{
  "best_metric": 0.702185081437324,
  "best_model_checkpoint": "hq_fer2013/checkpoint-3980",
  "epoch": 12.999372253609542,
  "global_step": 5174,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.03,
      "learning_rate": 1.9305019305019306e-07,
      "loss": 1.9428,
      "step": 10
    },
    {
      "epoch": 0.05,
      "learning_rate": 3.8610038610038613e-07,
      "loss": 1.9495,
      "step": 20
    },
    {
      "epoch": 0.08,
      "learning_rate": 5.791505791505792e-07,
      "loss": 1.9401,
      "step": 30
    },
    {
      "epoch": 0.1,
      "learning_rate": 7.722007722007723e-07,
      "loss": 1.9418,
      "step": 40
    },
    {
      "epoch": 0.13,
      "learning_rate": 9.652509652509653e-07,
      "loss": 1.9337,
      "step": 50
    },
    {
      "epoch": 0.15,
      "learning_rate": 1.1583011583011585e-06,
      "loss": 1.9323,
      "step": 60
    },
    {
      "epoch": 0.18,
      "learning_rate": 1.3513513513513515e-06,
      "loss": 1.9271,
      "step": 70
    },
    {
      "epoch": 0.2,
      "learning_rate": 1.5444015444015445e-06,
      "loss": 1.928,
      "step": 80
    },
    {
      "epoch": 0.23,
      "learning_rate": 1.7374517374517377e-06,
      "loss": 1.9251,
      "step": 90
    },
    {
      "epoch": 0.25,
      "learning_rate": 1.9305019305019305e-06,
      "loss": 1.9209,
      "step": 100
    },
    {
      "epoch": 0.28,
      "learning_rate": 2.1235521235521236e-06,
      "loss": 1.9163,
      "step": 110
    },
    {
      "epoch": 0.3,
      "learning_rate": 2.316602316602317e-06,
      "loss": 1.9107,
      "step": 120
    },
    {
      "epoch": 0.33,
      "learning_rate": 2.50965250965251e-06,
      "loss": 1.9087,
      "step": 130
    },
    {
      "epoch": 0.35,
      "learning_rate": 2.702702702702703e-06,
      "loss": 1.8944,
      "step": 140
    },
    {
      "epoch": 0.38,
      "learning_rate": 2.8957528957528956e-06,
      "loss": 1.8916,
      "step": 150
    },
    {
      "epoch": 0.4,
      "learning_rate": 3.088803088803089e-06,
      "loss": 1.8823,
      "step": 160
    },
    {
      "epoch": 0.43,
      "learning_rate": 3.281853281853282e-06,
      "loss": 1.8688,
      "step": 170
    },
    {
      "epoch": 0.45,
      "learning_rate": 3.4749034749034755e-06,
      "loss": 1.8619,
      "step": 180
    },
    {
      "epoch": 0.48,
      "learning_rate": 3.667953667953668e-06,
      "loss": 1.846,
      "step": 190
    },
    {
      "epoch": 0.5,
      "learning_rate": 3.861003861003861e-06,
      "loss": 1.8379,
      "step": 200
    },
    {
      "epoch": 0.53,
      "learning_rate": 4.0540540540540545e-06,
      "loss": 1.8181,
      "step": 210
    },
    {
      "epoch": 0.55,
      "learning_rate": 4.247104247104247e-06,
      "loss": 1.7913,
      "step": 220
    },
    {
      "epoch": 0.58,
      "learning_rate": 4.4401544401544405e-06,
      "loss": 1.771,
      "step": 230
    },
    {
      "epoch": 0.6,
      "learning_rate": 4.633204633204634e-06,
      "loss": 1.7484,
      "step": 240
    },
    {
      "epoch": 0.63,
      "learning_rate": 4.8262548262548266e-06,
      "loss": 1.7147,
      "step": 250
    },
    {
      "epoch": 0.65,
      "learning_rate": 5.01930501930502e-06,
      "loss": 1.6895,
      "step": 260
    },
    {
      "epoch": 0.68,
      "learning_rate": 5.212355212355213e-06,
      "loss": 1.6661,
      "step": 270
    },
    {
      "epoch": 0.7,
      "learning_rate": 5.405405405405406e-06,
      "loss": 1.6419,
      "step": 280
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.598455598455599e-06,
      "loss": 1.6045,
      "step": 290
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.791505791505791e-06,
      "loss": 1.5883,
      "step": 300
    },
    {
      "epoch": 0.78,
      "learning_rate": 5.9845559845559855e-06,
      "loss": 1.5143,
      "step": 310
    },
    {
      "epoch": 0.8,
      "learning_rate": 6.177606177606178e-06,
      "loss": 1.494,
      "step": 320
    },
    {
      "epoch": 0.83,
      "learning_rate": 6.370656370656371e-06,
      "loss": 1.5003,
      "step": 330
    },
    {
      "epoch": 0.85,
      "learning_rate": 6.563706563706564e-06,
      "loss": 1.4563,
      "step": 340
    },
    {
      "epoch": 0.88,
      "learning_rate": 6.7567567567567575e-06,
      "loss": 1.4393,
      "step": 350
    },
    {
      "epoch": 0.9,
      "learning_rate": 6.949806949806951e-06,
      "loss": 1.4276,
      "step": 360
    },
    {
      "epoch": 0.93,
      "learning_rate": 7.1428571428571436e-06,
      "loss": 1.38,
      "step": 370
    },
    {
      "epoch": 0.95,
      "learning_rate": 7.335907335907336e-06,
      "loss": 1.3721,
      "step": 380
    },
    {
      "epoch": 0.98,
      "learning_rate": 7.5289575289575304e-06,
      "loss": 1.3081,
      "step": 390
    },
    {
      "epoch": 1.0,
      "eval_accuracy": 0.5554857950086328,
      "eval_f1": 0.5136862225473102,
      "eval_loss": 1.313186764717102,
      "eval_precision": 0.5078982247553088,
      "eval_recall": 0.5554857950086328,
      "eval_runtime": 179.074,
      "eval_samples_per_second": 35.577,
      "eval_steps_per_second": 1.117,
      "step": 398
    },
    {
      "epoch": 1.01,
      "learning_rate": 7.722007722007722e-06,
      "loss": 1.3569,
      "step": 400
    },
    {
      "epoch": 1.03,
      "learning_rate": 7.915057915057915e-06,
      "loss": 1.3076,
      "step": 410
    },
    {
      "epoch": 1.06,
      "learning_rate": 8.108108108108109e-06,
      "loss": 1.2964,
      "step": 420
    },
    {
      "epoch": 1.08,
      "learning_rate": 8.301158301158302e-06,
      "loss": 1.266,
      "step": 430
    },
    {
      "epoch": 1.11,
      "learning_rate": 8.494208494208494e-06,
      "loss": 1.2416,
      "step": 440
    },
    {
      "epoch": 1.13,
      "learning_rate": 8.687258687258689e-06,
      "loss": 1.2614,
      "step": 450
    },
    {
      "epoch": 1.16,
      "learning_rate": 8.880308880308881e-06,
      "loss": 1.2231,
      "step": 460
    },
    {
      "epoch": 1.18,
      "learning_rate": 9.073359073359074e-06,
      "loss": 1.2225,
      "step": 470
    },
    {
      "epoch": 1.21,
      "learning_rate": 9.266409266409268e-06,
      "loss": 1.2033,
      "step": 480
    },
    {
      "epoch": 1.23,
      "learning_rate": 9.45945945945946e-06,
      "loss": 1.1699,
      "step": 490
    },
    {
      "epoch": 1.26,
      "learning_rate": 9.652509652509653e-06,
      "loss": 1.1767,
      "step": 500
    },
    {
      "epoch": 1.28,
      "learning_rate": 9.845559845559847e-06,
      "loss": 1.1941,
      "step": 510
    },
    {
      "epoch": 1.31,
      "learning_rate": 9.995704467353953e-06,
      "loss": 1.1508,
      "step": 520
    },
    {
      "epoch": 1.33,
      "learning_rate": 9.974226804123713e-06,
      "loss": 1.182,
      "step": 530
    },
    {
      "epoch": 1.36,
      "learning_rate": 9.952749140893471e-06,
      "loss": 1.1785,
      "step": 540
    },
    {
      "epoch": 1.38,
      "learning_rate": 9.931271477663231e-06,
      "loss": 1.1754,
      "step": 550
    },
    {
      "epoch": 1.41,
      "learning_rate": 9.90979381443299e-06,
      "loss": 1.1188,
      "step": 560
    },
    {
      "epoch": 1.43,
      "learning_rate": 9.88831615120275e-06,
      "loss": 1.1334,
      "step": 570
    },
    {
      "epoch": 1.46,
      "learning_rate": 9.86683848797251e-06,
      "loss": 1.0836,
      "step": 580
    },
    {
      "epoch": 1.48,
      "learning_rate": 9.84536082474227e-06,
      "loss": 1.1021,
      "step": 590
    },
    {
      "epoch": 1.51,
      "learning_rate": 9.823883161512028e-06,
      "loss": 1.0752,
      "step": 600
    },
    {
      "epoch": 1.53,
      "learning_rate": 9.802405498281788e-06,
      "loss": 1.0908,
      "step": 610
    },
    {
      "epoch": 1.56,
      "learning_rate": 9.780927835051546e-06,
      "loss": 1.1108,
      "step": 620
    },
    {
      "epoch": 1.58,
      "learning_rate": 9.759450171821306e-06,
      "loss": 1.0935,
      "step": 630
    },
    {
      "epoch": 1.61,
      "learning_rate": 9.737972508591066e-06,
      "loss": 1.0831,
      "step": 640
    },
    {
      "epoch": 1.63,
      "learning_rate": 9.716494845360826e-06,
      "loss": 1.0735,
      "step": 650
    },
    {
      "epoch": 1.66,
      "learning_rate": 9.695017182130586e-06,
      "loss": 1.074,
      "step": 660
    },
    {
      "epoch": 1.68,
