    },
    {
      "epoch": 0.76,
      "learning_rate": 7.585934413275385e-05,
      "loss": 0.5831,
      "step": 3840
    },
    {
      "epoch": 0.76,
      "learning_rate": 7.605689450809957e-05,
      "loss": 0.6703,
      "step": 3850
    },
    {
      "epoch": 0.76,
      "learning_rate": 7.625444488344529e-05,
      "loss": 0.6218,
      "step": 3860
    },
    {
      "epoch": 0.76,
      "learning_rate": 7.6451995258791e-05,
      "loss": 0.6795,
      "step": 3870
    },
    {
      "epoch": 0.77,
      "learning_rate": 7.664954563413671e-05,
      "loss": 0.6779,
      "step": 3880
    },
    {
      "epoch": 0.77,
      "learning_rate": 7.684709600948242e-05,
      "loss": 0.6174,
      "step": 3890
    },
    {
      "epoch": 0.77,
      "learning_rate": 7.704464638482814e-05,
      "loss": 0.6547,
      "step": 3900
    },
    {
      "epoch": 0.77,
      "learning_rate": 7.724219676017384e-05,
      "loss": 0.6003,
      "step": 3910
    },
    {
      "epoch": 0.77,
      "learning_rate": 7.743974713551956e-05,
      "loss": 0.6768,
      "step": 3920
    },
    {
      "epoch": 0.78,
      "learning_rate": 7.763729751086527e-05,
      "loss": 0.6117,
      "step": 3930
    },
    {
      "epoch": 0.78,
      "learning_rate": 7.783484788621098e-05,
      "loss": 0.618,
      "step": 3940
    },
    {
      "epoch": 0.78,
      "learning_rate": 7.803239826155671e-05,
      "loss": 0.6749,
      "step": 3950
    },
    {
      "epoch": 0.78,
      "learning_rate": 7.822994863690242e-05,
      "loss": 0.6358,
      "step": 3960
    },
    {
      "epoch": 0.78,
      "learning_rate": 7.842749901224814e-05,
      "loss": 0.6954,
      "step": 3970
    },
    {
      "epoch": 0.79,
      "learning_rate": 7.862504938759384e-05,
      "loss": 0.5934,
      "step": 3980
    },
    {
      "epoch": 0.79,
      "learning_rate": 7.882259976293955e-05,
      "loss": 0.6526,
      "step": 3990
    },
    {
      "epoch": 0.79,
      "learning_rate": 7.902015013828527e-05,
      "loss": 0.6171,
      "step": 4000
    },
    {
      "epoch": 0.79,
      "learning_rate": 7.921770051363097e-05,
      "loss": 0.6219,
      "step": 4010
    },
    {
      "epoch": 0.79,
      "learning_rate": 7.941525088897669e-05,
      "loss": 0.6421,
      "step": 4020
    },
    {
      "epoch": 0.8,
      "learning_rate": 7.96128012643224e-05,
      "loss": 0.6531,
      "step": 4030
    },
    {
      "epoch": 0.8,
      "learning_rate": 7.981035163966812e-05,
      "loss": 0.5801,
      "step": 4040
    },
    {
      "epoch": 0.8,
      "learning_rate": 8.000790201501384e-05,
      "loss": 0.63,
      "step": 4050
    },
    {
      "epoch": 0.8,
      "learning_rate": 8.020545239035954e-05,
      "loss": 0.6303,
      "step": 4060
    },
    {
      "epoch": 0.8,
      "learning_rate": 8.040300276570526e-05,
      "loss": 0.612,
      "step": 4070
    },
    {
      "epoch": 0.81,
      "learning_rate": 8.060055314105097e-05,
      "loss": 0.6031,
      "step": 4080
    },
    {
      "epoch": 0.81,
      "learning_rate": 8.079810351639669e-05,
      "loss": 0.6569,
      "step": 4090
    },
    {
      "epoch": 0.81,
      "learning_rate": 8.09956538917424e-05,
      "loss": 0.7017,
      "step": 4100
    },
    {
      "epoch": 0.81,
      "learning_rate": 8.119320426708811e-05,
      "loss": 0.6456,
      "step": 4110
    },
    {
      "epoch": 0.81,
      "learning_rate": 8.139075464243382e-05,
      "loss": 0.6752,
      "step": 4120
    },
    {
      "epoch": 0.82,
      "learning_rate": 8.158830501777954e-05,
      "loss": 0.5988,
      "step": 4130
    },
    {
      "epoch": 0.82,
      "learning_rate": 8.178585539312526e-05,
      "loss": 0.6224,
      "step": 4140
    },
    {
      "epoch": 0.82,
      "learning_rate": 8.198340576847097e-05,
      "loss": 0.6238,
      "step": 4150
    },
    {
      "epoch": 0.82,
      "learning_rate": 8.218095614381669e-05,
      "loss": 0.6953,
      "step": 4160
    },
    {
      "epoch": 0.82,
      "learning_rate": 8.237850651916239e-05,
      "loss": 0.675,
      "step": 4170
    },
    {
      "epoch": 0.83,
      "learning_rate": 8.25760568945081e-05,
      "loss": 0.6091,
      "step": 4180
    },
    {
      "epoch": 0.83,
      "learning_rate": 8.277360726985382e-05,
      "loss": 0.6527,
      "step": 4190
    },
    {
      "epoch": 0.83,
      "learning_rate": 8.297115764519952e-05,
      "loss": 0.6183,
      "step": 4200
    },
    {
      "epoch": 0.83,
      "learning_rate": 8.316870802054524e-05,
      "loss": 0.6101,
      "step": 4210
    },
    {
      "epoch": 0.83,
      "learning_rate": 8.336625839589096e-05,
      "loss": 0.6639,
      "step": 4220
    },
    {
      "epoch": 0.84,
      "learning_rate": 8.356380877123667e-05,