      "learning_rate": 9.673539518900344e-06,
      "loss": 1.0652,
      "step": 670
    },
    {
      "epoch": 1.71,
      "learning_rate": 9.652061855670104e-06,
      "loss": 1.0819,
      "step": 680
    },
    {
      "epoch": 1.73,
      "learning_rate": 9.630584192439863e-06,
      "loss": 1.0904,
      "step": 690
    },
    {
      "epoch": 1.76,
      "learning_rate": 9.609106529209623e-06,
      "loss": 1.0534,
      "step": 700
    },
    {
      "epoch": 1.78,
      "learning_rate": 9.587628865979383e-06,
      "loss": 1.0237,
      "step": 710
    },
    {
      "epoch": 1.81,
      "learning_rate": 9.566151202749143e-06,
      "loss": 1.0358,
      "step": 720
    },
    {
      "epoch": 1.83,
      "learning_rate": 9.544673539518901e-06,
      "loss": 1.0807,
      "step": 730
    },
    {
      "epoch": 1.86,
      "learning_rate": 9.52319587628866e-06,
      "loss": 1.0371,
      "step": 740
    },
    {
      "epoch": 1.88,
      "learning_rate": 9.501718213058419e-06,
      "loss": 1.0463,
      "step": 750
    },
    {
      "epoch": 1.91,
      "learning_rate": 9.480240549828179e-06,
      "loss": 1.0206,
      "step": 760
    },
    {
      "epoch": 1.93,
      "learning_rate": 9.458762886597939e-06,
      "loss": 1.0214,
      "step": 770
    },
    {
      "epoch": 1.96,
      "learning_rate": 9.437285223367698e-06,
      "loss": 1.0011,
      "step": 780
    },
    {
      "epoch": 1.98,
      "learning_rate": 9.415807560137458e-06,
      "loss": 0.991,
      "step": 790
    },
    {
      "epoch": 2.0,
      "eval_accuracy": 0.6331816041437764,
      "eval_f1": 0.6153430259954497,
      "eval_loss": 1.0141429901123047,
      "eval_precision": 0.6356481524669472,
      "eval_recall": 0.6331816041437764,
      "eval_runtime": 178.3829,
      "eval_samples_per_second": 35.715,
      "eval_steps_per_second": 1.121,
      "step": 796
    },
    {
      "epoch": 2.01,
      "learning_rate": 9.394329896907216e-06,
      "loss": 1.038,
      "step": 800
    },
    {
      "epoch": 2.04,
      "learning_rate": 9.372852233676976e-06,
      "loss": 0.9727,
      "step": 810
    },
    {
      "epoch": 2.06,
      "learning_rate": 9.351374570446736e-06,
      "loss": 1.0115,
      "step": 820
    },
    {
      "epoch": 2.09,
      "learning_rate": 9.329896907216496e-06,
      "loss": 0.9684,
      "step": 830
    },
    {
      "epoch": 2.11,
      "learning_rate": 9.308419243986256e-06,
      "loss": 0.9926,
      "step": 840
    },
    {
      "epoch": 2.14,
      "learning_rate": 9.286941580756015e-06,
      "loss": 0.959,
      "step": 850
    },
    {
      "epoch": 2.16,
      "learning_rate": 9.265463917525774e-06,
      "loss": 0.9983,
      "step": 860
    },
    {
      "epoch": 2.19,
      "learning_rate": 9.243986254295533e-06,
      "loss": 1.0044,
      "step": 870
    },
    {
      "epoch": 2.21,
      "learning_rate": 9.222508591065293e-06,
      "loss": 0.9497,
      "step": 880
    },
    {
      "epoch": 2.24,
      "learning_rate": 9.201030927835051e-06,
      "loss": 0.9326,
      "step": 890
    },
    {
      "epoch": 2.26,
      "learning_rate": 9.179553264604811e-06,
      "loss": 0.9778,
      "step": 900
    },
    {
      "epoch": 2.29,
      "learning_rate": 9.158075601374571e-06,
      "loss": 0.9504,
      "step": 910
    },
    {
      "epoch": 2.31,
      "learning_rate": 9.136597938144331e-06,
      "loss": 0.9304,
      "step": 920
    },
    {
      "epoch": 2.34,
      "learning_rate": 9.11512027491409e-06,
      "loss": 0.9475,
      "step": 930
    },
    {
      "epoch": 2.36,
      "learning_rate": 9.093642611683849e-06,
      "loss": 0.949,
      "step": 940
    },
    {
      "epoch": 2.39,
      "learning_rate": 9.072164948453609e-06,
      "loss": 0.9174,
      "step": 950
    },
    {
      "epoch": 2.41,
      "learning_rate": 9.050687285223368e-06,
      "loss": 0.9649,
      "step": 960
    },
    {
      "epoch": 2.44,
      "learning_rate": 9.029209621993128e-06,
      "loss": 0.9551,
      "step": 970
    },
    {
      "epoch": 2.46,
      "learning_rate": 9.007731958762888e-06,
      "loss": 0.9764,
      "step": 980
    },
    {
      "epoch": 2.49,
      "learning_rate": 8.986254295532646e-06,
      "loss": 0.9489,
      "step": 990
    },
    {
      "epoch": 2.51,
      "learning_rate": 8.964776632302406e-06,
      "loss": 0.9555,
      "step": 1000
    },
    {
      "epoch": 2.54,
      "learning_rate": 8.943298969072166e-06,
      "loss": 0.935,
      "step": 1010
    },
    {
      "epoch": 2.56,
      "learning_rate": 8.921821305841926e-06,
      "loss": 0.8946,
      "step": 1020
    },
    {
      "epoch": 2.59,
      "learning_rate": 8.900343642611684e-06,
      "loss": 0.962,
      "step": 1030
    },
    {
      "epoch": 2.61,
      "learning_rate": 8.878865979381444e-06,
      "loss": 0.9303,
      "step": 1040
    },
    {
      "epoch": 2.64,
      "learning_rate": 8.857388316151203e-06,
      "loss": 0.9676,
      "step": 1050
    },
    {
      "epoch": 2.66,
      "learning_rate": 8.835910652920963e-06,
      "loss": 0.9362,
      "step": 1060
    },
    {
      "epoch": 2.69,
      "learning_rate": 8.814432989690721e-06,
      "loss": 0.9708,
      "step": 1070
    },
    {
      "epoch": 2.71,
      "learning_rate": 8.792955326460481e-06,
      "loss": 0.9162,
      "step": 1080
    },
    {
      "epoch": 2.74,
      "learning_rate": 8.771477663230241e-06,
      "loss": 0.9267,
      "step": 1090
    },
    {
      "epoch": 2.76,
      "learning_rate": 8.750000000000001e-06,
      "loss": 0.9017,
      "step": 1100
    },
    {
      "epoch": 2.79,
      "learning_rate": 8.72852233676976e-06,
      "loss": 0.947,
      "step": 1110
    },
    {
      "epoch": 2.81,
      "learning_rate": 8.70704467353952e-06,
      "loss": 0.9243,
      "step": 1120
    },
    {
      "epoch": 2.84,
      "learning_rate": 8.685567010309279e-06,
      "loss": 0.8907,
      "step": 1130
    },
    {
      "epoch": 2.86,
      "learning_rate": 8.664089347079039e-06,
      "loss": 0.9137,
      "step": 1140
    },
    {
      "epoch": 2.89,
      "learning_rate": 8.642611683848798e-06,
      "loss": 0.9264,
      "step": 1150
    },
    {
      "epoch": 2.91,
      "learning_rate": 8.621134020618558e-06,
      "loss": 0.8905,
      "step": 1160
    },
    {
      "epoch": 2.94,
      "learning_rate": 8.599656357388318e-06,
      "loss": 0.9139,
      "step": 1170
    },
    {
      "epoch": 2.96,
      "learning_rate": 8.578178694158076e-06,
      "loss": 0.924,
      "step": 1180
    },
    {
      "epoch": 2.99,
      "learning_rate": 8.556701030927836e-06,
      "loss": 0.9099,
      "step": 1190
    },
    {
      "epoch": 3.0,
      "eval_accuracy": 0.6681839585622351,
      "eval_f1": 0.663091861602691,
      "eval_loss": 0.9256689548492432,
      "eval_precision": 0.6677315596229765,
      "eval_recall": 0.6681839585622351,
      "eval_runtime": 176.3264,
      "eval_samples_per_second": 36.132,
      "eval_steps_per_second": 1.134,
      "step": 1194
    },
    {
      "epoch": 3.02,
      "learning_rate": 8.535223367697594e-06,
      "loss": 0.8771,
      "step": 1200
    },
    {
      "epoch": 3.04,
      "learning_rate": 8.513745704467354e-06,
      "loss": 0.8795,
      "step": 1210
    },
    {
      "epoch": 3.07,
      "learning_rate": 8.492268041237114e-06,
      "loss": 0.915,
      "step": 1220
    },
    {
      "epoch": 3.09,
      "learning_rate": 8.470790378006874e-06,
      "loss": 0.8286,
      "step": 1230
    },
    {
      "epoch": 3.12,
      "learning_rate": 8.449312714776633e-06,
      "loss": 0.8646,
      "step": 1240
    },
    {
      "epoch": 3.14,
      "learning_rate": 8.427835051546393e-06,
      "loss": 0.8605,
      "step": 1250
    },
    {
      "epoch": 3.17,
      "learning_rate": 8.406357388316151e-06,
      "loss": 0.8941,
      "step": 1260
    },
    {
      "epoch": 3.19,
      "learning_rate": 8.384879725085911e-06,
      "loss": 0.8508,
      "step": 1270
    },
    {
      "epoch": 3.22,
      "learning_rate": 8.363402061855671e-06,
      "loss": 0.8926,
      "step": 1280
    },
    {
      "epoch": 3.24,
      "learning_rate": 8.34192439862543e-06,
      "loss": 0.8806,
      "step": 1290
    },
    {
      "epoch": 3.27,
      "learning_rate": 8.32044673539519e-06,
      "loss": 0.8458,
      "step": 1300
    },
    {
      "epoch": 3.29,
      "learning_rate": 8.29896907216495e-06,
      "loss": 0.8405,
      "step": 1310