      "loss": 0.6226,
      "step": 4230
    },
    {
      "epoch": 0.84,
      "learning_rate": 8.376135914658239e-05,
      "loss": 0.6441,
      "step": 4240
    },
    {
      "epoch": 0.84,
      "learning_rate": 8.39589095219281e-05,
      "loss": 0.5824,
      "step": 4250
    },
    {
      "epoch": 0.84,
      "learning_rate": 8.415645989727381e-05,
      "loss": 0.6669,
      "step": 4260
    },
    {
      "epoch": 0.84,
      "learning_rate": 8.435401027261952e-05,
      "loss": 0.6821,
      "step": 4270
    },
    {
      "epoch": 0.85,
      "learning_rate": 8.455156064796524e-05,
      "loss": 0.6243,
      "step": 4280
    },
    {
      "epoch": 0.85,
      "learning_rate": 8.474911102331094e-05,
      "loss": 0.5964,
      "step": 4290
    },
    {
      "epoch": 0.85,
      "learning_rate": 8.494666139865665e-05,
      "loss": 0.5887,
      "step": 4300
    },
    {
      "epoch": 0.85,
      "learning_rate": 8.514421177400237e-05,
      "loss": 0.6496,
      "step": 4310
    },
    {
      "epoch": 0.85,
      "learning_rate": 8.534176214934809e-05,
      "loss": 0.6563,
      "step": 4320
    },
    {
      "epoch": 0.86,
      "learning_rate": 8.553931252469381e-05,
      "loss": 0.6282,
      "step": 4330
    },
    {
      "epoch": 0.86,
      "learning_rate": 8.573686290003952e-05,
      "loss": 0.6189,
      "step": 4340
    },
    {
      "epoch": 0.86,
      "learning_rate": 8.593441327538523e-05,
      "loss": 0.6376,
      "step": 4350
    },
    {
      "epoch": 0.86,
      "learning_rate": 8.613196365073094e-05,
      "loss": 0.5814,
      "step": 4360
    },
    {
      "epoch": 0.86,
      "learning_rate": 8.632951402607665e-05,
      "loss": 0.5461,
      "step": 4370
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.652706440142237e-05,
      "loss": 0.6651,
      "step": 4380
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.672461477676807e-05,
      "loss": 0.6614,
      "step": 4390
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.692216515211379e-05,
      "loss": 0.604,
      "step": 4400
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.711971552745951e-05,
      "loss": 0.6378,
      "step": 4410
    },
    {
      "epoch": 0.87,
      "learning_rate": 8.731726590280522e-05,
      "loss": 0.6211,
      "step": 4420
    },
    {
      "epoch": 0.88,
      "learning_rate": 8.751481627815094e-05,
      "loss": 0.5943,
      "step": 4430
    },
    {
      "epoch": 0.88,
      "learning_rate": 8.771236665349664e-05,
      "loss": 0.6493,
      "step": 4440
    },
    {
      "epoch": 0.88,
      "learning_rate": 8.790991702884236e-05,
      "loss": 0.643,
      "step": 4450
    },
    {
      "epoch": 0.88,
      "learning_rate": 8.810746740418807e-05,
      "loss": 0.6311,
      "step": 4460
    },
    {
      "epoch": 0.88,
      "learning_rate": 8.830501777953379e-05,
      "loss": 0.6606,
      "step": 4470
    },
    {
      "epoch": 0.88,
      "learning_rate": 8.85025681548795e-05,
      "loss": 0.615,
      "step": 4480
    },
    {
      "epoch": 0.89,
      "learning_rate": 8.87001185302252e-05,
      "loss": 0.6847,
      "step": 4490
    },
    {
      "epoch": 0.89,
      "learning_rate": 8.889766890557093e-05,
      "loss": 0.6207,
      "step": 4500
    },
    {
      "epoch": 0.89,
      "learning_rate": 8.909521928091664e-05,
      "loss": 0.6094,
      "step": 4510
    },
    {
      "epoch": 0.89,
      "learning_rate": 8.929276965626236e-05,
      "loss": 0.6294,
      "step": 4520
    },
    {
      "epoch": 0.89,
      "learning_rate": 8.949032003160807e-05,
      "loss": 0.644,
      "step": 4530
    },
    {
      "epoch": 0.9,
      "learning_rate": 8.968787040695377e-05,
      "loss": 0.6265,
      "step": 4540
    },
    {
      "epoch": 0.9,
      "learning_rate": 8.988542078229949e-05,
      "loss": 0.6348,
      "step": 4550
    },
    {
      "epoch": 0.9,
      "learning_rate": 9.00829711576452e-05,
      "loss": 0.5913,
      "step": 4560
    },
    {
      "epoch": 0.9,
      "learning_rate": 9.028052153299092e-05,
      "loss": 0.6377,
      "step": 4570
    },
    {
      "epoch": 0.9,
      "learning_rate": 9.047807190833662e-05,
      "loss": 0.5929,
      "step": 4580
    },
    {
      "epoch": 0.91,
      "learning_rate": 9.067562228368236e-05,
      "loss": 0.6708,
      "step": 4590
    },
    {
      "epoch": 0.91,
      "learning_rate": 9.087317265902806e-05,
      "loss": 0.6314,
      "step": 4600
    },
    {
      "epoch": 0.91,
      "learning_rate": 9.107072303437377e-05,
      "loss": 0.6853,
      "step": 4610
    },
    {
      "epoch": 0.91,
      "learning_rate": 9.126827340971949e-05,
      "loss": 0.6196,
      "step": 4620
    },
    {
      "epoch": 0.91,
      "learning_rate": 9.146582378506519e-05,
      "loss": 0.5957,
      "step": 4630
    },
    {
      "epoch": 0.92,
      "learning_rate": 9.166337416041091e-05,
      "loss": 0.5991,
      "step": 4640
    },
    {
      "epoch": 0.92,
      "learning_rate": 9.186092453575662e-05,
      "loss": 0.6602,
      "step": 4650
    },
    {
      "epoch": 0.92,
      "learning_rate": 9.205847491110234e-05,
      "loss": 0.6852,
      "step": 4660
    },
    {
      "epoch": 0.92,
      "learning_rate": 9.225602528644804e-05,
      "loss": 0.5723,
      "step": 4670
    },
    {
      "epoch": 0.92,
      "learning_rate": 9.245357566179375e-05,
      "loss": 0.6218,
      "step": 4680
    },
    {
      "epoch": 0.93,
      "learning_rate": 9.265112603713948e-05,
      "loss": 0.625,
      "step": 4690
    },
    {
      "epoch": 0.93,
      "learning_rate": 9.284867641248519e-05,
      "loss": 0.6035,
      "step": 4700
    },
    {
      "epoch": 0.93,
      "learning_rate": 9.304622678783091e-05,
      "loss": 0.6119,
      "step": 4710
    },
    {
      "epoch": 0.93,
      "learning_rate": 9.324377716317662e-05,
      "loss": 0.6284,
      "step": 4720
    },
    {
      "epoch": 0.93,
      "learning_rate": 9.344132753852232e-05,
      "loss": 0.5956,
      "step": 4730
    },
    {
      "epoch": 0.94,
      "learning_rate": 9.363887791386804e-05,
      "loss": 0.5773,
      "step": 4740
    },
    {
      "epoch": 0.94,
      "learning_rate": 9.383642828921375e-05,
      "loss": 0.6735,
      "step": 4750
    },
    {
      "epoch": 0.94,
      "learning_rate": 9.403397866455947e-05,
      "loss": 0.6019,
      "step": 4760
    },
    {
      "epoch": 0.94,
      "learning_rate": 9.423152903990517e-05,
      "loss": 0.6271,
      "step": 4770
    },
    {
      "epoch": 0.94,
      "learning_rate": 9.442907941525089e-05,
      "loss": 0.6201,
      "step": 4780
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.462662979059661e-05,
      "loss": 0.6904,
      "step": 4790
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.482418016594232e-05,
      "loss": 0.6405,
      "step": 4800
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.502173054128804e-05,
      "loss": 0.5189,
      "step": 4810
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.521928091663374e-05,
      "loss": 0.623,
      "step": 4820
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.541683129197946e-05,
      "loss": 0.6278,
      "step": 4830
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.561438166732517e-05,
      "loss": 0.6264,
      "step": 4840
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.581193204267089e-05,
      "loss": 0.6407,
      "step": 4850
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.60094824180166e-05,
      "loss": 0.6766,
      "step": 4860
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.620703279336231e-05,
      "loss": 0.6492,
      "step": 4870
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.640458316870803e-05,
      "loss": 0.6087,
      "step": 4880
    },
    {
      "epoch": 0.97,
      "learning_rate": 9.660213354405374e-05,
      "loss": 0.5838,
      "step": 4890
    },
    {
      "epoch": 0.97,
      "learning_rate": 9.679968391939946e-05,
      "loss": 0.6389,
      "step": 4900
    },
    {
      "epoch": 0.97,
      "learning_rate": 9.699723429474516e-05,
      "loss": 0.664,
      "step": 4910
    },
    {
      "epoch": 0.97,
      "learning_rate": 9.719478467009087e-05,
      "loss": 0.6164,
      "step": 4920
    },
    {
      "epoch": 0.97,
      "learning_rate": 9.739233504543659e-05,
      "loss": 0.6112,
      "step": 4930
    },
    {
      "epoch": 0.98,
      "learning_rate": 9.75898854207823e-05,
      "loss": 0.5928,
      "step": 4940
    },
    {
      "epoch": 0.98,
      "learning_rate": 9.778743579612802e-05,
      "loss": 0.6131,
      "step": 4950
    },
    {
      "epoch": 0.98,
      "learning_rate": 9.798498617147372e-05,
      "loss": 0.6221,
      "step": 4960
    },
    {
      "epoch": 0.98,
      "learning_rate": 9.818253654681944e-05,
      "loss": 0.6302,
      "step": 4970
    },
    {
      "epoch": 0.98,
      "learning_rate": 9.838008692216516e-05,
      "loss": 0.6185,
      "step": 4980
    },
    {
      "epoch": 0.99,
      "learning_rate": 9.857763729751087e-05,
      "loss": 0.6397,
      "step": 4990
    },
    {
      "epoch": 0.99,
      "learning_rate": 9.877518767285659e-05,
      "loss": 0.6381,
      "step": 5000
    },
    {
      "epoch": 0.99,
      "learning_rate": 9.897273804820229e-05,
      "loss": 0.6274,
      "step": 5010
    },
    {
      "epoch": 0.99,
      "learning_rate": 9.917028842354801e-05,
      "loss": 0.5895,
      "step": 5020
    },
    {
      "epoch": 0.99,