    },
    {
      "epoch": 3.32,
      "learning_rate": 8.277491408934709e-06,
      "loss": 0.8426,
      "step": 1320
    },
    {
      "epoch": 3.34,
      "learning_rate": 8.256013745704467e-06,
      "loss": 0.8938,
      "step": 1330
    },
    {
      "epoch": 3.37,
      "learning_rate": 8.234536082474227e-06,
      "loss": 0.8767,
      "step": 1340
    },
    {
      "epoch": 3.39,
      "learning_rate": 8.213058419243986e-06,
      "loss": 0.8458,
      "step": 1350
    },
    {
      "epoch": 3.42,
      "learning_rate": 8.191580756013746e-06,
      "loss": 0.876,
      "step": 1360
    },
    {
      "epoch": 3.44,
      "learning_rate": 8.170103092783506e-06,
      "loss": 0.9005,
      "step": 1370
    },
    {
      "epoch": 3.47,
      "learning_rate": 8.148625429553266e-06,
      "loss": 0.8363,
      "step": 1380
    },
    {
      "epoch": 3.49,
      "learning_rate": 8.127147766323024e-06,
      "loss": 0.8437,
      "step": 1390
    },
    {
      "epoch": 3.52,
      "learning_rate": 8.105670103092784e-06,
      "loss": 0.8423,
      "step": 1400
    },
    {
      "epoch": 3.54,
      "learning_rate": 8.084192439862544e-06,
      "loss": 0.8654,
      "step": 1410
    },
    {
      "epoch": 3.57,
      "learning_rate": 8.062714776632303e-06,
      "loss": 0.8615,
      "step": 1420
    },
    {
      "epoch": 3.59,
      "learning_rate": 8.041237113402063e-06,
      "loss": 0.8762,
      "step": 1430
    },
    {
      "epoch": 3.62,
      "learning_rate": 8.019759450171823e-06,
      "loss": 0.8198,
      "step": 1440
    },
    {
      "epoch": 3.64,
      "learning_rate": 7.998281786941581e-06,
      "loss": 0.8521,
      "step": 1450
    },
    {
      "epoch": 3.67,
      "learning_rate": 7.976804123711341e-06,
      "loss": 0.8718,
      "step": 1460
    },
    {
      "epoch": 3.69,
      "learning_rate": 7.9553264604811e-06,
      "loss": 0.8447,
      "step": 1470
    },
    {
      "epoch": 3.72,
      "learning_rate": 7.933848797250859e-06,
      "loss": 0.8431,
      "step": 1480
    },
    {
      "epoch": 3.74,
      "learning_rate": 7.912371134020619e-06,
      "loss": 0.8561,
      "step": 1490
    },
    {
      "epoch": 3.77,
      "learning_rate": 7.890893470790379e-06,
      "loss": 0.8574,
      "step": 1500
    },
    {
      "epoch": 3.79,
      "learning_rate": 7.869415807560138e-06,
      "loss": 0.8346,
      "step": 1510
    },
    {
      "epoch": 3.82,
      "learning_rate": 7.847938144329897e-06,
      "loss": 0.8444,
      "step": 1520
    },
    {
      "epoch": 3.84,
      "learning_rate": 7.826460481099656e-06,
      "loss": 0.8253,
      "step": 1530
    },
    {
      "epoch": 3.87,
      "learning_rate": 7.804982817869416e-06,
      "loss": 0.794,
      "step": 1540
    },
    {
      "epoch": 3.89,
      "learning_rate": 7.783505154639176e-06,
      "loss": 0.8449,
      "step": 1550
    },
    {
      "epoch": 3.92,
      "learning_rate": 7.762027491408936e-06,
      "loss": 0.8291,
      "step": 1560
    },
    {
      "epoch": 3.94,
      "learning_rate": 7.740549828178696e-06,
      "loss": 0.8166,
      "step": 1570
    },
    {
      "epoch": 3.97,
      "learning_rate": 7.719072164948454e-06,
      "loss": 0.7915,
      "step": 1580
    },
    {
      "epoch": 3.99,
      "learning_rate": 7.697594501718214e-06,
      "loss": 0.8306,
      "step": 1590
    },
    {
      "epoch": 4.0,
      "eval_accuracy": 0.6765029037827657,
      "eval_f1": 0.6746608981965966,
      "eval_loss": 0.8831803202629089,
      "eval_precision": 0.6837870094377039,
      "eval_recall": 0.6765029037827657,
      "eval_runtime": 175.6547,
      "eval_samples_per_second": 36.27,
      "eval_steps_per_second": 1.139,
      "step": 1592
    },
    {
      "epoch": 4.02,
      "learning_rate": 7.676116838487974e-06,
      "loss": 0.8274,
      "step": 1600
    },
    {
      "epoch": 4.05,
      "learning_rate": 7.654639175257732e-06,
      "loss": 0.7948,
      "step": 1610
    },
    {
      "epoch": 4.07,
      "learning_rate": 7.633161512027491e-06,
      "loss": 0.7874,
      "step": 1620
    },
    {
      "epoch": 4.1,
      "learning_rate": 7.611683848797251e-06,
      "loss": 0.7902,
      "step": 1630
    },
    {
      "epoch": 4.12,
      "learning_rate": 7.59020618556701e-06,
      "loss": 0.8395,
      "step": 1640
    },
    {
      "epoch": 4.15,
      "learning_rate": 7.56872852233677e-06,
      "loss": 0.8212,
      "step": 1650
    },
    {
      "epoch": 4.17,
      "learning_rate": 7.54725085910653e-06,
      "loss": 0.7799,
      "step": 1660
    },
    {
      "epoch": 4.2,
      "learning_rate": 7.525773195876289e-06,
      "loss": 0.8153,
      "step": 1670
    },
    {
      "epoch": 4.22,
      "learning_rate": 7.504295532646049e-06,
      "loss": 0.7991,
      "step": 1680
    },
    {
      "epoch": 4.25,
      "learning_rate": 7.4828178694158086e-06,
      "loss": 0.7819,
      "step": 1690
    },
    {
      "epoch": 4.27,
      "learning_rate": 7.4613402061855675e-06,
      "loss": 0.7629,
      "step": 1700
    },
    {
      "epoch": 4.3,
      "learning_rate": 7.439862542955327e-06,
      "loss": 0.7433,
      "step": 1710
    },
    {
      "epoch": 4.32,
      "learning_rate": 7.418384879725087e-06,
      "loss": 0.7999,
      "step": 1720
    },
    {
      "epoch": 4.35,
      "learning_rate": 7.396907216494846e-06,
      "loss": 0.791,
      "step": 1730
    },
    {
      "epoch": 4.37,
      "learning_rate": 7.375429553264606e-06,
      "loss": 0.793,
      "step": 1740
    },
    {
      "epoch": 4.4,
      "learning_rate": 7.353951890034365e-06,
      "loss": 0.7949,
      "step": 1750
    },
    {
      "epoch": 4.42,
      "learning_rate": 7.332474226804124e-06,
      "loss": 0.7938,
      "step": 1760
    },
    {
      "epoch": 4.45,
      "learning_rate": 7.310996563573883e-06,
      "loss": 0.7626,
      "step": 1770
    },
    {
      "epoch": 4.47,
      "learning_rate": 7.289518900343643e-06,
      "loss": 0.7922,
      "step": 1780
    },
    {
      "epoch": 4.5,
      "learning_rate": 7.2680412371134026e-06,
      "loss": 0.7504,
      "step": 1790
    },
    {
      "epoch": 4.52,
      "learning_rate": 7.2465635738831615e-06,
      "loss": 0.7723,
      "step": 1800
    },
    {
      "epoch": 4.55,
      "learning_rate": 7.225085910652921e-06,
      "loss": 0.8179,
      "step": 1810
    },
    {
      "epoch": 4.57,
      "learning_rate": 7.203608247422681e-06,
      "loss": 0.808,
      "step": 1820
    },
    {
      "epoch": 4.6,
      "learning_rate": 7.18213058419244e-06,
      "loss": 0.8161,
      "step": 1830
    },
    {
      "epoch": 4.62,
      "learning_rate": 7.1606529209622e-06,
      "loss": 0.7688,
      "step": 1840
    },
    {
      "epoch": 4.65,
      "learning_rate": 7.13917525773196e-06,
      "loss": 0.7604,
      "step": 1850
    },
    {
      "epoch": 4.67,
      "learning_rate": 7.117697594501719e-06,
      "loss": 0.7663,
      "step": 1860
    },
    {
      "epoch": 4.7,
      "learning_rate": 7.096219931271479e-06,
      "loss": 0.8175,
      "step": 1870
    },
    {
      "epoch": 4.72,
      "learning_rate": 7.0747422680412384e-06,
      "loss": 0.7786,
      "step": 1880
    },
    {
      "epoch": 4.75,
      "learning_rate": 7.0532646048109974e-06,
      "loss": 0.7727,
      "step": 1890
    },
    {
      "epoch": 4.77,
      "learning_rate": 7.031786941580757e-06,
      "loss": 0.7718,
      "step": 1900
    },
    {
      "epoch": 4.8,
      "learning_rate": 7.010309278350515e-06,
      "loss": 0.8037,
      "step": 1910
    },
    {
      "epoch": 4.82,
      "learning_rate": 6.988831615120275e-06,
      "loss": 0.8355,
      "step": 1920
    },
    {
      "epoch": 4.85,
      "learning_rate": 6.967353951890034e-06,
      "loss": 0.796,
      "step": 1930
    },
    {
      "epoch": 4.87,
      "learning_rate": 6.945876288659794e-06,
      "loss": 0.7977,
      "step": 1940
    },
    {
      "epoch": 4.9,
      "learning_rate": 6.924398625429554e-06,
      "loss": 0.7848,
      "step": 1950
    },
    {
      "epoch": 4.92,
      "learning_rate": 6.902920962199313e-06,
      "loss": 0.7811,
      "step": 1960
    },
    {
      "epoch": 4.95,
      "learning_rate": 6.881443298969073e-06,
      "loss": 0.7819,
      "step": 1970
    },
    {
      "epoch": 4.97,
      "learning_rate": 6.8599656357388325e-06,
      "loss": 0.7992,
      "step": 1980
    },
    {