      "learning_rate": 9.936783879889372e-05,
      "loss": 0.6603,
      "step": 5030
    },
    {
      "epoch": 1.0,
      "learning_rate": 9.956538917423944e-05,
      "loss": 0.5812,
      "step": 5040
    },
    {
      "epoch": 1.0,
      "learning_rate": 9.976293954958514e-05,
      "loss": 0.6215,
      "step": 5050
    },
    {
      "epoch": 1.0,
      "learning_rate": 9.996048992493086e-05,
      "loss": 0.5478,
      "step": 5060
    },
    {
      "epoch": 1.0,
      "learning_rate": 9.996048992493086e-05,
      "loss": 0.5893,
      "step": 5070
    },
    {
      "epoch": 1.0,
      "learning_rate": 9.991110233109443e-05,
      "loss": 0.6172,
      "step": 5080
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.986171473725801e-05,
      "loss": 0.609,
      "step": 5090
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.981232714342158e-05,
      "loss": 0.5751,
      "step": 5100
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.976293954958514e-05,
      "loss": 0.6062,
      "step": 5110
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.971355195574872e-05,
      "loss": 0.5886,
      "step": 5120
    },
    {
      "epoch": 1.01,
      "learning_rate": 9.966416436191229e-05,
      "loss": 0.5286,
      "step": 5130
    },
    {
      "epoch": 1.02,
      "learning_rate": 9.961477676807587e-05,
      "loss": 0.611,
      "step": 5140
    },
    {
      "epoch": 1.02,
      "learning_rate": 9.956538917423944e-05,
      "loss": 0.5796,
      "step": 5150
    },
    {
      "epoch": 1.02,
      "learning_rate": 9.9516001580403e-05,
      "loss": 0.5771,
      "step": 5160
    },
    {
      "epoch": 1.02,
      "learning_rate": 9.946661398656658e-05,
      "loss": 0.5545,
      "step": 5170
    },
    {
      "epoch": 1.02,
      "learning_rate": 9.941722639273015e-05,
      "loss": 0.5793,
      "step": 5180
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.936783879889372e-05,
      "loss": 0.6067,
      "step": 5190
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.93184512050573e-05,
      "loss": 0.5685,
      "step": 5200
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.926906361122087e-05,
      "loss": 0.5841,
      "step": 5210
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.921967601738443e-05,
      "loss": 0.5445,
      "step": 5220
    },
    {
      "epoch": 1.03,
      "learning_rate": 9.917028842354801e-05,
      "loss": 0.5709,
      "step": 5230
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.912090082971158e-05,
      "loss": 0.607,
      "step": 5240
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.907151323587516e-05,
      "loss": 0.5702,
      "step": 5250
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.902212564203873e-05,
      "loss": 0.5739,
      "step": 5260
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.897273804820229e-05,
      "loss": 0.6375,
      "step": 5270
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.892335045436587e-05,
      "loss": 0.6164,
      "step": 5280
    },
    {
      "epoch": 1.04,
      "learning_rate": 9.887396286052944e-05,
      "loss": 0.6189,
      "step": 5290
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.8824575266693e-05,
      "loss": 0.5755,
      "step": 5300
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.877518767285659e-05,
      "loss": 0.5973,
      "step": 5310
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.872580007902015e-05,
      "loss": 0.6061,
      "step": 5320
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.867641248518372e-05,
      "loss": 0.5847,
      "step": 5330
    },
    {
      "epoch": 1.05,
      "learning_rate": 9.86270248913473e-05,
      "loss": 0.5368,
      "step": 5340
    },
    {
      "epoch": 1.06,
      "learning_rate": 9.857763729751087e-05,
      "loss": 0.5204,
      "step": 5350
    },
    {
      "epoch": 1.06,
      "learning_rate": 9.852824970367445e-05,
      "loss": 0.5737,
      "step": 5360
    },
    {
      "epoch": 1.06,
      "learning_rate": 9.847886210983801e-05,
      "loss": 0.5824,
      "step": 5370
    },
    {
      "epoch": 1.06,
      "learning_rate": 9.842947451600158e-05,
      "loss": 0.5648,
      "step": 5380
    },
    {
      "epoch": 1.06,
      "learning_rate": 9.838008692216516e-05,
      "loss": 0.5426,
      "step": 5390
    },
    {
      "epoch": 1.07,
      "learning_rate": 9.833069932832873e-05,
      "loss": 0.6269,
      "step": 5400
    },
    {
      "epoch": 1.07,
      "learning_rate": 9.82813117344923e-05,
      "loss": 0.5373,
      "step": 5410
    },
    {
      "epoch": 1.07,
      "learning_rate": 9.823192414065587e-05,
      "loss": 0.5895,
      "step": 5420
    },
    {
      "epoch": 1.07,
      "learning_rate": 9.818253654681944e-05,
      "loss": 0.5642,
      "step": 5430
    },
    {
      "epoch": 1.07,
      "learning_rate": 9.813314895298301e-05,
      "loss": 0.5318,
      "step": 5440
    },
    {
      "epoch": 1.08,
      "learning_rate": 9.808376135914659e-05,
      "loss": 0.6078,
      "step": 5450
    },
    {
      "epoch": 1.08,
      "learning_rate": 9.803437376531016e-05,
      "loss": 0.5631,
      "step": 5460
    },
    {
      "epoch": 1.08,
      "learning_rate": 9.798498617147372e-05,
      "loss": 0.6024,
      "step": 5470
    },
    {
      "epoch": 1.08,
      "learning_rate": 9.79355985776373e-05,
      "loss": 0.6337,
      "step": 5480
    },
    {
      "epoch": 1.08,
      "learning_rate": 9.788621098380087e-05,
      "loss": 0.5616,
      "step": 5490
    },
    {
      "epoch": 1.09,
      "learning_rate": 9.783682338996445e-05,
      "loss": 0.5906,
      "step": 5500
    },
    {
      "epoch": 1.09,
      "learning_rate": 9.778743579612802e-05,
      "loss": 0.6046,
      "step": 5510
    },
    {
      "epoch": 1.09,
      "learning_rate": 9.773804820229158e-05,
      "loss": 0.5795,
      "step": 5520
    },
    {
      "epoch": 1.09,
      "learning_rate": 9.768866060845516e-05,
      "loss": 0.6262,
      "step": 5530
    },
    {
      "epoch": 1.09,
      "learning_rate": 9.763927301461874e-05,
      "loss": 0.6531,
      "step": 5540
    },
    {
      "epoch": 1.1,
      "learning_rate": 9.75898854207823e-05,
      "loss": 0.6423,
      "step": 5550
    },
    {
      "epoch": 1.1,
      "learning_rate": 9.754049782694588e-05,
      "loss": 0.6074,
      "step": 5560
    },
    {
      "epoch": 1.1,
      "learning_rate": 9.749111023310946e-05,
      "loss": 0.5798,
      "step": 5570
    },
    {
      "epoch": 1.1,
      "learning_rate": 9.744172263927301e-05,
      "loss": 0.5806,
      "step": 5580
    },
    {
      "epoch": 1.1,
      "learning_rate": 9.739233504543659e-05,
      "loss": 0.5655,
      "step": 5590
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.734294745160016e-05,
      "loss": 0.523,
      "step": 5600
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.729355985776374e-05,
      "loss": 0.566,
      "step": 5610
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.72441722639273e-05,
      "loss": 0.5482,
      "step": 5620
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.719478467009087e-05,
      "loss": 0.5899,
      "step": 5630
    },
    {
      "epoch": 1.11,
      "learning_rate": 9.714539707625445e-05,
      "loss": 0.5508,
      "step": 5640
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.709600948241803e-05,
      "loss": 0.5866,
      "step": 5650
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.704662188858158e-05,
      "loss": 0.5878,
      "step": 5660
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.699723429474516e-05,
      "loss": 0.5563,
      "step": 5670
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.694784670090875e-05,
      "loss": 0.6075,
      "step": 5680
    },
    {
      "epoch": 1.12,
      "learning_rate": 9.68984591070723e-05,
      "loss": 0.6032,
      "step": 5690
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.684907151323588e-05,
      "loss": 0.5908,
      "step": 5700
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.679968391939946e-05,
      "loss": 0.5863,
      "step": 5710
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.675029632556303e-05,
      "loss": 0.5858,
      "step": 5720
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.670090873172659e-05,
      "loss": 0.5929,
      "step": 5730
    },
    {
      "epoch": 1.13,
      "learning_rate": 9.665152113789017e-05,
      "loss": 0.5891,
      "step": 5740
    },
    {
      "epoch": 1.14,
      "learning_rate": 9.660213354405374e-05,
      "loss": 0.618,
      "step": 5750
    },
    {
      "epoch": 1.14,
      "learning_rate": 9.65527459502173e-05,
      "loss": 0.6104,
      "step": 5760
    },
    {
      "epoch": 1.14,
      "learning_rate": 9.650335835638089e-05,
      "loss": 0.588,
      "step": 5770
    },
    {
      "epoch": 1.14,
      "learning_rate": 9.645397076254445e-05,
      "loss": 0.6054,
      "step": 5780
    },
    {
      "epoch": 1.14,
      "learning_rate": 9.640458316870803e-05,
      "loss": 0.5846,
      "step": 5790
    },
    {
      "epoch": 1.15,
      "learning_rate": 9.635519557487159e-05,
      "loss": 0.6161,
      "step": 5800
    },
    {
      "epoch": 1.15,
      "learning_rate": 9.630580798103517e-05,
      "loss": 0.5995,
      "step": 5810
    },
    {
      "epoch": 1.15,
      "learning_rate": 9.625642038719875e-05,
      "loss": 0.5678,
      "step": 5820
    },
    {
      "epoch": 1.15,
      "learning_rate": 9.620703279336231e-05,
      "loss": 0.6255,