      "epoch": 5.0,
      "learning_rate": 6.8384879725085914e-06,
      "loss": 0.7755,
      "step": 1990
    },
    {
      "epoch": 5.0,
      "eval_accuracy": 0.6892167634594255,
      "eval_f1": 0.6876383274434685,
      "eval_loss": 0.8583415746688843,
      "eval_precision": 0.6896076621388775,
      "eval_recall": 0.6892167634594255,
      "eval_runtime": 175.504,
      "eval_samples_per_second": 36.301,
      "eval_steps_per_second": 1.14,
      "step": 1990
    },
    {
      "epoch": 5.03,
      "learning_rate": 6.817010309278351e-06,
      "loss": 0.7808,
      "step": 2000
    },
    {
      "epoch": 5.05,
      "learning_rate": 6.795532646048111e-06,
      "loss": 0.7375,
      "step": 2010
    },
    {
      "epoch": 5.08,
      "learning_rate": 6.77405498281787e-06,
      "loss": 0.7564,
      "step": 2020
    },
    {
      "epoch": 5.1,
      "learning_rate": 6.75257731958763e-06,
      "loss": 0.7957,
      "step": 2030
    },
    {
      "epoch": 5.13,
      "learning_rate": 6.73109965635739e-06,
      "loss": 0.7293,
      "step": 2040
    },
    {
      "epoch": 5.15,
      "learning_rate": 6.709621993127148e-06,
      "loss": 0.7636,
      "step": 2050
    },
    {
      "epoch": 5.18,
      "learning_rate": 6.688144329896908e-06,
      "loss": 0.7105,
      "step": 2060
    },
    {
      "epoch": 5.2,
      "learning_rate": 6.666666666666667e-06,
      "loss": 0.7399,
      "step": 2070
    },
    {
      "epoch": 5.23,
      "learning_rate": 6.6451890034364265e-06,
      "loss": 0.7378,
      "step": 2080
    },
    {
      "epoch": 5.25,
      "learning_rate": 6.623711340206186e-06,
      "loss": 0.7255,
      "step": 2090
    },
    {
      "epoch": 5.28,
      "learning_rate": 6.602233676975945e-06,
      "loss": 0.7151,
      "step": 2100
    },
    {
      "epoch": 5.3,
      "learning_rate": 6.580756013745705e-06,
      "loss": 0.781,
      "step": 2110
    },
    {
      "epoch": 5.33,
      "learning_rate": 6.559278350515464e-06,
      "loss": 0.6941,
      "step": 2120
    },
    {
      "epoch": 5.35,
      "learning_rate": 6.537800687285224e-06,
      "loss": 0.7324,
      "step": 2130
    },
    {
      "epoch": 5.38,
      "learning_rate": 6.516323024054984e-06,
      "loss": 0.7305,
      "step": 2140
    },
    {
      "epoch": 5.4,
      "learning_rate": 6.494845360824743e-06,
      "loss": 0.7793,
      "step": 2150
    },
    {
      "epoch": 5.43,
      "learning_rate": 6.4733676975945025e-06,
      "loss": 0.691,
      "step": 2160
    },
    {
      "epoch": 5.45,
      "learning_rate": 6.451890034364262e-06,
      "loss": 0.7353,
      "step": 2170
    },
    {
      "epoch": 5.48,
      "learning_rate": 6.430412371134021e-06,
      "loss": 0.7474,
      "step": 2180
    },
    {
      "epoch": 5.5,
      "learning_rate": 6.408934707903781e-06,
      "loss": 0.7648,
      "step": 2190
    },
    {
      "epoch": 5.53,
      "learning_rate": 6.387457044673539e-06,
      "loss": 0.7562,
      "step": 2200
    },
    {
      "epoch": 5.55,
      "learning_rate": 6.365979381443299e-06,
      "loss": 0.7386,
      "step": 2210
    },
    {
      "epoch": 5.58,
      "learning_rate": 6.344501718213059e-06,
      "loss": 0.7719,
      "step": 2220
    },
    {
      "epoch": 5.6,
      "learning_rate": 6.323024054982818e-06,
      "loss": 0.7267,
      "step": 2230
    },
    {
      "epoch": 5.63,
      "learning_rate": 6.301546391752578e-06,
      "loss": 0.7246,
      "step": 2240
    },
    {
      "epoch": 5.65,
      "learning_rate": 6.2800687285223376e-06,
      "loss": 0.7013,
      "step": 2250
    },
    {
      "epoch": 5.68,
      "learning_rate": 6.2585910652920965e-06,
      "loss": 0.7186,
      "step": 2260
    },
    {
      "epoch": 5.7,
      "learning_rate": 6.237113402061856e-06,
      "loss": 0.7172,
      "step": 2270
    },
    {
      "epoch": 5.73,
      "learning_rate": 6.215635738831615e-06,
      "loss": 0.7641,
      "step": 2280
    },
    {
      "epoch": 5.75,
      "learning_rate": 6.194158075601375e-06,
      "loss": 0.7188,
      "step": 2290
    },
    {
      "epoch": 5.78,
      "learning_rate": 6.172680412371135e-06,
      "loss": 0.7574,
      "step": 2300
    },
    {
      "epoch": 5.8,
      "learning_rate": 6.151202749140894e-06,
      "loss": 0.7101,
      "step": 2310
    },
    {
      "epoch": 5.83,
      "learning_rate": 6.129725085910654e-06,
      "loss": 0.7314,
      "step": 2320
    },
    {
      "epoch": 5.85,
      "learning_rate": 6.108247422680414e-06,
      "loss": 0.7116,
      "step": 2330
    },
    {
      "epoch": 5.88,
      "learning_rate": 6.086769759450172e-06,
      "loss": 0.7059,
      "step": 2340
    },
    {
      "epoch": 5.9,
      "learning_rate": 6.0652920962199316e-06,
      "loss": 0.7328,
      "step": 2350
    },
    {
      "epoch": 5.93,
      "learning_rate": 6.0438144329896906e-06,
      "loss": 0.749,
      "step": 2360
    },
    {
      "epoch": 5.95,
      "learning_rate": 6.02233676975945e-06,
      "loss": 0.7107,
      "step": 2370
    },
    {
      "epoch": 5.98,
      "learning_rate": 6.00085910652921e-06,
      "loss": 0.7129,
      "step": 2380
    },
    {
      "epoch": 6.0,
      "eval_accuracy": 0.6931407942238267,
      "eval_f1": 0.6922092895697746,
      "eval_loss": 0.8442299962043762,
      "eval_precision": 0.6950647617596234,
      "eval_recall": 0.6931407942238267,
      "eval_runtime": 176.7871,
      "eval_samples_per_second": 36.038,
      "eval_steps_per_second": 1.131,
      "step": 2388
    },
    {
      "epoch": 6.01,
      "learning_rate": 5.979381443298969e-06,
      "loss": 0.7205,
      "step": 2390
    },
    {
      "epoch": 6.03,
      "learning_rate": 5.957903780068729e-06,
      "loss": 0.7511,
      "step": 2400
    },
    {
      "epoch": 6.06,
      "learning_rate": 5.936426116838489e-06,
      "loss": 0.7007,
      "step": 2410
    },
    {
      "epoch": 6.08,
      "learning_rate": 5.914948453608248e-06,
      "loss": 0.7468,
      "step": 2420
    },
    {
      "epoch": 6.11,
      "learning_rate": 5.893470790378008e-06,
      "loss": 0.6768,
      "step": 2430
    },
    {
      "epoch": 6.13,
      "learning_rate": 5.8719931271477675e-06,
      "loss": 0.6902,
      "step": 2440
    },
    {
      "epoch": 6.16,
      "learning_rate": 5.8505154639175264e-06,
      "loss": 0.673,
      "step": 2450
    },
    {
      "epoch": 6.18,
      "learning_rate": 5.829037800687286e-06,
      "loss": 0.693,
      "step": 2460
    },
    {
      "epoch": 6.21,
      "learning_rate": 5.807560137457045e-06,
      "loss": 0.6839,
      "step": 2470
    },
    {
      "epoch": 6.23,
      "learning_rate": 5.786082474226805e-06,
      "loss": 0.685,
      "step": 2480
    },
    {
      "epoch": 6.26,
      "learning_rate": 5.764604810996563e-06,
      "loss": 0.6983,
      "step": 2490
    },
    {
      "epoch": 6.28,
      "learning_rate": 5.743127147766323e-06,
      "loss": 0.7306,
      "step": 2500
    },
    {
      "epoch": 6.31,
      "learning_rate": 5.721649484536083e-06,
      "loss": 0.659,
      "step": 2510
    },
    {
      "epoch": 6.33,
      "learning_rate": 5.700171821305842e-06,
      "loss": 0.6822,
      "step": 2520
    },
    {
      "epoch": 6.36,
      "learning_rate": 5.678694158075602e-06,
      "loss": 0.6781,
      "step": 2530
    },
    {
      "epoch": 6.38,
      "learning_rate": 5.6572164948453615e-06,
      "loss": 0.6695,
      "step": 2540
    },
    {
      "epoch": 6.41,
      "learning_rate": 5.6357388316151204e-06,
      "loss": 0.6822,
      "step": 2550
    },
    {
      "epoch": 6.43,
      "learning_rate": 5.61426116838488e-06,
      "loss": 0.7029,
      "step": 2560
    },
    {
      "epoch": 6.46,
      "learning_rate": 5.59278350515464e-06,
      "loss": 0.6478,
      "step": 2570
    },
    {
      "epoch": 6.48,
      "learning_rate": 5.571305841924399e-06,
      "loss": 0.6929,
      "step": 2580
    },
    {
      "epoch": 6.51,
      "learning_rate": 5.549828178694159e-06,
      "loss": 0.7162,
      "step": 2590
    },
    {
      "epoch": 6.53,
      "learning_rate": 5.528350515463919e-06,
      "loss": 0.6372,
      "step": 2600
    },
    {
      "epoch": 6.56,
      "learning_rate": 5.506872852233678e-06,
      "loss": 0.6781,
      "step": 2610
    },
    {
      "epoch": 6.58,
      "learning_rate": 5.4853951890034375e-06,
      "loss": 0.7271,
      "step": 2620
    },
    {
      "epoch": 6.61,