      "step": 5830
    },
    {
      "epoch": 1.15,
      "learning_rate": 9.615764519952588e-05,
      "loss": 0.5903,
      "step": 5840
    },
    {
      "epoch": 1.16,
      "learning_rate": 9.610825760568946e-05,
      "loss": 0.5584,
      "step": 5850
    },
    {
      "epoch": 1.16,
      "learning_rate": 9.605887001185303e-05,
      "loss": 0.5297,
      "step": 5860
    },
    {
      "epoch": 1.16,
      "learning_rate": 9.60094824180166e-05,
      "loss": 0.5777,
      "step": 5870
    },
    {
      "epoch": 1.16,
      "learning_rate": 9.596009482418017e-05,
      "loss": 0.6187,
      "step": 5880
    },
    {
      "epoch": 1.16,
      "learning_rate": 9.591070723034374e-05,
      "loss": 0.5935,
      "step": 5890
    },
    {
      "epoch": 1.17,
      "learning_rate": 9.586131963650732e-05,
      "loss": 0.5897,
      "step": 5900
    },
    {
      "epoch": 1.17,
      "learning_rate": 9.581193204267089e-05,
      "loss": 0.5581,
      "step": 5910
    },
    {
      "epoch": 1.17,
      "learning_rate": 9.576254444883445e-05,
      "loss": 0.6178,
      "step": 5920
    },
    {
      "epoch": 1.17,
      "learning_rate": 9.571315685499804e-05,
      "loss": 0.5879,
      "step": 5930
    },
    {
      "epoch": 1.17,
      "learning_rate": 9.56637692611616e-05,
      "loss": 0.6004,
      "step": 5940
    },
    {
      "epoch": 1.18,
      "learning_rate": 9.561438166732517e-05,
      "loss": 0.5726,
      "step": 5950
    },
    {
      "epoch": 1.18,
      "learning_rate": 9.556499407348875e-05,
      "loss": 0.5794,
      "step": 5960
    },
    {
      "epoch": 1.18,
      "learning_rate": 9.551560647965232e-05,
      "loss": 0.586,
      "step": 5970
    },
    {
      "epoch": 1.18,
      "learning_rate": 9.546621888581588e-05,
      "loss": 0.5863,
      "step": 5980
    },
    {
      "epoch": 1.18,
      "learning_rate": 9.541683129197946e-05,
      "loss": 0.6062,
      "step": 5990
    },
    {
      "epoch": 1.19,
      "learning_rate": 9.536744369814303e-05,
      "loss": 0.6103,
      "step": 6000
    },
    {
      "epoch": 1.19,
      "learning_rate": 9.531805610430661e-05,
      "loss": 0.6179,
      "step": 6010
    },
    {
      "epoch": 1.19,
      "learning_rate": 9.526866851047018e-05,
      "loss": 0.5564,
      "step": 6020
    },
    {
      "epoch": 1.19,
      "learning_rate": 9.521928091663374e-05,
      "loss": 0.5664,
      "step": 6030
    },
    {
      "epoch": 1.19,
      "learning_rate": 9.516989332279732e-05,
      "loss": 0.5681,
      "step": 6040
    },
    {
      "epoch": 1.2,
      "learning_rate": 9.512050572896089e-05,
      "loss": 0.5789,
      "step": 6050
    },
    {
      "epoch": 1.2,
      "learning_rate": 9.507111813512446e-05,
      "loss": 0.59,
      "step": 6060
    },
    {
      "epoch": 1.2,
      "learning_rate": 9.502173054128804e-05,
      "loss": 0.5384,
      "step": 6070
    },
    {
      "epoch": 1.2,
      "learning_rate": 9.49723429474516e-05,
      "loss": 0.5677,
      "step": 6080
    },
    {
      "epoch": 1.2,
      "learning_rate": 9.492295535361517e-05,
      "loss": 0.6119,
      "step": 6090
    },
    {
      "epoch": 1.2,
      "learning_rate": 9.487356775977875e-05,
      "loss": 0.5626,
      "step": 6100
    },
    {
      "epoch": 1.21,
      "learning_rate": 9.482418016594232e-05,
      "loss": 0.5925,
      "step": 6110
    },
    {
      "epoch": 1.21,
      "learning_rate": 9.477479257210588e-05,
      "loss": 0.5993,
      "step": 6120
    },
    {
      "epoch": 1.21,
      "learning_rate": 9.472540497826946e-05,
      "loss": 0.5264,
      "step": 6130
    },
    {
      "epoch": 1.21,
      "learning_rate": 9.467601738443303e-05,
      "loss": 0.5545,
      "step": 6140
    },
    {
      "epoch": 1.21,
      "learning_rate": 9.462662979059661e-05,
      "loss": 0.5822,
      "step": 6150
    },
    {
      "epoch": 1.22,
      "learning_rate": 9.457724219676018e-05,
      "loss": 0.5915,
      "step": 6160
    },
    {
      "epoch": 1.22,
      "learning_rate": 9.452785460292374e-05,
      "loss": 0.5689,
      "step": 6170
    },
    {
      "epoch": 1.22,
      "learning_rate": 9.447846700908733e-05,
      "loss": 0.5721,
      "step": 6180
    },
    {
      "epoch": 1.22,
      "learning_rate": 9.442907941525089e-05,
      "loss": 0.5692,
      "step": 6190
    },
    {
      "epoch": 1.22,
      "learning_rate": 9.437969182141446e-05,
      "loss": 0.5578,
      "step": 6200
    },
    {
      "epoch": 1.23,
      "learning_rate": 9.433030422757804e-05,
      "loss": 0.5678,
      "step": 6210
    },
    {
      "epoch": 1.23,
      "learning_rate": 9.42809166337416e-05,
      "loss": 0.5738,
      "step": 6220
    },
    {
      "epoch": 1.23,
      "learning_rate": 9.423152903990517e-05,
      "loss": 0.5563,
      "step": 6230
    },
    {
      "epoch": 1.23,
      "learning_rate": 9.418214144606875e-05,
      "loss": 0.5703,
      "step": 6240
    },
    {
      "epoch": 1.23,
      "learning_rate": 9.413275385223232e-05,
      "loss": 0.5612,
      "step": 6250
    },
    {
      "epoch": 1.24,
      "learning_rate": 9.40833662583959e-05,
      "loss": 0.5578,
      "step": 6260
    },
    {
      "epoch": 1.24,
      "learning_rate": 9.403397866455947e-05,
      "loss": 0.609,
      "step": 6270
    },
    {
      "epoch": 1.24,
      "learning_rate": 9.398459107072303e-05,
      "loss": 0.5784,
      "step": 6280
    },
    {
      "epoch": 1.24,
      "learning_rate": 9.393520347688661e-05,
      "loss": 0.5322,
      "step": 6290
    },
    {
      "epoch": 1.24,
      "learning_rate": 9.38858158830502e-05,
      "loss": 0.5788,
      "step": 6300
    },
    {
      "epoch": 1.25,
      "learning_rate": 9.383642828921375e-05,
      "loss": 0.5519,
      "step": 6310
    },
    {
      "epoch": 1.25,
      "learning_rate": 9.378704069537733e-05,
      "loss": 0.5735,
      "step": 6320
    },
    {
      "epoch": 1.25,
      "learning_rate": 9.373765310154091e-05,
      "loss": 0.5542,
      "step": 6330
    },
    {
      "epoch": 1.25,
      "learning_rate": 9.368826550770446e-05,
      "loss": 0.5588,
      "step": 6340
    },
    {
      "epoch": 1.25,
      "learning_rate": 9.363887791386804e-05,
      "loss": 0.6296,
      "step": 6350
    },
    {
      "epoch": 1.26,
      "learning_rate": 9.358949032003162e-05,
      "loss": 0.592,
      "step": 6360
    },
    {
      "epoch": 1.26,
      "learning_rate": 9.354010272619519e-05,
      "loss": 0.54,
      "step": 6370
    },
    {
      "epoch": 1.26,
      "learning_rate": 9.349071513235875e-05,
      "loss": 0.5527,
      "step": 6380
    },
    {
      "epoch": 1.26,
      "learning_rate": 9.344132753852232e-05,
      "loss": 0.5999,
      "step": 6390
    },
    {
      "epoch": 1.26,
      "learning_rate": 9.33919399446859e-05,
      "loss": 0.5323,
      "step": 6400
    },
    {
      "epoch": 1.27,
      "learning_rate": 9.334255235084947e-05,
      "loss": 0.5933,
      "step": 6410
    },
    {
      "epoch": 1.27,
      "learning_rate": 9.329316475701303e-05,
      "loss": 0.5564,
      "step": 6420
    },
    {
      "epoch": 1.27,
      "learning_rate": 9.324377716317662e-05,
      "loss": 0.6013,
      "step": 6430
    },
    {
      "epoch": 1.27,
      "learning_rate": 9.31943895693402e-05,
      "loss": 0.5646,
      "step": 6440
    },
    {
      "epoch": 1.27,
      "learning_rate": 9.314500197550375e-05,
      "loss": 0.5512,
      "step": 6450
    },
    {
      "epoch": 1.28,
      "learning_rate": 9.309561438166733e-05,
      "loss": 0.5529,
      "step": 6460
    },
    {
      "epoch": 1.28,
      "learning_rate": 9.304622678783091e-05,
      "loss": 0.6227,
      "step": 6470
    },
    {
      "epoch": 1.28,
      "learning_rate": 9.299683919399448e-05,
      "loss": 0.5761,
      "step": 6480
    },
    {
      "epoch": 1.28,
      "learning_rate": 9.294745160015804e-05,
      "loss": 0.5036,
      "step": 6490
    },
    {
      "epoch": 1.28,
      "learning_rate": 9.289806400632162e-05,
      "loss": 0.5148,
      "step": 6500
    },
    {
      "epoch": 1.29,
      "learning_rate": 9.284867641248519e-05,
      "loss": 0.5503,
      "step": 6510
    },
    {
      "epoch": 1.29,
      "learning_rate": 9.279928881864876e-05,
      "loss": 0.5882,
      "step": 6520
    },
    {
      "epoch": 1.29,
      "learning_rate": 9.274990122481234e-05,
      "loss": 0.5803,
      "step": 6530
    },
    {
      "epoch": 1.29,
      "learning_rate": 9.27005136309759e-05,
      "loss": 0.5741,
      "step": 6540
    },
    {
      "epoch": 1.29,
      "learning_rate": 9.265112603713948e-05,
      "loss": 0.6181,
      "step": 6550
    },
    {
      "epoch": 1.3,
      "learning_rate": 9.260173844330304e-05,
      "loss": 0.594,
      "step": 6560
    },
    {
      "epoch": 1.3,
      "learning_rate": 9.255235084946662e-05,
      "loss": 0.5033,
      "step": 6570
    },
    {
      "epoch": 1.3,
      "learning_rate": 9.25029632556302e-05,
      "loss": 0.6282,
      "step": 6580
    },
    {
      "epoch": 1.3,
      "learning_rate": 9.245357566179375e-05,
      "loss": 0.5376,
      "step": 6590
    },
    {
      "epoch": 1.3,
      "learning_rate": 9.240418806795733e-05,
      "loss": 0.5705,
      "step": 6600
    },
    {
      "epoch": 1.31,
      "learning_rate": 9.235480047412091e-05,
      "loss": 0.5539,
      "step": 6610
    },
    {
      "epoch": 1.31,
      "learning_rate": 9.230541288028448e-05,
      "loss": 0.5966,
      "step": 6620