      "learning_rate": 5.463917525773196e-06,
      "loss": 0.6993,
      "step": 2630
    },
    {
      "epoch": 6.63,
      "learning_rate": 5.4424398625429555e-06,
      "loss": 0.6819,
      "step": 2640
    },
    {
      "epoch": 6.66,
      "learning_rate": 5.4209621993127145e-06,
      "loss": 0.6522,
      "step": 2650
    },
    {
      "epoch": 6.68,
      "learning_rate": 5.399484536082474e-06,
      "loss": 0.6682,
      "step": 2660
    },
    {
      "epoch": 6.71,
      "learning_rate": 5.378006872852234e-06,
      "loss": 0.6619,
      "step": 2670
    },
    {
      "epoch": 6.73,
      "learning_rate": 5.356529209621993e-06,
      "loss": 0.7058,
      "step": 2680
    },
    {
      "epoch": 6.76,
      "learning_rate": 5.335051546391753e-06,
      "loss": 0.7326,
      "step": 2690
    },
    {
      "epoch": 6.78,
      "learning_rate": 5.313573883161513e-06,
      "loss": 0.6588,
      "step": 2700
    },
    {
      "epoch": 6.81,
      "learning_rate": 5.292096219931272e-06,
      "loss": 0.6929,
      "step": 2710
    },
    {
      "epoch": 6.83,
      "learning_rate": 5.2706185567010315e-06,
      "loss": 0.6712,
      "step": 2720
    },
    {
      "epoch": 6.86,
      "learning_rate": 5.249140893470791e-06,
      "loss": 0.6928,
      "step": 2730
    },
    {
      "epoch": 6.88,
      "learning_rate": 5.22766323024055e-06,
      "loss": 0.6715,
      "step": 2740
    },
    {
      "epoch": 6.91,
      "learning_rate": 5.20618556701031e-06,
      "loss": 0.6951,
      "step": 2750
    },
    {
      "epoch": 6.93,
      "learning_rate": 5.18470790378007e-06,
      "loss": 0.6596,
      "step": 2760
    },
    {
      "epoch": 6.96,
      "learning_rate": 5.163230240549829e-06,
      "loss": 0.6738,
      "step": 2770
    },
    {
      "epoch": 6.98,
      "learning_rate": 5.141752577319588e-06,
      "loss": 0.6549,
      "step": 2780
    },
    {
      "epoch": 7.0,
      "eval_accuracy": 0.6951812902213154,
      "eval_f1": 0.6978348522900216,
      "eval_loss": 0.8494423031806946,
      "eval_precision": 0.7054213059137867,
      "eval_recall": 0.6951812902213154,
      "eval_runtime": 176.4527,
      "eval_samples_per_second": 36.106,
      "eval_steps_per_second": 1.133,
      "step": 2786
    },
    {
      "epoch": 7.01,
      "learning_rate": 5.120274914089347e-06,
      "loss": 0.7313,
      "step": 2790
    },
    {
      "epoch": 7.04,
      "learning_rate": 5.098797250859107e-06,
      "loss": 0.6489,
      "step": 2800
    },
    {
      "epoch": 7.06,
      "learning_rate": 5.077319587628866e-06,
      "loss": 0.6269,
      "step": 2810
    },
    {
      "epoch": 7.09,
      "learning_rate": 5.0558419243986255e-06,
      "loss": 0.6305,
      "step": 2820
    },
    {
      "epoch": 7.11,
      "learning_rate": 5.034364261168385e-06,
      "loss": 0.6454,
      "step": 2830
    },
    {
      "epoch": 7.14,
      "learning_rate": 5.012886597938144e-06,
      "loss": 0.6492,
      "step": 2840
    },
    {
      "epoch": 7.16,
      "learning_rate": 4.991408934707904e-06,
      "loss": 0.6347,
      "step": 2850
    },
    {
      "epoch": 7.19,
      "learning_rate": 4.969931271477664e-06,
      "loss": 0.6528,
      "step": 2860
    },
    {
      "epoch": 7.21,
      "learning_rate": 4.948453608247423e-06,
      "loss": 0.6789,
      "step": 2870
    },
    {
      "epoch": 7.24,
      "learning_rate": 4.926975945017183e-06,
      "loss": 0.6711,
      "step": 2880
    },
    {
      "epoch": 7.26,
      "learning_rate": 4.905498281786942e-06,
      "loss": 0.6095,
      "step": 2890
    },
    {
      "epoch": 7.29,
      "learning_rate": 4.884020618556702e-06,
      "loss": 0.6876,
      "step": 2900
    },
    {
      "epoch": 7.31,
      "learning_rate": 4.8625429553264606e-06,
      "loss": 0.6622,
      "step": 2910
    },
    {
      "epoch": 7.34,
      "learning_rate": 4.84106529209622e-06,
      "loss": 0.6514,
      "step": 2920
    },
    {
      "epoch": 7.36,
      "learning_rate": 4.81958762886598e-06,
      "loss": 0.6967,
      "step": 2930
    },
    {
      "epoch": 7.39,
      "learning_rate": 4.798109965635739e-06,
      "loss": 0.6579,
      "step": 2940
    },
    {
      "epoch": 7.41,
      "learning_rate": 4.776632302405499e-06,
      "loss": 0.6527,
      "step": 2950
    },
    {
      "epoch": 7.44,
      "learning_rate": 4.755154639175258e-06,
      "loss": 0.6347,
      "step": 2960
    },
    {
      "epoch": 7.46,
      "learning_rate": 4.733676975945017e-06,
      "loss": 0.6346,
      "step": 2970
    },
    {
      "epoch": 7.49,
      "learning_rate": 4.712199312714777e-06,
      "loss": 0.615,
      "step": 2980
    },
    {
      "epoch": 7.51,
      "learning_rate": 4.690721649484537e-06,
      "loss": 0.6616,
      "step": 2990
    },
    {
      "epoch": 7.54,
      "learning_rate": 4.669243986254296e-06,
      "loss": 0.6685,
      "step": 3000
    },
    {
      "epoch": 7.56,
      "learning_rate": 4.6477663230240554e-06,
      "loss": 0.6549,
      "step": 3010
    },
    {
      "epoch": 7.59,
      "learning_rate": 4.626288659793815e-06,
      "loss": 0.6587,
      "step": 3020
    },
    {
      "epoch": 7.61,
      "learning_rate": 4.604810996563574e-06,
      "loss": 0.6148,
      "step": 3030
    },
    {
      "epoch": 7.64,
      "learning_rate": 4.583333333333333e-06,
      "loss": 0.6231,
      "step": 3040
    },
    {
      "epoch": 7.66,
      "learning_rate": 4.561855670103093e-06,
      "loss": 0.6695,
      "step": 3050
    },
    {
      "epoch": 7.69,
      "learning_rate": 4.540378006872853e-06,
      "loss": 0.6499,
      "step": 3060
    },
    {
      "epoch": 7.71,
      "learning_rate": 4.518900343642612e-06,
      "loss": 0.6328,
      "step": 3070
    },
    {
      "epoch": 7.74,
      "learning_rate": 4.497422680412372e-06,
      "loss": 0.6388,
      "step": 3080
    },
    {
      "epoch": 7.76,
      "learning_rate": 4.4759450171821315e-06,
      "loss": 0.6338,
      "step": 3090
    },
    {
      "epoch": 7.79,
      "learning_rate": 4.4544673539518905e-06,
      "loss": 0.6098,
      "step": 3100
    },
    {
      "epoch": 7.81,
      "learning_rate": 4.4329896907216494e-06,
      "loss": 0.7013,
      "step": 3110
    },
    {
      "epoch": 7.84,
      "learning_rate": 4.411512027491409e-06,
      "loss": 0.623,
      "step": 3120
    },
    {
      "epoch": 7.86,
      "learning_rate": 4.390034364261169e-06,
      "loss": 0.6779,
      "step": 3130
    },
    {
      "epoch": 7.89,
      "learning_rate": 4.368556701030928e-06,
      "loss": 0.6209,
      "step": 3140
    },
    {
      "epoch": 7.91,
      "learning_rate": 4.347079037800688e-06,
      "loss": 0.6517,
      "step": 3150
    },
    {
      "epoch": 7.94,
      "learning_rate": 4.325601374570447e-06,
      "loss": 0.6249,
      "step": 3160
    },
    {
      "epoch": 7.96,
      "learning_rate": 4.304123711340207e-06,
      "loss": 0.6688,
      "step": 3170
    },
    {
      "epoch": 7.99,
      "learning_rate": 4.282646048109966e-06,
      "loss": 0.6246,
      "step": 3180
    },
    {
      "epoch": 8.0,
      "eval_accuracy": 0.6962800188353476,
      "eval_f1": 0.6977269225195705,
      "eval_loss": 0.8394434452056885,
      "eval_precision": 0.7022734754653476,
      "eval_recall": 0.6962800188353476,
      "eval_runtime": 176.2039,
      "eval_samples_per_second": 36.157,
      "eval_steps_per_second": 1.135,
      "step": 3184
    },
    {
      "epoch": 8.02,
      "learning_rate": 4.2611683848797255e-06,
      "loss": 0.6468,
      "step": 3190
    },
    {
      "epoch": 8.04,
      "learning_rate": 4.2396907216494845e-06,
      "loss": 0.6475,
      "step": 3200
    },
    {
      "epoch": 8.07,
      "learning_rate": 4.218213058419244e-06,
      "loss": 0.616,
      "step": 3210
    },
    {
      "epoch": 8.09,
      "learning_rate": 4.196735395189004e-06,
      "loss": 0.6218,
      "step": 3220
    },
    {
      "epoch": 8.12,
      "learning_rate": 4.175257731958763e-06,
      "loss": 0.6196,
      "step": 3230
    },
    {
      "epoch": 8.14,
      "learning_rate": 4.153780068728523e-06,
      "loss": 0.618,
      "step": 3240
    },
    {
      "epoch": 8.17,
      "learning_rate": 4.132302405498282e-06,
      "loss": 0.5797,
      "step": 3250
    },
    {
      "epoch": 8.19,
      "learning_rate": 4.110824742268042e-06,
      "loss": 0.5912,
      "step": 3260
    },
    {
      "epoch": 8.22,
      "learning_rate": 4.089347079037801e-06,
      "loss": 0.5914,
      "step": 3270
    },
    {
      "epoch": 8.24,
      "learning_rate": 4.0678694158075605e-06,
      "loss": 0.6207,
      "step": 3280
    },
    {
      "epoch": 8.27,
      "learning_rate": 4.04639175257732e-06,
      "loss": 0.6014,
      "step": 3290
    },
    {
      "epoch": 8.29,
      "learning_rate": 4.024914089347079e-06,
      "loss": 0.5937,
      "step": 3300
    },
    {
      "epoch": 8.32,
      "learning_rate": 4.003436426116839e-06,
      "loss": 0.6407,
      "step": 3310
    },
    {
      "epoch": 8.34,
      "learning_rate": 3.981958762886598e-06,
      "loss": 0.6087,
      "step": 3320
    },
    {
      "epoch": 8.37,
      "learning_rate": 3.960481099656357e-06,
      "loss": 0.5589,
      "step": 3330
    },
    {
      "epoch": 8.39,
      "learning_rate": 3.939003436426117e-06,
      "loss": 0.6039,
      "step": 3340
    },
    {
      "epoch": 8.42,
      "learning_rate": 3.917525773195877e-06,
      "loss": 0.6225,
      "step": 3350
    },
    {
      "epoch": 8.44,
      "learning_rate": 3.896048109965636e-06,
      "loss": 0.5764,
      "step": 3360
    },
    {
      "epoch": 8.47,
      "learning_rate": 3.8745704467353956e-06,
      "loss": 0.6121,
      "step": 3370
    },
    {
      "epoch": 8.49,
      "learning_rate": 3.853092783505155e-06,
      "loss": 0.6271,
      "step": 3380
    },
    {
      "epoch": 8.52,
      "learning_rate": 3.831615120274914e-06,
      "loss": 0.5951,
      "step": 3390
    },
    {
      "epoch": 8.54,
      "learning_rate": 3.8101374570446738e-06,
      "loss": 0.6312,
      "step": 3400
    },
    {
      "epoch": 8.57,
      "learning_rate": 3.788659793814433e-06,
      "loss": 0.5902,
      "step": 3410
    },
    {
      "epoch": 8.59,
      "learning_rate": 3.7671821305841926e-06,
      "loss": 0.6153,
      "step": 3420
    },
    {
      "epoch": 8.62,
      "learning_rate": 3.7457044673539524e-06,
      "loss": 0.6297,
      "step": 3430
    },
    {
      "epoch": 8.64,
      "learning_rate": 3.724226804123712e-06,
      "loss": 0.581,
      "step": 3440
    },
    {
      "epoch": 8.67,
      "learning_rate": 3.702749140893471e-06,
      "loss": 0.5827,
      "step": 3450
    },
    {
      "epoch": 8.69,
      "learning_rate": 3.6812714776632306e-06,
      "loss": 0.6585,
      "step": 3460
    },
    {
      "epoch": 8.72,
      "learning_rate": 3.6597938144329896e-06,
      "loss": 0.6341,
      "step": 3470
    },
    {
      "epoch": 8.74,
      "learning_rate": 3.6383161512027494e-06,
      "loss": 0.6031,
      "step": 3480
    },
    {
      "epoch": 8.77,
      "learning_rate": 3.616838487972509e-06,
      "loss": 0.6017,
      "step": 3490
    },
    {
      "epoch": 8.79,
      "learning_rate": 3.595360824742268e-06,
      "loss": 0.6274,
      "step": 3500
    },
    {
      "epoch": 8.82,
      "learning_rate": 3.573883161512028e-06,
      "loss": 0.6265,
      "step": 3510
    },
    {
      "epoch": 8.84,
      "learning_rate": 3.5524054982817874e-06,
      "loss": 0.6422,
      "step": 3520
    },
    {
      "epoch": 8.87,
      "learning_rate": 3.530927835051547e-06,
      "loss": 0.6203,
      "step": 3530
    },
    {
      "epoch": 8.89,
      "learning_rate": 3.509450171821306e-06,
      "loss": 0.6177,
      "step": 3540
    },
    {
      "epoch": 8.92,
      "learning_rate": 3.4879725085910652e-06,
      "loss": 0.6095,
      "step": 3550
    },
    {
      "epoch": 8.94,
      "learning_rate": 3.466494845360825e-06,
      "loss": 0.6035,
      "step": 3560
    },
    {
      "epoch": 8.97,
      "learning_rate": 3.4450171821305844e-06,
      "loss": 0.6032,
      "step": 3570
    },
    {
      "epoch": 8.99,
      "learning_rate": 3.423539518900344e-06,
      "loss": 0.6138,
      "step": 3580
    },
    {
      "epoch": 9.0,
      "eval_accuracy": 0.6995762046774446,
      "eval_f1": 0.7012607038931691,
      "eval_loss": 0.8420789241790771,
      "eval_precision": 0.7080354695530748,
      "eval_recall": 0.6995762046774446,
      "eval_runtime": 176.1589,
      "eval_samples_per_second": 36.166,
      "eval_steps_per_second": 1.135,
      "step": 3582
    },
    {
      "epoch": 9.02,
      "learning_rate": 3.4020618556701037e-06,
      "loss": 0.6226,
      "step": 3590
    },
    {
      "epoch": 9.05,
      "learning_rate": 3.380584192439863e-06,
      "loss": 0.5687,
      "step": 3600
    },
    {
      "epoch": 9.07,
      "learning_rate": 3.359106529209622e-06,
      "loss": 0.5863,
      "step": 3610
    },
    {
      "epoch": 9.1,
      "learning_rate": 3.3376288659793814e-06,
      "loss": 0.5831,
      "step": 3620
    },
    {
      "epoch": 9.12,
      "learning_rate": 3.316151202749141e-06,
      "loss": 0.5893,
      "step": 3630
    },
    {
      "epoch": 9.15,
      "learning_rate": 3.2946735395189007e-06,
      "loss": 0.5964,
      "step": 3640
    },
    {
      "epoch": 9.17,
      "learning_rate": 3.27319587628866e-06,
      "loss": 0.5342,
      "step": 3650
    },
    {
      "epoch": 9.2,
      "learning_rate": 3.2517182130584195e-06,
      "loss": 0.6017,
      "step": 3660
    },
    {
      "epoch": 9.22,
      "learning_rate": 3.2302405498281793e-06,
      "loss": 0.5704,
      "step": 3670
    },
    {
      "epoch": 9.25,
      "learning_rate": 3.2087628865979387e-06,
      "loss": 0.5659,
      "step": 3680
    },
    {
      "epoch": 9.27,
      "learning_rate": 3.1872852233676977e-06,
      "loss": 0.5769,
      "step": 3690
    },
    {
      "epoch": 9.3,
      "learning_rate": 3.165807560137457e-06,
      "loss": 0.5906,
      "step": 3700
    },
    {
      "epoch": 9.32,
      "learning_rate": 3.1443298969072165e-06,
      "loss": 0.6157,
      "step": 3710
    },
    {
      "epoch": 9.35,
      "learning_rate": 3.1228522336769763e-06,
      "loss": 0.588,
      "step": 3720
    },
    {
      "epoch": 9.37,
      "learning_rate": 3.1013745704467357e-06,
      "loss": 0.5753,
      "step": 3730
    },
    {
      "epoch": 9.4,
      "learning_rate": 3.079896907216495e-06,
      "loss": 0.5937,
      "step": 3740
    },
    {
      "epoch": 9.42,
      "learning_rate": 3.058419243986255e-06,
      "loss": 0.5785,
      "step": 3750
    },
    {
      "epoch": 9.45,
      "learning_rate": 3.036941580756014e-06,
      "loss": 0.5817,
      "step": 3760
    },
    {
      "epoch": 9.47,
      "learning_rate": 3.0154639175257733e-06,
      "loss": 0.5899,
      "step": 3770
    },
    {
      "epoch": 9.5,
      "learning_rate": 2.9939862542955327e-06,
      "loss": 0.5673,
      "step": 3780
    },
    {
      "epoch": 9.52,
      "learning_rate": 2.9725085910652925e-06,
      "loss": 0.5691,
      "step": 3790
    },
    {
      "epoch": 9.55,
      "learning_rate": 2.951030927835052e-06,
      "loss": 0.5782,
      "step": 3800
    },
    {
      "epoch": 9.57,
      "learning_rate": 2.9295532646048113e-06,
      "loss": 0.6209,
      "step": 3810
    },
    {
      "epoch": 9.6,
      "learning_rate": 2.9080756013745707e-06,
      "loss": 0.5828,
      "step": 3820
    },
    {
      "epoch": 9.62,
      "learning_rate": 2.8865979381443297e-06,
      "loss": 0.6037,
      "step": 3830
    },
    {
      "epoch": 9.65,
      "learning_rate": 2.8651202749140895e-06,
      "loss": 0.5255,
      "step": 3840
    },
    {
      "epoch": 9.67,
      "learning_rate": 2.843642611683849e-06,
      "loss": 0.5548,
      "step": 3850
    },
    {
      "epoch": 9.7,
      "learning_rate": 2.8221649484536083e-06,
      "loss": 0.55,
      "step": 3860
    },
    {
      "epoch": 9.72,
      "learning_rate": 2.800687285223368e-06,
      "loss": 0.5734,
      "step": 3870
    },
    {
      "epoch": 9.75,
      "learning_rate": 2.7792096219931276e-06,
      "loss": 0.5659,
      "step": 3880
    },
    {
      "epoch": 9.77,
      "learning_rate": 2.757731958762887e-06,
      "loss": 0.6138,
      "step": 3890
    },
    {
      "epoch": 9.8,
      "learning_rate": 2.7362542955326464e-06,
      "loss": 0.5592,
      "step": 3900
    },
    {
      "epoch": 9.82,
      "learning_rate": 2.7147766323024053e-06,
      "loss": 0.6158,
      "step": 3910
    },
    {
      "epoch": 9.85,
      "learning_rate": 2.693298969072165e-06,
      "loss": 0.5778,
      "step": 3920
    },
    {
      "epoch": 9.87,
      "learning_rate": 2.6718213058419246e-06,
      "loss": 0.5469,
      "step": 3930
    },
    {
      "epoch": 9.9,