    },
    {
      "epoch": 1.31,
      "learning_rate": 9.225602528644804e-05,
      "loss": 0.5605,
      "step": 6630
    },
    {
      "epoch": 1.31,
      "learning_rate": 9.220663769261162e-05,
      "loss": 0.6484,
      "step": 6640
    },
    {
      "epoch": 1.31,
      "learning_rate": 9.215725009877519e-05,
      "loss": 0.5461,
      "step": 6650
    },
    {
      "epoch": 1.32,
      "learning_rate": 9.210786250493877e-05,
      "loss": 0.6082,
      "step": 6660
    },
    {
      "epoch": 1.32,
      "learning_rate": 9.205847491110234e-05,
      "loss": 0.5517,
      "step": 6670
    },
    {
      "epoch": 1.32,
      "learning_rate": 9.20090873172659e-05,
      "loss": 0.5512,
      "step": 6680
    },
    {
      "epoch": 1.32,
      "learning_rate": 9.195969972342949e-05,
      "loss": 0.4835,
      "step": 6690
    },
    {
      "epoch": 1.32,
      "learning_rate": 9.191031212959305e-05,
      "loss": 0.5864,
      "step": 6700
    },
    {
      "epoch": 1.33,
      "learning_rate": 9.186092453575662e-05,
      "loss": 0.5513,
      "step": 6710
    },
    {
      "epoch": 1.33,
      "learning_rate": 9.18115369419202e-05,
      "loss": 0.5527,
      "step": 6720
    },
    {
      "epoch": 1.33,
      "learning_rate": 9.176214934808377e-05,
      "loss": 0.5251,
      "step": 6730
    },
    {
      "epoch": 1.33,
      "learning_rate": 9.171276175424733e-05,
      "loss": 0.5467,
      "step": 6740
    },
    {
      "epoch": 1.33,
      "learning_rate": 9.166337416041091e-05,
      "loss": 0.5539,
      "step": 6750
    },
    {
      "epoch": 1.34,
      "learning_rate": 9.161398656657448e-05,
      "loss": 0.5269,
      "step": 6760
    },
    {
      "epoch": 1.34,
      "learning_rate": 9.156459897273805e-05,
      "loss": 0.5198,
      "step": 6770
    },
    {
      "epoch": 1.34,
      "learning_rate": 9.151521137890163e-05,
      "loss": 0.5678,
      "step": 6780
    },
    {
      "epoch": 1.34,
      "learning_rate": 9.146582378506519e-05,
      "loss": 0.6333,
      "step": 6790
    },
    {
      "epoch": 1.34,
      "learning_rate": 9.141643619122877e-05,
      "loss": 0.5445,
      "step": 6800
    },
    {
      "epoch": 1.35,
      "learning_rate": 9.136704859739234e-05,
      "loss": 0.5594,
      "step": 6810
    },
    {
      "epoch": 1.35,
      "learning_rate": 9.131766100355591e-05,
      "loss": 0.5557,
      "step": 6820
    },
    {
      "epoch": 1.35,
      "learning_rate": 9.126827340971949e-05,
      "loss": 0.5797,
      "step": 6830
    },
    {
      "epoch": 1.35,
      "learning_rate": 9.121888581588305e-05,
      "loss": 0.5981,
      "step": 6840
    },
    {
      "epoch": 1.35,
      "learning_rate": 9.116949822204662e-05,
      "loss": 0.535,
      "step": 6850
    },
    {
      "epoch": 1.35,
      "learning_rate": 9.11201106282102e-05,
      "loss": 0.5275,
      "step": 6860
    },
    {
      "epoch": 1.36,
      "learning_rate": 9.107072303437377e-05,
      "loss": 0.5531,
      "step": 6870
    },
    {
      "epoch": 1.36,
      "learning_rate": 9.102133544053733e-05,
      "loss": 0.5669,
      "step": 6880
    },
    {
      "epoch": 1.36,
      "learning_rate": 9.097194784670091e-05,
      "loss": 0.5635,
      "step": 6890
    },
    {
      "epoch": 1.36,
      "learning_rate": 9.092256025286448e-05,
      "loss": 0.5911,
      "step": 6900
    },
    {
      "epoch": 1.36,
      "learning_rate": 9.087317265902806e-05,
      "loss": 0.5518,
      "step": 6910
    },
    {
      "epoch": 1.37,
      "learning_rate": 9.082378506519163e-05,
      "loss": 0.6044,
      "step": 6920
    },
    {
      "epoch": 1.37,
      "learning_rate": 9.07743974713552e-05,
      "loss": 0.5872,
      "step": 6930
    },
    {
      "epoch": 1.37,
      "learning_rate": 9.072500987751878e-05,
      "loss": 0.5214,
      "step": 6940
    },
    {
      "epoch": 1.37,
      "learning_rate": 9.067562228368236e-05,
      "loss": 0.5623,
      "step": 6950
    },
    {
      "epoch": 1.37,
      "learning_rate": 9.062623468984591e-05,
      "loss": 0.5604,
      "step": 6960
    },
    {
      "epoch": 1.38,
      "learning_rate": 9.057684709600949e-05,
      "loss": 0.5896,
      "step": 6970
    },
    {
      "epoch": 1.38,
      "learning_rate": 9.052745950217306e-05,
      "loss": 0.5493,
      "step": 6980
    },
    {
      "epoch": 1.38,
      "learning_rate": 9.047807190833662e-05,
      "loss": 0.4864,
      "step": 6990
    },
    {
      "epoch": 1.38,
      "learning_rate": 9.04286843145002e-05,
      "loss": 0.5621,
      "step": 7000
    }
  ],
  "logging_steps": 10,
  "max_steps": 25310,
  "num_train_epochs": 5,
  "save_steps": 7000,
  "total_flos": 1.8202371090127258e+17,
  "trial_name": null,
  "trial_params": null
}