      "learning_rate": 2.650343642611684e-06,
      "loss": 0.5407,
      "step": 3940
    },
    {
      "epoch": 9.92,
      "learning_rate": 2.628865979381444e-06,
      "loss": 0.5637,
      "step": 3950
    },
    {
      "epoch": 9.95,
      "learning_rate": 2.607388316151203e-06,
      "loss": 0.5735,
      "step": 3960
    },
    {
      "epoch": 9.97,
      "learning_rate": 2.5859106529209626e-06,
      "loss": 0.6139,
      "step": 3970
    },
    {
      "epoch": 10.0,
      "learning_rate": 2.5644329896907216e-06,
      "loss": 0.5824,
      "step": 3980
    },
    {
      "epoch": 10.0,
      "eval_accuracy": 0.7022445455972375,
      "eval_f1": 0.702185081437324,
      "eval_loss": 0.8437774181365967,
      "eval_precision": 0.7038651811268685,
      "eval_recall": 0.7022445455972375,
      "eval_runtime": 176.0815,
      "eval_samples_per_second": 36.182,
      "eval_steps_per_second": 1.136,
      "step": 3980
    },
    {
      "epoch": 10.03,
      "learning_rate": 2.542955326460481e-06,
      "loss": 0.5513,
      "step": 3990
    },
    {
      "epoch": 10.05,
      "learning_rate": 2.521477663230241e-06,
      "loss": 0.5145,
      "step": 4000
    },
    {
      "epoch": 10.08,
      "learning_rate": 2.5e-06,
      "loss": 0.5532,
      "step": 4010
    },
    {
      "epoch": 10.1,
      "learning_rate": 2.4785223367697596e-06,
      "loss": 0.5276,
      "step": 4020
    },
    {
      "epoch": 10.13,
      "learning_rate": 2.457044673539519e-06,
      "loss": 0.5481,
      "step": 4030
    },
    {
      "epoch": 10.15,
      "learning_rate": 2.4355670103092784e-06,
      "loss": 0.5673,
      "step": 4040
    },
    {
      "epoch": 10.18,
      "learning_rate": 2.4140893470790382e-06,
      "loss": 0.5783,
      "step": 4050
    },
    {
      "epoch": 10.2,
      "learning_rate": 2.3926116838487976e-06,
      "loss": 0.5852,
      "step": 4060
    },
    {
      "epoch": 10.23,
      "learning_rate": 2.3711340206185566e-06,
      "loss": 0.5597,
      "step": 4070
    },
    {
      "epoch": 10.25,
      "learning_rate": 2.3496563573883164e-06,
      "loss": 0.5156,
      "step": 4080
    },
    {
      "epoch": 10.28,
      "learning_rate": 2.328178694158076e-06,
      "loss": 0.5661,
      "step": 4090
    },
    {
      "epoch": 10.3,
      "learning_rate": 2.3067010309278352e-06,
      "loss": 0.5142,
      "step": 4100
    },
    {
      "epoch": 10.33,
      "learning_rate": 2.2852233676975946e-06,
      "loss": 0.5991,
      "step": 4110
    },
    {
      "epoch": 10.35,
      "learning_rate": 2.263745704467354e-06,
      "loss": 0.5411,
      "step": 4120
    },
    {
      "epoch": 10.38,
      "learning_rate": 2.242268041237114e-06,
      "loss": 0.5585,
      "step": 4130
    },
    {
      "epoch": 10.4,
      "learning_rate": 2.220790378006873e-06,
      "loss": 0.5741,
      "step": 4140
    },
    {
      "epoch": 10.43,
      "learning_rate": 2.1993127147766322e-06,
      "loss": 0.5793,
      "step": 4150
    },
    {
      "epoch": 10.45,
      "learning_rate": 2.177835051546392e-06,
      "loss": 0.5736,
      "step": 4160
    },
    {
      "epoch": 10.48,
      "learning_rate": 2.1563573883161515e-06,
      "loss": 0.5421,
      "step": 4170
    },
    {
      "epoch": 10.5,
      "learning_rate": 2.134879725085911e-06,
      "loss": 0.5545,
      "step": 4180
    },
    {
      "epoch": 10.53,
      "learning_rate": 2.1134020618556703e-06,
      "loss": 0.5386,
      "step": 4190
    },
    {
      "epoch": 10.55,
      "learning_rate": 2.0919243986254297e-06,
      "loss": 0.5526,
      "step": 4200
    },
    {
      "epoch": 10.58,
      "learning_rate": 2.0704467353951895e-06,
      "loss": 0.5206,
      "step": 4210
    },
    {
      "epoch": 10.6,
      "learning_rate": 2.0489690721649485e-06,
      "loss": 0.5299,
      "step": 4220
    },
    {
      "epoch": 10.63,
      "learning_rate": 2.027491408934708e-06,
      "loss": 0.5651,
      "step": 4230
    },
    {
      "epoch": 10.65,
      "learning_rate": 2.0060137457044677e-06,
      "loss": 0.5749,
      "step": 4240
    },
    {
      "epoch": 10.68,
      "learning_rate": 1.9845360824742267e-06,
      "loss": 0.5602,
      "step": 4250
    },
    {
      "epoch": 10.7,
      "learning_rate": 1.9630584192439865e-06,
      "loss": 0.5555,
      "step": 4260
    },
    {
      "epoch": 10.73,
      "learning_rate": 1.941580756013746e-06,
      "loss": 0.5262,
      "step": 4270
    },
    {
      "epoch": 10.75,
      "learning_rate": 1.9201030927835053e-06,
      "loss": 0.5816,
      "step": 4280
    },
    {
      "epoch": 10.78,
      "learning_rate": 1.8986254295532647e-06,
      "loss": 0.5511,
      "step": 4290
    },
    {
      "epoch": 10.8,
      "learning_rate": 1.8771477663230241e-06,
      "loss": 0.557,
      "step": 4300
    },
    {
      "epoch": 10.83,
      "learning_rate": 1.8556701030927837e-06,
      "loss": 0.5636,
      "step": 4310
    },
    {
      "epoch": 10.85,
      "learning_rate": 1.834192439862543e-06,
      "loss": 0.5939,
      "step": 4320
    },
    {
      "epoch": 10.88,
      "learning_rate": 1.8127147766323025e-06,
      "loss": 0.5513,
      "step": 4330
    },
    {
      "epoch": 10.9,
      "learning_rate": 1.791237113402062e-06,
      "loss": 0.5536,
      "step": 4340
    },
    {
      "epoch": 10.93,
      "learning_rate": 1.7697594501718215e-06,
      "loss": 0.5134,
      "step": 4350
    },
    {
      "epoch": 10.95,
      "learning_rate": 1.7482817869415807e-06,
      "loss": 0.5705,
      "step": 4360
    },
    {
      "epoch": 10.98,
      "learning_rate": 1.7268041237113403e-06,
      "loss": 0.5517,
      "step": 4370
    },
    {
      "epoch": 11.0,
      "eval_accuracy": 0.7002040495997489,
      "eval_f1": 0.7005239805464638,
      "eval_loss": 0.8496814370155334,
      "eval_precision": 0.7034423193747762,
      "eval_recall": 0.7002040495997489,
      "eval_runtime": 175.9126,
      "eval_samples_per_second": 36.217,
      "eval_steps_per_second": 1.137,
      "step": 4378
    },
    {
      "epoch": 11.01,
      "learning_rate": 1.7053264604810997e-06,
      "loss": 0.542,
      "step": 4380
    },
    {
      "epoch": 11.03,
      "learning_rate": 1.6838487972508594e-06,
      "loss": 0.5636,
      "step": 4390
    },
    {
      "epoch": 11.06,
      "learning_rate": 1.6623711340206185e-06,
      "loss": 0.504,
      "step": 4400
    },
    {
      "epoch": 11.08,
      "learning_rate": 1.6408934707903782e-06,
      "loss": 0.5493,
      "step": 4410
    },
    {
      "epoch": 11.11,
      "learning_rate": 1.6194158075601378e-06,
      "loss": 0.5387,
      "step": 4420
    },
    {
      "epoch": 11.13,
      "learning_rate": 1.597938144329897e-06,
      "loss": 0.5566,
      "step": 4430
    },
    {
      "epoch": 11.16,
      "learning_rate": 1.5764604810996564e-06,
      "loss": 0.491,
      "step": 4440
    },
    {
      "epoch": 11.18,
      "learning_rate": 1.554982817869416e-06,
      "loss": 0.5398,
      "step": 4450
    },
    {
      "epoch": 11.21,
      "learning_rate": 1.5335051546391756e-06,
      "loss": 0.5535,
      "step": 4460
    },
    {
      "epoch": 11.23,
      "learning_rate": 1.5120274914089348e-06,
      "loss": 0.5302,
      "step": 4470
    },
    {
      "epoch": 11.26,
      "learning_rate": 1.4905498281786942e-06,
      "loss": 0.5289,
      "step": 4480
    },
    {
      "epoch": 11.28,
      "learning_rate": 1.4690721649484538e-06,
      "loss": 0.5272,
      "step": 4490
    },
    {
      "epoch": 11.31,
      "learning_rate": 1.4475945017182134e-06,
      "loss": 0.5543,
      "step": 4500
    },
    {
      "epoch": 11.33,
      "learning_rate": 1.4261168384879726e-06,
      "loss": 0.5328,
      "step": 4510
    },
    {
      "epoch": 11.36,
      "learning_rate": 1.404639175257732e-06,
      "loss": 0.5413,
      "step": 4520
    },
    {
      "epoch": 11.38,
      "learning_rate": 1.3831615120274916e-06,
      "loss": 0.535,
      "step": 4530
    },
    {
      "epoch": 11.41,
      "learning_rate": 1.3616838487972508e-06,
      "loss": 0.5369,
      "step": 4540
    },
    {
      "epoch": 11.43,
      "learning_rate": 1.3402061855670104e-06,
      "loss": 0.5351,
      "step": 4550
    },
    {
      "epoch": 11.46,
      "learning_rate": 1.3187285223367698e-06,
      "loss": 0.5235,
      "step": 4560
    },
    {
      "epoch": 11.48,
      "learning_rate": 1.2972508591065294e-06,
      "loss": 0.5138,
      "step": 4570
    },
    {
      "epoch": 11.51,
      "learning_rate": 1.2757731958762886e-06,
      "loss": 0.5077,
      "step": 4580
    },
    {
      "epoch": 11.53,
      "learning_rate": 1.2542955326460482e-06,
      "loss": 0.4921,
      "step": 4590
    },
    {
      "epoch": 11.56,
      "learning_rate": 1.2328178694158076e-06,
      "loss": 0.5315,
      "step": 4600
    },
    {
      "epoch": 11.58,
      "learning_rate": 1.211340206185567e-06,
      "loss": 0.5415,
      "step": 4610
    },
    {
      "epoch": 11.61,
      "learning_rate": 1.1898625429553264e-06,
      "loss": 0.5123,
      "step": 4620
    },
    {
      "epoch": 11.63,
      "learning_rate": 1.168384879725086e-06,
      "loss": 0.5105,
      "step": 4630
    },
    {
      "epoch": 11.66,
      "learning_rate": 1.1469072164948454e-06,
      "loss": 0.5531,
      "step": 4640
    },
    {
      "epoch": 11.68,
      "learning_rate": 1.1254295532646048e-06,
      "loss": 0.5174,
      "step": 4650
    },
    {
      "epoch": 11.71,
      "learning_rate": 1.1039518900343642e-06,
      "loss": 0.5458,
      "step": 4660
    },
    {
      "epoch": 11.73,
      "learning_rate": 1.0824742268041239e-06,
      "loss": 0.5578,
      "step": 4670
    },
    {
      "epoch": 11.76,
      "learning_rate": 1.0609965635738833e-06,
      "loss": 0.5509,
      "step": 4680
    },
    {
      "epoch": 11.78,
      "learning_rate": 1.0395189003436427e-06,
      "loss": 0.5183,
      "step": 4690
    },
    {
      "epoch": 11.81,
      "learning_rate": 1.018041237113402e-06,
      "loss": 0.5328,
      "step": 4700
    },
    {
      "epoch": 11.83,
      "learning_rate": 9.965635738831617e-07,
      "loss": 0.4984,
      "step": 4710
    },
    {
      "epoch": 11.86,
      "learning_rate": 9.75085910652921e-07,
      "loss": 0.5227,
      "step": 4720
    },
    {
      "epoch": 11.88,
      "learning_rate": 9.536082474226805e-07,
      "loss": 0.5541,
      "step": 4730
    },
    {
      "epoch": 11.91,
      "learning_rate": 9.3213058419244e-07,
      "loss": 0.4869,
      "step": 4740
    },
    {
      "epoch": 11.93,
      "learning_rate": 9.106529209621994e-07,
      "loss": 0.5645,
      "step": 4750
    },
    {
      "epoch": 11.96,
      "learning_rate": 8.891752577319589e-07,
      "loss": 0.538,
      "step": 4760
    },
    {
      "epoch": 11.98,
      "learning_rate": 8.676975945017183e-07,
      "loss": 0.5154,
      "step": 4770
    },
    {
      "epoch": 12.0,
      "eval_accuracy": 0.7020875843666614,
      "eval_f1": 0.7018413131204047,
      "eval_loss": 0.8507698178291321,
      "eval_precision": 0.7030403631521157,
      "eval_recall": 0.7020875843666614,
      "eval_runtime": 175.5971,
      "eval_samples_per_second": 36.282,
      "eval_steps_per_second": 1.139,
      "step": 4776
    },
    {
      "epoch": 12.01,
      "learning_rate": 8.462199312714778e-07,
      "loss": 0.5475,
      "step": 4780
    },
    {
      "epoch": 12.04,
      "learning_rate": 8.247422680412372e-07,
      "loss": 0.5099,
      "step": 4790
    },
    {
      "epoch": 12.06,
      "learning_rate": 8.032646048109967e-07,
      "loss": 0.5053,
      "step": 4800
    },
    {
      "epoch": 12.09,
      "learning_rate": 7.817869415807561e-07,
      "loss": 0.5216,
      "step": 4810
    },
    {
      "epoch": 12.11,
      "learning_rate": 7.603092783505155e-07,
      "loss": 0.5515,
      "step": 4820
    },
    {
      "epoch": 12.14,
      "learning_rate": 7.38831615120275e-07,
      "loss": 0.481,
      "step": 4830
    },
    {
      "epoch": 12.16,
      "learning_rate": 7.173539518900344e-07,
      "loss": 0.5015,
      "step": 4840
    },
    {
      "epoch": 12.19,
      "learning_rate": 6.958762886597939e-07,
      "loss": 0.5338,
      "step": 4850
    },
    {
      "epoch": 12.21,
      "learning_rate": 6.743986254295533e-07,
      "loss": 0.5161,
      "step": 4860
    },
    {
      "epoch": 12.24,
      "learning_rate": 6.529209621993128e-07,
      "loss": 0.5135,
      "step": 4870
    },
    {
      "epoch": 12.26,
      "learning_rate": 6.314432989690722e-07,
      "loss": 0.5002,
      "step": 4880
    },
    {
      "epoch": 12.29,
      "learning_rate": 6.099656357388316e-07,
      "loss": 0.5295,
      "step": 4890
    },
    {
      "epoch": 12.31,
      "learning_rate": 5.884879725085911e-07,
      "loss": 0.5726,
      "step": 4900
    },
    {
      "epoch": 12.34,
      "learning_rate": 5.670103092783505e-07,
      "loss": 0.5028,
      "step": 4910
    },
    {
      "epoch": 12.36,
      "learning_rate": 5.4553264604811e-07,
      "loss": 0.5187,
      "step": 4920
    },
    {
      "epoch": 12.39,
      "learning_rate": 5.240549828178694e-07,
      "loss": 0.5372,
      "step": 4930
    },
    {
      "epoch": 12.41,
      "learning_rate": 5.02577319587629e-07,
      "loss": 0.4981,
      "step": 4940
    },
    {
      "epoch": 12.44,
      "learning_rate": 4.810996563573884e-07,
      "loss": 0.5199,
      "step": 4950
    },
    {
      "epoch": 12.46,
      "learning_rate": 4.596219931271478e-07,
      "loss": 0.5085,
      "step": 4960
    },
    {
      "epoch": 12.49,
      "learning_rate": 4.381443298969072e-07,
      "loss": 0.4946,
      "step": 4970
    },
    {
      "epoch": 12.51,
      "learning_rate": 4.1666666666666667e-07,
      "loss": 0.4826,
      "step": 4980
    },
    {
      "epoch": 12.54,
      "learning_rate": 3.951890034364261e-07,
      "loss": 0.5424,
      "step": 4990
    },
    {
      "epoch": 12.56,
      "learning_rate": 3.737113402061856e-07,
      "loss": 0.5226,
      "step": 5000
    },
    {
      "epoch": 12.59,
      "learning_rate": 3.5223367697594503e-07,
      "loss": 0.5175,
      "step": 5010
    },
    {
      "epoch": 12.61,
      "learning_rate": 3.307560137457045e-07,
      "loss": 0.4768,
      "step": 5020
    },
    {
      "epoch": 12.64,
      "learning_rate": 3.0927835051546394e-07,
      "loss": 0.5287,
      "step": 5030
    },
    {
      "epoch": 12.66,
      "learning_rate": 2.878006872852234e-07,
      "loss": 0.5527,
      "step": 5040
    },
    {
      "epoch": 12.69,
      "learning_rate": 2.6632302405498284e-07,
      "loss": 0.5047,
      "step": 5050
    },
    {
      "epoch": 12.71,
      "learning_rate": 2.448453608247423e-07,
      "loss": 0.5237,
      "step": 5060
    },
    {
      "epoch": 12.74,
      "learning_rate": 2.2336769759450175e-07,
      "loss": 0.5027,
      "step": 5070
    },
    {
      "epoch": 12.76,
      "learning_rate": 2.018900343642612e-07,
      "loss": 0.5102,
      "step": 5080
    },
    {
      "epoch": 12.79,
      "learning_rate": 1.804123711340206e-07,
      "loss": 0.536,
      "step": 5090
    },
    {
      "epoch": 12.81,
      "learning_rate": 1.5893470790378006e-07,
      "loss": 0.5341,
      "step": 5100
    },
    {
      "epoch": 12.84,
      "learning_rate": 1.3745704467353952e-07,
      "loss": 0.4976,
      "step": 5110
    },
    {
      "epoch": 12.86,
      "learning_rate": 1.1597938144329898e-07,
      "loss": 0.5259,
      "step": 5120
    },
    {
      "epoch": 12.89,
      "learning_rate": 9.450171821305842e-08,
      "loss": 0.5559,
      "step": 5130
    },
    {
      "epoch": 12.91,
      "learning_rate": 7.302405498281788e-08,
      "loss": 0.5687,
      "step": 5140
    },
    {
      "epoch": 12.94,
      "learning_rate": 5.1546391752577325e-08,
      "loss": 0.5046,
      "step": 5150
    },
    {
      "epoch": 12.96,
      "learning_rate": 3.006872852233677e-08,
      "loss": 0.4981,
      "step": 5160
    },
    {
      "epoch": 12.99,
      "learning_rate": 8.59106529209622e-09,
      "loss": 0.5318,
      "step": 5170
    },
    {
      "epoch": 13.0,
      "eval_accuracy": 0.7009888557526291,
      "eval_f1": 0.7013039636451874,
      "eval_loss": 0.8539798259735107,
      "eval_precision": 0.7028709661911645,
      "eval_recall": 0.7009888557526291,
      "eval_runtime": 176.5306,
      "eval_samples_per_second": 36.09,
      "eval_steps_per_second": 1.133,
      "step": 5174
    },
    {
      "epoch": 13.0,
      "step": 5174,
      "total_flos": 5.134539349854093e+19,
      "train_loss": 0.7936680297571396,
      "train_runtime": 34458.0815,
      "train_samples_per_second": 19.229,
      "train_steps_per_second": 0.15
    }
  ],
  "max_steps": 5174,
  "num_train_epochs": 13,
  "total_flos": 5.134539349854093e+19,
  "trial_name": null,
  "trial_params": null
}